Nuclear Engineering

Mathematical Modeling and Simulation

Nuclear Engineering
Mathematical Modeling and Simulation

Zafar Ullah Koreshi

Air University, Islamabad,
Pakistan
Academic Press is an imprint of Elsevier
125 London Wall, London EC2Y 5AS, United Kingdom
525 B Street, Suite 1650, San Diego, CA 92101, United States
50 Hampshire Street, 5th Floor, Cambridge, MA 02139, United States
The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, United Kingdom
Copyright © 2022 Elsevier Inc. All rights reserved.
No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including
photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher.
Details on how to seek permission, further information about the Publisher’s permissions policies and our arrangements with
organizations such as the Copyright Clearance Center and the Copyright Licensing Agency, can be found at our website:
www.elsevier.com/permissions.
This book and the individual contributions contained in it are protected under copyright by the Publisher (other than as may be
noted herein).
Notices
Knowledge and best practice in this field are constantly changing. As new research and experience broaden our understanding,
changes in research methods, professional practices, or medical treatment may become necessary.
Practitioners and researchers must always rely on their own experience and knowledge in evaluating and using any information,
methods, compounds, or experiments described herein. In using such information or methods they should be mindful of their
own safety and the safety of others, including parties for whom they have a professional responsibility.
To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors, assume any liability for any injury
and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of
any methods, products, instructions, or ideas contained in the material herein.
ISBN: 978-0-323-90618-0

For Information on all Academic Press publications

visit our website at https://www.elsevier.com/books-and-journals

Publisher: Charlotte Cockle

Editorial Project Manager: Sara Valentino
Production Project Manager: Niranjan Bhaskaran
Cover Designer: Christian J. Bilbow
Typeset by MPS Limited, Chennai, India
 Dedication

To
My Mother
Mahbano Begum
My Father
Ambassador Dr. Samiullah Mujahid Koreshi (PhD)
My Wife
Nasia
My Sons
Mekael, Saif, and Zain
Contents

About the author xiii 2.2 Types of neutron interactions 53

Foreword xv 2.2.1 Neutron scattering in the lab and
center of mass systems 55
1. The atom and nuclear radiation 1 2.3 The microscopic cross-section 58
2.4 The macroscopic cross-section 63
1.1 The atom 1 2.5 Flux measurement 64
1.1.1 Nuclear stability 5 2.6 Reaction rates 66
1.1.2 Binding energy 6 2.7 Neutron slowing down, diffusion and
1.2 Radioactive decay 6 thermalization 67
1.2.1 Alpha decay 8 2.8 Resonance cross-section 74
1.2.2 Beta decay 9 2.9 Nuclear fission 80
1.2.3 Gamma decay 10 2.9.1 The fission process 80
1.2.4 Radioactive nuclides in nuclear 2.9.2 Critical energy 81
technologies 11 2.9.3 Fission yield 84
1.3 Interaction of radiation with matter 12 2.9.4 Number of neutrons emitted in
1.3.1 Interaction of alpha rays with matter 12 fission 84
1.3.2 Interaction of beta radiation with 2.9.5 Fissile and fertile materials 85
matter 16 2.9.6 The fission spectrum 86
1.3.3 Interaction of gamma radiation with 2.10 Criticality 88
matter 19 2.10.1 Diffusion theory 90
1.4 Sources and effects of radiation 24 2.10.2 Transport theory 91
1.4.1 Radiation dose 26 2.10.3 Monte Carlo simulation 92
1.4.2 Absorbed dose 26 Problems 98
1.4.3 Equivalent dose 27 Nomenclature 98
1.4.4 Effective dose 28 References 100
1.4.5 Radiation safety limits 28
1.4.6 Radiation detection 29
1.5 Atomic densities of elements
3. Nuclear reactors and systems 103
and mixtures 30 3.1 Status of nuclear power 103
1.6 Mathematical modeling and simulation 34 3.1.1 Generations of nuclear power 103
1.6.1 Alpha particle transport simulation 35 3.1.2 Reactors shut down 106
1.6.2 Interaction of electrons with matter 35 3.1.3 The future of the nuclear power
1.6.3 Interaction of gamma radiation industry 106
with matter 40 3.2 Nuclear reactor systems 107
1.6.4 Radiation dose from Calfornium-252 3.2.1 Pressurized water reactor 108
gamma source in water 41 3.2.2 Boiling water reactor 110
Capabilities developed 44 3.2.3 Pressurized heavy water reactor 112
Nomenclature 44 3.2.4 Gas cooled reactor 113
Problems 46 3.2.5 Fast breeder reactor 114
References 47 3.3 Marine propulsion reactors 116
3.3.1 Introduction 116
3.3.2 US nuclear submarine program 116
2. Interactions of neutrons with matter 51 3.3.3 Former Soviet/Russian nuclear
2.1 Kinetic theory 51 submarine program 117

vii
viii Contents

3.3.4 Submarine programs: UK, France, 4.6 Approximate methods 185

China, India and Pakistan 117 4.6.1 The Ritz method 185
3.3.5 Modern-day submarines 117 4.6.2 The Rayleigh Ritz variational
3.3.6 Technical features 118 method 186
3.3.7 HEU/LEU submarine reactors 120 4.6.3 The weighted residual method 186
3.4 Plutonium production reactors 120 4.7 The adjoint function 186
3.5 Small modular reactors 121 4.8 Random processes, probability, and
3.5.1 Design features of SMRs 121 statistics 187
3.5.2 Very small modular reactor 125 4.8.1 Random processes 188
3.5.3 Generation-IV reactors 125 4.8.2 Markovian processes 188
3.5.4 Radiation source term 127 4.8.3 Population and sample 188
3.6 Nuclear fusion 128 4.8.4 Random variables, PDF, and CDF 189
3.6.1 The fusion reaction 128 4.8.5 Random numbers 195
3.6.2 Magnetic confinement fusion 129 4.8.6 Sampling from PDFs 196
3.6.3 Inertial confinement fusion 130 4.8.7 Kullback Leibler divergence for
3.7 Space propulsion 132 uniform random numbers 199
3.7.1 Conventional rocket designs 132 4.8.8 The law of large numbers 199
3.7.2 Space exploration 132 4.8.9 The central limit theorem 200
3.7.3 Nuclear rocket designs for deep 4.9 Evaluation of integrals 201
space exploration 134 4.9.1 The Monte Carlo method for
3.8 Nuclear power systems in space 138 numerical integration 203
3.8.1 Radioisotope thermal generators 138 Problems 206
3.8.2 Small nuclear auxiliary power Nomenclature 207
systems 138 References 208
3.9 Conclusions 141
Problems 141
Nomenclature 142 5. The neutron diffusion equation 211
References 144 5.1 The conservation equation 211
ANNEX: The physics of nuclear fusion 145 5.2 The one-group diffusion equation 213
5.2.1 Nonmultiplying systems 213
4. Mathematical foundations 149
5.2.2 Multiplying systems 215
4.1 Ordinary differential equations 150 5.2.3 One-group criticality 219
4.1.1 The Poisson equation: steady-state 5.3 The two-group diffusion equation 221
heat conduction in 1-D 153 5.3.1 Nonmultiplying systems 221
4.1.2 Coupled first-order ODEs: the point 5.3.2 Multiplying systems 227
kinetics equations 155 5.3.3 Two-group criticality 230
4.2 Partial differential equations 156 5.4 The multigroup diffusion equation 234
4.2.1 Equations of fluid dynamics 156 5.4.1 Numerical solution of the
4.2.2 The 1-D time-dependent heat multigroup diffusion equations 235
conduction 157 5.5 Effect of fuel concentration on critical
4.2.3 Laplace equation: 2-D steady-state mass 238
heat conduction 158 5.5.1 Goertzel’s theorem 239
4.2.4 Heat conduction in 2-D and 3-D 159 5.5.2 Nonuniform fuel distribution:
4.2.5 Flux formulation 164 a slab model 239
4.3 Integral equations 165 5.5.3 Nonuniform fuel distribution:
4.3.1 An important integral equation for a spherical model 244
neutron transport 169 5.5.4 Critical core with flat thermal flux
4.3.2 Integral equations in neutron loading 247
transport 169 5.6 The two-group adjoint diffusion equations 248
4.4 Integro-differential equations 170 5.7 Core neutronics with diffusion equations 251
4.5 Numerical methods 174 Problems 256
4.5.1 The Finite Difference Method 174 Nomenclature 257
4.5.2 The Finite Element Method 178 References 258
 Contents ix

6. The neutron transport equation 259 7.2.8 Processing a fission event 314
7.2.9 Processing a capture event 315
6.1 Structure of the neutron transport 7.2.10 Processing an escape-from-system
equation 260 event 315
6.1.1 An integro-differential form of the 7.2.11 Mean and variance 315
neutron transport equation 260 7.2.12 Batch, history, random walk and
6.1.2 The two-group transport equation 265 events 316
6.1.3 The integral form of the transport 7.3 Modeling the geometry 316
equation 266 7.3.1 Geometries for illustration of
6.1.4 Multigroup form of the integral Monte Carlo simulation 320
transport equation 268 7.4 Demonstration 328
6.2 Exact solutions of the transport equation 268 7.5 Variance reduction methods 332
6.2.1 The classic albedo problem 270 7.6 Estimating perturbations with Monte
6.2.2 Infinite medium with a plane Carlo simulation 333
isotropic source 270 7.7 Conclusions 333
6.2.3 Finite sphere with a point isotropic Problems 334
source 274 Nomenclature 334
6.3 Numerical methods for solving the References 335
transport equation 285
6.3.1 The discrete ordinates method 285
6.3.2 The Spherical harmonics method 287
8. Computer codes 337
6.3.3 The DP N method 293 8.1 Neutron and radiation transport codes 338
6.3.4 The B N method 295 8.1.1 ANISN 338
6.3.5 The finite element method 296 8.1.2 DOT 338
6.3.6 The nodal method with transport 8.1.3 TORT 338
theory 296 8.1.4 PARTISN 339
6.3.7 Hybrid methods 297 8.1.5 MCNP 339
6.3.8 Criticality estimates 297 8.1.6 TART 339
6.4 Transport theory for reactor calculations 298 8.1.7 MORSE 339
6.4.1 Collision probability method 299 8.1.8 KENO 340
6.4.2 Method of characteristics 299 8.1.9 Other Monte Carlo codes 340
Problems 302 8.2 Time-dependent reactor kinetics codes 340
Nomenclature 302 8.3 Thermal hydraulics codes 340
References 303 8.4 Radiological protection codes 341
8.5 Performance and safety analyses 341
7. The Monte Carlo method 305 8.6 Nuclear data 341
8.6.1. MCNP 344
7.1 Stochastic simulation 305 8.7 Conclusion 344
7.1.1 Markov processes 305 Problems 344
7.1.2 Events in a random walk 305 Nomenclature 345
7.1.3 The physics of interactions 306 References 345
7.1.4 Nuclear interaction data 306
7.1.5 How do we know an answer
is good? 306
9. Optimization and variational
7.2 Simulation of a random walk 308 methods 349
7.2.1 Monte Carlo simulation 308 9.1 Introduction 349
7.2.2 Estimators and tallies 309 9.2 Deterministic optimization 350
7.2.3 Sampling a source 312 9.2.1 Deterministic optimization
7.2.4 Sampling the “distance without constraints 350
to collision” 313 9.2.2 Deterministic optimization with
7.2.5 Determining the type of event 313 algebraic constraints 351
7.2.6 Determining the nuclide of 9.2.3 Optimal solution with a system of
interaction 314 first-order ordinary differential
7.2.7 Processing a scattering event 314 equation constraints 352
x Contents

9.2.4 Optimal solution with a system of Annex C MCNP input listing (BK10Shld,
first-order ordinary differential Section 10.5.1) 412
equation constraints 355 Annex D MCNP input listing (BK10AP10,
9.2.5 Optimal discrete control Section 10.5.1) 413
(Pontryagin maximum principle) 360
9.3 Controller design and optimization 361 11. Comparisons: Monte Carlo,
9.4 Dynamic programming 365 diffusion, and transport 417
9.5 Stochastic optimization 367
11.1 Introduction 417
9.5.1 Genetic algorithms 367
11.2 Criticality in a bare sphere 417
9.5.2 Particle swarm optimization 372
11.2.1 One-group diffusion theory
9.6 Applications of optimization
criticality 417
in reactors 373
11.2.2 Two-group diffusion theory
9.6.1 Multi-objective core optimization 373
criticality 418
9.6.2 Pressurized water reactor core
11.2.3 One-speed transport theory
pattern optimization 374
criticality 419
9.6.3 Controller proportional integral
11.3 The classic albedo calculation 421
derivative 374
11.4 Flux in a slab 423
9.6.4 Radiation shielding 374
11.4.1 Diffusion theory 423
9.6.5 Some other applications of
11.4.2 Transport theory 424
optimization 375
11.4.3 Monte Carlo simulation 425
Problems 375
11.4.4 Comparison 425
Nomenclature 375
11.5 Flux in a finite sphere with a point
References 376
isotropic source 428
11.5.1 Diffusion theory 428
10. Monte Carlo simulation in 11.5.2 Transport theory exact solution 430
nuclear systems 379 11.5.3 Monte Carlo simulation 431
Problems 433
10.1 Introduction 379
Nomenclature 433
10.2 Bare critical assemblies 381
References 434
10.2.1 Godiva 381
Annex A MATLAB Program
10.2.2 Jezebel 386
AlbedoSlabDiffTh.m (Section 11.3) 435
10.3 Criticality safety 388
Annex B MCNP Input File BK11Albd
10.3.1 Storage of interacting units 388
(Section 11.2) 438
10.3.2 Storage of uranium
Annex C MATLAB Program
hexafluoride cylinders 388
CH11ExactSolSlabJan03.m (Section 11.4.4) 440
10.4 Radiation moderation and shielding 389
10.4.1 Radiation moderation for a
neutron generator 389 12. Exercises in Monte Carlo simulation 449
10.4.2 Radiation shielding 390 12.1 Sampling from a distribution function 449
10.5 Nuclear fission applications 390 12.1.1 Sampling from a normal
10.5.1 Unit lattice cell and fuel distribution 450
assembly of the AP1000 12.1.2 Sampling from a Watt fission
reactor 390 spectrum 451
10.5.2 The Toshiba 4S reactor 394 12.2 Estimating the neutron flux in a non-
10.5.3 Micronuclear reactor 400 multiplying sphere 453
10.6 Nuclear fusion applications 401 12.2.1 The simulation process 453
Problems 405 12.2.2 MATLAB program for point source
Nomenclature 405 in a finite non-multiplying sphere 456
References 406 12.2.3 Results 460
Annex A MCNP listing for Godiva 12.3 Reflected assemblies 462
(Section 10.2.1) 408 12.4 Reactor core modeling 464
Annex B MCNP input listing (Jezebel, 12.4.1 Input file 464
Section 10.2.2) 410 12.4.2 Surrounding cells 465
 Contents xi

12.4.3 Source description 466 13.5.1 Optimal distribution: Pontryagin’s

12.4.4 Plotting the geometry 467 maximum principle 498
12.4.5 Tally cards 470 13.6 Radiation detection or optimization 501
12.4.6 Reaction rates 470 13.7 Controller design optimization 503
12.4.7 Plotting tallies 471 Problems 504
12.5 Radiation safety and shielding 473 Nomenclature 505
12.6 Perturbation calculations 474 References 506
12.7 MCNP geometry plotting in core
neutronics 476 14. Monte Carlo simulation in medical
Problems 480 physics 509
Conclusions 482
Nomenclature 482 14.1 Introduction 509
References 483 14.1.1 The production of radio-isotopes 510
Annex A: MATLAB Program 14.1.2 Alpha radiation therapy 511
CH12_NormalSampling.m 484 14.2 Brachytherapy 512
Annex B MATLAB Program CH12_Watt 14.2.1 Monte Carlo simulation in
Sampling.m 486 brachytherapy 512
14.2.2 Monte Carlo simulation to
calculate energy deposition
13. Optimization in nuclear systems 489
and dose distribution for
13.1 Introduction 489 brachytherapy 514
13.2 Reactor core design optimization 489 Nomenclature 517
13.3 Fusion neutronics design optimization 493 References 517
13.4 Radiation shielding design optimization 494
13.5 Fuel loading pattern optimization 495 Index 521
About the author

Zafar Ullah Koreshi [BSc (Hons) Nuclear Engineering, Queen Mary College, University of
London (UK); MS, Nuclear Engineering, University of Wisconsin, Madison (USA), PhD
Nuclear Engineering, University of Cambridge] is professor at Air University, having contrib-
uted as Dean Faculty of Engineering and is currently the Dean of Graduate Studies at Air
University. His experience includes positions at the Pakistan Atomic Energy Commission, Dr.
AQ Khan Research Laboratories, National University of Sciences and Technology, and at Air
University, Islamabad. He has published in the Annals of Nuclear Energy, Progress in Nuclear
Energy, Nuclear Technology and Radiation Protection, ASME Journal of Nuclear Engineering
and Radiation Science, and in several other leading international journals. Dr. Koreshi has
been Track Chair at the International Conference on Nuclear Engineering held in the United
States, China, and Japan and has presented his research at the American Nuclear Society
Annual Meetings. He is a member of the American Society of Mechanical Engineers (ASME), Professional Engineer
Pakistan Engineering Council, and a life member of the Pakistan Nuclear Society. He has also received commendations
as a reviewer for prestigious journals. Prof. Zafar Koreshi is also an associate editor of the ASME Journal of Nuclear
Engineering and Radiation Science.

xiii
Foreword

This book has evolved from over four decades of involvement in nuclear engineering, industry, and academia beginning
as an undergraduate at Queen Mary College, University of London, then at the University of Wisconsin, Madison
(USA), followed by an engineering career, then at the University of Cambridge for my PhD. The last 35 years of my
career have been at the Pakistan Atomic Energy Commission, at the National University of Sciences and Technology
and the last two decades as professor at Air University.
Even though my teaching and administrative positions kept me busy, I always stayed close to my first academic pas-
sion: nuclear engineering, for which I was lucky to have learnt from professors Mike Williams and Charlie Maynard,
Dr. Jeffery Lewins, and many other learned professors. To them, I owe all my knowledge and confidence developed as
a student. Professor Williams took us students to our first conference in London when we had not the slightest clue as
to what a conference was! From Charlie Maynard I learned not only from his lectures but from his personal values.
From Jeffery Lewins, at Cambridge, I received encouragement and exposure to some of the finest opportunities to listen
to great minds. When Jeffery Lewins was the president of the Institution of Nuclear Engineers, I met Dr. Otto Frisch
and briefly spoke to him. He was one of the first three credited with the discovery and first experiments of nuclear fis-
sion! And of course I must mention my friendship with Prof. Abdus Salam, an accomplished Pakistani scientist and
Nobel Laureate. In congratulating him at his office at Imperial College soon after his Nobel Prize was announced, I
was elated and had many opportunities to accompany him on conferences and to receive his guidance and advice.
In writing this book, I feel that I have completed one duty—transferring whatever knowledge I gained in this field
to the scientific community and especially to the young students studying nuclear engineering in universities all over
the world.
This book is my way of presenting things to the body of knowledge that encompasses nuclear engineering. In it is
the right proportion, I feel, of physics, mathematics, nuclear technology, diffusion theory, transport theory, Monte Carlo
simulation, optimization, and applications and exercises that are essential to an academic nuclear engineer who will
complete a PhD and continue in the profession.
The first two chapters of this book are the foundations, physics, and science of the atom (as known to an engineer);
I think it is an appropriate amount of information to comprehend the alpha, beta, gamma, x- and neutron radiations that
are modeled in nuclear engineering. A difficult concept, Doppler broadening, is explained in a relatively easy way!
Chapter three is an overview of nuclear technology deliberating on nuclear reactors, their types and generations, the
renaissance reactors, and their applications in marine propulsion, in rockets, and towards the exploration of deep space.
This year three missions went to Mars, and we should all be getting some more pictures and information on what the
red planet is like. This was a one-way trip so people could not be sent; in the future we can hope that nuclear rockets,
with far more power than today’s chemical fuels, will take and bring people to and from Mars. So, Chapter 3 gives a
picture on the big systems developed in the nuclear world. I have put in nuclear fusion, both magnetic and inertial, as
ITER could give humanity some hope and the next DEMO reactor could produce electricity from the technology of our
Sun and from fuel made out of water!
Chapter 4 is on mathematical foundations, both deterministic and stochastic; again, in the right proportion! With
this background, Chapter 5 shifts gears into modeling neutrons with simple diffusion theory which, in spite of its sim-
plicity, is quite useful in the design of nuclear reactors and systems. This is followed by neutron transport, which is a
step higher than diffusion, as it incorporates a full six-dimensional phase space formulation in Boltzmann’s transport
theory. This is where an advanced undergraduate student begins to take things a bit more seriously. And then, in
Chapter 7, we step into the world of Monte Carlo (MC) simulation, where the power of big computing comes to our
help in simulating processes to design nuclear reactors. Chapter 8 is a brief description of some well-known and exten-
sively used computer codes in nuclear engineering.
With the foundations more-or-less clearly defined and understood, Chapter 9 deliberates on the mathematics of opti-
mization and optimality as nuclear systems have not only to be designed but designed in the best way.

xv
xvi Foreword

The next three chapters, Chapters 10 12, focus on MC simulation and compare MC with diffusion and transport to
give the feel of a unified way of looking at the modeling of neutron transport and generalizing it to radiation transport.
Chapter 13 reviews current research in the applications of optimization methods in nuclear engineering. Finally,
Chapter 14 gives a little insight into MC simulations in medical physics; this is an area where humanity has benefitted
so much from medical radio-isotopes to cure cancer and to undergo medical treatments and diagnostics.
In this journey, I have to acknowledge the support and encouragement of friends and colleagues, most notably Prof.
Anil Prinja (QMC), Drs. Afzaal Malik, Tasneem Shah, Syed Arif Ahmad, and my colleagues at Air University and all
the other organizations I have had the privilege to work. From my students, too, I have learned a great deal; from Dr
Hamda Khan and Engr. Umair Aziz, and PG students and undergrads in all these years.
Even after all the effort I have put into this book, I must say, a lot remains and I hope this is just the first edition to
be improved into newer editions.
I give this book to the next generations; this is my best effort and I hope that you will do much better. The caravan
of knowledge moves slow but yes it moves.
In the end, the wealth of humans is the knowledge and values that we have and that has all come from our families,
schools, and universities that we must value, love, and sustain. The custodians of that knowledge are the professors
who have shaped this world and will take it to places that we cannot even imagine.
This book is addressed to the scientific community in radiation transport; as a text book, it would be suitable for a
graduate level course spread over two semesters. This would prepare a PhD scholar to be “fully-equipped” to undertake
research in big simulations for design optimization of present-day systems as well as for new systems, such as micro-
nuclear power reactors for space exploration and possibly power generation on a planet such as Mars.
It is necessary to understand that while a great amount of subject matter is covered in this book, several gaps and
shortcomings may appear. It is, in that sense, just a small contribution to the vast body of knowledge which has existed
in this area from which several generations have benefitted.
I would like to thank Elsevier Publishers, particularly Ms. Maria Convey, Acquisitions Editor, Ms. Sara Valentino,
Editorial Manager, Ms. Madeline Jones and Mr Niranjan Bhaskaran, Elsevier, for their continued support and
encouragement.
Finally, my thanks to my mother and father for giving me the wealth of values, support, and opportunities, to my
wife Nasia and to my sons Mekael, Saif, and Zain for their constant support and encouragement.
Zafar Koreshi
Air University
Islamabad, Pakistan
Chapter 1

The atom and nuclear radiation

This chapter comprises what can be considered as the essentials of nuclear engineering as a discipline that owes much
of its present knowledge to theoretical science, discoveries, inventions, and technology.
Nuclear systems are based on the science of atomic and subatomic particles, namely neutrons, protons, electrons
and the interaction of nuclear radiation (gamma and beta) and atomic radiation (X-rays) with matter.
These particles and radiations were discovered over many decades. Pioneering milestones include the discovery of
X-rays by Wilhelm Conrad Röntgen at the University of Würzburg in 1895, spontaneous radioactivity by Antoine
Henri Becquerel (University of Paris) in 1896, radioactivity by Pierre Curie and Marie Curie leading to the discovery of
polonium and radium in 1898 (University of Paris). In 1897, Sir Joseph John Thomson, at the University of Cambridge
discovered the electron. In 1919, one of his students Ernest Rutherford discovered the proton (Blackett, 1924;
7 N 1 2 H e-8 O 1 1 H and worked on radioactivity and named alpha
17
Rutherford, 1919) demonstrating transmutation: 14 4 1

and beta radiation. Another breakthrough was the reaction proposed as: 94 Be 1 42 H e-13 6 C 1 radiation, by Bothe and
Becker and by Mme. Curie but with brilliant arguments, theory and experimentation shown to be a neutron by
Chadwick (1932). In this paper Chadwick writes “If we suppose that the radiation is not a quantum radiation, but con-
sists of particles of mass very nearly equal to that of the proton, all the difficulties connected with the collisions disap-
pear, both with regard to their frequency and to the energy transfer to different masses.” Chadwick goes on to say, “We
may then proceed to build up nuclei out of a-particles, neutrons and protons, and we are able to avoid the presence of
uncombined electrons in a nucleus” and concluding, among other properties of the proposed neutron that “. . .the mass
of the neutron is about 1.006, just a little less than the sum of the masses of a proton and an electron.” Thus
Chadwick’s correct interpretation of these experiments gave us the neutron. Clearly, Chadwick had no knowledge of
quarks and thought that the neutron was an elementary particle. This chapter is on these particles and the nuclear and
atomic radiations as well as on the atomic models for which quantum physics emerged as a field which explained phe-
nomena based on particle-wave duality at very small dimensions.
In quantum physics, the “creators” of the field include Max Planck for the concept of the quanta, de Broglie for giv-
ing the de Broglie wavelength of matter waves, Bohr for the structure of the atom, Compton for demonstrating that
electromagnetic radiation behaves as particles during their interaction with electrons (Compton scattering), Heisenberg
for the uncertainty principle, Schrodinger for the wave function in the Schrodinger equation, Dirac for several develop-
ments in the field and significantly for quantum electrodynamics, and Pauli for the Pauli exclusion principle. Toward
the photoelectric effect, Einstein’s contribution demonstrated the absorption of an incident photon leading to the ejec-
tion of an orbital electron as one based on discrete values of energy rather than merely increasing the intensity of a
lower frequency photon.
The knowledge on science is far from complete; the challenges that remain and are being pursued are the under-
standing of elementary particles and the search for a grand unification theory.

1.1 The atom

The Bohr-Rutherford model, proposed in 1913, resembles our solar system with planets going around the sun. This
model was useful in explaining several observations but had serious deficiencies; one of them was that the release of
electromagnetic radiation by the orbiting electrons would lead to a spiraling-in movement and an ultimate collapse of
the atom. Thus improvements had to be made and three postulates were proposed regarding electron orbits, discretized
angular momentum, and the ability of electrons to move from one orbit to another by discretized energy exchange.
With all the model contributions and research into new directions for the search for the smallest “indivisible” parti-
cles we come to the present knowledge of the atom in which we understand that all atomic and subatomic particles are
built from elementary particles which are quarks and leptons.
Nuclear Engineering. DOI: https://doi.org/10.1016/B978-0-323-90618-0.00001-6
© 2022 Elsevier Inc. All rights reserved. 1
2 Nuclear Engineering

Thus an atom has a dense nucleus at its center, as shown in Fig. 1.1, consisting of neutrons with rest-mass mn and pro-
tons with rest-mass mp surrounded by a “cloud” of electrons with rest-mass me (Table 1.A1, Fundamental Physical
Constants). An atom has an atomic number Z which is the number of protons or electrons, and a mass number A which is
the number of protons and neutrons; thus A 5 N 1 Z. It is written as ZAX. The masses of atomic and subatomic particles
are so small that they are expressed in terms of the atomic mass unit 1 u is defined as 1/12th the mass of an unbound car-
bon C12 atom (1 amu 5 1:66054 3 10227 kg). Neutrons and protons are made from quarks, which are indivisible elemen-
tary matter particles assigned properties of up, charm, top, down, strange and bottom. A neutron, as shown in Fig. 1.1, is
two “up” quarks and 1 “down” quark while a proton is two “down” quarks and one “up” quark. The electron is itself an
elementary particle in the family of leptons which consists of six particles. Of the four forces of nature: electromagnetic,
strong, weak and gravitation, the nucleus is held together by the strong force carried by gluons. A more detailed insight
into elementary particles would fall beyond the scope of this book and the reader can consult books and papers in particle
physics (Kibble, 2015; Peskin, 2019). The neutron is stable as a bound particle within the nucleus but unstable as a free
neutron decaying to a proton, electron and antineutrino with a mean lifetime of 885.7 6 0.8 s.

FIGURE 1.1 Bohr’s model of an atom.

TABLE 1.A1 Fundamental physical constants.

Avogadro’s number Nav 6:022141 3 10223
Planck constant h 6:62609004 3 10234 m2/kg/s
Electron charge e 1:602176565 3 10219 C
Atomic mass unit amu 1:66054 3 10227 kg
Mass of neutron mn 1:67493 3 10227 kg
1.0086649156 amu
939.550 MeV
Mass of proton mp 1:67262 3 10227 kg
1.0072764663 amu
Mass of electron me 9:10938 3 10231 kg
5:485797 3 1024 kg
Bohr radius rB 5:291772 3 10211 m
Classical electron radius re 2:817940 3 10215
Rydberg constant R 10973731.568/m

Source: From https://physics.nist.gov/cgi-bin/cuu/Category?view 5 html&Atomic 1 and 1 nuclear.x 5 85&Atomic 1 and 1 nuclear.y 5 11

 The atom and nuclear radiation Chapter | 1 3

A measure of the size of the atom was obtained by Bohr from the radius of an electron orbit in the hydrogen atom
from (1) the discretized angular momentum, and (2) the coulomb and centripetal forces
h
L 5 me vr 5 n
2π
Zq2e me v2e
k 5
rn rn
for which the radius of the nth shell is
n2
rn 5 rB
Z
where the Bohr radius is
h2 o
rB 5 5 0:52918 10210 m 5 0:52918 A
4π2 m 2
e kqe

in units of an Angstrom (1 Å 5 10210 m) used for an atomic dimension. Similarly the unit of fermi (1 fm 5 10215 m) is
used for expressing the size of a nucleus. The radius of a nucleus, defined as the distance from its center to where its
density falls to a half of its value from the center, is R 5 R0 A1=3 where R0 5 1:2 fm.
For a hydrogen atom (Z 5 1, n 5 1) the Bohr radius is thus an estimate of the atomic radius. The energy of the shells
from similar arguments is
Z 2 e4 m0
En 5 2
2ℏ2 n2
The dual idea of particle and wave, and the nature of light is “connected” through the de Broglie wavelength
h
λ5
p
where h is Planck’s constant and p is the momentum given by p 5 mv:

Exercise 1:
1. From the definition of Avogadro’s number, estimate the diameter of a carbon atom (density of carbon ρ 5 2:26 g/
cm3, A 5 12 g (g-atom)21 and “packing fraction ε” ð0 , ε , 1).
2. Calculate the radius of a carbon nucleus and compare the dimensions of the nucleus and the atom. Can the nucleus
be considered to be a “point” nucleus as Rutherford did in his famous gold foil experiment?
In the Bohr model, there are four quantum numbers assigned to each electron: the principal quantum number n, the
azimuthal quantum number l 5 0; 1; 2; ?n 2 1, the magnetic quantum number ml 5 2 l; 2 l 1 1?0; l 2 1; l thus a total
of 2l 1 1 states, and the spin quantum number ms 5 2 1; =2; 1; =2. These numbers refer to the size of the orbit (n 5 1 is
the innermost orbit, the next is n 5 2 and so on) and its shape (l 5 0; 1; 2; 3,. . . called the s, p, d, f,. . . type have spheri-
cal, dumbbell or complex shapes and are related to the energy of the electron.)

At the level of an atom, the equivalent Newton’s second law of motion is the Schrodinger wave equation which
gives the wave function Ψðx; tÞ for an electron when the quantum numbers i.e. energy descriptors are specified. The
Schrodinger wave equation for nonrelativistic particle is


@ ℏ 2 @2
iℏ ½Ψðx; tÞ 5 2 1 V ð x; t Þ Ψðx; tÞ
@t 2m @x2
with classical and quantum “connections”
@
E ! iℏ ; P ! 2 iℏr
@t
The Schrodinger equation can be solved for a hydrogen atom, to show that the solution is a function of the radial
position r, orthogonal and azimuthal angles θ; ϕ and the three quantum numbers n; l; m. The closest shell, called the K
4 Nuclear Engineering

shell, has n 5 1 shell, for which the maximum number of electrons permissible, 2n2, is two electrons. Generally, the fill-
ing order for shells is 1s, 2s, 2p, 3s, 3p, 4s, 3d, 4p, 5s, . . . The Pauli exclusion principle states that electrons that occupy
the same orbital must have different spins, that is, (2l 1 1) electrons of one spin can fill an orbit while (2l 1 1) electrons
of another, so that 2(2l 1 1) electrons can fill an orbit; thus for l 5 1, 6 electrons can fill an orbit for n 5 2. The follow-
ing cases illustrate the filling order: hydrogen (Z 5 1) 5 1s1, helium (Z 5 2) 5 1s2, lithium (Z 5 3) 5 1s2 2s1 5 [He] 2s1,
and calcium (Z 5 20) 5 2s2 2p6 3s2 3p6 4s2. Uranium is an example where the total angular momentum filling-in rule
gives for (Z 5 92)1s 2 (2s2 2p6) (3s2 3p6 4s2 3d10) (4p6 5s2 4d 5p 6s 4f 5d 6p 5p6 4f14 5d10 6s2 6p6 5f3 6d1 7s2) 5
2,8,18,32,21,9,2 electrons.
Electrons can travel fast, close to the speed of light, for which their kinetic energy (KE) is related to their momen-
tum through the relativistic relation
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
KEr 5 ðmc2 Þ2 1 p2 c2 2 mc2

for which the wavelength is

2πℏc 2πℏc
λ5 5 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pc KEr 1 2KEr ðmc2 Þ
2

For a hydrogen atom, the MATLAB program below computes the energies and velocities of orbital electrons in the
ground state (n 5 1) and two excited states (n 5 2, 3).

MATLAB program 1: .

The energies of the electrons in the first three states are 213.6, 23.39, and 21.51 eV, respectively. The lowest
(ground) state has the largest negative value while the highest excited state would have the lowest value and a free elec-
tron would have a zero binding energy (BE). In these three states, the electrons have speeds less than 1% of the speed
of light and thus can be considered as nonrelativistic.

Example 1: Use both nonrelativistic and relativistic expressions to calculate the KE of an electron (me 5 9:11 3 10231
kg) with speed 0.01c and 0.9c. The classical KEc and relativistic KEr KE expressions are

1
KEc 5 mv2
2
and
KEr 5 ðγ 2 1Þmc2
 The atom and nuclear radiation Chapter | 1 5

where
1
γ 5 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2
1 2 v=c

v γ KEc ðJÞ KEr ðJÞ KEc =KEr

0.01c 1.0001 4:0995 3 10218 4:0998 3 10218 0.9999
0.9c 2.2942 3:3206 3 10214 1:0611 3 10213 0.3129

Clearly, the relativistic kinetic energy KEr becomes larger than the classical kinetic energy KEc and in the limit
v-c; KEr -N.
Nuclei, like atoms, have excited states at discrete levels and decay by emitting alpha, beta or gamma radiation. The
emitted radiation subsequently interacts with matter in a number of ways described below. Just as atoms take part in
atomic reactions (where the electrons participate), nuclei take part in nuclear reactions which will be discussed in detail
in Chapter 2. In nuclear engineering, two energy-producing reactions are fission and fusion. In nuclear fission energy is
released by the breaking-up of heavy nuclei such as uranium-235 and plutonium-239 into lighter nuclei. In nuclear
fusion, the reaction of the sun and the stars, energy is released by the “fusing” or joining of light nuclei such as isotopes
of hydrogen, into heavier nuclei such as helium. Fission is the basis of the present nuclear fission power reactors while
nuclear fusion has yet to be achieved. The biggest fusion experiment is the 35-nation International Tokamak
Experimental Reactor (ITER) located in France, which is expected to begin operation in 2025.

1.1.1 Nuclear stability

Inside the nucleus, the two competing forces that determine stability are the attractive nuclear force between neutrons
and protons and the repulsive electric force between protons. For Z . B20, N/P . 1 as seen in Fig. 1.2 and thus the
nuclei are “above” the straight line N 5 Z. Neutron-rich nuclides get rid of their excess neutrons by emitting beta parti-
cles and at the very high end, for high A, nuclei emit heavier alpha particles. Nuclei with “magic numbers” for Z or (A-
Z): 2, 8, 20, 28, 50, 82, 126 are strongly bound, more than their neighbors. Doubly magic numbers for A 5 4, 16, 40,. . .,
are bound even stronger than those with magic numbers.
In nature 176 of the 286 (over 60%) primordial stable nuclides have even mass number; 146 stable nuclides are
even-even Z, A others are mixed while very few (B3%) are odd-odd.

FIGURE 1.2 Nuclear stability (Z-N).

6 Nuclear Engineering

1.1.2 Binding energy

The nucleus is held together with a (negative) BE in which both pair up with opposite spins. This BE is the energy
equivalent of the difference, called the “mass defect Δm” between the mass-energy of constituent nucleons
ðZmp 1 Nmn Þc2 and the nuclear mass-energy Mn c2 ; thus by Einstein’s equivalence BE 5 Δmc2 .
For nuclides with A . 20, the BE can be calculated by semiempirical “liquid drop model”
2 Z2 ðZ 2N Þ2 ð21ÞZ 1 ð21ÞN aP
BEðA; Z Þ 5 aV A 2 aS A3 2 aC 2 a A 1
A1=3 A 2 A1=2
where aV 5 15:56MeV, aS 5 17:23MeV; aC 5 0:697MeV, aA 5 23:285MeV, aP 5 12:0 0. The first term, proportional to
the mass number A, represents the volume terms with a nucleon modeled as interacting only with its neighbors rather
than with all other nucleons. The following three negative terms reduce the BE. The surface reduction is proportional to
the surface area Br2, or A2/3. The next reduction term is due to coulombic repulsion between protons. The fourth term
is a reduction due to the Pauli exclusion principle which allows two protons or two neutrons of opposite spin in each
energy level. The last “pairing” term represents an experimentally observed phenomenon inversely proportional to the
square root of the atomic mass number. Binding energies re plotted in Fig. 1.3 which shows low values at the left end
where nuclear fusion, the reaction of the sun and the stars takes place. Both these reactions make the system “move”
toward an increase in the BE per nucleon and are accompanied by the release of energy. This is followed by an increase
up to 8.7542 MeV/nucleon for chromium, manganese, iron (8.8 MeV), cobalt, nickel, copper and zinc, then gradually
falling to 7.5855 MeV/nucleon for uranium-238.
A nuclear reaction
 has Q value,
  as the difference of the rest-mass energies of the reactants r1 ; r2 and pro-
defined
ducts p1 ; p2 Q 5 mr1 1 mr2 2 mp1 1 mp2 c2 5 Tp1 1 Tp2 2 ðTr1 1 Tr2 Þ 5 ½ðBEp1 1 BEp2 Þ 2 ðBEr1 1 BEr2 Þ KE T is exo-
thermal or exoergic if Q . 0 and with endothermic or endoergic if Q , 0.

Exercise 2: From the BE formula above, explain why the elements with odd Z are expected to be less abundant in
nature?

1.2 Radioactive decay

When an unstable radionuclide spontaneously emits energy, it is said to be radioactive to an extent measured by the
number of its disintegrations per second with 1 Becquerel (Bq) representing 1 disintegration per second. A commonly
used unit is the curie with 1 curie (Ci) denoting 37 billion Bq.
Decay data of over 3000 radionuclides are maintained by Brookhaven National Nuclear Data Services and published
monthly in Nuclear Data Sheets (Elsevier) which are produced mainly from the Evaluated Nuclear Structure Data File
(ENSDF) maintained by the US National Nuclear Data File Center http://www.nndc.bnl.gov.

FIGURE 1.3 Binding energy per nucleon from the liquid drop model.
 The atom and nuclear radiation Chapter | 1 7

Radioactivity arises in nuclear reactors due to fission reactions resulting in fission fragments such as cesium and
iodine. The number of such radioactive atoms can become significant, for example, about 6 atoms of cesium-137 are
produced per 100 fission events. Considering that one fission reaction produces 3.20 3 10211 J and that a 3000 MW
(th) reactor would be undergoing B 1020 fissions per second, the production of cesium-137 would be B 6 3 1018
atoms per second, or B 2 3 1026 atoms per full operating year. Later in this chapter, we will see how to estimate this
quantity of atoms into mass from the Avogadro number.
Radioactive decay results in a “decay chain” where further nuclides are produced. Consider the decay of iodine-135
atoms due to its decay constant λI
d
I ðt Þ 5 2 λI I ðt Þ
dt
At any instant, the number of iodine-135 atoms is then
I ðtÞ 5 Ið0Þe2λI t

 By the definition of half-life t1=2 is the time for half the atoms from the beginning to have remained; thus
I t1=2 5 Ið0Þ=2 so that
1 2λ
5 e It1=2
2
yielding a relationship between the decay constant and half-life:
λt1=2 5 ln2 5 0:693
The activity of a radionuclide (disintegrations per second) is then A 5 λN Bq.
Iodine-131, for example, decays to xenon-131 through two reactions, namely
131 8:02d 131
53 I - 54 Xe  1 β 1 νe
which is the predominant reaction, and
11:9d 131
131
54 Xe  - 54 XeðstableÞ 1 γð0:39% of reactions in 131
53 I decayÞ

In a radioactive chain, there is the progression of a nuclide, such as tellurium-135, shown below. This results in the
production of iodine-135, which successively decays to xenon-135, cesium-135, and barium-135.

52 T e-53 I -54 Xe-55 Cs-56 Ba

135 135 135 135 135

Just as the iodine-135 decay rate equation above, we can write down the equations for the number of atoms of each
of the radioisotopes in the tellurium decay chain to obtain their concentrations as a function of time.
We shall see the important consequences of this decay chain in Chapter 3 in the context of nuclear reactor control.
For now, we note that iodine has a “source” term (production by tellurium) and a “loss” term (decay to xenon-135).
This permits us to write an ordinary differential equation for the number of atoms of iodine-135 IðtÞ in terms of the
number of atoms of tellurium-135 T ðtÞas
d
I ðtÞ 5 λT T ðtÞ 2 λI I ðtÞ
dt
Similarly, for xenon and cesium,
d
X ðtÞ 5 λI I ðtÞ 2 λX X ðtÞ
dt
d
C ðtÞ 5 λX X ðtÞ 2 λC C ðtÞ
dt
The above first-order ordinary differential equations, also known as the Bateman equations, can be expressed in
matrix form and solved as an initial value problem
d
BðtÞ 5 λBðtÞ
dt
8 Nuclear Engineering

FIGURE 1.4 Radioactive decay of iodine-135 to xenon-135.

FIGURE 1.5 The process of spontaneous alpha decay.

The solution for I ðtÞ and X ðtÞ ignoring the initial concentration of T ðtÞ as well as the third equation for CðtÞ, an exact
solution can be obtained by standard methods.
With initial conditions I ð0Þ 5 100 atoms and X ð0Þ 5 0, the solution, obtained with MATLAB (Annex 7) is shown in
Fig. 1.4. Iodine-135 with a half-life of 6.6 h and xenon-135 with 9.2 h reaches a maximum xenon strength of 43.0370
atoms at 11.2 h. These decays and buildups are calculated in the event of a nuclear accident to determine the dynamics
of radiation effects.

1.2.1 Alpha decay

The radioactive decay of heavy unstable nuclides shown in Fig. 1.2, with alpha particles, as shown in Fig. 1.5, results
in the emission of a helium-4 nucleus and hence is associated with reduction of atomic number by 2 and of atomic
mass by 2. Early work on radiation by Becquerel, Rutherford and Curie showed that radiation is emitted by atoms
which change their form. The discovery of polonium and radium in 1898 from its presence in uranium ores, and the
subsequent isolation of radium in metallic form by Marie and Pierre Curie in 1911 led to several experiments.
Compounds of radium showed a faint blue light in the dark which was caused by the excitation, and subsequent de-
excitation, of electrons in elements of the compounds caused by the spontaneous emission of radiation from radium. It
was then discovered that the element has several isotopes which undergo alpha decay; for example, the isotopes
Ra223 ; Ra224 and Ra226 have a half-life 11.43 d, 3.66 d, and 1600 y each. The radium decay chain of Ra226 is
Ra226 !α;1600 y Rn222 !α;3:82d Po218 !α;3:05m Pb214 !β;26:8 m Bi214 !β;19:7 m Po214 !α;0:16 ms Pb210 . . .
!β;22y Bi210 !β;5d Po210 !α;138d Pb206 ðstableÞ
decaying ultimately to Pb-206 which is stable. All radioactive chains end with a stable product. In this chain, radon-222
is an odorless radioactive decay present in earth as background radiation due to its transient production from uranium
and thorium.
 The atom and nuclear radiation Chapter | 1 9

Other powerful alpha emitters include Polonium-210 emitting a 5.305 MeV alpha particle with the emission of one
milligram equal to that from 5 g of radium and Plutonium-238 emitting a 5.456 MeV alpha. The alpha decay reactions
of uranium-238 and plutonium-238 are

92 U-90 Th 1 2 He; Qα 5 4:268 MeV; T1=2 5 4:51 3 109 y

238 234 4

238 234
94 Pu-92 U 1 42 He; Qα 5 5:593 MeV; T1=2 5 87:7 y
The energy of alpha particles ranges from B1 to B 11 MeV, although most alpha emitters produce alpha particles
in the range 45 MeV. The number of alpha emissions is a function of the activity of an alpha emitter for which some
commonly used sources are: Am-241, Pu-238, Pu-239, Po-210, Ra-226; and the portable sources, for example, Am-Be,
Pu-Be, Ra-Be of 1050 mCi. Alpha particles led to several new developments including an understanding of the struc-
ture of an atom in Rutherford’s gold foil experiment and the discovery of the neutron by Chadwick by the alpha bom-
bardment of beryllium in the reaction 49Beðα; nÞ612C.

1.2.2 Beta decay

In beta decay, a neutron-rich nuclide decays by the conversion of a neutron into a proton and an electron accompanied
by the emission of an antineutrino. The antineutrino is an antiparticle (same mass but opposite electric charge and mag-
netic moment) of the neutrino. It is a lepton like an electron which is an elementary particle of spin 1/2. Unlike an elec-
tron the antineutrino has no electric charge. Antineutrinos cannot be easily detected as they pass the earth without any
interactions. The beta decay of carbon-14 is
14 6 14 7
6 C -7 N 1 2 10β 1 ν; T1=2 5 5730y

Exercise 3: How is the beta decay of carbon-14 used in “carbon-dating” to estimate the age of fossils?

For excess of neutrons (beta β 2 decay)

n-p 1 β 2 1 ν
while excess of protons (positron β1 decay) leads to the emission of a positron, which is an antiparticle of an electron
(Fig. 1.6)
p-n 1 β1 1 ν
Reactions follow conservation laws for charge and lepton number conservation. The radioactive decays of iodine-
131 and cesium-137, both powerful beta emitters with significant half-life and associated with cancer resulting
from nuclear tests and explosions are shown in Figs. 1.7 and 1.8, respectively. Some positron emitting radionuclides
with half-lives and mean energies are 611Cð20:38 min; 0:385MeVÞ; 713Nð9:96 min; 0:491MeVÞ; 917F;
ð64:5 s; 0:739MeV; 918F; ð109:77 min; 0:2498MeVÞ and 1938Kð7:636 min; 2:323MeVÞ:

FIGURE 1.6 The process of spontaneous beta decay.

10 Nuclear Engineering

FIGURE 1.7 Radioactive decay of iodine-131.

FIGURE 1.8 Radioactive decay of cesium-137.

FIGURE 1.9 The process of spontaneous gamma decay.

1.2.3 Gamma decay

Gamma decay takes place as a result of the relaxation process of a nucleus in its excited state just as X-rays are emitted
in the relaxation process of electrons. With discrete energy levels for nucleons, determined by quantum physics, each
excited state is characterized by quantum numbers. The nuclear Shell Model describes the motion of each nucleon
based on the forces it experiences from neighboring nucleons. Since the nucleus is tightly bound, the energy of gamma
rays are much higher than that of X-rays, typically MeV compared with keV. A comprehensive list of intense (.
1 keV) gamma rays, based on the ENSDF Brookhaven National Laboratory data, is given in Narita and Kitao (1992)
(Fig. 1.9)
One of the strongest gamma rays produced has an energy of 7.11515 MeV (4.9% intensity) from the beta decay of
716N (Bielajew, 1990). One method of determining uranium enrichment is from the alpha decay of uranium-235 to
thorium-231 which produces 4:3 3 104 gamma rays, of 185.7 keV, per gram uranium-235 per second.
 The atom and nuclear radiation Chapter | 1 11

It is important to note that energetic gamma rays can produce neutrons as in the photo-neutron ðγ; nÞ reactions
49Beðγ; nÞ48Be and 12Hðγ; nÞ11H.

1.2.4 Radioactive nuclides in nuclear technologies

Nuclear power reactors: In nuclear power reactors, the main source of nuclides is the nuclear fission reaction which
produces fission products as will be discussed in detail in the next chapter devoted to neutrons. These fission fragments
give rise to the radiations discussed here. Some fission fragments are a problem in the operations of nuclear reactor;
these are called poisons such as 53131Xe, which can cause a reactor to shut down for some time. This results in a loss
of power which is serious for reactors on ground but more so for propulsion reactors in submarines. In some cases,
“poisons” such as burnable boron solutions are used to control the reactor from staying within permissible power
bounds.
Radiation in nuclear weapons and accidents: In the early days of nuclear technology, weapons testing led to the
release of radioactive nuclides in the atmosphere. In the 1945 nuclear explosions over Hiroshima and Nagasaki, casual-
ties occurred due to the immediate blast and long-term radiation mainly from iodine and cesium isotopes. The iodine
isotope 53131I is harmful and had been found over in the Chernobyl accident, one of the four nuclear reactors of the
former Soviet RBMK-1000 Light Water Graphite (LWGR) were destroyed resulting in the release of B 5% of radioac-
tive core (Anspaugh, Catlin, & Goldman, 1988; World Nuclear Association, 2012) in the environment causing 28 fatali-
ties, within weeks, due to acute radiation syndrome. In the Fukushima disaster, caused by the Tohoku 9.0 magnitude
earthquake and tsunami, the radiation released was B18 1015 Bq of radioactive Cs-137 into the Pacific during the acci-
dent (Langlois, 2013). Water and soil samples nearby indicated the presence of strontium, iodine and cesium. The
Japanese Red Cross reported that as of 2011, the casualties consisted of almost 16,000 people confirmed dead, almost
4000 missing and that approximately 90% of the deaths were due to drowning (INPO, 2011).

Exercise 4: Estimate the fraction of cesium and iodine concentrations from the Hiroshima and Nagasaki blasts remaining
from 1945.

Space technologies: The exploration of space, starting from 1957 with increased interest in the 1960s followed by a
long gap, has revived this year (2020) with three missions to Mars launched in July by China (Mallapaty, 2020), UAE
(Gibney, 2020), and the United States (Witze, 2020). Traveling at about 20,000 kilometers per hour, they are expected
to complete the 500 million kilometer journey to Mars in about seven months. In the US mission, the Perseverance
Rover uses radioactive plutonium-238 for power. As described earlier, plutonium-238 is an alpha emitter but also emits
beta and gamma radiations as well as neutrons from spontaneous fissions (Matlack & Metz, 1967). Thus for future mis-
sions, possibly with personnel in returnable rockets, radiation shielding would be required. Plutonium-238 generating
0.55 W/g (th) with a half-life of 88 years and strontium-90 giving 0.93 W/g (th) with a half- life of 28 years as well as
others such as polonium-210, cesium-144 and curium-242 generate 141, 25, and 120 W/g (th) with lower half-lives of
0.378, 0.781, and 0.445 years, respectively, can all be used. In the Mars 2020 Perseverance Rover, plutonium-238 has
been used in the form of plutonium oxide as in several previous US space missions. The radioactive decay heat in space
systems can be converted into electrical energy by several methods such as the conventional ones used in power plants
based on the Rankine or Brayton cycles, or by conversion to mechanical energy in a Stirling engine or by directly con-
verting to electrical in a thermoelectric generator as in the Perseverance Rover. Such systems have a long history espe-
cially for space systems (Bennet, 1989; Demuth, 2003; Determan et al., 2011; El-Genk, 2008; Miskolczy & Lieb,
1990). NASA’s latest multimission radioisotope thermoelectric generator (MMRTG) is based on earlier SNAP (Small
Nuclear Auxiliary Power) systems uses General Purpose Heat Source (GPHS) Modules with a thermoelectric generator
and radiator funs, to produce electrical power without moving parts and to use heat for spacecraft operation and temper-
ature control.
Medical therapy: Over 200 radioisotopes are used in medicine such as technetium 99mTc, iodine 125I, palladium
103
Pd, iridium 192Ir, cesium 137Cs, and cobalt 60Co, for diagnostics and cancer therapy (Harkness-Brennan, 2018). They
are used both for the imaging of organs, such as the thyroid, bones, heart, liver, and for the treatment of cancer of lungs,
breast, colon and rectum, prostate, stomach, and liver, etc. The photons, protons, and positrons from these radiations are
also used in tomography such as in Positron Emission Tomography (PET), in medical imaging procedures (Gray;
Ulaner, 2018) using radio-tracers to determine changes such as in blood flow or in the condition of the heart muscle. In
the COVID-19 pandemic beginning from 2019, PET has also been used (Loforte, Gliozzi, Suarez, & Pacini, 2020) for
the detection of potential COVID-19 respiratory syndrome.
12 Nuclear Engineering

A relatively new therapy, boron neutron capture therapy (BNCT), based on the short range of alpha particles is
being developed for the treatment of brain tumors. Improvements in cancer therapy, such as dose enhancement by radia-
tion by nanoparticles and better computational methods for its estimation, are the focus areas of research. The reaction
10
5 B 1 10 n-73 Li 1 42 H e
is used in boron neutron capture therapy (Kumada, 2014; Nedunchezhian, Aswath, & Thiruppathy, 2016; Suzuki, 2020) with
the product ions producing very high linear energy transfer due to high stopping power and high ionization with  150 keV/
μm from the alpha particle and  175 keV/μm from the lithium ion in pathways ,10 μm which is comparable to the diame-
ter of one cell. One of the obstacles in the development of BNCT has been the thermal neutron flux required for the reaction;
for this the use of a tandem particle accelerator has been achieved with the required values of current and energy for the
nuclear reaction to take place with increased quality (Dymova, Taskaev, Richter, & Kuligina, 2020).

Exercise 5: In what medical applications would alpha radiation be preferred over beta radiation?

Nondestructive testing: In industry, due to the penetrating power of gamma radiation with an ability to go through
commonly used dense materials such as concrete and steels, nondestructive testing (NDT) is an established discipline
with widespread applications covering the testing of roads, bridges, pipes and structural materials. The use of radiation
for elemental identification and assaying, such as determining the purity of gold, is also carried out accurately with the
use of X-ray fluorescence (XRF) using the absorption edges of the photoelectric cross-section.

1.3 Interaction of radiation with matter

As highlighted above, radiation is an area of major concern in nuclear engineering and thus its interaction with matter
is central to calculations of activity of radionuclides as they are formed and as they decay in various radioactive chains.
Radiation, both ionizing and nonionizing, interacts in several ways. Some significant mechanisms are discussed
below (Fig. 1.10).

1.3.1 Interaction of alpha rays with matter

As described in Section 1.2.1, alpha particles are energetic ionizing radiation consisting of a helium nucleus with an
electrical charge of two units.
As alpha particles move in matter, they “see” atoms on the scale of a few Angstroms with “lots” of electrons and a
tiny point nucleus on the scale of a few fermis. It is thus more probable that they interact with the electrons. In their
movement, they constitute a charged particle with charge 12e moving in the vicinity of several electrons of charge e
moving in orbits of the target medium. The force between the alpha and an electron, in Newtons, is given by the cou-
lomb field with the force
kð2eÞðeÞ
F5
r2
where k is the conversion constant N/m2/C and r is the distance between them. In their interaction, three events are thus
likely: removal of electrons from their orbits (ionization), excitation of the orbital electrons, or transfer of energy to the

FIGURE 1.10 Interaction of radiation with matter.

 The atom and nuclear radiation Chapter | 1 13

“point” nuclei. Thus alpha particles can be understood to lose their energy via ionization, excitation and nuclear
collisions.
Since alpha particles are heavy (compared with electrons), they move in a relatively straight line between collisions
and undergo a series of “small-angle” deflections while the electrons receive energy from them and undergo deflection
themselves. The collision can be modeled with the usual kinematics governed by conservation of momentum and
energy. They lose energy by electronic excitation and ionization creating ion pairs. Due to their charge, they can be
deflected by electric and magnetic field.
The stopping power SðEÞ which is defined as the energy loss by collisions traveled, in units of energy/cm, is due to
energy lost to electrons Se and nuclei Sn both. Thus
dE
2 5 Se 1 Sn
dx
Clearly, the stopping powers would be expected to be functions of the alpha particle’s mass, energy and electrical charge,
and the atomic number of the target material as well as its density (which would determine the electron density).
An elementary two-body interaction with coulomb potential gives the electronic energy loss as

dE 4πNk2 e4 γ 2 me v3α f ðZÞ

2 5 ln
dx me v2α ke2
With the maximum energy transfer
Wmax 5 2me γ 2 v2α
where
1
γ2 5
1 2 β2
A quantum-mechanical relativistic expression is given by Bethe-Bloch as

2
dE 4πNz2 e2 Wmax δ C
SðEÞ 5 2 5 ln 2 β2 2 2
dx β2 4πε0 I 2 Z

in terms of the projectile atomic number z and speed v, and the density of electrons N and charge of electrons e and the
ionization potential I of the slowing-down medium. The two correction terms (1) the “density effect” accounts for the
shielding of distant electrons by the polarization of electrons caused by the electric field of the high-energy projectile,
and (2) the “shell correction” (given the symbol “C”) due to effects at low energy of tightly bound electrons.
The nonrelativistic Bethe formula
2
dE 4πNz2 e2 2me v2
52 ln
dx me v2 4πε0 I
The dependence of stopping power is 1/v2 so that low speed gives high energy loss (Table 1.1).
The collisions stopping power is the energy lost per cm due to coulomb collisions with the target atoms and elec-
trons which result in ionization and excitation, while the nuclear stopping power is the loss transferred to recoiling
atoms in an elastic collision. It is worth noting that the difference between the CSDA range and the projected range
reduces for high energy
The range, defined as
ð E0
dE
R5
0 SðEÞ

can be interpreted as the mean range if at each point in the ion’s track, the stopping power is the total of the collision
and nuclear terms.
Alpha particles, which are “heavy” compared with electrons by a factor of 7273, lose their energy very quickly and
thus travel small distances in typically straight lines before inducing ionization.
14 Nuclear Engineering

TABLE 1.1 Stopping power and range of alpha particles.

2
Material Energy (MeV) Stopping power ðMeV cmg Þ CSDA Range
 g 
electronic nuclear cm2
ðcmÞ
 
Dry air ρ 5 1:225 3 1023 cmg 3 1 3 1023 8:750 3 101 1:340 3 102 5:377 3 1026 4.389 3 1023
1 3 1021 1:018 3 103 1:368 3 101 1:665 3 1024 0.1359
1:00 1:922 3 103 2:103 6:698 3 1024 0.54722
 
10 4:634 3 102 2:817 3 1021 1:309 3 1022 10.686
Water ρ 5 1 cmg 3 1 3 1023 9:891 3 101 2:282 3 102 3:273 3 1026 3:273 3 1026
1 3 1021 1:131 3 103 1:921 3 101 1:425 3 1024 1:425 3 1024
1:00 2:190 3 103 2:898 5:931 3 1024 5:931 3 1024
10 5:340 3 102 3:844 3 1021 1:130 3 1022 1:130 3 1022

Source: ASTAR data from NIST.

FIGURE 1.11 Stopping power for alpha particles in oxygen and

nitrogen.

Example 2: Calculate the stopping power for alpha particles of KE 110 MeV in oxygen and nitrogen. From these,
obtain the stopping power in air.

The MATLAB program implementing the Beth-Bloch formula (without the density factor and shell correction) is
listed below, for the results shown in Fig. 1.11. Data used for mean ionization energy and density from NIST is
included in the listing.

MATLAB Program 2: .

The values for ASTAR from NIST are reproduced below, showing excellent agreement. The stopping power for air
can be obtained by incorporating the composition of air; thus (Table 1.2)
dE dE dE
jair 5 0:79 jN 1 0:21 jO
dx dx dx
Energy loss is a statistical quantity resulting in a distribution of path lengths giving a mean value higher than the
CSDA value and a straggling is thus found. This becomes important in experiments where “projected ranges” of actual
penetration of particles are measured and theoretically correlated with the CSDA range. The energy straggling is incor-
porated in simulation codes, as will be mentioned in Section 1.6 by, for example, a Gaussian distribution function. The
mean energy loss in ionization from the Rutherford scattering cross-section is
The atom and nuclear radiation Chapter | 1 15
16 Nuclear Engineering

TABLE 1.2 Stopping Power from ASTAR NIST for alpha particles in oxygen and nitrogen.
2
Stopping power ðMeV cmg Þ
Energy (MeV) 2 4 6 8 10
Nitrogen 1:406 3 103 8:985 3 102 6:785 3 102 5:523 3 102 4:691 3 102
Oxygen 1:329 3 103 8:602 3 102 6:519 3 102 5:313 3 102 4:520 3 102

 2πz2 e4 NZx
ΔE 5
m e v2
for which the fluctuations over a large number of collisions can be modeled with a Gaussian; for thin targets, that is,
few collisions, the Vavilov and Landau distributions can be used (Peralta & Louro, 2014). The Gaussian probability dis-
tribution function (PDF) for the fraction of particles f(E) 5 N(E)/N having energy in dE between E and E 1 dE is
 2 !
1 E2 ΔE
f ðEÞ 5 pffiffiffi exp 2
πσstr σ2str

in terms of the mean energy E and the standard deviation σstr which is given by
 1
σ2str 5 ΔE Emax 1 2 β 2
2
The methods and implementation of sampling methods for PDF’s will be discussed in Chapter 4.

1.3.2 Interaction of beta radiation with matter

Compared with alpha particles discussed in the previous section, beta particles have less mass and at energy 1 MeV
they are at almost twice their rest-mass energy; they move ‘swiftly’ in comparison and thus have a longer range. At low
energy, typically less than few MeV, the energy loss by electrons (and positrons) is mainly by excitation and ionization
with little contribution from Møller (and Bhabha scattering). At high energy, typically greater than 10 MeV, the losses
are mainly from bremsstrahlung, continuous electromagnetic braking radiation in the energy range of X-rays. The “criti-
cal energy” Ec defined as the energy at which ionization and bremsstrahlung losses are equal, decreases from about
300 MeV for hydrogen to B 20 MeV for iron varying as
610MeV
Ec 5
Z 1 1:24
The collision mechanism of electrons with electrons of the same mass implies that large energy transfer is possible.
At high energies, the fraction of beta energy converted into photons is directly proportional to the atomic number of the
target and energy of the electrons: fβBr 5 3:5 3 1024 ZE.
The stopping power for electrons is the sum of the electronic ionization Se and radiative loss Sr due to bremsstrah-
lung (Berger, Coursey, Zucker,& Chang, 2005; Taylor et al., 1970)
dE
2 5 Se 1 Sr
dx
The electronic part is similar to the ion-electron energy loss in the previous section. For electrons
8 0 1 9
>
> 2  p ffiffiffiffiffiffiffiffiffiffiffiffiffi>
>
> @
> m v E >
A 2 ln2 β 2 2 1 1 2 1 2 β 2 >
e
2 <
> ln  >
=
dE 2 2
2πre me c Nz Z 2I 1 2 β
2 2
Se ð E Þ 5 2 5 q ffiffiffiffiffiffiffiffiffiffiffiffiffi
dx β2 >
>   1 >
>
>
> 1 12β 2 1 1 2 1 2 β2 >
>
>
: 8 >
;
 The atom and nuclear radiation Chapter | 1 17

FIGURE 1.12 Stopping power for beta particles in sodium iodide

(NaI).

TABLE 1.3 Stopping Power from ESTAR NIST for beta particles in sodium iodide (NaI).
2
Stopping power ðMeV cmg Þ
Energy (MeV) 2 4 6 8 10
Collision 1:192 1:263 1:311 1:347 1:374
Radiative 0:1369 0:2857 0:4464 0:6145 0:7876

and the radiative loss is


dE ðz 1 1ÞzρN E 2E 4
Sr ðEÞ 5 2 jr 5 4ln 2
dx 137m2e c4 me c2 3
The stopping power Se ðEÞ for beta particles in sodium iodide is calculated from the above expression with the
MATLAB program listed below. For the KE of beta particles in the range 110 MeV, the stopping power is plotted as
a function of energy in Fig. 1.12, which shows good comparison with the ESTAR stopping powers from NIST. The
NIST values are plotted from 1 3 1022 MeV to 1000 MeV for which Se ðEÞ decreases from 11.16 MeV cm2/g to
1.173 MeV cm2/g at 1.5 MeV then increases slowly to 1.777 MeV cm2/g at 1000 MeV. In contrast, the radiative stop-
ping power Sr ðEÞ increases from 1:518 3 1022 MeV cm2/g at 1 3 1022 MeV, equaling Se ðEÞ at B 17.5 MeV, then
increasing to 103.8 MeV cm2/g at 1000 MeV.
The ESTAR NIST results are listed in Table 1.3 with the slow increase in Se ðEÞ which is still higher than the radia-
tive energy loss.

MATLAB program 3: .

Stopping power for beta particles in NaI

The range Rðmg=cm2 ) for beta particles of maximum energy E is given by the empirical relations
R 5 412 E1:26520:0954lnE ; 0:01 MeV # E # 2:5 MeV
R 5 530 E 2 106; E . 2:5 MeV
plotted for air, polyethylene, aluminum, iron, and lead in Fig. 1.13.
Thus energetic beta particles can travel more than a meter in air and are easily shielded by a few centimeters of
lead. The effect of bremsstrahlung, for beta particles is shown in Fig. 1.14.
18 Nuclear Engineering

FIGURE 1.13 The range of beta particles as a function of energy.

 The atom and nuclear radiation Chapter | 1 19

FIGURE 1.14 Stopping power of electrons in aluminum from ESTAR.

TABLE 1.4 Stopping power and range of beta particles.

2
Material Energy (MeV) Stopping power ðMeV cmg Þ CSDA range
 g 
collision radiative cm2
ðcmÞ
 
Dry air ρ 5 1:225 3 1023 cmg 3 1 3 1021 3:633 4:222 3 10 23
1:623 3 1022 13.249
1:00 1:661 1:271 3 1022 4:912 3 1021 400.98
10 1:979 1:795 3 1021 5:192 4238.4

Source: ESTAR data from NIST.

In order to compare the range of both alpha and beta in air, ASTAR (Table 1.1) and ESTAR (Table 1.4) values are
compared; the range is far greater for beta than for alpha by factors of B100, B800, and B400.
The specific ionization of beta particles (MeV/cm)
ð sn11
dE
En11 2 En 5 2 ds
sn ds
 4
dE 2πq4 NZ 3 3 109 Em Ek β 2
5   ln 2 β2
dx Em β 2 1:6 3 1026
2
I 2 ð1 2 β 2 Þ

can be used to estimate the number of ion pairs created by a beta particle as it slows down in an absorber in which the
mean energy required to create an ion pair is W electron volts. The straggling of electrons, that is, the random differ-
ences between electrons is found from a PDF since the energy loss in a step is not the same for every step.
The collisional stopping power of water (liquid) produced by ESTAR (Berger et al., 2005) for electrons gradually
decreases with energy, till about 1 MeV, to a constant value of B2.0 MeV cm2/g while the radiative loss becomes com-
parable at B90 MeV then gradually rises to B 2.7 MeV cm2/g. The total stopping power thus decreases with energy
until the radiative term increase and then rises.

1.3.3 Interaction of gamma radiation with matter

Electromagnetic gamma rays, as discussed in Section 1.2.3, have high energy (Bseveral MeV) and are thus very strong
in terms of penetration in matter. In fact, one of the major problems in nuclear systems is gamma radiation for which
detailed transport simulations are carried out.
20 Nuclear Engineering

FIGURE 1.15 The photoelectric effect.

Gamma rays interact with matter in many different ways of which the three most significant are the photoelectric
effect, Compton scattering and pair production. Some other forms of interaction are photoneutron reactions, quasi-
deuteron disintegration and pion production at very high energies (B. 1 GeV). The three main interactions combine to
give the overall parameter which is used to calculate the overall attenuation of gamma transport in matter.
In the photoelectric effect, shown in Fig. 1.15, a photon of frequency ν and energy E 5 hν interacts with an orbital
electron of an atom, gets absorbed by the electron, and if its energy is greater than the BE of the electron, then the elec-
tron gets ejected with a maximum KE Kmax 5 hðν 2 ν 0 Þ where hν 0 is the BE or the “work function” such that the
threshold frequency of the emissive surface is ν 0 . The photoelectric effect was explained by Einstein on the basis of
quantum physics in which Planck had earlier stated that the transfer of energy takes place in integral amounts of quanta
which depends on the frequency and not on the intensity of light. This was a revolutionary difference from classical
physics in which light was considered a wave. Thus in the photoelectric effect, it is the frequency of the incident photon
rather than its intensity which results in the ejection of an electron.
Ee 5 E γ 2 EB
When the host material is a high Z shield such as iron, lead, or tungsten, the number of electrons produced by the
photoelectric effect is large. The X-rays produced from the photoelectric absorption are called fluorescent radiation and
are used for element characterization. After the photoelectric effect has knocked out an electron, an outer electron can
fill the vacancy to produce the fluorescent emission. Subsequently the excess energy can eject another outer shell elec-
tron producing an Auger electron. Similar events can be triggered in subshells. XRF is used as a nondestructive tech-
nique (NDT) for multielemental identification such as determining the purity of gold (Handbook of Practical X-Ray
Fluorescence Analysis, 2006).
For the photoelectric effect to take place, the frequency of an incident photon must be higher than one of the fre-
quencies correspond to the binding energies of the “saw-tooth” absorption edges for K, L,.,. shell electrons shown for
lead in Figs. 1.16 and 1.17. Usually, the most important photoelectric effect is for the K shell electrons, for which the
approximate formula is EK 5 ðZ 21Þ2 13:5eVB88.573 keV (the square of the effective nuclear charge times the Bohr
energy orbit energy for the first orbit in hydrogen.
The probability of a photoelectric interaction σPE depends on the energy of the incident photon, the atomic number
of the element and the BE of the electrons it is proportional to Z n =E3 , n 5 44.8.
The shells and associated energies of elements from hydrogen to lawrencium are K, L, M1 to M7, N1 to N7, O1 to
O5, P1 to P5, Q1. BE in subshells of a free atom (eV) are listed in Table 1.5.
High Z materials would therefore have more photoelectric absorption and would thus be good gamma shields. Thus
one of the best shields is depleted uranium with a density of 19.1 g/cm3, about five times better than lead with a density
of 11.35 g/cm3.
 The atom and nuclear radiation Chapter | 1 21

FIGURE 1.16 The mass attenuation coefficient for the photoelectric

effect.

FIGURE 1.17 The photoelectric absorption edges for lead (Pb).

TABLE 1.5 Shell energies (eV).

Z Element Electronic configuration K L1 L2 L3 M1 ... Q1

1
1 H 1s 13.60
2 He 1s2 24.59
3 Li [He] 2s1 58 5.392
4 Be [He] 2s2 115 9.322
5 B [He] 2s22p1 192 12.93 8.298
^ ^ ^ ^ ^ ^ ^ ^ ^ ^
20 Ca 1s22s22p63s23p64s2 4041 441 353 349 46 . . .N2(6.113)
82 Pb [Xe] 4f14 5d10 6s2 6p2 88000 15860 15200 1304
92 U [Rn] 5f36d17s2 115611 21762 20953 17171 5553 ... 6
98 Cf [Rn] 5f107s2 134967 26008 25103 19907 6733 . . .N3 (1273)
^ ^ ^ ^ ^ ^ ^ ^ ^ ^
103 Lwa [Rn] 5f147s27p1 153040 30091 29101 22356 ... 7
a
Lawrencium symbol changed by IUPAC to Lr.
22 Nuclear Engineering

Forpaffiffiffi photon 2of 7=2high energy, in comparison with the K shell BE, the atomic absorption is σK
Z5
σK 5 4 2φ0 137 4
me c
hν
where
8π e2
φ0 5 5 0:6651 3 10224 cm2
3 me c2
and scattering is considered as that of scattering of light from free electrons given by the Thompson cross-section φ0 in
the nonrelativistic case hν{me c2 but modified to the Klein-Nishina cross-section for relativistic energies.
At higher energies than that compared with K shell binding energies, typically me c2 5 0:511MeV the photoelectric
effect becomes less important and scattering of photons, called Compton scattering, is considered as taking place with
free electrons. This is reasonable since the K shell edge goes to B0.11 MeV for the heaviest nuclei. Coherent scattering
is classical scattering or Thomson scattering with the cross-section φ0 for low-energy (in comparison with the ionization
energy) photons passing in the vicinity of an outer electron of an atom. This causes the electron to vibrate with the fre-
quency of the photon and to emit another photon of the same frequency and energy, but different direction, of the
incoming photon which no longer exists. Coherent scattering is similar to elastic scattering in the sense that energy is
not lost in the interaction. In contrast to coherent scattering, incoherent scattering changes the frequency and energy as
if the electrons were randomly fluctuating.
The kinematics of coherent scattering was first shown by Compton as a two-body collision (Fig. 1.18) where the
massless photon could be treated as a particle. In Compton scattering, at intermediate energies (few hundred keV to
low MeV) and predominantly at low-Z material, a photon transfers some of its energy to an electron and continues to
undergo multiple scattering. A photon of energy E 5 hν will have momentum p 5 E=c 5 hν=c with wavelength
λ 5 c=ν.
In Compton scattering, a collision results in scattering and recoil which changes the frequency of the photon.
Consider the collision of an energetic (massless) photon with an electron of rest-mass m0 with rest energy E0 5 m0 c2 in
which the photon is scattered to the right upwards an angle θ and the electron recoils to the right downwards an angle
ϕ.
The pre- and postcollision two-body parameters are shown below.
Collision parameters
“Particle” Energy Momentum

Before collision After collision Before collision After collision

0 0 0
Photon E 5 hν E 5 hν p 5 h=λ 0
p 5 h=λ
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Electron E0 5 m0 c 2
Ee 5 mc 2 pe 5 0 0
pe 5 1c Ee2 2 E02

With conservation of momentum and energy the equations are:

0 hc
λ 2λ5 ð1 2 cosθÞ 5 λc ð1 2 cosθÞ
E0
0
hν 1 E0 5 hν 1 Ee
where the Compton wavelength λc 5 2:426 3 10210 cm

FIGURE 1.18 Compton scattering for a photon.

 The atom and nuclear radiation Chapter | 1 23

FIGURE 1.19 Pair production for a photon.

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Ee 5 ðp0e cÞ2 1 E02
0
The energy of the photon after collision E is thus found to be related to its energy before collision E and it scatter-
ing angle θ by
1 1 1
2 5 ð1 2 cosθÞ
E0 E E0
0
As seen, λ 5 λ for forward scattering θ 5 0 and the maximum change is for backscattering θ 5 πc for which the
electron gets maximum energy. The probability of Compton scattering is given by the Klein-Nishina (K-N) differential
Ð dσ is, the probability that the photon gets scattered into solid angle dΩ 5 2πdθ, so that the total cross-
cross-section, that
section is σ 5 dΩ dΩ, and
0
!
dσ 1 2 λ 2 λ λ
5 r 0 0 1 2 sin θ
2
dΩ 2 e λ λ λ

with units of barns per steradian. At low incident energy, in comparison with the electron energy me c2 when the scatter-
ing is close to elastic, the final and initial wavelength are the same, while for high-energy scattering is inelastic. In the
Thomson regime, the K-N cross-section reduces to the Thomson “unpolarized” cross-section
dσ 1  
jT 5 re2 2 2 sin2 θ
dΩ 2
The total cross-section for an electron is


3 2ðE 1 1Þ 1 4 1 hν
σC 5 φ0 1 2 lnð2E 1 1Þ 1 1 2 ;E5
8E E 2 E 2ð2E11Þ 2 me c2
For an atom, the Compton scattering cross-section is thus NσC . The K-N cross-section reduces to the value from the
classical Thomson as forward scattering of the photon increases with increasing energy.
At yet higher energies, pair production is the dominant interaction mechanism for gamma resulting in energy depo-
sition and the subsequent creation of a beta particle and a positron as shown in Fig. 1.19. This ejected positron, in turn,
when it has a minimum energy 1.022 MeV, annihilates with an electron to produce two 0.511 MeV gammas.
The cross-section for pair production for sufficiently high energy hνc137me c2 Z 21=3 can be written in the asymptotic
form
28 2
σP 5 Z 2 re2 ln183 Z 21=3 2
9 27
The total cross-section (barns) for gamma interaction with matter is thus
σγt 5 σPE 1 NσC 1 σP
With nuclei, the photons can be energetic enough to be absorbed and eject protons, ðγ; pÞ reactions, or neutrons in
what are called the photoneutron ðγ; nÞ reactions. Gammas can also elastically or inelastically scatter for neutrons.
However neutron interactions will be discussed in a separate chapter. An excellent reference for photon cross-sections
is the photon cross-sections, Attenuation Coefficients, and Energy Absorption Coefficients from 10 keV to 100 GeV
(Hubbell, 1969).
24 Nuclear Engineering

With the cross-section defined above, the liner attenuation coefficient with units of cm21 together with the mass
attenuation coefficient μ~ with units of cm2 =g can be defined as
μ
μ 5 Nσγt ;~μ 5 ρ

The intensity of photons can then be found from the source intensity Io across a shield of thickness x as
IðxÞ 5 Io e2μx
The above expression would hold under ideal conditions, that is, a point anisotropic source of intensity Io incident
on a thin shield. For a point isotropic source, the intensity would be given as
e2μR
I ð RÞ 5 I o
4πR2
to include both geometrical and material attenuation. For a broad beam or thick shield, the effects of scattering of radia-
tion inside a shield is better represented by including a “buildup” factor B, a dimensionless number greater than one
representing the effect of secondary radiation over that of primary radiation in which the gamma radiation passes
straight through without appreciable scattering. The buildup factors for exposure rate X_ in place of intensity I above are
given in the Radiological Health Handbook (US Dept of Health Education and Welfare, 1970). For a point isotropic
source in water, the buildup factors, tabulated for values in the range E 5 0:255 2 10:0MeV show that at 0.255 MeV, B
increases from 3.09 for μx 5 1, to 982 for μx 5 20, thus the effect becomes more significant as μx increases. For fixed
thickness, μx 5 1 with increase in energy from 0.255 to 10 MeV, B decreases from 3.09 to 1.33, while at μx 5 10, for
the same energy increase, B decreases from 982 to 5.98. With the same trends, the maximum buildup factors are 141
for aluminum, and 55.6 for iron. These trends however diverge for tin as for μx 5 20, B increases from 18.8 at 1 to 33.4
at 10 MeV. Similar differences are reported for tungsten and lead, while for uranium the maximum buildup factor is
28.5 for 10 MeV and μx 5 20.
For a mixture or compound
X
~μ 5 wi ðμ=ρÞi
i

The coefficients for some materials are given in Table 1.6.

For gamma interactions with materials, the thickness of a shield is expressed as a Half Value Layer (HVL) as one
which would reduce the intensity by a half of its original intensity, or a Tenth Value Layer (TVL) which would reduce
the intensity to one-tenth. Thus for a 1 MeV gamma, the HVL of lead is 0.76 cm while that of iron is 1.52 cm.

Exercise 6: Compare the severity of radiation from alpha particles, beta particles, X-rays, and gamma rays in the con-
text of shielding personnel in the vicinity of a space station with a nuclear-powered radioisotope generation system.

1.4 Sources and effects of radiation

We are all exposed to natural radiation originating from cosmic radiation and decay chains of which the three main
ones are from thorium-238 called the thorium series, uranium-238 called the uranium or radium series, and uranium-
235 called the actinium series. These produce radon, an inert noble gas, which among other isotopes such as potassium-
39 and potassium-40 goes into the lungs and causes further damage by alpha decay with a half-life of 3.8 days.

TABLE 1.6 Mass attenuation coefficients ~μ for 1 MeV photons.

Material Air Water Concrete Iron Lead Glass, leada Polyethylene

~μðcm =gÞ
2 0.06358 0.07072 0.06495 0.05995 0.07102 0.06914 0.07262
a
Glass, lead has ,Z/A . 5 0.42101,I 5 526.4 eV, ρ 5 6.220 g/cm3, composition (weight fraction) oxygen: 0.156453, silicon: 0.080866, titanium: 0.008092,
arsenic: 0.002651, lead: 0.751938.
Source: From https://physics.nist.gov/PhysRefData/XrayMassCoef/ElemTab/z82.html.
 The atom and nuclear radiation Chapter | 1 25

In addition to α, β and γ radiations described above, neutron radiation is important in all fission processes in the
production of energy as well as in nuclear systems used for various applications in industry. Neutrons can be produced
from fission in nuclear reactors, spontaneous fission sources, accelerator-based neutron generators and from other neu-
tron producing reactions such as ðα; nÞ and ðγ; nÞ reactions.
A neutron producing nuclear fission reaction of uranium-235 is
01n 1 92235U-54140Xe 1 3894Sr 1 201n 1 200MeV
which is the basis for thermal energy in nuclear reactors.
Portable neutron sources include americium-beryllium (Am-Be) and californium Cf252 with properties listed in
Table 1.7 and neutron spectra shown in Fig. 1.17.
Californium-252 has a shorter half-life than Am-Be but is a more intense source; a small 10 μg pellet of Cf-252
emits 2.31 107 n/s with an average energy less than that from Am-Be. Both are intense sources with diverse and wide-
ranging applications. pffiffiffiffiffiffi
The energy spectrum of Cf-252 shown in Fig. 1.20 is of the form f ðEÞ 5 Ce2E=a sinh bE (a 5 1.025, b 5 2.926)
with a probable energy of 0.7 MeV and average energy 2.1 MeV, while Am-Be produces relatively more high-energy
neutrons.
The Am-Be source is an example of the production of neutrons from an alpha emitter (Am) in the presence of beryl-
lium from an ðα; nÞ reaction such as

6 1 n0 1 5:6MeV
B12 1
He42 1 Be94 -C13 -
C612 1 n10 1 γ4:4MeV 1 1:2MeV
The most intense source of neutrons are fission reactors, with a fast neutron (B2 MeV) yield B1012 n/MW though
small accelerator-based neutron generators, such as the van de Graff and Cockroft-Walton generators were used in the
early days with accelerated charged particles striking low-Z targets. Accelerator-based sources were developed mainly

TABLE 1.7 Am-Be and Cf-252 neutron sources.

252
Source Am-Be Cf
Type α-emitter Spontaneous
Half-life (years) 432.2 2.645
Specific activity (Ci/g)  532
Neutron yield (n s21 Ci21) 22.4 3 106 4.4 3 109
Average energy (MeV) B 4.2 MeV B2
Maximum energy (MeV) B 11 MeV B 10

FIGURE 1.20 Neutron energy spectra of Am-Be and Cf252sources.

26 Nuclear Engineering

for materials research (Granada, Santisteban, Dawidowski, & Mayer, 2012; Verbeke, Leung, & Vujic, 2000) and are
now being improved for specialized uses such as cancer therapy (Dymova et al., 2020; Letourneau et al., 2017;
Marchix et al., 2018). The latest technology for small intense portable neutron generators is based on the D-D and D-T
fusion reactions

1.4.1 Radiation dose

The first units used in radiation were the Becquerel (Bq 5 1 disintegration per second) and the curie (3.7 3 1010 disin-
tegrations per second). The amount of any radioisotope that emits 1 Ci depends on its half-life t1=2 from which its decay
constant λ is found; the number of atoms required to produce 1 Ci are then
3:7 3 1010
NCi 5
λ
from which the mass of that radioisotope required for an activity of 1 Ci can be found
NCi
MCi 5
NAv
The natural background radiation source from cosmic and decay chains is about 2 mrem/day, or about 0.730 rems
(7.30 mSv) in one year (Cacuci, 2010). As mentioned in the previous section, the natural background radiation exists
mainly due to decay chains from thorium and uranium; thus the human body contains radioactivity from isotopes such
as radon and potassium and has an activity of about 8000 Bq (a few micro Curies). For a neutron source, the half-life of
Cf-252, given in Table 1.7, is 2.645 years so that the amount of Cf-252 producing 1 Ci would be
3:7 3 1010 At12 3:7 3 1010 3 2:645 3 365 3 24 3 3600 3 252:08163
MCi 5 5 5 1:9mg
0:693 NAv 0:693 3 6:022 3 1023
Thus 1.9 milligram of Cf -252 gives 1 Ci which is a specific activity of B526 Ci/g.
The neutron dose from Cf252 is 2.22.3 3 103 rem m2/g/h (2223 Sv m2/g/h) so that a 25 μg source emitting 5.75
3 107 n/s would give a neutron dose at 1 m in air of 5557.5 mrem/h or 2.66 3 10215 Sv/s per source neutron.
The dose from 1 μg of 252Cf at 1 m in air is 0.0221 mSv/h (2.21 mrem/h) from fast neutrons and 0.0019 mSv/h from
gamma rays. Thus it is necessary to shield the neutrons as well as any secondary radiation produced from it; only then
it can be safely used.
Neutron dose 0.590.73 μSv/h at 1 m/GBq 2223 Sv m2/g/h
Gamma dose 0.68 μSv/h at 1 m/GBq 1.6 Sv m2/g/h

The gamma dose rate is quoted at R 5 100 cm in air, of 1.42368E-20 (0.1346) Sv/source. Assuming the source
strength of 1 g Cf252 to be 2.4 3 1012 n/s the gamma dose is 0.123 mSv m2/g/h from secondary photons produced by
neutron interactions. With a simulation in the photon-only mode, assuming an average gamma energy of 0.8 MeV, the
gamma dose is 3.40568 3 10217 (0.0001) Sv/source gamma which is 1.6249 Sv m2/g/h.
This has been estimated using an average neutron emission of 3.768 neutrons/fission and 2.314 3 1012 n/g.s i.e.
0.6141 3 1012 fissions/g.s and a gamma emission yield of 8.30 6 0.08/fission with an average energy of 0.80
6 0.01 MeV, Thus the gamma emission is estimated as 5.09703 3 1012 gamma/g.s 5 1.8349 3 1016 gamma/g.
h 5 0.6249 Sv/g.h.

Exercise 7: What would be the effect of radon gas Rn-222 on the lungs when it undergoes alpha decay?

The units used for quantifying the effects of nuclear radiation are the absorbed dose (D), the equivalent dose (H)
and the effective dose (E) defined as follows:

1.4.2 Absorbed dose

The SI unit of absorbed radiation dose is a Gray (Gy) (Commission, 2007; IAEA, 2014; Radiation Effects and Sources,
2016) defined as an absorbed energy dose of 1 J/kg. The absorbed dose (D) is the specific energy absorbed by a mate-
rial. Traditionally, it has been expressed as a rad which is defined as 100 erg/g. Thus
 The atom and nuclear radiation Chapter | 1 27

1 Gy 5 1 J=kg 5 107 erg=103 g 5 100 rad:

Historically, the unit of roentgen (R) has been used to represent the exposure (E) of radiation that results in the gen-
eration of one electrostatic unit of charge in 1 cm3 of air at STP. It is thus better to use the roentgen for electromagnetic
radiation.
Thus
1elementary chargeðcharge on an electronÞ 5 4:8 3 10210 esu
1
1esu 5 elementary charges
4:8 3 10210
One pair requires 32.5 eV
32:5
1esu 5 eV 5 6:76 3 1010 eV
4:8 3 10210
and 1 eV 5 1:6 3 10219 J, therefore
32:5
1esu 5 eV 5 1:0816 3 1028 J
4:8 3 10210
Thus 1:0816 3 1028 J is deposited in 1 cm3 dry air at STP and with air density ρ 5 0:00129 g/cm3 this is 0:0084J=kg
or 84erg=g, since 1 J 5 107 ergs.
Thus
esu J ergs C
1R 5 1 5 0:0084 5 84 5 0:258
cm3 kg g g
1 R 5 2.58 3 1024 C/kg. The exposure of 1 C/kg is equivalent to 33.8 J/kg, or 33.8 Gy.

Example 3: If dry air, with an ionization energy (I) of 32.5 ev/ion pair (B34 J/C), is subjected to an exposure of 1 R,
the absorbed dose is D 5 E I 5 2.58 3 1024 3 32.5 5 8.4 3 1023 J/kg 5 8.4 mGy.

From the roentgen, the CGS units of rem (roentgen equivalent man) is defined as a measure of the stochastic effect
of low level ionizing radiation on the body (1 roentgen 5 0.96 rem).

1.4.3 Equivalent dose

Since we would like to quantify the effect of radiation, two more factors must be considered viz (1) the type of radia-
tion, and (2) the response of a tissue of a living organism to that radiation. For these effects, the quantities defined are
the equivalent dose (H) and the effective dose (E).
Since the absorbed dose represents the amount of energy absorbed, another unit is required to represent the “effect”
of the dose. The SI unit of the “effect” called the “equivalent dose” is a Sievert (1 Sv 5 100 rems) defined as the
absorbed dose in grays multiplied by a “quality factor” Q which is a measure of the effect of a radiation. For X-rays,
beta and gamma radiation, QB1 while for neutrons an average value of QB10 can be used.
A chest X-ray which gives a dose of 0.1 mSv (0.01 rem 5 10 mrem) which is the equivalent of about one week of
natural background radiation. A procedure such as a CT scan of the abdomen and pelvis is about 20 mSv and a PET/CT
scan is B 25 mSv which are very high doses equivalent to over 200 chest X-rays.
The equivalent dose depends on the rate of energy deposition of a radiation as it collides with the host material in
its interactions. For a tissue T subjected to a radiation R with absorbed dose DT;R , the equivalent dose to that tissue ðHT Þ
is obtained by the cumulative effect of all the radiations on the tissue:
X
HT 5 WR DT;R;
R

where WR is the radiation weighting factor, formerly called the quality factor Q, resulting from the linear energy trans-
fer of a particular radiation. For X-rays, γ-rays, and generally for β-rays, WR B1, while for α-particles, heavy recoil
nuclei and neutrons, WR varies between 1 and 20 depending on the energy of the radiation.
28 Nuclear Engineering

Example 4: Two radiations viz an X-ray, and energetic neutrons, are incident on tissues of lung and skin with the data
provided below. Estimate the equivalent dose (H).

Radiation WR Dlungs;R (rads) Hlungs (rems) Dskin;R (rads) Hskin (rems)

X-rays 1 1 1 2 2
Neutrons 10 0.2 2 0.2 2

From the above, the equivalent dose to the lungs is Hlungs 5 1 3 1 1 10 3 0:2 5 3 rems (0.03 Sv). Similarly for skin,
Hskin 5 1 3 2 1 10 3 0:2 5 4 rems (0.04 Sv).
From a medical procedure in a hospital, a single X-ray of the chest, abdomen and pelvis would give an equivalent
dose of 10, 60 and 70 mrem while a full-body CT scan would give 10002000 mrem ( . 12 years of background
radiation).

1.4.4 Effective dose

To represent the tissue-specific response, the stochastic health risk is quantified by a weighting factor WT representing
the relative radio-sensitivity of a tissue. The overall effective dose E is thus expressed as
X
E5 WT HT;
T

The weighting factors WT , (Lamarsh and Baratta, p. 510) for some organs are listed below (Table 1.8).
The units of effective dose are also rems (Sv in SI).
Thus the overall effective dose is
X
E5 WT HT 5 Wlungs  Hlungs 1 Wskin  Hskin 5 0:12ð3Þ 1 0:01ð4Þ 5 0:40rems:
T

_ H_ and E_ with units of rem/h, mrem/s, Gy/hr, Gy/s, mSv/s, etc.

All the above quantities can be expressed as rates D;

1.4.5 Radiation safety limits

Radiation safety limits are prescribed by the International Commission on Radiological Protection (ICRP) and the US
Nuclear Regulatory Commission (USNRC) for the general public as well as professionals working with or exposed to
any form of radiation. According to US, Article 20.1201 Occupational annual dose limits for adults, the total effective
dose equivalent limit is set at 5 rems (0.05 Sv). The USNRC Code for Federal Regulations 10 CFR Part 19 requires
that” all individuals who, in the course of their employment, are likely to receive a dose of more than 100 mrem in a
year, must receive adequate training to protect themselves against radiation.”
No serious radiation effects have been normally seen for doses less than about 5 rems (0.05 Sv). Beyond this limits,
changes in blood chemistry have been observed. At 70 rems, vomiting takes place followed by hair loss, diarrhea, and

TABLE 1.8 Weighting factor WT for organs.

Tissue ICRP 26a ICRP 60a

Gonads 0.20 
Breast 0.15 0.05
Red bone marrow 0.12 0.12
Lung 0.12 0.12
Thyroid 0.03 0.05
Bone surfaces 0.03 0.01
Skin  0.01
Other organs 0.06 0.05
a
ICRP publications 26, 60.
 The atom and nuclear radiation Chapter | 1 29

FIGURE 1.21 The radiation dose in air from 1 mCi of iodine-131 and
cesium-137.

hemorrhage at about 100 rems. Possible death within two months can occur at 400 rems (4 Sv). Persons who receive doses
exceeding 5000 rems (50 Sv) die “within few hours of exposure” probably by the failure of the central nervous system.
As Fig. 1.21 shows, the dose from 1 mCi of iodine-131 and cesium-137 one meter away in air is 2.2 R/h and 3.4 R/
h respectively. Assuming a quality factor of 1, this still gives high dose rates which can be reduced to one-tenth of their
values with thin lead shields (12 cm thick).
With the above estimates, we can now get an idea of the magnitudes of radiation at Fukushima where the dose rates
were as high as 1.2 rems/h which is at the site boundary; compare this with the maximum (occupational) permissible
dose rate of 5 rems/y (0.05 Sv/y); it comes to the permissible dose of over 2000 years! Even thirty-seven miles (60 km)
away from the site, the dose rate was measured to be 0.8 mrem/h; much higher than the maximum permissible.

1.4.6 Radiation detection

Radiation detectors, both passive and active, are used extensively for routine monitoring of personnel in radiation envir-
onments in nuclear reactors and medical facilities. They are small passive dosimeters, lightweight and wearable on the
chest such as the common thermoluminescent dosimeter (TLD), a detector capable of measuring doses from 0.01 mGy
to 10 Gy, used for personal dose monitoring typically over a period of a few months. TLDs produce luminescence (visi-
ble light) by the thermal effect of ionizing radiation on a chemical compound such as calcium sulfate and lithium fluo-
ride. Other detectors are also used on the body for displaying radiation exposure or as a ring when radiation sources are
being handled. Several other uses include space monitoring to detect radiation, or further to identify radiation sources
from the measurements of the energy spectra. The basic purpose of these dosimeters is to detect and measure radiation
to ensure that, in accordance with regulations (10 CFR Part 20, and ICRP) whole body dose limits of 5 rems/y, or
50 mSv, as the total effective dose equivalent (TEDE) are not exceeded.
The modes of operation of detectors are based on the ionization caused by charged-particle alpha and beta radiation,
as well as by the photoelectric, Compton and pair production interactions of gamma rays. In its simplest form, a gas-
filled cylindrical ionization chamber with a central positively charged anode and a negatively charge wall cathode, both
with a potential, attract electrons and positive charges created by the energy deposited by the incident radiation. As the
potential is increased, the mobility of charges is also increased to give a flow of measurable current until the signal is
proportional to the intensity of the incident radiation. Proportionality is achieved when each electron, from the ion pair
created by the incident radiation, in turn produces its own secondary ionization and builds an avalanche with a high
multiplication. In this mode of operation, with an applied voltage typically of a few thousand volts, the ionization cham-
ber is called a proportional counter. At a higher voltage, the electric field surrounding the anode becomes weaker due
to a discharge caused by several avalanches spreading into the volume of the gas which reduces further secondary ioni-
zation. Proportional counters operate in a voltage range which inhibits this discharge. Beyond this voltage, a positive
space charge builds around the anode and the discharge terminates.
Proportional counters are better for alpha particle detection and measurements due to their short range, as described
in Section 1.3.1; they are also used for low-energy electrons produced by X-rays and for beta particles. In the Geiger-
Mueller mode, each pulse is counted and an audible signal indicates the presence of radiation; the GM tube is thus a
robust pulse counter.
Compact solid state detectors are used for measuring incident radiation the production of electron-hole pairs in a
semiconductor p-n junction due to the small band gap (in silicon 1.14 eV and in germanium 0.67 eV). These gaps, being
30 Nuclear Engineering

much less than the B32 eV energy required to create an ion pair in a gas detector, gives an edge to semiconductor
detectors which thus have better efficiency and resolution. Another very important application of germanium-based
semiconductor detectors is in nondestructive gamma ray spectroscopy which permits accurate measurement of excited
levels of nuclei (see Section 1.3.3). One important application of gamma ray spectroscopy is the determination of ura-
nium enrichment by measuring the 185.7 keV gamma rays emitted in the decay of uranium-235 to thorium-231.
The property of certain materials to “scintillate,” that is, to emit visible or ultraviolet light by energy deposition of
incident radiation is used for the detection of gamma rays. Though the resolution is lower than that for solid state detec-
tors, the long range of gamma rays makes scintillation detectors attractive.
Some commonly used inorganic scintillators with high photoelectric cross-sections, and hence high efficiency, are
sodium iodide and cesium iodide activated with thallium NaI(TI), CsI(TI), and silver activated zinc-sulfide. Organic
materials, such as plastics, are also used; however, due to their low atomic number, they are better suited to the detec-
tion of radioisotopes with low-energy beta emissions (Section 1.4.2).
Cherenkov detectors are used for the detection of very high velocity (β . 1=nÞ subatomic particles by photomulti-
plier tubes with light sensitive “photocathodes.” For electrons, this velocity is B 1 MeV which is easily reached by
electrons, while for protons and neutrons it is in the GeV range. The speed of light in a material of refractive index n
can be less than the speed of light in vacuum c since n 5 c=v; thus for vacuum, for example, n 5 1 and for water
n 5 1:33; thus light moves at B75% of the speed of light in vacuum; thus a 500 MeV muon (vB0:98c) would be travel-
ing faster than light and would produce a radiation such as the bluish white tinge of light observed by Cherenkov in the
water pool of a nuclear reactor.
For more detailed information on the energy spectrum of a radiation source, or of several sources, a multichannel
analyzers (MCA) is used to obtain “pulse height” tallies. The principle used in a MCA is that incident radiation which
produces charge Q would produced a voltage V 5 Q/C which could be “discriminated” by its’ electronics to permit a
measurement of the energy spectrum of the incident radiation. As will be illustrated in a simulation exercise
(Section 1.6), pulse height tallies are now available in simulation codes so that the energy spectrum can be identified
for various sources.
In nuclear forensics, and for security applications, activation techniques such as prompt neutron activation, are also
used for the detection and identification of nuclear material and explosives.
Neutrons are detected by the secondary ionization they produce during their interaction with matter, for example, in
their passage in air or a gas, they can create ionization which can be detected in the form of a voltage (Knoll & Kraner,
1981). In a commonly used gas detector containing boron trifluoride gas (BF3), the detector consists of a cylinder (of
aluminum, brass or copper) filled with BF3 gas at a pressure of 0.51 atmospheres.
The nuclear reaction with cross-section shown in Fig. 1.22 that takes place in the gas is
510B 1 01n-37Li 1 24He:
The products 37Li and α (24He) travel in opposite directions after the collision creating ion pairs in the BF3 gas.
When the lithium is left in the ground state (about 6% of the time) the particles have 2.792 MeV to create ion pairs. In
the other case where lithium is left in the excited state, the KE available is 2.310 MeV and hence a smaller resulting
signal. In the ground state of lithium, the energy carried by the α particle is 1.78 MeV while the remainder1.012 MeV
is carried by the lithium.
In naturally occurring boron, the B-10 percentage is 20% while B-11 is 80% and the boron ðn; αÞ cross-section is
attractive for the isotope B-10. Thus it is necessary to enrich the gas in B-10. Thus in practice the enrichments for B-10
is increased to 96%. In the detection system, the cylinder is the detector (cathode) while a single thin wire running
down the axis of the tube is the anode.

1.5 Atomic densities of elements and mixtures

Several elements and mixtures are used in standard nuclear engineering for various materials and processes used in the
front-end fuel cycle, in nuclear systems and in the water and reprocessing “back-end” fuel cycle. The first step in the
modeling process is thus preparing the atomic and molecular data. For criticality calculations, the atomic densities of
the following are calculated: uranium-235, uranium-238, plutonium, natural boron, water, boron carbide, uranium diox-
ide fuel, UO2F2 in solution with water and several other commonly used materials (Harmon et al., 1994). An excellent
sourcebook for modeling and simulation data is the Compendium of Material Composition Data for Radiation
Transport Modeling (McConn et al., 2011) which contains data for 372 materials from A-150 tissue equivalent plastic
to zirconium hydride including explosives and medical materials.
 The atom and nuclear radiation Chapter | 1 31

FIGURE 1.22 B(n,a) cross-section.

The atomic number density N is given by the expression N 5 ρNav =M, where ρ is the gram-density (g/cm3), Nav is
Avogadro’s Number (6.023 3 1023 atoms  gm-atom21 for an element, or molecules  g/mol for a molecule). This definition
of Avogadro’s number is crucial to the understanding of number density. As an example, consider the number density of water
molecules, of hydrogen atoms and oxygen atoms in such molecules. The number of water molecules can be found as
ρH2 O Nav 1 3 6:023 1023
NH2 O 5 5 5 0:3346 3 1023 molecules cm23
MH2 O 18
From the above, we can find the number of hydrogen and oxygen atoms: NH 5 2NH2 O , and NO 5 NH2 O with units of
atoms/cm3.

Example 5: Calculate the number density of an element given its density and molecular weight.

For a single element, we know that Avogadro’s number of atoms Nav , or one gram-atom, would weigh its atomic
weight A, so one gram would have Nav =A atoms, and for a density ρ g=cm3 , there would be
ρNav
N5 atomscm23 :
A
Calculate the number density of pure U238 with ρ 5 19:1 g=cm3 and A 5 238:0508g  ðg:atomÞ21
19:1X0:60231024 gcm23  atomsðg  atomÞ21
N5
238:0508 gðg  atomÞ21
32 Nuclear Engineering

which gives N 5 0:048331024 atomscm23 .

Example 6: Calculate the atomic fractions and atomic weight of an element given weight fractions of its constituent
elements.

Derivation from first principles: Consider two elements i and j, of atomic weight Ai g  ðg:atomÞ21 and Aj
g  ðg:atomÞ21 with weight fractions ε5wi =w; and12ε5wj =w, where w 5 wi 1 wj , respectively. Then, for Ni and Nj
atoms/cm3, since Avogadro’s number of any substance k weighs Ak grams, we can write
ρi Nav ρj Nav Ni ρ Aj w i Aj εAj
Ni 5 ; and Nj 5 ; so that 5 i 5 5
Ai Aj Nj ρ j Ai w j Ai ð1 2 εÞAi
and

Ai N i
ε5 :
Ai N i 1 Aj N j
The above can be readily used to express the atomic fractions αk of each element in terms of the “enrichment” ε.
Thus in this two-component mixture
Ni αi εAj
5 5
Nj αj ð1 2 εÞAi
from which

εAj
αi 5 ; withαj 5 1 2 αi :
ð1 2 εÞAi 1 εAj
From the above, we can write the mass of 1 g.atom, or its atomic weight A as
A 5 αi Ai 1 αj Aj
so that

1 ε 12ε
5 1 :
A Ai Aj
A relation between atomic fraction and weight fraction can be readily obtained as
αi A
5 :
wi Ai
Application: Consider U235 and U238 mixed with weight fractions ε 5 w5 5 0:03 and 1 2 ε 5 w8 5 0:97, respectively.
The average atomic weight of the mixture A and the atomic fractions α5 ; α8 are required to be determined.
The average atomic weight is found as

1 ε 12ε 0:03 0:97

5 1 5 1 ; A 5 237:9586;
A Ai Aj 235:04 238:05
and the atomic fractions are
α5 A 237:9586
5 5 ; α5 5 0:0304:
w5 A5 235:04
Similarly
α8 A 237:9586
5 5 ; α5 5 0:9696:
w8 A8 238:05
 The atom and nuclear radiation Chapter | 1 33

Example 7: Calculate the density of a mixture prepared from two elements of given densities and weight fractions.

Derivation from first principles: Consider a substance made by mixing two elements A and B, of density ρA g=cm3
and ρB g=cm3 with weight fractions wA ; andwB , respectively. The density of the mixture ρmix is found from the ‘first
principles’ approach as follows.
Consider a volume of 1 cm3, in which the volume fractions of elements A and B are VA and VB respectively. Then
VA 1 VB 5 1:
Thus for x grams of the mixture, which in this case is also ρmix , the individual amounts are ρA VA and ρB VB respec-
tively, and so
ρA VA 1 ρB VB 5 x:
Thus the volume fractions can be found, from which the density is
1 wA wB
5 1 :
ρmix ρA ρB
The above can be generalized, for a n 2 component mixture, to
1 X
n
wi
5 :
ρmix i51
ρi

Example 8: Calculate the number density of each element in a molecular substance of given density and weight
fraction.

Calculate the atomic densities of U235, U238, and O2 in U (4 wt. %) O2 fuel pellets. Assume that the density of UO2
is 10.9 g=cm3 .
Calculate the average atomic weight of U:
1 ε 12ε 0:04 0:96
5 1 5 1 ; A U 5 237:9281:
AU Ai Aj 235:04 238:05
The atomic fractions are found to be α5 5 0:0405; α8 5 0:9595.
The molecular weight of UO2 can now be determined, since one molecule of UO2 has one atom of U and one mole-
cule (2 atoms) of oxygen.
A UO2 5 A U 1 2A O 5 237:9281 1 2ð16Þ 5 269:9269:
Calculate the molecular density of UO2:

NUO2 5ρUO2 Nav =A UO2 5 0:02432 1024 molecules=cm3 :

The atomic density of U and O2 can now be determined as

NU 5 NUO2 5 0:02432 1024 atoms=cm3

and
NO 5 2NUO2 5 0:04864 1024 atoms=cm3 :
From the atomic fractions of U235 and U238, calculate the atomic densities of U235 and U238
The atomic fractions are α5 5 0:04049 and α8 5 0:95951, and from Eq.(-) the individual atomic densities of U235
and U238 are found to be

N5 5 0:00098 1024 atoms=cm3

34 Nuclear Engineering

and
N8 5 0:02334 1024 atoms=cm3 :
Preparing such “mixture” cross-sections is a crucial “preprocessing” exercise for retrieving elemental cross-section
data from a data library and multiplying by the constituent number densities.

1.6 Mathematical modeling and simulation

The interaction mechanisms of charged particles and photons discussed in the previous sections are the basis of mathe-
matical modeling and computer simulation using the relevant databases and computer programming. Their applications
are diverse such as in shielding, radiation therapy, plasma physics, and astrophysics. For mathematical modeling, the
most accurate representation of particle transport is by the Boltzmann Transport Equation (BTE) which describes the
transport of particles in phase space but suffers due to its complexity as an integro-differential equation making analyti-
cal solutions possible only for regular geometries restricted to simple scattering laws. In a deterministic approach, the
linear BTE has an asymptotic limit expressed as the Fokker-Planck equation (BT-FPE) (POMRANING, 1992) equation
with roots in quantum and statistical mechanics as an evolution of the probability distribution of a system with drift and
diffusion. In another formulation, the simplification of the “in-scattering” integration term resulted in a form of the
Fokker-Planck equation convenient for modeling electron, ion and photon transport.
The BTE is a central focus area, along with its stochastic counterpart: the Monte Carlo method which is an outcome
of the integral form of the BTE. These aspects form the core of this book in chapters developed to their development
and applications.
As will be highlighted, the integro-differential BTE was, in the pre-supercomputing era, the main model for neutron
transport in nuclear engineering, and thus a lot of effort has gone into analytical solutions for idealized models, and the
development of powerful theoretical and numerical models in 3D codes capable of solving realistic problems with dis-
cretization in a finite number of energy groups, angle discretization as in the discrete ordinates (SN) and spherical har-
monics (PN) methods, and space discretization with finite-element, finite volume, nodal and several advancements of
numerical methods.
In simulations, it is possible to model realistic configurations and use elaborate theoretical models with extensive
databases to obtain reliable answers. Real-world systems and engineering designs are not restricted to regular geometry
and hence the first step is to model the geometry. This is fortunately possible by methods, such as combinatorial geome-
try, which will also be described in detail. The second step is the use of elaborate physics models with the support of
voluminous data compiled from sophisticated models, empirical formulas and experiments. The power of simulation is
thus strong enough to undertake realistic design analysis in nuclear engineering, and it is precisely the strength of such
mathematical modeling enforced and advanced by simulation with modern high performance computing that is the
essence of education today.
For simulation, efficient numerical methods are continuously being developed for deterministic as well as stochastic
methods. These will be covered in detail in later chapters, especially the Monte Carlo method which gains its wide-
spread acceptance from powerful methods as well as from the features of high performance computing such as proces-
sing speed and vector and parallel processing.
Computer codes provided by contributing research and academic institutions are maintained at organizations includ-
ing the International Atomic Energy Agency (IAEA), the OECD Nuclear Energy Agency (NEA), and Radiation Safety
Information Computational Center (RSICC) at Oak Ridge National Laboratory. Extensive databanks are available from
several sources such as the National Institute for Science and Technology (NIST), the Korean Atomic Energy Research
Institute (KAERI), and the National Nuclear Data Center (Bielajew, 1990).
Of the several available particle transport codes, some are SRIM (Ziegler, Ziegler, & Biersack, 2010), electron
gamma shower EGS (Kawrakow & Rogers, 2003), the Monte Carlo particle simulation code GEANT (Research, 2020),
AlfaMC (Peralta & Louro, 2014), MCNPX (Waters et al., 2007), the electron and photon coupled Monte Carlo code
PENELOPE (Salvat et al., 2011), the multiparticle transport code FLUKA (Battistoni et al., 2016). For electron trans-
port ETRAN (Muraz et al., 2020; Seltzer, 1991) with the stopping power databases from ESTAR, ASTAR and PSTAR
for electrons, alpha particles and protons (Berger et al., 2005) are extensively used.
Both deterministic (based on the BTE) and stochastic (based on MC) methods are extensively used for charged-
particle simulations and have their advantages and disadvantages. Decisive factors are clearly, their ability to model
realistic design problems and the underlying physics, and to give reliable results with computational efficiency (Adams
& Martin, 1992; Patel, Warsa, & Prinja, 2020). In the former, this means acceleration schemes while in the latter it
 The atom and nuclear radiation Chapter | 1 35

means a careful selection of the simulation parameters which include several artifacts such as biasing, variance reduc-
tion, and importance sampling.

Exercise 8: What is the commonality between the Boltzmann Transport Equation, the Neutron Transport Equation, and
the Fokker-Planck equation?

1.6.1 Alpha particle transport simulation

The Monte Carlo simulation of alpha particles is based on simulating events in the transport of a particle from its
“birth” as a source particle to its “death” when it has lost all of its energy during interactions or has escaped from the
system. One birth-to-death sequence is called a random walk or a history. The word “random” is used because laws of
probability apply at each “event” in a history so the next one could consist of a different sequence. Many such histories
are followed to accumulate the tallies of interest. Each event in the random walk is modeled using the Bethe-Bloch
stopping power formula described in Section 1.3.1. The flowchart, in a very simplified form, is shown in Fig. 1.23.
Simulation parameters, such as the number of histories to be simulated, the number of materials to be used, and cut-off
parameters are specified. The tallies (e.g., energy deposited, transmitted energy, energy straggling, the CSDA range and
the projected range) are initialized. The energy loss and step size (% of energy allowed for each step) is specified so
that the step size is assigned.
A new particle with given parameters is started, and “located” in a volume and given a step length in its initial
direction. At the end of the step, its position and energy are updated and a check is carried out to determine if the new
position is in the same volume or whether it is a boundary crossing or system escape. The counters are updated
accordingly.
For charged-particle transport, the interaction physics is based on the coulomb electric force in collisions between
ions and electrons or ions and nuclei. One method used is based on the Continuous Slowing-Down Approximation
(CSDA) in which it is assumed that at every point in the trajectory, the energy loss by the alpha particles is gradual
modeled by the unrestricted (i.e. not cut-off energy to specify catastrophic events) stopping power.
For large energy losses in a target, the “thick target model” is used in which a Gaussian straggling model is used to
estimate the energy loss. Conversely, for intermediate and low energy losses the Vavilov and Landau distributions have
been used (Peralta & Louro, 2014; Ziegler, Ziegler, & Biersack, 2010). Stopping powers are used from various models
or from databases such as ASTAR (Berger et al., 2005) for both the nuclear and electron contributions.
Postcollision angles are determined from multiple scattering distributions. The change of angle in the scattering pro-
cess is by a series of many small angles to constitute “multiple scattering”; this is modeled by a Gaussian distribution.
In the case of large angle scattering, where the alpha particle would lose a large part of its energy the distribution of
angles would have a Rutherford scattering behavior with larger tails qffiffiffiffithan a Gaussian. In the AlphaMC code, the stan-
dard deviation of the deflection angles varies as θ0 5 13:6MeV pcβ Zα X0 so that the true path length is then t 5 s 1 4 s
s K 2
   2
Zα
13:6MeV pcβ
B θ0s ; this correction is important as alpha particle energy decreases and standard deviation θ0
2
where K 5 X0

increases as well as for high Zα .

The history is continued until the particle comes to a stop or escapes from the system. Many more such histories are
simulated to obtain reliable statistical results with standard deviation and variance.
As the alpha particles slow down, their stopping power increases due to the Bragg peak followed by a sudden end
of the trajectory.
The range of alpha particles is one of the several quantities estimated
Ð KE from such simulations. The range calculated
21
from CSDA is based on the evaluation of the integral RCSDA 5 0 α dEdE dx ; projected range is the mean value of the
maximum penetration depth computed from the distribution function of particles.

1.6.2 Interaction of electrons with matter

Following Section 1.3.2, the simulation of electron transport (Olbrant & Frank, 2010) is based on the interaction of beta
radiation incorporated through a realistic simulation of elastic and inelastic processes using forward-direction scattering
and transport coefficients extracted from the ICRU 77 database (Berger et al., 2007). This reduction introduces the stop-
ping dower into the BT-FPE. One application carried out with this formulation for light propagation in biological
36 Nuclear Engineering

FIGURE 1.23 Flowchart of alpha particle transport simulation.

tissues calculates reflectance and transmittance in the liver tissue (González-Rodrı́guez & Kim, 2008; Olbrant & Frank,
2010) and compares results with the transport equation discrete ordinates method taken as a benchmark. In another
application, the absorbed dose Dðr Þ, from 5- and 10-MeV electron beams in a semi-infinite water phantom, is calculated
via the stopping power and the angular flux from the transport equation. Good agreement is found with the standard
physics MC code GEANT and with PENELOPE. The compared results are for cancer therapy applications of an elec-
tron beam of 10-MeV on muscle (01.5 cm) followed by bone (1.53 cm) and then lungs (39 cm).
In the stochastic approach, the histories of a large number of electrons are followed and tallies gathered during the
simulation to obtain estimates based on means, variances and the laws of probability.
Electron transport is much more “complicated” than that for alpha particles due to the large number of collisions
dominated by the long-range coulomb force with a large number of continuous deflections. Due to the large number of
small-angle collisions, the simulation of electron transport can become computationally laborious if each interaction is
processed. Thus a “condensed history” approach is used in which a single event is taken to represent several, possibly
hundreds of thousands, small-angle collisions.
The classic paper on electron transport by Berger (1963), using Monte Carlo simulation written in FORTRAN and
implemented on an IBM704 computer, led to the development of the ETRAN code (Seltzer, 1991). In this paper,
Berger gives two classes of simulation strategies varying essentially in terms of their selection of slowing-down energy
and multiple scattering models. A good understanding of the simulation process, based on a random walk, can be
obtained from Berger’s “Class I” model based on the simplest approach as shown in Fig. 1.24. The process consists of
the following steps:
1. At the start of the simulation, the history counter is set to n 5 0 for a maximum number of histories N to be simu-
^ 0 ; t0 are given. Tallies and counters, such
lated, and the slab thickness, material and the source parameters r 0 ; E0 ; Ω
 The atom and nuclear radiation Chapter | 1 37

FIGURE 1.24 Flowchart of elec-

tron transport simulation.

as the number of steps in the random walk, absorptions, transmissions and reflections are initialized. Berger’s pro-
gram does not carry out variance reduction schemes which are an important feature of modern codes.
The energy loss index k 5 22m , for En11 5 kEn in a condensed step is specified for the number of steps in a logarith-
1
2.
mic spacing in which the electron loses half its energy; for example, m 5 4 would consist of four steps in which the
electron energy would be E0 ; E1 5 224 E0 ; E2 5 222 E0 ; E3 5 224 E0 and E4 5 221 E0 .
1 1 3

3. Angular deflections are computed considering kinematics with mean square deflections in multiple scattering calculated
from the Goudsmit-Saunderson model in conjunction with the Mott scattering cross-section; accounting for deflection of
electrons. The spatial displacement of the electron is computed by accounting for straggling through random variables.
4. The energy loss of the electron in the medium is computed once a collision has been processed.
5. Checks are made for boundary crossing to update counters and begin a new history until the maximum specified his-
tories are processed.
6. The tallies are estimated from arithmetic means, if variances are not calculated. Most codes now compute variances
and relative standard errors to give figures of merit to establish convergence.
The flowchart of a class II algorithm, used in EGS4, is different from the class I described above (Bielajew and
Rogers) as it samples the distance to an interaction rather than specifying the step size. It also calculates the energy loss
38 Nuclear Engineering

rather than sampling it from a distribution. Such differences have an effect on the results and thus simulation, especially
Monte Carlo simulation, is considered both an art and a science where parameters are selected based on some intuitive
sense of what is “more important” and otherwise. These aspects of simulation, such as importance sampling and vari-
ance reduction, will be covered later in this book. A similar selection of parameters is required in MCNP (Hughes,
1996) and its upgrade (Grady Hughes, 2014) which does not use restricted stopping powers and builds distribution
tables such as energy-dependent histograms for angular bins from Goudsmit-Saunderson for substep lengths and scatter-
ing angles, as well as for bremsstrahlung production probabilities. These parameters are selected in MCNP on the
PHYS:E card where default, or user-specified values are used for the simulation. The values include the upper limit for
electron energy, the choice of whether photons will produce electrons, whether electrons will produce photons, the use
of full bremsstrahlung or simple angular distribution approximations, the use of sampled or expected straggling, and the
number of electron-induced X-rays, the number of knock-on electrons, the number of photon-induced secondary elec-
trons and the option of producing bremsstrahlung at each substep.
The simulation used in MCNP for positrons is identical to that for electrons distinguishing the particles for purpose of tal-
lying only. For electron transport in aluminum, Berger sets m 5 16, that is, 16 steps for the electron to lose half its energy.
This covers the range of slowing-down from 2.0 to 0.03125 MeV in 97 steps as shown in Fig. 1.25 below. The step size is
ðk 2 1ÞEs
Δs 5  0 
Þ
dEðs
ds0


From Fig. 1.24 it is seen that that in the range 0.012 MeV the stopping power decreases; thus while slowing-
down, the energy Es and stopping power increases thus Δs decreases. In Berger’s simulation, the step size is
Δs 5 0.057476 g=cm2 at 2 MeV for the first energy interval in steps 116, while in the last interval (steps 8196),
Δs 5 0.000184 g=cm2 . In all intervals, the deflection angle increases from 11.4 degrees in the first interval to 16.5
degrees in the fourth interval and then decreases to 14.7 degrees.
For multiple scattering, Goudsmit and Saunderson (1940a, 1940b) derived the exact angular distribution for the scat-
tering angle as a Legendre series. Here

X
N ðs
1 0 0
AGS ðωÞsinω dω 5 l1 exp 2 Gl ðs Þds Pl ðcosωÞsinωdω
i50
2 0

where
ðπ
 
Gl ðsÞ 5 2πN σðθ; sÞ 1 2 Pl ðcosθ sinθdθ
0

The GS distribution peaks at small values of ωB103 2 253 as shown for a calculation for a pencil beam of 20 MeV
electrons passing through a 0.25 cm thick slab of water (Rogers & Bielajew, 1990).

FIGURE 1.25 Electron energy versus steps in simulation in aluminum

from 2.0 to 0.03125 MeV for m 5 16.
 The atom and nuclear radiation Chapter | 1 39

FIGURE 1.26 Continuous Slowing Down Approximation range in alu-

minum from ESTAR.

The problem considered by Berger was a one-dimensional slab with a given source in phase space Pðr 0 ; E0 ; Ω ^ 0 ; t0 Þ.
The simulation was carried out to estimate the transmission, reflection, and absorption of electrons in a slab of alumi-
num for which the range is shown in Fig. 1.26.
The effect of keeping the smallest step size for all energy values would thus be seen to be computationally
inefficient.
In other strategies for simulation, path length is modeled with log spacing so that energy loss is constant per step or
a mixed-log model depending on where the electron is (near a boundary or inside the medium), or fixed path length
making it shorter to keep angular deflections small.
Similarly, energy loss models can be continuous slowing-down or fluctuation models. Angular deflections are com-
puted considering kinematics with mean square deflections in multiple scattering as in GEANT4 for a detailed descrip-
tion of multiple scattering collisions calculated from single-scattering Rutherford scattering law, Gaussian distribution
or Molière distribution (considering the occasional large angle event which are not incorporated in the Gaussian model).
In GEANT4, the Lewis theory computes moments of the spatial distribution as well as described by Urban (2005,
2006) with results given for distributions of transmission, energy deposition and spectra of MeV electrons in aluminum
layers.
In a comparison between the Molière and Goudsmit-Saunderson model for the Rutherford and Mott models for elec-
trons and positrons, Berger has shown that the angular distribution for both electrons and positrons, slowing-down in
aluminum from 1 to 0.9576 MeV, has B40% each for small-angle scatterings (B41% for 015 degrees and B42% for
1530 degrees) and the rest as follows: B12%, 3%, B1%, B0.36%, 0.27%, 0.08% and 0.02% for each successive 15
degree interval to 180 degrees. Thus the chance for a “catastrophic” event in which a large deflection occurs with a
strong influence on the subsequent history has a probability of occurrence for a large number of collisions in a step.
Alternately, catastrophic collisions can be explicitly incorporated by specifying a cut-off energy loss beyond which
such an event occurs.
In more detailed models, such as in MCNP (Werner, 2017; X-5 Monte Carlo Team, 2008) simulations, production
of secondary electrons are also considered (electron-induced X-rays, electrons produced from electron-impact ionization
called “knock-on” electrons, and bremsstrahlung photons) particles shown in MCNP5 print Table 86 below.
In MCNP, the algorithms used are the Goudsmit-Saunderson theory for angular deflections, and the Landau theory
for modeling energy loss fluctuations. Straggling is accurately modeled when path length steps are less than 0.5 g/cm2
(McLellan, Med Phys, 1994).
There is considerable variation in commonly used codes. In the EGS4 code (Bielajew and Rogers) electron and posi-
tron models include standard models for low-energy Møller and Bhabha scattering as part of restricted (implying less
than a specified cut-off) collision stopping power, atomic excitation, soft bremsstrahlung and elastic particle multiple
scattering, as well as provisions for modeling “catastrophic” phenomena including large energy loss scatterings. The
parameters, such as cut-off thresholds, selected in a model have a bearing on the results. Clearly low thresholds will use
more computer time but would give better results which may not always be needed beyond some level of accuracy. In
energy deposition computations, both thresholds of secondary particle creation and bremsstrahlung production would
40 Nuclear Engineering

have an effect on the results and a user would need to model with carefully selected parameters considering whether
the nature of the problem corresponds to a thin or thick target.
Comparisons have been made between MCNP, GEANT, and PENELOPE (Archambault & Mainegra-Hing, 2015) to
highlight the effect of algorithms in the simulations. For beam energies of 0.5, 1.0, and 5.0 MeV on a water-filled
sphere, the estimated energy depositions showed good agreement for the single-scattering calculations. However, differ-
ences of up to 5% have been found between EGSnrc and PENELOPE for some problems such as the “in air” case for
decreasing radius at 0.5 MeV.

1.6.3 Interaction of gamma radiation with matter

In photon transport simulation shown in Fig. 1.27, a Monte Carlo simulation would begin, as for alpha and electron
transport, with a description of the geometry, materials, and photon source. Then the simulation parameters such as
number of photon histories (random walks) to simulate, “weight cutoffs,” the interaction physics treatment.
In a code like MCNP, the user has the option to use “simple photon transport,” where, for example, for high-energy
photons, coherent (Thomson) scattering and fluorescent photons production is ignored, or the “detailed physics trans-
port” in which coherent scattering and fluorescent photons are transported and form factors FðZ; E; θÞ are used to
account for binding effects. The Thomson ‘unpolarized’ differential cross-section (Section 1.3.3) is then written as
dσðZ; E; θÞ 1  
jT 5 re2 2 2 sin2 θ F ðZ; E; θÞ
dΩ 2
For which data is available for several material. The simple model reduces simulation effort and makes assumptions
that the photoelectric effect is an absorption process and its simulation continues or stops according to statistical criteria.
For Compton scattering, with appropriate probability, the collision is simulated and the photon transport continues while
the recoil energy is deposited at the collision site and can be used to generate a recoil electron for further transport.

FIGURE 1.27 Flowchart of photon transport simulation.

 The atom and nuclear radiation Chapter | 1 41

A typical photonuclear data library would contain the following interaction data for photons.
The simulation can be specified to run single mode or mixed-mode (photon-electron). Electron interaction data for
Z 5 1 to 94 contain stopping power parameters, bremsstrahlung data as well as data for K-edge energies, Auger
electrons.

Exercise 9: In the transport simulations for alpha particles, beta particles, and photons, what are common features in
Monte Carlo simulations that may contribute to better results and what would be their effect on their computation time?

1.6.4 Radiation dose from Calfornium-252 gamma source in water

Many practical problems require mixed-mode simulations where, for example, a photon transport simulation would
include the transport of electrons produced which could subsequently result in the production of X-rays. In such mixed-
mode “P-E” simulation, the electrons created by photons are banked and stored for later transport. When the electron
mode is not initiated, then a thick target bremsstrahlung (TTB) model is used in which electrons are generated but
locally stopped. When electrons are not transported, then their energy is assumed to be locally deposited.
This mini-simulation is a step toward understanding the elaborate physics that is incorporated into simulation tools.
Consider a point source emitting S gammas per second incident from the left in a straight line (anisotropic) striking
a rectangular material labeled a detector “D” as shown in Fig. 1.28. We carry out a simulation to estimate how many
gammas reach across the transverse water layers.
The gamma source, Cf252 placed at the center, has an energy spectrum shown in Fig. 1.29. This simulation is carried
out to tally the currents and dose inside the transverse water shells of radii 0.2, 0.4, 0.6, 0.8, 1.0, 1.2, 1.4, 1.6, 1.8 and
2.0 cm and length 1 cm.
The MCNP program Book1_06 is attached in Annex 1 and the MATLAB program from which these figures were
produced, is listed in Annex 3.
This simulation uses the photo-atomic data for hydrogen and oxygen from the Evaluated Nuclear Data File ENDF/
B-VI and electron data from the file e103. Each region is assigned equal statistical importance for photons and elec-
trons. In the simulation, the maximum photon energy is 100 MeV.
A typical code output describing the libraries used would read: The mcplib04 library used is 1000.04p 1898 ENDF/
B-VI Release 8 with photo-atomic data for hydrogen (1-H mat 100) and oxygen 8000.04p 3272 ENDF/B-VI Release 8
photo-atomic data for 8-O mat 800 02/07/03 indicating the release version and the element ID.
For the electron step size, MCNP uses the value k 5 221=8 to compute steps on a grid from a high value, in this case
E 5 100 MeV down to a given low value, 1.079 keV, which at an average loss of 8.3% per step requires 133 steps. At
these steps, the collision and radiation stopping powers are determined from the formulas described in Section 1.3
(Bethe, 1930). These major steps are broken down into m substeps where the value m ranges from 2, for Z , 6 to 15 for
Z . 91. In this simulation, m 5 3. The range table gives stopping power (collision, radiation, total), the range from
CSDA, the value of “drange” which, divided by the density gives the step size in cm. Five lines of the 133-line table 85
of MCNP, steps 1, 2, 10 and 100 give the following values for the energy, stopping powers, CSDA range and “drange.”

FIGURE 1.28 Source and detector configuration.

42 Nuclear Engineering

FIGURE 1.29 Probability distribution function of cobalt-60 gamma

source.

TABLE 1.9 MCNP simulation steps for electron transport.

Step Energy (MeV) Stopping Power (MeV cm2/g) Range (MeV cm2/g) drange (MeV cm2/g)
133 1.079 3 1023 1.140 3 1022 2.888 3 1023 4.873 3 1026 7.647 3 1027
100 1.8826 3 1022 13.8 3.938 3 1023 7.70 3 1024 1.095 3 1024
9 50 2.140 1.145 19.83 1.286
2 91.7 2.195 2.220 30.74 1.765
1 100 2.203 2.438 32.58 1.833

Table 1.9 shows that the energy from step 1 to step 2, and all successive steps, decreases by 8.3% due to the selec-
tion of the value 21=8 . Similarly, half the energy is lost in eight collisions from step 1 to step 9. The next two columns
give the collision and radiative stopping powers, while the fourth column gives the CSDA range and the last column
gives the “drange” value giving a step size of 1.833 cm in the first step, which reduces to 7.647 3 1027 at the 133rd
step where the electron energy is 1.079 keV.
As shown in the flowchart for electron transport, the condensed random walk is sampled using interpolated data
from the grid. Each path length s (of length “drange”) in the first major step consists of s/m minor steps for which the
Goudsmit-Saunderson model is used for angular deflection. At each collision point, the probability of secondary particle
(fluorescent radiation or knock-on electrons) production is estimated. Bremsstrahlung photons are produced according
to a Poisson distribution with energy sampled from distribution tables. The electron energy is subsequently reduced to
account for this production. Similarly, the ionization from K shell impact and Auger electrons is also included in the
cross-section library.
The electron secondary production table is generated for the same energy grid as above, with values for bremsstrah-
lung, X-rays and knock-on electron cross-sections. For energy straggling, the term responsible for the difference of the
actual path from a straight path, sampling is done from Landau’s distribution function which has been further extended
by Blunck and Leisegang to estimate the variance of the straggling Gaussian (Hughes, 1996).
The simulation is started by sampling photons from the specified source. In this case, all source photons are gener-
ated at the origin with direction cosine (1,0,0); the energy is sampled from the PDF (Fig. 1.28).
For 1 million particles simulated, the photon summary shows that the average source energy is 0.97764 MeV, for
which the production is mainly from bremsstrahlung (722 tracks, energy 2.2502 1025 MeV) and from photon annihila-
tion; thus for every track started 1.0008 tracks are simulated. The loss terms are mainly from escape (986466, average
energy 0.96453 MeV), capture (14224, 2.6118 1024 MeV) and pair production (98, 4.0133 1024 MeV). On the average
there were only 67256 (B6.7%) photon collisions.
 The atom and nuclear radiation Chapter | 1 43

Similarly, the electron summary table shows creation of electrons from pair production (196, 3.0117 1024 MeV),
from Compton recoil (47805, 1.2507 1022 MeV), photoelectric effect (14224, 2.6118 1024 MeV) and from knock-on
events (623758, 2.4104 1023 MeV). Thus total production was 685983 electrons with an energy of 1.5480 1022 MeV.
The total electron energy would thus be obtained by multiplying by the source term; for example if there are 106
photons per second emitted from the source, the electron energy produced would be 1.5480 104 MeV. The electrons
were “lost” in the simulation by escape (4642, 4.8246 1023 MeV) and by energy cut-off (681341, 6.6631 1024 MeV).
The above summary tables should be carefully analyzed to understand the essential phenomena taking place in the
problem of interest.
From the overview of the accounting described in the summaries, the activity table gives estimates of the average
track, that is, movement of photons in each transverse water region. From this table, we learn that there were 9554 colli-
sions with hydrogen atoms and 52381 collisions with oxygen atoms in the innermost water region (cell 1). We also see
that in the transverse direction, the number of collisions reduces drastically from innermost regions in the transverse
direction due to the “current” or boundary conditions reducing from 3.15% (0.0057) to 0.63% (0.0126) with the stan-
dard relative error increasing about three times, as seen in Fig. 1.30.
In later chapters, we will learn how we can reduce the errors in estimates to get better more reliable results.
Similarly, the photon flux across the right surface of the inner source cylinder region is 7.95881 1022 photons/cm2
(0.0005) with a surface area 12.5664 cm2. The volume averaged photon flux in the transverse regions is shown in
Fig. 1.31 to decrease from 3.9512 (0.0001) photons/cm2-s to 5.992 1024 (0.0126) photons/cm2-s. In the detector region,
the flux is 3.0178 1021 (0.0002) photons/cm2-s.

FIGURE 1.30 Surface current of photons across transverse surfaces.

FIGURE 1.31 Intensity of photons across transverse surfaces.

44 Nuclear Engineering

FIGURE 1.32 Energy deposition in transverse regions.

TABLE 1.10 Radiation dose rate (rem/h) in water regions.

Data set Inner region Outer region

23
H*10 6.15572 10 (0.0008) 2.68817 1027 (0.0132)
ICRP-21 5.95151 1023 (0.0009) 2.21992 1027 (0.0134)
ANSI-1977 6.78671 1023 (0.0008) 3.31944 1027 (0.0131)

Exercise 10: How would you interpret the reduction in the photon surface current versus distance in the context of
interaction of gamma rays with matter (Section 1.3.1)?

Another important quantity is the photon energy deposition shown in Fig. 1.32. This follows the same patterns as in
the previous tallies, that is, a decrease from 7.62 1023 (0.0072) MeV to 1.02176 1025 (0.0926) photons/cm2-s.
The dose estimates using three different data sets (Monte Carlo Team, 2005), H*10, ICRP-21and ANSI-1977, give
the radiation in rem/h shown in Table 1.10 for the inner and outer regions. The reduction by four orders of magnitude
is an important estimate of the amount of attenuation of a californium source due to water.
This simulation took 0.64 min on an Intel(R) Core(TM) i72620M CPU@ 2.70 GHz 32-bit operating system.

Capabilities developed
A basic understanding of
1. an atom
2. stability
3. binding energy
4. radioactivity
5. radiation and interaction mechanisms of alpha and beta particles and gamma rays
6. radiation in nuclear systems
7. mixed-mode simulation
8. skills for computing atomic densities for commonly used materials

Nomenclature
English
c speed of light
fβBr fraction of incident beta energy converted by Bremsstrahlung into photon energy
 The atom and nuclear radiation Chapter | 1 45

h Planck’s constant
ℏ reduced Planck’s constant ℏ 5 h=2π
k conversion factor
l azimuthal quantum number
ml magnetic quantum number
ms spin quantum number
n principal quantum number
n number of shell
n neutron
p momentum
p proton
qe charge on an electron
rB Bohr radius
re electron radius
rn radius of the nth shell
t1=2 half-life
u atomic mass unit
v speed of electron
wi weight fraction of the ith component in a mixture

English capital
A relative atomic mass
B binding energy as a function of A and Z
C coulomb
D absorbed dose
E exposure
E equivalent dose
Ek kinetic energy of beta particle
En energy of the nth shell
Em electron mass energy equivalent (0.511 MeV)
F form factor in Compton scattering
H equivalent dose
I intensity
I mean ionization and excitation potential of absorber atoms (MeV)
I ðtÞ atoms of iodine-135 at time t
KEc kinetic energy (classical)
KEr kinetic energy (relativistic)
L angular momentum
N number of absorber atoms/cm3
Nav Avogadro’s number
NZ number of absorber electrons/cm3
Q quality factor
R range
R roentgen
S target width
Sv Sievert
T ðt Þ atoms of tellurium-135 at time t
V volume
X ðt Þ atoms of xenon-135 at time t
X0 material radiation length
Z atomic number of absorber
46 Nuclear Engineering

Greek lower
α alpha particle
αi atomic fraction of ith constituent in mixture
β beta particle
β βv=c speed of beta particle relative to speed of light c
β1 positron (positively charged beta particle)
γ gamma ray
γ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
1 2 ðv=cÞ
2

ε packing fraction
ε0 permittivity of free space
θ orthogonal angle
θ0 standard deviation of deflection in alpha transport Gaussian model
λ decay constant
λ wavelength
μ linear attenuation coefficient
μ linear attenuation coefficient
ν neutrino
ν antineutrino
ρ density
σ cross-section
φ0 Thomson cross-section
ϕ azimuthal angle

Greek capital
Ω solid angle
Ψ wave function

Abbreviations
A angstrom
Bq Becquerel
Ci curie
eV electron volt
fm Fermi
Gy gray
ICRP International Commission on Radiological Protection
MeV million electron volt
Rad radiation absorbed dose
Rem Röntgen equivalent man
TLD thermoluminescent dosimeter
USNRC US Nuclear Regulatory Commission

Problems
1. Given atomic fractions: U234 (0.0057%), U235(0.72%), and U238(99.27%), find the average atomic weight and the
corresponding weight percentages.
2. Given that Q 5 ½ðBEp1 1 BEp2 Þ 2 ðBEr1 1 BEr2 Þ, if Q . 0 and the reaction is exothermic, what is the implied by
the statement that “the BE of the products is more that the binding energies of the reactants”?
3. In the tellurium-135 decay chain (Section 1.2), with a half-life of tellurium of 11 s, λI 5 2:874 3 1025 =s,
λX 5 2:027 3 1025 =s, assume initial conditions and solve the rate equations for iodine-135 and xenon-135.
4. To propose a power source for the Mars rover, compare three alpha emitters, radium, polonium-210 (140 W/g,
gamma dose 0.012 Gy/h), and plutonium-238. From their specific heat, estimate the temperatures they will attain
for a capsule of half a gram.
5. Consider an alpha particle with charge ze with KE moving ‘head on’ toward a nucleus with charge Ze and assume
both to be point particles having an elastic collision. As the alpha particle approaches the heavier nucleus it comes
to a point beyond which it can not continue so it “stops and turns back,” that is, scattering angle 180 degrees, write
 The atom and nuclear radiation Chapter | 1 47

TABLE 1.P1 Data for Godiva, Jezebel and Jezebel 23.

Model Godiva Jezebel Jezebel23

Radius (cm) 8.7407 6.3849 5.9838
Density (g/cm3) 18.74 15.61 18.424
Composition (atoms/barn-cm) U235 4.4994e2 Pu239 3.7047e2 U233 4.6712e2
U238 2.4984e3 Pu240 1.7512e3 U234 5.9026e4
U234 4.9184e4 Pu241 1.1674e4 U238 2.8561e4
Ga69 8.26605e4 U235 1.4281e5
Ga71 5.48595e4
Mass (g) 52419.98 17019.77 16534.98

the conservation of energy for kinetic and potential and show that the closest distance between the two is
2
d 5 4πε2Ze
0 ðKEÞ
5 197:32Z
137KE fm where the KE is in MeV.
6. Estimate the wavelength of the bremsstrahlung photons from a beta particle emitted from P32 with a maximum
energy of 1.71 MeV if all its energy is lost in a single collision.
7. Give reasons to support the choice of a low-Z shield for beta particles even when the range is shorter than in a
high Z absorber.
8. Given that the mean energy required to create an ion pair in air is 33.7 eV, estimate the number of ion pairs created
by a beta particle of KE 2 MeV. For the mean ionization and excitation potential, use the approximate formulas
IB19:0eV for Z 5 1, IB11:2 1 11:7Z eV for 2 # Z # 13, and IB52:8 1 8:71Z eV for Z . 13,.
9. Calculate atomic densities for the following:
a. Natural uranium with ρ 5 19.1 g/cm3 and atomic fractions U238 0.992745 U235 0.007200.
b. Bare Pu239 metal delta phase 100% Pu239 with ρ 5 15.8 g/cm3.
c. Given the following data for the fast critical assemblies Godiva, Jezebel and Jezebel23 (Cullen et al., 2007)
determine the weight fractions of each of the materials listed.
10. Calculate atomic densities of the fuels
a. UO2 of density 10.5 g=cm3 with a U235 enrichment of 17%.
239
b. PuO2 of density 11.46 g=cm3 , with weight fractions of O16 8 0.118055 and Pu94 0.881945.
c. U-10 wt% Zr alloy powder of density 15.48 g=cm with uranium consisting of 17wt.% U235 and the rest U238
3

(Table 1.P1).
11. Calculate atomic densities of the reflectors
a. Beryllium metal density 1.85 g=cm3
b. Beryllium oxide with a density of 3.01 g=cm3 , and weight fractions 4Be9 0.360320, 8O16 0.639680
12. Find the atomic densities in boron carbide given its density ρ 5 2:52 g=cm3 (weight fraction B10 5 5 0.782610,
5 5 0.217390) Atomic Weight 5 55.24, Answer: N(B4C) 5 0.0277, N(Bnat) 5 0.1108, N(B5 Þ 5 0.02205, N
B11 10

5 Þ 5 0.08875, N(C) 5 0.0277

(B11
13. Find the atomic densities in a solution of UO2F2 with a uranium enrichment of 5%, density of U235 of 0.04 g/cc
and a given ratio of hydrogen to fissile atoms (H/X) of 500.U(4.89)O2F2 solution N5 5 1.0889 3 1024,
N8 5 2.0909 3 1023, NF 5 4.3996 3 1023, NH 5 5.7058 3 1022, No 5 3.2929 3 1022, Nt 5 9.6586 3 1022
(atomic densities are in units of 1024 atoms/cm3
14. Calculate the atomic densities in the following structural materials
a. stainless steel consisting of Fe with 18% chromium by weight, 8% nickel and 0.08% carbon, find the atomic
densities
b. aluminum (atomic weight 26.9815, density ρ 5 2:7 g=cm3 )

References
Adams, M., & Martin, W. (1992). Diffusion synthetic acceleration of discontinuous finite element transport iterations. Nuclear Science and
Engineering. Available from https://doi.org/10.13182/NSE92-A23930.
Anspaugh, L., Catlin, R., & Goldman, M. (1988). The global impact of the chernobyl reactor accident. Science (New York, N.Y.), 242(4885),
15131519. Available from https://doi.org/10.1126/science.3201240.
48 Nuclear Engineering

Archambault, J. P., & Mainegra-Hing, E. (2015). Comparison between EGSnrc, Geant4, MCNP5 and penelope for mono-energetic electron beams.
Physics in Medicine and Biology. Available from https://doi.org/10.1088/0031-9155/60/13/4951.
Battistoni, G., Bauer, J., Boehlen, T. T., Cerutti, F., Chin, M. P. W., Dos Santos Augusto, R., Ferrari, A., et al. (2016). The FLUKA Code: An accurate
simulation tool for particle therapy. Frontiers in Oncology. Available from https://doi.org/10.3389/fonc.2016.00116.
Bennet, G.L. (1989). A look at the Soviet space nuclear power program.” In IECEC-89. Crystal City, Washington D.C. ,https://fas.org/nuke/space/
sovspace.pdf..
Berger, M.J. (1963). Monte Carlo calculation of the penetration and diffusion of fast charged particles. Methods in Computational Physics.
Berger, M.J., J.S. Coursey, M.A. Zucker, & J. Chang. (2005). ESTAR, PSTAR, and ASTAR: Computer programs for calculating stopping-power and
range tables for electrons, protons, and helium ions (Version 1.2.3). National Institute of Standards and Technology.
Berger, M. J., Jablonski, A., Bronic, I. K., Powell, C. J., Salvat, F., & Sanche, L. (2007). Elastic scattering of electrons and positrons. Journal of the
International Commission on Radiation Units and Measurements, 7(1), https://doi.org/10.1093%2Fjicru_ndm001.
Bethe, H. A. (1930). Zur Theorie Des Durchgangs Schneller Korpuskularstrahlen Durch Materie. Annalen der Physik, 5, 325.
Bielajew, A.F. & D.W.O. Rogers. Lecture notes: Electron Monte Carlo simulation. Ottawa, Canada. ,http://rchttp://www.kek.jp/research/egs/docs/
pdf/nrc-pirs0394.pdf..
Blackett, P. M. S. (1924). The ejection of protons from nitrogen nuclei, photographed by the Wilson method. Proceedings of the Royal Society,
349360.
Bielajew & Rogers (1990). Brookhaven National Brookhaven National Laboratory. (accessed 10 Jan 2021 Elsevier accessed 8 January 2021 ,https://
www.nndc.bnl.gov/nudat2/reCenter.jsp?z 5 7&n 5 9..)
Cacuci, D.G. (2010). Handbook of nuclear engineering, Volume 2. Engineering.
Chadwick, J. (1932). The existence of a neutron. Proceedings of the royal society. ,https://doi.org/10.1098/rspa.1932.0112..
Commission, E. (2007). ICRP: ICRP Publication 103. ICRP, 2007. The 2007 recommendations of the international commission on radiological protec-
tion. ICRP Publication 103. Ann. ICRP 37 (24).
Demuth, S. F. (2003). SP100 space reactor design. Progress in Nuclear Energy. Available from https://doi.org/10.1016/S0149-1970(03)90003-5.
Determan, W.R., G.A. Johnson, A.J. Zillmer, T.E. Dix, C.Y. Lu, M. Tosca, & L.A. Trager. (2011). SNAP-10A Safety Test Program. In Nuclear and
emerging technologies for space, NETS-2011.
Dymova, M. A., Taskaev, S. Y., Richter, V. A., & Kuligina, E. V. (2020). Boron neutron capture therapy: Current status and future perspectives.
Cancer Communications. Available from https://doi.org/10.1002/cac2.12089.
El-Genk, M. S. (2008). Space nuclear reactor power system concepts with static and dynamic energy conversion. Energy Conversion and
Management. Available from https://doi.org/10.1016/j.enconman.2007.10.014.
Elsevier. Nuclear data sheets. ,https://www.sciencedirect.com/journal/nuclear-data-sheets..
Gibney, E. (2020). Arab world’s first Mars probe takes to the skies. Nature. Available from https://doi.org/10.1038/d41586-020-02170-2.
González-Rodrı́guez, P., & Kim, A. D. (2008). Light propagation in tissues with forward-peaked and large-angle scattering. Applied Optics. Available
from https://doi.org/10.1364/AO.47.002599.
Goudsmit, S., & Saunderson, J. L. (1940a). Multiple scattering of electrons. Physical Review. Available from https://doi.org/10.1103/PhysRev.57.24.
Goudsmit, S., & Saunderson, J. L. (1940b). Multiple scattering of electrons. II. Physical Review. Available from https://doi.org/10.1103/
PhysRev.58.36.
Grady Hughes, H. (2014). Enhanced electron-photon transport in MCNP6. In. doi:10.1051/snamc/201403105.
Granada, J. R., Santisteban, J. R., Dawidowski, J., & Mayer, R. E. (2012). Neutron transmission: A powerful technique for small accelerator-based
neutron sources, InPhysics Procedia. Available from https://doi.org/10.1016/j.phpro.2012.03.015.
Gray, S. (Ed.). Handbook of positron emission tomography illustrated edition. Hayle Medical.
Handbook of Practical X-Ray Fluorescence Analysis. (2006). Handbook of practical X-Ray fluorescence analysis. doi:10.1007/978-3-540-36722-2.
Harkness-Brennan, L. (2018). An Introduction to the physics of nuclear medicine. doi:10.1088/978-1-6432-7034-0.
Harmon, C.D., II, R.D. Busch, J.F. Briesmeister, & R.A. Forster. (1994). Criticality calculations with MCNP: A primer.
Hubbell, J.H. (1969). Photon cross sections, attenuation coefficients, and energy absorption coefficients from 10 KeV to 100 GeV. National Standard
Reference Data System.
Hughes, H.G. (1996). Treating electron transport in MCNP. In Proceedings of the training course on the use of MCNP in radiation protection and
dosimetry held in Bologna-Italy, May 1316, 1996. Los Alamos National Laboratory. ,https://mcnp.lanl.gov/pdf_files/la-ur-96-4583.pdf..
IAEA. (2014). Radiation protection and safety of radiation sources: International basic safety standards general safety. IAEA Safety Standards.
INPO. (2011). Special report on the nuclear accident at the Fukushima Daiichi Nuclear Power Station. INPO 11-005.
Kawrakow, I. & D.W.O. Rogers. (2003). The EGSnrc code system: Monte Carlo simulation of electron and photon transport. System.
Kibble, T. W. B. (2015). The standard model of particle physics. European Review. Available from https://doi.org/10.1017/S1062798714000520.
Knoll, G. F., & Kraner, H. W. (1981). Radiation detection and measurement. Proceedings of the IEEE. Available from https://doi.org/10.1109/
PROC.1981.12016.
Kumada, H. (2014). Neutron sources. Comprehensive biomedical physics. Available from https://doi.org/10.1016/B978-0-444-53632-7.00613-4.
Langlois, L. (2013). IAEA action plan on nuclear safety. Energy Strategy Reviews, 1(4), 302306. Available from https://doi.org/10.1016/j.esr.2012.11.008.
Letourneau, A., Marchix, A., Tran, N. H., Chauvin, N., Menelle, A., Ott, F., & Schwindling, J. (2017). Development of compact accelerator neutron
source. EPJ web of conferences. Available from https://doi.org/10.1051/epjconf/201714603018.
Loforte, A., Gliozzi, G., Suarez, S. M., & Pacini, D. (2020). Contributory role of positron emission tomography in a left ventricular assist device recip-
ient at the time of COVID-19 pandemic. ASAIO Journal. Available from https://doi.org/10.1097/MAT.0000000000001176.
 The atom and nuclear radiation Chapter | 1 49

Mallapaty, S. (2020). China’s successful launch of mars mission seals global era in deep-space exploration. Nature. Available from https://doi.org/
10.1038/d41586-020-02187-7.
Marchix, A., Letourneau, A., Tran, H. N., Chauvin, N., Menelle, A., Ott, F., & Schwindling, J. (2018). Saclay compact accelerator-driven neutron
sources (SCANS), InJournal of Physics: Conference Series. Available from https://doi.org/10.1088/1742-6596/1046/1/012009.
Matlack, G.M. & C.s.F. Metz. (1967). Radiation characteristics of Plutonium-238. New Mexico. ,https://www.osti.gov/servlets/purl/4495451..
McConn, R. J., Jr, Gesh, C. J., Pagh, R. T., Rucker, R. A., & Williams, R. G., III (2011). Compendium of material composition data for radiation
transport modeling. Pacific Northwest National Laboratory.
Miskolczy, G. & D.P. Lieb. (1990). Radioisotope thermionic converters for space applications. In Proceedings of the intersociety energy conversion
engineering conference. doi:10.1109/iecec.1990.716874.
Monte Carlo Team. (2005). MCNP - A general Monte Carlo N-particle transport code, Version 5. Los Alamos Nuclear Laboratory. doi:LA-UR-031987.
Muraz, J.-F., Santos, D., Ghetta, V., Giraud, J., Marpaud, J., Hervé, M., . . . Forlino, M. (2020). Development of a regenerated beryllium target and a thermal
test facility for compact accelerator-based neutron sources. EPJ web of conferences. Available from https://doi.org/10.1051/epjconf/202023103003.
Narita, T. & K. Kitao. (1992). List of strong gamma-rays emitted from radionuclides.
Nedunchezhian, K., Aswath, N., Thiruppathy, M., & Thirugnanamurthy, S. (2016). Boron neutron capture therapy - A literature review. Journal of
Clinical and Diagnostic Research. Available from https://doi.org/10.7860/JCDR/2016/19890.9024.
Olbrant, E., & Frank, M. (2010). Generalized Fokker-Planck Theory for electron and photon transport in biological tissues: Application to radiother-
apy. Computational and Mathematical Methods in Medicine. Available from https://doi.org/10.1080/1748670X.2010.491828.
Patel, J. K., Warsa, J. S., & Prinja, A. K. (2020). Accelerating the solution of the SN equations with highly anisotropic scattering using the Fokker-
Planck approximation. Annals of Nuclear Energy. Available from https://doi.org/10.1016/j.anucene.2020.107665.
Peralta, L., & Louro, A. (2014). AlfaMC: A fast alpha particle transport Monte Carlo code. Nuclear Instruments and Methods in Physics Research,
Section A: Accelerators, Spectrometers, Detectors and Associated Equipment. Available from https://doi.org/10.1016/j.nima.2013.11.026.
Peskin, M. E. (2019). Concepts of elementary particle physics. Concepts of Elementary Particle Physics. Available from https://doi.org/10.1093/oso/
9780198812180.001.0001.
Pomraning, G. C. (1992). The Fokker-Planck operator As an asymptotic limit. Mathematical Models and Methods in Applied Sciences. Available from
https://doi.org/10.1142/s021820259200003x.
Radiation Effects and Sources. (2016). Radiation effects and sources. doi:10.18356/b1749f17-en.
Research, CERN: European Council for nuclear. (2020). GEANT4 physics reference manual.
Rogers, D. W. O., & Bielajew, A. F. (1990). Monte Carlo techniques of electron and photon transport for radiation dosimetry. The Dosimetry of
Ionizing Radiation. Available from https://doi.org/10.1016/b978-0-12-400403-0.50009-9.
Rutherford, E. (1919). LIV. Collision of α particles with light atoms. IV. An anomalous effect in nitrogen. The London, Edinburgh, and Dublin
Philosophical Magazine and Journal of Science. Available from https://doi.org/10.1080/14786440608635919.
F. Salvat, J.M. Fernández-Varea, J. Sempau. (2011). PENELOPE-2011, a code system for Monte Carlo simulation of electron and photon transport. Simulation.
Seltzer, S. M. (1991). Electron-photon Monte Carlo calculations: The ETRAN code. International Journal of Radiation Applications and
Instrumentation. Part. Available from https://doi.org/10.1016/0883-2889(91)90050-B.
Suzuki, M. (2020). Boron Neutron Capture Therapy (BNCT): A unique role in radiotherapy with a view to entering the accelerator-based BNCT era.
International Journal of Clinical Oncology. Available from https://doi.org/10.1007/s10147-019-01480-4.
Taylor, L.S. et al. (1970). ICRU Report 16: Linear energy tranfer. Washington D.C.
Ulaner, G. A. (2018). Fundamentals of oncologic PET/CT (1st ed.). Elsevier.
Urban, L. (2005). A model for multiple scattering in Geant4. In Proceedings of the Monte Carlo topical meeting.
Urban, L. (2006). A multiple scattering model in Geant4. In Proceedings of the CERNOPEN-2006-077.
US Dept of Health Education and Welfare (1970). In John C Villforth and George R Shultz (Ed.). Radiological health handbook. Maryland.
Verbeke, J. M., Leung, K. N., & Vujic, J. (2000). Development of a sealed-accelerator-tube neutron generator. Applied Radiation and Isotopes.
Available from https://doi.org/10.1016/S0969-8043(00)00262-1.
Waters, L.S., G.W. McKinney, J.W. Durkee, M.L. Fensin, J.S. Hendricks, M.R. James, . . . & D.B. Pelowitz (2007). The MCNPX Monte Carlo radia-
tion transport code. In AIP conference proceedings. doi:10.1063/1.2720459.
Werner, C.J. (ed.). (2017). MCNP (R) user’s manual code Version 6.2.
Witze, A. (2020). NASA has launched the most ambitious Mars Rover ever built: Here’s what happens next. Nature. Available from https://doi.org/
10.1038/d41586-020-02257-w.
World Nuclear Association. (2012). Chernobyl accident 1986. World Nuclear Association 1986 (April 1986): 114. ,http://www.world-nuclear.org/
info/chernobyl/inf07.html..
X-5 Monte Carlo Team. (2008). MCNP - A general Monte Carlo N-particle transport code, Version 5 volume I: Overview and theory. Los Alamos
Nuclear Laboratory. doi:LA-UR-031987.
Ziegler, J. F., Ziegler, M. D., & Biersack, J. P. (2010). SRIM - The stopping and range of ions in matter. Nuclear Instruments and Methods in Physics
Research, Section B: Beam Interactions with Materials and Atoms. Available from https://doi.org/10.1016/j.nimb.2010.02.091.
Chapter 2

Interactions of neutrons with matter

The interaction of a neutron with matter is influenced by the presence of atoms it encounters during its transport; the
electrons it comes across have a relatively smaller mass and are unable to cause any significant deflection while interac-
tions with nuclei are orders of magnitude less due to the smaller chance of “coming in the way.” Similarly, the proba-
bility of neutron-neutron collisions is also small. This leaves neutron-nuclei encounters which form the basis of neutron
interactions in matter. The single quantity that determines the probability of interaction is a microscopic cross-section
which is like the circular cross-section of a sphere. The higher the cross-section, the higher will be the probability of
interaction and vice versa. Further, neutrons carry no electrical charge and do not therefore cause direct ionization.
Neutron detectors, discussed in Chapter 1, thus use indirect ionization in a gas to induce a measurable current. The neu-
tron, as a free neutron, is unstable and decays with the formation of a proton and a beta particle 10 n-11 H 1 02 1 e.
The basic concept of neutron interaction in matter is the matter-wave duality, in which the analysis considers a neu-
tron as a wave function. This analysis will be briefly reviewed in this chapter. It is based on quantum physics, and
draws several concepts from optics, and one model is therefore called the optical model of neutron interaction.

2.1 Kinetic theory

We start our description by considering the kinetic theory of gases, for which Boltzmann formulated the Boltzmann
transport equation (BTE) described in Chapter 1 in the context of the Fokker-Planck equation for charged particles and
the photon transport equation for photons (gamma rays and X-rays). The BTE is applicable to neutron transport as well
and provides a ‘complete’ model of neutron transport in terms of a statistical distribution function and uses the micro-
scopic cross sections from their underlying quantum descriptions. The works of Boltzmann and Max Planck are both
early 20th century at which time kinetic theory of gases was refined as a basis for modeling transport phenomena and
quantum physics was born. There was no knowledge then of elementary particle physics or that a neutron was com-
posed of quarks. Thus present-day knowledge on neutron interaction with matter rests on quantum physics, which is the
basis of all data used in later chapters of this book for carrying out neutron and photon transport in nuclear systems.
This data is compiled in elaborate form in large ‘libraries’ that is used for simulating phenomena and obtaining engi-
neering design information.
For the distribution function, a kinetic theory description of a gas, containing n0 molecules, says that all molecules
do not have the same energy but a distribution nðEÞ, modeled as a Maxwellian distribution
2πn0 E1=2
nðEÞ 5 e2E=kT (2.1)
ðπkT Þ3=2
with units of molecules per unit energy.
In “velocity space,” the distribution function nðvÞ, with units for the speed of molecules v is obtained from the above
as E 5 1=2mv2 , so that nðvÞdv 5 nðEÞdE gives
4πn0 v2 2mv2 =kT
dnðvÞ 5
3=2 e dv (2.2)
2πkT=m
Thus what we understand as macroscopic quantities, such as temperature and pressure, represent the average of a
statistically large sample of gas molecules each moving about randomly with their own temperature.

Nuclear Engineering. DOI: https://doi.org/10.1016/B978-0-323-90618-0.00002-8

© 2022 Elsevier Inc. All rights reserved. 51
52 Nuclear Engineering

The most probable energy Ep and most probable speed vp are obtained by setting the first derivative for nðEÞ in
Eq. (2.1) and for nðvÞ in Eq. (2.2) to zero
dnðvÞ
j 50
dv vp
is
rffiffiffiffiffiffiffiffi
2kT
vp 5
m
Thus the most probable energy is Ep 5 12 kT. The average speed v and average energy E can be found as
ÐN rffiffiffiffiffiffiffiffi
vnðvÞdv 8kT 2
v 5 ÐN
0
5 5 pffiffiffi vp
0 nðvÞdv
πm π

ÐN
EnðEÞdE 3
E 5 Ð0 N 5 kT
0 nðE ÞdE
2

The distributions nðvÞ and nðEÞ are plotted in Fig. 2.1 with the transformations

v 4n0
qffiffiffiffiffiffi x2 e2x
2
x 5 qffiffiffiffiffiffi ; nðvÞ 5
2kT
m π1=2 2kT
m

and for nðEÞ

E 2n0
x5 ; nðEÞ 5 1=2 x1=2 e2x
kT π kT
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
so that the vertical axis is in units of n0 =kT for nðEÞ, and in units of n0 m=ð2kTÞ for nðvÞ so that vp 5 1 in Fig. 2.1. At
room temperature, T 5 300 K, the most probable speed of a neutron is thus
rffiffiffiffiffiffiffiffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2kT 2 3 1:38066 3 10223 3 300
vp 5 5 5 2239 m=s
m 1:67493 3 10227

FIGURE 2.1 Maxwellian distribution.

