[QE-users] k-point algorithm is not tested yet turbo davidson

Timrov Iurii iurii.timrov at epfl.ch
Sat Dec 21 15:04:53 CET 2019 

Dear Abhirup,


This is a completely different issue. You should open a new thread with a representative title in order to attract attention of the experts on this topic.


I am not an expert in OpenMP, so unfortunately I cannot help you. The only thing which I can suggest is to disable OpenMP and see if it works. Do you really need OpenMP for TDDFPT? Actually, TDDFPT does not have any improvement when OpenMP is enabled (as far as I know).


Greetings,

Iurii


--
Dr. Iurii Timrov
Postdoctoral Researcher
STI - IMX - THEOS and NCCR - MARVEL
Swiss Federal Institute of Technology Lausanne (EPFL)
CH-1015 Lausanne, Switzerland
+41 21 69 34 881
http://people.epfl.ch/265334
________________________________
From: users <users-bounces at lists.quantum-espresso.org> on behalf of Abhirup Patra <abhirupp at sas.upenn.edu>
Sent: Friday, December 20, 2019 6:44:24 PM
To: Quantum ESPRESSO users Forum
Subject: Re: [QE-users] k-point algorithm is not tested yet turbo davidson

Dear Iurii,

I have been trying to use QE-6.5 and the default QE-6.4.1  (compiled by myself with OpenMP) with mpi. It was working fine before, but now both the version of QE crashed without any error message or CRASH and empty output file. But it works fine for the serial version. It seems openmp works perfectly fine with other codes but for QE.

Any idea? It doesn't even start the scf calculation before the TDDFT calculation.

Best,
Abhirup
-------------------------------------------------------------------------------------------------------------------------------------
Abhirup Patra
Postdoctoral Research Fellow
Department of Chemistry
University of Pennsylvania


On Fri, Dec 20, 2019 at 12:32 PM Abhirup Patra <abhirupp at sas.upenn.edu<mailto:abhirupp at sas.upenn.edu>> wrote:
Dear Iurii,

Many thanks. It seems that turbo_eels does not work with nspin = 2. It gives an error:

Error in routine lr_readin (1):
LSDA is not implemented

But it does work without npsin = 2 but with smearing and tot_charge = 0/-1/+1

Best,
Abhirup
-------------------------------------------------------------------------------------------------------------------------------------
Abhirup Patra
Postdoctoral Research Fellow
Department of Chemistry
University of Pennsylvania


On Thu, Dec 19, 2019 at 11:49 AM Timrov Iurii <iurii.timrov at epfl.ch<mailto:iurii.timrov at epfl.ch>> wrote:

Dear Abhirup,


Thank you for clarifications!


Please note that the turbo_eels code works for the noncollinear spin-polarization (noncolin=.true.), and including the spin-orbit coupling (if needed). The turbo_eels code works for the general k-points implementation (k=0 is allowed), but it does not work for the "gamma_only" case (i.e. using tricks to speed up calculations).


The turbo_lanczos and turbo_davidson codes (which are for the absorption spectroscopy) do not work in the spin-polarized case (both collinear and noncollinear) - yes, you are right.


Best,

Iurii


--
Dr. Iurii Timrov
Postdoctoral Researcher
STI - IMX - THEOS and NCCR - MARVEL
Swiss Federal Institute of Technology Lausanne (EPFL)
CH-1015 Lausanne, Switzerland
+41 21 69 34 881
http://people.epfl.ch/265334
________________________________
From: users <users-bounces at lists.quantum-espresso.org<mailto:users-bounces at lists.quantum-espresso.org>> on behalf of Abhirup Patra <abhirupp at sas.upenn.edu<mailto:abhirupp at sas.upenn.edu>>
Sent: Thursday, December 19, 2019 5:40:47 PM
To: Quantum ESPRESSO users Forum
Subject: Re: [QE-users] k-point algorithm is not tested yet turbo davidson

Dear Iurii,

Thanks for the email. Sorry about not explaining it well. Here is what I meant: With spin-polarization and with gamma only kmesh none of the TDDFT code works (i.e. turbo_davidson, turbo_eels, turbo_lanczos). So it is difficult to simulate optical spectra for a "neutral (tot_charge = 0)" doped/defected structure. By "works", I meant that I could run the regular scf-calculation and TDDFT codes for any "non-zero (tot_charge = -1 or +1)" charged structure using "gamma only" kmesh and "non spin-polarization".

Please let me know if it's still confusing.

