[QE-users] hp.x - appropriate way to handle single dopant

Wed May 22 12:30:04 CEST 2024

Of course, I fully agree with you !Either you go with a 64 atoms cell (3.125 % doping) calculation which could take several weeks ... or just assume that the parameters obtained from 8 atoms unit cell (25% doping) are good enough and use them in the super cell.
For the U parameter, it could be safe to adopt the second strategy, since it is an on-site parameter. For the inter-site V, however, the local environment (the neighboring atoms) is important.  

A different method is hybrid functionals which are known to be accurate and can capture very well the missed physics.You can for example try HSE 06. One can starts with pure ZnS to see the fraction of the EXX (HF exchange) needed to match the well known experimental results, and then use this fraction in the doped system.
This is just a suggestion... I can't say, but I guess that the calculation time would less than the few weeks of DFT+U+V.
Best 
AH  

   On Wednesday, 22 May 2024 at 08:59:39 am GMT+1, Abdul Muhaymin <abdul.muhaymin at bilkent.edu.tr> wrote:  

Dear Dr. HOUARI,

 Thank you for your reply. I did not correct the band gap in any system. I am hoping if I apply some U parameter to Zn-3d, Co-3d and maybe to S-3p, I can capture the physics better. In this pursuit, some V parameter might also be an option. But this is not the problem I am facing. I am facing problem on how to obtain these U and V values. For undoped, periodic system, we can get the value using hp.x with simple unit cell. But the problem is for aperiodic doped system like the one I have. 

 As you said, the calculated U from the 8 atoms ZnS unit cell is probably okay to use for 64 atoms supercell. But if I dope the unit cell (to get the U value of Co), this system (25% doping) is much different than the 64 atoms supercell (3.125% doping). And when I am trying to run hp.x directly for the 64 atoms (U applied only to the single Co), only a single iteration of 1 q point takes an hour using 2240 CPUs. So, it would take weeks to complete a calculation! 

 I am hoping maybe I can calculate the U parameter separately for Co unit cell (i.e., Co-HCP) and then maybe for Zn and S from ZnS unit cell. But I have no rationale to support that it should work. And also the inter-site parameters cannot be obtained following this. 

 Sincerely,
 Abdul

 On 5/21/2024 7:03 PM, Abdesalem Houari wrote:

  Dear Abdul Muhaymen, 
  You say that your results are fine, except the band gap ! How did you correct it in pure ZnS, before Co-doping ? 
  The most commun way in DFT is hybrid functionals (like HSE 06), which obviously are very demanding in computational cost.  So the DFT+U (+V) could be a nice alternative. Here I think you might need on-site U not only for Co, but also for Zn (and as you said may be S). Since ZnS is a band insulator and covalent compound, the inter-site V could play an important role. 
  In principle U and V parameters are neither transferable nor universal, but in your case ( calculated U from 8 atoms unit cell to use in 64 atoms supercell), I guess it should be OK !

  Best regards

   =====================================
  Dr. Abdesalem HOUARI -------------------------------------------------------------------------------------------
 Department of physics, Theoretical Physics Laboratory
 University of Bejaia-06000. Algeria.
 E-mail: abdeslam.houari at univ-bejaia.dz & habdslam at yahoo.fr
 https://sites.google.com/site/houariabdeslam/homepage
 ===================================

      On Tuesday, 21 May 2024 at 04:17:30 am GMT+1, Abdul Muhaymin via users <users at lists.quantum-espresso.org> wrote:  

   Hello all,

  I am investigating single TM dopant in wide band gap semiconductors such 
  as Co in ZnS. I am using a 64 atoms supercell where I replaced one of 
  the Zn atom with a Co atom. I tested several convergence with respect to 
  the supercell size. My results seem fine except the band gap. Now I want 
  to apply the U correction to that Co-3d (and maybe to S-2p). For this, 
  is it possible to use the unit cell (8 atoms) and run hp.x to get the U 
  values? Or do I have to run hp.x with the large supercell (2*2*2 unit 
  cell=64 atoms)?

  Also, at the beginning of our studies, during the structural relaxation 
  phase, we first found our lattice parameter for the host semiconductor 
  from multiple scf calculations and subsequently running an eos analysis 
  (ev.x). Then we ran relax calculation but not vc-relax. We keep this 
  lattice parameter constant and when introducing new dopants, we only 
  vary the atomic positions (calculation='relax'). In this case, when 
  running hp.x, could we replace the vc-relax calculation with relax 
  calculation to self-consistently get the U values?

  Thanks,
  Abdul Muhaymin
  Graduate (MS) student, Materials Science and Nanotechnology
  Bilkent University, Ankara.

  _______________________________________________
  The Quantum ESPRESSO community stands by the Ukrainian
  people and expresses its concerns about the devastating
  effects that the Russian military offensive has on their
  country and on the free and peaceful scientific, cultural,
  and economic cooperation amongst peoples
  _______________________________________________
  Quantum ESPRESSO is supported by MaX (www.max-centre.eu)
  users mailing list users at lists.quantum-espresso.org
  https://lists.quantum-espresso.org/mailman/listinfo/users

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20240522/652a5c12/attachment.html>