Lecturer 02: One

Dimensional Discrete
Ordinates Formulation
Dany Sanchez Dominguez
[email protected]

INSTEC – Cuba
December - 2014
 Lecturer 02 - Topics
• One dimensional transport equation
• Discrete ordinate formulation (SN)
• Spatial discretization
• Diamond Difference
• Step and Step Characteristic

DMNT – INSTEC 2014 [email protected] 2 / 61

 1D Transport Equation
• Static BTE in one-speed approximation, Cartesian
geometry and external source appear as
  
   x, y , z ,       x, y , z ,       x, y , z ,   
x y z (1)

 t  x, y, z   x, y, z ,      s
  x , y , z  Y *
m    m
 x, y , z 
 0 m 

 Sext  x, y, z ,  

• Phase-space:  x, y, z,  ,  x, y, z,  , ,   ,  x, y, z, ,  

DMNT – INSTEC 2014 [email protected] 3 / 61

 1D Transport Equation
• Position an direction characterize a neutron

DMNT – INSTEC 2014 [email protected] 4 / 61

 1D Transport Equation
• One dimensional geometry = slab geometry
• Symmetric problems in the plane y,z
• Symmetry about a plane means the domain is
infinite in this plane
• When this assumption is robust?
• Dimensions in one direction are to bigger than
another directions (symmetry plane)
• Neutron density change stronger in one direction and
slow in another directions
• Nuclear reactor (axial and radial)
DMNT – INSTEC 2014 [email protected] 5 / 61
 1D Transport Equation
• Symmetric in the plane y,z → flux changes in this
plane are neglected
• Don’t have changes in the azimuthal angle
• Slab geometry = azimuthal symmetry
• Make a rotation in x-axis to coincident with axial
direction    cos  

x
 x, y, z ,  , ,   
    x,  
 x, y, z ,  ,   
DMNT – INSTEC 2014 [email protected] 6 / 61
 1D Transport Equation
• Now, we obtain each term of eq. (1) in a new
geometry
• Streaming term
 
   x, y , z ,       x, y , z ,   
x y (2)
 
   x, y , z ,       x,  
z x
• Scattering source term
 Y *m     Y *m   
  s

 0 m 
 x , y , z  Y *
m    m
 x, y , z 
 m  x, y , z    m  x 

DMNT – INSTEC 2014 [email protected] 7 / 61

 1D Transport Equation
• Expansion coefficients  m  x  appear as
d 
1
  x  
m
Y m      x,    (3)
1
2
• Considering spherical harmonic definition
Y m   ,   C P
 2  1  m  !
1/ 2 m
 e i m
Cm 
m
  m !
• in eq. (3) we have
d  m
1
 m
 x  C 1/2 i m
m e 1 2 P     x,  
• Considering properties of Legendre polynomials
0, m0
  x   0
m
(4)
  x     x, m0
DMNT – INSTEC 2014 [email protected] 8 / 61
 1D Transport Equation
• The previous result has two consequences
 Second sum in scattering source has one term
 We have only one coefficient define as
d 
1
  x   Y 0      x,    (5)
1
2
• Spherical harmonic of zero-order and the complex
conjugate
Y 0     Y *0      2  1 P  
1/ 2
(6)

• Substituting in eq. (4)

d 
1
  x    2  1 1 2 P     x,  
1/2
(7)

DMNT – INSTEC 2014 [email protected] 9 / 61

 1D Transport Equation
• The scattering source in one-dimensional geometry

S sc  x,      s
  x  Y *
m    m
 x
 0 m 

• Now considering results (4), (6) and (7) in previous

equation

S sc  x,    
2  1 1
 s  x  P     P      x,    d   (8)
0 2 1

• Convert the collisions and external source terms of

eq. (1) to one dimensional geometry is direct

DMNT – INSTEC 2014 [email protected] 10 / 61

 1D Transport Equation
• BTE in one dimensional geometry appear as

   x,     t  x   x,   
x

2  1 1

 0 2
 s  x  P     P      x,    d    Sext  x,   (9)
1

• BTE equation considering isotropic scattering?

  0, P0     1
1
1
    x,    d      x 
2 1

DMNT – INSTEC 2014 [email protected] 11 / 61

 1D Transport Equation
• BTE equation considering isotropic scattering . . .

   x,     t  x   x,     s 0  x    x   Sext  x,   (10)
x

• BTE equation considering linear anisotropic scattering?

