Photon spectrum difference between PHITS and OpenMC

Norbert_Schmidt · August 18, 2021, 6:18pm

Hi all,

I executed the Broomstick problem with OpenMC 0.12.2 and PHITS 3.24, using ENDF/B-
VIII.0, and found that the results for the photon spectrum differ, particularly for the U235 case with density 18.95 g/cm³ and 1MeV neutron source. For the neutron spectrum, no significant differences are observed.

I observed it when comparing the results obtained with an old version of OpenMC that I had downloaded a while ago (0.12.0-dev), and in this case they are consistent with the values reported by PHITS.

For all 0.12.X stable versions the same results are obtained, so I guess that the difference could be in the development between stable versions 0.11.0 and 0.12.0 (I tried to execute the model with OpenMC 0.11.0, but I get a Segmentation Fault).

Attached are two graphics where you can appreciate the differences mentioned.

Thanks in advance,

Norbert

paulromano · September 3, 2021, 4:34pm

Hi @Norbert_Schmidt. After thinking about this some, I believe the cause of the difference you are seeing is the result of a method used in OpenMC where the total photon yields from neutron reactions are scaled slightly higher to account for production of delayed photons from fission (hence why you are seeing this in U235). This method makes sense for reactor calculations when you assume the reactor is operating at steady state (and have a relatively stable population of fission products producing delayed photons). For other systems, it may not make sense. If you want to turn this off for the sake of comparison to PHITS or another code, add the following to your code:

settings = openmc.Settings()
settings.delayed_photon_scaling = False
...
settings.export_to_xml()

Here is the pull request where this was added to OpenMC:

github.com/openmc-dev/openmc

Fix/improve energy deposition for coupled neutron-photon calculations

openmc-dev:develop ← paulromano:photon-energydep-fixes

opened 10:23PM - 13 Jan 20 UTC

paulromano

+328 -91

This pull request makes two important changes: - In a coupled neutron-photon …calculation, tallying `heating-local` right now does not give the correct answer. The problem is that photons end up contributing to this score when they should not. The solution is to just ensure that when heating-local is being scored, only neutrons contribute. - Also in a coupled n-g calculation, right now OpenMC doesn't account for energy released as delayed photons when you ask for the `heating` score. The photon yield and energy spectrum that is provided in the data files only corresponds to the prompt photons. To account for the energy of delayed photons, what I've done here is to artificially increase the fission photon yields to account for the energy release of delayed photons. If EGP and EGD are the average energies released as prompt and delayed photons, the prompt yields are multiplied by (EGP + EGD)/EGP. This is an approximation because 1) the actual yield of delayed photons depends on the power history (at time 0 there are _no_ delayed photons) and 2) their spectrum is not the same as the prompt spectrum. But, usually it is better to include that energy than not. I have put in a new setting (`Settings.delayed_photon_scaling`) so that this behavior can be turned off if desired. One other unrelated change is that when messages are displayed to stdout, I've made sure to flush them (via `std::endl`) so that buffering doesn't result in delayed output (this tends to occur if you are running OpenMC from IPython/Jupyter notebook). ## Results I've done some simple tests looking at energy deposition for a PWR pin cell problem to confirm that the changes here give answers that make sense. For the full results, have a look at [this notebook](https://nbviewer.jupyter.org/gist/paulromano/89c78dc74d98215a8f2a406753b2215d?force_flush=true), but the summary is as follows: - `'heating` and `heating-local` both give reasonable answers when used to determine the average amount of fission energy deposited per fission (surprise: around 200 MeV) - In theory, `heating` and `heating-local` _should_ produce the same answer if there is no leakage in the problem, but they don't for a number of reasons: 1. In the ENDF/B-VII.1 data, there is a mismatch between the energy release data from MT=458 and the prompt fission yield given. [This paper](https://doi.org/10.1016/j.anucene.2019.02.003) discusses it in more detail. 2. Right now, our thick-target bremsstrahlung treatment does not conserve energy, so we end up producing a little more energy than we should. However, this ends up being a small effect compared with the effect from the data mismatch (in the notebook, I ran the calculations with TTB turned off). @drewejohnson You're probably most familiar with the motivation for this change; care to review?

Norbert_Schmidt · September 3, 2021, 6:54pm

Thank you very much @paulromano for your answer! Indeed that was the reason for the apparent differences between the results (I didn’t know that the delayed_photon_scaling option is declared as True by default). The new results can be seen in the next figure:

Thanks again and best regards,

Norbert

Topic		Replies	Views
Tally Uncertainty; OpenMC vs. MCNP User Support	13	1111	August 2, 2023
Photon transport discrepancies with MCNP and Serpent Validation and Verification	8	1384	December 17, 2021
Benchmark against MCNP and Tripoli4 unsatisfactory Validation and Verification	9	434	July 6, 2023
Photon flux discrepancies with EGSnrc, MCNP, and Serpent Validation and Verification	10	480	January 20, 2023
The energy spectra of neutron point sources calculated by mcnp and openmc are inconsistent User Support	4	147	July 31, 2023

Photon spectrum difference between PHITS and OpenMC

Related Topics