My Community

Hi,

I am trying to do a calculation for phonon modes using the IBRION =8 option in Vasp 5.2. I have optimized the structure, and I am using a 2 by 2 supercell of the original structure for the phonon calculation. I am using 32 nodes (processors) to run the job. I get the following error with NPAR=1:

"Linear response reoptimize wavefunctions to high precision
MPI process terminated unexpectedly"

The job is getting killed at the first cycle of energy calculation. It looks to me there is some problem with the communication bandwidth.

I tried to fix the issue by using NPAR = 2/4/8 etc. but it only seemed to be working (i.e ran for quite a few cycles) for NPAR=2 (other values of NPAR did not make any difference) but after a while I get the following error message:

-----------------------------------------------------------------------------
| |
| EEEEEEE RRRRRR RRRRRR OOOOOOO RRRRRR ### ### ### |
| E R R R R O O R R ### ### ### |
| E R R R R O O R R ### ### ### |
| EEEEE RRRRRR RRRRRR O O RRRRRR # # # |
| E R R R R O O R R |
| E R R R R O O R R ### ### ### |
| EEEEEEE R R R R OOOOOOO R R ### ### ### |
| |
| VASP internal routines have requested a change of the k-point set. |
| Unfortunately this is only possible if NPAR=number of nodes. |
| Please remove the tag NPAR from the INCAR file and restart the |
| calculations. |
| |
| ----> I REFUSE TO CONTINUE WITH THIS SICK JOB ..., BYE!!! <---- |
| |
-----------------------------------------------------------------------------

I tried NPAR=32 (i.e. total number of nodes) and NPAR=1, it did not work.

Can anyone help me to resolve this issue? Any help will be greatly appreciated.

I'd like to mention here that I obtained the phonon modes using the IBRION=5 option without any problem for the same system (i.e. the same 2 by 2 supercell but using selective dynamics option to displace only the unit cell atoms). And I got very reasonable results.

Thank you.

Best Regards,
Ferdows

---------------------------
Here are the input files I am using:

INCAR
-------

SYSTEM = DHT low energy relaxation
ISMEAR = 0 ! Gaussian smearing
SIGMA = 0.05
ISTART = 0 !1 means restart from wavecar
ENCUT = 400
ALGO = vFast
EDIFF = 1E-7
PREC = A
LREAL = A
IBRION = 8
ADDGRID = .TRUE.
EDIFFG = -0.005 !eV/Ang
NELMIN = 4
LEPSILON = .TRUE.

## for small communication bandwidth ##
NPAR = 2
#NSIM = 4
#IRMMD = 4


POSCAR
---------

SYSTEM: Bi2Te3 bulk vibration 2 by 2 by 1 supercell
1.000
7.707188205240295 4.187106074597922 19.68315992116976
0 8.771123491943904 19.68315992116976
0 0 10.7744998931885
12 8
direct
0.00000 0.00000 -0.00000
0.00000 0.50000 0.00000
0.50000 -0.00000 -0.00000
0.50000 0.50000 0.00000
0.10600 0.10600 0.21200
0.10600 0.60600 0.21200
0.60600 0.10600 0.21200
0.60600 0.60600 0.21200
0.39400 0.39400 0.78800
0.39400 0.89400 0.78800
0.89400 0.39400 0.78800
0.89400 0.89400 0.78800
0.20000 0.20000 0.40000
0.20000 0.70000 0.40000
0.70000 0.20000 0.40000
0.70000 0.70000 0.40000
0.30000 0.30000 0.60000
0.30000 0.80000 0.60000
0.80000 0.30000 0.60000
0.80000 0.80000 0.60000


KPOINTS
-----------

Auto K-Points
0
g
6 6 6
0 0 0


POTCAR
---------
paw_PBE for Bi and Te

If you still get the above error when you set NPAR = 32, you do not use 32 nodes. Check the header of your OUTCAR file. However, the best way is always to not specify NPAR in the INCAR file since VASP then defaults NPAR to the number of nodes.

Please also notice that when the k-point grid is changed, the number of irreducible k-points increases (due to the lower symmetry). This gives rise to larger memory consumption. Therefore in addition please check if your memory is sufficient.

Cheers,
/Dan
<span class='smallblacktext'>[ Edited Thu Jun 24 2010, 01:05PM ]</span>

Hi Dan,

Thank you very much for your reply.

I tried without specifying any value for NPAR i.e. using the defaults set-up but it did not work. That is the first thing I tried. I get the error:

"MPI process terminated unexpectedly
signal 15 received."

Then I tried to play with the NPAR tag but with NPAR=2 I get the other error (change in k-point set ...) mentioned above.

What is the reason behind the "MPI process termination"? Can that be due to memory issue? It does not look to me it is a memory problem.

Any help will be appreciated.

Thank you.

Best Regards,
Ferdows