Heating tally and nuclides does not match tally without nuclides

[jon-proximafusion]

When running a simulation with a heating score split by nuclides or not split by nuclides I am seeing different results in the total heat in the tally

I would expect the heat deposited in the tally is equal to the sum of the heat on all nuclides, this is what I see when I split up a score like '(n,Xt)' across nuclides. They sum to the same value as when there is no splitting by nuclides.

I notice this bug only when photon transport is enabled settings.photon_transport = True

Bug Description

minimal example, run this script, I have two tallies one with nuclides being split and the other without. The tally results are printed as a pandas dataframe and you can see that they don't have the same total heat

import openmc

material = openmc.Material()
material.add_nuclide('Cu63', 1.0)
material.add_nuclide('Cu65', 1.0)
material.set_density('g/cm3', 8.96)
materials = openmc.Materials([material])

sphere = openmc.Sphere(r=10, boundary_type='vacuum')
region = -sphere
cell = openmc.Cell(region=region, fill=material)
geometry = openmc.Geometry([cell])

settings = openmc.Settings()
settings.particles = 1000
settings.batches = 10
settings.inactive = 0
settings.photon_transport = True  # if this is set to False, then the heating tallies match
settings.run_mode = 'fixed source'

my_source = openmc.IndependentSource()
my_source.space = openmc.stats.Point((0, 0, 0))
my_source.angle = openmc.stats.Isotropic()

settings.source = my_source

heat_tally = openmc.Tally(name='heating-with-nuclides')
heat_tally.filters = [openmc.CellFilter(cell)]
heat_tally.scores = ['heating'] 
heat_tally.nuclides = ['Cu63', 'Cu65']

heat_tally2 = openmc.Tally(name='heating-without-nuclides')
heat_tally2.filters = [openmc.CellFilter(cell)]
heat_tally2.scores = ['heating']

my_tallies = openmc.Tallies([heat_tally, heat_tally2])

model = openmc.Model(geometry, materials, settings, my_tallies)
sp_filename = model.run()#apply_tally_results = True)

sp_tally = openmc.StatePoint(sp_filename).get_tally(name='heating-with-nuclides')

df = sp_tally.get_pandas_dataframe()
print(df)

print()
sp_tally = openmc.StatePoint(sp_filename).get_tally(name='heating-without-nuclides')

df = sp_tally.get_pandas_dataframe()
print(df)

Environment

openmc develop, also tried with last stable release
ubuntu 24.04

[shimwell]

Ah I just realised this is perhaps because photon cross sections are by element not per nuclide. This makes it a bit harder to divide up the heating deposited by a photon to nuclides. Assuming we can find the element that gets the heat deposited then perhaps we could just divide up the heat according to the relative nuclide number density within the element.