Cell Cards in MCNP

A cell card is a filled region of space: which material sits there, how dense it is, and which surfaces close it. Surfaces alone split the universe into half-spaces; the cell card is where you pick a side of each surface and say what lives in the resulting volume.

cell_number  material  density  geometry  parameters  $ comment

1           1        -10.4    -1        imp:n=1    $ Fuel region
2           2        -6.56     1 -2     imp:n=1    $ Cladding
3           3        -1.0      2 -3     imp:n=1    $ Water
4           0                  3        imp:n=0    $ Outside void

The first integer is a unique cell label for the rest of the deck. Field two is a material number from an m card — or 0 for void, which carries no density column. For a filled cell the third field is density, and the sign is the whole rule: negative means g/cm³, positive means atoms/barn·cm. Write water as 1.0 instead of -1.0 and the run still finishes; you just transported through 1 atom per barn·cm — roughly 10²⁴ atoms/cm³ — instead of liquid water at 1 g/cm³.

That sign is not the fraction sign on the material card. On an m card, positive is atom fraction and negative is weight fraction — a different convention on a different card. Keep them apart; Material Cards owns the fraction half.

After density comes the geometry expression, then optional parameters such as imp:n=1. A $ comment that names the physical region — fuel, clad, graveyard — is what you will want when a plot later shows the wrong annulus filled.

The sign in front of a surface number picks a half-space. For a closed surface (sphere, cylinder) negative means the interior and positive (or bare) means the exterior. For a plane, negative is the negative half-space of the plane equation. Flip one sense and you can still get a closed-looking model that fills the wrong side of every boundary.

Nested pin geometry lives on that rule. The expression 1 -2 is the region outside surface 1 and inside surface 2 — the annular shell between two cylinders. That is how a gap or clad cell is written without inventing a new surface type.

Three boolean operators build richer regions. A space is intersection (every condition true). A colon (:) is union (either side true). A pound (#) is complement — everything outside the named cell or parenthesized expression. MCNP still requires every point in the problem to belong to exactly one cell; an overlap is fatal, and a gap that nothing claims is an unbounded path out of the model.

c Simple and complex cell definitions
1  1  -10.4  -1           $ Inside cylinder 1
2  2  -6.56   1 -2        $ Annulus between cylinders 1 and 2
3  3  -7.8   (-3 4 -5):(6 -7 8)  $ Two separate regions
4  0         #1 #2 #3     $ Outside all other cells

The first two cells are plain intersections. Cell 3 unions two multi-surface chunks. Cell 4 uses complement to claim everything the others left — a common pattern for the exterior void, provided that void also carries imp:n=0.

imp:n decides whether a neutron history continues in a cell. A value of 1 means normal tracking; 0 kills the particle on entry. You want 0 on the exterior void that closes the problem, and nowhere else in a simple criticality deck. Omit it on a fuel or moderator cell and histories can die there with no fatal error — the run completes and tallies quietly lose particles. Separate keys exist for other particles (imp:p for photons).

Every deck still needs at least one imp:n=0 cell covering the complement of the problem. Without that graveyard, escaping neutrons have nowhere legal to go and the geometry check fails.

For repeated structures, u= tags a cell (or set of cells) as a universe you can reuse, and fill= places that universe inside another cell. A pin defined once as universe 1 can fill every lattice site instead of being rewritten hundreds of times.

c Universe definition
10  1  -10.4  -1  u=1      $ Define universe 1
c Using the universe
20  0  -10  fill=1         $ Fill region with universe 1

One consequence of fill= matters for tallies later: a cell that only exists as lattice fill often has no volume MCNP can compute for an F4 or F7. You will need vol or sd on those tallies — the cell card itself does not invent a volume for a filled universe.

The running example for this path is a PWR pin cell. Four materials fill four concentric annuli, and a fifth cell with imp:n=0 mops up everything outside. Material numbers and densities are the ones Material Cards defines. Hover each line for the field-by-field reading.

c Fuel Pin Cell Definitions
c cell mat  dens     surfaces     params
1     1    -10.4    -1           imp:n=1  $ UO2 fuel
2     2    -0.0001   1 -2        imp:n=1  $ Helium gap
3     3    -6.56     2 -3        imp:n=1  $ Zircaloy-4 clad
4     4    -0.998    3 -4        imp:n=1  $ Water moderator
5     0              4           imp:n=0  $ External void (kills neutrons)
c
c Density sign: negative = g/cm3, positive = atoms/barn-cm
c Surface sign: -N = inside surface N, +N = outside surface N
c Cell 5 has mat 0 and no density (void needs neither)
 
c Surface Cards (referenced above)
1     cz    0.4096   $ Fuel pellet radius
2     cz    0.4178   $ Clad inner radius (gap outer)
3     cz    0.475    $ Clad outer radius
4     cz    0.7108   $ Water boundary

Read outward: fuel inside surface 1, gap between 1 and 2, clad between 2 and 3, water between 3 and 4, then the exterior void outside 4 with imp:n=0. Shared surface numbers keep the annuli watertight — each interface appears once as the outer bound of one cell and the inner bound of the next.

Density signs are all negative, so every one is g/cm³. Only cell 5 omits the field, because material 0 is void and a void has nothing to be dense. The one simplification left in the viewer is z: these cylinders run to infinity, while the runnable deck on Example: Pin Cell adds two reflecting pz planes to close a 1 cm slice. That matters as soon as a tally needs a volume to divide by, which is why f4 waits until Tallies.

The pin cell below is missing neutron importances on three of its five cells. Importance decides whether MCNP keeps tracking a particle, and a cell at imp:n=0 kills everything that enters — which is what you want at the problem boundary and nowhere else. Leave the entry off a physical region and histories can vanish there without an error: the run completes and the answer is quietly wrong. Edit the deck and the checker will re-run as you type.

These are the same checks the OWEN editor extension and the GROVES desktop app run against your own decks, so a deck that comes back clean here comes back clean in the tools. They catch syntax and convention mistakes, not physics errors — a deck can pass every check and still model the wrong reactor.