SE energy calculation with PBC
shihao
shi... at gmail.com
Thu Sep 1 19:43:45 UTC 2011
Dear experts,
I'm trying to run semi-empirical energy calculation using CP2K, but
the calculation was very slow and the system consistency is not very
good. I've tried two systems: one with 125 water molecules (small
system), and the other with 250 water molecules (large system). To
prepare the large system, we duplicated and shifted the molecules from
the small system to generate two identical adjacent images. We expect
that the large system will give almost exactly twice the energy as the
small system, but the energies we got are around -1966.78 a.u. and
-3548.09 a.u., respectively, which deviates from 1:2 ratio. Besides,
the calculations are very slow. In the test of the small system on an
Intel Xeon E5462 quadri-core (3.0 GHz) cpu using one processor, it
took about 8 minutes to finish one inner SCF loop (50 steps). Is this
expected?
Any help or suggestions will be appreciated.
Thanks.
Shihao
P.S. Here's the input file for the small system test. I haven't
included the pdb and psf files because there's no upload function now,
but I can send them to you if needed.
@SET CURR_I 01
@SET SEED 1000
&GLOBAL
PROGRAM_NAME CP2K
PROJECT_NAME wt_bk_${CURR_I}
RUN_TYPE ENERGY
SEED ${SEED}
PREFERRED_FFT_LIBRARY FFTW
PRINT_LEVEL LOW
SAVE_MEM
&END GLOBAL
&FORCE_EVAL
METHOD QS
&DFT
&SCF
SCF_GUESS ATOMIC
EPS_SCF 1.0E-5
MAX_SCF 50
&OUTER_SCF
MAX_SCF 10
&END OUTER_SCF
&PRINT
&RESTART
&EACH
MD 20
&END EACH
&END RESTART
&RESTART_HISTORY OFF
&END RESTART_HISTORY
&END PRINT
&END SCF
&QS
METHOD AM1
EPS_DEFAULT 1.0E-10
EXTRAPOLATION ASPC
EXTRAPOLATION_ORDER 3
&SE
&END SE
&END QS
&MGRID
CUTOFF 200
&END MGRID
&POISSON
POISSON_SOLVER PERIODIC
PERIODIC XYZ
&END POISSON
&PRINT
&E_DENSITY_CUBE
&EACH
MD 20
&END EACH
&END E_DENSITY_CUBE
&END PRINT
&END DFT
&SUBSYS
&CELL
ABC 15.5516 15.5516 15.5516
PERIODIC XYZ
&END CELL
&TOPOLOGY
CONNECTIVITY UPSF
CONN_FILE_NAME ./build_small.xplor_psf
COORDINATE PDB
COORD_FILE_NAME ./tip125.pdb
PARA_RES T
&END TOPOLOGY
&END SUBSYS
&END FORCE_EVAL
More information about the CP2K-user
mailing list