NEUTRON SCATTERING APPLICATIONS AND TECHNIQUES

1 Neutron Imaging
and Applications
Neutron Imaging and Applications

A Reference for the Imaging Community

Ian S. Anderson | Robert L. McGreevy | Hassina Z. Bilheux

Editors

1 23
Neutron Imaging and Applications
Neutron Scattering Applications and Techniques

Series Editors:
Ian S. Anderson Alan J. Hurd
Neutron Sciences Directorate Lujan Neutron Scattering Center at LANSCE
Oak Ridge National Laboratory Los Alamos National Laboratory
Building 8600, MS 6477 PO Box 1663, MS H805
Oak Ridge, TN 37831 Los Alamos, NM 87545
USA USA
[email protected] [email protected]

Robert L. McGreevy
ISIS
Science and Technology Facilities Council
Rutherford Appleton Laboratory
Harwell Science and Innovation Campus
Chilton, Didcot OX11 0 QX
UK
[email protected]

Neutron Applications in Earth, Energy and Environmental Sciences

Liyuan Liang, Romano Rinaldi, and Helmut Schober, eds.
ISBN 978-0-387-09415-1, 2009
Neutron Imaging and Applications: A Reference for the Imaging Community
Ian S. Anderson, Robert L. McGreevy, and Hassina Z. Bilheux, eds.
ISBN 978-0-387-78692-6, 2009
Ian S. Anderson Robert L. McGreevy
l

Hassina Z. Bilheux
Editors

Neutron Imaging
and Applications
A Reference for the Imaging Community

13
Editors
Ian S. Anderson Robert L. McGreevy
Oak Ridge National Laboratory Rutherford Appleton Laboratory,
Oak Ridge, TN, USA Oxfordshire, UK
[email protected] [email protected]

Hassina Z. Bilheux
Oak Ridge National Laboratory
Oak Ridge, TN, USA
[email protected]

ISBN 978-0-387-78692-6 e-ISBN 978-0-387-78693-3

DOI 10.1007/978-0-387-78693-3

Library of Congress Control Number: 2008936128

# Springer ScienceþBusiness Media, LLC 2009

All rights reserved. This work may not be translated or copied in whole or in part without the written
permission of the publisher (Springer ScienceþBusiness Media, LLC, 233 Spring Street, New York,
NY 10013, USA), except for brief excerpts in connection with reviews or scholarly analysis. Use in
connection with any form of information storage and retrieval, electronic adaptation, computer
software, or by similar or dissimilar methodology now known or hereafter developed is forbidden.
The use in this publication of trade names, trademarks, service marks, and similar terms, even if they
are not identified as such, is not to be taken as an expression of opinion as to whether or not they are
subject to proprietary rights.

Cover illustration: Cover graphics courtesy of Dr John Katsaras, National Research Council, Chalk
River Laboratories, and Dr Taner Yildirim, NIST Center for Neutron Research.

Printed on acid-free paper

springer.com
Preface

The first successful experiments in Neutron Radiography were carried out (to
our knowledge) in 1935, just a few years after the discovery of the neutron, by
H. Kallmann and E. Kuhn using a small neutron generator [1]. Not
surprisingly, the field has developed and diversified over the last 70 years so
that neutron imaging, in the broad sense, is now routinely used in a wide range
of applications. The aim of this book (one of a series on the applications of
neutron scattering [2]) is to introduce to the reader, whether novice or
experienced researcher, the basic techniques used to image objects using
neutron beams and to give a flavor of the vast range of applications where
these imaging capabilities provide unique insight (Fig. 1).

Fig. 1 Left to right: Pressure gauge with metal backplate; fire hydrant and test tubes filled with H2O
and D2O imaged with gamma-rays (top) and neutrons (bottom) [3]

v
vi Preface

Traditionally, ‘‘neutron imaging’’ is used to describe the direct production of

images by transmitting a beam of neutrons through an object onto a detector
(e.g., film), i.e., exactly the same as is done with X-rays to image broken bones.
An extension of this two-dimensional method (radiography) is to take many
images of the same object in different orientations, and then to use the set of
images to reconstruct a three-dimensional image (tomography). Although this
may still seem to be ‘‘direct’’ imaging, in fact the image is ‘‘re-constructed’’ by
software in a computer and sophisticated mathematical processes can be used to
enhance particular features or generate virtual slices of the imaged object. We
refer to such approaches as ‘‘constructed’’ or indirect imaging methods.
There are many other ways in which virtual pictures or images of an object can be
rendered from more indirect measurements. Ultrasound imaging of an unborn baby
involves the reconstruction of an image from scattered sound waves. As the size of
the object being imaged decreases then simple direct imaging becomes effectively
impossible, as the wavelength of the radiation being used becomes of comparable
size. ‘‘Indirect’’ imaging methods such as electron diffraction can produce essentially
the same ‘‘images’’ of crystal structures as those produced by an apparently ‘‘direct’’
method such as transmission electron microscopy, and indeed it is common to use
the combination of these two techniques to enhance the images.
For example Fig. 2 shows a TEM/STEM image of Si3N4. The sample
progresses from crystalline at the bottom to amorphous structure at the top. In
addition there is a layer of lanthanum atoms on top of the crystalline part of the
sample. The crystalline structure is overlaid with a ball-and-stick image obtained
by diffraction from the crystal. Direct imaging resolves the crystal structure the
same way that diffraction (indirect imaging) does; it does not resolve the
amorphous structure, but amorphous structure can be determined by
diffraction [4].

Fig. 2 Image courtesy of David Cockayne (Oxford University)

Preface vii

The current extreme example of indirect imaging is probably the

reconstruction of three-dimensional images of single nanoparticles from
coherent X-ray diffraction.
In this book we have intentionally taken a broad view of neutron imaging to
include any process by which a picture or image of an object, or part of an
object, can be produced based on the interaction with a neutron beam. These
methods and their applications include
 Direct imaging methods such as neutron radiography being used to study
operating fuel cells
 Three-dimensional tomographical reconstructions of mechanical objects
 The use of indirect methods such as diffraction or small-angle scattering to
image strain patterns in materials
 Indirect imaging of the shapes of biological molecules by reconstruction from
small-angle neutron scattering data.
Clearly, neutron imaging is a less well-known technique than X-ray imaging;
most people know of the simple medical applications of X-ray radiography and
the more recent extension to tomography (CAT or computed axial tomography
scanning). This is largely due to the fact that it is simpler, and less costly, to
generate and manipulate high-intensity sources of X rays than of neutrons.
Hence the applications highlighted in this book rely heavily on the distinctive
properties of neutron beams, which allow useful and often unique information
to be derived from the image.
 Neutrons are weakly interacting neutral particles that penetrate deeply into
most materials, so they can be used to internally image large objects, e.g., a
full-size operating internal combustion engine, non-destructively.
 The amount of scattering or absorption of neutrons by atomic nuclei varies
in an apparently random fashion through the periodic table. Hydrogen in
particular has a very large scattering cross-section. Neutrons can therefore
provide good contrast for light atoms in the presence of heavy atoms, e.g.,
the ‘‘classical’’ neutron image of a rose inside a lead flask (Fig. 3). This
makes neutron imaging highly complementary to X-ray imaging (Fig. 1).
 The amount of scattering or absorption can also vary significantly between
isotopes of the same chemical element; e.g., hydrogen has a very different
scattering cross-section from that of its isotope deuterium. The contrast of
particular elements/materials in an image can therefore be enhanced by
substituting one isotope for another (Fig. 1).
 ‘‘Thermal’’ neutrons have wavelengths similar to inter-atomic distances, so
mechanisms such as refraction or diffraction can be used to enhance images
or to produce indirect images.
 Neutrons have a magnetic moment and a magnetic scattering cross-section
that is comparable to the nuclear cross-section for many atoms. They can
therefore be used to image magnetic structures.
viii Preface

Fig. 3 Neutron radiograph of a rose in a lead flask [5]

Hence the intrinsic properties of neutrons allow a wide range of objects to be

imaged, ranging from massive structures such as helicopter blades to the fine
details of crystal structures and passing through the delicate composition of
biological organisms and plants. Furthermore, these intrinsic properties pro-
vide for an extensive range of contrast enhancement mechanisms including
absorption, scattering, diffraction, refraction, magnetic interactions, and,
potentially, vibrations. These mechanisms can be used to determine the elemen-
tal compositions of objects, which may even be hidden, buried, or encapsulated
within an impenetrable environment. The possibility of studying objects in situ,
or in real operational environments, is promising for a range of industrial and
academic applications.
Neutron imaging techniques have a huge potential but in the past, applica-
tions have been slow to develop, mainly because of the weakness of the source
Preface ix

itself. Even the most powerful neutron sources in existence today have a source
brightness that is comparable to a simple X-ray tube and many orders of
magnitude lower than a third-generation synchrotron X-ray source. Hence,
while synchrotron X-ray sources provide the capability of imaging single nano-
particles with nanometer resolution, or dynamic images of larger objects with
micrometer spatial resolution and microsecond time resolution, neutrons are
presently limited to static images with spatial resolutions of the order of tens of
microns, or dynamic images of 100 microns and microsecond exposure times
for stroboscopic processes. Neutron imaging of smaller objects can only be
achieved indirectly, using scattering techniques from an ensemble of particles.
Despite these limitations, the following chapters give a flavor of the wide
range of applications that presently (or will potentially) benefit from neutron
imaging techniques. The book is organized into three major sections.
Section A provides a comprehensive overview of basic neutron techniques
aimed more specifically at a non-specialist audience. Frequent reference is made
to the two introductory chapters in the first book of this series [2] by Roger
Pynn (Neutron Scattering – a Non-Destructive Microscope for Seeing Inside
Matter), and Helmut Schober (Neutron Scattering Instrumentation). Both these
chapters are freely available at www.springerlink.com. In Chapter 1 Kenneth
Herwig summarizes the essential neutron properties and techniques which are
relevant to the majority of neutron imaging applications. Masatoshi Arai and
Kent Crawford provide, in Chapter 2, an excellent survey of the different types
and characteristics of neutron sources, including nuclear reactors, high-power
spallation sources, and portable generators, which are typically used nowadays
for neutron imaging. Although there is considerable overlap in the use of these
sources, each type of source has specific advantages for certain types of applica-
tions. Due to the low-intrinsic brightness of neutron sources, efficient optical
systems are imperative, so Ken Andersen’s chapter (Chapter 3) presents the basic
concepts of the neutron optics that are typically used on imaging beam lines.
Finally in this section, Lowell Crow (Chapter 4) examines modern neutron detec-
tion methods for imaging. The first section reviews neutron capture converters
which form the basis for thermal neutron detection, and the following sections
examine detector systems with an emphasis on imaging applications.
Section B focuses on the neutron beam implementation of some well-known
imaging techniques. Arthur Heller and Jack Brenizer (Chapter 5) present a
summary of the history, methods, and related variations of neutron radio-
graphy techniques including a section on the application of standards. Even
today, conventional film radiography remains the mainstay of high-resolution,
large field-of-view, neutron imaging. In Chapter 6, Wolfgang Treimer extends
the basic theories and applications to include three-dimensional tomography
and introduces some of the newer methods for enhancing contrast such as
wavelength dependent (‘‘Bragg edge’’) imaging and small-angle scattering.
Kenneth Tobin et al. (Chapter 7) provide a review of neutron image formation,
resolution analysis concepts, and methods for both the design and
characterization of radiography systems and conclude with a discussion of
x Preface

volumetric reconstruction techniques using analytic or iterative computed

tomography algorithms.
The next two chapters describe techniques that depend intrinsically on the
wave nature of the neutron. Franz Pfeiffer (Chapter 8) provides a fascinating
overview of neutron phase imaging and its natural extension to neutron phase
tomography. This technique offers the potential to image fundamental quan-
tum mechanical interactions. Bhaskar Sur et al. (Chapter 9) show initial results
from neutron holography experiments using both the internal and the external
source approaches. Although in its infancy, neutron holography has the poten-
tial to resolve to atomic resolution the structures of materials, which are difficult
to crystallize. Finally in this section Nikolay Kardjilov et al. (Chapter 10)
describe some novel imaging techniques using the magnetic properties of the
neutron. After demonstrating the power of neutrons to image magnetic fields in
and around objects, they go on to describe theoretically some tantalizing but
challenging potential applications of spin contrast imaging and neutron-based
magnetic resonance imaging.
Section C provides the reader with an excellent, though non-exhaustive,
overview of some specific applications of neutron imaging in diverse fields of
research. Muhammad Arif et al. (Chapter 11) describe in situ neutron radio-
graphy and tomography studies of operating fuel cells and hydrogen storage
systems, undoubtedly a high priority global research field for the foreseeable
future, where the ability of neutrons to ‘‘see’’ hydrogen provides essential
information to improve the practical chemical and mechanical engineering
design.
Dayakar Penumadu (Chapter 12) provides an overview of some recent
applications of neutron imaging methods to broad classes of materials science
and engineering studies including metal casting, strain imaging, and character-
ization of discrete particle systems.
Carla Andreani et al. (Chapter 13) discuss the growing use of neutrons to
image artifacts of interest in the domain of cultural heritage. Novel character-
ization methods allow determination of the provenance of ancient objects and
can shed light on the methods and tools used at the time of manufacture.
In Chapter 14 Kenneth Watkin et al. summarize past and recent research
efforts to apply neutron radiography to biological specimens, in the expectation
that clinical and medical research, as well as forensic science, may benefit from
advanced neutron imaging methods.
Moving on to prospective techniques for live imaging, Anuj Kapadia
describes (in Chapter 15) the development of a tomographic technique that
uses neutron inelastic scatter interactions to quantitatively identify the spatial
distribution of elements in the body. The technique, called Neutron Stimulated
Emission Computed Tomography (NSECT), uses a beam of fast neutrons to
excite stable isotopes of elements in the body to determine their concentration
and spatial distribution within the body. It has the potential to diagnose several
element-related disorders in humans that are characterized by a change in
element concentration in the diseased tissue.
Preface xi

In Chapter 16 William Heller and Gary Baker describe how small-angle

neutron scattering measurements when combined with advanced computer
modeling and contrast variation methods enable visualization of the structure
and function of biological macromolecules.
Chapter 17 by Tomoko Nakanishi shows some fascinating examples of
neutron imaging applied to plant physiology which allow in situ studies of
vital processes in plant growth to be visualized.
Finally in Chapter 18 Dick Lanza summarizes security-related applications
of neutron imaging, e.g., for the detection of illicit materials such as explosives
or nuclear materials. One particular technique, neutron resonance radiography,
is discussed in detail, as it illustrates many of the issues connected with imaging
and detection.
We have attempted to provide a broad overview of the potential of neutron
imaging methods. It is evident that in the process we will have omitted some
important areas of application. Nevertheless we hope that you enjoy this edition
and will be stimulated to read further.
The assistance of the Springer team during the various stages of the devel-
opment of this book is greatly appreciated.

Oak Ridge, TN Ian Anderson

Oak Ridge, TN Hassina Bilheux
Oxfordshine, UK Robert McGreevy

References
1. H. Kallman, Research 1, 254 (1947).
2. http: //www.springer.com/series/8141, this series of books.
3. O. Peter, Naturforsch. I(10), 557 (1946).
4. G. B. Winkelman, C. Dwyer, T.S. Hudson, D. Nguyen-Manh, M. Döblinger, R.L. Satet,
M.J. Hoffmann, D.J.H. Cockayne, Phil. Mag. Lett. 84, 755–62 (2004).
Contents

Section A

1 Introduction to the Neutron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

K.W. Herwig

2 Neutron Sources and Facilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

M. Arai and K. Crawford

3 Neutron Optics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
K. Andersen

4 Neutron Detectors for Imaging. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

L. Crow

Section B

5 Neutron Radiography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
A.K. Heller and J.S. Brenizer

6 Neutron Tomography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
W. Treimer

7 Mathematics of Neutron Imaging. . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

K.W. Tobin, P.R. Bingham, and J. Gregor

8 Neutron Phase Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129

F. Pfeiffer

9 Thermal Neutron Holography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153

B. Sur, R.B. Rogge, V.N.P. Anghel, and J. Katsaras

xiii
xiv Contents

10 Novel Imaging Techniques: Polarized Neutrons

and Neutron-Based Magnetic Resonance Imaging . . . . . . . . . . . . . . . 171
N. Kardjilov, W.T.H. Lee, and G.E. Granroth

Section C

11 Neutron Imaging for the Hydrogen Economy. . . . . . . . . . . . . . . . . . . 191

M. Arif, D.S. Hussey, and D.L. Jacobson

12 Material Science and Engineering with Neutron Imaging. . . . . . . . . . 209

D. Penumadu

13 Novel Neutron Imaging Techniques for Cultural

Heritage Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
C. Andreani, G. Gorini, and T. Materna

14 Probing the Potential of Neutron Imaging for Biomedical

and Biological Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
K.L. Watkin, H.Z. Bilheux, and J.F. Ankner

15 Neutron Stimulated Emission Computed Tomography: A New

Technique for Spectroscopic Medical Imaging . . . . . . . . . . . . . . . . . . 265
A.J. Kapadia

16 Visualizing Structures of Biological Macromolecules Through

Indirect Imaging with Small-Angle Neutron Scattering
and Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289
W.T. Heller and G.A. Baker

17 Neutron Imaging Applied to Plant Physiology . . . . . . . . . . . . . . . . . . 305

T.M. Nakanishi

18 Homeland Security and Contraband Detection . . . . . . . . . . . . . . . . . 319

R.C. Lanza

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339
Contributors

K. Andersen Institut Laue-Langevin, 6 rue Jules Horowitz, B.P. 156, 38042

Grenoble cedex 9, France
C. Andreani University of Milano Bicocca, Department of Materials Science,
Via Roberto Cozzi 53, 20125 Milano, Italy
V. N. P. Anghel Atomic Energy of Canada Limited, Chalk River Laboratories,
Chalk River, Ontario, Canada
J. F. Ankner Oak Ridge National Laboratory, Spallation Neutron Source, Oak
Ridge, TN, USA
M. Arai J-PARC Center, JAEA, 2-4 Shirakata-Shirame, Tokai Ibaraki, Japan
M. Arif National Institute of Standards and Technology, Gaithersburg, MD,
USA
G. A. Baker Oak Ridge National Laboratory, Oak Ridge, TN, USA
H. Z. Bilheux Oak Ridge National Laboratory, Spallation Neutron Source,
Oak Ridge, TN, [email protected]
P. R. Bingham Image Science and Machine Vision Group, Oak Ridge National
Laboratory, Oak Ridge, TN, USA
J. S. Brenizer Pennsylvania State University, University Park, PA, USA
K. Crawford Spallation Neutron Source, Ok Ridge National Laboratory, Oak
Ridge, TN, USA
L. Crow Neutron Facilities Development Division, Oak Ridge National
Laboratory, Oak Ridge, TN, USA
G. Gorini University of Milano-Bicocca–Piazza dell’Ateneo Nuova, Milano, Italy
G. E. Granroth Neutron Scattering Sciences Division, Oak Ridge National
Laboratory, Oak Ridge, TN, USA
J. Gregor Department of Electrical Engineering and Computer Science,
University of Tennessee, Knoxville, TN, USA

xv
xvi Contributors

A. K. Heller Department of Mechanical and Nuclear Engineering, The

Pennsylvania State University University Park, PA, USA
W. T. Heller Center for Structural Molecular Biology, Oak Ridge National
Laboratory, P.O. Box 2008, Oak Ridge, TN 37931, USA
K. W. Herwig Oak Ridge National Laboratory, Oak Ridge, TN, USA
D. S. Hussey National Institute of Standards and Technology, Gaithersburg,
MD, USA
D. L. Jacobson National Institute of Standards and Technology,
Gaithersburg, MD, USA
A. J. Kapadia Duke Advanced Imaging Laboratories, Department of
Radiology, Duke University Medical Center, Durham, NC, USA
N. Kardjilov Helmholtz Centre Berlin for Materials and Energy, Berlin, Germany
J. Katsaras National Research Council, Chalk River Laboratories, Chalk
River, ON, K0J 1J0, Canada
R. C. Lanza MIT Department of Nuclear Science and Engineering,
Cambridge, MA, USA
W. T. H. Lee Neutron Scattering Sciences Division, Oak Ridge National
Laboratory, Oak Ridge, TN, USA
T. Materna Institut für Kernphysik, Universität zu Köln, Köln, Germany
T. M. Nakanishi Graduate School of Agricultural and Life Sciences, The
University of Tokyo, Yayoi, Bunkyo-ku, Tokyo, Japan
D. Penumadu University of Tennessee-Knoxville, Joint Institute for Advanced
Materials (JIAM), Department of Civil and Environmental Engineering,
Knoxville, TN, USA, [email protected]
F. Pfeiffer Department of Physics (E17), Technical University Munich,
Munich, Germany
R. B. Rogge National Research Council of Canada, Chalk River Laboratories,
Chalk River, Ontario, Canada
B. Sur Atomic Energy of Canada Limited, Chalk River Laboratories, Chalk
River, Ontario, Canada
K. W. Tobin Image Science and Machine Vision Group, Oak Ridge National
Laboratory, Oak Ridge, TN, USA
W. Treimer Department of Mathematics, Physics and Chemistry, University of
Applied Sciences (TFH) Berlin, Luxemburger Str., Berlin, Germany, treimer@tfh-
berlin.de; Helmholtz Center Berlin for Materials and Energy, Department SF1,
Glienicker Str., Berlin, Germany, [email protected]
K. L. Watkin University of Illinois, College of Applied Health Science, 901 So.
Sixth, Champaign, IL, USA
Section A
Chapter 1
Introduction to the Neutron

K.W. Herwig

Abstract As a fundamental particle, the neutron has many unique attributes

that provide a variety of contrast mechanisms enabling many imaging techni-
ques that will be discussed throughout this book. This chapter introduces the
neutron with a short summary of its discovery and continues with a presenta-
tion of its fundamental properties. A physical picture of the attenuation of a
primary beam of particles and the mathematical formulae that govern that
attenuation are presented. This chapter concludes with a discussion of neutron
interactions with matter that provide the physical basis underlying neutron
imaging.

Keywords Neutron  Scattering  Absorption  Cross section

1.1 Introduction

As one of the constituents of the atom, the neutron was discovered well
after its electrically charged companions the electron, by J. J. Thomson in
1897 [1], and the proton, for which there is some debate over any individual
discoverer. A close combination of an electron and a proton was proposed
as a neutral doublet by E. Rutherford in 1920, which he postulated would
have ‘‘novel properties’’ including the ability to ‘‘move freely through mat-
ter,’’ be difficult to detect, and perhaps impossible to contain in a vessel [2].
The term ‘‘neutron’’ was apparently first recorded in the literature in 1921
by W. D. Harkins in a discussion on classifying isotopes [3]. However,
conclusive evidence for the existence of the neutron waited until 1932
when J. Chadwick reported first the ‘‘possible existence’’ [4] and shortly
thereafter ‘‘the existence’’ of the neutron [5]. Chadwick was recognized for
his discovery by receiving the 1935 Nobel Prize in Physics quite simply ‘‘for

K.W. Herwig (*)

Oak Ridge National Laboratory, P.O. Box 2909, Oak Ridge, TN 37831, USA
e-mail: [email protected]

I.S. Anderson et al. (eds.), Neutron Imaging and Applications, 3

Neutron Scattering Applications and Techniques, DOI 10.1007/978-0-387-78693-3_1,
Ó Springer ScienceþBusiness Media, LLC 2009
4 K.W. Herwig

the discovery of the neutron.’’ Electrical neutrality is perhaps the signature

fundamental property of the neutron that makes it so useful in both neutron
scattering and in neutron imaging applications. As Rutherford anticipated,
because the neutron is uncharged, it interacts primarily with nuclei; it is
highly penetrating and well able to investigate the interior of large assem-
blies; and as a significant additional benefit, it does so non-destructively.
Another fundamental property of the neutron is its mass, mn =
1.6749  10–27 kg [6], which gives the neutron a de Broglie wavelength on
the order of atomic distances for energies comparable to the excitations in
condensed matter at room temperature. Thus the neutron is simultaneously
sensitive to both length and time scales relevant to condensed matter. This is
particularly relevant to neutron scattering applications; but the microscopic
response to length scales also plays a significant role in imaging applications,
particularly for polycrystalline materials in which the primary attenuation is
due to Bragg scattering from the crystal lattice. The de Broglie wavelength, l
in units of nm, is given by

h 395:6
l¼ ¼ (1:1)
mn v v

where h = 6.6261  10–34 J  s is Planck’s constant and v is the neutron velocity

in m s–1. The neutron energy, E in units of meV, is given by

1
E ¼ mn v2 ¼ 5:2270  106 v2 : (1:2)
2

A neutron with an energy characteristic of room temperature, 300 K or

25.85 meV, will have a velocity of 2224 m s–1 and a wavelength of 0.18 nm.
The neutron carries momentum as expressed by its wave vector k. ~ The direction
of k~ is that of the neutron, while its magnitude jkj
~ ¼ 2p=l. Other basic proper-
ties of the neutron include its intrinsic spin of ½ and an associated nuclear
magnetic moment of –0.9662  10–26 J T–1. Thus in addition to nuclear inter-
actions, the neutron will respond to external magnetic fields and interact with
the magnetic moments of unpaired electrons in matter.
Neutrons are typically produced either by fission in nuclear reactors or by
spallation, in which a high-energy proton beam impinges a heavy metal target
and neutrons are ejected from the resulting excited nuclei. In both of these cases,
the neutrons have spectrums of energies that are too high to be useful for either
neutron scattering or imaging. Moderators are employed to cool the neutron
spectrum to lower energies that depend on the temperature of the moderating
medium, often hydrogen or hydrogenous materials. Neutrons are rather loosely
assigned to bands of energies associated with characteristic moderator tempera-
tures, as shown in Table 1.1, taken from [7]. See Chapter 2 for more details on
sources and moderators.
1 Introduction to the Neutron 5

Table 1.1 Neutron characteristics at various energy ranges

Neutron classification Energy (meV) Velocity (m/s) l (nm)
Ultra-cold 0.00025 6.9 57
Cold 1 437 0.9
Thermal 25 2187 0.18
Epithermal 1000 13,832 0.029

1.2 Interactions with Matter and Cross Sections

In neutron imaging applications, one is primarily interested in neutron interac-

tions that attenuate a well-defined incident beam of neutrons. Neutrons can be
removed from the incident beam either by absorption or by a change in direc-
tion as they interact with material in the beam. For the moment, consider the
neutron as a classical point-size particle and the atoms in a sample as classical
particles of fixed radius, r. Figure 1.1 illustrates the attenuation of a (well-
collimated) neutron beam incident perpendicularly on a thin sample element of
thickness dx located a distance x cm from the surface of the sample. Let dx be
thin enough that every atom in this sample element can be sampled by the
incident neutron beam (no shadowing). Define I(x) as the incident neutron flux
and I(x + dx) as the transmitted flux, both in units of neutrons cm–2s–1. Let n
be the number density of atoms in the sample in units of atoms cm–3. The

I(x) I(x+dx)

dx

Fig. 1.1 Illustration of the attenuation of an incident beam of idealized neutrons by a thin
sample
6 K.W. Herwig

attenuation of the incident neutron beam is then given by the fractional area
occupied by the classical atoms, dx n p r2 and
  
Iðx þ dxÞ ¼ IðxÞ 1  dx n p r2 ¼ IðxÞð1  dx nÞ : (1:3)

In this classical picture, the microscopic cross section, , is just the cross
sectional area of the individual sample particles, p r2, and has units of cm2.
More generally, the microscopic cross section is a measure of the effective
interaction area for the neutron with a nucleus.
From Eq. (1.3), the rate of change of I(x) is then

dIðxÞ
¼ n I ðxÞ ; (1:4)
dx

and the solution for I(x) is

IðxÞ ¼ Ii expðn  xÞ ; (1:5)

where Ii is the flux of neutrons incident on the sample. The product of n is

called the macroscopic cross section, S. For real samples with multiple elements
and isotopes, a sum over each species and all its associated cross sections must
be performed in order to calculate the sample transmission
X X
tot ¼ n j j ¼ j ; (1:6)
j j

where the sum is over all isotopes in the sample and the transmitted neutron flux
may be written as

IðxÞ ¼ Ii expðtot xÞ : (1:7)

The interaction of neutrons with nuclei is governed by quantum mechanics

and not the laws of classical physics; nonetheless, it is illustrative to explore the
magnitude of the microscopic cross section in a classical context. The typical
units for  are barns, and 1 barn = 10–24 cm2. (The whimsical name of the unit
originated with the Manhattan Project and was classified information until
after World War II. The typical effective area, 10–24 cm2, of a nuclear particle
during a collision was a relatively large interaction cross section and hence was
as ‘‘big as a barn’’ [8].) Consider a typical inter-atomic separation of 0.25 nm and
the radius determined from a cross section of 1 barn. The ratio of the inter-
atomic spacing to effective interaction radius is  44,000 and, as seen by a
neutron, there is a vast amount of effective space between scattering centers,
resulting in a correspondingly high penetration depth of neutrons into most
materials.
1 Introduction to the Neutron 7

A neutron may interact with a nucleus by either scattering or absorption and

it is typical to write a total microscopic cross section, t, as the sum of these
terms

t ¼ s þ a (1:8)

where s and a are the scattering and absorption cross sections, respectively.
An in-depth introduction to neutron scattering is provided by R. Pynn [9], so
only a few relevant terms are presented here. Because the neutron interaction
is with the nucleus, the strength of the scattering cross section varies non-
uniformly across the periodic table and depends not only on the individual
element but also on the particular isotope of that element. Figure 1.2 shows
the total neutron scattering and absorption cross sections for bound atoms
across the periodic table as derived from the tabulation in [10]. In practice, the
scattering cross sections are typically taken as constant in the epithermal,
thermal, and into the cold neutron energy ranges. In general, however, these
cross sections have an energy dependence with gadolinium (due to the close
proximity of a nuclear resonance) and bound hydrogen (e.g., in the case of
water) that exhibits significant variation. (See [11] for a short discussion of this
effect.)
The neutron scattering cross section can be further broken down into
coherent and incoherent components. In a coherent scattering process,
neutron waves that are scattered by different nuclei combine with one
another to produce an interference pattern that depends on the relative
locations of the atoms in the material. Incoherent scattering may arise
when more than one isotope is present in the sample (whether the same
element or not) or when a given isotope has a non-zero nuclear spin [12]. In

Fig. 1.2 Neutron scattering and adsorption cross sections for thermal neutrons (v = 2200 m/s)
as tabulated in [10]. Open circles are adsorption cross sections and are plotted on the logarithmic
scale of the right y-axis. Closed circles are the bound atom scattering cross sections plotted on
the linear scale of the left y-axis
8 K.W. Herwig

these cases, there is a randomization of the strength of the scattering cross

section, which results in a component of the neutron scattering that depends
only on the scattering from individual atoms and does not give rise to an
interference pattern. For each of these scattering processes, the neutron may
scatter elastically, exchanging no energy with sample atoms, or inelastically,
either gaining or losing energy. During the scattering process, the neutron
may also exchange momentum with the atom, Q ~ ¼ k~f  k~i , where k~f and k~i
are the final and initial neutron wave vectors, respectively. Elastic coherent
scattering gives rise to structural information on the relative locations of
atoms in the sample; in crystalline materials, it is called ‘‘diffraction.’’
Inelastic coherent scattering yields information on collective excitations in
the sample, such as phonons and spin waves. Elastic incoherent scattering
sums over the intensity from individual atoms, whereas inelastic incoherent
scattering conveys information on single-particle excitations such as diffu-
sion, methyl group rotations and, at high energy and momentum transfers,
momentum distributions.
All of these scattering processes attenuate the incident neutron beam and
give rise to contrast, which can be used for real-space imaging. As an extreme
but very useful example, consider the transmission through a sample of water.
Hydrogen has a very large incoherent scattering cross section of approximately
80 barns for bound atoms and thermal neutrons, 10–30 times greater than
typical values for other elements. At a density  of 1 g cm–3, the hydrogen
number density may be calculated as

2 H  atoms
n¼ NA ¼ 6:691  1022 ; (1:9)
M cm3

where M is the molecular weight (18 g mol–1 for water), 2 is for the two
hydrogen atoms/water molecule, and NA is Avogadro’s number, 6.022  1023
mol–1. The macroscopic cross section due to scattering from hydrogen in water
is SH = 5.353 cm–1. In comparison, the macroscopic scattering cross sections
are SFe = 0.99 cm–1and SAl = 0.091 cm–1 for iron and aluminum, respectively.
Figure 1.3 shows the attenuation of a neutron beam due to these scattering cross
sections as a function of depth x into a sample. What is very clear from Fig. 1.3
is that the presence of hydrogen in a specimen gives rise to a very rapid
attenuation of the neutron beam. This has been used to great advantage recently
in real-time identification of the formation and motion of water in operating
proton exchange membrane fuel cells (see Chapter 11 and, e.g., [13]). The high
penetration depth of the neutron in aluminum and iron is also illustrated
(Fig. 1.1 of [9]).
Real materials are much more complicated than the cases determined above
for simple attenuation; this is particularly evident in the energy dependence of
the total scattering cross section for polycrystalline materials. Figure 1.4 shows
the measured elastic cross section for polycrystalline iron [14]. The coherent
nature of the scattering and the regular positions of the iron atoms in the
1 Introduction to the Neutron 9

Fig. 1.3 Attenuation as a function of depth into samples of water, iron, or aluminum due to
neutron scattering

Fig. 1.4 The total neutron scattering cross section for polycrystalline iron

crystalline lattice give rise to abrupt changes in the cross section and associated
neutron transmission. Associated with a given spacing between planes of atoms
in the material, d is a Bragg reflection whose scattered intensity is determined by
the local arrangement of atoms within a unit cell. The condition for neutron
scattering is determined by Bragg’s law

nl ¼ 2d sinðÞ ; (1:10)
10 K.W. Herwig

where  is half the angular separation between the incident and scattered
neutron beams. There is clearly a maximum neutron wavelength lmax for any
given d at which the condition of Eq. (1.10) is satisfied, namely lmax = 2d at
 = 908. For shorter wavelengths and lower scattering angles, there will also be
neutron scattering that satisfies this condition. Thus there is a sharp edge at lmax
for each family of reflections allowed by the crystal structure, followed by a
decreasing tail at shorter l as seen in Fig. 1.4. This wavelength dependence of
the neutron cross section and associated transmission can be used to great
advantage in optimizing the conditions for high contrast from various materials
in a sample (Neutron Tomography (Chapter 6) and Material Science and
Engineering with Neutron Imaging (Chapter 12) and, e.g., [15]). Radiographs
at different neutron wavelengths can be obtained through either time-of-flight
means (spallation neutron sources) or monochromatization (reactor sources).
Neutron absorption, another neutron–nuclei interaction that attenuates an
incident neutron beam, is a mechanism whereby contrast can be obtained. In
this interaction, a neutron is absorbed by a nucleus, often resulting in an
unstable, radioactive nucleus that will decay with a particular half-life. The
half-life is the time required for the radioactivity to decay to half of its original
value. As the nucleus decays, it may emit a variety of secondary radiation,
including an -particle, which is identical to a helium nucleus and carries two
units of positive charge; a -particle, which is an energetic electron or positron;
-rays, which are high-energy photons originating in the nucleus; and others.
An -particle interacts strongly with matter and is rapidly attenuated even in
air, whereas the latter two can penetrate relatively deeply into matter. Emitted
–rays have a well-defined and catalogued energy spectrum that can act as a
signature for the decay of particular isotopes. One mature technique is neutron
activation analysis, whereby decay radiation, either delayed or prompt, is
detected from nuclei that have absorbed a neutron. Prompt decay gives rise to
an imaging paradigm whereby a sample rastered through a neutron beam can
enable a quantitative measure of the presence of particular isotopes in the
volume sampled by the neutron beam (Novel Neutron Imaging Techniques
for cultural Heritage Objects (Chapter 13)). An energy sensitive -detector can
register the energy of the promptly emitted , yielding a fingerprint of the
isotope. In the low epithermal through cold neutron energy ranges, these
absorption processes follow a 1/v probability law for most nuclei. The value
varies across the periodic table as demonstrated in Fig. 1.2 for a neutron
velocity of 2200 m/s. The probability of absorption then increases linearly
with increasing neutron wavelength. Exceptions include gadolinium, which
has many resonances; the lowest occurs at approximately 2 eV, with a corre-
sponding nonlinear response in the scattering cross section as well.
The half-life of the activated isotopes may be long enough that it is practical
to remove an object exposed to a neutron beam and use its residual activity to
expose either a photographic film or an image plate. The radiation emitted from
particular isotopes will be present to a lesser or greater degree after neutron
irradiation, depending on their respective half-lives and the intervening time
1 Introduction to the Neutron 11

before the film or image plate is exposed by the sample. Multiple exposures at
subsequent times will be more sensitive to the longer-lived isotopes as the
shorter-lived ones decay to low levels of activity, giving rise to an additional
means of obtaining isotopic contrast.
An additional neutron–nuclei interaction involves a high-energy neutron
(MeV) that scatters inelastically, imparting a relatively large energy to the
nucleus. The excited nucleus may give up the energy in a quantized fashion
through -emission. Similarly to prompt  detection, an energy-sensitive detec-
tor can be used to identify the signatures of particular isotopes. This is the
underlying principle of neutron stimulated emission computed tomography,
described in Chapter 15 of this book.
Finally, to this point, only neutron interactions with nuclei have been con-
sidered; but, as mentioned, the neutron also carries a magnetic moment that can
interact with magnetic fields. This provides an additional means by which
neutrons can be scattered out of an incident beam as they interact with the
magnetic fields produced by unpaired electrons in materials. This interaction
may be elastic, giving rise to magnetic Bragg peaks that yield information on the
magnetic order in a sample, or inelastic, giving insight into fluctuations. A
discussion of the use of magnetic scattering and the production and use of
polarized neutron beams is given by R. Pynn in his presentation on neutron
scattering techniques [9]. Some examples of applications to neutron imaging are
given in Chapter 10.
In conclusion, the neutron has several intrinsic properties that make it very
useful for imaging purposes and result in a variety of contrast mechanisms. It is
uncharged and is able to penetrate deeply into matter in a non-destructive
fashion. The neutron cross sections for interactions with nuclei are both isotope
and neutron energy dependent and include processes that attenuate a neutron
beam by either scatter or absorption. It has a wavelength/energy relationship
such that the neutron simultaneously responds to both length and time scales
relevant to condensed matter at room temperatures. Finally, the activation of a
nucleus upon neutron absorption can yield emission of -rays having energies
that are isotope dependent.

References
1. J.J. Thomson, Cathode Rays, The Electrician 39, 104–109 (1897).
2. E. Rutherford, Proc. Roy. Soc. A 97, 374–400 (1920).
3. W.D. Harkins, Natural systems for the classification of isotopes, and the atomic weights of
pure atomic species as related to nuclear stability, J. Am. Chem. Soc. 43, 1038–1060 (1921).
4. J. Chadwick, Possible existence of a neutron, Nature 129, 312 (1932).
5. J. Chadwick, The Existence of a neutron, Proc. Roy. Soc. A 136, 692–708 (1932).
6. All values for fundamental constants taken from the NIST Reference on Constants, Units,
and Uncertainty, available at http://physics.nist.gov/cuu/Constants/.
7. D.L. Price and K. Sköld in Neutron Scattering, edited by K. Sköld and D.L. Price
(Academic Press, Orlando, Fl., 1986), Vol. A.
12 K.W. Herwig

8. M. Perricone, Symmetry 3, 4 (2006), available on-line at http://symmetrymagazine.org.

9. http://www.springer.com/series/8141, this series of books.
10. V.F. Sears, Neutron News 3, 29–37 (1992).
11. J.R.D. Copley, Neutron News 18, 30–31 (2007).
12. See the following for discussions on the origin of coherent and incoherent scattering cross-
sections R.-J. Roe, Methods of X-Ray and Neutron Scattering in Polymer Science (Oxford
University Press, New York, 2000), G.L. Squires, Introduction to the Theory of Thermal
Neutron Scattering (Cambridge University Press, Cambridge, 1978), and V.F. Sears,
Neutron Optics: An Introduction to the Theory of Neutron Optical Phenomena and their
Applications (Oxford University Press, Oxford, 1989).
13. M.A. Hickner, N.P. Siegel, K.S. Chen, D. s. Hussey, D.L. Jacobson, and M. Arif, J.
Electrochem. Soc. 155, B294–B302 (2008).
14. J.A. Harvey, H.A. Mook, N.W. Hill, and O. Shahal in Nuclear Data for Science and
Technology: Proceedings of the International Conference, Antwerp, 6–10 September 1982,
edited by K.H. Bhockhoff (D. Reidel, Boston, 1983), p. 961, and available at http://www.
nndc.bnl.gov/exfor/exfor00.htm along with many other measured cross-sections.
15. W. Kockelmann, G. Frei, E.H. Lehmann, P. Vontobel, and J.R. Santisteban, Energy-
selective neutron transmission imaging at a pulsed source, Nucl. Instrum. Meth. Phys.
Res. Sect. A 578, 421–434 (2007).
Chapter 2
Neutron Sources and Facilities

M. Arai and K. Crawford

Abstract This chapter provides a brief survey of the types of neutron sources
that are available for imaging applications, primarily focusing on high-flux
sources such as reactors or spallation sources but also including smaller or
portable sources based on radioactive decay or small accelerators. Although
nearly all neutron imaging studies performed so far have used portable or
reactor-based sources, spallation sources will be of increasing importance in
the future. Their designs are less familiar and somewhat more complex, so the
chapter concentrates on them.

Keywords Reactor source  Spallation source  Portable source  Neutron

energy  Pulse  Pulse width  Thermal neutron  Moderators  Reflectors 
Neutron production  Fission  Spallation  Proton  Accelerator  Instrument 
Time-of-flight  Flux

2.1 Introduction

One might think it is not necessary to have any detailed knowledge of the
neutron source in order to use the techniques of neutron scattering or imaging.
To an extent this is true. However, the distribution of neutrons from the
source in terms of energy and time, and the distribution of ‘‘background’’
(fast neutrons, delayed neutrons, gammas, . . .), has a direct bearing on the
design of neutron scattering and imaging instruments and their performance,
on how measured ‘‘raw’’ data must be corrected to make them scientifically
meaningful, and on the types of measurements that can be undertaken. Even if
the average user does not design the instrument, the user needs to choose the
source/instrument to be used for a particular experiment or measurement, so

M. Arai (*)
Japan Proton Accelerator Research Complex Center, Japan Atomic Energy Agency,
Japan
e-mail: [email protected]

I.S. Anderson et al. (eds.), Neutron Imaging and Applications, 13

Neutron Scattering Applications and Techniques, DOI 10.1007/978-0-387-78693-3_2,
Ó Springer ScienceþBusiness Media, LLC 2009
14 M. Arai and K. Crawford

some knowledge of the different types of sources is advisable. There are two
kinds of neutron sources for neutron scattering and high-resolution/high-flux
imaging facilities: (1) reactor sources and (2) large accelerator-based neutron
sources. However, low-resolution/low-flux imaging can also be carried out
with small radioactive or accelerator-based sources, which can be fixed or
portable.
Nuclear reactors use the fission process to produce neutrons. Most of the
current reactor sources for scattering applications were built in the 1960s and
1970s and were primarily designed for materials testing for the nuclear
industry, providing medium flux. The best reactor source, optimized for
neutron scattering applications, is still the High-Flux Reactor (HFR) at the
Institut Laue-Langevin (ILL), built in 1972 in Grenoble, France [1, 2].
Reactors coming into operation more recently include JRR3 at the Japan
Atomic Energy Agency (1990) [3], HANARO at the Korea Atomic Energy
Research Institute (1997) [4], FRM-II in Munich, Germany (2004) [5], OPAL
at the Australian Nuclear Science and Technology Oganisation (2006) [6],
and the China Advanced Research Reactor in Beijing, China (potential
operation in 2008) [7]. With the exception of the HFR, these are all med-
ium-sized research reactor sources (typical power 10–20 MW) built with
advanced technology.
Electron accelerators can produce neutrons in a target material using the
Bremsstrahlung photo-neutron reaction. Although electron accelerators are
relatively inexpensive to construct, the large amount of heat dissipated in
the target per neutron produced severe limits on the potential performance.
Neutrons can also be produced by the spallation process, in which high-energy
protons strike a solid target. The development of proton accelerator technology,
driven by other applications such as particle physics, helped to increase the
potential power and hence the neutron flux of spallation sources. The pulsed
nature of most accelerator-based neutron sources can offer a significant advan-
tage in experiments using the time-of-flight (TOF) method, in which the speed of
the neutron is measured by timing its flight from the source to the detector. Pulsed
proton-driven neutron sources recently completed or under construction include
the Spallation Neutron Source (SNS) in the United States [8, 9], the Japanese
Spallation Neutron Source (JSNS) of the Japan Proton Accelerator Research
Complex (J-PARC) project in Japan (2008) [8, 9], and the China Spallation
Neutron Source (CSNS) in China (under construction) [10, 11].
Both fission and spallation produce neutrons in the megaelectron volt energy
range as discussed in Section 2.2. However, neutron scattering and most
neutron imaging applications require neutrons at electron volt or lower
energies. Moderators are used to slow the neutrons to these energies, as
described in the section ‘‘Moderation Mechanisms.’’ Reactor technology is
highly developed and has not changed significantly for many years. The tech-
nologies for accelerator-driven sources offer many more parameters that can be
varied to optimize the source for different purposes, so more space will be
devoted to describe these options in this chapter.
2 Neutron Sources and Facilities 15

2.2 Neutron Production

2.2.1 Reactors

At research nuclear reactors, neutrons are produced by the well-known fission

process

235
U þ n ! X þ Y þ 2:5 n ;
(2:1)
ð 200 MeV total energy release;  2 MeV per neutronÞ

where X and Y are fission fragments or atoms of smaller atomic weight.

Reactors are designed and optimized for different purposes. The exact design
is dependent on a number of features—the shape and size of the core, the
arrangement and type of the fuel elements, control rods, coolant, moderator,
reflectors, and beam tubes. Power reactors are optimized for heat extraction
and efficient use of fuel, so they have quite a different design from research
reactors that are optimized for high (external) thermal neutron flux. In a typical
research reactor design, one of the neutrons produced per fission is needed to
sustain the chain reaction, 0.5 is lost, and one is available for external use (i.e.,
200 MeV of heat is produced for each available neutron).
Recent design innovations have made a compact reactor core with high
enrichment capabilities practical, and this in turn produces very high neutron
fluxes outside the core for beam tube applications, as is the case at the FRM-II
reactor in Munich (Fig. 2.1). The highly optimized combination of core design
and moderator arrangement makes the thermal neutron flux available for
experiments comparable to that of the world’s preeminent research reactor
facility for neutron scattering at the ILL I in Grenoble—8  1014 ncm–2s–1for
20-MW reactor power compared with 1.5  1015 ncm–2s–1for the ILL power of
58 MW. Table 2.1 shows the reactor power and source flux for the operating
world-class research reactor sources (for scattering applications) along with
some other parameters that will be discussed later in the chapter.
The 1980s Advanced Neutron Source project in the United States attempted
to design a high-flux reactor of significantly higher power than ILL. However, it
became clear that the technical challenges, coupled with increasing safety
regulation, made this effort economically unviable [13, 14]. The future devel-
opment of higher-flux neutron sources will therefore be based on accelerators.

2.2.2 Proton Accelerator-Based Sources

High-energy protons can create large numbers of ‘‘spalled’’ neutrons from

bombardment of heavy nuclei. For example, a 1-GeV proton is capable of
producing approximately 25 neutrons from a lead target, with heat deposition
in the target of about half of the proton beam power—meaning one order of
16 M. Arai and K. Crawford

(a)

(b)

Fig. 2.1 (a) Layout of the reactor pool of FRM-II. The reactor core is very compact—24 cm in
diameter—so that the maximum neutron flux is achieved at the moderator position, as shown
in (b), giving a very high flux for instruments [5]

magnitude less heat must be dissipated than in a fission reaction producing the
same time-averaged neutron flux. Spallation reactions occur for proton ener-
gies above 100 MeV. High-energy neutrons, pions, and spalled nuclei cause
inter-nuclear cascades followed by low-energy neutron evaporation from the
excited nuclei, as illustrated in Fig. 2.2.
 Table 2.1 Existing medium- and high-flux reactor sources and their respective parameters
United United
Country States States Canada France France Germany Germany Australia Korea Japan
Neutron HFIR NBSR NRU HFR ORPHEE BENSC FRM-II OPAL HANARO JRR-3 M
source
Organization Oak Ridge National Atomic Institut Laue- Laboratoire Helmholtz- Technische Australian Korea Atomic Japan
National Institute of Energy of Langevin Léon Zentrum Universitat Nuclear Energy Atomic
Laboratory Standards and Canada Brillouin Berlin Munchen Science and Research Energy
Technology Limited Technology Institute Agency
Organization
Power (MW) 85 20 120 58 14 10 20 20 24 (present) 30 20
(designed)
Flux (ncm–2 1.51015 31014 31014 1.51015 31014 21014 81014 31014 21014 31014
2 Neutron Sources and Facilities

s–1)
Number of 1/0 1/0 0/0 2/1 1/1 1/0 1/1 1/0 1(planned)/0 1/0
cold/hot
sources
Number of 9(present) 24 5 26 22 22 20 (present) + 6 6 24
instruments +6 10 (under
(planned construction)
by 2012)
Existing BT-2 [14] [15] CONRAD ANTARES NR-port [18] TNRF [19]
neutron [16] [17] And
imaging TNRF-2
instrument [20]
Facility 1967 1970 1957 1972 1980 1973 2004 2006 1997 1990
operating (refurbished
since 1993)
HFIR: High-Flux Isotope Reactor [21]; NBSR: National Bureau of Standards Reactor [22]; NRU: National Research Universal Reactor, Chalk River, Canada [23]; HFR:
High-Flux Reactor at ILL [1, 2]; ORPHEE: reactor at LLB [24]; BENSC: Berlin Neutron Scattering Centre [25]; FRM-II: Forschungsneutronenquelle Heinz Maier-Leibnitz
[5]; OPAL: Open Pool Australian Light-water Reactor [6]; HANARO: High-flux Advanced Application Reactor [4]; JRR-3 M: Japan Research Reactor No. 3 Modified [3].
Consult the web sites for these facilities to obtain additional information and current details. A number of smaller research reactors, primarily at universities, are not listed here.
17
18 M. Arai and K. Crawford

Fig. 2.2 Nuclear spallation by high-energy proton bombardment [26]

Fig. 2.3 Neutron energy spectrum from mercury and PbBi alloy targets [27]

The energy of a small fraction of the neutrons produced in spallation

processes can be as high as the incident proton energy (these neutrons require
very thick shielding), but the spectrum reaches a maximum of around 2 MeV for
the evaporating neutrons, as shown in Fig. 2.3.
For efficient neutron production, as many protons as possible should
undergo high-energy collisions with nuclei rather than gradually losing energy
through other processes. The proton mean free path, dominated at high ener-
gies by energy-independent nuclear collision cross sections is approximately
200 g cm2 (dividing by the selected target material density gives units of
length). The proton stopping length, dominated at low energies by electron
excitation energy loss, depends on the material and the energy but is roughly
600 g cm2 (for lead and other heavy elements for 1-GeV protons). When the
2 Neutron Sources and Facilities 19

stopping length of a proton is greater than three times its mean free path,
neutron production efficiency becomes close to 100%, so the proton energy
should be as high as 1 GeV or greater.
In the early days of design work on spallation sources, proton energies were
less than 1 GeV because of the lack of experimental experience at higher energies.
However, more sophisticated codes and experiments in the 1990s demonstrated
that the neutron production rate is almost proportional to the accelerator power
even at 12 GeV [28, 29], as shown in Fig. 2.4. This resulted in flexibility in
optimizing accelerator and neutron target design. Proton current and energy
are equally optimized for beam experiments. Current and energy for spallation
sources are listed in Table 2.2 along with some other parameters to be discussed.
For a short-pulse neutron source, the repetition rate of the proton acceleration
is an important parameter to be considered. When a long neutron flight path is
used for improved TOF resolution, a slow repetition rate is important to mini-
mize frame overlap (where the fast neutrons from one pulse overlap with the slow
neutrons from the previous pulse). However, if the same time-averaged power is
maintained, a lower repetition rate requires more power per pulse, creating a
more difficult engineering problem for the accelerator and target.
Most accelerator-based neutron sources are pulsed, and heat is produced in
the target only during the pulses. This allows the heat to dissipate slowly in the
period between pulses, so the instantaneous power and neutron flux can be very
high. However, thermal shock in the target remains a problem to be overcome
at the highest levels of proton power. Building spallation neutron sources

Fig. 2.4 Neutron yield from a lead target as a function of proton energy. Neutron yield is
almost proportional to proton energy at up to 12 GeV. Here the symbols show experimental
results, whereas the lines show the results of various calculations [28, 29]
 20

Table 2.2 Past, existing, and future spallation source and their respective parameters
United United United
Country States States States U.K. Switzerland China Europe Japan Japan
Neutron source IPNS LANSCE SNS ISIS SINQ CSNS ESS KENS JSNS
Organization Argonne Los Alamos Oak Ridge Rutherford Paul Scherrer Institute Undecided High Energy Japan
National National National Appleton Institute of High Accelerator Atomic
Laboratory Laboratory Laboratory Laboratory Energy Research Energy
Physics Organization Agency
Proton energy (MeV)/ 450/15 800/70 1000/1400 800/200 590/1500 1600 1333/7500 500/9 3000/333
Current (mA)
Proton beam power 7 kW 56 kW 1.4 MW 160 kW 1 MW 100 kW 5 MW 4.5 kW 1 MW
Repetition rate (Hz) 30 20 60 50/10 (2 Continuous 25 16 (long 20 25
targets) pulse)
Target material Depleted Tungsten Mercury Tantalum Zircaloy Tungsten Mercury Tungsten Mercury
Uranium
Moderator S-CH4/L- L-H2/H2O L-H2/H2O L-H2/L-CH4/ L-D2/D2O H2OL- L-H2 S-CH4/H2O L-H2
CH4 H2O CH4L-
H2
Number of 12 7 24 (beam 22 (TS1) 15 20 (beam 15 23 (beam
instruments ports) 7 (TS2) ports) ports)
Existing neutron NEUTRA
imaging instrument [30] and
ICON [31]
Facility operating 1981 (closed 1983 2006 1985 (TS1) 1996 2014 Under 1980 (closed 2005) 2008
since or planned to 2008) 2008 (TS2) planning
operate in
IPNS: Intense Pulsed Neutron Source [32]; LANSCE: Los Alamos Neutron Science Center [33]; SNS: Spallation Neutron Source [8, 9]; ISIS: [34, 35]; SINQ: Swiss Spallation
Neutron Source [36, 37]; CSNS: Chinese Spallation Neutron Source [10, 11]; ESS: European Spallation Source [38, 39]; KENS: Koh-Energy-ken Neutron Source [40, 41];
JSNS: Japanese Spallation Neutron Source [8, 9]. Consult the websites for these facilities to obtain additional information and current details.
M. Arai and K. Crawford
2 Neutron Sources and Facilities 21

instead of reactors is therefore becoming a worldwide trend, as demonstrated by

the SNS in the United States [8, 9] and the J-PARC project in Japan [42]. China
has started construction of the CSNS [10, 11], and European countries are
planning a long-pulse neutron source, the European Spallation Source (ESS)
[38, 39], in the near future.
So far, short-pulse spallation neutron sources, typically delivering 1-ms pro-
ton pulse widths, have predominated because of the good timing resolution
provided for TOF measurements of the neutron energy. Ring structures (syn-
chrotrons or accumulator/storage rings) in the accelerator system are required
to provide high proton intensities in such short pulses. Figure 2.5 shows an
example of one such accelerator system using an accumulator ring.

Storage
ring

Linac

Target
Station

Fig. 2.5 A typical short-pulse spallation neutron source, the SNS facility, (Oak Ridge
National Laboratory, United States) [8, 9]

Long-pulse sources, typically having 1-ms proton pulse widths, are another
option rapidly gaining favor for neutron sources not requiring such high timing
resolution. No accumulator ring is required for such sources, so a very high-
intensity direct proton beam can be delivered from the linac to the neutron target
[43]. Continuous cyclotron-based spallation sources, such as that at the SINQ at
the Paul Scherrer Institute in Switzerland [36, 37], produce continuous neutron
beams; their performance is similar to that of a medium-flux reactor. Table 2.2
shows spallation neutron sources either in operation or being planned.

2.3 Moderation Mechanisms for Reactors and Spallation Sources

2.3.1 Reactor Neutron Sources

The cross section for neutron-induced fission is much higher for thermal
neutrons (100 meV) than for the fast neutrons (1–2 MeV) that are produced.
For a reactor to achieve a self-sustaining chain reaction from a small mass of
fissile material, and to obtain suitable neutrons for neutron scattering, the fast
neutrons within the core must be ‘‘slowed down.’’ This is done using a
22 M. Arai and K. Crawford

moderator, which is usually also the coolant. A moderator reduces the neutron
energy by inelastic scattering and so is preferably a material containing light
elements, e.g., H2O or D2O. Hydrogen has a large scattering cross section and a
large absorption cross section, so when H2O coolant is used, the core needs to be
relatively compact to achieve a high-flux density at the neutron beam tubes.
These beam lines are directed tangentially to the cylinder of maximum thermal
flux in order to reduce the background of high-energy neutrons and gammas, as
illustrated in Fig. 2.1. With D2O, a larger core can be used (meaning easier
cooling because heat can be dissipated more easily). Surrounding the moderator
there may be ‘‘reflector’’ materials that do not provide much moderation but
scatter or reflect some of the fast neutrons back into the moderator to enhance
the thermalized neutron flux output.
The neutron energy distribution can be altered from that produced by the
reactor coolant/moderator or reflector/moderator materials and design by stra-
tegic placement of special moderators, shifting the neutron energy spectrum to
either slightly lower energies (a ‘‘cold source’’—e.g., liquid hydrogen) or higher
energies (a ‘‘hot source’’—e.g., carbon heated by gamma radiation from the
reactor). Such special moderators expand the range of usable neutron energies
and hence expand the research capabilities of the reactor facility. Table 2.1 shows
which of the research reactor facilities provide such capabilities.

2.3.2 Pulsed Spallation Neutron Sources

In designing spallation sources, as much consideration should be paid to the

moderators as to the accelerator performance. The design of moderators is
strongly dependent on the kinds of instruments, resolution, and intensity that
are required. Among the most important parameters to be optimized for
moderators are (1) temperature, (2) neutronic structure, and (3) materials.
Neutrons in moderators reach thermal equilibrium after multiple scattering
events and have, in the case of sufficiently thick moderators, a Maxwellian
distribution in energy around the temperature of the moderator. On the other
hand, for pulsed sources, the time that neutrons spend in the moderator broad-
ens the pulse. Short-pulse spallation sources need to have a sharp pulse struc-
ture of thermalized neutrons, so the moderator dimensions need to be small and
are optimized at around 101010 cm3. The under-moderated neutrons result
in a rich ‘‘epithermal’’ flux, proportional to 1/E where E is the energy of the
neutrons, in the higher energy ‘‘slowing down’’ region above the Maxwellian
distribution (Fig. 2.6). Most pulsed neutron sources utilize moderators in a
‘‘wing’’ geometry so that the beam tubes are directed tangentially to the target to
minimize the flux of high-energy neutrons and gammas in the neutron beams.
The ‘‘flux trap’’ geometry is also effective, as is demonstrated by the Los
Alamos Neutron Science Center [44]. Low-temperature moderators can
extend the slowing-down region to shift the Maxwell distribution to
lower energy, although flux is sacrificed in the thermal energy region.
2 Neutron Sources and Facilities 23

Fig. 2.6 The neutron energy distribution (flux) of the J-PARC neutron source for coupled,
decoupled, and poisoned decoupled moderators. The flux consists of a Maxwell distribution
at low energies and a 1/E region at higher energies [45]

The pulse width is a key parameter for short-pulse spallation sources and is
directly influential on neutron beam instrument performance. It is almost
proportional to the neutron wavelength in the 1/E region of flux, is broadened
in the thermal equilibrium region, and then saturates in the very low-energy
region. The broadening starts to occur at about 300 meV (neutron wavelength
of 0.5 Å) for an ambient-temperature moderator and at about 15 meV (2.5 Å)
for a methane moderator at 20 K. This is clearly seen in Fig. 2.7. In the 1/E
region for each moderator, the pulse width, t, is proportional to wavelength,
l, as [46]

2
t½s  pffiffiffiffiffiffiffiffiffiffiffiffi  7l½— : (2:2)
E½eV
24 M. Arai and K. Crawford

Fig. 2.7 Pulse widths for typical moderators [45]

This proportionality between the pulse width and wavelength is of great

importance for high-resolution instruments, as will be discussed in a later section.
Pulse widths can also be reduced by surrounding the moderator with an
absorbing material (decoupler) such as cadmium on all sides except the side
from which the neutron beam emerges. This prevents the neutrons slowed down
in the reflector from entering the moderator and emerging as part of the neutron
beam. Since those neutrons would typically emerge into the beam later than
would the neutrons slowed down in the moderator, such decoupling prevents
the extra broadening of the neutron pulse that would result from such neutrons.
However, this decoupling also results in a reduced total intensity in the neutron
pulse. Still further reductions in pulse width can be achieved by placing an
absorbing material (cadmium or gadolinium) as ‘‘poisoning’’ in the moderator,
effectively reducing the moderator size for low-energy neutrons, but this again
results in a penalty in intensity. If no such absorbing materials are used, the
moderator is fully ‘‘coupled’’ to its surroundings and produces the highest
intensity but with relatively broad pulse widths.
The neutrons from small moderators are empirically understood to consist
of two components: (1) the thermal equilibrium component, the Maxwellian or
so-called storage component, and (2) the pre-equilibrium component caused
when neutrons undergoing a small number of scattering processes escape the
moderator before thermalizing, the so-called slowing down component [47].
The former component broadens the pulse width (Fig. 2.8) giving intense low-
energy flux as indicated in Fig. 2.6.
Moderator materials should be chosen that are suitable for thermalization
of neutrons with good neutron cross sections in the low-energy range.
2 Neutron Sources and Facilities 25

Fig. 2.8 Typical pulse structure from a short-pulse spallation source. The sharp rising edge
comes from the slowing down component followed by the storage component with a long time
tail [47]

Cold coupled moderators have been extensively optimized for pulsed sources
by Watanabe and Kiyanagi, greatly enhancing neutron flux from the cold
neutron moderator while minimizing heat deposition in the moderator
(Fig. 2.9) [48].

Fig. 2.9 Design of a coupled moderator for J-PARC. The hydrogen moderator is surrounded
by a water premoderator, which reduces neutron energy in the first stage and removes heat
before neutrons go into the hydrogen moderator [48]
26 M. Arai and K. Crawford

2.4 Comparison of Source Types

We can consider the case of crystalline diffraction to illustrate the differences

between reactor sources and pulsed spallation sources and the moderator
choice considerations for a pulsed source (this would also apply to, for example,
Bragg edge imaging, see Chapters 6 and12). When a diffraction measurement is
made at a reactor, the neutron energy is typically monochromatized by using
Bragg scattering from a single crystal, and neutrons scattered from the sample
are recorded as a function of angle. At a pulsed neutron source, the TOF
method naturally is used. Neutrons are counted as a function of flight time,
starting at neutron emission resulting from the proton pulse on the target.
Neutrons propagate along the primary flight path, L1; are scattered by a
sample; and are detected by a detector at a certain scattering angle and distance,
L2, as shown in Fig. 2.10. The peak pulse width is maintained during the
propagation from source to detector. Consequently, a sharp peak width from
a decoupled moderator is preferable for a high-resolution measurement with a
long flight path, leading to better separation of peaks at the detector. Alterna-
tively, a coupled moderator could be used to obtain higher flux, if a much longer
flight path could be used to maintain the resolution. Hence, essentially in a
pulsed spallation source, the longer the flight path the higher the resolution. The
transmission of modern neutron guides in the relevant wavelength range is
sufficiently good that flight paths of up to 250 m might be considered, but the
cost of the guides and associated shielding is significant.
Continuous reactor sources have the advantage of higher time-averaged
intensity. For instance, neither SNS nor JSNS will exceed the average neutron

Diffraction peaks
Flight length

Scattering
angle

Detector

L2
L2
Sample
Sample
Detectors
L1
L1

Source

Time of Flight

Fig. 2.10 Flight-length/time-of-flight diagram for a diffraction measurement

2 Neutron Sources and Facilities 27

flux of high-performance reactors such as the High-Flux Isotope Reactor at

Oak Ridge National Laboratory, the HFR, or FRM II. Furthermore, reactor
neutron sources offer high availability; for instance, FRM II operates 260 full-
power days per year. However, spallation neutron sources deliver the highest
peak flux, which, coupled with the advantages of instrumentation based on the
TOF principle (see Chapter 3 for further details), becomes the effective figure
of merit for many applications. Also, as noted earlier, spallation sources offer
the only serious opportunities for improvement of source intensities in the
future.

2.5 Neutron Facilities

Most neutron facilities are user facilities (i.e., open to external scientists). Pro-
spective users submit experimental proposals, which are evaluated by review
committees for approval. To attract a broad range of users, facilities should
offer a suitably wide variety of instruments, ease of access, good maintenance,
and high performance. The number of experiments/users that a facility can
accommodate depends on the power of the source (i.e., the speed of individual
measurements), the number of days of source operation, and the number of
instruments. Apart from relatively routine measurements that can be automated,
the minimum practical turnaround time for experiments tends to be 1–2 days, so
the capacity saturates regardless of the source power. However, higher-power
sources enable significantly more complex experiments to be carried out in the
same time (e.g., imaging at higher resolution, tomography, or the use of contrast
enhancement techniques), as described in Chapters 6 and 12.
Most modern instruments are equipped with a large number of detector
segments and easily produce nearly 1 GB or more of data in a single measure-
ment; analysis and storage of these large amounts of data are becoming a
limiting factor. Drastic improvements in accelerator or reactor performance
are not easy, technologically or financially. However, improvements in neutron
optics, detectors, and instruments have the potential to dramatically increase
the flux of useful neutrons for experiments, as has been the exemplary experi-
ence at ILL. Maintaining a suitable balance of effort among the accelerator,
target, instruments, and support activities (e.g., sample preparation, sample
environment, data analysis) is therefore an important factor in maintaining
facility capabilities.

2.6 Smaller Neutron Sources for Imaging and Other Applications

Neutrons can be produced by spontaneous fission, although this is unusual. For
example, 252Cf can decay by -decay with a half-life of 2.65 years (this produces
helium gas, which creates internal pressure in the source) and by n-decay with a
28 M. Arai and K. Crawford

half-life of 85.5 years. The average neutron energy is 2.14 MeV, and the rate is
2.34  1012 ns–1 g–1. Such sources are produced by irradiation in a high-flux
reactor.
A more common method is neutron production as a secondary process, as
used by Chadwick when he first discovered the neutron. For example, in an -n
source such as 241Am/Be, 241Am undergoes -decay; the -particle can be
absorbed by a light element such as beryllium, which then decays by neutron
emission. This can be written as

241
Am ! 237 Np þ 4 He ð5:6 MeVÞ followed by
(2:3)
9
Be þ 4 He ! 12 C þ n ðfew MeVÞ :

The half-life is 433 years, and these sources can produce 106–108 n s–1 g–1 Am.
An alternative is a -n source, for example

124
Sb !124 Te þ hð1:7 MeVÞ followed by 9 Beþ
(2:4)
h ! 2 4 He þ n ðfew MeVÞ :

The half-life is 60 days, and such sources can produce 109–1010 n s–1.
Radioactive decay sources have the advantage of being small and highly
portable, but they have low intensity and are always ‘‘on.’’ They can be used for
testing (e.g., of neutron detectors), in medicine (e.g., activation analysis, cancer
treatment with 252Cf needles), and for low-resolution/low-flux radiography.
Higher intensities can be produced by small accelerator-based neutron
sources. Over the years, these have evolved sufficiently that compact portable
sources are now commercially available from a range of vendors. They are
normally based on the ‘‘D-T’’ reaction:

2
Hð 150 keVÞ þ3 H ! 4 He þ n ð14:2 MeVÞ : (2:5)

Sealed tube sources with a typical length of 1 m and diameter of 10 cm,

operating at a power of 0.5 kW, can produce up to 3  1010 n s–1. At a distance
of 1 m, this gives a flux on the order of 2  105 n cm–2 s–1. Note that these are fast
neutrons; if they were moderated to be used for thermal neutron imaging, then
the flux would be considerably lower. They are widely used for industrial and
security applications based on fast neutron radiography (Chapter 18) and
prompt gamma activation analysis.
The Low Energy Neutron Source at Indiana University [49] is a small
cyclotron-based source that uses the reaction

pð7  13 MeVÞ þ 9 Be ! 9 B þ nð5  11 MeVÞ : (2:6)

2 Neutron Sources and Facilities 29

This source can produce a flux of moderated thermal neutrons comparable

to that produced by a small spallation source and is therefore suitable for both
scattering and imaging applications.

References
1. W.J. Drexel, IEEE Trans. Nucl. Sci. NS-29, 123–126 (1982).
2. C.J. Carlile, Physica B 385–386, 961–965 (2006).
3. Fumio Sakurai, Yoji Horiguchi, Shinsho Kobayashi, and Masaji Takayanagi, Physica B
311, 7–13 (2002).
4. Chang-Hee Lee, Young-Hwan Kang, and Il-Hiun Kuk, ICANS-XV Proceedings of the
15th Meeting of the International Collaboration on Advanced Neutron Sources, Tsukuba,
Japan, 146–154 (2000).
5. J. Neuhaus and W. Petry, Neutron News 18:2, 13–15 (2007).
6. Shane J. Kennedy, Physica B 385–386, 949–954 (2006).
7. D.F. Chen, Y.T. Liu, C. Gou, and C.T. Ye, Physica B 385–386, 966–967 (2006).
8. T.E. Mason, et al., Physica B 385–386, 955–960 (2006).
9. Tony A. Gabriel, John R. Haines, and Thomas J. McManamy, J. Nucl Mater. 318, 1–13
(2003).
10. J. Zhang, Q.W. Yan, C. Zhang, P.L. Zhang, S.N. Fu, F.W. Wang, Z. Zhang, and S.X.
Fang, J. Neutron Res. 13:1, 11–14 (2005).
11. Q.W. Yan, W. Yin, and B.L. Yu, J. Nucl. Mater. 343, 45–52 (2005).
12. Colin D. West, Physica B 174, 430–437 (1991).
13. A. Lawler, Science 279, 470–471 (1998).
14. D. Hussey et al., Nucl. Instrum. Methods Phys. Res. A 542 (2005) 9–15.
15. http://www-llb.cea.fr/neutrono/nr1.html.
16. A. Hilger et al., Physica B 385–386 (2006) 1213–1215.
17. E. Calzaba et al., Nucl. Instrum. Methods Phys. Res. A 542 (2005) 38–44.
18. I.-C. Lim, M.-S. Kim, K.-Y. Nam, C.-M. Sim, B.-C. Lee, H.-Y. Choi, and S.-Y.Hwang,
Characteristics of neutron beam from IR beam port of HANARO for its application to
dynamic neutron radiography, in Proceedings of the 2002 ANS RPD Topical Meeting, Int.
Conf. New Frontiers of Nuclear Technology, Seoul, Korea, Oct. 2002.
19. M. Matsubayashi and A. Tsuruno, JRR-3 Neutron Radiography Facility, Proceedings of
the 4th World onf. Neutron Radiography, San Francisco, California, May 10–16, 1992, p.
415, Gordon & Breach Science Publishers.
20. M. Matsubayashi et al., Nucl. Technol. 132 (2000) 309.
21. S.E. Nagler and H.A. Mook, Neutron News 19:2, 12–13 (2008); also other articles this
issue.
22. R.L. Cappelletti, C.J. Glinka, S. Krueger, R.A. Lindstrom, J.W. Lynn, H.J. Prask,
E. Prince, J.J. Rush, J.M. Rowe, S.K. Satija, B.H. Toby, A. Tsai, and T.J. Udovic,
J. Res. Natl. Instrum. Stand. Technol. 106, 187–230 (2001).
23. J. Root, Neutron News 14:2, 12–13 (2003).
24. P. Monceau and M. Alba, Neutron News 14:3, 10–12 (2003).
25. R. Michaelsen, Neutron News 12:3, 7–9 (2001).
26. N. Watanabe, Rep. Prog. Phys. 66:339–381 (2003).
27. M. Teshigawara et al., JAERI-Res. 99–0101 (1999).
28. M. Arai et al., J. Neutron Res. 8, 71–83 (1999).
29. J.M. Carpenter, et al., Physica B 270, 272–279 (1999).
30. E. Lehmann, et al., Nucl. Instrum. Meth. A 377 (1996) 11.
31. G. Kuehne et al., Nucl. Instrum. Methods Phys. Res. A 542 (2005) 264–270.
30 M. Arai and K. Crawford

32. Raymond G. Teller, James W. Richardson, and John M. Carpenter, Neutron News 15:3,
9–13 (2004).
33. J.A. Roberts, Neutron News 10:4, 11–14 (1999).
34. R. Eccleston and C. Wilson, Neutron News 15:1, 15–18 (2004).
35. Penfold, Neutron News 15:2, 9–12 (2004).
36. W. Wagner, J. Mesot, P. Allenspach, G. Kuehne, and H.M. Rønnow, Physica B 385–386,
968–971 (2006).
37. W. Wagner and H. Hayck, Proc. ICANS-XVI, ISSN 1433-559X, 53–66 (2003).
38. E. Cartlidge, Phys. World 10, June, 10 (2007).
39. D. Richter, Appl. Phys. A 74 [Supl.], S18–S22 (2002).
40. Ishikawa, ICANS-IV Proceedings of the 4th Meeting of the International Collaboration on
Advanced Neutron Sources, Tsukuba, Japan, 89–101 (1980).
41. Ikeda, ICANS-XV Proceedings of the 15th Meeting of the International Collaboration on
Advanced Neutron Sources, Tsukuba, Japan, 93–96 (2000).
42. Shoji Nagamiya, J. Neutron Res. 13:1, 7–10 (2005).
43. F. Mezei, Proceedings 12th meeting of International Collaboration on Advanced Neutron
Sources ICANS-XII, Abingdon, Oxfordshire, UK, I-377–384 (1993).
44. Takashi Ino, Motoki Ooi, Yoshiaki Kiyanagi, Yoshimi Kasugai, Fujio Maekawa,
Hiroshi Takada, Guenter Muhrer, Eric J. Pitcher, and Gary J. Russell, Nucl. Instrum.
Meth. A 525, 496–510 (2004).
45. Tamura et al., JAERI-Report 2003–008 (2003).
46. F.R. Mildner and R.N. Sinclair, J. Nucl. Energy 6, 225 (1979).
47. Ikeda and J. Carpenter, Nucl. Instrum. Meth. Phys. Res. A239 536–544 (1985).
48. Watanabe et al., Proceedings of the 10th meeting of International Collaboration on
Advanced Neutron Sources ICANS-X, Los Alamos, USA, pp.787–797 (1988).
49. Leuschner, D.V. Baxter, J.M. Cameron, V. Derenchuk, C. Lavelle, A. Lone, H. Nann,
T. Rinckel, and W.M. Snow, J. Res. Natl. Instrum. Stand. Technol. 110, 153–155 (2005).
Chapter 3
Neutron Optics

K. Andersen

Abstract An overview is given of neutron optics, aiming at a qualitative over-

view of the main concepts and components currently used for designing the
optics of neutron instruments. The section on source optics shows the bright-
ness curves of the Institu Laue-Langevin and ISIS cold sources and includes a
short discussion of Liouville’s theorem. Neutron guides and supermirrors are
dealt with in the section of reflection optics. In the diffraction optics section,
I describe the concepts of mosaic and bent-perfect crystals, giving reference
numbers for their performance and also attempting to cover some basic focus-
ing concepts. There is a very short section on refractive optics, covering both
materials and magentic lenses. The final section on polarisation optics is more
detailed, partly because of the current rate of progress in the field and partly
because it is an area with which the general reader may be less familiar.

Keywords Neutron optics  Neutron polarisation  Neutron instumentation

3.1 Introduction

Throughout this chapter, I restrict the descriptions to the qualitative ‘hand-

waving’ type. Neutron optics is a large field, which is covered in much more
detail elsewhere [1, 2]. I have attempted here to give an overview of the field and
have gone into more detail only where I felt that additional information was
useful or that significant progress had been made since the publication of the
Institut Laue-Langevin (ILL) Neutron Data Booklet [1]. In these cases, I have
also indicated recent references for further study.

K. Andersen (*)
Institut Laue-Langevin, 6 rue Jules Horowitz, B.P. 156, 38042 Grenoble Cedex 9,
France
e-mail: [email protected]

I.S. Anderson et al. (eds.), Neutron Imaging and Applications, 31

Neutron Scattering Applications and Techniques, DOI 10.1007/978-0-387-78693-3_3,
Ó Springer ScienceþBusiness Media, LLC 2009
32 K. Andersen

3.2 Source Optics

Neutron sources are typically 10–20 cm in size in all directions. They consist of
an efficient moderator material (usually containing hydrogen or deuterium) at
a well-defined temperature to produce a quasi-Maxwellian wavelength spectrum.
Their performance is characterised by their brightness (or brilliance) d2 =d dl
(i.e., flux  per unit solid angle and per unit wavelength l). It is usually given in
units of neutrons/sec/cm2/sterad/Å. Brightness curves for the ILL vertical cold
source are shown in Fig. 3.1. For a pulsed source, the brightness can be given as
the peak brightness (i.e., the instantaneous brightness at its maximum in time) or
integrated over the full time spectrum of the moderator. These curves are also
shown in Fig. 3.1 for the liquid H2 moderator on the first target station at ISIS.
The brightness is a property of the source and does not change with distance.
The neutron flux at the sample is obtained by integrating over the source solid
angle seen by the sample and over the desired wavelength range. It then needs to
be scaled down by any imperfections in the transmission optics such as guide or
monochromator reflectivities or chopper transmissions.
The efficiency of any neutron optical device is limited by Liouville’s theorem,
also known as the ‘no free lunch’ theorem, which states that phase space density
in a conservative system is conserved. What this means is that any compression
in one dimension, such as focusing the source onto a spot smaller than the
source, is accompanied by an equivalent defocusing in another dimension, in
this case the divergence. This can be verified by looking at the standard lens
equation illustrated in Fig. 3.2: w1 =L1 ¼ w2 =L2 (i.e., the size of the focal spot
scales with the distance to the lens).

ILL & ISIS: cold-source brightness

1012
Brightness (n/cm2/s/A/ster)

1010

108

106 ISIS H2 moderator peak brightness

ISIS H2 moderator time-integrated
ILL vertical cold source
Fig. 3.1 Brightness of the
ILL vertical cold source and 104
1 10
the ISIS liquid H2
moderator Wavelength (A)
3 Neutron Optics 33

Fig. 3.2 Source–sample

optics with focusing lens w1 w2

source L1 L2 sample

It can easily be verified that the reduction in the size of the focal spot is
accompanied by an increase in the divergence hitting the focal spot, in accor-
dance with Liouville’s theorem. Liouville’s theorem should really be treated as a
best-case scenario. No optics system is fully conservative: some neutrons are
always lost from absorption, scattering, or other imperfections.

3.3 Reflection Optics: Neutron Guides

A neutron beam impinging on a flat surface can be treated as a plane wave
hitting a potential step. The potential V is proportional to the scattering-length
density Nb – the product of the atomic number density and the average
coherent scattering length. Since the gradient of the potential is perpendicular
to the surface, this potential step should be compared to the component of
neutron kinetic energy which is perpendicular to the surface K? . When K? 5V,
the neutron beam is fully reflected. The condition K? ¼ V corresponds to the
critical angle for total reflection, which is given by
pffiffiffiffiffiffiffiffiffiffiffiffi
sin  ¼ l Nb=p : (3:1)

For natural nickel, which has a scattering-length density of 9.410–6 Å–2, it

occurs at an incidence angle  of 0.108 per Å neutron wavelength l.
Total reflection from a flat surface enables neutron beams to be transported
long distances from the source down glass guides with appropriate surface
coatings. Nickel is a common choice for this surface coating, as it is relatively
inexpensive and has a large scattering-length density. In order to significantly
increase the critical angle beyond that of nickel, the guide surfaces can be coated
with supermirrors. A neutron supermirror is an artificial multilayer consisting
of thin layers of alternating materials with a large contrast in scattering-length
density. Hundreds or even thousands of layers are deposited, usually by sput-
tering, to create a quasi-continuous range of d-spacings. The d-spacing is
the combined thickness of two adjacent layers, which we can treat using Bragg’s
law:

sin  ¼ l=2d : (3:2)

A supermirror is usually made by first depositing the thin layers, gradually

increasing the layer thicknesses in order to cover the range of d-spacings up to
34 K. Andersen
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
d ¼ p=4Nb, where Nb is the scattering-length density of the high-Nb material.
It is then finished off with a thick (several hundred Å) layer of the high-Nb
material, to ensure total reflection at the small angles. The maximum critical
angle is determined by the smallest d-spacing which is technically feasible. In
practice that tends to be limited by substrate roughness, interdiffusion between
the layers, and stress build-up, which can eventually destroy the periodicity of the
mirror when too many layers are deposited. The critical angle of the supermirror
is usually given relative to the critical angle of nickel, using the ‘m’ number: m=1
is nickel, m=2 is a supermirror with twice the critical angle of nickel, etc.
Presently, neutron supermirrors are commercially available with critical
angles of m=4, though values as high as m=6 have been reported by research
groups.
The reflectivity of a surface describes the probability of specular (i.e., out=in)
reflection as a function of incidence angle in. Below the total-reflection angle
of the material, the reflectivity can almost be perfect (i.e., greater than 99%),
while in the Bragg-scattering regime of a supermirror it tends to be rather lower,
typically around 80% at m=3.
Supermirror-coated glass guides (Fig. 3.3) can be used to transport neutrons
far from the neutron source, which has several desirable effects. There is room
for more instruments if they are not all crowded around the neutron source.
Signal-to-background ratios can also be significantly improved by transporting
the useful neutron beam far from the source. The flux of gamma rays and fast
neutrons emitted from the source decreases approximately by 1/r2 (where r is
the distance from the source). Gamma rays and fast neutrons can also be
substantially attenuated by curving the neutron guide so as to avoid direct
line-of-sight back to the source. This condition is given by
pffiffiffiffiffiffiffiffiffiffiffi
L4 8wRc ; (3:3)

where L is the guide length, w is the guide width, and Rc is the radius of
curvature. Curving the guide also preferentially reduces the transmission of
short-wavelength neutrons. The characteristic wavelength l* (in Å) of a curved
guide is given by

575 pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
l ¼ 2w=Rc ; (3:4)
m

and corresponds to two-third of the incident neutrons being transmitted. At

shorter wavelengths, the transmission falls rapidly to zero.
Guides can also be used to improve the resolution of pulsed-source
instruments. This is typically determined by the relative time uncertainty

t/t, where
t is mainly determined by the time-width of the pulse in the
moderator and t is the total time-of-flight from source to detector. Hence
by moving the instrument further from the source, the relative time uncer-
tainty can be reduced, thus improving the instrumental resolution. The
3 Neutron Optics 35

Fig. 3.3 A supermirror guide on the ISIS second target station. The guide is viewed end-on,
showing the multiple reflections from the side and top and bottom walls. Photo courtesy of the
Science & Technology Facilities Council (UK)

length of the instrument is limited by the high cost of the supermirror

guide and associated shielding and the need to avoid frame overlap. Frame
overlap can occur when the slowest neutrons from one pulse are overtaken
by the fastest neutrons from the subsequent pulse. Long instruments
usually incorporate mechanical choppers which limit the wavelength band-
width to avoid frame-overlap problems.
Finally, guides can be used to focus neutron beams to a smaller size, increas-
ing the flux at the sample, while at the same time increasing the divergence, in
accordance with Liouville’s theorem. Most guides comprise straight segments,
which can be placed so as to approximate a curved surface. However, in some
cases, it is sometimes worth the extra complication and cost to build them with
continuous curvature in elliptical or parabolic shapes.
36 K. Andersen

3.4 Diffraction Optics

Diffraction optics are based on Bragg’s law:

nl ¼ 2d sin  : (3:5)

Monochromators are used to select the wavelength of neutrons incident on

the sample, while analysers are used to select the wavelength after scattering.
Both are much more prevalent at steady-state sources, as alternatives to the use
of time-of-flight for labelling the neutron wavelength. For this chapter I will
refer to both as monochromators for simplicity. Most monochromators for
neutron instruments use the first order (n=1) reflection since it is the most
intense reflection. The higher-order reflections can be problematic as they
contaminate the beam with undesired shorter-wavelength neutrons, which
need to be dealt with by other means.
Monochromators usually consist of single crystals, but artificial multilayers
can also be made periodic to produce efficient Bragg diffraction. However,
these are more usually stacked aperiodically to produce supermirrors and are
considered in the section on reflection optics. A perfect single crystal accepts
only a very narrow range of wavelengths within a given range of scattering
angles, given by the Darwin width, which is of the order of arc-seconds. Though
the peak reflectivity can be very high, the integrated reflectivity is low, and to
increase the counting rate on the instrument, two types of crystals are habitually
used: bent perfect crystals and mosaic crystals.
When a perfect crystal is bent, the number of reflected neutrons increases
dramatically. This can be qualitatively understood as the combination of two
effects: when bending, the internal (concave) face is compressed in the direction
of the face. To conserve density, the crystal must expand slightly in the other
(radial) direction. The d-spacing is thus made to vary as a function of depth.
The other effect is the variation of orientation of the crystal along the face due
to the continuous curvature. The two effects can add to give a large increase in
integrated intensity compared to the flat crystal. A mosaic crystal can be
visualised as an assembly of small perfect crystals that are slightly misaligned
with respect to each other. The misalignment is known as the mosaicity and is
usually modelled by a Gaussian distribution of misorientations with the mosaic
spread given by the FWHM (full width at half maximum) of the Gaussian.
A monochromator crystal is characterised by its peak reflectivity and its
mosaic spread. For a high peak reflectivity, you need a large structure factor
associated with the reflection and a large scattering cross section. It is reduced
by extinction, absorption, parasitic reflections, and inelastic scattering. Extinc-
tion is caused by multiple first-order Bragg reflections either coherently within a
perfect crystal or mosaic crystal block (primary extinction) or incoherently
between mosaic blocks (secondary extinction). Parasitic reflections are
3 Neutron Optics 37

predominantly due to other Bragg reflections, attenuating the diffracted beam

as it leaves the crystal.
The peak reflectivity represents the reflectivity at the centre of the peak
measured in a rocking-curve scan. It is strongly wavelength dependent. For a
nonabsorbing crystal it increases with wavelength until the Bragg cutoff is
reached at l = 2d. The measured mosaic spread is usually given by the
FWHM of the rocking curve. It is wavelength-dependent and differs from the
true mosaic distribution of crystallites, primarily due to the finite thickness of
the crystal. The behaviour of mosaic crystals is treated in [3].
The external surface of the crystal can be cut at any angle with respect
to the crystal planes. Most monochromators are cut either in reflection
(Bragg) geometry or in transmission (Laue) geometry, with the reflecting
planes respectively parallel and perpendicular to the external surface. The
neutron reflectivity is always higher in reflection geometry, but transmis-
sion geometry may be chosen for convenience, particularly when the
Bragg angle is small. Table 3.1 gives typical performance numbers for com-
monly used monochromator crystals. The resolution of a single crystal mono-
chromator can be studied by differentiating Bragg’s law to give

 2  2
@l @l

l2 ¼
d þ
 ¼ ð2 sin 
dÞ2 þð2d cos 
Þ2
@d @ ; (3:6)
)
l=l  cot 


where we have dropped the first term which is usually small. From here we can
see that the diffracted wavelength spread is proportional to cot, which
becomes very large for small take-off angles. This means that when the crystal
is used to obtain wavelengths which are significantly shorter than the d-spacing,
the wavelength resolution becomes quite poor, which can be useful if you wish
to maximise flux at the sample at the expense of resolution. At the other end of

Table 3.1 Typical performance of some commonly used monochromator crystals and their
reflections. Heusler is the Cu2MnAl alloy, which provides a polarised beam. The peak
reflectivity is given for a crystal in reflection geometry
Peak reflectivity Peak reflectivity
Material Reflection d-spacing at l ¼ 2 Å at l ¼ 4 Å Mosaic
Silicon 111 3.135 Å 80% 90% bent
perfect
Graphite 002 3.355 Å 70% 80% 0.58
Germanium 111 3.266 Å 60% 80% 0.28
Copper 111 2.087 Å 60% 80% 0.38
Heusler 111 3.437 Å 25% 35% 0.38
38 K. Andersen

the range, when the Bragg angle approaches 908, the cot term becomes small
and very high resolution measurements are possible.
Monochromators usually comprise several crystals assembled together to
form a continuous surface, which can focus on to the sample. Large focusing
monochromators can give very large flux gains at the sample by accepting large
amounts of solid angle from the source. Figure 3.4 shows two large focusing
monochromators built for the IN8 triple-axis spectrometer at the ILL [4].
Vertical focusing contributes little to instrumental resolution as the vertical
tilt does not affect the Bragg angle to a first approximation. Horizontal focusing
of a parallel beam invariably entails varying the Bragg angle across the surface
of the monochromator and thus tends to degrade resolution to some extent. The
resolution degradation can be strongly reduced by arranging the instrument
optics in order to obtain focusing in Q or energy, as illustrated in Fig. 3.5.

