TURBOMOLE Users Forum
TURBOMOLE Modules => Ridft, Rdgrad, Dscf, Grad => Topic started by: ardie on April 23, 2009, 03:45:22 PM
-
Dear all,
Turbomole stopped with the following message in an optimization process:
========================================================================
wrote 42 tasks to file
200635485 2 e - integrals written in 97967 blocks requiring 1567472 k-byte
current damping : 0.700
ITERATION ENERGY 1e-ENERGY 2e-ENERGY NORM[dD(SAO)] TOL
1 -2388.6290013038 -12725.529139 5608.4942936 0.000D+00 0.465D-11
Na = 145.99927170 Nb = 144.99908861
Exc = -221.122990684091 N = 290.99836032
max. resid. norm for Fia-block= 9.796D-03 for orbital 109a beta
max. resid. fock norm = 8.164D-02 for orbital 651a beta
rdiag: n=678, sequential Div&Con
rdiag: n=678, sequential Div&Con
Delta Eig. = 0.7439767560 eV
current damping : 0.650
ITERATION ENERGY 1e-ENERGY 2e-ENERGY NORM[dD(SAO)] TOL
2 -2388.6298692974 -12725.491117 5608.4554035 0.372D-01 0.295D-11
Na = 145.99926747 Nb = 144.99908131
Exc = -221.117136918458 N = 290.99834878
Norm of current diis error: 0.19259E-01
max. resid. norm for Fia-block= 4.830D-03 for orbital 120a alpha
max. resid. fock norm = 3.375D-02 for orbital 649a beta
rdiag: n=678, sequential Div&Con
rdiag: n=678, sequential Div&Con
Delta Eig. = 0.2628570030 eV
current damping : 0.600
ITERATION ENERGY 1e-ENERGY 2e-ENERGY NORM[dD(SAO)] TOL
3 -2388.6299485150 -12725.632379 5608.5965867 0.197D-01 0.237D-11
Na = 145.99925772 Nb = 144.99907987
Exc = -221.120359233222 N = 290.99833759
Norm of current diis error: 0.22023E-01
max. resid. norm for Fia-block= 2.110D-03 for orbital 120a alpha
max. resid. fock norm = 1.425D-02 for orbital 645a alpha
rdiag: n=678, sequential Div&Con
rdiag: n=678, sequential Div&Con
Delta Eig. = 0.1196711102 eV
from getini : eof encountered at record 1
MODTRACE: no modules on stack
fatal error in getini.
dscf ended abnormally
=================================================
What does this mean and how to resolve it? Thank you.
Ardie
-
Hi,
looks like parallel dscf, right?
Did you set the location of the integral files (twoint) to a local directory? If not, all clients write on an NFS disk which will slow down your calculations (dramatically, if you are unlucky, can even make the calculation slower than the serial version...), and this might cause network problems either.
Please set either the size of the twoint files to zero by simply editing the control file at the $scfintunit and set size=0, or use Turbomole 6.0 which will not use the disk by default any more for integral storage.
Regards,
Uwe