Vibrational Analysis
flo
fsch... at pci.uzh.ch
Tue Jun 24 09:44:00 CEST 2008
Hi Ali,
yes you can get the frequency which is associated with these four
atoms but you forgot a keyword. maybe for to claryfy for everbody who
may read this topic I will just explain the different keywords in your
input file:
&VIBRATIONAL_ANALYSIS
! important!!! if you specify the keyword intensities you will have to
add the
! keyword for evaluating dipoles in the FORCE_EVAL section
FULLY_PERIODIC T
! Only important for a full vibrational analysis
!NPROC_REP 2
! Important for multiprocessor runs with n processors. It will
perform
! n(processors)/nproc_rep calculations at the same time.
! with mode_selective a block davidson algorithm is started.
&MODE_SELECTIVE
FREQUENCY 500
! frequency you want to track (as Ali said:
! using Frequency, Range and Involved
! atoms are mutually exclusive)
! in your case Ali, you will have to use involved_atoms 115 117 120
122
! to get the frequency associated with these atoms
ATOMS 115 117 120 122
! atom numbers for the initial guess (might help to converge faster).
! For the initial guess a random vector is construct only having
components
! on these atoms
EPS_MAX_VAL 1.0E-07
! convergence criterium for the Davidson algorithm.
! The default is already tight enough, since the second derivative is
evaluated analytically.
&END
For you it means jsut remove frequency and add INVOLVED_ATOMS 115 117
120 122.
One importtant point for your calculation might be:
Since I haven't removed the translations and rotations from the
algorithm (in this algorithm it's a bit more difficult), I restricted
the frequencies for the involved atoms part to values >400cm-1
(otherwise you will many times get the translations). Maybe your
frequency is lower. Just to be sure, you will get it, you can change
the code as follows:
grep '400' mode_selective.F
everywhere it occures just change the value to the threshold you want
and compile cp2k (should be fast, because it's a tiny change).
Hope this helps
Flo
More information about the CP2K-user
mailing list