Convergence failure and incorrect solvation energy with SCCS in cp2k-2022.1

[Wiko Ann]

Hello, everyone:

I was trying to apply solvation effect to the Ag(111)-H3O+ system, while I met several problems.  

1. The first problem is the convergence of Ag(111) surface. I built a 4*4*3 Ag(111) slab. The system could hardly converge when implicit solvation model was used. While in gas phase, it converged successfully.  Are there some tricks to make it converge?

2. Another problem is more confusing. The solvation energy of H3O+ differs a lot from the experimental value as well as the values calculated by other solvation models such as VASPsol. I've tried to change the parameters in SCCS sections according to the article  [Self-consistent continuum solvation (SCCS): The case of charged systems. J. Chem. Phys. 139, 214110 (2013) ], and try to change the Poisson solver, but I still couldn't obtain a reasonable result. Why is there such a huge difference? Did I misuse SCCS section? 

I'm new to CP2K, and I didn't know whether I set somethings wrong in the input file or I missed some subtle but important details. I'll appreciate it if someone can give me some advice on these problems, since solvation is vital in predicting the energy of H3O+ and the interaction of metal surface and adsorbates. 

The input file is attached below.

Thanks a lot!

Wiko

[Wiko Ann]

[Krack Matthias (PSI)]

Hi Wiko

 

1. Slab: SCCS in CP2K is not implemented for k points and not yet tested with smearing. I guess that you are using one of these features for the Ag(111) slab which might cause problems.
2. H3O+: The results are sensible to the choice of the parameters RHO_MIN and RHO_MAX. Their values differ especially for neutral and ionic systems. What do you get for the electrostatic part of the solvation energy?

 

Just as minor improvements (do not help for the issue above), I suggest to use for H3O+ the default diagonalization instead of OT (i.e. &OT off) and a smaller EPS_DEFAULT value (at least 1.0E-12 instead of 1.0E-10).

 

Best

 

Matthias

--
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/0420c53c-f0bb-49dc-85fa-2124328053d8n%40googlegroups.com.

[Krack Matthias (PSI)]

PS: Do not use with SCCS “PSOLVER MT” and “PERIODIC NONE” but the defaults.

 

M.

 

From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of Krack Matthias (PSI) <matthia...@psi.ch>
Date: Monday, 7 November 2022 at 16:53
To: cp...@googlegroups.com <cp...@googlegroups.com>
Subject: Re: [CP2K:17989] Re: Convergence failure and incorrect solvation energy with SCCS in cp2k-2022.1

Hi Wiko

 

1.       Slab: SCCS in CP2K is not implemented for k points and not yet tested with smearing. I guess that you are using one of these features for the Ag(111) slab which might cause problems.

2.       H3O+: The results are sensible to the choice of the parameters RHO_MIN and RHO_MAX. Their values differ especially for neutral and ionic systems. What do you get for the electrostatic part of the solvation energy?

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/ZRAP278MB08274F9BF3607A244882B3C3F43C9%40ZRAP278MB0827.CHEP278.PROD.OUTLOOK.COM.

[Wiko Ann]

Dear Matthias: 

Thanks a lot for your reply. The electrostatic energy is posted below. 

1. I did use k points and smearing in the calculation of slab. It should be the problem. 

2. Did you mean if I want to obtain the solvation energy of H3O+, I can use “PSOLVER MT” and “PERIODIC NONE” to calculate cluster energy in gas phase, and then use the default setting when applying SCCS?

Regards

Wiko

[Wiko Ann]

And if I use the default setting in gas phase, Will the additional interaction between charged periodic boxes be included in the total energy?

[Krack Matthias (PSI)]

Hi Wiko

 

1. That’s indeed a problem.
2. That’s correct, SCCS is meant to work with POISSON_SOLVER periodic and PERIODIC xyz.

 

Best

 

Matthias

 

From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of Wiko Ann <toda...@gmail.com>
Date: Tuesday, 8 November 2022 at 03:44
To: cp2k <cp...@googlegroups.com>
Subject: Re: [CP2K:17992] Re: Convergence failure and incorrect solvation energy with SCCS in cp2k-2022.1

Dear Matthias: 

 

Thanks a lot for your reply. The electrostatic energy is posted below. 

