[SOLVED] Orbitals rotating in CASSCF

LucaBabetto · 2019-05-16 13:24:48

Hello,

I want to run a CASSCF calculation on a not so small organic molecule, characterised by an extended delocalized pi system.

First, I ran a Hartree-Fock calculation with a minimal basis set (ANO-L-MB) and used the LOCALISATION module.

I checked the orbitals to put in the active space, and as expected I found 15 orbitals with 2pz (3pz for S) contributions, for a total of 18 electrons in pi orbitals.

I ran a CAS(18,15) calculation on this system, exploiting the EXPBAS feature to increase the basis set size to a VTZP quality, rotating the orbitals which weren't in the active space using the Alter keyword and putting them in the active space. The "intruder" orbitals were all of sigma character (2px/2py contributions).

This is confirmed by this section in the output file:

Molecular orbitals exchanged:
        In symmetry  1 :  45  54
        In symmetry  1 :  48  55
        In symmetry  1 :  52  56
        In symmetry  1 :  68  66
        In symmetry  1 :  71  67

At the end of the calculation, however, the "intruder" orbitals are put back in the order they were before. I was expecting only to have only pi orbitals in the active space. In the RASSCF iterations there is no mention of orbitals being rotated out of the active space.

Is this related to this warning, or is there a more fundamental issue underlying?

Note: transformation to natural orbitals
      has been made, which may change the order of the CSFs.

Ignacio · 2019-05-16 13:30:35

If the rotations happen slowly during the RASSCF calculation, you won't see any warning.

Check that you counted the orbitals correctly (see if the "total charge" in the output is what you expect). You may need to include other excited states (with CIRoot) ot stabilize your desired active space. If you are trying to include orbitals that are almost empty in the ground state, the system will prefer to include other orbitals with a somewhat larger occupation instead.

LucaBabetto · 2019-05-16 13:37:20

Yes the total charge is 0, which is the expected value.

The orbitals I'm trying to include in the active space are all the orbitals with pi character, of which about half are full and half are empty. In between the occupied pi orbitals, however, there are some sigma orbitals, so the program would rather try to include those instead?

How would including excited states help in this regard?

Ignacio · 2019-05-16 15:29:12

I don't meant empty in the HF determinant, but empty in what would be the CASSCF solution you are looking for. The algorithm finds out that there is very little gain (decrease in energy) by including these orbitals, and it is more favourable to include some sigma orbitals instead. But this is speculation.

You would include excited states by adding e.g. "CIRoot = 5 5 1", that is a state-averaged calculation with 5 roots (ground state plus 4 excited states). The idea is that the lowest-lying excited states will probably have pi-pi* excitations, which will generate some occupation in those empty orbitals, and make the algorithm decide that it is after all much better to use these orbitals in the active space, rather than sigma orbitals that would result in higher excitation energies.

But you said you used EXPBAS. I presume you did the CASSCF(18,15) calculation with MB and it went fine? If not, this what you should do, before doing EXPBAS. If yes, maybe your problem orbitals are not sigma, but Rydberg (try a smaller isovalue for plotting the orbitals, and see if they become significantly larger than other valence orbitals). In that case, you may need to actually enlarge your active space to include these orbitals (and compute more excited states if that's what you want).

LucaBabetto · 2019-05-17 15:35:26

I have not tried to run the CASSCF calculation with the minimal basis set, I will try that.

If I use CIRoot and include excited states via the state-averaged procedure, wouldn't it mean my ground state should also include those excited states?

I say this because I remember I had to use CIRoot and the state-averaged formalism for example in systems containing lanthanides. With Eu(III), for example, I included 7 state-averaged states in the ground state because there are 6 singly occupied 4f orbitals and 1 empty orbital.

I am a bit fuzzy on the formalism here.

Ignacio · 2019-05-17 15:41:33

State-averaged CASSCF only means the the orbitals are optimized with respect to the average energy of several states. Each state will still be "pure", but it won't have the best possible description, because the orbitals are not tailored for it exclusively.

