The ABINIT Users Mailing List ( CLOSED )

["Andrew M. Rappe" <rappe@sas.upenn.edu>]

Dear Chol-Jun,

I am afraid that you may have missed the point.  In "first principles"
calculations, we try to solve the governing equations as best we can.

In a pseudopotential calculation, this means that pseudopotentials are
made that reproduce all-electron theoretical values as well as
possible.

It is against the spirit of these methods to go back and change the
pseudopotential parameters to adapt solid-state values to a desired
value.  For better or for worse, our calculated solid-state values are
a prediction of our DFT calculations.

Pseudopotential-plane-wave and LAPW are two ways of solving what in
principle is the same equaiton.  Each has sources of error, and one
should attempt to reduce these errors.  But selecting a desired
solid-state outcome is not considered a valid way to use this
approach.

Sincerely,
Andrew M. Rappe

On Tue, Feb 05, 2008 at 06:45:37PM +0100 or thereabouts, Chol-Jun Yu wrote:
> Dear all,
>
> thanks a lot for all helps. Because ABINIT lattice constants for fcc Rh and 
> Pd agreed with experimental values well, I thought their parameters in 
> Opium would be optimized to reproduce the experimental ones, and there 
> would be another trick to produce the LAPW ones.
> It seems to have to consider more extensively to produce the LAPW ones!
>
> Thanks again.
>
> Best wishes,
> Chol-Jun
>
> Andrew M. Rappe wrote:
>> Dear Chol-Jun,
>> In the particular cases you mentioned, such as Rh,
>> the issue is the near-degeneracy of the s and d orbitals,
>> and the importance of achieving norm conservation
>> for a range of electronic configurations.
>> These pseudopotentials were designed with care to
>> reproduce s-d energy splittings and "Tail norms".
>> In fact we wrote a paper about the importance of this:
>> I. Grinberg, N. J. Ramer, and A. M. Rappe, "Quantitative criteria for 
>> transferable pseudopotentials
>> in density functional theory",
>> PRB 63, 201102(R) (2001).
>> I think that all computational methods have some sources of error.  So
>> to get closer agreement between psp and LAPW would require an
>> exhaustive analysis of every source of error in both particular codes.
>> Andrew
>> On Tue, Feb 05, 2008 at 05:26:12PM +0100 or thereabouts, Anglade 
>> Pierre-Matthieu wrote:
>>> Hi,
>>>
>>> Likely, the best would be to ask on the opium mailing list.
>>> The usual way to improve pseudopotentials is to play with all parameters 
>>> in
>>> order to improve the reproduction of the full electron calculation they 
>>> are
>>> based on. So the most obvious thing you can try is to
>>> 1) improve your opium test configurations set to cover more exotic
>>> situations (this may avoid you trying in Abinit some wrong
>>> pseudopotentials);
>>> 2) change the valance of the ionic configuration used to generate the
>>> pseudopotential;
>>> 3) create stiffer pseudopotentials;
>>> 4) play with all other parameters as described in opium tutorials.
>>>
>>> regards
>>>
>>> PMA
>>>
>>> On Feb 5, 2008 4:51 PM, Chol-Jun Yu <yucj@ghi.rwth-aachen.de> wrote:
>>>
>>>
>>>> Dear Marc Torent,
>>>>
>>>> thank you for your reply. Due to some other reasons, I want to and have
>>>> to use norm-conserving PP (using Opium package, optimized RRKJ type)
>>>> rather than PAW. I want to know how to control the parameters in Opium
>>>> code so as to produce the lattice constant and bulk modulus close to
>>>> FP-LAPW.
>>>>
>>>> Best regards,
>>>> Chol-Jun
>>>>
>>>> TORRENT Marc wrote:
>>>>
>>>>> Dear Chol-Jun,
>>>>>
>>>>> You could try with PAW "pseudopotentials" (you can find Rd and Pd in
>>>>> Abinit PAW tables).
>>>>> PAW always gives results to be compared with (FP)LAPW...
>>>>>
>>>>> Marc Torrent
>>>>>
>>>>>
>>>>>
>>>>> Chol-Jun Yu a ?crit :
>>>>>
>>>>>
>>>>>> Dear users,
>>>>>>
>>>>>> I would like to ask about the pseudopotential parameter file of Rh and
>>>>>> Pd. Using the Rh3 and Pd1 GGA
>>>>>> potential,(http://lorax.chem.upenn.edu/Research/psp_gga.html#Rh3) I
>>>>>> obtained the results close to experimental values with respect to the
>>>>>> lattice constant and bulk modulus of their fcc crystals with ABINIT
>>>>>> code (Ecut=30ha, kpoints=(10,10,10), ixc=11:PBE GGA). However, the
>>>>>> results are deviated from the LAPW values (Wien2k);
>>>>>>
>>>>>>        lattice constant (A)  |  Bulk modulus (Gpa)
>>>>>> ----------------------------------------------------
>>>>>>    |  Exp.    LAPW    PP-PW  |  Exp.  LAPW  PP-PW
>>>>>>    -------------------------------------------------
>>>>>> Rh3 |  3.803   3.832   3.806  |  270   260   262
>>>>>> Pd1 |  3.891   3.940   3.880  |  180   174   186
>>>>>>
>>>>>> Is it possible to make the results of PP-PW close to LAPW rather than
>>>>>> experimental values? If possible, could you provide the generating
>>>>>> parameter file.
>>>>>>
>>>>>> Thanks a lot in advance.
>>>>>> Chol-Jun
>
>
> -- 
> Chol-Jun Yu
> Computational Materials Engineering (CME)
> Center for Computational Engineering Science (CCES)
> Institute of Mineral Engineering (GHI)
> RWTH Aachen University
> Mauerstr. 5, D52064 Aachen
>
> email: yucj@ghi.rwth-aachen.de
> phone: 49-0241-80-94989
>