Vibrational frequencies (comparison with Gaussian 09)


Click here for full thread
Just Got Here
Hello,

I am trying to reproduce frequency calculations obtained in G09. My input is:

title "h2p s0"
 geometry units angstroms angstrom_to_au 1.889726132886
symmetry c1
C       0.684140000000000       -4.24575500000000       0.000000000000000E+000
C       1.12360200000000        -2.88439400000000       0.000000000000000E+000
N       0.000000000000000E+000  -2.10922200000000       0.000000000000000E+000
C       -1.12360400000000       -2.88434100000000       0.000000000000000E+000
C       -0.684159000000000      -4.24573100000000       0.000000000000000E+000
C       2.43141000000000        -2.41299700000000       0.000000000000000E+000
C       2.84124700000000        -1.07820400000000       0.000000000000000E+000
N       2.01702000000000        0.000000000000000E+000  0.000000000000000E+000
C       2.84124700000000        1.07817800000000        0.000000000000000E+000
C       4.23955900000000        0.676574000000000       0.000000000000000E+000
C       4.23955300000000        -0.676673000000000      0.000000000000000E+000
C       2.43146300000000        2.41298700000000        0.000000000000000E+000
C       1.12360400000000        2.88434100000000        0.000000000000000E+000
N       0.000000000000000E+000  2.10922200000000        0.000000000000000E+000
C       -1.12360200000000       2.88439400000000        0.000000000000000E+000
C       -0.684140000000000      4.24575500000000        0.000000000000000E+000
C       0.684159000000000       4.24573100000000        0.000000000000000E+000
C       -2.43141000000000       2.41299700000000        0.000000000000000E+000
C       -2.84124700000000       1.07820400000000        0.000000000000000E+000
C       -4.23955300000000       0.676673000000000       0.000000000000000E+000
C       -4.23955900000000       -0.676574000000000      0.000000000000000E+000
C       -2.84124700000000       -1.07817800000000       0.000000000000000E+000
N       -2.01702000000000       0.000000000000000E+000  0.000000000000000E+000
C       -2.43146300000000       -2.41298700000000       0.000000000000000E+000
H       -5.08682400000000       -1.35177200000000       0.000000000000000E+000
H       -5.08683500000000       1.35187900000000        0.000000000000000E+000
H       -1.34690300000000       5.10124000000000        0.000000000000000E+000
H       1.34693500000000        5.10120500000000        0.000000000000000E+000
H       5.08682400000000        1.35177200000000        0.000000000000000E+000
H       5.08683500000000        -1.35187900000000       0.000000000000000E+000
H       1.34690300000000        -5.10124000000000       0.000000000000000E+000
H       -1.34693500000000       -5.10120500000000       0.000000000000000E+000
H       -3.21052400000000       3.16994600000000        0.000000000000000E+000
H       3.21055300000000        3.16995600000000        0.000000000000000E+000
H       3.21052400000000        -3.16994600000000       0.000000000000000E+000
H       -3.21055300000000       -3.16995600000000       0.000000000000000E+000
H       -3.110000000000000E-004 1.09383400000000        0.000000000000000E+000
H       3.110000000000000E-004  -1.09383400000000       0.000000000000000E+000
 end

 basis cartesian
   * library 6-31g*
 end

 dft
  xc pbe0
  direct
  convergence energy 1e-6 density 1e-8 gradient 1e-6
  GRID lebedev 75 11 ssf euler
 end

# set cphf:thresh 1.0d-9
# set cphf:acc 1.0e-10

 task dft frequencies


Frequencies results for first few nodes:
NWChem: 55.117, 65.806, 95.066, 96.035, 130.044, 134.764,...
Gaussian: 54.9984, 64.8348, 95.6103, 97.5299, 129.8928, 134.0998,...
Most of frequencies are shifted around 0.5-1 [cm^-1] and the eigenvectors of vibrations are different.
The results are for further use in vibrational resolved optical spectra, so high precision and consistency between NWChem and Gaussian calculation is important for me.

Details:
Total DFT Energy was (still is) different:
-988.418737622 Gaussian
-988.4189460434 NWChem
, difference is in fourth digit (0.0002084214) so I added the following to NWChem:

angstrom_to_au 1.889726132886
Nuclear repulsion energy from Gaussian (1860.0533209599) and NWChem (1860.0534629997) was not the same. With angstrom_to_au 1.889726132886 nuclear repulsion energy difference lowers to 3.0E-10. Difference for One electron energy and Coulomb energy also lowers (forth digit), but the difference between total DFT Energies was still same (0.0002084214) so I ruled out this is not the problem.

basis cartesian
I used in gaussian 6D 10F for cartesian functions.

convergence energy 1e-6 density 1e-8 gradient 1e-6
To emulate similar SCF convergence. Both SCFs converge in 14 cycles (with change in energy between last cycles Gaussian DE= 1.82D-11 and NWChem DeltaE=5.46D-12).

GRID lebedev 75 11 ssf euler
For the same grid with gaussian (fine pruned (75,302) grid)

I also tried raising conditions in cphf section but the difference in frequencies and modes is still the same.

I would really appreciate your help or advice how to proceed.
Thanks,
Vedran Ivanic