[Wien] Coult not reproduced results of CoSb3

Tue May 16 08:31:35 CEST 2017

My earlier remark (probably it appears more than once or in different ways in the mailing list)
means that you cannot use the Engel-Vosko (EV) based functional for optimitzation of the crystal structure.

If you read the paper of Engel and Vosco, you find that they optimized the exchange part of the functional.
The method they used for that is the optimized potential model that is sometimes also called exact exchange method.
If you check the contributions to the energy you will find that the EV contributes a much different amount of exchange energy Ex to the exchange-correlation energy Exc
compared to other GGA's say Perdew-Burke-Ernzerhof (PBE).

Note that (nearly) all exchange-correlation functionals have an exchange and a correlation part.
(There might be methods where one or the other is not needed to be calculated, but that's a different story)
Exception appear if the two parts cannot be divided from each other. Examples may be the Slater X_alpha fuctional or the Kohn-Sham functional that is sometimes called an exchange-only functional.
If you read  the paper of Engel and Vosco carefully, you will find that the latter, the Kohn-Sham xc functional, still enters 
the exchange potential and energy (v_x and E_x) of the Engel Vosko GGA as v^LDA, E^LDA.
Indeed, one can use the Engel-Vosko functional as an exchange-only functional, however, in the old option 15 it was used together with the Perdew-Wang correlation part
and thus the results will be different if you compare the "exchange-only" calculation to an "exchange-correlation" calculation.

Note: The EV93 + PW91 functional may lead to strange magnetic states in spin polarized calculations.
I use it only for semiconductors  to "correct" the size of the band gap, as it is as a simple GGA much faster than other methods.

For more you should always read and understand the original papers on the different functionals before using them.

Ciao
Gerhard

DEEP THOUGHT in D. Adams; Hitchhikers Guide to the Galaxy:
"I think the problem, to be quite honest with you,
is that you have never actually known what the question is."

====================================
Dr. Gerhard H. Fecher
Institut of Inorganic and Analytical Chemistry
Johannes Gutenberg - University
55099 Mainz
and
Max Planck Institute for Chemical Physics of Solids
01187 Dresden
________________________________________
Von: Wien [wien-bounces at zeus.theochem.tuwien.ac.at] im Auftrag von fatima DFT [fatimadft5 at gmail.com]
Gesendet: Montag, 15. Mai 2017 14:49
An: A Mailing list for WIEN2k users
Betreff: Re: [Wien] Coult not reproduced results of CoSb3

1. For numeric options:

Let me do a more careful calculation and will update.

2. For EX_EV93 EC_NONE VX_EV93 VC_NONE:

I read one of your thread<http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/msg03993.html> where you mentioned that:  Note that the Engel-Vosko GGA is not designed for total energies but optimized for exact exchange only.

Then I looked to UG and in table 7.4 and 7.6 where I got the information for EX_xx and VX_xx only. So I considered EC_xx and VC_xx should be EC_NONE and VC_NONE, respectively.

Regards
Fatima

On Mon, May 15, 2017 at 5:22 PM, Fecher, Gerhard <fecher at uni-mainz.de<mailto:fecher at uni-mainz.de>> wrote:
I did a fast test with Si, and I have clearly different band gaps when switching between PBE (option 13) and  EV exchange parts of the GGA with different Correlation parts (15, 25, 26, or the corresponding EX, EC, VX, VC switches)

It seems to me that you do not realy know what you are doing when you are switching to different exchange-correlation functionals.

Who told you that EX_EV93 EC_NONE VX_EV93 VC_NONE  is supposed for EV approach ?

Georg Madsen pointed in his paper only on (E. Engel, S.H. Vosko, Phys. Rev. B 47 (1993) 13164–13174.) but there is no hint on the correlation part,
from the cited Engel Vosko paper one expects that this was taken from the Perdew Wang 1991 GGA, however, there are also hints on GEA.
Option 15 corresponds to the Engel Vosko exchange part combined with Perdew Wang 1991 correlation part
          25 corresponds to the Engel Vosko exchange part combined with the LSDA correlation term
          26 to a combination of 15 and 25 (roughly speaking).

