Fock space does not converge

40 views
Skip to first unread message

Maksim Shundalau

unread,
Aug 4, 2021, 10:47:26 AMAug 4
to dirac...@googlegroups.com
Dear Dirac experts,

I’m trying to calculate PECs of the low-lying doublet states of the
LiSr molecule, which belong to the first three dissociation limits:

the first is Li(^2S)+Sr(^1S) [X^2Sigma+, 0.0 cm-1];

the second is Li(^2S)+Sr(^3P) [(2)^2Sigma+, (1)^2Pi, 14317-14898 cm-1],

the third is Li(^2P)+Sr(^1S) [(3)^2Sigma+, (2)^2Pi, 14903-14904 cm-1].


I use FS-CCSD/cc-pCVTZ level of theory (sector 0,1).



In the range of the small internuclear distances (3.0-5.0 A) solution
converges, and PECs’ energies are close to the experimental/previous
theoretical ones. But for the larger internuclear distances (7.0 A)
solution diverges. If I use an intermediate Hamiltonian, the solution
converges very slowly. I think, if I will use other IH parameters, the
solution will converge faster.



But the main problem is:

The energies of three states are too overestimated (19049-19748 cm-1).
It looks like ostensibly the third dissociation limit (which arises
from the excitation of a lithium atom) is missed, and the calculation
reproduces the second dissociation limit and some of the high-lying
ones.


I have used different basis sets and IH schemes, but obtained the
similar results.


What am I doing wrong? Is there any way to fix it?



Thanks in advance,

Maksim.
lisr_lisr-3-500-large.out
lisr_lisr-7-000-large.out
lisr-ih1_lisr-7-000-large.out

Peterson, Kirk

unread,
Aug 4, 2021, 12:43:13 PMAug 4
to dirac...@googlegroups.com
Dear Marksim,

I think it is simply because your closed-shell sector (0,0) reference state cannot dissociate correctly for the 2 dissociation asymptotes that involve two open-shell fragments.

best regards,

-Kirk
--
You received this message because you are subscribed to the Google Groups "dirac-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dirac-users...@googlegroups.com.
To view this discussion on the web visit https://urldefense.com/v3/__https://groups.google.com/d/msgid/dirac-users/CAEbGO5i39So-z3YjqfhJ_5OyANxctLQawapOTHe_bxdhUcVmug*40mail.gmail.com__;JQ!!JmPEgBY0HMszNaDT!5LSh1b0t7Mm2BF0NBP4hVpJKS1jh5RJlYOkSCjDW_xL0d966TaRFdQxpkhDp1p0$ .

Ilias Miroslav, doc. RNDr., PhD.

unread,
Aug 4, 2021, 12:48:26 PMAug 4
to dirac...@googlegroups.com
Hello Maxim,

I would try combination (GAS) KRMCSCF + KRCI to get proper (MCSCF))orbitals for the whole dissociation curve as well as to include enough of correlation via (GAS) KRCI .

Miro


From: 'Peterson, Kirk' via dirac-users <dirac...@googlegroups.com>
Sent: Wednesday, August 4, 2021 18:43
To: dirac...@googlegroups.com <dirac...@googlegroups.com>
Subject: Re: [dirac-users] Fock space does not converge
 

Peterson, Kirk

unread,
Aug 4, 2021, 12:59:55 PMAug 4
to dirac...@googlegroups.com

Miro,

 

but isn't the KRMCSF code limited to a single state?

 

-Kirk

Ilias Miroslav, doc. RNDr., PhD.

unread,
Aug 4, 2021, 1:09:32 PMAug 4
to dirac...@googlegroups.com
Indeed, it is, but the following KRCI - upon good molecular spinors - can describe multiple states, http://diracprogram.org/doc/master/manual/wave_function/krci.html#ciroots  .

M


From: 'Peterson, Kirk' via dirac-users <dirac...@googlegroups.com>
Sent: Wednesday, August 4, 2021 18:59

Peterson, Kirk

unread,
Aug 4, 2021, 1:16:32 PMAug 4
to dirac...@googlegroups.com

Maybe, but I would be worried that as I neared dissociation my underlying spinors are going to be strongly biased towards a particular dissociation channel (one of three in the current case) of the state I chose for the KRMCSCF.  That's then asking a lot for the KRCI to make up for (particularly when you typically have to truncate the virtual space as well to make it manageable). You will certainly never recover the correct degeneracies at long range with this approach.

 

A state-averaged KRMCSCF would be really nice.  :)

 

best regards,

Maksim Shundalau

unread,
Aug 4, 2021, 3:35:41 PMAug 4
to dirac-users
Dear Kirk and Miroslav,

Thank you so much. Firstly I will try SA-CASSCF with 10 states (7 doublets and 3 quartets) and 3 active electrons via GAMESS/Firefly in order to see what's going on. 

Yours,
Maksim. 

среда, 4 августа 2021 г. в 20:16:32 UTC+3, Kirk Peterson:

Knecht Stefan

unread,
Aug 4, 2021, 6:26:32 PMAug 4
to <dirac-users@googlegroups.com>
Dear Kirk, dear Miro,

my few cents: I agree with Kirk that a state-specific KRMCSCF (for the ground state) will NOT be suitable in this case with an ensuing KRCI targeting an ensemble of states in the dissociation channel. 
To recover the correct degeneracies, one would probably have to run a state-specific KRMCSCF for ALL states in question + state-specific KRCI for the dynamical e-correlation 
(which would be possible to a certain extent with the existing code in DIRAC) or, as Kirk suggested, go for a state-averaged KRMCSCF in the first place. 
I have the latter on my radar and it might be something to enter DIRAC23 (certainly not earlier) in a simple-minded implementation. 

with best regards
Stefan


On 4 Aug 2021, at 19:16, 'Peterson, Kirk' via dirac-users <dirac...@googlegroups.com> wrote:

Maybe, but I would be worried that as I neared dissociation my underlying spinors are going to be strongly biased towards a particular dissociation channel (one of three in the current case) of the state I chose for the KRMCSCF.  That's then asking a lot for the KRCI to make up for (particularly when you typically have to truncate the virtual space as well to make it manageable). You will certainly never recover the correct degeneracies at long range with this approach.
 
A state-averaged KRMCSCF would be really nice.  :)
 
best regards,
-Kirk
 
From: <dirac...@googlegroups.com> on behalf of "Ilias Miroslav, doc. RNDr., PhD." <mirosla...@umb.sk>
Reply-To: "dirac...@googlegroups.com" <dirac...@googlegroups.com>
Date: Wednesday, August 4, 2021 at 10:09 AM
To: "dirac...@googlegroups.com" <dirac...@googlegroups.com>
Subject: Re: [dirac-users] Fock space does not converge
 
Indeed, it is, but the following KRCI - upon good molecular spinors - can describe multiple states,http://diracprogram.org/doc/master/manual/wave_function/krci.html#ciroots  .


--

---------------------------------------------------------------------------
PD Dr. Stefan Knecht
---------------------------------------------------------------------------
GSI Helmholtzzentrum für Schwerionenforschung GmbH
Abteilung SHE Chemie
Planckstr. 1
64291 Darmstadt
Germany

room: SB3 3.163
phone: +49 6159 71 2588
web: http://stefanknecht.xyz
email: s.kn...@gsi.de
---------------------------------------------------------------------------
GSI Helmholtzzentrum für Schwerionenforschung GmbH
Planckstraße 1, 64291 Darmstadt, Germany, www.gsi.de

Commercial Register / Handelsregister: Amtsgericht Darmstadt, HRB 1528
Managing Directors / Geschäftsführung:
Professor Dr. Paolo Giubellino, Dr. Ulrich Breuer, Jörg Blaurock
Chairman of the GSI Supervisory Board / Vorsitzender des GSI-Aufsichtsrats:
Ministerialdirigent Dr. Volkmar Dietz
---------------------------------------------------------------------------
Johannes-Gutenberg Universität Mainz
Departement Chemie
Duesbergweg 10-14
55128 Mainz
Germany
---------------------------------------------------------------------------

Johann Pototschnig

unread,
Aug 5, 2021, 4:54:14 AMAug 5
to dirac-users
Dear Maksim,

If you start from a closed shell and add one electron you cannot get a state with an open shell on the Sr.
I would suggest to first try to compute the transitions for the atom and consider which state you can compute with a certain reference / sector of the Fock space.
An way to get different states would be to start with the anion and use the (1h,0p) sector.

You could also use EA-EOM-CC which would allow for this state, but the deviations would be large.