Best,
Abhirup


-------------------------------------------------------------------------------------------------------------------------------------
Abhirup Patra
Postdoctoral Research Fellow
Department of Chemistry
University of Pennsylvania


On Thu, Dec 19, 2019 at 4:06 AM Timrov Iurii <iurii.timrov at epfl.ch<mailto:iurii.timrov at epfl.ch>> wrote:

Dear Abhirup,


Sorry, I miss the point. Could you clarify please what do you mean by saying "it works"? Do you refer to the k-points implementation of the turbo_davidson code or something else? The k-points algorithm must be disabled for all cases (neutral and charged cases), and if this not the case, then I will modify the code in order to disable it.


Best,

Iurii


--
Dr. Iurii Timrov
Postdoctoral Researcher
STI - IMX - THEOS and NCCR - MARVEL
Swiss Federal Institute of Technology Lausanne (EPFL)
CH-1015 Lausanne, Switzerland
+41 21 69 34 881
http://people.epfl.ch/265334
________________________________
From: users <users-bounces at lists.quantum-espresso.org<mailto:users-bounces at lists.quantum-espresso.org>> on behalf of Abhirup Patra <abhirupp at sas.upenn.edu<mailto:abhirupp at sas.upenn.edu>>
Sent: Wednesday, December 18, 2019 9:36:51 PM
To: Quantum ESPRESSO users Forum
Subject: Re: [QE-users] k-point algorithm is not tested yet turbo davidson

Forgot to mention, it only happens if I do not assign any charge to the structure i.e., for neutral case. However, it works for nonzero tot_charge, spin-unpolarized situation with fixed smearing and turb_davidson works fine.

Thanks for your comments.

Best,
Abhirup
-------------------------------------------------------------------------------------------------------------------------------------
Abhirup Patra
Postdoctoral Research Fellow
Department of Chemistry
University of Pennsylvania


On Wed, Dec 18, 2019 at 3:29 PM Abhirup Patra <abhirupp at sas.upenn.edu<mailto:abhirupp at sas.upenn.edu>> wrote:
Thanks, Iurii. Yes, the spectrum with gamma point is quite reasonable but do not match with the experimental observation. It could be some other issue.

However, I am wondering how to do the TDDFT calculation for a doped/defected structure with extra electron or hole. I tried to use turbo_davidson starting from i) scf calculation with npsin= 2 and total magnetization and turbo_davidson does not work since I believe it does not
for spin-polarized system, then, I tried ii) scf calculation without any spin-polarization and the scf calculation did not run since it cannot work with an extra charge and 'fixed' occupation, and, finally, iii) scf with smearing but turbo_davidson does not support scf calculation with smearing (ref: https://lists.quantum-espresso.org/pipermail/users/2017-January/037363.html)

I believe in this case I cannot use TDDFT of QE to calculate absorption spectra of a defected or doped structure. Am I right here?

Best,
Abhirup


-------------------------------------------------------------------------------------------------------------------------------------
Abhirup Patra
Postdoctoral Research Fellow
Department of Chemistry
University of Pennsylvania


On Wed, Dec 18, 2019 at 4:00 AM Timrov Iurii <iurii.timrov at epfl.ch<mailto:iurii.timrov at epfl.ch>> wrote:

Dear Abhirup,


> However, I am not sure with the number of atoms (~120 or more) GW-BSE calculations with supercell is actually feasible.


Yes, I understand the issue. I do not know if it is feasible. You may try to use e.g. Yambo and see how it goes. Otherwise, since you have already quite a big cell, maybe with k=0 you obtain quite a reasonably converged spectrum (to be checked w.r.t. the cell size)  at the TDDFT level in the adiabatic approximation.


Regards,

Iurii


--
Dr. Iurii Timrov
Postdoctoral Researcher
STI - IMX - THEOS and NCCR - MARVEL
Swiss Federal Institute of Technology Lausanne (EPFL)
CH-1015 Lausanne, Switzerland
+41 21 69 34 881
http://people.epfl.ch/265334
________________________________
From: users <users-bounces at lists.quantum-espresso.org<mailto:users-bounces at lists.quantum-espresso.org>> on behalf of Abhirup Patra <abhirupp at sas.upenn.edu<mailto:abhirupp at sas.upenn.edu>>
Sent: Tuesday, December 17, 2019 4:47:01 PM
To: Quantum ESPRESSO users Forum
Subject: Re: [QE-users] k-point algorithm is not tested yet turbo davidson

Thanks very much, Iurii. I really appreciate your answer and it does make sense to me. With gamma point only it works fine. And, you are  right, my system is a solid and I am
trying to simulate the optical spectrum. However, I am not sure with the number of atoms (~120 or more) GW-BSE calculations with supercell is actually feasible.