 Interactions of neutrons with matter Chapter | 2 53

for which E0 5 0.025 eV. Neutrons in thermal equilibrium with their surroundings at room temperature with a speed of
B2200 m/s and energy 0.025 eV are referred to as thermal neutrons. Neutrons with higher speeds, with energies on the
order of several MeV, are referred to as fast neutrons.

Exercise 2.1: Thermal flux show that the function φðEÞ 5 nv can be written as φðEÞ 5 φth MðE; TÞ where
E 2E=ðkTÞ
M ðE; T Þ 5 e
ðkT Þ2
qffiffiffiffiffiffi ÐN
where φth 5 8kT
πm n0  v M n0 and 0 M ðE; T Þ 5 1. Compare v M with vp found above.

Matlab Example 2.1: Distribution function plot the distribution functions nðvÞ and nðEÞ

k=1.38066e-23;% J/K
i=0;
for x=0:0.1:4
% for E, x=E/kT for v, x=v/sqrt(2kT/m)
i=i+1;
nE(i)=(2/pi^0.5)*sqrt(x)*exp(-x);
nv(i)=(4/pi^0.5)*x^2*exp(-x^2);
xx(i)=x;
end

It will be seen in subsequent chapters that the distribution function requires knowledge of the phase space compris-
ing seven variables (three variable of space, two of angle, one of energy and one of time) for a complete description of
transport. Clearly this is beyond the scope of deterministic treatment and thus moments of the distribution function are
often used. This is left as an exercise for the reader to appreciate the physical significance of moments which, in princi-
ple, can be used to reconstruct a distribution function.

Exercise 2.2: Moments of distribution function find expressions for the zeroth, first, and second moments of the distri-
bution function nðEÞ and describe their physical significance.
   
In nuclear engineering, the fundamental quantity of interest is the neutron flux φ r; E; Ω;^ t 5 n r; E; Ω;
^ t v with
units of neutrons per cm2 per second per unit energy interval per steradian. With some simplifications, the integrated
or “averaged” flux can also be defined as φ 5 nv where n is the number of neutrons per unit volume, in a domain, and
v is the neutron speed. Neutron flux has units of neutrons (cm/s) cm3 and is thus a ‘distance traveled’ in a volume, or a
track length (neutrons  cm2/s). The neutron current is the vector form of the flux, given by J 5 N^ ∙ vn where N^ is a unit
normal vector on area dA so that the net number of neutrons leaving dA is JdA neutrons per second. Note that v 5 Ωv ^
and N^ ∙ Ω^ . 0 implies that neutrons are leaving the surface and conversely N^ ∙ Ω ^ , 0 implies neutrons entering a
surface.

2.2 Types of neutron interactions

We now return to the particle-wave duality by which neutron interactions with matter are described. A neutron has a
de Broglie wavelength (Chapter 1) which varies from B 10214 m for a fast neutron to 10210 m for a thermal neutron
causing thermal neutrons to have a larger cross-section. At the quantum level, the discrete angular momentum L 5 lh̄
can have values for waves s; p; d; . . . for which l 5 0; 1; 2; . . . while at the classical level, the angular momentum of
the incident neutron is the product of its linear momentum and its impact parameter. Thus if an elastic collision is to
take place, it can happen at the lowest level of angular momentum for which the impact parameter must be of the
order of a fermion (nuclear size) which corresponds to the lowest wave scattering, that is, s-wave scattering ðkR{1Þ.
As shown in Fig. 2.2, for a neutron scattering off uranium 238, the angular momentum quantum number varies as
lB0:0019E1=2 (Ein eVÞ so that l 5 1 at EB277keV. Thus s-wave scattering in uranium 238 can be assumed to hold
for E&277keV, while p-wave scattering will become significant at E . 277keV. For neutrons colliding with light
nuclei, this ‘boundary’ will increase to the right i.e. s-wave scattering can be assumed to hold till higher energies as
depicted in Fig. 2.3.
54 Nuclear Engineering

FIGURE 2.2 Angular quantum number versus kinetic energy of neutron.

FIGURE 2.3 s-wave (l 5 0) and p-wave (l 5 1) scattering for neutron scattering.

Matlab Example 2.2: Angular quantum number plot the angular quantum number l versus energy E for elastic
scattering.

m=1;A=238;mu=m*A/(m+A);c=2.99e8;
h=4.1413e-15;% eV-s
m=1.6749e-27;% kg
mE=m*c^2*1e19/1.6; %eV
Radn=1.2e-15;% m
RadU238=Radn*A^(1/3);
KK=(2*pi/(h*c))*sqrt(2*mE)*(Radn+RadU238)
E=0:10:100;
l=KK*sqrt(E); % E in eV

The de Broglie wavelength is used to represent the neutron in the optical model (Hodgson, 1994; Lamarsh, 1966;
Reuss, 2008) as anpincident
ffiffiffiffi plane wave with amplitude ψBeikx and wave number k 5 2π=λ which is related to the neu-
tron energy as kB E. The incident wave is represented as a superposition of an infinite number of partial waves char-
acterized by the angular momentum and divided into cylindrical zones of radii which are multiples of h. It can then be
shown that elastic scattering is an s-wave phenomenon as presented with simplified reasoning above. These models
were presented at a time when the neutron was considered to be a particle and a wave but its composition as quarks
was unknown.
For the moment, theory presents neutron interactions as either potential scattering, in which there is no physical col-
lision but a scattering such that the neutron and nucleus both undergo deflections, or an actual collision where the neu-
tron enters the nucleus, stays in the compound nucleus for some time and then exits from the nucleus.
Elastic scattering, shown in Fig. 2.4, is thus considered to be an interaction, far from a resonance, where no com-
pound nucleus is formed but a two-body collision appears to take place. It can be modeled as potential scattering in
terms of the radius of the neutron (Rn 5 1:2 3 10215 m), and the radius of the target nucleus (RA 5 Rn A1=3 m) when it
 Interactions of neutrons with matter Chapter | 2 55

FIGURE 2.4 Selected neutron interactions.

takes the value σs 5 4πR2A . In the presence of a resonance, the elastic scattering cross-section has been modeled in terms
of a single resonance and potential scattering as well an “interference” term. All neutron interactions, such as inelastic
scattering, radiative capture and neutron multiplication ðn; 2nÞ shown in Fig. 2.4, other than elastic scattering away from
a resonance, undergo compound nucleus formation.
In an inelastic reaction an incident neutron enters the target nucleus adding its binding and kinetic energy to the
compound nucleus, resides for some time in the compound nucleus, and then exits with some energy retained in the
compound nucleus. The emitted particle can be a neutron itself or some other particle exiting from the nucleus. While a
neutron is in the compound nucleus, it exchanges energy with other nucleons; the outcome depends on whether or not it
imparts energy sufficient for a nucleon to be ejected. An ‘ejection’ takes place on the order of B10217 s, which is
instantaneous for all practical purposes, but large compared with the nuclear time, of the order of B10221 s. When a
neutron emerges from such an inelastic reaction, we describe it as a ðn; n0 Þ reaction. In case of high incident energy,
above a threshold, a neutron can result in multiplication ðn; 2nÞ or ðn; 3nÞ reactions.
A possible reaction is a direct interaction causing charged particles to be expelled from a nucleus as in the case of
such as ðn; αÞ and ðn; pÞ reactions.
In neutron interactions where a neutron enters the target nucleus and has insufficient energy to cause ejection of a
nucleon, it gets absorbed while the relaxation energy can be emitted from an isomer in the form of a gamma ray. When
the incident energy corresponds to one of the excited states of a nucleus, ‘resonance’ can take place where the cross-
section becomes very large. These states are called ‘metastable’ states if they exist for B10217 s.

2.2.1 Neutron scattering in the lab and center of mass systems

Neutron scattering with target nuclei is indeed a complicated phenomenon both theoretically and experimentally.
Scattering experiments are conducted in the laboratory (L) system with an incident neutron of velocity vL incident
on a stationary nucleus, while another system, called the center of mass (CM) system is used for mathematical simplic-
ity, as shown in Fig. 2.5.
56 Nuclear Engineering

FIGURE 2.5 Elastic scattering on a neutron and nucleus.

FIGURE 2.6 Scattering angle in Lab and center of mass systems.

Laboratory Center of mass

Neutron Nucleus Neutron Nucleus

Precollision Speed vL VL vC VC
Energy EL EAL  
Post collision Speed vL0 VL0 vC0 VC0
Energy EL0 0
EAL  
Scattering angle θL  θC 

Some simplicities of the CM system are as follows:

1. Velocities before and after collision: vC 5 vC0 ; VC 5 VC0
2. Momentum before collision and after collision is zero.
The CM system is useful for another property of scattering called isotropic scattering. The two systems are con-
nected, as shown in Fig. 2.6; the velocity of the center of mass VCM observable in the L-system. The incident neutron
^ in the L-system and scatters into the Ω
has direction Ω ^ 0 direction with Ω
^ ∙Ω
^ 0 5 cosθL ; notice that the azimuthal angle ϕ
is not shown in Fig. 2.6 as it is considered unchanged.
Applying the conservation of momentum
ðA 1 1ÞVCM 5 vL 1 AVL (2.3)
When VL 5 0, the velocities of the neutron and nucleus in the CM system are
A
vC 5 vL ;
A11
and
1
VC 5 2 vL:
A11
Energy conservation, assuming a target nucleus initially at rest, gives
EL 5 EL0 1 EAL
0
 Interactions of neutrons with matter Chapter | 2 57

which can be shown with the conservation of horizontal and vertical momentum, to yield a relationship between the
0
final energy of a neutron EnL and its initial energy EnL :
 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2
0 EL
EL 5 cosθL 1 A2 2sin2 θL :
ðA11Þ2
The above can be written as
1
EL0 5 EnL ½ð1 1 αÞ 1 ð1 2 αÞcosθC  (2.4)
2
where
2
A21
α5
A11
We can see from the above that if θC 5 0, EL0 5 EL 5 ELmax
0
(forward scattering) and when θC 5 π, EL0 5 ELmin
0
5 αEL
(back-scattering).
Thus for hydrogen, A 5 1, all the energy can be lost by a neutron in a single collision, while for a heavy nucleus
such as U238, α 5 0:9833 so that a neutron can lose just a small amount of energy since its minimum energy after colli-
sion is 0.9833EL ; a 1 MeV neutron striking a U238 nucleus will thus have an energy between 0.9833 and 1 MeV so that
the maximum energy transferred to the U238 nucleus can be 0.0167 MeV. The energy (in the Lab system) after a colli-
sion of a neutron with nuclei of hydrogen, carbon, iron and U238 is shown in the Fig. 2.7 below. For hydrogen, it is
seen that on average 50% of the neutron’s energy is lost in a collision.
From the horizontal and vertical components:
vL0 cosθL 5 VCM 1vC0 cosθC (2.5)
and
vL0 sinθL 5 vC0 sinθC (2.6)
yielding the relationship:
sinθC
tanθL 5 : (2.7)
1=A 1 cosθC

FIGURE 2.7 Neutron energy in Lab

system after a collision (initial energy
is 1 MeV).
58 Nuclear Engineering

Thus for heavy nuclei such as U238 (Ac1) θL  θC since the center of mass velocity VCM is much less than the inci-
dent neutron velocity in the laboratory system. This is a significant statement in the context of expressing isotropic scat-
tering in the C-system, which will differ from results of the L-system for light target nuclei.

Exercise 2.3: Scattering angles: from the scattering angles in the Lab and CM systems [Eq. (2.7)] find a relationship
between the solid angles dΩ 5 sinθ dθ dϕ in both systems.

The scattering cross-section has a smooth variation at low energies, like the absorption cross-section, but extends to
somewhat higher energies (typically a few MeV), then exhibits broad resonances, and then smoothens off.
At low energies, neutrons exhibit s-wave scattering for light as well as heavy nuclei, which is predominantly isotro-
pic in the center of mass system. At higher energies and for larger nuclear radius, the scattering is p-wave scattering
which is forward-biased rather than isotropic. Thus a neutron scattering off a U238 nucleus is bound to be forward-
biased i.e., favoring forward scattering rather than higher angles of back-scattering. Light nuclei, with a high threshold
for inelastic reactions, are more likely to undergo elastic collisions at low energies while heavy nuclei will mainly
undergo inelastic scattering.

Exercise 2.4: Solid angle: from Ðelementary

Ð 2π considerations
Ðπ of solid geometry, consider a volume element in a solid sphere
R
of radius R and show that dV 5 0 r 2 0 dϕ 0 sinθ dθ. Calculate the solid angle subtended by an electron on a nucleus.

2.3 The microscopic cross-section

The microscopic cross-section, with units of area, represents the measure of an interaction at the nuclear level. It can be
expressed as a probability of an interaction of a certain type. It is visualized as the cross-sectional area (cm2) of a sphere
which a neutron can “see” as it moves in matter. A small cross-section indicates a small probability of that particular
interaction. Since atoms and neutrons are very small, these cross sections are expressed in units of 10224 cm2 which
was code-named a “barn” during the Manhattan Project days.
Each interaction, such as an elastic scattering, inelastic scattering, absorption, fission, radiative capture etc. has an
associated cross-section. The sum of all cross sections is called the ‘total’ cross-section σt
σt 5 σs 1 σa 1 σf 1 σn;γ 1 σn;2n 1 . . .
The intensity of neutrons reduces as they interact with matter, as described in Section 1.3.3 for photons, since the
reduction dI is proportional to the intensity I, the atomic density N, the rate of interaction Nσt and the volume of mate-
rial dV 5 Adx, where A and dx are the cross-section area and thickness of the slab respectively as shown in Fig. 2.8.
The intensity of neutrons can then be found from the source intensity Io across a shield of thickness x from elemen-
tary considerations. The intensity across a thin strip dx is
I ðx 1 dxÞ 5 IðxÞð1 2 Nσt dxÞ
so that the change in intensity across the strip is dI ðxÞ given as
dI ðxÞ 5 I ðx 1 dxÞ 2 I ðxÞ 5 2 I ðxÞNσt dx
from which
dI ðxÞ
5 2 Nσt dx
I ðxÞ

FIGURE 2.8 Intensity reduction.

 Interactions of neutrons with matter Chapter | 2 59

Finally, integration over the width of the slab gives

ðX ðX
dI ðxÞ
5 2 Nσt dx
0 I ðxÞ 0

IðXÞ
ln 5 2 Nσt X
Ið0Þ
IðXÞ 5 Io e2Nσt X (2.8)
The above expression would hold under ideal conditions, that is, a point anisotropic source of intensity Io incident
on a thin shield. For a point isotropic source, the intensity would be given as
e2Nσt R
I ð RÞ 5 I o (2.9)
4πR2
to include both geometrical and material attenuation.

Exercise 2.5: Intensity: given an incident intensity of 108 neutrons/cm2/s on a composite slab consisting of a water
region 5 cm thick followed by an iron slab 3 cm thick, find the transmitted intensity.

Material At. Weight* Density (g/cm3) σa (b) σs (b)

Iron (Fe) 55.847 7.87 2.53 11
Water (H2O) 18.0153 1.00 0.664 103

1. Based on C12 (12 amu)

To obtain a probability of scattering, a normalized probability distribution function can be written as
1 dσ
f ðEÞdE 5 dE
σðEÞ dE
so that this probability, from initial energy E to final energy E0 (in the Lab system) is
σs ðθÞdΩðθ; ϕÞ
PðE-E0 ÞdE0 5 (2.10)
σs
Now, since dΩðθÞ 5 sinθC dθC dϕ 5 dμdϕ and with no preferential azimuthal angle, then
2πσs ðθÞdμ
PðE-E0 ÞdE0 5
σs
which, in the case of isotropic scattering in the center of mass system reduces to
dμ 1
PðE-E0 ÞdE0 5 5 2 sinθC dθC (2.11)
2 2
From Eq. (2.4)
dEL0 EL
5 ð1 2 αÞsinθC
dθC 2
giving
1
PðE-E0 Þ 5 (2.12)
Eð1 2 αÞ
This is a significant result in the context of s-wave scattering discussed in the previous section for which scattering
is isotropic in the CM system as is the case for light nuclei such as carbon 12 for low-energy neutrons. When kR , 1,
for light nuclei the scattering is always isotropic in the CM system while for kRc1, due to p-wave interference, the
scattering is anisotropic (in the forward direction).
60 Nuclear Engineering

Exercise 2.6: Probability distribution: Consider the forward and backward scattering cases for the differential scattering
cross-section by assuming linear forms and sketch the probability PðE-E0 Þ vs the final energy E0 which should be con-
stant for the isotropic case given above.

Exercise 2.7: Moments of scattering cross-section: Expand the angular scattering cross-section
  X M
2l 1 1  
Σs r; E0 -E; Ω ^ 0 -Ω
^ 5 Σsl ðr; E0 -EÞPl Ω ^
^ 0∙ Ω
l50
4π

to M 5 1 and explain what advantage there would be in using this expansion.

To further understand the process of slowing down, consider an infinite medium of hydrogen with a uniform distrib-
uted source emitting S neutrons/cm3/s with energy Eo : Assuming no absorption in hydrogen, we can write the probabil-
ity of scattering [Eq. (2.12)] into any energy E in dE as
dE
PðEo -EÞdE 5
Eo
Thus a neutron colliding with a hydrogen atom can lose all its energy as was shown in Fig. 2.7.

Exercise 2.8: Final energy after scattering: Calculate the probability that a neutron at an initial energy E 5 1MeV scat-
tering in hydrogen will have its final energy in the range 0.5 MeV1 MeV.

The resonance region has irregular peaks which can be described by models such as the Breit-Wigner model which
takes into account quantum-mechanical phenomenon and compound nucleus formation.
Fig. 2.9 shows the total ENDF/B-VIII.0 cross sections of selected light nuclei lithium-7, beryllium-9 and carbon-12,
in the energy range 1 3 1025 1 3 1028 eV plotted from http://atom.kaeri.re.kr/. In a low-energy range, 1 3 1025
eV1 3 1023 eV, the cross sections behave as

B
σt 5 A 1 pffiffiffiffi (2.13)
E
(A, B are constants). It is readily seen on a linear-linear scale that the carbon-12 cross-section, for example falls off
rapidly from B25 b at 1 3 1024 eV to B9 b at 1 3 1023 eV and then remains constant till B0.1 MeV. This low-
energy cross-section dependence is usually referred to as the 1=v-dependence and holds for light nuclei. At very low
temperature, below the thermal neutron energy B0.025 eV, the cross sections show temperature-dependent variations
due to crystalline effects.
In the energy range from B0.1 to B 10 MeV, resonances of varying widths and intensities are observed extend-
ing to about 20 MeV for carbon. At lower energies, the resonances are identifiable and are called resolved resonances
while at higher energies they become unresolved. For selected heavy nuclei, U-233, U-235, and Pu-239, the ENDF/
B-VIII.0 total fission cross sections are shown in Fig. 2.10 with the same three distinct regions for light nuclei but
with the resonance region having moved toward lower energies (few electron volts) and overlapping into ‘unre-
solved’ resonances.
Figs. 2.9 and 2.10 plotted from the point data evaluations available in ENDF/B files (Brown et al., 2018) are contin-
uous energy evaluations. From the figures, three energy groups, can be identified viz thermal, intermediate and fast, typ-
ically in the ranges 0 & 0.1 eV, 1 eV—few kilo electron volts, and few keV—MeV. In practice, these point cross
sections are processed and group averaged cross sections are prepared for use in nuclear design calculations. The
nuclear data processing system NJOY (Macfarlane, Muir, Boicourt, & Kahler, 2012) is used to prepare multigroup
cross-section data in general form or for specific codes and applications. For commonly used transport codes such as
ANISN (Engle, 1967), WIMS-D (Deen & Woodruff, 1995) and MCNP (Pelowitz et al., 2013), NJOY uses the modules
DTFR, WIMSR and ACER.
Some nuclear reactions, such as nuclear fission and multiplication take place when an incident neutron has energy
in excess of some threshold energy and are usually high energy reactions. One such reaction is the multiplication
ðn; 2nÞ reaction for which cross sections are shown in Fig. 2.11 for Be-9, U-238, and Pu-239.
Cross sections of selected elements are shown in Table 2.1 to indicate their probabilities of interactions. Thus
light nuclei such as carbon are predominantly scattering materials at thermal energy while plutonium-239 has very
high fission cross-section, and indium has very high absorption cross-section. These values have engineering design
 Interactions of neutrons with matter Chapter | 2 61

FIGURE 2.9 Total microscopic cross sections of Li-7, Be-9 and C-12.

applications. Similarly, boron-10 and xenon-135 have (thermal) absorption cross sections 3840 b and 2.65 Mb; thus
boron-10 is used as an absorber rod in nuclear reactors while xenon-135 is highly undesirable as it absorbs valuable
thermal neutrons which could otherwise have continued to produce fissions in the uranium fuel of a power reactor.
When microscopic cross sections are averaged over the energy spectrum in a nuclear reactor, they become
application-specific, and are more efficiently used than point cross sections, for carrying out large simulations.
Plutonium 239 cross sections are given in Table 2.2 averaged over a Maxwellian spectrum
Ð EU
2 EL σðE; T ÞM ðE; T ÞdE
, σM ðT Þ . 5 pffiffiffi Ð EU (2.14)
π
EL M ðE; T ÞdE

for the energy range EL 5 1025 eV; EU 5 10eV, a 1=E spectrum over ðEL 5 1025 eV; EU 5 10keVÞ
ð EU
1
, σri ðT Þ . 5 σðE; T Þ dE (2.15)
EL E

and a fission χðEÞ spectrum

Ð EU
EL σðE; T ÞχðEÞdE
, σ f ðT Þ . 5 Ð EU (2.16)
EL χðEÞdE
62 Nuclear Engineering

FIGURE 2.10 Total microscopic fission cross sections of U-233, U-235 and Pu-239.

where
rffiffiffiffiffiffiffiffiffiffi h i
4 ab E 1
χðEÞ 5 exp 2 2 sinh ðbEÞ2
πa3 b 4 a
ðEL 5 1025 eV; EU 5 20MeV; a 5 0:988MeV; b 5 2:249MeV21 Þ:
The Westcott g-factor is the ratio of the Maxwellian averaged cross-section to the 2200 m/s averaged cross-section
and has the value B1 for 1=v nuclides
, σM ðT Þ .
gð T Þ 5
σð0:0253eV; TÞ
Calculations of neutron thermal cross sections, Westcott factors, resonance integral averaged cross sections for 843
ENDF materials from major nuclear libraries are given at various temperatures (Pritychenko & Mughabghab, 2012).

Exercise 2.9: Generate group cross sections: Write a program to generate thermal group cross sections for Pu-239 by
curve-fitting the data in Fig. 2.10 (in the energy range 00.1 eV) and weighting them with a Maxwellian distribution
[Eq. (2.14)].

MT numbers are listed in the ENDF-6 Formats Manual (Trkov, Herman, & Brown, 2018). Nuclear cross sections from
the National Nuclear Data Center, Brookhaven National Laboratory released as ENDF/B-VIII.0 (Brown et al., 2018) are
the latest version with IAEA standards and benefit from experimental data and improvements in theory in simulation.
 Interactions of neutrons with matter Chapter | 2 63

FIGURE 2.11 (n,2n) reactions Be-9, U-238, and Pu-239.

TABLE 2.1 Thermal neutron absorption cross sections of selected elements.

Element σa (b) σs (b)

6 C
12 0.0034 4.8
238 7.6 8.3
92 U
Mn 13.3 2.3
In 197 2.2
94 Pu
239 1022 9.6

2.4 The macroscopic cross-section

The macroscopic cross-section Σ is an interaction rate per unit distance traveled by the neutron obtained by multiplying
the microscopic cross-section σ (a measure of the probability of interaction expressed as an area) by the atomic density
23
N. From Section 1.5, for the data given
P3 in Table 2.3, with ρ 5 19:1gcm , the thermal macroscopic absorption cross-
section Σa of natural uranium is Σa 5 Ni σa;i , where Ni 5 αi A. The atomic weight of natural uranium with the above
i51
64 Nuclear Engineering

TABLE 2.2 Averaged cross sections of Pu-239.

MTa Reaction 0.025 eV Maxwellian g-factor Resonance 14-MeV Fission spectrum

average integral average
, σM ðT Þ . g ðT Þ , σri ðT Þ . , σ f ðT Þ .
1 (n,total) 1.028 kb 1.110 kb 1.080  5.933 b 7.835 b
2 (n,elastic) 8.813 b 9.728 b 1.104  2.944 b 4.396 b
4 (n,inelastic) E-thr 5 7.894 keV 416.4 mb 1.581 b
16 (n,2n) E-thr 5 7.894 keV 229.6 mb 3.325 mb
17 (n,3n) E-thr 5 7.894 keV 2.318 mb 988.9 nb
18 (n,fission) 747.4 b 790.5 b 1.058 301 b 2.334 b 1.802 b
102 (n,gamma) 271.5 b 310.1 b 1.142 179.7 b 847.4 μb 52.61 mb
a
MT numbers from ENDF-6 Table formats.
Source: https://wwwndc.jaea.go.jp/cgi-bin/Tab80http://WWW.cgi?/data/JENDL/JENDL-4-prc/intern/Pu239.intern.

TABLE 2.3 Thermal absorption cross-section of some uranium isotopes.

Nuclide Atomic abundance (%) Atomic weight (u) σa ðbÞ

U-234 0.0057 234.04 103.47
U-235 0.72 235.04 680.8
U-238 99.27 238.05 2.73

specified atomic fractions is A 5 238:0179. This gives an atomic density

19:1 3 0:6023 1024
NU 5 5 0:0483 1023 atoms ∙ cm23 :
238:0179
from which the individual atomic densities N234 ; N235 ; N238 can be calculated. The thermal macroscopic cross-section is
Σa 5 0:3678/cm.

Exercise 2.10: Weight fractions of isotopes: From the atomic abundance, find the weight fractions of the uranium iso-
topes in Table 2.3.

The mean free path is the free-flight distance traveled between interactions and is the inverse of the macroscopic
cross-section; thus
1
λ5
Σ

Exercise 2.11: Half value layer thickness: For neutrons from Cf-252 calculate the Half Value Layer (HVL) in polyeth-
ylene, water, concrete and lead. From the HVL for gamma rays (Section 1.3) repeat your calculation to find the HVL
considering both neutron and gamma emissions from Cf-252.

Exercise 2.12: Average number of collisions: From the definition of the mean free path, estimate the average number
of collisions a neutron of 1 MeV and a gamma of 2 MeV would have in a 5-cm thick slab of iron. Justify any assump-
tions you make.

2.5 Flux measurement

In a nuclear reactor or any nuclear system with a source of neutrons, the energy spectrum is typically spread over sev-
eral orders of magnitude. Thus a wide range of neutrons are available for flux measurements based on the exposure or
fluence (IAEA, 1970) defined as the time integral of the neutron flux density with units of neutrons/cm2.
 Interactions of neutrons with matter Chapter | 2 65

TABLE 2.4 Transmutations of Mn-55, Ag-109, In-115, Au-197, and Dy-164 by neutrons.

Parent nuclide Reaction Radioactive nuclide

Identity % abundance σact (b) a

Identity T1=2 Radiations, MeVb
Mn 55 100 13.4 6 0.3 ðn; γÞ Mn 55 2.6 h β 2 ð2:81; 1:04; 1:0:65Þ; γð0:82; 1:77; 2:06Þ
Ag 109 48.65 2.8 6 0.5 ðn; γÞ Ag110m 270 days β 2 ðI:T Þ; γðcomplexÞ
Ag 109 48.65 110 6 20 ðn; γÞ Ag110 24.5 s β 2 ð2:24; 2:82Þ; γð0:66; 0:9Þ
In 115 95.8 145 6 15 ðn; γÞ In 116 m 54 min β 2 ð1:0; 0:87; 0:60Þ; γð0:14 2 2:19Þ
In 115 95.8 52 6 6 ðn; γÞ In 116 13 s β 2 ð2:95Þ; γðnoneÞ
Au 197 100 96 6 10 ðn; γÞ Au 198 2.7 days β 2 ð0:96; 0:29; 1:37Þ; γð0:41; 0:68; 1:09Þ
Dy 164 28.2 ,1000 ðn; γÞ Dy 165 2.4 hr β 2 ð1:25; 0:88; 0:42Þ; γð0:1 2 0:76Þ
a
for neutrons of 2200 m/s.
b
beta energies are maximum energies with average energy about 40% of maximum.
Source: Nuclear Engineering Handbook, 7-7, (1958), D. J. Hughes et al., “Neutron Cross Sections,” BNL-325, McGraw Hill Book Company, Inc., New York
1955; J. M. Hollander, I. Perlman and G. T. Seaborg, Table of Isotopes, Revs. Med. Phys. 25(2): 469651 (1953). Natl. Bur. Standards Circ. 499.

Thermal flux is measured by the activation of thin foils in a nuclear reactor for sufficient time to cause transmuta-
tion with beta emission, in most cases, which can subsequently be counted. For a flux φ0 for neutrons at a most proba-
ble speed of 2200 m/s corresponding to a temperature of T0 5 293.6 K, the integral value thermal flux for a reactor
temperature T can be found as
rffiffiffiffiffiffi
4T
φth 5 φ0
T0
Foils of manganese, silver, indium, gold, and dysprosium with isotopes and activation cross sections listed in
Table 2.4, are used.
The slowed down (epithermal) and thermal neutron components of the spectrum are separated using cadmium cov-
ers which discriminate between the two ranges, like a high-pass filter, due to the very high thermal absorption cross-
section of cadmium. Two measurements are carried out; in the first, a bare foil detector is used where the total activity
is measured and in the second, the foil is covered by cadmium so that it is subject to only epicadmium neutrons which
pass through the cadmium cover.
Fig. 2.12 shows the activation cross sections for Ag-109, Cd-113, In-115, and Au-197 with capture resonances, par-
ticularly of indium at 1.41 eV and gold at 4.9 eV, and high absorption cross-section of cadmium below B0.4 eV.
Cobalt-59 has also been used as a detector; its main advantages are that it exists in the form of this isotope with 100%
abundance, it is easy to make wires or thin disks and it can be used at high temperatures.
Fast neutron fluence is measured from threshold reactions such as ðn; αÞ and ðn; pÞ reactions using threshold activa-
tion detectors (Hyvönen-Dabek & Nikkinen-Vilkki, 1980) which have a good response to neutrons of a typical fission
spectrum. Commonly used detectors for dosimetry in materials use Fe54(n,p) and Ni58(n,p) monitors while for fission
monitors for very long irradiations, Ti46(n,p), Sc46 and Cu63(n, α) are used. Other methods include BF3 counters, ioniza-
tion chambers, U235 fission chambers, and semiconductor neutron spectrometers and silicon detectors (Shafronovskaia,
Zamyatin, & Cheremukhin, 2014; Zamyatin, Cheremukhin, & Shafronovskaya, 2017).
A thin foil of material such as Ag 109, In 115, and Au 197 is exposed to the neutron flux, which undergoes transmu-
tation to produce the radioactive nuclide. The amount of induced radioactivity will be proportional to the incident inten-
sity, exposure time and activation cross-section. Across the foil, the intensity will be given by Eq. (2.8) and the number
of nuclides PðtÞ can be found from the first-order rate equation (Section 1.2),
dPðtÞ
5 Σact φ 2 λP (2.17)
dt
In the production term, Σact φ, the neutron flux has been assumed to be time-dependent but space-independent due to
the assumption of a thin foil. Eq. (2.17) gives
Σact φ

Pð t Þ 5 1 2 e2λt
λ
When the foil has been exposed for T seconds, the number of atoms per unit volume of the foil, of radionuclide pro-
duced is Σact φT. The activity of the foil A 5 λP is subsequently measured in, for example, an ionization chamber
66 Nuclear Engineering

FIGURE 2.12 Cross sections of Ag-109, Cd-113, In-115, Au-197.

Exercise 2.13: Foil activation: A thin gold foil of mass 0.4 grams is placed in a nuclear reactor for one hour; subse-
quent measurements for the activity give 104 disintegrations per second. Estimate the thermal flux.

2.6 Reaction rates

The purpose of calculating the neutron flux in a system is to obtain reaction rates Rx where
Rx 5 Σx φV: (2.18)
all volumetric reaction rates and surface leakages are obtained from the flux.
In a more general form
ð ð
Rx 5 dx NðxÞ dE σx ðEÞφðx; EÞ (2.19)

The neutron flux can be used to estimate the reaction rate of a particular reaction within a region of interest. As an
example, the number of fission reactions taking place per second Rf in a volume V is given by
Rx 5 Nσx φ
In the Evaluated Nuclear Data File ENDF/B library, the ENDF/B-6 Formats Manual lists the MT numbers, from 1
to 999, of available reaction cross-section data. These are given for the incident neutron energy in the Laboratory sys-
tem. Each set has one (or more) Q value(s) depending on the possible number of outcomes of a reaction.
 Interactions of neutrons with matter Chapter | 2 67

TABLE 2.5 Commonly used reaction cross-section MT numbers.

MT Reaction Description
1 (n,total) Total cross-section
2 (z,z0) Elastic scattering cross-section for incident particle z 5 all particles (n, p, e. . .)
3 (z,nonelas.) Nonelastic neutron cross-section.
4 (z,n) Production of one neutron in the exit channel. Sum of the MT 5 50-91
16 (z,2n) Production of two neutrons and a residual1. Sum of MT 5 875-891, if they are present.
19-21 (n,f), (n,nf), (n,2nf) First-, second-, and third-chance neutron-induced fission cross-section
27 (n,abs) Absorption; sum of MT 5 18 and MT 5 102 through MT 5 117
102 (z,γ) Radiative capture
105 (z,t) Production of a triton, plus a residual. Sum of MT 5 700-749, if they are present.

Source: ENDF/B-6 Formats Manual.

FIGURE 2.13 Flux-to-dose conversion factors for neutrons and gammas.

In the following chapters in this book, simulations will be based on computing the neutron (or photon) flux in a sys-
tem and then using the reaction cross-section data, with the appropriate MT number, to compute reaction rates.
Some commonly used MT numbers are listed in Table 2.5.
Several other forms of data are also assigned MT numbers such as MT 5 500 for total charged particle stopping
power (Section 1.3), 502 for total photon interaction, 522 for photoelectric absorption, 527 for electro-atomic brems-
strahlung. Note that cross sections for charged particles and photons are also included.

Exercise 2.14: Reaction MT numbers: Which reaction rate MT numbers would be used to calculate the energy released
from fission reactions in UO2 fuel in the thermal, epithermal and fast groups?

Radiation dose can also be calculated from the neutron and photon flux-to-dose conversion factors shown in
Fig. 2.13; the H*10 dose due to neutrons and photons a distance 1 m away from a point isotropic source, are used to
estimate the radiation dose.

2.7 Neutron slowing down, diffusion and thermalization

In the transport process, a neutron at a source energy E0 , slows down to some energy, E, as it loses energy, or gains
lethargy u 5 lnE0 =E, in collisions by elastic and inelastic scattering. Since the threshold of inelastic reactions for low
mass nuclei is very high (BMeV) the energy loss by inelastic scattering is mainly from heavy nuclei. The fractional
energy loss of neutrons with light nuclei can be very high while only a small fraction of its energy is lost in collisions
with heavy nuclei such as uranium.
68 Nuclear Engineering

FIGURE 2.14 Typical neutron flux spectra in a micronuclear reactor.

FIGURE 2.15 Energy and lethargy scattering diagrams.

The neutron spectra in a micronuclear reactor (Aziz, Koreshi, Sheikh, & Khan, 2020) with a predominantly fast
spectrum in the matrix and a thermal spectrum in the water reflector is shown in Fig. 2.14. In fast reactors, for example,
the spectrum is “hard,” that is, predominantly high energy (BMeV) while in thermal reactors, it is “soft,” that is,
shifted toward lower energies. Since the fission cross-section is low at high energies and vice versa, fast reactors will
have a higher flux to maintain reactor power. Typical orders of magnitude of neutron flux in thermal and fast reactors
are B1012 n/cm2/s and B1015 n/cm2/s respectively.
Ignoring space dependence, the behavior of the energy-dependent flux φðEÞ in an infinite medium can be found
from elementary considerations. As shown in Fig. 2.15 the slowing down of a neutron from source (initial) energy E0
in the case of hydrogen medium (left fig) can take place to final energy E 5 0 as discussed in Section2.3. Note that the
source energy corresponds to zero lethargy since u 5 lnE0 =E since a source neutron has high energy and hence is the
opposite of a lethargic neutron. The energy range 0 , E , E0 thus corresponds to a lethargy range N . u . 0. The
figure on the right in Fig. 2.15 shows neutron collisions in a medium for which A . 1. In this case, a source neutron can
have energy in the range αE0 # E # E0 while a neutron at E0 can have energy Ev in the range αE0 # Ev # E0 .
Defining collision density FðEÞ5Σt ðEÞφðEÞdE as the total number of interactions at E in dE, the two contributions
will consist of a ‘direct’ contribution from the source and an ‘indirect’ contribution from collisions taking place at some
higher energy which lead to scattering into dE. Using Eq. (2.12), these two terms (for α 5 0) are
1. source neutrons falling into dE 5 SdE=Eo
Ð E0 5E Ð E0 5E 0 0
2. E0 5E o Σt ðE0 ÞφðE0 ÞdE0 PðE0 -EÞdE 5 dE E0 5E o Σt ðEEÞφ0 ðE Þ dE0
The balance equation for the collision density in dE is
ð E0 5Eo
SdE F ðE 0 Þ 0
F ðEÞdE 5 1 dE dE (2.20)
Eo E0 5E E0
 Interactions of neutrons with matter Chapter | 2 69

The above integral equation can easily be solved by differentiating to yield

dF F
52
dE E
from which the neutron flux is
S 1
φðEÞ 5 B (2.21)
Σt E E
Another quantity of interest is the slowing down density qðEÞ which is defined as the average number of neutrons at
some energy E0 . E falling below energy. The collision density is thus a reaction rate while the slowing down density
is analogous to the current J. It can be shown by similar reasoning that for the case of an infinite hydrogen medium
with a uniform source and in the absence of absorption, qðEÞ 5 S.
For neutron moderation for A . 1, the entire energy domain is not available for a neutron after collision, thus the
Placzek discontinuities shown in Fig. 2.16 are found for the collision density based on whether a neutron has had one
or more collisions while reaching the energy interval of interest. On the horizontal axis is the collision density for
hydrogen (A 5 1; α 5 0) while the infinitely heavy medium (A 5 N; α 5 1) has very little (or zero) energy transfer in a
collision hence it can have an infinite number of collisions in the first interval. For all other media, the wiggles or tran-
sients die out as shown and the collision density assumes a constant value (in lethargy).
When discontinuities appear only near the source, then in the asymptotic energy region, the solution can be shown to be
S
F5 (2.22)
ξE

FIGURE 2.16 Placzek wiggles for neutron scattering.

70 Nuclear Engineering

and subsequently
S 1
φ ðE Þ 5 B (2.23)
ξΣt E E
where ξ, or Δu, is the average change in lethargy per collision can be obtained using the probability for a neutron going
from E-E0
1
PðE-E0 Þ 5
Eð1 2 αÞ
so that
ðE
E
Δu 5 dE0 ln PðE-E0 Þ (2.24)
αE E0
where the integral has been carried out over final energy for the range permissible. This gives
ðA21Þ2 A11 2
Δu 5 ξ 5 1 2 ln D :
2A A21 A 1 2=3
Thus for neutrons colliding in hydrogen ξ 5 1 i.e. the change in lethargy is maximum while for A .. 1, ξ{1. The
number of collisions needed for a neutron to slow down from 1 MeV to 1 eV assuming isotropic scattering in the CM
system is shown in Table 2.6.

Exercise 2.15: Collision density: Write an expression for the collision density for neutrons reaching the energy interval
dE (Fig. 2.15) after one collision F1 ðEÞ.

The smallest number of collisions is for hydrogen (13.6) while the largest is for uranium (1649); these values have
are important for the selection of materials for achieving a neutron spectrum of interest.

Matlab Example 2.3: Number of collisions: Calculate the number of collisions in hydrogen, carbon, iron and uranium-
238 for a fission neutron at 1 MeV to slow down to energy 1 eV.

Exercise 2.16: Lethargy: Calculate the average gain in lethargy per collision ξ for light water and heavy water.

Extending the above analysis for hydrogen, it can be shown that for A .. 1
ð E0 5E=α
E 2 αE0
qð E Þ 5 F ðE0 ÞdE0 5 Cξ (2.25)
E0 5E ð1 2 αÞE0

TABLE 2.6 Number of collisions needed to slow down from 1 MeV to 1 eV.

Nucleus Mass no α ξ n
Hydrogen 1 0 1.0000 13.8
Deuterium 2 0.111 0.7253 19
Beryllium 9 0.640 0.2066 66.9
Carbon 12 0.716 0.1578 87.6
Oxygen 16 0.779 0.120 115.2
Sodium 23 0.840 0.0845 163.5
Iron 56 0.931 0.0353 391.5
Uranium 238 0.983 0.0084 1649
 Interactions of neutrons with matter Chapter | 2 71

where for no absorption in which case all neutrons eventually slow down past energy E, C 5 S=ξ, so that again
qðEÞ 5 S. Extensions to the case of nonmonoenergetic source show the same result for hydrogen as obtained for the
monoenergetic source considered above. For mixtures, again a similar dependence is found except that properties are
averaged over the elements of the mixture.
The above results can be used to estimate a moderating time tm defined as the time taken by a neutron to slow down
from some high energy, (e.g., fission energy) to some cutoff energy Em . If the cutoff is set at 1 eV, then it is above the
thermal range and generally ‘beyond’ the resonances. The energy group lower than Em , is the ‘thermal’ group; we now
have two distinct regions for slowing down and diffusion. In the first region, the neutrons will slow down or undergo
moderation; in the second, from 1 eV down to some equilibrium energy, neutrons will diffuse for a further diffusion
time td until they are captured or reach equilibrium.
The moderating time is easily found from Fermi’s “age” theory which assumes an “average” behavior for the slow-
ing down region and assumes that Fick’s law is valid for all energies, which is clearly a poor assumption especially for
hydrogenous media. The average number of collisions n is found using Δu [Eq. (2.24)] giving
du dE
n5 5
Δu Eξ
In Fermi’s continuous slowing down model, an average behavior is assumed for each neutron undergoing collisions.
A continuity equation for the neutrons crossing “boundaries” of a lethargy interval is written, using Fick’s Law which
introduces neutron flux φ in terms of the slowing down density qðuÞ. Using the relationship
ðu
qðr; uÞ 5 F ðr; u0 Þdu0
u2ξ

the “Fermi age equation” is written as

@qðr; τÞ
r2 qðr; τ Þ 5 (2.26)
@τ
where the neutron age τ (with units of area) is defined as
ðu
DðuÞ
τ5 du (2.27)
0 ξΣs ðuÞ

The interpretation of the age is possible by expressing the lethargy increase as du 5 ξΣs dx, i.e. the increase in leth-
argy per collision multiplied by the number of scattering collisions. Thus
ðu
DðuÞ
τ5 ξΣs ðuÞdx 5 DlðuÞ
0 ξΣ s ðuÞ

where the distance lðuÞ is a measure of the movement which results in a lethargy increase from 0 to u. Eq. (2.26) can be
solved to obtain the slowing down density q with boundary conditions for the source qðr; 0Þ 5 SðrÞ, flux and current
continuity at the interface conditions requiring, and at the surface where q vanishes, The flux continuity is
qðr; uÞ
φðr; uÞ 5
ξΣs ðuÞ
while the current continuity is to preserve
D
J 5 2 Drφðr; uÞ 5 2 rφðr; uÞ
ξΣs ðuÞ
For a plane source in an infinite slab, the solution to the Fermi age equation is very illustrative for the physical
understanding of neutron age. The slowing down density is
S
qðx; τ Þ 5 pffiffiffiffiffiffiffiffi e2x =4τ
2
(2.28)
4πτ
which gives a picture of neutrons moving outwards from a source in the form of a Gaussian distribution as they settle into
the host medium. In 3D, the number of neutrons that started at r 5 0 and have slowed down to within a shell dr at r is
S
e2r =4τ
2
qðr; τ Þ 5 4πr 2 dr (2.29)
ð4πτÞ3=2
72 Nuclear Engineering

The probability of finding a neutron in a spherical shell is thus

1
e2r =4τ
2
pðr Þ 5 4πr 2 (2.30)
ð4πτÞ 3=2

Since
ðN pffiffiffiffiffiffiffiffi
2 2r2 =4τ 4τ 4πτ
r e dr 5
0 4
the total probability is
ðN ðN pffiffiffiffiffiffiffiffi
4π 2 2r2 =4τ 4π
4τ 4πτ
drpðr Þ 5 drr e 5 51
0 ð4πτÞ3=2 0 ð4πτÞ3=2 4
With this probability, the variance or spread will be a measure of the age of neutrons in terms of the physical spread.
The second moment (measure of variance) is then
ðN
r 2 pðr Þdr 5 r 2 5 6τ
0

The physical significance of “age” is that

1 2
r τ5
6
The neutron age, shown in Fig. 2.17 is thus a measure of how far a neutron travels, or what its age becomes, in
going from a high energy to some low energy.
The measured age of fission neutrons slowing down to the indium resonance energy at 1.45 eV is given in
Table 2.7. Neutrons in water moderate in much shorter a distance than they do in graphite implying that neutrons reach
further into a nonhydrogenous material than into a hydrogenous material.
The moderation time for slowing down from E0 to E is obtained by integration:
2 1 1 2
tm 5 2 B ; ðv0 cvm Þ (2.31)
ΔuΣs vm v0 ΔuΣs vm
The moderating time for water, heavy water and beryllium is B1:0 μs, B8:1 μs, and B9:3 μs respectively. Estimates for
the diffusion time, obtained from the speed and mean free path as: td 5 vth =λa , give B100:0 μs, B1:5 3 105 μs,
B4:3 3 103 μs and B23 μs for water, heavy water, beryllium and graphite. The diffusion length L, of neutrons is defined as
D 1
L2 5 ; where D 5
Σa 3Σtr

FIGURE 2.17 Neutron slowing down density.

TABLE 2.7 Age-to-thermal (fission to 1.45 eV) of selected materials.

Moderator H2O D2O Be BeO Graphite

τ In (cm )
2
B 26 111 85 80 311

Source: Lamarsh, J. R. Introduction to Nuclear Engineering, p.203.

 Interactions of neutrons with matter Chapter | 2 73

TABLE 2.8 Diffusion length of selected materials.

Moderator Diffusion length L (cm)

H2O 2.85
D2O 116
C 54

Source: Murray R. L., Nuclear Energy, Elsevier, 6th (Ed.) p.55.

FIGURE 2.18 Neutron collision paths in water and carbon.

in terms of the transport cross-section Σtr 5 Σs ð1 2 μ 0 Þ for average cosine of the scattering angle μ 0 5 2=ð3AÞ.
Table 2.8 lists the values of some diffusion length in some moderating materials.
A schematic of the moderation and diffusion of neutrons in water and graphite is shown in Fig. 2.18.
Total cross-section ,65., ,, 35 . . carbon and water, good picture ,, p.38 .., theory ,, 39 ...
When neutrons reach thermal energies, they are in equilibrium with the host nuclei which have similar energies, and
hence neutrons can receive energy from the nuclei, or up-scatter in contrast to their slowing which was down-scattering
only. The transport and energy exchange of thermal neutrons takes into account the binding of nuclei in the host
medium. It is found that the cross sections of water are substantially different from the cross-section obtained by the
sum σwater 5 2σH 1 σO . The distribution is

1=2 ðN
2πnN 2
Fa 5 σs ðEÞe2E=kT dE
ðπkTÞ3=2 μ 0

ð
N
X
Fa 5 a
ðEÞφμ ðEÞdE (2.32)
0
P
where a ðE Þ 5 Nσa ðE Þ and φμ ðEÞ is given by

1=2
2πn 2
φ μ ðE Þ 5 Ee2E=kT (2.33)
ðπkTÞ3=2 μ

The thermal corrections for thermal flux at temperature Tand cross sections with 2200 m/s neutrons is:

1=2
νT T
5
νo To

1=2 pffiffiffi 1=2

2 T π To
φT 5 pffiffiffi φo ; and Σ5 gð T Þ Σ ð Eo Þ (2.34)
π To 2 T

where gðTÞ is the non-1=v Westcott factor shown in Table 2.9 for U-235, U-238, and Pu-239.

Exercise 2.17: Gold foil activation: Repeat the gold foil calculation of Exercise 2.13 with the thermal cross-section
assuming the temperature in the reactor was 100 C.
74 Nuclear Engineering

TABLE 2.9 Non-1=v factors.

T, oC U235 U238 Pu239

ga gf ga ga gf
20 0.9780 0.9759 1.0017 1.0723 1.0487
100 0.9610 0.9581 1.0031 1.1611 1.1150
200 0.9457 0.9411 1.0049 1.3388 1.2528
400 0.9294 0.9208 1.0085 1.8905 1.6904
600 0.9229 0.9108 1.0122 2.5321 2.2037
800 0.9182 0.9036 1.0159 3.1006 2.6595
1000 0.9118 0.8956 1.0198 3.5353 3.0079

2.8 Resonance cross-section

Resonances for light and heavy nuclei cross sections were seen in Figs. 2.9 and 2.10 to occur as peaks where the cross
sections jump to very high values. An understanding of the phenomenon of resonance is based on the quantum-
mechanical structure of the nucleus where excited levels exist, and when an incident neutron adds energy to the nucleus
which matches the level of an excited state, the cross-section shows a sudden jump. The terminology of resonance is
similar to that in conventional mechanical and electrical systems where the amplitude of a response suddenly increases
when an induced signal matches a natural frequency of the system.
The probability of an interaction in which a compound nucleus is formed is the product of the probability that a
compound nucleus is formed, and subsequently that it relaxes in a particular emission such as a neutron emission or a
gamma emission. A reaction α-β where α 5 n 1 A and β 5 b 1 B thus proceeds in two stages as α-CN-β.
2
This probability PðtÞ 5 ΨðtÞ is obtained from the wave function
ΨðtÞ 5 Ψð0Þe2iE0 t=h̄ e2t=ð2τÞ
To find the probability of finding a state at energy E, taking the Fourier transform
ð
1 N
ΦðEÞ 5 pffiffiffiffiffiffi dt eiEt=h̄ ΨðtÞ
2π 2N
yields
ih̄Ψð0Þ 1
ΦðEÞ 5 pffiffiffiffiffiffi
ih 
2π ðE2E0 Þ2 1 2τ
¯

Now
2
PðEÞ 5 ΦðEÞ
h̄2 jΨð0Þj2
which gives PðEÞ 5 1
-h Þ2
2π ðE2E0 Þ2 1 ð2τ

The width of a resonance, denoted by Γ (energy) is given by the Heisenberg uncertainty relation
h̄
ΓD
τ
where τ is the lifetime. Thus
N0
Pð E Þ 5
ðE2E0 Þ2 1 Γ2
where the normalization constant is
- 2 jΨð0Þj2
N0 5 h
2π
 Interactions of neutrons with matter Chapter | 2 75

The cross-section for compound nucleus formation is thus

1
σCN ðEÞB (2.35)
ðE2E0 Þ2 1 Γ2 =4

Γ 1
σCN E 5 E0 6 5 σCN ðE 5 E0 Þ
2 2

With nuclei at rest (T 5 0oK), the ith cross-section is given by the BreitWigner formula

1=2
E0 Γn Γ i
σi E 5 π λ~ 1 g
2
(2.36)
E ðE2E0 Þ2 1 Γ2 =4

where Γ 5 Γn 1Γi and

h̄ ð2JC 1 1Þ
λ~ 1 5 pffiffiffiffiffiffiffiffiffiffiffi ; g 5
2μE0 ð2Jn 1 1Þð2JA 1 1Þ

with

4π λ~ 1 gΓn
2
2 σ 1 Γi 1
x5 ðE 2 E0 Þ; σ1 5 ; σ i ðxÞ 5
Γ Γ Γ 1 1 x2
For radiative capture i 5 γ, for fission i 5 f , . . .
For the scattering cross-section, the BreitWigner formula has i 5 n with the potential scattering p 5 4πR2 and an
interference term
Γ2n pffiffiffiffiffiffiffiffi Γn ðE 2 E0 Þ
σs 5 πγg 1 2 πgp 1p (2.37)
ðE2E0 Þ2 1 Γ2 =4 ðE2E0 Þ2 1 Γ2 =4

σ 1 Γn 1 2σ1 R x
σ s ðxÞ 5 1 1 4πR2 (2.38)
Γ 11x 2
λ~ 1 1 1 x2

The total cross-section is

Γn Γ pffiffiffiffiffiffiffiffi Γn ðE 2 E0 Þ
σt 5 πg 1 2 πgp 1p (2.39)
ðE2Eo Þ 1 Γ =4
2 2
ðE2Eo Þ2 1 Γ2 =4

At higher energies, for the above light nuclei * 20 MeV, the total cross-section falls off rapidly as 1=v. This behav-
ior for light nuclei is also observed for some magic-number heavy nuclei.
The total cross-section for uranium-238 shown in Fig. 2.19 has resonances in the intermediate region similar
to those for U-233, U-235, and Pu-239 (Fig. 2.10). Comparing the resonance regions of Figs. 2.5 and 2.6, it is seen
that the resonances of uranium-238 extend in the region from B 6 eV to B 20 keV are sharper and resolved at the
lower limits but become unresolved at higher energies. The cross sections for heavy nuclei generally exhibit this
trend.
The first few resonances of uranium-238 are at 6.67, 20.8, 36.7, and 66.03 eV with relative widths 27.5, 31.8, 63,
and 49 meV, respectively. These correspond to excited states of the nucleus which occur at discrete levels similar to the
excited states of atoms; the difference being that these nuclear excited states are separated at energies of MeV rather
than in eV for atoms.
When ΓB27meV, τB2:44 3 10214 s which is a very ‘long’ time and hence a gamma would be emitted as in the
time in between, no nucleon had sufficient energy to be emitted. Sharp peaks are thus radiatively captured peaks while
broad resonances represent resonance scattering.
Each excited state is called an isomer and with the ground state, it makes an “isomer pair.” The return to the
ground state is by gamma emission which could be “delayed” due to spin effects. This delay is orders of magnitude
76 Nuclear Engineering

FIGURE 2.19 Total microscopic cross-section of U-238.

of compound nucleus formation which is B10214 s (compared with B10222 s for a new nuclear bond to form). The
distance between the excited states decreases, going to higher energies, till the isomers overlap in the “continuum”
domain. States which do not return to the ground state immediately and ‘hold on’ for a while are called metastable
isomers (Fig. 2.20).
Fig. 2.21 shows the total and scattering cross sections of U-238 showing the relative magnitudes and asymmetry.
The potential scattering cross-section is the value of σs far from the resonance [Eq. (2.38)].
A simple model to understand a resonance cross-section is the BreitWigner one-level model [Eq. (2.36)] demon-
strated in Matlab example 2.4.

Matlab Example 2.4: Breit-Wigner Formula: Demonstration of BreitWigner single resonance formula.
 Interactions of neutrons with matter Chapter | 2 77

FIGURE 2.20 First few resonances of U-238.

The above discussion has assumed that nuclei are at rest, or at absolute zero K. when the cross-section depends on
the energy of the neutron, the nuclide and quantum-mechanical phenomena. As temperature increases, the thermal
vibration of nuclei increases and within the nucleus more states are available for capture by the neutron. When relative
motion between neutrons and nuclei becomes important, the radiative capture cross-section is
1
σ 1 Γγ E1 2
σ 1 Γγ
σ γ ðE; T Þ 5 ψðς; xÞB ψðς; xÞ (2.40)
Γ E Γ

where
ðN
ς exp 2 14 ς 2 ðx2yÞ2
ψðς; xÞ 5 pffiffiffi dy
2 π 2N 1 1 y2

This effect of Doppler broadening is shown in Fig. 2.22 with the zero K value plotted from the BreitWigner for-
mula (Matlab example 2.4).
From Fig. 2.22, plotted for Eq. (2.40) and demonstrated in Matlab Example 2.5, it is seen that the area under the
curves remains the same as they reduce in value but increase in width with an increase in the probability of resonance
capture.
78 Nuclear Engineering

FIGURE 2.21 Total and scattering microscopic cross-section at first resonance of U-238.

FIGURE 2.22 Doppler broadening due to temperature.

 Interactions of neutrons with matter Chapter | 2 79

Matlab Example 2.5: Doppler Broadening: Doppler broadening due to temperature.

The resonance escape probability can be written as

ð Ei1
Σa dE
pi 5 1 2 (2.41)
Ei2 Σ s 1 Σa ξE
but this expression cannot be solved analytically since the flux across a resonance is not known analytically. In the
above, scattering can be considered to take place due to the moderator while absorption would be due to the presence
of absorber (fuel). A dilution ratio representing the moderator to fuel ratio is used in the analysis, with a dilution cross-
section σd defined as
σd 5NM σsM =NaA
Thus approximate methods are used starting from a single resonance model with infinite and finite dilution with the
narrow resonance (NR) and the narrow resonance infinite mass (NRIM) models with and without the effect of tempera-
ture. When resonances are separated and narrow, they can be treated separately, and the resonance escape probability
can be expressed as the product of individual probabilities pi as
N
p 5 L pi
i51 2 3
ð Ei1 (2.42)
Σa dE5
p 5 exp4 2
Ei2 Σ s 1 Σa ξE

with Σa 5 NA σγA and ΣP 5 NA σPA 1 ΣsM 5 constant. The scattering cross-section is written as
σ 1 Γn 1 2σ1 R x 0
σ s ðxÞ 5 1 1 4πR2 5 σsA 1 σPA
Γ 1 1 x2 λ~ 1 1 1 x2
with
σ 1 Γγ 1
σγA ðxÞ 5
Γ 1 1 x2
giving
 ð Ei1 
Σa dE
p 5 exp 2 0
Ei2 NM σ sM 1 N A ðσ sA 1 σ PA Þ 1 Σa ξE

or in terms of the dilution cross-section σd 5 Σp =NA

 ð Ei1 
NA σγA dE
p 5 exp 2 0 (2.43)
ξΣP Ei2 1 1 ðNA =ΣP ÞðσsA 1 σγA Þ E
80 Nuclear Engineering

The resonance escape probability is then

 
NA I
p 5 exp 2 (2.44)
ξΣP
with the effective resonance integral, for resonance in the range ðEi2 , Ei1 Þ defined as
ð Ei1
σγA dE
I5 0 (2.45)
Ei2 1 1 ðNA =ΣP ÞðσsA 1 σ γA Þ E

In Eq. (2.44) the average lethargy gain averaged over the moderator and absorber is
ξ M ΣsM 1 ξA ΣpA ξ ΣsM 1 ξA ΣpA
ξ ðE Þ 5 5 M
ΣsM 1 ΣpA ΣP
Thus ξ ðEÞΣP 5 ξM ΣsM 1 ξ A ΣpA and finally
 
NA I
p 5 exp 2 (2.46)
ξM ΣsM 1 ξA ΣpA
Eq. (2.46) is used for calculating the resonance escape probability in reactors.

Exercise 2.18: Resonance escape probability: Given a graphite moderator-uranium (U238) absorber system with
the data: NM 5 7:96 3 1022 cm23, NM 5 3:98 3 1020 cm23, σSM 5 4:8 b, σγA 5 7:64 b, I 5 73 b, calculate the resonance
escape probability.

2.9 Nuclear fission

Nuclear fission, first observed by Otto Hahn and Fritz Strassmann in December 1938, named and conceptualized by
Lisa Meitner and Otto Frisch as depicted in Fig. 2.23, was a revolutionary breakthrough that led to the production of
sustainable energy from a nuclear reaction. This discovery laid the basis for the Manhattan Project in which nuclear
weapons were designed, tested and used at the end of the second world-war in 1945. Subsequently, nuclear fission was
developed for several useful applications including naval propulsion, electricity generation from nuclear reactors, and
the production of radioisotopes for medicine and agriculture.

2.9.1 The fission process

The nuclear fission reaction takes place when a neutron incident on a nucleus gets absorbed in it leading to the breakup
of the nucleus into two fission fragments accompanied with the release of two or more energetic neutrons, radiation and
a burst of energy. Several combinations of fission fragments can be emitted; one possible reaction shown in Fig. 2.23 is

0 n 1 92 U-54 Xe 1 38 Sr 1 20 n 1 200 MeV

1 235 140 94 1

with a neutron striking a 92U235 atom resulting in a compound nucleus 92U236 which fissions almost spontaneously with
energy released of the order of 200 MeV (3.2 10211 J). The mean lifetime of the compound nucleus formed was initially

FIGURE 2.23 Nuclear fission.

 Interactions of neutrons with matter Chapter | 2 81

TABLE 2.10 Spontaneous fission rates.

Nuclide Half-life (years) Spontaneous fission rate (n/g-s)

Uranium 235 7:04 3 108
7:43 3 1024
Uranium 238 4:50 3 109 1:80 3 1022
Plutonium 238 87:70 2:64 3 103
Plutonium 239 2:40 3 104 2:30 3 1022
Plutonium 240 7:00 3 103 8:43 3 102
Plutonium 242 7:04 3 108 1:80 3 103
Curium 242 7:04 3 108 1:80 3 107
Curium 244 7:04 3 108 1:17 3 107

estimated (Feather, 1939) to be of the order of B 10213 s. By B 10212 s, the fission fragments have lost their energy
and come to rest with the subsequent release of radiation.
This is indeed a very small amount of energy compared with that involved in applying a 10 N force and moving an
object by 1 m (10 J). The power produced by a “small” nuclear reactor capable of supplying electricity to a small city
is B300 MW which means energy of B9.46 1015 J in one year. To produce this much energy, we would need to fission
B2.96 1026 atoms, or about 490 gram-atoms (B 115 kg) of U235. But that would require a 100% enrichment of ura-
nium and the fission of every single atom in the material which is not possible because natural uranium has only 0.71%
of the isotope U-235 and the absorption cross-section of U-238 is much lower than that of U-235 (Table 2.3). Due to
this and other reasons, a nuclear reactor producing 300 MW of power would require about 30,000 kg of uranium fuel
enriched to only 3% of the isotope U235. With this much fuel, it would only be a cylinder of height and diameter 2 m
which means that very high energy density is achievable, and that nuclear fission is a great source of energy.
During fission. there is also a probability that three fragments (two main and a light particle) and neutrons are pro-
duced. This emission occurs in less than one in hundred fissions. The process is sometimes called ternary fission as a
third fragment is produced. In Table 2.5, cross-section MT numbers in ENDF/B files are given for first-, second- and
third-chance fissions for reactions ðn; f Þ, ðn; nf Þ, and ðn; 2nf Þ reactions, respectively.
While fission can be induced as discussed above, it also takes place on its own; such spontaneous fission is due to a prob-
abilistic arrangement where tunneling leads to the occurrence of fission. The nuclear fission yields listed in ENDF/B-VIII.0
library released in 2018 by the National Nuclear Data Center Brookhaven National Laboratory have sub-libraries No.5 for
spontaneous fission yields with nine elements (uranium-238, curium-244,246,248, californium-250,252, einsteinium-253 and
fermium-243,256) and No. 11 for induced fission yields with 31 elements (from thorium-227 to fermium-255). Table 2.10
lists the spontaneous fission rates of some uranium, plutonium and curium isotopes. The high spontaneous fission of Pu-238,
together with its half-life of B 88 years, makes it an attractive radioisotope power source (Section1.0)
The reason that spontaneous fission takes place is the instability caused by a high N/P ratio (stability curve
Section 1.1); this ratio remains 1.0 till calcium 40 and then increases to B1.5 causing an excess of neutrons for heavy
nuclides which is “corrected” by alpha decay. In fission, neutrons are released but the ratio remains high so that beta
decay takes place in which N is reduced while mass number remains the same charge increases (Fig. 2.24).
The N/P ratio is then reduced by successive isobaric (same number of A) beta decays β 2 with, in most cases,
increasing half-lives until a stable configuration is achieved such as in the Krypton 93 decay to stable niobium 93
(Hanson et al., 2016) (Fig. 2.25).

2.9.2 Critical energy

Fission takes place with the compound nucleus formation as an intermediate stage in which the binding energy as well
as the kinetic energy of the absorbed neutron is added to the target nucleus. This energy is then shared with the
nucleons while the compound nucleus readjusts to a new form depending on the energy sharing process. To break the
nucleus, energy must be sufficient enough to give it potential energy larger than it had in its original form; the threshold
energy is the height of the coulomb barrier
Z1 Z2 ke2
EC 5
ðR1 1 R2 Þ
82 Nuclear Engineering

FIGURE 2.24 An isobaric decay chain.

FIGURE 2.25 Isobaric decay chains in uranium and plutonium fission.

The difference between this energy and the Q value of fission is then a minimum requirement called the critical
energy for fission to take place.
The Q value of a nuclear reaction was defined in Section1.1 as the difference of the rest mass energies of the reac-
tants r1 ; r2 and products p1 ; p2 :


Q 5 mr1 1 mr2 2 mp1 1 mp2 c2
with Q . 0 for endothermic or endoergic and Q , 0. For fission, the Q value varies depending on the fission fragments;
it is thus measured from the energy deposited over all fissions. For the fission reaction in Example 2.1, the Q value is
B182 MeV.

92 U 1 0 n-56 Ba 1 36 Kr 1 0 n
Example 2.1: Q value: calculate the Q value of the fission reaction 235 1 144 90 1

Reactants Products

Reactant Mass (a.m.u) Product Mass (a.m.u)

92 U 56 Ba
235 235.044 144 143.92(11.5 s)
0n 36 Kr
1 1.0082 90 89.92 (32.32 s)
2 3 10 n 2.0164
Total 236.0522 235.8564
Q 5 0:1958 3 931:5 5 182:39 MeV

Source: Isotope mass was obtained from the Commission on Isotopic Abundances and Atomic Weights. http://www.ciaaw.org, 92235U https://ciaaw.org/
uranium.htm, 56144Ba https://pubchem.ncbi.nlm.nih.gov/element/Bariumsection 5 Atomic-Mass-Half-Life-and-Decay.
 Interactions of neutrons with matter Chapter | 2 83

Critical energies of some fissile nuclides is shown in Table 2.11.

Matlab Example 2.6: Critical energy.

The energy released in fission of U235 is shown in Table 2.12. With the exception of the energy associated with
neutrinos, all energy is recoverable. Most of the fission energy, over 80%, is carried by fission fragments; this is
released within the fuel as will be discussed in detail in Chapter 3. Charged particles such as beta rays, as discussed
in Chapter 1, will also have a short range. Gamma rays from fission, as well as those produced from radiative cap-
ture, would travel a few centimeters into the surroundings of the fuel, while neutrons would be expected to go little
further.
From elementary considerations, the energy produced in fission can be estimated from the binding energy curve.
For U238 the binding energy is 7.5 MeV/nucleon; if the nucleus breaks into two nuclei of mass 238/2 5 119 each, for
which the binding energy per nucleon is 8.4 MeV/nucleon then the gain is 0.9 MeV/nucleon so that the energy
released would be 238*0.9B214 MeV. These numbers compare well with Table 2.12 where a detailed breakdown is
given.

TABLE 2.11 Critical energy.

Nuclide Critical energy (MeV)

Th 232 5.9
U 233 5.5
U 235 5.75
U 238 5.85
Pu 239 5.5

Source: Lamarsh, Introduction to Nuclear Reactor Theory, p.85.

TABLE 2.12 Energy released in a nuclear fission reaction.

Product MeV
Fission fragment (kinetic energy) 166168
Fission neutrons (kinetic energy) 5
Prompt gamma rays 7
Fission product gamma rays 7
Fission product beta particles 58
Neutrinos 1012
Total 200207
84 Nuclear Engineering

2.9.3 Fission yield

The fission yield is defined in terms of the independent fission yield which is the number of atoms of a specific nuclide
produced “immediately” in 100 fission reactions, the cumulative fission yield which is the total number of atoms of a
nuclide produced in 100 fission reactions, and the chain yield, which is the total chain yield, after the emission of
prompt neutrons, for fission is cumulative yield of the last chain member. There are at least 500 nuclides which can be
produced in fission directly or by beta decay of precursors depending on the incident energy and the mass and charge
distributions. Several models have been proposed to estimate the fission yield. One such model is based on the partition
of the nuclear yield on isomeric states (Madland & England, 1977; Okumura, Kawano, Jaffke, Talou, & Chiba, 2018)
and the spin distribution of the fragments.
We cannot say with certainty what the fission fragments will be as there is a probability of emission of fragment
fragments with conservation of atomic number and atomic mass (before and after reaction). For U235, the distribution of
fission fragments is shown in Fig. 2.26. This shows that there is a high probability of getting two fission fragments of
mass numbers 95 and 140. Some nuclides close to A 5 140 are Tellurium 135, Iodine 135, neon 135, Cesium 135 and
Barium 135.
A possible nuclear fission reaction is 10 n 1 235
92 U-54 Xe 1 38 Sr 1 20 n 1 200 MeV, with both fission fragments xenon
140 94 1

and samarium being unstable like other fission fragments and undergo beta decay.

2.9.4 Number of neutrons emitted in fission

The number of neutrons emerging from a fission reaction, ν, varies between zero and seven; for U235 the average num-
ber of neutrons is ν B2:5. As a function of energy,
ν ðEÞ 5 ν 0 1 aE
where ν 0 5 2.43, a 5 0.065/MeV in the range 01 MeV, and ν 0 5 2.45, a 5 0.150/MeV for E . 1 MeV. In U 233,
ν 0 5 2.48, a 5 0.075/MeV in the range 0 2 1 MeV, and ν 0 5 2.41, a 5 0.136/MeV for E . 1 MeV. In Pu 239,
ν 0 5 2.87, a 5 0.148/MeV in the range 0 2 1 MeV, and ν 0 5 2.91, a 5 0.133/MeV for E . 1 MeV. These prompt
214
neutrons are ejected by evaporation from the fission fragments, such as 87 36 Kr, instantaneously, i.e. within B10
212
to 10 35 Br ðdecay time55sÞ emit neutrons with a little delay
s of a fission reaction. Some fission fragments, such as 87
which can be from a fraction of a second to a minute as shown in Table 2.13 for thermal fission in uranium 235. These
delayed neutrons, though less than one percent of the total number of neutrons emitted in fission, play a beneficial role

FIGURE 2.26 Fission fragment yield for U233, U235 and Pu239.
 Interactions of neutrons with matter Chapter | 2 85

TABLE 2.13 Prompt and delayed neutrons for thermal fission in U235.

Group i Half-Life (s) Decay constant (λi , s) Energy (keV) Yield, neutrons per fission Fractions (β)
1 55.72 0.0124 250 0.00052 0.000215
2 22.72 0.0305 560 0.00346 0.001424
3 6.22 0.111 405 0.00310 0.001274
4 2.30 0.301 450 0.00624 0.002568
5 0.610 1.14 0.00182 0.000748
6 0.230 3.01 0.00066 0.000273
Total yield: 0.0158
Total Delayed fraction(β): 0.0065

Source: Lamarsh and Baratta, Introduction to Nuclear Engineering, p88. Based in part on G. R. Keepin, Physics of Nuclear Kinetics, Reading, Mass: Addison-
Wesley 1965.

in the control of nuclear reactors. The delayed neutron yield was measured by Keepin in 1957 (Keepin, Wimett, &
Zeigler, 1957) and measured by Krick and Evans (Evans, 1974) for fast fission of U-233, U-235, U-238, Pu-239, Pu-
240 and Th-232 averaged over 0.1-1.8 MeV. For U-235, the measured value by Krick and Evans is 0.0163 6 0.0013.
Fission is also accompanied with prompt gamma ray, within the first 69 ns of thermal neutron fission from U-235,
with an energy distribution nγ ðEÞ neutrons emitted per MeV per fission (Peele & Maienschein, 1970).
8
>
<
6:6 0:1 , E , 0:6MeV
nγ ðEÞ 5 20:2e21:78E 0:6 , E , 1:5MeV
>
: 7:2e21:09E 1:5 , E , 10:5MeV

where E is in MeV.

2.9.5 Fissile and fertile materials

Nuclides, such as uranium-233, uranium-235, and plutonium-239, which undergo fission directly are called fissile
nuclides while nuclides that need to be transmuted (change of the nucleus) to become fissile nuclides are called fertile
nuclides. As uranium resources are finite, and plutonium does not exist in nature, the potential role of fertile nuclides
has remained an important concern.
Two important “fertile” nuclides found in nature are thorium-232 and uranium-238, which can be induced to
undergo transmutation by two-stage nuclear reactions

90 Th 1 0 n-90 Th 1 γ
232 1 233

β2ð22minÞ 233 β2ð27dayÞ 233

233
90 Th 













 ! 91 Pa 













! 92 U

and
238
92 U 1 10 n-239
92 U 1 γ

β2ð23:5minÞ 239 β2ð2:3dayÞ 239

90 U 












! 93 Np 













! 94 Pu
239

The thermal and fast cross sections for some fissile and fertile nuclides are given in Tables 2.14 and 2.15, respec-
tively (Nuclear Energy Agency, 2015). These tables use a thermal spectrum for an averaged Pressurized Water Reactor
(PWR) and a fast spectrum for a reference fission spectrum. The data of Tables 2.14 and 2.15 for U-233, U-235 and
Pu-239 shown in Fig. 2.10 show very high fission cross sections for these fissile nuclides with U-233 and Pu-239
86 Nuclear Engineering

TABLE 2.14 Averaged cross sections for fissile/fertile nuclides (thermal spectrum).

Thermal spectrum Th232 U233 U235 U238 Pu238 Pu239 Pu240 Pu241
Neutron capture (n,g) (b) 5.010 11.270 14.380 14.898 37.052 83.656 479.200 51.600
Fission (n,f) (b) 0.009 79.710 53.310 0.040 2.767 142.644 0.370 139.300

Source: NEA, OECD, 2015.

TABLE 2.15 Averaged cross sections for fissile/fertile nuclides (fast spectrum).

Fast spectrum Th232 U233 U235 U238 Pu238 Pu239 Pu240 Pu241
Neutron capture (n,g) (b) 0.093 0.068 0.085 0.067 0.096 0.057 0.099 0.117
Fission (n,f) (b) 0.078 1.893 1.229 0.321 1.997 1.793 1.376 1.649

Source: NEA, OECD, 2015.

formed from Th-232 and U-238, respectively; in fact, both U-233 and Pu-239 have higher fission cross sections, than
for U-235, for both thermal and fast spectra.
The data of these tables is valid for a fresh core as it does not consider the changes in the spectra due to changes in
the fuel composition as the fuel is consumed.

2.9.6 The fission spectrum

Neutrons emerging from a fission reaction can have energies ranging from 0 to 10 MeV; a typical average value is
2 MeV. The number of neutrons emerging from fission with energies in the range E to E 1 dE is ν ðEÞχðEÞdE, where
χðEÞ is the fission spectrum. The following empirical expression, shown is the Watt fission spectrum

πa3 b
1=2 h i
eab=4 e2E=a sinh ðbEÞ2
1
χðEÞ 5 (2.47)
4

where the constants a, b are listed in Table 2.16 for thermal and fast fissions at 1 and 14 MeV in thorium-232,
uranium-233, uranium-235, and plutonium-239.
The shapes of the fission spectra for the nuclides is shown in Fig. 2.27.
Some other fits to the experimental fission energy spectrum are the Maxwell spectrum

2π pffiffiffiffi E
χðEÞdE 5 E exp 2 dE
ðπT Þ 3=2 T

(T 5 1.33 MeV for uranium) for which the average energy is: Eav 5 3=2TB2MeV, and the Cranberg spectrum

2exp ð2 AB=4Þ E pffiffiffiffiffiffi

χðEÞdE 5 pffiffiffiffiffiffiffiffiffiffiffi exp 2 sinh BE dE
πA3 B A

(A 5 0.965 MeV, B 5 2.29 MeV21 for uranium). For some other fission spectra used for modeling and simulation,
the reader is referred to the MCNP Manual (Werner, 2017).

Exercise 2.19: Watt Spectrum: Find the most probable and average energy of neutrons from the Watt fission spectrum.
TABLE 2.16 Constants a and b for the Watt fission spectrum.

Nuclide energy Th-232 U-233 U-235 Pu-239

th 1 MeV 14 MeV th 1 MeV 14 MeV th 1 MeV 14 MeV th 1 MeV 14 MeV

a MeV 1.0888 1.1096 1.17 0.977 0.977 1.0036 0.988 0.988 1.028 0.966 0.966 1.055
b MeV221 1.6871 1.6316 1.461 2.546 2.546 2.6377 2.249 2.249 2.084 2.842 2.842 2.383

Source: ENDF/B.
88 Nuclear Engineering

FIGURE 2.27 Watt fission energy spectrum.

2.10 Criticality
When fission takes place, the number of neutrons can rapidly multiply leading to the release of a large explosive
energy. It thus needs to be controlled so that the population remains steady as in the case of a nuclear power reactor.
Three such regions can be identified in terms of the system multiplication kðgÞ in the gth generation defined as the
neutron population in a generation divided by the neutron population in a previous generation
nðgÞ
kðgÞ 5 (2.48)
nðg21Þ
The regions are
8
< , 1 subcritical
kðgÞ 51 critical
:
. 1 supercritical
In a nuclear reactor, the objective is to produce steady electrical power so that the objective is to keep the system
critical. It is important to understand that nuclear criticality can be achieved at different power levels, for example, a
300 MW reactor, a 600 MW reactor and a 1200 MW reactor are all kept critical. When an experiment in a lab is being
carried out, it is usually subcritical for safety purposes while when a system is designed to release an explosive burst of
energy, such as a nuclear weapon, then it is designed to become supercritical.
Nuclear fission can result in the uncontrolled multiplication of neutrons as fission proceeds if neutrons are not cap-
tured or do not escape from the system. If we assume that a fission reaction is accompanied with the release of two neu-
trons, then N fissions will result in 2 N neutrons. Thus for N 5 10 fission reactions, there will be 210 5 1024 neutrons
and ten “generations” later, this number will have grown to 220 5 1,048,576, and subsequently by the 30th generation,
there will be 230 5 1.0737 109 neutrons. Such a runaway situation is called super-criticality as it is accompanied by the
uncontrolled release of energy as in a nuclear bomb or a nuclear accident.

Exercise 2.20: Generations in fission: Estimate the number of generations needed for the release of energy equivalent
to the energy from 1 ton of TNT explosive.

Neutrons ‘born’ as fission neutrons at energies of the order of 1 MeV, with energy distribution as modeled by the
Watt spectrum (Fig. 2.27) slow down by collisions, as they travel in a medium, as discussed in Section 2.7, until they
achieve equilibrium with their surroundings at ‘thermal’ energies. To understand criticality, we consider a simplified
picture shown in Fig. 2.28.
For S “fast” neutrons injected into a system, PF S neutrons leak out of the system boundaries; of the remaining
ð12PFÞ S, the fast fission multiplication increases the number to εð12PFÞ S. These fast neutrons, while colliding with
host nuclei, lose their energy gradually and reach the “resonance” region where εpð12PF ÞPT S escape the resonances,
 Interactions of neutrons with matter Chapter | 2 89

FIGURE 2.28 The processes contributing to the system

multiplication.

while εð12pÞð12PFÞ S are “captured,” and continue to slow down to thermal energies. At this “lower-energy,” some
neutrons are captured in moderator and structural material while εpf ð12PFÞ ð1 2 PT ÞS are absorbed in fissile material
and thus εpηf ð12PFÞ ð1 2 PT ÞS neutrons are produced by fission. In summary, the cycle began with S neutrons and
ended with εpηf ð12PFÞ ð1 2 PT ÞS neutrons. This picture gives us an estimate of the ‘multiplication’ of the system,
viewed as the number of neutrons emerging in the “next generation” divided by the corresponding number in the previ-
ous generation. This is called the effective multiplication of the system and is written from Eq. (2.48) as
εpηf ð1 2 PF Þð1 2 PT ÞS
keff 5 (2.49)
S
The above can be simplified for an “infinite” system from which there is no leakage. In that case, keff 5 kN PNL ,
where kN is the infinite multiplication factor
kN 5 εpηf (2.50)
and PNL is the nonleakage probability due to the finite size of the system expressed as the product nonleakage probabili-
ties PNL 5 ð1 2 PF Þð1 2 PT Þ. For a finite system, the effective multiplication factor is
keff 5 kN PNL;F PNL;T (2.51)
From the cycle in Fig. 2.28, the quantities are defined as follows:
1. epsilon ε, the fast fission factor is a measure of the number of fissions caused by “fast” neutrons; thus we would
need to specify the cutoff energy Eth above which neutrons are classified as fast neutrons. For neutrons in the energy
range (0; E0 ) where E0 is the maximum energy of neutrons, with the fast group (Eth ; E0 ) and the thermal group
(0, Eth ), ε is defined as
Ð ÐE Ð ÐE
no:of fission reactions dV 0 0 Σf ðr; RÞφðr; EÞdE dV Eth0 Σf ðr; RÞφðr; EÞdE
ε5 5 Ð ÐE 511 Ð ÐE
no:of fission reactions by thermal neutrons dV 0 th Σf ðr; RÞφðr; EÞdE dV 0 th Σf ðr; RÞφðr; EÞ dE
(2.52)
2. the resonance escape probability p [Section 2.8, Eq. (2.46)] is defined as the probability of escaping resonance cap-
ture while slowing down from fast to thermal energies. In terms of the absorbing (fuel) atoms and the scattering
(moderator) atoms, and the effective resonance integral I,
 
NA I
p 5 exp 2
ξM ΣsM 1 ξA ΣpA
3. eta η is defined as
Ð
no:of fission neutrons produced υΣf φth dV υΣf
η5 5 Ð 5 (2.53)
no of thermal neutrons aborbed in fuel ΣaF φth dV ΣaF
90 Nuclear Engineering

TABLE 2.17 One-Group cross sections for Uranium-235.

υσf
U-235 υ σf (b) σc (b) σs (b) σt (b) σa

(A) 2.50 1.3  4.0 5.3 2.25

(B) υσf 5 5:297 σa 5 2:844 σtr 5 8:246  1.8625

Source: (A) Bell and Glasstone, Nuclear Reactor Theory: p.126; (B) Reactor Physics constants ANL-5800.

4. and the thermal utilization f as

ÐV
thermal neutrons absorbed in fuel 0 ΣaF φth dV
f5 5 ÐV (2.54)
0 ðΣaF 1 Σa;other Þφth dV
total no of thermal neutrons absorbed

where Σa;other 5 Σa;M 1 Σa;structural materials 1 . . . is the thermal macroscopic thermal cross-section for absorption in fuel,
moderator with absorption cross-section Σa;M and structural material etc. If absorption is considered to take place only
in fuel and moderator, then for a homogeneous mixture of fuel and moderator (so that they occupy the same volume)
ΣaF
f5 (2.55)
ΣaF 1 ΣaM
In a most elementary analysis, the criticality for an ‘infinite’ system thus depends on four factors viz ε; p; η and f.

Example 2.2: Six-factor formula: One-group criticality of a bare sphere of uranium-235 using the six-factor formula,
one-group diffusion theory, one-speed neutron transport and Monte Carlo simulation.

For a one-group calculation, the data of Table 2.17 is used.

2.10.1 Diffusion theory

Based on the four-factor formula [Eq. (2.50)], kN can be calculated from the data. A sphere of U-235 will have a fast
spectrum since there is no light-weight material; the fast fission the neutron spectrum will be fast. The fission factor,
resonance escape and thermal fuel utilization,εB1; pB1; f B1, so that kN 5 η 5 2:25. This says that an infinitely large
system will have no leakage and its multiplication will be very high; in fact supercritical. The nonleakage probabilities
for fast and thermal neutrons will be derived in Chapter 4 (diffusion theory) for one- and two-group equations. For
now we accept the results to demonstrate calculation of keff ; the critical equation. The fast and thermal nonleakage
probabilities are

1
PNL;F 5 (2.56)
1 1 B2 τ
and

1
PNL;T 5 (2.57)
1 1 L2 B2
respectively, so that

kN
keff 5 1 5 (2.58)
ð1 1 B2 τ Þð1 1 L2 B2 Þ
in terms of the buckling B, that is, the curvature of the neutron flux, given by
2
kN 2 1 π
B2m 5 5 B2g 5 (2.59)
L2 Rc
 Interactions of neutrons with matter Chapter | 2 91

for a sphere with critical radius Rc . Criticality requires that the material buckling Bm and geometrical buckling Bm are
equal. The critical radius is then found as
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
L2
Rc 5 π (2.60)
kN 2 1
In Eq. (2.55), the diffusion length is interpreted as the fast diffusion length Lf since there is no thermalization of
neutrons.

D 1
L2 5 5
Σa 3Σtr Σa
The results are: D 5 0:8413 cm; L2 5 6:1568cm2 ; critical radius Rc 5 8:3935 cm, and the critical mass
Mc 5 43 πRc 3 5 46:443 kg. This is to be compared with the critical radius and mass of a pure U-235 bare sphere (critical
radius 8.46 cm and critical mass 48 kg) given by de Volpi (DeVolpi, 1982) listed in Table 2.20.
Diffusion theory is an elementary model and useful for obtaining approximate solutions; a better formalism is pro-
vided by transport theory which is covered in some detail in Chapter 6. Here, we use some results from the one-speed
transport equation which has a diffusion approximation little better than the one-group diffusion model.

2.10.2 Transport theory

The one-speed transport equation (Bell & Glasstone, 1979) reduced from the continuous energy neutron transport
equation has been solved analytically for several idealized geometries and scattering models. Exact form solutions for
criticality in slabs and spheres (Bell & Glasstone, 1979; Kaper, Lindeman, & Leaf, 1974; Öztürk, 2012; Rawat &
Mohankumar, 2011) have been obtained which are both elegant and useful but of course need validation with elaborate
computer simulations.
The transport theory asymptotic flux in a sphere of radius R is
A r
φas ðr Þ 5 sin (2.61)
r jν 0 j
where ν 0 is a solution of the transcendental equation

1 cν 0 ν 0 1 1
1 5 cν 0 tanh21 5 ln (2.62)
ν0 2 ν0 2 1
For the mean number of secondaries c, that is, the number of neutrons emerging from an interaction

υσf 1 σs 1 σn;2n 1 ?
c5 (2.63)
σt
B
B 
The sphere is approximately critical when the asymptotic flux is zero at the extrapolated radius R ; thus φas R 5 0
B
at R 5 Rc 1 x0 5 πjν 0 ðcÞj.
The solutions of Eq. (2.62) are real for c , 1 and complex for c . 1; an approximation for ν 0 is
 
1 3ð1 2 cÞ 912c
5 12 2? (2.64)
ν02 c 5 c

To find the critical radius ν 0 is determined form the transcendental equation [Eq. (2.62)]
Rc 5 πjν 0 j 2 x0 (2.65)
and the extrapolation distance x0 can be calculated from the Mark P1 boundary condition
 
1 1 1
x0 5 pffiffiffi 1 2 ðc 2 1Þ 1 ðc21Þ 1 ?2
(2.66)
3 3 5

From the data of Table 2.17,

92 Nuclear Engineering

In the spherical harmonics P1 “diffusion approximation,” the neutron flux is

rffiffiffiffiffiffiffiffiffiffiffi
1 3
φðxÞ 5 e2 jxj=L (2.67)
2 12c
where the diffusion length is
1
L 5 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi (2.68)
3ð1 2 cÞ
Criticality verification benchmark sets for one-group U-235 and Pu-239 bare spheres are listed in Table 2.17 (Sood,
Arthur Forster, & Kent Parsons, 2003); these give the value k 5 1 for the critical radii Rc listed.

2.10.3 Monte Carlo simulation

For criticality calculations, as for most other nuclear design calculations, Monte Carlo (MC) simulation is one of the
most elaborate methodologies due to its capability of handling realistic 3D geometries and interaction models. Details
of the Monte Carlo method will be covered in Chapter 8; here only the results are presented and discussed in the con-
text of this example.
The objectives of the MC simulation are i- to estimate keff , ii- to obtain the values of ε; p; η and f in the four-factor
formula, and iii- the leakage of neutrons.
The composition of the Godiva sphere is: (wt.%) U-235 93.71, U-238 5.27, U-234 1.02 with a density 18.74 g/cm3
and radius 8.7407 cm. The Godiva assembly was one of the first critical experiments for the experimental validation of
calculations as well as for kinetics experiments; it has continued since the 1950s with the present Godiva IV assembly
for research on nuclear as well as high energy lasers (Thompson et al., 2020). Several hand calculation methods have
also been developed for critical assemblies as well as for reactor configurations (Bowen & Busch, 2005; Caplin, Duluc,
& Richet, 2015); these are usually based on one- and two-group diffusion theory as well as interacting array models
and validating with codes such as MCNP (Brown, 2009; Harmon, Busch, Briesmeister, & Forster, 1994).
In MC simulation, you only get the answer you ask for; in deterministic codes, the solution i.e. neutron flux and
reactions rates are given in the problem domain specified in the input files. Thus the tallies and group boundaries need
to be specified. For this, we need to have a basic understanding of the physics and engineering of the problem. That
was the purpose of the preceding sections of this chapter.
So, now we take a look at Fig. 2.29, in the context of the underlying theory covered in Section 2.8 and
Figs. 2.192.21, to ‘see’ the resonance structures of both U-235 and U-238. The impressions we get are: i- both iso-
topes have similar resonance structures, ii- there are three distinct energy regions, iii- resonances are spread over the
energy range, few eV1 keV for U-235 and few eV to B 2 keV for U-238, and iv- U-238 has higher resonance cross-
section values at lower energies. Fig. 2.29 shows resolved resonances at lower energies and unresolved resonances at
higher energies.
In order to select bin boundaries for energy groups, another detailed picture of the resolved resonances is required;
for this we take a look at Figs. 2.30 and 2.31.
From Fig. 2.31, we are in a position to select bin boundaries to tally absorptions in the low-energy resonance regions.
While Fig. 2.29 suggests group boundaries: 0 2 1eV, 1eV 2 2keV, 2keV 2 14MeV, Fig. 2.31 suggests “finer”
boundaries 1; 1:6; 2:5; 4; 5:2; 5:7; 6:6; 7; 7:5 and 10 eV and 2keV; 1MeV, and 14MeV. This process tells that MC simula-
tion is both an art and a science.
The simulation was carried out for 5000 neutrons per batch, 200 batches, and 20 “skip” cycles; all these terms will
be covered in detail in this book. The interactions of one million neutrons in the uranium sphere were simulated with a
computational effort of 1.95 min on an Intel (R) Core i7-2620M CPU @ 2.70 GHz processor with a 32-bit Operating
System and 8.00 GB RAM.
The energy-dependent neutron flux obtained from simulation is shown in Fig. 2.32. It appears to be a “hard” spec-
trum, essentially all above 1 keV which is understandable as there is no light material present in Godiva.
From Fig. 2.32, it can be expected that resonances will play no part in the neutron population as neutrons will most
probably not reach the resonance energies during their slowing down process. The energy group boundaries for neutron
flux: 300 equal-lethargy bins from 0-10 MeV. This would not have been expected in the presence of water as in a Light
Water Reactor.
 Interactions of neutrons with matter Chapter | 2 93

FIGURE 2.29 Radiative capture cross sections of U-235 and U-238.

The average flux for one source neutron is φ 5 2.4213 3 1023 (0.0006) neutrons per cm2 per second. The tallies,
their bin structures and results from simulation are given in Table 2.18.
Reactions tallied: Σt φV, υΣf φV, Σf φV, Σðn;2nÞ φV, Σðn;3nÞ φV, Σðn;fxÞ φV, Σðn;f Þ φV in energy groups 0 2 1eV,
1 eV 2 2 keV, 2 keV 2 14 MeV;
Results showed no radiative captures Σðn;γÞ φV in groups with bin boundaries at 1, 1.6 2.5, 4, 5.2 5.7 6.6, 7, 7.5 and
10 eV since there were no neutrons at these energies. The radiative captures in the range 10 eV2 keV were found to
be 3.58282 3 1025 (0.1048) and in the range 1 MeV14 MeV there were 8.68463 3 1023 (0.0014); the total captures
being 4.48660 1023 (0.0014). Thus in the four-factor formula, it is justifiable to set ε 5 p 5 1. Similarly there is no
moderator so that fuel utilization f 5 1 and keff 5 ηPNL;f . The system multiplication based on the average of three esti-
mators is keff 5 0:99739. From the tallies, ηB2:3319 and PNL;f B0:4263 giving keff B0:9941:

Exercise 2.21: Godiva: From the results of the above Monte Carlo simulation, write a conservation equation balancing
the total number of interactions with all other interactions. What information can you obtain from this balance
equation?

Several benchmarks are given for criticality validation calculation (Sood et al., 2003) with one-group data for U-235
and Pu-239 given in Table 2.19. Notice that these values give a critical radius for a U-235 sphere which is different
from that of Godiva (Table 2.22). The same is true for Pu-239 from the values in Tables 2.21 and 2.22.
94 Nuclear Engineering

FIGURE 2.30 Low-energy range radiative capture cross sections of U-235 and U-238.

It is important to understand that criticality depends on a number of factors such as the enrichment of fissile materi-
als, their geometrical shapes, and the environment which they are placed in. The first step in a nuclear system such as a
nuclear reactor is to determine its critical composition and size. The data that must be used for such calculations is
strongly dependent on the neutron spectrum. Only a basic understanding of the system can be obtained from generalized
data such as that shown in Table 2.20, for example, the difference in the fission cross-section for three situations viz
thermal energy σ(fs,th), a reactor spectrum σ(fs,rs) and a fast spectrum σ(fs,f). Some basic questions that can be
answered by looking at the data in Table 2.20 are:
1. Would it more hazardous to sit in the vicinity of a U-235 or a U-238 sphere in terms of neutron emissions from
spontaneous fission?
2. Which isotope of plutonium would be most suitable for a thermoelectric generator?
3. What would be the effect of water placed in the vicinity of a Godiva sphere?
In the early days of nuclear technology, the most important challenge was to determine the critical masses of fissile
nuclides and experimentally verified for bare and reflected spheres of uranium and plutonium listed in Table 2.21. It is
important to note that criticality can be achieved by significant reductions of fissile material when the core is sur-
rounded with a “reflector” such as U-238 though with a penalty of increased mass.
In Table 2.21, the critical mass for U-235 is 48 kg for a pure U-235 sphere; this is different from the Godiva bare
sphere (Rowlands et al., 1999) which has a weight composition given in Table 2.22.
A study of neutronic systems can be carried out with the use of diffusion theory, transport theory or Monte Carlo
simulation which are based on group averaged cross-section data and spectrum calculations described in this chapter.
 Interactions of neutrons with matter Chapter | 2 95

FIGURE 2.31 Low-energy resolved resonances of U-235 and U-238.

FIGURE 2.32 Energy-dependent neutron flux in Godiva.

TABLE 2.18 Monte Carlo simulation results for Godiva.

Tally Quantity estimated Reaction MT number Energy Bins

0 2 1eV 1eV 2 2keV 2keV 2 14MeV Total

Neutron current J    0.5737 (0.0006)
Total interactions Σt φV 1  2.3219 3 1024 (0.1002) 2.6531 (0.0007)a 2.6534 (0.0007)
Fission reactions Σf φV 26  7.1939 3 1025 (0.1023) 3.8385 3 1021 (0.0006) 3.8392 3 1021 (0.0006)
Neutrons produced from fission υΣf φV 26 7  1.7509 3 1024 (0.1023) 9.9697 3 1021 (0.0006) 9.9715 3 1021 (0.0006)
Multiplication (n,2n) Σðn;2nÞ φV 16    2.6531 3 1023 (0.0076)
Multiplication (n,2n) Σðn;3nÞ φV 17    4.5917 3 1026 (0.1428)
Total fissionb Σðn;fxÞ φV 18    3.7727 3 1021 (0.0006)
fission Σðn;f Þ φV 19    6.5429 3 1023 (0.0010)

relative standard error; read as 2.6531(1 6 0.0007).

a
b
sum of (n,f),(n,n’f), (n,2nf),(n,3nf).
 Interactions of neutrons with matter Chapter | 2 97

TABLE 2.19 One-Group isotropic cross sections and critical radius of bare spheres.

Material υ Σf (cm) Σc (cm) Σs (cm) Σt (cm) υΣf

(cm) Rc (cm)
Σa

U-235 (a) 2.70 0.065280 0.013056 0.248064 0.32640 2.25 7.428998

Pu-239 (a) 2.84 0.081600 0.019584 0.225216 0.32640 2.612903 6.082547

Source: Sood, A., Forster, A., and Parsons, D. K. Analytical Benchmark Test Set for Criticality Code Verification, Progress in Nuclear Energy, Vol. 42, No. 1,
pp. 55-106, 2003.

TABLE 2.20 Some useful actinide physical properties.

Nuclide α α ν ν σ σ σ σ σ τ N Q
(th) (f) (th) (f) (fs, th) (fs, rs) (fs, f) (c, th) (c, rs) (total)
b/b b/b n/fs n/fs b b b b b y n/s-kg W/kg
Th-232 2.10 0 0.08 7 1.4E10 0.1
U-233 0.09 2.45 2.61 528 1.90 50 1.6E5
U-234 2.60 0 3.9 1.50 281 2.5E5 1.7
U-235 0.17 0.27 2.40 2.56 580 409 1.30 100 105 7.1E8 0.2
U-236 2.50 3.2 0.80 6 74 2.4E7 1.7
U-236 2.30 3.7E7 2E4
Np-237 7 2.85 0 2.5 1.4 170 359 2.2E6 B1
U-238 2.50 0 0.6 0.3 3 8 4.9E9 17
Pu-238 (3) 19 2.5 500 366 8.8E1 2.6E6 567
Pu-239 0.36 0.22 2.86 3.08 742 1060 1.8 243 433 2.4E4 30 1.9
Pu-240 1.6 2.8 3.10 0 2.5 1.4 300 897 6.6E3 1E6 7.1
Pu-241 0.38 0.13 2.96 3.0 1010 1110 1.8 420 334 1.4E1 3.4

th 5 thermal, f 5 fast, rs 5 reactor spectrum, fs 5 fission, c 5 capture.

α 5 capture-to-fission ratio, v 5 neutron yield, σ 5 cross-section (in barns).
τ 5 half-life, Q 5 specific power (heat), N 5 spontaneous fission neutron rate.
E 5 exponent (to power of ten).
Neutron emission rate from (α, n) reaction in 238Pu oxide is 1.4E7 n/s-kg.
Some infinity-dilute resonance integers (0.6eV-10 MeV):
α(239Pu) 5 0.6; σ(c) (240Pu) 5 8500 b; σ(c) (242Pu) 5 1120b.
Source: DeVolpi, A., Denaturing Fissile Materials, Progress in Nuclear Energy, Vol. 10, No. 2, pp. 161-220: from Table 2.

TABLE 2.21 Critical solid spherical systems: bare and reflected with 10 cm U.

Material Density (g/cc) Configuration Core Ref. Total mass (kg)

Radius (cm) Core mass (kg) Ref. mass (kg)

233
U 18.9 Bare 5.87 16  16
Ref. 4.2 5.7 221 227
U235 18.9 Bare 8.46 48  48
Ref. 5.8 15.7 300 316
Pu239 α 15.9 Bare 5.49 11  11
Ref. 3.8 4.5 204 209
Pu239 δ 19.5 Bare 5.68 15  15
Ref. 4.7 7.0 243 250

Source: DeVolpi, A., Denaturing Fissile Materials, Progress in Nuclear Energy, Vol. 10, No. 2, pp. 161-220: Table 3.
98 Nuclear Engineering

TABLE 2.22 Bare critical assemblies.

Model Godiva Jezebel Jezebel23

Radius (cm) 8.7407 6.3849 5.9838
Density (g/cm3) 18.74 15.61 18.424
Composition (atoms/barn-cm) U235 4.4994e-2 Pu239 3.7047e-2 U233 4.6712e-2
U238 2.4984e-3 Pu240 1.7512e-3 U234 5.9026e-4
U234 4.9184e-4 Pu241 1.1674e-4 U238 2.8561e-4
Ga69 8.26605e-4 U235 1.4281e-5
Ga71 5.48595e-4
Mass (g) 52419.98 17019.77 16534.98

Problems
2.1. Describe the Boltzmann equation for the kinetic theory of gases and the significance of its moments.
2.2. From the size of aU-238 nucleus, calculate the potential scattering cross-section and compare with the
figures for U-238 cross sections in this chapter.
2.3. How would you find the average lethargy gain for a mixture consisting of four elements with given atomic num-
ber densities Ni ; i 5 1; 2; 3; 4?
2.4. Write a formula for the intensity reduction of a spherical multilayered shield of neutrons.
2.5. What is the significance of the cadmium ratio in the measurement of neutron flux?
2.6. How can the neutron age used in a “modified” one-group formula to compute neutron leakage in a finite
system?
2.7. On an energy-flux figure, for a narrow resonance sketch the flux and the absorption cross-section and describe
the phenomenon known as self-shielding. Show that the lethargy dependent flux decreases inversely with the
total cross-section.
2.8. For the first resonance of U-238 at 6.67 eV calculate the total cross-section given in Eq. (2.39) ignoring the inter-
ference term.
2.9. Write a formula for estimating the critical energy of fission in Pu-239.
2.10. Calculate the value of c for the Godiva one-speed data given in Table 2.17 and solve the transcendental equation
[Eq. (2.62)].
2.11. Calculate the extrapolation distance x0 from the Mark boundary condition [Eq. (2.63)] and estimate the critical
radius for Godiva.
2.12. Given a homogenous mixture of a fissile material and water, outline the steps for calculating the critical dimen-
sions of slab system. The given data is i-concentration of the fissile material cF grams per liter, and ii- the non-
B
1/v gi , factors and cross sections σi for both materials, i 5 abs; f . For a slab system, the buckling is B 5 π= a ,
B B
where the physical slab thickness is a 5 a 2 2d, a is the extrapolated thickness and d 5 0:71λtr . Justify any
other assumptions you make.
2.13. In the above problem (criticality for a homogeneous system) consider the effect of slowing down (Section 2.8)
including both neutron age τ and the diffusion length L to calculate the nonleakage probabilities in the six-factor
formula.
2.14. In the above problem (criticality for a homogeneous system), outline the procedure if you are required to find
the concentration of the fissile material to make a slab of given dimensions critical.

Nomenclature
English lower case
c mean number of secondary neutrons per collision
f thermal utilization factor
keff effective multiplication
kN infinite system multiplication
m mass of neutron
n(E) number of molecules with energy E per unit interval dE
 Interactions of neutrons with matter Chapter | 2 99

n0 total number of molecules

tm moderating time
vc velocity of neutron in center of mass system
Vc velocity of nucleus in center of mass system
vl velocity of neutron in laboratory system
v M average Maxwellian neutron velocity
vp most probable velocity
vr relative speed of neutron and nucleus in the laboratory system
x0 extrapolation distance

English upper case

A relative mass of nucleus (relative to neutron mass)
B buckling
D diffusion coefficient
E0 resonance center of mass energy
EC critical energy
Ec kinetic energy of neutron and nucleus center of mass energy
EAL kinetic energy of target nucleus before collision (Lab system)
0
EAL kinetic energy of target nucleus after collision (Lab system)
EL kinetic energy of neutron before collision (Lab system)
EL0 kinetic energy of neutron after collision (Lab system)
Ep most probable energy
F collision density
I effective resonance integral
J neutron current
L diffusion length
P potential scattering
P resonance escape probability
PF,T leakage probability (F 5 fast, T 5 thermal)
PNL,F,T nonleakage probability (F 5 fast, T 5 thermal)
Rx reaction rate type x
Rc critical radius
S source
T temperature
V volume
VCM velocity of center of mass
Vl velocity of nucleus in the laboratory system
J spin

Greek lower case

μ reduced mass m1 m2 =ðm1 1 m2 Þ
α capture-to-fission ratio

A212
α A11
ε
η number of fission neutrons emerging per thermal neutron absorbed
ξ average gain in lethargy per collision
λ mean free path
λ~ 1 reduced wavelength
ν0 transport relaxation length
ν neutrons emerging from fission reaction
χ fission distribution function
θL scattering angle in the lab system
θCM scattering angle in the CM system
τ neutron age
100 Nuclear Engineering

τ half-life
τ lifetime of compound nucleus
φas asymptotic flux
φ average flux
ϕ azimuthal angle

Greek upper case

Γ measure of the width of a resonance (energy)
Γi =Γ probability of ith particle emission from a resonance
Γγ =Γ probability of gamma emission from a resonance
Σa macroscopic transport cross-section
Σf macroscopic fission cross-section
Σ(n,γ) macroscopic radiative capture cross-section
Σtr macroscopic transport cross-section
Σ(n,2n) macroscopic ðn; 2nÞcross-section
Σ(n,3n) macroscopic ðn; 3nÞcross-section
Σ(n, fx) sum of ðn; f Þ; ðn; n0 f Þ; ðn; 2nf Þ; ðn; 3nf Þ cross sections
Σ(n, f) macroscopic fission cross-section
Φ probability
Ψ wave function

Abbreviations
L Laboratory system
CM center of mass system
CN compound nucleus
MT material reaction number

References
Aziz, U., Koreshi, Z. U., Sheikh, S. R., & Khan, H. (2020). A study of criticality and thermal loading in a conceptual micronuclear heat pipe reactor
for space applications. Nuclear Technology and Radiation Protection, 35. Available from https://doi.org/10.2298/NTRP2003208A.
Bell, G. I., & Glasstone, S. (1979). Nuclear reactor theory. New York: Robert E. Krieger Publishing Co., Inc.
Bowen, D G. & Busch, R D. (2005). Hand calculational methods for criticality safety - A Primer. In Proceedings of the NCSD topical meeting
(American Nuclear Society Nuclear Criticality Safety Division).
Brown, D. A., Chadwick, M. B., Capote, R., Kahler, A. C., Trkov, A., Herman, M. W., Sonzogni, A. A., et al. (2018). ENDF/B-VIII.0: The 8th major
release of the nuclear reaction data library with CIELO-project cross sections, new standards and thermal scattering data. Nuclear Data Sheets.
Available from https://doi.org/10.1016/j.nds.2018.02.001.
Brown, F B. (2009). A review of best practices for Monte Carlo criticality calculations. In Proceedings of the American Nuclear Society - nuclear crit-
icality safety division topical meeting on realism, robustness and the nuclear renaissance.
Caplin, G., Duluc, M., & Richet, Y. (2015). On the benefit of using both hand calculation methods and computer experiments algorithms to solve
complex nuclear criticality safety problems. In Proceedings of the international conference on nuclear criticality safety.
Deen, J. R., & Woodruff, W. L. (1995). WIMS-D4M user manual Rev.0. Illinois: Champaign-Urbana.
DeVolpi, A. (1982). Denaturing fissile materials. Progress in Nuclear Energy. Available from https://doi.org/10.1016/0149-1970(82)90022-1.
Engle, W. W. Jr. (1967). A users manual for ANISN: A one dimensional discrete ordinates transport code with anisotropic scattering. Tennessee.
Evans, A. E. (1974). Delayed neutrons from fission: Present status of measurements of yields, group half-lives and abundances and spectra. Los
Alamos, New Mexico, USA. ,https://inis.iaea.org/collection/NCLCollectionStore/_Public/06/197/6197971.pdf..
Feather, N. (1939). The time involved in the process of nuclear fission [1]. Nature. Available from https://doi.org/10.1038/143597a0.
Hanson, S K., Pollington, A, D., Waidmann, C. R., Kinman, W. S., Wende, A. M., Miller, J. L., Selby, H. D. et al, (2016). Measurements of extinct
fission products in nuclear bomb debris: Determination of the yield of the trinity nuclear test 70 y later. In Proceedings of the National Academy
of Sciences of the United States of America. Available from https://doi.org/10.1073/pnas.1602792113.
Harmon, C. D., II, Busch, R. D., Briesmeister, J. F., & Forster, R. A. (1994). Criticality calculations with MCNP: A primer.
Hodgson, P. E. (1994). The nucleon optical model. ,https://doi.org/10.1142/9789812815538_0003..
Hyvönen-Dabek, M., & Nikkinen-Vilkki, P. (1980). Determination of neutron generator fast neutron spectra using threshold detectors. Nuclear
Instruments and Methods. Available from https://doi.org/10.1016/0029-554X(80)90824-1.
IAEA. 1970. Neutron fluence measurements, Technical Reports Series No. 107. Vienna, Austria.
Kaper, H. G., Lindeman, A. J., & Leaf, G. K. (1974). Benchmark values for the slab and sphere criticality problem in one-group neutron transport the-
ory. Nuclear Science and Engineering. Available from https://doi.org/10.13182/nse54-94.
 Interactions of neutrons with matter Chapter | 2 101

Keepin, G. R., Wimett, T. F., & Zeigler, R. K. (1957). Delayed neutrons from fissionable isotopes of uranium, plutonium, and thorium. Physical
Review. Available from https://doi.org/10.1103/PhysRev.107.1044.
Lamarsh, J. R. (1966). Introduction to nuclear reactor theory. Addison-Wesley Publishing Company.
Macfarlane, R. E., Muir, D. W., Boicourt, R. M., & Kahler, A. C. (2012). The NJOY nuclear data processing system, Version 2012. LA-UR-12-
27079, Los Alamos National Laboratory.
Madland, D. G., & England, T. (1977). Influence of isomeric states on independent fission product yields. Nuclear Science and Engineering.
Available from https://doi.org/10.13182/NSE77-A14501.
Nuclear Energy Agency, OECD. 2015. Introduction of thorium in the nuclear fuel cycle short- to long-term considerations.
Okumura, S., Kawano, T., Jaffke, P., Talou, P., & Chiba, S. (2018). 235U(n, f) independent fission product yield and isomeric ratio calculated with the
statistical HauserFeshbach Theory. Journal of Nuclear Science and Technology. Available from https://doi.org/10.1080/00223131.2018.1467288.
Öztürk, H. (2012). The criticality calculations for one-speed neutrons in a reflected slab with anisotropic scattering using the modified UN Method.
Kerntechnik. Available from https://doi.org/10.3139/124.110272.
Peele, R. W., & Maienschein, F. C. (1970). The absolute spectrum of photons emitted in coincidence with thermal-neutron fission of Uranium-235.
Oak RIdge, TN.
Pelowitz, D. B., Goorley, J. T., James, M. R., Booth, T. E., Brown, F. B., Bull, J. S., Cox, L. J., et al. (2013). MCNP6 User’s Manual. La-Cp-13-00634.
Pritychenko, B., & Mughabghab, S. F. (2012). Neutron thermal cross sections, Westcott factors, resonance integrals, Maxwellian averaged cross sec-
tions and astrophysical reaction rates calculated from the ENDF/B-VII.1, JEFF-3.1.2, JENDL-4.0, ROSFOND-2010, CENDL-3.1 and EAF-2010
evaluated data Librari. Nuclear Data Sheets. Available from https://doi.org/10.1016/j.nds.2012.11.007.
Rawat, A., & Mohankumar, N. (2011). Benchmark results for the critical slab and sphere problem in one-speed neutron transport theory. Annals of
Nuclear Energy. Available from https://doi.org/10.1016/j.anucene.2011.02.011.
Reuss, P. (2008). Neutron physics. Cedex, France: EDP Sciences.
Rowlands, J, Bosq, J. C., Brinckman, J. J., Fort, E., Giffard, F., Jacqmin, R., Rimpaul, G., et al. (1999). Intercomparison of calculations for Godiva
and Jezebel.
Shafronovskaia, A. I., Zamyatin, N. I., & Cheremukhin, A. E. (2014). Method of measuring fast neutron fluence using the planar silicon detectors. In
Proceedings of the twenty-fourth Russian particle accelerator conference, RuPAC 2014.
Sood, A., Arthur Forster, R., & Kent Parsons, D. (2003). Analytical benchmark test set for criticality code verification. Progress in Nuclear Energy.
Available from https://doi.org/10.1016/S0149-1970(02)00098-7.
Thompson, N., Hutchinson, J., Bahran, R., Hayes, D., Myers, W., Arthur, J., Bounds, J., et al. (2020). National Criticality Experiments Research Center
(NCERC) - Capabilities and recent measurements. EPJ Web of Conferences. Available from https://doi.org/10.1051/epjconf/202023918003.
Trkov, A., Herman, M., & Brown, D. A. (2018). ENDF-6 formats manual, data formats and procedures for the evaluated nuclear data files ENDF/B-
VI, ENDF/B-VII and ENDF/B-VIII. Upton, NY 11973-5000. ,http://www.nndc.bnl.gov..
Werner, C. J. (ed.). (2017). MCNP (R) User’s manual code Version 6.2.
Zamyatin, N. I., Cheremukhin, A. E., & Shafronovskaya, A. I. (2017). Measuring fluence of fast neutrons with planar silicon detectors. Physics of
Particles and Nuclei Letters. Available from https://doi.org/10.1134/S1547477117050156.
Chapter 3

Nuclear reactors and systems

A nuclear reactor is based on the utilization of energy from a nuclear reaction such as fission or fusion. In a nuclear
power reactor, the thermal energy from nuclear reactions is converted into mechanical energy, which can subsequently
be converted into electrical energy which we know as electricity that powers our homes, cities, and industries. Energy
is what makes our 21 century world so different from earlier centuries. It is in our interest to sustain the production of
energy and develop better, cleaner, and more sustainable forms of energy that will make it possible for human beings to
live high-quality lives.
There has been renewed worldwide interest in nuclear energy due to its success, resilience and innovations leading
to competitive high-density power systems compatible with sustainable development goals.
Nuclear reactors have been operating since the first power reactor went critical on December 1, 1942 producing
only 60 W just enough to light a bulb. Since then, the nuclear industry has undergone great advancements; the first gen-
eration reactors have mostly retired and advanced third generation reactors are being commissioned and several are con-
nected to the grid. The contribution to global electricity now exceeds 11% while the nuclear industry faces competition
from renewable solar and wind energy technologies.
In the last 80 years of nuclear power production, the nuclear industry has seen a very high safety record with three
exceptions namely Three Mile Island (USA, 1979) Chernobyl (Ukraine, 1986) and Fukushima (Japan, 2011). In the
Three Mile Island (TMI) accident, partial core melt resulted due to a loss of primary coolant; the reactor shut down
within a second, there was no injury or loss of life by significant radioactive release into the environment. Chernobyl
was a major accident due to both design and operational failures; the radiation release was significant and the accident
caused a number of deaths and thyroid cancers. The Fukushima accident was also a major accident resulting from a
large earthquake followed by a Tsunami causing damage to the cores of reactors and large-scale evacuations. With the
knowledge that followed from detailed analyses, nuclear reactor design and practices have improved to reduce the prob-
ability of such occurrences significantly.
This chapter presents a review of the current status of nuclear reactors leading to the next generation, Gen-IV,
designs aimed for the coming decades, followed by a birds’-eye view of the modeling and simulation requirements of
the nuclear industry. The technology of nuclear reactors is based on particles and radiation (electrons, protons, neutrons,
X-rays and gamma-rays) and nuclear reactions discussed in the preceding chapters. Here, the overall systems will be
discussed and, in their context, engineering requirements and challenges to be addressed by scientists, engineers and
designers will be highlighted.

3.1 Status of nuclear power

Nuclear power contributes to over 11% of the world’s electricity with generation units established in the 1950s. This
section reviews the generations of nuclear power reactors from the early 1950s to the next generation reactors.

3.1.1 Generations of nuclear power

Nuclear reactors are classified in several ways and can now, with hindsight, be categorized in generations with the first
generation beginning in the early 1950s to the current beginning of what will soon be called the fourth generation.
According to the neutron spectrum, coolant and moderator a reactor is classified as a thermal reactor for a thermal
(low energy) spectrum and conversely as a fast reactor when it has a fast (relatively high energy) spectrum, with little
neutron moderation. Going by the coolant and moderator, a water cooled and moderated reactor is called a light water
reactor (LWR), which can be subclassified as a pressurized water reactor (PWR) or a boiling water reactor (BWR), a
Nuclear Engineering. DOI: https://doi.org/10.1016/B978-0-323-90618-0.00003-X
© 2022 Elsevier Inc. All rights reserved. 103
104 Nuclear Engineering

heavy water (D2O) cooled and moderated reactor is called a pressurized heavy water reactor (PHWR), when gas is used
as coolant, it is called a gas cooled reactor (GCR).
Reactors are also classified according to their capacity as shown in Table 3.1.
In earlier generations, several variants were tested and developed with materials and moderators such as beryllium
and beryllium oxide to verify their compatibility with water, sodium and lead-bismuth alloys. The supercritical water
reactor, operating beyond the critical point of water, was tested at high temperatures to enhance the thermodynamic
efficiency. Organic coolants were tested for their moderating properties and corrosion reduction. The behavior of mate-
rials under intense radiation was a deciding factor in their selection. The latest developments in reactor design,
Generation-IV systems, aim for compact architecture, the use of liquid coolants (instead of water) for higher operating
temperatures instead of water, enhanced safety systems and new proliferation-resistant fuel technologies. Small and
modular reactors further ease the factory manufacturing process and since they require less water than larger reactors,
so they can be located in off-grid locations with limited water supply.
The nuclear submarine program of the Navy supported and benefited commercial land-based PWR development;
the same is true for space nuclear systems that benefited the development of compact systems.
The first power-producing nuclear reactor at the Calder Hall plant, in the United Kingdom, started generating elec-
tricity in 1951. These reactors, for example, Calder Hall (GCR), Dresden-1 (BWR), Shippingport (PWR), Fermi-I
(SFR), Kola-I (PWR) and Kola-II Russian Vodo-Vodyanoi Energetichesky PWR (VVER) reactors were part of the first
generation reactors (IAEA, 2020).
Generation I (1950s
1970s) refers to these earliest nuclear reactors. These reactors were the products of military-
and submarine-inspired efforts taking place with huge financial inputs, infrastructure developments in the post-Second
Word War era. This knowledge and experience, with massive financial expenditures, found useful application in the
construction and subsequent grid connections to the grid. Between 1954 and 1960, there were fifteen (15) operational
reactors with capacity 1087 MW(e).
Generation II (1970s
1990s) reactors, at higher capacity, were being deployed on a large scale with the objectives of
improving the performance, increasing capacity and fuel burnups with better safety and increased lifetime from 30
40 year
to 50
60 years. The number of reactors increased to 84 units with a capacity of 16,656 MW(e). Canada had developed the
PHWR from its Chalk River experience and was looking forward to its first exports. The UK had improved the gas-cooled
reactors, and the Soviet Union had upgraded the RBMK (Reactor Bolshoy Moshchnosty Kanalny) graphite moderated reac-
tors. In the next decade 161 reactors were added to the grid taking the number of NPPs to 245 with a capacity exceeding
133 GW(e). The US had 47 commercial reactors, USSR had 10, followed by Canada (4), India (3), Argentina (3) and Japan
(2), while China, Germany, and South Korea were the countries that had none. The size of reactors had increased to
B3000 MW(t)/1000 MW(e), with a few B3500 MW(t)/1200 MW(e) in the US, 1500 MW(t)/655 MW(t) in the UK and
going up to B3200 MW(t)/925 MW(e) in Russia. The United States had 31 PWRs and 16 BWRs, while the UK had
GCRs, and Russia had four types namely the large LWGR (RBMK-1000) 3200 MW(t) to the small LWGR Bilibino
62 MW(t)/11 MW(e), and intermediate PWRs (VVER) B1375 MW(t)/440 MW(e) reactors. The emphasis in these systems
was to go toward large designs for favorable economies of scale. These were large-scale power plants, for example, Bruce
(CANDU), Kalanin (PWR), Kursh 1
4 (LWGR), Palo Verde (PWR), Grand Gulf (BWR) and Fukushima II (BWR).
In the 1980s, the Republic of Korea and Germany added seven (7) and six (6) reactors, respectively, to their com-
mercial grids while China still had none.
The average capacity of nuclear reactors had increased from B200 MW(e) in 1970 to B540 MW(e) by 1980 and to
B760 MW(e) by 1990. Thus, the fivefold increase in the number of commercial NPPs had undergone a capacity
increase of about four times. These numbers indicate the sharp rise of the nuclear industry and the operational confi-
dence that led to the capacity increase.

TABLE 3.1 Nuclear reactor classifications by size.

Reactor size Standard terminology Power (MWe) Typical applications

Large NPP .1000 Baseload power
Medium
700
1000 Baseload power
Small SMR 100
700 Flexible grid power
Very small vSMR ,100 Remote locations
Micro MNR ,10 Propulsion
Battery Nuclear battery ,1
 Nuclear reactors and systems Chapter | 3 105

The two events that affected the awareness levels in the nuclear industry in this time period were TMI and
Chernobyl.
In 1975, the “Reactor Safety Study” report WASH-1400, known as the Rasmussen Report (Bartel, 2016) was the
“first full-scope use of Probabilistic Risk Assessment PRA techniques” that contributed to the awareness of quantifying
the risk of what can go wrong, its likelihood, and consequences. Published 4 years before TMI, the WASH-1400 Report
could correctly identify that a small scale Loss of Coolant Accident (LOCA) could initiate core damage which is a
design basis accident that occurred at TMI.
Generation III (1990 onwards) nuclear reactors were considered to have begun, with the purpose of design improve-
ments producing simpler and rugged designs which would lead to standardization and more efficient licensing and
construction processes. Technological improvements included enhanced passive safety features, higher operating
lifetimes of up to 60 years, higher burnup for better fuel utilization and reduced waste, reduced core melt and damage
probability, and reduced power generation cost. These reactors required certifications from the European Utility
Requirements (EUR) in Europe and by EPRI (Electric Power Research Institute)/URD (Utility Requirements
Document) in the United States.
In Europe, the European Pressurized Reactor, also called the Evolutionary Power Reactor (EPR) was the first to
obain the EUR license; the first EPR unit became operational in CHina (Taishan 1) in 2018. Anotehr large Generation
III Advanced Power Reactor APR1400, Section 3.3, is a 1400 MW(e) large PWR designed with safety barriers and pro-
tection on “defense in depth” implemented to five levels of protection. These include successive barriers to prevent the
release of radioactive material to the environment and an integral design (steam generators inside the Reactor Pressure
Vessel) to prevent a LOCA.
wGeneration III 1 had realized innovative and evolutionary designs in which safety features were enhanced,
radiation exposures reduced and the maintenance and operations made simpler. In 1994, China had its first connec-
tion to the grid; a 300 MW(e) PWR Qinshan-I. Globally, between 1990 and 2000, fifty-two (52) new reactors
were connected to the grid. The number of reactors increased in number to 416 by 1990 with a capacity of
B318 GW(e).
The Gen-III 1 , Advanced Passive Reactor AP1000, Section 3.3, is an upgrade of the Gen-III AP600 at 1100 MW
(e) is also a relatively large reactor. Its distinguishing features include a relatively large pressurizer connected to one
of the hot leg pipes variable fuel enrichment (2.3% 2 4.8%), fuel burnable absorbers for better lifespan, high average
burnup of 60 GWd/t, and thermodynamic efficiency B34%. A modular design is applied to the containment and for
the fabrication of equipment such as the reactor internals and the condenser. The advancements to the AP600 include
low maintenance and high availability. The AP1000 has good safety margin for shutdown and core damage frequency
B 5 3 1027/reactor year (RY). It meets very high standards of reliability and cost affordability and satisfies sustain-
able development goals, mitigating climate change and reducing carbon emissions associated with fossil fuel thermal
plants. These reactors included the CANDU 6, AP600, the Advanced Power Reactor APR1400, and the advanced
boiling water reactor (ABWR).
The Fukushima earthquake in March 2011, led the nuclear industry to take a fresh look at the nuclear industry.
The evolution of nuclear power reactors and designs (IAEA, 2020) from the early prototypes of the first generation
reactors of the 1950s thus led to large-scale deployment with second generation designs in the 1970s.
In the period 1990
2000, there were 52 new reactors connected on their grids. This was the period of the beginning
of evolutionary Generation III 1 designs, such as the Advanced Passive AP1000 Reactor. In the next two decades,
from 2000 to 2010 there were 32 new reactors and from 2010 to end of December 2019, there were 58 new reactors
connected to the grid.
The most notable programs were those of China with forty-eight (48) new reactors, India with sixteen (16), Japan
twenty-six (26) until 2010, South Korea twenty-one (21) and Russia seventeen (17). Thus, from 2000 until the begin-
ning of 2020, out of the 90 new reactors, 71 were in Asia and twelve (12) in Russia.
As of the end of 2019, there were 448 operational nuclear reactors in the world. Most of the reactors in the world
are PWRs (68%), followed by BWRs (14%), PHWRs (11%), GCRs (3%), LWGR (3%), and Fast Breeder Reactors
(FBRs) (1%). There are 305 operating PWRs in the world (IAEA, 2020) out of which 42 produce less than 600 MWe
and 263 produce greater than or equal to 600 MWe. Out of 54 new constructions, 44 are PWRs. By far, the PWR pro-
gram is the most successful in the world contributing to over 60% of the total nuclear power reactors.
The number of nuclear reactors by region indicates the rising trend of nuclear energy in Asia, the stagnation in the
United States, and the decline in Europe and the United Kingdom. Asia (Far East) is leading with both operational and
under-construction reactors counted, followed by Northern America, Western Europe, Central and Easter Europe, then
Middle East and South Asia and Africa.
106 Nuclear Engineering

In China, nuclear power development is taking place at a phenomenal level; with its 49 reactors and a B48 GW(e)
capacity, 16 reactors are under construction. This will modernize China’s power infrastructure and reduce its heavy
dependence on coal which presently contributes to over 65% of the electricity generation. The carbon dioxide emissions
of China at 28% of the world’s total places it at first position and hence its’ move to nuclear power is bound to in line
with sustainable development goals. The new builds are mainly large reactors such as AP1000, Hualong One (based on
the GenII 1 CPR1000 and AP1000 technologies) and the CAP1400 whose extensions will be 1400 and 1700 MW reac-
tors with China holding intellectual property rights. With reactors under construction, as well as 43 planned reactors,
China with 107 nuclear power plants (NPPs) is set to become the world’s number one nuclear power program with a
capacity exceeding 100 GW(e).
It is important to consider the number of permanent shutdowns, and its long-term effect on nuclear power at the
global level. As of December 2019, there have been 186 permanent shutdowns with Europe seeing 62, USA 37, UK 30,
Japan 27 and Canada 6. By February 2020, the number of shutdowns had increased in Europe with closures of 30 reac-
tors in Germany, six in Sweden, four each in Italy, Ukraine and Bulgaria, and three each in Spain and Slovakia. Many
of these have been due to the expiration of design life. The number of shutdowns peaked in 2011 with the Fukushima
disaster in March 2011 leading to the closure of four units.
Another peak in shutdowns was in 2019 mostly due to the aging of reactors and expiration of licenses. The average
age of nuclear reactors in the world is about 20 years while the US reactors on the average have an age of about 40 years.

3.1.2 Reactors shut down

The design life of Generation-1 nuclear reactors was generally about 30
40 years; several reactors were shut down
before the expiry of this design life due to economics of their electricity generation or other operational problems. In
spite of this design life span, the operational life span, as in improved reactors, can be much longer as there are no
strong technical limits to prevent extending reactor operations by another 20 years, and possibly up to 80 years. In the
United States, for example, with 96 operational reactors, as of January 1, 2021, providing about 20% of the electricity,
it is likely that most will get life extensions to go up to possibly 80 years.
Of the shut down reactors across the world, 57 are PWRs, 50 are BWRs, and 38 are graphite-moderated GCRs; this
probably marks the end of the UK-type GCRs and a decline in BWRs although four are under construction and nine are
planned. The PWRs, however, are on the ascendance, with 44 under construction and 57 reactors planned.
The UK program of early Magnox GCRs, the first in the world to have generated electricity, had to shut down reducing
its fleet to AGRs and a PWR With these shutdowns, The UK is considering the Chinese Hualong Design; similar to the two
1100 MWe units which China has constructed in Pakistan out of which one was connected to the grid in March 2021.
With its quick progress, China has also entered into foreign agreements most notably with Pakistan and fourteen
other countries including Romania, Argentina, the UK, Iran and Turkey. In Pakistan, 300 MW PWRs are in operation
at Chashma as the C-1 to C5 series and in December 2020, the fuel loading of the Karachi-2 (K2) Hualong 1100 MW
was followed by criticality in March 2021 and grid connection the same month. Out of 8 nuclear reactors in Pakistan,
seven PWRs (B3700 MW) have been set up by China.

3.1.3 The future of the nuclear power industry

The future of the nuclear industry is thus of vital concern to several countries. As the largest operator of nuclear reac-
tors, the United States currently has 96 operational reactors at 58 NPPs; these have an average age of B40 years which
approaches the limit of the design lifespan. These reactors provide B20% of the total electricity generation of the US.
After the Fukushima earthquake disaster of March 11, 2011, the nuclear energy prospects have suffered a setback.
In earlier days, Japan had 61 NPPs, out of which 27 were shut down and one was under construction. Out of the
remaining operational 33 NPPs, 16 were PWRs and 17 were BWRs. Currently 31 NPPs have a capacity of 31.7 GW(e).
Japan’s energy situation is heavily dependent on oil imports and hence it has been looking toward a reduced depen-
dency on fossil fuels, as well as on nuclear energy after Fukushima; leaving renewables as the preferred source. In spite
of this, the share of nuclear is planned to be B20% by 2030.
Since its inception in the 1950s, the nuclear industry has had an excellent safety record; it has been a reliable clean
energy baseload power source contributing to B16% of the electricity at one time and reducing now to B 11%. From
the three accidents that have taken place, TMI, Chernobyl and Fukushima, considerable knowledge has been gained
and incorporated into new designs especially the Gen-IV designs.
Based on the statistics and trends discussed above, there are conflicting signals on the growth of the nuclear industry.
 Nuclear reactors and systems Chapter | 3 107

The key advantages of nuclear energy are:

1. clean and low-carbon energy generation; if col or gas would be burnt instead of nuclear, hundreds of millions of
tons of carbon dioxide woud be released in to the atmosphere, worsening the environmental problems,
2. high-density energy,
3. safe track record, and
4. reliability.
The nucler industry faces challenges due to
1. high capital cost and
2. radioactive high-level waste management.
Some sustainable growth strategies for the nuclear industry are
1. technology improvements and enhanced safety in large power reactors,
2. development of small modular reactors (SMRs) to become commercially viable by the 2030s, and
3. hydrogen production from nuclear reactors.

3.2 Nuclear reactor systems

Nuclear reactors are huge and capital intensive power infrastructures integrating technologies that are both conventional
and hi-tech. Each NPP as a system consists of several subsystems. The International Atomic Energy Agency has classi-
fied the major systems as
1. The primary system
2. Balance of plant
3. Spent fuel storage
4. Nonelectrical systems
Essentially, a nuclear reactor is a system that produces thermal energy from fissions in the fuel elements placed in
its core. The thermal energy can be converted into any other form of energy from which “useful” work can be extracted.
In the form of a mathematical statement, this is the First Law of Thermodynamics which states that energy can neither
be created nor destroyed but it can be converted from one form to another. In the PWR core, the heat produced Q raises
the enthalpy H of the water entering at some temperature and pressure Ti ; Pi , and the heat is used to produce work W in
the turbine from which electricity is generated. After the steam has been used, it loses its energy, flows through a con-
denser and is pumped back into the steam generator. The efficiency η of a reactor is a measure of the ratio of net work
to heat generated. Reactors are designed in a way that maximizes their efficiency; for this it should produce heat at the
highest temperature possible and reject heat at the lowest temperature possible.
The subsystems (IAEA, 2007) of each system are classified as follows:
The reactor vessel (RV) which includes the reactor core and the control systems, the coolant system, the pressurizer
and the steam generator with safety and protection systems.
In the reactor core, the components include the fuel assembly with fuel rods and cladding surrounded by the moder-
ator and coolant. The engineering design of the reactor core covers the areas of neutronics, radiation transport, thermal
hydraulics, structural mechanics and control systems. Detailed calculations are used to obtain design and operating para-
meters for the amount and placement of fissile material in the core, the flow of coolant, and the control of reactivity.
The system pressure, in a PWR for example, is maintained by a pressurizer which ensures the state of the coolant and
maintains the high pressures required.
At the heart of the control system is the control rod mechanism and the calculation of operational load following
movements as well as maintaining reactivity margins for safe shutdown in the event of any departure from the pre-
scribed set points. In addition to control rods, nuclear reactors use burnable absorbers such as boron, gadolinium, euro-
pium in the form of compounds. Boric acid, for example, is dissolved in the coolant to reduce the reactivity in case it
increases. Another operational feature, to “flatten the flux” is to provide the fuel rods with a burnable poison at the
beginning of a cycle to enhance the performance of the reactor.
A steam generator is a major component where heat transfer from the nuclear system takes place to the working fluid
such as water which gets converted to hot “superheated” steam which is intended to drive a turbine. Here, the shell and
108 Nuclear Engineering

tube exchange process is dependent on the flow rate and thermal capacity of the fluid as well as the physical condition of
the system. The design and operation parameters of a steam generator are based on thermodynamics and structural analysis.
Safety systems are based on several independent and interdependent subsystems to ensure operational functionality
in case of any minor or major departures from set points. These utilize containment pressure suppression systems, core
spray pumps, flowrate regulators, and other components, which includes the turbine and condenser.
In the “secondary” or conventional part of a nuclear plant the major components are turbine(s), generator(s), con-
denser, feed-water system, emergency power supply systems and fire protection systems.
For spent fuel storage, the reactor has subsystems comprising spent fuel pool and an interim storage facility.
The major areas are discussed in the context of specific reactor designs in the following subsections.

3.2.1 Pressurized water reactor

The PWR, illustrated in Fig. 3.1, is classified as a light water (thermal) reactor since it uses light (ordinary) water as the
moderator and coolant. As discussed in Chapter 2 (interaction of neutrons with matter) fission neutrons colliding with a
hydrogenous light material are able to slow down and thermalize in a small number of collisions. This gives a “thermal”
flux for which the fission cross-section is higher than that for fast neutrons.
The major components in the primary side of a PWR are the pressure vessel, the core, the pressurizer, the steam
generator, the coolant pump and the control system shown in Fig. 3.1 with the residual heat removal system in a sepa-
rate building. The secondary side is the “balance of plant” which is common with conventional power plants.
These are common to all PWRs though the number of loops may vary depending on the size of the plant. A “four-
loop” Westinghouse PWR core, shown in Fig. 3.2, has four steam generators and a pressurizer within the pressure ves-
sel. The terminology refers to the number of steam generators and primary circuit coolant pumps. The internals of the
core are shown in detail in Fig. 3.2B.
The fuel rods and control rods are vertically placed within the reactor while “cold” coolant fluid enters from the top
of the core on one side flows downwards then upwards and exits as a hot fluid from the top on the other side. In the
steam generator, heat is transferred to the secondary side fluid (water in a PWR) from which superheated steam flows
to the turbine where electricity is generated. The secondary loop is closed through the condenser and return pump.
Some representative figures for the Westinghouse Advanced Passive reactors AP600 and AP1000, and the KEPCO
Advanced Power Reactor APR1400 are shown in Table 3.2. AP1000 and APR1400 are typical of the “big” league of
reactors, with the European Pressurized Reactor (EPR) 4590 MW(t)/1660 MW(e) Taishan units 1 and 2 as mentioned in
the previous section.

FIGURE 3.1 Schematic diagram of a pressurized water reactor.

 Nuclear reactors and systems Chapter | 3 109

STEAM GENERATOR
ROD TRAVEL
HOUSING
CONTROL ROD INSTRUMENTATION
DRIVE MECHANISM PORTS
THERMAL SLEEVE
UPPER SUPPORT
PLATE LIFTING LUG
MAIN COOLANT PUMP
INTERNALS CLOSURE HEAD
SUPPORT ASSEMBLY
LEDGE HOLD-DOWN SPRING

CORE BARREL
CONTROL ROD
GUIDE TUBE
SUPPORT COLUMN
CONTROL ROD
DRIVE SHAFT
UPPER CORE
PLATE
OUTLET NOZZLE INLET NOZZLE

BAFFLE RADIAL CONTROL ROD

SUPPORT CLUSTER (WITHDRAWN)
BAFFLE

CORE SUPPORT ACCESS PORT

COLUMNS
INSTRUMENTATION REACTOR VESSEL
THIMBLE GUIDES
RADIAL SUPPORT

PRESSURIZER CORE SUPPORT LOWER CORE PLATE

REACTOR

(A) (B)
FIGURE 3.2 (A) A four-loop PWR, (B) PWR core. https://www.nrc.gov/reading-rm/basic-ref/students/for-educators/04.pdf. Courtesy of
Westinghouse Electric Company (c) 2021. All rights reserved.

TABLE 3.2 Overall characteristics of AP-PWR and APR1400 nuclear reactors.

Parameter AP600 AP1000a APR1400b

Thermal power (MW(t)) 1940 3400 3983
Net electrical power (MW(e)) 600 1100 1400
Efficiency (%) 31 32 35
Plant life span (y) 60 60 60
Average plant factor (%) 93 93 . 90
Equivalent core diameter (m) 2.921 3.04 3.63
Active core height (m) 3.658 4.267 3.81
a
AP1000 https://aris.iaea.org/PDF/AP1000.pdf.
b
APR1400 Advanced Power Reactor https://aris.iaea.org/PDF/APR1400_2020May.pdf https://aris.iaea.org/sites/core.html.
Source: AP1000 iaea.org https://www.osti.gov/etdeweb/servlets/purl/20256613; AP600 https://aris.iaea.org/PDF/AP-600.pdf.

As Table 3.2 shows, the reactors are “high power density,” for example, 3983 MW(t) in a volume of B 40 m3
is B100 MW(t)/m3. With such high power densities, PWRs require very high levels of safety and control.
The efficiencies of these reactors is still a maximum of 35% due to thermodynamic limits of the Rankine cycle. These
efficiencies are optimized by minimizing the energy losses in the system. The lifetimes of the PWRs given in Table 3.2
are B60 years although there is no technical limit other than the usual wear and tear of components. These plants have
to be safe to avert any unplanned departure from design conditions. Some safety goals of the AP1000 are: core damage
frequency , 5.09 3 1027 per RY, occupational radiation exposure , 0.7 person-Sv/RY, and operator action time of
half an hour.
A design comparison of a typical 1000 MW(e) PWR with a small 300 MW(e) reactor is given in Table 3.3 illustrat-
ing significant differences in size and fuel and coolant flow.
110 Nuclear Engineering

TABLE 3.3 Physical design parameters of AP1000 and PWR 300 MW(e).

Core AP1000a CHANSNUPPb (300 MW)

Active height (m) 4.267 2.9
Equivalent diameter (m) 3.04 2.486
Fuel material Sintered UO2 UO2
Fuel enrichment (%) 2.35/3.40/4.45 2.4/2.67/3.0
Fuel inventory (t) 96.176 35.917
No. of assemblies 157 121
Rods per assembly 264 204
Assembly pitch (cm)
20.03
Rod pitch (cm) 1.26 1.33
Geometry (no. of rods) 17 3 17 15 3 15
Pin outer dia. (mm) 9.5 10
Pellet height (mm) 9.83 10
Pin height (mm)
3210
Clad/thickness (mm) 0.5715 0.70
Clad material ZIRLO Zircalloy
Moderator H2O (liquid) H2O (liquid)
Moderator inventory (MT)
57
Coolant H2O (liquid) H2O (liquid)
Primary pumps 4 2
No. of loops 2 2
a
https://www.westinghousenuclear.com/new-plants/ap1000-pwr.
b
Mian, Z. and Nayyar, A. H. (1999). Pakistan’s Chashma Nuclear Power Plant, PU/CEES Report. No. 321, Center for Energy and Environmental Studies,
Princeton Environmental Institute, Princeton University, Princeton, NJ 08544
5263, Dec. 1999, p. 62.
Source: AP1000 fuel data from https://www.nrc.gov/docs/ML0715/ML071580895.pdf.

TABLE 3.4 Process variables of AP1000 and CHASNUPP.

Core AP1000* CHANSNUPP** (300 MW)

Burnup (MWD/t) 60,000 30,000
Max. centerline temperature ( CÞ 2593 1806
Max. clad temperature ( CÞ
345.7
Coolant Water Water
Coolant flow rate (tph) 14300 kg/s, (51480 tph) 24,000
Coolant inlet temperature ( CÞ 279.4 288.5
Coolant outlet temperature ( CÞ 324.7 315.5
Pressure (kg/cm2) 155.13 155
Geometry 17 3 17 15 3 15
Moderator temperature at full load ( CÞ 303 302

*https://www.nrc.gov/docs/ML0715/ML071580895.pdf.
**Mian, Z., and Nayyar, A. H., Pakistan’s Chashma Nuclear Power Plant, PU/CEES Report. No. 321, Center for Energy and Environemntal Studies, Princeton
Environemental Institute, Princeton University, Princeton, NJ 08544-5263, Dec. 1999, p. 62.

The scale-up in process variables of both systems is given in Table 3.4 showing a double discharge burnup, higher
operating centerline temperature and higher coolant flow rate.
The control systems of both reactors use Ag-In-Cd control rods and borated dilution

3.2.2 Boiling water reactor

In a BWR (Fig. 3.3), heat is carried by the coolant (water) flowing upwards through the core; this steam-water mixture
then passes through separation systems at the top of the core, and hot dry steam leaves the core. The hot dry steam goes
 Nuclear reactors and systems Chapter | 3 111

FIGURE 3.3 Schematic design of a Boiling Water Reactor.

TABLE 3.5 Physical design parameters of BWR (Fukushima).

Main Electric output (MW) 1100
specifications Start of construction May-73
Start of commercial operation Oct-79
Reactor type BWR5
Containment type Mark II
Main contractor GE/Toshiba
Nuclear reactor Heat output (MW) 3293
Number of fuel assemblies 764
Full length of fuel assemblies (in.) 176
Number of control rods 185
Reactor pressure vessel (RPV) Inner diameter (in.) 252
Height (in.) 906
Total weight (short ton) 827
Design pressure (psi) 1249.9
Design temperature ( F) 576
Primary containment vessel Height (ft) 157.5
(PCV) Diameter of cylindrical portion (ft) 32.8 (top)
Diameter of spherical portion (ft) 82.0 (bottom)
Suppression pool water amount 845.4
(kgal)
Design pressure (psig) 40.6
Design temperature ( F) 340 (DW) 221 (SC)
Steam turbine Number of revolutions (rpm) 1500
Steam temperature ( F) 540
Steam pressure (psig) 950
Fuel Type Uranium dioxide (Unit 3 contains
MOX)
Uranium (ton) 132

Source: https://www.tepco.co.jp/en/nu/fukushima-np/outline_f1/index-e.html.

to the turbine where thermal energy is converted into mechanical energy and eventually to the generator where mechan-
ical energy is converted into electrical energy.
On exit from the turbine, the “low quality” steam and water is condensed into water and pumped back into the core.
The physical design parameters of Unit 6 of the Fukushima BWRs are shown in Table 3.5. The units 1
4 of the
Fukushima Daiichi NPP were abolished on April 19, 2012, and units 5, 6 were abolished on January 31, 2014 in accor-
dance with the Electric Utility Law. The Fukushima Daiichi BWRs reactors are GE designs of the 1960s known as
Mark I containment. They came into commercial operation from 1971 onwards; the earlier units were 460 and 780 MW
(e) while unit 6 given in the Table was 1100 MW(e) During normal operation they all had a core outlet temperature of
112 Nuclear Engineering

TABLE 3.6 The advanced boiling water reactor (ABWR).

Parameter ABWR ABWR-II

(GEHa, HGNEb, and Toshiba) GE-Hitachi
Thermal power (MWt) 3926 4960
Net electrical power (MWe) 1350 1638
Efficiency (%) 34 33
Plant life span (y) 60 60
Average plant factor (%) 90 93
Equivalent core diameter (m) 5.163 5.41
Active Core height (m) 3.81 3.71
a
General Electric Hitachi.
b
Hitachi General Electric Nuclear Energy.
Source: ABWR https://aris.iaea.org/PDF/ABWR(Hitachi-GE)_2020.pdf, ABWR-II https://aris.iaea.org/PDF/ABWR-II.pdf, Core data https://aris.iaea.org/sites/
core.html.

286 C under a pressure of 6930 kPa and with 115
130 kPa pressure in dry containment. The operating pressure was
about half that in a PWR. A drawback of the safety system was that the reactors had analog instrumentation in a ground
floor building instead of digital controls on an elevated floor; thus, information could not be accessed remotely.
For improved designs, the 1350 MW(e) advanced ABWR, with overall parameters listed in Table 3.6, has an
improved plant life of 60 years, plant availability of 87%, longer refueling period of 24 month intervals, and reduced
probability of core damage of 1025 per RY. The enhanced safety systems include an emergency core cooling system
designed to ensure emergency cooling in the event of a design basis LOCA, so that the cladding temperature does not
reach the limit of 2200 F (1204 CÞ in accordance with USNRC 10CFR50.46. Concerns of radiation have been
addressed by reducing both the radiation exposure (limited to 100 man-rem/year) and radiation waste.
As of December 2019, there were 65 operational BWRs while 50 were shut down, four were under construction,
and nine are planned. Most of the operational units were of capacity greater than 600 MW(e).
In Japan, 16 BWRs (out of which 10 were in Fukushima) were permanently shut down. In the US, 13 BWRs, mostly
connected to the grid in the 1960s, (Big Rock Point US-155, Bonus US-014, Dresden-1 US-10, Elk River US-11, GE
Vellecitos US-018, Humboldt Bay US-133, Lacrosse US-409, Millstone-1 US-245, Oyster Creek US-219, Pathfinder US-
130, Pilgrim-1 US-293, Shoreham US-322, Vermont Yankee US-271) were permanently shut down. In Europe 19 BWRs
were permanently shut down: 10 in Germany, 2 in Italy, 4 in Sweden, and 1 each in Switzerland, Spain and Netherlands.
Based on current information (IAEA, 2020), the only countries planning BWRS are Japan (six ABWRs, 3926 MW(t)
each) and USA (1 ESBWR 4500 MW(t) and 2 ABWR 3926 MW(t)) while two ABWRs are under construction in Japan.

3.2.3 Pressurized heavy water reactor

The design of the CANDU PHWR can be traced back to the Manhattan Project when the Chalk River Laboratory was
established in 1944 for supplying spent fuel from the plutonium production reactor NRX for the US weapons program.
The NRX suffered a partial meltdown and was replaced by the higher flux NRU to increase production.
The experience from NRX and NRU led to the CANDU design starting with a small 20 MW(e) reactor, called the
Nuclear Power Demonstration (NPD). Subsequently, the 200 MW(e) Douglas Point reactors were built and commis-
sioned in 1966. The reactor power was increased to 500 MW by increasing the diameter of the calandria from B 8 to
B10 cm with a corresponding increase in the number of fuel elements per fuel bundle from 19 to 28.
The PHWR uses natural uranium fuel rods, heavy water cooled and moderated “horizontal” reactor with several
simple and safe features. The overall plant schematic is shown in Fig. 3.4.
The fuel is bundled in a pressure tube typically 8
10 cm in diameter and placed horizontally in pressure tubes within
a calandria (a horizontal cylindrical tube). Several calandria tubes are inserted into the horizontal moderator tank. Coolant
flows though horizontal channels between fuel rods as shown in Fig. 3.4. Overall physical design parameters of CANDU
variants are listed in Table 3.7 from the relatively low power Indian PHWR to the Enhanced CANDU 6 design.
Another variant of the PHWR design is the 300 MW(e) thorium-fueled Indian advanced heavy water reactor (AHWR)
designed and developed by the Bhabha Atomic Research Center (BARC). This uses a vertical pressure tube with boiling
light water coolant and heavy water moderator reactor The design incorporates passive safety systems working on natural
 Nuclear reactors and systems Chapter | 3 113

FIGURE 3.4 Schematic diagram

of a pressurized heavy water reactor.

TABLE 3.7 Physical design parameters of PHWR designs.

Parameter IPHWR-220 APHWR (BARC) Enhanced CANDU 6

Thermal power (MWth) 754.50 920 2084
Net electrical power (MW(e)) 235.81 284 740
Efficiency (%) 26.5
35.5
Plant life span (y) 40 100 60
Average plant factor (%) . 90 90 90
Equivalent core diameter (m) 4.51 6.9 7.595
Active Core height (m) 4.95 3.5 6.28

Source: IPHWR-220: https://aris.iaea.org/PDF/IPHWR-220.pdf, EC6: https://aris.iaea.org/PDF/EC6.pdf, APHWR:Bhabha Atomic Research Center http://www.
barc.gov.in/reactor/ahwr.pdf.

laws such as force of gravity and heat absorption in a gravity-driven water pool, independent shutdown systems, passive
poison injection into the moderator for a wired-system failure. Due to boiling in the coolant, a steam generator is replaced
by a simpler steam drum; similarly a number of major components such as primary coolant pumps are not required. All
reactivity temperature coefficients (fuel, channel, void and moderator) are negative. Minor actinide production, including
plutonium production, is reduced due to the thorium fuel making the fuel proliferation-resistant.
As of December 2019, there are 49 PHWRs operational in world; with 19 in Canada, 18 in India, 3 in South Korea,
2 in China, 2 in Romania, 3 in Argentina, and 1 in Pakistan, and 1 is under construction.

3.2.4 Gas cooled reactor

Gas cooled Magnox (magnesium-aluminum alloy cladded fuel rods) reactors, as depicted in Fig. 3.5 below, were con-
structed in the 1950s for both commercial and military use. The first Magnox reactors at Calder Hall, commissioned in
August 1956, and decommissioned in 2003, were also the first nuclear reactors to produce electricity on a commercial
basis, though their main purpose was plutonium production. Each of the four 268 MW(t)/60 MWh reactors used natural
uranium fuel, graphite moderator, and carbon dioxide gas as the coolant.
The first generation Calder Hall reactors were decommissioned and succeeded by the second generation Windscale
Advanced Gas Cooled Reactor. As pioneer of the civil nuclear program, the UK had 26 Magnox units, with grid connections
between 1962 and 1971, including Berkely, Calder Hall, Chapelcross, Dungeness, Hinkley Point, Hunterston, Oldbury,
Sizewell, Trawsfynydd and Wylfa. These have all been decommissioned by Magnox Ltd. on behalf of the UK government.
The Calder Hall reactors had natural uranium metal fuels, in the form of one cast bar, with element length 1.016 m
and 6 elements in a channel. The reactor had an active core height of 6.4 m and an active core diameter of 9.45 m. The
last Magnox reactor, Wylfa, produced power about seven times more thermal power, but with about 10 times more elec-
trical output than that from Calder Hall. Thus efficiency had been significantly increased mainly by increasing both the
mass of uranium in the core and the total gas flow.
114 Nuclear Engineering

FIGURE 3.5 Schematic diagram

of a gas cooled reactor.

TABLE 3.8 Physical design parameters of the Torness GCR.

Parameter Torness AGR

Thermal power (MW(t)) 2 3 1623
Net electrical power (MW(e)) 2 3 660
Efficiency (%) 40.7
Plant life span (y) Grid connection 1989; lifespan increased to 2030a
Average plant factor (%)
Equivalent core diameter (m) 9.5
Active core height (m) 8.3
a
https://www.world-nuclear-news.org/C-EDF-Energy-extends-lives-of-UK-AGR-plants-1602164.html.
Source: Nonbel, E. (1996). Description of the advanced gas cooled type of reactor (AGR), Riso National Laboratory, Roskilde, Denmark, Nov. 1996. https://
www.osti.gov/etdeweb/servlets/purl/448871.

Currently there are 14 operational reactors, all in the UK (IAEA) at 6 sites and 7 stations. Table 3.8 shows the over-
all data for Torness GCR which is based, as the previous Magnox reactors, on graphite moderator and CO2 coolant but
with a change of fuel from metal natural uranium to 2.2% enriched uranium in UO2 fuel. These changes have reduced
the physical size of the plant, raised outlet operating pressure and gas flow rate as well as the gas temperature, resulting
in 40.7% efficiency.
The AGRs need to improve in their design to be competitive with LWRs particularly PWRs to which they have lost
the market, even though GCRs were the first nuclear reactors to produce commercial power.

3.2.5 Fast breeder reactor

The FBR program began in the 1940s with the objective of using high enriched uranium to produce small-size reactors
with liquid metal coolants. This would increase the overall plant efficiency as well as the neutron economy by “breeding”
plutonium, which does not occur in nature, from a fast neutron spectrum. These efforts led to the experimental breeder reac-
tor (EBR-I) which in 1951 was the first reactor to demonstrate the production of electricity. The EBR-I, followed by EBR-
II and the French Phoenix FBR were all shut down after just a few years of operation due to several technical difficulties.
At present the only two fast reactors in operation, BN-600 and BN-800, are in Beloyarsk in Russia (IAEA, 2017).
As shown in Fig. 3.6, a liquid metal cooled fast reactor has a primary circuit in which high temperature liquid metal,
such as sodium, flows the core and exchanges heat in an intermediate heat exchanger with a coolant, such as water.
From there on, the conventional part of the power conversion unit is the same as discussed in the previous sections for
the PWR, BWR and GCR.
The BN-600 reactor with overall parameters and design evolutions shown in Table 3.9, has been in operation in
Beloyarsk since the last 30 years. The design improvements were made due to lessons learned from operational experiences
 Nuclear reactors and systems Chapter | 3 115

FIGURE 3.6 Schematic diagram

of a Liquid Cooled Fast Reactor.

TABLE 3.9 Characteristics of the BN-600 Sodium cooled Fast Breeder Reactor (FBR).

Parameter Reactor Type I Type M Type M-1

Thermal power (MWth) 1470 1470 1470
Net electrical power (MWe) 600 600 600
Efficiency (%)
Plant life span (y)
Average plant factor (%)



Equivalent core diameter (m) 2.058 2.058 2.058
Active core height (m) 0.750 1.000 1.030
Fuel inventory in core (kg) 8260 11630 12090
Average fuel burnup (MWd/kg U) 42.5 44.5 60.0
Fuel enrichment (U-235)
Low enrichment zone 21 17 17
Medium enrichment zone
21 21
High enrichment zone 33 26 26

LEZ, low enrichment zone; MEZ, medium enrichment zone; HEZ, high enrichment zone.
Fast Reactors and Related Fuel Cycles: Next Generation Nuclear Systems for Sustainable Development FR17, Proceedings of an International Conference
Yekaterinburg, Russian Federation, 26
29 June 2017. https://www-pub.iaea.org/MTCD/Publications/PDF/STIPUB1836web.pdf.
Source: IAEA Tecdoc 1569, p. 123.

the power of the plant was kept the same while increasing the active core height from 750 mm to 1030 mm, “flattening” the
flux by incorporating three enrichment zones, and increasing the fuel inventory of the core from B8.3 tons to B12 tons; all
these modifications improved fuel utilization enhancing the burnup by almost 50%, to 60 MWd/kg U.
A power upgrade to BN-600, the sodium cooled BN-800 reactor was commissioned for commercial operation in
2016 at the Beloyarsk NPP. In addition to these two operational reactors, the BN-1200 design is under development for
technical and economic feasibility.
The directions for possible improvements in fast reactor design are under investigation; several material selection
decisions are yet to be taken such as the selection of sodium, lead, gas, molten salt, supercritical water cooled systems,
as well as hybrids, such as accelerator-driven systems. As work continues toward innovative Gen-IV designs for the
next decades, safety and sustainability are being enhanced and being incorporated in new designs.
Out of 54 new constructions, one 470 MW FBR was under construction as of December 2019 in India and one
1500 MW(t)/600 MW(e) CFR-600 sodium cooled fast reactor (SFR) is under construction in Fujian province, China.
In spite of the major advantages of FBRs including their efficiency, compact design, and plutonium breeding, their
safety and proliferation resistance are issues that must be addressed for better public acceptance.
116 Nuclear Engineering

3.3 Marine propulsion reactors

3.3.1 Introduction
The use of nuclear reactors extend far beyond production of electricity for cities; they are used for numerous other
applications such as marine propulsion for submarines and icebreakers, radioisotope production for medicine and agri-
culture, testing of materials, desalination and plutonium production. This section presents an overview of the applica-
tions of nuclear reactors for marine propulsion that includes the propulsion systems of merchant ships, icebreakers, and
the nuclear navies with submarines, aircraft carriers and cruisers (Fig. 3.7).
Nuclear reactors for marine propulsion are attractive, in comparison with fossil fuels since they do not need air to
operate and hence submarines can stay under water for long periods, they have high power density, long operational
periods, and low operational costs.

3.3.2 US nuclear submarine program

Lobner (2018) presents the timeline for development of marine nuclear power in the US from Enrico Fermi’s briefing to
the Navy Department in 1939, followed by the first controlled critical pile (CP-1) established under Fermi’s leadership on
2nd December 1942, to Manhattan Project, Alvin Weinberg’s first description of the PWR, the Gunn-Abelson 1946 nuclear
plant design concept for a submarine, involvement of Argonne National Lab and Westinghouse to the beginning of the
construction work on the Submarine Thermal Reactor (STR) Mark 1 Prototype in 1950. The contractors included Bechtel
(B), Combustion Engineering (C), General Electric (G), and Westinghouse (W). Their codes appear in the naming of ship
power systems; for example, A1W represent aircraft carrier generation 1 built by Westinghouse. Other first alphabetic
codes are C for cruiser, D for destroyer and S for submarine. It was on the 1st of August 1946 that the US Atomic Energy
Commission was established. Two reactor designs actively pursued in the 1950s were based on the liquid metal cooled
fast reactor and the STR series.
After the second world war, the US Navy was quick to replace diesel and battery systems on their fleet with nuclear
power systems, starting from the nuclear-powered submarine USS Nautilus in 1955, powered by the STR Mark II
(S2W) reactor. The three big projects at that time were USS Nautilus in January 1955 STR Mark II Thermal PWR, USS
Seawolf in 1955 SSN-575 Liquid Metal Cooled Reactor S2G and USS Skipjack in May 1956, SSN-585 Advanced
Submarine Fleet Reactor (ASFR). The poor performance of the liquid metal reactor S2G on USS Seawolf led to its
removal and replacement by the Nautilus PWR. This was probably the turning point that gave the field to PWRs which
Westinghouse spearheaded for commercial reactor designs.
In 1958, keels were laid for the first US civilian nuclear-powered merchant ships NS Savannah, the SSN USS
Scorpion, and the USS George Washington.
Work on a GCR program began in the US in 1956. By 1959, the liquid metal design for submarine reactors had
been discarded in favor of PWRs.
By 1962, the US Navy had rapidly built up a fleet of 26 nuclear submarines which included the aircraft carrier USS
Enterprise powered with eight PWRs and the cruiser USS Long Beach powered with two PWRs. At that time, 30 more
nuclear submarines were under construction in a large nuclear submarine industrial infrastructure.

FIGURE 3.7 USS Gerald R Ford aircraft carrier. https://upload.

wikimedia.org/wikipedia/commons/b/b2/USS_Gerald_R._Ford_
%28CVN-78%29_underway_on_8_April_2017.JPG (in public
domain, permitted use).
 Nuclear reactors and systems Chapter | 3 117

3.3.3 Former Soviet/Russian nuclear submarine program

In Russia, work on a submarine using a nuclear propulsion reactor was initiated in 1952 (Reistad & Olgaard, 2006).
Within 4 years after USS Nautilus, the first Soviet nuclear submarine Leninsky Komsomol was commissioned in June
1958 and became operational in 1959 powered with two water cooled reactors 1.4 kW each in addition to two diesel
generators and two auxiliary electric motors. Around the same time, the first nuclear-powered civilian surface ship and
icebreaker Lenin, was constructed and commissioned in 1959 powered with three PWR nuclear reactors.

3.3.4 Submarine programs: UK, France, China, India and Pakistan

The S5W (Submarine 5th Generation Westinghouse) nuclear reactor propulsion system was provided to the UK, under
agreement, to power the UK’s first submarine HMS Dreadnought commissioned in 1960. S5W provided 15,000 shaft
horsepower (shp) equivalent to 11 MW for the submarine to achieve speeds of 20 knots (37 km/h) on surface and 28
knots (52 km/h) submerged. The S5W Westinghouse design enabled Rolls Royce to develop its first PWR in 1965
which powered the UK’s second nuclear submarine.
Apart from the UK which had nuclear submarines together with the US in the 1950s, France, China and India had
developed significant programs independently.
The first French nuclear submarine Le Redoutable (S611), commissioned in December 1971 was also powered by
a PWR.
The first Chinese nuclear submarine was commissioned in 1974; at present four ballistic missile submarines
(SSBNs) with two additional Jin-class SSBNs added in 2020, and six attack submarines (SSNs) form the nuclear part
of the total submarine fleet estimated at 60 (NTI: https://www.nti.org/analysis/articles/china-submarine-capabilities/)
The first Indian nuclear submarine INS Arihant (S2 for Strategic Strike) SSBN was commissioned in August 2016
and is powered by an 83 MW(e) PWR with 40% enriched uranium. This was followed by the INS Arighat ballistic mis-
sile submarine (S3) which will be part of a fleet of four nuclear-powered submarines to be armed with ballistic missiles
carrying nuclear warheads (Mian, Ramana, & Nayyar, 2019). In addition to these nuclear submarines, India has 15
diesel-electric attack submarines.
The submarine fleet of Pakistan has five diesel-electric attack submarines and three mini submarines; collaborative
efforts with China are underway for the construction of eight Type 39 Yuan-Class attack submarines (Mian et al., 2019;
Pakistan Navy, 2021) including two Agosta-70 commissioned in 1979
80 and three Agosta-90B, all diesel-electric ves-
sels purchased from France and subsequently assembled and made entirely in Pakistan.

