Posted by Björn Roos on February 22, 2004 at 10:42:04:

In Reply to: molecular properties posted by rajendra prasad on February 17, 2004 at 12:39:02:

: The following is mentioned in the MOLCAS-5.4 manual:

: "The CASPT2 method can also be used in combination with the FFPT program to compute dynamic correlation contributions to properties with good results in most cases."

: How these results are different from the expectation value of the respetive operator in the CASPT2 property section?

: Thanks.

Dear Prasad,

Of course they are. The CASPT2 method is non-variational with the coefficients of the first order wave function determined by perturbation theory. Therefore, the expectation values and FFPT results disagree. The FFPT are certainly most accurate. They depend only on the energy, which is correct to second order. The most important problem with the expectation values is the lack of single excitations in the first order wave functions. They contribute linearly to one-electron properties. One should therefore treat CASPT2 proeprties with some care. I have recently computed

Best

Björn Roos