Huge difference in adsorption energies between CP2K and VASP with PBE

[Sharma SRK Chaitanya Yamijala]

Dear CP2K members,

I am calculating the adsorption energies of CO on ZnO (4x4-slab with 4 layers of ZnO(10-10) surface). The adsorption energy obtained from VASP is ~ 0.3 eV (without dispersion) and it is ~ 0.45 eV (with dispersion). However, when calculating the same system with CP2K (geometry is re-optimized), I am obtaining the energies 0.688 eV and 0.863 eV for without and with dispersion correction, respectively. 

The literature value is close to the VASP result. Here is the reference (kindly, see Table 2, C-down, 2x1 cell, Full).

B Meyer and D Marx 2003 J. Phys.: Condens. Matter 15 L89

http://iopscience.iop.org/article/10.1088/0953-8984/15/2/112/pdf

Could anyone give an advice on why there is such a discrepancy in the adsorption energies irrespective of using the same functional?

I am attaching the geometry and the input file for your reference.

Thanks,

Sharma.

--------------------------------------------------------------------
SRK Chaitanya Sharma, Yamijala.

http://www.bmwong-group.com/group-members/

[hut...@chem.uzh.ch]

Hi

my first guess: you didn't calculate the BSSE correction.

other possible contributions: size of basis sets
k-point sampling

regards

Juerg Hutter
--------------------------------------------------------------
Juerg Hutter Phone : ++41 44 635 4491
Institut für Chemie C FAX : ++41 44 635 6838
Universität Zürich E-mail: hut...@chem.uzh.ch
Winterthurerstrasse 190
CH-8057 Zürich, Switzerland
---------------------------------------------------------------

-----cp...@googlegroups.com wrote: -----
To: cp...@googlegroups.com
From: "Sharma SRK Chaitanya Yamijala"
Sent by: cp...@googlegroups.com
Date: 07/12/2018 01:48AM
Subject: [CP2K:10507] Huge difference in adsorption energies between CP2K and VASP with PBE

--
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 post to this group, send email to cp...@googlegroups.com.
Visit this group at https://groups.google.com/group/cp2k.
For more options, visit https://groups.google.com/d/optout.


[attachment "1CO_on_opt_4x4_slab.xyz" removed by Jürg Hutter/at/UZH]
[attachment "zno_bulk.inp" removed by Jürg Hutter/at/UZH]

[Sharma SRK Chaitanya Yamijala]

Dear Prof. Hutter,

Thanks. I will try the BSSE correction with CP2K. Regarding k-point sampling, in both the softwares, I have used Gamma point only. Could you please elobarate your point on "size of basis sets" here? I don't know how should I compare VASP basis (plane waves) with CP2K (both PW and Gaussians)? I have converged the CUTOFFs in both the cases such that the energy differences are < 1 meV.

Also, yesterday, I have performed similar calculations with FHIAims and I have obtained the adsorption energy (withOUT dispersion and BSSE corrections) as 0.412 eV which is again closer to the VASP result and to the result in the reference.

Seeking your advice.

Thanks,

Sincerely,

Sharma.

[hut...@chem.uzh.ch]

Hi

you are using a DZVP basis, this is rather small. Check the number
of basis functions. Do the same for the FHIaims calculation
and compare.
The cutoff of VASP cannot be compared to a local basis set.

regards

Juerg Hutter
--------------------------------------------------------------
Juerg Hutter Phone : ++41 44 635 4491
Institut für Chemie C FAX : ++41 44 635 6838
Universität Zürich E-mail: hut...@chem.uzh.ch
Winterthurerstrasse 190
CH-8057 Zürich, Switzerland
---------------------------------------------------------------

-----cp...@googlegroups.com wrote: -----

To: "cp2k" <cp...@googlegroups.com>

From: "Sharma SRK Chaitanya Yamijala"
Sent by: cp...@googlegroups.com

Date: 07/12/2018 07:55PM
Subject: Re: [CP2K:10514] Huge difference in adsorption energies between CP2K and VASP with PBE

I am calculating the adsorption energies of CO on ZnO (4x4-slab with 4 layers of ZnO(10-10) surface). The adsorption energy obtained from VASP is ~ 0.3 eV (without dispersion) and it is ~ 0.45 eV (with dispSECTION_PARAMETERS TRUEersion). However, when calculating the same system with CP2K (geometry is re-optimized), I am obtaining the energies 0.688 eV and 0.863 eV for without and with dispersion correction, respectively.

