CASSCF Excited State Geometry Optimization - Error

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
renjith
Posts: 3
Joined: 18 Mar 2016, 22:34
First name(s): Renjith
Last name(s): B Nair
Affiliation: Texas A&M University
Country: United States

CASSCF Excited State Geometry Optimization - Error

Post by renjith » 19 Mar 2016, 17:11

Dear Dalton Users,

I am new to DALTON. I recently tried to optimize O2 molecule for the second excited state (symmetry = 1, spin multiplicity = 1) at the CASSCF level using 8 electrons in 10 orbitals. The input files (O2.dal and O2.mol) are attached herewith for more details.

However, the program ends with severe error which is stated as follows. (I attached the output file, O2.out, herewith).

------------------------------------------------------------------------------------------------
" --- CICHCK option 2, cannot proceed:
number of trial vectors of right symmetry is 1
which is less than current reference state no. 3

--- SEVERE ERROR, PROGRAM WILL BE ABORTED ---

Reason: --- CICHCK: too few trial CI vecs. of right symmetry "

------------------------------------------------------------------------------------------------

I kindly request you to please help me to figure out this issue. I searched throughout the topics in the forum and have not found a proper solution. Looking forward to hearing from you.

Thanks in advance,
Regards,
Renjith
Attachments
O2.out
Output file with the error message
(30.45 KiB) Downloaded 292 times
O2.mol
Molecule Specification
(195 Bytes) Downloaded 332 times
O2.dal
Dalton Input file. Second excited state geometry optimization at CASSCF (8,10). Two orbitals were interchanged. Overall spin multiplicity is 1.
(319 Bytes) Downloaded 357 times

taylor
Posts: 589
Joined: 15 Oct 2013, 05:37
First name(s): Peter
Middle name(s): Robert
Last name(s): Taylor
Affiliation: Tianjin University
Country: China

Re: CASSCF Excited State Geometry Optimization - Error

Post by taylor » 19 Mar 2016, 19:50

Your Email is not entirely clear, because you start by saying you want an excited state of symmetry 1, yet you have specified "Nosymmetry" (almost always a confusing and therefore bad thing to do with a diatomic) and so it is not possible to get any other states than symmetry 1! If I take a guess, you are saying you are trying for the 1Sigma_g^+ state (the second excited singlet state)? In that case you should come to terms with the symmetry. Specify the generators as X Y Z, giving you D2h as the point group, and specify one symmetry-independent oxygen atom (where did you get your bizarre O2 coordinates from ?!) at a point 0 0 z, where "z" is half the bond-length. You should be able to now figure out your desired occupancy but I give you here, for the ground state, the HF occupancy as

Code: Select all

.DOUBLY OCCUPIED
  3  1  1  0  2  0  0  0
.SINGLY OCCUPIED
  0  0  0  0  0  1  1  0
which corresponds to the 3Sigma_g^- ground state.

Now, in your MCSCF you want the second excited state in symmetry 1, with spin multiplicity 1. Not a third state, which would be some sort of excited singlet, because the ground state is a triplet and not part of this manifold. And the use of full symmetry will avoid any complications with singlet states of other symmetries.

Finally, a ten-orbital active space for O2 is almost certain to be not size-consistent. That is, it cannot display the correct behaviour at dissociation. Irrespective of whether you want to study dissociation, and irrespective of how many authors may have used such an active space, it is not size-consistent (and let me say parenthetically I am very well aware of the distinction between size-consistent and size-extensive, and am quite confident about my usage here) and should be viewed with some suspicion.

And finally finally, if you use symmetry, there should be no need for any "REORDER"ing of MOs.

Best regards
Pete

hjaaj
Posts: 371
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: CASSCF Excited State Geometry Optimization - Error

Post by hjaaj » 19 Mar 2016, 22:45

Dear Renjith,

As Peter Taylor writes, I would also do the calculation using symmetry. However, you can do it without using symmetry as you tried to do, BUT, when you ask for an excited state, the program makes a symmetry check anyway (unless you ask it not to). You can see in your output that the initial CI gave
@ Final CI energies and residuals in symmetry 1 (irrep A )
@ 1 -149.663971638221000 3.95D-02
@ 2 -149.663909089510298 4.12D-02
@ 3 -149.630847527014339 4.54D-02
@ 4 -149.399464553753575 7.30D-02
I note that roots 1 and 2 are degenerate, fitting with the Delta state, and the third is then indeed the Sigma state you were looking for. If you look in the Dalton manual (page 287, read also the comment on page 289) you will see that you can get the desired root with

Code: Select all

*OPTIMIZATION                                     
.STATE                                            
3                                                 
.SYM CHECK
 1
in your input. (You can also disable symmetry check with ".SYM CHECK" option -1, but then you let Dalton use more CPU time because it then includes the two irrelevant Delta components in the optimization iterations.) The default is chosen because that was what we most often wanted, but not so in your case.

-- Hans Jørgen

renjith
Posts: 3
Joined: 18 Mar 2016, 22:34
First name(s): Renjith
Last name(s): B Nair
Affiliation: Texas A&M University
Country: United States

Re: CASSCF Excited State Geometry Optimization - Error

Post by renjith » 21 Mar 2016, 21:09

Thanks to both Prof. Taylor and Prof. Jorgen for your kind reply.

@Prof. Taylor

First of all, I prepared the input geometry using AVOGADRO software and I kept the distance between oxygen atoms to 1.1 Angstrom. I did a series of geometry optimization for ground state geometry using CASSCF algorithm. I have copied one of those input (and therefore, I didn't notice the "Nosymmetry" option) for the second excited state calculation. Thanks for pointing out the this mistake. Further, I want to optimize for the 1Sigma_g^+ state (the second excited singlet state) that you have guessed correctly.

However, I am not following your statement " Specify the generators as X Y Z, giving you D2h as the point group, and specify one symmetry-independent oxygen atom." As I stated earlier, I used the same input geometry for both the ground state and excited calculations. D2h point group is identified in ground state calculations.

@Prof. Jorgen

I will go through the DALTON manual as suggested by you and will let you know the status of the calculation.

Regards,
Renjith

taylor
Posts: 589
Joined: 15 Oct 2013, 05:37
First name(s): Peter
Middle name(s): Robert
Last name(s): Taylor
Affiliation: Tianjin University
Country: China

Re: CASSCF Excited State Geometry Optimization - Error

Post by taylor » 22 Mar 2016, 19:56

Whether you are studying a ground state or excited state has nothing to do with the symmetry analysis or handling in the program, which is done much earlier. I was suggesting that for a small, highly symmetrical system it is much better to take charge oneself (there are multiple forum postings and a complete chapter in the manual about this, some of it written by me so I'm not prepared to repeat it here!) in treating the symmetry. If you specify the symmetry using the Generators keyword you need only (and indeed you must only) specify a single O atom, since the program will then generate its symmetry-related partner. Convention would dictate your unique O having coordinates 0.0 0.0. X.YYYYY, where X.YYYYY is half the bond distance.

As we have stated, in D2h symmetry the two lowest singlets, 1Deltag and 1Sigmag+ will appear in symmetries 1 and 4 for the former, and symmetry 1 for the latter (this is all explained elsewhere and you can read about it there --- one posting is actually "sticky" at the top of the forum...). As HJJ pointed out if you run in no symmetry the code detects the two degenerate components of the 1Deltag and immediately tries to ensure that it maintains that symmetry, which is not what you want. If you calculate a root in symmetry 1, the first root will be one component of the Delta, but you will not see the other and the program will not find any degeneracy.

I assume that you are trying something simple before starting a more complicated study. i mention this because it is almost always less efficient to use gradient (or gradient/Hessian) based geometry optimization for a diatomic. Calculating energies and fitting three points by hand to a polynomial in the reciprocal of the bond distance and cycling through this twice is likely to use less computer time, although again for a diatomic this probably hardly matters.

Best regards
Pete

renjith
Posts: 3
Joined: 18 Mar 2016, 22:34
First name(s): Renjith
Last name(s): B Nair
Affiliation: Texas A&M University
Country: United States

Re: CASSCF Excited State Geometry Optimization - Error

Post by renjith » 22 Mar 2016, 20:08

In continuation with my previous e-mail, I now would like to inform you that I modified the input geometry for the O2 molecule as suggested by Prof. Taylor which is attached herewith (O2_new.mol) for your quick reference. At the same time, I also tried to incorporate the CAS space with its symmetry (O2_new.dal) concerns. I believe the CAS spaces are now in correct symmetry. I re-run the calculations and got the following error

" --- SEVERE ERROR, PROGRAM WILL BE ABORTED ---
Date and time (Linux) : Tue Mar 22 13:46:55 2016

Reason: Sorry, backstep only implemented for NEO
"

However, I activated the .NEO ALWAYS option and the error still remains. For more details, the output file (O2_new.out) is attached below.

