keff different in MCNP5, MCNPX, and MCNP6

Anomaly: | In an input file with more than 100
duplicate unique surfaces, the resulting keff value will be different
in MCNP5, MCNPX, and MCNP6. |

History: | This was a problem submitted to Randy
by a user. That input file was proprietary so this is one
Randy created that illustrates the problem. |

Versions affected: | MCNP5 gives an accurate response,
MCNPX aborts with an error, MCNP6 gives a different response. |

Impact: | In this example, keff for MCNP5 is
1.85133. For MCNP6, keff is 0.69175. MCNPX aborts with an
error. |

Work Around: | Don't create input files that have more than 100 duplicate unique surfaces. |

Steps required to reproduce this anomaly:

Note: Run this example in the RSICC release of each of these codes.

About the Input file:

The geometry consists of 192 cells. 96 cells containing 235U/water mix and 96 cells containing 239Pu.

The Uranium oxide cells are at (-100) and the 239Pu at 100.

The density in each set of cells is set to randomly change from cell to cell.

The complexity of this example is necessary to provide a realistic example of the use of over 200 duplicate surfaces. While not as common in non-macrobody geometries, duplicate surfaces are very common in geometries that use macrobodies heavily.

Click here to download the input file, irand.dat (27 kb).

Calculate keff:

In MCNP6, calculate keff.

In MCNP5, calculate keff.

Notice the different result.

Calculate the Total Fission Energy F7:

In MCNP6, calculate the Total Fission Energy (F7) in all cells.

In MCNP5, calculate the Total Fission Energy (F7) in all cells.

Notice the different result.

Results obtained from our test:

MCNP5:

keff:

| the final estimated combined collision/absorption/track-length keff = 1.85133 with an estimated standard deviation of 0.01088

F7:

cell union total 1.46966E-05 0.0065

MCNP6:

keff:

| the final estimated combined collision/absorption/track-length keff = 0.69175 with an estimated standard deviation of 0.00185

F7:

cell union total 6.71120E-06 0.0056

Digging deeper:

To further investigate this problem, try the following changes to the input file:

Change the number of particles per cycle from 100 to some larger number.

Move the source to the Plutonium 239

Use and SDEF volume source to sample all fissionable cells.