How to read CSFs in the output file of CASSCF calculations

Find answers or ask questions regarding Dalton calculations.
Please upload an output file showing the problem, if applicable.
(It is not necessary to upload input files, they can be found in the output file.)

Post Reply
footballplayer
Posts: 14
Joined: 01 May 2016, 16:51
First name(s): Nahn
Last name(s): Mahp
Affiliation: HCMUNS
Country: Viet Nam

How to read CSFs in the output file of CASSCF calculations

Post by footballplayer » 01 May 2016, 18:41

Dear all,

I am running some CASSCF calculations. I do not know how to read Configuration state functions in the CASSCF output. For example, this is what I got from the output file after doing calculation (C2v point group)

Printout of CI-coefficients abs greater than 0.05000 for root 1
*** NOTE: this root is the reference state ***

Printout of coefficients in interval 3.1623E-01 to 1.0000E+00
===============================================================

Coefficient of CSF no. 1221 is 0.95115466 9.51154657E-01
Orbital 1 2 3 7 9 4 5 10 12
Spin coupling 2 2 2 2 2 1 1 1 1

Printout of coefficients in interval 1.0000E-01 to 3.1623E-01
===============================================================

Coefficient of CSF no. 6423 is -0.10957103 -1.09571026E-01
Orbital 1 2 7 9 11 4 5 10 12
Spin coupling 2 2 2 2 2 1 1 1 1

Printout of coefficients in interval 5.0000E-02 to 1.0000E-01
===============================================================

I would like to know the irrep of each orbital, spin (alpha, beta) of singly occupied orbitals, the meanings of digits after the label "Orbital" as in the above excerpt.

For the case of Molcas, I can read these CSFs easily.
printout of CI-coefficients larger than 0.05 for root 1
energy= -1023.956646
conf/sym 11111 22 333 44 Coeff Weight
271 22uu0 20 u00 u0 -0.96625 0.93365
3836 u2du0 u0 uu0 ud 0.05263 0.00277
5814 u2uu0 d0 ud0 uu -0.06398 0.00409
6475 22uu0 00 u00 u2 0.09798 0.00960
8113 02uu0 20 u20 u0 -0.11194 0.01253
15807 u2uu0 u0 ud0 ud 0.05381 0.00290

hjaaj
Posts: 333
Joined: 27 Jun 2013, 18:44
First name(s): Hans Jørgen
Middle name(s): Aagaard
Last name(s): Jensen
Affiliation: Universith of Southern Denmark
Country: Denmark

Re: How to read CSFs in the output file of CASSCF calculatio

Post by hjaaj » 01 May 2016, 20:02

Spin coupling 2: doubly occupied, not changing spin (Molcas: 2)
Spin coupling 1: singly occupied, raising spin with 0.5 (Molcas: u)
Spin coupling -1: singly occupied, lowering spin with 0.5 (Molcas: d)

Your CSF's thus correspond to S=2 (quintet).

footballplayer
Posts: 14
Joined: 01 May 2016, 16:51
First name(s): Nahn
Last name(s): Mahp
Affiliation: HCMUNS
Country: Viet Nam

Re: How to read CSFs in the output file of CASSCF calculatio

Post by footballplayer » 01 May 2016, 20:20

Thanks for your answer hjaaj. And what about the irreducible representations of orbitals?

In my case, I would like to extract the dominant configuration (including irreducible representations of occupied orbitals) in order to do a single reference calculation on the basis of this dominant CSF. For example, the above CSFs are 5B1 of the C2v point group. Which orbitals are a1, b1, b2 and a2?

Coefficient of CSF no. 1221 is 0.95115466 9.51154657E-01
Orbital 1 2 3 7 9 4 5 10 12
Spin coupling 2 2 2 2 2 1 1 1 1

Coefficient of CSF no. 6423 is -0.10957103 -1.09571026E-01
Orbital 1 2 7 9 11 4 5 10 12
Spin coupling 2 2 2 2 2 1 1 1 1

hjaaj
Posts: 333
Joined: 27 Jun 2013, 18:44
First name(s): Hans Jørgen
Middle name(s): Aagaard
Last name(s): Jensen
Affiliation: Universith of Southern Denmark
Country: Denmark

Re: How to read CSFs in the output file of CASSCF calculatio

Post by hjaaj » 01 May 2016, 20:36

If you have a .CAS SPACE of 5 2 3 2 then orbital no. 1-5 are from symmetry 1, orbital no. 6-7 from symmetry 2, orbital no. 8-10 from symmetry 3, and orbital no. 11-12 from symmetry 4. The totally symmetric symmetry is always no. 1 (for C2v: a_1), the order of the remaining three irreps depend on your input, see symmetry information in the beginning of the output. CSF no. 1221 would look like this in Molcas style (if you have 5 2 3 2):

1221 222uu 02 00u 0u 0.951154 0.951154**2
(It looks like your CAS space really was 6 3 2 1 ...:
1221 222uu0 200 u0 u 0.951154 0.951154**2 )

-- Hans Jørgen.

