MCNP--Serpent Benchmark Issue

Old bugs, solved problems and other topics that are no longer relevant are moved here

MCNP--Serpent Benchmark Issue

Postby bherman » Sun Jun 06, 2010 8:58 pm

Dear Serpent Users,

I am currently working on a simple benchmark between MCNP5 and Serpent so that I can use Serpent for generating homogenized parameters for a nodal code. I am running into an issue with the multiplication factor for a simple two-zone axial pin-cell arrangement (infinite 3D geometry). The lower axial region is a fissile zone (20% Pu) and the upper axial region is a blanket (Dep. UO2). MCNP k-eff is ~1.25 while Serpent gives ~1.08. Interestingly, when each zone is run separate in 2D or 3D geometry, Serpent and MCNP5 agree very well. However, when I stack the two zones on top of each other, this huge error arises. The two files are listed below, please let me know if I missed something in the construction of these files.

Thank you,

Bryan Herman

Serpent Input File

Code: Select all
set title "Pin-Cell Hex Geometry"

% --- Surfaces

surf  1000  hexxprism  0.0 0.0 0.5675,,0.0 71.4  % pin cell boundary
surf  1001  cyl        0.0 0.0 0.5025            % outer clad
surf  1002  cyl        0.0 0.0 0.4425            % inner clad
surf  1003  cyl        0.0 0.0 0.4350            % outer fuel
surf  1004  pz         19.4                      % top of fissile zone

cell  110   0  fuel     -1003 -1000 -1004       % Lower Fissile Mat
cell  111   0  blanket  -1003 -1000  1004       % Inner Blanket
cell  112   0  gap      1003 -1002  -1000       % Helium Gap
cell  113   0  clad     1002 -1001  -1000       % Cladding Mat
cell  114   0  coolant  1001 -1000              % Water Coolant
cell  115   0  outside  1000                    % outside region

% --- Materials

mat blanket -10.5
8016.09c      -0.118495
92235.09c     -0.001763
92238.09c     -0.879742

mat fuel -10.6
8016.09c      -0.118175
92235.09c       -0.0013
92238.09c     -0.518008
93237.09c    -0.0015951
94238.09c     -0.010332
94239.09c     -0.159508
94240.09c     -0.130507
94241.09c      -0.01867
94242.09c     -0.017981
95241.09c     -0.012942
95242.09c    -0.0005438
95243.09c     -0.004785
96243.09c    -7.25e-005
96244.09c     -0.003879
96245.09c    -0.0012326
96246.09c    -0.0003625
96247.09c    -7.25e-005
96248.09c   -3.625e-005

mat clad -6.5
40090.09c     -0.505548
40091.09c     -0.110248
40092.09c     -0.168516
40094.09c     -0.170776
40096.09c     -0.027513
50112.09c     -0.000141
50114.09c     -9.4e-005
50115.09c     -4.9e-005
50116.09c     -0.002108
50117.09c     -0.001114
50118.09c     -0.003512
50119.09c     -0.001244
50120.09c     -0.004726
50122.09c     -0.000671
50124.09c      -0.00084
26054.09c     -7.9e-005
26056.09c     -0.001239
26057.09c     -2.9e-005
26058.09c       -4e-006
24050.09c     -4.3e-005
24052.09c     -0.000838
24053.09c     -9.5e-005
24054.09c     -2.4e-005
28058.09c     -0.000374
28060.09c     -0.000144
28061.09c       -6e-006
28062.09c       -2e-005
28064.09c       -5e-006

mat gap -0.002089
2004.09c 1.000

mat coolant -0.736  moder lwtr 1001
1001.06c       -0.11189
8016.06c       -0.88811

therm lwtr lwe7.12t

% --- Cross section library file path

set acelib "/opt/xs_data/endfb7/sss_endfb7u.xsdata"

% --- Reflective boundary conditionn

set bc 2

% --- Neutron population and criticality cycles

set pop 6000 500 20

% --- Geometry and mesh plots

plot 3 1000 1000
mesh 3 1000 1000

% --- Unresolved Resonance

set ures 1


MCNP5 Input File

Code: Select all
C  Pin-Cell Hex Geometry--MCNP5
C
C  CELLS
C
  1   1  -10.6      -4 -1 -5   $ LF pellet
  2   2  -10.5      -4 -1 5    $ IB pellet
  3   3  -0.002089  4 -3 -1    $ He gap
  4   4  -6.5       3 -2 -1    $ Clad
  5   5  -0.736     2 -1       $ Water
  6   0             1          $ outside
 
C  SURFACES
C
*1   rhp 0 0 0 0 0 71.4 0.5675 0 0  $ pin cell boundary
  2   cz  0.5025                     $ outer clad
  3   cz  0.4425                     $ inner clad
  4   cz  0.4350                     $ outer fuel
  5   pz  19.4                       $ top of LFZ
 