  0,1; P0     1; P1     
1 1
1 1
     x,    d     j  x,    d   J  x 
2 1 2 1

   x,     t  x   x,     s 0  x    x  
x (11)
3 s1  x  J  x   Sext  x,  
DMNT – INSTEC 2014 [email protected] 12 / 61
 1D Discrete Ordinates
• The next step is discretize angular dependence 
• Methods to discretize angular dependence
 Discrete Ordinates Formulation (SN methods)
 Spherical harmonics (PN methods)
• SN methods has become the dominant means for
deterministic solution of BTE
• SN has its earliest roots in radiation transport
calculations in stellar atmospheres (Chandrasekahar, 1960)
• These concepts was evolved and applied to nuclear
energy in later 60s by (Carlson e Lathrop, 1968)
DMNT – INSTEC 2014 [email protected] 13 / 61
 1D Discrete Ordinates
• In synthesis SN methods
 Replace continuous angular domain by a subset of
selected directions
 Selected directions = discrete ordinates
   m , m  1: N
 In addition, integral in source term (moments of angular
flux) must be approximated using a numerical quadrature
• Illustrate the fundamentals of SN methods using one
dimensional BTE with isotropic scattering

1
1
   x,     t  x   x,     s 0  x     x,    d    Sext  x,   (12)
x 2 1
DMNT – INSTEC 2014 [email protected] 14 / 61
 1D Discrete Ordinates
• The integral in scattering term represent scalar neutron
flux
1
1
  x      x,    d   (13)
2 1
• Considering an slab domain of longitude L
  cos  
0 xL

0
x 1    1
L
• To solve equation (12) we need boundary conditions in left
and right boundaries

DMNT – INSTEC 2014 [email protected] 15 / 61

 1D Discrete Ordinates
• Prescribed boundary conditions
  0,     0    ,  0
(14)
  L,     L    ,  0

• Reflexive boundary conditions

  0,      0,    ,   0,     
(15)
  L,      L,    ,   0,     

• Now constrain neutrons to only travel in certain

directions   m , m  1: N
• m are the selected directions, and N is the number
of directions (order of formulation)
DMNT – INSTEC 2014 [email protected] 16 / 61
 1D Discrete Ordinates
• Then we approximate
  x,      x,  m    m  x 
•  m  x  are the neutrons traveling in direction m
• In addition, the integral of source term is
approximate as
1
1 1 N
   x,    d     n  x  n (16)
2 1 2 n 1

• That’s use a discrete quadrature formula to

approximate integral

DMNT – INSTEC 2014 [email protected] 17 / 61

 1D Discrete Ordinates
• Then we write BTE eq.(12) for each m direction, and
regard integral approximation (16) obtain
N
d 1
m  m  x    t  x  m  x    s 0  x   n  x  n  S m  x  (17)
dx 2 n 1

m  1: N

• Eq. (17) represent a system of DOE coupled by

scattering source,
• SN formulation convert integro-differential equation
(12) in a system of differential equations
• Now, we have only one dependent variable x

DMNT – INSTEC 2014 [email protected] 18 / 61

 1D Discrete Ordinates
• In one-dimensional SN formulation  m , m  are the
angular quadrature parameters
 m are the direction or ordinates 1  m  1
 m are the weight m  1: N

• The weights m must attend the normalization condition

N


n 1
n 2

• The quadrature formulas fulfill the following conditions

 Even number of ordinates (N – even)
 Symmetric in relation with  = 0

DMNT – INSTEC 2014 [email protected] 19 / 61

 1D Discrete Ordinates
• The use of odd number of N is undesired for two reasons
 For m = 0 the streaming term in eq. (17) disappears and
mischaracterize neutron balance
 Discontinuity problems for m = 0 in vacuum boundary
conditions

1 - incident,    0
1

 2 - incident,    0
2

 Discontinuity when
1  0, and  2  0

DMNT – INSTEC 2014 [email protected] 20 / 61

 1D Discrete Ordinates
• The symmetric condition about  = 0 in angular
quadrature considers that the right and left neutrons
fluxes are of equal importance
• The symmetry leads to  N /21  N /2
 N 1
k  0 2
 N 1 k   k , k  1: N 2 N 1
 N 1 k  k m  0, m  1: N / 2 x
m  0, m  N / 2  1: N

  N /2  2  1 1 2  N /2 
0
 N /21  N 1 N
DMNT – INSTEC 2014 [email protected] 21 / 61
 1D Discrete Ordinates
• Design 1D discrete ordinates for S2, S4 and S8
• Now, we write a boundary conditions for a subset of
discrete ordinates
• Prescribed boundary conditions
 m  0  fm ,   0, m  1: N / 2
(18)
 m  L   gm ,   0, m  N / 2  1: N

