Posted by Valerij Gurin on March 18, 2005 at 14:43:51:

In Reply to: Re: TDM posted by Roland Lindh on March 17, 2005 at 16:41:04:

Dear Roland,

Thanks a lot for explanation.

Should I understand that MOLCAS principally cannot keep this sign or

I did not examined yet in its capabilities?

: the determination of the wave function is always arbitrary with respect to the sign of the wave function. This is no problem since any observable is independent of the sign. In energies and electron densities, for example, wavefunction is squared. Hence, no problem here if the sign flip-flops.

OK. You are right, TDM is not observable, and usually the problem with sign does not appear from this reason. However, in the works those calculate TDMs they are given with certain sign, and for some sequence of geometries, e g potential curve of diatomics,

TDMs are plotted by smooth curves even crossing the zero. Thus, software can calculate correct sign. I saw publication with MOLCAS also, the other packages are MOLPRO, COLUMBUS (Gaussiano3, by the way, results also in flip-flop signes...). Possible, some special care need for building these sequences of tasks.

The TMDs are used in the computation of the transition intensities, here one use the square of TDM, i.e. no problem where either.

Of course.

A problem could be if you want to create a plot of the TDM as the geometry change. Here you would not like the associated wave functions to change sign. You could either change these by hand, however, when the TDM goes through zero you must be careful. You could try to avoid this by using the CI-vectors from the previous point and let the RASSCF restart from these (remember to use both the JOBIPH and the CIRESTART options).

Yes, I'l try, but CI-vectors from previous point correspond to previous geometry.

New geometry (in SEWARD) can provide change of sign again. Or I'm not right in this understanding of sequenceof of MOLCAS actions?

This should take care of most of your problems. Of course there would still be a problem if you would have the TDM being close to zero or going through zero along a reaction path as you pass an avoided crossing. Here the two states will mix and you might lose track of your reference sign, but good judgement would probably help you here.

This is the next step - to analyse carefully close-to-zero TDMs. At the first approx.

these transitions could be neglected (I~(TDM)^2).

Regards,

Valerij