Fig. 3.4 Graphite (left) and copper crystals on the IN8 monochromator. Both faces consist of
99 mosaic crystals and have variable horizontal and vertical curvature
3 Neutron Optics 39

θA θA
A A

θB θB
B B

Fig. 3.5 Q-focusing and time-focusing. The left-hand frame shows Q-focusing as employed on
powder diffractometers on continuous sources. The right-hand frame shows time-focusing as
employed on time-of-flight spectrometers at continuous sources

In both cases, the monochromator focuses on the sample in space and the
wavelength selected by the monochromator varies with position on the mono-
chromator. As a result, the incident angle on the sample is strongly correlated
with the wavelength. In the case of Q-focusing, the angular difference between
the short- and long-wavelength neutrons coming from the monochromator is
cancelled out by the Bragg scattering at the sample, and the result is a
scattered beam which is as well collimated as the beam incident on the
monochromator. Q-focusing works only when the direction of scattering at
the sample is the opposite of that at the monochromator, and is most efficient
when the d-spacing measured at the sample is similar to that of the
monochromator.
In the case of time-focusing, a Fermi chopper (a rotating collimator) pulses
the beam before the sample. It spins in the direction which lets through
the slower neutrons (B) before the faster ones (A), so that if they transfer
the same amount of energy at the sample, they arrive at the detector at the
same time.
When Q- or time-focusing are properly optimised they can provide the same
energy resolution as what would be obtained with a flat monochromator, but
with a large gain in flux at the sample.
When using mosaic crystals, the reflected beam is somewhat smeared by
the misorientation of the crystallites of the monochromator crystals. The
left-hand panel of Fig. 3.4 illustrates Q-focusing starting from a parallel
beam, which can be considered as a point-source at an infinite distance. In
general, a focusing monochromator can be optimised for a point-source at
any distance by adjusting the curvature. The finite size of the real source
then needs to be optimised to match the mosaic spread of the crystals and
the sample size. Neutrons originating outside this optimum source size will
not hit the sample and therefore only contribute to background. Hence the
concept of the ‘virtual source’ in which a diaphragm is used to reduce the
beam size at the source, giving significant gains in signal-to-background
ratios.
40 K. Andersen

3.5 Refraction Optics

Neutron beams can be deviated by a change of the refractive index of the

medium. This is associated with a change in the potential energy either from a
magnetic field (the Zeeman energy) or from the refractive index n of the
material:
 
n ¼ 1  Nb l2 =2p : (3:7)

Since the refractive index is smaller than that of vacuum, a focusing lens must
have a concave shape, unlike for the case of light.
A magnetic field gradient can also deviate the neutron trajectory:

@2r  ~
¼  rB ; (3:8)
@t2 m

where  is the magnetic moment of the neutron and m is its mass. The plus/
minus sign refers to the two spin states of the neutron: antiparallel or parallel to
the magnetic field, respectively.
Unfortunately for neutron optics purposes, the amount of focusing which
can be achieved by refraction and magnetic-field gradients is usually quite
small. Currently, used lens devices are restricted to small beam sizes and tend
to suffer from long focal lengths (of the order of tens of metres) and chromatic
aberrations, though work is underway to address these problems both for
refractive [5] and for magnetic [6] lenses.

3.6 Polarisation Optics

Considerable technical advances have been made in polarisation optics over the
last 10 years or so. For this reason, I go into a little more detail here than in the
other sections. The reader is referred to [7] for a good overview of the technique.
Each individual neutron in a neutron beam has a spin of ½. When placed in a
magnetic field, this defines a quantisation axis with respect to which there are
only two possible orientations: parallel and antiparallel or ‘up’ and ‘down’. The
beam polarisation can be expressed as a scalar P which is usually defined to vary
between 0 and 1:

N"  N#
P¼ ; (3:9)
N" þ N#

where there are N" (N#) neutrons with spin up (down). The beam polarisation is
often expressed in terms of the flipping ratio F given by
3 Neutron Optics 41

N" 1 þ P F1
F¼ ¼ )P¼ : (3:10)
N# 1  P Fþ1

Bragg reflection from a single crystal can provide a well-polarised beam

when the nuclear and magnetic structure factors are equal in magnitude, so
that the combined structure factor is approximately zero for one of the two spin
states. Several crystals have reflections which satisfy this criterion. The most
commonly used are the 200 reflection of Co0.92Fe0.08 (d=1.76 Å) and the 111
reflection of the Heusler alloy Cu2MnAl (d=3.43 Å). Of these, the Heusler alloy
has the highest reflectivity and is available in good quality. Heusler crystals can
be assembled into large focusing arrays, similar to nonpolarising crystals,
provided they are designed so that a sufficiently large magnetic field (about
1,000 Oe [8]) can be applied to properly magnetise the crystals. A good-quality
magnetised Heusler crystal will have a reflectivity for an unpolarised beam at
l = 2 Å of about 25%. Seeing as the reflection is polarising, the maximum
reflectivity which can be expected is 50%, corresponding to reflection of all the
neutrons of one spin state and none of the ‘wrong’ spin state.
If you do not wish to polarise and monochromate at the same time, two types
of wide-bandwidth polarisers are currently in common use: polarising super-
mirrors and polarised 3He gas.
Polarising supermirrors consist of alternating layers of magnetic and
nonmagnetic materials. The scattering-length density of a ferromagnetic mate-
rial depends on the spin state of the neutron and can be written as N(bp) for
the two neutron spin states: parallel and antiparallel to the magnetisation of the
layers. N is the atomic number density and b and p are the nuclear and magnetic
scattering lengths, respectively. The combination of the two materials is chosen
so that (1) they have a large contrast in scattering-length density for the ‘up’ spin
state and (2) there is no contrast for the ‘down’ spin state. The result is a
supermirror which reflects only one spin state and transmits the other. At
very small angles, however, the ‘down’ spin state will also be reflected, as it
encounters the interface with the substrate material underneath the supermir-
ror. The substrate is usually made of glass or Si, which gives a critical angle for
total reflection of about m=0.5 for both spin states. The reflection at the
substrate interface can be suppressed by adding an absorbing layer between
the substrate and the supermirror. Presently, the largest critical angle which can
be achieved for a polarising supermirror is about m=4.
The supermirror must be kept magnetised in order to polarise effectively.
This is usually done by applying a magnetisation field of a few hundred Gauss
using permanent magnets. Under particular deposition conditions, a large
magnetisation anisotropy can be induced in the magnetic layers, resulting in a
remanent magnetisation after the external field is removed.
Supermirror polarisers can be used either in transmission or reflection geo-
metry. In reflection geometry, the undesired spin state is transmitted through
the supermirror and then absorbed in the substrate or elsewhere. In
42 K. Andersen

transmission geometry, the supermirror must be deposited on a nonabsorbing

substrate. The undesired spin state is reflected out of the beam direction and
then absorbed, while the other spin state is transmitted through both the super-
mirror and the substrate.
Supermirrors are a very efficient means of polarising cold-neutron beams
with low divergence. For the case of small-angle scattering and reflectometry
instruments, where the beam divergence is typically of the order of a few tenths
of a degree and wavelengths tend to be long, supermirror polarisers can provide
a transmission of more than 40% with a polarisation in excess of 95%. In these
cases, a single supermirror deposited on a flat substrate and inserted in the beam
at a grazing angle can do the job.
For a wide beam or when the divergence is large, the polarising supermirrors
need to be assembled into packages. The most common form is the supermirror
bender. These devices usually consist of thin glass plates with air gaps in between.
The supermirrors are deposited on the glass plates and the neutrons are trans-
mitted through the air channels. The assembly is curved, so as to avoid direct line
of sight [Eq. (3.3)]. For a channel width of 1 mm and a length of 300 mm, we can
see that a radius of curvature of 11 m is required. The transmission of the two
spin states can be evaluated by considering each channel as a small curved guide.
Keeping these dimensions with a critical angle of m=3 for the ‘up’ spin state and
m=0.5 for the ‘down’ spin state, we find from Eq. (3.4) that the characteristic
wavelengths for the two spin states are 2.6 Å and 15.5 Å. This is roughly the
wavelength range over which such a polarising bender will work efficiently:
transmitting one spin state but not the other. Supermirror packages can also
be assembled from thin supermirror-coated Si wafers without air gaps, in which
the neutrons are transmitted inside the Si. Supermirror benders typically have a
transmission of 20–30%, depending on wavelength and divergence.
More recently, nuclear-spin-polarised 3He gas has emerged as a viable alterna-
tive to supermirror-based devices. Helium-3 has a very large absorption cross
section and only absorbs neutrons in the opposite spin state to the nucleus. The
transmitted beam thus emerges polarised but has otherwise not been reflected,
diffracted, or otherwise scattered. For this reason, 3He-based polarisers and analy-
sers are often referred to as neutron spin filters (NSFs). There are two mechanisms
by which the 3He nuclear polarisation can be created in a gas in a magnetic field.
In spin-exchange optical pumping (SEOP) [9], an alkali metal vapour,
usually Rb, is mixed in with the 3He gas. One of the ground-state electrons of
the alkali metal is optically pumped with circularly polarised laser light of the
appropriate wavelength, creating an electronic polarisation. During atomic
collisions the polarisation of the alkali atom is transferred to a 3He nucleus
via the hyperfine interaction.
In metastability-exchange optical pumping [10], an RF discharge is created
in order to populate a particular excited electronic state. This state is then
pumped with circularly polarised laser light to preferentially populate one of
the Zeeman-split energy levels, which decays into a metastable state upon
emission of a photon. The electronic polarisation of the metastable state is
3 Neutron Optics 43

then transferred to a 3He nucleus by collisional mixing. In order to satisfy the

rate balance between the depopulation of the metastable states and the atomic
collisions, this mechanism is most efficient at low pressures, of the order of
1 mbar. The gas thus needs to be compressed by mechanical means in order to
reach the density required to have sufficient column density to polarise a
neutron beam.
Both optical pumping methods can result in gas polarisations PHe of up to
80% at pressures of the order of a few bars. The transmission of the two spin
states is given by

T"ð#Þ ðlÞ ¼ expðOðlÞ½1  PHe Þ ; (3:11)

where O is the dimensionless opacity of the gas, given by

O ¼ Nla ¼ 7:28  102 O0 ðbar  cm  —Þ at T ¼ 295K : (3:12)

N is the 3He number density = 2.4551019 cm–3  pressure [bar] at

T=295 K, l (cm) is the length of the 3He gas along the beam direction, a =
2,96610–24 cm2  l [Å] is the absorption cross section for the unwanted
neutron spin state, and O’ is the opacity, given by the product of the cell
pressure (bar), the cell length (cm), and the neutron wavelength (Å).
The transmission of an unpolarised beam is given by

1 
Tn ðlÞ ¼ T" þ T# ¼ expðOðlÞÞ coshðOðlÞPHe Þ ; (3:13)
2

and the resulting beam polarisation can be expressed as

T"  T#
Pn ðlÞ ¼ ¼ tanhðOðlÞPHe Þ : (3:14)
T" þ T#

The performance of a polariser or analyser is often given in terms of a figure

of merit which combines the polarisation and the transmission. The most
commonly used figure of merit is P2T and is shown in Fig. 3.6 together with
the transmission and the beam polarisation as a function of opacity, calculated
for a 3He polarisation of 75%.
The figure of merit is seen to exhibit a maximum at an opacity of about
27 bar.cm.Å. Depending on the wavelength range of interest the length or
pressure of the 3He polariser should be adjusted in order to always work in
the vicinity of this performance maximum. The polarisation performance is
seen to be very similar to that of Heusler crystals and polarising benders.
The nuclear polarisation created by the optical pumping is very far from
thermal equilibrium. If the gas is polarised using SEOP, it is possible to
continuously polarise the gas on the neutron beam. However, for most current
applications, the gas is polarised off-line and gradually depolarises while on the
44 K. Andersen

Fig. 3.6 Performance of a 1

3
He NSF cell as a function of
opacity for a 3He
polarisation of 75%. 0.8
Transmission T
Neutron beam polarisation Beam Polarization P
(P) and transmission (T) of Figure of Merit P2T
0.6

P, T, P2 T
an unpolarised beam are
shown, as is the standard
figure of merit P2T, which 0.4
display a maximum at the
optimum opacity
0.2

0
0 10 20 30 40 50 60
Opacity (bar.cm.A)

instrument with a characteristic time constant T1. The polarisation decay is

exponential with a 1/e time constant which is a combination of three terms:
1 1 1 1
¼ þ þ : (3:15)
T1 Td Tm Tw
The first term is caused by the dipole–dipole interactions between the 3He
nuclei. It scales inversely with pressure and is approximately 800 h/bar. The
second term is caused by magnetic field gradients, and the third term is due to
collisions with the cell walls. Current best efforts allow T1 relaxation times on
the instrument of the order of 100–200 h.
To minimise the relaxation due to interactions with the cell walls, containers
for polarised 3He gas are usually made of glass or Si, coated with Cs or Rb
metal, as displayed in Fig. 3.7. Different types of magnetic field cavities have

Fig. 3.7 Examples of glass 3He polariser cells with silicon entrance and exit windows. Photo-
graph courtesy of David Jullien (ILL)
3 Neutron Optics 45

been developed to maximise the relaxation time due to field gradients. Their
geometries are optimised for the particular instrument and are usually based on
solenoids, Helmholtz coils, or magnetised mu-metal. Screening from external
fields can be provided by steel, mu-metal, or superconducting Meissner screens.

References
1. ILL Neutron Data Booklet, Albert-José Dianoux and Gerry Lander (ed.), OCP Science
(July 2003).
2. V.F. Sears, Neutron Optics: An Introduction to the Theory of Neutron Optical Phenomena
and Their Applications, Oxford University Press, (1989).
3. V.F. Sears, Acta Cryst. A 53, 35 (1997).
4. A. Hiess et al., Physica B 385–386, 1077 (2006).
5. S.-M. Choi et al., J. Appl. Cryst. 33, 793 (2000).
6. H.M. Shimizu, Physica B 385–386, 989 (2006).
7. W.G. Williams, Polarized Neutrons, Oxford University Press, 1988.
8. A. Freund et al., Physica B 120, 86 (1983).
9. T.R. Gentile et al., J. Res. N.I.S.T. 110, 299 (2005).
10. A.K. Petoukhov et al., Physica B 385–386, 1146 (2006).
Chapter 4
Neutron Detectors for Imaging

L. Crow

Abstract This chapter examines neutron detection methods for imaging. The
first section reviews neutron capture converters, which form the basis for
thermal neutron detection, and the following sections examine various detector
systems with an emphasis on imaging applications and the choices available for
imaging at time-of-flight facilities.

Keywords Neutron detector  Scintillation neutron detector  Neutron

converter  Gas neutron detector  Solid state neutron detector  Imaging
neutron detector

4.1 Overview of Neutron Converters

As neutral particles, neutrons reveal themselves through interactions with
atomic nuclei. Neutrons with sufficient kinetic energy can produce signals
when nuclei, recoiling from elastic collisions, ionize the surrounding material.
Proton recoil in particular is a standard detection mechanism for fast neutrons
[1, 2] and enabled initial discovery of the neutron [3]. Neutrons with kinetic
energy less than a few electron volts cannot, however, produce ionization
directly; detection requires neutron capture reactions, which release large
amounts of energy in the form of charged particles and gamma rays. These
are the epithermal, thermal, and cold neutrons of greatest interest in neutron
imaging applications. In practice, a neutron converter, an isotope chosen for its
neutron capture reaction, forms the basis for a neutron detector. Desirable
converter characteristics include large neutron absorption cross section, mini-
mal gamma-ray cross section, and energetic charged particle products. Limited
or absent gamma-ray production is also desirable. Table 4.1 shows some con-
verters currently used in neutron scattering or imaging instrumentation.

L. Crow (*)
Neutron Facilities Development Division, Oak Ridge National Laboratory,
Oak Ridge, TN 37831, USA
e-mail: [email protected]

I.S. Anderson et al. (eds.), Neutron Imaging and Applications, 47

Neutron Scattering Applications and Techniques, DOI 10.1007/978-0-387-78693-3_4,
Ó Springer ScienceþBusiness Media, LLC 2009
48 L. Crow

Table 4.1 Thermal neutron cross sections and charged particle production for some materials
used as neutron converters [4–6]
Charged
Neutron absorption cross particles and Gamma-ray
Isotope Reaction section (barns) (at 2,200 m/s ) energies (keV) production?
3 3
He He(n,p)3H 5333 p: 573, 3H: 191 No
6 6
Li Li(n, )3H 940 3
H: 2727, : 2055 No
10 10
B B(n, )7Li 3835 : 1472, 7Li: 480 Yes
nat nat
Gd Gd(n, ) 49700 Conversion Yes
electron
29–191
157 157
Gd Gd(n, ) 158Gd 259000 Conversion Yes
electron
29–182
235 235
U U(n,f) 681 Various fission Yes
products

Both the nuclear properties and the material characteristics affect the choice
of converter for a particular application. Helium-3 and 6Li have relatively large
cross sections, but the most important feature is that these capture reactions
release all of their energy in charged particles [6]. This makes these converters
the best choices for low-noise detectors which need to operate at high rates in
accessible areas. Helium-3 gas proportional counters have modest position
resolution, but very low gamma-ray sensitivity, and are the most widely used
detectors for neutron scattering. Lithium-6 is a very efficient scintillation con-
verter and appears in the most widely used thermal neutron scintillators, 6Li
glass and 6LiF:ZnS [7]. Boron-10 has a larger cross section than 6Li, and unlike
3
He can be in solid materials with very high atom density. The 10B capture
reaction has the disadvantage of producing a gamma-ray and heavier, shorter-
ranged charged particle products ( and 7Li). The compound 10BF3 has long
been used in proportional counters, and 10B is used in surface conversion gas
detectors. Boron-10 is also used in semiconductor detectors, microchannel plate
detectors, and a few scintillators.
Gadolinium has a number of disadvantages as a converter but is of interest in
imaging because of the enormous resonance at thermal neutron energies of
157
Gd (155Gd also has a very large cross section). The 157Gd capture results in a
spectrum of high energy gamma rays, which are difficult to count or shield,
along with conversion electrons. Gadolinium-based detectors rely on the inter-
nal conversion electrons to transfer the energy to a phosphor, gas, or lattice.
The cross section does not have the inverse velocity dependence of most con-
verters. This feature makes comparison of energy-dependent neutron response
difficult, but with white-beam imaging, the preferential response to thermal
neutrons helps suppress fast and epithermal neutron background.
Neutron capture reactions where the primary products are gamma rays may be
used for applications such as prompt or decay gamma activation analysis. The
4 Neutron Detectors for Imaging 49

characteristic gamma rays identify the isotopes which have absorbed the neutrons.
This method may be used for chemical analysis or for neutron integrated flux
measurements. Isotopes with higher energy absorption resonances have applica-
tions in energy-specific neutron detection [8]. A significant effort for chemical
spectroscopy and imaging is discussed in Chapter 13 of this volume. Isotopes
which capture neutrons and produce short-lived radioactive products maybe used
in indirect imaging or autoradiography, using the decay beta particles.

4.2 Data Acquisition for Continuous and Time-Dependent

Applications

Some neutron detection systems operate in current or integrating mode, where

the individual neutron events are not distinguished. This can work well when
the time dependence of the signal is not an important parameter within the
measurement period. With intense sources and fast gating, integrating detectors
can be used at relatively fast frame rates, but cannot provide the time resolution
(within a few microseconds) required for time-of-flight methods.
Counting rather than integration allows use of statistical methods at low
data rates, especially when signal and background are low. The vast majority
of neutron scattering measurements use counting detection schemes. In addi-
tion, counting detectors facilitate time stamping or binning for time-of-flight
measurements using event mode readout. Neutron time-of-flight with a pulsed
source (or a chopped continuous beam) offers a method of measuring neutron
interactions as a function of neutron kinetic energy. In addition to applica-
tions in diffraction and inelastic neutron scattering, time-of-flight offers
opportunities in imaging for energy-dependent or element-dependent contrast
variation.

4.3 Photographic Film Detectors

Photographic film provides an analogue readout, always in signal integrating

mode. Film has the advantages of compact detector assembly and immediate
image generation. Most film methods can also provide position resolution on
the order of tens of microns, which was difficult until recently with electronic
readout. The frame rate is governed by the relatively long exposure times, due to
the low detection efficiency, and the requirement to physically remove and
replace the film. Film continues in wide use at continuous source installations
[9] but is not adaptable to time-of-flight techniques.
A direct capture camera uses prompt particle emission from neutron capture
to illuminate the film (Fig. 4.1). Gadolinium foil/X-ray film detectors have
served as the standard tool during much of the development of neutron radio-
graphy [9], and continue in use today [10]. The resolution of such detectors,
50 L. Crow

(a)

(b)

Fig. 4.1 The figure illustrates the dominant processes in direct and indirect screen capture for
two common screen choices. (a) A direct Gadolinium screen and X-ray film neutron imaging
process is shown schematically. The neutrons are captured by the Gd nuclei [157Gd (n, )
158
Gd]. The high energy prompt gamma rays escape, but some of the internal conversion
electrons deposit energy in the film and contribute to the photographic image. (b) In an
indirect process, a dysprosium screen captures neutrons during irradiation [164Dy (n, )
165
Dy]. The screen is removed from the neutron beam and placed in contact with film. An
image is produced by - particles from dysprosium decay [165Dy ( -, 165Ho), t1/2 ¼ 2.33 h]

usually a few tens of microns, is determined by the grain sizes in the converter
screen and the film, and the efficiency is dependent on the gadolinium conver-
sion electron capture in the film. Newer methods are more efficient, but the
position resolution of these screens remains competitive [11]. This detector
system requires a mechanism to open and close the incident beam to define
the exposure time and allow removal and replacement of the film and converter
screen and a facility to develop the film. This works well at continuous sources
with relatively long exposure times but is difficult to adapt for fast sample
changes.
4 Neutron Detectors for Imaging 51

The direct screen and film method requires suppression of gamma-ray,

X-ray, and charged particle radiation from both source and sample. If the
sample produces significant prompt emissions or is already radioactive, a screen
consisting of a short-lived activation material, typically a -emitter such as In,
Dy, or Au [9, 12], can serve as an indirect imaging device. The screen activates
when the sample and screen are exposed to the neutron beam. The screen is then
moved from the beam and placed in contact with the film. The decay from
neutron activation produces the image, and incident gamma effects are absent.
The indirect method needs long exposure times and image recording time, but
the gamma insensitivity allows high-resolution imaging of difficult samples
such as spent nuclear fuel elements (Chapter 5).
A specialized technique known as neutron autoradiography uses activation
products within the sample to produce images. This technique works in princi-
ple with any object with variations in composition. Direct autoradiography
couples the sample, usually containing boron, to an alpha sensitive film and
records the prompt alpha emissions [13, 14]. It has been used to monitor boron
compound uptake for boron neutron capture therapy [15, 16]. Indirect auto-
radiography works with activation of components within a sample. Autoradio-
graphy can provide details of paint pigment distribution in historical paintings
[17]. The activation gamma-ray energies reveal pigment composition, which
indicates the age and authenticity of the painting. The resultant images illus-
trate the artist’s composition technique and may show earlier drafts of the
painting or even earlier paintings on the same canvas.
The track-etch method uses charged particle products of neutron capture to
record tracks directly in an emulsion film. The track density, viewed through a
microscope, forms the image. For 6Li or 10B, the typical particle ranges are
between 5 and 30 mm, which effectively sets the resolution limit. The most
established track-etch film is boron-loaded nitrocellulose [18]. The track-etch
method requires long exposure times.
Neutron scintillators may be used with film but are more commonly used
with faster electronic light detection systems.

4.4 Scintillator and Storage Phosphor Detectors

Neutron Scintillators: A scintillator is a material that absorbs ionizing radiation
and emits low-energy photons. A neutron scintillator may operate by recoil [19],
but a thermal neutron scintillator requires a converter that emits charged particles,
either nuclear products from the capture reaction or internal conversion electrons.
For a number of years, the most widely used thermal neutron scintillators
have been 6Li glass and 6LiF/ZnS [20]. Cerium-activated 6Li glass is an intrinsic
scintillator (i.e., the converter atom and phosphor are part of the same mate-
rial). Lithium-6 glass, used in some diffraction detectors [21–23], is not often
used in imaging because a thickness of a millimeter or more is required for
good efficiency, gamma-ray sensitivity is higher than desired, and the 8,000
52 L. Crow

photons/neutron light output is relatively small. The binary scintillator 6LiF/

ZnS is created by mixing the converter particles with phosphor particles, where
both materials have sizes of a few microns (consistent with charged particle
ranges). The original formulation of 6LiF/ZnS [20] provides about 160,000
photons/neutron light output. Continuing efforts to increase efficiency for
scattering detectors [24] by increasing thickness and lithium density have not
proven advantageous to imaging, but recent work for imaging detectors [11]
improves position resolution by reducing thickness to about 50 mm at the cost of
lower neutron capture efficiency.
Several additional 6Li and 10B scintillators have been proposed and tested
[25–29]. Gadolinium-based scintillators, for example Gd2O2S (GOS), Gd2SiO5
(GSO), and a microcolumnar gadolinium-cesium iodide hybrid [30–33], have
limited light yield because only the conversion electron energy is available, but
the very large cross section means that high-capture efficiency can be achieved
with only a few microns of material. For this reason, gadolinium scintillators
have potential applications in high-resolution (tens of microns) neutron imaging.
Photographic Scintillation Detectors: The optical part of a scintillation detec-
tor collects the scintillation light, records it, and uses the light collection pattern to
locate the neutron. Photographic film offers a very small package in the beam
location, but has far lower light collection efficiency and requires a film change
for readout. Faster readout can be achieved using a television camera viewing the
scintillator; sensitivity can be increased using an image intensifier [34].
The advent of high-performance charge-coupled device (CCD) cameras has
transformed this technology into a competitive high resolution imaging system
[11] that can collect and read out images far faster than film. A typical system
consists of a scintillator screen, a 458 front surface mirror, a focusing lens, and a
CCD camera [35–39]. Lehmann et al. have performed an optimization of these
components for the ICON facility at SINQ [11] (Fig. 4.2). Thick scintillator
screens optimized for efficiency degrade the spatial resolution; standard 250 mm
or more thick 6LiF/ZnS can be replaced with 50 mm or less 6LiZnS layers, or
with 10 mm GOS. A front surface mirror is used to eliminate multiple reflection
effects. This high resolution system uses a cooled CCD camera with a custom
lens system and achieves a resolution of 50 mm.
Most photographic readouts, whether film or CCD, effectively integrate
over a large number of neutron events. Such readouts give good imaging results
in steady-state source instruments but are less well suited to time-of-flight
measurements. Fast gated cameras are a possible method for using the time
structure from pulsed sources and for imaging dynamic processes [40]. The use
of camera systems with pulsed sources will be simplified as more fast cameras
become available.
A mirror keeps the CCD camera outside the neutron path to protect it from
radiation damage, but the increased distance decreases the light collection
efficiency. A fiber optic taper [41, 33] greatly increases the light collection
efficiency. The glass in the taper also provides partial shielding for the camera
sensor; however, it is not yet known whether a CCD can operate for long
4 Neutron Detectors for Imaging 53

Fig. 4.2 The figure shows a scale drawing and a perspective view of the microtomography
scintillator and CCD detector system at the Paul Scherrer Institut, Villigen, Switzerland. The
scintillator screen, a 458 mirror, the lens system, and the cooled CCD camera form the neutron
detector system. (This image was provided by Eberhard Lehmann of PSI by private
communication)

periods in this geometry. Amorphous silicon flat panel photodiode arrays

provide a fast readout system which could address dynamic measurement and
time-of-flight requirements while providing a large solid angle for light collec-
tion. Preliminary tests at operating neutron beam lines [42–44] have successfully
demonstrated the concept with good readout performance, and the test results
for tomography are promising. Further development requires improved
dynamic range.
Counting Scintillation Detectors: For neutron counting with scintillators,
photomultipliers (PMTs) provide the necessary fast photon counting capability.
Avalanche photodiodes are a possible alternative with higher quantum efficiency,
but at present they have higher dark counts at higher cost. Although a scintillator
with a PMT can be operated as a single pixel detector, most scattering applica-
tions use multiple PMT elements to build position-sensitive detectors. Anger
cameras with neutron scintillators [21–23] can now achieve area resolution of
1.3 mm or smaller while counting (and time stamping) neutrons at tens of kilo-
hertz, and possibly faster (Fig. 4.3). Fiber light collection using end-on fibers [45,
46] or wavelength-shifting fibers [24] gives light collection over large areas with a
relatively small investment in light detectors, appropriate for diffraction instru-
ments which require pixel sizes of a few millimeters. Fiber collection can, like the
54 L. Crow

(a) (c)

(b) (d)

Fig. 4.3 (a) The figure shows an optical detector assembly of an Anger camera module for
the SNAP high pressure diffractometer at the Spallation Neutron Source (SNS) at Oak
Ridge National Laboratory. Nine 88 channel position-sensitive photomultiplier tubes are
assembled to form a single Anger module. (b) A SNAP module is shown with the bezel,
which contains a 15 cm square 6Li glass scintillator. This detector has a two-dimensional
position resolution of 1.3 mm full width at half maximum (FWHM). (c) A crossed
wavelength-shifting fiber assembly for a detector module to be mounted on the Vulcan
engineering diffractometer at SNS is shown. The wavelength-shifting fibers collect the blue
scintillation light from a 6LiF/ZnS(Ag) scintillator screen and guide the shifted green light
to a coded photomultiplier array. (d) A completed Vulcan module operates for initial
calibration. This 80 cm  40 cm detector module has 5 mm wide by 37.5 mm tall pixels,
with resolution matched to the diffractometer

Anger camera, achieve resolution approaching 1 mm [47], but it is not likely that
these technologies can achieve the submillimeter resolution needed to compete in
radiography or tomography.
Storage phosphors and image plate readers: Storage phosphors are function-
ally similar to scintillators, except that instead of a prompt light emission, energy
is stored in the lattice. The phosphor is read by photostimulation (typically by
scanning with a laser), inducing light emissions. Neutron image plates using a
gadolinium converter with a storage phosphor (Fig. 4.4), developed in the 1990s
4 Neutron Detectors for Imaging 55

(a)

(b)

Fig. 4.4. (a) A 157Gd-based image plate accumulates energy from internal conversion
electrons produced in neutron capture in its storage phosphor. The image plate is
transferred to a reading system. A laser scans the image plate and photostimulates the
release of blue photons from the storage phosphor in the image plate. (b) A 25 cm square
neutron image plate is shown

[48, 49], have become commercially available [50] in the last decade and are now
in regular use for diffraction [51] and radiography [52] at steady-state sources.
Image plates are much more efficient than any of the photographic film
56 L. Crow

techniques. Spatial resolution is a function of image plate thickness but is on the

order of 25–100 mm [11]. In addition, the automated readout is much faster than
conventional film processing (and in electronic form), and the image plates
themselves are reusable. Still, image plates are integrating detectors and cannot
be gated for use at time-of-flight sources. While some facilities use image plate
readers developed for X-ray image plates [52], development of readers optimized
for neutron applications is continuing [53].

4.5 Gas Detectors

In a gas detector, ions produced within the gas accelerate due to a charge bias
applied within the gas volume. If the acceleration is large enough, the signal is
amplified as the ions collide with gas atoms and produce additional ionization.
In this way, an ionization event, such as a neutron capture, will produce an
electron ‘‘charge cloud,’’ which drifts to the biased anode. Depending on the
bias and the gas, the charge may be equivalent to the original energy deposition
(ionization mode), amplified proportional to the original energy (proportional
mode), or a nonproportional avalanche (Geiger-Müller mode). Neutron detec-
tors with gas converters such as 10BF3 [54] and 3He [55, 56] usually operate in
the proportional range to take advantage of the excellent discrimination
allowed by the energetic charged particles.
Helium-3 proportional counters are the most widely used detector systems
for neutron scattering, where resolution can be matched to the requirements
of the instruments. The development of position-sensitive proportional coun-
ters has led to imaging detectors for neutron-scattering applications such as
small-angle neutron scattering (SANS), neutron reflectometry, and diffrac-
tion. Readout systems for area detectors include delay line [57], charge divi-
sion [58], and the now more widely used multiwire centroid coincidence
method [59–61]. Linear position-sensitive detectors using charge division
[62] may be arranged to form area detectors for SANS [63, 64] and a wide
variety of neutron-scattering applications [65, 66] (Fig. 4.5). The counting
readout systems are readily adaptable to time-of-flight work. Gas converter
detectors cannot compete easily for submillimeter position resolution because
of the thick gas volume and the long capture product ranges (typically on the
order of a millimeter or longer). The best wire chamber detector resolution to
date is about 400 mm [67]. Gas detector development continues with efforts to
overcome the present count rate limitations of wire chambers, typically
between 40,000 and 100,000 counts per second per chamber due to drift
time. A new high-performance multiwire gas chamber detector known as
MILAND1 [68] is expected to achieve 1 mm resolution at a count rate of up

1
MILAND (for Millimeter Resolution Large Area Neutron Detector) is a joint research
activity of the European Union supported under the Fifth Framework Program (FP5).
4 Neutron Detectors for Imaging 57

(a) (b)

(c)

Fig. 4.5 (a) An array of 8 mm diameter  1 m long linear position-sensitive 3He proportional
tubes forms an area detector for small angle neutron scattering, with resolution of about
8 mm. The modules shown here will be installed on EQSANS at SNS. (b) A 20 cm  17 cm
multiwire proportional counter manufactured by Brookhaven National Laboratory operates
in the SNS liquid reflectometer. (c) This multiwire detector, with 6.5 bar of 3He, has a
resolution of 1.3 mm FWHM, as shown in this calibration mask image

to 1,000,000 counts per second over the entire detector area using readout with
separate amplifiers on each wire.
Strategies for higher performance gas detectors use micropattern elec-
trode structures to effectively reduce drift distances and times and in the
process divide the readout of a detector chamber into smaller segments.
58 L. Crow

These detectors may use gas converters or solid converters within the
counting gas volume. In microstrip gas counters (MSGC) [69–72], the
anode and cathode strips are tens of microns apart on a glass substrate,
reducing the drift times to a few hundred nanoseconds. Gas scintillation
proportional counters are variants of the MSGC that use the light from the
charge amplification to localize the neutron events [73–75]. Gas electron
multipliers [76–78] use biased stacks of perforated kapton foils to achieve
high gas amplification in a very small distance. Another in this same
subclass of gas detectors, the Micromegas (for MicroMEsh Gaseous
Structure), has a converter layer and a very small amplification gap of
about 100 mm [79–80] between two parallel glass or epoxy plates.

4.6 Solid-State Detectors

Semiconductor detectors: If the charged particles from a neutron capture enter a
biased semiconductor device, the resultant ionized region will produce a pulse.
If the pulse is large enough, the device becomes a neutron detector that can be
coupled to fast readout counting electronics. Potential advantages of semicon-
ductor detectors include high data rate, small size, and very low power require-
ments. With very short charged particle paths and close integration to digital
readout systems, these detectors should be capable of very high data rates and
high resolution. A neutron converter can be intrinsic to the semiconductor (part
of the semiconductor structure) or extrinsic (adjacent to the semiconductor
region). Intrinsic semiconductors proposed or tested include boron phosphide
[81] and boron nitride [82], but no intrinsic semiconductor has been successfully
developed as a neutron detector. McGregor [83] has reviewed semiconductor
neutron detectors in detail and demonstrated the efficiency limits for flat layers
of 6Li and 10B compounds on semiconductor surfaces. Successful detection
requires that one of the neutron capture product charged particle tracks end
within the semiconductor. The track lengths of 30 mm or less place severe
restrictions on the usable converter thickness and restrict the detection effi-
ciency to less than about 12% even for cold neutrons. Flat layers on charged
particle detector application-specific integrated circuits (ASICs) may increase
readout performance [84] but will not address the efficiency issue. Modifica-
tions to the converter geometry such as placing the converter in wells or
channels in perforated semiconductor structures (Fig. 4.6) have produced
higher efficiency [85].
A recent MgB2 superconducting detector prototype makes use of mean-
der traces of MgB2 a few microns wide on an insulating substrate. With the
MgB2 at superconducting temperature, single neutron capture events can
produce normal conducting regions for very short times [86] and produce
signal pulses. The resulting detector can operate at high counting rates.
4 Neutron Detectors for Imaging 59

(a) (b)

Fig. 4.6 (a) A silicon wafer detector element has been perforated with grooves at 100 mm
spacing. Each groove has been filled with 6LiF. The pixel size is 100 mm by 4 cm (the detector
height). (b) The detector is mounted at the center of the circuit board; the application-
specific integrated circuits to operate the detector are shown above and below the detector
window

Resolution of tens of microns is achievable in principle with such a detector,

but multiple layers would be needed to reach good efficiency for thermal
neutrons.
Glass microchannel plates (MCPs), long used in optical image intensifiers,
amplify electronic signals by accelerating electrons through evacuated micro-
scopic channels, typically with diameters of 5–10 mm and spacing 20 mm or less.
Microchannel plates become neutron detectors if the glass material contains a
converter, typically 10B, in sufficient concentrations [87]. The charged particle
capture product ranges are a few microns in glass so that the resolution is
potentially comparable to the channel spacing. Nova Scientific has carried
out a development program to optimize the microscopic geometry and compo-
sition of MCP neutron detectors and is continuing development of readout
electronics. Amorphous silicon readout [88] provides good position resolution
but is an integrating system. The newer delay line readout [89] gives the detector
system counting capability with 25 ns time resolution and possible rates up to
1 MHz, with spatial resolution better than 40 mm FWHM, the best yet achieved
with a counting neutron detector (Fig. 4.7). Another development, also based
on Nova MCPs, uses the Medipix multipurpose ASIC readout which can
operate as a counting system or as a higher resolution centroiding system [90]
that can approach 15 mm position resolution with gating that should be fast
enough for time-of-flight applications.
60 L. Crow

Fig. 4.7 A 40 mm diameter microchannel plate detector built by NOVA Scientific, similar to
the detector described in [91], is shown. The microchannel plate assembly is visible through the
sapphire neutron window. The delay line readout enables counting operation at high rates
with 40 mm resolution

4.7 Conclusion
The current state of neutron imaging detectors presents a range of options for
continuous sources and some promising paths for achieving the timing capabil-
ities needed for pulsed sources. The neutron storage phosphor image plates
yield excellent resolution at modest operating costs. The CCD camera systems
are fast and flexible and are currently the primary detection technique for most
major imaging user facilities. The potential for improvement in scintillators or
cameras puts pulsed source operation within reach, but gating and camera
readout at high rates are big challenges. The microchannel plate detectors
operate at very high count rates, are operable as counting detectors now, and
may have the shortest path to operation as timing detectors.

References
1. J.A. Harvey and N.W. Hill, Nucl. Instrum. Meth. 162, 507 (1979).
2. G.F. Knoll, Radiation Detection and Measurement, 3rd ed. Wiley, New York (2000).
4 Neutron Detectors for Imaging 61

3. J. Chadwick, Proc. Roy. Soc. A136, 703 (1932).

4. V.F. Sears, Neutron News 3, 26 (1992).
5. C.M. Lederer and V.S. Shirley (eds.); E. Browne, R.E. Doebler, and J.M. Dairiki
(principal authors); A. Shihab-Eldin, L.J. Jardine, J.K. Tuli, and A.B. Buyrn, Table of
Isotopes, 7th edition. Wiley, New York (1978).
6. B. Gebauer, Comprehensive summary session 4, Proceedings of the International work-
shop on Position-Sensitive Neutron Detectors—Status and Perspectives, (PSND2001),
http://www.hmi.de/bensc/psnd2001.
7. A.R. Spowart, Brit. J. Nondestr. Test. 11, 2 (1969).
8. G. Gorini, E. Perelli-Cippo, M. Tardocchi, C. Andreani, A. D’Angelo, A. Pietropaolo, R.
Senesi, S. Imberti, A. Bracco, E. Previtali, G. Pessina, N.J. Rhodes, E.M. Schooneveld.
Nucl. Instr. Meth. A 529, 293 (2004).
9. H. Berger, Neutron Radiography – Methods, Capabilities and Applications. Elsevier,
Amsterdam (1965).
10. R. Ilic and M. Najer, Nucl. Instrum. and Meth. 180, 635 (1981); R. Pugliesi and M. L.G.
Andrade, Appl. Radiat. 1sot. 48, 339 (1997).
11. E.H. Lehmann, G. Frei, G. Kühne, P. Boillat, Nucl. Instrum. Meth. A 576, 389 (2007).
12. M. Mullner and H. Jex, Nucl. Instrum. Meth. 103, 229 (1972).
13. Z.B. Alfassi, and T.U. Probst, Nucl. Instrum. Meth. A 428, 502, (1999).
14. C.P. Bean, R.L. Fleischer, P.S. Swartz and H.R. Hart. J. Appl. Phys. 37, 2218 (1966).
15. M. Takagaki, and Mishima, Yu., Nucl. Tracks Radiat. Meas. 17, 531–535, (1990).
16. J.E. Woollard, T.E. Blue, J.F. Curran, T.F. Mengers, and R.F. Barth, Nucl. Instr. Meth.
A 299, 600 (1990).
17. C. Laurenze-Landsberg, C. Schmidt, B. Schroder-Smeibidl, and L.A. Mertens, Notiziario
Neutroni e Luce di Sincrotone 11, 24 (2006).
18. A.A. Harms and M.S. Moniz, Nucl. Instrum. Meth. 122, 567 (1974).
19. R. Batchelor, W.B. Gilboy, J.B. Parker and J.H. Towle, Nucl. Instrum. Meth. 13, 70
(1961).
20. A.R. Spowart, Nucl. Instrum. Meth. 75, 35 (1969).
21. M. Heiderich, R. Reinartz, R. Kurz and J. Schelten, Nucl. Instrum. Meth. A 305, 423
(1991).
22. M.G. Strauss, R. Brenner, H.P. Chou, A.J. Schultz and C.T. Roche, Position-Sensitive
Detection of Thermal Neutrons. Academic Press, London, p. 175 (1983).
23. R. G. Cooper, J. D. Richards, C. Donahue, T. Visscher, IEEE Nucl. Sci. Symp. (2007).
24. M.L. Crow, J.P. Hodges, and R.G. Cooper, Nucl. Instrum. Meth. A 529, 287 (2004).
25. J.P. Chaminade, O. Viraphong, F. Guillen, C. Fouassier and B. Czirr, IEEE Trans. Nucl.
Sci. 48, 1158 (2001).
26. J.P. Hodges, M.L. Crow, and R.G. Cooper, patent pending (2007).
27. T. Kojima, M. Katagiri, N. Tsutsui, K. Imai, M. Matsubayashi, and K. Sakasai, Nucl.
Instrum. Meth. A 529, 325 (2004).
28. M.J. Knitel, New Inorganic Scintillators and Storage Phosphors for Detection of Thermal
Neutrons, Ph.D. Thesis, TU Delft, (1998).
29. C.W.E. van Eijk, A. Bessière and P. Dorenbos, Nucl. Instrum. Meth. A 529, 260 (2004).
30. K. Takagi, T. Fukazawa, Appl. Phys. Lett. 42,43 (1983).
31. C.L. Melcher, J.S. Schweitzer, R.A. Manente, and C.A. Peterson, IEEE Trans. Nucl. Sci.
38, 506 (1991).
32. J.S. Brenizer, H. Berger, C.T. Stebbings, and G.T. Gillies, Rev. Sci. Instrum. 68, 3371
(1997).
33. I. Shestakova, V. Gaysinskiy, J. Antal, L. Bobek, and V.V. Nagarkar, Nucl. Instrum.
Meth. B 263, 234 (2007).
34. Harold Berger, Wilfrid F. Niklas, and Adolph Schmidt, J. Appl. Phys. 36, 2093 (1965).
35. T. Hibiki, K. Mishima, K. Yoneda, S. Fujine, A. Tsuruno, and M. Matsubayashi, Nucl.
Instrum. Meth. A351, 423 (1994).
62 L. Crow

36. Helena Pleinert, Eberhard Lehmann, and Sonja Körner, Nucl. Instrum. Meth. A399, 382
(1997).
37. D.S. Hussey, D.L. Jacobson, M. Arif, P.R. Huffman, R.E. Williams, and J. C. Cook,
Nucl. Instrum. Meth. A542, 9 (2005).
38. B. Schillinger, E. Calzada, F. Grunauer, and E. Steichele, Appl. Radiat. Isot. 61, 253 (2004).
39. M. Dierick, J. Vlassenbroeck, B. Masschaele, V. Cnudde, L. Van Hoorebeke, and A.
Hillenbach, Nucl. Instrum. Meth. A 542, 296 (2005).
40. B. Schillinger, H. Abele, J. Brunner, G. Frei, R. Gähler, A. Gildemeister, and A.
Hillenbach, E. Lehmann, P. Vontobel, Nucl. Instrum. Meth. A 542, 142 (2005).
41. Masahito Matsubayashi and Kazuhiko Soyama, Nucl. Instrum. Meth. A 529, 384 (2004).
42. E. Lehmann and P. Vontobel, Appl. Radiat. Isot. 61, 567 (2004).
43. Mirko Estermann and Joe Dubois, IEEE Trans. Nucl. Sci., 52, 356 (2005).
44. Mirko Estermann, Gabriel Frei, Eberhard Lehmann, and Peter Vontobel, Nucl. Instrum.
Meth. A 542, 253 (2005).
45. N.J. Rhodes, A.G. Wardle, A.J. Boram, and M.W. Johnson, Nucl. Instrum. Meth. A 392,
315 (1997).
46. N.J. Rhodes, E.M. Schooneveld, and R.S. Eccleston, Nucl. Instrum. Meth. A 529, 243 (2004).
47. M. Katagiri, T. Nakamura, M. Ebine, A. Birumachi, S. Sato, E.M. Schooneveld, and N.J.
Rhodes, Nucl. Instrum. Meth. A 573, 149, (2007).
48. Nobuo Niimura,Yuuko Karasawa, Inchiro Tanaka, Junji Miyahara, Kenji Takahashi,
Hiroki Saito, Satoshi Koizumi, and Masanori Hidaka, Nucl. Instrum. Meth. A 349, 521
(1994).
49. Jens Hofmann and Christian Rausch, Nucl. Instrum. Meth. A 355, 494 (1995).
50. Fuji Film, I & I-Imaging Inform. (1995).
51. D.A.A. Myles, C. Bon, P. Langan, F. Cipriani, J.C. Castagna, M.S. Lehmann, and C.
Wilkinson, Physica B 241-243, 1122 (1998).
52. Shigenori Fujine, Kenji Yoneda, Masahiro Kamata, and Masahiro Etoh, Nucl. Instrum.
Meth. A 424, 200 (1999).
53. B. Schillinger, J. Baumann, H. Gebele, R. Schaetzing, H. Schaller, and M. Schuster, App.l
Radiat. Isot. 61, 451 (2004).
54. S.A. Korff and W.E. Danforth, Phys. Rev. 55, 980 (1939).
55. R. Batchelor, R. Aves and T.H.R. Skyrme, Rev. Sci. Instrum. 26, 1037 (1955).
56. W.R. Mills, R.L. Caldwell, and I.L. Morgan, Rev. Sci. Instrum. 33, 866 (1962).
57. C.J. Borkowski and M.K. Kopp, Rev. Sci. Instrum. 46, 951 (1975).
58. V. Radeka, IEEE Trans. Nucl. Sci. 21, 51 (1974).
59. G. Charpak and F. Sauli, Nucl. Instrum. Meth. 113, 381 (1973).
60. Veljko Radeka and Robert A. Boie, Nucl. Instrum. Meth. 178, 543 (1980).
61. R. Kampmann, M. Marmotti, M. Haese-Seiller, and V. Kudryashov, Nucl. Instrum.
Meth. A 529, 342 (2004).
62. R. Berliner, J.S. King and D.F.R. Mildner, Nucl. Instrum. Meth. 152, 431 (1978).
63. K. Abend, W. Schmatz, J. Schelten and K.D. Müller, Nucl. Instrum. Meth. 83, 111 (1970).
64. R. Berliner, D.F.R. Mildner, O.A. Pringle and J.S. King, Nucl. Instrum. Meth. 185, 481
(1981).
65. R.K. Crawford, J.R. Haumann, J.E. Epperson, P. Thiyagarajan, A.J. Schultz, G.P.
Felcher, D.G. Montague and R.J. Dejus, Nucl. Instrum. Meth. A 299, 17 (1990).
66. R.G. Cooper, Nucl. Instrum. Meth. A 529, 394 (2004).
67. V. Radeka, N.A. Schaknowski, G.C. Smith, and B. Yu, Nucl. Instrum. Meth. A 419, 642
(1998).
68. B. Guerard, http://jra2.neutron-eu.net/jra2/ (2008).
69. A. Oed, Nucl. Instrum. Meth. A 263 351(1988).
70. J.F. Clergeau, P. Convert, D. Feltin, H.E. Fischer, B. Guerard T. Hansen, G. Manzin,
A. Oed, and P. Palleau, Nucl. Instrum. Meth. A 471, 60 (2001).
4 Neutron Detectors for Imaging 63

71. Hiroyuki Takahashi, Prasit Siritiprussamee, Kaoru Fujita, Hisako Niko, Masaharu
Nakazawa, Shunji Kishimoto, Takashi Ino, Michihiro Furusaka, and Hidenori Toyo-
kawa, Nucl. Instrum. Meth. A 573, 175 (2007).
72. B. Gebauer, S.S. Alimov, A.Yu. Klimov, F.V. Levchanovski, E.I. Litvinenko, A.S.
Nikiforov, V.I. Prikhodko, G. Richter, V.V. Rogov, Ch. Schulz, V.I. Shashkin, M.
Wilhelm, and Th. Wilpert, Nucl. Instrum. Meth. A 529, 358 (2004).
73. Masayo Suzuki, Tan Takahashi, Hidekazu Kumagai, Masayasu Ishihara, and Hisao
Kobayashi, Nucl. Instrum. Meth. A 333, 484 (1993).
74. G. Manzin, B. Guerard, F.A.F. Fraga, and L.M.S. Margato, Nucl. Instrum. Meth. A 535,
102 (2004).
75. J.C. Buffet, J.F. Clergeau, R.G. Cooper, J. Darpentigny, A. De Laulany, C. Fermon, S.
Fetal, F. Fraga, B. Guerard, R. Kampmann, A. Kastenmueller, G.J. Mc Intyre, G. Manzin,
F. Meilleur, F. Millier, N. Rhodes, L. Rosta, E. Schooneveld, G.C. Smith, H. Takahashi, P.
Van Esch, T.L. Van Vuure, and K. Zeitelhack, Nucl. Instrum. Meth. A 554, 392 (2005).
76. M. Klein, H. Abele, D. Fiolka, and Chr. J. Schmidt, AIP Conf. Proc. 596, 193 (2001).
77. D.D. DiJulio, A.I. Hawari, and R. Berliner, Nucl. Instrum. Meth. A 579, 71 (2007).
78. F.A.F. Fraga, L.M.S. Margato, S.T.G. Fetal, M.M.F.R. Fraga, R. Ferreira Marques, A.
J.P.L Policarpo, B. Guerard, A. Oed, G. Manzini, and T. van Vuure, Nucl. Instrum. Meth.
A 478, 357 (2002).
79. S. Andriamonje, D. Cano-Ott, A. Delbart, J. Derre, S. Diez, I. Giomataris, E.M.
Gonzalez-Romero, F. Jeanneau, D. Karamanis, A. Lepretre, I. Papadopoulos, P. Pavlo-
poulos, and D. Villamarin, Nucl. Instrum. Meth. A 481, 120 (2002).
80. Charles L. Britton, Jr, William L. Bryan, Alan L. Wintenberg, Robert J. Warmack,
Timothy E. McKnight, Shane S. Frank, Ronald G. Cooper, Nancy J. Dudney, Gabriel
M. Veith, and Andrew C. Stephan, IEEE Trans. Nucl. Sci. 51, 1016 (2004).
81. Y. Kumashiro, and Y. Okada, Appl. Phys. Lett. 47, 64 (1985).
82. D.S. McGregor, T.C. Unruh, and W.J. McNeil, Nucl. Instrum. Meth. A 591, 530 (2008).
83. D.S. McGregor, M.D. Hammig, Y.-H. Yang, H.K. Gersch, and R.T. Klann, Nucl.
Instrum. Meth. A 500, 272 (2003).
84. J. Uher, T. Holy, J. Jakubek, E. Lehmann, S. Pospisil, and J. Vacik, Nucl. Instrum. Meth.
A 542, 283 (2005).
85. W.J. McNeil, S.L. Bellinger, B.J. Blalock, C.L. Britton, Jr., J.L. Britton, S.C. Bunch, S.A.
Cowley, C.M. Henderson, T.J. Sobering, R.D. Taylor, and D.S. McGregor, Nuclear
Science Symposium Conference Record, 2007. NSS ’07. IEEE 3, 2340 (2007).
86. Takekazu Ishida, Masatoshi Nishikawa, Shigehito Miki, Hisashi Shimakage, Zhen
Wang, Kazuo Satoh, Tsutomu Yotsuya, Masahiko Machida, and Masaru Kato, Physica
C 460–462, 618 (2007).
87. G.W. Fraser and J.F. Pearson, Nucl. Instrum. Meth. A 293, 569 (1990).
88. R.M. Ambrosi, G.W. Fraser, R.A. Street, J.I.W. Watterson, R.C. Lanza, J. Dowson, A.
F. Abbey, B. Feller, G. Downing, P. White, and T. Stevenson, Nucl. Instrum. Meth. A 542,
271 (2005).
89. Oswald H.W. Siegmund, John V. Vallerga, Adrian Martin, Bruce Feller, Muhammad
Arif, Daniel S. Hussey, and David L. Jacobson, Nucl. Instrum. Meth. A 579, 188 (2007).
90. A.S. Tremsin, J.V. Vallerga, J.B. McPhate, O.H.W. Siegmund, W.B. Feller, L. Crow, and
R.G. Cooper, Nucl. Instrum. Meth. A 592, 374 (2008).
Section B
Chapter 5
Neutron Radiography

A.K. Heller and J.S. Brenizer

Abstract Neutron radiography and its related two-dimensional (2D) neutron

imaging techniques have been established as invaluable nondestructive inspec-
tion methods and quantitative measurement tools. They have been used in a
wide variety of applications ranging from inspection of aircraft engine turbine
blades to study of two-phase fluid flow in operating proton exchange mem-
brane fuel cells. Neutron radiography is similar to X-ray radiography in that the
method produces a 2D attenuation map of neutron radiation that has pene-
trated the object being examined. However, the images produced differ and are
often complementary due to the differences between X-ray and neutron inter-
action mechanisms. The uses and types of 2D neutron imaging have expanded
over the past 15 years as a result of advances in imaging technology and
improvements in neutron generators/sources and computers. Still, high-intensity
sources such as those from reactors and spallation neutron sources, together with
conventional film radiography, remain the mainstay of high-resolution, large
field-of-view neutron imaging. This chapter presents a summary of the history,
methods, and related variations of neutron radiography techniques.

Keywords Neutron imaging  Neutron radiography  Neutron converter 

Direct radiographic method  Neutron imaging standards  Neutron computed
radiology

5.1 Introduction

Neutron radiography is a powerful nondestructive imaging technique that pro-

duces a two-dimensional (2D) attenuation map of neutrons that have penetrated
an object being examined. The method was initially developed after X-ray radio-
graphy, and the techniques share many similarities in setup and practice. X-ray

A.K. Heller (*)

Department of Mechanical and Nuclear Engineering,
The Pennsylvania State University, University Park, PA 16802, USA
e-mail: [email protected]; [email protected]

I.S. Anderson et al. (eds.), Neutron Imaging and Applications, 67

Neutron Scattering Applications and Techniques, DOI 10.1007/978-0-387-78693-3_5,
Ó Springer ScienceþBusiness Media, LLC 2009
68 A.K. Heller and J.S. Brenizer

and neutron radiography are often complementary techniques, especially when

low-energy neutrons are used. X-rays interact with orbital electrons and are
strongly tied to the physical density of the examined object. Neutrons interact
with an object’s nucleus rather than its orbital electrons, so there is usually no tie
to the object’s electron density, but rather its elemental composition. Because the
technique is based on attenuation from a well-collimated beam, either scattering
or absorption will result in intensity variations to create an image. Low-Z
materials such as hydrogen are easily imaged due to scattering, while boron
and cadmium are readily imaged due to their strong absorption.
New methods similar to radiography, but using nonfilm image detectors, have
been developed. To remove the confusion originally created by also referring to
nonfilm methods as neutron radiography, new terms such as radioscopy and
computed radiology have been adopted. These other methods are dealt with in
more detail in Chapter 6 of this volume.

5.2 A History of Neutron Radiography

Although the discovery of the neutron by English physicist James Chadwick

[1] occurred in 1932, the history of neutron radiography begins in 1935 when, in
Germany, Kallmann and Kuhn [2] performed the first experiments specifically
concerned with generating images using neutrons. The neutron radiographs
produced by these experiments were not of high quality and took 4 h due to the
small accelerator neutron source available during the study, which yielded a
low-intensity thermal neutron beam. Regardless, these early experiments did
provide insight into some of the possible uses of neutron radiography and the
detection methods used to generate neutron radiographs.
Around the same time, another experimenter in Germany, Peter [3], was con-
ducting a similar set of experiments using a much more intense neutron source.
Because of this, the neutron radiographs produced by Peter were of fair quality and
could be obtained in 1–3 min. Because of the Second World War, further devel-
opment of neutron radiography did not occur until the mid-1950s. Indeed, Peter
had to wait until 1946 to publish his results and Kallmann and Kuhn until 1948.
Despite the War, nuclear reactor technology development continued, with
advances during and after the War that would increase the intensity of neutron
fluxes available to researchers by many orders of magnitude. The first instance of a
neutron radiograph being produced by a beam of thermal neutrons from a reactor
occurred in 1956. Thewlis and Derbyshire [4] used the 6 MW BEPO1 graphite
reactor at Harwell in England for their neutron source, producing radiographs of
better quality than those made previously by Peter and Kallmann. This work
helped to illustrate a number of possible applications for the technique, specifically
the inspection of neutron shielding materials and the study of organic specimens.

1
British Experimental Pile ‘0,’ the UK’s second reactor.
5 Neutron Radiography 69

Beginning in 1960, several independent studies to further the technique were

undertaken at various laboratories by individuals such as Berger [5, 6], Watts
[7], Barton [8], Criscuolo and Polansky [9], and Schultz and Leavitt [10]. It was
during this time period that neutron radiography established itself as a viable
nondestructive inspection method.
In the 1970s, the majority of research reactors throughout Europe and Amer-
ica had facilities in place capable of taking quality neutron radiographs. It soon
became evident, however, that standardization was needed if further develop-
ment of neutron radiographic techniques was to be achieved. Initial steps in this
direction were taken in 1973 at Birmingham, England [11], and at Gaithersburg,
Maryland (USA), in 1975 [12], but the pivotal moment occurred in 1979 when,
through The Commission of European Communities, a Neutron Radiography
Working Group was formed. This group initiated an extensive research program
aimed at determining the best techniques for producing neutron radiographs and
the accuracy of the dimensional measurements associated with them. This Eur-
opean effort, coupled with American and Canadian interest in the same area, led
to the first of many World Conferences on Neutron Radiography [13] in 1981, the
same year the group published its Neutron Radiography Handbook [14].
In the early 1980s, electronic, real-time neutron radiography emerged [15],
which made use of neutron image intensifier tubes to increase image production
rates and videocassettes as the storage medium. The dynamic nature of this
technique found a wide variety of applications from investigations of internal
combustion engines [16] to analysis of two-phase flow [17]. By the mid-1980s,
images were being digitized and stored on computers [18], allowing the applica-
tion of advanced imaging processing techniques for quantitative analysis [19].
From the 1990s to the present day, the imaging systems used to perform neutron
radiography and the techniques used to analyze the images have continued to
advance. The enhancement of existing detectors with, for example, thinner scintil-
lation screens, and the development of new detectors like the microchannel plate,
have yielded continual increases in resolution. Digital cameras with faster readouts
and higher dynamic range are now used, and the analysis of uncertainty associated
with measured optical density has provided more accurate results. These advance-
ments have allowed neutron radiography to evolve from a nondestructive inspec-
tion method into a measuring method using neutrons as microscopic probes [20].

5.3 Basic Principles

5.3.1 Sources

Neutron radiography requires, of course, a source of neutrons. There are three

general types of neutron sources: accelerator, radioisotope, and nuclear reactor.
Accelerator-based neutron sources are ones that accelerate and direct a beam of
charged particles such as protons, deuterons, and alphas onto a target, which
then results in the emission of a neutron. There are a variety of combinations of
70 A.K. Heller and J.S. Brenizer

incident particle and target material that can be used, a list of which may be found in
von der Hardt [14]. Accelerators produce thermal neutrons in the range from 107 to
1010 n/cm2/s and offer the benefit of intermittent operation and portability. Sealed-
neutron generator tubes are easy to use but generally produce low fluxes compared
with high current accelerators, which are moderately complex to operate and have a
target output that may deteriorate with use. Large spallation sources can achieve fluxes
of more than 1014 n/cm2/s, but such sources are very complex and not portable.
A radioisotope-based neutron source makes use of radioactive isotope decay
to generate its neutron flux in much the same way as accelerators: by allowing
the gamma rays or alpha particles emitted by a radioactive isotope to bombard
a neutron emitting target. A number of (,n) and (,n) radioisotopic neutron
sources exist, tables of which may be found in von der Hardt [14] and Berger [8].
Radioisotope-based neutron sources have low thermal neutron fluxes, within
the range of 105–109 n/cm2/s, and also suffer from decreasing output, from both
deterioration of the target and decay of the source. The inability to turn off the
radiation is often considered a disadvantage. The major advantages these
systems possess are their simple operation and ease of portability.
Nuclear reactor-based neutron sources provide intense neutron beams and
thus high-quality neutron radiographs. Thermal neutron fluxes obtainable in
such facilities range from 1010 to 1015n/cm2/s or even higher. It should be noted
that large spallation neutron facilities may produce beams with intensities equal
to or higher than those produced by nuclear reactors. However, reactors
typically provide a lower cost per neutron than accelerators. Their major dis-
advantages are the high costs associated with construction; licensing and regula-
tory requirements; complexity to operate; and lack of portability.

5.3.2 Moderation

The neutrons born in each of the sources discussed above possess high energies
with a continuous spectrum of energies peaking from 0.85 MeV from fission (in
reactors) up to 14 MeV (in accelerators). Conventional neutron radiography,
however, requires neutrons in the thermal/epithermal energy range of
0.025 eV–10 keV. Thus, some form of moderator with low neutron absorption
cross section (to maximize flux) and high scattering cross section (to maximize
energy loss) is required to slow down the neutrons to this energy range. The
often-used moderator materials of water, heavy water, graphite, beryllium, and
polyethylene meet these criteria. In this, the nuclear reactor has an inherent
advantage: the moderation of its core already produces a low-energy spectrum
resulting in fewer neutrons lost in the moderation process.

5.3.3 Collimation
Once low-energy neutrons are produced, they must be formed into a usable
beam. Neutrons are emitted and then scattered randomly in the moderator and,
5 Neutron Radiography 71

because of their neutral charge, they cannot be focused like electrons. Those
neutrons traveling in the desired solid angle can, however, be selected by the
introduction of a tube into or adjacent to the moderator. This has the effect of
allowing neutrons to stream down the tube axis toward the object being radio-
graphed. The walls of the collimator tube are lined with a neutron opaque
material having a high absorption cross section (such as boron, gadolinium,
and cadmium), which prevents stray neutrons from entering and also reduces
low-angle scattering within the collimator. The most common collimator design
is a divergent collimator (Fig. 5.1) with a small entrance aperture and a larger
exit. This maximizes the neutron flux and permits a larger field at the imaging
plane. The angular spread of the emerging beam is dependent upon the ratio of
the collimator tube length (L) to its aperture diameter (D), referred to as the L/D
ratio. A higher L/D results in a narrower beam spread at the expense of a lower
neutron flux. This ratio is a characteristic parameter of each collimator. Extensive
discussions regarding collimator design may be found in von der Hart [14]
and Domanus [21].

Fig. 5.1 Divergent

collimator illustrating the
ratio of the collimator tube
length (L) to its aperture
diameter (D) (the L/D ratio)

5.3.4 Detectors

After a neutron’s birth in the source, its moderation to thermal energies, and its
escape along the collimator tube, it will encounter the object to be radiographed.
Any neutrons that successfully penetrate the object must then be detected to
produce the radiograph. In neutron radiography, a detector collectively refers to
both an intermediate medium, called a converter (which emits an alpha, beta,
gamma, or light when neutrons are absorbed) and the sensor used to detect this
emitted radiation, called the image recorder. The converter material is used
because it emits a much more readily detectable radiation.
When the image recorder is film, one possible converter material is a gado-
linium foil which emits an electron with every absorbed neutron. The converter
foil is placed in direct contact with the film’s emulsion and the emitted electrons
expose the emulsion, producing an image. A typical spatial resolution using a
single-coated fine-grain radiographic film, a vapor deposited gadolinium con-
verter, and a vacuum cassette is 10 mm.
Another possible converter is a scintillation screen, which will expose the
film’s emulsion with light. A scintillation screen is 30–100 times faster at
producing an image on radiographic film than a gadolinium foil. However,
72 A.K. Heller and J.S. Brenizer

due to light spread within the scintillator, the spatial resolution is reduced.
A typical spatial resolution for a scintillator and film is 100 mm. Both of these
converters continuously emit radiation for the duration of neutron exposure
and can therefore be used in low neutron flux environments with long integra-
tion exposures.
An electronic form of imaging where a scintillation converter is optically
coupled with an analog or digital camera image recorder can also be used. This
allows the rapid capture of successive neutron radiographic images that can
be viewed directly and stored on videocassette or digital media, preserving
dynamic information. This electronic form of producing radiographic images
is referred to as radioscopy.
Other types of analog and electronic detectors exist and vary depending
upon the imaging technique used. They are discussed in later sections of this
chapter.
To summarize, a neutron radiography system consists of a neutron source, a
moderator to thermalize the neutrons, an aperture and a collimator to organize
the neutrons into a beam, and a detector to visualize the image (Fig. 5.2).

Moderator
Image
Aperture Recorder

Object
Neutrons
Neutron Source
Converter

Fig. 5.2 Illustration of a typical neutron imaging system

5.4 Image Analysis

Any analysis of a radiographic image, be it film or electronic, begins with an

understanding of how the image is formed. The relationship between the incident
neutron intensity upon an object to be radiographed and the transmitted neutron
intensity (ignoring scattering) is the simple exponential attenuation law

 ¼ o ei t : (5:1)
5 Neutron Radiography 73

The transmitted neutron intensity, , is a function of the incident neutron

intensity, 0 , and the product of the total macroscopic cross section and thick-
ness of the object, t t [14].
In the case of film, the degree of film darkening (photographic density, De ) is
related to the neutron exposure by the film’s characteristic response curve. De
will have a logarithmic nature as described by

De ¼ G ðlog EÞ ; (5:2)

where E is the exposure of the film (transmitted neutron intensity multiplied by

time, T) and G is the slope in the linear portion of the characteristic response
curve for the film being used; it is a parameter describing the manner in which a
particular film responds to an exposure. This is the manner in which images are
formed on film. One must bear in mind that when the image is being viewed, the
processed film’s photographic density is described by
 
Io
De ¼ In ; (5:3)
I

where Io is the incident light (such as from a light box) and I is the transmitted
light through the film.
In nearly all forms of digital imaging, the resulting grey level value of any
pixel making up the image may be described by

G ¼ C   þ Goffset ; (5:4)

where G is the numerical grey level value of the pixel within an image, C is the
electronic gain of the camera or imaging system (a constant),  is the trans-
mitted neutron intensity and Goffset is the dark current, an additive offset due to
electronic noise.
These equations form the basis of all radiographic image analysis. With
them, one may manipulate images to isolate terms and perform quantitative
analyses or provide the basis for qualitative comparisons. Extensive descrip-
tions and applications of these equations may be found in [14, 20–22].

5.5 Direct Radiographic Method

The direct radiographic method refers to the technique by which a radiogra-

phic image is generated when the converter is in direct contact with the image
recorder. Traditionally, this has referred to detectors using film. The converter
screen used more often than not is a gadolinium foil, which absorbs neutrons
and emits gamma rays that are internally converted to low-energy electrons.
These electrons expose the film’s emulsion, after which the film must be
74 A.K. Heller and J.S. Brenizer

developed. Converter screens of this type are usually 25-mm thick and are either
gadolinium foils laminated to aluminum or vapor-deposited gadolinium on
aluminum for ease of handling. Vacuum cassettes are used to ensure good contact
between the converter and image recorder, which is vital in reducing image blur.
The resolution of this converter–imager recorder combination is 10–20 mm. This
resolution is quite impressive, considering it is obtainable over very large areas
using films up to 35.5 cm by 43 cm.
In a similar fashion, a light-emitting scintillation screen may be used as the
converter. In these screens, a neutron-absorbing material that yields a charged
particle is mixed with a phosphor material that produces light. An example is a
mixture of lithium-6 and ZnS(Ag). Lithium-6 emits an alpha particle when
absorbing a neutron, and it is the kinetic energy of the alpha particle that causes
the ZnS(Ag) to emit light. This light then exposes the film emulsion. Scintilla-
tion screens are thicker than gadolinium foils and, because of the light spread
within them, the obtainable resolution (75–100 mm with the best screens) is less
than that of gadolinium foils.
The direct imaging method has a major disadvantage when it involves
nuclear applications. A radioactive object emitting gamma rays as it decays
or a neutron beam contaminated with gamma rays will directly produce an
image on the film. Since the gamma rays are from a different source than the
neutrons, the images will be different and the film will be blurred or what is
known as ‘‘gamma fogged.’’ Another disadvantage of the method is the time
associated with the film development, which inherently prohibits the investiga-
tion of dynamic processes.
There are, however, a wide variety of applications for which this method
is well suited. In the area of turbine manufacturing, the direct radiographic
method can easily perform a quality control check on the presence of residual
materials used in the blade’s production (Fig. 5.3). In fact, the technique has
been used for quality control inspections for moisture, corrosion, adhesive
defects, proper lubrication, and quality of seals in the aerospace and automotive
industries. It is also used in the health monitoring and maintenance of in-service
components such as aircraft flight control surfaces [23]. The direct method also

Fig. 5.3 Neutron

radiograph of a turbine
blade using a gadolinium
contrast agent to accentuate
blocked channels
5 Neutron Radiography 75

plays a major role in the field of research, where it has been used to investigate
two-phase flow behavior in heat pipes, water distribution in fuel cells, and water
permeability in concrete, to list a few examples.

5.6 Indirect Radiographic Method

The indirect radiographic method applies when the converter screen is not in
direct contact with the image recorder. This is also referred to as the transfer
technique because of the manner in which the image is produced. A metal foil
converter is placed in the beam independent of the image recorder. The foil
builds up a radioactivity through neutron absorption, producing what is
called an activation image. After being removed from the beam, the activated
foil is placed in contact with the image recorder and the decay radiation
emitted (low-energy electrons) ‘‘transfers’’ the activation image to the image
recorder. In this method, the image recorder is film, just as with the direct
radiographic method, but the converter screen used is indium, dysprosium, or
gold because of the need for the foil to rapidly activate for image transfer and
rapidly decay for reuse. The image quality produced by these techniques is the
same as that of the direct techniques and provides spatial resolutions of
10–20 mm.
The technique is slower than the direct method, but one major advantage is
that the converter foil used is insensitive to gamma radiation from the object
or beam, making it an ideal candidate for use in nuclear applications such as
the investigation of spent fuel rods. Gamma rays produced by radioactive
decay in these applications will not fog up the radiographic image. The fact
that this technique is slower than the direct method does not, however,
preclude it from being applied to the same fields as the direct radiographic
method. Indeed, the indirect radiographic method is equally applicable to the
same industries, research, and quality control as the direct radiographic
method.

5.7 Track-Etch Method

Technically a form of the direct radiographic method because of the direct

contact between the converter and image recorder, the track-etch method is
discussed independently here because of its uniqueness. The image recorder is
nitrocellulose film, a dielectric material capable of detecting charged particles
by the radiation damage caused within it, and it is placed between two alpha-
particle emitting converter screens, such as boron or lithium. Unlike the low-
energy electrons emitted by the metal foil converter screens used in the direct
and indirect radiographic methods, the alpha particles take short and relati-
vely straight paths through the nitrocellulose film, giving good resolution. The
76 A.K. Heller and J.S. Brenizer

radiation damage within the film is made visible by etching it in a hot base
solution, such as sodium hydroxide. A vacuum cassette is used to ensure tight
contact between the film and the converter screens.
The track-etch method is insensitive to gamma rays and also to visible light,
allowing the etching to be performed in daylight. However, the exposure time
needed for this technique is longer than the direct and indirect methods, and the
contrast is also weaker. A unique advantage is the ability to stop the etching
process at intermediate stages, which can then be continued after evaluating
the resultant neutron radiograph. In the end, several radiographs of varying
densities and contrasts can be had. Because of the gamma-ray insensitivity,
radiography of radioactive objects, such as spent nuclear fuel, is the primary
application for this method.

5.8 Electronic Imaging Methods

Film is not the only form of image recorder possible. There are techniques that use
electronic means, both analog and digital, for the image recorder. The production
of radiographic images via electronic image recorders, initially referred to as real-
time neutron radiography, is now known as radioscopy [24]. Electronic image
recorders depend upon the capture of light to record an image and, because they
have the ability to rapidly acquire these images, they are universally coupled with
a converter that has an equally rapid (if not faster) light response: a scintilla-
tion screen. Thus, these methods are frequently used to view and store dynamic
radioscopic images.
In electronic imaging systems, the image recorder can be a video camera if an
analog system, or a charge-coupled device (CCD) camera if a digital one. In the
case of an analog system, the camera is connected to a television and VCR for
the real-time viewing and recording of the dynamic radioscopic images. In a
digital system, the CCD camera is connected to a computer that displays and
stores the radioscopic images in a digital format. A combination of these systems
may also exist wherein the analog television camera can be connected to a
computer via an analog-to-digital converter card which will digitize the camera’s
analog signal and store it.
The camera, be it analog or digital, is coupled to the scintillation screen via a
light-tight box to prevent polluting the radioscopic image with outside light.
The intensity of light produced by the scintillation screen is linear with the
intensity of incident neutron flux, and there are a number of reactor facilities
where the neutron flux intensity is sufficiently large for the light produced by the
scintillation screen to be seen on a TV monitor. However, there are many
facilities that do not possess a neutron flux of such intensity, and in these
circumstances an image intensifier is placed within the light-tight box between
the scintillation screen and the camera. The image intensifier boosts the light
from the scintillation screen by about 104, still providing good image definition
5 Neutron Radiography 77

and allowing the radiographic image to be viewed. In addition, a mirror is often

placed between the camera and image intensifier (or scintillation screen if an
intensifier is not needed), forming a right angle and allowing the camera to be
positioned outside the beam path to reduce exposure.
In certain circumstances, an electronic image recorder may be used, but not
with the intent of dynamic imaging. If very high signal-to-noise ratios are
needed, such as in computed tomography applications, the integration time
can be increased to obtain the required ratio. The use of a digital camera makes
the image acquisition easier by eliminating the need to develop a film and,
because the image is digital, provides ease of image manipulation.

5.9 Nonfilm Imaging Methods

Recently new sensors called photostimulable luminescence (PSL) imaging

plates have been developed that can be used for X-ray and neutron imaging.
Photostimulated luminescence is a phenomenon where a phosphor is first
exposed to light or charged particle radiation [25]. After removal from the
exposing radiation, the PSL plate retains a stored image which can be read
out by later exciting the previously exposed phosphor to longer wavelength
light. The 2D flexible storage phosphor that can store a latent image from
radiation is called an imaging plate. The overall imaging process—where an
imaging plate is first exposed to gamma-ray, X-ray, or neutron beams in a
manner similar to traditional radiography and is subsequently read by means of
photo stimulation to obtain a radiographic image—is called computed radiol-
ogy (CR) [24].
Imaging plates can be used as a direct replacement for radiographic film.
There are several advantages of using CR instead of traditional film. Imaging
plates can be erased and reused for potentially thousands of exposures. Physical
damage from handling is the common limiting factor. Readout is accomplished
in a scanner-type device that reads out the latent image, stores a digital image,
and erases the imaging plate for the next use. Imaging plates have been shown to
have a much greater sensitivity range than photographic film, allowing a wide
linear dynamic range of eight orders of magnitude. This is very advantageous
for neutron imaging, where neutron intensity at the imaging plane is usually
lower than desired. CR has the additional advantage of eliminating film proces-
sing, thereby making the imaging process faster, independent of processing
chemicals, and more environmental friendly than film processing. While not
inexpensive compared with simple radiography with film and hand processing,
CR systems offer significant cost savings over film when large volumes of
radiographs are needed.
The major disadvantage of CR is the spatial resolution, typically 100 mm,
compared with the 15 mm obtainable with film [26, 27, 28].
78 A.K. Heller and J.S. Brenizer

5.10 Standards

As early as 1969 it was recognized that some standardization in neutron imaging

was needed. Haskins presented two reviews of neutron radiography standards in
the United States [29, 30], and a later paper by Newacheck and Tsukimura
updated these earlier papers [31]. The American Society for Testing and Materi-
als International Committee E07 on Nondestructive Testing has developed a
suite of standards for neutron radiography. There are currently no standards for
nonfilm neutron imaging.
E748, Standard Practices for Thermal Neutron Radiography of Materials,
provides a good introduction to neutron radiography [32]. The document is
tutorial in nature, describing common practices, facilities, and necessary equip-
ment. This standard also provides example applications and some basic gui-
dance for determining the practicality of the method.
E545, Standard Test Method for Determining Image Quality in Direct Ther-
mal Neutron Radiographic Examination, has become the world standard for
determining the relative overall quality of neutron radiographs [33]. It is not
intended to be used for controlling the acceptability or quality of materials or
components. Radiographic quality is based upon the evaluation of images
obtained from two different indicators, the beam purity indicator and the
sensitivity indicator. The information obtained from radiographic images
using these devices is used to determine a facility’s neutron radiographic
category.
E803, Standard Test Method for Determining the L/D Ratio of Neutron
Radiography Beams, is a method widely used by radiographers to characterize
neutron beams [34]. Knowledge of the L/D ratio is important for understanding
the geometric ‘‘unsharpness’’ of imaged objects as a function offset from the
imaging detector.
A fourth standard, E1496, Standard Test Method for Neutron Radiographic
Dimensional Measurements, presents a method that can be used to obtain
quantitative length dimensions reproducibly from a radiographic image [35].

5.11 Conclusions
Neutron radiography and its related 2D neutron imaging techniques have
established themselves as invaluable nondestructive inspection methods and
quantitative measurement tools. They have been used in a wide variety of
applications ranging from inspection of aircraft engine turbine blades to study-
ing two-phase fluid flow in operating proton exchange membrane fuel cells.
Advances in digital cameras, CR using imaging plates, and the development of
more intense and robust neutron generator tubes will expand the use of neutron
imaging. These advances will also increase portability, making it possible to take
imaging systems to the field rather than bringing objects to a fixed imaging
5 Neutron Radiography 79

facility. Improvements in digital image processing coupled with the increased

computational power of modern processors provide the ability to collect and
analyze images rapidly. This allows imaging of dynamic events at spatial and
temporal resolutions sufficient to provide the qualitative and quantitative infor-
mation needed for many new applications. However, high-intensity sources, such
as those from reactors and spallation neutron sources, with conventional film
radiography will remain the mainstay of high-resolution, large field-of-view
neutron imaging into the foreseeable future.

References
1. J. Chadwick, Possible Existence of a Neutron, Nature 129, 312 (1932).
2. H. Kallmann, Neutron Radiography Research 1, 254 (1948).
3. O.Z. Peter, Neutronen-Durchleuchtung, Naturforsch. 1, 557 (1946).
4. J. Thewlis and R.T.P. Derbyshire, Report AERE M/TN 37, U. K. Atomic Energy
Research Establishment, Harwell England (1956).
5. H. Berger, Proceedings Symposium on Physics and Nondestructive Testing, Argonne,
Illinois ANL-6346, Argonne National Laboratory, Argonne Illinois (1960).
6. H. Berger, Neutron Radiography: Methods, Capabilities and Applications, Elsevier Pub-
lishing Company, New York (1965).
7. H.V. Watts, Report ARF 1164-6, Armour Research Foundation, Chicago, Illinois (1960).
8. J.P. Barton, Introductory Remarks Neutron Radiography: Proceedings at the Second World
Conference on Neutron Radiography, Paris, France (June 16–20, 1986), J. Barton, J. Farny,
J. Person, H. Röttger, D. Reidel, eds. Pub. Co., Dordrecht, Holland, p. 13 (1987).
9. E.L. Criscuolo and D. Polansky, Proceedings Missiles and Rockets Symposium, Concorde,
California, 1961, U. S. Naval Ammunition Depot, Concord, California, p. 112 (1961).
10. A.W. Schultz and W.Z. Leavitt, Report WALTR 142.67,Watertown Arsenal Labora-
tories, Watertown Massachusetts (1961).
11. M.R. Hawkesworth (technical editor) Radiography with Neutrons, Conference held
10–12 September, 1973, at the University of Birmingham, London, British Nuclear
Energy Society (1975).
12. H. Berger (editor) Practical Applications Of Neutron Radiography And Gauging, ASTM
STP 586 (1976).
13. J.P. Barton and P. von der Hardt (editors) Neutron Radiography Proceedings of the First
World Conference, San Diego, California, U.S.A. , December 7–10, 1981. D. Reidel
Publishing Company, Dordrecht: Holland/Boston: U.S.A./London, England (1983).
14. P. von der Hardt and H. Röttger (editors), Neutron Radiography Handbook, D. Reidel
Publishing Company, Dordrecht: Holland/Boston: U.S.A./London, England (1981).
15. J.S. Brenizer and M.F. Sulcoski Real-Time Neutron Radiography at the University of
Virginia in Use and Development of Low and Medium Flux Research Reactors Harling,
O.K., Clark, L., and von der Hardt, P., eds., Karl Theiemig Graphische, W. Germany,
p. 958 (1984).
16. J.D. Jones, J.T. Lindsay, C.W. Kaufmann, A.T. Vulpeti and B. Peters, Real Time
Neutron Imaging Applied to Internal Combustion Engine Behavior SAE Technical
Paper Series No. 850560, International Congress and Exposition (1985).
17. M. Tamaki, K. Ohkubo, Y. Ikeda, and G. Matsumoto Analysis of Two Phase Flow in
Heat Pipe by Neutron Radiography Proceedings of the Second World Conference on
Neutron Radiography, Paris, France, June 16–20, 1986 Barton, G. Farny, J. Person, H.
Röttger, D. Reidel, eds. Pub. Co., Dordrecht, Holland, pp. 609–616 (1987).
80 A.K. Heller and J.S. Brenizer

18. J.S. Brenizer, B. Hosticka, R.W. Jenkins Jr and D.D. McRae An Advanced Video System
for Real-Time Neutron Radiography Proceedings of the Second World Conference on
Neutron Radiography, Paris, France, June 16–20, 1986, Barton, G. Farny, J. Person, H.
Röttger, D. Reidel, eds. Pub. Co., Dordrecht, Holland, pp. 571–578 (1987).
19. R. Polichar and D. Shreve Processing of Real-Time Images for Quantitative Neutron
Radiography Proceedings of the Second World Conference on Neutron Radiography, Paris,
France, June 16–20, 1986, J. Barton, G. Farny, J. Person, H. Röttger, D. Reidel, eds. Pub.
Co., Dordrecht, Holland, pp. 587–593 (1987).
20. T. Hibiki and T. Mishima Prediction of Measurement Error Due to Low Gray Scale
and Spatial Resolution of an Imaging System on Quantification of Neutron Radiogra-
phic Image Nuclear Instruments and Methods in Physics Research Section A, vol. 338,
pp. 204–211 (1997).
21. J.C. Domanus (editor) Practical Neutron Radiography, Kluwer Academic Publishers
Dordrecht: Holland/Boston: U.S.A./London: England (1992).
22. R.A. Quinn and C.C. Sigl (editors) (1980) Radiography in Modern Industry, 4th edition,
Eastman Kodak Company, Rochester NY.
23. T.R. Chalovich, L.G.I. Bennett, W.J. Lewis, and J. Brenizer, Jr., Development of
Neutron Radioscopy for the Inspection of CF188 Flight Control Surfaces, Applied
Radiation and Isotopes, vol. 61, pp. 693–700 (2004).
24. E07 on Nondestructive Testing ASTM E1316 Standard Terminology for Nondestructive
Examinations, Annual Book of ASTM Standards vol 03.03 (2007).
25. E07 on Nondestructive Testing E2007 Standard Guide for Computed Radiology (Photo-
stimulable Luminescence (PSL) Method) Annual Book of ASTM Standards, vol. 03.03
(2007).
26. M. Thoms, D. Myles and C. Wilkinson Neutron Detection with Imaging Plates Part I.
Image Storage and Readout, Nuclear Instruments and Methods in Physics Research A, vol.
424, pp. 26–33 (1999).
27. M. Thoms Neutron Detection with Imaging Plates Part II. Detector Characteristics
Nuclear Instruments and Methods in Physics Research A, vol. 424, pp. 34–39 (1999).
28. H. Kolbe, E Lehmann, W. Gunia and S. Körner Applications and Characteristics of
Imaging Plates as Detector in Neutron Radiography at SINQ Nuclear Instruments and
Methods in Physics Research A, vol. 424, pp. 40–47 (1999).
29. J.J. Haskins ASTM Activities in Neutron Radiography Practical Applications of Neutron
Radiography and Gauging, ASTM STP 586, H. Berger (editor) American Society for
Testing and Materials, Philadelphia, pp. 106–113 (1976).
30. J.J. Haskins Neutron Radiography Standards in the United States of America Neutron
Radiography Proceedings of the First World Conference, San Diego, California, U.S.A. ,
December 7–10, 1981, Barton JP and von der Hardt P (editors) D. Reidel Publishing
Company. Dordrecht: Holland/Boston: U.S.A./London: England, pp. 985–991 (1983).
31. R.L. Newacheck and R.R. Tsukimura Current Status of the ASTM E545 Image Quality
Indicator System Neutron Radiography (3) Proceedings of the Third World Conference on
Neutron Radiography Osaka, Japan (May 14–18, 1989), S. Fujine, K. Kanda, G. Matsumoto,
J. Barton, eds. Pub. Co. Kluwer Academic, Dordrecht, The Netherlands, pp. 875–883 (1990).
32. E07 on Nondestructive Testing E748 Standard Practices for Thermal Neutron Radio-
graphy of Materials Annual Book of ASTM Standards, vol. 03.03 (2007).
33. E07 on Nondestructive Testing E545 Standard Test Method for Determining Image
Quality in Direct Thermal Neutron Radiographic Examination Annual Book of ASTM
Standards, vol. 03.03 (2007).
34. E07 on Nondestructive Testing E803 Standard Test Method for Determining the L/D
Ratio of Neutron Radiography Beams Annual Book of ASTM Standards, vol. 03.03 (2007).
35. E07 on Nondestructive Testing E1496 Standard Test Method for Neutron Radiographic
Dimensional Measurements Annual Book of ASTM Standards, vol. 03.03 (2007).
Chapter 6
Neutron Tomography

W. Treimer

Abstract This chapter provides an introduction to neutron tomography. The

basic factors influencing instrument design and capability are discussed and the
mathematical methods used for image reconstruction are explained. Some of
the more promising new techniques, which are likely to have an increasing range
of applications, are described in more detail. These include energy (wave-
length)-dispersive radiography and tomography, real-time radiography, phase
contrast, refraction, and small-angle tomography. In the latter methods, there is
an interesting overlap/complementarity between the real-space imaging aspects
and the Fourier space imaging aspects, particularly as this gives rise to the
possibility of imaging to a resolution of 10 mm or better.

