Possible problem tallying photon heating with weight windows enabled

[pshriwise]

Bug Description

I was staring at the section on photon heating today and noticed that we may be subtracting too much energy from photon heating tallies when weight windows are applied.

openmc/src/tallies/tally_scoring.cpp

Lines 958 to 981 in 2d988a6

	case HEATING:
	if (p.type() == Type::neutron) {
	score = score_neutron_heating(
	p, tally, flux, HEATING, i_nuclide, atom_density);
	} else {
	if (i_nuclide == -1 || i_nuclide == p.event_nuclide()) {
	// The energy deposited is the difference between the pre-collision
	// and post-collision energy...
	score = E - p.E();
	// ...less the energy of any secondary particles since they will be
	// transported individually later
	const auto& bank = p.secondary_bank();
	for (auto it = bank.end() - p.n_bank_second(); it < bank.end();
	++it) {
	score -= it->E;
	}
	score *= p.wgt_last();
	} else {
	score = 0.0;
	}
	}
	break;

Based on how the n_bank_second attribute is incremented in the Particle::create_secondary method and that this method is applied when splitting particles in weight window application, it seems particles split in application of weight windows will be incorporated into the correction, which I don't believe is correct.

openmc/src/weight_windows.cpp

Lines 116 to 118 in 2d988a6

	for (int l = 0; l < i_split - 1; l++) {
	p.create_secondary(weight / n_split, p.u(), p.E(), p.type());
	}

I'll update this issue when I have more information.

[giovanni-mariano]

Hello Patrick! I faced this issue as well during my tests but I haven't had much time to spend on it.
My conclusion (I think I wrote something on slack) was that it affect all the tallies which are not using the track-length estimator: basically all the tallies using the pre-collision weight (p.wgt_last()) in the score calculation instead of the current weight (p.wgt()). But I'm not sure on 100%.
Best,
Giovanni

[pshriwise]

Yeah I need to dig into this more soon. I'll try to come up with a minimal working example to start -- hopefully a problem where this effect is particularly pronounced.

[SteSeg]

I think the problem stands. A while ago I was running some fusion benchmarks with nuclear heating as tally and I remember I got weird results. Photon heating with weight windows was indeed negative in almost all positions.

Is it possible that it only happens when also photon transport is affected by ww? I think I got the negative results only when I tried to generate ww for both neutrons and photons while ww for only neutrons seem to work fine.

Here the results without weight windows (track-length estimator):

Here the results for the same tally but with weight windows for only neutrons (track-length estimator):

Here same tally with weight windows for both neutrons and photons (track-lenght estimator):

However, I'm not sure it is an estimator issue, these are results for the same tally with ww for heating and a collision estimator (only neutron ww, collision estimator):

Lastly, in the benchmark the case withouth ww and the case with ww only for neutrons behave very well when compared with experimental results:

[pshriwise]

Super helpful!! Thank you @SteSeg.