[Pw_forum] Logarithmic Derivative Energies

[Andrea Dal Corso]

On Wed, 2005-06-29 at 10:23 -0400, Stephen Daunheimer wrote:
> Dear users,
> I was trying to make an ultrasoft pseudopotential for Ge and I am having 
> trouble.  This is the output that I get for the all-electron run:
> 
> s(4S/4S) =  1.000000  <r> =   2.1065  <r2> =    5.1396  r(max) =   1.7277
> s(4P/4P) =  1.000000  <r> =   2.8791  <r2> =    9.9061  r(max) =   2.1908
> 
> If I put the rcut for the pseudopotential input at about 1.73 I get an 
> error saying that there is a node for 0 < r < 1.749.  If I increase the 
> rcut to 1.90 though the program will run fine and give me a 
> pseudopotential.  This is what the input file looks like for the 
> pseudopotential:
> &input
>        title='Ge',
>        zed=32.0,
>        rel=1,
>        isic=0,         lsd=0,         beta=0.3,
>        rlderiv=2.4,
>        rmax=250,
>        eminld=-4.0,
>        emaxld=4.0,        deld=0.02d0,
>        nld=2,
>        config='[Ar] 3d10 4s1 4p3',
>        iswitch=3,
>        dft='PZ',
>        file_logderae='geUS.PZ.AELOG.dat',
> /
> &inputp
>        lloc=0,
>        zval=4.d0,
> !       rho0=1d0,
>        file_pseudopw='geUS.PP.dat',
>        file_logderps='geUS.PP.LOG.dat',
> /
> 2
> 4P  2  1  3.00  0.00  2.20  2.30
> 4S  1  0  1.00  0.00  1.90  1.93
> 
> The problem with this is that while it will run, the pseudopotential is 
> not very good.  When I graph the logarithmic derivatives they are off by 
> about .02 ryd.  Any ideas on why the program cannot take the input file 
> with the rcut set to 1.73?
> 
In the atomic code it is not possible to use a small core radius if the
pseudization is done with the RRKJ recipe. If you really want a very
small core radius, for norm conserving pseudopotentials, you can use
tm=.true.. 
Moreover, in the US approach you need at least two energies per l
channel.
The pseudopotential does not work if you use two different core radii
but only one energy per channel. 
You can look at examples for similar elements in the pseudo-gen
directory. 
si_nc.in, or al.in for generating a norm conserving pseudo-potential,
or as.in for a US-PP. 








> Best,
> Stephen Daunheimer
> 
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum
-- 
Andrea Dal Corso                    Tel. 0039-040-3787428
SISSA, Via Beirut 2/4               Fax. 0039-040-3787528
34014 Trieste (Italy)               e-mail: dalcorso at sissa.it