C MATERIALS
C
C  LF material
m1    8016.72c      -0.118175
      92235.72c       -0.0013
      92238.72c     -0.518008
      93237.72c    -0.0015951
      94238.72c     -0.010332
      94239.72c     -0.159508
      94240.72c     -0.130507
      94241.72c      -0.01867
      94242.72c     -0.017981
      95241.72c     -0.012942
      95242.72c    -0.0005438
      95243.72c     -0.004785
      96243.72c    -7.25e-005
      96244.72c     -0.003879
      96245.72c    -0.0012326
      96246.72c    -0.0003625
      96247.72c    -7.25e-005
      96248.72c   -3.625e-005
C  IB material
m2    8016.72c      -0.118495 
      92235.72c     -0.001763
      92238.72c     -0.879742
C  Helium
m3     2004.72c 1.000
C  Zircaloy
m4    40090.72c     -0.505548
      40091.72c     -0.110248
      40092.72c     -0.168516
      40094.72c     -0.170776
      40096.72c     -0.027513
      50112.72c     -0.000141
      50114.72c     -9.4e-005
      50115.72c     -4.9e-005
      50116.72c     -0.002108
      50117.72c     -0.001114
      50118.72c     -0.003512
      50119.72c     -0.001244
      50120.72c     -0.004726
      50122.72c     -0.000671
      50124.72c      -0.00084
      26054.72c     -7.9e-005
      26056.72c     -0.001239
      26057.72c     -2.9e-005
      26058.72c       -4e-006
      24050.72c     -4.3e-005
      24052.72c     -0.000838
      24053.72c     -9.5e-005
      24054.72c     -2.4e-005
      28058.72c     -0.000374
      28060.72c     -0.000144
      28061.72c       -6e-006
      28062.72c       -2e-005
      28064.72c       -5e-006
C  Water
m5    1001.71c       -0.11189
      8016.71c       -0.88811
mt5 lwtr.16t
C
C DATA
mode n
imp:n 1 1 1 1 1 0
kcode 6000 1.0 20 500
ksrc  0.0 0.0 2.0
bherman
 
Posts: 30
Joined: Wed May 19, 2010 7:27 pm
Location: Massachusetts Institute of Technology - Cambridge, MA

Re: MCNP--Serpent Benchmark Issue

Postby Jaakko Leppänen » Sun Jun 06, 2010 11:12 pm

Bryan,

You have discovered a bug. Here's the story...

Because of the Woodcock delta-tracking method, Serpent handles all repeated boundary conditions using coordinate transformations, rather than actual reflections or translations at the site of crossing (see Sec. 3.7 in the User's Manual for some additional info). The difference between the two modes is that the directional vector is inverted if an uneven number of reflective boundaries is crossed.

The problem with reflective boundary conditions is that they do not make physical sense with hexagonal surfaces, unless there is a 30 (or 60?) degree symmetry in the geometry. In this case, however, reflectional boundary conditions produce the same result as periodic boundary conditions. This is why Serpent actually uses periodic condition with all hexagonal surface types. When I added the two hexprismatic surface types, I didn't change the coding and failed to consider the fact that there may be reflections in the axial direction as well, in which case the two conditions produce a different result.

So basically the problem in your case is that the boundary conditions are actually periodic, rather than reflective. You can probably fix the problem by changing the model in such way that there is an axial symmetry, and reflective and periodic boundary conditions produce the same result (add another fuel region below the model and another blanket region below that). This shouldn't change your geometry, just the way it is defined.

I will fix the coding in the next update. As a quick fix, you can add lines:

Code: Select all
/* Handle Reflection */
       if ((int)DATA[DATA_GEOM_BC] == BC_REFLECTIVE)
         {
      if (k % 2)
        {
          /* Odd number of z-surface crossings. Swap vectors. */
      
          *z = -*z;
          *w = -*w;
        }
         }


just before:

Code: Select all
       /* Transfer */

       *x = *x + x0;
       *y = *y + y0;
       *z = *z + z0;


at lines 288 and 348 in source file atboundary.c and recompile the code. When I made this change, I got a k-eff close to 1.25.

Let me know if this fixes your problem.
- Jaakko
User avatar
Jaakko Leppänen
Site Admin
 
Posts: 1989
Joined: Thu Mar 18, 2010 10:43 pm
Location: Espoo, Finland

Re: MCNP--Serpent Benchmark Issue

Postby bherman » Wed Jun 09, 2010 5:30 am

Dear Jaakko,

The fix you suggested did work, the two input files previously posted now yield a difference of ~0.02%. The fix should be inserted at lines 288 and 362 (not 348).

Thank you,

Bryan
bherman
 
Posts: 30
Joined: Wed May 19, 2010 7:27 pm
Location: Massachusetts Institute of Technology - Cambridge, MA

Re: MCNP--Serpent Benchmark Issue

Postby Jaakko Leppänen » Thu Jul 01, 2010 3:46 pm

The bug fix was included in update 1.1.12.
- Jaakko
User avatar
Jaakko Leppänen
Site Admin
 
Posts: 1989
Joined: Thu Mar 18, 2010 10:43 pm
Location: Espoo, Finland


Return to Old posts

Who is online

Users browsing this forum: No registered users and 1 guest

cron