BRMIX and FEXCP errors

[ericsun]

Hi,

I'm trying to run VASP on an IBM BlueGene machine. I tried two systems, one is defect in ZnO with about 300 atoms, another Si(111)-(7x7) surface with about 450 atoms. Both stopped at the switching point from non-selfconsistent to self-consistent steps (see below the error message I received). I have searched this forum and tried several tricks with no success. I use the DAV algorithm and the code is compiled with -O0. Actually, these jobs are for benchmarking purpose. So they have been finished successfully on all other machines. The POSCAR files were set up with fully relaxed structures. So, unreasonable small interatomic distances should not be a problem. Any other possible reason for this error?

running on 256 nodes
distr: one band on 256 nodes, 1 groups
vasp.4.6.28 25Jul05 gamma-only
POSCAR found : 2 types and 299 ions
LDA part: xc-table for Ceperly-Alder, standard interpolation
POSCAR, INCAR and KPOINTS ok, starting setup
FFT: planning ... 1
reading WAVECAR
entering main loop
N E dE d eps ncg rms rms(c)
DAV: 1 -0.446945275248E+11 -0.44695E+11 0.21099E+10 3088 0.107E+09
DAV: 2 0.154771935708E+05 0.44695E+11 -0.43620E+05 3072 0.875E+02
DAV: 3 0.123640880992E+04 -0.14241E+05 -0.13743E+05 3112 0.336E+02
DAV: 4 -0.180067487837E+04 -0.30371E+04 -0.31667E+04 3264 0.168E+02
DAV: 5 -0.373941003973E+07 -0.37376E+07 -0.37363E+07 3632 0.503E+01BRMIX: very serious problems
the old and the new charge density differ
old charge density: 2694.00000 new***********
0.117E+06
ERROR FEXCP: supplied Exchange-correletion table
is too small, maximal index : 1281270