3.3.5 Modern-day submarines

Today, forty countries have attack submarines, but only have five nuclear-weapon states
China, France, Russia, the
United Kingdom, United States
have nuclear-powered attack submarines while India has two SLBMs.
For naval submarines, they offer the added advantage of enhanced capability of stealth movement and quiet propul-
sion under water enabling them to undertake combat and intelligence-gathering activities. However, nuclear ships are
expensive; for example, the USS Gerald R. Ford aircraft carrier, length 1,092
1,106 ft and height nearly 250 ft and dis-
placement B 100,000 tons, with two nuclear reactors on board, would cost about $12.9 billion which is roughly the
cost of four 1000 MW nuclear power reactors. The total naval reactors budget request in 2018 was in excess of US
$1.47 billion in three heads namely naval reactors operations and infrastructure (US$ 467 million), naval reactors devel-
opment (US$ 473 million), and S8G prototype refueling (US$ 190 M). The weapons carried on submarines include
Tomahawk missiles, Trident missiles (SSBN), antiship cruise missiles, nuclear-tipped missiles, cruise missiles, subma-
rine launched ballistic missiles (SLBMs), intercontinental ballistic missiles (ICBMs), and multiple independently
targetable reentry vehicles (MIRVs).
Today, the US has 83 nuclear submarines including 10 aircraft carriers and 1 research vessel, with over 50 attack
submarines (SSNs), 14 strategic ballistic missile submarines (SSBNs), 11 nuclear aircraft carriers (CVNs) including 10
Nimitz-class CVNs and 1 Ford-class CVN.
The ship service life is 30
33 years and reactors are mainly S5W, S6W, S9G (Submarine platform 9th Generation
design by General Electric). The size of the subs varies from the smallest Virginia-class with a length of 377 ft, Beam
34 ft to the largest Ohio class with a length of 560 ft, beam 33 ft. The subs travel at speeds from 20 1 knots for the
large subs and from 25 1 knots for the smaller subs.
118 Nuclear Engineering

Most marine reactors are PWRs while some are liquid metal cooled reactors; in both cases, the steam produced
from the reactor heat is used either directly to power the propulsion system or indirectly by producing electricity from
turbo-electric generators.
The former Soviet Union produced 246 nuclear submarines, each with two PWRs, with most decommissioned in the
1990s, leaving Russia with 58. Typical mission times for first generation submarines were 60 days with underwater (at
depths of 200
300 m) speeds of 20
25 knots. At the height of the cold war in the early 1980s, both the US and the
then USSR were actively pursuing their programs with the total estimated to be over 460 nuclear-powered submarines.
Between 1955 and 1995, three generations and classes of nuclear submarines have been classified (Reistad & Olgaard,
2006) for attack, cruise missile and ballistic missile classes including 56, 102 and 78 submarines built with a total of
246. A few others were built between 1974 and 1996 bringing the total built to 264. Russia has continued with the
development of nuclear-powered icebreakers, for maintaining the Northern Sea Route in the frozen Arctic coast of the
North Pole region which is mostly free of ice only 2 months of the year and connects the Atlantic and Pacific Oceans.
From the first icebreaker Lenin, several are operational with 5 let Pobedy being the world’s largest icebreaker powered
by two OK-900A (171 MW(t) each) nuclear reactors. The design parameters of Lenin’s first generation LEU OK-150
nuclear reactor and the HEU KLT-40 reactor are given in Table 3.10.
In today’s scenario, nuclear ships including submarines, aircraft carriers and cruisers armed with hundred-kiloton or
more nuclear warheads and ballistic missiles as well as stealthy intelligence-gathering mobile platforms are part of mili-
tary strategy for global dominance. All of this has been made possible by fast moving nuclear-powered ships capable of
maneuvering battle conditions to defend themselves as well as to operate in the attack mode. In future, this list is bound
to expand, with Canada, Pakistan, Australia, Iran, Israel, and Brazil taking interest in the development of nuclear
submarines.

3.3.6 Technical features

There are major differences in the requirements of land-based and water-based nuclear systems such as the availability
of space, longer refueling period, power maneuverability, physical movement, vibrations, and corrosion. Space require-
ments necessitate compactness of nuclear systems which translates immediately into the need for higher fuel enrich-
ment. Fortunately, the power requirements are much less than that for typical land-based 1000 MW power plants.
Some distinguishing features of submarine nuclear reactors (Ragheb, 2011) are
1. The use of HEU, as high as 93%
96% enriched weapon grade uranium, in PWRs.
2. The use of high burnup fuels such as U-Zr, U-Al and ceramics.
3. The requirement of “lifetime cores” which means no refueling for B50 years.
4. The requirement of overcoming dead time from xenon 54135Xe (half life 9.2 hours) poisoning.
5. Internal radiation (neutron and gamma) shielding.
6. Operating temperatures B600
700  F (maximum fuel temperatures) in earlier U-Zr fuel elements, increasing to
B1700 F in UO2 fuel clad in stainless steel with beryllium moderator and reflector to give an intermediate neutron
spectrum and hence a higher thermodynamic efficiency.
7. Noise reduction from coolant pumps and speed reduction gearboxes.
8. Power levels typically 10
200 MW(e) compared with 300
1200 MW(e) for commercial plants.
9. Containment vessel design to withstand pressures (B13 atmospheres for earlier designs) that could be attained in
accident conditions.
Some variations in operating parameters are also mentioned in the literature; these could reflect changes in operating
conditions. Similarly, design improvements were made from OK-150 due to the LOCA; these included moving the inlet
and outlet tubes of the reactor tank to the top of the tank to avoid operator error resulting in loss of coolant from drain-
age, as had occurred in the Lenin reactor.
From Table 3.10, the total number of fuel rods is 7,704 with 189 fuel elements containing 36 fuel rods each and 30
elements containing 30 rods each. The diameter of each rod was 6.1 mm. With the burnup of 1 kg equivalent to B
1000 MWd, OK-150 achieved B20% full burnup. The control rods were inserted vertically, and flux flattening was
achieved by boron-10 burnable poison in the shroud tubes of the central rods.
For the present KLT-40 design, similar to OK-900, the burnable poison used is natural gadolinium with fuel ele-
ments. Shielding for most reactors is steel-water layers on the sides (radially) and concrete on top.
In Russian submarine reactors, the first generation water cooled and water-moderated reactors had enrichments of
B20% power limited to 70 MW(t) with separated primary and secondary system. These plants all had U-Al alloy fuels
 Nuclear reactors and systems Chapter | 3 119

TABLE 3.10 Nuclear submarine reactors OK-150 and KLT-40.

OK-150, first core load KLT-40

General
Time period 1950s Present
Reactor power (MW(t)) 90a 135c
Burnup (MWd) 18,000
20,000b 62,000
68,000
Pressure vessel diameter/height (m) 2/5
Core
Core height (m) 1.58 1.0
Core diameter (m) 1.0 1.21
Fuel
U-enrichment (%) 5 90
Mass of U235 in core (kg) 85 150.7
Fuel material UO2 U-Zr-alloy
Number of fuel elements 219 241
Fuel element lattice pitch (mm) 64 72
Fuel element lattice type Triangular Triangular
Shroud, outer diameter (mm) 54 60
Shroud material Zr-alloy? Zr-alloy?
Number of fuel pins per element 36 53
Fuel pin diameter (mm) 6.1 5.8
Fuel pin lattice pitch (mm) Not available 7.2
Cladding thickness (mm) 0.75 Not available
Cladding material Zr-alloy or SS Zr-alloy
Helium gas gap fuel and cladding (mm) 0.05

Fuel pellet diameter (mm) 4.5 Not given
Coolant
Loops 2

Steam generator per loop 1

Coolant pump per loop 2
Pressurizers (total number); in Russian, called volume compensators 4

Inlet temperature ( C ) 248 278
Outlet temperature ( C ) 325 318
Pressure (bar) 200 130
Coolant flow rate (m3/h) 1000
Steam
Output (t/h) 360
Pressure (bar) 29 40
Temperature ( C ) 310 290
a
Modified to 159 MWth for OK-900 in 1967
70, then to 171 MWt for nuclear-powered icebreakers Arktika, Sibir, Rossia, Sovetskiy, Soyus, and Yamal.
b
Burnup was increased to 29,000
38,000 MWd for OK-900 lin 1967
70, later increased to 88,000
96,000 MWtd.
c
In Sevmorput; for icebreakers Taimyr and Vaigatch, 171 MWt.
Source: Reistad, Nuclear Plants Russia, p. 18.

with enrichment varying from 5.45 to 21% with 30
50 kg U-235 in the core, except for K-140 which had a higher
power (90 MW(t)), 21% enrichment and 116.3 kg U-235. With fuel densities of the order of 10 g/cm3, the core height
suggested is B 1 m. Improvements for second generation systems, such as the reorganization of reactor systems and
reducing the number of reactors from two to one as well as the use of lighter material made the reactors more compact
with more assemblies at B20% enrichment or less assemblies with B40% enrichment; shaft power and mission time
were increased. Third generation systems requiring more power contain more fissile material than earlier designs.
These reactors used heavy biological (steel-water-concrete) shielding amounting to B50% of the plant weight. For con-
trol systems, europium rods, in the form of Eu2O3 were used.
120 Nuclear Engineering

3.3.7 HEU/LEU submarine reactors

The use of HEU makes it possible to meet the space constraints of submarines as well as to overcome, to a significant
extent, the “dead time” caused by the buildup of xenon poison.
The present life of US submarine HEU cores is 33 years for the Virginia-class attack submarines being extended to
42 years for the cores of the Columbia-class ballistic missile submarines. All UK submarine reactors are HEU fueled
PWRs based on earlier Westinghouse designs with typical power levels of 15,000 shaft horsepower (78 MWth/11 MW
(e)) for the PWR-1 design, and 27,500 shaft horsepower (145 MW(t)/20.5 MW) for the PWR-2 design with improve-
ments planned for PWR-3. (Lobner, 60 years of marine power https://www.lynceans.org/wp-content/uploads/2015/09/
Part-4_UK-France-Others-60-yrs-of-marine-nuc-power.pdf)
Commercial power reactors use low enriched uranium (LEU) fuel; for example, LWRs typically use 3%
5%
enriched fuel while fast reactors use 13%
19.5% enriched fuel. Enrichment below 20% U235 concentration in ura-
nium, is called LEU while above 20%, called Highly Enriched Uranium (HEU) is prohibited for commercial nuclear
reactors as well as research reactors which were earlier based mainly on HEU. Another classification for uranium
enrichment is WG-HEU (Weapons Grade Highly Enriched Uranium) which is above 85% enrichment.
The role of PWRs for nuclear propulsion seems to be as large as that for commercial power; it may further
strengthen SMR technologies and extend into higher power densities for micronuclear reactors (MNRs) making it possi-
ble to enable return journeys to Mars. However, as more countries aspire for nuclear submarines, the use of weapons
grade uranium will increase as it is not prohibited by international law. At present HEU is used by submarines of the
USA, Russia, and the United Kingdom (Philippe & von Hippel, 2016). The use of HEU by the US nuclear submarines
is B2.5 tons which is B60% of global navel HEU and sufficient for manufacturing 100 nuclear weapons every year.
This may be slow due to the long experience particularly of the US Navy, with over 60 years of operational experience,
30 different designs and a good safety record, the conversion of HEU cores to LEU cores would be inevitable just as in
the case of research reactors. The technical difficulties that would need to be surmounted would be the decrease in
power of LEU cores of similar size to the HEU cores; this would indeed require design changes in the space allocations
for submarine reactors.
As the lifespan of older submarines has come to an end, several have been decommissioned and dismantled in elab-
orate and integrated programs. Out of B465 nuclear submarines built in the world, over 300 have most likely been
decommissioned and largely dismantled. It has been reported that in the dismantling effort, in most cases “20
45 per-
cent uranium-235 has been used as fuel. . .,” that all but one US submarine has one reactor while most Russian submar-
ines had two reactors, that the “cores of these reactors typically hold between 200 and 300 fuel assemblies each
containing up to a few tons of fuel rods,” and that the thermal power varies from 10 MW in older submarines to
200 MW in newer classes of submarines (Kopte, 1997). There have been serious issues and concerns in the dismantling
process due to the radiation levels of the spent fuel, and other liquid and solid radioactive waste, and the safety of pro-
cedures as well as the waste disposal procedures, and long-term ecological effects, in land sites or in deep oceans.

3.4 Plutonium production reactors

The term “production” for a reactor represents the production of plutonium; since plutonium does not occur in nature
(except for in trace amounts in naturally occuring uranium ores), it must be produced. Plutonium is more fissile than
uranium-235 (Chapter 1) resulting in smaller-sized weapons. In nuclear power reactors, plutonium is produced during
normal operation from the transmutation of uranium-238. Typically, 5 kg of Pu-239 and 2 kg are produced ton of fuel
in a 1000 MW(e) PWR over a period of about 3 years. It is separated from the spent fuel in a reactor by reprocessing
methods. Some of its main isotopes are Pu-238, Pu-239 (fissile), Pu-240 (fertile), Pu-241 (fissile), and Pu-242.
The purpose of production reactors has been to produce weapons grade plutonium (Pu-239 with less than 8% Pu-240).
As discussed in the previous section, the NRX reactor with natural uranium and heavy water, and Magnox reactors
with natural uranium metal, graphite, and CO2 were production reactors. Similarly, breeder reactors, fast or thermal,
have been utilized for breeding plutonium in a fast reactor such as the Liquid Metal Fast Breeder Reactor (LMFBR),
the gas cooled fast reactor (GCFR) and the Lead Cooled Fast Reactor (LFR).
The present reactor grade plutonium (55%
70% Pu-239, 30%
35% nonfissile Pu, and more than B19% Pu-240)
production from power reactors in the world is about 70 tons annually. It is estimated that there are hundreds of tons of
weapons grade plutonium in the world. This is clearly a tremendous radiation and proliferation hazard and a major area
of concern for the public acceptance of the nuclear industry. Similarly, thermal breeders can be used with thorium 232
to breed the fissile fuel uranium 233.
 Nuclear reactors and systems Chapter | 3 121

The EBR-I built at Idaho (USA) became the first plant to produce electricity on the December 20, 1951 and to dem-
onstrate in 1953 that a reactor could breed fissile fuel. The EBR-I reactor operated for 12 years but was shut down in
1963. EBR-I was followed by the SFR EBR-II which achieved criticality in 1965, produced 62.5 MWt/20 MWe with
65% enriched uranium fuel and demonstrated several safety systems and procedures for a HEU reactor. EBR-II was
shut down after about 30 years operations in 1994. A similar breeder reactor Phoenix was built and commissioned in
France in December 1973 and operated at 590 MW(t)/233 MW(e) demonstrating a breeder ratio 1.16. The plant was
shut down in 2009. The reasons for closing down breeders were most probably technical difficulties such as their shut-
down for refueling resulting in low capacity factors, and safety and proliferation concerns. Two power reactors, similar
to breeders operating in Russia are BN-600 and BN-800 which are discussed in the next section.
In thermal reactors, plutonium is produced as a by-product from the radiative capture of uranium-238 in the fuel.
It is separated by reprocessing spent fuel and is then fed again into reactors as a mixed oxide fuel (MOX). The amount
of plutonium produced in a 1000 MW(e) LWR with B25 tons spent fuel per fuel is B290 kg, of which more than half
is the fissile plutonium 239. This fraction is higher for CANDU and Magnox reactors; for the former with low burnup,
it can be B80%.
The early-day production reactors in the US have been Hanford and Savannah River estimated to have produced
103.4 tons of plutonium in the 50 years from 1944 to 1994. In Russia, according to IAEA (IAEA TECDOC-1591) esti-
mates the spent fuel reprocessed till 2002 exceeded 300,000 metric tons. Today’s stockpiles of plutonium are “sufficient
for tens of thousands of nuclear weapons” https://vcdnp.org/plutonium-stockpiles-causes-and-solutions/ and efforts are
underway in several countries to limit their production and seek sustainable solutions.

3.5 Small modular reactors

As described in Section 3.1, a SMR is defined by the IAEA as a NPP with a capacity in the range 100
700 MW(e)
(Bari, 2014; IAEA, 2018).
As of December 2019 there were 97 small and medium operational reactors with gross capacity less than 700 MW
out of 443 operational reactors worldwide (IAEA, 2020, p. 65). The rationale to develop SMRs includes specialized
applications, such as powering of production facilities, desalination, hydrogen production, district heating, reaching off-
grid remote locations, and providing incremental demands to large nuclear reactors.
The technology of SMRs is mostly well-known and derived from the earlier designs of the 1970s and 1980s. Out of
the 45 SMR designs being pursued, most are in the conceptual or detailed design phase and none of the innovative SMR
designs are commercially available. Some designs (e.g., NuScale 50 MW/module; mPower 125 MW; Westinghouse
200 MW; PRISM 311 MW; SMART 100 MW; 4S 10 MW; HTR-10 10 MW; HTR-PM 250 MW) are moving toward
licensing. The USNRC issued a Standard Design Approval to NuScale in September 2020 https://www.nuscalepower.
com/about-us/faq#ES1 making it possible to move ahead with the development of small plants such as the 77 MW(e)
NuScale Power Module which can be integrated into a 12 module plant with an output of 924 MW(e).

3.5.1 Design features of SMRs

The design features of SMRs, some of which are listed for land-based reactors in Table 3.11 and for marine-based and
other designs in Table 3.12, are based on innovative designs with scaling-down of large NPPs essentially integrating all
primary components inside a RV resulting in a small size, in addition to passive safety systems. The water cooled
SMRs are limited to outlet core temperatures of the 300 C range as compared to the high temperature reactors with
enhanced efficiencies in Table 3.12.
The Westinghouse lead fast reactor (LFR), listed in Table 3.12 and shown in Fig. 3.8, with a power capacity of
.450 MW(e), is classified as a medium-output modular, passively-safe plant uses MOX fuel. It is cooled by lead
which has a high boiling point of 174 C. It has a fast spectrum with high burnup core and a high operating temperature.
It has innovative design features that include the use of advanced materials to enable high temperature operation leading
to good thermodynamic efficiency; of the order of 40%
50% at temperatures in the range 600 C
700 C.
The safety features of the WLFR include DC power capacity for over 72 hours and no events for which immediate
operator action is needed. The integral design eliminates concerns of LOCAs. Lead has better performance than that
observed with sodium. Reactor shutdown is achievable with automatically-actuated passive cooling systems, without
the need for instrumentation and control signals or moving parts and, with lessons learned from Fukushima, passive
heat removal of decay heat is possible by radiation transfer between the RV and the guard vessel (GV) as well as by
the use of naturally
circulating air. Another advantage is the use of smaller size turbomachinery as no condenser is
TABLE 3.11 Overall parameters of some water cooled SMRs.

No. SMR Power Design En. % Fuel Core In P RPV H/D Burnup Life Design Primary
MW t/ /Out  C (MPa) (m) GWD/t (y) Circ.
MWe
1 CAREM A 100/B30 Int. PWR 3.1 UO2 284/326 12.25 11/3.2 24 40 UCP NC/PSC
2 ACP100 C 385/125 -do- ,4.95 -do- 286/319.5 15 10/3.5 ,52 60 BDF FC
3 CAP200 C 660/ . 200 4.2 -do- 289/313 15.5 8.845/3.280 37 60 CDF FC
4 DHR C 400/none Pool type ,5.0 -do- 68/98 30 60 BDF FC
5 IRIS I 1000/335 Int. PWR 4.95 -do- 92/330 15.5 21.3/6.2 65 (max) 60 BDF FC
6 DMS J 840/ 300 BWR ,5 -do- 186/287 7.17 15/4.8 , 60 60 BDF NC
7 IMR J 1000/350 Int. PWR 4.8 -do- 329/345 15.51 17/6 . 40 60 CDF NC
8 SMART K 330/100 Int. PWR ,5 -do- 296/323 15 15.5/6.5 ,60 60 LC FC
9 ELENA R 3.3/0.068 PWR 15.2 -do- 311/328 19.6 3.7/1.25 57600/273900 - CDF NC
10 KARAT-45 R 180/45
50 BWR 4.5 -do- 180/286 7 11.5/3.10 45.9 80 CDF NC
11 KARAT-100 R 360/100 BWR 4 -do- 104/286 7 13.25/4 45.9 80 CDF NC
12 RITM-200 R 175/50 Int. PWR ,20 -do- 277/313 15.7 8.5/3.3 - 60 UD FC
13 RUTA-70 R 70/NA Pool type 3 -do- 75/102 atm 17.25/3.2 25
30 60 CDF FC
14 UNITHERM R 30/6.6 PWR 19.75 -do- 249/330 16.5 9.8/2.9 1.15 25 CDF NC
15 VK-300 R 750/250 SP BWR 4 -do- 190/285 6.9 13.1/4.535 41.4 60 DD NC
16 UK SMR UK 1276/443 PWR ,4.95 -do- 296/327 15.5 11.3/4.5 55
60 60 MC FC
17 mPower U 575/195 Int. PWR ,5 -do- 290.5/318.9 14.8 27.4/4.15 ,40 60 UD FC
18 NuScale U 160/50 Int. PWR ,4.95 -do- 258/314 12.8 17.8/3.0 .30 60 URR NC
19 SMR-160 U 525/160 PWR 4.95 -do- 209/321 15.5 15/3 45 80 PD NC
20 WSMR U 800/ . 225 Int. PWR ,5 -do- 294/324 15.5 28/3.7 .62 30 CDF FC

A, Argentina; C, China, IIC-Iris International Consortium; J, Japan; K, Republic of Korea; RF, Russian Federation; U, America.
Design status: UCP, under construction as prototype; BDF, basic design finished; LC, license certified; UD, under development; DD, detailed design; MC, mature concept; URR, under regulatory review; PD, preliminary
design in progress; Primary circulation: NC PSC, natural circulation, pressure suppression containment; FC, forced circulation.
Source: Advances in Small Modular Reactor Technology Developments, A Supplement to: IAEA Advanced Reactors Information System (ARIS), 2018 Edition.
TABLE 3.12 SMRSs: marine-based LWRs, HTGCR, Molten Salt, eVinci.
No. SMR Power MW Th/e Design En. % Fuel Core In/ P (MPa) RPV H/D B Life (y) Design Primary
Out,  C (m) GWD/t Circ.

1 ACPR50S C 200/50 Loop type PWR ,5 UO2 299.3/321.8 15.5 7.2/2.2 ,52 40 CDF FC
2 KLT-40S R 150/35 PWR 18.6 -do- 280/316 12.7 4.8/2 45.4 40 UT FC
3 RITM-200M R 175/50 Int. PWR ,20 -do- 277/313 15.7 9.2/3.5 - 60 UD FC
4 VBER-300 R 917/325 Int. PWR 4.95 -do- 292/328 16.3 9.3/3.9 50 60 LS FC

Gas cooled

1 HTR-PM C 2x250/210 GCR 8.5 Sph. 250/750 7 25/5.7 90 40 UC FC

(inner)
2 GTHTR300 J ,600/100B300 Prismatic HTGR 14 UO2 TRISO 587
633/ 7 23/8 120 60 BDF FC
850
950
3 GT-MHR R 600/288 Mod. He LEU or WPu Coated 490/850 7.2 29/8.2 100
720 60 PDF FC
particle fuel
4 PBMR-400 S 400/165 high 9.6% Pebble bed with 500/900 9 30/6.2 92 40 PD FC
temperature coated particle fuel
GCR

Liquid Metal Cooled

1 4S Toshiba J 30/10 (pool type) , 20 (U-Zr alloy) 355/510 atm 24/3.5 34 60 DD FC

2 BREST-OD-300 R 700/300 B13.5 U-Pu /N 420/535 Low pressure 17.5/26 61.45 30 DDS FC
3 LFR-TL-X N 15/5; 30/10; 60/20 (pool type) 19.75 LEU, 360/420 atm 3.5/2 40 30 CD FC
cylindrical cassette
4 WLFR a
U 950/ . 450 Pool type # 19.75 Oxide, with 420/600 Nearly atm 8.0/7.5 $ 100 60
100 CD FC
provision for
transition to UN

Molten Salt

1 IMSRb Ca 400/190 MSR ,5 Molten salt fuel 625
660/ , 0.4 7.0/3.6 - 60 CDF FC
670
700
2 CMSRc D 250/100
115 MSR (U 1.1%, Pu 69% 600/700 or 1 2.5/2.1 250 60 CD FC
fissile) Sodium- 700/900
actinide fluoride
3 CAWBd D 50/20 MSR _ LiF-ThF4 _ _ _ _ 3
5 CD FC
4 ThorCone I 557/250 Th. MSR 19.7 LiF-ThF4 565/704 1.2 12/8 220 80 CBD FC
1 eVinci U 0.6
4
40/0.2
15 Monolithic core 19.5 UO2 or UN NA NA NA ,10 10 TRL 5 HP
HP

UT, under test; LS, licensing stage; PDF, preliminary design completed; S, South Africa; DDS, detailed design startup in early 2023; P, heat pipe; D, Denmark; Ca, Canada.
a
Westinghouse Lead Fast Reactor.
b
Integral Salt Molten Reactor, Canada.
c
Compact Molten Salt Reactor, Seaborg Technologies, Canada.
d
Copenhagen Atomics Waste Burner, Denmark.
e
ThorCon International, Indonesia, 557 mWth/250 MWe (per module).
Source: Advances in Small Modular Reactor Technology Developments, A Supplement to: IAEA Advanced Reactors Information System (ARIS), 2018 Edition.
124 Nuclear Engineering

FIGURE 3.8 Westinghouse liquid cooled fast reactor (WLFR). Courtesy of Westinghouse Electric Company (c) 2021. All rights reserved.

TABLE 3.13 Physical design parameters of LFR and Evinci Westinghouse designs.

Core WLFRa Evincib

Reactor type Lead cooled fast reactor Monolith core with heat pipes
Power (MW(th)/MW 950/ . 460 0.6
4
40/0.2
15
(e))
RPV height/diameter Approx. 8.0/7.5 N/A
(m)
In-house plant B15 MWe N/A
consumption
Fuel Oxide (UO2 or MOX) UO2 or UN
Coolant Pb Heat pipes/metal hydride moderator
Core discharge $ 100 MWd/kgHM , 10 GWd/ton for a 1.2 MWth reactor
burnup
Fuel material UO2, MOX then UN UO2, UN
Fuel enrichment # 19.75% 19.5
Fuel cycle (months) 24 120
Main reactivity Rods Ex-core control drums
control
Approach to Passive, inherent Passive, inherent
engineering safety IAEA passive safety category B goal IAEA passive safety category B goal
Design life (years) 60 for components, 100 for structures 10
Plant footprint (m2) N/A , 1500
Distinguishing Low pressure reactor vessel; high BP coolant, no Transportable reactor (vSMR) operating
features moving parts or actuation for decay heat semiautonomously, no moving parts or actuation for
removal; high efficiency decay heat removal; high efficiency
Design status Conceptual design TRL-5 (Technology Readiness Level 5 with USNRC and
CNSC) Canadian Nuclear Safety Commission
a
https://www.westinghousenuclear.com/new-plants/lead-cooled-fast-reactor.
b
https://www.westinghousenuclear.com/new-plants/evinci-micro-reactor.

required with supercritical CO2 in the power conversion system. The Westinghouse LFR is anticipated to start construc-
tion of full-scale prototype by B2030 and convert to a first of a kind (FOAK) commercial unit by about 2035.
Another new very small, portable and advanced Westinghouse design eVinci (Table 3.13) is modular, transportable,
factory-built, fueled and assembled configuration with installation time of less than 30 days. With minimal moving
 Nuclear reactors and systems Chapter | 3 125

parts and high reliability, it has the advantages of a small footprint and a near-zero Emergency Planning Zone (EPZ).
Some overall design parameters of the WLFR and eVinci designs are compared in Table 3.13.
A step forward for eVinci was its selection, on the March 10, 2021 by the Department of Energy (US DOE) with a
DOE funding of BUS$ 12.9 million out of a total value of BUS$ 28.6 million, to “prepare for the Nuclear
Demonstration Unit (NDU) through design, analysis, testing and licensing to manufacture, site and test the NDU
by 2022” Source: SMR-Book, IAEA (Status of eVinci on March 10, 2021, https://www.energy.gov/ne/articles/us-
department-energy-further-advances-nuclear-energy-technology-through-industry-awards).
The SMRs at various stages are: one mature concept, six basic design, four preliminary design, 24 conceptual
design, five under development, four detailed design, one final design, one certified design, one licensing stage, three
under construction (CAREM, KLT-40S, HTR-PM). Both NuScale and eVinci are set to progress for the commercial
markets.

3.5.2 Very small modular reactor

At the low power level, a very small modular reactor (vSMR) such as the Toshiba 4S (Super Safe Small and Simple)
10 MW(e) reactor (Table 3.13) with active core height: 2.5 m, and diameter 1.16 m has a small core size with. 1.69 t
fissile U-235 inventory, out of 9.23 t U, has high burnup (average 34 GWd/t) and low source term for radiation release
in case of an accident.
Both the 4S designs (30 and 50 MW) are small and address Gen-IV requirements including nonproliferation, essen-
tially due to the absence of blanket breeders typically incorporated in fast reactors, and passive safety in LEU reactor
systems. One of the most attractive features of the 4S reactor is that refueling is not required in the entire 30-year
operation.
Key design and technology issues of SMRs are based on their inherent features which include:
G Small factory-built systems leading to reduction in on-site work
G A single (integrated) pressurized vessel containing steam generator so that a large break LOCA is eliminated from
DBAs
G A small core will have less risk due to less accident radioactive release
G Passive safety systems, for example, natural gravity-driven circulation in LWR SMRs
The above factors lead to reductions in the EPZ from 5 to 25 km, suggested by IAEA for 100
1000 MWth reactors,
to possibly 1000 ft as listed by Babcock and Wilcox for its mPower reactor. SMRs can thus be shifted closer to the
location requiring power in contrast to present-day NPPs. Thus resilient and isolated micro grids of high reliability can
be set up.
As mentioned in Section 3.4, one issue of concern for SMRs is the amount of plutonium production which over the
30 year operation period of 4S could be as high as 11 kg after 2 years, 82 kg after 15 years and 159 kg after 30 years.
Unfortunately, this is a negative aspect of a SMR of the 4S type which means that proliferation resistance is an issue
that would require careful security standards and practices.

3.5.3 Generation-IV reactors

With increased global competitiveness from conventional fossil-based and renewable energy systems, the nuclear indus-
try finds itself in a position in which it has to offer compact, reliable, flexible and safe systems addressing additional
public concerns on the long-term disposal of nuclear waste and proliferation of sensitive nuclear material. A nuclear
renaissance depends on the ability of the nuclear industry to meet the challenges of free markets. As discussed in
Section 3.1.3, the present developments indicate growth in Asia and a down-swing in Europe and the United States.
To address these challenges, new reactor designs have been considered as Generation-IV reactors for the coming
years. These are “new” in the sense of having improved designs and technologies on old systems.
Essentially, the capability of nuclear reactors to respond to natural and operational events, potentially leading to
accident conditions, has been increased. The lessons learned from over 60 years of nuclear power generation with three
accidents (TMI, Chernobyl, Fukushima) have contributed to design changes. These have led to enhanced performance
objectives with improved safety, passive and self-actuated shutdown systems, better decay heat remove, and severe
accident prevention.
The overall objectives translate into a shift from water systems to nonwater systems with higher operational tem-
peratures and higher power densities.
126 Nuclear Engineering

The Generation-IV International Forum (GIF), created in January 2000 to study potential systems, consists of four-
teen partners (Argentina, Australia, Canada, People’s Republic of China, Euratom, France, Japan, Republic of Korea,
Russian Federation, Republic of South Africa, Switzerland, United Kingdom, the United States and Brazil as a nonac-
tive member). This Forum has selected six reactor technologies “to support the next generation of innovative nuclear
systems.” The systems are
1. Gas Cooled Fast Reactor (GCFR)
2. Lead Cooled Fast Reactor (LFR)
3. Molten Salt Reactor (MSR)
4. Supercritical Water Cooled Reactor (SCWR)
5. Sodium Cooled Fast Reactor (SFR)
6. Very High Temperature Reactor (VHTR)
These systems, as stated above, were selected on the basis of their enhanced safety, improved economics and new
products such as hydrogen for transport applications, reduced nuclear waste and increased proliferation resistance.
While Gen-III 1 systems were evolutionary, Gen-IV systems are revolutionary in terms of their energy conversion sys-
tems as well as the nuclear fuel cycle technologies.
Some highlights from the Technology Roadmap Update for Generation-IV Nuclear Energy Systems, (Nuclear
Energy Agency [NEA], 2014), are:
1. the importance given to fast reactors is mainly due to their breeding capability reducing the magnitude of waste
disposal,
2. the development of advanced separation technologies will avoid separation of sensitive fissile material for any unde-
sirable diversion,
3. significant resources dedicated to the development of the SFR and VHTR, are due to the considerable historical
effort associated with these technologies,
4. materials testing and corrosion, and lead chemistry, core instrumentation, fuel development (MOX for first core,
then minor actinide MA-bearing fuels); and possibly nitride fuel for lead cooled reactors will be carried out with the
Russian reactor BREST-OD-300 as the reference design, and that,
5. efforts will be dedicated toward improving safety and operation with advanced instrumentation and control systems,
prevention and mitigation of sodium fires, and the overall prevention and mitigation of severe accidents with poten-
tial of large energy releases.
For the 2400 MW(t) Gas Cooled Fast Reactor, the 75 MW(t) two-loop He/water ALLEGRO experimental demon-
strator in Czech Republic is a test-bed to demonstrate the viability of the thermal tolerance of (U-Pu)C fuel, the func-
tionality of the He/gas heat exchangers and of safety systems. Based on utilizing the advantages of both fast and high
temperature reactors, ALLEGRO also tests the utilization of process gas for hydrogen production.
Similarly, the Pb-Bi-cooled Fast Reactor (LFR) SVBR-100 test-bed BREST-300 in Russia will carry out tests and
study functionality of components.
The MSR is an old concept from the 1950s to 1970s; at that time they were mainly thermal-neutron-spectrum
graphite-moderated reactors but since 2005 the liquid-fueled MSR research and development efforts have focused on
fast spectrum designs.
For the SCWR two baseline concepts are the pressure-vessel-based and pressure-tube-based LWR designs operating
at high temperature and high pressure above the thermodynamic critical point of water (374 C, 22.1 MPa). The neutron
spectrum depends on the core design.
For the SFR, there is an extensive body of knowledge operational reactors operational, for example, in China, India,
Japan and Russia. The SFR uses liquid sodium as the reactor coolant, at low pressure and high outlet temperature
500
550 C. Much of the basic technology for the SFR has been established in former fast reactor programs and was
further confirmed by the EBR-I and EBR-II in the US, from Phe´nix in France, the BN-600 reactor in Russia and the
Monju fast reactor in Japan.
The VHTR is a graphite-moderated, helium-cooled thermal reactor with high core exit temperatures, in the range
700 C
950 C. In the Japanese HTTR, the reactor core is a prismatic-block type such as the Japanese HTTR, while the
Chinese HTR-10 is of a pebble bed type. For these reactors, TRISO fuel is used at high operating temperature (up to
1250 C) giving high burnups (up to 200 GWd/tHM).
Generally SMRs and Gen-IV reactors can be classified based on their core exit temperature; a temperature of
B300 C is a water cooled SMR, B500
600 C is a liquid cooled fast reactor, B600
700 C is a MSR, and greater
 Nuclear reactors and systems Chapter | 3 127

than B700 C is a gas cooled fast reactor. As a rule of thumb, the higher the exit temperature, the higher will be the
thermal efficiency of the reactor power plant.

3.5.4 Radiation source term

Although the nuclear industry has generally had a good record in terms of radiation release to the environment, with the
exceptions of TMI, Chernobyl and Fukushima, the dose in the vicinity of a reactor is a public concern. A person living
within 80 km from a NPP would have an annual dose of about 400 mrem (4 mSv), out of which half would be from
radon.
In case of an accident at a plant with a release of radiation in the environment, the “source term” estimates for a
power plant would be mainly strontium, iodine and xenon. For a 300 MW(e), the inventory of fission products is given
in Table 3.14.
Source: CHASNUPP (p. 67 Zia Mian).
In the case of a core melt, for rupture of cladding, the assumed release fraction from fuel is generally assumed to be
1% for xenon, krypton, cesium and iodine. Gen-IV reactors such as the Lead Cooled Fast Reactor lead will have better
shielding to retain iodine and cesium at temperatures up to 600 C, thereby reducing the source term in case of release
of volatile fission products from the fuel.

TABLE 3.14 Source term for a 300 MW(e) PWR.

Fission product Inventory (Ci/MW(e))

Kr-85 560
Kr-85m 24,000
Kr-87 47,000
Kr-88 68,000
Sr-89 94,000
Sr-90 3700
Sr-91 110,000
Y-91 120,000
Mo-99 160,000
Ru-103 110,000
Ru-106 25,000
Te-129m 5300
Te-131m 13,000
Te-132 120,000
Sb-127 6100
Sb-129 33,000
1
131 85,000
1
132 120,000
1
133 170,000
1
134 190,000
1
135 150,000
Xe-131m 1000
Xe-133 170,000
Xe-133m 6000
Xe-135 34,000
Xe-138 170,000
Cs-134 7500
Cs-136 3000
Cs-137 4700
Ba-140 160,000
La-140 160,000
Ce-144 85,000
Np-239 1.64 3 106

Source: CHASNUPP (p. 67 Zia Mian).

128 Nuclear Engineering

3.6 Nuclear fusion

Our planet earth revolves around the sun, at the center of our solar system, which is our source of light, heat and gravi-
tation. The sun is a massive star by our standards, but otherwise of moderate size in comparison with other stars, about
50 times larger than the earth consisting of over 90% (atomic) hydrogen and about 9% helium, with a core temperature
understood to be over 15 million  C and a surface temperature of B5500 C. At its core, it undergoes nuclear fusion
reactions where hydrogen is converted into helium. Clearly, when all the hydrogen has been consumed in such reac-
tions, the life of the sun will have come to an end and it would no longer be capable of sustaining its role in the solar
system.
The big question for humans is whether nuclear fusion, of the same type happening in the sun and stars, can be rep-
licated on earth in our scientific laboratories. Can such high temperatures be maintained; do we have materials strong
enough to withstand such excessive heat loads, and can this energy be harnessed to produce electricity? If all these
questions can be addressed, then we have a virtually inexhaustible source of energy since hydrogen and its isotopes
occur in the vast oceans of earth.
First, we need to understand how the sun undergoes nuclear fusion reactions. Before it became a star, the sun was a
collection of hydrogen and helium gases which gained density and formed gravity. The gravity pulled nuclei into a core
and became heavy enough to form into a star. As the nuclei came close together, they fused and reactions were initi-
ated, that continue today.
On earth, we first need the fusion fuels: hydrogen and its isotopes deuterium and tritium. Hydrogen is available in
(light) water in the form of H2O while deuterium is available in the ratio 1:8000 in sea water. Tritium does not occur in
nature and has to be produced from nuclear reactions such as the bombardment of lithium by neutrons.

3.6.1 The fusion reaction

From the mass defect and binding energy it is evident that energy can be produced from nuclear reactions in two cases
(Section 1.3); for heavy nuclei breaking into smaller nuclei, called nuclear fission, and for light nuclei combining or
fusing into heavier nuclei, called nuclear fusion.
Fig. 3.9 illustrates the nuclear fusion reaction of two isotopes of hydrogen namely deuterium and tritium. The fusion
reaction results in the production of alpha particles (helium) and a neutron
2
1H 1 31 H-42 He 1 10 n ð17:6MeVÞ
The energy released in this reaction, 17.6 MeV; the alpha particle carries 3.6 MeV which is deposited “locally” due
to the relatively short range of an alpha particle in matter (Section 1.1) while the neutron carries 14 MeV which can
escape from the system as neutrons have a relatively larger range (Section 2.1). The physics of nuclear fusion is
reviewed in Annex.
The figure of merit is the Lawson criterion, which for DT fusion is
nτ E $ 1:5 3 1020 m23 s
The goal of achieving nuclear fusion by plasma confinement is to achieve breakeven by meeting this criterion.
Of the three confinement schemes namely gravitational confinement, magnetic confinement and inertial confine-
ment, the latter two have been investigated in scientific laboratories. Gravitational confinement is an appropriate
description of nuclear fusion that takes place in the sun where very large confinement times are evident. Here, on earth
it is only possible to try magnetic confinement fusion (MCF) and inertial confinement fusion (ICF) for which, so far, it
has not been possible to achieve even the short confinement times which would enable nuclear fusion reactors to oper-
ate in the way that nuclear fission reactors can for years.
To further appreciate the technological challenges of advanced nuclear systems, and before that, the modeling and
simulation efforts necessary, both MCF and ICF are briefly reviewed.

FIGURE 3.9 DT nuclear fusion reaction.

 Nuclear reactors and systems Chapter | 3 129

3.6.2 Magnetic confinement fusion

When a plasma, that is, free ions and electrons, is in a magnetic field, the charged particles experience a Lorentz force,
which causes them to gyrate about a magnetic field line. The force is proportional to the magnetic field, and the radius
of gyration is directly proportional to the perpendicular velocity (perpendicular to the magnetic field) and inversely pro-
portional to the applied field. A plasma could be pinched by the application of a strong electric field too. Thus a strong
field leads to a gyration of very small radius and a charged particle is confined about the field. Both electrons and ions
move in circular orbits about a field line. Due to their longitudinal speed they drift away from their orbit. One of the
first attempts of magnetic confinement were based on the pinch and mirror concepts; thus in a magnetic mirror, in
which both ends had strong field lines would bounce the plasma back and forth between the two ends. The concept of
mirrors was one of the earliest for confinement schemes from the 1960s; the Mirror Test Facility at Lawrence
Livermore Laboratory was designed in the 1970s but research work by then had shifted to several other devices such as
the stellerator and the Soviet Tokamak design. By the 1980s, the Joint European Torus (JET) at Culham, England, and
the Tokamak Fusion Test Reactor at Princeton as well as other tokamaks around the world were operated to demon-
strate scientific feasibility.
Currently the Tokamak project is the leading effort with work underway on the International Thermonuclear
Experimental Reactor (ITER), illustrated in Fig. 3.10, an international project with 35 states, to demonstrate scientific
and technological feasibility of fusion.
The basic principle of plasma confinement is the use of magnetic coils placed around a vacuum vessel in the shape
of a toroidal chamber. These produce toroidal and poloidal fields which close upon themselves. In the vacuum of the
chamber, the plasma is injected in the form of gaseous fuel. With the induced current, the gas is ionized into the plasma
state. The ions and electrons are confined along magnetic field lines and energized to fusion temperatures of up to 300
million  C.
The first plasma is scheduled to be introduced in ITER when its construction phase will be over by December 2025.
36Li 1 01n-24He 1 13T 1 4:8MeV
37Li 1 01n-24He 1 13T 1 01n 2 2:46MeV
A schematic of a nuclear fusion power reactor is shown in Fig. 3.11; the toroidal chamber where the hot plasma
undergoes fusion temperatures at few hundreds of million  C is the source of thermal energy. The core is surrounded
by a lithium blanket where tritium breeding is achieved by the 14 MeV neutrons produced from the DT reaction. The

FIGURE 3.10 The ITER Tokamak for

magnetic fusion. Courtesy ITER @iter.org
https://www.iter.org/newsline/-/3037.
130 Nuclear Engineering

FIGURE 3.11 Schematic of a

nuclear fusion power plant.

thermal energy is carried out of the reactor into a conventional power conversion system very similar to that described
in previous sections for nuclear fission reactors.
The fusion reactor, though a power plant, has a virtually inexhaustible fuel supply which comes from water in the
form of deuterium; tritium is bred within the reactor. There is no fissile fuel or radiation hazard with the exception of
some neutron and gamma radiation, but no safety, waste disposal, or proliferation issues. It is thus the ultimate energy
source which would most probably not find any competitor.

3.6.3 Inertial confinement fusion

Fusion can also take place when a deuterium and tritium pellet is heated to very high temperatures, as illustrated in
Fig. 3.12 for direct- and indirect-drive mechanisms, and held together merely by its inertia for sufficient time to achieve
Lawson’s criterion nτ E $ 1:5 3 1020 m23 s. In this approach, called inertial confinement fusion, the high temperatures
are obtained by intense heating of a tiny pellet to burn its surface, causing ablation, or evaporation, of a thin layer which
expands and, by newton’s third law of motion, causes an inward moving shock wave. Thus the outwards explosion
results in an inwards implosion which, if carefully shaped to be symmetric, results in the compression of a spherical tar-
get and subsequent heating and confinement to enable fusion.
The dynamics of ICF, in the context of external heating, requires a slightly different form of the Lawson criterion to
model the energy and its transport.
Since ICF involves compression, the figure of merit is based on the density ρ 5 mn of the imploding material and
the speed of the imploding material
rffiffiffiffiffiffi
kT
v5
mi
The time scale would then be, in terms of the pellet radius R
R
τEB
v
and the Lawson criterion could be written as
T 3=2
ρRBC
hσvi
where
12 k3=2 mi 1=2
C
Ec
With the appropriate “best values” the Lawson criterion is expressed as
ρr . 3g=cm2
 Nuclear reactors and systems Chapter | 3 131

FIGURE 3.12 Laser compression

of a pellet by direct- and indirect-
drive.

FIGURE 3.13 Stages of implosion in a

direct-drive laser ICF.

ÐR
The confinement properties depend on the areal density of the compressed core, ρR 5 0 ρðr Þdr
 2:2
T
ρR .1
4:7
Research into ICF has been going on for decades with significant investments in facilities for delivering large amounts
of laser energy (BMJ) into tiny DT pellets (B1 mm diameter) over very short periods of time driving imploding DT
shells up to over 360 km/h compressing it to initiate ignition in a “hot spot” at the center as illustrated in Fig. 3.13.
The idea is to hold this hot spot for sufficient time to cause a thermonuclear burn wave in its vicinity which is sur-
rounded with cold and dense material and still collapsing. As the hot plasm burns, its pressure and hence its expansive
force begins to increase; it competes with the imploding surrounding until the pressure is so much that the entire assembly
explodes; whatever fusion energy is produced is that between the ignition of the hot spot and disassembly. These types of
experiments comprise direct-drive ignition and have the best chance for succeeding for spherical pellets. There are several
other “requirements” to ensure success, such as slow isentropic compression and perfect implosion symmetry.
The gain G is expressed as
Eout
G5
Ein
where fusion output Eout is achieved typically from a high-powered laser providing the input energy Ein . The gain value
of the JET has been B0.67 while ITER aims for a value of 10 producing 500 MW(e) for an input of 50 MW(e).
132 Nuclear Engineering

Research efforts in the US (Betti & Hurricane, 2016), Europe (Tikhonchuk, 2020), Japan (Horioka, 2018), China
(He, 2016; He & Zhang, 2007) and several other countries, are focused on achieving larger values for the Lawson
criterion.
The best-performing indirect-drive implosions with a 1.9 MJ laser have “accelerated the DT mass to about
360
380 km/s, reaching a fuel kinetic energy of about 12 kJ and producing about 26 kJ of fusion energy” (Betti &
Hurricane, 2016).
Some of today’s challenges in ICF are the development of advanced schemes to reduce losses and mitigate instabil-
ities, and improve implosion technologies.

3.7 Space propulsion

Space travel must have fascinated humans from times immemorial; history records the proposal for rockets over 100
years ago by the Russian theoretician Konstantin Ziolkovsky. The then Soviet work, finding its way to Europe, culmi-
nated in Germany in the 1930s. There, under the leadership of Werner von Braun, the German rocket program produced
the first rocket to fly into space, the V-2. Since the 1960s spacecraft with nuclear heaters and power systems have been
operating
Space propulsion can be achieved by a number of routes including chemical combustion, nuclear thermal energy to
a propellant and direct nuclear fission and nuclear fusion energy. Liquid and solid propellants based on combustion are
most common for missile and space systems from the first ballistic missiles to NASA’s 2020 Mission to Mars. Before
the present mission, the most notable has been the Apollo landing of three US astronauts on the Moon in 1969. These
were followed by the Viking, Pioneer, Voyager, Galileo, Ulysses, and Cassini missions. The first use of Pu-238 as a
source for radioisotope power was for Pioneer 10.

3.7.1 Conventional rocket designs

Ballistic missiles and rockets today are based on the German V-2 Vergeltungswaffe-2 rocket, technical name A-4 shown
in Fig. 3.14 designed by the group led by Wehrner von Braun and used in large quantities in the second world war. On
its first test flight on October 3, 1942, the V-2 flew 80 km high, and as such, was the first rocket ever to reach such a
height. The V-2 was a short-range ballistic, surface-to-surface, missile about 300 km, liquid-propelled missile that car-
ried a 1000 kg chemical warhead and was used in Europe against Allied targets. It was fired at an angle to reach its tar-
get as control systems were not very advanced in those days. Still, it had gyroscopes to determine angles and an
accelerometer for engine cutoff. It could reach altitudes of about 80 km during its powered flight; if fired vertically it
could go as high as about 200 km and achieve a maximum speed of 2880 km/h compared with 1100 km and
24,100 km/h for the most advanced long-range intercontinental ballistic missiles (ICBM), Minuteman III, developed in
1964
70 for the US land-based nuclear strategic force. The submarine launched ballistic missile (SLBM) Trident D-5,
with two other SLBMs Polaris (IRBM) and Poseidon (IRBM) is also an intercontinental submarine launched ballistic
missile (SLBM), deployed in 1990 with a unit cost of US$ 29.1 million, launched from SSBN submarines of the types
discussed in the previous section.
The overall design parameters of the V-2 rocket, up-scaled to the present Minuteman III and Trident D5 missile
designs is shown in Table 3.15. With the same basic design concept, there are differences in the size, the propellant and
casing materials, and in the thrust. Minuteman III has three rocket motors of length and diameter 18.6/5.5, 9.1/4.3 and
5.4/4.3 ft with thrusts 200,400, 60,700, and 34,500 lbf, respectively.
The Trident II D5, first deployed in 1990 is part of missile inventory on the US Ohio class and British Vanguard
class submarines also has a three-stage rocket motor system.

3.7.2 Space exploration

The V2 design was based on ethanol and liquid oxygen with ethanol having a higher heat of vaporization (B846 kJ/kg)
than kerosene (B251 kJ/kg) which means that ethanol would take longer to vaporize keeping the engine from melting.
These technologies were improved to design better performing systems.
From missiles to rockets, the thrust and size changes, for example, The Saturn V rocket that carried Apollo 11,
the first human mission to the Moon, generated 34.5 million Newtons (about 7.7 million lbf) thrust at launch
https://solarsystem.nasa.gov/news/382/10-things-rockets-we-love-saturn-v/ which is about twenty times the thrust of
Minuteman III.
 Nuclear reactors and systems Chapter | 3 133

FIGURE 3.14 The V-2 rocket. Courtesy NASA grc.nasa.

gov 2021.

Warhead (Explosive
charge)

Automatic gyro control

Guidebeam and radio

command receivers

Container for
alcohol-water
mixture

Container for
liquid oxygen

Container for
turbine propellant
(hydrogen peroxide)
Propellant turbopump

Vaporizer for turbine

propellant (propellant
turbopump drive) Steam
exhaust
Oxygen main
from turbine
valve

Rocket motor

Alcohol main
valve

Jet vane Air vane

German V-2 (A-4) Missile

134 Nuclear Engineering

TABLE 3.15 Physical design of V-2, Minuteman III and Trident D5 missiles.

Design parameter V-2 Minuteman III Trident D5

Classification SRBM ICBM ICBM (SLBM)
Range (km) 320 13,000 12,000
Length (m) 14 18.2 13.42
Diameter (m) 1.65 1.85 2.11
Fuel 75% ethanol, 25% water Three-stage Solid Three-stage solid
Oxidizer Liquid oxygen Combined in solid Combined in solid
Warhead chemical thermonuclear Thermonuclear MIRV
Launch weight (kg) B13,000 34,467 59.090
Specific impulse (s) 203
239 Varies by stage
Thrust (kN) 264.90 935
Control system Gyroscope/accelerometer INS/IGS INS, SRS

Source: V-2: http://www.worldwar2facts.org/v2-rocket-facts.html http://www.astronautix.com/v/v-2.html ICBM: https://missilethreat.csis.org/country/united-

states/.

The rocket, launched from Cape Kennedy on July 16, 1969 carrying three astronauts, Neil Armstrong, Michael
Collins, and Edwin Aldrin, landed on the Moon from where on July 20, 1969, Armstrong voice was heard with the his-
torical sentence “. . . on small step for a man, one giant leap for mankind.” The specific impulse of Saturn V with
Liquid Oxygen (LOX)/Rocket Propellant (RP) powered with Rocketdyne engines was 265 seconds for the first stage
and 424 seconds for the second and third stages with LOX/LH2.
Fifty years later, NASA’s mission to Mars with the Perseverance Rover was launched on a two-stage Atlas V-541
rocket from Cape Canaveral, Florida, on July 30, 2020 landed on February 18, 2021. Weighing 531,000 kg fully fueled
with the spacecraft on top, the 58 m high rocket fueled with type RP-1 kerosene fuel/ liquid oxygen, in the first stage,
and liquid hydrogen/liquid oxygen in the second stage Centaur RL-10 engine, can provide a thrust of 3.8 million
Newtons. The thrust-to-weight ratio of the system at B0.8 is better than that for previous space rockets.
The NASA Mars Mission 2020 to the Red Planet along the trajectory shown in Fig. 3.20 took 6 months and 19 days
to travel a distance of about 470 million km with an average speed of 98,000 km/h (B27 km/s)
The RD-180 engine, used in the Atlas V rocket of the Mars 2020 mission is a high performance engine LOX/kero-
sene engine built and marketed 50
50 by NPO Energomash (Russia) and Pratt and Whitney (USA). Some technical
features of the RD-180 engine, used in Atlas III and Atlas V launchers, are its specific impulse of 337.8 seconds in vac-
uum, thrust of 0.9 million lbf, and a chamber pressure of 257 atmospheres. The launch vehicle has stage 1 and 2 Atlas
V and Centaur engines with four solid rocket boosters; their combined thrust is over 5 million lbf. Before the 2020
Mars mission, Apollo missions 4, 6, 8 used 1.5 million lbf thrust while Apollo missions 9
17 used 1.522 million lbf
thrusts.
These are the limits of chemical propulsion systems, and thus define the limits of deep space travel with severe con-
straints of delivering payloads with reusable rockets.

3.7.3 Nuclear rocket designs for deep space exploration

To improve the capability of rockets to perform better than the chemical propulsion systems described above, there has
been an awareness for decades that nuclear propulsion may hold the key to the future of deep space exploration.
The work on nuclear rockets began in the 1950s with the Project Rover and the Nuclear Engine for Rocket Vehicle
Application (NERVA) shown in Fig. 3.15; design improvements and testing continued with development of state of the
art systems. Nuclear thermal propulsion (NTP) engines were demonstrated from the 1950s to 1970s during Project
Rover and later by the NERVA (Nikitaev & Thomas, 2020). Fuel is uranium nitride (UN) (supplies the heat) and pro-
pellant (that carries the heat) is hydrogen. Recently, efforts in deep space exploration have intensified with several mis-
sions to the Moon and the most recent (2020) missions to Mars (NASA, 2021).
The nuclear option for rockets is an attractive one due to its very high heating value compared with other fuels, as
shown in Table 3.16.
The values in Table 3.16 show that the nuclear advantage is orders of magnitude higher in comparison to chemical
fuels. Further, as concluded from an overview of Gen-IV reactors, high temperature systems would become available in
the near future, with liquid metal and GCRs ideal for safe and efficient power. With these advantages, a nuclear rocket
 Nuclear reactors and systems Chapter | 3 135

TABLE 3.16 Heating values of some conventional fuels compared with uranium fuels.

Fuel Heating value (MJ/kg)a

Hydrogen (H2) 120
142
Petrol/gasoline 44
46
Crude oil 42
47
Hard black coal 25
Natural uranium in LWR 500 GJ/kg
Uranium enriched to 3.5% in LWR 3900 GJ/kg
Natural uranium in FNR 28,000 GJ/kg
a
A Giga-Joule is a large amount of energy; the explosive yield of one ton of chemical explosive trinitrotoluene is TNT 4.18 GJ.
Source: World Nuclear Association https://www.world-nuclear.org/information-library/facts-and-figures/heat-values-of-various-fuels.aspx.

NOZZLE SKIRT EXTENSION FIGURE 3.15 NERVA rocket

INTERNAL engine. Courtesy NASA grc.nasa.
CONTROL NOZZLE
SHIELD gov 2021.
DRUM

REACTOR CORE
TURBOPUMPS
PROPELLANT LINE
EXTERNAL REFLECTOR
DISC SHIELD

FIGURE 3.16 The concept of a nuclear propulsion rocket system.

would essentially be used a thermal energy source for heating a propellant, as shown in Fig. 3.16 giving it high exhaust
velocity and thrust.
NTP offers an alternative to CPT since more thermal energy could be provided to propellants; this is likely to result
in higher nozzle exit velocities and hence higher thrust. If realized, NTP could reduce the travel time for outer space
exploration.
136 Nuclear Engineering

It is worth estimating the comparative advantage of NTP over CPT with regards to propulsion efficiency and spe-
cific impulse for deep space exploration and, specifically, the possibility of reducing flight time to Mars and then the
possibility of a return flight as well. In the present NASA mission, the rocket is used for launching the spacecraft carry-
ing the Perseverance Rover and descent systems.
The NTP engine, unlike chemical systems which depend on combustion to produce thrust, uses a propellant pumped
into a nuclear reactor heated to high temperatures and with thermal energy converted to kinetic energy, the monopropel-
lant exits the nozzle with very high velocity.
LEU NTP engines have been extensively studied by Aerojet Rocketdyne (AR) with the objective of providing a sys-
tem to NASA for payload mass and high thrusts to enable crewed missions from the 2030s through the 2050s (Joyner
et al., 2020). These include moving heavy cargo within 200 days to Mars.
LEU thermal spectrum systems based on ceramic-metallic (CERMET) fuel were moderated using materials in a
structural assembly [tie-tubes (T/T)] similar to the NERVAs design. The LEU NTP approach has been shown to be
technically feasible based on work from 2016 to 2019.
Significant conceptual design evaluations have been completed for the LEU NTP design approach at 25,000-lbf thrust and
have shown high specific impulse and thrust-to weight ratios at or above 3:1. Analyses has been carried out with results show-
ing that a NTP can be used for Mars architectures for crewed missions using a 25,000-lbf (111-kN) NTP engine system.
A nuclear thermal rocket does not require oxygen; it is thus a monopropellant system. It needs the highest specific
impulse for high temperature, low molecular weight, and chamber pressure; thus a high temperature reactor and hydrogen
fuel are the best combination. For hydrogen, M 5 2:016 g/mol, while for on the average propellants, MB13:8 g/mol with
TB2800K; for all other conditions fixed, the increase in specific impulse is more than twice, from Isp B300 to B770 sec-
onds. A nuclear rocket (Fig. 3.17) could give Isp B900 seconds with significant advantage in deep space exploration mis-
sions, such as reducing mission time and carrying crew and equipment. The power of a nuclear thermal rocket depends
most significantly on the temperature achievable; the higher the better. Thus, the SMRs and Gen-IV technologies reviewed
in the previous section are all candidates, provided space and weight constraints are satisfied. These are the same chal-
lenges characteristic of Gen-IV systems such as the technological development of advanced fuels and materials as well as
the optimization of design and operational parameters.
Bimodal nuclear thermal rocket engines, working for both propulsion and power generation, have been shown
(Clough) to reduce the weight of space vehicles to the Moon, Mars, and beyond. The NERVAs NERVA reactor can be
used for both propulsion and power; a 316 MW(t) reactor that produces a thrust of 66.6 kN and specific impulse a spe-
cific of 917 seconds can be run at 73.8 kW(t) to produce the necessary 16.7 kW electric power with a Brayton cycle
generator with a flow as illustrated in Fig. 3.16.
Four engines, using HEU and liquid hydrogen, have been simulated for the RL-10, RL-60 and CERMET engines.
The results from these simulations are given in Table 3.17. In all cases, the specific impulse far exceeds the capabilities
of chemical fuels while providing high thrust-to-weight ratio.
The improvements considered to the above HEU designs have included optimization of LEU core configurations
(Joyner) to maximum specific impulse and minimum the core reactor mass.
Better fuel materials such as UN with a matrix of molybdenum and tungsten to raise the melting temperature led to
an improvement in performance. Several other candidate materials, for example, zircaloy, zirconium hydride, and zirco-
nium carbide have been considered for cladding uranium dioxide fuels to raise the operating conditions.
Similar to the issue of HEU in submarine reactors, space systems will require LEU fuels in NTP systems. It has
been shown that modifications can be made to fast systems in the NERVA to achieve a thermal spectrum for which
LEU-based nuclear thermal rockets can offers distinct advantages over existing HEU systems.

TABLE 3.17 Engine comparisons for NERVA and CERMET systems.

Description Small NERVA Small CERMET Large NERVA Large CERMET

Engine diameter (cm) 87.7 55 98.5 68
U-235 mass (kg) 27.5 177.3 36.8 376.7
Reactor mass (kg) 1435 1215 2645 2722
Engine total mass (kg) 1730 1450 3305 3327
Thrust (kN) 33.2 33.8 113.3 111.2
Specific impulse (s) 899.6 885.4 911.5 824.2
Thrust-to-weight 1.96 2.38 3.49 3.41
 Nuclear reactors and systems Chapter | 3 137

FIGURE 3.17 Engine flow in NERVA-

type nuclear propulsion rocket engine.

TABLE 3.18 Design characteristics of some SNAP systems.

Characteristic SNAP-2A SNAP-8 SNAP-10A

Power (kW) 3 35 0.58
Design lifetime (a) 1 1 1
Reactor power (kW) 55 600 43
Reactor outlet (K) 920 975 833
Fuel and spectrum U-ZrH thermal U-ZrH thermal U-ZrH thermal
Coolant Na-K-78 Na-K-78 Na-K-78
Power conversion Rankine (Hg) Rankine (Hg) Thermoelectric (Si-Ge)
Turbine inlet (K) 895 950 777 K/610 K
Unshielded weight (kg) 545 4545 295

Source: IAEA.
138 Nuclear Engineering

3.8 Nuclear power systems in space

The two kinds of nuclear power systems used in space have been the radioisotope thermal generators (RTGs) and small
nuclear auxiliary power (SNAP) reactor systems. The new systems under development are more efficient and are classi-
fied as nuclear batteries or MNRs. These are briefly described in the following sections.

3.8.1 Radioisotope thermal generators

As described in Chapter 1, the alpha decay of heavy nuclides can be used as a thermal source of energy which can be
directly converted to electrical energy in, for example, a thermoelectric generator. Such systems, called RTGs have
been used since the 1960s for satellite and space probes. The most recent is a Pu-238 powered RTG to supply electrical
power to the Perseverance Rover in the NASA Mission to Mars 2020. The thermal energy from the radioisotope Pu-
238 is used to charge the batteries as well as to keep the various systems in the rover at their correct operating tempera-
tures. The hot side from such a source and the cold side at space temperatures are both connected to a thermocouple
which converts thermal energy into electricity. For the Perseverance Rover, about 110 W(e) is produced from a 2 ft by
2 ft, 45 kg power system containing 4.8 kg Pu-238 with an operational life of about 14 years.

3.8.2 Small nuclear auxiliary power systems

In addition to the RTGs used for heat and very small amounts of power, nuclear power reactors were also developed and
used in space systems. The Romashka reactor developed and used by the former Soviet Union, in the 1950s and 1960s,
was a nuclear reactor, with 49 kg HEU in a core of diameter 241 mm and height 351 mm, with an electrical output of
460
475 W produced from a thermoelectric generator weighing 635 kg. By 1970, the Topaz thermionic reactor converter

FIGURE 3.18 The SNAP-10A reactor. Courtesy NASA grc.nasa.gov 2021.

 Nuclear reactors and systems Chapter | 3 139

was developed based on the “boiling” of electrons, used as the working fluid, from a very hot static emitter surface typi-
cally at 1800K collected at a cooler surface typically at about 1000K. This electrical plant had electrons as the working
fluid. From the 1970s to the late 1980s, the former Soviet Union operated a series of the Cosmos spacecraft.
SNAP power systems, such as the SNAP-10A shown in Fig. 3.17, based on the Soviet program, were developed in
the 1960s.
Design characteristics of some SNAP (Fig. 3.18) models are listed in Table 3.18; these systems were intended for
use in the kilowatt electrical range but were not successful.
Small power systems, in the range of 3
20 kW(e), have also been proposed for converting nuclear fission thermal
energy using thermoelectric generators with heat pipes instead of the conventional coolant with pump flow. The Heat
pipe Operated Exploration Reactor is based on this concept.

3.8.2.1 Nuclear reactors for space

With missions to the Moon and Mars (Fig. 3.19), several designs of MNRs for space have been investigated with
renewed vigor. These represent improvements on the early designs of the 1950s and 1960s.
Very small, micronuclear (MNR) reactors have been designed for space and marine applications. One such design is
shown in Fig. 3.18 with a solid matrix core comprising fuel and heat pipes surrounded by radial and axial reflector,
absorber, and shielding region. The reactor core is embedded with cylindrical fuel rods of UN and heat pipes in a mono-
lith, surrounded with radial and axial reflectors and radiation shields, with boron carbide (B4C)-tipped control drums
embedded within the radial reflector. Thermal energy produced from nuclear fission in the critical core is transported
by heat pipes from the core to the TEG. The objective of this design is power conversion from nuclear thermal energy
in the core to electrical energy in the TEG without the use of rotating machinery such as pumps and turbines. The
shielding is designed so that permissible safety levels are maintained.

FIGURE 3.19 Core of a micronuclear

heat pipe reactor.
140 Nuclear Engineering

The main features of the overall core design are listed in Table 3.19 and Table 3.20. The thermal power is 700 kW
with a thermoelectric conversion system giving 35 kW(e) which gives an efficiency of only 5%.
The total uranium fuel is B149 kg, out of which B104 kg is U-235, with a compact core within a barrel of radius
35 cm and height 40 cm. The thermal output of this reactor can be as high as B700 kW(t) with lithium in heat pipes; with

TABLE 3.19 Overall characteristics of a MNR.

Parameter Material
Power kW(t)/kW(e) 700/35
Fuel/enrichment UN (70%)
Heat pipe Li (wick Mo-12Re)
Reflector/configuration BeO/side and top
Control system Control drums in reflector
Radiation shield/configuration Water and tungsten/side, top, and bottom

TABLE 3.20 MNR design parameters.

Design parameter Data

Fuel rod
Diameter (cm)/Height (cm), Density (g/cm3) 2.03/40, 13.6
Cross-section area of fuel rod (cm2) 3.2365
Mass fuel rod (kg) 1.7607
Mass fraction U235/ U238 0.6608/0.2743
Mass fraction N 0.0560
Mass of U (kg)/Mass of U235 (kg) 1.6621/1.1635
Fuel and fissile material
No. of fuel rods 90
Area of fuel rods (cm2) 291.2893
Mass of fuel rods (kg)/Mass of fuel rods/length (kg/cm) 158.46/3.9615
Mass U (kg)/Mass of U235 (kg) 149.5890/104.7150
Heat pipes
No. of heat pipes 37
RadiiVapor/Wick/Liq/Struc (cm) 0.90/1.00/1.02
Area of heat Pipe (cm2) 3.2685
Total area of heat pipes (cm2) 120.9345
Den of vapor (g/cm3) 0.01
Den of wick, structure (g/cm3) 12.0, 12.0
Mass (kg), Mass/length (kg/cm) 12.8538/0.321345
Matrix
Material, density (g/cm3) Nb-1Zr, 6.55
Area of Hexagon (cm2) 1585.1
Area (Hexagon-Fuel-HP) (cm2) 1172.9
Mass (kg), Mass/length (kg/cm) 331.9531
Core
Mass (kg) 503.2669
Mass/length (kg/cm) 12.5817
Axial shield
Top and Bottom shield water 35 cm
Top and Bottom shield tungsten 40 cm
 Nuclear reactors and systems Chapter | 3 141

electrical output depending on the thermoelectric efficiency. With typically B5%, the MNR is thus capable of providing
B35 kW(e). The shielding is the only major disadvantage in terms of weight when tungsten becomes necessary for
gamma dose reduction. It is necessary to minimize the shield weight to suit space and permissible dose requirements. The
consequences of heavy shield will be acceptable for on-ground specialized power applications but unfavorable for space
missions where a compromise on the dose might be necessary for conforming to any propulsion constraints.
Further, the use of HEU with over 90% enriched fuel will not be able to satisfy the requirements of criticality safety
in case of a launch abort; thus fuel enrichment in the range 55%
80% will be preferable.
In the kilopower range, 1
10 kW(e), reactors can be smaller and useful for space and specialized surface applica-
tions. Most recently, a 5 kW(t)/1 kW(e) prototype called KRUSTY (Kilopower Reactor Using Stirling Technology) was
fueled with a solid core HEU (Poston, Gibson, Godfroy, & McClure, 2020) U-8Mo, 32.2 kg fuel (27.7 kg U235) was
demonstrated at Los Alamos National Laboratory at fuel temperatures greater than 800 C. The heat removal system
uses sodium as the fluid and nickel wick in heat pipes connected to a Stirling engine. Thus, there are no movable parts.
This compares with a larger 2400 kW(t)/120 kW(e) 70% enriched UN fueled reactor with a 310 kg core and a thermo-
electric generator (Sun et al., 2018). An advantage of using high-density UN fuel is a reduction in the core size In another
conceptual design by Wang et al. (Chenglong, Sun, Simiao, Wenxi, & Suixheng, 2020), a 500 kW(t), 25 kW(e) 65%
enriched UN fueled HPR is given with a core weight also of 310 kg. Heat pipes of sodium, lithium and potassium have
been studied for a number of designs (El-Genk & Tournier, 2011) and appear promising for near future space applications.

3.9 Conclusions
This chapter reviewed nuclear reactor systems and designs that have evolved over the last seven decades. Some, like
MCF and ICF have yet to be demonstrated with technological and commercial feasibility.
The objective of this chapter was to develop a realization of the nature and scale of nuclear technologies.
For nuclear power reactors, the need to demonstrate economically competitive reactors with enhanced safety and
proliferation-resistant features were identified as the most important goals of Generation-IV reactors.
Nuclear propulsion reactors, especially for submarine, require a transition from HEU to LEU designs which must be
addressed but may not be achievable for several reasons.
For nuclear fusion reactors, ICF appears to be at its very early stages of development while MCF is anticipated to
show technological promise with success in ITER by 2025.
Nuclear propulsion is an area that appears both necessary and promising due to the limits of chemical propulsion
technologies that restrict deep space exploration and habitation of Mars.
In all these areas, there is a need for more R&D efforts, with more extensive modeling and computationally power-
ful simulation as well as validations with new technologies such as advancements in fuel and separation technologies.
Some identifiable challenges in modeling and simulation include
1. The development of advanced neutronic and thermal-hydraulic coupling models
2. A detailed safety analysis with coupled nuclear models
3. An improvement of computational tools
4. Validation of existing calculation tools and nuclear data libraries
5. Simulation of fuel burnup for developing high burnup minor actinide (neptunium, americium, curium) bearing fuels
6. Modeling the clad behavior at high temperature
7. Modeling the core-cooling capability
8. Modeling passive heat removal system
9. Modeling hydrogen production methods
10. Modeling the effect of high core outlet temperatures 700
950 C
11. Carrying out system integration studies
12. Carrying out a full design optimization

Problems
1. Why is it not necessary to use heavy water in a PWR?
2. What are the significant design differences between a BWR and a PWR?
3. What is the purpose of breeding in a FBR?
142 Nuclear Engineering

4. When did enriched uranium become available to NPP designers and why was the STR Mark II designed as a ther-
mal reactor rather than a fast reactor?
5. If you had to choose from a PWR or a liquid metal cooled reactor (LMCR) for a submarine, with the objective of
space constraints and enhanced power, which would you select and for what reasons? Consider the following fac-
tors: energy spectrum, size, pressure vessel, moderator, coolant, xenon poisoning, refueling, coolant melting point,
coolant-water interaction.
6. Comment on the following Russian design for RM-1 and VM-40A reactors in terms of their advantages and
disadvantages.
RM-1
Fuel:90% enriched U-Be13 dispersed in a beryllium matrix
Amount of U-235 in core: 90 kg
Fuel rod pellet diameter B10 mm covered with 0.1 mm MG clad in SS with 0.5 mm thickness
Number of fuel rods: B3000
Coolant: eutectic lead-bismuth alloy (44.5 wt.% Pb, 55.5 wt.% Bi), melting pointB125 C
VM-40A
Fuel:90% enriched U-Be13 dispersed in a beryllium matrix
Amount of U-235 in core: 200 kg
Coolant: eutectic lead-bismuth alloy (44.5 wt.% Pb, 55.5 wt.% Bi), melting pointB125 C
7. In view of the critical mass of Godiva (Section 3.2), comment on the amount of U-235 in a submarine reactor core.
8. How is high temperature achieved in Gen-IV reactors?
9. Neutronics of a submarine core: For a U-Zr fuel consisting of 15 wt.% Zr and 85 wt.% U enriched to 93% in U-
235, calculate the fast and thermal cross-sections and comment on the size of a submarine reactor core using this
fuel. How much would the size differ from that of a commercial PWR?
10. Estimate the natural uranium required for a submarine reactor 150 MW(t) operating at an average of 25% power for 1
year. Assume that 50% of all U235 atoms undergo fission and that 60% of the natural uranium goes into the submarine.

Nomenclature
English (lower case)
cp specific heat at constant pressure
cv specific heat at constant volume
e internal energy
k thermal conductivity
k Boltzmann constant
mi mass of ion
p pressure

English (upper case)

A cross-section area at throat
Ae cross-section area at exit (diverging part)
Ec energy of charged particle
G gain
B
H extrapolated height
H enthalpy
Isp specific impulse
M molecular weight
Pe exit pressure
Pe inlet pressure
Q heat
R universal gas constant
T absolute temperature
Q heat
T temperature
 Nuclear reactors and systems Chapter | 3 143

U internal energy
W work

Greek lower case

β delayed neutron fraction
E expansion ratio
γ ratio of specific heats cp =cv
η efficiency
ρ density
τE energy confinement time

Abbreviations and acronyms

Ci curie
kN kilo newton
Sv Sievert
Zr zirconium
ABWR advanced boiling water reactor
AP advanced power
BARC Bhabha Atomic Research Center
BWR boiling water reactor
CANDU Canadian Deuterium Uranium
CERMET ceramic metal
CHASNUPP Chashma nuclear power plant
D2O deuterium oxide
FBR Fast Breeder Reactor
GCFR gas cooled fast reactor
GCR gas cooled reactor
GW(e) gigawatt electric
HEU highly enriched uranium
HTR high temperature reactor
IAEA international atomic energy agency
ICBM intercontinental ballistic missile
ICF inertial confinement fusion
INS inertial navigation system
ITER international thermonuclear experimental reactor
LEU low enriched uranium
LOCA loss of coolant accident
LFR lead cooled fast reactor
LMFBR liquid metal cooled Fast Breeder Reactor
LWGR light water gas cooled reactor
MIRV multiple independently targetable reentry vehicle
MCF magnetic confinement fusion
MNR micronuclear reactor
MOX mixed oxide
MSR molten salt reactor
MW(e) megawatt electric
MW(t) megawatt thermal
MWd megawatt-day
NASA National Aeronautics and Space Administration
NERVA Nuclear Engine for Rocket Vehicle Application
NTP nuclear thermal propulsion
NPP nuclear power plant
PHWR pressurized heavy water reactor
144 Nuclear Engineering

PWR pressurized water reactor

RTG radioisotope thermal generator
SCR sodium cooled fast reactor
SCWR supercritical water cooled reactor
SHP shaft horsepower
SLBM submarine launched ballistic missile
SMR small modular reactor
SNAP small nuclear auxiliary power
SSBN submersible ship, ballistic, nuclear
SRS stellar reference system
TNT trinitrotoluene
USNRC United States Nuclear Regulatory Commission
VHTR very high temperature reactor
vSMR very small modular reactor
VVER water-water energetic reactor

References
Bari, R. A. (2014). Proliferation resistance and physical protection (PR&PP) in small modular reactors (SMRs). Handbook of Small Modular Nuclear
Reactors, 219
236. Available from https://doi.org/10.1533/9780857098535.2.219.
Bartel, R. (2016). WASH-1400 the reactor safety study. NUREG/KM-0010. Washington, DC.
Betti, R., & Hurricane, O. A. (2016). Inertial-confinement fusion with lasers. Nature Physics. Available from https://doi.org/10.1038/NPHYS3736.
Chenglong, W., Sun, H., Simiao, T., Wenxi, T., & Suixheng, Q. (2020). Thermal-hydraulic analysis of a new conceptual heat-pipe cooled small
nuclear reactor system. Nuclear Engineering and Technology, 52, 19
26.
El-Genk, M. S., & Tournier, J.-M. P. (2011). Uses of liquid-metal and water heat pipes in space reactor power systems. Frontiers in Heat Pipes.
Available from https://doi.org/10.5098/fhp.v2.1.3002.
He, X. T. (2016). The updated advancements of inertial confinement fusion program in China. Journal of Physics: Conference Series. Available from
https://doi.org/10.1088/1742-6596/688/1/012029.
He, X. T., & Zhang, W. Y. (2007). Inertial fusion research in China. European Physical Journal D. Available from https://doi.org/10.1140/epjd/
e2007-00005-1.
Horioka, K. (2018). Progress in particle-beam-driven inertial fusion research: Activities in Japan. Matter and Radiation at Extremes. Available from
https://doi.org/10.1016/j.mre.2017.08.002.
Huba, J. D. (2011). NRL PLASMA FORMULARY Supported by the office of naval research. Plasma Physics.
IAEA. (2007). Nuclear power plant design characteristics: Structure of nuclear power plant design characteristics in the IAEA power reactor informa-
tion system (PRIS). Vienna: International Atomic Energy Agency Nuclear Po. Available from https://www-pub.iaea.org/mtcd/publications/pdf/
te_1544_web.pdf.
IAEA. (2017). Fast reactors and related fuel cycles: Next generation nuclear systems for sustainable development FR17. In Proceedings of an
International Conference Yekaterinburg, Russian Federation, 26
29 June 2017. Yekaterinburg, Russian Federation: International Atomic Energy
Agency. Available from https://www-pub.iaea.org/MTCD/Publications/PDF/STIPUB1836web.pdf.
IAEA. (2018). Advances in small modular reactor technology developments. Vienna.
IAEA. (2020). Nuclear power reactors in the world. Vienna: International Atomic Eneergy Agency, Edited by Series No. 2. IAEA-RDS-2.
Joyner, C. R., Eades, M., Horton, J., Jennings, T., Kokan, T., Levack, D. J. H., . . . Reynolds, C. B. (2020). LEU NTP engine system trades and mission
options. Nuclear Technology. Available from https://doi.org/10.1080/00295450.2019.1706982.
Kopte, S. (1997). Nuclear submarine decommissioning and related problems. Bonn.
Lobner, P. (2018). Marine nuclear power: 1939
2018 Part 2A: United States-Submarines.
Mian, Z., Ramana, M. V., & Nayyar, A. H. (2019). Nuclear submarines in South Asia: New risks and dangers. Journal for Peace and Nuclear
Disarmament. Available from https://doi.org/10.1080/25751654.2019.1621425.
NASA. (2021). MARS 2020 mission perseverance rover. https://mars.nasa.gov/mars2020/timeline/cruise/.
Nikitaev, D., & Thomas, L. D. (2020). Seeded hydrogen in nuclear thermal propulsion engines. Journal of Spacecraft and Rockets. Available from
https://doi.org/10.2514/1.a34711.
Nuclear Energy Agency [NEA]. (2014). Technology roadmap update for generation IV nuclear energy systems. Gen IV International Forum.
Pakistan Navy. (2021). Pakistan submarine capabilities. Nuclear Threat Initiative. Available from https://www.nti.org/analysis/articles/pakistan-submarine-
capabilities/.
Philippe, S., & von Hippel, F. (2016). The feasibility of ending HEU fuel use in the U.S. Navy. Arms Control Today, November.
Poston, D. I., Gibson, M. A., Godfroy, T., & McClure, P. R. (2020). KRUSTY reactor design. Nuclear Technology. Available from https://doi.org/
10.1080/00295450.2020.1725382.
Ragheb, M. (2011). Nuclear naval propulsion. Nuclear Power
Deployment, Operation and Sustainability. Available from https://doi.org/10.5772/19007.
Reistad, O., & Olgaard, P. L. (2006). Russian nuclear power plants for marine applications. Roskilde, Denmark.
 Nuclear reactors and systems Chapter | 3 145

Sun, H., Wang, C., Ma, P., Liu, X., Tian, W., Qiu, S., & Su, G. (2018). Comnceptual design and analysis of a multipurpose micro nuclear reactor
power source. Annals of Nuclear Energy, 121, 118
127.
Tikhonchuk, V. T. (2020). Progress and opportunities for inertial fusion energy in Europe. Philosophical Transactions of the Royal Society A:
Mathematical, Physical and Engineering Sciences. Available from https://doi.org/10.1098/rsta.2020.0013.

ANNEX: the physics of nuclear fusion

For fusion to take place, the nuclei must be brought close to each other. As their energy is increased, the atoms are
stripped of their electrons and ionize into the plasma state. To get ions closer to fuse, the coulomb barrier must over-
come the repulsive forces. Once the barrier is crossed, the nuclear strong attraction forces enable fusion.
Z1 Z2 e2
V ðr Þ 5
4πε0 r
where

1
k5 B9 3 109 Nm2 =C2
4πε0
The potential energy is then
 2
Z1 Z2 e2 1:6 3 10219 106
V ðr Þ 5 59 3 10 9
Z1 Z2 ∙ MeVðrin fmÞ
4πε0 r r 10215
Thus

1:44Z1 Z2
V ðr Þ 5 MeVðrin fmÞ
r
For D-D (with Z1 5 Z2 5 1), and r 5 10 fm, V ðr Þ 5 0:144MeV. Taking the radius of a nucleus as
r 5 1:2 A1=3 fm
the radius of a deuterium nucleus is 1.5119 fm; at twice this distance, that is, if two D nuclei are touching each other,
V 5 0:4762 MeV. This energy corresponds to a temperature found from E 5 kB T from which

E 0:4762 3 106
T5 5 B5:5 3 109 K
kB 8:617 3 1025
If the coulombic barrier is the only dominant barrier, then the thermal energy is calculated to be in excess of 109 K
for D-D. This is the theoretical critical ignition temperature below which fusion should not be expected to be possible.
However, fusion does take place at temperatures lower than this critical estimate. Two phenomena make fusion possible
below this critical temperature. The first is the Maxwellian distribution of energies and the second is the tunneling
of ions through the coulomb barrier; both these phenomena lower the critical temperature. If this were not the case,
then fusion would not have been possible on the sun, as its core temperature is of the order of 1:5 3 107 K which is
lower than the critical temperature for D-T fusion.
At a given temperature, or energy, all the ions do not have the same energy; in fact, most have a probable energy
while a few have much lower and much higher energies. This is illustrated in Fig. 3.A1 for the distributions nðEÞ; nðvÞ;
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
for nðEÞ; x 5 E=kT; for nðvÞ; x 5 v m=2kT . pffiffiffi
The probability of quantum-mechanical tunneling, given by the Gamow factor, is Be2EG = E while the number of
particles at high energies due to their Maxwellian distribution (Section 2.1) is Be2E=kT ; thus at high energies the num-
pffiffiffiffiffiffiffiffi
ber of ions tunneling through the barrier is Be2 EG =E e2E=kT . This is a result obtained from the Schrodinger equation
(Section 1.1). Here, the Gamow energy EG is given by

EG 5 2 mr c2 ðπαZ1 Z2 Þ2

where EG , Gamow energy; mr , reduced mass; α, fine structure constant α 5 e2 =ð4πε0 h̄ÞB137
146 Nuclear Engineering

FIGURE 3.A1 Distribution functions n(v) and n(E).

The transmission probability has a maxima at

 2=3
kT
E M 5 EG
2EG
Several other fusion reactions are also possible but the reactions between isotopes of hydrogen are more likely than
the reactions between heavier nuclei such as helium due to the stronger coulomb barrier in case of the latter. The D-D
fusion reactions possible are
(
2 He 1 0 n
3 1
2
1 H 1 2
1 H-
1H 1 1H
3 1

The fusion reaction rate for two species with n1 and n2 atoms cm23 is given by
R5nD nT hσvi cm23 s21
where hσvi, the Maxwellian-averaged cross-section is a function of the temperature T of the reaction. For DD and DT
reactions,
 
hσviDD 5 2:33 3 10214 T 22=3 exp 2 18:76 T 21=3 cm3 s21

and
 
hσviDT 5 3:68 3 10212 T 22=3 exp 2 19:94T 21=3 cm3 s21

where T is in keV (1 eVB11,600K). The functions hσviDT and hσviDT , have maximum values of 3.6 3 10
17 cm3/s at
825.3 keV and 5 3 10
15 cm3/s at 991 keV respectively, that is, B 100 million degrees.
The fusion cross-sections, from the Naval Research Lab’s plasma physics formulary are
 21
A5 1 ðA4 2A3 EÞ2 11 A2
σ F ðE Þ 5   1 
E exp A1 E22 2 1
with the coefficients listed in Table 3.A1
The fusion cross-sections for DD and DT reactions are shown in Fig. 3.A2.
The DD cross-sections are orders of magnitude lower than the DT reaction and so the DT reaction is preferred.
The power density released in the form of charged particles (Huba, 2011) is
PDD 5 3:3 3 10213 n2D hσviDD W=cm3
and
PDT 5 5:6 3 10213 nD nT hσviDT W=cm3
 Nuclear reactors and systems Chapter | 3 147

TABLE 3.A1 Coefficients of the fusion cross-sections.

Coefficient D(T,n)He4 D(D,p)T3 D(D,n)He3

A1 45.95 46.097 47.88
A2 50200 372 482
A3 1.368 3 1022 4.36 3 1024 3.08 3 1024
A4 1.076 1.22 1.177
A5 409 0 0

FIGURE 3.A2 D-D and D-T fusion cross-section.

Assuming nD 5 nT 5 n=2, the rate of reactions is

1 2
R5 n hσvi
4
The energy of the hot plasma is
E 5 3nkB T
As a first estimate, equating the power produced by fusion, with the assumption that the neutron energy is lost and
the charged particle energy Ec remains (due to the short range of ions) to the power lost during the energy confinement
time, τ E , gives an energy balance
1 2 3nkB T
n hσviEc $
4 τE
which is rearranged to give
12kB T
nτ E $
Ec hσvi
For D-T, the minimum value of T=hσvi at 26 keV (B260 million K) and the values of kB and Ec gives the Lawson
criterion for DT fusion
nτ E $ 1:5 3 1020 m23 s
The Lawson criterion is fundamental figure of merit to the realization of fusion as source of energy. In simple
words, it says that the product of plasma density n and confinement τ E must exceed the quantity 1:5 3 1020 m23 s
which means the higher the density, the lower the confinement time and vice versa. To achieve breakeven through
Lawson’s criterion is the goal of all experiments.
Chapter 4

Mathematical foundations

This chapter reviews the mathematical foundations and knowledge required for understanding the formulation, performing a
simulation, and coding for numerical implementation of problems in nuclear engineering. The material is presented in the
following order: general mathematics covering ordinary and partial differential equations, mathematics specific to the neutron
diffusion and transport equations followed by probability and statistics for Monte Carlo methods used in simulation. The
objective is, again, to provide a comprehensive review of the mathematics applicable to nuclear engineering, usually spread
over different subjects, in one resource available for study before or during the phase when problems are encountered.
Modeling in simulation in nuclear engineering covers several areas including neutron diffusion and transport, fluid
dynamics, heat transfer, dynamics and control, and structural mechanics as depicted in Fig. 4.1.
The mathematical formulation of the governing equations is mostly in the form of differential equations, integral
equations and integro-differential equations. Other than the deterministic formulations, simulation is carried out based
on probability distributions and estimates in a stochastic formalism.
The objectives of mathematical modeling and simulation in nuclear engineering incude the design and performance
anlayses of nuclear systems in many cases to compute the power produced in a nuclear reactor and to ensure it remains
within safe critical bounds. For example, in a model of a reactor core shown in Fig. 4.2, the innovative nuclear reactor
design has a core at the center, contains fuel rods and control rods placed vertically in assemblies, moderator and coolant
and is surrounded by radiation shielding. The fluid removes the heat from the fissioning fuel rods and, as described in
Section 4.3 for reactor designs, transfers its thermal energy to produce steam for a turbine and an electrical generator.
The mathematical models described in this chapter are intended to be applied for the computation of neutron and
photon flux inside the core, carrying out a dynamic and safety analysis as well as to estimate the radiation leaving the

FIGURE 4.1 Modeling and simulation in nuclear engineering.

Nuclear Engineering. DOI: https://doi.org/10.1016/B978-0-323-90618-0.00004-1

© 2022 Elsevier Inc. All rights reserved. 149
150 Nuclear Engineering

FIGURE 4.2 Model of a micronuclear reac-

tor (side view).

core for which radiation shielding is designed as shown by the presence of water and tungsten shields in Fig. 4.2. The
underlying physics of nuclear radiation and radiation transport was covered in Chapter 1 for charged particles and elec-
tromagnetic radiation and in Chapter 2 for neutron interactions with matter.

4.1 Ordinary differential equations

Ordinary differential equations (ODEs) are used to represent systems with one independent variable such as the spatial
variable x, and can, in many cases, be solved exactly (Baker, 2016; Polyanin & Manzhirov, 2006; W. G. & D.
Zwillinger, 1993). ODEs can be first-, second-, or higher order, where the order represents the power of the highest
derivative of the dependent variable with respect to the independent variable.
A linear ODE is of the nth-order if it can be expressed as
yðnÞ 1 f1 ðxÞyðn21Þ 1 . . . 1 fn21 ðxÞy0 1 fn ðxÞy 5 RðxÞ (4.1)
where yðnÞ is the nth derivative of the function y. If the RHS 5 0, the ODE is called a homogeneous linear ODE. If any
of the coefficients are a function of the dependent variable y, the ODE is nonlinear.
Thus, first-order differential equations areÐ linear if they can be expressed as yð1Þ 1 f ðxÞy 5 RðxÞ. They are easiest
solved by the use of an integrating factor exp ð f ðxÞdxÞ; thus, the ODE yð1Þ 1 2xy 5 x can be written as
d x2 2
ye 5 xex
dx
and by integrating both sides
ð
2 2
yðxÞex 5 xex 1 C
 Mathematical foundations Chapter | 4 151

giving the results

1
1 Ce2x :
2
yðxÞ 5
2
Solving Nth-order ODEs is possible by reducing them to N first-order ODEs which, when linear, are easily cast into
state-space equations amenable to standard matrix methods of linear algebra.
Examples of nonlinear first- and second-order ODEs are
dy
y 2x50
dx
and
d2 y   dy
a1 2
1 a2 y 2 2 1 1 a3 y 5 0:
dx dx
If an ODE has coefficients that are functions of the independent variable x,
dy
1 2xy 5 x
dx
it is called a variable coefficient linear ODE.
Equations of the form
dy
1 pð xÞy 5 gð xÞ
dx
have an exact solution
Ð ð Ð
yðxÞ 5 e2 pðxÞdx
gðxÞe pðxÞdx
dx:

In standard form, the SturmLiouville theory applies to second-order linear ODE boundary value problems (BVPs)
that can be written as
 
d dy
pð x Þ 1 qðxÞy 5 2 λwðxÞyð xÞ (4.2)
dx dx
where the functions pðxÞ; qðxÞ; wðxÞ are the coefficient functions and λ, if it exists, is called the eigenvalue which is a
scalar for which the equation is balanced and has eigenfunctions associated with the eigenvalue.
As an example, the steady-state neutron diffusion (Section 4.5) equation
r ∙ Drφ 2 Σa φ 1 S 5 0 (4.3)
and the steady-state heat conduction equation
r:krT 1 S 5 0 (4.4)
are both second-order ODEs which can be expressed in the form of the SturmLiouville equation.
Let us therefore review a method of obtaining an exact solution of a simple form of such an equation: the linear
nonhomogeneous form
d2 y dy
a1 ð x Þ
2
1 a2 ð xÞ 1 a3 ð xÞy 5 Fext
dx dx
which arises in several areas of science and engineering such as the Schrodinger equation discussed in Chapter 1,
forced vibrations in mechanical systems, including the Euler equation, the Bessel equation, the Airy equation. The
Legendre equation. Eq. (4.2) also represents the 1-D motion in the mass-spring-damper problem, when φ represents
the displacement xðtÞ, and x represents time t, where the restoring force of the spring is represented by Hooke’s Law
and an external time-dependent force Fext is applied. The procedure of solving this equation is to seek a complemen-
tary solution yc ðxÞ from the homogeneous equation and a particular solution yðxÞ from the inhomogeneous term, and
express the solution as
yðxÞ 5 yc ðxÞ 1 yp ðxÞ:
152 Nuclear Engineering

Thus, in operator form

 
a1 D^ 1 a2 D^ 1 a3 yc ðxÞ 5 0
2

The operator D^ isPtreated as a scalar and roots are obtained for a1 m2 1 a2 m 1 a3 5 0. For roots m1 ; m2 , the solution
is expressed as yc 5 2i51 Ai emi x . For the particular solution
1
yp ðxÞ 5   Fext :
a1 D^ 1 a2 D^ 1 a3
2

The solution
X
yp ðxÞ 5 Ai e m i x
i

where mi are found from the particular solution and the coefficients Ai are determined from the given boundary condi-
tions which can be Dirichlet (with the dependent variable specified on the boundaries), Neumann (with the normal
derivative specified on the boundaries), or mixed boundary conditions.

Exercise 4.1: Find the exact solution of the equation.

d2 φ
D 2 Σa φ 5 0
dx2
for a planar source at x 5 0 emitting S neutrons cm22 s21 with the boundary conditions: (1) Dirichlet b. c: finite flux
SL 2x=L
φðxÞ, and (2) source Neumann condition: limx-0 J ðxÞ 5 S=2; ðL2 5 D=Σa ). Answer: φ 5 2D e .

Example 4.1: Find the exact solution of the second-order ODE.

d2 φ
D 2 Σa φ 5 2 S
dx2
for a planar source at x 5 0 emitting S neutrons cm22 s21 in a finite slab of thickness 2d. With the boundary conditions:
(1) φða 1 d Þ 5 φð 2 a 2 dÞ 5 0, and (2) source condition: limx-0 J ðxÞ 5 S=2; ðL2 5 D=Σa ). First obtain the complemen-
tary solution using the homogeneous equation
 2 
D^ 2 1=L2 φc ðxÞ 5 0:

The roots are m1 5 1=L, m2 5 2 1=L, giving the complementary solution

φc ðxÞ 5 C1 ex=L 1 C2 e2x=L
For the particular solution
1
φp 5    ð 2 SÞ
D^ 2 1=L D^ 1 1=L

which can be written in partial fractions as

" #
1 1 1  
φp 5  2  2 SL2
2 LD^ 2 1 LD^ 1 1

and expressed as
1 h 21  21 i 2 
12LD^
φp 5 1 11LD^ SL
2
 21  2
Now expanding 12LD^ 5 1 1 LD^ 2 LD^ 1 ?, gives
φp 5 SL2
 Mathematical foundations Chapter | 4 153

The solution is thus

φðxÞ 5 A1 coshx=L 1 A2 sinhx=L 1 SL2
and the coefficients A1 ; A2 can be found from the boundary conditions. Using the boundary conditions gives

SL sinh ða 1 d 2 jxjÞ=L
φðxÞ 5  :
2D cosh a 1L d
In case there are a number of sources, placed at different locations, the Green’s function can be used for calculating the
flux in the medium.

4.1.1 The Poisson equation: steady-state heat conduction in 1-D

The steady-state heat conduction equation (Eqs. 4.3 and 4.4) with Fick’s law q 5 2 krT, gives the steady-state heat
conduction (as for neutron diffusion) equation with heat generation for constant k in regular geometry:
d2 T qw
rectangular 2
1 50
dx k

1d dT qw
cylindrical r 1 50
r dr dr k
1 d dT qw
spherical r2 1 50
r 2 dr dr k
of the form r2 φ 5 S, which is called the Poisson equation; in the homogeneous case ðS 5 0Þ, the equation reduces to
the Laplace equation. These equations can be solved by the standard procedure given above.

Exercise 4.2: For a fuel rod considered as a plate of thickness 2a and clad of thickness b with the boundary conditions:

dT
T ð0Þ 5 Tm ; 50
dx x50

obtain the temperature distribution, in the fuel and clad, and the heat flow using the AP1000 data from Chapter 3
(Table 3.3): fuel radius 5 4.095 mm, gap thickness 0.082 mm, clad thickness 0.5715 mm.
The AP1000 fuel rods are of a cylindrical configuration with cross section. In this model, we can assume that the
“half-thickness” of a fuel plate is a equal to the radius of the fuel (Fig. 4.3).
Assume that the thermal conductivity of UO2 (IAEA, 2008) is given as

100 6400 16:35

kf ðτ Þ 5 1 5=2 exp 2 W=m K
7:5408 1 17:692τ 1 3:6142τ 2 τ τ
where
T
τ5
1000
and for the cladding, kc 5 36W=mK.
Recall that the steady-state heat equation
qw
r2 T 1 50
k
is valid for constant thermal conductivity while kf is generally a function of temperature or space in the case of a com-
posite wall. Thus to solve the equation, it will be easier to assume that kf is constant. When this assumption is valid,
the above ODE can be easily solved to yield the temperature distribution
qw 2
T ðxÞ 5 Tm 2 x
2kf
154 Nuclear Engineering

FIGURE 4.3 AP1000 fuel rod in a moderator cell.

Ða Ða
and the heat flow is found by integrating from 0 q dx 5 2 k 0 dTðxÞ. Also

Tm 2 Ts
q5
Rf

where T ðaÞ 5 Ts and the “resistance” (analogous to Ohms Law V 5 IR) is

a
Rf 5 :
2kf A

In the cladding, the temperature is found by integrating r2 T 5 0 to get

TðxÞ 5 C1 x 1 C2

in terms of the constants of integration C1 and C2 .

With the outer surface temperature T ða 1 bÞ 5 Tc , the temperature in the cladding is
x2a
T ðxÞ 5 Ts 2 ðTs 2 Tc Þ:
b
Thus the overall heat flow is
Tm 2 Tc
q5
Rf 1 Rc
where the resistance of the cladding is
b
Rc 5 :
kc A
 Mathematical foundations Chapter | 4 155

In the above exercise, the thermal conductivity is a function of the dependent variable TðxÞ; thus the governing
equation in plane geometry is the nonlinear equation
d2 T
kf ðT Þ
1 qw 5 0:
dx2
In some cases, it is possible to obtain exact solutions of such nonlinear equations, but generally numerical methods are
required as will be discussed in this chapter.

4.1.2 Coupled first-order ODEs: the point kinetics equations

The Point Kinetics Equations (PKE) in a lumped model of a nuclear reactor, for the neutron density nðtÞ and the con-
centration of the ith nuclide Ci ðtÞ from Chapter 2, are
B
dnðtÞ ρðtÞ 2β X6
5 nð t Þ 1 λi C i ðt Þ
dt Λ i51

and
dCi ðtÞ β i ðtÞ
5 nðtÞ 2 λi Ci ðtÞ; i 5 1; ?6
dt Λ
with initial conditions nð0Þ and Ci ð0Þ. These are two coupled first-order ODEs that can be solved by combining them
into one second-order ODE, or by taking Laplace transforms
ðN
N ð sÞ 5 nðtÞe2st dt
0

and
ðN
Γi ð s Þ 5 Ci ðtÞe2st dt
0

which yields
h P Ci ð0Þi
l nð0Þ 1 i λsi 1 λi
N ðsÞ 5 P β i λi :
ls 1 β 2 ρo 2 i s 1 λi

The solution, obtained by inversion, is

X
7
nðtÞ 5 Ai e ω i t
i51

with each root ωi satisfying the “inhour equation”

X β i λi
ρ0 5 β 1 lω 2 iω1λ
:
i
P
Since the total delayed fraction β 5 i βi
X βiω
ρ0 5 lω 1 i ω1λ
:
i

Thus the roots ωi can be found explicitly or by solving the transcendental equation giving the solution.

Exercise 4.3: Solve the coupled equations PKE to show, for one group of delayed neutrons, that the solution is.
2 3 2 3
  1 1
^
nðtÞ β β
^ 5 a0 4 5 e ω 0 t 1 a1 4 5 eω 1 t
CðtÞ Λðλ 1 ω0 Þ Λðλ 1 ω1 Þ
156 Nuclear Engineering

where the roots are

2 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 3
2
1 ρ2β ρ2β ρ
ω0;1 5 4 2λ 6 2λ 1 4λ 5:
2 Λ Λ Λ

Evaluate the roots for ρ 5 0:2β given the thermal fission U235 data
β
λ 5 0:0767s21 ; β 5 0:00650; Λ 5 5 0:0847s
λ
and plot the solutions. In general, there is one real root with the same sign of the applied reactivity while all other roots
are negative and real.

4.2 Partial differential equations

Partial differential equations (PDEs) represent systems where the differential equation has two or more independent variables
such as spatial dimensions x; y; z, orthogonal and azimuthal angles θ; ϕ, energy E, and time t. A linear second-order PDE

@2 φ @2 φ @2 φ @φ @φ
aðx; yÞ 1 b ð x; y Þ 1 cð x; y Þ 1 dð x; yÞ 1 eð x; yÞ 1 f ð x; yÞφ 5 0 (4.5)
@y2 @x@y @x2 @y @x
is classified as elliptic, parabolic, or hyperbolic if b2 2 4ac is less than, equal to, or greater than zero, respectively.
Using the classification criterion specified above, we can classify the PDE:
1 @φðr; tÞ @2 φ
5 D 2 2 Σa φðr; tÞ 1 Sðr; tÞ
v @t @x
with a 5 b 5 0; c 5 D; thus b2 2 4ac 5 0, as a parabolic equation.
This implies a certain kind of boundary conditions: a Dirichlet or Neumann boundary condition on an open surface to
be specified for a stable solution. Similarly, elliptical PDEs require Dirichlet or Neumann boundary conditions on a closed
surface surrounding the region of interest, while hyperbolic PDEs require Cauchy boundary conditions on an open surface.
The Handbook of Nonlinear Partial Differential Equations (Polyanin & Zaitsev, 2004) gives solutions for nonlinear
parabolic, hyperbolic, and elliptic equations, nonlinear equations with exponential, hyperbolic, logarithmic, and trigono-
metric nonlinearities in one, two, or more space variables, as well as for higher order equations. The exact methods
given for solving nonlinear PDEs include characteristic equations and canonical forms.
The solution procedure is usually to convert PDEs to ODEs and then solved to obtain exact analytical solutions
5 5
where possible, or a system of algebraic equations Ax 5 B where A is a matrix, x is the vector being solved, and B is a
known “force” vector. The solution is thus obtained, from standard numerical techniques, such as Gaussian elimination,
5
GaussSiedel iterative methods, LU decomposition methods, etc., by inversion of the matrix A giving x 5 A21 B.

4.2.1 Equations of fluid dynamics

Examples of PDEs used in thermal hydraulics are the continuity equation
@ρ
1 r ∙ ðρu Þ 5 0;
@t
the conservation of momentum equation
@ρu
1 r ∙ ðρuu Þ 5 2 rp 1 rτ 1 ρg;
@t
and the conservation of energy equation
@ρe @ρ  
2 1 r ∙ ðρueÞ 5 r ∙ krT 2 pr ∙ ðu Þ 1 Φ 1 S
@t @t
To “close” the above system, an equation of state
ρ 5 ρ ðP; TÞ
is used for solving for the six independent variables ρ; u; p; e.
 Mathematical foundations Chapter | 4 157

4.2.2 The 1-D time-dependent heat conduction

A 1-D time-dependent form of Eq. (4.4) for temperature, with constant thermal conductivity k, from an energy balance, is
@2 T qw 1 @T
1 5
@x 2 k α @t
where α 5 k=ρc is the thermal diffusivity of the material and can be solved with given boundary conditions for a 1-D
slab shown in Fig. 4.4.
Assuming separability T ðx; tÞ 5 XðxÞΓðtÞ and setting equal to a separability constant 2β 2
1 d 2 XðxÞ 1 dΓðtÞ
5 5 2 β2
XðxÞ dx2 αΓðtÞ dt
the PDEs are converted to a set of ODEs
dΓðtÞ
1 αβ 2 ΓðtÞ 5 0
dt
d 2 XðxÞ
1 β 2 X ðxÞ 5 0
dx2
with solutions ΓðtÞ 5 e2αβ τ and Xðβ m xÞ. This is an eigenvalue problem, as it satisfies a boundary condition such as
2

0 0
X 0 5 0 and for X ðxÞ 5 cosβx, X 0 5 5 2 βsinβx 5 0, so either β 5 mπ, m 5 0; 1; 2; . . .. The solution is thus
X
N
cm X ðβ m xÞe2αβ m τ :
2
T ðx; tÞ 5
m51

The spatial part of the solution is

X
N
F ðxÞ 5 cm X ðβ m xÞ
m51

for which the coefficients are determined using orthogonality

ðL 
0 for m 6¼ n
X ðβ m ; xÞX ðβ n ; xÞdx 5
0 N ðβ m Þ for m 5 n
where the normalization integral N ðβ m Þ is defined as
ðL
 2
N ðβ m Þ 5 Xðβ m ; xÞ dx
0

and
ðL
1
cm 5 5 Xðβ m ; xÞFðxÞdx
N ðβ m Þ 0
to yield the time-dependent solution
X
N ðL
1 0 0 0
e2αβ m τ
2
T ð x; tÞ 5 X ðβ m ; xÞ Xðβ m ; x ÞFðx Þdx :
m51
N ðβ m Þ 0

Here F(x) is an arbitrary function and the X functions depend on the boundary conditions.

FIGURE 4.4 Boundary conditions on a 1-D slab.

158 Nuclear Engineering

4.2.3 Laplace equation: 2-D steady-state heat conduction

For 2-D steady-state conduction with constant k and no heat generation, that is, qw 5 0, Eq. (4.4) is the Laplace equation
@2 @2
1 T ðx; yÞ 5 0
@x2 @y2
for a square plate of width W and height H. Given the Dirichlet boundary conditions T ðx; H Þ 5 Tu and
T ðW; yÞ 5 Tð0; yÞ 5 Tðx; 0Þ 5 Ts the equation is solved by assuming separability T ðx; yÞ 5 XðxÞYðyÞ; this gives
1 d2 X 1 d2 Y
2
1 1 λ2 5 0
X dx Y dy2
with two second-order ODEs
1 d2 X 1 d2 Y
5 2 β2; 5 2 γ2 ;
X dx2 Y dy2
where β 2 1 γ 2 5 λ2 . The solutions of X and Y are trigonometric functions with four constants determined by the bound-
ary conditions and from the orthogonality of Legendre polynomials. The exact solution is
nπy
2X N
ð 21Þn11 1 1 nπx sinh W
T ðx; yÞ 5 ðTu 2 Ts Þ sin   1 Ts :
π n51 n W sinh nπHW

The above solutions were considered for steady-state and constant thermal conductivity k. A numerical finite-
difference solution for this 2D plate is given in Section 4.5.1. In general, if kðT Þ 5 f ðTÞ, the equation on a square plate
is an elliptic nonlinear Poisson equation.
@2 @2
k ðT Þ 1 2 T ðx; yÞ 1 qw 5 0:
@x 2 @y
Some homogeneous cases for which numerical methods are compared with exact solutions (Filipov & Faragó, 2018;
Yeih, 2020) are for kðT Þ 5 T on a circle of unit radius and on an annular region; kðT Þ 5 eT and on a square plate
0 # x # 1, 0 # y # 1.
For kðT Þ 5 eT , the exact solution is T ðx; yÞ 5 logðxy 1 5Þ, shown in Fig. 4.5, which is compared with numerical solu-
tions which are generally used for nonlinear PDEs.
Since it is not always possible to obtain exact solutions, numerical methods are used as discussed later in this
chapter.

FIGURE 4.5 Temperature field T ðx; yÞ on a square plate

(kðT Þ 5 eT ).
 Mathematical foundations Chapter | 4 159

4.2.4 Heat conduction in 2-D and 3-D

In rectangular, cylindrical, and spherical coordinate systems, the governing heat conduction equation can be written as
1 @ @T 00 0 @T
rn k 1 q 5 ρc
r n @r @r @t
with n 5 0 (slab), 1 (cylinder), an 2 (sphere).
Thus, in rectangular geometry,
@2 @2 00 0 @T
k 1 2 T ð x; yÞ 1 q 5 ρc
@x 2 @x @t
This equation is solved assuming separability, similar to the 1-D case. In the steady-state heat conduction equation,
for 2-D or 3-D problems, the governing equation becomes a PDE with the appropriate forms of the Laplacian.
The Laplace equation is solved by assuming separability when the resulting equations can be written as ODEs of the
form for harmonic motion. With the relevant boundary conditions, solutions can be expressed in terms of modes, or
eigenfunctions, for eigenvalues. The solutions can be expressed in terms of an infinite series and have orthogonal
properties.
In 3-D, the rectangular coordinate equation for thermal conductivity k which may be space- or temperature-
dependent, the equations are
@ @T @ @T @ @T 0 00 @T
k 1 k 1 k 1 q 5 ρc
@x @x @y @y @z @z @t
For a cylinder x 5 rcosφ; y 5 rsinφ
1@ @T 1 @T @T @ @T 00 0 @T
k 1 2 k 1 k 1 q 5 ρc
r @r @r r @φ @φ @z @z @t
For a sphere, x 5 rcosφsinθ; y 5 rsinφsinθ, and z 5 cosθ (Fig. 4.6)
1 @ @T 1 @ @T 1 @ @T 00 0 @T
kr 2 1 2 2 k 1 2 ksinθ 1 q 5 ρc
r 2 @r @r r sin θ @φ @φ r sinθ @θ @θ @t
Some boundary conditions and initial conditions need to be specified to solve the above: Neumann, Dirichlet, or
mixed are given in Table 4.1.

FIGURE 4.6 Solid angle.

160 Nuclear Engineering

TABLE 4.1 Dirichlet, Neuman and mixed boundary conditions.

Boundary type Boundary condition

Temperatures at surfaces T ð0; t Þ 5 T1 , T ðL; t Þ 5 T2
Heat fluxes ðW=m2 Þ q 5 2 k @T @x
Insulated boundary @T ð0;t Þ
@x 50
Reflecting boundary @T ðL=2;t Þ
@x 50
Convection boundary condition 2k @T@x ð0;t Þ
5 h1 ½TN1 2 T ð0; t Þ
h i
Radiation boundary condition 2k @T@x ð0;t Þ
5 ε1 σ Tsurr;1
4
2 T 4 ð0; tÞ
ðx0 ;t Þ ðx0 ;t Þ
Interface boundary conditions TA ðx0 ; t Þ 5 TB ðx0 ; t Þ, 2kA @TA@x 5 2 kB @TB@x

In the time-dependent case, assuming separability T ðr; tÞ 5 ψðrÞΓðtÞ in the usual manner
1 1 dΓðtÞ
r 2 ψ ðr Þ 5 2 5 2 λ2
ψðr Þ αΓðtÞ dt
to get the Helmholtz equation
r2 ψðr Þ 1 λ2 ψðr Þ 5 0
Which can be reduced to ODEs in coordinate systems including the rectangular, cylindrical, and spherical systems.

Exercise 4.4: For 3-D Cartesian geometry, extend the analysis of Sections 4.4.2.3 and 4.2.3 to find the exact solution
for the time-dependent temperature given the initial conditions.

In cylindrical geometry, with the same procedure followed above, for

1 @2 ψ 1 @ψ 1 @ 2 ψ @2 ψ 1 dΓðtÞ
1 1 2 2 1 2 5 5 2 λ2
ψ @r 2 r @r r @φ @z αΓðtÞ dt
and ðr; φ; zÞ 5 RðrÞΦðφÞZðzÞ, the ODEs are
1 d2 Z 1 d2 Φ 1 d 2 Rv 1 dRv v2
5 2 η2 ; 5 2 v 2
; and 1 2 5 2 β2
Z dz2 Φ dφ2 Rv dr 2 r dr r2
The separated ODEs and their solutions are
d2 Z
1 η2 Z 5 0Z ðη; zÞ:sinηzandcosηz
dz2
d2 Φ
1 v2 Φ 5 0Φðv; φÞ:sinυφandcosυφ
dφ2
d 2 Rv 1 dRv v2
1 1 β 2
2 Rv 5 0Rv ðβ; r Þ:Jv ðβr ÞandYv ðβr Þ
dr 2 r dr r2
β 2 1 γ 2 1 η2 5 λ 2
dΓ
1 αλ2 Γ 5 0ΓðtÞ:e2αλ τ
2

dt
Bessel functions are solutions of the νth order (ν is a nonnegative real number) second-order ODEs
d 2 y 1 dy ν2
2
1 1 12 2 y50
dx x dx x
There are functions of the first kind, which are regular at the origin (e.g., I0 ; I1 ; J0 ; J1 ), and functions of the second
kind which are singular at the origin (e.g., Y0 ; Y1 ; K0 ; K1 ). The solutions which are linear combinations of Jν ðxÞ and
 Mathematical foundations Chapter | 4 161

Yν ðxÞ. These are related as

cosπυJυ ðxÞ 2 J2υ ðxÞ
Yυ ðxÞ 5
sinπυ
for J2ν 5 ð 21Þν Jν . When the differential equation is written in a modified form:
d 2 y 1 dy υ2
1 2 1 1 y50
dx2 x dx x2
the solutions are a linear combination of Iν ðxÞ and Kν ðxÞ: Kv ðxÞ 5 π2 I2v ðxÞ 2 lv ðxÞ
sinπν . Functions of the first and second kind are
shown in Fig. 4.7.
Now, consider the 1-D cylindrical form of the neutron diffusion equation (Eq. 4.3)
d2 1d
D 1 φðr Þ 2 Σa φðr Þ 1 Sðr Þ 5 0
dr 2 r dr
For a “line source” where the source extends along a line on the z-axis at r 5 0, the homogeneous form of the equation
will apply to all r excluding the source. This will be the modified Bessel equation form given above with ν 5 0 and a

FIGURE 4.7 (AD) Bessel functions of the first and second kind.
162 Nuclear Engineering

solution of the form

φðr Þ 5 A1 I0 ðr Þ 1 A2 K0 ðrÞ
where A1 and A2 are constants. Since we require a solution at r 6¼ 0, and the function I0 ðrÞ grows with increasing r
while K0 ðrÞ remains bounded and decreases, the solution will be φðr Þ 5 A2 K0 ðrÞ. For the Bessel functions, the series
forms are:
x2 x4 x6  x x2 3x4 11x6
I0 ðxÞ 5 1 1 1 1 1 . . . and K0 ðxÞ 5 2 γ 1 ln I0 ð xÞ 1 1 1 1 ...
4 64 2304 2 4 128 13; 824
where γ 5 2 0:577215665 . . . is Euler’s constant.

4.2.4.1 Spherical geometry

In spherical coordinates, the temperature field, in general, is T 5 Tðr; θ; φ; tÞ given by the governing equation
@2 T 2 @T 1 @ @T 1 @2 T 1 @T
1 1 sinθ 1 5
@x 2 r @r r sinθ @θ
2 @θ r sin θ @φ
2 2 2 α @t
with μ 5 cosθ
 
@2 T 2 @T 1 @  
2 @T 1 @2 T 1 @T
1 1 1 2 μ 1 5
@r 2 r @r r 2 @μ @μ r 2 ð1 2 μ2 Þ @φ2 α @t
defining a new variable V 5 r 1=2 T and simplifying for azimuthal symmetry, temperature is a function of r; μ; t:
 
@2 V 1 @V 1V 1 @  
2 @V 1 @V
1 2 1 1 2 μ 5
@r 2 r @r 4 r2 r 2 @μ @μ α @t
With separability
V ðr; μ; φ; tÞ 5 ΓðtÞRðrÞMðμÞ
the ODEs are
 
d  
2 @M
12μ 1 nðn 1 1ÞM 5 0
dμ @μ
" #
d 2 R 1 dR 1 21
1 1 λ 2 n1
2
R50
dx2 2 dr 2 r2

with elementary solutions

Pn ðμÞ and Qn ðμÞ
Jn11=2 ðλr Þ and Yn11=2 ðλr Þ
where Pn ðμÞ and Qn ðμÞ are Legendre functions of the first and second kinds, respectively.
dΓ
1 αλ2 Γ 5 0ΓðtÞ:e2αλ τ
2

dt
The above can also be written as, with x 5 rsinθcosϕ, y 5 rsinθsinϕ, z 5 rcosθ, and f ðr; θ; [Þ 5 RðrÞΘðθÞΦðϕÞ, the
equations are
@2 Φ
5 2 m 2 Φðϕ Þ
@ϕ2
1 @ @Θ m2
sinθ 2 2 ΘðθÞ 5 2 lðl 1 1ÞΘðθÞ
sinθ @θ @θ sin θ
1 @ @R l ð l 1 1Þ
r2 2 R 5 λ α Rð r Þ
r @r
2 @r r2
 Mathematical foundations Chapter | 4 163

where ΘðθÞΦðϕÞ 5 Yl;m ðθ; ϕÞ. Here, l; m are integers, and l $ jmj, or 2m # l # m, and
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2l 1 1 ðl 2 mÞ! m 
Yl;m ðθ; ϕÞ 5 P cosθÞeimϕ
4π ðl 1 mÞ! l

where Yl;m are spherical harmonics and Pm

l are associated Legendre functions. Note that
 
2 m=2 d
m
Pm
l ð x Þ 5 ð 21 Þ m
12x Pl ð x Þ
dxm
For azimuthal symmetry, m 5 0 (azimuthal symmetry) Pm l ðxÞ 5 Pl ðxÞ. The orthogonality condition gives the orthogonal-
ity condition
ð1
2 ðn 1 mÞ!
l ðxÞPn ðxÞdx 5
Pm δln
m
21 2n 1 1 ðn 2 mÞ!
The spherical harmonics functions are orthogonal and normalized; being complete, any function can be expressed in
terms of spherical harmonics. The solution can thus be written in the form:
1 
fα ðr; θ; [Þ 5 Yl;m ðθ; [Þ pffiffi Jl112 jl11=2;n r=aÞ
r
where jl11=2;n represents the nth zero of the Bessel function Jl112 . α 5 ðn; l; mÞ; n 5 1; 2; 3; ?.
Legendre’s equation is
1 @ @Θ
sinθ 1 lðl 1 1ÞΘðθÞ 5 0;
sinθ @θ @θ
which, for x 5 cosθ, can be expressed as an eigenvalue equation
d  d
1 2 x2 ΘðxÞ 5 2 lðl 1 1ÞΘð xÞ
dx dx
for 21 # x # 1, has solutions Pl ðxÞ, called Legendre polynomials, defined by Rodriguez’ formula (Abramowitz &
Stegun, 1964)
1 d l  2 l
Pl ðxÞ 5 x 21 :
2l l! dxl
These polynomials are complete and have the orthogonality condition:
ð1
2
Pl ðxÞPm ðxÞdx 5 δlm
21 2l 1 1
Pl ð 2 xÞ 5 ð 21Þl Pl ðxÞ:
Note that for the normalization Pn ð1Þ 5 1 for Legendre polynomials given in Table 4.2, it is easily seen for these
cases. The recurrence relation (Abramowitz & Stegun, 1964)
ðn 1 1ÞPn11 ðxÞ 2 ð2n 1 1ÞxPn ðxÞ 1 nPn21 ðxÞ 5 0

TABLE 4.2 Legendre polynomials.

l Pl ðμÞ l Pl ðμÞ
0 1 1 μ
   
2 3μ 2 1 2 5μ 2 3μ
2 1 2 3 1 3
 
4 1
8  35μ4 2 30μ2 1 3 5 1
63μ5 2 70μ3 1 15μ
 8  
6 1
16  231μ6 2 315μ4 1 105μ2 2 5 7 16 429μ 2 639μ 1 315μ
1 7 5 3
2 35μ

128 6435μ 2 12012μ 1 6930μ 2 1260μ2 1 35
8 t 8 6 4
164 Nuclear Engineering

can be used to obtain higher terms. Since the Legendre polynomials form a complete set, a function such as gðz; μÞ can
be expressed as
X
N
gðz; μÞ 5 gl ðzÞPl ðμÞ
l50

ð1
2l 1 1
gl ð z Þ 5 gðz; μÞPl ðμÞdμ
2 21

with moments gl ðzÞ. We can thus express gðz; μÞ as:  gðz; μÞ5g0 ðzÞ; First-order angular: gðz; μÞ5g0 ðzÞ 1 μg1 ðzÞ;
 Isotropic:
Second-order angular: gðz; μÞ5g0 ðzÞ 1 μg1 ðzÞ 1 12 3μ2 2 1 g2 ðzÞ and higher orders.

Exercise 4.5: Show the connection between the spherical geometry equation for 1-D spherical geometry.

@2 T 2 @T 1 @T
1 5
@r 2 r @r α @t
and the one-dimensional rectangular geometry equation.

4.2.5 Flux formulation

In the flux formulation of the 1-D time-dependent equation (Ozisik, 1993)
@2 T 1 @T ðx; tÞ
5
@x 2 α @t
can be written in terms of the heat flux
@Tðx; tÞ
qðx; tÞ 5 2 k
@x
as
@2 q 1 @qðx; tÞ
5
@x 2 α @t
ð0 , x , N; t . 0Þ with boundary conditions
qðx; tÞ 5 f0 5 constant at x 5 0; t . 0
qðx; tÞ 5 0 for t 5 0With Qðx; tÞ 5 qðx; tÞ 2 fo
@2 Q 1 @Qðx;tÞ
@x2 5 α @t in 0 , x , N; t . 0
Qðx; tÞ 5 2f0 5 constant at x 5 0; t . 0
Qðx; tÞ 5 0 for t 5 0 and

x
Qðx; tÞ 5 2fo erf pffiffiffiffiffiffiffi
4αt
 
x
qðx; tÞ 5 fo 1 Qðx; tÞ 5 fo 1 2 erf pffiffiffiffiffiffiffi
4αt

and

x
qðx; tÞ 5 fo erfc pffiffiffiffiffiffiffi
4αt

From qðx; tÞ, the temperature distribution is found by integrating the above
ðN
fo y
T ðx; tÞ 5 erfc pffiffiffiffiffiffiffi dy
k x 4αt
 Mathematical foundations Chapter | 4 165

and
" #
1=2
2f0 αt 2x2 =4αt x x
T ðx; tÞ 5 e 2 erfc pffiffiffiffiffiffiffi
k π 2 4αt

Exercise 4.6: Compare the heat flux formulation with the Fermi Age formulation and comment on the similarities or
differences.

4.3 Integral equations

Several problems in science and engineering, such as the Boltzmann transport equation, are formulated in terms of
integral equations where the unknown appears in an integral (Kanwal, 2013; Polyanin & Manzhirov, 2008). In general,
a differential equation can be converted into an integral equation. It is important to note that initial value problems as
well as boundary value problems can be converted into equations.
Integral equations are classified according to the limits on the integral, the occurrence of the unknown function and
the homogeneity of the equation. Some equations with their classification are listed in Table 4.3.
In the above equations, Kðx; yÞ is called the “kernel of the integral operator.”

Exercise 4.7: Show that the IVP.

d2 y dy 0

2
2 5 1 6y 5 0; yð0Þ 5 0; y ð0Þ 5 2 1
dx dx
can be converted to a Volterra equation of the second kind
ðx
gðxÞ 5 ð6x 2 5Þ 1 ð5 2 6x 1 6tÞgðtÞdt
a

Exercise 4.8: Show that the BVP.

d2 y
1 xy 5 1; yð0Þ 5 0; yð1Þ 5 1
dx2
can be converted to a Fredholm integral equation of the second kind.
ð1
1
yðxÞ 5 xð1 1 xÞ 1 K ðx; tÞyðtÞdt
2 0

Obtain the kernel K ðx; tÞ.

TABLE 4.3 Classification of integral equations.

Equation Name Type

Ðb    
Fredholm Homogeneous I
uðx Þ 5 a K x; y f y dy
Ðb    
Fredholm Homogeneous II
f ðx Þ 5 a K x; y f y dy
Ðb    
Fredholm Inhomogeneous II
f ðx Þ 5 a K x; y f y dy 1 gðxÞ
Ðx    
uðx Þ 5 a K x; y f y dy Volterra Homogeneous I
Ðb    
Volterra Homogeneous II
f ðx Þ 5 x K x; y f y dy
Ðx    
f ðx Þ 5 a K x; y f y dy 1 gðxÞ Volterra Inhomogeneous II
166 Nuclear Engineering

The above equations are called nonsingular linear integral equations since the unknown function appears inside one
or more integrals and linear operations are performed on it. Examples of nonlinear inegral equations are
ðb
yðxÞ 5 K ðx; tÞ½yðtÞ2 dt
a
ðb ðb
yðxÞ 5 K ðx; s; tÞyðsÞyðtÞdsdt; a # s # b
a a

If the function to be determined in the above equations, f ðxÞ, appears as hðxÞf ðxÞ and hðxÞ is neither zero or one, then
the equation is called a linear integral equation of the third kind.
In this section, the integrals considered are nonsingular, linear, and either homogeneous or nonhomogeneous.
Some of the methods to solve integral equations are:
1. differentiating under the integral sign using Leibnitz’s rule
2. using successive approximations
3. using a resolvant kernel
4. Laplace transform for an integral of the convolution type
The formula for Leinnitz’s rule is
ð ð bðxÞ
d bðxÞ @F db da
F ðx; tÞdt 5 dt 1 F ½x; bðxÞ 2 F ½x; aðxÞ
dx aðxÞ aðxÞ @x dx dx
A Laplace transform method is used for an integral equation of the convolution type
ðb
yðxÞ 5 f ðxÞ 1 K ðx 2 tÞyðtÞdt
a
ðx
yðxÞ 5 f ðxÞ 1 λ K ðx 2 tÞyðtÞdt
a

where the convolution theorem holds for transforms such as Laplace and Fourier transforms and algebraic properties
apply, that is, additive, associative, distributive and multiplicative.
One way of solving integral equations is by integral transforms, for example, for Fredholm integrals, when
K ðx; yÞ 5 Kðx 2 yÞ, the Fourier transform can be used while for Volterra equations, the Laplace transform can be used.

Example 4.2: Fredholm Equation by successive approximations.

ðb
f ðxÞ 5 gðxÞ 1 kðx; yÞf ðyÞdy
a

Write as a fixed-point equation to define a transformation operator

Tf 5 f
where the operator T is
ðb
Tf ðxÞ 5 gðxÞ 1 kðx; yÞf ðyÞdy
a

By the contraction Mapping Theorem, the integral equation has a unique solution in C([a, b]).
The successive approximations are
Tf0 5 g 1 Kf0
T f0 5 T ðTf0 Þ 5 T ðg 1 Kf0 Þ 5 g 1 K ðg 1 Kf0 Þ 5 g 1 Kg 1 K 2 f0
2

 
T 3 f0 5 T T 2 f0 5 g 1 Kg 1 K 2 g 1 K 3 f0
^
T f0 5 g 1 Kg 1 K g 1 ? 1 K n21 gn21 1 K n f0
n 2 2
 Mathematical foundations Chapter | 4 167

Since
lim T n f0 5 0
n-N

the function is obtained as

X
N
f 5 lim T n f0 5 K n g:
n-N
n50

As an example, consider the inhomogeneous Fredholm equation

ð1
f ðxÞ 5 1 1 xf ðyÞdy
0

Write the upper limt as α to demonstrate convergence; then

ðα
f ðxÞ 5 1 1 xf ðyÞdy
0

for 0 , α , 1.
With any starting approximation f0 Acð½0; 1Þ successive approximations
 ðα
Tf xÞ 5 1 1 xf ðyÞdy
0

are obtained. Let f0 ðxÞ 5 1. Then

ðα ðα  
1
f2 ðxÞ 5 Tf1 ðxÞ 5 1 1 xf1 ðyÞdy 5 1 1 xð1 1 αyÞdy 5 1 1 x α 1 α3
0 0 2
ðα ðα
1
f3 ðxÞ 5 Tf2 ðxÞ 5 1 1 xf2 ðyÞdy 5 1 1 x 1 1 y α 1 α3 dy
0 0 2
 
1 3 1 5
511x α1 α 1 2 α
2 2
Continuing, the nth approximation is
 
1 1 1
fn ðxÞ 5 1 1 x α 1 α3 1 2 α5 1 . . . 1 n21 α2 11
n21

2 2 2
So that the solution may be reconstructed as
XN 1
f ðxÞ 5 limn-N fn ðxÞ 5 1 1 x n50 2n
α2n11

XN α2
n
1
5 1 1 xα 5 1 1 xα ∙
n50 2n 1 2 α2 =2

which gives the solution

2α
f ð xÞ 5 1 1 x 5 1 1 2x:
2 2 α2
 pffiffiffiffipffiffiffi
This satisfies the starting integral equation. In the solution, the denominator
pffifficonverges
ffi for αA 2 2; 2 ; the solu-
tion is not valid for and furthermore that the solution satisfies for any α 5 6 2.

Example 4.3: Solve the Fredholm integral equation.

ð1
yðxÞ 5 e 1 λ
x
2ex et yðtÞdt
0
168 Nuclear Engineering

Let
ð1
C5 et yðtÞdt
0

Then yðtÞ 5 et ð1 1 2CλÞ, and using this, the value of C can be found as
C 5 2½1 1ðeλðe
2 1Þ
2
2 2 1Þ ; where λ 6¼ 1=ðe 2 1Þ
2

The solution is
x
yðxÞ 5 1 2 λðe
e
2 2 1Þ ; where λ 6¼ 1=ðe 2 1Þ
2

A resolvent kernel Rðx; t;λÞ can be obtained for both nonhomogeneous Fredholm and Volterra integrals to remove
the unknown function from under the integral sign; for a Fedholm integral equation
ðb
yðxÞ 5 f ðxÞ 1 λ kðx; tÞyðtÞdt
a

the equation then becomes

ðb
yðxÞ 5 f ðxÞ 1 λ Rðx; t;λÞf ðtÞdt
a

which is easily solved since f(x) is given.

Then this resolvent kernel satisfies the following relationship:
ðb
Rðx; t;λÞ 5 K ðx; tÞ 1 λ kðx; sÞRðs; t;λÞds
a

giving
XN
Rðx; t;λÞ 5 m51
λm21 Km ðx; tÞ

Exercise 4.9: Classification of an integral equation.

The integral equation for Chandrasekhar’s H function (Chandrasekhar, 1960) appears in the exact solution of the
neutron transport equation which will be used in Chapter 6. How would you classify this equation?
ð1  0 
ψ μ  0 0
H ðμÞ 5 1 1 μH ðμÞ 0 H μ dμ
0 μ1μ

Exercise 4.10: Fredholm equation by successive approximations. Consider the integral equation
ð
3 x 1 x 1 1 1
yðxÞ 5 e 2 xe 2 1 tyðtÞdt:
x 2 2 2 0
Let

K1 ðx; tÞ 5 K ðx; tÞ 5 t

so that the kernel is

ð1
Km ðx; tÞ 5 Kðx; sÞKm21 ðs; tÞds
0

show that
ð1 m21
1
Km ðx; tÞ 5 K ðx; sÞKm21 ðs; tÞds 5 t
0 2
 Mathematical foundations Chapter | 4 169

and the resolvant kernel is

XN XN 1 m21
1 m21
Rðx; t;λÞ 5 λm21 Km ðx; tÞ 5 t
m51 m51 2 2
" #
XN 1 m21
1 1 2
4t
5 t5t 11 1 1? 5
m51 4 4 4 3

from which
ð1
yðxÞ 5 f ðxÞ 1 λ Rðx; t;λÞf ðtÞdt
0

and the solution is easily obtained as

3 1 e
yðxÞ 5 ex 2 xex 2 1 1:
x 2 3
Ðx
Exercise 4.11: Solve the Volterra equation uðxÞ 5 0 ex2y uðyÞdy, given the boundary condition uð0Þ 5 0 by differentiat-
ing with respect to x.

Exercise 4.12: Solving the Volterra equation using the convolution theorem.
Ðx
Solve the Volterra equation uðxÞ 5 f ðxÞ 1 0 kðx 2 yÞuðyÞdy by taking the Laplace transform.

4.3.1 An important integral equation for neutron transport

In the previous section, the integral equation of a Volterra form was obtained from a PDE
ð
1 x 2x2y
φðx; μÞ 5 e μ f ðxÞdy 1 Ce2x=μ (4.6)
μ 0
When this equation is integrated over μ, the equation can be written in operator form as

f ðxÞ 5 K^ ðx 2 yÞf ðxÞ 1 gðxÞ (4.7)

where
ð ð
1 x 2x2y
K^ ðx 2 yÞf ð xÞ  dμ e μ f ðxÞdy
μ 0
and
ð
gðxÞ  dμCe2x=μ :

We will return to this form in the chapter on Transport Theory to obtain a numerical solution to an important bench-
mark in nuclear engineering.

4.3.2 Integral equations in neutron transport

Several “special” functions (Abramowitz & Stegun, 1964) are used to represent source distributions and neutron flux
especially for the separation of variables to facilitate exact solution or numerical procedures.
The delta function δðxÞ is visualized to be zero everywhere and have a large value at x 5 0. In engineering, espe-
cially in the context of a neutron source, it is used to represent a source at some spatial position, angle or time. As an
example a source defined as
Sðx; E; Ω; tÞ 5 δðx 2 x0 ÞδðE 2 E0 ÞδðΩ 2 Ω0 Þδðt 2 t0 Þ
170 Nuclear Engineering

represents a unit source located at position x0 with energy E0 , emitting neutrons in the solid angle dΩ0 about Ω0 at
time t0 . The response to this unit point instantaneous source is the Green’s function which can be used to find the flux
distribution from a general source.
Among the several definitions of the δ function, consider the normal distribution with an infinitesimal variance σ2 :
1 x2
δðxÞ 5 lim pffiffiffiffiffiffi e22σ2
σ-0 σ 2π

and a variation, used in complex integration, of the form:

1 ε
δðxÞ 5 lim 2
π x 1 ε2
ε-0

One of the properties of the δ functions is

ðN
f ðxÞδðx 2 aÞdx 5 f ðaÞ
2N

4.4 Integro-differential equations

An integro-dfifferential equation has the unknown function in the derivative term as well as under the integral sign. An
example is the Volterra integro-differential equation
ðx
dn φðxÞ
5 f ð x Þ 1 λ K ðx; yÞφðyÞdy (4.8)
dxn 0

From Table 4.3 this is classified as an Volterra equation of the second kind since the unknown function φðxÞ is on both
sides of the equation in the derivative and under the integral sign, and the presence of f ðxÞ makes it nonhomogeneous.
Similar to the successive approximation methods described above, the Adomian decomposition method can be used to
determine the solution from recurrence relations when an exact solution is not possible.
Consider the equation
ðx
dφðxÞ
5 1 2 φðyÞdy
dx 0

with the initial condition φð0Þ 5 0. The simplest method is to apply a Laplace transform

1 ΦðsÞ
sΦðsÞ 2 Φð0Þ 5 2
s s
Rearrranging

1
ΦðsÞ 5
ðs2 1 1Þ
giving the solution as φðxÞ 5 sinx. There are several other methods such as the variational iteration method and the
Laplace transform method as applied here as well for the convolution integral. Now consider the first-order PDE (Bell
& Glasstone, 1952)

@
μ φðx; μÞ 1 φðx; μÞ 5 f ð xÞ 1 Sð x; μÞ
@x
Ð1
where f ðxÞ 5 2c 21 dμ0 φðx; μ0 Þ. A solution for this can be obtained by multiplying both sides by the integration factor
e2x=μ and integrating over x:
ð
1 x 2x2y
φðx; μÞ 5 e μ f ðxÞdy 1 Ce2x=μ
μ 0
where C, a constant of integration, can be obtained from the given boundary condition.
 Mathematical foundations Chapter | 4 171

To prepare for the mathematical formulation to come in the following chapters, consider an integro-differential
equation of the form
ð
@ c 1
μ 1 1 φðx; μÞ 5 dμ0 φðx; μ0 Þ 1 Sðx; μÞ:
@x 2 21
A general solution is obtained in the classical albedo problem for a semiinfinite half-space xAð0; NÞ, shown in Fig. 4.6,
for an incident source φð0; μÞ 5 1 at the free boundary z 5 0. For c # 1 (nonmultiplying medium) and without an extraneous
source, that is, Sðx; μÞ 5 0, this problem was considered originally for radiative transfer (Chandrasekhar, 1960) (Fig. 4.8).
Note that transport is along the s direction although the equation is written for the x variable, that is, it is a one-
dimensional representation of two-dimensional transport. The connection between the variables is
x 2 x0
s5
μ
and, since the contribution to the flux [ðx; μÞ will come from all x0 , we need to connect the variable through the
transformation
d @x0 @ @
5 52μ 0
ds @s @x0 @x
The equation thus becomes
 
d c
2 1 φðx0 ; μÞ 5 2 φð x0 Þ
ds 2
where
ð1
φðx0 Þ  dμ0 φðx0 ; μ0 Þ
21
Multiplying by the integrating factor e2s and integrating from 0 to s
ð
2x2x
0 c x
0
φðtÞe2 μ dt
x2t
φðx; μÞ 5 φðx ; μÞe μ 1
2μ x0
The neutron transport equation, describing the angular flux distribution in a nuclear reactor is an integro-differential
equation in phase space
ðN ð
 0
Ω ∙ rφð~
r ; Ω; EÞ 1 Σt ð~
r ; E; tÞφð~
r ; Ω; EÞ 5 dE0 dΩ0 Σs ð~
r ; E0 ; Ω0 ∙ ΩÞφ ~ r ; Ω0 ; E
0
ðN ð (4.9)
1 0 0 0 0 0
1 χð E Þ dE dΩ νΣf ðE Þφð~ r ; Ω ; E Þ 1 Sð~
r ; Ω; EÞ:
4π 0

FIGURE 4.8 Neutron transport in 1-D.

172 Nuclear Engineering

During reactor operation, the composition Ni ðtÞ of isotopes at any time t is found by solving the integro-differential
burnup equation coupled with the point kinetics equation described earlier in this chapter. This time-dependent integro-
differential equation is
ð 
dNi ðtÞ X N
5 γ ji ðE; tÞσfj ðE; tÞφðE; tÞdE Nj ðtÞ 1 σci21 ðE; tÞφðE; tÞNi21 ðtÞ 1 λi0 Ni0 ðtÞ
dt j 0
 ð N   (4.10)
 
2 σfi ðE; tÞ 1 σci ðE; tÞφðE; tÞdE 1 λi Ni ðtÞ:
0

The composition of an ith nuclide changes with time due to creation in processes a to c and destruction to process d
and e as follows:
a. fission of some isotope j which produces the ith by an amount γ ji ,
ð N 
γ ji ðE; tÞσfj ðE; tÞφðE; tÞdE Nj ðtÞ
0

b. radiative capture of an ði 2 1Þth nuclide:

σci21 ðE; tÞφðE; tÞNi21 ðtÞ
c. radiative decay of another nuclide λi0 Ni0 ðtÞ
d. fission and capture in which some other nuclide is produced
ð N 
 
σfi ðE; tÞ 1 σci ðE; tÞφðE; tÞdE Ni ðtÞ
0

e. its radioactive decay λi Ni ðtÞ

These calculations are coupled with reactor kinetics on the short-term (Bseconds and minutes) as described by the
PKE as well as on the long-term (Bdays and months) with the burnup equations. Such coupled kinetic-dynamics,
which incorporate dynamics with fuel burnup, as illustrated in Figs. 4.9 and 4.10, are used for reactor start-up simula-
tions. The short-term changes in reactivity affect the reactor power during start-up which could take about 5 days to

FIGURE 4.9 Coupling of neutronics with thermal hydraulics.

FIGURE 4.10 Coupling of neutronics with ther-

mal hydraulics and point kinetics.
 Mathematical foundations Chapter | 4 173

reach 100% full power. Consider a model (Johnson, Lucas, & Tsvetkov, 2010) based on the point kinetic equations,
Bateman equations for nuclide transmutation, and a lumped thermal model computing the dynamics of the power and
fuel temperature (Bell & Glasstone, 1979; Hetrik, 1993; Lamarsh & Baratta, 2001). This is a simplified lumped model
which considers the reactor as a point, that is, spatial dependence is ignored.
The flow of the calculation procedure is as follows
1. Initial conditions specified include the fuel vector consisting of atomic densities of actinides Ni ðtÞ, fuel TF and mod-
erator temperatures TM , feedback coefficients αF , αM and source term QðtÞ, and the input reactivity ρ0 ; this gives
the reactivity ρðtÞ in Eq. (4.11).
2. The neutron density nðtÞ and precursor concentrations Ci ðtÞ are determined from Eqs. (4.12) and (4.13), giving the
flux φðtÞ 5 nðtÞv, and reactor power PðtÞ (Eq. 4.14).
3. The new system multiplication k is found from Eq. (4.15), which gives the reactivity ρðtÞ and the change in reactiv-
ity ΔρðtÞ from Eq. (4.16).
4. The fuel temperature TF ðtÞ is updated from the reactor power PðtÞ, using the heat balance Eq. (4.17); the temperature
is then used to calculate the coefficient of reactivity αF ðtÞ from Eq. (4.18).
5. The fuel vector N is updated by solving the Bateman equations (Eq. 4.19) for all the actinides considered in the
model [24 actinides in (Johnson et al., 2010)].
6. The generation time Λ is calculated from the time for absorption ΛN , for absorption or leakage Λ0 , based on the
new fuel vector and system multiplication k, from Eq. (4.20).

ρðtÞ 5 ρ0 1 αF ðTF ðtÞ 2 TF0 Þ 1 αM ðTM ðtÞ 2 TM0 Þ (4.11)

B
dφðtÞ ρðtÞ 2β X6
5 φ ðt Þ 1 λi Ci ðtÞ 1 QðtÞ (4.12)
dt Λ i51

where
dCi ðtÞ β i ðtÞ
5 φðtÞ 2 λi Ci ðtÞ; i 5 1; ?6 (4.13)
dt Λ
and
PðtÞ 5 ER Σf φðtÞV (4.14)
The system multiplication is
υΣf ðtÞ 1
keff ðtÞ 5 (4.15)
Σa ðtÞ 1 1 L2 B2
Where the numerator is calculated from the atomic densities Ni ðtÞ in the fuel vector N
X
υΣf ðtÞ 5 υi σf ;i Ni ðtÞ
i

The reactivity and the change in reactivity are calculated (in $) from
k21 k0 2 k
ρ5 ; Δρð$Þ 5 (4.16)
k βkk0
and
dTF ðtÞ
mcp 5 PðtÞ 2 hAðTF ðtÞ 2 TM ðtÞÞ (4.17)
dt
The temperature coefficients of reactivity are
dρ
αi 5 ; i 5 F; M (4.18)
dTi
The Bateman equations are
dNZA  A21 
5 NZ σc 2 NZA σc φðtÞ 2 λNZA 1 λN (4.19)
dt
174 Nuclear Engineering

and
1 ΛN Λ0
ΛN 5 ; Λ0 5 ;Λ5 (4.20)
vΣa 11L B
2 2 keff
The procedure described above constitutes a large set of ODEs which can be solved using standard methods in
Matlab.
For a more elaborate analysis, the full phase space-dependence of the neutron flux is required and the computational
effort increased by orders of magnitude.

4.5 Numerical methods

Several numerical and computational methods have been used and developed (Bell & Glasstone, 1979; Case & Zweifel,
1967; Lewis & Miller, 1984) to solve neutron diffusion and transport problems and in thermal hydraulics, structural
mechanics and other areas (Hutton, 2004; Lewis, Nithiarasu, & Seetharamu, 2004; Rao, 2017) with the complexity of
realistic engineering systems. These include the Finite Difference Method (FDM), the Finite Element Method (FEM),
the Finite Volume Method (FVM), nodal methods, spectral methods, quadrature-based methods, and hybrid methods.
This section considers two widely used methods viz the FDM and the FEM.

4.5.1 The Finite Difference Method

In a discretization of the differential equations, the entire domain is divided into elements, or meshes, and the solution
is obtained at the node points. In a 1-D model, the domain a # x # b is divided into a finite number of elements N (e.g.,
N 5 3) for which there will be M (e.g., M 5 4) mesh points while in a 2-D model, for example, in the Cartesian x 2 y
domain the node points will be in the x and y dimensions as shown in Fig. 4.7 (Fig. 4.11).
The FDM is applied to the 2-D steady-state heat conduction equation
@2 T @2 T q_
1 1 50
@x 2 @y 2 k
Applying the first-order forms for the first and second derivatives
@T Ti;j 2 Ti21;j
j 5
@x i21=2;j Δx
@T Ti11;j 2 Ti;j
j 5
@x i11=2;j Δx
The second derivative is thus
@T
@2 T @x ji11=2;j 2 @T
@x ji21=2;j
j 5
@x2 i21=2;j ðΔxÞ2
written as a “three node” term
@2 T Ti11;j 22Ti;j 1 Ti21;j
ji;j 5
@x 2
ðΔxÞ2

FIGURE 4.11 Elements and nodes.

 Mathematical foundations Chapter | 4 175

Similarly, in the y-direction

@T Ti;j 2 Ti;j21
ji;j21=2 5
@y Δy
@T Ti;j11 2 Ti;j
j 5
@y i;j11=2 Δy
The second derivative is thus
@T
@2 T @x ji11=2;j 2 @T
@x ji21=2;j
j 5
@y2 i;j ðΔxÞ2
written as a “three node” term
@2 T Ti;j11 22Ti;j 1 Ti;21j
j 5
@y2 i;j ðΔyÞ2
the FDM is, for Δx 5 Δy,
Ti11;j 22Ti;j 1 Ti21;j Ti;j11 22Ti;j 1 Ti;21j q_
1 1 50
ðΔxÞ 2
ðΔyÞ 2 k
which reduces to

q_
Ti11;j 22Ti;j 1 Ti21;j 1 Ti;j11 22Ti;j 1 Ti;21j 1 ðΔxÞ2 5 0
k
or
 
1 q_
Ti;j 5 Ti11;j 1 Ti21;j 1 Ti;j11 1 Ti;21j 1 ðΔxÞ2
4 k

In a square with temperatures given at the four sides and no heat generation (q_ 5 0), as shown in Fig. 4.12.
1
T1 5 ½ TL 1 TU 1 T2 1 T3 
4
with TL 5 300 C; TU 5 200 C; TR 5 400 C; TD 5 100 C the four simultaneous equations in matrix form are
2 32 3 2 3 2 3
4 21 21 0 T1 TL 1 TU 500
6 21 4 0 2 1 76 T2 7 6 TR 1 TU 7 6 600 7
4 21 5 5
0 4 2 1 54 T3 5 4 TL 1 TD 5 4 700 5
0 21 21 4 T4 TR 1 TD 500

FIGURE 4.12 Square plate with Dirichlet boundary conditions.

176 Nuclear Engineering

Solving these equations in Matlab: A 5 [4 21 21 0; 2 1 4 0 21; 2 1 0 4 21;0 21 21 4]; B 5 [500;600;750;500];

T 5 (A\B) gives the results
2 3 2 3
T1 279:17
6 T2 7 6 289:58 7
4T 5 5 4 327:08 5
3
T4 279:17
For a larger number of nodes, the surface temperature can be obtained from the Matlab program listed below, as shown
in Fig. 4.9 for 400 (20 3 20) node points (18 3 18 5 324 interior nodes) (Fig. 4.13) and compared with the exact solu-
tion given in Section (4.2.1).
The boundary conditions are visualized better from Fig. 4.10 (Fig. 4.14).

FIGURE 4.13 Temperature field Tðx; yÞ surface in a square plate.

FIGURE 4.14 Temperature field Tðx; yÞ projection on a square plate.

 Mathematical foundations Chapter | 4 177

4.5.1.1 Matlab program: Finite Difference Method

Steady-state heat conduction in a 2-D square plate with boundary temperatures specified.

% Ch4_FinDif2DSquarePlate.m
% Finite Difference
N = 20;% N-2 interior points
x = linspace(0,1,N);
dx = x(2)-x(1);
y = x; dy = dx;
eps = 1e-6; % convergence criterion
T = zeros(N,N);
T(1,1:N) = 100; %left
T(N,1:N) = 200; %right
T(1:N,1) = 300; %bottom
T(1:N,N) = 400; %top
dt = dx^2/4;
Qdot=zeros(N,N); % heat generation rate
k=1; % thermal conductivity
err = 1; kk = 0;
while err > eps
kk = kk+1;
Told = T;
for i = 2:N-1
for j = 2:N-1
a=(Told(i+1,j)-2*Told(i,j)+Told(i-1,j))/dx^2;
b=(Told(i,j+1)-2*Told(i,j)+Told(i,j-1))/dy^2;
T(i,j) = dt* (a+b+Qdot(i,j)/k)+ Told(i,j);
end
end
error = max(max(abs(Told-T)));
end
x0=0;delx=1/(N-1);
y0=0;dely=1/(N-1);
for j=1:N
y(j)=y0+(j-1)*dely;
for i=1:N
x(i)=x0+(i-1)*delx;
end
end
surf(x,y,T)
set(gca,'FontSize',12)
xlabel('\bf X (cm)','fontsize',14)
ylabel('\bf Y (cm)','fontsize',14)
zlabel('\bf T ({}^oC)','fontsize',14)

Exercise 4.13: Steady-state head conduction (FDM).

With the given Matlab program, obtain the temperature profile in a 4 3 4 cm copper plate (thermal conductivity
k 5 385 W/(m-K)) with heat generation 2000 W at the center.
The greater the number of elements, the more will the computational effort generally resulting in more accuracy.
For large 3-D models of reactor cores, for example, the FDM is used with refinements for both “structured” and
“unstructured” grids. As problems become large, methods such as the FEM and FVM are used.
The resulting algebraic system can be solved by standard matrix methods such as the Gaussian elimination method,
Choleski’s method, Jacobi’s method, the GaussSeidel method, and the Standard Over Relaxation method.
178 Nuclear Engineering

4.5.2 The Finite Element Method

In the FEM (Hutton, 2004; Lewis et al., 2004; Rao, 2017) the domain of interest is first discretized into a finite number
of elements which fit into the specified (regular or irregular) geometry. For illustration, consider a 1-D domain consist-
ing of three elements and four nodes (Fig. 4.15).
The field function, for example, T(x), in an element with nodes i and j is expressed in terms of nodal values Ti and
Tj and shape functions Ni ðxÞ and Nj ðxÞ
 
 Ti
TðxÞ 5 Ni ðxÞTi 1 Nj ðxÞTj 5 Ni ðxÞ Nj ðxÞ
Tj
where the shape functions are shown in Fig. 4.16.
The field equation is then reduced to a set of linear algebraic equations which are solved, by standard methods, for
the nodal values. The equation for each local element is combined or assembled into a global system expressed mathe-
matically as
½KfTg 5 ff g
where ½K is called the global stiffness matrix and ff g is called the force function. The temperature at the nodes is found as
fTg 5 ½K21 ff g
and from the nodal values, the temperature distribution is reconstructed using the shape functions. If the problem is in
2-D, the square and triangular elements shown in Fig. 4.13 can be used (Fig. 4.17).
Some commonly used elements in 1-D are the linear element (Fig. 4.11) with two nodes and the quadratic element
with three nodes. In 2-D, the linear triangular element with three nodes (Fig. 4.13), the quadratic triangular element
with six nodes, the quadrilateral element with four, eight or nine nodes, and the isometric element with eight nodes for
curvilinear coordinates, that is, coordinates which change direction with position, are used. In 3-D, the linear tetrahe-
dron element with four nodes, the linear hexahedron element with eight nodes, the linear prism element with six nodes
and the quadratic tetrahedron element with ten nodes (Fig. 4.14) are used (Fig. 4.18).With TðxÞ 5 a1 1 a2 x applied to
both nodes of the element, coefficients a1 ; a2 can be found for the shape functions of Fig. 4.12.

FIGURE 4.15 A 3-element, 4 node linear domain.

FIGURE 4.16 Shape functions NðxÞ.

FIGURE 4.17 2-D quadrilateral and triangular

elements.
 Mathematical foundations Chapter | 4 179

FIGURE 4.18 The 3-quadratic tetrahedron with ten nodes.

At x 5 xi , Ti 5 a1 1 a2 xi , and at x 5 xj , Tj 5 a1 1 a2 xj , giving the coefficients and the shape functions listed in

Tables 4.3 and 4.4 respectively. The space-dependence of T ðxÞ is
   
Tj 2Ti xj Ti 2 xi Tj
T ðxÞ 5  x1  
xj 2 xi xj 2 xi
For other elements, the coefficients and shape functions are listed in Tables 4.4 and 4.5 respectively.
The gradient variable B for the 1-D linear element is
" #
dT 1 1  Ti 
5g  2 5 ½B½T
dx l l Tj

We first consider the procedure from a variational formulation where the stationary function of a functional (Lewis,
Vrabie, & Syrmos, 2012) is the governing equation. In the case of the 2-D heat conduction equation, for example, the
functional
ð " # ð ð
1 @T 2 @T 2 @T 2 1
I ðT Þ 5 kx 1 ky 1 kz 2 2GT dΩ 1 qT ds 1 hðT 2Ta Þ2 ds
2 Ω @x @y @z S2 S3 2

is stationary for
@ @T @ @T @ @T
kx 1 ky 1 kz 1G50
@x @x @y @y @z @z
with the following boundary conditions
T 5 TB on surface S1
B @T B @T B
kx @T
@x l 1 ky @y m 1 kz @z n 1 q 5 0 on surface S2
B @T B @T B
kx @T
@x l 1 ky @y m 1 kz @z n 1 hðT 2 Ta Þ 5 0 on surface S3
where l, m, and n are surface normal, h is the heat transfer coefficient, k is the thermal conductivity and q is the heat
flux.
The procedure is then as follows:
1. The temperature in each element with r nodes is written as
X
r
Te 5 Ni Ti 5 ½NfTg
i51

2. Nodal values are to be selected to make the integral I(T) stationary

Xn
@I
δI ðT Þ 5 50
i51
@T i
@I
thus for each element @T i
5 0 for i 5 1; 2; ...;n
TABLE 4.4 Coefficients of linear and quadratic polynomials.

Element Form of T(x) a1 a2 a3

1-D T ðx Þ 5 a1 1 a2 x Ti xj 2 Tj xi Tj 2 Ti 
xj 2 xi xj 2 xi
linear
   
1-D T ðx Þ 5 a1 1 a2 x 1 a3 x 2 Ti 1
l 2 3Ti 1 4Tj 2 Tk 2
l2 Ti 2 2Tj 1 Tk
quadratic
        
2-D T x; y 5 a1 1 a2 x 1 a3 y 1
2A ðxj yk 2 xk yj ÞT i 1 xk yi 2 xi yk Tj 1 ðxi yj 2 xj yi ÞTk 1
2A ðyj 2 yk ÞT i 1 yk 2 yi Tj 1 ðyi 2 yj ÞTk 1
2A ðxk 2 xj ÞT i 1 ðxi 2 xk ÞTj 1 ðxj 2 xi ÞTk
linear
triangular
TABLE 4.5 Shape functions for linear and quadratic polynomials.

Element Form of T(x) Ni Nj Nk

h i h i
xj 2 x x 2 xi
1-D T ðx Þ 5 a1 1 a2 x xj 2 xi xj 2 xi 
linear h i h i h i
2x 2 4x 2 2x 2
1-D T ðx Þ 5 a1 1 a2 x 1 a3 x 2 12 3x
l 1 l2
4x
l 1 l2 l2 2 x
l
quadratic       
 
2-D T x; y 5 a1 1 a2 x 1 a3 y 1
2A ðxj yk 2 xk yj ÞT i 1 xk yi 2 xi yk Tj 1 ðxi yj 2 xj yi ÞTk 1
2A ðyj 2 yk ÞT i 1 yk 2 yi Tj 1 ðyi 2 yj ÞTk 1
2A ðxk 2 xj ÞT i 1 ðxi 2 xk ÞTj 1 ðxj 2 xi ÞTk
linear
triangular
182 Nuclear Engineering

3. For an element,
ð " 2 2 2
# ð ð
1 @T @T @T 1
I 5
e
kx 1 ky 1 kz 2 2GT e
dΩ 1 qT ds 1 hðT e 2Ta Þ2 ds
2 Ω @x @y @z S2e S3e 2

4. The first derivative with respect to nodal temperatures is

ð ð ð ð ð
@I e 1 1
5 2½BT ½D½B½TdΩ 2 dΩΩ 2G½NT ½T 1 q½NT ½T 1 h½NT ½Tds 2 h½NT Ta ds 5 0
@½T Ω2 2 S2e S3e S3e

5. The terms are written in compact form as

@T e 2
@T e 2
@T e 2
 T  
kx 1 ky 1 kz 5 g ½D g
@x @y @z
where
8 e9
> @T >
>
> >
>
3>
> @x >
>
>
>
( )2 > e>
> >
 T   @T e @T e @T e kx 0 0 < @T =
g ½D g 5 40 ky 0 5 @y :
@x @y @z > >
0 0 kz >
>
>
>
e>
>
>
> @T >
>
>
> >
: @z >;

The algebraic system obtained is

½KfTg 5 ff g
where
ð ð
½K 5 ½BT ½D½BdΩ 1 h½NT ½Ns
Ω S3

and
ð ð ð
½f  5 G½NT dΩ 2 q½NT ds 1 hTa ½NT ds:
Ω S2 S3

This procedure is demonstrated for a three-element, four-node wall of material, with a Neumann boundary condition
on the left wall and a convective boundary condition on the right wall, the various terms are:
2 3
1
ðl ðl 2 " #  
6 l7
6 7 21 1 kA 1 21
dΩ½BT k½B 5 kA dx6 7 5 :
0 0 4 1 5 l l l 21 1
l

Since the heat generation is zero, G 5 0 and

ð ð
½f  5 2 q½NT ds 1 hTa ½NT ds
S2 S3

the first term will appear only in the first element at which the Neumann boundary condition applies; similarly, the sec-
ond term will appear only in the third element since the convective heat loss condition appears at the last node.
The element matrices are:for the first element:
   
kA 1 21   qA
½K 1 5 ; f 15
x1 2 1 1 0
 Mathematical foundations Chapter | 4 183

for the second element:

   
kA 1 21   0
½K 2 5 ; f 25
x2 21 1 0
and for the third element:
   
kA 1 21   0
½K 3 5 ; f 35
x3 21 1 hATa
giving a 4 3 4 global matrix
2 32 3 8 9
1 21 0 0 T1 > qA >
kA 6 2 1 < =
2 21 0 76 T2 7 0
4 0 21 5 4 5 5
x 2 21 T3 : 0 >
> ;
0 0 21 1 1 hA T4 hATa
which can be solved to compute the temperatures Ti at the nodes.

Exercise 4.14: For a rod of length 10 cm, the temperature at the left and right nodes is Ti 5 100 C and Tj 5 200 C.
Using the shape functions above, sketch the temperature variation in the element and calculate the temperature at a
point 6 cm from the left node of the bar.

Exercise 4.15: For the above rod and end temperatures, use a 1-D quadratic element for which i, j, k are at 0, 5 10 cm
from the left node to calculate the temperature at a point 6 cm from the left node of the bar. Compare the two tempera-
ture profiles.

Exercise 4.16: For the triangular element specified below, write the expression for T(x, y) and plot the surface
temperatures.
Coordinate T (C) X (cm) Y (cm)
i 100 1 2
j 200 2 1
k 300 3 3

From the above elements and shape functions, it is clear that if the nodal values are known, then the function can be
found for the entire element. From a single element, describe how you would solve the 8-element square plate problem
illustrated in Fig. 4.19.
For the neutron diffusion equation, the functional
ðb (  2 )
dφ 2
J ðφÞ 5 2 DðxÞ 1 Σa ðxÞ φðxÞ 2 2Sð xÞφð xÞ dx
a dx

FIGURE 4.19 2-D triangular elements: square plate with Dirichlet boundary conditions.
184 Nuclear Engineering

gives an extremum that satisfies the governing equation. This extremum can be determined by varying φ to form
the quantity J ðφ 1 δφÞ 2 JðφÞ. The procedure is, as described above, to express the flux in terms of “shape
P
functions” MðxÞ, so that φðxÞ 5 Ni51 Ai Mi ðxÞ. The unknown coefficients Ai are determined from the extremum
condition
@J
50
@Ai
for all elements i 5 1; 2; 3; . . . N.
An important equation in nuclear reactor theory is the eigenvalue equation

^ 5 1 νΣf φ
Lφ
ki
which represent the loss terms on the left hand side and the gain terms on the right hand side; their are several values
or modes for which the equation is balanced. The largest eigenvalue is the system multiplication keff of a reactor for
which the corresponding eigenfunction is the stationary flux distribution.
In a simplified one-group model, the eigenvalue form of the neutron diffusion equation is
 
d d 1
2D 1 Σa φðxÞ 5 νΣf φðxÞ 1 SðxÞ; 0,x,L
dx dx k

with albedo boundary conditions

J ð0Þ 5 2 αL φð0Þ; J ðLÞ 5 2 αL φðLÞ
where
dφ
J 52D :
dx
By multiplying the above with Legendre polynomials for linear shape functions Pl ðNi ðxÞÞ and integrating over an
element results in spatial balance equations for each element. The resulting system of simultaneous equations
5
Ax 5 B
is solved for the vector x comprising nodal and cell-averaged fluxes and currents. The numerical results of this linear
FEM have been determined to be more accurate for coarser grids than FDM for fixed-source problems (Brandao,
Dominguez, & Iglesias, 2011).
In nuclear reactors with hexagonal cells, discretization with FEM or spectral methods is based on dividing hexagons
into equilateral triangles and assuming that the neutron cross sections remain constant on each triangle. In the colloca-
tion method (González-Pintor, Ginestar, & Verdú, 2008) the flux on each triangle φe , in the eigenvalue equation, is
expanded as
X
φe ðx; yÞ 5 φe;i gij ðx; yÞ
i;j

where the spatial dependence is in the gij ðx; yÞ polynomial basis functions. The pseudo-spectral methods, based on mod-
ified Dubiner’s polynomials for the basis functions, is applied to a 2-D IAEA reactor benchmark problem to compute
eigenvalues and power distribution in the core.
Spectral methods can be used to obtain an approximate solution to differential equations in terms of a finite number
of basis functions. Both spectral and FEMs are similar in the sense that some shape functions are used; the difference
is that in spectral methods, the basis functions are global while in FEM, they are local, that is, within the nodes of an
element.
One application of spectral methods is illustrated for the 2-D linear steady-state transport equation
 pffiffiffiffiffiffiffiffiffiffiffiffiffi @  ð1
@ 1
μ 1 12μ 2 1 Σt ðx; yÞ φðx; y; μÞ 5 ½Σs 1 νΣf  dμ0 φðx; y; μ0 Þ 1 Sð x; μÞ
@x @y 2 21

in a rectangular domain 21 # x; y # 1, 21 # μ # 1 with given boundary conditions.

