[Wien] lapw1 crash with SECLIT - Error in Cholesky

[Pavel Ondračka]

Laurence Marks píše v Pá 10. 10. 2014 v 11:59 -0500:
> Thanks.
> 
> 
> I don't see anything obviously problematic. The only way I know to
> "check" RMT values is to minimize the energy with a constant
> RKMAX*min(RMT) (Peter might know a better one). My observation is that
> is approximately when the last term on the right of the output, the
> step in the gradient, is roughly the
> same for different atom types. It should also be not too large.
> 
> 
> Your value of 0.1 for Ti & O looks fine to me, both small enough and
> suggesting that your RKMAX is big enough. There is a reasonable amount
> of charge at the RMT for O ( ~1.3) which is to be expected, less for
> the Ti, again fine. The value for the gradient for Si of 0.576 is
> slightly surprising to me, I guess most of the Si valence density is
> outside the RMT except for the 2s (guessing).
> 
> 
> Without real proof I would be slightly cautious of how the Si is being
> treated. It is probably fine.

OK, ATM I'm running the calculation again with different RMTs just to be
sure and if the both results are consistent, I will continue with mBJ.

Thanks a lot for your help.