The ABINIT Users Mailing List ( CLOSED )

[Masayoshi Mikami <mmikami@yk.rim.or.jp>]

Hello,

I found a post (about two and half years ago) regarding
this point in my mail-box. Apparently this old e-mail 
cannot be traced from the Web site, so please let me post this. 
(I just wish this post would ease followers' writing, saving 
 their time to find this reference ...
 Doug and Xavier, I would appreciate your kind understanding,
 although I did not confirm your permission to do this.)

More specific detail/summary that you had with the TM PPs
(and the other PPs) on the ABINIT WEB site would be very much 
appreciated, as mentioned in the cited e-mail. 
Your cases were just metallic bulks, or compounds ? etc.

In passing, you might want to refer to 
"~ABINIT/Infos/Psp_infos/psp1.data" as well.
Your contribution might be included in this database 
(or somewhere).

Good continuation,
Masayoshi

---------------------------------------------------------------------------
Subject: RE: (ABINIT) Pseudopotentials
From: "Allan, Douglas C Dr" <AllanDC@corning.com>
To: abinit@pcpm.ucl.ac.be, "'Xavier Gonze'" <gonze@pcpm.ucl.ac.be>
Date: Mon, 02 Apr 2001 14:07:27 -0400
X-Mailer: Internet Mail Service (5.5.2653.19)

To further amplify this point: when we generated pseudopotentials for the
whole periodic table, we only used rough trends across rows and columns of
atoms to choose core radii and other adjustable constants.  For a few of our
favorite elements, these choices were well made and well tested (Si, O, Ge,
C, ...).  For a handful of other elements, we did run typical tests of
lattice constants, etc..  For most of the periodic table, no such tests were
run by us at all.  Our goal was to have full coverage, but not to test and
optimize each potential.

Thus: test every potential before you use it.  Report failures right away
(with details please) to the abinit web site.  Also, I would encourage all
users to report successes as well, with details, so that we may imagine
compiling such test results into a "pedigree" that might accompany each
pseudopotential.  If anyone would care to suggest a specific set of test
results that might be used as a standard pedigree, we might better organize
the test results for each potential.  The goal, of course, is to avoid
duplicating the test runs of others.  (And please accept my apology for
potentials that stink.  I fixed ones that I tested and found wanting.)

Happy computing,
Doug Allan

> ----------
> From:         Xavier Gonze[SMTP:gonze@pcpm.ucl.ac.be]
> Sent:         Monday, April 02, 2001 12:49 PM
> To:   abinit@pcpm.ucl.ac.be
> Subject:      (ABINIT) Pseudopotentials
> 
> Dear Abinitioners,
> 
> Despite being available for a long time, there were still
> a few black sheeps among the pseudopotentials available on the
> Web site (the full table of Troullier-Martins psps) ...
> 
> You might remember that the Nb pseudopotential was 
> withdrawn at the end of 2000. The Cr pseudopotential
> had been criticized by A. Filipetti earlier in 2000, and 
> also recently by Jay Sullivan. The In pseudopotential
> was tested by Steve Erwin, and failed miserably to
> give the correct lattice parameters of bulk Indium.
> 
> These three pseudopotentials are no more available 
> in this table. However, for Cr and Nb, new TM pseudopotentials,
> build using the FHI code, are available. They have been 
> tested and seem OK, unlike the black sheeps ...
> 
> Please, have a look at the updated version of the
> web page :
> http://www.pcpm.ucl.ac.be/ABINIT/Psps/LDA_TM/lda.html
> (note that you might have to reload the graphics).
> 
> Also, please note that pseudopotentials should always
> be tested in well-known situations before using
> them for predictions.
> 
> Do not hesitate to contact us if you notice a problem
> with some pseudopotential ...
> 
> Xavier
> 
---------------------------------------------------------------------------

From: David Clatterbuck <clatterbuck1@llnl.gov>
Subject: [abinit-forum] LDA overbinding and TM pseudopotentials
Date: Fri, 17 Oct 2003 04:50:41 -0700
Message-ID: <5.2.1.1.2.20031017042244.00a663d0@popcorn.llnl.gov>

> Fellow Abinitioners,
> 
> It is my understanding that in most cases LDA will "overbind" compared to 
> experiment, causing the lattice parameter to be too small and the bulk 
> modulus too big.
> 
> In trying out different pseudopotentials for several different problems, 
> I've begun to notice a pattern that the Troulier-Martins pseudopotentials 
> available on the Abinit website (the ones generated by A. Khein and D.C. 
> Allan) tend to produce lattice parameters that are bigger than experiment 
> as well as bulk moduli that are bigger than experiment. With the HGH 
> pseudopotentials, I tend to get the expected LDA overbinding (lattice 
> parameters too small, bulk moduli too big). Most of my work in this area 
> has been on on metals (Al, Cu, Ta, Mo, W, ....). All of the calculations 
> have been thoroughly tested for convergence (ecut, ngkpt, tsmear, nband).
> 
> Has anyone else noticed this? Any ideas as to why there is this systematic 
> difference between the two types of pseudopotentials? Is it related to the 
> fact that the TM pseudopotentials tend to be softer (lower ecut necessary 
> to converge)? Based on this, is the error between 
> all-electron-full-potential LDA calculations and pseudopotential LDA 
> calculations larger for the case of TM pseudopotentials?
> 
> Thanks for any thoughts,
> 
> David Clatterbuck
> Postdoctoral Researcher
> Lawrence Livermore National Lab