Keywords Foundations of CT  Geometry  Point spread function 

Reconstruction  Phase contrast  Refraction and USANS tomography 
Neutron cross section  Scintillator  Neutron flux  Neutron intensity  Spatial
resolution  Time resolution  Detector  Point spread function  Fourier
transform  Fourier space  Modulation transfer function  Attenuation 
Absorption  Scattering  Image reconstruction  Projection  Filtered-back
projection  Grey value  Energy-dispersive radiography  Energy-dispersive
tomography  Bragg-edge radiography  Bragg-edge tomography 
Wavelength  Wedge transmission  Real-time radiography  Phase contrast 
Refraction  Small-angle tomography

W. Treimer (*)
Department of Mathematics, Physics and Chemistry, University of Applied Sciences
(TFH) Berlin, Luxemburgerstr. 10, D-13353 Berlin; Helmholtz Center Berlin for
Materials and Energy, Department SF1, Glienicker Str. 100, D-14106, Berlin
e-mail: [email protected], [email protected]

I.S. Anderson et al. (eds.), Neutron Imaging and Applications, 81

Neutron Scattering Applications and Techniques, DOI 10.1007/978-0-387-78693-3_6,
Ó Springer ScienceþBusiness Media, LLC 2009
82 W. Treimer

6.1 Introduction

Tomography and radiography are well known from their uses in medicine, and
many people have already come into contact with one or more of these meth-
ods. They are closely related because tomography is based on radiography,
which is a two-dimensional (2D) attenuation coefficient distribution of a ray-
path-integrated projection from a three-dimensional (3D) object (Chapter 5). In
the past, X-ray film-converter systems were used for neutron imaging; but since
the mid-1990s digital imaging systems, e.g., image plates or charge-coupled
device (CCD) cameras with scintillator screens, have become increasingly avail-
able (Chapter 4). The quality of digital images is now comparable to that
obtained using film. The advantages of CCD camera-based imaging systems
are easy image processing, high bit depth (optical density), and real-time
imaging and information handling (delivery, storage, post-processing). Thus
radiographic and tomographic techniques for neutron imaging have become
easier to handle and to apply with higher performance since the advent of PC-
based electronic media. It is only a short step from a radiograph to a tomogram,
so some fundamental parameters for both techniques are briefly summarized
here.
For neutron radiography and tomography, four parameters characterizing
the neutron beam are of principal interest: the incident flux intensity , the
wavelength spectrum spread
l/l, the divergence , and the given spin state of
the neutron beam in the case of polarized neutrons (Chapter 10). The geometry
required for a radiography setup, for example the distance between sample and
detector, is influenced by the size of the sample and by the pixel size (smallest
scanning part) of the detector. The flux and the wavelength spread are directly
connected because the smaller the
l/l, the lower the neutron flux. This also
holds true in principle for the achievable spatial resolution—the smaller the
pixel size (i.e., the higher the spatial resolution), the smaller the integrated flux
at the pixel. Therefore, for conventional radiography and tomography (i.e.,
attenuation-based neutron imaging techniques), the full spectrum (white beam)
is used to get acceptable results in short times. For various other cases
of tomography, the use of monochromatic neutrons is necessary, so a certain

l/l, typically 1–5%, is required to determine structures in samples that are not
revealed by conventional tomography.
The basic properties of the neutron, for example, the relationships between
wavelength, energy, and velocity, and the scattering and absorption cross
sections, have already been introduced in Chapter 1. Scattering processes are
unwanted effects in a classical radiograph or tomogram because they cause a
decrease of sharpness in the image. However, it must be noted that scattering
processes can also be used as imaging signals using different imaging techniques
(Section 6.4). X-ray scattering cross sections generally have an atomic
number–dependent relationship, while neutron cross sections typically have a
‘‘statistical’’ behavior, as shown in Fig. 6.1 (see also Fig. 2.2).
6 Neutron Tomography 83

Fig. 6.1 X-ray and neutron cross sections for different elements. The sizes of the circles
symbolize the strengths of the scattering cross sections

6.2 Geometric Considerations and Spatial Resolution

For tomography (and radiography), neutrons are collimated by slits, apertures,

or collimator systems to reduce the range of directions in which the radiation
propagates and to obtain a defined point-to-point image. For a (virtual) source
of a given size, a point P in the object (sample) plane is enlarged to have a
diameter d as a result of the incident divergence   L/D, where L is the distance
of the point from the source and D the source dimension, and of the distance ld
from the detector to the object (Fig. 6.2).
However, in order to obtain an image in a reasonable exposure time, the
available neutron flux is also important. The exposure time that is reasonable
depends on the problem to be solved and can range from fractions of seconds to
several minutes. In every case, the exposure time must be optimized to the
available neutron flux.
The neutron intensity incident on a sample to be imaged, I0, is determined by the
neutron flux Isource coming from the source decreased by l/L2. The horizontal
collimation h and vertical collimation v define a solid angle
¼ h  v so
that the intensity I0 is proportional to the area of the surface of the sphere
A= dsh 
dsv. The elements dsh,v can be determined as h,v R. So for a given divergence  (which

Fig. 6.2 Geometric point broadening: A point P in the object plane is enlarged (blurred) up to
a size d as a result of a given L/D
84 W. Treimer

can be measured), the incident intensity I0 in front of a sample can be calculated from
a given source flux Isource. To realize a point-to-point image, the size of the source and/
or the beam divergence must be kept small. This means that a large L/D ratio
corresponds to a small horizontal and vertical divergence. Furthermore, one must
consider that a point in an object is also enlarged and blurred to an area in the image
by the object-detector (e.g., film, CCD camera) distance, and by scattering effects.
Every point in a sample is enlarged to an area with diameter d on the detector
as d ¼ l/(L/D) if L/D defines the (inverse) divergence h,v (Fig. 6.2). One can
calculate that even for L/D ¼ 1000 (divergence h,v 0.0578 ¼ 0.001 rad), a
point 50 mm from the detector is enlarged to 50 mm. Additional broadening of
an object point is also caused by the process of converting neutrons into visible
light; for example, the thickness of the scintillator material and the optical
system focusing the point onto the CCD chip might inherently decrease the
sharpness of the image. However, it must be noted that in the case of refraction,
or of phase radiography and tomography, the general rule that a shorter
distance between object and detector means a sharper image does not hold.
The imaging system, i.e., the imaging of a given point in a sample, can be
described by the so-called point spread function (PSF) because it involves not
only the interaction with the sample but also the detector resolution (pixel size),
the spread of radiation due to divergence and scintillation (detection) processes,
and optical systems (CCD cameras). The PSF is also known as the impulse
response (Chapter 7). The spatial 2D resolution of a system is best measured by
determining its modulation transfer function (MTF; Fig. 6.3). The MTF can be
derived from the Fourier transform of the PSF, i.e., |MTF| ¼ FT{PSF(x,y)},
the PSF being calculated as [2]

2 =2p ;
PSFðx; yÞ ¼ (6:1)
 3
1 þ d2 fx2 þy2 g 2

Fig. 6.3 The point spread function of an imaging system can be derived from the first derivative
of a ‘‘real’’ (blurred) edge (left). The MTF is derived from the Fourier transform of the PSF as
jMTFj ¼ FTfPSFg (right)
6 Neutron Tomography 85

where  ¼ 1/dconv, dconv being the distance of the converter foil to the image
plane. Usually dconv can be taken as the thickness of the converter if the film is
attached to it or the CCD camera is focused on it.
So the measured image I(x,y) is the convolution of the 2D PSF with the 2D
projection of the object function OP(x,y) (the 2D Radon transform of the 3D
object), plus background (bg), normalized to the pixel-efficiency "pixel (x,y) of
the detector:

Iðx; yÞ ¼ fPSFðx; yÞ  OP ðx; yÞ þ bgðx; yÞg  "pixel ðx; yÞ : (6:2)

Because of different efficiencies of the detector pixel and background, these

two parameters must be determined very accurately to achieve definite images
or reconstructions. They are derived from flat field and background images, the
first obtained without the sample in the beam and the second when the beam
shutter is closed.
The physics underlying the detection of neutrons using CCD cameras in
conjunction with converter screens is quite sophisticated. Neutrons must be
detected and finally converted into visible light (which can also be done by
imaging plates, converter foils, or semiconductor devices). This process is based
on the absorption of a neutron by a Li6 atom, which then decays into H3 and H4
atoms. These react with ZnS-Ag particles and the reaction is accompanied by
the emission of visible light. This light must be focused onto the chip of the CCD
camera by a mirror system in the neutron beam (to avoid radiation damage —
Fig. 6.4) and processed for data acquisition.
The CCD is connected to a conventional PC that steers the experiment and
collects the data to be reconstructed (Section 6.3). The 2D image can be
processed and stacked into a 3D image so that details can be extracted and
visualized as 2D or 3D images. The use of modern CCD cameras as 2D neutron

Fig. 6.4 Left: Neutrons hitting the ZnS screen are converted into visible light that is collected
on the CCD of a camera. Right: Light-proofed box containing mirror and CCD optics and
camera. For details see, for example, Lehmann et al. [3]
86 W. Treimer

detectors has nearly superseded the film-converter technique, because all the
image-processing techniques known from medical imaging can also be applied
to neutron radiography and tomography [3].
The number of applications of neutron tomography is quite large, and there
are several papers and books on this topic. However, some new developments
are significant. The experimental conditions have improved, and there is an
increased need for applications of neutron tomography related to industry and
cultural heritage. The new developments are energy-dependent radiography,
(rapid) real-time radiography, refraction or phase contrast radiography and
tomography, small-angle scattering radiography and tomography, and radio-
graphy and tomography using polarized neutrons (the latter is covered in
Chapter 10). However, these are currently not routine methods, and their
potential must be further evaluated.

6.3 Mathematical Foundations of (Neutron) Tomography

6.3.1 Scanning

The methodological foundations underlying the reconstruction of 2D and 3D

images of objects have been described in detail by Kak and Slaney [4] and
Rosenfeld and Kak [5]. Other important text books, for example by G. Herman
[6], describe different methods such as the algebraic reconstruction technique
(ART). This chapter follows the method of Kak and Slaney and explains the
essential steps for tomography from the measurement to the reconstruction of a
single slice. Stacking of such slices yields a 3D image. By means of image
processing, parts of the 3D reconstruction can then be extracted and displayed
separately from the rest of the data set.
An arbitrary object O(x,y,z) is considered to comprise n slices, of equal
thickness
z, that all lie in planes parallel to the (x,y) plane and perpendicular
to the z-axis (Fig. 6.5). Each slice represents a cut through the object O, which

Fig. 6.5 Scanning geometry of an object (I ) in the {x-y-z} coordinate system. A slice in the
(x–y) plane is scanned through angle  (II) and the transmitted intensity is registered in the t-s
rotated coordinate system (II, III). For details, see text
6 Neutron Tomography 87

has to be reconstructed. It is considered as a 2D function fn(x,y) that describes,

for example, the position-dependent attenuation coefficient mn(x,y), or any
other 2D function that can be measured and the signal described as a line
integral. Any function f (x,y) can be taken into account for tomography as
long as it is bounded and finite in a given region and zero outside this region.
(Note that the object can be larger than the scanning region, but the reconstruc-
tion of the outer part is not unique). The condition of finite size is easy to realize
for solid or liquid samples or gas components in a box, but it is not so easy to
handle tomography of electric and magnetic fields. The problem in generalizing
this function for tomographic measurements is in determining the interaction
that a (pencil) beam (e.g., neutron, X-ray) experiences as it passes through the
object. The goal of tomography is to reconstruct this 2D function (i.e., slice)
from projections measured in a unique way.
The slice is scanned through angles  with 0 <  < 1808 (3608 in the case
of refraction tomography), and the transmitted intensity (e.g., of neutrons)
is registered as a function of the translation, i.e., the position parameter t
(Fig. 6.6). The transmitted intensity is given by Lambert’s law as
0 1
Z
B C
Iðx; yÞ ¼ I0 exp@ ðx; yÞ  dsA ; (6:3)
path

where I0 is the intensity of the incident neutron beam and f ¼ m(x,y) is the 2D
function to be reconstructed. A new rectangular coordinate system (t,s) is

In(
I0 /
I)

θ
t

Pa
ral
lel S
sca
no
f th
es t = x.cos(θ) + y.sin(θ)
ou
rce

Fig. 6.6 Scanning of a slice in the x–y plane (compare with Fig. 6.5). Note that z is the axis of
rotation, which coincides with the z-axis of the {s-t-z} system
88 W. Treimer

defined that expresses the rotated detector system (t,s) with respect to the fixed
sample system (x,y) (or vice versa, if the sample is rotated as is done in neutron
tomography). Transforming the (x,y) system into the (t,s) system, t can be
expressed as

t ¼ x  cosðÞ þ y  sinðÞ : (6:4)

The path across the sample, given in terms of t and , by eliminating s, is

ðt  x cos  þ y sin Þ; i.e., the  function guarantees that only points fulfilling
Eq. (6.4) that lie on the pencil beam will contribute to the projection P(t).
One can define a so-called projection P (t) as
  Z
I0
P ðtÞ ¼ ln ¼ ðx; yÞ  ds : (6:5)
I
path

The set of all projections P (t) of m(x,y) is called the Radon transform of
m(x,y). From these projections, a 2D image can be reconstructed using the
‘‘Fourier slice theorem.’’ This states that the Fourier transform of a projection
P(t) of m(x,y) is a subset of a function M(u,v), which is the Fourier transform of
m(x,y), i.e.,

FTfP ðtÞg ¼ S ðwÞ  FTfðx; yÞg ¼ Mðu; vÞ ; (6:6)

where u ¼ w cos () and v ¼ w sin(). To demonstrate this relation, we again

follow the method given in [4] and [5]. Assume m(x,y) to be bounded and finite,
so that m(x,y) has a Fourier transform M(u,v):

Z1 Z1
Mðu; vÞ ¼ ðx; yÞ  e2piðuxþyvÞ  dxdy: (6:7)
1 1

To prove the slice theorem, consider the simple case of  ¼ 0. In this case, the
two coordinate systems (x,y) and (u,v) coincide, so the projection P(t) is simply
the integral (Eq. [6.10]):

Z1
P¼0 ðtÞ ¼ ðx; yÞ  dy : (6:8)
1

The Fourier transform of the object becomes (as a result of  ¼ 0)

6 Neutron Tomography 89

Z1 Z1
Mðu; 0Þ ¼ ðx; yÞ  e2piðuxÞ  dxdy
1 1
8 9 (6:9)
Z1 < Z1 =
¼ ðx; yÞ  dy e2piðuxÞ  dx :
: ;
1 1

Now, one can substitute into Eq. (6.7) and get

Z1
Mðu; 0Þ ¼ P¼0 ðxÞ  e2piux  dx : (6:10)
1

If the Fourier transform of P(t) is given by

Z1
S ðwÞ ¼ P ðtÞ  e2piwt  dt ; (6:11)
1

then S¼0(w) is equal to M(u,0). Because the orientation of the object in the (u,v)
system is arbitrary with respect to the (x,y) system, this relation allows it to be
used for all angles , keeping in mind that the Fourier transform conserves the
rotation made in real space. The Fourier transform of P(t) yields the values of
the Fourier transform of m(x,y). A dense set of values in the Fourier space (given
in polar coordinates because of the polar scanning of the object) can be
approximated to rectangular coordinates and back-transformed into real
space. This may sometimes lead to diffuse reconstructions as a result of incom-
plete fitting for high frequencies in the Fourier domain (Fig. 6.7).

V V

u u

Fig. 6.7 Plot of Fourier-transformed projection P (left) and interpolation of the data onto a
rectangular coordinate system (right). The interpolation for higher frequencies (large (u,v)
values) becomes imprecise
90 W. Treimer

How many projections yield the best result? This is answered by the Shannon
theorem, which states that a unique reconstruction from a sampled object is
obtained if the object is sampled with a frequency twice as high as the highest
frequency in the Fourier-transformed image. For parallel mode scanning, using
N steps for one projection, it can easily be derived that M projections (angles )
are required to fulfill the Shannon theorem for tomography. If D is the diameter
of the object to be scanned, and
x is the difference between two scanning
points, then the number of scanning points per projection is
D
N¼ ; (6:12)

x
fulfilling the Shannon condition. The rotation of the sample around 1808 means
a path length y for a point sitting on the circle with the radius D/2. We require
that
x ¼
y, so that the number of projections M must be

D
2 p Dp
M¼ ¼ : (6:13)

y 2 
y

Using
x ¼
y results in a relation between the number of scanning points

of a projection and the number of angles M

p
M N : (6:14)
2

N and M determine the quality of the reconstruction. If M does not fulfill the
inequality, i.e., if fewer than M projections are registered, then the reduced D ¼
D* that fulfills the Shannon condition can be calculated for a smaller number of
projections.

6.3.2 Image Reconstruction

As mentioned in the previous section, the back-transform (inverse Fourier

transform) of the Fourier transform S(w) can yield a 2D reconstruction of
the slice m ¼ m(x,y). However, a more efficient (and elegant) method is the
filtered back-projection (FBP), which is described in detail in [4] and [5]. The
basic idea is embedded in the scanning mode: the sample (slice) is scanned from
08 to 1808, which involves P(t) ¼ P+180(–t) and suggests the use of a polar
coordinate system rather than a rectangular one. The projections are registered
in polar coordinates, so that the Fourier-transformed projections of m(x,y), the
S(w) data, are discrete polar function values. So it seems to be appropriate to
write m(x,y) as a Fourier transform representation in polar coordinates. The
function m(x,y) written as the Fourier representation with M(u,v) in rectangular
coordinates is
6 Neutron Tomography 91

Z1 Z1
ðx; yÞ ¼ Mðu; vÞ  e2piðuxþvyÞ dxdy : (6:15)
1 1

The same representation in polar coordinates is

Z2p Z1
ðx; yÞ ¼ Mðw; Þ  eþ2piwðx cosðÞþy sinðÞÞ w  dw  d : (6:16)
0 1

The transformation of the (x,y) system into the (w,) system uses the relations

u ¼ w cosðÞ
v ¼ w sinðÞ (6:17)
du  dv ¼ w  dw  d :

Substituting these formulae in the m(x,y) Fourier representation yields (with

the slice theorem)

Zp Z1
 
ðx; yÞ ¼ Mðw; Þ  eþ2piwt jwj  dw d
0 1
2 3 (6:18)
Zp Z1
¼ 4 S ðwÞ  eþ2piwt jwj  dw5d :
0 1

The integral given in the square brackets can be considered as the Fourier
transform of P(t). However, S(w) is multiplied by the |w| function, which
is a special filtering function of the frequencies in the frequency space. We
define

Z1
Q ðtÞ ¼ S ðwÞ  jwj  eþ2piwt  dw : (6:19)
1

With t ¼ x.cos() + y.sin(), m(x,y) then becomes simply

Zp
ðx; yÞ ¼ Q ðx cosðÞ þ y sinðÞÞ d : (6:20)
0

A product in Fourier space, when inverse Fourier-transformed, corresponds

to the convolution of the inverse Fourier-transformed functions, i.e., we use the
relation
92 W. Treimer

FT1 fS ðwÞ  jwjg ¼ P ðtÞ  FT1 fjwjg : (6:21)

The convolution operation is denoted by the  symbol. The function |w| is

not a square integrable function, so it has no inverse Fourier transform. How-
ever, FT1 fjwjg can be approximated by several filter response functions (con-
volution kernels, filter functions). The reconstruction of m(x,y) from projections
P (t) now works in the following way.  ! Each P(t) is convoluted with a proper
filter function (e.g., Shepp-Logan, 2p sin cð!Þ) and the resulting values are
‘‘smeared’’ over the (x,y) planes as demonstrated in Fig. 6.8.
To control the tomographic measurement, the recorded projections are
arranged in a ‘‘sinogram,’’ as shown in Fig. 6.9. Each vertical line represents
one projection at  and is plotted in gray levels as a function of the pixel number.
Figure 6.10 shows four different reconstructions. The projections P(t) are
converted into gray levels, convoluted with a filter kernel, and back-projected
over the total (x,y) plane. Adding up all measured projections yields the desired
2D reconstruction of the slice. The more the projections overlap, the more the
star-artifacts are suppressed. With such reconstructions, 3D images can be
made by stacking up all the reconstructed slices and 3D volume data details
(e.g., having different attenuation values) can be extracted from the rest of the
body (Fig. 6.11).
A number of striking applications of neutron absorption tomography can be
found in [3] and [7].

Fig. 6.8 Scanning of a simple 2D object (square and circle) and corresponding projections
(figure courtesy of T. Donath, GKSS)
6 Neutron Tomography 93

Fig. 6.9 The sinogram of Fig. 6.8. The y-axis shows the projection as gray values of Fig. 6.8

Fig. 6.10 Four 2D projections to illustrate the filtered back-projection procedure

Fig. 6.11 Three-

dimensional reconstructions
of parts of a fuel injector
device (height 115 mm, base
45 by 60 mm, ring outer
diameter 35 mm). Metallic
components (here not
specified) can be extracted
or suppressed by image
processing
94 W. Treimer

6.4 Experimental Techniques and Results

This section summarizes some techniques that are becoming increasingly rele-
vant in neutron tomography and that will be used for a range of applications in
the future.

6.4.1 Energy-Dispersive and Bragg-Edge Radiography

and Tomography

The transmission (attenuation) of neutrons through matter depends on the

composition of the material itself, the particular absorption (and scattering)
cross sections, the wavelength (energy) of the neutrons used, temperature,
isotopic composition and, also quite important, the crystalline structure and
its orientation in the material. Materials consisting of polycrystalline com-
ponents attenuate neutrons in a unique way because the thermal and cold
neutron wavelength spectrum (l ¼ 0.05 nm to 1 nm) spans the typical
crystal lattice constants. Neutrons having wavelengths of the order of the
lattice constants d can fulfill the well-known Bragg condition and be
scattered away from the incident direction.
If certain regions or parts of a material contain aluminum, for example, then
the dependence of the attenuation can be observed from the wavelength that
affects a radiograph, as shown in Fig. 6.12a and b (see also Fig. 1.4). This
behavior is caused by the superposition of scattering interactions with the
sample crystallites with a well-defined lattice spacing, but in random orienta-
tion. The integrated reflecting power increases with the Bragg angle, so that
 ¼ 908 and l¼ 2d result in sharp jumps in the attenuation spectrum. This

(a) (b)
Fig. 6.12 (a) Attenuation of an element (polycrystalline) as a function of neutron wavelength
(b) attenuation coefficient for aluminum in the range 2–6 Å [8]
6 Neutron Tomography 95

behavior can be enhanced if well-oriented crystalline structures (textures) are

present. The spectrum in Fig. 6.12 was measured with the CONRAD setup at
the Helmholtz Center Berlin (HZB) (formerly the Hahn-Meitner Institut
[HMI]), using a graphite double monochromator [8].
A similar behavior of the attenuation cross section is found for Fe, Cu, brass,
and other crystals having lattice spacings of the order of the neutron wave-
length. In the close neighborhood of a so-called Bragg edge, the attenuation
(scattering away from the incident beam direction) rises as a result of the
increasing integrated reflectivity [9] and then falls very quickly because the
Bragg condition for a certain set of atomic planes is no longer fulfilled. This
can be used to detect regions of strong deformations, as shown in Fig. 6.13.
Here a steel plate was deformed and radiographs were taken at wavelengths in
the range of 2.2 to 6.4 Å in steps of 0.1 Å. A 2D mapping of the Bragg edge for
each pixel in the deformed plate is shown in the bottom image. This map can
be related to the residual stress distribution in the plate.
This dependence of the attenuation upon the neutron wavelength enables
sophisticated radiographies. If a material has the same linear attenuation coeffi-
cient S for two different wavelengths, it can be made transparent in a normalized
image as follows. If I1(l1) and I2(l2) are the transmitted intensities at l1 and l2

I1 ¼ I01  eSðl1 Þd ; I2 ¼ I02  eSðl2 Þd ; (6:22)

then the ratio of the normalized intensities is

I1
I01 I1  I02 eSðl1 Þd
I2
¼ ¼ ¼ e½Sðl2 ÞSðl1 Þ d : (6:23)
I02
I2  I01 eSðl2 Þd

In the case of S(l1)  S(l2), for example, as shown in Fig. 6.12 (a), the
exponential function becomes nearly unity and there will be no contribution of
this material to the normalized image; i.e., this material becomes transparent in this
image.

Original sample

Radiograph with
2.62 Å
Fig. 6.13 Radiographs of a
bent metal plate (50  10 
5 mm, top picture) Radiograph with
performed at two different 3.55 Å
wavelengths (2.62 and
3.55 Å). For details see [8]
96 W. Treimer

Such measurements have been performed, for example, using CONRAD, a

radiography and tomography instrument for cold neutrons at the BER II
reactor (HZB). To be able to make radiographs at different wavelengths, a
double monochromator system of two graphite crystals was used so that the
wavelength could be tuned within the range of 2.8 to 6.2 Å [8]. It must be taken
into account that the incident intensities I01 and I02, corresponding to different
wave lengths l1 and l2, are different owing to the wavelength spectrum. A set of
step wedges of known composition and thickness (5, 10, 15, and 20 mm) was
imaged at different neutron energies and the radiographs properly normalized
to each other. The different behaviors of the attenuation coefficients are plotted
for 3.14 and 6.40 Å and the transmitted intensities compared with each other as
shown in Fig. 6.14.
This wavelength dependence of the neutron attenuation enables other sophis-
ticated radiographies, similar to the examples shown in Fig. 6.15, to enhance or

(a) (b) (c)

Fig. 6.14 Different wedge transmissions for two wavelengths, (a) 3.14 Å and (b) 6.40 Å, for
aluminum brass, steel, copper, and polyethylene (PE). (c) Step wedges: 5–20 mm thickness. To
visualize the effect of normalization, the transmitted intensities have been converted into gray
values (compare with Fig. 6.15)

(a) (b)
Fig. 6.15 Ratios of properly normalized radiographs of polyethylene and metal wedges for
different wavelengths. Some materials become transparent for particular wavelength ratios,
for example, copper in (a) and steel in (b) [8, 9]
6 Neutron Tomography 97

suppress components in a sample under investigation. This technique of normal-

ized intensities yields an exponential expression, which can significantly increase
or decrease parts in an image of a sample and so improve the interpretation of
images.
Some other examples of energy-dispersive radiography are shown in
Chapter 12.

6.4.2 Real-Time Radiography

A number of technical processes must or should be directly controlled on-line.

Engineers and scientists want to see, for example, how oil flows in machines and
how water flows in fuel cells (Chapter 11) or is taken up in plants (Chapter 17)
and so on, because from these observations they can modify their theories or
designs in order to optimize machines, to save energy, and to understand
natural processes. The process to be visualized determines the method of
collecting data, because one must distinguish between repetitive and simple
emerging nonrepetitive processes. In the case of nonrepetitive processes, each
frame must image (by absorption) an instantaneous state of the process within a
certain time interval
t. The process within a time interval
t itself can be
visualized by registering as many frames as possible. In the case of an oscillating
piston of a model aircraft engine, the motion was triggered so that different
positions of the piston could be imaged at different times [8, 9].
The sharpness of the images depends on the number of neutrons that can be
collected within one pixel (
x) within the time interval
t (i.e., per frame) and
how well defined the triggering is. It can easily be understood that the exposure
time
t must be small compared with the velocity of the process for a given
x;
i.e., any detail of the moving object being imaged should move less than half
a pixel within
t [11, 12]. See Chapter 12 for examples of what is currently
achievable.

6.4.3 Phase Contrast, Refraction, and Small-Angle Tomography

In considering neutrons not as particles but as de Broglie waves, all properties

that are connected with the nature of these waves must be taken into account.
The most important feature of this representation is the coherence property of
the waves. One can adapt from optics the description of Fresnel and calculate
the coherent area of a source emitting coherent (neutron) waves. This area is
given by the first Fresnel zone and its size AFZ is

Rr 1
AFZ ¼ pl : (6:24)
Rþr sinðÞ
98 W. Treimer

Here R is the distance from the point source to the object, r is the distance
between object and detector, and  is the angle between the ray propagation
vector and the plane of observation (usually  ¼ 908). (Compare with draw-
ings in classical optic books). The size AFZ determines the coherent source
for any interference effects to be observed. For standard geometries in
neutron tomography (and radiography), R  500 cm, r  1 cm, and l 
0.5 nm, so AFZ is  1.6  10–7 cm2 and the diameter of the coherent source
can roughly be estimated as 4 mm. To use the coherent properties of such a
beam, a point-like source must be created by a small pinhole, which deter-
mines—together with the divergence of the beam—the coherent area of the
real neutron beam. The diameter dcoh characterizing the coherent size of a
source can be estimated by simple geometric considerations as given by [13].

1L
dcoh ¼ l   (6:25)
2D

To achieve a reasonable dcoh, the source (i.e., D) must be kept small and
L must be large (meters).
To understand the phase contrast, consider a (plane) wave traveling through
an object that could be smaller or larger than the coherent wave front of the
wave (Fig. 6.16).
Phase contrast originates from a coherent superposition of diffracted
waves; the resulting intensity variation can be observed only in the coherent
part of the wave that propagates through the object and experiences position-
dependent phase shifts caused by different indices of refraction in the sample.
This modulates the wave front, and a change of propagation direction will be
observed. The strongest phase contrast is observed in the near field region, z <
dcoh2/l, where the modulation of the amplitude is best; it broadens with
increasing z [13]. (z is the distance from object to detector and dcoh  dimen-
sion of the coherence width.) Fringes can be registered that are caused, for
example, by an edge that would be washed out if the beam divergence and/or

(a) (b)
Fig. 6.16 An object is illuminated by a coherent wave front, which can be (a) larger or (b) smaller
than the object
6 Neutron Tomography 99

source size were too large. It can be calculated that the optimal contrast occurs
at z ¼ dcoh2/2l [13]. One approach for realizing phase contrast was performed
by neutron radiography some years ago [14]. The first experiments based on
phase effects (refraction) were performed in 1989 using a double crystal
diffractometer [15]. The first neutron radiography based on phase contrast
was performed in 2000 at the National Institute for Standards and Technol-
ogy (USA) using a pinhole and a wavelength of 4.43 Å and with the sample
placed l.8 m from the source [16, 17]. However, phase contrast imaging is
much more often used with synchrotron radiation because of the higher
brilliance of the source.
One can calculate the phase contrast, determining the Fourier-transform of
the structure or of the phase boundary of two structures. This can be done by
ray tracing, assuming that a certain width of the edge is coherently illuminated.
Here we assume that an incident plane wave hits a phase step (boundary of
two different materials with a difference of 10 mm in thicknesses). However, in
general, one can also use ray-tracing, adding up all amplitudes emerging from
the structure that cause the phase gradient that contributes to a measuring point
showing an intensity distribution as plotted in Fig. 6.17, top.

Si: 60µm - Al: 50µm

5

1
500 550 600 650 700

Fig. 6.17 Calculated phase contrast (top; x-scale in mm, y-scale in arbitrary units) for a phase
step consisting of silicon and aluminum (bottom) [12]
100 W. Treimer

The phase development quite close behind the sample yields strong inten-
sity oscillations that can be measured at different distances from the object
[18]. Very often, another concept is used to describe this phase contrast
correctly via geometric considerations of refraction by edges [19]. The latest
developments use a phase grating-based shearing interferometer; expressed
simply, this senses the phase-modulated (refracted) beam behind the sample
using an analyzing grating. The measured intensity depends on the differ-
ence in phase shift introduced by two adjacent regions in the sample that are
coherently illuminated. Several images must be measured with varying phase
grating positions and the phase change due to the object then calculated.
Additional details concerning this technique can be found in Chapter 8 and
in [20, 21].

6.4.4 Refraction and Small-Angle Scattering Tomography

As was shown earlier, the size of the coherence area of the neutron (X-ray) wave
with respect to the object determines the kind of interaction—refraction or
small-angle scattering. If the coherent width (lateral coherence length) is smaller
than the object, then refraction occurs (Fig. 6.16). If the object is smaller than
the coherent width, small-angle scattering occurs. To study this behavior, we
consider a function n(x,y) that describes the position-dependent index of refrac-
tion n(x,y) 6¼ 1 inside the sample and n(x,y) ¼ 1 outside it. The Snellius law
states that n1sin("1) ¼ n2sin("2), "1 being the angle of incidence and "2 the angle
of refraction; n1 and n2 are the indices of refraction of medium 1 and medium 2.
From geometrical optics it is also well known that the larger the "1 is, then
the larger "2, up to the case of total reflection. Neutrons leaving the sample will
have a final angle of deflection , which is the sum of all individual deflections
 ( ¼ " – "+1) along the path through the sample (Fig. 6.18):

X
n
¼ i : (6:26)
i¼1

A deflection  from the preceding direction is present for adjacent volumes

with different indices of refraction and if k~ rnðx; yÞ (k~ is the neutron wave
vector). One can determine {rn(x,y)}, and from that n(x,y), by measuring the 
of each path (line integral) and varying the orientation  of the sample from 0 to
3608. If k~ ¼ ðkk ; k? Þ is represented in terms of components parallel and perpen-
dicular tornðx; yÞ, and rnðx; yÞis the component of rnðx; yÞ perpendicular to
k? , then a point P(t) of a projection is given by [22]
Z
P ðtÞ ¼ P ðt; Þ ¼ ?  ds :
r n ðx; yÞ  k (6:27)
path
6 Neutron Tomography 101

Fig. 6.18 Neutrons are refracted several times in the sample (e.g., by inhomogeneities) and
exit with a deviation angle 

The change of the phase
’(x,y) (here
is not the Laplacian operator) can
be expressed as the sum of all deviations that the ray experiences:
Z

’ðx; yÞ ¼ k  ðx; yÞ  ds : (6:28)
path

The data corresponding to the phase changes are measured as refraction

angles  ¼ ðt; Þ, which represent the orientation of the object in the (x,y)
system.  is large (i.e., can be measured) if the incident ray is nearly parallel to
the interface that separates two regions with different indices of refraction; so 
can be written as

1
ð; tÞ ¼ r? ð
jðx; yÞÞ : (6:29)
k

So P(t) becomes
Z t
P ðtÞ ¼ ðx; yðx; ; t0 ÞÞ  dt0
0
Z 1 Z 1
(6:30)
¼ Dirac ðx cosðÞ þ y sinðÞ  tÞ  dx  dy :
1 1

To measure the (resulting) refraction angle , the incident direction must be

known;  and P(t) can be determined from the experiment. The total set of
functions P(t) is then another form of the Radon transform of n(x,y), so the
102 W. Treimer

reconstruction of n(x,y) from projections can follow well-known mathemati-

cal procedures. An estimation of the total angle of deflection is given by the
calculation of the index of refraction n, which is expressed for neutrons by

l2 Nbc
n¼1  (6:31)
2p

In this expression, l  5  10–10 [m], N ¼ number of unit cells m–3  5  1028,

and bc ¼ coherent scattering length  10–15 [m] of the sample; i.e., 1 – n  10–5 to
10–6  mrad gives the size of the angle of refraction that must be measured. This
can be done only with a special double crystal instrument, as described in [22].
In a case in which the extent of the object illuminated by the neutron wave is
smaller than the lateral coherence width, the whole object acts as a phase shifter
and small-angle scattering is observed (compare Figs. 6.16 and 6.19).
In the case of (ultra) small-angle neutron scattering (USANS) tomography,
m(x,y) must be replaced by a suitable function that represents the USANS
effects in the sample. One feature of small-angle scattering—single scattering
events assumed—is the individual scattering function of the particles in the
sample. From this scattering function, the size and shape of the particle can be
deduced. The measured scattering function is the convolution of the particle
shape function with the instrumental resolution function, which is the rock-
ing curve (RC) or 2D scattering pattern without the sample. The broadening
B(t) [nm–1] of a scattering function due to (ultra) small-angle scattering
(t again means the scanning parameter defined in Eq. [6.4]) can be (roughly)
approximated by a Gaussian distribution and, involving multiple scattering,
written as [24, 25]
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u Z ðx; yÞ  Nðx; yÞ
u
B ðtÞ ¼ t  ds ; (6:32)
Rðx; yÞ2
path

where  is the scattering cross section, N(x,y) is the particle density distribution,
and R is a parameter with the dimension of a length specifying an average size or
correlation length in the scattering object (corresponding to the Gaussian
approximation). Now one can use B or B2 as an imaging signal to use (ultra)

Fig. 6.19 The coherent

incident wave front is larger
than the object. The
scattered wave contains all
phase changes due to the
object and therefore yields
its 2D Fourier transform
6 Neutron Tomography 103

small-angle scattering for image reconstruction. A function fB(x,y) can be

defined as
ðx; yÞ  Nðx; yÞ
fB ðx; yÞ ¼ ; (6:33)
R2 ðx; yÞ
so that a projection P (t) can be rewritten as P ðtÞ ¼ B2 ð; tÞ
Z
2
P ðtÞ ¼ B ð; tÞ ¼ fB ðx; yðx; ; tÞÞ  ds : (6:34)
path

With this proper description of P(t) ¼ B2(,t), the standard methods of

reconstruction such as the FBP can be used. Note that, in both cases, the wave
properties of a neutron are responsible for the phase change due to the inner
sample structure; i.e., the individual phase shifts in the sample add coherently to,
and result in a final phase change of, the wave behind the sample. So, in both
cases, we can call this phase contrast because the fundamental interaction is the
change of the phase front due to structures in the sample. In the case of refraction,
the wave is refracted many times before it leaves the object and has a different
direction from the incident direction in front of the sample. The change of the
wave front (mean direction) of the beam can be measured for each path of the
neutron through the sample, forming (for a given ) a projection P(t) that
indicates for each t a certain angular shift of the intensity distribution.
The same occurs in the case of small-angle scattering. Here the neutron wave
is scattered behind the sample as a result of phase shifts in the sample. This
causes a deformation of the wave front behind the sample (Figs. 6.16 and 6.19)
and is measured as a broadening B(t) of the intensity distribution I(x,y).
Usually one must deal with three different effects—absorption, refraction,
and small-angle scattering—because all of them will occur simultaneously in a
sample. The signals can be represented as follows:
1. Refraction shifts the total RC without a decrease in the total (integrated)
intensity and leaves the shape unchanged (Fig. 6.20a).
2. Small-angle scattering also has no influence on the total scattered intensity;
however, the shape of the RC is changed, the peak intensity is decreased, and
the wings of the curve are enhanced (Fig. 6.20b).

(a) (b) (c)

Fig. 6.20 Rocking curves due to refraction (a), small-angle scattering (b), absorption (c). A is
the rotation angle of the analyzer crystal
104 W. Treimer

3. As a result of absorption caused by the sample, the total intensity is decreased,

but the shape of the RC remains nearly unchanged (Fig. 6.20c).
Note that the angular shift due to refraction is on the order of mrad, the
broadening of the order of 10–3 to 10–5 nm–1, so both signals are rather difficult
to measure with standard tomography setups. Moreover, the fundamental
problem with such imaging signals (refraction and small-angle scattering) is
to separate absorption, refraction, and small-angle scattering clearly from one
another. In a real experiment, all three interactions may happen simultaneously,
so a single measurement should register all of them at the same time. In principle,
one can assess them separately; however, the best way is to determine all three
signals simultaneously as is done with a special double crystal diffractometer,
shown in Fig. 6.21.
The sample under investigation is placed between the silicon monochroma-
tor and the silicon analyzer crystal, as shown in Fig. 6.21. Both are perfect
crystals; the analyzer is bent so that the reflecting net planes continuously
change their orientation with respect to the incident beam, which enters perpen-
dicular to the front face of the analyzer crystal (Fig. 6.21). Neutrons traveling
parallel to the surface are reflected by those net planes that fulfill the
Bragg condition, i.e., entering within the so-called Darwin (angular) width
( 10–60 mrad). The strong angular correlation between both crystals guarantees
an angular sensitivity of order mrad or, in terms of momentum transfer q units,
some 10–4 nm–1. Any deviation of the incident beam direction is measured with a
1D or 2D position-sensitive detector. Each detector channel corresponds to a

certain deviation angle or momentum transfer q ¼ 4p k  sinðÞ, where k=2p/l
and  ¼ half scattering angle, defining the peak position of the RC at q ¼ 0 and
as well for  ¼ 0 (cp Fig. 6.20).
The sample is scanned by steps perpendicular to the neutron beam. For each
step, the total RC, and so the shift (center of gravity of the curve), is measured.
The worst cases yield no absorption contrast, and only refraction contrast
yields a good reconstruction, as demonstrated in Fig. 6.22.
With this technique a 3D-refraction tomography could also be realized,
stacking reconstructed slices upon each other (Fig. 6.23) [26].

Fig. 6.21 The double crystal

diffractometer V12a at
the HZB
6 Neutron Tomography 105

Fig. 6.22 Reconstructions

of (a) the gradient and
(b) the integrated gradient of
the index of refraction
(sample aluminum, diameter
6 mm, holes 2 mm,
60 projections)

(a) (b)

Fig. 6.23 Three-dimensional reconstruction based on pure phase (refraction) contrast

(aluminum sample, diameter 15 mm, central hole 1 mm, oblique hole 3 mm)
106 W. Treimer

The most astonishing result was the detection of clusters of particles (ß-
carotene) having a size of approximately 150 nm, dissolved in D2O ( 5–12%).
In this experiment all effects—refraction, USANS, and absorption—were mea-
sured simultaneously with the double crystal diffractometer. The shift of the RC
yields the final deviation and determines the refraction, the broadening of the
small-angle scattering, and the decrease of the integral intensity of the absorp-
tion of the neutron beam. All these data are derived from a single curve.
Combining USANS and refraction data yields a complete reconstruction.
With pure attenuation, and only incomplete refraction or USANS data, imper-
fect images are reconstructed [24, 27].
Chapter 16 describes the use of modeling to reconstruct the shapes of biolo-
gical macromolecules with a typical size of <100 nm based on small-angle
scattering data. In combination with the methods described here, this offers the
possibility of covering the full range of length scales from macroscopic to nano-
scopic with neutron imaging methods.

6.5 Outlook

There are several other fields of application of neutron tomography that should
be mentioned here. In stress and strain investigations (Fig. 6.13), the investiga-
tion time can be drastically reduced by 3D scanning of a sample using Bragg-
edge and energy-resolved techniques (Chapter 12). There seems to be a great
interest in investigating quasi-in-situ water distribution in operating fuel cell
stacks by neutron tomography (Chapter 11 and e.g., [30]), and this also
holds true for applications in the fields of moisture detection and geological
sampling (Chapter 17) and noninvasive investigation of cultural heritage obj-
ects (Chapter 13 and, e.g., [3]). Another interesting topic will be the use of
polarized neutrons (Chapter 10). The imaging signal is the change of the
spin direction with respect to the orientation in front of the sample; i.e., the
2D-depolarization image of a neutron beam will be measured. The neutron
depolarization technique is well established; the problem is to calculate a unique
reconstruction, for example, of two-dimensionally (three-dimensionally) ran-
domly arranged small magnetic fields in samples, such as complicated magnetic
flux distributions in solids, magnetic domains, etc. Several attempts to solve this
problem have been published [31]; however, a reconstruction technique applied
to refraction or small-angle tomography similar to that discussed in [22] or
iterative techniques seem to be most challenging and promising.

References
1. S.G. Glasstone, M.C. Edlund, Kernreaktortheorie, Wien, Springer-Verlag, p. 30 (1961).
2. A.A. Harms, D.R. Wyman, Mathematics and Physics of Neutron Radiography, D. Reidel
Publishing Company, Dordrecht, Holland (1986).
6 Neutron Tomography 107

3. E. Lehmann, P. Vontobel, R. Hassanein, Neutron Tomography as Tool for Applied

Research and Technical Inspection, Adv. Solid St. Phys. 45, 389–405 (2005).
4. A.C. Kak, M. Slaney, Principles of Computerized Tomographic Imaging, IEEE Ed., N.Y.
(1999).
5. A. Rosenfeld and A.C. Kak, Digital Picture Processing, Computer Science and Applied
Mathematics, Academic Press Inc. (1982).
6. G.T. Herman, Image Reconstruction from Projections: The Fundamentals of Computerized
Tomography, Academic Press, New York (1980).
7. E.H. Lehmann, P. Vontobel, G. Frei, The non-destructive study of museums objects by
means of neutrons imaging methods and results of investigations, Il Nuovo Cimento, 30 C,
N. 1, 93–104 (2007).
8. W. Treimer, A. Hilger, M. Strobl, N. Kardjilov, I. Manke, a wavelength tunable device
for neutron radiography and tomography, Appl. Phys. Lett. 89, 203504 (2006).
9. R.W. James, The optical principles of the diffraction of X-rays, The Crystalline State –
Vol. II, ed. Sir L. Bragg, G. Bells and Sons Ltd. 1967.
10. N. Kardjilov, A. Hilger, I. Manke, M. Strobl, W. Treimer, J. Banhart, Industrial applica-
tions at the new cold neutron radiography and tomography facility of the HMI, Nucl.
Instrum. Methods Phys. Res. A 542, 16–21 (2005).
11. W. Treimer, A. Hilger, N. Kardjilov, M. Strobl, Review about old and new imaging
signals for neutron computerized tomography, Nucl. Instrum. Methods Phys. Res. A 542,
367–375 (2005).
12. W. Treimer, N. Kardjilov, U. Feye-Treimer, A. Hilger, I. Manke, M. Strobl Absorption-
and phase-based imaging signals for neutron tomography, In Bernhard Kramer (Ed.)
Adv. Solid St. Phys. 45, 407–420, ISSN 1438-4329 (2005).
13. X. Wu, H. Liu, Clinical implementation of x-ray phase-contrast imaging: Theoretical
foundations and design considerations, Med. Phys. 30, 8, 2169 (2003).
14. A. Pogany. D. Gao, S.W. Wilkins, Contrast and resolution in imaging with a microfocus
x-ray source, Rev. Sci. Instrum. 68, 2774 (1997).
15. K.M. Podurets, V.A. Somenkov, S.Sh. Shiĺshtein; Neutron radiography with refraction
contrast, Physica B, 156, 691–693 (1989).
16. B.E. Allman, P.J. McMahon, K.A. Nugent, D. Paganin, D.L. Jacobson, M. Arif,
S.A. Werner, Phase radiography with neutrons, Nature 408, 158 (2000).
17. P.J. McMahon, B.E. Allman, D.L. Jacobson, M. Arif, S.A.Werner, K.A. Nugent, Quan-
titative phase radiography with polychromatic neutrons, Phys. Rev. Lett. 91, 14 (2003).
18. N. Kardjilov, E. Lehmann, E. Steichele, P. Vontobel, Phase contrast radiography with a
polychromatic neutron beam, Nucl. Instrum. Methods Phys. Res. A 527, 519–530 (2004).
19. M. Strobl, W. Treimer, N. Kardjilov, A. Hilger, S. Zabler, On neutron phase contrast
imaging, Nucl. Instrum. Methods Phys. Res. B 266, 181–186 (2008).
20. F. Pfeiffer, T. Weitkamp, O. Bunk, C. David, Phase radiography with neutrons, Nat.
Phys. 2, 258 (2006).
21. F. Pfeiffer, C. Grünzweig, O. Bunk, G. Frei, E. Lehmann, C. David, Neutron phase
imaging and tomography. Phys. Rev. Lett. 96, 215505 (2006).
22. W. Treimer, M. Strobl, A. Hilger, A.C. Seifert, U. Feye-Treimer, Refraction as imaging
signal for computerized neutron tomography. Appl. Phys. Lett. 83, 398 (2003).
23. M. Strobl, W. Treimer, A. Hilger, U. Feye-Treimer, Neutron tomography in double
crystal diffractometers, Physica B 350, 1–3, 155–158 15 July, (2004).
24. M. Strobl, W. Treimer, A. Hilger, Small angle scattering signals for (neutron) computer-
ized tomography, Appl. Phys. Lett. 85, 3, 488–490 (2004).
25. T.M. Sabine, W.K. Bertram, The use of multiple-scattering data to enhance small-angle
neutron scattering experiments, Acta Cryst. A 55, 500 (1999).
26. M. Strobl, W. Treimer, A. Hilger, First realisation of a three-dimensional refraction
contrast computerised neutron tomography, Nucl. Instrum. Methods Phys. Res. B
222, 653–658 (2004).
108 W. Treimer

27. W. Treimer, M. Strobl, A. Hilger, H.J. Peschke, Neutron tomography using small angle
scattering data, IEEE 52:1, 386–288 (2005).
28. N. Kardjilov, I.Manke, M. Strobl, A, Hilger, W. Treimer, M. Meissner, T. Krist,
J. Banhart, Three-dimensional imaging of magnetic fields with polarized neutrons. Nat.
Phys. 4, 399–403 (01 May 2008).
29. W. Treimer, BMBF project 03TR7TFH 2007–2010.
30. I. Manke, Ch. Hartnig, M. Grünerbel, J. Kaczerowski, W. Lehnert, N. Kardjilov,
A. Hilger, J. Banhart, W. Treimer and M. Strobl, Quasi-in situ neutron tomography on
polymer electrolyte membrane fuel cell stacks, Appl. Phys. Lett. 90, 184101 (2007).
31. E. Jericha, R. Szeywerth, H. Leeb, G. Badurek, Reconstruction techniques for tensorial
neutron tomography, Physica B 397, 1–2, 159–161 (2007).
Chapter 7
Mathematics of Neutron Imaging

K.W. Tobin, P.R. Bingham, and J. Gregor

Abstract Imaging with neutrons at macro-world scales (e.g., >10 mm) requires
particular understanding of the non-diffracting or refracting interactions that these
electrically neutral particles have with their environment. While image formation
with neutrons shares some commonality with other radiation sources such as X-rays
and gamma rays, neutrons provide complementary interaction mechanisms to these
techniques that can uniquely probe materials and structure. Neutron sources are
also historically fraught with issues regarding low source intensities, challenging
beam configurations and resolution limitations that are all closely linked. In this
chapter we will present a mathematical construct and methods compatible with the
design and characterization of radiography systems for volumetric imaging. We
begin with a review of neutron image formation, provide resolution analysis con-
cepts and methods for both the design and the characterization of radiography
systems, and conclude with a review and discussion of volumetric reconstruction
techniques using analytic or iterative computed tomography algorithms.

Keywords Neutron radiography  Computed tomography  Modeling 

Resolution  Contrast

7.1 Introduction

Neutron imaging has long been known to provide complementary, nondestructive

imaging capabilities to X-ray and gamma-ray imaging methods [3]. Today,
neutron imaging with conventional reactor-based sources enables the interro-
gation of complex, multicomponent systems for many applications, such as
nuclear material nondestructive testing [9], characterizing flight control sur-
faces on aircraft [4], testing heat transfer in porous materials [30], examining

K.W. Tobin (*)

Image Science and Machine Vision Group, Oak Ridge National Laboratory, Oak
Ridge, Tennessee 37831-6075
e-mail: [email protected]

I.S. Anderson et al. (eds.), Neutron Imaging and Applications, 109

Neutron Scattering Applications and Techniques, DOI 10.1007/978-0-387-78693-3_7,
Ó Springer ScienceþBusiness Media, LLC 2009
110 K.W. Tobin et al.

heat exchanger systems [2, 10], development of hydrogen fuel cells (Chapter 11),
inspection of cultural heritage objects (Chapter 13), and interrogating biologi-
cal systems [28]. Newer, intense sources of neutrons from spallation facilities are
providing the potential to interrogate time and energy-dependent phenomena
as well. Therefore, a mathematical representation of the neutron image forma-
tion process can provide insight into the potential application of time- and
energy-dependent, volumetric radiography to the characterization and quanti-
tation of complex material structures.
The mathematics of neutron imaging is similar to that of other non destruc-
tive methods that use non-refracting or diffracting radiation to interrogate
materials or composites of materials [3]. Figure 7.1 represents the basic elements
of a generic neutron radiography system beginning with a source of neutrons,
beam conditioning and shaping, neutron conversion and detection, and sam-
pling and signal generation.
For the purposes of this discussion, neutron image formation will be treated
as an encoding of the material characteristics of a three-dimensional (3D) object
projected onto a two-dimensional (2D) detection array, typically comprising a
converter followed by an imaging sensor as described in Chapter 5. Material
characteristics of the object are represented by the total macroscopic interaction
cross-section, T, which is a sum of the absorption and scattering cross-
sections, T = a+s (i.e., ignoring the potential for neutron capture and
fission cross-sections in some materials). The interaction of the neutron beam,
o, with the 3D object produces a total 2D field at the converter plane, T, that
can be decomposed into the uncollided flux, u, and the scattered flux, s, as
 T = u+ s. The uncollided flux arises from neutrons that arrive at the con-
verter plane unperturbed by the imaged object while the scattered flux considers
only neutrons that have been scattered by the object toward the converter plane.
To understand the dependencies of the scattered flux on the image formation
process, we introduce a buildup factor [20, 24] defined as B = T/u= 1+s/u.
Rearranging to solve for T, we can represent the total flux at the converter
plane as T = Bu, i.e., a multiplicative build up of the uncollided flux due
to object scattering. This is represented using the well-known neutron

Object Converter Imager

Source Aperture
ΣT (E,x,y,z,t)

θ
z
n
n γ
Beam conditioning/shaping φ T (E,θ,ΣT ,x,y,t) I(E,θ,ΣT ,x,y,t)
Projected flux Signal

Fig. 7.1 Basic elements of a neutron radiography-imaging environment. A source of neutron

is typically shaped and collimated, interacts with the material constituents of a 3D object and
is projected onto a 2D imaging array
7 Mathematics of Neutron Imaging 111
R
attenuation law as T= oBexp{ Tdz}, or with space-, time-, and energy-
dependencies as,
R
T ðE; ; T ; x; y; tÞ ¼ o ðE; x; y; tÞBðE; ; s ; x; y; tÞe T ðE;x;y;t;zÞdz (7:1)

where x and y are spatial position in the plane perpendicular to the ray from
source to detector, z represents position along ray from source to detector and E is
the neutron energy. In this relationship,  corresponds to the angular orientation
of the object under test (Fig. 7.1) and is included as a prelude to later discussions
regarding volumetric reconstruction from projections. Also note that the buildup
factor is a value B  1. It is functionally dependent on the material composition
and geometry presented by the object under test and is therefore difficult to
predict or accommodate under typical radiographic conditions. Buildup can be
significant for materials containing, e.g., H, Si, Ni, Cu, and other highly scattering
metals, and should be considered a source of error in quantitative analysis. For
example, imaging geometries that move the object away from the detection plane
can reduce the number of neutrons scattered into the detector, but this may also
reduce the system resolution due to the characteristics of the beam aperture. The
beam aperture and length is typically characterized by the L/D ratio, where L
denotes the distance from the aperture to the object and D is the diameter of the
aperture. This effect will be detailed further in Section 7.2.
Note that a primary goal of image analysis can be formulated as one of
measuring the interaction cross-section of the materials in the imaged object,
T(E,x,y,z). Direct measurement of T, for example, provides indirect informa-
tion regarding other material properties since  = N = NA/M, where  is
the microscopic cross-section, N is the number of atoms per cm2,  is the material
density, NA is Avagadro’s number, and M is the atomic weight [25].
The electronic signal produced by the system is further impacted beyond the
geometry effects of L/D by the response of the converter, optical components,
and sensor (e.g., a charge-coupled device, CCD). Making the practical assump-
tion of a linear shift invariant (LSI) system [14], we can describe the time-,
space-, and energy-dependent 2D electronic image, I, in terms of a convolution
of the modulated (by the object) neutron field at the converter, T, with a system
impulse response function, h, to give,
Z Z
IðE; ; T ; x; y; tÞ ¼  T ðE; ; T ; ; ; tÞhðx  ; y  Þdd
(7:2)
¼ T ðE; ; T ; x; y; tÞ  hðx; yÞ

In Section 7.2 we will model and analyze the impulse response of a radio-
graphy system to estimate system design resolution and characterize resolution
of an existing radiography system through empirical means. In Section 7.3 we
will describe considerations and methods required to perform volumetric recon-
struction and estimation of T using projection data and computed tomogra-
phy based on both analytic and iterative techniques.
112 K.W. Tobin et al.

7.2 Neutron Image Formation and Resolution Analysis

In this section, we will describe the image formation characteristics of a neutron

radiography system from both a modeling and an empirical perspective. A
model of resolution is important when designing a system to ensure that realistic
resolution goals can be established and achieved. Once a system is operational,
an empirical methodology is required to verify performance.

7.2.1 Resolution Modeling

For modeling purposes, Fig. 7.2 represents an idealized radiography system

containing the basic elements described earlier in Fig. 7.1. The goal of a resolu-
tion model is to understand the impact of various system components on signal
degradation. For this analysis we will neglect the impact of electron noise and
counting statistics, and focus instead on the effects of the neutron aperture, D,
the optical diffusion response of a scintillator, , the optical component, and
sampling at the sensor, s (e.g., by a CCD camera).
The effects of these system elements are fundamentally described through the
system impulse response, h(x), introduced in Eq. (7.2), which will be expressed
as 1D for this discussion. In an LSI system, the response, I(x), is a convolution
of the real-world signal, T (x) in our case, with an impulse function, h(x), as
I(x) = T(x)*h(x). The total impulse response can be decomposed into elements
representing the major components of the system. For our case, based on
Fig. 7.2, h(x) = hD(x)*h(x)*hCCD(x), corresponding to the neutron beam
aperture, scintillator, and CCD sensor, respectively. We model these compo-
nents of the system using idealized functions for the physical neutron beam
aperture, converter optical diffusion, and sensor sampling,

Scintillator

Aperture Object δ CCD

D sensor
Δs

z
n
Optical
Beam line system
L d
x '' x' x

Fig. 7.2 Representation of a basic radiographic image formation system used to estimate
resolution performance during the design stage. Model accounts for geometry, aperture size,
scintillator resolution, and optical system resolution
7 Mathematics of Neutron Imaging 113

   0 x
x 00 x
hD ðx 00 Þ ¼ rect ; h ðx 0 Þ ¼ Gauss ; hCCD ðxÞ ¼ rect ; (7:3)
D  s

where rect() is a rectangle function and Gauss() is a Gaussian function as

defined in [14].
Note in Eq. (7.3) and Fig. 7.2, the introduction of different scales x00 , x0 , and x.
The aperture in plane x00 is magnified at the scintillator plane, x0 , which is in turn
magnified (or de-magnified) in the CCD plane, x. The aperture-to-scintillator
magnification is given by MD=d/L, whereas the scintillator-to-CCD magnifica-
tion is defined as MCCD= (object in scintillator plane/object in CCD sensor
plane). The transformation, x = MCCDx0 = MDMCCDx00 , results in the follow-
ing analytical model for the impulse response, h, at the CCD imaging plane, x,
    x
x x
hðxÞ ¼ rect  Gauss  rect : (7:4)
MD MCCD D MCCD  s

The normalized modulus of the Fourier transform of the impulse response is

known as the modulation transfer function [7], defined as MTF(u) = |H(u)|/
|H(0)|, where H(u) is the system transfer function calculated by taking the
Fourier transform of the impulse response. The MTF describes the magnitude
of the frequency response of the system and is useful for depicting and quantify-
ing system resolution as discussed in Chapter 6.
From Eq. 7.4, we can analytically produce MTF(u) through Fourier trans-
formation and simplification to yield,
   
 dMCCD 

MTFðuÞ ¼ sinc u  GaussðMCCD uÞ  sincðs uÞ  ; (7:5)
L=D

where sinc(x) = sin(x)/x. Note that this expression has been put into a form
explicitly containing the L/D ratio.
In Eq. (7.5), the sinc () term drives much of the response typically observed
in a radiography system. For example, as L/D !1, this term broadens out in
frequency representing improved resolution. Also, as the object moves closer to
the detection plane, i.e., as d! 0, resolution also improves. These are both
effects that are readily observed in radiography systems.
An example of the MTF is shown on the left-hand-side of Fig. 7.3 for a system
with the model parameters: L/D=150, d=10 cm, =50 mm, MCCD=0.25. MTF
is a measure of system response versus frequency content of the object being
imaged. In this figure, the X axis represents frequency, with units of cycles per
millimeter often referred to as line pairs per millimeter. The Y axis represents the
transmission of this frequency to the image. In this example, the response to a
sinusoid with 10 cyc/mm drops to approximately 55% of the sinusoid’s amplitude.
Applying the Rayleigh criterion of 10% MTF [16] to this example, indicates
that the highest resolvable frequency is 17 cyc/mm. System resolution is defined
114 K.W. Tobin et al.

0.8 0.8
L /D = 150
0.6 0.6
MTF (μ)

0.4 Total MTF 0.4

Beam aperture
Scintillator L /D 200
0.2 0.2
CCD/Optics L /D 150
10% Rayleigh criterion L /D 100
0 0
0 5 10 15 20 0 5 10 15 20
μ (cyc /mm) μ (cyc /mm)

Fig. 7.3 MTF modeling results for typical radiography system parameters. The left plot shows
the frequency response of constituent system components at a specified L/D ratio, while the
right plot shows the system response for a variety of L/D ratios

as the period of this frequency. Therefore, resolution is 1/(17 cyc/mm) = 60 mm

at the CCD imaging plane. Due to magnification in the system, the resolution at
the object plane is 231 mm.
Note also for this geometry that the beam aperture (or equivalently the L/D
ratio) is the primary factor impacting resolution. The scintillator screen and the
optical system (defined by the CCD sample rate) produce only minor relative
degradation. On the right-hand side of Fig. 7.3, the MTF is shown for a variety
of L/D ratios, indicating the strong effect this factor has on the resolution of a
neutron radiography system. Of course, the obvious result of this analysis is to
increase the L/D ratio either by moving farther from the beam aperture or by
reducing the size of the aperture. Unfortunately, both of these choices also
result in a rapid reduction in the available neutron flux, requiring further signal-
to-noise analysis to optimize both the geometry and the counting statistics.

7.2.2 System Performance Measurement

Measurement of system performance allows system designers to determine if

the imaging system is performing as expected and to give potential users
information useful in determining whether the imaging system is appropriate
for their application. The key measurements of system performance are resolu-
tion and contrast discrimination. Resolution is a measure of the details that can
be seen within an image while contrast discrimination measures the contrast
required between an object and its background to resolve the object.
Resolution measurement methods can be categorized as direct and indirect.
For direct methods, a series of test objects over a range of sizes near the
resolution limits of the system are used. These test objects take many forms
from line pairs to spheres to radial patterns and provide a direct visual display
of the resolution on the radiographs [7].
7 Mathematics of Neutron Imaging 115

Indirect resolution measurements use the edge response of the system to

provide a more quantitative view of resolution. These measurements result in an
MTF for the system. MTF can be calculated for radiography systems through
measurement of the response of the system to an edge object. These edges are
tilted to provide appropriate edge sampling by a pixilated imager [37, 32]. Tilted
thin wires can also be used for this method [29]. The MTF is calculated from an
edge image by first locating the edge within the radiograph and binning the edge
response perpendicular to the line over the length of the line to get the edge
response function, Er(x). Taking a derivative of the edge response function
produces the edge spread function, which is then Fourier transformed to produce
F (u), which is used to produce the MTF, where F (u) is given explicitly by,
Z
FðuÞ ¼ Er0 ðxÞ eiux dx ; (7:6)

where Er0 (x) = dE (x)/dx represents a suitable discrete derivative process.

Applying these methods to a radiograph will measure the MTF of the imaging
system and verify the expected results based on the calculations from the
previous section.
Neutron radiography systems are naturally extended to perform computed
tomography (CT) for 3D reconstruction. Resolution of the reconstructed results
will depend on the system, data collection, and reconstruction parameters as well
as contrast in the object. Measurement of the empirical resolution of the 3D
results after tomographic (CT) reconstruction can also be obtained from the
determination of the edge response. The American Society for Testing and
Materials (ASTM) has developed a standard test method for measurement of
CT system performance designated E1695. In this standard, a cylindrical phan-
tom is used to determine both spatial resolution and contrast discrimination
measurements on the reconstructed data from a CT system as shown in Fig. 7.4.
The shown example is a result from X-ray CT imaging; however, neutron
tomography systems will produce the same type of data and the ASTM standard
is directly applicable.

0.8

0.6
MTF

0.4

0.2

Projection Reconstruction 0
0 0.1 0.2 0.3 0.4 0.5
Frequency (lp/pixel)

Fig. 7.4 Spatial resolution from CT reconstruction using ASTM E1695 standard
116 K.W. Tobin et al.

In the ASTM E1695 standard method, a cylindrical phantom is scanned and

a slice normal to the central axis of the cylinder is reconstructed. A normal slice
through the cylinder contains a circular cross-section of the cylinder. Just as in
the tilted edge method, the MTF is calculated from the line edge response. The
line edge response is found by locating the center of this circle and circularly
integrating the pixels around this center point. In the tilted edge method, the tilt
provides subpixel resolution across the edge. The pixel sampling on the circle
edge will also provide this higher resolution edge response assuming the circle is
of adequate size. Guidelines for the size selection are provided in the standard.
The MTF characterizes the spatial resolution of the system as described earlier.
The second key performance parameter for CT systems is also included in ASTM
E1965 and is the contrast discrimination function (CDF). The CDF measures the
ability to discriminate an object from its background and is calculated using the
same cylindrical phantom or spherical phantom used for the MTF calculation.
For calculation of the CDF, the region within the reconstructed slice through
the phantom is subdivided into neighborhoods of size N  N as shown on the
left side of Fig. 7.5. For CDF, the mean value for each neighborhood and the
standard deviation of the means are calculated. This process is repeated for
N=1,2,. . ., M, until there are fewer than 25 neighborhoods within the cylinder
slice. Finally, for confidence in discrimination of the object from the
background, the contrast needs to be three times greater than the standard
deviation in the noise. Therefore, the standard deviations are multiplied by 3
and graphed as a function of neighborhood size to produce the CDF.
Since the cylinder is a uniform material, CDF can be interpreted as the
average noise in the reconstructed image for various levels of pixelization. An
example CDF is given on the right side of Fig. 7.5. In this example, the CDF
shows that a contrast of 70% is required to discriminate a single pixel object
from the background, and a contrast of 10% enables discrimination of objects
larger than 33 pixels.

75
Central
Contrast sensitivity (%)

reconstructed 60
slice
Tiling 45

30

15
1/3 radius
0
1 2 3 4 5 6 7
Region size in pixels (N×N)

Fig. 7.5 Contrast discrimination measurement geometry (left) and CDF results (right)
7 Mathematics of Neutron Imaging 117

The 3D MTF measurement method has been extended to use a spherical

phantom [6] for two reasons. First, any tilt in the cylinder will produce an
elliptical cross-section and degrade the calculated edge response, whereas any
slice through a sphere will produce the appropriate circular cross-section.
Second, resolution will vary over the reconstruction space for CT systems
depending on the sampling and the reconstruction method used. This spherical
phantom implementation locates the center of the sphere and integrates over
spherical surfaces centered on this point to develop an edge response function
that is an average over all directions on the spherical surface. Again using
Eq. (7.6), the MTF is calculated over all directions.
With the resolution and contrast measurements presented in the graphical
forms as shown in Figs. 7.4 and 7.5, a user can determine whether the imaging
system is suited for a particular measurement.

7.3 Volumetric Imaging

Neutron imaging systems have followed the path of x-ray imaging systems in
the introduction of rotational systems and CT reconstruction algorithms to
perform volumetric imaging. An introduction to neutron tomography was
provided in Chapter 6. This section builds upon the introduction by providing
a mathematical description of CT and the classical methods for reconstruction
of volumetric data.
Given a collection of projection images of an object, a 3D volumetric image
can be reconstructed using either an analytic or an iterative algorithm. The
former is based on some variant of filtered backprojection. The latter is based
on optimization of a criterion function that relates the fit of the volumetric
image with the observed projection data. In this section, we review the mathe-
matics behind these conceptually different approaches and comment on advan-
tages and disadvantages. To simplify the discussion, we apply the following
model of the projection data:
 Z 
I ¼ I0 exp  mðÞ d ; (7:7)
K

where I0 and I refer to the beam intensity before and after interaction with
the object, K denotes the path of the beam, and m represents the material-
dependent macroscopic cross-sections. To reconstruct m, which corresponds to
T, in the preceding section for neutron radiography, we apply log-normal-
ization to isolate the cross-section (Eq. (7.1), assuming B1),
Z
I
mðÞ d ¼  log : (7:8)
K I0
118 K.W. Tobin et al.

We note that in practice, the result of a dark field scan (where the source is
turned off) is often subtracted from both I0 and I to compensate for noise
generated by physical processes that occur inside the imager (CCD).
Different imaging geometries require different reconstruction algorithms.
We consider both parallel and divergent beam collimation of the neutron
beam as shown in Fig. 7.6. We assume that the entire object is visible in each
projection image and that it is placed on a turntable to allow for a circular orbit-
type scan. Generalization to a helical scan is possible but beyond the scope of
the present exposition.

I(θ,s) I(θ,s(1+d/L))
s s
y y

x x
θ θ
u u

L d

Fig. 7.6 Parallel beam (left) and fan beam (right) sampling geometries. Cone beam sampling
extends the fan beam geometry axially

7.3.1 Filtered Backprojection

For 2D parallel beam data, the Fourier-slice theorem says that the Fourier trans-
form of a 1D projection equals a comparably angled slice through the 2D Fourier
space of the object being imaged [23]. Let p (,s) denote the log-normalized
projection data and let P (,w) be the Fourier transform thereof. Mathematically,
Z1
pð; sÞ ¼ mðs cos   u sin ; s sin  þ u cos Þ du
1
(7:9)
Z1
Pð; wÞ ¼ pð; sÞ expfi 2pwsg ds
1

Merging these expressions, while introducing the variable substitutions

x = s cos  u sin and y = s sin + u cos, leads to,
Z Z 1
Pð; wÞ ¼ mðx; yÞ expfi2p wðx cos  þ y sin Þgdx dy
1 (7:10)
¼ Mðw cos ; w sin Þ
7 Mathematics of Neutron Imaging 119

where M denotes the Fourier transform of m. While it would be tempting to

superimpose appropriately rotated 1D Fourier transforms of a collection of
projections and then obtain m through 2D Fourier inversion of the thus
approximated version of M, the interpolation needed to map the radial data
into the required Cartesian format would likely introduce artifacts in the
resulting image.
Filtered backprojection is a better approach. The mathematical derivation
can be summarized as follows [35]. When rewriting the inverse Fourier trans-
form of M using polar coordinates, the resulting Jacobian introduces an abso-
lute value-based frequency multiplier. That is,
ZZ 1
mðx; yÞ ¼ Mðu; vÞ expfi 2p ðux þ vyÞ gdu dv
1
Z p Z 1
(7:11)
¼ jwj Mðw cos ; w sin Þexpf i2p ðx cos þ y sin Þg dw d
0 1

We can apply the Fourier-slice theorem to replace M (w cos , w sin ) by

P (,w). For simpler notation, we can furthermore express the inverse Fourier
transform of |w| P (,w) as a convolution of p (,s) with a kernel that implements
the ramp-filter corresponding to |w|. The result is the well-known algorithm,
Z p
mðx; yÞ ¼ p~ð; x cos  þ y sin Þ d
0 (7:12)
p~ð; sÞ ¼ pð ; sÞ  hRAMP ðsÞ

where the ramp-filter is often band-limited to suppress high-frequency noise

and sampling based aliasing.
The extension to a 2D fan-beam geometry is not difficult. Essentially, geo-
metric weighting is introduced to compensate for the divergent nature of the
projection rays. The algorithm for equiangular sampling seen in connection
with a curved ring-detector is very similar to that associated with the equidistant
sampling geometry of a linear detector. The latter algorithm is given by,

Z2p
1
mðx; y; zÞ ¼ g2 ð; x; yÞ~
pð; gð; x; yÞðx cos  þ y sin ÞÞ d
2
0
(7:13)
gð; x; yÞ ¼L=ðL þ x sin   y cos Þ
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
p~ð; sÞ ¼L= L2 þ s2  pð; sÞ  hRAMP ðsÞ

where L refers to the distance from the neutron beam aperture to the isocenter
of the object (c.f. Fig. 7.6). The detector is assumed to be located at the isocenter
for mathematical simplicity (i.e., d=0). Notice that the backprojection takes
place over 2p instead of just p as used for the parallel beam geometry. Also note
that L! 1 leads to the parallel beam algorithm.
120 K.W. Tobin et al.

The extension to 3D parallel beam data is even more straightforward. In fact,

the reconstruction algorithm is identical to the one above for 2D parallel beam
data except for the increase in dimensionality.
The extension to a 3D cone-beam geometry based on a flat-panel detector is
more involved. Tuy [40] established a sufficient condition for reconstruction
that says that every image plane must contain at least one source point. While
this is satisfied for helical scanning and other trajectories, it is not satisfied for a
circular-orbit scan, which is considered here. This leads to an approximate
algorithm of which Feldkamp’s without a doubt is the most widely used.
Developed for industrial x-ray cone-beam CT, the Feldkamp algorithm [13]
can be viewed as a generalization of the fan-beam algorithm. The central slice of
a cone-beam volume is reconstructed in exactly the same manner as it would
have been, had it been reconstructed from single-slice fan-beam data. Slices
further out in the cone, however, correspond to tilted fans and are thus handled
differently. While the mathematical details are too extensive for inclusion here,
the end result is the following algorithm,
Z2p
1
mðx; y; zÞ ¼ g2 ð; x; yÞ~ pð; gð; x; yÞ½x cos  þ y sin ; zÞ d
2
0
(7:14)
gð; x; yÞ ¼L=ðL þ x sin   y cos Þ
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
p~ð; ½s; tÞ ¼L= L2 þ s2 þ t2  pð; ½s; tÞ  hRAMP ðsÞ

Notice that z=0 leads to the fan-beam algorithm. Also note that filtering is
not applied in the axial dimension.
We mentioned that the fan-beam backprojection equation indicates that the
source trajectory data must cover 2p. However, by introducing a weight func-
tion that effectively limits each projection ray to be considered just once, Parker
[34] showed that it is possible to stop when p+ has been covered where
denotes the fan angle. The idea, which is heuristic, can be extended to apply to
cone-beam data as well.
Filtered backprojection algorithms, whether exact or approximate, are based on
continuous mathematics. When implemented for practical use, all integrals
are replaced by Riemann sums and nonintegral projection data lookup indices are
handled using some form of interpolation. The filtering can take place in either the
spatial domain or the frequency domain. When based on fast fourier transforms, the
latter can be done very efficiently. In that case, backprojection along with interpolation
dominates the computational cost. Still, given recent advances in computer technol-
ogy, it is possible to reconstruct even large images in a very short amount of time.
The image quality depends on taking enough projections and having good
count statistics to work with. Not taking enough projections invariably leads to
streak artifacts. Indeed, they are almost always present to some degree. Poor
count statistics results in a low-contrast, grainy image. Under normal imaging
conditions, however, good images are produced.
7 Mathematics of Neutron Imaging 121

7.3.2 Iterative Reconstruction

Henceforth, let vectors x and b refer to the image representing the macroscopic
cross-sections and the log-normalized projection data, and let matrix A denote
the discretization of the projection integral connecting the two. Reconstruction
can then be viewed as a matter of solving Ax=b. Numerous applicable iterative
algorithms exist ranging from Gauss–Seidel, Jacobi, and other stationary
methods to conjugate gradient (CG) and other Krylov subspace methods.
Stationary methods can be derived using matrix splitting [36]. For example,
A=MN leads to the iteration scheme,

xðkþ1Þ ¼ M1 NxðkÞ þ M1 b (7:15)

where x(0) is given. Convergence to the desired solution is guaranteed if

(M–1N) < 1 where  denotes the spectral radius of the matrix involved. Let
matrices D, L, and U represent the diagonal, strictly lower triangular, and
strictly upper triangular portions of A. Then Gauss–Seidel is the result of setting
M=DL and N=U. Likewise, Jacobi follows from setting M=D and
N=L+U. Both methods converge if A is strictly diagonally dominant.
Krylov subspace methods monotonically minimize the residual error using
orthogonality conditions [36]. They are nonstationary in the sense that the
computations involve information that changes each iteration. The prototypi-
cal example is CG, which is based on the assumption that matrix A is sym-
metric, positive-definite. The iterative update scheme is given by

 ¼ jrðkÞ j2 =dðkÞ j2A

xðkþ1Þ ¼ xðkÞ þ  dðkÞ
rðkþ1Þ ¼ b  Axðkþ1Þ (7:16)
 ¼ jrðkþ1Þ j2 =jrðkÞ j2
dðkþ1Þ ¼ rðkÞ þ  dðkÞ

where x(0) is given, d(0)=r(0) = b Ax(0), |r|2 = rTr, and |d|A2= dTAd.
Whether a stationary method or a Krylov subspace method is used, the
search for x is terminated either when the residual norm associated with the
current iterate is deemed to be small enough, or when a fixed number of
iterations have been executed.
In most imaging applications, the linear system is under-or overdetermined
and a least squares problem must be solved instead. Let this be denoted by

x ¼ argminjAx  bj2 ; (7:17)

where |Axb|2 = (Axb)T(Axb). By taking the derivative of the squared

residual norm with respect to x and equating the result to zero, we obtain the
122 K.W. Tobin et al.

normal equations which form a linear system that can be solved by any of the
above-mentioned methods, namely,

AT Ax ¼ AT b (7:18)

Other methods include SIRT, which is a stationary method for solving a

weighted least squares problem that has been widely used in medicine and
biology [15;18]) and LSQR, which is a CG-like method with better numerical
properties [33].
The reconstructed image should not contain any negative values, as they
have no physical interpretation. Their absence is only guaranteed if the least
squares problem is constrained accordingly. This leads to,

x ¼ argminjAx  bj2 subject to x  0 : (7:19)

Note that depending on the neutron imaging system, a highly scattering

object can cause neutron detection in detectors just outside the shadow of the
object under test. These additional neutrons can result in negative attenuation
values. Therefore, modifications to the typical constraints may be required
along these edges.
A number of methods exist for solving constrained quadratic optimization
problems, one example being to embed the least squares algorithm in a pro-
jected gradient framework [5, 19]. Computationally, an outer algorithmic layer
iteratively (i) uses an inexact line search based on safeguarded quadratic inter-
polation to determine the set of free variables, i.e., those that are strictly positive
and (ii) invokes the least squares algorithm to compute a feasible solution in
that subspace. Control is passed back from the least squares solver to the outer
algorithmic layer either at convergence or when a variable is about to violate the
nonnegativity constraint.
Computational steering can be introduced through the addition of a penalty
term [26]. For example, suppose that the image is known to consist of regions of
uniform macroscopic cross-sections. In that case, it might be advantageous to
discourage neighboring voxels from taking on different values except near
region boundaries. This could be achieved by applying an edge-preserving
Markov random field that has smoothing properties [8]. Other structural con-
straints are possible. Generally speaking, the resulting optimization problem
can be written as

x ¼ argminjAx  bj2 þ  RðxÞ subject to x  0 (7:20)

where R(x) denotes the penalty term and  is a free parameter that controls the
influence exerted thereby on the solution. From a computational point of view,
linear penalty terms are preferable as they can be handled by straightforward
row-expansion of A and b.
7 Mathematics of Neutron Imaging 123

Bayesian reconstruction forms an alternative to algebraic methods. Com-

monly used in the field of medical imaging, this iterative approach is statistical
by nature. The image formation is described probabilistically, thereby making
reconstruction a matter of model parameter estimation. Specifically, let
p(x,b|) denote the data likelihood for x, the unobservable image data, and b,
the observable projection data, given , the associated model parameters. As
above, Ax=b, only now A is a matrix of detection probabilities. The goal is then
to solve the penalized maximum likelihood problem

 ¼ argmax log pðx; bjÞ þ  RðÞ : (7:21)

where R, strictly speaking, is a log-prior; although in practice a heuristic

smoothing term is often used.
Methods for maximum likelihood estimation include gradient descent, con-
jugate gradient, and similar algorithms. These typically require first and/or
second derivatives of the likelihood function to be evaluated. This is not the
case for the two-step EM algorithm [11]. First, an E-step is carried out to
compute the expectation of the data log-likelihood given the observed data
and the current parameter estimate. Then, an M-step is used to re-estimate the
model parameters by means of maximization of the expectation. The result is
the following iteration scheme

kþ1 ¼ argmax E½log pðx; bjÞjb; ðkÞ  þ  RðÞ (7:22)

In some applications, e.g., maximum likelihood estimation for independent

Poisson events, the M-step is trivial as it has a closed-form solution. This allows
the E-step and the M-step to be combined. For other applications, the M-step
has to be implemented as an iterative computation of its own. This may also be
the case for penalized maximum likelihood estimation, as the penalty term may
preclude true maximization from taking place. The M-step is then turned into
an iterative parameter update computation that ensures a nondecrease of the
likelihood [21].
There are several advantages associated with taking an iterative approach to
image reconstruction. For one, it is possible to quite accurately model the system
geometry including nonuniformities in the sampling pattern as well as sensitivity
variations across the detector. As pointed out above, it is also possible to steer the
computation away from structurally undesirable images. Poor count statistics
tend not to be a problem either, as the objective of both algebraic and statistical
reconstruction methods is to find a solution that best fits the data. That is, there
is no built-in assumption of data consistency across projections. Iterative
approaches also allow reconstruction with few projections and irregular projec-
tion patterns and enable the inclusion of more accurate physics models up to full
Monte Carlo models. This flexibility comes at a high cost with respect to both
implementation complexity and computational cost at run-time. However, if
these can be overcome, the payoff is a high-quality image.
124 K.W. Tobin et al.

7.3.3 Computer Platforms

Many platforms exist that can provide the compute power needed in order for
a volumetric image to be reconstructed in a reasonable amount of time. A
relatively small cluster of multi-core PCs may suffice to solve even a rather
large problem in a matter of minutes [18]. From a programming point of view
this is conceptually straightforward as it merely involves writing multi-
threaded code for a distributed memory environment. In recent years, com-
modity graphics processing units (GPUs) have seen a rapid increase in both
capability and programmability to the point that they are now being used for
general purpose computing and thus also tomographic image reconstruction
[31]. Papers often report order-of-magnitude performance gains over opti-
mized CPU applications. However, GPU programming is not for the novice.
Also, floating-point arithmetic may not follow IEEE standards, which could
be a problem. Field-programmable gate-arrays (FPGAs) provide a different
but equally powerful platform from a data throughput point of view. How-
ever, FPGAs are notoriously difficult to program, especially since floating-
point arithmetic may not be supported to the extent needed. Finally, we
mention the cell processor, which conceptually is somewhat similar to a
multi-core CPU only orders-of-magnitude faster. High-level programming is
possible, e.g., in C, but some thought must be devoted to memory handling as
the processing units only have a limited amount of memory available. None-
theless, the cell may represent the fastest platform currently on the market for
volumetric imaging [22].

