The ABINIT Users Mailing List ( CLOSED )

[Konstantin Rushchanskii <Konstantin.Rushchanskii@physik.uni-regensburg.de>]

Dear Bohdan,

I think, there are two reasons for your problem:

1. Be sure your results are well converged with respect to k-points mesh 
and Ecuts
2. When you use LDA for the band structure investigations, be carefull, 
because LDA typically underestimates the band gap. For ionic crystals 
the underestimation is about 2/3 of experimental value, but for covalent 
crystals with strong anisotropy and with the presence of Wan der Vaals 
interaction and/or hydrogen bonding (I suppose this all is present in 
your crystal), the underestimation of the LDA results are not 
predictible. You have at least two possibility to treat this incostistance:
   a. If you want a purely ab initio results, you can obtain the band 
structure in GW approach. ABINIT can do this, but (as far as I know) 
only for centrosymmetrical crystals. But for your crystal it is 
computationally VERY COMPLICATED procedure.
   b. If you have some experimental data about Eg, you can apply so 
called "scissors" operator (rigid shift of all conduction bands with 
respect to experimentally observed Eg). This correction is very usefull 
for dielectric properties of the crystals. In ABINIT you can apply this 
correction by "sciss" input variable.
       Any way, you cannot expect very good coincidense between 
experimental and theoretical results for these quantities.

I hope, this helps.

Sincerely,
K.Rushchanskii

> Dear Abinit Users,
> When I have calculated the conductivity spectrum g(E) of my crystal (Eg > 5 
> eV) using the 'abinis' and 'conducti' exe-files, I have calculated the imaginary 
> part of the dielectric permittivity epsilon'' using the relation:
> g=2*pi*f*epsilon0*epsilon'' , where f=E/h (h=6.62E-34 is Planck constant), 
> and epsilon0=8.85E-12.
> The structure of the spectrum epsilon''(E) obtained agrees well with the 
> corresponding experimental data. Then I have calculated the real part of the 
> dielectric permittivity epsilon'(E) using the known Kramers-Kronig relation:
> epsilon'(E)-1=(2/pi)*integral{[t*epsilon''(t)*dt]/[t^2-E^2]}.
> The magnitude of the epsilon'(0)=1.4 obtained by this calculation is much 
> smaller than the experimental one and the analogous magnitude obtained from 
> the RF calculations (2.5) using the Abinit.
> Maybe anyone already has done similar analysis? I will be thankful for any 
> advices to resolve my problem.
>
> Regards,
> Bohdan Andriyevsky
>
>  


--
K.Rushchanskii, Dr.
Theoretische Physik, Universität Regensburg
93040 Regensburg Germany
tel: +49 (0) 941-943-2466  fax: +49 (0) 941-943-4382