 Mathematical foundations Chapter | 4 185

The angular flux is expanded in terms of Chebyshev polynomials of the first kind Ti ðyÞ which are orthogonal to a
weight function wðxÞ
8
ð1 < 0: i 6¼ j
Ti ðxÞTj ðxÞwðxÞdx 5 π=2; i 5 j 6¼ 0
21 :
π; i 5 j 5 0
P
The expansion is a function of space (y variable), φðx; y; μÞ 5 Ni50 φi ðx; μÞTi ðyÞ is then used, after some algebraic
to reduce the 2-D equation into a set of N one-dimensional equations.
Another numerical approximation for the integral (Section 4.4.10) is the use of the quadrature rule, as in the discrete
ordinates (SN Þ method, to separate μ from the angular flux φðx; y; μÞ
ð1 X
0  0
N
dμ φ x; y; μ B wn φi ðx; yÞ
21 i51

The convergence of such a combined spectral-SN method has been illustrated for the 2-D linear isotropic scattering
transport equation (Asadzadeh & Kadem, 2006) which reduced to a system of 1-D equations.
For unstructured meshes, in the case of complex geometries, FEM and the FVM are used in 2-D and 3-D core calcu-
lations. In the FVM, the transported quantity is conserved within a volume; partial currents and surface fluxes are con-
tinuous at the boundaries of adjacent volumes. This method has been applied for 2-D and 3-D Light Water Reactors by
Bernal, Miró, Ginestar, and Verdú (2014) demonstrating better accuracy and low computational times for 2-D results as
compared to 3-D results.

4.6 Approximate methods

4.6.1 The Ritz method
The heat balance equation for a fin (Lewis et al., 2004) written as
d2 θ
2 μ2 θ 5 0
dζ 2
With boundary conditions: x 5 0; dT=dx 5 0 and at x 5 L; T 5 Tb , with ðT 2 Ts Þ 5 θ; ζ 5 x=L; kA hP
5 m2 and m2 L2 5 μ2 .
Pn
At ζ 5 0; dθ=dζ 5 0 and at ζ 5 1; θ 5 θ0 and with ðT 2 Ts Þ 5 θ is solved with the approximation T 5 i51 ai Ni ðxÞ
substitute into the equation and residual R.
Then a function θðζÞ which satisfies the b.c. is selected

θðζÞ  
5 1 2 1 2 ζ2 B
θb

where B is an unknown parameter to be determined by requiring that

ð1
d2 θðζÞ
2 μ2 θ dζ 5 0:
0 dζ 2

This determines the constant B

μ2
B5 2
μ2
:
11 3

to yield the approximate solution

θðζÞ 3 
5 1 2 1 2 ζ2 :
θb 4
186 Nuclear Engineering

4.6.2 The RayleighRitz variational method

Consider the second-order ODE
d2 θðζÞ
2 μ2 θ 5 0
dζ 2
with boundary conditions
dθð0Þ
5 0 and θð1Þ 5 θb :
dζ
The variational functional for the govening equation is
ð1 " 2
#
1 dθ
I5 1 μ θ dζ
2 2
02 dζ

The θ profile that minimizes the integral I is the solution to the governing equation. The procedure is the same as
 
above; a solution is assumed θðζÞ @I
θb 5 1 2 1 2 ζ B, and setting @B 5 0, gives
2

μ2
B5 2
1 1 25 μ2

and the temperature profile

μ2
θðζÞ  
5 1 2 1 2 ζ2 2
:
θb 1 1 25 μ2

Note that the solution found is different from that with the Galerkin method.

4.6.3 The weighted residual method

P
With T  T 5 ni51 ai Ni ðxÞ, the coefficients ai are determined by substituting the approximation and using the residual
with weighting functions wi to get
Ð
Ω ωi ðxÞRdx 5 0 with i 5 1; 2; ...;n
Several weighting functions can be used, for example, in the collocation method ωi 5 δðx 2 xi Þ, in the Galerkin
method ωi ðxÞ 5 Ni ðxÞ and in the least-squares method, ωi 5 @R=@ai

Exercise 4.17: Use an approximate result for θðxÞ from above in the Galerkin method to obtain a solution. Compare all
three solutions for the given heat transfer equation.

4.7 The adjoint function

Mathematical operators have “adjoints.” Thus an operator
@2 @
L^  2 1 13
@x @x
would have an adjoint operator written as
1 @2 @
L^  2 2 13
@x @x
The adjoint of a derivative is its negative. A “forward” equation of the form
^ ðxÞ 5 SðxÞ
Lφ
 Mathematical foundations Chapter | 4 187

would have a “backward,” or “adjoint” form expressed as

1
L^ φ1 ðxÞ 5 S1 ðxÞ

Both equations would be solved with the appropriate boundary conditions.

In nuclear engineering, φ1 is the adjoint neutron flux and is also called the “importance” function. For a critical
reactor, the diffusion equation reads
^ ðxÞ 5 0
Lφ
where
d2
L^  2 1 B2
dx
1
The adjoint operator L^ is defined as
ð a=2 ð a=2
1
^
uLvdx 5 vL^ udx
2a=2 2a=2

where u and v are any two functions which vanish at the physical boundaries. The above is also written in the form of
an inner product as
 
^ 5 ðv; L^ 1 uÞ
u; Lv
or equivalently as
 
^ 5 ðL^ 1 u; vÞ
u; Lv
For the second-order “diffusion” operator, it is readily shown, by carrying out an integration by parts, that the operator
is self-adjoint, that is
1
L^ 5 L^
The solutions, ðxÞand φ1 ðxÞ, to both the forward and backward homogeneous equations respectively, are thus pro-
portional to each other.

4.8 Random processes, probability, and statistics

Modeling and simulation in nuclear engineering requires mathematical formulation of coupled systems with the central
quantity being the neutron flux φðPÞ in phase space Pðr; E; Ω;tÞ ^ where for a Cartesian coordinate system, r 5 rðx; y; zÞ
^ ^
and Ω 5 Ωðθ; ϕÞ and thus P is defined by seven variables. Thus, a neutron transport FDM, FVM or FEM computation
with 103 points along each x; y; z axis, 10 angular bins, and 10 energy groups would require 1012 node point calculations
in each time interval. Such intensive computational efforts led to alternate ways of modeling neutron transport. Another
way of looking at nature is the probabilistic way where things happen randomly. The debate between scientists such as
Einstein with his famous comment to Max Born that “God doesn’t play dice” supported a deterministic model of nature
an opposed to quantum physics where Planck, Heisenburg, Schrodinger, Dirac and others were creating a probabilistic
interpretation. A similar view of transport phenomena took shape as a random process where particles interact in a
probabilistic manner. The Monte Carlo (MC) method, based on laws of probability, was developed in the 1940s for the
simulation of neutron transport and with advancements in computer hardware and computation, MC simulation has
since been extended to several areas science, engineering, and the social sciences.
The following sections are intended to cover the concepts and mathematics as described for MC simulation of alpha
particle transport (Section 4.1.61), electron transport (Section 4.1.62), gamma transport (Section 4.1.63), and neutron
transport for criticality (Section 4.2.10). The later part of this book focuses on the applications of MC simulation in neu-
tron and photon transport applied to nuclear reactors and systems.
The topics covered in these sections are random processes, random variables, probability distribution functions
(PDFs), cumulative distribution functions (CDFs), random numbers, sampling methods, and accuracy and precision of
MC methods.
188 Nuclear Engineering

4.8.1 Random processes

Random, or stochastic, processes are considered as an abstraction of events that happen and are observed as a collection
of random variables such as Brownian motion (named after Robert Brown) of particles suspended in a liquid. Such pro-
cesses, as shown below, are observed across science and engineering in signal processing, telecommunications, particle
transport and in financial systems such as the fluctuations of the stock market.
Particle transport is considered to consist of random events as particles are “born” and travel in matter colliding
with host nuclei, exchanging energy and undergoing nuclear interactions until they “die” by some capture or leakage
process. The events are considered to be random as they are determined from probabilistic interaction data (Fig. 4.20).

4.8.2 Markovian processes

A Markovian process is stochastic in nature and has a future state that depends only on the present state without any
dependence on its’ past history. In the above Wiener process, for example, the “standard” Wiener continuous process
Wt begins at zero, has independent increments in which the future is independent of the past, and has Gaussian incre-
ments with mean zero and variance 1 as will be explained in the following sections. In the above, the time step is taken
pffiffiffiffiffiffi
Δt 5 T=N where T 5 1s and N 5 10; 000 steps. With j as the time index and W0 5 0, Wt 5 Wt21 1 ξ Δt. The ran-
dom variable ξ is independent and drawn from a normal distribution. Several other characterizations of Wt have been
made for modeling Brownian motion. Processes with drift are also used to model fluctuations with a stochastic differen-
tial equation of the form dXðtÞ 5 aðt; XÞdt 1 dWðtÞ.
For particle transport, this is a natural way of looking at events in history consisting of random events. In the sto-
chastic sense, a history does not repeat itself and a random walk has a probability of occurrence. In this description of
particle transport, nature is understood to behave in a stochastic rather than in a deterministic manner.

4.8.3 Population and sample

In a stochastic formulation, estimates are obtained from quantities from a finite sample of histories as it is not possible
to consider the whole population which can be of the order of Avogadro’s number of particles. When simulating parti-
cles in nuclear systems, it is typical to consider a sample of B106 for which reasonably acceptable results are obtaine-
able. After reviewing some very powerful theorems of probability and statistics, it will be shown that the variance of a
sample mean can be reduced in comparison with the population mean.

FIGURE 4.20 A random fluctuation depicted as a

Wiener process.
 Mathematical foundations Chapter | 4 189

4.8.4 Random variables, PDF, and CDF

In nuclear transport, modeled through stochastic simulations, the outcomes of an experiment are random; if the simula-
tion is repeated the same result may not be obtained. It is thus considered a random event, just as radioactive transmuta-
tion and the compound nucleus formation described in Chapter 2 were based on one of possibly several outcomes.
Random variables are broadly classified as discrete and continuous; in the former, there are a finite number of reali-
zations of a random variable while in the latter the relaizations are given by continuous functions.
For a discrete random variable X taking values xi ; i 5 1; 2; 3; ?; N with probability pi 5 PfX 5 xi g has an expecta-
tion value hX i
X
N
hX i 5 E ð X Þ 5 pi xi 5 μ: (4.21)
i51

Similar expressions can be written for powers of the random variable X as

X
N
hX n i 5 E ð X n Þ 5 pi xni :
i51

Consider for example, the second power, which is

 2   X N
X 5 E X2 5 pi x2i :
i51

If we want a metric for estimating how far each realization of the random variable is from its expectation value hX i,
it is useful to take the second moment
    XN
ðX2μÞ2 5 E ðX2μÞ2 5 pi ðxi 2μÞ2
i51

The right hand side can be further simplied as

X
N  
pi ðxi 2μÞ2 5 X 2 2 hX i2
i51
2
which reads “the expectation value of X minus the mean squared,” a large value of this would mean that the values are
far from the mean and thus vary to a large extent. This is the reason for this quantity to be called the variance, formally
defined as
 
varðX Þ 5 σ2 5 X 2 2 hX i2 (4.22)
Eqs. (4.21) and (4.22) define the mean μ and variance σ2 which are very important charactersitics of distributions of
random variables.
As an example, consider the binomial distribution in which the outcome of an experiment can be zero or one. Thus
in ten experiments, the outcome could vary from 0 to 10. If the probability of getting a 1 is p, then the probability of
getting a 0 must be 1 2 p since there are two possible outcomes and the probabilities therefore must add up to 1. Using
the above analysis, the expectation values for an outcome is
X
N
hX i 5 EðX Þ 5 pi x i 5 p 3 1 1 ð 1 2 pÞ 3 0 5 p
i51

and
    X N
X2 5 E X2 5 pi x2i 5 p 3 12 1 ð1 2 pÞ 3 02 5 p
i51

giving the variance

 
varðX Þ 5 σ2 5 X 2 2 hX i2 5 p2p2 5 pð1 2 pÞ:
190 Nuclear Engineering

From the above, the value

X
N
S5 Xi
i21

which is the sum of outcomes of N experiments, each of which has an outcome Xi is found if we knew how many of
the exeriments had an outcome of 1; say n experiments had an outcome 1 and therefore N 2 n would have outcomes 0.
We can find the probability of having n outcomes as
N n
Pn 5 p ð12pÞN2n
n
If we carried out N 5 100 experiments and p 5 0:2, then
100 100!
Pn 5 0:2n 0:81002n 5 0:2n 0:81002n
n n!ð100 2 nÞ!
Thus P20 5 0:0993, P40 5 2:316 3 1026 ,P100 5 1:2677 3 10270 . The probability Pn versus the number of outcomes n
with score 1, plotted with the Matlab program CH4_BinomialPDF.m listed below, is shown in Fig. 4.21.

% Program name CH4_BinomialPDF.m

p=0.2;i=0;N=100;
for n=0:1:100
i=i+1;
a=factorial(N)/(factorial(N-n)*factorial(n));
P(i)=a*0.2^n*0.8^(100-n);
x(i)=n;
S(i)=(x(i)*P(i))*1;
end
OutcomeOfAllNexpts=sum(S)

We can thus find the total outcome of experiments

X
N
N n
S5 Xi 5 nPn 5 n p ð12pÞN2n 5 Np
i21
n

For N 5 100; p 5 0:2, S 5 20 which is given from the Matlab program as OutcomeOfAllNexpts.

FIGURE 4.21 Probability of outcomes Pn versus the number of

events n with score 1.
 Mathematical foundations Chapter | 4 191

If we have N sets of two discrete random variables X and Y, and a joint distribution function fX;Y ðx; yÞ 5 1=N, that is,
each set is equiprobable, as an example, then the expectation (mean), variances, covariance and correlation can be
obtained as:
X
N
EfX ½X 5 xi f X ðxi Þ
i51

X
N
EfY ½Y 5 yi f Y ðyi Þ
i51

where
X
f X ðxi Þ 5 fX;Y ðx; yÞ
y

The variance, covariance, and correlation are defined as

  2
VarfX ½X  5 EfX X 2 2 EfX
CovfX;Y ½X; Y  5 EfX;Y ½XY  2 EfX EfY
and
CovfX;Y ½X; Y 
CorrfX;Y ½X; Y  5 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
VarfX ½X  3 Var fX ½X 
The correlation coefficient lies between 21 and 1, the two limits at which the variables are uncorrelated and fully
correlated.

Example 4.4: Given six pairs of values for two random variable, calculate their mean values, variances, covariance and
correlation coefficient.

Let ðx; yÞ 5 fð0; 0Þ; ð1; 2Þ; ð 2 1; 2Þ; ð2; 1Þ; ð 2 1; 4Þ; ð2; 3Þg, fX;Y ðx; yÞ 5 1=6 Given
8
> 2
>
> ;x521
>
> 6
>
>
X >
<1
fX ðxi Þ 5 fX;Y ðx; yÞ 5 6 ; x 5 0; 1
>
>
y >
>
>
> 2
>
: 6;x52
>

the expectation values are

X
N
2 1 1 2 3
EfX ½X  5 xi f X ðxi Þ 5 2 1 3 103 113 123 5
i51
6 6 6 6 6

 XN
2 1 1 2 11
Ef X X 2 5 xi 2 f X ðxi Þ 5 1 3 1 0 3 1 1 3 1 4 3 5
i51
6 6 6 6 6

and the variance is

  11 2 3 2 57
Var fX ½X  5 EfX X 2 2 EfX 2 5 5 5 1:5833
6 6 36
8
> 1
>
> ; y 5 0; 1; 3; 4
X < 6
fY ðyi Þ 5 fX;Y ðx; yÞ 5
>
> 2
x >
: ;y52
6
192 Nuclear Engineering

X
N
1 1 2 1 1 12
Ef Y ½ Y  5 yi fY ðyi Þ 5 0 3 113 123 133 143 5 52
i51
6 6 6 6 6 6

 XN
1 1 2 1 1 34
Ef Y Y 2 5 Yi 2 fX ðyi Þ 5 02 3 1 12 3 1 22 3 1 32 3 1 42 3 5
i51
6 6 6 6 6 6

The variance, covariance, and correlation coefficient are given by

  34 2
VarfY ½Y  5 EfY Y 2 2 EfY 5
2 22 5 1:6667;
6
1 1 1 1 1 2
CovfX;Y ½X; Y  5 EfX;Y ½XY  2 EfX ½X EfY ½Y  5 0 3 1 2 3 2 2 3 1 2 3 1 6 3 2 1 5 5
6 6 6 6 6 6
and
CovfX;Y ½X; Y  2
CorrfX;Y ½X; Y  5 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi 5 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
6
5 0:2052:
Var fX ½X  3 Var fX ½X  1:5833 3 1:6667
For a random variable X that can take on a continuous set of realizations, the pdf f ðxÞ is a density function that repre-
sents the probability that X takes on a realization x. Thus, with the requirement that
ðN
f ðxÞdx 5 1;
2N

as for discrete random variables, the probability P that X has value that lies between a and b is expressed in terms of an
integral
ðb
Pða # X # bÞ 5 f ðxÞdx:
a

An associated quantity is the cdf

ðx
F ðxÞ 5 PðX # xÞ 5 f ðtÞdt
2N

Thus, FðxÞ is a montonically increasing function, that is, it continues to increase.  

An elementary discrete pdf is the uniformly distributed integer between 1 and n obtained from: 1 1 nðrndÞ .
Some widely used continuous pdf’s are the uniform, exponential, the binomial, and the Poisson distribution func-
tions. Consider the uniform pdf (Fig. 4.22) f ðxÞ 5 1; xAð0; 1Þ, and the exponential pdf (Fig. 4.23) f ðxÞ 5 2e22x shown in
Figs. 4.17 and 4.18 (Fig. 4.22).
The Statistics Toolbox of Matlab has GUI-based tools disttool for plotting distribution function (pdf’s and cdf’s) and
randtool for randomly sampling from distribution functions. Of the over twenty distribution functions available, the
ones used here is the uniform, exponential and normal distribution functions. Plots can also be obtained from the com-
mands, for example, x 5 0:0.1:3; y 5 exppdf(x,0.5); plot(x, y), mean 5 sum(x.*y)/sum(y)
The exponential pdf is y 5 f ðxj; μÞ 5 μ1 e2x=μ with mean μ Ðb Ðb Ðb P
The computed mean, obtained from the expression x 5 a xyðxÞdx= a yðxÞdx, and I1  a xyðxÞdxD M i50 xi yi ,
Ðb PM
I2  a yðxÞdxD i50 yi is shown in Table 4.3. In the limit b-N, I2 -1. Table 4.2 shows that the computed mean
tends to the exact mean μ 5 1=2 (Table 4.6).
The mean and variances for three continuous PDFs are given in Table 4.7.where
ð
2 x 2
erf ðxÞ 5 pffiffiffi e2t dt:
π 0
The PDF and CDF of a normally distributed random variable, with 5 5; σ2 5 2, are shown in Fig. 4.24 and 4.25
respectively.
The Matlab program for Figs. 4.24 and 4.25 is listed below.
 Mathematical foundations Chapter | 4 193

FIGURE 4.22 PDF and CDF of a uniformly distributed ran-

dom variable.

TABLE 4.6 Estimated mean from an exponential distribution.

Range and intervals Computed mean

0:0.10:3 0.4454
0:0.01:3 0.4877
0:0.001:3 0.4921
0:0.0001:3 0.4925
0:0.0001:3 0.4999

TABLE 4.7 PDF and CDF of the uniform, exponential and normal distribution functions.

Distribution Uniform Exponential Normal

f ðxÞ 0; x , a; x .b b2
1
a;a#x #b 0; x , 0λe 2λx
;x $0 ðx2μÞ2
p1ffiffiffiffi e 2 2σ2
σ 2π
μ a1b
2
1
λ
μ
σ 2 ðb2aÞ2 1
λ2
σ2
12
x h  i
FðxÞ 0; x , 0 1 2 e 2λx ; x $ 0 1 1 erf x2
pffiffiμ
0; x , a b 2 a ; a # x # b 1; x . b 1
2 σ 2
194 Nuclear Engineering

FIGURE 4.23 PDF and CDF of an exponentially distributed

random variable.

FIGURE 4.24 PDF of a normally distributed random variable.

FIGURE 4.25 PDF of a normally distributed random variable.

 Mathematical foundations Chapter | 4 195

% Program name CH4_NormalPDF.m

gid = fopen('out.txt','w');
mu=5;
sigma=2;
N=1000;
del=10/(N-1);
i=0; sum=0;sum2=0;
for x=0:del:10
i=i+1;

PDF(i)=(1/(sigma*sqrt(2*pi)))*exp(-(x-mu)^2/(2*sigma^2));
CDF(i)= (1/2)*(1+erf((x-mu)/(sigma*sqrt(2))));
xx(i)=x;
sum=sum+x*PDF(i);
sum2=sum2+PDF(i);
end
mean=sum/sum2

The mean value of the random variable is computed in the program to verify the validity of sample size N 5 1000
which was set arbitrarily.

4.8.5 Random numbers

Numbers drawn randomly from some underlying population distribution function, using an algorithm, are called
pseudo-random numbers. Since the number of random numbers drawn from a population can be a finite, though large,
sample, they need to satisfy statistical tests to ensure they are truly representative of the population from which they are
drawn. For this, we will generate numbers and compute their means and variances. The correlation coefficient lies
between 21 and 1, the two limits at which the variables are uncorrelated and fully correlated.
The Matlab functions generating uniform, exponential and normal random numbers are rand, exprnd, and randn.
One of the oldest and most reliable methods used for generating random numbers is the linear congruential random
number generator (RNG)
Sk11 5 ðSk g 1 cÞmodp
where p 5 2m , and the random number ξ k 5 Sk =p. Here, S0 is the seed, g and c are multipliers, and modðpÞ implies the remain-
der after division by p; thus 4 mod 2 5 0. In Matlab the command is mod(4,2). The integers determine the period of the num-
bers and depend on the computer hardware; a large NB36 ensures a large period. The Los Alamos Monte Carlo code uses
m 5 48 giving a period of 246 B7:03 3 1013 , g 5 519 , c 5 0. These limits are dependent on the word length of a computer.
4.8.5.1 Matlab program

s(1)=1;c=47;p=100;N=100;Nbins=5;
for i=1:Nbins count(i)=0; end
for i=1:N
t=c*s(i)+1;
s(i+1)=mod(t,p);
end
trand=s/p;
for i=1:N
if (trand(i)<0.2) count(1)=count(1)+1; end
if ((trand(i)>=0.2) && (trand(i)<0.4))
count(2)=count(2)+1; end
if ((trand(i)>=0.4) && (trand(i) <0.6))
count(3)=count(3)+1; end
if ((trand(i)>=0.6) && (trand(i) <0.8))
count(4)=count(4)+1; end
if ((trand(i)>=0.8) && (trand(i) <=1))
count(5)=count(5)+1; end
end
196 Nuclear Engineering

TABLE 4.8 Random number generation.

Seed g c p Period
1 47 1 100 20
1 2 0 9 6
3 2 0 9 2

The output is: count 5 20 20 20 20 20 so that 20 random numbers fall into each bin.
When c 5 0, the RNG is called a multiplicative congruential generator or the Lehmer RNG; when c ¼ 6 0, the generator
is called a mixed congruential generator. The period depends on the combination of values as shown below (Table 4.8).
The “holes” left by pseudo-random numbers in sampling space uniformly, is filled by quasi-random numbers which
are used for numerical integration for sampling space more efficiently.

Exercise 4.18: Use the Matlab random number generator rand and generate 100 numbers and check their distribution.
Using the tic and toc functions, get an idea of the time taken by your computer to generate 1,000,000 uniform random
numbers.

Exercise 4.19: Generate a set of 1,000,000 uniform random numbers ðξ1 ; ξ2 Þ, test how many of these lie in the first
quadrant of a unit circle and estimate π from the answer π 5 4  ϖ, where ϖ is the fraction of points falling inside the
first quadrant. Now use the value from π 5 4  tan21 1 to estimate the accuracy of a random number estimate of π.

4.8.6 Sampling from PDFs

4.8.6.1 Sampling from analytic PDFs
Sampling can be carried out from a discrete or continuous PDF by the direct method if the CDF can be analytically
obtained, and by the rejection method otherwise.
For a discrete probability, the domain of interest is divided into a finite number of intervals, or bins, each bin having
a probability fi . As an example, consider the case of three bins. Since the sum of all probabilities is 1,
X
3
fi 5 1:
i51

If the bin probabilities are f1 5 0:2; f2 5 0:7; f3 5 0:1; as shown in Fig. 4.26, the CDF is obtained by adding the bin
probabilities. A uniform number ξ in the interval (0,1) is generated; as an example, if ξ 5 0:7 as shown in Fig. 4.26, the
dotted lines connect down into the second group.
The selection for a group is done by comparing ξ with the sums
Xm21 Xm
i51 i
f ,ξ# f ; m 5 1; 2; 3
i51 i

It is seen that ξ 5 0:7 falls in the second interval as the sums are 0.2, 0.9, 1.0. Thus the second interval is selected.
In transport simulation, discrete sampling is carried out to decide, for example, which event to select from a number of
possible events, and to select a nuclide for a collision from a compound.
For an example of direct sampling, consider the uniform pdf for which the cdf is

x; 0 , x , 1
F ðxÞ 5
1; x $ 1
can be used to represent the uniform pdf. A simple proof is the “conservation of probability” as we move from one vari-
able to another. Consider two random variables X and Y with pdf’s f ðxÞ and gðyÞ. Since probability must be conserved
regardless of the variable used, f ðxÞdx 5 gðyÞdy, and if X  f ; Y  F, then since dF dx 5 f , the distribution function
gðF Þ 5 1, that is, the distribution of the cdf is uniform. It is for this reason that F can be used as a pseudo-random
number.
 Mathematical foundations Chapter | 4 197

FIGURE 4.26 Sampling from a discrete distribution.

FIGURE 4.27 Sampling from a uniform distribution.

Fig. 4.27 shows the PDF and CDF for f ðxÞ 5 1=2; 2 # x # 4. The PDF is rectangular while the CDF is a monotoni-
cally increasing function. The dotted lines show that sampling for a random number ξ generated from a uniform distri-
bution. If, for example, ξ 5 0:7, then the intersection with the CDF is projected down to the x 2 axis to get a sample
x 5 3:40. Ðx
For an exponential PDF, shown in Fig. 4.28, f ðxÞ 5 2e22x , F ðxÞ 5 0 f ðxÞdx 5 1 2 e22x the random variable X can be
sampled using x 5 F 21 ðξÞ, so that x 5 ð2 1=2Þlnð1 2 ξÞ (Table 4.9).
From Fig. 4.27, for ξ 5 0:7, the random variable sampled is x 5 0:6020: This procedure is done on a computer where
thousands or possibly millions of random numbers may be required depending on the problem.
198 Nuclear Engineering

FIGURE 4.28 Sampling from an exponential distribution function.

TABLE 4.9 Sampling from uniform, exponential and normal PDFs.

Distribution Uniform Exponential Normal

2λx
f ðxÞ 0; x , a; x .b b2
1
a;a#x
#b 0; x , 0λe ;x $0 ðx2μÞ2
p1ffiffiffiffi e 2 2σ2
x σ 2π
h  i
FðxÞ 0; x ,a b2a ; a # x # b 1; x .b 0; x , 0 1 2 e 2λx ; x $ 0 1
2 1 1 erf xσ2 pffiffiμ
2
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
x x 5 a 1 ðb 2 aÞξ 2 λ1 log ð1 2 ξÞ p 5 σ 2 2logξ1 θ 5 2πξ2 μ 1 psinθ

pffiffiffi
For the case of a linear PDF, ðxÞ 5 2x; 0 # x # 1, the CDF is F ðxÞ 5 x2 ; for which x is sampled as x 5 ξ :
Another useful case is the PDF f ðr Þ 5 re2r =2 ; p can be integrated easily to get the CDF F ðr Þ 5 1 2 e2r =2 from
2 2
which
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
which the random variable can be sampled as r 5 2 2log ð1 2 ξÞ:
An approximate method to sample from a Gaussian is to form the sum
Z 5 ξ 1 1 ξ2 1 ?ξk ; largek

from uniform random numbers ξi with mean and variance E ξi 5 1=2; σ2 5 1=12. The mean and variance of Z are thus
E½Z  5 k=2; σ2 5 k=12:
With the variable for a standardized normal (mean 0 variance 1), x and Z are related; here μ and σ are the mean and
standard deviation of the Gaussian from which sampling is done. Then
x 2 μ Z 2 k=2
5 pffiffiffiffiffiffiffiffiffiffi
σ k=12
from which the sampled variable is
!
Z 2 k=2
xDσ pffiffiffiffiffiffiffiffiffiffi 1 μ;
k=12
This method works well for large ð*12Þ.
 Mathematical foundations Chapter | 4 199

4.8.6.2 Sampling from nonanalytic PDFs

Such a direct inversion is not possible for nonintegrable pdf’s such as a rejection scheme in which a simple bounding
function gðxÞ is chosen ðf ðxÞ # cgðxÞÞ; the variable x is sampled from G and the value x^ is selected if ξ2 cgðxÞ
^ # f ðxÞ,
^ oth-
erwise it is rejected.

Exercise 4.20:
 
1. Use both direct and rejection schemes for sampling from the PDF: f ðxÞ 5 2x, xE 0; 12 .
2. Use the rejection scheme for sampling from the normal PDF:

1
f ðxÞ 5 pffiffiffiffiffiffi e22ð σ Þ
1 x2μ 2

2πσ

4.8.7 KullbackLeibler divergence for uniform random numbers

An important metric for estimating convergence in a sampling process is the KullbackLeibler divergence,
X pðxÞ X X
DðpjðqÞ 5 pðxÞlog 5 pðxÞlogpðxÞ 2 pðxÞlogqðxÞ  2 H ðX Þ 1 Hc ðXÞ
x
qðxÞ x x

between a true pdf represented by p and an estimate q, where Hc ðXÞ is the cross-entropy which is a measure of the inef-
ficiency. Ideally, the divergence ðpjðqÞ, also called the relative entropy, should be zero so that no information is lost.
Let us consider this metric for deciding how many uniform random numbers are enough for convergence. From
Fig. 2.9, it is clear that less than N 5 100 numbers is inadequate and NB500 is a “good” sampling size.
The KL distance is seen in Fig. 4.26 to decrease as the sample size is increased. The convergence is also dependent
on the number of bins where a larger sample size is required for 20 bins compared with that for 10 bins (Fig. 4.29).

4.8.8 The law of large numbers

For random numbers generated as independent and identically distributed, that is, each with the same mean μ then as
their number increases the sample mean μs will tend to the theoretical mean μ.
Chebyshev’s Inequality tells us that the deviation from the mean of a random variable more than k standard devia-
tions is less than or equal to 1/k2. Thus the probability of being out of two standard deviations is 1/4, and
  V ðxÞ 1
P X 2 μ . kσ # 2 5 2
ε k

FIGURE 4.29 (A and B) KL distance for 10 and 20 bins.

200 Nuclear Engineering

According to the law of large numbers if X1, X2, . . ., Xn are independent trials, with an expectation μ 5 EðXj Þ and
finite variance σ2 5 VðXj Þ, then for an average S n  Snn 5 X1 1 X2n1 ?Xn , for any ε . 0
Sn
lim P j 2 μj $ ε -0
n-N n
Sn
lim P 2 μ , ε -1
n-N n

Example 4.5: Consider n Bernoulli trials processes and take a probability of success
 p 5 0:3. If an outcome
  is a success, let
the random variable be 1 otherwise it is zero. The expectation value E Xj 5 0:3 and variance V Xj 5 pq 5 0:21. Let
S n  Snn 5 X1 1 X2n1 ?Xn , then the expectation and variance of the average can be found as , S n $ 0:3, and VS n 5 VnS2n 5 0:21
n ,
 
and according to Chebyshev’s Inequality, P jS n 2 0:3j $ 0:1 # 21=nð0:1Þ 5 n . If we have n 5 1000 trials, then we can
2 21

thus show that the probability that this estimate lies between 0.2 and 0.4 is greater than or equal to 0.979; or

Pð0:2 , S 1000 , 0:4Þ $ 0:979

Thus, for a very large number of trials the probability that the number will lie in the range 0.2 to 0.4 will be one. We
know that the number should be 0.3 for a single trial; this states that it will tend to 0.3 for a large number of trials as well.

4.8.8.1 Application of the law of large numbers

For uniform random numbers between 0 and 1, the mean value and variance are 1/2 and 1/12 respectively. Thus if n
numbers are drawn from this distribution, the EðSn =nÞ 5 1=2 and V(Sn/n) 5 1/12n so that
1 1
P jSn 2 j $ ε #
2 12nε2
For ε 5 0.1, P # 100=12n, so that a sample size n 5 100, would give P # 8%, for n 5 1000, P , 5 0.8%, so if want an
error of less than 1%, we must have a sample size of nB1000 (Grinstead & Snell, 1997). Thus Chebyshev’s Inequality
tells us the sample size required for the desired accuracy.

4.8.9 The central limit theorem

Chebyshev’s Inequality gives an idea of the sample size while the stronger Central Limit Theorem tells us the effect of
increasing sample size on the range of estimates.
Let Sn 5 X1 1 X2 1 X3 ?Xn be the sum of n discrete independent random variables with common distribution having
expected value μ and variance σ2 . Then for a , b,
ð
Sn 2 nμ 1 b
lim P a , pffiffiffiffiffiffiffiffi , b 5 pffiffiffiffiffiffi e2x =2 dx
2

n-N nσ2 2π a
! ð
Sn =n 2 μ 1 b
lim P a , pffiffiffiffiffiffiffiffiffiffi , b 5 pffiffiffiffiffiffi e2x =2 dxNð0; 1Þ
2

n-N σ =n
2 2π a
In words, the above says that whatever the population distribution of the random variable X, with mean and vari-
ance, the sampled mean X will be normal as the sample size becomes very large. The sampled mean will have a mean
equal to the population mean; however its variance will have reduced to σ2 =n. Thus, as the sample size becomes larger,
the range becomes narrower.

Example 4.6: Population distribution of a random variable X is normal μp ; σp ;sample size N, sample of the mean has
pffiffiffiffi
μ s ; σs ; sample mean X is normally distributed with μ s 5 μp and σs 5 σp = N
 Mathematical foundations Chapter | 4 201

4.8.9.1 Matlab program: mean and variance

mu=27;std=12;
fid=fopen('OutSample.txt','w');
N=36;
fprintf(fid,'\n N=%8.0f',N)
for i=1:100
ct=0; % count people less than 30 years old
for j=1:N
age(j)= mu+std*randn;
if (age(j)<30)
ct=ct+1;
end
end
count(i)=ct;
SampleAvg(i)=mean(age);
SampleStd(i)=sqrt(var(age));
fprintf(fid,'\n %6.0f %6.0f %8.4f
%8.4f',i,count(i),SampleAvg(i),SampleStd(i))
end
% now find the standard error of the SampledMean
AvgSampleMean=mean(SampleAvg);
StdSampledMean=sqrt(var(SampleAvg));
fprintf(fid,'\n %8.4f %8.4f',AvgSampleMean,StdSampledMean)

Table 4.10 shows results for N batches with sample sizes 36, 3600, and 360,000, that is, an increase of 100 in each
computation. The decrease in the standard deviation can be seen to be by a factor of 10 while the sample mean tends to
the population mean (μp 5 27; σp 5 12).

4.9 Evaluation of integrals

There are several methods for evaluating integrals; the most basic of ideas is to break the region into a large number of
rectangles and sum the areas of the rectangles. This would give an error at the curves of the function under which the
area is being evaluated. Such methods (Abramowitz & Stegun, 1964) include the trapezoidal rule, Simpson’s rule,
EulerMaclaurin formulas, NewtonCotes formulas, and Gauss’ formula (based on Legendre polynomials) for 1-D
integrals. As the dimensionality increases, these methods become inefficient and random sampling begins to look attrac-
tive. In this section, the basics of the sampling methods are described.
A definite integral for a function f(x) is thus expressed as a finite sum
ð1 Xn
I5 f ðxÞdx  ci f ðxi Þ;
21 i51

The reactangles are shown in Fig. 4.30.

TABLE 4.10 Sample mean μ s and standard deviation σs .

N (batches) 36 3600 360,000

100 26.8202(1.8477) 26.9864(0.2020) 27.0013(0.0189)
1000 27.1001(1.9830) 26.9950(0.2005) 27.0008(0.0192)
202 Nuclear Engineering

FIGURE 4.30 Numerical integration.

TABLE 4.11 Abscissa and weights of the Gaussian quadrature.

n xi ci
2 0.5773502691896257 1.0000000000000000
2 0.5773502691896257 1.0000000000000000
3 0.7745966692414834 0.5555555555555556
0 0.8888888888888888
2 0.7745966692414834 -0.555555555555556
4 0.8611363115940525 0.3478548451374544
0.3399810435848563 0.6521451548625460
2 0.3399810435848563 0.6521451548625460
2 0.8611363115940525 0.3478548451374544

In the Gaussian quadrature, the abscissa and weights arise out of requiring exact results for ðn 1 1Þth order polyno-
mials. For the case n 5 2, an integral I is evaluated at to points; thus
ðb X2
I 5 f ðxÞdx 5 wi f ðxi Þ 5 w1 f ðx1 Þ 1 w2 f ðx2 Þ:
a i51

For a third-order polynomial

f ðxÞ 5 a0 1 a1 x 1 a2 x2 1 a3 x3
the integration is carried out resulting in four simultaneous linear equations for the four unknowns w1 ; w2 ; x1 ; x2 ; the
solution gives
   
b2a b2a 1 b1a b2a b2a 1 b1a
I5 f 2 pffiffiffi 1 1 f pffiffiffi 1
2 2 3 2 2 2 3 2
Thus, if a 5 2 1; b 5 1
   
1 1
I5f 2 pffiffiffi 1f pffiffiffi
3 3
which requires a change of limits on the integral
ð1
I5 f ðxÞdx 5 w1 f ðx1 Þ 1 w2 f ðx2 Þ
21

yielding the weights and abscissae of Table 4.11

1 1
w1 5 1; w2 5 1; x1 5 2 pffiffiffi ; x2 5 pffiffiffi :
3 3
Similarly, higher order polynomials are used to determine quadrature parameters for n 5 2, 3 and 4 .
 Mathematical foundations Chapter | 4 203

Ð5 Ð 2π Ð 10
Example 4.7: Sketch the Integrals I 1 5 0 x2 e2x dx, I2 5 0 e2x sin4xdx, and I3 5 0 e2x dx, and evaluate them using
2

Simpson’s Rule, and Gaussian Quadrature.

4.9.1 The Monte Carlo method for numerical integration

The computational effort for quadrature methods discussed above increases with the dimensionality of the integral thus
a 1000 point evaluation in each dimension requires one billion evaluation points. For such large problems, Monte Carlo
methods are more efficient. An integral evaluation of
ðb
I5 gðxÞf ðxÞdx
a

with the Monte Carlo Method (Dunn & Shultis, 2012; Kalos & Whitlock, 2008; Reiher, 1966) is carried out by sam-
pling from the PDF f ðxÞ and estimating gðxÞ as

  1X N
ID gðxÞ 5 gðxi Þ
N i51

With the Central Limit Theorem, there is justification to anticipate that as the sample size N-N, that is, a large num-
ber, then the domain of the result will become smaller and smaller with the sample mean tending to the population
mean and the variance reducing inversely with the sample size N.
Ð2 Ð2 Ð2
1. Sketch theiIntegrals, I1 5 0 dx e2x , I2 5 0 xe2x dx and I3 5 0 e2x pdxffiffi
2

Ð2  
2. Obtain an exact solution for each of the above (I3 5 0 exp 2 x2 dx 5 2π erf ð2Þ 5 0:8821)
3. With a calculator, taking N 5 10 points, estimate the integrals and the standard deviations
4. Write a program which can be used to estimate the integrals using the MC method.
The functions are plotted in Fig. 4.31.

FIGURE 4.31 Functions for MC integration.

204 Nuclear Engineering

4.9.1.1 Matlab program

fid=fopen('outMain.txt','w');
x=0:0.001:2;
y1=exp(-x);
y2=x.*exp(-x);
y3=exp(-x.*x);
plot(x,y1,'-k','LineWidth',1.5)
hold on
plot(x,y2,'-.k','LineWidth',1.5)
hold on
plot(x,y3,':k','LineWidth',1.5)
h = legend('y=exp(-x)','y=x*exp(-x)','y=exp(-x^2)',1);
grid on
xlabel('\bf x','FontSize',12)
ylabel('\bf y','FontSize',12)
N=100;
% integral 1
F=@(x)exp(-x);
Q=quad(F,0,2);
sum=0.0; sum2=0.0;
for i=1:N
t = rand;
r = exp(-2.0*t);
sum = sum + r ;
sum2=sum2 + r^2;
end
Int1 = 2.0*sum/N;
var1 = sum2/N - (sum/N)^2;
std1 = sqrt(var1);
PercErr = 100.0*abs((Int1 - Q)/Q);
fprintf(fid,'\n Integral 1');
fprintf(fid,'\n N= %6.0f',N);
fprintf(fid,'\n quad Int1 var1');
fprintf(fid,'\n %8.4f %8.4f %12.4e ',Q,Int1,std1);
fprintf(fid,'\n Percentage Error = %8.4f',PercErr);
% integral 2

F=@(x)x.*exp(-x);
Q2=quad(F,0,2);
sum=0.0; sum2=0.0;
 Mathematical foundations Chapter | 4 205

for i=1:N
t = rand;
r = t*exp(-2.0*t);
sum = sum + r ;
sum2=sum2 + r^2;
end
Int2 = 4.0*sum/N;
var2 = sum2/N - (sum/N)^2;
std2 = sqrt(var2);
PercErr = 100.0*abs((Int2 - Q2)/Q2);
fprintf(fid,'\n Integral 2');
fprintf(fid,'\n N= %6.0f',N);
fprintf(fid,'\n quad Int1 std2' );
fprintf(fid,'\n %8.4f %8.4f %12.4e ' ,Q2,Int2,std2);
fprintf(fid,'\n Percentage Error = %8.4f' ,PercErr);

% integral 3
F=@(x)exp(-x.*x);
Q3=quad(F,0,2);
sum=0.0; sum2=0.0;
for i=1:N
t = rand;
r = exp(-4.0*t*t);
sum = sum + r ;
sum2=sum2 + r^2;
end
Int3 = 2.0*sum/N;
var3 = sum2/N - (sum/N)^2;
std3 = sqrt(var3);
PercErr = 100.0*abs((Int3 - Q3)/Q3);
fprintf(fid,'\n Integral 3');
fprintf(fid,'\n N= %6.0f',N);
fprintf(fid,'\n quad Int3 std3' );
fprintf(fid,'\n %8.4f %8.4f %12.4e ' ,Q3,Int3,std3);
fprintf(fid,'\n Percentage Error = %8.4f' ,PercErr);
fclose(fid);

Results from MC evaluation of integrals are given in Table 4.12, giving the estimate and relative error with refer-
ence to the exact.
In Monte Carlo simulation, we say an estimate is accurate if its mean is close to the mean of the population, which
is generally unknown. Precision is a measure of the spread or variance of the estimates from which the mean is com-
puted. An estimate is not “good” if it is inaccurate but precise, or accurate and not precise; it is desirable to have an
answer that is both accurate and precise. The “best” MC estimate would have the “correct” mean and “zero-variance.”

TABLE 4.12 Values of integrals from Gaussian Quadrature and random sampling.

Integral Matlab (quad) Exact N 5 100 N 5 100,000

Ð2 21
I1 5 0 dxe ,2x
0.8647 0.8647 0.9176 2.4359 3 10 6.13%
b
0.8649 2.4176 3 1021 0.027%
Ð2
I 2 5 0 xe 2x dx 0.5940 0.5940 0.5924 4.4340 3 1022 0.2656% 0.5944 4.1886 3 1022 0.0656%
Ð 2 2x 2
I3 5 0 e dx 0.8821 0.8821a 0.8827 3.2023 3 1021 0.0651% 0.8792 3.4455 3 1021 0.322%
Ð2   pffiffi
a
I3 5 0 exp 2 x 2 dx 5 2π erfð2Þ 5 0:8821.
b
Mean, standard deviation and percentage error (with quad taken as ref).
206 Nuclear Engineering

There are several methods to obtain better and faster results such as importance sampling and other variance reduc-
tion techniques which will be discussed in the sequel in the context of simulations for nuclear systems.

Problems
1. Show that for a system in equilibrium with nðtÞ 5 n0 , subject to a step input ρ0 at t 5 0, the response nðtÞ is given by
no jρ   ρo  ω 2 t k
nðtÞ 5 o
2 ω2 eω 1 t 1 ω1 2 e
ω1 2 ω 2 l l
For ρ0 . 0 plot the response and comment on the effect of the roots on the response. Use the data for thermal
fission U235.
β
λ 5 0:0767 s21 ; β 5 0:00650; Λ5 5 0:0847s
λ
2. Solve the steady-state heat conduction equation to plot the temperature in a fuel rod using data of Chapter 3. The
volumetric heat generation is
πz
qw 5 qmax
w cos B :
H
From Chapter 3, use data for fuel rod diameter, clad thickness, fuel rod length, centerline fuel temperature.
3. Solve the steady-state heat conduction equation
d2 T 1 dT qw
1 1 50
dr 2 r dr kf
with boundary conditions (1) T is nonsingular within the rod, (2) T(0) 5 Tm. Show that the temperature within the
rod is
0 00
q r2
T 5 Tm 2
4kf
Now solve for the temperature in the clad. Show that the heat flow is
2πHkc ðTs 2 Tc Þ
q5
ln ð1 1 b=aÞ
where T(a) 5 Ts , T(a 1 b) 5 Tc
4. Given the advection-diffusion equation
dT d2 T
u 2 k 2 5 Qð xÞ
dx dx
in the domain a # x # b, with the boundary conditions T ðaÞ 5 Ta and T ðbÞ 5 Tb , write a finite difference equation
and describe the computational strategy to solve for the temperature TðxÞ.
In the above equation, write the finite difference equation for the first two nodes for a given heat flux boundary
condition at the boundary x 5 a. Ðπ
5. Solve the Fredholm integral equation ðxÞ 5 1 1 λ 0 sinðx 1 tÞyðtÞdt
Ð1   λ

6. Show that the resolvent kernel for yðxÞ 5 1 1 λ 0 ð1 2 3xtÞyðtÞdt I Rðx; t;λÞ 5 4 24 λ2 1 1 λ 2 3λx
2 2 3t x 1 2 2 λx
and the solution is yðxÞ 5 4 14λð2 2 3xÞ
2 λ2
; forjλj , 2
Ð 5 2 2x
7. Estimate the integral I 5 0 x e dx with x2 as the estimator and e2x as the pdf.
8. Use the linear congruential RNG Sk11 5 Sk g 1 cmodp with g 5 20, c 5 1, So 5 1, p 5 100 to generate random
numbers.
9. In the previous question, increase the value of g to g 5 40 and comment on the period of the random numbers.
What general conclusion can be drawn?
10. Sample random numbers from a linear pdf f ðxÞ 5 27 ð 2 1 3xÞ in the range ð0; 1Þ.
 Mathematical foundations Chapter | 4 207

Nomenclature
English lower case
d extrapolation distance
e internal energy
f probability distribution function
ff g the force function
g acceleration due to gravity
fxy joint distribution function
h heat transfer coefficient
k thermal conductivity
k effective multiplication
kc thermal conductivity of cladding
keff effective multiplication
kf thermal conductivity of fuel
n number density
p probability
p pressure
q heat flow (W)
000
q volumetric heat generation (W/m3)
u velocity vector

English upper case

Ai area
B buckling
Ci concentration of precursor
D diffusion coefficient
D^ derivative operator
D KullbackLeibler divergence
E energy
F force
F cumulative distribution function
HðμÞ Chandrasekhar’s H function
J variational functional
K stiffness matrix
K kernel
L diffusion length
1
L^ adjoint operator
Ni atomic density of the ith nuclide
Ni shape function in the ith element
Pml associated Legendre functions
Rc cladding resistance
Rf fuel resistance
S source
Sk kth random number
T absolute temperature
Tm moderator temperature
TC clad temperature

Greek lower case

α thermal diffusivity
αF temperature coefficient of reactivity of fuel
αM temperature coefficient of reactivity of moderator
β separability constant
208 Nuclear Engineering

β delayed neutron fraction

B
β delayed neutron fraction for a mixture
γ separability constant
γji production fraction of the ith isotope due to fission of isotope j
ε emissivity
η separability constant
λ decay constant
μ cosine of angle of scattering
μ expectation, mean
ξ random number
ρ reactivity
ρ density
σ StefanBoltzmann constant
σ standard deviation
τ stress
φ flux
φC complementary solution
φP particular solution
χ fission spectrum

Greek upper case

Λ neutron generation time (s)
Ω solid angle
P
a macroscopic absorption cross section

Abbreviations
CDF cumulative distribution function
ODE ordinary differential equation
PDE partial differential equation
PDF probability distribution function

References
Abramowitz, M., & Stegun, I. A. (1964). Handbook of mathematical functions with formulas, graphs and mathematical tables. Washington, DC:
National Bureau of Standards, U.S. Department of Commerce.
Asadzadeh, M., & Kadem, A. (2006). Chebyshev spectral-SN method for the neutron transport equation. Computers and Mathematics with
Applications.
Baker, G. (2016). Systems of differential equations. In Differential equations as models in science and engineering.
Bell, G., & Glasstone, S. (1952). Nuclear reactor theory. van Nostrand Reinhold Company.
Bell, G. I., & Glasstone, S. (1979). Nuclear reactor theory. New York: Robert E. Krieger Publishing Co., Inc..
Bernal, Á., Miró, R., Ginestar, D., & Verdú, G. (2014). Resolution of the generalized eigenvalue problem in the neutron diffusion equation discretized
by the finite volume method. Abstract and Applied Analysis.
Brandao, M. A., Dominguez, D. S., & Iglesias, S. M. (2011). Linear finite element method for one-dimensional diffusion problems. In International
Conference on Mathematics and Computational Methods Applied to Nuclear Science and Engineering (M&C 2011).
Case, K. M., & Zweifel, P. F. (1967). Linear transport theory. Reading, MA: Addison-Wesley Pub. Co..
Chandrasekhar, S. (1960). Radiative transfer. New York: Dover Publications.
Dunn, W. L., & Shultis, J. K. (2012). Exploring Monte Carlo Methods.
Filipov, S., & Faragó, I. (2018). Implicit euler time discretization and fdm with Newton method in nonlinear heat transfer modeling. arXiv.
González-Pintor, S., Ginestar, D., & Verdú, G. (2008). Continuous pseudo spectral method for the neutron diffusion equation on a triangular mesh. In
International Conference on the Physics of Reactors 2008, PHYSOR 08.
Grinstead, C. M., & Snell, J. L. (1997). Introduction to probability: Second revised edition. American Mathematical Society.
Hetrik, D. (1993). Dynamics of nuclear reactors. La Grange Park, IL: American Nuclear Society.
Hutton, D. V. (2004). Fundamentals of finite element analysis (ISBN: 0072395362, ISSN: 1098-6596). Textbook (Important).
IAEA. (2008). Thermophysical properties of materials for nuclear engineering: A tutorial and collection of data. Vienna, Austria: IAEA.
Johnson, M., Lucas, S., & Tsvetkov, P. (2010). Modeling of reactor kinetics and dynamics.
Kalos, M. H., & Whitlock, P. A. (2008). Monte Carlo methods. Wiley-VCH Verlag GmbH & Co. KGaA.
 Mathematical foundations Chapter | 4 209

Kanwal, R. P. (2013). Linear integral equations: Theory and technique.

Lamarsh, J. R., & Baratta, A. J. (2001). Introduction to nuclear engineering (3rd ed., p. 783)Prentice Hall.
Lewis, E. E., & Miller, W. F. Jr (1984). Computational methods of neutron transport. New York: John Wiley and Sons.
Lewis, F. L., Vrabie, D. L., & Syrmos, V. L. (2012). Optimal control (3rd ed.).
Lewis, R. L., Nithiarasu, P., & Seetharamu, K. N. (2004). Fundamentals of the finite element method for heat and fluid flow. John Wiley & Sons, Inc.
Ozisik, M. N. (1993). Heat conduction (2nd edn.). John Wiley & sons, Inc.
Polyanin, A. D., & Manzhirov, A. V. (2006). Handbook of mathematics for engineers and scientists.
Polyanin, A. D., & Manzhirov, A. V. (2008). Handbook of integral equations (2nd edn.).
Polyanin, A. D., & Zaitsev, V. F. (2004). Handbook of nonlinear partial differential equations.
Rao, S. S. (2017). The finite element method in engineering.
Reiher, W. (1966). Hammersley, J. M., D. C. Handscomb: Monte Carlo Methods. Methuen & Co., London, and John Wiley & Sons, New York, 1964.
VII 1 178 S. Preis: 25 s, Biom. Z.
Yeih, W. (2020). Solving the nonlinear heat equilibrium problems using the local multiquadric radial basis function collocation method. Mathematics.
Zwillinger, D. (1993). Handbook of differential equations. Mathematics of Computation.
Chapter 5

The neutron diffusion equation

A simple yet representative model of neutron transport is the neutron diffusion equation (NDE) for which analytical
solutions can be readily found for nuclear systems modeled as regular geometries (rectangular, spherical, and cylindri-
cal). For irregular geometries, numerical methods such as the Finite Difference Method (FDM), the Finite Element
Method (FEM) and Finite Volume Method (FVM) are used. In reactor calculations, the Nodal Expansion Method
(NEM) is used with the NDE to calculate the flux and power distribution over a full core. In such calculations, the
NDE is easier to use with less computational effort than the neutron transport equation or the Monte Carlo (MC)
method.
In this chapter, neutron diffusion theory is introduced for nuclear systems with nuclear fuel materials such as U235
and Pu239 as well as nonfissile materials such as aluminum, beryllium, boron, carbon (graphite), and iron used for struc-
tural applications.
The NDE is developed for one-dimensional regular geometries, and extended to three-dimensional geometries, for
obtaining the neutron flux and associated reaction rates (absorption, scattering, etc.) in systems.

5.1 The conservation equation

Deterministic models that describe neutron behavior in a system are based on the diffusion equation (Jevremovic, 2009;
Lamarsh, 2005), the integro-differential Boltzmann equation, and the integral equation. An elementary form of the dif-
fusion equation can be derived from first principles by writing a conservation equation in a volume element assuming
that neutrons do not lose energy during interactions with matter. Consider the volume element ΔV 5 ΔxΔyΔz with net
inward and outward currents in the x direction Jin 5 Jx and Jout 5 Jx1Δx shown in Fig. 5.1.
The change in neutron population nðx; tÞ, for this one-group model, can be written from the conservation equation
expressing the net boundary
crossings and losses and gains in the volume element. With the net boundary crossings
taken only in the x-direction J in 2 J out A\ n/s, the absorptions Σa φðx; tÞΔV n/s and the source production Sðx; tÞΔV,
the rate of change of neutron population is:
@nðx; tÞ

ΔV 5 J in 2 J out A\ 2 Σa φðx; tÞΔV 1 Sðx; tÞΔV: (5.1)
@t
Using a Taylor series first-order expansion for the current, we can write
J out A\  J in A\ 1 r  JΔV
so that the diffusion equation, with φ 5 nv, becomes
1 @φðx; tÞ
5 2 r  J 2 Σa φðx; tÞ 1 Sðx; tÞ:
v @t
Using Fick’s law J 5 2 Drφ, the diffusion equation reads
1 @φðx; tÞ
5 Dr2 φ 2 Σa φðx; tÞ 1 Sðx; tÞ: (5.2)
v @t
The diffusion coefficient D is given by
1 1
D 5 λtr 
3 3Σtr

Nuclear Engineering. DOI: https://doi.org/10.1016/B978-0-323-90618-0.00005-3

© 2022 Elsevier Inc. All rights reserved. 211
212 Nuclear Engineering

FIGURE 5.1 A volume element.

in terms of the macroscopic transport cross-section

Σtr 5 Σt 2 μ0 Σs :
0
where the average scattering angle Ω^ Ω^ for isotropic scattering in the center of mass system is μ 5 2=ð3AÞ.
0
The loss terms are the leakage r  J and the removal Σr φ while the gain terms are the direct source S and the fission
production νΣf φ; thus in general the steady-state diffusion equation can be expressed as
r  J 1 Σr φ 5 S 1 νΣf φ:
In a two-group model, the neutrons produced by fission (per cm3/s)
ν 1 Σf ;1 φ1 1 ν 2 Σf ;2 φ2
would contribute to both groups in proportion to the fission spectrum χi ; i 5 1; 2:
For steady-state, the one-group balance equation reduces to
Dr2 φ 2 Σa φ 1 S 5 0 (5.3)
with absorption being the only removal process. If a two-group model is considered, then the processes have to be mod-
eled in the groups separately. Thus the net leakage rate of the group-1 neutrons would be balanced by the removal rate
due to capture in group-1 and to out-scattering to group-2 ΣR;1 φ1 5ΣC;1 φ1 1ΣS;1-2 φ1 as well as to the source produc-
tion in group 1. The two-group equations are
D1 r2 φ1 2 Σr;1 φ1 1 S1 5 0 (5.4)
D2 r φ2 2 Σr;2 φ2 1 S2 5 0
2
(5.5)
where S1 could be a “direct” source and a fission source χ1 ðν 1 Σf ;1 φ1 1 ν 2 Σf ;2 φ2 Þ and S2 5 pΣS;1-2 φ1
1 χ2 ðν 1 Σf ;1 φ1 1 ν 2 Σf ;2 φ2 Þ, a “down-scattered” contribution plus a fission contribution. The eigenvalue form of the dif-
fusion equation, in a 1-D slab of thickness L, as briefly discussed in Section 4.5, is written as
   
d d 1
2D 1 Σr φðxÞ 5 νΣf φðxÞ 1 SðxÞ; 0 , x , L
dx dx k
where k is the eigenvalue that balances the loss and gain terms.
The multigroup form of the diffusion equation for energy group i is
X ðsÞ X
2r  Di ðr Þrφi ðr Þ 1 Σi ðr Þφi ðr Þ 5 Σi0 -i φi0 ðr Þ 1 χi νΣf :i0 φi0 ðr Þ 1 Si ðrÞ (5.6)
i0 6¼i i0
where Si ðrÞ is the independent source.
The forms taken by the r  Di ðr Þr were discussed in Chapter 4;for Cartesian
 
@ @ @ @ @ @
D 1 D 1 D 2 Σr φðx; y; zÞ 1 νΣf φðx; y; zÞ 1 S 5 0;
@x @x @y @y @z @z
cylindrical
 
1@ @ 1 @ @ @ @
rD 1 2 D 1 D 2 Σr φðr; ϕ; zÞ 1 νΣf φðr; ϕ; zÞ 1 S 5 0;
r @r @r r @ϕ @ϕ @z @z
 The neutron diffusion equation Chapter | 5 213

and spherical coordinates

 
1 @ 2 @ 1 @ @ 1 @ @
r D 1 sin θ D 1 D 2 Σ r φðr; θ; ϕÞ 1 νΣf φðr; θ; ϕÞ 1 S 5 0:
r 2 @r @r r 2 sinθ @θ @θ r 2 sinθ @ϕ @ϕ

5.2 The one-group diffusion equation

The time-dependent diffusion Eq. (5.2), for a one-group energy model, can be solved analytically for simple problems
in slab, cylindrical and spherical geometry. In this section, some elementary steady-state solutions are obtained to illus-
trate the flux distributions.

5.2.1 Nonmultiplying systems

Although the one-group model is too simple for practical applications, it gives considerable insight into important neu-
tron characteristics of nuclear systems. For systems which are predominantly “one-group,” that is, when the neutron
spectrum is predominantly “fast” or “thermal,” it is used with the appropriate “averaged” nuclear data to obtain the neu-
tron flux and associated quantities such as criticality parameters and thermal power.
The NDE is a second-order ordinary differential equation (ODE) which can be solved given two boundary condi-
tions: Neumann, Dirichlet or mixed, depending on the physical conditions. A bare assembly, for example, will have the
condition of vanishing flux at the extrapolated boundary (Dirichlet boundary condition) while a reflecting surface will
have a zero flux-gradient, or zero current, (Neumann boundary condition) condition. An interface will have both flux
and current continuity conditions.
The NDE, based on Poisson’s equation, is mathematically much simpler than the integro-differential transport equa-
tion and hence is used for simple problems. Even then, numerical solutions are required for large problems such as
“whole-core” reactor design. In the multigroup form, the diffusion equation is used to obtain “first estimates” and can
serve as a useful step for providing guesses to full transport calculations. A limitation of diffusion equation is near
sources and boundaries where the angular flux requires detailed consideration as provided, for example, in the discrete
ordinates and spherical harmonics methods.

5.2.1.1 Finite slab

In a graphite slab of thickness 2a in the x-direction and infinite in the y and z directions, with uniformly distributed
sources emitting S neutrons/cm3/s the neutron flux is given by
!
S cosh Lx
φðxÞ 5 12
 : (5.7)
Σa cosh a 1L d

Another method for obtaining the flux for a distributed source is by expanding in eigenfunctions and using the
orthogonality property to determine the coefficients. This formulations results in a Green’s function which is useful for
more general source distributions. From Eq. (5.7), the net current in the slab, from Fick’s Law, is
sinhx=L
J ðxÞ 5 2 DrφðxÞ 5 SL
:
cosh a 1L d
From J ðxÞ, the escape probability Pes ðaÞ, defined as the fraction of neutrons leaking out of the system across the sur-
face x 5 a, is
Jðx 5 aÞ L sinha=L
Pes ðaÞ 5 5
:
aS a cosh a 1L d
The total escape probability from the system is Pes 5Pes ð 2 aÞ 1 Pes ðaÞ; due to the symmetry of the system
2Jðx 5 aÞ L sinha=L
Pes 5 5

2aS a cosh a 1L d
From L’Hopital’s rule it is readily seen that in the limit a-0, Pes -1 (for d 5 0) that is, all the neutrons escape
from an intesimally thin slab.
214 Nuclear Engineering

The diffusion length defined as

D
L2 5
Σa
was discussed in Chapter 2 in the context of the slowing down of neutrons.

Example 5.1: .

Calculate the escape probability in a graphite slab of thickness 1 mean free path (mfp).
The first step is to find the atomic density N of graphite:

ρNav 1:60 X 0:6023 1024

N5 5 5 0:0802 3 1024 atoms cm23
A 12:01115
The next step is to determine the macroscopic cross-sections Σa ; Σs ; Σtr :
Σa 5 Nσa 5 0:0802 X 0:0034 5 2:7268 1024 cm021
; Σs 1
5 Nσs 5 0:0802 3 4:75 5

 2
0:3809 cm21 ; Σtr 5 Nσtr 5 Nσs 1 2 μo 5 Nσs @1 2 A 5 0:3599 cm21
3A

These are used to find the diffusion coefficient D and the diffusion length L:
rffiffiffiffiffiffi rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 1 D 0:9262
D5 5 5 0:9262cm; L 5 5 5 58:2804cm
3Σtr 3ð0:3599Þ Σa 2:72681024
and extrapolation distance d 5 0:71λtr 5 0:71
0:3599 5 1:9728cm. The mean free path λ, is found as
1 1 1
λ5 5 5 5 2:6233 cm:
Σt Σs 1 Σa 0:3812
For a graphite slab of thickness 1 mfp, the escape probability is

L sinha=L 58:2804 sinh1:3116=58:2804

Pes 5
a 1 d  5 1 1:9728
5 0:9985:
a cosh L 2:6233=2 cosh 1:3116
58:2804

Thus, in this case there is a 99.85% probability that a neutron will escape from the surfaces.
Fig. 5.2 shows the escape probability in the above graphite slab as a function of its thickness. The probability is
“high” since graphite is a weak absorber and is thus used for moderating neutrons in a reactor.

FIGURE 5.2 Escape probability in a graphite slab.

 The neutron diffusion equation Chapter | 5 215

5.2.1.2 Finite cylinder

The governing equation
 
1@ @ @ @
rD 1 D 2 Σr φðr; zÞ 1 S 5 0
r @r @r @z @z
can be solved to obtain exact solutions. For an infinitely long cylinder with a uniformly distributed source as
for the finite slab case, the solution is similar to Eq. (5.7) with the modified Bessel function of the first kind I 0
replacing the cosh x function. From the method of eigenfunctions, the flux in a finite cylinder can be found in
the form
X xn r mπz
φðr; zÞ 5 Amn J0 cos ; n; m 5 1:23; . . . (5.8)
m;n
R H

in terms of Bessel functions, eigenvalues and eigenfunctions as described in Chapter 4. The coefficients are then deter-
mined from the diffusion equation using the orthogonality condition.

5.2.1.3 Point source in an infinite medium

The flux in a finite sphere with a uniformly distributed source is of the same form as shown for the slab case in
Eq. (5.7) with the sinh function replacing the cosh function as shown in Table 5.1.
For a point source in an infinite medium, the (thermal) neutron flux is
Se2r=L
ϕðr Þ 5 (5.9)
4πDr
where L  LT , the thermal diffusion length.
Table 5.1 lists some solutions with source conditions, boundary conditions and one-group fluxes for nonmultiplying
systems. Source conditions for infinite plane, point-like and infinite linear sources can also be derived from Gauss’
divergence theorem (Carrillo, 2001) using Fick’s law.

5.2.2 Multiplying systems

When the source is given by Sðr Þ 5 kN Σa ϕðrÞ, with the infinite system multiplication factor kN 5 εpηfT , the diffusion
equation can be written as
r2 φ 1 B2m φ 5 0 (5.10)
where the material buckling Bm is
kN 2 1
B2m 5 :
L2
Some elementary solutions are given here for slab, cylindrical, and spherical critical reactors.

5.2.2.1 Slab reactor

The 1-D slab reactor equation
d2 φ
1 B2m φð xÞ 5 0 (5.11)
dx2

for 2 a=2 # x # a=2 with the boundary conditions

   
a~ a~ dφ
φ 2 5φ 5 0; j0 5 0; (5.12)
2 2 dx

is critical for the flux

φðxÞ 5 AcosBx (5.13)
216 Nuclear Engineering

TABLE 5.1 Exact solutions: 1-group nonmultiplying media.

Medium Source Boundary conditions Flux

Slab
Infinite Planar source Finite flux φðx Þ φðx Þ 5 SL 2x=L
2D e I
at x 5 0 Source condition
S n/cm2/s limx-0 J ðx Þ 5 S=2
Infinite slab (in y and z) of Planar source at x 5 0 φða 1 d Þ 5 φð 2 a 2 d Þ 5 0 φðx Þ 5 SL sinh½ða 1 d 2 jx jÞ=L
2D cosh½a 1L d 
thickness 2a S n/cm2/s Source condition
limx-0 J ðx Þ 5 S=2
 
Infinite slab (in y and z) of Uniformly distributed φða 1 d Þ 5 φð 2 a 2 d Þ 5 0 coshx=L
dφ φðx Þ 5 S
12 coshða 1L d Þ
thickness 2a sources S n/cm3/s dx 5 0 at x 5 0 Σa

Cylinder
P
x r 
Infinite cylinder (in z axis) as given φðR Þ 5 0 m An J0 n
R ; n 5 1:23; . . .
dφ
finite radius R dr 5 0
at r 5 0
P
x r 
Finite cylinder radius R, φðR Þ 5 φðR Þ 5 0 m;n Amn J0 R
n
H ; n; m 5 1:23; . . .
cos mπz
0 0
height H φr50 5 φz5H 5 0
Sphere
Infinite Point isotropic source at Finite flux φðr Þ φðr Þ 5 Se2r=L
4πDr
r 5 0 S n/s Source condition
limr-0 4πr 2 J ðr Þ 5 S

 flux φðr Þ and

Finite sphere of radius R Point isotropic source at Finite sinhκðR 1 d 2 rÞ
φðr Þ 5 S
4πDsinhðR 1L d Þ r
r 5 0 S n/s φ R~ 5 0
Finite sphere of radius R Uniformly distributed
 flux φðr Þ and
Finite φðr Þ 5 S
12 R1d sinhκr
φ R~ 5 0
Σa r sinhκðR 1 d Þ
sources S n/cm3/s

where the material and geometrical buckling are

kN 2 1 π
Bm 5 ; Bg 5 :
L2 a1d

5.2.2.2 Cylindrical reactor

For an infinite cylindrical reactor (i.e., infinite in height) which is critical and has radius R, the flux is space-dependent.
The reactor equation
1 d dφ
r 1 B2 φ 5 0 (5.14)
r dr dr
has two independent solutions J0 ðBrÞ and Y0 ðBrÞ which are ordinary Bessel functions of the first- and second-kind. The
general solution is
φ 5 AJ0 ðBrÞ 1 CY0 ðBrÞ (5.15)
where A and C are constants. Since Y0 ðxÞ is infinite at x 5 0, while J0 ð0Þ 5 1, for φ to remain finite, C must be taken to
zero and the flux is
φðr Þ 5 AJ0 ðBrÞ: (5.16)
The boundary condition therefore takes the form


φ R~ 5 AJ0 ðBrÞ 5 0: (5.17)
The function J0 ðxÞ is zero at a number of values of x, so that φ is satisfied if B has eigenvalues
xn
Bn 5 : (5.18)
R~
 The neutron diffusion equation Chapter | 5 217

In a critical system, the lowest eigenvalue has physical significance; thus

 2
x1
2
B1 5
2
5 2:405R~ :
R~
The one-group flux then becomes


φ 5 AJ0 2:405r R~ : (5.19)
The constant A is determined after using from the reactor power given by
ð
P 5 ER Σf φðr ÞdV:

For a cylinder, dV 5 2πrdr and

ðR ðR  
2:405r
P 5 2πER Σf φðr Þrdr 5 2πER Σf A J0 rdr
0 0 R

0 0 0
Ð
which is evaluated using J0 x x dx 5 xJ1 ðxÞ to get
2πER Σf R2 AJ1 ð2:405Þ
P5 5 1:35ER Σf R2 A:
2:405
After rearranging the equation for A, the flux is
 
0:738P 2:405r
φðr Þ 5 J 0 : (5.20)
ER Σf R2 R
From Section 3.1, the flux in a critical finite cylindrical reactor of radius R and height H is

 πz
φðr; zÞ 5 AJ0 2:405r R~ cos (5.21)
BH
where normalization gives
3:63P
A5 :
ER Σ f V
For a reactor such as the pressurized water reactor (PWR) AP1000 described in Chapter 3, the average value of the
fission cross-section in the core of a reactor can be obtained as
Σ f jr nr Vr
Σ f jc 5
Vc
where Σ f jc is the average value of thermal fission cross-section in the core, Σ f jr is the average value of thermal fission
cross-section in a fuel rod, nr is the number of fuel rods, Vr is the volume of a fuel rod, and Vc is the core volume
The flux, plotted using the Matlab program listed below, is shown in Fig. 5.3 for a cylindrical reactor; the maximum
power is at the origin of the reactor decreasing axially for a given radius. Physically, the flux signifies the power distri-
bution in the reactor core. Thus, fuel rods at the center of a core are much “hotter” than those at the periphery.

FIGURE 5.3 Flux in a critical finite cylindrical reactor of radius R and height.
218 Nuclear Engineering

Matlab program for flux in a cylindrical reactor

gid = fopen('out.txt','w');
C:\Users\Administrator\Desktop\Elsevier2020\PROGRAMS\Ch5_FluxPlotCyl.m
R=1.52;H=4.267/2;% half height
[X,Y] = meshgrid(0:0.1:R,0:0.1:H);
Z = besselj(0,2.405*X/R).*cos(pi*Y/H);
figure(1)
set(gca,'FontSize',12)
surf(X,Y,Z)
hold on
xlabel('\bf Radius (m)','fontsize',14)
ylabel('\bf Height (m)','fontsize',14)
ylim([0,H])
%legend('\bf n(E), n_0 kT','\bf n(v), n_0 {(m/(2kT))}^{1/2}')
grid on
fclose(gid)

Exercise 5.1: .

Plot the neutron flux in an assembly with H 5 2R 5 5ft: and S 5 107 neutrons/s, kN 5 0:980; p 5 0.83, L2T 5 2:8 cm2,
and Σ a 5 0:16/cm.

5.2.2.3 Spherical reactor

In a critical spherical reactor, the flux can be obtained by solving
 
1 @ 2 @
r 1 B φðr Þ 5 0
2
(5.22)
r 2 @r @r


and applying the boundary condition φ R~ 5 0 and the requirement of finite flux inside the reactor. The critical reactor
flux is
sinBr
φðr Þ 5 A (5.23)
r


~ since φ R~ 5 0. The constant A, as in the cylindrical case, is determined
where the geometric buckling B5Bg 5 π=R,
from the power of the reactor P:
P
A5 :
4ER Σf R2

Exercise 5.2: .

Using Eq. (5.23), obtain the (1) average flux in the reactor, and (2) the neutron current at the surface of the reactor.
Calculate and plot the quantity 4πr 2 ϕðrÞ for r=RAð0; 1Þ and describe what it physically represents.

Example 5.2: .

Consider a spherical reactor of radius R 5 8.7407 cm operating at a power of 100 W. Find the constant A in the
expression for neutron flux.
Use the data: Σf 5 0:0672cm21 ; R~ 5 R 1 d; d 5 1=3λtr 5 1:0212cm. The recoverable energy from fission ER is
assumed to be 200 MeV. Then
100 1 n
A5 5 1:52171011
20010 1:610 0:0672 4ð8:7407Þ2
6 19 cm2  s
 The neutron diffusion equation Chapter | 5 219

and since
π
Bg 5
R1d
For Bm 5 Bg 5 B, the flux is
sinBr n
φðr Þ 5 1:52171011 :
r cm2  s
Fig. 5.4 shows the flux for two power levels 100 and 200 W; the magnitude is determined by the operating power
while the shape is determined by the “buckling,” or the curvature.
Table 5.2 lists some interesting source conditions, boundary conditions and one-group fluxes for multiplying systems.

5.2.3 One-group criticality

From an elementary consideration that all neutrons have the same energy and are therefore classified as neutrons of one
group, it is possible to estimate the criticality of slab, cylindrical and spherical reactors. These are inaccurate due to the
assumptions made but are nonetheless useful to generate an idea of the size of critical systems.
For criticality, equating the material and geometrical bucklings Bm 5 Bg 5 B, gives
π 2 kN 2 1
B2 5 5
R L2
in a spherical system from which
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
L2
Rc 5 π
kN 2 1
gives an estimate for the critical radius Rc . Rearranging the above into a form used for the leakage probability, the
effective multiplication for a critical system keff 5 1 is
kN
keff 5 5 1: (5.24)
1 1 B2 L2
Eq. (5.24) can be used, in an elementary analysis, to calculate the critical composition of a slab, cylinder or sphere.

Example 5.3: .

Given a U235-Na fast spherical reactor with 1% weight fraction U235 and kN 5 1:49, determine the critical radius
and the amount of U235 required.

FIGURE 5.4 Neutron flux in a bare spherical reactor.

220 Nuclear Engineering

TABLE 5.2 Exact solutions: 1-group diffusion multiplying media.

Medium Source Boundary conditions Flux

Slab
Infinite critical slab reactor of thickness a; Bare critical φða 1 d Þ 5 φð 2 a 2 d Þ 5 0 φðx Þ 5 πP
2aER Σf cos πx
a
dφ
and d {a reactor dx 5 0 at x 5 0

 πy
Bare cubical reactor of sides a Bare critical φða 1 d Þ 5 φð 2 a 2 d Þ 5 0 φT x; y; z 5 Acos πx πz
a~ cos a~ cos a~
reactor a~  a 1 d
dφ
dx 5 0 at x 5 y 5 z 5 0

Cylinder


Finite cylinder Bare critical φ
R~ ; z  5 0 φðr; z Þ 5 AJ0 2:405r
cos πz
~ R~ ~
reactor φ r; H 5 0 H



 flux φðr Þ and φðr Þ 5 0:738P 2:405r
Infinite cylinder Bare critical Finite ER Σf R 2 J0 R
reactor φ R~ 5 0
Sphere
sinπr=R~

 flux φðr Þ and
Bare spherical reactor of radius R Bare critical Finite φðr Þ 5 P
4ER Σf R 2

reactor ϕ R~ 5 0 r

P is the reactor power, ER is the energy recoverable from fission (B200 MeV).

The microscopic transport and absorption cross-sections are σtr;F 5 6:8b; σtr;M 5 3:3b; σa;F 5 1:65b; σa;M 5 0:0008b;
~ the critical radius Rc 5 R~ 2 d is
With B 5 π=R,
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
π2 L2 π 1
Rc 5 2 d 5 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 :
kN 2 1 3Σtr Σa ðkN 2 1Þ 3Σtr
Step 1: Calculate the atomic densities of fuel NF and moderator NM
From the mass ratio it is possible to find the atomic ratio:
NM ρ AF
5 M 5 1011:7
NF ρ f AM
To get another equation for NF and NM calculate the atomic weight of the “mixture” A from the atomic fractions
αU 5 1=1012:7 and αNa 5 1011:7=1012:7
1 1011:7
A 5 α F AF 1 α M AM 5 235:04 1 23 5 23:2094:
1012:7 1012:7
Calculate the mixture density ρmix from the given weight fractions
1 wF wM 0:01 0:99
5 1 5 1 5 1:0212;
ρmix ρF ρM 18:7 0:97
ρmix 5 0:9792g  cm23 , and calculate the atomic density of the mixture
ρmix Av 0:9792 0:6023 1024
Nmix 5 5 5 0:0254 1024 atomsUcm23 :
A 23:2094
From this the individual atomic densities can be determined using the atomic fractions
1
NF 5 Nmix 5 2:50811019 atoms  cm23
1012:7
1011:7
NM 5 Nmix 5 2:54001022 atoms  cm23
1012:7
 The neutron diffusion equation Chapter | 5 221

Step 2: Calculate the macroscopic cross-sections Σtr ; Σa



Σtr 5 NF σtr;F 1 NM σtr;M 5 2:50811025 6:8 1 2:541022 3:3 5 0:0840cm21
and the extrapolation distance is d 5 0:71λtr 5 0:71=ð0:0840Þ 5 8:3333cm ; the fuel and moderator macroscopic
absorption cross-sections are:


Σa;F 5 NF σa;F 5 2:50811025 1:65 5 4:13841025 cm21


Σa;M 5 NM σa;M 5 2:54001022 0:0008 5 2:0321025 cm21
from which the mixture absorption cross-section is
Σa 5 NF σa;F 1 NM σa;M 5 6:17041025 cm21 :
Thus
1 1
L2 5 5 5 0:6431 3 105 cm2
3Σtr Σa 3ð0:0840Þð6:17041025 Þ
and
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
π2 L2 253:5942π
Rc 5 2 d 5 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 8:3333 5 1130 cm:
kN 2 1 ð1:49 2 1Þ
The volume of this spherical reactor is 6044 m3. With the mixture density calculated above, this has a weight of
about 6000 tons, of which about 60 tons is U235 and the rest, B5940 tons, is Na.

5.3 The two-group diffusion equation

5.3.1 Nonmultiplying systems
The two-group NDE for the energy-averaged “fast” and “thermal” flux φ1 ðrÞ and φ2 ðrÞ are

D1 ðrÞr2 φ1 ðrÞ 2 Σ1 ðrÞφ1 ðrÞ 5 0 (5.25)

D2 ðrÞr2 φ2 ðrÞ 2 Σ2 ðrÞφ2 ðrÞ 1 Σ1 ðrÞφ1 ðrÞ 5 0: (5.26)

For uniform material distribution, that is, constant macroscopic cross-sections, exact solutions can be obtained, from
the above, in both finite and infinite media.
These are illustrated by writing a conservation equation and then going through a detailed derivation of the solution
using complimentary and particular solutions for the above coupled second-order system.
To obtain the two-group fluxes in a nonmultiplying system, we begin with the elementary conservation of neutrons
depicted in Fig. 5.5. In this simple illustration, there is a source of group-1 neutrons at the center of the sphere giving
rise to a group-1 neutron flux, as well providing the slowing-down “source” of neutrons into group-2 in the form of

FIGURE 5.5 Conservation of neutrons in a two-group model.

222 Nuclear Engineering

slowed-down neutrons Σ1 Φ1 V and neutrons crossing into the volume element J2 x A1 . The “loss” terms of group-2 neu-
trons are the boundary crossings J2 x1Δx A1 and the absorptions Σ2 Φ2 V.
For group-1 neutrons,
ðD1 r2 Σ1 ÞΦ1 ðxÞ 5 0 (5.27)
with the boundary conditions
 
dΦ1
S5 4πr 2 ; Φ1 ðR 1 d 1 Þ 5 0
dr r-0
The balance equations for group-2 neutrons is
J2 x A1 1 Σ1 Φ1 V 5 Σ2 Φ2 V 1 J2 x1Δx A1
which, expanded into a first-order Taylor series
d x
J2 x A1 1 Σ1 Φ1 V 5 Σ2 Φ2 V 1 ½J2 x 1 J2 ΔxA1
dx
simplifies to
d x
J2 V 5 Σ1 Φ1 V 2 Σ2 Φ2 V:
dx
Using Fick’s law, this becomes
 
d d
2D2 Φ2 V 5 Σ1 Φ1 V 2 Σ2 Φ2 V
dx dx
Expressed, for constant diffusion coefficient, in the form
d2
D2 Φ2 2 Σ2 Φ2 1 Σ1 Φ1 5 0 (5.28)
dx2
or in operator form
D2 r2 Φ2 2 Σ2 Φ2 1 S2 5 0:
The boundary conditions are
0
2D2 4πr 2 Φ2 j0 5 0; Φ2 ðṜ2 Þ 5 0
In a spherical coordinate system, with angular symmetry
1 @ 2 @ 2 0 00
r2 5 r D 5 Φ 1Φ
r 2 @r @r r
0 0 00
00 0
With the substitution w 5 rΦ, w 5 rΦ 1 Φ; and w 5 rΦ 1 Φ 1 Φ0
 
00 00 2 0
w 5r Φ 1 Φ
r
so that the group-1 equation
1
r2 Φ1 ðr Þ 2 Φ1 ðr Þ 5 0;
τ
where τ 5 D1 =Σ1 ; becomes
00 1
w 2 w50
τ
with the solution
prffi 2prffi
wðr Þ 5 A1 e τ 1 A2 e τ
 The neutron diffusion equation Chapter | 5 223

giving the group-1 flux

A1 prffiτ A2 2prffiτ
Φ1 ðr Þ 5 e 1 e :
r r
With the identities
1 x
coshx 5 ðe 1 e2x Þ; coshx 1 sinhx 5 ex
2
1
sinhx 5 ðex 2 e2x Þ; coshx 2 sinhx 5 2 ex
2
we can write
   
A1 r r A2 r r
Φ1 ðr Þ 5 cosh pffiffiffi 1 sinh pffiffiffi 1 cosh pffiffiffi 2 sinh pffiffiffi ;
r τ τ r τ τ
p rffiffi prffiffi
cosh τ sinh τ
Φ1 ðr Þ 5 ðA1 1 A2 Þ 1 ð A1 2 A2 Þ ;
r r
or
cosh prffiffiτ sinh prffiffiτ
Φ1 ðr Þ 5 B1 1 B2
r r
For the group-2 flux,
D2 ðr Þr2 Φ2 ðr Þ 2 Σ2 Φ2 ðr Þ 5 2 Σ1 Φ1 ðr Þ;
We seek complimentary and particular solutions
Φ2 ðr Þ 5 Φ2 c ðr Þ 1 Φ2 p ðr Þ:
For the complimentary solution,
D2 ðr Þr2 Φ2 ðr Þ 2 Σ2 Φ2 ðr Þ 5 0
gives
cosh Lr sinh Lr
Φ2 e ðr Þ 5 B3 1 B4 :
r r
For the particular solution,
1 2 Σ1 Φ1 ðr Þ
r 2 Φ2 ðr Þ 2 Φ2 ðr Þ 5 ;
L2 D2
and in operator form
1 Σ1
Φ2 p ðr Þ 5 2 Φ1 ðr Þ
^ 2
D 2 L12 D2
!
L 1 1 2Σ1 ðrÞ
5 2 Φ1 ðr Þ
2 D^ 2 L1 D^ 1 L1 D2 ðrÞ

L2 ^ 21 1 ð11LDÞ
^ 21 2Σ1 ðrÞ
52 ð12LDÞ Φ1 ðr Þ
2 D2 ðrÞ
L2 ^ 2 L2 D2 1 . . . 1 1 2 LD 1 L2 D2 1 . . . 2Σ1 ðrÞ
52 1 1 LD Φ1 ðr Þ
2 D2 ðrÞ
h i 2Σ ðrÞ
Φ2 p ðr Þ 5 2 L2 1 1 L2 D^ 1 L4 D^ 1 : . . . :
2 4 1
Φ1 ðr Þ
D2 ðrÞ
224 Nuclear Engineering

For constant cross-sections, the infinite series

"  2 2 #
L2 L L1
Φ2 ðr Þ 5 2 L
p 2
11 1 1 ... Φ1 ðr Þ;
τ τ D2

is written in compact form, for L2 {τ,

 21
L2 L1
Φ2 ðr Þ 5 L
p 2
12 Φ1 ðr Þ:
τ D2
Simplifying further,
Σ1
Σ2
Φ2 p ðr Þ 5 Φ1 ðr Þ:
1 2 Lτ
2

Thus
P
P 1
coshr=L sinhr=L
Φ2 ðr Þ 5 B3 1 B4 1 2
Φ1 ðr Þ:
r r 12 L2
τ

We now apply the boundary conditions to get the four coefficients B1 ; B2 ; B3 ; B4 :

The current boundary condition gives
prffiffi sinh prffiffiτ 2 cosh prffiffiτ prffiffi cosh prffiffiτ 2 sinh prffiffiτ
dΦ1 τ τ
5 B1 1 B 2 ;
dr r2 r2
 
dΦ1
4πr 2 5 ½ðB1 4πÞð 2 1Þ 1 4πB2 ð0Þð 2 D1 Þ;
dr r-0
S 5 2 B1 ð4πÞð 2 D1 Þ;
so that
S
B1 5 :
ð4πÞ ðD1 Þ
Equating the group-1 flux at the extrapolated boundary condition to zero, Φ1 ðR 1 d1 Þ 5 0; gives
R 1 d1 R 1 d1
B1 cosh pffiffiffi 1 B2 sinh pffiffiffi 5 0
τ τ
1ffiffid1
cosh R p τ
B2 5 2 1ffiffid1
B1 :
sinh R p τ

Simplifying,
" #
cosh prffiffiτ 1ffiffid1
cosh R p τ
sinh prffiffiτ
Φ 1 ð r Þ 5 B1 2 1ffiffid1
;
r sinh R p τ
r

gives
" #
B1 cosh pffiffiτ sinh pffiffiτ 2 sinh pffiffiτ cosh pffiffiτ
r R 1 d1 r R 1 d1
Φ1 ðr Þ 5 1ffiffid1
;
r sinh R p τ

which, upon using the identities,

coshðA 6 BÞ 5 cosh A cosh B 6 sinh A sinh B
sinhðA 6 BÞ 5 sinh A cosh B 6 cosh A sinh B
 The neutron diffusion equation Chapter | 5 225

r R 1 d1 R 1 d1 r Ṝ2r
‘cosh pffiffiffi sinh pffiffiffi 2 cosh pffiffiffi sinh pffiffiffi 5 sinh pffiffiffi
τ τ τ τ τ
Similarly, applying the boundary condition to the group-2 flux,

R1 2 r
sinh pffiffiffi
τ
Φ1 ðr Þ 5 α
r
(5.29)
R2 2 r
sinh pffiffiffi
τ
Φ2 ðr Þ 5 β 1 γΦ1 ðr Þ
r

where
8
> S 1
>
> α5 pffiffiffi
>
> 4πD1 sinhṜ1 = τ
>
>
>
< S 1 L2
β5
>
> 4πD2 sinhṜ2 =L τ 2 L2
>
>
>
> Σ1 τ
>
> γ5
: Σ2 τ 2 L2


n a finite medium, the boundary conditions, for the “fast” group: limr-0 4πr 2 J1 ðr Þ 5 S, and φ1 R~ 5 0 yield

 pffiffiffiffiffi
S sinh R~ 2 r = τ T pffiffiffiffiffi
φ 1 ðr Þ 5 = τT : (5.30)
4πD1 r sinhR~

 the “thermal” group, there is no “direct” source so that the boundary conditions: 4πr J2 ðrÞjr50 5 0, and
2

For
~
φ2 R 5 0 yield the solution


S L2 sinh R~ 2 r =L Σ1 τ
φ 2 ðr Þ 5 =L 1 φ ðr Þ: (5.31)
4πD2 r τ 2 L2 sinhR~ Σ2 τ 2 L2 1

For an infinite medium, it can readily be shown that the fluxes are

S exp 2 prffiffiffiffi
τT
φN
1 ðr Þ 5 (5.32)
4πD1 r
and
SL22 pffiffiffiffi
φN
2 ðr Þ 5

2  e2r=L2 2 e2r= τ T : (5.33)
4πrD2 L2 2 τ T

5.3.1.1 Computer programming example

Eqs. (3.26) and (3.27) for a slab of thickness, with an incident fast neutron source S=2 neutrons/cm3/s on the left, are
written as
d2 φ1
D1 2 Σr1 φ1 5 0
dx2
d2 φ2
D2 2 Σr2 φ2 1 Σr1 φ1 5 0
dx2
226 Nuclear Engineering

with boundary conditions

dφ1
D1 j 5 S=2
dx x50
φ1 ðx 5 T Þ 5 φ2 ðx 5 0Þ 5 φ1 ðx 5 T Þ 5 0
and the following data:
D1 5 1:1302 cm Σr1 5 0:0418 cm21 S 5 1 cm-3 s21
D2 5 0:1107 cm Σr2 5 0:0167 cm21
These yield the exact solutions:
pffiffi pffiffi
τ
φ1 ðxÞ 5 A1 ex= 1 A2 e2x= τ

and
φ2 ðxÞ 5 A3 ex=L 1 A4 e2x=L 1 βφ1 ðxÞ
where
D1 2 D2 Σr1 τ
τ5 ;L 5 ;β5
Σr1 Σr2 Σr2 τ 2 L2
and τ 5 27:0080 cm2 ; L2 5 6:6128 cm2 ; β 5 3:3100:
The constants found from the boundary conditions are: A 5 ½ 2 0:0010; 2:2981; 0; 2 7:6034.
Fluxes are evaluated with the Matlab program listed below:

den=1.0;MolWt=18;AvNo=0.6022e24;
No=den*AvNo/MolWt; %number density of water
sigr1=1.2508e-24; sigtr1=8.8158e-24; D1=1/(3*No*sigtr1);
sigr2=0.5004e-24; sigtr2=90.000e-24; D2=1/(3*No*sigtr2);
S=1;
Sigr1=No*sigr1;Sigr2=No*sigr2;tau=D1/Sigr1;Lsq=D2/Sigr2;L2=sqrt(Lsq);
beta=(Sigr1/Sigr2)*(tau/(tau-Lsq))
L=20; % slab thickness (cm)
AA=[ 1 -1 0 0;
exp(L/sqrt(tau)) exp(-L/sqrt(tau)) 0 0;
beta beta 1 1;
beta*exp(L/sqrt(tau)) beta*exp(-L/sqrt(tau)) exp(L/L2) exp(-L/L2)]
B= [-S*sqrt(tau)/(2*D1);0;0;0]
A= inv(AA) * B
flux1=@(x) A(1)*exp(x/sqrt(tau))+A(2)*exp(-x/sqrt(tau));
flux2=@(x) A(3)*exp(x/L2)+A(4)*exp(-x/L2)+
beta*(A(1)*exp(x/sqrt(tau))+A(2)*exp(-x/sqrt(tau)));
x=0:0.1:20;
FF1=feval(flux1,x); % group 1 flux
FF2=feval(flux2,x); % group 2 flux
figure(1)
plot(x,FF1,'k-','LineWidth',2);
hold on
plot(x,FF2,'k--','LineWidth',2)
grid on
xlabel('\bf x (cm)','fontSize',14)
ylabel('\bf Flux (n cm^{-2} s^{-1})','fontSize',14)
legend('\phi_1','\phi_2','Location','best','fontSize',12)

The resulting fluxes φ1;2 ðxÞ are shown in Fig. 5.6. As expected, the fast flux φ1 ðxÞ gradually decreases as neutrons
undergo collisions in the host medium giving rising to the thermal flux φ2 ðxÞ which subsequently rises and decays. Both
 The neutron diffusion equation Chapter | 5 227

FIGURE 5.6 Two-group fluxes in a slab.

fluxes fall to zero at the physical (rather than the extrapolated) boundary as required by the specified boundary
conditions.

Exercise 5.3: .

Estimate the time a person can stand a distance R from a Cf-252 point source of given strength S n/s shielded by
water. The maximum permissible dose is 5 rem/y (B100 mrem/week). Use the infinite medium solutions Eqs. (5.32)
and (5.33) and the flux-to-dose conversion factors (Fig. 2.13); justify any other assumptions you make.

5.3.1.2 Temperature effects on neutron flux

As discussed in Chapter 2, thermal properties of materials, such as diffusion coefficient and diffusion length, are
temperature-dependent and corrections must be made to account for this dependence. For the thermal data given in
Table 5.3, the corrections are made with the expressions
  m

 ρ0 T
D ðρ; T Þ 5 D ρ0 ; T0
ρ T0
to evaluate the properties at any desired temperature.
 2 m11=2
ρ0
Similarly, the diffusion length is evaluated as L2T ðρ; T Þ 5 L2T ρ0 ; T0 ρ
T
T0 with m 5 0:470 for H2 O and
m 5 0:112 for D2 O and zero otherwise.

5.3.2 Multiplying systems

In general, multiplying systems are designed in a way that neutron losses are minimized; this leads to efficient designs.
Thus bare assemblies are surrounded by some scattering material, like water or beryllium, from which some of the leak-
ing neutrons are reflected back into the core.
In such a core-reflector system, the two-group equations for the core are

D1c r2 φ1c 2 Σ1c φ1c 1 ηΣ2c φ2c 5 0 (5.34)

and
D2c r2 φ2c 2 Σ2c φ2c 1 Σ1c φ1c 5 0: (5.35)
228 Nuclear Engineering

TABLE 5.3 Thermal data (at 20 C) for diffusion theory neutron flux: infinite medium.

Moderator Density D Σa LT

g=cm3 cm cm cm
H2 O 1.00 0.16 0.0197 2.85
a 25
D2 O 1.10 0.87 9.3 3 10 97
23
Be 1.85 0.50 1.04 3 10 21
24
Graphite 1.60 0.84 2.4 X 10 59
a
D2O containing 0.25 weight/percent H2O.

In the reflector, the equations are

D1r r2 φ1r 2 Σ1r φ1r 5 0 (5.36)
and
D2r r2 φ2r 2 Σ2r φ2r 1 Σ1r φ1r 5 0 (5.37)
where the subscripts c; r refer to core and reflector, respectively. The boundary conditions are continuity of flux and
current at the interface
φ1c 5 φ1r (5.38)
φ2c 5 φ2r (5.39)
0 0
D1c φ1c 5 D1r φ1r (5.40)
0 0
D2c φ2c 5 D2r φ2r (5.41)
and at the boundaries, for a spherical system,
0 0
2D1c φ1c jr50 5 2 D2c φ2c jr50 5 0 (5.42)
and



φ1c R~ 5 φ2c R~ 5 0: (5.43)
We follow the solution procedure given in Lamarsh (1966), where two coupled second-order ODEs are converted
into one fourth-order ODE.
From Eq. (5.34), φ2c can be obtained in terms of φ1c to get a fourth-order ODE for φ1c

2 

r 1 μ2 r2 2 λ2 φ1c 5 0 (5.44)
with
 ffi
1
 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
μ 5
2
2 τ 1 1 L 1 ðτ 1 1L Þ 1 4ðkN 2 1Þτ 1 L
2 2 2 2 (5.45)
2τ 1 L2
and
 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1

λ2 5 τ 1 1 L 2
1 ðτ 1 1L2 Þ2 1 4ðkN 2 1Þτ 1 L2 : (5.46)
2τ 1 L2
For a uniformly distributed fuel, the coefficients in the above equations are constant and solutions are obtained by
expressing the core equations as two 4th-order ODEs to yield the flux
φ1c 5 AX 1 CY (5.47)
 The neutron diffusion equation Chapter | 5 229

φ2c 5 AS1 X 1 CS2 Y (5.48)

φ1r 5 FZ1 (5.49)
φ2r 5 S3 φ1r 1 GZ2 (5.50)
where the “coupling coefficients” S1 ; S2 ; S3 are
Σ1c 1
S1 5 ; (5.51)
Σ2c 1 1 μ2 L2c
Σ1c 1
S2 5 ; (5.52)
Σ2c 1 2 λ2 L2c
and
D1r 1
S3 5 : (5.53)
Σ2r τ T 2 L2r
The core and reflector functions X; Y for slab, cylinder and spherical geometries are given in Table 5.4 for the core
and in Table 5.5 for the reflector.
After applying the interface conditions, the resulting equations are four linear homogenous equations, for which the
only nontrivial solution, by Cramer’s Rule, will be found when the determinant vanishes, that is,

X Y 2 Z1 0
0 0 0
D1c X D1c Y 2 D1r Z1 0
5 0: (5.54)
S1 X S2 Y 2 S3 Z 1 2 Z2
0 0 0 0
D2c S1 X D2c S2 Y D2r S3 Z1 2 D2r Z2

Eq. (5.54) is called the two-group critical determinant. When the material properties are given, the critical dimension
can thus be determined; conversely, when the dimension is given, the material properties (such as enrichment) can be
determined. Consider a U235-H2O dilute homogenous solution with fuel-water ratio 1:500 and two-group data as given
below (Table 5.6).

TABLE 5.4 Two-group core functions.

Geometry X Y
Infinite slab cosμx coshλx
Infinite cylinder J0 ðμrÞ I0 ðλrÞ
Sphere sinμr sinhλr
r r

TABLE 5.5 Two-group reflector functions.

Geometry Z

Reflector thickness b Reflector thickness infinite


Infinite slab sinhκ 2a 1 b 2 jxj e 2κjxj

Infinite cylinder I0 ðκr ÞK0 ½κðR 1 b Þ 2 I0 ½κðR 1 bÞK0 ðκr K0 ðκrÞ

Sphere sinhκðR 1 b 2 rÞ e2κr

r r
230 Nuclear Engineering

TABLE 5.6 Two-group data for homogeneous spherical reactor.

Data Group 1 Group 2

D (cm) 1.13 0.16
Σc (cm) 0.0419 0.060
224
σaW (10 2
cm )  0.664
224
σa2;5 (10 2
cm )  678

For uniform fissile distribution, that is, constant values for D and Σ, the fluxes obtained in the core and reflector for
the data in are
 
sin0:113r sinh0:651r
φ1c ðr Þ 5 A 2 4:671028 (5.55)
r r
 
sin0:113r 27 sinh0:651r
φ2c ðr Þ 5 0:676A 1 3:6410 (5.56)
r r
and
e20:192r
φ1r 5 39:6A (5.57)
r
 20:192r 
e e20:351r
φ2r 5 120A 2 21:0 (5.58)
r r
The core and reflector fluxes are shown in Fig. 5.7. The group-1 fast flux decreases steadily from the core into the
reflector while the group-2 thermal flux also decreases within the core but has a hump at the core-reflector interface
due to the reflection of slowed-down neutrons from the reflector.
For a uniform fissile distribution, with the flux shown in Fig. 5.8, the critical radius and mass obtained from diffu-
sion theory are 21.9 cm and 1.15 kg U235, respectively.

5.3.3 Two-group criticality

The two-group critical determinant results in a transcendental equation from which the critical radius R is determined.
This is written in the form
α1 ζ Y ζ Z2 1 α2 ζ Y ζ Z1 1 α3 ζ Z1 ζ Z2
ζX 5 (5.59)
α4 ζ Y 1 α5 ζ Z1 1 α6 ζ Z2
where
0  
X 1
ζX  52μ 2 cotμR
X μR
0  
Y 1
ζ Y  5 λ cothλR 2
Y λR
0  
Z 1
ζZ  5 2 κ 1 cothκb
Z κR
and
α1 5 D1c D2r ðS3 2 S1 Þ; α2 5 D1r D2c S2 2 D1c D2r S3 ; α3 5 D1r D2r ðS1 2 S2 Þ
α4 5 D1c D2c ðS2 2 S1 Þ; α5 5ðD1r D2c S1 2 D1c D2r S3 Þ; α6 5 D1c D2r ðS3 2 S2 Þ:
 The neutron diffusion equation Chapter | 5 231

The roots μR 5 2:472; R 5 21:91 cm and μR 5 5:61; R 5 49:72 cm are determined from the Matlab program listed
below and shown in Fig. 5.8. The critical radius is thus established to be 21.91 cm, in agreement with Lamarsh.
The Matlab program uses the function
0  
Z 1
ζZ  5 2 κ 11
Z κR
for an infinite reflector.

% Lamarsh Spherical Reflected Critical Reactor

% LamarshSphereCriticalityBookReactor.m
fid=fopen('Out.txt','w');
Nav=0.6023;denW=1.0;MolWt=18.02;AtDenW=denW*Nav/MolWt;
AtDen235=(1/500)*AtDenW;
fprintf(fid,'\n AtDen235=%12.4e AtDenW=%12.4e',AtDen235,AtDenW)
MuFunc = @(a,b) 0.5* ( -a + sqrt(a^2 + 4*b));
LambdaFunc = @(a,b) 0.5* ( a + sqrt(a^2 + 4*b));

% data
sigaW=0.664; sigaTh5=678; Sig1c=0.0419;Sig1r=0.0419;
Tauc=27; Taur=27;D1c=1.13; D2c=0.16;D1r=1.13; D2r=0.16;
p=1.0;
Sig2c=0.886*(AtDen235*sigaTh5+AtDenW*sigaW);
Sig2r=0.886*AtDenW*sigaW;
LsqC=D2c/Sig2c;eta=2.07;
LsqR=D2r/Sig2r; K1r=1/sqrt(Taur);K2r=1/sqrt(LsqR);
f=AtDen235*sigaTh5/(AtDen235*sigaTh5+AtDenW*sigaW);
kinf=eta*f;
a=LsqC+Tauc;b=(kinf-1)*Tauc*LsqC;
Mu2 =(1.0/(LsqC*Tauc))*MuFunc(a,b); Mu= sqrt(Mu2);
Lambda2=(1.0/(LsqC*Tauc))*LambdaFunc(a,b); Lambda = sqrt(Lambda2);
S(1)=(p*Sig1c/Sig2c)/(1+Mu2*LsqC);S(2)=(p*Sig1c/Sig2c)/(1-Lambda2*LsqC);
S(3)=(D1r/Sig2r)/(Taur-LsqR);
Rhs1=D1c*D2r*(S(3)-S(1));
Rhs2=D1r*D2c*S(2)-D1c*D2r*S(3);
Rhs3=D1r*D2r*(S(1)-S(2));
Rhs4=D1c*D2c*(S(2)-S(1));
Rhs5=D1r*D2c*S(1)-D1c*D2r*S(3);
Rhs6=D1c*D2r*(S(3)-S(2));
fprintf(fid,'\n sigaW= %8.5f b Tauc=%8.4f cm^2 sigaTh5=%8.4f b
Sig1c=%8.4f D1c=%8.4f cm D2c=%8.4f cm ',sigaW,Tauc,sigaTh5,Sig1c,D1c,D2c)
fprintf(fid,'\n Taur=%8.4f cm^2
Sig1r=%8.4f D1r=%8.4f cm D2r=%8.4f cm ',Taur,Sig1r,D1r,D2r)
fprintf(fid,'\n Rhs1=%8.4f Rhs2=%8.4f Rhs3=%8.4f',Rhs1,Rhs2,Rhs3)
fprintf(fid,'\n Rhs4=%8.4f Rhs5=%8.4f Rhs6=%8.4f\n',Rhs4,Rhs5,Rhs6)
fprintf(fid,'\n Sig2c= %8.4f cm^-1 Sig2r=%8.4f cm^-1 LsqC= %8.4f cm^2
LsqR=%8.4f K1r=%8.4f cm^-1 K2r=%8.4f cm^-1
',Sig2c,Sig2r,LsqC,LsqR,K1r,K2r)
fprintf(fid,'\n f= %8.4f kinf= %8.4f Mu= %12.4e Lambda= %12.4e
',f,kinf,Mu,Lambda)
fprintf(fid,'\n p= %8.4f ',p)
fprintf(fid,'\n S(1)= %12.4e S(2)=%12.4e S(3)=%12.4e ',S(1),S(2),S(3))
232 Nuclear Engineering

%b=22.5; % ref thickness

fprintf(fid,'\n Reflector thickness = %8.4f cm',b)
fprintf (fid,'\n\n i MuR -R*LHS -R*RHS
Diff ')
Rmin=0;Rmax=60;Nsteps=1000;del=(Rmax-Rmin)/Nsteps;
R=Rmin;RCrit=1e10;MinDif=1e10;
for i=1:Nsteps
R=R+del;
XprimeOverX= Mu*(cot(Mu*R)-1/(Mu*R));
YprimeOverY= Lambda*(coth(Lambda*R)-1/(Lambda*R));
% for a finite reflector
%Z1primeOverZ1= -K1r*coth(K1r*b);
%Z2primeOverZ2= -K2r*coth(K2r*b);
% for an infinite reflector
Z1primeOverZ1= -K1r*(1+1/(K1r*R));
Z2primeOverZ2= -K2r*(1+1/(K2r*R));
%
LHS=XprimeOverX;
r1=Rhs1*YprimeOverY*Z2primeOverZ2;
r2=Rhs2*YprimeOverY*Z1primeOverZ1;
r3=Rhs3*Z1primeOverZ1*Z2primeOverZ2;
RHSNum=r1 + r2 + r3;
r4=Rhs4*YprimeOverY;
r5=Rhs5*Z1primeOverZ1;
r6=Rhs6*Z2primeOverZ2;
RHSDen=r4+r5 + r6 ;
RHS=RHSNum/RHSDen;

X(i)=Mu*R;
Y(i)=-R*LHS;
Z(i)=-R*RHS;
Difference=Z(i)-Y(i);
Diff(i)=Difference;
if (abs(Difference)<MinDif)
MinDif=abs(Difference);
RCrit=R; %this is the critical radius found
end
%fprintf(fid,'\n %6.0f %8.4f %12.4e %12.4e
%12.4e',i,R,LHS,RHS,Diff(i))
fprintf(fid,'\n %6.0f %8.4f %12.4e %12.4e
%12.4e',i,X(i),Y(i),Z(i),Diff(i))
end
 The neutron diffusion equation Chapter | 5 233

figure(1)
set(gca,'FontSize',14)
plot(X,Y,'k-','LineWidth',2)
hold on
plot(X,Z,'k-','LineWidth',2)
xlabel('\bf \mu R','fontSize',14)
ylabel('\bf ','fontSize',14);
text(4,8,'\bf LHS')
text(2.0,12.5,'\bf RHS')
text(5.0,12.5,'\bf RHS')
XL1=[2.472 2.472];YL1=[-20 4.12];
line(XL1,YL1,'LineWidth',0.5,'Color','k','LineStyle','--')
XL2=[5.61 5.61];YL2=[-20 8.08];
line(XL2,YL2,'LineWidth',0.5,'Color','k','LineStyle','--')
grid off
xlim([0 6.2])
ylim([-20 20])

Partial Output
AtDen235= 6.6848e-05 AtDenW= 3.3424e-02
sigaW= 0.66400 b Tauc= 27.0000 cm^2 sigaTh5=678.0000 b Sig1c= 0.0419
D1c= 1.1300 cm D2c= 0.1600 cm
Taur= 27.0000 cm^2 Sig1r= 0.0419 D1r= 1.1300 cm D2r= 0.1600 cm
Rhs1= 0.4283 Rhs2= -1.5021 Rhs3= 1.0738Rhs4= -1.0738 Rhs5= -0.4283
Rhs6= 1.5021

Sig2c=0.0598 cm^-1 Sig2r=0.0197 cm^-1 LsqC=2.6747 cm^2 LsqR=8.1369

K1r= 0.1925 cm^-1 K2r= 0.3506 cm^-1
f=0.6713 kinf=1.3896 Mu=1.1284e-01 Lambda=6.5088e-01 p= 1.0000
S(1)= 6.7737e-01 S(2)= -5.2617e+00 S(3)= 3.0465e+00

FIGURE 5.7 Two-group fluxes in a spherical reactor.

234 Nuclear Engineering

FIGURE 5.8 Transcendental equation for two-group criticality in a reflected

spherical reactor.

The two-group expression for the system multiplication keff with nonleakage probabilities is an extension of the
one-group expression; here the fast and thermal probabilities are both incorporated, based on Fermi’s slowing-down
equation where the neutron age τ T , with units of cm2 is introduced as the area “traveled” by a neutron as it slows down
from source energy to thermal energy. Thus
kN e2B τ T
2

keff 5 5 1: (5.60)
1 1 B2 L2T
In order to find the critical radius, we can set x  B2 τ T , and solve the transcendental equation
 2
2x L
kN e 5 1 1 T x (5.61)
τT
to find x, and hence B from which the critical radius R is found.
Eq. (5.61), from Eq. (1.27), compares with
keff 5 kN ð12PF Þð12PT Þ
where 12PF and 12PT were defined to be the fast and thermal nonleakage probabilities, from the finite system,
respectively. Thus
1
PF 5 e2B τ T D
2

1 1 B2 τ T
where the approximation holds for large reactors for which B2 τ T , 1, and
1
PT 5
1 1 B2 L2T
An alternate expression for large reactors, called the modified one-group critical equation, is
kN
15 (5.62)
1 1 B2 MT2
where MT2 5 L2T 1 τ T is called the thermal migration area.
We now have three expressions to compute critical compositions from one-group theory Eq. (5.24), two-group the-
ory Eq. (5.60) and a modified one-group theory Eq. (5.62).

5.4 The multigroup diffusion equation

When a two-group formulation is inadequate to represent a system, such as for the case of modeling fast, epithermal
and thermal regions, or a larger number of groups, a multigroup formulation is used.
 The neutron diffusion equation Chapter | 5 235

Neutron energy during transport may vary from a few MeV down to a fraction of an electron-volt (thermal
energy ET BkTB0:025 eV). This indeed is a broad range spanning over six orders of magnitude. It is thus conve-
nient to divide the entire range into a finite number of groups within which neutronic properties are averaged and
the neutron “group” equations are written from which group flux are subsequently obtained. While two-group cal-
culations are extensively used for preliminary computations, elaborate MC simulations can be carried out for as
many as 262 groups. Clearly, the more the number of groups, the better the results as fine details, such as cross-
sections in the resonance region, may be represented. For the critical radius and corresponding mass, a detailed
Monte Carlo simulation, with the Monte Carlo N-Particle (MCNP) code, for the 1:500 U235-H2O homogeneous
solution described in the previous section gives a radius of 22.5 cm and mass of 1.2477 kg U235respectively. Thus,
the two-group results can be seen to be fairly accurate.
As an example, in a 6-group diffusion formulation to compute the neutron flux and critical radius of a bare sphere
of Pu239 the group-constants data are given in Table 5.7 (Lamarsh, 1966).
For realistic neutronic analysis, a large number of groups is generally used for which data cross-section libraries are
“generated” using group flux φi and associated reaction rates , σx φ . (where the inner product ,. . .. implies inte-
gration over the energy group of interest), and setting the group cross-section, for group i, as σðiÞ
x 5 , σ x φ . / φi . The
existing methodology for obtaining multigroup cross-sections is based on reading ENDF pointwise cross-section data,
by processing codes such as NJOY (LANL) and MC2 -3 (ANL) to produce binned cross-sections for use in multigroup
deterministic codes. This is achieved by the group flux-weighting mentioned above and accounting for resonances and
self-shielding.

5.4.1 Numerical solution of the multigroup diffusion equations

In case of nonuniform material loading, the flux can be obtained by a numerical solution of the diffusion equations. A
first step for a numerical procedure is the FDM for which a general iterative form for the multigroup diffusion equa-
tions, for φg ,
X
2r  Dg ðr ÞrφðgiÞ ðr Þ 1 Σr;g ðr ÞφðgiÞ rÞ 2 Σs;g0 -g φðgi0Þ ðrÞ 5 Sði21Þ
g ðrÞ (5.63)
i0 6¼i

where, Dg ðr Þ 5 diffusion coefficient for energy group g at spatial position r, φðiÞ g 5 “scalar” flux for energy group g at
spatial position r at the ith iteration, Σr;g ðr Þ 5 removal cross-section for energy group g at spatial position r at the ith
iteration, Σs;g0 -g 5 ’in-scattering’ cross-section for energy group g from groups at the ith at spatial position r,
Sðgi21Þ ðrÞ 5 source for energy group g at spatial position r at the (i 2 1) iteration.
The source term consists of a fission contribution and the contribution from an independent source S0 ðrÞ
X
Sðgi21Þ ðr Þ 5 χg νΣf ;g0 ðr Þφðgi21Þ ðr Þ 1 S0 ðr Þ: (5.64)
g0

TABLE 5.7 Multigroup cross-sections for Pu239.

g E MeV ν χ σtr (b) σf (b) σr (b) σðg-hÞ

h52 3 4 5 6
1 3.0N 3.48 0.204 4.25 1.90 0.03 0.20 0.27 0.45 0.31 0.04
2 1.43.0 3.09 0.344 4.50 1.95 0.05  0.18 0.50 0.35 0.05
3 0.91.4 2.99 0.168 4.80 1.83 0.07   0.45 0.30 0.06
4 0.40.9 2.93 0.180 5.70 1.70 0.11    0.29 0.05
5 0.10.4 2.88 0.090 8.40 1.67 0.17     0.05
6 0.00.1 2.86 0.014 12.0 2.05 0.50     

Source: From Lamarsh J. R., Introduction to Nuclear Reactor Theory. Addison-Wesley Publishing Company, 1966, p. 368; ANL-5800.
236 Nuclear Engineering

In this iterative procedure, the first iteration gives φð0Þ

g from the source distribution S0 ðrÞ, that is, by solving the equation
X
2r  Dg ðr Þrφð0Þ ð0 Þ
g ðr Þ 1 Σr;g ðr Þφg ðrÞ 2 Σs;g0 -g φðg00 Þ ðr Þ 5 S0 ðr Þ: (5.65)
i0 6¼i

With this initial distribution the iterations for i 5 1; 2; 3; ?; N are carried out from the fission source
X X
2r  Dg ðr ÞrφðgiÞ ðr Þ 1 Σr;g ðr ÞφðgiÞ ðr Þ 2 Σs;g0 -g φðgi0Þ ðrÞ 5 χg νΣf ;g0 ðr Þφðgi21
0
Þ
ðr Þ
i0 6¼i g0

Thus for i 5 1
X X
2r  Dg ðr Þrφðg1Þ ðr Þ 1 Σr;g ðr Þφðg1Þ ðr Þ 2 Σs;g0 -g φðg10 Þ ðrÞ 5 χg νΣf ;g0 ðr Þφðg00 Þ ðr Þ
0 0
i 6¼i g

and for i 5 2
X X
2r  Dg ðr Þrφðg2Þ ðr Þ 1 Σr;g ðr Þφðg2Þ ðr Þ 2 Σs;g0 -g φðg20 Þ ðrÞ 5 χg νΣf ;g0 ðr Þφðg10 Þ ðr Þ
i0 6¼i g0

and so on. The group flux φg ðrÞ is then

X
N
φg ðr Þ 5 φðgiÞ ðrÞ; (5.66)
i50

which converges, for a critical system for some large N. The system multiplication keff is determined by balancing the
lloss and gain terms from the eigenvalue equation
X χg X
2r  Dg ðr ÞrφðgiÞ ðr Þ 1 Σr;g ðr ÞφðgiÞ ðr Þ 2 Σs;g0 -g φðgi0Þ ðr Þ 5 νΣf ;g0 ðr Þφðgi21
0
Þ
ðr Þ
0 k eff 0
i 6¼i g

where
ÐP
g0 νΣf ;g0 ðr Þφðgi0Þ dr
ðiÞ
keff 5 Ð P : (5.67)
g0 νΣf ;g0 ðr Þφðgi21
0
Þ
dr

The flux is normalized at each iteration, to satisfy the power output of a reactor;
1
φðgi0Þ ðr Þ 5 φðgi21
0
Þ
ðr Þ: (5.68)
keff ðiÞ
can be written in the discretized form by equating the currents at the interfaces allowing for different materials in each
mesh interval (Sekimoto, 2007) from Fig. 5.9
0
2r  Drφjk 5 ck21 φk21 1 dk φk 1 ek φk11 (5.69)
for meshes numbered k 5 1; 2; 3; . . . ; M: With

2eð11Þ 1ΣðsÞ ð1Þ ð1Þ ð1Þ

1-2 φ1 1 e1 φ2 5 1

the coefficients are given as

p
2ð p 1 1Þrk2 1
ck21 5 2  2 ;
p11 p11 Δrk21 Δrk
rk11 2r ðk21Þ 1
k21 2 2 Di DiðkÞ

p
2ð p 1 1Þrk1 1
ek 5 2  2 ;
p11 p11 Δrk Δrk11
rk11 2r ðkÞ 1
k21
2 2 Di Dðk11Þ
i
 The neutron diffusion equation Chapter | 5 237

FIGURE 5.9 Mesh intervals for finite difference computation.

and
0
dk 5 2 ck21 2 ek
where p 5 0; 1; 2 for slab, spherical and cylindrical geometries, respectively. In the above, the origin is at r1=2 and the
outer radius is R5rM11=2 . The radii at the midpoints of each mesh interval are r1 ; r2 ; r3 ; ?rK so that the mesh width of
the kth mesh is: Δrk 5 rk11=2 2 rk21=2 .
Further, c0 5 0, and for the first mesh, the point source boundary condition yields the equation

2 e1 1 ΣðsÞ
1-2 φ1 1 e1 φ2 5 1

where subscripts indicate energy and superscripts indicate the mesh interval. At the extrapolated boundary, it can be
shown (Park et al., 2020) that eiðKÞ is
p
2ð p 1 1ÞrK1 1
eiðKÞ 5 2  2 
p11 p11 ΔrK
rK11 2 rK21 ðKÞ 1 4:26
2 2 Di

The procedure is as follows

1. Discretize Eq. (5.63) to obtain a form for φk;g for mesh interval k and energy group g.
2. Apply boundary conditions.
3. Write the equations in matrix form

̿ 5 1 Bφ:
Aφ (5.70)
keff
4. The total number of mesh intervals for a one-dimensional problem is K; the surrounding nodes for the kth element
being xk21=2 and xk11=2 . Thus element 1 is from x1=2 to x3=2 . The column vector φ is of size K 3 G; thus for 100
mesh intervals and 2 groups, there are 200 fluxes to compute.
5. The matrix A̿ 5 L̿ 2 S̿ represents the three terms
X
2r  Dg ðr ÞrφðgiÞ ðr Þ 1 Σr;g ðr ÞφðgiÞ ðr Þ 2 Σs;g0 -g φðgi0Þ ðr Þ
i0 6¼i

the first two terms, the “loss terms,” have element matrices L̿ g ; g 5 1; 2; 3 . . . G as diagonal elements while the third ter-
minal S̿ is the scattering-in (lower triangular) matrix since down-scattering occurs from higher energy groups. Each ele-
ment matrix L̿ g is a tridiagonal matrix of size K 3 K.
238 Nuclear Engineering

6. The matrix B represent the RHS fission term

X
χg νΣf ;g0 ðr Þφðgi21
0
Þ
0
g

7. In the power method, for eigenvalues k1 ; k2 ; k3 ; . . . N, the solution at the ith iteration is

1
ðiÞ
φðiÞ 5 i21
A̿ 21B φð0Þ (5.71)
L kðiÞ
i0 51

as i-N, the flux φðiÞ converges to the flux φ and the largest eigenvalue is the value of keff .
Finally, the matrix can be solved by standard methods such as Gaussian elimination, Jacobi or Gauss-Siedel meth-
ods, or to reduce the memory advantage is taken of the sparse matrix and Choleski’s method is more efficient.
Computation is accelerated by the use of the Standard Overrelaxation Method (SOM) or a number of other methods.
As an example, for K mesh intervals and 2 groups, the matrices are
 
̿ A1 0
A5
0 A2

0 
in terms of the matrices A g , where the “group” coefficients c; d; e dk 5 d k 1 Σk are evaluated for each mesh interval.
The group matrices A g are
2 3
d1 e 1 0
6 c 1 d2 e 2 7
6 7
6 : : 7
6 7
6 : : : 7
6 7 (5.72)
6 c d e 7
6 k21 k k 7
6 : : : 7
6 7
4 0 : : 5
cK21 dK
The K 3 2 flux, for each group and mesh interval, are
T
φ 5 φ1;1 φ2;1 . . . φK;1 φ1;2 φ2;2 . . . φK;2 (5.73)
and the “source” term is a vector of length 2K (for the case of 2 groups) and B are the fission terms.

5.5 Effect of fuel concentration on critical mass

A research area of interest in nuclear engineering has been the effect of nonuniform fuel distribution on the critical
mass of nuclear systems.
The concept of minimum critical mass (MCM) by Goertzel (2003) is a classical and seminal problem in nuclear
engineering; over the last few decades, it has retained a strong interest and led to further work on the requirement of
flat thermal flux (FTF) as a necessary condition. Starting from a multigroup diffusion model by Goertzel and extending
to a two-group transport model, Williams (2003) has shown that the condition for MCM is that the integral of the prod-
uct of the thermal angular flux and a function related to the adjoint function is a constant and that an unsatisfactory
aspect of the Goertzel theorem (that of adding delta functions at the core-reflector interface) disappears in transport the-
ory. A number of models in reflected systems have been considered by Lewins (2004) in one- and two-group diffusion
formulations to further establish maximum and minimum loading conditions.
There is also an anomaly (Van Dam, 2015) that “the critical size of a MCM system with continuous in-core fuel dis-
tribution can take the same value for two different reflector thicknesses.” The thermal flux flattening has also been dem-
onstrated for heterogeneous systems of thin fuel “foils” in a moderator (Van Dam, 2013).
The objective is to distribute the fissile material in such a way as to achieve criticality with minimum fissile mate-
rial. Theoretical works have necessarily been based on assumptions and artifacts that need to be validated with elaborate
simulations; hence the motivation to include this work in this chapter.
 The neutron diffusion equation Chapter | 5 239

5.5.1 Goertzel’s theorem

The two-group uniformly distributed fuel spherical reactor considered in Section 5.3.2 had a critical radius of 21.9 cm
with a critical mass 1.15 kg U-235. Goertzel’s theorem states that the critical mass can be reduced if a FTF is achieved
in the core. This corresponds to a nonuniform fuel distribution which is covered in this section for slab and spherical
reflected reactors. Calculations are compared with MC simulations.

5.5.2 Nonuniform fuel distribution: a slab model

Two-group diffusion theory is used for obtaining flux and criticality estimates for both uniform fuel loading and for
loading resulting in a FTF which corresponds to MCM. An intriguing question regarding the “delta function” at the
core-reflector interface and critical conditions arising out of analytical studies based on one- and two-group diffusion as
well as two-group transport theory are illustrated for a “practical” core in which both minimum and maximum fuel
loadings are required for achieving a FTF which corresponds to MCM.
Here, MC simulation, with MCNP5 (Briesmeister, 2000) is used for validation of two-group diffusion results.
Consider a 1D slab of a dilute U235-H2O solution reflected by water. For an exact solution with uniform fuel load-
ing, followed by a FTF requirement, two-group diffusion equations are used. An “optimal” fuel loading is thus obtained
and used in a MC simulation. The criticality conditions are obtained for both cases and used to reverse the larger core
size by the addition of fuel at the core periphery.
These are discussed with Goertzel’s result that more fissile material should be added to the central regions and pro-
gressively less material toward the end of the core region. The work by Williams (2003) concludes that “the major dif-
ference between diffusion and transport theories arises from the origin of the fuel mass required at the edges,” and that
“without the ad hoc addition of delta functions, diffusion theory would fail. In the transport equation on the other hand
these features arise naturally.” A detailed MC simulation is thus a natural extension of this study to compare the validity
of both diffusion and transport models with results obtained from a detailed simulation where none of their assumptions,
such as constant scattering, are used.
In Eqs. (5.34)(5.37) if a FTF is assumed, that is, φ2c 5 φO , then Eq. (5.35) is simplified to 2Σ2c φo 1 Σ1c φ1c 5 0.
In the MCM problem, the objective is to find the fuel distribution NF ðxÞ which will result in minimum fuel mass Bm
ð a=2
Bm 5 α NF ðxÞdV (5.74)
0

where α 5 2A235 =Nav .

For a uniformly distributed fuel, described by Eqs. (5.34)(5.37) the coefficients in Eq. (5.35) are constant and solu-
tions are obtained by expressing the core equations as two 4th-order ODEs to yield the fluxes (Lamarsh & Baratta,
1955) with the resulting (transcendental) criticality equation for half-thickness ao
α1 λ tanhλao =2 1 α2
μ tanμao =2 5 (5.75)
α3 λ tanhλao =2 1 α4
where
α1 5 2κ2r D1c D2r ðS3 2 S1 Þ2κ1r ðD1r D2c S2 2 D1c D2r S3 Þ
α2 5 κ1r κ2r D1r D2r ðS1 2 S2 Þ
α3 5 2κ1r D1c D2c ðS2 2 S1 Þ
α4 5 2κ2r D1c D2r ðS3 2 S2 Þ2κ1r ðD1r D2c S1 2 D1c D2r S3 Þ
where and μ2 and λ2 and the coupling coefficients are defined in are defined in Section 5.3.2.
For FTF φO in the core, Eqs. (5.34) and (5.35) yield
Σ2Mc
φ1c ðxÞ 5 C1 cosγx 1 η φ (5.76)
γ 2 D1c o


φ1r ðxÞ 5 φ1r a=2 e2κ1r ðx2ja=2jÞ (5.77)
240 Nuclear Engineering

φ2r ðxÞ 5 fe2κ2r ðx2ja=2jÞ 1gφ1r ðxÞ (5.78)

where the constant C1 in Eq. (5.76) can be found from the reflector interface condition

 Σ2Mc
C1 cosγa 5 φ1r a=2 2 η 2 φ :
γ D1c o
Here

 D2r
φ1r a=2 5 ðκ1r 1 κ2r Þκ2r (5.79)
Σ1r
and
κ1r Σ1r 1
f5 φ andg 5 (5.80)
κ1r 2 κ2r o D2r κ2r 2 κ21r
2

with a=2 being the half-width of the core.

The criticality condition is then
κ1r
γκ1r S tanγa 5 1 1 (5.81)
κ1r
where
κ1r 1
S5 2 :
κ1r η21
Thus the critical half-thickness is
 
2 21 γD1c γτ c κ1r γΣ2r
a 5 cot 2 (5.82)
γ D1r κ1r ðη 2 1ÞD2r κ2r ðκ1r 1 κ2r Þ
leading to fuel loading
 
κ1r 1 κ2r cosγx Σ2Mc
Σ2F ðxÞ 5 Σ2r 1 : (5.83)
κ2r cosγa=2 η 2 1
The fuel mass for both cases, uniform loading and FTF, are then
ao
mF;o 5 αΣF;o (5.84)
2
and
 
αΣ2r κ1r a
mF 5 ðκ1r 1 κ2r Þτ c 1 : (5.85)
η 2 1 κ2r 2
The two-group data for uniform loading used is given in Table 5.6.
With a H2O-U235 ratio 5001 for uniform fuel loading (N5 5 6.6848 3 1019 atoms/cm3, NW 5 3.342 3 1022 mole-
cules/cm3) the criticality equation is
0:8011tanhλao =2 1 0:6416
tanμao =2 5 (5.86)
0:6989tanhλao =2 1 0:4442
ðμ 5 0:1128; λ 5 0:6509Þ from which the critical thickness was found to be 16.1319 cm (half-thickness 8.06595 cm)
with the two-group flux shown in Fig. 5.10 (mass 0.2106 g /cm2).
Subsequently, for a slab of thickness 8.06 cm, height and width 20 cm with reflecting surfaces, a MCNP5 simulation
with N5 5 6.7 3 1019, NH 5 6.7 3 1022 and NO 5 3.35 3 1022 atoms/cm3 (total 0.100567 3 1024) in kcode with
1000 histories per cycle and 500 cycles with 10 skip cycles and tallies accumulated in two groups: group 1 (0.625 eV
1 MeV) and group 2 (00.625 eV) gives the fluxes shown in Fig. 5.11.
The result for this simulation gives keff 5 0:99544ð0:00080Þ with a simulation time of 7.00 min on an Intel (R) Core
(TM) i72620M CPU @2.70 GHz processor, 32-bit Operating System.
 The neutron diffusion equation Chapter | 5 241

FIGURE 5.10 Two-group diffusion flux for a critical slab.

FIGURE 5.11 MCNP Two-group fluxes for uniform fuel dis-

tribution in a slab reactor. MCNP, Monte Carlo N-Particle.

The atom and mass fractions of U235 in the solution are 6.66223 3 1024 and 2.54368 3 1022 respectively. Each
core mesh has a volume of 32.56 cm3, and a mass of 33.4736 g, so that for 99 mesh intervals, the mass is 3.3139 kg.
The total fuel mass is thus 84.2947 g, equivalent to 0.2107 g/cm2, which matches very well with the diffusion estimate.
For FTF, the critical half-thickness, from the criticality equation
tanγa=2 5 2 3:8833 (5.87)
was found to be 9.1567 cm with flux shown in Fig. 5.12. Thus the ratio of critical dimensions for uniform and FTF can
be estimated as
ao γ 0:9007
 5 0:8719 (5.88)
a μ π 2 1:3188
The core region (half-thickness 9.1567 cm) is divided into 20 mesh intervals of equal width (0.4578 cm) and the
reflector of thickness 22.5 cm is also divided into 20 mesh intervals of equal thickness (1.1250 cm). The decrease in
Σ2F is from 4.9002 3 1022 to 1.3522 3 1022/cm over the core.
242 Nuclear Engineering

FIGURE 5.12 Two-group flux for flat thermal flux.

FIGURE 5.13 Fuel absorption cross-section.

The atomic density of U235, N5 , is found from N5 5 Σ2F =σa5 where the thermal fission cross-section for U235 is
taken to be σf ;5 5 0:886σa;Th5 b. The decrease in N5 is from 8.1574 3 1019 to 2.2510 3 1019 atoms/cm3 per mesh
while the H2O-to-U235 ratio increases from B410 to B1485 at the periphery. With this, the fuel mass decreases per
mesh from 14.572 to 4.02 mg/cm2 with a total fuel mass 0.2118 g/cm2 compared with a fuel mass of 0.2106 mg/cm2
for the case of uniform loading; hence no significant mass reduction.
With a cross-section area (depth times height) 1 cm2 the volume of the core would be 4.0787 cm3 so that the fuel
concentration would be 3.3589 mg/cm3 or 3.3589 g/L. These results are shown in Figs. 5.135.16.
A MCNP5 simulation was carried out with fuel loading of Fig. 5.16 shown in Fig. 5.17.
For a slab of thickness 9.1567 cm, with reflector thickness 22.5 cm, height and width 20 cm with reflecting surfaces,
with varying N5, NH 5 6.7 3 1022 and NO 5 3.35 3 1022 atoms/cm3 with 1000 histories per cycle and 500 cycles with
10 skip cycles. Tallies were accumulated in two groups: group 1 (0.625 eV1 MeV) and group 2 (00.625 eV) for
which the flux are shown in Fig. 5.17. The result for this simulation was: keff 5 1:00282ð0:0008Þ, with a simulation time
of 4.52 min on an intel processor.
 The neutron diffusion equation Chapter | 5 243

FIGURE 5.14 Fuel atomic density.

FIGURE 5.15 Water-to-fuel ratio.

The total fuel mass is 84.5594 g, equivalent to 0.2114 g/cm2, which matches very well with the diffusion estimate.
The effect of reducing the core size was found with the truncated flat flux fuel distribution taken till 7.7832 cm,
then continued for mesh intervals at 8.0, 8.02, and 8.06 cm. The mass reduction was 4.5066 g resulting in
keff 5 0:98171ð0:00087Þ. Addition of material toward the end is thus required. With the atomic density shown in
Fig. 5.18, that is, with a total fuel mass of 84.6945 g equivalent to 0.2117 g/cm2, the system has a multiplication
keff 5 1:00074ð0:00083Þ with flux shown in Fig. 5.19.
The addition of fuel at the edges are estimated by MC reruns which is inefficient and recommended to be carried
out by sampling for derivatives in a MC perturbation simulation (Koreshi & Lewins, 1990; Rief, 1984).
Thus, a uniform fuel distribution gave a critical half-thickness and fuel mass of 8.06595 cm, 0.2107 g/cm2 with diffusion
and 8.06 cm, 0.2106 g/cm2 keff 5 0:99544ð0:00080Þ with MCNP5 with the characteristic “hump” in the thermal flux at the
core-reflector interface. With FTF in the core, it was found that the critical half-thickness and fuel mass were 9.1567 cm,
0.2118 g/cm2 with diffusion and for the same thickness, fuel mass was 0.2106 g/cm2 keff 5 1:00282ð0:00080Þ with MCNP5.
244 Nuclear Engineering

FIGURE 5.16 Fuel loading for flat thermal flux.

FIGURE 5.17 MCNP flux with flat thermal flux fuel loading.
MCNP, Monte Carlo N-Particle.

5.5.3 Nonuniform fuel distribution: a spherical model

In a spherical homogenous U235-H2O core reflected with water, two-group diffusion theory is used for estimating the
critical size given an initial composition. Subsequently, solutions are obtained for a FTF in the core to obtain a fuel dis-
tribution which gives the diffusion estimate of MCM. These estimates are validated with MC simulations using the
MCNP5 code which shows the requirement of a “δ-source” at the core-reflector interface. In practical terms, the amount
of fuel mass to be added in a finite region at the interface is estimated by sensitivity coefficients sampled from MC per-
turbation derivatives. These are used to add fuel at the interface to estimate the MCM. Diffusion theory with uniform
composition gives a critical core of 21.0 cm radius with 1.15 kg fuel while a flat thermal flux gives 15.7668 cm com-
pared with MCNP which gives a critical mass of 718.067 g with a “corrected” fuel distribution. While Goertzel’s origi-
nal work found the MCM to be 30% less than that for uniform fuel distribution, detailed simulations find a reduction
by almost B38%.
This section considers a spherical core surrounded with water reflector. The core is a dilute U235-H2O solution ini-
tially with a water-to-fuel atomic ratio 500 to 1. For an exact solution with uniform fuel loading, followed by a flat
 The neutron diffusion equation Chapter | 5 245

FIGURE 5.18 Atomic density with addition at periphery.

FIGURE 5.19 Thermal flat flux in core with reduced size.

thermal flux (FTF) requirement, two-group diffusion equations are used. An “optimal” fuel loading is thus obtained and
used in a MC simulation.
For a uniformly distributed fuel, Section 5.3.2, the transcendental criticality equation for radius R is
α1 ζ Y ζ Z2 1 α2 ζ Y ζ Z1 1 α3 ζ Z1 ζ Z2
ζX 5 (5.89)
α4 ζ Y 1 α5 ζ Z1 1 α6 ζ Z2
with
0  
X 1
ζX  52μ 2 cotμR
X μR
0  
Y 1
ζ Y  5 λ cothλR 2
Y λR
0  
Z 1
ζZ  5 2 κ 1 cothκb
Z κR
246 Nuclear Engineering

and
α1 5 D1c D2r ðS3 2 S1 Þ; α2 5 D1r D2c S2 2 D1c D2r S3 ; α3 5 D1r D2r ðS1 2 S2 Þ
α4 5 D1c D2c ðS2 2 S1 Þ; α5 5ðD1r D2c S1 2 D1c D2r S3 Þ; α6 5 D1c D2r ðS3 2 S2 Þ:
For flat thermal flux φ2c 5φO in the core
Σ2Mc
φ1c ðr Þ 5 AX 1 η φ (5.90)
γ 2 D1c o
where
sinγr
X5 :
r
The constant A in Eq. (5.90) can be found from the core-reflector interface conditions
Σ2Mc
AX 1 η φ 5 FZ1 (5.91)
γ 2 D1c o
and
0 0
D1 C1 X 5 D1r FZ1 : (5.92)
Also, F 5 α1 G; andG 5 α2 φo where
Z2 ζ Z2
α1 5 2
Z1 ζ Z1
and
Z1 ζ Z1
α2 5 :
Z2 ζ Z1 2 ζ Z2
The criticality condition is then
1 γD1c
R 5 tan21 (5.93)
γ D1c 1 D1r f ðRÞ
where
ζ Z1 ζ Z2
f ð RÞ 5 (5.94)
βζ Z1 1 ð1 2 β Þζ Z2
and
Σ2Mc
β 5 S3 η
γ 2 D1c
leading to a fuel loading
sinγr Σ2Mc
Σ2F ðr Þ 5 A 1 : (5.95)
r η21
The fuel mass for both cases, uniform loading and FTF, are then
mF;o 5 αΣF;o Vc (5.96)
and
 
α 2 4πD1r
mF 5 AsinγRf ðRÞ 1 Σ2Mc Vc : (5.97)
η21 φo
The two-group data for uniform loading used is: D1c 5 1:13 cm, Σ1c 5 0:0419/cm, D2c 5 0:16 cm, Σ2c 5 0:060/cm,
σaW 5 0:664 b, σa25 5 678 b, τ c 5 27 cm2, ηT 5 2:07. It is also assumed that D1c 5 D1r ; D2c 5 D2r ; τ c 5 τ r . Then,
μ 5 0:1128; λ 5 0:6509.
 The neutron diffusion equation Chapter | 5 247

All MC simulations were performed on an Intel (R) Core (TM) i72620M CPU @2.70 GHz processor, 32-bit
Operating System with 1000 histories per cycle and 500 cycles with 10 skip cycles. Tallies were accumulated in two
energy groups: group 1 (0.625 eV1 MeV) and group 2 (00.625 eV).
With a H2O-U235 ratio 500:1 for uniform fuel loading (N5 5 6.6848 3 1019 atoms/cm3, NW 5 3.342 3 1022 mole-
cules/cm3) the criticality equation gives critical radius 21.9 cm with the two-group flux shown in Fig. 5.8 for which the
fuel mass is 1.15 kg.
Subsequently, for a core of radius 22.5 cm with a 22.5 cm thick water reflector a MCNP5 simulation with N5 5 6.7
3 1019, NH 5 6.7 3 1022 and NO 5 3.35 3 1022 atoms/cm3 for which the flux are shown in Fig. 5.20. The result for
this simulation was: keff 5 1.001364 (0.0011) with a simulation time of 6.52 min.
For a uniform fissile distribution, the critical radius and mass obtained from diffusion theory are close to the esti-
mates obtained from MCNP: radius 22.5 cm and mass 1.2477 kg U235 (0.0262 kg/L).

5.5.4 Critical core with flat thermal flux loading

For FTF, the critical radius, from the criticality equation was found to be 15.7668 cm with flux shown in Fig. 5.21.
The core and reflector regions are divided into 20 mesh intervals of equal volume in-core and reflector, respectively.
The decrease in Σ2F is from 0.1725 to 0.0266/cm over the core.
The atomic density of U235, N5 , is found from N5 5 Σ2F =σa5 where the thermal fission cross-section for U235 is
taken to be σf ;5 5 0:886σa;Th5 b. The decrease in N5 is from 2.8721 3 1020 to 4.4355 3 1019 atoms/cm3 per interval
while the H2O-to-U235 ratio increases from B116 to B754 (B1081 at the interface) at the periphery. These results are
shown in Figs. 5.225.25.
MC simulation with MCNP5 with the above fuel loading discretized as shown in Fig. 5.26 gives the fluxes shown
in Fig. 5.27.
For a sphere of radius 15.7668 cm, with reflector thickness 22.5 cm, with varying N5, the flux are shown in Fig. 5.9.
For this simulation, keff 5 0:99447ð0:0011Þ, with a simulation time of 4.02 min. The value of keff indicates a require-
ment of more fuel in the core.
The total fuel mass is 702.8952 g, or 0.0428 kg/L.
Now consider the effect on the criticality of addition of fuel toward the interface to correct the keff . To estimate the
amount of fuel to be added at the interface, sensitivity coefficients are obtained for a relative mass change of
δm=mo 5 6 5% with the uniform fuel distribution taken as the “reference” design point. The relative importance of the
intervals, for a mass increase, is maximized at the center and steadily decreases toward the interface where it is seen to
increase, suggesting the addition of mass in the last interval, the δ-source addition. For a 5% fractional increase, in the
interface interval, the increase in system multiplication is δkeff B9.66 3 1024.

FIGURE 5.20 MCNP flux uniform distribution. MCNP,

Monte Carlo N-Particle.
248 Nuclear Engineering

FIGURE 5.21 Two-group flux for flat thermal flux (spherical

reactor).

FIGURE 5.22 Fuel absorption cross-section.

The addition of 15.1695 g fuel in the last interval increases the critical mass to 718.067 g U235 and the system multi-
plication from keff 5 0:99447ð0:0011Þ to 0.999495 (0.0013), that is, δkeff B5 3 1023 which corresponds to a B25%
fractional increase in the mass of the interval at the interface.
It is also reported in the literature that for “optimum moderation” the smallest critical mass in a water reflected
spherical U235 aqueous solution is 0.784 kg. Table 5.8 shows the results from diffusion and MC simulation; with MC
giving 1.2477 kg U-235 compared with 1.150 kg from two-group diffusion theory. The flat thermal flux (FTF) reduces
the critical mass to 0.7029 kg U-235.

5.6 The two-group adjoint diffusion equations

1
For the one-group diffusion equation which is a second-order ODE the operator L^ is self-adjoint, that is, L^ 5 L.
^ In
contrast, the two-group equations, with χ1 5 1; χ2 5 0; Σf ;1 5 0 are



2 r  D1 r 1 Σr;1 φ1 2 νΣf ;2 φ2 5 0 (5.98)
2 r  D2 r 1 Σr;2 φ2 2 Σs;12 φ1 5 0 (5.99)
 The neutron diffusion equation Chapter | 5 249

FIGURE 5.23 Fuel atomic density.

FIGURE 5.24 Water-to-fuel ratio (spherical reactor).

and with Σs;12 5 pΣs1 , the above are written as

  
div D1 grad 2 Σr;1 νΣf ;2 φ1
50 (5.100)
pΣs1 div D2 grad φ2
or, in matrix form,
̿ 50
Mφ (5.101)
1
In this case, it is observed that the adjoint operator M̿ is the transpose of M , that is, the rows and columns are
̿

interchanged. Thus
 T  
1 div D1 grad2Σr;1 νΣf ;2 div D1 grad 2 Σr;1 pΣs1
M̿ 5 5 (5.102)
pΣs1 div D2 grad νΣf ;2 div D2 grad
250 Nuclear Engineering

FIGURE 5.25 Fuel-to-water ratio.

FIGURE 5.26 Fuel loading for flat thermal flux (spherical

reactor).

and the set of equations is thus

  
divD1 grad 2 Σr;1 pΣs1 φ1
1 50 (5.103)
νΣf ;2 divD2 grad φ1
2

The adjoint equations are thus, like the forward equations, coupled ordinary second-order differential equations
which satisfy the same boundary conditions as the flux, that is, flux and current continuity at a physical interface
and vanishing flux at the extrapolated boundary. The adjoint flux is an importance function which has physical sig-
nificance in sensitivity analysis as will be demonstrated for variational formulations used in sensitivity analysis
and optimization.
 The neutron diffusion equation Chapter | 5 251

FIGURE 5.27 MCNP flux with flat thermal flux fuel

loading (spherical reactor). MCNP, Monte Carlo N-
Particle.

TABLE 5.8 Criticality estimates for a spherical reactor.

Criticality MCNP 2-group diffusion

FTF δ 2 source Flat thermal flux (FTF) fuel dist. Uniform Uniform
Rc (cm) 15.7668 15.7668 22.5 21.9
ΔRr (cm) 22.5 22.5 22.5 18.1
keff 0.999495 (0.0013) 0.99447 (0.0011) 1.001364 (0.0011) 1.0
M (kg) 0.718067 0.7028952 1.2477 1.150

5.7 Core neutronics with diffusion equations

Core neutronics calculations are performed over the whole core consisting of fuel assemblies which have fuel rods as
well as control rods in the presence of a moderator or coolant. The most basic unit of an assembly is the unit lattice
cell. This cell is representative of the fuel and moderator; as an entity it repeats itself within an assembly.
For the AP1000 reactor (Table 3.3) the core, shown in Fig. 5.28 has an equivalent diameter 3.04 m and an active
core height 4.267 m.
Within the core are 157 assemblies with 17 3 17 rods in an arrangement represented in Fig. 5.29.
Each assembly, Fig. 5.30, is of dimensions 21.40204 3 21.40204 cm.
Within an assembly a unit cell of dimensions 1.26 3 1.26 cm, shown in Fig. 5.31, typically consists of fuel, IFBA
(integral fuel burnable absorber), a helium gap, and cladding.
The fuel pellet is of radius 4.09575 mm, the IFBA is of thickness 0.00518 mm, the helium gap is of thickness
0.59202 mm, and the clad thickness is 0.05715 mm. Thus the outer fuel radius is 4.7475 mm (diameter 0.94950 cm). The
unit cell represents basic nuclear characteristics of the fuel assembly; fluxes and reaction rates are determined within the unit
cell. Lattice parameter programs are written to carry out calculations within this cell. The cell is typically divided into three
regions, namely the fuel, the gap and the coolant. Within a cell, detailed calculations are performed to obtain its overall para-
meters which, in the second stage are used to carry out whole-core calculations. If fuel elements are placed at the corners of
squares, the lattice is called a square lattice. In some configurations, hexagonal lattices are modeled when fuel elements are
placed at vertices of a hexagon. Heterogeneity within a reactor core, for example when strong absorbers are placed within a
252 Nuclear Engineering

FIGURE 5.28 AP1000 core (equivalent radius 3.04 m, active core

height 4.267 m).

FIGURE 5.29 A typical PWR assembly with 17 3 17 rods. PWR, pressurized water reactor.
 The neutron diffusion equation Chapter | 5 253

FIGURE 5.30 Fuel assembly of AP1000.

FIGURE 5.31 Unit cell of AP1000.

core, is modeled by the use of macro cells. A detailed transport calculation is carried out for a cell and in the second stage, a
quarter model of the core, as shown in Fig. 5.32, is used with a less cumbersome diffusion model.
Codes produce lattice parameters to study the sensitivity of burnup, for example, to changes in operating conditions.
In a detailed calculation, starting from an initial configuration, a transport equation solution normalized to the given
reactor power, is used for small time scales using the point reactor kinetics equations and for the long time scale using
the burnup equation to calculate the isotopes produced (Section 4.4), and subsequently the new overall reactivity. A
code such as WIMSD (Deen & Woodruff, 1995) is a typical reactor lattice code which at the first level, calculates the
pin cell parameters in a 69-group computation. These are used for few-group calculations with detailed geometry mod-
els, followed by burnup calculations in an iterative procedure. In WIMSD, unit cell calculations are carried out with the
integral transport equation using collision probability methods. One such code is Serpent (Leppänen, Pusa, Viitanen,
Valtavirta, & Kaltiaisenaho, 2015), a multipurpose 3D continuous energy MC particle transport code distributed by
OECD/NEA for neutron/photon transport with burnup calculation capability and coupled multiphysics simulations to
generate homogenized few-group reaction cross-sections among several other parameters.
Here, an insight is given on the use of the NDE to obtain core fluxes and power distribution for which methods
include the FDM, the coarse mesh FDM, and the NEM for structured meshes. In the case of unstructured meshes
required for complex geometries, FEM and the FVM are used.
In the FDM methods for the NDE, the DIF3D code developed at Argonne National Laboratory and used for decades comes
upgraded with diffusion and transport solvers (Nelson, Smith, & Heidet, 2021). In the nodal methods, for example, a coarse
mesh analysis is carried out, with one node representing a fuel assembly; this is preferred over the use of FDM and FEM due
to the computational efficiency of nodal methods though at the expense of detailed information especially if the heterogeneity
254 Nuclear Engineering

is large, that is, when the number of control rods in an assembly is large. Some of the widely used nodal codes developed for
thermal reactors are KIKO3D (Keresztúri et al., 2003), DYN3D (Rohde et al., 2016) and RAST-K (Park et al., 2020).
In one approach (Trkov, Najžer, & Škerget, 1990), the 2D (xy) diffusion equation for the quarter core of Fig. 5.32 is
integrated over the y axis to obtain a 1D equation for each energy group. The nodal fluxes and currents are then
obtained analytically. In another approach, the Green;s function is used for a solution of the NDE in which continuity
conditions result in a system of equations giving nodal fluxes and currents.
The multigroup NDE for energy group g
2r  Dg ðr Þrφg ðr Þ 1 Σr;g ðr Þφg ðr Þ 5 fg ðr Þ (5.104)
with the “source” term
X χg X
fg ðr Þ 5 Σs;g0 -g φg0 ðr Þ 1 νΣf ;g0 ðr Þφg0 ðr Þ: (5.105)
keff 0
g0 6¼g g

in 2D rectangular homogeneous zones, are integrated over the transverse direction y and using Fick’s law, a set of one-
dimensional equations are obtained for the quantities
ð
1 Vj
φ xð j Þ ð x Þ 5 φðx; yÞdy: (5.106)
hð jÞ Vj21
For a node Vði;jÞ for column i, xi21 # x # xi width hi and row j, yj21 # y # yj height hj , integration gives the transverse
leakage
ð 


1 Vj @2 ny J x; yj 2 J x; yj21
Lyð jÞ ðxÞ 5 2 φðx; yÞdy 5 (5.107)
hð jÞ Vj21 @y2 hð jÞ Dð jÞ
resulting in the set of equations
d 2 φ xð j Þ
2DðijÞ 1ΣðijÞ φxð jÞ 5fxð jÞ 2DðijÞ ðr ÞLyðjÞ : (5.108)
dx2
Eq. (5.108) are then solved to obtain the nodal fluxes φxð jÞi and current Jxð jÞi on the node boundaries i. These give the
average leakage in the x direction
 
1 Jxð jÞi 2 Jxð jÞi21
LxðijÞ 5 : (5.109)
hðiÞ DðijÞ

FIGURE 5.32 One-fourth model of a reactor core with 52

assemblies.
 The neutron diffusion equation Chapter | 5 255

The integrations are performed in the x direction over a node, with a weighting function wðxÞ which is the Green’s
function to the problem. This reduces the integrals to an algebraic expression involving the flux φi21 ; φi and average
current Ji21 and Ji in the node and at boundaries. All such equations for a a row along the x direction, applying continu-
ity conditions for interfaces and external boundaries, are assembled into a global matrix and solved for fluxes and cur-
rents. The same procedure is repeated by integrating over the transverse direction.
In this method, there are no trial functions or a priori assumptions about the shape of the neutron flux.
These methods have been demonstrated for IAEA 2D and 3D PWR BSS-11 and BIBLIS benchmarks with a 20 cm
coarse mesh giving overall system multiplication and power distribution within 4%5%.
An application of an intermediate mesh method, such as FEM, with a coarse method, such as the Spectral Green’s
Function (SGF) Nodal Method (Rocha, Dominguez, Iglesias, & de Barros, 2016), has been demonstrated to solve the
one-group multiplying media NDE for solving a benchmark problem with a high degree of heterogeneity. The
equations
dJðxÞ 1
1 Σa ðxÞφðxÞ 5 νΣf ðxÞφðxÞ (5.110)
dx keff
and
dφðxÞ
J ðxÞ 5 2 Dð xÞ (5.111)
dx
are discretized over a spatial domain 0 # x # L, with albedo boundary conditions J1=2 5 2 αL φ1=2 , and
JI11=2 5 αR φI11=2 on both ends of a grid of I mesh intervals. The above balance equations are multiplied by the
Legendre polynomial
 
2ð x 2 x i Þ
Pl
hi
integrated over an element xi21=2 # x # xi11=2 to get averaged flux φ i ; φ^ i and current J i ; J^i equations for the zeroth and
first moments
ð
1 xi11=2
φi 5 φðxÞdx; (5.112)
hi xi21=2
ð  
3 xi11=2 2ðx 2 xi Þ
φ^ i 5 φðxÞdx; (5.113)
hi xi21=2 hi
ð
1 xi11=2
Ji 5 J ðxÞdx; (5.114)
hi xi21=2

and
ð xi11=2  
3 2ð x 2 x i Þ
J^i 5 J ðxÞdx: (5.115)
hi xi21=2 hi

The spectral parameters αi ; β i are incorporated into the flux and current formulation through the FEM linear
functions
2β i
φðxÞ 5 αi φ i 1 ðx 2 xi Þφ^ i (5.116)
hi
and
2β i
J ðxÞ 5 αi J i 1 ðx 2 xi ÞJ^i : (5.117)
hi
The resulting algebraic equations contain the spectral parameters which are found by analytic solutions first pro-
posed by Case & Zweifel (1967). The linear system of equations for ð6I 1 2Þ unknowns, that is, I values each of
φ i ; φ^ i ; J i ; J^i ; αi ; β i and two end points, are reduced to ð2I 1 2Þ unknowns, and the same number of equations, by elim-
inating φ i ; φ^ i ; J i ; J^i , since they are expressed in terms of the eigenvalues and αi ; β i are subsequently determined. Thus
256 Nuclear Engineering

neutron flux distribution is calculated by direct or iterative techniques and the system multiplication keff is calculated
by the power method described in the previous section.
Rocha et al. (2016) applied the hybrid FEM-SGF method described above to a six-region benchmark and compared
results with a reference 960 element FEM calculation to find that, in the presence of strong heterogeneities, the accu-
racy in flux distribution was within 0.01% for 30 elements in a coarse mesh. This method appears to be promising for
multigroup and higher dimensional eigenvalue calculations.
For unstructured meshes, the FVM has been demonstrated (Bernal, Miró, Ginestar, & Verdú, 2014) for 2D and 3D
two-group NDE applicable to LWRs with the requirement of complex geometry modeling. For 2D reactor models, the
computational time is reported to be in seconds while for 3D models, the time is in hours for fine meshes. A full 3D
Nodal Diffusion Code for PWRs (Park et al., 2020) performs steady-state and transient analyses with generated micro-
scopic cross-sections for 37 isotopes, and can handle a depletion chain with 22 actinides (uranium, plutonium. neptu-
nium, americium and curium isotopes) and 15 fission products (xenon, iodine, samarium, promethium, and
neodymium) and burnable poisons. The multigroup NDE are also solved with parallel computing using the two-node
nodal method with the Coarse Mesh FDM (Song, Yu, & Kim, 2018) reducing the time for quarter-core and whole-core
calculations. For multiphysics full 3D calculations, the computer code, RAST-K-v2, performs calculations with four
nodes for a fuel assembly with 2446 axial nodes. It can carry out core design, incorporating neutronics, thermal
hydraulics, critical heat flux, departure from nucleate boiling (DNBR) and depletion calculations for prediction of iso-
tope inventory in spent nuclear fuel. The RAST-K-v2 has been verified and validated for several Korean PWRs includ-
ing the OPR-1000, APR-1400 and Westinghouse 2-loop and 3-loop reactors. The diffusion equation models are thus
valuable for carrying out realistic whole-core calculations with parallel computing capabilities.

Problems
1. Consider slabs of aluminum, boron, beryllium, carbon, and iron of thickness 3 m.f.p each with a planar source at
x 5 0 emitting 106 neutrons/s. Plot the flux and estimate the surface leakages from their sides (Table 5.9).
2. Show that the neutron flux in a sphere of radius R with a point isotropic source at r 5 0 emitting S neutrons/s is
given by
S sinhκðR 1 d 2 rÞ
ϕ5
d
4πDsinh R 1
L
r
Find the leakage probability from the surface of a graphite sphere of radius 2 m.f.p. with a 1 Ci 226Ra-Be point
source at its center emitting 107 neutrons/s.
3. Compare the leakage probability of the above sphere if instead of graphite, the material was boron.
4. What is the magnitude of the flux at the center of a bare cubical reactor of sides 3 m operating at a power of
100 kW(th)?
5. Calculate the geometric buckling for a cylinder of radius 1 m and height 2 m. What information does this give on
the infinite multiplication of the system?
6. How would you calculate the number of absorption and scattering reactions in a graphite sphere if the atomic den-
sity of the graphite is N ðr Þ 5 No ð1 2 r=RÞ where No is the nominal density?

TABLE 5.9 Thermal cross-sections.


