[molpro-user] VSCF calculation
Guntram Rauhut
rauhut at theochem.uni-stuttgart.de
Thu Feb 8 11:19:12 CET 2018
Dear Fabian,
your calculation may have failed for different reasons:
a) A calculation with ngrid=8 usually does not provide meaningful
results. ngrid is NOT the number of ab initio points in one dimension,
but the number of fine grid points, which is used in subsequent
calculations for the wave function. ngrid controls also (to some extent)
the maximum elongations along the normal coordinates. As ngrid=8 is far
too low, this will lead to very small surfaces. Note that the default is
ngrid=16. Once your surfaces are too small, the effective potentials
within the VSCF calculations may not be high enough in the outer regions
to harbor the VSCF wave function.
b) It is very disturbing that your 1D vibrational frequencies within the
SURF program show meaningful results, while the diagonal frequencies in
the VSCF are completely wrong. Essentially both calculations should give
(more or less) the same answer. The differences are (1) that the 1D
calculations within the SURF program do not include the Watson
correction term, while this term will be added in the POLY calculation
and thus it is included in the 1D calculation in the VSCF program.
However, this effect should by small and cannot explain the huge
differences, which you can see here. (2) The 1D calculations in the VSCF
program are based on polynomials, while the 1D calculations in the SURF
program just request the grid points and use a DVR procedure for
calculating the energies. Most likely, it is the polynomial fit, which
causes you the trouble in your calculations. Did you check the chi^2
values in the POLY program? Are they meaningful? How do the polynomials
look like?
c) You did neither shift nor scale your individual normal coordinates
(SCALNM,AUTO=ON), which might lead to the situation, that the 1D
potential of individual normal coordinates is very unbalanced. This may
be the reason, why the polynomial fit runs into trouble. Since you have
plotted the potentials, you can easily control this by visual inspection.
d) What can you do to rescue your existing surface? (1) You may leave
out the polynomial fitting (POLY) and may use the grid based VSCF
program instead. (2) You may transform the potential to a grid
representation with more grid points using the VGRID program. Then you
may use this representation to enter the grid-based VSCF code. But I am
afraid that you need to rerun the calculations with ngrid=16, the
automated scaling a.s.o.
I hope this helps.
Best wishes,
Guntram
On 02/07/18 10:28, Fabian Berger wrote:
> Dear molpro user,
> I have calculated the attached VSCF calculation. This calculations
> finished properly but the results are quite suspicious.
> Please have a look at the frequencies (which are printed after the
> full VSCF calculation) below:
>
> Mode Harmonic Diagonal VSCF E(abs)
> 21 3895.46 35100.93 94654.42 *********
> 20 2287.49 11388.67 9714.29 *********
> 19 2263.76 2242.11 3415.39 *********
> 18 1952.70 2036.93 2797.70 *********
> 17 1930.40 8892.65 13269.76 *********
> 16 1041.72 582.15 257.85 *********
> 15 990.83 620.67 1839.88 *********
> 14 937.69 747.23 1553.29 *********
> 13 894.56 88.28 12701.39 *********
> ...
>
> While the harmonic frequencies appear to be reasonable, the 1D
> frequencies (Diagonal) and VSCF frequencies deviate significantly.
> Interestingly, the frequencies which are printed during the PES
> calculation (after the 1D part) seem to be much more similar to the
> harmonic values:
>
> Mode Harmonic Diagonal Intens
> 21 A 3895.46 3726.10 167.28
> 20 A 2287.49 2214.86 149.16
> 19 A 2263.76 2191.54 158.09
> 18 A 1952.70 1895.77 246.07
> 17 A 1930.40 1874.12 287.72
> 16 A 1041.72 1035.21 342.11
> 15 A 990.83 985.04 122.05
> 14 A 937.69 949.79 39.75
> 13 A 894.56 933.24 426.63
> ...
>
> I have the following questions:
>
> 1) What is the difference between the 1D frequencies after calculating
> the 1D potentials and the full PES calculation?
> 2) Why do such non-senseful frequencies arise after the calculation of
> the coupling terms in the 1D frequencies?
> 3) Which input parameters could I adjust to fix this problem?
>
> Thank you very much!
>
> Best regards,
> Fabian Berger
>
>
> Input file:
>
> ***,anharm
> memory,960,m
> file,1,surfdebug.int
> file,2,surfdebug.wf
> angstrom
> geometry={
> O, 1.37395990000000, 3.12849230000000, -2.72838240000000
> Al, 0.11308370000000, 1.80597310000000, -3.67800270000000
> O, -1.01719110000000, 2.68772110000000, -2.58192110000000
> Si, 0.01089600000000, 3.62802360000000, -1.76857130000000
> H, 2.24328820000000, 2.96029440000000, -2.35571480000000
> H, 0.59049360000000, 0.37640650000000, -3.19723810000000
> H, 0.05245040000000, 2.19748800000000, -5.20403290000000
> H, -0.17319630000000, 5.08032320000000, -1.91244630000000
> H, 0.29860050000000, 3.28203280000000, -0.36344240000000
> }
>
> ! structure relaxation
> basis,default=cc-pvtz
> {hf
> accu,14}
> ccsd(t)
> optg
>
> ! hessian
> mass,iso
> frequencies,symm=no
>
> ! PES calculation
> label1
> int
> {hf
> start,atden
> accu,14}
> ccsd(t)
> goto,label3
>
> label2
> {df-rks,pbe0;disp}
> goto,label3
>
> label3
> {surf,start1D=label1,sym=no,ndim=2,ngrid=8,plot=2
> vmult,start2D=label2,Multi=1
> disk,save=5600.2,where=home,dump='surfdebug.pot'}
>
> ! VSCF
> poly,dipole=0
> vscf,type=poly,thermo=1,dipole=0
> _______________________________________________
> Molpro-user mailing list
> Molpro-user at molpro.net
> http://www.molpro.net/mailman/listinfo/molpro-user
--
********************************************************************************
Apl. Prof. Dr. Guntram Rauhut
Institut f. Theoretische Chemie
Universitaet Stuttgart
Pfaffenwaldring 55
D-70569 Stuttgart
Germany
Tel. : +49/(0)711/685-64405
FAX : +49/(0)711/685-64442
E-Mail : rauhut at theochem.uni-stuttgart.de
HTTP : www.uni-stuttgart.de/theochem/rauhut/
********************************************************************************
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.molpro.net/pipermail/molpro-user/attachments/20180208/ed8978f2/attachment.html>
More information about the Molpro-user
mailing list