1. I did use k points and smearing in the calculation of slab. It should be the problem. 

2. Did you mean if I want to obtain the solvation energy of H3O+, I can use “PSOLVER MT” and “PERIODIC NONE” to calculate cluster energy in gas phase, and then use the default setting when applying SCCS?

 

Regards

 

Wiko

 

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/ea085a09-02d3-425f-ac2c-cfcf2f1564e6n%40googlegroups.com.

[Anton Lytvynenko]

Dear Krack, dear all,

I'd like to ask another related question -- if SCCS is capable to model water without addition of some explicit water molecules in such situation?

Yours,

Anton

08.11.2022 10:17, Krack Matthias (PSI) пише:

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/ZRAP278MB0827413D164197DF78B9DC76F43F9%40ZRAP278MB0827.CHEP278.PROD.OUTLOOK.COM.

[Krack Matthias (PSI)]

The box is not charged with the defaults “POISSON_SOLVER periodic” and “PERIODIC xyz” due to the automatically added neutralizing background charge. The periodic box cannot be charged in that case as this would add an infinite interaction energy term. In any case, the box should be chosen large enough like for gas phase calculations without periodicity.

Charged systems like H3O+ need special care and additional correction terms are required. Currently, CP2K provides only an approximative correction (Makov and Payne) for such systems.

 

Matthias

 

From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of Wiko Ann <toda...@gmail.com>

Date: Tuesday, 8 November 2022 at 04:13
To: cp2k <cp...@googlegroups.com>

Subject: Re: [CP2K:17993] Re: Convergence failure and incorrect solvation energy with SCCS in cp2k-2022.1

And if I use the default setting in gas phase, Will the additional interaction between charged periodic boxes be included in the total energy?

 

在2022年11月8日星期二 UTC+8 10:44:23<Wiko Ann> 写道：

Dear Matthias: 

 

Thanks a lot for your reply. The electrostatic energy is posted below. 

1. I did use k points and smearing in the calculation of slab. It should be the problem. 

2. Did you mean if I want to obtain the solvation energy of H3O+, I can use “PSOLVER MT” and “PERIODIC NONE” to calculate cluster energy in gas phase, and then use the default setting when applying SCCS?

 

Regards

 

Wiko

 

.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/d4cca932-39ee-47ac-8a78-3dbf619d0588n%40googlegroups.com.

[Wiko Ann]

Dear Matthias: 

Many thanks for your patient replies and advice. They really help me a lot! 

Best

Wiko 

[Krack Matthias (PSI)]

Hi Anton

 

You are free to add explicit water molecules when using SCCS, but that’s not required, though it might be better to have some explicit water.

 

Best

 

Matthias

 

From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of Anton Lytvynenko <anton.s.l...@gmail.com>
Date: Tuesday, 8 November 2022 at 10:28
To: cp...@googlegroups.com <cp...@googlegroups.com>
Subject: Re: [CP2K:17995] Re: Convergence failure and incorrect solvation energy with SCCS in cp2k-2022.1

Dear Krack, dear all,

I'd like to ask another related question -- if SCCS is capable to model water without addition of some explicit water molecules in such situation?

Yours,

Anton

08.11.2022 10:17, Krack Matthias (PSI) пише:

Hi Wiko

 

1.       That’s indeed a problem.

2.       That’s correct, SCCS is meant to work with POISSON_SOLVER periodic and PERIODIC xyz.

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/37c2571e-99f6-0535-d8e7-e8d3813118c8%40gmail.com.

[Anton Lytvynenko]

Dear Matthias,

thank your very much for the explanation.

Yours,

Anton

09.11.2022 9:46, Krack Matthias (PSI) пише:

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/ZRAP278MB082703DA579071D939E44035F43E9%40ZRAP278MB0827.CHEP278.PROD.OUTLOOK.COM.

[Krack Matthias (PSI)]

Hi Wiko

 

After running checks using SCCS with POISSON_SOLVER MT and “PERIODIC none” for charged and uncharged molecules (solutes), I have correct my previous comment. CP2K/SCCS with MT seems to work fine which makes it the suggested setup especially for charged systems with SCCS. Note, that you need relatively high cutoff values (>800 Ry) and large cells for numerically accurate forces.

