# MCLR convergence

[ Molcas user's WWWBoard ]

Posted by Puja Goyal on November 15, 2013 at 22:08:45:

Hi,

I am trying to compute the SA-CASSCF gradient for one of the excited electronic states of my system.

After the RASSCF section, my input is as follows:

========
&mclr
iter = 1000
sala = 2

pnew
verbose
show
=======

I keep getting the following message:
=======

Warning perturbation number 1 will diverge
****************************************
* *
* WARNING!! *
* Elements in the E^[2] matrix small!! *
* The calculation might diverge. *
* *
* Check your active space!!!! *
* *
* Make sure degenerate orbitals do not *
* belong to different spaces. *
* Note that no LR code can handle *
* 2.0d0 occupancy in active orbitals!! *
****************************************
No convergence for perturbation no: 1. Increase Iter.
========

I am not sure how to handle this error. The occupancy of the active orbitals from the RASSCF output is as follows:

=====================================

Natural orbitals and occupation numbers for root 1 sym 1: 1.934400 1.999587 1.935544 1.944444 1.927387 0.001640 0.066298 0.072392 0.053870 0.064439

Natural orbitals and occupation numbers for root 2 sym 1: 1.934790 1.028756 1.932995 1.943376 1.923652 0.972574 0.067077 0.076069 0.055910 0.064801

Natural orbitals and occupation numbers for root 3 sym 1: 1.922291 1.996076 1.872671 1.939476 1.146554 0.856009 0.114374 0.049028 0.037084 0.066437

Natural orbitals and occupation numbers for root 4 sym 1: 1.920818 1.998743 1.150596 1.934432 1.804314 0.855403 0.161962 0.061368 0.044091 0.068273
====================================

Is the correct spot in the output where I should be checking the active orbital occupancies? For root 2, for which I am trying to get the gradient, none of the active orbitals has an occupancy close to 2.0. Even if I consider the active orbital occupancies averaged over all the roots, none of them would be close to 2.0.

I will really appreciate it if someone can guide me regarding what I can do to make MCLR converge.

Thanks a lot.

Puja

Name:
E-Mail:

Subject:

if B is 1s22s22p1, what is Li?

Passfield: