MOF optimization

[K.AK]

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

[Anton Lytvynenko]

Dear K.AK,

I see here at least a few reasons which could impair your calculation.

1. Hybrid functional - do you really need it? Geometries even of 3d complexes are quite good with GGA, which is tremendously faster in periodic calculations. Wouldn't PBE (with dispersion correction) fit?

2. GAPW+6-31G? Any specific reason? I would take typical GPW together with MOLOPT bases.

3. Your cell size is >30x30x30 A quite loaded with atoms. Surely it will be slow and memory-consuming.

4. Convergence issues are quite common with metal complexes, some efforts could be necessary to overcome.

Yours,

Anton

--
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/ac837fd8-7833-4871-a8b1-acf6099ccee5n%40googlegroups.com.

[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...@googlegroups.com <cp...@googlegroups.com> on behalf of K.AK <kosh...@gmail.com>
Sent: Thursday, November 23, 2023 10:10 AM
To: cp2k
Subject: [CP2K:19549] MOF optimization

--
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/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>.

[K.AK]

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:

[Jürg Hutter]

Hi

Use GPW not GAPW for this setup.
Use DZVP-MOLOPT-SR-GTH for all elements (SR!)

regards
JH

________________________________________
From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of K.AK <kosh...@gmail.com>

Sent: Thursday, November 23, 2023 2:48 PM
To: cp2k
Subject: Re: [CP2K:19556] MOF optimization

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>.

[K.AK]

Dear cp2k users

Thank you, Jürg Hutter.

I had forgotten about conditions such as GPW. Thank you for pointing these out.

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 22:53:56 UTC+9 Jürg Hutter:

[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

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/9ba670b6-c5bc-4b0e-89c9-d6db28733753n%40googlegroups.com.

[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:

[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

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/277b24d4-dca6-4ce7-b40d-cc9c14fcfe36n%40googlegroups.com.

[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:

[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:

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/410e5bf6-f122-46ee-ad86-b719d316bb77n%40googlegroups.com.

[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:

[Gupta Dhananjay]

Dear K.AK,

If possible use any other software to change CIF to XYZ, sometimes VESTA gives wrong XYZ file (sometimes it happened with me while dealing with MOFs).

Thanks 

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/277b24d4-dca6-4ce7-b40d-cc9c14fcfe36n%40googlegroups.com.

[K.AK]

Dear cp2k experts

Thank you for your reply, Gupta Dhananjay.

I would try to change CIF to XYZ by mercury etc.software, and I simulated it.

Also, under previous conditions, SCF did not converge, and energy change was approximately -1.0E-07-1.0E06.

The input and output files (Default conditions & FULL_SINGLE_INVERSE )have been attached below.

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27日月曜日 2:04:58 UTC+9 Gupta Dhananjay:

[K.AK]

Dear cp2k experts

In addition, the cif  files have been attached below.

Please advice me what I should modify in the input file? 

Many thanks,
K.AK

2023年11月27日月曜日 9:13:32 UTC+9 K.AK:

[Anton Lytvynenko]

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

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/9434ce5d-32de-47fc-875b-5702d511042en%40googlegroups.com.

[K.AK]

cp2k users,

Thank you, Anton Lytvynenko

I would try to simulate the optimization under recommended condition.

Many thanks,

K.AK.

2023年11月28日火曜日 20:32:45 UTC+9 Anton Lytvynenko: