Problem of sign of nonadiabatic coupling matrix

[Takao Kobayashi]

Dear MOLPRO users,

Although I have partially solved the inconsistency problem between sign of
nonadiabatic coupling matrix (NACM) and phase of wave function, there still
remains one problem that 

< (phi1) | d(phi2)/dR > is NOT equal to - < (phi2) | d(phi1)/dR > but
                        is     equal to   < (phi2) | d(phi1)/dR >. 

If (phi1) and (phi2) in the NACM element are commuted, the sign should be
changed, whereas in my MOLPRO calculation the sign does NOT seem to be
changed. Did anybody experience this kind of problem ?

Takao Kobayashi

PS.
The cause of the inconsistency problem is that the printed-out CI vectors
(and the corresponding orbitals) are not corresponding to those used in the
following calculation of NACM. So I have solved this problem by doing
two-times CASSCF calculation (first CASSCF for saving orbitals and second
just started by reading the previous orbitals without orbital optimization,
proceeding to the NACM calculation).