You might take a look at https://arxiv.org/abs/2107.11234 in order to see which state you can get with which approach and what accuracy you can expect, although is a more complicated system.

best,
Johann Pototschnig

Maksim Shundalau

unread,
Aug 5, 2021, 5:15:33 AMAug 5
to dirac-users
Dear  Johann,

Thanks a lot. I will try. 
Yours,
Maksim.
четверг, 5 августа 2021 г. в 11:54:14 UTC+3, pototschn...@gmail.com:

Maksim Shundalau

unread,
Oct 8, 2021, 3:58:45 PM (7 days ago) Oct 8
to dirac-users

Dear Dirac experts,

 

I’m still trying to calculate PECs of the low-lying doublet states of the LiSr molecule, which belong to the first three dissociation limits (see above).

 Now I use KRCI/cc-pCVTZ level of theory (spin-free), but I have the same problem: the energies of three states, which should belong to the third dissociation limit (Li(^2P)+Sr(^1S), 14903-14904 cm-1), are too overestimated (20037-20684 cm-1).

 Could this be related to incorrect choice of the GAS shells? Or is it something else?

 

Thanks in advance,

Maksim.


четверг, 5 августа 2021 г. в 12:15:33 UTC+3, Maksim Shundalau:
lisr_lisr-10-000.out

Johann Pototschnig

unread,
Oct 11, 2021, 2:40:25 AM (5 days ago) Oct 11
to dirac-users
1. I would recommend running without ECP.

2. If you look at your output you have the following states:
    1S(Sr)/2S(Li) _1/2   0          -37.7795980343  0.000000000000            0.000000000000              0.000000000000      (1   )         
    1S(Sr)/2P(Li)_1/2    1          -37.7142594371  0.065338597186            1.777953621312          14340.164525753595      (1   )
    1S(Sr)/2P(Li)_3/2    2          -37.7140755141  0.065522520255            1.782958422475          14380.530973583311      (3   )
    1S(Sr)/2P(Li)_3/2    3          -37.7140755140  0.065522520295            1.782958423568          14380.530982397384      (1   )
    3P(Sr)/2S(Li)_1/2    4          -37.7131849826  0.066413051735            1.807191016035          14575.980041752462      (1   )
    3P(Sr)/2S(Li)_5/2     5          -37.7131849825  0.066413051808            1.807191018033          14575.980057864814      (5   )
    3P(Sr)/2S(Li)_5/2     6          -37.7131849822  0.066413052123            1.807191026588          14575.980126869405      (1   )
    3P(Sr)/2S(Li)_5/2     7          -37.7131849822  0.066413052147            1.807191027254          14575.980132240189      (3   )
    3P(Sr)/2S(Li)_3/2     8          -37.7131771113  0.066420923066            1.807405205841          14577.707599219888      (1   )
    3P(Sr)/2S(Li)_3/2     9          -37.7131771112  0.066420923115            1.807405207172          14577.707609959896      (3   )
   ?  10          -37.6883006635  0.091297370798            2.484327763651          20037.456792803321      (3   )
   ? 11          -37.6883006635  0.091297370799            2.484327763666          20037.456792926518      (1   )
   ? 12          -37.6853529361  0.094245098214            2.564539504631          20684.408180491184      (1   )

 For 2P(Li) the splitting of the states is overestimated, while for Sr you see none at all due to the use of ECP.


I hope this helps,
Johann

Maksim Shundalau

unread,
Oct 11, 2021, 9:41:59 AM (5 days ago) Oct 11
to dirac-users

Dear Johann,

 

Thanks for your answer.

The above calculation was performed with spin-free ECP for the Sr atom especially in order to see main regularities. I think that origin of the 20037-20684 cm-1 splitting is rather splitting between Pi (first two values) and Sigma (last value) molecular terms, than atomic (Li) SOC splitting. At least, for a distance of 20 A, this splitting is absent.

 

I am attaching two more files. The first of them is a calculation with spin-orbit ECP for the Sr atom, and the second one is calculation within an all-electron basis set.

 

 

Thanks in advance,

Maksim.


понедельник, 11 октября 2021 г. в 09:40:25 UTC+3, pototschn...@gmail.com:
lisr-all-el_lisr-all-el-10-000.out
lisr_lisr-so-10-000.out
Reply all
Reply to author
Forward
0 new messages