[gpaw-users] Dipole correction with implicit solvent

[Georg Kastlunger]

Dear Alexander,


On 02/09/2017 04:57 AM, Alexander Held wrote:
> Dear Georg,
>
> It might be the case that the analytical solution of the dipole 
> correction
> is only valid in vacuum and not for the Poisson equation with 
> dielectric (I
> did not check that). Maybe one should figure this out with pen and paper
> first. Otherwise it could be that this combination will produce wrong
> results. The monopole-correction is already adapted to the case with
> dielectric, but I never worked with the dipole-correction in the 
> solvation
> model.
I found out that, in principle, the implementation of the dipole 
correction works with implicit solvent too. However, in it's original 
implementation the countering dipole is added in the center of the unit 
cell and is quite wide. In the work we are conducting we changed the 
solvation model in the sense, that only parts of the unit cell are 
filled with the dielectric. Therefore applying the original dipole 
correction leads to a countering dipole that is partly solvated making 
it arbitrary. Since in our model we have the region behind the slab not 
being filled with dielectric, I decided to shift the countering dipole 
into this region and narrowing it, which does not change the result. 
Doing so the correction can be used, since the correcting potential 
(dvHt_g in the code) is constant in the solvated regions, meaning 
fieldless. However, an effective dipole of the surface system has to be 
calculated for the potential (or electric field) reduction due to the 
solvent. Unfortunately an analytical epsilon dependance is still missing 
to my knowledge at the time, but I guess figuring this out is part of 
the fun of the subject.
> I guess that you are working with slabs and the solvation model. As a 
> hint:
> In this case one has also to carefully check that the point-like solvent
> does not enter between the atoms inside the slab. I am currently 
> working on
> implementing wall potentials to protect slabs from the solvent (untested,
> not merged to master, see here:
> https://gitlab.com/aheld84/gpaw/tree/solvation_slab_tools ).
Yes that happens in our case too. We countered this issue by changing 
the definition of the effective potential u_eff near the surface by 
changing from a three dimensional to a two dimensional situation. This 
means that the LJ potential raises near the surface plane instead of the 
surface atoms and inside the slab (and also behind it in our work) u_eff 
is just set to be infinity.

I will take a look at your implementation. Thank you for sending the 
link. I suppose we will try to combine our efforts on the implicit 
solvent model with the original implementation after it is fully grown 
and tested. Given that you agree on that, of course.

Best wishes,
Georg
> Best regards
> Alexander
>
>> -----Ursprüngliche Nachricht-----
>> Von: gpaw-users-bounces at listserv.fysik.dtu.dk [mailto:gpaw-users-
>> bounces at listserv.fysik.dtu.dk] Im Auftrag von Georg Kastlunger via gpaw-
>> users
>> Gesendet: Samstag, 4. Februar 2017 15:05
>> An: gpaw-users at listserv.fysik.dtu.dk
>> Betreff: [gpaw-users] Dipole correction with implicit solvent
>>
>>
>> Dear gpaw community,
>>
>> I am currently working on a dipole correction combined with the provided
>> implicit solvent model in GPAW. This has proven to be quite a 
>> challenge in
>> terms of electrostatics and I am afraid I am not able to capture the
> physics
>> correctly, unless I can understand the routine for the dipole correction
>> completely.
>>
>> Therefore, I was wondering if there is any literature reference which
> explains
>> the reasoning behind the dipole correction in fd mode.
>>
>> One question upfront:  Why is the potential representing the dipole
>> contribution (dvHt_g) subtracted before the poisson equation is solved?
>> Does that mean that the starting guess for the potential is changed?  I
>> noticed that if I change dvHt_g it changes the slope at the outer 
>> parts of
> the
>> resulting electrostatic potential, even though dvHt_g is always constant
> in
>> those regions.
>>
>> Best wishes,
>> Georg
>>
>> _______________________________________________
>> gpaw-users mailing list
>> gpaw-users at listserv.fysik.dtu.dk
>> https://listserv.fysik.dtu.dk/mailman/listinfo/gpaw-users