The ABINIT Users Mailing List ( CLOSED )

[lan haiping <lanhaiping@gmail.com>]

Dear Don,

Thanks for this detailed explanation.  
It seems that i came to  problems for SCQP parallel calulation too. I just found that the  results of parallel executation is quite different from serial results.  
My primary examination found this difference may be due to overflow errors in epsilon calculations. The version i used is 5.7.3. But when i turn to 5.6.3, both serial and parallel executation give the same result.  is there any change  due to MLWFs implementation for SCQP in 5.7.3 ?

Another question is about the criterion for SCQP calculation.  I browsed the all variables for GWA calculations, and didnot find any parameter to give the convergence  thread for SCQP calculations.  Therefore, i wonder how can we do judgement for SCQP calculations ?   My test on silicon with SCQP calculation give me  1.53 eV for GW gap after two iterations. This result is quite larger than the experimental value and G0W0 result.  Is it due to no enough iterations in this test calculation ? Therefore, how can i find the proper settings for the number of iterations ?
Thanks again for your time and kind help.
Best,
Hai-Ping

On Mon, Mar 23, 2009 at 12:32 AM, D. R. Hamann <drhamann@mat-simresearch.com> wrote:

Dear Hai-Ping,

The wannier90 code which supplies the library routines is strictly a sequential code, so you should do your GW in parallel, and then do the wannier function run as a separate sequentail run.  We probably need a test to stop things earlier, as well as clearer documentation.

Best regards,
Don

lan haiping wrote:

Dear Don,

I tried to run your  GW+MLWFs scheme with $abinit/tests/wannier90/Input/t03.in <http://t03.in> parallelly, but it always failed after  GW calculations and gave  error information below

" mlwfovlp_qp : ERROR
 orthonormality error for quasiparticle wave functions.
 spin=   1  k point=     5  ortho_err= 2.6E+02 >1E-6
 Action : Be sure input nband>=maxval(bndgw)
"
But i can finish this test by running serially.
I also try to  perform this test in another workstation, and it gave me the same behavior. The version i used for tests is 5.7.3, and the compilor is ifort 10.1 with open-mpi, would you please give me some comments and advices for this test ?

Regards,
Hai-Ping

On Fri, Feb 6, 2009 at 10:15 PM, D. R. Hamann <drhamann@mat-simresearch.com <mailto:drhamann@mat-simresearch.com>> wrote:



   -------- Original Message --------
   Subject:    Re: [abinit-forum] generate the GW band structure
   Date:       Fri, 06 Feb 2009 09:14:03 -0500
   From:       D. R. Hamann <drhamann@mat-simresearch.com>

   <mailto:drhamann@mat-simresearch.com>
   To:         yelinhui@hotmail.com <mailto:yelinhui@hotmail.com>
   References:         <sympa.1233887840.2684.463@abinit.org>

   <mailto:sympa.1233887840.2684.463@abinit.org>



   Dear Linhui,

   For documentation of the methods, see D. R. Hamann and D.
   Vanderbilt, Phys. Rev. B *79*, 045109 (2009).  An example with
   reasonably good notes in 5.7.2 is tests/wannier90/Input/t03.in

   <http://t03.in>.  Also see the README, and be sure to go through

   the wannier90 tutorial (LDA-based) first.


   Don Hamann

   yelinhui@hotmail.com <mailto:yelinhui@hotmail.com> wrote:

   Thanks for the reply. Is there any document on how to do this? Or maybe an
   example input?

   Linhui

     

   --    D. R. Hamann
   Mat-Sim Research LLC    | Department of Physics
   P.O. Box 742            |  and Astronomy
   Murray Hill, NJ 07974   | Rutgers University
   phone: 908-370-8079     | 732-445-5500, x4381

   email: drhamann@mat-simresearch.com <mailto:drhamann@mat-simresearch.com>

       

   --    D. R. Hamann
   Mat-Sim Research LLC    | Department of Physics
   P.O. Box 742            |  and Astronomy
   Murray Hill, NJ 07974   | Rutgers University
   phone: 908-370-8079     | 732-445-5500, x4381

   email: drhamann@mat-simresearch.com <mailto:drhamann@mat-simresearch.com>

       



--
Hai-Ping Lan
Department of Electronics ,
Peking University , Bejing, 100871

lanhaiping@gmail.com <mailto:lanhaiping@gmail.com>, hplan@pku.edu.cn <mailto:hplan@pku.edu.cn>

--

D. R. Hamann
Mat-Sim Research LLC    | Department of Physics
P.O. Box 742            |  and Astronomy
Murray Hill, NJ 07974   | Rutgers University
phone: 908-370-8079     | 732-445-5500, x4381

email: drhamann@mat-simresearch.com

-- 
Hai-Ping Lan 
Department of Electronics ,
Peking University , Bejing, 100871
lanhaiping@gmail.com, hplan@pku.edu.cn