 

Sorry for confusing

 

Matthias

 

From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of Krack Matthias (PSI) <matthia...@psi.ch>
Date: Tuesday, 8 November 2022 at 10:17
To: cp...@googlegroups.com <cp...@googlegroups.com>
Subject: Re: [CP2K:17994] Re: Convergence failure and incorrect solvation energy with SCCS in cp2k-2022.1

Hi Wiko

 

1.       That’s indeed a problem.

2.       That’s correct, SCCS is meant to work with POISSON_SOLVER periodic and PERIODIC xyz.

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/ZRAP278MB0827413D164197DF78B9DC76F43F9%40ZRAP278MB0827.CHEP278.PROD.OUTLOOK.COM.

[Wiko Ann]

Dear  Matthias

Thank you for your comment! 

1. I've tried to set "POISSON_SOLVER MT" and "PERIODIC none", but the calculated solvation energy is different from the experimental value. For example, The solvation energy of H3O+ is -108kcal/mol in experiment. With "POISSON_SOLVER MT" and "PERIODIC none", it's -138.4 kcal/mol. But if I set "POISSON_SOLVER PERIODIC" and "PERIODIC XYZ", The value is -107.5kcal/mol. Another settings are all the same. The parameters RHO_MIN and RHO_MAX are set to 0.0002 and 0.0035 respectively, which are suitable for cations according to the article  [Self-consistent continuum solvation (SCCS): The case of charged systems. J. Chem. Phys. 139, 214110 (2013) ].  

The main discrepancy is from the energy in gas phase (the energies are posted below). But I still don't understand why there's such a big difference.

PERIODIC
  Total energy(without SCCS):                                               -17.54128600763402
  Total energy(with SCCS):                                               -17.71255174117677
MT
  Total energy(without SCCS):                                               -17.49054517094935

  Total energy(with SCCS):                                               -17.71112388917346

2. And another question arose. When I used SCCS to study a 6-atom Ag cluster with H3O+ along side, there were different distance between H3O+ and Ag cluster, also between O and H. I found the solvation energy of the system is positive. And the energies between adjacent structures changed rapidly (below). I suppose there are something settings still not appropriate, maybe RHO_MIN and RHO_MAX. I'd love to know how to decide RHO_MIN and RHO_MAX in such a system?  The input file is attached below. 

 -239.471083048784806
 -238.953979533671856
 -238.946448820809621
 -239.402925734069186
 -239.260639793435388
 -239.180932339281611
 -239.014590347465059
 -239.163178351021770
 -239.194699170551246
 -239.162140349715088

[Wiko Ann]

Dear  Matthias

Can Constrained DFT work well with SCCS in cp2k?

Wiko

[Krack Matthias (PSI)]

Hi

 

I get the following results with the current CP2K trunk version (epsilon(H2O) = 78.36, cutoff = 1200 Ry)

 

System   Method   alpha [mN/m]   gamma [mN/m]   beta [GPa]   rho(min)   rho(max)   G(Sol) [kcal/mol]

H2O      0d  MT    50.0            0.0        -0.350           0.0001     0.0050    -7.16

H2O      3d FFT    50.0            0.0        -0.350           0.0001     0.0050    -7.12

H3O+     0d  MT     5.0            0.0         0.125           0.0002     0.0035  -108.48

OH-      0d  MT     0.0            0.0         0.450           0.0024     0.0155  -112.14

 

using the SCCS parameters from the literature fitted for a plane wave code (Quantum ESPRESSO).

In my opinion, given all the differences between the two code packages CP2K and QE (basis set etc.), these results compare quite well with the experimental values for the free solvation energies

 

G(Sol, H2O)  =   -6.3 kcal/mol

G(sol, H3O+) = -108.7 kcal/mol

G(Sol, OH-)  = -106.3 kcal/mol

 

The agreement can be improved, especially for anions, most likely by tuning the SCCS parameters specifically for CP2K.

 

HTH

.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/4752fc67-7eae-4c54-aef0-c6439b0990b3n%40googlegroups.com.

[Krack Matthias (PSI)]

Hi Wiko

 

No clue, never tried combining these features.

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/1930068a-0761-4db8-8e79-e10e7c7599f9n%40googlegroups.com.

[Wiko Ann]

Dear Matthias

Really thanks for your comments!

Wiko