Page 1 of 1

D-field freezes in parallel (and tips on converging?)  [SOLVED]

Posted: Sat Mar 21, 2020 10:55 am
by dbennett1994
Hi all,

I'm trying to run a set of finite (reduced) D-field calculations for a 5 atom unit cell perovskite and am having some problems. I am doing a relaxation of the atomic coordinates and the lattice vector that lines up with the D-field. When I run the calculation in parallel (4 cores), the first calculation with zero field runs fine, but the finite field calculation freezes in the SCF cycle, always on the 3rd step.

lao.in
(6.75 KiB) Downloaded 199 times
lao_parallel.out
(51.98 KiB) Downloaded 207 times
log_parallel.out
(146.02 KiB) Downloaded 212 times

When I run in serial, this doesn't happen. Similar problems were reported a few years ago here, but I don't think a solution was found.

As a separate issue, the finite D-field calculations are not converging in serial, and I would appreciate any advice on getting them to converge. I was using a red_dfield of around 10% of the reduced polarisation and the program suggested to try using a smaller one. Even when I go to 1% of the reduced polarisation it still doesn't converge.

lao_serial.out
(52.04 KiB) Downloaded 184 times
log_serial.out
(177.83 KiB) Downloaded 209 times

Thank you

Re: D-field freezes in parallel (and tips on converging?)

Posted: Mon Apr 20, 2020 8:30 am
by ebousquet
Dear dbennett1994,
The applied D-field has not been implemented properly and the developer did not make it correctly for parallel.
Same applies for phonons under applied electric field, only sequential up to somebody put his hands in...
Best wishes,
Eric

Re: D-field freezes in parallel (and tips on converging?)

Posted: Wed Apr 22, 2020 12:07 pm
by dbennett1994
Dear Eric,

Thanks for letting me know. I'll mark this as solved now

Danny