Nuclide Atomic weight Density g=cm3 σa ðbÞ σs ðbÞ
Aluminum 26.9815 2.70 0.235 1.4
Boron 10.811 2.30 759 4.0
Beryllium 9.0122 1.85 0.0092 6.14
Carbon (graphite) 12.01115 1.60 0.0034 4.75
Iron 55.847 7.87 2.53 11.0
 The neutron diffusion equation Chapter | 5 257

Nomenclature
English lower case
d extrapolation distance
keff effective multiplication
kN infinite multiplication
n number density
000
q volumetric heat generation (W/m3)

English upper case

Ai area
Bg geometrical buckling
Bm material buckling
Ci concentration of precursor
D diffusion coefficient
D^ derivative operator
J neutron current
L diffusion length
1
L^ adjoint operator
Ni atomic density of the ith nuclide
Ni shape function in the ith element
Pes escape probability
Pl Legendre functions
Rc critical radius
S source

Greek lower case

λ decay constant
μ cosine of angle of scattering
μ0 average cosine of scattering angle
φ flux
φ1 adjoint flux
φC complementary solution
φP particular solution
τ neutron age
ν number of neutrons produced per fission
χ fission spectrum

Greek upper case

Ω solid angle
P
macroscopic absorption cross-section
Pa
macroscopic fission cross-section
Pf
macroscopic removal cross-section
Pr
macroscopic scattering cross-section
Ps
macroscopic transport cross-section
Ptr
t macroscopic total cross-section

Abbreviations
FD finite difference method
FEM finite element method
FVM finite volume method
IFBA integral fuel burnable absorber
NDE neutron diffusion equation
NEM nodal expansion method
258 Nuclear Engineering

References
Bernal, Á., Miró, R., Ginestar, D., & Verdú, G. (2014). Resolution of the generalized eigenvalue problem in the neutron diffusion equation discretized
by the finite volume method. Abstract and Applied Analysis.
Briesmeister, J. F. (2000). MCNP  A general Monte Carlo N-particle transport code,” Los Alamos National Laboratory.
Carrillo, H. R. V. (2001). The source condition in neutron diffusion theory. Revista Mexicana de fı́sica.
Case, K. M., & Zweifel, P. F. (1967). Linear transport theory. Reading, MA: Addison-Wesley Pub. Co.
Deen, J. R., & Woodruff, W. L. (1995). WIMS-D4M user manual Rev.0. Illinois: Champaign-Urbana.
Goertzel, G. (2003). Minimum critical mass and flat flux. Journal of Nuclear Energy.
Jevremovic, T. (2009). Nuclear principles in engineering (2nd ed.).
Keresztúri, A., Hegyi, G., Marázcy, C., Panka, I., Telbisz, M., Trosztel, I., & Hegedüs, C. (2003). Development and validation of the three-
dimensional dynamic code—KIKO3D. Annals of Nuclear Energy, 30(1), 93120. Available from https://doi.org/10.1016/S0306-4549(02)00043-9,
ISSN 0306-4549.
Koreshi, Z. U., & Lewins, J. D. (1990). Two-group Monte Carlo perturbation theory and applications in fixed-source problems. Progress in Nuclear
Energy, 24, 2738.
Lamarsh, J. R. (1966). Introduction to nuclear reactor theory. Addison-Wesley Publishing Company.
Lamarsh, J. R. (2005). Introduction to nuclear reactor theory. Interactions.
Lamarsh, J. R., & Baratta, A. J. (1955). Introduction to nuclear engineering. Prentice Hall.
Leppänen, J., Pusa, M., Viitanen, T., Valtavirta, V., & Kaltiaisenaho, T. (2015). The Serpent Monte Carlo code: Status, development and applications
in 2013. Annals of Nuclear Energy.
Lewins, J. (2004). Minimum critical mass and flat flux in a 2-group model, Annals of Nuclear Energy, 31, pp. 541-576.
Nelson, A. G., Smith, M. A., & Heidet, F. (2021). Verification of the diffusion and transport solvers within DiF3D for 3D hexagonal geometries. In
Proceedings of the EPJ web of conferences.
Park, J., Jang, J., Kim, H., Choe, J., Yun, D., Zhang, P., Cherezov, A., & Lee, D. (2020). RAST-K v2—Three-dimensional nodal diffusion code for
pressurized water reactor core analysis. Energies, 13(23), 6324. Available from https://doi.org/10.3390/en13236324.
Rief, H. (1984). Generalized Monte Carlo perturbation algorithms for correlated sampling and a second-order Taylor series approach. Annals of
Nuclear Energy.
Rocha, R. V. M., Dominguez, D. S., Iglesias, S. M., & de Barros, R. C. (2016). Finite Element Method with spectral Green’s function in slab geome-
try for neutron diffusion in multiplying media and one energy group. TEMA (São Carlos).
Rohde, U., Kliem, S., Grundmann, U., Baier, S., Bilodid, Y., Duerigen, S., . . . Mittag, S. (2016). The reactor dynamics code DYN3D  models, vali-
dation and applications. Progress in Nuclear Energy, 89, 170190. Available from https://doi.org/10.1016/j.pnucene.2016.02.013, ISSN 0149-
1970.
Sekimoto, H. (2007). Textbook nuclear reactor theory. Tokyo Institute of Technology.
Song, S., Yu, H., & Kim, Y. (2018). One-node and two-node hybrid coarse-mesh finite difference algorithm for efficient pin-by-pin core calculation.
Nuclear Engineering and Technology.
Trkov, A., Najžer, M., & Škerget, L. (1990). Variant of Green’s function nodal method for neutron diffusion. Journal of Nuclear Science and
Technology, 27, 766777.
Van Dam, H. (2013). Minimum critical mass in a heterogeneous thermal system. Annals of Nuclear Energy.
Van Dam, H. (2015). On the anomaly in sizes of minimum critical mass cores. Annals of Nuclear Energy.
Williams, M. M. R. (2003). Some further considerations on Goertzel’s minimum critical mass problem. Annals of Nuclear Energy.
Chapter 6

The neutron transport equation

The neutron transport equation (NTE) is a mathematical statement on the conservation of neutrons for the angular,
rather than scalar, neutron flux in phase space. It is a more accurate description of the underlying phenomena of neutron
transport than the neutron diffusion equation (NDE) described in Chapter 5.
There are several mathematical formulations of the NTE such as the integral form (Section 4.3) and the
integro-differential form (Section 4.4) extensively covered in standard nuclear engineering text books (Bell &
Glasstone, 1979; Case & Zweifel, 1967; Clark & Hansen, 1964; Davison, 1957; Duderstadt & Hamilton, 1976;
Henry, 1975; Williams, 1971). The NTE can be interpreted as the zeroth moment of the Boltzmann transport equa-
tion which Boltzmann had derived for the kinetic theory of gases (Cercignani, 1988; Harris, 1971; Robson, White,
& Hildebrandt, 2017) and is also applicable to radiation transport including charged particles, photons and neutrons
(Section 1.6).
This chapter begins with a derivation of the integro-differential form of the NTE and shows some simple forms of
the NTE, such as the one-speed equation with isotropic scattering, the one- and two-group transport equations similar to
the two-group diffusion equations, the integral form, the collision and transport formulations, and the multigroup
equations.
In Section 6.2, some classical exact, analytical, closed-form solutions obtained in the early days of neutron transport,
the 1950s
70s, under idealized conditions with the Fourier and Laplace transforms are discussed. Classical solutions
for the for infinite, semiinfinite, and some finite media were found with simplified scattering models. These are
described here for the purpose of familiarizing the reader with the early approaches for obtaining the neutron flux and
eigenvalues.
The compact exact solutions are then followed by the development and application of numerical methods described
in Section 6.3. The most extensively used methods are the discrete ordinates (SN ) method, and the spherical harmonics
(PN ) method. These are followed by the double PN , or the DPN method, the BN method and Finite Element Method
(FEM) described in Section 4.5, nodal and hybrid methods. Such methods are used extensively for neutron transport
computations (Kang & Hansen, 1973; Lewis & Miller, 1984) for the design of nuclear reactor cores by calculations on
lattice cells, in fuel assemblies and in whole-core neutronic analyses as discussed in the context of the NDE
(Section 5.6).
In Section 6.4, two widely used methods for reactor calculations, namely the Collision Probability Method (CPM)
and the Method of Characteristics (MOC), are described for lattice cell calculations and for whole-core eigenvalue cal-
culations for the design of nuclear systems.
Current research and applications of transport theory are focused on the development of efficient computational
methods and algorithms on multiprocessors to carry out multidimensional core neutronics calculations for the design of
Gen III 1 and Gen-IV nuclear reactors described in Chapter 3.
It is important to realize that neutron transport contains some of the most elegant mathematics applied in nuclear
engineering, and there are several methods and techniques, each impressive in its own right. It is not possible to include
a more detailed introduction in a single chapter; indeed several books are devoted to neutron transport. The reader is
advised to look into the research publications in journals such as Progress in Nuclear Energy, Annals of Nuclear Energy
and Nuclear Science and Engineering for the latest trends and developments. The body of research contains new devel-
opments on making transport calculations more robust, applicable and fast. The gap between deterministic and stochas-
tic computation is thus bound to become narrower and large reactor design calculations will become easier in the years
to come.
Whatever numerical and computational advancements take place, the role of elegant mathematics will never end, as
the foundations of neutron transport will always lie in the compact analytical expressions of the early days which will
serve as standards for huge computational codes.

Nuclear Engineering. DOI: https://doi.org/10.1016/B978-0-323-90618-0.00006-5

© 2022 Elsevier Inc. All rights reserved. 259
260 Nuclear Engineering

An understanding of neutron diffusion, neutron transport, and in the later chapters, Monte Carlo simulation are the
bedrock upon which nuclear engineering will always stand. It is with this objective that the following sections must be
read to understand the equations.

6.1 Structure of the neutron transport equation

There are at least eight forms in which the NTE, in addition to its most basic integro-differential form, is used
(Ganapol, 2008) viz (1) integral, (2) even-odd parity, (3) slowing down kernel, (4) multiple collision, (5) invariant
embedding, (6) singular integral, (7) pseudoflux, and (8) Green’s function, in addition to the Monte Carlo form which,
though a stochastic simulation of the transport phenomena, can be interpreted to be a Neumann series solution of the
integral equation. Each one of these forms has mathematical properties that enable a class of solutions.
In this chapter, two of these eight (deterministic) forms will be considered in some detail.

6.1.1 An integro-differential form of the neutron transport equation

For a derivation of the NTE from first principles, consider a neutron with energy E at time t at some position
r 5 x^i 1 y^j 1 zk^ where i;^ j;^ k^ are unit vectors along the x; y; z axes moving in a direction defined by two angles θ; ϕ such
~
that the unit vector is Ω ^ 5 Ωx^i 1 Ωy^j 1 Ωz k, ^ where Ωx 5 sin θ cos ϕ, Ωy 5 sin θ sin ϕ, Ωz 5 cos θ. These parameters
define its “position” Pðr;Ω; E; tÞ in phase space. Let’s say in the next Δt seconds, it moves a distance vΔt, during which
the total number of interactions it is likely to have is Σt vΔt; conversely, the number of interactions it is likely not to
have are 1 2 Σt vΔt, which means that if there were N ðr;Ω; E; tÞdVdEdΩ neutrons in a packet initially, there are
r; Ω; E; tÞ½1 2 Σt ð~
N ð~ r; E; tÞvΔtdVdEdΩ neutrons left in the packet (Fig. 6.1).
The change in the number of neutrons is thus
Change 5 N ðr 1 vΩΔt;Ω; E; t 1 ΔtÞ 2 N ðr;Ω; E; tÞ:
This can be written as
Change 5 N ðr 1 vΩΔt;Ω; E; t 1 ΔtÞ 2 N ðr;Ω; E; t 1 ΔtÞ 1 N ðr;Ω; E; t 1 ΔtÞ 2 N ðr;Ω; E; tÞ
where the term N ðr;Ω;E;t 1 ΔtÞ has been added and subtracted to be able to express the four terms as a space deriva-
tive and a time derivative, respectively. In the limit as Δt-0, the first term is vΩ∙rN ðr; Ω; E; tÞ; the change in the
packet, per unit volume interval, per unit energy interval and per unit solid angle interval, is thus
@
v∙ΩrN ðr;Ω; E; tÞ 1 N ðr;ΩE; tÞ:
@t
We can write v∙ΩrN ðr;Ω; E; tÞ as r∙ΩvN ðr;Ω; E; tÞ and with angular flux φðr;Ω; E; tÞ and the neutron current
J ðr;E; tÞ defined as
φðr;Ω; E; tÞ 5 vN ðr;Ω; E; tÞ (6.1)

FIGURE 6.1 Coordinates in three-dimensional Cartesian axes.

 The neutron transport equation Chapter | 6 261

and
ð 4π
J ðr;E; tÞ 5 Ωφðr;Ω; E; tÞdΩ (6.2)
0

the leakage term is

ð 4π
r∙ Ωφðr;Ω; E; tÞdΩ 5 r∙J ðr;E; tÞ: (6.3)
0

The change in the neutron population is due to the gains and losses expressed as
Change 5 Gains 2 Losses (6.4)
The gains are due to in-scattering, fissions, an independent source, and any other gains due to inelastic scattering or
other nuclear reactions:
ðN ð ðN ð
1
dE0 dΩ0 Σs ðr; Ω0 ∙Ω; E0 Þφðr;Ω0 ;E0 ; tÞ 1 χð E Þ dE0 dΩ0 νΣf ðE0 Þφðr;Ω0 ;E0 ; tÞ 1 Sðr;Ω; E; tÞ (6.5)
0 4π 0

Eq. (6.5) can be modified to separate the prompt and delayed fission neutrons by expressing the fission contribution
as
ðN ð
1 1 X6
χ ðE Þ dE0 dΩ0 ½1 2 β ðE0 ÞνΣf ðE0 Þφðr;Ω0 ;E0 ; tÞ 1 χ ðEÞλj C j ðtÞ: (6.6)
4π p 0 4π i51 j

The losses due to the leakage of neutrons and their scattering out from the phase space of interest Pðr;Ω; E; tÞ are
r∙J ðr;Ω; E; tÞ 1 Σt ðr; EÞφðr;Ω; E; tÞ: (6.7)
Now with the gain and loss terms substitued into Eq. (6.4) gives the time-dependent balance equation
ðN ð
1@
φðr;Ω; E; tÞ 5 dE0 dΩ0 Σs ðr; Ω0 ∙Ω; E0 Þφðr;Ω0 ;E0 ; tÞ
v @t 0
ðN ð
1
1 χ ðE Þ dE dΩ0 ½1 2 β ðE0 ÞνΣf ðE0 Þφðr;Ω0 ;E0 ; tÞ
0

4π p 0

1 X6
1 χ ðEÞλj Cj ðtÞ 2 ½r∙J ðr;Ω; E; tÞ 1 Σt ðr; EÞφðr;Ω; E; tÞ 1 Sðr;Ω; E; tÞ: (6.8)
4π i51 j

Eq. (6.8) is written with leakage and out-scattering on the LHS and gains on the RHS; rearranging gives the familiar
form of the Linear Transport Equation (Bell & Glasstone, 1979)
ðN ð
1@
φðr;Ω; E; tÞ 1 r∙J ðr;Ω; E; tÞ1Σt ðr; EÞφðr;Ω; E; tÞ 5 dE dΩ0 Σs ðr; Ω0 ∙Ω; E0 Þφðr;Ω0 ;E0 ; tÞ
0

v @t 0

ðN ð
1 1 X6
1 χ ðE Þ dE 0
dΩ0 ½1 2 β ðE0 ÞνΣf ðE0 Þφðr;Ω0 ;E0 ; tÞ 1 χ ðEÞλj C j ðtÞ 1 Sðr;Ω; E; tÞ (6.9)
4π p 0 4π i51 j

where (Section 6.2) the scattering cross section is written in terms of the pre-collision energy and a function based on
the collision dynamics.

Σs ðr; Ω0 ∙Ω; E0 Þ 5 Σs ðr; E0 Þf ðΩ0 -Ω;E0 -EÞÞ:

Since scattering depends on the dot products of the two directions, rather than the angles themselves, a further simplifi-
cation is
f ðΩ0 -Ω;E0 -EÞ 5 f ðΩ0 ∙Ω;E0 -EÞ
262 Nuclear Engineering

with the normalization

ðN ð
dE dΩf ðΩ0 ∙Ω;E0 -EÞ 5 1;
0

where integration is over the solid angle 4π steradians. The conservation of energy and momentum is expressed as
 
  1
f ðΩ0 ∙Ω;E0 -EÞ 5 p μ0 ; E0 δ E 2 ½ð1 1 αÞ 1 ð1 2 αÞ cos θc E0
2
with μ 5 cos θc in the center-of-mass system.
In Eq. (6.9), the time-dependent precursor concentration equation (Section 6.2) has a decay term as well as a flux-
dependent production term due to the delayed fission neutrons; for the jth precursor, the concentration is
ðN ð
dC j ðtÞ 1
5 2 λj C j ðt Þ 1 χ ðE Þ dE dΩ0 β ðE0 ÞνΣf ðE0 Þφðr;Ω0 ;E0 ; tÞ:
0
(6.10)
dt 4π j 0

To look at a simpler form, we neglect delayed neutrons, for which the source term is
ðN ð ðN ð
1
qðr;Ω; E; tÞ 5 dE0 dΩ0 Σs ðr; Ω0 ∙Ω; E0 Þφðr;Ω0 ;E0 ; tÞ 1 χð E Þ dE0 dΩ0 νΣf ðE0 Þφðr;Ω0 ;E0 ; tÞ 1 Sðr;Ω; E; tÞ
0 4π 0
(6.11)
which can be justified since delayed neutrons account for less than one percent of the fission neutrons. For this model,
the transport equation can be written in terms as
1@
φðr;Ω; E; tÞ 1 Ω∙rφðr;Ω; E; tÞ1Σt ðr; EÞφðr;Ω; E; tÞ 5 qðr;Ω; E; tÞ (6.12)
v @t
With the usual boundary conditions of flux and current continuity at an interface and appropriate surface conditions.
^ the quantity N must be constant across an interface from both sides,
Thus, for distance s traveled in the direction sΩ,
i.e.
   
N r 1 sΩ; E; t 1 s=v j2 5 N r 1 sΩ; E; t 1 s=v j1 : (6.13)
Similarly for a non reentrant or free surface denoted by a boundary b, and a free surface condition
 
- ^ ,0
N r ; E; t jb 5 0 for n^ UΩ (6.14)

that is, zero neutrons returning from a boundary.

Consider now the forms of Ω∙r for slab, spherical and cylindrical geometry. In a one-dimensional slab, the distance
traveled s is related to x by x 5 s cos θ, so that
@
Ω∙r 5 μ (6.15)
@x
This is extended to rectangular geometry as
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi @ @

@
Ω∙r 5 1 2 μ2 cosϕ 1 sinϕ 1μ ; (6.16)
@x @y @z
while in spherical geometry Ω∙r 5 μ and

d @ dr @ dμ
Ω∙r 5 5 1 :
ds @r ds @μ ds

With r 5 s cos θ 5 sμ,

dμ dμ dθ sin 2 θ 1 2 μ2
5 5 
ds dθ ds r r
 The neutron transport equation Chapter | 6 263

so that

@ 1 2 μ2 @
Ω∙r 5 μ 1 : (6.17)
@r r @μ

In cylindrical geometry,
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi  
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi @ 1 2 μ2 @ @ @
Ω∙r 5 1 2 μ2 cos χ 1 sin χ 2 1μ (6.18)
@r r @ϕ @χ @z
where the vector r in the Cartesian coordinates has height z azimuthal angle ϕ and χ is the angle between the planes of
Ω and z vectors. The direction of the neutron Ω is μ 5 Ω∙z and azimuthal angle χ. Thus, for an infinite (in height) cyl-
inder with azimuthal symmetry
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi @ 1 2 μ2 @
Ω∙r 5 1 2 μ2 cos χ 2 sin χ :
@r r @χ
Another useful relationship between the flight path and variables xðsÞ; tðsÞ is the change when expressed in terms of
ds i.e.
d @ dt @ dx
5 1
ds @t ds @x ds
which is obtained with the relations (given initial conditions) x 5 x0 1 sμ, and t 5 t0 1 s=v as
d 1@ @
5 1μ : (6.19)
ds v @t @x
With the above forms for the leakage term, the NTE can be explicitly written for regular geometry. Now consider
another simplified form of the NTE known as the one-speed form, for which the steady state one-dimensional angular
flux φðx; μ; EÞ in a slab would reduce the terms of Eq. (6.12) as follows:
@ @
μ φðx; μ; EÞ-μ φðx; μÞ
@x @x
Σt ðx; EÞφðx; μ; EÞ-Σt ðxÞφðx; μÞ
ðN ð1 ð1
1 1
dE0 Σs ðx; E0 Þ dμ0 φðx; μ0 ; E0 Þ- Σs ðxÞ dμ0 φðx; μ0 Þ
2 0 21 2 21

and

ðN ð1 ð1
1 1
χðEÞ dE0 νΣf ðx; E0 Þ dμ0 φðx; μ0 ; E0 Þ- νΣf dμ0 φðx; μ0 Þ:
2 0 21 2 21

The steady-state one-dimensional transport equation is then

@
μ 1 Σt ðxÞ φðx; μÞ 5 qðx; μÞ (6.20)
@x

where
ð ð
1
qðx; μÞ 5 dμ0 Σs ðx; μ0 Þφðx; μ0 Þ 1 dμ0 νΣf φðx; μ0 Þ 1 Sðx; μÞ: (6.21)
2

Notice here that χðEÞ has disappeared as it is understood that neutrons emerging from fission will fall in this energy
group (for which the equation is being simplified). Mathematically what we are saying is that χðEÞ 5 1; there is no
other energy into which an emerging fission neutron can go.
264 Nuclear Engineering

With the above, a one-dimensional (azimuthal symmetry) one-energy NTE in plane geometry is
ð1
@ 1
μ φðx; μÞ 1 Σt ðxÞφðx; μÞ 5 ½Σs 1 νΣf  dμ0 φðx; μ0 Þ 1 Sðx; μÞ: (6.22)
@x 2 21

With the quantity c defined as the mean number of secondary neutrons emerging from a collision by any process,
for example, scattering (c 5 1), fission (c 5 ν ), the (n,2n) reaction for which c 5 2,
Σs 1 ν Σf 1 Σn;2n 1 ?
c5 : (6.23)
Σt
Eq. (6.22) is written as
ð1
@ c
μ φðz; μÞ 1 φðz; μÞ 5 dμ0 φðz; μ0 Þ 1 Sðz; μÞ (6.24)
@z 2 21

where the unit of distance has changed to z 5 Σt x, the “optical distance,” that is, unit of distance measured in terms of
the mean free path λ 5 1=Σt .
In another simplified form of the one-speed transport equation the angular flux is expanded in spherical harmonics
with Legendre polynomials (Section 4.2)
X
N
2l 1 1
φðz; μÞ 5 φl ðzÞPl ðμÞ (6.25)
l50
2
where the moments φl ðxÞ are defined as
ð1
φ l ð zÞ 5 dμφðz; μÞPl ðμÞ: (6.26)
21

Expanding to N 5 1, Eq. (6.25) gives the angular flux in terms of the zeroth and first moments:
1
φðz; μÞ 5 φ ðzÞ 1 3μφ1 ðzÞ (6.27)
2 0
defined as
ð1 ð1
φ0 5 φðz; μÞdμ; φ1 5 μφðz; μÞdμ: (6.28)
-1 -1
These equations, for a plane isotropic source δðzÞ, are
dφ1 ðxÞ
1 ð1 2 cÞφ0 ðzÞ 5 δðzÞ (6.29)
dx
and
dφ0 ðzÞ
1 3φ1 ðzÞ 5 0: (6.30)
dz
Eq. (6.30) is recognized as Fick’s law and Eq. (6.29) is written as
dJ
1 Σa φ 5 S (6.31)
dx
or, in the form of the NDE (Section 4.1)
d2 φ
D 2 Σa φ 1 S 5 0: (6.32)
dx2
We can also see an equivalence for the diffusion length (with D 5 1=3) in units of optical thickness,
1
L 5 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi :
3ð1 2 cÞ
 The neutron transport equation Chapter | 6 265

For a nonmultiplying medium, this gives

rffiffiffiffiffiffi
1 D
L 5 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 5 :
3Σa =Σt Σa

Thus, for fission

ð
N ð 4π
1
dEχðEÞ dΩ 51
0 4π
0

and to consider, for a moment, the steady-state equation, we have

ðN ð
Ω∙rφð~
r; Ω; EÞ 1 Σt ð~
r; E; tÞφð~
r; Ω; EÞ 5 dE dΩ0 Σs ð~
0
r; E0 ; Ω0 ∙ΩÞφð~
r; Ω0 ; E0 Þ
0
ðN ð
1
1 χð E Þ dE0 dΩ0 νΣf ðE0 Þφð~
r; Ω0 ; E0 Þ 1 Sð~
r; Ω; EÞ (6.33)
4π 0

The NTE in the above form is an integro-differential equation; it is linear in the sense that superposition applies and
Green’s functions can be used. This means that if the flux is φ1 for a source S1 and φ2 for a source S2 ; then it must be
φ1 1 φ2 for S1 1 S2 ; which in turn means that a Green’s function Gðr0 ; Ω0 ; E0 -r;Ω;E) obtained for the steady-state
transport equation with a point isotropic source δðr 2 r0 ÞδðΩ 2 Ω0 ÞδðE 2 E0 Þ would give the flux for any source from
 ð ð ð
φ r; Ω; EÞ 5 dV 0 dΩ0 dE0 Sðr0 ; Ω0 ; E0 Þ G ðr0 ; Ω0 ; E0 -r; Ω; EÞ: (6.34)

6.1.2 The two-group transport equation

The two-group form of the NTE, similar to the two-group NDEs described in Section 5.3 are obtained from Eq. (6.9).
The neutrons are divided into two energy groups; for group-1 and group-2 energies E are in the range E1 # E # E0
and E2 # E # E1 where E2 5 0 with the assumption that there is no up-scattering. The group-averaged neutron flux is
φg and the group-averaged neutron current is J g defined as
ð ð E0 ð ð E1
φ1 ðrÞ 5 dΩ φðr; Ω;EÞdE; φ2 ðrÞ 5 dΩ φðr; Ω;EÞdE; (6.35)
E1 0

and
ð ð E0 ð ð E1
J 1 ð rÞ 5 dΩ Ωφðr;Ω; EÞdE; J 2 ðrÞ 5 dΩ Ωφðr;Ω; EÞdE: (6.36)
E1 0
 
The fission contribution in group 1 is χ1 νΣf 1 φ1 ðrÞ 1 νΣf 2 φ2 ðrÞ . Assuming an isotropic source S0;1 ðrÞ in group 1
defined as

ð ð E0
S0;1 ðrÞ 5 dΩ Sðr;Ω; EÞdE
E1

the balance equation is

 
r∙J 1 ðrÞ1Σt0;1 ðrÞφ1 ðrÞ 5 χ1 νΣf 1 φ1 ðrÞ 1 νΣf 2 φ2 ðrÞ 1 S0;1 ðrÞ (6.37)

with the flux-current relation Fick’s law as follows:

rφ1 ðrÞ 1 3Σt;1 ðrÞJ 1 ðrÞ 5 0: (6.38)

266 Nuclear Engineering

In the above, the group-averaged total and fission macroscopic cross-sections are defined as
Ð
Σðr; EÞ φ ðr;EÞ dE
Σ g ð rÞ 5 Ð : (6.39)
φ ðr;EÞ dE
Similarly for Group 2, the balance equation
 
r∙J 2 ðrÞ1Σt0;2 ðrÞφ2 ðrÞ 5 Σs;1-2 φ1 ðrÞ 1 χ2 νΣf 1 φ1 ðrÞ 1 νΣf 2 φ2 ðrÞ (6.40)
has an in-scattering contribution from group 1 neutrons, The scattering cross-section is also defined as a flux-averaged
quantity
Ð 0 Ð
dE φðr;E0 Þ Σðr; E0 -EÞdE
Σg0 -g 5 Ð (6.41)
φðr;E0 ÞdE0
where it is assumed that scattering is isotropic. The Fick’s law statement for Group 2 is then
rφ2 ðrÞ 1 3Σt;2 ðrÞJ 2 ðrÞ 5 0: (6.42)
We will see this form of the equation obtained in the spherical harmonics form of the multigroup transport equation
used for obtained numerical solutions for the angular flux.

6.1.3 The integral form of the transport equation

There are two ways possible to readily convert the integro-differential form; one is the MOC, using the derivative terms
of Eq. (6.19) in which a PDE is converted into an ODE, and the other is an integration of the PDE.
with the derivative terms of the integro-differential transport equation in Cartesian geometry
 
^ ðr;Ω;E; tÞ  1 @ 1 Ωx @ 1 Ωy @ 1 Ωz @ φðr;Ω;E; tÞ
Lφ (6.43)
v @t @x @y @z
which can be expressed as a total derivative
dφ @φ dt @φ dx @φ dy @φ dz
5 1 1 1
ds @t ds @x ds @y ds @z ds
from which the “characteristics” are identified as t 5 t0 1 s=v, and r 5 r0 1 sΩ. In three-dimensional Cartesian
geometry,
x 5 s sin θ cos ϕ 5 sΩx
y 5 s sin θ sin ϕ 5 sΩy
z 5 s cos θ 5 sΩz
so that

dt 1 dx dy dz
5 ; 5 Ωx ; 5 Ωy ; 5 Ωz
ds v ds ds ds

then

^ ðr;Ω;E; tÞ 5 dφ :
Lφ
ds

The transport equation can thus be expressed as a first-order partial differential equation

d    
1 Σt ðr0 1 sΩ; EÞ φ r0 1 sΩ; Ω; E; t0 1 s=v 5 q r0 1 sΩ; Ω; E; t0 1 s=v (6.44)
ds
 The neutron transport equation Chapter | 6 267

and integrated to yield the integral form of the transport equation (Bell & Glasstone, 1979)
ð N Ð s0
2 Σ ðr2s0 0 Ω;EÞds0 0  
φðr; Ω; E; tÞ 5 e 0 t q r 2 s0 Ω; Ω;E; t 2 s0 =v0 ds0 : (6.45)
0

If we put the explicit form for qð~

r; Ω; E; tÞ in the above, we get the integral equation expressed in the form of a scat-
tered contribution and a direct contribution from the extraneous source. Thus, with
ðN ð ðN ð
1
qðr; Ω; E; tÞ 5 dE0 dΩ0 Σs ðr; E0 ; Ω0 ∙ΩÞφðr; Ω0 ; E0 Þ 1 χð E Þ dE0 dΩ0 νΣf ðE0 Þφðr; Ω0 ; E0 Þ 1 Sðr; Ω0 ; E0 Þ
0 4π 0
(6.46)
we can write the terms iteratively as φ0 5 S0 , φ1 5 Kφ0 ,. . .,φn11 5 Kφn . The interpretation of these terms is clear:φ0
represents the “uncollided” flux, that is, the flux of source neutrons, multiplied by the attenuation factor, φ1 is the flux
of neutrons that have had one collision, and φn is the flux of neutrons that have had n collisions. The Neumann series
PN
φn is a solution to the integral equation if it converges.
n50
In this section, the integral formulation of the NTE for collision density ψðr;Ω; E; tÞ 5 Σt ðr; EÞ φ ðr;Ω; E; tÞ is
expressed as
ð
ψ ðPÞ 5 SðPÞ 1 K ðP0 -PÞψ ðP0 Þ dP0 (6.47)

where ψ ðPÞ refers to the collision density in phase space P. This form is obtained by integrating the differential form of
the transport equation described in the previous section. It is also “obvious” intuitively by considering that the collision
density in some phase space P can come from either a direct source into P or by a collision density at some other phase
space P0 that makes a transition into P. This integral form is amenable to computational methods and is “naturally”
suited for analog simulation as will be explained in Chapter 7 (The Monte Carlo Method).
We will later see that this approach is followed in the Monte Carlo simulation of neutrons as they are transported
from the source in a successive series of events. Thus we can write (Rief, 1984)

N ð
X ð
ψ ðPÞ 5 ? K ðP; un Þ K ðun ; un21 Þ? K ðu2 ; u1 Þ S0 ðu1 ; u0 Þ dun . . . du0 (6.48)
n51

where the collision density is the sum of collision densities from the direct source, those that have had one collision,
two collisions and so on, up to n collisions
ψ ðPÞ 5 ψ0 1ψ1 1 ψ2 1 ?ψn :
The direct contribution from the source is the integral over all source neutrons that contribute to the phase space of
interest P
ð
ψ0 5 SðP; u0 Þdu0 ;

The once-collided contribution is the source neutrons born in phase space u0 that have a collision that takes them to
phase space u1 and then to phase space P
ð ð
ψ1 5 K ðP; u1 Þdu1 Sðu1 ; u0 Þdu0 ;

Thus, a source particle born at u0 ðr0 ; Ω0 ; E0 ; t0 Þ moves to ðr1 ; Ω0 ; E0 ; t1 Þ where it has its first collision that changes
its energy and angle into phase space ðr1 ; Ω1 ; E1 ; t1 Þ; we thus express the change as a kernel
Kðr0 ; Ω0 ; E0 ; t0 -r1 ; Ω1 ; E1 ; t1 Þ in terms of a transport process Tðr0; t0 -r1 ; t1 ;Ω0 ; E0 Þ followed by a collision process
CðΩ0 ; E0 -Ω1 ; E1 ;r1 ; t1 Þ, or simply
K i21 5 C i21 ∙T i21 :
268 Nuclear Engineering

Eq. (6.40) can be written as

N ð
X ð n

ψ ðPÞ 5 ? L C i ∙T i dui S0 du0 (6.49)
n51 i51

with the scattering kernel K, expressed as the product of a collision operator C, in which the precollision energy and
angle will transform to the postcollision parameters, and a transport operator T.
Thus a collision will be represented by the term Kφ  C∙Tφ where the collision and transport operators
C^  Cðr
^ 0 ; E0 -E; Ω^ 0 -ΩÞ
^ T^  Tðr ^ 0 Þ represent the scattering process.
^ 0 -r; E0 ; Ω

6.1.4 Multigroup form of the integral transport equation

The integral form of the transport equation can be “solved” in a continuous way which we can call analog simulation,
or in a discrete way, by dividing the energy spectrum in a finite number of energy groups as was shown for diffusion
equation in Chapter 3. In a multigroup form, where the continuous energy spectrum is discretized into a finite number
of groups, as done in the MORSE code (Emmett, 2000), the integral form of the transport equation is obtained from the
multigroup form of Eq. (6.12):
1 @
1Ω∙r 1Σðt gÞ ðrÞ φðgÞ ðr; Ω; tÞ 5 qðgÞ ðr; Ω; tÞ (6.50)
vg @t
to yield the collision density integral form
  X1 ð    
ψðgÞ ~ ^ t 5
r; Ω;
0
^ 0 -ΩÞψ
dΩ0 Σs g -g ðr; Ω ^ ðg0 Þ ~ ^ t 1SðgÞ ~
r; Ω; ^ t
r; Ω; (6.51)
g0 5g

and with the collision and transport operators:

  0   0   
ψðgÞ ~ ^ t 5 C^ g -g ~
r; Ω; r;Ω ^ T^ g r 0 -r;Ω
^ 0 -Ω ^ 0 ψðg0 Þ r 0 ; Ω
^ 0 ; t 1 SðgÞ (6.52)

where the collision operator C^ is

  X 1 ð 0
^ 0 -Ω
^
g0 -g
^ 5
^ 0 -Ω Σs g -g ðr; Ω
C^ ~
r;Ω dΩ0 0 (6.53)
g0 5g
Σt ðrÞ
g

and the transport operator T^ is defined as

g0
  ðN
^0 5
T^ r 0 -r;Ω
ð gÞ
dR Σðt gÞ ðr Þ e2z ðr;r2RΩÞ : (6.54)
0

In the integral form, the analog simulation this begins from a source neutron being transported to a collision site,
where the collision process results in “postcollision” energy and angle of travel. The collision density is tallied and the
simulation continues until a “history” ends due to an escape from the system or some other termination criteria as will
be shown in the next chapter.

6.2 Exact solutions of the transport equation

The complexity of the transport equation makes it impossible to obtain exact solutions for all but idealized configura-
tions such as infinite medium and regular geometry. A number of exact solutions are given in Bell and Glasstone
(1979), and the following benchmarks by Ganapol (Ganapol, 2008) are available for comparing with numerical solu-
tions: (1) Infinite medium slowing down, (2) Slowing down/Laplace transform solution in the BL approximation, (3)
Slowing down/the multigroup solution in the BL approximation, (4) Slowing down and thermalization in an infinite
medium/the embedded multigroup approximation, (5) Monoenergetic transport in an infinite medium/the Fourier trans-
form solution, (6) Monoenergetic transport in a semiinfinite medium/the Laplace transform solution, (7) Monoenergetic
transport in a 1-D slab/the Fn solution, (8) Monoenergetic transport in a 1-D cylinder/the Fn solution, (9) Multigroup
transport in infinite media/the Fourier transform solution, (10) Multigroup slab transport/the Green’s function method,
 The neutron transport equation Chapter | 6 269

(11) Monoenergetic transport in a two-dimensional semiinfinite medium/the searchlight problem (SLP), and (12)
Multgroup transport in a three-dimensional infinite medium/the point kernel method.
For the source-free infinite medium case, analytical solutions can be obtained, assuming separability
φðz; μÞ 5 χðzÞψðμÞ, in terms of eigenfunctions ψν ðμÞ:
φν ðz; μÞ 5 e2z=μ ψν ðμÞ: (6.55)
It can then be shown (Bell & Glasstone, 1979) that for all values of μAð2 1; 1Þ for ν not both real and in the interval
(21,1), that
c ν
ψν ðμÞ 5 (6.56)
2ν 2μ
where the eigenvalues satisfy the transcendental equation
1 cν 0 ν 0 1 1
1 5 cν 0 tan h21 5 ln : (6.57)
ν0 2 ν0 2 1
The above uses the relation
1 11x
tan h21 x 5 ln :
2 12x
When c , 1 the roots are real (Section 2.10), as shown in Fig. 6.2, while for c . 1, they are imaginary.
The value of ν o determines the rate of decrease of the asymptotic flux in a source-free infinite medium and is there-
fore called the asymptotic relaxation length.
cν 0
φo6 ðx; μÞ 5 e7x=υ0 :
2ðν 0 7μÞ
When c is near unity,
1 1 2
5 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 1 ð1 2 cÞ 1 ?
ν0 2
3ð1 2 cÞ 5

For weakly absorbing media, this relaxation length is the same as the diffusion length L of diffusion theory
1 1
L 5 pffiffiffiffiffiffiffiffiffiffiffiffiffi 5 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi :
3Σt Σa 3Σt ð1 2 cÞ
Comparisons between diffusion and transport theory, and Monte Carlo simulation yet to be covered, will be demon-
strated for some simple cases.

FIGURE 6.2 Relaxation length in m.f.p.’s for isotropic scattering.

270 Nuclear Engineering

It was shown above that for a source-free infinite medium for the asymptotic flux, for real values of ν o is of the
form
cν 0
φo6 ðx; μÞ 5 e7x=υ0
2ðν 0 7μÞ
For the monoenergetic transport solution in a semiinfinite medium, for example, Ganapol (Ganapol, 2008) has
obtained a “classical solution” for the scalar flux φðxÞ in terms of the inverse Laplace transform
2 φð0; 2 1=sÞ
φðxÞ 5 L21 (6.58)
c x s
as
8c μ    
>
>
0
H μ0 H 1=s ; Beam source
>
< 2 1 1 sμ0
2 pffiffiffiffiffiffiffiffiffiffiffi  
φðxÞ 5 L21 1 2 1 2 cH 1=s (6.59)
c x >>
>
: ; Isotropic source
s

Chandrasekhar’s H function satisfies the integral equation

ð1
ψðμ0 Þ
H ðμÞ 5 1 1 μH ðμÞ H ðμ0 Þdμ0
0 μ1μ
0

Ð1
where 0 ψðμÞdμ 5 1. In case of isotropic scattering, ψðμÞ 5 c=2. The equation is solved iteratively (Matlab Program
6.1) with
" #21
c XN
H km0
Hm k11=2
5 12 μm ω m0
2 m0 51 μm 1μm0
α0
k11
Hm 5 H k11=2
k11=2 m
α0
k11=2 P Ð1 pffiffiffiffiffiffiffiffiffiffiffi
with α0  Nm0 51 ωm0 H m k11=2
0 , and a normalizationα0  0 dμH ðμÞ 5 2c 12 12c .
This will be referred to in the following sections, especially for exact solution for finite media configurations.

6.2.1 The classic albedo problem

The classic albedo problem, of determining reflectivity of a surface, was addressed originally in radiative transfer
(Chandrasekhar, 1960) and exact solutions have been obtained (Ganapol, 2008) using the Laplace transform. For emerg-
ing radiation, μ , 0; x0 -N, the solution is
ð
  c N 0 2x2xμ 0
φ x; 2 μ 5 dx e φðx0 Þ (6.60)
2jμj x
which, at the surface x 5 0, reduces to
ðN
  c x0
φ 0; 2 μ 5 dx0 e2 μ φðx0 Þ: (6.61)
2jμj 0

The exact solution, for an isotropic medium with a point isotropic source at the surface separating the half-space
from the vacuum, in terms of Chandrasekhar’s H functions is
pffiffiffiffiffiffiffiffiffiffiffi
φð0; 2 μÞ 5 1 2 1 2 cH ðμÞ: (6.62)
Fig. 6.3 shows the albedo for monoenergetic neutrons incident on a surface x 5 0 as a function of c.
 The neutron transport equation Chapter | 6 271

PROGRAM 6.1 Chandrasekhar’s H function.

% Program C:\Users\User1\Documents\MATLAB\TrTheory2012\Hfunction
% give a range of values of c
% prog calculates H(mu) vs mu and plots
% open output file
resl=fopen('outTT.txt','w');

N=32;
if (N==32)
t(1)=0.0483076656877383162348126; w(1)=0.0965400885147278005667648;
t(2)=0.1444719615827964934851864; w(2)=0.0956387200792748594190820;
t(3)=0.2392873622521370745446032; w(3)=0.0938443990808045656391802;
t(4)=0.3318686022821276497799168; w(4)=0.0911738786957638847128686;
t(5)=0.4213512761306353453641194; w(5)=0.0876520930044038111427715;
t(6)=0.5068999089322293900237475; w(6)=0.0833119242269467552221991;
t(7)=0.5877157572407623290407455; w(7)=0.0781938957870703064717409;
t(8)=0.6630442669302152009751152; w(8)=0.0723457941088485062253994;
t(9)=0.7321821187402896803874267; w(9)=0.0658222227763618468376501;
t(10)=0.7944837959679424069630973; w(10)=0.0586840934785355471452836;
t(11)=0.8493676137325699701336930; w(11)=0.0509980592623761761961632;
t(12)=0.8963211557660521239653072; w(12)=0.0428358980222266806568786;
t(13)=0.9349060759377396891709191; w(13)=0.0342738629130214331026877;
t(14)=0.9647622555875064307738119; w(14)=0.0253920653092620594557526;
t(15)=0.9856115115452683354001750; w(15)=0.0162743947309056706051706;
t(16)=0.9972638618494815635449811; w(16)=0.0070186100094700966004071;
j=16;
for jj=1:16
tt(jj)=-t(j); ww(jj)=w(j);
j=j-1;
end
k=1;
for kk=17:32
tt(kk)=t(k); ww(kk)=w(k);
k=k+1;
end
end

%%% Begin
Ncases=6; Niters=50;
cc=[0.1;0.2;0.4;0.6;0.8;0.9];

for icase=1:Ncases
c=cc(icase);
fprintf (resl,'\n c = %6.4f ',c);
Alpha0=(2.0/c)*(1.0-sqrt(1.0-c));
fprintf ('\n Alpha0 = %6.4f ',Alpha0);

for ih1 = 1:N

mu(ih1)=0.5*(tt(ih1)+1.0);
Hf(ih1) = 1.0;
end

(Continued )
272 Nuclear Engineering

PROGRAM 6.1 (Continued)

for ih6 = 1:Niters
for ih2 = 1:N
sum =0.0;
for ih3 = 1:N
sum = sum + ww(ih3)*( Hf(ih3)/(mu(ih2)+ mu(ih3)));
end
sum = 0.5*sum;
Hf(ih2) = 1.0/( 1.0 - (c/2.0)*mu(ih2)*sum );
end

Alpha0k=0.0;
for ih4=1:N
Alpha0k = Alpha0k + ww(ih4)*Hf(ih4);
end
Alpha0k=0.5*Alpha0k;

Error = Alpha0-Alpha0k;

for ih5=1:N
Hf(ih5)=(Alpha0/Alpha0k)*Hf(ih5);
end

end
%keep the H values for this case to be plotted later
for ih7=1:N
HH(icase,ih7)=Hf(ih7);
end

end
for muVal=1:N
H1(muVal)=HH(1,muVal);
H2(muVal)=HH(2,muVal);
H3(muVal)=HH(3,muVal);
H4(muVal)=HH(4,muVal);
H5(muVal)=HH(5,muVal);
H6(muVal)=HH(6,muVal);
end
figure(1)
plot(mu,H1,':k','LineWidth',1.5)
hold on
plot(mu,H2,':k','LineWidth',1.5)
hold on
plot(mu,H3,':k','LineWidth',1.5)
hold on
plot(mu,H4,':k','LineWidth',1.5)
hold on
plot(mu,H5,':k','LineWidth',1.5)
hold on
plot(mu,H6,'-k','LineWidth',1.5)
xlabel('\mu','FontSize',12)
ylabel('H(\mu)','FontSize',12)
text(0.4,1.7,'c','FontSize',12)

(Continued )
 The neutron transport equation Chapter | 6 273

PROGRAM 6.1 (Continued)

h = legend('c = 0.1','c = 0.2','c = 0.4','c = 0.6','c = 0.8','c = 0.9',2);
grid on

figure(2)
% plot albedo here
for muVal=1:N
H1(muVal)=(1-sqrt(1-cc(1)))*HH(1,muVal);
H2(muVal)=(1-sqrt(1-cc(2)))*HH(2,muVal);
H3(muVal)=(1-sqrt(1-cc(3)))*HH(3,muVal);
H4(muVal)=(1-sqrt(1-cc(4)))*HH(4,muVal);
H5(muVal)=(1-sqrt(1-cc(5)))*HH(5,muVal);
H6(muVal)=(1-sqrt(1-cc(6)))*HH(6,muVal);

end
plot(mu,H1,':k','LineWidth',1.5)
hold on
plot(mu,H2,':k','LineWidth',1.5)
hold on
plot(mu,H3,':k','LineWidth',1.5)
hold on
plot(mu,H4,':k','LineWidth',1.5)
hold on
plot(mu,H5,':k','LineWidth',1.5)
hold on
plot(mu,H6,'-k','LineWidth',1.5)
xlabel('\mu','FontSize',12)
ylabel('H(\mu)','FontSize',12)
text(0.4,1.7,'c','FontSize',12)
h = legend('c = 0.1','c = 0.2','c = 0.4','c = 0.6','c = 0.8','c = 0.9',2);
grid on

fclose(resl);

6.2.2 Infinite medium with a plane isotropic source

The steady-state transport equation exact solution for a plane isotropic source located at x0 has been obtained by Case’s
method and by the Fourier Transform method (Bell & Glasstone, 1979). The angular flux, for x0 5 0, is given by
" x2x ð1 #
2ν0 x2x0
1 ψ1 0 ðμÞe
0 ψν ðμÞe2 ν
φðx; μÞ 5 1 dν (6.63)
4π N10 0 Nν

for x . 0, where
c ν0
ψ1
0 5
2 ν0 2 μ
c 3 c 1
N1
0 5 ν0 2 2 2
2 ν0 2 1 ν0
c ν
ψν ðμÞ 5 P 1 λðν Þδðμ 2 ν Þ
2 ν2μ
274 Nuclear Engineering

FIGURE 6.3 Albedo for monoenergetic neutrons for

isotropic medium.

π2 c2 2
N ν 5 ν λ2 ðν Þ 1 ν
4
λðν Þ 5 1 2 cνtanh21 ν:
The angular fluxes, calculated from Matlab Program 6.2, are shown in Figs. 6.4
6.6 for the cases c 5 0:2; 0:4; 0:8,
that is, increasing scattering in the medium. In all cases, the “forward” flux ðμ . 0Þ is larger in magnitude than the
“backward” flux ðμ , 0Þ while the magnitude of the angular flux increases as the material becomes less absorbing
(higher c).
Integrating Eq. (6.63) gives the total flux
" x2x0 ð 1 2x2x #
2 0
1 e ν0 e ν0
φðxÞ 5 1 dν : (6.64)
2 N1 0 0 Nν

The flux given in Eq. (6.64) is the sum of an asymptotic φas and a transient part φtr with the asymptotic part repre-
senting the collision equilibrium component. The ratio of the asymptotic flux to the total flux φ 5 φas 1 φtr is shown in
Fig. 6.7 where a high-scattering medium (high c) achieves equilibrium faster than low scattering media. A pure
absorber (c 5 0) would thus never achieve “collisional” equilibrium. The calculation for Fig. 6.7 is given in Matlab
Program 6.3.

6.2.3 Finite sphere with a point isotropic source

For a sphere with a point isotropic source, an exact solution has been obtained by Erdmann and Siewert (1968), and by
(Siewert & Grandjean, 1979) as
  ð1 r 
1 r
φðr Þ 5 φN ðr Þ 2 Eðν 0 Þe2R=ν 0 sin h 1 Eðν Þe2R=ν 0 sin h dν (6.65)
4πr ν0 0 ν
where
  ð1  r
1 1 r 1
φN 5 exp 2 1 exp 2 dν (6.66)
4πr ν 0 Nðν 0 Þ ν0 0 νNðνÞ ν
 The neutron transport equation Chapter | 6 275

PROGRAM 6.2 Angular flux in an infinite medium with a planar isotropic source.
% Program C:\Users\User1\Documents\MATLAB\TrTheory2012\AngFluxInfMed

% open output file

resl=fopen('outTT.txt','w');

N=32;

if (N==2)
t(1)=0.5773502691896257645091488;
t(2)=-t(1);
w(1)=1.0;
w(2)=1.0;
end

if (N==4)
t(1)=0.3399810435848562648026658;
t(2)=-t(1);
w(1)=0.6521451548625461426269361;
w(2)=w(1);
t(3)=0.8611363115940525752239465;
t(4)=t(3);
w(3)=0.3478548451374538573730639;
w(4)=w(3);
end

if (N==8)
t(1)=0.1834346424956498049394761; t(2)=-t(1);
t(3)=0.5255324099163289858177390; t(4)=-t(3);
t(5)=0.7966664774136267395915539; t(6)=-t(5);
t(7)=0.9602898564975362316835609; t(8)=-t(7);

w(1)=0.3626837833783619829651504; w(2)=w(1);
w(3)=0.3137066458778872873379622; w(4)=w(3);
w(5)=0.2223810344533744705443560; w(6)=w(5);
w(7)=0.1012285362903762591525314; w(8)=w(7);
end

if(N==16)
t(1)=0.0950125098376374401853193; w(1)=0.1894506104550684962853967;
t(2)=-t(1); w(2)=w(1);

t(3)=0.2816035507792589132304605; w(3)= 0.1826034150449235888667637;

t(4)=-t(3); w(4)=w(3);

t(5)=0.4580167776572273863424194; w(5)= 0.1691565193950025381893121;

t(6)=-t(5); w(6)=w(5);

t(7)=0.6178762444026437484466718; w(7)= 0.1495959888165767320815017;

t(8)=-t(7); w(8)=w(7);

t(9)=0.7554044083550030338951012; w(9)= 0.1246289712555338720524763;

t(10)=-t(9); w(10)=w(9);

(Continued )
276 Nuclear Engineering

PROGRAM 6.2 (Continued)

t(11)=0.8656312023878317438804679; w(11)= 0.0951585116824927848099251;
t(12)=-t(11); w(12)=w(11);
t(13)=0.9445750230732325760779884; w(13)= 0.0622535239386478928628438;
t(14)=-t(13); w(14)=w(13);

t(15)=0.9894009349916499325961542; w(15)= 0.027152459411754094;

t(16)=-t(15); w(16)=w(15);
end

if (N==32)
t(1)=0.0483076656877383162348126; w(1)=0.0965400885147278005667648;
t(3)=0.1444719615827964934851864; w(3)=0.0956387200792748594190820;
t(5)=0.2392873622521370745446032; w(5)=0.0938443990808045656391802;
t(7)=0.3318686022821276497799168; w(7)=0.0911738786957638847128686;
t(9)=0.4213512761306353453641194; w(9)=0.0876520930044038111427715;
t(11)=0.5068999089322293900237475; w(11)=0.0833119242269467552221991;
t(13)=0.5877157572407623290407455; w(13)=0.0781938957870703064717409;
t(15)=0.6630442669302152009751152; w(15)=0.0723457941088485062253994;
t(17)=0.7321821187402896803874267; w(17)=0.0658222227763618468376501;
t(19)=0.7944837959679424069630973; w(19)=0.0586840934785355471452836;
t(21)=0.8493676137325699701336930; w(21)=0.0509980592623761761961632;
t(23)=0.8963211557660521239653072; w(23)=0.0428358980222266806568786;
t(25)=0.9349060759377396891709191; w(25)=0.0342738629130214331026877;
t(27)=0.9647622555875064307738119; w(27)=0.0253920653092620594557526;
t(29)=0.9856115115452683354001750; w(29)=0.0162743947309056706051706;
t(31)=0.9972638618494815635449811; w(31)=0.0070186100094700966004071;
for jj=2:2:32
t(jj)=-t(jj-1); w(jj)=w(jj-1);
end
end

%%% Begin
c=0.8;
fprintf (resl,'\n c = %6.4f ',c);
% step 1 calc Nu0
TNu1=0.5; TNu2=10.0; NPTS=50000; del=(TNu2-TNu1)/NPTS;
TNu = TNu1;
for ip=1:NPTS
rt(ip)=TNu;
FN(ip) = 1.0-c*(TNu/2)*log((TNu+1)/(TNu-1));
a(ip)=sign(FN(ip));
TNu=TNu+del;
end
% find the root where the sign changes
root = 1.0;
for ir = 1:NPTS-1
if (a(ir)~=a(ir+1))
root = rt(ir);
end
end
fprintf (resl,'\n nu0 = %6.4f ',root);

(Continued )
 The neutron transport equation Chapter | 6 277

PROGRAM 6.2 (Continued)

% step2 calc N0+
N0Plus = 0.5*c*root^3*( (c/(root^2-1.0)) - (1/root^2));
fprintf (resl,'\n N0Plus = %12.4e ',N0Plus);

%xLo = 1.0; xHi = 10.0; DelX=1.0; Nx=10;

xLo = 0.1; xHi = 10.0; Nx=100; DelX=(xHi-xLo)/Nx;
muLo=-1.0; muHi=1.0; DelMu=0.4; Nangles=6;
keyX=0; iX=0; countX=0;
for iX=1:Nx
x=xLo+(iX-1)*DelX;

xx(iX)=x;

for iM=1:Nangles
mu=muLo+(iM-1)*DelMu;

mm(iM)=mu;

% step 3 calc psi0+

psi0Plus = 0.5*c * root/(root-mu);
phi1 = (psi0Plus*exp(-x/root))/N0Plus; % first term of angular flux

% step 4 calc lambda and N

sum=0.0;
for i=1:N
nu=0.5*(1+t(i)); ww=w(i);
lambda=1-c*nu*atanh(nu);
Nnu = nu*( lambda^2 + (1.0/4.0)*c^2*nu^2*3.14159^2 );
psiNu=0.5*c*(nu/(nu-mu));
value=(psiNu*exp(-x/nu))/Nnu;
sum = sum + ww*value;
end
beta=0.5*sum;

gamma=0.0;
if (mu>0)
% step 5 calc 2nd term part 2(gamma)
lambdaMu =1-c*mu*atanh(mu);
NnuMu = mu *( lambdaMu^2 + (1.0/4.0)*c^2*mu^2*3.14159 );
gamma = (lambdaMu/NnuMu)*exp(-x/mu);
end

phi2=beta+gamma;% second term

phi(iX,iM)=(1.0/(4.0*pi))*(phi1+phi2);
end
end

% print the results

% there are countX values of x and countM values of mu
fprintf(resl,'\n x values are\n');
for ix = 1:Nx
fprintf(resl,'\n x(%2.0f)=%6.4f',ix,xx(ix));

(Continued )
278 Nuclear Engineering

PROGRAM 6.2 (Continued)

end
fprintf(resl,'\n mu values are\n');
for im = 1:Nangles
fprintf(resl,'\n mm(%2.0f)=%6.4f',im,mm(im));
end
fprintf (resl,'\n flux values are');
for ix = 1:Nx
fprintf(resl,'\n flux at x = %6.4f ',xx(ix));
for im = 1:Nangles
fprintf(resl,'\n phi(%2.0f %2.0f)=%12.4e',ix,im,phi(ix,im));
end
end
% now collect, for a single angle, all the fluxes as a function of x
for ix1=1:Nx
phiMu1(ix1)=phi(ix1,1);
phiMu2(ix1)=phi(ix1,2);
phiMu3(ix1)=phi(ix1,3);
phiMu4(ix1)=phi(ix1,4);
phiMu5(ix1)=phi(ix1,5);
phiMu6(ix1)=phi(ix1,6);
end

figure(1)
set(gca,'FontSize',12)
semilogy(xx,phiMu1,'-k','LineWidth',2)
hold on
%semilogy(xx,phiMu2,'-.k','LineWidth',1.5)
%hold on
semilogy(xx,phiMu3,':k','LineWidth',1.0)
hold on
semilogy(xx,phiMu4,':k','LineWidth',1.0)
hold on
%semilogy(xx,phiMu5,'-k','LineWidth',1.5)
%hold on
semilogy(xx,phiMu5,'-k','LineWidth',2)
hold on
xlabel('\bf x (mfp)','fontsize',14)
ylabel('\bf Angular Flux (cm^{-2} s^{-1} steradian^{-1})')
%h = legend('\mu=-1.0','\mu = -0.6','\mu = -0.2','\mu = 0.2','\mu =
0.6','\mu=1',2);
%h = legend('\mu=-1.0','\mu = -0.2','\mu = 0.2','\mu=1',2);
grid off
xlim([1 10])
ylim([1e-5 2e-1])
% etime(clock,t0)
% t=cputime;
% fprintf(resl,'\n\n CPU Time is %9.2f \n\n',t)
text(8.5,1e-1,'\bf c=0.8','fontsize',14)
text(1.3,1e-2,'\mu= -1')
text(1.1,3e-2,' -0.2')
text(1.8,3.5e-2,'0.2')
text(2,1e-1,'\mu=+1')
fclose(resl);
 The neutron transport equation Chapter | 6 279

FIGURE 6.4 Angular flux for c 5 0.2, infinite medium.

FIGURE 6.5 Angular flux c 5 0.4, infinite medium.

FIGURE 6.6 Angular flux c 5 0.8, infinite medium.

280 Nuclear Engineering

FIGURE 6.7 Total flux, infinite medium isotropic source.

is the infinite medium solution. Ð1

The expansion coefficient EðνÐ0 Þ is given by Eðν 0 Þ 5 Gðν 0 Þ 1 0 K ðxÞEðxÞdx where EðνÞ satisfies the Fredholm equa-
1
tion (Section 4.3) Eðν Þ 5 Gðν Þ 1 0 K ðx
!ν ÞEðxÞdx. The four known functions G and K are:
(  )   ð1  
R1z
22 ν 0 2 R12z
2 ν 0 cν 0 exp 2 Rν dν
12e 0
G ðν 0 Þ 5 e 0
1 (6.67)
ν 0 Nðν 0 Þ H ðν 0 ÞNðν 0 Þ 0 H ðν ÞNðνÞ ν 1 ν 0
 
2ð R 1 z 0 Þ cxν 0 
1 2 exp 2 K ðxÞ 5 exp 2 2R=xÞ (6.68)
ν0 2ðν 0 1 xÞH ðν 0 ÞN ðν 0 ÞHðxÞ
ð1
cνν 0   2 1  
H ðν ÞN ðν ÞGðν Þ 5 exp 2 R=ν 0 Gðν 0 Þexpð2R=ν 0 Þ 1 1 cν exp 2 R=x dx
2ðν 1 ν 0 ÞH ðν 0 Þ ν 0 N ðν 0 Þ 0 ðx 1 ν ÞH ðxÞNðxÞ
(6.69)
and

cνν 0 cνx 
H ðν ÞN ðν ÞK ðx
!ν Þ 5 expð2 2R=ν 0 ÞK ðxÞ 1 exp 2 2R=xÞ (6.70)
2ðν 1 ν 0 ÞH ðν 0 Þ 2ðν 1 xÞH ðxÞ

with
c 3 c 1
Nðν 0 Þ  N 1
0 5 ν 2
2 0 ν 20 2 1 ν 20
π2 c2 2
N ν 5 ν λ2 ðν Þ 1 ν
4
λðν Þ 5 1 2 cν tan h-1 ν
and z0 , the extrapolation distance, given by

1 4N ðν 0 ÞH 2 ðν 0 Þ
z0 5 ν 0 ln : (6.71)
2 cν 0
 The neutron transport equation Chapter | 6 281

PROGRAM 6.3 Angular flux in an infinite medium with a planar isotropic source.
% C:\Users\User1\Documents\MATLAB\TrTheory\TotFluxInfMed
% infinite medium exact solution
% give 4 c values;
% get the total flux and plot

resl=fopen('outTT.txt','w');

N=32;
if (N==2)
t(1)=0.5773502691896257645091488;
t(2)=-t(1);
w(1)=1.0;
w(2)=1.0;
end

if (N==4)
t(1)=0.3399810435848562648026658;
t(2)=-t(1);
w(1)=0.6521451548625461426269361;
w(2)=w(1);
t(3)=0.8611363115940525752239465;
t(4)=t(3);
w(3)=0.3478548451374538573730639;
w(4)=w(3);
end

if (N==8)
t(1)=0.1834346424956498049394761; t(2)=-t(1);
t(3)=0.5255324099163289858177390; t(4)=-t(3);
t(5)=0.7966664774136267395915539; t(6)=-t(5);
t(7)=0.9602898564975362316835609; t(8)=-t(7);

w(1)=0.3626837833783619829651504; w(2)=w(1);
w(3)=0.3137066458778872873379622; w(4)=w(3);
w(5)=0.2223810344533744705443560; w(6)=w(5);
w(7)=0.1012285362903762591525314; w(8)=w(7);
end

if(N==16)
t(1)=0.0950125098376374401853193; w(1)=0.1894506104550684962853967;
t(2)=-t(1); w(2)=w(1);

t(3)=0.2816035507792589132304605; w(3)= 0.1826034150449235888667637;

t(4)=-t(3); w(4)=w(3);

t(5)=0.4580167776572273863424194; w(5)= 0.1691565193950025381893121;

t(6)=-t(5); w(6)=w(5);

t(7)=0.6178762444026437484466718; w(7)= 0.1495959888165767320815017;

t(8)=-t(7); w(8)=w(7);

t(9)=0.7554044083550030338951012; w(9)= 0.1246289712555338720524763;

t(10)=-t(9); w(10)=w(9);

(Continued )
282 Nuclear Engineering

PROGRAM 6.3 (Continued)

t(11)=0.8656312023878317438804679; w(11)= 0.0951585116824927848099251;
t(12)=-t(11); w(12)=w(11);
t(13)=0.9445750230732325760779884; w(13)= 0.0622535239386478928628438;
t(14)=-t(13); w(14)=w(13);

t(15)=0.9894009349916499325961542; w(15)= 0.027152459411754094;

t(16)=-t(15); w(16)=w(15);
end

if (N==32)
t(1)=0.0483076656877383162348126; w(1)=0.0965400885147278005667648;
t(3)=0.1444719615827964934851864; w(3)=0.0956387200792748594190820;
t(5)=0.2392873622521370745446032; w(5)=0.0938443990808045656391802;
t(7)=0.3318686022821276497799168; w(7)=0.0911738786957638847128686;
t(9)=0.4213512761306353453641194; w(9)=0.0876520930044038111427715;
t(11)=0.5068999089322293900237475; w(11)=0.0833119242269467552221991;
t(13)=0.5877157572407623290407455; w(13)=0.0781938957870703064717409;
t(15)=0.6630442669302152009751152; w(15)=0.0723457941088485062253994;
t(17)=0.7321821187402896803874267; w(17)=0.0658222227763618468376501;
t(19)=0.7944837959679424069630973; w(19)=0.0586840934785355471452836;
t(21)=0.8493676137325699701336930; w(21)=0.0509980592623761761961632;
t(23)=0.8963211557660521239653072; w(23)=0.0428358980222266806568786;
t(25)=0.9349060759377396891709191; w(25)=0.0342738629130214331026877;
t(27)=0.9647622555875064307738119; w(27)=0.0253920653092620594557526;
t(29)=0.9856115115452683354001750; w(29)=0.0162743947309056706051706;
t(31)=0.9972638618494815635449811; w(31)=0.0070186100094700966004071;
for jj=2:2:32
t(jj)=-t(jj-1); w(jj)=w(jj-1);
end
end

%%% Begin
Ncases=4;
cc=[0.2;0.4;0.6;0.8];

for ic =1:Ncases
c=cc(ic);
fprintf (resl,'\n c = %6.4f ',c);
% step 1 calc Nu0
TNu1=0.5; TNu2=10.0; NPTS=50000; del=(TNu2-TNu1)/NPTS;
TNu = TNu1;
for ip=1:NPTS
rt(ip)=TNu;
FN(ip) = 1.0-c*(TNu/2)*log((TNu+1)/(TNu-1));
a(ip)=sign(FN(ip));
TNu=TNu+del;
end
% find the root where the sign changes
root = 1.0;
for ir = 1:NPTS-1
if (a(ir)~=a(ir+1))
root = rt(ir);
end

(Continued )
 The neutron transport equation Chapter | 6 283

PROGRAM 6.3 (Continued)

end
Nu0(ic)=root;
fprintf (resl,'\n nu0 = %6.4f ',root);

% step2 calc N0+

N0Plus = 0.5*c*root^3*( (c/(root^2-1.0)) - (1/root^2));
fprintf (resl,'\n N0Plus = %12.4e ',N0Plus);
xLo = 0.2; xHi = 20.0; Nx = 100; DelX=(xHi-xLo)/Nx;
Nx=Nx+1;
keyX=0; iX=0; countX=0;
for iX=1:Nx
x=xLo+(iX-1)*DelX;
xx(iX)=x;
% step 3 calc psi0+
phi1 = (exp(-x/root))/N0Plus; % first term
% step 4 calc lambda and N
sum=0.0;
for i=1:N
nu=0.5*(1+t(i)); ww=w(i);
lambda=1-c*nu*atanh(nu);
Nnu = nu*( lambda^2 + (1.0/4.0)*c^2*nu^2*3.14159^2 );
value=(exp(-x/nu))/Nnu;
sum = sum + ww*value;
end
beta=0.5*sum;
gamma=0.0;

phi2=beta+gamma;% second term

phi(iX)=(1.0/(2.0))*(phi1+phi2);
phiAs(iX)=(1.0/2.0)*phi1;
ratio(iX)=phiAs(iX)/phi(iX);
end

% print the results

% now calc the total flux

fprintf(resl,'\n Total Flux Asymp FLux Ratio PhiAs/Phi \n\n');
fprintf(resl,'\n i x Phi(x) PhiAs(x) Ratio ');
for k = 1:Nx
fprintf(resl,'\n %4.0f %8.4f %12.4e %12.4e
%12.4e',k,xx(k),phi(k),phiAs(k),ratio(k));
phiC(ic,k)=phi(k); % for each c value
phiAsC(ic,k)=phiAs(k);
end

% now write all these values for this value of c

for k1 = 1:Nx
ratioC(ic,k1)=ratio(k1);
end

end

(Continued )
284 Nuclear Engineering

PROGRAM 6.3 (Continued)

fprintf(resl,'\n\n c Nu0');
for in=1:Ncases
fprintf(resl,'\n %6.4f %6.4f',cc(in),Nu0(in));
end

iplot=2;

if(iplot==1)
set(gca,'FontSize',12)
semilogy(xx,phi,'-k','LineWidth',2)
hold on
semilogy(xx,phiAs,'-r','LineWidth',2)
xlabel('\bf x (mfp)''fontsize',14)
ylabel('\bf Total Flux cm^{-2} s^{-1}','fontsize',14)
h = legend('\phi','\phi_{as}',2);
grid on
end

if(iplot==2)
set(gca,'FontSize',12)
for ip=1:Nx
ratio1(ip)=ratioC(1,ip); ratio2(ip)=ratioC(2,ip);
ratio3(ip)=ratioC(3,ip); ratio4(ip)=ratioC(4,ip);
end
plot(xx,ratio1,'-k','LineWidth',1.5)
hold on
plot(xx,ratio2,'-k','LineWidth',1.5)
hold on
plot(xx,ratio3,'-k','LineWidth',1.5)
hold on
plot(xx,ratio4,'-k','LineWidth',1.5)
xlabel('\bf x (mfp)','FontSize',14)
ylabel('\bf \phi_{as}/\phi ','FontSize',14)
xp1=5; yp1= 0.05 ;text(xp1,yp1,'c=0.2','FontSize',12);
xp2=5; yp2= 0.64 ;text(xp2,yp2,'c=0.4','FontSize',12);
xp3=5; yp3= 0.88 ;text(xp3,yp3,'c=0.6','FontSize',12);
xp4=3; yp4= 0.95 ;text(xp4,yp4,'c=0.8','FontSize',12);
%h = legend('c=0.2','c=0.4','c=0.6','c=0.8',2);
grid off
ylim([0 1.05])
end
fclose(resl);

The Chandrasekhar H-function satisfies the integral equation

ð1
ψðμ0 Þ
H ðμÞ 5 1 1 μH ðμÞ H ðμ0 Þdμ0 : (6.72)
0 μ 1 μ 0

The above equations are solved with the MATLAB program SpherePtSrcMain giving the total flux shown in
Figs. 6.8
6.10 for various values of c
The values of flux given in Table 6.1 agree well with the results for c 5 0:3 and c 5 0:9 given by (Siewert &
Grandjean, 1979).
Fig. 6.11 shows that transport theory flux for values of c approaching the critical case (c 5 1) from which the cosine
buckling is observable.
 The neutron transport equation Chapter | 6 285

FIGURE 6.8 Transport theory flux in a finite sphere (c 5 0.3).

FIGURE 6.9 Transport theory flux in a finite sphere (c 5 0.9).

6.3 Numerical methods for solving the transport equation

The first idealizations of infinite medium, semiinfinite medium and homogeneous half-spaces are useful for obtaining
exact analytical solutions using Laplace and Fourier transforms, Green’s functions, etc., but found very limited practical
application. This was followed by semianalytical and numerical solutions methods such as SN and PN methods. Later,
finite element (Section 4.5), boundary-element and finite volume methods were incorporated with tremendous numeri-
cal effort to extend the capabilities of deterministic codes to full 3D modeling.

6.3.1 The discrete ordinates method

The discrete ordinates method, first used by Wick and Chandrasekhar for radiative transfer analysis in stellar atmo-
spheres (Bell & Glasstone, 1952), was developed for neutron transport by Carlson and Lathrop (1968). In the discrete
ordinates, SN , method, the angular distribution for solid angle Ω is discretized into a quadrature scheme as shown in
Fig. 6.12 (Reuss, 2008) for the cases N 5 4:8. Thus each octant has 3 and 10 elements, respectively.
The μ-interval (21,1) is divided into n intervals with μ0 5 2 1; μN 5 1. Thus, In the S4 approximation,
μ0 5 2 1; μ1 5 2 1=2; μ2 5 0; μ3 5 1=2; μ4 5 1.
286 Nuclear Engineering

FIGURE 6.10 Transport theory flux in a finite

sphere.

TABLE 6.1 Neutron flux in a finite sphere.

r c 5 0:3a c 5 0:6b c 5 0:9c

4πr 2 ρN ðrÞ 4πr 2 ρðrÞ 4πr 2 ρN ðrÞ 4πr 2 ρðrÞ 4πr 2 ρN ðrÞ 4πr 2 ρðrÞ
0 1.0 1.0 1.0 1.0 1.0 1.0
0.1 0.96427 0.96404 1.03384 1.03273 1.12076 1.11470
0.2 0.91944 0.91851 1.04871 1.04420 1.23423 1.20988
0.3 0.87072 0.86856 1.05026 1.03994 1.33831 1.28317
0.4 0.82055 0.81657 1.04180 1.02300 1.43252 1.33356
0.5 0.77035 0.76382 1.02562 0.99527 1.51681 1.36018
0.6 0.72103 0.71102 1.00350 0.95789 1.59138 1.36195
0.7 0.67318 0.65840 0.97682 0.91126 1.65659 1.33720
0.8 0.62717 0.60576 0.94672 0.85480 1.71285 1.28277
0.9 0.58325 0.55204 0.91410 0.78578 1.76064 1.19143
1.0 0.54155 0.49139 0.87972 0.68833 1.80046 1.02745
a
c 5 0.3, v0 5 1.0026, Z0 5 2.50310.
b
c 5 0.6, v0 5 1.1021, Z0 5 1.19240.
c
c 5 0.9, v0 5 1.9031, Z0 5 0.78964.

In the limits ðμj21 ; μj Þ, the angular flux is approximated by a linear function

μ 2 μj21 μj 2 μ
φðr; μÞ 5 φðr; μj Þ 1 φðr; μj21 Þ: (6.73)
μj 2 μj21 μj 2 μj21

This approximation is then substituted in the transport equation and after integration over μ and N equations are sub-
sequently obtained. To obtain the final equations, an additional equation is required for μ 5 2 1.
For details on the mathematics and iterative algorithms on the SN method, the reader is referred to Clark and
Hansen (1964) and Bell and Glasstone (1979).
 The neutron transport equation Chapter | 6 287

FIGURE 6.11 Transport theory flux for values

approaching c 5 1.

FIGURE 6.12 Discrete ordinates with N 5 4,8.

In the transport equation, the integral term is represented by weighted angular flux at discrete directions. The repre-
sentation is
ð1 X
N
φðx; μ0 Þdμ0 D wi φðx; μi Þ (6.74)
21 i51
 
where the quadrature weights, for a special choice of μi corresponding to PN μi 5 0, given in Table 6.2 (Bell &
Glasstone, 1979)
The discretized transport equation is

@     cX N    
μ φ z; μj 1 φ z; μj 5 wi φ z; μi 1 S z; μj
@z 2 i51

for angular bins specified for

j 5 1; 2; 3; ?; N:
The discrete ordinates method is extensively applied in reactor calculations for lattice cells and whole-core eigen-
value calculations and is being made faster by using parallel algorithms in supercomputing environments.

6.3.2 The Spherical harmonics method

In this section, the spherical harmonics method, mentioned for the two-group transport equations, is described for a
one-speed infinite slab with isotropic scattering followed by anisotropic scattering and then for a critical finite slab.
288 Nuclear Engineering

TABLE 6.2 Quadrature weights and angle for S2 ; S4 ; S6 expansions.

N wi μi
2 w1 5 w2 5 1:000 μ1 5 2 μ2 5 0:57735
4 w1 5 w4 5 0:65215 μ1 5 2 μ4 5 0:33998
w2 5 w3 5 0:34785 μ2 5 2 μ3 5 0:86114
6 w1 5 w6 5 0:46791 μ1 5 2 μ6 5 0:23862
w2 5 w5 5 0:36076 μ2 5 2 μ5 5 0:66121
w3 5 w4 5 0:17132 μ3 5 2 μ4 5 0:93247

Consider now the case of an isotropic plane source SðxÞ 5 δðxÞ=4π. The one-speed equation from Eq. (6.22) is
ð
@ c 1 δðxÞ
μ φðx; μÞ 1 Σt ðxÞφðx; μÞ 5 dμ0 φðx; μ0 Þ 1 : (6.75)
@x 2 21 4π
Expanding the angular flux φðx; μÞ in terms of Legendre polynomials Pm ðμÞ
X
N
2m 1 1
φðx; μÞ 5 φm ðxÞPm ðμÞ (6.76)
m50
4π

with moments defined by

ð1
φm ðxÞ 5 2π φðx; μÞPm ðμÞdμ (6.77)
21

so that the zeroth moment, as defined earlier, is the scalar flux

ð1
ð Þ
φ0 x 5 2π φðx; μÞdμ (6.78)
21

and the first moment is the current

ð1
φ1 ðxÞ 5 2π μφðx; μÞdμ 5 J ðxÞ: (6.79)
21

In the spherical harmonics, PN , method, the angles are written in terms of spherical harmonic functions which are
orthogonal in the range μEð2 1; 1Þ where μ 5 cos ϑ. The first few functions (Table 4.2) are
1 2 
Po ðμÞ 5 1; P1 ðμÞ 5 μ; P2 ðμÞ 5 3μ 2 1 ; . . .
2
and orthogonality gives
ð1
2δn;m
Pn ðμÞPm ðμÞdμ 5 (6.80)
21 ðn 1 1Þ
where

1; n 5 m
δn;m 5
0; n 6¼ m
Substituting Eq. (6.76) in Eq. (6.75)
X
N
dφm XN
μ ð2m 1 1Þ Pm ðμÞ 1 ð2m 1 1Þφm ðxÞPm ðμÞ 5 cφ0 ðxÞ 1 δðxÞ (6.81)
m50
dx m50

Using the recurrence relation for spherical harmonics:

ð2m 1 1ÞμPm ðμÞ 5 ðm 1 1ÞPm11 ðμÞ 1 mPm21 ðμÞ (6.82)
 The neutron transport equation Chapter | 6 289

and the orthogonality of Legendre polynomials, Eq. (6.81) reduces to a numerically useful form:
dφn11 ðxÞ dφ ðxÞ
ðn 1 1Þ 1 n n21 1 ð2n 1 1Þð1 2 cδ0n Þφn ðxÞ 5 δ0n δðxÞ (6.83)
dx dx

1; n 5 0
from which moments can be computed. In Eq. (6.83), φ21 ðxÞ 5 0, and the Kronecker delta function δ0n 5 .
0; n 6¼ 0
Eq. (6.83) is truncated at n 5 N, and called the PN approximation, with the condition
dφN11 ðxÞ
5 0:
dx
An accurate answer is obtained for large N. The first step towards practical application is a slab finite in one dimension con-
sidered infinite in the other two dimensions. Such a configuration is an idealization of a critical slab for reactor calculations as
described for the NDE in Chapter 5. This would require boundary conditions on both sides for incoming and outgoing flux or
current. Spherical harmonics, based on Legendre polynomials have orthogonality for angles in the range 21 # μ # 1 and thus
are inadequate to represent a free boundary condition at the left face of a slab. The inward current at the left face of a slab
(x 5 0), J 1 ð0Þ is zero if there are no incoming neutrons at this face. Such a condition is written as
ð1
J 1 ð0Þ 5 μφð0; μÞdμ 5 0: (6.84)
0

The above is not to be confused with

ð0
2
J ð0Þ 5 μφð0; μÞdμ 6¼ 0 (6.85)
21

such that the net current

J ð0Þ 5 J 1 ð0Þ 2 J 2 ð0Þ 6¼ 0: (6.86)
Similarly on the right side of a critical slab (x 5 a), zero incoming current is expressed as
ð1
2
J ðaÞ 5 μφða; 2 μÞdμ 5 0; (6.87)
0

while
ð1
J 1 ð aÞ 5 μφða; μÞdμ 6¼ 0; (6.88)
0

and

J ðaÞ 5 J 1 ðaÞ 2 J 2 ðaÞ 6¼ 0: (6.89)

The generalized expression of such boundary conditions are the Marshak boundary conditions
ð1 ð1
Pi ðμÞφð0; μÞdμ 5 Pi ð 2 μÞφða; 2 μÞdμ 5 0; (6.90)
0 0

for i 5 1; 3; 5; . . . N:

Exercise 6.1: Marshak conditions

Apply the Marshak boundary conditions for i 5 1, and use Fick’s Law (P1 equations) to show that the linear extrapo-
lation distance d for which φðdÞ 5 0 is

2
d5 ðmfpÞ (6.91)
3
290 Nuclear Engineering

In the P1 approximation, Marshak’s boundary conditions for c 2 1 # 1, that is, c # 2; lead to the extrapolation
distance
2 4 16
x0 5 1 2 ð c 2 1Þ 1 ðc21Þ2 1 ? : (6.92)
3 9 45
Recalling from Eq. (6.31) that c is the mean number of neutrons emerging from a collisions, assumed constant,
c 5 0 for an absorber, c 5 1 for a pure scattering material, c . 1 for neutron multiplication, c 5 ν for fission. Thus, for
the case c 5 1, x0 5 d, as in the case for diffusion theory.
The Mark boundary conditions
 
φ 0; μi 5 0 (6.93)
for i 5 1; 2; 3; . . . ðN 1 1Þ=2; with N odd are another possibility for representing the boundary conditions at the two sur-
faces of a critical slab. Thus
     
φ 0; μ1 5 φ 0; μ2 5 φ 0; μ3 . . . 5 0
where μ1 are obtained from the roots of
PN11 ðμi Þ 5 0: (6.94)
As an example, for N 5 1,
  1 
P2 μ1 5 3μ21 2 1 5 0
2
gives
1
μ1 5 pffiffiffi : (6.95)
3
which is one of the two Gauss quadratures for N 5 2 used in numerical integration also (Section 6.4).
The extrapolation distance for the P1 Mark boundary conditions are
1 1 1
x0 5 pffiffiffi 1 2 ðc 2 1Þ 1 ðc21Þ2 1 ? : (6.96)
3 3 5
For c 5 1; the Marshak and Mark extrapolations length differ by about 12%.
The discrete ordinates method, using Gauss quadrature sets, and spherical harmonics method with Mark’s boundary
conditions are equivalent.
The preceding formulation for isotropic scattering has been extended to the case of anisotropic scattering
ð 2π ð1
@
μ φðx; μÞ 1 φðx; μÞ 5 c dϕ0
dμ0 f ðΩ0 -ΩÞφðx; μ0 Þ 1 Qðx; μÞ: (6.97)
@x 0 21

Since f ðΩ0 -ΩÞ 5 f ðΩ0 ∙ΩÞ 5 f ðμ0 Þ, a spherical harmonics expression

  X N
2m 1 1  
f ðΩ0 -ΩÞ 5 f μ0 5 f l Pl μ0 (6.98)
l50
4π

is used, with moments

ð1
   
f l 5 2π f μ0 Pl μ0 dμ: (6.99)
21
ð1
f 0 5 2π f ðμ0 Þdμ 5 1 (6.100)
21
ð1
φm ðxÞ 5 2π φðx; μÞPm ðμÞdμ (6.101)
21
 The neutron transport equation Chapter | 6 291

With the addition theorem

  Xl
ðl 2 mÞ! m
Pl μ0 5 Pl ðμÞPl ðμ0 Þ 1 2 P ðμÞPm
l ðμ Þ cos mðϕ 2 ϕ Þ
0 0
(6.102)
m51
ðl 1 mÞ! l

where Ω 5 Ωðμ; ϕÞ: The angular flux and angular source φðx; μÞ and Qðx; μÞ are expanded as in Eq. (6.68)
X
N
2m 1 1
Qðx; μÞ 5 Qm ðxÞPm ðμÞ (6.103)
m50
4π

with
ð1
Qm ðxÞ 5 2π Qðx; μÞPm ðμÞdμ: (6.104)
21

With the expansions (.) inserted into (..) and multiplying both sides by
ð1
1
ð2n 1 1Þ Pm ðμÞdμ (6.105)
2 21

the set of coupled ODEs is

dφn11 ðxÞ dφ ðxÞ  
ðn 1 1Þ 1 n n21 1 ð2n 1 1Þ 1 2 cf n φn ðxÞ 5 ð2n 1 1ÞQn ðxÞn 5 0; 1; 2; . . . (6.106)
dx dx
For a P3 solution, for illustration, the four first-order coupled ODEs for a one-dimensional slab are
dφ1 ðxÞ  
1 1 2 cf 0 φ0 ðxÞ 5 Q0 ðxÞ (6.107)
dx
dφ2 ðxÞ dφ0 ðxÞ  
2 1 1 3 1 2 cf 1 φ1 ðxÞ 5 3Q1 ðxÞ (6.108)
dx dx
dφ3 ðxÞ dφ1 ðxÞ  
3 12 1 5 1 2 cf 2 φ2 ðxÞ 5 5Q2 ðxÞ (6.109)
dx dx
and
dφ4 ðxÞ dφ ðxÞ  
4 13 2 1 7 1 2 cf 3 φ3 ðxÞ 5 7Q3 ðxÞ (6.110)
dx dx
These can be solved by standard methods, such as the finite difference method (Section 6.4) to obtain the moments
in the expression
1
φðx; μÞ 5 φ ðxÞP0 ðμÞ 1 3φ1 ðxÞP1 ðμÞ15φ2 ðxÞP2 ðμÞ17φ3 ðxÞP3 ðμÞ (6.111)
4π 0
The angular flux will then be
1 5  7 
φðx; μÞ 5 φ0 ðxÞ 1 3μφ1 ðxÞ1 3μ2 21 φ2 ðxÞ1 5μ3 23μ φ3 ðxÞ (6.112)
4π 2 2
Note that for a P3 formulation, there are four coupled ODEs; in general there are N 1 1 equations with N 1 1 equa-
tions; thus, N 1 1 boundary and interface conditions are required. For the critical slab described above, free-surface
boundary conditions, that is, zero incoming current at the left and right boundaries gives
1 1
J ð0Þ 5 2 φð0Þ; J ðaÞ 5 φðaÞ (6.113)
2 2
which, using Fick’s law can be expressed as

^
φ 1 2Dn∙rφ 50 (6.114)
292 Nuclear Engineering

where n^ is an outward unit normal to a boundary surface.

For a unit lattice cell, a reflective boundary condition, shown in Fig. 6.3, would imply zero (net) current at the point
of reflection xr , i.e J ðxr Þ 5 0 (Fig. 6.13).
The general boundary condition is written in a mixed form, consisting of Dirichlet, Neumann or mixed boundary
conditions.
In spherical geometry, the one-speed equation, using Eqs. (6.11), (6.12) and (6.17),
ð 2π ð1
@ 1 2 μ2 @  
μ 1 1 Σt ðr; EÞ φðr;Ω; E; tÞ 5 dϕ0 dμ0 Σs r; μ0 φðr; μ0 Þ 1 Qðx; μÞ (6.115)
@r r @μ 0 21

and the spherical harmonics expansions (Bell & Glasstone, 1979) leads to the coupled ODEs
   
d n12 d n21
ð n 1 1Þ 1 φn11 ðr Þ 1 n 1 φn21 ðr Þ 1 ð2n 1 1ÞΣs;n φn ðr Þ 5 ð2n 1 1ÞQn ðr Þ (6.116)
dr r dr r
for n 5 0; 1; 2; . . .
  X N
2m 1 1  
Σs r; μ0 5 Σs;m ðr ÞPm μ0 (6.117)
m50
4π

As for the slab case, the N 1 1 equations require the same number of boundary conditions, for which half are speci-
fied at the outer boundary, as the free-surface conditions given above, and the other half requiring all fluxes φn ðr 5 0Þ
at the origin to be finite.
Another set of boundary conditions, as was applied in Chapter 5, is to use a free-surface boundary condition at the
radius and a zero-current symmetry condition Jð0Þ 5 0 at the origin.

Exercise 6.2: Spherical harmonics for an isotropic source

Show that the P1 equations in spherical geometry for an isotropic source Q0 ðr Þ reduce to the diffusion equation
1 d 2 dφðrÞ
r D 2 Σs;0 ðr Þφðr Þ 1 Q0 ðr Þ 5 0 (6.118)
r 2 dr dr

Exercise 6.3: The diffusion coefficient

From the P1 equations, Eqs. (104) and (106), with an isotropic source, show that the diffusion coefficient is
1
D5  : (6.119)
3 1 2 cf 1

FIGURE 6.13 Reflective boundary condition on a unit lattice cell.

 The neutron transport equation Chapter | 6 293

such that the transport cross-section is

 
Σtr 5 Σt 1 2 cf 1 : (6.120)
The average cosine of the scattering angle μ 0 is
Ð1  
21 μ0 f μ0 dμ0
μ0 5 Ð1   5f1 (6.121)
21 f μ0 dμ0

For a scattering, nonfissile, material Eq. (6.120) is

Σtr 5 Σt 2 Σs μ 0 : (6.122)
The forms of r∙J for plane, infinetly long cylinder and sphere are (Section 5.1) of the form,
d
r∙J 5 J x ;
dx
 
d 2
r∙J 5 1 J;
dr r
 
d 1
r∙J 5 1 Jr ;
dr r
which can be expressed as
 
d n
r∙J 5 1 Jr
dr r
with n 5 0; 1; 2 for plane, cylinder and sphere. Similarly the diffusion equation, from the P1 equations is
 
1 d n dφðr Þ
2 n r D 1 Σt φðr Þ 5 Q0 ðr Þ: (6.123)
r dr dr

6.3.3 The DP N method

The double spherical harmonics DPN method, proposed by Yvon (1957; Ziering & Schiff, 1958), gives an expansion of
angular flux in the forward and backward hemispheres of angle, due to a singularity at μ 5 0, that is, at an angle of inci-
dence θ 5 π=2 on a free surface. The method gives good accuracy for a planar free surface or interface between differ-
ent materials where the flux could be anisotropic.
The differences between the PN and the DPN methods are shown in Table 6.3. In the PN method, the angular flux
and source are expanded in terms of a complete set of Legendre polynomials with angular moments φl ðxÞ and orthogo-
nality in the range Að2 1; 1Þ. In the DPN method, the expansion is for the half range at the left hemisphere 21 # μ # 0

TABLE 6.3 Expansions, moments and orthogonality relations for PN and DPN methods.

PN DPN
P
N P
N
φðx; μÞ 5 2n 1 1
2 Pn ðμÞφn ðx Þ φðx; μÞ 5 ð2n11ÞPn ð2μ 2 1Þφ1
n ðx Þ; 0 # μ # 1
l50 n50

PN
φðx; μÞ 5 ð2n11ÞPn ð2μ 1 1Þφ2 n ðx Þ; 2 1 # μ # 0
n50
Ð1 1
Ð1
φn ðx Þ 5 21 Pn ðμÞφðx; μÞd μ φn ðx Þ 5 0 Pn ð2μ 2 1Þφðx; μÞd μ
Ð0

φ2n ðx Þ 5 21 Pn ð2μ 1 1Þφðx; μÞd μ
Ð1 2
Ð1
21 Pn ðμÞPm ðμ0 Þd μ 5 2n 1 1 δnm Pn ð2μ 2 1ÞPm ð2μ 2 1Þd μ 5 2n11 1 δnm
Ð00

1
21 Pn ð2μ 1 1ÞPm ð2μ 1 1Þd μ 5 2n 1 1 δ nm
294 Nuclear Engineering

and the half range at the right hemisphere 0 # μ # 1 of the angular domain with angular moments in the positive range,
φ1 2
l and in the negative range φl . The orthogonality of the Legendre polynomials is separately for both angular
domains.
When the flux is not heavily angle-dependent a few terms are sufficient in the spherical harmonics PN method; con-
versely a high order PN expansion is required for good accuracy when the angular dependence is prominent as could be
the case for an interface between two very different materials.
Both the DP0 and the P1 are approximations of the diffusion equation. It is better to apply Yvon’s method near the
layer boundaries and for optically thin media while the PN method would be expected to be better at large distance
from boundaries or for optically thick media.
A finite difference form for the DPN expansion can be obtained by mutilying Eq. (6.98) with Pm ð2μ 2 1Þ and inte-
grating over the interval 0 # μ # 1 and then by Pm ð2μ 2 1Þ and integrating over 21 # μ # 0. The first term is then
ð1
@φðx; μÞ
dμμPm ð2μ 2 1Þ
0 @x
which becomes
ð1
@X N
ð2n11ÞPn ð2μ 2 1Þφ1
n ðxÞ dμμPm ð2μ 2 1Þ: (6.124)
@x n50 0

The integration over μ is carried out using the recurrence relation

1
μPn ðμÞ 5 ½ðn 1 1ÞPn11 ðμÞ 1 nPn21 ðμÞ
2n 1 1
So that Eq. (6.125) becomes
m dφ1m21 ðxÞ m 1 1 dφ1
m11 ðxÞ
1
2m 1 1 dx 2m 1 1 dx
The full equation becomes
6 6
m dφm21 ðxÞ m 1 1 dφm11 ðxÞ dφm6 ðxÞ
1 6 1 2Σt ðxÞφm6 ðxÞ
2m 1 1 dx 2m 1 1 dx dx
X
L N 
X
6 1 2 2
5 ð2l 1 1Þplm Σs ðxÞ 2n 1 1Þ p1
lm φn ðxÞ 1 plm φn ðxÞ (6.125)
l50 n50

where
ð1
p1
lm 5 Pl ðμÞPm ð2μ 2 1Þdμ
0

and
ð0
p2
lm 5 Pl ðμÞPm ð2μ 1 1Þdμ
21
6
so that p0n 5 δ0n : Eq. (6.125) are a set of coupled ODEs which can be solved by standard methods.
The DPN method has been applied by Ziering and Schiff and by Gelbard, Davis, and Pearson (1959) for plane
geometry with four angular intervals to calculate the angular flux accurately. By the 1960s, the method had been
extended to eigenvalue problems in two-dimensional geometries, and by the 1980s, it had been applied to PHWRs using
the one-group and 27-group steady-state integral equation with isotropic scattering and constant source (Krishnani,
1982). A cylindrical lattice cell is used for modeling the 7-, 19- and 28-rod PHWR fuel clusters (Section 3.3) for hexag-
onal lattices with DP1 and DP2 expansions to calculate the flux distribution and overall keff . The DPN method for slab
geometry was incorporated into a multigroup transport code (Stepanek, 1981) and found to be more efficient than com-
parable SN methods. Using a variational principle, the DPN method has been extended to 2D plane geometry (Ghazaie,
Zolfaghari, & Abbasi, 2017) and applied to several benchmarks including a 2D lattice cell and a miniature reactor
problem.
 The neutron transport equation Chapter | 6 295

6.3.4 The B N method

In the BN method, where the B denotes buckling (for spherical or cylindrical bare reactors for example, Section 5.2),
the spatial part of the angular flux is assumed to be of a cosine form for a multiplying medium and an imaginary num-
ber for a nonmultiplying such as a reflector surrounding a core. The flux is assumed to be of the form
φðx; μ; EÞ 5 e2iBx φðμ; EÞ: (6.126)
and similar expansions are assumed for the source term
Qðx; μ; EÞ 5 e2iBx Qðμ; EÞ
For a bare reactor B is the buckling, and the real part of the above is
φðx; μ; EÞ 5 cos Bxφ ðμ; EÞ:
With the expansion given by Eq. (6.126), the “within group” flux equation becomes
XN ð ð1
2n 1 1 1
½Σt ðEÞ 2 iBμ φ ðμ; EÞ 5 Pn ðμÞ dE Σs;n ðE -EÞ
0 0
dμ0 φðE0 ; μ0 ÞPl ðμ0 Þ 1 SðEÞ (6.127)
n50
2 21 2

Dividing by ½Σt 2 iBμ and multiplying the above by Pl ðμÞ and integrating, with
ð1
dμPn ðμÞφðμ; EÞ 5 φn ðEÞ
21

and defining
ð1
1 Pl ðμÞPn ðμÞ
dμ 5 Al;n
2 21 1 2 μgðEÞ
½
and
gðEÞ 5 iB=Σt ðEÞ
gives
X
N ð
Σt ðEÞφl ðEÞ 5 ð2n 1 1ÞAl;n dE0 Σs;n ðE0 -EÞφn ðE0 Þ 1 SðEÞAl;n : (6.128)
l50

For an isotropic source S0 ðEÞ;

X
N ð
Σt ðEÞφðμ; EÞ 5 ð2n 1 1ÞAl;n dE0 Σs;n ðE0 -EÞφn ðE0 Þ 1 S0 ðEÞAl;0 (6.129)
l50

Then
ð1
1 1 1   1 j1 1 gj 1
A0;0 ðgÞ 5 dμ 5 2 ln 1 2 g 2 ln 1 1 g 5 ln 5 tanh21 g
2 21 ½1 2 μgðEÞ 2g 2g j1 2 gj g
since
 
1 x11
ln 5 tanh21 x:
2 x21
Also
ð1
1 μ 1   1 
A1;0 5 dμ 5 2 2 2g 2 ln 1 2 g 1 ln 1 1 g 5 A0;0 ðgÞ 2 1
2 21 ½1 2 μgðEÞ 2g g
1 δl;n
ð2n 1 1ÞAl;n ðgÞ 2 ðn 1 1ÞAl;n11 ðgÞ 2 Al;n21 ðgÞ 5 :
gð E Þ g
Note that Al;n 5 An;l .
296 Nuclear Engineering

Consider the B0 approximation; with l 5 0, the scattering moments are truncated at Σs;0 ðE0 -EÞ to give the integral
equation
ð
Σt ðEÞφ0 ðEÞ 5 A0;0 dE0 Σs;0 ðE0 -EÞφ0 ðE0 Þ 1 S0 ðEÞA0;0 : (6.130)

For higher moments φl ,

X
N ð
Σt ðEÞφl ðEÞ 5 ð2l 1 1ÞAl;n dE0 Σs;l ðE0 -EÞφn ðE0 Þ 1 S0 ðEÞAl;n
l50

and the angular flux φðμ; EÞ is easily found as

ð
A0;0
φðB; μ; EÞ 5 dE0 Σs;0 ðE0 -EÞφ0 ðE0 Þ 1 S0 ðEÞ (6.131)
½Σt ðEÞ 2 iBμ
It is also possible to find all the φl ðEÞ for n 5 0 from
X
N ð
Σt ðEÞφl ðEÞ 5 ð2n 1 1ÞAl;n dE0 Σs;n ðE0 -EÞφn ðE0 Þ 1 S0 ðEÞAl;0 (6.132)
n50

In the B1 approximation, there will be two equations, one for n 5 0 and the other for n 5 1: Group constants, as
defined in the two-group transport equation (Section 6.1) can be determined with this method. The general procedure
for the BN method is that a buckling is assumed for which a group-1 calculation is carried out. From this, a multigroup
calculation gives the spatial distribution of the flux; if this differs considerably from the assumed distribution, then a
series of iterations is carried out until the fluxes converge. When multigroup fluxes are found, the overall system criti-
cality can be found. It is found that the converges of the BN method is better than that of the PN method.

6.3.5 The finite element method

The first and simplest numerical methods for computing fluxes were based on easy-to-use finite difference methods
(Section 4.5.1) using spatial discretization but these were suited to square-type grids. These were refined in several
ways, such as the FEM using “shape functions” (Section 4.5.2) to model irregular geometry as well as polynomial-type
variations in the flux. The process of using these finite “elements” which are first solved at a unit level and then com-
bined into a “global” system reduces the coupled differential equations into a system of linear algebraic equations of
the form ̿A X 5 B which are subsequently solved by traditional methods of linear algebra. Techniques based on continu-
ous and discontinuous Galerkin FEMs have also been used for improving the computational efficiency. For large pro-
blems, iterative schemes and convergence techniques have been used so that the FEM is widely used in reactor physics
(Ackroyd, 1992; Kang & Hansen, 1973; Rothenstein & de Oliveira, 1991).

6.3.6 The nodal method with transport theory

In a typical 1000 MW PWR, there could be B3 M zones for which multiphysics computations would be required for
carrying out stead state core analysis and other transient and safety studies. If in each zone a fine mesh structure would
be used, the computational burden would become prohibitively large for supercomputers. Thus, in the first step a lattice
cell calculation can be done using a fine mesh and reflective boundary conditions with a 2D transport code.
Subsequently, in the second step, the resulting fluxes can be used to obtain spatially homogenized cross-sections for a
simplified core model with nodes as each of the B3 M zones which could then be simulated with a less computation-
ally expensive diffusion code (Smith, 1986). The nodal fluxes are then given a reconstructed shape from the nodal
expansion function (Lawrence, 1986; Putney, 1986). High speedups with parallel algorithms have been demonstrated
for Open multiprocessors for a pin-resolved 3D seven-group NuScale modular reactor core neutron transport calcula-
tions (Wang et al., 2019). Similarly, high accuracy and attractive computational speed has been demonstrated for pin
power reconstruction by the high order triangle-based polynomial expansion nodal (TPEN) algorithm for a VVER-1000
water-water Russian energetic reactor (Safarzadeh, 2020).
 The neutron transport equation Chapter | 6 297

6.3.7 Hybrid methods

There are several problems in nuclear engineering when a deterministic, or stochastic, approach in itself is not sufficient
to give accurate answers within the limits of realistic computational effort. One such area is radiation shielding in ducts
for which a hybrid computation, coupling a deterministic approach such as a 1D SN transport computation with Monte
Carlo simulation can be beneficial. An initial run of a deterministic code could give the biasing parameters favorable
for a detailed Monte Carlo simulation to get “good” results which otherwise may not be possible due to large variances
from analog simulations.
Another class of problems is the classical source-detector configuration where the “forward” or standard transport
equation represents a solution at a detector given the source and conversely the “backward” or adjoint transport equa-
tion represents the importance function, that is, the importance of a particular phase space to the detector. Carrying out
such coupled forward-backward computations can give the biasing parameters for a Monte Carlo simulation. These
have also been used for design optimization in a variational formulation.

6.3.8 Criticality estimates

The single most important characteristic of a multiplying system is its criticality. In Chapter 2, the criticality of the
Godiva and Jezebel assemblies was described, with critical radii given in Tables 2.19 and 2.22. The one-group critical-
ity equation (Section 5.2.3) was the first expression given for estimating the keff of a bare slab, sphere or cylinder. The
next step was to obtain a slightly better two-group criticality formula (Section 5.3.2) in which both fast and thermal lea-
kages were incorporated. From a two-group, the formulation of a multigroup diffusion model was discussed in
Section 5.4. The eigenvalue formulation of a steady-state neutron balance equation amounted to a problem in which the
largest eigenvalue found had the physical significance of being related to the critical configuration. This was demon-
strated for a two-group two-region spherical core surrounded by a reflector; a deterministic model resulted in the critical
determinant from which the critical dimension was obtained. It was then understood that a criticality problem could be
to determine the critical size of a system given its material composition, or alternately, to determine the material com-
position that would result in criticality for a certain shape. These were central to the description of nuclear power reac-
tors, propulsion reactors and space reactors in Chapter 3. Whether it was PWRs, BWRs, PHWRs or any other reactor
type, the most important design parameter was its critical configuration. We thus came to a description of neutron trans-
port, in the present chapter. The improvement of the formulations in this chapter, over those in the previous chapter,
was essentially in the description of the angular flux. This added a complexity to the mathematical description. We now
realize that exact solutions of transport theory, done in the early 1960s and 1970s could also give criticality estimates,
for example, from the asymptotic flux using the Mark boundary conditions (Eq. 6.96) and requiring the flux to be zero
at the extrapolated distance gave the first estimates from this end point method of critical radius for a bare sphere as
a 5 π ν 0 ðcÞ 2 x0 :
As an example, for 5 1:20; ν 0 ðcÞ 5 1:198, These are listed in Table 6.4 for some values of c, the mean number of
secondary neutrons emerging from a collision; the exact method rnsesults are from a transport theory solution by varia-
tional method.
From the exact solutions (Section 6.2) the description of numerical methods such as the SN and PN methods gave a
better capability to solve eigenvalue multigroup transport problems to determine the systems criticality. Comparisons of
transport theory estimates for the Godiva, Jezebel, Topsy and other assemblies (Bell & Glasstone, 1979; Bowen &
Busch, 2005; Rawat & Mohankumar, 2011; Rowlands et al., 1999; Thompson et al., 2020) show good results between
theory and experiments.

TABLE 6.4 Critical radius estimates of a sphere (in mean free paths); exact vs PN method.

c End point Exact P1 P3 P5

1.05 7.277 7.2772 7.543 7.296 7.284
1.20 3.172 3.1720 3.513 3.204 3.181
1.60 1.476 1.4761 1.850 1.550 1.497

Source: Bell, G. I., & Glasstone, S. (1979). Nuclear reactor theory. New York: Robert E. Krieger Publishing Co., Inc., 1979, Table 2.7, p. 101.
298 Nuclear Engineering

It is seen that a multigroup (typically 6, 12, or 24 groups) S4 ; P3 calculation is sufficient to estimate the criticality of
fast systems such as Godiva with certain data libraries. To close this chapter, we also describe methods, such as the
MOC which have the capability of performing lattice cell calculations leading to whole-core neutronic eigenvalue cal-
culations useful for the design of nuclear systems described in Chapter 3. Later in this book, we will cover the Monte
Carlo method which is another powerful method as well as a simulation tool for the design of nuclear systems.

6.4 Transport theory for reactor calculations

The NDE, based on Poisson’s equation, is mathematically much simpler than the integro-differential transport equation
and hence is used for simple problems. Even then, numerical solutions are required for large problems such as “whole-
core” reactor design. In the multigroup form, the diffusion equation is used to obtain “first estimates” and can serve as
a useful step for providing guesses to full transport calculations. A limitation of diffusion equation is near sources and
boundaries where the angular flux requires detailed consideration as provided, for example, in the discrete ordinates
and spherical harmonics methods.
Reactor core calculations are carried out primarily to determine the overall core flux and distributions during
steady-state operation as well as in the case of operational perturbations and in the determination of safety margins in
case of accidents.
A typical 1000 MW PWR has an equivalent core diameter of B3.37 m, heightB 3.66 m with 193 fuel assemblies,
each 17 3 17 assembly containing 264 fuel rods. Thus there are 50,952 fuel rods in the core. If each fuel rod is parti-
tioned, for computational purposes, into 50 zones axially then the total number of zones for which computation is
required becomes 2,547,600. Within the assembly, we can identify a repeating pattern consisting of a cell with a fuel
rod, clad and water coolant.
While it is desirable to carry out calculations for a full reactor core, it is generally quite difficult to do this due to
the complicated geometry of the core, consisting of fuel pellets in fuel pins, cladding and structural material. Further,
the heterogeneity of the core is due to various types of materials such as variable enrichment fuel, absorbers, and guide
tubes. There is also the energy dependence of interaction cross-sections with resonances, as mentioned in Chapter 1.
Thus, a two-step procedure is used. In the first step, lattice calculations over unit lattice cells, typically cylindrical,
square, hexagonal (Fig. 6.4) or plate-type, are carried out with an ultra-fine and fine energy mesh structure for a fuel
pin-cell with reflective boundary conditions. The purpose here is to generate energy- and space-averaged cross-sections,
weighted over the neutron flux, from evaluated nuclear data files. In the second step, these “averages” are used as
homogenized fuel assemblies with a coarse mesh in an integrated whole-core calculation (Fig. 6.14).
In obtaining “averages” for a whole-core calculation, a considerable amount of data processing and reactor physics
is required to accurately obtain multigroup cross-sections particularly accounting for the resonances in the cross-section
data. The number of groups used depends on the level of accuracy required; as an example, it is typical to take a 70-
group structure for fast reactor cores and over 100 groups for thermal reactors which can be further group-collapsed
into two-group cross-sections. The “computational burden” in whole-core calculations is reduced by using the diffusion
equations for scalar flux instead of the transport equation for angular flux.
In deterministic methods, computations for the neutron flux are carried out by discretizing in space, energy and
angle so that the entire domain is divided into a finite number of spatial zones, each of uniform nuclear properties, a

FIGURE 6.14 Cylindrical, square and hexagonal lattice cells.

 The neutron transport equation Chapter | 6 299

finite number of energy groups and a finite number of angles. Thus, each discretization has its own limitations such as
truncation errors.
In a nuclear reactor core there are typically several assemblies consisting of fuel, control rods, grid spacers, modera-
tor, coolant and supporting structures. These assemblies are placed together in a lattice which could be rectangular as in
the case of water-cooled reactors, and hexagonal as in the case of sodium cooled reactors. Thus, the generation of a
reactor core mesh can become very complicated and could require large memory storage for a complete description.
For this purpose, open-source mesh generation software is available for efficiently modeling realistic cores which could
require millions of elements defined by mesh vertices (Jain & Tautges, 2014) requiring a mesh file size of a few GB.
Modeling a fast reactor core, such as the Japanese MONJU reactor, comprising eight assembly types of 715 assemblies
requires 101 million hexahedral elements generated by 712 processors in parallel. In realistic simulations involving the
modeling of neutron transport with thermal hydraulics over possibly more than 1 billion hexahedral elements, the task
of mesh generation is cumbersome as is the task of computation necessitating the most powerful supercomputers with
hundreds of thousands of CPU cores.

6.4.1 Collision probability method

The collision probability Pij;g is defined as the probability that a neutron born, isotropically in the lab system and with a
uniform spatial probability, in any region V i of the lattice has its first collision in some region of interest V j of a unit
cell. The neutron flux and collision density are found from the spatially discretized multigroup integral form of the
transport equation by tracking a neutron in a straight line to its next collision. After tracking a large number of trajecto-
ries the collision probabilities for all cells are computed and the scalar fluxes are then determined from a relationship of
the form (Cacuci, 2010)
P
Qi;g V i Pij;g
φj;g 5 i (6.133)
V j Σj;g
where Qi;g is the source term (including fission) for in-scattering into region i and energy group g, Σj;g is the macro-
scopic total cross-section in region j and energy group g.
In a computer code such as DRAGON (Marleau, Hébert, & Roy, 2011), the integral form of the transport equation
is written in terms of collision probabilities for which symmetry and reciprocity relations can be derived. Together with
conservation relations (Stoke’s theorem across boundaries), and with further assumptions of isotropic and uniform
source, simple and compact expressions can be obtained for scalar flux and angular currents.
In the CPM, there are two main disadvantages: (1) full square matrices are produced, for example, a 20 3 20 matrix
for 20 spatial regions, and (2) scattering is restricted to isotropic collisions in the L system. The CP method is thus pre-
ferred for problems where the number of regions is small and meshes are unstructured.

6.4.2 Method of characteristics

In the MOC, the solution is found for a fixed angle along a “straight path” across regions divided into uniform zones.
In the integral transport (Section 6.1.2)
ð N Ð s0
2 Σ ðr2s0 0 Ω;EÞds0 0  
φðr; Ω; E; tÞ 5 e 0 t q r 2 s0 Ω; Ω;E; t 2 s0 =v0 ds0 (6.134)
0

with
ðN ð ðN ð
1
qðr; Ω; E; tÞ 5 dE0 dΩ0 Σs ðr; E0 ; Ω0 ∙ΩÞφðr; Ω0 ; E0 Þ 1 χð E Þ dE0 dΩ0 νΣf ðE0 Þφðr; Ω0 ; E0 Þ 1 Sðr; Ω0 ; E0 Þ
0 4π 0
(6.135)
the source term is written as
qki:j 5 QkF;i 1 QkS;i 1 Qki (6.136)
where the indices i; j; k refer to the mesh, characteristic line (CL) and energy group respectively (Fig. 6.15).
A MOC computation flowchart, based on Eqs. (6.136
6.140), is shown in Fig. 6.16. At the first level, a unit lattice
cell calculation begins with the selection of a representative cell such as the one shown in Fig. 6.15.
300 Nuclear Engineering

FIGURE 6.15 Characteristic lines in a lattice cell.

The cell is divided into zones within which the data is uniform, such as a fuel zone, a cladding zone and a coolant
zone. The group cross-section data for each of these zones is given. The first step is an initialization of scalar fluxes,
angular fluxes and the system multiplication keff .
A CL, represented by the index j, has a specified polar angle θ and a specified azimuthal angle ϕ; parallel lines are
chosen by varying the initial point.
In the first step, with the initialized values, the fission source for zone i, energy group k. QkF;i is calculated from the
0
initialized scalar fluxes ϕki , as

1 χðEÞ X G
0 0
QkF;i 5 νΣk φk :
4π keff k0 51 F i

The fission source is calculated from Eq. (6.137) for all energy groups k in all meshes i.
In the second step, the scattering source QkS;i is calculated from all in-scattering cross-sections and scalar fluxes.
In the next step the outgoing angular flux φi;j out;k
from a mesh is calculated from the incoming angular flux along a
CL ðjÞ, φi;j
in;k

2Σki Δs Qki  2Σki Δs


φi;j
out;k
5 φin;k
i;j e 1 1 2 e (6.137)
Σki
where Qki 5 QkF;i 1 QkS;i and Δs is the distance traveled along the CL across the mesh.
The average angular flux, along this CL is

k Qki φin;k
i;j 2 φi;j
out;k
φ i;j 5 1 (6.138)
Σki Σki Δs
k
When all such CLs are processed, their average is the cell averaged group scalar flux φ i from which the number of
fissions, summed over all I meshes and G energy groups, is calculated as
X
I X
G
νΣkF φki (6.139)
i51 k51
 The neutron transport equation Chapter | 6 301

FIGURE 6.16 Flowchart for the method of characteristic in a lattice cell.

If the “new” scalar fluxes for the mth generation φki 5 φm;k
i are used for the present generation and the “old” scalar
fluxes φki 5 φm21;k
i are used for the previous, or m 2 1th, generation, then the ratio of the mth and m 2 1th generation of
fissions neutrons would be the mth estimate of the system multiplication kðmÞ
eff ; thus

P
I P
G
νΣkF φm;k
i
kðmÞ 5 i51 k51
(6.140)
eff
P
I PG
νΣkF φm21;k
i
i51 k51

Inner and outer iterations are carried out until the scalar fluxes and the kðmÞ
eff converge. This is the essence of an
eigenvalue calculation at the level of the unit lattice cell.
There are several solvers used in MOC applications in reactor physics such as the cyclic tracking concept using infi-
nite tracks with periodic characteristics and finite tracks with explicit boundary conditions. The latter are “faster sol-
vers” and therefore more widely used in production codes. The numerical computation in the MOC is based on the
ODE form, called the characteristic form, of the NTE and a track is computed as it traverses across homogenized spatial
302 Nuclear Engineering

zones, crossing boundaries in a fixed solid angle, through the entire domain. Integrals are cast using quadrature schemes
to compute scalar fluxes and boundary currents.
Several computation strategies have been developed and demonstrated for acceleration with advantages in solving
unstructured 2D and structured 3D reactor core configurations. The MOC has been implemented in WIMS-E (Lindley
et al., 2017) and DRAGON codes. The MOC has been used for fundamental lattice calculations in several applications
such as in an ACR-type cell (Le Tellier & Hébert, 2005) consisting of a cluster geometry with light water coolant and
heavy water moderator. Algorithms have been successfully implemented for full core simulations with modern single
instruction multiple data (SIMD) computer architectures for high performance computing (Tramm et al., 2016).

Problems
1. Compare the neutron flux obtained by transport theory with that obtained by diffusion theory in a finite sphere with
a point isotropic source at its center. Plot the flux for a source of 107 neutrons s21.
2. The asymptotic flux in a source-free sphere is given as rϕas ðr Þ 5 A sin jνro j. In end point theory an estimate for the
critical radius R is obtained by putting the asymptotic flux equal to zero at the extrapolated boundary. This gives
R 5 π ν o ðcÞ 2 d, where d is the extrapolation radius. Using the data for Godiva estimate the critical radius and com-
pare with that obtained with diffusion theory.
3. Compare the asymptotic flux of Problem 2 with the exact transport theory asymptotic flux and with the diffusion
theory flux in Chapter 5. pffiffiffiffiffiffiffiffiffiffiffi
4. Compare the reflectivity, or albedo, φð0; 2 μÞ 5 1 2 1 2 cHðμÞ obtained from transport theory: with the corre-
sponding estimate from diffusion theory and calculate the albedo of a slab of graphite 3 m.f.p. thick.

Nomenclature
English lower case
d extrapolation distance
keff effective multiplication
kN infinite multiplication
n number density

English upper case

Ai area
Bg geometrical buckling
Bm material buckling
BN the BN method
C^ collision operator
Ci concentration of precursor
D
D^ derivative operator
DPN the DPN (double spherical harmonics) method
G Green’s function
J neutron current
L diffusion length
1
L^ adjoint operator
Ni atomic density of the ith nuclide
Ni shape function in the ith element
Pes escape probability
Pm l associated Legendre functions
PN the PN (spherical harmonics) method
Rc critical radius
S source
SN the SN (discrete ordinates) method
T^ transport operator
 The neutron transport equation Chapter | 6 303

Greek lower case

λ decay constant
μ cosine of angle of scattering
μ0 average cosine of scattering angle
φ flux
φ1 adjoint flux
φC complementary solution
φP particular solution
τ neutron age
ν
χ fission spectrum
ψ collision density

Greek upper case

Ω solid angle
P
macroscopic absorption cross-section
Pa
macroscopic fission cross-section
Pf
macroscopic removal cross-section
Pr
macroscopic scattering cross-section
Ps
macroscopic transport cross-section
Ptr
t macroscopic total cross-section

Abbreviations
FD finite difference method
FEM finite element method
FVM finite volume method
IFBA integral fuel burnable absorber
MOC method of characteristics
NDE neutron diffusion equation
NEM nodal expansion method

References
Ackroyd, R. T. (1992). Finite element methods for neutron transport based on maximum and minimum principles for discontinuous trial functions.
Annals of Nuclear Energy.
Bell, G., & Glasstone, S. (1952). Nuclear reactor theory. van Nostrand Reinhold Company.
Bell, G. I., & Glasstone, S. (1979). Nuclear reactor theory. New York: Robert E. Krieger Publishing Co., Inc..
Bowen, D. G., & Busch, R. D. (2005). Hand calculational methods for criticality safety
A primer. In 2005 NCSD topical meeting (American
Nuclear Society Nuclear Criticality Safety Division).
Cacuci, D. G. (2010). Handbook of Nuclear Engineering, 2.
Carlson, B. G., & Lathrop, K. D. (1968). Transport theory the method of discrete ordinates. Computing methods in reactor physics. .. .
Case, K. M., & Zweifel, P. F. (1967). Linear transport theory. Reading, MA: Addison-Wesley Pub. Co..
Cercignani, C. (1988). The Boltzmann equation and its applications. Springer.
Chandrasekhar, S. (1960). Radiative transfer. Dover Publications.
Clark, M., & Hansen, K. F. (1964). Transport methods. Numerical methods of reactor analysis. .. .
Davison, B. (1957). Neutron transport theory. New York: Oxford University Press.
Duderstadt, J. J., & Hamilton, L. J. (1976). Nuclear reactor analysis. John Wiley and Sons.
Emmett, M. B. (2000). MORSE: Present capabilities and future directions. Applied radiation and isotopes. .. .
Erdmann, R. C., & Siewert, C. E. (1968). Green’s functions for the one-speed transport equation in spherical geometry. Journal of Mathematical
Physics.
Ganapol, B. D. (2008) Analytical benchmarks for nuclear engineering applications case studies in neutron transport theory.
Gelbard, E., Davis, J., & Pearson, J. (1959). Iterative solutions to the P 1 and double- P 1 Eq. 1. Nuclear Science and Engineering.
Ghazaie, S. H., Zolfaghari, A., & Abbasi, M. (2017). An extended half-range spherical harmonics method for first-order neutron transport equation
based on variational treatment. Progress in Nuclear Energy.
Harris, S. (1971). An introduction tot he theory of the Boltzmann equation. New York: Dover Publications, Inc.
Henry, A. F. (1975). Nuclear reactor analysis. The M.I.T. Press.
304 Nuclear Engineering

Jain, R., & Tautges, T. J. (2014). Generating unstructured nuclear reactor core meshes in parallel. Procedia Engineering.
Kang, C. M., & Hansen, K. F. (1973). Finite element methods for reactor analysis. Nuclear Science and Engineering.
Krishnani, P. D. (1982). Interface current method for PHWR cluster geometry with anisotropy in the angular flux at interfaces. Annals of Nuclear
Energy.
Lawrence, R. D. (1986). Progress in nodal methods for the solution of the neutron diffusion and transport equations. Progress in Nuclear Energy.
Le Tellier, R., & Hebert, A. (2005). Application of the characteristics method combined with advanced self-shielding models to an ACR-type cell. In
26th Annual Canadian Nuclear Society Conference and 29th CNS/CNA Student Conference.
Lewis, E. E., & Miller, J. W. F. (1984). Computational methods of neutron transport. New York: John Wiley and Sons.
Lindley, B. A., et al. (2017). Current status of the reactor physics code WIMS and recent developments. Annals of Nuclear Energy.
Marleau, G., Hébert, A., & Roy, R. (2011). A user guide for DRAGON version 4. Inst. ge´nie nucle´aire, De´partement ge´nie me´canique, E´cole
Polytech. Montre´al. (Report IGE
294).
Putney, J. M. (1986). A hexagonal geometry nodal expansion method for fast reactor calculations. Progress in Nuclear Energy.
Rawat, A., & Mohankumar, N. (2011). Benchmark results for the critical slab and sphere problem in one-speed neutron transport theory. Annals of
Nuclear Energy.
Reuss, P. (2008). Neutron physics. Cedex, France: EDP Sciences.
Rief, H. (1984). Generalized Monte Carlo perturbation algorithms for correlated sampling and a second-order Taylor series approach. Annals of
Nuclear Energy.
Robson, R. E., White, R. D., & Hildebrandt, M. (2017). Fundamentals of charged particle transport in gases and condensed matter.
Rothenstein, W., & de Oliveira, C. R. E. (1991). Discrete energy or multigroup finite element transport calculations in lattice physics. Progress in
Nuclear Energy.
Rowlands, J., et al. (1999). Intercomparison of calculations for Godiva and Jezebel.
Safarzadeh, O. (2020). Development of the triangle-based nodal algorithm for reconstructing pin power distributions. Progress in Nuclear Energy.
Siewert, C. E., & Grandjean, P. (1979). Three basic neutron transport problems in spherical geometry. Nuclear Science and Engineering, 70, 96
98.
Smith, K. S. (1986). Spatial homogenization methods for light water reactor analysis. Progress in Nuclear Energy.
Stepanek, J. (1981). The DP N surface flux integral neutron transport method for slab geometry. Nuclear Science and Engineering.
Thompson, N., et al. (2020). National Criticality Experiments Research Center (NCERC) - Capabilities and recent measurements. EPJ Web of
Conferences.
Tramm, J. R., Gunow, G., He, T., Smith, K. S., Forget, B., & Siegel, A. R. (2016). A task-based parallelism and vectorized approach to 3D Method of
Characteristics (MOC) reactor simulation for high performance computing architectures. Computer Physics Communications.
Wang, Y., Zhang, T., Lewis, E. E., Yang, W. S., Smith, M. A., & Wu, H. (2019). Three-dimensional variational nodal method parallelization for pin
resolved neutron transport calculations. Progress in Nuclear Energy.
Williams, M. M. R. (1971). Mathematical methods in particle transport theory. New York: John Wiley & Sons.
Yvon, J. (1957). La diffusion macroscopique des neutrons une methode d’approximation. Journal of Nuclear Energy, 4(3), 305
318.
Ziering, S., & Schiff, D. (1958). Yvon’s method for slabs. Nuclear Science and Engineering.
Chapter 7

The Monte Carlo method

The Monte Carlo (MC) method (Dunn & Shultis, 2012; Kalos & Whitlock, 2008; McClarren, 2018) for radiation trans-
port simulation is a powerful methodology for modeling realistic and often complex engineering systems. Though the
method, with all its mathematical rigor and its strong foundations in the laws of probability, was fairly established, it
was not until the Manhattan Project that it attracted the attention of nuclear scientists challenged with the estimation of
the composition, dimensions and configurations of nuclear weapon systems.
The name “Monte Carlo” was given by Nicholas Constantine Metropolis (1915
99), of the Theoretical Physics
Division of the Manhattan Project, in 1942, (Keys & Groves, 1963) to this statistical process of simulation (Metropolis &
Ulam, 1949). His colleagues included Teller, John von Neumann, Stanislaw Ulam, and Robert Richtmyer. Those were the
days of MANIAC, The Mathematical and Numerical Integrator and Computer, designed according to von Neumann’s prin-
ciple of the stored program (Anderson, 1986). It is said that Fermi used statistical sampling techniques as early as 1934,
when he was working on neutron diffusion in Rome. When, in 1944, the development of the atomic-bomb project entered
its final phase; the three problems at the center of the nuclear weapons program: neutron transport, behavior of materials
under very temperatures and pressures, and fluid dynamics, required mathematical computation for design and prediction.
Earlier work was developed by Hammersley and Handscombe and Spanier and Gelbard (Hammersley &
Handscomb, 1964; Hammersley, Handscomb, & Weiss, 1965; Spanier, Gelbard, & Bell, 1970). It does not per se
involve the solution of any equation but can be interpreted to follow the integral form of the transport equation where
the flux is expressed in terms of the uncollided and collided terms.
Several mathematical methods were developed, along with powerful computing hardware, that led to the successful
testing of a nuclear weapon and the advent of the first electronic computer during World War II.
Since then, MC methods have been important in many areas of science and engineering, as well as to medicine and fore-
casting. In nuclear engineering, MC simulation is an essential component of the curriculum, starting preferably at the
advanced undergraduate level. A picture of radiation transport in the mind of a student who has not been exposed to MC
could be restrictive in several ways. Here, it is easier to relate to concepts such as flux, current, and the overall transport
flow in complex geometries.
This chapter builds upon the fundamentals described for charged particles and gamma transport in Chapter 1, neu-
tron interactions with matter in Chapter 2, mathematical foundations given in Chapter 4, and diffusion and transport
modeling in Chapters 5 and 6. Here the simulation process is applied to the nuclear systems covered in Chapter 3 with
emphasis on modeling fission and fusion systems.

7.1 Stochastic simulation

7.1.1 Markov processes
In Section 4.9.2, a Markovian process was described as a stochastic process in which the next state Pi11 depends on its
present state Pi but there is no memory of the
g0

past. 
The transport kernel (Section 6.1.4) T^ r0 -r; Ω ^ 0 represents a neutron
at r0 ; Ω ^ 0 ; E0as shown in Fig. 7.1 being trans-
g0 -g
ported to position r where an interaction takes place represented by C ^ ^ 0
r; Ω -Ω ^ which changes the phase space
0 ^ 0 0 ^ 0 ^ 0
of the neutron from Pðr ; Ω ; E Þ to Pðr; Ω; EÞ; there is thus no dependence

on any state prior to the state Pðr ; Ω ; E0 Þ.
0 ^ 0 0
Since the interaction has no dependence on another state prior to P r ; Ω ; E , it is classified as a Markovian process.

7.1.2 Events in a random walk

As described in Chapter 1, a neutron source such as californium-252 can emit millions of neutrons per second. Each
neutron is thus born as a source neutron at a position r0 with a certain energy E0 at an angle Ω0 ; it undergoes
Nuclear Engineering. DOI: https://doi.org/10.1016/B978-0-323-90618-0.00007-7
© 2022 Elsevier Inc. All rights reserved. 305
306 Nuclear Engineering

FIGURE 7.1 A neutron interaction.

FIGURE 7.2 Random walks.

interactions in a random walk consisting of events. Two such random walks are shown in Fig. 7.2; in the first instance,
a source neutron undergoes scattering collisions at locations 1
5 until it is absorbed at position 6. The second random
walk shows a source neutron undergoing five interactions; the first four are scattering collisions with a change of angle
(and of course energy) while the fifth interaction is an absorption which terminates its random walk. An absorption can
be termed as the death of a neutron as it ends the random walk. In fact any event that takes a neutron out of the domain
of interest, such as a leakage across the physical boundary of a nuclear system, is called a death. A random walk is a
birth-to-death process with millions or several more such random walks considered in a simulation.
In Chapter 2, we discussed the several types of neutron interactions that are possible such as elastic scattering,
inelastic scattering, radiative capture, fission and many more.

7.1.3 The physics of interactions

In MC simulation, the physics of interactions is modeled with very elaborate methods as it is crucial to have accurate
knowledge of exactly what takes place in an interaction. As described in Section 2.2 (neutron interactions), radiative cap-
ture is understood to be a process in which a compound nucleus is formed for a very short while before the emission of a
gamma ray. The details of such an interaction are part of a MC simulation for the calculation of postcollision parameters.

7.1.4 Nuclear interaction data

Data files such as the evaluated nuclear data file (ENDF/B) (Brown et al., 2018) briefly described in Chapter 6 contain
elaborate neutron cross section data for Z 5 0 2 95 (neutron to Amercium-241), neutron reaction sublibrary with 557
materials and photon data for 163 materials mostly with evaluations up to 140 MeV. There are fifteen sublibraries with
photonuclear, photoatomic, radioactive decay, spontaneous fission yields and charged-particle cross sections. These
libraries are used for the charged particle and gamma transport MC simulations described in Chapter 1 and neutron
transport simulations described in Chapter 2.
ENDF/B files are part of the MCNP and SCALE packages but for other codes, group cross sections are prepared
with codes such as NJOY and WIMSD as described in Chapters 5 and 6 for diffusion and transport codes. In MC simu-
lation, the ENDF/B cross sections are used directly rather than going through the process of group-averaging.

7.1.5 How do we know an answer is good?

In MC simulation, quantities such as flux and current are estimated from an average of several random walks or histories.
Generally, the larger the number of histories simulated, the greater will be the confidence level of an estimate, as predicted
by the Central Limit Theorem. However, for a tally that is not well-behaved, such as in a rare event simulation, it is not
 The Monte Carlo method Chapter | 7 307

necessary that the confidence level will increase with the number of histories. It is thus important to understand how reli-
able an estimate is. Two statistical quantities play a fundamental role here namely the mean of an estimate x and its vari-
ance σ2 . Actually we know the sample mean and the sample variance rather than the population mean and the population
variance. The relationship between the sample mean and population mean is through the Law of Large Numbers, which
states that as the sample size N-N, the sample mean x tends to the true mean, or the population mean.
So that leaves us with the question: how is the sample variance σ2s related to the population variance σ2 ?
The sample variance of the mean σ2s ðx) is another quantity that is possible to calculate; this is expected to get smal-
ler as the sample size increases. Thus
σ2s
σ2s ðx Þ 5 ;
N
which means that we get continually better estimates of the sampled mean.
Our objective, to have as small as possible a value of σ2s ðx Þ, can be achieved if we reduce the sample variance or
increase the sample size. The first is possible by variance reduction techniques such as importance sampling in which
we bias a simulation to give favorable estimates. As an example, if a quantity is known to be reducing at an exponential
rate, then we would give more importance to lower values in a simulation. Thus instead of using uniformly distributed
random numbers, we would use exponentially distributed random numbers. A simple demonstration of such an impor-
tance sampling will be demonstrated for the evaluation of integrals in Section 7.5.
It is now important to define what we mean by the quality of an estimate. If we get a bad estimate with a small vari-
ance then the answer can be classified as inaccurate but precise. For example, if the true value of a quantity is 32, and
our estimate is 26 6 0:0001, then this estimate is inaccurate but in terms of standard deviation, it is a precise estimate.
Increasing the sample size N would be futile. On the other hand, an estimate of 31 6 2 is a better estimate as the mean
is closer to the true mean, while the standard deviation is much higher than that of the previous estimate. Here, it would
be reasonable to expect an improvement in the estimate by increasing the sample size.
In MC simulations, the underlying model uses physics as well as cross section data which are in-built and we have
no control on them. The dynamics of a neutron interaction that determine, for example, compound nucleus formation,
are based on quantum physics (Section 2.8). We thus determine Doppler broadening from a given model. The two fac-
tors in our control are the coding and modeling of a problem. Coding, especially in the case of large computer codes,
relies heavily on mathematical techniques used such as finite-difference, FEM, etc., to solve governing ODEs, PDEs
and integro-differential equations. The modeling of a problem is in the hands of a user; free choices include the use of
a particular method in preference to another, the use of a variance reduction scheme and the number of histories simu-
lated. These choices determine the precision of an estimate.
In Chapter 4, the Central Limit Theorem was used to show that the distribution of the estimated means x would be
approximately normal; the area under such a distribution would be B0.68 within one standard deviation and B0.95
within two standard deviations. These would serve as confidence levels for estimates obtained from a MC simulation. A
useful metric would then be a relative error defined as the ratio of the standard deviation of the mean divided by the mean
pffiffiffiffiffiffiffiffiffiffiffi
σ2s ðx Þ
R5 :
x
As described in MCNP Volume I, in the best estimate, R would be zero, while in a bad estimate R would be as large
as the estimated mean x. The relative error would thus lie in the range zero to one. In MCNP, the relative error is classi-
fied as generally reliable when R , 0:10 for a detector other than a point detector, and R , 0:05 for a point detector.
From the definition of the mean and standard deviation, the relative error is
2 31=2
PN
2
6 xi
17
6 7
R 5 6 i51 2 2 7
4 P N N5
xi
i51

which means that if a simulation is carried out for which each xi is the same xi 5 μ, then
2 31=2
PN
6 μ
2
17
1
6 7
R 5 6 i51 2 2 7 5 0:
4 PN N5
μ 1
i51
308 Nuclear Engineering

An easy example is the integral

ð1
I5 e2x dx
0

for which the PDF can be the uniform PDF f ðxÞ 5 1, and the estimator is gðxÞ 5 e2x (in which case xi 0 s are all differ-
ent) or alternately, in which the PDF is f ðxÞ 5 e2x and the estimator is gðxÞ 5 1 (in which case each history gives the
same estimate). This choice of a PDF-estimator combination gives a zero-variance result. So, even one sample would
be sufficient to give the correct answer.
When a zero-variance result is not foreseen, the value of the relative error squared is expected to decrease with an
increase of the sample size N while the computer time would increase more-or-less linearly with N. It would thus be
fair to expect that the quantity R2 T would remain constant; this quantity, in MCNP, is listed in the output as a Figure of
Merit, FOM 5 1=ðR2 T), from which the relative error is given as
1
R 5 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi :
FOM 3 T
Another useful indicator for the quality of the results given by the MCNP code, is the variance of the variance
(VoV) which is the estimated relative variance of the relative error R; a value less than 0.1 is taken to be indicative of a
good result. Again, the VoV is expected to decrease as the sample size N increases. For a much more rigorous discus-
sion the user is referred to the MCNP Volume I.

7.2 Simulation of a random walk

The integral equation “simulated” in MC for the collision density χ in a multigroup formulation with the collision and
transport operators as discussed in Chapter 6 is

 0
 0


χðgÞ ~ ^ t 5 C^ g -g ~
r ; Ω; r ;Ω ^ T^ g r 0 -r;Ω
^ 0 -Ω ^ 0 χðg0 Þ r 0 ; Ω
^ 0 ; t 1 SðgÞ (7.1)

where C^ and T^ are the collision and transport operators. The MC procedure consists of starting a “source” neutron,
transporting it to a collision site, processing the collision, making the appropriate tallies and continuing the simulation
until the neutron is lost by some terminating process. Many such neutrons are simulated, and at the end of the simula-
tion, tallies are averaged with both mean and variance computations. The MC process, which can be understood to be a
simulation of the integral formulation, is also known as a “Neumann series” solution as will be outlined in more detail
in the next section.

7.2.1 Monte Carlo simulation

In the MC method, a “large sample” of neutrons, representative of the total “population” is simulated to infer on quanti-
ties of interest using laws of probability. The MC approach allows modeling complexity of geometry and collision
physics and hence is not limited to idealizations of the sort that limit deterministic approaches. One of its’ drawbacks is
heavy computational effort; this can be addressed by powerful hardware and by efficiency-improving techniques which
result in considerable “speed-up.”
Fig. 7.3 shows the transport of neutrons in matter. Three starting neutrons, A, B, and C, are shown here though in
actual situations there can be a large number such as 1014 neutrons in a nuclear power reactor. Neutron A collides with
a host nucleus which undergoes the nuclear fission reaction resulting in the prompt emission of three “first generation”
neutrons A-1, A-2, and A-3 while the nucleus recoils to a new position. Neutron A-1 is captured by a host nucleus, while
A-2 collides with another nucleus which undergoes fission and emission of two neutrons A-2
1 and A-2
2. The third
neutron A-3 collides with a host nucleus and scatters “off” it, that is, the neutron transfers some of its’ energy to the
nucleus. Thus the number of first generation neutrons is three. Each of these neutrons continue to be transported and
collide with host nuclei undergoing one of a large possibilities of nuclear reactions, until they are either captured or
they escape from the physical domain by crossing the boundary. In a straightforward simulation of neutron transport,
each neutron is followed from its’ birth as a source neutron to its’ death when it is captured or has escaped. The birth-
to-death process is called a “history” and the straight paths in between interactions during a history is called a “random
walk” as it each interaction is determined randomly from one of several possible interactions. The random walk is “sto-
chastic” in nature, that is, it may not repeat itself, and it is Markovian in the sense that the next event is not related to
 The Monte Carlo method Chapter | 7 309

FIGURE 7.3 Neutron transport in matter.

the entire history of a neutron, just to its immediately preceding event. During a history, “tallies” are updated from
which useful information on the neutron flux and subsequent reaction rates can be obtained.
In the MC simulation of neutron transport, several such histories are simulated, and their tallies are processed using laws
of probability and statistics to infer on the averages of quantities obtained from the simulated sample. The usefulness of the
MC method is largely due to the high-speed computing possible with modern processors capable of gigaflop computations.
As a result, nuclear reactors, nuclear weapons and other nuclear systems are designed with great precision using MC
simulation. Since the 1940s MC applications in the Manhattan Project, the techniques have been adapted to a wide
range of areas in computational mathematics and physics, econometrics and forecasting, to biology and medicine.

7.2.2 Estimators and tallies

The fundamental quantity of interest is the flux φðr; Ω; E; tÞ which can be obtained in exact analytical form from the
transport equation for idealized conditions (Section 6.1) but in greater detail through numerical methods and simulation.
In general, this quantity is obtained by dividing the phase space into a number of bins in which tallies are scored. In
MC simulation, a result is obtained in a specified region of interest rather than over the whole domain as for diffusion
and transport calculations. Thus the bin tallies amount to a multiangle multigroup approach which, as we will see, is a
convenient approach for obtaining reliable estimates of the angular flux and from it, any reaction of interest.
Consider the three current and flux tallies (Werner, 2017)
ð ð ð ð
J 5 dt dE dΩ dAjΩ ∙ n^ j φðr; Ω; E; tÞ; (7.2)

the surface flux

ð ð ð ð
1
φS 5 dt dE dΩ dA φðr; Ω; E; tÞ; (7.3)
A
and the volume-averaged cell flux
ð ð ð ð
1
φV 5 dt dE dΩ dA φðr; Ω; E; tÞ: (7.4)
V
The track-length and collision estimators (CE) are used to score the scalar neutron flux φ  nv, where n is the num-
ber density (neutrons/cm3) and v is the neutron speed (cm/s). The flux, with units of neutrons/cm2/s, represents the rate
of neutrons crossing a unit area perpendicular to their direction of motion.
310 Nuclear Engineering

Typical tallies required in neutron/gamma transport calculations of nuclear fission reactors and systems are the
system multiplication keff , neutron and gamma fluxes as well as their distribution in space and associated reaction
rates such as radiative capture (gamma production), radiation dose, energy deposition. The flux loading determines
the effect of radiation on materials in the reactor such as structural materials, coolants, moderators, control and
instrumentation materials and wiring. The effect of radiation streaming through ducts and ports as well as the radia-
tion environment external to the reactor/nuclear systems is an integral part of all neutron/gamma transport
calculations.
For fusion reactors, such as ITER (International Thermonuclear Experimental Reactor), briefly described in
Section 3.10, and the European DEMO reactor (Demonstration Fusion Power Reactor) beyond ITER, the tallies of inter-
est are similar with little differences from nuclear fission reactors due to the differences in basic engineering design.
In fusion reactors, very high temperatures and their loading on the vacuum vessel first wall (FW) as well as beyond
the FW, the effect on superconducting coils, the tritium breeding in blanket are required in addition to the usual
neutron/gamma transport calculations in fission systems. Radiation damage to steel affects its life and operation; it is
calculated in terms of displacements per atom (dpa) per full power year (FPY).
Tallies for water in PWRs as well as in fusion reactor designs utilizing water as coolant, such as DEMO water-
cooled lithium lead, are similar as they are based on the low water density at high temperatures and pressures.
Two main differences between fission and fusion power reactors are the neutron energy which is B 1 MeV on the
average from fission neutrons while a D-T fusion reactor would be a powerful source of 14 MeV neutrons. The other
difference is that in a fusion reactor, the neutrons are born in the central vacuum chamber and stream around relatively
easily due to gaps between surrounding blanket modules and more ducts and channels.
All tallies are based on the neutron/gamma flux, that is, on the current, surface fluxes, and volume-averaged fluxes
[Eqs. (7.2)
(7.4)] and associated reaction rates Rx (Section 2.6) of the form
ð ð
Rx 5 C dt dEσx ðEÞ φðr; E; tÞ: (7.5)

where σx ðEÞ is a microscopic cross section for reaction type x (taken from a cross section library) and C is a normaliza-
tion constant.
These reaction rates are calculated from a simulation using tally multiplier input commands as will be described for
input files.
In Fig. 7.4 below, for a neutron history consisting of the four scattering events shown, the track lengths are
d1 ; d2 ; d3 ; d4 . For better efficiency, nonanalog MC uses a weight modification factor p 5 Σs =Σt , so that a history is con-
tinued after every collision with a reduced weight. The track-length estimator (TLE) for flux is

φTLE 5 nv 5 ðN:dÞ=ðV:tÞ 5 wd=V (7.6)

so that the flux estimate for this history is, for total starting weight W and volume V,
  4
1 X
φTLE 5 wi di (7.7)
WV i51

FIGURE 7.4 Random walk of a neutron.

 The Monte Carlo method Chapter | 7 311

TABLE 7.1 Estimators for the system multiplication keff .

Estimator Tally
2P 3
Collision estimator K
fk ν k Σf ;k
P
Ncol
6 7
1
N Wi 4 k51
P K 5
i51 fk Σt;k
k51

Absorption estimator (analog simulation) P

Nabs
Σ
1
N Wi ν k Σc;k 1f ;kΣf ;k
i51
Absorption estimator (nonanalog simulation) P
Nabs 0 Σ
1
N Wi ν k Σc;k 1f ;kΣf ;k
i51
Track-length estimator P
Npaths P
K
1
N Wi ρd fk ν k Σf ;k
i51 k51

The same quantity can be found using the collision estimator; in this case the estimate is made only at a collision
site and not on an escape event, unlike the TLE.
  4
1 X wi
φCE 5 (7.8)
WV i51 Σt;i

Finally, the surface flux Js , that is, the number of neutrons crossing the surface per unit area, can be estimated from
the weight crossing a surface divided by the total starting weight and the volume to get
 
1 X
Js 5 wi (7.9)
WV i

The summation is over all the neutrons that cross the surface.
There are three commonly estimators namely the collision estimator (CE), the absorption estimator (AE) and the
TLE with each having their own advantages such as when a large number of collisions are anticipated, the CE will give
a good answer, that is, bot accurate and precise. When fewer collisions are anticipated, then nonanalog simulation is
preferred since biasing a simulation, while preserving overall quantities, will give more tallies. Similarly, when the
medium is optically thin then the TLE is bound to give the best estimate as the number of collisions in both analog and
nonanalog simulations will be low or nonexistent while there will always be a track-length even in the absence of any
collision.
Four estimators for the system multiplication keff
CE A;as A;nas TLE
keff , keff , keff and keff are listed in Table 7.1.
In the collision estimator, the estimate is made at each collision with the actual (analog) particle weight while in the
AE the tally is updated at each absorption event. With implicit capture, a particle history is continued at an absorption
but with reduced weight (equal to the survival probability). The TLE, in contrast to both, updates a tally regardless of a
collision such as when it is crossing a region without undergoing a collision.

Exercise 7.1:

Assume a starting weight of one, K 5 2, (U-235 and U-238) with f5 5 0:2, f8 5 0:8, assume values for ν k and for
cross sections, calculate values for a history with three collisions.
There is no fixed number for simulation parameters, but typical reactor simulations require a few million histories in
over 100 cycles with thousands or hundreds of thousands of neutrons simulated per cycle for accurate results. Such a
calculation takes several hours on multiprocessors with several cores.

Exercise 7.2:

From Fig. 7.5, estimate the TLE flux and the collision estimator flux. Are they the same? Also estimate the surface
flux if line B represents the surface (assume unit surface area) (Table 7.2).
312 Nuclear Engineering

FIGURE 7.5 Forward scattering of neutrons in a slab.

TABLE 7.2 Tally estimators in MCNP.

Quantity Estimator Units

Current W Particles
Surface flux W
jWd
μjA
Particles/cm2
Cell flux Particles/cm2
Wp ðΩV
p Þe t
2Σ R
Point flux R2 Particles/cm2
ρ
Energy deposition m WdΣt H ðE Þ MeV/g
ρ
Heating m WdΣf Q MeV/g
Pulse tally W Weight accumulated in energy bins

7.2.3 Sampling a source

The source in a simulation is specified according to the problem of interest, which are considered here in three
categories:
1. fixed-source nonmultiplying medium,
2. criticality calculation in nuclear systems, and
3. mixed-mode neutron, photon, electron transport.
In the first category, the source would be specified in terms of its position energy, direction, and time. There could
be one or several sources at fixed points or distributed over a surface or within a volume. A fusion source, for example,
would have energy sampled from a fusion spectrum such as a Gaussian spectrum
h  2 i
pðEÞ 5 CE1=2 exp 2 ðE2bÞ=a

where a is the spread in MeV and b is the average neutron energy in MeV; for DT fusion at 10 keV, a 5 2 0:01 MeV
and b 5 2 1.
A fission energy spectrum, such the neutrons from a californium-252 source would be specified with a Watt
spectrum
  pffiffiffiffiffiffi
pðEÞ 5 Cexp 2 E=a sinh bE
where the parameters are given for the appropriate nuclide as shown for a few cases in Table 7.3.
In the second category, a criticality calculation starts by specifying a starting source at specified points or reading a
source from a previous simulation. The system multiplication is calculated over a number of specified cycles for a spec-
ified number of source histories N, with some cycles skipped for using better estimates for calculating means and var-
iances. The number of source histories is kept constant in each cycle by adjusting the weight of neutrons since one
cycle may produce 1000 neutrons while another may produce 1200 neutrons from fission. The procedure for calculating
keff is as dscribed in the previous section.
 The Monte Carlo method Chapter | 7 313

TABLE 7.3 Watt fission spectrum parameters.

Fission type Nuclide Energy (MeV) a b

Neutron induced U-235 Thermal 0.988 2.249
1 0.988 2.249
14 1.028 2.084
Pu-239 Thermal 0.966 2.842
1 0.966 2.842
14 1.055 2.383
Spontaneous Cf-252
1.025 2.926
Cm-244
0.906 3.848

In the third category, the simulation can be run as a neutron, photon, electron or as a mixed-mode simulation
accounting for photon produced from nuclear interactions and secondary electrons produced from photons as described
in Chapter 1.
MC simulations can handle a great variety of source specifications such as several point or distributed sources
defined by discrete probabilities in energy bins. The source(s) can be biased in angle and energy as the problem may
require. In shielding calculations, where very few tallies are scored in regions of interest, an MC simulation can be car-
ried out to write a surface source where a good tally is scored. In a subsequent run, the surface source can be used to
generate histories to score tallies in detector or regions of interest.
In this chapter, a fixed-source MC simulation is described. In a fission criticality problem, an initial fission source
distribution is specified and the next generation fission points are obtained. This gives a first estimate of the system
ð1Þ
multiplication keff ; these fission points are started in the second generation to give the next generation estimate and so
on until convergence is reached. The energy of the source neutrons is sampled from a fission spectrum
For a point isotropic source of neutrons, the angles of emission are sampled using random numbers (Section 4.9)
μi 5 2ξi 2 1; ϕi 5 2πξ i (7.10)
In Chapters 9 and 11, simulations will be carried out to demonstrate the versatility of MC simulation.

7.2.4 Sampling the “distance to collision”

The probability Y that a neutron at r1 , at time t1 , traveling with energy E1 in the direction θ1 ; ϕ1 has its next interaction
within dr 2 located at r2 at a later time t 1 dt is the product of two probabilities viz (1) the probability Y1 that it does not
have an interaction between r1 and r2 , and (2) the probability Y2 that, having reached r2 , it has its next collision in dr 2 . Thus
2
Y 5 Y1 Y2  L Yi 5 e2Σt jr2 2r1 j Σt dr2 (7.11)
i51

For a one-dimensional case, the pdf Y is written as

f ðxÞ 5 e2Σt x Σt dx (7.12)
and the cdf can be readily found as
ðs
f ðxÞdx 5 1 2 e2Σt x
F ð sÞ 5 (7.13)
0

  

distance
The   to collision is thus sampled as si 5 Σ1t ln 1 2 ξ i , which will have the same distribution as
si 5 Σ1t ln ξ i since ξi is uniformly distributed in (0,1) as will be 1 2 ξi .

7.2.5 Determining the type of event

At a collision site, determined by the mean free path, a MC simulation requires a determination of the type of event.
This is selected by sampling from a possible chain of events as shown in Fig. 7.6. A random number ξ is used to deter-
mine whether the event is an absorption or a scattering event. In case of an absorption, another random number is used
314 Nuclear Engineering

FIGURE 7.6 Chain of possible events.

to determine whether it is a fission or a capture. Similarly, in the case of a scattering event, a random number is used to
determine whether it is elastic or inelastic scattering.

7.2.6 Determining the nuclide of interaction

In a MC simulation which handles nuclides independently, it is required to determine the nuclide with which a neutron
collides at a collision site, for example, a neutron making a collision in water will collide with either a hydrogen atom
or an oxygen atom. This decision is made from the probability of an interaction with nuclide i at incident energy E
given as
Σt;i ðEÞ
pi ðEÞ 5 : (7.14)
Σt;5 ðEÞ 1 Σt;8 ðEÞ

7.2.7 Processing a scattering event

In an elastic scattering collision, a neutron changes its direction with conservation of kinetic energy and momentum;
there is no internal energy change or loss in friction.
The energy of a scattered neutron after collision E0 can be found in terms of its energy before collision E, its angle
of scattering in the laboratory system θ as
E h pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi i2
E0 5 cosθ1 A2 2sin2 θ (7.15)
ð11AÞ2
The energy of a 1 MeV neutron colliding with a U235 nucleus is shown as a function of the scattering angle is shown
in Fig. 7.7. It is seen that there is very little energy loss due to the high mass number ratio between the target and pro-
jectile. It is easy to understand that maximum energy loss will occur when the collision is between equal-mass particles;
thus a neutron colliding with a hydrogen atom, in water for example, may easily lose most of its energy which essen-
tially means that a “fast” (MeV) neutron can thermalize (B0.025 eV).

7.2.8 Processing a fission event

When a fission reaction (Section 2.9) takes place, two quantities needed are the number of prompt neutrons emitted and
their associated energies. The average number of prompt neutrons emitted in fission is given by ν 5 ν o 1 αE [Lamarsh
(p. 95)] where the constants are given in Table 7.4.
From experiments carried out, it has been determined that neutron multiplication is a stochastic process and there is
a probability distribution function assigned with ν (Schmidt & Jurado, 2018). For U235, the mean values measured were
2.41 for thermal fission, 2.49 for 0.5 MeV fission and 3.19 for 5.55 MeV fission.
The Gaussian distribution is used for the probability Pn, that n neutrons are emitted, given by
ð n11=2  
1 ðx2ν Þ2
Pn 5 pffiffiffiffiffiffiffiffiffiffi exp dx (7.16)
2πσ2 n21=2 2σ2

For the case n 5 0, the lower limit on the integral is replaced by 2N. Sampling is done from this normal distribu-
tion function using a rejection algorithm. The width of the distribution σ is taken as 1.088 for U235 and 1.116 for U238.
The mean ν is energy dependent: for U235 it is 2.4 1 0.12 E for E ,5 MeV and 2.2 1 0.16 E for E $ 5 MeV, while for
U238 it is 2.4 1 0.0666 E for E ,3 MeV and 2.1538 1 0.1654 E otherwise. The returned value for ν is the rounded inte-
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ger ν 5 ν 1 σx 2 2lnz=z.
 The Monte Carlo method Chapter | 7 315

FIGURE 7.7 Postcollision energy for elastic

scattering of a 1 MeV neutron-U235 nucleus.

TABLE 7.4 Neutrons emerging from fission in U235 and U238.

Isotope νo α (MeV21) E (MeV)

235
U 2.43 0.065 0,E,1
2.35 0.150 E.1
U238 2.30 0.160 All E

Another method is to estimate ν from the number of neutrons produced by fission νσf =σt and to round off the num-
ber to the nearest integer.
The energy of fission neutrons can be sampled from the empirical, Maxwell or Cranberg fission spectra (Section 2)
or others in the literature. The sampled values will be discussed in detail in a later chapter on MC simulation.

7.2.9 Processing a capture event

In analog simulation, a capture ends a history, as is done in the program written for this paper. However, a better way
is to continue the history with a reduced “weight” until it is ended due to very low weight or escape from the system.

7.2.10 Processing an escape-from-system event

When a neutron escapes from the system, it is not added to the updated number of neutrons generated and thus does not
contribute to the system multiplication.

7.2.11 Mean and variance

P
The mean and variance of an estimate (Section 4.8) is given as 5x 5 N1 Ni51 xi , and the variance of the population is esti-
P
mated from the sample as σ2 5 N1 Ni51 x2j 2 x 2 , from which the variance of the mean is estimated as σ52x 5 σN .
2
316 Nuclear Engineering

7.2.12 Batch, history, random walk and events

The procedure for neutrons follows the same sequence as for alpha, beta and gamma transport (Section 1.6) depicted in
the flow charts (Figs. 1.23, 1.24, and 1.27).
A simulation for a multiplying system begins with a batch of source neutrons N simulated for G generations for
each source neutron.
The pseudocode is described below:
i51
1. Begin source neutron i
2. Initialize n 5 1
3. Initialize neutron generation counter ng 5 0; k 5 n 5 1; j 5 1
4. Begin generation j
5. Begin source neutron k for this generation
6. Sample the distance to collision d
7. Sample the type of event
8. Process the event
9. Compute postcollision parameters
10. Determine if neutron still inside system
11. Update ng
12. Update k 5 k 1 1
13. If k , n go to v (continue the same generation)
14. Comes here when all “mother neutrons” for this gen processed Update n 5 ng ; k 5 k 1 1, for the next generation
15. Score for this source neutron i and generation j, Mði; jÞ5ng
16. If k , G 1 1 go to iv (begin a new generation)
17. Comes here when all generations processed
18. Update source neutron counter i 5 i 1 1
19. If i # N go to step I, begin a new source neutron
20. All neutrons have been processed
Estimate keff from the following steps: PN
a. Compute the number of neutrons produced in each generation: PðjÞ 5 Mði; jÞ, for generations j 5 1, 2, 3,. . .G
b. Compute the keff for each generation: i51

keff ðjÞ 5 PðjÞ=Pðj 2 1Þ;

j 5 2,3,. . .G, with keff ð1Þ 5 1
c. Compute the error for each keff:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 1
εðjÞ 5 keff ðjÞ 1 (7.17)
PðjÞ Pðj 2 1Þ

d. The estimate is

keff ðjÞ 1 εðjÞ (7.18)

7.3 Modeling the geometry

One of the strengths of the MC method for simulation of neutron transport in nuclear systems is its ability to model
complex geometry typical of practical systems containing a variety of materials in regular and irregular shapes as well
as ducting for instrumentation and biological shielding. An example of regular geometry is shown in Fig. 7.8 consisting
of four boxes generated from the MCNP input file listed below. Another combination of primitive configurations is
shown in Fig. 7.9.
 The Monte Carlo method Chapter | 7 317

FIGURE 7.8 Four boxes with sur-

face and cell numbers in MCNP.

FIGURE 7.9 Unions and intersections of volumes.

318 Nuclear Engineering

A convenient way of describing a volume is in terms of its enclosing surfaces. Fig. 7.10 shows a cylinder and a trun-
cated cone inside a sphere. The surface numbers from 10 to 16 represent the sphere of radius 8 cm centered at the ori-
gin, a cylinder of radius 3 cm on the z-axis within planes (surfaces 12 and 13) perpendicular to the z-axis, a cone with
its vertex on the z-axis at the point (0,0,4) within its outside sheet, truncated by two surfaces perpendicular to the z-axis
(surfaces 15 and 16). This way, large geometrical configurations can be modeled. Similarly Fig. 7.11 shows an XY view
of a cylinder inside a sphere. Better visualizations are possible, and especially helpful for configurations such as the
ITER design (Chapter 3) by the integration of CAD and MC input files.
 The Monte Carlo method Chapter | 7 319

FIGURE 7.10 A cylinder and trun-

cated cone within a sphere (YZ plane).

The above representations use constructive solid geometry to describe surfaces and regions through equations of
primitive surfaces and combinatorial geometry based on the Boolean operators of intersection and unions.
Surfaces are defined by mnemonics and associated parameters or by specifying points explicitly. Some of the sur-
faces used in MCNP are listed in Table 7.5.
A cell is a physical region in space defined by combinatorial geometry using the Boolean operators for union and
intersection. In Fig. 7.12, for two concentric spheres, the three cells are defined as:

In the above, the first cell is enclosed by surface 1, cell 2 is described as 1
2 which represents the intersection of a
region with a positive with respect to surface 1 and a negative sense with respect to surface 2.
The concentric spheres centered at the origin are represented by the equation

f ðx; y; zÞ  x2 1 y2 1 z2 2 R2 5 0
^ 5 Ωx i^ 1 Ωy j^ 1 Ωz k.
A ray is defined by its origin Ro 5 ðXo ; Yo ; Zo Þ and its direction (unit) vector Ω ^ The equation of a plane
is Ax 1 By 1 Cz 1 D 5 0. The normal vector N is ðA; B; C Þ. To find d, the distance from the origin of the ray to the plane
where it intersects, we first need to find the point of intersection. Along the ray, the point is: Xd 5 Xo 1 Ωx t; Yd 5 Yo 1 Ωy t,
Zd 5 Zo 1 Ωz t. The point Rd 5 ðXd ; Yd ; Zd Þ lies on the plane; therefore AXd 1 BYd 1 CZd 1 D 5 0. Thus the scalar t can be
320 Nuclear Engineering

FIGURE 7.11 A cylinder within

a sphere (XY plane).

found as
ðAXo 1 BYo 1 CZo 1 DÞ N ∙ Ro 1 D
t52 52
AΩx 1 BΩy 1 CΩz N ∙Ω^
as depicted in Fig. 7.13. If N ∙ R o 5 0, the ray and plane are parallel and there is no intersection; otherwise t is found
from
N ∙ Ro 1 D
t52 :
N ∙Ω^
For t , 0, there is no intersection with the plane.

Exercise 7.3:

^ 5 Ωx i^ 1 Ωy j^ 1 Ωz k^ where the orthogo-

In Fig. 7.14, consider the ray at Ro 5 ðXo ; Yo ; Zo Þ  ð2; 1; 3Þ in the direction Ω
 
nal angle is θ 5 60 and the azimuthal angle is ϕ 5 30 . Determine whether there is a point of intersection
Rd 5 ðXd ; Yd ; Zd Þ with the plane y 2 2 5 0, and if there is, then find the distance to surface from Ro .

7.3.1 Geometries for illustration of Monte Carlo simulation

Some representative and practically important models considered in this chapter are fixed-source benchmarks, criticality
benchmarks, a PWR unit lattice cell model and fusion reactor neutronics.
TABLE 7.5 Surface mnemonics for combinatorial geometry.

Mnemonic Type Description Equation Card entries

P Plane General Ax 1 By 1 Cz 2 D 5 0 ABCD
PX Normal to X 2 axis x 2D 50 D
PY Normal to Y 2 axis y 2D50 D
PZ Normal to Z 2 axis z 2D 50 D
SO Sphere Centered at Origin x 2 1 y 2 1 z 2 2 R2 5 0 R
 2
S General ðx 2x Þ2 1 y 2y 1 ðz 2z Þ2 2 R 2 5 0 x y zR
SX Centered on X 2 axis 2
ðx 2x Þ 1 y 1 z 2 R 5 0
2 2 2 xR
 2
SY Centered on Y 2 axis x 2 1 y 2y 1 z 2 2 R 2 5 0 yR
SZ Centered on Z 2 axis x 2 1 y 2 1 ðz2z Þ2 2 R 2 5 0 zR
 2
C/X Cylinder Parallel to X 2 axis y 2y 1 ðz 2z Þ2 2 R 2 5 0 y zR
C/Y Parallel to Y 2 axis ðx 2x Þ2 1 ðz 2z Þ2 2 R 2 5 0 x zR
 2
C/Z Parallel to Z 2 axis ðx 2x Þ2 1 y 2y 2 R 2 5 0 x yR
CX on X 2 axis y 2 1 z 2 2 R2 5 0 R
CY on Y 2 axis x 2 1 z 2 2 R2 5 0 R
CZ on Z 2 axis x 2 1 y 2 2 R2 5 0 R
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2
K/X Cone Parallel to X 2 axis y 2y 1 ðz2z Þ2 2 t ðx 2 x Þ 5 0 x y z t2 6 1
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  
Parallel to Y 2 axis ðx 2x Þ2 1 ðz 2z Þ2 2 t y 2 y 5 0 x y z t2 6 1
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2
K/Z Parallel to Z 2 axis ðx 2x Þ2 1 y 2y 2 t ðz 2 z Þ 5 0 x y z t2 6 1
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
KX on X 2 axis y 2 1 z 2 2 t ðx 2 x Þ 5 0 x t2 6 1
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  
KY on Y 2 axis x2 1 z2 2 t y 2 y 5 0 y t2 6 1
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
KZ on Z 2 axis x 2 1 y 2 2 t ðz 2 z Þ 5 0 z t 2 6 1 6 1 used only for
 2   one-sheet cone
SQ Ellipsoid Hyperboloid Axis parallel to X-, Y-, Aðx2x Þ2 1 B y 2y 1 C ðz2z Þ2 1 2D ðx 2 x Þ 1 2E y 2 y 1 2F ðz 2 z Þ 1 G 5 0 ABCDEFGx y z
Paraboloid or Z-axis
GQ Cyl, Cone Ell, Hyp, Axis not parallel to X-, Ax 2 1 By 2 1 Cz 2 1 Dxy 1 Eyz 1 Fzx 1 Gx 1 Hy 1 Jz 1 K 5 0 ABCDEFGHJK
Par Y-, or Z-axis
TX Elliptical or circular Axis parallel to X-, Y-, qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 ffi 2 x y z ABC
torus or Z-axis ðx 2x Þ2 =B 2 1 y 2y 1 ðz 2z Þ2 2A =C 2 5 1

TY  2 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 x y z ABC
y 2y =B 2 1 ðx 2x Þ2 1 ðz 2z Þ2 2A =C 2 5 1

TZ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2ffi 2 x y z ABC
ðz 2z Þ2 =B 2 1 ðx 2x Þ2 1 y 2y 2A =C 2 5 1
322 Nuclear Engineering

FIGURE 7.12 Concentric spheres.

FIGURE 7.13 Flowchart for a ray-plane intersection.

FIGURE 7.14 Ray-plane intersection.

 The Monte Carlo method Chapter | 7 323

FIGURE 7.15 Carbon-Beryllium

sphere (dimensions in cm).

Two simple fixed-source models, depicted in Fig. 7.15, considered are

1. a carbon bare sphere with a point isotropic source at its center, and
2. a carbon sphere reflected by a thin layer of beryllium.
The illustrative models for criticality benchmarks are bare spheres of the Godiva and Jezebel (Section 2.10,
Table 2.21 for uranium and plutonium bare and reflected critical systems and Table 2.22 for Bare critical
assemblies Godiva and Jezebel) assemblies. The MC simulation of critical assemblies is described in detail
in Chapter 9.
For nuclear power reactors (Section 3.2), nuclear propulsion reactors (Section 3.3), SMRs and space
reactors (Section 3.5), the overall design characteristics and core neutronics were described in Section 5.6 for neutron
diffusion models and in Section 6.4 for neutron transport models for the unit lattice cell, fuel assembly, and the reac-
tor core. The most basic unit for calculations in reactor core neutronics is the unit lattice cell for AP1000
(Section 3.2) shown in Fig. 7.16 represented by concentric cylinders in a bounding box with reflective surfaces.
The unit lattice cell is part of a fuel assembly and the reactor core as described for a PWR in Section 5.6. Its dimen-
sions are very small as shown in the figure; the cell shown above has dimensions of 1.26 3 1.26 3 2 cm. This repre-
sents a part of a typical PWR fuel assembly containing 17 3 17 such cells of dimensions 21.40202 3 21.40204 cm
with an active fuel height of 4.2 m.
For a Gen IV SMR (Section 3.5.3), Fig. 7.17 shows the core of the Toshiba 4S design (Koreshi & Hussain, 2014;
Ueda et al., 2005) with eighteen hexagonal fuel assemblies and a single central control rod. Note the dimensions of
the 4S core; the reflector outer radius is B 60 cm as compared to B3 m for the AP1000. These differences arise out
of different fuel enrichments and overall power requirements of both reactors.
324 Nuclear Engineering

FIGURE 7.16 AP1000 unit lat-

tice cell (dimensions in cm).

