9
$\begingroup$

I have tried to calculate state-averaged gradients with MOLPRO using the default basis set cc-pvdz.

However, I was met with the error message:

1PROGRAM * FORCE (Gradient of the energy)
 SA-MCSCF information from    5101.1
 Orbitals from record         2140.2
 Number of closed-shell orbitals: 29 ( 29 )
 Number of active orbitals:       12 ( 12 )
 Number of occupied orbitals:     41 ( 41 )
 Number of electrons= 72     Singlet     Space symmetry=1      Wavefunction type: SA-MCSCF
 GENERAL BASIS CONTRACTION NOT IMPLEMENTED IN CADPAC GRADIENTS
 Please try with segmented basis set

Can MOLPRO really not calculate excited-state gradients with cc-pvdz?? That's the default basis-set and the correlation consistent basis set.

The gradient does work with the Pople basis sets, however. I'm wondering if this is something that I'm doing wrong or is intrinsic to this version of MOLPRO (2012.1)

Bonus question: what does the error message mean general basis set contraction? I didn't consider Pople and correlation consistent basis functions much different.

$\endgroup$
2
  • $\begingroup$ Very interesting question @Cody Aldaz. Have you tried upgrading your version of MOLPRO? $\endgroup$ May 14, 2020 at 20:08
  • 1
    $\begingroup$ I think it costs money to upgrade. If any one could check if the newer version works that would be a good answer. $\endgroup$
    – Cody Aldaz
    May 14, 2020 at 20:09

1 Answer 1

11
$\begingroup$

The MOLPRO 2012.1 manual is no longer online, although there is an archive of at least some of the pages and you can try your luck for finding the pages you want on the archive, but it might not be necessary since the 2019 manual seems to answer your question.

MOLPRO has 3 different programs for calculating gradients: CADPAC, ALASKA/SEWARD, and AIC.

The 2019 manual clearly says "CADPAC does not work with generally contracted basis functions" and this is the error you got in your 2012 version (i.e. the functionality was not there then, and it's not here now). To do gradients with CADPAC, you need to use a basis set with a segmented contraction.

AIC does do gradients for SA-MCSCF, but it seems to be meant for use with density-fitting: "This program is used for computing gradients for the DF-MP2, DF-LMP2, DF-CCSD, DF-CCSD(T), DF-MP2-F12, DF-CCSD-F12, DF-CCSD(T)-F12, state averaged DF-MCSCF, and DF-CASPT2 methods".

The option you want to use for SA-MCSCF gradients with generally contracted basis sets is SEWARD: "SA-MCSCF gradients can be computed using segmented or generally contracted basis sets using SEWARD and the RS2 gradient program."

The MOLPRO 2012.1 manual makes no reference to AIC, and while it does refer to RS2, I was not able to find the above detailed description about the various programs that do gradients for SA-MCSCF. Functionality might have been added for MOLPRO 2015 and later, in which case your best options are either:

  • get the 30-day free-trial of MOLPRO 2019 to get you through this paper,
  • do gradients by finite differences (which I don't recommend),
  • do not do gradients, instead just optimize the geometry manually,
  • use a basis set with segmented contraction,
  • for cc-pVDZ, you can also uncontract the basis set and CADPAC may actually be able to do the gradients.
$\endgroup$
1
  • $\begingroup$ Actually getting the trial software to check whether it can do what you want, then apply for funding once you know it does should be a good way to do it. Or collaborate, or use a different program; there have to be some others capable of this. $\endgroup$ Mar 12, 2021 at 23:20

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .