Chapter 2. Geometry, Data, Physics, and Mathematics 2.4.

Physics

To generate cross-section tables for electron/photon transport problems that will use the multigroup Boltzmann-
Fokker-Planck algorithm [60], the CEPXS [61–63] code developed by Sandia National Laboratory and available
from RSICC can be used. The CEPXS manuals describe the algorithms and physics database upon which the
code is based; the physics package is essentially the same as ITS version 2.1. The keyword “MONTE-CARLO”
is needed in the CEPXS input file to generate a cross-section library suitable for input into CRSRD; this
undocumented feature of the CEPXS code should be approached with caution.

2.4 Physics

The physics of neutron, photon, and electron interactions is the very essence of the MCNP code. A review
of charged particle transport capabilities in the MCNP code can be found in [43]. For a description of all
high-energy event generators used by the MCNP code, see [64]. This section may be considered a software
requirements document in that it describes the equations the MCNP code is intended to solve. All the
sampling schemes essential to the random walk are presented or referenced. But first, particle weight and
particle tracks, two concepts that are important for setting up the input and for understanding the output,
are discussed in the following sections.

2.4.1 Statistical Weight

At the most fundamental level, weight is a tally multiplier. That is, the tally contribution for a weight w
is the unit weight tally contribution multiplied by w. Weight is an adjustment for deviating from a direct
physical simulation of the transport process. Note that if a Monte Carlo code always sampled from the same
distributions as nature does, then the Monte Carlo code would have the same mean and variance as seen
in nature. Quite often, the natural variance is unacceptably high and the Monte Carlo code modifies the
sampling using some form of “variance reduction” [§2.7]. The variance reduction methods use weighting
schemes to produce the same mean as the natural transport process, but with lower calculational variance
than the natural variance of the transport process.

With the exception of the pulse height tally ( F8 ), all tallies in the MCNP code are made by individual
particles. In this case, weight is assigned to the individual particles as a “particle weight.” The manual
discusses the “particle weight” cases first and afterward discusses the weight associated with the F8 tally.

2.4.1.1 Particle Weight

If the MCNP code were used only to simulate exactly physical transport, then each MCNP particle would
represent one physical particle and would have unit weight. However, for computational efficiency, the MCNP
code allows many techniques that do not exactly simulate physical transport. For instance, each MCNP
particle might represent a number w of particles emitted from a source. This number w is the initial weight
of the MCNP particle. The w physical particles all would have different random walks, but the one MCNP
particle representing these w physical particles will only have one random walk. Clearly this is not an exact
simulation; however, the true number of physical particles is preserved in the MCNP code in the sense of
statistical averages and therefore in the limit of a large number of MCNP source particles (of course including
particle production or loss if they occur). Each MCNP particle result is multiplied by the weight so that the
full results of the w physical particles represented by each MCNP particle are exhibited in the final results
(tallies). This procedure allows users to normalize their calculations to whatever source strength they desire.
The default normalization is a weight of one per MCNP source particle. A second normalization to the
number of Monte Carlo histories is made in the results so that the expected means will be independent of the
number of source particles actually initiated in the MCNP calculation.

LA-UR-22-30006, Rev. 1 67 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

The utility of particle weight, however, goes far beyond simply normalizing the source. Every Monte Carlo
biasing technique alters the probabilities of random walks executed by the particles. The purpose of such
biasing techniques is to increase the number of particles that sample some part of the problem of special
interest (1) without increasing (and sometimes actually decreasing) the sampling of less interesting parts of
the problem, and (2) without erroneously affecting the expected mean physical result (tally). This procedure,
properly applied, increases precision in the desired result
√ compared to an unbiased calculation taking the
same computing time. For example, if an event √ is made 2 times as likely to occur (as it would occur without
biasing), the tally ought to be multiplied by 1/ 2 so that the expected average
√ tally is unaffected. This tally
multiplication can be accomplished by multiplying the particle weight by 1/ 2 because the tally contribution
by a particle is always multiplied by the particle weight in the MCNP code. Note that weights need not be
integers.

In short, particle weight is a number carried along with each MCNP particle, representing that particle’s
relative contribution to the final tallies. Its magnitude is determined to ensure that whenever the MCNP
code deviates from an exact simulation of the physics, the expected physical result nonetheless is preserved in
the sense of statistical averages, and therefore in the limit of large MCNP particle numbers. Its utility is in
the manipulation of the number of particles, sampling just a part of the problem to achieve the same results
and precision, obviating a full unbiased calculation which has a longer computing time.

2.4.1.2 Pulse-height Tally ( F8 ) Weight

Unlike other tallies in the MCNP code, the pulse height tally depends on a collection of particles instead of
just individual particles. Because of this, a weight is assigned to each collection of tallying particles. It is this
“collective weight” that multiplies the F8 tally, not the particle weight.

When variance reduction is used, a “collective weight” is assigned to every collection of particles. If variance
reduction techniques have made a collection’s random walk q times as likely as without variance reduction,
then the collective weight is multiplied by 1/q so that the expected F8 tally of the collection is preserved.
The interested reader should consult [65, 66] for more details.

2.4.2 Particle Tracks

When a particle starts out from a source, a particle track is created. If that track is split 2 for 1 at a splitting
surface or collision, a second track is created and there are now two tracks from the original source particle,
each with half the single track weight. If one of the tracks has an (n,2n) reaction, one more track is started for
a total of three. A track refers to each component of a source particle during its history. Track length tallies
use the length of a track in a given cell to determine a quantity of interest, such as fluence, flux, or energy
deposition. Tracks crossing surfaces are used to calculate fluence, flux, or pulse-height energy deposition
(surface estimators). Tracks undergoing collisions are used to calculate multiplication and criticality (collision
estimators).

Within a given cell of fixed composition, the method of sampling a collision along the track is determined
using the following theory. The probability of a first collision for a particle between l and l + dl along its line
of flight is given by
p(l)dl = exp(−Σt l)Σt dl, (2.2)
where Σt is the macroscopic total cross section of the medium and is interpreted as the probability per unit
length of a collision. Setting ξ the random number on [0, 1), to be

ˆl
ξ= exp(−Σt s)Σt ds = 1 − exp(−Σt l), (2.3)
0

LA-UR-22-30006, Rev. 1 68 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

it follows that
1
l=− ln(1 − ξ). (2.4)
Σt
However, because 1 − ξ is distributed in the same manner as ξ and hence may be replaced by ξ, we obtain
the well-known expression for the distance to collision,
1
l=− ln(ξ). (2.5)
Σt

2.4.3 Neutron Interactions

When a particle (representing any number of neutrons, depending upon the particle weight) collides with a
nucleus, the following sequence occurs:

1. the collision nuclide is identified;

2. either the S(α, β) treatment is used or the velocity of the target nucleus is sampled for low-energy
neutrons;

3. photons are optionally generated for later transport;

4. neutron capture (that is, neutron disappearance by any process) is modeled;
5. if the energy of the neutron is low enough and an appropriate S(α, β) table is present, the collision is
modeled by the S(α, β) treatment;

6. otherwise, either elastic scattering or an inelastic reaction (including fission) is selected, and the new
energy and direction of the outgoing track(s) are determined.

2.4.3.1 Selection of Collision Nuclide

If there are n different nuclides forming the material in which the collision occurred, and if ξ is a random
number on the unit interval [0, 1), then the k th nuclide is chosen as the collision nuclide if
k−1
X n
X k
X
Σt,i < ξ Σt,i ≤ Σt,i , (2.6)
i=1 i=1 i=1

where Σt,i is the macroscopic total cross section of nuclide i. If the energy of the neutron is low enough
(below about 4 eV) and the appropriate S(α, β) table is present, the total cross section is the sum of the
capture cross section from the regular cross-section table and the elastic and inelastic scattering cross sections
from the S(α, β) table. Otherwise, the total cross section is taken from the regular cross-section table and is
adjusted for thermal effects [§2.4.3.2].

2.4.3.2 Free Gas Thermal Treatment

A collision between a neutron and an atom is affected by the thermal motion of the atom, and in most cases,
the collision is also affected by the presence of other atoms nearby. The thermal motion cannot be ignored in
many applications of the MCNP code without serious error. The effects of nearby atoms are also important
in some applications. The MCNP code uses a thermal treatment based on the free gas approximation to
account for the thermal motion. It also has an explicit S(α, β) capability that takes into account the effects

LA-UR-22-30006, Rev. 1 69 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

of chemical binding and crystal structure for incident neutron energies below about 4 eV, but is available for
only a limited number of substances and temperatures. The S(α, β) capability is described in §2.4.3.6.

The free gas thermal treatment in the MCNP code assumes that the medium is a free gas and also that, in
the range of atomic weight and neutron energy where thermal effects are significant, the elastic scattering
cross section at zero temperature is nearly independent of the energy of the neutron and that the reaction
cross sections are nearly independent of temperature. These assumptions allow the MCNP code to have a
thermal treatment of neutron collisions that runs almost as fast as a completely non-thermal treatment and
that is adequate for most practical problems.

With the above assumptions, the free gas thermal treatment consists of adjusting the elastic cross section and
taking into account the velocity of the target nucleus when the kinematics of a collision are being calculated.
The MCNP free gas thermal treatment effectively applies to elastic scattering only.

Cross-section libraries processed by NJOY already include Doppler broadening of elastic, capture, fission, and
other low-threshold absorption cross-sections (< 1 eV). Inelastic cross sections are never broadened by NJOY.

2.4.3.2.1 Adjusting the Elastic Cross Section

The first aspect of the free gas thermal treatment is to adjust the zero-temperature elastic cross section by
raising it by the factor

√

F = 1 + 0.5/a2 erf(a) + exp −a2 / a π , (2.7)
p
where a = AE/kT , A is the atomic weight of the nucleus, E is the incident neutron energy, and T is the
material temperature. For speed, F is approximated by F = 1 + 0.5/a2 when a ≥ 2 and by linear interpolation
in a table of 51 values of aF when a < 2. Both approximations have relative errors less than 0.0001. The
total cross section also is increased by the amount of the increase in the elastic cross section.

The adjustment to the elastic and total cross sections is done partly in the setup of a problem and partly
during the actual transport calculation. No adjustment is made if the elastic cross section in the data library
was already processed to the temperature that is needed in the problem. If all of the cells that contain a
particular nuclide have the same temperature, which is constant in time, that is different from the temperature
of the library, the elastic and total cross sections for that nuclide are adjusted to that temperature during the
setup so that the transport will run a little faster. Otherwise, these cross sections are reduced, if necessary,
to zero temperature during the setup and the thermal adjustment is made when the cross sections are used.
For speed, the thermal adjustment is omitted if the neutron energy is greater than 500 kT /A. At that energy
the adjustment of the elastic cross section would be less than 0.1%.

Note that this adjustment of the nuclear data is less accurate than the one used within NJOY, as NJOY
will handle more reactions and does not assume constant data. As such, it is recommended to use datasets
Doppler-broadened to the temperature of interest, rather than relying on this adjustment. See the discussion
in the TMP card for more information.

2.4.3.2.2 Sampling the Velocity of the Target Nucleus

The second aspect of the free gas thermal treatment takes into account the velocity of the target nucleus
when the kinematics of a collision are being calculated. The target velocity is sampled and subtracted from
the velocity of the neutron to get the relative velocity. The collision is sampled in the target-at-rest frame
and the outgoing velocities are transformed to the laboratory frame by adding the target velocity.

There are different schools of thought as to whether the relative energy between the neutron and target, Er ,
or the laboratory frame incident neutron energy (target-at-rest), Eo , should be used for all the kinematics of

LA-UR-22-30006, Rev. 1 70 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

the collision. Eo is used in the MCNP code to obtain the distance-to-collision, select the collision nuclide,
determine energy cutoffs, generate photons, generate fission sites for the next generation of a KCODE criticality
problem, for S(α, β) scattering, and for capture. Er is used for everything else in the collision process, namely
elastic and inelastic scattering, including fission and (n, xn) reactions. It is shown in Eq. (2.8) that Er is
based upon vrel. that is based upon the elastic scattering cross section, and, therefore, Er is truly valid
only for elastic scatter. However, the only significant thermal reactions for stable isotopes are absorption,
elastic scattering, and fission. 181 Ta has a 6 keV threshold inelastic reaction; all other stable isotopes have
higher inelastic thresholds. Metastable nuclides like 242m Am have inelastic reactions all the way down to zero,
but these inelastic reaction cross sections are neither constant nor 1/v cross sections and these nuclides are
generally too massive to be affected by the thermal treatment anyway. Furthermore, fission is very insensitive
to incident neutron energy at low energies. The fission secondary energy and angle distributions are nearly
flat or constant for incident energies below about 500 keV. Therefore, it makes no significant difference if Er
is used only for elastic scatter or for other inelastic collisions as well. At thermal energies, whether Er or Eo
is used only makes a difference for elastic scattering.

If the energy of the neutron is greater than 400 kT and the target is not 1 H, the velocity of the target is set
to zero. Otherwise, the target velocity is sampled as follows. The free-gas kernel is a thermal interaction
model that results in a good approximation to the thermal flux spectrum in a variety of applications and can
be sampled without tables. The effective scattering cross section in the laboratory system for a neutron of
kinetic energy E is ¨
1 dµt
eff
σs (E) = σs (vrel. )vrel. p(V )dv . (2.8)
vn 2
Here, vrel. is the relative velocity between a neutron moving with a scalar velocity vn and a target nucleus
moving with a scalar velocity V , and µt is the cosine of the angle between the neutron and the target
direction-of-flight vectors. The equation for vrel. is

1/2
vrel. = vn2 + V 2 − 2vn V µt . (2.9)

The scattering cross section at the relative velocity is denoted by σs (vrel. ), and p(V ) is the probability density
function for the Maxwellian distribution of target velocities,
4 3 2

p(V ) = β V exp −β 2 V 2 , (2.10)
π 1/2
with β defined as
 1/2
AMn
β= , (2.11)
2kT
where A is the mass of a target nucleus in units of the neutron mass, Mn is the neutron mass in MeV-sh2 /cm2 ,
and kT is the equilibrium temperature of the target nuclei in MeV.

The most probable scalar velocity V of the target nuclei is 1/β, which corresponds to a kinetic energy of kT
for the target nuclei. This is not the average kinetic energy of the nuclei, which is 3kT /2. The quantity that
the MCNP code expects on the TMP n input card is kT and not just T [§5.7.5]. Note that kT is not a function
of the particle mass and is therefore the kinetic energy at the most probable velocity for particles of any mass.

Equation (2.8) implies that the probability distribution for a target velocity V and cosine µt is
σs (vrel. )vrel. p(V )
P (V, µt ) = . (2.12)
2σseff. (E)vn
It is assumed that the variation of σs (v) with target velocity can be ignored. The justification for this
approximation is that (1) for light nuclei, σs (vrel. ) is slowly varying with velocity, and (2) for heavy nuclei,
where σs (vrel. ) can vary rapidly, the moderating effect of scattering is small so that the consequences of the
approximation will be negligible. As a result of the approximation, the probability distribution actually used
is p

P (V, µt ) = vn2 V 2 − 2V vn µt V 2 exp −β 2 V 2 . (2.13)

LA-UR-22-30006, Rev. 1 71 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

Note that the above expression can be written as

p
vn2 V 2 − 2V vn µt  3


P (V, µt ) = V exp −β 2 V 2 + vn V 2 exp −β 2 V 2 . (2.14)
vn + V
As a consequence, the following algorithm is used to sample the target velocity.

√
1. With probability α = 1/(1 + ( πβvn /2)), the target velocity V is sampled from the distribution


P1 (V ) = 2β 4 V 3 exp −β 2 V 2 . (2.15)
√
The transformation V = y/β reduces this distribution to the sampling distribution P (y) = y exp(−y).
The MCNP code actually codes 1 − α.

2. With probability 1 − α, the target velocity is sampled from the distribution

√


P2 (V ) = 4β 3 / π V 2 exp −β 2 V 2 . (2.16)

Substituting V = y/β reduces the distribution to the sampling distribution for y to

√


P (y) = 4/ π y 2 exp −y 2 .

