[CP2K-user] [CP2K:19582] MOF optimization

Anton Lytvynenko anton.s.lytvynenko at gmail.com
Tue Nov 28 11:32:38 UTC 2023


Dear K.AK,

my recommendation would be to loose the scf convergence tolerance 
(eps_scf) to 1e-6 instead of 1e-7 (2x times in the input file).

Your output demostrate quite good convergence behavior, actually, and I 
wonder if you can really benefit from tight scf tolerance.

If you really believe you need so tight convergence, try first to 
optimize the geometry with 1e-6 and then re-run with 1e-7, but starting 
from the pre-optimized geometry rather than initial one (poor geometry 
impairs scf convergence).

Yours,

Anton


Am 26.11.2023 um 17:42 schrieb K.AK:
> Dear cp2k experts
>
> Thank you for your reply, Anton Lytvynenko.
>
> The input and output files (Default conditions & FULL_SINGLE_INVERSE 
> )have been attached below.
>
>
> Many thanks,
> K.AK
> 2023年11月27日月曜日 1:23:10 UTC+9 Anton Lytvynenko:
>
>     Dear K.AK,
>
>     Please show us the output of these two efforts.
>
>     Anton
>
>     Am 26.11.23 um 07:31 schrieb K.AK:
>>     Dear cp2k experts
>>
>>     Thank you for your help, Anton Lytvynenko.
>>
>>     Now, I simulated MOF cell under the default conditions.
>>
>>     But, SCF did not converged, and energy change was approximately
>>     -1.0E-07-1.0E06.
>>
>>     Also, under the suggested conditions (PRECONDITIONER
>>     FULL_SINGLE_INVERSE ; MAX_SCF 150 ; LINESEARCH GOLD), SCF did not
>>     converged, and energy change was approximately -1.0E06.
>>
>>     Could you advice me what I should modify in the input file?
>>     Should I change the cell structure and how do I modify it?
>>
>>
>>     Many thanks,
>>     K.AK
>>
>>     2023年11月25日土曜日 4:03:28 UTC+9 Anton Lytvynenko:
>>
>>         Dear K.AK,
>>
>>         your SCF convergence is unstable from the very beginning. I
>>         would question the choice of FULL_ALL preconditioner and DIIS
>>         minimizer.
>>
>>         I wonder if DIIS is about to behave well in complicated
>>         systems. FULL_ALL may be sensitive to ENERGY_GAP parameter,
>>         AFAIR.
>>
>>         Have you tried to allow the default OT settings?
>>
>>         Plainly, something like that:
>>
>>>            &SCF
>>>               EPS_SCF 1.0E-07
>>>               &OT
>>>               &END OT
>>>               &OUTER_SCF
>>>                  EPS_SCF 1.0E-07
>>>               &END OUTER_SCF
>>>             &END SCF
>>         If this does not work, I would first consider change to
>>         PRECONDITIONER FULL_SINGLE_INVERSE ; MAX_SCF 150 ; LINESEARCH
>>         GOLD.
>>
>>         Regards,
>>
>>         Anton
>>
>>         Am 24.11.2023 um 16:40 schrieb K.AK:
>>>         Dear cp2k experts
>>>
>>>         Thank you for your advice, Anton Lytvynenko.
>>>
>>>         In this case, CIF file which I used was converted from real
>>>         CIF to XYZ coordinations by VESTA.
>>>
>>>         UiO-66 was a miswritten, I would like to optimize different
>>>         MOF structure.
>>>
>>>         Also, I have simulated MOF optimization under suggested
>>>         conditions.
>>>
>>>         But, the SCF did not converged for 400 cycles.
>>>
>>>         Could you advice me what I should modify in the input file?
>>>
>>>         The input and output files have been attached below.
>>>
>>>         Many thanks,
>>>         K.AK
>>>
>>>         2023年11月24日金曜日 0:27:52 UTC+9 Anton Lytvynenko:
>>>
>>>             Dear K.AK,
>>>
>>>             Am 23.11.2023 um 15:04 schrieb K.AK:
>>>>             Also, this is a general question, but what should I do
>>>>             if the calculation does not converge in the periodic
>>>>             boundary calculation?
>>>
>>>             I wonder if this question can be answered shortly and
>>>             via definitive set of instructions, it worth a long
>>>             manual addressing various sources of the solution
>>>             instability.
>>>
>>>             In your case, you should first start with OT optimizer,
