[CP2K-user] [CP2K:12951] transition_metal_oxide-all_electron_GAPW-Preconditioner_SCF_convergence

[Krack Matthias (PSI)]

Hi Min

The electron count is quite poor. I suggest to reduce EPS_FIT, since the soft basis is quite small, however, this might not solve necessarily your convergence problem.

Best

Matthias

From: cp... at googlegroups.com <cp... at googlegroups.com> On Behalf Of ??
Sent: Donnerstag, 5. März 2020 16:57
To: cp2k <cp... at googlegroups.com>
Subject: [CP2K:12951] transition_metal_oxide-all_electron_GAPW-Preconditioner_SCF_convergence


Dear cp2k experts,



I am doing the single point calcualtions of 3d transition metal oxide, using all electron GAPW method.



When I using preconditioner= FULL_ALL, some cases converge very fast, in several tens of scf cycles, some cases can not converge, the charge distance becoming constant, like this:



grep 'OT CG' waste1/output_75532 |tail

    21 OT CG       0.18E-11    7.3     0.01617359    -29825.8903476335  1.12E-08

    23 OT CG       0.15E-12    7.3     0.01617359    -29825.8903476348 -1.27E-09

    25 OT CG       0.84E-14    7.3     0.01617359    -29825.8903476335  1.25E-09

    27 OT CG       0.34E-13    7.3     0.01617359    -29825.8903476348 -1.34E-09

    29 OT CG       0.67E-13    7.3     0.01617359    -29825.8903476331  1.73E-09

    31 OT CG       0.13E-12    7.3     0.01617359    -29825.8903476329  2.33E-10

    33 OT CG       0.50E-14    7.3     0.01617359    -29825.8903476336 -6.91E-10

    35 OT CG       0.10E-13    7.3     0.01617359    -29825.8903476325  1.05E-09

    37 OT CG       0.40E-13    7.3     0.01617359    -29825.8903476306  1.89E-09

    39 OT CG       0.16E-12    7.3     0.01617359    -29825.8903476311 -4.33E-10




I feel like that it's very strange, because those case have same chemical componment,different space group, all other setup keep the same. (see the attchment)



What is the reason ?



When I use preconditioner= FULL_KINETIC, both cases get converge, but very slow,(several thousand scf cycles).



So it looks like that FULL_ALL can get very fast converge, but sometimes is promatic, how to solve it?

Best,



Min

--
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 cp... at googlegroups.com<mailto:cp... at googlegroups.com>.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/befac638-36c1-4f8e-9220-c5b2c565a27e%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/befac638-36c1-4f8e-9220-c5b2c565a27e%40googlegroups.com?utm_medium=email&utm_source=footer>.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20200305/0348e9a1/attachment.htm>