[Sharma SRK Chaitanya Yamijala]

Dear Prof. Hutter,

I have performed the BSSE corrections and the results are given below. Clearly, the BSSE corrected interaction energy is huge (~ 1 eV). If I consider this CP-corrected energy as the adsorption energy (I have calculated adsorption energy as Eads = EAB - EA -EB), then it will be too off from the VASP (-0.466 eV) or FHIAIMs (-0.595 eV) results. As mentioned, without BSSE the calculated adsorption energy is -0.863 eV. All the energies given are with Grimme's-D3 dispersion. Kindly, let me know if you have any comments about this.

To check the accuracy of my basis set, I have started a run with TZVP basis (instead of DZVP) for adsorption energy calculation and I will update here.

On a different note, the number of basis functions used for the calculation in FHIAIMS (6172) and CP2K (5660) is roughly the same.

 -------------------------------------------------------------------------------
 -                                                                             -
 -                                 BSSE RESULTS                                -
 -                                                                             -
 -                 CP-corrected Total energy:        0.039208                  -
 -                                                                             -
 -                       1-body contribution:    -9804.468056                  -
 -                       1-body contribution:      -21.667247                  -
 -                                                                             -
 -                       2-body contribution:     9826.174511                  -
 -                 BSSE-free interaction energy:     9826.174511               -
 -------------------------------------------------------------------------------

Thanks,

Sharma.

--------------------------------------------------------------------
SRK Chaitanya Sharma, Yamijala.

http://www.bmwong-group.com/group-members/

 To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+unsubscribe@googlegroups.com.

 To post to this group, send email to cp...@googlegroups.com.
 Visit this group at https://groups.google.com/group/cp2k.
 For more options, visit https://groups.google.com/d/optout.

--
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/tjXJMpIfVx8/unsubscribe.
To unsubscribe from this group and all its topics, send an email to cp2k+unsubscribe@googlegroups.com.

[Sharma SRK Chaitanya Yamijala]

And I believe the increase in the adsorption energy could be because of the non-ground-state geometry of the "slab" during the BSSE calculation. The slab's absolute energy when considered in the monomer basis (for the ground state geometry) is -9804.47553 au and it is -9804.468056 au for the geometry in the dimer configuration (i.e. for the slab geometry in the "slab + CO" optimized geometry).

--------------------------------------------------------------------
SRK Chaitanya Sharma, Yamijala.

http://www.bmwong-group.com/group-members/

[Sharma SRK Chaitanya Yamijala]

I just want to add that, the energy of the CO obtained through the one body contributions (listed in the output of a BSSE calculation) and the energy of the optimized CO (in the same cell size but without the slab) is found to be exactly the same. To me, this suggests the current basis should be accurate enough as one should expect a larger effect of the BSSE on CO rather than on slab (as slab contributes to the large portion of basis functions of the total system). Kindly, let me know if my logic is wrong.