• Symmetric boundary conditions

 m  0    N 1 m  0  ,   0, m  1: N / 2
(19)
 m  L    N  m 1  L  ,   0, m  N / 2  1: N

DMNT – INSTEC 2014 [email protected] 22 / 61

 1D Discrete Ordinates
• Although we have restrictions for discrete ordinates
(even and symmetric) exists some alternatives to choose
(m, m)
• The most common set of discrete ordinates in 1D SN
 Gauss-Legendre Quadrature
 Double Legendre Quadrature
• Gauss-Legendre Quadrature
• Discrete ordinates (m, m=1:N) are the zeros of Legendre
polynomial of order N
PN  m   0, m  1: N (20)

DMNT – INSTEC 2014 [email protected] 23 / 61

 1D Discrete Ordinates
• Gauss-Legendre Quadrature …
• The weights of quadrature (m, m=1:N) are determined
such at the quadrature formula correctly integrates the
Legendre polynomials P  m  ,  0 : N 1
• This condition generate the algebraic system equation
1 N

2 n 1
n P   n    0 ,  0 : N 1 (21)

• Using eqs. (20) and (21) we determine parameters of

Gauss Legendre quadrature (m, m)
• Example: Determine the parameters of S2 quadrature.
DMNT – INSTEC 2014 [email protected] 24 / 61
 1D Discrete Ordinates
• Example . . .
N 2
 Unknowns:
Ordinates: 1 , 2
Weight: 1 , 2
 Using eq. (20)

P2      3  1  0  1,2  
1 3
2
 0.57735
2 3
 Using eq. (21)
1
 0, 1 P0  1   2 P0   2     00
2
 1  2  1
1
 1, 1 P1  1   2 P1   2    10
2
DMNT – INSTEC 2014 [email protected] 25 / 61
 1D Discrete Ordinates
• Homework 1: Determine the parameters of SN
quadrature for N=4, 6 and 8.
• The quadrature parameters for a mots used orders

DMNT – INSTEC 2014 [email protected] 26 / 61

 1D Discrete Ordinates
• Finally, discrete ordinates formulation for one
dimensional problem with isotropic scattering
N
d 1
m  m  x    t  x  m  x    s 0  x   n  x  n  S m  x  (22)
dx 2 n 1

 m  0  fm ,   0, m  1: N
 m  L   gm ,   0, 0 xL

 The meaning of each term in equation?

 Consequences of uses higher orders quadrature?
 Consequences of uses lower order quadrature?

DMNT – INSTEC 2014 [email protected] 27 / 61

 1D Discrete Ordinates
• Exact solution, high- and low-order GL approximations in
slab boundary (angular dependence)

• The next step is discretized spatial dependence x.

DMNT – INSTEC 2014 [email protected] 28 / 61
 1D Spatial Discretization
• First, we considering the next grid Ωx in the domain

0 1 2 i I 1 I L
x1/2 x3/2 xi 1/2 xi 1/2 xI 1/2 xI 1/2

I  spatial cells, i  1: I
I  1  edge cells, xi 1/2
L
hi  xi 1/2  xi 1/2 
I

• The σt (x), σs0 (x) , Sm(x) are piecewise constant functions of x,

constant inside the spatial cell

DMNT – INSTEC 2014 [email protected] 29 / 61

 1D Spatial Discretization
• Now we write eq. (22) inside an spatial cell
N
d 1
m  m  x    t ,i m  x    s 0,i  n  x  n  Sm ,i (23)
dx 2 n 1

• To obtain spatial moment balance equations we apply

next operator in eq. (23)
2  1
xi 1/2
 2  x  xi   xi 1/2  xi 1/2
hi 
xi 1/2
P
 hi
 dx,

xi 
2
(24)

• To obtain zeroth spatial moment balance equation,

consider  0 in eq. (24) and apply in (23)
xi 1/2
1
hi 
xi 1/2
dx (25)
DMNT – INSTEC 2014 [email protected] 30 / 61
 1D Spatial Discretization
• Zeroth spatial moment balance equation
 m ,i 1/2  m ,i 1/2 C0,i N S m ,i
 m ,i
x
  m ,i  
2 n 1
 n ,i n 
 t ,i
(26)

