Re: [dirac-users] GAS SHELLS for MCSCF vs KRCI

[Hans Jørgen Aagaard Jensen]

How big are the spikes? If you email 2 outputs from neighboring geomtries, one wirh a spike and one without, I can take a look at it. 

Hans Jørgen 

Sendt fra Outlook til iOS

---

Fra: dirac...@googlegroups.com <dirac...@googlegroups.com> på vegne af Ruairi Graham <ruairig...@gmail.com>
Sendt: Tuesday, December 2, 2025 11:38:31 PM
Til: dirac-users <dirac...@googlegroups.com>
Emne: [dirac-users] GAS SHELLS for MCSCF vs KRCI

 

Du får ikke ofte mails fra ruairig...@gmail.com. Få mere at vide om, hvorfor dette er vigtigt

Hi Dirac users,

I am trying to form PECs for a diatomic hydride. I have still not fully grasped the concept of gas shells. I have tried to follow a similar example to the water tutorial and my PEC curve from the MCSCF calculation looks very smooth. However, when i try and do KRCI on top of that with the +Q correction it completely messes up my PEC.

For MCSCF.inp 

**WAVE FUNCTIONS
.SCF
.PHCOEF
.RESOLVE
.KRMCSCF
*SCF
.MAXITR
300
.CLOSED SHELL
20
.OPEN SHELL
1
1/2
.FOCC
.AUTOCC
.EVCCNV
1.0d-5 1.0d-4
.ERGCNV
1.0d-6 1.0d-5
.FCKCNV
1.0d-5 1.0d-4
.MOSTART
SCRPOT
.DAMPFC
0.7
.LSHIFT
1.5
*KRMCSCF
.CI PROGRAM
LUCIAREL
.SYMMETRY
1
.PRINT
3  
.INACTIVE
9
.GAS SHELLS
1
3 3 / 4  
.THRESH
1.0D-7

and for my krci.inp

*KRCI
.CI PROGRAM
LUCIAREL  
.INACTIVE
9
.GAS SHELLS
2
1 3 / 4
3 3 / 63         ! GAS3: 63 virtual pairs (MRCISD into full virtual space)  
.CIROOTS
1 1
.MAX CI
200
.MXCIVE
50

.NOOCCN

.OMEGAQ
*QCORR
.REFFILE
.CVORB
9 0

The calculation works although it no longer makes my PEC look smooth and there are lots of different spikes in it. Is it a problem with how im setting the gas space in the KRCI step? My aim is to do MRSD after MCSCF. 

--
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 visit https://groups.google.com/d/msgid/dirac-users/691e9851-80aa-46b5-9a01-a17ee55fe39an%40googlegroups.com.

[Ruairi Graham]

Hi Hans,

Thank you for your quick response.
I will attach the 3.2 A and 3.4 A KRCI output files.

However, one thing to mention: I have tried a new active space which seems to have solved the problem with the spikes. 
.INACTIVE
0
.GAS SHELLS
3
18 18 / 9
19 21 / 4
21 21 / 63

However i do not really understand how this is much different from  

.INACTIVE                                                                                          
9                                                                                                  
.GAS SHELLS                                                                                        
2                                                                                                  
1 3 / 4                                                                                            
3 3 / 63  

Would you be able to explain why the three GAS shells setup works better because my current perception was that setting the first gas shell to 18 18 / 9 is the equivalent to setting 9 inactive orbitals. 

To view this discussion visit https://groups.google.com/d/msgid/dirac-users/DB9PR08MB7746E607163CA4CEACD87CB7D1D9A%40DB9PR08MB7746.eurprd08.prod.outlook.com.

[Hans Jørgen Aagaard Jensen]

Hi Ruairi

Neither of the two KRCI calculations you attached did converge, in fact, they are far from converged. Thus, no wonder there are spikes among from these non-converged energies. Both CI residuals are > 1.0, and therefor you have zero converged decimal places in the total energies!!  (Remember that for CI you have quadratic accuracy in the CI energy, if the wave function gradient (i.e. the residual) norm is 1.e-3, you will have approximately 6 decimal places converged in the energy.)

The convergence is extremely show. What orbitals do you have in the secondary space? Usually when one observes such a slow convergence, it is because the CI Hamiltonian is far from diagonally dominant, and then the Davidson algorithm will be very slow.

For an MR-CISD the best orbitals for the SD space to get a diagonally dominant CI matrix should be Fock-type orbitals, at least that is what Jeppe Olsen once told me.

Hans Jørgen.

To view this discussion visit https://groups.google.com/d/msgid/dirac-users/CAFO5yz5cKjYLg3hHWLzUuoL5o%2BLbeez%2BpFEsurjSsoenSNpGAw%40mail.gmail.com.