--------------------------------------------------------------------
SRK Chaitanya Sharma, Yamijala.

http://www.bmwong-group.com/group-members/

[hut...@chem.uzh.ch]

Hi

assuming you did all calculations correctly, you should have the following 7 energies ready

System Geometry from Basis

1 A+B A+B A+B
2 A A A
3 B B B
4 A A+B A
5 A A+B A+B
6 B A+B B
7 B A+B A+B

The uncorrected adsorption energy is : E1 = (1) - (2+3)

Counterpoise correction energy is : E2 = ((4)-(5)) +((6)-(7))

BSSE corrected adsorption energy is : Ea = E1 + E2


E1 is too negative because of BSSE.
Counterpoise correction is positive because larger basis stabilises systems.
Typically the counterpoise correction is 0-50% of the non-corrected energy
(depending on the system and the basis)

regards

Juerg

--------------------------------------------------------------
Juerg Hutter Phone : ++41 44 635 4491
Institut für Chemie C FAX : ++41 44 635 6838
Universität Zürich E-mail: hut...@chem.uzh.ch
Winterthurerstrasse 190
CH-8057 Zürich, Switzerland
---------------------------------------------------------------

-----cp...@googlegroups.com wrote: -----
To: cp...@googlegroups.com
From: "Sharma SRK Chaitanya Yamijala"
Sent by: cp...@googlegroups.com

Date: 07/14/2018 01:59AM
Subject: Re: [CP2K:10528] Huge difference in adsorption energies between CP2K and VASP with PBE

--
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/tjXJMpIfVx8/unsubscribe.

To unsubscribe from this group and all its topics, send an email to cp2k+uns...@googlegroups.com.

[Sharma SRK Chaitanya Yamijala]

Dear Prof. Hutter,

Thank you very much for your help. Following the last equations which you wrote for energy, I have E1(uncorrected adsorption energy) = -0.863 eV and E2 (counterpoise correction) = +1.0669. So, from the way the BSSE corrected adsorption energy is defined, I am getting +0.204 eV which suggests that the adsorption is thermodynamically not feasible. I am getting negative values with VASP and FHIAIMS.

However, I would like to point that, when I consider the results of a BSSE calculation (majorly the one-body contribution results) and calculate the counterpoise correction (E2 = ((4)-(5)) +((6)-(7))), then I obtain zero correction to the adsorption energy (because the energy of 4 = 5 and 6 = 7). Even following eq13 of this document by Prof. Sherrill, I am obtaining basically zero correction. Thus, I believe the printed value of counterpoise correction in the CP2K output is (1 - 5 - 7) and not ((4)-(5)) +((6)-(7))).

My runs with TZVP calculation are not finished yet and I will try to update the results when they are finished. If you get time, could you please suggest anything else which I may need to change in the input or anything you feel suspicious in the input?

Here are the energies for all the terms which you have mentioned in the earlier email.

System      Geometry from      Basis     Energy (eV)

1  A+B          A+B                    A+B      -267383.806
2      A               A                        A           -589.596
3      B               B                        B      -266793.347
4      A           A+B                        A           -589.596 (Take the A geometry from A+B and do a single point calculation)
5      A           A+B                   A+B            -589.596 (From one body contribution result of BSSE calculation)
6      B           A+B                       B       -266793.143 (Take the B geometry from A+B and do a single point calculation)
7      B           A+B                   A+B       -266793.143 (From one body contribution result of BSSE calculation)

Thanks,

Sharma.

--------------------------------------------------------------------
SRK Chaitanya Sharma, Yamijala.

http://www.bmwong-group.com/group-members/

  To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+unsubscribe@googlegroups.com.

 To post to this group, send email to cp...@googlegroups.com.
  Visit this group at https://groups.google.com/group/cp2k.
  For more options, visit https://groups.google.com/d/optout.


 --
 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/tjXJMpIfVx8/unsubscribe.

 To unsubscribe from this group and all its topics, send an email to cp2k+unsubscribe@googlegroups.com.

 To post to this group, send email to cp...@googlegroups.com.
 Visit this group at https://groups.google.com/group/cp2k.
 For more options, visit https://groups.google.com/d/optout.




  --
 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@googlegroups.com.

 To post to this group, send email to cp...@googlegroups.com.
 Visit this group at https://groups.google.com/group/cp2k.
 For more options, visit https://groups.google.com/d/optout.


--
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/tjXJMpIfVx8/unsubscribe.

To unsubscribe from this group and all its topics, send an email to cp2k+unsubscribe@googlegroups.com.