• where, spatial moment of zeroth order the neutron flux

(average angular flux in a cell) is define as
xi 1/2
1
 m ,i 
hi    x  dx
xi 1/2
m
(27)

• and
hi t ,i  so ,i
 x
 (28) C0,i  (29)
m ,i
m  t ,i
DMNT – INSTEC 2014 [email protected] 31 / 61
 1D Spatial Discretization
• Eq. (26) must be write for any spatial cell i=1:I and any
direction m=1:N
 m ,i 1/2  m ,i 1/2 C0,i N S m ,i
 m ,i
x
  m ,i 
2

n 1
n ,i n 
 t ,i
(26)

• Which is the number of unknowns and equations?

 m ,i • Unknowns
 N(I+1) – edge angular fluxes
 m,i 1/2  m,i 1/2
 NI - average angular fluxes
 Total: N(2I+1)
DMNT – INSTEC 2014 [email protected] 32 / 61
 1D Spatial Discretization
• Equations
 NI – balance equations
 N – boundary conditions (N/2 in each boundary)
 Total: N(I+2)

Equations  Unknowns
N  I  2  N  2 I  1

• The system of balance equations and boundary

conditions is soluble and undetermined
• What can I do?
DMNT – INSTEC 2014 [email protected] 33 / 61
 1D Spatial Discretization
• We added I auxiliary equations
• The balance equations
• We obtained without approximations (exact)
• Are present in all methods of numerical solutions
• The difference between methods is in the type of
auxiliary equation
• Auxiliary equations introduce approximation of the
method and spatial truncation errors
• Finite difference methods to solve 1D SN formulations
 Diamond Difference
 Step
 Step Characteristic
DMNT – INSTEC 2014 [email protected] 34 / 61
 Diamond Difference
• Auxiliary equations
 m,i 1/2  m,i 1/2
 m ,i  (27)
2
• Average spatial flux is the arithmetic average between
incident and emergent fluxes in the same direction
• Interpretation of these approximation?
 m ( x)  m,i 1/2

 m ,i
 m,i 1/2

xi 1/2 xi 1/2
DMNT – INSTEC 2014 [email protected] 35 / 61
 Diamond Difference
• Consider linear behavior of angular flux inside spatial cell
• Auxiliary equations is decoupling in angular directions
• In which conditions this is a good approximation?
• Finer spatial grids
• DD method is a fine mesh method obtain good results in
problems discretized with a high number of spatial cell
• Balance equations. (22), boundary conditions (22) and
auxiliary equations (27) formed an closed algebraic
system for DD method
• The next step is derive an algorithm to implement
iterative scheme to obtain solutions
DMNT – INSTEC 2014 [email protected] 36 / 61
 Diamond Difference
• Implicit methods
 Construct matrix system
 Solve using numerical methods
 Lower computational performance (I→)
• Explicit methods
 Obtain sweeps expressions
 Solve using iterative numerical scheme
 High computational performance

• Is common uses explicit method called source iteration

DMNT – INSTEC 2014 [email protected] 37 / 61

 Diamond Difference
• Construct marching schemes that follow the direction of
neutrons travel
 We solve using sweeps,
 For each spatial cell we obtain emergent flux using incident
flux and average flux (in previous iteration)
 1 sweep in right direction m>0
 1 sweep in left direction m<0
 Update scattering sources (average fluxes)
 Verify convergence criteria
• Representation of one iteration
 11  12  13  14  0
x1/2  21  12 x  13 x7/2  14 x9/2  0
x3/2 5/2
DMNT – INSTEC 2014 [email protected] 38 / 61
 Diamond Difference
• First, we obtain the equations for right sweep (m>0 )
• Emergent fluxes are  m,i 1/2
• Solve eq. (27) for  m,i 1/2 we obtain
 m,i 1/2  2 m,i  m,i 1/2 (28)

• Substituting eq. (28) into zeroth spatial moment balance

equation (26) and solve for cell average angular flux
 S m ,i 
2 m ,i 1/2   x
 S sc ,i  
m ,i
  t ,i  (29)
 m ,i 
2   mx ,i

DMNT – INSTEC 2014 [email protected] 39 / 61

Diamond Difference
• In eq. (29) the scattering source term is grouped in
C0,i N
S sc ,i 
2

n 1
n ,i n  C0,ii (30)

• Second, we obtain the equations for left sweep (m<0 )

