The ABINIT Users Mailing List ( CLOSED )

["Scott P. Beckman" <sbeckman@ices.utexas.edu>]

Hi Andrew,

There was a paper published in 2003 or 2004 in which the scissors 
operator was applied to turn the gap from negative to slightly 
positive.  This was used as a starting point for the GW correction.  
The results looked reasonable, although this step seems somewhat ad 
hoc.

I don't remember the authors, but I vaguely remember that they were 
affiliated with a university in Japan.

I'm not certain how you would do this in Abinit, or if you would even 
want to.

If you're interested in studying transitions between metals and 
semiconductors, it might be simpler to begin with a semiconductor and 
apply pressure to the crystal until it becomes metallic.  This would be 
easier to control, and if you selected an material such as Si you don't 
have to worry about core corrections.

Scott

On Sep 9, 2005, at 1:00 PM, Andrew Johnson wrote:

> Dear all,
>
> I am a newbie of GW calculation. Recently I've been trying to do some 
> GW calculation in InN but it was not successful. There is a tutorial 
> about Si which I followed up quite smoothly but InN seems to be a 
> tougher system to play with.
>
>  It is to my understanding that in ABINIT GWA is initialized from the 
> LDA KS orbitals and energies, and the problem arises in InN since LDA 
> in fact predicts a negative band gap due to an overestimation of the 
> pd repulsion. That in consequence gives a bad initial starting point 
> for the rest GW calculation. Moreover, a significant change in the 
> dielectric properties from metal to semiconductors is expected. The 
> dielectric constant (since it's "metal") is too large to be printed 
> out in the output file (instead, it only prints *******). I even had 
> trouble with convergence test on the number of planewave to calculate 
> the screening. I wonder if GW calculation in ABINIT code is capable of 
> dealing with such a case where the substantial change from a metal to 
> semiconductor is anticipated. By the way, I use the FHI 
> pseudopotentials for N and In (4d semi-core included). Thanks.
>
> Andrew