What part of the pseudopotential is used as Vlocal in SIESTA? Is it the
one with largest l? and can one control which l to use as local part or
to provide other way to generate Vlocal?
Is that PS.lmax and PS.KBprojectors that controls what l to use as
Vlocal? - the description is not very clear in manual.
Where one can read about multiple KB projectors per angular momentum?
-------------
I obtain ghost states for As l=0 if I use default lmax=3, but with
lmax=2 everything is OK (except that pseudopotential for l=3 is not used
at all - it is even not read from PSF - that means that just because of
not constructing separable form for l=3 it won't be used in calculations
at all?)
[See also example for Ga on page 16 in manual - it will generate KB
projectors only for p pseudopotential]
SIESTA proposes to change parameters of pseudopotential generation but
changing lamx or KBprojectors block helps too.
So what solution to avoid ghost states is better? - adding additional KB
projector or changing cutoff radii for pseudopotentials?
--------------
Although SIESTA shows Ekb and kbcos for successfull potentials it
doesn't show them if ghost states are present so it is difficult enough
to fix the potential parameters with closed eyes.
If generation of KB potentials is not realted to the structure isn't it
better to have all that info in ATOM directly?
Log derivatives of the wavefunctions for KB potential (not for the
original potential) would be also very helpful because the absence of
ghost states doesn't yet guarantee a good potential - if Ekb is high
enough the quality of separable potential will not be as good as of
original pseudopotential. So it would be nice to have in ATOM at least
Ekb and kbcos.
Also probably graphs of deltaVl would be helpful to find pseudopotential
parameters to eliminate ghost states.
-------------
Is there an error in .OUT file of SIESTA?:
by default SIESTA uses 1 projector per angular momentum, so why
for lmax=2 (i.e. for 3 values of l) we see
KBgen: Total number of Kleinman-Bylander projectors: 9
for lmax=3 (4 l values) we see 16
for 2 projectors for l=0 and 1 projector per l=1,2,3 we see 17
?
Sincerely,
Alexander Vozny
University of Sherbrooke, Canada
