Why are the IR-Active Frequencies Wrong in MOLPRO 2000.1

[Jun Li]

 >The likely reason is that by default Molpro uses spherical harmonics,
 >while Gaussian uses cartesian basis functions (d,f,g..). Please add
 >the keyword CARTESIAN to your input (e.g., before basis) and try the
 >calculation again. Make sure that you get the same number of basis functions
 >in Gaussian and Molpro and an identical initial scf energy.
 >Should you still get unexpected results, please let me know.
 >H.-J. Werner

Dear Professor Werner,

Happy New Year!

Thanks for your instruction. I have carefully compared the calculations of 
UF6 vibrational frequencies using MOLPRO 2000.1 and Gaussian 98 (as well as 
NWChem).  I tried to use both pure (5d7f) or Cartesian (6d10f) basis 
functions and carefully compare the initial guess and final electron 
occupations.  The results seem to indicate that MOLPRO frequencies might be 
wrong for the IR-modes.  I am listing here the results (see previous 
message for the input file):

Vibrational Frequencies (cm-1) of UF6  RHF/LANL2DZ (Hay) for U, 6-31G(d) 
for F

            T2u  T2g   T1u   T1u   Eg   A1g    bond length  E(total)
5d7f basis functions
  MOLPRO    157  217   261  1044   583   761   1.9838     -647.42569878 

  G98 (JL)  157  216   209   703   582   761   1.9838     -647.42569865
  G98 (Hay) 157  216   209   702   582   761   1.9838     -647.42570



6d10f basis functions
   MOLPRO    155  219   257  1035   583  754    1.9875     -647.48110684 

   G98 (JL)  155  218   206   697   583  754    1.9875     -647.48110644

   exptl.    143  200   186   626   534  667    1.999 or 1.996


Thank you very much for your attention to this questions.

Sincerely,

Jun Li