[Pw_forum] Applying vdW to only part of a system

[Alberto Otero de la Roza]

Hi Ian,

Your results sound reasonable to me, but unless I'm misunderstanding
what you are doing it sounds a bit like comparing apples and oranges.

A some-small-% underestimation of the equilibrium lattice parameter
relative to the experimental lattice constant is more or less the
correct result because of thermal expansion. If you want to compare
the two, you need to have some way to model vibrational effects and
how those effects change with pressure (the Grüneisen gamma). Several
options: i) Ruszinszky et al. have an approximate formula that uses
the Debye temperature, ii) you could use a simplified Debye model like
the one implemented in gibbs2 (only the E(V) curve is needed), iii)
you could go the whole hog and calculate phonon frequencies on a set
of volumes around equilibrium or, iv) if you know the experimental
thermal expansion coefficient you could use it directly. Either way,
in absence of thermal effects, you can't say much about the
performance of a functional for lattice parameters other than "they
should be a tad smaller than experiment".

Best,

Alberto

* Ian Shuttleworth <shuttleworth.ian at gmail.com> [2017-09-12 11:54:42 +0100]:
> Hi all
> 
> First, thanks for all the replies.
> 
> My thoughts are that I want vdW for the adsorbate-adsorbate interaction,
> the adsorbate-substrate interaction but not for the bulk
> substrate-substrate interaction.
> 
> When setting up a surface simulation, you can first determine a bulk
> lattice constant. Then, using this lattice constant, you define a slab
> which is the truncated bulk and relax the surface layers and add the
> adsorbates. So, for a material whose bulk lattice constant is reasonably
> well described by non-vdW (e.g. PBEsol) you then overbind if you add vdW
> to bulk component of the simulation, and contract the surface lattice
> constant as a result.
> 
> I've found that if I get a +1 to +2% bulk lattice parameter overestimation
> with PBE, I get an equivalent underestimation if I 'turn on' the vdW
> correction. I haven't looked at any other parameters such as formation
> energies yet, as I was interested to get some feedback about this lattice
> constant problem first,
> 
> Does this sound reasonable? Or am I thinking about the problem incorrectly?
> 
> Again, very kindest thanks for the responses,
> 
> Ian
> 
> 
> 
> 
> 
> Hi Ian, It is possible with dispersion corrections based on the asymptotic
> formula (like XDM or D2, as described by Evren). Why would you want to do
> that, though? In general, dispersion improves the description of metals and
> ionic solids and, one would guess, surfaces. There was a recent paper by
> Perdew and Ruzsinszky on the topic, if memory serves. And I also know for a
> fact that, for instance, B86bPBE-XDM reproduces alkali halide and alkaline
> earth oxide cubic cell lengths with errors of 0.1 angstrom of less once
> vibrational effects are taken into account.
> 
> Best,
> 
> Alberto
> 
> --
> Dr. Alberto Otero de la Roza
> Department of Chemistry,
> University of British Columbia, Okanagan
> Kelowna, British Columbia, Canada V1V 1V7
> 
> * Ian Shuttleworth <shuttleworth.... at gmail.com> [2017-09-11 16:33:30 +0100]:
> > Is it possible to apply vdW forces to only a part of system?
> >
> > I am thinking of examples where, for example, a vdW adlayer (e.g. benzene)
> > sits on a surface, and surface doesn't need vdW corrections.
> >
> > If I add the vdw_corr toggle (or if I set an appropriate value
> > to 'input_dft') I've understood that vdW are applied to the entire system -
> > if this is right, how could I make this option more selective?
> >
> > With thanks
> >
> > Ian
> 
> > _______________________________________________
> > Pw_forum mailing list
> > Pw_forum at pwscf.org
> > http://pwscf.org/mailman/listinfo/pw_forum
> 
> _______________________________________________
> Pw_forum mailing
> listPw_forum at pwscf.orghttp://pwscf.org/mailman/listinfo/pw_forum

> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://pwscf.org/mailman/listinfo/pw_forum