3. The cosine of the angle between the neutron velocity and the target velocity is sampled uniformly on
the interval –1 ≤ µt ≤ 1.
4. The rejection function R(V, µt ) is computed using
p
vn2 + V 2 − 2V vn µt
R(V, µt ) = ≤ 1. (2.17)
vn + V

With probability R(V, µt ), the sampling is accepted; otherwise, the sampling is rejected and the procedure
is repeated. The minimum efficiency of this rejection algorithm corresponding to assuming V = vn = vrel.
averaged over µt is
´1 ˆ1 p √ ˆ1
−1
R(vrel. , µt )dµt 1 2 + v 2 − 2v 2 µ
vrel. rel. t 2 p 2
´1 = rel.
dµt = 1 − µt dµt = , (2.18)
dµt 2 2vrel. 4 3
−1 −1 −1

which approaches 100% as either the incident neutron energy approaches zero or becomes much larger than
kT .

For more accuracy, the probability distribution in Equation 2.12 can be directly sampled without the constant
cross-section approximation. This is enabled through the DBRC card. This is not enabled by default.

2.4.3.3 Optional Generation of Photons

Photons are generated if the problem is a combined neutron/photon run and if the collision nuclide has a
nonzero photon production cross section. The number of photons produced is a function of neutron weight,
neutron source weight, photon weight limits (entries on the PWT card), photon production cross section,
neutron total cross section, cell importance, and the importance of the neutron source cell. No more than 10
photons may be born from any neutron collision. In a KCODE calculation, secondary photon production from
neutrons is turned off during the inactive cycles.

LA-UR-22-30006, Rev. 1 72 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

Because of the many low-weight photons typically created by neutron collisions, Russian roulette is played
for particles with weight below the bounds specified on the PWT card, resulting in fewer particles, each having
a larger weight. The created photon weight before Russian roulette is
Wn σγ
Wp = , (2.19)
σt
where

Wp is the photon weight,

Wn is the neutron weight,

σγ is the photon production cross section, and
σt is the total neutron cross section.

Both σγ and σt are evaluated at the incoming neutron energy without the effects of the thermal free gas
treatment because nonelastic cross sections are assumed independent of temperature.

The Russian roulette game is played according to neutron cell importances for the collision and source cell.
For a photon produced in cell i where the minimum weight set on the PWT card is Wimin. , let Ii be the neutron
importance in cell i and let Is be the neutron importance in the source cell. If Wp > Wimin. Is /Ii , one or more
photons will be produced. The number of photons created is Np , where

Wp I i
Np = + 1 , Np ≤ 10. (2.20)
5Wimin. Is

Each photon is stored in the bank with weight Wp /N
p . If Wp < Wimin. Is /Ii , Russian roulette is played and
the photon survives with probability Wp Ii / Wimin. Is and is given the weight Wimin. Is /Ii .

If weight windows are not used and if the weight of the starting neutrons is not unity, setting all the Wimin.
on the PWT card to negative values will make the photon minimum weight relative to the neutron source
weight. This will make the number of photons being created roughly proportional to the biased collision
rate of neutrons. It is recommended for most applications that negative numbers be used and be chosen to
produce from one to four photons per source neutron. The default values for Wimin. on the PWT card are −1,
which should be adequate for most problems using cell importances.

If energy-independent weight windows are used, the entries on the PWT card should be the same as on the
WWN 1:p card. If energy-dependent photon weight windows are used, the entries on the PWT card should be the
minimum WWN n:p entry for each cell, where n refers to the photon weight window energy group. This will
cause most photons to be born within the weight window bounds.

Any photons generated at neutron collision sites are temporarily stored in the bank. There are two methods
for determining the exiting energies and directions, depending on the form in which the processed photon
production data are stored in a library. The first method has the evaluated photon production data processed
into an “expanded format” [67]. In this format, energy-dependent cross sections, energy distributions, and
angular distributions are explicitly provided for every photon-producing neutron interaction. In the second
method, used with data processed from older evaluations, the evaluated photon production data have been
collapsed so that the only information about secondary photons is in a matrix of 20 equally probable photon
energies for each of 30 incident neutron energy groups. The sampling techniques used in each method are
now described.

LA-UR-22-30006, Rev. 1 73 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

2.4.3.3.1 Expanded Photon Production Method

In the expanded photon production method, the reaction n responsible for producing the photon is sampled
from
n−1
X N
X Xn
σi < ξ σi ≤ σi , (2.21)
i=1 i=1 i=1
where ξ is a random number on the interval [0, 1), N is the number of photon production reactions, and σi
is the photon production cross section for reaction i at the incident neutron energy. Note that there is no
correlation between the sampling of the type of photon production reaction and the sampling of the type of
neutron reaction described in §2.4.3.5.

Just as every neutron reaction (for example, (n, 2n)) has associated energy-dependent angular and energy
distributions for the secondary neutrons, every photon production reaction (for example, (n, pγ)) has associated
energy-dependent angular and energy distributions for the secondary photons. The photon distributions are
sampled in much the same manner as their counterpart neutron distributions.

All non-isotropic secondary photon angular distributions are represented by either 32 equiprobable cosine
bins or by a tabulated angular distribution. The distributions are given at a number of incident neutron
energies. All photon-scattering cosines are sampled in the laboratory system. The sampling procedure is
identical to that described for secondary neutrons in §2.4.3.5.1.

Secondary photon energy distributions are also a function of incident neutron energy. There are two
representations of secondary photon energy distributions allowed in ENDF-6 format: tabulated spectra and
discrete (line) photons. Correspondingly, there are two laws used in the MCNP code for the determination of
secondary photon energies. Law 4 provides for representation of a tabulated photon spectra possibly including
discrete lines. Law 2 is used solely for discrete photons. These laws are described in more detail beginning in
§2.4.3.5.4.1.

The expanded photon production method has clear advantages over the original 30 × 20 matrix method
[§2.4.3.3.2]. In coupled neutron/photon problems, users should attempt to specify data sets that contain
photon production data in expanded format. Such data sets are identified by “yes” entries in the GPD column
in [45]. However, it should be noted that the evaluations from which these data are processed may not include
all discrete lines of interest; evaluators may have binned sets of photons into average spectra that simply
preserve the energy distribution.

2.4.3.3.2 30 × 20 Photon Production Method

For lack of better terminology, we will refer to the photon production data contained in older libraries as
“30 × 20 photon production” data. In contrast to expanded photon production data, there is no information
about individual photon production reactions in the 30 × 20 data. This method is not used in modern tables
and is only included to maintain backwards compatibility for very old data libraries.

The only secondary photon data are a 30×20 matrix of photon energies; that is, for each of 30 incident neutron
energy groups there are 20 equally probable exiting photon energies. There is no information regarding
secondary photon angular distributions; therefore, all photons are taken to be produced isotropically in the
laboratory system.

There are several problems associated with 30 × 20 photon production data. The 30 × 20 matrix is an
inadequate representation of the actual spectrum of photons produced. In particular, discrete photon lines
are not well represented, and the high-energy tail of a photon continuum energy distribution is not well
sampled. Also, the multigroup representation is not consistent with the continuous-energy nature of the
MCNP code. Finally, not all photons should be produced isotropically. None of these problems exists for
data processed into the expanded photon production format.

LA-UR-22-30006, Rev. 1 74 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

2.4.3.4 Absorption

Absorption is treated in one of two ways: analog or implicit. Either way, the incident incoming neutron
energy does not include the relative velocity of the target nucleus from the free gas thermal treatment because
nonelastic reaction cross sections are assumed to be nearly independent of temperature. That is, only the
scattering cross section is affected by the free gas thermal treatment. The terms “absorption” and “capture”
are used interchangeably for non-fissile nuclides, both meaning (n, 0n). For fissile nuclides, “absorption”
includes both capture and fission reactions.

2.4.3.4.1 Analog Absorption

In analog absorption, the particle is killed with probability σa /σt , where σa and σt are the absorption and
total cross sections, respectively, of the collision nuclide at the incoming neutron energy. The absorption cross
section is specially defined for the MCNP code as the sum of all (n, x) cross sections, where x is anything
except neutrons. Thus σa is the sum of σn,g , σn,a , σn,d , σf , etc. For all particles killed by analog absorption,
the entire particle energy and weight are deposited in the collision cell.

2.4.3.4.2 Implicit Absorption

For implicit absorption, also called survival biasing, the neutron weight Wn is reduced to Wn0 as
 
0 σa
Wn = 1 − Wn . (2.22)
σt

If the new weight Wn0 is below the problem weight cutoff (specified on the CUT card), Russian roulette is
played, resulting overall in fewer particles with larger weight.

For implicit absorption, a fraction σa /σt of the incident particle weight and energy is deposited in the collision
cell corresponding to that portion of the particle that was absorbed. Implicit absorption is the default method
of neutron absorption in the MCNP code.

2.4.3.4.3 Implicit Absorption Along a Flight Path

Implicit absorption also can be done continuously along the flight path of a particle trajectory as is the
common practice in astrophysics. In this case, the distance to scatter, rather than the distance to collision, is
sampled. The distance to scatter is
1
l = − ln(1 − ξ). (2.23)
Σs
The particle weight at the scattering point is reduced to account for the expected absorption along the flight
path,
W 0 = W exp(−Σa l), (2.24)
where

W0 is the reduced weight after expected absorption along flight path,

W is the weight at the start of the flight path,
σa is the absorption cross section,

LA-UR-22-30006, Rev. 1 75 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

σs is the scattering cross section,

σt is the total cross section (σa + σs ),
l is the distance to scatter, and
ξ is a uniformly sampled random number.

Implicit absorption along a flight path is a special form of the exponential transformation coupled with
implicit absorption at collisions. See the description of the exponential transform in §5.12.7. The path length
is stretched in the direction of the particle, µ = 1, and the stretching parameter is p = Σa /Σt . Using these
values the exponential transform and implicit absorption at collisions yield the identical equations as does
implicit absorption along a flight path.

Implicit absorption along a flight path is invoked in the MCNP code as a special option of the exponential
transform variance reduction method. It is most useful in highly absorbing media, that is, Σa /Σt → 1. When
almost every collision results in absorption, it is very inefficient to sample distance to collision. However,
implicit absorption along a flight path is discouraged. In highly absorbing media, there is usually a superior
set of exponential transform parameters. In relatively non-absorbing media, it is better to sample the distance
to collision than the distance to scatter.

2.4.3.5 Elastic and Inelastic Scattering

If the conditions for the S(α, β) treatment are not met, the particle undergoes either an elastic or inelastic
collision. The selection of an elastic collision is made with the probability
σel σel
= , (2.25)
σin + σel σt − σa
where

σel is the elastic scattering cross section.

σin is the inelastic cross section, including any neutron-out process such as (n,n’), (n,f), (n,np), etc.
σa is the absorption cross section; Σa (n, x), where x 6= n , that is, all neutron disappearing reactions.
σt is the total cross section, σt = σel + σin + σa .

Both σel and σt are adjusted for the free gas thermal treatment at thermal energies.

The selection of an inelastic collision is made with the remaining probability,

σin
σt − σa
If the collision is determined to be inelastic, the type of inelastic reaction, n, is sampled from
n−1
X N
X n
X
σi < ξ σi ≤ σi , (2.26)
i=1 i=1 i=1

where ξ is a random number on the interval [0, 1), N is the number of inelastic reactions, and σi is the ith
inelastic reaction cross section at the incident neutron energy.

LA-UR-22-30006, Rev. 1 76 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

Directions and energies of all outgoing particles from neutron collisions are determined by sampling data from
the appropriate cross-section table. Angular distributions are provided and sampled for scattered neutrons
resulting from either elastic or discrete-level inelastic events; the scattered neutron energy is then calculated
from two-body kinematics. For other reaction types, a variety of data representations is possible. These
representations may be divided into two types: (a) the outgoing energy and outgoing angle are sampled
independently of each other, or (b) the outgoing energy and outgoing angle are correlated. In the latter case,
the outgoing energy may be specified as a function of the sampled outgoing angle, or the outgoing angle may
be specified as a function of the sampled outgoing energy. Details of the possible data representations and
sampling schemes are provided in the following sections.

2.4.3.5.1 Sampling of Angular and Energy Distributions

The cosine of the angle between incident and exiting particle directions, µ, is sampled from angular distribution
tables in the collision nuclide’s cross-section library. The cosines are either in the center-of-mass or target-at-
rest system, depending on the type of reaction. Data are provided at a number of incident neutron energies.
If E is the incident neutron energy, if En is the energy of table n, and if En+1 is the energy of table n + 1,
then a value of µ is sampled from table n + 1 with probability (E − En )/(En+1 − En ) and from table n
with probability (En+1 − E)/(En+1 − En ). There are two options in the MCNP code for representing and
sampling a non-isotropic scattering cosine. The first method involves the use of 32 equally probable cosine
bins. The second method is to sample a tabulated distribution as a function of µ.

When the method with 32 equiprobable cosine bins is employed, a random number ξ on the interval [0, 1) is
used to select the ith cosine bin such that I = 32 + 1. The value of µ is then computed as

µ = µi + (32ξ − i)(µi+1 − µi ). (2.27)

The method of 32 equiprobable cosine bins accurately represents high-probability regions of the scattering
probability; however, it can be a very crude approximation in low-probability regions. For example, it
accurately represents the forward scattering in a highly forward-peaked distribution, but may represent all
the back angle scattering using only one or a few bins.

A new, more rigorous angular distribution representation was implemented in MCNP4C. This new repre-
sentation features a tabulation of the probability density function (PDF) as a function of the cosine of the
scattering angle. Interpolation of the PDF between cosine values may be either by histogram or linear-linear
interpolation. The new tabulated angular distribution allows more accurate representations of original
evaluated distributions (typically given as a set of Legendre polynomials) in both high-probability and
low-probability regions.

Tabular angular distributions are equivalent to tabular energy distribution (as defined using ENDF Law 4)
except that the sampled value is the cosine of the scattering angle, and discrete lines are not allowed. For
each incident neutron energy Ei there is a pointer to a table of cosines µi,k , probability density functions pi,k ,
and cumulative density functions ci,k . The index i and the interpolation fraction r are found on the incident
energy grid for the incident energy Ein such that

Ei < Ein < Ei+1 (2.28)

and
Ein = Ei + r(Ei+1 − Ei ). (2.29)
A random number, ξ1 , on the unit interval [0, 1) is used to sample a cosine bin k from the cumulative density
function
cl,k < ξ1 < cl,k+1 , (2.30)

LA-UR-22-30006, Rev. 1 77 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

where l = i if ξ2 > r and l = i + 1 if ξ2 < r, and ξ2 is a random number on the unit interval. For histogram
interpolation the sampled cosine is
ξ1 − cl,k
µ0 = µl,k + . (2.31)
pl.k
For linear-linear interpolation the sampled cosine is
r h i 

 pl,k+1 −pl,k 
 Pl,k + 2 µl,k+1 −µl,k (ξ1 − cl,k ) − pl,k 
2

0
µ = µl,k + h i (2.32)

 pl,k+1 −pl,k 

 µl,k+1 −µl,k 

If the emitted angular distribution for some incident neutron energy is isotropic, µ is chosen from µ = ξ 0 ,
where ξ 0 is a random number on the interval [−1, 1). Strictly, in the MCNP code, random numbers are always
furnished on the interval (0, 1). Thus, to compute ξ 0 on (−1, 1) we calculate ξ 0 = 2ξ − 1, where ξ is a random
number on (0, 1).

The ENDF-6 format also has various formalisms to describe correlated secondary energy-angle spectra. These
are discussed later in this chapter.

For elastic scattering, inelastic level scattering, and some ENDF-6 inelastic reactions, the scattering cosine is
chosen in the center-of-mass system. Conversion must then be made to µlab , the cosine in the target-at-rest
system. For other inelastic reactions, the scattering cosine is sampled directly in the target-at-rest system.