-- Hans Jørgen.

footballplayer
Posts: 14
Joined: 01 May 2016, 16:51
First name(s): Nahn
Last name(s): Mahp
Affiliation: HCMUNS
Country: Viet Nam

Re: How to read CSFs in the output file of CASSCF calculatio

Post by footballplayer » 01 May 2016, 20:45

Dear Hans,

That is what I really need to know.
Thank you very much for your detailed explanation.

Nhan

taylor
Posts: 545
Joined: 15 Oct 2013, 05:37
First name(s): Peter
Middle name(s): Robert
Last name(s): Taylor
Affiliation: Tianjin University
Country: China

Re: How to read CSFs in the output file of CASSCF calculatio

Post by taylor » 02 May 2016, 05:15

It looks like your issues have been resolved, but for further reference you might want to look at the chapter in the manual that deals specifically with symmetry (there's also a "sticky" posting at the top of this "Running Dalton" forum). I say this because it is important to understand, as HJJ mentioned, that the labelling of some irreducible representations is a matter of convention. Take C2v as an example. The A1 irrep is uniquely defined, and so is A2 (antisymmetric with respect to both planes). But how to label B1 and B2? For a planar C2v molecule, like water, the convention introduced by Mulliken more than 60 years ago is that B2 is symmetric with respect to the molecular plane, and B1 is antisymmetric. But not everyone follows this convention! And then how to generalize it? What if we consider the molecule CH2Cl2? Which plane is which? It is essential (a) to define this clearly when you present results and (b) to be clear what convention other people, such as experimentalists you may be interacting with, are using. Do not just take the labels from the program (any program, not only Dalton) as "definitive", because they are not uniquely defined.

For the groups of order 2 (Cs, C2, Ci) there is obviously a unique definition. Further, for C2h there is a unique definition because A is symmetric and B antisymmetric with respect to the plane and g and u labels are always uniquely definable. But for C2v, D2, and D2h itself one has to define a convention. Dalton does that to the best it can, but I repeat what I said above, you have to look at the output yourself to see whether Dalton's "B1" is the same as your experimentalist colleague's "B1", or indeed what someone said was "B1" in a paper she published that you are reading.

Best regards
Pete
P.S. I'm always delighted to see people taking advantage of the fact that Dalton exploits symmetry the way it does, which is why I decided to post. But honesty requires me to admit that both the manual chapter and the sticky posting I referred you to were written by me...

footballplayer
Posts: 14
Joined: 01 May 2016, 16:51
First name(s): Nahn
Last name(s): Mahp
Affiliation: HCMUNS
Country: Viet Nam

Re: How to read CSFs in the output file of CASSCF calculatio

Post by footballplayer » 02 May 2016, 11:16

Dear Pete,

Thanks lots for your words. I'll pay attention to this notification when applying symmetry to the systems I am working with.

Best regards,

Nhan

hjaaj
Posts: 333
Joined: 27 Jun 2013, 18:44
First name(s): Hans Jørgen
Middle name(s): Aagaard
Last name(s): Jensen
Affiliation: Universith of Southern Denmark
Country: Denmark

Re: How to read CSFs in the output file of CASSCF calculatio

Post by hjaaj » 02 May 2016, 13:00

BY THE WAY.

Always when your run an MCSCF on the ground state of a molecule I strongly recommend that you start from MP2 orbitals. Input:

Code: Select all

.HF
.MP2
.MCSCF
also for open-shell cases. You should specify the singly occupied open shells in the HF input. The MP2 code will then freeze the open shell orbitals in the MP2 calculations, and give you very good starting orbitals for the MCSCF. If you did not do that before, it is very likely that you now will converge to a lower energy ! And thus a better description.

The MP2 start guess works well for many open shell cases in spite of the frozen open-shell orbitals (the implementation is really a closed shell MP2). This is because the biggest correlation effects are on the doubly occupied HF orbitals.

Hartree-Fock orbitals are generally very poor starting orbitals for MCSCF, which often will lead to convergence to local minimum and not the optimal MCSCF. This is because the lowest virtual orbitals are the most diffuse orbitals, which are not the correlating orbitals needed in the active space. The more diffuse your basis set (aug, daug, taug, ...) the worse is the HF start guess for MCSCF.

footballplayer
Posts: 14
Joined: 01 May 2016, 16:51
First name(s): Nahn
Last name(s): Mahp
Affiliation: HCMUNS
Country: Viet Nam

Re: How to read CSFs in the output file of CASSCF calculatio

Post by footballplayer » 02 May 2016, 15:47

Dear Hans,

This is a very useful tip. I am going to try this recommendation in future calculations. Usually, a starting from HF orbitals is used in almost my calculations, and then the orbitals in the CAS SPACE are checked manually.