>>>             it is quite robust algorithm and should converge
>>>             reasonably for your system.
>>>
>>>             If you see any problems, try to check your geometry
>>>             first - it can be tricky to prepare input files from
>>>             real CIF data for a number of reasons. If there are
>>>             still problems, please show us the problematic output.
>>>
>>>             By the way, UiO-66 is a Zr MOF, isn't it? But I can't
>>>             see Zr atoms in your input. Is it correct?
>>>
>>>             Yours,
>>>
>>>             Anton
>>>
>>>
>>>>             2023年11月23日木曜日 22:53:56 UTC+9 Jürg Hutter:
>>>>
>>>>                 Hi
>>>>
>>>>                 Use GPW not GAPW for this setup.
>>>>                 Use DZVP-MOLOPT-SR-GTH for all elements (SR!)
>>>>
>>>>                 regards
>>>>                 JH
>>>>
>>>>                 ________________________________________
>>>>                 From: cp... at googlegroups.com
>>>>                 <cp... at googlegroups.com> on behalf of K.AK
>>>>                 <kosh... at gmail.com>
>>>>                 Sent: Thursday, November 23, 2023 2:48 PM
>>>>                 To: cp2k
>>>>                 Subject: Re: [CP2K:19556] MOF optimization
>>>>
>>>>                 Thank you for your helps, Anton Lytvynenko and Jürg
>>>>                 Hutter.
>>>>
>>>>
>>>>                 Following your advices, I have reviewed the
>>>>                 calculation conditions. I have created a new input
>>>>                 file, and Could you please verify this?
>>>>
>>>>                 I would proceed with the calculation under these
>>>>                 conditions. Now, the calculation time is
>>>>                 approximately 120 s per step.
>>>>
>>>>                 Also, this is a general question, but what should I
>>>>                 do if the calculation does not converge in the
>>>>                 periodic boundary calculation?
>>>>
>>>>
>>>>                 Many thanks,
>>>>
>>>>                 K.AK
>>>>
>>>>                 2023年11月23日木曜日 22:07:19 UTC+9 Jürg Hutter:
>>>>                 Hi
>>>>
>>>>                 a few general remarks:
>>>>
>>>>                 For the use of hybrid functionals in PBC, you
>>>>                 should use an interaction
>>>>                 potential with a cutoff (6 Angstrom should be ok).
>>>>
>>>>                 Make sure that you have enough memory to keep the
>>>>                 integrals in core
>>>>                 (print HF_INFO, set MEMORY).
>>>>
>>>>                 EPS_SCHWARZ can most likely be set to 1.E-08.
>>>>
>>>>                 Restart from a GGA wavefunction and set
>>>>                 SCREEN_ON_INITIAL_P true.
>>>>
>>>>                 EPS_DEFAULT 1.E-10 is most likely good enough.
>>>>
>>>>                 Don't set EPS_FIT and EPS_PGF_ORB, defaults will be
>>>>                 good enough.
>>>>                 Ignore the HFX warning about occupation or follow
>>>>                 advice in comment.
>>>>
>>>>                 REL_CUTOFF should be 60
>>>>
>>>>                 GAPW with 6-31G basis might not be a good choice.
>>>>                 Consider GPW/MOLOPT basis sets. In the latter case
>>>>                 use ADMM!
>>>>
>>>>                 For MOFs I would advice to use vdW corrections.
>>>>
>>>>                 I don't know what the reason behind the fixed K
>>>>                 atom is. But this will cause
>>>>                 convergence problems in the geometry optimization.
>>>>                 It might also
>>>>                 interfere with CENTER_COORDINATES (why use it for a
>>>>                 periodic system?).
>>>>
>>>>                 regards
>>>>                 JH
>>>>
>>>>                 ________________________________________
>>>>                 From: cp... at googlegroups.com
>>>>                 <cp... at googlegroups.com> on behalf of K.AK
>>>>                 <kosh... at gmail.com>
>>>>                 Sent: Thursday, November 23, 2023 10:10 AM
