Continuing Relax

[aruth]

Hello, the computer that I am working on has a time limit for which a job can run, and thus I need to restart a large job several times. I've normally restarted by copying CONTCAR to POSCAR and resubmitting, but the relax I am working on right now is so large, that this results in identical behavior week after week.

I have tried using job continuation through the ISTART tag, but so far, nothing I have entered has caused vasp to continue from where it left off and read WAVECAR. Instead it always starts with new wavefunctions.

Here is my INCAR

Code: Select all

System = MoSx20052x1x1
ISTART = 2
ISMEAR = 0 ! default 1: 0 GAUSSIAN, N>0 method of Methfessel-Paxton order N.
ISYM = 0
NSW    = 200 !Number of ionic steps
IBRION    = 2
ISIF = 3
NPAR= 4 !number of cores per band

And here is the logfile, I had to trim it down because of size limits.

Code: Select all

 running on   64 nodes
 distr:  one band on   16 nodes,    4 groups
 vasp.5.2.12 11Nov11 complex                                                    
  
 POSCAR found type information on POSCAR  Mo N  S  C  H  O 
 POSCAR found :  6 types and     336 ions

 ----------------------------------------------------------------------------- 
|                                                                             |
|  ADVICE TO THIS USER RUNNING 'VASP/VAMP'   (HEAR YOUR MASTER'S VOICE ...):  |
|                                                                             |
|      You have a (more or less) 'large supercell' and for larger cells       |
|      it might be more efficient to use real space projection opertators     |
|      So try LREAL= Auto  in the INCAR   file.                               |
|      Mind: At the moment your POTCAR file does not contain real space       |
|       projectors, and has to be modified,  BUT if you                       |
|      want to do an extremely  accurate calculation you might also keep the  |
|      reciprocal projection scheme          (i.e. LREAL=.FALSE.)             |
|                                                                             |
 ----------------------------------------------------------------------------- 

 LDA part: xc-table for Pade appr. of Perdew
 POSCAR, INCAR and KPOINTS ok, starting setup
 WARNING: small aliasing (wrap around) errors must be expected
 FFT: planning ...(          11 )
 WAVECAR not read
 WARNING: random wavefunctions but no delay for mixing, default for NELMDL
 entering main loop
       N       E                     dE             d eps       ncg     rms          rms(c)
DAV:   1     0.523761741006E+04    0.52376E+04   -0.39868E+05 49824   0.105E+03
DAV:   2    -0.146401411305E+04   -0.67016E+04   -0.64763E+04 62712   0.258E+02
DAV:   3    -0.213188855422E+04   -0.66787E+03   -0.66088E+03 60000   0.875E+01
DAV:   4    -0.215155787608E+04   -0.19669E+02   -0.19569E+02 62888   0.158E+01
DAV:   5    -0.215221256906E+04   -0.65469E+00   -0.65367E+00 64536   0.270E+00    0.169E+02
DAV:   6    -0.200901127268E+04    0.14320E+03   -0.48257E+02 59232   0.267E+01    0.728E+01
DAV:   7    -0.199600822871E+04    0.13003E+02   -0.23661E+02 64024   0.244E+01    0.739E+01
DAV:   8    -0.196887342792E+04    0.27135E+02   -0.48818E+01 58896   0.102E+01    0.166E+01
DAV:   9    -0.196661397192E+04    0.22595E+01   -0.18425E+01 58680   0.654E+00    0.864E+00
DAV:  10    -0.196629223497E+04    0.32174E+00   -0.33056E+00 54944   0.231E+00    0.566E+00
DAV:  11    -0.196631630845E+04   -0.24073E-01   -0.86856E-01 56944   0.118E+00    0.334E+00
DAV:  12    -0.196632066199E+04   -0.43535E-02   -0.12156E-01 57872   0.595E-01    0.161E+00
DAV:  13    -0.196632176331E+04   -0.11013E-02   -0.20831E-02 57176   0.271E-01    0.642E-01
DAV:  14    -0.196632504656E+04   -0.32832E-02   -0.15348E-02 57096   0.155E-01    0.298E-01
DAV:  15    -0.196632536777E+04   -0.32121E-03   -0.14070E-03 57904   0.610E-02    0.170E-01
DAV:  16    -0.196632556811E+04   -0.20035E-03   -0.77936E-04 48224   0.352E-02    0.896E-02
DAV:  17    -0.196632563008E+04   -0.61968E-04   -0.18711E-04 36224   0.211E-02
   1 F= -.19663256E+04 E0= -.19663256E+04  d E =-.196633E+04
 curvature:   0.00 expect dE= 0.000E+00 dE for cont linesearch  0.000E+00
 trial: gam= 0.00000 g(F)=  0.140E+00 g(S)=  0.405E-01 ort = 0.000E+00 (trialstep = 0.100E+01)
 search vector abs. value=  0.181E+00
 bond charge predicted

I am out of ideas at this point because I tried leaving out ISTART, setting ISTART =1 and setting ISTART = 2. For some reason the program overrides the last two. I see that on the wiki there is some mention of restarting jobs and parallelization, but I could not make sense of, from a practical standpoint, what was necessary in order to achieve continuation. Do I have to copy CONTCAR to POSCAR? Do I have not not copy CONTCAR to POSCAR?

[admin]

The WAVECAR is written when the relaxation is finished. When your job crashes because
of the time-limit the WAVECAR is not written.
Set such number of ionic steps that your job finishes before the time limit.
Then CONTCAR = POSCAR, and ISTART = 1
will read the WAVECAR and continue the relaxation.
Many systems have the symmetry of the cell clearly defined by XRD experiments.
That means the variation of the cell-shape is not necessary. The use of IBRION = 2
instead of 3 can save large number of relaxation steps.