The incident particle direction cosines (uo , vo , wo ) are rotated to new outgoing target-at-rest system cosines
(u, v, w) through a polar angle whose cosine is µlab , and through an azimuthal angle sampled uniformly. For
random numbers ξ1 and ξ2 on the interval [−1, 1) with rejection criterion ξ1 ξ2 ≤ 1, the rotation scheme is
[page 54 of 20]
p
1 − µ2lab (ξ1 uo wo − ξ2 vo )
u = uo µlab + p , (2.33a)
(ξ12 + ξ22 )(1 − wo2 )
p
1 − µ2lab (ξ1 vo wo + ξ2 uo )
v = vo µlab + p , (2.33b)
(ξ12 + ξ22 )(1 − wo2 )
p
ξ1 (1 − µ2lab )(1 − wo2 )
w = wo µlab − p . (2.33c)
(ξ12 + ξ22 )

If 1 − wo2 ∼ 0, then
p
1 − µ2lab (ξ1 uo vo + ξ2 wo )
u = uo µlab + p , (2.34a)
(ξ12 + ξ22 )(1 − vo2 )
p
ξ1 (1 − µ2lab )(1 − vo2 )
v = vo µlab − p , (2.34b)
(ξ12 + ξ22 )
p
1 − µ2lab (ξ1 wo vo − ξ2 uo )
w = wo µlab + p . (2.34c)
(ξ12 + ξ22 )(1 − vo2 )

If the scattering distribution is isotropic in the target-at-rest system, it is possible to use an even simpler
formulation that takes advantage of the exiting direction cosines, (u, v, w), being independent of the incident

LA-UR-22-30006, Rev. 1 78 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

direction cosines, (uo , vo , wo ). In this case,

u = 2ξ12 + 2ξ22 − 1, (2.35a)

s
1 − u2
v = ξ1 , (2.35b)
ξ12 + ξ22
s
1 − u2
w = ξ2 , (2.35c)
ξ12 + ξ22

where ξ1 and ξ2 are rejected if ξ12 + ξ22 > 1.

2.4.3.5.2 Energy from Elastic Scattering

Once the particle direction is sampled from the appropriate angular distribution tables, then the exiting
energy, Eout , is dictated by two-body kinematics,
1
Eout = Ein [(1 − α)µcm + 1 + α]
2 " #
1 + A2 + 2Aµcm
= Ein 2 , (2.36)
(1 + A)

where Ein is the incident neutron energy, µcm is the center-of-mass cosine of the angle between incident and
exiting particle directions,
 2
A−1
α= , (2.37)
A+1

and A is the mass of nuclide nucleus in units of the mass of a neutron (atomic weight ratio).

2.4.3.5.3 Inelastic Reactions

The treatment of inelastic scattering depends upon the particular inelastic reaction chosen. Inelastic reactions
are defined as (n, y) reactions such as (n, n0 ), (n, 2n), (n, f), (n, n0 α) in which y includes at least one neutron.

For many inelastic reactions, such as (n, 2n), more than one neutron can be emitted for each incident neutron.
The weight of each exiting particle is always the same as the weight of the incident particle minus any
implicit capture. The energy of exiting particles is governed by various scattering laws that are sampled
independently from the cross-section files for each exiting particle. Which law is used is prescribed by the
particular cross-section evaluation used. In fact, more than one law can be specified, and the particular one
used at a particular time is decided with a random number. In an (n, 2n) reaction, for example, the first
particle emitted may have an energy sampled from one or more laws, but the second particle emitted may
have an energy sampled from one or more different laws, depending upon specifications in the nuclear data
library. Because emerging energy and scattering angle is sampled independently for each particle, there is no
correlation between the emerging particles. Hence energy is not conserved in an individual reaction because,
for example, a 14-MeV particle could conceivably produce two 12-MeV particles in a single reaction. The net
effect of many particle histories is unbiased because on the average the correct amount of energy is emitted.
Results are biased only when quantities that depend upon the correlation between the emerging particles are
being estimated.

Users should note that the MCNP code follows a very particular convention. The exiting particle energy
and direction are always given in the target-at-rest (laboratory) coordinate system. For the kinematical

LA-UR-22-30006, Rev. 1 79 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

calculations in the MCNP code to be performed correctly, the angular distributions for elastic, discrete
inelastic level scattering, and some ENDF-6 inelastic reactions must be given in the center-of-mass system,
and the angular distributions for all other reactions must be given in the target-at-rest system. The MCNP
code does not stop if this convention is not adhered to, but the results will be erroneous. In the checking of
the cross-section libraries prepared for the MCNP code at Los Alamos, however, careful attention has been
paid to ensure that these conventions are followed.

The exiting particle energy and direction in the target-at-rest (laboratory) coordinate system are related to
the center-of-mass energy and direction as [19]:

" p #
E + 2µcm (A + 1) 0
EEcm
0 0
E = Ecm + 2 (2.38)
(A + 1)
and r r
0
Ecm 1 E
µlab = µcm + , (2.39)
E0 A+1 E0
where

E0 is the exiting particle energy (laboratory),

E0cm is the exiting particle energy (center-of-mass),
E is the incident particle energy (laboratory),
µcm is the cosine of center-of-mass scattering angle,

µlab is the cosine of laboratory scattering angle, and

A is the atomic weight ratio (mass of nucleus divided by mass of incident particle).

For point detectors it is necessary to convert

dµcm
p(µlab ) = p(µcm ) (2.40)
dµlab
where s s
E0 1 E
µcm = µlab 0
− 0
(2.41)
Ecm A+1 Ecm
and
dµcm E 0 /Ecm
0
=q q
dµlab E0
0
µcm
− A+1 E
Ecm E0 cm
q
E0
0
Ecm
= q . (2.42)
µcm E
1− A+1 E0

LA-UR-22-30006, Rev. 1 80 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

2.4.3.5.4 Non-fission Inelastic Scattering and Emission Laws

Non-fission inelastic reactions are handled differently from fission inelastic reactions. For each non-fission
reaction Np particles are emitted, where Np is an integer quantity specified for each reaction in the cross-
section data library of the collision nuclide. The direction of each emitted particle is independently sampled
from the appropriate angular distribution table, as was described earlier. The energy of each emitted particle
is independently sampled from one of the following scattering or emission laws. Energy and angle are
correlated only for ENDF-6 Laws 44 and 67. For completeness and convenience, all the laws are listed
together, regardless of whether the law is appropriate for non-fission inelastic scattering (for example, Law 3),
fission spectra (for example, Law 11), both (for example, Law 9), or neutron-induced photon production (for
example, Law 2). The conversion from center-of-mass to target-at-rest (laboratory) coordinate systems is
given in the above equations.

2.4.3.5.4.1 Law 1 (ENDF Law 1): Equiprobable Energy Bins

The index i and the interpolation fraction r are found on the incident energy grid for the incident energy Ein
such that
Ei < Ein < Ei+1 (2.43)
and
Ein = Ei + r(Ei+1 − Ei ). (2.44)
A random number on the unit interval ξ1 is used to select an equiprobable energy bin k from the K
equiprobable outgoing energies Ei,k where
k = ξi K + 1. (2.45)

Then scaled interpolation is used with random numbers ξ2 and ξ3 on the unit interval. Let
E1 = Ei,1 + r(Ei+1,1 − Ei,1 ) (2.46)
and
EK = Ei,K + r(Ei+1,K − Ei,K ) (2.47)
and (
i ξ3 > r
l= (2.48)
i + 1 ξ3 < r
and
E 0 = El,k + ξ2 (El,k+1 − El.k ) (2.49)
then
(E 0 − El,1 )(EK − E1 )
Eout = E1 + . (2.50)
El,k − El,1

2.4.3.5.4.2 Law 2: Discrete Photon Energy

The value provided in the library is Eg . The secondary photon energy Eout is either
Eout = Eg (2.51)
for non-primary photons or
Eout = Eg + [A/(A + 1)]Ein (2.52)
for primary photons, where A is the atomic weight to neutron weight ratio of the target and Ein is the
incident neutron energy.

LA-UR-22-30006, Rev. 1 81 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

2.4.3.5.4.3 Law 3 (ENDF Law 3): Inelastic Scattering (n, n0 ) From Nuclear Levels

The value provided in the library is Q and as a result

 2  
A Q(A + 1)
Eout = Ein − . (2.53)
A+1 A

2.4.3.5.4.4 Law 4 (ENDF Law 4): Tabular Distribution

For each incident neutron energy Ei there is a pointer to a table of secondary energies Ei,k , probability
density functions pi,k , and cumulative density functions ci,k . The index i and the interpolation fraction r are
found on the incident energy grid for the incident energy Ein such that

Ei < Ein < Ei+1 (2.54)

and
Ein = Ei + r(Ei+1 − Ei ). (2.55)

The tabular distribution at each Ei may be composed of discrete lines, a continuous spectra, or a mixture
of discrete lines superimposed on a continuous background. If discrete lines are present, there must be the
same number of lines (given one per bin) in each table. The sampling of the emission energy for the discrete
lines (if present) is handled separately from the sampling for the continuous spectrum (if present). A random
number, ξ1 , on the unit interval [0, 1) is used to sample a second energy bin k from the cumulative density
function.

If discrete lines are present, the algorithm first checks to see if the sampled bin is within the discrete line
portion of the table as determined by

ci,k + r(ci+1,k − ci,k ) < ξ1 < ci,k+1 + r(ci+1,k+1 − ci,k+1 ).

If this condition is met, then the sampled energy E 0 for the discrete line is interpolated between incident
energy grids as
E 0 = Ei,k + r(Ei+1,k − Ei,k ). (2.56)

If a discrete line has been sampled, the energy sampling is finished. If a discrete line has not been sampled,
the secondary energy is sampled from the remaining continuous background.

For continuous distributions, the secondary energy bin k is sampled from

cl.k < ξ1 < cl.k+1 , (2.57)

where l = i if ξ2 > r and l = i + 1 if ξ2 < r, and ξ2 is a random number on the unit interval. For histogram
interpolation the sampled energy is
ξ1 − cl.k
E 0 = El,k + . (2.58)
pl,k

For linear-linear interpolation the sampled energy is

LA-UR-22-30006, Rev. 1 82 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

r h i 
 2 + 2 pl,k+1 −pl,k (ξ − c ) − p
 Pl,k 

 El,k+1 −El,k 1 l,k l,k 
0
E = El,k + h i . (2.59)

 pl,k+1 −pl,k 

 El,k+1 −El,k 

The secondary energy is then interpolated between the incident energy bins i and i + 1 to properly preserve
thresholds. Let
E1 = Ei,1 + r(Ei+1,1 − Ei,1 ) (2.60)
and
EK = Ei,K + r(Ei+1,K − Ei,K ) (2.61)
then
(E 0 − El,1 )(EK − E1 )
Eout = E1 + . (2.62)
(El,K − El,1 )

The final step is to adjust the energy from the center-of-mass system to the laboratory system, if the energies
were given in the center-of-mass system.

Law 4 is an independent distribution, i.e. the emission energy and angle are not correlated. The outgoing angle
is selected from the angular distribution as described in §2.4.3.5.1. Data tables built using this methodology
are designed to sample the distribution correctly in a statistical sense and will not necessarily sample the
exact distribution for any specific collision.

2.4.3.5.4.5 Law 5 (ENDF Law 5): General Evaporation Spectrum

The function g(x) is tabulated versus χ and the energy is tabulated versus incident energy Ein . The law is
then  
Eout
f (Ein → Eout ) = g . (2.63)
T (Ein )
This density function is sampled by Eout = χ(ξ)T (Ein ), where T (Ein ) is a tabulated function of the incident
energy and c(ξ) is a table of equiprobable χ values.

2.4.3.5.4.6 Law 7 (ENDF Law 7): Simple Maxwell Fission Spectrum

The law is  
p Eout
f (Ein → Eout ) = C Eout exp − . (2.64)
T (Ein )

The nuclear temperature T (Ein ) is a tabulated function of the incident energy. The normalization constant
C is given by " √  ! r
r  #
π E in − U Ein − U E in − U
C −1 = T /2 (2.65)
3
erf − exp − ,
2 T T T
where U is a constant provided in the library and limits Eout to 0 ≤ Eout ≤ E − U . In the MCNP code this
density function is sampled by the rejection scheme
 2 
ξ1 ln(ξ3 )
Eout = −T (Ein ) 2 + ln(ξ4 ) , (2.66)
ξ1 + ξ22

where ξ1 , ξ2 , ξ3 , and ξ4 are random numbers on the unit interval. ξ1 and ξ2 are rejected if ξ12 + ξ22 > 1.

LA-UR-22-30006, Rev. 1 83 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

2.4.3.5.4.7 Law 9 (ENDF Law 9): Evaporation Spectrum

The law is  
Eout
f (Ein → Eout ) = CEout exp − , (2.67)
T (Ein )
where the nuclear temperature T (Ein ) is a tabulated function of the incident energy. The energy U is provided
in the library and is assigned so that Eout is limited by 0 ≤ Eout ≤ Ein − U . The normalization constant C
is given by    
Ein − U Ein − U
C −1 = T 2 1 − exp − 1+ . (2.68)
T T

In the MCNP code this density function is sampled by

Eout = −T (Ein ) ln(ξ1 ξ2 ), (2.69)
where ξ1 and ξ2 are random numbers on the unit interval, and ξ1 and ξ2 are rejected if Eout > Ein − U .

2.4.3.5.4.8 Law 11 (ENDF Law 11): Energy Dependent Watt Spectrum

The law is   p 
Eout
f (Ein → Eout ) = C exp − sinh b(Ein )Eout . (2.70)
a(Ein )
The constants a and b are tabulated functions of incident energy and U is a constant from the library. The
normalization constant C is given by
r
3b
 " r r ! r r !#
1 πa ab Ein − U ab E in − U ab
C −1 = exp erf − + erf +
2 4 4 a 4 a 4
  p 
Ein − U
− a exp − sinh b(Ein − U ) , (2.71)
a
where the constant U limits the range of outgoing energy so that 0 ≤ Eout ≤ Ein − U . This density function
is sampled as follows. Let s 2  
ab ab
g= 1+ −1+ 1+ . (2.72)
8 8
Then Eout = −ag ln(ξ1 ). Eout is rejected if
2
[(1 − g)(1 − ln(ξ1 )) − ln(ξ2 )] > bEout , (2.73)
where ξ1 and ξ2 are random numbers on the unit interval.

2.4.3.5.4.9 Law 22 (UK Law 2): Tabular Linear Functions of Incident Energy Out

Tables of Pi,j , Ci,j , and Ti,j are given at a number of incident energies Ei . If Ei ≤ Ein < Ei+1 then the ith
Pi,j , Ci,j , and Ti,j tables are used and
Eout = Ci,k (Ein − Ti,k ), (2.74)
where k is chosen according to
k
X k+1
X
Pi,j < ξ ≤ Pi,j ,
j=1 j=1

where ξ is a random number on the unit interval [0, 1).

LA-UR-22-30006, Rev. 1 84 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

2.4.3.5.4.10 Law 24 (UK Law 6): Equiprobable Energy Multipliers

The law is
Eout = Ein T (Ein ). (2.75)
The library provides a table of K equiprobable energy multipliers Ti,k for a grid of incident neutron energies
Ei . For incident energy Ein such that
Ei < Ein < Ei+1 .

The random numbers ξ1 and ξ2 on the unit interval are used to find T with

k = ξ1 K + 1 (2.76)

and
T = Ti,k + ξ2 (Ti,k+1 − Ti,k ) (2.77)
so
Eout = Ein T (2.78)

2.4.3.5.4.11 Law 44 Tabular Distribution (ENDF Law=1 Lang=2): Kalbach-87 Correlated

Energy-angle Scattering)

Law 44 is an extension of Law 4. For each incident energy Ei there is a pointer to a table of secondary energies
Ei,k , probability density functions pi,k , cumulative density functions ci,k , pre-compound fractions Ri,k , and
angular distribution slope values Ai,k . The secondary emission energy is found exactly as stated in the Law 4
description in §2.4.3.5.4.4. Unlike Law 4, Law 44 includes a correlated angular distribution associated with
each incident energy Ei as given by the Kalbach parameters Ri,k , and Ai,k . Thus, the sampled emission
angle is dependent on the sampled emission energy.

The sampled values for R and A are interpolated on both the incident and outgoing energy grids. For discrete
spectra,
A = Ai,k + r(Ai+1,k − Ai,k ) (2.79)
and
R = Ri,k + r(Ri+1,k − Ri,k ). (2.80)

For continuous spectra with histogram interpolation,

A = Ai,k (2.81)

and
R = Ri,k (2.82)

For continuous spectra with linear-linear interpolation,

E 0 − El,k
A = Al,k + (Al,k+1 − Al,k ) (2.83)
El,k+1 − El,k

and
E 0 − El,k
R = Rl,k + (Rl,k+1 − Rl,k ) . (2.84)
El,k+1 − El,k

LA-UR-22-30006, Rev. 1 85 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

The outgoing neutron center-of-mass scattering angle µ is sampled from the Kalbach density function
1 A
p(µ, Ein , Eout ) = [cosh(Aµ) + R sinh(Aµ)] (2.85)
2 sinh(A)

using the random numbers ξ3 and ξ4 on the unit interval as follows. If ξ3 > R, then let
T = (2ξ4 − 1) sinh(A) (2.86)
and √

ln T + T2 + 1
µ= , (2.87)
A
or if ξ3 < R, then
ln[ξ4 exp(A) + (1 − ξ4 ) exp(−A)]
µ= . (2.88)
A
As with Law 4, the emission energy and angle are transformed from the center-of-mass to the laboratory
system as necessary.