7.4 Conclusions and Application

The ability to model, design, and measure the resolving power of a radiography
system is critically important to the experimental design required to achieve
relevant measurement characteristics of complex, material structures. The focus
of this chapter has been on methods to design and characterize radiography
systems according to experimental goals for volumetric measurements (i.e., CT
recording and reconstruction methods). This focus has been primarily related to
spatial resolution and the estimation of macroscopic material interaction cross-
sections.
New or modified radiography facilities that generate cold neutrons [38]
provide wavelength tunable monochromaters [39] or provide time-of-flight
(TOF) measurements [27], facilitating the radiographic capture of images that
take I(,T,x,y,z) to I(E,,T,t,x,y,z). The ability to better characterize
the energy–time–space dependencies of materials, T(E,t,x,y,z), will produce
better material resolution.
As an example of how to adapt this discussion to higher resolution measure-
ments, consider the example MTF shown in Fig. 7.7. Using Eq. (7.5), we have
7 Mathematics of Neutron Imaging 125

Fig. 7.7 MTF for a system

design with L/D = 750,
corresponding to an
0.8
aperture size of 5 mm. the
resolution (71 cyc/mm or
14 mm at the detector) is now 0.6

MTF (μ)
comparable to the
scintillator limit. How do we
achieve sufficient neutron 0.4 Total MTF
counting statistics? Beam Aperture
Scintillator
0.2
CCD/Optics
10% Rayleigh criterion
0
0 20 40 60 80
μ (cyc/mm)

hypothesized a radiography system with L/D = 750, corresponding to a beam

aperture of 5 mm at a distance of 3.75 m from the object—a seemingly achiev-
able design. For a scintillator material with an optical diffusion parameter of
= 10 mm, the resolution of such a system would be 14 mm at the sensor, moving
from the macro-analysis scale to the microscopy scale. Reducing the beam
aperture from 25 mm (our earlier example) to 5 mm will increase system
resolution by a factor of 4, yet it is likely that neutron-counting statistics and
background noise will result in measurement challenges that are difficult to
overcome.
Overcoming low neutron counts may be achievable by encoding the beam
aperture, i.e., implementing an array of 5 5 mm (or smaller) apertures that
constitute a significant total area (e.g., >25 25 mm). In such a case, the MTF
of Eq. (7.5) could be replaced by a new geometric transfer function, HCA() that
describes a coded aperture [17], e.g.,
   
 d MCCD 

MTFðuÞ ¼ HCA u  Gaussð MCCD uÞ  sin cðs uÞ (7:23)
L=D

The point is that these analysis tools can be adapted to new, more stringent
resolution requirements by applying creative ideas and methods.
For another example, the steady-state flux of a conventional reactor cannot
readily interrogate time-dependent phenomena as deeply as the high-intensity
and energy-dependent beams available at a spallation neutron source. Using a
spallation source, the potential exists to develop new time-synchronized,
energy-resolved methods to better resolve subtle material differences. This is
potentially achievable by taking advantage of the Bragg edges in materials [27]
or of the fact that absorption cross-sections, a at low neutron energies
p
are typically proportional to 1/ E. Using prior knowledge of the material
characteristics under test should facilitate the collection of a series of
126 K.W. Tobin et al.

energy-windowed images, IE(,T,x,y,z), whose linear combinations, e.g.,

using principle component or linear discriminant analysis [12], may provide
improved contrast (i.e., CDF) between materials.
Finally, any of these methods that compartmentalize the total neutron beam
into small spatial- or energy-dependent units will make volumetric recovery of
material characteristics more challenging due to reduced signal-to-noise and the
corresponding artifacts that conventional CT methods produce in low-signal
environments. In these situations, the iterative reconstruction methods intro-
duced in Section 7.3.2 become more relevant to achieving adequate signal
recovery and reconstruction.

References
1. R. Accorsi, F. Gasparini, et al., A Coded aperture for high-resolution nuclear medicine
imaging with conventional anger camera. IEEE Trans. Nucl. Sci. 48:6, 2411 (2001).
2. H. Asano, N. Takenaka, et al., Visualization and void fraction measurement of gas-liquid
two-phase flow in a commercial plate heat exchanger by thermal neutron radiography.
IEEE Trans. Nucl. Sci. 52:1, 280–284 (2005).
3. M. Balasko, E. Svab, et al., Comparison of neutron radiography with other nondestruc-
tive methods. IEEE Trans. Nucl. Sci. 52:1, 330–333 (2005).
4. L.G.I. Bennett, T.R. Chalovich, et al., Comparison of neutron radiography with other
non-destructive techniques for the inspection of cf188 flight control surfaces. IEEE Trans.
Nucl. Sci. 52:1, 334–337 (2005).
5. M. Bierlaire, P.L. Toint, et al., On iterative algorithms for linear least squares problems
with bound constraints. Linear Algebra Appl. 143, 111–143 (1991).
6. P. Bingham, L. Arrowood, et al., Calibration and performance testing for reconfigurable
computed tomography systems. Mat. Eval. 65, 1102–1107 (2007).
7. G.D. Boreman, Modulation Transfer Function in Optical and Electro-Optical Systems.
Bellingham, WA, SPIE Press (2001).
8. C.Bouman and K. Sauer, A generalized Gaussian image model for edge-preserving MAP
estimation. IEEE Trans. Image Process. 2, 296–310 (1993).
9. S. Casalta, G.G. Dawuino et al., Digital image analysis of X-ray and neutron radiography
for the inspection and the monitoring of nuclear materials. NDT&E Int. 36, 349–355
(2003).
10. J.M. Cimbala, J.S. Brenizer, et al., Study of a loop heat pipe using neutron radiography.
Appl. Radiat. Isot. 61, 701–705 (2004).
11. A. Dempster, N. Laird, et al., Maximum likelihood from incomplete data via the EM
algorithm. J. R. Stat. Soc. B 39, 1–38 (1977).
12. R.O. Duda, P.E. Hart, et al., Patt. Class.. New York, John Wiley & Sons, Inc. (2001).
13. L.A. Feldkamp, L.C. Davis, et al., Practical cone beam algorithm. J. Opt. Soc. Am. A 1,
612–619 (1984).
14. J.D. Gaskill, Linear Systems, Fourier Transforms, and Optics, John Wiley & Sons, Inc.
(1978).
15. P. Gilbert, Iterative methods for the three-dimensional reconstruction of an object from
projections. J. Theor. Biol. 36, 105–117 (1972).
16. J.W. Goodman, Introduction to Fourier Optics, Roberts & Company Publishers (2004).
17. S.R. Gottesman, and E.E. Fenimore, New family of binary arrays for coded aperture
imaging. Appl. Opt. 28:20, 4344 (1989).
18. J. Gregor and T. Benson, Computational analysis and improvement of SIRT. IEEE
Trans. Med. Imaging 27, 918–924 (2008).
7 Mathematics of Neutron Imaging 127

19. J. Gregor and F.R. Rannou, Least squares framework for projection MRI reconstruction.
Medical Imaging, San Diego, CA, Proceedings of SPIE (2001).
20. A.A. Harms and D.R. Wyman, Mathematics and Physics of Neutron Radiography.
Boston, D. Reidel Publishing Company (1986).
21. T. Hebert and R. Leahy, A generalized EM algorithm for 3-D Bayesian reconstruction
from Poisson data using Gibbs priors. IEEE Trans. Med. Imaging 8, 194–202 (1989).
22. M. Kachelriess, M. Knaupp, et al., Hyperfast parallel-beam and cone-beam backprojec-
tion using the Cell general purpose hardware. Med. Phys. 34, 1471–1486 (2007).
23. A. Kak and M. Slaney. Principles of Computerized Tomographic Imaging, IEEE Press
(1988).
24. G.F. Knoll, Radiation Detection and Measurement, John Wiley & Sons, Inc. (2000).
25. J.R. Lamarsh, Introduction to Nuclear Engineering, Addison-Wesley Publishing Com-
pany (1983).
26. C.L. Lawson and R.J. Hanson, Solving Least Squares Problems, Prentice-Hall (1974).
27. E. Lehmann, P. Vontobel, et al., Neutron imaging – present status and options with tof
methods – TOF neutron imaging. J. Neutron Res. 13, 27–31 (2005).
28. U. Matsushima, Y. Kawabata, et al., Measurement of changes in water thickness in plant
materials using very low-energy neutron radiography. Nucl. Instrum. Meth. Phys. Res.
A 542, 76–80 (2005).
29. S.L. Meeks, J.F. Hamon, et al., Performance characterization of megavoltage computed
tomography imaging on a helical tomotherapy unit. Med. Phys. 32, 2673–81 (2005).
30. J.J. Milczarek, A. Czachor, et al., Dynamic neutron radiography observations of water
migration in porous media. Nucl. Instrum. Meth. Phys. Res. A 542, 232–236 (2005).
31. K. Müller and F. Xu, Accelerating popular tomographic reconstruction algorithms on
commodity PC graphics hardware. IEEE Trans. Nucl. Sci. 3, 654–663 (2005).
32. M. Ohkubo, S. Wada, et al., An effective method to verify line and point spread functions
measured in computed tomography. Med. Phys. 33, 2757–64 (2006).
33. C.C. Paige and M.A. Saunders, LSQR: An algorithm for sparse linear equations and
sparse least squares. ACM Trans. Math. Software 8, 43–71 (1982).
34. D. Parker, Optimal short-scan convolution reconstruction for fanbeam CT. Med. Phys. 9,
254–257 (1982).
35. A. Rosenfeld and A. Kak, Digital Picture Processing, Academic Press (1976).
36. Y. Saad, Iterative Methods for Sparse Linear Systems, PWS Publishing Co (1996).
37. E. Samei and M.J. Flynn, A method for measuring the presampled MTF of digital
radiographic systems using an edge test device. Med. Phys. 25, 102–113 (1998).
38. M. Tamaki, Conceptual monochromatic digital neutron radiography using continuous
cold neutron beam. Nucl. Instrum. Meth. Phys. Res. A 542, 32–37 (2005).
39. W. Treimer, M. Strobl, et al., Wavelength tunable device for neutron radiography and
tomography. Appl. Phys. Lett. 89, 203504 (2006).
40. H. Tuy, An inversion formula for cone-beam reconstruction. SIAM J. Appl. Math. 43,
546–552 (1983).
Chapter 8
Neutron Phase Imaging

F. Pfeiffer

Abstract This chapter is dedicated to neutron phase imaging. We first review

the underlying physical principles describing the phase shift that neutrons
experience when they propagate through matter. Based on these principles we
describe the three main approaches for neutron phase imaging and discuss their
extension to three-dimensional neutron phase tomography. The chapter ends
with a final discussion of the concept of coherence and its implications con-
cerning practical neutron phase-imaging applications.

Keywords Neutron phase-contrast tomography  Neutron phase-contrast

imaging  Neutron phase imaging

8.1 Introduction

From a classical physics point of view, neutrons are small massive particles with a
confinement radius of about 1015 m and a distinct internal quark-gluon structure.
In quantum mechanics, neutrons are described by de Broglie [1] wave packets
whose spatial extent may be large enough to show interference effects similar to
what can be observed with visible laser light or highly brilliant X-rays from
synchrotron sources. Measurements of the neutron wave packet’s phase shift
induced by different interaction potentials have a long and distinguished history
in the exploration of the fundamental principles of quantum mechanics [2, 3]. If
such phase-sensitive measurements are further combined with neutron-imaging
approaches, two or even three dimensionally resolved spatial information on the
quantum mechanical interactions of massive particles with matter can be obtained.
In the following, the term ‘‘phase imaging’’ is used to describe not only phase
contrast methods, which simply record some form of contrast generated due to
a phase shift induced by the object, but also methods that determine the phase

F. Pfeiffer (*)
Department of Physics (E17), Technical University Munich,
85747 Garching/Munich, Germany
e-mail: [email protected]

I.S. Anderson et al. (eds.), Neutron Imaging and Applications, 129

Neutron Scattering Applications and Techniques, DOI 10.1007/978-0-387-78693-3_8,
Ó Springer ScienceþBusiness Media, LLC 2009
130 F. Pfeiffer

shift quantitatively. The research activities in the direction of quantitative phase

imaging are particularly interesting because they open the way for more sophis-
ticated methods such as neutron phase tomography or quantitative phase
imaging of fundamental quantum mechanical interactions.
Conventional attenuation-based neutron computed tomography (CT) is well
developed as a three-dimensional (3D) imaging method. Neutron phase
tomography was attained by introducing the technique of neutron phase-sensitive
imaging into neutron tomography. In this case, because the neutron phase shift is
expressed as a projection of the real part of the refractive index, a reconstructed 3D
image corresponds to a map of the real part of the refractive index. Neutron phase
tomography, of course, inherits its high and complementary sensitivity from
neutron phase imaging.
Since phase information is generally lost in measuring neutron transmission
images, several techniques have been developed to convert the neutron phase shift
to neutron intensity modulation. Because of the short wavelength of neutrons,
comparatively low coherence of neutron beams, and limited numerical aperture of
neutron optical elements, the construction of phase-sensitive optics is not as flexible
as in the visible-light region. Nevertheless, recent developments, specifically in
neutron optics, have opened the door for several approaches to carry out neutron
phase imaging. In this chapter we describe the basic characteristics and principles
of phase-sensitive imaging and tomographic methods for extending quantitative
phase imaging into three dimensions and present an overview of the existing
experimental approaches for phase-sensitive imaging and tomography.

8.2 Principles

8.2.1 Phase Shift Versus Attenuation

Let us consider an object characterized by a complex neutron refractive index

distribution

nðx; y; zÞ ¼ 1  ðx; y; zÞ þ iðx; y; zÞ ; (8:1)

where  (x,y,z) describes the decrement from unity of the real part of the
refractive index and  (x,y,z) is the imaginary part of the refractive index. A
conventional neutron attenuation-based projection image maps the imaginary
part of n and is given by
 Z   Z 
Iðy; zÞ
¼ exp  ðx; y; zÞdx ¼ exp  4pðx; y; zÞ=ldx ; (8:2)
I0 ðy; zÞ

where l is the wavelength of the neutron radiation. Moreover, we have assumed

that the neutron beam projects the object along the x-direction onto the (y,z)
detector plane and that the linear coefficient  is related to the imaginary part of
8 Neutron Phase Imaging 131

n with  = l/4p . Generally  contains not only the specific neutron capture
cross-section of the material, but also the effect that neutrons are removed from
the beam traversing the specimen to coherent or incoherent scattering
processes.
In analogy to the description of attenuation-based projections, the projected
phase shift of an object can be written as [3]
Z
2p
ðy; zÞ ¼ ðx; y; zÞdx : (8:3)
l

A phase shift projection image thus can give direct information on the real part
of the refractive index distribution in the object and carries additional and
complementary information if compared to an attenuation image.
Unlike X-rays, where the values of  and  are linked through the X-ray
optical properties of the materials, the corresponding values for neutrons can be
very different. Table 8.1 shows values for several materials for a neutron
wavelength of 4Å. It is interesting to note that, for example, titanium and
copper have very similar  values, but opposite signs in . Moreover, we observe
that aluminum and silicon are characterized by very small  values and thus
yield very little contrast in a conventional attenuation image. Their relatively
large  values, on the other hand, make them easily detectable in phase-sensitive
imaging (see below). A graphical overview of the distribution of  values and
corresponding attenuation coefficients  is shown in Fig. 8.1.

Table 8.1 Values for the complex neutron refractive index n = 1-+i of several representa-
tive materials for neutrons with a wavelength of 4Å [4, 5]
Material   /
Al 5.3  106 9.8  1011 5.4  104
Si 5.3  106 6.0  1011 8.7  104
Ti 5.0  106 2.4  109 2.0  103
Mn 7.8  106 7.7  109 1.0  103
Cu 1.7  105 2.3  109 7.3  103
Gd 5.0  106 1.1  10 5
4.8  101
Pb 7.9  106 4.0  1011
2.0  105

In a more fundamental way, the decrement of the real part of the neutron
refractive index is related to the interaction potential that causes the phase shift.
For most materials, the nuclear interaction potential produces the strongest
phase shift and thus determines the  value. For an idealized slab of homo-
genous material with a thickness d, the phase shift due to the nuclear interaction
potential can be expressed by [3]

 ¼ Nbc ld ; (8:4)
132 F. Pfeiffer

where N is the atom number density and bc the neutron nuclear scattering
length. This is consistent with Eq. (8.3), when  is expressed by

l2 Nbc
¼ : (8:5)
2p

Fig. 8.1 Semilogarithmic plot of linear attenuation coefficient  versus the real part of the
refractive index  for different elements for thermal neutrons with a wavelength l ¼ 1.8Å [4, 5]
(courtesy of K. Lorenz)

The second most important interaction potential causing a phase shift is the
magnetic interaction. In this case  can be written as
Bmld
 ¼  ; (8:6)
h2
2p
where  is the neutron dipole moment, B the applied magnetic field, m the
neutron mass, and d the sample thickness.
In addition to the phase shift effect caused by the nuclear and magnetic
interactions, several other interaction potentials exist that cause a measurable
phase shift. Among those are the phase shift effects caused by the Coriolis [6],
the Aharanov–Casher [7], the scalar Aharanov–Bohm [8], the magnetic Joseph-
son [9], the Fizeau [10], and the geometric (Berry) [11] interactions.
8.2.2 Phase Tomography

Computed tomography enables the reconstruction of a 3D object from projec-

tion data measured over a range of projection angles. In conventional
8 Neutron Phase Imaging 133

attenuation-based neutron CT, a projection image corresponds essentially to a

projection line integral over the 3D distribution of the imaginary part of the
neutron refractive index of the object (Chapter 6).
In phase tomography, the aim is to quantitatively reconstruct the 3D distribution
of the real part of the refractive index of the object. To do so, a series of projection
images that exhibit some sort of phase contrast are recorded and used as input for
adequately matched CT reconstruction algorithms. The high sensitivity or comple-
mentary information present in the phase contrast images is of course inherited.
There are essentially three modalities for measuring phase contrast projection
images. On the fundamental physics level, they differ in the way they represent a
phase change induced by the object in the measured projection image. They can
be classified into (1) methods measuring the phase shift  directly, (2) methods
that yield one or both components of the phase gradient r, or (3) methods that
measure a projection image containing information on the Laplacian of the wave
front phase r2. In the following sections we will briefly discuss how tomo-
graphic reconstruction can be achieved in all three cases.

8.2.2.1 Reconstruction from Phase Shift Projections

Consider a 3D object described by g (x,y,z) as shown in Fig. 8.2. The projections
in a plane defined by z ¼ z0 through the object at an angle  to the x-axis are
given by the Radon transform [12]:

Fig. 8.2 Projection geometry for tomographic reconstruction. The three different projection
functions correspond to the three main modalities for obtained phase contrast images. G (y 0 , ):
direct measurement on the projected phase shift  induced by the object, D (y 0 , ): measurement
of the gradient of the phase shift r, and L (y 0 , ): measurement of a projection image
containing information on the Laplacian of the wave front phase r2
134 F. Pfeiffer

Z 1
Gðy 0 ; Þ ¼ gðx 0 ; y 0 Þdx 0 ; (8:7)
1

where x 0 and y 0 denote a coordinate system which is rotated by an angle 

around the z-axis with respect to x and y. Note that we have omitted the
variable z because it does not affect further derivation. A fast and accurate
algorithm for reconstructing the original object function g (x,y) from the Radon
transform Gðy 0 ; Þ is the filtered backprojection (FBP) algorithm, which can be
represented by the convolution backprojection integral [13, 14]:
Z p
gðx; yÞ ¼ F T 1 ½Gðv e 0 Þd ;
e 0 ; Þ  kðv (8:8)
0

where Gðve 0 ; Þ is the Fourier transform of Gðy 0 ; Þ, v 0 the reciprocal space
coordinate corresponding to y 0 ¼ x sin  þ y cos , kðve 0 Þ the Fourier represen-
1
tation of the filter function in the FBP, and F T the inverse Fourier trans-
form. For continuously sampled projection functions Gðy 0 ; Þ, the filter is
e 0 Þ  jv 0 j. In practice, where the projections are sampled with a
defined as kðv
finite resolution, y 0 , the filter should be truncated at the Nyquist frequency,
1=2y 0 , and becomes [13]

e 0Þ  jv 0 j; jv 0 j  1=2y 0
kðv : (8:9)
0; jv 0 j41=2y 0

(See also Fig. 8.3a.)

8.2.2.2 Reconstruction from Phase Gradient Projections

Let us now consider the case where the experimental arrangement does not yield
the ordinary line projection of the object function as defined by Eq. (8.7), but
the line projection of the partial derivative of the object function

Fig. 8.3 Backprojection filter functions for (a) simple projection line integrals, (b) gradient
projection integrals, and (c) projection image containing information on the Laplacian of the
object
8 Neutron Phase Imaging 135

Z 1
@gðx 0 ; y 0 Þ 0
Dðy 0 ; Þ ¼ dx : (8:10)
1 @y 0

Then tomographic reconstruction based on the conventional FBP using the

above-mentioned linear filter function [Eq. (8.9)] will not result in a correct
reconstruction of the original object function g (x,y). In analogy to what has
been suggested by Faris and Byer [15], in the case of deflection tomography
using visible laser light, the problem can be solved by adapting the filter
function in the FBP algorithm accordingly. Using the Fourier derivative theo-
e 0 ; Þ, where De is the Fourier transform of D.
e 0 ; Þ ¼ 2piv 0  Gðv
rem, we find Dðv
Changing the integration and differentiation order in Eq. (8.10) and substitut-
ing into Eq. (8.8), we obtain
Z p
gðx; yÞ ¼ F T 1 ½Dðv e 0 Þd ;
e 0 ; Þ  hðv (8:11)
0

with the differential phase contrast (DPC)-FBP filter function:

(
e Þ0
1
2pi  sgnðv 0 Þ; jv 0 j  1=2y 0
hðv ; (8:12)
0; jv 0 j41=2y 0

where sgnðv 0 Þ is the sign function (Fig. 8.3b).

8.2.2.3 Reconstruction from Laplacian Phase Projections

In the case where the projection images contain information on the second
derivative, or more precisely the Laplacian of the phase shift induced by the
object, r2, the projections are given by
Z 1 2 
@ gðx 0 ; y 0 ; zÞ @ 2 gðx 0 ; y 0 ; zÞ
Lðy 0 ; z; Þ ¼ þ dx 0 : (8:13)
1 @2y 0 @2z

We note that the z-axis, which was omitted for reasons of simplicity for the
treatment above, was reintroduced here. This is because projection images
containing the Laplacian of the phase shift cannot simply be factored into
single tomographic slices. The filter applied to the individual projections before
the back projection intrinsically becomes two-dimensional (2D) in this case.
Analogous to the derivation of the correct filter function, using the Fourier
derivative theorem above, we find that the tomographic reconstruction of the
object is given by
Z p
gðx; y; zÞ ¼ e 0 ; w; Þ  e
F T 1 ½Lðv lðv 0 ; wÞd ; (8:14)
0
136 F. Pfeiffer

e 0 ; w; Þ is the 2D Fourier transform of Lðy 0 ; z; Þ and w the reciprocal

where Lðv
counterpart of the z-coordinate. The filter function e
lðv 0 ; wÞ is of the form
( 0

e 0
1
4p2
 v 02jvþwj 2 jv 0 j  1=2y 0
lðv ; wÞ  ; (8:15)
0; jv 0 j41=2y 0

where sgn(v 0 ) is the sign function (Fig. 8.3c).

8.3 Experimental Methods

8.3.1 Crystal Interferometer

The most direct way of measuring a phase shift is interferometry, where a

coherent reference beam interferes with a beam that has been passed through
the sample. Because of the short wavelength of neutrons, however, the
construction of neutron interferometers is not straightforward. To form
neutron beam paths generating a stable interference, optical elements of a
neutron interferometer should be sufficiently stable so that the deviation of
the optical path length is smaller than the wavelength. Although it is a
challenging task, various neutron crystal interferometers have been
developed and used for neutron phase imaging. The stability of a crystal
interferometer is ensured by cutting its entire body monolithically out of a
single-crystal ingot.
Experimentally, the first single-crystal X-ray interferometer was reported by
Bonse and Hart in 1965 [16], and the technology was later transferred to
neutrons [17, 18]. As can be seen in Fig. 8.4, these devices typically consist of
three parallel lamellae separated by a constant spacing. The lamellae function as
beam splitters when a neutron beam is incident at the Bragg diffraction condi-
tion on a lattice plane perpendicular to the surface of the lamellae. The ampli-
tude of the incident neutron wave front is coherently divided into diffracted and

Fig. 8.4 Sketch of a monolithic crystal interferometer

8 Neutron Phase Imaging 137

forward-diffracted beams outgoing from the opposite side of the lamella.

Neutrons thus divided by the first lamella are divided again by the second
lamella in the same manner. Two beams overlapping at the third lamella are
also divided and interference is observed in the beams outgoing from the third
lamella. Because of its monolithic configuration, no mechanical tuning is
needed except in the arrangement of the interferometer so that the Bragg
diffraction condition is satisfied.
When an object is placed into one arm of the interferometer, it generates an
interference pattern, which in general is given by

Iðy; zÞ ¼ a0 ðy; zÞ þ a1 cos½ðy; zÞ þ inst: ðy; zÞ ; (8:16)

where a0 and a1 are the average intensity and fringe contrast, respectively. In
practice the interferometer has some residual strain that normally generates a
built-in fringe pattern, and the influence is involved in inst.. Often this effect is
not negligible, making the interference pattern too complicated for grasping
structural information by eye. One way to separate the effect of the sample from
the characteristics of the interferometer is using a ‘‘phase-stepping’’approach.
This phase-stepping method processes multiple interference patterns observed
by varying the phase difference between the two arms, for instance by using a
tunable phase shifter. When interference patterns

Im ðy; zÞ ¼ a0 ðy; zÞ þ a1 ðy; zÞ cos½ðy; zÞ þ inst: þ 2pm=M;

(8:17)
ðm ¼ 1; 2; :::; MÞ ;

are obtained with M 3, both with and without the sample in place, (y,z) and
inst. can uniquely be processed out from the data set. For phase tomography,
this measurement is repeated while rotating the sample.
Figure 8.5 shows an example [19] where an aluminum screw has been
imaged in an aluminum block of slightly different composition by using
neutron crystal interferometry. From the image without sample (Fig. 8.5a),
we can deduce that the phase of the empty interferometer is reasonably con-
stant over the total projection area of about 10 mm width and 25 mm height.
With sample (Fig. 8.5b), the modulation of the phase induced by the sample is
clearly recognizable.
By essentially taking the difference and correcting for phase-wrapping
errors, one obtains a projected phase map of the specimen, as shown in
Fig. 8.6a. Because the measured phase projection images are of a quanti-
tative nature, phase tomography reconstructions can be performed using
Eq. (8.8). One such reconstruction is shown in Fig. 8.6b, in the form of a
3D rendering of the object (see [19] for more details). Other examples of
the application of neutron interferometry for phase imaging are reported
in [20, 21, 22, 23, 24].
138 F. Pfeiffer

Fig. 8.5 Raw images behind a neutron crystal interferometer of an aluminum screw in an
aluminum block. (a) Reference image without sample and (b) interferometer image with
sample in the beam path (images from [19])

a b

Fig. 8.6 Processed neutron interferometer images of an aluminum screw in an aluminum

block. (a) Projection phase map of the sample after processing the data. (b) 3D rendering of
the reconstructed phase object (images from [19])
8 Neutron Phase Imaging 139

8.3.2 Refraction-Based Techniques

8.3.2.1 Crystal Analyzer-Based Techniques

Similar to the refraction of visible light that is governed by Snell’s law, the
refraction of a neutron beam can be caused by a spatially varying phase shift
induced by a specimen. Thus, by measuring these deflection angles a, a contrast
related to the phase shift gradient (DPC) is generated, which can be expressed
by [25]
l
a/ r ; (8:18)
2p

where r denotes the gradient operator.

The deflection angle caused by refraction is extremely small, and neutrons
are assumed conventionally to go straight through materials. To generate the
contrast, an angular sensor sufficiently sensitive to the neutron beam refraction
is needed, and the incident neutron beam should furthermore be correspond-
ingly collimated. Perfect crystal analyzers are well suited for that purpose
because the angular width of the Bragg diffraction is on the order of
10–100 rad. Figure 8.7 shows a typical arrangement for generating a contrast
based on the refraction of neutron beams using perfect crystals. When a neutron
beam collimated by a crystal passes through a sample, the wave front is
deformed by the phase shift at the sample and the neutron propagation direc-
tion varies depending on the phase gradient. The analyzer crystal placed
downstream of the sample reflects only the neutrons that meet the Bragg
diffraction condition. As a result, a contrast can be seen in the reflected beam.
By changing the angular setting of the analyzer slightly, the contrast can be
varied in a very sensitive way.

Fig. 8.7 A typical setup for

analyzer-based phase
contrast imaging and
tomography

Given the reflection curve R(g) of the analyzer and the complex transmis-
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
sion function Iðy; zÞ=I0 ðy; zÞ exp [i(y,z)] of the specimen, the image
I(y,z,g0) of the beam reflected by the analyzer is given by
140 F. Pfeiffer

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Iðy; z; g0 Þ ¼ Iðy; zÞ=I0 ðy; zÞR½g0  aðy; zÞ ; (8:19)

where g0 is the incident angle without the sample. Similar to the phase-stepping
procedure in single-crystal interferometry, the values a (y,z) can be determined
more precisely by measuring several points along the crystal rocking curve and
subsequent data processing. Tomographic reconstruction of an object from a
data set containing measured projections of g (y,z) is also possible. Since these
projections basically represent gradient projections of the object phase shift, CT
reconstruction can be achieved by applying Eq. (8.11).
An example of neutron phase tomography using such a crystal analyzer-
based method is shown in Fig. 8.8 (from [26]). We see tomographic images of an
aluminum test sample imaged with conventional attenuation-based neutron
tomography (Fig. 8.8a) and in phase tomography (Fig. 8.8b). Because alumi-
num exhibits only a very small neutron capture cross-section, it produces
negligible contrast in absorption. In the phase reconstruction, however, the
internal details are clearly visible. Moreover, it has recently been shown that
these reconstructions are in quantitative agreement with the expected literature
values for the neutron refractive index of materials (see [27] for more details).
Other examples and possible future applications of analyzer-based neutron
phase imaging are reported in [28, 29, 30, 31].

8.3.2.2 Grating Interferometers

Similar to the crystal analyzer-based method described, grating interferometers
can also be classified as a refraction-based technique. The main advantage,
however, is that grating interferometers are considerably more efficient because
they can accept a much larger neutron energy spread and incoming divergence.
Figure 8.9 shows the basic setup of a grating interferometer. It consists of a
source grating G0, a phase grating G1, and an analyzer absorption grating G2.

Fig. 8.8 Reconstructed tomographic slices obtained for an aluminum test sample with holes.
(a) Conventional attenuation-based reconstruction; (b) phase contrast-based reconstruction
(images from [26]). The size of the sample is approximately 25  25 mm2
8 Neutron Phase Imaging 141

Fig. 8.9 (a,b) Principle setup of a grating-based neutron shearing interferometer. (c) The
phase grating (G1) forms a periodic interference pattern in the plane of the analyzer grating. A
phase object in the incident beam will cause a slight refraction, which results in changes of the
locally transmitted intensity through the analyzer

The source grating (G0), an absorbing mask with transmitting slits, typically
placed close to the neutron beam exit port, creates an array of individually
coherent but mutually incoherent sources. Each individual line source provides
enough spatial coherence for the image formation process [32, 33, 34]. As the
source mask G0 can contain a large number of individual lines, each creating a
virtual source, efficient use can be made of typical neutron source sizes of more
than a few square centimeters. To ensure that each of the line sources contributes
constructively to the image formation process, the geometry of the setup should
satisfy the condition p0 = p2 l/d, where p0 (p2) is the period of G0 (G2), l the
distance between G0 and G1, and d the distance between G1 and G2 (Fig. 8.9b).
142 F. Pfeiffer

The DPC image formation process achieved by the two gratings G1 and G2
is similar to Schlieren imaging [25] or to a Shack–Hartmann wave front sensor
[35]. The second grating (G1) acts as a beam splitter and divides the incoming
beam essentially into the two first diffraction orders. Since the wavelength l of
the illuminating neutron radiation ( 1010 m) is much smaller than the grating
pitch ( 106 m), the angle between the two diffracted beams is so small that
they overlap almost completely.
Downstream of the beam-splitter grating, the diffracted beams interfere and
form linear periodic fringe patterns in planes perpendicular to the optical axis
[33, 36]. For a phase grating with a phase shift of p, the periodicity p2 of the
fringe pattern equals half the period of G1. Neither the period nor the lateral
position of these fringes depends on the wavelength of the radiation used.
Perturbations of the incident wave front, such as those induced by refraction
on a phase object in the beam, lead to local displacement of the fringes
(Fig. 8.9c).
The fundamental idea of the method presented here is to detect the local
positions of the fringes and determine from these the phase shift induced by the
object. However, since the spacing of the interference fringes does not exceed a
few microns, an area detector placed in the detection plane will generally not
have sufficient resolution to resolve the fringes, let alone the exact position of
their maxima. Therefore, a grating G2 with absorbing lines and the same
periodicity and orientation as the fringes is placed in the detection plane,
immediately in front of the detector. This analyzer grating acts as a transmis-
sion mask for the detector and transforms local fringe positions into signal
intensity variations. The detected signal profile thus contains quantitative
information about the phase gradient of the object.
To separate this phase information from other contributions to the signal
such as attenuation in the sample, inhomogeneous illuminations, or imperfec-
tions of the gratings, the phase-stepping approach used in other interferometry
methods [3] was adapted to this setup. When one of the gratings is scanned
along the transverse direction yg (Fig. 8.9c), the intensity signal I (y,z) in each
pixel (y,z) in the detector plane oscillates as a function of yg (Fig. 8.10e). The
interferometer phases #ðy; zÞ of the intensity oscillations in each pixel are
related to the wave front phase profile (y,z), the neutron wavelength l, and
the period p2 of the absorption grating by [25]

ld @
#¼ : (8:20)
p2 @y

Since # contains no other contributions, particularly no attenuation contrast,

the phase profile of the object can be retrieved from #ðy; zÞ by a simple one-
dimensional [1D] integration.
Figure 8.11 shows processed transmission (a), DPC (b), and integrated phase
(c) images of a sample comprising titanium (Ti), lead (Pb), molybdenum (Mo),
and magnesium (Mg) metal rods with square profiles (Fig. 8.9c). The results
8 Neutron Phase Imaging 143

Fig. 8.10 Principle of phase stepping. (a–d) Raw image data of a square molybdenum (Mo)
metal rod, taken at different positions yg ¼ y1, . . ., y4 of the analyzer grating G2. (e) Intensity
oscillation in two detector pixels as a function of yg (images from [33])

were obtained at the ICON beam line of the SINQ (Paul Scherrer Institut,
Switzerland). Due to their specific neutron capture cross-sections and incoher-
ent scattering cross-sections, differences in the transmitted intensity through the
rods are observed in Figure 8.11a. Because the vector component of the phase
gradient in the direction perpendicular to the grating lines is constant, but of
opposite signs for both sides of the rods, the rods appear black and white in the
DPC image. It is interesting to note that titanium has a negative neutron
scattering length density [5], and consequently, a negative phase shift is mea-
sured in the material (see Fig. 8.11c). Furthermore, we note that particularly for
weakly absorbing materials, like magnesium and lead (Fig. 8.11a), the DPC
(Fig. 8.11b) and the integrated phase shift (Fig. 8.11c) signal yield a higher
contrast in the image.
As the phase contrast obtained with a grating interferometer can be
described quantitatively, it can further be used for neutron phase tomography.
Figure 8.12 shows one exemplary processed DPC image (a), a 3D rendering of
the reconstructed coherent scattering length density distribution (b), and 2 (out
of 512) reconstructed tomographic slices (c,d) of a sample comprising titanium
144 F. Pfeiffer

a b
10 mm Ti Pb Mo Mg

Fig. 8.11 Linear contour plots and section profiles of processed data for a test object
comprising titanium (Ti), lead (Pb), molybdenum (Mo), and magnesium (Mg) metal rods
with a square profile. (a) Conventional neutron transmission image. (b) Differential phase
contrast image d /d y. (c) Integrated phase shift (y,z) retrieved from d/d y by integration.
(d,e,f) Section profiles through the corresponding image data (images from [33])

Fig. 8.12 Projection images and tomographic reconstruction for a sample comprising lead
(black) and titanium (white) rods with diameters of 6.35 mm. (a) Differential phase contrast
projection image. (b) 3D rendering of the reconstructed coherent scattering length
distribution in the sample. (c,d) 2 (out of 512) reconstructed tomographic slices (images
from [33])
8 Neutron Phase Imaging 145

and lead rods of 6.35 mm diameter. The corresponding CT reconstruction

algorithm used 180 differential phase projections and 1 reference projection
without the sample as input (see Eq. (8.12) and [37]).

8.3.3 Propagation-Based Techniques

In the classical sense, the distance between an object and a neutron image
detection plane should be as short as possible to avoid blurring. However,
under (partially) coherent illumination, Fresnel or Fraunhofer diffraction is
observed at a distance away from an object even with neutrons. Then, an edge-
enhanced contrast is observed without optical elements even for a pure phase
object, as illustrated in Fig. 8.13. The choice of the propagation distance has a
direct influence on the characteristics of the observed fringe contrast produced
by the edges of the object.
The effect can be described by considering the propagation of a wave field
along a certain direction in space (Fig. 8.13). It can be written under the paraxial
approximation by [38, 39]

lx 2
Iðx; y; zÞ ¼ 1 þ r ð0; y; zÞ ; (8:21)
2p

where we have assumed a propagation along the x direction and r denotes the
2D gradient operator acting in the (y,z) plane. This expression can be derived
from the so-called transport of intensity equation [38, 39], assuming a nonab-
sorbing, weak-phase object. The character of the Laplacian in Eq. (8.21)
suggests the generation of the contrast particularly outlining surfaces and
structural boundaries, where the refractive index changes abruptly.
In propagation-based phase contrast radiography, the distance between the
sample and the image detector is large so that the weak lateral deflections due to

Fig. 8.13 Illustration showing the generation of an edge-enhanced contrast of a phase object
due to propagation
146 F. Pfeiffer

nonabsorbing, refractive sample components can form, bringing these fine

structures into sharp relief. For this method, the collimation ratio, or angular
spread of incident beam, is necessarily on the order of 104–105, so that the detail
is not washed out in source size blurring.
The development of phase contrast with propagation is shown in Fig. 8.14.
Here a hollow lead sinker has been mounted on an aluminum screw. In the
contact image (at 20 mm), scattering/absorption within the microcrystalline
structure of the lead leads to little image contrast, but affords no detail. The
aluminum screw is invisible. Images for propagation steps of approximately
220, 450, 900, and 1800 mm are then shown. In the final phase contrast image,
the hollow structure and flaws within the sinker, the aluminum screw, and
shavings in the axial hole from the threading process are all clearly visible
[40]. Other examples of the application of propagation-based phase contrast
imaging with neutrons are reported in [41, 42, 43, 44, 45, 46, 47].

Fig. 8.14 A series of neutron radiographs showing the development of the phase contrast of a
lead sinker mounted on an aluminum screw. The images, counterclockwise from top left to
top right (a–e), correspond to propagation distances of approximately 0.02, 0.22, 0.45, 0.90,
and 1.8 m, respectively. The visibility of the edge detail within the sinker, screw, and shavings
trapped in the axial hole are greatly enhanced (images from [40])
8 Neutron Phase Imaging 147

It is conceivable that a number of these projections could be used to generate

a quantitative tomographic reconstruction based on Eq. (8.15) in a manner
similar to how it is done in the case of X rays [48].

8.4 Coherence Requirements

To generate contrast with phase information, a wave front with a defined phase is
required. This means that a certain degree of coherence is required in the neutron
beam. The coherence can be evaluated perpendicular to the optical axis and along
the optical axis. The first case refers to the so-called spatial or transverse coher-
ence and the second to the temporal or longitudinal coherence [49, 50].
Let us first consider the coherence properties of a source in a plane transverse
to the propagation direction of the radiation. A good measure for the degree of
coherence is the interference contrast recorded in the far-field diffraction pattern
of an object in the beam. Let us take as an object Young’s double slit arrangement
[25]. Illuminating the double slit with a monochromatic point source results in the
far field in a set of interference fringes with fully developed maxima and zero
intensity in the minima, corresponding to a fringe visibility of 100%.
In practice, however, the source is not point-like, but has some lateral
extension. It is easy to see that this reduces the visibility of the interference
fringes. Although the fringe patterns resulting from different points of the
extended source are identical, they emerge at slightly shifted diffraction angles.
Their superposition, therefore, causes the observed pattern to be somewhat
washed out (i.e., the fringe visibility to be reduced). The effect of the extension
of the source can be readily quantified. Let us consider two narrow slits at
distance d, illuminated by a uniform line source of height w (Fig. 8.15).
The interference fringes produced by an infinitesimally small emitting source
element located on the optical axis appear at diffraction angles a = ml /d, with
m = 0,  1,  2, . . . for the maxima and m ¼  12,  32 ;  52    for the minima.
Here, the small-angle approximation has been made. The fringes produced by
an emitting element at the border of the source are found at the shifted angles
ml /d + w/2R, with R the distance between the source and the slits. It is easy to
see that the maxima from a border element of the source coincide with the
minima from the central element at a slit separation d = l R/w. This distance is
defined as the transverse coherence length

lR
t ¼ : (8:22)
w

Considering now a uniform rectangular source with horizontal and vertical

widths wh and wv, which illuminates two pairs of pinholes at right angles at
distances dh and dv, we arrive at the corresponding transverse coherence lengths
h ¼ lR=wh and v ¼ lR=wv .
148 F. Pfeiffer

Fig. 8.15 Diffraction patterns from two narrow slits at distance d, originating from the central
part of the source (solid curve) and from the edge of the source at height w/2 (dashed curve).
The slit distance d is such that the two patterns are in antiphase. This occurs for d¼l R/w.
Angles and distances are not to scale (adapted from [50])

The degree of coherence of the radiation along its propagation direction (i.e.,
its longitudinal [or temporal] coherence) enters the imaging process in a different
way. Let us consider two wave fronts, one at wavelength l and the other at a
slightly different wavelength l+l, which simultaneously depart from a
single-point. Let us assume that after some distance l both wave fronts are in
antiphase (Fig. 8.16).

Fig. 8.16 Propagation of two waves with wavelengths l and l+l. The longitudinal coher-
ence length l is defined as the distance over which the phase difference of the two waves has
become p (adapted from [50])
8 Neutron Phase Imaging 149

If the first wave has made N oscillations over that distance, the second wave
must have made N  12 oscillations. One therefore has Nl ¼ ðN  12Þ (l+l).
Solving for N and substituting in l ¼ Nl we find for this distance

1 l2
l ’  : (8:23)
2 l
The precise prefactor is dependent on the spectral power density of the source;
for a Lorentzian spectrum one has l ’ ð2=pÞ  l2 =l [51].
In practice the intrinsic coherence properties of a neutron spallation or
reactor source are relatively low. This is because such sources typically are a
few centimeters in size and emit a broad polychromatic spectrum. But the
coherence can be improved by adapting the experimental setup accordingly.
As can easily be seen from Eq. (8.22), the transverse coherence can be
increased by inserting a small aperture into the neutron beam at a large distance
upstream from the sample. For example, if we insert an aperture with a 1 mm
diameter at a distance of 10 m upstream from the sample, we will obtain a
transverse coherence length on the order of a few microns for thermal neutrons
with a typical wavelength of a few angstroms. The propagation-based phase
contrast imaging method, for example, typically requires this amount of trans-
verse coherence (see also Section 8.3.3).
As the longitudinal coherence depends on the wavelength spread of the neutron
energy spectrum [Eq. (8.23)], it can be improved by reducing the bandwidth of the
wavelength distribution. This can be achieved, for example, by using a double-crystal
silicon monochromator. With a typical bandwidth of l=l 104 , a longitudinal
coherence length of a few microns can be achieved for thermal neutrons with a
typical wavelength of a few angstroms. Such a beam preparation is necessary, for
example, in the case of the crystal analyzer-based method (see Section 8.3.2.1).
Table 8.2 contains a comparison of the coherence requirements for the phase
imaging methods described above.

Table 8.2 Comparison of the coherence requirements of several phase-sensitive imaging

methods. Note that, strictly speaking, the requirements of a single-crystal interferometer on
the coherence of the incident neutron beam are relatively low. It is more an intrinsic property
of the device that it only accepts the spatially and temporally coherent fraction of the incident
beam
Requirements on Requirements on
Method spatial coherence longitudinal coherence
Crystal High High
interferometer
Crystal analyzer High High
Grating Low Moderate
interferometer
Propagation- High Low
based imaging
150 F. Pfeiffer

References
1. A. F. Kracklauer, On the theory of quanta (a translation of the thesis of de Broglie,
‘‘Recherches sur la théorie des quanta’’), Ann. de Phys. 10, t. III (1925).
2. A. G. Klein, S. A. Werner, Rep. Prog. Phys. 46, 259 (1983).
3. H. Rauch, S. A. Werner, Neutron Interferometry, Oxford University Press, Oxford, (2000).
4. http://www.ncnr.nist.gov/resources/sldcalc.html.
5. V. F. Sears, Neutron News, 3, 26 (1992).
6. M. Zawisky, U. Bonse, F. Dubus, et al., Europhys. Lett. 68, 337–343 (2004).
7. A. Cimmino, G. I. Opat, A. G. Klein, H. Kaiser, Phys. Rev. Lett. 68, 380 (1989).
8. B. E. Allman, A. Cimmino, A. G. Klein, Phys. Rev. Lett. 68, 2409 (1992).
9. G. Badurek, H. Rauch, D. Tuppinger, Phys. Rev. A 34, 2600 (1986).
10. A. G. Klein, G. I. Opat, A. Cimmino, et al., Phys. Rev. Lett. 46, 1551 (1981).
11. A. G. Wagh, V. C. Rakhecha, J. Summhammer et al., Phys. Rev. Lett. 78, 755 (1997).
12. J. Radon, Ber. Verh. Sächs. Akad. Wiss. Leipzig, Math–Nat. 69, 262 (1917).
13. G.T. Herman, Image reconstruction from projections, Academic Press, New York (1980).
14. A. C. Kak M. Slaney, Principles of Computerized Tomography IEEE Press, New York,
(1987).
15. G. W. Faris, R. L. Byer, Appl. Opt., 27, 5202–5212 (1988).
16. U. Bonse, M. Hart, Appl. Phys. Lett. 6, 155 (1965).
17. H. Rauch, W. Treimer, U. Bonse, Phys. Lett. A 47, 369–371 (1974).
18. W. Treimer, Krist. Tech. Crystal Res. Technol. 13, 1105–1116 (1978).
19. F. Dubus, U. Bonse, M. Zawisky, et al., IEEE Trans. Nucl. Sci. 52, 364–370 (2005).
20. G. Badurek, M. Hochhold, H. Leeb, Physica B 234, 1171–1173 (1997).
21. G. Badurek, R. J. Buchelt, H. Leeb, et al., Physica B 335, 114–118 (2003).
22. E. Jericha, R. Szeywerth, H. Leeb, et al., Physica B 397, 159–161 (2007).
23. J. Rehacek, Z. Hradil, M. Zawisky, et al., Phys. Rev. A, 71, 023608 (2005).
24. S. A. Werner, J. L. Staudenmann, R. Colella, Phys. Rev. Lett. 35, 1053 (1979).
25. M. Born, E. Wolf, Principles of Optics, Pergamon Press, Oxford (1980).
26. W. Treimer, M. Strobl, A. Hilger, et al., Appl. Phys. Lett. 83, 398–400 (2003).
27. M. Strobl, K. Staack, W. Treimer, et al., Meas. Sci. Technol. 19, 034020 (2008).
28. M. Strobl, W. Treimer, A. Hilger, Nucl. Instrum. METHODS Phys. Res. B 222, 3–4
(2004).
29. M. Strobl, W. Treimer, A. Hilger, Phys. B, 385–386, 1209–1212 (2006).
30. M. Strobl, W. Treimer, P. Walter, et al., Appl. Phys. Lett. 91, 254104 (2007).
31. W. Treimer, M. Strobl, A. Hilger, et al., IEEE Trans. Nucl. Sci., 52, 386–388 (2005).
32. F. Pfeiffer, O. Bunk, C. Schulze–Briese, et al., Phys. Rev. Lett. 94, 164801 (2005).
33. F. Pfeiffer, C. Grunzweig, O. Bunk, et al., Phys. Rev. Lett., 96, 215505 (2006).
34. F. Pfeiffer, T. Weitkamp, O. Bunk, et al., Nat. Phys. 2, 258–261 (2006).
35. J. Hartmann, Z. Instrumentenkd. 8, 2 (1990).
36. T. Weitkamp, A. Diaz , C. David, et al., OPT. Express, 13, 6296–6304 (2005).
37. F. Pfeiffer, C. Kottler, O. Bunk, et al., Phys. Rev. Lett. 98, 108105 (2007).
38. J. M. Cowley, Diffraction Physics, Elsevier, Amsterdam (1995).
39. M. R. Teague, J. Opt. Soc. Am. 73, 1434 (1983).
40. B. E. Allman, K. A. Nugent, Physica B, 385–86, 1395–1401 (2006).
41. B. E. Allman, P. J. McMahon, K. A. Nugent, et al., Nature 408, 158–159 (2000).
42. J. W. Chen, H. Y. Gao, H. F. Zhu, et al., Acta Phys. Sin. 54, 1132–1135 (2005).
43. F. Fiori, A. Hilger, N. Kardjilov, et al., Meas. Sci. Technol. 17, 2479–2484 (2006).
44. D.L. Jacobson, B.E. Allman, J. McMahon, et al., Appl. Radiat. Isot. 61, 547–550 (2004).
45. E. Lehmann, K. Lorenz, E. Steichele, et al., Nucl. Instrum. Methods Phys. Res. A 542,
95–99 (2005).
46. P. J. McMahon, B. E. Allman, D. L. Jacobson, et al., Phys. Rev. Lett. 91, 145502 (2003).
47. B. Schillinger, E. Calzada, K. Lorenz, Mat. Trans., Proc. 112, 61–71 (2006).
8 Neutron Phase Imaging 151

48. S. C. Mayo, T. J. Davis, T. E. Gureyev, et al. Opt. Express 11, 2289–2302 (2003).
49. J. Als-Nielsen, D. McMorrow, Elements of Modern X–Ray Physics, John Wiley & Sons
Ltd, New York (2001).
50. F. van der Veen, F. Pfeiffer, J. Phys. Condens. Matter 16, 5003–5030 (2004).
51. J. W. Goodman, Statistical Optics, Wiley, New York, (1985).
Chapter 9
Thermal Neutron Holography

B. Sur, R.B. Rogge, V.N.P. Anghel, and J. Katsaras

Abstract X-ray and neutron diffraction techniques have for almost a century
produced results that provide important insights into materials of interest to
a wide range of scientific and technological disciplines. However, traditional
diffraction techniques have their limits, and these limits are best exemplified
by the fact that certain important materials (e.g., integral proteins) are
difficult if not impossible to crystallize — diffraction techniques usually
require high-quality single crystals. Recently developed atomic resolution
X-ray and neutron holography techniques offer the promise to resolve the
structures of difficult-to-crystallize materials to atomic resolution. This chap-
ter will discuss the latest developments in neutron holography and the
challenges that must be overcome to make the technique a viable tool.

Keywords X-ray holography  Neutron holography  Inside source

holography  Inside detector holography  Protein crystallography

9.1 Introduction

Holography can be traced to Bragg’s [1] X-ray work and Gabor’s [2] electron
interference microscope. Although the technique lay dormant for a number of
decades, there has been increasing interest in its development as implied by the
increasing number of publications on atomic-resolution holography using
either electrons [3–7] or hard X rays [8–13]. Moreover, the feasibility of
atomic-resolution thermal neutron holography, either the ‘‘inside source’’ or
‘‘inside detector’’ type, was experimentally demonstrated [14–16], and a theory
was developed showing that the diffraction pattern of plane waves incident on
a sample with a uniformly random distribution of incoherent scatterers

J. Katsaras (*)
National Research Council, Chalk River Laboratories, Chalk River, ON K0J 1J0,
Canada
e-mail: [email protected]

I.S. Anderson et al. (eds.), Neutron Imaging and Applications, 153

Neutron Scattering Applications and Techniques, DOI 10.1007/978-0-387-78693-3_9,
Ó Springer ScienceþBusiness Media, LLC 2009
154 B. Sur et al.

(e.g., 1H in a polymeric material) per unit cell is the same as that from a sample
with a single incoherent scatterer per unit cell [17]. There is interest in the
possibility of using neutron holography to reconstruct, to atomic resolution,
the three– dimensional (3D) structure of materials rich in hydrogen that are
presently difficult to crystallize (e.g., membrane-associated proteins).
Dennis Gabor (1900–1979) developed holography as a technique to improve
the resolution of the electron microscope while working for the British Thomson-
Houston Company (Rugby, England). A patent was filed in 1947 [18], but the
general field did not advance until the invention of the laser [19, 20] (the process
of stimulated emission was first theorized by Einstein in 1917) [21]. The first 3D
holograms were made by Denisyuk [22] in 1962 (reflection hologram) and later
that year by Leith and Upatnieks at the University of Michigan (transmission
hologram) [23]. Since then, optical holography has been extensively used in
optical devices (e.g., holographic lens for heads-up displays), security (e.g., credit
cards), scanners, computer memory storage, medicine, and a variety of other
applications. For example, holographic interferometry is a precise technique for
measuring changes in the dimensions of an object, and digital cameras with
holographic capabilities are able to focus in poor contrast conditions.
For many decades, X-ray diffraction has been relied upon to determine the
3D structure of a number of crystalline materials, the prerequisite being a very
high-quality single crystal. However, even when the requisite sample was made
available, the technique suffered from the so-called ‘‘phase problem’’ — that is,
the phase information was not recorded. As a result, a number of isomorphous
crystals had to be grown to allow for the direct solution of molecular structure —
at increased cost and time. It should also be pointed out that, in some cases,
isomorphous derivatives were impossible to produce. Recently, however, syn-
chrotron sources with their tunable X-ray beams have revolutionized crystal-
lography by allowing the development of multiple-wavelength anomalous
dispersion (MAD), a technique used to obtain reliable phase information.
Now a single crystal serves the purpose of several different metal-containing
crystals and has the added advantage of faster data collection times. Despite
these advances, the need for high-quality crystals remains.
Both holography and crystallographic techniques, such as MAD, solve the
phase problem in the case of 3D materials. Holographic techniques, however,
have the added advantage that only orientational order is required; that is,
translational symmetry is not necessary. In addition to imaging crystalline
materials, holography is also suitable for imaging noncrystalline materials (e.g.,
nematic liquid crystals) or poorly crystallized membrane-associated proteins.
Although the idea of atomic-resolution holography has its roots in Bragg’s
[1] X-ray work and Gabor’s [2] electron interference microscope, atomic-
resolution holography did not evolve until Szöke [24] pointed out that photo-
excited atoms within a sample emit highly coherent outgoing electrons or
fluorescent X rays. The interference between the unperturbed wave from the
photoexcited atom (reference wave) and waves scattered from neighboring
atoms (object waves) thus permitted the realization of atomic-resolution
9 Thermal Neutron Holography 155

holography (Fig. 9.1). Of course, evidence for atomic-resolution holography

existed as far back as the early twentieth century in the experiments of Kikuchi
[25] and Kossel [26].

ik r −r
b(r1 )
i kin ⋅r1 e 2 1
e
r2 − r1
ik R−r
× b(r2 )
e 2

R − r2
r2 = ΨII (R )
ΨI (r2 ) =
ik r −r
b(r1 )
i kin ⋅r1 e 2 1
e
r2 − r1
r1 ΨI (R ) =
ik R−r
b(r1 )
i kin ⋅r1 e 1

e
R − r1

Ψinc (r1 ) = ei kin ⋅r1

Fig. 9.1 (A) Schematic depiction of second-order plane-wave scattering. The incident neutron
plane wave inc interacts with an atom at r1, producing a primary spherical wave (S wave) I.
This primary S wave interacts with a second atom at r2, producing a secondary S wave II.
The interference between the primary and secondary S waves leads to an intensity modulation
(hologram) detected at R. The detected intensity (evaluated  for incident and detected wave
vectors, kin and kout) is given as I ¼ I I þ 2Re I II þ II II . Figure adapted from Sur
et al. [17]. See Ref. [17] or Sect. 9.5 of this paper for details

9.2 Atomic-Resolution Holography

There are two kinds of holographic methods capable of atomic resolution:

inside source and inside detector. The inside source method [24] uses some of
the atoms making up the sample as independent sources of coherent illumina-
tion. In the case of X-ray holography and certain types of electron holography
(e.g., Auger holography), the radiation generated by atomic de-excitation
manifests itself in the form of nearly spherical waves (S waves). This radiation
156 B. Sur et al.

can either pass through the sample unperturbed (reference wave) or be scattered
by the surrounding atoms (object wave). The interference between the reference
and object beams produces the hologram.
In the inside detector method developed by Gog et al. [9, 27], the inside source
holography beam paths can be reversed and the source and detector exchanged. In
other words, an incident plane wave can reach the photoexcitable atom in a crystal
unperturbed (reference wave) or after it has scattered off other atoms in the sample
(object wave). The interference between these two beams dictates the fluorescence
intensity given off by the photoexcitable atom, which is now the detector. When the
incident beam is scanned, the phase relation between the reference and object
waves leads to oscillations of the fluorescent intensity [28] and forms a hologram.

9.3 Neutrons
In 1932, James Chadwick [29, 30] discovered the neutron — first postulated by
Rutherford [31]—a neutral, subatomic elementary particle found in all atomic
nuclei except hydrogen nuclei. A neutron’s mass is similar to that of a proton,
but a neutron possesses a nuclear spin of ½ and a magnetic moment. Neutrons
are stable only when bound in an atomic nucleus; an unstable free neutron has a
mean lifetime of approximately 900 s and decays into a proton, an electron, and
an antineutrino [32].
Even though the neutron’s interaction with atomic nuclei is weak, the
scattering power (cross section) of an atom is not related to its atomic number.
Therefore, neighboring elements in the periodic table can have substantially
different scattering cross sections. It is important to note that the interaction of
a neutron with the nucleus of an atom allows for an element’s isotopes to be
differentiated. An example is the isotopic substitution of 1H (hydrogen) for 2H
(deuterium), a technique commonly used in the study of polymeric and bio-
logically relevant materials. Hydrogen has a negative coherent scattering length
(bcoh = –3.74  10–15 m), lending it ‘‘contrast’’ when surrounded by other,
positive-scattering-length atoms. For biological samples, intrinsically rich in
hydrogen, judicious substitution of 2H (bcoh = +6.67  10–15 m) for 1H
provides a powerful method for selectively tuning the ‘‘contrast’’ of a given
macromolecule. One can therefore accentuate or nullify the scattering from
particular parts of a macromolecular complex by selective deuteration [33]. Of
importance to holography is that 1H has a large incoherent cross section
for thermal neutrons (i = 80 barns), producing S waves similar to the de-
excitation of a photoexcitable atom in X-ray holography. Unlike forms of
electromagnetic radiation that interact primarily with electrons, neutrons inter-
act directly with nuclei and are, to first order, equally capable of imaging light
atoms and heavy atoms. Consequently, atomic-resolution holographic images
of materials comprising entirely light atoms can be obtained, a great advantage
in the study of polymers and biologically relevant materials.
9 Thermal Neutron Holography 157

Neutrons suitable for scattering experiments are presently being produced

either by nuclear reactors (e.g., the 120-MW National Research Universal
[NRU], Chalk River, Ontario) via the fission of uranium nuclei or by spallation
sources (Spallation Neutron Source, Oak Ridge, Tennessee), in which acceler-
ated subatomic particles (e.g., protons) strike a target (e.g., tungsten, lead, or
liquid mercury), spalling neutrons from the target material’s nuclei (for further
details, refer to the instrumentation chapter, which is online at http://
www.springer.com/series/8141).

9.4 Neutron Holography and K Lines

The feasibility of atomic-resolution neutron holography was discussed by Cser

et al. [34] on the basis of the large incoherent thermal neutron cross section
exhibited by 1H atoms. Experimentally, neutron holography was demonstrated
by Sur et al. [14] using a single crystal of natural simpsonite, a rare oxide mineral
of aluminum and tantalum first discovered in western Australia [35] in 1939.
The mineral, with the chemical formula Al4Ta3O13(OH), was examined using
X-ray diffraction [36] in the same year as its discovery; more recently, X-ray
diffraction has been used to determine the crystal structure in the trigonal space
group P3 with unit cell parameters a = 7.386(1) Å and c = 4.516(1) Å [37].
More important, simpsonite contains only one 1H atom per crystallographic
unit cell and was shown to lend itself ideally to holographic reconstruction.
The following year, Sur et al. [15] reported the observation of K lines
(representing, collectively, Kossel and Kikuchi lines) produced by monochro-
matic thermal neutrons interacting with a potassium dihydrogen phosphate
(KDP) single crystal. K lines contain phase information, further establishing the
experimental basis for direct crystallographic phasing of atomic structures
containing incoherent scatterers via thermal neutron inside source holography.
The interaction of single-frequency plane waves with a periodic potential,
formulated in the Born approximation, produces diffraction intensities in
discrete directions commonly known as Bragg reflections. Similarly, single-
frequency S waves interacting with a 3D periodic potential (e.g., an atomic
lattice) give rise to sharp conical intensity variations. Viewed on a surface, the
conic sections appear as lines.
Nearly spherical single-frequency electromagnetic waves are obtained by
electronic de-excitations of atoms (e.g., fluorescence, photoemission) within
the crystal sample itself, an ‘‘inside source,’’ or a source in close proximity to
the sample. In 1922 Clark and Duane [38] predicted the resulting line structures
for X rays, which were experimentally observed using a single crystal of copper
in 1934 by Kossel, Loeck, and Voges [26]. For X rays, such lines are therefore
commonly referred to as Kossel lines. A theoretical explanation was subse-
quently provided in 1935 by von Laue [39]. Spherical waves can also be gener-
ated by dynamical effects (e.g., multiple and inelastic scattering); the resulting
158 B. Sur et al.

lines were first observed by Kikuchi [25] in electron scattering studies using mica
plates of varying thicknesses. Thus such lines are generally referred to as
Kikuchi lines. In the present case, thermal neutron S waves are generated by
incoherent elastic scattering from hydrogen. The fundamental scattering pro-
cess that gives rise to the characteristic lines — S wave interaction with the
crystal lattice — is the same as for Kossel and Kikuchi lines; therefore, they are
collectively referred to as K lines.
The K line intensity pattern depends on the relative position of the periodic
scatterers with respect to the S wave source (i.e., the crystallographic phase of
the scattering structure function). Thus K lines provide the signal for direct 3D,
atomic-resolution imaging techniques (holography) for single crystals via the
inside source or inside detector concept.
Atomic nuclei are, to a very good approximation, point or pure S wave
scatterers of thermal neutron waves. Hydrogen atoms have a large bound, spin-
incoherent cross section, i = 80 barns, for unpolarized thermal neutrons. An
unpolarized, parallel, thermal neutron beam interacting with a single crystal
containing hydrogen atoms should therefore give rise to both Bragg reflections,
from the incident plane-wave neutron beam, and K lines, from the S waves
produced by the incoherent scattering from the 1H atoms.

9.5 Theory
To gain an understanding of the process, consider a single-point S wave source
of amplitude a at the origin surrounded by a distribution of weak point or
purely S wave scatterers described by a coherent scattering length density b(r).
Then, in the Born approximation, the unperturbed or reference wave is

aeijkjjrj
ðrefÞ ðrÞ ¼ ; (9:1)
j rj

and the scattered or object wave is

Z
bðr0 Þeijkjðjr0 jþjrr0 jÞ
obj ðrÞ ¼ a d r0 : (9:2)
jr0 jjr  r0 j
r0

The total wave amplitude, tot for a detector at R is given by ref (R) + obj
(R). The detected intensity, I1S (R), is then *tot tot. Using the far field
diffraction condition (i.e., R >> ^r) and defining the outgoing wave vector in
the detector direction as kout ¼ k R, the detected intensity is then given as

a a
I1S ffi ½1 þ 2Re½ðkout Þ þ jðkout Þj2  ; (9:3)
R2
9 Thermal Neutron Holography 159

where the intensity modulation function is

Z
bðr0 Þeiðkr0 kout r0 Þ
ðkout Þ ¼ d r0 : (9:4)
r0
r0

In the case of inside source holography, the scattering length distribution b(r)
is reconstructed by a Fourier inversion of the observed modulation intensity [14].
To gain an understanding of K lines, consider this expression [Eq. (9.4)]. In
reciprocal space, the modulation function (aside from constants) is given by
Z
BðqÞ
ðkout Þ ¼ 2
dq ; (9:5)
jqj 2kout  q

where B(q) is the Fourier transform of b(r).

In the case of a crystal lattice, the 3D scattering potential is completely
specified by the magnitude and phase of the structure function Fhkl at discrete
points in reciprocal space. The reciprocal lattice points are given by t hkl = hb1
+ kb2 + lb3, where b1, b2, and b3 are three reciprocal lattice basis vectors. The
integral in Eq. (9.5) then becomes a sum over all hkl, namely
X Fhkl
ðkout Þ ¼ 2
: (9:6)
hkl j hkl j 2kout   hkl

For every pair of discrete points  t hkl located within a sphere of radius 2k,
the zeros in the denominator of Eq. (9.6) cause the modulation to go to infinity
in kout directions whose locus is a cone with axis along  t hkl and full opening
angle 2, satisfying the relation 2k cos =  hkl. Viewed on the surface of a
hemisphere of radius kout, the intersections of the cones with the hemisphere are
K line rings centered on t hkl. These K line rings can readily be calculated for any
hkl relative to the crystal coordinate system of a known structure. For example,
a selection of K lines expected for the structure of KDP in a particular sample
orientation is shown in Fig. 9.2A on the surface of a hemisphere kout/| kout| with
| kout| = 4.8 Å–1 [15]. The incident plane wave, kin, will also produce Bragg
peaks, which will appear as discrete points on the surface of the hemisphere,
when condition kout – kin = t hkl is satisfied (Fig. 9.2A).
In a high-resolution experiment, K lines are readily identified by their character-
istic fine structure [27]. However, K line fine structure is affected by factors such as
sample size and shape;
dynamics or higher-order partial wave scattering in both the coherent and
incoherent scatterers, leading to a broadening of the signal and a nonuni-
form reference wave intensity; and
wavelength spread and angular resolution.
All of these effects conspire to mute the sharp modulations characteristic of
K lines, but they still produce modulations along the lines (Fig. 9.2B).
160 B. Sur et al.

Fig. 9.2 (A) Calculated K lines for the various KDP lattice planes for l ¼ 1.30 Å neutrons. In
addition to the K lines, the locations of Bragg reflections are shown as white spots. The
orientation of the crystal with respect to the sphere-of-scattering (kin || kz) is depicted by the
crystallographic axes a, b, and c. (B) KDP holographic data corrected for background,
Debye-Waller factor, and sample shape. K lines appear as bands of scattering
corresponding to the white lines in (A). Figure adapted from Sur et al. [15]

9.6 Reconstruction of the Diffraction Patterns from S Wave

Scatterers

According to Eqs. (9.3) and (9.4), the experimentally observed hologram mod-
ulation function (k) produced by the interaction of a weak S wave coherent
scattering length distribution b(r) with either a single incoherent point scatterer
or with a uniform distribution of incoherent scatterers is given, aside from
constants, by
9 Thermal Neutron Holography 161

Z  
bðrÞ ½iðkrkrÞ  
ðkÞ Re e dr þ O b2 : (9:7)
r
all space

The conditions for pure S wave weak scattering are particularly well satisfied
for unpolarized thermal neutron scattering from nuclei. Additionally, for most
nuclei, the thermal neutron coherent scattering lengths are almost purely real
and have the same sign. The real part of the coherent scattering length distribu-
tion can be directly reconstructed from the measured hologram modulation
function using the following relations.
Z
breal ðrÞ þ breal ðrÞ
b~even ðrÞ ¼ cosðkrÞ / ðkÞ cosðk  rÞdk : (9:8)
kr
constantjkj
Z
breal ðrÞ  breal ðrÞ
b~odd ðrÞ ¼ sinðkrÞ / ðkÞ sinðk  rÞdk : (9:9)
kr
constantjkj

Strictly speaking, for a single wavelength or constant k, the even and odd
parts are reconstructed at nonoverlapping points in space (i.e., the zeros of the
even part reconstruct at turning points of the odd part and vice versa). There-
fore, it is possible to eliminate the conjugate image breal (r) from the recon-
struction of a single hologram data set by a suitable combination of b~odd ðrÞ and
b~even ðrÞ. For instance, b~odd ðrÞ sin(kr) and b~even ðrÞ cos(kr) can be summed with a
moving box average of dimensions l/2 to give b(r). A consequence of this
procedure is that the resolution of the reconstructed image is broadened by
l/2. Alternatively, the quality of the reconstructed image can be improved by
combining holograms obtained at several different wavelengths.
For a single-point incoherent scatterer per unit cell, the above formulation
will reconstruct the coherent scattering length with the incoherent scatterer
located at the origin. For a uniformly random distribution of point incoherent
scatterers, the reconstruction origin will be the ‘‘center of illumination,’’ which
by definition occupies exactly the same volume as the sample.

9.7 Inside Source Neutron Holography

The first inside source neutron holography experiments were carried out at the
N5 triple-axis spectrometer (Fig. 9.3A) located at the NRU reactor at Chalk
River Laboratories, Ontario. Those experiments used 1.3-Å neutrons with a
l/l 1.5% using the (113) reflection from a germanium (Ge) single-crystal
monochromator. The sample-to-detector angular resolution was limited by
distance collimation, defined by a combination of an aperture in neutron-
absorbing cadmium masks and the size of the sample. Sample orientation ()
162 B. Sur et al.

Fig. 9.3 (A) The N5 triple-axis spectrometer on which the first neutron holography
experiments were carried out, located at the National Research Universal reactor, Chalk
River Laboratories. (B) Plan view photograph and schematic of the neutron holography
experimental setup at the N5 triple-axis spectrometer. The incident 1.3-Å neutron beam
(originating from the bottom of the photograph) was parallel to the axis of rotation, . The
angle between the detector and the  axis is denoted by . In order to efficiently obtain the
hemisphere of scattering, the crystal was rotated, in optimal  steps of 28/sin , from 0 to 2p for
a given  (178  838). The lower value for  (178) was dictated as a result of the physical
interference between the scattered beam collimator and the sample rotation spindle. The
scattered neutrons were recorded using a 3He detector. Figure adapted from Sur et al. [14]
9 Thermal Neutron Holography 163

and detector angle () were manipulated in such a way that nearly a hemisphere
of scattering was recorded (Fig. 9.3B). However, as a result of poor scattered
intensity and a single-point detector, collecting good statistical quality data
(total scattering and background) over 1.7p radians in scattering wavevector
(kout) took 10 days (Fig. 9.4A).
The structure of simpsonite contains two layers of oxygen atoms perpendi-
cular to the c axis of the crystal lattice. The first layer, which is closest to the 1H
atom, has an oxygen atom located directly above (positive direction) the 1H

Fig. 9.4 (A) Raw hologram data containing 4,334 28 by 28 pixels plotted on a surface of
constant scattered wave vector magnitude (|kout| = 4.8 Å–1). Because of the experimental
geometry, data collection was limited between 17 and 838 in , and about 13% of the pixels
containing Bragg reflections were excluded from the reconstruction. The remaining data
contained the hologram and -dependent background from the vanadium sample holder
and aluminum wires (e.g., aluminum powder lines at  = 32, 37, 54, and 648) used to secure
the simpsonite crystal to the vanadium pedestal. (B) Reconstructed plane located
approximately + 0.9 Å from the hydrogen atom at the origin showing the seven oxygen
atoms. The plane is roughly coplanar to the basal plane of the crystal. The central spot is the
oxygen atom located directly above the origin. The triangles are used to indicate the two
triplets of oxygen atoms located slightly below this central oxygen atom. (C) Reconstructed
plane located approximately –1.4 Å from the hydrogen atom at the origin showing the
positions of the three oxygen atoms. The distortion of the triangles was attributed to
limitations in the quality of the demonstration data. Figure adapted from Sur et al. [14]
164 B. Sur et al.

atom along the direction of the c axis. This oxygen atom is surrounded by two
triangular sublattices of oxygen atoms slightly below the position of the central
oxygen atom. In the second layer, located below (negative direction) the 1H
atom, the oxygen atoms form triangular lattices about the perpendicular pro-
jection of the 1H atom onto this layer.
Figure 9.4B shows a reconstructed plane, which is approximately coplanar to
the basal plane, at a distance of +0.9 Å from the 1H atom, which is at the origin
of the reconstruction. There is a systematic distortion of the triangular lattices
(Fig. 9.4B) owing to the just adequate counting statistics. Nevertheless, all of
the expected oxygen atoms, six in the two triangles and the central atom, are
present in the reconstruction. Figure 9.4C shows the next closest plane of
oxygen atoms found at a distance of –1.4 Å from the origin. From the recon-
struction, it is easily seen that these oxygen atoms form a triangle of nearest
neighbors, albeit distorted.

9.8 Inside Detector Neutron Holography

The example discussed used a large incoherent neutron scatterer (i.e., hydrogen)
to produce nearly S waves, ‘‘imaging’’ its surrounding atoms. By using
the principle of optical reciprocity, the inside detector concept interchanges
the positions of source and detector. In 2002, Cser et al. [16] demonstrated the
feasibility of the inside detector technique to image lead (Pb) nuclei in a single
crystal of Pb doped with the strong neutron absorber cadmium (Cd). In this
case, Cd atoms act as point-like detectors, with a neutron wave reaching these
detector nuclei either directly (reference beam) or after scattering from the Pb
atoms in the crystal (object beam). The capture of neutrons by Cd results in
the emission of  radiation. The intensity of the  radiation is directly propor-
tional to the neutron intensity experienced by the detector nuclei. When
recorded as a function of sample orientation with respect to the incident
neutron beam, this fluctuating  radiation intensity is effectively the
hologram.
To demonstrate inside detector neutron holography, Cser et al. [16] used a
7-mm-diameter spherically shaped single crystal of Pb0.9974Cd0.0026 (the use of
spherical samples greatly simplifies the data analysis). Cadmium atoms are
strong neutron absorbers with a thermal absorption cross section over four
orders of magnitude greater than that of Pb atoms [40]; this accounts for their
ability to act as efficient detectors. The crystal structure of the Pb0.9974Cd0.0026
alloy is the same (face-centered cubic, fcc) as for pure lead [16], with the Cd
atoms randomly distributed. As Cd was used in low concentrations, on average,
every Cd atom was surrounded by 12 nearest-neighbor Pb atoms.
The (220) reflection from a Cu single crystal was used to select 0.8397-Å
neutrons. Prompt  radiation from the Cd nuclei was recorded using two
scintillation counters shielded by 20 cm of lead. The sample was rotated
9 Thermal Neutron Holography 165

about its axis () in 38 steps through 3508 and tilted (), also in 38 steps,
through 1208. Figure 9.5A shows the reconstruction from the hologram of the
12 Pb atoms surrounding the detector Cd atom plotted in  and  coordinates,
and Fig. 9.5B shows the Cd atom and its 12 neighboring Pb atoms plotted on
the surface of a sphere of radius 3.49 Å. The x-axis is the direction of the
incident neutron beam.

Fig. 9.5 (A) Reconstruction showing the positions of the 12 Pb atoms surrounding the detector
Cd atom plotted in  and  coordinates. (B) The 12 Pb atoms surrounding the Cd detector atom
plotted on the surface of a sphere of radius 3.49 Å. The x-axis denotes the direction of the
incident neutron beam. Data were collected using the D9 diffractometer located at the Institute
Laue-Langevin (Grenoble, France). Figure adapted from Cser et al. [16]
166 B. Sur et al.

The most recent inside detector neutron holography study by Cser et al. [41]
demonstrated for the first time the atomic positions with precision down to
1 picometer. To date, this is the only work based on a neutron holography
experiment that contains information not previously made available by other
experimental techniques.

9.9 Holographic Reconstruction from Multiple

Incoherent Scatterers
The mathematical formulations for atomic structure holography have, until
recently, been limited to samples with one S wave scatterer per unit cell.
However, such samples are extremely rare. On the other hand, materials rich
in hydrogen (e.g., biological or polymeric) are ubiquitous. Recently, Sur et al.
[17] derived a kinematical formulation for the diffraction pattern of mono-
chromatic plane waves scattering from a mixed coherent and incoherent
scattering length distribution. From this second-order kinematical formula-
tion, it became evident that samples with a uniformly random distribution of
incoherent scatterers can be reconstructed to atomic resolution. Also, Sur
et al. [17] provided a formulation for holographic reconstruction eliminating
the so-called conjugate, or twin image problem, inherent in the holographic
inversion of single-wavelength data.

9.10 Holography and Poorly Crystallized Proteins

Proteins are linear heteropolymers that play crucial roles in virtually all biolo-
gical processes. The linear chain folds into an intricate 3D structure, which is
determined by the sequence of amino acids, and is unique to each protein. It is
this 3D structure that is the critical determinant of a protein’s biological func-
tion. In principle, if one has knowledge of the amino acid sequence, it should be
possible to predict the protein’s 3D structure. However, a number of problems
make this ab initio approach very difficult.
Traditionally, protein structures have been determined by X-ray diffraction
(and to a much lesser extent by neutron diffraction) studies of crystallized
samples and more recently, in the case of small (< 25 kDa) proteins in solution,
by 2D and 3D nuclear magnetic resonance spectroscopy. In order to decipher
the 3D structure using X-ray diffraction, well-ordered protein crystals must be
available. However, the availability of quality crystals constitutes a major
obstacle for traditional diffraction methods, particularly for those proteins
integral to the cell’s membrane. As a result, practically all of the approximately
34,000 known 3D protein structures are water soluble as opposed to membrane
proteins. Fewer than 100 membrane proteins have been registered in the RCSB
(Research Collaboratory for Structural Bioinformatics) protein data bank.
9 Thermal Neutron Holography 167

This is a critical shortfall since membrane-associated proteins are known to

constitute approximately one-third of all known proteins [42], and approxi-
mately 40% of the human genome encodes for membrane proteins.
The first membrane-associated protein structure determined to near-
atomic resolution (3 Å) was that of the photosynthetic reaction center of
Rhodopseudomonas viridis, a purple sulfur photosynthetic bacterium whose
3D structure was solved by Deisenhofer et al. [43, 44]. In the case of the
photoreaction center and other membrane-associated proteins, [45–47] deter-
gents were essential to producing high-quality crystals suitable for Bragg and
Laue diffraction studies. At present there does not appear to be a systematic
method for obtaining high-quality crystals for the majority of membrane
proteins. However, low-quality membrane protein crystals, which do not
diffract to high resolution, may be more readily obtainable using the present
crystal growing techniques. The concept in holography is to reconstruct the
local atomic structure in the vicinity of the source or the detector atoms. Thus
long-range translational order (i.e., periodicity) of the protein unit cell struc-
ture is not required for a solvable hologram. The only requirement is that the
unit cells be oriented in the same direction (within an acceptable mosaic
spread, consistent with the unit cell dimensions). Thus poor crystals or qua-
sicrystals are especially useful candidates for the determination of atomic-
resolution 3D structure via holography.
Depending on the protein’s chemistry, thermal neutron and X-ray hologra-
phy techniques may be used to resolve, to atomic resolution, the 3D structure of
proteins forming either poor crystals or quasicrystals. One strategy may be to
start with a low-resolution ( 6 Å data) 3D X-ray structure that can be refined
to atomic resolution using the appropriate X-ray or neutron hologram of the
protein. Assuming that the quasi-protein crystal contains only positional and
not orientational disorder, this strategy, if successful, will enable atomic-resolu-
tion studies of an entire class of macromolecules with structures that are
inaccessible presently and for the foreseeable future. However, there are poten-
tial difficulties.
Until now the mathematical formulations for atomic structure holography
have been limited to samples with one source atom per unit cell (not common).
As mentioned, Sur et al. [17] have mathematically formulated the reconstruc-
tion of samples with multiple source atoms per unit cell. In practice, for thermal
neutron holography, this condition can be realized when the samples used have
a uniformly random distribution of hydrogen atoms, an approach very much
analogous to that of the commonly used diffuser in optical holography. How-
ever, the experimental feasibility of this proposed technique has yet to be
demonstrated. On the other hand, if this approach proves to be feasible, holo-
graphic techniques will drive a worldwide expansion of knowledge about the 3D
structure of membrane-associated proteins and other macromolecules, leading
to better insights into biological processes and improved designs of existing
drugs.
168 B. Sur et al.

9.11 The Future

The technique of atomic-resolution holography is not yet established for rou-

tinely determining atomic structure. It is, however, a promising and technically
feasible method for solving problems that cannot be resolved with traditional
diffraction methodologies. Presently, there has yet to be an atomic-resolution
holography study using a sample that could not be successfully studied using
traditional crystallography (i.e., a poor crystal or quasicrystal).
A recent technological advance challenging the holographic techniques
described is the development of hard X-ray free electron lasers (XFELs), mak-
ing coherent X-ray diffraction imaging a possibility. The first XFEL, the Linac
Coherent Light Source at the Stanford Linear Accelerator Center, is slated to
begin experiments in 2009. The European XFEL Facility [48] (Germany) will
follow in 2013. Unlike third-generation synchrotron sources, which use lenses
to produce coherent X rays, the inherently coherent X-ray beams produced by
XFELs have the potential to image materials with 10 times better spatial
resolution. XFEL sources thus offer the possibility to resolve structural details
of materials to atomic resolution with high-brilliance femtosecond coherent
X-ray pulses. In a study headed by scientists from Lawrence Livermore
National Laboratory, a soft XFEL (l = 320 Å) was used to demonstrate that
a coherent diffraction pattern from a nonperiodic object could be obtained
prior to destroying the sample at 60,000 K [49].

9.12 Concluding Remarks

This brief review has presented the current state of neutron holography.
Neutron holography is a nondestructive, in situ technique capable of atomic-
resolution measurements of biologically relevant materials and appears to be an
excellent option for the solution of poorly crystallized or quasicrystal proteins.
However, an experimental demonstration of ‘‘diffuse source’’ atomic-resolution
holography, utilizing incoherent scattering from the large numbers of hydrogen
atoms in biological materials, is key to future developments. Future neutron
holography experiments will benefit greatly from spallation source instruments
capable of acquiring neutron scattering data with simultaneous large solid angle
coverage and at multiple wavelengths, capabilities that will become available at
the Spallation Neutron Source.

References
1. W.L. Bragg, The X-ray microscope. Nature 149, 470 (1942).
2. D. Gabor, A new microscopic principle. Nature 161, 777–778 (1948).
3. G.R. Harp, D.K. Saldin, and B.P. Tonner, Atomic resolution holography in solids with
localized sources. Phys. Rev. Lett. 65, 1012–1015 (1990).
9 Thermal Neutron Holography 169

4. M. Zharnikov, M. Weinelt, P. Zebish, S M. tichler, and H. -P. Steinrück, First experi-

mental determination of an adsorption site using multiple wave number photoelectron
diffraction patterns. Phys. Rev. Lett. 73, 3548–3551 (1994).
5. M.T. Sieger, J.M. Roesler, D. -S. Lin, T. Miller, and T. -C. Chiang, Holography of
Ge(111)-c(28) by surface core-level photoemission. Phys. Rev. Lett. 73, 3117–3120 (1994).
6. A. Orchowski, W. D. Rau, and H. Lichte, Electron holography surmounts resolution
limit of electron microscopy. Phys. Rev. Lett. 74, 399–402 (1995).
7. T. Matsushita, F.Z. Guo, F. Matsui., Y. Kato., and H. Daimon, Three-dimensional
atomic-arrangement reconstruction from an Auger-electron hologram. Phys. Rev. B 75,
085419(1)–085419(5) (2007).
8. M. Tegze and G. Faigel, Atomic resolution X-ray holography. Europhys. Lett. 16, 41–46
(1991).
9. T. Gog., P.M. Len, G. Materlik., D. Bahr., C.S. Fadley, C. Sanchez-Hanke, Multiple-
energy holography: Atomic images of hematite (Fe2O3). Phys. Rev. Lett. 76, 3132–3135
(1996).
10. M. Tegze and G. Faigel. X-ray holography with atomic resolution. Nature 380, 49–51
(1996).
11. M. Tegze, G. Faigel, S. Marchesini, M. Belakhovski, and A.I. Chumakov, Three dimen-
sional imaging of atoms with isotropic 0.5 Å resolution. Phys. Rev. Lett. 82, 4847–4850
(1999).
12. S. Eisebitt, J. Lüning, W.F. Schlotter, M. Lörgen, O. Hellwig, W. Eberhardt, and J. Stöhr,
Lensless imaging of magnetic nanostructures by X-ray spectro-holography. Nature 432,
885–888 (2004).
13. H.N. Chapman, et al. Femtosecond time-delay X-ray holography. Nature 448, 676–670
(2007).
14. B. Sur, R.B. Rogge, R.P. Hammond, V.N.P. Anghel, and J. Katsaras, Atomic structure
holography using thermal neutrons. Nature 414, 525–527 (2001).
15. B. Sur, R.B. Rogge, R.P. Hammond, V.N.P. Anghel, and J. Katsaras Observation of
Kossel and Kikuchi lines in thermal neutron incoherent scattering. Phys. Rev. Lett. 88,
065505(1)–065505(4) (2002).
16. L. Cser, Gy. Török, G. Krexner, I. Sharkov, and B. Faragó, Holographic imaging of
atoms using thermal neutrons. Phys. Rev. Lett. 89, 175504(1)–175504(4) (2002).
17. B. Sur, V.N.P. Anghel, R.B. Rogge, and J. Katsaras, Diffraction pattern from thermal
neutron incoherent elastic scattering and the holographic reconstruction of the coherent
scattering length distribution. Phys. Rev. B 71, 014105(1)–014105(12) (2005).
18. P.G. Tanner, and T.E. Allibone, The patent literature of Nobel laureate Dennis Gabor
(1900 – 1979). Notes Rec. R. Soc. Lond. 51, 105–120 (1979).
19. A.L. Schawlow, and Townes C.H. Infrared and optical masers. Phys. Rev. 112,
1940–1949 (1958).
20. A.L. Schawlow, and Townes C.H. U.S. patent no. 2,929,222 (March 22, 1960).
21. A. Einstein, Zur quantentheorie der strahlung. Physik. Zeitschr. 18, 121–128 (1917).
22. Y.N. Denisyuk, On the reflection of optical properties of an object wave field scattered by
it. Doklady Akademii Nauk SSSR 144, 1275–1278 (1962).
23. E.N. Leith, and J. Upatnieks, Reconstructed wavefronts and communication theory.
J. Opt. Soc. Am. 52, 1123–1130 (1962).
24. A. Szöke, X-ray and electron holography using a local reference beam, in Short Wave-
length Coherent Radiation: Generation and Applications, eds. D.J. Attwood, and
J. Boker AIP Conf. Proc. No. 147, pp. 361–467 (American Institute of Physics, 1986).
25. S. Kikuchi, Diffraction of cathode rays by mica. Jpn. J. Phys. 5, 83–96 (1928).
26. W. Kossel, V. Loeck, and H. Voges, Die richtungsverteilung der in einem kristall
entstandenen charakteristichen Röntgenstrahlung. Z. Phys. 94, 139–142 (1935).
27. T. Gog, D. Bahr, and G. Materlik, Kossel diffraction in perfect crystals: X-ray standing
waves in reverse. Phys. Rev. B 51, 6761–6764 (1995).
170 B. Sur et al.

