[Wien] setrmt_lapw #3

[John Rundgren]

Peter Blaha:

In the year 2012 all LEED programs use a spherical KKR-ASA
representation, working well for interpreting LEED experiments. That's
why I insist about a spherical .vcoul and a spherical .cdlm for radii
beyond setrmt_lapw.

What my LEED community has in mind is supercell surface slabs.

As you said earlier, WIEN2k contains a transformation of Fourier
components into multipoles, of which LM=0,0 concerns LEED.

v(j,lm1,jatom) on line 1587 in lapw0.F (with usual factors) gives
correctly VC(0,0) in Rydbergs for r<rmt, as checked by r0*VC(0,0)=-2*Z.

At which lines shall I put in the Bessel transformation of exp(iKr) in
order to generate spherical .vcoul and .cdlm ? My printing out CVOUT
(r>rmt) after line 1591 is perhaps inappropriate.

TiO2 from lapw0.F after line 1591:
Ti
rmt,    v/sqfp  1.920000000000E+00 -1.303086273225E-01 (Ryd)
rmt,cvout/sqfp  1.920000000000E+00  2.538710687746E-02 
rmt,cvout/r**2  1.920000000000E+00  2.441269278879E-02
O
rmt,    v/sqfp  1.740000000000E+00  4.493324211710E-01 (Ryd)
rmt,cvout/sqfp  1.740000000000E+00  5.567839465700E-01
rmt,cvout/r**2  1.740000000000E+00  6.519182522264E-01

Best regards,
John Rundgren 


On Sat, 2012-09-22 at 18:20 +0200, Peter Blaha wrote:
> 
> Am 22.09.2012 15:10, schrieb John Rundgren:
> > Dear Peter Blaha,
> >
> > Here is a new test run on TiO2_rutile.vcoul extended into the Fourier
> > domain on the assumption of the following units,
> >
> > v(j,jm1,jatom) in units of Rydberg*sqrt(4*pi),
> 
> As I said, I'm not sure where your v comes from, but most of the time it should
> also have a r^2 factor !?
> 
> 
> > CVOUT(LM1) in units of Rydberg*radius(Bohr)**2.
> >
> > As a result the kinks in extended vcoul come out smaller than in the
> > previous email "setrmt_lapw #2". See Attached file VC-TiO2-bis.ps .
> >
> > I shall be glad for help about units of potential and elimination of
> > kinks. Is a meaningful max radius for the vcoul extension stored in the
> > program?
> 
> Our program NEVER have any extensions of vcoul (spherically symmetrized) above RMT.
> You have to find out, until what distance a spherical approximation is still ok.
> 
> For me it is always clear: If some results depend crucially on the size of a reasonably
> chosen sphere, the spherical averaging procedure is problematic and one should consider
> to extend the formalism to full (non-spherical potentials.
> 
> 
>