2.4.3.5.4.12 Law 61 Tabular Distribution (ENDF Law=1 Lang=0, 12, or 14): Correlated
Energy-angle Scattering

Law 61 is an extension of Law 4. For each incident energy Ei there is a pointer to a table of secondary
energies Ei,k , probability density functions pi,k , cumulative density functions ci,k , and pointers to tabulated
angular distributions Li,k . The secondary emission energy is found exactly as stated in the Law 4 description
in §2.4.3.5.4.4. Unlike Law 4, Law 61 includes a correlated angular distribution associated with each incident
energy Ei as given by the tabular angular distribution located using the pointers Li,k . Thus, the sampled
emission angle is dependent on the sampled emission energy.

If the secondary distribution is given using histogram interpolation, the angular distribution located at Li,k
is used to sample the emission angle. If the secondary distribution is specified as linear interpolation between
energy points, Li,k is chosen by selecting the bin closest to the randomly sampled cumulative distribution
function (CDF) point. If the value of Li,k is zero, the angle is sampled from an isotropic distribution as
described on page 78. If the value of Li,k is positive, it points to a tabular angular distribution which is then
sampled as described on page 78.

As with Law 4, the emission energy and angle are transformed from the center-of-mass to the laboratory
system as necessary.

2.4.3.5.4.13 Law 66 (ENDF Law 6): N-body Phase Space Distribution

The phase space distribution for particle i in the center-of-mass coordinate system is:
√ 3n/2−4
Pi (µ, Ein , T ) = Cn T (Eimax − T ) , (2.89)
where all energies and angles are also in the center-of-mass system and is the maximum possible energy
Eimax
for particle i, µ, and T . T is used for calculating Eout . The Cn normalization constants for n = 3, 4, 5 are:
4
C3 = 2, (2.90a)
π(Eimax )
105
C4 = 7/2
, (2.90b)
32(Eimax )
256
C5 = 5. (2.90c)
14π(Eimax )

LA-UR-22-30006, Rev. 1 86 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

Eimax is a fraction of the energy available, Eavail , given as

M − mi
Eimax = Eavail , (2.91)
M
where M is the total mass of the n particles being treated, mi is the mass of particle i, and
mt
Eavail = Ein + Q, (2.92)
mp + mt

where mt is the target mass and mp is the projectile mass. For neutrons,
mt A
= (2.93)
mp + mt A+1
and for a total mass ratio Ap = M/mi ,
M − mi Ap − 1
= . (2.94)
M Ap
Thus,  
Ap − 1 A
Eimax = Ein + Q . (2.95)
Ap A+1

The total mass Ap and the number of particles in the reaction n are provided in the data library. The
outgoing energy is sampled as follows.

Let ξi , i = 1, 10 be random numbers on the unit interval. Then from rejection technique R28 from the Monte
Carlo Sampler [68], accept ξ1 and ξ2 if ξ12 + ξ22 ≤ 1 and accept ξ3 and ξ4 if ξ32 + ξ42 ≤ 1.

Then let


−ξ1 ln ξ12 + ξ22
x= − ln(ξ9 ) (2.96)
ξ12 + ξ22
 −ξ ln ξ2 +ξ2
 3 ( 3 4)

 ξ32 +ξ42
− ln(ξ5 ) n=3
y = − ln(ξ5 ξ6 ξ7 ) n=4 (2.97)


 −ξ3 ln(ξ32 +ξ42 )
ξ 2 +ξ 2
− ln(ξ5 ξ6 ξ7 ξ8 ) n = 5
3 4

and
x
T = . (2.98)
x+y
Then
Eout = T Eimax (2.99)
The cosine of the scattering angle is always sampled isotropically in the center-of-mass system using another
random number ξ10 on the unit interval as

µ = 2ξ10 − 1. (2.100)

2.4.3.5.4.14 Law 67 (ENDF Law 7): Correlated Energy-angle Scattering

For each incident neutron energy, first the exiting particle direction µ is sampled as described in §2.4.3.5.1.
In other law data, first the exiting particle energy is sampled and then the angle is sampled. The index i and
the interpolation fraction r are found on the incident energy grid for the incident energy Ein such that

Ei < Ein < Ei+1

LA-UR-22-30006, Rev. 1 87 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

and
Ein = Ei + r(Ei+1 − Ei ). (2.101)
For each incident energy Ei there is a table of exiting particle direction cosines µi,j and locators Li,j . This
table is searched to find which ones bracket µ, namely,

µi,j < µ < µi,j+1 . (2.102)

Then the secondary energy tables at Li,j and Li,j+1 are sampled for the outgoing particle energy. The
secondary energy tables consist of a secondary energy grid Ei,j,k , probability density functions pi,j,k , and
cumulative density functions ci,j,k . A random number ξ1 on the unit interval is used to pick between incident
energy indices: if ξ1 < r then l = i + 1; otherwise, l = i. Two more random numbers ξ2 and ξ3 on the unit
interval are used to determine interpolation energies. As such,
( µ−µ1,j
Ei,j+1,k , m = j + 1 ξ2 < µ1,j+1 −µ ,l=i
Ei,k = µ−µ1,j
i,j
. (2.103)
Ei,j,k , m = j ξ2 ≥ µ1,j+1 −µi,j , l = i

Similarly, ( µ−µi+1,j
Ei+1,j+1,k , m = j + 1 ξ3 < µi+1,j+1 −µi+1,j , l =i+1
Ei+1,k = µ−µi+1,j . (2.104)
Ei+1,j,k , m = j ξ3 ≥ µi+1,j+1 −µi+1,j , l =i+1

A random number ξ4 on the unit interval is used to sample a secondary energy bin k from the cumulative
density function
cl,m,k < ξ4 < cl,m,k+1 . (2.105)
For histogram interpolation the sampled energy is
ξ4 − cl,m,k
E 0 = El,m,k + . (2.106)
pl,m,k

For linear-linear interpolation the sampled energy is

r h i 

 pl,m,k+1 −pl,m,k 
 Pl,m,k + 2 El,m,k+1 −El,m,k (ξ4 − cl,m,k ) − pl,m,k 
2

0
E = El,m,k + h i . (2.107)

 pl,m,k+1 −pl,m,k 

 El,m,k+1 −El,m,k 

The final outgoing energy Eout uses scaled interpolation. Let

E1 = Ei,1 + r(Ei+1,1 − Ei,1 ) (2.108)

and
EK = Ei,K + r(Ei+1,K − Ei,K ) (2.109)
then
(E 0 − El,1 )(EK − E1 )
Eout = E1 + . (2.110)
El,K − El,1

2.4.3.5.5 Emission from Fission

For any fission reaction a number of neutrons, n, is emitted according to the value of ν(Ein ). Depending on
the type of problem (fixed source or KCODE ) and on user input ( TOTNU card), the MCNP code may use either

LA-UR-22-30006, Rev. 1 88 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

prompt ν p (Ein ) or total ν t (Ein ). For either case, the average number of neutrons per fission, ν(Ein ), may be
a tabulated function of energy or a polynomial function of energy.

If DATA entry on the FMULT card is zero (default), then n is sampled between I (the largest integer less than
ν(Ein )) and I + 1 by (
I + 1 ξ ≤ ν(Ein ) − 1
n= , (2.111)
I ξ > ν(Ein ) − 1
where ξ is a random number drawn from the unit interval.

If more microscopically correct fission neutron multiplicities are desired for fixed source problems, the DATA
entry on the FMULT card can be used to select which set of Gaussian widths are used to sample the actual
number of neutrons from fission that typically range from 0 to 7 or 8 [69]. For a given fission event, there is a
probability Pn that n neutrons are emitted. This distribution is generally called the neutron multiplicity
distribution. Fission neutron multiplicity distributions are known to be well reproduced by simple Gaussian
distributions [70],
ˆ1/2 !
2
1 (x − ν + b)
P0 = √ exp − dx, (2.112)
2πσ 2 2σ 2
−∞

and
ˆ 1/2
n+ !
2
1 (x − ν + b)
Pn6=0 = √ exp − dx, (2.113)
2πσ 2 2σ 2
n−1/2

where ν is the mean multiplicity, b is a small adjustment to make the mean equal to ν, and σ is the Gaussian
width. For the range of realistic widths, the adjustment b can be accurately expressed as a single smooth
function of (ν + 0.5)/σ [2]. To determine the value of σ from experimental data, many authors have minimized
the chi-squared
X P exp − Pn (σ) 2
2
χ (σ) = n
, (2.114)
n
∆Pnexp
where ∆Pnexp is the uncertainty in the experimentally measured
P multiplicity distribution Pnexp . The factorial
moments of the neutron multiplicity distribution (ν i = Pn n!/(n − i)!) emitted by a multiplying sample
can be expressed as a function of the factorial moments for spontaneous and induced fission [71]. Therefore,
for many applications it is not necessary to know the details of the neutron multiplicity distribution, but it
is more important to know the corresponding first three factorial moments. A reevaluation of the existing
spontaneous fission and neutron induced fission data has been performed [2] where the widths of Gaussians are
adjusted to fit the measured second and third factorial moments. This reevaluation was done by minimizing
the chi-squared
X3  2
νi (Pnexp ) − νi (Pn (σ))
χ2 (σ) = . (2.115)
i=2
∆νiexp
These results are summarized in Table 2.1. Despite the change in emphasis from the detailed shape to the
moments of the distributions, the inferred widths are little changed from those obtained by others. However,
by minimizing the chi-squared in Eq. (2.115) the inferred widths are guaranteed to be in reasonable agreement
with the measured second and third factorial moments. The widths obtained using Eq. (2.115) give Gaussian
distributions that reproduce the experimental second and third factorial moments to better than 0.6%. The
adjustment parameter b ensures that the first moment (ν) is accurately reproduced. If the chi-squared in
Eq. (2.114) is used, then the second and third factorial moments can differ from the experimental values by
as much as 10%.

Assuming that the widths of the multiplicity distributions are independent of the initial excitation energy
of the fissioning system [2], the relationship between different factorial moments is easily calculated as a
function of ν. The corresponding calculated relationships between the first three factorial moments are in

LA-UR-22-30006, Rev. 1 89 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

Table 2.1: Recommended Gaussian Widths [2] from Eq. (2.115)

Reaction σ
233
U(n,f) 1.070
235
U(n,f) 1.088
238
U(n,f) 1.116
239
Pu(n,f) 1.140
241
Pu(n,f) 1.150
238
Pu SF∗ 1.135
240
Pu SF 1.151
242
Pu SF 1.161
242
Cm SF 1.091
244
Cm SF 1.103
246
Cm SF 1.098
248
Cm SF 1.108
250
Cf SF 1.220
252
Cf SF 1.245
254
Cf SF 1.215
254
Fm SF 1.246
∗
SF: Spontaneous Fission

good agreement with experimental neutron induced fission data up to an incoming neutron energy of 10 MeV
[2]. This implies that an energy independent width can be used with confidence up to an incoming neutron
energy of at least 10 MeV. The Gaussian widths in Table 2.1 are used for fission multiplicity sampling in the
MCNP code when the DATA entry on the FMULT card is 1. Induced fission multiplicities for isotopes not listed
in Table 2.1 use a Gaussian width that is linearly dependent on the mass number of the fissioning system [2].

The direction of each emitted neutron is sampled independently from the appropriate angular distribution
table by the procedure described in §2.4.3.5.1.

The energy of each fission neutron is determined from the appropriate emission law. These laws are discussed
in the preceding section. The MCNP code then models the transport of the first neutron out after storing all
other neutrons in the bank.

2.4.3.5.6 Prompt and Delayed Neutron Emission

If (1) the MCNP code is using ν t , (2) the data for the collision isotope includes delayed-neutron spectra, and
(3) the use of detailed delayed-neutron data has not been preempted (on the PHYS :n card), then each fission
neutron is first determined by the MCNP code to be either a prompt fission neutron or a delayed fission
neutron. Assuming analog sampling, the type of emitted neutron is determined from the ratio of delayed
ν(Ein ) to total ν t (Ein ) where a delayed neutron is produced if

ν d (Ein )
ξ≤ (2.116)
ν t (Ein )

and a prompt neutron is produced if

ν d (Ein )
ξ> , (2.117)
ν t (Ein )
where ν d is the expected number of delayed neutrons.

If the neutron is determined to be a prompt fission neutron, it is emitted instantaneously, and the emission
laws (angle and energy) specified for prompt fission are sampled.

LA-UR-22-30006, Rev. 1 90 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

If the neutron is determined to be a delayed fission neutron, then the MCNP code first samples for the decay
group by using the specified abundances. Then, the time delay is sampled from the exponential density with
decay constant specified for the sampled decay group.

Finally, the emission laws (angle and energy) specified for that decay group are then sampled. Since the
functionality in the MCNP code to produce delayed neutrons using appropriate emission data is new, we
include next a somewhat more expanded description.

A small but important fraction (~1%) of the neutrons emitted in fission events are delayed neutrons emitted
as a result of fission-product decay at times later than prompt fission neutrons. the MCNP code users have
always been able to specify whether or not to include delayed fission neutrons by using either ν t (prompt
plus delayed) or ν p (prompt only). However, in versions of the MCNP code up through and including 4B, all
fission neutrons (whether prompt or delayed) were produced instantaneously and with an energy sampled
from the spectra specified for prompt fission neutrons.

For many applications this approach is adequate. However, it is another example of a data approximation
that is unnecessary. Therefore, Versions 4C and later of the MCNP code allow delayed fission neutrons to be
sampled (either analog or biased) from time and energy spectra as specified in nuclear data evaluations. The
libraries with detailed delayed fission neutron data are listed in [45] with a “yes” in the “DN” column.

The explicit sampling of a delayed-neutron spectrum implemented in MCNP4C has two effects. One is that
the delayed neutron spectra have the correct energy distribution; they tend to be softer than the prompt
spectra. The second is that experiments measuring neutron decay after a pulsed source can now be modeled
with the MCNP code because the delay in neutron emission following fission is properly accounted for. In
this treatment, a natural sampling of prompt and delayed neutrons is implemented as the default and an
additional delayed neutron biasing control is available to the user via the PHYS :n card. The biasing allows the
number of delayed neutrons produced to be increased artificially because of the low probability of a delayed
neutron occurrence. The delayed neutron treatment is intended to be used with the TOTNU card in the MCNP
code, giving the user the flexibility to use the time-dependent treatment of delayed neutrons whenever the
delayed data are available.

The impact of sampling delayed-neutron energy spectra on reactivity calculations has been studied [72]. As
expected, most of the reactivity impacts are very small, although changes of 0.1-0.2% in keff were observed
for certain cases. Overall, inclusion of delayed-neutron spectra can be expected to produce small positive
reactivity changes for systems with significant fast neutron leakage and small negative changes for some
systems in which a significant fraction of the fissions occurs in isotopes with an effective fission threshold
(e.g., 238 U and 240 Pu).

2.4.3.6 The S(α, β) Treatment

The S(α, β) thermal scattering treatment is a complete representation of thermal neutron scattering by
molecules and crystalline solids. Two processes are allowed: (1) inelastic scattering with cross section σin and
a coupled energy-angle representation derived from an ENDF S(α, β) scattering law, and (2) elastic scattering
with no change in the outgoing neutron energy for solids with cross section σel and an angular treatment
derived from lattice parameters. The elastic scattering treatment is chosen with probability σel /(σel + σin ).
This thermal scattering treatment also allows the representation of scattering by multi-atomic molecules (for
example, BeO).