28. G. Faigel, and M. Tegze, X-ray holography. Rep. Prog. Phys. 62, 355–393 (1999).
29. J. Chadwick, Possible existence of a neutron. Nature 129, 312 (1932).
30. J. Chadwick, The existence of a neutron. Proc. Roy. Soc. (London) A136, 692–708 (1932).
31. E. Rutherford, Nuclear constitution of atoms. Proc. Roy. Soc. A97, 374–400 (1920).
32. W. Mampe, P. Ageron, C. Bates, J.M. Pendelbury, and A. Steyerl, Neutron lifetime
measured with stored ultracold neutrons. Phys. Rev. Lett. 63, 593–596 (1989).
33. B. Jacrot, The study of biological structures by neutron scattering from solution. Rep
(sp. nov.) from Tabba Tabba, western Australia. J. R. Soc. W. Aust. Prog. Phys. 39,
911–953 (1976).
34. L. Cser, G. Krexner, and Gy. Török, Atomic resolution neutron holography. Europhys.
Lett. 54, 747–752 (2001).
35. H. Bowley, Simpsonite (sp. nov) from Tabba Tabba, western Australia. J. R. Soc. W. Aust.
25, 89–92 (1939).
36. L.E.R. Taylor, X-ray studies of Simpsonite. J. R. Soc. W. Aust. 25, 93–97 (1939).
37. T.S. Ecrit, Ćerný, P., and F.C. Hawthorne, The crystal chemistry of Simpsonite. Can.
Mineral. 30, 663–671 (1992).
38. G.L. Clark, and W. Duane, A new method of using x-rays in crystal analysis. Proc. Natl.
Acad. Sci. U.S.A. 8, 90–96 (1922).
39. M. v. Laue, Die fluoreszensröntgenstrahlung von einkristallen. Ann. Physik 23, 705–746
(1935).
40. V.F. Sears Neutron scattering lengths and cross sections. Neutron News 3, 26–37 (1992).
41. L. Cser, G. Krexner, Markó, M., I. Sharkov, and Gy. Török, Direct observatotion of
local distortion of a crystal lattice at picometer accuracy using atomic resolution neutron
holography. Phys. Rev. Lett. 97, 255501(1)–255501(4) (2006).
42. A. Goffeau, Life with 482 genes. Science 270, 445–446 (1995).
43. J. Deisenhofer, O. Epp, K. Miki, R. Huber, and H. Michel, X-ray structure analysis of a
membrane protein complex: electron density map at 3 Å resolution and a model of the
chromophores of the photosynthetic reaction center from Rhodopseudomonas viridis.
J. Mol. Biol. 180, 385–398 (1984).
44. J. Deisenhofer, O. Epp, K. Miki, R. Huber and H. Michel, Structure of the protein
subunits in the photosynthetic reaction center of Rhodopseudomonas viridis at 3 Å resolu-
tion. Nature 318, 618–624 (1985).
45. M. Garavito and J.P. Rosenbusch, Isolation and crystallization of bacterial porin. Meth.
Enzymol. 125, 309–328 (1986).
46. S.W. Cowan, et al., The structure of OmpF porin in a tetragonal crystal form. Structure
3, 1041–1050 (1995).
47. R. Dutzler, et al., Crystal structure and functional characterization of OmpK36, the
osmoporin of Klebsiella pneumoniae. Structure 7, 425–434 (1999).
48. I.A. Vartanyants, I.K. Robinson, I. McNulty, C. David, P. Wochner, and Th.
Tschentscher, Coherent X-ray scattering and lensless imaging at the European XFEL
Facility. J. Synch. Rad. 14, 453–470 (2007).
49. H.N. Chapman, et al., Femtosecond diffractive imaging with a soft-X-ray free-electron
laser. Nat. Phys. 2, 839–843 (2006).
Chapter 10
Novel Imaging Techniques: Polarized Neutrons
and Neutron-Based Magnetic Resonance Imaging

N. Kardjilov, W.T.H. Lee, and G.E. Granroth

Abstract This chapter describes three rather novel neutron-based imaging

methods all of which involve magnetic fields. The first two take advantage of
the neutron magnetic moment which makes it sensitive to the presence of
magnetic fields. In the first application the Larmor precession of the neutron
in a magnetic field is used to produce two- and three-dimensional (3D) visua-
lization of magnetic field distributions in free space and in bulk materials.
In the second potential application, the Larmor precession is used as a
measure of the material density through which the beam passes. Although
this technique of spin contrast imaging is still very much in the concept stage,
initial estimates show that it is feasible with new sources and may open up the
possibility of obtaining 3D images without moving the sample.
Polarization of the nuclei in a material can also provide unique ways of
imaging a system. By partially polarizing the nuclei with a sufficiently large
magnetic field, the Zeeman splitting of the nuclei can be detected by high-
resolution neutron spectroscopy. If the field is then reduced, the time for the
Zeeman splitting to relax to the equilibrium value is a measure of the nuclear
relaxation time; this is the physical quantity measured in magnetic resonance
imaging. A full image of the relaxation times with respect to position could then
be obtained by rastering the sample in the neutron beam.

Keywords Polarized neutrons  Magnetic fields  Magnetic resonance imaging 

Spin-echo  Time-of-flight  three-dimensional imaging

10.1 Neutron Spin Polarized Imaging of Magnetic Field

Neutrons are sensitive to magnetic fields due to their magnetic moment, i.e., spin.
Therefore, similar to the conventional attenuation contrast image, the magnetic
field inside and around a sample can be visualized independently by detection of

N. Kardjilov (*)
Helmholtz Centre Berlin for Materials and Energy, Berlin, Germany
e-mail: [email protected]

I.S. Anderson et al. (eds.), Neutron Imaging and Applications, 171

Neutron Scattering Applications and Techniques, DOI 10.1007/978-0-387-78693-3_10,
Ó Springer ScienceþBusiness Media, LLC 2009
172 N. Kardjilov et al.

the polarization changes in the transmitted beam. Polarized neutron radiography

is based on the spatially resolved measurement of the final precession angles of a
collimated and polarized monochromatic neutron beam that is transmitted
through the magnetic field, which is present inside and outside of a sample.
This new technique can be described using theoretical considerations related
to the interaction of the magnetic moment of the neutron with the magnetic field.
~ of the
~ the magnetic moment (i.e., the spin S)
In the presence of a magnetic field B,
neutron will perform Larmor precession with a frequency !L . For a monochro-
matic neutron beam having a uniform neutron velocity v, the precession angle ’ is
~ along a certain path:
proportional to the integral of the magnetic field B

Z
g
’ ¼ !L t ¼ L Bds (10:1)
v
path

where gL is the neutron gyromagnetic ratio. ’ can be measured experimentally

using a spin polarizer-analyzer arrangement as shown in Fig. 10.1. Here the
neutrons are first polarized (Chapter 3) and, after the interaction with the
magnetic field, the final spin orientation is analyzed with respect to the initial
state. The signal transmitted through the spin analyzer depends on the spin
rotation angle (Eq. 10.1) and has a maximum for parallel and a minimum for
anti-parallel spin orientation. In this case the detected image behind the analy-
zer is determined by a superposition of conventional attenuation contrast
Ia(x,y) and the contrast variations due to spin rotation Im(x,y):
Z
1
Iðx; yÞ ¼ I0 ðx; yÞ  expð ðsÞdsÞ ð1 þ cos ’ðx; yÞÞ (10:2)
2
|fflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
path
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} Im ðx;yÞ
Ia ðx;yÞ

Fig. 10.1 Schematic of the set-up used for imaging magnetic materials on CONRAD. The
neutron beam is first polarized, then precesses around a magnetic field is analyzed and finally
detected. Note how the intensity (represented by the height of the dark arrow) behind the
analyzer is smaller than the intensity behind the polarizer [1]
10 Novel Imaging Techniques 173

where I0(x,y) is the incident beam intensity, (x,y) is the linear attenuation
coefficient of the sample and (x,y) are the coordinates in the detector plane. The
cosine implies a periodic transmission function for the analyzed precession
angles and complicates a straightforward quantification with respect to the
traversed magnetic fields. By use of a calculation model based on the Biot-
Savart law, it was shown that quantitative analysis is possible [1].
Imaging experiments using the setup shown in Fig. 10.1 were carried out at
the neutron tomography facility CONRAD at HMI [1]. For this purpose the
instrument was equipped with solid-state polarizing benders [2] providing a
beam with a cross-section of 15 mm width and 45 mm height. For investigation
of larger samples, up to 20 cm width, a scanning arrangement was adapted. A
double crystal monochromator [3] was used to select a defined wavelength from
the cold neutron spectrum. The spatial resolution achieved in the radiography
images was around 500 mm for the given experimental geometry.
The potential of the method was demonstrated by visualization of the
magnetic field around a simple dipole magnet (Fig. 10.2). This shows the

Fig. 10.2 A radiograph showing the field lines surrounding a bar magnet. The magnetic field
decreases in strength with distance from the magnet, resulting in a series of maxima and
minima, where the beam polarization is sequentially parallel or antiparallel to the analyzer.
Very close to the magnets (where the field is strongest) the field lines are too close together to
be spatially resolved
174 N. Kardjilov et al.

decay of the magnetic field strength with increasing distance from the
magnet, resulting in an annular structure around the sample with an increas-
ing period due to the changing precession angles of the neutron spins on
their path through the strongly decaying field. The gray level scale used in
the images is related to the intensity variations induced by the sample and
the magnetic field (from black = minimum to white = maximum, repre-
senting the periodic 2p rotation of the neutron spin). The structure in the
center part cannot be seen due to the limited spatial resolution of
approximately 500 mm [4].
This imaging method can in some instances be extended into three dimen-
sions by a standard tomographic technique. Figure 10.3 shows the distribution
of a magnetic field trapped inside a polycrystalline lead cylinder, with a dia-
meter of 1 cm and length of 3 cm, that becomes superconducting when cooled
below the critical temperature, Tc = 7.2 K. The sample was cooled down to
6.8 K in a homogenous magnetic field of 10 mT. After this the magnetic field
was switched off, resulting in partially trapped magnetic fields in the

Fig. 10.3 A 3D reconstruction of trapped flux (solid regions left and right) inside a
polycrystalline cylinder of lead. When cooled to below its critical temperature (7.2 K) in the
presence of a weak magnetic field, some flux is present inside due to defects and grain
boundaries and this remains trapped even after the field is switched off [1]
10 Novel Imaging Techniques 175

superconductor due to grain boundaries and other defects [5, 6]. The tempera-
ture dependence of the residual field distribution inside the sample was visua-
lized by recording radiographic images during the heating process. The images
(Fig. 10.3) show an inhomogeneous residual field which decreases during heat-
ing and vanishes completely when the critical temperature Tc is reached. For the
weak trapped residual field at 7.0 K, a tomographic investigation was per-
formed by rotating the sample around its vertical axis. The beam attenuation
for each pixel can be related to absorption and magnetic contrast (Eq. 10.2)
assuming that the trapped magnetic field conserves its main orientation per-
pendicular to the beam polarization and is weak enough to cause spin rotations
smaller than p for all recorded projections. The volumetric data set was recon-
structed from 60 2D images, collected through rotation over 1808,
using a filtered back projection algorithm (Chapter 6). The results show the
three-dimensional (3D) representation of the flux trapped in the sample. Flux
concentrations could be found close to the end surfaces of the cylinder and at
the position where the sample was held with a screw.
This imaging method has many potential applications. The presence and
controlled application of magnetic fields are essential in many fields of science
and technology as well as in fundamental physics. For example, the flux
distribution and flux pinning in large superconducting samples, the skin effect
[7] in conductors, or magnetic domain distributions in bulk ferromagnets could
be visualized and studied in detail.

10.2 Spin Echo Imaging

The Larmor precession described in the previous section forms the base of the
neutron spin-echo technique which is typically used to determine small energy or
momentum changes upon scattering [7], see also Neutron Scattering Instrumen-
tation [8]. In a typical spin-echo setup (Fig. 10.4), two identical regions of
magnetic field are applied before and after a ‘‘p-flipper.’’ A neutron with polar-
ization perpendicular to the field undergoes Larmor precession in the first field
region and its spin rotates by an angle of  radians. Due to the energy spread of
the neutron beam, the spins of different energy neutrons acquire different pre-
cession angles and the neutron beam appears to be depolarized after the first field
region. The p-flipper between the two magnetic field regions rotates the neutron
spins about an axis perpendicular to the applied field by 1808. This has the
effect of reversing the angle of the neutron spin from  to –. Since the two
magnetic field regions are identical, the neutron spin rotates an additional
angle of  radians in the second field region. The spin precession in the two
field regions cancels and the beam ‘‘recovers’’ its polarization after going through
the field regions. This setup constitutes a neutron spin-echo instrument, and
the process through which the neutron beam recovers its polarization is called
‘‘spin echo.’’
176 N. Kardjilov et al.

Fig. 10.4 A basic neutron spin-echo instrument setup. The spin of a neutron undergoes
Larmor precession in the upstream magnetic field. A p-flipper between the magnetic field
regions inverted the spin rotation angle from  to –. The neutron spin rotates an additional
angle equal to  in the identical downstream field region and returns to its starting direction

In neutron spin-echo spectrometers, a sample is placed between the two

magnetic field regions in the neutron flight path. Inelastic scattering of the
neutron beam changes the energy (velocity) of the neutrons, so that the times
that a neutron spends in the two field regions are no longer equal. Since the spin
precession is proportional to the time spent in the magnetic field, this prevents
the beam from completely recovering its polarization. The rate of depolariza-
tion is proportional to the difference in the times spent in the two field regions;
hence the neutron energy change due to inelastic scattering by the sample can be
determined.
In a previous development [9], experiments were carried out to measure the
neutron optical potential in nonmagnetic materials by applying the neutron
spin-echo technique. When a neutron beam goes through a material without
being scattered, it experiences an optical potential and changes its momentum.
In neutron optics, the optical potential manifests as the neutron index of
refraction n,

Nbl2
n¼1 ; (10:3)
2p

where N is the number density of the material, b is the coherent scattering

length, and l is the neutron wavelength. The velocity of a neutron in the
material is related to its velocity in vacuum v0 by v ¼ nv0 . Due to the change
in the neutron velocity, neutrons that go through a material along a direction x
over a distance X acquire an additional spin angle , given by
Z X Z X Z X
1 1 m n gL 3
¼ gL B dx  gL B dx ¼ Bl NðxÞbðxÞdx ; (10:4)
0 nðxÞv0 0 v0 2ph 0

where mn is the neutron mass, h is the Planck’s constant, and b is the coherent
scattering length. The quantity N(x)b(x) is more commonly known as the
neutron scattering length density. The magnitude of the neutron scattering
length density ranges from 10–7 Å–2 to 10–5 Å–2. Hence the spin angle change
 measures the path integral of the scattering length density.
In a modification of the neutron spin-echo instrument discussed above,
placing a material in one arm of the pair of magnetic field regions results in a
10 Novel Imaging Techniques 177

net neutron spin angle that corresponds to the path integral of the scattering
length density. With a proper neutron beam arrangement, this effect can be
applied to measure the optical potential as a function of position in a material –
an imaging technique. A typical benchmark for an imaging technique is a 1 mm
spatial resolution. A common neutron polarization analyzer such as a polarized
3
He analyzer (Chapter 3 in this book) may resolve a spin angle of 108. If the
magnetic field strength is strong enough to rotate the neutron spin by 108 over
1 mm of its flight path, a 1 mm imaging resolution in the direction along the
beam path can be achieved. Using a typical scattering length density of
Nb=10–6 Å–2, Fig. 10.5 shows the magnetic field required to reach = 108
over a 1 mm flight path as a function of the wavelength. A potential imaging
device using this spin-echo technique would therefore use cold neutrons with
wavelengths ranging from 20 to 100Å and an applied field from several gauss
for 100Å neutrons to 1 Tesla for 30Å neutrons. The magnetic potential energy
U ¼ ~ ~ where 
  B,  is the neutron magnetic moment, is equal to 0.06 meV in a 1
Tesla field. Hence, compared with a kinetic energy of 8.181 meV for 100Å
neutrons, the optical effect of the magnetic potential can be ignored.
One main difference between the imaging application and the neutron spin-
echo spectrometer is that in imaging, the additional change in the spin angle
occurs only when the neutron is inside the material, whereas the change accu-
mulates over the entire field region after the sample in the case of spin echo
spectrometer. This makes the spin angle change insensitive to the beam diver-
gence. The beam divergence, however, is still important in that neutrons
through different parts of a sample may end up striking the same detector
pixel. This blurring of the image is, however, a common concern for all imaging

Fig. 10.5 Applied field strength necessary to achieve a 1 mm resolution

178 N. Kardjilov et al.

techniques rather than being specific to the spin contrast imaging. In neutron
imaging, collimation is used to reduce the divergence and anti-grids can be used
between the sample and the detector system. The same setup can be applied in
the techniques discussed here.
To acquire a three-dimensional image of an object using imaging techniques
based on direct neutron absorption, measurements with the neutron beam
passing the object at different angles are first carried out and tomography
reconstruction is then applied to render the 3D image. The advantage of the
imaging technique discussed here is that it can be used with a time-of-flight
(TOF) technique to achieve 3D imaging without the need to measure the object
at many different angles and without using tomographic reconstruction.
The idea behind the technique is illustrated in Fig. 10.6. Initially the neutron
spins are aligned parallel to the applied field. At time t=0, a p/2 pulse is applied
over the length of the sample to rotate the neutron spins to a perpendicular
direction with respect to the field, which starts the precession of the neutron
spins. Along the neutron flight path, neutrons at different position along a flight
path go through a different path length through the material and therefore
acquire a different spin angle change . For instance, the spin change angle is 1
for neutrons starting at point 1 and 2 for neutrons starting at point 2, respec-
tively. The spin angle change is converted to intensity change by a polarization
analyzer before the detector. Since neutrons at point 1 at t=0 reach the detector
later than neutrons at point 2 at t=0, by recording the time-of-flight of each
neutron at the detector, the position of each neutron at t=0 can be determined.
Subtracting 2 from 2 gives the spin angle change that corresponds to the
scattering length density from point 1 to point 2. Applying this analysis to a
series of images taken at consecutive times-of-flight, a slice-by-slice 3D distri-
bution of the scattering length density of an object can be obtained. The speed
of 30Å neutrons is 131.87 m/s. It takes 7.6 ms to travel a distance of 1 mm. A
timing resolution of 1 ms would be sufficient to resolve the initial location of the
neutrons at the time the precession is started.

Fig. 10.6 Technique for direct 3D imaging. At time t=0, a p/2 flip starts the neutron
precession; neutrons at point 1 precess through a longer path in the sample than neutrons at
point 2. Subtracting the two gives the precession angle from point 1 to point 2
10 Novel Imaging Techniques 179

In order to demonstrate the technique, a Monte Carlo simulation was carried

out that takes into account bandwidth, beam divergence, and placements of the
neutron source and neutron detectors. The sample model is a 1  1 cm plate. It
is 2 mm thick in the beam direction with one-quarter of the sample cross-section
only 1 mm thick. A nominal scattering length density Nb=10–6 Å–2was used for
the calculations. The field strength at the sample was 1 Tesla. For the purposes
of the simulation a 5 mm diameter, isotropic neutron source located at 4 m from
the sample was assumed to produce a 30Å neutron beam with a typical wave-
length spread Dl/l=10%. This wavelength spread allows for both 1 mm reso-
lution and sufficient neutron intensity. A detector was located 1m away from
the sample. For simplicity, only the additional amount of spin precession in the
material was computed instead of the spin precessions in the two magnetic field
regions. Also for simplicity, the incident neutron polarization efficiency was
assumed to be 1 and the efficiency of the polarization analyzer was also
considered to be ideal. In a real instrument, the polarization and the efficiency
of the analyzer is typically 95–98%. The neutron intensity after the analyzer is
given by I=cos2(). In this setup, 11.58 for a passage of 1 mm through the
material.
The intensity profile at the image plate detector for different TOF neutrons
is shown in Fig. 10.7. The upper-right corner is the region where the material
is thinner (1 mm thick compared with 2 mm for the rest of the sample). In
Fig. 10.7(a), the neutrons began to precess before entering the sample. In
Fig 10.7(b), the neutrons are half-way through the sample when the precession
begins. Due to the finite size of the source and the beam divergence, there is
blurring of the edges of the modeled sample. The diverging beam also gives rise
to an amplified image. To further evaluate the simulation results, the intensity
across the x-direction of the image plate at y=+3 mm to +3.2 mm was
extracted to calculate the spin angle change . The intensity results are
shown in Fig. 10.7(c) and the spin angle change results are shown in Fig.
10.7(d). The simulation verifies that  238 for neutrons passing the 2 mm
thick area and  11.58 for those passing the 10 mm thick area. Subtracting
the two spin angle changes gives the spin angle change through the first 1 mm
thick area and it gives the expected  11.58 as shown in Fig. 10.7(d). This
demonstrates that, by extracting the spin angle change  from the intensity
map at different TOFs, the distribution of the scattering length density in an
object can be mapped in a slice-by-slice fashion to give a 3D rendering of the
composition.
There are several technological challenges to reaching 1 mm resolution with
this method. The first is the requirement to flip the neutron spins by p/2, i.e., from
being parallel to being perpendicular to the 1 Tesla field. The Larmor precession
frequency for a 1T field is 29.164 MHz. The neutron spin will rotate 908 in 8.6 ns.
To avoid a broadening of the radio-frequency pulse spectrum, a pulse of 2 np +
p/2 can be used. However, a setup accurate to 0.5 ns in the application of the pulse
is needed to reach accuracy better than 58 in the p/2-flipping. A second challenge
is the precision required to cancel the spin precession (the spin-echo point) using
180 N. Kardjilov et al.

(a) (b)

(c) (d)

Fig. 10.7 Neutron intensities at the detector if spin precession starts (a) when the neutrons are
right before the sample, and (b) when the neutrons are passing the middle of the sample. The
color scales are relative; (c) The normalized integrated intensity at the detector along the
x-direction of the detector from y=+2 mm to y=+3 mm from the two detector images
shown in (a) and (b); (d) The spin angle change  calculated from the two time-of-flight
intensities

both high magnetic field and very cold neutrons: From Eq. (10.1) above, 30Å
neutrons in a 1 T field precess 221 times every 1 mm flight path. If identical
upstream and downstream field regions are used, the location where the p-
flipping occurs will need to be accurate to 0.1 mm for a 108 accuracy. This
would be a formidable task using currently available technology. However,
only a small region that covers the sample will need the 1 T field. The rest of
the field regions can use a field that can be substantially lower. This is because the
spin precession angle depends upon both the field strength and the path length
along which the field is imposed. A smaller field over a long distance can result in
the same spin rotation as a larger field over a shorter distance. If a nominal 1 T
field is applied over a 1 cm length where the sample is located and the rest of the
setup is in a 20 gauss field, the second field region can extend 5 m and require an
10 Novel Imaging Techniques 181

accuracy of 0.06 mm in the p-flipping location for a 108 accuracy. The third
challenge results from the fact that the precession of the neutrons starts at different
locations along the beam. Hence either the downstream p/2 pulse or the p-flipping
will need to be applied to different locations along the flight path. This would
require a more sophisticated p-flipper setup than those currently used at neutron
beam lines, which typically apply p-flipping over a small distance along the beam.
Hence although we have shown that spin contrast tomography is theoreti-
cally feasible, there are some interesting technical challenges to be overcome
before the method can be applied in practice.

10.3 Prospects for Neutron-Probed Magnetic Resonance Imaging

Magnetic resonance imaging (MRI) is a powerful tool used primarily for

internal imaging in biological systems. It is basically a position-dependent
map of the nuclear spin relaxation time. There are many other systems that
one would like to image using MRI, but they have a negligible penetration
depth for radio frequency (RF) radiation. Therefore traditional MRI techni-
ques cannot be used, and an alternative method to perform the nuclear mag-
netic resonance (NMR) measurement that is the core technology behind MRI
must be found. The main classes of samples that are of interest are metallurgical
samples, hydrogen storage materials, and biological samples in ionic salt solu-
tions. As discussed elsewhere in this book, many of these systems can be
addressed with conventional neutron radiography. Where MRI brings addi-
tional information is in systems where there are magnetic interactions of inter-
est. For example in a magnetic alloy, MRI would provide a position-dependent
measure of the interaction between one magnetic region and its environment,
where conventional imaging only provides the positional dependence of the
constituents of the alloy.
Neutrons are a promising probe to overcome the aforementioned difficulty.
They penetrate deeply into materials. However, neutrons present their own
challenges. Several of these are discussed. First and foremost is performing
the NMR measurement.
Fundamentally, NMR and neutron spectroscopy are MHz and THz techni-
ques, respectively. Therefore, although the Zeeman splitting of nuclear mag-
netic moments can be observed with conventional spectroscopy, the additional
fine splittings of interest in NMR measurements cannot. Alternatively, a time-
dependent measurement can be used to measure the decay of a nuclear spin (t)
and has been demonstrated in some limited cases. These studies used low
temperature (mK) [10], or microwaves [11], to saturate the nuclear spins.
These oriented spins changed the intensity of Bragg peaks. Then studies the
temperature was raised, or the microwaves turned off, and the intensity decay of
the Bragg peak was recorded as a function of time. The resultant decay is
proportional to t. However, mK temperatures are difficult to attain, would
182 N. Kardjilov et al.

freeze liquid samples, and the microwaves suffer from the same penetration
depth problems as the lower-frequency RF. An alternative approach to mea-
suring the decay time must be found to make this technique accessible to more
types of samples. A potential solution is to use high magnetic fields.
With the availability of high magnetic fields, a simple extension of the
aforementioned techniques would be to use the field to raise the temperature
necessary for the saturation of the nuclear spins. Figure 10.8 examines this
possibility by looking at the nuclear polarization as a function of temperature
for different fields. As can be seen from the figure, T  10 mK is required for the
current state of the art field for neutron-scattering experiments of 15 T. If the
field was increased to 55 T, the required temperature is in the hundreds of mK
range. Such an increase makes the measurement simpler but does not relieve the
freezing problem. To use this technique at room temperature requires  105 T,
an unrealizable static field.
So the main challenge is to find another way to measure t. The approach
described below eliminates the need for microwaves and/or mK temperatures.
The general idea is to use spectroscopy to measure the Zeeman splitting of
relaxing nuclear spins at a high field, then quickly drop the field to a lower value
and watch the Zeeman splitting relax to the equilibrium value at the lower field.
A large static field is required to sufficiently separate the peak associated with
the Zeeman splitting from other incoherent processes. Protons are the nuclei of
choice for this study, as they are the most familiar NMR active nuclei. The
protons are assumed to be provided by a water sample at room temperature.
Therefore the nuclear relaxation time is assumed to be 200 ms. A static field of

1.0

0.8
H = 5x105 T
Polarization

H = 4.5T
H = 15T
H = 55T

0.6

0.4

0.2

0.0

10–4 10–3 10–2 10–1 100 101 102 103

T(K)

Fig. 10.8 The normalized proton polarization as a function of temperature for different
values of magnetic field
10 Novel Imaging Techniques 183

15 T is sufficient for this system. Extending this technique to significantly more

nuclei would require 2–4 times the static field, e.g.,  40 T [12]. The simulations
detailed below are for a 10 T pulsed field on top of the 15 T static field.
To demonstrate the feasibility of this technique, one must show that, with
current magnet technology, the field can cycle fast enough to allow for mea-
surement of the nuclear relaxation time. The current state of the art in pulsed
magnetic fields is the long-pulsed magnet at the National High Magnetic Field
Laboratory in the United States [13]. A calculation was performed to observe
the response of the resonance energy of the aforementioned proton system to a
10 T pulse (solid curve in Fig. 10.9) having the shape of the peak of the long
pulse magnet [14]. The dashed curve is the resultant solution to the Bloch
equation [15]. The equation was solved numerically with MathCAD [16].
This resonance energy will be the center value for the splitting of the scattering
away from zero energy transfer. Given the dashed curve, a pulsed neutron
source, operating at 60 Hz, will provide 80 pulses over the observed nuclear
decay time.

Fig. 10.9 The resonance energy for a system of hydrogen nuclei (dashed curve) for a given
applied field (solid curve)

An estimate of the neutron count rate is also necessary to demonstrate the

feasibility of this type of experiment. Approximations of the cross-sections for
scattering between spin states were estimated elsewhere [17]. The assumed flux
on a 40 mm3 sample is of the order 105 n cm–2 s–1. The scattering from the
sample will illuminate 0.379 sr of detector area. The energy resolution of the
instrument close to the elastic line is 3 meV. (Detailed balance and absorption
are ignored in these calculations.) Given these assumptions, an example trace is
simulated and shown in the inset of Fig. 10.10. A similar trace is calculated
every 16.7 ms out to 1.7 s after the field is applied. The uncertainty in the curve is
simulated Poisson statistics. To obtain sufficient counting statistics, each point
184 N. Kardjilov et al.

8.5
2000

8.0 1500

cnts
Resonance Energy (µeV)

7.5 1000

500
7.0
0
–30 –20 –10 0 10 20 30
6.5
E(µeV)
simulated data
6.0
fit

5.5

5.0
0.4 0.6 0.8 1.0 1.2 1.4 1.6
t(s)
Fig. 10.10 The resonance energy determined from the simulated spectra. The resonance
energy is fitted to an exponential decay to determine the relaxation time. The inset shows an
example simulated spectrum (circles) and a fit (solid)

on the decay curve should be counted for a total of 10 s. Therefore the measure-
ment would be repeated many times and the aggregated results analyzed. In
practice this means a  0.6 Hz pulsed magnet operated for 600 pulses. A full
relaxation time measurement would require  20 min.
Each of these resultant spectra are analyzed by fitting 3 Lorentzians of 1 meV
width, convoluted with the instrumental energy resolution of 3 meV, at each
time interval. The parameter to be tracked, as a function of time, to determine
the relaxation time is the splitting of the finite energy transfer peaks from the
center. The results from the simulated data for this parameter and its uncer-
tainties are plotted as points in Fig. 10.10. An exponential decay is then fit to the
points to determine the relaxation rate. The specific function used for this fit is
E ¼ Aet=t þ E0 with E0 the field splitting at the lower field value and A a
normalization constant. The results are as follows: t = 196  5 ms, A = 25 
2 meV and Eo = 5.33  0.02 meV. The fitted relaxation or Zeeman splitting time,
t, is in excellent agreement with the input time of 200 ms. Therefore, simulations
show that neutron measurements of the relaxation time could be performed by
watching the timed decay of the Zeeman splitting.
10 Novel Imaging Techniques 185

To test the limits of this method, similar calculations were performed with a
10 T static field. However, for this configuration the Zeeman splitting was too
small at the longest times to be resolved from the central incoherent peak. This
resulted in an artificially large value of t from the fit. In principle the t
measurement will be influenced at longer times by the spins already flipped at
shorter times. However, less than 106 spins are flipped before the next field pulse
is applied, which is a negligible quantity when compared with the total number
of spins in the system.
In order to do this measurement, a high-resolution spectrometer such as
BASIS at the U.S. Spallation Neutron Source (SNS), a high (static) field
magnet in excess of 30 T such as that proposed for Zeemans at the SNS, and a
10 T pulsed magnetic field are required [18, 19].
Though the core technology behind neutron-probed MRI has been
described, other significant challenges remain. Position encoding is one of
them. In traditional MRI a field gradient is placed across the sample [20, 21].
Then the resonance condition is only met at one spot in the sample. The field
gradient is then varied to change this location. A field gradient would only
smear out the neutron-probed measurement as it measures all energies. Instead,
we would raster the sample through the beam so only a single-point is measured
at a time. We assume the state of the art pixel size of 1  1 mm to 0.1  0.1 mm,
as defined for the Vulcan instrument at SNS [22]. An additional restriction is
that the bore in the proposed magnet [19] is only 50 mm. Therefore rastering a
sample within this bore requires a small sample 1  1 cm and would provide
between 100 and 10000 unique points depending on the beam size. Extending
this technique to larger samples is a motivation for increasing the bore size in
the next generation of superconducting hybrid magnets. Another approach
would be to use small pulsed magnets that can achieve 30 T, but have a
repetition rate of 1 pulse per 7–8 minutes [23]. A sample sandwiched between
two of these magnets and rastered independently of the magnets provides a
way to increase the sample size. For this approach one would have to stop
measuring the relaxation after the peaks associated with the Zeeman splitting
coalesce with the main incoherent peak. The main restriction of this approach is
the long time between pulses. This time is controlled by internal heating of the
magnet. The time between pulses could be reduced to the time limit of the
capacitor bank power supply by rastering additional magnets while the first
pair cools.
For both styles of magnets, field gradients are important. Though for tradi-
tional MRI homogeneity on a few PPM [20] is expected, 100 ppm is accep-
table. The pulsed magnets are quite configurable so field gradients of this order
are reasonable. For the Zeemans style of magnet, a high-homogeneity (100 ppm)
resistive insert would have to be designed. This insert design would drop the peak
field to 28 T (M. D. Bird, private communications). Nevertheless, one could
envision running the magnet for a series of cycles in a high field mode and then
changing for another series of cycles to a high homogeneity mode.
186 N. Kardjilov et al.

Finally, counting times for the MRI measurement must be considered.

Assuming the counting described for the NMR measurement, but reducing
the sample to a 1  1 mm cross-section, means that beam focusing of 15,000 is
required so that a measurement can be carried out in 1 week. This is assuming
that the static and pulsed field method is used. An order of magnitude may be
gained by increasing the pulse rate of the pulsed magnet. If the fully pulsed field
approach is used, this number would increase, as the repetition rate is even
slower. Therefore beam flux is the biggest challenge to realizing neutron-probed
MRI.
In summary the major challenges to realizing neutron-probed MRI are tech-
nical and not fundamental to the science. The core technique of neutron NMR
can be demonstrated with the next generation of high magnetic field beamlines.
Nevertheless, flux and the limitations on neutron focusing are the major chal-
lenges to using the neutron NMR technique for neutron-probed MRI.

Acknowledgments Garrett Granroth is grateful for useful discussions with J.-K. Zhao,
G. Greene, E. Iverson, M. W. Meisel, N. Sullivan, and B. Halperin. His work was performed
at Oak Ridge National Laboratory, managed for the U.S. Department of Energy by UT-
Battelle, LLC, under Contract No. DE-AC05-00OR22725.

References
1. N. Kardjilov, I. Manke, M. Strobl, A. Hilger, W. Treimer, M. Meissner, T. Krist, J.
Banhart, Three-dimensional imaging of magnetic fields with polarised neutrons, Nat.
Phys. 4, 399–403 (2008).
2. Th. Krist, S. J. Kennedy, T.J. Hick, F. Mezei, New compact neutron polarizer. Physica B
241–243, 82–85 (1998).
3. W. Treimer, M. Strobl, N. Kardjilov, A. Hilgar, I. Manke, Wavelength turable device for
neutron radiography and temography, Appl. Phys. Lett, 89, 203504 (2006).
4. A. Hilger, N. Kardjilov, M. Strobl, W. Treimer, J. Banhart, The new cold neutron
radiography and tomography instrument CONRAD at HMI Berlin, Physica B
385–386, 1213 (2006).
5. P. Gammel, D. Bishop, Fingerprinting vortices with smoke, Science 279, 410–411 (1998).
6. Ch. Jooss, J. Albrecht, H. Kuhn, S. Leonhardt, H. Kronmüller, Magneto-optical studies
of current distributions in high-Tc superconductors. Rep. Prog. Phys. 65, 651–788 (2002).
7. I. Manke, N. Kardjilov, M. Strobl, A. Hilger, J. Banhart, Investigation of the skin effect
in the bulk of electrical conductors with spin-polarized neutron radiography, J. Appl.
Phys, 104, 076109 (2008).
8. http://www.springer.com/series/8141, this series of books
9. A.I. Frank, I. Anderson, I.V. Bondarenko, A.V. Kozlov, P. Hoghoj, and G. Ehlers, Phys.
Atom. Nucl. 65, 2009 (2002).
10. T.A. Jyrkkiö, M.T. Huiku, K.N. Clausen, K. Siemensmeyer, K. Kakurai, and M. Steiner,
Calibration and applications of polarized neutron thermometry at milli- and microkelvin
temperatures, Z. Phys. B 71, 139 (1988).
11. J.B. Hayter, G.T. Jenkin, and J.W. White, Polarized-Neutron Diffration from Spin-
Polarized Protons: A Tool in Structure Determination? Phys. Rev. Lett. 33, 696 (1974).
12. M. Dobhal, NMR Data for Carbon-13. Part 4: Natural Products, Springer, Heidelberg
(2006).
10 Novel Imaging Techniques 187

13. http://www.lanl.gov/orgs/mpa/nhmfl/60TLP.shtml
14. http://www.lanl.gov/orgs/mpa/nhmfl/images/60TLPpulse.jpg
15. C.P. Slichter, Principles of Magnetic Resonance, Springer-Verlag, Heidelberg (1990).
16. MathCad, Parametric Technology Corporation.
17. G.E. Granroth, Prospects for Neutron probed NMR in The proceedings of ICANS–XVII,
J. Wei, S. Wang, W.L. Huang, J.S. Zhao, ed. http://www.icans-xviii.ac.cn/proceedings/
indexed/copyr/15. pdf.
18. http://www.sns.gov/instrument_systems/beamline_02_basis/index.shtml, (2007).
19. G.E. Granroth, C. Broholm, F. Klöse, G. Srajer, Z. Islam, Y. Lee, and J. Lang, Report
from Probing Matter at High Magnetic Fields with X-Rays and Neutrons, SNS Report
SNS107000000TR-005-R00, (2005).
20. J. Jin, Electromagnetic Analysis and Design in Magnetic Resonance Imaging, CRC Press,
Boca Raton (1999).
21. P. Sprawls, Magnetic Resonance Imaging, Medical Physics Publishing, Madison (2000).
22. X.-L. Wang, T.M. Holden, G.Q. Rennich, A.D. Stoica, P.K. Liaw, H. Choo, and C.R.
Hubbard, Physica B 385–386, 673 (2006).
23. K. Ohoyama, N. Katoh, H. Nojiri, Y.H. Matsuda, H. Hiraka, K. Ikeda, H.M. Shimizu,
Neutron diffraction under 30 T pulsed magnetic fields, J. Mag. Mag. Mat. 310, e974
(2007).
Section C
Chapter 11
Neutron Imaging for the Hydrogen Economy

M. Arif, D.S. Hussey, and D.L. Jacobson

Abstract The development of fuel cells and hydrogen storage materials will be one
of the highest global research and development priorities for the foreseeable future.
The particular abilities of neutrons to penetrate materials and to image hydrogen
will make neutron imaging a key technique, allowing the in-situ study of real
operational devices. This chapter describes the current state of the art in this field.

Keywords Fuel cell  PEM  Hydrogen storage  Water dynamics

11.1 Introduction
Compared to most other forms of radiation, neutrons are highly efficient means of
probing complex structures because of their tremendous of penetrating capability
many common materials and their unique capability of distinguishing different
materials with similar chemical and physical properties. Consequently, neutron
imaging has always been regarded as one of the most potentially important tools
for many areas of industrial research. However, these promises have remained
largely unfulfilled, and neutron imaging techniques have seen limited applications,
primarily due to traditionally poor spatial resolution and slow processing time as
well as lack of identification of critical, high-impact areas of research. Most
industrial and academic research that can significantly benefit from this technique
involves dimensions that are a few tens of micrometers in length and interaction
times that last at longest a few seconds. The current practical limitations for
neutrons are about 100 mm in spatial resolution and tens of seconds in time
resolution. These limitations are mainly due to the relatively weak neutron sources
(nuclear reactors that are many orders of magnitude weaker than even table-top
photon sources), large mean free paths of neutrons and secondary radiation in

M. Arif (*)
National Institute of Standards and Technology, 100 Bureau Drive, MS 6100,
Gaithersburg, MD 20899-6100, USA
e-mail: [email protected]

I.S. Anderson et al. (eds.), Neutron Imaging and Applications, 191

Neutron Scattering Applications and Techniques, DOI 10.1007/978-0-387-78693-3_11,
Ó Springer ScienceþBusiness Media, LLC 2009
192 M. Arif et al.

materials (limits spatial resolution), and the inability to process data in real-time
while being statistically significant (limited by weak source).
However, in the past decade enough gains have been made to enable neutron
imaging to be applied in different fields of science and technology. One of the most
important prospects is the understanding of the interface physics, chemical kinetics,
and proper modeling of the processes that govern the operation of the future
generation of energy production and storage devices based on hydrogen. The
sensitivity of neutrons to hydrogen, together with other unique properties, makes
them an ideal probe for characterizing the structure, morphology, and hydrogen
dynamics that are key to the future use of these promising devices. The neutron
imaging technique, in particular, has emerged as one of the most important tools in
the development of robust and efficient fuel cells accompanying hydrogen storage
materials and devices. This article will focus on the application of neutron imaging
research in the development of low-temperature fuel cells that may power homes,
consumer electronics, and transport vehicles in the emerging hydrogen economy.

11.2 Neutron Imaging of Fuel Cells and Hydrogen

Storage Devices

11.2.1 Neutron Imaging of Fuel Cells

Fuel cells are efficient, silent, and virtually non-polluting electrochemical

devices that have immense potential for powering homes, businesses, and
transport vehicles of the future. Major energy-related and automobile compa-
nies worldwide are currently engaged in extensive research to perfect fuel cells
for everyday, cost-efficient use.
Fuel cells containing solid proton exchange membranes (PEM) are of special
interest, as they enable fuel cell operation at lower temperatures (< 1008C) than
conventional solid oxide fuel cells (> 5008C). PEM fuel cells convert the
chemical energy of hydrogen-containing fuels directly into electricity in the
presence of a catalyst, without combustion and without any moving parts.
Fuel cells provide a quantum leap in fuel efficiencies compared to traditional
heat engines because conversion to electricity does not require the intermediate
generation of heat. Similar to batteries, fuel cells generate power electrochemi-
cally and thereby produce extremely low to zero emissions, depending on the
fuel supply and operating conditions.
Current PEM fuel cells use perfluorosulfonic acid-based membrane technol-
ogy such as Nafion1 and Gore-Select1 [1]. Platinum catalyst is applied to both
sides of the membrane to form the membrane electrode assembly (MEA).
During operation, hydrogen gas is brought to the anode side of the MEA
while air is brought to the cathode side. H2 is oxidized to protons and electrons.
The protons migrate through the membrane (Fig. 11.1) to the cathode side and
react with O2 in the air to form water. The electrons, on the other hand, conduct
externally to provide electricity. In this process, water is ‘‘dragged’’ from the
11 Neutron Imaging for the Hydrogen Economy 193

Fig. 11.1 Schematic of a fuel cell. The platinum (Pt) catalyst is generally applied to both sides
of the membrane to form the membrane electrode assembly. Formation of water at the
membrane can cause the cell to perform poorly

anode side of the MEA to the cathode side with the proton transport. In the
meantime, water is generated on the cathode side of the MEA. A water gradient
thus exists within the membrane. The PEM water content is extremely critical as
it impacts the ionic conductivity and mechanical properties such as strength and
gas permeability. Depending on the water concentration, the membrane per-
formance and lifetime can be significantly reduced. In order to maintain the
proton conductivity and preserve membrane life, PEM fuel cells typically
require humidified reactant gases. This vapor, together with the water gener-
ated during operation, poses significant challenges for water management. A
hydrophobic porous gas diffusion layer (GDL) is typically used to help manage
water distribution. Furthermore, the electrode composition may be hydropho-
bic or hydrophilic, which is another area of water management difficulty.
PEM fuel cells have gone through significant evolution in recent years, but
one of the remaining key challenges is efficient water management. The control
194 M. Arif et al.

of incoming humidity and the product water within the MEA and GDL is
generally acknowledged as one of the most crucial aspects of operating a PEM
fuel cell. Too much water within the MEA or GDL will result in flooding
conditions that impede gas diffusion; too little water will reduce the membrane
proton conductivity, thereby decreasing the cell performance; and cycling
between water levels will reduce membrane life. Thus, proper water manage-
ment is the key to a stable and long-life PEM fuel cell and must be achieved by
properly designing the flowfield, the GDL, the MEA, and their interfaces. This
requires a fundamental understanding of the in situ water distribution in an
operating fuel cell.
It is critically important for both optimal design and efficient operation of
the fuel cells to study and quantify these adverse effects in situ using non-
destructive means [2]. At present, neutron imaging is the only technology that
has the ability to look inside a standard, commercially viable PEM fuel cell
while the cell is operating and characterize the water content non-destructively.
Since neutrons are highly penetrating in most materials, they are ideal probes in
the investigation of internal structures of bulk objects like the fuel cell. Thermal
neutrons (low-energy neutrons) passing through matter experience the follow-
ing interactions: (a) absorption, (b) scattering (elastic, inelastic, small angle
scattering), (c) refraction, (d) change in the phase of the neutron wave,
(e) depolarization of an incident spin direction.
These interactions are used for signals in thermal neutron two-dimensional
(2D) (radiographic) and 3D (tomographic) imaging (Chapters 5 and 6, in this
volume). The most common processes are absorption and scattering. The ther-
mal cross-sections for absorption and scattering events are complex functions of
atomic mass, atomic number, nuclear isotope, and neutron energy. The contrast
obtainable between materials in radiographic images is a function of their differ-
ent cross-sections. For thermal neutrons, hydrogen-based compounds (such as
water) would be highly visible against aluminum, iron, or carbon, which are
commonly used in fuel cell construction (Fig. 6.1). This is due to the fact that the
so-called mass attenuation coefficient of water is nearly two orders of magnitude
larger than that of the background materials. This unique signature of hydrogen
allows for the study of water transport phenomena in fuel cells with sub-micro-
gram precision.

11.2.2 Neutron Imaging of Hydrogen Storage Devices

Vehicles powered by PEM fuel cells require, among other things, the presence of
onboard hydrogen storage vessels. The design of metal-hydride beds with
suitably low weight and volume to meet the demands of automotive applica-
tions is no trivial matter. The characterization of prototype technologies would
benefit greatly from an in situ imaging technique such as neutron radiography
to aid in understanding the hydrogen concentration gradients across these beds
11 Neutron Imaging for the Hydrogen Economy 195

during absorption/desorption cycling. As a demonstration, the hydrogen

uptake in Sieverts’ Reactor was imaged in situ during a hydrogen charging
step. Figure 11.2 shows a photo of the Sievert’s Reactor and neutron images at
two different points during the hydrogen charge. Since hydrogen has such a
large total neutron scattering cross-section, it is straightforward to measure the
real-time metal hydride uptake of gaseous hydrogen.

Fig. 11.2 Neutron radiography study of hydrogen storage in LaNi4.78Sn0.22. (a) Photo of the
Sievert’s reactor. (b,c) Colorized (bottom) and raw greyscale (middle) images of the reactor and
the hydrogen content (top) in the uncharged and charged states, respectively

Future research on hydrogen storage will correlate local hydrogen reaction

rates and contents with temperature profiles numerically simulated by detailed
thermal models based on designs such as those previously developed [3,4] at the
Jet Propulsion Laboratory (JPL) for hydride compressor beds to be used in
sorption cryocoolers on the spacecraft of the Planck Mission [5,6]. A cross-
sectional drawing of such a highly integrated hydride bed is shown in Fig. 11.3.

Fig. 11.3. Cross-sectional drawing of a prototype hydrogen storage bed. The porous alumi-
num is filled with LaNi4.78Sn0.22. With neutron tomography, the effects of hydrogen uptake,
such as swelling and temperature gradients, can be revealed
196 M. Arif et al.

This prototype hydrogen storage vessel consists of an outer aluminum shell

surrounding an inner bed enclosed by a 3 cm diameter stainless steel tube
(30 cm long) that contains an annular, open-pore Al foam matrix filled with
LaNi4.78Sn0.22 hydrogen-absorbing powder. The center of the annulus contains
a porous 316L stainless steel (SS) filter tube for hydrogen delivery.
In such a study, real-time images of concentration gradients within the
LaNi4.78Sn0.22/Al foam hybrid matrix would need to be collected as a function
of controlled stepwise hydrogen addition and removal (hydrogen coming from
an adjacent control volume). Also, neutron tomography would be performed
for each step after steady state had been reached. In this way, it may be possible
to characterize the presence and nature of any hydrogen concentration
gradients.

11.2.3 Primary Areas of Applications

11.2.3.1 Efficient Flow Field Design

Flow channels are the conduit through which hydrogen and air are drawn into a
fuel cell. They also act as conduits to expel excess water. Effective design of the
flow channels is essential to achieve the desired proper balance of incoming
humidity and outgoing water. The design should ensure proper hydration of the
MEA that in turn determines the fuel cell operating efficiency and durability.
The flow channels have a typical dimension of a few tenths of a millimeter to a
few millimeters and can be arranged in many different geometric patterns. By
optimizing the channel geometry, efficient, passive water removal from the fuel
cell can be realized, reducing the need for energy-intensive gas purges [7].
Neutron imaging has played a critical part in describing the correlation between
current density and total measured water content and its distribution in operat-
ing fuel cells, since the spatial resolution of scintillators easily resolves the
channel separation, as shown in Fig. 11.4.

11.2.3.2 Water Dynamics Through the PEM

Fuel cell performance depends on the complex interplay of several chemical and
mass transport processes that occur at interfaces. A typical MEA consists of
electrodes with catalytic particles on larger particles of a support medium,
which are embedded on an electrode. Two electrodes sandwich a PEM. The
structure of the electrode involves pores that have a typical length scale of
10–50 nm. For a fuel cell to be efficient, these pores must remain open for
reactant transport during the operating cycle.
The first reported fuel cell neutron imaging measurement was carried out
more than a decade ago [8]. It was a pioneering experiment attempting to
improve the understanding of water mass transport and its effects on the
electrochemical performance of PEM fuel cells via detailed knowledge of the
11 Neutron Imaging for the Hydrogen Economy 197

Fig. 11.4 A sub-image of a neutron radiograph of a 50 cm2 active area fuel cell clearly showing
the water content (dark areas) of the separate channels. In this case the anode and cathode
channels were oriented perpendicularly to permit discrimination of anode vs. cathode channel
water.

through-plane water distribution. The measurements, as shown in Fig. 11.5,

involved direct neutron imaging and quantification of the variation of water
content normal to the MEA surface cross sections. The resolution and effi-
ciency of the detector used for the experiment were poor and the available
neutron flux was low. As a result, the measurement was actually done on a
thermally bonded membrane assembly that was nearly ten times thicker than
that in modern fuel cells. For commercially competitive materials, a similar
measurement requires a detector resolution of 10 mm or better in order to
spatially resolve the water distribution. A future generation of neutron imaging
techniques and technologies are being developed to achieve this.

11.2.3.3 GDL Characteristics and Water Retention

The structure of the GDL plays a large role in determining the water retention
of the fuel cell. Placing a highly hydrophobic microporous layer (MPL) near the
cathode catalyst helps to reduce flooding by presenting a barrier to condensed
water [9]. Adding PTFE to the bulk GDL substrate can decrease or increase
198 M. Arif et al.

Fig. 11.5 One of the first neutron radiographs of an operating fuel cell, showing the water
gradient across a 0.5 mm thick (wide) membrane, comprising several thermally bonded
Nafion 117 membranes. The darker region in the membrane center indicates more water

water retention [10]. Having a broad range of pore sizes can ensure that there
are sufficient pathways for reactant gas to reach the catalyst layer [11].

11.2.3.4 Characterization of Two-Phase Flow Phenomena in the Flow Field

The water removal from the fuel cell GDL and channels is a difficult process to
model since it is a non-isothermal, two-phase flow system. For instance, the
waste heat produced in the cathode can drive evaporation so that the fuel cell
retains significantly less water at higher current production, despite the corre-
sponding increase in water production [12]. Additionally, the operating pres-
sure and inlet gas flows impact the condensation rate and the convection of
water in the flow channels [13].

11.2.3.5 Membrane Durability

One of the failure modes of PEM fuel cells stems from dimensional changes of the
polymer layers resulting from repeated drying and wetting cycles or freeze–thaw
cycling. Using high-resolution neutron imaging, the effects of freezing on the
formation of ‘‘frost heaves’’ between the GDL and MEA can be resolved [14].
11 Neutron Imaging for the Hydrogen Economy 199

Also, it may be possible to study interface defects by neutron phase contrast

imaging techniques that are currently being developed (Chapter 8).

11.2.3.6 Materials for Hydrogen Storage

The success of any future hydrogen economy depends, in no small part, on the
ability to develop inexpensive, state-of-the-art materials with high hydrogen
storage capacity. Such targets require the search for new and improved materi-
als. The utility of a particular material for these applications depends critically
on ionic transport. Thus, the fundamental questions that must be answered in
order to understand and tailor the performance of these systems are ‘‘Where do
the ions go?’’ and ‘‘How do they get there?’’ Many types of materials have been
tried or suggested for use as hydrogen storage media. Proper characterization of
these materials requires information regarding hydrogen diffusion properties as
well as the macroscopic morphologies that govern their useful properties. As
shown in Fig. 11.2, neutron imaging can play a key role in assessing device
performance.

11.3 Neutron Imaging Facilities and Techniques

11.3.1 Basic Principles

The quality of the image depends on a complex interplay of beam properties,

sample material, neutron intensity, and temporal and spatial resolution of the
detector system. For hydrogen or water, the primary form of interaction is
scattering with a cross section of 0.810–22 cm2. This is nearly two orders of
magnitude larger than that of aluminum, a common material used in fuel cell
construction. The total transmission of neutron flux is given by the Lambert-
Beer law:
P
I  ðNtÞi
¼e i ; (11:1)
I0

where I is the transmitted and I0 the incident neutron flux, N is the atom
number density, t is the sample thickness, and  is the total scattering cross-
section responsible for the neutron loss in the forward direction passing through
the sample. Since the primary goal is to image and quantify the water content,
all measurements involve the basic steps of taking a dry and wet image of the
device. The wet image is normalized with respect to the dry image to determine
the profile of the water thickness. The detection system typically consists of a
scintillator viewed by either a charge-coupled device CCD or an amorphous
silicon flat panel. The spatial resolution, which is a convolution of detector
200 M. Arif et al.

hardware pixel resolution, geometric blur, and scintillator blooming, deter-

mines which components of the fuel cell can be identified.
The water volume can be obtained by combining the water thickness from
Eq. (1) with knowledge of the area A and the liquid water attenuation coefficient,
mw = Nww. Since w depends on neutron energy, mw must be measured for each
beamline or spectrum and detector system used. The total uncertainty in the
water thickness, tw, comes from both random and systematic contributions.
The random uncertainty, from Poisson counting statistics, is the fundamental
limit to the measurement accuracy and is estimated by:
sffiffiffiffiffiffiffiffiffiffiffiffiffi
1 2
tw  ; (11:2)
w I0 AT

where I0 is the incident fluence rate (cm–2 s–1), T is the acquisition time, and  is
the detection efficiency [15]. Typical random uncertainties achieved at the
national institute for standards and technology (NIST) neutron imaging facility
range from 1 to 10 mm.

11.3.2 Imaging Techniques

Both 2D and 3D neutron imaging are used to study water dynamics in fuel cells.
Neutron tomography can provide important information about an object’s
internal construction and about its constituent materials. The neutron tomo-
graphy technique allows one to reconstruct a 3D image of an object from the 2D
neutron radiographs taken for several rotations of the object about a rotation
axis (Chapter 6). Using the technique of filtered back projection (FBP), these
projection images are Fourier filtered and then back-projected across planes
orthogonal to the rotation axis. The result of this process is a stack of 2D
reconstructions of the planes of the object, orthogonal to the rotation axis. The
result of the FBP is a 3D map of the object’s neutron attenuation coefficient.
Subtracting the dry map from the wet map yields the 3D water profile of the fuel
cell. The spatial resolution of the map is determined by the resolution of the
detector system, on the order of 100 mm. Figure 11.6 shows the 3D water
content of an operating fuel cell at steady state.
Because of the time required to rotate the fuel cell by 1808, neutron tomo-
graphy can be used only as a steady state measurement. Often the desired
information is the average water gradient going from the anode to the cathode.
This ‘‘through-plane’’ water gradient must be imaged with higher resolution
than is typical of scintillator-based detectors, since the total thickness of the
PEM sandwich is only on the order of 0.5 mm. Various companies are devel-
oping new detector technologies aimed at achieving 10 mm or better resolution
in neutron detection in the near future [16]. These detectors are based on micro-
channel plates and have been demonstrated to have resolution better than
25 mm [17] (Chapter 3). With a smaller pixel size, image acquisition time must
11 Neutron Imaging for the Hydrogen Economy 201

Fig. 11.6 A small fuel cell (4 cm2) was used to demonstrate neutron tomography. The water
distribution in each slice plane from the anode to the cathode shows that anode flooding is
occurring in this fuel cell as a result of the thick (1 mm) gas diffusion layers. The data set was
acquired within 20 min

be increased in order to maintain signal-to-noise ratios similar to those achieved

with scintillators, and a typical time is similar to that of a tomography scan.
While a high-resolution through-plane image yields only 2D information, it
enables direct measurement of the water gradient in commercial-grade GDLs.

11.3.2.1 Phase Imaging and Other Emerging Techniques

The neutron tomography technique can be complemented by two other
related imaging techniques, refraction imaging (Chapter 6) and phase contrast
imaging (Chapter 8). The refraction imaging technique uses a mono-energetic
neutron beam that is refracted by constituent elements in the sample. The
refracted beams from different elements are spatially separated from each
other by a very small amount. A single-crystal analyzer redirects this spatially
separated beam away from the original direction, and a high-resolution
position-sensitive detector detects the redirected beam. Since the refracted
beam is now separated from the direct transmitted beam, the inherent back-
ground present in normal radiographs is eliminated, and an image with very
sharp contrast is formed.
In phase contrast imaging, an image of the locally varying phase shift of a
neutron beam caused by the sample is formed. Phase variation may occur
because of dimensional and constituent variations, the presence of a magnetic
domain structure, the presence of precipitates, or a spatial distribution of stress.
The image is recorded with a high-resolution position-sensitive detector. Not
202 M. Arif et al.

only are the bulk properties of the sample measured, but the elemental distribu-
tions of local inhomogenities are also observable. This technique is typically a
few hundred times more sensitive than normal radiography. It is ideally suited
for detecting laminate separation and minute thickness variations of polymer
electrolyte layers. NIST is developing a cold neutron imaging facility that will
enable research to measure the phase gradient, using a variety of such techni-
ques, including a grating method based on the Talbot effect [18].
Ultimately, these technologies can be used as a quality control tool for fuel
cell manufacturers. Fuel cells can be subjected to routine scans for the verifica-
tion of proper assembly and operational characteristics. Additional resolution
may be achieved using neutron point sources to magnify the neutron image of
the sample.

11.4 Neutron Imaging Facilities for Fuel Cell Research

Successful applications of neutron imaging of fuel cells require access to high-
intensity neutron sources and a detection system with fast temporal and high
spatial resolution. Also, any large-scale comprehensive research with fuel cell
and hydrogen storage devices requires an associated hydrogen support infra-
structure capable of safely providing large quantities of hydrogen to these
devices. The latter requirement is especially difficult for conventional steady
state reactor-based neutron sources. There are facilities in the United States
(NIST, PSU), Japan (JAERI, KURRI), and Europe (PSI, HMI, FRM2) that
are carrying out fuel cell related research. Among these, the facility at NIST has
the most developed and focused infrastructure for hydrogen economy related
research. In order to give an idea of the practical considerations, a brief
description of this facility follows.
The NIST neutron imaging facility is located at beam tube 2, with a direct
view of the reactor core at the NIST Center for Neutron Research; a schematic
layout is shown in Fig. 11.7. An aperture of diameter D (variable diameter from
0.1 cm to 2 cm) defines the beam collimation, characterized by the ratio L/D,
where L  6 m is the distance from the aperture to the detector. A 10 cm section
of Bi is used to reduce the gamma-ray and fast neutron background, though it
also reduces the thermal neutron fluence rate by about a factor of four from that
with no filter. The typical fluence rates used range from about 2106 cm–2 s–1 to
3107 cm–2 s–1. The flight path is evacuated to reduce the neutron scattering in
air. The sample position sits directly in front of the detector, and there is a 5-axis
manipulation table to align the object and perform tomography. There are
several penetrations through the radiation shielding to allow gas lines and
electrical connections to be made between the fuel cell and fuel cell test stand.
The NIST test stand accurately controls the flow of hydrogen, air, nitrogen,
oxygen, and helium over a wide range of flow rates [15]. The maximum current
that the test stand can control is about 500 A. The inlet gases are humidified in
11 Neutron Imaging for the Hydrogen Economy 203

Fig. 11.7 (a) Photograph, and (b) schematic of the NIST neutron imaging facility. Also shown
is (c) the fuel cell test stand and (d) the sample area with connections to a typical fuel cell

sparger bottles, and heated inlet gas lines prevent condensation. The hydrogen is
supplied by a hydrogen generator so there is no stored inventory of hydrogen in
the confinement building. There is an extensive set of hydrogen safety features,
including several hydrogen sensors, a line pressure sensor, and flame detection in
the hydrogen vent that automatically shuts down all hydrogen systems in the
event of a detected emergency. Also, there is a freeze chamber, capable of
controlling the temperature between –408C and 508C for use in freeze/thaw
studies.

11.5 Examples of PEM Fuel Cell Neutron Imaging

To demonstrate the capability of 3D neutron imaging, we describe a study of a small

PEM fuel cell, with an active area of 4 cm2 (Fig. 11.6). The tomography was
performed with a scintillator-based detector and, in order to measure water gradi-
ents, a 1.0 mm thick GDL on the anode and cathode were used, as well as a Nafion
117 membrane. Because of this large thickness, the diffusion through the GDL was
significantly slower than in typical fuel cells, which impacts the water content. In
particular, the water gradient shows that there is significant flooding in the anode.
Because of the surprising amount of anode flooding, a second experiment
using the same small fuel cell was conducted to measure the water diffusion
204 M. Arif et al.

times under different conditions. The fuel cell was operated at 0.4 V with both
gas streams humidified to a relative humidity (RH) of 75 % for 2 h to ensure
that the GDL saturation was at equilibrium. After this precondition, the fuel
cell load was dropped. In the first case, the humidified gas flows were main-
tained, and in the second case the gas flows were stopped. By so doing, the
efficacy of the competing GDL water transport mechanisms of capillary action
versus evaporation and advection could be compared. Figure 11.8 shows the
first case in which the fuel cell load was stopped but the gas flow was main-
tained. The water content in the GDL is seen to decline rapidly to the comple-
tely dry state within about 15 min. Figure 11.9 shows the second case where

Fig. 11.8 Water transport in the GDL when the fuel cell load is stopped, but the under-
saturated flows are maintained. (A–D) Through-plane water profiles are shown from four of
the repeat measurements. The grayscale neutron images show the water profile change from a
time after the load was stopped
11 Neutron Imaging for the Hydrogen Economy 205

Fig. 11.9 Water transport in the GDL when the fuel cell load and the gas flows are stopped. (A–D)
Through-plane water profiles are shown from four of the repeat measurements. The grayscale
neutron images show the water profile change from a time after the load was stopped. Compared
with the evaporation/advection case, the change in water content of the GDL is negligible

both the fuel cell load and gas flows were stopped. The water content of the GDL
barely changes, even after 70 min, indicating that capillary action plays a smaller
role than evaporation and advection in the water transport through the GDL.

11.6 Conclusion

Neutron imaging has become a powerful tool in the development of hydrogen-

based alternate energy devices. It is the only technique that can study the
critically important in situ water transport in commercial-grade fuel cells.
206 M. Arif et al.

Today this technique is used by almost all automobile manufacturers and major
fuel cell developers. By all accounts, neutron imaging research has shortened
the development time of fuel cells and is providing reliable, real-time data
important for optimum design and performance that was almost impossible
to get only a few years ago. Though the primary focus of neutron imaging
research has been on fuel cells during the last few years, it has now started to
play an expanded role in the development of hydrogen storage materials and
devices. It is expected that this trend will continue in the coming years. In spite
of all the success that neutron imaging has enjoyed in recent years, even higher-
impact research is limited by the current temporal and spatial resolution of
neutron detection devices, the lack of very accurate analytical techniques, and
the lack of availability of wide infrastructures for hydrogen research at various
neutron facilities around the world. Fortunately, all of these issues are being
aggressively pursued by various government laboratories and industrial and
academic researchers. It is expected that in the very near future, both temporal
and spatial resolution will improve by an order of magnitude. This, coupled
with advanced methods such as phase contrast imaging, will provide much
needed enhanced capabilities and open new avenues of research that are not
possible today. Consequently, in the foreseeable future, neutron imaging will
continue to significantly contribute to enhanced efforts to develop and com-
mercialize pollution-free, hydrogen-powered energy devices that are the cen-
terpieces of the emerging hydrogen economy.

References
1. Certain trade names and company products are mentioned in the text or identified in an
illustration in order to adequately specify the experimental procedure and equipment used.
In no case does such identification imply recommendation or endorsement by NIST, nor
does it imply that the products are necessarily the best available for this purpose.
2. J. St-Pierre, J. Electrochem. Soc.154, B724 (2007).
3. P. Bhandari, M. Prina, M. Ahart, R. C. Bowman, and L. A. Wade, Sizing and Dynamic
Performance Prediction Tools for 20 K Hydrogen Sorption Cryocoolers in Cryocoolers 11,
Edited by R. G. Ross, Jr. Kluwer Academic/Plenum Press, New York, pp. 541–549 (2001).
4. M. Prina, P. Bhandari, R.C. Bowman, L. A. Wade, D. P. Pearson, and G. Morgante,
Performance Prediction of the Planck Sorption Cooler and initial Validation, Advances in
Cryogenic Engineering, Vol. 47, edited by S. Breon, et al., Am. Inst. Phys., New York,
pp. 1201–1208 (2002).
5. P. Bhandari, M. Prina, R.C. Bowman-Jr, C. Paine, D. Pearson, A. Nash, Sorption Coolers
using a Continuous Cycle to Produce 20 K for the Planck Flight Mission, Cryogenics 44,
395–401 (2004).
6. D. Pearson, R. Bowman, M. Prina, P. Wilson, The Planck Sorption Cooler: Using Metal
Hydrides to Produce 20 K, J. Alloys Compounds 446–447, 718–722 (2007).
7. J. P. Owejan., T. A. Trabold, D. L. Jacobson, M. Arif, and S. G. Kandlikar, Int. J.
Hydrogen Energy 32, 4489 (2007).
8. R. J. Bellows, M. Y. Lin, M. Arif, A. K. Thompson, and D. Jacobson, J. Electrochem. Soc.
146, 1099 (1999).
11 Neutron Imaging for the Hydrogen Economy 207

9. M. Mathias, J. Roth, J. Fleming, and W. Lehnert, ‘Diffusion Media for PEM Fuel Cells,’
in Handbook of Fuel Cells – Fundamentals, Technology and Applications. Fuel Cell
Technology and Applications, Vol. 3, W. Vielstich et al.(Eds.), John Wiley & Sons, Chapter
46 (2003).
10. R. Mukundan, J. R. Davey, T. Rockward, J. S. Spendelow, B. S. Pivovar, D. S. Hussey,
D. L. Jacobson, M. Arif, and R. L. Borup, ’Imaging of Water Profiles in PEM Fuel Cells
Using Neutron Radiography: Effect of Operating Conditions and GDL Composition’,
ECS Trans. 11(1), 411 (2007).
11. K. Yoshizawa, K. Ikezoe, Y. Tasaki, D. Kramer, E. H. Lehmann, and G. G. Scherer,
ECS Trans. 3, 397 (2006).
12. M. A. Hickner, N. P. Siegel, K. S. Chen, D. N. McBrayer, D. S. Hussey, D. L. Jacobson,
and M. Arif, J. Electrochem. Soc.153, A902 (2006).
13. M. A. Hickner, N. P. Siegel, K. S. Chen, D. N. McBrayer, D. S. Hussey, D. L. Jacobson,
and M. Arif, J. Electrochem. Soc. 155(4), B427–B434 (2008).
14. S. Kim, A.K. Heller, M.C. Hatzell, M.M. Mench, D.S. Hussey, D.L. Jacobson, High
Resolution Neutron Imaging of temperature-driven flow in polymer electrolyte fuel cells,
American Nuclear Society National Meeting, Anaheim, CA (2008).
15. D.S. Hussey, D.L. Jacobson, M. Arif, K.J. Coakley, and D.F. Vecchia, In situ fuel cell
water metrology at the NIST neutron imaging facility, Proceedings of the ASME Fuel Cell
Conference, New York, June 18–20, 2007.
16. O.H.W. Siegmund, J.V. Vallerga, A. Martin, B. Feller, M. Arif, D.S. Hussey, and D.L.
Jacobson, Nucl. Instrum. Meth. A 579, 188 (2007).
17. W.B. Feller, P.L. White, and P.B. White, Gamma Insensitive Highly Borated Micro-
channel Plates for Neutron Imaging, Proceedings of the 8th World Conference on Neutron
Radiography, 583–591 (2008).
18. F. Pfeiffer, C. Grünzweig, O. Bunk, G. Frei, E. Lehmann, and C. David, Phys. Rev. Lett.
96, 215505 (2006).
Chapter 12
Material Science and Engineering
with Neutron Imaging

D. Penumadu

Abstract This chapter summarizes some of the results related to the use of
neutron imaging (radiography and tomography) as applied to the broad area of
materials science and engineering research. These include multi-phase flow
visualization in metal casting techniques, energy-selective imaging of materials
and its use for texture and stress imaging in polycrystalline materials, charac-
terization of discrete particle systems, flow through porous media, and strobo-
scopic imaging. The importance of spatial resolution and neutron detector type
for given engineering applications is also addressed.

Keywords Metals  Composites  Granular  Flow  Pyrolysis  Casting 

Numerical  Modeling  Imaging  Neutrons  MultiPhase  Functional 
Structural  Materials

12.1 Introduction

Film and image plate-based thermal neutron imaging for nondestructive inspec-
tion has found many useful engineering applications in the past three decades, as
demonstrated by the establishment of an American Society for Testing and
Materials (ASTM) standard (E1316) for its routine implementation at various
research reactors (Chapter 5). However, in the past decade the use of thermal and
cold neutrons for material science and engineering applications has rapidly
evolved due to a dramatic improvement in the availability of large-area amor-
phous silicon flat panels, high-resolution (20482048) and large charge-coupled
device (CCD) chip-based digital cameras, sophisticated and inexpensive optical
lenses, and bright and thin (50–200 mm) neutron scintillation screens. This has

D. Penumadu (*)
University of Tennessee-Knoxville, Joint Institute for Advanced Materials (JIAM)
Chair of Excellence, Department of Civil and Environmental Engineering, Knoxville,
TN 37996-2010, USA
e-mail: [email protected]

I.S. Anderson et al. (eds.), Neutron Imaging and Applications, 209

Neutron Scattering Applications and Techniques, DOI 10.1007/978-0-387-78693-3_12,
Ó Springer ScienceþBusiness Media, LLC 2009
210 D. Penumadu

also enabled imaging of materials in three dimensions using tomography, which

requires obtaining a series of radiographic projections of the object as it is
rotated in fixed angular steps through 1808 (Chapter 6).
Imaging using X rays has had an impact on many fields of science and
engineering and is now used routinely for studying materials. There are many
user facilities in the United States and Europe that can provide a significant
number of options for both source and detector configurations for various
types of imaging possibilities. In general, the imaging includes the following
modes of operation: transmission, diffraction, fluorescence, spectroscopy, and
polarization. In transmission mode, the attenuation or phase shift of the X-ray
beam by the sample is measured. Absorption contrast is then used to map the
sample density or features of interest. Particular elemental constituents can be
located using measurements on each side of an absorption edge to give an
element-specific difference image with moderate sensitivity. Phase-contrast ima-
ging can be sensitive to internal structure even when absorption is low and can be
enhanced by tuning the X-ray energy. In X-ray tomography, one of these modes is
combined with sample rotation to produce a series of two-dimensional projection
images to be used for reconstructing the internal three-dimensional structure of
the sample. This is particularly important for observing the morphology of
microstructures. Many textbooks are currently published dealing with the vast
amount of literature on this rapidly evolving topic. These are too extensive to cite
in this chapter.
Imaging with neutrons is an evolving and powerful imaging technique for the
internal evaluation of materials that complements the X-ray technique very
well. The interactions of neutrons and X rays with matter are fundamentally
different, and this difference forms the basis of many unique and complemen-
tary applications. As is discussed in Chapter 11, the sensitivity of neutrons to
hydrogen and other unique properties make it an ideal probe for characterizing
the structure, morphology, and hydrogen content in the development of next-
generation fuel cells and hydrogen-storage materials. A comparison of the
interaction of X rays and neutrons with matter has already been covered in
earlier chapters; the following particular observations are important for mate-
rial science and engineering applications:
 sensitivity to light elements and contrast for neighboring elements in the
periodic table,
 variation of attenuation coefficients as a function of wavelength for thermal
and cold energy spectrum,
 the ability to distinguish isotopes (for example hydrogen and deuterium),
 large penetration distance for many commonly used structural elements,
 the ability to detect small amounts of water, and
 interaction with magnetic moments.
The complementary aspect of imaging with both neutrons and X rays is
demonstrated in Fig. 12.1 for a computer floppy disk. The option of using both
X rays and neutrons now exists at recently established neutron imaging facilities
12 Material Science and Engineering with Neutron Imaging 211

(a) (b)

Fig. 12.1 Radiograph of a computer floppy disk using neutrons (a) shows the polymeric
components clearly while X-rays (b) are sensitive to the metallic components

such as that at the Paul Scherrer Institute in Switzerland (http://neutra.web.

psi.ch/).
As is discussed in other chapters in this book, the use of neutron imaging is finding
important applications for nondestructive evaluation (NDE) inspections, studying
water flow management and coupled performance of fuel cells, studying structural
materials such as steel and concrete, many applications in geology and paleontology,
studying working engines using stroboscopic imaging, and more recently in biome-
dical and agricultural engineering. This chapter summarizes the use of neutron
imaging for additional novel material science and engineering problems.

12.2 Phase Change and Transport During Metal Casting

Casting is one of the oldest methods of manufacturing, dating back to 3000

B.C., and is often used because of its ability to produce large and durable
volumes of complex shaped objects with desirable features such as internal
cavities and hollow sections. The lost foam casting (LFC) method, also
known as expandable polystyrene (EPS) casting, began high-production appli-
cation in the early 1980s. The LFC process has unique advantages for metal
casting compared with competing technologies. It has no size limit and the near-
net-shaped casting products can be obtained; complex shapes can be cast in one
piece by using EPS foam patterns glued together. It is starting to be widely used
by many casting companies as a replacement for conventional processes such as
green sand and investment castings, especially in the automobile industry. For
example, General Motors has nine different parts made using LFC, including
aluminum cylinder head, aluminum cylinder block, aluminum-driven sprocket
212 D. Penumadu

support for automatic transmission, iron crankshaft, iron-differential case, and

iron clutch housing for automatic transmission. Castings from this process can
have surface finishes from 5 to 20 mm; dimensional accuracy is better than
traditional sand castings. Almost no further modification is required after the
LFC process, the machining and assembly being eliminated or reduced [1].
Unlike other casting processes, no parting lines, cores, or riser systems are
needed. The process is inexpensive, with the polystyrene, sand, and containing
units being relatively cheap. Compared with conventional sand casting using
resin-bonded sand, LFC reduces environmental waste because it can use
unbonded and recycled sand.
Two major steps are involved in the LFC process. The first is to produce a
polymeric foam pattern typically using EPS. The second step is the process of
casting, i.e., replacing the polymer pattern by molten metal. After the coated
polymer pattern is ready, it is placed in a mold around which silica sand is poured
to fill all the cavities. Molten metal is then introduced into the mold through the
gating, causing the polymer foam pattern to decompose into gaseous and liquid
pyrolysis products which escape into the loose sand. The molten metal flow fills the
space originally occupied by the foam pattern, thus yielding the target shape. The
casting product is obtained by removing the sand and refractory coating. During
the process of metal filling, a gas layer, also called the pyrolysis front, exists between
the metal and the polymer fronts; it mainly consists of the thermal degradation
products of styrene. The behavior of the pyrolysis front during the casting process
is important and has a significant impact on the quality of the casting products [2].
However, the pyrolysis front is difficult to identify using X-ray-based radiography
techniques and consequently has not been studied in the past.
Radiography has long been considered a valuable tool for detecting internal
features and flaws. Shrinks, blowholes, gas or slag inclusions, and other flaws
could be identified quickly using X-ray radiography, and gating and risering
appropriately adjusted [3]. Traditional radiography using X rays has also been
successfully used in practice to image aluminum flow to aid gating design and
turbulence reduction and determine flow patterns [4, 5]. However, it is also
crucial to characterize the behavior of the pyrolysis fronts of polymer foam
patterns as they recede during the LFC process. X-ray radiography has difficulty
in visualizing the polymer foam and pyrolysis products because they have low
attenuation, but for neutron imaging they give high attenuation. Then other
issues involved in LFC, such as the effect of the properties of the refractory
coatings and unbonded granular materials on the filling behavior of molten metal
flows, can also be investigated. In addition, an in-situ quality control approach
for LFC can be developed based on neutron imaging.
The macroscopic cross section and the calculated thickness for 50% neutron
attenuation of the typical materials involved in LFC are listed in Table 12.1.
This shows that there is remarkable difference in the attenuation between
aluminum and styrene, which creates the contrast mechanism in neutron radio-
graphy for studying the LFC process in situ.
12 Material Science and Engineering with Neutron Imaging 213

Table 12.1 Material properties and the calculated thickness for 50% attenuation for thermal
neutrons
Molecular Macroscopic Thickness for
Chemical Density weight cross-section 50% attenuation
Material formula (g cm–3) (g mol–1) (cm–1) (cm)
Aluminum Al 2.70 27 0.0984 7.05
Styrene C8H8 0.9 106 3.12 0.22
Water H2O 1.0 18 2.62 0.26
Silica SiO2 2.65 60 0.259 2.68

Neutron radiography experiments were performed at the research facility

located in the McClellan Nuclear Radiation Center (MNRC) in the United
States [6]. As shown in Fig. 12.2, MNRC has a two megawatt TRIGA reactor
with four neutron beam tubes all tangential to the reactor core. The neutron
beam is approximately 9 inches (22.5 cm) in diameter and has an intensity of
approximately 107 n cm–2 s–1. The neutron spectrum is highly thermalized, which

(a) (b)

(c) (d) (e)

Fig. 12.2 (a) Neutron radiography facility at MNRC. (b) Casting setup and image intensifier.
Three types of gating systems: (c) Top-gating, (d) side-gating, and (e) bottom-gating
214 D. Penumadu

allows high-quality imaging. The neutron beams are collimated with an L/D ratio
of 50–400, resulting in good spatial resolution for digital still radiographs. The
radioscopic systems use a CF Thompson tube as the neutron camera, which
combined with a SIT (silicon intensified tube) camera allows image capture at a
rate of 30 frames/s. The radiography setup for LFC is shown in Fig. 12.3. The
polymer foams used in the study are EPS foams with a density of 22.4 kg/m3.
The casting metal used in the study was aluminum alloy. This was heated to
14008 F/7608C before pouring into the EPS patterns. Real-time neutron radio-
graphy was used to visualize the molten metal and pyrolysis foam fronts during
the whole LFC process. The casting samples were plates of size 100  150 mm
with varying thicknesses (4, 12, and 24 mm), and various gating orientations
(bottom, side, and top), as shown in Fig. 12.2.

(a) (b)

Fig. 12.3 Bottom-gating casting with pattern thickness of 4 mm. (a) Neutron radiograph
(at 19 s). (b) Optical picture of the final casting

Real-time digital videos were recorded during the entire process of metal entry,
foam decomposition, and casting solidification. Specific images of interest were
then extracted from the digital videos, further processed, and analyzed using
Image-Pro Plus V4.5 [7]. Figure 12.3 shows an example neutron radiograph of a
bottom-gated casting that did not fill completely at 19 s since the hot aluminum
metal was introduced to the down-sprue. There is a wealth of information in such
radiographs as a function of time that is valuable for understanding the pyrolysis
aspects of foam and the associated phase transitions of both polystyrene (from
foam to liquid to gaseous monomer) and metal (liquid to solid state).
Only neutrons provide this outstanding ability to visualize and model complex
multiphase and multiscale processes, penetrating metal and at the same time
clearly showing light elements such as the byproducts of polymer degradation.
The neutron images shown in Fig. 12.4 indicate the effect of foam thickness on
the degradation process during LFC. It is of great interest to note that for thinner
12 Material Science and Engineering with Neutron Imaging 215

Fig. 12.4 The effect of EPS foam pattern thickness on degradation; 4-mm thick plate (top) and
24-mm thick (bottom)

foam patterns, one does not see diffusion pathways, as was noted for thicker
foam, and often thicker castings have defects from incomplete pyrolysis of the
foam in certain regions. Data of this nature are of profound significance for
understanding the physics of the problem related to multiphase dynamic flow and
transport. These data can be used to develop suitable engineered solutions to
produce near-net-shaped castings with lower scrap rate. The spatial and temporal
information on a given phase during casting solidification is valuable for model-
ing and verification of the dynamic process using computational fluid dynamic
(CFD) codes such as Flow3D1 and Magma1.

12.3 Particulate Materials and Porous Media

The stress–strain relationship of particulate materials is complex and

depends on the initial state of packing, the past stress history, and the
applied stress path (compression, shear, or extension). This section illustrates
216 D. Penumadu

the development of an experimental approach using neutron imaging to

provide a fundamental understanding of the macroscopic stress and strain
(including the meaning of the terms ‘‘stress’’ and ‘‘strain’’ in particulate
materials) and its relation to the microstructure (particle contacts and void
size distribution). Images were acquired using the CONRAD instrument at
the Berlin Neutron Scattering Centre in tomography mode with a white
beam cold source [8].
The response of a granular assembly to a global applied force (usually
applied via a loading platen) is an important topic of interest in the study of
particulate materials. The commonly observed global stress–strain response
of granular materials must be related to the local force and displacement
variations at the scale of the particles. For continuum materials, the stress–
strain relation can be well described macroscopically, based on continuum
mechanics. However, for a granular assembly it is well known that the
macroscopic stress–strain relation is quite complex and depends on the
initial state of the assembly (local and global porosity and particle coordina-
tion numbers), the past stress history, and the loading path. Granular
assemblies show anisotropy, inhomogeneity, and nonlinearity in the distri-
bution of both the contact network and the contact forces among particles
(see references [9–12]). Typical X rays (from table-top sources rather than
synchrotrons) have difficulty in penetrating silica sand for more than a few
millimeters and are not sensitive to the degree of saturation with water.
Neutrons are ideal for such applications, where the intent is to characterize
the initial geometric arrangement of particles and its evolution as a function
of applied stress.
Figure 12.5 shows images of two slices of a silica sand core, obtained from a
tomographic reconstruction using a high-resolution neutron imaging setup. It is

Fig. 12.5 Neutron images of a sand core specimen ( 1  1.24 cm)

12 Material Science and Engineering with Neutron Imaging 217

relatively easy to identify individual grains; the particles shown are mostly
smaller than 800 mm. The sand core, obtained from a foundry, was produced
by a specialized technique which utilizes a vaporized or gaseous chemical that
operates as a catalyst or co-reactant in order to cure (polymerize) a resin binder
film. The resin is spread uniformly over the surface of the sand and compacted
into a mold (rectangular in the study illustrated). When the coating has fully
cured, each sand grain is firmly joined to the others in contact with it. An
understanding of the stability and durability of sand cores is important for
the foundry industry, but a realistic fundamental model of core behavior
from particle scale mechanics and numerical modeling is still a dream.
Neutron imaging offers the unique possibility to provide actual information
on particle contacts and voids with appropriate resolution. Using suitable
thresholding techniques with advanced image processing algorithms, it will
in the future be possible to describe the three-dimensional void and solid
particle fabric and its evolution as a function of externally applied mechan-
ical or thermal stress.
Many practical issues, such as the fate of contaminants in the sub-surface of
the earth, or of nutrients for plant systems (Chapter 17), can benefit from
information that is uniquely obtainable from neutron imaging. Using H2O
and D2O alternately, or in clever miscible modes, neutron imaging experiments
can be designed to extract information on porous media filled with fluid and gas
phases, and the arrangement of solid phases, with relatively good precision and
ease. They can also provide much needed information for the study of partially
saturated porous media and suction characteristic curves that are presently
unresolved.
The micromechanical analysis of granular assemblies using computer
simulations, through numerical methods such as the distinct element
method (DEM), can be evaluated by comparing the results for discrete
particulate model systems with carefully controlled neutron imaging
experiments that involve a mechanism to apply a known state of stress
or strain using a portable loading system. For simulations, the material
microproperties such as particle shear and tangential stiffness can be
evaluated from nanoindentation test results on individual particles. The
particle contact behavior, together with the nature of force chain structure
and other related micromechanics associated with one-dimensional com-
pression on granular materials having different shape and mechanical
properties, is then analyzed in conjunction with experimental data gener-
ated using both neutron diffraction and imaging. The goal of the DEM is
not to match the macroscopic response but to use the fundamental proper-
ties of the particles to predict the behavior of the assembly, and to
evaluate the basic mechanisms contributing to the strength and deforma-
tion behavior of particulate systems in terms of particle force-chain struc-
ture (Fig. 12.6).
218 D. Penumadu

Fig. 12.6 Example force distribution of a granular assembly under a punch obtained using a
particle flow code

12.4 Stroboscopic and Dynamic Neutron Imaging

There are some common engineering applications that require continuous fast
imaging with neutrons to obtain detailed information related to a fast process.
To solve such problems, major difficulties must be overcome related to the low
number of neutrons detected in one measurement time frame and the number of
pixels that can be read. To meet this challenge requires a high-flux neutron
source, low decay time neutron scintillation materials, and the use of strobo-
scopic imaging for fast but periodic processes. Using the stroboscopic
approach, the number of neutrons in one time window is still very low, but
many exposures of the same time window of a periodic process may be accu-
mulated on the detector before readout, thus increasing the available intensity
(though not improving the signal-to-noise ratio).
The stroboscopic technique can be illustrated by a study of fuel injection and oil
flux in a combustion engine. A commercial combustion engine was stripped of the
intake and exhaust system and coupled to a 2 kW electric motor with a transmis-
sion belt [13]. The spark plugs were also removed, so that the engine would not
generate compression in its cylinders. The electric motor was capable of turning the
engine at low rotation speeds and, due to the reduced heat production, the water
cooling could be omitted. A Hall sensor close to the transmission wheel on the
camshaft generated a pulse which was used to trigger the image intensifier of a
cooled CCD camera. The trigger signal could be delayed in order to shift the time
window within the cycle of the engine. Measurements were made at two facilities:
the Neutrograph facility at the Institut Laue-Langevin (ILL) reactor in Grenoble
with a high thermal neutron flux of 3 109 n cm–2 s–1and low L/D value of 170,
12 Material Science and Engineering with Neutron Imaging 219

and the ANTARES facility at the FRM-II reactor of the Technische Universitat
Munchen with a lower thermal neutron flux of 1108 n cm–2 s–1 but a high L/D
value of 400. At ILL, the detection system was a MCP-intensified CCD camera
with a Peltier-cooled chip (13001024 pixels) with 16 bit digitization.
The full cycle of the four-stroke engine running at 1000 rpm was split into 120
individual frames over two rotations; 150 individual images were recorded as an
on-chip accumulation of 200 ms exposure each. In this way, the total exposure
time for the full run was only of order 18 min. The field of view was 2424 cm.
At FRM-II the measurements were repeated with an Andor MCP intensified
CCD camera with 10241024 pixels. Because of the lower available flux, the
time window was extended to 1 ms and the rotation speed was reduced to
600 rpm. The images thus show more motional blurring but better definition
of the stationary parts. Figure 12.7 shows typical results that demonstrate the
differences in spatial and temporal resolution for two facilities related to strobo-
scopic imaging. As can be seen, the lubrication of a running engine can be
visualized; this can have important applications related to the design and
modeling of a future automotive powertrain.
Possible options for detection and triggering systems for short-time strobo-
scopic neutron imaging include LCD shutters, interline CCDs, CCDs with a
gated image intensifier, and electronically shuttered CCDs [14]. Stroboscopic
neutron radiography with a time resolution down to a few microseconds, as well
as dynamic tomography on a sub-minute time scale, can allow the visualization
of thick and strongly absorbing materials [15], thus providing the ability to also
investigate large samples with extremely low contrast as required in some fields
of engineering.

