[Q-e-developers] on the calculation of Hubbard U values

[Timrov Iurii]

Dear All,


There were some discussions outside of the QE developers mailing list about this issue, and we would like to share with you our thoughts. José proposed his changes to the files set_hubbard_l.f90, set_hubbard_n.f90, tabd.f90, where all elements (we haven't checked that really all) are tabulated with the corresponding channel to be effected by the Hubbard correction. We think we should not hurry and not include this in QE 6.2 which is going to be released very soon (as far as we know).


The main problem is that some pseudos may not comply with the default choice of states made in the aforementioned routines. For example, for La, in José's version of set_hubbard_l.f90, it is specified that hubbard_l = 3 (i.e. the f state). While some pseudos include empty 4f states (e.g. Pslibrary, GBRV), some pseudos do not (e.g. Dojo library, La.pbe-nsp-van.UPF from the QE web site). So, in the latter case the code will crash because the f channel is not present in the pseudo. Thus, we have to be careful. As Paolo and we discussed, it  is probably better to have a possibility to specify hubbard_l from the input (e.g. if you choose the pseudo which contains 4f for La and you want to apply a Hubbard correction to 4f, and not 5d, then just specify hubbard_l = 3 from the input; if nothing is specified in the input, the code can use the default value which should work for all pseudos, e.g. hubbard_l =2 for La). Another option might be to keep as the default hubbard_l  = 3 for La, and if the code does not find the f state in the pseudo then stop smoothly and ask the user to change hubbard_l by specifying it in the input. As we said, probably it would be safer to implement this things after the release of QE 6.2 in order to avoid possible bugs.


Best regards,

Iurii and Matteo

Swiss Federal Institute of Technology Lausanne (EPFL)
Laboratory of Theory and Simulation of Materials (THEOS)
CH-1015 Lausanne, Switzerland
+41 21 69 34 881

________________________________
From: q-e-developers-bounces at qe-forge.org <q-e-developers-bounces at qe-forge.org> on behalf of Paolo Giannozzi <p.giannozzi at gmail.com>
Sent: Thursday, October 12, 2017 7:20 PM
To: General discussion list for Quantum ESPRESSO developers
Cc: jcconesa at icp.csic.es
Subject: Re: [Q-e-developers] on the calculation of Hubbard U values

I also think that a better solution would be providing parameters of Hubbard U from input. Something like a "Hubbard_U" card with
Mn(3d)=4.0
O(2p)=5.0
...

Paolo

On Thu, Oct 12, 2017 at 1:50 PM, Timrov Iurii <iurii.timrov at epfl.ch<mailto:iurii.timrov at epfl.ch>> wrote:

Dear Jose,


Thanks for the remark, we will keep this in mind when working on the porting of the DFpT code to compute U to the official release of QE. Perhaps it would be useful to introduce input variables which allow user to specify which atomic manifold to associate with the Hubbard correction.


Best regards,

Iurii


--
Dr. Iurii Timrov
Postdoctoral Researcher
Swiss Federal Institute of Technology Lausanne (EPFL)
Laboratory of Theory and Simulation of Materials (THEOS)
CH-1015 Lausanne, Switzerland
+41 21 69 34 881
http://people.epfl.ch/265334
________________________________
From: q-e-developers-bounces at qe-forge.org<mailto:q-e-developers-bounces at qe-forge.org> <q-e-developers-bounces at qe-forge.org<mailto:q-e-developers-bounces at qe-forge.org>> on behalf of José C. Conesa <jcconesa at icp.csic.es<mailto:jcconesa at icp.csic.es>>
Sent: Thursday, October 12, 2017 1:29 PM
To: q-e-developers at qe-forge.org<mailto:q-e-developers at qe-forge.org>
Subject: [Q-e-developers] on the calculation of Hubbard U values


Dear QE developers,

I have seen in the last Psi-K highlight about the advances in the QE code that version 6.2 (now at a release candidate status) will allow computing ab initio U values for the DFT+U scheme using DFpT, so that supercells would not be needed and the process would be much easier. Then, if one wants to find U values considering the response not only for a specific element but for all of them present in the material, i.e. computing the complete (inverse) susceptibility matrices, it is necessary that Hubbard parameters be considered for all elements in the material. But in its present version, the code in files Modules/set_hubbard_l.f90 Modules/set_hubbard_n.f90 and PW/src/tabd.f90 only allows using Hubbard parameters for part of the elements in the periodic table.

For this reason I wish to make you a request: please include in those files, in the final qe-6.2 version, all elements in the periodic table (well, you may omit noble gases) so that each user does not need to modify those files, and compile the executables again, every time that (s)he has to consider a new element in the material for which U values should be computed. This should not be difficult to do. I am aware that in some cases it may be not clear which atomic subshell should be affected by the Hubbard correction; for example, La might need to do it for d or f orbitals depending on the compound, and In or Zn might be in a similar problem for p and d orbitals. This could be solved just by adding a word of caution in the manual.

Hoping that this modification may be made,

Best wishes,

--
José C. Conesa
Instituto de Catálisis y Petroleoquímica, CSIC
Marie Curie 2, Cantoblanco
28049 Madrid, Spain
Tel. (+34)915854766<tel:+34%20915%2085%2047%2066>

_______________________________________________
Q-e-developers mailing list
Q-e-developers at qe-forge.org<mailto:Q-e-developers at qe-forge.org>
http://qe-forge.org/mailman/listinfo/q-e-developers




--
Paolo Giannozzi, Dip. Scienze Matematiche Informatiche e Fisiche,
Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
Phone +39-0432-558216, fax +39-0432-558222

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/developers/attachments/20171013/9725ac2b/attachment.html>