For the inelastic treatment, the distribution of secondary energies is represented by a set of equally probable
final energies (typically 16 or 32) for each member of a grid of initial energies from an upper limit of typically
4 eV down to 10-5 eV, along with a set of angular data for each initial and final energy. The selection of a
final energy E 0 given an initial energy E can be characterized by sampling from the distribution
N
1 X
p(E 0 |Ei < ξ < Ei+1 ) = δ[E 0 − ρEi,j − (1 − ρ)Ei+1,j ], (2.118)
N j=1

LA-UR-22-30006, Rev. 1 91 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

where Ei and Ei+1 are adjacent elements on the initial energy grid,

Ei+1 − E
ρ= , (2.119)
Ei+1 − Ei

N is the number of equally probable final energies, and Ei,j is the j th discrete final energy for incident energy
Ei .

There are two allowed schemes for the selection of a scattering cosine following selection of a final energy and
final energy index j. In each case, the (i, j)th set of angular data is associated with the energy transition
E = Ei → E 0 = Ei,j .

In the first scheme, the data consist of sets of equally probable discrete cosines µi,j,k for k = 1, . . . , ν with ν
typically 4 or 8. An index k is selected with probability 1/ν, and µ is obtained by the relation

µ = ρµi,j,k + (1 − ρ)µi+1,j,k . (2.120)

In the second scheme, the data consist of bin boundaries of equally probable cosine bins. In this case, random
linear interpolation is used to select one set or the other, with ρ being the probability of selecting the set
corresponding to incident energy Ei . The subsequent procedure consists of sampling for one of the equally
probable bins and then choosing µ uniformly in the bin.

For elastic scattering, the above two angular representations are allowed for data derived by an incoherent
approximation. In this case, one set of angular data appears for each incident energy and is used with the
interpolation procedures on incident energy described above. For elastic scattering, when the data have been
derived in the coherent approximation, a completely different representation occurs. In this case, the data
actually stored are the set of parameters Dk , where
(
Dk /E Ebk ≤ E < Ebk+1
σeI = (2.121)
0 E < EB1

and EBk are Bragg energies derived from the lattice parameters. For incident energy E such that EBk ≤
E ≤ EBk+1 ,
Pi = Di /Dk , i = 1, . . . , k (2.122)
represents a discrete cumulative probability distribution that is sampled to obtain index i, representing
scattering from the ith Bragg edge. The scattering cosine is then obtained from the relationship

µ = 1 − 2EBi /E. (2.123)

Using next-event estimators such as point detectors with S(α, β), scattering cannot be done exactly because
of the discrete scattering angles. The MCNP code uses an approximate scheme [73, 74] that in the next-event
estimation calculation replaces discrete lines with histograms of width δµ < 0.1.

See also §2.5.6.4.7.

2.4.3.7 Probability Tables for the Unresolved Resonance Range

Within the unresolved resonance range (e.g., in ENDF/B-VI, 2.25–25 keV for 235 U, 10–149.03 keV for 238 U,
and 2.5–30 keV for 239 Pu), continuous-energy neutron cross sections appear to be smooth functions of energy.
This behavior occurs not because of the absence of resonances, but rather because the resonances are so
close together that they are unresolved. Furthermore, the smoothly varying cross sections do not account

LA-UR-22-30006, Rev. 1 92 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

for resonance self-shielding effects, which may be significant for systems whose spectra peak in or near the
unresolved resonance range.

Fortunately, the resonance self-shielding effects can be represented accurately in terms of probabilities based
on a stratified sampling technique. This technique produces tables of probabilities for the cross sections in
the unresolved resonance range. Sampling the cross section in a random walk from these probability tables is
a valid physics approximation so long as the average energy loss in a single collision is much greater than the
average width of a resonance; that is, if the narrow resonance approximation [75] is valid. Then the detail in
the resonance structure following a collision is statistically independent of the magnitude of the cross sections
prior to the collision.

The utilization of probability tables is not a new idea in Monte Carlo applications. A code [76] to calculate
such tables for Monte Carlo fast reactor applications was utilized in the early 1970s. Temperature-difference
Monte Carlo calculations [77] and a summary of the VIM Monte Carlo code [78] that uses probability tables
are pertinent early examples. Versions of the MCNP code up through and including 4B did not take full
advantage of the unresolved resonance data provided by evaluators. Instead, smoothly varying average cross
sections were used in the unresolved range. As a result, any neutron self-shielding effects in this energy range
were unaccounted for. Better utilizations of unresolved data have been known and demonstrated for some
time, and the probability table treatment has been incorporated [79] into MCNP4C and its successors. The
column “UR” in [45] lists whether unresolved resonance probability table data is available for each nuclide
library.

Sampling cross sections from probability tables is straightforward. At each of a number of incident energies
there is a table of cumulative probabilities (typically 20) and the value of the near-total, elastic, fission, and
radiative capture cross sections and heat deposition numbers corresponding to those probabilities. These
data supplement the usual continuous data; if probability tables are turned off ( PHYS :n card), then the usual
smooth cross section is used. But if the probability tables are turned on (default), if they exist for the nuclide
of a collision, and if the energy of the collision is in the unresolved resonance energy range of the probability
tables, then the cross sections are sampled from the tables. The near-total is the total of the elastic, fission,
and radiative capture cross sections; it is not the total cross section, which may include other absorption or
inelastic scatter in addition to the near-total. The radiative capture cross section is not the same as the usual
MCNP capture cross section, which is more properly called “destruction” or absorption and includes not
only radiative capture but all other reactions not emitting a neutron. Sometimes the probability tables are
provided as factors (multipliers of the average or underlying smooth cross section) which adds computational
complexity but now includes any structure in the underlying smooth cross section.

It is essential to maintain correlations in the random walk when using probability tables to properly model
resonance self-shielding. Suppose we sample the 17th level (probability) from the table for a given collision.
This position in the probability table must be maintained for the neutron trajectory until the next collision,
regardless of particle splitting for variance reduction or surface crossings into various other materials whose
nuclides may or may not have probability table data. Correlation must also be retained in the unresolved
energy range when two or more cross-section sets for an isotope that utilize probability tables are at different
temperatures.

The impact of the probability-table approach has been studied [80] and found to have negligible impact for
most fast and thermal systems. Small but significant changes in reactivity may be observed for plutonium and
233
U systems, depending upon the detailed shape of the spectrum. However, the probability-table method
can produce substantial increases in reactivity for systems that include large amounts of 238 U and have
high fluxes within the unresolved resonance region. Calculations for such systems will produce significantly
nonconservative results unless the probability-table method is employed.

2.4.4 Photon Interactions

Sampling of a collision nuclide, analog capture, implicit capture, and many other aspects of photon interactions
such as variance reduction, are the same as for neutrons. The collision physics are completely different.

LA-UR-22-30006, Rev. 1 93 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

The MCNP code has two photon interaction models: simple and detailed.

The simple physics treatment ignores coherent (Thomson) scattering and fluorescent photons from photoelectric
absorption. It is intended for high-energy photon problems or problems where electrons are free and is also
important for next-event estimators such as point detectors, where scattering can be nearly straight ahead
with coherent scatter. The simple physics treatment uses implicit capture unless overridden with the CUT :p
card, in which case it uses analog capture.

The detailed physics treatment includes coherent (Thomson) scattering and accounts for fluorescent photons
after photoelectric absorption. Form factors and Compton profiles are used to account for electron binding
effects. Analog capture is always used. The detailed physics treatment is used below energy EMCPF on the
PHYS :p card, and because the default value of EMCPF is 100 MeV, that means it is almost always used by
default. It is the best treatment for most applications, particularly for high-Z nuclides or deep penetration
problems.

The generation of electrons from photons is handled three ways. These three ways are the same for both the
simple and detailed photon physics treatments.

1. If electron transport is turned on ( MODE P E), then all photon collisions except coherent scatter can
create electrons that are banked for later transport.
2. If electron transport is turned off (no E on the MODE card), then a thick-target bremsstrahlung model
(TTB) is used. This model generates electrons, but assumes that they are locally slowed to rest. Any
bremsstrahlung photons produced by the non-transported electrons are then banked for later transport.
Thus electron-induced photons are not neglected, but the expensive electron transport step is omitted.
The TTB production model contains many approximations compared to models used in actual electron
transport. In particular, the bremsstrahlung photons inherit the direction of the parent electron.
3. If IDES = 1 on the PHYS :p card, then all electron production is turned off, no electron-induced photons
are created, and all electron energy is assumed to be locally deposited.

The TTB approximation is the default for MODE P problems. In MODE P E problems, it plays a role when
the energy cutoff for electrons is greater than that for photons. In this case, the TTB model is used in the
terminal processing of the electrons to account for the few low-energy bremsstrahlung photons that would be
produced at the end of the electrons’ range.

2.4.4.1 Simple Physics Treatment

The simple physics treatment is intended primarily for higher energy photons. It is inadequate for high-Z
nuclides or deep penetration problems. The physical processes treated are photoelectric effect, pair production,
Compton scattering from free electrons, and (optionally) photonuclear interactions (described in §2.4.4.3).
The photoelectric effect is regarded as an absorption (without fluorescence). The kinematics of Compton
scattering is assumed to be with free electrons (without the use of form factors or Compton profiles). The
total scattering cross section, however, includes the incoherent scattering factor regardless of the use of simple
or detailed physics. Thus, strict comparisons with codes using only the Klein-Nishina differential cross section
are not valid. Highly forward coherent Thomson scattering is ignored. Thus the total cross section σt is
regarded as the sum of three components:

σt = σpe + σpp + σs . (2.124)

LA-UR-22-30006, Rev. 1 94 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

2.4.4.1.1 Photoelectric Effect

This is treated as a pure absorption by implicit capture with a corresponding reduction in the photon weight
WGT, and hence does not result in the loss of a particle history except for Russian roulette played on the
weight cutoff. The non-captured weight W GT (1 − σpe /σs ) is then forced to undergo either pair production or
Compton scattering. The captured weight either is assumed to be locally deposited or becomes a photoelectron
for electron transport or for the TTB approximation.

2.4.4.1.2 Pair Production

In a collision resulting in pair production [probability σpp /(σt − σpe )], either an electron-positron pair is
created for further transport (or the TTB treatment) and the photon disappears, or it is assumed that the
kinetic energy W GT (E − 1.022) MeV of the electron-positron pair produced is deposited as thermal energy
at the time and point of collision, with isotropic production of one photon of energy 0.511 MeV headed in
one direction and another photon of energy 0.511 MeV headed in the opposite direction. The rare single
1.022-MeV annihilation photon is ignored. The relatively unimportant triplet production process is also
ignored. The simple physics treatment for pair production is the same as the detailed physics treatment that
is described in §2.4.4.2.4.

2.4.4.1.3 Compton Scattering

The alternative to pair production is Compton scattering on a free electron, with probability σs /(σt − σpe ).
In the event of such a collision, the objective is to determine the energy E 0 of the scattered photon, and
µ = cos(θ) for the angle θ of deflection from the line of flight. This yields at once the energy W GT (E–E 0 )
deposited at the point of collision and the new direction of the scattered photon. The energy deposited at
the point of collision can then be used to make a Compton recoil electron for further transport or for the
TTB approximation. The differential cross section for the process is given by the Klein-Nishina formula [19]
 0 2  0 
α α α
K(α, µ)dµ = πro2 + 0 + µ2 − 1 dµ, (2.125)
α α α
where ro is the classical electron radius 2.817938

× 10
−13
cm , α and α0 are the incident and final photon
energies in units of 0.511 MeV [ α = E/ mc , where m is the mass of the electron and c is the speed of
2

light], and α0 = α/[1 + α(1 − µ)].

The Compton scattering process is sampled exactly by Kahn’s method [81] below 1.5 MeV and by Koblinger’s
method [82] above 1.5 MeV as analyzed and recommended by Blomquist and Gelbard [83].

2.4.4.2 Detailed Physics Treatment

The detailed physics treatment includes coherent (Thomson) scattering and accounts for fluorescent photons
after photoelectric absorption. Again, photonuclear interactions may (optionally) be included [§2.4.4.3]. Form
factors are used with coherent and incoherent scattering to account for electron binding effects. Photo-neutron
reactions can also be included for select isotopes. Analog capture is always used, as described in §2.4.4.2.3.
The detailed physics treatment is used below energy EMCPF on the PHYS :p card, and because the default
value of EMCPF is 100 MeV, that means it is almost always used by default. It is the best treatment for most
applications, particularly for high-Z nuclides or deep penetration problems.

The detailed physics treatment for next-event estimators such as point detectors is inadvisable, as explained
in §2.4.4.2.5, unless the NOCOH = 1 option is used on the PHYS :p card to turn off coherent scattering.

LA-UR-22-30006, Rev. 1 95 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

1.0

0.8

0.6
I (Z, v) /Z

0.4

0.2

Z = 10
0.0 Z = 80

0 1 2 3 4 5 6 7 8
v

Figure 2.5: Scattering factor modifying the Klein-Nishina cross section from [1].

2.4.4.2.1 Incoherent (Compton) Scattering

To model Compton scattering it is necessary to determine the angle θ of scattering from the incident line of
flight (and thus the new direction), the new energy E 0 of the photon, and the recoil kinetic energy of the
electron, E − E 0 . The recoil kinetic energy can be deposited locally, can be transported in MODE P E problems,
or (default) can be treated with the TTB approximation.

Incoherent scattering is assumed to have the differential cross section

σI (Z, α, µ)dµ = I(Z, v)K(α, µ)dµ, (2.126)

where I(Z, v) is an appropriate scattering factor modifying the Klein-Nishina cross section in Eq. (2.112).

Qualitatively, the effect of I(Z, v) is to decrease the Klein-Nishina cross section (per electron) more extremely
in the forward direction, for low E and for high-Z independently. For any Z, I(Z, v) increases √ from
I(Z, 0) = 0 to I(Z,
√
∞) = Z. The parameter v is the inverse length v = sin(θ/2)/λ
√ = κα 1 − µ, where
κ = 10−8 mo c/ h 2 = 29.1445 cm−1 . The maximum value of v is vmax = kα 2 = 41.2166α at µ = −1.
The essential features of I(Z, v) are indicated in Figure 2.5.

For hydrogen, an exact expression for the form factor is used [84], which is
1
I(1, v) = 1 −
, (2.127)
1 2 2 4
1+ 2f v
√
where f is the inverse fine structure constant, f = 137.0393, and f ⁄ 2 = 96.9014.

The Klein-Nishina formula is sampled exactly by Kahn’s method [81] below 1.5 MeV and by Koblinger’s
method [82] above 1.5 MeV as analyzed and recommended by Blomquist and Gelbard [83]. The outgoing
energy E 0 and angle µ are rejected according to the form factors.

LA-UR-22-30006, Rev. 1 96 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

For next-event estimators such as detectors and DXTRAN, the probability density for scattering toward the
detector point must be calculated as
 0 2  
1 πro2 α α α0
p(µ) = I(Z, v)K(α, µ) = I(Z, v) + 2
+µ −1 , (2.128)
σ1 (Z, α) σ1 (Z, α) α α0 α
where πro = 0.2494351 and σ1 (Z, α) and I(Z, v) are looked up in the data library.

The new energy, E 0 , of the photon accounts for the effects of a bound electron. The electron binding effect
on the scattered photon’s energy distribution appears as a broadening of the energy spectrum due to the
pre-collision momentum of the electron. This effect on the energy distribution of the incoherently scattered
photon is called Doppler broadening.

The Hartree-Fock Compton profiles, J(pz ), are used to account for the effects of a bound electron on the
energy distribution of the scattered photon. These Compton profiles are a collection of orbital and total atom
data tabulated as a function of the projected pre-collision momentum of the electron. Values of the Compton
profiles for the elements are published in tabular form by Biggs et al. [53] as a function of pz .

The scattered energy of a Doppler broadened photon can be calculated by selecting an orbital shell, sampling
the projected momentum from the Compton profile, and calculating the scattered photon energy, E 0 , from
E − E 0 − EE 0 (1 − cos(θ))/mc2
pz = −f p . (2.129)
E 2 + E 02 − 2EE 0 cos(θ)
The Compton profiles are related to the incoherent scattering function, I(Z, v), by
max

X ˆ
pz

I(Z, v) = Jk (pz , Z)dpz , (2.130)

k −∞

where k refers to the particular electron subshell, Jk (pz , Z) is the Compton profile of the k th shell for a given
element, and pmax
z is the maximum momentum transferred and is calculated using E 0 = E − Ebinding .

2.4.4.2.2 Coherent (Thomson) Scattering

Thomson scattering involves no energy loss, and thus is the only photon process that cannot produce electrons
for further transport and that cannot use the TTB approximation. Only the scattering angle θ is computed,
and then the transport of the photon continues.

