VASP on NERSC - LAPACK ZPOTRF error

Message

ashinde · #1 Post by **ashinde** » Sat Sep 16, 2006 4:27 pm

I am running VASP on Nersc (in CA, USA), and I got a new error saying :
LAPACK: Routine ZPOTRF failed! 13

Could this be a problem with my input files? with VASP? or with NERSC's LAPACK?
I do not get any errors in OUTCAR that say anything out of the ordinary related to the ZPOTRF subroutine, but the OUTCAR ends abruptly and an error output file was generated with that message.
Thanks.

#2 Post by **admin** » Mon Sep 18, 2006 3:05 pm

ZPOTRF computes the Cholesky factorization of a complex Hermitian positive definite matrix A. The reasons of the error might be
a) there is some problem with your LAPACK installation
-- try to check it by using the lapack package provided with the code in vasp.4.lib
b) if you use the RMM-DIIS algorithm for electronic convergence (IALGO=48 or ALGO=V): this may fail for this step
-- switch to a different algorithm (ALGO = Normal or IALGO = 38, blocked Davidson may be more stable)
c) if the error occurs in the first ionic step: please check your input geometry again, have a look at the smallest nearest-neighbor distances as written in OUTCAR, maybe there is a printing error in your POSCAR file
d) If the error did not show up at the very first ionic step but after having done a few steps of ionic optimization: maybe, the first (or some other prior) ionic relaxation step lead to an unreasonable geometry
-- compare the input and output geometries of the first ionic iteration step and/or have a look at the "history" of ionic relaxations in XDATCAR.
-- in that case it can be helpful to switch to a different relaxation algorithm (IBRION-tag)
-- reduce the step size of the first step by setting POTIM= (0.1)

ashinde · #3 Post by **ashinde** » Mon Sep 18, 2006 4:17 pm

I changed IALGO to 38, I double checked POSCAR for positions and I set POTIM = .1. I get the same error, I think it does not even go through the first ionic step.

ashinde · #4 Post by **ashinde** » Mon Sep 18, 2006 4:27 pm

I'm not sure I understand how to check the LAPACK installation using the code in vasp.4.lib. LAPACK is already installed on the NERSC computers. Is there something I should tell them to check?

ashinde · #5 Post by **ashinde** » Fri Sep 29, 2006 5:06 pm

I've put up my POSCAR, can someone double check it ? I have the Pt(111) unit cell (3 layers, 4 atoms/layer) with an O atom on top.
Pt 111, O atom fcc site
3.99
.70710678 .0000000 .000000
-0.35355339 0.6123724 .000000
.000000 .000000 7.113
12 1
Selective Dynamics
Cartesian
0 0 0 F F F
2.821 0 0 F F F
1.410 2.443 0 F F F
4.232 2.443 0 F F F
1.410 .8140 2.303 F F T
2.821 3.257 2.303 F F T
0 3.257 2.303 F F T
4.232 .8140 2.303 F F T
0 1.628 4.607 F F T
2.821 1.628 4.607 F F T
1.410 4.072 4.607 F F T
4.232 4.072 4.607 F F T
1.410 2.443 5.807 T T T