I'm attempting to run an anisotropic Migdal–Eliashberg calculation with EPW. When I use temperature values of 20 K, 40 K, 60 K ... 140 K, everything seems to be written correctly (see the scf.imag_aniso_* files for T = 140 K for reference). However, when I try to restart the calculation with T = 160 K, a file named "scf.imag_aniso_0*****" is written to the disk, and the calculation fails with the error log:
Can you make sure you have 'scf.imag_aniso_160.00' file present? It looks like the code did not find the file or could not read it. If you do not have the file for 160 k, you can read (restart) from 140 k and calculate at 160 k.
Thanks for your response! Shouldn't the code generate scf.imag_aniso_160.00 when I restart from 160 K (as in epw2r.in)? The problem seems to be that it writes the file name incorrectly as scf.imag_aniso_0***** and then has trouble reading it, as you point out. Why do you think I should restart from 140 K (which is already done) and not 160 K? Thanks very much!
Can you confirm that the 'epw2r.out' is produced from 'epw2.in'? They do not look consistent to me, because the first temp is 20 k in your input and 160 k in the output. Also, I see "Read from file delta and znorm on imaginary-axis" in your output file, which is only possible if you have "laniso = .true. and imag_read = .true." in your input.
If you use "imag_read = .true.", code reads the delta and znorm on imaginary-axis for the first temp (in your input) and calculates for the next temp (I meant this type of calculation is a restart calculation in anisotropic calculation).
In your case, you have already done 140 k, so it is recommended to read (not recalculate) the delta and znorm from 'scf.imag_aniso_140.00' file, (which will be the starting point for the next temperature, otherwise it has to start from the estimated BCS gap).
I apologize, it looks like I forgot to put some of the files in the folder. Indeed, epw2r.out is the output generated by epw2r.in, which is a restart calculation (imag_read = .true.). From what I understand, if I restart from 140 K, with all the files already present for 140 K, it will not recalculate them, but will calculate starting from the next temperature, which is 160 K. Is that correct? Thank you once again for your help!
As you suggested, I attempted to restart this calculation from T = 140 K, but it seems that the same error occurred as soon as the calculation started (see epw2r.in and epw2r.out). I listed all the files that have "140" in their names, and none of them seem to be modified in the attempted calculation. I added all of these files to the Dropbox folder. Also, I listed all the files modified in the attempt, which are the following:
I also added these files to the Dropbox folder for your reference. Overall, my guess is still that the scf.imag_aniso_0***** file is incorrectly created and messes up everything. But I don't understand why, since most of my other calculations using the same EPW compilation work fine. I'd greatly appreciate it if you could take another look at my files and let me know where I'm going wrong. Thank you so much!
Overall, my guess is still that the scf.imag_aniso_0***** file is incorrectly created and messes up everything.
You are right. I guess that the index of itemp or the array of gtemp containing temperatures has wrong values for some reasons.
You can confirm it by printing temp, gtemp and itemp in the subroutine of eliashberg_read_aniso_iaxis and eliashberg_write_aniso_iaxis.
( See https://gitlab.com/QEF/q-e/-/blob/devel ... g.f90#L111 )
Thank you very much for your comments. Indeed, the problematic file seems to be written to disk where you point out in io_eliashberg.f90. I ran other calculations with temperatures above 100 K (as well as this calculation which had run properly up to 140 K), so the problem shouldn't be about the number of digits in the temperature. Could you please instruct me more specifically about which lines to add to the code and where? I'm not experienced in editing Fortran code so I'm afraid I might mess up the program. Thank you very much!
I wanted to put this question back onto your radar. I want to be able to compute the full anisotropic SC properties of this material, and I'm stuck here. I was wondering if there's been any relevant updates to the EPW code in the five months since we had our last correspondence in this thread (so far I've tried with QE-6.6 and QE-6.7). Otherwise, could you please instruct me specifically on how to print the relevant parameters to pin down the issue? Thank you very much!