Fig. 12.7 Dynamic radiography of a running engine, (a) acquired at the Neutrograph facility
and (b) acquired at the ANTARES facility [13]
220 D. Penumadu

The requirements for dynamic imaging are clearly more severe since the
number of neutrons recorded in a single time frame is very low, even for the
highest flux sources, so a low background is extremely important as well as a
fast frame rate. A recently developed prototype detector for this type of appli-
cation is based on electron multiplying charge-coupled device (EMCCD) tech-
nology [16]. This CCD technology evolved from the original development of
the high-resolution -ray imaging detector by RMD [17], which uses a back-
illuminated, thermoelectrically cooled, 512  512 pixel EMCCD as a readout
sensor coupled to a LiF/ZnS:Ag scintillator. The scintillator screen is 225 mm in
thickness and provides bright emission for short exposure times (2 ms) with
an imaging area of  25  25 mm2 and an effective pixel size of  50 mm.
The camera permits a minimum of 2 ms data acquisition and can provide
30 frames per second (fps) imaging at full pixel resolution. In binned mode,
the frame rate can be increased to 225 fps. While the internal gain effectively
offsets read noise, reducing it to < 1 e– pixel–1 s–1, the dark noise is also
minimized to < 1 e– pixel–1 s–1 by cooling the EMCCD to –358C.
Using this detector, dynamic imaging in integrated mode was conducted at
the National Institute of Standards and Technology reactor facility using the
thermal neutron port BT2. Figure 12.8 shows neutron images of water droplets
from a plastic bottle with a small hole, imaged at 30 fps to compensate for the
low flux and relatively low detection efficiency of the LiF/ZnS:Ag screen. It is
important to consider the scattering effects for such images using tools such as

Fig. 12.8 Dynamic imaging (single frames) of a water droplet using the EMCCD detector
12 Material Science and Engineering with Neutron Imaging 221

the Monte Carlo method to extract contrast from linear attenuation asso-
ciated with absorption versus scattering.
The ability to use EM CCD technology now opens up many possibilities for
pseudo real-time neutron imaging applications that require the study of flow or
transport of fluids through porous media or channels. The following are examples:
 optimization of the channel geometry for microfluidic-based devices for
thermal management of high-performance electronic chips and related
applications,
 optimization of catalytic converters,
 understanding of physico-thermal issues related to the conversion of low-
yield hydrocarbon-rich geo-materials for next-generation oil extraction,
 study of the environmental degradation of polymeric composites due to
moisture and humidity exposure, and
 development of novel drainage materials for hydraulic management and
filtration applications.

12.5 Residual Stress Imaging

Residual stress is one of the more common phenomena occurring in materials.

It arises during processes such as machining, grinding, rolling, heat treatment,
and welding. Residual stress can be defined as self-equilibrating internal stress
existing in a free body without constraints or external force added on the
boundary, or as an elastic response of the material to an inhomogeneous
distribution of inelastic strains such as plastic strains, precipitation, phase
transformation, misfit, or thermal expansion strain. It can originate from
shape misfit between the unstressed shapes of different parts, regions, or phases
of components. Residual stress has a significant impact on the mechanical
behavior and durability of materials. It is not immediately apparent in many
cases, as it is difficult to measure or predict, and can lead to premature failure of
materials if not considered in the design.
The application of neutron diffraction to measure/image stresses and strains
in metallic and other materials under in situ and ex situ loading conditions is
now well established and has found important applications in many fields of
engineering (see, e.g., [18]). The approach is to use a fixed-wavelength neutron
source (typically at a reactor) or a multi-wavelength pulsed source (typically at a
pulsed spallation source) to obtain a diffraction pattern for a gauge volume
(
x
y
z) which is typically in the range of 0.5–1000 mm3. Analysis of the
diffraction pattern, using single Bragg peak fit or full pattern Rietveld refine-
ment, gives an accurate value for the inter planar spacing, d, from which the
lattice strain e can be determined:

dc ðx; y; zÞ  d0
"c ðx; y; zÞ ¼ ; (12:1)
d0
222 D. Penumadu

where d0 is the inter planar spacing under stress free conditions. e is

then determined as a function of position (x,y,z) and orientation ( ,’) in
the specimen by translation/rotation through the neutron beam, and the
strain tensor
2 3
"xx "xy "xz
6 7
4 "xy "yy "yz 5 (12:2)
"xz "yz "zz

at each location can be determined from measurements in several orientations

( ,’) (a minimum of six in the general case) by applying the relation

"c ¼ "xx cos2  sin2 c þ "xy sin 2 sin2 c

þ "yy sin2  sin2 c þ "zz cos2 c (12:3)
þ "xz cos  sin 2c þ "yz sin  sin 2c :

The stress tensor is then derived from the strain tensor using Hooke’s law.
A particular advantage of neutrons for residual stress imaging is the ability
to penetrate deep into large components, nondestructively, as illustrated in
Figs. 12.9 and 12.10. Hard (high energy) X rays can now offer similar penetra-
tion (depending on the material) but the relevant diffraction angles are then
very small, so the gauge volumes are ‘‘needle-like,’’ which is problematic for

Fig. 12.9 Wingspar from an Airbus A340 aircraft mounted for stress imaging on the Engine-X
diffractometer at the ISIS spallation neutron source
12 Material Science and Engineering with Neutron Imaging 223

Fig. 12.10 Image of the stress distribution in the head of a railway rail (see e.g., [19])

measuring one component of the stress. The two techniques are therefore highly
complementary and are increasingly used in combination.

12.6 Bragg Edge-Based Energy Selective Neutron Imaging

The principles of energy-dispersive and Bragg edge neutron imaging have been
described in Section 4.1 in Chapter 6 of this book. When these are fully imple-
mented at pulsed neutron sources, they will enable new and exciting applications
of energy-selective neutron radiography and tomography for various aspects of
materials science and engineering. However, there are significant challenges for
detector development since both high spatial resolution and high time resolution
(to measure the neutron wavelength via its time of flight) are required; to refine
the position of a Bragg edge to an accuracy of 5  10–5, equivalent to 50 micro-
strain, needs a time resolution of the order of 10 ms. Current detectors with an
acceptable spatial resolution of 50 mm and reasonable duty cycle can only
achieve 100–1000 ms. Also, no beamlines have yet been built that are optimized
specifically for this type of measurement, but some experiments have been
carried out that fully demonstrate the potential of the technique.
The transmission cross section as a function of wavelength shown in Fig. 1.4 of
Chapter 1 is for an ‘‘ideal’’ polycrystalline iron powder. In a ‘‘real’’ solid metal
component, the sharp Bragg edges can either be shifted due to residual stress, or
change in height due to texture (nonisotropic distribution of crystallite orientation).
If the edge shifts can be measured with sufficient spatial and wavelength resolution,
then they can be used to image the strain or stress, as is shown in Fig. 12.11.
224 D. Penumadu

Fig. 12.11. Image of the residual elastic strain distribution around a cold expanded hole in a
12-mm thick steel plate. The measured strains correspond to the through-thickness average of
the out-of-plane strains. The in-plane spatial resolution  22 mm2. [20]

In practice there can often be a combination of edge shift and height change,
averaged over the crystallites within the gauge volume, leading to significant
broadening and distortion of the sharp Bragg edges. Two samples of the same
material can therefore have quite different transmission spectra, which can be used
as a signature of the method used to create the samples, as illustrated in Fig. 12.12.
The Bragg edges also obviously differ between the different materials that
might make up a macroscopic system (Fig. 12.13). The ratio of images mea-
sured below and above a Bragg edge for one of the component materials gives a
good contrast for that material, while the other material effectively disappears
(provided that it does not also have a Bragg edge in the same wavelength
region).
In principle, full tomographic imaging can be carried out as a continuous
function of wavelength, with the possibility to, then, reconstruct images with
highlighted contrast to illustrate stress, texture, composition, or even complex
combinations. Currently, the available neutron fluxes and detector technolo-
gies mean that the resolution has to be relaxed in some respect to achieve
reasonable images; for example, either positional resolution is relaxed to
achieve high time resolution for high Bragg edge contrast, as in Fig. 12.13,
or wavelength resolution is relaxed to achieve higher spatial resolution, as in
Fig. 12.14. However, with the advent of higher-flux neutron sources, such as
12 Material Science and Engineering with Neutron Imaging 225

Fig. 12.12 Neutron

transmission spectra of the
following two bronze
samples: one an
archaeological specimen and
the other a cast piece. The
significant differences
indicate that the ‘‘Celtic
handle’’ has been worked
after casting [21]

Fig. 12.13 Neutron transmission images for a copper cored iron slug, with Bragg edge
filtering used to enhance the contrast for either copper or iron [22]
226 D. Penumadu

Fig. 12.14 Neutron transmission radiographs of a steel welding as a function of neutron

energy, which was measured using two different detectors systems, PI-MAX (upper row) and
ANDOR (lower row) [22]

the Spallation Neutron Source in the United States, and/or the design of
specialized instruments using advanced neutron optics and detectors, such
as the IMAT instrument proposed for the ISIS source in the United Kingdom,
these restrictions are likely to be overcome within the next 5 years. Then, this
will open the way for a much broader range of applications of neutron
imaging in materials science and engineering.

References
1. J. Campbell, The concept of net shape for castings. Mater. Design 21(4), 373–380 (2000).
2. X. Yao and S. Shivkumar Mould filling characteristics in lost foam casting process. Mater.
Sci. Technol. 13, 841–846 (1997).
3. R. E. Turner, Radiography: Major contribution to the metalcasting industry. Modern
Casting 61(1), 44–45 (1972).
4. B. Sirrell, M. Holliday, et al., The Benchmark Test 1995. 7th Conference on Modelling of
Casting. Welding and Advanced Solidification Processes, London, UK.
5. M. Cox, R. A. Harding, et al., Optimised running system design for bottom filled alumi-
nium alloy 2L99 investment castings. Mater. Sci. Technol. 19(5), 613–625 (2003).
6. X. Luo. Study on Infrastructure Materials by Using Neutron Radiography and Diffraction.
PhD dissertation, University of Tennessee, Knoxville, TN, USA (2007).
7. X. Luo and D. Penumadu, Visualization of the multiphase flow in the lost foam casting
process by neutron radiography and image processing. Automotive Light Metal Castings:
Technology and Applications, Materials Science and Technology (MS&T), Detroit, MI,
USA (2007).
8. D. Penumadu and N. Kardjilov, High resolution neutron tomography of sandcores used in
metal casting (2008). To be submitted.
12 Material Science and Engineering with Neutron Imaging 227

9. T.S. Majmudar and R.P. Behringer, Contact force measurements and stress-induced
anisotropy in granular materials. Nature 435(7045), 079–1082 (2005).
10. N.P. Kruyt, Contact forces in anisotropic frictional granular materials. Int. J. Solids
Struct. 40(13–14), 3537–3556 (2003).
11. M. Otto et al., Anisotropy in granular media: Classical elasticity and directed-force chain
network. Phys. Rev. E 67(3), 031302-1 to 24 (2003).
12. P. Lehmann, P. Wyss, et al., Tomographical imaging and mathematical description of
porous media used for the prediction of fluid distribution. Vadose Zone J. 5(1), 80–97
(2006).
13. B. Schillinger, J. Brunner, and E. Calzada, A study of oil lubrication in a rotating engine
using stroboscopic neutron imaging. Physica B 385–386, 921–923 (2006).
14. B. Schillinger, H. Abele J. Brunner, G. Frei, R. Gahler, A. Gildemeister, A. Hillenbach,
E. Lehmann, and P. Vontobel, Detection systems for short-time stroboscopic neutron
imaging and measurements on a rotating engine. Nucl. Instrum. Meth. Phys. Res. A 542,
142–147 (2005).
15. A. Hillenbach, M. Engelhardt, H. Abele, and R. Gahler, High flux neutron imaging for
high-speed radiography, dynamic tomography and strongly absorbing materials. Nucl.
Instrum. Meth. Phys. Res. A 542, 116–122 (2005).
16. V.V. Nagarkar, I. Shestakova, S. C. Thacker, S. R. Miller, D. Penumadu, J. F. Ankner,
H. Z. Bilheux, and C. E. Halbert, Time-resolved high resolution neutron imaging studies
at the ornl spallation neutron source. Session N14-4, Nuclear Science Symposium,
Honolulu, Hawaii, USA, Oct. 27-Nov. 3 (2007).
17. V.V. Nagarkar, I. Shestakova, V. Gaysinskiy, S.V. Tipnis, B. Singh, W. Barber,
B. Hasegawa, and G. Entine, CCD-Based Detector for SPECT. IEEE Trans. Nucl. Sci.
53(1), 54–58 (2006).
18. M.T. Hutchings, P. J. Withers, et al., Introduction to the Characterization of Residual
Stress by Neutron Diffraction. Boca Raton, CRC Press (2005).
19. P. J. Webster, X. Wang, G. Mills, G. A. Webster, Residual stress changes in railway rails.
Physica B 180–181, 1029 (1992).
20. J.R. Santisteban, L. Edwards, M.E. Fizpatrick, A. Steuwer, and P.J. Withers, Appl. Phys.
A 74, S1433–S1436 (2002).
21. J.R. Santisteban, S. Siano, and Daymond, M.R, Neutron strain scanning of archaeolo-
gical bronzes, Mater. Sci. Forum 524–525, 975–980 (2006).
22. W. Kockelmann, G. Frei, E.H. Lehmann, P. Vontobel, and J.R. Santisteban, Energy-
selective neutron transmission imaging at a pulsed source. Nucl. Instrum. Meth. Phys.
Res. A 578, 421–434 (2007).
Chapter 13
Novel Neutron Imaging Techniques for Cultural
Heritage Objects

C. Andreani, G. Gorini, and T. Materna

Abstract The use of neutrons for cultural heritage (CH) research is illustrated
with special reference to neutron tomography (NT) methods, providing three-
dimensional (3D) images of neutron attenuation, and the analysis techniques
known as prompt gamma-ray activation analysis (PGAA) and neutron reso-
nance capture analysis (NRCA), providing the elemental composition of
an object. PGAA and NRCA are well-established nondestructive methods
for bulk analysis of CH objects, with sensitivities that can reach the
parts-per-million range. By improving the spatial resolution of PGAA and
NRCA it will be possible to measure the composition of small parts inside a
large object or even to provide a full 3D map of the elemental composition of
an artifact. The imaging techniques under development are called prompt
gamma-ray activation imaging (PGAI), neutron resonance capture imaging
(NRCI) and neutron resonance transmission (NRT) tomography. The
NRCA experience at the GELINA neutron source is the starting point for
the development of NRCI/NRT now taking place at the 100 times more power-
ful ISIS pulsed neutron source.
PGAI, NRCI, and NRT are unlikely to achieve the spatial resolution
obtained by conventional NT. It is the combined use of these new imaging
techniques with neutron and X-ray tomographies that is proposed here as the
way forward in neutron imaging. There are parallel developments in the use of
Bragg edges to improve the elemental sensitivity of NT (dichromatic tomogra-
phy). Another foreseen development is 3D phase imaging (neutron diffraction
imaging, NDI), which can essentially be undertaken using existing diffraction
instruments. The ongoing developments are promising and suggest that neu-
tron-based techniques will provide the basis for integrated analysis protocols
leading to important advances in the scientific characterization of the materials
that constitute a CH object.

C. Andreani (*)
University of Milano-Bicocca, Department of Materials Science, Via Roberto
Cozzi 53, 20125 Milano, Italy

I.S. Anderson et al. (eds.), Neutron Imaging and Applications, 229

Neutron Scattering Applications and Techniques, DOI 10.1007/978-0-387-78693-3_13,
Ó Springer ScienceþBusiness Media, LLC 2009
230 C. Andreani et al.

Keywords X-ray and neutron scattering  Neutron diffraction  Neutron

imaging  Ancient Roman marbles  Textures  Experimental methods of
materials testing and analysis

13.1 Imaging, Neutrons, and Cultural Heritage

The use of X-ray and neutron facilities for cultural heritage (CH) research is
rapidly growing [1, 2]. Detailed three-dimensional (3D) maps of neutron and
X-ray attenuation in CH objects are produced by the tomographic methods
described earlier in this book. X-ray and neutron facilities also provide a
number of nondestructive analytical techniques that are used in the CH
field. X-ray fluorescence (XRF) is a well-known example, in particular
because of the widespread use of portable XRF devices, a clear advantage
for analysis of valuable objects at (or close to) museums. More generally, the
application of nuclear-based techniques to CH research has expanded in recent
years as laboratories with small ion accelerators have been installed close to
museums, facilitating the study of CH objects by particle-induced X-ray and
gamma-ray emission (PIXE and PIGE) and Rutherford backscattering [3].
With XRF and PIXE, imaging is possible down to micrometer resolution; but
because both techniques are based on the detection of low-energy X-rays, only
information on the surface of an object is retrieved. Bulk analysis techniques are
alternatively available (see below) but do not offer a spatial resolution adequate
for measuring the composition of small parts inside a large object well. Ideally,
one would like to combine the spatial resolution of tomographic methods with the
power of nondestructive analytical techniques to provide a 3D map of the
elemental composition of a large object. Can neutrons be used for such a challen-
ging task? The question is addressed in this chapter by first reviewing some general
features of neutron (and X-ray) techniques as relevant for CH studies.
X-rays and neutrons conveniently complement each other in CH research. This
is because CH objects can be made from diverse materials sampling the Mende-
leev table at both low and high atomic numbers, making the attenuation of X-rays
and neutrons very different. X-rays have a lower penetration depth at high atomic
numbers. Thus X-ray radiography (XR) is suitable for probing the interior of
objects made from wood or bone, but it has problems with bulky metals such as
gold, silver, and lead, which are nearly opaque to X-rays. On the other hand,
neutron radiography (NR) can probe metal artifacts (e.g., copper, tin, iron,
bronze, lead) but does not penetrate thick layers of organic materials because of
strong neutron beam attenuation by hydrogen. An introduction to NR and its
applications can be found elsewhere in this book.
Neutron tomography (NT) is performed by taking a large set of radiographs
of an object at different orientations and reconstructing all data offline in a
virtual 3D image. For further details, see Chapters 6 and 7 in this book. Only a
few NT setups exist in the world (see [4] for a review of facilities in Europe)
because of the high neutron fluxes required to perform a full scan of an object in
13 Novel Neutron Imaging Techniques for Cultural Heritage Objects 231

a reasonable time. This has not prevented the successful application of NT to

CH objects. Generally CH objects benefit from NT developments driven,
example, by engineering applications. Figure 13.1 shows an example where
the image contrast of an ancient Roman brooch is improved by using a cold
neutron beam: the proportionality of neutron capture cross section to neutron
wavelength is exploited to select the neutron wavelength that best matches the
thickness and composition of the object. Imaging CH objects with neutrons
offers unique opportunities: it permits the search for hydrogenous materials
inside sealed metal objects, which is impossible with X-rays; it allows differ-
entiation between elements close in the Mendeleev table, which would show the
same contrast to X-rays; and finally, it enables differentiation between isotopes.
The elemental information is inferred from the neutron attenuation. A typical
example is a metal artifact, such as a statuette or a vase, in which different
materials can be found in inserts or welds and provide useful information on the
fabrication methods.

Fig. 13.1 Radiographs of an

ancient Roman brooch taken
with (a) cold neutron (5.5 Å)
using a velocity selector at
the Prompt Gamma-Ray
Activation (PGA) station of
the Paul Scherrer Institut
(PSI), Switzerland [5] and
(b) standard thermal (1.8 Å)
neutron radiography
technique at the Neutron
Transmission Radiography
(NEUTRA) station at PSI [6] (a) (b)
232 C. Andreani et al.

In the cold/thermal neutron energy range, the elemental sensitivity of

neutron imaging can be enhanced using Bragg cutoffs (or Bragg edges). At a
so-called Bragg cutoff wavelength, the neutron attenuation coefficient drops
drastically. If the drop in attenuation is important and one can monochroma-
tize the beam and select its wavelength, this behavior can be exploited to change
the material contrast. Indeed, if two radiographs are taken, one with a wave-
length just below the Bragg cutoff of a selected element and one with a wave-
length just above it, the ratio will be sensitive to this element only. On the other
hand, if two wavelengths are selected where the attenuation coefficient is equal,
the element disappears in the ratio of the two radiographs as well as in the
tomographic result (Fig. 13.2) [5]. This dichromatic tomography method has
been demonstrated both at reactor sources and at pulsed neutron sources. It
is similar to what can be done with X-rays using the rise in attenuation at the
K-edge energy [7] (Fig. 13.3). Both methods can in principle provide a 3D map
of a selected element. However, not all elements have Bragg edges, and only
elements with high atomic number, which have the K-edge at high X-ray
energy, can be imaged in large samples. Thus the use of Bragg edges and
K-edges represents a step forward, but with limited scope.

(a) (b) (c)

Fig. 13.2 (a) Neutron radiograph of a spark plug at 6.9 Å; (b) result of the division of two
radiographs, one at 6.9 Å and the other at 3.2 Å, showing that the steel cladding (central part)
became transparent; (c) 3D result of dichromatic tomography. Detailed explanations can be
found in [5]
13 Novel Neutron Imaging Techniques for Cultural Heritage Objects 233

(a) (b) (c)

Fig. 13.3 A gold-sensitive tomography of an ore of Ashanti (Ghana) origin performed at the
high-energy beam line (ID15) of the European Synchrotron Radiation Facility (Grenoble,
France) shows that thin (about 40 mm) but extended layers (gold veins) appear in the quartz
beads [7]: (a) shows a horizontal cut in the tomography made at 80 keV, where the gold veins
and the quartz matrix are visible on top of ring-shaped reconstruction noise; (b) and (c) are
two different 3D views of the result of the dichromatic tomography, sensitive to gold only.

Bragg edges may be more useful for CH applications because of the texture
information embedded in their detailed shapes. Texture in the phases distorts
the Bragg edges, so it can be used as a fingerprinting technique with the
potential for texture mapping [8]. Appropriate transmission detectors need to
be developed before this technique can be used routinely at pulsed neutron
sources. On the other hand, phase-sensitive neutron diffraction (ND) tech-
niques are in regular use for phase and microstructure characterization of
ceramic and bronze artifacts, including texture analysis (see below). Texture
provides important clues to the deformation history and thus to historic pro-
duction steps, a prominent example being the neutron texture analysis of the
early Copper Age Iceman axe [9].
Three features of neutron interactions with nuclei are exploited to provide
analytical tools to determine the elemental compositions of CH objects:
1. radioactive nuclei produced by thermal/cold neutron capture,
2. prompt gamma rays from thermal/cold neutron capture, and
3. resonant neutron absorption at element-specific neutron energies in the
epithermal energy range.
The first feature has long been exploited by a method known as neutron
activation analysis (NAA). It started with the detection of radioactive isotopes
through their half-lives. Now, these isotopes are observed and quantified with
high-resolution gamma-ray detectors on the basis of gamma-ray energies as
well as gamma-ray half-lives. This technique, known as instrumental neutron
activation analysis (INAA), is now a valuable analytical tool for CH studies in
use at many facilities [10, 11]. INAA may be invasive because it may require the
234 C. Andreani et al.

preparation of a small sample (e.g., powder) of the object before irradiation in a

reactor. A related noninvasive imaging method is neutron activation autora-
diography (NAAR) [12], in which two-dimensional (2D) objects are exposed to
cold neutrons and the map of the neutron-induced gamma activation is
recorded at different times by radiographic methods. NAAR is limited to
2D objects but is a unique tool for the investigation of paintings.
For the noninvasive elemental imaging of 3D objects, features (2) and (3) in
the list above, which provide the basis for recently developed analytical meth-
ods, should be considered. Prompt gamma-ray activation analysis (PGAA) is
based on the detection of characteristic gamma-ray energy spectra during
thermal/cold neutron capture. It provides a nondestructive determination of
the elemental (isotopic) composition of CH objects with good sensitivity to both
major and trace elements. Mainly developed by G. Molnár and coworkers at
the Budapest research reactor [13], PGAA has found many applications and has
reached a high level of sophistication. Neutron resonance capture analysis
(NRCA) was developed as a new analytical tool by H. Postma and coworkers
at the GELINA pulsed neutron source [14]. NRCA makes use of the unique
resonance absorption properties of epithermal neutrons. Many elements found
in common CH materials have neutron absorption resonances in the energy
range below 1 keV, including copper, cobalt, silver, tin, antimony, and gold.
The gamma rays from resonant neutron capture are detected as a function of
time, that is of neutron velocity. The time spectrum has emission peaks at
element-specific resonance energies. Calibration of the gamma-ray detector
array using reference samples is used to convert the resonant gamma-ray yield
into absolute elemental composition. An alternative approach is also possible in
which the neutron transmission through an object is measured instead of the
gamma-ray emission. The transmitted neutron spectrum of a pulsed neutron
beam has characteristic resonance dips in the time spectrum that can be used to
determine the elemental composition of the object in the same way as the
resonance peaks of NRCA.
PGAA and NRCA are well-established nondestructive methods for bulk
analysis of CH objects, with sensitivities that can reach the parts-per-million
range. There is no important reason why the spatial resolution of PGAA,
NRCA, and neutron transmission methods should not be improved, eventually
leading to nondestructive measurement with good spatial resolution of the
composition (major and trace elements) of small parts inside a large object.
These 3D extensions of PGAA and NRCA are presented in this chapter as a
new development that can fill the gaps between 3D imaging, surface analyses,
and bulk analyses. In the following discussion, the 3D extensions will be called
prompt gamma-ray activation imaging (PGAI), neutron resonance capture
imaging (NRCI), and neutron resonance transmission (NRT) tomography.
PGAI, NRCI, and NRT are unlikely to achieve the spatial resolution obtained
by conventional NT. It is the combined use of these new imaging techniques
with neutron and X-ray tomographies that we foresee as the way forward in
neutron imaging. Another development foreseen is 3D imaging of crystalline
13 Novel Neutron Imaging Techniques for Cultural Heritage Objects 235

phases (neutron diffraction imaging, NDI), which can essentially be undertaken

using existing diffraction instruments. These methods together are expected to
provide a new, comprehensive neutron-based approach that could be used for
3D imaging of the elemental and phase compositions of CH objects.
Interestingly, this is a rare case of a new development in neutron instrumen-
tation that is motivated by CH applications. The new imaging techniques are
being developed within the Ancient Charm (AC)1 project, a collaboration of
ten universities, central laboratories, and cultural heritage institutions across
Europe [15]. A technical challenge of the project is represented by the need to
perform high-efficiency gamma-ray measurements in a neutron environment,
with the associated background, and to extend spatially resolved neutron
transmission measurements into the epithermal range. The underlying tech-
niques are presented in more detail in this chapter together with highlights of
recent developments and CH applications.

13.2 Two Neutron Beam Analytical Techniques: Neutron

Resonance Capture Analysis and Prompt Gamma-Ray
Activation Analysis
13.2.1 Neutron Resonance Capture Analysis

Neutron resonance capture analysis makes use of the unique resonance absorp-
tion properties of epithermal neutrons for analyzing the bulk composition of
materials and objects. The absorption cross sections of most elements are well
known; they vary from one element to another and actually vary between
isotopes of the same element [16]. Neutron absorption is followed by the prompt
emission of a gamma-ray cascade with total cascade energies of up to about
8 MeV. The detailed energy spectrum of the gamma emission is also well known
in the case of thermal neutron capture, but it can vary among resonances. NRCA
is used at the GELINA pulsed neutron source for analysis of CH artifacts to
determine the concentration of neutron-sensitive elements [17–25].
So far NRCA has been applied to a large number of artifacts, mainly bronze
objects. A typical NRCA spectrum looks like the one in Fig. 13.4 [25] obtained
by exposing an Etruscan votive object to the GELINA neutron beam.
A number of resonance peaks of different elements (copper, lead) are clearly
visible, corresponding to neutron energies in the range from 3,250 to 3,650 eV.
The energies and intensities of the resonance peaks characterize the elemental
composition of the sample. Figure 13.4 shows the 3,357 eV resonance of
lead-206 and three copper resonances at 3,310, 3,503, and 3,588 eV. Because

1
Ancient Charm (Analysis by Neutron Resonant Capture Imaging and other Emerging
Neutron Techniques: Cultural Heritage and Archaeological Research Methods), http://
ancient-charm.neutron-eu.net/ach.
236 C. Andreani et al.

Fig. 13.4 Example of neutron resonance capture analysis NRCA spectrum featuring three
copper resonances and one lead resonance in the region of 3,250–3,650 eV used to determine
the Pb/Cu weight ratio of an Etruscan votive object [25]

these four resonances are weak, the Pb/Cu weight ratio can be obtained after
small self-shielding corrections. A set of Etruscan statuettes from a collection
originally owned by Earl Corazzi of Cortona (Tuscany), now at the National
Museum of Antiquities in Leiden (Netherlands), are among the objects inves-
tigated by NRCA. In this case, NRCA was used to distinguish suspected fakes
from genuine statuettes, solely on the basis of the elemental compositions of the
artifacts [20]. The conclusion was based on the observation that minor quan-
tities (up to several percent) of zinc occurred in some of these statuettes. Given
the melting techniques available to the Etruscan blacksmiths, no more than a
fraction of a percent of zinc would occur in a genuine artifact (Fig. 13.5).

13.2.2 Prompt Gamma-Ray Activation Analysis

Prompt gamma-ray activation analysis is a nondestructive nuclear method for
performing both qualitative and quantitative multielement analysis of major,
minor, and trace elements in samples. The technique is useful for the analysis of
light elements such as H, B, C, N, Si, P, S, and Cl, as well as for heavy elements such
as Cd, Sm, Gd, and Hg [26]. When a thermal or cold neutron is absorbed by a
target nucleus, the compound nucleus is in an excited state with energy equal to the
binding energy of the added neutron. Then the compound nucleus will almost
instantaneously (in less than 10–14 s) de-excite into a more stable configuration
through the emission of characteristic prompt gamma rays. In many cases, this new
configuration yields a radioactive nucleus that also de-excites (or decays) by
13 Novel Neutron Imaging Techniques for Cultural Heritage Objects 237

Fig. 13.5 Photograph of two Etruscan statuettes, one genuine and one false [20]

emission of characteristic delayed gamma rays. PGAA is based on the detection

and quantification of the prompt gamma rays emitted by the sample during
neutron irradiation, while NAA uses the delayed gamma rays from the radioactive
daughter nucleus (Fig. 13.6). Consequently, PGAA complements NAA by allow-
ing determination of elements that do not form radioactive products after irradia-
tion (e.g., hydrogen and boron) and elements for which the half-life is too long to be
conveniently measured by NAA (e.g., carbon). Boron is the only exception to the
238 C. Andreani et al.

Fig. 13.6 Nuclear processes involved in prompt gamma-ray activation analysis (PGAA) and
neutron activation analysis (NAA): neutron capture is immediately followed by emission of
prompt gamma rays. The compound nucleus decays by beta emission followed by the emis-
sion of delayed gamma rays

usual prompt gamma-ray measurements in that the gamma measured is not due to
the (n, ) reaction. Indeed, boron (like lithium-6) reacts with neutrons by emission
of -particles [i.e., via the reaction 10B(n, )7Li]. Most of the lithium-7 is formed in
an excited state and de-excited instantaneously (within 10–14 s) by the emission of
477 keV gamma rays.
The PGAA technique requires mainly a source of thermal or cold neutrons
and a high-resolution spectrometer for measurement of gamma rays with
energies over a range from about 100 keV to 11 MeV. The energies of the
prompt gamma rays identify the neutron-capturing elements, while the inten-
sities of the peaks at these energies are used to determine their concentrations
(Fig. 13.7). An advantage of the PGAA method is that nothing special is
required in the way of sample preparation: the sample can be in a solid, liquid,
or gaseous state. Depending on the amount of the investigated material,

Fig. 13.7 Example of prompt gamma-ray activation analysis (PGAA) spectrum from an
estuarine sediment
13 Novel Neutron Imaging Techniques for Cultural Heritage Objects 239

element, and target matrix, the detection limit for a PGAA installation can be as
low as 10 ppb (for B, Sm, Gd). The activity of the object after irradiation is low
or decreases rapidly in most cases; therefore, the object can be returned within a
few days of the measurement. Compared with NAA, where the sample is placed
close to the core of the reactor, the neutron beam intensity at the sample is six
orders of magnitude lower.
A typical setup for PGAA consists of a sample box, a shielding, and a
detection system [13, 27–30]. The object is exposed to an intense thermal or,
preferably, cold neutron beam on the order of 106–108 n/cm2s, originating from a
neutron guide. The neutron guide is preferably curved to avoid a direct view to
the moderator or cold neutron source, thus reducing the fast-neutron and
gamma-ray background. The huge number of peaks in a spectrum (up to about
a thousand) requires a high-purity germanium (HPGe) detector with high-energy
resolution (<2 keV at 1,332 keV). Moreover, many high-energy gamma rays will
not deposit their full energy inside the HPGe detector and thus build up large
Compton tails that will hide peaks at lower energies or generate (unwanted)
additional escape peaks. To reduce these effects, the HPGe detector is sur-
rounded by a fast gamma-ray detector, usually made of an annular bismuth
germanate oxide (BGO) or NaI(Tl) scintillator that serves to reject events simul-
taneously detected in the HPGe detector and the scintillator. This anti-Compton
shield acts as an active shield against unwanted, external high-energy gamma rays
that deposit part of their energy in the scintillator before scattering in the HPGe
detector. The whole detection system is then covered with lead bricks and addi-
tional neutron shielding to further improve the signal/background ratio. The
sample box is usually covered for the same reasons with lithium-6, except at the
entry of the neutron beam, to prevent scattered neutrons from reaching the
detector. The detection system is normally placed at 908 to the neutron beam.

Fig. 13.8 Main composition of 53 brooches studied by S. Baechler et al. [31] (left) and a
photograph of one of them (right)
240 C. Andreani et al.

The count rate of the acquisition system is limited to about 10 kHz by adjusting
the neutron beam size with lithium-6 collimators.
PGAAs have been performed on various samples, in a wide range of fields,
including material science, medicine, nuclear science, geology, and archeology.
In one of the first CH applications, S. Baechler and coworkers [31] studied 53 La
Tene and Roman brooches from western Switzerland and identified four dif-
ferent groups of copper-based alloys (Fig. 13.8).

13.3 Combined Prompt Gamma-Ray Activation Analysis Scanning

and Neutron Tomography
The scope for upgrading a PGAA station to PGAI can be assessed by an
estimate of the experimental time needed for a full scan of an object. The time
needed to measure a 5  5  5 cm3 object with a 1-mm3 spatial resolution would
be equal to 125,000 times the acquisition time to measure a 1-mm3 sample.
Limiting the acquisition time to 4 s—which is already a huge compromise
because it takes from minutes to a few hours to analyze most samples using
PGAA (to detect trace elements)—and estimating the time to move the object at
1 s, the total beam time would be more than a week. Unfortunately, objects with
cultural importance are typically manufactured objects, made of numerous
small parts, each of them being to some extent homogeneous. A complete
scan at high spatial resolution of such objects would be not only impossible
but also of little use. A better approach to the problem is to use another
technique to distinguish the different parts of the object and then to measure
them by PGAI, with an adequate spatial resolution smaller than an ideal level of
detail but large enough to measure the main and minor elements in a reasonable
length of time. The full process can be optimized further in the following way:
1. the object is imaged by X-ray tomography with a high spatial resolution,
typically 10 mm;
2. an NT of the object is performed with a resolution down to 100 mm;
3. a bulk PGAA of the whole object is performed to identify all the elements
present; and
4. the regions of interest are measured by PGAI.
The acquisition time is regulated by the requirements for quantifying the
elements seen in the bulk analysis. Another limitation may come from the need
to keep the induced long-lived activation of the object to an acceptable level; for
CH objects this means returning the object to a curator after a few weeks. This is
not a problem for PGAA but should be reconsidered in the case of trace
elements exposed to large neutron fluences in PGAI.
Alternative geometrical arrangements can be considered in the implementation
of PGAI. The first and preferred one consists in collimating the neutron beam
using a 1-mm-thick lithium-6 foil with a hole that lets a neutron beam of the order
13 Novel Neutron Imaging Techniques for Cultural Heritage Objects 241

of 1 mm2 to pass. A thick gamma-ray collimator made of lead, placed in front of

the spectrometer, defines a pencil beam of gamma rays allowed to enter the
spectrometer. The intersection of the neutron beam and the gamma-ray beam
delimits a voxel, a region of the object measured by PGAI. With this solution, a
3D map of the composition of the object is obtained by scanning the object across
the neutron beam and in front of the HPGe detector in three dimensions (XYZ).
For each voxel, a gamma-ray spectrum is acquired and its analysis yields the
composition of the small portion of the object, defined by the intersection of
neutron beam collimation and HPGe collimation. An advantage of this solution
is that one can in a simple way select the voxel and its size. Provided an X-ray and/
or a neutron tomography of the object is performed before the PGAI measure-
ments, the experiment can be as simple as the following: the user selects with the
3D visualization software the 3D coordinates of a region of interest (ROI) and a
motorized 3D positioning stage places this ROI in front of the collimators for a
PGAI measurement. In principle, the best orientation of the sample in the set of
collimators can be found by minimizing the attenuation of neutrons in the path to
the ROI and of the gamma rays in the path from the ROI to the detector.
This scanning solution presents the following drawbacks.
1. Neutrons scatter inside the object, especially if the object contains hydrogen,
so the irradiated region is no longer a beam of 1 mm2.
2. Collimation of high-energy gamma rays (>1 MeV) requires at least 20-cm-thick
lead brick. This increases the distance between the object and the detection
system, reducing its efficiency and thereafter increasing the acquisition time.
A compromise again must be chosen between best collimation and measuring
time. Fortunately, most emitted gamma rays have energies below 1 MeV.
3. Reducing the viewing cone of the spectrometer reduces not only the size of
the ROI but also the efficiency of the detector. It does not, however, reduce
the background.
An alternative solution that would improve the signal/background ratio of
PGAI consists of removing the gamma-ray collimation. The object would be
scanned in two dimensions with the neutron beam and rotated from 0 to 1808.
This solution corresponds to prompt gamma-ray activation tomography in
which the HPGe acquires spectra coming from lines though the object. Measur-
ing the set of lines at different orientations of the object yields a 3D map of the
composition of the object. The number of positions is larger than with the
scanning solution. Indeed, keeping in mind our 5  5  5 cm3 object and a final
spatial resolution of 1 mm3, the required number of projections is slightly larger
than for the scanning solution: 50  50  p/2  50 = 196,350. The acquisition
time is, on the other hand, more than 100,000 times shorter in a first approxima-
tion because the spectrometer views a 50  1  1 mm3 portion of the object, and
the collimation in front of the detector can be released. The total measurement
time would then be less than 3 days, dominated by the time to move the object.
The real disadvantage of this method is that it is not possible to determine the
composition of a part of the object without measuring at least a full slice of it, as
242 C. Andreani et al.

required by the reconstruction algorithm. The a priori knowledge of the object

structure from neutron or X-ray tomography is not in this case directly
useable — the whole object must be measured regardless. A third option for
implementation of PGAI would consist of keeping the gamma-ray collimation
and not the neutron collimation. This solution reproduces single photon emis-
sion computed tomography, with spectra as projections; the reconstruction of
the composition maps consists of reconstructing each element separately. This
solution, however, does not make use of the main advantage of using cold
neutrons (i.e., that one can easily collimate them and irradiate only a selected
part of the object without reducing the total measurement time).
S. Baechler and coworkers performed the first PGAI experiment at the
PGA station of Paul Scherrer Institut (PSI), Switzerland, to study a piece of
the 2-billion-year-old Oklo natural reactor [32]. The cold neutron beam was
focused using a neutron lens and reduced to 1 mm2 by a lithium-6 collimator at
the focal point. No collimator was placed in front of the HPGe detector, and
the sample was scanned only in two dimensions in front of the neutron beam
without rotations. Although they gave the composition integrated over the
neutron path only, the results agreed with the uranium-sensitive tomography
(Fig. 13.9) taken at the European Synchrotron Radiation Facility (Grenoble,
France) using high-energy X-rays [7]. Besides the combinative use of two
techniques to interpret PGAI results, this experiment showed that one could
correct PGAI data for gamma-ray attenuation using the tomography results.
It is obvious from the simple calculation on the required measuring time that
an intense cold neutron beam should be available to study large objects. A new
PGAA facility [33] is now available at FRM-II2 in Germany [34]. One of its major
features is a super-mirror elliptic tapered neutron guide that should focus cold

(a) (b) (c)

Fig. 13.9 (a) X-ray radiograph at 87 keV of a piece from the Oklo natural reactor (Gabon);
(b) 3D distributions of lead and uranium resulting from tomographies taken at the high-
energy beam line (ID15) of the European Synchrotron Radiation Facility (Grenoble, France)
[7]; (c) spatial distributions of selected elements in the same piece, obtained by prompt
gamma-ray activation imaging at the Paul Scherrer Institut (Villigen, Switzerland) [32]

2
Forschungsreaktor München II (translation: Munich Research Reactor II), Munich
Technical University.
13 Novel Neutron Imaging Techniques for Cultural Heritage Objects 243

neutrons to a spot size of 4 7 mm2 at 10 cm from the end of the guide. The cold
neutron flux is expected to exceed 1010 n/cm2s at the focal point [35]. One of the
aims of the facility is to test different PGAI geometry options on CH objects.

13.4 Imaging with Neutron Resonances

The NRCA experience at the GELINA neutron source [17–25] is the starting
point for the development of an imaging method based on resonant neutron
capture. Imaging applications require high neutron fluxes, and the 100 times
more powerful ISIS pulsed neutron source (Chilton, UK) is now used for the
development of NRCI and NRT [15]. Figure 13.10 shows a sketch of the NRCI/
NRT setup that is under construction at ISIS. In the case of NRCI, the object is
scanned in front of a collimated neutron beam while a set of scintillator
detectors surrounding the object record the capture gamma rays. The composi-
tion of the object must be retrieved by a tomographic reconstruction corrected
for neutron self-absorption. In practice the spatial resolution is limited by the
need to collimate the beam of epithermal neutrons while keeping the signal/
background at an acceptable level. The best compromise at the moment seems

400 mm

Gamma
400 mm detectors

Collimator
300 mm

XYZ-w stage Neutron transmission detector

Fig 13.10 Sketch of a setup for performing neutron resonance capture imaging and neutron
resonance tomography. The sample is attached to a frame allowing translation and rotation
motion relative to the beam and detectors, which are fixed. The gamma-ray detectors for
NRCI are placed around the sample. A position-sensitive detector for NRT is placed in the
neutron beam behind the sample
244 C. Andreani et al.

to be a collimated beam diameter of 5 mm, which is also the anticipated spatial

resolution to be achieved by NRCI in the ISIS experiment.
For NRT measurements, the neutron collimator is removed and the object is
scanned and rotated in front of the NRT detector. Indeed the main challenge in
NRT is the realization of a position-sensitive detector for epithermal neutrons
suitable for time-of-flight (TOF) measurements. It is obvious that the spatial
resolution of such a detector will not reach, in the near future, the resolution of
the ones used in thermal or cold NT. The detector prototype tested at ISIS [36]
features 2  2 mm2 pixels made from lithium glass. A length of 8 mm for each
pixel provides adequate efficiency for epithermal neutron detection, and an array
of 10  10 pixels can be managed. Thus NRT is expected to provide better spatial
resolution than NRCI in a much shorter time. It may, however, have lower
sensitivity to trace elements, as suggested by the GELINA experience, where
transmission measurements are in regular use for total neutron cross section
measurements. Compared with NT, NRCI and NRT have a coarse spatial
resolution but provide direct information on the elemental composition; NT
provides composition information only indirectly via energy analysis of neutrons.
An interesting observation is that the information acquired by the NRT detector
is much richer than a radiograph: a full TOF spectrum per pixel with several
resonances per element is measured, yielding several element-sensitive images at
once. The NRT potential is better understood in the case of large samples. With
thermal or cold NT, when the sample is ‘‘black’’ (i.e., too thick or too absorbing),
there is no way to image it. With NRT, there are often many resonances for one
or more elements in the object. An object may be black for one resonance but
could still be imaged with another resonance in the spectrum. The NRT detector
can also be used in the thermal energy range to provide, for example, strain and
texture information using Bragg edge analysis to complement neutron diffraction
imaging (NDI) studies. The ISIS experience will tell us to what extent the
potential of the new NRCI/NRT device under construction can be exploited to
provide 3D maps of the elemental and phase composition of CH objects.
From the descriptions of the different techniques PGAA, NRCA, PGAI,
NRCI, NT, and NRT, one may conclude that an appropriate combination of
all of them, along with NDI, can be the basis for integrated analysis protocols
for the study of CH objects. The integrated use of imaging and other neutron-
based techniques for CH studies is indeed an area of ongoing research, as
illustrated by the applications discussed below.

13.5 Applications to Cultural Heritage Studies

In the past few years, neutron techniques have increasingly been used for the
study of CH materials, in particular to address their nature and authenticity,
their provenance and diffusion, the manufacturing techniques used in their
fabrication, and their state and conservation [37]. In the following discussion,
we show first results from the AC project [38–41] and examples of applications
13 Novel Neutron Imaging Techniques for Cultural Heritage Objects 245

of thermal and cold neutron tomographic techniques to the study of phase and
microstructure characterization of ceramic artifacts, ancient bronzes, and mar-
ble artifacts. A common feature in the selected examples is that the use of
neutrons was motivated either by the general aim of developing new imaging
techniques or by the need to perform a nondestructive analysis. Cast and
restored metal objects were inspected to elucidate the production technology
and the corrosion damage. Marble and ceramic artifacts were inspected to
probe the presence of adhesives, glued parts, epoxy filler materials, or epoxy
protective coatings.
For testing purposes, the AC archaeologists prepared ‘‘black boxes.’’ These
closed metal cubes contain objects of various shapes, forms, and elemental
composition representative of the composition of real archaeological objects.
The task was to recognize the elemental composition and the rest–strain of the
inner objects by PGAI, NRCI, and NDI [38–41]. Neutron and X-ray radio-
graphy and tomography were used to visualize the inner content of the black
boxes and to help with targeting the coordinates of the interesting spots for the
new 3D neutron techniques. NT was performed at the ANTARES3 facility [42] at
FRM-II, and the X-ray tomographies were performed at the Centre for X-ray
Tomography at the Ghent University [43]. Figure 13.11 shows a 3D visualization
(tomograph) of the reconstructed X-ray data (a) and the corresponding neutron
data (b) for one of the black boxes. Using solely the image content, it can be seen
that neither X-ray tomography nor NT alone allows the derivation of both types
of materials and the elemental composition of the scanned object. However, the

Fig. 13.11 X-ray tomograph (a) and neutron tomograph (b) of black box H-IX from the
Ancient Charm project [39]

3
Advanced Neutron Tomography and Radiography Experimental System.
246 C. Andreani et al.

combined use of the two techniques allows areas of differing materials and their
shapes to be distinguished and separated. At a later stage, to determine precisely
the elemental composition of these objects, PGAI analysis was performed
directly at the coordinates provided by the NT image. Figure 13.12 shows an
X-ray tomograph and the corresponding neutron tomographs for another black
box. Here three separate layers of materials can be visually distinguished. The
upper row shows corresponding horizontal slices of the object (looking ‘‘from
above’’); the bottom row shows vertical slices (looking ‘‘from the side’’). A
suitable isocontour of the X-ray slice helps in visualizing the initial misalignment
of the NT image and its correct alignment after the registration. This example
demonstrates in various ways the value added by the registration and the sub-
sequent point-to-point comparison.
The neutron tomograph in Fig. 13.13 refers to laminated bronze artifacts
from rich sepulchral complexes in the necropolis of Osteria-Poggio Mengarelli
and Cavalupo (eighth century B.C.) [44]. Quantitative phase analysis was per-
formed by neutron diffraction at the INES4 diffractometer installed at the ISIS
neutron source. This result was complemented by the analysis of the spatial
distribution of the components, obtained with an NT experiment (Fig. 13.13),

Fig. 13.12 (a) X-ray tomograph of black box D-VIII from the Ancient Charm project [39];
corresponding neutron tomograph in the initial position (b) and after automatic registration (c) [38]

4
Italian Neutron Experimental Station, a collaborative project between ISIS and the Italian
National Research Council [Consiglio Nazionale delle Ricerche].
13 Novel Neutron Imaging Techniques for Cultural Heritage Objects 247

Fig. 13.13 Neutron

tomography from bronzes
[44] provided information
on the spatial distribution of
the components forming the
samples. Measurements
were performed at the COld
Neutron RADiography
(CONRAD) instrument at
the Hahn-Meitner Institut
(Berlin, Germany)

and additional information on elemental composition through XRF. A surface

photomicrograph and infrared scan were used to complete the characterization
of the bronzes.
The results in Fig. 13.14 refer to marbles from Villa Adriana [45, 46] for
which the geological setting is a priori unknown. Marble is typically composed
of carbonate minerals (such as calcite and dolomite), silicates (such as quartz,
plagioclase and micas), oxides (such as rutile and magnetite), and phosphates
(such as apatite) can be present as trace or minor phases. Natural calcite rocks
and marbles commonly exhibit a preferred orientation (texture), which forms
during growth or deformation and which is modified during recrystallization
and plastic tectonic deformations. These geological textures, formed over mil-
lions of years, are not affected by quarrying or stone sculpting. Thus, in the
archaeological context, the texture analysis of marble helps in attribution and
provenance determination. It may further be useful in the field of conservation:
for instance, the weathering of marble by anisotropic thermal expansion is also
influenced by the crystallographic textures [47]. Based on references [45, 46], it
was possible to distinguish among the different marble types purely on the basis
of the mineral phase compositions and the crystallographic textures (geological
stress). In particular, the phase analysis performed on the Villa Adriana marble
fragments allows for a clear separation into four different groups. The same
marbles were also investigated by thermal NT (Fig. 13.15). From this study, it
was possible to distinguish distinct components and their relative spatial dis-
tributions in the marbles. These are indicated with different colors in the
tomographic picture in Fig. 13.15. (a) darkest shade (most abundant compo-
nent), (b) medium shade (intermediate component), and (c) lightest shade
(smaller component) (courtesy of R. Triolo et al. [48]).
Finally, ceramic artifacts dated from the second to the third century (A.D.)
from a large collection found in recent excavations (1988) in Puglia (Italy) were
investigated. Their provenance is different areas of the central Mediterranean.
Neutron diffraction and NT measurements [48] performed respectively at the
ISIS pulsed neutron source and at PSI [49] provided valuable information on
composition, provenance, and fabrication methods. Samples were found to be
made of the same type of clay, with similar percentages of calcite, quartz, and
248 C. Andreani et al.

Fig. 13.14 (a) Example of (a)

texture analysis of a copper
reference sample derived (hkl) lattice plane= 111
Bragg peak
from diffraction data on the
ROTAX instrument
installed at ISIS (courtesy
W. Kockelmann); (b) Maps Texture strength:
Max
of the c-axis distributions, maximum m.r.d. 1.76

represented by the (006) pole

figures, of eight distinct
marble samples from Villa
Adriana [45] compared with
a geological calcite reference (111)
sample from Syros, Greece.
For each pole figure, the sa
m (200)
maximum multiples of
ple

random distribution are (220)

or
ien
given
ta
2.2
tio

1.2 1.4 1.6 1.8 2.0

1.0
n

α-spacing
(b)
006 3 006 18 006 22

Max Max Max

1.64 1.78 1.42

006 9 006 11 006 28

Max Max Max

1.57 1.22 1.28

008 16 008 17 008 ref

Max Max Max

1.17 13.80 7.23

plagioclase. The examples in Fig. 13.16 demonstrate that a 3D inside view of

objects a few centimeters thick can be achieved within a relatively short time
(from a few minutes to a few hours) with spatial resolution of as low as 100 mm.
The characterization of a ceramic artifact and the analysis of the composition of
alloys in a metallic object provided an additional piece of information to aid in
establishing whether the object belonged to the history of the local people or
was imported from another culture [50].
13 Novel Neutron Imaging Techniques for Cultural Heritage Objects 249

Fig. 13.15 Example of neutron tomography of marble fragments from Villa Adriana [48]

Fig. 13.16 Example of

neutron tomography from
ceramic artifacts [48]

The characterization by neutron diffraction of ancient bronze, marble, and

ceramic artifacts illustrates how the penetrating power of neutrons, combined
with imaging analysis, leads to advances in the scientific characterization of the
materials that constitute a CH object. The techniques used are truly nonde-
structive, an important advantage if a unique and precious artifact is to be
investigated without risk of damaging it in the analysis. The ongoing develop-
ments in applications of neutrons in CH research are promising and suggest that
neutron-based techniques will have a lasting impact on the way CH analyses are
performed. This will be especially the case if comprehensive analysis protocols
are developed and adopted as best practices by the CH community.

Acknowledgment This work was performed with financial support from the European Com-
munity under contract HPRI-2001-50043 and within the CNR-Council for the Central
Laboratory of the Research Councils (UK) agreement. Carla Andreani and Giuseppe Gorini
acknowledge Consiglio Nazionale delle Ricerche (CNR), Italy, for financial support for the
experiments performed at ISIS. Thomas Materna acknowledges financial support from the
250 C. Andreani et al.

Swiss National fund, the University of Fribourg (Switzerland), and the University of Cologne
(Germany) for the experiments performed at PSI and the European Synchrotron Radiation
Facility. We also wish to thank Giulia Festa for critical revision of the text.

References
1. Portal (News and developments from the CCLRC). Looking forward to the past: science
and heritage. Special issue Autumn/winter (2006). http://www.scitech.ac.uk/Publications/
nl/Portal/.
2. ESRF Newsletter. Science and art. 44 (December 2006).
3. M. Menu et al., The dedicated accelerator-based IBA facility AGLAE at the Louvre,
Nucl. Instrum. Meth. Phys. Res. B 45, 610–614 (1990).
4. E.H. Lehmann, P. Vontobel, B. Schillinger, T. Bücherl, S. Baechler, J. Jolie, W. Treimer,
R. Rosa, G. Bayon, S. Legoupil, S. Körner, V. Micherov, M. Balasko, in Proceeding of
the 15th World Conf. on Non-Destructive Testing, Rome, Italy, October 15–21, (2000).
Available from http://www.ndt.net/article/wcndt00/papers/idn804/idn804.htm.
5. N. Kardjilov, S. Baechler, M. Bastürk, M. Dierick, J. Jolie, E. Lehmann, T. Materna, B.
Schillinger, P. Vontobel, Nucl. Instrum. Meth. Phys. Res. A 501, 536 (2003).
6. NEUTRA: Thermal neutron facility for imaging at the Paul Scherrer Institute, Villigen,
Switzerland http://neutra.web.psi.ch/facility/index.html.
7. T. Materna et al., Uranium-sensitive tomography with synchrotron radiation, J. Syn-
chrotron Rad. 6, 1059–1064 (1999).
8. W. Kockelmann, G. Frei, E.H. Lehmann, P. Vontobel and J.R. Santisteban, Nucl.
Instrum. Meth. Phys. Res. A 578, 42 (2007).
9. G. Artioli, M. Dugnani, I. Angelini, L. Lutterotti, A. Pedrotti, A. Fleckinger, Proc.
Archaeometallurgy Europe 2, 19–27, (2003); G. Artioli, M. Dugnani, T. Hansen, L.
Lutterotti, A. Pedrotti, G. Sperl, in: A. Fleckinger (ed.), La mummia dell’età del rame. 2.
Nuove ricerche sull’uomo venuto dal ghiaccio, Vol. 3, Folio Verlag, Bolzano, 9–22, (2003).
10. E. Ciliberto and G. Spoto (eds.) Modern Analytical Methods in Art and Archaeometry,
John Wiley & Sons, Inc. New York (2000).
11. D.R. Brothwell and A.M. Pollard (eds.) Handbook of Archaeological Sciences, John
Wiley & Sons, Inc., New York (2001).
12. C.O.Fischer et al, Neues zur Neutronen-Aktivierungs-Autoradiographie, Restauro 105,
426 (1999) (in German).
13. Zs. Révay and T. Belgya, Principles of the PGAA method, in G.L. Molnár ed., Hand-
book of Prompt Gamma Activation Analysis with Neutron Beams, Kluwer Academic
Publishers, pp. 1–305 (2004).
14. H. Postma, M. Blaauw, P. Bode, P. Mutti, F. Corvi and P. Siegler, Neutron-resonance
capture analysis of materials, J. Radioanal. Nucl. Chem. 248(1), 115–120 (2001).
15. G. Gorini for the Ancient Charm collaboration, Ancient Charm: A research project for
neutron-based investigation of cultural-heritage objects, Il Nuovo. Cimento. C, 30, 47–58
(2007).
16. S.F. Mughabghab et al., Neutron Cross-sections, vol.1, Academic Press, New York
(1981). Available online at http://atom.kaeri.re.kr/cgi-bin/endfplot.pl.
17. H. Postma, M. Blaauw, P. Schillebeeckx, G. Lobo, R. Halbertsma and A.J. Nijboer,
Non-destructive elemental analysis of copper-alloy artefacts with epithermal neutron-
resonance capture, Czech. J. Phys. 63(Suppl. A), A233–A240 (2003).
18. M. Blaauw, H. Postma, P. Mutti, An attempt to date an antique Benin bronze using
neutron resonance capture analysis, Appl. Radiat. Isot., 62, 429–433 (2005).
19. M. Blaauw, H. Postma and P. Mutti, Quantitative neutron capture resonance analysis
verified with instrumental neutron activation analysis, Nucl. Instrum. Meth. Phys. Res. A
505, 508–511 (2003).
13 Novel Neutron Imaging Techniques for Cultural Heritage Objects 251

20. H. Postma, P. Schillebeeckx, and R.B. Halbertsma, Neutron resonance capture analysis
of some genuine and fake etruscan copper-alloy statuettes, Archaeometry 46 635–646
(2004).
21. H. Postma and P. Schillebeeckx, Non-destructive analysis of objects using neutron resonance
capture, memorial issue Gabor Molnár, J Radioanal. Nucl. Chem. 265, 297–302 (2005).
22. R. C. Perego, H. Postma, M. Blaauw, P. Schillebeeckx, A. Borella, Neutron Resonance
Capture Analysis: improvements of the Technique for resonances above 3 keV and new
applications, J. Radioanal. Nucl. Chem. 271(1) 89–94 (2007).
23. H. Postma, R.C. Perego, P. Schillebeeckx, P. Siegler, and A. Borella, Neutron resonance
capture analysis and applications, J. Radioanal. Nucl. Chem. 271(1) 95–99 (2007).
24. H. Postma, D. Fontijn, P. Schillebeeckx, R.C. Perego, and J.J. Butler, Neutron Reso-
nance Capture Analysis of Geistingen axes, LUNULA, Archaeologia protohistorica XIII,
41–46, (2005).
25. H. Postma and P. Schillebeeckx, Neutron-resonance capture as a tool to analyse the
internal compositions of objects non-destructively, Notiziario Neutroni e Luce di Sincro-
trone 11(2), 14–18 (2006).
26. Z.B. Alfassi, C. Chung (Ed.), Prompt Gamma Neutron Activation Analysis, CRC Press,
Boca Raton, Florida, 1995.
27. M. Crittin et al. The new prompt gamma-ray activation facility at the Paul Scherrer
Institute, Switzerland, Nucl. Instr. Meth. Phys. Res. A 449, 221 (2000).
28. Zs. Revay et al., Cold neutron PGAA facility at Budapest, Nucl. Instrum. Meth. Phys.
Res. B 213, 385–388 (2004).
29. H-J Cho et al., Study on prompt gamma-ray spectrometer using Compton suppression
system, Nucl. Instrum. Meth. Phys. Res. B 229, 499–507 (2005).
30. Zs. Revay et al., Construction and characterization of the redesigned PGAA facility at the
University of Texas at Austin, Nucl. Instrum. Meth. Phys. Res. A 577, 611–618 (2007).
31. S. Baechler et al. The k0-method in cold-neutron prompt gamma-ray activation analysis,
J. Radioanal. Nucl. Chem. 256, 239–245 (2003).
32. S. Baechler et al., Non-destructive analysis of a bulky sample from a natural fossil reactor,
J. Radioanal. Nucl. Chem. 250, 39–45 (2001).
33. P. Kudejova et al., On the construction of a new instrument for cold-neutron prompt
gamma-ray activation analysis at the FRM II, J. Radioanal. Nucl. Chem. 265, 221–227
(2005).
34. Research reactor II (FRM-II), Technische Universität München, 2007 http://www.frm2.
tum.de/en/index.html.
35. P. Kudejova, Two new installation for non-destructive sample analysis: PIXE and PGAA,
Ph.D. Thesis, University of Cologne (2005).
36. E. Perelli Cippo et al, Meas. Sci. Technol. 19, 034027 (2008).
37. W. Kockelmann, L.C. Chapon, R. Engels, j. Schelten, C. Neelmejer, H.M. Walcha, G.
Artioli, S. Schalev, E. Perelli Cippo, M. Tardocchi, G. Gorini, and P. Radaelli, Neutrons
in cultural heritage research, J. Neutron Res. 14(1), 37–42 (2006).
38. P. Kudejova, J. Cizek, R. Schulze, J. Jolie, B. Schillinger, K. Lorenz, M. Mühlbauer, B.
Masschaele, M. Dierick, J. Vlassenbroeck, A marker-free 3D image registration for the
Ancient Charm project. Case study with neutron and X-ray tomography datasets, Noti-
ziario Neutroni e Luce di Sincrotrone 12(2), 6–13 (2007).
39. P. Kudejova for the Ancient Charm collaboration, Neutron and X-ray imaging of the
‘black boxes’ for the Ancient Charm project, Archeometriai Mũhely, 35–40 (2008/1).
40. G. Festa, W. Kockelmann, A. Kirfel and the Ancient Charm collaboration, Neutron
diffraction analysis of ‘black boxes’, Archeometriai Mũhely 2008/1 61–72.
41. Z. Kis et al., Prompt gamma activation imaging on ‘black boxes’ in the ‘Ancient Charm’
project, Archeometriai Mũhely. 41–60 (2008/1).
42. ANTARES Facility at the research reactor II (FRM-II), Technische Universität
München (2007). http://www.physik.tu-muenchen.de/antares/.
252 C. Andreani et al.

43. X-ray Radiography and Tomography Facility at the Ghent University (UGCT) (2007)
http://www.ugct.ugent.be.
44. G. Festa, P.A. Caroppi, A. Filabozzi, C. Andreani, M.L. Arancio, R. Triolo, F. Lo Celso,
V. Benfante, S. Imberti, Composition and corrosion phases of Etruscan Bronzes from
Villanovan Age, Measur. Sci. Technol. 19(3) 034004 (7 pp) (2008).
45. A. Filabozzi, A. Pietropaolo, C. Andreani, M. P. De Pascale, G. Gorini,.W. Kockelmann,
L.C. Chapon, Non invasive neutron diffraction analyses of marbles from the ‘‘Edificio
con Tre Esedre’’ in Villa Adriana, Il Nuovo Cimento C 29, 237–252 (2006).
46. A. Filabozzi, C. Andreani, M.P. De Pascale, G. Gorini, A. Pietropaolo, E. Perelli Cippo,
R. Senesi, M. Tardocchi and W. Kockelmann, Texture and structure studies on marbles
from Villa Adriana via neutron diffraction technique, J. Neutron Res. 14, 55–58 (2006).
47. S. Siegesmund and K. Ullemeyer, T. Weiss, and E.K. Tschegg, Int. J. Earth Sci. 89, 170
(2000).
48. R. Triolo, V. Bonfante, and F. Lo Celso, private communication (2007); see also N.
Kardijlov et al, Neutron tomography in modern archaeology, Notiziario Neutroni e Luce
di Sincrotrone 13(2), 6–9 (2008).
49. G. Kühne et al., CNR—the new beamline for cold neutron imaging at the Swiss spallation
neutron source SINQ, Nucl. Instrum. Meth. Phys. Res. A 542, 264–270 (2005); ICON:
Cold neutron facility for imaging at the Paul Scherrer Institute, Villigen, Switzerland,
described in http://neutra.web.psi.ch/facility/index.html.
50. G. Festa for the Ancient Charm collaboration, ANCIENT CHARM: A new project for
neutron-based 3D imaging with applications to CH research. Poster Section (Ist prize
winner): Looking Forward to the Past: Science and Heritage, Tate Modern London
(28 November 2006). Available online at http://ancient-charm.neutron-eu.net/ach/
conferences.
Chapter 14
Probing the Potential of Neutron Imaging
for Biomedical and Biological Applications

K.L. Watkin, H.Z. Bilheux, and J.F. Ankner

Abstract Neutron imaging of biological specimens began soon after the dis-
covery of the neutron by Chadwick in 1932. The first samples included tumors
in tissues, internal organs in rats, and bones. These studies mainly employed
thermal neutrons and were often compared with X-ray images of the same or
equivalent samples. Although neutron scattering is widely used in biological
studies, neutron imaging has yet to be exploited to its full capability in this area.
This chapter summarizes past and current research efforts to apply neutron
radiography to the study of biological specimens, in the expectation that clinical
and medical research, as well as forensic science, may benefit from it.

Keywords Biology  Neutron imaging  X-ray imaging  Biomedical 

Contrast agent  Thermal neutrons  Epithermal neutrons  Small
gadolinium particulate oxide (SPGO)

14.1 Introduction

Although carbon dominates the chemistry of biological tissues, hydrogen atoms

are the most abundant. Protons (1H) scatter thermal neutrons more strongly than
any other nuclei. These two facts ensure that hydrogen is the primary contributor
to neutron contrast in biological materials. In addition, since neutrons interact
with the nucleus, deuterium (2H) exhibits quite different scattering properties
from 1H. These contrast differences are already exploited in neutron scattering
and imaging and can readily be applied in a biomedical context.
The principle of neutron imaging is based on the attenuation, via both scattering
and absorption, of a directional neutron beam by the matter through which
it passes. Neutron imaging is complementary to, rather than competitive with, X-
ray techniques. X-rays are scattered and absorbed by electrons, so absorption and

K.L. Watkin (*)

University of Illinois, College of Applied Health Science, 901 So. Sixth, Champaign, IL
61821, USA
e-mail: [email protected]

I.S. Anderson et al. (eds.), Neutron Imaging and Applications, 253

Neutron Scattering Applications and Techniques, DOI 10.1007/978-0-387-78693-3_14,
Ó Springer ScienceþBusiness Media, LLC 2009
254 K. Watkin et al.

scattering increase monotonically with atomic number. Neutrons, on the other

hand, interact with nuclei and their scattering power does not scale in any regular
way with atomic number, as explained in Chapter 1, in this volume.
Neutron radiography and tomography have flourished for engineering and
materials science applications, as illustrated in Chapters 11 and 12. Recent
studies have been carried out in more biological contexts, such as water trans-
port in plants, as referenced in Chapter 17, and resin uptake in wood [1].
Current probes for medical imaging such as X-rays, ultrasound, magnetic
resonance imaging (MRI), and positron emission tomography (PET) princi-
pally determine differences in structural anatomy. Ultrasound and functional
magnetic resonance imaging (fMRI) also enable the study of functional activity
such as changing blood flow, whereas PET highlights specific metabolic func-
tions by measuring the concentration and distribution of a radiopharmaceuti-
cal tracer similar to glucose. Neutron imaging is essentially similar in spatial
resolution to these techniques, but it exhibits different elemental and isotopic
contrasts, which may be advantageous in many cases.

14.2 First Experimental Measurements on Biological Tissues

Initial neutron imaging experiments were conducted in Germany in the mid-

1930s [2], soon after the discovery of the neutron by Chadwick in 1932. Thermal
neutron imaging of biological tissues was first conducted in the mid 1950s, and
work continued through the 1960s. Due to a lack of efficient detectors for
epithermal and fast neutrons, experiments during this period were limited to
thermal neutron imaging and so used thin biological samples. These were often
imaged by passing a beam of thermal neutrons through the sample and expos-
ing a dysprosium- or indium-loaded plate that created beta-emitting radioactive
isotopes following neutron capture. The activated plate was then developed
using conventional autoradiographic techniques, by placing the plate in close
contact with an X-ray-sensitive photographic film.
As a proof-of-principle, plant tissues were successfully imaged with thermal
neutrons [3], as illustrated in Fig. 14.1. This work demonstrated the ability of
neutrons to detect 1H nuclei in biological tissues. Neutron radiographs of a
boral (boron–aluminum alloy) sheet were also produced and compared with their
X-ray counterparts to illustrate the complementarity of the two techniques.
Several-millimeter-thick biological samples were imaged with thermal neu-
trons, as referenced in [4–6]. However, when a much thicker sample such as a
human chest was studied, the image (Fig. 14.2) was seriously distorted by both
multiple scattering, arising mainly from 1H nuclei, and poor detector spatial
resolution. Furthermore, the resolution of the image was mediocre compared
with that of an equivalent X-ray radiograph [7].
Although the strong attenuation of thermal neutrons drastically limited both
the choice of biological tissues and specimen thickness, various animal tissues
were exposed to thermal and cold neutrons and compared with X-ray
14 Probing the Potential of Neutron Imaging 255

Fig. 14.1 Neutron radiography of plant tissues demonstrates the sensitivity of neutrons
to hydrogenous materials [3]

Fig. 14.2 An early thermal neutron radiograph of a human chest was limited by multiple
scattering and poor detector spatial resolution [7]

radiographs [6]. Thermal neutron dose was estimated to be equivalent to X-ray

dose, whereas incident epithermal dose was predicted to be lower for the same
neutron exposure time – epithermal neutrons generally have lower elastic
scattering cross sections than thermal neutrons. Gadolinium and indium were
commonly used as contrast agents for thermal and epithermal neutron imaging,
respectively. A brief discussion of radiation damage due to neutrons can be
found in a later section of this chapter.
256 K. Watkin et al.

Figure 14.3 shows images of thin tumor specimens: (A) adenocarcinoma, (B)
liposarcoma, and (C) leiomyoma [6]. Thermal neutron images are displayed in
the upper row. The control images for each tumor, 20-kV roentgenograms
(X-ray images), are shown in the lower row. The greater contrast seen in the
neutron images probably arises from differences in hydrogen density, since
many tumors appear to have greater hydrogen content than healthy tissues [6].

Fig. 14.3 Comparison of tumors imaged with thermal neutrons (top) and 20-kV X-rays
(bottom). The contrast in the neutron radiographs is more pronounced because of higher
hydrogen concentrations in tumorous tissue [6]

The complementarity between neutrons and X-rays is better illustrated by

images of a rat paw and tail, shown in Fig. 14.4. Although bones are clearly
visible in the X-ray radiograph, blood vessels are more prominent in the neutron
image. The nasopharynx, which is obscured by bone in the X-ray radiograph, is
clearly visible in the epithermal neutron image, as illustrated in Fig. 14.5.
Depending on the biological sample and the region(s) of interest, there are
significant differences in visible features between X-ray and neutron radiography.
Much of the early interest in neutron imaging of biological tissues was
generated by the need to determine the distribution of boron delivery agents
being developed for treatment of cancers using boron neutron capture therapy
(BNCT). In this method, boron atoms are loaded into cancerous cells, which
are then irradiated by neutrons; boron absorbs a neutron and produces an
energetic alpha particle and 7Li nucleus, which kill the cell. The thermal
neutrons used were either strongly absorbed by the body before reaching the
cancerous cells, resulting in high radiological dose, or required highly invasive
surgery to expose the cancer more directly to the neutron beam [8]. In the
14 Probing the Potential of Neutron Imaging 257

Fig. 14.4 Radiographs of a rat paw and tail using (A) thermal neutrons and (B) 42-kV X-rays.
Gd2O3 in oil was used as the contrast agent [6]

Fig. 14.5 Radiographs of a rat head using (A) epithermal (i.e., 0.5 to 2 eV) neutrons and (B)
42-kV X-rays. The rat nasopharynx is visible in (A), as indicated by the arrow. Although
In2O3 was used as a contrast agent, the epithermal neutron flux was not sufficient to
demonstrate its contrast capabilities [6]

1990s it was realized that epithermal beams should be more favorable for
BNCT, since body tissues act as a natural neutron moderator. Interest in
clinical applications reawakened, as did interest in thermal and epithermal
imaging of biological tissues.
258 K. Watkin et al.

14.3 Neutron Dosimetry: A Short Overview

As with any imaging method, there is a need to understand and evaluate the
benefits and risks of exposing a living subject to neutrons. In order to estimate
the potential radiological hazard that neutron imaging poses to soft biological
tissues, Monte Carlo simulations of neutron exposure of a simulated human
arm were conducted [9]. The conclusions were quite pessimistic. The simula-
tions predicted no effective contrast at given neutron energies of 1 keV, 120 keV,
fission spectrum, and 14 MeV. However, the study provided useful information
on dose deposition between tissue, bone cortex, and marrow. For example,
the calculated neutron dose at 120 keV energy to bone was twice the dose to
contiguous soft tissue.
While epithermal (1 eV–0.1 MeV) and fast (>0.1 MeV) neutrons are required
to penetrate thick tissues, they do not pose a direct radiological hazard. Rather,
hydrogenous tissue moderates these neutrons down to the thermal energy range
(<1 eV) at which they can be captured and thereby cause biological damage.
Fast neutrons also knock off protons from hydrogenous tissues, creating an
additional radiological hazard.
Recently, Monte Carlo simulations for neutron stimulated emission com-
puted tomography (NSECT), an in vivo tomographic spectroscopy using fast
neutrons (Chapter 15), demonstrated that the radiation dose deposited in a
human abdomen or breast using NSECT would be comparable to the dose from
an abdominal CT (computed tomography) exam or a mammogram, respec-
tively [10–12].
Clinical, i.e., small-animal, and pharmaceutical research may benefit the
most from epithermal neutron imaging. The exposure of the specimen is a lesser
concern in those fields, and imaging may aid medical diagnosis without the risk
of radiation exposure.
Recent data on the biological effects of low radiation exposures can be found
in Ref. [13]. Information on the biological effects of fast neutrons on biological
tissues can be found in Refs. [14, 15].

14.4 Recent Trends

Neutron beams generated by high-energy medical linacs have recently been

shown to be sufficiently intense to produce images exhibiting soft-tissue contrast
[16]. It was reported that direct thermal neutron imaging was not as contrast-
sensitive as indirect neutron imaging (Chapter 5) or fast neutron imaging. Later,
neutron radiography of rabbit VX-2 liver cancer using thermal and cold neutron
radiography permitted clear observation of hepatic vessels and the VX-2 tumor,
with cold neutron imaging being especially effective. The image contrast of
this modality enhanced tumor identification compared with absorption-contrast
X-ray radiography [17].
14 Probing the Potential of Neutron Imaging 259

Very recently (Figs. 14.6 and 14.7), a study of rat femurs demonstrated the
relative strengths of X-ray and neutron imaging. Heavy elements (predomi-
nantly calcium and phosphorus in bone) determine X-ray contrast, so the X-ray
images highlight the hard bone framework, with the hydrogenous matter being
essentially invisible. Neutrons, on the other hand, highlight a relatively uniform
distribution of hydrogen throughout the interior and exterior regions of the
bone and do not resolve the filaments of trabecular hard bone running through
the marrow. Interestingly, the neutron images (Fig. 14.7) show an obvious bone
reduction in the decalcified samples – after inorganic compounds are dissolved
by soaking in a 15% ethylene diamine tetra acetic acid solution, as expected

Fig. 14.6 Images made by 40-keV X-ray radiography of (a) calcified bones (i.e., heated at
6008C for 5 h) and (b) natural bones. The distinction between the cortical (surface) and the
trabecular (inner) bone regions is evident [18]

Fig. 14.7 Neutron radiographs of rat bones from (a) decalcified and (b) natural bones [18]
260 K. Watkin et al.

from the egression of mineral material from inside the bone. Post-mortem bone
desiccation was also evaluated as a possible forensic tool [18].
Several recent neutron imaging experiments on biological samples have been
published, e.g., the remains of an ichthyosaur embryo imbedded in a shell [19],
arachnids [20], and insects [21]. Examples of radiographs of other biological
specimens can be found at http://neutra.web.psi.ch/gallery/biological.html.
Adult male porcine lung tissue (1.5 mm thick, 1 cm by 1 cm sample size) immersed
in formalin was successfully imaged using thermal neutrons. These slices were first
extracted from frozen samples from the anatomical region illustrated in Fig. 14.8.
Standard histological techniques were used to allow comparison of neutron
and optical images (Fig. 14.9). After neutron exposure, the samples were stained
using hematoxylin and eosin stain, which is known to provide good resolution
of lung tissue components such as muscle and cartilage.
Although the images in Fig. 14.9 look very similar, the distribution of their
pixel intensities can supply a true quantitative comparison. Recently developed
software [22] enables automatic integration of multiscale volumes from differ-
ent imaging modalities through a cyber-based portal imaging system. It was
found that 92% of the pixel intensities of these two images match, meaning the
pixel-by-pixel intensity between the histological and the neutron images is very
similar. These initial results for fixed porcine lung tissue provide significantly
more detail than any previously reported images taken with a thermal neutron
beam [7]. To compare thermal neutron imaging with X-ray imaging, the
specimen was also imaged using microCT, as shown in Fig. 14.10. The microCT
measurement was conducted before histological staining and subsequent opti-
cal imaging. The gross morphology in all the images is very similar. Cartilage

Fig. 14.8 Approximate location of the thin porcine lung specimen imaged using thermal
neutrons. Only the lung airways are illustrated
14 Probing the Potential of Neutron Imaging 261

Fig. 14.9 Histological and thermal neutron images of the same section of porcine lung, with a
92% pixel agreement (see text)

Fig. 14.10 Multimodal images of the same specimen of porcine lung tissue

supporting the airway is clearly visible in each modality, along with differentia-
tion of the alveolar regions, although the greatest definition is seen in the
histology images.
In a recent study, the anatomical changes in corn kernels affected by Asper-
gillus flavus fungus were analyzed quantitatively using neutron tomography
[23], demonstrating the effectiveness of the technique in identifying the degra-
dation of the biological tissues compared with tissues in uninfected control
kernels. Resistant kernels showed no sign of anatomical change. This interest-
ing research shows the sensitivity of neutron radiography to the effects of
diseases on biological tissues. The most recent and promising studies are illu-
strated by tomography of a rat lung [24].
262 K. Watkin et al.

14.5 Small Particulate Gadolinium Oxide (SPGO) Nanoparticles

for Targeted and Nontargeted Contrast Enhancement for
Future Biomedical Neutron Imaging Applications

There is a great and largely untapped potential for the use of neutrons in
biomedical imaging using isotopically labeled marker compounds. Medical
compounds incorporating neutron-absorbing isotopes can be targeted to spe-
cific organs, such as the heart, to enable neutron imaging of anatomical struc-
ture and organ function as well as to deliver radiation doses specifically to
tumor sites. Such a system could provide higher-resolution images than PET
and more precision in targeting disease. Neutrons may also be used with water-
stable SPGO (small particulate gadolinium oxide) nanoparticles [25–27] for
targeted or non-targeted contrast enhancement. Neutron imaging through
metal – e.g., stents and other metallic objects used for biological purposes – is
possible. Small-animal imaging studies, isotopic imaging, and targeted
contrast-enhanced imaging are all potential applications.
Recent experiments highlight the biomedical potential of the neutron-sensitive
SPGO contrast agent. Dextranized SPGO (10 nm) was injected into the portal
vessel of a porcine liver specimen. A thermal neutron radiograph of the thin
specimen (1.5 mm) with the dextranized SPGO highlighting the lumen of the
vessel is shown in Fig. 14.11. SPGO particles appear dark because of the high
thermal neutron capture cross section of gadolinium.

Fig. 14.11 Neutron images of SPGO-doped liver tissue. A: original liver specimen; B: color-
enhanced image of A; C: distribution of the SPGO nanoparticles created using simple thresh-
olding techniques
14 Probing the Potential of Neutron Imaging 263

14.6 Summary

Many recent advances in medical imaging are based upon technological advances
in computer visualization, refined image projection techniques, and advanced
charge-coupled device systems, among others. Although there have been several
recent attempts to create small neutron generating systems that may be useful at
the laboratory level, the use of larger reactor- and accelerator-based neutron
sources with dedicated imaging systems remains the most effective and efficient
way to use thermal neutrons for biological tissue imaging.
Neutron radiographs of biological tissues accompanied the first neutron
imaging experiments. Neutron imaging of biological samples flourished during
the 1960s and 1970s and then largely disappeared until the 1990s. The recent
emergence of new techniques and improved detection (e.g., for epithermal
neutrons – Chapter 4) are leading to a rejuvenation of the field of neutron
imaging.
Neutron imaging of living human subjects still provokes obvious radiologi-
cal concerns. However, its potential for applications such as clinical research,
small-animal imaging, and forensic science needs to be investigated further. The
high sensitivity of neutrons to hydrogen atoms should be of benefit for forensic
applications, where quantification of water changes in dead tissues is highly
important.
Acknowledgments This work is supported by the U.S. Department of Energy. The Oak Ridge
National Laboratory Spallation Neutron Source is managed by UT-Battelle, LLC, for the
Department of Energy under contract DE-AC05-00OR22725. We would like to thank Dr.
John Nemeth of Oak Ridge Associated Universities for financial support of part of this
research, under contract DE-AC05-06OR23100. We would also like to thank our many
colleagues, especially Daniel Hussey at the National Institute of Standards and Technology
Center for Neutron Research and Candice Halbert from Oak Ridge National Laboratory.

References
1. E. Lehmann, S. Hartmann, P. Wyer, Nucl. Instrum. Meth. A 542, 87 (2005).
2. H. Kallman, Research 1, 254 (1947).
3. J. Thewlis et al., Br. J. Appl. Phys. 7, 345 (1956).
4. H.L. Atkins, Biological application of neutron radiography, Mat. Eval. 23, 453–458
(1965).
5. Barton, J. P., Some possibilities of neutron radiography, Phys. Med. Biol. 9, 33–42 (1964).
6. M. Brown and P.B. Parks, Am. J. Roentg. 106, 472 (1969).
7. J. Anderson et al., Br. J. Radiol. 37, 957 (1964).
8. R.F. Barth, Coderre, M.G.H.Vicente and T.E.Blue, Clin. Can. Res. 11, 3987–4002 (2005).
9. T.F. Budinger et al., Phys. Med. Biol. 16(3), 439–450 (1971).
10. C.E. Floyd et al., Phys. Med. Biol. 51, 3375–3390 (2006).
11. J.E. Bender, A.J. Kapadia, A.C. Sharma, G.D. Tourassi, B.P. Harrawood, and C.E. Floyd,
Med. Phys. 34, 3866–3871 (2007).
12. A.J. Kapadia, A.C. Sharma, J.E. Bender, G.D. Tourassi, C.R. Howell, A.S. Crowell,
M.R. Kiser, B.P. Harrawood, R.S. Pedroni, and C.E. Floyd, IEEE Trans. Nucl. Sci.
55(1), 501–509 (2008).
264 K. Watkin et al.

13. F. Zolzer, C. Streffer, Radiat. Res. 169, 207–213 (2008).

14. Health effects of exposure to low levels of ionizing radiation: BEIR V, National Academic
Press, ISBN 0-309-03997-5.
15. Introduction to Radiological Physics and Radiation Dosimetry, Wiley-VCH, ISBN
0-471-01146-0.
16. N. Adnani and B.G. Fallone, Neutron imaging using Medical Linacs, Proceedings of the
22nd Annual EMBS International Conference, July 23–28, Chicago IL. (2000).
17. Y. Tsuchiya, M. Matsubayashi, T. Takeda, T. Lwin, J. Wu, A. Yoneyama, A. Matsumura,
T. Hori and Y. Itai, Imaging of rabbit VX-2 hepatic cancer by cold and thermal neutron
radiography Jpn. J. Appl. Phys. 42, 7151–7153 (2003).
18. K.K. Moghaddam, T.T. Taheri, and M. Ayubian, Bone structure investigation using
X-ray and neutron radiography techniques, Appl. Radiat. Isot. 66, 39–43 (2008).
19. D. Schwarz, P. Vontobel, E.H. Lehmann, C.A. Meyer, and G. Bongartz, Neutron
tomography of internal structures of vertebrate remains: A comparison with X-ray
computed tomography, Palaeontol. Electron. 8 (2): Art. No. 30A (2005).
20. J.T. Cremer, M.A. Piestrup, H. Park, C.K. Gary, Appl. Phys. Lett. 87, 161913 (2005).
21. B. Allman et al., Nature 408, 158–159 (2000).
22. http://www.ncsa.uiuc.edu/Conferences/2007Meeting/agenda.html, see Bajcy and Watkin
presentation.
23. T.E. Cleveland et al., J. Cereal Sci. 1 (2008).
24. B. Schillinger et al., private communication.
25. M.A. McDonald and K.L. Watkin, Acad. Radiol. 13(4), 407–536 (2006).
26. M.A. McDonald, K.L. Watkin, Invest. Radiol. 38(6), 305–310 2003.
27. K.L. Watkin, M.A. McDonald, Acad. Radiol. 9(S2), 285–289 (2002).
Chapter 15
Neutron Stimulated Emission Computed
Tomography: A New Technique for Spectroscopic
Medical Imaging

A. J. Kapadia

Abstract Neutron stimulated emission computed tomography (NSECT) is being

developed as a new medical-imaging technique to quantify spatial distributions of
elements in a sample through inelastic scattering of fast neutrons and detection of
the resulting gamma rays. It has the potential to diagnose several disorders in the
human body that are characterized by changes in element concentration in
the diseased tissue. NSECT is sensitive to several naturally occurring elements
in the human body that demonstrate concentration changes in the presence of
diseases. NSECT, therefore, has the potential to noninvasively diagnose such
disorders with radiation dose that is comparable to other ionizing imaging
modalities. This chapter discusses the development and progress of NSECT
and presents an overview of the current status of the imaging technique.