The differential cross section is σ2 (Z, α, µ)dµ = C 2 (Z, v)T (µ)dµ, where

C(Z, v) is a form factor modifying
the energy-independent Thomson cross section T (µ) = πro2 1 + µ2 dµ.

The general effect of C 2 (Z, v)/Z 2 is to decrease the Thomson cross section more extremely for backward
scattering, for high E, and low Z. This effect is opposite in these respects to the effect of I(Z, v)/Z on K(α, µ)
in incoherent (Compton) scattering. For a given Z, C(Z, v) decreases from C(Z, 0) = Z to C(Z, ∞) = 0. For
example, C(Z, v) is a rapidly decreasing function of µ as µ varies from +1 to −1, and therefore the coherent
cross section is peaked in the forward direction. At high energies of the incoming photon, coherent scattering
√
is strongly forward and √ can
be ignored. The parameter v is the inverse length v = sin(θ/2)/λ √ = κα 1 − µ,
where κ = 10−8 mo c/ h 2 = 29.1445 cm−1 . The maximum value of v is vmax = κα 2 = 41.2166α at
µ = −1. The square of the maximum value is vmax = 1698.8038α2 . The qualitative features of C(Z, v) are
shown in Figure 2.6.

For next-event estimators, one must evaluate the probability density function


p(µ) = πro2 1 + µ2 C 2 (Z, v)/σ2 (Z, α) (2.131)

LA-UR-22-30006, Rev. 1 97 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

1.0 Z = 10
Z = 80

0.8

0.6
C (Z, v) /Z

0.4

0.2

0.0
0 1 2 3 4 5 6
v

Figure 2.6: Form factor modifying the energy-dependent Thomson cross section from [1].

√
for a given µ. Here σ2 (Z, α) is the integrated coherent cross section. The value of C(Z, v) at v = κα 1 − µ
must be interpolated in the original C 2 (Z, vi ) tables separately stored on the cross-section library for this
purpose.

Note that at high energies, coherent scattering is virtually straight ahead with no energy loss; thus, it appears
from a transport viewpoint that no scattering took place. For a point detector to sample this scattering,
the point must lie on the original track (µ ∼ = 1), which is seldom the case. Thus, photon point detector
variances generally will be much greater with detailed photon physics than with simple physics unless coherent
scattering is turned off with NOCOH = 1 on the PHYS :p card, as explained in §2.4.4.2.5.

2.4.4.2.3 Photoelectric Effect

The photoelectric effect consists of the absorption of the incident photon of energy E, with the consequent
emission of several fluorescent photons and the ejection (or excitation) of an orbital electron of binding energy
e < E, giving the electron a kinetic energy of E − e. Zero, one, or two fluorescent photons are emitted. These
three cases are now described.

(1) Zero photons greater than 1 keV are emitted. In this event, the cascade of electrons that fills up the
orbital vacancy left by the photoelectric ejection produces electrons and low-energy photons (Auger effect).
These particles can be followed in MODE P E problems, or be treated with the TTB approximation, or be
assumed to deposit energy locally. Because no photons are emitted by fluorescence (some may be produced
by electron transport or the TTB model), the photon track is terminated. This photoelectric “capture” of the
photon is scored like analog capture in the summary table of the output file. Implicit capture is not possible.

(2) One fluorescent photon of energy greater than 1 keV is emitted. The photon energy E 0 is the difference in
incident photon energy E, less the ejected electron kinetic energy E − e, less a residual excitation energy e0

LA-UR-22-30006, Rev. 1 98 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

that is ultimately dissipated by further Auger processes. This dissipation leads to additional electrons or
photons of still lower energy. The ejected electron and any Auger electrons can be transported or treated
with the TTB approximation. In general,

E 0 = E − (E − e) − e0 = e − e0 . (2.132)

These primary transactions are taken to have the full fluorescent yield from all possible upper levels e0 , but
are apportioned among the x-ray lines Kα1 , (L3 → K); Kα2 , (L2 → K); Kβ10 , (mean M → K); and kβ20 ,
(mean N → K).

(3) Two fluorescence photons can occur if the residual excitation e0 of process (2) exceeds 1 keV. An electron
of binding energy e00 can fill the orbit of binding energy e0 , emitting a second fluorescent photon of energy
E 00 = e0 –e00 . As before, the residual excitation e00 is dissipated by further Auger events and electron production
that can be modeled with electron transport in MODE P E calculations, approximated with the TTB model, or
assumed to deposit all energy locally. These secondary transitions come from all upper shells and go to L
shells. Thus the primary transitions must be Kα1 or Kα2 to leave an L shell vacancy.

Each fluorescent photon born as discussed above is assumed to be emitted isotropically and can be transported,
provided that E 0 , E 00 > 1 keV. The binding energies e, e0 , and e00 are very nearly the x-ray absorption edges
because the x-ray absorption cross section takes an abrupt jump as it becomes energetically possible to eject
(or excite) the electron of energy E ∼
= e00 , then e0 , then e, etc. The jump can be as much as a factor of 20 (for
example, K-carbon).

A photoelectric event is terminal for elements Z < 12 because the possible fluorescence energy is below 1 keV.
The event is only a single fluorescence of energy above 1 keV for 31 > Z ≥ 12, but double fluorescence (each
above 1 keV) is possible for Z ≥ 31. For Z ≥ 31, primary lines Kα1 , Kα2 , and Kβ10 are possible and, in
addition, for Z ≥ 37, the Kβ20 line is possible.

In all photoelectric cases where the photon track is terminated because either no fluorescent photons are
emitted or the ones emitted are below the energy cutoff, the termination is considered to be caused by analog
capture in the output file summary table (and not energy cutoff).

2.4.4.2.4 Pair Production

This process is considered only in the field of a nucleus. The threshold is 2mc2 [1 + (m/M )] ∼
= 1.022 MeV,
where M is the nuclear mass and m is the mass of the electron. There are three cases:

1. In the case of electron transport ( MODE P E), the electron and positron are created and banked and the
photon track terminates.
2. For MODE P problems with the TTB approximation, both an electron and positron are produced but not
transported. Both particles can make TTB approximation photons. The positron is then considered to
be annihilated locally and a photon pair is created as in case (3).

3. For MODE P problems when positrons are not created by the TTB approximation, the incident photon of
energy E vanishes. The kinetic energy of the created positron/electron pair, assumed to be E − 2mc2 ,
is deposited locally at the collision point. The positron is considered to be annihilated with an electron
at the point of collision, resulting in a pair of photons, each with the incoming photon weight, and each
with an energy of mc2 = 0.511 MeV. The first photon is emitted isotropically, and the second is emitted
in the opposite direction. The very rare single-annihilation photon of 1.022 MeV is ignored.

LA-UR-22-30006, Rev. 1 99 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

2.4.4.2.5 Caution for Detectors and Coherent Scattering

The use of the detailed photon physics treatment is not recommended for photon next-event estimators
(such as point detectors and ring detectors) nor for DXTRAN, unless coherent scatter is turned off with
the NOCOH = 1 option on the PHYS :p card. Alternatively, the simple physics treatment (EMCPF < 0.001 on the
PHYS :p card) can be used. Turning off coherent scattering can improve the figure of merit (FOM) [§2.6.5] by
more than a factor of 10 for tallies with small relative errors because coherent scattering is highly peaked in
the forward direction. Consequently, coherent scattering becomes undersampled because the photon must
be traveling directly at the detector point and undergo a coherent scattering event. When the photon is
traveling nearly in the direction of the point detector or the chosen point on a ring detector or DXTRAN
sphere, the PSC term, p(µ), of the point detector [§2.5.6.1] becomes very large, causing a huge score for the
event and severely affecting the tally. Remember that p(µ) is not a probability (that can be no larger than
unity); it is a probability density function (the derivative of the probability) and can approach infinity for
highly forward-peaked scattering. Thus the under-sampled coherent scattering event is characterized by many
low scores to the detector when the photon trajectory is away from the detector (p(µ) = small) and a very
few, very large scores (p(µ) = huge) when the trajectory is nearly aimed at the detector. Such under-sampled
events cause a sudden increase in both the tally and the variance, a sudden drop in the FOM, and a failure
to pass the statistical checks for the tally as described in §2.6.9.2.3.

2.4.4.3 Photonuclear Physics Treatment

Photonuclear physics may be included when handling a photon collision. A photonuclear interaction begins
with the absorption of a photon by a nucleus. There are several mechanisms by which this can occur. The
nuclear data files currently available focus on the energy range up to 150 MeV incident photon energy. The
value of 150 MeV was chosen as this energy is just below the threshold for the production of pions and the
subsequent need for much more complicated nuclear modeling. Below 150 MeV, the primary mechanisms for
photoabsorption are the excitation of either the giant dipole resonance or a quasi-deuteron nucleon pair.

The giant dipole resonance (GDR) absorption mechanism can be conceptualized as the electromagnetic
wave, the photon, interacting with the dipole moment of the nucleus as a whole. This results in a collective
excitation of the nucleus. It is the most likely process (that is, the largest cross section) by which photons
interact with the nucleus. Expected peak cross sections of 6–10 millibarns are seen for the light isotopes and
600–800 millibarns are not uncommon for the heavy elements. Thus, photonuclear collisions may account for
a theoretical maximum of 5–6% of the photon collisions. The GDR occurs with highest probability when the
wavelength of the photon is comparable to the size of the nucleus. This typically occurs for photon energies
in the range of 5–20 MeV and has a resonance width of a few MeV. For deformed nuclei, a double peak is
seen due to the variation of the nuclear radius. Outside of this resonance region, the cross section for a GDR
reaction becomes negligible. A complete description of this process can be found in the text by Bohr and
Mottelson [85].

The quasi-deuteron (QD) absorption mechanism can be conceptualized as the electromagnetic wave interacting
with the dipole moment of a correlated neutron-proton pair. In this case, the neutron-proton pair can be
thought of as a QD having a dipole moment with which the photon can interact. This mechanism is not as
intense as the GDR but it provides a significant background cross section for all incident photon energies
above the relevant particle separation threshold. The seminal work describing this process was published by
Levinger [86, 87]. Recent efforts to model this process include the work of Chadwick et al. [88].

Once the photon has been absorbed by the nucleus, one or more secondary particle emissions can occur. For
the energy range in question (that is, below 150 MeV) these reactions may produce a combination of gamma
rays, neutrons, protons, deuterons, tritons, helium-3 particles, alphas, and fission fragments. The threshold
for the production of a given secondary particle is governed by the separation energy of that particle, typically

LA-UR-22-30006, Rev. 1 100 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

a few MeV to as much as a few 10s of MeV. Most of the these particles are emitted via pre-equilibrium and
equilibrium mechanisms though it is possible, but rare, to have a direct emission.

Pre-equilibrium emission can be conceptualized as a particle within the nucleus that receives a large amount
of energy from the absorption mechanism and escapes the binding force of the nucleus after at least one but
very few interactions with other nuclei. This is in contrast to a direct emission where the emission particle
escapes the nucleus without any interactions. Typically this occurs from QD absorption of the photon where
the incident energy is initially split between the neutron-proton pair. Particles emitted by this process tend
to be characterized by higher emission energies and forward-peaked angular distributions.

Equilibrium emission can be conceptualized as particle evaporation. This process typically occurs after the
available energy has been generally distributed among the nucleons. In the classical sense, particles boil
out of the nucleus as they penetrate the nuclear potential barrier. The barrier may contain contributions
from coulomb potential (for charged particles) and effects of angular momentum conservation. It should
be noted that for heavy elements, evaporation neutrons are emitted preferentially as they are not subject
to the coulomb barrier. Particles emitted by this process tend to be characterized by isotropic angular
emission and evaporation energy spectra. Several references are available on the general emission process
after photoabsorption [89–91].

For all of the emission reactions discussed thus far, the nucleus will most probably be left in an excited state.
It will subsequently relax to the ground state by the emission of one or more gamma rays. The gamma-ray
energies follow the well known patterns for relaxation. The only reactions that do not produce gamma-rays
are direct reactions where the photon is absorbed and all available energy is transferred to a single emission
particle leaving the nucleus in the ground state.

Reactions at higher energies (above the pion production threshold) require more thorough descriptions of the
underlying nuclear physics. The delta resonance and other absorption mechanisms become significant and the
amount of energy involved in the reaction presents the opportunity for the production of more fundamental
particles. While beyond the scope of this current work, descriptions of the relevant physics may be found in
the paper by Fasso et al. [92].

New photonuclear data tables are used to extend the traditional photon collision routines. Because of the
sparsity of photonuclear data, the user is allowed to toggle photonuclear physics on or off (with the fourth
entry on the PHYS :p card) and the code defaults to off. Once turned on, the total photon cross section,
photoatomic plus photonuclear (i.e. the photonuclear cross section is absent from this calculation when
photonuclear physics is off), is used to determine the distance to the next photon collision. For simple
physics, this implies the sum of the photoelectric, pair production, incoherent and photonuclear cross sections.
Detailed physics includes the additional coherent cross section in this sum.

The toggle for turning on and off photonuclear physics is also used to select biased or unbiased photonuclear
collisions. For the unbiased option, the type of collision is sampled as either photonuclear or photoatomic
based on the ratio of the partial cross sections. The biased option is similar to forced collisions. At the
collision site, the particle is split into two parts, one forced to undergo photoatomic interaction and the
other photonuclear. The weight of each particle is adjusted by the ratio of their actual collision probability.
The photoatomic sampling routines (as described in §2.4.4.1 and §2.4.4.2) are used to sample the emission
characteristics for secondary electrons and photons from a photoatomic collision. The emission characteristics
for secondary particles from photonuclear collisions are handled independently.

Once it has been determined that a photon will undergo a photonuclear collision, the emission particles
are sampled as follows. First, the appropriate collision isotope is selected based on the ratio of the total
photonuclear cross section from each relevant table. Note that photoatomic collisions are sampled from a
set of elemental tables whereas photonuclear collisions are sampled from a set of isotopic tables. Next, the
code computes the ratio of the production cross section to the total cross section for each secondary particle
undergoing transport. Based on this ratio, an integer number of emission particles are sampled. If weight

LA-UR-22-30006, Rev. 1 101 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

games (i.e. weight cutoffs or weight windows) are being used, these secondary particles are subjected to
splitting or roulette to ensure that the sampled particles will be of an appropriate weight. The emission
parameters for each secondary particle are then sampled independently from the reaction laws provided in
the data. Last, tallies and summary information are appropriately updated, applicable variance reduction
games are performed, and the emitted particle is banked for further transport.

Note that photonuclear physics was implemented in the traditional Monte Carlo style as a purely statistical
based process. This means that photons undergoing a photonuclear interaction produce an average number of
emission particles. For multiple particle emission, the particles may not be sampled from the same reaction;
for example, if two neutrons are sampled, one may be from the (γ, 2n) distributions and the second from the
(γ, np) distributions. Note that the photonuclear data use the same energy/angle distributions that have
been used for neutrons and the same internal coding for sampling. See §2.4.3.5.4. This generalized particle
production method is statistically correct for large sampling populations and lends itself to uncomplicated
biasing schemes. It is (obviously) not microscopically correct. It is not possible to perform microscopically
correct sampling given the current set of data tables.

Because of the low probability of a photon undergoing a photonuclear interaction, the use of biased photonuclear
collisions may be necessary. However, caution should be exercised when using this option as it can lead to
large variations in particle weights. It is important to check the summary tables to determine if appropriate
weight cutoff or weight windows have been set. That is, check to see if weight cutoffs or weight windows are
causing more particle creation and destruction than expected. It is almost always necessary to adjust the
default neutron weight cutoff (when using only weight cutoffs with photonuclear biasing) as it will roulette a
large fraction of the attempts to create secondary photoneutrons.

More information about the photonuclear physics included in the MCNP code can be found in White [93, 94].

2.4.5 Electron Interactions

The transport of electrons and other charged particles is fundamentally different from that of neutrons and
photons. The interaction of neutral particles is characterized by relatively infrequent isolated collisions, with
simple free flight between collisions. By contrast, the transport of electrons is dominated by the long-range
Coulomb force, resulting in large numbers of small interactions. As an example, a neutron in aluminum slowing
down from 0.5 MeV to 0.0625 MeV will have about 30 collisions, while a photon in the same circumstances
will experience fewer than ten. An electron accomplishing the same energy loss will undergo about 105
individual interactions. This great increase in computational complexity makes a single-collision Monte Carlo
approach to electron transport infeasible for most situations of practical interest.

