[Wien] negative grr values in mBJ calculation causing lapw0 to fail

[tran at theochem.tuwien.ac.at]

In principle, GRR should not be negative because it is the average of
|∇ρ|/ρ (note that it's the norm) in the unit cell volume.
In output0_grr, there is
TOTAL=  *************
for the PLANE WAVE CONTRIBUTION, which means that the problem comes from
the interstitial region. I would guess that this comes from some numerical
instabilities.

Such problems car arise close to the nucleus of very atoms or in regions
of space with low electron density. Is it the case in your system?
For the moment, I can not say much more.

F. Tran

On Tue, 10 Jul 2012, Neil Johnson wrote:

> Hello all,
> I am attempting to perform an mBJ calculation using the steps listed in sec
> 4.5.8 of the user's manual. However, lapw0 keeps failing in the final scf
> cycles after switching to IXC=28. We tracked the problem down to a negative
> GRR (∇ρ/ρ = -36.0544), the square root of which is taken to calculate "c for
> IXC=28" in lapw0. Taking the square root of a negative number is generally
> frowned upon, and the program spits out a lot of "NaN"s before failing.
> 
> Has anyone experienced this problem before, or know what would cause lapw0
> -grr to produce a negative GRR value?
> 
> I've posted the output0 and output0_grr files to
> http://homepage.usask.ca/~nwj781/lapw0/
> <http://homepage.usask.ca/%7Enwj781/lapw0/> for reference. Thanks!
> 
> Cheers,
> 
> Neil Johnson
> University of Saskatchewan
> Saskatoon, Canada
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
> 
>