Keywords Inelastic scatter  Medical imaging  NSECT  Neutron imaging 

Spectroscopy  Tomography

15.1 Introduction and Background

Neutron spectroscopic techniques have been widely used in medical and biomedi-
cal research to detect the presence of elements in the body. Techniques, such as
instrumental neutron activation analysis (INAA), prompt-gamma neutron activa-
tion analysis (PGNAA), and delayed neutron activation analysis (DNAA), have
been used since the 1960s to identify elements in healthy and diseased tissue, and
techniques based on inelastic scattering of fast neutrons have been used frequently
in body composition studies to measure whole body carbon and oxygen content.
The use of NAA techniques for medical applications was first reported in
1964 by Anderson et al. [1] for measurement of sodium in the body. Between

A.J. Kapadia (*)

Duke Advanced Imaging Laboratories, Department of Radiology, Duke University
Medical Center, Durham, NC 27705, USA
e-mail: [email protected]

I.S. Anderson et al. (eds.), Neutron Imaging and Applications, 265

Neutron Scattering Applications and Techniques, DOI 10.1007/978-0-387-78693-3_15,
Ó Springer ScienceþBusiness Media, LLC 2009
266 A.J. Kapadia

1968 and 1972, Chamberlain reported the measurement of body calcium and
sodium in the body [2–4] and described techniques for whole-body NAA [5–7]
and pulsed NAA [8]. In 1971, Cohn and Dombrowski reported the measure-
ment of calcium, sodium chlorine, nitrogen, and phosphorus in the human
body through in vivo NAA [9]. Since then, NAA and PGNAA have been used
for a variety of applications, such as the measurement of nitrogen [10], carbon
and oxygen [11], cadmium [12], and manganese [13] in the body and in trace
element research to identify cancerous tissue [14–17]. The development of
NAA has been documented in several review articles [18–21], which the reader
may refer to.
Inelastic neutron scatter analysis (INSA) using fast neutrons was initially
reported by Kyere et al. in1982 [22], who used 14 MeV neutrons from a (d,T)
sealed-tube neutron generator to determine whole body carbon content as a
measure of energy expenditure in the body. Following this experiment,
Kehayias et al. reported methods to determine fat content in the body through
measurement of whole body carbon and oxygen through inelastic neutron
scatter [23–25], which they later combined with hydrogen and nitrogen mea-
surement through in vivo NAA [26]. Until recently, most experiments utilizing
fast-neutron inelastic scatter focused primarily on in vivo measurement of
body fat.
The use of nuclear resonance scattering (NRS) was reported by Wielopolski
et al. for detection of iron in the liver [27] and in the heart [28] using an indirect
method of nuclear excitation by gamma rays generated through neutron cap-
ture. Recently, these authors reported the use of 14 MeV neutrons for in vivo
measurement of liver iron through INSA and NRS [29].
In this chapter, we would discuss the development of a tomographic tech-
nique that uses inelastic scattering of fast neutrons and detection of the
resulting gamma rays to quantify spatial distributions of elements in a sample.
The technique, called neutron stimulated emission computed tomography
(NSECT) [30, 31], was pioneered at Duke University in 2003 by the late
Dr. Carey E. Floyd Jr. for the purpose of diagnostic medical imaging.
NSECT uses a beam of fast neutrons to excite stable isotopes of elements in
a sample to determine their concentration and spatial distribution within the
sample. Since it is sensitive to a wide variety of elements that naturally occur in
the human body, it has the potential to noninvasively diagnose several dis-
orders that are characterized by changes in element concentrations in the
diseased tissue [32–34].
Here we present an overview of the NSECT technique and discuss the
current status of the research. The reader may note that while medical imaging
of the human body is the ultimate goal of NSECT, the technique is still in
development stages, with most experiments being performed on phantoms and
excised tissue specimens. The methods and results presented in the chapter are
mainly for the purpose of demonstrating feasibility and proof-of-concept for
future medical applications.
15 Neutron Stimulated Emission Computed Tomography 267

15.1.1 Principle

The principle of NSECT can be described briefly as follows. A neutron incident

on a sample travels freely along its projected path until it collides with an atomic
nucleus of an element present in the sample. If the collision with the atomic
nucleus results in inelastic scatter, the nucleus can get excited to one of its
quantized higher-energy states. The excited nucleus is often unstable and will
rapidly decay to a lower energy state, emitting a gamma-ray photon with energy
equal to the difference of the two states. These energy states are well established
for most elements and isotopes and are mostly unique for the elements com-
monly found in the body. Therefore, the energy of the emitted gamma photon
can be treated as a unique signature of the emitting element. Tomographic
detection and analysis of gamma lines in the emitted spectrum provide quanti-
tative information about the spatial distribution of the element in the sample.
While the principle of NSECT is similar to that of gamma-ray emission
computed tomography (G-ECT) [35], NSECT provides the advantage of
being able to detect stable isotopes of elements that naturally occur in the
body. Unlike G-ECT, NSECT does not require to inject radioactive tracers to
the patient to produce emission decay. Stable isotopes of naturally occurring
elements can be excited through inelastic scatter and detected to obtain spatial
and quantitative information about their distribution in the body.
While NSECT can be used over a wide range of elements and isotopes, the
elements of interest for medical applications show characteristic gamma lines
with energies ranging from 350 to 6500 keV [34]. To induce gamma emission
from these transitions, the energy of the incident neutrons is required to be
greater than the largest energy state to be excited in the target. For most medical
applications, neutron energies of approximately 7–8 MeV are sufficient to
image elements found commonly in the body. NSECT applications at Duke
University typically use neutron energies approximately 1 MeV greater than the
highest excited state desired to compensate for energy loss in propagating
through the body toward the targeted element in the tissue. For example, an
investigation of 12C at 4.4 MeV can be performed with a 5–6 MeV beam,
whereas detection of 56Fe at 847 keV can be performed with a 1.5–2 MeV beam.
As the name suggests, NSECT is a computed tomography application of
neutron inelastic scatter spectroscopy. Tomography is performed using a sin-
gle-beam, fixed-detector geometry similar to the acquisition mode of first-
generation computed tomography systems where a single beam is translated
and rotated around the object in discrete steps until the entire volume of the
object is sampled [36]. In NSECT, a thin neutron beam is used to illuminate a
single line through the object being scanned. Each line, also called a projection,
represents the location of characteristic gamma intensities within the object.
Tomography is performed by translating the object through the beam, rotating
it through a fixed angle, and repeating the translation and rotation until the
entire volume of the object has been scanned. The set of projections obtained
268 A.J. Kapadia

from the tomographic scan can then be reconstructed to obtain a two-dimen-

sional image of the spatial distribution and concentration of each element in the
object. It can be noted that while first-generation CT translated the beam and
detector, the use of Van-de-Graaff sources in NSECT requires that the sample
be moved. In both cases, the final outcome is the same (i.e., the entire volume of
the sample object is scanned with the beam).
Owing to the high gamma energies observed in NSECT, it is impractical to use
standard gamma cameras that are normally used in G-ECT studies. The G-ECT
gamma cameras are unable to provide sufficient spatial or energy resolution for
NSECT imaging. NSECT therefore uses high purity germanium (HPGe) detec-
tors, which exhibit excellent energy resolution within the range of energies
observed in body elements [37]. Unfortunately, HPGe detectors with adequate
efficiency are usually large single-crystal detectors that do not have the ability to
provide spatial information. Spatial information is obtained by tracking the
position of the neutron beam as it illuminates a single projection path at a time
through the sample. Consequently, the current implementation of NSECT
requires a collimated neutron beam to facilitate tomographic acquisition.

15.1.2 Applications in Medical Imaging

NSECT has the potential for use in a variety of applications in both medical and
biological imaging research. It is sensitive to a wide variety of elements that naturally
occur in the human body and has the ability to provide a three-dimensional map
of the elemental distribution within the target object from a single scan. These
characteristics of NSECT have the potential to diagnose several disorders in the
human body that are characterized by changes in element concentrations in
the diseased tissue. Disorders such as iron and copper overload in the liver
[38–41] and several types of cancers including breast [14, 16, 17, 42–48], prostate
[49–51], and brain [52, 53] have been associated with changes in element concentra-
tions. NSECT has the potential to diagnose these disorders through a noninvasive
in vivo scan without the use of radioactive isotope tracers. Three potential medical
and biological applications of NSECT are discussed below.

15.1.2.1 Cancer Diagnosis

Several studies have indicated that changes in trace element concentrations in
human tissue may be a precursor to malignancy in several organs, such as the
brain, prostate, and breast. In vivo measurement of such cancer-marker trace
elements has the potential to differentiate between malignant and benign tissue
at very early stages and enable in vivo cancer diagnosis through element
quantification. Concentration differences measured through NAA have been
reported in the breast [16, 17, 42, 45, 46, 48], colon [54–61], prostate [49, 51,
62–66], kidney [54], liver [50, 54, 67, 68], stomach [54], urinary bladder [54],
15 Neutron Stimulated Emission Computed Tomography 269

testis, and female reproductive organs [54]. NSECT cancer diagnosis can be
performed by measuring the trace element composition of the tissue and analyzing
this information to identify combinations of elements that are markers of cancer.
Owing to NSECT’s tomographic ability, given adequate sensitivity, it can identify
both the location of the tumor as well as its cancer state from a single scan.

15.1.2.2 Noninvasive Measurement of Liver Iron and Copper

A second application of NSECT is in the diagnosis of liver disorders such as
hemochromatosis [38–40] and Wilson’s disease [41], which exhibit changes in
iron and copper concentrations in the liver, respectively. Iron overload is a
serious concern for patients with thalassemia, sickle cell anemia, and hemo-
chromatosis mutations and for other patients who receive chronic blood
transfusions. Similarly, copper overload is observed in patients with Wilson’s
disease. The current definitive diagnostic technique for these disorders is
invasive liver biopsy followed by chemical analysis of iron to determine the
severity of the disorder [38, 40, 69]. NSECT provides a noninvasive alternative
with the potential to obtain the same information without the need for a
biopsy. Several different techniques (reported in [33]) have attempted to detect
and quantify iron concentration in the liver in vivo, for example, X-ray
computed tomography (X-ray CT), MRI, magnetic resonance spectroscopy
(MRS), super-conducting quantum interference susceptometry (SQUID),
and nuclear resonance scattering (NRS) [27, 70–88]. While each has had
varying degrees of success, few have found widespread acceptance in the
clinical environment. Both X-ray CT and MRI lack the sensitivity required
for detecting mild to moderate degrees of iron overload, and MRI also shows
reduced sensitivity in cases of severe iron overload [75, 76, 80, 89]. In MRS, the
low signal from bound iron makes quantification studies difficult. MRS also
has limited resolution (larger than 1 cm) [90]. While SQUID has found some
success in detecting liver iron, there are only about four SQUID facilities in
the world that offer iron overload testing, making it difficult to translate into a
widely used clinical alternative. NRS has achieved some success in patients
with iron overload but with high levels of dose [27]. NSECT has the potential
to overcome many of these limitations and develop into a clinical technique
for noninvasive measurement of liver iron and copper [33].

15.1.2.3 Small Animal Imaging

A third potential area of application for NSECT is in the spectroscopic imaging
of small animals [91, 92] to understand the structural and functional mechanisms
of metabolism and transport of elements in normal and disease models. For
example, NSECT can be used to follow the metabolic migration of nutrients in
normal and in genetically modified mice, measurement of iron and calcium levels in
the cardiac wall, and molecular transport and drug delivery studies where mole-
cules of interest are tagged with stable isotopes rather than radioactive tracers.
270 A.J. Kapadia

NSECT represents a technique that combines both anatomic and functional

imaging. Through CT acquisition, it is possible to obtain a three-dimensional
map of element distribution within a specific organ, thereby providing
anatomic information about the location and extent of the diseased tissue.
Simultaneously, information about elements within the tissue can provide
information about the functional (disease) state of the tissue.

15.2 Imaging Facility and Apparatus

All NSECT experiments performed so far have been conducted at the Triangle
Universities Nuclear Laboratory (TUNL) accelerator facility located at Duke
University. TUNL is a low energy nuclear physics research laboratory with a
10 MeV tandem Van-de-Graaff accelerator capable of producing 20 MeV proton
or deuteron beams and a host of other light ion beams. The shielded neutron
source provides researchers with an intense pulsed collimated mono-energetic
neutron beam with energies between 5 and 23.2 MeV. A detailed description of
the facilities and apparatus used in NSECT has been provided in [32, 34].
Three key components are required in an NSECT acquisition system:
(a) neutron source, (b) gamma detector, and (c) tomographic acquisition and
reconstruction technique. Each of these components is readily available at
TUNL. Figure 15.1 shows a schematic of an NSECT system with the major
components. A description of the key components is given below.

Fig. 15.1 Schematic of an NSECT system used in tomographic measurements. Major components
of the system are labeled as A – Neutron Collimator; B – Instantaneous Neutron Flux Monitor;
C – Sample Manipulation Tomographic Gantry; D – HPGe detector; E – Anticoincidence
Compton Shield; and F – Inelastically Scattered Neutron. The neutron beam (shown by the line
of dots) travels from the source to the sample (C) and scatters inelastically. Gammas emitted in the
inelastic scatter interaction (shown as line of dashes) are detected by the HPGe detector, while the
scattered neutron (F) continues along another trajectory (shown as a solid line).
(Figure modified from [94] pp. 2313–2326. # Institute of Physics and IOP Publishing
Limited 2008)
15 Neutron Stimulated Emission Computed Tomography 271

15.2.1 Neutron Source

NSECT can be performed with a variety of neutron sources, some of which are
available commercially, while others are specialized sources available only at
state-of-the-art neutron facilities. An ideal source for an NSECT application
should produce a narrow collimated beam of fast neutrons with high neutron
flux and allow beam pulsing and energy tuning with a nearly monochromatic
energy profile. High neutron flux is required to achieve minimal scan durations,
while a collimated beam is required to attain spatial resolution in the tomo-
graphic acquisition. Beam pulsing or chopping is necessary to identify the time-
resolved inelastic scatter signal and separate it from other neutron and gamma
effects that occur later in time. Without beam pulsing (i.e., with a continuous-
beam source), the inelastic scatter signal can be obscured by radiation
detected from room and sample-related background effects that are observed
at different times, such as radioactive decay following neutron capture in the
sample. These effects can be reduced through time-of-flight (TOF) correction,
which needs beam pulsing. The TOF correction technique for NSECT has been
described in detail in a separate publication [93].
Energy tuning, while highly desirable, is not an essential requirement for a
neutron source. A requirement of inelastic scatter spectroscopy is that the energy
of the neutron beam must be higher than the maximum energy state to be excited.
However, unnecessarily high neutron energies will simply excite higher unwanted
states in both desired and undesired elements, leading to a busy spectrum
cluttered with noise. Further, as NSECT detectors are susceptible to Compton
noise from high gamma energies, exciting unwanted high energy states in the
sample will also produce higher Compton background noise signals in the
detectors, which can potentially obscure elements with low-lying characteristic
gamma lines. A monochromatic energy source producing neutrons with energy
slightly higher than the highest desired excited state is the most suitable. With a
tunable neutron source, a beam can be produced with energy that is sufficient to
excite the elements of interest and avoid other unwanted elements in the sample.
For example, to image carbon in a biological sample (at 4.439 MeV), a neutron
energy of 5.0 – 5.5 MeV is ideal as it excites the 4.439 MeV energy state in 12C but
avoids excitation of the 6.13 MeV excited state in 16O. A monochromatic source
also facilitates lower patient dose by restricting neutron illumination to the
required range of energy. Higher neutron energies that contribute to dose by
depositing more energy per interaction and lower thermal energies that contri-
bute to dose through neutron capture and subsequent radioactive emission are
eliminated in a true monochromatic source.
Table 15.1 lists some neutron sources that have been identified for use in
NSECT: Van-de-Graaff accelerators, deuterium-deuterium (DD) neutron
tubes, and deuterium-tritium (DT) neutron tubes. Van-de-Graaff accelerators
provide tunable collimated neutron beams that are suitable for tomography.
NSECT experiments performed thus far have used a Van-de-Graaff accelerator
272 A.J. Kapadia

Table 15.1 A comparison of three neutron sources suitable for NSECT applications
Characteristic Van-de-Graaff DD Neutron Tube DT Neutron Tube
Energy 5 MeV to 23.2 MeV 3.2 MeV 14 MeV
Flux 104 – 106 n/s.cm2 108 – 1011 n/s 1010 – 1013 n/s
Pulsing 2 ns @ 2.5 MHz 100 ms 100 ms
Collimation Available Available Available
Monochromatic Beam Yes Yes Yes
Energy Tuning Yes No No

at Duke University as the source of neutrons. The accelerator produces a

tunable, monochromatic, pulsed neutron beam, which can be tuned to energies
between 5 and 23.2 MeV and produces collimated fluxes of 104–106 n/s.cm2.
The beam can be collimated to any desired shape and size using collimator
inserts and is pulsed to provide 2 ns wide bunches at the target to allow
measurement of neutron and gamma TOF. The repetition rate of the beam
pulses is adjustable from 2.5 MHz down to 39 kHz [32]. Beams measuring
0.8 cm at full width at half maximum have been used to obtain image resolution
of approximately 1 cm [94].
DD and DT neutron tubes produce high-flux monochromatic beams and
possess the advantage of being commercially available. DD tubes produce
neutrons at 3.2 MeV, while DT tubes produce neutrons at 14.2 MeV. Owing
to the required neutron energy constraint, DD tubes are not suitable for
applications that require the imaging of carbon or oxygen. Both tubes can be
pulsed down to 100 ms or operated in continuous mode, and collimation is
available at the cost of neutron flux.
A thin neutron plastic scintillator attached to a photomultiplier tube acts as
an instantaneous neutron flux monitor to measure fluctuations in the neutron
beam current. The attenuation of neutrons by the flux monitor in the energy
range of current NSECT studies is typically less than 0.1%.

15.2.2 Gamma Detector

As mentioned earlier, most elements of interest for medical NSECT applica-
tions show characteristic gamma lines with energies ranging from 350 to
6500 keV. In some elements, gamma lines differ by as little as 4 keV (e.g., 27Al
at 843.7 keV and 56Fe at 846.7 keV). NSECT gamma detectors must therefore
possess excellent energy resolution as well as high efficiency at high gamma
energies. Standard gamma camera detectors used in nuclear medicine studies
are unable to provide the energy resolution and efficiency required and, there-
fore, cannot be used for NSECT applications.
15 Neutron Stimulated Emission Computed Tomography 273

Several gamma-ray detectors have been explored and evaluated for use in
NSECT, including single-crystal and clover HPGe detectors and the compara-
tively inexpensive options NaI-Tl and BGO. For multielement applications,
such as cancer diagnosis, only HPGe detectors are found to be suitable, provid-
ing the most desirable combination of energy resolution and efficiency. HPGe
detectors typically possess energy resolution of 0.1%, providing approximately
1 keV resolution at 1 MeV. BGO and NaI-Tl detectors provide typical energy
resolutions of approximately 7 and 12%, respectively, which is insufficient to
separate close-lying gamma lines in a multielement analysis. However, these
detectors are suitable for single-element evaluations where the element of interest
lies in a region, which is free from close-lying gamma lines from other elements.
NSECT experiments performed at Duke University have used two HPGe
segmented detectors in the two fold segmented clover detector configuration. In
this configuration, each detector consists of four coaxial n-type germanium
crystals mounted together in the shape of a 4-leaf clover. Each detector exhibits
minimum relative efficiency of 22% (relative efficiency compares the efficiency
of the detector at 1332 keV to that of a 3 inch cubic NaI detector) and full width
at half maximum less than or equal to 2.25 keV for 1.332 MeV gamma rays of
60
Co. Detectors are typically calibrated against a 22Na source and positioned at
135 degrees from the incident neutron beam. A majority of the elements of
interest to NSECT decay through electric quadrupole transitions, whose dis-
tribution has maximum intensity at 45 and 135 degrees. Placing the detectors at
the 135 degree orientation helps to maximize signal intensity and simulta-
neously prevent detector activation and damage from forward scattering neu-
trons. In Figure 15.1, the HPGe detector (labeled D) is visible surrounded by an
anticoincidence Compton shield (labeled E), which is used to reduce the effects
of Compton scattering in the detectors. These shields have a minimum peak-to-
Compton ratio of 41 for 60Co gamma rays at 1.332 MeV [32].

15.2.3 Tomographic Acquisition Gantry

As with all other tomographic imaging systems, NSECT performs tomography
by acquiring a complete set of projections of line integrals through the sample.
For NSECT, these projections are defined by the angular and spatial sampling
intervals between successive positions of the beam on the sample. Acquisition is
performed in a manner similar to first generation CT with a stationary beam
and single element detector located at fixed positions with respect to each other.
The sample is translated through the beam, rotated through a finite angle, and
translated through the beam again. The process is repeated until the volume of
interest in the sample has been scanned. While ideal tomography systems would
acquire projections between 0 and 360 degrees, due to limitations in acquisition
time and concerns about patient dose, the current NSECT tomography system
scans the sample from 0 to 180 degrees and employs the maximum likelihood
expectation maximization (MLEM) technique [95] for reconstruction. The
274 A.J. Kapadia

MLEM technique is able to take into account parameters such as nonuniform

and partial sampling when reconstructing an image.
Tomography is performed using the single-beam, fixed-detector geometry of
first-generation computed tomography systems where a thin neutron beam is
used to illuminate a single line (projection) through the object being scanned.
Tomography is performed by translating the object through the beam, rotating
it through a fixed angle, and repeating the translation and rotation until the
entire volume of the object has been scanned. The set of projections obtained
from the tomographic scan are then reconstructed to obtain a two-dimensional
image of the spatial distribution and concentration of each element in the
object.
Figure 15.1 shows the tomographic computer-controlled gantry (labeled C),
which enables tomographic acquisition by translating and rotating the sample
with respect to the neutron beam.
Reconstruction can be performed through any available reconstruction
algorithm. While a variety of algorithms are available, the MLEM algorithm
is selected due to its superiority of performance in undersampled acquisition
systems and its ability to incorporate the Poisson nature of photon counting
[95], both characteristics important to NSECT. An analysis of the MLEM
algorithm and its advantages over other iterative and Fourier techniques for
image reconstruction in NSECT has been described in [96].
Data acquisition can be performed using any nuclear spectroscopy acquisi-
tion system that allows acquisition and storage of raw data and offline retrieval
for data visualization and analysis. Examples of such software that has been
used at Duke include CODA [97] and SpecTcl [98].

15.3 Current Applications

There are many diverse and exciting applications where NSECT can potentially
be used to make a clinical diagnosis through noninvasive tomography. Dis-
orders, such as cancer, hemochromatosis (liver iron overload), and Wilson’s
disease (liver copper overload), can be diagnosed by identifying the concentra-
tions of the respective disease-marking elements in the tissue or organ. Several
experiments have been performed to investigate the feasibility of diagnosing
these disorders through NSECT, using both tomographic acquisition as well as
nontomographic (single projection) spectroscopy. The results of some of these
experiments are described below.

15.3.1 Experiment 1: Tomographic Acquisition

of Multielement Phantom
Figure 15.2 shows an image of a solid iron-copper phantom containing bars of
natural iron and natural copper arranged in the shape of the letter ‘N.’ Each bar
15 Neutron Stimulated Emission Computed Tomography 275

Fig. 15.2 Geometry of the phantom imaged in the tomography experiment. The vertical outer
bars represent copper while the diagonal inner (gray) bars represent iron. Each bar measures
0.6 cm by 6 cm by 2.5 cm.
(Figure from Floyd et al, ‘‘Neutron Stimulated Emission Computed Tomography of a Multi-
Element Phantom,’’ Phys Med Biol, vol. 53, pp. 2313–2326. # Institute of Physics and IOP
Publishing Limited 2008)

measures 6 mm in width, 60 mm in height, and 25 mm in depth. The phantom

was scanned with an 8 mm wide 7.5 MeV neutron beam with 11 spatial projec-
tion at 8 angles. As can be seen in Fig. 15.3, the corresponding NSECT
reconstructed image shows excellent agreement with the spatial locations of
iron and copper within the image. Images were reconstructed individually for
each of the two elements and then combined to generate the figure shown.
Reconstructed resolution was observed to be approximately 1 cm, which is
reasonable for a beam that measured 8 mm at full width at half maximum. The
experiment demonstrates the tomographic ability of NSECT in identifying
spatially individual elements in a multielement sample.
Figure 15.4 shows the corresponding gamma spectrum from the NSECT
scan of the iron-copper phantom. Six spectral lines can be identified for energy
transitions from iron and copper in the phantom.
Detailed analysis and a complete description of the experiment are available
in [94].

15.3.2 Experiment 2: Diagnosis of Iron Overload in Human

Liver Phantom
Figure 15.5 shows a human torso phantom with two chambers – an outer
chamber corresponding to an adult torso and an inner chamber corresponding
276 A.J. Kapadia

Fig. 15.3 Reconstructed image from the NSECT acquisition of the sample. The vertical outer
regions represent copper while the diagonal inner region represents iron. Each element was
reconstructed separately and then combined.
(Figure from Floyd et al, ‘‘Neutron Stimulated Emission Computed Tomography of a Multi-
Element Phantom,’’ Phys Med Biol, vol. 53, pp. 2313–2326. # Institute of Physics and IOP
Publishing Limited 2008)

Fig. 15.4 Gamma energy spectrum from the iron-copper phantom showing spectral lines
from six transitions in 56Fe and 63Cu:
1. 63Cu from 1st excited state to ground state; energy 660 keV
2. 56Fe from 1st excited state to ground state; energy 847 keV
3. 63Cu from 2nd excited state to ground state; energy 962 keV
4. 56Fe from 3rd to 2nd excited state; energy 1239 keV
5. 56Fe from 4th to 2nd excited state; energy 1811 keV
6. 63Cu from 6th to 1st excited state; energy 1864 keV
(Figure from [94] pp. 2313–2326. # Institute of Physics and IOP Publishing Limited 2008)
15 Neutron Stimulated Emission Computed Tomography 277

Fig. 15.5 A phantom of the human torso with two chambers – outer chamber corresponding
to an adult torso, and inner chamber corresponding to an adult human liver. Both chambers
can be filled separately with any desired material.
(Figure from Kapadia et al. [32], pp. 2633–2649. # Institute of Physics and IOP Publishing
Limited 2008)

to an adult human liver. The phantom was filled with bovine liver tissue with
artificially induced iron overload and was scanned using a 5 MeV neutron
beam. The resultant spectrum shown in Fig. 15.6 was analyzed to quantify
the concentration of iron in the liver. The gamma line at 847 keV was detected
for 4.18 g of 56Fe. Based on this spectrum, a clinically relevant projected
sensitivity of 6 mg/g was obtained for iron overload diagnosis. In addition,
gamma lines were identified for several other elements in the liver, including Cl,
Cu, K, Na, and Zn, which were confirmed through NAA, and for 12C from the
phantom. Gamma lines were also identified for 74Ge and 76Ge from the gamma-
ray detectors, 42K from the room background, and 1H (neutron capture) from
the phantom. The dose from the scan was calculated using a Monte-Carlo
simulation as 0.375 mSv [33], which is significantly lower than an abdominal
X-ray exam that typically delivers 2 mSv [99]. The experiment demonstrates
that NSECT has the potential to detect clinically relevant concentrations of
iron in the human body through a noninvasive scan with reasonable dose.
Although the scan time for this experiment was unreasonable at over 24 h, it
278 A.J. Kapadia

Fig. 15.6 Gamma energy spectrum from the uniform iron overload torso phantom, showing
peaks corresponding to elements detected in the liver. The peak at 847 keV corresponds to
56
Fe. Peaks are also seen for Ge from the detector and 12C and 1H from the tissue.
(Figure from Kapadia et al. [32], pp. 2633–2649. # Institute of Physics and IOP Publishing
Limited 2008)

can be brought down to a few seconds using high-flux neutron sources (such as
the aforementioned DD and DT tubes) and increasing the number of detectors.
Detailed analysis and a complete description of the experiment are available
in [33].

15.3.3 Experiment 3: Detection of Breast Cancer

To investigate the feasibility of breast cancer detection through NSECT

(Figs. 15.7 and 15.8), samples of benign and malignant breast tissue obtained
through breast biopsy were scanned with a 6 MeV neutron beam for a cumulative
dose of 1 mSv each. After TOF and sample-out background subtraction, the
resultant spectra were analyzed to identify elements that showed statistically
significant differences between the two samples. Concentration differences were
found for Al, Br, Cl, Co, Fe, K, Rb, and Zn, of which Cl, Fe, K, and Zn were
verified through NAA. The experiment demonstrates that NSECT can poten-
tially be used to detect several cancer-marking elements in human tissue.
Table 15.2 gives a list of elements that showed statistically significant differ-
ences between the benign and malignant spectra. While several elements were
identified with concentration differences in each of the two samples, NAA
analysis of the samples determined that their concentration was of the order
15 Neutron Stimulated Emission Computed Tomography 279

Fig. 15.7 NSECT spectrum of a benign breast sample showing elements identified through
NSECT spectroscopy.
(Figure from Kapadia et al. [32], pp. 501–509. # 2008 IEEE)

Fig. 15.8 NSECT spectrum of a malignant breast sample showing elements identified through
NSECT spectroscopy.
(Figure from Kapadia et al. [32], pp. 501–509. # 2008 IEEE)
280 A.J. Kapadia

Table 15.2 List of elements showing statistically significant differences between benign and
malignant spectra. Elements with negative differences showed a decrease in concentration in
the malignant sample. Statistical significance was calculated using a z-score test for difference
of means. (Table from Kapadia et al. [32], pp. 501–509. # 2008 IEEE)
Energy keV Element match Counts benign Counts malignant Diff p-val
79
219 Br 6 19 13 0.01
59
397 Co, 79Br 16 2 –14 0.01
81
1028 Br 13 29 16 0.05
39
1128 K, 68Zn 0 13 13 0.001
56
1306 Fe 10 0 –10 0.01
27
2299 Al 0 13 13 0.001
37
2469 Cl, 56Fe, 66Zn 5 15 10 0.05
35
3635 Cl 3 14 11 0.01

of a few hundred micrograms, which is below the sensitivity of the current system,
which is a few hundred milligrams [32]. Therefore, accurate quantification of these
microgram concentrations will require a significant improvement in sensitivity.
Detailed analysis and a complete description of the experiment are available
in [32].

15.4 Dose Analysis

Patient dose in NSECT is of significant concern due to the use of fast neutrons, which
have a dose quality factor (Q-factor) of 10. For the technique to translate successfully
to the clinical environment, NSECT dose must be comparable to the dose levels of
other ionizing imaging modalities. As the technique is still in early stages of develop-
ment, dose analysis in NSECT is currently performed using Monte-Carlo simulations.
The process, described in [96, 100], can be summarized in the following three steps:
(a) a Monte-Carlo simulation is used to estimate two parameters for an inci-
dent neutron beam – the number of neutrons that interact in the volume of
interest and the average energy deposited per interacting neutron,
(b) the resulting energy deposited in the volume is converted from MeV to J/kg
using the known mass of the volume to give the absorbed energy in Gray (Gy),
and
(c) the absorbed energy is multiplied by the quality factor for neutrons (10) and
the weighting factor for the organ of interest to give the effective dose
equivalent in Sieverts (Sv).
This technique has been used to calculate the patient dose for NSECT scans
of the abdomen [96], liver [33, 96, 100], and breast [100, 101]. Table 15.3
summarizes the results of these dose-analysis simulations.
In comparison, an abdominal X-ray scan typically delivers 2 mSv, a mam-
mogram delivers 0.7 mSv, while an abdominal CT exam delivers approximately
10 mSv [99]. The doses from NSECT scans appear comparable to, or even lower
than, the other techniques. This is largely due to the modest number of neutrons
15 Neutron Stimulated Emission Computed Tomography 281

Table 15.3 NSECT dose delivered to organs in the body from NSECT scans
Organ Spectroscopic scan Tomography scan
Abdomen 1–2 mSv 1–5 mSv
Liver 0.02–1 mSv 0.5–3 mSv
Breast 0.02–0.5 mSv 0.5–1 mSv

required to create an NSECT image or extract sufficient information from the

sample. Preliminary simulation studies indicate that as few as 10 million
incident neutrons are required to obtain quantitative accuracy of 95% in a
tomographic image [102]. A fluence of 10 million incident neutrons on the
breast, for example, corresponds to a dose of less than 1 mSv. Dose levels in
NSECT can potentially be reduced even further by increasing the number of
gamma-ray detectors and using high-flux neutron sources to reduce scan time.

15.5 Summary

NSECT represents an exciting new imaging modality that has the potential for
application in both medical and biological research. Several human disorders
characterized by element changes can be diagnosed through noninvasive in vivo
scanning using this technique. NSECT has the ability to obtain tomographic
information about the spatial distribution of elements within a tissue or organ to
make a quantitative and spatial diagnosis. A prototype of the NSECT acquisition
system has been developed and built at Duke University using a Van-de-Graaff
accelerator and HPGe detectors. As demonstrated through experiments with the
prototype system, NSECT has the sensitivity to detect concentrations of iron that
represent a clinically relevant liver iron overload condition. Sensitivity evaluation
experiments indicate that concentrations as low as 3 mg/g may be quantifiable
through NSECT. Although MRI is able to quantify moderate concentrations of
iron overload, it suffers from a loss of accuracy for concentrations above 6 mg/g
wet weight (20 mg/g dry weight) due to a reduction in the signal intensity caused by
high concentrations of iron [89]. NSECT, on the other hand, shows an increase in
the signal with increasing iron concentration. This facilitates iron overload detec-
tion in patients with severe overload, where MRI begins to lose accuracy.
The image resolution observed with the prototype system is approximately
1 cm, which is passable for imaging large organs such as the liver. However, as
the resolution depends primarily on the width of the neutron beam, it is possible to
improve the resolution significantly by using a narrow collimated beam (at the cost
of additional scan time). Simulation experiments have demonstrated a resolution
of 5 mm when scanning was performed with a 5 mm rectangular beam [96]. While
the resolution of the NSECT system depends primarily on the size of the incident
neutron beam, the best resolution achievable in a hydrogen-rich biological sample
will also be limited by the noise component generated from neutron scatter by
hydrogen. Neutrons that are meant to illuminate a certain voxel in the sample may
282 A.J. Kapadia

scatter onto adjacent voxels containing an element of interest and induce inelastic
gamma emission from that voxel. Detection of the adjacent-voxel gammas may
lead to a spread in the tails of the resulting sinograms, which can lead to reduced
resolution. The contribution of this effect to the limits of resolution is currently
being investigated through Monte-Carlo simulations.
Patient dose for NSECT examinations of different organs have been found
to vary between 0.02 mSv for a breast spectroscopic exam [101] to under 1 mSv
[96] for a liver spectroscopic exam, evaluated through Monte-Carlo simula-
tions. Tomographic doses are slightly higher depending on the number of
projections used but are generally lower than the dose delivered from a corre-
sponding X-ray CT exam [96].

15.6 Future

The current focus of NSECT has been on quantitative diagnosis of large-organ

disorders (e.g., those of the liver), largely due to the limits of resolution and
sensitivity observed in the prototype system. As these limiting effects are under-
stood better, efforts to improve the sensitivity and resolution of the system are
being made. Improved sensitivity will facilitate the diagnosis of a greater number
of disorders, including several types of cancer. The use of portable neutron
sources such as pulsed DD and DT tubes is being explored to develop a portable,
high-flux scanning solution for use in a clinical environment. Such portable
tomographic scanning systems with improved resolution and sensitivity will
present an attractive method for human and small-animal imaging and for
diagnosis and screening of several disorders in the human body.

Acknowledgment The author would like to acknowledge and thank all the members of the
Duke Advanced Imaging Laboratories (DAILabs) and Triangle Universities Nuclear
Laboratory (TUNL) who have been involved in the development of NSECT, especially
Georgia Tourassi, Amy Sharma, and Janelle Bender for their analytical contribution and
deep involvement in NSECT; Brian Harrawood for his unparalleled computing support;
Calvin Howell, Alexander Crowell, Matthew Kiser, and Robert Macri for their help and
guidance with NSECT acquisition experiments; and Anton Tonchev and Anthony Hutcheson
for their help with gamma detector setup and management. Finally, the author would like to
express his deep gratitude to Dr. Carey Floyd, the pioneer of NSECT, in whose memory and
name this research continues.

References
1. J. Anderson, S.B. Osborn, R.W. Tomlinson, D. Newton, J. Rundo, L. Salmon, and J.W.
Smith, Neutron-Activation Analysis in Man in Vivo. a New Technique in Medical Inves-
tigation, Lancet 2, 1201–1205, (Dec 5 1964).
2. M.J. Chamberlain, J.H. Fremlin, D.K. Peters, and H. Philip, Total body sodium by whole
body neutron activation in the living subject: further evidence for non-exchangeable
sodium pool, Br. Med. J. 2, 583–585 (Jun 8, 1968).
15 Neutron Stimulated Emission Computed Tomography 283

3. M.J. Chamberlain, J.H. Fremlin, D.K. Peters, and H. Philip, Total body calcium by
whole body neutron activation: new technique for study of bone disease, Br. Med. J. 2,
581–3, Jun 8 1968.
4. M.J. Chamberlain, J.H. Fremlin, D.K. Peters, and H. Philip, Measurement of whole body
calcium and sodium by neutron activation analysis in the living subject, Strahlentherapie
[Sonderb], 67, 178–85, 1968.
5. M.J. Chamberlain, Whole-body neutron activation analysis, Proc. R. Soc. Med. 62, 370–3
(Apr 1969).
6. M.J. Chamberlain, J.H. Fremlin, I. Holloway, and D.K. Peters, Use of the cyclotron for
whole body neutron activation analysis: theoretical and practical considerations, Int. J.
Appl. Radiat. Isot. 21, 725–34 (Dec 1970).
7. M.J. Chamberlain, J.H. Fremlin, D.K. Peters, and H. Philip, Applications of the whole-
body counter in total-body neutron activation analysis, Br. J. Radiol. 43, 287–8 (Apr
1970).
8. M.J. Chamberlain and J.H. Fremlin, Measurement of whole-body nitrogen by pulsed
neutron activation analysis, Strahlentherapie [Sonderb] 72, 88–93 (1972).
9. S.H. Cohn and C.S. Dombrowski, Measurement of total-body calcium, sodium, chlorine,
nitrogen, and phosphorus in man by in vivo neutron activation analysis, J. Nucl. Med. 12,
499–505 (Jul 1971).
10. D. Vartsky, K.J. Ellis, and S.H. Cohn, In vivo measurement of body nitrogen by analysis
of prompt gammas from neutron capture, J. Nucl. Med. 20, 1158–1155 (Nov 1979).
11. C.L. Hollas, L.E. Ussery, K.B. Butterfield, and R.E. Morgado, A method for in vivo
determination of carbon and oxygen using prompt gamma radiations induced by
14.7-MeV neutrons, Basic Life Sci. 55, 395–400 (1990).
12. H.C. Biggin, N.S. Chen, K.V. Ettinger, J.H. Fremlin, W.D. Morgan, R. Nowotny,
M.J. Chamberlain, and T.C. Harvey, Cadmium by in vivo neutron activation analy-
sis, in J. Radioanal. Chem, v. 19, no. 2, pp. 207–214; International colloquium on
activation analysis of very low amounts of elements; 2 Oct 1972; Saclay, France United
Kingdom (1974).
13. M.L. Arnold, F.E. McNeill, I.M. Stronach, A. Pejovic-Milic, D.R. Chettle, and A.
Waker, An accelerator based system for in vivo neutron activation analysis measurements
of manganese in human hand bones, Med. Phys. 29, 2718–24 (Nov 2002).
14. A. Garg, V. Singh, et al., An elemental correlation study in cancerous and normal breast
tissue with successive clinical stages by neutron activation analysis, Biol. Trace Element
Res. 46, 185–202 (1994).
15. A. Danielsen and E. Steinnes, A study of some selected trace elements in normal and
cancerous tissue by neutron activation analysis, J. Nucl. Med. 11, 260–4 (Jun 1970).
16. K.H. Ng, D.A. Bradley, and L.M. Looi, Elevated trace element concentrations in
malignant breast tissues, Br. J. Radiol. 70, 375–82 (Apr 1997).
17. K.H. Ng, D.A. Bradley, L.M. Looi, C.S. Mahmood, and A.K. Wood, Differentiation of
elemental composition of normal and malignant breast tissue by instrumental neutron
activation analysis, Appl. Radiat. Isot. 44, 511–6 (Mar 1993).
18. K.J. Ellis, Human body composition: in vivo methods, Physiol. Rev. 80, 649–80 (Apr
2000).
19. J.F. Sutcliffe, A review of in vivo experimental methods to determine the composition of
the human body, Phys. Med. Biol. 41, 791–833 (May 1996).
20. E. Witkowska, K. Szczepaniak, and M. Biziuk, Some applications of neutron activation
analysis: A review, J. Radioanal. Nucl. Chem. 265, 141–150 (2005).
21. S. Mattsson and B.J. Thomas, Development of methods for body composition studies,
Phys. Med. Biol. 51, R203–28 (Jul 7 2006).
22. K. Kyere, B. Oldroyd, C.B. Oxby, L. Burkinshaw, R.E. Ellis, and G.L. Hill, The
feasibility of measuring total body carbon by counting neutron inelastic scatter gamma
rays, Phys. Med. Biol. 27, 805–17 (Jun 1982).
284 A.J. Kapadia

23. J.J. Kehayias, S.B. Heymsfield, A.F. LoMonte, J. Wang, and R.N. Pierson, Jr., In vivo
determination of body fat by measuring total body carbon, Am. J. Clin. Nutr. 53, 1339–44
(Jun 1991).
24. J.J. Kehayias, Aging and body composition: possibilities for future studies, J. Nutr. 123,
454–8 (Feb 1993).
25. J.J. Kehayias, H. Zhuang, V. Hughes, and L. Dowling, Assessment of body fat and lean in
the elderly by measuring body carbon and oxygen: validation against hydrodensitometry,
Appl. Radiat. Isot. 49, 723–5 (May-Jun 1998).
26. J.J. Kehayias, S. Valtuena, A.B. Waitekus, C.A. Sheahan, and M. O’Neill, In vivo
elemental partition analysis using fast neutrons. A tool for testing the efficacy of new
clinical interventions, Ann. N Y. Acad. Sci. 904, 140–7 (May 2000).
27. L. Wielopolski, R.C. Ancona, R.T. Mossey, A.N. Vaswani, and S.H. Cohn, Nuclear reso-
nance scattering measurement of human iron stores, Med. Phys. 12, 401–4 (Jul-Aug 1985).
28. L. Wielopolski and E.C. Zaino, Noninvasive in-vivo measurement of hepatic and cardiac
iron, J. Nucl. Med. 33, 1278–82 (Jul 1992).
29. L. Wielopolski, Feasibility of measuring iron in vivo using fast 14 MeV neutrons, Cooley’s
Anemia Foundation Inc. BNL-7980–2005 (2005).
30. C.E. Floyd, C.R. Howell, B.P. Harrawood, A.S. Crowell, A.J. Kapadia, R. Macri, J.Q.
Xia, R. Pedroni, J. Bowsher, M.R. Kiser, G.D. Tourassi, W. Tornow, and R. Walter,
Neutron Stimulated Emission Computed Tomography of Stable Isotopes, in SPIE
Symposium on Medical Imaging, San Diego, CA, pp. 248–254 (2004).
31. C.E. Floyd, J.E. Bender, A. Sharma, A. Kapadia, J. Xia, B. Harrawood, G.D. Tourassi,
J. Lo, and C.R. Howell, Introduction to Neutron Stimulated Emission Computed Tomo-
graphy. Phys. Med. Biol. 50 (14), 3375–90 (2005).
32. A.J. Kapadia, A.C. Sharma, J.E. Bender, G.D. Tourassi, C.R. Howell, A.S. Crowell,
M.R. Kiser, B.P. Harrawood, and C.E. Floyd, Neutron Stimulated Emission Com-
puted Tomography for Diagnosis of Breast Cancer, IEEE Trans Nucl. Sci. 55, 501–509
(2008).
33. A.J. Kapadia, G.D. Tourassi, A.C. Sharma, A.S. Crowell, M.R. Kiser, and C.R. Howell,
Experimental detection of iron overload in liver through neutron stimulated emission
spectroscopy, Phys. Med. Biol. 53, 2633–2649 (2008).
34. C.E. Floyd, J.E. Bender, A.C. Sharma, A.J. Kapadia, J.Q. Xia, B.P. Harrawood, G.D.
Tourassi, J.Y. Lo, A.S. Crowell, and C.R. Howell, Introduction to neutron stimulated
emission computed tomography, Phys. Med. Biol. 51, 3375–3390 (2006).
35. M.N. Wernick and J.N. Aarsvold, Emission Tomography: The Fundamentals of PET and
SPECT: Academic Press, San Diego (2004).
36. A. Kak and M. Slaney, Principles of computerized tomographic imaging vol. 33. Philadel-
phia: society for industrial and applied mathematics (2001).
37. J.E. Bender, C.E. Floyd, B.P. Harrawood, A.J. Kapadia, A.C. Sharma, and J.L. Jesneck,
The effect of detector resolution for quantitative analysis of neutron stimulated emission
computed tomography, in SPIE Medical Imaging, pp. 1597–1605 (2006).
38. L.W. Powell, Diagnosis of hemochromatosis, Semin. Gastrointest Dis. 13, 80–8 (Apr 2002).
39. S. Joffe, Hemochromatosis, N. Lamki, B. Coombs, U. Schmiedl, R.M. Krasny, and J. Karani,
Eds., Medscape emedicine, emedicine.medscape.com/article/369012-overview, Mar 11, 2005.
40. L. Powell, Hemochromatosis, in D. Kasper, Fawci, AS, Longo, DL, Braunwald, E,
Hauser, SL, Jameson, JL, Ed.,Harrison’s Principles of Internal Medicine, 16 ed. vol. 2,
McGraw Hill, New York, NY, pp. 2298–2303 (2005).
41. G. Brewer, Wilson Disease, in D. Kasper, Fawci, AS, Longo, DL, Braunwald, E, Hauser,
SL, Jameson, JL, Ed., Harrison’s Principals of Internal Medicine, 16 ed. vol. 2, NY:
McGraw Hill, pp. 2313–2315 (2005).
42. H. Mussalo-Rauhamaa, S. Piepponen, J. Lehto, R. Kauppila, and O. Auvinen, Cu, Zn,
Se and Mg concentrations in breast fat of Finnish breast cancer patients and healthy
controls, Trace Elements Med. 10, 13–15 (1993).
15 Neutron Stimulated Emission Computed Tomography 285

43. K. Geraki, M.J. Farquharson, and D.A. Bradley, X-ray fluorescence and energy disper-
sive x-ray diffraction for the quantification of elemental concentrations in breast tissue,
Phys. Med. Biol. 49, 99–110 (Jan 7 2004).
44. U. Majewska, D. Banas, J. Braziewicz, S. Gozdz, A. Kubala-Kukus, and M. Kuchar-
zewski, Trace element concentration distributions in breast, lung and colon tissues, Phys.
Med. Biol. 52, 3895–911 (Jul 7 2007).
45. K. Geraki, M.J. Farquharson, and D.A. Bradley, Concentrations of Fe, Cu and Zn in
breast tissue: a synchrotron XRF study, Phys. Med. Biol. 47, 2327–39 (Jul 7 2002).
46. S.L. Rizk and H.H. Sky-Peck, Comparison between concentrations of trace elements in
normal and neoplastic human breast tissue, Cancer. Res. 44, 5390–4 (Nov 1984).
47. S. Rizk and H. Sky-Peck, Comparison between Concetrations of Trace Elements in
Normal and Neoplastic Human Breast Tissue, Cancer. Research. 44, 5390–539 (1984).
48. P.M. Santoliquido, H.W. Southwick, and J.H. Olwin, Trace metal levels in cancer of the
breast, Surg. Gynecol. Obstet. 142, 65–70 (Jan 1976).
49. M. Brys, A.D. Nawrocka, E. Miekos, C. Zydek, M. Foksinski, A. Barecki, and W.M.
Krajewska, Zinc and cadmium analysis in human prostate neoplasms, Biol. Trace. Elem.
Res. 59, 145–52 (Winter 1997).
50. A. Feustel, R. Wennrich, D. Steiniger, and P. Klauss, Zinc and cadmium concentration in
prostatic carcinoma of different histological grading in comparison to normal prostate
tissue and adenofibromyomatosis (BPH), Urol. Res. 10, 301–3 (1982).
51. V. Zaichick, T.V. Sviridova, and S.V. Zaichick, Zinc in the human prostate gland:
normal, hyperplastic and cancerous, Int. Urol. Nephrol. 29, 565–74 (1997).
52. E. Andrasi, M. Suhajda, I. Saray, L. Bezur, L. Ernyei, and A. Reffy, Concentration of
elements in human brain: glioblastoma multiforme, Sci. Total. Environ. 139/140, 399–402
(Nov 1 1993).
53. M. Persigehl, H. Schicha, K. Kasperek, and H.J. Klein, Trace element concentration in
human organs in dependence of age, Beitr. Path. 161, 209–220 (1977).
54. E.J. Margalioth, J.G. Schenker, and M. Chevion, Copper and zinc levels in normal and
malignant tissues, Cancer 52, 868–72 (Sep 1 1983).
55. P. Ghadirian, P. Maisonneuve, C. Perret, G. Kennedy, P. Boyle, D. Krewski, and A.
Lacroix, A case-control study of toenail selenium and cancer of the breast, colon, and
prostate, Cancer. Detect. Prev. 24, 305–13 (2000).
56. I. Kato, A.M. Dnistrian, M. Schwartz, P. Toniolo, K. Koenig, R.E. Shore, A. Zele-
niuch-Jacquotte, A. Akhmedkhanov, and E. Riboli, Iron intake, body iron stores and
colorectal cancer risk in women: a nested case-control study, Int. J. Cancer. 80, 693–8
(Mar 1 1999).
57. R.L. Nelson, F.G. Davis, E. Sutter, L.H. Sobin, J.W. Kikendall, and P. Bowen, Body iron
stores and risk of colonic neoplasia, J. Natl. Cancer. Inst. 86, 455–60 (Mar 16 1994).
58. K. Witkowski, A. Kozlowski, M. Pardela, J. Piecuch, and P. Walichiewicz, [Level of
copper in plasma and tissue of patients with esophageal and large bowel cancer], Wiad.
Lek. 46, 586–8 (Aug 1993).
59. K.Q. Xiao and W.J. Henderson, [Electron microscopy microanalysis and quantitative
detection of trace elements in carcinoma of the colon], Zhonghua Bing Li Xue Za Zhi 21,
142–5 (Jun 1992).
60. W.P. Banner, J.J. DeCosse, Q.H. Tan, and M.S. Zedeck, Selective distribution of sele-
nium in colon parallels its antitumor activity, Carcinogenesis 5, 1543–6 (Dec 1984).
61. G.C. Gregoriadis, N.S. Apostolidis, A.N. Romanos, and T.P. Paradellis, A comparative
study of trace elements in normal and cancerous colorectal tissues, Cancer 52, 508–19
(Aug 1 1983).
62. Z.M. Bataineh, I.H. Bani Hani, and J.R. Al-Alami, Zinc in normal and pathological
human prostate gland, Saudi. Med. J. 23, 218–20 (Feb 2002).
63. A. Feustel, R. Wennrich, and H. Dittrich, Investigations of trace elements in metastases
and primary carcinoma of the prostate, Urol. Res. 17, 107–9 (1989).
286 A.J. Kapadia

64. J.O. Ogunlewe and D.N. Osegbe, Zinc and cadmium concentrations in indigenous blacks
with normal, hypertrophic, and malignant prostate, Cancer 63, 1388–92 (Apr 1 1989).
65. I. Romics and L. Katchalova, Spectrographic determination of zinc in the tissues of
adenoma and carcinoma of the prostate, Int. Urol. Nephrol. 15, 171–6 (1983).
66. M. Yaman, D. Atici, S. Bakirdere, and I. Akdeniz, Comparison of trace metal concen-
trations in malign and benign human prostate, J. Med. Chem. 48, 630–634 (2005).
67. F.K. Habib, G.L. Hammond, I.R. Lee, J.B. Dawson, M.K. Mason, P.H. Smith, S.R.Y.K.
Stitch, A.G. Meade, E.P. Rack, and A.J. Blotcky, Metal-androgen interrelationships in
carcinoma and hyperplasia of the human prostate, J. Endocrinol., 61 (1), 133–41 (1976).
68. H. Kubo, S. Hashimoto, and A. Ishibashi, Simultaneous determinations of Fe, Cu, Zn,
and Br concentrations in human tissue sections, Med. Phys. 3, 204–9 (Jul-Aug 1976).
69. G.M. Brittenham and D.G. Badman, Noninvasive measurement of iron: report of an
NIDDK workshop, Blood 101, 15–9 (Jan 1 2003).
70. J.M. Alustiza, J. Artetxe, A. Castiella, C. Agirre, J.I. Emparanza, P. Otazua, M. Garcia-
Bengoechea, J. Barrio, F. Mujica, and J.A. Recondo, MR quantification of hepatic iron
concentration, Radiology 230, 479–84 (Feb 2004).
71. W.F. Avrin and S. Kumar, Noninvasive liver-iron measurements with a room-tempera-
ture susceptometer, Physiol. Meas. 28, 349–61 (Apr 2007).
72. H.L. Bonkovsky, R.B. Rubin, E.E. Cable, A. Davidoff, T.H. Rijcken, and D.D. Stark,
Hepatic iron concentration: noninvasive estimation by means of MR imaging techniques,
Radiology 212, 227–34 (Jul 1999).
73. G.M. Brittenham, S. Sheth, C.J. Allen, and D.E. Farrell, Noninvasive methods for
quantitative assessment of transfusional iron overload in sickle cell disease, Semin.
Hematol. 38, 37–56 (Jan 2001).
74. E. Cecchin, S. De Marchi, F. Querin, M.G. Marin, R. Fiorentino, and F. Tesio, Efficacy
of hepatic computed tomography to detect iron overload in chronic hemodialysis, Kidney.
Int. 37, 943–50 (Mar 1990).
75. R.W. Chapman, G. Williams, G. Bydder, R. Dick, S. Sherlock, and L. Kreel, Computed
tomography for determining liver iron content in primary haemochromatosis, Br. Med. J.
280, 440–2 (Feb 16 1980).
76. J.L. Chezmar, R.C. Nelson, J.A. Malko, and M.E. Bernardino, Hepatic iron overload:
diagnosis and quantification by noninvasive imaging, Gastrointest. Radiol. 15, 27–31
(Winter 1990).
77. R.M. Dixon, P. Styles, F.N. al-Refaie, G.J. Kemp, S.M. Donohue, B. Wonke, A.V.
Hoffbrand, G.K. Radda, and B. Rajagopalan, Assessment of hepatic iron overload in
thalassemic patients by magnetic resonance spectroscopy, Hepatology 19, 904–10 (Apr
1994).
78. Y. Gandon, D. Olivie, D. Guyader, C. Aube, F. Oberti, V. Sebille, and Y. Deugnier, Non-
invasive assessment of hepatic iron stores by MRI., Lancet 363, 357–362 (Jan 31, 2004 2004).
79. D. Guyader, Y. Gandon, J.Y. Robert, J.F. Heautot, H. Jouanolle, C. Jacquelinet, M.
Messner, Y. Deugnier, and P. Brissot, Magnetic resonance imaging and assessment of
liver iron content in genetic hemochromatosis, J Hepatol 15, 304–8 (Jul 1992).
80. J.M. Howard, C.N. Ghent, L.S. Carey, P.R. Flanagan, and L.S. Valberg, Diagnostic
efficacy of hepatic computed tomography in the detection of body iron overload, Gastro-
enterology 84, 209–15 (Feb 1983).
81. P. Liu, M. Henkelman, J. Joshi, P. Hardy, J. Butany, M. Iwanochko, M. Clauberg, M.
Dhar, D. Mai, S. Waien, and N. Olivieri, Quantification of cardiac and tissue iron by
nuclear magnetic resonance relaxometry in a nvel murine thalassemia-cardiac iron over-
load model, Can. J. Cardiol. 12, 155–64 (Feb 1996).
82. P. Nielsen, R. Engelhardt, M. Duerken, G.E. Janka, and R. Fischer, Using SQUID
biomagnetic liver susceptometry in the treatment of thalassemia and other iron loading
diseases, Transfus. Sci. 23, 257–8 (Dec 2000).
15 Neutron Stimulated Emission Computed Tomography 287

83. P. Nielsen, R. Engelhardt, J. Dullmann, and R. Fischer, Non-invasive liver iron quanti-
fication by SQUID-biosusceptometry and serum ferritin iron as new diagnostic para-
meters in hereditary hemochromatosis, Blood Cells Mol. Dis. 29, 451–8 (Nov-Dec 2002).
84. P. Nielsen, U. Kordes, R. Fischer, R. Engelhardt, and G.E. Janka, [SQUID-biosuscep-
tometry in iron overloaded patients with hematologic diseases], Klin Padiatr 214, 218–22
(Jul-Aug 2002).
85. H. Perrimond, C. Chagnon, I. Moulanier, G. Michel, H. Guidicelli, and P.J. Bernard, The
value of nuclear magnetic resonance in the study of iron overload in thalassemia patients,
Ann. Pediatr. (Paris) 38, 175–84 (Mar 1991).
86. S. Sheth, SQUID biosusceptometry in the measurement of hepatic iron, Pediatr. Radiol.
33, 373–7 (Jun 2003).
87. D. Vartsky, K.J. Ellis, D.M. Hull, and S.H. Cohn, Nuclear resonant scattering of gamma
rays – a new technique for in vivo measurement of body iron stores, Phys. Med. Biol. 24,
689–701 (Jul 1979).
88. D. Vartsky, L. Wielopolski, K.J. Ellis, and S.H. Cohn, The Use of Nuclear Resonant
Scattering of Gamma-Rays for Invivo Measurement of Iron, Nucl. Instrum. Meth. Phys.
Res. 193, 359–364 (1982).
89. E. Angelucci, A. Giovagnoni, G. Valeri, E. Paci, M. Ripalti, P. Muretto, C. McLaren, G.
M. Brittenham, and G. Lucarelli, Limitations of magnetic resonance imaging in measure-
ment of hepatic iron, Blood 90, 4736–42 (Dec 15 1997).
90. Z.J. Wang, J.C. Haselgrove, M.B. Martin, A.M. Hubbard, S. Li, K. Loomes, J.R. Moore,
H. Zhao, and A.R. Cohen, Evaluation of iron overload by single voxel MRS measure-
ment of liver T2, J. Magn. Reson. Imaging 15, 395–400 (Apr 2002).
91. A.J. Kapadia, A.C. Sharma, G.D. Tourassi, J.E. Bender, A.S. Crowell, M.R. Kiser, C.R.
Howell, and C.E. Floyd, Neutron Spectroscopy of Mouse Using Neutron Stimulated
Emission Computed Tomography (NSECT), in IEEE Nuclear Science Symposium, Med-
ical Imaging Conference, San Diego, CA, pp. 3546–3548 (2006).
92. A.C. Sharma, G.D. Tourassi, A.J. Kapadia, A.S. Crowell, M.R. Kiser, A. Hutcheson, B.
P. Harrawood, C.R. Howell, and C.E. Floyd, Elemental Spectrum of a Mouse Obtained
via Neutron Stimulation., in 2007 SPIE Symposium on Medical Imaging, San Diego, CA,
p. 65100 K (2007).
93. C.E. Floyd, A.C. Sharma, J.E. Bender, A.J. Kapadia, J.Q. Xia, B.P. Harrawood, G.D.
Tourassi, J.Y. Lo, M.R. Kiser, A.S. Crowell, R.S. Pedroni, R.A. Macri, S. Tajima, and C.
R. Howell, Neutron Stimulated Emission Computed Tomography: Background Correc-
tions, Nucl. Instrum. Meth. Phys. Res. Sect. B 254, 329–336 (2007).
94. C.E. Floyd, A.J. Kapadia, J.E. Bender, A.C. Sharma, J.Q. Xia, B.P. Harrawood, G.D.
Tourassi, J.Y. Lo, A.S. Crowell, M.R. Kiser, and C.R. Howell, Neutron Stimulated
Emission Computed Tomography of a Multi-Element Phantom, Phys. Med. Biol. 53,
2313–2326 (2008).
95. K. Lange and R. Carson, EM reconstruction Algorithms for Emission and Trans-
mission Tomography, Journal of Computer Assisted Tomography 8, 306–316 (1984).
96. A.J. Kapadia, Accuracy and Patient Dose in Neutron Stimulated Emission Computed
Tomography for Diagnosis of Liver Iron Overload: Simulations in GEANT4, in Biome-
dical Engineering. vol. PhD, Duke University, Durham, NC (2007).
97. W.A. Watson, III, J. Chen, G. Heyes, E. Jastrzembski, and D. Quarrie, CODA: a
scalable, distributed data acquisition system, in IEEE Transactions on Nuclear Science,
Vancouver, BC, Canada, pp. 61–68 (1994).
98. R. Fox, C. Bolen, K. Orji, and J. Venema, NSCLSpecTcl Meeting the Needs of Pre-
liminary Nuclear Physics Data Analysis in 11’th Annual Tcl/Tk Conference New Orleans,
Louisiana (2004).
99. RSNA, Radiation Exposure in X-ray Examinations, in American College of Radiology
(ACR) and the Radiological Society of North America (RSNA) (2007).
288 A.J. Kapadia

100. A.C. Sharma, B.P. Harrawood, J.E. Bender, G.D. Tourassi, and A.J. Kapadia, Neutron
stimulated emission computed tomography: a Monte Carlo simulation approach, Phys.
Med. Biol. 52, 6117–31 (Oct 21 2007).
101. J.E. Bender, A.J. Kapadia, A.C. Sharma, G.D. Tourassi, B.P. Harrawood, and C.
E. Floyd, Breast cancer detection using Neutron Stimulated Emission Computed
Tomography: prominent elements and dose requirements, Med. Phys. 34,
3866–3871 (2007).
102. A.J. Kapadia, C.E. Floyd, C.R. Howell, and B.P. Harrawood, Sampling Requirements
for Neutron Stimulated Emission Computed Tomography, in RSNA, Physics (Digital
Imaging, PACS) session Chicago, IL (2004).
Chapter 16
Visualizing Structures of Biological
Macromolecules Through Indirect Imaging
with Small-Angle Neutron Scattering
and Modeling

W.T. Heller and G.A. Baker

Abstract The molecular machinery of living organisms is built of systems

whose length scales range from nanometers to microns. While neutron imaging
is a powerful tool for studying macroscopic systems, neutron optics is not
capable of effectively resolving these macromolecular systems. Small-angle
neutron scattering when combined with modeling and contrast variation meth-
ods enables indirect imaging for visualization of structure and function over
these length scales. A brief overview of small-angle neutron scattering with
contrast variation is presented with a description of available modeling meth-
ods. Applications of the methods are also presented to provide a survey of the
wide variety of systems that are suitable for study.

Keywords Small-angle neutron scattering  Structural biology  Contrast

variation  Modeling

16.1 Introduction

Neutron imaging has great potential in the biological sciences because it is a

sensitive probe of hydrogen. Unfortunately, the lack of effective optics, such as
those available for light and electron microscopes, limits the application of the
technique to macroscopic systems, as opposed to the study of molecular systems
where truly unique information can be obtained. While it is possible to focus
neutrons, the focal lengths of neutron lenses are very long because neutrons
do not interact strongly with matter. Therefore, it is difficult to refocus the
neutrons to resolve the image of the object being studied as one does in a
conventional microscope. Still, the unique properties of the neutron make it a
very attractive probe of biological materials over a wide range of length scales.

W.T. Heller (*)

Center for Structural Molecular Biology Oak Ridge National Laboratory
P.O. Box 2008, Oak Ridge, TN 37931, USA
e-mail: [email protected]

I.S. Anderson et al. (eds.), Neutron Imaging and Applications, 289

Neutron Scattering Applications and Techniques, DOI 10.1007/978-0-387-78693-3_16,
Ó Springer ScienceþBusiness Media, LLC 2009
290 W.T. Heller and G.A. Baker

Two closely related alternative approaches to extract structural information

from macromolecular systems involve small-angle scattering (SAS) of either
neutrons [small-angle neutron scattering (SANS)] or X-rays [small-angle X-ray
scattering (SAXS)]. Small-angle scattering probes distances ranging from 1 to
100 nm, giving it a potential maximum resolution comparable to transmission
electron microscopy. When working with dilute, homogeneous solutions, SAS
becomes a powerful tool for indirectly imaging the structures and functions
of proteins, nucleic acids, and their complexes, particularly when combined
with computational methods for constructing models. Advances in modeling
methods have generated considerable interest among researchers in the field
of biological sciences who seek to understand complex biological systems,
particularly those that have proven resistant to high-resolution structure deter-
mination by other means.

16.1.1 Theory

Small-angle scattering of dilute solutions of identical particles results from the

interference of neutrons that interact with the nuclei of the atoms in the
particles, much like a diffraction measurement. Unlike diffraction, however,
the samples contain no crystalline ordering that produces well-defined intensity
peaks at locations determined by the crystal lattice. Neutrons, which can be
described as a propagating plane wave, are incident on the sample and scattered
as shown in Fig. 16.1.
The scattered neutrons are then collected at an angle 2 from the incident
beam. The resulting pattern is due to the interference of neutrons, scattered
from different locations, within the particle and therefore is a function of the
particle shape. The scattering is described as a function of the momentum
~ which has a magnitude Q = 4psin()/l. The neutron wavelength,
transfer, Q,
l, is h/mv, where h is Plank’s constant, m is the mass of the neutron and v is the
neutron velocity. The neutrons commonly used for scattering experiments
have wavelengths ranging from 2 to 20 Å, which corresponds to velocities

Fig. 16.1 Schematic of small-angle scattering. Neutrons as plane waves are incident on a
sample particle and are scattered. The pattern observed results from the interference of
neutrons, scattered from different positions, within the particle
16 Visualizing Structures of Biological Macromolecules 291

ranging from a few hundred to a few thousand meters per second. For an
isotropic sample, such as for a protein tumbling freely in solution, the pattern
of scattered neutrons is symmetric about the beam and can be described
according to Eq. (16.1) [1, 2].

*Z 2 +

    
~R ~ 3 
IðQÞ ¼ n  ~
 R  s e iQ
d R ; (16:1)
 
V

where I(Q) is the scattered intensity, n is the number of particles per unit
volume,  R ~ is the scattering length density of the particle at position R, ~
and s is the scattering length density of the solvent. The integral is taken over
the volume of the particle V [1, 2]. The integral is averaged over time, all
orientations, and the ensemble of structures present in the solution.

16.1.2 Contrast Variation

The fact that Eq. 16.1 describes the scattering signal relative to the background
solvent is incredibly important because it enables the use of SANS with contrast
variation [3]. The dramatic difference between the scattering lengths of hydro-
gen (–3.71 fm) and deuterium (6.67 fm) makes it possible to greatly alter the
scattering length density of a biological macromolecule or the aqueous back-
ground by substituting for hydrogenated material ones that are isotopically
labeled with deuterium. Contrast variation is a particularly powerful technique
for studying multisubunit complexes in which one subunit is labeled with
deuterium because it enables the separation of the scattering from the two
components within the complex. In a contrast variation experiment, a complex
containing a subunit with a different average scattering length density, such as a
deuterium-labeled protein or nucleic acid, is measured in a series of solutions
consisting of mixtures of H2O and D2O. Conceptually, the effect of changing the
H2O/D2O mixture in the solution is illustrated in Fig. 16.2. The intensity profiles
in the contrast variation series can be written in the following manner (Eq. 16.2):

IðQ;
1 ;
2 Þ ¼
21 I1 ðQÞ þ
22 I2 ðQÞ þ
1
2 I12 ðQÞ (16:2)

Here,
1 and
2 are the differences between the hydrogenated and deu-
terated components of the complex relative to the solvent, respectively, and are
often called the contrasts. The three functions, I1(Q), I2(Q), and I12(Q) are
known as the basic scattering functions [3]. The basic scattering functions
are the scattering signals from the components within the complex having two
different scattering length densities and the cross term that provides informa-
tion on the relative disposition of the components. The measured contrast
variation series data define a set of linear equations that can be used to extract
the basic scattering functions for further analysis.
292 W.T. Heller and G.A. Baker

Fig. 16.2 Contrast variation. The scattering length density of the background solvent is varied
relative to the two subunits having different scattering length densities, thereby differentially
highlighting the structures of the subunits within the whole

16.1.3 Information Content and Basic Data Analysis

There are two common, basic methods of SAS data analysis. For a dilute,
monodisperse solution of homogenous particles, the intensity I(Q) can be fit
for the radius of gyration (Rg) according to Guinier [2]:
 2
QRg
ln½IðQÞ ¼ ln½Ið0Þ   (16:3)
3
Here, I(0) is the scattered intensity at Q = 0, which is a shape-independent
function of the molecular weight of the particle and the average scattering
length density relative to the background solution. The radius of gyration, Rg,
is a shape-dependent function of the contrast-weighted distance from the center
of a particle to scattering elements within the particle. While Guinier analysis
may be used for any particle of unknown size or shape, there are instances when
it is not valid. To use Guinier analysis, the following three conditions must be
satisfied: (1) the system must be rotationally isotropic, (2) the particles must not
interact with each other, and (3) Q  Rg must be less than 1.3, a condition
resulting from the approximation used to derive Eq. (16.3) for globular parti-
cles. Condition (3) must be further restricted for other shapes, with Q  Rg less
than 1.0 being commonly applied.
Additional analysis can be performed to obtain more information on the
shape of the scattering particle. Equation (16.1) can be inverted to provide the
distance distribution function P(R) as a function of the experimental I(Q) using
the Fourier transform in Eq. (16.4).
Z1
1
PðRÞ ¼ 2 IðQÞQR sinðQRÞdQ : (16:4)
2p
0
16 Visualizing Structures of Biological Macromolecules 293

The actual approach uses inverse Fourier methods and includes the Glatter
algorithm [4, 5], the Moore algorithm [6], and the method implemented in the
software GNOM [7]. Fitting SAS data for P(R) provides a measure of dmax, the
maximum linear dimension of the particle.
An example SANS intensity profile calculated from the crystal structure of
hemoglobin (PDB ID: 4HHB [8]) using the software ORNL_SAS [9] is shown
in Fig. 16.3. As can be seen, the data have a relatively simple appearance that
suggests they do not contain a great deal of information, in contrast to high-
resolution crystallography data. Researchers have constructed arguments
based on the Shannon Sampling Theorem [10] to estimate the appropriate
number of degrees of freedom, nf, suitable for fitting SAS data [6, 11–13]. The
number is a function of the Q-range of the data collected and the size of the
scattering particle. If data are collected for a particle of size dmax over a range
that extends from Qmin to Qmax, then nf  dmax(Qmax-Qmin)/ p.

Fig. 16.3 Small-angle

scattering profile calculated
from the atomic-resolution
structure of hemoglobin
(PDB ID: 4HHB [8]) using
the program ORNL_SAS
[14]. The intensity is
displayed in arbitrary units

16.2 Modeling for Visualization

Structural biology is a very visual science, but the basic analysis of SAS data
does not provide a structure that can be used to understand the interplay
between structure and function in complex macromolecular systems. Hence,
methods have been developed for constructing models from SAS data. Much
of the effort has focused on SAXS data, due to the greater availability of
instruments. Still, tools for building models from SANS data have been
developed that can be applied to contrast variation data of multisubunit
complexes. The kind of modeling possible is often a function of the amount
of existing structural information, particularly high-resolution structural
information.
294 W.T. Heller and G.A. Baker

16.2.1 Shape Restoration

In cases where no structural information exists for the protein or complex being
studied, several methods exist for building structural models that are referred
to as shape restoration methods. These techniques encompass a broad range of
approaches spanning simple shapes, spherical harmonics, and methods using
aggregates of shapes. Each approach has its strengths and weaknesses. A large
number of these methods are freely available to the scientific community,
making them attractive choices for building models from SAS data.
Some of the most established modeling methods for SAS data on proteins
and protein complexes use simple shapes. Model fitting using these approaches
employs a limited number of degrees of freedom, providing a degree of con-
fidence that the answer is uniquely determined by the data. Modeling using a
simple shape has the further advantage that the associated intensity profile is
quite often an analytical function [15], making it attractive for fitting data by
hand. More importantly, many of these geometrical shapes are implemented in
freely available software packages [16, 17]. For modeling using multiple shapes,
the only analytical function available is the Debye formula for aggregates of
spheres [18]. Other multishape structures lack analytical functions for the
scattering curves, although methods have been developed for building models
in this manner [19–27]. Such approaches are very well suited to modeling
contrast variation series of SANS data from biological macromolecular com-
plexes, where each subunit can be assigned to a single shape.
Shape reconstruction methods using spherical harmonics were among the
first ab initio methods developed for extracting structural information from
SAS data [11, 28–34]. In this approach, the structure is defined by a series
superposition of spherical harmonics, an infinite set of orthogonal functions. It
is trivial to limit the number of terms used in the reconstruction, which is
typically assigned to nf. The ability to readily restrict the number of free
parameters in the fit provides a measure of confidence that the structure
found is reasonable. Still, the resulting structures are restricted to those which
can be represented by spherical harmonics, and significantly more complicated
structures cannot be effectively modeled. The approach is readily applied to
SAXS data, but it is also applicable to contrast variation experiments. These
shape restoration methods can also be applied directly to basic scattering
functions extracted from such data sets.
Currently, the most popular methods for shape restoration from SAS data
are the aggregate ab initio techniques that use large sets of simple, relatively
small shapes to define the volume of the scattering particles. Spheres are the
most common base shape for aggregate techniques [35–38], but ellipsoids have
also been used [14]. The programs DAMMIN [39] and GASBOR [40] use
dummy atoms and dummy residues, respectively, for the shape reconstructions
and are the most popular software packages currently available for this pur-
pose. A related grid-based method has also been developed [41–43]. In general,
16 Visualizing Structures of Biological Macromolecules 295

any shape can be reproduced using these methods up to the limits of the
resolution of the data and the base shape used to build the structure, in principle
avoiding the limitations inherent in using spherical harmonics to define shapes.
Unfortunately, the large number of degrees of freedom inherent in such model-
ing makes degenerate solutions possible that fit the SAS data equally well. For
this reason, multiple independent structures are often generated and averaged
into a consensus shape [35–37, 44]. An implementation of DAMMIN exists that
is capable of directly modeling against contrast variation series of SANS data
from complexes with selective deuterium labeling [39]. The other techniques can
be applied to SAXS data, SAS data from structures that are not selectively
deuterated, or to basic scattering functions extracted from contrast variation
series data.

16.2.2 Modeling with High-Resolution Structures

The dramatic increase in the number of high-resolution structures deposited in

the Protein Data Bank [45] has led to the development of methods that
can leverage this data to build models from SAS data. The basic idea of
comparing a measured SAS profile to the SAS intensity simulated from an
atomic-resolution structure is not new, and several methods for doing so have
been described throughout the literature [9, 28, 46–56]. Determining whether
the solution structure of a protein is consistent with an available high-resolution
structure is a form of modeling that allows one to visualize the structure. If the
SAS data are inconsistent with the existing high-resolution structure, then
additional modeling methods must be applied to understand the nature of
these differences. For example, it is possible to understand the nature of the
conformational changes by using the existing high-resolution structure as a
starting point.
There are a variety of approaches that can use high-resolution structures
as starting points. Rigid body modeling methods without advanced docking
algorithms are the simplest, relying only on the SAS data to build models [21,
54, 57–60]. These relatively basic rigid body methods can also be applied to
model building using low-resolution structures produced employing the
ab initio methods described above. More recently, increased interest in the use
of high-resolution structures for rigid body modeling of complexes has led to
the development of methods that use more formal docking algorithms and
potentials to produce more realistic models [61–64]. Mixed model building in
which an ab initio density is added to an existing high-resolution structure to
simulate missing mass is also a very useful tool [65].
It is important to point out that SAS modeling methods can be developed to
address specific problems, and the amount of other structural information that
is available truly dictates the best choice for the problem at hand. Time,
creativity, and programming expertise are the only limiting factors.
296 W.T. Heller and G.A. Baker

16.3 Examples

Broader reviews of the applications of SAS and modeling to the study of

protein structures have been published recently [66–67]. Rather than trying
to provide a comprehensive listing of examples where SAS and modeling
have been applied to visualize the structure of biological macromolecules
in solution, some specific examples demonstrating the use of SANS with
contrast variation are presented. Such studies are excellent showcases of
the various ways in which models are built from scattering data and of the
power of neutron scattering when combined with specific isotope labeling
and contrast variation methods.
The structures of various ribosomes have been studied on multiple occa-
sions using SANS with contrast variation, and these studies represent some of
the earliest uses of modeling to visualize macromolecular complexes. For
example, neutron scattering studies have been used to map out the relative
distances between various proteins within the 30S ribosome complex from
E. coli [68–70]. The culmination of these extensive studies was a low-resolu-
tion model showing the relative spatial relationships of the protein elements
within the ribosome [71]. More advanced modeling methods were later
applied to contrast variation experiments on larger ribosomal complexes.
Small-angle neutron scattering with contrast variation experiments with the
E. coli 50S ribosome, a larger complex than the 30S ribosome, resulted in a
map of the distribution of the protein and RNA within the complex through
the use of a two-phase model built using a spherical harmonic shape restora-
tion approach [31, 32, 72]. A series of papers also investigated the even larger
70S ribosome from E. coli using the combined approach of SANS with
contrast variation and selective deuteration. A total of 42 SANS data sets
were collected and modeled with a four-phase spherical harmonic shape
restoration approach [73, 74]. Later, this same data set was subjected to ab
initio modeling using DAMMIN [75], providing a much more detailed struc-
ture of the complex macromolecular machine.
Other protein complexes have been subjected to extensive analysis using
SANS with contrast variation and modeling. The activation of myosin
light chain kinase by calmodulin was studied using selectively deuterated
complexes [76, 77]. Simple shape modeling using ellipsoids was employed
to visualize the structure and the conformational transitions that occur
upon calcium binding to calmodulin. The interpretation of these results
and subsequent X-ray scattering studies [78, 79] was further enhanced by
docking high-resolution structures for the two proteins that showed the
location of the calmodulin relative to the catalytic cleft of the myosin light
chain kinase. Changes in the dimensions of the ellipsoids were used to
infer that the cleft of the kinase opened in response to the binding of
calmodulin.
16 Visualizing Structures of Biological Macromolecules 297

The skeletal and cardiac variants of troponin complexes have also been
studied by SANS with contrast variation. A variety of modeling approaches
have been used to visualize the structure of the complex, including mixing
high- and low-resolution structural information [80, 81], simple shapes [24],
ab initio modeling of the basic scattering functions followed by rigid body
modeling using the shapes [37, 58], as shown in Fig. 16.4, and investigating
the function of the complex from an available high-resolution structure [82].
Protein kinase A is another dynamic multisubunit complex for which SANS
plus contrast variation and modeling was used to obtain structural and function
information [26, 83], as shown in Fig. 16.5. Protein–nucleic acid complexes
benefit from the inherently different scattering length densities of hydrogenated
protein and nucleic acids, which was leveraged in a study of the assembly of the
methionine repressor protein, MetJ, on different lengths of DNA taken from
regulatory sequences of different genes within the met regulon [84], which is
shown in Fig. 16.6. By no means do these examples comprise an exhaustive
listing of the multisubunit complexes that have been studied to great benefit
by SANS with contrast variation, as other systems have also been studied
recently [26, 85–90], a trend certainly to continue as interest in the method
grows.

Fig. 16.4 Models of the ternary cardiac troponin complex (TnC–TnI–TnT(198–298) derived
from SANS with contrast variation using two different isotopic labeling patterns [91]. In the
left model, the TnC was deuterated (large dark spheres), while the TnI subunit was labeled
(small light spheres) for the contrast variation series used to derive the model on the right. The
density near the middle of the structure was inferred to be the TnT(198–298). The longest
dimension of the structure is 125 Å
298 W.T. Heller and G.A. Baker

Fig. 16.5 Model of the protein kinase A structure derived using SANS with contrast variation
showing the catalytic domains (gray), cyclic AMP binding domains (light gray), the dimeriza-
tion-docking domain (dark gray), and unstructured linker domains (black) [83]. The longest
dimension of the structure is 155 Å

Fig. 16.6 Models of MetJ–DNA complexes using different lengths of DNA. MetJ dimers
along the length of the DNA (black spheres) are colored light and dark gray in an alternating
pattern to show differences in the length-dependent packing of the protein dimers on the DNA
[84]. The radius from the axis of the DNA of the structure on the left is 40 Å

16.4 Conclusions

Small-angle neutron scattering with contrast variation, when combined with

modeling methods, is a powerful tool for visualizing the structure and internal
organization of multisubunit biological complexes. While it is not a direct
imaging method, it has the advantages of relatively simple sample preparation
16 Visualizing Structures of Biological Macromolecules 299

and very broad applicability to biological systems in near-native, solution

environments. The methods presented here are readily applicable to protein–
nucleic acid complexes, which benefit from an inherent difference in scattering
length between the protein and nucleic acid constituents. Selective deuteration
adds the powerful capability of introducing a protein subunit into a complex
whose scattering can be separated from the remainder of the protein compo-
nents. The continued development of modeling algorithms will further enhance
the accessibility of SANS with contrast variation for nonexpert users and
improve the value of the resulting models for further interpretation of the
functions of the complexes studied.

Acknowledgments This work was supported by the Oak Ridge Center for Structural
Molecular Biology (KP1102010) of the Office of Biological and Environmental Research of
the U. S. Department of Energy under contract No. DE-AC05-00OR22725 with Oak Ridge
National Laboratory, managed and operated by UT-Battelle, LLC. The submitted manu-
script has been authored by a contractor of the U.S. government under Contract DE-AC05-
00OR22725. Accordingly, the U.S. government retains a nonexclusive royalty-free license to
publish or reproduce the published form of this contribution, or allow others to do so, for U.S.
government purposes.