>>>>                 To: cp2k
>>>>                 Subject: [CP2K:19549] MOF optimization
>>>>
>>>>                 Dear experts,
>>>>
>>>>                 I try to optimize MOF structure by performing
>>>>                 periodic DFT calculations.
>>>>
>>>>                 I have performed the calculations under the
>>>>                 periodic conditions of the attached input file.
>>>>
>>>>                 But, the computation time is very long and the
>>>>                 convergence is not good.
>>>>
>>>>                 Please advise about the cp2k simulation method and
>>>>                 input file.
>>>>
>>>>                 The input files have been attached below.
>>>>
>>>>
>>>>                 Many thanks,
>>>>                 K.AK
>>>>
>>>>                 -- 
>>>>                 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... at googlegroups.com<mailto:cp2k+uns... at googlegroups.com>.
>>>>
>>>>                 To view this discussion on the web visit
>>>>                 https://groups.google.com/d/msgid/cp2k/ac837fd8-7833-4871-a8b1-acf6099ccee5n%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/ac837fd8-7833-4871-a8b1-acf6099ccee5n%40googlegroups.com?utm_medium=email&utm_source=footer
>>>>                 <https://groups.google.com/d/msgid/cp2k/ac837fd8-7833-4871-a8b1-acf6099ccee5n%40googlegroups.com?utm_medium=email&utm_source=footer>>.
>>>>
>>>>
>>>>                 -- 
>>>>                 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... at googlegroups.com<mailto:cp2k+uns... at googlegroups.com>.
>>>>
>>>>                 To view this discussion on the web visit
>>>>                 https://groups.google.com/d/msgid/cp2k/0c14f2c9-3fe3-441b-853c-ef037665fdbcn%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/0c14f2c9-3fe3-441b-853c-ef037665fdbcn%40googlegroups.com?utm_medium=email&utm_source=footer
>>>>                 <https://groups.google.com/d/msgid/cp2k/0c14f2c9-3fe3-441b-853c-ef037665fdbcn%40googlegroups.com?utm_medium=email&utm_source=footer>>.
>>>>
>>>>
>>>>             -- 
>>>>             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... at googlegroups.com.
>>>>             To view this discussion on the web visit
>>>>             https://groups.google.com/d/msgid/cp2k/9ba670b6-c5bc-4b0e-89c9-d6db28733753n%40googlegroups.com
>>>>             <https://groups.google.com/d/msgid/cp2k/9ba670b6-c5bc-4b0e-89c9-d6db28733753n%40googlegroups.com?utm_medium=email&utm_source=footer>.
>>>
>>>         -- 
>>>         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... at googlegroups.com.
>>>         To view this discussion on the web visit
>>>         https://groups.google.com/d/msgid/cp2k/277b24d4-dca6-4ce7-b40d-cc9c14fcfe36n%40googlegroups.com
>>>         <https://groups.google.com/d/msgid/cp2k/277b24d4-dca6-4ce7-b40d-cc9c14fcfe36n%40googlegroups.com?utm_medium=email&utm_source=footer>.
>>
>>     -- 
>>     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... at googlegroups.com.
>>     To view this discussion on the web visit
>>     https://groups.google.com/d/msgid/cp2k/410e5bf6-f122-46ee-ad86-b719d316bb77n%40googlegroups.com
>>     <https://groups.google.com/d/msgid/cp2k/410e5bf6-f122-46ee-ad86-b719d316bb77n%40googlegroups.com?utm_medium=email&utm_source=footer>.
>
> -- 
> 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+unsubscribe at googlegroups.com.
> To view this discussion on the web visit 
> https://groups.google.com/d/msgid/cp2k/9434ce5d-32de-47fc-875b-5702d511042en%40googlegroups.com 
> <https://groups.google.com/d/msgid/cp2k/9434ce5d-32de-47fc-875b-5702d511042en%40googlegroups.com?utm_medium=email&utm_source=footer>.

-- 
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+unsubscribe at googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/43fe5c7b-c451-49c1-ad64-d649fa79e473%40gmail.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20231128/788dde71/attachment-0001.htm>


More information about the CP2K-user mailing list