Odd EnergyFilter behaviour

User Support
1

I’ve noticed the following odd behaviour when plotting energy spectra using defined energy grids (or I am fundamentally misunderstanding this filter). I will demonstrate with the pincell example:

tallies = openmc.Tallies()

#create two sets of energy tallies: 
#one set with uniform log energy grids with np.logspace and 
#the second set with pre-defined grids

#evenly-spaced grids
num_groups = [75, 150, 300, 600, 1200]
for n in num_groups:
    groups = np.logspace(-5, log10(20e6), n)
    tally = openmc.Tally(name=str(n))
    tally.scores = ['flux']
    efil_in = openmc.EnergyFilter(groups)
    tally.filters = [efil_in]
    tallies.append(tally)

#pre-defined grids
scale_energy_grid = [1.000E-5, 1.000E-4, ... 1.733E+7, 2.000E+7 ] #shortened for easier viewing
grids = ['CASMO-70', 'XMAS-172', 'SCALE-238', 'SHEM-361', 'UKAEA-1102']

for grid in grids:
    if grid=='SCALE-238':
        groups = scale_energy_grid
    else:
        groups = openmc.mgxs.GROUP_STRUCTURES[grid]
    tally = openmc.Tally(name=grid)
    tally.scores = ['flux']
    efil_in = openmc.EnergyFilter(groups)
    tally.filters = [efil_in]
    tallies.append(tally)

tallies.export_to_xml()
openmc.run()

When plotting the results of all these EnergyFilter tallies we see fairly smooth results for the set of tallies using np.logspace, but wildly oscillating spectra from the set which used defined group structures:

image
image774×627 94 KB

image
image772×613 96.2 KB

Again I might be misunderstanding how the filter works, but some clarification would be much appreciated!

Thanks,

Sourena

2

@sourena Did you divide each flux value by its energy width?

energies = openmc.mgxs.GROUP_STRUCTURES['XMAS-172']

ax.loglog(energies[:-1], mean_7/np.diff(energies),  label='XMAS-172')
1 Like
3

@Pranto’s suggestion is right on. In fact, I recently put together a Jupyter notebook showing how to tally and plot a flux spectrum and this is mentioned (the need to divide by energy bin width).

1 Like
4

Thank you both, you are absolutely correct. The plots shown above were a straight flux plot, and were from a simplified version of the model in which the odd spectra were originally encountered. I revisited the original model after seeing your notes and the energy width calculation was being performed incorrectly. Thank you again for your help.

5

I am a new user of OpenMC. Very nice job, and thanks to all creators!

I did my first simulations. However, I have some misunderstandings with the interpretation of data. My question was: When using an Energy filter with non-equidistant bin size, what quantity do OpenMC scores, dN or dN/dE? Does the corresponding bin size divide data in each bin, i.e., the distribution is created? Or are data not divided by bin size, i.e., dN is scored?

However, after reading this thread and Jupyter Notebook Viewer
I think that I know the answer - only dN is scored. Please, am I right?

6

Yes number of scores in each bin dN

Not bin density dN/dE

Perhaps related, we do have some methods on the EnergyFilter like .lethargy_bin_width if you want to do some lethargy plotting and need to divide by the bin width. Example here

1 Like
7

Thank you so much for the answer!

My best regards,
Karel