Page 1 of 1

BSE soveler havs troubles in convergence

Posted: Mon Nov 16, 2015 9:37 am
by liusjtc
Dear yambo developers


I have some trouble in solving the BSE equation. When I set the BSHayTrs=-0.070, then start the job, everything seems going well at first. However, after several iterations, I found the job is still running but the files(l.optical_...) never update anymore.

For example <19s>[Haydock] Iteration 15 Accuracy :0.07541|-0.07000
<20s>[Haydock] Iteration 16 Accuracy :0.07741|-0.07000
then it didn't update anymore.
I have no idea whats going on, please help me out. So what should I do to make it complete the calculation?

Thanks
Best regards
Liu SHijie
Shanghai Jiaotong university

Re: BSE soveler havs troubles in convergence

Posted: Mon Nov 16, 2015 10:48 am
by Daniele Varsano
Dera Liu,
can you please post your input/l_* and report files?
We will have a look and try to understand what is happening.

Best,
Daniele

Re: BSE soveler havs troubles in convergence

Posted: Mon Nov 16, 2015 11:22 am
by liusjtc
Please,thanks.

Re: BSE soveler havs troubles in convergence

Posted: Mon Nov 16, 2015 12:02 pm
by myrta gruning
Dear Liu

it may be a problem of communication among the nodes, e.g. the calculation is converged, but the master node did not receive the 'OK' from all other nodes for some reason (we will look at that)
Anyway the Haydock updates (if everything is fine) the output file o_* at every iteration. You can check if those were written and at which iteration. As well you can compare the spectra at the last with the last minus one iteration (e.g. column 2 and column 6) in e.g. o.eps_q1_haydock_bse
to see if the spectrum is converged.
Note as well that by restarting the calculation, yambo should complete the iterations.

Let us know. Best,
Myrta