[Wien] 'SECLR4' - POTRF failed

Michael Gurnett michael.gurnett at kau.se
Thu Sep 15 19:29:11 CEST 2005


What R-MT*K-MAX value are you using. Seems strange (atleast in my experiance) that Rmt values larger than 2 would require NMATMAX of 20000 for 64 atoms. 

Michael 
  ----- Original Message ----- 
  From: Moon, Chang-Youn 
  To: A Mailing list for WIEN2k users 
  Sent: Thursday, September 15, 2005 7:08 PM
  Subject: RE: [Wien] 'SECLR4' - POTRF failed


  Well, I found similar prescription like yours on Wien2k user's guide, but I'm not sure if this is the case for my situation. My last Rmt's I tried are 2.5 a.u. for In, 2.2 for P, and a value in between for Fe. Actually I also tried 2.3 for In and 2.1 or 2.2 for P, but the problem seemed to persist..

   

  In the meanwhile, after I increased NMATMAX to 20000, finally lapw1 ran properly without errors for spin up, but again, the similar error occurred for spin down in this time. So I wonder if I should adhere to increasing NMATMAX, or there is another way, more relevant, to solve the problem, like adjusting Rmt's, etc. 

   


------------------------------------------------------------------------------

  From: wien-bounces at zeus.theochem.tuwien.ac.at [mailto:wien-bounces at zeus.theochem.tuwien.ac.at] On Behalf Of Michael Gurnett
  Sent: Thursday, September 15, 2005 10:34 AM
  To: A Mailing list for WIEN2k users
  Subject: Re: [Wien] 'SECLR4' - POTRF failed

   

  I have found that errors like these can be caused by Rmts that differ too much. What Rmt values are you using

   

  Michael

    ----- Original Message ----- 

    From: Moon, Chang-Youn 

    To: wien at zeus.theochem.tuwien.ac.at 

    Sent: Thursday, September 15, 2005 6:04 PM

    Subject: [Wien] 'SECLR4' - POTRF failed

     

    Dear Wien2k users,

     

    I'm trying to perform defect calculations using wien2k (version 3, released on Jan 18, 2003). My systems are III-V compound semiconductors containing a Fe atom as a defect, 64-atom cubic supercell. I had no problems with GaAs, GaP, and AlAs, but for InP,  I met an error during running lapw1, as following:

     

    Cholesky INFO = 4053

    'SECLR4' - POTRF (Scalapack/LAPACK) failed.

     

    in uplapw1_1.error file.

     

    I run the spin-polarized calculation k-point parallel with 4 k points, and this error occur only with the gamma point, and for other k-points there's no problem. So I guess the error is for executable lapw1 (for real matrices), but I can't figure out exactly what I'm doing wrong. After I referred to some websites about the potrf  subroutine in LAPACK library, I  understand this error is related to not positive definite matrix element (for above case, 4053rd matrix element). The actual number of Cholesky INFO in the error file depends on muffin-tin radius of constituent atoms, the separation energy for semi-core and valence states, etc. Moreover, after I realized that I had used too small NMATMAX value in param.inc file for lapw1 (my RKMAX had been reduced to smaller values than the one I originally set to), I increased NMATMAX value to 10000 and compiled again, then now lapw1 runs longer than before, but eventually the same error occurs with larger Cholesky INFO value than before. 

     

    What should I do? Shall I just increase NMATMAX further?  Or is 64 atom calculation too big for wien2k?

     

    Thanks,

     

    Chang-Youn Moon

    National Renewable Energy Laboratory

     


----------------------------------------------------------------------------

    _______________________________________________
    Wien mailing list
    Wien at zeus.theochem.tuwien.ac.at
    http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien



------------------------------------------------------------------------------


  _______________________________________________
  Wien mailing list
  Wien at zeus.theochem.tuwien.ac.at
  http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20050915/5f3aa8df/attachment.html


More information about the Wien mailing list