References
1. P. Debye and A. Bueche, Scattering by an Inhomogeneous Solid. J. Appl. Phys.
20, 518–525 (1949).
2. A. Guinier and G. Fournet, Small-angle Scattering of X-rays. New York, Wiley (1955).
3. K. Ibel and H.B. Stuhrmann, Comparison of Neutron and X-Ray-Scattering of Dilute
Myoglobin Solutions. J. Mol. Biol. 93, 255–265 (1975).
4. O. Glatter, New Method for Evaluation of Small-Angle Scattering Data. J. Appl. Crystallogr.
10, 415–421 (1977).
5. O. Glatter, Interpretation of Real-Space Information from Small-Angle Scattering
Experiments. J Appl Crystallogr 12: 166–175 (1979).
6. P.B. Moore, Small-Angle Scattering - Information-Content and Error Analysis. J. Appl.
Crystallogr. 13, 168–175 (1980).
7. D.I. Svergun, Determination of the Regularization Parameter in Indirect-Transform
Methods Using Perceptual Criteria. J. Appl. Crystallogr. 25, 495–503 (1992).
8. G. Fermi, M.F. Perutz, B. Shaanan, and R. Fourme, The Crystal-Structure of Human
Deoxyhemoglobin at 1.74a Resolution. J. Mol. Biol. 175, 159–174 (1984).
9. E. Tjioe and W.T. Heller, ORNL_SAS: software for calculation of small-angle scattering
intensities of proteins and protein complexes. J. Appl. Crystallogr. 40, 782–785 (2007).
10. C.E. Shannon, A Mathematical Theory of Communication. Bell Sys. Tech. J. 27,
379–423, 623–656 (1948).
11. D.I. Svergun, V.V. Volkov, M.B. Kozin, and H.B. Stuhrmann, New developments in
direct shape determination from small-angle scattering .2. Uniqueness. Acta. Crystallogr.
Sect. A 52, 419–426 (1996).
12. D. Taupin and V. Luzzati, Information-Content and Retrieval in Solution Scattering
Studies .1. Degrees of Freedom and Data Reduction. J. Appl. Crystallogr. 15, 289–300
(1982).
13. V.V. Volkov and D.I. Svergun, Uniqueness of ab initio shape determination in small-
angle scattering. J. Appl. Crystallogr. 36, 860–864 (2003).
300 W.T. Heller and G.A. Baker

14. D. Vigil, S.C. Gallagher, J. Trewhella, and A.E. Garcia, Functional dynamics of the
hydrophobic cleft in the N-domain of calmodulin. Biophys. J. 80, 2082–2092 (2001).
15. J.S. Pedersen, Analysis of small-angle scattering data from colloids and polymer solutions:
modeling and least-squares fitting. Adv. Colloid. Interface. Sci. 70, 171–210 (1997).
16. S.R. Kline, Reduction and analysis of SANS and USANS data using IGOR Pro. J. Appl.
Crystallogr. 39, 895–900 (2006).
17. M.V. Petoukhov, P.V. Konarev, A.G. Kikhney, and D.I. Svergun, ATSAS 2.1 – towards
automated and web-supported small-angle scattering data analysis. J. Appl. Crystallogr.
40, S223–S228 (2007).
18. P. Debye, Zerstreuung von Röntgenstrahlen. Annalen der Physik 351, 809–823 (1915).
19. S. Fujiwara, F.J. Kull, E.P. Sablin, D.B. Stone, and R.A. Mendelson, The Shapes of the
Motor Domains of 2 Oppositely Directed Microtubule Motors, Nod and Kinesin – a
Neutron-Scattering Study. Biophys. J. 69, 1563–1568 (1995).
20. S. Hansen Calculation of Small-Angle Scattering Profiles Using Monte-Carlo Simula-
tion. J Appl Crystallogr 23, 344–346 (1990).
21. S.P. Harris, W.T. Heller, M.L. Greaser, R.L. Moss, and J. Trewhella, Solution structure
of heavy meromyosin by small-angle scattering. J. Biol. Chem. 278, 6034–6040 (2003).
22. H. Hartmann, A. Bongers, and H. Decker, Small-angle neutron scattering reveals an
oxygen-dependent conformational change of the immunogen keyhole limpet hemocyanin
type 1 (KLH1). Eur. Biophys. J. Biophys. Lett. 30, 471–475 (2001).
23. W.T. Heller, ELLSTAT: shape modeling for solution small-angle scattering of proteins
and protein complexes with automated statistical characterization. J. Appl. Crystallogr.
39, 671–675 (2006).
24. S.J. Henderson, Monte Carlo modeling of small-angle scattering data from non-
interacting homogeneous and heterogeneous particles in solution. Biophys. J. 70,
1618–1627 (1996).
25. D.B. Stone, P.A. Timmins, D.K. Schneider, I. Krylova, C.H.I. Ramos, F.C. Reinach, and
R.A. Mendelson, The effect of regulatory Ca2+ on the in situ structures of troponin C
and troponin I: A neutron scattering study. J. Mol. Biol. 281, 689–704 (1998).
26. J.K. Zhao, E. Hoye, S. Boylan, D.A. Walsh, and J. Trewhella, Quaternary structures
of a catalytic subunit-regulatory subunit dimeric complex and the holoenzyme of the
cAMP-dependent protein kinase by neutron contrast variation. J. Biol. Chem. 273,
30448–30459 (1998).
27. J. Zhou, A. Deyhim, S. Krueger, and S.K. Gregurick, LORES: Low resolution shape
program for the calculation of small angle scattering profiles for biological macromole-
cules in solution. Comput. Phys. Commun. 170, 186–204 (2005).
28. J.G. Grossmann, Z.H.L. Abraham, E.T. Adman, M. Neu, R.R. Eady, B.E. Smith, and
S.S. Hasnain, X-Ray-Scattering Using Synchrotron-Radiation Shows Nitrite Reductase
from Achromobacter-Xylosoxidans to Be a Trimer in Solution. Biochemistry 32,
7360–7366 (1993).
29. F. Spinozzi, F. Carsughi, and P. Mariani, Particle shape reconstruction by small-angle
scattering: Integration of group theory and maximum entropy to multipole expansion
method. J. Chem. Phys. 109, 10148–10158 (1998).
30. H.B. Stuhrman, Interpretation of Small-Angle Scattering Functions of Dilute Solutions
and Gases – a Representation of Structures Related to a One-Particle-Scattering
Function. Acta Crystallogr. Sect. A-Crystal Phys. Diff. Theor. Gen. Crystallogr. A 26,
297–306 (1970).
31. D.I. Svergun, M.H.J. Koch, J.S. Pedersen, and I.N. Serdyuk, Structural Model of the 50-S
Subunit of Escherichia-Coli Ribosomes from Solution Scattering .2. Neutron-Scattering
Study. J. Mol. Biol. 240, 78–86 (1994).
32. D.I. Svergun, M.H.J. Koch, and I.N. Serdyuk, Structural Model of the 50-S Subunit of
Escherichia-Coli Ribosomes from Solution Scattering .1. X-Ray Synchrotron-Radiation
Study. J. Mol. Biol. 240, 66–77 (1994).
16 Visualizing Structures of Biological Macromolecules 301

33. D.I. Svergun and H.B. Stuhrmann, New Developments in Direct Shape Determination
from Small-Angle Scattering .1. Theory and Model-Calculations. Acta. Crystallogr. Sect.
A 47, 736–744 (1991).
34. D.I. Svergun, V.V. Volkov, M.B. Kozin, H.B. Stuhrmann, C. Barberato, and M.H.J.
Koch, Shape determination from solution scattering of biopolymers. J. Appl. Crystallogr.
30, 798–802 (1997).
35. P. Chacon, J.F. Diaz, F. Moran, and J.M. Andreu, Reconstruction of protein form with
X-ray solution scattering and a genetic algorithm. J. Mol. Biol. 299, 1289–1302 (2000).
36. P. Chacon, F. Moran, J.F. Diaz, E. Pantos, and J.M. Andreu, Low-resolution structures
of proteins in solution retrieved from X-ray scattering with a genetic algorithm. Biophys. J.
74: 2760–2775 (1998).
37. W.T. Heller, N.L. Finley, W.J. Dong, P. Timmins, H.C. Cheung, P.R. Rosevear, and
J. Trewhella, Small-angle neutron scattering with contrast variation reveals spatial
relationships between the three subunits in the ternary cardiac troponin complex and
the effects of troponin I phosphorylation. Biochemistry 42, 7790–7800 (2003).
38. D. Walther, F.E. Cohen, and S. Doniach, Reconstruction of low-resolution three-dimen-
sional density maps from one-dimensional small-angle X-ray solution scattering data for
biomolecules. J. Appl. Crystallogr. 33, 350–363 (2000).
39. D.I. Svergun, Restoring low resolution structure of biological macromolecules from
solution scattering using simulated annealing. Biophys. J. 76, 2879–2886 (1999).
40. D.I. Svergun, M.V. Petoukhov, and M.H.J. Koch, Determination of domain structure of
proteins from X-ray solution scattering. Biophys. J. 80, 2946–2953 (2001).
41. H. Hartmann, A. Bongers, and H. Decker, Small-angle X-ray scattering-based three-
dimensional reconstruction of the immunogen KLH1 reveals different oxygen-dependent
conformations. J. Biol. Chem. 279, 2841–2845 (2004).
42. H. Hartmann and H. Decker. Small-angle scattering techniques for analyzing conforma-
tional transitions in hemocyanins. Energ. Biol. Macromol. Pt D. 379, 81–106 (2004).
43. H. Hartmann, T. Muller, and H. Decker, Modeling techniques for analysing conforma-
tional transitions in hemocyanins by small-angle scattering of X-rays and neutrons.
Micron 35, 11–13 (2004).
44. M.B. Kozin and D.I. Svergun, Automated matching of high- and low-resolution
structural models. J. Appl. Crystallogr. 34, 33–41 (2001).
45. H.M. Berman, J. Westbrook, Z. Feng, G. Gilliland, T.N. Bhat, H. Weissig, I.N. Shindyalov,
and P.E. Bourne, The Protein Data Bank. Nucl. Acids. Res. 28, 235–242 (2000).
46. B.A. Fedorov, O.B. Ptitsyn, and L.A. Voronin, X-Ray Diffuse Scattering of Globular
Protein Solutions – Consideration of Solvent Influence. FEBS Lett. 28, 188–190 (1972).
47. S.C. Gallagher, A.J. Callaghan, J.K. Zhao, H. Dalton, and J. Trewhella, Global
conformational changes control the reactivity of methane monooxygenase. Biochemistry
38, 6752–6760 (1999).
48. R. Langridge, D.A. Marvin, W.E. Seeds, H.R. Wilson, C.W. Hooper, M.H.F. Wilkins,
and L.D. Hamilton, The Molecular Configuration of Deoxyriboucleic Acid: Molecular
Models and their Fourier Transforms. J. Mol. Biol. 2, 38–64 (1960).
49. F. Merzel and J.C. Smith, SASSIM: a method for calculating small-angle X-ray and
neutron scattering and the associated molecular envelope from explicit-atom models of
solvated proteins. Acta. Crystallogr. Sect. D-Biol. Crystallogr. 58, 242–249 (2002).
50. J. Ninio, M. Yaniv, and V. Luzzati, Comparative Small-Angle X-Ray Scattering Studies
on Unacylated, Acylated and Crosslinked Escherichia-Coli Transfer Rna Val(I). J. Mol.
Biol. 71, 217–229 (1972).
51. M.Y. Pavlov, M.A. Sinev, A.A. Timchenko, and O.B. Ptitsyn, A Study of Apo-Forms
and Holo-Forms of Horse Liver Alcohol-Dehydrogenase in Solution by Diffuse-X-Ray
Scattering. Biopolymers 25, 1385–1397 (1986).
52. C.A. Pickover and D.M. Engelman, On the Interpretation and Prediction of X-Ray-
Scattering Profiles of Biomolecules in Solution. Biopolymers 21, 817–831 (1982).
302 W.T. Heller and G.A. Baker

53. D. Svergun, C. Barberato, and M.H.J. Koch, CRYSOL – A program to evaluate x-ray
solution scattering of biological macromolecules from atomic coordinates. J. Appl.
Crystallogr. 28, 768–773 (1995).
54. D.I. Svergun, I. Aldag, T. Sieck, K. Altendorf, M.H.J. Koch, D.J. Kane, M.B. Kozin, and
G. Gruber, A model of the quaternary structure of the Escherichia coli F-1 ATPase from
X-ray solution scattering and evidence for structural changes in the delta subunit during
ATP hydrolysis. Biophys. J. 75, 2212–2219 (1998).
55. D.I. Svergun, Richard S, M.H.J. Koch, Z. Sayers, S. Kuprin, and G. Zaccai, Protein
hydration in solution: Experimental observation by x-ray and neutron scattering. Proc.
Natl. Acad. Sci. USA 95, 2267–2272 (1998).
56. J. Zhou, S. Krueger, and S.K. Gregurick, A coarse graining approach to determine
nucleic acid structures from small angle neutron scattering profiles in solution. Nucl.
Acids. Res. 33, 6361–6371 (2005).
57. A.W. Ashton, M.K. Boehm, J.R. Gallimore, M.B. Pepys, and S.J. Perkins, Pentameric
and decameric structures in solution of serum amyloid P component by X-ray and
neutron scattering and molecular modelling analyses. J. Mol. Biol. 272, 408–422 (1997).
58. W.T. Heller, E. Abusamhadneh, N. Finley, P.R. Rosevear, and J. Trewhella, The solution
structure of a cardiac troponin C-troponin I-troponin T complex shows a somewhat
compact troponin c interacting with an extended troponin I-Troponin T component.
Biochemistry 41, 15654–15663 (2002).
59. P.V. Konarev, M.V. Petoukhov, and D.I. Svergun. MASSHA – a graphics system
for rigid-body modelling of macromolecular complexes against solution scattering data.
J. Appl. Crystallogr. 34, 527–532 (2001).
60. M.B. Kozin and D.I. Svergun, A software system for rigid-body modelling of solution
scattering data. J. Appl. Crystallogr. 33, 775–777 (2000).
61. F. Gabel, B. Simon, and M. Sattler, A target function for quaternary structural refinement
from small angle scattering and NMR orientational restraints. Eur. Biophys. J. Biophys.
Lett. 35, 313–327 (2006).
62. A. Grishaev, J. Wu, J. Trewhella, and A. Bax, Refinement of multidomain protein
structures by combination of solution small-angle X-ray scattering and NMR data.
J. Am. Chem. Soc. 127, 16621–16628 (2005).
63. M. Nollmann, W.M. Stark, and O. Byron, A global multi-technique approach to study
low-resolution solution structures. J. Appl. Crystallogr. 38, 874–887 (2005).
64. M.V. Petoukhov and D.I. Svergun, Global rigid body modeling of macromolecular
complexes against small-angle scattering data. Biophys. J. 89, 1237–1250 (2005).
65. M.V. Petoukhov, N.A.J. Eady, K.A. Brown and D.I. Svergun, Addition of missing loops and
domains to protein models by X-ray solution scattering. Biophys. J. 83, 3113–3125 (2002).
66. M.H.J. Koch, P. Vachette, and D.I. Svergun, Small-angle scattering: a view on the
properties, structures and structural changes of biological macromolecules in solution.
Q Rev. Biophys. 36, 147–227 (2003).
67. D.I. Svergun and M.H.J. Koch, Small-angle scattering studies of biological macromole-
cules in solution. Rep. Prog. Phys. 66, 1735–1782 (2003).
68. J.A. Langer, D.M. Engelman, and P.B. Moore, Neutron-Scattering Studies of Ribosome
of Escherichia-Coli – a Provisional Map of Locations of Proteins S3, S4, S5, S7, S8 and S9
in 30-S Subunit. J. Mol. Biol. 119, 463–485 (1978).
69. V. Ramakrishnan, M. Capel, M. Kjeldgaard, D.M. Engelman, and P.B. Moore,
Positions of Protein-S14, Protein-S18 and Protein-S20 in the 30-S Ribosomal-Subunit
of Escherichia-Coli. J. Mol. Biol. 174, 265–284 (1984).
70. V.R. Ramakrishnan, S. Yabuki, I.Y. Sillers, D.G. Schindler, D.M. Engelman, and
P.B. Moore, Positions of Proteins S6, S11 and S15 in the 30-S Ribosomal-Subunit of
Escherichia-Coli. J. Mol. Biol. 153, 739–760 (1981).
71. M.S. Capel, D.M. Engelman, B.R. Freeborn, M. Kjeldgaard, J.A. Langer,
V. Ramakrishnan, D.G. Schindler, D.K. Schneider, B.P. Schoenborn, I.Y. Sillers,
16 Visualizing Structures of Biological Macromolecules 303

S. Yabuki, and P.B. Moore, A Complete Mapping of the Proteins in the Small
Ribosomal-Subunit of Escherichia-Coli. Science 238, 1403–1406 (1987).
72. D.I. Svergun, J.S. Pedersen, I.N. Serdyuk, and M.H.J. Koch, Solution Scattering from
50s Ribosomal-Subunit Resolves Inconsistency between Electron-Microscopic Models.
Proc. Natl. Acad. Sci. USA 91, 11826–11830 (1994).
73. D.I. Svergun, N. Burkhardt, J.S. Pedersen, M.H.J. Koch, V.V. Volkov, M.B. Kozin,
W. Meerwink, H.B. Stuhrmann, G. Diedrich, and K.H. Nierhaus, Solution scattering
structural analysis of the 70 S Escherichia coli ribosome by contrast variation .1. Invar-
iants and validation of electron microscopy models. J. Mol. Biol. 271, 588–601 (1997).
74. D.I. Svergun, N. Burkhardt, J.S. Pedersen, M.H.J. Koch, V.V. Volkov, M.B. Kozin,
W. Meerwink, H.B. Stuhrmann, G. Diedrich, and K.H. Nierhaus, Solution scattering
structural analysis of the 70 S Escherichia coli ribosome by contrast variation .2. A model
of the ribosome and its RNA at 3.5 nm resolution. J. Mol. Biol. 271, 602–618 (1997).
75. D.I. Svergun and K.H. Nierhaus, A map of protein-rRNA distribution in the 70 S
Escherichia coli ribosome. J. Biol. Chem. 275, 14432–14439 (2000).
76. J.K. Krueger, G.A. Olah, S.E. Rokop, G. Zhi, J.T. Stull, and J. Trewhella, Structures of
calmodulin and a functional myosin light chain kinase in the activated complex:
A neutron scattering study. Biochemistry 36, 6017–6023 (1997).
77. J.K. Krueger, G. Zhi, J.T. Stull, and J. Trewhella, Neutron-scattering studies reveal
further details of the Ca2+/calmodulin-dependent activation mechanism of myosin
light chain kinase. Biochemistry 37: 13997–14004 (1998).
78. J.K. Krueger, N.A. Bishop, D.K. Blumenthal, G. Zhi, K. Beckingham, J.T. Stull, and
J. Trewhella, Calmodulin binding to myosin light chain kinase begins at substoichio-
metric Ca2+ concentrations: A small-angle scattering study of binding and conforma-
tional transitions. Biochemistry 37, 17810–17817 (1998).
79. J.K. Krueger, S.C. Gallagher, G. Zhi, R. Geguchadze, A. Persechini, J.T. Stull, and
J. Trewhella, Activation of myosin light chain kinase requires translocation of bound
calmodulin. J. Biol. Chem. 276, 4535–4538 (2001).
80. G.A. Olah, S.E. Rokop, C.L. WangA, S.L. Blechner, and J. Trewhella, Troponin-I
Encompasses an Extended Troponin-C in the Ca2+-Bound Complex – a Small-Angle
X-Ray and Neutron-Scattering Study. Biochemistry 33, 8233–8239 (1994).
81. G.A. Olah and J. Trewhella, A Model Structure of the Muscle Protein Complex 4ca2+-
Center-Dot-Troponin-C-Center-Dot-Troponin-I Derived from Small-Angle Scattering
Data – Implications for Regulation. Biochemistry 33, 12800–12806 (1994).
82. W.A. King, D.B. Stone, P.A. Timmins, T. Narayanan, von A.A.M. Brasch,
R.A. Mendelson, and P.M.G. Curmi, Solution structure of the chicken skeletal
muscle troponin complex via small-angle neutron and x-ray scattering. J. Mol.
Biol. 345, 797–815 (2005).
83. W.T. Heller, D. Vigil, S. Brown, D.K. Blumenthal, S.S. Taylor, and J. Trewhella,
C Subunits binding to the protein kinase a RI alpha dimer induce a large conformational
change. J. Biol. Chem. 279, 19084–19090 (2004).
84. A.M. Augustus, P.N. Reardon, W.T. Heller, and L.D. Spicer, Structural basis for the
differential regulation of DNA by the methionine repressor protein MetJ. J. Biol. Chem.
281, 34269–34276 (2006).
85. P. Callow, A. Sukhodub, J.E. Taylor, and G.G. Kneale, Shape and subunit organisation
of the DNA methyltransferase M.AhdI by small-angle neutron scattering. J. Mol. Biol.
369, 177–185 (2007).
86. D. Comoletti, A. Grishaev, A.E. Whitten, I. Tsigelny, P. Taylor, and J. Trewhella,
Synaptic arrangement of the neuroligin/beta-neurexin complex revealed by X-ray and
neutron scattering. Structure 15, 693–705 (2007).
87. D.L. Ho, W.M. Byrnes, W.P. Ma, Y. Shi, D.J.E. Callaway, and Z.M. Bu, Structure-
specific DNA-induced conformational changes in Taq polymerase revealed by small
angle neutron scattering. J. Biol. Chem. 279, 39146–39154 (2004).
304 W.T. Heller and G.A. Baker

88. S. Krueger, S.K. Gregurick, J. Zondlo, and E. Eisenstein, Interaction of GroEL and
GroEL/GroES complexes with a nonnative subtilisin variant: a small-angle neutron
scattering study. J. Struct. Biol. 141, 240–258 (2003).
89. A.E. Whitten, D.A. Jacques, B. Hammouda, T. Hanley, G.F. King, J.M. Guss,
J. Trewhella, and D.B. Langley, The structure of the KinA-Sda complex suggests an
allosteric mechanism of histidine kinase inhibition. J. Mol. Biol. 368, 407–420 (2007).
90. J. Zimmer, D.A. Doyle, and J.G. Grossmann, Structural characterization and
pH-induced conformational transition of full-length KcsA. Biophys. J. 90, 1752–1766
(2006).
91. W.T. Heller, J.K. Krueger. and J. Trewhella, Further insights into calmodulin-myosin light
chain kinase interaction from solution scattering and shape restoration. Biochemistry 42,
10579–10588 (2003).
Chapter 17
Neutron Imaging Applied to Plant Physiology

T.M. Nakanishi

Abstract Until recently, the in situ study of roots and dynamic measurements
of water uptake in plants or trees has relied on physical, and therefore invasive
and often destructive, sampling or sensor installation in the surrounding soil.
These methods can, at the very least, alter the equilibrium of the rhizosphere of
the plant or require the removal of the tree or plant. The sensitivity of neutrons
to hydrogen atoms provides a noninvasive, nondestructive experimental
approach to understand real-time processes in plants (e.g., water uptake).
This chapter attempts to illustrate the effectiveness of neutron imaging techni-
ques in studying plant physiology, soils, and wood.

Keywords Botany  Neutrons  Cold Neutrons  Thermal Neutrons 

2D imaging  3D imaging  Charge-coupled device  in situ  Soil  Water
uptake  Wood  Flower  Carnation  Plant physiology  Roots  Growth 
Sand

17.1 Introduction

As discussed in Chapter 1, neutrons are particularly sensitive to hydrogen

atoms and can therefore provide information on their location with high spatial
resolution (50 mm). When a neutron beam passes through a sample, the
attenuation coefficient of some light elements — including hydrogen as well
as rare earth elements such as gadolinium, dysprosium, and samarium —
is extremely high, that is, 100–1000 times higher than those of the other
common elements [1, 2]. In the case of X-rays, the attenuation coefficient
changes by factors rather than orders of magnitude, from a light element to a
heavier one, making it difficult to distinguish the change in the contrast due to a

T.M. Nakanishi (*)

Graduate School of Agricultural and Life Sciences, The University of Tokyo,
1-1-1 Yayoi, Bunkyo-ku, Tokyo 113-8657, Japan
e-mail: [email protected]

I.S. Anderson et al. (eds.), Neutron Imaging and Applications, 305

Neutron Scattering Applications and Techniques, DOI 10.1007/978-0-387-78693-3_17,
Ó Springer ScienceþBusiness Media, LLC 2009
306 T.M. Nakanishi

specific element from other neighboring ones. Neutron imaging techniques are
naturally suited for materials containing hydrogen atoms and other low atomic
weight attenuating materials, permitting nondestructive in situ measurements
of the spatial and temporal distribution of water (and other fluids rich in
hydrogen) in plants and trees as well as in soils.

17.2 Nondestructive Water Observation in Plants and Soils

Using Neutron Transmission Imaging Techniques
Nondestructive techniques for analyzing processes within tissues of plants have
the potential to provide extremely important information for studying plant
activity. Neutrons are efficient at probing plant roots, which have greater water
content (between 70 and 95%) than their surrounding media (between 5 and
30%), as explained in [3]. Clearly the size of the root is limited to ensure a
‘‘reasonable’’ transmission thickness for neutrons, which may be up to several
millimeters depending on the neutron wavelength (for an example of neutron
transmission calculations, see Chapter 1). Water movement in particular, which
plays an important role both in biochemical processes and in physical structure,
has not been studied in detail. To acquire nondestructive in situ images of plant
or soil water distribution, the use of X-ray computed tomography (CT) mea-
surements has been reported [4]; however, the resolution obtained with this
method was insufficiently high to acquire microscopic-level images. X-ray
imaging has been able to resolve spatial patterns of root distribution in soil,
although this requires destructive collection of soil slices [5]. The most promis-
ing approach to obtaining images of water, as well as images of other biological
chemicals, is nuclear magnetic resonance spectroscopy. Although the resolution
of these images can theoretically reach 10 mm [6], the sample size is limited and
measurements are time-consuming.
Plant morphological development is reflected by water distribution within
tissues, which can be monitored using neutron imaging techniques [7]. Neutron
imaging of intact plant tissues can indeed yield exquisite microscopic images from
a unique and novel perspective, and without causing much radiation damage.
Although neutron imaging is the most suitable method for the study of living
plant behavior based on water uptake, historically, a few researchers in plant
physiology have employed it [8–10]. Recently, however, with the development
of neutron beam techniques such as cold neutron radiography [11–13], new
technological advances such as thin neutron sensitive imaging scintillators [14],
and advances in charge-coupled device cameras (Chapter 4), the use of neutrons
for plant research has increased significantly. Dynamic imaging (i.e., the study
of water movement) has especially increased in application but still remains
limited compared to other techniques.
In thin plant tissues such as flower petals or leaves, changes in the water
content are too small to image with thermal neutrons; therefore, cold neutrons
can be applied to a thin tissue to enhance the contrast, as, for instance, in the
17 Neutron Imaging Applied to Plant Physiology 307

case of a chrysanthemum leaf [13]. Although cold neutrons are limited in their
use because of a low transmission, small veins in the leaf can be observed, and
the contrast is approximately 70% higher than that of a thermal neutron image.
In addition, the range of changes in water thickness observed in the leaf is
smaller than 30 mm [15].
Neutron beam is perfectly suitable for analysis of the tissue level of the
sample, such as water distribution within a cut flower [3, 16–20] or a wood
sample [21–24]. However the resolution afforded by the technique, typically tens
of microns, is insufficient to study the process taking place within a single cell
because much higher resolution is needed for observation within a cell.
To obtain 2D neutron transmission images of living plants and their root
system in soil, samples are often grown in thin aluminum containers (2–3 mm
thick). The total cross section (absorption and scattering) of aluminum is
approximately 1/50 that of hydrogen; therefore, aluminum is almost transpar-
ent to thermal neutrons, and images of samples are not much affected by it.

17.2.1 Plant Roots Imbedded in Soil

Understanding of root growth is very limited because of the difficulty in observing
root systems without obvious damage of the surrounding media and/or the root
system itself. Often, destructive physical sampling or invasive sensor or minirhizo-
tron installation is necessary and ultimately perturbs the rhizosphere (i.e., immedi-
ate area around the root system). However, observation of growth rates and
root systems can be successfully imaged with neutron [25] or X-ray [26] imaging.
Figure 17.1 shows a neutron radiograph of a soybean plant root, after (A) 8
and (B) 15 days of growth in an aluminum container (of size 10  150  2 mm3)
packed with standard sand (silt type, 197–203 mm grain size, in pore size),
containing 18% (by weight) of water. Since the water content in the root (in
white in Fig. 17.1) is much higher than that in the surrounding sand, the root
image was clearly shown even in the presence of soil [27].
The water-deficient part of the plant (dark in the image) is clearly visible near
the upper part of the main root. To distinguish the water content of the root
more clearly, a part of the main root initiating the growth of lateral roots is
magnified in Fig. 17.1C. The apparent 3D image in Fig. 17.1D is produced by
converting the degree of whiteness, which is directly related to the water content
and to the height. The lower area near the root indicates a large amount of water
uptake by the root. The activity of the root is well seen in a 1-mm area from the
root surface. In this region, not only nutrients or water uptake, but also a
corporative interaction between root and microorganisms occurs. From
the neutron radiograph, it can be seen that the amount of water was very
low around some parts of the root, suggesting the possibility of water vapor
absorption rather than water solution. Possible future studies include the
measurement of heterogeneity in water distribution and uptake at the
rhizosphere. The latter phenomenon is expected to play an important role in
308 T.M. Nakanishi

Fig. 17.1 A neutron radiograph of a soybean plant root embedded in soil after (A) 8 and
(B) 15 days of growth; (C) magnification of the root; and (D) conversion of (B) to the
3D image

understanding the fundamental activities of the plant root system. Further-

more, neutron imaging may help explain the discrepancy between the amount
of nutrients the plant actually absorbs from its environment and the amount
‘‘invasively’’ measured by soil sample extraction.

17.2.2 Observation of Root Development Under Different

Soil Conditions
Noninvasive neutron imaging methods play an important role in the observa-
tion of growth rates and root development under different soil conditions, as
discussed in [25]. In general, roots seek the most favorable place to grow, and
adapt their growth orientation or pattern to the soil condition. For example,
when water-rich soil is provided at one lateral side of the main root, the
secondary roots preferably develop on that side. Similarly, Fig. 17.2A displays
changes in root development due to the presence of vanadium in the soil. A
soybean seedling is grown in an aluminum container where water-absorbing
polymer material [28] containing 50 mM of vanadium is embedded in the soil.
When vanadium is placed below the main root, the growth of the root ceases at
17 Neutron Imaging Applied to Plant Physiology 309

Fig. 17.2 Root development pattern. A: Soybean root development in the presence of 50 mM
of vanadium (white aggregate). B: Radish root development. (Left) Low-pH soil containing
23 mM of aluminum; (right) Modified lignin mixed with the soil shown in the image on the left
to reduce inhibition of growth by the aluminum

the polymer, but the side roots grows larger to compensate. On the other hand,
when the polymer ‘‘doped’’ with vanadium is placed on one side of the root, the
lateral root facing the polymer side does not grow, but the root system on the
other side is longer. The position at which growth stops and the direction of root
growth are dependent on the vanadium concentration inside the polymer [29]. It
is also shown that the amount of water uptake at a certain height of the main root
increases before side root growth begins.
Since root development is an indicator of the soil condition, neutron imaging
can be applied to evaluate soil conditioning agents as well as the effect of a
chemically modified fertilizer. To study lignin derivatives used as a soil con-
ditioning agent, a radish plant is employed to observe the root profile in an
acidic soil [30]. The root growth pattern captured in the neutron image in
Fig. 17.2B clearly shows that the modified lignin is effective as a soil condition-
ing agent. Subsequently, the total root length is measured through image
analysis, and the soil conditioning effect of different chemicals is analyzed.
The root pattern shown in Fig. 17.2B is similar to patterns found in lightning
or in the production of cracks in a pane of glass. There is no reproducibility in
the pattern—a fundamental feature, especially for roots, for maintaining bio-
logical diversity. Because of this, experimental errors cannot be quantified
properly when analysis of root growth is measured using invasive techniques
such as the line intersection method, in which a lattice board is placed on the
root and then removed from the soil, allowing the number of intersections
between the root and the lattice to be counted. Neutron imaging, because it is
nondestructive, can use the same sample to trace the root activity over time,
which reduces the fundamental errors that arise from using different samples.
Recent laboratory experiments reproducing water infiltration in soil, plant,
and the atmosphere have reported water redistribution processes as a function
310 T.M. Nakanishi

of different conditions such as, for example, root structure, soil structure, and
moisture [31].

17.2.3 Neutron Tomography of Root Systems

Since water movement in a plant is slow (approximately several minutes to

hours), computed tomography can be used to evaluate ‘‘static’’ spatial water
distribution inside and around root systems. For example, Fig. 17.3 illustrates a
soybean root grown in a 3-cm diameter aluminum container packed with
standard sand [27].

Fig. 17.3 In situ neutron tomography of the root system of a soybean during different days

Six CT reconstructions corresponding to different heights of the root system

permit comparison of water content changes. Within a few days, the amount of
water in the upper part of the container decreases, corresponding to an
increased formation of roots. Using tomographic root images, positions and
lengths of the side roots, as well as water movement around them, can be
quantified (Fig. 17.4) [31].
There is an obvious increase in side root growth around the upper part of the
container (up to 20 mm down from the air/soil interface) compared to side roots
located farther down the container (Fig. 17.4, left). This correlates to a decrease
in the water amount found in the soil after several days (6 days, for example) in
the same region (Fig. 17.4, right).
Root surfaces and volumes can also be calculated using neutron imaging
techniques. In the case of a soybean root, when 10 mM of AlCl3 solution is applied
to the soil, a decrease in both root surface area and root volume is observed [32].
The presence of aluminum ions is one of the main factors inhibiting plant growth
17 Neutron Imaging Applied to Plant Physiology 311

Fig. 17.4 Quantitative measurements of (left) side root growth according to their position
along the main root and (right) water distribution around the main root days after watering
the specimen. Depth is in mm

in an acidic soil. Moreover, root development in soil contaminated with heavy

metals has also been analyzed via neutron imaging, as illustrated in [33].

17.2.4 The Aboveground Portion of the Plant

Neutron imaging is commonly used to analyze the aboveground portion of a

plant. For example, neutron imaging of a cowpea plant can provide informa-
tion on stem internode tissue, whose function is to store water [34]. Under
water-deficient conditions, water is primarily moved from the internode to
other tissues. A complementary study of water labeled with a positron emitter,
15
O, has measured real-time movement of water in a plant, as referenced in [35].
Extending the life of cut flowers is a key issue in the floral industry, and the
amount of water in a flower plays a key role in maintaining its flowering stage.
Neutron imaging of cut flowers shows detailed images of water content in
flowers such as lilies, morning glories, and chrysanthemums. Figure 17.5
shows an example of carnation flowers imaged with thermal neutrons.

Fig. 17.5 Images of a carnation flower. A photograph is on the left, a neutron image in the
center, and a magnification of the neutron image on the right. In the left and center images,
the two flowers on the inside are controls and the two on the outside were dried before the
neutron radiograph was taken [36]
312 T.M. Nakanishi

Increasing the viscosity of water (i.e., xenon is dissolved under high pressure
into water) can help prolong the life of a carnation flower after it has been
harvested, as illustrated in Fig. 17.6A [36]. This helps control the metabolism of
the flower, which slows down the deterioration process mediated by an enzy-
matic reaction. An example of a 3D neutron imaging study is shown in
Fig. 17.6B. The flower respiration rate while conducting neutron imaging
during the senescent (i.e., aging) stage has been evaluated [3, 18, 19]. Further-
more, the analysis of 3D images of a carnation flower demonstrates the impor-
tance of water inside the ovule to maintain the flowering stage longer. It is

Fig. 17.6 Neutron image of a carnation flower. (A) Neutron radiographs of a carnation
flower after being supplied with water containing xenon. (B) CT images at different heights
along the flower head [36]
17 Neutron Imaging Applied to Plant Physiology 313

shown that the ovary and pistil areas continue to increase in size after the
respiration rate stops increasing [36].

17.3 Investigation of Wood Samples

The study of physiological activity of a tree mainly relies on knowing the water
distribution inside the wood. The example below illustrates the use of neutron
imaging on wood disks extracted from a tree cut only a few hours before
experiments, which therefore permits the observation of the green moisture
image [23, 24]. A stem section, covered as a precaution to prevent water loss, is
further cut to obtain 1 cm thick wood disks just prior to being exposed to
neutrons. Figure 17.7 shows the neutron images of disks from a Japanese
cypress (Chamaecyparis obtusa; Fig. 17.7A) and a Sugi (Cryptomeria japonica;
Fig. 17.7B), whose ages are estimated to be about 19 and 24 years, respectively.
The upper images in the figure are photographs of wood disks, and the

Fig. 17.7 Images of wood disks, each 1 cm thick. Photographs of disks are shown in the top
row; neutron images in the center row; and reconstructions of the middle images, with high
points corresponding to water-rich areas, in the bottom row
314 T.M. Nakanishi

center images are the corresponding neutron images. The lower images are
reconstructed ones based on the images in the center row with the degree of
whiteness indicated as height.
In the Japanese cypress, the neutron image shows higher water content in the
outer rings, adjacent to the bark. This zone includes the living phloem tissue and
recently developed xylem tissue that transports water from the roots to the
leaves. Both tissues have high water content. Water content declines radially
from the bark into the older, less functional, air-filled xylem tissue. Water
distribution in some areas shows a wavy pattern that might indicate tissue
damage or buildup of specific chemicals within the rings. In the heartwood
region, rings appear to have much lower water content, likely indicating buildup
of decay-resistant extractive chemicals. Besides the large hydrogen-rich outer
part of the disk, many rings corresponding to the annual rings of the wood are
shown to contain a higher level of water than their immediate surroundings.
Sugi is a popular wood in Japan for building houses and furniture. Although
there is only one species of Sugi, neighboring trees of the same cultivar show
different water content at the heartwood. It is not known what causes the
differences in moisture content at the heartwood. When lumber is processed,
residual moisture causes serious problems. When the moisture content at the
heartwood is high in green lumber, it is difficult to remove water completely
during the drying process. However, moisture is gradually lost after the lumber
is used in houses or furniture, causing the shape to warp over the years.
When neutron radiographs of wood discs are taken during the drying pro-
cess, the mechanism by which moisture is lost can be directly analyzed [24]. In
the case of Sugi, another feature in the disk is observed in the heartwood. As
shown in Fig. 17.7B, there is always a water-deficient area between heartwood
and sapwood. Because of the lighter color of this zone, it is called the white
zone. As the tree grows, the area of heartwood increases. However, the water
content throughout the heartwood is maintained at a high level as the tree
grows. It is not known how the water in sapwood moves into heartwood across
the white zone.
A rapid decrease in water content can be observed when the plant tissue
is injured. For example, neutron imaging shows that when Sugi wood is
inoculated with a canker fungus, a water deficiency occurs and spreads
from the inoculated site [30]. The dynamic process of water absorption in
a wood sample can be evaluated with neutron imaging techniques [22].
Imaging of solvent absorption and water-loss processes in small wood samples
(4  4  2.7 cm) has been reported [21] when acryl resin solvents are applied
to protect the surface of the wood, which is of importance for the development
of wood-protective materials for preserving wood art with historical and
cultural value.
Several other neutron imaging studies of wood can be found in [37–41].
17 Neutron Imaging Applied to Plant Physiology 315

17.4 Other Agricultural Applications

Besides the examples presented above, neutron imaging has been used in many
other agricultural applications. The water uptake of seeds during germination
provides useful information, for example, in investigating the method of
storing seeds because in some seeds, physical damage or the storage condition
may change the germination rate (Fig. 17.8) [42]. Even at an early stage of
germination, the germination rate of seeds can be estimated using neutron
images of water absorption.
Recently, a tomographic study of corn kernels has aimed at understanding
the effect of Aspergillus flavus (A-flavus fungus) to resistant species. Anatomic
changes are observed when resistant inoculated and uninoculated species are
compared to their nonresistant inoculated counterparts [43].

Fig. 17.8 Water absorption process in five kinds of seeds. Photographs of the seeds are shown
in the upper left corner (the black bar is 5 mm long). Seeds are soaked in water and neutron
images are taken after 0, 2, 4, and 6 h (top to bottom). For the broad bean, images taken after 0,
2, and 4 h are shown
316 T.M. Nakanishi

17.5 Concluding Remarks

Nondestructive imaging methods for imaging water or H-rich fluids in living

plants, soils, and wood using neutrons provide the highest resolution yet
obtained for in situ water content analysis. Due to the high sensitivity to H
atoms, this technique is capable of providing high-resolution images of water
movement in seeds, in roots embedded in soil, in wood, and in meristems during
development. The images presented in this chapter provide a taste of the
intriguing new areas of investigation in the field of plant physiology that are
being opened by new developments in neutron imaging methods.
Recently, the powerful methods derived from molecular genetics have
resulted in a tendency to focus research on the molecular aspects of biology
and to ignore important aspects of the intact plant. However, the intact plant
itself has a high potential to integrate many functions and to respond to many
diverse environmental conditions. Nondestructive techniques are essential to
studying and understanding the activity of a living plant, its development, and
its adaptation to environmental changes. The method reported here seems to be
the most promising new tool for doing so.

References
1. K.A. Garrett and H. Berger, Atom. Rev. 15, 125 (1977).
2. J.F.W. Markgraf and R. Matfield, Neutron beam, in J.C. Domanus, ed., Practical
Neutron Radiography, Kluwar Academic Publishers, Dordrecht, Boston, London,
pp. 26–50 (1992).
3. U. Matsushima, Y. Kawabata, C-M. Sim, K-Y. Nam, and T. Nishizawa, Proc. VIIIth IS
Postharvest Phys. Ornamentals, Eds. N. Marissen et al., Acta Hortic. 669, 111 ISHS
(2005).
4. J.M. Hainsworth and L.A.G. Alymore, Soil Sci. Soc. Am. J. 50, 841 (1986).
5. C.J. Moran, A. Pierret, and A.W. Stevenson, Plant Soil 223, 101 (2000).
6. P.T. Callaghan, Diffusion-limited resolution, in P.T. Callaghan ed., Principles of Nuclear
Magnetic Resonance Microscopy, Oxford Science, Oxford, pp. 203–206 (1991).
7. T.M. Nakanishi and M. Matsubayashi, J. Plant Phys. 151, 442 (1997).
8. S.T. Willatt et al., Agron. J. 70(4) 581–586 (1978).
9. S.T. Willatt et al., Ann. Bot. 43, 415–422 (1979).
10. P. Couchat et al., Agron. J. 72, 321–324 (1980).
11. A. Hilger, N. Kardjilov, M. Strobl, W. Treimer, and J. Banhart, Physica B 385(86), 1213
(2006).
12. Y. Kawabata, M. Hino, T. Nakano, H. Sunohara, U. Matsushima, and P. Geltenbort,
Nucl. Instrum. Meth. Phys. Res. A 542, 61 (2005).
13. Y. Kawabata, U. Matsushima, T. Horie, T. Nakano, and R. Maruyama, J. Radioanal.
Nucl. Chem. 264, 319 (2005).
14. G. Frei and E. Lehmann, Proceedings of 8th World Conference of Neutron Radiography.
Gaithersburg, USA, p. 21, October 16–19 (2006).
15. U. Matsushima, K. Kawabata, M. Hino, P. Geltenbort, and B. Nicolai, Nucl. Instrum.
Meth. Phys. Res. A 542, 76 (2005).
16. U. Matsushima, T. Ooshita, T.M. Nakanishi, M. Matsubayashi, Y. Seo, and Y. Kawagoe,
J. Jpn. Soc. Agr. Machin. 62, 70 (2000).
17 Neutron Imaging Applied to Plant Physiology 317

17. U. Matsushima, K. Kawabata, T. Nakano, M. Hino, P. Geltenbort, and B. Nicolai, Proc.

5th Int. Postharvest Symp. 682, pp. 1411 ISHS (2005).
18. U. Matsushima, Y. Kawabata, and T. Horie, J. Radioanal. Nucl. Chem. 264, 325 (2005).
19. U. Matsushima, E. Lehmann, P. Vontobel, G. Fre, B.M. Nicolai, T. Nishizawa, and
Y. Kawamitsu, Proceedings of the Asia-Pacific Symposium on Quality Management for
Agri-Foods in Supply Chains. Bangkok, Thailand, Aug. 7–10, p. 119 (2006).
20. T.M. Nakanishi, J. Furukawa, and M. Matsubayashi, Nucl. Instrum. Meth. Phys. Res.
A 424, 136 (1999).
21. E. Lehmann, S. Hartmann, and P. Wyer, Nucl. Instrum. Meth. Phys. Res. A 542, 87 (2005).
22. D. Mannes, E. Lehmann, S. Oswald, and P. Niemz, Proceedings of 5th Plant Biomechanics
Conference, Stockholm, August 28–September 1, pp. 393 (2006).
23. T.M. Nakanishi, I. Karakama, T. Sakura, and M. Matsubayashi, Radioisotopes 47, 387
(1998).
24. T.M. Nakanishi, T. Okano, I. Karakama, T. Ishihara, and M. Matsubayashi,
Holzforschung 52, 673 (1998).
25. M. Menon, B. Robinson, S.E. Oswald, A. Kaestner, K.C. Abbaspour, E. Lehmann, and
R. Schulin, Euro. J. Soil Sci. 58, 802 (2007).
26. A. Pierret, M. Kirby, and C. Moran, Plant Soil 255, 361–373 (2003).
27. T.M. Nakanishi, Y. Okuni, J. Furukawa, K. Tanoi, H. Yokota, N. Ikeue, M. Masubayashi,
H. Uchida, and A. Tsuji, J. Radioanal. Nucl. Chem. 255, 149 (2003).
28. T.M. Nakanishi, S. Matsumoto, and H. Kobayashi, Radioisotopes 42, 26 (1993).
29. J. Furukawa, T.M. Nakanishi, and M. Matsubayashi, Nucl. Instrum. Meth. Phys. Res. A
424, 116 (1999).
30. T. Yamada, Y. Aoki, M. Yamamoto, M. Komatsu, D. Kusumoto, K. Suzuki, and
T.M. Nakanishi, J. Radioanal. Nucl. Chem. 264, 329 (2005).
31. J. Furukawa, T.M. Nakanishi, and M. Matsubayashi, Nondestr. Test Eval. 16, 335 (1999).
32. K. Saito and T.M. Nakanishi, Mokuzai Gakkaishi 43, 669 (1997).
33. T.M. Nakanishi, S. Matsumoto, and H. Kobayashi, Radioisotopes 41, 638 (1992).
34. T.M. Nakanishi, K. Don-Jin, T. Kitamura, R. Ishii, and M. Matsubayashi, 242,
353 (1999).
35. Y. Okuni, J. Furukawa, M. Matsubayashi, and T.M. Nakanishi, Anal. Sci. 17(Supple-
ment), i1499 (2001).
36. T.M. Nakanishi, J. Furukawa, and M. Matsubayashi, Nucl. Instrum. Meth. Phys. Res.
A 424, 136 (1999).
37. E.H. Lehmann, P. Vontobel, and P. Niemz, Investigation of Moisture Distribution in
Wooden Structures by Neutron Radiography. PSI Annual Report Annex VI, p. 53 (1999).
38. E.H. Lehmann, P. Vontobel, P. Niemz, and P. Haller, The method of neutron radio-
graphy and its use for wood properties analysis. In Proceedings of the International
Conference on Wood and Wood Fibre Composites, Stuttgart, Germany (2000).
39. E.H. Lehmann, P. Vontobel, P. Scherrer, and P. Niemz, Application of neutron radio-
graphy as method in the analysis of wood. Holz als Rohund Werkstoff 59(6), 463–471
(2001).
40. P. Niemz, E.H. Lehmann, P. Vontobel, P. Haller, and S. Hanschke, Investigations using
neutron radiography for evaluations of moisture ingress into corner connections of wood.
Holz als Roh-und Werkstoff 60(2), 118–126 (2002).
41. V. Bucur, Nondestructive Characterization and Imaging of Wood, Springer Series in Wood
Science, (Hardcover), ISBN 3-540-43840–8 (2003).
42. T.M. Nakanishi and M. Matsubayashi, Bioimages 5, 45 (1997).
43. T.E. Cleveland et al., J. Cereal Sci. 48, 517 (2008).
Chapter 18
Homeland Security and Contraband Detection

R.C. Lanza

18.1 Introduction

18.1.1 Nature of Problem

Detection of contraband and illicit materials has become increasingly impor-
tant, especially since the terrorist attacks in the United States on September
11, 2001. The nature of the detection problem embodies both physics issues
and a set of operational constraints that limit the practical application of
neutrons. The issue under consideration is detection of materials that are
considered serious threats; these may include explosives, radioactive materi-
als, fissile materials, and other materials associated with nuclear weapons,
often referred to as special nuclear material (SNM). The overriding con-
straint is in the physics: systems must be based on clean physics; but unlike
physics experiments, detection systems work under the limitation that mate-
rials must be identified nonintrusively, without interrupting the normal flow
of commerce and with a high probability of detection and a low probability
of false alarms. A great deal of work has been reported in the literature on
neutron-based techniques for detecting explosives and drugs. The largest
impetus by far for detecting explosives comes from aviation industry
requirements for inspecting luggage and, to a lesser extent, cargo. The
major alternative techniques are either X-ray–based or chemical trace detec-
tion methods that look for small traces of explosive residues. The limitations
of the X-ray and trace methods in detecting explosives are well known, but
currently (2008) it is safe to say that no neutron- or nuclear-based technique
is being used routinely for security inspection, despite extensive development
of these methods. Smuggling of nuclear materials has become a concern,
and neutron techniques are particularly attractive for detecting them. Given

R.C. Lanza (*)

Department of Nuclear Science and Engineering, MIT, NW13-221, Cambridge,
MA 02139-4307, USA
e-mail: [email protected]

I.S. Anderson et al. (eds.), Neutron Imaging and Applications, 319

Neutron Scattering Applications and Techniques, DOI 10.1007/978-0-387-78693-3_18,
Ó Springer ScienceþBusiness Media, LLC 2009
320 R.C. Lanza

the limitations of X-ray techniques and the need for SNM detection, it is now
useful to reexamine neutron methodologies, particularly imaging. A significant
number of neutron-based techniques have been proposed and are under devel-
opment for security applications, especially SNM detection, but describing how
they work is beyond the scope of the chapter. Instead, one particular approach
to neutron imaging, neutron resonance radiography (NRR), is discussed in
detail as it illustrates many of the issues connected with imaging and detection.

18.1.2 Operational Issues

An important part of any proposed system design, beyond physics, is a descrip-

tion of how the actual measurement and detection will be done, how the system
integrates into end-users’ operations, and a means for testing systems under
realistic conditions. The ability to reliably inspect containers without interrupt-
ing normal commerce is essential for any system to be deployed. It is necessary
to have a reasonably detailed knowledge of various environments where a
technology will operate to ensure that it will actually be usable.
Another important point is that the goal of these inspection systems is not to
identify the contents of every container but rather to determine if the materials
of interest are not present in the container. Consider the sheer volume of
inspections compared with the number of actual threats that are present.
There are estimates that, for example, the number of checked bags to be
examined may be as high as 109 per year. The number of actual threats will be
much smaller, perhaps ultimately only 103 or so. The difference,106, means that
system performance is dominated by false alarms; most ‘‘threat alarms’’ are in
fact false alarms. Given this, we must devise means to clear containers rapidly;
hence the idea is to determine that a threat is not present, rather than making a
complete analysis of the container.
One approach is to simply consider a system with a single decision variable
that is used to distinguish threats from nonthreats. In Fig. 18.1, the results of
a series of measurements are shown. Clearly, to the extent that there are
overlaps, it is not possible to have a decision variable with a high detection
probability without generating false alarms. One way of expressing this
tradeoff is to use the receiver–operator curve (ROC), Fig. 18.2, which
shows the tradeoff in a clearer manner. This curve can be generated
theoretically from assumptions about the distribution of the decision para-
meters in Fig. 18.1, but it is often necessary to generate it instead by experi-
mental measurement. The line at 458 in Fig. 18.2 also illustrates the result of
a purely random measurement. Unfortunately, the ratio of measurements
(inspections) with no threat to measurements (inspections) with threats may
be as high as 106 to 1, resulting in high false alarm rates. The goal, then, is to
develop inspection methods that are specific to a given threat so as to make
the separation as large as possible, as shown in Fig. 18.1. This is the driving
force behind the use of neutrons in imaging.
18 Homeland Security and Contraband Detection 321

Fig. 18.1 Influence of decision threshold on false positive rate relative to true positive rate

Fig. 18.2 Receiver–operator curve (ROC). The dashed line represents random performance
322 R.C. Lanza

18.1.3 Practicality

The goal of all systems is their eventual use in a nonlaboratory environment, a

goal that presents issues of safety, ease of operation, and reliability.
Generally, almost all systems share the requirement that they be relatively
compact and operable by personnel relatively unskilled in nuclear instrumenta-
tion. This point is well known to designers of, for example, baggage X-ray
scanners such as those used at airports and X-ray- and gamma-ray–based cargo
container scanners. It is especially significant in techniques based on large
accelerators and detector assemblies, which present a new level of complexity
and complication. Most airports and ports are exceedingly crowded and either
lack space for large instruments or would prefer to put the space for other uses
that more directly impact the economic or operational aspects of the port.
All of these systems rely on the use of neutron sources, generally either a
small accelerator [radio frequency quadrupole (RFQ) or electrostatic] or a
sealed deuterium–tritium (DT) generator [1]. With the exception of the DT
(or deuterium–deuterium) sealed-tube generators, accelerators remain a critical
problem in deploying these techniques. Issues of size, weight, and general
complexity must be solved before these systems can be placed in a typical
inspection environment.
Practical systems also require sophisticated software for control and
calibration and for automatic or semiautomatic threat detection algorithms.
Most of these have been developed by vendors over years and form a
substantial part of the cost of deploying scanners, even after the hardware
has been built.
Performance specifications generally require specific detection probabilities,
spatial resolution, scanning speed, and reliability. These are often couched in
terms such as the ability to see a wire of a certain diameter behind a 25-cm steel
plate or the ability to detect, for example, 100 cc of SNM behind 25 cm of steel.
Safety issues are especially important, since these systems are often large
radiation producers that are used in semipublic areas, or at least areas in which
it may be anticipated that there will be members of the general public and non
radiation workers [2].

18.1.4 Special Nuclear Materials

The SNMs are the basic ingredients for nuclear weapons, and their control
and interdiction are highly important. The current prevailing model for
addressing threats associated with cargo-container-borne weapons of mass
destruction, especially nuclear weapons, could be characterized as a cus-
toms-based approach. Under this scheme, new systems using either active or
passive techniques to detect nuclear signatures, from hidden fissile material
would be integrated into the existing customs infrastructure of ports. The
18 Homeland Security and Contraband Detection 323

customs-based approach is a natural extension of conventional strategies for

finding and seizing incoming contraband. Nuclear weapons, however, are
utterly unlike any other kind of contraband in that their power is so great,
and their effects are so far-reaching, that even discovering them in port (or
allowing them to enter port at all) cannot provide adequate protection from
their destructive reach.
Despite decades of research and development and considerable expenditure
on alternative approaches, the systems used to examine cargo are almost the
same as those used in the 1970s with a few minor improvements. The most
common nonintrusive inspection systems use X-ray transmission to form an
image of the interior of a cargo container and search for anomalies. Some add
X-ray backscattering to detect materials, such as drugs, hidden in the walls of
containers. The major application of this technology is the verification of cargo
manifests and the detection of large amounts of smuggled materials. There is,
however, one overarching issue: no current X-ray system, whether transmission
or transmission enhanced with backscattering, can distinguish fissile material
from ordinary materials. Thus, X-ray systems are practically useless in inter-
cepting nuclear weapons. The detection of fissile material requires a system
designed for that purpose, and neutron imaging provides a particularly
powerful tool.

18.1.5 Explosives

Explosive materials come in a large variety of physical and chemical forms,

and their density and elemental composition vary widely [3]. Military-style
high explosives have been used in the past; but in recent times the use of
improvised and nonconventional explosives has increased, and the spectrum
of potential threats has therefore expanded. Although some have argued
that the primary common characteristic of these materials is that they
explode, there remain characteristics of explosives that can be detected.
Many commercial and military explosives have high concentrations of nitro-
gen; but recent experience has revealed a new set of nonnitrogenous explo-
sives often fabricated by terrorists from readily available and fairly ordinary
materials that can be purchased without arousing suspicion. It is essential
that the designer of detection equipment recognize the vast variety of
potential threats and thus clearly understand what the limitations of a
particular technique may be. The implication of the widespread use of
these improvised explosives is that detection of, for example, nitrogen
alone will often be inadequate; indeed, the use of assumed elemental ratios
for explosive composition may lead to difficulties in detection. Neutron
imaging is an approach to detection that can be used to image elemental
composition and hence has the potential to detect these newer explosives.
324 R.C. Lanza

18.1.6 Others

Depending on circumstances, there may be other materials of interest for

detection, such as drugs and other contraband or even smuggling of illicit
goods to avoid customs taxes. Although these materials are of great concern,
the current prevailing worry appears to have shifted from illicit drugs to
materials likely to be associated with explosives and SNM.

18.1.7 Physics of Detection

Neutron techniques are based on detecting the nuclear properties of possible

threats rather than their density or chemical properties. This has several advan-
tages over X-ray techniques, such as the ability to determine elemental compo-
sition rather than just density, a high degree of penetration and potential for
cargo inspection, the difficulties of shielding contraband against probing radia-
tion, and the ability to detect SNM. However, in the interest of completeness,
one should recognize that the presence of large absorbers in an object under
inspection will also trigger a ‘‘shield alarm’’ in many X-ray imaging systems and
thus is in itself an indication of a potential threat.
The capability to determine the nuclear or elemental properties of objects under
inspection is the major advantage of nuclear techniques compared with conven-
tional X-ray scanners, which generally measure the apparent density and shape of
objects. This capability can be enhanced in the case of dual-energy X-ray systems.
An approximation to the average atomic number (Z) of the material also can be
obtained and thus used, along with spatial properties, to identify suspicious
material. Neutron imaging techniques go beyond this capability to offer the
possibility of determining the unique elemental composition of a suspicious object.
Imaging provides spatial and geometric information that can be used in
recognizing and identifying suspect objects, and reduces the background due
to extraneous objects in the container. Imaging often reduces the signal, but it
reduces the ‘‘noise’’ by a much larger factor. Consequently, the signal-to-noise
ratio improves; the ability to actually detect objects in the presence of back-
ground improves; and it is possible to note spatial relationships between SNM,
explosives, and high-Z containers and shields.

18.2 Representative Technical Approaches

A number of approaches to detect explosives and SNM using neutron techni-

ques exist, some of which can also be applied to imaging. Reviews of these
applications for detecting explosives have been provided by Buffler [4] and
Lanza [5]. Few of these methods have actually reached the prototype stage
and some are beyond prototypes.
18 Homeland Security and Contraband Detection 325

18.2.1 Pulsed Fast Neutron Analysis

One highly developed neutron approach that has been tested is pulsed fast
neutron analysis (PFNA) [6–9]. In this system, shown in Figs. 18.3 and 18.4,
fast neutrons are used to produce gamma rays from neutron reactions with
elements of interest; for explosives, this means particularly C, N, and O, but
other elements are also detectable. The approach in principle is also capable of
detecting SNM by detection of delayed gammas and neutrons between pulses,
but this has not been its primary design goal.
The operation is as follows: Neutrons in the range of 8 MeV are generated
by an electrostatic accelerator by the reaction D(d,n)He. These neutrons
are collimated into a beam and the beam is mechanically scanned across a
container. At the same time, the accelerator is pulsed to produce 1-ns pulses
of neutrons with a repetition rate of 1 MHz. Gamma rays from inelastic
interactions are detected by an array of scintillation detectors. The time differ-
ence between the accelerator pulse and gamma detection is used to determine
the position along the beam where the interaction took place. The pulsed
approach also reduces background and clutter for another reason: the number
of neutrons produced per pulse is low enough so that there is typically only one

Fig. 18.3 Pulsed fast neutron analysis system; from Brown [7]
326 R.C. Lanza

Fig. 18.4 Schematic of pulsed fast neutron analysis system

neutron interaction in the container during a given 1-ns pulse; the next potential
event comes during the next beam pulse, 1 microsecond later.
The beam line is mechanically scanned so as to move the beam through the
object, therefore the gammas can be correlated with the position of interaction
in three dimensions, and a three-dimensional image is formed. The gamma-ray
spectrum obtained for each voxel is used to determine relative elemental content
in the object under inspection. For 8-MeV neutrons, a time-of-flight measure-
ment of 1 ns yields an accuracy of 5 cm in space. The physics for this
particular system has been refined over a number of years. Note that the choice
of neutron energy,  8 MeV, means that scattered neutrons will generally lose
enough energy so that they fall below the threshold for gamma production and
thus do not contribute to spatial clutter. Another point to observe is that the
detectors, are conventional scintillators (NaI) rather than high-resolution
detectors, such as HPGe. The choice of detectors was made after both experi-
mental observation and cost analysis. NaI has considerably higher efficiency
and count rate capability than does HPGe, but it has lower energy resolution.
This disadvantage is overcome by the use of the entire spectrum rather than
single peaks to determine elemental composition. Rather than fit the data to
individual peaks, this approach fits the entire spectrum to a linear combination
of whole spectra from each of the elements, a method long used in areas such as
well-logging and online coal and mineral analysis.
18 Homeland Security and Contraband Detection 327

18.2.2 The ‘‘Nuclear Car Wash’’

Lawrence Livermore National Laboratory (LLNL) has developed a neutron

system specifically for the detection of SNM in cargo containers, naming it
after the system’s resemblance to car-washing systems [10–12]. Cargo is inspected
as it moves through a fan-beam of high-energy neutrons, and SNM is detected by
the presence of fission neutrons and gammas. The neutron source is an RFQ that
produces 7-MeV neutrons through the reaction D(d,n)3He. The gammas are
detected by means of large plastic scintillators that form walls on either side of
a container as it moves through the inspection system (Fig. 18.5). Fission neu-
trons, either prompt or delayed, are often used as the signature of fission;
however, in this system, the primary signature is the detection of gammas from
decays of short-lived isotopes. These gammas are almost an order of magnitude
more abundant than the delayed neutrons and are emitted on a time scale of
1 min. The distribution of energies and times produces a unique signature of
fission, and further, high-energy gammas are more likely to escape from the
container to be detected. It is estimated that the intensity of such delayed gammas
may be orders of magnitude greater than the intensity of delayed neutrons.

Fig. 18.5 The Lawrence Livermore National Laboratory (LLNL) nuclear ‘‘car wash’’

Extensive modeling and preliminary experiments have confirmed the perfor-

mance potential of this approach as a detector for SNM in cargo containers.
Figure 18.6 shows the high-energy gamma-ray spectrum with a U target absent
(lower curve) and present (upper curve).

18.2.3 Combined Neutron and Gamma Radiography

Combined neutron and gamma radiography has been extensively developed by
the Commonwealth Scientific and Industrial Organization and is currently in
328 R.C. Lanza

Fig. 18.6 Delayed gamma spectrum with special nuclear material (SNM) (black) and without
SNM (gray)

production and deployment as an air cargo inspection system [13, 14]. Its basic
approach is to image a container using a combination of 14-MeV neutrons
generated from a sealed DT neutron generator and gamma rays from a 60Co
source (1.17, 1.33 MeV). The ratio, R, of the neutron and gamma attenuation
coefficients can be determined on a pixel by pixel basis. This ratio shows a
significant correlation with material type and can thus be used to discriminate
among various materials. Figure 18.7 shows this ratio for various materials of
interest in cargo. The ratio is color-coded and displayed as an image for visual
inspection by the operator (Fig. 18.8).

Fig. 18.7 Ratio (R) of neutron to gamma attenuation coefficients for various materials
18 Homeland Security and Contraband Detection 329

Fig. 18.8 Optical picture (left) and R-ratio image of aircraft container (right)

18.2.4 Neutron Resonance Radiography

Fast neutron resonance radiography (NRR) is a neutron imaging technique
proposed as a method for elemental imaging [15–19]. NRR exploits the
fact that the total neutron cross section exhibits unique resonance structure
for many elements (Fig. 18.9). In a two-dimensional projection image, the

Fig. 18.9 Total neutron cross section for H, C, N, and O in the MeV energy region. With the
exception of hydrogen, clear and unique resonance structures exist for each element
330 R.C. Lanza

attenuation image is the sum of elemental contents. Attenuation images

taken at different neutron energies form a set of linear equations that can
be solved to map individual elemental contents; in principle, this makes
detection possible based on the presence of specific elements or combina-
tions of elements.
The attenuation of a beam of neutrons traversing a slab of element a with
thickness in the beam direction xa is governed by

 
Ix
 ln ¼ Na a ðEÞxa ;
Iair

where  lnðIx =Iair Þ represents the ratio of the neutron beam attenuated by
the object to the unattenuated beam, Na represents the atom density of
element a, and a(E) represents the total neutron cross section of material a
at energy E. All equations in this section apply to a single pixel. Aside from
the intrinsic energy variation in the neutron cross section, factors related to
the neutron production mechanism will introduce a smearing in the cross
section as a function of energy. The notation 1a is used to represent the
effective neutron cross section for material a at energy setting 1. Notation-
ally, superscripts will represent the energy setting while subscripts represent
the material. The measurement  lnðIx =Iair Þat energy setting 1 is defined as
b1. Finally, Na  xa is combined into a . In a two-dimensional projection
image, the attenuation image is the sum of elemental contents. For a
material comprising several different elements (a, b, c, d), the observed
attenuation becomes

1a a þ 1b b þ 1c c þ 1d d ¼ b1 :

Several attenuation images taken at different neutron energies form a set

of linear equations that can be solved to map individual elemental contents.
After performing measurements at, for example, six energies, the set of
equations is

1a a þ 1b b þ 1c c þ 1d d ¼ b1

2a a þ 2b b þ 2c c þ 2d d ¼ b2
3a a þ 3b b þ 3c c þ 3d d ¼ b3
:
4a a þ 4b b þ 4c c þ 4d d ¼ b4
5a a þ 5b b þ 5c c þ 5d d ¼ b5
6a a þ 6b b þ 61 6
c c þ  d d ¼ b
6
18 Homeland Security and Contraband Detection 331

This can be cast into a matrix form A ¼ b as

0 1 0 11
1a 1b 1c 1d b
B 2 2 C 0 1 B 2C
B a 2b 2c d C a Bb C
B 3 C B C
B a 3b 3c 3d C B b C
C B b3 C
B CB
B C B 4C
¼ B
B 4 4d C C :
B a 4b 4c C @ c A B b C
B 5 C B 5C
@ a 5b 5c 5d A d @b A
6a 6b 6c 61
d b6

As long as the number of equations is greater than the number of elements,

these equations can be solved by a least squares method. Measurement of the
matrix A and the set of observed attenuations b will allow determination of .
This in principle makes detection possible based on the presence of specific
elements and enables the formation of a two-dimensional image with elemental
separation.
In addition to its multielement detection capability, NRR offers the distinct
advantage that attenuation images are governed by a 1/R2 falloff in the neutron
flux, whereas other methods (‘‘neutron in, photon out’’) are governed by 1/R2
*1/R2 = 1/R4. Using the kinematics of the reaction D(d,n)3He [20], a single
energy deuteron beam can be used to generate 2- to 6-MeV neutrons, as shown
in Fig. 18.9. An alternative approach to obtain energy-resolved neutron spectra
uses a pulsed ‘‘white’’ neutron source with a broad range of neutron energies
and separates them out by time of flight [21].
Despite its conceptual simplicity, NRR is technically difficult in practice.
The challenges include producing a variable-energy neutron source with suffi-
cient flux and minimal energy spread at each energy, detecting fast neutrons in
the presence of photons, and producing basic measurements of pure elements of
sufficient quality so that reliable fits can be obtained in a complex image
environment typical of actual cargo. This work was the result of a collaboration
between L-3 Communications, the Massachusetts Institute of Technology
(MIT), and LLNL.

18.2.5 Apparatus

The primary system components are shown in Fig. 18.10. A fixed-energy

deuteron beam is provided by a commercial 3-MeV RFQ accelerator. The
beam is focused and transported to the deuterium gas target by the high-
energy beam transport system. An adjustable source collimator comprising
borated-polyethylene wedges surrounds the deuterium gas target. The entire
neutron source sits on a platform that rotates from 0 to 1208 to allow
selection of the neutron energy. Following the source collimator are a set
of fixed front and rear object collimators. The front object collimator
332 R.C. Lanza

Detector
source
Position: 60 Deg. rotated

Detector array
Neutron beam

RFQ accelerator

LD-3
Cont
ainer

Object
Collimators

Fig. 18.10 Neutron resonance radiography imaging system. The deuteron accelerator on the
rotating platform (in gray) rotated to the 608 position. Front and rear object collimators
are shown with the object being imaged (typically an LD-3 air cargo container) between
the collimators, resting on the object transport system. The detector array is behind the
rear object collimator

prevents stray radiation from the source from reaching the object, and the
rear object collimator prevents radiation scattered from the object from
reaching the detectors. The neutron beam is defined by aligning the
adjustable source collimator with the fixed object collimators and the detec-
tor array. The object to be imaged (typically a standard aircraft LD-3
container) is transported through the beam on a motorized trolley. The
system is housed in a warehouse with minimal climate control (heat only)
and no running water. The prototype occupies a radiation safety exclusion
zone of about 25 by 15 m.
The detector system is based on a combination of plastic scintillator and
photomultiplier tube in which the fast neutron is detected by proton recoil in the
plastic. Typically, gamma background signals are larger than those produced
by neutrons, and the electronics are based on a window discriminator for
gamma rejection.
The success of the NRR technique is largely determined by the accuracy of
the elemental cross sections. Initially, Monte Carlo methods were used to
18 Homeland Security and Contraband Detection 333

simulate neutron scattering on elements. The cross sections extracted from the
simulations were then used to study the possibility of creating elemental maps.
However, it was quickly realized that the ideal detector in the type of back-
ground-free environment assumed in the Monte Carlo studies does not reflect
actual experimental conditions. Hence, a precise experimental extraction of the
elemental cross sections was necessary as photon background can significantly
change the extracted neutron cross section. Overwhelming photon background
washes out the resonant features in the cross sections and makes NRR difficult;
reducing the photon-to-neutron ratio is essential to the success of the method.
Figure 18.11 shows experimentally determined carbon and nitrogen cross
sections. All elemental cross sections were lower than those calculated using
Monte Carlo simulations, largely as a result of gamma contamination. How-
ever, the resonant structure of these elemental cross sections is still observed,
allowing for a good elemental separation in compounds.
Once the elemental calibration had been completed, a set of angles was
selected for performing radiographic analysis. The best set of angles is the one
that allows the most precise determination of elemental composition in com-
pounds. Analytically, these angles would correspond to the most pronounced
features on elemental cross sections. For example, the oxygen cross section
exhibits the most unique behavior at high angles (100 through 1158), whereas
the nitrogen cross section has a distinct structure at lower angles (around 358).
The carbon cross section is most distinctive at around 908. Hydrogen and, to a
lesser extent, silicon cross sections exhibit no distinct resonance structure. A non-
trivial relationship among these five cross sections in determining elemental
composition is best studied quantitatively using a computer optimization code.
To find the best set of angles, compounds composed solely of the basis
materials (such as melamine, H6C3N6) were scanned at 58 increments in
accelerator angle. To resolve a set of five basis elements, scans of six or
more angles were required. Since a larger number of angles linearly increase
the scan time, it was found that seven angles represented a good balance

Fig. 18.11 Measurement of total neutron cross section for carbon (left) and for nitrogen
(right) as a function of accelerator angle (degrees)
334 R.C. Lanza

between quality and scan time. For each unique set of 7 angles picked from
a set of 24, the square of the difference between radiographically determined
and known elemental composition in each compound was calculated. The
sum of these differences was then minimized to produce the best set of
angles. Figure 18.12 (left) shows the results of the angle optimization
studies, comparing the theoretical atom fraction for melamine with the
one determined from the set of optimal angles. The resulting calibration
was also tested with various compounds not included in the optimization. A
scan of one such compound, ammonium nitrate (NH4NO3), in air is shown
in Fig. 18.12 (right). The relative amounts of hydrogen and nitrogen are in
good agreement with expected values. The amount of oxygen is somewhat
lower than the expectation, but still pronounced in comparison to carbon
and ‘‘other’’ represented by silicon.

Fig. 18.12 Left: Atom fraction of melamine sample as determined from the calibration
procedure. The basis elements are H, C, N, O, and Si. The black (right) bars are the
expected values and the gray (left) bars represent the best values obtained during
the optimal angle selection. The correctness of the melamine atom fraction was one of
the constraints of the angle selection. Right: Atom fraction of ammonium nitrate sample.
The basis elements are H, C, N, O, and Si. The black (right) bars are the expected values
and the gray (left) bars represent the best values obtained after the optimal angle
selection. The correctness of the ammonium nitrate atom fraction was not constrained
by the angle selection

18.2.6 Elemental Maps

The results in Fig. 18.12 are derived from blocks of pixels on isolated
objects rather than from a more realistic situation with multiple elements.
Figures 18.13 and 18.14 show images developed by the system together with
their elemental maps. Images presented are rotated by 908 from their actual
scan direction: top to bottom in the image is left to right in the scan; left to
right in the image is bottom to top in the scan. The leftmost part of
Fig. 18.13 represents the attenuation image at one of the 7 energies scanned
18 Homeland Security and Contraband Detection 335

Fig. 18.13 Identification of

water sample from 11 other
targets. Shown in the panes
(left to right) are a single
energy attenuation image, a
hydrogen elemental map, an
oxygen elemental map, and
identification of water target
through the coincident
presence of hydrogen and
oxygen

Fig. 18.14 Identification of

melamine sample from 11
other targets. Shown in the
panes (left to right) are a
single energy attenuation
image and the nitrogen
elemental map. The
melamine sample is clearly
distinguished from the other
samples in the nitrogen
elemental map

with the presence of 12 different targets. Also shown are the hydrogen and
oxygen elemental maps and finally, on the right, the result of a simple
algorithm that correctly identifies the water sample by identifying the coin-
cident presence of hydrogen and oxygen. Similarly, Fig. 18.14 shows the
identification of the nitrogen-rich melamine.
In the next set of images, Figs. 18.15 and 18.16, a series of targets are
superimposed into typical air cargo contents. As an example of algorithm
development in this project, a preliminary algorithm was developed to detect
graphite. The carbon elemental image appears to be very stable and have low
noise. The carbon block is detected with high efficiency in an environment of
typical air cargo contents is seeded with other potential targets.
336 R.C. Lanza

Fig. 18.15 Automated detection of carbon object in an air cargo of cut flowers. The left pane is
the carbon image and the right pane is the sum of the elemental map for all five basis elements.
The carbon block alone is correctly identified by the algorithm among the clutter (100 cm
hydrangeas) and other seeded targets

Fig. 18.16 Automated

detection of carbon object in
air cargo. The left pane is the
carbon image and the right
pane is the sum of the
elemental map for all five
basis elements. The carbon
block alone is correctly
identified by the algorithm
amongst the clutter
(electronics) and other
seeded targets
18 Homeland Security and Contraband Detection 337

18.3 Summary and Future Outlook

All of the methods described in this chapter have demonstrated the

potential, but final deployment will require a significant amount of engi-
neering to make them faster, more reliable, and capable of automatic, or at
least semi-automatic, operation with a minimal number of skilled technical
level operators. This work has concentrated only on one aspect of this
problem—those applications that involve both neutrons and imaging.
Until recently, much of the work was concentrated in the area of detection
of contraband and explosives; however, more recently, there has been
increased interest in the detection of SNM. The NRR system is now being
modified to enable it to detect SNM as well as explosives; this work will be
reported in the future. Active interrogation techniques generally have relied
on relatively large accelerators, and that has limited their practical deploy-
ment. However, if more compact and less expensive neutron sources were
developed, there would be a greater impetus toward deployment, especially
for SNM detection. Solving the source problem will not in itself lead to
deployment, as the issues of cost, complexity, and capability to rapidly
detect threats in complex backgrounds still must be overcome.

Acknowledgments The work on NRR was a collaboration between the MIT Department of
Nuclear Science and Engineering, the MIT Laboratory for Nuclear Science, L3-Communica-
tions, and LLNL. The author especially thanks David Perticone (L-3), who was the principle
investigator on the NRR project and was responsible for much of the progress made on it, as
well as Vitalyi Ziskin (L-3 and MIT), Gongyin Chen (L-3 and Varian), Whitney Raas (MIT),
and Gordon Kohse (MIT) for their contributions throughout the project. Jim Hall (LLNL),
Brian Rusnak (LLNL), and John Watterson (University of the Witwatersrand) made invalu-
able contribution in the early parts of the project.
Others include the staff of the MIT-Bates Laboratory for their excellent contributions to
the construction and operation of the NRR prototype, especially Bob Fisk, Peter Binns, Jim
Kelsey, Peter Goodwin, Christoph Tschalaer, and Jan Van Der Laan; and the following
engineering staffs of L-3 Communications for the design of system components: Tim Hart,
Ken Prather, Jim Karon, Tony Antoniou, John Price, Keith McClelland, Brain Pearson,
Steve McDevitt, Jeff Stillson, and Sal Gargiulo.

References
1. D.L. Chichester and J.D. Simpson, Compact accelerator neutron generators, The Indus-
trial Physicist, Dec-Jan 2003/2004 9(6), 22–25 (2004).
2. ANSI 43.14, Radiation Safety Guidelines for Active Interrogation Systems for Security
Screening of Cargo. American National Standards Institute (2008).
3. J. Zukas and W. Walters, eds., Explosive Effects and Applications, Springer-Verlag, New
York, NY (1997).
4. A. Buffler, Contraband detection with fast neutrons, Rad. Phys. Chem. 71(3–4), 853–861
(2004).
5 R. Lanza, Neutron techniques for detection of explosives in Y. Yinon, ed., Counterterrorist
Detection Techniques of Explosives, Elsevier, Amsterdam, pp. 131–155 (2007).
338 R.C. Lanza

6. T. Gozani, Novel applications of fast neutron interrogation methods, Nuc. Instrum.

Meth. A 353(1–3), 635–640 (1994).
7. D.R. Brown and T. Gozani , Cargo inspection system based on pulsed fast neutron
analysis, Nucl. Instrum. Meth. B 99(1–4), 753–756 (1995).
8. R. Loveman, J. Bendahan, T. Gozani and J. Stevenson ,Time of flight fast neutron
radiography, Nucl. Instrum. Meth. B 99(1–4), 765–768 (1995).
9. D.R. Brown, T. Gozani, R. Loveman, J. Bendahan, P. Ryge, J. Stevenson, F. Liu and M.
Sivakumar, Application of pulsed fast neutrons analysis to cargo inspection, Nucl.
Instrum. Meth. A 353(1–3), 684–688 (1994).
10. D. Slaughter, M. Accatino, A. Bernstein, J. Candy, A. Dougan, J. Hall, A. Loshak, D.
Manatt, A. Meyer, B. Pohl, S. Prussin., R. Walling,and D.Weirup, Detection of special
nuclear material in cargo containers using neutron interrogation, UCRL-ID-155315 (2003).
11. D.R. Slaughter, M.R. Accatino, A.Bernstein, A.D. Dougan, J. M. Hall, A. Loshak,D.R.
Manatt, B.A. Pohl, R.S. Walling, D.L. Weirup, and S.G. Prussin, The ‘‘Nuclear Car
Wash’’: A Scanner to Detect Illicit Special Nuclear Material in Cargo Containers, IEEE
Sens. J. 5(4), 560–564 (2005).
12. J.M. Hall, S. Asztalos, P. Biltoft, J. Church, M.-A. Descalle, T. Luu, D. Manatt, G.
Mauger, E. Norman, D. Petersen, J. Pruet, S. Prussinb, and D. Slaughter, The Nuclear
Car Wash: Neutron interrogation of cargo containers to detect hidden SNM, Nucl.
Instrum. Meth. B 261, 337–340 (2007).
13. J.E. Eberhardt, Y. Liu, S. Rainey, G.J. Roach, R.J. Stevens, B.D. Sowerby and J.R.
Tickner, Fast Neutron and Gamma-Ray Interrogation of Air Cargo Containers in
Proceedings of International Workshop on Fast Neutron Detectors and Applications
(FNDA2006) in Proceedings of Science http://pos.sissa.it/
14. J.E. Eberhardt, Y. Liu, S. Rainey, G.J. Roach, R.J. Stevens, B.D. Sowerby and J.R.
Tickner, Fast neutron radiography scanner for the detection of contraband in air cargo
containers, Appl. Rad. Isot. 63, 179 (2005).
15. G. Chen, Fast Neutron Resonance Radiography for Elemental Imaging: Theory and
Applications, Ph.D. Thesis, MIT, April (2001).
16 G. Chen, R.C. Lanza, Fast neutron resonance radiography for elemental imaging: Theory
and applications, IEEE Trans. Nucl. Sci. 49(4), 1919–1924 (2002).
17. W. L. Raas, B.W. Blackburn, E.M. Boyd, G.E. Kohse, and R.C. Lanza, Neutron
interrogation for threat interdiction: technical challenges, IEEE Nuclear Science Sympo-
sium Conference Record (2005).
18. B.W Blackburn, G.Chen, W.A Franklin, E. E Ihloff, G. E Kohse, R.C. Lanza, B.
McAllister, D. Perticone, V. Ziskin, Fast Neutron Resonance Radiography for Home-
land Security, IEEE Nuclear Science Symposium Conference Record (2007).
19. B.W. Blackburn, G. Chen, W.A. Franklin, E.E Ihloff, G.E Kohse, R.C. Lanza, B.
McAllister, D. Perticone, and V. Ziskin, Fast Neutron Resonance Radiography for
Elemental Imaging: First Results, IEEE Trans. Nucl. Sci. (in press) (2008).
20. H. Lisikien and A. Paulson, Neutron Production Cross Sections and Energies for the
Reactions T(p,n)3He, D(d,n)3He, and T(d,n)4He, Nucl. Data Tables 11, 569–619 (1973).
21. P.K. Van Staagen, T.G. Miller, B.C.Gibson, R.A. Krauss, High speed data acquisition
for contraband identification using neutron transmission, AIP Conference Proceedings, n
392, pt.2, pp. 853–856 (1997).
Index

A E
Absorption, 5, 7, 10, 11, 22, 33, 36, 43, 47, 48, Epithermal neutrons, 48, 234, 235, 243–244,
49, 68, 70, 71, 75, 81, 82, 85, 92, 94, 97, 255, 256, 257, 258, 263
103–106, 110, 125–126, 140–142, 146,
164, 175, 178, 183, 194–195, 210, 221, 233,
234, 235, 243, 253, 258, 307, 314, 315 F
Accelerator, 14, 15–21, 22, 27, 28, 68–70, Fission, 4, 14, 15, 16, 21–22, 27–28, 48, 70,
168, 230, 263, 270–272, 281, 322, 325, 110, 157, 258, 327
331–333, 337 Flow, 69, 75, 78, 97–98, 196, 198, 202–203,
204–205, 211, 212, 215, 218, 221, 254, 319
Flower, 306–307, 311, 312–313, 336
B Flux, 5–6, 13–14, 15, 16, 17, 19–21, 22, 23, 24,
Biology, 122, 293, 316 25, 26, 27, 28, 29, 32, 34, 35, 37, 39, 49, 68,
Biomedical, 211, 253, 262, 265 70, 76–77, 82, 83, 106, 110, 114, 125–126,
Botany, 305–316 174, 175, 183–184, 186, 197, 199–200, 218,
219–220, 224, 243, 257, 270–271, 272, 278,
C 281, 282, 331
Carnation, 311–312 Fuel cell, 8, 75, 78–79, 97–98, 106, 109–110,
Casting, 209–226 192–199, 200–201, 202–205, 206, 210, 211
Charge-coupled device (CCD), 52–53, 76, 82, Functional, 254, 269–270, 314
84–85, 111–114, 118, 199, 209, 218, 219,
220, 221, 220, 306
G
Cold Neutrons, 47, 58, 94, 124, 177, 179–180,
Gas neutron detector, 56–58
209–210, 231, 236, 242, 254–255, 306–307
Granular, 212, 216, 217–218
Composites, 110, 221
Growth, 247, 307–311
Contrast agent, 74, 255, 257, 262
Cross section, 5–11, 18–19, 21–22, 24–25, 36,
42, 47, 48, 52, 70, 71, 73, 82, 94–95, 102, H
110, 111, 116, 117, 121, 122, 125, 130–131, Hydrogen storage, 181, 192–199, 202,
140, 143, 156, 157, 158, 164, 173–174, 206, 210
182–183, 186, 194, 195, 197, 199, 212, 213,
223, 231, 235, 244, 255, 262, 307, 329–330,
332–333 I
Imaging, 4, 8, 10, 11, 13, 26, 27, 29, 67–68, 69,
69, 72, 73, 76–78, 82, 85, 104, 109–110,
D 117–124, 129–130, 131, 136, 142, 149, 156,
2D imaging, 85, 86, 88 164, 168, 171–186, 194–195, 199–203,
3D imaging, 85, 86, 92, 130, 200, 230, 234, 211, 297–299, 306–313, 314, 315–316,
307–308, 312 320, 324, 329, 332, 337

339
340 Index

Imaging neutron detector, 47–60 69–70, 72, 73–74, 82, 83–84, 85, 86–87,
Inelastic scatter, 22, 36, 157–158, 176, 265, 93–94, 95, 97, 100, 101, 104, 106, 109–110,
266, 267–268, 271–272 111, 122, 124, 129, 130, 131, 132, 136–137,
In situ, 106, 168, 194–195, 205, 212, 221, 139, 145, 146, 149, 156–157, 158, 160,
306, 310 161–163, 164–165, 171–186, 191–192,
Instrument, 13–14, 16, 17, 20, 22, 23, 24, 27, 194, 209–211, 213, 214–215, 216, 218,
34, 35, 36, 38, 42, 44, 45, 47, 52, 53–54, 56, 219–220, 222, 230–240, 241, 242,
96, 102, 157, 168, 173–174, 175, 176–177, 243–244, 249, 253, 254–255, 256–257,
179, 183–184, 185, 216, 226, 229, 233, 258, 260, 262, 263, 265, 266, 270–272, 280,
234–235, 247, 248, 265, 293, 322 281, 289–290, 305–307, 310, 315–316,
325–326, 327, 332, 337
Neutron stimulated emission computed
M tomography (NSECT), 258, 266,
Magnetic fields, 4, 11, 87, 106, 171–172, 173, 267–268, 269–270, 272, 273–274, 275,
175, 182, 185 276, 277–278, 279, 280–282
Magnetic resonance imaging, 171–186, 254 Numerical, 73, 122, 130, 183, 195–196,
Materials, 4, 6, 8–10, 11, 14, 22, 24–25, 33, 217–218
40–41, 48, 52, 68, 70, 74–75, 78, 94, 96, 99,
109–110, 111, 115, 124, 125–126, 131, 139,
140, 143, 154, 156, 166, 168, 172, 176, 181, P
191–192, 194, 197, 199, 200, 206, 209–210, Plant physiology, 305–316
211, 212, 215–218, 219, 221, 223, 224–225, Polarized neutrons, 11, 82, 86, 106,
229, 230, 231, 234, 235, 244, 245–246, 249, 171–186
253, 254, 255, 289, 291, 305–306, 314, Portable source, 28
319–320, 322–323, 324, 328, 333–334 Proton, 3–4, 8, 14, 15–21, 26, 47, 78, 156, 182,
Medical imaging, 86, 123, 254, 262, 263, 265, 183, 192, 193, 194, 270, 332
266, 268–270 Proton Exchange Membranes (PEM),
Metals, 4, 42, 44, 75–76, 95, 96, 111, 142–143, 192–194, 196–197, 200–201, 203–205
144, 154, 194–195, 211–215, 223, 230, 245, Pulse, 19, 20, 21, 22, 25, 26, 35, 58, 178,
262, 311 179–180, 181, 183, 185, 186, 218, 325–326
Modeling, 106, 112–114, 192, 215, 217, 219, Pulse width, 23, 24, 26
289–299, 327 Pyrolysis, 212, 214, 215
Moderators, 4, 14, 15, 16, 20, 22, 23, 24, 25,
26, 32, 34, 70–71, 72, 239, 257
Multi-Phase, 214–215 R
Reactor source, 10, 14, 15, 17, 26–27, 149, 232
Reflectors, 15, 22, 24
Roots, 154–155, 306, 307–309, 310–311,
N 314, 316
Neutron converter, 47–49, 58
Neutron detector, 28, 47, 53, 56, 58, 59–60,
85–86, 179 S
Neutron energy, 4, 7, 10, 11, 18, 21, 22, 23, Sand, 211–212, 216–217, 307, 310
25, 26, 28, 111, 140, 149, 176, 194, 200, Scattering, 4, 6, 7–10, 11, 13–14, 15, 17,
226, 231–232, 271, 272, 326, 331–332 21–22, 26, 29, 34, 36, 39–40, 41–42, 48, 49,
Neutron imaging, 4, 5, 11, 14, 17, 20, 28, 47, 56, 68, 82, 94, 100–106, 110, 111, 122, 131,
50, 52, 72, 77, 78–79, 82–83, 106, 109–126, 143–145, 155, 156, 157–158, 159, 160, 161,
129, 179, 191–206, 209–226, 229–249, 162, 163, 164, 166, 168, 175, 176, 177, 183,
253–263, 289, 305–316, 319, 323, 324, 194, 239, 253–254, 255, 273, 289–290, 291,
329–330 293, 296, 297, 333
Neutron production, 15–21, 28, 330 Scintillation neutron detector, 52–53, 84,
Neutrons, 4, 5–6, 7, 8, 11, 13–14, 15–16, 18, 325–326
19, 21–22, 24–25, 26, 27–28, 31, 32, 33, Small particulate gadolinium oxide
35–36, 39, 40–41, 42, 43, 44, 47–60, 67–68, (SPGO), 262
Index 341

Soil, 306, 307–311, 316 Time-of-flight (TOF), 10, 14, 19, 21, 26, 27,
Solid state neutron detector, 58–60 36, 39, 47, 52, 53, 56, 59, 124, 178, 180,
Spallation, 4, 14, 16, 18, 110 244, 271, 326
Spallation source, 14, 19, 20, 21–25, 26, 27, Tomography, 11, 27, 53, 54, 77, 81–106, 111,
29, 70, 125–126, 157, 168, 221 115, 117, 130, 132–136, 140, 178, 181, 195,
Spectroscopy, 49, 166–167, 181–183, 210, 196, 200–201, 202–203, 210, 216, 219, 223,
258, 267–268, 269, 271, 274, 279, 306 230, 232, 233, 234, 240–242, 245, 247, 249,
Spin-echo, 175, 176–177 254, 258, 261, 265–282, 306, 310
Structural, 8, 122, 123, 137, 145, 166–167,
168, 210, 211, 254, 269–270, 289–290, 293,
294, 295, 297 W
Water dynamics, 196–197, 200
Water uptake, 306, 307, 309, 315
T Wood, 230, 254, 307, 313–316
Thermal Neutrons, 7, 8, 15, 21–22, 28, 29, 47,
51–52, 59, 68, 70, 78, 132, 149, 153–168,
194, 202–203, 209–210, 213, 218, 220, 232, X
253, 254–255, 256–257, 258, 260, 262, 307 X-ray imaging, 117, 260, 306, 324