The ABINIT Users Mailing List ( CLOSED )

[Sebastien Hamel <hamel2@llnl.gov>]

Thank you for the clarification.

Sebastien

On May 31, 2006, at 1:41 PM, Martin Fuchs wrote:

> Dear ***,
>
> I have some comments. First on the Si case you discuss, then on
> the pseudopotentials on the abinit website in general.
>
> Using the present fhipp version I can reverse-engineer this potential,
> and conclude:
>
> 1) the abinit website Si *potentials* for l=s,p,d,f were generated  
> from
> a neutral atom
>
> 2) the 3s and 3p *wavefunctions* were too
>
> 3) the 4f wavefunction corresponds to a scattering state (at the 3p  
> energy),
> obtained from the neutral atom; this is just the construction  
> proposed by
> D. Hamann in his 1989 paper and is well defined, even if it looks ugly
>
> 4) the 3d wavefunction that for a Si++ with 3s1 3p0.75 3d0.25  
> occupation,
> generated by a separate run; the occupation is like in the BHS paper;
> this yields a bound eigenstate, i.e. a localized function, to be  
> prefered
> e.g. for the Kleinman-Bylander form of the pseudopotential (of course
> irrelevant if d is used as local component, see below too!)  or for
> projected DOS purposes
>
> 5) you can't get this abinit file from a single run of fhipp, but  
> need to
> paste the l-components of the potentials and wavefunctions from
> two different runs (Si and Si++); that's done by hand or using a
> script like in fhipp/bin/tools/multiconfig-script, try yourself  
> with the
> attached for Si
>
> 6) thus this is not a problem of changing fhipp versions but
> of how fhipp has been used:)
>
> ... and perhaps more importantly
>
> 7) the abinit files for Si and N (and presumably other main group
> elements) contain f components. If I understand alright, those f's
> are included in an actual abinit run!? Usually, s,p,d should be  
> sufficient
> for Si and N, and d is an appropriate local potential (this holds for
> many other main group elements too, not for transition metals).
> Why the extra computational effort (5 more pseudopotential  
> projections)
> for including also the f's???
>
> Best wishes,
> Martin
>
> Sebastien Hamel wrote:
>
> > Matthieu Verstraete wrote:
> >
> > > Aaaaah! But your curve isn't just smoothed, it's ugly! It looks  
> > > like the state is unbounded, explodes, and gets cut off at r =  
> > > 2.1 If it depends on the exact energy of the unbounded exploding  
> > > state it could be machine dependent. At this stage, without  
> > > looking into the innards of fhipp (which I will not do again for  
> > > any amount of chocolate) we don't know how it treats the unbound  
> > > states and whether this cutoff has changed from version to  
> > > version. At any rate mine is prettier... and you should be using  
> > > local d psp anyway for first row(s) elements.
> > >
> > > Matthieu
> > I know! This what I thought too. But yes i am using d local anyway  
> > so this is not really an issue.
> > I was just curious as to how your curve was obtained since our  
> > inputs are identical.
> >
> > Thanks,
> > Seb
>
>
> --
> Martin Fuchs
> Phone:  +30 8413 4802           | Fritz-Haber-Institut der MPG
> Fax:    +30 8413 4701           | Faradayweg 4-6
> E-Mail: fuchs@fhi-berlin.mpg.de | D-14195 Berlin (Germany)
>
> 1-0.0735-11600-23.05
>
> #! /bin/csh -f
> #
> # generic script for collecting pseudopotential components or
> # wavefunctions from two different reference states, e.g.
> # to obtain
> # - a bound pseudo wavefunction for the KB potential
