Téléchargement

ORNL/NUREG/CSD-3
A Monte Carlo Method of Solving Heat

Conduction Problems
S. K. Fraley
T. J. Hoffman
P. N. Stevens
% &
Prepared for the U.S. Nuclear Regulatory Commission

Office of Nuclear Material Safety and Safeguards
Under Interagency Agreement DOE 40-549-75
ORNL/NUREG/CSD-3
Distribution Category
UC-32
Contract No. W-7405 eng 26
A MONTE CARLO METHOD OF SOLVING HEAT

CONDUCTION PROBLEMS*
S. K. Fraley
T. J. Hoffman
Computer Sciences Division
P. N. Stevens
University of Tennessee
Submitted to the University of Tennessee by S. K. Fraley as a

doctoral dissertation in the Department of Engineering Science
and Mechanics.
Date Published: December 1977
PREPARED FOR THE U.S. NUCLEAR REGULATORY COMMISSION

OFFICE OF NUCLEAR MATERIAL SAFETY AND SAFEGUARDS
UNDER INTERAGENCY AGREEMENT DOE 40-549-75
UNION CARBIDE CORPORATION, NUCLEAR DIVISION

operating the
Oak Ridge Gaseous Diffusion Plant . Oak Ridge National Laboratory
Oak Ridge Y-12 Plant . Paducah Gaseous Diffusion Plant
for the
DEPARTMENT OF ENERGY
1/ISTKIBUriON Of THIS OOLUVLV

BLANK PAGE
TABLE OF CONTENTS
ABSTRACT
CHAPTER PAGE
I. INTRODUCTION 1
II. TRANSPORT THEORY APPROXIMATION TO THE DIFFERENTIAL
EQUATION OF HEAT CONDUCTION 4
III. APPLICATION OF MONTE CARLO TO THE TRANSPORT EQUATION ... 12
IV. COMPARISONS OF MONTE CARLO TRANSPORT SOLUTIONS TO

ANALYTICAL SOLUTIONS 18
Two-Media Sphere 18
Two-Media RPP Green's Function 23
Two-Media RPP with Sinusoidal Internal Source 31
Two-Media RPP with Uniform Internal Source 43
Two-Media RPP with Given Surface Temperature
Distribution 45
Three-Media RPP Green's Function 48
Three-Media RPP with Sinusoidal Internal Source 60
Two-Media RCC Green's Function 64
Two-Media RCC with Functional Internal Source 71
Two-Media RCC with Uniform Internal Source 74
Two-Media RCC with Given Surface Temperature
Distribution 78
Two-Media RCC with Functional Internal Source and
Given Surface Temperature Distribution 81
Sphere with Time Dependent Source 83
Sphere with Time Dependent Source and Convective
Boundary Condition 84
V. CORRECTIONS TO THE TRANSPORT EQUATION APPROXIMATION . . . . 90
VI. MONTE CARLO SOLUTION OF A COUPLED GAMMA RADIATION AND
HEAT CONDUCTION PROBLEM 98
VII. CONCLUSIONS AND RECOMMENDATIONS Ill
BIBLIOGRAPHY 115
APPENDIXES 121
A. VERIFICATION OF ANALYTICAL SOLUTIONS 124
B. DERIVATION OF A BOUNDED NEXT EVENT ESTIMATOR 147
iii
iv
PAGE
C. LISTING OF COMPUTER INPUT DATA AND SELECTED OUTPUT .... 151

LIST OF TABLES
TABLE PAGE
1. Temperature Comparisons for a Two-Media Sphere with a

Point Source 22
2. Temperature Comparisons for a Two-Media RPP with a

Sinusoidal Internal Source Using the Forward
Formulation 37
3. Temperatures at a Point Using a Bounded Next Event

Estimator 40
4. Temperature Comparisons for a Two-Hedia RPP with a

Sinusoidal Internal Source Using the Adjoint
Formulation 42

Uniform Internal Source 46

7. Temperature Comparisons for a Three-Media RPP with a

Sinusoidal Internal Source 65
8. Temperature Comparisons for a Two-Media RCC with a

Functional Internal Source 75
9. Temperature Comparisons for DOT Runs with Different

Scaling Factors 76

Uniform Internal Source .- 79

12. Temperature Comparisons for a Sphere with a Time

Dependent Source 85
13. Temperature Comparisons for a Sphere with a Time

Dependent Source and Convective Boundary Condition .... 89
14. Temperature Comparisons for Problem Two with the

Negative Albedo Correction Factor 97
vi
TABLE PAGE
C
15. Teaperatures for 9=0 Coapared to Analytical Estimate . . . 102
16. Teaperatures for 8=45° Coapared to Analytical Estimate . . 102
17. Teaperatures for 8=90° Coapared to Analytical Estimate . . 103
18. Teaperatures for 8=135° Compared to Analytical Estimate . . 103
19. Temperatures for 8=180° Compared to Analytical Estiaate . . 104

2
20. d / a Versus Ea 150
LIST OF FIGURES
FIGURE PAGE
1. Two-Media Sphere with Point Source at Center 19
2. Two-Media RPP with Point Source in Medium One 24
3. Two-Media RPP with Distributed Source in Medium One . . . . 32
4. Three-Media RPP with Point Source in Mediun One 50
5. Three-Media RPP with Distributed Source in Medium One . . . 61
6. Two-Media RCC with a Ring Source in Medium One 66
7. Two-Media RCC with a Distributed Source in Medium One . . . 72
8. Diffusion Theory vs Transport Theory with Different
Boundary Conditions 95
9. One-Medium Sphere with Incident Gamma Beam 99
10. Temperatures tor 8 = 0 ° Compared to Analytical

Estimates 105
11. Temperatures for 8 = 45° Compared to Analytical
Estimates 106
12. Temperatures for 8 = 90* Compared to Analytical

Estimates 107

Estimates 108

Estimates 109
vii
ABSTRACT
A new approach in the use of Monte Carlo to solve heat conduction
problems was developed using a transport equation approximation to the
heat conduction equation. The method was shown to be applicable to the
solution of multimedia problems in complex geometries with no inherent
limitations as to the geometric complexity of problems which can be
solved. Nuclear radiation transport and heat conduction problems can
be calculated with the same computer code; and, in the case where the
nuclear reactions are the heat source, the problem can be performed in
a coupled mode with a single computer run.
The method was demonstrated for problems involving multimedia,
internal sources, fixed temperature boundary conditions, convective
boundary conditions, time-dependent sources, and for a coupled nuclear
radiation-heat conduction problem. Comparisons were made to analytical
solutions when they were available or could be generated.
The adjoint formulation was found to be the most suitable mode of
solution for the class of problems which was considered and therefore
the method appears to be more useful for calculating temperatures at
specific points rather than temperature distributions.
ix
CHAPTER I
INTRODUCTION
Monte Carlo methods are particularly useful in the solution of
neutron and gamma-ray transport problems which involve complex
geometries. Recent work has suggested that the Monte Carlo approach
might also be useful in the solution of certain classes of heat
conduction problems.
In this paper a new approach to the use of Monte Carlo in the
solution of heat conduction problems is developed. This method, which
uses a transport equation approximation, will allow the solution of
problems with as much geometric complexity as is necessary to adequately
describe the physical model. The method is also completely compatible
with Monte Carlo radiation transport programs. Thus, criticality,
shielding, and heat transfer calculations can be performed with one
geometric description and one computer code. Coupled radiation transport
and heat conduction problems can also be performed in a single computer
run where the heat generated by the nuclear radiation (for example,
through fission reactions or by energy absorption) is the heat source
for the heat transfer calculation.
There are many ways to solve the basic heat conduction problem both
numerically and analytically. This method is not intended to replace
existing methods. Numerical techniques such as finite difference or
finite element will in general provide a much faster solution. The
proposed method will, however, be able to solve certain problems for
1
2
which solutions were not previously feasible. Since the method uses the
same logic as existing Monte Carlo radiation transport computer codes,
it was not necessary to generate completely new computer programs for
its implementation. The solution to the problems solved in this paper
were performed using existing computer codes.
Other continuous Monte Carlo techniques for solving heat conduction
problems have been reported, and a summary of these methods is given
in Reference 1. However, all published results using these methods have
been limited to single medium problems. Discrete Monte Carlo approaches
have also been used. ~ These methods are Monte Carlo methods which
solve the finite difference equations and have not proven to be
competitive with other finite difference techniques. Hybrid Monte Carlo

14 15
approaches have also been developed, ' but these require the use of
both analog and digital systems and appear to be limited in their poten
tial applications.
The equations of heat conduction and their relationship to the
transport equation approximation in its solution are described in
Chapter II. Chapter III provides a description of the application of
Monte Carlo to the solution of the transport equation.
The validity of the method is established in Chapter IV where
Monte Carlo solutions are compared to analytical results for a wide
range of problems. Since a major area of interest is the demonstration
of the applicability of the transport equation approximation to
multimedia problems, and analytical solutions were available for only
a very limited number of multimedia problems, Chapter IV also includes

3
the development of the analytical solutions for those problems for which
solutions were not previously available. The verification of these
solutions is given in Appendix A. A derivation of a bounded next event
(flux at a point) estimator which was used in a Monte Carlo solution
in Chapter IV is given in Appendix B.
Methods o correct for the differences between transport theory
and diffusion theory as applied to heat conduction are developed in
Chapter V.
A coupled gamma-ray, heat conduction problem is solved in
Chapter VI as an example of the type of problem for which this method
is applicable. A listing of the input data and output for this
problem is given in Appendix C.
Conclusions and recommendations are given in Chapter VII.

CHAPTER II
TRANSPORT THEORY APPROXIMATION TO THE DIFFERENTIAL
EQUATION OF HEAT CONDUCTION
The differential equation of heat conduction through a stationary,
isotropic solid is given by:
V • K(r,T)VT(r,t) • S(r,t) = pc | I , (2-1)
where
K(r,T) = thermal conductivity
p = density
c = specific heat
S(r,t) = internal heat source
If the thermal conductivity is not a function of position or
temperature, this simplifies to:
2
KV T(r,t) • S(r,t) = pc |£ . (2-2)
It is desired to solve the above differential equation subject to
the following types of standard boundary conditions.
1. Boundary condition of the first kind:
The temperature is specified along a boundary surface.
T = f(r ,t) ,
s (2-3)
where r is on a boundary surface.

5
2. Boundary condition of the second kind:
The normal derivative of the tetp^rature is specified along a
boundary surface.
= f
£ <v*> • (2-4)
where r is on a boundary surface, and 3/3n represents
differentiation along the outward normal to the surface.
3. Boundary condition of the third kind:
A combination of the temperature and normal derivative is
specified at a boundary.
3T
K — • hT = f(r-,t) , (2-5)
* 3n
where r is on a boundary surface, 3/3n represents differentiation
along the outward normal to the surface, K is the thermal
conductivity, and h is the convective heat transfer
coefficient. This type of boundary condition is normally
associated with forced convection and written as
K |£ = h(T.-T) , (2-6)
where T^ is the temperature of the surroundings.
4. Natural convection boundary condition:
5 7 4
Kf^hCToo-T) (2-7)
6
S. Thermal radiation boundary condition:
4
K |J = a e d ^ - T ) , (2-8)
where a is the Stefan-Boltzmann constant
e is the emissivity of the surface.
Across interfaces between two different media
3Ti 3T 2
K = K ( 2 9 )
1 ^ T 2 3n~ ' "
where K. and T. are near the boundary in the first medium and K_
and T- are near the boundary in the second medium and
T. = T 2 at boundary.
The neutron diffusion equation can be written as
2
DV <Kr,t) - E *(r,t) • S(r,t) = ^ § £
a . (2-10)
where <J» is the neutron flux
D is the diffusion coefficient
£ is the absorption cross section, and
V is the neutron's speed.
D is normally given by
D 2 11
"3(1^1 ' C- )
where £ is the total cross section,
jj is the average cosine of the scattering angle, and
I is the scattering cross section.

7
If Z = 0, then £ = Z and
a s
D = ^ - (2-12)
3Z(l-u)
If the scattering is isotropic, y = 0, and
D--JT • (2-13)
Therefore, under these condition, the diffusion equation becomes
^•S(r,t)=ifi . (2-14)
and by comparing the above to the heat conduction equation, the two
equations are equivalent if
and
v =— . (2-16)
v
pc •*
In neutron transport theory problems, the diffusion equation is

frequently used as an approximation to the Boltzmann transport equation.
Monte Carlo methods are available for solving the transport equation in
problems with very complex geometries and these methods may then be
used to approximate the solutionr to the equivalent heat conduction
problems.
Transport theory differs from diffusion theory when scattering is
highly anisotropic or the angular distribution of particles (angular
8
flux) is highly anisotropic. Transport theory also differs from
diffusion theory in the application of various boundary conditions. For

1
example, diffusion theory allows the specification of a zero f ux
boundary condition whereas transport theory allows only a zero incoming
flux boundary condition.
In trying to adjust neutron diffusion theory to approximate the
transport equation, various methods of correction have been developed.
For example, to account for anisotropic scattering the diffusion
coefficient is modified by the average scattering angle and a zero flux

18
boundary condition at an extrapolated distance from the boundary is
used to simulate a zero incoming flux at the boundary. In optically
thin regions (very few mean free paths thick), in the vicinity of strong
absorbers or sources o when scattering is strongly anisotropic, the
differences between the neutron diffusion approximation and transport
theory may be significant and in these cases, the use of neutron
diffusion theory is inappropriate. These differences are more easily
corrected when transport theory is used to approximate diffusion theory.
Diffusion theory is based en isotropic scattering and the scattering
used in the transport solution can always be made isotropic. To correct
for the anisotropic flux which transport theory will give in the
vicinity of strong absorbers or sources or in optically thin regions,
the actual cross sections used in the transport calculation can be
scaled so that the regions of interest are no longer optically thin.
An anisotropic flux will therefore exist only over a small region of the
9
problem and the transport theory approximation to the diffusion equation
will be appropriate.
Scaling of cross sections is accomplished by letting T = 86 where
6 is a constant.
Then
34
Z
K0V (J) • S(r,t) = Bpc (2-17)
3tT '
1. Boundary condition of first kind:
f(r ,t)
s
<D = (2-18)
6
2. Boundary condition of second kind:
3<j> f(rs.t)
= (2-19)
3n 8
3. Boundary condition of third kind:
K£ | £ + n8* = f ( r , t ) s
(2-20)
4. Natural convection boundary condition:
K8||=h8 5 / 4
(^) 5 / 4
(2-21)
5. Thermal radiation boundary condition:
4
KB | J - ote* (*i-0) ) (2-22)
10
across interface
V *r - * * *r
2 » ^2-23)
and
B ^ = 0* 2 . (2-24)
Therefore, by letting T = B$ and comparing to the diffusion
equation
1 (2 251
-m - -
and
V . J j . (2-26)
The source doesn't change, S = S(r,t) and the above revised boundary
conditions should be used.
The method of solution using the transport equation approximation
is then to scive the Boltzmann transport equation for the flux $ under
conditions of isotropic scattering and no absorption. The actual cross
sections and velocities used in the calculation may be scaled using
Equations (2-25) and (2-26) so as to make the differences between the
transport equation solution and the diffusion equation solution as small
as desired (with a penalty of increasing computation time). The boundary
conditions used during the calculation are obtained from
Equations (2-18) through (2-24). Since the diffusion equation is
equivalent to the heat conduction equation under the above conditions,

11
the transport equation approximation of the solution to the heat
conch;*.tion equation is then obtained from
T = B+ . (2-27)
The next chapter describes the application of Monte Carlo to the
solution of the Boltznann transport equation.

CHAPTER III
APPLICATION OF MONTE CARLO TO THE TRANSPORT EQUATION
As shown in the previous chapter, the solution of a heat conduction
problem can be approximated by the solution of an appropriate transport
problem. Once the appropriate transport problem has been defined, any
of the methods which would normally be used in its solution can be used
without additional justification. Monte Carlo methods have been
developed for solving transport problems and the previously developed
methodology is thereby appropriate. A mathematical treatment of the
application of the Monte Carlo method to solve the Boltzmann transport

19
equation is given in a book by Spanier and Gelbard, and a review of
the Monte Carlo method as applied to particle transport is given by

20
Carter and Cashwell.
Only a brief discussion of the application of Monte Carlo methods to
the solution of the transport equation will be given here. The
formulation given is based primarily on the article by Coveyou, Cain,
and Yost.
The integral form of the transport equation can be written as
i|»(x) * S(x) • /*(y)G(v,x)dy , (3-1)
ft
where <|»(x) is the event density (collisions and leakage are considered
events)
S(x) is the first event source distribution function
G(y,x) is the transition kernel (the probability of an event
at x given an event at y), and

12
13
x,y are vectors in the space fl of interest and may include
position, direction, time, and energy dependence.
/G(x,y)dy represents the nonabsorption probability at x.

n
The flux at any point in a medium is obtained from the event density
by dividing by the cross section of the medium.
For nonmultiplying media
/G(x,y)dy $ 1 for all xefl. (3-2)
For the case of heat transfer all media will be nonabsorbing and
for all x at collision sites,
ft •i:
/G(x,y)dy = f
f 0 for all x at leakage sites.
(3-3)
The usual problem is to calculate an effect of interest A, which
is given by the functional
\ = /if;(x)R(x)dx , (3-4)
ft
where R(x) is the response function.
The adjoint formulation of the same problem can be written
•*(x) = R(x) • /G(x,y)*(y)dy , (3-5)

ft
and
A = /*(x)S(x)dx. (3-6)
ft
Note that in the adjoint formulation the integration of the
transition kernel is with respect to the second variable instead of the
first. The integral equation for <J» will be referred to as the forward
formulation to differentiate it from the adjoint formulation.

14
Let
•(x) - f * (x) , (3-7)

v
p*0
where
• „ W « S(x) , (3-8)
and
( x ) = ; W C C j r x ) d y ( 3 _ 9 )
Vi *p ' •
Then
/*(y)G(y,x)dy = / ? * (y)G(y,x)dy . (3-10)

p
a flp=o
The suanation, if it exists (existence is discussed in Chapter Three of

Reference 19) is absolutely convergent since <j* (y) — 0 and G(y,x) — 0
and therefore sumation and integration can be interchanged giving
/*(y)G(y,x dy = ?_ /*
J (y)G(y,x)dy = p2+ 1* ^ 0 0 = p ? *
p p W -
n p
P =o a P =o p«i
(3-11)
Therefore, aJding S(x) (or i|» (x)) to both sides gives
r
S(x) + fAy) >(y,x)dy = *(x) , (3-12)
a
and )|>(x) as defined by Equation (3-7) satisfies the integral form of
the transport equation.

15
Rewriting gives:
<|/(x,) = S(xj) • /S(x )G(x2,x )dx2 2 1
a
• fG{x ,xi) 5 /S(X2)G(x ,x )dx2 d x 2 3 3
Q ft
• /G(x ,x ) /G(x ,x ) /S(x )G(x2,x )dx dx3dx
4 1 3 4 2 3 2 4
n ft ft
• . .. . (3-13)
And X is given by
X = /ij;(x )R(x )dx •

1 1 1 (3-14)
Or
X * /SCxORCx^dx, • //R(x )S(x )G(x ,x )dx dx
1 2 2 1 2 1
ft ftft
• /^(XJJGCX^.XJ) /S(x )G(X2 x )dx dx3dx

2 > 3 2 1
ftft ft
• //R(x )G(x ,x ) /G(x ,x ) /S(x )G(x2,X3)dx dx3dX4dx
1 4 1 3 4 2 2 1
ftft ftft
• . . . . (3-15)
Or, interchanging dummy integration variables,
X = /SCxORCxOdx. • //R(x_)S(x )G(x ,x.)dx dx

x x z l
1 1 1 9
SI ftft L i i i
• //R(x )G(x2,x ) /S(x )G(x ,x )dx dx dx3

3 3 1 1 2 1 2
ftft ft
• //R(x )G(x ,x j /G(x ,x ) /S(x )G(x ,x )dx dx dX3dx
4 3 4 2 3 1 1 2 1 2 4
ftft ft ft
• ... . (3-16)
16
The above formulation may be used to construct a Monte Carlo
random walk. 1) A position x. is picked from S(x) and R(x.) is scored.
This is an estimate of the first integral. 2) Starting at x., a new
position x- is picked using G(x.,y) and a value w is assigned by
w = /G(x y)dy . (3-17)

1 l
fi
The product w -R(x_) is then scored as the estimate of the second
integral. 3) Starting at x- a new position x, is picked using G(x ,y)
and a value w_ is assigned by
w 2 = W ] L /G(x ,y)dy .
2 (3-18)
u
The product w *R(x ) is then scored as the estimate of the third

2 3
integral, 4) . . . .
This process is continued until an external boundary is reached or
until the value of w is so small that additional scores of w-R(x) would
be negligible. In this case Russian roulette may be used to determine
if the random walk is to be continued.
The random walk which comprises the series of points x., x~, . . .
in ft will be called a history.
The cumulative score from the history is then an estimate of A.
If a large number of histories are performed, the mean will be a
reasonable estimate of X. The variance of the estimations should also
be computed to provide an indication of the statistical uncertainty of
the estimation.
17
Similarly, the adjoint formulation of the solution for X can be
written
X = /R(x )S(x )dx

1 1 1 + //S(x )R(x )G(x ,x )dx dx
2 1 2 1 1 2 • . . . . (3-19)
Note that this is identical to the forward formulation with the order
of integration interchanged.
The Monte Carlo solution of the adjoint formulation is accomplished
in a similar manner except that x. is chosen from R(x) and x 2 is chosen
from G(x,x.), i.e., given a collision at x,, G(x,Xj) is the probability
that it was preceded by a collision at x.
A number •>:' techniques have been developed to reduce the variance
of the estimate of X for given computational times. Some of the more
common techniques include source biasing, Russian roulette, splitting,
and exponential transform (path length stretching). Standard techniques

are also available for obtaining an estimate of A from the Monte Carlo
histories. Although only a few of these techniques are used in the
following sections in the solution of the demonstration problems, all
of the techniques which have been developed for the Monte Carlo solution
of the transport equation are available for use.

