[Pw_forum] failure with NCPP

[Luke Shulenburger]

Dear Ian,

In regards to the kink in the ionic potential near the origin, this is
often caused because the GGA functionals can be ill behaved in regions
of small charge density.  Opium has two options for handling this
problem.  The first is to switch to LDA inside of some radius and the
second is to extrapolate the potential from outside of some radius
into the core.  These two options are controlled by adding a line
after the functional name in the [XC] section of your input.  If a
positive number is given, this is the radius at which the code
switches to LDA.  If a negative number, this gives the radius at which
the extrapolation begins.

About your tests, it appears you have everything under control.  I
would just make sure that your results are well converged with respect
to the plane wave cutoff.  When I ran opium with your input file, it
suggested that very high cutoffs (>400Ry) would be necessary to get
energy convergence within 1mRy (or even 10mRy).  This may not be
entirely accurate (and indeed, I have made changes to this part of my
version of opium so it may not even be correct), but it is always a
good idea to test this.

Luke
(lshulenburger at ciw.edu)
Geophysical Laboratory
Carnegie Institution of Washington
5251 Broad Branch Rd NW
Washington, DC 20015

On Mon, Jul 20, 2009 at 11:11 AM, lan haiping<lanhaiping at gmail.com> wrote:
> Dear Luke,
>
> I tested this pseudopotential for Zinc  bulk (Hex), and got   well
> comparable lattice constants 2.65 , 4.97 \AA,  and the bulk module is 74.2
> KBar by fitting Murnaghan Equation (Energy cutoff was set to 50 Hartree) .
> Opium's test also gives small differences for eigenvalues from different
> configurations.
>
> Is there any way to improve the quality or remove the kink you mentioned ?
> I forgot to check ionic potential, and just found it indeed has a large kink
> .
>
> thanks.
>
> regards,
> On Mon, Jul 20, 2009 at 10:38 PM, Luke Shulenburger <lshulenburger at ciw.edu>
> wrote:
>>
>> Hello,
>>
>> I took a quick look at these pseudopotentials.  The first thing I
>> found is that the problem seems to be related to  the augmentation
>> functions.  I removed the second and third line after KBdesign in the
>> Zn input and the potential loads into pwscf without issue.  I haven't
>> used augmentation functions much, so I'm not sure quite what the
>> problem is.  I know Eric Walter looks at this list from time to time,
>> perhaps he could comment on the reason for this.  The other thing is
>> that the the output from opium for the Zn pseudopotential looks
>> somewhat strange.  As an example, the plots of the ionic potential are
>> not very smooth at all and there are large kinks near the origin which
>> will likely result an a very high plane wave cutoff being necessary to
>> get converged results.  Has this potential been tested?
>>
>> Sincerely,
>> Luke Shulenburger
>> (lshulenburger at ciw.edu)
>> Geophysical Laboratory
>> Carnegie Institution of Washington
>> 5251 Broad Branch Rd. NW
>> Washington, DC 20015
>>
>>
>> On Mon, Jul 20, 2009 at 8:28 AM, lan haiping<lanhaiping at gmail.com> wrote:
>> > Dear Paolo,
>> > Thanks .
>> > I then check input settings for opium ,  and find the wavefunctions are
>> > already arranged by increasing 'l'.
>> >  The setting for Zn  pseudopotential  is following:
>> > "
>> > [Atom]
>> > Zn
>> > 8                   # norb: number of orbitals
>> > 100  2.00  -        # nlm occ eigen(- means auto-generate)
>> > 200  2.00  -
>> > 210  6.00  -
>> > 300  2.00  -
>> > 310  6.00  -
>> > 400  1.50  -
>> > 410  0.00  -
>> > 320  10.00  -
>> >
>> > [Pseudo]
>> > 3  1.87 1.85 1.97
>> > opt
>> >
>> > [Optinfo]
>> > 7.07 10
>> > 7.07 10
>> > 7.07 10
>> > [XC]
>> > gga               # lda[PZ] or gga[PBE]
>> >
>> > [KBdesign]
>> > 0                 # local orb
>> > 1                # number of aug. operators
>> > au 0.1 1.60 -3.05  # unit(gp or au)  left  right  height(Ry)
>> > [Loginfo]
>> > 0
>> > 2.0 -4.0 4.0
>> > "
>> > and for O is
>> > "
>> > [Atom]
>> > O
>> > 3
>> > 100   2.00   -
>> > 200   2.00   -
>> > 210   4.00   -
>> > [Pseudo]
>> > 2  1.34 1.53
>> > opt
>> > [Optinfo]
>> > 7.07  10
>> > 7.07  10
>> > conmax
>> > [XC]
>> > gga
>> > "
>> >
>> > I donot know why these PP cannot be read by pw.x .
>> > By the way, is there any simple method to use  opium's settings  in
>> > ld1.x   ?
>> >
>> >
>> > Regards,
>> >
>> > Hai-Ping
>> >
>> >
>> > On Mon, Jul 20, 2009 at 4:49 PM, Paolo Giannozzi
>> > <giannozz at democritos.it>
>> > wrote:
>> >>
>> >> lan haiping wrote:
>> >>
>> >> > I wonder  whether QE still support NCPP format
>> >>
>> >> it does, otherwise it would have issued a different message
>> >>
>> >> > or other special consideration needs to take into for opium's pp ?
>> >>
>> >> atomic wavefunctions should be in order of increasing l.
>> >> Not sure why this is needed and whethere it is really needed.
>> >>
>> >> P.
>> >> --
>> >> Paolo Giannozzi, Democritos and University of Udine, Italy
>> >> _______________________________________________
>> >> Pw_forum mailing list
>> >> Pw_forum at pwscf.org
>> >> http://www.democritos.it/mailman/listinfo/pw_forum
>> >
>> >
>> >
>> > --
>> > Hai-Ping Lan
>> > Department of Electronics ,
>> > Peking University , Bejing, 100871
>> > lanhaiping at gmail.com, hplan at pku.edu.cn
>> >
>> > _______________________________________________
>> > Pw_forum mailing list
>> > Pw_forum at pwscf.org
>> > http://www.democritos.it/mailman/listinfo/pw_forum
>> >
>> >
>> _______________________________________________
>> Pw_forum mailing list
>> Pw_forum at pwscf.org
>> http://www.democritos.it/mailman/listinfo/pw_forum
>
>
>
> --
> Hai-Ping Lan
> Department of Electronics ,
> Peking University , Bejing, 100871
> lanhaiping at gmail.com, hplan at pku.edu.cn
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum
>
>