axiSEM crashing on some heterogeneity models and periods

This question is reposted from the cig-seismo mailing list.


I’ve been using axiSEM to simulate PKP waves through models with lateral heterogeneities. I’ve found that on some runs, while the code completes without errors, the seismogram outputs (in the Data directory) are almost full of NANs. I’m running simulations at 2s period, 2000 s length, and with a fluid scaling factor of 0.7 and wavelengths per element of 2.25 to make the PKP waves clearer. With one heterogeneity model the code ran correctly and the resulting seismogram files were good (run #1). However, then I changed the heterogeneity model (keeping everything else the same) and the code ran but the seismograms were full of NANs (run #2). I then ran the same heterogeneity model but at 10 s period and the code ran correctly and the seismograms were good (run #3). Could you please direct me to the possible source of the problem? I’ve attached the output files for failed 2s run (run #2) and the working 10s run (run #3). The only thing that I changed is the dominant period. I would have included an output from the working 2s run (run #1) but I accidentally overwrote the OUTPUT file. I’ve also included one of the failed seismogram outputs.

OUTPUT_ak135_2s_np544_2000s_2.25EL_FSF0.7_sngl_SLAB_AlaskaSR_65_-155to_58_-145.txt (139.6 KB)
OUTPUT_ak135_10s_np128_2000s_sngl_SLAB_AlasSR_70_-168to_53_-140.txt (42.2 KB)
S100.025_A1_disp.dat.txt (3.8 MB)