Using electronic wavefunction after NEVPT2 correction for response

[ankit7540]

Dear All,
After applying NEVPT2 correction to mcscf wavefunctions, significant reduction in energy is observed. However, the subsequent response calculation falls back to the mcscf wavefunction giving no improvement in the properties. Hence, the question.
Can we use the electronic wavenfunction after the NEVPT2 correction in the RESPONSE module.

Attached, an example file.

After NEVPT2 section,

Code: Select all

     ********************************
     TOTAL ENERGY CONTRIBUTIONS
     ********************************


 STATE   DIAGONALISATION        PERTURBATION SC-D        PERTURBATION PC-D
    1    -149.71095598           -150.05854638           -150.06388075

In the RESPONSE section, output is

Code: Select all

 Linear Response calculation
 ---------------------------

 Print level                                    : IPRLR  =   2
 Maximum number of iterations                   : MAXITL = 125
 Threshold for relative convergence             : THCLR  = 1.000D-05
 Maximum iterations in optimal orbital algorithm: MAXITO =  18
  .
  .
  .
  .
 MCSCF energy         :     -149.710955983285999
 -- inactive part     :     -128.176372296813440
 --   active part     :      -44.391726543615420
 -- nuclear repulsion :       22.857142857142858


                        **********************************
                        *** MCSCF response calculation ***
                        **********************************

Thank you.

[kennethruud]

The NEVPT2 energy correction has only been implemented as that, an energy correction, and not as correction to the wave function. No properties have been implemented for the NEVPT2 approximation.


Best regards,

Kenneth

[ankit7540]

Thanks for the information.

[taylor]

If you wish to calculate (static) polarizabilities at the NEVPT2 level you could consider adding a small finite electric field to the Hamiltonian (this is documented in the manual), calculating the polarizability components by finite differences of energies. But this is only feasible for the static case.

Best regards
Pete

[ankit7540]

Thank you for this suggestion. I will definitely check about this. ( Usually we are interested in frequency dependent properties since we have experimental data to compare. )