This is different from the mixture of states you get with RASSI or MS-CASPT2.

LucaBabetto · 2019-05-20 07:22:13

Okay, I understand now, thank you for clarifying this! I will try including some of the lower excited states (which should all be pi-pi transitions) and see if this does the trick.

LucaBabetto · 2019-05-20 09:35:13

But you said you used EXPBAS. I presume you did the CASSCF(18,15) calculation with MB and it went fine? If not, this what you should do, before doing EXPBAS. If yes, maybe your problem orbitals are not sigma, but Rydberg (try a smaller isovalue for plotting the orbitals, and see if they become significantly larger than other valence orbitals). In that case, you may need to actually enlarge your active space to include these orbitals (and compute more excited states if that's what you want).

I tried running the same calculation with a minimal basis set, and in the active space now there are only pi orbitals. In the large basis set calculations, the "intruder" orbitals do not seem to be Rydberg states, as they do not seem to be much larger than the other valence orbitals at the same isovalue).

As a reference, this is the last orbital in the active space after the optimization with the minimal basis set (clearly an antibonding pi orbital):

While this is the same orbital with the larger basis set (all other settings left equal), which definitely looks like an antibonding sigma orbital:

Ignacio · 2019-05-20 09:53:20

Since the MB orbitals are "right", did you try running EXPBAS from those (not from the localized orbitals, but from the final RASSCF optimized orbitals)?

LucaBabetto · 2019-05-20 09:57:34

So, running a first DMRG/RASSCF calculation with a minimal basis set, then the "real" calculation using EXPBAS on the RASSCF optimized orbitals? No I have not tried that, and I have to say I feel silly for not having thought of that myself cry

I'm still puzzled by what is causing the sigma orbitals to rotate in the active space with a larger basis set. In theory those orbitals are already present even with the MB set, since it's all 2pz vs. 2px+2py.

LucaBabetto · 2019-05-21 14:46:23

I tried using the RASSCF wavefunction as input for the larger basis set calculation, and the resulting orbitals are still very similar to the "original" calculation with the larger basis set. It seems those orbitals really want to get in the active space.

Now I'm trying including excited states for a state-averaged calculation as you suggested.

LucaBabetto · 2019-05-22 08:19:48

I did a run using the CIRoot keyword, including 5 excited states 4 excited states (+ the ground state). All the orbitals in the active space remained of pi character, so this is what I'll be using from now on, thank you for the tips Ignacio!

Last edited by LucaBabetto (2019-05-22 08:20:24)

Molcas Forum

Announcement

#1 2019-05-16 13:24:48

[SOLVED] Orbitals rotating in CASSCF

#2 2019-05-16 13:30:35

Re: [SOLVED] Orbitals rotating in CASSCF

#3 2019-05-16 13:37:20

Re: [SOLVED] Orbitals rotating in CASSCF

#4 2019-05-16 15:29:12

Re: [SOLVED] Orbitals rotating in CASSCF

#5 2019-05-17 15:35:26

Re: [SOLVED] Orbitals rotating in CASSCF

#6 2019-05-17 15:41:33

Re: [SOLVED] Orbitals rotating in CASSCF

#7 2019-05-20 07:22:13

Re: [SOLVED] Orbitals rotating in CASSCF

#8 2019-05-20 09:35:13

Re: [SOLVED] Orbitals rotating in CASSCF

#9 2019-05-20 09:53:20

Re: [SOLVED] Orbitals rotating in CASSCF

#10 2019-05-20 09:57:34

Re: [SOLVED] Orbitals rotating in CASSCF

#11 2019-05-21 14:46:23

Re: [SOLVED] Orbitals rotating in CASSCF

#12 2019-05-22 08:19:48

Re: [SOLVED] Orbitals rotating in CASSCF

Board footer