eon @ vasp

[markus]

Hi,

I'd like to perform some dimer calculations with eon and VASP. I've seen that there is an MPI and a 'over files' interface to VASP. I'd like to use the 'over files' interface.

Could you provide a set of input files as an example, please? Are there some special options in INCAR that are important?


My current INCAR looks like this:

Code: Select all

SYSTEM = test
LINTERACT = .TRUE.
Startparameter for this run:
 ISYM = 0
 PREC = Accurate
 NWRITE = 2;  LPETIM = F write-flag & timer
 ISPIN = 1
Electronic Relaxation 1
 NELM = 300
 EDIFF = 1.E-07 stopping-criterion for ELM
 LREAL = .FALSE. A real-space projection
Ionic relaxation
 ALGO = Normal
 DOS related values:
 ISMEAR = 0 ; SIGMA = 0.1 broadening in eV -4-tet -1-fermi 0-gaus
 Electronic relaxation 2 (details)
 Write flags
  LAECHG = .FALSE.
  LCHARG = .FALSE.
  LWAVE = .FALSE.
!NBANDS = 350
 ADDGRID = .TRUE.

MAXMIX = 20
AMIX     =   0.05
BMIX     =   0.0001
AMIX_MAG     =   0.3
BMIX_MAG     =   0.0001




LPLANE = .TRUE.
NPAR  =  4
LSCALU = .FALSE.
NSIM   = 4
ENCUT =  600
IBRION = -1
POTIM = 0

NSW =   0
PSTRESS =     0.000000000000000E+00
EDIFFG =    -1.000000000000000E-06
ISIF   = 0

and my config. ini like this

Code: Select all

[Main]
job = saddle_search
finite_difference = 0.01

[Potential]
potential = vasp

[Optimizer]
opt_method = lbfgs
lbfgs_memory = 20
max_move = 0.2
converged_force = 0.01
max_iterations = 600

[Saddle Search]
min_mode_method = dimer

[Dimer]
converged_angle = 0.1
rotations_max = 20

[Debug]
write_movies = False

using these files I get the following message: vasp died unexpectedly!
From the OUTCAR file, I can see that VASP does not crash, as it terminates properly by showing the final timings (General timing and accounting informations for this job)
It seems as if EON can call VASP only once.

Here are the contents of client.log:

Code: Select all

nohup: ignoring input
EON Client
VERSION: rexported
BUILD DATE: Sun Oct 12 16:02:23 CEST 2014

Hostname: titan
OS: Linux
Arch: x86_64
PID: 63021
DIR: /home/markus/tests

Loading parameter file config.ini
* [Main] job: saddle_search
* [Main] finite_difference: 0.01
* [Potential] potential: vasp
* [Debug] write_movies: False
* [Optimizer] opt_method: lbfgs
* [Optimizer] converged_force: 0.01
* [Optimizer] max_iterations: 600
* [Optimizer] max_move: 0.2
* [Optimizer] lbfgs_memory: 20
* [Dimer] converged_angle: 0.1
* [Dimer] rotations_max: 20
* [Saddle Search] min_mode_method: dimer
Saddle point search started from reactant with energy -135.950321 eV.
[Dimer]       Step   Step Size      Delta E            ||Force||   Curvature    Torque    Angle   Rots
vasp died unexpectedly!

Best,
Markus

[chill]

I think the reason you get that message is because NSW=0. This means VASP quits after the first calculation. That is why Eon complains about VASP dying. Could you try making NSW larger than the expected number of "steps" in your calculation? You can just set it to a huge number if you like.