MCNP vs Serpent pin-cell burnup discrepancy in keff evolution

[emilmammadzada]

TL;DR

MCNP vs Serpent pin-cell burnup discrepancy in keff evolution

Hello Dear Experts,
I am performing a simple pin-cell burnup (depletion) calculation using MCNP and Serpent and comparing the keff evolution as a function of time and burnup.


Despite modeling the same physical system in both codes (geometry, materials, initial isotopic composition, power normalization, burnup steps, boundary conditions, and thermal scattering treatment), I observe non-negligible differences in keff values and depletion behavior between MCNP and Serpent.


The general trend of keff decrease with burnup is consistent, but the absolute keff levels and burnup-dependent behavior do not match quantitatively.


I would appreciate insights on:

• Common physical or numerical sources of MCNP–Serpent discrepancies in depletion problems
• The impact of depletion solvers, normalization methods (power vs flux), or nuclear data (decay, fission yields)
• Key input parameters that must be carefully aligned to obtain consistent burnup results
• Recommended benchmarking practices for pin-cell depletion validation

I will provide both MCNP and Serpent input files for direct comparison.


Any suggestions or references would be greatly appreciated.

 

[emilmammadzada]

@Alex A

 

[Astronuc]

Achieving 30 GWd/tU within 250 days is impressive. I must read the text files before asking further questions or making comments. It would be interesting to compare with other codes like CASMO, WIMS and OpenMC.

 

[emilmammadzada]

Achieving 30 GWd/tU within 250 days is impressive. I must read the text files before asking further questions or making comments. It would be interesting to compare with other codes like CASMO, WIMS and OpenMC.

Thank you for your interest. I have attached the MCNP and Serpent input files for direct comparison. I would appreciate any feedback after your review.

 

[Alex A]

I usually avoid commenting on something so far away from what I know, that said, I think I've spotted something. The main difference is that SERPENT predicts the reactivity increasing, something consistent with burning off a poison. The MCNP prediction doesn't have this, and the input file is only told to burn material 1, with the boron poison being in the moderator material 4 and treated as a constant for the burn. Neither scenario is wrong, a reactor in a large pool of circulating water with the boron being replenished is just as realistic and valid, but these are different scenarios. I don't know SERPENT so I might just be wrong.

 

[emilmammadzada]

I usually avoid commenting on something so far away from what I know, that said, I think I've spotted something. The main difference is that SERPENT predicts the reactivity increasing, something consistent with burning off a poison. The MCNP prediction doesn't have this, and the input file is only told to burn material 1, with the boron poison being in the moderator material 4 and treated as a constant for the burn. Neither scenario is wrong, a reactor in a large pool of circulating water with the boron being replenished is just as realistic and valid, but these are different scenarios. I don't know SERPENT so I might just be wrong.

Thank you, that makes sense.
For a physically consistent MCNP–SERPENT comparison, how would you recommend defining the burn card in MCNP (which materials to include in depletion and how to define it)?