[Sharma SRK Chaitanya Yamijala]

Dear Prof. Hutter,

Here are the results with TZVP, TZV2P basis. For Zn, I got the basis from BASIS_MOLOPT_UCL and for all the other atoms, from BASIS_MOLOPT.

Clearly, the results with larger basis seem better (though not yet matching with the literature reported values). Could you please suggest anything else which I could change to improve these adsorption energies? I am trying to get the correct binding energies because I would like to use CP2K to study the desorption dynamics of these adsorbates (using MD runs). If the binding energies are not reasonable (without BSSE), then the results with the MD will be less meaningful (Here, I am supposing that there is no way to include BSSE while running MD).

Seeking your advice,

Sincerely,

Sharma.

in 4x4 cell	with dispersion
1CO (DZVP)	-589.596
1CO (TZVP)	-589.879
1CO (TZV2P)	-589.951
1CO2 (DZVP)	-1027.520
1CO2 (TZVP)	-1027.697
1CO2 (TZV2P)	-1027.710

(opt param: A=3.28, C=5.293)	with dispersion
4x4 + DZVP	-266793.347
4x4 + TZVP	-266798.943
4x4 + TZV2P	-266799.744

Basis Set effect	Total energy (eV)	Adsorption energy (eV)
4x4+1CO (DZVP)	-267383.806	-0.863
4x4+1CO (TZVP)	-267389.552	-0.730
4x4+1CO (TZV2P)	-267390.371	-0.676
4x4+1CO2 (DZVP)	-267821.300	-0.434
4x4+1CO2 (TZVP)	-267827.040	-0.399
4x4+1CO2 (TZV2P)	-267827.857	-0.404

--------------------------------------------------------------------
SRK Chaitanya Sharma, Yamijala.

http://www.bmwong-group.com/group-members/

[Gabriele]

Hello,

I had a look at the  paper you mentioned http://iopscience.iop.org/article/10.1088/0953-8984/15/2/112/pdf. It seems that the number you gave is 0.32 eV. This is for a 2x1 structure, which is 1/2ML coverage as quoted in the table. 1ML coverage is defined by 1 CO molecule adsorbed per surface Zn atom. However, the coverage in your input structure is much smaller, i.e. 1/16ML, as there is only one CO molecule per 16 ZnO surface unit cells. Is the input you posted the same input you used to get the values of the adsorption energies reported above? Did you indeed adsorb 8 CO molecules in a 2x1 periodicity?

On a probably less important note, the paper you quoted has also applied a dipole correction. Although CO has a small dipole, this might play a role, so you might want to consider removing the periodicity along z or  check the results running also a larger vacuum region.

Cheers,

Gabriele

 --
 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/tjXJMpIfVx8/unsubscribe.

 To unsubscribe from this group and all its topics, send an email to cp2k+uns...@googlegroups.com.

 To post to this group, send email to cp...@googlegroups.com.
 Visit this group at https://groups.google.com/group/cp2k.
 For more options, visit https://groups.google.com/d/optout.




  --
 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 post to this group, send email to cp...@googlegroups.com.
 Visit this group at https://groups.google.com/group/cp2k.
 For more options, visit https://groups.google.com/d/optout.

--
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/tjXJMpIfVx8/unsubscribe.

To unsubscribe from this group and all its topics, send an email to cp2k+uns...@googlegroups.com.

[Sharma SRK Chaitanya Yamijala]

Hi Gabriele,

Thanks for your suggestions. I too have thought about the vacuum space and I have already submitted some jobs with higher vacuum (Till now, I have used 15 Ang vacuum. Now, I have submitted with 20 Ang vacuum space along with TZV2P basis) and I will try to update them when they complete. Can you please suggest which Poisson solvers are best for the XY periodicity and may I know whether there is any possibility to include such dipole-correction schemes with CP2K?

