MCNP Syllabus

1
MCNP Syllabus

• Introduction
• Input File Basics
• Geometry Definition
• Source Definition
• Tally Definition
• Variance Reduction
• Criticality

2
Variance Reduction in MCNP

• Syntax for Variance Reduction
• Truncation
• Population Control
• Modified Sampling
• Partially Deterministic
• Pathological Problems

3
Energy Time Cut-off

• CUTn T E WC1 WC2 SWTM
• n particle type N,P,E
• T upper time cutoff shakes
• E lower energy cutoff MeV
• Kill particles when their time is above T
• Kill particles when their energy is below E
• Defaults
• n T very large, E0.0
• p T neutron T, E0.001
• e T neutron T, E0.001

4
Weight Cut-off

• CUTn T E WC1 WC2 SWTM
• n particle type N,P,E
• WC1, WC2 weight cut-offs
• SWTM Minimum source weight
• If WGT
• Survival probability WGT/(WC1.R)
• Survival weight WC1.R
• R ratio of cell importance to source cell
  importance
• If WC1 WC2
• WC1WC1.Ws WC2WC2.Ws
• Ws minimum source weight

5
Geometry Splitting/Roulette

• Based on importance of cell geometries
• Comments, tips and guidelines in MCNP
• Splitting and roulette always work together
• Set cell importances with goal of maintaining
  constant population from cell to cell between
  source and detector (see Print Table 126)
• Avoid very large/small values of N (few)
• Unreliable in problems with extreme angular
  dependence

6
Population Control w/ Geometry Splitting
IMP Tracks
New IMP
Source Tally
1 300
1
2 200
3 (2/1) x (300/200) x 1
4 100
12 (4/2) x (200/100) x 3
8 25
96 (8/4) x (100/25) x 12
EX vr1
7
Energy Splitting/Roulette

• ESPLTn N1 E1 N2 E2 N5 E5
• n particle type
• Ni of tracks into which particle will be split
• Ei Energy at which particles undergo splitting
• Ni define splitting/roulette as particles drop
  below Ei
• can be non-integer
• Ni1 for splitting
• 0
• Maximum 5 pairs
• Opposite game performed if energy increases
  past Ei (unless Ni

8
Weight Windows

• WWEn E1 E2 Ei EI (I
• WWNin wi1 wi2 wij wiJ
• n particle type
• Ei upper energy bound of ith energy window
• i energy window index
• wij lower weight bound for ith energy window and
  jth problem cell

9
Weight Window Parameters(1)

• WWPn WUPN WSURVN MXSPLN MWHERE SWITCHN MTIME
• n particle type
• WUPN ratio between upper and lower weight bound
  5
• WSURVN ratio between survival weight and lower
  weight 0.6 x WUPN
• MXSPLN
• upper limit of particle splitting 5
• 1/MXSPLN lower limit on roulette probability
• MWHERE where check weight windows 0
• -1 collisions only
• 0 collisions and surfaces
• 1 surfaces only

10
Weight Window Parameters(2)

• WWPn WUPN WSURVN MXSPLN MWHERE SWITCHN MTIME
• SWITCHN where to find lower weight bound 0
• -1 from external WWINP file
• 0 from WWNi cards
• 0 lower weight window SWITCHN/IMP
• Common to use SWITCHN0
• MTIME type of weight windows on WWE card 0
• 0 energy
• 1 time

EX vr2
11
Weight Window Generator

• MCNP will automatically generate optimized weight
  windows
• Generated for a specific tally relative to a
  specific reference cell (source cell)
• Generated weight windows can be based on cells or
  based on a mesh that overlays the geometry
• Writes to file that is then referenced with WWP
  card (SWITCHN-1)

12
Weight Window Generator

• WWG It IC Wg J J J J IE (p. 3-43)
• It problem tally number for which to optimize
• IC reference cell for weight windows
• 0 problem cell number
• 0 MESH generator
• Wg value of generated lower weight window for
  reference cell (IC) or for MESH
• 0 lower bound will be ½ the average source
  weight
• J unused
• IE toggles energy(0) or time(1) dependent

13
Using the Weight Window Generator

• Add WWG info to input file in1
• Run MCNP with this file
• mcnp iin1 ncase1_
• Change the WWP info to read from a WWINP file in
  input file in2
• Run MCNP with this input file
• mcnp iin2 ncase2_ wwinpcase1_e

EX vr3
14
MESH Generator(1)

• MESH mesh variablespecification
• GEOM (p. 3-44)
• xyz, rec Cartesian
• rzt, cyl Cylindrical
• REF x,y,z coordinates of reference point
• ORIGIN x,y,z coordinates of superimposed mesh
  MCNP geometry
• AXS vector giving the direction of the axis of
  the cylindrical mesh
• VEC vector defining, along with AXS, the plane
  for q0

15
MESH Generator(2)

• MESH mesh variablespecification
• IMESH, JMESH, KMESH locations of coarse mesh
  points in the x,y,z (r,z,q) direction
• IINTS, JINTS, KINTS number of fine mesh
  intervals within corresponding coarse meshes in
  the x,y,z (r,z,q) direction default 10 fine
  mesh per coarse mesh
• Mesh boundaries should be outside problem
  geometry
• See 3-44 for more

