[CP2K:9891] Very short metal-H distances from IC-QM/MM

[Dorothea Golze]

Dear Maxime,

>>So I'll conclude that this behaviour is a shortcoming of
>>the Siepmann-Sprik potential (at least in the current,
>>canonical parametrization). As I'm not really comfortable
>>with such a discrepancy between the QM/MM and full-QM
>>results, I think I will be trying to amend the potential.

In the Siepmann-Sprik potential there is a term that depends on the
orientation of the dipole vector of the water molecule; Equation (1) and
Equation(3) in  https://doi.org/10.1063/1.469429 The purpose of this term
is to direct the hydrogens away from the surface.
If you want to change the behavior of the potential, then you have to
modify this term. But as I said, during MD runs it is a rare event that the
hydrogen point toward the surface.

>>Could you further clarify why the cell size would be too
>>small for IC-QM/MM calculations? Do you mean too small
>>to be relevant to real interfaces? Then I agree, but this is
>>just a test, and I would expect these short Pt-H distances
>>to also occur in larger cells (as you found yourself) --
>>if I replicate the structure in the X and Y directions, for
>>example, I assume I would be getting the same result.
>>Or do you mean something else?

Well, if you just want to see if the code runs through it's fine. For
anything else (e.g. testing of bonding distances) I would not trust the
results. But yes, these short distances can occur  for "H-down" water
molecules. However, these "H-down" structures have been also observed by
full DFT calculations and experiments, see
https://journals.aps.org/prl/pdf/10.1103/PhysRevLett.89.276102.

Best regards,
Dorothea

2018-01-19 12:43 GMT+02:00 Maxime Van den Bossche <
maxime.cp.v... at gmail.com>:

> Dear Dorothea,
>
> Thanks a lot for attending this issue (and for your efforts in
> developing and implementing this method)! I hadn't noticed
> that the figure in your paper was also showing the presence
> of such short Pt-H distances.
>
> So I'll conclude that this behaviour is a shortcoming of
> the Siepmann-Sprik potential (at least in the current,
> canonical parametrization). As I'm not really comfortable
> with such a discrepancy between the QM/MM and full-QM
> results, I think I will be trying to amend the potential.
>
> Could you further clarify why the cell size would be too
> small for IC-QM/MM calculations? Do you mean too small
> to be relevant to real interfaces? Then I agree, but this is
> just a test, and I would expect these short Pt-H distances
> to also occur in larger cells (as you found yourself) --
> if I replicate the structure in the X and Y directions, for
> example, I assume I would be getting the same result.
> Or do you mean something else?
>
> Best,
> Maxime
>
>
>
> On Thursday, January 18, 2018 at 7:13:52 PM UTC, Dorothea Golze wrote:
>
>> Hi Maxime,
>>
>> there is most likely nothing wrong with the force field implementation
>> etc. I ran quite a few simulations with the Siepmann-Sprik forcefield +
>> image charge setup and did not encounter any problems so far.
>> Some explanations:
>> The Pt-O distance is approximately 2.4 Angstrom with the Siepmann-Sprik
>> force field (see https://doi.org/10.1063/1.469429 Tab. 1). In most
>> cases, the O-H bonds are roughly parallel to the surface or point away from
>> it. However, in some cases the H atoms point towards the surface, and in
>> this  case the Pt-H bond is only 1.5 Angstrom. This is demonstrated in
>> Figure 11 (d) in https://doi.org/10.1021/ct400698y .The onset of the red
>> curve is already at 1.5 Angstrom (but the peak is at 2.4 Angstrom, so most
>> Pt-H distances are 2.4 Angstrom).
>> Another important point, your "super-cell" is only 3x4 gold atoms. That's
>> way too small. Might be that you also see some more funny effects due to
>> this.
>>
>> Your points above
>>
>> >a) ... something is wrong in my input? I also got the
>>        same results with narrower Gaussians for the image
>>        charge distributions (5 Å^-2 instead of 3.5 Å^-2).
>>
>> No, it is also correct that you see the same results with different IC
>> Gaussian widths (that's good) . Keep the default.
>>
>> >b) ... this is related to the Siepmann-Sprik potential
>>        parameters? I couldn't find previous works, though,
>>        (such as this forum, the original paper, or the
>>        implementation paper) reporting such behaviour.
>>        But if this is the cause, I could just add a short-
>>        ranged repulsive potential to the Pt-H interaction
>>        or maybe modify the Phi function in Eq 3 of the
>>        original paper. Could it be that one really needs
>>        to reparametrize the potential because of the
>>        differences between DFT-water and SPC/E-water?
>> Of course you can always reparametrize if you wish, however, I think
>> there is not actually a problem, see above.
>>
>> >c) ... this is related to the implementation of the
>>        Siepmann-Sprik potential and/or image charges
>>        in CP2K?
>> >d) ... something else?
>> see above
>>
>>
>> Best regards,
>> Dorothea
>>
>> 2018-01-18 16:30 GMT+02:00 Maxime Van den Bossche <
>> maxime.... at gmail.com>:
>>
>>> Dear all,
>>>
>>> I've been interested in applying the IC-QM/MM approach
>>> implemented in CP2K to investigate certain metal-water
>>> interfaces. I would like to describe the water with DFT,
>>> the metal using some forcefield, and the metal-water
>>> interactions via the Siepmann-Sprik potential plus image
>>> charge electrostatics.
>>>
>>> During my initial testing of H2O layers on a Pt(111)-c(3x4)
>>> substrate, I found some of the structures (with the
>>> bottom H2O molecules adsorbed H-down) to display overly
>>> short Pt-H bond lengths (1.6 Å). I've attached the
>>> coordinates of such a structure, and the input with
>>> which this (relaxed) structure was obtained. For simplicity /
>>> familiarity, the input file is essentially identical to
>>> that of the online Pt(111)-H2O how-to example.
>>>
>>> The output (with CP2K version 6.0, commit 8b033c3)
>>> is also attached. As can be seen from the 'run-r-1.out'
>>> file, the Pt atoms underneath the down-pointing H atoms
>>> acquire a negative image charge of around -0.07 a.u.,
>>> which still seems to be reasonable.
>>>
>>> When running a full DFT (PBE-D3) optimization of the same
>>> structure, the Pt-H bond lengths expand to more believable
>>> values of around 2.15 Å.
>>>
>>> So, I'm wondering what is going on, because there seem
>>> to be different possibilities. Do you think ...
>>>
>>> a) ... something is wrong in my input? I also got the
>>>        same results with narrower Gaussians for the image
>>>        charge distributions (5 Å^-2 instead of 3.5 Å^-2).
>>>
>>> b) ... this is related to the Siepmann-Sprik potential
>>>        parameters? I couldn't find previous works, though,
>>>        (such as this forum, the original paper, or the
>>>        implementation paper) reporting such behaviour.
>>>        But if this is the cause, I could just add a short-
>>>        ranged repulsive potential to the Pt-H interaction
>>>        or maybe modify the Phi function in Eq 3 of the
>>>        original paper. Could it be that one really needs
>>>        to reparametrize the potential because of the
>>>        differences between DFT-water and SPC/E-water?
>>>
>>> c) ... this is related to the implementation of the
>>>        Siepmann-Sprik potential and/or image charges
>>>        in CP2K?
>>>
>>> d) ... something else?
>>>
>>>
>>> The online how-to:
>>> https://www.cp2k.org/howto:ic-qmmm
>>>
>>> The original Siepmann-Sprik paper:
>>> https://doi.org/10.1063/1.469429
>>>
>>> The IC-QM/MM implementation paper:
>>> https://doi.org/10.1021/ct400698y
>>>
>>> Best,
>>> Maxime
>>>
>>> --
>>> 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+... 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.
>>>
>>
>> --
> 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 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.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20180124/51e3ea86/attachment.htm>