[Pw_forum] ld1.x Questions

Paul M. Grant w2agz at pacbell.net
Sat Jan 19 01:26:42 CET 2008

Thanks to all who replied.  Additional comments follow:

1) Negative occupations.  I should have guessed...or scanned INPUT_LD1 more
thoroughly.  Let me see if I got this right.  In constructing a PP for
sulfur, I should "unoccupy" the 3d states?  But what about heavily oxidized
sulfur molecules, like in SF6 where S3d participates in the bonding?  I
guess this is why PP cooking requires a good chef.

2) TM PP examples.  This was, as we Yanks say, a purposely "loaded
question."  Or casting a net seeking pearls of wisdom.  Those following my
LDA+U thread know I'm looking for a Cu PP (and probably an O partner) that
will yield a gap in cuprate compounds (nominally divalent Cu-O bonds). I
strongly suspect given the notoriety of this family of structures, many have
tried and failed.  The only formalism that I know which has succeeded in
this endeavor has been LDA+U on an LMTO basis set.  The Cu PP issue may
perhaps be the most interesting challenge facing PP "theory and practice"
today.  Thanks for pointing out that the CVS /pseudo_library/... does
contain some TM examples, but they're 4d and 5d.  As far as I know, none of
their monoxides are Hubbard insulators (if indeed they can be formed
chemically at all...during the pandemonium of early Spring, 1987, my group
at IBM Almaden mined the whole 4 and 5 series searching for HTSCs with no
hits.  Those were very heady and naïve days.).

2a) I'm going to try to "reverse engineer" Fe.pz-nd-rrkjus.UPF used in
example 25 into an ".in" file (does anyone have one?) to try and understand
better how ld1.x works (an O.pz-rrkjus.in file already exists in
/atomic_doc/pseudo_library/LDA/SR, but the O.pz-rrkjus.UPF produced by CVS
ld1.x is "somewhat" different than that contained in the distro library
/pseudo and used in example 25 (different total energies by a few tenths of
a percent, no E pseu  and no wfc and rho suggested cutoffs in the latter and
fewer mesh points in the former ???)).  Simplistically, Fe++ is the minority
spin electron analog of the minority spin hole state expected in Cu++.

3) ld1.x - CVS vs Espresso.  Espresso ld1.x screws up the &input namelist
sequence, especially the "author" entry (is "ADC" the mysterious Andrea Dal
Corso?), and about half the time fails to properly format the total energy
(all asterisks in the UPF output, likely a Fortran format placeholder limit

4) Grace output.  Another loaded question. I anticipated Paolo's answer, but
was hoping that some saintly soul in a state of grace...usually a
postdoc...in the PWscf community had taken this task on.

Paul M. Grant, PhD
Principal, W2AGZ Technologies
Visiting Scholar, Applied Physics, Stanford University
EPRI Science Fellow (Retired)
IBM Research Staff Member Emeritus
w2agz at pacbell.net

-----Original Message-----
From: pw_forum-bounces at pwscf.org [mailto:pw_forum-bounces at pwscf.org] On
Behalf Of Lorenzo Paulatto
Sent: Friday, January 18, 2008 1:28 AM
To: pw_forum at pwscf.org
Subject: Re: [Pw_forum] ld1.x Questions

On Thu, January 17, 2008 22:20, Paul M. Grant wrote:
> 1)	What is the purpose of the "negative ocs(nl)" settings one sees
> frequently in the input files in /atomic_doc/pseudo_library? (not covered
> that I can see in INPUT_LD1)  It doesn't seem to make physical sense.

Orbitals with negative occupation are excluded from the all-electron
calculation; you need to do this when they are unbounded.

Putting a negative occupation means: I don't really want to compute this
orbital, but I'd like to use it's label later for a specific orbital
identified by an arbitrary, but well defined, reference energy, and L and
M indexes.

There is probably a more formal way to say that.

> 2)	Why doesn't this library contain some transition metal examples?

The examples in the library are the pseudopotentials that Andrea Dal Corso
(and only he) has done during the years. There are a few transition metals
among them in the CVS version.

> 3)	I've observed ld1.x from O-sesame (CVS) performs much better the
> executable in the present distribution  espresso-3.2.3, which yields a
> number of formatting errors.

There have been a few improvements on that front recently, nothing should
really change but the number of things that can will cause crashes has
been reduced a bit. Keep in mind that "input files which produce a good
pseudo" are a tiny subset of "input files which doesn't cause a crash."

> 4)	It's stated that ld1.x output can be plotted with Grace, but I see
> no options to produce or launch .agr files such as exists in Opium.  I
> guess it has to be done "by hand."

I think you are right.

> -Paul

Lorenzo Paulatto
+39 040 3787 511

  SISSA Webmail https://webmail.sissa.it/
  Powered by SquirrelMail http://www.squirrelmail.org/

Pw_forum mailing list
Pw_forum at pwscf.org

More information about the users mailing list