16
Tips for MESH Generation

• Enter only the upper mesh bounds
• At least 2 coarse meshes required per dimension
• Geometry should be entirely inside mesh
• Source should not be on mesh boundaries (or on
  cylindrical mesh axis)
• Avoid too many mesh cells (beware the defaults)
• KMESH for q in units of revolutions (0

EX vr4
17
BOOKMARK FOR REVIEW OF NOTES
18
Exponential Transform

• EXTn A1 A2 Ai AI (p.
  3-36)
• VECT Va xa ya za Vb xb yb zb
• n particle type (n,p,e)
• Ai entry for cell i format QVc
• Q 0 no transform
• Q p 0
• Q S pSa/St equiv. to implicit capture (Sa
  is capture cross section)
• Vc corresponds to vector described in VECT
• I number of cells in problem
• a,b any numbers to unique identify vectors Va,
  Vb, etc.
• xa,ya,za coordinate triplet defining vector Va

19
Forced Collisions

• FCLn x1 x2xixI (p. 3-39)
• n particle type (n,p)
• xi forced collision control for cell i xi ?1
• If xi survival probability xi
• I number of cells in problem
• Weight-window game not played on surfaces when
  particles enter a FCL cell
• xi
• xi 0, force collisions repeatedly until killed

20
Source Biasing Syntax

• SBn option B1 B2 Bk (p 3-57)
• SBn f a b
• n distribution number
• option (same as for SPn)
• same as D for H or L type, pointwise PDF
  for A type
• D bin probabilities for an H or L type
• C cumulative probabilities for an H or L type
• V for cell distributions
• Probabilities are proportional to cell volumes
  (times Pi)
• Bi probabilities
• f code (negative integer) for built-in function
• a b parameters for built-in function

21
Basic Source Biasing

• Energy biasing
• SDEF ERGD1SI1 0 1 2 3 4 5SP1 0
  1 2 1 2 1SB1 0 1 1 1 1 10
• Direction biasing
• SDEF VEC1 0 0 DIRD2SI2 -1 0 1SP2 0 1
  1SB2 0 1 5

22
Biasing Built-in Functions

• Without biasing, usually no SI card
• SI card can be provided for biasing
• MCNP approximates function as table with up to
  300 equiprobable bins
• SDEF ERGD1SP1 5 aSI1 0.005 0.1
  20SB1 C 0 0.5 1

23
Point Ring Detectors

• Fn5p X Y Z ?R0
• Fn5ap a0 r ?R0
• n5 tally number ending in 5
• p particle type (n,p,e)
• R0 radius of exclusion in cm (0) or mean free
  paths (
• Point detectors
• X,Y,Z location of detector
• Ring detectors
• a the letter X,Y, or Z defining axis of detector
• a0 distance along axis a
• r radius of ring detector

24
DXTRAN Spheres

• DXTn x1 y1 z1 RI1 RO1 x5 y5 z5 RI1 RO5
• n particle type
  (p. 3-103)
• xi,yi,zi center of sphere i
• RIi inner radius of sphere i
• ROi outer radius of sphere i
• Inner sphere should enclose region of
  interest/concern
• Outer sphere should enclose regions that scatter
  into inner sphere
• DXTRAN spheres should not overlap

25
Speeding Up MCNP

• Point Detectors and DXTRAN spheres slow down the
  computer time per history
• Calculate average mean free path from every
  collision to detector/sphere location
• All contributions from some cells may be
  unimportant
• Some contributions from any cell may be
  unimportant

26
Ignoring by Cell

• PDn P1 P2 Pi PL (p. 3-47)
• DXCmq P1 P2 Pi PL
• n tally number
• m which DXTRAN sphere
• q particle type
• Pi probability of contribution to detector n
  (sphere m) from cell i (with L cells in problem)
• Contributions from cell i to detector n (sphere
  m) will only be counted with probability Pi and
  weight increased by 1/Pi

27
Ignoring by Weight

• DDn k1 m1 k2 m2 (p. 3-102)
• n detector or sphere
• 1 neutron DXTRAN spheres
• 2 photon DXTRAN spheres
• else tally number (ending in 5)
• blank default for all detectors and spheres
• ki roulette criterion (default 0.1)
• mi print criterion for large contributions (def
  ault 1000)
• Speeds up calculations
• Indicates origin of large contributions

3-103
28
Roulette for DD

• If ki0
• First 200 contributions are not affected
• Average contribution per history is calculated
• If contribution of history is less than ki times
  the average contribution, play roulette
• If ki
• If contribution is less than ki, play roulette
• If ki0
• No roulette is played

29
Print Output for DD

• If ki0
• First 200 contributions are not affected
• Average contribution per history is calculated
• For first 100 histories with contribution more
  than mi times the average contribution, print
  diagnostics
• If ki
• For first 100 histories with contribution more
  than miki, print diagnotiscs

30
MCNP VR Strategy

• Best techniques and parameters for variance
  reduction can be difficult to determine
• What is special about particles that tally?
• Direction, energy, collision location?
• What techniques will increase number of special
  particles?
• What parameters for technique?
• Too little better than too much
• Use short MCNP runs to determine effectiveness of
  VR

EX wellProb