CHAPTER IV
COMPARISONS OF MONTE CARLO TRANSPORT SOLUTIONS
TO ANALYTICAL SOLUTIONS
In this chapter Monte Carlo solutions are formulated for a number
of multimedia problems and compared against analytical solutions.

Since analytical solutions were not available in the literature for
multimedia problems involving more than ^ne dimension, 22 the development
of the analytical solutions for these problems is also presented.
I. TWO MEDIA SPHERE
Problem 1: To calculate the temperature at several points within

a one-dimensional, two-media sphere (Figure 1 ) .
Assume:
1. A point source of strength s at the center
2. K. is the thermal conductivity for 0 S r < a
3. K is the thermal conductivity for a - r - b

2
A. No contact resistance between media
5. T(r*b) = 80°C
6. steady state.
22
The analytical solution can be obtained from Carslaw and Jaeger.
^ ^ - ^ y - v ^ ^ (4-D
T
« - 4i^r" - 4i^b * V a5r5b. (4-2)
18
19
K,
Vigure 1. Two-media sphere with point source at center.

20
The parameters for the probiea are:
K = .3333 cal/sec OB°C

x
K 2 = .1667 cal/sec ca°C
a = 5 cm
b - 10 a
s = 238.8 cal/sec (1000 watts).
This gives:
1
57.0 + 80.0 0 5 r 55
114.0 • 68.6 5 5 r < 10 (4-3)
r
where the temperature is given in °C.
The Monte Carlo calculations were performed using the MORSE-SGC

23
Monte Carlo neutron and gamma-ray transport code, which provides a
Monte Carlo solution to either the forward or adjoint Boltzmann
transport equation.
The parameters used in the transport equation solution (using the
conversion in Chapter II) were
1. Z l C m
x
3Kp
2. 1 2 I p - = 2 cm
i.
3. a - 5 cm
4. b = 10 cm
5. s = 238.8/sec
21
-2 -1
and the flux $ has units of cm sec . The flux was calculated at 1 cm
radial intervals in the sphere using a boundary crossing estimator. The
solution was obtained using the forward formulation. The results of
these calculations are given in Table 1 along with the analytical
results. (The estimate of the standard deviation of the Monte Carlo
solution is also given). The Monte Carlo results were obtained using
4.5 min of CPU time on the IBM 369/91.
To more accurately approximate the diffusion equation by the
transport equation, an extrapolation distance correction could have been
made to account for edge effects. The extrapolation distance d is
given by:
d = j — = .355 cm. (4-4)
This distance would be subtracted from the radius giving an outer
radius of 19.645 cm. The Monte Carlo solution as performed corresponds
to the solution of a problem with outer radius of 20.355 cm, if the
extrapolation distance correction is incorporated.
The analytical solution for a composite sphere with an outer radius
of 20.355 cm is included in the table to show the effects of including
the correction. The Monte Carlo solution to the transport equation
compares very well to the analytical solution once this correction is
made, except in the vicinity of the point source.

22
TABLE 1
TEMPERATURE COMPARISONS FOR A TWO-MEDIA SPHERE

WITH A POINT SOURCE
Analytic Temperature Estimated

Analytic with Extrapolation Monte Carlo Standard
Radius Temperature Distance Correction Temperature Deviation
°C °C °C C*
1.0 137.0 137.4 142 1.8
2.0 108.5 108.9 108 0.70
3.0 99.00 99.39 98.6 0.54
4.0 94.25 94.64 93.6 0.44
5.0 91.40 91.79 91.2 0.26
6.0 87.6 87.99 87.8 0.17
7.0 84.89 85.28 85.2 0.14
8.0 82.85 83.24 83.2 0.083
9.0 81.27 81.66 81.7 0.047
10.0 80.00 80.39 80.3 0.0004
Note: M l numbers calculated with Monte Carlo in t h i s and subse

quent t a b l e s w i l l be given with an estimation of the standard deviation
of the numbers.
23
II. TWO-MEDIA RPP GREEN'S FUNCTION
The next series of problems requires analytical solutions for three-

dimensional , two-media rectangular parallelepipeds. The first step in
generating the solutions to these problems is to derive an appropriate
Green's function (Figure 2).

From Carslaw and Jaeger, 22 the one-medium Green's function for a
rectangular parallelepiped (RPP) 0 5 x - a, 0 5 y 5 b, 0 5 z 5 c, with

zero surface temperature and a point source at (x',y',z',T) is
_. 8 « 9 » f . fcirx . fcirx' . rnny . amy'_._Mrz_._mrz'

T = -r— Z Z Z \ sin sin sin "' sin "' sin sin
a b c L
*=lm=ln=l a a b b c c
To obtain the time independent Green's function integrate /* dx and take
lim t -*• », to get
T 8 « f ? 1 . fcirx . 8.TTX' . miry . miry' . nirz . nirz'

T= =• Z l z — _ Sln __sin-——sin Y/ sin .' sin „ sin——
2 2 2 2 a a b C
abcK* Ufc-m-i H m n ^ ^
a b C
(4-6)
Also from Carslaw and Jaeger,22 if the RPP face at x=0 is maintained at
•(yfZ.t), other faces at zero, with initial conditions T(x,y,z,0)»0
C
T . - p ! ! ? [lsiiîiîn^/dT/ /Iin5Jîii^(y',z' T) f
-K7T (t-T) (-y • J J • -j) , , J

e d y d z (4 . 7)
+-X
-a
POINT SOURCE ATU;/,/*l
Figure 2. Two-media RPP with point source in m?di
urn one.
25
Let $ ( y , z , t ) = • ( y . z ) and t a k e lim t •*• °° t o g e t t h e s t e a d y - s t a t e
solution
2£ff . £irx . miry . nirz

T-±-
be
t ? ? (1 2T
*m sin , / sin
£=lm=ln=l I .2 2 2 2
£ IT +a IT m n
2 2
b c
/^( ^ .) l!^ np^ . . .

y z sin sin y dz (4-8)
C
>o )
But
1/2
s i n h L ( a - x ) 5L • 4
00 2£ir £TTX
E sir (4-9)
f 2 ? 2 2 T77T
4=1 2 2 ^ 2 2 m n
£ IT + a ir
0
sinhl IT a m n
b c J +
~2
b c
Therefore
1 / 2
m 2 2l
s i n h Tr(a-x) 2 2
4 oo oo [b cJ j . miry . nirz
T = r- Z * -1 2TI77
b c
m=ln=l
s i n h Tra m n I
b c .
cb
//<(> ( y ' , z ' ) sin™^—sin^-^-dy' dz' (4-10)
oo
To o b t a i n a Green's f u n c t i o n f o r a c o m p o s i t e , two-media r e g i o n , first
assume an RPP - a - x - 0 , 0 5 y 5 b , 0 5 z - c , K=K 2> <f>(y,z) a t x=0,
other faces at z e i o , then

26
-» 2) 1/2
sinh ir(a+x) • • n
. any . nirz
T =— Z 1 2 ^sin—r*-sin
•=1 n=l
» 1,h
2
2 1/2 " o c
sinh ir a • • n
.2 ?
[b c _
c b
, ,
/ /•(y ,z )sin^-sin2^dy dz' ,
(4-11)
c
o o
Now assuae RPP 0 * x < z, 0 * y S b, 0 * z S c, K = K <fr(y,z) at x = 0,
other faces zero, and point source at (x'.y'.z'), then
2 2 1/2
1 be
{ sinhlir(a-x) "-•£_
2
2
2
211/21
] . wry . nirz
sin-j^-sin
b c
B=I n=i ) . . r m n I
I sinh lira
I? ?i J
c b
• / My\ ')siTîn2?Vdz
2
oo KjTr abc
1 - —=sm. £irx . fcirx' . miry . nrrry' . nirz . nirz' (

. ? ?? .2
-
2 2 a
sin
a
s i n "' s i n "' s i n „ sin—-—/ •
b b c c
4=lm=ln=l I fm n
2 , 2 * 2
a b c
(4-12)
For continuity of heat flow at x = 0
c ^ =K
i dx
(4-13)
^2 3x x=0 x=0
27
From Equation (4-11)
r 2 2^ 1/2
1 / 2
2 cosh wa m n
3T,
n2] . miry . mrz
IT b c
2 ' 2 b c 1/2 sin—|f*-sm
b J
c
i
m=l n=l ,
m2 2
x=0 c sinh ira n
[b cJ
c b
/ /•(/•, ')sinîn2îydz' Z . (4-14)
o o
f 1/2
\ 2 2]
cosh ira m • n
°2i 4rr a>
m °° n ,b 2 2
c J J/2 . nnrz . mrz
be , ,, v2 5
3x m=l n=l b c2 .2 n2
x=0 sinh ira +
2 2
b
c
c b
//*(y',z')sinîn2l^y.dz'
0 0 C K be
CO 00 00
2 Air . £TTX' . nnry . miry' . nirz . nirz'
I I I , or-sm s i n , / sin—ir*—sin sin
Jl=lm=ln=l .2 2 2 2 m2 p
2) a b b c c
I ir + a ir
b c
(4-15)
Using
1/2
fm2 n2l
sinh ir(a-- X ' )
2S.TT . ilTx' b c .
7—z TTS 1 0
(4-16)
£=1 2 2 2 2 m2 • n 2 a fnr2 2l l / r
n*
0
sinh ira
£ IT • a if 2 2
I J c b c j
28
and s e t t i n g
3T 2 3 T
1
C = K (4-17)
2 3T l ST
x=0 x=0
gives
1/2 1/2
2 2 2 2
Alt °* ** • n . miry .
cosh ira o n sin .'si nirz
b c
m=ln=l b c j > c
c b
/ /•^z')sinHîn2£Vdz'
o o
be
m=l n=l
t
sinhma-x')
. .r
m
b
2
f«2
n
c
n
2
"1
2] 1 / Z
l
. miry . nrrry' . mrz . mrz'
sin ,' sin "' sin sin
b b c c
sinhlna —=• • - ^ I
L [b cJ J (4-18)
Assuming termwise equality gives
cb
/ /<Ky' .z'Jsin-—*—sin dy'dz'
o o
1
,[,
sinhhr(a-x')
b
2
ra
c
n
2 1/2-,
sin "' sin

n.2 n2 172 ?
m*
2117?
n* b c
^(K K ) cosh lira
[•
1+ 2
b c b c I (4-19)
Using this in Equations (4-11) and (4-12) gives the solution as

29
00 CD
. mry . ni:z
Z Z sin—•i*s\\\
bc(K *K.) " " b c
1 z. m=l n=l
2
4 » I?*?)'' ]
•m2 n2
sinhlTrCa-x')!^*^ I
i /-? r- i /o-. s i n
1/ s i n (4-20)
2 2
m n_ 1/2 f f -> Vjl/2l b c
.2 * 2 cosh lira
and
sinh
oo oo
1
i - T r b c ( K ) :z :zj
T, =
V 2 m l n f m 2 —^7172
b
•j * —j\
c
sinh
r 7*7 J
1/2-
2 2
m n
.inhlir ( a - x ' )
,b c , . miry' . nirz*
s i n ,' sin
b c
K IT abc
cosh hra
r (7*7) J
00 00 00 1
1 . 9,vx . B-irx . nvrry . miry* . nirz . nirz'
Z E E •== =sin sin sin . / s i n .' s i n sin
22 2 a a b b c c
i=lm=ln=l I m n
2 . 2 2
a b c
(4-21)
To show t h a t t h e s e equations reduce t o the one-media equations when
K. = K , consider an RPP 0 5 x 5 2a, 0 < y < b , 0 < z < c , with zero

2
surface temperature and a point source at ( x ' , y ' , z ' ) . The s t e a d y - s t a t e
s o l u t i o n (Equation (4-6)) i s then

30
00 00 0O
. £ITX . fcirx* . wiry . nnry' . mrz . rorz'
T = Z Z Z •z =jsm=—siite-—sin J s i n " s i n sin—— » f
2 Z 2 Za Za b b c c
KTT abc i = l « = l n = l i m n
- • 2 . 2 2
4a b c
(4-22)
but
£irx . iwx'
- - -2 2 2" 2a 2a
£=1 i m n
2
^ ** . 2 * 2
4a b c
2i 1/2-,
1 / 2
r f 2
r f 2 2i i
sinh sinh hr(2a-x*)
TO
2 2
m n
K 2
2
b c
(4-23)
therefore.
1/2-
T =
Kirbc
m=ln=l 2 2 172 r (2 2] 1 / 2
i
m n
| c o s h
b c r ?*7 J
. miry . miry' . nirz . nirz'
sin-r^-sin J sin sin (4-24)
b b c c
which is Equation (4-20) with K = K~ ~ » K a n c l

translated by a.
31
III. TWO-MEDIA RPP WITH SINUSOIDAL

INTERNAL SOURCE
Problem 2: To calculate the temperature distribution within a

three-dimensional, two-media rectangular parallelepiped (Figure 3).
Assume:
1. K. is the thermal conductivity for 0 < x ^ a, 0 < y < b,

0 5 z <c
2. K is the thermal conductivity for -a < x < 0, 0 < y < b,

0 < z <c
3. No contact resistance between media
4. All external surface temperatures are zero

5. Steady state
6. S = S sin irx/a sin iry/b sin irz/c for O S x S a , 0<y<b,

0 < z <c
The analytical solution for this problem can be obtained using the
previously generated Green's function, Equations (4-20) and (4-21).
Denoting the Green's function solution by T (x,y,z,x',y',z')

where x',y* ,z' is the source point, the temperature distribution due to
a distributed source S(x,y,z) is then given by
T(x,y,z) = /// T (x,y,z,x',y',z')S(x',y',z')dx'dy'dz'

g . (4-25)
The solution for the Green's function in medium two (no source) was
DISTRIBUTED SOURCE
IN MEDIUM ONE
Figure 3. Two-media RPP with distributed source in median one.

33
1/2
sinh * ( a x ) | - 2 * + 4i i
c 2
2g -bc(K K )
Z Z L J ittg£*si
S
nirz
172
1 + 2 m = l n = b ^ T 2 2
m n
sinh Tra
b c
sinh
[--'i? • ?n I 7 _ 5 i n _nmy'
2 L ^. nirz'
__ i n (4-26)
[••
cosh lira
N l
Using the following i n t e g r a l s
D
b/2 , m= 1 ,
i i
.- . iry* . wiry' . (4-27)
} sin-^-sin—i-dy' =
M l ,
1 / 2
a T f 2 2] -|
1
sin——dx
a
(4-28)
"- v
' ir + a IT
n 2i
m n
b c ;
gives
Sa 1
ToU.y.z) = ^ 2 0
K
* <V 2> i. 2 f1 * 1 ' a
2 2
lb
i / 2
r fi i i i
1 sinh jTr(a-x) -=• • —\ . ny . ITZ
2 sinc^-sin— . (4-29)
1 1
I l b c-1 J b c
1 / 2
1/2 cosh[,a
f C,[ ^ ^, ) ^1/21]
b c ;
34
The solution for the Green's function in medium one was
T ( x , y , z , x ' , y \ z ' ) = T- ( - x , y , z , x ' , y ' , z ' ) •

l 0 \
l g i g
K^ abc
CO QO 00
. Z I Z
£=lm=ln=lf £ m n
2 * 2 * 2
a b c
. iirx . iirx' . nnry . miry' . rorz . nirz'

sin sin sin J/sin ,' sin sin (4-30)
a a b b c c
Integrating Equation (4-30) gives
TjCx.y.z) = T (-x,y,z) • -j— 2 j j j-sirMin^sin^- (4-31)

K
* l 1a *b72 c* T
To verify the solution, i t i s necessary that
7%(x,y,z) = 0 , (4-32)
VTjCx.y.z) S(x,y,z) (4-33)

K
l
3T. 3 T
2
= -K (4-34)
hJT 2 3x~
x=0 x=0
and T(x,y,z) = 0 for x, y, z on outer boundaries. The verification of
this solution is shown in Appendix A.
The parameters for the problem are:
1. K x = .1667 cal/sec cm°C
2. K_ = .3333 cal/sec cm*C

35
3. a = 1 cm
4. b = 1 cm
5. c = 1 cm
3 „,, 3
6. S = 2 3 8 . 8 Or/2) c a l / s e c cm
o
with an external surface temperature of 80°C.
This g i v e s :
„. , . 238.8ir s i n h [ / 2 ir(l+x)l . _ . _. f A , .
c
l ltx
T (x,y,z) =
? -* s i m r y sirnrz • 80, (4-35)
12/2 c o s h [ / 2 it]
for -1 < x < 0, o < y < 1, 0 < z * 1, and
T r •> 238.8TT s i n h [ / 2 T T ( 1 - X ) ] _ .
l
T,(x,y,z) = -* '-^sxn y s i n z
1
12/2 cosh[/2 IT]
238 8ir
+ ———sinirx siniry sirnrz • 80, (4-36)
for 0 * x * 1, 0 * y * 1, and 0 * z * 1 .
The source parameter S is chosen to give a source of 1000 watts
(238.8 cal/sec) when the source is integrated over the source volume.
The Monte Carlo solution was attempted using the forward formulation
and a next-event flux at a point estimator. The parameters for the
transport equation solution were
1 E = = 2 C m
' l 3K7
i v 1 . -1
l 1 c m
2
' 2 ' 3KJ *
3. a = b = c = l c m
and the source S is given by

36
3
firl
S « 238.8 \j\ sinirx siniry sirnrz, (4-37)
for 0 S x * 1, 0 - y - 1, 0 - y - 1, and zero elsewhere.
Since these parameters define a problem which is only one mean free
path thick in one dimension, scaling was necessary. A scaling factor of
10 was used. This gives:
1
Ej « 20 cm' , (4-38)
and
1
E 2 = 10 cm" , (4-39)
and the correct temperature estimate T is given by
(4 40)
T « fo ' "
-2 -1
where $ is the flux in units of cm sec which is found using the
scaled parameters. The results of performing a one-hour Monte Carlo
calculation on the IBM 360/195 are compared to the analytical resrlts
in Table 2. The temperature was calculated at .1 cm intervals along the
y = .5 cm, 2 = .5 cm centerline.
Although the results are not unreasonable, the standard
deviations are high for a one-hour run on the IBM 360/195. Since the
standard deviations vary significantly from point to point, as for
example, 1.52 C* to .34 C° to 2.97 C° at x = -.7, -.6, and -.5, the
possibility existed that the fluctuations and poor statistics were the
result of using an unbounded estimator.