Considerable theoretical work has been done to develop a variety of analytic and semi-analytic multiple-
scattering theories for the transport of charged particles. These theories attempt to use the fundamental cross
sections and the statistical nature of the transport process to predict probability distributions for significant
quantities, such as energy loss and angular deflection. The most important of these theories for the algorithms
in the MCNP code are the Goudsmit-Saunderson [95] theory for angular deflections, the Landau [96] theory
of energy-loss fluctuations, and the Blunck-Leisegang [97] enhancements of the Landau theory. These theories
rely on a variety of approximations that restrict their applicability, so that they cannot solve the entire
transport problem. In particular, it is assumed that the energy loss is small compared to the kinetic energy
of the electron.

In order to follow an electron through a significant energy loss, it is necessary to break the electron’s
path into many steps. These steps are chosen to be long enough to encompass many collisions (so that
multiple-scattering theories are valid) but short enough that the mean energy loss in any one step is small (so
that the approximations necessary for the multiple-scattering theories are satisfied). The energy loss and
angular deflection of the electron during each of the steps can then be sampled from probability distributions
based on the appropriate multiple-scattering theories. This accumulation of the effects of many individual

LA-UR-22-30006, Rev. 1 102 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

collisions into single steps that are sampled probabilistically constitutes the “condensed history” Monte Carlo
method.

The most influential reference for the condensed history method is the 1963 paper by Berger [98]. Based on
the techniques described in that work, Berger and Seltzer developed the ETRAN series of electron/photon
transport codes [99]. These codes have been maintained and enhanced for many years at the National Bureau
of Standards (now the National Institute of Standards and Technology). The ETRAN codes are also the basis
for the Integrated TIGER Series [100], a system of general-purpose, application-oriented electron/photon
transport codes developed and maintained by Halbleib and his collaborators at Sandia National Laboratories
in Albuquerque, New Mexico. The electron physics in the MCNP code is essentially that of the Integrated
TIGER Series, Version 3.0. The ITS radiative and collisional stopping power and bremsstrahlung production
models were integrated into MCNP4C.

2.4.5.1 Electron Steps and Substeps

The condensed random walk for electrons can be considered in terms of a sequence of sets of values

(0, E0 , t0 , u0 , r0 ), (s1 , E1 , t1 , u1 , r1 ), (s2 , E2 , t2 , u2 , r2 ), . . .

where sn , En , tn , un , and rn are the total path length, energy, time, direction, and position of the electron
at the end of n steps. On the average, the energy and path length are related by
ˆsn
dE
En−1 − En = − ds, (2.133)
ds
sn−1

where −dE/ds is the total stopping power in energy per unit length. This quantity depends on energy and
on the material in which the electron is moving. ETRAN-based codes customarily choose the sequence of
path lengths {sn } such that
En
= k, (2.134)
En−1
for a constant k. The most commonly used value is k = 2−1/8 , which results in an average energy loss per
step of 8.3%.

Electron steps with (energy-dependent) path lengths s = sn − sn−1 determined by Eqs. (2.133)–(2.134)
are called major steps or energy steps. The condensed random walk for electrons is structured in terms
of these energy steps. For example, all pre-calculated and tabulated data for electrons are stored on an
energy grid whose consecutive energy values obey the ratio in Eq. (2.134). In addition, the Landau and
Blunck-Leisegang theories for energy straggling are applied once per energy step. See §2.4.5.6 for a more
detailed option. For a single step, the angular scattering could also be calculated with satisfactory accuracy,
since the Goudsmit-Saunderson theory is valid for arbitrary angular deflections. However, the representation
of the electron’s trajectory as the result of many small steps will be more accurate if the angular deflections
are also required to be small. Therefore, the ETRAN codes and the MCNP code further break the electron
steps into smaller substeps. A major step of path length s is divided into m substeps, each of path length
s/m. Angular deflections and the production of secondary particles are sampled at the level of these substeps.
The integer m depends only on material (average atomic number Z). Appropriate values for m have been
determined empirically, and range from m = 2 for Z < 6 to m = 15 for Z > 91.

In some circumstances, it may be desirable to increase the value of m for a given material. In particular, a
very small material region may not accommodate enough substeps for an accurate simulation of the electron’s
trajectory. In such cases, the user can increase the value of m with the ESTEP option on the material card M .
The user can gain some insight into the selection of m by consulting PRINT Table 85 in the MCNP output.
Among other information, this table presents a quantity called DRANGE as a function of energy. DRANGE

LA-UR-22-30006, Rev. 1 103 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

is the size of an energy step in g/cm2 . Therefore, DRANGE/m is the size of a substep in the same units, and
if ρ is the material density in g/cm3 , then DRANGE/(mρ) is the length of a substep in centimeters. This
quantity can be compared with the smallest dimension of a material region. A reasonable rule of thumb is
that an electron should make at least ten substeps in any material of importance to the transport problem.

2.4.5.2 Condensed Random Walk

In the initiation phase of a transport calculation involving electrons, all relevant data are either precalculated
or read from the electron data file and processed. These data include the electron energy grid, stopping powers,
electron ranges, energy step ranges, substep lengths, and probability distributions for angular deflections and
the production of secondary particles. Although the energy grid and electron steps are selected according to
Eqs. (2.133)–(2.134), energy straggling, the analog production of bremsstrahlung, and the intervention of
geometric boundaries and the problem time cutoff will cause the electron’s energy to depart from a simple
sequence sn satisfying Eq. (2.134). Therefore, the necessary parameters for sampling the random walk will be
interpolated from the points on the energy grid.

At the beginning of each major step, the collisional energy loss rate is sampled (unless the logic described in
§2.4.5.6 is being used). In the absence of energy straggling, this will be a simple average value based on the
nonradiative stopping power described in the next section. In general, however, fluctuations in the energy loss
rate will occur. The number of substeps m per energy step will have been preset, either from the empirically
determined default values, or by the user, based on geometric considerations. At most m substeps will be
taken in the current major step with the current value for the energy loss rate. The number of substeps may
be reduced if the electron’s energy falls below the boundary of the current major step, or if the electron
reaches a geometric boundary. In these circumstances, or upon the completion of m substeps, a new major
step is begun, and the energy loss rate is resampled.

With the possible exception of the energy loss and straggling calculations, the detailed simulation of the
electron history takes place in the sampling of the substeps. The Goudsmit-Saunderson [95] theory is used to
sample from the distribution of angular deflections, so that the direction of the electron can change at the end
of each substep. Based on the current energy loss rate and the substep length, the projected energy for the
electron at the end of the substep is calculated. Finally, appropriate probability distributions are sampled for
the production of secondary particles. These include electron-induced fluorescent X-rays, “knock-on” electrons
(from electron-impact ionization), and bremsstrahlung photons.

Note that the length of the substep ultimately derives from the total stopping power used in Eq. 2.133,
but the projected energy loss for the substep is based on the nonradiative stopping power. The reason for
this difference is that the sampling of bremsstrahlung photons is treated as an essentially analog process.
When a bremsstrahlung photon is generated during a substep, the photon energy is subtracted from the
projected electron energy at the end of the substep. Thus the radiative energy loss is explicitly taken into
account, in contrast to the collisional (nonradiative) energy loss, which is treated probabilistically and is not
correlated with the energetics of the substep. Two biasing techniques are available to modify the sampling of
bremsstrahlung photons for subsequent transport. However, these biasing methods do not alter the linkage
between the analog bremsstrahlung energy and the energetics of the substep.

The MCNP code uses identical physics for the transport of electrons and positrons, but distinguishes between
them for tallying purposes, and for terminal processing. Electron and positron tracks are subject to the
usual collection of terminal conditions, including escape (entering a region of zero importance), loss to time
cutoff, loss to a variety of variance-reduction processes, and loss to energy cutoff. The case of energy cutoff
requires special processing for positrons, which will annihilate at rest to produce two photons, each with
energy mc2 = 0.511008 MeV.

LA-UR-22-30006, Rev. 1 104 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

2.4.5.3 Collisional Stopping Power

Berger [98] gives the restricted electron collisional stopping power, i.e., the energy loss per unit path length
to collisions resulting in fractional energy transfers  less than an arbitrary maximum value m , in the form
    2  
dE E (τ + 2) −
− = N ZC ln + f (τ,  m ) − δ , (2.135)
ds m 2I 2

where
 2
− τ 2m 2τ + 1 1
f (τ, m ) = −1 − β + 2
+ 2 ln(1 − m ) + ln[4m (1 − m )] + 1 −  . (2.136)
τ +1 2 (τ + 1) m

Here  and m represent energy transfers as fractions of the electron kinetic energy E; I is the mean ionization
potential in the same units as E; β is v/c; τ is the electron kinetic energy in units of the electron rest mass; δ
is the density effect correction (related to the polarization of the medium); Z is the average atomic number
of the medium; N is the atom density of the medium in cm−3 ; and the coefficient C is given by

2πe4
C= (2.137)
mv 2
where m, e, and v are the rest mass, charge, and speed of the electron, respectively. The density effect
correction δ is calculated using the prescriptions of Sternheimer, Berger and Seltzer [101] when using data
from the el03 library and using the method of Sternheimer and Peierls [102] when using data from the el
library.

The ETRAN codes and the MCNP code do not make use of restricted stopping powers, but rather treat
all collisional events in an uncorrelated, probabilistic way. Thus, only the total energy loss to collisions is
needed, and Eqs. (2.135)–(2.136) can be evaluated for the special value m = 1/2. The reason for the 1/2 is
the indistinguishability of the two outgoing electrons. The electron with the larger energy is, by definition,
the primary. Therefore, only the range  < 1/2 is of interest. With m = 1/2, Eq. (2.136) becomes
  2
− 1 τ
2
f (τ, m ) = −β + [1 − ln(2)] + + ln(2) . (2.138)
8 τ +1

On the right side of Eq. (2.135), we can express both E and I in units of the electron rest mass. Then E can
be replaced by τ on the right side of the equation. We also introduce supplementary constants


C2 = ln 2I 2 , (2.139a)
C3 = 1 − ln(2), (2.139b)
1
C4 = + ln(2), (2.139c)
8
so that Eq. (2.135) becomes
  (  2 )
dE 2πe4   τ
− = N Z2 ln τ 2 (τ + 2) − C2 + C3 − β 2 + C4 −δ . (2.140)
ds mv 2 τ +1

This is the collisional energy loss rate in MeV/cm in a particular medium. In the MCNP code, we are actually
interested in the energy loss rate in units of MeV barns (so that different cells containing the same material
need not have the same density). Therefore, we divide Eq. (2.140) by N and multiply by the conversion
factor 1024 barns/cm2 . We also use the definition of the fine structure constant

2πe2
α= , (2.141)
hc

LA-UR-22-30006, Rev. 1 105 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

where h is Planck’s constant, to eliminate the electronic charge e from Eq. (2.140). The result is as follows:
  (  2 )
dE 1024 α2 h2 c2  2  τ 1
− = 2
Z ln τ (τ + 2) − C2 + C3 − β + C4 −δ . (2.142)
ds 2πmc2 τ +1 β2

This is the form actually used in the MCNP code to preset the collisional stopping powers at the energy
boundaries of the major energy steps.

The mean ionization potential and density effect correction depend upon the state of the material, either gas
or solid. In the fit of Sternheimer and Peierls [102] the physical state of the material also modifies the density
effect calculation. In the Sternheimer, Berger and Seltzer [101] treatment, the calculation of the density effect
uses the conduction state of the material to determine the contribution of the outermost conduction electron
to the ionization potential. The occupation numbers and atomic binding energies used in the calculation are
from Carlson [103].

2.4.5.4 Radiative Stopping Power

The radiative stopping power is

dE

(n)
− = 1024 Z(Z + η) αre2 T + mc2 Φrad , (2.143)
ds rad

(n)
where Φrad is the scaled electron-nucleus radiative energy-loss cross section based upon evaluations by Berger
and Seltzer for data from either the el or the el03 library (details of the numerical values of the data on
the el03 library can be found in [104–106]); η is a parameter to account for the effect of electron-electron
bremsstrahlung (it is unity when using data from the el library and, when using data from the el03 library, it
is based upon the work of Seltzer and Berger [104–106] and can be different from unity); α is the fine structure
constant; mc2 is the mass energy of an electron; and re is the classical electron radius. The dimensions of the
radiative stopping power are the same as the collisional stopping power.

2.4.5.5 Energy Straggling

Because an energy step represents the cumulative effect of many individual random collisions, fluctuations in
the energy loss rate will occur. Thus the energy loss will not be a simple average ∆; rather there will be a
probability distribution f (s, ∆)d∆ from which the energy loss ∆ for the step of length s can be sampled.
Landau [96] studied this situation under the simplifying assumptions that the mean energy loss for a step
is small compared with the electron’s energy, that the energy parameter ξ defined below is large compared
with the mean excitation energy of the medium, that the energy loss can be adequately computed from the
Rutherford [107] cross section, and that the formal upper limit of energy loss can be extended to infinity.
With these simplifications, Landau found that the energy loss distribution can be expressed as

f (s, ∆)d∆ = φ(λ)dλ (2.144)

in terms of φ(λ), a universal function of a single scaled variable

 
∆ 2ξmv 2
λ= − ln + δ + β 2 − 1 + γ. (2.145)
ξ (1 − β 2 )I 2

Here m and v are the mass and speed of the electron, δ is the density effect correction, β is v/c, I is the
mean excitation energy of the medium, and γ is Euler’s constant (γ = 0.5772157 . . . ) . The parameter ξ is
defined by
2πe4 N Z
ξ= s, (2.146)
mv 2

LA-UR-22-30006, Rev. 1 106 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

where e is the charge of the electron and N Z is the number density of atomic electrons, and the universal
function is
ˆ
x+i∞
1
φ(λ) = exp(µ ln(µ) + λµ)dµ, (2.147)
2πi
x−i∞
where x is a positive real number specifying the line of integration.

For purposes of sampling, φ(λ) is negligible for λ < –4, so that this range is ignored. Börsch-Supan [108]
originally tabulated φ(λ) in the range –4 ≤ λ ≤ 100, and derived for the range λ > 100 the asymptotic form
1
φ(λ) ≈ , (2.148)
w2 + π 2
in terms of the auxiliary variable w, where
3
λ = w + ln(w) + γ − . (2.149)
2

Recent extensions [57] of Börsch-Supan’s tabulation have provided a representation of the function in the
range –4 ≤ λ ≤ 100 in the form of five thousand equally probable bins in λ. In the MCNP code, the
boundaries of these bins are saved in the array eqlm(mlam), where mlam = 5001. Sampling from this tabular
distribution accounts for approximately 98.96% of the cumulative probability for φ(λ). For the remaining
large-λ tail of the distribution, the MCNP code uses the approximate form φ(λ) ≈ w, which is easier to
sample than (w2 + π 2 )−1 , but is still quite accurate for λ > 100.

Blunck and Leisegang [97] have extended Landau’s result to include the second moment of the expansion of
the cross section. Their result can be expressed as a convolution of Landau’s distribution with a Gaussian
distribution: " #
ˆ∞ 2
∗ 1 0 (∆ − ∆0 )
f (s, ∆) = √ f (s, ∆ ) exp d∆0 . (2.150)
2πσ 2σ 2
−∞

Blunck and Westphal [109] provided a simple form for the variance of the Gaussian:
2
= 10 eV · Z (2.151)
4/3
σBW ∆.

Subsequently, Chechin and Ermilova [110] investigated the Landau/Blunck-Leisegang theory, and derived an
estimate for the relative error "  3 #−1/2
10ξ ξ
CE ≈ 1+ , (2.152)
I 10I
caused by the neglect of higher-order moments. Based on this work, Seltzer [111] describes and recommends
a correction to the Blunck-Westphal variance as
σBW
σ= . (2.153)
1 + 3CE
This value for the variance of the Gaussian is used in the MCNP code.

Examination of the asymptotic form for φ(λ) shows that unrestricted sampling of λ will not result in a finite
mean energy loss. Therefore, a material- and energy-dependent cutoff λc is imposed on the sampling of
λ. In the initiation phase of an MCNP calculation, the code makes use of two preset arrays, flam(mlanc)
and avlm(mlanc), with mlanc = 1591. The array flam contains candidate values for λc in the range
–4 ≤ λc ≤ 50000; the array avlm contains the corresponding expected mean values for the sampling of λ. For
each material and electron energy, the code uses the known mean collisional energy loss ∆, interpolating in
this tabular function to select a suitable value for λc , which is then stored in the dynamically allocated array
flc. During the transport phase of the calculation, the value of flc applicable to the current material and
electron energy is used as an upper limit, and any sampled value of λ greater than the limit is rejected. In
this way, the correct mean energy loss is preserved.

LA-UR-22-30006, Rev. 1 107 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

2.4.5.6 Logic for Sampling Energy Straggling

The Landau theory described in the previous section provides an energy-loss distribution determined by
the energy E of the electron, the path-length s to be traversed, and the properties of the material. Let us
symbolize a sampling of this distribution as an application of a straggling operator L(E, s, ∆) that provides a
sampled value of the energy loss ∆. In the MCNP code earlier than version 5.1.40, all parameters needed
for sampling straggling were precomputed and associated with the standard energy boundaries En and the
corresponding ranges s
n . In effect the code was restricted to calculations based on discrete arguments of the
operator L En , sn , ∆n . As a result, the proper assignment of an electron transport step to an energy group
n required a rather subtle logic. Eventually, two algorithms for apportioning straggled energy loss to electron
substeps were made available. With MCNP code version 5.1.40, a third algorithm is provided, as discussed in
§2.4.5.6.3.

2.4.5.6.1 Energy Indexing Algorithm in the MCNP Code

The first energy indexing algorithm (also called the “bin-centered” treatment) developed for the MCNP code
is arguably the less successful of the two existing algorithms, but for historical reasons remains the default
option. It was an attempt to keep the electron substeps aligned as closely as possible with the energy groups
that were used for their straggling samples. A simplified description of the MCNP algorithm is as follows.
An electron of energy E is assigned

to the group n such that En > E ≥ En+1 . A straggled energy loss ∆ is
sampled from L En , sn , ∆n . The electron attempts to traverse m substeps, each of which is assigned the
energy loss ∆/m. If m substeps are completed, the process starts over with the assignment of a new energy
group. However, if the electron crosses a cell boundary, or if the electron energy falls below the current group,
the loop over m is abandoned, even if fewer than m substeps have been completed, and the energy group is
reassigned.

Since the straggling parameters are pre-computed at the midpoints of the energy groups, this algorithm
does succeed in assigning to each substep a straggled energy loss based on parameters that are as close as
possible to the beginning energy of the substep. However, there are two problems with the current MCNP
approach. First, there is a high probability that the electron will not actually complete the expected range
sn for which the energy loss was sampled, in which case the energy loss relies on a linear interpolation in a
theory that is clearly

nonlinear. Second, the final substep of each sequence using the sampled energy loss
from L En , sn , ∆n will frequently fall partially in the next-lower energy group n + 1, but no substep using


the sample from L En , sn , ∆n will ever be partially in the higher group n–1.

o Caution
This results in a small, but potentially significant, systematic error.

See for example the investigations of Schaart et al. [112] and references therein.

2.4.5.6.2 Energy Indexing Algorithm in the ITS Code

Developed for the ITS codes earlier than the MCNP algorithm, this method (also called the “nearest-group-
boundary” treatment) was added to the MCNP code in order to explore some of the energy-dependent
artifacts of the condensed history approach, and in order to offer more consistency with the TIGER Series
codes. This algorithm differs from the default treatment in two ways. First, the electron is initially assigned
to a group n such that
(En−1 + En )/2 > E ≥ (En + En+1 )/2. (2.154)

LA-UR-22-30006, Rev. 1 108 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

In other words, the electron is assigned to the group whose upper limit is closest to the electron’s energy.
Second, although the electron will be reassigned when it enters a new geometric cell, it will not be reassigned
merely for falling out of the current energy group. These differences serve to reduce the number of times
that unwanted imposition of linear interpolation on partial steps occurs, and to allow more equal numbers of
excursions above and below the energy group from which the Landau sampling was made. As [112] shows, these
advantages make the ITS algorithm a more accurate representation of the energy loss process, as indicated
in comparisons with reference calculations and experiments. Nevertheless, although the reliance on linear
interpolation and the systematic errors are reduced, neither is completely eliminated. It is straightforward to
create example calculations that show unphysical artifacts in the ITS algorithm as well as in the MCNP logic.

The “nearest-group-boundary” treatment is selected by setting the 18th entry of the DBCN card to 1. For
example, the card “ DBCN 17J 1” selects this straggling logic without affecting any of the other DBCN options.

2.4.5.6.3 New Energy- and Step-specific Method

It is easy to express what we would like to see in the straggling logic. For an electron with energy

E about to
traverse a step of length s, we would like to sample the straggling from the operator L E, s, ∆ without regard
to the prearranged energy boundaries En . In the MCNP code, version 5.1.40, we have now brought this
situation about. A new Fortran 90 module has been installed to deal with straggling data. Those parameters
that are separate from the individual straggling events are still precomputed, but each electron transport
step can now sample its energy loss separately from adjacent steps, and specifically for its current energy
and planned step length. Using this approach, we largely eliminate the linear interpolations and energy
misalignments of the earlier algorithms and obviate the need for a choice of energy group. As of the MCNP
code, version 5.1.40, the new straggling logic is included in the code, but is still being tested. Preliminary
results [113] indicate that a more accurate and stable estimate of the straggling is obtained, and a variety of
unphysical artifacts are eliminated.

The new logic is selected by setting the 18th entry of the DBCN card to 2, for example with the card “ DBCN 17J 2”.

2.4.5.7 Angular Deflections

The ETRAN codes and the MCNP code rely on the Goudsmit-Saunderson [95] theory for the probability
distribution of angular deflections. The angular deflection of the electron is sampled once per substep
according to the distribution
∞ 
X 
1
F (s, µ) = l+ exp(−sGl )Pl (µ), (2.155)
2
l=0

where s is the length of the substep, µ = cos(θ) is the angular deflection from the direction at the beginning
of the substep, Pl (µ) is the lth Legendre polynomial, and Gl is
ˆ1
dσ
Gl = 2πN [1 − Pl (µ)]dµ, (2.156)
dΩ
−1

in terms of the microscopic cross section dσ/dΩ, and the atom density N of the medium.

For electrons with energies below 0.256 MeV, the microscopic cross section is taken from numerical tabulations
developed from the work of Riley [114]. For higher-energy electrons, the microscopic cross section is
approximated as a combination of the Mott [115] and Rutherford [107] cross sections, with a screening
correction. Seltzer [99] presents this “factored cross section” in the form
 
dσ Z 2 e2 (dσ/dΩ)Mott
= 2 , (2.157)
dΩ p2 v 2 (1 − µ + 2η) (dσ/dΩ)Rutherford

LA-UR-22-30006, Rev. 1 109 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

where e, p, and v are the charge, momentum, and speed of the electron, respectively. The screening correction
η was originally given by Molière [116] as
 2 "  2 #
1 αmc αZ
(2.158)
2/3
η= Z 1.13 + 3.76 ,
4 0.885p β

where α is the fine structure constant, m is the rest mass of the electron, and β = v/c. The MCNP code now
follows the recommendation of Seltzer [99], and the implementation in the Integrated TIGER Series, by using
the slightly modified form

 2 "  2 r #
1 αmc αZ τ
(2.159)
2/3
η= Z 1.13 + 3.76 ,
4 0.885p β τ +1
where τ is the electron energy in units of electron rest mass. The multiplicative factor in the final term is an
empirical correction which improves the agreement at low energies between the factored cross section and the
more accurate partial-wave cross sections of Riley.

2.4.5.8 Bremsstrahlung

When using data from the el library, for the sampling of bremsstrahlung photons, the MCNP code relies
primarily on the Bethe-Heitler [117] Born-approximation results that have been used until rather recently
[104] in ETRAN. A comprehensive review of bremsstrahlung formulas and approximations relevant to the
present level of the theory in the MCNP code can be found in the paper of Koch and Motz [118]. Particular
prescriptions appropriate to Monte Carlo calculations have been developed by Berger and Seltzer [119]. For
the ETRAN-based codes, this body of data has been converted to tables including bremsstrahlung production
probabilities, photon energy distributions, and photon angular distributions.

For data tables on the el03 library, the production cross section for bremsstrahlung photons and energy
spectra are from the evaluation by Seltzer and Berger [104–106]. The evaluation uses detailed calculations
of the electron-nucleus bremsstrahlung cross section for electrons with energies below 2 MeV and above
50 MeV. The evaluation below 2 MeV uses the results of Pratt, Tseng, and collaborators, based on numerical
phase-shift calculations [120–123]. For 50 MeV and above, the analytical theory of Davies, Bethe, Maximom,
and Olsen [124, 125] is used and is supplemented by the Elwert-Coulomb [126] correction factor and the
theory of the high-frequency limit or tip region given by Jabbur and Pratt [127, 128]. Screening effects
are accounted for by the use of Hartree-Fock atomic form factors [1, 129]. The values between these firmly
grounded theoretical limits are found by a cubic-spline interpolation as described in [104, 105]. Seltzer reports
good agreement between interpolated values and those calculated by Tseng and Pratt [130] for 5- and 10-MeV
electrons in aluminum and uranium. Electron-electron bremsstrahlung is also included in the cross-section
evaluation based on the theory of Haug [131] with screening corrections derived from Hartree-Fock incoherent
scattering factors [1, 129]. The energy spectra for the bremsstrahlung photons are provided in the evaluation.
No major changes were made to the tabular angular distributions, which are internally calculated when using
the el library, except to make finer energy bins over which the distribution is calculated.

The MCNP code addresses the sampling of bremsstrahlung photons at each electron substep. The tables of
production probabilities are used to determine whether a bremsstrahlung photon will be created. For data
from the el03 library, the bremsstrahlung production is sampled according to a Poisson distribution along
the step so that none, one or more photons could be produced; the el library allows for either none or one
bremsstrahlung photon in a substep. If a photon is produced, the new photon energy is sampled from the
energy distribution tables. By default, the angular deflection of the photon from the direction of the electron
is also sampled from the tabular data. The direction of the electron is unaffected by the generation of the
photon because the angular deflection of the electron is controlled by the multiple scattering theory. However,

LA-UR-22-30006, Rev. 1 110 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.4. Physics

the energy of the electron at the end of the substep is reduced by the energy of the sampled photon because
the treatment of electron energy loss, with or without straggling, is based only on non-radiative processes.

There is an alternative to the use of tabular data for the angular distribution of bremsstrahlung photons. If
the fourth entry on the PHYS :e card is 1, then the simple, material-independent probability distribution

1 − β2
p(µ)dµ = 2 dµ, (2.160)
2(1 − βµ)

where µ = cos(θ) and β = v/c, will be used to sample for the angle of the photon relative to the direction of
the electron according to the formula
2ξ − 1 − β
µ= , (2.161)
2ξβ − 1 − β
where ξ is a random number drawn from the unit interval. This sampling method is of interest only in the
context of detectors and DXTRAN spheres. A set of source contribution probabilities p(µ) consistent with
the tabular data is not available. Therefore, detector and DXTRAN source contributions are made using
Eq. (2.160). Specifying that the generation of bremsstrahlung photons rely on Eq. (2.160) allows the user to
force the actual transport to be consistent with the source contributions to detectors and DXTRAN.

2.4.5.9 K-shell Electron Impact Ionization and Auger Transitions

Data tables in the el03 library use the same K-shell impact ionization calculation (based upon ITS1.0) as
data tables on the el library, except for how the emission of relaxation photons is treated; the el03 evaluation
model has been modified to be consistent with the photo-ionization relaxation model. In the el evaluation, a
K-shell impact ionization event generated a photon with the average K-shell energy. The el03 evaluation
generates photons with energies given by Everett and Cashwell [51]. Both el03 and el treatments only take
into account the highest Z component of a material. Thus inclusion of trace high Z impurities could mask
K-shell impact ionization from other dominant components.

Auger transitions are handled the same for data tables from the el03 and el libraries. If an atom has
undergone an ionizing transition and can undergo a relaxation, if it does not emit a photon it will emit an
Auger electron. The difference between el and el03 is the energy with which an Auger electron is emitted,
given by EA = EK or EA = EK − 2EL for el or el03, respectively. The el value is that of the highest energy
Auger electron while the el03 value is the energy of the most probable Auger electron. It should be noted
that both models are somewhat crude.

2.4.5.10 Knock-on Electrons

The Møller cross section [132] for scattering of an electron by an electron is

(  2 )
dσ C 1 1 τ 2τ + 1 1
= + 2 + − 2 , (2.162)
d E 2 (1 − ) τ +1 (τ + 1) (1 − )

where , τ , E, and C have the same meanings as in Eqs. (2.135)–(2.138). When calculating stopping powers,
one is interested in all possible energy transfers. However, for the sampling of transportable secondary
particles, one wants the probability of energy transfers greater than some c representing an energy cutoff,
below which secondary particles will not be followed. This probability can be written

ˆ1/2
dσ
σ(c ) = d. (2.163)
d
c

LA-UR-22-30006, Rev. 1 111 of 1078 Theory & User Manual

Chapter 2. Geometry, Data, Physics, and Mathematics 2.5. Tallies

The reason for the upper limit of 1/2 is the same as in the discussion of Eq. (2.138). Explicit integration of
Eq. (2.162) leads to
(  2    )
C 1 1 τ 1 2τ + 1 1 − c
σ(c ) = + + − c − 2 ln . (2.164)
E c 1 − c τ +1 2 (τ + 1) c

Then the normalized probability distribution for the generation of secondary electrons with  > c is given by
1 dσ
g(, c )d = d. (2.165)
σ(c ) d

At each electron substep, the MCNP code uses σ(c ) to determine randomly whether knock-on electrons will
be generated. If so, the distribution of Eq. (2.165) is used to sample the energy of each secondary electron.
Once an energy has been sampled, the angle between the primary direction and the direction of the newly
generated secondary particle is determined by momentum conservation. This angular deflection is used for
the subsequent transport of the secondary electron. However, neither the energy nor the direction of the
primary electron is altered by the sampling of the secondary particle. On the average, both the energy loss
and the angular deflection of the primary electron have been taken into account by the multiple scattering
theories.

2.4.5.11 Multigroup Boltzmann-Fokker-Planck Electron Transport

The electron physics described above can be implemented into a multigroup form using a hybrid multigroup
and continuous-energy method for solving the Boltzmann-Fokker-Planck equation as described by Morel [60].
The multigroup formalism for performing charged particle transport was pioneered by Morel and Lorence
[61–63] for use in deterministic transport codes. With a first-order treatment for the continuous slowing down
approximation (CSDA) operator, this formalism is equally applicable to a standard Monte Carlo multigroup
transport code as discussed by Sloan [133]. Unfortunately, a first-order treatment is not adequate for many
applications. Morel, et al. have addressed this difficulty by developing a hybrid multigroup/continuous energy
algorithm for charged particles that retains the standard multigroup treatment for large-angle scattering,
but treats exactly the CSDA operator. As with standard multigroup algorithms, adjoint calculations are
performed readily with the hybrid scheme.

The process for performing an MCNP/MGBFP calculation for electron/photon transport problems involves
executing three codes. First the CEPXS [61–63] code is used to generate coupled electron-photon multigroup
cross sections. Next the CRSRD code casts these cross sections into a form suitable for use in the MCNP
code by adjusting the discrete ordinate moments into a Radau quadrature form that can be used by a Monte
Carlo code. CRSRD also generates a set of multigroup response functions for dose or charge deposition
that can be used for response estimates for a forward calculation or for sources in an adjoint calculation.
Finally, the MCNP code is executed using these adjusted multigroup cross sections. Some applications of this
capability for electron/photon transport have been presented in [134].

2.5 Tallies

The MCNP code automatically creates standard summary information that gives the user a better insight into
the physics of the problem and the adequacy of the Monte Carlo simulation including: a complete accounting
of the creation and loss of all tracks and their energy; the number of tracks entering and reentering a cell plus
the track population in the cell; the number of collisions in a cell; the average weight, mean free path, and
energy of tracks in a cell; the activity of each nuclide in a cell (that is, how particles interacted with each
nuclide, not the radioactivity); and a complete weight balance for each cell.

LA-UR-22-30006, Rev. 1 112 of 1078 Theory & User Manual