Many thanks,

Nhan.

taylor
Posts: 545
Joined: 15 Oct 2013, 05:37
First name(s): Peter
Middle name(s): Robert
Last name(s): Taylor
Affiliation: Tianjin University
Country: China

Re: How to read CSFs in the output file of CASSCF calculatio

Post by taylor » 02 May 2016, 17:39

To add a bit more to HJJ's very wise advice here, one should think about what is needed to describe correlation and to describe near-degeneracy effects in the valence space. First, it is most likely that if we are, for example, breaking bonds, we should have both the bonding and the corresponding antibonding orbitals in the active space. Second, the simplest (HJJ suggests MP2) description of correlation, for two electrons in an occupied orbital "i", will be given by an expression that involves a sum over terms (ia|ia)2/E0, where the numerator here is the square of an exchange integral with virtual orbital "a" and the denominator is a sum of orbital energies. To maximize this energy contribution we want the numerator to be as large as possible and the denominator to be small. The latter is usually fulfilled for H-F calculations --- at least for the atomic case the "H-F limit" or "complete basis" answer will produce a virtual orbital space of continuum functions, starting at an orbital energy of zero (because the H-F virtuals are defined in terms of an N-electron potential and are thus basically anion orbitals; note that this is not the case for DFT). However, to make the numerator large we have to ask what is likely to give us a large exchange integral? The virtual orbital "a" should occupy the same spatial region as the occupied "i", but will necessarily, for reasons of orthogonality, have more nodes. This is almost never, in terms of spatial localization, the case for the H-F virtuals, because they're going to be more like plane waves, and the bigger and more diffuse the basis the worse this gets. At least if the system is neutral, and if it's an anion things get even worse!

In Dalton one approach that is straightforward to use, as HJJ has written, is to use the MP2 natural orbitals. But it may be useful to point to other strategies, partly because they may be useful with other codes. Many (i.e., >50...) years ago people realized that one approach to getting better "correlating orbitals" was to define the virtual space using an (N-1)electron potential. This gives much less diffuse, plane wave-like functions, and several programs can do this. Sometimes this approach is called "improved virtual orbitals" (IVOs). But a better tactic along these lines was suggested (more than 30 years ago...) by Charlie Bauschlicher --- run a calculation for starting orbitals that excludes all the valence orbitals, or a significant fraction. And this is usually straightforward to run as a closed shell. For example, for O2, start the calculation by getting SCF orbitals for O22+, and then use those for the MCSCF step. As I say, Dalton offers more flexibility than this, but it can be useful when one is using An Other Program...

Best regards
Pete

footballplayer
Posts: 14
Joined: 01 May 2016, 16:51
First name(s): Nahn
Last name(s): Mahp
Affiliation: HCMUNS
Country: Viet Nam

Re: How to read CSFs in the output file of CASSCF calculatio

Post by footballplayer » 04 May 2016, 15:04

Dear Pete,

Thanks a lot for your suggestion. It is so nice. I would love to apply all in my calculations. I'll see if I can simplify the procedure for obtaining the initial guess of orbital set. I should admit that at the beginning of a project, it usually takes me about couples of weeks for finding the best initial guess of orbitals.

And, I have one more thing would like to ask you here. I am doing some DMRG calculations. Is there pre-test of DMRG interface between Dalton and one other DMRG program. I would be happy if there is such an interface or a way to perform DMRG with Dalton then NEVPT2 on the basis of DMRG wavefunctions, because Dalton has a very efficient way to specify active orbitals in active space with regard to symmetry.

Thanks

Nhan
Last edited by footballplayer on 04 May 2016, 16:47, edited 2 times in total.

erik
Posts: 14
Joined: 12 Dec 2013, 12:43
First name(s): Erik
Middle name(s): Donovan
Last name(s): Hedegård
Affiliation: University of Southern Denmark
Country: Denmark
Contact:

Re: How to read CSFs in the output file of CASSCF calculatio

Post by erik » 04 May 2016, 16:26

Dear Nhan

Although this is now getting a bit off the original topic, I can say that
we are developing an interface between DMRG and DALTON in our group, and NEVPT2 is also on the list. The interface is working for DMRG. It might be a part of the next release, but this i cannot guarantee. The perturbation treatments will take longer, and will almost certainly not be part of next release.

with best regards
Erik

footballplayer
Posts: 14
Joined: 01 May 2016, 16:51
First name(s): Nahn
Last name(s): Mahp
Affiliation: HCMUNS
Country: Viet Nam

Re: How to read CSFs in the output file of CASSCF calculatio

Post by footballplayer » 04 May 2016, 16:47

Dear Erik,

Thanks for the useful information. I would be waiting for the next release.

Regards

Nhan

Post Reply

Who is online

Users browsing this forum: No registered users and 4 guests