37
TABLE 2
TEMPERATURE COMPARISONS FOR A TWO-MEDIA RPP WITH A SINUSOIDAL

INTERNAL SOURCE USING THE FORWARD FORMULATION
Escimated
Analytic Monte Carlo Standard
X Temperature Temperature Deviation
(cm) C
C °C C°
-1.0 80.000 80.21 .080
- .9 80.478 80.96 .38
- .8 81.046 81.96 .55
- .7 81.832 84.30 1.52
- .6 82.985 82.39 .34
- .5 84.728 87.40 2.97
- .4 87.451 90.27 1.72
- .3 91.653 95.0 2.57
- .2 98.149 106.3 3.93
- .1 108.417 113.4 3.29
0.0 124.178 132.5 7.82
.1 166.207 192. 19.0
.2 208.713 195. 12.7
.3 243.817 238. 18.8
.4 266.025 264. 15.8
.5 272.473 276. 19.3
.6 261.488 254. 15.2
.7 233.548 226. 16.0
.8 191.042 195. 10.7
.9 138.506 146.9 12.0
1.0 80.000 91.5 1.37
38
Methods have been developed which produce a bounded next event flux
24-26
at a point estimate, but these schemes are usually difficult to
implement and require a modification of the random walk procedure. To
eliminate these problems, a new bounded flux at a point estimator was
developed.
For a monoenergetic flux, with isotropic scattering, the flux at a
point r * 0, is given by
-Z r -E r t t
2 2
• (r=0) = /E * .(r)4irr ^—=- dr • /S(r)4irr ^ - dr
avi (4-41)
K
4irr 4irr
where
/ / <|>dA
r ) = (4 42)
W TT^T "
A
and A is the surface area at r. A Monte Carlo estimate of
Equation (4-41) is obtained using each collision as an estimate of

-E r 2 2
r 4 i r r an<
£ • a v e ( ) * scoring e
* /4irr at each collision. The source
-E r 2 l
term may be added on analytically o r e /4irr is scored at each source
location chosen.
The large or unbounded estimates occur from collisions very near
the point at which the flux is being estimated. In this case the
-E r
t 2
e" /4ffr contribution is very large. (Note, however, that the total
2
integral is bounded since the r terms cancel.)
39
To eliminate this problem, consider the contributions to the riux
estimator from collisions which occur within a sphere of radius 'a'
about the point
L
C =
/ s *ave ( r ) 4 7 r r
~ T d r
• ( 4
- 4 3 )
o * 4mr
If <l>avg(r) i s known, then a *d' can be c a l c u l a t e d such t h a t
-E r -Z r
a _ t a t
Z
/ E * avg ( r ) 4 T r r ^ ^—^- dr = / Z (r)4irr ^ - dr (4-44)
o 4irr o 4ird
-Z r 2
where e * /4ird is now scored at each collision which occurs within
radius *a*. This estimate is bounded and gives the correct estimate.
But, the calculation of the correct *d' depends upon knowing $_„_(r).
avg
For spheres with a radius **' up to a mean free path, 'd* is fairly
independent of the flux shape within the sphere. Calculations of 'd*
are performed in Appendix B for a flux which varies linearly across the
sphere and for a flux which varies exponentially across the sphere.
Using these results, the calculation of the temperature at a ooint
was performed using several different size spheres to bound the
estimator. The results of these calculations are shown in Table 3.
These calculations are of the temperature at (.1, .5, .5), and were
made using 4.5 min of CPU time on the IBM 360/91. The analytical
result for this point is 166.207°C.
these calculations demonstrate that the poor statistics were not
primarily the result of using an unbounded estimator. No collisions

40
TABLE 3
TEMPERATURES AT A POINT USING A BOUNDED NEXT EVENT ESTIMATOR
Estiuated
Total Standard Response in
Bounding Radius Response Deviation Bounding Sphere Collisions* in
(mean free paths) ^C C^ ^C Bounding Sphere
0.0 160.7 17.7 0.0 0
0.1 160.7 17.7 0.0 0
0.2? 176.0 26.4 34.4 3
0.50 167.9 14.9 38.0 15
1.00 159.8 9.6 46.3 86
2.00 167.2 7.9 74.0 870
'There was a total of 151,387 collisions in the problem.

41
occurred within .1 mean free path of the point detector and only 870 out
of 151387 collisions occurred within two mean free paths. Note, however,
that the collisions within two mean free paths contribute approximately
85* of the answer due to collisions (.SO C° of the answer is due to the
external boundary condition) and that collisions within one mean free
path contributes over 50% of the answer due to collisions. The poor
statistics are a result then of not having a sufficient number of
collisions in the area of interest. These results suggest that the
adjoint formulation would be more appropriate for this problem.
In the adjoint formulation the forward response becomes the adjoint
source (all particles are started at the point of interest) and the
forward source becomes the adjoint response (a collision density
estimator was used to score the source at each collision site). With
these changes a Monte Carlo calculation of the temperature at (.1, .5,
.5) was made using 4.5 min of CPU time on the IBM 360/91. The answer
obtained was 175.O^C with an estimated standard deviation of 1.52 C •
The standard deviation is approximately a factor of 10 lower than the
forward calculation using the same computing time. A reduction of the
standard deviation by a factor of 10 is roughly equivalent to a factor
of 100 in reduction of computing time. The disadvantage of the adjoint
calculation is that a separate computer run must be made for each point
at which a temperature calculation is to be made.
Adjoint calculations were made for the temperatures at four points
along the y * .5, z * .5 centerline. The results are shown in
Table 4. Each calculation represents 4.5 min of IBM 360/91 CPU time.
42
TABLE 4
TEMPERATURE COMPARISONS FOR A TWO-MEDIA RPP WITH A SINUSOIDAL

INTERNAL SOURCE USING THE ADJOINT FORMULATION
Estimated
(cm) °C °C C°
-T
81.83 82.75 .18
- .5 84.73 86.50 .31
.1 166.2 175.0 1.52
.5 272.5 284.3 3.94

43
The Monte Carlo results are high. The magnitude of the error is
that expected due to the edge effects (not including an extrapolation
distance correction). The effects of not including this correction
factor and ways of minimizing this error are discussed in more detail
in two-media right circular cylinder problems later in the section, in
which case deterministic computer codes could be used to calculate the
transport theory solution and therefore eliminate the uncertainty
associated with the Monte Carlo results. One method to reduce the error
associated with the edge effects is to increase the scale factor used
in the calculations. The calculations of the temperature at (.1, .5, .5)
were performed using a scale factor of 40. The calculation represents
20 min of CPU time on the IBM 360/195. The Monte Carlo result was
171.0°C with an estimated standard deviation of 2.05 C°. An even better
answer would be obtained if a larger scale factor (requiring more CPU
time) was used.
IV. TWO-MKDIA KPP WITH UNIFORM INTI.RNAL SOURCi:
Problem 3: To calculate the temperature distribution in a thrce-
dimeisional, two media-rectangular parallelepiped as in the previous
problem except with a source distribution of
S f x . y . z ) = 21H.H cal/scc cm', M -45)
for 0 1 / - a, 0 '. / ' h, 0 - z c , and / . f r o v 1 scwlicrc .
Tin; a n a l y t i c a l '.oJul. ion i \ o b t a i n e d ir.inj.', I In- f wo-med i a KIT •irrrn':
f uri'.T. ion.
44
Using the following i n t e g r a l s
2b AA
—~ » m oaa >
nflr
(4-
o
0 , m even .
and
c sh
f ° H^?l}
HfV
a T
/ sinh[7r(a- x) (4-
° L •( 2o 2^ 1 / 2
,r m n
gives
T (x,y,z) = - j
16S
E E
00 00
1 [-Ml
ft • fl
2
2
fm n2 1
IT (Kj+K-) m odd n odd
- b * ? si„ h
c nirz
"t * JAlig,! (4-
1 / 2
cosh lira
r f 2 2l 1
and
64S 00 00 00
Tj(x,y,z) T (-x,y,z) •
2 1 I I I 2
KJTT &odd m o d d n o d d i » 9 m 2
2
x. m n
- + 72 + -1
a b c
. fcirx . nnry . m r z
(4-
sin snir^^-sin
a D c
This solution is verified in Appendix A
45
This solution still requires the evaluation of an infinite series
to obtain numerical results. However, the series is absolutely
convergent and converges very quickly.
The Monte Carlo solution was performed using the adjoint
formulation. A scaling factor of 10 was used. Adjoint calculations
were made for the temperature at four points along the y = .5, z = .5
centerline. The results are shown in Table 5. Each Monte Carlo
calculation represents 4.5 min of IBM 360/91 CPU time.
V. TWO-MEDIA RPP WITH GIVEN SURFACE

TEMPERATURE DISTRIBUTION
Problem 4: To calculate the temperature distribution in a
three-dimensional, two media rectangular parallelepiped as in the
previous problem except:
1. No internal sources
2. T(x=a) = sin "ny/b sin TTZ/C .
The analytical solution is obtained using the two-media RPP
Green's function. Denoting the Green's function solution by
T (x,y,z,x',y',z') where (x*,y' z*) is the source point, the temperature

f
distribution due to a surface temperature 0(x,y,z) is given by
3T (x,y,z,x\y\z')
T(x\y',z') = K ff^± 4>(x,y,z)dxdydz , (4-SO)
where 3/3n denotes differentiation with respect to the x,y,z variables
along the inward normal to the surface.

46
TABLE 5
TEMPERATURE COMPARISONS FOR A TWC-MEDIA RPP WITH
A UNIFORM INTERNAL STJRCE
Estimated
x Temperature Temperature Deviation
(em) °_C °_C C^
- .7 si.ss s:.s .i:
- .5 8-1.07 So.O .-I
.1 ISo.oT US.: .~S
.5 I b-4.01 1~4.S .So

47
Performing the indicated integration using the following integral:
b/2 , m=l ,
14-51)
J sin^sin^y =
"*b b nj«l ,
and the following Fourier expansion:
1/2
_ co £.(-1) s i n — - sinh rrx
_ r a lb c )
(4-52)
* fc=l
. . JT
, 2 +. a_2fl 1 1/2
2 sinh ira —=• • —=•
b^c ] lb c
gives
1/2"
sinh ir(a+x) 1 1 . Try . TTZ
T (x,y,z) = j
7*7 siit^-sm—
2 ( K + K
1/2" 1/2T'
1 1
sinh ira
7*7
cosh
["(?• 7 ) 1
(4-53)
for -a * x * 0, 0 * y * b, 0 * z * c
and
\l/2
sinh TTX 1 1
b c J . Try . TTZ
T ^ x . y . z ) =• T ( - x , y , z ) •
2 in in
(4-54)
i7r^ b^ r'
sinh ira 1 1
b c
for 0 5 x 5 a, 0 5 y < b, and 0 < z < c.
This solution i s verified in Appendix A. For numerical r e s u l t s a
temperature of 80°C was used on a l l external boundaries except for the
face at x - 1 which was maintained at T(x = 1) = 2200 sin Try sin rri • 80.
48
The Monte Carlo solutions were performed using the adjoint formulation.
The external temperature was scored for each particle which escaped. A
scaling factor of 10 was used. Adjoint calculations were made for the
temperatures at four points along the y - .5, z = .5 centerline. The
results of the Monte Carlo solutions are compared to the analytical
results in Table 6. Each Monte Carlo calculation was made using
4.5 min of CPU time on the IBM 360/91.
VI. THREE-MEDIA RPP GREEN'S FUNCTION
To solve the next problem, it is first necessary to derive a Green's
function for a three-media rectangular parallelepiped (Figure 4 ) .
The same geometry as used in the two-media rectangular
parallelepiped Green's function will be retained with an additional
medium having a conductive coefficient K, added. This gives K~ for
-a * x « 0, 0 * y * b, 0 * z * c, Kj for 0 * x * , 0 * y * ,
a D
0 * z * c, and K 3 for a * x * 2a, 0 * y * b, and 0 * z * c.
The Green's function will be calculated for a point source in the
medium with conductive coefficient K.. Assume that the temperature at
the interface at x = 0 is given by <|>(y,z) and the temperature at the
interface at x = a is given by i|>(y,z). Assuming no contact resistance
at media interfaces and that all external boundaries are at zero
temperatures, the temperatures in the three media can then be obtained
from the single medium Green's function and are given by:
49
TABLE 6
TEMPERATURE COMPARISONS FOR A TWO-MEDIA RPP WITH A

GIVEN SURFACE TEMPERATURE DISTRIBUTION
Estimated
(cm) °C °C C°
- .7 80.716 80.94 .25
- .5 81.849 82.86 .51
.1 114.826 117.3 3.17
.5 317.666 312.5 8.13

O
-a 0 Za
Figure 4. Three-media RPP with point source in mcdi urn one.
51
inh x)
T (x,y,z) = —
2 Z Z
h [^?] ] 2 2 1/2
. nnry . nirz
s i n J/ s i n
b c
-H.J f. n ] l
c b
(4
• / / *(y',z')sin^în2^tiydz'
o o
for -a 5 x < 0, 0 5 y < b, 0 < z < c
1 / 2
I [ (2 2l l
sinh ir(2a-x) ==. • 3 J
ib c J
T,(x,y,z) = T - £ E L .Jsi imry . nirz
[- N"l
D C
m=ln=l
sinh lira
c b
/ / ^(y',z*)sin^^—sin^-^-dy'dz* (4
o o
f o r a 5 x < 2a, 0 < y < b , 0 < z < c ,
and
T.(x,y,z) = £ - Z Z
h-Ayif]
sinh
. miry . nirz
sin-r*-sin
,4* j^p]
1 b c
m=ln=l
c b i i i
• / / My',z')sin-jp—sin dy'dz'j
o o
1/2
2
n 2]
sinh ITX
2
c J . miry . nirz
• — Z Z i n s i n _
b c
m-ln*ll
f
2 T72^ b r
m n 2]
sinh ira
.7* 2
52
c b 8
/ / ip(y'z')sinîn2îy'dz'
o o K-ir abc
1 . iirx . Ivx' . amy . mry . nirz . mrz'

Z •=—=—rsin
Z Z sin sin—i^-sm if s i n sin »
2 2 2 a a b b c c
l=lm=ln=l 1* • * n'
2 . 2 2
a b c
(4-57)
for 0 5 x 5 a, 0 5 y 5 b, o S z S c ,
where $(y,z) and i|»(y,z) are still to be determined. At the interface
at x = 0 continuity of heat flow gives
3T,
(
k — = K (4-58)
2 3x 1 3x
x=o x=o
8 T
2
Solving for •=— gives
x=o
2 2^ 111
3T - 2\^f^ c o s h
""* m 2
n
2
2 K
z l . miry . nirz
W bT \S+h b
2
c )
2
x=o m=ln=llb c m n
sinh ira
b c
c b
/ / .Ky'^sinîn^y'dz' (4-59)
o o
For simplicity of notation the integral term w i l l be denoted F(m,n).
3T,
Solving for -r— gives
x=o
53
' 2 2' 1/2

1/2 cosh wa m n
fm2 n2}
Us b c .
l/T^
miry . nirz ,
r^^sin—1
b e
.
F(m,n) c
f 2
ax m=ln=l b c n 2]
x=o sinh ira m
b c
1/2 . miry . nirz
2 2 sin . s i n
4ir " " b c
• T— Z Z m n
m=ln=l K 2
2 2 2 1/2
m n
b c sinh ira
b c
c b 4ff
/ / 'I'Cy' f z')sin-r<--sin dy'dz' K be
o o
00 00 00
. Z E E 2S.TT ATTX' _...miry^. miry*, nirz . r.ifz'
•sin sin—r*-sin—r*—sin sin-
lV*.V n—2i
*=lm=ln=l
m n
c2
2
K
b
(4-60)
where the remaining integral term w i l l be denoted G(m,n).
Using
2 .\ 1/2-
r
sinhhr(a-x')
f * 1
2£TT £irz'
•sin- (4-61)
m
£=1 . 2 2 2 2fm n
£ ir +a ir —r- + - T
2 2
r \2 2 \ \
2 2
lb c
3T,
and s e t t i n g K~ ~— 1 3x gives
x=o x=0
54
1/2
2 2
i n
1/2 cosh ira
4 * OB OB 2 2
lb c J - m y . mrz _ , ,,
5 1 n 5 i n F ( B n )
•*ln«l b cJ 2 2
l72~ b -c- '
. n
sinh ira
b c
1 / 2 n , r z
2 2 -n^si-
sir b C
- F. Z
be n*l.i=l
b 2
- 2
"" 2
172- <*»•"> * fe
J- n2
sinh w a l , • —5-
lb^ c
OB
I Z
•=ln=
OB
l s i n h
HiHn
sinh N C a - x ' ) ! ^
. nrny . nmy* . nnz . nirz'
s i n "' s i n "' s i n sin—— -
, . ,_.
(4-62)
r [7*7] J
For termwise equality, which follows from the orthogonality of s i n e
functions,
1/2 1/2
2 2 2 2
ir(K K )1+ 2
cosh Tra m n F(m,n)
c2
2
K
b v2 2
b c
1/2 1/2
2 2 . miry* . nirz'
» 2
n 2
m n
= uK, G(m,n) • sinh ir(a-x') + sin "' sin
b c Jb "2c b c
(4-63)
For continuity of heat flow at the interface at x = a
3T,
c3 3 3x
(4-64)
1 3x
x»a x*a
55
3T.
S o l v i n g f o r -g— gives
x=a
1/2
f 2 2)
1/2 c o s h rra m n
2 2 ? ?
4TT m n lb
L
c J . miryin . nirzG „ , .
Z Z sin n
3x be \ TfT l > ~c~~ (»» ) •
m=ln=l b c 2 2
x=a s i n h ira m
—=•+ n—
,2 c
2
b .
(4-65)
3T
Solving for gives
3x"
x=a
1/2 . miry . nirz

3T, 2 2 sin-rf^sin
= - ibel l , Z b c
m n F ( m n )
3x
m=ln=l K 2 2
2 2
,I7T ' *b7
x=a b c m n
s i n h ira
2 2
b c
1 / 2
2 2^
m n
1/2 cosh ira
00 CD 2 2
ra n .b c ) . miry . rnrz .. 4
z z b c 2 2
sin
T/2 -b^ T- 1Tt
r r
G ( m
' n ) +
Oc"
m=ln=l
s i n h ira m n
2 2
K
b c
OO OO OO
2&ir(-l)' iirx' . miry . miry' . nirz . nirz'
Z E E — sin- s i n "^sin .' s i n sin
f 2 2
JL=lm=ln=l . 2 2 2 2 m n
£ IT *a ir
(b c
(4-66)
Using
1 / 2
2 2l
s i n h irx m n
2
* 2ir&(-l) i
fcirx u 2
(4-67)
2 2) sir b c J 1/2
A=l , 2 J 2 2 ro_ n 2 2
x. v +a ir 4
w 2 2 +
s i n h ira m n
b c i.2 2
b c
56
3T,
and setting K, -5— = K gives
1 ax
x=a x=a
2 2
1/2
n
1/2 cosh ira -•2-
2 2 K 2
2
m n i n 5 s i n G(m n)
i7? F ir '
•2 n2
m=ln=l x 2
2
b c sinh Fa
b c
2 2 1/2 sin-înSEx
b c
F(m,n) - |j£-
be 1 , . K 2
2 1/2
m=ln=l b c 2 2
m n
sinh ira K 2
2
b c
1/2
2 2
sinh irx' m n
0 00
lb c J ,,_—sin-ji^-sin-^—sin sin (4-68)
• Z Z
m=ln=l m2 n2'•\l/2 b b c c
sinh ira K 2 2
b c k 4
For termwise equality
1/2 1/2
m 2
n 2 2 2
^(Kj*K ) 3
+
cosh iTa m n G(m,n)
^b "2
c u 2 2
b c
1/2 1/2
2 2 m2 n 2l . miry' . nirz* ,, , , rt
= 1TK m n F(m,n) • sinh irx' +
s m "' s i n . (4-69)
u 2 2 ^
b "2
c
b c
This gives two equations for F(m,n) and G(m,n). Solving these
gives
57
. miry' . nirz'
s i n "' s m
F(m,n) = b c
2 2 1/2
m n
b c
1/2 1 / 2
i 2 2) f [2 2] 1
(K K )cosh iraSj-*—-
1+ 3 sinh i r C a - x ' ) — * ^
2
( K )(K K )cosh [Tra|^4j
V 3 1 + 2 J V
1/2
K sinh ITX* p j + ^jl
(4-70)
Ki
and
(K^KK^K^cosh'
Hr ;(T
1/2 1/2
2 2 • 2 2
. miry' . mrz* (Kj+K^cosh T T a l ^ * ^ . . , m n
sin- J sin sinh TTX' l - j + ô
b c b c b c .
G(m,n) =
1/2
2 2
m n 2
+ 2 ( K )( K )cosh [rraj^n!j J -^
Jb ~c V 3 V 2
K x sinhUa-x')^*^
(4-71)
2
(Kj+^JCKj^Jcosh'|ira |^y
Hs^fT^
Note that
58
and
lis G(m,n) = 0 > (4-73)
which gives the two-media RPP Green's function derived previously.
Using F(m,n) and G(m,n) then gives
T (x,y,z)
2
irbc
m=ln=lj
.miry' . miry . mrz . mrz' . .
, s i n "' sin—rf-sm
b b
2
c
sin
c
sinh ir(a+x)
2 1/2
Ml
m
K
b
2
n
c
2
sinh ira
[r 4)1
h-Wl
1 / 2
2 2'
+ I
(K +K )cosh ^ a F j
1 3 7l sinh
m
(K^K )(K *K Jcosh'
1/2
HH h
2
K sinh TTX'
J
m_ n^
K 2 2
(4-74)
lb c 2 2 1/2
m n
(K +K )(K +K )cosh' ira
1 3 1 2
2 2
-K,
K
b c
for -a 5 x < 0, 0 5 y 5 b, 0 < z 5 c,

