[CP2K-user] [CP2K:11379] Re: hybrid functional revPBE0-D3 gets stuck with no errors outputted

Luis Ruiz Pestana lar... at lbl.gov
Thu Mar 7 17:21:31 UTC 2019


Thanks for the pointers! I finally made it work starting from a previously
converged meta-GGA calculation, using EPS_SCHWARZ 1.0E-6, and EPS_PGF_ORB
1.0E-15 (wouldn't work for higher values of this parameter, complaining
that the Kohn-Sham matrix is not 100% occupied). Here are the timings:

Step     Update method      Time    Convergence         Total energy
Change

------------------------------------------------------------------------------
     1 OT DIIS     0.15E+00  555.7     0.00082967     -2039.1909073186
-2.04E+03
     2 OT DIIS     0.15E+00  354.5     0.00032946     -2039.4948432943
-3.04E-01
     3 OT DIIS     0.15E+00  355.1     0.00015380     -2039.5448765937
-5.00E-02
     4 OT DIIS     0.15E+00  354.9     0.00007709     -2039.5495586907
-4.68E-03
     5 OT DIIS     0.15E+00  355.2     0.00002709     -2039.5501113855
-5.53E-04
     6 OT DIIS     0.15E+00  355.2     0.00001723     -2039.5501685709
-5.72E-05
     7 OT DIIS     0.15E+00  356.3     0.00000431     -2039.5502010572
-3.25E-05
     8 OT DIIS     0.15E+00  354.9     0.00000239     -2039.5502031502
-2.09E-06
     9 OT DIIS     0.15E+00  354.8     0.00000112     -2039.5502038116
-6.61E-07
    10 OT DIIS     0.15E+00  355.0     0.00000048     -2039.5502039566
-1.45E-07

  *** SCF run converged in    10 steps ***

Btw, I also had to play a little bit with the MEMORY parameter because for
values above ~2000 the simulation was crashing (e.g. WARNING: New heap
segment map at 0x1004be00000 failed: Cannot allocate memory). I contacted
my system administrator and mentioned that it may be an issue of compiling
cp2k using craype-hugepages. I'm not sure this is an issue many people will
have, but I wanted to document it.

Thanks!
Luis

On Wed, Mar 6, 2019 at 1:54 AM Matt W <mattwa... at gmail.com> wrote:

> I'd suggest starting from a converged GGA calculation. If you screen on
> that initial density you can get a big speed up (though rerun to check).
>
> EPS_PGF_ORB is very tight, you could increase it a bit. EPS_SCHWARZ is
> also quite strict, you could increase it to 10^-7 or even -6 and then check
> convergence when you have a good wavefunction.
>
> + Juergs advice.
>
> Matt
>
> On Tuesday, March 5, 2019 at 8:17:53 PM UTC, lar... at lbl.gov wrote:
>>
>> Hello,
>>
>> I am trying to run a simulation using the hybrid functional revPBE0-D3.
>> The simulation runs but doesn't output anything, including any error
>> messages (I have attached here the files for the simulation). I have been
>> able to simulate the same system using a meta-GGA functional with no
>> issues, so I am pretty confident there is nothing intrinsically wrong with
>> the system. I don't have a lot of experience using hybrid functionals,
>> however, so I am not sure what may be causing this issue.
>>
>> Thanks in advance!
>> Luis
>>
> --
> 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/zT-OSyja9gY/unsubscribe.
> To unsubscribe from this group and all its topics, send an email to
> cp2k+uns... at googlegroups.com.
> To post to this group, send email to cp... at googlegroups.com.
> Visit this group at https://groups.google.com/group/cp2k.
> For more options, visit https://groups.google.com/d/optout.
>


-- 
Luis A. Ruiz Pestana, PhD
Postdoctoral Fellow
University of California, Berkeley
Lawrence Berkeley National Laboratory
Berkeley, CA 94720
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20190307/307d57da/attachment.htm>


More information about the CP2K-user mailing list