Inorganic optimization SCF convergence

89 views
Skip to first unread message

nat zip

unread,
Dec 11, 2023, 9:25:51 PM12/11/23
to cp2k
Hello all,

I am having trouble getting an inorganic supercell of LaCl to optimize. I have tried using diagonalization, DIIS and CG minimizers with OT, and varying ALPHA and EPS_SCF. It is important to note that the system is +1 charge due to a missing Cl atom. None of the SCF iterations converge before reaching 300 steps. Thank you for any suggestions. 

Best regards,
Kyle
4344186_iopt.inp
4344186_iopt4.out

Jürg Hutter

unread,
Dec 12, 2023, 10:02:49 AM12/12/23
to cp...@googlegroups.com
Hi

I would
1) set EPS_DEFAULT 1.E-12
2) run a maximum of 50 SCF steps e.g. 20 x 2 (inner/outer SCF)
3) If your gradient at the end of the first unconverged SCF is small
enough, e.g. 0.001, restart with DIIS method.
4) if nothing works, try first some GEO_OPT steps with EPS_SCF 1.E-6 or
even higher, then switch back for final optimization.

regards
JH

________________________________________
From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of nat zip <natzi...@gmail.com>
Sent: Tuesday, December 12, 2023 3:25 AM
To: cp2k
Subject: [CP2K:19648] Inorganic optimization SCF convergence
--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com<mailto:cp2k+uns...@googlegroups.com>.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/f07c4dac-3566-4adb-ad6d-f366663fda30n%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/f07c4dac-3566-4adb-ad6d-f366663fda30n%40googlegroups.com?utm_medium=email&utm_source=footer>.

nat zip

unread,
Dec 13, 2023, 10:03:25 AM12/13/23
to cp...@googlegroups.com
Hello Jürg,

Thank you for reaching out. After changing EPS_DEFAULT, I found that RMS Gradient was 0.58E-3 and restarted with DIIS. The SCF calculations begin to converge, but the calculation ends once 200 optimization steps are reached. Should I increase MAX_ITER or attempt a GEO_OPT first?

Best regards,
Kyle

To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/ZR0P278MB0759C722AA5E7CCC6F7695339F8EA%40ZR0P278MB0759.CHEP278.PROD.OUTLOOK.COM.
Reply all
Reply to author
Forward
0 new messages