The ABINIT Users Mailing List ( CLOSED )

[Masayoshi Mikami <mmikami@yk.rim.or.jp>]

Dear Nichols,

Sorry to have the splitted e-mail, but I just had to add:
- Do you have well-converged DEN/WFK as well as geometry
  before RF-calculation ? 

- "isecur" has no meaning for RF-calculation:
  http://www.abinit.org/ABINIT/Infos_v4.3/vardev.html#isecur

If Gaussian is available, it might also be interesting to 
see the difference (numerical difference, CPU-time/resource, ...)?!

Good continuation,
Masayoshi

From: Masayoshi Mikami <mmikami@yk.rim.or.jp>
Subject: Re: [abinit-forum] response function calculations
Date: Fri, 04 Jun 2004 07:47:49 +0900
Message-ID: <20040604074749D.mmikami@yk.rim.or.jp>

> Dear Nichols,
> 
> diemac:
> "For molecules in an otherwise empty big box, try 1.5 ... 3.0"
> ( http://www.abinit.org/ABINIT/Infos_v4.3/vargs.html#diemac )
> 
> diemix:
> "For molecules, a value on the order 0.5 or 0.33 is rather usual."
> ( http://www.abinit.org/ABINIT/Infos_v4.3/vargs.html#diemix )
> 
> Still, I think that some trials and errors might be needed 
> for dimiex/diemac for your system ...
> 
> Regarding parallel calculation, kindly see 
> http://www.abinit.org/ABINIT/Infos/Features/features.html
> and, in particular, "B. Speed in the parallel version (B.2)."
> 
> In your case, nkpt = 1/non-spin-polarization (implicitly),
> so parallization of bands (MPI) or FFT (OpenMP) might be tried.
> 
> In passing, I might have comments after just skimming the input:
> - typat may be set as "28*1 2"
> - k-point sampling may be carefully tested for such a cluster.
>   If more than one k-points could not be practical on your platform,
>   still you might want to try Chadi-Cohen's special k-point for 
>   the Simple Cubic, i.e. (1/4,1/4,1/4). So the comparison with 
>   kpt=(0,0,0),(1/4,1/4,1/4) and (1/2,1/2,1/2) might be interesting.
> 
> Good continuation,
> Masayoshi
> 
> From: "Nichols A. Romero" <nromero@uiuc.edu>
> Subject: [abinit-forum] response function calculations
> Date: Thu, 03 Jun 2004 13:54:21 -0500
> Message-ID: <1086288861.26342.50.camel@nromero.physics.uiuc.edu>
> 
> > Hi,
> > 
> > I'm doing a straightforward phonon calculation for fullerenes in a SC
> > box, just doing q=0,0,0.
> > 
> > For each perturbation, I need about 44 SCF cycles to reach convergence.
> > This seems like quite a bit considering that WFK are pre-conditioned
> > with those of the unperturbed groundstate calculation.
> > 
> > I'm starting to wonder if I don't have some parameter set incorrectly,
> > e.g. dimiex or diemac.
> > 
> > Also, when doing response calculation in parallel does it parallelize
> > trivially? or is it restricted to by the number of kpts and spins.
> > 
> > I'm attaching my inputfile.
> > 
> >  acell 3*25
> >  ecut 50 Ry
> >  enunit 2  
> >  natom  29
> >  nband  70
> >  nstep 500
> >  ntypat 2
> >  prtvol 3
> >  rprim   1.0 0.0 0.0 0.0 1.0 0.0 0.0 0.0 1.0
> >  tolwfr 1.0d-16
> >  typat
> >  1
> >  1
> >  1
> >  1
> >  1
> >  1
> >  1
> >  1
> >  1
> >  1
> >  1
> >  1
> >  1
> >  1
> >  1
> >  1
> >  1
> >  1
> >  1
> >  1
> >  1
> >  1
> >  1
> >  1
> >  1
> >  1
> >  1
> >  1
> >  2
> >  xcart
> >   4.4435614233E+00  1.0045927297E+00 -9.9912245547E-01
> >   3.2965031371E+00  3.2924841976E+00  9.9445554601E-02
> >   3.2952950752E+00  1.2483220506E-01  3.3088090978E+00
> >   2.7922564088E+00  2.8094949302E+00  2.7904077156E+00
> >   4.4427561909E+00 -1.0040210357E+00  1.0361334566E+00
> >   1.0186213957E-01  3.3084170519E+00  3.2820400623E+00
> >  -1.0124523765E+00  1.0318305743E+00  4.4363209562E+00
> >  -4.4301068320E+00  1.0038928391E+00  1.0062795413E+00
> >  -3.2909476236E+00 -9.5028092171E-02  3.2980952239E+00
> >  -3.2900531739E+00  3.2808172295E+00 -1.2229033865E-01
> >  -1.0114343590E+00  4.4332709764E+00  9.9011297272E-01
> >   3.3082585306E+00 -1.2797379812E-01 -3.2762804606E+00
> >   1.0262468582E+00  9.7665942078E-01 -4.4295823097E+00
> >  -9.5186324265E-02  3.2648857676E+00 -3.3051643836E+00
> >   1.0144251158E+00  4.4231692966E+00 -1.0281255165E+00
> >   1.0126929243E+00 -9.8689022357E-01  4.4532604958E+00
> >   3.3070563126E+00 -3.2953604332E+00 -6.6569108040E-02
> >  -9.7204936859E-02 -3.2783537919E+00  3.3248643914E+00
> >   1.0245289739E+00 -4.4330279788E+00  1.0523009777E+00
> >   2.8115460348E+00 -2.8141042671E+00 -2.7595757545E+00
> >  -2.7858702543E+00 -2.7861620080E+00  2.8167813569E+00
> >  -3.2799766672E+00 -3.3046640998E+00  1.3053960501E-01
> >  -9.9365225072E-01 -4.4497583973E+00 -9.7341920617E-01
> >  -4.4231575807E+00 -1.0314542403E+00 -1.0024239518E+00
> >  -2.7841756646E+00  2.7638646311E+00 -2.8068517890E+00
> >  -3.2790891114E+00  7.1009957027E-02 -3.2901353093E+00
> >  -9.9264928550E-01 -1.0485811504E+00 -4.4200535856E+00
> >   1.2460251418E-01 -3.3216696602E+00 -3.2612500206E+00
> >   7.4755374696E-03 -5.9688480643E-03  9.0676144597E-03
> >  znucl    6.0 40.0
> > # ionmov    2
> > # optcell   1
> > # ntime   20
> >  diemix   0.99
> >  diemac   1.0
> >  nline 10
> >  isecur 1
> > # needed for later use with phonon calculation
> >  istwfk 2
> > # phonon calculation parameters
> >  irdwk 1
> >  rfphon 1
> >  rfatpol 1 29
> >  rfdir 1 1 1
> >  nqpt 1
> >  qpt 0 0 0
> > # may solve mem problems
> > # mkmem  0
> > # mffmem 0
> > # mkqmem 0
> > # mk1mem 0
> > # save disk space
> >  prtwf  0
> > 
> > Bests,
> > -- 
> > Nichols A. Romero
> > 
> > 4-133 ESB, MC 704                           508 East Michigan Avenue
> > 1110 West Green Street                      Apartment 12
> > Urbana, IL 61801-3080                       Urbana, IL 61801-5176
> > (217) 333-6108                              (217) 417-5210