[Wien] Plotting surface potential

[Peter Blaha]

Hi,

The numerical problems of PW91 are well known. They are inherent to the
functional itself and these problems were the reason for the development
of PBE96. Thus, PW91 is an old and outdated functional which should not be
used anymore. (most likely I should remove it from the options...)

LDA looks more or less ok. (Small oszillations at large distances ?)

PBE96 is ok almost everywhere, execpt in a small region at 12 Ang, where there
is a sudden step of 0.1 Ry and VXC is even positive. I suppose that this
could be avoided by choosing better cutoffs in lapw0 (see below)   OR  by
a larger GMAX (to get better gradients in this region).


PS: Version sept.2003 is "ok".

If larger GMAX do nor help for PBE96, you can play around with proper
cutoffs yourself in SRC_lapw0/vxclm2.f:

Right at the beginning there is a limit for rho-up or dn lt. 10^-6
In this case LDA is used (and VXC should be smooth). (in addition densities
lt. 10-8 are set to 10-8 to avoid even LDA-numerical problems)

I suppose the problems happen when rhoup is just a little larger than 10-6.
On could reduce this 10^-6  (eg. to 5*10^-5) and check if it is ok now.

Alternatively, better but more complicated, one could look directly into
pbe1.f, subroutines

exchpbe and corpbe;

where the enhancement factors  FxPBE (probably due to S)  and  DVCup
could cause the problem (and should be restricted/smoothed,....)

Regards
                                      P.Blaha
--------------------------------------------------------------------------
Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
Phone: +43-1-58801-15671             FAX: +43-1-58801-15698
Email: blaha at theochem.tuwien.ac.at    WWW: http://info.tuwien.ac.at/theochem/
--------------------------------------------------------------------------