[molpro-user] SA-CASSCF optimisation with ALASKA

[Michael Banck]

Hi,

On Fri, Jun 06, 2014 at 03:04:46PM +0200, benjamin gonon wrote:
> I'm actually encountering a problem in performing optimization with
> state specific gradient including cpmcscf. If i well understand Molpro
> Manual, this need to include a CPMCSCF card in MCSCF group and use
> ALASKA gradient code (supposed to be the default) I think i did this
> well:

[...]
 
The 2012.1 manual says this in section 44.1: "The CADPAC gradient
program [...] works for [...] (state averaged) MCSCF" and further: "The
ALASKA gradient program [...] allows the calculation of gradients of
generally contracted basis functions for [...] MCSCF.[...] Gradients for
state averaged MCSCF wave functions can be evaluated using the RS2
gradient program".

So ALASKA does *not* work for state-averaged MCSCF wave functions via
the MULTI program.

> Unfortunately this doesn't seem to work and get me this too relevant
> error lines in output:
> VARIABLE ALASKA UNDEFINED, ASSUMING 0
>  GENERAL BASIS CONTRACTION NOT IMPLEMENTED IN CADPAC GRADIENTS

As mentioned above, CADPAC does not allow basis contraction.

> I know the same gradient can be computed using the rs2 program and
> that's probably what i will do in the end. But i would prefer to use
> this straight method than use rs2 code, as i will need to compute
> non-adiabatic coupling and optimize conical intersection.

What I have done in the past when RS2 was not an option was to use the
respective uncontracted basis set, by specifying the basis via the
primitive set definition (see section 11.8 in the 2012.1 manual).

I believe in your case it would look like this:

basis={
sp,h,vdz
spd,c,vtz
spd,o,vtz}

Using this as basis, the CADPAC gradient is getting evaluated.

Be aware that (i) I am not really sure this is a sound approach, (ii)
the calculation will take longer and (iii) the energies will differ
slightly.


Best regards,

Michael