59
2 2
. mry . mwy' . nirz . nirz

T f- n ] 1
. _ _ . s i n "'sin . / s i n s i n irz' sinh|Tr(2a-x) - j * — j " I
4 - | ) b b
T (x,y,z)
3 c L U> c J J
TTbc i=ln=l ' 2 2) 1/2
sinhlira
b c
1 / 2 1 / 2
f 2 2l [2 2l
+ I1
(K «K )cosh irapJ + ^J
1 sinh T T X ' F j 2
2
(K 1 + K 3 )(K K )cosh [ira|^^]
1 + 2 J -K*
C inh[,(a-x.)[^4J ]
lS
(4-75)
(K *K )CK *K )cosh* : K
rp*7j ] " '
for a - x - 2 a , 0 - y - b , 0 - z - c ,
and
TjCx.y.z) = T ( - x , y , z ) • T ( 2 a - x , y , z ) •
2 3 j
K TT abc
00 CO 00
1 £TTx. _g.Trx'_. miry . miry* . nirz . rorz'
;
I Z Z 2 15 I 2 "——sin
" " a "sin
S 1
a " b" "sin-^—sin
" b sin »
£=lm=ln=l j ^ + 5L. + D_
2 2
a b c2
(4-76)
for 0 * x * a, O ^ y ^ b , 05 z < .
c
60
VII. THREE-MEDIA RPP WITH SINUSOIDAL

INTERNAL SOURCE
Problem 5: To calculate the temperature distribution in a three-
media, rectangular parallelepiped with no contact resistance between
media and with a fixed, uniform temperature on all external boundaries
(Figure 5 ) .
Assume:
1. K is the thermal conductivity for 0 ^ x - a , 0-y-b,
0 * z * c
2. K_ is the thermal conductivity for -a * x * 0, 0 * y * b,
0 < z * c
3. K, is the thermal conductivity for a * x ^ 2a, 0 < y < b,
0 5 z 5 c
4. Steady state
5. S = S sin irx/a sin iry/b sin TTZ/C for 0 - x - a , 0 S y < b,
previously generated Green's function.
T(x,y,z) * /// T Cx,y,z,x',y',z») S(x',y',z')dx'dy'dz' (4-77)

V *
Using
b/2 , m=l ,
b
/ siifc^- sin-r^dy * (4-78)
° \ 0 , m*l ,
and
z
A
' /
DISTRIBUTED SOURCE
IN MEDIUM ONE
^"7r
/
/
/
/
/
> /
/
/ / /
/ / /
/
/ / *K /
/ / /
Z.
-a 0 Zo
Figure 5. Three-media RPP with distributed source in medium one.
62
1/2
2 2
m n
1 / 2 a sinh ira
a •2 2l 1 . irx , ,b c , (4-79)
/ sinh ir(a-x) s i n — ax =
a
o 2 2
2 m n
ir • a ir 2 + 2
u
b c
gives
S a
T (x,y,z) = — •
2
172
IT .^.4-
^
b c j
(K +K )cosh 3
["W"l' '•
(Kj+yCK^Kpcosh ! ra
[ H -K,
TTZ
sinh
inhma+x) - j ^ - j | s i n
i) (4-80)
fOT - a * X * 0, 0 5 y S b ) 0 * Z * C,
S a'
o
Tj(x,y,z)
b cl 2 2
'-M
1 + K
CVV"^"^?] 'J i
(K *K )(K *K )cosh'
1 3 1 2
FBTF
63
inh ir(2a-x) — j * ~ T sin

>) I? (4-81)
c
for a * x * 2a, 0 * y * b, 0 * z * c.
and
TjCx.y.z) = T (-x,y,z) + T (2a-x,y,z)

2 3
o 1 .irx.iry.irz
(4-82)
~1T I i r "V* "^"^ 1
K +
* i T * 72 T
a b c
for 0 5 x 5 , 0 * y * b, 0 i z * c.
a
This solution is verified in Appendix A. The parameters for the
problem are:
1. Kj = .1667 cal/sec cm°C
2. K_ = .3333 cal/sec cm°C
3. K = .3333 cal/sec cm°C
4. a = 1 cm
5. b = 1 cm
6. c = 1 cm
3 3
7. S = 238.8 (IT/2) cal/sec cm ,
with an external surface temperature of 80°C.
formulation. A scaling factor cf 10 was used. Adjoint calculations
were made for the temperature at four points along the y = .5, z = .5
64
centerline. Each Monte Carlo calculation represents 4.5 min of
IBM 360/91 CPU time. The results are shown in Table 7.
VIII. TWO-MEDIA RCC GREEN'S FUNCTIONS
The next series of probleas requires analytical solutions for two-
aedia right circular cylinders. The first step in generating the
solutions to these probleas is to derive an appropriate Green's
function (Figure 6 ) .
22
Froa Carslaw and Jaeger, the one-medium Green's function for a
right circular cylinder (RCC) 0 * z * £, 0 * 9 * IT, 0 * r * a with zero
surface temperature and a unit instantaneous point source at
( z ' , 8 ' , r ' , t = 0) i s
2 2 2
2 » |_ -Km ir t/* _,_mirz_,_nflrz'
T • —=—»=1L
2
i,|e 1L
" sin—=-sin- l l
ra *—
where a are positive roots of
J (aa) = 0. (4-84)
For a source with symmetry about the z axis, the integration over
6 can be performed using

65
TABLE 7
TEMPERATURE COMPARISONS FOR A THREE-MEDIA RPP WITH

A SINUSOIDAL INTERNAL SOURCE
Estimated
(cm) °C °C C°
- .7 81.849 82.6 .16
- .5 84.89 86.7 .32
.1 166.9 176.7 1.93
.5 276.9 290.6 2.30

RING SOURCE IN MEDIUM ONE—,
\ I
Y
r/\ f
\ i
/
Kr K*
Figure 6. Two-media RCC with a ring source in medium one.

67
2TT 2TT , n=0 ,

/ cos n(6-e»)d0 = (4-85)
o 0 , nj«) ,
to give
2 1 1 2 J
T . -£- E e - ^ ^ / s i n ^ i n ^ Z *'** W \ ^ , (4-86)
a £ m=l a [Jo (aa)]
where a are p o s i t i v e roots of Jo(ota) = 0, o r , since Jo'(x) = - J , ( x )
_2_2 2 W0
T = JL- le'**" t / £
s i n S p s i n ^ Z e"**' ? W MW) J
, .
( 4 8 7 )
a I m=l n=l J. (oâ)
where c^ are positive roots of Jo(a a) = 0. n
For steady state the solution is given by lim / T(r,z,t-T)dt.

J
V x*> o
Using
? ? 2 2
1 2 2
-K(mV/£%a )t ,
Jr e-K(mV/fc *an )(t-T),
v
' 6x = v J 1
2 2 2 2
t
1-e n ' '
r(nV/r*a ) (4-88)
n
taking lim gives -j^l 5~
or, for a steady-state source symmetric about the z axis
4T? . rnirz . mirz' ? 1 Jn("nr) J (<*nr') n ,„ „„,

Z sin %o'-sin-^;— _E, 2 2
—7-^— u s o s a L
g 2
a £ m=l % n=l m TT 2r *V— - — •
Jj ( ° a ) n
(4-89)
2 n
68
If the RCC face is maintained at f(r), other faces at zero, the
temperature is given by
a 3T
T = K / r'f(r') ^ dr' . (4-90)
3z'
o
This gives
T = _«2E_ z a s i n -p j i— Joiai£l_/ , r f ( r , ) J o ( a n r . ) d r ,
a I m=l
-f^-^l
n=lfm n . _ 21 J (oâ) o
(4-91)
Substituting
• ,_ , sinha (£-z)
E Ê 1 5Ji . . » , (4-92)
11 2
* (M
s i n
r 2_2 <,-, £ sinha I

n
into Equation (4-91) gives
i » T <•„ •»-* sinha (i-z) a ,. ._.
a n=l J. (an*) n o
To obtain a Green's function for a composite two-media region,
first assume an RCC -£ - z - 0, 0 S r - a, K = K 2 and f(r) at z = 0,
with the other faces at zero, then
a
T 2 J Jolonrj, sinha (Uz) n ; x , f ( r , ) J o ( a | | r . ) d r , . ( 4 . 9 4 )
a n=l J- (câ) sinncî o

69
Now assume an RCC 0 - z - f c , 0*r*a, K = K . , and f(r) at z = 0
with the other faces at zero and a symmetric ring source at z', r', then
a *> m= lh
00
L. (4-95)
2 2
n=l m ir
, ta2
Jâ) J
,2 n
For continuity of heat flow at z = 0
3T„
(4-96)
1 3z
z=o z=o
3T.
1
2 J J ^ o ^ onco'ft J" r ' f C r ^ J o C V J d r ' • (4-97)
"3T =
sinncÂ
a n=l J (câ) x
z=o
, 00
4ir Jo(anr)J (anr')
m s .i nimrz Q
3z 3z 2
0 2
1
n=l. •V •a 2
z=o z=o a Jl m=l
,2 n
(4-98)
Using
2mn . miTz _ sinh(fl.-z)otq

(4-99)
i
m=l
fcM
«2 2 2
* [m TT
r £ ~ sinhdn?.
70
and setting
3T.
K
= K (4-100)
2 3z 1 3z
z=o z=o
gives
a n=l J. (ana) o
2_ Jo(anr)J (a r') sinh(&-z')an

E 0 n
(4-101)
2 . _ 2, , sinhAa«
a n=l J (ana) "
Assuming ternwise equality gives
/" r.f(r')J a r')dr. = ^

o( n ^ £ ^ Jo(a r') .
n (4-102)
Substituting Equation (4-102) into Equations (4-94) and (4-95) gives the
solution as
f y Jo(Qnr)Jo(qnr') sinhan(&+2)sinhan(fl--2')
T (r,z,r ,z') = - j
2
On sinhan^ coshanfc
a (Kj+K^n-l
(4-103)
and
T^r.z.r'.z') * T (r,-z,r\z') 2
4 ? . irmz . miTz ^ 1
1 JntOnrjJânr')
• -s— £ sin „ s i n - ; — E — T 2 "» 1
a * m=l n=l. J x (ana)
,2 n
(4-104)
71
With the form of the Green's function solution Tg(r,z,r',z'), if
S(r,z,8) is any arbitrary source in medium one symmetric about z, i.e.,
S(r,z,8) = S(r,z), then the temperature is given by
T = / / S(r'.z') TVCr.z^'.z^r'dr'dz' . (4-105)

z'r' *
IX. TWO-MEDIA RCC WITH FUNCTIONAL

INTERNAL SOURCE
Problem 6: To calculate the temperature distribution in a two-
media, right circular cylinder (Figure 7).
Assume:
1. K. ir- the thermal conductivity for 0 * z * £, 0-r-a
2. K ? is the thermal conductivity for - 4 - z - O , O ^ r ^ a
3. Steady state
4. No contact resistance between media
5. All external surface temperatures are zero
6. S = S Jotraj) sin Vz/l, 0*r*a, 0 $ z * £ .
previously generated Green's function.
Using
2 2
la
n = m
a \j- VV) ' '
J rJo^rJJotOmrJdr = (4-106)
° [0 , n/m ,
1/2 , m=l ,
I sin -^-sinj-dz (4-107)
0 , iMl
DISTRIBUTED SOURCE IN MEDIUM ONE —i
Figure 7. Two-media RCC with a distributed source in medium one.

73
and
/ TtA
sinha (£.-z) sinirzdz = —~—-~ sinncÎ ,
n (4-108)
TT *l a 2
n
gives
l
T ( r ^ = So * Jo(<»ir) sinhaT (i+z) (4-109)
v 3 2 2 cosh
' '\ *i\ VW V '
for 0 * r * a, -«. * z * 0,
and
TjCr.z) = T (r,-z) • 2 Jo(a.r) s irz (4-110)

2 TT
i '
K,
for 0 S r 5 a > o S< z, S< I .
This solution is verified in Appendix A.
The parameters for the problem are

x
2. K- = .3333 cal/sec cm°C
3. a = .5 cm
4. I = 1 cm
0 ^ r * .5
5. S = 238.8 (4.632259) J ( r a ) sin irz c a l / s e c cm'
0
0 S z *1
with an external surface temperature of 80°C. The constant
4.632259 is the normalization factor for a unit source.

74
formulation and a scaling factor of 10. Adjoint calculations were made
for the temperature at four points along the r = 0 centerline. Each
Monte Carlo calculation required 4.5 min of IBM 360/91 CPU time. The
results are shown in Table 8.
The use of the scale factor to allow the transport equation to more
closely approximate the diffusion equation was studied for this

27
problem. Calculations were made using the DOT discrete ordinates
radiation transport code. These results are shown in Table 9. (The DOT
diffusion theory calculations include an extrapolation distance
correction to approximate transport theory.) As can be seen, the
transport equation approximation can be made as close as desired by
increasing th; scale factor used in the calculations. However, as the
scale factor is increased, the time involved in performing the
calculation increases also. For example, a calculation of the
temperature at x = -.7 cm was calculated by Monte Carlo using a scale
factor of 40. The results obtained were 81.44°C with an estimated
standard deviation of .38 C°. This run required 4.5 min of IBM 360/91
CPU time.
X. TWO-MEDIA RCC WITH UNIFORM INTERNAL SOURCE
media, right circular cylinder as in the previous problem except with a
source distribution of
75
TABLE 8
TEMPERATURE COMPARISONS FOR A TWO-MEDIA RCC

WITH A FUNCTIONAL INTERNAL SOURCE
Estimated
(cm) °C °C C°
- .7 81.426 82.69 .21
- .5 83.921 86.19 .35
.1 169.209 180.0 2.16
.5 285.031 301.9 2.63

76
TABLE 9
TEMPERATURE COI-PARISONS FOR DOT RUNS WITH

DIFFERENT SCALING FACTORS
DOT DOT DOT DOT DOT

z Diffusion Diffusion Transport Transport Transport
Scale Factor Analytic 1000 100 40 20 10
(cm) °C °C °C °C °C °C
- .9 80.556 80.361 80.396 80.418 80.554 80.857
- .8 80.798 80.805 80.860 80.876 81.089 81.543
- .7 81.426 81.438 81.?16 81.545 81.855 82.500
- .6 82.388 82.412 82.531 82.579 83.009 83.916
- .5 83.92 83.94 84.11 84.20 84.82 85.99
- .4 86.38 86.40 86.62 86.78 87.62 89.19
- .3 90.34 90.36 90.65 90.94 92.01 94.02
- .2 96.72 96.76 97.12 97.60 98.96 101.37
- .1 107.0 107.0 107.5 108.2 109.9 113.0
0.0 123.7 123.9 124.4 125.8 127.5 131.8
.1 169.2 168.8 169.6 171.7 174.6 180.3
.2 214.9 214.4 215.4 218.3 221.1 228.3
.3 253.1 252.4 253.6 257.2 260.3 268.4
.4 277.7 277.0 278.2 282.5 285.6 294.4
.5 285.0 284.4 285.6 289.9 293.2 302.3
.6 273.7 273.0 274.4 278.4 281.5 290.9
.7 244.1 243.6 245.0 248.6 251.9 260.8
.8 198.9 198.7 200.1 202.7 206.6 215.4
.9 142.6 142.6 144.0 146.1 150.9 159.5
77
S(x,y,z) = S Q , (4-111)
for 0 < r 5 a, 0 < z ^ £.

The analytical solution is obtained using the two-media, right
circular cylinder Green's function.
Using the following integrals:
/ r J
o(°n ) r d r =
|r l " J ( Q a ) (4-112)
and
/ sinha (£-z)dz = £2^Z&±

n , (4-113)
gives
o ™ cosho^H-l Jo(Qnr) sinho^(£*z)

T
2 ( r
' z ) =
a(K KJ - 3 (4-114)
1 ++ K s i n h a c o s h a
1 2> n-1 c t ^ J ^ a ) n
n* n*
for 0 * r - a, -1 - z - 0,
and
8S «> . „ oc
T r,z) - T ( r . - z ) • -
l( I ^xn-j- Z- Jo("nT) ,
modd n=l K.ct m w •a 2
1 n ,2 n
(4-115)
for 0 5 r * a, 0 5 5 %, z
This solution is verified in Appendix A.

78
For a unit source, S = l/iraî.

o
This solution still requires the evaluation of an infinite series
to obtain numerical results. The series is absolutely convergent,
however, and converges very quickly. A source of
So « ^ ^ - cal/sec , (4-116)
with an external botJidary temperature of 80°C was used for the numerical
calculations.
formulation and a scaling factor of 10. Adjoint calculations were made
at four points along the r = 0 axis. Each Monte Carlo calculation
required 4.S min of IBM 360/91 CPU time. The results of these
calculations are shown in Table 10.
XI. TWO-MEDIA RCC WITH GIVEN SURFACE

TEMPERATURE DISTRIBUTION
media right circular cylinder as in the previous problem except:
1. No internal sources
2. T(z=£) » J (a,r) .
0
The analytical solution is obtained using the two-media, RCC
Green's function. Denoting the Green's function solution by
Tg(r,z,r',z') for a ring source at (r',z'), the temperature distribution
due to a surface temperature <Kr,z) is given by

79
TABLE 10
TEMPERATURE COMPARISONS FOR A TWO-MEDIA RCC

WITH A UNIFORM INTERNAL SOURCE
Estimated
z Temperature Temperature Deviation
(cm) °C °C C°
- .7 80.716 80.94 .25
- .5 81.849 82.86 .51
.1 114.826 117.3 3.17
.5 317.666 312. 5 8.13

80
(r z r, Z,)
T(r'/'} * K J J ! ^ ' ' ' <Kr,2)rdrdz , (4-117)
3n
r z
where 3/3n denotes differentiation with respect to the r,z variables
along the inward normal to the surface over which the integration is
taken. Note that the integration over $ is already incorporated into
the Green's function.
Performing the indicated integration using the following integral:
2
' a 2
J- [Jl(a a)l » n n=m ,
a
/ rJo(Onr)Jo(otmr)dr = (4-118)
o
n/m ,
and the Fourier expansion
m s i n
nnrz
2 ~ m(-l) "r sinhaiz
(4-119)
%=1 2 , £ Oi 2 2 sinhajil
gives
K sinha («.*z)
1
T
2 ( r
' Z ) = J
K~H(^ o ( a , r ) ^ ^ ^ > c o s h x
(4-120)
for 0 * r * a, -£, * z * a,
and
,_ , sinha^z
Tjf^z) • T (r,-z) • J > ' slnhoÎ
2
( a r ) (4-121)
for 0 S x ^ i , 0 S x i
81
where a. i s the f i r s t zero of J (ot a ) . This solution i s v e r i f i e d in

1 ° n
n
Appendix A.
A temperature of 80°C used on all external boundaries except for
the face at z = 1 which was maintained at T(z = 1) = 2200 J (a,r) • 80. Q
The Monte Carlo solutions were performed using the adjoint formulation
and a scaling factor of 10. Adjoint calculations were made for the
temperatures at four points along the r = 0 axis. The results of the
Monte Carlo solutions compared to the analytical results in
Table 11. Each Monte Carlo calculation was made using 4.5 min of CPU
time on the IBM 360/91.
XII. TWO-MEDIA RCC WITH FUNCTIONAL INTERNAL SOURCE

AND GIVEN SURFACE TEMPERATURE DISTRIBUTION
Problem 9: To calculate the temperature distribution in a two
media right circular cylinder with both an internal source and a given
external temperature distribution.
For convenience, the parameters to be used for this problem are
those which will give the same analytic solution as problem 6. This
is accomplished by setting the outer radius of the right circular
cylinder at a/2 and using the solution for problem 6 at r = a/2 as
the given external temperature distribution with all other parameters
remaining the same.
formulation and a scale factor of 10. Adjoint calculations were made
at two points along the r = 0 axis. At z = -.7 the analytical solution

82
TABLE 11
TEMPERATURE COMPARISONS FOR A TWO-MEDIA RCC WITH

A GIVEN SURFACE TEMPERATURE DISTRIBUTION
Estimated
z Temperature Temperature Deviation
(cm) ^C *_£ C°
- .7 80.389 80.66 .21
- .5 81.070 81.75 .57
.1 105.31 108.2 2.71
.5 278.08 279.2 6.98

83
is 81.426°C. The Monte Carlo solution using 4.2 min of IBM 360/91 CPU
time was 8l.91°C with an estimated standard deviation of .03 C°. At
z = .5 the analytical solution is 285.031°C. The Monte Carlo solution
using 4.5 min of IBM 360/91 CPU time was 302.66°C with an estimated
standard deviation of .82 C°. The Monte Carlo solution with a scale
factor of. 40 using 4.2 min of IBM 360/91 CPU time was 287.76°C with an
estimated standard deviation of 4.0 C°.
XIII. SPHERE WITH TIME DEPENDENT SOURCE
Problem 10: To calculate the time dependent temperature
distribution in a one medium sphere with a uniform, time dependent
source.
Assume:
1. A sphere of radius a
2. K is the thermal conductivity for 0 - r - a
3. 6 is the diffusivity for 0 - r - a
4. Surface and initial temperature of 0°C
5. A uniform source of S = S s~ 0 for t > 0.
The analytical solution for this problem can be obtained from

22
Carslaw and Jaeger.
1 / 2
Tfr t i - Soi -Xt asinr(X/6) l
Hr.tJ - K Xe , *,l/2
w "
rsina(A/6)
+ 2aiSo_» — f c i t - T - s i n MI -«nVt/.
e ^
a
nTK n=l n(n -X / 6 T T )
a
84
For numerical results, the parameters used were:
1. S =23.88 cal/sec cm
o
3. a = 20 cm
4. X = ,01/sec
5. c = .22 cal/g°C
6. p = 2 . 7 g/cm
The temperature was calculated at various times for the center of
the sphere both analytically and using Monte Carlo. The results of
these calculations are shown in Table 12. The Monte Carlo calculations
were made using the adjoint formulation. No scaling was used. For a
time dependent calculation a velocity of 1/pc is used during the
tracking. An extrapolation distance correction was made for the Monte
Carlo calculation. The Monte Carlo results were obtained with 2.5 min
of IBM 360/91 CPU time. An advantage of the adjoint formulation for
problems with time dependent sources and boundary conditions is that the
solution car. be generated for many different times with the same
calculation. For example, the above 2.5 min run performed the
calculation for 59 different times.
XIV. SPHERE WITH TIME DEPcNDENT SOURCE AND

CONVECriVE BOUNDARY CONDITION
Problem 11: To calculate the temperature distribution in a one-
medium sphere with a convective boundary condition.

85
TABLE 12
TEMPERATURE COMPARISONS FOR A SPHERE WITH

A TIME DEPENDENT SOURCE
Estimated
t Temperature Temperature Deviation
(sec) °C °C C°
1 40.09 41. 1.8
5 196.53 199. 3.6
10 385.48 589. 6.2
50 1435.92 1444. 17.3
70 1603.05 1610. 29.0
80 1619.85 1621. 34.0
90 1605.10 1598. 38.4
IOC 1566.85 1546. 40.2
110 151 J.61 1483. 43.0
150 1210.86 1164. 47.7
200 830.41 787. 45.6
500 48.355 47. .5

86
Assume:
1. A sphere of radius a
2. K is the thermal conductivity for 0 - r - a
3. 6 is the diffusivity for 0 - r - a
4. Initial temperature is 0°C

>
5. A uniform source of S for t 0
o
6. Convection at r = a into a medium at 0°C
7. h is the coefficient of convection transfer

22
The analytical solution can be obtained from Carslaw and Jaeger.
The temperature for r = 0 is given by
2 2 2
S a S a 2hS a „ -6a t
TCr-o.t) . - § - • - § - - — 2 - E / , a h (4-123)
K n = l a a a + ( 1 ) S i n a a
n n lT r- n
where a are the roots of aa cot aa = 1 - ah/K. This series converges
very quickly when evaluated numerically.
To perform the Monte Carlo calculation, it is necessary to develop
a method for applying the convective boundary condition to transport
theory. The convective boundary condition for heat conduction problems
can be written:
K = h (Too T)
l In ' ' (4-124)
iliis can be rewritten as

87
If it is assumed that T is the temperature at a distance An away

from the surface, then hAn has the form of a conduction coefficient and
(T^-TJ/An represents the change in temperature per unit distance. Let
K_ = hAn. Then this boundary condition could be approximated in

transport theory by adding a layer An thick onto the boundary with cross
section E given by
The thickness of this layer in mean-free-paths is given by
Z A n 1 1 2 7
= %=3K' C - )
The thickness in mean-free-paths does not depend upon the An used

in the approximation. If the limit as An approaches 0 is taken, then
it is apparent that the convective boundary condition is equivalent to a
physically infinitely thin layer of material which is l/3h mean-free-
paths thick. This type of boundary condition can be treated as an albedo
problem. Letting 6 be the albedo (the ratio of the current returning
from the boundary layer to the current entering the boundary layer) and
6 = l/3h be the thickness of the boundary layer in mean-free-paths, a
simple diffusion theory solution (based on Fick's law) for the albedo
gives
1 -|y
s
3 = 5- • (4-128)
1 +
3*
88
The parameters for the problem are
1. I « ,5 cal/sec cm'C
2#
2. h • .050 cal/sec c m C
# 3
3. pc « .5 C/cal cm
3
4. S * 1 cal/cm sec
o
5. r « 10 cm
formulation, k scaling factor was not used. The temperature was
calculated at the center of the sphere as a function of time. Using
the adjoint formulation, the temperatures for all time desired can be
made with the same computer run. The results are shown in Table 13.
The Monte Carlo calculation required 4.2 min of IBM 360/91 CPU time.
The above problems have shown that the method is applicable to
multimedia problems in a variety of geometries. The adjoint formulation
appears to be the most suitable for most problems.
The major source of error in most of the problems was not due to
statistics but to the use of the transport equation approximation. In
the next chapter the differences between the approximate solution and
the exact solution are examined to determine if corrections can be made.

89
TABLE 13
TEMPERATURE COMPARISONS FOR A SPHERE W i n A TIME

DEPENDENT SOURCE AND CONVECTIVE
T
BOUNDARY CONDI ION
Estimated
t Temperature Temperature Deviation
(sec) °C °C C°
1 2.000 2.05 .074
2 4.000 4.08 .11
3 6.000 6.15 .15
4 8.000 8.18 .18
5 9.996 10.08 .17
10 19.775 19.57 .23
20 37.039 36.17 .33
30 50.775 49.49 .59
40 61.535 60.08 .78
SO 69.945 68.79 .89
100 91.248 91.46 1.55
ISO 97.451 98.63 1.87
200 99.258 100.99 2.02
Note: The analytic steady solution as t-*» is 100°C.

CHAPTER V
CORRECTIONS TO THE TRANSPORT
EQUATION APPROXIMATION
In the problems solved in Gutter IV the major source of inaccuracy
was the approximation associated with using radiation particle
transport theory to solve heat conduction problems. In some of the
problems the inaccuracy was reduced by the use of an extrapolation
distance correction as suggested by neutron diffusion theory.
Problem 11, however, which involved a convective boundary condition, had
no obvious inaccuracies associated with its solution. These results
suggested that the transport equation approximation should be examined
to determine if additional refinements could be made.
The differences between transport theory and diffusion theory can
be identified as two distinct types: 1) the differences in the
description of the heat flow process, and 2) the differences in the
boundary conditions which can be used.
The transport description of the heat flow process will be
different from that provided by the heat conduction equation when the
angular flux calculated by the transport equation is highly
anisotropic. This condition will occur at points which are within
several mean free paths of boundaries or nonuniformly distributed
sources and can be regarded as a localized effect. If an estimate of
the temperature is desired at points within these regions, it will often
90
91
be possible to s_a!e up the cross sections used in the transport

calculation so that the point of interest is no longer close to a
boundary or strong source when the distance is measured in mean free
paths. An alternate approach is also possible if the anisotropic
angular flux has azimuthal symmetry about a given direction. In this
case a localized one-dimensional calculation can be employed on the
basis of the following analogy. The one-dimensional transport
equation for the angular flux in a pure scattering medium with
isotropic scattering can be written as
u f P * ^ • E*(x,ii) = | /Mx,u')duU I S(x) (5-1)
where ty(x,v) is the angular flux

u is the cosine of angle between the angular flux and the
x-axis
Z is the scattering cross section, and
S(x) is an isotropic source.
Defining
1
¥-(x) - /vty(x y)du ,
f (5-2)
* -1
and integrating Equation (S-l) with respect to u gives
Î-S(x). (5-3)
92
Multiplying Equation (5-1) by V and integrating with respect to u

gives
^ j M + Z* (x) = 0.
x (5-4)
Taking the derivative of Equation (5-4) with respect to x and combining

with Equation (5-3) gives
fÎ*S(x)-0. (5-5)
l dx
Note that this is the one dimensional diffusion equation. Therefore,

the second aoaent of the solution of the transport equation is a
solution of the diffusion equation for one dimensional problems with
isotropic scattering and isotropic sources regardless of the
anisotropic shape of the angular flux. This approach should therefore
be useful in making calculations of temperatures in the vicinity of
boundaries or nonuniformly distributed sources.
The second kind of inaccuracy is caused by the differences between

the boundary conditions which are associated with the transport
equation and those which can be used with the diffusion equation.
These effects are not localized and represent the major source of
inaccuracy in most of the problems solved in Chapter IV.
As was shown in problem 11, the convective boundary condition can

be represented by an equivalent albedo condition in transport theory;
and results using this boundary condition are accurate. However, the
fixed temperature boundary condition of heat conduction (which is
93
equivalent to the neutron diffusion theory zero flux condition) does
not have an equivalent boundary condition in transport theory. The
most common aethod used is to treat the zero flux diffusion theory
boundary condition as being represented by a zero incoming flux
condition in the transport equation. To account for the differences
in the boundary conditions, the diffusion equation flux is assumed to be
zero at an extrapolated distance beyond the boundary at which the zero
incoming angular flux condition is imposed. The extrapolation distance
used (.71 mean free paths) is determined from neutron transport theory.
This correction was made in several of the problems in Chapter IV and
resulted in excellent agreement with the analytical solutions.
However, this type of correction is not always practical for problems
with complex shapes or with several different types of materials on the
external boundaries. The effect of the different boundary conditions
can still be minimized in cases where an extrapolation distance
correction is not practical by scaling up the cross sections used in the
transport calculation (with a resultant penalty of increased computer
running time).
A better boundary condition for the transport solution to simulate
the diffusion theory zero flux condition is obtained by considering the
limit of the convective boundary condition as the thickness of the
convective layer (in mean free paths) goes to zero. This limit is
equivalent to the fixed temperature boundary condition. Taking this
limit in Equation (4-128) (pg. 87) gives a transport condition of an
albedo of minus one on the boundary. Fractional negative albedo

94
conditions can be used in transport solutions but a negative albedo of
minus one on all external boundaries is nonconvergent. To test the
negative albedo boundary condition, various negative albedo conditions
were applied at the boundary of the Milne problem and solution obtained
28
with ANISN. The Milne problem is a on*, dimensional infinite half-space
problem (material from 0 to <*>) with a unit source at infinity. The
albedo condition was imposed at the boundary x = 0. The results of the
ANISN calculations for the total flux are given in Figure 8 for albedo
conditions of 0 and -.4, and are compared to the diffusion theory
solution with a zero flux boundary condition. Based on these
calculations, if the change associated with going from a vacuum boundary
condition (zero incoming angular flux) to a -.4 albedo condition is
multiplied by 1.89 and added to the vacuum boundary transport solution,
the results are the same as the diffusion theory solution at points
which are not close to the boundary. The results are also very good
even at points close to the boundary. An albedo of -.4 was chosen
because it gave a significant correction (over 50% of the difference
between the vacuum boundary transport theory solution as compared to
the zero flux diffusion theory solution) and is still small enough so
as to not prohibitively increase tho running time for problem solutions.
Test problems with a -.4 albedo condition ran 1.5 to 2.0 times as long
as those with a vacuum boundary condition. This type of boundary
condition is easily implemented in Monte Carlo transport codes and can
be implemented so as to give the solution with a vacucm boundary
condition and the negative albedo boundary condition in the same run.
95
3 4
DISTANCE (mean free paths)
Figure 8. Diffusion theory vs transport theory

with different boundary conditions.
96
The negative albedo boundary condition was used to recalculate
problea 2. The results of these calculations are shown in Table 14.
The Monte Carlo solution was performed using the «djoint foraulation
and a scaling factor of 10. With the exception of the negative albedo
boundary condition, the transport theory results were performed in the
same Banner as the calculations in Table 4 (p. 42). Since the
estimated standard deviation for the results at x = -.7 en were larger
than desired to make an adequate evaluation of the effect of the
negative albedo boundary condition, a second calculation was performed
using 19 min of IBM 360/91 CPU time. The results of this calculation
was a Monte Carlo estimate of 81.86°C with an estimated standard
deviation of .18 C* as compared to the analytical results of 81.83*C.
Using the above correction factor, the major source of error is now
the statistical error associated with the Monte Carlo process.

97
TABLE 14
TEMPERATURE COMPARISONS FOR PROBLO! TWO KITH

Tl!F. NEGATIVE ALBEDO CORRECTION FACTOR
Estimated
x Temperature Temperature Deviation
fen) °C C°
81.83 81.79 .31
- .-•> 84.73 34.79 .62
.1 166.2 169.5 2.6
.5 271.6 3.3
CHAPTER VI
MONTE CARLO SOLUTION OF A COUPLED GAMMA RADIATION
AND HEAT CONDUCTION PROBLEM
One of the major advantages of this method of solution of the heat
conduction equation is that it will allow the solution of coupled
nuclear radiation transport and heat <onduction problems where nuclear
radiation interactions with matter are the heat source for the heat
conduction problem. Another important advantage is the capability
of solving these problems in complex geometries. In this section the
Monte Carlo method is applied to the solution of a coupled gamma
radiation and heat conduction problem.
Problem 12: To calculate the temperature at several points in a
sphere when the heat source is an incident parallel beam of gamma
radiation (Figure 9).
Assumptions and Parameters:
1. Outer boundary of the sphere is maintained at a constant
temperature.
2. Radius of sphere is 10 cm.
3. K = .3333 cal/cm sec °C.
4. Z for gammas (all energies) = .1 cm .
5. Z for gammas = .03 cm" .

2
6. 1 cal/cm sec energy in the incident gamma radiation.
7. Steady-state condition.
8. Isotropic scattering of gamma radiation.

98
X's
vO
Figure 9. One-medium sphere with incident gamna beam.

100
The values chosen are typical of those for a metal such as
aluminum.
To provide a chec<c on the Monte Carlo solution, the average
temperature as a function of radius of the sphere is estimated using
previously available techniques. For calculating the average gamma

29
flux as a function of radius, Kalos has shown that the incident
2
parallel beam of $ particles/cm -sec can be replaced by an isotropic
2
flux of <f> particles/cm -sec over the entire surface of the sphere.
Using this transformation, the average gamma fiux as a function of
radius was solved using the discrete ordinates code ANISIT and the
average energy deposition was calculated. With an incident gamma

2
radiation beam of 1 cal/cm -sec, the radius averaged gamma flux was
2
fairly uniform in the sphere varying from a low of .685 cal/cm -sec at
2
the center to a high of .848 cal/cm -sec at the edge. The volume
2
averaged flux was .773 cal/cm -sec, The average heat source in the
3
sphere is then given by <J> Z and is .0232 cal/cm -sec. Assuming that
avg a
this heat source is uniform across the sphere, an estimation of the
temperature as a function of radius can then be made analytically.
The Monte Carlo calculation of the coupled gamma radiation and
heat conduction problem was performed using MORSE-SGC. The logic used
for performing a coupled gamma radiation and heat conduction problem
with MORSE-SGC is the same as that used for solving a coupled neutron-
gamma problem. For the forward formulation of the problem, a gamma
starting location is chosen from the parallel beam distribution and the
gamma particle is tracked until it escapes or until the weight of the

101
particle his been reduced by nonabsorption weighting at collision sites
and its survival determined by Russian roulette. At each collision site
energy is absorbed and represents a heat source. This heat source is
treated as a secondary particle. In the actual calculation a secondary
particle was generated with a probability of .1 at each collision site
with particle weighting to maintain a fair game. The secondary
particles (the heat source) are then tracked to determine temperatures.
As was expected, an adjoint formulation of the problem was
necessary to obtain good statistics. (The adjoint calculation obtained
fractional standard deviations of less than 1% in the same time that the
forward calculation obtained fractional standard deviations of over 30%).
A next event estimator was used in the adjoint formulation at each gamma
collision site (the secondary particles in the adjoint formulation) and
the probability of the gamma having reached this collision site from
the incident beam was scored. By scoring to different detectors, where
for each detector a different location of the incoming beam was used,
the temperature at a number of points could be calculated with the same
computer run. Each of these points will have the same radius but are
at locations which have the same relative locations to the incoming
beams as was used during the estimator process.
Comparisons of the analytical estimation of the average
temperature as a function of radius and the Monte Carlo calculation of
the temperature at various locations in the sphere are shown in
Tables 15-19 and Figures 10-14. An adjoint Monte Carlo calculation
was performed for each of the seven radii shown in the figures and
102
TABLE 15
TEMPERATURES FOR 8=0° COMPARED TO ANALYTICAL ESTIMATE
Estimated
Analytical Monte Carlo Standard
r Estimate Calculation Deviation
Ccm) °C °C C*
1.05 .021
1. 1.13 .026
2. 1.13 .024
3. 1.15 .023
4. 1.15 .033
5. 1.03 .023
7. .640 .019
9.0 .312 .019
Note: The Monte Carlo temperatures are calculated along a radius

which makes an angle 9 with the direction towards the incident gamma
radiation.
TABLE 16
Estimated
(cm) °C C°
0 1.16 1.05 .021
1.15 1.12 .024
1.11 1.14 .023
1.056 1.11 .023
.974 1.12 .032
.871 .988 .021
.508 .611 .018
9.0 .220 .297 .018

radiation.
103
TABLE 17
Estimated
V Estimate Calculation Deviation
(cm) °C °C C°
0 1.16 1.05 .021
1.0 1.15 1.09 .025
2.0 1.11 1.07 .021
3.0 1.056 1.00 .024
4,0 .974 .987 .031
S.O .371 .353 .020
7.5 .508 .497 .015
9.0 .220 .238 .016

radiation.
TABLE 13
Estimated
(cm) °C °C C°
0 1.16 1.05 .021
1.0 1.15 1.05 .025
2.0 1.11 .993 .021
3.0 1.056 .904 .024
4.0 .974 . 33? .027
5..0 .871 .697 .020
•'.5 .508 .3^8 .013
9.0 .220 .154 .012

radiation.
104
TABLE 19
TEMPERATURES FOR e=180° COMPARED TO ANALYTICAL ESTIMATE
Estimated
(cm) °C °C C°
0 1.16 1.05 .021
1.0 1.15 1.04 .026
2.0 1.11 .963 .022
3.0 1.056 .867 .024
4.0 .974 .760 .025
5.0 .871 .630 .023
7.5 .508 .289 .012
9.0 .220 .127 .011

which makes an angle e with the direction towards the incident gamma
radiation.
105
1.25
Figure 10. Temperatures for 6 = 0° compared to analytical estimates.

106
Figure 11. Temperatures for 8 « 45° compared to analytical estimates.

107
1.25
Figure 12. Temperatures for P = 90° compared to analytical estimates.

108
1.25
1.00
o
o°_ 0.75
cc
<
w
a.
S 0.50
0.25
Figure 13. Temperatures for 6 « 135° coapared to analytical estimates.

109
1.25
Figure 14. Temperatures for 8 = 180° compared to analytical estimates.

110
tables. Each Monte Carlo calculation required 4.2 sin of CPU time on
the IBM 360/91. A listing of the input to MORSE-SGC and LAVA, and
the relevant output is given in Appendix C for the calculation at
r * 2.0 c«. LAVA was used to create the necessary cross sections for
MORSE-SGC.
The analytical estimate of the average temperature at a given
radius differs froa the Monte Carlo calculation of the sane quantity
because the analytical results are based on the assumption that the
heat generation within the sphere is uniform. This is a fairly good
approximation but is not exact. Since the heat generation is higher
near the edges than at the center, the correct temperature at the center
is lower than the analytical estimate.
The Monte Carlo results along various radii show the variation of
the temperature from the average over the sphere. As is expected
intuitively, the hottest point in the sphere is not at the center but
is several centimeters off center in the direction towards the incident
beam of gamma radiation.

CHAPTER VII
CONCLUSIONS AND RECOMMENDATIONS
The transport theory approximation has been shown to be appropriate
for use in solving many types of heat conduction problems. As was
demonstrated in the problems, the transport approximation is not limited
to Monte Carlo methods. However, since most other techniques can be
used to solve the heat conduction equation directly, the approximation
is not necessary with these other methods.
As is typical with current Monte Carlo methods, other methods of
problem solutions appear to be preferable when they may be used.
However, there are many heat conduction problems for which the solution
is not feasible using other methods. The application of the Monte Carlo
method to the transport equation approximation will provide solutions to
many of these rroblems. There are no inherent limitations as to the
geometric complexity of the heat conduction problems which can be
solved using this method. It will also simplify, or make possible, the
solution of coupled problems which involve nuclear reactions as the heat
source for a heat conduction problem.
The Monte Carlo solution of the transport equation approximation
is more suitable for calculating temperatures at individual points within
a system rather than calculating complete temperature distributions.
The adjoint formulation appears to be the most suitable mode of
solution for most problems. The method is not applicable to problems
where the thermal conductivity is temperature dependent.
Ill
112
The advantages and disadvantages of the Monte Carlo solution of the
transport equation approximation are illustrated by the following
comparison to the surface density walk and floating random walk
techniques:
1. There are no published results on the use of either the surface
density walk or floating random walk in multimedia problems.
2. The tracking algorithm used by the floating random walk is
difficult since it requires the determination of the nearest boundary
point. Both the transport equation approximation and the surface
density walk use a simpler method which requires the finding of the
intersect point between a surface and a given line.
3. The surface density walk tracks directly to surface points,
with the number of tracks for a given walk being determined by the
number of terms in the truncated Fredholm series being evaluated. The
floating random walk tracks over large distances until the walk
approaches a boundary in which case the distances become smaller and
smaller. An approximation must then be used to reach the boundary.
The transport equation approximation uses relatively small steps
throughout the walk with the step size becoming smaller as a scale
factor is increased to make the transport equation approximation better.
4. The surface density walk and the transport equation
approximation may be used with or readily adapted to existing Monte
Carlo radiation transport codes.

113
5. All three methods may be used to solve heat conduction
problems with internal heat sources, given surface temperature
distributions and time dependence.
6. The floating random walk and the transport equation
approximation may be used to solve problems with convective boundary
conditions.
7. The transport equation approximation may be used to solve
coupled nuclear radiation and heat conduction problems.
The Monte Carlo solution of the transport equation approximation
has been shown to be useful in the solution of many types of heat
conduction problems. However, the following areas still require
further research.
1. Contact resistance between media.
This type of problem should be solvable by the use of an
appropriate albedo medium at the contact surface to transmit or reflect
a fixed ratio of particles striking the surface. This technique still
requires demonstration, however.
2. Other boundary conditions.
In Chapter II, many different types of standard boundary
conditions are listed for heat conduction problems. Only the fixed
surface temperature distribution and the convective boundary condition
problems were solved. Further research is needed to determine
effective means of handling the remaining types of boundary conditions.

114
3. Biasing techniques.
All of the standard biasing techniques available for use in
solving transport problems with Monte Carlo are applicable to this
method. Further research is still needed, however, to determine how
these techniques may be effectively applied in the use of transport
problems which approximate heat conduction problems.
4. Coupling with deterministic codes.
The method should also be useful when used in conjunction with
deterministic codes to perfonn perturbation calculation or to couple the
Monte Carlo calculation to deterministic codes as in the adjoint

31
difference method. Further research is needed to establish effective
and convenient methods to accomplish the coupling with heat conduction
codes.
BIBLIOGRAPHY
BLANK PAGE
BIBLIOGRAPHY
1. T. J. Hoffman and N. E. Banks, "Monte Carlo Surface Density

Solution to the Dirichlet Heat Transfer Problem," Nucl. Sci. S
Eng., 59, 205-214 (1976).
2. A. Haji-Sheikh and E. M. Sparrow, "The Solution of Heat Conduction

Problems by Probability Methods," J. Heat Transfer, Trans. ASME,
89_, 121 (1967).
3. A. Haji-Sheikh and E. M. Sparrow, "The Floating Random Walk and

Its Application to Monte Carlo Solutions of Heat Equations," J.
SIAM, Appl. Math., 14_, 370 (1966).
4. E. S. Troubetzkoy and N. E. Banks, "Solution of the Heat Diffusion

Equation by Monte Carlo," Trans. Am. Nucl. S o c , lj), 163 (1974).
5. T. J. Hoffman and N. E. Banks, "Monte Carlo Solution to the

Dirichlet Problem with the Double Layer Potential Density," Trans.
Am. Nucl. S o c , _1£, 136 (1974).
6. N. t. Banks and T. J. Hoffman, "Continuous Monte Carlo Solution to

the Dirichlet Heat Transfer Problem Using the Double Layer
Potential," 20th Army Math. Conf., Boston, Mass. (1974).
7. N. E. Banks and T. J. Hoffman, "Extension of the Surface Density

Monte Carlo Technique to Heat Transfer in Non-Convex Regions,"
Trans. Am. Nucl. S o c , 19, 164 (1974).
8. T. J. Hoffman and N. E. Banks, "Monte Carlo Solutions to Transient

Heat Conduction Problems," Trans. Am. Nucl. S o c , 21, 204 (1975).
9. T. J. Hoffman, "Monte Carlo Solution to Heat Conduction Problems

with Internal Sources," Trans. Am. Nucl. Soc., 24, 181 (1976).
10. M. E. Muller, "Some Continuous Monte Carlo Methods for the

Dirichlet Problem," Ann. Math. Stat., 2_7, 569-589 (1956).
11. J. J. Thompson and P. Y. P. Chen, "Heat Conduction with Internal

Sources by Modified Monte Carlo Method," Nucl. Engr. and Design,
J2, 207-214 (1970).
12. J. Todd, "Experiments in the Solution of Differential Equations by

Monte Carlo Methods," J. of the Wash. Acad, of Sci., 44_ (1954).
13. S. Hoshino and K. Ichida, "Solution of Partial Differential

Equations by a Modified Random Walk," Numer. Math., 18, 61-72
(1971).
117
118
14. W. D. Little, "Hybrid Computer Solutions of Partial Differential

Equation by Monte Carlo Method," Ph.D. thesis, University of
British Colimbia (1965).
15. L. Fahrmeir, "Extension of the Hybrid Monte Carlo Method for

Boundary-Value Problems," Simulation, 2 H 6 ) , 477 (1973).
16. M. N. Ozisik, "Boundary Value Problems of Heat Conduction,"

Scranton, International Textbook Co. (1968).
17. R. L. Murray, "Nuclear Reactor Physics," Princeton Hall Publishing

Co., Inc. (1957).
18. K. M. Case, F. de Hoffman and G. P. Placzek, "Introduction to the

Theory of Neutron Diffusion," U.S. Gov't Printing Office,
Washington, D.C. (1953).
19. J. Spanier and E. M. Gelbard, "Monte Carlo Principles and Neutron

Transport Problems," Addison-Wesley Publishing Co., Inc. (1969).
20. L. L. Carter and E. D. Cashwell, "Particle-Transport Simulation

with the Monte Carlo Method," USERDA Tech. Info. Center, TID-26607
(1975).
21. R. R. Coveyou, V. R. Cain and K. J. Yost, "Adjoint and Importance

in Monte Carlo Application," Nucl. Sci. 5 Eng., 2_7, 219-234 (1967).
22. H. S. Carslaw and J. C. Jaeger, "Conduction of Heat in Solids,"

Oxford, 2nd ed. (1959).
23. S. K. Fraley, "User's Guide to MORSE-SGC," ORNL/CSD-7 (1976).
24. M. H. Kalos, "On the Estimation of Flux at a Point by Monte Carlo,"

Nucl. Sci. 6 Eng. 16, 111-117 (1963).
25. H. A. Steinberg and M. H. Kalos, "Bounded Estimators for Flux-at-

a-Point in Monte Carlo," Nucl. Sci. S Eng., 44, 406 (1971).
26. H. A. Steinberg and H. Lichtenstein, "Implementation of Bounded

Point Estimators in Point Cross-Section Monte Carlo Codes," Trans.
Am. Nucl. Soc., 17., 259 (1973).
27. W. A. Rhoades and F. R. Mynatt, "The DCT-III Two Dimensional

Discrete Ordinates Transport Code," ORNL-TM-4280 (1973).
28. W. W. Engle, Jr., "A User's Manual for ANISN, A One-Dimensional

Discrete Ordinates Transport Code with Anisotropic Scattering,"
UCC-ND, K-1693 (1967).
119
29. M. H. Kalos, "ANTE 2, Adjoint Monte Carlo Time-Dependent Neutron

Transport Code in Combinatorial Geometry," Appendix G, MAGI Inc.
(1970).
30. S. K. Fraley, "A User's Guide for LAVA as Issued with MORSE-SGC,"
0RNL/CSD-11 (1976).
31. T. J. Hoffman, J. C. Robinson and P. N. Stevens, "The Adjoint

Difference Method and Its Application to Deep Penetration Radiation
Transport," Nucl. Sci. 6 Eng., 48, 179-188 (1972).
APPENDIXES
APPENDIX A
VERIFICATION OF ANALYTICAL SOLUTIONS
Once an analytical solution has been developed for a heat conduction
problem, the solution nay be verified by showing that it satisfies the
heat conduction equation and the boundary conditions. Many of the
problens presented in Chapter IV involved the derivation of an appropriate
Green's function to obtain the analytical solution. Since a large number
of manipulations (which could introduce errors) were involved in the
development of the solutions, a verification of the more difficult
solutions is given in this appendix.
The analytical solution to problem 2 is
S a ,
T /•„„,•» - ° 1 1 1 . iry . irz
T (x,y,z) - — • _ - g -
2 . 2 j - . ^ j - 3 7 2 • s i n - f sxn T
b • c b + c
1/2
sinh 11 [(a • x){±* -j) ]
1 (A 1}
r r Vî
h "
c o s h [ i r a ( - j • —jj J
b c
for - a < x < 0 , 0 < y < b , 0 < z < c , and
S
~ . -. _ , 1 0 1 s i n
. i T s Xi n . i r ys i . n T r z
T j ( x , y , z ) = T ( - x , y , z ) • --—
2 . - j - ' -~- 5 —
71 K +
l T * 72 "7
3 b C
(A-2)
for 0 < x < a , 0<y<b, and 0 < z < c.
123
124
Since the problea is steady state, the solution must satisfy the
following:
S x z
CD v^T ^ - { rY' ) , (A-3)
where
S(x,y,z) = S sin —- sin ^ sin ^ , CA-4)

o • o c
for 0 £ x ^ a, 0 < ^ y < ^ b , 0 < ^ z < ^ c , zero elsewhere,
(2) T(x,y,z) = 0 , (A-5)
for all (x,y,z) on outer boundaries, and
3Tj 3T 2
( 3 ) K
l 3x~ 2 JT '
= K
(A-6)
at the interface between the two media.
By inspection it is obvious that T(x,y,z) = 0 at outer boundaries.

To evaluate v^T (x,y,z), let 2
1 / 2
T (x,y,z) = a s i n ^ sin ^- sinh[ir(a* x ) ( - ^ * - ~ )
2 ]. (A-7)
b c
Then
2
3T
2 ,w.»2 . iry irz . . , , .,1 1.1/2, ,. „
— y - -oC-g-) sin ^ s m — smh[ir(a • x) ( - j -• -y) ' ] , (A-8) %
3y b e
125
2
2 ,ir.2 . iry . irz . . , . , 1 . 1 ^1/2, r ,. ,
n
— j - = -<x(-) sin -£- sin — sinh[ir(a + x ) ( - j + —) ' ], (A-9)

3z b e
and
2
a !.2 _
22 ,, 11 11 .., . iry
iry . irz
irz .. . . ,,
r .., ,11 11,.1/z,
J
.1/2 ,. l n ,
.2 = air \2
(-^r- 2' " " b " " c " — i " * - -^2 2
3x b • -zf)
c sin -if- s i n — sinh[ir(a + x) (-=•b + e- 7 ) ]• (A-10)
Therefore,
anr. an .2.
3'T
-
V^T (x.y.z) = — y - + —f- * —f- « 0. (A-ll)

3x^ 3y 2z
2
To evaluate V TjCx.y.z), i is only necessary to evaluate the t
second term of T.(x,y,z), which will be denoted T (x,y,z), since

2
v T2(-x,y,z) = 0. This gives
S
T r i ° 1 .irx.iry.irz ,, „ ,
T ( x
s ' ' y z ) =
-T7 1 1 r sin
T sin s i n
b T ' {A_12)
a b c
This gives
2
3 T S
S -IK 2 0 1 irx iry irz
= s i n s i n s i n (A_13)
71- - 1 7 1 — 1 — r T b T -
a b c
2
3 T S
S ,TT 2 0 1 .irx.iry.irz ,. ,.,
-2"--( ) -T--! F ; j-sin — s i n -^ sin — , (A-14)
3y K + +
* l -a J b - c
and
126
2 1 *x *y rz
-c-) - a T-
1
sin — sin -i*- sin —
a b c
(A-15)
3E V _2„
¥ K
i T+172+."51
a b c
Therefore,
_ .irx.iry.irz
S sin — - sin -sf- sin -—
o a b c
v*T (x.y.z) = (A-16)
and
^(x.y.z) = - *l£g*lL . (A-17)
3T.
The final condition will be checked by evaluating K- ^-~ and
3T x=0
. and setting then equal to each other to see if the equality
^2 3x
is verified. Performing these operations gives
1/2 2
K,acosh[ira(-V
.2 • ±)
2' ]H±"-, 2 • 2-'
V ' J »» ? sin 2-
b e b e
„ ,1u l%l/2, , 1
r 1 .1/2 . iry . irz
K. ocosh[ira(-^ + - j ) J^C-j • —j) s i n -£- sin —
b e b e
S
o 1 ,ir. iry . irz (A-18)
( s i n s i n
T i — 1 — r i> b r
a b c
where
a=
V 1 1 1
2 K K 2 1 1 l
ir
1 + 2
1 • *\Ar + - j)
b c
1
(A-19)
/2
cosh[ira(-J=- • -~y )
b c*
127
or
S a
o 1 (A-20)
» -L • -L • -L '
b c a 2 . b 2 2c
which checks. Therefore the analytical solution to problem 2 is
verified.
2 2
m
vr_ r ^ n
-.1/2 ,
16 S
T (x,y,z) I I cosh[ira(-2 • —~) - 1-
2
4 «. t. 2 2
ir (IL+K-)raodd n odd /•• n
•»
b c
• t. r_ /•"! . H -wl/2,
v
mn(-_- • -^) sinh[ira(-j * —) 1
b e b e
2 2 .,_
smh[ir(a + x)(-^- • - j ) ]
b b miry nirz
- 2 j — sin - ^ sin — (A-21)
1 / 2
cosh[»a(2 * % )
f ]
b c
for - a < x < 0 , 0 < y < b , 0<z<c, and
00
64 S °° °°
T^x.y.z) = T (-x,y,z) • 1 \ \ \ 1
2
K- TI I odd m odd n odd „2 2 2
a b c
i-nx miry . nirz

sin sin —r*- s i n (A-22)
a b c
for 0 ^ x _< a, Q <_y <_b, and 0 <_ z <_ c.
The solution must satisfy the following:

128
2 S x z
(!) V ! = - <. >p ) , CA-23)
where
S(x,y,z) = S Q , (A-24)
for 0 < x < a , 0<y<b, 0<z<c, zero elsewhere ,
;2) T(x,y,z) = 0 , (A-25)
for all (x,y,z) on outer boundaries, and
(3) K ~ = K ^2
1 3x 2-^- , CA-26)
By inspection it is obvious that T(x,y,z) = 0 on outer boundaries.
To evaluate 7 T (x,y,z) let 2
1 / 2
. . <4 * 4 •
1,111
\ D2 JZ
c
(A-27)
Then
2
16 S cosh(aira ) - 1
I I mn
sinh(airo )
3x< if (K.+K-) m odd n odd mna mn
mn'
sinh[n(a • x)o ]
miry . mrz
*• mn sin —r*- sin (A-28)
cosh (aira ) b c
mn
2
3 T„ •16 S coshfaTra ) -
m
2 2
I I „K 2 2
sinhfaira )
3y i r ( K + K ) m odd n odd
1 2
v
mn'
nb a
mn
smhfirfa • x)o 1
1 v J
mn miry nirz
r-f—i =; s i n -r*- sin (A-29)
cosh (aira ) b e
129
and
2
3 T„ -16 S cosh(a-av ) - 1
cur
2 2 I I 2 2 sinh(aira )
mc a
3z IT (K.+K^) m odd n odd mn ni j
smhfirfa + x)a 1 J
mn mirv mr: (A-30)
r-i ^ s i n ——- s i n —
coshfa-rra )
v b e
mn'
Using the definition of a gives

" mn "
in n = 0 , (A-31)
mn ,2 2 22
nb a mc a
mn rn
and
7 T (x,y,z) = 0.
2 (A-32)
To evaluate V T (x,y,z), it is only necessary to evaluate the second
term of T^x.y.z), which will be denoted T (x,y,z), since V"T,(-x,y,z) = 0.
This gives
64 S r 1
T (x,y,z)
s
K TT I
I I I
odd m odd n odd r 2 2 2
1
«xin(— • — • -j]
b" c
JITTX . miry nir

sin s i n -—- sin — (A-33)
a b c
This gives
130
2
3*T 64 S .2/ 2
4 /a
I I I .2 2 2
3x K. IT I odd m odd n odd £m(~=- + -j • ~2)
a b c
iirx airy nirz

sin sin -r»- sin (A-34)
a b c
2 2
3 T j T
and similar expressions for s_ and s_. Combining these gives
2 2
3y 3z
o4 S °° <*> <*> ,
o r r r 1 . iwx rawy nirz
7 T 2
) > > - — sin s i n —zf- sin .
K. IT *• . m odd
I odd »• £.
n odd Inn a b
n cr
(A-35)
To evaluate t h i s expression, consider a Fourier expansion of the
number "1" over the region 0 < x < a , 0<y<b, 0<z<c. This gives
, r r r . JUx miry nirz

(A-36)
1 = 1 I I A «™ — s i n - L s i n —
£=1 m=l n=l
where
a b C
O 6
.
A =
8 r t r s i n
. Jtirx miry nuz . . . (A-37)
» ru~ i l l a
sin ~rf- sin dzdydx
c
£mn abc 0 n 0 ^
Using
odd
f *
I sin dx (A-38)
0 I even
gives
64
£,m,n odd
Imnit"
(A-39)
Iran otherwise
0
131
Therefore
2
VT = - -5- (A-40)
s K * x
The final condition will be checked by evaluating K

3T„
1 3x and 5
x=0
E
K —- and setting them equal to each other to see if the equality
2 3x
x=0
16 K„ cosh(aira ) - 1
nn miry mrz
I I sinhfaira ) s i n —rr- s i n
ir (Kj+K_) in odd n odd mn v
inn'
b c
-16 K coshfaira ) - 1
1 mn miry . nirz
n (Kj+K-)
I I
m odd n odd mn sinh(aira )
ran'v
sin —Tf- sin
b c
64 1
— . I I I s i n
miry
S i n
nirz ,. . , ,
A 4 1
air % odd m odd n odd , £2 m2 n~T~

, "b ~c~ ' ( ~ )
m n { _ + + _)
a b c
To verify t h i s equation, i t i s necessary to verify that
jr_ ;osh irB - 1

7 2 ' (A-42)
4g sinh TT8 L
I odd I +B
where
= a o
8 - rrm •
m (A-43)
To verify Eq. (A-42), consider a Fourier cosine expansion of

cosh B(TT-X)
6 s i n h BIT
•. U s i n g
132
JcoshBOr-x) c o s ) u d x s i_^ A . 4 4 )
B s i n h B w 2 2
0 i • 6
gives
C S h g ( x )
° . ;- = -iy • I l _ i _ cos .tx . (A-45)
B S i n B 2 W 2 2
* , 1=1 £ * 6
B
Setting x = 0 gives
CD
cosh Bn 1_
B s i n B
* =
*B 2 +
I TT^-
»-i * i + e 2 2
(A" ) 46
Setting x = ir gives
_ _ J _ - J _ . ? 2 (-1)*
B s i n h B 2 2 2
* *B J l " £ * B "
Subtracting Eq. (A-47) from Eq. (A-46) and multiplying by -j gives
00
It COSh 1T& - 1 r 1
(A-48)
40 sinh *B % ^ %2 + 6 2 »
which verifies the equality. Therefore the analytical solution to

problem 3 is verified.
. , , . , 1 1 .1/2, ny irz
„ smn[ir(a • x) (-=• • - j ) ] sin -£- sin —
T , . 1 b c
2 K + K 1 / 2 1 / 2
l 2 sinh[ira(4r • 4 ^ ) ] cosh[ira(4=- • 4 r ) ]
b e b e
(A-49)
for -a < x < 0, 0 < y < b, 0 < z < c, and
133
. , , 1 1 ,1/2, r
sinh[i:x(-j + —) J
T (x..;,z) = T (-x,y,z) +
2 \ ^-172- " n f sin ^ (A-50)
smh[Tra(-2 + - j ) J
b c
for 0 < x < a , 0 < y < b , and 0 < z < c .
(1) 7 T=0 , (A-51)
itz
IL sin ^— , on surface at x = a,
sin -£-
b e
(2) T(x,y,z) = (A-SO
, on all other external
boundaries
and
3T 3T
(A-53)
™ l IT ' IT '
K
By inspection it is obvious that the external boundary conditions
as given in Eq. (A-52) are met. To evaluate 7 l„(x,y,z) let
T_(x,y,z) =asinh[ir(a + x) (—- + —~) ] sin ^ - sin — , (A-54)

2 2
K
b c
then from Eqs. (A-7) and (A-ll)
7 T (x,y,z) = 0
2 (A-55)
To evaluate 7 Tj(x,y,z)^ it is only necessary to evaluate the second
term of T,(x,y,z) which will be denoted T (x,y,z). This gives
T ( x , y , z ) = B sinh[TTx(-y + —=•) J s i n -£- s i n — . (A-56)

134
Therefore
2
3T
2 1 / 2
— ^ = B * ( ^ • -^) sinh[wx(^ + - ^ ) ] sin ^ sin ^ , (A-57)
3x b e b e
2
3 T 2
s „ w . . ,\ 1,1/2, iry irz ,,
— s - = -0 -=• sinh[irxC-=- • -_-) ] s i n -£~ sin — ,
r
(A-58
c o
3y b b e
and
2
3 T 2
s 1 / 2
= -6 ?j s i n h f r x ^ • - 3 ) ] s i n ^ - sin ^ . (A-59)
I?"
This gives
v^Tjfx.y.z) = 0 . (A-60)
3 T
The final condition w i l l be checked by evaluating K. -?— and

H
1 3 X
>x=0
K
2 3x and s e t t i n g them equal t o each other t o see i f the equality
x=0
i s v e r i f i e d . Performing these operations gives
, 1 1 .1/2 iiy . TTZ

K K +
2 l 7 7 sin-^sin = T
1/2
V*I sinh[L(-L*-7)
b c
]
, ., i • y n .u a,,. 2
1 b c
1 2 sinhliraC-j • —j) J
b c
„ , 1 1 .1/2 iry . irz
K * ( - j * —) sin -f sin —
• £ c (A-61)
. . . 1 1 .1/2,
r
sinhtitaC-j • — ) J
b c
135
which checks. Therefore the analytical solution to problem 4 i s
verified.
The a n a l y t i c a l solution to problem 5 i s
i \.*,y,7.)
2 2 \ 1 1/2 , 2 . 1 l .
( + 1 + a ( +
* 72 T> 72 I*
b e b e
1/2
(Kj+K.JcoshtTvat— + .L) ]*K X
b c
2 1 1 1/7 2
1
K *K )(K *K )cosh^[rra(-L • ^ . ) ^ ] _
1 3 1 7 K ^
b e
. ., , .,1 1 -.1/2, . iry . irz , . ,_..