Best,
Abhirup
-------------------------------------------------------------------------------------------------------------------------------------
Abhirup Patra
Postdoctoral Research Fellow
Department of Chemistry
University of Pennsylvania


On Tue, Dec 17, 2019 at 7:50 AM Timrov Iurii <iurii.timrov at epfl.ch<mailto:iurii.timrov at epfl.ch>> wrote:

Dear Abhirup,


As you can see from the error message of the turbo_davidson code, the k-points algorithm is not supported. Only Gamma point sampling can be used.


If I understand correctly, your system is a solid, right? And you want to use a unit cell with a k-points mesh, right? The turbo_davidson cannot be used in this case. You may try to use a supercell with a Gamma point sampling. But be careful when using TDDFT in the adiabatic approximation for the absorption spectroscopy of solids. I recommend that you check literature on this topic. Maybe you will need to resort to the BSE approach.


HTH


Iurii Timrov


--
Dr. Iurii Timrov
Postdoctoral Researcher
STI - IMX - THEOS and NCCR - MARVEL
Swiss Federal Institute of Technology Lausanne (EPFL)
CH-1015 Lausanne, Switzerland
+41 21 69 34 881
http://people.epfl.ch/265334
________________________________
From: users <users-bounces at lists.quantum-espresso.org<mailto:users-bounces at lists.quantum-espresso.org>> on behalf of Abhirup Patra <abhirupp at sas.upenn.edu<mailto:abhirupp at sas.upenn.edu>>
Sent: Monday, December 16, 2019 6:12:08 PM
To: Quantum ESPRESSO users Forum
Subject: [QE-users] k-point algorithm is not tested yet turbo davidson

Hello,

I am trying to use turbo_davidson starting from a SCF calculation using PBE without smearing and spin-polarization. This system is a non-metal with bandgap of 2.6 eV. I gave uniform k-mesh as follows for the bcc structure with lattice constant of 10.34 angstrom.

 K_POINTS {crystal_b}
 9
0.0 0.0 0.0 0
0.0 -0.5 0.0 0
0.0 -0.5 0.5 0
0.5 -0.5 0.5 0
0.5 -0.5 0.0 0
0.5 0.0 0.0 0
0.0 0.0 0.0 0
0.0 0.0 0.5 0
0.5 0.0 0.5 0

&lr_input
    prefix = 'pristine_pbe'
    outdir = './'
!    wfcdir = './'
/
&lr_dav
  num_eign = 35,
  num_init = 70,
  num_basis_max = 200,
  start = 0.0d0
  finish = 3.5d0
  step = 0.001d0
  broadening = 0.005d0
/

However, I am getting the following error:

     Error in routine lr_readin (1):
     k-point algorithm is not tested yet

I am wondering if there is a simple way out to this problem other than explicitly testing different k mesh combinations.

Best,
Abhirup
-------------------------------------------------------------------------------------------------------------------------------------
Abhirup Patra
Postdoctoral Research Fellow
Department of Chemistry
University of Pennsylvania
_______________________________________________
Quantum ESPRESSO is supported by MaX (www.max-centre.eu/quantum-espresso<http://www.max-centre.eu/quantum-espresso>)
users mailing list users at lists.quantum-espresso.org<mailto:users at lists.quantum-espresso.org>
https://lists.quantum-espresso.org/mailman/listinfo/users
_______________________________________________
Quantum ESPRESSO is supported by MaX (www.max-centre.eu/quantum-espresso<http://www.max-centre.eu/quantum-espresso>)
users mailing list users at lists.quantum-espresso.org<mailto:users at lists.quantum-espresso.org>
https://lists.quantum-espresso.org/mailman/listinfo/users
_______________________________________________
Quantum ESPRESSO is supported by MaX (www.max-centre.eu/quantum-espresso<http://www.max-centre.eu/quantum-espresso>)
users mailing list users at lists.quantum-espresso.org<mailto:users at lists.quantum-espresso.org>
https://lists.quantum-espresso.org/mailman/listinfo/users
_______________________________________________
Quantum ESPRESSO is supported by MaX (www.max-centre.eu/quantum-espresso<http://www.max-centre.eu/quantum-espresso>)
users mailing list users at lists.quantum-espresso.org<mailto: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/20191221/542c8174/attachment.html>

More information about the users mailing list