Fission neutronics give the following results

1. kN in a unit lattice cell and fuel assembly
2. keff for full-core
3. Neutron flux distribution (n/cm2/s)
4. Power density distribution (W/cm3)
5. Peak power (W)
6. Average power (W)
7. Control rod worth (pcm)
8. Equilibrium xenon worth (pcm)
9. Boron coefficient of reactivity with/without xenon (pcm/ppm)
10. Doppler coefficients with/without xenon (pcm/K),
11. Cold shutdown (coolant heat removal after a reactor has shut down) calculation.
Typical calculations for an AP1000 (de Stefani, Losada Moreira, Maiorino, & Russo Rossi, 2019) are shown in Table 7.6.
Power density distribution is illustrated in a plot showing the power density (W/cm3) in the fuel assemblies typically
in a one-fourth core model.
Boron concentration is shown as a function of burnup.
Coupled analyses are carried out by integrating neutronics, activation and thermal hydraulics as described in
Chapter 4 such as by combining MCNP (Werner, 2017), FISPACT (Sublet et al., 2017) and thermal hydraulics codes
(Yang, Liu, Xiong, Chai, & Cheng, 2018).
An example of a reactor smaller than the Toshiba 4S reactor is the micronuclear reactor (MNR) shown in Fig. 7.18
(Aziz, Koreshi, Sheikh, & Khan, 2020; Sun et al., 2018) with overall physical design characteristics described in
 The Monte Carlo method Chapter | 7 325

FIGURE 7.17 Core of the Gen

IV Toshiba 4S reactor (FA 5 fuel
assembly, CR 5 control rod,
dimensions in cm).

TABLE 7.6 Some quantities of interest in fission neutronics.

Quantity Result Comments

Analyses at cold zero power (CZP), hot
These are distinct states of a power reactor
zero power (HZP), hot full power (HFP)
kN for unit cell and FA R1:58 1:21029 6 0; 00003, Different fuel enrichments at CZP (defined Tf ; Tm
R2:35 1:32863 6 0; 00004, moderator specific mass and temperature in cross section
R3:20 1:40462 6 0; 00004, library)
R3:40 1:41697 6 0; 00004,
R4:45 1:46858 6 0; 00003,
Neutron flux (n/cm2/s) E $ 1:0MeV 1:37 3 1014 6 0:09 3 1014
5:53keV , E , 1:0MeV 1:93 3 1014 6 0:09 3 1014
0:625eV , E , 5:53keV 1:18 3 1014 6 0:13 3 1014
E # 0:625eV 5:01 3 1014 6 0:21 3 1014
Power density distribution Shown on a surface plot
Power peak factor 2.71
Peak power density 299.73 W/cm3
Average power density 110.6 W/cm3
Fuel depletion
Radioactive waste production Thermal output 2
20 kW/
m3

Note: Uranium consumption, and plutonium build-up are shown as functions of burnup; at B 1 MWd/gU, 1 GWd would consumeB 1 kgU.
326 Nuclear Engineering

FIGURE 7.18 Core of a micronuc-

lear reactor (FR 5 fuel rod, CD 5
control drum, dimensions in cm).

Section 3.8. Note again, the dimensions compared with the dimensions of the AP1000 and the 4S reactors. There are no
fuel assemblies in the MNR due to the higher fuel enrichments and lower power requirements.
The geometry models for fusion Tokamak systems (Section 3.6) can be more complicated than for fission systems.
ITER, for example is D-shaped torus with toroidal and poloidal field coils and considerable geometrical complexity of
its systems.
In a Tokamak, with high neutron flux B 1014 n/cm2/s causing high loadings and subsequent materials challenges,
some tallies of interest are
1. the FW loading (MW/m2) to determine the rate of energy transfer through the wall (flux tally),
2. neutron (fast, E. 0.1 MeV) fluence to toroidal field coils (TFC),
3. nuclear heating (W/m3); over 80% of the DT energy is carried by a 14 MeV neutron and is deposited in the breeder
blanket; the rest (B16%) in the FW, and tungsten armor, manifolds and surrounding structure,
4. the tritium breeding rate (flux multiplier tally)
01n 1 36Li-24He 1 13T 1 4:8MeV
01n 1 37Li-24He 1 13T 1 01n 2 2:5MeV
The TBR is calculated for candidate materials such as liquid metal, eutectics such as lithium leads, solids such
as lithium oxide and ternary oxides (e.g., LiAlO2, Li4SiO4) and ceramics such as lithium titanate; design calcula-
tions are performed to optimize the lithium enrichment, mix of neutron multiplier such as beryllium and lead, cool-
ant such as water, helium and molten salt, neutron reflector and structural material (steel),
5. helium production in steel structure and in reweldable zones; flux multiplier tally ðn; αÞ reaction; calculated in
ppm/FPY (atomic parts per million per FPY),
6. radiation damage (dpa) to copper (flux multiplier tally) and steel; since this degrades the material reducing its life;
MC simulations thus give the dpa/FPY (displacements per atom per FPY),
 The Monte Carlo method Chapter | 7 327

7. radiation streaming from plasma to superconducting magnet TFC,

8. radiation dose (Gy) to epoxy in the winding pack facing the central cell,
9. radiation dose with reactor shutdown to permit maintenance,
10. radiation across ducts and shielding, and
11. activation of steel (effect of low activation materials in steel such as Eurofer), coolants and components to mini-
mize activation and radioactive waste due to B1021 neutrons per second during a power pulses.
The pressure and temperature operating conditions of tokamaks are of the same order as in PWRs, that is,
15.5 MPa, B280 C; thus water density is significantly reduced.
Table 7.7 lists some estimates for an engineering study of the water-cooled lithium lead 1998 MW DEMO fusion
reactor with net electrical power 500 MW (Moro et al., 2020) with 16 TFCs and a minor and major radius of 2.883 and
8.938 m, respectively, using mainly tungsten armor as the FW, LiPb, H2O, and Eurofer steel as the breeder, and LiPb
and H2O manifolds.
Neutron studies lead to optimal designs, such as in the study quoted above (Moro et al., 2020), a coupled
neutronics-thermal hydraulics analysis showed that FW cooling ws achievable by reducing the number of water chan-
nels which subsequently led to an increase in the Eurofer steel content and a resulting TBR increase to 1.138. The rea-
son for such an increase is that water acts both as a moderator enhancing the Li6 reaction and as a shield reducing the
intensity of radiation.
Thus breeder design is an active optimization research area with the objective of enhancing tritium breeding from
the above reactions of lithium (Del Nevo et al., 2019; Hernández & Pereslavtsev, 2018; Moro et al., 2020).
A simplified 1-D representative geometrical representation of a breeder blanket (Colling, 2016; Shimwell et al.,
2016) of a spherical Tokamak reactor, is shown in Fig. 7.19.
A mockup experiment for tritium breeding in ITER is depicted in Fig. 7.20 (Jakhar et al., 2015; Mandal, Shenoi, &
Ghosh, 2010) consists of a central D-T fusion source surrounded by lead (Pb) multiplier and reflected by high density
polyethylene (HDPE) reflector.
When a geometrical model is made, the next step in MC simulation is the description of the physics of interactions,
the material descriptions, the neutron/gamma source, tallies and simulation parameters.

TABLE 7.7 Some quantities of interest in fusion neutronics.

Quantity Desired/limits Objective

2
Average wall loading 1.04 MW/m
TBR (desired) .1.10 Highest TPR B 1.07 3 1012
tritium atoms/cm3/s
TBR (assessed) 1.119 With reduced water density
(ρ 5 0:725g=cm3, TBR assessed is
1.118
Damage on vacuum vessel steel , 2.75 dpa to ensure fracture toughness is
over 6 FPR (lifetime) reduced by no more than 30%)
Assessed radiation damage (I/B) Eurofer FW B 8.1 dpa/FPY; SS316L VV inner

shellB8.67 3 1023 dpa/FPY
Assessed Radiation damage (O/B) Eurofer FW B 9.5 dpa/FPY; SS316L VV inner

shellB2.46 3 1023 dpa/FPY
He concentration , 1 ppm To permit welding in reweldable
zones
He concentration in FW I/B Eurofer in breeder blanket 94 ppm/FPY, VV inner shell over a period of 6 FPY, 0.48 ppm
(assessed) 8.08 3 1022 ppm/FPY, TFC 1.21 3 1026 dpa/FPY
He concentration in FW O/B Eurofer in breeder blanket 114 ppm/FPY, VV inner shell over a period of 6 FPY, 0.24 ppm
(assessed) 2.43 3 1022 ppm/FPY, TFC 1.21 3 1026 dpa/FPY
Fast neutron fluence on TFC , 109 n/cm2/s shielding capability should ensure
a fluence within the limits
Nuclear heating on winding pack , 50 W/m3

Total neutron (inboard) flux 5.13 3 1014 n/cm2/s

Power deposition W armor: B 3%, FWB16%, Breeder zone B 75%,

ManifoldsB3%
Heat load (O/B) W armor B 28.4 W/cm3, FW EuroferB 9.59 W/cm3

328 Nuclear Engineering

FIGURE 7.19 A simplified spherical model of a fusion reactor blanket for neutronics; vac-
uum, purple 5 breeder, blue 5 helium, yellow 5 beryllium multiplier, green 5 EUROFER steel
structure.

FIGURE 7.20 TBR ITER Mockup

Experiment (dimensions in cm).

7.4 Demonstration
This section takes the reader through a fixed-source problem to introduce an input and output of a MC neutron transport
simulation. Multiplying systems for simple critical assemblies and reactor lattices and nuclear reactor core calculations
are described in chapters nine and eleven respectively.
Consider the carbon-beryllium sphere configuration shown in Fig. 7.15. There are three spheres in the model; the first
sphere of radius 0.02 cm contains a point isotropic source of energy 1 MeV located at the origin. For such an idealized prob-
lem, exact analytical diffusion and transport solutions were described in Chapters 4 and 5. It is a good practice to learn MC
simulation by first carrying out simulations for simple problems for which exact solutions are known. A time-of-flight
benchmark for a carbon sphere is a useful starting point for comparing MC simulation results with experimental results.
 The Monte Carlo method Chapter | 7 329

Some benchmarks will be considered for critical systems in Chapter 9.

The MCNP input file (X-5 Monte Carlo Team, 2008) consists of three blocks separated by a blank line between
blocks at the end of the third block. The first line of MCNP describes the simulation. In Block 1, the cells (volume
regions) are defined in terms of their bounding surfaces and the material present in them.
In this example, Block 1 has four lines since there are four cells; source, carbon, beryllium, and the outside world.
The four lines, after the first description line, define the cell number followed by the material number, material den-
sity (negative sign meaning a gram density and positive meaning an atomic density) followed by the sense of the cell
with respect to the surface number(s) given.
There are three surfaces defined in the second block, each is a sphere with its center at the origin; surfaces 1, 2 and
3 represent spheres of radii 0.02, 4.187, and 5.187 cm, respectively.
The first cell is defined to be within the region bounded by surface 1, that is, every point in cell 1 has a negative
sense with respect to surface 1. Mathematically, this condition is stated as f ðx; yÞ 5 x2 1 y2 1 z2 2 ð0:01Þ2 , 0 for a point
x; y in cell 1 which contains material 1 with atomic density 0.001288 g/cm3; the material entry is given in Block 3 of
the input file. Similarly cells 2 and 3 are defined each between their two bounding surfaces. The last cell, cell 4, the
outside world is everything beyond surface 3. A particle is considered to have left the system when it enters into the
outside world and can not rerenter the system.
MCNP input file for a carbon-beryllium configuration
330 Nuclear Engineering

The comment lines added to Block 3 describe the input lines for the mode of the simulation, the cell importance,
the material descriptions, the bin structures for energy and angular distributions, tallies, tally multipliers, source descrip-
tion and simulation parameters.
The tally multiplier

calculates quantities of the form described by Eq. (7.5). The FM24 follows a tally F24 which is a flux tally of type
4 (each succeeding tally is defined distinctly such as F4, F14, F24, F34,. . .). The form of this tally multiplier is

where the constant C is a multiplier, M is a material number and R1,. . .R4 are Reaction Numbers defined in the ENDF/B
files.
In the present case, C is the atomic density of carbon multiplied by the volume of the carbon region and the
Reaction Numbers are -1 (total cross section), -2 (absorption scattering cross section), -3 (elastic scattering cross sec-
tion) and 102 (radiative capture cross section). In this demonstration only four reaction rates were specified; this
depends on the user’s requirements (limits are given in the MCNP manual).
Following an overall summary of the interactions and gains and losses, the required tallies are listed as shown in
Table 7.8 for three simulations with 100, 104 and 106 histories.
The neutron current tally (F1) remains 0.99999 (0.0000) for all three simulations since the absorption is very small
as shown in Table 7.8. The current is only in the forward cosine bin, that is, in the range ð0; π=2Þ, in one energy bin
(1 eV
1 MeV). This tally tells us that all postcollision energies lie in the range 1 eV
1 MeV; this range could have
been made smaller to get finer detail.
Table 7.8 lists seven tallies per source neutron per second (surface flux, cell flux, point flux total and uncollided,
total reactions, absorptions, elastic scatterings and radiative captures).
The results show that
1. For each tally, the quality of the results gets progressively better with the number of histories; this is due to the large
number of collisions in the carbon sphere.
 The Monte Carlo method Chapter | 7 331

2. Relative error decreases by an order of magnitude as the simulations increase by two orders of magnitude.
3. The variance of the variance (VoV) also decreases in the same way as the relative error.
4. The (total) point flux improves with the number of histories while the uncollided point flux has zero relative error
as it is a deterministic quantity.
5. Reactions are predominantly elastic scattering reactions; absorption is due to radiative capture alone. Since the absorptions
include the ðn; pÞ, ðn; tÞ, ðn; dÞ, ðn; He3 Þ and ðn; γÞ reactions, it is observed that absorption is only to the ðn; γÞ reactions.
Finally, the energy-dependent neutron flux, shown in Fig. 7.21 for one million histories, indicates that there is very
little spectrum degradation due to the absence of any low-Z hydrogenous materials.
Such experiments, important as they are for an understanding of the simulation, will be left as exercises for the MC user.

TABLE 7.8 Tallies for a carbon sphere fixed-source Monte Carlo simulation.

Tally N 102 104 106

Surface flux (F2) 5:81407 3 1023 5:4334 3 1023 5:39525 3 1023
Relative error 0:0824 0:0055 0:0005
VoV 0:4931 0:0409 0:0004
Cell flux (F4) 1:74752 3 1022 1:80857 3 1022 1:80573 3 1022
Relative error 0:0329 0:0044 0:0004
VoV 0:0587 0:0018 0
Point flux (F5) 1:02127 3 1022 5:40671 3 1023 5:39618 3 1023
Relative error 0:3408 0:0428 0:0045
Point flux (F5) uncollided 1:81729 3 1023 1:81729 3 1023 1:81729 3 1023
Relative error 0 0 0
VoV 0:5535 0:1618 0:0028
Reactions/cm3
Total reactions 4:13757 3 1023 4:231955 3 1023 4:30813 3 1023
Relative error 0:0414 0:0056 0:0006
VoV 0:0748 0:0026 0
Absorptions 1:80589 3 1029 1:85912 3 1029 1:85759 3 1029
Relative error 0:0271 0:0036 0:0004
VoV 0:0748 0:0026 0
Elastic 4:13756 3 1023 4:31955 3 1023 4:30813 3 1023
Relative error 0:0414 0:0056 0:0006
VoV 0:0748 0:0026 0
Radiative capture 1:80589 3 1029 1:85912 3 1029 1:85759 3 1029
Relative error 0:0271 0:0036 0:0004
VoV 0:0748 0:0026 0

FIGURE 7.21 Neutron flux in a carbon sphere.

332 Nuclear Engineering

7.5 Variance reduction methods

In the previous section, a natural or straightforward MC simulation, called analog simulation, was carried out since it
was anticipated, and demonstrated by simulation, that a large number of collisions took place in the carbon sphere.
Subsequently, the results were both accurate and precise.
In nuclear fission reactor calculations, this is usually the case within the core where neutrons and gamma interac-
tions take place frequently. However, as we go far from the source of activity the flux falls by orders of magnitude.
By design, radiation shielding has to be thick enough to attenuate the radiation so that very few neutrons or gammas
are able to penetrate its thick layers. If, for example, one million neutrons are started as source neutrons and only two
neutrons get through the shield to contribute to the tally of interest, then it will be an unreliable tally.
The validity of the Law of Large Numbers or the Central Limit Theorem would then become questionable and it
would not make sense to compute a sample estimate or a variance.
As we saw in y7:1:5, there are two options for getting a better estimate; the first is to increase the histories to be simu-
lated. An increase of 100 would reduce the standard deviation by 10. But simulating 100 million histories instead of 1 mil-
lion would requires hours and hours of computer time, with no guarantees that a good enough tally would be possible. The
second option is to force things to happen (bias the simulation) in such a way that more histories contribute to the tally.
This approach, of modifying the simulation in an unrealistic way would not be correct unless the overall weight is con-
served. Thus in a small solid angle of interest, if there would have been 10 particles by geometrical considerations alone,
then we can force 50 particles to travel toward the detector but with a weight reduced by a factor of five. Such a strategy,
called a nonanalog simulation has several such variance reduction methods which help in obtaining reliable estimates.
Some methods are used in analog simulations to reduce the computational effort on particles which are likely to
have a small contribution to a tally. For example, when the weight of a particle, after several interactions, has reduced
below a small number such as 1.0e-6, then Russian Roulette is played to determine whether to let it continue, with
increased weight. Say if the weight falls below the prescribed cutoff and it is given a small probability of survival, then
a random number is generated to decide its fate. The probability that the random number would favor survival is also
very low so in most cases, the particle would be terminated. However, there is small chance that it survives, in that case
its weight is increased by the ratio of survival probability to nonsurvival probability and it continues. Conversely, if its
weight becomes too large, then it is split into many particles of reduced weight. This process ensures that particles
remain within a favorable weight window. Codes such as MCNP have a weight window generator which suggests best
weights to specify. This is an art as well as a science as weight can be intuitively assigned as well.
Such splitting techniques can be carried out on the basis of space, energy and angle. When some regions are more
important than others it is desirable to invest more time in simulating them as there is reason to believe that they will
contribute toward the quality of a tally.
Similarly, when a particle moves into an important region, it is split into more particles, while conserving the overall
weight. Conversely, when a particle moves into a region of less importance, a decision has to be made whether to let it
continue at the expense of computational effort and time, or to terminate its history. This decision is based on a statisti-
cal process known as Russian Roulette. A random number is generated and with the prescribed probability of termina-
tion, the appropriate decision is made.
Geometry importances are assigned in a data file at the beginning of a MC simulation. If region A has importance one
and region B has importance two, then the particle will be split into twice as many particles. In practice it is advisable to
manually input cell importance and inspect the results to see whether they were set properly; this remains an art more than a
science and is bound to rely heavily on intuition. A formal method is to run an adjoint problem but a code such as MCNP,
which is regarded as industry-standard now, does not have this option in the “continuous energy” simulation mode.
For shielding problems where variance reduction can be necessary and most useful, a step-wise approach can be
used with several runs, starting from an initial guess of cell importance and using generated weight windows progres-
sively in regions closer to the shield.
Some more nonanalog techniques are exponential transform, forced collisions and source biasing where a probability
distribution function is modified, and accounted for later, to give a large number of nonzero contributions from histories.
All these techniques can result in improved tallies, but judgment based mostly on intuition is the guide.
Importance biasing has mathematical foundations in the adjoint equation which is also called the backward equation
as it represents a detector-to-source formalism rather than one based on the conventional source-to-detector formalism.
The adjoint flux or “importance” function has then represents the importance of particles in phase space; these have
also been called contributions. This amounts to knowing the solution and biasing it in such a way that only those parti-
cles are simulated that contribute to the tally.
 The Monte Carlo method Chapter | 7 333

A simple example of knowing the solution to carry out importance sampling is the integral evaluation deterministic
problem described in Chapter 4. An altered estimator is used so that
ðb ðb
f ðxÞ
I5 gðxÞdx 5 f =g dx 5 f =g ðb 2 aÞ
a gðxÞ
 a

estimates f ðxÞ=gðxÞ  hðx while the PDF is gðxÞ. It can be shown that by a “suitable” choice of the “importance
function” gðxÞ this process leads to a “zero-variance”
Ð1 solution.
Consider the estimation of the integral I 5 0 e2x dx, first with the estimator f ðxÞ  e2x with sampling from the uni-
form PDF, and then with the estimator hðxÞ 5 1, and of course the PDF normalization factor ð1 2 e21 Þ outside the inte-
gral, with random number samped from the PDF
e2x
gðxÞ 5
1 2 e21
Clearly, this leads to
ð1
I5 gðxÞdx
0

with a precise (zero-variance) answer of ð1 2 e21 Þ.

While it was easy to demonstrate a “zero-variance” result for the simple problem considered above, it is not always
easy to “know” a priori the importance function gðxÞ since the “best” gðxÞ is the answer itself which we want to find.
Variance reduction based on intuition, or modified sampling or with the help of the adjoint equation has been found to
be necessary in nuclear engineering for shielding and streaming problems to determine, with reasonable accuracy, the
shutdown activation for example. In Chapter 9, variance reduction will be demonstrated for a representative problem.

7.6 Estimating perturbations with Monte Carlo simulation

In the design and operation of nuclear reactors and systems, it is often important to be able to estimate the effect of
small material density perturbations arising out of the burnup of fuel or the insertion of control rods. Additionally, it
may be important to estimate the effect of geometry perturbations due to moving components.
A drawback of MC methods is that the uncertainty of an estimate may be of the same order of magnitude as the per-
turbation itself and thus the effect of the design or operational change may become masked. This implies that the differ-
ence of two MC simulations may be inadequate to estimate perturbations. Thus other methods had been developed
initially along the traditional route of deterministic formulations based on the adjoint function, or Lagrange multiplier,
as a sensitivity coefficient. Such analyses have been used extensively to estimate perturbations from two MC runs viz a
forward run and an adjoint run from codes such as MCNP and MORSE. New developments in this area came with “cor-
related tracking” and “derivative sampling” techniques (Rief, 1984) and implemented in KENO and TIMOC codes. The
techniques were extended by Watkins for material perturbations in a research reactor and demonstrated for two-group
transport (Koreshi & Lewins, 1990) and later extended to fusion reactor neutronics optimization studies. The perturba-
tion feature is now part of several MC codes and is used extensively. Comparisons of the adjoint versus derivative
approach have been discussed by Kiedrowski & Brown (2011). Applications of MC perturbation theory are given in
Chapter 11 for estimating changes in keff due to material density pertrubations in a bare sphere such as Godiva. Such
studies have also been carried out to quantify the sensitivity of estimates to uncertainties in cross section data. With rea-
sonable confidence for small systems, perturbation can be carried out for design optimization in bigger simulations thus
saving computational effort.

7.7 Conclusions
In this chapter, the fundamentals of the Monte Carlo (MC) method used for particle transport were presented. The use
of random numbers, event probabilites, and probability distribution functions, given in Chapter 4, was described
for simulating a random walk and tallying quantities using the collision, absorption, and track length estimators.
The simulation for an elementary fixed-source problem was used to introduce the reader to prepare an input for carrying
out a neutron transport simulation giving basic tallies and interpreting the estimated quantities.
334 Nuclear Engineering

Problems
1. Use the acceptance-rejection technique to sample from an exponential distribution and compare its efficiency by a
straightforward analytical inversion.
2. Use the same acceptance-rejection technique to braw a bounding box around a normal distribution to get samples.
3. For Exercise 7.2, write a program to estimate neutron flux in a carbon slab of thickness 3 mfp with thermal cross
sections.
4. Repeat the above for isotropic scattering.
5. Modify your program to use the collision estimator for flux rather than the TLE.
6. In the AP1000 unit lattice cell, using the models for alpha and beta transport (Chapter 1), what would be the signif-
icance of both radiations on the energy deposition and on the coolant activation?
7. Based on the topics covered in Chapter 4, how would the neutron fluxes calculated from a MC simulation be used
to calculate the source term for spent nuclear fuel from a nuclear reactor?
8. With reference to quantities of interest in fission neutronics, list a few activation reactions in constituent elements
of steel that could pose a long-term storage problem. How would these quantities be calculated using cross section
fields?
9. With reference to TableÐ7.7 comment on the effect of water reduction leading to an increase in TBR.
10. In Section 7.5, for I 5 0 e2x dx, show that a zero-variance solution is obtained for an exponential PDF. Compare
1

the estimate with a PDF representing the first quadrant of a circle.

Nomenclature
Greek lower case
ρ density
θ orthogonal angle
μ cosine of orthogonal angle
μ mean value
νk average number of fission neutrons emerging from a fission
σ standard deviation
σx standard deviation of x
τ time constant
ϕ azimuthal angle
φS surface-averaged scalar flux
φV volume-averaged scalar flux
χ collision density

Greek upper case

Σt macroscopic total cross section
Ω solid angle

English lower case

d distance
fk fraction of nuclide k
keff system multiplication
m mass
n atomic density
p resonance escape probability
p probability
s distance
ξ random number
w weight

English upper case

AE absorption estimator
C^ collision kernel
 The Monte Carlo method Chapter | 7 335

Cf californium-252
Cm curium-244
HðEÞ heating value as a function of energy
J current
N number of histories
Q Q-value of a reaction
R relative error
S source
T^ transport kernel
Tf fuel temperature
Tm moderator temperature
ppm atomic parts per million
dpa displacements per atom
pcm percent mille (1025)

Abbreviations and Acronyms

CE collision estimator
CZP cold zero power
FOM figure of merit
FPY full power years
FW first wall
HFP hot full power
HZP hot zero power
I/B inboard
ENDF evaluated nuclear data files
O/B out-board
TBR tritium breeding ratio
TFC toroidal field coil
TLE track-length estimator
TPR tritium production rate
VV vacuum vessel
MC Monte Carlo

References
Anderson, H.L. (1986). Metropolis, Monte Carlo, and the MANIAC. Los Alamos Science.
Aziz, U., Koreshi, Z. U., Sheikh, S. R., & Khan, H. (2020). A study of criticality and thermal loading in a conceptual micronuclear heat pipe reactor
for space applications. Nuclear Technology and Radiation Protection. Available from https://doi.org/10.2298/NTRP2003208A.
Brown, D. A., Chadwick, M. B., Capote, R., Kahler, A. C., Trkov, A., Herman, M. W., & Zhu, Y. (2018). ENDF/B-VIII.0: The 8th major release of
the nuclear reaction data library with CIELO-project cross sections, New Standards and Thermal Scattering Data. Nuclear Data Sheets. Available
from https://doi.org/10.1016/j.nds.2018.02.001.
Colling, B. R. (2016). Blanket Performance and Radioactive Waste of Fusion Reactors: A Neutronics Approach. England: Lancaster University.
Dunn, W. L., & Shultis, J. K. (2012). Markov chain Monte Carlo. Exploring Monte Carlo Methods. Available from https://doi.org/10.1016/b978-0-
444-51575-9.00006-3.
Hammersley, J.M. & Handscomb, D.C. (1964). Monte Carlo methods. ,https://doi.org/10.1007/978-94-009-5819-7.
Hammersley, J. M., Handscomb, D. C., & Weiss, G. (1965). Monte Carlo methods. Physics Today. Available from https://doi.org/10.1063/1.3047186.
Hernández, F. A., & Pereslavtsev, P. (2018). First principles review of options for tritium breeder and neutron multiplier materials for breeding blan-
kets in fusion reactors. Fusion Engineering and Design. Available from https://doi.org/10.1016/j.fusengdes.2018.09.014.
Jakhar, S., Abhangi, M., Tiwari, S., Makwana, R., Chaudhari, V., Swami, H. L., & Chaudhuri, P. (2015). Tritium breeding mock-up experiments con-
taining lithium titanate ceramic pebbles and lead irradiated with DT neutrons. Fusion Engineering and Design. Available from https://doi.org/
10.1016/j.fusengdes.2015.04.003.
Kalos, M. H., & Whitlock, P. A. (2008). Monte Carlo methods. Wiley-Blackwell.
Keys, D. A., & Groves, L. R. (1963). Now it can be told. The story of the Manhattan project. International Journal. Available from https://doi.org/
10.2307/40198796.
336 Nuclear Engineering

Kiedrowski, B. C. ,, & Brown, F. B. (2011). Comparison of the Monte Carlo adjoint-weighted and differential operator perturbation methods.
Progress in Nuclear Science and Technology. Available from https://doi.org/10.15669/pnst.2.836.
Koreshi, Z. U., & Hussain, S. Z. (2014). A neutronic pin cell and full-core design analysis of the 4 S reactor. Annals of Nuclear Energy, 72, 145
150.
Available from https://doi.org/10.1016/j.anucene.2014.05.015.
Koreshi, Z. U., & Lewins, J. D. (1990). Two-group Monte Carlo perturbation theory and applications in fixed-source problems. Progress in Nuclear
Energy. Available from https://doi.org/10.1016/0149-1970(90)90020-6.
Mandal, D., Shenoi, M. R. K., & Ghosh, S. K. (2010). Synthesis & fabrication of lithium-titanate pebbles for ITER breeding blanket by solid state
reaction & spherodization. Fusion Engineering and Design. Available from https://doi.org/10.1016/j.fusengdes.2010.06.018.
McClarren, R. G. (2018). Introduction to Monte Carlo methods. Computational Nuclear Engineering and Radiological Science Using Python.
Available from https://doi.org/10.1016/b978-0-12-812253-2.00024-8.
Metropolis, N., & Ulam, S. (1949). The Monte Carlo method. Journal of the American Statistical Association. Available from https://doi.org/10.1080/
01621459.1949.10483310.
Moro, F., Colangeli, A., Del Nevo, A., Flammini, D., Mariano, G., Martelli, E., & Villari, R. (2020). Nuclear analysis of the water cooled lithium lead
DEMO reactor. Fusion Engineering and Design. Available from https://doi.org/10.1016/j.fusengdes.2020.111833.
Del Nevo, A., Arena, P., Caruso, G., Chiovaro, P., Di Maio, P. A., Eboli, M., & Martelli, E. (2019). Recent progress in developing a feasible and inte-
grated conceptual design of the WCLL BB in EUROfusion project. Fusion Engineering and Design. Available from https://doi.org/10.1016/j.
fusengdes.2019.03.040.
Rief, H. (1984). Generalized Monte Carlo perturbation algorithms for correlated sampling and a second-order Taylor series approach. Annals of
Nuclear Energy. Available from https://doi.org/10.1016/0306-4549(84)90064-1.
Schmidt, K. H., & Jurado, B. (2018). Review on the progress in nuclear fission - Experimental methods and theoretical descriptions. Reports on
Progress in Physics. Available from https://doi.org/10.1088/1361-6633/aacfa7.
Shimwell, J., Kovari, M., Lilley, S., Zheng, S., Morgan, L. W. G., Packer, L. W., & McMillan, J. (2016). A parameter study of time-varying tritium
production in solid-type breeder blankets. Fusion Engineering and Design. Available from https://doi.org/10.1016/j.fusengdes.2016.01.011.
Spanier, J., Gelbard, E. M., & Bell, G. (1970). Monte Carlo principles and neutron transport problems. Physics Today. Available from https://doi.org/
10.1063/1.3022338.
de Stefani, G. L., Losada Moreira, J. M., Maiorino, J. R., & Russo Rossi, P. C. (2019). Detailed neutronic calculations of the AP1000 reactor core
with the Serpent code. Progress in Nuclear Energy. Available from https://doi.org/10.1016/j.pnucene.2019.03.030.
Sublet, J. C., Eastwood, J. W., Morgan, J. G., Gilbert, M. R., Fleming, M., & Arter, W. (2017). FISPACT-II: An advanced simulation system for acti-
vation, transmutation and material modelling. Nuclear Data Sheets. Available from https://doi.org/10.1016/j.nds.2017.01.002.
Sun, H., Wang, C., Ma, P., Liu, X., Tian, W., Qiu, S., & Su, G. (2018). Comnceptual design and analysis of a multipurpose micro nuclear reactor
power source. Annals of Nuclear Energy, 121, 118
127.
Ueda, N., Kinoshita, I., Minato, A., Kasai, S., Yokoyama, T., & Maruyama, S. (2005). Sodium cooled small fast long-life reactor “4 S.”, . Progress in
Nuclear Energy (47, pp. 222
230). . Available from https://doi.org/10.1016/j.pnucene.2005.05.022.
Werner, C. (ed.) (2017). MCNP (R) User’s Manual Code Version 6.2.
X-5 Monte Carlo Team. (2008). MCNP - A general Monte Carlo N-particle transport code, Version 5 Volume I: Overview and theory. Los Alamos
Nuclear Laboratory. ,https://doi.org/LA-UR-03-1987..
Yang, D., Liu, X., Xiong, J., Chai, X., & Cheng, X. (2018). Coupling of neutronics and thermal-hydraulics codes for the simulation of reactivity inser-
tion accident for LFR. Progress in Nuclear Energy. Available from https://doi.org/10.1016/j.pnucene.2018.02.023.
Chapter 8

Computer codes

Computer codes have become the mainstay of nuclear engineering and are abundantly used for the design and analyses
of nuclear systems. The use of codes, developed in universities, companies and government organizations, is indispens-
able as a first step in the design of a reactor and a requirement in regulatory processes. The USNRC has been using
computer codes since the development of early nuclear reactors. Similarly other international efforts include the
Institute de Radioprotection et de Surete Nucleaire (IRSN) in Europe and more recently, the Chinese nuclear regulatory
institution through the National Nuclear Safety Administration (NNSA).
With the progression followed in this book, earlier chapters have covered the radiation transport processes covered
charged-particle and gamma transport, neutron transport, a description of nuclear reactors, mathematical foundations,
neutron diffusion followed by neutron transport and Monte Carlo simulation and optimization in the later chapters. All
these are included in the computer codes used for the design and performance analyses of nuclear reactors and systems.
In Chapter 4, the mathematical treatment of coupled analyses, such as neutronics coupled with thermal hydraulics, was
introduced to demonstrate the interdisciplinary nature of nuclear engineering simulations. In these chapters, we saw that
reactor core analyses requires such modeling of phenomena that take place within, for example, a reactor core typically
consisting of several assemblies of fuel rods and control rods surrounded by moderator, coolant and protective structure.
Thus mathematical formulations are necessarily coupled and the required computer codes are not restricted to any one
field. The nuclear industry thus does not have a single code to handle all the requirements of nuclear systems analyses.
Therefore, we will look separately at codes in the following areas which are fairly representative of the broader picture
1. Neutron and radiation transport,
2. Time-dependent reactor kinetics,
3. Thermal hydraulics,
4. Radiological protection, and
5. Performance and Safety analyses.
Over time, there has been a shift toward the usage of these codes utilizing the power of supercomputing architecture.
The emphasis is thus on the development of algorithms and away from the basic mathematical rigor that was included
earlier.
The strength of a computer code lies in its capability to model the interaction physics as well as the three-
dimensional geometries of nuclear fission and fusion power reactors. The earlier codes, written for nuclear fission reac-
tors and being restricted to “regular” geometry, that is, slab, circular or cylindrical systems were capable only of idealiz-
ing complex geometry features intrinsic to core internals and radiation as well as biological shielding systems
surrounding the reactor. Over the years, with the development of better numerical schemes such as finite element meth-
ods in deterministic codes and combinatorial geometry in stochastic codes, it became possible to adequately describe
many physical complexities. The same applies to complexities in the collision data, which was used in multi-group
form rather than in continuous energy form, scattering models, fission spectra and so on.
With better capability in modeling real systems, the computational effort also increases and hence computer proces-
sing is a crucial requirement. Thus, computer codes have benefited from the processing speeds achievable by computer
systems as they evolved from mechanical and electrical computers in the 1940s to electronic computers, mainframes
such as IBM and CDC in the 1970s and 1980s, to workstations, RISC processors, Cray supercomputers with vector and
parallel architecture, to modern-day portable and efficient computers (Schober, 2018). Looking to the future, it is possi-
ble that optical data communication may replace the present electronic circuitry making large computational codes even
more attractive.

Nuclear Engineering. DOI: https://doi.org/10.1016/B978-0-323-90618-0.00008-9

© 2022 Elsevier Inc. All rights reserved. 337
338 Nuclear Engineering

The software and hardware features available in present-day computing permits simulation codes to cover complex
features to a level sufficient for reliable design. A computational challenge in reactor neutronics has been to estimate
“rare events” such as the transmission of radiation across the shielding.
Some of the widely used computer codes in nuclear systems are briefly described below.

8.1 Neutron and radiation transport codes

Early codes, which have been updated and are still used for reactor analysis including fuel depletion, include ANISN
(Engle, 1967), CITATION, WIMS, and ORIGEN. The diffusion theory code CITATION (Fowler, Vondy, &
Cunningham, 1971) released in 1972, solves the 3D neutron diffusion equation with the finite difference method and
has been used together with WIMS-D and ORIGEN for calculating neutronics coupled with burn-up and depletion of
fuel in nuclear systems. It is distributed by RSICC. WIMS (Winfrith Improved Multi-group Scheme) is a general code
for reactor physics lattice cell calculations for nuclear reactor systems and has been used for the calculation of group
constants (Section 6.1) for fuel assemblies such as for analysis of Loss of Flow transients (Noori-Kalkhoran,
Minuchehr, Shirani, & Rahgoshay, 2014), for core optimization simulation of a PWR (Hussain, Xinrong, & Ahmed,
2008) and a wide range of neutronics.
The cross section data library has a number of resonance groups for thermal neutron analysis. WIMSD is distributed
by the RSICC and it has been updated with several benchmarks (Lindley et al., 2017). ORIGEN (ORNL Isotope
Degeneration and Depletion Code), a fuel cycle depletion code developed by ORNL, performs nuclide transmutation
calculations based on neutron flux in a system (Bell, 1973; Yesilyurt, Clarno, Gauld, & Galloway, 2011). Other deter-
ministic codes are described below.

8.1.1 ANISN
One of the first codes used for neutron/photon transport, ANISN (Anisotropic SN method), is based on an advanced dis-
crete ordinates method proposed by Carlson in 1953 and developed by Wick and Carlson (Carlson & Lathrop, 1968) at
Los Alamos Scientific Laboratory.
ANISN calculates angular flux in sufficient detail to solve deep-penetration problems in addition to regular calcula-
tions of flux and reaction rates in regions of nuclear systems. Older versions of ANISN-ORNL, which were operable on
VAX and IBM mainframes, are now extended with improvements for Cray and IBM RISC versions as well as for
Pentium IV computers under Windows XP operating systems.

8.1.2 DOT
The DOT code (Rhoades & Mynatt, 1973), as version DOT-4, calculates multi-group angular flux by solving the
Boltzmann transport equation using the SN method, diffusion theory or a special P1 method, for two-dimensional geo-
metric systems. It can solve for criticality also, but the principal application is to the more difficult deep-penetration
transport of neutrons and photons for shielding problems. Cross sections are modeled using Legendre expansion to arbi-
trary order allowing for anisotropic scattering. A vectorized version, DOT-4/VE has increased computational efficiency.
The CPU time required by DOT depends on the Flux Work Units (FWUs) which are dependent on the number of space
mesh cells, energy groups, and the number of iterations per group. Typical of large deterministic codes, big problems
could take over an hour on modern platforms.

8.1.3 TORT
TORT (2D, 3D) and DORT (1D, 2D) are Discrete-Ordinate Neutron and Photon transport codes (Kirk, 2009) with
eigenvalue and deep-penetration calculation capabilities. TORT calculates the flux of particles due to particles incident
upon the system from extraneous sources or generated internally as a result of interaction with the system. The
Boltzmann transport equation is solved using the discrete ordinates method to treat the directional variable and the finite
difference methods to treat spatial variables. Energy dependence is treated using a multi-group formulation. Time
dependence is not treated.
 Computer codes Chapter | 8 339

8.1.4 PARTISN
PARallel Time-Dependent SN (Alcouffe, 2001), a successor of ONEDANT, TWODANT, and THREEDANT is a par-
allel time-dependent deterministic code that can solve three-dimensional problems in regular geometry using the dis-
crete ordinates method.
Some of the extensively used Monte Carlo codes are described below.

8.1.5 MCNP
Monte Carlo methods for Neutrons and Photons (Pelowitz et al., 2013) were developed at Los Alamos National Labs
(LANL) in the 1940s by the group that included von Neumann. Its’ name later implied Monte Carlo N-Particle transport
as it included charged particles. MCNP is used for a diverse range of problems involving neutron and radiation transport
in nuclear reactor engineering with the capability of carrying out “whole-core” analyses, as well as medical physics,
oil-well logging, and radiography. The first Monte Carlo code appeared as a Los Alamos report in 1963. By the early
1980s, MCNP begain to be widely distributed and used as MCNP3A and MCNP3B written in Fortran 77 with neutron
and photon transport. The next version, MCNP4 had improvements such as electron transport capability as well as new
tallies. It had parallel processing to speed up the computations. The perturbation capability was added in MCNP4B in
the late 1990s. This was followed by MCNP4C which had some better physics such as the treatment of nuclear reso-
nances. MCNP5 released in 2003 was re-written in Fortran 90 and had several improvements. The latest version,
MCNP6.2 (2017) can transport 36 different particles from neutrons, photons, electrons to muons, anti neutrons various
baryons, anti-particles, deuterons, tritons, alpha and heavy ions. The long list of improvements includes the modeling of
physics, sources, data, tallies, unstructured meshes as well as code enhancements. This has led to a greatly increased
versatility going far beyond the early neutron and photon transport to areas such as medical physics, accelerator-driven
energy source research, high-energy dosimetry and futuristic areas such as charged-particle propulsion. Stochastic
geometry has been incorporated to model the random arrangements in a geometry such as the high-temperature gas
cooled reactor (HTGR) with its spherical fuel pebbles.
New Monte Carlo codes such as OpenMC (Romano et al., 2015) have improved combinatorial geometry models
that have been demonstrated for photon simulation (Kargaran, Jafari, & Minuchehr, 2021) and are openly accessible.
MCNP benchmarks are available for neutron and problems (Whalen, Cardon, Uhle, & Hendricks, 1991), criticality
safety problems (Wagner, Sisolak, & McKinney, 1992) and LWR criticality. MCNP has been used extensively for core
neutronics; some representative latest work is the study on the performance analysis of thorium-based mixed oxide fuel
in a PWR (Tucker & Usman, 2018), a conceptual design of a micronuclear reactor using heat pipes (Aziz, Koreshi,
Sheikh, & Khan, 2020; Hao Sun et al., 2018) and the study of accident tolerant fuels in the upcoming NuScale SMR
(Yu, Cai, He, & Li, 2021).

8.1.6 TART
TART 2005 (Kirk, 2009) is a coupled neutron photon Monte Carlo particle transport code written in FORTRAN and
developed by Dermott E. Cullen (McKinley, Cullen, Latkowski, Procassini, & Skulina, 2006) and maintained by
Lawrence Livermore Laboratories (LLL). It is time-dependent and uses combinatorial geometry allowing the modeling
of large and complex problems. It is LLL’s equivalent of LANL’s MCNP and is said to be the fastest MC code.

8.1.7 MORSE
MORSE (Multi-group Oak Ridge Stochastic Experiment), succeeded by MONACO, is a Monte Carlo code developed
by Oak Ridge National Laboratories (ORNL) (Emmett, 2000). Initially, it was also based on regular geometry and
although it uses the same integral formulation of the transport equation as a continuous code such as MCNP, MORSE
uses a multi-group approach. A neutron thus collides with an atom of the homogenized mixture prepared from the mac-
roscopic cross sections rather than with individual atoms. Such a multi-group MC approach has served for quick com-
parisons with deterministic codes such as ANISN but for more elaborate computation, MCNP is preferred.
MORSE-SGC (Emmett, 1975) is run in stand-alone mode and is part of the SCALE licensing evaluation system. It
uses the MARS (Multiple Array System) geometry package. The MORSE code is distributed by RSICC. The multi-
group energy handling capability of MORSE has been useful for comparisons with ANISN and DOT computations.
MORSE has also been coupled with discrete ordinates codes with the DOMINO-II code.
340 Nuclear Engineering

Dose rates from a cask array have been calculated with MAVRIC (MONACO with Automated Variance Reduction
using Importance Calculations) for the safety of workers.

8.1.8 KENO
The KENO multi-group Monte Carlo code for criticality computations (Goluoglu, Bowman, & Dunn, 2007) was devel-
oped in the 1960s at ORNL. It is now part of the SCALE system and its use is as widespread as that of other MC codes.
KENO is a distributed by the Nuclear Energy Agency (NEA) with generalized geometry capability allowing easy
description of systems composed of cylinders, spheres, and cuboids (rectangular parallelepipeds) arranged in any order.

8.1.9 Other Monte Carlo codes

MONK and MCBEND (Long et al., 2015; Richards et al., 2015) are UKAEA’s standard criticality and radiation shield-
ing codes (Richards et al., 2015).
TRIPOLI (Nimal & Vergnaud, 1990, 2001; Vergnaud & Nimal, 1990) is a Monte Carlo code that was written at
CEA, France, and used for “whole-core” neutronic reactor analyses, radiation shielding, and protection as well as for
criticality calculations.
Other neutron/photon transport codes are VIM, a Monte Carlo code developed by Argonne National Labs, and
SERPENT, a continuous energy MC code for neutron transport, criticality and reactor physics burn-up calculations
developed by VTT Technical Research Center, Finland. Charged-particle codes that handle electron/photon transport
include EGS5 and PENELOPE, while for high-energy transport codes include GEANT4, FLUKA and MCNPX.

8.2 Time-dependent reactor kinetics codes

PARCS (The Purdue Advanced Reactor Core Simulator), developed by the USNRC, solves for the transient neutron
flux distribution and is used for the analysis of possible reactivity-initiated accidents in light water reactors. It can also
be coupled to other USNRC thermal hydraulics codes such as RELAP.

8.3 Thermal hydraulics codes

Thermal hydraulics studies are carried out coupled with neutronics to study transients such as LOFA in WWER (Noori-
Kalkhoran et al., 2014), and reactivity insertion accidents in Gen IV Liquid Metal Cooled Fast Reactor with the
COBRA code (Yang, Liu, Xiong, Chai, & Cheng, 2018).
A thermal hydraulics code, such as RELAP, carries out a transient two-phase two-fluid hydrodynamic analysis for a
nuclear reactor modeled in sufficient detail to analyze transient conditions for safety studies. Detailed studies are carried
out in a suite such as the SCALE code system (Rearden & Jessee, 2016). A full-scale neutronic analysis for a sodium-
cooled fast reactor core carried out using the lattice code HELIOS-2 with the Monte Carlo code Serpent to generate
few-group constants used by the 3D deterministic code DYN3D was shown to give reliable results (Rachamin,
Wemple, & Fridman, 2013).
For the analysis of transients and large/small LOCAs in PWRs and BWRs, TRACE/RELAP (TRAC/RELAP
Advanced Computational Engine) is used. RELAP (The Reactor Excursion and Leakage Analysis Program) will be
replaced by TRACE in a few years. Some representative calculations for new reactor systems are: a LOCA analysis for
NuScale carried out to model an extremely severe accident caused by loss of decay heat removal (Skolik et al., 2021;
Skolik, Trivedi, Perez-Ferragut, & Allison, 2019), and a study on the graphite oxidation behavior for reliability studies
of a HTGR (Chai, Wu, & Okamoto, 2020).
Full 3D reactor core simulators, with the capability of neutronic-thermal hydraulics coupling are used to solve for
the reactor power and flow field for steady-state as well as transient calculations. Free open source codes such as
OpenMOC (Boyd, Shaner, Li, Forget, & Smith, 2014) and OpenMC (Romano et al., 2015) are available for carrying
out deterministic computations with the method of characteristics and Monte Carlo simulations for neutronics and
OpenFOAM (OpenFOAM, 2014) for computational fluid dynamics computations. Monte Carlo methods have been
combined with the method of characteristics and demonstrated that the hybrid model is better than each one alone (Lee,
Choi, & Lee, 2015). Coupling of neutronics with CFD has been carried out for Multiphysics simulations (Castagna
et al., 2020; Seubert et al., 2012; Tuominen, Valtavirta, & Lepp, 2018; Tuominen, Valtavirta, Peltola, & Leppänen,
2016) of 3D deterministic as well as Monte Carlo codes for core analysis.
 Computer codes Chapter | 8 341

8.4 Radiological protection codes

For source term analysis (Section 3) and radionuclide releases under accident conditions, the USNRC codes
RADTRAD (RADionuclide Transport and Removal And Dose Estimation) and RASCAL (Radiological Assessment
Systems for Consequence AnaLysis) are used.

8.5 Performance and safety analyses

For the performance of a single fuel rod under near-normal reactor operating conditions, codes developed by the
USNRC are FRAPCON-3, for steady-state and mild transient analysis and FRAPTRAN for design basis accident analy-
sis. These codes are very useful for the design of new SMRs, as described in Chapter 3. Two such examples are the
work on a small modular 150 MWt PWR (Mirian & Ayoobian, 2020) and on studying aa accident tolerant fuel (He,
Shirvan, Wu, & Su, 2019). In the first case, using FRAPCON, the effect of varying fuel diameter and pitch on the
excess reactivity, neutron flux and burn-up were calculated by MCNP and coupled with FRAPCON to compute the fuel
centerline temperature, stresses and pressures with the objective of enhancing the fuel cycle lifetime. In the second,
using FRAPTRAN, a new silicon carbide (SiC) fuel was studied during a LOCA for its mechanical integrity.
For severe accident analysis the USNRC codes MELCOR for core meltdown analysis, such as for the modeling of
Fukushima Daiichi Unit 1 (Herranz & López, 2020) and for a study on the AP1000 (Malicki, Pieńkowski, Skolik, &
Trivedi, 2019) an integral severe accident analysis code, and SCDAP/RELAP5 are used.
To perform analyses on the possible atmospheric release of radionuclides with environmental consequences, the
MACCS (The MELCORE Accident Consequence Code) Code has been widely used.
Probabilistic Risk Assessment: SAPHIRE (Systems Analysis Programs for Hands-on Reliability) developed by the
USNRC
Typical simulation results from the codes described above include neutron and photon flux, criticality computations,
and reaction rates in regions of interest. Deep-penetration calculations are required for the design of radiation and bio-
logical shielding around areas of high radiation.
For ITER, for example, elementary one-dimensional calculations are useful before proceeding to two-dimensional
and full three-dimensional models. A simple 1D model of ITER in MCNP has been used to get energy-dependent neu-
tron flux and dose rates.
Such 1D results are often useful to draw important engineering design parameters and limits. For example, in this
1D analysis, the authors conclude that “the magnitude of the dose rate on the outside hall of bioshield during normal
ITER operation cannot be considered low in accordance with the result found in the simulation performed in this work,
that is, 1 μSv=h. It is important to remember that the contributions due to the presence of slits between blanket mod-
ules and also the various holes in the cryostat and bioshield were not considered in the calculation.”
These are the very important findings of such simulations. Consider these with the acceptable radiation levels of the
order of 0.05 Sv described in Chapter 1 and a preliminary estimate can be made to conclude that safety radiation levels
will be exceeded by about five years. In the expression that “. . . the presence of slits between blanket modules . . . were
not considered in this calculation,” it is clear that a 2D, or possibly a full 3D simulation is required.
To model geometrical complexity such as the ducts mentioned above, codes such as MCNP, KENO and DOT can
be used to carry out two- or three-dimensional calculations for the streaming analysis.
A typical ducting problem that must be addressed in nuclear reactors viz the streaming of radiation in ports and ducts
designed to place instrumentation at locations in the reactor. Results are found for levels of radiation that can subse-
quently be used to determine safety for workers and radiation damage to instrumentation.

8.6 Nuclear data

Computer codes come with nuclear data libraries which may need to be processed before they can be used. There are
several nuclear data libraries such as ENDF/B (Evaluated Nuclear Data File, Brookhaven National laboratory) (Brown
et al., 2018), ENDL (Evaluated Nuclear and Atomic Reaction Data Library, Lawrence Livermore Laboratory, USA),
JEFF (European Joint Evaluated Fission and Fusion (JEFF) Library, OECD Nuclear Energy Agency), JENDL (Japanese
Evaluated Nuclear Data Library), CENDL (Chinese Evaluated Nuclear Data Library) etc. which contain data obtained
from experiments (Ge, Wu, Chen, & Xu, 2017; Herman & Trkov, 2010; Shibata et al., 2011).
342 Nuclear Engineering

These cross sections are available in the “old” formats with tabulations described below. The ENDF data is retrieved
by specifying material numbers containing the atomic number Z and the atomic mass number A, the reaction of interest,
the energy of interest, the version of the evaluation and several other details.
ENDF/B evaluations have a MAT number for all materials, an MF number for files, and a MT number for reaction
type. For U238, for example, the information file has the info: U-238 MAT 5 9237 MF 5 1 MT 5 451 Library: ENDF/
B-VII.0
Thus U238 evaluations have 10 information files viz MF 5 1, 2, 3, 4, 5,6, 12, 13, 14 and 15, each of which carry
information on the MT data available, for example, MT1 is the total neutron microscopic cross section σtot and MT16
is the ðn; 2nÞ data. A glance over the MF files is useful to appreciate the data “pointers”, that is, the MT reaction data in
each file.
MF 5 1 General Information
MT 5 452: total average neutron multiplicity per fission (total nubar)
MT 5 455: average delayed neutron multiplicity per fission
MT 5 456: average prompt neutron multiplicity per fission
MT 5 458: energy release from fission
MF 5 2 Resonance Parameters
MT 5 151 resolved and unresolved resonance parameters
MF 5 3 Neutron Cross sections
MT 5 1: total cross section
MT 5 2: elastic scattering cross section
MT 5 3: nonelastic cross section
MT 5 4: inelastic cross section
MT 5 16: (n,2n) cross section
MT 5 17: (n,3n) cross section
MT 5 18: fission neutron cross section
MT 5 19,20,21,38: multi-chance fission cross sections
MT 5 37: (n,4n) cross section
MT 5 5171 discrete inelastic level cross sections
MT 5 7290: discrete inelastic level cross sections (direct reactions to groups of states) MT 5 91: inelastic contin-
uum neutron cross section
MT 5 102: neutron radiative capture cross section
MF 5 4 Angular Distributions of Secondary Particles
MT 5 2: neutron elastic scattering angular distributions
MT 5 5153: discrete inelastic neutron angular distributions for the ground-state rotational band
MT 5 5471: discrete inelastic neutron angular distributions
MT 5 7290: discrete inelastic neutron angular distributions for direct reactions
MF 5 5 Energy Distributions of Secondary Particles
MT 5 18: prompt fission neutron spectrum matrix
MT 5 455: delayed neutron emission spectra from fission
MF 5 6 Product Energy-Angular Distributions
MT 5 16,17,37: (n, xn) continuum distributions
MT 5 91: (n, n0 ) continuum distributions
MF 5 12 Photon Prod Multiplicities & Transition Probabilities
MT 5 18,102: photon multiplicities from inelastic scattering, fission, and radiative capture
MF 5 13 Photon Production Cross Sections
MT 5 3: photon production cross sections from nonelastic reactions
MF 5 14 Photon Angular Distributions
MT 5 3,18,102: photon angular distributions from nonelastic, fission, and radiative capture reactions
MF 5 15 Continuous Photon Energy Spectra
MT 5 3,18,102: photon energy spectra from nonelastic reactions, fission and radiative capture
Reactions are also available with MT numbers in MF files.
In the table above, for MF 5 1, the information in this file holds only five reaction types, that is, MT451, 452, 455,
456, and 458. MT451 is the INFO associated with the data, while MT452 holds ν t , the average total number of neutrons
 Computer codes Chapter | 8 343

(prompt plus delayed) neutrons released per fission event. More details are in the ENDF-6 Formats Manual on the BNL
website. The first ten lines in this data file are

This section is reproduced from the “Formats Manual”: To read the data, the position of the “records” has to be
known. All records are one of six types: TEXT, CONT (has six special cases called DIR, HEAD, SEND, FEND,
MEND, and TEND), LIST, TAB1, TAB2, and INTG. For example, SEND indicates a section end, FEND indicates a
file end, MEND indicates a material end, and TEND indicates a “tape end.” The counter NS is reset for every section
in the file; the last is SEND which has NS 5 99,999.
Every line toward the end has 92,237 which is the MAT number, followed by a one which is the file number, then
451 which is the MT (reaction) number, then the record number NS.
From the ENDF/B data, it can be verified that the first file has the first section (MT451) with 796 lines; the second
section (MT452) has 07 lines; the third section (MT455) has 07 lines; the fourth section (MT456) has 07 lines; the fifth
section (MT458) has 05 lines; then file 2 begins with MT151.
For specific applications where temperature-dependent data, for example, is required, nuclear data processing codes,
such as NJOY, are used to process the cross sections in data libraries such as ENDF/B to prepare multi-group and mix-
ture cross sections. Needless to state, data handling is itself a large field and professionals involved in the compilation,
maintenance and processing of nuclear data most often stay within this domain.
As mentioned earlier, in the two step process, the first step is to use a lattice physics codes with a cross section gen-
eration code to provide the cross sections input to a neutronics code which computes the fluxes to get the power distri-
bution. This power distribution is input to a thermal hydraulics code
One such system uses the GENPMAXS (Porhemmat, Hadad, & Faghihi, 2015) for the Generation of the Purdue
Macroscopic XS set code module for the 3D Purdue Advanced Reactor Core Simulator PARCS. The generation code uses lat-
tice codes such as HELIOS-2 (Wemple & Villarino, 2008), CASMO (Rhodes, Smith, & Lee, 2006), and TRITON (DeHart,
2006).
The use of codes, developed in universities, companies, and government organizations, is indispensable as a first
step in the design of a reactor and a requirement in regulatory processes. The USNRC has been using computer codes
since the development of early nuclear reactors. Similarly other international efforts include the Institute de
Radioprotection et de Surete Nucleaire (IRSN) in Europe and more recently the Chinese nuclear regulatory institution
through the National Nuclear Safety Administration (NNSA).
344 Nuclear Engineering

8.6.1. MCNP
The first Monte Carlo code appeared as a Los Alamos report in 1963. By the early 1980s, MCNP began to be widely
distributed and used as MCNP3A and MCNP3B written in Fortran 77 with neutron and photon transport. The next ver-
sion, MCNP4, had improvements such as electron transport capability as well as new tallies. It had parallel processing
to speed up the computations. The perturbation capability was added in MCNP4B in the late 1990s. This was followed
by MCNP4C which had some better physics such as the treatment of nuclear resonances. MCNP5 released in 2003 was
rewritten in Fortran 90 and had several improvements. The latest version, MCNP6.2 (2017), can transport 36 different
particles from neutrons, photons, electrons to muons, antineutrons various baryons, antiparticles, deuterons, tritons,
alpha, and heavy ions. The long list of improvements includes the modeling of physics, sources, data, tallies, unstruc-
tured meshes, as well as code enhancements. This has led to a greatly increased versatility going far beyond the early
neutron and photon transport to areas such as medical physics, accelerator-driven energy source research, high-energy
dosimetry, and futuristic areas such as charged particle propulsion. Stochastic geometry has been incorporated to model
the random arrangements in a geometry such as the high-temperature gas-cooled reactor with its spherical fuel pebbles.

8.7 Conclusion
Many of the computer codes used during the various stages of a nuclear reactor lifecycle are discussed in this chapter,
from the concept and design stage to licensing, performance, regulatory, safety and accident analysis, to the decommis-
sioning of a reactor. With advancements in computer hardware as well as the development of powerful algorithms,
computer simulations and codes are most likely to play an increasingly important role in the nuclear industry leading to
better, safer, and more competitive designs giving greater eminence to nuclear energy in the future.

Problems
1. Describe briefly how the ANISN code computes neutron flux in a cylinder. (a) What would be the difference in a
computational strategy in ANISN for radiation shielding compared with a boron concentration criticality search? (b)
How are neutron and photon dose rates obtained in ANISN?
2. For a discrete ordinates ANISN-S4 computation, what angular bins would be specified in a MCNP tally to com-
pare results from both?
3. What is the fundamental difference between MORSE and MCNP (both Monte Carlo codes) in modeling neu-
tronnuclei interactions?
4. What is common between the deterministic code ANISN and the Monte Carlo code MORSE?
5. For a situation in which a pipe failure in the Spent Fuel Pool (SFP causes a loss of coolant accident (LOCA)
describe what useful results the FRAPTRAN code could give. From the FRAPTRAN model, answer the following
questions: (a) Why would the spent fuel rods stored in the SFP remain hot even though they have been removed from
the reactor? (b) What model is used for thermal conductivity of uranium oxide fuel and how does it relate to the expres-
sion given in Chapter 4? (c) In the finite-difference scheme for the time-dependent conduction equation what quantities
are obtained? (d) How is the stored energy (in a fuel rod) modeled in the code? (e) Which model for convection would
be most appropriate to apply in SPF conditions? (f) What mechanical response does the code model? (g) What makes
the modeling in the code multidisciplinary (as mentioned in Chapter 4)? (h) What would occur in the reactor building if
the cladding were to burst?
6. How could the perturbation feature in MCNP be used for carrying out a boron criticality search in an LWR?
7. What makes the MC method particularly suitable for parallel computation? How would you carry out an MC sim-
ulation of particle histories using the parallel computing feature in Matlab and what computational speedup would you
expect?
8. In an optimization study of a fusion reactor blanket design what factors could favor the use of a deterministic
code such as ANISN over a Monte Code such as MORSE?
9. With reference to Chapter 4 and the description of the MCNP and ORIGEN codes given above, if the two codes
are coupled for a burnup calculation in an LWR, (a) how will MCNP and ORIGEN be coupled, that is, what quantity
will be given from MCNP to ORIGEN and which input data of MCNP will be modified for the next run in the loop, (b)
which nuclide compositions will be of interest, and (c) when will a MCNP-ORIGEN coupling be favored over a
ANISN-ORIGEN coupling?
 Computer codes Chapter | 8 345

10. In the MCNPORIGEN coupling discussed in 8.9, how would the reactor power be used in data transfer from
MCNP to ORIGEN?

Nomenclature
Abbreviations and acronyms
ANISN anisotropic SN method
CENDL Chinese evaluated nuclear data library
DORT discrete-ordinate neutron and photon transport
ENDF/B evaluated nuclear data file, Brookhaven National Laboratory
ENDL evaluated nuclear and atomic reaction data library
FORTRAN formula translation
FRAPTRAN fuel rod analysis program transient
GEANT geometry and tracking
HTGR high temperature gas cooled reactor
IBM International Business Machines
ITER International Thermonuclear Experimental Reactor
JEFF European Joint Evaluated Fission and Fusion (JEFF) Library, OECD Nuclear Energy Agency
JENDL Japanese evaluated nuclear data library
IRSN Institute de Radioprotection et de Surete Nucleaire
LOCA loss of coolant accident
LWR light water reactor
MARS multiple array system
MC Monte Carlo
MCNP Monte Carlo N-particle
MACCS The MELCORE Accident Consequence Code
MF material file
MORSE Multigroup Oak Ridge Stochastic Experiment
MT material type
NEA Nucear Energy Agency
NNSA National Nuclear Safety Administration
ORIGEN ORNL Isotope Degeneration and Depletion Code
RELAP The Reactor Excursion and Leakage Analysis Program
RADTRAD RADionuclide Transport and Removal And Dose Estimation
RASCAL Radiological Assessment Systems for Consequence AnaLysis
RISC reduced instruction set computer
RSICC radiation safety information computational center
SAPHIRE systems analysis programs for hands-on reliability
TORT a three-dimensional discrete ordinates neutron/photon transport code
USNRC United States Nuclear Regulatory Commission
WIMS Winfrith improved multigroup scheme
WWER water-water energetic reactor

References
Alcouffe, R. E. (2001). PARTISN calculations of 3D radiation transport benchmarks for simple geometries with void regions. Progress in Nuclear
Energy.
Aziz, U., Koreshi, Z. U., Sheikh, S. R., & Khan, H. (2020). A study of criticality and thermal loading in a conceptual micronuclear heat pipe reactor
for space applications. Nuclear Technology & Radiation Protection.
Bell, M. J. (1973). ORIGEN—The ORNL isotope generation and depletion code.
Boyd, W., Shaner, S., Li, L., Forget, B., & Smith, K. (2014). The OpenMOC method of characteristics neutral particle transport code. Annals of
Nuclear Energy.
Brown, D. A., et al. (2018). ENDF/B-VIII.0: The 8th Major Release of the Nuclear Reaction Data Library with CIELO-project Cross Sections, New
Standards and Thermal Scattering Data. Nuclear Data Sheets.
346 Nuclear Engineering

Carlson, B. G., & Lathrop, K. D. (1968). Transport theory the method of discrete ordinates. Computing methods in reactor physics. Gordon and
Breach.
Castagna, C., et al. (2020). A Serpent/OpenFOAM coupling for 3D burnup analysis. European Physical Journal - Plus.
Chai, P., Wu, Y., & Okamoto, K. (2020). Improvement and validation of RELAP5/SCDAP code on evaluating the graphite behavior during oxidation
process. Progress in Nuclear Energy.
DeHart, M. D. (2006). Lattice physics capabilities of the SCALE code system using TRITON. PHYSOR-2006 - American Nuclear Society’s topical
meeting on reactor physics. American Nuclear Society.
Emmett, M. (1975). Morse Monte Carlo radiation transport code system.
Emmett, M. B. (2000). MORSE: Present capabilities and future directions. In Applied radiation and isotopes.
Engle, W. W. J. (1967). A users manual for ANISN: A one dimensional discrete ordinates transport code with anisotropic scattering. Tennessee.
Fowler, T. B., Vondy, D. R., & Cunningham, G. W. (1971). Nuclear reactor core analysis Code-CITATION.
Ge, Z., Wu, H., Chen, G., & Xu, R. (2017). CENDL project, the Chinese evaluated nuclear data library. EPJ web of conferences. EDP Sciences.
Goluoglu, S., Bowman, S. M., & Dunn, M. E. (2007). KENO monte carlo code capabilities. Transactions of the American Nuclear Society. American
Nuclear Society.
Hao Sun, G. S., Wang, C., Ma, P., Liu, X., Tian, W., & Qiu, S. (2018). Comnceptual design and analysis of a multipurpose micro nuclear reactor
power source. Annals of Nuclear Energy, 121, 118127.
He, Y., Shirvan, K., Wu, Y., & Su, G. (2019). Integrating a multi-layer deformation model in FRAPTRAN for accident Tolerant fuel analysis. Annals
of Nuclear Energy.
Herman, M., & Trkov, A. (2010). ENDF-6 formats manual.
Herranz, L. E., & López, C. (2020). Challenges and sensitivities in the modelling of Fukushima Daiichi Unit 1 unfolding with MELCOR 2.2. Annals
of Nuclear Energy.
Hussain, A., Xinrong, C., & Ahmed, R. (2008). The conceptual design study of a TRISO fueled compact PWR core. International conference on
nuclear engineering, proceedings, ICONE. ASME.
Kargaran, H., Jafari, H., & Minuchehr, A. (2021). A novel approach based on improved combinatorial geometry (ICG) model for photon transport in
Monte Carlo codes. Progress in Nuclear Energy.
Kirk, B. (2009). Overview of neutron transport codes. 11th Neutron and ion dosimetry symposium-NEUDOS-11. NEUDOS.
Lee, H., Choi, S., & Lee, D. (2015). A hybrid Monte Carlo/method-of-characteristics method for efficient neutron transport analysis. Nuclear Science
and Engineering: The Journal of the American Nuclear Society.
Lindley, B. A., et al. (2017). Current status of the reactor physics code WIMS and recent developments. Annals of Nuclear Energy.
Long, D., et al. (2015). MONK10: A Monte Carlo code for criticality analysis. ICNC 2015 - International Conference on Nuclear Criticality Safety.
ICNC.
Malicki, M., Pieńkowski, L., Skolik, K., & Trivedi, A. K. (2019). Development of MELCOR thermal hydraulic model of AP1000 and its verification
for a DECL break. Annals of Nuclear Energy.
McKinley, S., Cullen, D. E., Latkowski, J. F., Procassini, R., & Skulina, K. M. (2006). Application of MCNP, MERCURY and TART to calculation of
the National Ignition Facility (NIF) shielding. Transactions of the American Nuclear Society. American Nuclear Society.
Mirian, S. F., & Ayoobian, N. (2020). Investigations on a typical small modular PWR using coupled neutronic-thermal-mechanical evaluations to
achieve long-life cycle-length. Progress in Nuclear Energy.
Nimal, J. C., & Vergnaud, T. (1990). TRIPOLI: A general Monte Carlo code, present state and future prospects. Progress in Nuclear Energy.
Nimal, J. C., & Vergnaud, T. (2001). TRIPOLI-3: A neutron/photon Monte Carlo transport code. Advanced Monte Carlo for radiation physics, parti-
cle transport simulation and applications. Springer.
Noori-Kalkhoran, O., Minuchehr, A., Shirani, A. S., & Rahgoshay, M. (2014). Full scope thermal-neutronic analysis of LOFA in a WWER-1000 reac-
tor core by coupling PARCS v2.7 and COBRA-EN. Progress in Nuclear Energy.
OpenFOAM. (2014). OpenrFOAM - The open source CFD toolbox - User guide. OpenFOAM Found.
Pelowitz, D. B., et al. (2013). MCNP6 user’s manual. La-Cp-1300634.
Porhemmat, M. H., Hadad, K., & Faghihi, F. (2015). PARCS cross-section library generator; Part one: Development and verification. Progress in
Nuclear Energy.
Rachamin, R., Wemple, C., & Fridman, E. (2013). Neutronic analysis of SFR core with HELIOS-2, Serpent, and DYN3D codes. Annals of Nuclear
Energy.
Rearden, B. T., & Jessee, M. A. (2016). SCALE code system.
Rhoades, W. A., & Mynatt, F. R. (1973). DOT III two-dimensional discrete ordinates transport code.
Rhodes, J., Smith, K., & Lee, D. (2006). CASMO-5 development and applications. PHYSOR-2006 - American Nuclear Society’s topical meeting on
reactor physics. American Nuclear Society.
Richards, S. D., et al. (2015). MONK and MCBEND: Current status and recent developments. Annals of Nuclear Energy.
Romano, P. K., Horelik, N. E., Herman, B. R., Nelson, A. G., Forget, B., & Smith, K. (2015). OpenMC: A state-of-the-art Monte Carlo code for
research and development. Annals of Nuclear Energy.
Schober, P. (2018). Supercomputers. High-performance computing in finance: Problems, methods, and solutions. Chapman and Hall/CRC.
Seubert, A., Sureda, A., Bader, J., Lapins, J., Buck, M., & Laurien, E. (2012). The 3-D time-dependent transport code TORT-TD and its coupling with
the 3D thermal-hydraulic code ATTICA3D for HTGR applications. Nuclear Engineering and Design.
 Computer codes Chapter | 8 347

Shibata, K., et al. (2011). JENDL-4.0: A new library for nuclear science and engineering. Journal of Nuclear Science and Technology.
Skolik, K., Trivedi, A., Perez-Ferragut, M., & Allison, C. (2019). LOCA-type scenario simulation for NuScale-SMR with RELAP/SCDAPSIM/
MOD3.4. 18th International topical meeting on nuclear reactor thermal hydraulics, NURETH 2019. American Nuclear Society.
Skolik, K., Allison, C., Hohorst, J., Malicki, M., Perez-Ferragut, M., Pieńkowski, L., & Trivedi, A. (2021). Analysis of loss of coolant accident without
ECCS and DHRS in an integral pressurized water reactor using RELAP/SCDAPSIM. Progress in Nuclear Energy.
Tucker, L. P., & Usman, S. (2018). Thorium-based mixed oxide fuel in a pressurized water reactor: A burnup analysis with MCNP. Annals of Nuclear
Energy.
Tuominen, R., Valtavirta, V., & Lepp, J. (2018). Application of the serpent  openfoam coupled code system to the sealer reactor core. PHYSOR
2018: Reactor physics paving the way towards more efficient systems. American Nuclear Society.
Tuominen, R., Valtavirta, V., Peltola, J., & Leppänen, J. (2016). Coupling serpent and openfoam for neutronics - CFD multi-physics calculations.
Physics of reactors 2016, PHYSOR 2016: Unifying theory and experiments in the 21st century. American Nuclear Society.
Vergnaud, T., & Nimal, J. C. (1990). TRIPOLI capabilities proved by a set of solved problems. Progress in Nuclear Energy.
Wagner, J. C., Sisolak, J. E., & McKinney, G. (1992). MCNP: Criticality safety benchmark problems, LA-12415. New Mexico, United States.
Wemple, C. A., & Villarino, E. A. (2008). The HELIOS-2 lattice physics code. In Physics. College Park, MD.
Whalen, D. J., Cardon, D. A., Uhle, J. L., & Hendricks, J. S. (1991). MCNP: Neutron benchmark problems. LA12212 DE92 004710.
Yang, D., Liu, X., Xiong, J., Chai, X., & Cheng, X. (2018). Coupling of neutronics and thermal-hydraulics codes for the simulation of reactivity inser-
tion accident for LFR. Progress in Nuclear Energy.
Yesilyurt, G., Clarno, K. T., Gauld, I. C., & Galloway, J. (2011). Modular origen-S for multi-physics code systems. American Nuclear Society.
Yu, H., Cai, J., He, S., & Li, X. (2021). Analysis of neutron physics and thermal hydraulics for fuel assembly of small modular reactor loaded with
ATFs. Annals of Nuclear Energy.
Chapter 9

Optimization and variational methods

9.1 Introduction
In engineering design three very important goals are safety, design optimization, and economic optimization.
This is true for nuclear technologies as nuclear power reactors are designed to operate safely, designed and built on
the best knowledge of the day, and provide economically competitive energy. These features have been the hallmark of
nuclear energy over the last 60 years of providing safe reliable and competitive energy (Section 3.1).
This last requirement will be most severely tested in the decades to come. Nuclear energy will have to compete with
fossils and renewables to make a renaissance possible. The competitiveness will most likely come from the Gen IV
reactor designs which have some very attractive features described in Chapter 3 (Section 3.5.3).
The present challenge for nuclear technology is optimization in proposed designs to win the licensing formalities
and gain public acceptance to provide energy for the decades to come.
The main difference between design optimization in nuclear engineering and in other power systems is the high
nuclear energy density that minimizes the tolerances. Both fission and fusion power systems contain some of the most
sophisticated concepts about the way nature works as well as some of the most advanced technologies developed in the
scientific community; and with it have come derivative benefits in the development of mathematical models and com-
puting platforms.
Today, optimization in nuclear engineering is performed mostly with heuristic and meta-heuristic (meaning more
efficient search techniques) methods with artificial intelligence expert knowledge-based systems to make a machine
learn while it calculates, to mimic the workings of the human brain by neural networks rendering a machine towards a
human way of thinking. It has thus become possible, for example, to optimize the fuel loading in 121 assemblies of a
pressurized water reactor (PWR) core (Section 3.2.1) with the fuel of the right enrichment and the right burnup to be
placed at the right place in the core at the right time of its power producing several-year schedule. A great achievement
considering that the factorial of 121 is B 10200 which means that out of so many possible solutions, the best must be
found. How does one do that!
Reactor core fueling optimization is a huge search space even by modern computational capabilities, so a reduction in
complexity is achieved by using a 1/8 core symmetric model to search from an optimal configuration from 15! B 1012
permutations and combinations, which becomes possible.
In nuclear engineering, optimization is applied in areas including overall reactor core design (Bae, Betzler,
Chandler, & Ilas, 2021; Betzler, Chandler, Cook, Davidson, & Ilas, 2019), reactor operations (Kumar & Tsvetkov,
2015; Mousakazemi, 2021), safety-related management strategies, in-core fuel management (Chham et al., 2021), the
design of criticality experiments, centrifuge cascade optimal configurations, radiation shielding design (Ahmad, Chang,
Li, Yang, & Liu, 2021; Cai, Hu, Lu, & Jia, 2018), controller design (Mousakazemi, 2021; Mousakazemi, Ayoobian, &
Ansarifar, 2018; Wan & Zhao, 2017), and many more.
Most of the present-day optimization is multi-objective and multi-modal and, in many cases, requires combinatorial
optimizations made efficient by the use of machine learning algorithms and expert-based systems.
This chapter focuses on the applicable mathematics of optimization followed by a cursory review of some application areas.
Historically, this field has its origins in the works of Euler (1707
83) and Lagrange (1736
1813), following
Newton’s (1642
1726) optimization, and was further demonstrated by Bernoulli (1667
1748) in the classical brachis-
tochrone problem. Two centuries later, Pontryagin (1908
88) developed what is now known and widely applied as the
“bang-bang” control based on Pontryagin’s Maximum Principle. Such variational methods were mainly focused on
mechanics with considerable contributions by mathematicians including Hamilton (1805
65). The bedrock of optimiza-
tion and optimal methods is thus considered to be Newton’s optimization followed by Hamiltonian and Lagrangian

Nuclear Engineering. DOI: https://doi.org/10.1016/B978-0-323-90618-0.00009-0

© 2022 Elsevier Inc. All rights reserved. 349
350 Nuclear Engineering

formulations and the Euler-Lagrange equations as equivalent forms for the stationarity of a functional (Lewis, Vrabie,
& Syrmos, 2012; Rao, 2009).
At the simplest level, one or many maxima or minima of a function are found by setting its first derivative to zero
and then using the second derivative, or the Hessian matrix, to determine the nature of an extremum.
In the next step, with a quest to seek the best function a new calculus was developed by considering an independent
function as a variable and variational calculus gave, in the simplest Euler-Lagrange equation, the capability to find the
best function which minimized or maximized some functional. One common-sense illustration is that of a straight line
corresponding to the shortest distance between two points on a flat surface (Section 9.2.2). Any other function would
give a longer distance. Similarly, the function that gives the shortest distance between two points is a geodesic which
lies on the larger circle on the surface of that sphere. The terminologies commonly used for the best value or function
from calculus and variational calculus are optimization and optimal analysis.
Continuing the sequel in this book, with Chapter 4 covering the essential mathematics for a nuclear engineering
basic degree curriculum, and Chapters 5
7 covering formulations in the diffusion and transport equations and the math-
ematics and tools for Monte Carlo simulation, the simple understanding is that nuclear, and in a wider sense, radiation
transport is by no means easy to solve for realistic designs. Thus in Chapter 8, some illustrative and extensively used
codes and databases were described.
With all the learning from Chapters 4 to 8, optimization would require several computations based on intuition and
some learning. That would, of course, be a very inefficient way, maybe an impossible way, to obtain a nuclear reactor
design. One can therefore imagine the large amount of experimentation that must have gone into the design of the first
generation nuclear reactors described in Chapter 3.
Fortunately, strong and powerful computers were made in the 1960s and 1970s, notably the Control Data
Corporation (CDC), International Business Machines (IBM), and Cray computers. With the hardware came vector and
parallel processing capabilities and with all that the capability of performing huge computations became so simple. And
that opened the door for stochastic and heuristic optimization techniques which are so widespread these days.
Traditional optimization methods are based on the adjoint function with a variational formulation and deterministic
methods such as Dynamic Programming (DP) developed by Richard Bellman in the 1950s and widely used in optimal
control as the Hamilton-Jacobi-Bellman equation (Bellman, 1952; Bertsekas, 2010). A major thrust however is on heu-
ristic optimization methods using artificial intelligence, expert systems, neural networks, genetic algorithms (GAs), and
several other methods (Lee & El-Sharkawi, 2007; Nissan, 2019; Pardalos & Resende, 2002). The optimization field is
so vast and developed that it is not possible to consider more than just a few methods in this chapter, but the content
here is sufficient to give a reasonable insight into the methods and applications in nuclear engineering.

9.2 Deterministic optimization

9.2.1 Deterministic optimization without constraints
The simplest method of finding the minima or maxima of a continuous function is from its first and second derivatives.
As an example, for a two-variable function
1 2
f ðx1 ; x2 Þ 5 x 1 x1 x2 1 x22 1 x2 (9.1)
2 1
the partial derivatives are
@f @f @2 f @2 f
5 x1 1 x2 ; 5 x1 1 2x2 1 1; and 5 5 1: (9.2)
@x1 @x2 @x1 @x2 @x2 @x1


Setting both first partial derivatives equal to zero and solving for the extremum x , gives x1 ; x2 5 ð1; 2 1Þ. The
Hessian matrix is
2 2 3
@f @2 f
6 @x2 @x1 @x2 7
6 1 7
H56 26 7 (9.3)
@ f @ 2
f 7
4 5
@x2 @x1 @x22
 Optimization and variational methods Chapter | 9 351

with a determinant
 2 2 
@ f @ f @2 f @2 f 
jHj 5  2 2 2 5 ð1Þð2Þ 2 ð1Þð1Þ 5 1 (9.4)
@x1 @x2 @x1 @x2 @x2 @x1 
so that the extremum is a local minima. For a function of two independent variables, it is possible to plot the functions
as shown in Fig. 9.1 for the surface f ðxÞ and in Fig. 9.2 for the contours illustrating that at the minima f ðx Þ 5 0.
From Fig. 9.2, the minima is seen at x 5 ð1; 2 1Þ for which f ðx Þ 5 2 0:5.

9.2.2 Deterministic optimization with algebraic constraints

For optimizing a function f ðxÞ with N constraints gi ðxÞ 5 0; we write a Lagrangian
X
N
Lðf ; g; x; λÞ 5 f ðxÞ 1 λi gi ðxÞði 5 1; 2; 3?; NÞ: (9.5)
i51

As an example of a two-variable function with a single algebraic constraint, consider the optimization problem:
maxVðR; HÞ subject to the constraint of available surface area As for a cylinder of radius R and height H. The volume is
VðR; HÞ 5 πR2 H subject to a constraint of given surface area
As ðR; H Þ 5 2πR2 1 2πRH 5 K: (9.6)
An elementary consideration leads to a reduction of the two-variable problem to a single variable problem by substi-
tuting for the height of the cylinder
K
H ð RÞ 5 2R (9.7)
2πR
when the volume can be written as
1
V ð RÞ 5 KR 2 πR3 : (9.8)
2
The extremum
pffiffiffiffiffiffiffiffiffiffiffiffi can be found by differentiating the above to get the optimum for which the optimal radius and height
are R 5 K=6π, and H  5 2Ro , respectively, and the volume is
 3=2
 K
V 5 2πR 5 2π
3
:
6π

FIGURE 9.1 Surface plot of f ðxÞ 5 12 x21 1 x1 x2 1 x22 1 x2 showing

x1 ; x2 contours.
352 Nuclear Engineering

FIGURE 9.2 Contour plot of 12 x21 1 x1 x2 1 x22 1 x2 5 C.

With the Lagrangian (Eq. 9.5), L, expressed as the function to maximize (or minimize) V, and the constraint condi-
tion As 2 K 5 0, multiplied by a Lagrange multiplier λ (Lewis et al., 2012)
rLj 5 rVj 1 λrAs j 5 0 (9.9)

since, at the extremum, rLj 5 0, so that the equations are
@V  @As 
j 1λ j 50
@R @R
and
@V  @As 
j 1λ j 5 0:
@H @H
This gives H  5 2R , and the Lagrange multiplier
rV  R
λ52 j 52 : (9.10)
rAs 2
Fig. 9.3 shows the permissible R 2 H contours for surface constraints (almost straight lines) and the volume curves.
In Fig. 9.2, there were no constraints but here the maximization or minimization must be performed within the specified
constraints. The lines and curves move towards the right as values are increased.
Table 9.1 shows the optimum volumes V  for values of As 5 K 5 10; 20; 30; 40; and 50 cm2.
The optimal values R ; H  are indicated with crosses. Clearly, as the surface area increases, the optimal volume
increases from 2.43 to 27.14 cm3.
The area constraint, shown in Fig. 9.4, along with the volume contours athey move upwards for varying K show
that the curves become tangential, that is, the normal vectors on both coincide with each other at the optimal point
R 5 1:0301 cm, H  5 2:0601 cm with volume V  5 6:8671 cm3 which is a maxima as shown in Fig. 9.5.
It is seen that the volume contours shift towards the right touching each other tangentially at the optimal value for
the height for which the difference in the two values HV 2 HA goes to zero at the optimal radius. This represents the
geometrical interpretation of the Lagrange multiplier as the point where volume and surface area contours are tangential
to each other.

9.2.3 Optimal solution with a system of first-order ordinary differential equation constraints
In the previous section, the best values of independent variables were found to maximize a function subject to an alge-
braic constraint.
 Optimization and variational methods Chapter | 9 353

FIGURE 9.3 Objective function (volume) and constraint (area) curves.

TABLE 9.1 Constrained maximization of the volume of a cylinder.

K (cm2) 10 20 30 40 50

R (cm) 0.7284 1.0301 1.2616 1.4567 1.6287
H  (cm) 1.4567 2.0601 2.5231 2.9135 3.2574
V  (cm3) 2.4279 6.8671 12.6157 19.4231 27.1446

FIGURE 9.4 Surface area constraint for As 5 20 cm2 and a family of

objective function (volume) curves.

Consider now the problem of finding a best function which maximizes an objective function
ðb
ε 5 Fðx; yðxÞ; y0 ðxÞÞdx (9.11)
a

An example is to find a function that minimizes the distance between two points Aðxa ; ya Þ and Bðxb ; yb Þ on a flat sur-
face. Clearly there are an infinite number of functions gðxÞ that pass through these two points and satisfy the boundary
354 Nuclear Engineering

FIGURE 9.5 Objective function V versus radius R for a surface area

constraint.

conditions at A and B. One such example is the function gðxÞ 5 p 1 qx 1 rx2 where p; q; r are constants found by satisfy-
ing the boundary conditions. For an arc on gðxÞ the objective function is
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
0
Fðx; yðxÞ; y ðxÞÞ 5 1 1 ðy0 Þ2 : (9.12)

Since the distance along the curve, using Pythagoras’ Theorem for a right angle triangle relating the arc length ds to
the other two sides of the triangle dx and dy, is
ðb ð b qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
s 5 Fðx; yðxÞ; y0 ðxÞÞdx 5 1 1 ðy0 Þ2 dx (9.13)
a a

Just like Newton’s calculus for varying the independent variable to find the extremum of a function, here the func-
tion yðxÞ is varied so that a new function considered is gðxÞ 5 yðxÞ 1 εhðxÞ; then Eq. (9.11) is written as
ðb
ε 5 Fðx; gðxÞ; g0 ðxÞÞdx (9.14)
a

In terms of the small parameter ε, differentiating ε with respect to ε gives

ðb
dε dF
5 dx (9.15)
dε a dε

which is
ðb  
dε @F dx @F dg @F dg0
5 1 1 0 dx (9.16)
dε a @x dε @g dε @g dε
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Note that for Fðx; gðxÞ; g0 ðxÞÞ 5 1 1 ðg0 Þ2, the first two partial derivatives are zero while
@F g0
0
5 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
@g 1 1 ðg0 Þ2
and Eq. (9.16) becomes
ðb !
dε g0 0
5 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi h ð xÞ dx (9.17)
dε a 1 1 ðg0 Þ2
 Optimization and variational methods Chapter | 9 355

Integrating by parts and setting the derivative ε0 5 0,

0 1 0 1 0 1
ðb 0 0 ðb 0
dε B g C B g C B g C
5 @qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiA dhðxÞ 5 @qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiAhð xÞjba 2 hðxÞd@qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiA 5 0 (9.18)
dε 0 2 0 2 0 2
a 1 1 ðg Þ 1 1 ðg Þ a 1 1 ðg Þ

The first part of Eq. (9.18) can be made zero by choosing hðxÞ such that hðaÞ 5 hðbÞ 5 0 while the second term says
that
0 1
d B g0 C
@qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiA 5 0 (9.19)
dx 0 2
1 1 ðg Þ
0
which means that the term in the brackets is a constant in x. Thus g ðxÞ is also a constant and a function that has a con-
stant slope is a straight line. Therefore our assumption that gðxÞ 5 p 1 qx 1 rx2 was a trial function could only be true if
the constant r 5 0:
The purpose of doing this exercise from Eq. (9.16) onwards for gðxÞ considered as a second-order polynomial was
to make the understanding easy. In general, a functional F can be written as in Eq. (9.11); for that case, Eq. (9.16)
leads to the general form of Eq. (9.18) which is known as the Euler-Lagrange equation:
@F d @F
2 50 (9.20)
@y dx @y0
To conclude: the Euler-Lagrange equation is the stationarity condition in variational calculus for an unconstrained
optimization similar to the first derivative being zero in Newton’s calculus.

9.2.4 Optimal solution with a system of first-order ordinary differential equation constraints
Now consider the next step in which the aim is to find a best function u of one or several variables to maximize a func-
tional JðuÞ subject to N first-order ordinary differential equation (ODE) constraints x_ i ðtÞ 5 fi ðx; u; tÞ for i 5 1; 2; 3; ?; N.
This optimization of a functional rather than a function requires variational calculus and is classified as an optimal con-
trol problem.
A system is described by the equation of state
x_ 5 f ðx; u; tÞ (9.21)
with a performance index (PI)
ðt
ε5 Fðx; u; tÞdt: (9.22)
0

A control u is implemented causing the system to move to a state u; subsequently, a small change is made in the
control δu; such that jδuj # ε is made over the time domain ð0; tÞ. This causes a change in x and x_ since the time rate of
x changes from dx=dt to dðx 1 δxÞ=dt with a change δx_ which is related to the change in
@f @f
δx_ 5 δf 5 δx 1 δu 1 OðεÞ (9.23)
@x @u
causing a change in the PI
ðt 
@F @F
δε 5 δx 1 δu dt 1 Fðx; tÞjt5t δt 1 OðεÞ (9.24)
0 @x @u
Multiplying Eq. (9.12) by λðtÞ integrating over t and adding to Eq. (9.13) gives
ð t    
@F @f @F @f
δε 5 1λ δx 2 λδx_ 1 1λ δu dt 1 Fjt5t δt 1 OðεÞ (9.25)
0 @x @x @u @u
356 Nuclear Engineering

Integrating λδx_ by parts,

ð t    
@F @f _ @F @f
δε 5 1 λ 1 λ δx 1 1λ δu dt 1 Fjt5t δt 2 λðt Þδxðt Þ 1 λð0Þδxð0Þ 1 OðεÞ (9.26)
0 @x @x @u @u
Therefore λ should satisfy the differential equation
@F @f @H
1 λ 1 λ_ 5 1 λ_ 5 0: (9.27)
@x @x @x
Then, the change in the PI is
ðt  
@F @f
δε 5 1λ δudt 1 Fjt5t δt 2 λðtÞδxðtÞ 1 λð0Þδxð0Þ 1 OðεÞ: (9.28)
0 @u @u
Consider now the boundary conditions.
When time t is specified then there is no uncertainty in the end points and δt 5 0, so that the coefficient of δt does
not need to be zero. In that case
δε 5 2 λðt Þδxðt Þ 1 λð0Þδxð0Þ 1 OðεÞ (9.29)
Further if the end-point is given, then δxðt Þ 5 0 and λðt Þ is not known. In case the end-point is free, then δxðt Þ 6¼ 0
so that λðt Þ 5 0.
When time t is not specified then xðt 1 δtÞ 5 xðt Þ 1 δxðt Þ 1 OðεÞ. If x is free then
Fjt5t δt 2 λðt Þδxðt Þ
can be simplified
 using the state equation x_ 5 f which can be understood to be δx 5 f δt, so that
Fjt5t δt 2 λðt Þδxðt Þ 5 Fjt5t 2 λðt Þf δt. Free and fixed boundary conditions are applied as specified in the problem
(Table 9.2).
Eq. (9.18) becomes
ðt  
@F @f
δε 5 1λ δudt 1 OðεÞ (9.30)
0 @u @u
For δε $ 0, the value of δu can be chosen to be such that
 
@F @f
δu 5 2 ε 1λ : (9.31)
@u @u
With a positive definite square term in the integral
ðt  
@F @f 2
δε 5 2 ε 1λ dt $ 0: (9.32)
0 @u @u
The above condition is written as a minimum
ðt  
@F @f 2
δε 5 ε 1λ dt # 0 (9.33)
0 @u @u

TABLE 9.2 Adjoint boundary conditions at initial and final time ð0; tÞ.

x Condition Time

Specified Unspecified
xðtÞ Free λð t Þ 5 0 λð t Þ 5 0


xðtÞ Fixed λðt Þ unspecified Fjt5t 2 λ t f
xð0Þ Free λð0Þ 5 0 Initial time specified
xð0Þ Fixed λð0Þ unspecified Initial time specified
 Optimization and variational methods Chapter | 9 357

and since δε cannot be negative, therefore the Hamiltonian function written as

H 5 F 1 λf (9.34)
where the stationarity conditions are the state equation
@H
5f (9.35)
@λ
the adjoint, or co-state, equation
@H
5 2 λ_ (9.36)
@x
and the optimal control equation
@H
5 0: (9.37)
@u
In nuclear engineering, an optimization problem could be the maximization
Ð of power for a given amount
Ð of fuel in
some configuration. This would be as formulated as: maximize P 5 Σf ðxÞφðxÞdx subject to No 5 N ðxÞdx and
x_ i ðtÞ 5 fi ðx; u; tÞ for i 5 1; 2:
The Lagrange multiplier is a sensitivity coefficient and represents a powerful method to estimate the change in an
optimal cost function per unit change in the constraint. It has thus been used for design sensitivity studies leading to
optimization in nuclear reactor design.
The second-order neutron diffusion equation could be expressed as two first-order equations with state variables
x1 5 φ; x2 5 φ_ giving two first-order ODEs.
X
Lðφ; λ; xÞ 5 Σf ; φ 1 λi fi: (9.38)
i

Consider a system described by the coupled first-order ODEs

x_ 1 ðtÞ 5 x2 ðtÞ  f1 ðx; u; tÞ (9.39)
x_ 2 ðtÞ 5 2x2 ðtÞ 1 uðtÞ  f2 ðx; u; tÞ (9.40)
with boundary conditions x ð0Þ 5 0; x ð2Þ 5 ½5 2T , Find a function uðtÞ such that the objective function, or PI,
ð
1 tf 2
J ð uÞ 5 u ðtÞdt (9.41)
2 0
is maximized. The optimality analysis consists of the following steps:
1. Write the Lagrangian in terms of the objective function, the Lagrange multiplier(s), and the constraint(s),
2. Apply the stationarity conditions, and
3. Solve the equations to obtain the optimal control.
In this problem, the Lagrangian is
1
Lðx; u; tÞ 5 u2 ðtÞ 1 λ1 f1 1 λ2 f2 (9.42)
2
with the stationarity conditions (state equations) giving the constraints
@L
x_ 1 ðtÞ 5 5 f1 (9.43)
@λ1
@L
x_ 2 ðtÞ 5 5 f2 (9.44)
@λ2
and the adjoint (Lagrange multipliers) or co-state equations
@L
2 λ_ 1 ðtÞ 5 50 (9.45)
@x1
358 Nuclear Engineering

@L
2 λ_ 2 ðtÞ 5 5 λ1 ðtÞ 2 λ2 ðtÞ: (9.46)
@x2
The above are solved to give
λ1 ðtÞ 5 C1 (9.47)
λ2 ðtÞ 5 C1 1 C2 e t
(9.48)
and the optimal control u
@L
5 u 1 λ2 ðtÞ 5 0: (9.49)
@u
Therefore, the optimal control is


u 5 2 λ2 ðtÞ 5 2 C1 1 C2 et : (9.50)
All that is required now is to find the constants C1 and C2 . The optimal “trajectories” are then
C2 t
x1 ðtÞ 5 2 C1 t 2 e 2 C3 e2t 1 C4 (9.51)
2
and
C2 t
x2 ðtÞ 5 2 C1 2 e 1 C3 e2t : (9.52)
2
From the boundary conditions, the constants are:
C1 5 2 7:2918; C2 5 1:1870; C3 5 2 6:6983; and C4 5 2 6:1048:
Then, the PI is found as
ð tf 
 1  1 2 1


J ðu Þ 5 ½u ðtÞ dt 5
2
C1 tf 1 C22 e2tf 2 1 1 2C1 C2 etf 2 1 5 16:7507: (9.53)
2 0 2 2
The optimal functions are plotted in Fig. 9.6; this represents the optimal solution.

FIGURE 9.6 Optimal control u ðtÞ, trajectories x1 ðtÞ; x2 ðtÞ and adjoint
function λ2 ðtÞ.
 Optimization and variational methods Chapter | 9 359

The same analysis is carried out using symbolic computing in MATLABs using the dsolve function (Wang, 2007)
with the steps below:
% step 1 declare the symbolic variables
syms x1 x2 p1 p2 u; % two equations so two Lagrange multipliers p1 p2% step 2 give the state equations
Dx1 5 x2;
Dx2 5 -x2 1 u;
% step 3 write the cost function (inside the integral)
syms g;
g 5 0.5*u^2;
% step 4 write the Hamiltonian
syms p1 p2 H;
H 5 g 1 p1*Dx1 1 p2*Dx2;
% step 5 Adjoint (Co-state) equations
Dp1 5 -diff(H,x1);
Dp2 5 -diff(H,x2);
% step 6 solve for control u
du 5 diff(H,u);
sol_u 5 solve(du, 0 u0 );
The solution has the following variables
sol_a 5
x2: [1x1 sym]
x1: [1x1 sym]
p1: [1x1 sym]
p2: [1x1 sym]
and the optimal control u is found as: sol_u 5 -p2
% Substitute u in the state equations
Dx2 5 subs(Dx2, u, sol_u);
% convert symbolic objects to strings for using 0 dsolve0
eq1 5 strcat(0 Dx1 5 0 ,char(Dx1));
eq2 5 strcat(0 Dx2 5 0 ,char(Dx2));
eq3 5 strcat(0 Dp1 5 0 ,char(Dp1));
eq4 5 strcat(0 Dp2 5 0 ,char(Dp2));
sol_h 5 dsolve(eq1,eq2,eq3,eq4);
% BC x1(0) 5 x2(0) 5 0; x1(2) 5 5; x2(2) 5 2;
conA1 5 0 x1(0) 5 00 ;
conA2 5 0 x2(0) 5 00 ;
conA3 5 0 x1(2) 5 50 ;
conA4 5 0 x2(2) 5 20 ;
sol_a 5 dsolve(eq1,eq2,eq3,eq4,conA1,conA2,conA3,conA4)
x1sol 5 sol_a.x1;
x2sol 5 sol_a.x2;
p2sol 5 sol_a.p2;
g1 5 matlabFunction(x1sol);
g2 5 matlabFunction(x2sol);
p2 5 matlabFunction(p2sol);
%u 5 -p2;
x 5 0:0.1:2;
xx1 5 feval(g1,x); xx2 5 feval(g2,x); pp2 5 feval(p2,x); u 5 -pp2;
plot(x,xx1,0 g0 ,0 LineWidth0 ,2);
hold on
plot(x,xx2,0 b0 ,0 LineWidth0 ,2)
hold on
plot(x,u,0 k0 ,0 LineWidth0 ,2)
legend(0 x10 ,0 x20 ,0 u0 )
grid on

The above program gives the functions x1 ; x2 ; u of Fig. 9.6.
360 Nuclear Engineering

Exercise 9.1: A first-order system is described as

dθðtÞ
5 2 aðθðtÞ 2 θa Þ 1 buðtÞ
dt
where a and b are constants. Find the optimal control uðtÞ which minimizes J
ð
1 T 2
J5 u ðtÞdt
2 0
θðt 5 0Þ 5 θa 5 25 C, and the final state is required to be θðt 5 T Þ 5 20 C.
1. Sketch the optimal control u ðtÞ and the temperature θ ðtÞ, for a 5 0:25; b 5 0:5 and T 5 4 s and calculate the opti-
mal value J  .
2. Sketch the Lagrange λðtÞ and explain how it can be used to find another optimal for different conditions.
Ð1
Exercise 9.2: Maximize J 5 0 ½xðtÞ 1 uðtÞdt subject to x_ ðtÞ 5 1 2 uðtÞ2 ; xð0Þ 5 1 by writing the Hamiltonian and apply-
ing the stationarity conditions. Sketch uðtÞ; λðtÞ; xðtÞ for the optimal solution and obtain the value of J.

9.2.5 Optimal discrete control (Pontryagin maximum principle)

In the previous section, the optimal control was a continuous function of the independent variable. When the control is
discrete, the best control u is that which gives the maximum value of the Hamiltonian.
Given that the functional to be maximized is
ð3
J 5 ðx 2 uÞdt (9.54)
0

for a system described by

x_ 5 x 1 u; (9.55)
with initial and final conditions xð0Þ 5 2, xð3Þ free, with admissible controls (minimum and maximum)
u 5 ½ um u M  5 ½ 0 2 
the Hamiltonian is written in the same way as before:
H 5 ðx 2 uÞ 1 λðx 1 uÞ: (9.56)
It is rearranged as H 5 ð1 1 λÞx 1 ð1 2 λÞu to show that a sign change takes place for λ . 1. Thus in order to maximize
H, um should be used for λ . 1. Note that the stationarity condition, Eq. (9.37), cannot be applied since the control is
not continuous now.
Applying the stationarity condition for the adjoint equation
@H
5 2 λ_ (9.57)
@x
gives the adjoint equation 1 1 λ 5 2 λ, _ which is easily solved to give λðtÞ 5 Ce2t 2 1; applying the boundary condition
λð3Þ 5 0, C 5 expð3Þ: This gives λð0Þ 5 19:0855, λð3Þ 5 2 1; at the switching point, the time T is found from λðT Þ 5 1;
which gives T 5 2:3069:
In the range 0 # t # T, the control u 5 uM 5 2 and in T # t # 3, u 5 um 5 0 would maximize J: This discrete optimal
control solution is referred to as the bang-bang solution since one value is applied for some time and then the control is
abruptly switched to another value.

Exercise 9.3: Show that the solutions for Eq. (9.44) x_ 5 x 1 u for u 5 0 and u 5 2 are xðtÞ 5 0:1991et and xðtÞ 5 2 and
calculate J:

An application of bang-bang control is in the placement of fuels of different enrichments in various spatial domains.
 Optimization and variational methods Chapter | 9 361

9.3 Controller design and optimization

Nuclear reactor control, achieved by the movement of control rods or by burnable poison distributed in the core, can be
modeled by the point kinetic equations (PKE) introduced in Section2.1 which is a set of two non-linear first-order equa-
tions that can be combined into one second-order ODE. The reactor can be controlled by implementing reactivity
changes which alter the neutron density nðtÞ and bring the reactor back to the critical state. The manner in which a reac-
tor is controlled has to ensure that no unsafe conditions are attained. Another situation is where a reactor is producing
power according to some required load which may vary with time. In that case the reactor must be operated in a load-
following mode. In both the situations, a controller must be designed and implemented accordingly.
In engineering systems, a control system would typically have some overshoot η and a finite (non-zero) response
(settling) time τ s , and will not therefore respond instantly. While an ideal controller cannot be designed, an optimized
controller can be designed to give a response within acceptable safety limits.
In this section, the mass-spring-damper (MSD) equation, a second-order ODE, is considered to represent the PKE
and the response is obtained for a given step (reactivity) input. The response is analyzed in terms of η and τ s .
The solution of the MSD
d2 x dx
m 2
1 b 1 kx 5 f ðtÞ (9.58)
dt dt
for a step response uðtÞ 5 1; t $ 0, is found by taking the Laplace transform, so that Eq. (9.58) becomes

 1
ms2 1 bs 1 k X ðsÞ 5 : (9.59)
s
Now defining the natural frequency ωn , the critical damping coefficient Cc and the damping coefficient ζ as
rffiffiffiffi
k pffiffiffiffiffiffiffiffiffi b
ωn 5 ; Cc 5 4km; ζ 5
m Cc
Eq. (9.59) is written in terms of partial fractions and the coefficients in the numerator are then determined to get a
useful form for inversion:
"
 #
1 1 s b ω2n 1 2 ζ 2
X ð sÞ 5 2
1

 : (9.60)
k s ðs1ζωn Þ2 1 ω2n 1 2 ζ 2 mω2n 1 2 ζ 2 ðs1ζωn Þ2 1 ω2n 1 2 ζ 2

XðsÞ is readily inverted to give the compact time-dependent displacement (after some algebraic simplifications)
" ( qffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffi )#
1 2ζωn t ζ
xðtÞ 5 12e cosωn 1 2 ζ t 1 pffiffiffiffiffiffiffiffiffiffiffiffi2ffi sinωn 1 2 ζ 2 t :
2
(9.61)
k 12ζ
To understand this solution, consider some special cases. First, for a mass-spring system without any damping 5 0,
1
xðtÞ 5 ½1 2 fcosωn tg (9.62)
k
which means that a unit step force would set the system into an oscillatory never ending motion as shown in the peri-
odic function of constant amplitude in Fig. 9.7. For ζ , 1, the displacement oscillates with an overshoot until it settles
down to a steady gain as shown for the solutions for ζ 5 0:2; 0:4; 0:8; 1:0 in Fig. 9.7. Each of these cases has an over-
shoot η which decreases in magnitude as the damping ratio increases. Similarly, the settling time τ s decreases with an
increasing damping ratio. For higher damping when ζ . 1 the square roots in Eq. (9.61) are complex and since
siniθand cosiθ are real numbers, there is no oscillatory motion. All these cases are shown in Fig. 9.7 for normalized
spring constant k:

Exercise 9.4: A system is described by the state equations

ρ2β
n_ 5 nðtÞ 1 λCðtÞ
l
362 Nuclear Engineering

FIGURE 9.7 Response xðtÞ versus ωn t of a second-order system due to

a unit step function.

FIGURE 9.8 A proportional-integral-derivative

controller.

β
C_ 5 nðtÞ 2 λCðtÞ
l
where n 5 nðtÞ; c 5 cðtÞ β, l and λ are constants given below. The system is in equilibrium for t # 0 when cðtÞ 5 βn0 =λl.
A step input ρðtÞ 5 ρ0 . 0; is applied at t 5 0; ρðtÞ 5 0; t # 0 and causing a jump in nðtÞ. Solve for the system response
to compute nðtÞ for 0 , t , 200 s for ρ0 ðρ0 =β 5 0:1; 0:2; 0:5Þ and plot your results. The data for this exercise is:
β 5 0:0079, l 5 0:0001 s, and λ 5 0:0767 s21 .
Now consider one of the simplest controllers, a proportional (P), integral (I), derivative (D), or proportional integral
derivative (PID), controller which provides corrective action from the feedback of the plant (nuclear reactor) to take a
feedback from the output signal yðtÞ, compare it with a desired signal r ðtÞ to produce an error signal eðtÞ which is fed
into the controller which subsequently produces a control signal uðtÞ to drive the plant as desired, as depicted in
Fig. 9.8. The plant output yðtÞ is then optimized with respect to the control parameters.
The PID control signal
ð
de
uðtÞ 5 KP eðtÞ 1 KI eðtÞdt 1 KD (9.63)
dt
has a transfer function (TF)
KD s2 1 KP s 1 KI
U ð sÞ 5
s2
which modifies the TF (Eq. 9.59) to
XðsÞ KD s2 1 KP s 1 KI
5 3 : (9.64)
RðsÞ s 1 ðb 1 KD Þs2 1 ðk 1 KP Þs 1 KI
The open-loop system response for ðm 5 1; b 5 3; k 5 46:5; ζ 5 0:2Þ is shown in Fig. 9.9.
The objective is now to reduce the overshoot η, decrease the settling time τ s , and increase the steady gain of the
reactor (plant).
 Optimization and variational methods Chapter | 9 363

FIGURE 9.9 Open-loop system response for a unit step

function.

FIGURE 9.10 Unit step responses: open-loop response TF, P, PD, and
PID responses.

For a preliminary analysis, the closed-loop parameters (Eq. 9.64) KP ; KI and KD are taken randomly. As KP is
increased, the overshoot increases and the steady value also increases; this is followed by increasing KD which reduces
τ s . Finally, KI is increased to improve the steady gain. For KP 5 500; KI 5 100 and KD 5 200, the rise time τ r , the
time at which the response first crosses the amplitude-one line, is 0.0111 s: All these cases namely, the open-loop
response, TF and the responses for P, PD, and PID are shown in Fig. 9.10.
The PD response rises sharply to a response of B1.0 then falls and stabilizes to a steady value of 1.0 by B2 s:
Even the PID response has not been able to rise to a steady value of 1.
There are several methods of tuning a PID controller, the oldest being the Nichols-Ziegler Method (Burns, 2001)
which can readily give reasonably good response parameters. Here, the MATLAB tuning command is used to remove
the overshoot and give the response shown in Fig. 9.11 with tuned parameters listed in Table 9.3.
The tuned response for the baseline parameters gave KP 5 178:0441, rise time ðτ r Þ 0.0735 s, settling time ðτ s Þ 2.51
s, overshoot ðηÞ 73.4%, and peak amplitude 1.32. The final improvement gives the parameters listed in Table 9.3 with a
steady peak amplitude 1.099 and rise and settling times τ r 5 0:944s, settling time τ s 2.27 s, and no overshoot.
364 Nuclear Engineering

FIGURE 9.11 PID-tuned response compared with the baseline response.

TABLE 9.3 Proportional integral derivative controller tuning parameters.

Parameter Tuned Baseline

KP 35.4172 500
KI 129.7872 100
KD 2.4162 200
τ r (s) 0.944 0.0111
τ s (s) 2.27 —
η (%) 0 0
Peak 1.099 NaN

Exercise 9.5: The closed-loop system for a PI controller based on an error signal yðtÞ and input gðtÞ is given by

d2 y
 dy dg
1 0:02 1 kp 1 ki y 5 10 :
dt2 dt dt
When compared with a MSD
d2 x dx
M 1 D 1 Kx 5 0
dt2 dt
for which the characteristic polynomial is s2 1 2ξω0 s 1 ω20 where ξ and ω0 denote relative damping and the undamped
natural frequency
1. draw a block diagram of the system,
2. determine the controller parameters kp ; ki for critical damping ξ 5 1 with assumed values of ω0 Að0; 1Þ, and
3. sketch the response and comment on the dynamic response of the system.
Another set of optimal tools is the linear quadratic controller (LQR) for optimizing the control when the plant model
is linear and the PI is a quadratic function of the state variables and control. The plant can then can follow a reference
trajectory (Burns, 2001; Lewis et al., 2012) as applied to a PWR with an optimal linear quadratic Gaussian (LQG)
 Optimization and variational methods Chapter | 9 365

control with the robust integral sliding mode technique (Vajpayee et al., 2021a, 2021b) to track a reference trajectory
with external disturbances and parametric uncertainties.
With optimized techniques on a PID, LQR, LQG a nuclear reactor can adjust itself to efficiently follow the load
requirement.

9.4 Dynamic programming

DP is used in this chapter as a deterministic optimization method in which an optimal problem is structured into a num-
ber of multiple stages and considered as smaller sub-optimal problems. It is a powerful method used for deterministic
problems, such as the Traveling Salesman Problem, to find an optimal policy and has been applied to decision-making
under uncertainty as well.
In nuclear reactor refueling, the objective is to find the optimal refueling pattern at the end of each cycle when
once- or twice- burnt fuel is replaced and the remaining is shuffled. This has a bearing on the revenue requirement for
producing energy in a cycle.
This section considers a simple example to illustrate DP; a more detailed description with relevance to fuel reload-
ing in nuclear reactors is discussed in Section 13.2.
The reloading optimization problem can thus be expressed as
X
N X
N
minTc 5 min C ½xi ; ei ; fi  5 min Ci (9.65)
i51 i51

where Ci is the total discounted revenue requirement for producing energy Ei in the ith cycle, xi ei , and fi are the state
of the reactor, the reload enrichment and batch fraction at the beginning of the ith cycle.
In refueling optimization, a stage would represent a cycle at the end of which several possible states, defined by the
fuel types (in the sense of enrichment and burnup) are available and a decision has to be made on the reloading pattern.
The essence of DP is that
1. A problem is configured into N stages consisting of various states Pi;k where i can denote the fuel types available
and k represents the stage number.
2. At each stage, a decision is made on which state Pi;k11 to move to in the ðk11Þth stage; this decision will influence
the power distribution, criticality control (possibly by way of poison distribution), burnup behavior and cycle length
in the next cycle. Of course, this decision will be based on several detailed coupled neutronics-depletion simula-
tions, and hence will be computation-intensive.
3. An optimal path is found by starting at the end (the Nth stage), in an adjoint or backward manner with defined goals,
moving to find the sub-optimal solution at the ðN 21Þth stage, and proceeding to the initial state.
The DP computation is thus expressed as a recursive algorithm; a simple example being to calculate the factorial of
an integer or Fibonacci numbers.
function x 5 fact(n)
if n , 5 1
x 5 1;
else
x 5 n.* fact(n-1);
end
end

Consider the five-stage optimization problem, with the starting node as Stage 1 and the terminal node as Stage 5,
with three intermediate stages as depicted in Fig. 9.12.
In this shortest path problem, the objective is to find the path from start to end which minimizes the total distance
traveled. Such problems have been solved by Djikstra’s algorithm by generating a minimum path tree by scanning all
neighboring nodes from a starting node.
In this section, consider the recursive algorithm
fs ðiÞ 5 min ci;j 1 fs11 ðjÞ (9.66)
366 Nuclear Engineering

FIGURE 9.12 A five-stage forward network for finding the shortest route.

TABLE 9.4 Distances cij between nodes i and j in a network.

Node 1 2 3 4 5 6 7 8 9 10
1 0 2 5 6 — — — — — —
2 — 0 — — 3 4 6 — — —
3 — — 0 — 3 2 4 — — —
4 — — — 0 4 3 5 — — —
5 — — — — 0 — — 1 4 —
6 — — — — — 0 — 6 3 —
7 — — — — — — 0 3 3 —
8 — — — — — — — 0 — 3
9 — — — — — — — — 0 4
10 — — — — — — — — — 0

Starting from the last stage N, the algorithm Eq. (9.64) gives the distances as.

where fs ðiÞ is the cost associated with a transition from state i to the end state node N in the stage s and ci;j is the cost
associated in taking the system from state i to state j, listed in Table 9.4. In this problem, the assumption is that only
forward transitions are allowed.
Stage s 5 4
f4 ð8Þ 5 3
f4 ð9Þ 5 4
Moving back another stage:
Stage s 5 3

c5;8 1 f4 ð8Þ 5 1 1 3 5 4
f3 ð5Þ 5 min
c5;9 1 f4 ð9Þ 5 4 1 4 5 8

c6;8 1 f4 ð8Þ 5 6 1 3 5 9
f3 ð6Þ 5 min
c6;9 1 f4 ð9Þ 5 3 1 4 5 7
and

c7;8 1 f4 ð8Þ 5 3 1 3 5 6
f3 ð7Þ 5 min
c7;9 1 f4 ð9Þ 5 3 1 4 5 7
At Stage 3, the minimum cost paths from states 5, 6 and 7 are: 5
8
10, 6
9
10, and 7
8
10, respectively.
Moving now to Stage 2, the calculations are as follows:
 Optimization and variational methods Chapter | 9 367

For stage s 5 2
8
< c2;5 1 f3 ð5Þ 5 3 1 4 5 7
f2 ð2Þ 5 min c2;6 1 f3 ð6Þ 5 4 1 7 5 11
:
c2;7 1 f3 ð7Þ 5 6 1 6 5 12
At this stage, the minimum node is thus found to be 2
5 and 2
6 and 2
7 are rejected.
8
< c3;5 1 f3 ð5Þ 5 3 1 4 5 7
f2 ð3Þ 5 min c3;6 1 f3 ð6Þ 5 2 1 7 5 11
:
c3;7 1 f3 ð7Þ 5 4 1 6 5 10
Similarly the route 3
5 is the best out of 3
5, 3
6, and 3
7.
8
< c4;5 1 f3 ð5Þ 5 4 1 4 5 8
f2 ð4Þ 5 min c4;6 1 f3 ð6Þ 5 3 1 7 5 10
:
c4;7 1 f3 ð7Þ 5 5 1 6 5 11
This calculation again shows that the intermediate node (node 5) is the best from node 2.
Stage s 5 1
8
< c1;2 1 f2 ð2Þ 5 2 1 7 5 9
f1 ð1Þ 5 min c1;3 1 f2 ð3Þ 5 5 1 7 5 12
:
c1;4 1 f2 ð4Þ 5 6 1 11 5 17
This stage calculation says that the rout 1
2 is the best of all options. The optimal path is thus found to be
1
2
5
8
10 with a minimum cost of 9 units.

9.5 Stochastic optimization

Stochastic methods, equipped with the elaborate mathematical and physical foundations of transport phenomena, and
capability of modeling realistic 3D configurations, have powerful techniques which efficiently search for an optimal
solution from a very large search space. This section illustrates GAs for constrained optimization of a simple two-
variable function (Section9.1) and describes some other methods and applications.

9.5.1 Genetic algorithms

The GA method is used as an optimization method for large problems where better computational efficiency, in com-
parison with deterministic methods, can be achieved.
The GA method (Haupt & Haupt, 2004; Holland, 1975) is based on the “survival of the fittest” biological evolution-
ary processes and has been adapted in computational algorithms that carry out a random-search for the best values of
one or several variables that maximize (or minimize) an “objective” function. A typically large search space is scanned
with the help of algorithms incorporating concepts such as genes (bits), chromosomes (bytes), fitness function (objec-
tive function), and “selection” tests of genetic crossover and mutation. Thus, “fitter” successive generations are obained
to ultimately “converge” to an optimal as shown.
The cylinder volume optimization is repeated with the GA method using the MATLAB program given in Haupt and
Haupt (2004).
It is useful to go through the calculation details for the first two iterations given below. The simulation parameters
are listed in Table 9.5. Note that simulations will have a much larger number of variables, population size, number of
iterations, and length of a chromosome.
After the simulation parameters are fixed, as shown in Fig. 9.13, the initial population of chromosomes is generated
and evaluated.
The maximization problem (Section 9.2.2, Eq. 9.8) is converted to a minimization problem with a sign change and
each chromosome is evaluated for its fitness. The chromosomes in binary form with 4 bits, are converted into real num-
bers in the search space for radius RAð0; 2Þ; thus a chromosome 1111 with a value 15 would correspond to the real
number 2:0. Thus for a chromosome in the range 0
15, the linear mapping in the range 0
2 would give the required
368 Nuclear Engineering

TABLE 9.5 Simulation parameters for genetic algorithm optimization.

Parameter Value/description
Population size (no. of chromosomes) 10
No. of iterations 10
Length of a chromosome (bits) 4
Selection fraction 0.50
Crossover strategy Single bit
Mutation rate μ 0.15


Objective function (Eq. 9.8) 2 12 KR 2 πR 3
Search range for radius R RAð0; 2Þ

FIGURE 9.13 Flowchart of the genetic algorithms

optimization method.

transformation giving a four digit accuracy. A 16-bit chromosome would give 216
1 5 65,535 mapping with R 5 2 to
give an accuracy 1025.
i chromosome number cost fn
1 1010 1.3333 2 5.8866e 1 00
2 1101 1.7333 2 9.7285e-01
3 1010 1.3333 2 5.8866e 1 00
4 1111 2.0000 2 5.1327e 1 00
5 0101 0.6667 2 5.7358e 1 00
 Optimization and variational methods Chapter | 9 369

6 0111 0.9333 2 6.7791e 1 00

7 1000 1.0667 2 6.8539e 1 00
8 1011 1.4667 2 4.7551e 1 00
9 1111 2.0000 5.1327e 1 00
10 0101 0.6667 2 5.7358e 1 00

The fitness of the chromosomes is sorted with the best being ranked first out of 10 chromosomes.
Sorted Cost (lowest obj. fn. first)
i number obj. fn.
1 1.0667 2 6.8539e 1 00
2 0.9333 2 6.7791e 1 00
3 1.3333 2 5.8866e 1 00
4 1.3333 2 5.8866e 1 00
5 0.6667 2 5.7358e 1 00
6 0.6667 2 5.7358e 1 00
7 1.4667 2 4.7551e 1 00
8 1.7333 2 9.7285e-01
9 2.0000 5.1327e 1 00
10 2.0000 5.1327e 1 00

The chromosomes generated above give a minimum, or best value, 26.8539 and a mean value 23.2340.
The next step in the optimization is to improve upon the initial population by selected the best and creating the next
generation by crossover and mutation. In this computation the number of matings is 3 obtained from ceil (popsize-
keep)/2 where the ceiling of the difference between the population size and the chromosomes to keep (fraction 0.5 as
set in Table 9.5). The probability density function (PDF), f ðiÞ for chromosome i is calculated by its value of the objec-
tive function divided by the sum of all values. From the PDF, the cumulative distributive function (FðiÞ )is calculated
(Section 4.9) as shown below. For 10 chromosomes, 5 were kept and used to calculate PDF and cumulative distribution
function (CDF) values.
i f(i) F(i)
1 0.3333 0.3333
2 0.2667 0.6000
3 0.2000 0.8000
4 0.1333 0.9333
5 0.0667 1.0000

Note, as stated before, that the CDF is a monotonically increasing function rising to the value one.
The mates are selected by three uniform random numbers ξ1 ; ξ 2 generated in the range (0,1):
i ξ1 ξ2
1 0.9941 0.6520
2 0.3977 0.9061
3 0.6533 0.1331

A mate is selected from the random numbers compared with the CDF values, for example, for the first value
ξ1 5 0.9941, the last bin of the CDF (chromosome 5) is selected.
i Mate1 Mate2
1 5 3
2 2 4
3 3 1

The next generation of chromosomes are obtained from crossover and mutation:
Mating using single-point crossover

i chromosome
1000
0111
1010
1010
370 Nuclear Engineering

0101

Mutation using mutrate ðμÞ

i chromosome
1000
0010
1010
1010
0101

The new chromosomes are re-evaluated for their fitness:

i chromosome number obj. fn.
1 1000 1.0667 2 6.8539e 1 00
2 1000 1.0667 2 6.8539e 1 00
3 1001 1.2000 2 6.5713e 1 00
4 0110 0.8000 2 6.3915e 1 00
5 1010 1.3333 2 5.8866e 1 00
6 1010 1.3333 2 5.8866e 1 00
7 1010 1.3333 2 5.8866e 1 00
8 0101 0.6667 2 5.7358e 1 00
9 1011 1.4667 2 4.7551e 1 00
10 0010 0.2667 2 2.6071e 1 00

Sorted Cost (lowest cost first)

i number obj. fn.

1 1.0667 2 6.8539e 1 00
2 1.0667 2 6.8539e 1 00
3 1.2000 2 6.5713e 1 00
4 0.8000 2 6.3915e 1 00
5 1.3333 2 5.8866e 1 00
6 1.3333 2 5.8866e 1 00
7 1.3333 2 5.8866e 1 00
8 0.6667 2 5.7358e 1 00
9 1.4667 2 4.7551e 1 00
10 0.2667 2 2.6071e 1 00
minimum cost of population minc(1) 5 2 6.8539
mean cost of population meanc(1) 5 2 5.7428

This process continues until the maximum number of iterations specified in the simulation parameters.
Each iteration ends with a minimum cost and a mean cost (plotted in Figs. 9.14
16):
minimum cost of population minc(1) 5 -6.8539
mean cost of population meanc(1) 5 -5.3445

Table 9.5 shows GA results for As 5 20 cm2 for which the deterministic results (Table 9.1) are R* 5 1.0301 cm,
H* 5 2.0601 cm, V* 5 6.8671 cm3. For all the GA runs, the selection fraction was 0.5, mutation rate was 0.15 and each
“chromosome” was of length 8 bits. The search space specified for the radius was (0, 2). The GA results for the optimal
R and H are in good agreement with the ‘exact’ result and convergence is seen in the fourth generation while the mean
fluctuates randomly. For the fourth row of results listed in Table 9.6, the GA result for optimum volume (6.8670)
matches the deterministic exact result (6.8671) to three decimal places.
The “quality” of results from a GA simulation is dependent on the specified simulation parameters for which there
is no formal method and the selection remains an “art” rather than a science.
Such random methods are feasible when the number of variables is large. In principle, GA is efficient in a “parallel”
computing environment with benefits of computational speed-up.
 Optimization and variational methods Chapter | 9 371

FIGURE 9.14 Genetic algorithm results for minimization (N 5 10,

M 5 100, Nbits 5 4).

FIGURE 9.15 Genetic algorithm results for minimization (N 5 100,

M 5 100, Nbits 5 4).

FIGURE 9.16 Genetic algorithm results for minimization (N 5 100,

M 5 100, Nbits 5 8).
372 Nuclear Engineering

TABLE 9.6 Genetic algorithm results for optimized cylinder volume (V  K 5 20) cm2.

Population size No. of bits Iterations ðR ; H ; V  Þ

10 4 100 1.0667, 2.1334, 6.8539
100 4 100 Same as above
100 8 100 1.0275, 2.0350, 6.8670

9.5.2 Particle swarm optimization

Particle swarm optimization (PSO), first introduced by Kennedy and Eberhart (Kennedy & Eberhart, 1995) in 1995 is a
swarm intelligence (meta-heuristic) algorithm inspired by the movement of a flock of birds with an objective to survive
by looking for a safe place to land with the availability of food. Since the time PSO was introduced, several variants of
the method have been developed. In nuclear engineering the method has been extensively applied in-core design optimi-
zation, controller design optimization (Domingos, Schirru, & Pereira, 2006; Mousakazemi, 2021; Mousakazemi et al.,
2018) and several other areas
The PSO algorithm, in the standard form, is a simple two step procedure; in the first, the position x and velocity v
of particles in a swarm are initialized and evaluated using the objective function. The best values pbt for each particle in
the tth generation and the best values for all particles in the tth and previous generations, pg , are obtained. With this
information and a set of weighting factors wi ; c1 ; c2 with randomness modeled by random numbers ξ, the velocity is
updated for the next iteration as:


vt11 5 wi vt 1 c1 ξ1 pbt 2 xt 1 c2 ξ2 ðpg 2 xt Þ: (9.67)
It is worth mentioning that this is not truly a velocity in the sense of distance divided by time; it is given this name
most probably since it is a parameter that travels to the next iteration where the position of each particle is updated as
follows:
xt11 5 xt 1 vt11 : (9.68)
The iteration continues until the objective function converges to some prescribed tolerance ε. The variables in
Eq. (9.67) and Eq. (9.68) are formally defined as
c1;2 correction factors
pg global best (combination of values)
pbt particle best position at the tth iteration
vt velocity at iteration step t
vt11 velocity at iteration step t 1 1
wi inertia weight
ξ1;2 uniform random number Eð0; 1Þ.
The PSO method is simple in its standard form and is best illustrated by an exercise.

Exercise 9.6: Consider a minimization problem for the objective function given as Eq. (9.1) with

f ðx1 ; x2 Þ 5 12 x21 1 x1 x2 1 x22 1 x2 for a search space 24 # x1 ; x2 # 4. For a hand calculation of two iterations, use the
following values: Population size (Pop) 5 5, c1 5 c2 5 1:5, No. of iterations 5 10, w 5 0:8 and fill Table 9.7.
Steps for the first iteration
Step 1. Generate 5 random numbers ξ in the range ð0; 1Þ first for v1 then for v2
Step 2. Generate 5 random numbers ξ in the range ð0; 2Þ first for x1 andx2 .
Step 3. For each pair x1 x2 calculate f ðx1 ; x2 Þ
Step 4. For the first iteration of each pair x1 :x2 is the best value since there are no values from the previous
iteration.
Step 5. Find the global best pair, that is, the x1 x2 pair with the lowest f value
 Optimization and variational methods Chapter | 9 373

TABLE 9.7 Optimization by the particle swarm optimization method (first iteration).

Particle Velocity Position Obj. Fn. Particle best

v1 v2 R H f ðR; HÞ pb1 pb2

1
2
3
4
5

This completes the first iteration.

 
Pbg 5
Steps for the second iteration
Step 1: update v1 , then v2 from Eq. (9.2).
Step 2: update R from Eq. (9.3), then update H from Eq. (9.3).
Step 3: find the particle best value pb1 from b
 this and previous iteration, then find p2
Step 4: find the best global value Pg 5
b
from this and the previous iteration.
The calculated values can be filled in a Table similar to Table 9.7.
The procedure in the PSO method is simpler than that in the GA method since there are no chromosomes, cross-
overs, or mutations.
Other optimization methods used are Simulated Annealing and the algorithms based on the observation of ants,
birds, fish and animals as they behave for survival: Ant Colony Optimization, Grey Wolf Optimization, and Cuckoo
optimization.

9.6 Applications of optimization in reactors

In nuclear engineering, the refinement of mathematical models and the development of powerful algorithms enabling
efficient computer simulation toward the design and operations of present and future nuclear power reactors makes it
possible to optimize designs and strategies to a high level of accuracy.
This section, building upon the introduction given in Section 9.1 provides further insight into the areas where opti-
mization methods are being applied and indeed, have considerable scope for refinement.

9.6.1 Multi-objective core optimization

Optimization for reactor design optimization is an extensive and challenging multi-discipline process involving coupled
phenomena (Section 4.4, Section 4.8) and several interdependent parameters mostly requiring multi-objective optimiza-
tion for a wide spectrum (Section 3.2) covering the primary system (core and associated systems), Balance of Plant,
back-end fuel cycle, and safety analyses.
The optimization methods require a search for the optimal design parameters for which a very basic and elegant the-
ory is provided by variational calculus especially Pontryagin’s bang-bang control (Section 9.2.5) while large problems
can be efficiently simulated by stochastic heuristic methods (Section 9.5) such as GA, PSO, and simulating annealing.
Optimization broadly falls into two categories namely design and operations. The design optimization process
begins from the conceptual and feasibility studies as for vSMR’s (Section 3.2.5) and Gen IV reactors (Section 3.5.3)
going through the licensing procedures and freezing of a final design. This is followed by the operational phase from
startup to normal operations, maintenance and unplanned shutdowns and periodic refueling.
For large operating baseload reactors mainly in the .700 MWe category (Section 3.1), optimization is performed
for better understanding operating parameters such as Doppler broadening due to temperature (Section 2.8) improving
operating parameters and to carry out safety-related simulations including source-term calculations (Section 3.5.4) such
as for the Fukushima Boiling Water Reactors (BWRs). The most suitable techniques are stochastic heuristic coupled
374 Nuclear Engineering

with numerical solutions such as nodal methods and the method of characteristics of the diffusion or transport equations
(Section 5.4.1, Section 5.6).
Typical illustrations of operational areas for optimization are prediction of the critical heat flux (CHF) and moisture
carryover (MCO) in a BWR. For both CHF and MCO prediction, the attractive optimization tools include arificial intel-
ligence (AI) data-driven models, Support Vector Regression, and Gaussian Process Regression for producing lookup
tables and AI-based predictive models in the face of non-linearity and uncertainty.
Several present-day optimization challenges need to be addressed for design changes in nuclear submarine reactors
(Section 3.3), basic design configurations of micronuclear reactors and space propulsion reactors (Section 3.7) and
nuclear fusion reactors (Section 3.6, Section 10.6). Some of these challenges for submarine reactors (Section 3.3.6)
could be itemized as follows:
1. minimize the overall reactor size and weight,
2. design long “lifetime cores” without refueling for B50 years,
3. design better, lighter and more effective neutron and gamma shielding,
4. increase the operating temperatures B600 F
700 F for better thermodynamic efficiency,
5. explore the use of advanced high burnup fuels such as U-Zr, U-Al and ceramics,
6. optimize the power conversion system,
7. plan for overcoming dead time from xenon 54135Xe (half life 9.2 hours) poisoning, and
8. simulate off-normal operating conditions with a potential for accident scenarios.

9.6.2 Pressurized water reactor core pattern optimization

A fuel reloading strategy is necessary to remove spent fuel, add new fuel, and shuffle the remaining fuel within the
core at the end of a cycle. This requires placement of fuels of differing enrichment as well as distributing the poison to
ensure criticality with changing nuclide concentrations. Load Pattern Optimization (LPO) is best achieved by stochastic
heuristic, and AI-driven meta-heuristic, optimization (Section 9.5) with GA (Section 9.4), PSO, Ant Colony
Optimization, Evolutionary Algorithms, Differential Evolutionary Algorithms (DEA) and DP (Section 9.5.1).
AI-driven meta-heuristics for LPO have shown significant speedups for the first core of the 1000 MWe PWR with
optimization to manage the radial power peaking factor (RPPF) under Xe equilibrium conditions. The maximum RPPF
of the obtained loading pattern (LP) is decreased more than that of the LP by the designer.
By similar methods, automated and optimized control rod pattern design in a BWR has been demonstrated to calcu-
late the axial power distribution to calculate keff and the shutdown margin.

9.6.3 Controller proportional integral derivative

An optimized controller design for nuclear power reactors is used to optimize the power output and follow a load pat-
tern in the presence of external disturbances and statistical uncertainties. Some designs are based on the PID (Section
9.3), LQR, and LQG controllers.

9.6.4 Radiation shielding

Radiation shielding design (Section 10.4.2) optimization requires that the best combination and configuration of materi-
als is found providing shielding from both neutron and gamma radiation. Thus, stochastic heuristic methods are again
an attractive choice and have been used to give optimal designs.
Extensive work in shield design optimization has been carried out using GAs (Cai, Hu, Pan, Hu, & Zhang, 2018;
Cai, Hu, Pan, Sun, & Yan, 2018) while multi-grid algorithms with better computational efficiency have been used by
Asbury (Asbury, Holloway, Fleming, Gallimore, & Martin, 2012).
GA combined with diffusion theory has been used for shielding design optimization (Abdul Rahman, Lee, &
Franceschini, 2018; Israeli & Gilad, 2018).
Simulation with the Monte Carlo N-Particle code has also been carried out in the area of detector shielding design,
Boron Neutron Capture Therapy and neutron source imaging (Cai, Hu, Pan, Hu, et al., 2018; Hegazy, Skoy, & Hossny,
2018; Li et al., 2016; Williamson, 2010).
In other studies (Yadollahi, Nazemi, Zolfaghari, & Ajorloo, 2016), the effect of boron in concrete has been studied
to find that the optimal mixture of concrete, used for neutron shielding, has a water
cement composition of 38%,
cement content of 400 kg/m3, 50% aggregate with Colemanite and 15% silica fume
cement. Similarly, the effect of
 Optimization and variational methods Chapter | 9 375

another strong absorber, gadolinium, has been studied (Park, Kim, & Yi, 2017) to find that in concrete Gd content up to
10 at.% concrete composite (10 cm 3 5 cm thick) shielding efficiency of around 86%. For multi-layered iron-water
shields (Fuse, Yamaji, & Miura, 1970), it was found that the optimum arrangement was a composite iron-water-iron
configuration, rather than a homogeneous mixture.

9.6.5 Some other applications of optimization

Other areas in nuclear engineering where optimization is applied is in gas centrifuge configurations, in the design of
criticality, in condenser control, and in the optimal placement of instrumentation and sensors in a reactor core.
In a centrifuge hall, for example, the design of an optimal cascade which gives the minimum specific cost of
enriched uranium, is a problem in the domain of combinatorial optimization.
As a final remark, optimization in nuclear engineering is an area of significant research output with emphasis
towards AI-based meta-heuristic stochastic methods and machine learning based on classification and regression.

Problems
1. A recursive algorithm
 for the
 minimum value, in this case the distance traveled to node j after n intermediate points
is Vn ðjÞ 5 min cij 1 Vn21 ðiÞ ; i 5 1; 2; 3; ?; N. Evaluate V1(4) 5 shortest distance from node 1 to node 4 with at most
1 intermediate point, the algorithm will use V1 ðj 5 4Þ 5 min cij 1 V0 ðiÞ ; j 5 1; 2; 3; ?; N.
2. Consider the well-known Bounded Knapsack problem where the weight wi and value vi of N items is given and the objec-
tive is to fill a bag with the maximum value of items while not going beyond a permissible weight W. P It is also allowed to
M
fill
PM x i copies of each item but not exceeding c copies This optimization problem can be written as max i51 vi xi subject to
wi xi # W and 0 # xi # c. Comment on the classification of this problem in terms of computational complexity as a
i51
NP
complete or as a NP
hard problem. How does it compare with the use of DP for solving a NP
hard problem.
3. How would the shortest route problem be solved by the GA method described in Section 9.5.1? Would the results
be the same if DP follows a backwards calculation while GA follows a forward calculation?
4. In Section 9.3 (Controller design and optimization) comment on the objective function and its relevance to tracking
a given load requirement.
5. With the PSO method described in Section 9.5.2, maximize the two-variable function f ðR; H Þ 5 2πR2 1 2πRH sub-
ject to the constraint condition Eq. (9.6). The search space is 0 # R; H # 2. Compare your results with the results of
Section 9.5.1 obtained with the GA method.

Nomenclature
English lower case
b damping coefficient
k spring constant
l neutron generation time
n neutron density

English upper case

C precursor concentration
Cc critical damping coefficient
KD derivative gain
KI integral gain
KP proportional gain

Greek lower case

β delay fraction
ε error
ζ damping ratio
η overshoot
λ decay constant
φ neutron flux
376 Nuclear Engineering

ρ reactivity
τr rise time
τs settling time
ω frequency

Greek upper case

F objective function
Σf (macroscopic) fission cross section
H Hamiltonian
J integral performance index
L Lagrangian

Letter-like symbols
E integral performance index
H Hessian matrix

Abbreviations and acronyms

CDC Control Data Corporation
IBM International Business Machines
LQR linear quadratic control
PID proportional integral derivative

References
Abdul Rahman, F. B., Lee, J. C., & Franceschini, F. (2018). A deterministic method for PWR loading pattern optimization. Annals of Nuclear Energy,
117, 131
144.
Ahmad, S., Chang, B., Li, B., Yang, Q., & Liu, C. (2021). Mass optimization of the radiation shadow shield for space nuclear power system. Progress
in Nuclear Energy.
Asbury, S. T., Holloway, J. P., Fleming, R. F., Gallimore, A. D., & Martin, W. R. (2012). Multi-grid genetic algorithms for optimal radiation shield
design. PhD Thesis, Nuclear Engineering and Radiation Sciences, University of Michigan.
Bae, J. W., Betzler, B. R., Chandler, D., & Ilas, G. (2021). Automated fuel design optimization for high flux isotope reactor low enriched uranium
core design. EPJ Web Conference.
Bellman, R. (1952). On the theory of dynamic programming. Proceedings of the National Academy of Sciences of the United States of America.
Bertsekas, D. P. (2010). (3rd ed.). Dynamic programming and optimal control, (Vol. II). Athena Scientific.
Betzler, B. R., Chandler, D., Cook, D. H., Davidson, E. E., & Ilas, G. (2019). Design optimization methods for high-performance research reactor core
design. Nuclear Engineering and Design.
Burns, R. S. (2001). Advanced control engineering. Butterworth-Heinemann.
Cai, Y., Hu, H., Pan, Z., Sun, W., & Yan, M. (2018). Metaheuristic optimization in shielding design for neutrons and gamma rays reducing dose
equivalent as much as possible. Annals of Nuclear Energy.
Cai, Y., Hu, H., Lu, S., & Jia, Q. (2018). Optimization of radiation shielding material aiming at compactness, lightweight, and low activation for a
vehicle-mounted accelerator-driven D-T neutron source. Applied Radiation and Isotopes: Including Data, Instrumentation and Methods for Use in
Agriculture, Industry and Medicine.
Cai, Y., Hu, H., Pan, Z., Hu, G., & Zhang, T. (2018). A method to optimize the shield compact and lightweight combining the structure with compo-
nents together by genetic algorithm and MCNP code. Applied Radiation and Isotopes: Including Data, Instrumentation and Methods for Use in
Agriculture, Industry and Medicine.
Chham, E., et al. (2021). Fuel reloads optimization for TRIGA research reactor using Genetic Algorithm coupled with neutronic and thermal-
hydraulic codes. Progress in Nuclear Energy.
Domingos, R. P., Schirru, R., & Pereira, C. M. N. A. (2006). Particle swarm optimization in reactor core design. Nuclear Science and Engineering:
The Journal of the American Nuclear Society.
Fuse, T., Yamaji, A., & Miura, T. (1970). The optimum arrangement of laminated iron-water shields. Nuclear Engineering and Design.
Haupt, R. L., & Haupt, S. E. (2004). Practical genetic algorithms (2nd ed). John Wiley & Sons.
Hegazy, A. H., Skoy, V. R., & Hossny, K. (2018). Optimization of shielding- collimator parameters for ING-27 neutron generator using MCNP5. EPJ
Web Conference, 177, 02003.
Holland, J. H. (1975). Adaptation in natural and artificial systems: An introductory analysis with applications to biology, control, and artificial intelli-
gence, University of Michigan Press.
Israeli, E., & Gilad, E. (2018). Novel genetic algorithm for loading pattern optimization based on core physics heuristics. Annals of Nuclear Energy.
 Optimization and variational methods Chapter | 9 377

Kennedy, J., & Eberhart, R. (1995). Particle swarm optimization. IEEE International Conference on Neural Networks - Conference Proceedings.
IEEE.
Kumar, A., & Tsvetkov, P. V. (2015). A new approach to nuclear reactor design optimization using genetic algorithms and regression analysis. Annals
of Nuclear Energy.
Lee, K. Y., & El-Sharkawi, M. A. (2007). Modern heuristic optimization techniques: Theory and applications to power systems. John Wiley & Sons.
Lewis, F. L., Vrabie, D. L., & Syrmos, V. L. (2012). Optimal control (3rd ed.). John Wiley & Sons.
Li, H., Zhang, G., You, Z., Li, H., Zhang, G., & You, Z. (2016). Optimization design and simulation of a multi-source energy harvester based on solar
and radioisotope energy sources. Micromachines, 7(12), 228.
Mousakazemi, S. M. H. (2021). Control of a pressurized light-water nuclear reactor two-point kinetics model with the performance index-oriented
PSO. Nuclear Engineering and Technology.
Mousakazemi, S. M. H., Ayoobian, N., & Ansarifar, G. R. (2018). Control of the pressurized water nuclear reactors power using optimized proportio-
nal
integral
derivative controller with particle swarm optimization algorithm. Nuclear Engineering and Technology.
Nissan, E. (2019). An overview of AI methods for in-core fuel management: Tools for the automatic design of nuclear reactor core configurations for
fuel reload, (re)arranging new and partly spent fuel. Designs.
Pardalos, P., & Resende, M. (2002). Handbook of applied optimization. Oxford University Press.
Park, J.-S., Kim, J., & Yi, S.-H. (2017). Effects of Gadolinium in Fe based amorphous ribbons with high boron contents on the neutron shielding effi-
ciency. Annals of Nuclear Energy, 109, 365
369.
Rao, S. S. (2009). Engineering optimization: Theory and practice (4th ed.). Wiley.
Vajpayee, V., Becerra, V., Bausch, N., Deng, J., Shimjith, S. R., & Arul, A. J. (2021a). LQGI/LTR based robust control technique for a pressurized
water nuclear power plant. Annals of Nuclear Energy.
Vajpayee, V., Becerra, V., Bausch, N., Deng, J., Shimjith, S. R., & Arul, A. J. (2021b). Robust-optimal integrated control design technique for a pres-
surized water-type nuclear power plant. Progress in Nuclear Energy.
Wan, J., & Zhao, F. (2017). Optimization of AP1000 power control system setpoints using genetic algorithm. Progress in Nuclear Energy.
Wang, X. (2007). Solving optimal control problems with Matlab - Indirect methods. Raleigh, NC: North Carolina State University.
Williamson, M. R. (2010). Multivariate optimization of neutron detectors through modeling. Philosophy (London, England).
Yadollahi, A., Nazemi, E., Zolfaghari, A., & Ajorloo, A. M. (2016). Optimization of thermal neutron shield concrete mixture using artificial neural
network. Nuclear Engineering and Design.
Chapter 10

Monte Carlo simulation in nuclear

systems

10.1 Introduction
Criticality is central to all nuclear designs from the first critical assemblies, Godiva and Jezebel, and the first reactor
core calculations for nuclear power reactors (Section 3.1). Criticality calculations are relevant to the nuclear industry
where fissile material is transported, processed, and stored in the form of gases, liquids and solids, and for the design of
nuclear reactors and systems to determine the materials required, their amounts, and configurations.
This chapter takes the reader through representative and illustrative situations in the following areas:
1. criticality of bare fissile assemblies (Section 10.2),
2. criticality safety in the front-end of the fuel cycle (Section 10.3),
3. radiation moderation and shielding (Section 10.4),
4. fission neutronics (Section 10.5), and
5. fusion neutronics (Section 10.6).
A detailed simulation of the legacy Godiva bare assembly is demonstrated to give physical insight into its neutro-
nics; the energy- and space dependence of the flux and associated reaction rates.
In the front-end of the nuclear fuel cycle, the ore is processed using uranyl nitrate solution UO2(NO3)2 to produce
hexafluoride (UF6). In the back-end of the fuel cycle, decladded spent fuel is stored and processed in which it is dis-
solved in nitric acid to form uranyl nitrate used for reprocessing. MC simulations are carried out to determine the safety
of storage configurations,
Nuclear thermal power reactors (Section 3.1) require uranium enriched to typically 2
5 weight % while fast reac-
tors (Section 3.2) require higher enrichments and naval propulsion reactors could require highly enriched uranium
(HEU) (B90% enriched) just as in weapons grade uranium (WPG). Uranium enrichment is therefore a central part of
the nuclear fuel cycle for all applications of nuclear technology. From several demonstrated enrichment technologies,
commercial processes use gaseous diffusion (mechanical flow through a porous membrane) and gas centrifuge (gas sep-
aration by centripetal force in a rotating cylinder). In an enrichment plant, uranium hexafluoride (UF6) is transported
and stored at various stages in its processing from gas with composition varying from that of natural uranium composi-
tion to possible HEU. In the front-end of the nuclear fuel cycle, uranium ore mining and milling takes place, followed
by conversion from uranium oxide in the form of U3O8, commonly known as yellow cake to UF6 in a form (solid, liq-
uid, or gas) depending on the process conditions, followed by enrichment and fabrication.
UF6 gas is used in an enrichment plant to separate U235F6 from U238F6 by a process such as gas diffusion or a gas centri-
fuge. In gaseous diffusion, the gas is heated and passed through a porous membrane in which U235F6 passes faster than U238F6
due to their weight difference. Similarly in a fast-spinning centrifuge, typically 1150 Hz, the heavier gas flows to the sides
while the lighter gas stays closer. Each stage thus gives a small separation which, accumulated over a cascade with thousands
of stages, results in significant enrichment. At room temperature, UF6 is in the form of white crystalline material (density
5.09 g/cm3); its triple point is at a pressure of about 1.5 times atmospheric pressure, that is, 22 psia and at 64.05 C. Thus at
room temperature with normal atmospheric pressure, UF6 is in solid phase and sublimes to the gas phase (directly). Its liquid
phase occurs above the triple point so that in an enrichment plant, just the solid and gas phases are typically observed.
UF6 is transported to a cascade facility in large steel containers with about 12 tons of material stored as liquid but
cooled to solid form during transport. At the cascade facility, it is heated to the liquid phase and transferred through
pipes to the enrichment site. Since UF6 is corrosive, it is generally transported in nickel or stainless steel pipes. In cylin-
ders during storage, it forms a layer which inhibits corrosion. In enriched form, the UF6 is shipped out of the cascade

Nuclear Engineering. DOI: https://doi.org/10.1016/B978-0-323-90618-0.00010-7

© 2022 Elsevier Inc. All rights reserved. 379
380 Nuclear Engineering

facility in cylinders of varying capacity depending on the amount of enrichment. For a reactor fuel grade gas, the cylin-
ders contain typically about two and half tons material.
Depleted UF6 (waste stream from an enrichment process) is stored in large cylindrical tanks containing typically up
to 14 tons of material, filled as a liquid which cools and solidifies, occupying about 65% of the cylinder volume. At the
bottom of the cylinder, there is solid crystals and above it is the gas. When small cylinders are moved, the noise made
by the crystals can be heard just like that of table salt.
During its transportation and storage, the issues of criticality safety and impact to environment are of vital
importance.
Very large amounts of UF6 are transported and stored in enrichment plants. For example, the amount of UF6
required to produce 1 kg product gas to 4.4% U235 with feed assay 0.711% (natural uranium) and tails assay 0.23% is
about 8.7 kg U as UF6 requiring about 6.7 SWU. Thus a 100-ton production would require about 870 tons of reactor-
grade enriched material to be transported and at 6700 SWU/t, a total of 670,000 SWU. Compare this with the total
world enrichment capacity of 66,700,000 SWU/y which is sufficient for the fuel of 100 such reactors every year. To
produce 1 kg 90% U (WGU) in UF6 would require about 187 kg UF6 at 0.711% feed and 0.23% tails, with about 215
SWU. Thus 20 kg WGU/y would require 4300 SWU/y; thus is each centrifuge gives 5 SWU, then about 1000 centri-
fuges are required, and the material flow for 20 kg HEU/y is about 3.7 t/y. The largest enrichment facilities in the world
are Rosatom (27,654 SWU/y), Urenco (18,320 SWU/y), Orano (7500 SWU/y) and CNNC (6,750 SWU/y). All these
facilities require criticality analyses.
The quantities of interest are the keff of a single cylinder or hundreds or thousands of cylinders in their storage facility,
and the radiation dose. The system multiplication is a criticality calculation carried out in much the same way as for fissile
materials; standard analytical and computational tools are based on the diffusion, transport and Monte Carlo methods.
MC simulations are carried out for the design of detection systems, for example, in Prompt Neutron Activation
Analysis (PNAA) or in Thermal Neutron Activation Analysis (TNAA). A PNAA detection system is described in
Section 10.4 for determining the best moderator resulting in an optimal energy spectrum and a best physical configura-
tion to get a good signal and to protect the detector from radiation.
In radiation shielding, MC simulation is used for the design of shields from α, β, γ and neutron radiation.
Alpha emitters such as U-238 and Pu-239 may contain trace impurities which could possibly emit other radiations.
However, the range of α radiation is small and shielding is possible with a thin sheet of paper. Similarly the pene-
trating power of β radiation is not strong, typically just a few millimeters in tissue and hence they can be shielded
with a few mm thickness of a low-Z material such as plastic. Gamma radiation pose a serious radiation risk due to
their penetration power. Typical radiation shield materials in the nuclear industry are water, concrete, and iron.
The half value layer (HVL), the thickness of a material that reduces the intensity of a radiation by half, is typically
in the range 0.5
2 cm for a lead shield for 0.5
2 MeV gamma radiation. The next best shield is iron, followed by
concrete and water. As described in Section 1.3, gamma radiation follows an exponential attenuation in a shield
and a 1=R2 intensity reduction at a distance R. However, simple theoretical models use a buildup factor to account
for multiple scattering which is best estimated by Monte Carlo simulation. Neutron radiation also poses a serious
problem in the design of shields due to the penetration power of neutrons and the several possible reactions. A typ-
ical strategy used for the design of neutron shields is to first slow the fast neutrons with a hydrogenous material
and then absorb the neutrons. This can produce gamma rays which then need to be shielded by lead, iron or con-
crete as discussed. In nuclear reactors, openings are kept for piping, ducts and instrumentation; this provides chan-
nels for radiation steaming which is a challenge for MC simulation as discussed in Section 7.5 (variance
reduction).
In fission neutronics, MC simulations are very effective for core neutronics starting from the basic unit of a lattice
cell to a fuel assembly followed by a full-core analysis.
In this chapter, the three core designs considered are
1. Westinghouse AP-1000
2. Toshiba Gen-IV 4S, and the
3. Micronuclear heat pipe reactor (MNR).
Section 10.4 describes unit lattice cell MC simulations, extended to infinite assembly simulations and whole-core
simulations for these designs.
Finally, in Section 10.6, fusion neutronics are described with particular reference to the ITER design for which MC
simulations are useful in determining several engineering design parameters, for example, the First Wall (FW) loading,
the tritium breeding rate (TBR), and the displacements per atom (dpa) in the superconducting coils.
 Monte Carlo simulation in nuclear systems Chapter | 10 381

10.2 Bare critical assemblies

10.2.1 Godiva
A detailed Monte Carlo simulation for Godiva, as a follow-up to Section 2.10 and Section 6.3, is of great educational
value as it gives physical insight into the flux and associated reaction rates. A problem generally associated with large
simulations is that one can easily lose track of the physical significance of voluminous results.
The physical model of Godiva is a bare sphere of radius 8.741 cm with a composition of 1.02 wt.% U-234, 93.7 wt.
% U-235, and 5.27 wt.%U-238, density 18.74 g/cm3, mass 52.4254 kg, atomic density 5 4.79838 3 1022/cm3. The total
volume (2.7951 3 103 cm3) is divided into 50 zones of equal volume.
The input file BK10Gdva is listed in Annex A. The weight fractions and atomic fractions of uranium nuclides are
given in Table 10.1.
The nuclear data used is from endf66c, 92234.66c for uranium nuclides at temperature 293.6K. Simulation para-
meters are given in Table 10.2.
With 3000 particles per history, a total of 150 cycles out of which 60 cycles are skipped, and an initial guess
keff 5 1 for the purpose of estimating keff and the prompt removal lifetime τ p . As described in Section 7.2, M is the
number of source points generated in each simulation cycle, shown in Fig. 10.1.
Each cycle gives an estimate of the prompt removal lifetime τ ðaÞ p (abs) and the source points generated:
28
τ ðaÞ
p 5 0.967664 shakes (1 shake 5 10 s), τ pðcÞ 5 0.969156 shakes, and keff shown in Table 10.3.

TABLE 10.1 Material composition of Godiva.

Composition/nuclide U-234 U-235 U-238

22 21
Weight fraction 1.02010 3 10 9.37094 3 10 5.27053 3 1022
Atomic fraction 1.02511 3 1022 9.37677 3 1021 5.20719 3 1022

TABLE 10.2 MC simulation parameters.

M N K Ks keff
Varies (Fig. 10.1) 3000 150 60 1.0

FIGURE 10.1 Fission points simulated in each cycle.

382 Nuclear Engineering

TABLE 10.3 Estimates of keff with collision-, absorption and track length-estimators.

Estimate Average of 90 cycles

ðcÞ
keff 0.994040 (0.0014)
ðaÞ
keff 0.993941 (0.0014)
ðT Þ
keff 0.994835 (0.0012)
ðCAT Þ
, keff . 0.994272 (0.0012)

ðcÞ ðaÞ ðTÞ

keff 5 system multiplication (collision estimate), keff 5 system multiplication (absorption estimate), keff 5 system
ðCATÞ
multiplication (track length estimate), and , keff . 5 average system multiplication (collision, absorption, track
length estimate).
The above results compare with keff 5 0.9976 (0.0011) by Whalen et al. (Whalen, Cardon, Uhle, & Hendricks,
1991). Some important observations from the simulation results are:
1. The average neutron energy is 2.0621 MeV
2. Particles created per particle due to weight cutoff was 0.032885 at an energy 1.1819 3 1022 MeV
3. The number of neutron collisions per source particle is 4.063
4. The random numbers generated are 43,579,980 (B44 million)
5. The source tracks were 269,862 each with a weight 1
6. Total source tracks 272,187 with weight 1.0383 energy 2.0776 MeV
7. Source multiplication (n, xn) was 2325 tracks weight per source particle 5.4134 3 1023 and energy 3.7340 3 1023 MeV.
The number of collisions in each zone is fairly large, varying from 8369 in the outermost cell to 43,082 in the inner-
most cell. About 93.6% of the collisions are with the U-235 nuclide; this is also the atomic abundance of U-235. Since
the number of collisions is high in every cell, there is no need for variance reduction except for implementing a low
weight cutoff to avoid “wasting” time on low weight particles.
The average track mean free path varies between 2.6251 cm in the innermost cell increasing to 2.6676 cm in the out-
ermost cell. These numbers tell us that this is a simulation where we can expect good results due to the large number of
nonzero histories.
To get an idea of the gains and losses in the system, Table 10.4 gives two contributions, namely the source itself
and the small contribution from weight cutoffs. The total weight of 1.0383 must be balanced in the losses table.
Table 10.5 illustrates where the losses are taking place: the escapes, the weight cutoffs, captures, multiplication and
fission treated as a loss mechanism.
The convergence of keff is seen in Fig. 10.2; for 150 cycles, we can see that it is a good idea to skip a few cycles at
the beginning of a simulation.
After the first B ten cycles, Fig. 10.2 gives a satisfactory result. Of course, the number of skip cycles is more or
less a guess given at the beginning of a simulation.
Deciding which estimator to rely on for the keff estimate, Fig. 10.3 shows that the absorption and collision estimators
underestimate the tally; the collision estimator is above the absorption estimate and hence closer to the track length esti-
mate. It would be fair to conclude that since this is not a strongly absorbing medium, the TLE and collision estimators
would be expected to give more reliable results.
In the tallies, the standard and nonstandard ENDF/B reaction numbers for the reactions tallied in the simulation are
listed below
standard ENDF/B reaction numbers
1 total cross-section
2 elastic scattering cross-section
16 n,2n reaction
17 n,3n
18 n,f
102 (n,γ)
103 (n,p)
104 (n,d)
 Monte Carlo simulation in nuclear systems Chapter | 10 383

TABLE 10.4 Gains per source particle.

Mechanism Tracks Weight Energy (MeV)

source 269862 1.00000 2.06210
weight cutoff
0.032885 0.011819
Total 272187 1.03830 2.07760

TABLE 10.5 Losses per source particle.

Mechanism Tracks Weight Energy (MeV)

Escape 231,765 0.574920 0.925520
Weight cutoff 39,260 0.033177 0.011845
Capture 0 0.044765 0.027477
Down scattering 0 0 0.525060
(n,xn) 1162 0.0027056 0.021057
(n,f) 0 0.382730 0.566660
Total 272,187 1.038300 2.077600

FIGURE 10.2 System multiplication keff (TLE).

105 (n,t)
106 (n,He3)
107 (n,α)
nonstandard ENDF/B reaction numbers
2 2 absorption cross-section
2 6 total fission cross-section
2 7 fission nu
The tallies specified in the input file are the neutron current (F1), the neutron flux (F4), the energy deposition (F6),
the fission energy deposition (F7) and the reaction rates (FMn); the tally multipliers are the product of the atomic num-
ber density and the volume. The units of the tally multipliers are thus reactions/s.
The simulation gives the number of neutrons escaping as (F1:N 5 ) 5.75122 3 1021 0.0011, energy deposition
(F6:N 5 ) 1.23121 3 1023 0.0011 MeV/g and fission energy deposition (F7:N 5 ) 1.32154 3 1023 0.0011 MeV/g
(includes gamma ray heating; it is assumed that fission photons deposit energy locally).
384 Nuclear Engineering

FIGURE 10.3 System multiplication (average) keff after 60 skip cycles.

TABLE 10.6 Tallies for Godiva.

Group/Tally Thermal Intermediate Fast Total

E , 0:625 eV 0:625eV 2 100keV E . 100keV

21
Total reactions 0 1.54931 3 10 0.0094 2.49234 2.64727
0.0012 0.0012
Elastic scattering 0 0.126275 1.53455 1.66082
0.0094 0.0014 0.0014
Absorption(22) 0 6.50478 3 1023 3.82374 3 1022 4.47422 3 1022
0.0096 0.0019 0.0021
Radiative capture (102) 0 6.50478 3 1023 3.82374 3 1022 4.47422 3 1022
0.0096 0.0019 0.0021
Fission reactions (26) 0 2.04106 3 1022 3.62598 3 1021 3.83008 3 1021
0.0095 0.0012 0.0011
Fission neutrons (26) (27) 0 4.9529 3 1022 9.45185 3 1021 9.94714 3 1021
0.0095 0.0012 0.0011
(n,2n) 1 (n,3n) 0 0 2.59771 3 1023 2.59771 3 1023
0.0139 0.0139

The tallies in Table 10.6, for the thermal, intermediate, and fast neutron energy groups clearly show that most of the
activity is in the fast group.
The flux falls off with radius as shown in Fig. 10.4, as a classic one-group flux.
The center of the sphere has the highest flux as shown in Fig. 10.5.
The same picture of flux is shown in Fig. 10.6 with the central values about five times higher than the peripheral
values.
The value of the flux, normalized to the power P, is
1 νPφF4
φ5 211 k
3:1724 3 10 eff

where the energy from fission is 3:1724 3 10211 J and φF4 is the F4 scalar flux tally of MCNP.
The energy-dependent flux, shown in Fig. 10.7, exhibits high values at near source energies and is therefore a hard
spectrum. There is no Maxwellian at lower energies due to the absence of any hydrogenous medium.
Monte Carlo simulation in nuclear systems Chapter | 10 385

FIGURE 10.4 Neutron flux φðrÞ versus radius r.

FIGURE 10.5 Neutron flux φðrÞ versus xand y.

FIGURE 10.6 Neutron flux ðrÞ.

386 Nuclear Engineering

FIGURE 10.7 Neutron flux φðEÞ versus energy E.

FIGURE 10.8 Standard deviation versus cycle number.