• sinh[ir(a • x) (—=• • —=•) Jsin ~ sin — , (A-62)
b c
for - a < x < 0 , 0<y<b, 0<z<c,
S a
o 1 1
T (x,y,z) = 2 , 1
3 K,„ , 2,1 1
1 / 2 1 + a
* <72*-2> ^ 2 * ^
b e b e
1/2
(K^K^coshhaC-j • -^-) ]*K 1
b c
1/2 2
(K •K )(K *K )coshM*a(-^ •
3 1 2 -^ )~K
b c
1 / 2
•sinh[ir(2a - x)(-^- * - ^ ) ]sin ^ sin ^ , (A-63)
b ';
for a < x < 2a, 0 ' , < b, 0 < z < c,
and
136
Tj(x,y,z) = T (-x,y,z) • Tj(2a-x,y,z)

2
S
o 1 .1rx.1ry.1rz ,. ,.
+ sin s i n s i n (A 64
ST -i—i—r
" l 1 72 T
K + + T b T - "
a b c
for O ^ x ^ a , 0 < ^ y < ^ b , and 0 <^ z <^ c.
The solution mist satisfy the following:
2 S ( X Z )
(1) v T = - ^ , (A-65)
where
S(x,y,z) = S sin — sin -#- sin — (A-66)

"* '
v
o a b c
for 0 < x ^ a , 0 £ y ^ b, 0 < ^ z < ^ c , and zero elsewhere,
(2) T(x,y,z) = 0 , (A-67)
for all (x,y,z) on outer boundaries,
3T 3T 7
at the x = 0 interface, and
3T 3T,
(4 (A 68)
> hir-hiT' -
at the x = a interface.
By inspection it is obvious that T(x,y,z) = 0 on the outer

boundaries. From the verification of problen 2 it follows that
Eq. (A-65) is also verified. The condition at the x = 0 interface will
3Tj 3T 2
be checked by evaluating K. — - and settinq them 3 n d K
x=0 2 3T n
=0
137
equal to each other to see if the equality is verified. Performing
these operations gives
1,1/2 1
(K +K )cosh[ ra(-^- + - y )
1 2 1
1 / 2
] (Kj+KjDcoshliraC-L *-j)
S
2 1/2 2
(K K )(K K )cosh [Tra(-^ * ±) )
1+ 3 1+ 2 - ^
1 1 1/2 2
1 / Z
K^Kj+K^cosMnaC-y + - j ) ] • Kj
b c + (A-70)
V 2
(K^^CKj+K^cosh^TraC-^ • \ ) ] -K^ '
which checks. The final condition at the x = a interface will be

3T
checked by evaluating K and K and setting them equal
1 3x 3 9x
x=a x=a
to each other to see if the equality is verified. Performing these
operations gives
l/2
(K K )cosh[Tra(-y • - j )
1+ 3
1 7 2
] (K^Kpcoshhaf- .-* -j) ] 1
*K
b c
7 1 11II
(K +K )(K +K )cosh'>a(-y • \ } ' } - K^
1 3 1 2
b c
1/2 2
^(K^KJcoshlnaC-^ • -^-) ] • K j
v i "3^ b c
+1 (A-71)
(K K )(K K )cosh [ira(-^
+ 3 1+ 2
2
+ -Lj 1 / 2
] - K^
b c
which checks. Therefore the analytical solution to problem S is
verified.

138
S nl J (a.r) sinh a.(1 + z)

T-(r,z) = -= =-^-— ° * i— , (A-72)
2
2
w •l\ l l 22
V a (K + K } C O S n
for 0 <_ T <_ a., -I <_ z <_ 0,
and
S
Tjd-.z) = T (r,-z) • 2 2 _ J ( r)sin ^
o ai , (A-73)
K 2
l^l * ^
for 0 <_ r <_ a, and 0 <_ z <_ t.
The solution mist satisfy the following:
2
(1) 7 T=-i£ii, (A-74)
where
S(r,z) = S J ( ) sin { £ ^
o o V ( A ? 5 )
for 0 <_ r <_ a, 0 < : £ I, and zero elsewhere,
(2) T(r,z) = 0 , (A-76)
for all (r,z) on outer boundaries, and

3T 3T_
3 (A 77)
<> hir-hir* -
By inspection and from the definition of a. as the first zero of

J (aa)» It is obvious that T(r,z) • 0 on the outer boundaries.
139
For cylindrical geometries with angular symmetry
„2 , J. j_ J, il_ i r
(r J
+ (A-78)
r 3r 3r ^3z2 '
Using
l r J a
- l o °l J ( r )
' (A-79)
r 3r 3r
and
3 2
—2 s
^ n h a
i( a + z
) = a
i Si"* 1 a
i( A + z
) (A-80)
3z
gives
v^Cr.z) = 0 . (A-81)
To evaluate 7 T (r,z), it is only necessary to evaluate the second
term of T.(r,z), which will be denoted T (r,x). This gives
*2y c i =
v T s (r,z)
S
o o l n J ( a r ) s i n
V (A-82)
* x
Therefore
S(r,z)
v'TjCr.z) (A-83)
3T,
The final condition will be checked by evaluating K. 5 — and
3T z=0 1 a Z
K- K—I and setting them equal to each other to see if the equality
!
z=0
140
S K, Hi S irfc K.
o 2 , . v _ o 1 , . r l
2 , 2 2 K, + K, " o l ^ T
2 . 2 2 It + L " o 1
A
IT + t a . 1 2 ir • J. a . 1 2
S *
o
* _ 2
J" o(°,r)
*l
, (A-84)
»^ 2 IT ,
(
* °1 *^

verified.
2S„o • cosh anI - 1 Jo(a

n r) sinh an (JUz)
v^'-ii^^J, ¥ , . sinh a nI cosh a n£
an3 J,(a
l na) '
v
(A-85)
for 0 < r < a , -«. < z < 0
and
8S J (a r)
o* n '
T , . „ . T (r,-z, • ^
l( 2 I i sin H i I
m odd n=l „ ,m T +a 2 )J. fa a)
K,a (—=— K
1 n 2 n ' \ n v 7
(A-86)
for 0 <_ r <_ a, and 0 £ z <_ i.
The solution must satisfy the following
( 1 ) 2
V T= - *&£L (A-87)
where
S(r,z) = S, (A-88)
for 0 < r < a , 0 < z < «., and zero elsewhere

141
(2) T(r,z) = 0 (A-89)
for a l l ( r , z ) on o u t e r b o u n d a r i e s , and
3T 3T_
( 3 ) K K f A 9 0 )
1 3 T - 2 i r "
a t t h e i n t e r f a c e b tween the two media.
By i n s p e c t i o n and from the d e f i n i t i o n of t h e a as t h e zeroes of
J ( a a ) , i t i s obvious t h a t T ( r , z ) = 0 on o u t e r boundaries. Using
3J ( o r )
r 2
~ T- C a ) = -or J (a r ) (A-91)
r 3 r 3 r n o n ' *• '
and
2
3 s m h a (I + z)
l
= = a sinh a (£ + z) (A-92)
.2 n n
3z
gives
2
V T (r,z) = 0 .
2 (A-93)
2
To evaluate V T ( r , z ) , i t i s only necessary to evaluate the second
term of T . ( r , z ) which w i l l be denoted T ( r , z ) . Using
2 2 2
3 . mirz m TT rnirz , . ,._
— j sin - j - = =- sin - - 1 (A-94)
3z JT
gives
- 8S » » J (a r )
V p z = s i n ( A 9 5 )
V ' > " IT" * Z mKa J , (a a) T" ' '
v
m odd n=l In l n •*
To evaluate this, consider a series expansion of the number "1"
over the region 0 <_ T <_ a, 0 ^ z £ «.. This is given by

142
1 Si J ( (A-96)
'111Ji Ji *"" " " ° V '
=1 n=l
where
a £
n m 2
A__
Bin
- -^—i=
2 2
/ '/ sin
J si- -9— Jn *(ar)rdzdr
n* w v >
(A-97)
z n
a * J x (a a) 0 0 * °
Using
— m odd
miT
J sin - j - dz = (A-98)
0 m even
and
J r J ( a r ) d r = f J a a)
o n l ( n
(A-99)
gives
mn mira J, fa a) (A-100)
l n ' v
Therefore
V^.(r,:
T (r,2) - - - *
s
(A-101)
3T,
The final condition will be checked by evaluating K, ""
- —1 and
3T
C
z=0
2 TT and setting them equal to each other to see if the equality
z=0
143
2S K_ - cosh a i - 1 J (a r)
o 2 r n _ o n
sinh a I
a (
W „ = l a J ra a) 2
n
n I n
2S K. - cosh a - 1 J (a r )
oo 1
1 y
r nn o n
, . s i n a. I
1
< 1 2 > n i l a j„ r
K +K
. (a
. a) n 2
y
n \ n '
8S °° "> J (a r ) J
o^ n
T (A-102)
•IT m odd n = l
I
a (—— + a ) J , ( a a) v
n „2 n ' l n
or
I y cosh a , ! - 1 J ( a r ) o n
4 , 2 , T. sinh a I
n=l a J , ( a a) yn
n l n •*
00 00 J (o r )
o n '
(A-103)
2 2
m odd n=l
a + a J
n^T n ) i(V)
To v e r i f y t h i s e q u a l i t y , i t i s n e c e s s a r y t o v e r i f y that
cosh a I - 1
£ 0
n 1
(A-104)
4a sinh a I ' 22
n n m odd m TT 2
+a
T2~ n
la
or, l e t t i n g 8 =
TT cosh TT8 - 1
(A-105)
46 s i n h Tf8 ' 2 2
m odd m + 8
But, t h i s i s p r e c i s e l y Eq. (A-42) which was v e r i f i e d in checking t h e

144
solution to problem 3. Therefor* the analytical solution to problem 7

is verified.
K 1 sinh a. (J. + z)
T_(r,z) = V " 4 P - J
(°i ) -*-z
r
? i: T (A-106)
2 ''
V
i* 2 K K
1 s i n n
Oi* c o s n
"î*
for 0 < r < a, -I < z < 0,
and
sinh o z
V r . z ) = T (r,-z) • J ^ a ^ ) ^ ^
2 (A-107)
for 0 <_ r £ a, and 0 _< z £ 4.
2
(1) 7 T = 0 (A-108)
a
J ( ,r), on the surface at z=£
(2) T(r,z) = o 1 '' v
(A-109)
0 , oi» all other external boundaries
and
3T 3T„
( 3 ) * i.* ' ( A . U 0 )
1 3z 2 3z '
By inspection it is obvious that the external boundary conditions
as given in Eq. (A-109) are met. Using
( r J (a r)) 2 } A U1
7-k £ o i • -°i W • ( - )
145
and
2
3 2
s n n 2, + z = a s i n h a i + z
—_- * "it ) i i( ) > (A.-112)
3z
gives
2
V T = 0 . (A-113)
3T ,
The final condition will be checked by evaluating K, - — i and
1 32 'z=0
K and setting them equal to each other to see if the equality
2 3z
z=0
K K K 2 J j a r
i 2 W ) i o<v> K o t k „..
a
Kj+K l sinh a ^ "
2 Kj K 2
a
l sinh a^ * 1°1 sinh a.^ ' ( U
*'
verified.
1
BLANK PAGE
£
APPENDIX B
DERIVATION OF A BOUNDED NEXT EVENT ESTIMATOR
The following derivation will be made assuming a monoenergetic flux
and isotropic scattering. These assumptions are made for simplicity of
notation only. The results are not changed if the flux is assumed to be
energy dependent and if the scattering is anisotropic.
The flux at a point r=0 can be represented by
• (r=0) = / E s * a v g ( r ) 4*T 5 _ _ dr • / S(r) 4»r ^ dr (B-l)

r 4nr r 4irr
where
J / • dA
fr) = — (B-2)
avg' //dA
and A is the surface area at r.
The standard Monte Carlo unbounded next event estimator uses each
fcr
2 Es e"
collision as an estimate of E A (r)4irr and scores =• at each v
t avg ' _ 2
E 4irr
collision site. The estimator is considered unbounded since the score
is unbounded as r approaches zero.
This problem can be eliminated by scoring all collisions within a

-Er
sphere of radius "a" by -—=• where d is calculated from
J
4iTr
-Er -Er
z 2 d r l ( r ) 5 d r ( B 3 )
/ s * ( r ) *™ ^-J
a v eg z
= / s » a v Bg
4 I I T 2
—;z • "
0 4nd 0 4nr
147
148
This calculation depends upon the flux shape assumed for <f> (r). In
this appendix "d" will be calculated as a function of "a" for (1) a flat
flux and (2) an exponentially decreasing flux.
Assume •(r) = $ , a constant. Then 4>
0 (r) = $ and
Q
J r2e -Er,
dr
2
d = °- (B-4)
e dr
Evaluating the above integrals gives
e (1 • ) r
(B-5)
1 -e
where \ - la is the radius of the bounding sphere in mean free paths.

This can also be written in series form as
( n 1 } 2 ? n
i C-D" : ^ >
, 2 ^t n (n+3)[ (B-6)
d = a <*> _ ,n
n A
I
n=0 c-u (n+1)!
-Ex
r i s t n e n
Next assume $ = <J»e 0 •~„„( ) given by
w zx (B-7)
•«« • i? y *°*' "
2 +y
where A is the area of the sphere with radius r and r = (x 2 +z2 )1/2.
Evaluating this integral gives

149
r ) =
slnh(Ir) , (B-8)
W *° Er
and
*r 2 sinh(Ir) -£r.
1J r —=—»—'- e dr
Er
,2 0
(B-9)
sinMIrl e -Er d r
Zr
Evaluating these integrals gives
— + fl+c)
\22 2 2 v J
d =a (B-10)
Ei(c) + in(c) • y
t
where c = 2la,
Y is Euler's constant, and
Ei(c) is the exponential integral function of c.
This can also be written in series form as:
y , n n (n+2)c n
K
>- ' fn+3V
{ n Jm
2
A 2 n=0 *
d = a CB-11)
n
y r-n £
n =o (n*l)(n*l)!
2 2
Table 20 g i v e s d / a v e r s u s Z f o r both a s s u m p t i o n s , i . e . , 4>=4> and
n
$=<<> e
o
Note that for bounding spheres with radii up to one mean path in
2
length the flux shape used has very little effect on d , and therefore
on the score used in the estimator. In estimating the flux at a point

150
which is not near a strong point source or an external boundary, the
above aethod should therefore be applicable for bounding spheres up to
one aean free path in radius since the flux shape will usually be
soaewhere between a constant and exponentially decaying.
TABLE 20
2
d /a 2
VERSUS l a
la 2
d /a 2
cfor L *
$=4> e -Ex
(•fp) for <J>=<|> 0 0
.1 .3250 .3252
.25 .3127 .3136
.50 .2925 .2960
1.00 .2541 .2665
a - a — ^ a—r=
This procedure does not require a modification of the random walk
procedure and may be implemented very easily into Monte Carlo programs
which use next event estimators.

APPENDIX C
LISTING OF COMPUTER INPUT DATA AND SELECTED OUTPUT
The input to MORSE-SGC and its output are given on the following
pages for the coupled gamma radiation and heat conduction problem solved
in Chapter VI.
The listings of the analysis routines are given on the next three
pages. The input data to LAVA to generate the cross sections and to
MORSE-SGC is given on the following page, and the remaining pages in
this Appendix are MORSE-SGC output.
151
LtVlL 21.6 <0<C »2I 0»/)6J fOW»«N H 1 » Tt T».I 16/OV.JS.•0
COMPILS* OPTIONS -
NANta MAlN,OPTP02...INCCNTa60.SI*E'OOOOX a
SOURCI . « e C O I C . N O H S r . N O O C C K . L O A O , . ) A P . N O e D I T , < . 3 I O . N O X » E F
ISM 000 2 SUBROUTINE NfLCOHO.XO.YO.fO.VEL.CXTRO.EXTROI
c
C SUBROUTINE RCLCOL I S CALLCO AFT I B ( A C M C O L L I S I O N OJA I M O
C T H t RANDOM M I X
C
c
ISM 0003 COMMON/INPUT/I AOJN.NSTRT.NMGST.NITS.NOUIT.NCQLTP. I l f AT. N S P . T ,
2 NftlLL.M»AST NOLCAK IEalAS NKCAlC NOIIM*.
a a a a
• MCDIA.NNI X.MEDALB.MXREG.MPlSTP.NNGA.NaCA. NN6TP.

