Nonlinear susceptibility calculated by HSE06

Dear everyone,
I’m trying to calculate the nonlinear susceptibility using the HSE06, so I set ixc to -428 in dataset 2. However, the non self-consistent calculation cannot proceed and gets stuck at this step. Did I make mistake somewhere? Thanks!
tnlo_6.abi (7.0 KB)
tnlo_6.abo (83.2 KB)

Hi Jiang,

in principle dtset 2 should work : if the SCF cycle goes through all the ingredients are there for the non scf as well. For the further dtsets with response function calculations and hybrid functionals this will probably not work (not implemented afaik @gonze @torrent??). You are running the GS with LDA then going for HSE. Why not start the GS with HSE as well?

From your abo file there is not much information about what is failing. Perhaps you have a log file, the MPI ERROR and LOG files? As usual following the netiquette (see the main header pages of this forum): you need to provide us with much more information on version, platform, compilers etc to give us a chance of finding the issue.

From the abo it seems the run stops when reading in the WFK file. It could be an issue of re-distribution of the information when running in parallel from one DTSET to the next, or it is missing some HF pieces which were not calculated in ds1. Try running with reduced ecut and kpts in serial - if that goes through the problem is probably in the parallelization of successive dtsets, or reading back in a WFK with a different distribution (in principle this should be agnostic, but not quite always). The number of kpt changes as well, so the WFK file has to be scanned for each new k. If you look at mkmem the number of k in memory is not distributed for dtset 2 (goes to 432), so it may simply be exploding the memory.

BTW your ecut 80 is a bit insane - this is probably not needed.

good luck

Dear mverstra,
Thank you very much. I have used some parallel parameters (nkpthf nbandhf np_spkpt nphf) and waited for some time to obtain the output. I will follow your advice and use HSE06 for GS calculation, and perform convergence test.
Best regards,