The quality of the simulation results is illustrated in Fig. 10.8; the standard deviation of the collision-absorption-
TLE averaged keff decreases by a factor B10 by the 90th active cycle. pffiffiffiffi
In this simulation, the relative error was found to decrease as 1= N , the variance of the variance was found to
decrease as 1=N, and the FOM is constant as seen in Fig. 10.9.
The Godiva simulation of this section is meant to understand the qualitative aspect of the results as much as the
quantitative aspects. In any nuclear system, it is most important to have an idea of the overall space- and energy-
dependence of the flux. In a fast reactor, for example, the spectrum is similar to the Godiva spectrum while in a water
reactor, two Maxwellian peaks will be found; one for the source neutrons and the other for thermal neutrons. There rela-
tive magnitudes will indicate the reaction rates when the cross-section behavior of the predominant nuclides is also
understood.

10.2.2 Jezebel
The input file for the Jezebel (Chapter 2, Section 6.3) assembly (Rowlands et al., 1999) is listed in Annex B. The bare
assembly is modeled as a plutonium sphere (95.5 wt.% Pu-239 and 4.5 wt.% Pu-240) with a radius 5 6.385 cm,
density 5 15.61 g/cm3 and atomic fractions shown in Table 10.7.
Table 10.8 shows the tallies for Jezebel for the same MC simulation parameters and energy bins as for Godiva indi-
cating the same overall behavior, that is, hard spectrum and predominance of the fast neutrons.
 Monte Carlo simulation in nuclear systems Chapter | 10 387

FIGURE 10.9 Figure of merit versus cycle number.

TABLE 10.7 Material composition of Jezebel.

Composition/nuclide Pu-239 Pu-240

Weight fraction 0.955 0.045
Atomic fraction 0.955179 0.0448206

TABLE 10.8 Tallies for Jezebel.

Group/Tally Thermal Intermediate Fast Total

E , 0:625 eV 0:625eV 2 100keV E . 100keV

22
Total reactions 0 5.90774 3 10 1.41514 1.47422
0.0143 0.0012 0.0013
Elastic scattering 0 4.83311 3 1022 8.22303 3 1021 8.70634 3 1021
0.0143 0.0014 0.0014
Absorption (22) 0 1.76497 3 1023 8.76700 3 1023 1.05320 3 1022
0.0165 0.0032 0.0037
Radiative capture (102) 0 1.76497 3 1023 8.76700 3 1023 1.05320 3 1022
0.0165 0.0012 0.0037
Fission reactions (26) 0 6.78738 3 1023 3.11815 3 1021 3.18602 3 1021
0.0143 0.0011 0.0011
Fission neutrons (26) (27) 0 1.96806 3 1022 9.86400 3 1021 1.00608
0.0143 0.0012 0.0011
(n,2n) 1 (n,3n) 0 0 1.12502 3 1023 1.12502 3 1023
0.0140 0.0140

The tallies are: neutrons escaping from sphere (F1:N 5 ) 6.73781 3 1021 0.0008, energy deposition (F6:N 5 )
3.26504 3 1023 0.0011 MeV/g, fission energy deposition (F7:N 5 ) 3.54458 3 1023 0.0011 MeV/g (includes gamma
ray heating; it is assumed that fission photons deposit energy locally).
388 Nuclear Engineering

The results of the simulation are: keff 5 1.006261 0.0012 (combined average col/abs/tle), removal lifetime 5 0.37120
0.0023 (combined average col/abs/tle).

10.3 Criticality safety

10.3.1 Storage of interacting units
Typical benchmarks for critical assemblies (Whalen et al., 1991) and their storage as interacting units are as follows:
1. Enriched uranium metal (93.2% enriched) cylinders 2 3 2 3 2 unreflected array critical experiment (Wagner,
Sisolak, & McKinney, 1992) with cylinders 10.765 cm in height and 11.496 cm in diameter. These cylinders are
contained in cuboids referred to as a 2C8 unit. The spacing between the units is 2.244 cm in the x and y directions
and 2.245 cm in the z direction. The TLE keff of this system was found to converge to 1.0000 with MCNP version
4.2 with 3000 particles per cycle in 200 cycles and 20 skip cycles.
2. In the second experiment, the 2C8 unit with increased separation of approximately 11.98 cm, is reflected on six
sides of the cube by 15.24 cm of paraffin and gives a MC simulation result converged to 1.0000.
3. SP12: A composite array of four HEU (93.2%) cylinders and four cylindrical plexiglass containers filled with HEU
(92.6%) uranyl nitrate solution UO2(NO3)2 are critical.
4. SP16: an infinite number of slabs of uranyl fluoride solution UO2F2 contained in Pyrex glass and separated by
borated UO2F2.

10.3.2 Storage of uranium hexafluoride cylinders

All uranium hexafluoride operations in nuclear facilities are carried out within the regulations of the International
Atomic Energy Agency (IAEA), International Standards Organization (ISO), and United States Nuclear Regulatory
Commission (USNRC). In the United States, operations are conducted under the Nuclear Regulatory Commission
(Code of Federal Regulations, Section 10), American National Standards Institute (ANSI), Department of
Transportation and the Department of Energy.
During the handling, transportation and storage of fissile material, subcriticality within margins of safety is main-
tained. These safety limits are derived on the basis of one of two types of criteria: criteria based on the value of keff for
the system under analysis and criteria based on the critical value of one or more control parameters, such as mass, vol-
ume, concentration, geometry, moderation, reflection, interaction, isotopic composition and density, and with account
taken of neutron production, leakage, scattering and absorption (IAEA, 2014).
Cylinder sizes vary (USEC (United States Enrichment Corporation), 1995) from the 1S Model with a diameter of
1.5 inches (minimum volume 0.15 L) for 100 w/o U-235 enrichment and a maximum shipping limit of 0.45 kg, to the
48 H Model with a diameter of 48 inches and minimum volume 3964 L for maximum 1 w/o enriched U-235 in UF6
and maximum shipping limit 12,261 kg.
For radiation dose calculations, the source term is calculated from the modeling and simulation of transmutation of
nuclides for both neutron and gamma sources. The direct neutron source depends on the enrichment and is assumed
independent of the decay chain while the photon source depends on the decay (Su, 2015); the UF6 neutron source term
for a 48Y cylinder (12501 kg UF6) with 0.711 weight percent U-235 is 4.978 3 105 n/s with B62% in the range 0.4
MeV- 1.42 MeV. The neutron source term decreases with decreasing tails U-235 enrichment. The photon source term is
due to Th-231, Th-234, Pa-234, and Pa-234m in addition to the uranium nuclides. It is a factor B105 times greater than
the neutron source but the spectrum is towards lower energies B10 keV
0.2 MeV increasing steadily over a two-year
time period. Dose calculations were carried out using MCNP5 and ENDF/B-VII.0 with ANSI/ANS-6.1.1
1977 flux-to-
dose conversion factors.
For a filled cylinder, with 0.711 wt. % U-235 the dose rate from both neutrons and photons is 5.317 mrem/h inside
(B94% photon dose) while outside the cylinder it is 1.739 mrem/h (B88% photons). With distance, the dose rate falls
off by about four orders of magnitude at 100 m and another two orders by the next 100 m. At the boundaries of such
storage sites, the dose rates, are below the limits prescribed in 10 CFR Section 20.1301 (Dose Limits for individual
members of the public). When such filled feed (natural uranium in UF6) cylinders are stacked one or two layers up and
thousands of cylinders are placed on the ground in a storage yard, their combined effect, calculated from MC simula-
tions, is summarized as follows:
In a 10 3 10 array, the single, double and triple stack MC neutron (and photon) dose rates are 0.165, 0.183 and
0.191 (and 1.13, 1.19 and 1.21) mrem/h falling off in the radial direction to 2.58 3 1023, 4.01 3 1023 and
 Monte Carlo simulation in nuclear systems Chapter | 10 389

TABLE 10.9 Atom density of UF6 (atoms/b-cm).

Density (g/cc) U-235 U-238 F H

5.1 4.4168E-4 8.2860E-3 5.31340E-2 7.6800E-4

4.96 3 1023 mrem/h (and 8.19 3 1023, 1.39 3 1022 and 1.85 3 1022) mrem/h by 50 m and to 1.76 3 1025, 2.17 3 1025
and 2.70 3 1025 (and 3.70 3 1025, 5.14 3 1025 and 7.54 3 1025) mrem/h by 395 m. These MC estimates, with no cell
source have relative error within 1%, 1%
4% and 5%
8% for neutrons and within 1%
2%, 1%
3%, 3%
4% for
photons at 0, 50, and 395 m respectively.
In a 100 3 100 array, the single, double and triple stack MC neutron (and photon) dose rates are 0.209, 0.248, and
0.258 (1.16, 1.25 and 1.27) mrem/h falling off in the radial direction to 4.02 3 1022, 5.07 3 1022 and 5.56 3 1022
(6.19 3 1022, 7.58 3 1022 and 8.89 3 1022) mrem/h by 50 m and to 8.13 3 1024, 9.96 3 1024 and 1.04 3 1023
(1.22 3 1023, 1.45 3 1023 and 1.55 3 1023) mrem/h by 395 m.
These MC estimates, with no cell source have relative error within 1%
2%, 1%
3% and 2%
4% for neutrons and
within 1%, 3%
5%, 3%
5% for photons at 0, 50, and 395 m, respectively.
The effect of ground scattering is also considered for calculating radial and axial dose rates. This reduces the dose
rate.
The above results show that the photon dose forms the major component of the total dose; its relative importance
decreases with distance due to increased neutron skyshine (emitted radiation reaching a facility by scattering from the
atmosphere).
These calculations can be used to design storage facilities in terms of arrays and stacking for estimating the distance
to the boundary for safe and compliant radiation dose. A 10 3 10 array size with triple stacking, for example, would
require a site boundary of 80 m while a 100 3 100 array size with the same triple stacking would require 270 m to the
boundary.
Criticality calculations carried out for a 21/2 ton UF6 steel cylinder with 5% enrichment UF6 (material composition
listed in Table 10.9) at Oak Ridge National Laboratory using the SCALE system with ENDF/B-IV cross sections
(Broadhead, 1991) give a single unit keff 5 0.453 6 0.003 with an effectively infinite water reflector (SG 5 1).
The keff varies with the specific gravity of water; from 0.72 at zero SG increasing to a maximum of B0.82 for SG
0.02 then falling off to B0.445 at 0.5 SG and slightly increasing to B0.453 at SG 1.0.
For single 10- and 14-ton UF6 cylinders in an infinite water-reflected array (SG 5 1), keff 5 0.526 6 0.002 and
keff 5 0.533 6 0.003.
Sensitivity studies are carried out to investigate the effect of temperature due to density and resonance-capture varia-
tions and fuel location patterns. It is reported that keff is insensitive to temperature.
These studies seek optimization to get the best separation distance and to estimate the density and temperature
effects.

10.4 Radiation moderation and shielding

10.4.1 Radiation moderation for a neutron generator
In Prompt Gamma Neutron Activation Analysis (PGNAA), a sample is activated with a neutron source, characteristic
gamma rays are subsequently emitted, and measurement is performed simultaneously (IAEA, 2012). In some applica-
tions, such as the detection of explosives, thermal neutron activation analysis (TNAA) is used to enhance radiative cap-
ture reactions.
The placement of the sample, between the source and the detector determines the quality of the signal. MC simula-
tions are carried out to select the best moderator and to determine the minimum size and optimal configuration.
Powerful neutron sources are used such as californium-252 with a mixed energy spectrum and D-D fusion monoenergetic
source of energy 2.5 MeV emitting 1011 n/s. In such cases, an important consideration is the protection of a detector since
efficiency is degraded due to the effect of fast neutron bombardment on scintillators and germanium detectors.
When there is no medium in between a source and detector, the dose can be calculated by the geometrical attenua-
tion of the neutron flux, for example, a 14 MeV flux of 1.7 3 108 n/cm2/h gives a dose of 1 Sv/h (100 rem/h); compared
with the highest permissible dose of 10 μSv/h (1 mrem/h).
390 Nuclear Engineering

TABLE 10.10 Comparison of moderators for obtaining maximum flux at a detector.

Moderator Min. R (cm) Max. φth 108 n/cm2/s

Light water 78 8.3
Polyethylene 63 12
Heavy water 85 4.08
Graphite 130 1.8

For attenuation of fast neutrons in the range 1
2.5 MeV, MC simulations show that the best shielding is by a hydroge-
nous material such as polyethylene and the worst would be by a high scattering material such as graphite. Table 10.10 shows
results for three moderators to produce the lowest thermal flux on a detector with minimum thickness.
MC simulations have been used for the design optimization of a neutron source generator for PNAA (Ref) by
repeated simulations to show that an optimal design with lowest fast neutron and photon fluence at a detector for a
1011 n/s D-D generator heavy water is more effective than light water, polyethylene, beryllium, and graphite and that
the optimal radius of moderator of about 86 cm with a minimum lead thickness of 25 cm. This optimal design reduces
the thermal neutron flux to 2.9 3 107 n/cm2/s.

10.4.2 Radiation shielding

For neutrons, as mentioned above, the best shield will be a hydrogenous material such as water, followed by lead, then
low Z and high Z (water-lead) materials. Fast neutrons slowed down by water, for example, will produce
2.2 MeV gamma rays from the radiative capture reactions in hydrogen. These energetic gammas will require shielding
by a high-Z material such as iron or lead.
The HVL for 0.5, 1.0, 1.5 and 2.0 MeV photons is 0.51, 0.76, 1.27, and 1.52 cm for lead, and 3.30, 4.57, 5.84, and
6.60 cm for concrete, respectively.
A simple and practical demonstration of shielding illustrated in Fig. 10.10 shows a concrete shell, labeled cell “2,”
located 10 ft. (304.8 cm) from a 14 MeV point isotropic D-D source at the center of the sphere labeled “S” emitting 109
neutrons per second. In a neutron transport MC simulation, MCNP input file BK10Shld (Annex C), the neutron current,
surface flux, and radiation dose are estimated.
Results of the simulation for 400,000 histories and a lower energy cutoff 1 MeV are listed in Table 10.11.
For a void, the surface flux 5.37428 3 1027 n/cm2/s from simulation is equal to the exact value at the surface labeled
“1” in Fig. 10.10.
1
φðRÞ 5
4πR2
at R 5 304:8 cm which gives a dose of 2.79462 3 10216 Sv for one source neutron. The neutron flux-to-dose coeffi-
cients are used from the ICRP conversion factors (ICRP, 1996). For 1011 n/s, the dose is 100.6063 mrem/h. With a con-
crete shell, this dose is reduced to 1.2008 and 0.5117 mrem/h for shield thickness 70 and 80 cm, respectively. The
relative errors of the estimates are within 2.4%.
The calculation is done as follows:
1 rem 5 0.01 Sv, D-D fusion source 14 MeV 109 n/s
Dose 5 109 n=s 3 2.79462 3 10216 n=s Sv
3 3600 hs 3 105 mrem
Sv 5 100.6063 mrem/h.

10.5 Nuclear fission applications

In this section, MC simulations are demonstrated for unit lattice cells and reactor core models of the AP1000, Toshiba
4S and the micronuclear heat pipe cooled reactor (MNR).

10.5.1 Unit lattice cell and fuel assembly of the AP1000 reactor
The AP1000 unit lattice cell (Fig. 7.16) has fuel of three enrichments as shown in Table 10.12.
 Monte Carlo simulation in nuclear systems Chapter | 10 391

FIGURE 10.10 Concrete shell with surface and cell numbers.

The atomic densities and volume occupied by water (at 300K), fuel, helium gap and the cladding are listed in
Table 10.13. The quantity NV is used in the MCNP tally multiplier to convert the product of microscopic cross-section
and flux into a reaction rate.
In a PWR such as the AP1000, the water density at 600K decreases appreciably to B70% of its value at 300K so
that atomic densities used in a MC simulation are as given in Table 10.14.
The integral fuel burnable absorber (IFBA) consists of a mixture of zirconium and boron with atomic densities listed
in Table 10.15 (Laranjo de Stefani, Losada Moreira, Maiorino, & Russo Rossi, 2019).
The MCNP input file for the unit lattice cell, BK10AP10 is listed in Annex D. The simulation is carried out for
5000 histories per cycle for 150 cycles with 5 skip cycles and an initial guess keff 5 1. The neutron flux is fairly steady
as seen in Fig. 10.11. This is due to the small dimensions of the lattice cell; otherwise the flux would have the same
shape as for Godiva.
392 Nuclear Engineering

TABLE 10.11 Radiation dose across a concrete shield.

Tally Units Void Concrete

70 cm 80 cm
23
Neutron current Neutrons 1.0 9.7329 3 10 4.22892 3 1023
0.0122 0.0180
Surface flux n/cm2/s 5.37428 3 1027 8.08496 3 1029 3.50463 3 1029
0.0148 0.0244
Radiation dose Sv 2.79462 3 10216 3.33562 3 10218 1.42129 3 10218
0.0146 0.0239
mrem/h 100.6063 1.2008 0.5117

TABLE 10.12 AP1000 fuel atomic densities.

Enrichment w/o 2.35 3.4 4.45

3
Density (g/cm ) 10.47635 10.47635 10.47635
Mass fraction
U-235 0.0207146 0.0299696 0.0392243
U-238 0.8607574 0.8514884 0.8422196
O-16 0.1185280 0.1185421 0.1185561
Total 1.0000000 1.0000000 1.0000000
MCNP input N
U-235 5.56014e-04 8.04432e-04 1.05284e-03
U-238 2.28120e-02 2.25663e-02 2.23207e-02
O-16 4.67360e-02 4.67416e-02 4.67471e-02
Total 7.01040e-02 7.01123e-02 7.01206e-02

AP1000 atomic densities, are obtained from the MATLAB program % \Elsevier\Programs\Ch7_AP1000_MCNPinput.m.

TABLE 10.13 Atomic densities and volumes of cells in AP1000 unit lattice cell.

Region Atomic density ðNÞ atom/b Volume ðVÞ NV (atoms cm2/b)

Water 1.00367 3 1021 3.51510 0.35280
Fuel as in Table 10.12 2.10829 0.1478
24 22
Gap 1.00000 3 10 8.52588 3 10 8.5259e-6
22 24
Clad 4.34418 3 10 6.41741 3 10 0.0279

Fig. 10.12 shows the energy spectrum in the lattice cell; this is very different from the Godiva hard spectrum. In this
case, thermalization occurs due to neutrons scattering with water, and two peaks of comparable magnitudes are
observed. These correspond to thermal neutrons due to moderation and the fission spectrum due to source neutrons.
 Monte Carlo simulation in nuclear systems Chapter | 10 393

TABLE 10.14 AP1000 water atomic densities.

Water temp (K) 300 600

Density (g/cm3) 1 0.7
Mass fraction
H 0.1111111 0.1111111
O-16 0.8988889 0.8988889
Total 1.0000000 1.0000000
MCNP input N
H 6.69111e-02 4.68378e-02
O-16 3.34556e-02 2.34189e-02
Total 1.00367e-01 7.02567e-02

TABLE 10.15 AP1000 IFBA (zirconium diboride) atomic densities. Material density 5.42 g.cm3.

Isotope Weight fraction Atomic density (atoms/cm3)

B-10 0.0187 5010
B-11 0.1713 5011
Zr-90 0.416745 40090
Zr-91 0.090882 40091
Zr-92 0.138915 40092
Zr-94 0.140778 40094
Zr-96 0.02268 40096

FIGURE 10.11 Neutron flux φðrÞ versus x and y.

In the subsequent analysis, the following tallies listed in Table 10.16, are obtained in the fuel and water regions
1. Total reaction rate
2. Elastic scattering rate
394 Nuclear Engineering

FIGURE 10.12 Neutron flux φðEÞ versus E.

3. Absorption rate (in water and fuel regions)

4. Radiative capture
and the fission reaction rate and number of fission neutrons emitted in the fuel. Results are binned in three energy
groups corresponding to thermal E , 0:625 eV, intermediate 0:625eV 2 100keV and fast E . 100keV energies.
Neutrons undergo about nine times more collisions in water than in fuel regions, with the energy distribution shown
in Table 10.17. Elastic and absorption reactions are predominantly in the thermal and intermediate energy groups. In
fact, radiative capture is the absorption mechanism in these two lower groups.
Table 10.17 shows that fissions are predominantly in the thermal group, as expected (since the fission cross-section
is orders of magnitude higher at thermal energies). Subsequently, the number of fission neutrons emitted is also mainly
in the thermal group. We see that the number of neutrons emitted per fission, obtained as
 
νΣf φ
ν5  
Σf φ
is 2.4367, 2.4342, and 2.7579 for the thermal, intermediate, and fast groups, respectively. The cell energy deposition
tallies give 1.57493 (0.0012) MeV/g in water and 9.86552 (0.0011) MeV/g, in fuel. The fission energy deposition is
6.11183 (0.0011) MeV/g in the fuel region.
The unit lattice cell is part of a fuel assembly as shown in Fig. 10.13.
A detailed MCNP simulation for the AP1000 by Stefani et al. (Laranjo de Stefani et al., 2019) estimates kN for the
Beginning of Life (BOL) and End of Life of the core, and reactivity coefficients.
For a lattice with fuel rod including gap, Zirlo cladding and pitch, for various enrichments kN is 1.21029 6 0.00003
(1.58%), 1.32863 6 0.00004 (2.35%), 1.40462 6 0.00004 (3.20%), 1.41697 6 0.00004 (3.40%) and 1.21029 6 0.00003
(4.45%).
For the AP1000 assembly kN is reported for a number of configurations; for 4.45% enriched fuel with 24 Pyrex and
72 IFBA burnable absorbers, kN 5 1.26524 6 0:00008.

10.5.2 The Toshiba 4S reactor

The 4S Gen-IV design (Section 3.5) is based on the idea of a small low-power high temperature metal-cooled fast reac-
tor with no moving parts.
To realize such a design, enriched U-10Zr fuel is placed in hexagonal fuel assemblies in a liquid sodium pool mod-
erator with a single central absorber rod and a beryllium reflector (Koreshi & Hussain, 2014; Tsuboi, Arie, Ueda, &
Grenci, 2012).
In this thirty-year two-phase operation, with minimal hands-on maintenance, the first phase of fifteen years of opera-
tion is achieved by the gradual withdrawal of the central rod. Thus the beginning of cycle (BOC) reactivity should be
high enough such that the reactor is subcritical with the rod fully inserted and sufficiently critical to produce power.
TABLE 10.16 Reaction rates in an AP1000 unit cell.

Water Fuel

Thermal Intermediate Fast Thermal Intermediate Fast

E , 0:625 eV 0:625eV 2 100keV E . 100 keV E , 0:625 eV 0:625eV 2 100keV E . 100 keV
Total 3.75361E 1 01 0.0012 3.91317E 1 01 0.0005 1.40565E 1 01 0.0008 2.89720E 1 00 0.0012 4.29777E 1 00 0.0007 3.45535E 1 00 0.0009
Elastic 3.73608E 1 01 0.0012 3.91139E 1 01 0.0005 1.40468E 1 01 0.0008 1.51791E 1 00 0.0012 3.77071E 1 00 0.0007 2.81159E 1 00 0.0009
Abs 1.75349E-01 0.0013 1.77668E-02 0.0010 7.21758E-03 0.0053 3.06725E-01 0.0013 3.73804E-01 0.0019 2.48070E-02 0.0012
ðn; γÞ 1.75349E-01 0.0013 1.77668E-02 0.0010 9.11486E-05 0.0009 3.06725E-01 0.0013 3.73804E-01 0.0019 2.05502E-02 0.0009
396 Nuclear Engineering

TABLE 10.17 Fissions and multiplication in an AP1000 unit cell.

Thermal Intermediate Fast

E , 0:625 eV 0:625eV 2 100keV E . 100 keV

 
Fission Σf φ 1.07256E 1 00 0.0013 1.47861E-01 6.54692E-02 0.0016
0.0013
 
Fission neutrons νΣf φ 2.61351E 1 00 0.0013 3.59921E-01 1.80569E-01 0.0017
0.0013
ðn; 2nÞ 1 ðn; 3nÞ 0 0 2.58042E-03 0.0137

FIGURE 10.13 An AP1000 fuel assembly with 25 Pyrex burnable absorbers.

 Monte Carlo simulation in nuclear systems Chapter | 10 397

In the second phase, the gradual upward movement of the external beryllium reflector keeps the reactor critical. The
movement is upwards, against gravity to ensure passive safety. In the event of power failure, the reflector would fall
downwards and the reactor would return to a subcritical safe state.
The neutronic design challenge is thu s the BOC excess reactivity, the worth of the central absorber and the capacity
of the beryllium reflector.
MC simulation is the perfect tool for carrying out such an analysis. Compared with the AP1000 in the previous sec-
tion, the 4S is smaller and has fewer assemblies. Its electrical output is 10 MW in the uranium-fueled design and
50 MW in the plutonium-fueled design.
The movement of the liquid sodium metal is also upwards, by stationary electromagnetic pumps. This movement of
the coolant is another passive safety design feature. Again, a failure of the electrical power in the plant will lead to
subcriticality.
The issue in that case would be the heat removal from the core to prevent it from achieving melt-down conditions.
The 4S core model for a Monte Carlo code MCNP simulation, shown in Fig. 10.14 from the top and in Fig. 10.15
from the side, fits into a reflector cylinder with outer radius 60 cm and height 260 cm.
A standard MC simulation would thus carry out a preliminary pin-cell simulation to obtain the neutron spectrum,
the infinite multiplication kN and the effect of temperature and moderator density followed by a full-core simulation
for the effective multiplication factor keff with a central absorber control rod.

FIGURE 10.14 Top view of the 4S reactor core model.

398 Nuclear Engineering

FIGURE 10.15 Side view of the 4S

reactor core model (R 5 Reflector,
F 5 Fuel, A 5 Absorber).

TABLE 10.18 Physical parameters of the 4S reactor core.

Core U-10Zr

17% 19%
No. of assemblies 6 12
Mass of uranium (t) 3.0903 6.1805
Mass of uranium-235 (t) 0.5253 1.1743
24 3
Atomic density (10 atoms/cm )
U-235 6.1329 3 1023 6.8545 3 1023
U-238 2.9565 3 1022 2.8852 3 1022
22
Zr 1.0328 3 10 1.0328 3 1022

The 4S 10 MW(e) design, Table 10.18, consists of a cylindrical core with active core height 2.5 m, and diameter
1.16 m. The fuel (diameter 10.50 mm, length 2.50 m) is metallic U/10% Zr 17%
19% enrichment, density 15.6425 g/
cm3 with 169 (13 3 13) fuel rods in a triangular fuel pin arrangement (hexagonal array) fuel assembly with pitch
1.5 cm, and 18 assemblies, with pitch 20.2073 cm, comprising a heavy metal (U) inventory of 9.2708 t and a fissile
U235 inventory of 1.6996 t. The moderator/coolant is sodium metal (8t) operating in the temperature range 355
510 C.
 Monte Carlo simulation in nuclear systems Chapter | 10 399

Three unit-cells in the core configuration are:

1. the central cell with the control rod,
2. the 169-fuel pin 17% enriched fuel cell, and
3. the 169-fuel pin 19% enriched fuel cell.
Each lattice fuel cell has a hexagonal structure with fuel surrounded by liquid sodium moderator.
The source is sampled using the KSRC card at the center of each assembly. The KCODE card is used with 2000 his-
tories and 510 cycles with 10 skip cycles. The MCNP cross-section files used were: for sodium, endf66a (ENDF-6.1);
for zirconium, endf66b (ENDF-6.1); for natural hafnium endf60 (ENDF-6); for U235 and U238, endf66c (ENDF-6.5).
All evaluations used are at a temperature of 293.6K.
A preliminary pin-cell simulation for the 1693-fuel pin 17% enriched fuel cell gave kN 5 1:49739ð0:00067Þ for a
fairly “hard” spectrum in the range 1022 2 10 MeV. For a change in temperature from 300K to 3000K, kN reduces to
1.48583.
Similarly as moderator density varies from 0.93
0.83 g/cm3 in the range 97.80 C
500 C, the change in kN was
from 1.50010 to 1.49415 (Table 10.19).
In a full-core simulation for the effective multiplication factor keff , two absorber materials are considered viz boron
carbide (B4C) and hafnium (Hf) to determine the corresponding fuel inventory to balance the BOC excess reactivity.
For the given fuel rod, keff decreases from 0.98208 (0.00186) without rod, to 0.89988(0.00150) with a fully inserted
rod of 8 cm radius. Thus the core can not attain criticality even with the rod fully removed. The worth of a control rod
is found from keff with and without the rod; when keff is 0.98274 (0.00061) without a control rod, a 1 cm radius fully
inserted hafnium reduces keff to 0.98179 (0.00058) while for a similar B4C control rod the change is to 0.98057
(0.00056).
It is found that for a single central absorber rod of radius 9 cm, the maximum fuel rod radius permissible is 6 mm
for hafnium and 6.2 mm for B4C control rods.
Without a reflector and 10000 histories, 210 cycles, 10 skip cycles, keff 5 1:00783ð0:00043Þ for the same B4C
(21.89% B10) rod of radius 6 cm.
The flux is generally flat as shown in Fig. 10.16.

TABLE 10.19 Atomic densities of boron carbide.

AvNo (at./gm-atom) 0.6022 E24
Density (g/cc) 2.54
At. Wt. B10 10.01
At. Wt. B11 11.01
AtFrB10 0.199 0.2189 0.2587
AtFrB11 0.801 0.7811 0.7413
Abar B (g/g-atom) 10.811 10.7911 10.7513
At. Wt. C-12 (g/g-atom) 12
Abar B4C [g/(g-mol)] 55.244 55.1644 55.0052
N B4C (atom/b-cm) 0.0276879 0.0277278 0.0278081
B10 enrich (%) 19.9 21.89 25.87
NB 0.1107514 0.1109112 0.1112322
N B10 2.2040E-02 2.4278E-02 2.8776E-02
N B11 8.8712E-02 8.6633E-02 8.2456E-02
NC 2.7688E-02 2.7728E-02 2.7808E-02
N total (atom/b-cm) 1.3844E-01 1.3864E-01 1.3904E-01
400 Nuclear Engineering

FIGURE 10.16 Neutron flux in the 4S reactor core.

10.5.3 Micronuclear reactor

With renewed interest in space applications, micronuclear reactors (Section 3.8, Section 7.3) with ,100 kWe genera-
tion based on HEU (60%
70% enriched U-235) and heat pipes with liquid metals for high temperature operation are
being considered (Aziz, Koreshi, Sheikh, & Khan, 2020; Chenglong, Sun, Simiao, Wenxi, & Suixheng, 2020; Hao Sun,
Pan Ma, Wenxi Tian, & Suizheng Qiu, 2018).
The MNR described in shown in Figs. 3.18 and 7.18 with a compact core within a cylinder of radius of radius
35 cm and height 40 cm, listed in Table 3.20, is much smaller than the 4S reactor discussed in the previous section.
This portable core with mass B503 kg has 90 fuel rods and 37 heat pipes in a monolithic Nb-1Zr matrix. As described
in Section 3.8, the total uranium fuel is B149 kg containing B104 kg U-235.
Monte Carlo whole-core simulations for such advanced compact fast reactors have been carried out, with materials
and atomic densities listed in Table 10.20, to determine neutronic parameters such as fuel design, power distribution
and control rod worth.
MC simulations with 1000 histories per cycle and 5000 cycles give keff 5 0.955739 0.0003 and 1.026546 0.0003 for
absorbers in the front and back positions respectively. The source was sampled uniformly in each of the 90 fuel rods
with a Watt fission spectrum. The delay fraction β from two simulations (with and without delayed neutrons)
kp
β512
keff
was found to be 0.006951. The maximum excess reactivity is thus $ 3.66 (dollars), and the shutdown margin is $ 5.445
(dollars); one dollar being the reactivity normalized with the delayed neutron fraction.
The fluxes in the core and water radial shield are shown in Fig. 10.17. The fluxes can be seen to be dominated by
the higher energy range in the fuel and matrix in contrast to the fluxes in the water shield. Thus power in the core is
expected to have more contribution from high energies.
With the 1/6th symmetry in the core, as shown in Fig. 10.18, the power distribution was estimated using the fission
reaction tally.
Fig. 10.19 show the power distribution with absorber rods facing the core. The radial power peaking (RPP) is pre-
dominantly in the inner fuel cells showing a maximum of 11% with respect to the average power distribution. These
decrease gradually towards the periphery and show the trend, as anticipated, of lowest RPP (0.86) in the fuel cells
directly facing the absorber and 1.09 each in cells influenced by the reflector.
For rods facing the core and away from the core the power variations can be up to B20% which requires appropri-
ate heat removal from the heat pipes with sodium, potassium or lithium as heat transfer fluids. In this design, a heat
pipe has the capacity to remove power in 14
17 kW range. Thus if the reactor operates at a thermal power of
2
3 MW, each fuel rod will be producing B20
30 kWt which would exceed the capabilities of sodium and potassium.
Even lithium with a maximum envisaged of B18 kWt would be working at its maximum capacity. For a power
 Monte Carlo simulation in nuclear systems Chapter | 10 401

TABLE 10.20 Material data for micronuclear reactor simulation.

Material Temp (K) Density (g/cm3) Atoms/cm3/atom fraction*

Fuel: UN (70%) 2000 K 13.6 U235 0.023025, U238 0.009743, N14 0.032768
25
Helium
3.7 3 10 He 1
Lithium 1200 0.4 Li6 0.01 Li7 0.99
Lithium Vapor
0.001 Li6 0.01 Li7 0.99
Mo-14Re
12 Mo 0.064777 Re 0.0054332
Reflector BeO 900 3.01 Be9 20.360320, O16 20.639680
Absorber B4C
2.52 B10 20.782610, C12 20.217390
Matrix Nb-1Zr 1200K 6.55 Nb93 4.2031e-2 Zr 4.3239e-4
Water
1.00 H1 2, O16 1
Tungsten
19.3 W74 1

when not specified, T 5 2.5300E-8 MeV; negative fractions indicate weight fractions.

FIGURE 10.17 Core and shield fluxes.

conversion system, a thermo electric generator (TEG) operating at 5%
10% conversion efficiency would be able to
generate typically B150 kWe.

10.6 Nuclear fusion applications

Following a basic introduction to nuclear fusion and its confinement schemes given in Section 3.6 introducing the
ITER design, and the introduction to the Monte Carlo method in Chapter 7, an overview of the parameters of interest in
fission and fusion neutronics was given in Section 7.3 and Table 7.7. In this section, the geometrical complexity of the
Tokamak design and requirements and challenges for MC applications are reviewed.
We first take a look at the overall ITER design in Fig. 10.20. The size and geometrical complexity are far more dif-
ficult to model than in fission power reactors.
402 Nuclear Engineering

FIGURE 10.18 Micronuclear reactor core 1/6th model.

FIGURE 10.19 Power distribution for absorber rods facing core.

FIGURE 10.20 A view of international thermonuclear experimental

reactor. Courtesy http://www.iter.org.
 Monte Carlo simulation in nuclear systems Chapter | 10 403

The ITER experiment has a fusion power of 500 MW and its main purpose is to demonstrate significant energy gain
with pulses of the order of B400 s which are called long pulses due to its promise for the production of steady power
in the next design. Presently, the low coolant temperature is not meant to produce electrical power. It is evident that
special techniques would be required for accurate modeling of the complex geometry of ITER; for this purpose, the
fusion community has combined the MCNP code with CAD software (Juarez et al., 2021; López-Revelles et al., 2018).
In Fig. 10.21, the Toroidal Field Coils represent a superstructure within ITER; they are huge, complex and provide
the high magnetic field (5.3 T) necessary for plasma confinement.
There are 18 D-shaped superconducting toroidal field coils, and six poloidal field coils, operating at cryogenic tem-
peratures cooled by supercritical helium; a TFC has coil height 16.5 m, width 9 m and weighs 310 t. Each TFC has 134
turns with B4 miles of conductor constructed from 40 tons of niobium-tin superconducting strands. The material
niobium-tin (Nb3Sn) in the magnets becomes superconducting at low temperatures down to 4K while the boiling point
of liquid helium is 4.2K at atmospheric pressure. The TFCs weigh B6000 tons which is over one-fourth of the total
weight of the Tokamak and have a total magnetic energy of 41 GJ.
The performance of TFCs is thus dependent on parameters such as heat deposition and radiation damage which a
MC simulation should be capable of providing with confidence.
In addition to the TFCs, some other critical components are the FW, the divertor the blanket, the shielding and the
various ducts and openings for cables and instrumentation. Another feature of the ITER Tokamak is its modularity
which present added technical challenges for modeling and ensuring protection of equipment.
With these design features, some parameters needed from an elaborate MC simulation are:
1. FW loading (BMW/m2) is calculated mainly by the fast fluence (E . 0.1 MeV). The FW receives very high neu-
tron fluxes as well as the bombardment of energetic ions. MC simulations need to assess the performance of high
melting point high-strength metals such as W, Re, Ta, Mo, Nb, V.
2. Total nuclear heating to the inboard leg of TFC; about 70% of the total heating is due to neutrons and photons.
Heating causes thermal and mechanical stresses.
3. The prompt dose outside the biological shield needs to be estimated to ensure safety limits of occupational dose
compliant with regulatory standards such as the CFR 10 Part 20; this is an enormous modeling challenge for MC
simulations due to the large size and the strong shielding material. Since analog simulation would yield unreliable
results, variance reduction (source biasing and WWG, automated procedures for variance reduction) and hybrid
MC/deterministic techniques such as the Consistent Adjoint Driven Importance Sampling (CADIS) are used to
optimize particle population for reliable tallies. This has shown attractive enhancement in the quality of the results
with more precision and higher FOM (Ibrahim et al., 2011, 2015).
4. Radiation dose to materials due to neutrons and photons. The integrity of components is threatened by fast neutron
radiation damage due to radiation displacing atoms from there sites. The dislocation of a Primary Knock-on Atom

FIGURE 10.21 Toroidal field coils of the international thermonuclear

experimental reactor machine. Courtesy http://www.iter.org.
404 Nuclear Engineering

(PKA) creates a vacancy; the PKA moves about until it becomes an interstitial stationary atom at some location
creating a Frenkel pair. The pair leads to embrittlement
ÐE of structural materials such as iron and steel. MC simula-
tions give the displacement reactions RDPA 5 N EmM σD φðEÞdE where the displacement per atom (DPA) is an
important quantity to estimate the degradation of mechanical and electrical properties of materials such as copper
in the electrical wiring (Mascitti & Madariaga, 2011; Rajput, Subhash, & Srinivasan, 2020; Valentine, Colling,
Worrall, & Leppänen, 2021).
5. Tritium production in the blanket models. The TBR is required as an essential performance parameter.
6. Hydrogen and helium production. As described earlier, the presence of these gases makes welding difficult if the
gas concentrations exceed ppm levels
7. Shutdown dose due to neutron and photons. A regulatory compliance and an ITER safety limit of 100μSv/
h 5 10 mrem/h twelve days after shutdown is a requirement for the shielding around ITER. The biological dose is
calculated at the position of the person standing in Fig. 10.21.
8. The effectiveness of beryllium, lead or any other multiplier for enhancing the TBR.
9. The effectiveness of beryllium as a plasma-facing material being a low-Z material with low radiation losses and
enhanced oxygen removal from the plasma.
10. Neutrons produce activation in coolants such as water by the reaction O16 ðn; pÞN 16 :
11. High-strength materials such as molybdenum and niobium are high activation elements and are thus replaced by
vanadium and tungsten.
12. Nb-93 (n,2n), Au-197 ðn; γÞ, Al (n,α), Ni(n,p) foil.
13. Neutron/photon streaming through ducts.
14. Radiation damage to the tungsten tiles in the divertor assemblies at the bottom of the vacuum vessel.
The displacement cross-section for stainless steel decreases from 100 b to B0.015 b from 1 3 10210 to
B1023 MeV then increases to B 104 b till 14 MeV; thus fast neutrons produce the highest displacements in the
spectrum.
Elementary 1-D models have been extensively used in early fusion design studies. A typical 1-D slab, cylindrical or
spherical geometry simulation has two main parts of the configuration, the inboard and the outboard regions to model
the TF Coils, FW, vacuum vessel, breeder blanket units shown in Fig. 10.22 (Section 7.3), and shielding.
Inboard side: TFC with layers of structural material (e.g.,316SS), coolant (e.g., H2O), multiplier (Pb, Be, etc.) insu-
lator, gaps and experimental ports.
In the simulation, the 14.1 MeV source neutrons are generated in the vacuum vessel.

FIGURE 10.22 International thermonuclear experimental reactor blanket modules. Courtesy ITER: https://www.iter.org/mach/blanket.
 Monte Carlo simulation in nuclear systems Chapter | 10 405

Better 3-D MC simulations model millions of cells and may require several hours of run-time on fast super-
computing platforms. MC simulation is carried out with codes including MCNP, OpenMC, KENO, TRIPOLI, Serpent
and FLUKA for elaborate physical and geometrical modeling. Fusion-specific cross-section library such as the Fusion
Evaluated Nuclear Data Library (FENDL) with continuous energy pointwise data with MCNP or with multigroup 46-
neutron/21-gamma data which includes data from ENDF/B-VIII for neutrons; for photons the MC library MCPLIB is
used.
In terms of neutronics the main difference in fission and fusion systems lies in the neutron spectrum; in nuclear fis-
sion the average neutron energy is B 1 MeV while in fusion the DT reaction gives a 14.1 MeV neutron.
In terms of operating temperature, PWRs and ITER have lower temperatures than metal salt reactors (MSR) and
Gas-Cooled Fast Reactors (GFR) and of course much lower than the Very High Temperature Reactors or the MNR
using liquid metals win heat pipes.
Due to the very high plasma temperature in the central chamber of the Tokamak, special materials with compatible
thermo-mechanical properties, such as niobium and tungsten (melting point 3414 c), are used in fusion reactors.
The radiation damage in fusion reactors will generally be higher than that in fission power rectors due to the harder
spectrum in the former.
Due to the difference in neutron energies in fission and fusion, the primary knock-on atom (PKA) has an energy up
to a few keV while in fusion the energy is up to B 1 MeV. The radiation damage in fusion reactors due to the higher
fluence is also expected to be significantly larger than in fission reactors.
Special problems in fusion reactors are the high flux environment in the FW and divertor giving rise to much higher
mechanical and thermal stresses.
The radiation waste, handling and storage from fusion reactors will be much less of a challenge than that from fis-
sion reactors. MC simulations are capable of giving good estimates of activation to determine the levels of radiation in
the waste and spent fuel in fission reactors.
Overall, the technical challenges in fusion are arguably much steeper than those for fission reactors.
Once, MC simulations are carried out and reliable estimates are obtained with all the computational effort described
above, the greater challenge of optimization will remain.
Such studies are in progress as developments continue with proposed designs for better materials utilization and con-
figurations. Parametric studies will require efficient tools for estimating the design sensitivity of replacing structural
materials as in the case of low activation EUROFER structural steel, and tungsten and its alloys such as W-Re for
divertor surfaces, compounds of lithium such as ceramic titanates for enhancing the TBR and optimized shielding
designs for minimizing the biological dose.

Problems
10.1 In Sec. (10.2) for the Godiva simulation results briefly describe how the mean free path can be estimated from
the total number of collisions.10.2 For the Godiva spectrum shown in Fig. 10.7 can you justify the use of a one-group
diffusion model for estimating its critical radius? What advantages would one-speed transport have over a one-speed
diffusion model?10.3 How would you model the simulation when the Godiva assembly is surrounded by people?
What effects would you anticipate regarding the safety of such a configuration?10.4 For a uranium hexafluoride stor-
age facility, how would you estimate the relative effects of the neutron and photon skyshine some distance away?10.5
In Sec. (10.4) how would you design a multi-layered low-Z high-Z shield for best attenuation? How would you expect
this to compare with a composite shield of varying composition?10.6 In Sec. (10.5) what effect does the reduced den-
sity at high temperature have on the criticality?10.7 Compare the neutron flux spectra of Godiva (Fig. 10.7) and
AP1000 (Fig. 10.12) to comment on the radiation dose from both systems.10.8 For the 4S reactor core in Sec. (10.5)
what would be effect on criticality and power distribution if the positions of the low enriched (17%) and higher
enriched (19%) fuel rods are interchanged?

Nomenclature
d grain size/sphere diameter
g acceleration due to gravity
h wick thickness
keff effective multiplication factor
kp multiplication factor with prompt neutrons
406 Nuclear Engineering

le evaporator length
lc condenser length
la adiabatic region length
lt total length of heat pipe
ri inner heat pipe radius

Greek letters
Ψ tilt angle from vertical axis
ε porosity
φ flux
σDPA microscopic cross-section for displacements per atom
τp prompt removal lifetime

Abbreviations and acronyms

BOC beginning of cycle
DPA displacement per atom
FOM figure of merit
HEU highly enriched uranium
HVL half value layer
IAEA International Atomic Energy Agency
ISO International Standards Organization
ITER International Thermonuclear Experimental Reactor
Gen-IV Generation IV
MNR Micronuclear reactor
PFC Poloidal field coil
PGNAA Prompt gamma neutron activation analysis
RPP Radial power peaking
SWU Separative work units
TEG Thermo electric generator
TFC Toroidal field coil
TVL Tenth value layer
USEC United States Enrichment Corporation
WGU Weapons grade uranium

References
Aziz, U., Koreshi, Z. U., Sheikh, S. R., & Khan, H. (2020). A study of criticality and thermal loading in a conceptual micronuclear heat pipe reactor
for space applications. Nuclear Technology and Radiation Protection, 35, 208
215. Available from https://doi.org/10.2298/NTRP2003208A.
Broadhead, B. L. (1991). Criticality safety review of 21/ 2-, 10-, and 14-ton UF6 cylinders. Tennessee.
Chenglong, W., Sun, H., Simiao, T., Wenxi, T., & Suixheng, Q. (2020). Thermal-hydraulic analysis of a new conceptual heat-pipe cooled small
nuclear reactor system. Nuclear Engineering and Technology, 52, 19
26.
IAEA. (2012). Neutron generators for analytical purposes. IAEA Radiation Technology Reports Series.
IAEA. (2014). Criticality safety in the handling of fissile material, Specific Safety Guide SSG-27. Vienna, Austria.
Ibrahim, A. M., Mosher, S. W., Evans, T. M., Peplow, D. E., Sawan, M. E., Wilson, P. P. H., . . . Heltemes, T. (2011). Iter neutronics modeling using
hybrid Monte Carlo/ deterministic and CAD-based Monte Carlo methods. In Nuclear technology. ,https://doi.org/10.13182/NT175-251..
Ibrahim, A. M., Wilson, P. P. H., Sawan, M. E., Mosher, S. W., Peplow, D. E., Wagner, J. C., & Grove, R. E. (2015). Automatic mesh adaptivity for
hybrid Monte Carlo/deterministic neutronics modeling of difficult shielding problems. Nuclear Science and Engineering, 181. Available from
https://doi.org/10.13182/NSE14-94.
ICRP. (1996). ICRP Publication 74. Conversion coefficients for use in radiological protection against external radiation. In Annals of the ICRP.
Juarez, R., Pedroche, G., Loughlin, M. J., Pampin, R., Martinez, P., De Pietri, M., & Sanz, J. (2021). A full and heterogeneous model of the ITER
tokamak for comprehensive nuclear analyses. Nature Energy. Available from https://doi.org/10.1038/s41560-020-00753-x.
Koreshi, Z. U., & Hussain, S. Z. (2014). A neutronic pin cell and full-core design analysis of the 4S reactor. Annals of Nuclear Energy, 72, 145
150.
Available from https://doi.org/10.1016/j.anucene.2014.05.015.
 Monte Carlo simulation in nuclear systems Chapter | 10 407

Laranjo de Stefani, G., Losada Moreira, J. M., Maiorino, J. R., & Russo Rossi, P. C. (2019). Detailed neutronic calculations of the AP1000 reactor
core with the Serpent code. Progress in Nuclear Energy. Available from https://doi.org/10.1016/j.pnucene.2019.03.030.
López-Revelles, A. J., Catalán, J. P., Kolšek, A., Juárez, R., Garcı́a, R., Garcı́a, M., & Sanz, J. (2018). MCNP model of the ITER Tokamak complex.
Fusion Engineering and Design. Available from https://doi.org/10.1016/j.fusengdes.2018.04.023.
Mascitti, J. A., & Madariaga, M. (2011). Method for the calculation of DPA in the reactor pressure vessel of Atucha II. Science and Technology of
Nuclear Installations. Available from https://doi.org/10.1155/2011/534689.
Rajput, M., Subhash, P. V., & Srinivasan, R. (2020). Displacement damage study in tungsten and iron for fusion neutron irradiation. Fusion
Engineering and Design. Available from https://doi.org/10.1016/j.fusengdes.2019.111370.
Rowlands, J., Bosq, J. C., Brinckman, J. J., Fort, E., Giffard, F., Jacqmin, R., . . . Pelloni, S. (1999). Intercomparison of calculations for Godiva and
Jezebel.
Su, S.-D. (2015). Environmental radiation dose assessment and validation for uranium hexafluoride storage management. University of New Mexico.
,https://digitalrepository.unm.edu/ne_etds?utm_source 5 digitalrepository.unm.edu%2Fne_etds%2F11&utm_medium 5 PDF&utm_campaign 5
PDFCoverPages..
Sun, H., Wang, C., Ma, P., Liu, X., Tian, W., Qiu, S., & Su, G. (2018). Conceptual design and analysis of a multipurpose micro nuclear reactor power
source. Annals of Nuclear Energy, 121, 118
127. Available from https://doi.org/10.1016/j.anucene.2018.07.025.
Tsuboi, Y., Arie, K., Ueda, N., & Grenci, T. (2012). Design of the 4S reactor. Nuclear Technology, 178, 201
217.
USEC (United States Enrichment Corporation). (1995). Uranium hexafluoride: A manual of good handling practices.
Valentine, A., Colling, B., Worrall, R., & Leppänen, J. (2021). Benchmarking of the Serpent 2 Monte Carlo code for fusion neutronics applications.
EPJ Web of Conferences. Available from https://doi.org/10.1051/epjconf/202124704015.
Wagner, J. C., Sisolak, J. E., & McKinney, G. (1992). MCNP: Criticality safety benchmark problems, LA-12415. New Mexico, USA.
Whalen, D. J., Cardon, D. A., Uhle, J. L., & Hendricks, J. S. (1991). MCNP: Neutron benchmark problems. LA- 2 12212 DE92 004710.
408 Nuclear Engineering

Annex A MCNP listing for Godiva (Section 10.2.1)

BK10Gdva Godiva with 50 equi-volume regions

c this gives f(r)
1 1 -18.74 -1 imp:n=1
2 1 -18.74 1 -2 imp:n=1
…
… lines removed for space
…
48 1 -18.74 47 -48 imp:n=1
49 1 -18.74 48 -49 imp:n=1
50 1 -18.74 49 -50 imp:n=1
51 0 50 imp:n=0

1 SO 2.37267244
2 SO 2.98937995
3 SO 3.42198581
…
… lines removed for space
46 SO 8.50139893
47 SO 8.56256218
48 SO 8.62286390
49 SO 8.68233380
50 SO 8.74100000

kcode 3000 1.0 60 150 $ criticality calculation

c in kcode fission is treated as an absorption
ksrc 0 0 0
m1 92234 -1.02
92235 -93.7
92238 -5.27
m2 92234 -1.02 92235 -93.7 92238 -5.27
m3 92234 1
m4 92235 1
m5 92238 1
f1:n 1
fc4 tallies in bare sphere
f4:n 1 2 3 4 5 6 7 8 9 10
11 12 13 14 15 16 17 18 19 20
21 22 23 24 25 26 27 28 29 30
31 32 33 34 35 36 37 38 39 40
41 42 43 44 45 46 47 48 49 50
fmesh24:n GEOM=xyz ORIGIN = -8.741 -8.841 -0.5
IMESH=8.741 IINTS=40 JMESH=8.741 JINTS=40 KMESH=0.5 KINTS=1
c
E34 0.625e-6 0.1 14
c fc34 tallies in bare sphere (N=0.0479838,R=8.741,
c V=2.7975e3, NV=134.2353)
 Monte Carlo simulation in nuclear systems Chapter | 10 409

f34:n 1 2 3 4 5 6 7 8 9 10
11 12 13 14 15 16 17 18 19 20
21 22 23 24 25 26 27 28 29 30
31 32 33 34 35 36 37 38 39 40
41 42 43 44 45 46 47 48 49 50
fc34 NV nu f n,xn a f n,g
fm34 (134.2353 2 (-6 -7) (16:17) (-2) (-6) (102) )
c
c fc44 tallies in bare sphere
f44:n 1
fc44 NV m n,g
fm44 (134.2353 4 (102))
c
c
FC54:N Energy dependent scalar flux to plot f(E)
F54:N 1
e54 3.77e-9 4.05e-9 4.36e-9 4.69e-9 5.04e-9 5.42e-9 5.83e-9 6.27e-9
6.74e-9 7.25e-9 7.79e-9 8.38e-9 9.01e-9 9.69e-9 1.04e-8 1.12e-8
…….. lines removed for space
3.01e-2 3.24e-2 3.48e-2 3.74e-2 4.02e-2 4.33e-2 4.65e-2
5.e-2 5.38e-2 5.78e-2 6.22e-2 6.69e-2 7.19e-2 7.73e-2
8.32e-2 8.94e-2 9.61e-2 1.03e-1 1.11e-1 1.20e-1 1.29e-1
1.38e-1 1.49e-1 1.60e-1 1.72e-1 1.85e-1 1.99e-1 2.14e-1
2.30e-1 2.47e-1 2.66e-1 2.86e-1 3.07e-1 3.30e-1 3.55e-1
3.82e-1 4.10e-1 4.41e-1 4.75e-1 5.10e-1 5.49e-1 5.90e-1
6.34e-1 6.82e-1 7.33e-1 7.89e-1 8.48e-1 9.12e-1 9.81e-1
1.05 1.13 1.22 1.31 1.41 1.52 1.63
1.75 1.88 2.03 2.18 2.34 2.52 2.71
2.91 3.13 3.37 3.62 3.89 4.19 4.50
4.84 5.20 5.60 6.02 6.47 6.96 7.48
8.04 8.65 9.30 10
Print
410 Nuclear Engineering

Annex B MCNP input listing (Jezebel, Section 10.2.2)

BK10Jzbl Jezebel concentration

1 1 -15.61 -1 imp:n=1
2 0 1 imp:n=0

1 so 6.385

kcode 3000 1.0 60 150

ksrc 0 0 0
m1 94240 -4.5 94239 -95.5
m3 94239 1
m4 94240 1
m5 94240 1
f1:n 1
fc4 tallies in bare sphere
f4:n 1
fmesh24:n GEOM=xyz ORIGIN = -6.385 -6.385 -0.5
IMESH=6.385 IINTS=40 JMESH=6.385 JINTS=40 KMESH=0.5 KINTS=1
c
E34 0.625e-6 0.1 14
c fc34 tallies in bare sphere (N=0.0393163,R=6.385,
c V=1.0904e3, NV=42.8691)
f34:n 1
fc34 NV tot el nu f n,xn a f n,g
fm34 (42.8691 2 (1) (2)(-6 -7) (16:17) (-2) (-6) (102) )
c fc44 tallies in bare sphere
f44:n 1
fc44 NV m n,g
fm44 (42.8691 3 (102))
c
FC54:N Energy dependent scalar flux to plot f(E)
F54:N 1
e54 3.77e-9 4.05e-9 4.36e-9 4.69e-9 5.04e-9 5.42e-9 5.83e-9 6.27e-9
6.74e-9 7.25e-9 7.79e-9 8.38e-9 9.01e-9 9.69e-9 1.04e-8 1.12e-8
1.20e-8 1.29e-8 1.39e-8 1.50e-8 1.61e-8 1.73e-8 1.86e-8
2.e-8 2.15e-8 2.31e-8 2.49e-8 2.68e-8 2.88e-8 3.09e-8
3.33e-8 3.58e-8 3.85e-8 4.13e-8 4.45e-8 4.78e-8 5.14e-8
5.53e-8 5.94e-8 6.39e-8 6.87e-8 7.39e-8 7.95e-8 8.54e-8
9.19e-8 9.88e-8 1.06e-7 1.14e-7 1.23e-7 1.32e-7 1.42e-7
1.53e-7 1.64e-7 1.77e-7 1.90e-7 2.04e-7 2.19e-7 2.36e-7
2.54e-7 2.73e-7 2.93e-7 3.15e-7 3.39e-7 3.65e-7 3.92e-7
4.22e-7 4.53e-7 4.88e-7 5.24e-7 5.64e-7 6.06e-7 6.52e-7
7.01e-7 7.54e-7 8.10e-7 8.71e-7 9.37e-7 1.01e-6 1.08e-6
1.16e-6 1.25e-6 1.35e-6 1.45e-6 1.56e-6 1.67e-6 1.80e-6
1.94e-6 2.08e-6 2.24e-6 2.41e-6 2.59e-6 2.78e-6 2.99e-6
3.22e-6 3.46e-6 3.72e-6 4.e-6 4.30e-6 4.62e-6 4.97e-6
5.35e-6 5.75e-6 6.18e-6 6.65e-6 7.15e-6 7.69e-6 8.26e-6
8.89e-6 9.56e-6 1.03e-5 1.10e-5 1.19e-5 1.28e-5 1.37e-5
1.48e-5 1.59e-5 1.71e-5 1.84e-5 1.97e-5 2.12e-5 2.28e-5
2.45e-5 2.64e-5 2.84e-5 3.05e-5 3.28e-5 3.53e-5 3.79e-5
4.08e-5 4.39e-5 4.72e-5 5.07e-5 5.45e-5 5.86e-5 6.31e-5
6.78e-5 7.29e-5 7.84e-5 8.43e-5 9.06e-5 9.75e-5 1.05e-4
 Monte Carlo simulation in nuclear systems Chapter | 10 411

1.13e-4 1.21e-4 1.30e-4 1.40e-4 1.51e-4 1.62e-4 1.74e-4

1.87e-4 2.01e-4 2.17e-4 2.33e-4 2.50e-4 2.69e-4 2.89e-4
3.11e-4 3.35e-4 3.60e-4 3.87e-4 4.16e-4 4.47e-4 4.81e-4
5.17e-4 5.56e-4 5.98e-4 6.43e-4 6.91e-4 7.43e-4 7.99e-4
8.60e-4 9.24e-4 9.94e-4 1.07e-3 1.15e-3 1.24e-3 1.33e-3
1.43e-3 1.54e-3 1.65e-3 1.78e-3 1.91e-3 2.05e-3 2.21e-3
2.37e-3 2.55e-3 2.75e-3 2.95e-3 3.17e-3 3.41e-3 3.67e-3
3.95e-3 4.24e-3 4.56e-3 4.91e-3 5.28e-3 5.67e-3 6.10e-3
6.56e-3 7.05e-3 7.58e-3 8.15e-3 8.77e-3 9.43e-3 1.01e-2
1.09e-2 1.17e-2 1.26e-2 1.36e-2 1.46e-2 1.57e-2 1.68e-2
1.81e-2 1.95e-2 2.09e-2 2.25e-2 2.42e-2 2.60e-2 2.80e-2
3.01e-2 3.24e-2 3.48e-2 3.74e-2 4.02e-2 4.33e-2 4.65e-2
5.e-2 5.38e-2 5.78e-2 6.22e-2 6.69e-2 7.19e-2 7.73e-2
8.32e-2 8.94e-2 9.61e-2 1.03e-1 1.11e-1 1.20e-1 1.29e-1
1.38e-1 1.49e-1 1.60e-1 1.72e-1 1.85e-1 1.99e-1 2.14e-1
2.30e-1 2.47e-1 2.66e-1 2.86e-1 3.07e-1 3.30e-1 3.55e-1
3.82e-1 4.10e-1 4.41e-1 4.75e-1 5.10e-1 5.49e-1 5.90e-1
6.34e-1 6.82e-1 7.33e-1 7.89e-1 8.48e-1 9.12e-1 9.81e-1
1.05 1.13 1.22 1.31 1.41 1.52 1.63
1.75 1.88 2.03 2.18 2.34 2.52 2.71
2.91 3.13 3.37 3.62 3.89 4.19 4.50
4.84 5.20 5.60 6.02 6.47 6.96 7.48
8.04 8.65 9.30 10
FC6 Energy deposition averaged over cell (MeV/g)
F6:N 1
FC7 Fission energy deposition averaged over cell (MeV/g)
F7:N 1
Print
412 Nuclear Engineering

Annex C MCNP input listing (BK10Shld, Section 10.5.1)

BK10Shld Concrete shell 30 cm thick

c Shield neutron 2 june 2021
c cell cards
1 0 -1 imp:n=1
2 1 -2.3 1 -2 imp:n=1 $ concrete shield
3 0 2 imp:n=0

c surface cards
1 so 304.8
2 so 374.8

c geometry cards
mode n
c concrete (rho = 2.3 g/cm^3)
m1 1001 1.68765e-1
8016 5.62493e-1
11023 1.18366e-2
12000 1.39951e-3
13027 2.14316e-2
14000 2.04076e-1
19000 5.65495e-3
20000 1.86720e-2
26054 2.47295e-4
26056 3.91067e-3
26057 9.38014e-5
26058 1.19384e-5
6012 1.41730e-3
sdef erg=14
f1:n 1 2
f2:n 1 2
E0 1 12i 14.0
f45:n 0 374.8 0 0
cut:n j 1.0 $ cut neutrons at 1 MeV
f12:n 2
c ambient neutron dose equiv. H*(10mm) ICRP Sv cm2
de12 2.500E-08 1.000E-07 1.000E-06 1.000E-05 1.000E-04 1.000E-03
1.000E-02 2.000E-02 5.000E-02 1.000E-01 2.000E-01 5.000E-01
1.000E+00 1.500E+00 2.000E+00 3.000E+00 4.000E+00 5.000E+00
6.000E+00 7.000E+00 8.000E+00 1.000E+01 1.400E+01 1.700E+01
2.000E+01
df12 8.000E-12 1.040E-11 1.120E-11 9.200E-12 7.100E-12 6.200E-12
8.600E-12 1.460E-11 3.500E-11 6.900E-11 1.260E-10 2.580E-10
3.400E-10 3.620E-10 3.520E-10 3.800E-10 4.090E-10 3.780E-10
3.830E-10 4.030E-10 4.170E-10 4.460E-10 5.200E-10 6.100E-10
6.500E-10
nps 400000
dd 0.1 1e100
print
Annex D MCNP input listing (BK10AP10, Section 10.5.1)

BK10AP10 Unit Lattice Cell AP1000 23 MAY 2021

c
1 1 -1.0 (1 -2 3 -4 5 -6) (7) imp:n=1 tmp=2.5300e-8 $ water
2 2 -10.47635 (-7 5 -6) imp:n=1 tmp=2.5300e-8 $ fuel pin
3 3 1.0e-4 (7 -8 5 -6) imp:n=1 tmp=2.5300e-8 $ gap
4 4 4.295e-2 (8 -9 5 -6) imp:n=1 tmp=2.5300e-8 $ clad
5 0 (-1:2:-3:4:-5:6) imp:n=1 tmp=2.5300e-8 $ outside world

C surface cards
*1 PX -0.63
*2 PX 0.63
*3 PY -0.63
*4 PY 0.63
*5 PZ -2
*6 PZ 2
7 CZ 0.4096
8 CZ 0.4178
9 CZ 0.4750

mode n
c ---------WATER (den 1 g/cm3)
M1 8016.50C 1 $ rmccs 2.53e-8
1001.50C 2 $ rmccs 2.53e-8
MT1 lwtr.01t $ tmccs 2.53e-8
c --------- Fuel (den 10.4 g/cm3 4.5%) endf66a 2.53e-8
c m2 8016.66c 4.64149e-2 92234.66c 8.49269e-6 92235.66c 1.05705e-3
c 92238.66c 2.21413e-2
c --------- Fuel (den 10.4 g/cm3 4.45%) endf66a 2.53e-8
m2 8016.66c 4.67471e-2 92235.66c 1.05284e-3 92238.66c 2.23207e-2
c ---------- Helium4 Gap (den 0.1785e-3 g/cm3)
m3 2004.50c 1 $ rmccs 2.53e-8
c --------clad(6.506 g/cm^3) at den=4.295e-2 PNNL ----------
m4 40000.66c 1 $ endf66b
c
c tallies
m5 92235.66c 1
m6 92238.66c 1
c
FC4 neutron flux in cells 1 2 3 4
F4:N 1 2 3 4
c sd4
E4 0.625e-6 0.1 14
VOL 1 1 1 1
c ----------------
FC14 TALLIES IN WATER
C ----------------
E14 0.625e-6 0.1 14
F14:N 1
C NV tot el a n,g
fm14 (0.3526 1 (1) (2) (-2) (102) )
C fM14 (1 5 (-6))(1 6 (102))
C ---------------
FC24 TALLIES IN FUEL
C ---------------
E24 0.625e-6 0.1 14
F24:N 2
414 Nuclear Engineering

C NV tot el nu f n,xn a f n,g

fm24 (0.1478 2 (1) (2)(-6 -7) (16:17) (-2) (-6) (102) )
C
C ---------------------
FC34 TALLIES IN HELIUM GAP
C ---------------------
E34 0.625e-6 0.1 14
F34:N 3
C NV tot el a n,g
fm34 (8.5259e-6 3 (1) (2) (-2) (102) )
C
C ---------------
FC44 TALLIES IN CLAD
C ---------------
E44 0.625e-6 0.1 14
F44:N 4
C NV tot el nu f n,xn a f n,g
FM44 (0.0279 4 (1) (2)(-6 -7) (16:17) (-2) (-6) (102) )
fmesh54:n GEOM=xyz ORIGIN = -0.63 -0.63 -0.5
IMESH=0.63 IINTS=40 JMESH=0.63 JINTS=40 KMESH=0.5 KINTS=1
FC64:N Energy dependent scalar flux to plot f(E) in water (1) and fuel (2)
F64:N 1 2
e64 3.77e-9 4.05e-9 4.36e-9 4.69e-9 5.04e-9 5.42e-9 5.83e-9 6.27e-9
6.74e-9 7.25e-9 7.79e-9 8.38e-9 9.01e-9 9.69e-9 1.04e-8 1.12e-8
1.20e-8 1.29e-8 1.39e-8 1.50e-8 1.61e-8 1.73e-8 1.86e-8
2.e-8 2.15e-8 2.31e-8 2.49e-8 2.68e-8 2.88e-8 3.09e-8
3.33e-8 3.58e-8 3.85e-8 4.13e-8 4.45e-8 4.78e-8 5.14e-8
5.53e-8 5.94e-8 6.39e-8 6.87e-8 7.39e-8 7.95e-8 8.54e-8
9.19e-8 9.88e-8 1.06e-7 1.14e-7 1.23e-7 1.32e-7 1.42e-7
1.53e-7 1.64e-7 1.77e-7 1.90e-7 2.04e-7 2.19e-7 2.36e-7
2.54e-7 2.73e-7 2.93e-7 3.15e-7 3.39e-7 3.65e-7 3.92e-7
4.22e-7 4.53e-7 4.88e-7 5.24e-7 5.64e-7 6.06e-7 6.52e-7
7.01e-7 7.54e-7 8.10e-7 8.71e-7 9.37e-7 1.01e-6 1.08e-6
1.16e-6 1.25e-6 1.35e-6 1.45e-6 1.56e-6 1.67e-6 1.80e-6
1.94e-6 2.08e-6 2.24e-6 2.41e-6 2.59e-6 2.78e-6 2.99e-6
3.22e-6 3.46e-6 3.72e-6 4.e-6 4.30e-6 4.62e-6 4.97e-6
5.35e-6 5.75e-6 6.18e-6 6.65e-6 7.15e-6 7.69e-6 8.26e-6
8.89e-6 9.56e-6 1.03e-5 1.10e-5 1.19e-5 1.28e-5 1.37e-5
1.48e-5 1.59e-5 1.71e-5 1.84e-5 1.97e-5 2.12e-5 2.28e-5
2.45e-5 2.64e-5 2.84e-5 3.05e-5 3.28e-5 3.53e-5 3.79e-5
4.08e-5 4.39e-5 4.72e-5 5.07e-5 5.45e-5 5.86e-5 6.31e-5
6.78e-5 7.29e-5 7.84e-5 8.43e-5 9.06e-5 9.75e-5 1.05e-4
1.13e-4 1.21e-4 1.30e-4 1.40e-4 1.51e-4 1.62e-4 1.74e-4
1.87e-4 2.01e-4 2.17e-4 2.33e-4 2.50e-4 2.69e-4 2.89e-4
3.11e-4 3.35e-4 3.60e-4 3.87e-4 4.16e-4 4.47e-4 4.81e-4
5.17e-4 5.56e-4 5.98e-4 6.43e-4 6.91e-4 7.43e-4 7.99e-4
8.60e-4 9.24e-4 9.94e-4 1.07e-3 1.15e-3 1.24e-3 1.33e-3
1.43e-3 1.54e-3 1.65e-3 1.78e-3 1.91e-3 2.05e-3 2.21e-3
2.37e-3 2.55e-3 2.75e-3 2.95e-3 3.17e-3 3.41e-3 3.67e-3
3.95e-3 4.24e-3 4.56e-3 4.91e-3 5.28e-3 5.67e-3 6.10e-3
6.56e-3 7.05e-3 7.58e-3 8.15e-3 8.77e-3 9.43e-3 1.01e-2
1.09e-2 1.17e-2 1.26e-2 1.36e-2 1.46e-2 1.57e-2 1.68e-2
1.81e-2 1.95e-2 2.09e-2 2.25e-2 2.42e-2 2.60e-2 2.80e-2
3.01e-2 3.24e-2 3.48e-2 3.74e-2 4.02e-2 4.33e-2 4.65e-2
5.e-2 5.38e-2 5.78e-2 6.22e-2 6.69e-2 7.19e-2 7.73e-2
8.32e-2 8.94e-2 9.61e-2 1.03e-1 1.11e-1 1.20e-1 1.29e-1
 Monte Carlo simulation in nuclear systems Chapter | 10 415

1.38e-1 1.49e-1 1.60e-1 1.72e-1 1.85e-1 1.99e-1 2.14e-1

2.30e-1 2.47e-1 2.66e-1 2.86e-1 3.07e-1 3.30e-1 3.55e-1
3.82e-1 4.10e-1 4.41e-1 4.75e-1 5.10e-1 5.49e-1 5.90e-1
6.34e-1 6.82e-1 7.33e-1 7.89e-1 8.48e-1 9.12e-1 9.81e-1
1.05 1.13 1.22 1.31 1.41 1.52 1.63
1.75 1.88 2.03 2.18 2.34 2.52 2.71
2.91 3.13 3.37 3.62 3.89 4.19 4.50
4.84 5.20 5.60 6.02 6.47 6.96 7.48
8.04 8.65 9.30 10
FC6 Energy deposition averaged over cell (MeV/g)
F6:N 1 2 3 4 T
FC7 Fission energy deposition averaged over cell (MeV/g)
F7:N 1 2 3 4 T
kcode 5000 1 5 150
prdmp 150 150 150
ksrc 0 0 -1
0 0 0
0 0 1
print
Chapter 11

Comparisons: Monte Carlo, diffusion, and

transport

11.1 Introduction
A university nuclear engineering curriculum typically places neutron diffusion in the early to middle part of undergraduate
studies while transport theory is delayed to the final year. It is not uncommon for students to feel uncomfortable with the
neutron transport equation and usually one does not muster the courage to go deeper and explore whether these two for-
mulations have common ground.
By the time the final year is about to end, one gets to hear of Monte Carlo (MC) simulation but since it is not cov-
ered in sufficient detail, one typically graduates without seeing the connection.
It is in a Master’s or PhD program that one takes a serious look at the transport equation and appreciates the power
of MC simulation for a thesis and beyond. That power is often comforting, especially to see voluminous results pre-
sented in colorful pictures.
In the 1970s and early 1980s, professors and students would spend a great amount of effort and time in getting the
mathematics right. The power of theory, its challenges, and elegance, were considered the power of academia. With the
demands of industry, that focus on theory probably weakened.
This chapter is meant to present a unified picture of neutron diffusion, transport, and MC plus the realization that it
comes under the more general scope of radiation transport. This picture comes towards the end of this book when the
reader has covered all three with a fair amount of depth and breadth.
Once a unified picture has been made in the mind, it should be natural to view things as one; whether deterministic
or stochastic, nuclear engineering as a profession will cherish both and see their strengths.
To present a unified picture, some elementary models are reviewed and solutions with diffusion, transport and MC
simulation are presented for each.
This chapter comes after diffusion theory (DT) (Chapter 5), transport theory (Chapter 6), and the basics of MC
method (Chapter 7) have been covered in a fair amount of detail.
A comparison must begin with benchmarks which are fortunately available in nuclear engineering to a good extent. In
the preceding chapters, these benchmarks have been cited and used on occasions; in criticality, diffusion, transport and MC.
The early work, nicely covered by Bell and Glasstone (1979), Clark and Hansen (1964), Ganapol (2008) begin from
the early work in radiative transfer and use powerful mathematics such as Fourier transforms, Laplace transforms,
and the Green’s functions. Pioneering work by Case and Zweifel, the Weiner-Hopf method, and Chandrasekhar’s
H-function are the solid bedrock upon which nuclear engineering stands firm.
The following four sections cover system multiplication in a bare sphere, 1D slab, and spherical systems and a com-
parison of fluxes for a one-group flux calculation.

11.2 Criticality in a bare sphere

11.2.1 One-group diffusion theory criticality
This section is based on Chapter 5 (The neutron diffusion equation). From Section 5.2.2, the one-group diffusion equa-
tion (Eq. 5.11) expressed as an eigenvalue equation is
1
Dr2 φ 2 Σr φ 1 νΣf φ 5 0 (11.1)
k
gives the infinite medium equation, setting the leakage (r2 φÞ term to zero, for which the infinite multiplication factor kN is
Nuclear Engineering. DOI: https://doi.org/10.1016/B978-0-323-90618-0.00011-9
© 2022 Elsevier Inc. All rights reserved. 417
418 Nuclear Engineering

TABLE 11.1 One-group data for Godiva.

Nuclide ν σa (b) σf (b) σtr (b)

235
U 2.60 1.65 1.40 6.80
U238 2.60 0.255 0.095 6.90

Source: Criticality Hand calculations LA14244-M p. 45/181.

TABLE 11.2 One-group macroscopic cross sections for Godiva.

Nuclide N atoms/b-cm Σa cm21 Σf cm21 Σtr cm21

U-235 4.489 3 1022 7.407 3 1022 6.285 3 1022 3.0530 3 1021
U-238 3.081 3 1022 7.857 3 1024 2.927 3 1024 2.130 3 1022

νΣf
kN 5 (11.2)
Σr
and for the finite system
kN
keff 5 : (11.3)
1 1 L2 B2
The corresponding neutron flux is
sinBr
φðr Þ 5 A (11.4)
r
where A is a constant depending on the power P. The Godiva data in Tables 11.1 gives the macroscopic quantities listed
in Table 11.2.
From the above, the diffusion coefficient is D 5 1=3Σtr 5 1:0209 cm, the diffusion length (squared) is
L2 5 D=Σa 5 13:6382 cm2, from which kN 5 νΣf =Σa 5 2:1932.
The criticality equation gives B2 5 0:0875 cm2, from which the extrapolated radius is R 5 10.6213 cm. With
d 5 2:13 3 D 5 2:1745 cm, the critical radius is R 5 R 2 d 5 8:4468 cm.

11.2.2 Two-group diffusion theory criticality

From Section 5.3.2, the two-group diffusion equations are
1

D1 r2 φ1 2 Σ1r φ1 1 νΣf 1 φ1 1 νΣf 2 φ2 5 0 (11.5)
k
and
D2 r2 φ2 2 Σ2r φ2 1 Σs1-2 φ1 5 0 (11.6)
with the assumption that there is no fission contribution into group 2.
The two-group (core) fluxes are given by Eqs. (5.47) and (5.48) for functions X and Y and coupling coefficients.
For an infinite system, setting the leakage (r2 ::Þ terms to zero, gives kN as
Σ2r νΣf 1 1 Σs1-2 νΣf 2
kN 5 (11.7)
Σ1r Σ2r
and the finite mutiplication as
νΣf 1 Σs1-2 νΣf 2
keff 5 1 : (11.8)
D1 B2 1 Σ1r ðD1 B2 1 Σ1r ÞðD2 B2 1 Σ2r Þ
 Comparisons: Monte Carlo, diffusion, and transport Chapter | 11 419

The critical size, given the composition, is found from the 4X4 criticality determinant (Eq. 5.54) with the appropri-
ate two-group data as for the reflected spherical reactor in Section 5.3.2. A simpler expression, the six-factor formula,
was obtained in Section 2.10 for a single generation, as
kN
keff 5 1 5 : (11.9)
ð1 1 B2 τ Þð1 1 L2 B2 Þ
In Eq. (11.9), the effects of the fast and thermal groups appear through the leakage terms with neutron age and diffusion
length respectively.

11.2.3 One-speed transport theory criticality

To compare diffusion estimates with the transport result, consider the asymptotic flux (2.10, Section 6.3.8) in a sphere
of radius r
A r
φas ðr Þ 5 sin (11.10)
r jν 0 j
where ν 0 is a solution of the transcendental equation
1 cν 0 ν 0 1 1
1 5 cν 0 tanh21 5 ln : (11.11)
ν0 2 ν0 2 1
Recall that c is the number of neutrons emerging from an interaction such that 0 , c , 1 for non-multiplying media
and c . 1 for a multiplying medium such
as Godiva. The sphere is approximately critical when the asymptotic flux is
zero at the extrapolated radius R; thus φas R 5 0 at R 5 Rc 1 x0 5 πjν 0 ðcÞj.
The solutions of (Eq. 11.11) are real for c , 1 and complex for c . 1; an approximation for ν 0 is
 
1 3ð1 2 cÞ 912c
5 12 2? : (11.12)
ν02 c 5 c
For c near unity, another approximation is
 
1 2
ν 0 5 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 1 ð1 2 cÞ 1 ? : (11.13)
3ð1 2 cÞ 5

To find the critical radius ν 0 is determined from the transcendental equation and the critical radius is found from
Rc 5 πjν 0 j 2 x0 , where the extrapolation distance x0 can be calculated from the Mark P1 boundary condition (Eq. 2.65).
In the spherical harmonics P1 “diffusion approximation,” the neutron flux is given by Eq. (2.54), re-written here for
quick reference:
rffiffiffiffiffiffiffiffiffiffiffi
1 3
φðxÞ 5 e2jxj=L
2 12c
where the diffusion length is
1
L 5 ν 0 5 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi : (11.14)
3ð 1 2 c Þ
Values of ν 0 for c , 1 are plotted in Fig. 6.2; others are reproduced from Bell and Glasstone (Bell & Glasstone,
1979) (Table 11.3).
For a one-group calculation, the data of Table 2.17(a) (Bell and Glasstone) is used for U-235 (density 18.8 g/cm3)
assuming isotropic scattering:
ν 5 2.50, σf 5 1.3 b, σs 5 4.0 b. This gives c 5 1.3679, for which Eq. (11.11) gives
jiν 0 j 5 0:8359:
Note that the transcendental equation for c , 1 is straightforward as the roots are real numbers while for c . 1, with
complex roots, the MATLABs program listed below was written.
The extrapolation
 radius, from the Mark P1 boundary conditions is x0 5 0.5181 cm. The critical radius,
Rc 5 πν 0 ðcÞ 2 x0 5 8.2558 cm which is to be compared with the computed value of the critical radius 8.710 cm
with keff 5 0.9912.
420 Nuclear Engineering

TABLE 11.3 Relaxation lengths for isotropic scattering (in mean free paths).

c Exact (Eq. 11.11) Approximation (Eq. 11.13) Diffusion theory (Eq. 11.14)
c,1
jν 0 j
0.99 5.797 5.797 5.774
0.90 4.116 4.115 4.083
0.80 1.408 1.394 1.291
0.50 1.044 0.979 0.816
0 1.000 0.808 0.577
c.1
jiν 0 j
1.01 5.750 5.751 5.774
1.05 2.532 2.531 2.582
1.10 1.757 1.756 1.826
1.20 1.198 1.195 1.291
1.50 0.689 0.680 0.816

The relaxation length in one-speed transport theory P1 diffusion approximation is ν 0 , and it determines the shape of
the flux just like the buckling in one group DT.
MATLAB program for computing the eigenvalue v0 from Eq. (11.11)

% calculate eigenvalue from transcendental equation

% Bell and Glasstone p.126 q.11
nu=2.5; sigF=1.3;
sigC=0;sigS=4;sigT=sigF+sigC+sigS;A=235;mu0=2/(3*A);sigTR=sigS*(1-mu0);
c=(nu*sigF+sigS)/sigT; k_inf=nu*sigF/(sigT-sigS);
c=1.3679; % godiva
z= 0.8:0.0001:0.96; x=z*i; % convert to a complex number
f = @(x) (2./(c*x)) - log((x+1)./(x-1)); y=f(x); yabs=abs(y);
plot(z,yabs);

% find root
n=size(z);
ymin=10;
for j=1:n(2)
if(yabs(j)<ymin)
Nu0=z(j); ymin=yabs(j);
end
end

% extrapolated radius Rc
x0=(1/sqrt(3))*(1-(1/3)*(c-1)+(1/5)*(c-1)^2-(1/7)*(c-1)^3); % extrap dist
rho=18.8;AvNo=0.6022e24;M=235;N=rho*AvNo/M;

SIGT=N*sigT*1e-24; lambda=1/SIGT;

Rc=pi*Nu0-x0; % critical radius in mfp

Rc_cm=Rc*lambda; % critical radius in cm
 Comparisons: Monte Carlo, diffusion, and transport Chapter | 11 421

Exercise 11.1: One-speed transport theory critical radius.

Use Eq. (11.14) (DT approximation) to calculate the eigenvalue and the critical radius of Godiva with the data given
in this section.

11.3 The classic albedo calculation

In the classic albedo calculation, the reflection is estimated at a surface due to an incident source as shown in Fig. 11.1.
In DT from the infinite slab one-group flux (Table 5.1), the albedo
Jout
β5 (11.15)
Jin
is given by


1 2 2D=L cotha=L
β5
 : (11.16)
1 1 2D=L cotha=L
where a is the slab thickness.
Fig. 11.2 shows the albedos for light water (LW), beryllium (Be), aluminum (Al), graphite (C), and heavy water
(HW) using thermal data listed in the MATLAB program CH11_AlbedoSlabDiffTh.m (Annex A). The highest albedo
(0.7369) is for a water slab, saturated at a thickness B10 cm, with progressively lower values for Be, Al, C and the
lowest for heavy water.
In Section 6.2.1, the albedo from transport theory was given as
pffiffiffiffiffiffiffiffiffiffiffi
φð0; 2 μÞ 5 1 2 1 2 cH ðμÞ:

FIGURE 11.1 Reflection of an incident beam of neutrons.

FIGURE 11.2 Albedos for light water, beryllium (Be), aluminum (Al),
graphite (C), and heavy water (HW) due to an anisotropic source inci-
dent on the left face. [Annex A: CH11_AlbedoSlabDiffTh.m]
422 Nuclear Engineering

FIGURE 11.3 Albedo for 1 MeV neutrons incident on the left face of
a water slab of thickness 10 cm. (μ1 5 2 1; μ2 5 2 0:8; μ3 5 2 0:6; μ4 5
2 0:4, μ5 5 2 0:2, μ6 5 2 1).

FIGURE 11.4 Albedo for a water slab of thickness 10 cm for 1 MeV inci-
dent neutrons.

The MC results for albedo were obtained from a MCNP simulation (input file ANNEX B: BK11albd) for 1 MeV
mono-energetic neutrons incident on the left face of a slab of water of thickness 10 cm. The number of neutrons simu-
lated was 400,000 giving a relative error of , 1%. The albedos obtained from a MCNP simulation are shown in
Fig. 11.3 as a function of the reflection angle (μ1 5 2 1; μ2 5 2 0:8; μ3 5 2 0:6; μ4 5 2 0:4, μ5 5 2 0:2, μ6 5 2 1).
The highest reflection in the range 21 , μ , 2 0:8 is for low energy neutrons (E , 0:2 MeV) since higher energy neu-
trons penetrate deep into the medium.
The energy-integrated albedos in water are shown in Fig. 11.4; the large-angle reflection is most prominent falling
off with decreasing angle.

Exercise 11.2:

From the steady-state one-group diffusion equation (Eq. 5.3), show that for an incident source S n/s on the left side
of a slab of thickness a, and extrapolated distance d, the flux φðxÞ and albedo β are:
x x
φðxÞ 5 A cosh 1 Bsinh (11.17)
L L
where the constants A and B are
a1d
A 5 2 Btanh
L
 Comparisons: Monte Carlo, diffusion, and transport Chapter | 11 423

TABLE 11.4 Albedo for water and graphite slabs (MC simulations).

μ Water SS316
21:0; 2 0:8 0.246371 0.0086 0.207984 0.0031
20:8; 2 0:6 0.172610 0.0107 0.169504 0.0035
20:6; 2 0:4 0.104693 0.0144 0.130964 0.0040
20:4; 2 0:2 0.0529103 0.0208 0.086881 0.0051
20:2; 0 0.0120528 0.0446 0.026080 0.0096

and the albedo is

a1d
Jout p 2 2D
L coth L
β5 5 a1d
;
Jin p 1 2D
L coth L

where p 5 Σs =Σt .
Note: The incident and reflected currents are
Σs D dφ
Jin ð0Þ 5 φ ð 0Þ 2 j
4Σt 2 dx x50
and
Σs D dφ
Jout ð0Þ 5 φð0Þ 1 j
4Σt 2 dx x50
and the net current is J 5 Jin 2 Jout .
The energy-integrated albedos are listed in Table 11.4; in water, the back-scattering is predominant for a mono-
energetic anisotropic source.
Comparing the diffusion values of albedo in Fig. 11.2 with MC simulation values in Fig. 11.4 for 1 MeV neutrons, a
large difference is observed. This is due to the energy of incident neutrons. For water, the albedo increases from
0.588636 (0.0040) to 0.74175 (0.0023) as the energy of the incident neutrons decreases from 1 MeV to 0.1 eV.

11.4 Flux in a slab

Three idealized and illustrative cases in slab geometry are infinite medium with an incident source, finite medium with
incident source on the left boundary, and finite medium with an isotropic source.

11.4.1 Diffusion theory

Analytical solutions for infinite media fixed source problems are typically in the form of decaying exponentials while
finite media have algebraic, trigonometric, or hyperbolic functions.
Some elementary cases for non-multiplying media in DT for slab geometry (Section 5.2.1; Table 5.1) are:
G infinite medium with an incident source S0 neutrons/s, for which the boundary conditions of limAx-0 J 5 S0 =2 and
x-N give the flux

So
φ5 expð 2 κjxjÞ; x 6¼ 0 (11.18)
2κD
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
where κ 5 Σa =D the inverse of the diffusion length expressed in terms of the total cross-section
 21=2
L 5 ð3Σt Σa Þ21=2 5 3ð12cÞ
which, is the same as the eigenvalue ν 0 given by the transcendental equation (Eq. 11.11) in Table 11.3.
G finite slab xAð0; ΔÞ, with an impinging flux at the left surface of a slab Jð0Þ 5 FL and φðΔÞ 5 0
424 Nuclear Engineering

 source at the center, the boundary conditions are φð0Þ 5 φðΔÞ 5 0 and
G finite slab with
an isotropic
limx-0 J 5 limx-0 2Ddφ=dx 5 S0 =2. For the point isotropic source, the flux

So L h i21
φ ð xÞ 5 11e2Δ=L e2jxj=L 2 eðjxj2ΔÞ=L (11.19)
2D
which can also be expressed in the following forms:
So L sinh½ðΔ 2 2jxjÞ=2L
φ ð xÞ 5 (11.20)
2D coshðΔ=2LÞ
and
2So X 1 lπx
φðxÞ 5 lodd cos : (11.21)
ΔΣa 1 1 B2l L2 Δ
The transport correction in the diffusion model is usually made by considering the slab to be of physical dimension
Δ 2 2d where the extrapolation distance is d 5 0:71λtr . The extension of the eigenfunction form of the solution is easily
made to the 3D case where the flux φðx; y; zÞ can be found as
8So X 1 lπx mπy nπz
φ5 3 l; m; n;odd cos cos cos : (11.22)
Δ Σa 1 1 B 2
l;m;n L2 Δ Δ Δ
where

 π 2
B2l;m;n 5 l2 1 m2 1 n2 :
a

11.4.2 Transport theory

In transport theory, some elementary cases (Section 6.2) for slab geometry are
G source-free infinite medium, when separation of variables is assumed with eigenfunctions ψυ ðμÞ and eigenvalues ν
(Table 11.3) to give an asymptotic solution

φðx; μÞ 5 e2x=ν ψν ðμÞ (11.23)

based on the 1D transport equation with isotropic scattering.
  ð
@ c 1 0
0
μ 1 1 φðx; μÞ 5 dμ φ x; μ (11.24)
@x 2 21

G infinite medium spherical harmonics method in the P1 approximation

rffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffi
1 3 2 3ð12cÞx
φðxÞ 5 e (11.25)
2 12c

G infinite medium with plane isotropic source has a transient behavior (in the sense of collisional equilibrium) in the
P3 approximation, for which a Fourier transform of Eq. (11.23) for a plane source δðxÞ is
Ð
1 N eikx
φðxÞ 5 2π 2N ð1 2 cÞ 1 gðkÞwhere

7 k 1 15k
9 4 2
gð k Þ 5 :
7 k 1 15
55 2

Evaluating the residues at the simple poles, yields

 
7 2 55
7 m 1 15 2mjxj
2
2 55
7 n 1 15 2njxj
2
φðxÞ 5 e 2 e (11.26)
9 2mðn2 2 m2 Þ 2nðn2 2 m2 Þ
 Comparisons: Monte Carlo, diffusion, and transport Chapter | 11 425

where m and n, and the eigenvalues, which are the relaxation lengths, υ0  1=m and υi  1=n, are given in Table 11.5.
Spatial attenuation is, as for the cases above, defined by eigenvalues that include ν 0 . Far from the source, the asymp-
totic part will dominate (Bell & Glasstone, 1979). Conversely, in strongly absorbing media, the transient part will domi-
nate as there will hardly be any collisional equilibrium. The source appears as a normalization factor for collisional
equilibrium while it dominates for weak equilibrium.
G finite 1D slab, Ganapol (2008) with the impinging (on left surface) and vacuum (on right surface) boundary condi-
tions for: and μ . 0; φð0; μÞ 5 FL ðμÞ and φðΔ; μÞ 5 0: The exact FN solution, based on the integral equations, for
this case has been given as
ð1 X
N 21 ð1
2x=μ
φðxÞ 5 dμFL ðμÞe 1 cα xÞ dμφα ðμÞ (1)
0 α50 0

where φα ðμ) are basis functions and the functions cα ðxÞ are determined from the procedure defined.

11.4.3 Monte Carlo simulation

The MC simulation uses the collision estimator (CE) to estimate the total number of collisions in a region, that is, the
weight participating in collisions in a region s ψ 5 Σt ϕ 5 , pi . so that ϕCE 5 , pi . =Σt , and the track-length estima-
tor (TLE) For the TLE, in which the flux is the mean weighted track-length expressed as
N s 1
φTLE 5 nv 5   , pi s i . :
V t Vt

11.4.4 Comparison
The simulation is carried out for strong absorbing materials, such as boron and gadolinium, to strong scattering materi-
als such as iron and aluminum using one-group data (Lamarsh & Baratta, 1955) listed in Table 11.6.
For a unit isotropic source located in the center of a cube of 5 m.f.p. the thermal flux, obtained from DT (ANNEX
C, MATLAB Program: ExactSolSlabJan03) is shown in Fig. 11.5.
It can be seen that the flux in a high scattering material such as aluminum will attenuate slower than that in a highly
absorbing material. Both gold (Au) and boron (B) are highly absorbing and hence the flux drops rapidly.
The leakage, per unit source neutron, computed for the slabs is as follows: Al: 7.802 3 1022, Fe: 4.832 3 1022, U:
5.063 3 1023, Au: 5.146 3 1024, B: 4.648 3 1024, Gd: 3.532 3 1024.

TABLE 11.5 Infinite medium eigenvalues from P3 transport theory.

c 0.5 0.6 0.8 0.9 0.95 0.99

υ0 1.0114 1.0838 1.4047 1.9029 2.6350 5.7971
υi 0.4094 0.4271 0.4660 0.4866 0.4969 0.5051

TABLE 11.6 Thermal cross-section data.

Element σa (b) σs (b) p 5 σs =σt

Aluminum 0.235 1.4 0.8563
Boron 759 4.0 0.0052
Iron 2.53 11.0 0.8130
Gold 98.8 9.3 0.0860
Gadolinium 46,000 4.0 8.695 3 1025
Uranium 7.5 8.3 0.5253
426 Nuclear Engineering

FIGURE 11.5 One-speed flux in a 1D slab.

FIGURE 11.6 One-speed flux in 1D slab (DT vs P1 ).

A comparison of DT and spherical harmonics P1 approximation for U and Au is shown in Fig. 11.6. The significant
difference is near the boundaries where the validity of Fick’s law is doubtful.
Fig. 11.7. shows the scalar flux in a 1D slab of gadolinium with an isotropic source at the center. It is seen that for a strong
absorber, there is significant difference between the P3 result and the DT result near the source and near the boundaries. Thus,
for a strong absorber, the sharper rise due to lack of collisional equilibrium is better modeled by the P3 approximation.
For a slab with an incident source Jin on the left side of the surface (Fig. 11.8), the DT fluxes are plotted for light
water (LW), beryllium (Be), aluminum (Al), graphite (C), and heavy water (HW). The sharpest attenuation is for water
which is the most absorbing of the five materials considered; while heavy water has the slowest attenuation. Beyond a
slab thickness of B11 cm, the flux has become negligible due which the albedo in Fig. 11.2 saturates to a constant value.
To consider a more detailed flux behavior, a simulation carried out with MCNP (ANNEX B input file) for an aniso-
tropic 1 MeV mono-energetic source incident from the left face of water and steel (SS-316) slabs, shows a rapid drop
in Fig. 11.9. At the interface, the flux in both cases is high (B100 times higher than in the next 1 cm layer). The flux
in water falls off faster than that in SS-316 (higher c than that for water).
It is also important to visualize the energy spectrum of the flux in a hydrogenous medium such as water and a high
c material such as SS-316.
As shown in Fig. 11.10, the energy dependent flux in SS-316 has a hard spectrum while a Maxwellian is observed
in water at thermal energies (B0.025 eV).
Now consider the MC collision density in a 5 m.f.p thick aluminum slab with a point isotropic source at the center.
Comparisons: Monte Carlo, diffusion, and transport Chapter | 11 427

FIGURE 11.7 One-speed flux in Gd: DT, P1 ; P3 comparisons.

FIGURE 11.8 Neutron flux for light water, beryllium (Be), aluminum
(Al), graphite (C), and heavy water (HW) due to an anisotropic source
incident on the left face. [Program CH11_AlbedoSlabDiffTh.m].

FIGURE 11.9 Neutron flux versus distance (x) in water and SS-316
slabs of thickness 10 cm due to a 1-MeV mono-energetic anisotropic
source incident from the left face. (ANNEX B input file).
428 Nuclear Engineering

FIGURE 11.10 Neutron flux versus energy E in water and SS-316

slabs of thickness 10 cm due to a 1-MeV mono-energetic anisotropic
source incident from the left face.

FIGURE 11.11 Comparison of collision density:

Monte Carlo versus diffusion theory.

Fig. 11.11 shows the thermal flux computed with the collision density estimator (CDE) and the one-group diffusion
expression. It can be seen that DT over-estimates the flux, especially near the source. However, the difference is smaller
further from the source.
The difference between the CDE and TLE flux estimates for a sample size 1000 3 5 (1000 histories and 5 batches)
is ,2%. The transmission from the left (and right) surfaces is ,4% for aluminum. For iron and boron the transmission
was B3.5% and 0.6.% respectively.

11.5 Flux in a finite sphere with a point isotropic source

11.5.1 Diffusion theory
In Section 5.2.2, the neutron DT flux is given by
S sinhκðR 1 d 2 rÞ
φðr Þ 5
4πDr sinhκðR 1 dÞ
 Comparisons: Monte Carlo, diffusion, and transport Chapter | 11 429

and the leakage from the sphere is given by

SκðR 1 d Þ
L 5 4πR2 J ðRÞ 5 2 4πR2 Drφ 5 :
sinhκðR 1 d Þ
Consider an aluminum sphere of radius R 5 25 cm, ρ 5 2:70 g/cm3 (mass 5 176.7 kg) with the data: σs 5 1:4 b,
σa 5 0:235 b, N 5 0:06027 nuclei (b.cm)21, Σt 5 0:0985 cm21.
With this data, the mean free path is λ 5 1=Σt 5 10:1478 cm, so that the sphere is of radius 2.4636 mean free paths.
From theffi data:μ0 5 2=3A 5 0:0247, Σtr 5 0:0823cm21, d 5 0:71=Σtr 5 8:6275 cm, d 5 0:8502 fp, D 5 4:0505 cm,
pffiffiffiffiffiffiffiffiffiffiffi
L 5 D=Σa 5 16:9108 cm, L  L=λ 5 1:6664, c  Σs =Σt 5 0:8563.
From the one-speed transport theory spherical harmonics P1 approximation, the quantity LT is equivalent to the
diffusion length (since 1 2 c is the macroscopic p ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi cross-section) with the diffusion coefficient D 5 1=3 in
absorption
units of mean free paths. For this data, LT 5 1= 3ð1 2 cÞ 5 1:5229. However, transport theory gives the (normal-
ized) relaxation length as ν 0 5 1:6115. The DT flux shown in Fig. 11.12 becomes very large at r 5 0 where the source
is located.
Fig. 11.13 is the thermal flux (DT) for a point isotropic source at the center of a sphere; compared with Fig. 11.8 it
has the same shape except that light water is at the top. This shows the effect of geometry on the attenuation of flux.

FIGURE 11.12 Diffusion Theory flux in an aluminum sphere.

FIGURE 11.13 Thermal flux (diffusion theory) for a point isotropic

source at the center of a sphere.
430 Nuclear Engineering

11.5.2 Transport theory exact solution

In Section 6.2, the flux in a sphere with point isotropic source equations is obtained with a program written in
MATLAB using a Gaussian quadrature with N 5 32. The expansion coefficients are obtained by solving the Fredholm
equations iteratively. For c 5 0:8563 (the one-group data for aluminum) (Lamarsh & Baratta, 2001), gives ν 0 5 1:6183,
N ðν 0 Þ 5 0:2669, H ðν 0 Þ 5 1:903, Gðν 0 Þ 5 1:1601, Eðν 0 Þ 5 1:1627, and the extrapolation distance, from Eq. (11.7),
z0 5 0:83024.
Table 11.7 lists the results, from DT and transport theory. Clearly, DT does not give good results near the source
and boundary. The ratio of the transport-to-diffusion flux is close to unity in the range 0.742.2 m.f.p. This will have
an important bearing when choosing a reference for the Kullback-Leibler estimates.
To better understand the fluxes in Fig. 11.13, a MCNP simulation was carried out to compare fluxes in water, ura-
nium and aluminum. The simulation used 50,000 histories with a point isotropic mono-energetic source of 1 MeV
located at the center of the sphere of radius 8.4710 cm (the same as for Godiva).
The neutron fluxes shown in Fig. 11.14 validate Fig. 11.13 with the scalar flux in water being higher than that for
aluminum and uranium.
Although, in Fig. 11.14, the neutron flux is higher in water, the emergent spectrum, given in Table 11.8, shows that
water moderates the high incident energy while aluminum and uranium do not result in comparable shielding of neutron
radiation.

TABLE 11.7 Diffusion theory flux compared with transport theory.

r=R r r φðrÞ 4πr 2 φðrÞ

cm m.f.p n/cm2/s DT TT
0.0 0.0 0.0 — — 1.0
0.1 2.5 0.2464 6.7341e-003 0.5289 1.1017
0.2 5.0 0.4927 2.8785e-003 0.9043 1.1823
0.3 7.5 0.7391 1.6353e-003 1.1559 1.2413
0.4 10.0 0.9854 1.0406e-003 1.3076 1.2788
0.5 12.5 1.2318 7.0198e-004 1.3783 1.2946
0.6 15.0 1.4782 4.8905e-004 1.3827 1.2882
0.7 17.5 1.7245 3.4613e-004 1.3321 1.2586
0.8 20.0 1.9709 2.4557e-004 1.2344 1.2032
0.9 22.5 2.2172 1.7214e-004 1.0951 1.1157
1.0 25.0 2.4636 1.1678e-004 0.9172 0.9643

FIGURE 11.14 Neutron flux versus cell number in a sphere of radius

8.7410 cm divided into 50 equal-volume cells.
 Comparisons: Monte Carlo, diffusion, and transport Chapter | 11 431

TABLE 11.8 Emerging current spectrum from a sphere of radius 8.471 cm.

Energy bin Water Aluminum Uranium

01 eV 1.65416 3 1021 0.0092 0 0
1 eV1 keV 1.70841 3 1021 0.0098 0 0
1 keV1 MeV 6.26264 3 1021 0.0035 9.99463 3 1021 0.0000 8.99460 3 1021 0.0008
Total 9.62522 3 1021 0.0005 9.99463 3 1021 0.0000 8.99460 3 1021 0.0008

FIGURE 11.15 Neutron flux versus radius in a sphere of radius

8.7410 cm divided into 50 equal-volume cells.

A brief discussion on Fig. 11.9 in which a very high flux appears at the interface where source neutrons are incident
is helpful to clarify this artefact. In both slab and spherical geometry, when equal volume regions are specified, the first
volume has a larger radius (difference) and the fluxes appear to take a shape illustrated in Fig. 11.15 which is the same
result as shown in Fig. 11.14. So, the magnitude of the flux is very high in a region adjacent to the source in both slab
and spherical geometries.

11.5.3 Monte Carlo simulation

Fig. 11.16 shows a comparison of the exact DT flux with the TLE flux in the 30 equal-volume regions of an aluminum
sphere of radius 25 cm (2.4636 m.f.p) with a mass of 176.7 kg. To comment on the accuracy of the results, we need to
consider two parameters viz the mean free path λ and the scattering probability ps . For this one-speed model,
λB10:15cm and ps B0:86. We can thus expect that there will be a sufficient number of collisions for the CE flux esti-
mate to agree well with the TL estimate. For this case, the average number of collisions per source neutron is B1.88 in
the first region gradually decreasing to B0.17 in the last region which is about 2.4 m.f.p’s away from the source. For a
sample size NB10; 000 this means that we will have of the order of 19,000 collisions in the first region and B1700
collisions in the last region. Similarly, the relatively high scattering probability makes this material far from an
absorber. It is seen that these two factors contribute to the good agreement between DT and the MC estimate from the
mid-point of the first region which is half a mean free path away. Similarly, good agreement is observed for the leak-
age, which is 0.54319 from the MC simulation, compared with 0.5548 from the exact solution.
Comparisons of diffusion and transport flux show good agreement far from the boundaries. The convergence can be
verified using the Kullback-Leibler divergence indicating that NB10,000 is a sufficient sample size for simulation.
Consider now the individual cell balances shown in Table 11.9. The steady-state balance requires

JinðcÞ 1 SðcÞ 5 Jout

ðcÞ
1 AðcÞ
432 Nuclear Engineering

FIGURE 11.16 Monte Carlo compared with transport and diffusion

theory.

TABLE 11.9 Cell balances using diffusion theory flux.

Cell Sources Sinks Δ

SðcÞ JðcÞ
in
Total In AðcÞ JðcÞ
out
Total Out

1 1 0 1.0000 0.1467 0.8533 1.0000 0

2 0 0.8533 0.8533 0.0590 0.7943 0.8533 0
3 0 0.7943 0.7943 0.0411 0.7532 0.7943 0
4 0 0.7532 0.7532 0.0317 0.7215 0.7532 0
5 0 0.7215 0.7215 0.0256 0.6958 0.7215 0
6 0 0.6958 0.6958 0.0213 0.6745 0.6958 0
7 0 0.6745 0.6745 0.0181 0.6564 0.6745 0
8 0 0.6564 0.6564 0.0155 0.6409 0.6564 0
9 0 0.6409 0.6409 0.0134 0.6276 0.6409 0
10 0 0.6276 0.6276 0.0116 0.6160 0.6276 0

Ð
The reaction rate RkðcÞ for a reaction k in a cell c between radii r ðcÞ and r ðc11Þ RðcÞ
k 5 Σck φðr ÞdV is
 
SΣk r ðc11Þ
 r ðcÞ
 1
 1
RðcÞ
k 5 2 coshκ R 1 d 2 r ðc11Þ 1 coshκ R 1 d 2 r ðcÞ 2 2 sinhκ R 1 d 2 r ðc11Þ 1 2 sinhκðR 1 d 2 rðcÞ Þ
DsinhκðR 1 dÞ κ κ κ κ
 
S sinhκðR 1 dÞ 2 sinhκd 2 κRcoshκd
RðTÞ 5 Σk
k
Σa sinhκðR 1 dÞ
where RðTÞ ðcÞ
k , obtained by setting r 5 0; r
ðc11Þ
5 R, is the total reaction rate of reaction type k in the sphere. With the
leakage given by
ð ð  
dφ r1dr
L 5 r ∙ J dV 5 2 Dr2 φ dV 5 2 4πD r 2 jr
dr
ðc Þ
5 4πr22 J ðr2 Þ 2 4πr12 J ðr1 Þ 5 Jout 2 JinðcÞ
where
 
dφ S κrcosh½κðR 1 d 2 r Þ 1 sinh½κðR 1 d 2 r Þ
J ðr Þ 5 2 D 5 :
dr 4πr 2 sinhκðR 1 dÞ
 Comparisons: Monte Carlo, diffusion, and transport Chapter | 11 433

Elementary solutions and simulations show that

1. DT, based on Fick’s law, is valid in the “interior” of a medium and not near a source or boundary;
2. the asymptotic relaxation length υ0 for a source-free infinite medium transport solution is identical to that from DT,
when it is called the diffusion length L, for a pure scattering medium (c 5 1), and
3. the relaxation length, expressed by the eigenvalues, has a “transient” and an “asymptotic” part which is a function
of the medium properties (c).

Problems
1. In Section 11.1, the eigenvalues νo are defined by the transcendental equation (Eq. 11.11). Calculate νo for
c 5 1:0; 1:2; 1:4 and compare with values in Table 11.3.
2. Repeat Q.1 using values of for c 5 0:2; 0:4; 0:6; 0:8; 1:0 and comment on your results, especially in the limits of
low c and the case c 5 1.
3. Explain how the Kullback-Leibler Divergence estimates can be used to comment on the quality of the results in
Fig. 11.16 (MC compared with transport and diffusion estimates).
4. For the slab and spherical geometry results, use a different material from aluminum with a lower c value, such as
boron, and again compare the results. Does MC sample size depend on the scattering power of the material?
5. Plot the relative errors for the above and again, comment on the trend you observe.

Nomenclature
English lower case
c number of neutrons emerging from an interaction
d extrapolation distance
keff effective multiplication
kN infinite multiplication
N atomic number density

English upper case

B buckling
D diffusion coefficient (xx)
D^ derivative operator
H Chandrasekhar H-function
J neutron current
L diffusion length
Ni atomic density of the ith nuclide
Rc critical radius
R extrapolated radius
S source

Greek lower case

β albedo
μ cosine of angle of scattering
μ0 average cosine of scattering angle
ν0 eigenvalue
ν number of neutrons produced per fission
φ flux
φas asymptotic flux
σa microscopic absorption cross section
σf microscopic fission cross section
σr microscopic removal cross section
σs microscopic scattering cross section
σtr microscopic transport cross section
434 Nuclear Engineering

σt microscopic total cross section

τ neutron age
χ fission spectrum

Greek
P
upper case
macroscopic absorption cross section
Pa
macroscopic fission cross section
Pf
macroscopic removal cross section
Pr
macroscopic scattering cross section
Ps
macroscopic transport cross section
Ptr
t macroscopic total cross section

Abbreviations
m.f.p mean free path
CDE collision density estimate
DT diffusion theory
MC Monte Carlo
TLE track-length estimator
SS-316 Stainless Steel Type 316

References
Bell, G. I., & Glasstone, S. (1979). Nuclear reactor theory. New York: Robert E. Krieger Publishing Co., Inc..
Clark, M., & Hansen, K. F. (1964). Transport methods. Numerical methods of reactor analysis. Elsevier.
Ganapol, B. D. (2008). Analytical benchmarks for nuclear engineering applications case studies in neutron transport theory. Nuclear Energy Agency.
Lamarsh, J. R., & Baratta, A. J. (2001). Introduction to nuclear engineering (3rd ed., p. 783)Prentice Hall.
Lamarsh, J. R., & Baratta, A. J. (1955). Introduction to nuclear engineering. McGraw-Hill.
 Comparisons: Monte Carlo, diffusion, and transport Chapter | 11 435

Annex A MATLAB Program AlbedoSlabDiffTh.m (Section 11.3)

% CH11_AlbedoSlabDiffTh.m
% Aluminum data Lamarsh (INE)
N=0.06027e23; siga=0.235e-24;Siga=N*siga; D=4.0505;
L=sqrt(D/Siga);kappa=1/L;a=80; % slab thickness
d=8.6275;den=2.70;S=1.0;
% Diffusion coefficient and thermal diffusion length (source L&B)
% water heavy water beryllium graphite
Dw=0.16;Lw=2.85; Dhw=0.87;Lhw=97; Dbe=0.50;Lbe=21; Dc=0.84;Lc=59;

% slab thickness a
% S=1 incident from left side (J+)
% phi(x) =B1 cosh x/L + B2 sinh (x/L)

B2Al =(-4*S*cosh(a/L))/( (2*D/L)*cosh(a/L) + sinh(a/L)); B1Al =4*(S+(D/

(2*L))* B2Al);
B2w =(-4*S*cosh(a/Lw))/((2*Dw/Lw)*cosh(a/Lw) + sinh(a/Lw));B1w =4*(S+(Dw/
(2*Lw))* B2w);
B2hw =(-4*S*cosh(a/Lhw))/((2*Dhw/Lhw)*cosh(a/Lhw) +
sinh(a/Lhw));B1hw=4*(S+(Dhw/(2*Lhw))*B2hw);
B2be =(-4*S*cosh(a/Lbe))/((2*Dbe/Lbe)*cosh(a/Lbe) +
sinh(a/Lbe));B1be=4*(S+(Dbe/(2*Lbe))*B2be);
B2c =(-4*S*cosh(a/Lc))/( (2*Dc/Lc)*cosh(a/Lc) + sinh(a/Lc));B1c
=4*(S+(Dc/ (2*Lc))* B2c);

Steps=1000;Del=a/Steps;Len=0;
for i=1:Steps
Len=Len+Del; x(i)=Len;
phi(i) = B1Al*cosh(x(i)/L) + B2Al *sinh(x(i)/L);
phiw(i) = B1w *cosh(x(i)/Lw) + B2w *sinh(x(i)/Lw);
phihw(i) = B1hw*cosh(x(i)/Lhw) + B2hw *sinh(x(i)/Lhw);
phibe(i) = B1be*cosh(x(i)/Lbe) + B2be *sinh(x(i)/Lbe);
phic(i) = B1c *cosh(x(i)/Lc ) + B2c *sinh(x(i)/Lc);
end

for j=1:8
SlabTh(j)=10*j+5;
ggAl=(2*D/L)*coth(SlabTh(j)/L); albedoAl(j) = (1-ggAl)/(1+ggAl);
ggw =(2*D/L)*coth(SlabTh(j)/Lw); albedow(j) = (1-ggw)/(1+ggw);
gghw=(2*D/L)*coth(SlabTh(j)/Lhw); albedohw(j) = (1-gghw)/(1+gghw);
ggbe=(2*D/L)*coth(SlabTh(j)/Lbe); albedobe(j) = (1-ggbe)/(1+ggbe);
ggc=(2*D/L)*coth(SlabTh(j)/Lc); albedoc(j) = (1-ggc)/(1+ggc);
end

figure(1)
set(gca,'FontSize',12)
plot(x,phiw,'r-','LineWidth',1.5)
hold on
plot(x,phibe,'b-','LineWidth',1.5)
hold on
plot(x,phi,'k-','LineWidth',1.5)
hold on
plot(x,phic,'m-','LineWidth',1.5)
hold on
plot(x,phihw,'g-','LineWidth',1.5)
hold on
436 Nuclear Engineering

legend('LW','Be','Al','C','HW',12,'Location','NorthEast')
%text(0.1,0.7,'\bf Water','fontsize',12)
%xlim([1e-8 2])
ylim([0 4])
%hold on
grid off
xlabel('{\bf x} (cm)','fontsize',14)
ylabel('{\bf \phi(x)} (n cm^{-2} s^{-1})','fontsize',14)

figure(2)
set(gca,'FontSize',12)
plot(x,phiw,'k--','LineWidth',1.5)
hold on
plot(x,phibe,'k:','LineWidth',1.5)
hold on
plot(x,phi,'k-','LineWidth',1.5)
hold on
plot(x,phic,'k-.','LineWidth',1.5)
hold on
plot(x,phihw,'k--','LineWidth',1)

legend('LW','Be','Al','C','HW',12,'Location','NorthEast')
%text(0.1,0.7,'\bf Water','fontsize',12)
%xlim([1e-8 2])
ylim([0 4])
%hold on
grid off
xlabel('{\bf x} (cm)','fontsize',14)
ylabel('{\bf \phi(x)} (n cm^{-2} s^{-1})','fontsize',14)

figure(3)
set(gca,'FontSize',12)
plot(SlabTh,albedow,'r-','LineWidth',1.5)
hold on
plot(SlabTh,albedobe,'b-','LineWidth',1.5)
hold on
plot(SlabTh,albedoAl,'k-','LineWidth',1.5)
hold on
plot(SlabTh,albedoc,'m-','LineWidth',1.5)
hold on
plot(SlabTh,albedohw,'g-','LineWidth',1.5)
hold on

legend('LW','Be','Al','C','HW',12,'Location','SouthEast')
%text(0.1,0.7,'\bf Water','fontsize',12)
%xlim([1e-8 2])
%ylim([1e-3 2e-1])
%hold on
grid off
xlabel('{\bf Slab Thickness} (cm)','fontsize',14)
ylabel('{\bf Albedo} (n/s)','fontsize',14)
 Comparisons: Monte Carlo, diffusion, and transport Chapter | 11 437

figure(4)
set(gca,'FontSize',12)
plot(SlabTh,albedow,'k--','LineWidth',1.5)
hold on
plot(SlabTh,albedobe,'k:','LineWidth',1.5)
hold on
plot(SlabTh,albedoAl,'k-','LineWidth',1.5)
hold on
plot(SlabTh,albedoc,'k-.','LineWidth',1.5)
hold on
plot(SlabTh,albedohw,'k--','LineWidth',1)

legend('LW','Be','Al','C','HW',12,'Location','SouthEast')
%text(0.1,0.7,'\bf Water','fontsize',12)
%xlim([1e-8 2])
%ylim([1e-3 2e-1])
%hold on
grid off
xlabel('{\bf Slab Thickness} (cm)','fontsize',14)
ylabel('{\bf Albedo} (n/s)','fontsize',14)
438 Nuclear Engineering

Annex B MCNP Input File BK11Albd (Section 11.2)

BK Ch11 Albedo comparison with H function

1 12 -1.0 1 -2 3 -4 5 -6 imp:n=1
2 0 (-1:2:-3:4:-5:6) imp:n=0

1 py 0
2 py 10
*3 px -0.5
*4 px 0.5
*5 pz -0.5
*6 pz 0.5

c BLOCK 3 .............................................
SDEF POS=0 0 0 AXS=0 1 0 EXT=0 RAD=d1 PAR=1 ERG=1 VEC=0 1 0 DIR=1
SI1 0 0.05 $ radial sampling range: 0 to Rmax (=0.05cm)
SP1 -21 1 $ radial sampling weighting: r^1 for disk
c
mode n
c STEEL(316)(7.92 g/cm3)
m3 14000 -0.010 24000 -0.170 25055 -0.020 26000 -0.655
28000 -0.120 42000 -0.025
c Water (1.0 g/cm3)
m12 1001 2 8016 2
mt12 lwtr.02t
FC1 neutrons crossing surface 1, 2
F1:N 1 2
E1 0.2 0.4 0.6 0.8 1.0
C1 -0.8 -0.6 -0.4 -0.2 0 1.0 $ six bins
c ENERGY SPECTRUM OF FLUX IN SLAB
fc64 energy spectrum of flux in slab
f64:n 1
e64 3.77e-9 4.05e-9 4.36e-9 4.69e-9 5.04e-9 5.42e-9 5.83e-9 6.27e-9
6.74e-9 7.25e-9 7.79e-9 8.38e-9 9.01e-9 9.69e-9 1.04e-8 1.12e-8
1.20e-8 1.29e-8 1.39e-8 1.50e-8 1.61e-8 1.73e-8 1.86e-8
2.e-8 2.15e-8 2.31e-8 2.49e-8 2.68e-8 2.88e-8 3.09e-8
3.33e-8 3.58e-8 3.85e-8 4.13e-8 4.45e-8 4.78e-8 5.14e-8
5.53e-8 5.94e-8 6.39e-8 6.87e-8 7.39e-8 7.95e-8 8.54e-8
9.19e-8 9.88e-8 1.06e-7 1.14e-7 1.23e-7 1.32e-7 1.42e-7
1.53e-7 1.64e-7 1.77e-7 1.90e-7 2.04e-7 2.19e-7 2.36e-7
2.54e-7 2.73e-7 2.93e-7 3.15e-7 3.39e-7 3.65e-7 3.92e-7
4.22e-7 4.53e-7 4.88e-7 5.24e-7 5.64e-7 6.06e-7 6.52e-7
7.01e-7 7.54e-7 8.10e-7 8.71e-7 9.37e-7 1.01e-6 1.08e-6
1.16e-6 1.25e-6 1.35e-6 1.45e-6 1.56e-6 1.67e-6 1.80e-6
1.94e-6 2.08e-6 2.24e-6 2.41e-6 2.59e-6 2.78e-6 2.99e-6
3.22e-6 3.46e-6 3.72e-6 4.e-6 4.30e-6 4.62e-6 4.97e-6
5.35e-6 5.75e-6 6.18e-6 6.65e-6 7.15e-6 7.69e-6 8.26e-6
8.89e-6 9.56e-6 1.03e-5 1.10e-5 1.19e-5 1.28e-5 1.37e-5
1.48e-5 1.59e-5 1.71e-5 1.84e-5 1.97e-5 2.12e-5 2.28e-5
2.45e-5 2.64e-5 2.84e-5 3.05e-5 3.28e-5 3.53e-5 3.79e-5
4.08e-5 4.39e-5 4.72e-5 5.07e-5 5.45e-5 5.86e-5 6.31e-5
6.78e-5 7.29e-5 7.84e-5 8.43e-5 9.06e-5 9.75e-5 1.05e-4
1.13e-4 1.21e-4 1.30e-4 1.40e-4 1.51e-4 1.62e-4 1.74e-4
1.87e-4 2.01e-4 2.17e-4 2.33e-4 2.50e-4 2.69e-4 2.89e-4
3.11e-4 3.35e-4 3.60e-4 3.87e-4 4.16e-4 4.47e-4 4.81e-4
5.17e-4 5.56e-4 5.98e-4 6.43e-4 6.91e-4 7.43e-4 7.99e-4
8.60e-4 9.24e-4 9.94e-4 1.07e-3 1.15e-3 1.24e-3 1.33e-3
 Comparisons: Monte Carlo, diffusion, and transport Chapter | 11 439

1.43e-3 1.54e-3 1.65e-3 1.78e-3 1.91e-3 2.05e-3 2.21e-3

2.37e-3 2.55e-3 2.75e-3 2.95e-3 3.17e-3 3.41e-3 3.67e-3
3.95e-3 4.24e-3 4.56e-3 4.91e-3 5.28e-3 5.67e-3 6.10e-3
6.56e-3 7.05e-3 7.58e-3 8.15e-3 8.77e-3 9.43e-3 1.01e-2
1.09e-2 1.17e-2 1.26e-2 1.36e-2 1.46e-2 1.57e-2 1.68e-2
1.81e-2 1.95e-2 2.09e-2 2.25e-2 2.42e-2 2.60e-2 2.80e-2
3.01e-2 3.24e-2 3.48e-2 3.74e-2 4.02e-2 4.33e-2 4.65e-2
5.e-2 5.38e-2 5.78e-2 6.22e-2 6.69e-2 7.19e-2 7.73e-2
8.32e-2 8.94e-2 9.61e-2 1.03e-1 1.11e-1 1.20e-1 1.29e-1
1.38e-1 1.49e-1 1.60e-1 1.72e-1 1.85e-1 1.99e-1 2.14e-1
2.30e-1 2.47e-1 2.66e-1 2.86e-1 3.07e-1 3.30e-1 3.55e-1
3.82e-1 4.10e-1 4.41e-1 4.75e-1 5.10e-1 5.49e-1 5.90e-1
6.34e-1 6.82e-1 7.33e-1 7.89e-1 8.48e-1 9.12e-1 9.81e-1
1.05 1.13 1.22 1.31 1.41 1.52 1.63
1.75 1.88 2.03 2.18 2.34 2.52 2.71
2.91 3.13 3.37 3.62 3.89 4.19 4.50
4.84 5.20 5.60 6.02 6.47 6.96 7.48
8.04 8.65 9.30 10
print
c cutoff the neutrons at 0 MeV
cut:n j 0.0
nps 400000
440 Nuclear Engineering

Annex C MATLAB Program CH11ExactSolSlabJan03.m (Section 11.4.4)

Comparisons: Monte Carlo, diffusion, and transport Chapter | 11 441
442 Nuclear Engineering
Comparisons: Monte Carlo, diffusion, and transport Chapter | 11 443
444 Nuclear Engineering
Comparisons: Monte Carlo, diffusion, and transport Chapter | 11 445
446 Nuclear Engineering
Comparisons: Monte Carlo, diffusion, and transport Chapter | 11 447
448 Nuclear Engineering
Chapter 12

Exercises in Monte Carlo simulation

This chapter comprises hands-on exercises in:

1. random sampling from the Watt fission spectrum,
2. Monte Carlo (MC) simulation for the one-group neutron flux in a non-multiplying sphere,
3. search for a reflected spherical critical configuration,
4. reactor core modeling,
5. radiation shielding,
6. perturbation analysis, and
7. geometry modeling.

12.1 Sampling from a distribution function

In the MC simulation of a multiplying medium, the neutron source energy is sampled from a spectrum such as the Watt
fission spectrum (Section 2.9). As described in Chapter 4, random sampling is straightforward when the probability dis-
tribution function (PDF) is simple and analytic; in that case the cumulative distributive function (CDF) is readily
inverted. When the CDF is not easy to obtain, or when the PDF is tabulated, other methods, such as the acceptance-
rejection method, must be used.
For an exponential PDF
f ðEÞ 5 λe2λE (12.1)
the CDF is
F ðEÞ 5 1 2 e2λE (12.2)
and the energy Ei can be sampled as
1

Ei 5 2 ln 1 2 ξi : (12.3)
λ
where ξi is a uniform random number.
In MATLABs, the commands for Eq. (12.3) are

and the mean energy can be found as

As an example, if λ 5 1 MeV, with N 5 1000 Ebar 5 E 5 1:0398 MeV.

Nuclear Engineering. DOI: https://doi.org/10.1016/B978-0-323-90618-0.00012-0

© 2022 Elsevier Inc. All rights reserved. 449
450 Nuclear Engineering

12.1.1 Sampling from a normal distribution

For neutron energy E distributed normally with mean μ and standard deviation σ
1
f ðEÞ 5 pffiffiffiffiffiffi e22ð σ Þ :
1 E2μ 2
(12.4)
2πσ
Fig. 12.1 shows a plot of f ðEÞ for μ 5 MeV and σ 5 MeV. The total area of the curve is 1.0; in MATLAB this is
easily found with the example below

The MATLAB(R) function normpdf can also be used to plot the function

and the function normcdf(x,5,2) gives the CDF at x 5 2.

The energy is sampled from
Ei 5 σξ ðNÞ
i 1μ (12.5)
giving a sampled PDF f1 ðEÞ and
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Ei 5 μ 1 σ 2 2lnξi;1 cos2πξ i;2 (12.6)
giving f2 ðEÞ, also shown in Fig. 12.1. The random numbers ξi;1 and ξi;2 are generated from a uniform distribution while
ξiðNÞ are generated from a normal distribution.
The MATLAB program CH12_NormalSampling.m used for sampling from the normal PDF is listed in Annex A.
In Fig. 12.1, both sampled distributions f1 ðEÞ and f2 ðEÞ are for a sample size N 5 10; 000 which show
acceptable agreement, except for at the ends, with the population distribution f ðEÞ.

Exercise 12.1: From the matlab program given above, compute the sample standard deviation from each sampling
scheme.

FIGURE 12.1 Sampling from a normal distribution with

μ 5 5; σ 5 12.
 Exercises in Monte Carlo simulation Chapter | 12 451

Exercise 12.2: Show that the CDF of the normal distribution is

  
1 x2μ
F ðE Þ 5 1 1 erf pffiffiffi (12.7)
2 σ 2

12.1.2 Sampling from a Watt fission spectrum

The energy of the prompt neutron(s) emitted from fission, E, is found from the Watt Spectrum (Cullen, 2004; Froehner
& Spencer, 1980; Monte Carlo Team, 2005)
pffiffiffiffiffiffi
W ða; b; EÞ 5 Ce2E=a sinh bE (12.8)
where
 3 21=2
πa b
C5 e2ab=4 :
4
and the constants a, b are weak functions of the energy. For neutron-induced fission in U-235 at thermal and 1 MeV
neutron energy a 5 0.988 MeV and b 5 2.249 MeV21.
At low energies E&1MeV
W ða; b; EÞBAE1=2 (12.9)
and at high energies E*1 MeV
pffiffiffiffi
bE2E=a
W ða; b; EÞ 5 Ce (12.10)
From Eq. (12.8), the Watt fission spectrum is shown, on a log-log scale, in Fig. 12.2. Figs. 12.212.4 are obtained
from the MATLAB Program CH12_WattSampling.m listed in Annex B.
Energy is sampled from (Everett, Turner, & Cashwell, 1973)
ab pffiffiffiffiffiffiffiffiffiffiffiffiffiffi L
K 511 ; L 5 a K 1 K2 2 1 ; M 5 2 1
8 a
x 5 2 lnξ1 ; y 5 2 lnξ2 (12.11)
If ðy2M ðx11ÞÞ2 # bLx, then the numbers are accepted and the energy is E 5 Lx; otherwise a new set of random
numbers is generated (program listed in Annex B).and from a Maxwellian in the center of mass system transformed to
the Lab system (Brown; Froehner & Spencer, 1980)

FIGURE 12.2 Watt fission spectrum for U-235 induced fission

(a 5 0.988 MeV, b 5 2.249 MeV21).
452 Nuclear Engineering

FIGURE 12.3 Sampled Watt fission spectra probabilities f 3 ðEÞ and

f 4 ðEÞ.

FIGURE 12.4 PDF and CDF of the U-235 Watt spectrum.

π
w’a 2 lnξ 1 2 lnξ2 cos2 ξ 3
2
a b

2 pffiffiffiffiffiffiffiffiffiffi
E’w 1 1 2ξ 4 2 1 a2 bw (12.12)
4
The sampled spectrum using Eq. (12.11), f3 ðEÞ, and Eq. (12.12), f4 ðEÞ, are shown in Fig. 12.3. For 100,000 points
simulated, the sampled mean energies for f3 ðEÞ and f4 ðEÞ were 2.0343 and 2.0310 MeV respectively. The acceptance-
rejection sampling scheme had a 75.9% acceptance.
The PDF f ðEÞ and CDF FðEÞ are shown in Fig. 12.4. The CDF can be curve-fitted and used for indirect sampling.
Alternatively, the acceptance-rejection scheme can be used for sampling.

Exercise 12.3: Run the above program for sample size N varying from 100 to 1,000,000 and note the accuracy and
sampling time.

Exercise 12.4: For the low- and high-energy approximations given by Eqs. (12.2) and (12.3) find expressions for the
cumulative distribution function to cary out analytical sampling. Again, comment on the accuracy of using these
approximations.
 Exercises in Monte Carlo simulation Chapter | 12 453

Exercise 12.5: Apply the acceptance-rejection scheme for sampling the Watt energy spectrum. Recall from Chapter 4,
the scheme for sampling from a difficult PDF f is based on considered a simple PDF g and choosing a constant c such
that f ðyÞ # cgðyÞ. For a random number ξ (uniform in 0; 1) the point is selected if

f ðξÞ
ξ# :
cgðξÞ

Exercise 12.6: Use Eqs. (12.9) and (12.10) to carry out direct sampling and compare the efficiencies.

12.2 Estimating the neutron flux in a non-multiplying sphere

This section considers a non-multiplying sphere of radius R with a point isotropic source located at the origin emitting
S neutrons/s. Recall that in Section 11.5 the neutron flux was calculated from diffusion theory, transport theory, and
MC simulations. Comparisons were also illustrated for diffusion theory and MC results obtained from a modest pro-
gram MCFSoneProg.m which is listed here as a hands-on exercise.
The main program has 162 lines and calls three functions:
Function and arguments returned Line called from
[RG, RS, Vol] 5 MCFSShells(Rad_sphere, NV); 6
IREG 5 MCFSregion(POS); 28, 34, 77
[Ncols, FLUX] 5 MCFSsameR(FREG, WGT, DTC, Ncols, 38
FLUX);

This program could have been written in fewer lines; the purpose here is to illustrate the philosophy of having many
subroutines and functions to do their bits separately rather than as one main program; this is the way large codes are
written.

12.2.1 The simulation process

Step 1: Enter the basic data
The data entered is:
G Avogadro’s number
G molecular weight
G density
G radius
G number of regions NREG
G nuclear data (σs :σa ; σt Þ
G number of histories N to simulate.
Step 2: Initialize counters
The following counters are initialized
G scattering, absorption and collision counters
G weight participating in the collision WtPartCol
G collision estimator (CE) flux (CEflux)
G track-length estimator (TLE) flux (FLUX).

Step 3: Begin simulation

Source histories are simulated by generating source neutrons at the origin with a starting weight of one isotropically
454 Nuclear Engineering

μ 5 2ξ 1 2 1; θ 5 cos21 μ; ϕ 5 2πξ 2
The distance to collision (DTC) is used to calculate the final position.
1

DTC 5 ln 1 2 ξ3 (12.13)
Σt

Exercise 12.7: What is the purpose of defining bounding surfaces in this program?

The function MCFSregion determines the regions IREG and FREG for the initial and final positions.
 Exercises in Monte Carlo simulation Chapter | 12 455

There are two possibilities viz (1) where the final and initial regions are the same, (IREG 5 FREG, then
MCFSsameR is called) and (2) where the regions are different.

Exercise 12.8: What is the purpose of updating Ncols, FLUX and FLUXsq in MCFSsameR?

In the former, a collision is tallied by updated the number of collisions in this region Ncols, the track-length flux
FLUX, and the weight participating in the collision WtPartCol (which will give the collision estimate of flux). The sec-
ond case, where the initial and final regions are different, has to determine the path of the neutron as it crosses interme-
diate regions to ultimately collide in the final region or when it escapes from the system.
A region number of 1000 is used to represent the region external to the sphere. To account for double precision, a
number Small 5 1.0e6 is used to avoid small numbers which may misrepresent a neutron as not lying on a surface
when it has been transported to a surface in its path. Note the distances DTC, DTS and Dleft which are used to update
the distance left Dleft from the distance to collision DTC when a distance to surface (DTS) is used to determine
whether the history continues or is ended in a final region. It is crucial to determine which of the bounding surfaces in
a region a neutron will be transported to. For this, the MATLAB function isreal is used to reject a complex root.
Similarly, a negative real root is not selected and thus only one real root for the parameter DTS is used to determine the
surface of interest. During a surface crossing, only the track-length estimator (TLE) is updated while in a final region
both the CE and TLE counters are updated. When a neutron escapes from the physical domain, that is, the sphere, the
counters Nescapes, WGT_escaping and TotalWGTlost are updated to later contribute to the leakage estimates. In a col-
lision occurring in the final region FREG, the weight is reduced in this non-analog simulation and the history is contin-
ued. The parameter key takes the value zero until a history is inside the domain, and one when it terminates due to an
escape from the domain.
Step 4: Calculate mean values
The mean values are obtained by dividing the estimates by the number of neutrons started to get the value for one
source neutron.
456 Nuclear Engineering

TABLE 12.1 Some of the variables used in the Monte Carlo program.

Quantity Variable
Avogadro’s number AvNo
Radius Rad_sphere
Nuclear data sigma_s, sigma, sigma_t
Distance left (remaining) Dleft
Distance to collision DTC
Distance to surface DTS
Number of histories N
Present region IREG
Final region FREG
Number of regions NREG
26
A small number (1 3 10 ) Small
Weight participating in a collision WtPartCol
Collision estimator flux CEflux
Flux variable updated at each interaction Flux
Number of scatterings Nscats
Number of absorptions Nabs
Number of collisions Ncols
Number of escapes Nescapes
Update total (statistical) weight of neutron TotalWGT
Weight lost by escape TotalWGTlost
Position x; y; z XPOS, YPOS, ZPOS
Direction cosines Uu, vv, ww
Particle weight WGT

Step 5: Calculate exact flux

The exact flux is calculated (line 118) with μ0 5 2=ð3AÞ, extrapolation distance d 5 0:71λtr , Σtr 5 Σs ð1 2 μ0 Þ, diffu-
pffiffiffiffiffiffiffiffiffiffiffiffi
sion coefficient D 5 1=ð3Σtr Þ, diffusion length L 5 D=Σa (Table 12.1).

12.2.2 MATLAB program for point source in a finite non-multiplying sphere

A MATLAB program is given below for the steps described in the previous section. It is recommended that these lines
are “copied and pasted” into a MATLAB.m file and executed.
Exercises in Monte Carlo simulation Chapter | 12 457
458 Nuclear Engineering
Exercises in Monte Carlo simulation Chapter | 12 459
460 Nuclear Engineering

FIGURE 12.5 Neutron flux (Monte Carlo and diffusion theory) in a

sphere.

12.2.3 Results
The MC TLE flux for N 5 1000 histories is shown in Fig. 12.5. For each region, the TLE flux is estimated to be , x .
where

1X N
,x.  wi si
N i51

It is also necessary to estimate the standard deviation of , x . , written as σ , x . . For a random variable x drawn
from a population with PDF f ðxÞ, the variance σ2 is
ð
σ2 5 ðx2, x . Þ2 f ðxÞdx

and subsequently, from the Central Limit Theorem, the variance of the estimated mean σ2, x . is given by
σ2
σ2, x . 5
N
An estimate of the variance of the estimated mean is thus
σ2 1 2
σ2, x . 5 5 x 2 x2
N N
Thus
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 2
σ,x. 5 x 2 x2
N
and another quantity, the relative standard deviation (RSD) R , x . of the estimated mean is
σ,x.
R,x. 5
,x.
Fig. 12.2 shows the neutron flux from MC with N 5 1000 histories sampled (green box) compared with the exact
diffusion theory solution (dashed line). In this case, the comparison is seen to be acceptable. It is good practice to com-
pare a simulation result with an “exact” analytical result but this is not always possible, as has been stated in earlier
chapters.
The output of the program given above is summarized below:
 Exercises in Monte Carlo simulation Chapter | 12 461

Consider now the estimate of the standard deviation of the estimated mean σ , x . and the relative estimated mean
R , x .  σ , x . = , x . shown in Table 8.5.
The additions to the code are then an estimate of the square of the estimator FLUXsq at every event in the region,
that is,
462 Nuclear Engineering

and the standard deviation (STD) σ , x . is obtained for each region:

from which the RSD R , x . is obtained for each region:

It is important to note that R , x . reduces an order of magnitude when N increases by two orders of magnitude in
accordance with the variation of the variance following the central limit theorem: σ2 B1=N. It can be concluded that
with a sample size N 5 100; 000 the flux have “converged” since the RSD is less than 0:5% for the “least precise” TLE
flux estimate in region 10, farthest from the source (Table 12.2).

12.3 Reflected assemblies

In Section 10.1, MC simulation of the Highly Enriched Uranium (HEU) Godiva sphere with radius 8.741 cm and mass
ðCATÞ
52.4254 kg respectively gave , keff . 5 0:994272 (0.0012) with the Monte Carlo N-Particle code (MCNP5) which
close to the Los Alamos value of keff 5 0:9976 (0.0011) (Whalen, Cardon, Uhle, & Hendricks, 1991).
An important exercise in nuclear engineering is to carry out simulations for reduction of fissile material to achieve
criticality by the use of a reflector such as beryllium, graphite or water, as was demonstrated in Section 5.3.2 for a
spherical solution of U-235 in water surrounded by a water reflector using two-group diffusion theory.
With MC simulations, much better designs in the sense of smaller size and efficient performance can be achieved.
In this section, simulations are demonstrated for reflected neutron systems with 93.5% U-235 surrounded by graph-
ite and 97.6% U-235 surrounded by water (Whalen et al., 1991). There are several references for critical systems
(Bowen & Busch, 2005; DeVolpi, 1982; Snood, Forster, & Parsons, 2003; Wagner, Sisolak, & McKinney, 1992) which
can be used to test simulation results.

TABLE 12.2 Mean , x . and relative standard deviation R , x . of the Monte Carlo TLE flux.

Region N 5 100 N 5 10; 000 N 5 100; 000

,x. R,x. ,x. R,x. ,x. R,x.

1 15.128 0.0000 14.853 0.0000 14.8950 0.0000
2 4.0574 0.0000 4.3861 0.0000 4.4056 0.0000
3 3.2868 0.0660 2.9959 0.0050 3.0262 0.0015
4 2.2036 0.0542 2.2855 0.0069 2.3241 0.0023
5 1.5420 0.0951 1.8662 0.0091 1.8830 0.0028
6 1.3609 0.0747 1.5436 0.0103 1.5610 0.0033
7 1.1447 0.0892 1.3125 0.0122 1.3298 0.0037
8 0.9845 0.1190 1.1435 0.0135 1.1466 0.0042
9 0.8948 0.1193 0.9924 0.0147 0.9814 0.0045
10 0.7337 0.1011 0.8526 0.0154 0.8582 0.0049
 Exercises in Monte Carlo simulation Chapter | 12 463

The search for a critical configuration begins with a bare sphere of radius 8.74 cm for which MC simulation (3000
histories for 50 cycles with 10 skip cycles) gives keff 5 0:98513 0.00180 listed in Table 12.3; this is not critical due to
the slightly decreased density compared with that for the Godiva sphere in Section 10.1. The mass of the HEU core is
52.0159 kg. To reduce the core size, a surrounding graphite layer of thickness 5.0 cm gives keff 5 01:10789 0.00170
which is encouraging as it offers the possibility of reducing the core. In the next simulation, the core radius is decreased
to 8.50 cm keeping the reflector thickness the same. The result shows that the system is supercritical with a reduction
of about 4.2 kg in the core. A further reduction to a core radius of 8 cm is found to be slightly supercritical and another
simulation is carried out with a core radius of 7.5 cm (HEU mass reduced by about 19 kg which is a B37% mass reduc-
tion) with a 5 cm thick graphite reflector which gives keff 5 1:00058 0.00160. Decreasing the core radius to 7.20 cm
with the same reflector thickness results in a sub-critical system.
Based on the above results, it is possible to conclude on an approximate critical design that has a core radius in the
range 7.207.50 cm with a 5.0 cm graphite reflector. Compare this with the experimental result of 7.39840 cm with a
5.1 cm graphite reflector (Whalen et al., 1991) for which keff 5 0:9981 0.0010 for a sphere of radius 7.39840 cm.
Another design exercise is the selection of a “best” reflector. Table 12.4 shows MC simulation results for the critical
HEU sphere of radius 7.39840 cm (31.5511 kg) described above with three reflectors-graphite, beryllium and light
water showing beryllium to be the best.
Thus maximum core savings are possible with beryllium and the least effective of the three is graphite.
The effectiveness of a reflector is dependent on its atomic mass number and scattering and absorption cross-
sections. Hydrogen has the lowest atomic number but it also has a high absorption cross-section and thus beryllium
emerges as the most effective out of the three materials considered.
The critical masses for plutonium (density 19.5 g/cc α and 15.9 g/cc δ phase) are 11 and 15 kg respectively; for ura-
nium (U233, U235 density 18.9 g/cc) 16g and 48 kg respectively. For the Godiva and Jezebel criticality experiments, see

TABLE 12.3 Random search for criticality of a graphite-reflected spherical core.

Core Reflector Total mass keff

Radius (cm) Mass (kg) Thickness (cm) Mass (kg) (kg)

8.74 52.0159 Nil nil 52.0159 0.98513 0.00180
8.74 52.0159 5.0 9.79459 61.8105 1.10789 0.00172
8.50 47.8474 5.0 12.9150 60.7624 1.10315 0.00168
8.00 39.8907 5.0 11.787 51.6777 1.04749 0.00166
7.50 32.8689 5.0 10.7115 43.5804 1.00058 0.00171
7.20 29.0803 5.0 10.0914 39.1717 0.96719 0.00195
7.20 29.0803 6.0 13.4779 42.5583 0.97962 0.00203
7.20 29.0803 7.0 17.4185 46.4988 0.99462 0.00197

TABLE 12.4 System multiplication keff for a reflected core.

Reflector keff

Material Density (g/cm) Thickness (cm) Mass (kg)

Graphite 1.67 5.0 10.8246 0.9920 0.00229
Beryllium 1.85 5.0 11.9913 1.09337 0.00221
Light Water 1.0 5.0 6.48179 1.01864 0.00219
464 Nuclear Engineering

TABLE 12.5 Critical spherical systems reflected with 10 cm U.

Material Density (g/cc) Radius (cm) Core mass (kg) Total Mass (kg)
U233 18.9 4.2 5.7 227
235
U 18.9 5.8 15.7 316
Pu 239
α 15.9 3.8 4.5 209
239
Pu δ 19.5 4.7 7.0 250

problem no. 4 at the end of this chapter. The reduction in critical masses with a 10 cm natural uranium reflector is
shown in Table 12.5 (DeVolpi, 1982).
The search for criticality is an exercise in which the critical dimensions are calculated for a specified core composi-
tion or the critical core composition is determined for a specified size.
One drawback in MC simulations has been the requirement of performing simulations for each separate design; this
has been addressed to some extent by the perturbation algorithms which give the design sensitivities due to independent
parametric changes. With this feature, the search for a critical design can be simplified.

Exercise 12.9: With the one-group effective multiplication given by

νΣf
keff 5 (12.14)
DB2 1 Σa

1. How would a decrease in the core radius affect keff ?

2. How would the change in keff be estimated due to a 10% increase in the atomic number density of U-238 in
Godiva?

12.4 Reactor core modeling

In this section, MCNP5B is used to model a Pressurized Water Reactor (PWR) fuel assembly starting from an input file
by Hideki Matsumoto. Changes were made to obtain the assembly which this neutronic analysis is meant for.

12.4.1 Input file

Details of the input file can be found in the MCNP Documentation (Werner, 2017) (Volume I, Chapters 3 and 4). Here,
parts of the input with reference to the PWR fuel assembly only will be described. The same geometry can be described
(MCNP Vol II, Example 8, pp. 437) using LIKE m BUT and TRCL cards. The procedure is to define universes, fill a
lattice cell, and then define the bounding window in which as many lattices will fit in as the bounding region would
permit.

12.4.1.1 Defining the bounding window

The window, in which the lattice cell (universe 1) is to be filled, is defined as cell 1 with a void bounded by the sur-
faces 21 and 22 in the x direction (210.7525, 10.7525, length 21.5050 cm), 23 and 24 in the y direction (same as in the
x direction, i.e., 21.5050 cm), and surfaces 5 and 6 in the z direction (length 1 cm). The input line is:
1 0 21 222 23 24 5 6 fill 5 1

12.4.1.2 Defining the universe(s)

The universes are defined which will fill the lattice
Here, we have 3 universes, u 5 1, u 5 2, and u 5 4
Universe 1 will be defined in the following section.
Universe 2 has cells 3, 4, and 5 with materials 1 (UO2), 2 (zirconium), and 3 (water)
 Exercises in Monte Carlo simulation Chapter | 12 465

Exercise 12.10: Universes 2 and 4 are described radially. Where in the input would they be bounded axially?

12.4.1.3 Defining the lattice

Lattice 1 is defined, as cell 2 filled with universe 1 containing material 3 (water) and filled by itself with 19 3 19 3 1
elements in the x-y-z directions. In the x direction, it is bounded by the surfaces 11 and 12 and is of length 1.2650 cm;
in the y direction by surfaces 13 and 14 also of length 1.2650 cm, and in the z direction by surfaces 5 and 6 and is of
length 1 cm.
The input lines, describing cell 2 as the lattice 1 and filled with universe 1 defined itself with universes 1, 2, and 4,
(with 19 3 19 elements) are:

12.4.2 Surrounding cells

Cell 10 surrounds cell 1 (PWR fuel assembly) and is filled with material 3 (water), while cell 9 is the outside universe
surrounding the entire assembly and water; defined as everything in the universe that excludes cells 1 and 10.
The input lines are:
466 Nuclear Engineering

12.4.2.1 Surface cards

12.4.3 Source description

The source description in MCNP can be done in one of three ways (MCNP Vol II, pp. 352) viz and explicit value, a
distribution, or a distribution of a dependent variable. Here, the SDEF card is used to describe source location in a cell
whose spatial distribution is represented by d1, and energy (erg) represented by Distribution 2.
For Distribution 1, the source information card SI1 is used with the l option indicating discrete source variable values.
The source cell path is described as the location of the cell from the highest level to the lowest level. Here the lattice has 8
cells which have material 1 (fuel) so we need to locate cell 3 wherever it needs to be specified. So, since cell 3 is inside cell
2 which is inside cell 1, the “address” of this cell is through cell 1, and since cell 2 is a lattice, each position needs a location
identifier such as (21 21 0) indicating the cell in the lowest row and most left column of the 3 3 3 lattice.
The energy Distribution 3 in MCNP represents the Watt fission spectrum described in Section 12.1.
The input parameters for source definition cards (SDEF), source information (SI) and source probability (SP) define
the phase space parameters of the source. In the following lines, SDEF defines a source located in CEL 5 D1 with a
space location defined by a probability Distribution 1, and with an energy distribution given by Distribution 2; SI is the
source information card which gives the location of the source as Distribution 1 in the repeated geometry structure.
There are eight cells in which the source appears and it will have equal probability as given in the SP1 card. Similarly,
SP2 samples from the Watt fission spectrum with default parameters as stated above.

Exercise 12.11: Is the cell sampling of the source adequate or can it be improved? What will be the effect of the
cel 5 d1 specification on the quality of tallies?
 Exercises in Monte Carlo simulation Chapter | 12 467

12.4.4 Plotting the geometry

With the above surfaces, let’s first calculate the length of the lattice cell which is bounded by surfaces 11 and 12 in the
x direction, of length 0.6325 2 (20.6325) 5 1.2650 cm. Cell 2 has 19 such-sized units in the x direction since it goes
from 29:9 in the x direction, and 19 units in the y direction. They are numbered as

Thus, all 19 3 19 cells are of length 24.035 cm and extend from 212.0175 to 112.0175, but the bounding surfaces
on cell 1 are 210.7525 to 110.7525. Similarly 18 3 18 cells are of length 22.77 cm extending from 211.385 to
111.385, and 17 3 17 cells are of length 24.035 cm extending from 212.0175 to 112.0175. Thus, the above input
should show only 17 3 17 of the 19 3 19 defined cells.
The geometry is obtained by running mcnp5 in the ip (input and plot) mode; Fig. 12.6 below.
To obtain 3 3 3 cells from the above, the bounding window surfaces (21, 22, 23, 24) would be changed and the total
length would now be: 21.8975 to 11.8975 5 3.795 5 3 3 2 3 0.6325.

FIGURE 12.6 A typical PWR fuel assembly.

(ZK1pwrA).
468 Nuclear Engineering

The 3 3 3 assembly is shown in Fig. 12.7 plotted with the commands

with the input file ZKpwr3 3 3 listed below the figure.

Exercises in Monte Carlo simulation Chapter | 12 469
470 Nuclear Engineering

Exercise 12.12: How would you modify the water density to account for the temperature changes in the reactor from
cold to hot states?

12.4.5 Tally cards

MCNP gives information “demanded” by the user; thus tallies need to be specified. In most cases, the tallies required
are volume averaged flux, current, surface flux, energy deposition, and neutron heating. These are obtained from stan-
dard MCNP tallies F1, F2, F4, F5, F6 and associate energy values *F1, *F2, *F4, *F4, *F5, and *F6, as described
below.
The flux tally is TALLY F14:N in the cells defined below:

Tallies available (for neutrons) include F1:N (particle current integrated over a surface, that is, no. of particles), F2:
N (surface flux particles cm22), F4:N (flux at point or ring detector particles cm22), F6:n (energy deposition averaged
over a cell MeV/g), F7:N (fission energy deposition averaged over a cell MeV/g). Together with these tallies, *F1 gives
energy (MeV) integrated over a surface, *F2, *F4 and *F5 are energy quantities (MeV cm22), while *F6 and *F7 give
jerks/g (1 jerk  1028 sÞ. In MCNP, adding 10 to a tally gives the same tally quantity; for example, F4, F14, F24, F34,
. . . are all flux quantities. Each tally can be obtained in bins of time, energy and direction.

12.4.6 Reaction rates

Ð
Any quantity of the form C φðEÞRm ðEÞdE, where φðEÞ is the energy-dependent fluence (units particles cm22) and the
(microscopic) reaction of interest Rm for material m is taken from the cross-section file. The material m needs to be
specified on a separate Mm card and is not required to be present in any cell defined in the geometry. The reaction num-
bers R are taken from the Evaluated Nuclear Data File (ENDF/B) manual (where it is called the material type (MT)
number), some commonly used MT numbers are
 Exercises in Monte Carlo simulation Chapter | 12 471

FIGURE 12.7 A 3 3 3 assembly (scales shown

are centimeters).

For MT18 the condition is: if and only if MT 5 18 is used for fission in the original evaluation
Total fission cross-section; equal to MT 5 18 if MT 5 18 exists (otherwise sum of MTs 19,20,21 and 38).

12.4.7 Plotting tallies

For plotting tallies and cross-sections, see MCNP5 Vol II Appendix B.III. The command mcnp5 z options invokes
MCPLOT. Fig. 12.8 was plotted with the commands kcode 1 and file which gave a postscript file plotm.ps which was
then converted to the pdf file shown below. The average of the collision, absorption, and track-length estimates is
keff 5 1:44256ð0:00062Þ. The standard deviation 0.00062, is good; it can be further improved as described in Chapter 4.
Fig. 12.9 shows the Figure of Merit for the average CE/absorption estimator (AE)/TLE of keff , which appears to
have reasonably converged beyond 400 cycles.
472 Nuclear Engineering

FIGURE 12.8 Collision estimate of keff versus

Cycle number.

FIGURE 12.9 FOM of average keff versus

Cycle number.

The energy-integrated neutron flux, plotted in Fig. 12.10 as a surface and in Fig. 12.11 on a xy plane, show a dip at
the center of the 3 3 3 assembly.
These were obtained by using the fine mesh (FMESH) input “card” (Vol II, pp. 3112) which gives an output in
file MESHTAL processed in MATLAB.
 Exercises in Monte Carlo simulation Chapter | 12 473

FIGURE 12.10 Neutron flux in a 3 3 3 PWR cell.

FIGURE 12.11 Neutron flux in a 3 3 3 PWR cell (xy plane).

12.5 Radiation safety and shielding

The basics of radiation and radiation shielding to protect people working in radiation environments, were highlighted in
Chapter 1. The design of such shields is readily carried out by MC codes since detailed interaction models are used by
sampling histories in a considered sample.
Nuclear criticality safety analysis, described in Chapter 10, is required at various stages in the fuel cycle from pro-
cessing of uranium ore to the handling of process gases such as uranium hexafluoride (UF6) as it is enriched, for exam-
ple in a centrifuge plant Section 10.3.2, to its final stages when it takes the shape of enriched fissile fuel with reactor
grade.
For analysis of nuclear criticality safety, simple cylinders containing solutions of uranium and plutonium to repeated
structures in hexahedral (square) and hexagonal (triangular) lattices are easily modeled in MCNP.
In this section, MCNP simulations are carried out for a 20 cm thick slab of water in the x-direction, and 10 cm thick
in the y, z-directions and with front, back, top, and bottom surfaces taken as reflecting surfaces. A point anisotropic neu-
tron source of unit strength and energy 1 MeV is incident on the left face of the slab.
474 Nuclear Engineering

Exercise 12.13: A person is standing in front of a Cf252 source which has an activity of 532 Curies/g with a neutron
dose B22 Sv m2/g/h and a gamma dose 1.6 Sv m2/g/h.

If the maximum permissible dose allowed is 5 rem/y (0.05 Sv/y), how much would have been accumulated more
than the annual limit in 1 hour from 1 g of material at a distance of 1 m? [Answer:B400 times]
Consider a 1 MeV neutron source for which a water and an iron (Fe) slab are compared, each of thickness 20 cm.
Results from the MCNP simulation, listed below, show that the total neutron dose across the iron shield
7.68320 3 10213 Sv/(n/s) is about 100 times higher while the total photon dose 4.74296 3 10216 Sv/(n/s) is about 3%
less than that from a comparable water shield. MCNP mixed mode (N/P) simulations for 20 cm 3 10 cm 3 10 cm slabs
of water (2 kg) and iron (15.720 kg) are given in Table 12.6.
So, it can be concluded that water is a much more effective neutron shield while iron is a much better gamma shield.
The total dose for both is given below
Material Neutron (Sv/(n-s)) Gamma (Sv/(n-s)) Total (Sv/(n-s))
Water 7.82087E15 0.0477 1.74176E14 0.0124 2.5239E14
Iron 7.68320E13 0.0100 4.74296E16 0.0920 7.6879E13

The total dose across the iron shield is thus about 30 times more. To estimate the effect of radiation, neutron and
gamma radiation must be considered separately due to the different nature of both radiations.
For an annual limit of 0.05 Sv/y (B1.6 nSv/s), the maximum permissible neutron source would thus be B2000 n/s.
Comparing this with the neutron emission of B1012 n/g from Cf252, even a microgram would be far too dangerous for
short durations

Exercise 12.14: Compare the half value layer thickness (HVL), at 1 MeV for iron and water to estimate the attenuation
from 20 cm for both shields.

A detailed simulation shows (F1 current tally) gives a leakage of 5.85004E03 gammas/(n, g) gamma compared to
0.1721 from water. The (n, γ) reactions are about ten times greater in water than in iron, thus requiring an effective
iron shield to stop secondary gammas.
Relaxation length (width of material for intensity reduction to 1/e B37%) and HVL for 1 MeV photons is given in
Table 12.7.

12.6 Perturbation calculations

One of the weaknesses of MC methods has been considered to be their inability to estimate the effect of small changes
in independent parameters such as enrichment, material density or material composition (Section 6.6). This is due to the
magnitude of the perturbation in the independent parameter which may be of the order of uncertainty in a stochastic
estimate and hence the estimated change may be masked in the error itself.
As discussed in a previous chapter, perturbation computations can be carried out in schemes such as derivative sam-
pling, correlated sampling or adjoint weighting in a Lagrangian variational formulation. This section considers the per-
turbation capability of MC simulation to estimate such design changes.
In Fig. 12.12 MCNP Godiva simulations (1000 3 130 with 30 skip cycles) show the increase in keff and decrease in
J (particles crossing surface) as the relative density ρr  ρ=ρo increases (reference density ρo 5 18:75g=cm3 ). Each run
is an independent estimate.
Consider now an elementary “one-group” diffusion estimate of the change in keff due to a material density change.
The one-group diffusion equation gives
 Exercises in Monte Carlo simulation Chapter | 12 475

TABLE 12.6 Current, flux and dose for water and iron slabs.

Tally*/Mat/ Surface/ Description 05 eV 5 eV14 MeV Total

reaction cell
F1:N 2 Current (n) 2.31323E02 3.26898E03 2.64013E02
0.0146 0.0552 0.0144
4.18566E05 2.34209E01 2.34251E01
0.2921 0.0056 0.0056
F4:N 1 Cell flux (n) 1.14378E02 5.81316E03 1.72509E02
0.0033 0.0020 0.0026
6.73521E06 1.67608E02 1.67676E02
0.1129 0.0037 0.0037
F2:N 2 Dose (n) Sv 3.57166E15 4.24921E15 7.82087E15
0.0190 0.0864 0.0477
6.64665E18 7.68314E13 7.68320E13
0.3187 0.0100 0.0100
FM14:N 1 21 90 102 3.15975E04 1.12125E06 3.17097E04
0.0034 0.0034 0.0034
21 190 102 1.68707E07 9.18275E06 9.35145E06
0.1265 0.0094 0.0101
FM24 1 1 90 102 3.15003E03 1.11780E05 3.16121E03
0.0034 0.0034 0.0034
1 190 102 1.99052E06 1.08344E04 1.10334E04
0.1265 0.0094 0.0101
FM34 1 1 92 102 2.10062E03 7.45418E06 2.10808E03
0.0034 0.0034 0.0034
— — — —
F11:P 2 Current (p) 0 1.72100E01 1.72100E01
0.0084 0.0084
0 5.85004E03 5.85004E03
0.0420 0.0420
F12:P 2 Dose (p), Sv 0 1.74176E14 1.74176E14
0.0124 0.0124
0 4.74296E16 4.74296E16
0.0920 0.0920

Upper and lower rows for each tally are for water and iron slabs respectively.

TABLE 12.7 Half value layer thickness for some shielding materials.

Material Density ðg=cm3 Þ μ=ρ ðcm2 =gÞ 1=μρ (cm) HVT (cm)
Air 1.225 3 1023 0.0636 1.2835 3 104 1.2835 3 104
Water 1 0.0707 14.1443 9.8020
Concrete 2.4 0.0637 6.5411 4.5330
Iron 7.874 0.0599 2.1202 1.4693
Lead 11.34 0.0680 1.2968 0.8987
476 Nuclear Engineering

FIGURE 12.12 Effect of perturbation in material density on keff .

νΣf
keff 5 (12.15)
DB2 1 Σa
which can be used to estimate the perturbations, with respect to material density ρ for example, from the derivatives in
a Taylor series of the form

 @keff 1 @2 keff
keff 5 keff 1 δρ 1 ðδρÞ2 1 . . . (12.16)
@ρ 2! @ρ2
The first and second derivatives are obtained as:
1 @keff 2β 1
5 (12.17)
keff @ρ ρ β 1 γρ2
and
1 @2 keff 2β β 2 3γρ2
5 2 (12.18)
keff @ρ 2 ρ ðβ1γρ2 Þ2
B2

2
where β  3σ tr
and γ  NMav σa .For Godiva, consider a pure U235 sphere of radius 8.37 cm of density
ρ 5 18:75 g=cm3 (m 5 46.054 kg) and use the fast spectrum data from Wirtz (1982): νσf 5 5:297 atom b21,
σa 5 2:844 atom b21 and σa 5 8:246 atom b21. The first and second derivatives are: 0:0494 and 20:0030 respectively.
The above values are used to estimate keff for full re-runs (solid curve) compared with second order perturbation
estimate from derivative sampling with MCNP, and with one-group exact analysis. Relative errors are shown for both
sets of (MCNP) estimates.
Fig. 12.13 shows estimates for keff for full re-runs (solid curve) compared with second order perturbation estimate
from derivative sampling and the elementary model. Relative errors are shown for both sets of estimates. It can be seen
that the predicted derivative sampling estimates are good for up to a 60% decrease, and 40% increase, in the relative
density. The lack of symmetry is due to the neutron histories simulated; as the density increases the mean free path
decreases and more collisions take place, thus the sampling size has to be appropriately modified for statistical accuracy
of an estimate. The elementary diffusion estimates have a larger error but indicate the trend very well.
Similar estimates, with similar conclusions, are shown for the surface current in Fig. 12.14.

12.7 MCNP geometry plotting in core neutronics

One of the strengths of MC simulations, particularly of MCNP, is the ability to represent realistic geometry of the con-
figurations considered for systems such as the 4S reactor, the micronuclear reactor and the International Thermonuclear
Reactor (ITER) tokamak.
In Fig. 12.15 a 4S-type configuration is shown; this has hexagonal cells with fuel elements and a central control rod.
 Exercises in Monte Carlo simulation Chapter | 12 477

FIGURE 12.13 Perturbation estimates using

one-group derivative sampling.

FIGURE 12.14 Effect of perturbation in mate-

rial density on current J.

Exercise 12.15: The configuration given in Fig. 12.15 is of the 4S Gen-IV type reactor with metallic uranium fuel, sodium
coolant and a beryllium reflector. Model the geometry with MCNP and carry out a simulation for the keff and flux tallies.

Consider a reactor core and reflector as shown below red being the fuel rods, blue the heat pipes, yellow the matrix,
green the reflector and the dark strips in the six surrounding control drums representing the absorber material.

Exercise 12.16: Model the absorber regions in the control rod to face toward or away from the core.

This Exercise is thus to model the control drums with the absorbers spanning an angle 45 degrees to face the core or
to be on the “opposite” side.
The first step is to generate the equations of the planes for the absorber regions. These are shown below; equations
are mx 2 y 1 c 5 0 in the MCNP format Ax 1 By 1 Cz 2 D 5 0. The drums are labeled P to U with centers at (x, y). The
478 Nuclear Engineering

FIGURE 12.15 Reactor 4S-type core arrange-

ment. (MCNP inp 5 ZK4Sfil).

