How to determine occupied and closed orbitals for a Molpro CASSCF calculation?

I am currently working with Molpro, to which I am completely new, trying to run some calculations on $$\ce{PdO}$$. From literature I have gathered that the ground state is given by $$^3\Sigma^-$$ state, which translates to a $$^3A_2$$ state in the $$C_{2\mathrm{v}}$$ point group, or a wave function card {wf,54,4,2,0}. Moreover, the literature gives a possible active space of ($$4a_1$$, $$2b_1$$, $$2b_2$$, $$1a_2$$). I now want to run a CASSCF calculation using this active space, but I am not sure how to determine the occupied (OCC) and closed shell (CLOSED) orbitals.

How do I determine these values in general?

Edit

So I have been working on this a little more, and maybe I have found an approach that could work. I can run a simple HF calculation and get a set of orbitals whereby Molpro gives me the corresponding IRREPS. In the case for $$\ce{PdO}$$ I get a set of occupied orbitals ($$13a_1$$, $$7b_1$$, $$6b_2$$, $$2a_2$$), whereby two are singly occupied. To get my desired active space of ($$4a_1$$, $$2b_1$$, $$2b_2$$, $$1a_2$$), I would simply have to close orbitals according to ($$9a_1$$, $$5b_1$$, $$4b_2$$, $$1a_2$$).

Is this a sensible approach?

• Mar 11, 2021 at 23:10
• Thanks for the comment. I have read this post+answer and while very helpful, my problem is, unfortunately, primarily concerned with the direct implementation. I have played around with the occ and closed spaces in molpro, but I usually just get errors. The only successful calculation I ran was with the default CASSCF (no specifications on occ and closed), but this yielded terrible energies in comparison with lit. values.
– ABCCHEM
Mar 11, 2021 at 23:48
• @ABCCHEM, perhaps you can also tell us your basis set and the bond length you're using. Mar 14, 2021 at 3:07
• The basis sets I have been using: Pd=cc-pvdz-pp and O=cc-pvdz Mar 14, 2021 at 11:39

The occupied orbital pattern of 13 7 6 2 that you got from your Hartree-Fock calculation is not unique. For example, I've just run an RHF calculation on PdO with the ANO-RCC basis set and got a different occupancy pattern compared to you:

 Final alpha occupancy:  14   6   6   2
Final beta  occupancy:  14   5   5   2


I would recommend to find the experimental ground state configuration, which is something that for most diatomic molecules is not too hard to find. For example, here's an unpublished table from a project I worked on with an undergraduate student:

There is another complication, which is that, while the best basis sets for diatomic molecules like PdO are arguably the correlation-consistent basis sets, there are no correlation-consistent basis sets for Pd, so many of the electrons will be treated by a pseudopotential (and therefore not treated in the normal way).

If you want to pick the occupied orbitals based on your (non-unique!) HF calculation, then you can proceed with:

{casscf
closed,9,5,4,1
occ,4,2,2,1
wf,54,4,2,0};


But I would highly recommend surveying the literature first to see if the experimental ground state configuration is known, and otherwise doing a systematic study to decide which theory calculation is the one on which you wish to rely (already your HF calculation disagrees with mine!).

I may be able to help you more if it turns out we need to determine the occupation numbers based on theory alone, but we'd have to also make sure we're using the same basis set and bond length.

• Thank you very much, this has helped a great deal and I am now at least able to run some programs without getting errors or poor results. I have two follow-ups, if I may: how does Molpro determine the occupied orbital pattern that you and I got different results for, in other words why is it not unique? Also, is there a good choice of starting orbitals, better than HF, that one could use for CASSCF? With regards to the literature configuration, I found that in the literature, and it corresponds to the active space $(4a_1, 2b_1, 2b_2, 1a_2)$ as far as I could tell. Mar 14, 2021 at 8:50
• Getting different HF ground state configurations is a common feature of strongly correlated systems. If you have one-electron states that are very close to each other, it may be a game of luck which your calculation ends up converging on. It might be that a simple change of the basis set will land you on a different solution. DFT has been found in many papers to yield better initial orbitals than HF; however, the question is mainly about the active space. If the active space is reasonable, you should have few problems converging onto the right CAS orbitals. Mar 15, 2021 at 21:46
• @ABCCHEM I would be happy to answer these follow-up questions if you ask them in separate posts. We are extremely limited in terms of what we can say in comments, and it's not really appropriate to ask follow-up questions like these in the comments. Comments can be used to try to clarify what's in an answer, but not to ask a completely new question. Mar 15, 2021 at 21:52