EPW Forum

Hello!
I am testing the calculations by EPW. The problems is like below, when doing the procedure "Solve anisotropic Eliashberg equations on imaginary-axis", it output like "BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES". I also try changing the etf_mem from 0 to 2, it does not work. Can anyone give me any suggestions? Thanks a lot.
_________
intel18+mkl18+impi, etf_mem=0 give the following output:


temp( 1) = 0.0005 K

Solve anisotropic Eliashberg equations on imaginary-axis

Total number of frequency points nsiw ( 1 ) = ******
Cutoff frequency wscut = 1.0000


===================================================================================
= BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES
= PID 13981 RUNNING AT compute1511
= EXIT CODE: 9
= CLEANING UP REMAINING PROCESSES
= YOU CAN IGNORE THE BELOW CLEANUP MESSAGES
===================================================================================

_____
when using intel12+mkl12+openmpi, the output is:
temp( 1) = 0.0005 K

Solve anisotropic Eliashberg equations on imaginary-axis

Total number of frequency points nsiw ( 1 ) = ******
Cutoff frequency wscut = 1.0000

4 total processes killed (some possibly by mpirun during cleanup)

__________
by playing with some input parameters, I now get the following message from the output:
Total number of frequency points nsiw ( 1 ) = 184691
Cutoff frequency wscut = 1.0000


Size of required memory per pool : ~= 5.7537 Gb

AKeri is calculated on the fly since its size exceedes max_memlt

The pbs jobs are still there, but the calulations was stuck without updating;
I also tried to give max_memlt with bigger values (>5.8 Gb), but the calculations then give "= BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES" again.
I will appreciate it very much if someone can help me fix the problem.

Hi,

I believe the code is crashing because the input temperature at which the anisotropic Eliashberg equations need to be solved is very small (0.0005 K) and then the number of imaginary frequency points nsiw is very large (it this case more than one million since it is printed as ****).

WRITE(stdout,'(5x,a,i6,a,i6)') 'Total number of frequency points nsiw ( ', itemp, ' ) = ', nsiw(itemp)

The equations should be solved in the vicinity of 0 K, but I wouldn't go below 1 K. You may also want to reduce wscut. What is the estimated critical temperature with McMillan formula in your case?

Best,
Roxana

roxana wrote:Hi,

I believe the code is crashing because the input temperature at which the anisotropic Eliashberg equations need to be solved is very small (0.0005 K) and then the number of imaginary frequency points nsiw is very large (it this case more than one million since it is printed as ****).

WRITE(stdout,'(5x,a,i6,a,i6)') 'Total number of frequency points nsiw ( ', itemp, ' ) = ', nsiw(itemp)

The equations should be solved in the vicinity of 0 K, but I wouldn't go below 1 K. You may also want to reduce wscut. What is the estimated critical temperature with McMillan formula in your case?

Best,
Roxana

Dear Roxana,
Thanks a lot for your kind reply. In my case, the Tc is very very low (e.g. using mu=0.05, the Tc is ~0.001 K). I know my system is not physically meaningful, I just want to test if the code can fit my systems for calculations. Now I guess I know where does the problem come from, following you comments above. Thanks a lot.