PS.: I would wonder if for CoSb3 all different functionals would give the same result as PBE.

Ciao
Gerhard

DEEP THOUGHT in D. Adams; Hitchhikers Guide to the Galaxy:
"I think the problem, to be quite honest with you,
is that you have never actually known what the question is."

====================================
Dr. Gerhard H. Fecher
Institut of Inorganic and Analytical Chemistry
Johannes Gutenberg - University
55099 Mainz
and
Max Planck Institute for Chemical Physics of Solids
01187 Dresden
________________________________________
Von: Wien [wien-bounces at zeus.theochem.tuwien.ac.at<mailto:wien-bounces at zeus.theochem.tuwien.ac.at>] im Auftrag von fatima DFT [fatimadft5 at gmail.com<mailto:fatimadft5 at gmail.com>]
Gesendet: Montag, 15. Mai 2017 13:04
An: A Mailing list for WIEN2k users
Betreff: Re: [Wien] Coult not reproduced results of CoSb3

Sir, See my pointwise answer:

Did you save case.in0 after you changed to option 15 ?

Yes, I saved after case.in0 after replacing PBE with 15.

and did you NOT run again the initialisation (for PBE) after you changed to option 15 in case.in0 ?

I did not run init_lapw again after replacing PBE with 15.

If you use the new switches EX_... etc, and the results differ from earlier work, may be you have to set to the same correlation energy and potential  (EC_xxx, VC_xxx) , that was used by Madsen

Sir, I do not know which switched Pof. Madsen used for EC_xx and VC_xx. It is not mentioned in the paper so simpley I used NONE. Could you please tell me what switched I should use with EC/VC_xxx?

Regards
Fatima

Ciao
Gerhard

DEEP THOUGHT in D. Adams; Hitchhikers Guide to the Galaxy:
"I think the problem, to be quite honest with you,
is that you have never actually known what the question is."

====================================
Dr. Gerhard H. Fecher
Institut of Inorganic and Analytical Chemistry
Johannes Gutenberg - University
55099 Mainz
and
Max Planck Institute for Chemical Physics of Solids
01187 Dresden
________________________________________
Von: Wien [wien-bounces at zeus.theochem.tuwien.ac.at<mailto:wien-bounces at zeus.theochem.tuwien.ac.at><mailto:wien-bounces at zeus.theochem.tuwien.ac.at<mailto:wien-bounces at zeus.theochem.tuwien.ac.at>>] im Auftrag von fatima DFT [fatimadft5 at gmail.com<mailto:fatimadft5 at gmail.com><mailto:fatimadft5 at gmail.com<mailto:fatimadft5 at gmail.com>>]
Gesendet: Montag, 15. Mai 2017 11:56
An: A Mailing list for WIEN2k users
Betreff: Re: [Wien] Coult not reproduced results of CoSb3

Sir, I did not test the options 25 and 26 but I tested option 15 which is exactly same as PBE. So, the option 15 is not the correct option.
 I initialised the case with option 15. when I saw case.in0 file PBE was printed there. I replaced that PBE with "15" and calculated further properties.
In another directory, I ran the case with PBE option and compared results obtained from both and both are exactly same.

Option EX_EV93 EC_NONE VX_EV93 VC_NONE gave me quite different results but then mu results are not in agreement with the reported in the test case of Boltztrap.

Please correct me where I have mistaken? My thermoelectric properties are quite different that reported in test case.

Warm regards
Fatima

On Mon, May 15, 2017 at 12:24 PM, Fecher, Gerhard <fecher at uni-mainz.de<mailto:fecher at uni-mainz.de><mailto:fecher at uni-mainz.de<mailto:fecher at uni-mainz.de>><mailto:fecher at uni-mainz.de<mailto:fecher at uni-mainz.de><mailto:fecher at uni-mainz.de<mailto:fecher at uni-mainz.de>>>> wrote:
If I read the code correctly then options 15, 25, and 26 should still be related to the functional of Engel and Vosco

Ciao
Gerhard

DEEP THOUGHT in D. Adams; Hitchhikers Guide to the Galaxy:
"I think the problem, to be quite honest with you,
is that you have never actually known what the question is."