S N 6 O P T , l 0 « O P T . N O S N . N 0 S e > N C 0 H * , N I C T , M A X 0 P , IROSO* 1 S T * ,
• | P M U . I M 0 M , I P R I N . I P U N . I X T A P < , , M T A P C . 1 0 6 R , IOOPT,
7 I SOUR.NGPPS.I s a i A S . N S O U f t .
• NO.NNE.NE .NT.NA.NACSP.NCX . N C X N O . I F L A G I 2 0 I ,
I TMAX.TCUT.TJTSIRI.AGJTRT.X.TAT.Y&TAT.miT,
I 1>INP.VINP,«INP.NXPN,NH|STR.NH|SMX.NSINOO«' .NCO^.SI I I I .
I N 4 . I P T e . I G O . N N U C . I O T . N R r . N i M . M M . N S 6 P S . T I T L f i 1 0 I . OAT « * >
I . JFTG.KPT&.LFTi
ISM 0004 COMMON / P O I N T / L F P I . L P P 2 . L P P J L P P * . L F P 6 . L P P 6 . L P P 7 . . F » » , . PPT),
a
I LPPt O . L P P t l .L»P»» . L F P I J . L P P I * . L P P | S . L F P | 6 . L P P | T , . F » I S . . F P I 9 .
I LPP2 0 . L P P 2 l L ' P 2 2 . L P P 2 3 . L P P 2 * . L A P | . L A P 2 L A P 3 L A P 4 . . A P S
a a a a
I LAP6.LAPT.LiPl.LSPa.LSP3,
I L'P2S.LPP26..LPP2V a LPPa« LPP2'>,LPPM..r»3l,LPPJ*.
a
000 S I LXI.LX2.LXJ.LX« , L l l .LT9

ISM COMMON / M U T f t O M / N A M t .NAMCX < l 0 . 1 0 0 . N M t O t MCOOL 0 , NRSO, J . V . « . UCK. 0 , V 3 . 0
ISM 0006 i . • C H . o . x . v . i . x o i s . v o L o . j O L O . a A r e . O L O p r . « r 0 0 « . * M T . B . 2 3 N , A « I . SLOAOC
C ONMON/APOLLO* P .OOP . 0 1 A0«T IS I « ST A. CT AT H. CT AJSO. XT RA < 10 I .
I I T C R 3 | T | « e . | T S T R . M A X T I P.MGPftCG.NALe.NOCADt*),NC«NM.NGCON,
a
I NtAST.NMZM.NMdP.NMTS.
ISM OOOT 3 NPSCLOJt .NSIOL.MXTRAtlO.
COMMON/OOI T/i.ocRSP.kOCAO.i.ocia i,occo «ocT Locjo.LOcao. a a a
ISM 000* I LOCae.LOCOT.LOCOTi.LOCQAe.LMM.ePIMT.KiTOP

COMMON/GOMLOC/ KMA - M P O • K L C f t . K N a O . K I OR. K R I Z . K R C Z . < M I I . X R C I .
ISM 0009 I KfcRI . K K R a . K N S R . K V U L . N A D O . L O AT A . LT M A . L P P O . NJMR. I R T R J . M J M R . N l 4 M
ISN 0010 I N T E G E R CXTRO Cn
o i M t N S i O N o i 11 . x o i i . . T O O I . ^ O I I t . Y E L I I I . e x r a G i N M T G . i t . K>
ISM 00 i I I CXTROINO.lt
c CNTRV S 5 A R I O . X O . Y O . i O t
c I N TRY K M I S C A L I C O » H C N A SCCONOAHY P * R . H.L.E ! * G t M t r f A T C O
c
c
ISM 001* I F M O . N C . 2 ) R! TURN
c
c ONLY SECONDARY P A R T I C L E S I T t I C GAMMAS ARe | N GROUP a IS THIS
c A D J O I N T F O R M U L A T I O N ! ARE SCURCO
c
c
ISN 0016 OO 10 1-1.ND
c
c DETECTOR L O C A T I O N S ARC USCO TO I N O I C A T C T HC DIRECT l > :
c TO PAR 0 THC I N C O M I N G B E A M . THE O I F F C R C N T D I R E C T I O N S A t e
c C H O S i M TO G I V E O I F P C R E N T O R I E N T A T I O N S P I T H RESPECT TO THE
c STARTING LOCATION
c
ISM OOIS m»»*xnn
ISM 0016 Ytt"V«YD-•
I ]
So
rflh-
i a j n
Ouiws
"»
W
z -
•
tfl^C
IfvJOW
o •o • < -o
••• - J •> • — 1
O N . 4} Z h *
*
n*
© *
»-—o
•J«
• azo
a
•• n— Oa i
\3<W
—u» •
-iJzC
*»
•* ^•
•*-5
n T
« 0 — •'S
W
~o~ l u »*u
«JC»- — DO
^> rt•<•
?*•-! OO—OJ
. . J —
*- o
«
—o
o -
• u :
• M
w u«^—a
J J « 0 •
r:a*
-—WZ
J W
y^
j»-o
• NS.Z->
<n°..J>.2Z'a
«« oo—a
- — < < ? * * l l — O t L J O "WV.V. *f r > O Q O i u 2

sasss> * # * u * > v i t — u I — — u — Ol u j l k i
"* n o
1
rt »"» A #« A «% rt rt A ^ * A A rt rt M J * rt * \ rt A ^* rtrtrt*%
zzzzzzzzz ssssssssssss ssss yyfy-r-Jt

154
ou
-ll»
o<
>o
•1
m
K
O
•
O • »
•z
>o
• »i
-» I
.a
<3 .
I t Ul4l
H
OZ
Z"*
oft 1
XT
MM
«t
z • < ••
M a *
oc
z
a< 53 »*
•#
• O O
u
zz
MM
so O J
b
••
& * ; o•o.
III
!S I t
In
>>n o •>ai -•)
?! • « fi
X
•• • VS»
6 • M *™oi n &
«~
© — «l w»W
•O
>-z 5 o z
12 a
« < MMX
• • f
z1 ^
u <<«!_•
SJ
»••
• J l
UO •
_«i *•*••
•
IC *o
ILU M M
z o * > z • ISI «*
« •>
sis• O X 0 3 Z
ZuO ZoJ •l o z z
• • •r>«f
->Z«
**-
ooa*
«»•• •U*
•I3Z I l l l rtT«i<
M0ZKX* b»-<» r »-»-•-• —_m>
4uO ».z 0 •IrlfciiLi
• uJO.J •
5£2i • « S x x 0 > « 3
llMMw
1 •• «X
»•• till •>*M>S««0
. . .••o»
033u(
azoza !i | !0 <l •;•=:•
> O** :**«s
1- 3iHK u ' v i i ' t i u ' * ' • • • • 5z
X~x" l i i x i - j x ^ o
•• Mu
o -Oil
us sssi J U ; ^ 5*£O-O0-5" • l l l l l l l '
-a n>« <iauii«>i3«3>li«
v
w J
W "•
o a
" S
• oo <>'> io^<*^<«^—MrfnnffliAjrifiAiflfl(«nKin*)nnn
- oo o<< > o o o o o o o o o o o o o o o o o o o o o o o o o o o
«| OO Oo JOOOOOOOOOOOOOOOOOOOOOOOOOOO
i MM MMMMMMMM^MMMMMMMMMMMMMMMMMMMMM
AMPX 4 0 D U - E LOG AND L I S T OF C A R ) INPUT DATA
MOOU.E LAVA HAS BEEN CALLED
ott s e
It* I I J 4 9 I I 0 IT
2*1 I MS 0 *»• I T*« 2 . 0 1.0 a** 2.0 I.J it
2 S O I OUMMV f0# HEAT • ! TM GAMMA AOJOINT
0. 0. 1. I. C.
0.03 0.0 .1 .07 .03
MOOU-E LAVA IS FINI3HEO.
C»U T I M E . 0 3 (SECONDS*. 1/O'S 156 USER CODE JJOO SYSTEM CODE 000
MODULE MORSE HAS BEEN CALLED
TEST OF GAMMA HEATING
!•» 0 100 200 100 I 0 0 0 1 0 0 0 0 0
2*1 I I 1000 I 0
Ml I I O 42 1 0 2 I SI • 32
«»» 1 0 0 0
9*1 3 0 0 0 0 I I t
72 I I FO
IT
GAMMA H E A T I N G TEST
42S » M 32 10
JBM 32 i.E*9
S»M 32 3.C«9
END
2NI 1
2S2 2 -1
2SS 3 -2
e>»o
i i i
1 1000 0
6*« 4.2 0 1. 0 «.0 0. 0. 0. 0. 0
6» • 1.3 1 .5 .03 .OS .1 .1 .5 .S
1144F . 1
1211 1 130* 1 14*4 |.
2T
!«•• I . E » 3 22 1.E44 I.E44 0 0 1 .E*5 •I.E»4 i.e»4

-1.E43 22
20** FI2.SAC3T
3T
MOOULE M3RSE IS FINISHED.
C»J TIME 2 3 8 . 4 0 ISECUNOSI. I/O'S 17|7 USER COOc. 1 3 0 0 SYSTEM CODI 000
156
< z o
to
s - o
o
to
o
Z
M
to
<n to T
o K " O
3
i
•»
< •H
P» 5 to z
- a« s u a
o«
4 u5
Q O v> 3 oit
W Z O to to
U
J to« <
r *" t * o ki ••
•1
to «l
o
•i j i 2
u 3
o
5
O
«• U
5 to
J _
kl M ™
ir a *l y kl u v«
n
•> a
< vj
- « <
•"
o
w
to
< k.
o
z
—
<n
•
iw
i
*J
«i
a
o
3= o •*
l»
a_ i
i3 < 5
•. j -w a
5
a«
< toB<
Z
•• v» « O
to A.
J a •a• •» to to to
to
3
oJ O «
v» «I MI
a «i
to
— <• a T a
•i
VI
to
7 z Z Z
•»
Z 3 a,
to 2
toIff a
to
j 3m a* z a z «•- a
1) O z o 3 a•« a•• a z 3 « < X
««
V X
v> * ki z w z 7. k. a n
a a n O O J kl
V n U a u kl to
to J to 4 k. o •. Z a a to a
. J M kl < 4 z < < a a q
*n -v *.
? rfa a
«i w 3
o
a —
X «.
to
•n
i 5
k. Z 12 X
to to « 9 3
•r I z T X O O <3
z a a
o o o — o o o — o «
"> o o
-. <M —
>
o o •»
J
aki y s < -!
HI to
-• a
6
•I
kl O 3
S K - O
z - to a
to" J z • k,
O «.
« *
S * % s° s
o a o o 2 «
» z
« — k.
a o
ss & * z &s M
O
a
to
w
3
« t > akl kla (J Z

B s I kl 8 6 J k. to «
!i
* £ 6 t i n a a z at
Ill O
• O B 5 S 8 5 5 5 O
L>
3
Z
i 3 >»
a x
to
>> z u a u
-35 •I
P to to < a <L a o t.
-» « *i ~ X «
•- to n < 4 t o z " k i t o » i n n
« < a o 8 5 o - o
w a ki r d z O o u t i ) 4 a 0
•• i. Z Z Z T z z z z z z a - z
a 2 a
157
vi u — v z a
« - » * • z
- o z «
a z •! o J «
- -i a z a
• <a * z >-
19 kl Z « *
u O
or w 3 a
3 X
!•» «•
VI ».
 s 5
VI
a
VI
a
VI
a
v»
a
VI
or
TEC
19 «i
s s s * > $ $
>
19
ai » 3 ti
a
1
w
z
gr
Hi
V*
•>
•i
>
• ^ > z z j
4
z a. a k. j» u u o
a w *.
o <
a o O «.
O o <
K
a u
a a z 3 4 o^ j
4 HI
U
a
u < aMl a s
w UJ
a
u < a
r
S
z ^
71
3
VI
ut
or
dr s a 8 B B o
•• 3 • £*
5z vt 3
s sr 3
T INE
T INE
T INE
TINE
T INE
I/I y T T
O
i
J
IT l/l ok 3 3 3 3 3
O O O n O
"» «.
n w ^ .3 fr
<n 1z a
z sz z z
w
I z 23
3
B
3 z
3
l/l V) VI VI VI
158
S ¥ ¥
kl u O O O
•» •• U U U
z 3 z z z
- - o o o
o o — - -
« « •>• fc »-
o n u <J yj
n rt w w tu
a a o
u u w u w
a a a at a
3 3 3 3 3
n n o ~
o o i
<oo
>eo
oo
Zoo
woo
•.«!«>
VI *
• "
nn —
3 >
Ul • •
soo
u
o
«•>
*j
o
o
UJ
o
m
o o o
o a
• ^ 1 n
K
o o o o• o o
lw s??
• 511
VI
o
«ss
a
oo
>
a X "So
a
u 8 •-VIVI a
* ^ O 4
*» 3!a i 5 *^ J-3
1 a3 V) w o « :
•• 3
Ul
»<*• u uJ a^ S
o
a Z - - v
5! a4> o^> h- a4
•*
ki
v»
u
>
4
•3»
> oM
a
4
z
4 A a * a < » < a o
* 5 1i n<or ulu w Hi a a Ul (9
T u u 4 w
4 a •^ s
o • s VI33 5 • ^•
•e
o3 oID •w
11. « « w
«
in * a n3 - m
O
4 3 VI VI f
9 U » V V)
•> * B « K
GAMMA H E A T I N G TEST
I VOPT a 0 I DBG • 0
dOOY OAT A
S>M I 0.0 0.0 0.0 0.10000000 02
SPH 2 0.0 0.0 0.0 0.20000000 06
S»H 3 0.0 0.0 0.0 0.30030000 06
NUMBER OF B O D I E S 3
^ E N G T H OF F » 0 - A R R A V IB
INPUT ZONE OAT A

l*0UT ZONE COOE ZONE BOOY NUMBERS
ZNI I I I
ZN2 2 2 2 - 1
ZN3 3 3 3 -?
NUMBER OP INPUT ZONES 3
NUMBER OF COOE ZONES 3
L E N G T H OF INTEGER ARRAY 74
M
COOt tOte INPUT ZONE ZONE OATA LOC . NO. JF BODIES REGION N O . KEOIA NO. *-£
1 I 22 I I I *G
2 2 27 i | 1000
3 3 36 2 1 0
I K* l< I * KR2U I
1 I I
2 2 2
3 3 3
MORSE R E G I O N IN INPCT Z ONE U I ARRAY 4 iW»l Z ( I t . I « I . 3)

1 I I
404SE MEDIA IN INPUT ZONtdl AfcRAY < MMI Z l I ) . I • I . 31

I 1000 0
TEST OP GAMMA MATING THUVSOAV. MAT 1 9 . 19TT
OEYECT34 X HAD TO
O. IOO0B c* 0.0 0.0 0.9999E OS 0.333SE-0S

O.IOOOE OS O.IOOOE OS 0.0 O.I*l«E 00 0.4TISE-06
0.0 O.IOOOE 06 0.0 O.IOOOE oe O.3336E-0S
C9 O.IOOOE OS 0.0 0.4719E-06
-o.ioooe C6 0.0 0.0 O.MISE 09
0.3336E-03
O.IOOOE 06
-O.IOOOE
RESP31SE FUNCTIONS
CROUP RESPI II
0 . I 2 S T C C2
0.I2S7C (>
T I M E R E O U I R E O FOR I N P U T MAS L E S S THAN ONE SECONO.
•••START BATCH I RANOOM*000?IAPD«90D
SOURCE DATA
VOU ARE U S I * « THE DEFAULT V E R S I O N 0» SOURCE MHICH SETS BATE TO OOF ANO P 4 3 V I O E S AX E<*E*GV I G .
• Ta I.OOOOE 0 2 UAVEa -3.T939E-02 VAVEa » .0694E-02 WAVE* -4.6966E-02 AGEAVEa 0.0
I AVE* 1.00 AAVE-. «.0OOOE 0 0 VAVE* 0.0 2AVE* 0.0
VOU ARE U S I N G THE D E F A U L T V E R S I O N O" GTMEO V H I C H ASSUMES THAT THE «EOMETRV ANO MSEC MEOIA ARK THE SAME
NUMBE4 OF C O L L I S I O N S OF TYPE NCOLL
S3URCE S P L I T t O 1 FIM4N GAMGEN REALCOLL ALBEOQ BORVX ESCAPE E - C U T TIMEX7.LL R R H I L L R R SURV CAMLOST
I0O 0 0 6T6 6936 664 3 0 0 332 160 0
0
T I M E R E Q U I R E D FOR THE P R E C E O I N G BATCH M l 6 SECONDS.
• ••START BATCH 2 RANOOM'2B2?0* A04EF3
C*»
S0U4CE DATA O
UAVEa T.I316E-03 VAVEa -3.4663E-02 MAVE> I.62S3E-02 0.0
• T> I.OOOOE 02
iVE* 9.0000E 00 VAVEa J .0 ZAVEa 0.0
IAVE* 1.00 >
NUMBER OF C O L L I S I O N S OF TVPE NCOLL
S3URCE S P L l T I O I FISHN GAMGEN REALCOLL ALdEOO BORVX ESCAPE E-CUT T IME< I L L R 4 K I L L R R sunt GAMLOST
100 0 0 3T9 396* T*S 0 0 0 0 3T0 IAS 0
T I M E R E O U I R E O FOR THE P R E C E D I N G BATCH MAS T SECONDS.
• ••START BATCH 3 R ANO0N»S&0»»4 S 9 4 0 4 I .
SOURCE OATA
AGEAVE> 0.0
• T. I.OOOOE 0 2 UAVE" 7.2SI9E-02 VAVE» 5.TJ63E-02 VAVE" A.3TJ0E-03
I Ave- 1.00 XAVEa « . OOOOE 0 0 VAVE* 3.0 ZAVEa 0.0
NUMBER OF C O L L I S I O N S OF TVPE NCOLL
S3URCE S P L I T I O ) F I SHN GAMGEN REALCOLL ALdEOO BORVX ESCAPE E-CUT T I M E i I L L R 4 M I L L R R SURV GAMLOST
100 0 0 632 6B99 B33 0 0 0 0 3*6 I9T 0
T I M E R E Q U I R E O FOR THE P R E C E D I N G BATCH MAS 8 SECONDS.

• ••START BATCH • RANOOM*eF230293000l
SOURCE OATA
• Ta I.OOOOE 02 UAVEa 9.461 IE-03 VAVEa 3.6692E-03 SAVEa -I.9069C-02 AGEAVEa
UVE« 1.00 XAVEa 9 . OOOOE 0 0 VAVEa j ,0 ZAVEa 0.0
;(>:
NMB9S
8 «o—«o«
LOQOOoq
ooooo
uooooo
V) • I I I I
i Z WMlUIUiyJ
Ô.— * * * «
^ W — 0">*<IM
ct * • • • •
ANN- -•
_J*.P>i>K(>
I T • * » « - . * "»
3
W 3 • • • * •
bV OOOOO
*• I f I I t
X
- 9t
J • u*
z o-*o-•-••"•«»
* T<
ooo
..* 1 1 t 1 1
^ > dS 38*£2
tia 0«*">* *
2 <
O Si 3u — 0«wO
a
* •
3
5X • ••••
1 1 1 1 1
u ^
m
*
W
z
0
> •
rx
O
M
k> *> 1
< a
<
r W *
^
v —*!*»#*>
a *) UJ
a
ui 0« ^
Ui
0
»- «/t
y
3 X
<r ^

Téléchargement

Transféré par

Informations du document

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Téléchargement

Transféré par

Droits d'auteur :

Formats disponibles

ORNL/NUREG/CSD-3

A Monte Carlo Method of Solving Heat

Prepared for the U.S. Nuclear Regulatory Commission

Contract No. W-7405 eng 26

A MONTE CARLO METHOD OF SOLVING HEAT

Computer Sciences Division

Submitted to the University of Tennessee by S. K. Fraley as a

Date Published: December 1977

PREPARED FOR THE U.S. NUCLEAR REGULATORY COMMISSION

UNION CARBIDE CORPORATION, NUCLEAR DIVISION

1/ISTKIBUriON Of THIS OOLUVLV

II. TRANSPORT THEORY APPROXIMATION TO THE DIFFERENTIAL

EQUATION OF HEAT CONDUCTION 4

III. APPLICATION OF MONTE CARLO TO THE TRANSPORT EQUATION ... 12

IV. COMPARISONS OF MONTE CARLO TRANSPORT SOLUTIONS TO

V. CORRECTIONS TO THE TRANSPORT EQUATION APPROXIMATION . . . . 90

VI. MONTE CARLO SOLUTION OF A COUPLED GAMMA RADIATION AND

HEAT CONDUCTION PROBLEM 98

VII. CONCLUSIONS AND RECOMMENDATIONS Ill

A. VERIFICATION OF ANALYTICAL SOLUTIONS 124

B. DERIVATION OF A BOUNDED NEXT EVENT ESTIMATOR 147

C. LISTING OF COMPUTER INPUT DATA AND SELECTED OUTPUT .... 151

1. Temperature Comparisons for a Two-Media Sphere with a

2. Temperature Comparisons for a Two-Media RPP with a

3. Temperatures at a Point Using a Bounded Next Event

4. Temperature Comparisons for a Two-Hedia RPP with a

5. Temperature Comparisons for a Two-Media RPP with a

6. Temperature Comparisons for a Two-Media RPP with a

7. Temperature Comparisons for a Three-Media RPP with a

8. Temperature Comparisons for a Two-Media RCC with a

9. Temperature Comparisons for DOT Runs with Different

10. Temperature Comparisons for a Two-Media RCC with a

11. Temperature Comparisons for a Two-Media RCC with a

12. Temperature Comparisons for a Sphere with a Time

13. Temperature Comparisons for a Sphere with a Time

14. Temperature Comparisons for Problem Two with the

16. Teaperatures for 8=45° Coapared to Analytical Estimate . . 102

17. Teaperatures for 8=90° Coapared to Analytical Estimate . . 103

18. Teaperatures for 8=135° Compared to Analytical Estimate . . 103

19. Temperatures for 8=180° Compared to Analytical Estiaate . . 104

1. Two-Media Sphere with Point Source at Center 19

2. Two-Media RPP with Point Source in Medium One 24

3. Two-Media RPP with Distributed Source in Medium One . . . . 32

4. Three-Media RPP with Point Source in Mediun One 50

5. Three-Media RPP with Distributed Source in Medium One . . . 61

6. Two-Media RCC with a Ring Source in Medium One 66

7. Two-Media RCC with a Distributed Source in Medium One . . . 72

8. Diffusion Theory vs Transport Theory with Different

9. One-Medium Sphere with Incident Gamma Beam 99

10. Temperatures tor 8 = 0 ° Compared to Analytical

12. Temperatures for 8 = 90* Compared to Analytical

13. Temperatures for 6 = 135° Compared to Analytical

14. Temperatures for 8 = 180° Compared to Analytical

A new approach in the use of Monte Carlo to solve heat conduction

problems was developed using a transport equation approximation to the

heat conduction equation. The method was shown to be applicable to the

solution of multimedia problems in complex geometries with no inherent

limitations as to the geometric complexity of problems which can be

solved. Nuclear radiation transport and heat conduction problems can

a coupled mode with a single computer run.

The method was demonstrated for problems involving multimedia,

internal sources, fixed temperature boundary conditions, convective

boundary conditions, time-dependent sources, and for a coupled nuclear