TABLE 12.8 Surface coefficients.

x y ϕ m c surface
P Top Right a b 30 2 θ 0.131652497587396 10.881235923901318 40
30 1 θ 1.303225372841206 217.721544252545193 41
Q Top 0 28.1908 90 2 θ 2.414213562373095 28.190799999999999 48
90 1 θ 22.414213562373095 28.190799999999999 49
R Top Left -a b 150 2 θ 21.303225372841206 217.721544252545193 45
150 1 θ 20.131652497587396 10.881235923901318 44
S Bottom Left 2a 2b 30 1 θ 1.303225372841206 17.721544252545193 47
30 2 θ 0.131652497587396 210.881235923901318 46
T Bottom 0 2 28.1908 90 1 θ 22.414213562373095 228.190799999999999 50
90 2 θ 2.414213562373095 228.190799999999999 51
U Bottom Right a 2b 150 1 θ 20.131652497587396 210.881235923901318 43
150 2 θ 21.303225372841206 17.721544252545193 42

a 5 24.4140, b 5 28.1908, θ 5 (45/2) degrees, numbers obtained with format “long” in MATLAB.

angles ϕ are relative to the horizontal axis, and the slope m and intercept c are calculated for each plane. The surface
numbers are given in the “right-most column” below (Table 12.8).
The second step is to write the above in MCNP input format. Here, material 8 (density 3.01 g/cc) and material 9
(density 2.52 g/cc) are the reflector and absorber material, respectively. Cells are numbered 404436.
 Exercises in Monte Carlo simulation Chapter | 12 479

The third step is to run MCNP in the ip mode with the commands:
480 Nuclear Engineering

FIGURE 12.16 Absorbers on control drums in

core barrel.

mcnp5 ip inp 5 inpfile plot. pz 0 plot . extent 35 35 plot. label 0 0 plot. scales 1plot. color on plot. color by
mat plot. shade 1 red plot. shade 4 blue plot. shade 7 yellow plot. shade 8 green plot. shade 9 black
plot . file
This gives a plotm.ps (or appropriately named file) which must be converted to pdf format.
Figs. 12.1612.18 show the geometry with absorbers as required for the simulation.

Problems
1. Carry out a MC simulation for a graphite sphere of specified radius to determine the escape probability and compare
with the diffusion theory estimate.
2. It was shown that the neutron flux in a sphere of radius R with a point isotropic source at r 5 0 emitting S neutrons
s21 is given by
S sinhκðR 1 d 2 rÞ
φðr Þ 5
R 1 d 
4πDsinh L r
Carry out a MC simulation and determine the shape of the neutron flux and
the leakage probability from the surface of a beryllium sphere of radius 2 m.f.p. with a
1 Ci 226Ra-Be point source at its center emitting 107 neutrons s21.
3. Carry out a simulation to determine the albedo of a graphite slab of thickness 3 m.f.p and compare with previous
estimates from diffusion theory and transport theory.
4. Compare MC simulation values for the keff values of some fast critical assemblies (Bell and Glasstone, 1979) given
below (Table 12.9).
Exercises in Monte Carlo simulation Chapter | 12 481

FIGURE 12.17 Absorbers on control drums fac-

ing toward the core.

FIGURE 12.18 Absorbers on control drums fac-

ing away from core.
482 Nuclear Engineering

TABLE 12.9 Criticality estimates.

Assembly Core radius (cm) Calculated k

Core Reflector
Uranium-233 None 5.965 1.0115
Uranium-235 (Godiva) None 8.710 0.9912
Plutonium-239 (Jezebel) None 6.285 1.0039
Uranium-235 (16.7%) 7.6 cm uranium 20.32 0.9893
Uranium-235 8.9 cm uranium 6.391 0.9939
Uranium-235 1.6 cm thorium 7.80 0.9905
Uranium-235 (Topsy) Thick uranium 6.045 0.9907

Conclusions
This chapter gives the necessary skills to apply the concepts learnt in the fundamentals covered in the earlier chapters
of this book. Each section forms part of the underlying foundations of a Monte Carlo simulation. Though random sam-
pling is done in a “black box” mode within a code, and quantities such as the average source energy and sampled ener-
gies of the first 50 particles appears in the output file, it gives confidence to the user to appreciate how these values
were generated inside the code. The section on a fixed-source MC simulation was intended to show the easy steps of
the simulation process; in eigenvalue problems, a source distribution is written at the end of each generation simulated
and subsequently used in the next generation. In many cases, a large number of generations and source particles are
specified in the input file assuming that convergence will take place. The statistics to determine stationarity are gener-
ally not well understood by a beginner. Thus, convergence is illustrated in these exercises. Other areas include geometry
modeling, which typically takes a substantial part of the effort in setting up an input file, and the perturbation feature
which is a very useful capability of MC codes such as MCNP5 and can help in carrying out design sensitivity studies
leading to optimization. There is much more versatility in present MC codes to handle a far wider range of problems in
terms of the number of particles that can be simulated, the detailed modeling and interfaces than earlier versions. The
basics covered here are at the core foundation level.

Nomenclature
English lower case
d extrapolation distance
f probability distribution function (PDF)
keff effective multiplication
kN infinite multiplication
s distance
w statistical weight

English upper case

B buckling
D diffusion coefficient
F cumulative distributive function (CDF)
J neutron current
L diffusion length
Ni atomic density of the ith nuclide
Rc critical radius
R~ extrapolated radius
R relative standard deviation
 Exercises in Monte Carlo simulation Chapter | 12 483

S source

Greek lower case

λtr transport mean free path
μ cosine of angle of scattering
μ mean value
μ0 average cosine of scattering angle
ν0 eigenvalue
ν number of neutrons produced per fission
ξ random number
ρ density
σ standard deviation
σa microscopic absorption cross-section
σf microscopic fission cross-section
σr microscopic removal cross-section
σs microscopic scattering cross-section
σtr microscopic transport cross-section
σt microscopic total cross-section
τ neutron age
ϕ azimuthal angle
φ neutron (photon) flux
χ fission spectrum

Greek upper case

P
macroscopic absorption cross-section
Pa
macroscopic fission cross-section
Pf
macroscopic removal cross-section
Pr
macroscopic scattering cross-section
Ps
macroscopic transport cross-section
Ptr
t macroscopic total cross-section

Abbreviations
HVL half value layer
MC Monte Carlo
m.f.p mean free path
Sv sievert
TLE track-length estimator

References
Bell, G. I., & Glasstone, S. (1979). Nuclear reactor theory. New York: Robert E. Krieger Publishing Co., Inc.
Bowen, D. G., & Busch, R. D. (2005). Hand calculational methods for criticality safety—A primer. 2005 NCSD Topical Meeting (American Nuclear
Society Nuclear Criticality Safety Division).
Brown, F. B. Fundamentals of Monte Carlo particle transport.
Werner, C. ed. (2017). MCNP (R) user’s manual code version 6.2.
Cullen, D. E. (2004). Sampling ENDL Watt fission spectra. Energy.
DeVolpi, A. (1982). Denaturing fissile materials. Progress in Nuclear Energy.
Everett, C., Turner, G., & Cashwell, E. (1973). A method of sampling certain probability densities without inversion of their distribution functions.
Froehner, F. H., & Spencer, R. R. (1980). Method for sampling from fission neutron energy spectra.
Monte Carlo Team. (2005). MCNP—A general Monte Carlo N-particle transport code, Version 5. Los Alamos National Laboratory.
Sood, A., Forster, R. A., & Parsons, D. K. (2003). Analytical benchmark test set for criticality code verification. Progress in Nuclear Energy.
Wagner, J. C., Sisolak, J. E., & McKinney, G. (1992). MCNP: Criticality safety benchmark problems. LA-12415. New Mexico, United States.
Whalen, D. J., Cardon, D. A., Uhle, J. L., & Hendricks, J. S. (1991). MCNP: Neutron benchmark problems. LA-12212 DE92 004710.
Wirtz, K. (1982). Lectures on fast reactors. American Nuclear Society.
484 Nuclear Engineering

Annex A: MATLAB Program CH12_NormalSampling.m

Exercises in Monte Carlo simulation Chapter | 12 485
486 Nuclear Engineering

Annex B MATLAB Program CH12_Watt Sampling.m

Exercises in Monte Carlo simulation Chapter | 12 487
488 Nuclear Engineering
Chapter 13

Optimization in nuclear systems

13.1 Introduction
The canvas of optimization in nuclear engineering is vast, diverse, and interconnected. It is by no means possible for an
author to convey more than a glimpse to a reader in a book chapter, but it is hoped that the few descriptions and exam-
ples given here will give some insight into the work over the last few years and the emerging methods.
In nuclear fission and fusion system design optimization, the overall objective is to design the best systems which
meet the requirements. This is more aptly described as multi-objective, multi-module, multi-constraints optimization in
which the modules are inter-dependent (Rozin, Rubin, & Soboĺ, 1990; Stewart, Palmer, & DuPont, 2021). Thus, it is a
fairly big challenge from the perspective of mathematical modeling and computing (Sections 9.1 and 9.6). The domains
covered in it include fission and fusion neutronics, fusion reactor blanket design, thermal hydraulics (TH), control sys-
tems, and radiation shielding and since the constraints cannot be expressed in algebraic forms, the deterministic optimi-
zation techniques of Section 9.2.2 (differentiable functions in the Lagrangian) and Section 9.2.3 (Euler-Lagrange
equation) cannot be applied to obtain analytical derivative expressions. The optimal formulation in the form of
Pontryagin’s Maximum (or Minimum) Principle (Section 9.2.5) can be applied as shown in this chapter. The
suitable models are Dynamic Programming (Section 9.4) but more suitable are the stochastic meta-heuristic methods
(Section 9.5) such as Genetic Algorithms (Section 9.5.1) and Particle Swarm Optimization (Section 9.5.2).
The analysis and particularly the engineering design results will be the focus for optimization calculations in reactor
core design, fusion neutronics, radiation shielding, fuel loading pattern (LP), a thermal neutron activation analysis
(TNAA)-based explosives detection system.

13.2 Reactor core design optimization

Important parameters related to core design are materials, dimensions and operating parameters. In a preliminary analy-
sis, the overall design is achieved by carrying out a neutronic simulation coupled with thermal hydraulics (TH); this
gives the basic initial design.
In a neutronic simulation depicted in Fig. 13.1, the materials and dimensions are input on the basis of some past
experience and knowledge of a similar system to obtain the overall system multiplication keff . A number of runs are typ-
ically required to adjust the materials and dimensions to get keff 5 1. This would be a starting design for a coupled
neutronic-TH simulation leading to more detailed and elaborate design parameters in a full optimization calculation.
In a preliminary model, the materials and geometry of the fuel rods, fuel assemblies and the whole core are specified
as shown in Fig. 13.1 on the basis of a comparison with other similar designs and a neutronics solver, such as Monte
Carlo N-Particle (MCNP), can be used to estimate keff and vary the parameters until a feasible design is obtained.
An example of a first-step neutronic calculation of a homogeneous model of the Korean KORI-1 nuclear reactor
(Kim, 2019; Lee, 1973) can be carried out with a basic two-group model with cylindrical geometry in a neutronic code
such as ANISN or DOT (Section 8.1).
A simulation carried out with MCNP5, with the input listed below, considers a homogeneous water-fuel mixture in
a cylinder of radius 122.6 cm and height 365.8 cm.
MCNP input file: KORI-1 homogeneous reactor model
1 1 0.079394
1 3
4
2 1 0.079393 1
2 3
4
3 0 2:
3:4
1 cz 86.69
2 cz 122.6
Nuclear Engineering. DOI: https://doi.org/10.1016/B978-0-323-90618-0.00013-2
© 2022 Elsevier Inc. All rights reserved. 489
490 Nuclear Engineering

FIGURE 13.1 Basic geometry and materials feasibility for a neutronic

calculation.

TABLE 13.1 KORI-1 reactor homogeneous model with MCNP5.

Enrichment (%) k eff

1.200 1.03204 (0.00109)
1.100 1.00168 (0.00109)
1.098 1.00446 (0.00105)
1.090 0.99965 (0.00106)
1.050 0.98414 (0.00110)
1.000 0.96907 (0.00107)
0.900 0.92754 (0.00098)

3 pz
182.9
4 pz 182.9
mode n
imp:n 1 1 0
c ------- enrich 5 1.1% total 0.079394
m1 92235.66c 7.6934e-05 92238.66c 6.9171e-03 8016.66c 3.3460e-02 1001.66c 3.894e-2
kcode 500 1.0 10 500
ksrc 0 0 0
A number of such simulations have been carried out for fuel enrichment in the range 0.9%
1.2% and given the
results shown in Table 13.1. The bare system with atomic densities given is critical at an enrichment of 1.09%. This
matches the one-zone criticality result from the determinant (Section 5.3.2).
The atomic densities and masses of uranium are listed in Table 13.2
For criticality, the atomic densities and mass ratio are: NH 5 3:8940 3 1022 , NO 5 3:236 3 1022 atoms cm23,
N5 =NU 5 0:01099, M5 =NU 5 0:01086. The H/U235 ratio for criticality is 506.15 with 1.1% enrichment,
keff 5 0:99974ð0:00052Þ with a total uranium mass B47.741 t out of which the fissile fuel U-235 is B519 kg.

Exercise 13.1: Two-group criticality for a bare cylindrical reactor.

For a bare homogeneous nuclear reactor core (Lee, 1973) with the two-group data in Table 13.3, solve the criticality
determinant to find the critical dimensions.
The atomic densities are:
N U 5 N 235 1 N 238 5 6:994 3 1021 atoms cm23
U
NOxygen 5 number of oxygen atoms in uranium 5 1:399 3 1022 atoms cm23
Nw 5 1:947 3 1022 molecules cm23
 Optimization in nuclear systems Chapter | 13 491

TABLE 13.2 Atomic density and mass of uranium in KORI-1 homogenous model.

Nuclide N (atoms/cm3) N (atoms) Mass (kg)

U-235 7:6934 3 10 19
1:3289 3 10 27 518.5958
U-238 6:9171 3 1021 1:1948 3 1029 47223.19
U 6:9940 3 1021 1:2081 3 1029 47741.79

TABLE 13.3 Two-group diffusion data.

Data Group 1 Group 2

D (cm) 1.367 0.2294
σ235
a (10224 cm2) — 600.62
Σs (cm21) 0.03922 —
Σothers
a (cm21) — 0.028058

At the next level, a coupled neutronic-TH analysis, as depicted in Fig. 13.2, is carried out to obtain key parameters
of thermal and flow characteristics. Such analysis typically gives the following set of design parameters:
1. Core dimensions (e.g., cylindrical radius and height)
2. Fuel road diameter ðDÞ and length (L)
3. Fuel pitch to diameter ratio ðP=DÞ
4. Size and pitch of fuel assemblies
5. Total thermal power
6. Channel flow characteristics such as coolant velocity ðVc Þ
7. Convective heat transfer
8. Pressure drop in coolant flow
9. Temperature differences (radial and axial).
The neutronic and TH modules are solved until the constraints are satisfied and a feasible solution is obtained. In
case the initial parameters obtained from a preliminary neutronic analysis are not favorable for a feasible solution then
a new set of values is obtained, if possible.
Once the neutronic-TH coupled simulations are performed, an optimization is performed to obtain the optimal solu-
tion. In case of genetic algorithms (GA), the method described in Section 9.5.1 is followed iteratively. This implies that
a set of chromosomes, in a population, is used for each of the modules. Such a scheme can be prohibitively inefficient
due to the large computation effort for each of the modules. For example, a GA code calling MCNP 100 times and then
calling each other code the same number of times would be far too much to be realistic. Thus a more efficient scheme,
shown in Fig. 13.3, utilizes regression and machine learning. In such a scheme, several runs are made initially with trial
data to obtain output which trains a regression model which subsequently is able to produce predictive output.
The GA chromosomes are tested for fitness in the regression models rather than directly into the modules thus
reducing the effort. Such schemes have been successfully implemented to give reliable optimized parameters.
Consider two optimizations where GA has been used for multi-objective multi-module multi-constraint optimization
for a Gen-IV lead-cooled fast reactor (LFR) (Section 3.5.3) by Luo, Zhang, Wang, Jiang, and Chen (2021), a gas-
cooled fast breeder reactor core (Kumar & Tsvetkov, 2015).
The optimization for a 1000 MWt medium-power modular lead-cooled fast reactor M2LFR the optimization is:
min J ðyÞ; s:t: x # x0 ; keff # keff;0 ; PFF # PFF0 (13.1)
where
X
4
J5 ai yi ðxÞ; (13.2)
i51
492 Nuclear Engineering

FIGURE 13.2 Coupled neutronics and thermal hydraulics calculation.

x 5 ðx1 ; x2 ; x3 Þ and the objective functions yi are the system multiplication keff ; the peak power factor ðPPFÞ; the clad-
ding temperature Tc and the fuel pellet temperature Tp with weights ai , and x0 5 ðx1;0 ; x2;0 ; x3;0 Þ are given maximum
values.
The fixed quantities are the number of fuel assemblies, control rods (CR) and safety rods (SR), fuel rod layers in an
assembly, the fuel enrichment, the linear power density and the materials; these are specified at the beginning of the
optimization.
In the first part of the neutronic optimization, the geometric design parameters are obtained for the fuel rods and
overall core design. Then, in a coupled neutronic-TH optimization, the core is modeled as consisting of six layers
namely the inner fuel zone, the SR zone with 12 safety rod assemblies, the outer fuel zone, the CR zone with 12 CR
assemblies, the reflector zone and the pure lead (Pb) zone. The layers are homogenized, and cross sections are prepared
for MCNP. The feasibility of the scheme is verified by comparing the maximum coolant velocity and fuel, cladding
and coolant temperatures with the design constraints, For optimization variables: D, P=D and Vc , the search space for

each independent variable is specified to obtain the optimum solution D 5 10mm; P=D 5 1:3; Vc 5 1:4ms21 . Thus,
the optimized geometric design and overall thermal and flow parameters are found.
For a 100 MWt gas-cooled fast breeder reactor core optimization is carried out (Kumar & Tsvetkov, 2015) with the
GA method in a scheme with a strategy similar to that depicted in Fig. 13.3. The four modules used in the optimization
are for the fuel pin, the whole core, thermofluids and heat transfer, and Brayton cycle energy conversion. The multi-
objective function comprises ten objectives namely high breeding of U-233 and Pu-239, desired radial and axial PPF,
multiplication constants keff and kN within prescribed limits, high fast fission factor, high energy conversion, limited
pressure drop and high fuel burnup. The constraints and constants are prescribed similarly to the previous case.
 Optimization in nuclear systems Chapter | 13 493

FIGURE 13.3 Coupled genetic algorithm-regression with

multi-module optimization.

The independent variables are the fuel radius ðrF Þ, the enrichment ðεÞ, the coolant inlet temperature ðTin Þ and the
_
mass flow rate m.
The simulation parameters are: number of chromosomes in every generation 5 40, number of bits per gene 5 16,
number of generations 5 100. Crossover probability 5 0.4, mutation probability 5 0.05 and 2 elite elements preserved.
The solutions seem to have converged within B70 iterations, giving the optimized solution: rF 5 0:217cm,
ε 5 18:467%, Tin 5 63:476 and m_  5 35:937 kg/s.
The two simulations just described are representative of current research in reactor core design optimization by GA;
the second uses multivariate regression for improved computational efficiency.
To optimize fuel pin burnable poison in UO2/Gd2O3 fuel placed in pressurized water reactor (PWR) fresh assem-
blies, Yilmaz, Ivanov, Levine, and Mahgerefteh (2006) have used GA to maximize the end of cycle (EOC) keff .
In several other works, core optimization has been successfully carried out with the particle swarm optimization
(PSO) method (Domingos, Schirru, & Pereira, 2006), for multi-objective sodium-cooled fast test reactors (Zeng, Stauff,
Hou, & Kim, 2020), for Small Modular Reactors (Akbari, Rezaei Ochbelagh, Gharib, Maiorino, & D’Auria, 2020), for
design optimization for a High Flux Isotope Reactor conversion project (Bae, Betzler, Chandler, & Ilas, 2021; Betzler,
Chandler, Cook, Davidson, & Ilas, 2019) and for a data-driven predictive model for predicting moisture carryover in a
boiling water reactor using Artificial Intelligence (AI) technique (Wang, Gruenwald, Tusar, & Vilim, 2021).
Considerable research effort is focused on the development of AI-based predictive models using data, look-up
tables and regression for addressing non-linearities and uncertainties.

13.3 Fusion neutronics design optimization

In a fusion reactor such as the International Thermonuclear Experumental Reactor (ITER, Section 10.6), the neutron
radiation from the plasma in the central cell produces secondary gamma rays during its interaction with structural and
blanket materials. Two important factors in fusion, distinct from fission reactors, are activation and the requirement of a
shield to ensure protection to the copper material in the surrounding superconducting coils which operate at cryogenic
temperatures (to be superconducting).
In a fusion reactor, the inboard (IB) and outboard (OB) blanket modules serve to produce tritium to sustain the
fusion reaction and to shield the superconducting coils. The blanket is exposed to intense radiation heating which is
494 Nuclear Engineering

removed by the flow of coolant. Some blanket designs under study for ITER and the Demonstration Fusion Power
Plant (DEMO) and the intermediate China Fusion Engineering Test Reactor are the helium-cooled solid breeding blan-
ket, the water-cooled lithium lead (WCLL) breeding blanket and the water-cooled ceramic breeding blanket. The objec-
tive of blanket optimization studies is to maximize the tritium production, with a tritium breeding ratio (TBR)
exceeding one, that is, more tritium produced than consumed and provide shielding for the blankets.
Some blanket/shield design optimization studies are:
1. optimization of the first wall armor, to maximize the TBR (Dai et al., 2021),
2. full 3D neutronic-TH coupled optimizations for TBR in IB and OB equatorial blankets, and the effect of beryllium
and tungsten armor on the shielding performance of the blanket (Cui et al., 2017),
3. tritium self-sufficiency and shielding effectiveness in the WCLL breeder blanket of the DEMO design (Moro et al.,
2020),
4. the design of an optimized WCLL mock-up experimental layout based on MCNP calculations (Flammini et al.,
2020), and
5. the development of a coupled neutronics-TH code with multi-objective optimization, Neutronics/Thermal-hydraulic
Coupling Optimization Code (NTCOC), for a radial build of the water-cooled ceramic breeder blanket (Li et al.,
2016).

13.4 Radiation shielding design optimization

The design of radiation shielding in large nuclear power reactors consists mostly of structural steel surrounding the core
and a large volume of concrete as a biological shield around the reactor. In swimming-pool type reactors, the water in
the pool acts as a shield.
For small compact and portable micronuclear reactors, particularly for space and underwater applications, the design
of a shield requires optimization of size, weight and cost. It is thus both desirable and necessary to find the optimal
materials and configurations.
Shielding is thus a multi-objective, multi-constraint optimization for which heuristic, meta-heuristic and AI-based
optimization methods are being applied and further developed for computational efficiency.
In the design of neutron and gamma radiation shielding, GA has also been used for design optimization with GA-
MCNP calculations (Cai, Hu, Pan, Hu, & Zhang, 2018) for multi-region and composite shields as illustrated by
Fig. 13.4.
In Fig. 13.4, the first material (labeled A) is typically iron (Fe) or tungsten (W) as a structural material and capable
of attenuating fast neutrons and gammas. This is followed by a hydrogenous material such as water (labeled A) which
further slows down the neutrons while producing gammas of B2 MeV which need a third layer of a material such as
lead (Pb) which is effective in attenuating energetic gammas. Optimization studies would thus typically give the dimen-
sions of each of the iron, water, and lead layers obtained by a calculation which searches for the optimal configuration.
Several other candidate materials and impurities include stainless steel SS316, tungsten, polyethylene, borated polyeth-
ylene, and B4C.
A detailed calculation would use energy-dependent cross sections for radiative capture ðn; γÞ and alpha production
ðn; αÞ reactions. For example, for removal through the capture ðn; γÞ reaction, iron has an order of magnitude higher at
low energies with prominent resonances at intermediate ( . 10 keV) energies extending well toward the high
(B1 MeV) energies.
Optimization can also be carried out to determine the optimal compositions of a composite homogenized shield as
shown in Fig. 13.4. A number of innovative combinations, such as mixing nanomaterials in concrete, have been investi-
gated (Norhasri, Hamidah, & Fadzil, 2017) to make new and better radiation shielding materials. A lightweight and
compact sixteen-layer gradient composite shield (Hu et al., 2020) has been designed by a GA-MCNP coupled simula-
tion based on epoxy resin, B4C, lead and a small amount of graphene oxide. The densities of three materials for the
composite are 11.34 g/cm3 for Pb, 1.20 g/cm3 for epoxy resin and 2.52 g/cm3 for B4C. In the 30 cm composite shield
the optimal density is found to be a parabolic shape starting from B10 g/cm3 in the beginning falling gradually to

FIGURE 13.4 Radiation shield design optimization.

 Optimization in nuclear systems Chapter | 13 495

slightly above B1 g/cm3 by about 20 cm and then gradually increasing to B10 g/cm3. This shield was developed and
tested and showed better performance than a uniform composite shield. In another GA-MCNP optimization for a shield
subject to a U-235 spectrum with mixed neutrons and gamma rays (Cai, Hu, Pan, Sun, & Yan, 2018), Cai et al. have
also shown that a composite multilayer shield optimized for volume and weight has the best performance. Extending
the study to volume, weight as well as minimum activation for a vehicle carrying a D-T neutron source, a model using
MCNP for neutronic calculations and a predictive model for mechanical and thermal properties a synergistic optimiza-
tion model based on evolutionary algorithms has been developed using two multi-objective methods (priori and poster-
iori) for the design of composite shields (Cai, Hao, Yu, Wang, & Hu, 2020).
For small and portable micronuclear reactors, the emphasis is on both safety and lightweight effective shields
(Ahmad, Chang, Li, Yang, & Liu, 2021; Kim & Moon, 2010). Coupling GA and neural network models (Song et al.,
2020) with improved efficiency over Monte Carlo (MC) simulations for multi-objective and multi-constrained radiation
shielding design optimization problem have been demonstrated it by the non-dominated sorting genetic algorithm
(NSGA-II) and mini-batch stochastic gradient descent. An adaptive mutation rate operator has been proposed to
improve the global searching capability. The optimization results show that the adaptive scheme can find the Pareto-
optimal front of reactor shielding designs precisely and its calculation speed is two orders of magnitude faster than the
pure MC method and similar computational accuracy.
Such an optimization based on GA has been used (Kim & Moon, 2010) for the search of an optimal shield subject
to a set of constraints. A number of innovative combinations, such as mixing nanomaterials in concrete, have been
investigated (Norhasri et al., 2017) to make new and better radiation shielding materials.

13.5 Fuel loading pattern optimization

A nuclear reactor begins with a fresh core with fuel that burns unevenly. Tin an ideal case, flux would have a more-or-
less cosine distribution in the radial and axial directions. The variation in fluxes gives rise to a peaking factor described
in the previous section in the context of optimization. With time, the fuel has burned unevenly and requires removal,
shuffling, or replacement. This is the area of fuel loading where the best strategy is desired for both technical and eco-
nomic reasons.
A utility that owns a nuclear reactor would typically have a planning horizon of a few years as shown in Fig. 13.5.
As an example, consider a 7-year horizon where it is desired to produce, in the first 5 years, a certain amount of (elec-
trical) energy Ei ; within cycles of length τ i ði 5 1; 2; . . . 5Þ months, with each cycle having a beginning of cycle (BOC)
and an EOC as shown in Fig. 13.6.

FIGURE 13.5 Possible paths include A-B-D-C-D and B-A-

B-A-C.
496 Nuclear Engineering

FIGURE 13.6 A fuel cycle with enrichment ei and fuel loading fraction fi at beginning of cycle.

At every BOC, the utility has a choice of adding fresh fuel of enrichment ei for a fraction fi of the total fuel in the
core. As an example, a utility might decide to add 3.2% enriched fuel for one-third of the core.
A fuel cycle has a front-end (uranium ore to fuel fabrication), a middle part (reactor operation and refueling), and a
back-end (open-cycle long-term storage or closed-cycle reprocessing).
During fuel reloading, typically one-third to one-half of the fuel may be removed and new fuel loaded in a LP while
the remaining fuel is shuffled. A plant is shut down during the reloading period. It is desirable to have as long as possi-
ble operation between refueling. An outage may last several days or possibly up to a month.
The question then arises is how to best place it, how does this decision affect the burnup and cycle length, and how
much electricity is planned in the coming cycle.
The reactor core would thus operate in cycles of different time durations during the planning horizon. Any decision
would then require a consideration of the costs and benefits; this would be formulated as an optimization problem with
the following parameters:
1. the number of fuel assemblies to remove
2. the location of fuel assemblies to remove
3. the next-cycle length
4. the reload fuel enrichment and batch fraction
5. the poison distribution to manage criticality, and
6. the pattern of remaining fuel for shuffling.
The constraints in this optimization problem would be the core power density ratios and the fuel burnup.
In the five cycles shown in Fig. 13.5, the nodes A, B, C, and D represent fuel types which are classified by their
enrichment and burnup; finding the optimal path from the first stage to the terminal stage would thus constitute a
Dynamic Programming (shortest path) problem solved by a recursive algorithm in Section 9.4.
Fig. 13.6 shows a 3-batch fuel LP in a 1/4th core with fresh fuel, once- and twice-burned fuel loaded in the
assemblies.
Alternatively, a path such as. A-B-D-C-D or B-A-B-A-C could be considered a chromosome for a GA optimization
scheme and its fitness could be evaluated. Both Dynamic Programming (DP) are GA and variants and advancements of these
methods are applicable for performing optimization analyses to find the optimal path in a fuel loading strategy (Fig. 13.7).
A DP application of fuel loading (Kearney, 1973) is formulated in terms of minimizing the total cost TC which is
the sum over all cycles N of the revenue requirement in a refueling strategy
X
N X
N
minTc 5 min C ½xi ; ei ; fi  5 min Ci (13.3)
i51 i51

subject to the constraints that the energy produced in cycle i is Ei , the power peaking in the ith cycle PPi # PPmax and
that the burnup at discharge Bi # Bmax .
In Eq. (13.3), Ci is the total discounted revenue requirement for producing energy Ei in the ith cycle, and xi , ei and fi
denote the state, reload enrichment and reload batch fraction of the reactor at the beginning of the ith cycle.
The energy production in the reactor is dependent on the mass of the fissile nuclides uranium-235, plutonium-239
and plutonium-241 in each cycle.
In a preliminary model, a single decision variable can be taken to be the batch fraction, and with possible values [A]
0.37, [B] 0.33, [C] 0.29 and [D] 0.25 and the fitness of a path selected as a GA chromosome can be evaluate. This
would of course require a coupled neutronic and depletion analysis for the burnup at the end of each cycle given a
BOC composition.
A sample optimization problem carried out for the Zion-I 1065 Mwe reactor (Kearney, 1973) for a 3-zone refueling
with a cycle energy production of 7363.2 GWHe and a discharge constraint of 50 GWD/T, gave C-A-D-D-C as the
optimal path for reload batch fraction.
 Optimization in nuclear systems Chapter | 13 497

FIGURE 13.7 A nuclear reactor core with a 3-batch fuel loading pattern.

TABLE 13.4 Optimal refueling policy.

 
Cycle Ei ðGWHe=cycleÞ fi ei w=o Ci ð106 $Þ P
N
Ci ð$106 Þ
i51

1 7831.2 0.293 4.2145 12.420355 12.420355

2 9784.8 0.373 3.3637 14.446574 26.866923
3 7180.8 0.253 4.3932 9.472094 36.339008
4 6420.0 0.253 2.8467 7.428905 43.767904
5 6859.2 0.293 3.2650 7.537723 51.305616

In a more detailed model, for each cycle of the optimal path, the cycle energy production Ei , reload batch fraction
fi , reload enrichment ei and the discounted cycle cost Ci were calculated using the two-group code and the DP algorithm
for each stage beginning from the final stage.
Some of the results for the optimal path from stage 1 to 5 reproduced from Kearney are listed in Table 13.4. This is
a simple illustration from legacy calculations when computing power was modest, but the design was based on elegant
mathematics.
Maximum fuel utilization within the power demand requirement can be achieved by removing the fuel when it has
achieved a maximum burnup, that is, there is minimum amount of fuel left in it. De Klerk et al. (1997) have demon-
strated a nonlinear mixed integer formulation for calculating the reload optimization pattern with minimum discharge
burnup. In this model, the two parameters used are the power density vector and the kN values for fuel elements from n
batches. Within a cycle, a linear decreasing function for the burnup equations is assumed for kN obtaining values at
each BOC ðt 5 0Þ and EOC ðt 5 TÞ such that the EOC value of kN at a cycle becomes the BOC value of the fuel at the
next cycle.
The GA method has been used with MCNP coupled with the PARET code (a point kinetics and TH to predict transi-
ents and Departure from Nucleate Boiling (DNB)) to replace five most burned fuel elements in a Training Research
Reactor Idaho General Atomics research reactor (Chham et al., 2021) with reloaded fresh fuel on the basis of maximiz-
ing kN , minimizing the central fuel temperature (CFT) and maximizing the Departure from Nucleate Boiling Ratio
(DNBR) as a safety margin. In this scheme, the chromosomes are input into MCNP where the neutronics calculation
gives the excess reactivity and PPF. The PPF is input in to the PARET model to give the CFT; the fitness of the chro-
mosomes is subsequently evaluated and updated for obtaining the converged solution. A similar fuel LP optimization
for a materials test reactor with 29 fuel assemblies has been carried out using PSO (Section 9.5.2). The group constants
were generated from Winfrith Improved Multigroup Scheme/D4S coupled with a diffusion theory code with objectives
of maximizing keff while incorporating a penalty function to keep the PPF low (Ahmad & Ahmad, 2018).
In the present scenario, optimal strategies for fuel loading are obtained from stochastic heuristic and meta-heuristic
methods with the help of expert-based systems. AI and Artificial Neural Networks (ANN) are designed on rule-based
expert systems on the workings of the human brain. A strong feature in computer architecture favoring both AI and
498 Nuclear Engineering

ANN is the speedup based on vector processing and parallel processing. In ANN, for example, the structure of the brain
is modeled in the form of layers of neutrons (decision points) connected by axons in which chemical ions carry signals
to connect neurons. The complexity of the brain is so vast that of the order of ten billion neurons are interconnected for
processing knowledge and information by which humans recognize, make decisions, and pass the command for action
through signals across the spinal cord. Automated LP search tools may be attractive to solve the huge combinatorial LP
optimization problem. Rules are developed based on the experience of people operating systems and are formulated in
the form of if this then do that. This is the way humans operate by separating knowledge from action. Thus, modules
such as neutronics, depletion models and TH are combined with heuristic rulesets on the placement of a particular type
of fuel in terms of enrichment and burnup at a certain location in the core. As an example, consider two rules (Nissan,
2019): an elimination rule which says Do not load a fresh assembly in such a position that it is adjacent to another
position where there is another assembly of the same kind, except when one of these two positions is in a corner posi-
tion, and a preference rule which says If it is a once-burned assembly that is currently being considered, then choose
for it—from among those positions that were not forbidden by Rules 1
6 (the elimination rules)—that position whose
distance from the center of the core is minimal.
Rules are coded into a program for neural processing to make a rule-based expert system or an intelligent system
based on GA.
Surrogate models encoding a set of heuristics have been proposed for use in quantum annealers, which can find
global minima by tunneling through a barrier (as in quantum physics), applied to PWR fuel loading (Whyte & Parks,
2021).

13.5.1 Optimal distribution: Pontryagin’s maximum principle

Pontryagin’s Maximum Principle (Section 9.2.5) was described in Chapter 9 to obtain an optimal control when the
admissible controls are discrete. Such a model can be used for the optimal placement of different enrichment fuels. A
simple and elegant application is given by Lee (1973) for optimal placement of fuel to minimize the mass in a reactor.
In a two-group diffusion formulation, fluxes and currents are the state variables x1 5 φ1 , x2 5 r φ_ 1 , x3 5 φ2 and
x4 5 r φ_ 2 expressed as ~
x 5 ðx1 ; x2 ; x3 ; x3 Þ
dxi
5 fi ð~
x ; uÞ (13.4)
dr
ði 5 1; 2; 3; 4Þ:
The independent control parameter u is the fuel enrichment. In this model, there are fuels of two enrichments
umin ; umax and the objective is to place them in such a way that the functional
ðR
J 5 2π uðrÞrdr (13.5)
0

is minimized; J represents the mass of fissile material in a cylindrical reactor of radius R and unit height. Another state
variable x0 is defined as
ðR
x0 5 2π uðrÞrdr (13.6)
0

A Hamiltonian is constructed as:

 Xn54
~ ~
H Ψ; x ; uÞ 5 Ψ f ð~
x ; uÞ (13.7)
i50 i i

where Ψi are the auxiliary functions defined as

dΨi Xn54 @fi ð~
x ; uÞ
52 Ψi (13.8)
dr i50 @xi
for ði 5 0; 1; 2; 3; 4Þ. The above is equivalent to the system:
dxi @H
5 ; (13.9)
dr @Ψi
 Optimization in nuclear systems Chapter | 13 499

and
dΨi @H
52 (13.10)
dr @xi
ði 5 0; 1; 2; 3; 4Þ
The Hamiltonian is then arranged as H 5 uϕ 1 ϑ so that the extremum on H can be determined by the sign change
and the appropriate control can be implemented. With the normalization condition 2πΨ0 5 2 1, gives
ϕðrÞ 5 2 r½1 1 αx3 ðrÞΨ2 ðrÞ 2 βx3 ðrÞΨ4 ðrÞ; (13.11)
1 1 1 r
ϑðr Þ 5 x 2 Ψ 1 1 r 2 x 1 Ψ2 1 x 4 Ψ3 2 x 1 Ψ 4 (13.12)
r τ1 r τ2
where α 5 kN =pD1 , β 5 1=D2 , τ 1 5 D1 =Σ1 and τ 2 5 D2 =pΣ1 .
The objective is to ensure that H remains maximum for all rAð0; RÞ; thus uðrÞ 5 umin for ϕ , 0 and uðrÞ 5 umax for
ϕ . 0.
Further, it is evident that φð0Þ 5 0 and φðRÞ 5 2 R2 , 0, since x3 , the thermal flux, vanishes at the extrapolated
boundary r 5 R. The fluxes (Section 5.3.2 for cylindrical geometry) are
sinμr cosμr sinhλr coshλr
x1 5 AS1 1 BS2 1 CS3 1 DS4 (13.13)
r r r r
and
sinμr cosμr sinhλr coshλr
x3 5 A 1B 1C 1D (13.14)
r r r r
and the auxiliary functions are
sinμr cosμr sinhλr coshλr
Ψ2 5 a 1b 1c 1d (13.15)
r r r r
and
sinμr cosμr sinhλr coshλr
Ψ4 5 aT1 1 bT2 1 cT3 1 dT4 (13.16)
r r r r
The above are sufficient to determine the minimum critical mass by using appropriate values for the control vari-
able, u depending on the sign of ϕ. This is a powerful tool, to determine the minimum critical mass but only after speci-
fying the number of zones in the core.
Based on the numerical results and the intuitive fuel shuffling of the previous section, now consider an optimal con-
trol method which will take us to the desired configuration. The state variables x1 (fast flux) and x3 (thermal flux) can
be used, with the boundary conditions (zero flux at the outer boundary r 5 R and zero current at r 5 0) to yield the criti-
cal determinant. The transversality conditions are used to obtain the boundary conditions of the auxiliary functions
ðΨ1 ð0Þ 5 Ψ3 ð0Þ 5 Ψ2 ðRÞ 5 Ψ4 ðRÞ 5 0Þ.
The flux with appropriate constants A; B; C; D, and coupling coefficients SðjÞ
i , for a homogeneous system, given by

φðjÞ ðjÞ ðjÞ ðjÞ ðjÞ ðjÞ ðjÞ ðjÞ ðjÞ

1 5 A S1 X1 1 B S2 X2 1 C S3 Y1 1 D S4 Y2 (13.17)
and

φðjÞ ðjÞ ðjÞ ðjÞ ðjÞ

2 5 A X1 1 B X2 1 C Y1 1 D Y2 (13.18)
where
sinμr cosμr sinhλr coshλr
X1 5 ; X2 5 ; Y1 5 ; Y2 5
r r r r
S1 5 S2 5 τ 2 ½μ2 1 βΣ2Mc ð1 1 uÞ
S3 5 S4 5 2 τ 2 ½λ2 2 βΣ2Mc ð1 1 uÞ
500 Nuclear Engineering

and
 
1 1 αu βΣ2Mc ð1 1 uÞ
μ2 5 ða 1 bÞ; and 2λ2 5 ;
2 μ2 τ 2 τ1
(   )12  
1 2 αu βΣ2Mc ð11uÞ 1
a βΣ2Mc ð11uÞ1 14 2 ; b  2 βΣ2Mc ½1 1 u 1
τ1 τ2 τ1 τ1

Thus μ 5 μðuÞ; λ 5 λðuÞ, that is, both the variables are functions of the control variable, and the coupling coeffi-
cients have to be written separately for each physical zone in which an optimal control is sought.
In the one-zone arrangement and two-zone arrangements, the boundary conditions are insufficient to determine an
optimal condition. The three-zone arrangement has 24 unknown constants and 24 boundary conditions and can therefore
be considered an optimal arrangement. This gives a ½umin ; umax ; umin  optimal fuel placement arrangement shown in
Fig. 13.8.
The critical determinant, set equal to zero for the optimal condition (minimum critical mass) can have only one vari-
able for which the criticality search can be carried out. Thus, for a nuclear reactor, when one value is given for the fuel
enrichment, the other value can be determined from this analysis.
First consider that the enrichment is specified in Zone 2, and it is required to determine the enrichment in Zones 1
and 3. The control problem is then: “given u 5 uM , r1 , r , r2 , find u 5 um for 0 # r # r1 and r2 # r # R.” Alternately,
we can have the control variables all given and carry out a search for the critical radius of the core.
The control variable u is defined as
Σ2Fc N5 σ2;a5
u 5 5 u γ
Σ2Mc Nw σ2;aw
The optimization goal is to determine a Hamiltonian for which
Hðy  ; u ; xÞ $ Hðy  ; u; xÞ

where umin # u # umax . As earlier stated, this discrete form readily yields “optimal” discrete values of u constant in sub-
domains of the problem. Thus where g is minimum (either sign), umax is applied for an extremum. The zeros of the
switching function will thus determine the controls applied.
The criticality determinant for the three-zone problem has ten simultaneous algebraic equations. As in Chapter 5,
since these equations are homogenous, they are solved by Cramer’s rule. The determinant is set equal to zero, the
dimensions are entered and for a maximum control umax in the central zone, the minimum control umin is found from
the determinant. These give pairs ½umin ; umax  of feasible solutions for the Kori-1 data considered by Lee (1973).
umax 0.9 1.0 1.1 1.2 1.3 1.4 1.5
umin 0.89 0.80 0.71 0.65 0.61 0.58 0.56

The combination ½umin 5 0:65; umax 5 1:2 gives a minimum-mass combination.

The minimum-mass optimal placement PMP formulation was reviewed for its elegance and its capability to solve
optimal LPs also. A variational formulation based on the adjoint formulation for higher-order perturbation theory has
also been used for fusion design optimization (Graca et al., 1988). The preferred optimization schemes at present are
AI-based meta-heuristic schemes with comparatively less use of deterministic or stochastic perturbation for
optimization.

FIGURE 13.8 Optimal discrete (bang-bang) control.

 Optimization in nuclear systems Chapter | 13 501

13.6 Radiation detection or optimization

Radiation detection (Section 1.4.6) is used extensively for radiation monitoring of people, vehicles, cargo, and at air-
ports to ensure security (Kouzes, Ely, Lintereur, Siciliano, & Woodring, 2009). Gamma radiation is monitored using
scintillators such as NaI(TI) and for better resolution, HPGe solid state detectors, while neutron radiation is measured
using gas proportional counters such as the He3 and BF3 detectors. Both gamma and neutron counting has to be done
efficiently to be able to accurately classify materials and to raise alarms only when necessary.
The efficiency of radiation counters depend on a number of factors such as the moderation needed for obtaining a
good signal from a gas proportional counter.
Consider, for example, an explosives detection system based on TNAA counting gammas with a NaI(TI) detector
and neutrons with a BF3 detector (Khan, Koreshi, & Yaqub, 2017).
Explosives such as TNT (trinitrotoluene) have carbon, oxygen, nitrogen, and hydrogen in varying fractions which
allows their identification by neutron bombardment and subsequent analysis of the scattered and emitted signal.
Fast (14 MeV) neutron bombardment of hydrogen, carbon, oxygen, and nitrogen atoms produce the following reac-
tions which produce neutrons and gammas:
612C 1 01nð14MeVÞ-612C 1 01n 1 γ ð4:4MeVÞ
714N 1 01nð14MeVÞ-714N 1 01n 1 γ ð5:1MeVÞ
816O 1 01nð14MeVÞ-816O 1 01n 1 γ ð6:1MeVÞ
while thermal neutron bombardment reactions are significant for H and N in comparison with reactions for C and O
producing characteristic gammas from hydrogen and nitrogen
11H 1 01nðthermalÞ-12H 1 γ ð2:2MeVÞ
714N 1 01nðthermalÞ-715N 1 γ ð10:8MeVÞ
For the ðn; γÞ reactions, the disappearance cross sections shown in Fig. 13.9 illustrate the extent to which thermal
neutrons are beneficial in comparison with fast neutrons (KAERI Nuclear Data Center, 2019). Thus, the H and N ther-
mal neutron ðn; γÞ reactions present an important method for the detection and characterization of explosives. Further,
as there are no portable thermal neutron sources, the fast neutrons from a source such as Am-Be or Califorium-252 will
be required to be thermalized before activating a sample under investigation. This non-destructive method of characteri-
zation is called thermal neutron activation (TNA).

FIGURE 13.9 Disappearance cross

sections.
502 Nuclear Engineering

For a sufficient signal to be produced from TNA, a detector needs to be optimized with a moderator. The design of
the optimal moderator is obtained by optimization methods on GA, PSO, coupled GA-MCNP and AI methods described
in the previous sections.
The functionals to maximize are the radiative capture reaction rates in the explosive for both hydrogen and nitrogen
are JHðn;γÞ and JNðn;γÞ defined as
ððð
JHðn;γÞ 5 ΣH n;γ φðr; E; ΩÞdVdEdΩ
ððð
JNðn;γÞ 5 ΣNn;γ φðr; E; ΩÞdVdEdΩ:

The independent variables are the moderator material, its size and placement, the source energy spectrum, boron
enrichment and gas pressure in the detector.
The optimization effort for moderator optimization is considerably less than that for a shielding optimization since
one material is sufficient. Typical candidate moderators are low-Z materials such as wax, paraffin wax and
polyethylene
The Bðn; αÞLi reaction rate in a single BF3 (cylindrical) detector with a Cf-252 source is shown in Table 13.5 for a
wax moderator of thickness in the range 1
30 cm, showing an optimum thickness B4 cm. The first five MCNP results
have relative standard error within 2% while the last two are within 5% and 19%, respectively.
The energy spectrum of the neutron flux in the detector tube in shown in Fig. 13.10 for moderators of thickness 2, 4
and 6 cm together with the B10 ðn; αÞ cross section. With an increase from a thickness of 2 to 4 cm, there is a slight
increase in the low energy flux due to which there is an increase in the reaction rate shown in Table 13.5 while an

TABLE 13.5 Bðn; αÞLi reactions for a wax-moderated BF3 detector.

ΔR (cm) 1 2 4 6 8 15 30
B ðn; αÞLi 3.20627 8.96149 12.4770 9.72714 8.23782 1.31617 0.07004
26 3
10 /cm /s 0.0305 0.0204 0.0172 0.0200 0.0213 0.0489 0.1930

FIGURE 13.10 Neutron flux energy spectrum in the wax-

moderated BF3 detector.
 Optimization in nuclear systems Chapter | 13 503

TABLE 13.6 Bðn; αÞ reaction rate as a function of pressure and 10B weight fraction.
10
B in B Gas density ρ0 1023 g/cm3 1025 ðn; αÞ reactions cm23 per source neutron s21
P 5 1 atm P 5 2 atm P 5 3 atm
ρ5ρ0 ρ 5 2ρ0 ρ 5 3ρ0
0.2 2:7527 3 1023 1.15580 1.89182 2.4132
0.5 2:7406 3 1023 2.16983 3.07464 3.5792
0.9 2:7243 3 1023 3.89159 5.51797 6.4283

FIGURE 13.11 Characteristic H; N gamma

peaks from trinitrotoluene.

increase in moderator thickness to 6 cm there is a reduction in the flux due to over-moderation which results in
increased attenuation.
Simulations for polyethylene moderator shows an optimized thickness of 6 cm which is also reported as an experi-
mental result in the literature (Waheed et al., 2017).
MCNP5 simulations were carried out to estimate the ðn; αÞ reaction rate, at T 5 300K as a function of pressure over
the range 0.5
3 atm in a BF3 tube and the isotopic enrichment of 10B. The results of Table 13.6 show that the reaction
rate is best for highest 10B enrichment as well as for highest gas pressure giving a fourfold increase. While high enrich-
ment is possible, there is limit on increasing gas pressure due to the high operating voltage that would be required at
increased pressure.
In an explosives detection system, a source such as Californium-252 can be used to scan an area for mines by firing
fast neutrons on the ground. Due to their penetrating power, the neutrons would undergo interactions with any subsur-
face material. In the case of concealed TNT, the back-scattered neutron and gamma radiation would be detected in the
BF3 and NaI (TI) detectors; from the resulting spectrum such as the pulse height of the gamma radiation shown in
Fig. 13.11, it would be possible to estimate characterize an explosive.

13.7 Controller design optimization

Nuclear reactors have traditionally operated as baseload power plants with some maneuvering capability. Looking
ahead, it is possible that a load-following capability might become necessary as renewables are added to electricity
grids. Thus, a load-following mode would likely improve the operational costs. The technical issues with a strong
504 Nuclear Engineering

bearing on load-following would be the speed by which temperatures and pressures in the moderator and coolant and
the subsequent reactivity changes would be adequately controlled by the available mechanisms. This would mean that
CR and poison management would require a careful re-appraisal. As an example, the ratio of gray to black CR would
require re-adjustment. It is likely that with new modular reactors, load-following might not be as difficult as with a sin-
gle large unit.
At present, countries such as France, with B70% nuclear, and a few others have a significant nuclear share of elec-
tricity. For such countries, a load-following capability would become increasingly important.
One technical challenge for load-following would be an optimized controller which could respond in a good way.
That translates into an optimization problem that requires minimization of a cost function that incorporates the over-
shoot, settling time, and stabilization time. Several optimization methods, such as GA, PSO, AI and fuzzy logic, have
been designed and optimized. The Proportional Integral Derivative (PID) controller (Section 9.3) has been optimized
for a load-following PWR using PSO (Mousakazemi, 2021) to minimize the overshoot, settling time and stabilization
time. The agreement between the desired signal and the PID controller signal has been within a good range as con-
firmed by changes in the control rod movements, changes in precursor density, xenon concentration and coolant
temperature.
In another load-following mode simulation of a PWR (Abdulraheem & Korolev, 2021), the nonlinear Integral slid-
ing mode control has been used in combination with an optimal control Linear Quadratic Gaussian to show the robust-
ness of the controller against disturbances incorporating the point kinetics equations with xenon reactivity. In
simulation experiments the core is modeled with point kinetics with xenon reactivity feedbacks. The effectiveness of
the hybrid controller is demonstrated by successful load-following under parametric variations. The issue of chattering
disturbance, due to frequent switching of the controller, has been addressed by using a sliding mode controller for
robust load-following in the presence of disturbances and uncertainties (Hui & Yuan, 2021). The proposed controller
follows the load with a maximum power error of less than B104 W, compared with B5 3 105 W for a PID controller
and 4 3 105 W for a conventional sliding mode controller.
There are several areas in nuclear engineering where optimization, particularly meta-heuristics are being applied and
developed. Some of these areas have earlier been mentioned in chapter none. As stated in the introduction, optimization
is indeed a vast area full of interesting ideas yet to be explored and developed.

Problems
1. Describe briefly how the micronuclear reactor described in Section 10.5.3 can be optimized in terms of weight and
volume.
2. Describe a possible application of MC perturbation theory (Section 12.6) for calculating sensitivity coefficients of
candidate materials in a shielding design optimization.
3. Considering the various breeder blankets for DEMO and the China Fusion Engineering Test Reactor (Section 13.3)
describe how a blanket could be optimized for maximizing the TBR with a GA-MCNP coupled scheme.
4. Consider the composite gradient shield (Section 13.4) and write down a two-group diffusion model with variable
atomic density (spatial dependence) for two candidate materials; iron and water to minimize the radiation dose
across the shield.
5. With the two-group diffusion equation criticality equation (Section 5.3.2), show that a one-zone model does not
have an optimal arrangement. Then consider the two-zone bare cylinder to find critical pairs for the fuel enrichment
in each zone with the criticality determinant

S11 J0 ðμRÞ S13 I0 ðλRÞ

50
J0 ðμRÞ I0 ðλRÞ
     
2μ1 S11 J1 μ1 r1 μ2 S21 J1 μ2 r1 μ2 S22 Y1 μ2 r1 λ1 S13 I0 ðλ1 r1 Þ 2λ2 S23 I1 ðλ2 r1 Þ λ2 S24 K1 ðλ2 r1 Þ
2 μ1 J1 μ1 r1 μ2 J1 μ2 r1  μ2 Y1 μ2 r1  λ 1 I 1 ð λ 1 r1 Þ 2λ2 I1 ðλ2 r1 Þ λ 2 K 0 ð λ 2 r1 Þ
S11 J0 μ1 r1 2 S21 J0 μ2 r1 2 S22 Y0 μ1 r1 S13 I0 ðλ1 r1 Þ 2 S23 I0 ðλ2 r1 Þ 2 S24 K0 ðλ2 r1 Þ
J0 μ 1 r1 2 J0 μ2 r1 2 Y0 μ1 r1 I0 ðλ1 r1 Þ 2 I0 ðλ2 r1 Þ 2 K0 ðλ2 r1 Þ
0 S21 J0 μ2 R S22 Y0 μ2 R 0 S23 I0 ðλ2 RÞ S24 K0 ðλ2 RÞ
0 J0 μ2 R Y0 μ2 R 0 I 0 ð λ 2 RÞ K0 ðλ2 RÞ
 Optimization in nuclear systems Chapter | 13 505

where, Sij are the coupling coefficients for zone i and index j with
S11 5 τ 2 ðμ2 1 βuÞ 5 S21 5 S22 ; S13 5 2 τ 2 ð λ2 2 βuÞ 5 S23 5 S24
 
1   qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
μ 5
2
2 τ 1 1 L 1 ðτ 1 1L Þ 1 4ðkN 2 1Þτ 1 L
2 2 2 2
2τ 1 L2
 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1  
λ2 5 τ 1 1 L2
1 ðτ 1 1L2 Þ2 1 4ðkN 2 1Þτ 1 L2 :
2τ 1 L2

Nomenclature
English Lower Case
e enrichment
f fraction
keff effective multiplication factor
kN infinite multiplication factor
n neutron density
r radius
u control parameter

English Upper Case

C precursor concentration
D diffusion coefficient
D diameter
E energy
J integral performance index
P pitch
P pressure
S coupling coefficient
T temperature
Tc total cost

Greek Lower Case

β delay fraction
ε enrichment
λ decay constant
φ neutron flux
ρ reactivity
σ microscopic cross section
τ cycle time

Greek Upper Case

F objective function
Σx macroscopic cross section (reaction x)
H Hamiltonian
Ψ auxiliary function

Abbreviations and acronyms

AI Artificial Intelligence
ANN Artificial Neural Networks
BOC beginning of cycle
CR control rod
EOC end of cycle
GA genetic algorithm
506 Nuclear Engineering

LFR Lead-cooled fast reactor

LP loading pattern
PPF peak power factor
PSO particle swarm optimization
PWR pressurized water reactor
SR safety rod
TH thermal hydraulics
TNAA thermal neutron activation analysis
WCLL water-cooled lithium lead

References
Abdulraheem, K. K., & Korolev, S. A. (2021). Robust optimal-integral sliding mode control for a pressurized water nuclear reactor in load following
mode of operation. Annals of Nuclear Energy, 158, 108288.
Ahmad, A., & Ahmad, S. U. L. I. (2018). Optimization of fuel loading pattern for a material test reactor using swarm intelligence. Progress in
Nuclear Energy.
Ahmad, S., Chang, B., Li, B., Yang, Q., & Liu, C. (2021). Mass optimization of the radiation shadow shield for space nuclear power system. Progress
in Nuclear Energy.
Akbari, R., Rezaei Ochbelagh, D., Gharib, A., Maiorino, J. R., & D’Auria, F. (2020). Small modular reactor full scope core optimization using
Cuckoo Optimization Algorithm. Progress in Nuclear Energy.
Bae, J. W., Betzler, B. R., Chandler, D., & Ilas, G. (2021). Automated fuel design optimization for high flux isotope reactor low enriched uranium
core design. EPJ Web Conference.
Betzler, B. R., Chandler, D., Cook, D. H., Davidson, E. E., & Ilas, G. (2019). Design optimization methods for high-performance research reactor core
design. Nuclear Engineering and Design.
Cai, Y., Hao, R., Yu, S., Wang, C., & Hu, G. (2020). Comparison of two multi-objective optimization methods for composite radiation shielding mate-
rials. Applied Radiation and Isotopes: Including Data, Instrumentation and Methods for Use in Agriculture, Industry and Medicine.
Cai, Y., Hu, H., Pan, Z., Hu, G., & Zhang, T. (2018). A method to optimize the shield compact and lightweight combining the structure with compo-
nents together by genetic algorithm and MCNP code. Applied Radiation and Isotopes: Including Data, Instrumentation and Methods for Use in
Agriculture, Industry and Medicine.
Cai, Y., Hu, H., Pan, Z., Sun, W., & Yan, M. (2018). Metaheuristic optimization in shielding design for neutrons and gamma rays reducing dose
equivalent as much as possible. Annals of Nuclear Energy.
Chham, E., et al. (2021). Fuel reloads optimization for TRIGA research reactor using Genetic Algorithm coupled with neutronic and thermal-
hydraulic codes. Progress in Nuclear Energy.
Cui, S., et al. (2017). Evaluation and optimization of tritium breeding, shielding and nuclear heating performances of the helium cooled solid breeder
blanket for CFETR. International Journal of Hydrogen Energy.
Dai, T., Cao, L., He, Q., Wu, H., Feng, K., & Cao, Q. (2021). Neutronics optimization study on the first wall design of CFETR for TBR enhancement.
Fusion Engineering and Design, 172.
De Klerk, E., et al. (1997). Optimization of nuclear reactor reloading patterns. Annals of Operations Research.
Domingos, R. P., Schirru, R., & Pereira, C. M. N. A. (2006). Particle swarm optimization in reactor core design. Nuclear Science and Engineering:
The Journal of the American Nuclear Society.
Flammini, D., et al. (2020). Pre-analysis of the WCLL breeding blanket mock-up neutronics experiment at the frascati neutron generator. Fusion
Engineering and Design.
Graca, G. T., Lewins, C. O., Parks, J. D., Graca, G. T., Lewins, C. O., & Parks, J. D. (1988). Optimization of blanket design for fusion reactors by
higher order perturbation theory. Nuclear Energy, 27(121), 121
130.
Hu, G., Shi, G., Hu, H., Yang, Q., Yu, B., & Sun, W. (2020). Development of gradient composite shielding material for shielding neutrons and gamma
rays. Nuclear Engineering and Technology.
Hui, J., & Yuan, J. (2021). Chattering-free higher order sliding mode controller with a high-gain observer for the load following of a pressurized water
reactor. Energy, 223, 120066.
Kearney, J. P. M. E. A. (1973). Simulation and optimization techniques of nuclear in-core fuel management decisions. Massachusetts Institute of
Technology.
Khan, H., Koreshi, Z. U., & Yaqub, M. (2017). The sensitivity studies of a landmine explosive detection system based on neutron back scattering
using Monte Carlo simulation. Nuclear Technology & Radiation Protection.
Kim, D. W. (2019). Site-specific PSR process for Kori-1 permanent shut down NPP. In PBNC 2018—Pacific Basin nuclear conference, San
Francisco, California, USA, 30 September - 4 October 2018. ISBN: 978-1-5108-7472-5. http://toc.proceedings.com/42142webtoc.pdf
Kim, B. S., & Moon, J. H. (2010). Use of a genetic algorithm in the search for a near-optimal shielding design. Annals of Nuclear Energy, 37(2),
120
129.
Kouzes R. T., Ely J. H., Lintereur A. T., Siciliano E. R. & Woodring M. L. BF3 neutron detector tests. Pacific Northwest National Laboratory http://
www.pnl.gov/main/publications/external/technical_reports/PNNL-19050.pdf (Date of access: 03/12/2016) (2009).
 Optimization in nuclear systems Chapter | 13 507

KAERI Nuclear Data Center. (2019). https://atom.kaeri.re.kr/nuchart/ last accessed 20 June 2021.
Kumar, A., & Tsvetkov, P. V. (2015). A new approach to nuclear reactor design optimization using genetic algorithms and regression analysis. Annals
of Nuclear Energy.
Lee, C. K. (1973). Critical mass minimization of a cylindrical geometry reactor by two-group diffusion equation. Journal of the Korean Nuclear
Society, 5(2), 115
131.
Li, J., Jiang, K., Zhang, X., Nie, X., Zhu, Q., & Liu, S. (2016). Nuclear-thermal-coupled optimization code for the fusion breeding blanket conceptual
design. Fusion Engineering and Design.
Luo, X., Zhang, X., Wang, S., Jiang, R., & Chen, H. (2021). Automated core design code development for a lead-cooled fast reactor and its core opti-
mization. International Journal of Energy Research.
Moro, F., et al. (2020). Nuclear analysis of the water cooled lithium lead DEMO reactor. Fusion Engineering and Design.
Mousakazemi, S. M. H. (2021). Control of a pressurized light-water nuclear reactor two-point kinetics model with the performance index-oriented
PSO. Nuclear Engineering and Technology.
Nissan, E. (2019). An overview of AI methods for in-core fuel management: Tools for the automatic design of nuclear reactor core configurations for
fuel reload, (re)arranging new and partly spent fuel. Designs.
Norhasri, M. S. M., Hamidah, M. S., & Fadzil, A. M. (2017). Applications of using nano material in concrete: A review. Construction and Building
Materials.
Rozin, S. G., Rubin, I. E., & Soboĺ, I. M. (1990). On multicriterial optimization in reactor problems using Monte Carlo calculations. Progress in
Nuclear Energy.
Song, Y., et al. (2020). Research on fast intelligence multi-objective optimization method of nuclear reactor radiation shielding. Annals of Nuclear
Energy.
Stewart, R. H., Palmer, T. S., & DuPont, Bryony (2021). A survey of multi-objective optimization methods and their applications for nuclear scientists
and engineers. Progress in Nuclear Energy, 138.
Waheed, A., et al. (2017). Optimization of moderator assembly for neutron flux measurement: Experimental and theoretical approaches. Nuclear
Science and Techniques.
Wang, H., Gruenwald, J. T., Tusar, J., & Vilim, R. (2021). Moisture-carryover performance optimization using physics-constrained machine learning.
Progress in Nuclear Energy.
Whyte, A., & Parks, G. (2021). Quantum annealing optimization of a heuristic suurogate model for PWR fuel loading. EPJ Web Conference.
Yilmaz, S., Ivanov, K., Levine, S., & Mahgerefteh, M. (2006). Genetic algorithm to optimize the UO2/Gd2O3 fuel pin designs in a pressurized water
reactor. Nuclear Technology.
Zeng, K., Stauff, N. E., Hou, J., & Kim, T. K. (2020). Development of multi-objective core optimization framework and application to sodium-cooled
fast test reactors. Progress in Nuclear Energy.
Chapter 14

Monte Carlo simulation in medical

physics

14.1 Introduction
In medical physics (MP), radiation plays a very important role for diagnostics, pathological investigations and therapy,
with Monte Carlo (MC) simulation appearing as a high-preference methodology just as it has become in nuclear engi-
neering (Section 7.2). One measure of the number of research journals, the Web of Science Master Journal List (2021)
lists 5,121 journals relevant to a search on “nuclear medicine”, 2,278 for “medical physics” and 1,404 for “biomedical
physics” out a total of 24,930 results. The growth in the number of papers in Physics in Medicine and Biology in the
initial years of computing was more than 20-fold in the years from 1970 to 2005 (Rogers, 2006) with electron-photon
transport as the foundation of the versatile Electron Gamma Shower (EGS) system of codes (Bielajew & Rogers, 1989;
Hirayama, 2005) just as the Monte Carlo N-Particle (MCNP) code was in the nuclear community for criticality and
nuclear reactor calculations. A search on “Monte Carlo Simulation” in the ScienceDirect database of 2021 lists 1,013
results for the number of research articles published in Medicine and Dentistry. In MC simulation, the MCNP code is
extensively used in medical physics as given in the primer on Medical Physics Calculations with MCNP (Lazarine &
Goorley, 2005; Reed, 2007) covering the modeling of tumors in tissue, phototherapy, and brachytherapy as well as an
elaborate description on the use of the Zubal phantom (a Computerized Tomography(CT)-based torso and a CT-based
head for use in MC simulations of realistic nuclear medicine (NM) imaging geometries).
For imaging, liquids containing radio-isotopes are injected in a patient and used as radiotracers giving images as
they collect in the vicinity of an organ or move in an available channel. The images are produced from the emitted
radiations and captured in a camera such as the Single-Photon Emission Computed Tomography (SPECT/CT) 3D scan.
Similarly, a positron emitter such as fluorine-18 is used as a radiotracer in a Positron Emission Tomograhy (PET)/CT
for imaging by the production of gamma rays formed by positron-electron annihilation.
Cancer cases globally are expected to grow from 14.1 million in 2012 to 24 million by 2035 with the top three: lung
(Yu, Lewis, Luisa Trejos, Patel, & Malthaner, 2011), prostate and colorectal cases accounting for over 47% of all can-
cers in men (Blanchard et al., 2018; Rice et al., 2019). The treatments for cancer, in order of general preference, are
chemotherapy, surgery, and radiation therapy. However, in the case of recurrent cancer, brachytherapy is the preferred
treatment due to the localized and non-invasive effects.
Another use of radiation is for the treatment of cancer by applying radiation from an external source placed some
distance away from a person or by placing seeds of a radio-isotope inside the body of a person close to the organ where
a tumor is to be destroyed. An example of the former, called tele-therapy is the cobalt knife where gamma rays from a
cobalt-60 source focus on an organ which is dangerous to be operated upon by surgery such as a tumor inside the brain.
The treatment by placing a source within the body is called brachytherapy from the Greek work brachus meaning
short.
In NM, radiation nuclide therapy is used for the imaging of the thyroid, bones, heart, liver and for the treatment of
cancer in organs such as the lungs, breast, colon and rectum, prostate, stomach, and liver etc. Nuclear radiation, in the
form of X-rays, gamma rays and charged particles, can damage the DNA (deoxyribonucleic acid) molecules of cancer
cells that carry the genetic information and pass it from one generation to the next, to stop their further division.
More than two hundred radio-isotopes are used in NM common radio-isotopes include technetium 99mTc, iodine
125
I, palladium 103Pd, iridium 192Ir, cesium 137Cs, and cobalt 60Co. Important radionuclides for brachytherapy include
the conventional encapsulated (Perez-Calatayud et al., 2019) low dose rate (LDR) sources (,200 cGy per hour) Ir-192,
and high dose rate (HDR) sources ( . 1200 cGy per hour). Permanent brachytherapy sources (Lechtman et al., 2013)

Nuclear Engineering. DOI: https://doi.org/10.1016/B978-0-323-90618-0.00014-4

© 2022 Elsevier Inc. All rights reserved. 509
510 Nuclear Engineering

with energy ,50 keV, such as 125I and 103Pd are commonly used LDR sources, with typical implants of 50
80 metallic
seeds encasing isotopes, used for the treatment of cancer.
There are several radiations which are used for radiotherapy according to the energy emitted (Section 1.2) and the
ways in which they interact with matter (Section 1.3), for example α, have a short range and high energy with a high
linear energy transfer (LET) which makes them favorable for short-range treatments. The β rays are, in contrast, longer
range radiation so they would not be expected to kill tumor cells at a short distance; the same would hold true for
gamma radiation.

14.1.1 The production of radio-isotopes

In nuclear power reactors, as fission takes place, there are several fission fragments such as in the case of fission from
Uranium-235, molybdenum is produced in about 6.1% of the nuclear fission reactions. Molydenum (Mo-99) is
unstable and decays with a half-life of 66 years which means that in 33 years its activity gets reduced to half of what it
was; so it is fairly stable. During its decay, it produces technetium-99m, a metastable state, with a half-life of six hours
which compared with Mo-99 is very short. If a patient had to undergo a procedure with technetium in the evening and
the radio-isotope was procured the previous day then by the time of the procedure it would have only a quarter of its
strength left. These are the typical problems associated with radio-isotopes; they are unstable.
The other places where radio-isotopes are produced are nuclear research reactors, cyclotrons and generators where a
particle is bombarded and after transmutation, it becomes the desirable radio-isotope.
Coming back to technetium-99m, a radio-isotope used in over 85% of all 30 million patient examinations every year
worldwide, is so useful because of the gamma radiation it gives off during its decay

43 -Tc43 1 γ
Tc99m ð140:5keVÞ:
99

Technetium-99m (Tc-99m) is used in NM for the diagnostic scans of the organs. In the United States alone, with
about 10 million procedures/year over half are cardiac-related. Together with Canada, Germany, France, Japan, Italy,
Spain, Belgium, and UK, these countries account for about 9% of the procedures for diagnostic scans of a broad range
of body parts and for the diagnoses of cancer, heart disease and neurological disorders including dementia and move-
ment disorders (OECD/NEA, 2019).
The gamma radiation coming out of a technetium radio-isotope as it stays placed near or in an organ, is captured in
a gamma camera such as the NM technologies SPECT/CT and PET/CT with little slits to differentiate parts of space
where the intensity is greater than that in other parts. This spatial distribution in slices is put together into a 3D image.
Some other imaging techniques are based on strong magnetic fields and radio waves in the Magnetic Resonance
Imaging (MRI) machine, and the X-ray based CT scans; these are fundamentally different from NM imaging technolo-
gies. The NM techniques are functional as they give information on an organ as well as on its activity while the non-
NM technologies give anatomical information without organ activity.
The objective in this chapter is to understand how MC simulations play such an important role in estimating the
effect of radiations so that a doctor may be able to plan a surgery for a patient according to the size and condition of
the tumor.
The calculation schemes for MC simulations of α; β and γ radiation transport (Section 1.6) are suitable for medical
physics because they can model a great number of relevant mechanisms. At the same time, it would be true to say that
while MC simulations for nuclear engineering calculations may be excellent for performing dosimetry calculations and
give very accurate estimates of the dose (Section 1.4) from a given amount of a radio-isotope, they may be insufficient
to model the interaction of radiation with biology particularly the DNA which has a biological structure that is still not
completely understood.
In spite of a remaining biology interface, MC simulation has been successfully applied for modeling ionizing radia-
tion (IR) useful for dosimetry and therapy (Chatzipapas et al., 2020) by codes such as MCNP (Pelowitz et al., 2013),
EGS, FLUKA (Fluctuating Kaskade), and PENELOPE (Penetration and ENErgy LOss of Positrons and Electrons
(Archambault & Mainegra-Hing, 2015) at the tissue and organ level.
Efforts in the modeling of IR-induced DNA Damage, DNA Damage Response (DDR) and the DNA Repair mechan-
isms referred to as computational radiobiology is still in the developing and emerging stage. For simulations in biologi-
cal matter at small scales (nm-μm), Monte Carlo Track Structure (MCTS) codes are used while the GEometry ANd
Tracking code (Geant4)-DNA incorporates radiobiological functionalities (Chatzipapas et al., 2019; Engels et al., 2020)
capable of modeling interactions of gold nanoparticles (GNPs) to quantify their dose enhancement. Inter-comparisons,
using seven well-known MC codes have been carried out to estimate dose enhancement effects (Li et al., 2020) by
 Monte Carlo simulation in medical physics Chapter | 14 511

irradiation of single GNPs of 50 and 100 nm diameter by X-rays generated from 50 to 100 kilovolts. The results have
showed large variations and uncertainties in the electron energy spectra in the range 100
500 eV indicating that model-
ing effort is still required.
In this chapter, the discussion and use of MC will be largely connected with the calculation of the angular flux
^ E; tÞ in an organ or a tissue of given composition. The appropriate reaction rate yielding the dose D in medically
φðr; Ω;
relevant quantities as the gray, rad, or rem will be used to make the connection with the dosage for a medical treatment.
In MCNP, this is the F6 tally defined as
ð ð ð ð
ρ ^ E; tÞMeV=g
D5 dE dt dV dΩ σt ðEÞHðEÞφðr; Ω;
m
where all quantities are defined in the preceding chapters of neutron diffusion, transport and MC simulation. Recall that
a specific energy in MeV/g is converted to yield the dose rate Gy/h with 1 Gy (1 gray) 5 1 J/kg 5 100 rad. Also recall
that the background radiation dose is of the order 2
3 mSv/year, or 0.2
0.3 rem/year (1 rem 5 0.01 Sv 5 10 mSv). A
typical NM treatment could expose a person to a few millisievert while to destroy a tumor more than a few hundred Gy
could be required. Regulators use the linear non-threshold dose model (LNT) according to which there is no safe level
of radiation as cancers and heredity affects are understood to be stochastic, rather than deterministic, events. Therefore,
there is no safe limit as regards the amount of radiation that can be considered safe. Technically, it is preferred that the
amount of radiation should satisfy the ALARA (As Low As Reasonably Achievable) approach which amounts to saying
that no amount of radiation is safe. Thus, a radiation of about 6 millisievert, though small, is still undesirable when
accumulated in a hospital treatment.
The following sections consider MC applications for alpha and gamma radiations in some MP treatments.

14.1.2 Alpha radiation therapy

As mentioned above, the use of α particles in MP is preferable for short-range (70
100 μm) target tumor cells due to
their high LET. The relative biological effectiveness (RBE) of α radiation is about 20 times higher than for photons,
electrons and protons; thus, a higher dose is delivered. Ra-223, for example, with the reaction

88 -Rn86 1 He2
Ra223 219 4

emits a 5.77 MeV α and a 0.141 MeV gamma ray, has a half-life of 11.4 days and is therefore a suitable radio-isotope
for cancer treatment such as in boron neutron capture Bðn; αÞ therapy (Section 1.2.4). Some high energy alpha emitters
good for targeted radiotherapy are astatine At-211 (E 5 6 MeV, t1/2 5 7.2 hours), bismuth Bi-213 (E 5 6 MeV, t1/
2 5 46 minutes) and actinium Ac-225 (E 5 5.9 MeV, t1/2 5 10 days) (Zalutsky & Pruszynski, 2012).
A boron neutron capture therapy (BNCT) would function as depicted in Fig. 14.1. Consider a group of tumor cells
in the vicinity of normal cells in the human brain with blood flowing and an intricate mechanism operating in the brain;
so intricate that no physical surgery would be considered free of a big risk.
With BNCT, an amount of boron, in whatever form, would be injected in the brain and neutrons would be fired
from outside initiating the nuclear reaction

5 1 n0 -He2 1 Li3
B10 1 4 7

which would give off α (He42 Þ radiation which would travel to the desired site and destroy the tumor. Of course, there
are still many mechanisms by which the IR could find its way to the tumor cells rather than to the normal cells; that is

FIGURE 14.1 Boron neutron capture therapy.

512 Nuclear Engineering

another technology beyond the scope of this chapter but as a mention, it uses one of several schemes such as polymer
coated nanoparticles on which a magnetic field is applied to recognize some characteristic of a tumor cell and a drug is
released to penetrate the surface of the tumor cell.
BNCT has been assessed (Dymova, Taskaev, Richter, & Kuligina, 2020) in view of the development of new accel-
erators and has been found to be useful if performed alone or with chemotherapy or radiotherapy provided more selec-
tive boron delivery agents are developed to transport boron compounds into the brain, and an epithermal neutron beam
with definite characteristics can be developed. MC simulations have been used with the Particle and Heavy Ion
Transport Code System (Matsuya, Fukunaga, Omura, & Date, 2020; Sato et al., 2015) to show good agreement with
in vitro experimental data for BNCT with Boron-10 concentrations of 10 ppm. Several MC simulations have been car-
ried out for modeling the epithermal flux for BNCT treatment (Darda, Soliman, Aljohani, & Xoubi, 2020; Hassanein,
Hassan, Mohamed, & Abou Manoury, 2018; Kasesaz, Khalafi, & Rahmani, 2014; Monshizadeh, Kasesaz, Khalafi, &
Hamidi, 2015; Shaaban & Albarhoum, 2015) including the development of a head phantom for accurate calculations of
the dose requirement (Bavarnegin, Khalafi, Sadremomtaz, & Kasesaz, 2016). To determine the place for a patient to be
treated by BNCT, a MC simulation of the BAEC TRIGA reactor (Darda et al., 2020) using the OpenMC concludes that
the thermal column with an epithermal flux of B109 near the thermal column graphite is a suitable location. In another
study, with a miniature source neutron reactor (MNSR), using MCNP4C (Monshizadeh et al., 2015), the thermal and
epithermal neutron fluxes are reported as 1.39 3 109 n/cm2/s and 0.635 3 109 n/cm2/s, respectively, with a treatment
time estimated to be about 70 minutes.

14.2 Brachytherapy
Consider now a brachytherapy procedure in which MC simulations are used to calculate the dose absorbed by a tissue
in an organ.
An MC simulation could give useful information in the size, composition and configuration of GNPs for example,
to maximize the dose since that is the ultimate goal of brachytherapy. The constraint is that there should be minimum
collateral damage to the surrounding normal tissue and hence this could be formulated as an optimization problem ame-
nable to the solutions described in Chapter 9, particularly the meta-heuristic applications discussed Chapter 13. In
brachytherapy, there is still no consensus on what the optimal size, shape and distribution of GNPs should be. There is
some experimental and pre-clinical evidence for mouse tumors showing a 1-year survival rate of 86% mice given a
dose of 26 Gy with 1.9 nm intravenously administered GNPs versus 20% for tumors not laden with GNPs (Jain, Hirst,
& O’Sullivan, 2012). This is the motivation for performing detailed MC simulations for brachytherapy to reduce the
human exposure with maximum benefit of whatever radiation is administered.

14.2.1 Monte Carlo simulation in brachytherapy

MC simulation has been used in several studies for brachytherapy (Jangjoo, Ghiasi, & Mesbahi, 2019; Khan, Aziz, &
Koreshi, 2019; Yu et al., 2017). For brachytherapy, radio-isotopes with energy ,50 keV, such as 125I and 103Pd sources
with typical implants of 50
80 metallic seeds encasing isotopes, are used as LDR therapies for the treatment of prostate
cancer, uveal melanomas and brain tumors.
The dose is estimated by a number of computer codes such as EGSnrc, GEANT, PENELOPE, and MCNP based on
MC methods.
In order to enhance the effectiveness of brachytherapy the injection of GNPs through fenestrations of cancer cells is
being considered. The requirements of particle size necessitate the range down to nanoscales (B1029 m) which com-
pares with the diameter of an atom (B10210 m). For MC simulations in brachytherapy, it has been demonstrated that
the dose enhancement factor (DEF) depends on the source energy and concentration of GNP solution, while the size of
GNPs, according to some research results, is not significant above the K-edge energy (Chatterjee et al., 2013; Jain
et al., 2012; Jangjoo et al., 2019; Lechtman et al., 2011; Mesbahi, Jamali, & Gharehaghaji, 2013; Sharabiani, Vaez-
zadeh, & Asadi, 2016).
Several studies have thus used a homogenous model with considerable savings on the computational effort required
for a full heterogeneous model, is used to extract crucial information on the DEF. Calculations are performed for the
dose resulting from the presence of gold in small concentration, MC codes such as MCNP come with a capability of
modeling very detailed heterogeneous configurations so this assumption is not a limitation of MC simulations.
The dose delivered to a prostate tumor. For example, by radio-isotopes 125I, 103Pd, and 131Cs are assessed particu-
larly in the context of dose enhancement in the presence of GNPs.
 Monte Carlo simulation in medical physics Chapter | 14 513

The radio-isotope cesium-131 (131Cs) with ðE 5 29 2 30:4keV; t1=2 5 9:7dÞ is found to yield the required dose in the
least less time compared with 125I ðE 5 35:49keV; t1=2 5 59:4dÞ and 103Pd ðE 5 20:8keV; t1=2 5 17dÞ due to its shorter
half-life and higher energy.
Typically, iodine 125I is used with a radiation dose of 145 Gy or more in accordance with the prescribed dose of
145 Gy suggested by the American Association of Physicists in Medicine Task Group 64 (Pons-Llanas et al., 2018; Yu
et al., 1999).
One of the main advantages of 131Cs is that it offers an initial dose rate of B32 cGy/h at the periphery which is 1.5
and 4 times higher than that from 103Pd and 125I, respectively. This initial dose rate advantage is a vital radiobiological
parameter for a tumor that is growing at a high speed in comparison with slow-growing tumors such as prostate
adenocarcinomas.
The MC code MCNP5 has been used (Khan et al., 2019) to estimate the rose distribution in a coupled photon-
electron simulation in the range 1 keV
100 MeV for prostate tumor brachytherapy. The radio-isotope sources consid-
ered are 125I, 103Pd, and 131Cs in the form of “seeds” modeled as point sources.
As shown in Fig. 14.2, 98 125I, seeds each of activity 0.31 mCi and 115 103Pd seeds each of activity1.4 mCi were
considered spatially distributed in tumor tissue. The placement of needles considers a number of factors including the
location of vital organs surrounding the prostate, such as the bladder, urethra and rectum. For simulating the effect of
131
Cs, the 103Pd seeds were replaced by 131Cs seeds of the same initial activity and spatial distribution. The energy
deposition track length and the pulse height tallies are both used for estimating energy deposition to get reliable esti-
mates in case of a few interactions in a region of interest. The simulation is repeated with gold-tissue solution. The pho-
ton and electron data for both tissue and gold are based on Evaluated Nuclear Data File ENDF/B-VI (Release 8).
Several MC codes have been used and compared (Šı́dlová & Trojek, 2010) on the basis of calculating bremsstrah-
lung, energy deposition in matter, electron ranges and production of secondary electrons by gamma radiation.
The activity of a radio-isotope is AðtÞ 5 Ao e2λt and the number of transformations in an interval τ 5 t2 2 t1 gives an
estimate for the absorbed dose
Ao NE
Abs Dose 5 1:6 3 10210 Gy
λ m
where Ao is the initial activity of a radio-isotope, N is the number of radio-isotope seeds, and E is the energy pulse
height tally in MeV for a tumor of mass m grams.
Thus, the absorbed dose varies directly with the energy deposition and source, which in turn depends on the number
of transformations, that is, the product of initial activity and half-life. The longer decay time for 125I while faster decay
rates of 103Pd and 131Cs give the number of transformations (integrated over time) of 2.74 3 1014, 7.84 3 1013, and
4.47 3 1013 for 125I, 103Pd, and 131Cs. Thus, iodine has a slower build-up, but a higher dose from 125I after B6 months
compared with the much faster effects of 103Pd and 131Cs. The absorbed dose calculated from a MC simulation is
shown in Fig. 14.3 showing the extent of dose enhancement resulting from GNPs. This makes it possible to achieve an
absorbed dose as high as about 300 Gy with a GNP assisted brachytherapy.

103
FIGURE 14.2 Placement of 29 needles of 115 seeds of Pd/131Cs in
the x-y plane.
514 Nuclear Engineering

FIGURE 14.3 Absorbed dose (Gy) for 125I, 103Pd and 131Cs.

Iodine 125I starts with an activity 0.31 mCi, with a distribution shown in Fig. 14.2 while 103Pd and 131Cs both have
initial activity 1.4 mCi. The results indicate that 131Cs is clearly the best in terms of delivering the highest and fastest
dose to the tumor, reaching 50 Gy in the first 10 days and 100 Gy when GNPs in a solution of 25 mg/g tissue are
injected in the tumor while on a longer timescale, 125I gives the highest dose.
The initial dose 131Cs over the first B300 hours is the highest exceeding 5.7 times and 1.5 times that of 125I and
103
Pd respectively in the initial period and falling gradually about 300 hours for 103Pd and about 400 hours for 125I.
In MCNP the “detailed physics” simulation incorporates coherent (Thomson) scattering and fluorescent photons pro-
duced from photoelectric absorption. Electrons produced from photon collisions are transported in a “condensed his-
tory” method that accumulates the effects of many individual collisions into single steps sampled probabilistically.
The effects of such artifacts for electron transport have been investigated by inter-comparisons (Section 14.2) with
other MC codes, such as EGSnrc, GEANT and PENELOPE codes.
In some cases “large discrepancies” ( . 3%) have been found between MCNP5 dose distributions and the “reference
codes” concluding that MCNP5 electron transport calculations are not accurate at all energies and in every medium by
general clinical standards (Almansa, Guerrero, Al-Dweri, Anguiano, & Lallena, 2007; Archambault & Mainegra-Hing,
2015; Koivunoro et al., 2012). It can thus be anticipated that MCNP5 may differ due to its inadequate low-energy treat-
ment of electron transport. The differences with are reported to have been reduced with improved electron transport in
MCNP6.
These results are in line with the decay rates of the three isotopes which favor 125I in terms of energy but result in a
slow dose delivery. The dose enhancement with gold GNP-tissue found in these simulations is the effect of both source
energy of each radio-isotope and the solution concentration. The photoelectric effect plays a dominant for high-Z mate-
rials, such as gold, for which the cross-section varies as μPE BρZ 3 =E3 so that low energy and high-Z are desirable for
dose enhancement which is localized to the tumor due to the short range of photoelectrons and Auger electrons in the
surrounding medium which for electrons of energy 0.1 MeV is B100 microns in water and B15 microns in gold. Thus
GNPs are used in thin layers of , 100 nm thickness to utilize the energy of photoelectrons in water.

14.2.2 Monte Carlo simulation to calculate energy deposition and dose distribution for
brachytherapy
In continuation of the previous section, the dose distribution and the subsequent DEF by the use of gold, a high-Z bio-
compatible element, in solution are both estimated by MC codes as a function of source energy typical of brachytherapy
sources (40 keV
1 MeV), solution concentration (5
25 mg Au/g H2O) and solution placement (1
2 cm concentric
shells).
Results from the MC simulation code MCNP5 are compared with other widely used MC codes such as PENELOPE
and GEANT, to validate the dose estimates which may vary considerably due to artifacts and data libraries. MC pertur-
bation estimates for radiation oncology, are yet to be used extensively to carry out sensitivity studies which can be used
to obtain optimal experimental parameters such as radiation energy, concentration of gold in solution and the optimal
 Monte Carlo simulation in medical physics Chapter | 14 515

distribution of material for maximizing an objective function of interest. MC simulation, using MCNP5, has been car-
ried out for simulating coupled photon-electron radiation transport from X-rays emanating from a radiation source
implanted in a cancer cell, modeled by a spherical water phantom, to estimate the energy deposited and the subsequent
DEF using a water sphere of radius 15 cm to represent tissue.
Since the dose enhancement is mainly due to photoelectron production from gold, the use of water for soft tissue, as
defined by the International Committee for Radiological Protection (ICRP) is justifiable due to similar density and pho-
ton interaction cross-sections. For carrying out sensitivity studies, the MC perturbation feature, with material perturba-
tions, was used to sample derivatives in a single run which were used in a Taylor series to estimate subsequent dose
and enhancement.
From the validation studies, good agreement is found between MCNP5 yields dose estimates, distributions and
enhancements which values reported for the MC codes PENELOPE and GEANT with lower energy cutoffs for electron
transport as compared with MCNP5. While the dose increases with source energy, the DEF was found to increase
inversely with the source energy, for a given concentration, achieving a value as high as 1.8
2.5 for a source of
40 keV and concentration 5
10 mg Au/g H2O.
This enhancement was found to occur at energies near the K- and L-shell electrons of gold and no significant
enhancement was found at higher (MeV) energies. The validity of first- and second-order perturbation theory was con-
firmed for small changes (B2%) in the material density of gold-water solution enabling a single run to estimate the
DEF from concentrations in the range 5
25 mg Au/g H2O with a considerable computational speedup.
For calculations for dose distribution MC methods and general purpose codes, such as EGSnrc, GEANT,
PENELOPE, and MCNP have been extensively used for simulating the transport of radiation from radiation seeds in
the medium consisting of tissue, and material in the vicinity of the cancer. These codes have produced fast and accurate
results which have been experimentally validated and benchmarked.
At the small scales of nanotechnology, comparable with the atomic scale, the behavior of materials and subsequently
their electrical and thermal properties also depend on size and shape. Nanotechnology, driven largely by the opportu-
nities in electronics and semiconductors, has emerged as one of the frontiers of science capable of revolutionizing tech-
nology in areas including communications and computing, materials and medicine.
The source energy of interest is 40 keV
1 MeV typical of brachytherapy sources, while the gold particles in solu-
tion form with water were assumed to comprise a homogeneous mixture with concentrations ranging form 5
25 mg
Au/g H2O. While MCNP has the capability of modeling very detailed heterogeneous configurations, this paper consid-
ers a homogeneous model solely for the purpose of demonstrating (1) results from benchmarked problems and (2) for
obtaining MC sensitivity estimates for a “bulk” material to demonstrate validity of MC perturbation and quantifying
increased computational efficiency.
MC perturbation analysis has applications in brachytherapy for estimating dose perturbations when implements are
present in the vicinity of an organ receiving radiation from an implanted source (see e.g., Yu et al., 1999). However, it
is yet to be used to simulate brachytherapy studies and can provide great computational efficiency leading to optimal
designs based on “best” experimental parameters such as radiation energy, concentration of gold in solution and mate-
rial placement for maximizing an objective function of interest.
The MC code MCNP5 is used to carry out a coupled photon-electron simulation of radiation transport in the range
1 keV
100 MeV to estimate the dose distribution, from a point isotropic photon source typical of brachytherapy
sources (30 keV for 125I 21.25 MeV for 60Co) located at the origin of a sphere, in concentric shells of water of thick-
ness 0.5 mm up to a radius of 15 cm. A solution of gold and water is then considered in the shells located 1
2 cm from
the center and the F6 and *F8 tallies are used for phantom dosimetry. Photon and electron data for air, water and gold
are based on ENDF/B-VI (Release 8). In the MCNP “detailed physics” simulation, coherent (Thomson) scattering is
included and fluorescent photons produced from photoelectric absorption are included and electrons produced from
photon collisions are transported in a “condensed history” method that accumulates the effects of many individual colli-
sions into single steps sampled probabilistically.
Perturbation algorithms in MC simulation, developed extended the capability of MC methods to sensitivity studies
and optimization. The change in a response function such as dose D, due to a variation in an independent parameter,
such as material density ρ, such as the dose, expressed as a Taylor series
  0  1
DðρÞ 5 D ρo 1 D ρo δρ 1 Dv ðδρÞ2 1 ?
2!
0
can be used with first- and second-order derivatives D and Dv from a single run. Thus, when the change in a parameter
is small enough to be of the order of the statistical uncertainty of a MC estimate, then the difference from two
516 Nuclear Engineering

independent runs may mask the actual difference. The energy deposition for a “perturbed” design ðDnew Þ can thus be
estimated from a Taylor series in terms of a “reference” design ðDref Þ.
The material used in this work is water which has density ρ 5 1:04 g/cm3 very close to that of soft tissue ρ 5 1:04 g/
3
cm (with a four-component simplified composition hydrogen (H), carbon (C), nitrogen (N) and oxygen (O) with
weight percentages: 10.454%, 22.663%, 2.49%, 63.525%, respectively) and the photon electric cross-sections for both
are almost indistinguishable so that conclusions drawn for energy deposition would hold for both materials to a reason-
ably acceptable order. While water phantoms are used for representing configurations similar to soft tissue, full anthro-
pomorphic phantoms are also used to carry out detailed simulation for the dose.
To compare MCNP5 results for energy deposition from a mono-energetic point isotropic source at the center of a
sphere of radius 15 cm, the F6 tally in 0.5 mm thick concentric shells gives the energy distribution. When normalized
to a reference dose from a 1 MeV source, it is seen that the lowest energy, 15 keV, has an energy deposition about 65%
of the reference which falls off rapidly to less than 10% in 1.5 cm of water. With a further increase in energy, the rela-
tive intensities drop but the deposition sustains to greater depths which is less important in brachytherapy than to give a
localized deposition from a low-energy source.
The DEF shown in Fig. 14.4 gives the extent and magnitude as a function of source energy and concentration for
E 5 40 keV, 50 keV, 90 keV, 1 MeV for varying concentrations 5, 10, 25 mg Au/g H2O confirming the trend of direct
variation of DEF with concentration and inverse variation with source energy. The spatial effect of high energy photons
extends further in the medium as compared with low-energy photons as observed from results in Fig. 14.5 when the
1 MeV photons are able to persist to longer distances with a rising slope for GNP size.
MC simulations have also been used for ophthalmic brachytherapy with 125I with mean energy 35.49 keV with a
full three-dimensional heterogeneous model simulating 107
109 histories for 50 nm GNPs (Asadi et al., 2015) estimat-
ing the DEF for 7, 10, 18 and 30 mg GNP concentration/g as varying from 1.9, 2.2, 3.2 to 4.6 for the tumor phantom
compared with 1.9, 2.3, 3.3 and 4.8 in the water phantom, respectively, which indicates that the tumor and water phan-
toms are very similar.
Preliminary MC simulations with a homogenous mixture model for a water phantom with 106 simulations gives, for
40 keV and concentrations of 5, 10, 25 mg/g maximum DEFs of 1.8, 2.5. 4.6.
Mesbahi et al. (2013) report DEF from 1.4 to 3.7 with the highest DEF for 90 keV; they also concluded, as did
Lechtman et al. (2011), that the effect of GNP size was not considerable while concentration and energy were impor-
tant. Of their simulations carried out for 7 and 18 mg Au/g H2O for 30, 50, 100 nm GNPs and energies between 50 to
120 keV , they reported highest DEF of 3.5 at 90 keV followed by 3.0 at 50 keV both for 30 nm GNPs and 18 mg/g
concentration.
For high energy sources (Banoqitah & Djouider, 2016) a maximum DEF of 1.45 is found within the tumor when
implanted with 70 mg/g Gd for the lowest energy (192Ir). Thus MC simulations find that the DEF varies directly with
concentration and inversely with source energy, for example, for a concentration of 30 mg/g, it decreases from 1.24 to
1.09 when the photon energy increases from 0.380 MeV (192Ir) to 1.20 MeV (60Co).

FIGURE 14.4 Energy deposition for mono-energetic photons of

energy E: 15 keV
1 MeV.
 Monte Carlo simulation in medical physics Chapter | 14 517

FIGURE 14.5 Dose enhancement factor (DEF) versus distance.

When full heterogeneities are modeled, MC sensitivity analysis can be useful for calculating sensitivity coefficients
for the design optimization of GNP size with varying concentration.

Nomenclature
English
m mass
t time
D dose
E energy
H heating number (MeV/collision)
N number of particles simulated
V volume
Z atomic number

Greek
μPE attenuation coefficient (photoelectric)
ρ density
σt microscopic total cross-section
φ scalar flux scored over volume (track length/volume) cm22
Ω solid angle

Abbreviations
BAEC Bangladesh Atomic Energy Commission
DEF dose enhancement factor
GNP gold nanoparticle
Gy gray
HDR high dose rate
LDR low dose rate
TRIGA Training Research Isotopes General Atomics

References
Almansa, J. F., Guerrero, R., Al-Dweri, F. M. O., Anguiano, M., & Lallena, A. M. (2007). Dose distribution in water for monoenergetic photon point sources
in the energy range of interest in brachytherapy: Monte Carlo simulations with PENELOPE and GEANT4. Radiation Physics and Chemistry.
Archambault, J. P., & Mainegra-Hing, E. (2015). Comparison between EGSnrc, Geant4, MCNP5 and Penelope for mono-energetic electron beams.
Physics in Medicine and Biology.
518 Nuclear Engineering

Asadi, S., Vaez-zadeh, M., Masoudi, S. F., Rahmani, F., Knaup, C., & Meigooni, A. S. (2015). Gold nanoparticle-based brachytherapy enhancement
in choroidal melanoma using a full monte carlo model of the human eye. Journal of Applied Clinical Medical Physics/American College of
Medical Physics.
Banoqitah, E., & Djouider, F. (2016). Dose distribution and dose enhancement by using gadolinium nanoparticles implant in brain tumor in stereotac-
tic brachytherapy. Radiation Physics and Chemistry.
Bavarnegin, E., Khalafi, H., Sadremomtaz, A., & Kasesaz, Y. (2016). Construction of a head phantom for mixed neutron and gamma field dosimetry
in TRR. Measurement: Journal of the International Measurement Confederation.
Bielajew, A. F., & Rogers, D. W. O. (1989). Lecture notes: Electron Monte Carlo simulation. Canada: National Research Council of Canada Ottawa.
Blanchard, P., et al. (2018). Patient-reported health-related quality of life for men treated with low-dose-rate prostate brachytherapy as monotherapy
with 125-iodine, 103-palladium, or 131-cesium: Results of a prospective phase II study. Brachytherapy.
Chatterjee, D. K., et al. (2013). Convergence of nanotechnology with radiation therapy-insights and implications for clinical translation. Translational
Cancer Research.
Chatzipapas, K. P., et al. (2019). Quantification of DNA double-strand breaks using Geant4-DNA. Medical Physics.
Chatzipapas, K. P., et al. (2020). Ionizing radiation and complex dna damage: Quantifying the radiobiological damage using monte carlo simulations.
Cancers.
Darda, S. A., Soliman, A. Y., Aljohani, M. S., & Xoubi, N. (2020). Technical feasibility study of BAEC TRIGA reactor (BTRR) as a neutron source
for BNCT using OpenMC Monte Carlo code. Progress in Nuclear Energy.
Dymova, M. A., Taskaev, S. Y., Richter, V. A., & Kuligina, E. V. (2020). Boron neutron capture therapy: Current status and future perspectives.
Cancer Communications.
Engels, E., et al. (2020). Advances in modelling gold nanoparticle radiosensitization using new Geant4-DNA physics models. Physics in Medicine and
Biology.
Hassanein, A. M., Hassan, M. H., Mohamed, N. M. A., & Abou Mandour, M. A. (2018). An optimized epithermal BNCT beam design for research
reactors. Progress in Nuclear Energy.
Hirayama, H., et al. (2005). The EGS5 code system.
Jain, S., Hirst, D. G., & O’Sullivan, J. M. (2012). Gold nanoparticles as novel agents for cancer therapy. British Journal of Radiology.
Jangjoo, A. G., Ghiasi, H., & Mesbahi, A. (2019). A Monte Carlo study on the radio-sensitization effect of gold nanoparticles in brachytherapy of
prostate by 103Pd seeds. Polish Journal of Medical Physics and Engineering.
Kasesaz, Y., Khalafi, H., & Rahmani, F. (2014). Design of an epithermal neutron beam for BNCT in thermal column of Tehran research reactor.
Annals of Nuclear Energy.
Khan, H., Aziz, U., & Koreshi, Z. U. (2019). Radiation dose delivered by 125I, 103PD and 131CS and dose enhancement by gold nanoparticle (gnp)
solution in prostate brachytherapy: A comparative analysis by monte carlo simulation. IIUM Engineering Journal.
Koivunoro, H., Siiskonen, T., Kotiluoto, P., Auterinen, I., Hippeläinen, E., & Savolainen, S. (2012). Accuracy of the electron transport in mcnp5 and
its suitability for ionization chamber response simulations: A comparison with the egsnrc and penelope codes. Medical Physics.
Lazarine, A. D., & Goorley, T. (2005). Medical physics calculations with MCNPTM: A primer. Transactions of the American Nuclear Society.
Lechtman, E., Chattopadhyay, N., Cai, Z., Mashouf, S., Reilly, R., & Pignol, J. P. (2011). Implications on clinical scenario of gold nanoparticle radio-
sensitization in regards to photon energy, nanoparticle size, concentration and location. Physics in Medicine and Biology.
Lechtman, E., et al. (2013). A Monte Carlo-based model of gold nanoparticle radiosensitization accounting for increased radiobiological effectiveness.
Physics in Medicine and Biology.
Li, W. B., et al. (2020). Intercomparison of dose enhancement ratio and secondary electron spectra for gold nanoparticles irradiated by X-rays calcu-
lated using multiple Monte Carlo simulation codes. Physica Medica.
Matsuya, Y., Fukunaga, H., Omura, M., & Date, H. (2020). A model for estimating dose-rate effects on cell-killing of human melanoma after boron
neutron capture therapy. Cells.
Mesbahi, A., Jamali, F., & Gharehaghaji, N. (2013). Effect of photon beam energy, gold nanoparticle size and concentration on the dose enhancement
in radiation therapy. BioImpacts.
Monshizadeh, M., Kasesaz, Y., Khalafi, H., & Hamidi, S. (2015). MCNP design of thermal and epithermal neutron beam for BNCT at the Isfahan
MNSR. Progress in Nuclear Energy.
OECD/NEA. (2019). The supply of medical isotopes: An economic diagnosis and possible solutions.
Pelowitz, D. B., et al. (2013). MCNP6 user’s manual. La-Cp-13
00634.
Perez-Calatayud, J., et al. (2019). GEC-ESTRO ACROP recommendations on calibration and traceability of LE-LDR photon-emitting brachytherapy
sources at the hospital level. Radiotherapy and Oncology: Journal of the European Society for Therapeutic Radiology and Oncology.
Pons-Llanas, O., et al. (2018). Permanent seed implant brachytherapy in low-risk prostate cancer: Preoperative planning with 145 Gy vs real-time
intraoperative planning with 160 Gy. Reports of Practical Oncology and Radiotherapy.
Reed, A. (2007). Medical physics calculations with MCNP: A primer. Boston, MA: Los Alamos Natl. Lab. X-3 MCC.
Rice, S. R., et al. (2019). A phase II study evaluating bone marrow-sparing, image-guided pelvic intensity-modulated radiotherapy (IMRT) with
Cesium-131 brachytherapy boost, adjuvant chemotherapy, and long-term hormonal ablation in patients with high risk, nonmetastatic prostate can.
American Journal of Clinical Oncology: Cancer Clinical Trials.
Rogers, D. W. O. (2006). Fifty years of Monte Carlo simulations for medical physics. Physics in Medicine and Biology.
Sato, T., et al. (2015). Overview of particle and heavy ion transport code system PHITS. Annals of Nuclear Energy.
 Monte Carlo simulation in medical physics Chapter | 14 519

Shaaban, I., & Albarhoum, M. (2015). Design calculation of an epithermal neutronic beam for BNCT at the Syrian MNSR using the MCNP4C code.
Progress in Nuclear Energy.
Sharabiani, M., Vaez-zadeh, M., & Asadi, S. (2016). Size dependence of GNPs dose enhancement effects in cancer treatment
Geant4 and MCNP
code. Radiotherapy and Oncology: Journal of the European Society for Therapeutic Radiology and Oncology.
Šı́dlová, V., & Trojek, T. (2010). Testing Monte Carlo computer codes for simulations of electron transport in matter. Applied Radiation and Isotopes:
Including Data, Instrumentation and Methods for use in Agriculture, Industry and Medicine.
Yu, E., Lewis, C., Luisa Trejos, A., Patel, R. V., & Malthaner, R. A. (2011). Lung cancer brachytherapy: Robotics-assisted minimally invasive
approach. Current Respiratory Medicine Reviews.
Yu, P. C., et al. (2017). CT-based MCNPX dose calculations for gynecology brachytherapy employing a Henschke applicator. Radiation Physics and
Chemistry, 140, 392
397.
Yu, Y., et al. (1999). Permanent prostate seed implant brachytherapy: Report of the American Association of Physicists in Medicine Task Group No.
64. Medical Physics.
Zalutsky, M. R., & Pruszynski, M. (2012). Astatine-211: Production and availability. Current Radiopharmaceuticals.
Index

Note: Page numbers followed by “f,” “t,” and “b” refer to figures, tables, and boxes, respectively.

A reaction rates in an AP1000 unit cell, 395t Bare homogeneous nuclear reactor core, 490
Absorbed dose (D), 26 27 unit lattice cell of, 390 394 Bare sphere, criticality in, 417 421
Absorption estimator (AE), 311, 382 AP1000 fuel atomic densities, 392t Bateman equations, 173 174
Accelerator-based sources, 25 26 AP1000 IFBA atomic densities, 393t BE. See Binding energy (BE)
Acceptance-rejection sampling scheme, 449, AP1000 water atomic densities, 393t Beginning of cycle (BOC), 394, 495
452 atomic densities and volumes of cells in Beginning of Life (BOL), 394
Actinium series, 24 AP1000 unit lattice, 392t Beryllium (Be), 72 73, 104, 404, 421, 426,
Adaptive mutation rate operator, 495 Approximate methods, 185 186 462
Adjoint equation, 357 358, 360 Rayleigh Ritz variational method, 186 Beryllium oxide, 104
Adjoint function, 186 187 Ritz method, 185 Bessel equation, 161 162
Advanced boiling water reactor (ABWR), 105 weighted residual method, 186 Beta decay, 9
Advanced heavy water reactor (AHWR), AR. See Aerojet Rocketdyne (AR) radioactive decay
112 113 Artificial Intelligence (AI), 493, 502, 504 of cesium-137, 10f
Advanced Submarine Fleet Reactor (ASFR), 116 AI data-driven models, 374 of iodine-13, 10f
Aerojet Rocketdyne (AR), 136 AI-based meta-heuristic schemes, 500 Beta radiation with matter, interaction of,
“Age” theory, 71 AI-based meta-heuristic stochastic methods, 16 19
ALARA. See As Low As Reasonably 374 375 stopping power for beta particles, 17f, 19t
Achievable (ALARA) Artificial Neural Networks (ANN), 497 498 stopping power from ESTAR NIST, 17t
Algebraic constraints, deterministic As Low As Reasonably Achievable (ALARA), stopping power of electrons, 19f
optimization with, 351 352 511 Bethe-Bloch formula, 13 14
constrained maximization of volume of ASFR. See Advanced Submarine Fleet Reactor Bhabha Atomic Research Center (BARC),
cylinder, 353t (ASFR) 112 113
Algebraic equations, 500 Asymptotic relaxation length, 269 Bimodal nuclear thermal rocket engines, 136
Alpha Atom, 1 6 Binding energy (BE), 4 6, 6f
decay, 8 9 binding energy, 6 Biological shielding systems, 337
emitters, 380 Bohr’s model of, 2f Birth-to-death process, 305 306, 308 309
particle transport simulation, 35 fundamental physical constants, 2t Blanket/shield design optimization studies, 494
radiation therapy, 511 512 nuclear stability, 5 BN method, 295 296
rays interaction with matter, 12 16 Atom, mathematical modeling and simulation Bohr model, 3
stopping power and range of alpha, 14t of, 34 44 Bohr radius, 3
stopping power from ASTAR NIST, 16t Atomic density, 242 Bohr-Rutherford model, 1
Aluminum (Al), 38, 421, 425 426 of elements and mixtures, 30 34 Boiling water reactor (BWR), 103, 110 112
American Association of Physicists in of fuel, 220 advanced boiling water reactor, 112t
Medicine Task Group 64, 513 Atomic fractions, 33 34 physical design parameters of BWR, 111t
American National Standards Institute (ANSI), Atomic number density, 31 Boltzmann equation, 211
388 Atomic ratio, 220 Boltzmann transport equation (BTE), 34, 51,
Americium-beryllium (Am-Be), 25, 501 Atoms, 58 259, 338
Analog simulation, 332 Avogadro’s number, 3, 31 Boltzmann Transport Fokker-Planck equation
Ananisotropic 1 MeV mono-energetic source, (BTFPE), 34
426 Boric acid, 107
Angular deflections, 37 B Boron (B), 107, 324, 425
Angular flux, 260 261, 273 274 BAEC TRIGA reactor (BTRR), 512 Boron-10, 60 61
Anisotropic SN method (ANISN), 338 339, 489 Bang-bang control, 373 Boron carbide (B4C), 30, 139
ANISN-ORNL, 338 Bang-bang solution, 360 Boron neutron capture therapy (BNCT), 12,
ANN. See Artificial Neural Networks (ANN) BARC. See Bhabha Atomic Research Center 374, 511, 511f
Ant colony optimization, 373 374 (BARC) Brachus, 509
AP1000 reactor Bare assembly, 386 Brachytherapy, 509, 512 517
fissions and multiplication in an AP1000 Bare critical assemblies, 381 388 Monte Carlo simulation in, 512 514
unit cell, 396t Godiva, 381 386 to calculate energy deposition and dose
fuel assembly of, 390 394 Jezebel, 386 388 distribution for, 514 517

521
522 Index

“Breeding” plutonium, 114 Collisions, 267 268, 382 of nuclear fission, 81 83, 83t
Breit-Wigner model, 60 density, 68, 267 Critical half-thickness, 240
Breit Wigner formula, 77 parameters, 22 Critical heat flux (CHF), 374
Bremsstrahlung photons, 42 stopping power, 13 Critical radius, 419
Broglie wavelength, 3 Compound nucleus, 74 Critical reactor flux, 218
Brownian motion, 188 Compton scattering, 22 23, 40 Criticality, 88 97, 379, 464
Computational radiobiology, 510 511 in bare sphere, 417 421
Computer codes, 512 one-group diffusion theory criticality,
C neutron and radiation transport codes, 417 418
CADIS. See Consistent Adjoint Driven 338 340 one-speed transport theory criticality,
Importance Sampling (CADIS) ANISN, 338 419 421
Calcium, 3 4 DOT, 338 two-group diffusion theory criticality,
Calder Hall reactors, 113 KENO, 340 418 419
Califorium-252, 25, 389, 501, 503 MCNP, 339 condition, 240
radiation dose from Calfornium-252 gamma MORSE, 339 340 diffusion theory, 90 91
source, 41 44 PARTISN, 339 equation, 241, 246, 418
Californium Cf252, 25 TART, 339 estimates method, 297 298
Cancer, 509 TORT, 338 Monte Carlo simulation, 92 97
Carbon (C), 60 61, 516 nuclear data, 341 344 safety, 388 389
nucleus, 3 in nuclear systems, 338 atom density of UF6, 389t
Carbon dioxide, 106 performance and safety analyses, 341 storage of interacting units, 388
Carbon-14 radiological protection codes, 341 storage of uranium hexafluoride cylinders,
beta decay of, 9 thermal hydraulics codes, 340 388 389
CASMO (Lattice codes), 343 time-dependent reactor kinetics codes, 340 transport theory, 91 92
CDC. See Control Data Corporation (CDC) Computer programming, 225 227 uranium-235, 90t
CDE. See Collision density estimator (CDE) Computer systems, 337 Cross-section
CDFs. See Cumulative distribution functions Computerized Tomography (CT), 510 macroscopic, 63 64
(CDFs) “Condensed history” method, 36, 514 515 microscopic, 58 62
CE. See Collision estimator (CE) Conservation equation, 211 213 resonance, 74 80
Center of mass systems (CM systems), volume element, 212f CSDA. See Continuous Slowing-Down
55 56 Consistent Adjoint Driven Importance Approximation (CSDA)
neutron scattering in lab and, 55 58 Sampling (CADIS), 403 CT. See Computerized Tomography (CT)
Central fuel temperature (CFT), 497 Constraints, deterministic optimization without, Cuckoo optimization, 373
Central limit theorem, 200 201, 306 307, 350 351 Cumulative distribution functions (CDFs), 187,
460, 462 CONT record, 343 449
Matlab program, 201 Continuous Slowing-Down Approximation Cumulative fission yield, 84
Ceramic titanates, 405 (CSDA), 35 Cylinder volume optimization, 367
Ceramic-metallic (CERMET), 136 Contributions, 332 Cylindrical reactor, 216 218
CERMET. See Ceramic-metallic (CERMET) Control Data Corporation (CDC), 337, 350
Cesium 137Cs, 11 Control rods (CR), 492
CFT. See Central fuel temperature (CFT) Controller design optimization, 372 D
Chemotherapy, 509 Controller proportional integral derivative, 374 D2O. See Heavy water (HW)
CHF. See Critical heat flux (CHF) Conventional rocket designs, 132 DDR. See DNA Damage Response (DDR)
China Fusion Engineering Test Reactor, physical design, 134t De Broglie wavelength, 54
493 494 Copper, 403 404 Death of neutron, 305 306
Chinese Evaluated Nuclear Data Library Core neutronics, 380 “Decay chain”, 7
(CENDL), 341 Core neutronics with diffusion equations, DEF. See Dose enhancement factor (DEF)
Choleski’s method, 238 251 256 Delta function, 169 170
Chromosomes, 367 Core-reflector system, 228 Dementia, 510
CITATION codes, 338 Coupled first-order ODEs, 155 156 Demonstration Fusion Power Plant, 493 494
Classic albedo Coupled kinetic-dynamics, 172 173 Demonstration Fusion Power Reactor (DEMO
calculation, 421 423, 423t Coupled neutronic-TH analysis, 491 reactor), 310
problem, 270 272 Coupling Deoxyribonucleic acid (DNA), 509
Closest shell, 3 4 GA models, 495 Department of Energy, 388
CM systems. See Center of mass systems (CM of neutronics, 340 Department of Transportation, 388
systems) Cramer’s rule, 229, 500 Departure from Nucleate Boiling (DNB), 497
Co-state equation, 357 358 Cray supercomputers, 337 Departure from Nucleate Boiling Ratio
Cobalt (60Co), 11 Critical assemblies, 388 (DNBR), 497
COBRA code, 340 bare, 381 388 Depleted UF6, 380
Coherent scattering, 22, 515 benchmarks for, 388 Derivative (D), 362
Collision density estimator (CDE), 428 Critical core with flat thermal flux loading, Design and operation of nuclear reactors, 333
Collision estimator (CE), 309, 311, 382, 425 247 248 Design optimization, 349, 373
Collision probability method (CPM), 259, 299 Critical energy, 16 “Detailed physics” simulation, 514 515
Collisional equilibrium, 425 energy released in nuclear fission reaction, 83t Detector shielding design, 374
 Index 523

Deterministic codes, 339 Electronic computers, 337 Finite non-multiplying sphere, MATLAB
Deterministic methods, 298 299, 350 Electronic energy loss, 13 program for point source in, 456 459
Deterministic optimization method, 365 Element matrices, 182 183 Finite slab, 423
Deterministic problems, 365 Elementary 1-D models, 404 Finite sphere with point isotropic source,
Differential Evolutionary Algorithms (DEA), 374 Elementary diffusion, 474 476 274 284
“Diffusion approximation”, 419 Emergency Planning Zone (EPZ), 124 125 neutron flux in finite sphere, 286t
Diffusion coefficient, 227, 418 Emitted particle, 55 transport theory flux in, 285f
Diffusion equations, core neutronics with, End of cycle (EOC), 493 Finite Volume Method (FVM), 174, 211
251 256 Energy deposition, 383 First controlled critical pile (CP-1), 116
Diffusion process, 67 73 Monte Carlo simulation to energy deposition First wall (FW), 310, 380
Diffusion theory (DT), 90 91, 269, 338, 374, calculation, 514 517 First-order ordinary differential equation
417, 453 Energy loss, 14 16 constraints, 352 360
CITATION, 338 of electron, 37 Fissile materials, 85 86
flux index, 37 Fissile nuclides, 85
in finite sphere with point isotropic Energy-dependent flux, 68, 384 Fission, 5, 85
source, 428 429 Energy-dependent neutron flux, 92, 331 Fission energy
in slab, 423 424 Energy-integrated albedos in water, 422 deposition, 383
Discrete optimal control solution, 360 Energy-integrated neutron flux, 472 spectrum, 312
Discrete ordinates method, 34, 285 287, 338 Engineering systems, 361 Fission neutronics, 324, 379 380, 489
quadrature weights and angle, 288t Enrichment plant, 379 Fission neutrons, 72, 394
Discrete-ordinate neutron transport codes, 338 Equivalent dose, 27 28 Fission power systems, 349
Displacement per atom (DPA), 310, 380, Estimators, 309 311 Fission reaction, 314
403 404 Euler-Lagrange equation, 349, 355 Fission spectrum, 86 87
Distance to collision (DTC), 454 Euler’s constant, 161 162 watt fission spectrum, 87t
sampling, 313 European Joint Evaluated Fission and Fusion Fission yield of nuclear fission, 84
Distribution function, 53 (JEFF) Library, 341 Fitness function, 367
sampling from, 449 453 European Pressurized Reactor (EPR), 108 Fixed-source MC simulation, 313
normal distribution, 449 451 Europium, 107 Flat thermal flux (FTF), 238, 245 246
Watt fission spectrum, 451 453 Evaluated Nuclear and Atomic Reaction Data critical core with FTF loading, 247 248
Djikstra’s algorithm, 365 Library (ENDL), 341 Fluid dynamics equations, 156
DNA Damage Response (DDR), 510 511 Evaluated Nuclear Data File, Brookhaven FLUKA (Fluctuating Kaskade) code, 510
DNA Repair mechanisms, 510 511 National laboratory (ENDF/B), 306, Fluorescent photons, 515
DNB. See Departure from Nucleate Boiling 341, 343 Fluorine-18, 509
(DNB) Evaluated Nuclear Structure Data File Flux
DNBR. See Departure from Nucleate Boiling (ENSDF), 6 in finite sphere with point isotropic source,
Ratio (DNBR) Evolutionary Algorithms, 374 428 433
Doppler broadening, 77 Exact flux, 456 diffusion theory, 428 429
Dose distribution calculation, Monte Carlo Expansion coefficient, 280 Monte Carlo simulation, 431 433
simulation to, 514 517 Experimental breeder reactor (EBR-I), 114 transport theory exact solution,
Dose enhancement factor (DEF), 512 Extrapolation radius, 419 430 431
DOT code, 338, 341, 489 flux-current relation Fick’s law, 265 266
DP. See Dynamic Programming (DP) formulation, 164 165
DPA. See Displacement per atom (DPA) F loading determines, 310
DPN method, 293 294 Fast breeder reactors (FBRs), 105, 114 115 measurement, 64 66
DT. See Diffusion theory (DT) BN-600 Sodium cooled fast breeder reactor, transmutations, 65t
DTC. See Distance to collision (DTC) 115t in slab, 423 428
Dynamic Programming (DP), 350, 365 367, Fast neutron, 52 53, 65, 89, 390 comparison, 425 428
374, 489, 496 Fast-spinning centrifuge, 379 diffusion theory, 423 424
distances between nodes, 366t Fedholm integral equation, 165, 168 Monte Carlo simulation, 425
FENDL. See Fusion Evaluated Nuclear Data transport theory, 424 425
Library (FENDL) Flux tally, 470
E “Fermi age equation”, 71 Flux Work Units (FWU), 338
Effective dose, 28 Fertile Force function, 178
weighting factor for organs, 28t materials, 85 86 Former soviet nuclear submarine program, 117
Effective multiplication of system, 88 89 nuclides, 85 4S Gen-IV design, 394
EGS code, 510 Fick’s law, 71, 153, 211 213, 222, 254, 264, 4S reactor, 476
EGSnrc codes, 512, 514 515 266, 433 Fourier transform, 259, 417, 424
Eigenvalue equation, 151, 417 418 Finite 1D slab, 425 Fractional energy loss of neutrons, 67
Elastic scattering cross-section, 54 55 Finite cylinder, 215 FRAPCON, 341
Electromagnetic gamma rays, 19 Finite difference method (FDM), 174 177, FRAPTRAN, 341
Electron(s), 2, 4, 16 17, 514 211, 338 Fredholm equation, 280
interaction data, 41 Matlab program, 177 FTF. See Flat thermal flux (FTF)
interaction of electrons with matter, 35 40 Finite element method (FEM), 174, 178 185, Fuel, 112
transport, 36 211, 296, 337 assembly of AP1000 reactor, 390 394
524 Index

Fuel (Continued) Gen IV Liquid Metal Cooled Fast Reactor, 340 Heterogeneity, 253
concentration effect on critical mass, General Purpose Heat Source (GPHS), 11 HEU. See Highly enriched uranium (HEU)
238 248 Generation of Purdue Macroscopic XS set code Heuristic optimization techniques, 350
critical core with flat thermal flux loading, (GENPMAXS set code), 343 Hexafluoride (UF6), 379
247 248 Generation-IV International Forum (GIF), 126 High Flux Isotope Reactor conversion project,
Goertzel’s theorem, 239 Generation-IV reactors, 125 127 493
slab model, 239 243 Genetic algorithms (GAs), 350, 367 371, High-energy transport codes, 340
spherical model, 244 247 373 374, 489, 491, 496, 504 Highly enriched uranium (HEU), 120, 379
cycle, 496 results for optimized cylinder volume, 372t submarine reactors, 120
depletion code, 338 simulation parameters for genetic algorithm Hooke’s Law, 151 152
depletion, 338 optimization, 368t Hybrid methods, 297
loading pattern optimization, 495 500 Geometry modeling, 316 327 Hydrogen (H), 3 4, 57, 463, 516
optimal distribution, 498 500 for illustration of Monte Carlo simulation,
optimal refueling policy, 497t 320 327
reloading strategy, 374, 496 surface mnemonics, 321t I
Fukushima earthquake, 105 Germanium-based semiconductor detectors, In-core design optimization, 372
Full power year (FPY), 310 29 30 Inboard (IB), 493 494
Full-core simulation, 399 GFR. See Gas-Cooled Fast Reactors (GFR) Independent fission yield, 84
Fusion Evaluated Nuclear Data Library GIF. See Generation-IV International Forum Independent variables, 502
(FENDL), 405 (GIF) Inelastic reaction, 55
Fusion neutronics, 379 380, 489 Global stiffness matrix, 178 Inertial confinement fusion (ICF), 128, 130 132
design optimization, 493 494 Godiva, 381 386, 419 Infinite medium
Fusion power systems, 349 data, 418 with plane isotropic source, 273 274
Fusion reactor, 310, 493 estimates of with collision-, absorption and angular flux in, 275b, 281b
Fusion reactor blanket design, 489 track length estimators, 382t point source in, 215
Fusion system design optimization, 489 gains per source particle, 383t 1-group nonmultiplying media, 216t
Fusion Tokamak systems, 326 losses per source particle, 383t solutions, 225 226
Fuzzy logic, 504 material composition of, 381t Infinite medium equation, 417 418
FVM. See Finite Volume Method (FVM) MC simulation parameters, 381t Inorganic scintillators, 30
FWU. See Flux Work Units (FWU) simulation, 386 Integral (I), 362
spectrum, 386 evaluation of, 201 206
tallies for, 384t Integral equation, 69, 165 170, 425
G Goertzel’s theorem, 239 classification of, 165t
Gadolinium, 107, 425 Gold (Au), 425, 514 important integral equation for neutron
GA-MCNP optimization, 494 495 Gold nanoparticles (GNP), 510 511 transport, 169
Gamma camera, 510 Goudsmit-Saunderson model, 42 integral equations in neutron transport,
Gamma decay, 10 11 GPHS. See General Purpose Heat Source 169 170
Gamma dose rate, 26 (GPHS) for neutron transport, 169 170
Gamma radiation, 380, 501, 510 GPR. See Gaussian Process Regression (GPR) Integral form of transport equation, NTE,
interaction with matter, 19 24 Graphite, 390, 421, 426, 462 266 268
mass attenuation coefficients, 24t Green’s function, 265, 417 Integral fuel burnable absorber (IFBA), 391
photoelectric effect, 20f Grey Wolf optimization, 373 Integral transport equation, 253
shell energies, 21t Ground scattering, 389 multigroup form of, 268
with matter, interaction of, 40 41 Group-averaged neutron flux, 265 Integrated flux, 53
shielding, 494 Guard vessel (GV), 121 124 Integro-differential equations, 170 174, 265
Gamma rays, 20, 509 Integro-differential form, 260
from fission, 83 of neutron transport equation, 260 265, 260f
Gamma transport, 337 H Interaction
GAs. See Genetic algorithms (GAs) H-function, 280, 417 determining the nuclide of, 314
Gas centrifuge configurations, 375 Half Value Layer (HVL), 24 physics of, 306
Gas centrifuge process, 379 Hamilton-Jacobi-Bellman equation, 350 Intercontinental ballistic missiles (ICBM), 117,
Gas cooled fast reactor (GCFR), 120, 126 Hamiltonian function, 357, 360, 498 499 132
Gas cooled reactor (GCR), 103, 113 114 Handbook of Nonlinear Partial Differential Intermediate node, 367
physical design parameters of Torness GCR, Equations, 156 International Atomic Energy Agency (IAEA),
114t Heat balance equation, 185 34, 388
Gas-Cooled Fast Reactors (GFR), 405 Heat conduction in 2-D and 3-D, 159 164 International Business Machines (IBM), 350
Gaseous diffusion process, 379 Dirichlet, Neuman and mixed boundary International Commission on Radiological
Gauss-Siedel methods, 238 conditions, 160t Protection (ICRP), 28, 515
Gaussian distribution, 39, 71 72, 314 spherical geometry, 162 164 International Standards Organization (ISO),
Gaussian elimination, 238 Heat deposition, 403 388
Gaussian PDF, 14 16 Heavy nucleus, 57 58 International Thermonuclear Experimental
Gaussian Process Regression, 374 Heavy water (HW), 72 73, 103, 421, 426 Reactor (ITER), 129, 310, 493
Gaussian spectrum, 312 HELIOS-2 (Lattice codes), 343 International Tokamak Experimental Reactor
GEANT codes, 512, 514 515 Hessian matrix, 350 351 (ITER), 5
 Index 525

INTG record, 343 Linear quadratic Gaussian (LQG), 364 365, interaction of gamma radiation with matter,
Iodine (125I), 11, 512 514 504 40 41
Iodine-135 Linear Transport Equation, 261 262 radiation dose from Calfornium-252 gamma
atoms, 7 Liner attenuation coefficient, 24 source in water, 41 44
decay rate equation, 7 Liquid drop model, 6 MATLAB program, 177, 195 196, 359,
Ionizing radiation (IR), 510 Liquid Metal Fast Breeder Reactor (LMFBR), 419 421, 449 450
Iridium (192Ir), 11 120 for point source in finite non-multiplying
Iron (Fe), 403 404, 425, 474, 494 Liquid Oxygen (LOX), 134 sphere, 456 459
Isotropic scattering, 56, 424 Liquid sodium metal, 397 random number generation, 196t
Liquids containing radio-isotopes, 509 Maxwellian distribution, 51
LIST record, 343 MC. See Monte Carlo (MC)
J Lithium, 3 4, 405 MCA. See Multichannel analyzers (MCA)
Jacobi-Siedel methods, 238 Load Pattern Optimization (LPO), 374 MCBEND, 340
Japanese Evaluated Nuclear Data Library Loading pattern (LP), 374, 489 MCF. See Magnetic confinement fusion (MCF)
(JENDL), 341 Los Alamos National Labs (LANL), 339 MCO. See Moisture carryover (MCO)
Jezebel, 386 388 Loss of Coolant Accident (LOCA), 105, 340 MCTS. See Monte Carlo Track Structure
material composition of, 387t Low dose rate (LDR), 509 (MCTS)
tallies for, 387t Low enriched uranium (LEU), 120 Mean values, 455
Joint European Torus (JET), 129 submarine reactors, 120 Medical physics (MP), 509
Medical therapy, 11
MELCORE Accident Consequence Code
K M (MACCS), 341
K shell, 3 4 Machine learning, 375 Meta-heuristic methods, 489
KCODE card, 399 Macroscopic cross-section, 63 64, 64t, 221 Metal salt reactors (MSR), 405
KENO codes, 340 341 Magnetic confinement fusion (MCF), 128 130 Metastable isomers, 75 76
“Kernel of the integral operator”, 165 Magnetic Resonance Imaging (MRI), 510 Method of characteristics (MOC), 259
Kilopower Reactor Using Stirling Technology Marine propulsion reactors, 116 120. See also Micronuclear heat pipe reactor, 380
(KRUSTY), 141 Space propulsion reactors Micronuclear reactor (MNR), 120, 139,
Kinetic energy (KE), 4 former soviet/Russian nuclear submarine 324 326, 374, 400 401, 476
Kinetic theory, 51 53 program, 117 material data for micronuclear reactor
Klein-Nishina differential cross-section (K-N HEU/LEU submarine reactors, 120 simulation, 401t
differential cross-section), 23 modern-day submarines, 117 118 Microscopic cross-section, 58 62
Korean Atomic Energy Research Institute submarine programs, 117 averaged cross sections of Pu-239, 64t
(KAERI), 34 technical features, 118 119 thermal neutron absorption cross sections,
KORI-1 homogeneous reactor model, 489 490 US nuclear submarine program, 116 63t
Kullback Leibler divergence for uniform Mark I reactor, 111 112 Miniature source neutron reactor (MNSR), 512
random numbers, 199 Markov processes, 305 Minimum critical mass (MCM), 238
Markovian processes, 188 Minimum path tree, 365
MARS. See Multiple Array System (MARS) MIRVs. See Multiple independently
L Marshak boundary conditions, 289 290 targetable reentry vehicles (MIRVs)
L’hopital’s rule, 214 Mass attenuation coefficient, 24 Mixed oxide fuel (MOX), 121
Laboratory system, 314 Mass defect, 6 “Mixture” cross-sections, 34
Lagrange multiplier, 352, 357 Mass-spring-damper equation (MSD equation), MOC. See Method of characteristics (MOC)
Laplace equation, 153, 158 159 361 Moderator macroscopic absorption cross-
Laplace transform method, 166, 259, 270, 361, solution of, 361 sections, 221
417 Material densities, 393t, 474 Modern-day submarines, 117 118
Lattice cell, 464, 467 Mathematical and Numerical Integrator and nuclear submarine reactors, 119t
Lattice codes, 343 Computer (MANIAC), 305 Moisture carryover (MCO), 374
Lattice fuel cell, 399 Mathematical foundations in nuclear Molière distribution, 39
Lattice physics codes, 343 engineering Molten Salt Reactor (MSR), 126
Lawrence Livermore Laboratories (LLL), 339 adjoint function, 186 187 Molydenum (Mo-99), 510
Lead (Pb), 492, 494 approximate methods, 185 186 MONACO (Monte Carlo code), 339
Lead Cooled Fast Reactor, 120, 126 evaluation of integrals, 201 206 MONACO with Automated Variance
Lead fast reactor (LFR), 121 integral equations, 165 170 Reduction using Importance
Lead-cooled fast reactor, 491 integro-differential equations, 170 174 Calculations (MAVRIC), 340
Legendre polynomials, 163, 184, 255 numerical methods, 174 185 Monte Carlo (MC), 92
Legendre’s equation, 163 ODEs, 150 156 code, 339 340, 512, 514 515
Lehmer RNG, 196 PDEs, 156 165 exercises in Monte Carlo simulation
Leinnitz’s rule, 166 random processes, probability, and statistics, estimating neutron flux in non-multiplying
Light water (LW), 421, 426 187 201 sphere, 453 462
Light water reactor (LWR), 103 Mathematical modeling and simulation of atom MCNP geometry plotting in core
Linear energy transfer (LET), 510 and radiation, 34 44 neutronics, 476 481
Linear integral equation, 166 alpha particle transport simulation, 35 perturbation calculations, 474 476
Linear quadratic controller (LQR), 364 365 interaction of electrons with matter, 35 40 radiation safety and shielding, 473 474
526 Index

Monte Carlo (MC) (Continued) MSD equation. See Mass-spring-damper Neutron diffusion equation (NDE), 184, 211,
reactor core modeling, 464 472 equation (MSD equation) 259
reflected assemblies, 462 464 MSR. See Metal salt reactors (MSR); Molten conservation equation, 211 213
sampling from distribution function, Salt Reactor (MSR) Core neutronics with diffusion equations,
449 453 Multi-grid algorithms, 374 251 256
method, 34, 187, 203, 211, 305, 340, 401, Multi-group Oak Ridge Stochastic Experiment fuel concentration effect on critical mass,
417 (MORSE), 339 340 238 248
demonstration, 328 331 MORSE code, 339 multigroup diffusion equation, 234 238
estimating perturbations with Monte Carlo MORSE-SGC code, 339 one-group diffusion equation, 213 221
simulation, 333 Multi-objective core optimization, 373 374 two-group adjoint diffusion equations,
modeling geometry, 316 327 pressurized water reactor core pattern 248 250
for numerical integration, 203 206 optimization, 374 two-group diffusion equation, 221 234
simulation of random walk, 308 316 Multichannel analyzers (MCA), 30 Neutron DT flux, 428 429
stochastic simulation, 305 308 Multigroup diffusion equation of NDE, Neutron energy, 235, 450
variance reduction methods, 332 333 234 238 Neutron flux, 53, 66, 92, 383, 418, 430, 460
nuclear fission applications, 390 401 multigroup cross-sections, 235t estimation in non-multiplying sphere,
micronuclear reactor, 400 401 numerical solution of, 235 238 453 462
Toshiba 4S reactor, 394 399 Multigroup form of integral transport equation, MATLAB program for point source
unit lattice cell and fuel assembly of 268 in finite non-multiplying sphere,
AP1000 reactor, 390 394 Multimission radioisotope thermoelectric 456 459
nuclear fusion applications, 401 405 generator (MMRTG), 11 results, 460 462, 462t
perturbation, 515 Multiple Array System (MARS), 339 simulation process, 453 456
analysis, 515 Multiple independently targetable reentry temperature effects on, 227
simulation, 92 97, 260, 308 309, 337, 350, vehicles (MIRVs), 117 thermal data, 227
381, 390, 397, 417, 495, 509 Multiplicative congruential generator, 196 Neutron generator, radiation moderation for,
actinide physical properties, 97t Multiplying systems. See also Nonmultiplying 389 390, 390t
alpha radiation therapy, 511 512 systems Neutron interaction, 51, 306
bare critical assemblies, 98t one-group diffusion equation, 215 219 criticality, 88 97
brachytherapy, 512 517 cylindrical reactor, 216 218 flux measurement, 64 66
to calculate energy deposition and dose slab reactor, 215 216 kinetic theory, 51 53
distribution for brachytherapy, 514 517 spherical reactor, 218 219 macroscopic cross-section, 63 64
critical solid spherical systems, 97t two-group diffusion equation, 227 230 microscopic cross-section, 58 62
estimating perturbations with, 333 two-group core functions, 229t nuclear fission, 80 87
flux in finite sphere with point isotropic two-group data, 230t reaction rates, 66 67
source, 431 433 two-group reflector functions, 229t resonance cross-section, 74 80
flux in slab, 425 slowing down, diffusion and thermalization,
geometries for illustration of, 320 327 67 73
production of radio-isotopes, 510 511 N types of, 53 58
quantities of interest in fission neutronics, Nanotechnology, 515 angular quantum number vs. kinetic
325t Narrow resonance (NR), 79 80 energy, 54f
simulation in nuclear systems Narrow resonance infinite mass models (NRIM neutron scattering in lab and center of
bare critical assemblies, 381 388 models), 79 80 mass systems, 55 58
criticality safety, 388 389 National Institute for Science and Technology s-wave scattering, 54f
radiation moderation and shielding, (NIST), 34 selected neutron interactions, 55f
389 390 National Nuclear Data Services, 6 Neutron transport equation (NTE), 259, 417
Monte Carlo for Neutrons and Photons Natural boron, 30 exact solutions of transport equation,
(MCNP), 339 NDT. See Nondestructive testing (NDT) 268 284
code, 339, 341, 510, 512, 515 NDU. See Nuclear Demonstration Chandrasekhar’s H function, 271b
documentation, 464 Unit (NDU) classic albedo problem, 270 272
geometry plotting in core neutronics, NEA. See Nuclear Energy Agency (NEA) finite sphere with point isotropic source,
476 481 NEM. See Nodal Expansion Method 274 284
problems, 480 481 (NEM) infinite medium with plane isotropic
surface coefficients, 478t NERVA. See Nuclear Engine for Rocket source, 273 274
Godiva simulations, 474 Vehicle Application (NERVA) numerical methods for solving transport
MCNP5 simulations, 503 “Neumann series”, 308 equation, 285 298
simulation, 422, 473 474 Neumann’s principle of stored program, 305 structure of, 260 268
tallies, 470 Neural network models, 495 integral form of transport equation,
Monte Carlo N-Particle (MCNP), 339, 489 Neurological disorders, 510 266 268
Monte Carlo Track Structure (MCTS), Neutron integro-differential form of, 260 265
510 511 age, 72 multigroup form of integral transport
Monte Carlo whole-core simulations, 400 current, 260 261, 383 equation, 268
Movement disorders, 510 density, 173 two-group transport equation, 265 266
MOX. See Mixed oxide fuel (MOX) Neutron A-1, 308 309 transport theory for reactor calculations,
MRI. See Magnetic Resonance Imaging (MRI) Neutron diffusion, 260, 337, 417 298 302
 Index 527

Neutron(s), 2, 25, 30, 52 53, 58, 88, 265, NRIM models. See Narrow resonance infinite small modular reactors, 121 127
338 340, 390, 394, 404 mass models (NRIM models) for space, 139 141
colliding in hydrogen, 70 NTCOC. See Neutronics/Thermal-hydraulic space propulsion, 132 137
emerging, 86 Coupling Optimization Code (NTCOC) status of nuclear power, 103 107
neutron/photon transport codes, 340 NTE. See Neutron transport equation (NTE) systems, 107 115, 338
population, 211 NTP. See Nuclear thermal propulsion (NTP) boiling water reactor, 110 112
radiation, 380, 493 Nuclear criticality safety analysis, 473 fast breeder reactor, 114 115
scattering in lab and center of mass systems, Nuclear cross sections, 62 gas cooled reactor, 113 114
55 58 Nuclear data, 341 344, 381 pressurized heavy water reactor,
shielding, 494 processing 112 113
source imaging, 374 codes, 343 pressurized water reactor, 108 110
spectra, 68, 103 system, 60 theory, 184
thermal cross sections, 62 Nuclear Demonstration Unit (NDU), 125 Nuclear Regulatory Commission, 388
transport, 260, 337 Nuclear energy, 105, 107, 349 Nuclear rocket designs for deep space
Neutronic[s], 405 Nuclear Energy Agency (NEA), 34, 126, 340 exploration, 134 137
code, 489 Nuclear Engine for Rocket Vehicle Application engine comparisons, 136t
coupled with thermal hydraulics, 337 (NERVA), 134 heating values, 135t
MCNP geometry plotting in core, 476 481 Nuclear engineering, 5, 53, 187, 337, 349, 357, Nuclear Shell Model, 10
neutronic-TH coupled simulations, 491 372 373, 509 Nuclear stability, 5, 5f
optimization, 492 optimization in, 489 Nuclear submarines, 118, 120
parameters, 400 Nuclear fission, 80 87, 128, 489 reactors, 374
simulation, 489 critical energy, 81 83 Nuclear system, 1, 64, 305 306, 337
Neutronics/Thermal-hydraulic Coupling fissile and fertile materials, 85 86 Nuclear technologies, 349, 379
Optimization Code (NTCOC), 494 fissile/fertile nuclides, 86t radioactive nuclides in, 11 12
Newton’s second law of motion, 3 fission spectrum, 86 87 Nuclear thermal power reactors, 379
Nichols-Ziegler Method, 363 fission yield, 84 Nuclear thermal propulsion (NTP), 134
Niobium, 405 number of neutrons emitted in fission, Nuclear thermal rocket, 136
Niobium-tin (Nb3Sn), 403 84 85 Nuclei, 5
NIST. See National Institute for Science and process, 80 81 Numerical methods, 174 185, 211, 259
Technology (NIST) spontaneous fission rates, 81t FDM, 174 177
Nitrogen (N), 516 reaction, 84 FEM, 178 185
NJOY (Nuclear data processing codes), 343 reactors, 337 for solving transport equation, 285 298
NM. See Nuclear medicine (NM) Nuclear fuel cycle, 379 BN method, 295 296
Nodal Expansion Method (NEM), 211 Nuclear fusion, 5, 128 132 criticality estimates, 297 298
Nodal method with transport theory, 296 fusion reaction, 128 discrete ordinates method, 285 287
Non-dominated sorting genetic algorithm ICF, 130 132 DPN method, 293 294
(NSGA-II), 495 MCF, 129 130 finite element method, 296
Non-multiplying media, 419 reactors, 374 hybrid methods, 297
Non-multiplying sphere, neutron flux Nuclear interaction data, 306 nodal method with transport theory, 296
estimation in, 453 462 Nuclear medicine (NM), 509 spherical harmonics method, 287 293
Nonanalog simulation, 332 Nuclear power
Nondestructive testing (NDT), 12, 20 generations of, 103 106
Nonleakage probability, 89 nuclear reactor classifications, 104t O
Nonlinear first-and second-order ODEs, 151 industry, 106 107 Oak Ridge National Laboratories (ORNL), 339,
Nonlinear integral equations, 166 nuclear power industry, 106 107 389
Nonmultiplying systems. See also Multiplying reactor shut down, 106 One dimension (1D)
systems status of, 103 107 slab reactor equation, 216
one-group diffusion equation, 213 215 systems in space, 138 141 time-dependent heat conduction, 157
finite cylinder, 215 RTGs, 138 transport equation, 424
finite slab, 213 214 SNAP systems, 138 141 One-group balance equation, 212
point source in infinite medium, 215 Nuclear Power Demonstration (NPD), 112 One-group critical equation, 234
two-group diffusion equation, 221 227 Nuclear power plants (NPPs), 106 One-group criticality, 219 221
computer programming example, Nuclear power reactors, 11, 323, 349, 374, 510 One-group diffusion equation, 417 418,
225 227 Nuclear propulsion reactors, 323 474 476
temperature effects on neutron flux, 227 Nuclear radiation, 509 of NDE, 213 221
Nonsingular linear integral equations, 166 Nuclear reaction, 6, 60 multiplying systems, 215 219
Nonstandard ENDF/B reaction numbers, 383 Nuclear reactor(s), 64, 103, 337, 380, 495, nonmultiplying systems, 213 215
Nonuniform fuel distribution 503 504 one-group criticality, 219 221
slab model, 239 243 control, 361 One-group diffusion theory criticality,
spherical model, 244 247 marine propulsion reactors, 116 120 417 418
Normal distribution, sampling from, 449 451 nuclear fusion, 128 132 one-group data for Godiva, 418t
NPD. See Nuclear Power Demonstration (NPD) nuclear power systems in space, 138 141 one-group macroscopic cross sections for
NPPs. See Nuclear power plants (NPPs) plutonium production reactors, 120 121 Godiva, 418t
NR. See Narrow resonance (NR) refueling, 365 One-speed transport equation, 91, 264
528 Index

One-speed transport theory criticality, PARallel Time-Dependent SN (PARTISN), Positron Emission Tomography (PET), 11
419 421, 420t 339 Postcollision angles, 35
One-zone criticality, 490 PARET code, 497 Power conversion system, 374
ONEDANT (parallel time-dependent Partial differential equations (PDEs), 156 165. Power density distribution, 324
deterministic code), 339 See also Ordinary differential equations Power plants, 104
Open-loop system, 362 (ODEs) Power-producing nuclear reactor, 104
OpenMOC codes, 339 340 equations of fluid dynamics, 156 PPF. See Peak power factor (PPF)
Ophthalmic brachytherapy, 516 flux formulation, 164 165 Pressurized heavy water reactor (PHWR),
Optical data communication, 337 heat conduction in 2-D and 3-D, 159 164 85 86, 103, 112 113, 349, 493
Optical model of neutron interaction, 51 Laplace equation, 158 core pattern optimization, 374
Optimal analysis, 350 Particle and Heavy Ion Transport Code System, physical design parameters of PHWR, 113t
Optimal control equation, 357 512 Pressurized water reactor, 108 110
Optimal discrete control, 360 Particle swarm optimization (PSO), 372 373, AP1000 and PWR300 MW, 110t
Optimal distribution, 498 500 489, 493 P-PWR and APR1400 nuclear reactors, 109t
Optimal energy spectrum, 380 optimization by, 373t Primary Knock-on atom (PKA), 403 405
Optimal path, 367 Pb-Bi-cooled Fast Reactor, 126 Probabilistic Risk Assessment PRA techniques,
Optimal solution with system of first-order Peak power factor (PPF), 492 105
ordinary differential equation Penetration and ENErgy LOss of Positrons and Probability distribution function (PDF), 14 16,
constraints, 352 360 Electrons (PENELOPE) code, 510, 512, 59, 187, 314, 369, 449
Optimization method, 337, 350, 502, 504 514 515 sampling from, 196 199
applications of optimization in reactors, Performance index (PI), 355 Prompt Gamma Neutron Activation Analysis
373 375 Perturbation (PGNAA), 389
controller design and optimization, 361 365 algorithms, 515 516 Prompt Neutron Activation Analysis (PNAA),
proportional integral derivative controller calculations, 474 476 380
tuning parameters, 364t Perturbations with Monte Carlo simulation, 333 Proportional counter, 29
deterministic optimization, 350 360 PET. See Positron Emission Tomography Proportional Integral Derivative controller (PID
with algebraic constraints, 351 352 (PET) controller), 362, 374
without constraints, 350 351 PGNAA. See Prompt Gamma Neutron Prostate adenocarcinomas, 513
optimal discrete control, 360 Activation Analysis (PGNAA) Protons, 2
optimal solution with system of first-order Photoelectric effect in photon, 20 Pseudo-random numbers, 195
ordinary differential equation Photon collisions, 514 PSO. See Particle swarm optimization (PSO)
constraints, 352 360 Photoneutron reactions, 23 Purdue Advanced Reactor Core Simulator
dynamic programming, 365 367 PHWR. See Pressurized heavy water reactor (PARCS), 340, 343
in nuclear engineering, 375 (PHWR) Pythagoras’ Theorem, 354
in nuclear systems Physical model of Godiva, 381
controller design optimization, 503 504 PI. See Performance index (PI)
fuel loading pattern optimization, PID controller. See Proportional Integral Q
495 500 Derivative controller (PID controller) Quality factor, 27
fusion neutronics design optimization, Pin-cell simulation, 399
493 494 PKA. See Primary Knock-on atom (PKA)
radiation detection or optimization, PKE. See Point kinetics equations (PKE) R
501 503 Placzek discontinuities, 69 Radial power peaking (RPP), 400
radiation shielding design optimization, Planck’s constant, 3 Radial power peaking factor (RPPF), 374
494 495 Plane isotropic source, 424 Radiation, 509
reactor core design optimization, 489 493 finite sphere with, 274 284 damage, 403
stochastic optimization, 367 373 infinite medium with, 273 274 detection, 501 503
variables, 492 Plastic, 380 B(n, α)Li reactions for wax-moderated
Optimized controller design, 374 Plutonium, 30, 463 464 BF3 detector, 502t
“Optimum moderation”, 245 246 Plutonium-238, 11 dose, 67
Ordinary differential equations (ODEs), Plutonium-239, 5, 60 61, 85 energy, 514 515
150 156, 355. See also Partial production reactors, 120 121 with matter, interaction of, 12 24, 12f
differential equations (PDEs) sphere, 386 interaction of alpha rays with matter,
coupled first-order ODEs, 155 156 PNAA. See Prompt Neutron Activation 12 16
Poisson equation, 153 155 Analysis (PNAA) interaction of beta radiation with matter,
Organic coolants, 104 Point isotropic mono-energetic source, 430 16 19
ORNL Isotope Degeneration and Depletion Point isotropic source, flux in finite sphere interaction of gamma radiation with
Code (ORIGEN), 338 with, 428 433 matter, 19 24
Outboard (OB), 493 494 Point isotropic source equations, 430 mathematical modeling and simulation of,
Oxygen (O), 516 Point kinetics equations (PKE), 155 156, 361 34 44
Poisson’s equation, 153 155, 213, 298 alpha particle transport simulation, 35
Polyethylene, 390, 502 interaction of electrons with matter,
P moderator, 503 35 40
Palladium (103Pd), 11 Polymer coated nanoparticles, 511 512 interaction of gamma radiation with
Paraffin wax, 502 Pontryagin’s maximum principle, 498 500 matter, 40 41
 Index 529

radiation dose from Calfornium-252 Random processes, probability, and statistics, source description, 466
gamma source in water, 41 44 187 201 surrounding cells, 465 466
moderation, 389 390 central limit theorem, 200 201 surface cards, 466
for neutron generator, 389 390 Kullback Leibler divergence for uniform tally cards, 470
nuclide therapy, 509 random numbers, 199 Reactor design optimization, 373
oncology, 514 515 law of large numbers, 199 200 Reactor Excursion and Leakage Analysis
optimization, 501 503 application of, 200 Program (RELAP), 340
B(n, α) reaction rate as function of Markovian processes, 188 RELAP code, 340
pressure and 10B weight fraction, 503t population and sample, 188 Reactor Safety Study” report, 105
safety, 473 474 random numbers, 195 196 Reactor vessel (RV), 107
current, flux and dose for water and iron random processes, 188 Reactors, optimization applications in,
slabs, 475t random variables, PDF, and CDF, 189 195 373 375
half value layer thickness for some sampling from PDFs, 196 199 controller proportional integral derivative,
shielding materials, 475t Random sampling, 449 374
limits, 28 29 Random variables, 189 195 multi-objective core optimization, 373 374
source term, 127, 127t Random walk, 35, 308 309 radiation shielding, 374 375
sources and effects of, 24 30 batch, history, random walk and events, 316 Rectum, 513
absorbed dose, 26 27 determining the nuclide of interaction, 314 Recurrent cancer, 509
Am-Be and Cf-252 neutron sources, 25t determining the type of event, 313 314 Recursive algorithm, 365
effective dose, 28 events in, 305 306 “Reference codes”, 514
equivalent dose, 27 28 mean and variance, 315 Reflected neutron systems, 462
neutron energy spectra, 25f processing Refueling optimization, 365
radiation detection, 29 30 capture event, 315 Relative biological effectiveness (RBE), 511
radiation dose, 26 escape-from-system event, 315 Relative entropy, 199
radiation safety limits, 28 29 fission event, 314 315 Relative standard deviation (RSD), 460
therapy, 509 scattering event, 314 Relativistic kinetic energy, 5
Radiation Safety Information Computational sampling “distance to collision”, 313 Relaxation length, 420, 474
Center (RSICC), 34 sampling source, 312 313 Resolved resonances, 60
Radiation shielding, 374 375, 380, 389 390, watt fission spectrum parameters, 313t Resonance
473 474, 489 simulation of, 308 316 cross-section, 74 80
design optimization, 494 495, 494f estimators and tallies, 309 311 escape probability, 79, 89
radiation dose across concrete shield, 392t Monte Carlo simulation, 308 309 integral averaged cross sections, 62
Radiation transport, 305 RASCAL. See Radiological Assessment Ritz method, 185
codes, 338 340 Systems for Consequence AnaLysis RNG. See Random number generator (RNG)
process, 337 (RASCAL) Rocket Propellant (RP), 134
Radiative capture, 394 Rasmussen Report, 105 Rodriguez’ formula, 163
cross-section, 77 Rayleigh Ritz variational method, 186 RPP. See Radial power peaking (RPP)
Radio-isotope cesium-131 (Radio-isotope Reaction Numbers, 330 RPPF. See Radial power peaking factor (RPPF)
131
Cs), 513 Reaction rates, 66 67, 383 RSD. See Relative standard deviation (RSD)
Radio-isotopes, 509, 512 age-to-thermal, 72t RSICC. See Radiation Safety Information
production of, 510 511 diffusion length of selected materials, 73t Computational Center (RSICC)
Radioactive atoms, 7 non-factors, 74t RTGs. See Radioisotope thermal generators
Radioactive decay, 6 12 reaction cross-section MT numbers, 67t (RTGs)
alpha decay, 8 9 Reactor core design optimization, 489 493 Russian nuclear submarine program, 117
beta decay, 9 atomic density and mass of uranium in Russian Roulette, 332
gamma decay, 10 11 KORI-1 homogenous model, 491t Russian submarine reactors, 118 119
radioactive nuclides in nuclear technologies, basic geometry and materials feasibility for Rutherford scattering cross-section, 14 16
11 12 neutronic calculation, 490f
Radioactivity arises in nuclear reactors, 7 coupled genetic algorithm-regression with
Radioisotope thermal generators (RTGs), 138 multi-module optimization, 493f S
Radiological Assessment Systems for coupled neutronics and thermal hydraulics Safety rods (SR), 492
Consequence AnaLysis calculation, 492f Sampling from PDFs, 196 199
(RASCAL), 341 KORI-1 reactor homogeneous model with sampling from analytic PDFs, 196 198
Radiological protection codes, 341 MCNP5, 490t Sampling from nonanalytic PDFs, 199
RADionuclide Transport and Removal And two-group diffusion data, 491t SCALE code system, 340
Dose Estimation (RADTRAD), 341 Reactor core fueling optimization, 349 Scattering cross-section, 79 80
Radionuclides for brachytherapy, 509 Reactor core modeling, 464 472 Schrodinger wave equation, 1, 3 4, 151 152
Radiotracers, 509 input file, 464 466 Scintillators, 501
Radium bounding window, 464 Second derivatives, 476
decay chain of Ra226, 8 lattice, 465 Second-order neutron diffusion equation, 357
series, 24 universe, 464 465 “Selection” tests, 367
Random number generator (RNG), 195 plotting geometry, 467 470 Silicon carbide (SiC), 341
Random numbers, 195 196 plotting tallies, 471 472 Simple photon transport, 40
MATLAB program, 195 196 reaction rates, 470 471 Simulated Annealing optimization, 373
530 Index

Simulation Steady-state one-group diffusion equation, neutron diffusion equation, 338

of legacy Godiva bare assembly, 379 422 423 PARCS, 343
parameters, 493 Steam generator, 107 108 reactor core simulators, 340
process, 453 456 Steel slabs (SS-316 slabs), 426 Three-dimensional heterogeneous model, 516
variables used in Monte Carlo program, 456t Stochastic heuristic methods, 374 THREEDANT (parallel time-dependent
Single-Photon Emission Computed Stochastic optimization, 350, 367 373 deterministic code), 339
Tomography (SPECT/CT), 509 genetic algorithms, 367 371 Threshold energy, 60
Six-factor formula, 419 PSO, 372 373 Time-dependent precursor concentration
Slab model, 239 243 Stochastic simulation, 305 308 equation, 262
Slab reactor, 215 216 events in random walk, 305 306 Time-dependent reactor kinetics codes, 340
Slowing down process, 67 73 Markov processes, 305 Toroidal field coils (TFC), 326
Slowing-down equation, 234 nuclear interaction data, 306 Toroidal Field Coils, 403
Small modular reactors (SMRs), 107, physics of interactions, 306 TORT code, 338
121 127, 493 Sturm Liouville theory, 151 Toshiba 4S reactor, 394 399
design features of, 121 125 Submarine 5th Generation Westinghouse atomic densities of boron carbide, 399t
parameters of, 122t (S5W), 117 physical parameters of the 4S reactor core,
SMRSs, 123t Submarine launched ballistic missile (SLBM), 398t
generation-IV reactors, 125 127 117, 132 Toshiba Gen-IV 4S, 380
radiation source term, 127 Submarine programs, 117 Total effective dose equivalent (TEDE), 29
very small modular reactor, 125 Submarine Thermal Reactor (STR), 116 Total fuel mass, 243
Small nuclear auxiliary power systems (SNAP Super-criticality, 88 TRAC/RELAP Advanced Computational
systems), 11, 138 141 Supercritical Water Cooled Reactor (SCWR), Engine, 340
nuclear reactors for space, 139 141 126 Track length estimator (TLE), 309 310, 425,
“Small-angle” deflections, 13 Support Vector Regression, 374 455
SN method, 338 Surface cards, 466 Traditional optimization methods, 350
Sodium cooled fast reactor (SFR), 115, 126, 493 Surgery, 509 Training Research Reactor Idaho General
Solid state detectors, 29 30 Swarm intelligence algorithm, 372 Atomics research reactor, 497
Solving transport equation, numerical methods Systems Analysis Programs for Hands-on Transcendental equation, 419, 423
for, 285 298 Reliability (SAPHIRE), 341 Transport codes, 60
Source-free infinite medium, 269 neutron and radiation, 338 340
Space exploration, 132 134 Transport correction in diffusion model, 424
Space propulsion reactors, 132 137. See also T Transport equation, 339, 417
Marine propulsion reactors TAB1 record, 343 integral form of NTE, 266 268
conventional rocket designs, 132 TAB2 record, 343 Transport simulations for alpha particles, 41
nuclear power systems in, 138 141 Tallies, 309 311 Transport theory, 91 92, 417
nuclear rocket designs for, 134 137 Tally cards, 470 flux in finite sphere with point isotropic
space exploration, 132 134 TALLY F14, 470 source, 430 431
Space propulsion reactors, 374 TART codes, 339 flux in slab, 424 425
Space technologies, 11 Taylor series, 474 476 infinite medium eigenvalues from
Spatial attenuation, 425 Technetium (99mTc), 11, 510 transport theory, 425t
Spectral Green’s Function Nodal Method (SGF technetium radio-isotope, 510 for reactor calculations, 298 302
Nodal Method), 255 Tele-therapy, 509 collision probability method, 299
Spherical geometry, 162 164 Tellurium-135, 7 method of characteristics, 299 302
Legendre polynomials, 163t Tenth Value Layer (TVL), 24 Transport-to-diffusion flux, 430
Spherical harmonics method, 34, 163, TEXT record, 343 Traveling Salesman Problem, 365
287 293 Thermal column graphite, 512 Trinitrotoluene (TNT), 501
Spherical model, 244 247 Thermal conductivity, 155 TRIPOLI code (Monte Carlo code), 340
Spherical reactor, 218 219 Thermal flux, 65, 425, 429 Tritium, 493 494
1-group diffusion multiplying, 220t Thermal hydraulics code (TH code), 340, 489 Tritium breeding rate (TBR), 380, 493 494
neutron flux, 219f Thermal macroscopic cross-section, 63 64 TRITON (Lattice codes), 343
Splitting techniques, 332 Thermal neutron activation (TNA), 501 Tungsten (W), 405, 494
Spontaneous fission, 81 Thermal neutron activation analysis (TNAA), Two-group adjoint diffusion equations,
Square lattice, 253 380, 389, 489 248 250
Standard deviation (STD), 461 462 Thermal neutrons, 52 53, 73 Two-group critical determinant, 230
Standard ENDF/B reaction numbers, 382 383 Thermal reactor, 103 Two-group criticality, 230 234
Standard Overrelaxation Method Thermalization process, 67 73, 392 Two-group diffusion equation of NDE,
(SOM), 238 Thermo electric generator (TEG), 221 234
State equation, 356 400 401 multiplying systems, 227 230
Statistical distribution function, 51 Thermoluminescent dosimeter (TLD), 29 nonmultiplying systems, 221 227
Steady-state heat conduction equation, 151, Thick target bremsstrahlung model (TTB two-group criticality, 230 234
153 154 model), 41 Two-group diffusion theory, 239
Steady-state neutron diffusion equation, 151 Thorium series, 24 criticality, 418 419
Steady-state one-dimensional transport Three dimension (3D) Two-group equations, 212, 418
equation, 263 MC simulations model, 405 Two-group fluxes, 418
 Index 531

Two-group transport equation of NTE, Uranium nitride (UN), 134 Weapons Grade Highly Enriched Uranium
265 266 Uranium oxide, 379 (WG-HEU), 120
2-D steady-state heat conduction, 158 Uranyl nitrate solution (UO2(NO3)2), 379 Weapons grade uranium (WPG), 379
TWODANT (parallel time-dependent Urethra, 513 Weighted residual method, 186
deterministic code), 339 US nuclear submarine program, 116 Weiner-Hopf method, 417
Westcott factors, 62
Westcott g-factor, 62
U V Westinghouse AP-1000, 380
Unified picture, 417 Variable coefficient linear ODE, 151 Wiener continuous process, 188
Unit lattice cell, 394 Variance, 189 WIMSD code, 253
of AP1000 reactor, 390 394 reduction methods, 332 333 Windscale Advanced Gas Cooled
MC simulations, 380 Variance of variance (VoV), 308 Reactor, 113
United States Nuclear Regulatory Commission Very High Temperature Reactor (VHTR), 126, Winfrith Improved Multi-group Scheme
(USNRC), 28, 341, 388 405 (WIMS), 338
thermal hydraulics codes, 340 Very small modular reactor (vSMR), 125 WtPartCol, 455
Uranium, 3 4, 375, 463 464 Volterra integro-differential equation, 165, 170
enrichment, 379 Volume-averaged cell flux, 309
series, 24 X
Uranium-233, 85 X-ray fluorescence (XRF), 12
Uranium-235, 5, 30, 85 W X-rays, 1, 20
Uranium-238, 30 Water, 30, 72 73, 390, 426, 462, 494 Xenon-135, 60 61
Uranium dioxide fuel, 30 Water phantoms, 516
Uranium hexafluoride (UF6), 379, 388, 473 Water-cooled lithium lead (WCLL), 493 494
storage of uranium hexafluoride cylinders, Watt fission spectrum, 86, 449, 466 Z
388 389 sampling from, 451 453 Zion-I 1065 reactor, 496