Regarding 2x1 cell, yes, I have indeed started with this cell but I couldn't optimize this cell using CP2K (even with 10x10x1 k-mesh) and 1000 Ry cutoff. It always goes to very high energies right from the first geometry cycle (I have started from an already optimized geometry. Also, the same geometry, when extended to form 4x4 or 6x3 runs fine with CP2K. So, I believe the problem of convergence could be because of the k-point related issues for the non-cubic cells in CP2K). So, I switched to a 4x4 cell, directly (which I could optimize).

Regarding coverages, yes, I have performed many coverage dependent studies on a 6x3 cell (with DZVP basis) and I found the energy per adsorbate to be almost the same for all of them (-0.83 eV for 9 COs to -0.86 eV for 1 CO). A 6x3 cell with 9 COs will be equivalent to a 2x1 cell with 1 CO (we confirmed this by calculating the energy per adsorbate values for 2x1 (-0.265 eV) and 6x3 (-0.267 eV) cells using VASP. With VASP, I could converge both 2x1 cell and 6x3 cells).

Thanks,

Sharma.

--------------------------------------------------------------------
SRK Chaitanya Sharma, Yamijala.

http://www.bmwong-group.com/group-members/

To unsubscribe from this group and all its topics, send an email to cp2k+unsubscribe@googlegroups.com.

[Gabriele]

Hi,

You can use MT with a vacuum region larger than the slab thickness and a centered cell. You can also use an orthorhombic cell for this system. It should not make up for 0.4 eV of difference though.

Previous adsorption studies on ZnO(10-10) using DZVP-MOLOPT did not show more than a few meV difference in adsorption energies compared to plane wave codes.

Can you please upload the inputs, output and .restart file only for the calculation at 0.5ML coverage on the 6x3 cell, that for a clean slab and a CO molecule in vacuum?

Cheers,

Gabriele

To unsubscribe from this group and all its topics, send an email to cp2k+uns...@googlegroups.com.

[Sharma SRK Chaitanya Yamijala]

Hi Gabriele,

Thanks a lot for trying to help me. Can I send all these files to you off the list?

Also, can you send any references of the papers which published the adsorption energies with ZnO(10-10) using CP2K. I really, couldn't find anything for CO on ZnO from the google scholar search!

Are you by any chance the first author of this paper? "Solvent-Induced Proton Hopping at a Water–Oxide Interface"

Thanks,

Sharma.

--------------------------------------------------------------------
SRK Chaitanya Sharma, Yamijala.

http://www.bmwong-group.com/group-members/

To unsubscribe from this group and all its topics, send an email to cp2k+unsubscribe@googlegroups.com.

[Sharma SRK Chaitanya Yamijala]

Hi Gabriele,

I have sent all the files, you have asked for, to your email.

I just want to add that, after your last email, I have checked my other calculations with CO2 as an adsorbate and I indeed found that such a huge discrepancy (0.4 eV) between plane wave codes and CP2K is observed only for CO (a molecule with dipole moment) and not with CO2 molecule (zero dipole moment).

For CO2 on a 4x4 cell, results of CP2K (-0.434 and -0.2 eV with and without dispersion, respectively) and VASP (-0.383 and -0.152 eV with and without dispersion, respectively) are quite close. The difference of 0.05 eV (when using DZVP basis set in CP2K) is further reduced to 0.02 when using TZV2P basis.

So, I believe, the presence of an inherent dipole moment for CO could be the main reason for such a huge difference. However, the question of how to achieve consistency between the packages for CO adsorption still remains the same!

Thanks for your time,

Sharma.

--------------------------------------------------------------------
SRK Chaitanya Sharma, Yamijala.

http://www.bmwong-group.com/group-members/

[Sharma SRK Chaitanya Yamijala]

For 4x4 + 1CO system, the absolute energy differs by just 2 meV between the 20 Ang and 15 Ang space calculations. So, I think even the vacuum is not the problem!

--------------------------------------------------------------------
SRK Chaitanya Sharma, Yamijala.

http://www.bmwong-group.com/group-members/