cutoff test didn't converged

[LEI CHEN]

Hi ,

  I 'm doing the test of optimising the cutoff test according to the cp2k page (periodic MOF structure ). what I  tried is from 300 to 500. however, I found that there got some value that is hard to reach SCF convergence (300, 420, 430, 460, 470) or converged at a strange value that don't follow the trend (when you enhance the cutoff value ,it would converge near one value, 410, 450, 480 490) 

310    -538.32547
320    -538.32534
330    -538.32417
340    -538.32417
350    -538.32287
360    -538.32287
370    -538.32287
380    -538.32287
390    -538.32287
400    -538.32287
410    -538.33887 
440    -538.32259
450    -538.33843
480    -538.33901
490    -538.33901
500    -538.32268

  could anybody help me with that. the input file is attached. thanks a lot for any help.

[Anna Hehn]

Dear Lei Chen,

I would recommend to try out larger cutoffs when dealing with MOFs, starting e.g. with a CUTOFF of 900 and a REL_CUTOFF of 80.

Best regards,

Anna

--
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.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/95e1ec3a-6aea-463f-ab05-f89f2fce70c0n%40googlegroups.com.

[bty...@gmail.com]

Hi Anna,

 

  Thanks for your help. I still got some question about input files:

 

1. Do I need to add smear section when calculating the spe or optimising the cell and structure?
2. Should ngrids parameter be changed after the cutoff and rcutoff increase?

 

Thanks for your help again.

 

Sincerely,

 

Lei Chen

--
You received this message because you are subscribed to a topic in the Google Groups "cp2k" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/cp2k/x2hXheHwV6o/unsubscribe.
To unsubscribe from this group and all its topics, send an email to cp2k+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/CACM8JBkt48gCJOQqyoWOLAuxHHVeDPhn-0dFC0YqPmAkDMpZ_A%40mail.gmail.com.

[Sam Broderick]

Hi Lei Chen

1. Smearing is for metals without an energy gap, if I understand it correctly. Do MOF's have gaps?

2. Whatever gives you the most efficiency. See the tutorial on a convergence study (https://www.cp2k.org/howto:converging_cutoff).

Regards

Sam

[Lucas Lodeiro]

When you use smoothers for density and XC potentials, the convergence of total energy with respect to PW cutoff is not asymptotic... it gets a "noisy" behaviour when you approximate the convergence point. They enables to use lower cutoff values with good results, but the convergence cutoff test does not behave as is expected.

Regards - Lucas

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/da828bb8-0f08-4d5f-8af0-ad20db74cd2bn%40googlegroups.com.