Thank you in advance,
with regards,
Renjith
Attachments
O2_new.out
The latest output file terminated with an error for the second excited state geometry optimization for O2 molecule
(416.36 KiB) Downloaded 310 times
O2_new.dal
The modified Dalton Input file for the O2 molecule for second excited geometry optimization at CASSCF (8,10) level
(404 Bytes) Downloaded 313 times
O2_new.mol
The modified input for O2 Molecule geometry specification
(185 Bytes) Downloaded 282 times

taylor
Posts: 589
Joined: 15 Oct 2013, 05:37
First name(s): Peter
Middle name(s): Robert
Last name(s): Taylor
Affiliation: Tianjin University
Country: China

Re: CASSCF Excited State Geometry Optimization - Error

Post by taylor » 22 Mar 2016, 20:33

Your active space is completely symmetry broken and if you look at the CASSCF natural orbital occupation numbers you can see they are not remotely plausible for the 1Sigmag+ state of O2. That state is derived from an occupation of pi_g^2, so one would naively expect that the occupation pattern would very much echo that from Hartree-Fock considerations, with doubly occupied orbitals that are g and u combinations of 1s and of 2s, taking care of eight electrons between them. Then we have a sigma bonding pair and a pi bonding quartet (degenerate pi orbitals), with the last two electrons going into the pi_g. Again, this would have a pattern of doubly occupied orbitals

Code: Select all

3 1 1 0 2 0 0 0
and singly occupied

Code: Select all

 0 0 0 0 0 1 1 0
Your active space has two active orbitals in symmetry 3 but only one in symmetry 2: since these are the pi_u orbitals this is immediately a problem with the symmetry --- the pi_x and pi_y components are not treated equivalently. The same is true for the pi-g orbitals in symmetry 6 and 7. Finally, you have orbitals in the active space with occupation numbers like 0.001: this is an immediate sign of problems with the choice of active space. If the occupation number is even 0.01 or less the orbital should not be in the active space, and if the occ no is much greater than 1.98 it should be inactive, not active.

I suggest as a starting point --- you can get fancier later if need be --- you make the oxygen-derived 2p orbitals active and the 1s and 2s inactive. This would require

Code: Select all

.INACTIVE SPACE
  2  0  0  0  2  0  0  0
.CAS SPACE
  1  1  1  0  1  1  1  0
.ELECTRONS
  8
As I say, once this works, you can if desired start to extend this. But you must do this in a symmetry-balanced way.

Incidentally, the radius of convergence of the Dalton MCSCF algorithm is huge and the rate of convergence excellent. A calculation in a simple diatomic, in the lowest state of its particular symmetry (in the full point group), that takes 27 iterations to converge is another sign that something is wrong.

Best regards
Pete

taylor
Posts: 589
Joined: 15 Oct 2013, 05:37
First name(s): Peter
Middle name(s): Robert
Last name(s): Taylor
Affiliation: Tianjin University
Country: China

Re: CASSCF Excited State Geometry Optimization - Error

Post by taylor » 23 Mar 2016, 12:28

I have only just noticed you still have

Code: Select all

.STATE
 3
specified in your input. This is also a problem (but note that everything I said about your choice of active space in the previous posting is, I believe, still valid --- what I am about to write does not change any of it). The 1Deltag state is doubly degenerate with one component in symmetry 1, and one is (with the usual conventional choice of specifying the generators as X Y Z and the nuclei on the z-axis) in symmetry 4. The 1Sigmag+ is therefore the second singlet state in symmetry 1, the one component of the 1Deltag being the first singlet state in symmetry 1. I have no idea for certain what the third state might be, but I am convinced it is not anything you want! And anyway the use of an active space that is not balanced across the irreps makes it largely hypothetical. The symmetry chapter in the Dalton manual will show you which higher-symmetry irreps correspond to which D2h irreps, both for atoms and diatomics, and should help ensure you treat both components of degenerate orbitals equally.

Best regards
Pete

Post Reply

Who is online

Users browsing this forum: No registered users and 6 guests