====================================
Dr. Gerhard H. Fecher
Institut of Inorganic and Analytical Chemistry
Johannes Gutenberg - University
55099 Mainz
and
Max Planck Institute for Chemical Physics of Solids
01187 Dresden
________________________________________
Von: Wien [wien-bounces at zeus.theochem.tuwien.ac.at<mailto:wien-bounces at zeus.theochem.tuwien.ac.at><mailto:wien-bounces at zeus.theochem.tuwien.ac.at<mailto:wien-bounces at zeus.theochem.tuwien.ac.at>><mailto:wien-bounces at zeus.theochem.tuwien.ac.at<mailto:wien-bounces at zeus.theochem.tuwien.ac.at><mailto:wien-bounces at zeus.theochem.tuwien.ac.at<mailto:wien-bounces at zeus.theochem.tuwien.ac.at>>>] im Auftrag von fatima DFT [fatimadft5 at gmail.com<mailto:fatimadft5 at gmail.com><mailto:fatimadft5 at gmail.com<mailto:fatimadft5 at gmail.com>><mailto:fatimadft5 at gmail.com<mailto:fatimadft5 at gmail.com><mailto:fatimadft5 at gmail.com<mailto:fatimadft5 at gmail.com>>>]
Gesendet: Samstag, 13. Mai 2017 23:18
An: A Mailing list for WIEN2k users
Betreff: [Wien] Coult not reproduced results of CoSb3

Dear Sir

I tried to reproduce the case.trace and case.condtens provide with Boltztrap test case.
I took CoSb3.struct<https://sites.google.com/site/dftfatima5/letter/mailing-list-queries/CoSb3_ex.struct> file and run it with  EX_EV93 EC_NONE VX_EV93 VC_NONE (which is supposed for EV approach), RKMax=5.5, numk 300 shifted, non-sp,  -ecut -6 Gmax=12,   and then after scf I took  24000 non-shifted mesh after as mentioned in the calculation part of the original paper of Boltztrap: Madsen et al. Commun. 2006 under section: 3.1.2. Test case: CoSb3.

scf was run using:  run_lapw -p -i 300 -ec 0.00001 -cc 0.0001

I see in the case of CoSb3 the authors reported FER: 0.55475(Ry) for a thermoelectric part but here I am getting 0.6497801651 with a gap 0.0554 Ry.
.
A complete scf run is here: CoSb3_15.tar.gz<https://sites.google.com/site/dftfatima5/letter/mailing-list-queries/CoSb3_15.tar.gz>,

I also tested scf with option 15 (EV-GGA as suggested by Prof Fecher Fecher<http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/msg04252.html> and Dr. <http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/msg03990.html> Tran<http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/msg03990.html>  got FER:  0.4971964326,  GAP 0.0443(Ry). If I run init_lapw for option 15 then it runs for PBE-GGA while prof Fecher and Dr. Tran suggested option 15 for EV-GGA.

Could you please correct me where I did a mistake?

Warm regards

Fatima
_______________________________________________
Wien mailing list
Wien at zeus.theochem.tuwien.ac.at<mailto:Wien at zeus.theochem.tuwien.ac.at><mailto:Wien at zeus.theochem.tuwien.ac.at<mailto:Wien at zeus.theochem.tuwien.ac.at>><mailto:Wien at zeus.theochem.tuwien.ac.at<mailto:Wien at zeus.theochem.tuwien.ac.at><mailto:Wien at zeus.theochem.tuwien.ac.at<mailto:Wien at zeus.theochem.tuwien.ac.at>>>
http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
SEARCH the MAILING-LIST at:  http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html

_______________________________________________
Wien mailing list
Wien at zeus.theochem.tuwien.ac.at<mailto:Wien at zeus.theochem.tuwien.ac.at><mailto:Wien at zeus.theochem.tuwien.ac.at<mailto:Wien at zeus.theochem.tuwien.ac.at>>
http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
SEARCH the MAILING-LIST at:  http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html

_______________________________________________
Wien mailing list
Wien at zeus.theochem.tuwien.ac.at<mailto:Wien at zeus.theochem.tuwien.ac.at>
http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
SEARCH the MAILING-LIST at:  http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html