Photon out of range error

[hvbish]

I have encountered an error which only seems to happen when repeating boundary conditions are turned on. Most often this will occur many hours into a long run with >1e8 photons, but has occasionally happened as soon as 50 minutes into a run. There is nothing in the .err file, but the run fails and the .out file looks something like this:

Result of optional subdivide overdense cells
total number of cells = 86908736
number of orig cells = 86908736
number of orig subdiv cells = 8001
number of new subdiv cells = 0
working on wavelength [micron] = 1565
tau = 1 albedo = 0.6 g = 0.7 
frac_miss-- = 2.77556e-17
frac_miss++ = 1.11022e-16
frac_miss++ = 1.11022e-16
frac_miss++ = 1.11022e-16
frac_miss-- = 1.11022e-16
frac_miss++ = 5.55112e-17
frac_miss++ = 4.33681e-19
frac_miss-- = 1.11022e-16
frac_miss++ = 1.11022e-16
frac_miss++ = 5.55112e-17
frac_miss-- = 1.11022e-16
frac_miss-- = 5.55112e-17
frac_miss-- = 3.46945e-18
frac_miss++ = 1.11022e-16
frac_miss++ = 4.33681e-19
frac_miss-- = 1.73472e-18
frac_miss++ = 1.11022e-16
out_of_range Matrix::operator(), element 1
photon # = 1983267
num scat = 1
classify_scattered_photon - out of range.

I have found that using a different random number seed often solves the issue - but sometimes I need to try several different seeds before finding one that doesn't result in the error.

Attaching example param file here:
edenhofer_repeatbounds_star1_params_1.txt

Example input grids can be retrieved here:
https://www.dropbox.com/scl/fi/h2d9m74ffzaa2rlh9g34z/DIRTY_Edenhofer_output_grid_pos_edenhofer_March2025.fits
https://www.dropbox.com/scl/fi/yi0jo9s38di0uc1a4td8w/DIRTY_Edenhofer_output_grid_tau_ref_per_pc_edenhofer_March2025.fits

[karllark]

Thanks for the info. Will think about it carefully and hope to come up with a fix.

[hvbish]

Update: I have now encountered this error with repeat_boundary_xy set to 0 and a single star placed at x,y,z=0,0,0. In this case, however, DIRTY only ran for about one minute before failing with the photon out of range error. If I set the star location to x,y,z=1,1,1 I do not encounter the error.

So repeating boundary conditions does not seem to be the culprit after all...

Here is the param file used:
artifact_test_tau_params_0.txt

And the grid files:
https://www.dropbox.com/scl/fi/kdm6tac1hsy6w683y97yf/DIRTY_Edenhofer_output_grid_pos_edenhofer_TAU_TEST.fits?rlkey=wpf3ilso6gjq780izh2rmxwvg&st=9nfh6f0x&dl=0
https://www.dropbox.com/scl/fi/r5se8c9e4horwwq5jrrvj/DIRTY_Edenhofer_output_grid_tau_ref_per_pc_TAU_TEST.fits?rlkey=h4tdtsjsfc9mofzrhpp94tptm&st=7dcw51l6&dl=0

[karllark]

This latest error is unrelated. It happens when the star at 0, 0, 0 and the internal observer are at the same location. When deciding where in the sky to put the stellar photon (direct photons from the star), it crashes as it is actually anywhere in the sky image. As long as the internal observer and star are not at the same location, then this issue will not appear and that is why moving the star to 1,1,1 means no crash.

So, this is a separate bug and I should add a check on input for this case and not allow it as the math just doesn't work.

[karllark]

The latest error was not unrelated, but it was a different error. But it did provide a valuable clue! Also thanks for posting the output with the error details. That was also critical for debugging the issue.

PR #54 should fix both related bugs. Please re-open this issue if that is not the case.

[hvbish]

Still experiencing this same 'photon out of range' error with the latest version, unfortunately. I apparently do not have permission to reopen the issue.

[karllark]

I have reopened this issue. Is it exactly the same error as in the issue description? If different, please post the new error message and how to reproduce it.

Have you confirmed that it is the repeating boundary condition by running the same model with the same number of photons w/ and w/o the repeating boundary condition and shown that it doesn't happen w/o the repeating boundary condition? Really, really helpful to know the answer to this question as it gives an important clue to where to look in the code.

[hvbish]

The output is slightly different in that each line starting frac_miss now lists the photon number as well (a change that I believe you implemented in the latest version). But yes, the error I encounter is the same (output ends with classify_scattered_photon - out of range. and then the run fails).
The last four lines output from my last run are:

out_of_range Matrix::operator(), element 1
photon # = 12933986
num scat = 0
classify_scattered_photon - out of range.

I will test whether the same issue occurs when I turn off repeating boundary conditions and let you know.

[karllark]

Thanks for this.

Can you put in a few lines that show frac_miss? I believe I changed the format and just want to see it.

I look forward to the test w/o repeating boundary conditions.

[hvbish]

Yep, here you go:

frac_miss++ = 1.11022e-16; photon # = 12643257
frac_miss-- = 2.77556e-17; photon # = 12711087
frac_miss-- = 1.11022e-16; photon # = 12711320
frac_miss-- = 1.11022e-16; photon # = 12765207
frac_miss++ = 1.11022e-16; photon # = 12862916
frac_miss-- = 2.77556e-17; photon # = 12922832
out_of_range Matrix::operator(), element 1
photon # = 12933986
num scat = 0
classify_scattered_photon - out of range.

I'll update you with the results of the boundary conditions test.