The ABINIT Users Mailing List ( CLOSED )

[Fabio Finocchi <finocchi@gps.jussieu.fr>]

Dear Conor,
I read your email just now (!!!!). I hope you succeded in managing polar 
slabs...
In any case, there are few hints below.
I have some experience in dealing with polar slabs. See, e.g. Bottin et al, 
PRB 68 (2003) 035418, and references therein. I can also give
you, on request, the text of an article that recently appeared in the 
proceedings of a NATO-ASI school [F.Finocchi, F. Bottin and C.
Noguera, "Computer simulations of polar oxide surfaces", in Computational 
Materials Science, edited by C.R.A. Catlow and E. Kotomin
(NATO-ASI Series E, IOS Press, 2003), pages 196-217], where a detailed 
analysis of the numerical problem is carried out.
First of all, the physics: if your slab does not fulfill the electron 
counting rule, it is likely metallic (but not necessarily, since some
symmetry can be broken, which may open a small gap around EFermi). Therefore, 
I would suggest you to use some smearing of the Fermi surface,
eventually using different values for tsmear (two are generally enough to 
extract the T=0 total energy). Also, start a test calculation from
a slightly perturbed configuration with a low symmetry.
Second, the Broyden method is not very stable and can diverge whenever the 
electronic structure is very sensitive to atomic relaxations. A
quasi-continuum damped dynamics (ionmov=7, with equal atomic masses amu~4, 
vis~10, and dtion~50) is better suited when atomic forces are
less than, let's say, some mHa/bohr.
Third, we have a home-made version of the code in which a compensating dipole 
in the vacuum region is used. However, due to the lack of
human time, nobody included it in a abinit version so far, to my knowledge.
Nevertheless, some quantities, such as the total energy, are not extremely 
sensitive to the use of a compensating dipole, provided that a
reasonable vacuum region is added in the supercell. This is because the 
electrostatic potential is mostly biased in the slab outer regions,
where the electron density is small.
Others, like the work function, do actually depend on the use of a dipole.

Buona fortuna,
Fabio

cdhogan@roma2.infn.it a écrit :

> Dear Abinites,
>
> Apologies in advance for the long post:  this will perhaps concern people 
> with
> experience in using slab geometries or with the internal workings of the
> code.
>
> I'm trying to optimize the geometry of a polar slab: GaAs(001) [back surface
> passivated by pseudohydrogens, Z=1.25], and I'm quickly running out of
> options, as the calculation refuses to converge, at least with the Broyden
> scheme.
> The slab is ~11 layers thick, and I've tried
> 7 and 11 layers of vacuum (the latter =15 Angs!). I'm using geometries
> converged with another code to start, and the same pseudopotentials
> (Hamann type from FHI code).
> For odd reasons, Im using Abinit 4.0.3, compiled with IFCv7.1, on a Dell 
> Xeon.
>
> Typical input parameters I've tried are:
> ionmov 3 or 2
> iscf 5 or 3
> diemac 12 or 8
> diemix 0.5 or 0.2
> iprcell 45 or undefined
> tolmxf 5e-4, toldff 5e-5 or lower
> k-points: Gamma pt, 1/2/4 points in IBZ (1/4 SBZ)
> The convergence of the SCF part is usually without problem, converging
> in 10-20 steps.
>
> However, the total energy and forces, while initially decreasing,
> jump to high values after a few steps
> and oscillate; the atomic positions overshoot, and the system never
> really seems to be relaxing smoothly - and even after this system has 
> returned
> to the original atomic coordinates the same instability occurs.
>
> I noticed that these jumps coincide with the one-dimensional electrostatic
> potential showing a constant gradient in the vacuum (perp to surface), not 
> to a
> constant value. This I recognise as being associated with a macroscopic
> electric field in the cell appearing due to the periodic boundary
> conditions, which people have solved with dipole layers in the vacuum.
> In the absence of such a correction, I would expect to see such a constant
> gradient at all times, even for fixed geometry, but this does not occur.
>
> So my questions are:
> 1) Can Abinit handle polar slabs?
> 2) Does Abinit have, or plan to, incorporate such a dipole layer in the 
> vacuum?
> (or do I have to tackle it myself...?!)
> 3) Has anyone managed to converge geometry for polar slabs, without just 
> making the slab symmetric? (this would make my slabs too large)
> 4) Has anyone any further suggestions or tricks for converging such
> structures,
> other than whats in the FAQ, etc.? Am I simply missing something?
> It was suggested to me that it might be a problem with the pseudo...
> 5) Where does the "sensitivity to the vacuum thickness" come from?
> Would a molecular dynamics calculation help?
>
> Any comments, criticisms or full solutions (!) are most welcome.
> All the best,
> Conor
>
> ---
> Dr. Conor Hogan
> Dipartimento di Fisica
> Universita' di Roma "Tor Vergata"
> Tel: +39 06 72594548
> Fax: +39 06 2023507
>
> What is it that makes a complete stranger dive into an icy river to save a
> solid gold baby? Maybe we'll never know.  - Jack Handey
>
> To me, it's always a good idea to always carry two sacks of something when
> you walk around. That way, if anybody says, "Hey, can you give me a
> hand?," you can say, "Sorry, got these sacks." - Jack Handey

--

      Fabio Finocchi

       Groupe de Physique des Solides  (UMR CNRS 7588)
  Campus de Boucicaut, 140 rue de Lourmel, 75015 Paris (France)
  tel: (33) 1 44275116    fax: (33) 1 43542878    bureau: 2.1.5