• Emergent fluxes are  m,i 1/2
• Solve eq. (27) for  m,i 1/2 we obtain
 m,i 1/2  2 m,i  m,i 1/2 (31)
• Substituting eq. (31) in (26) and solve
 S m ,i 
2 m ,i 1/2   x
 S sc ,i  
m ,i
  t ,i  (32)
 m ,i 
2   mx ,i 40 / 61
 Diamond Difference
• Equations (28-29) and (31-32) represent equations
sweeps for positives and negatives directions
respectively
• Is possible simplify notation considering signal of
negatives ordinates in (30-31), the general equations

 m,i 1/2  2 m,i  m,i 1/2 (33)

 S m ,i 
2 m ,i  x
 S sc ,i  
1/2 m ,i
  t ,i 
 m ,i 
2   mx ,i (34)

DMNT – INSTEC 2014 [email protected] 41 / 61

 Diamond Difference
• Convergence criterion
i(iter 1)
max 1  (iter )   (35)
i 1:I i
• Maximum difference between successive iterations not
exceed pre-defined value
• At this point is possible write a computational code for
solving fixed source transport problems using
 One dimensional discrete ordinate formulation
 Gauss Legendre quadrature
 DD method
DMNT – INSTEC 2014 [email protected] 42 / 61
 Diamond Difference - Algorithm
1. Input
N, I
  n , n 
 t ,i ,  s 0,i , Sm ,i

 m ,1/2 BC for m  1: N / 2
 m , I 1/2 BC for m  N / 2  1: N

2. Initialization
S sc ,i  0 for m  1: N , i  1: I
0

i(0)  0 for i  1: I
DMNT – INSTEC 2014 [email protected] 43 / 61
 Diamond Difference - Algorithm
3. Rightgoing sweep
For i  1: I
sum1  0
For m  1: N / 2
Calculate  m ,i , eq.(29) using S (sciter,i )
Calculate  m ,i 1/2 , eq.(28)
sum1  sum1   m ,im
End
sum1
 i
( iter 1)

2
End

DMNT – INSTEC 2014 [email protected] 44 / 61

 Diamond Difference - Algorithm
4. Leftgoing sweep
For i  I :1
sum2  0
For m  N / 2  1: N
Calculate  m ,i , eq.(32) using S (sciter,i )
Calculate  m ,i 1/2 , eq.(31)
sum2  sum2   m ,im
End
sum2
 i
( iter 1)

i
( iter 1)

2
End

DMNT – INSTEC 2014 [email protected] 45 / 61

 Diamond Difference - Algorithm
5. Verify convergence criteria
For i  I :1
i(iter 1)
 cal  max 1  (iter )
i 1:I i
End
If ( cal   )
1)
Calculate S sc(iter
,i using eq. (30)
Else
Terminate program
End

DMNT – INSTEC 2014 [email protected] 46 / 61

 Step Method
• Auxiliary equations
 m ,i 
1  sgn( m )  m ,i 1/2  1  sgn( m )  m ,i 1/2
(36)
2
• Eq. (36) result in
 m ,i 1/2 , m  0

 m ,i   (37)
 m ,i 1/2 , m  0


• Interpretation?  m ( x)  m,i 1/2

 m ,i
m  0

 m,i 1/2

DMNT – INSTEC 2014 xi 1/2 xi 1/2 47 / 61

 Step Characteristic Method
• Auxiliary equations
1   m ,i  m ,i 1/2  1   m ,i  m ,i 1/2
 m ,i  (38)
2
• In eq. (38) we define
  mx ,i  2
 m ,i  coth    x (39)
 2   m ,i
• Interpretation?  m ( x)  m,i 1/2
 m ,i

 m,i 1/2

DMNT – INSTEC 2014 xi 1/2 xi 1/2 48 / 61

 DD, Step, Step Characteristic
• Comparison between approximations
DD Step Step Characteristics

 m ,i  m ,i
 m ,i  m,i 1  m,i 1
 m,i 1

xi 1/2 xi 1/2 xi 3/2 xi 1/2 xi 1/2 xi 3/2 xi 1/2 xi 1/2 xi 3/2

 Step characteristics is better approximation

 All methods obtain good results on finer mesh
 Non-physics negative flux appear in coarse mesh
 Easy implementation
 Poor computational performance
DMNT – INSTEC 2014 49 / 61
 DD, Step, Step Characteristic
• Homework 2. Obtain sweeps equations for step method.

• Homework 3. Obtain sweeps equation for step

characteristics method.

• Course project. Solve homogeneous and heterogeneous

transport one dimensional problems .

DMNT – INSTEC 2014 50 / 61