> # - a semilocal pseudopotential component with the BHS recipe
> #
> # here for Si:
> # 1. run 1 - generate 3s,p,d,4f components from the neutral atom
> # 2. run 2 - generate the from an the Si 3s1 3p0.75 3d0.25 f0 atom  
> (see BHS paper table II)
> # 3. replace the 3d pseudo wavefunction from run 1 by that from run 2
> #    and put together the usual .cpi file
>
> cd  $cwd
>
> set INI      = si-test
> set COMMENT  = "u_d from ionic configuration"
> set RC       = 0.0d0
> set XC       = 6
>
> #---------------- neutral configuration
>
> rm fort.* coredensity
>
> cat << otto >! $INI.ini
> 14.00 3 2 $XC $RC      : z nc nv
>     1  0   2.00        : n  l   f
>     2  0   2.00
>     2  1   6.00
>     3  0   2.00
>     3  1   2.00
> 3 "t"                  : lmax  s_pp_def
> 0 0.0 0    "t"
> 1 0.0 0    "t"
> 2 0.0 0    "t"
> 3 0.0 0.01 "t"
> otto
>
> psgen  -o $INI $INI.ini -g
>
> #---------------- store pseudopotential components
>
> mv fort.40 1.s
> mv fort.41 1.p
> mv fort.42 1.d
> mv fort.43 1.f
>
> #---------------- ionic configuration
>
> cat << akka >! wp.ini
> 14.00 3 4 $XC $RC      : z nc nv
>     1  0   2.00      : n  l   f
>     2  0   2.00
>     2  1   6.00
>     3  0   1.00
>     3  1   0.75
>     3  2   0.25
>     4  3   0.00
> 3 "t"                  : lmax  s_pp_def
> 0 0.0 0 "t"
> 1 0.0 0 "t"
> 2 0.0 0 "t"
> 3 0.0 0 "t"
> akka
>
> #a frozen core calculation (use this if potentials are from ionic  
> config)
> #psgen -xv -o wp wp.ini -g -fc $INI.fc
> #a relaxed core calculation
> psgen -xv -o wp wp.ini -g
>
> #---------------- store pseudopotentials from ionic configuration
>
> mv fort.40 2.s
> mv fort.41 2.p
> mv fort.42 2.d
> mv fort.43 2.f
> cp fort.27 coredensity
>
> #---for p-component: use potential from calculation 1
> #   and wavefunction from calc. 2
> #   the result is put into file 1.d
> #   in fort.40 etc:  $1 ... index, $2 ... radius, $3 ... wavefct,  
> $4 ... potential
>
> paste 2.d 1.d | awk '{if (NR == 1) {print $0} else {print $1,$2,$3, 
> $NF}}' >! junk
> mv junk 1.d
>
> rm fort.*
>
> #---------------- assemble .cpi file for fhi96md code
>
> echo `grep 'integrated valence' $INI.dat | awk '{print $NF"E+00"}'`  
> `ls 1.* | wc -l` >! ${INI}.cpi
> cat << unused_fit_parameters >>! ${INI}.cpi
>   0.0000    0.0000    0.0000   0.0000
>   0.0000    .00e+00   .00e+00
>   0.0000    .00e+00   .00e+00
>   0.0000    .00e+00   .00e+00
>   0.0000    .00e+00   .00e+00
>   0.0000    .00e+00   .00e+00
>   0.0000    .00e+00   .00e+00
>   0.0000    .00e+00   .00e+00
>   0.0000    .00e+00   .00e+00
>   0.0000    .00e+00   .00e+00
> unused_fit_parameters
>
> awk < 1.s '{if (NR == 1) {print $2,$3} else {print $0}}' >>! $ 
> {INI}.cpi
> awk < 1.p '{if (NR == 1) {print $2,$3} else {print $0}}' >>! $ 
> {INI}.cpi   # substitute only wfct
> #awk < 2.p '{if (NR == 1) {print $2,$3} else {print $0}}' >>! $ 
> {INI}.cpi  # substitute pot & wfct
> awk < 1.d '{if (NR == 1) {print $2,$3} else {print $0}}' >>! $ 
> {INI}.cpi
> awk < 1.f '{if (NR == 1) {print $2,$3} else {print $0}}' >>! $ 
> {INI}.cpi
> cat coredensity >>! ${INI}.cpi
>
> if( $?COMMENT ) echo "comment: " $COMMENT >>! $INI.dat
> cat wp.dat    >>! ${INI}.dat
> exit
>
> rm fort.* coredensity 1.* 2.* *wp* *fc
>
> exit

================================
Sebastien Hamel, PhD
Lawrence Livermore National Laboratory
Physics and Advanced Technologies
H-division
Bldg. 211, Rm. 214, L-415
Phone: (925) 423-8048
Email: hamel2@llnl.gov
================================