Direct way to perform FCI calculations?

[ststober]

Hello,

I am new to Dalton... I typically use Psi4. I need to run some reference calculations on small molecules using FCI. There is a simple way to do this in Psi4 (there is an FCI keyword to use). However, I can't seem to find a direct way to do this in Dalton.

What I have attempted so far is below (note the ????? in the .GAS SHELL section, as I am not sure what to put here for this example).

Any help on the correct (and most direct) input to use to execute an FCI calculation on the *.mol input provided would be very much appreciated.

Thanks in advance, Spence

Input is below:

--- beh2.mol ---
BASIS
STO-3G
BeH2 using the STO-3G basis
Using automatic symmetry detection.
Atomtypes=2 Angstrom
Charge=4.0 Atoms=1
Be 0.0 0.0 0.0
Charge=1.0 Atoms=2
H_a 0.0 0.0 1.2906
H_b 0.0 0.0 -1.2906
---------------------------

--- beh2_fci.dal ---
**DALTON INPUT
.OPTIMIZE
**WAVE FUNCTIONS
.HF
.GAS-CI
*LUCITA
.INIWFC
HF_SCF
.CITYPE
GASCI
.MULTIP
1
.GAS SHELLS
?????
?????
**END OF DALTON INPUT
-----------------------

[kennethruud]

I am not too familar with the GAS-CI module, and in particular whether geometry optimization would work with this module.

What I would normally do in order to perform a FCI calculation with geometry optimization would be to run a regular MCSCF calculation, use no inactive orbitals (or keep the Be 1s orbitals inactive, keyword .INACTIVE), have 6 (or 4) active electrons (.ELECTRONS), placing all (or all except Be 1s) orbitals in the .CAS SPACE. See pages 13/14 in the Dalton manual for a valence FCI example if the CAS space contains all orbitals except the core orbitals.


Best regards,

Kenneth

[reinholdt]

You can try the FULLCI option under *CI INPUT,
though I am not sure if geometry optimizations will work.

Code: Select all

**DALTON INPUT
.RUN WAVEFUNCTIONS
**WAVE FUNCTIONS
.CI
*CI INPUT
.FULLCI
**END OF DALTON INPUT

[ststober]

Thanks you both for the ideas. I will give these suggestions a try. Best regards, Spence

[hjaaj]

The suggestion by Peter uses the CI module for MCSCF, which is not efficient for large number of active orbitals.

That is the GAS-CI / LUCITA as you specified in your example. No one have implemented a .FULLCI or .FCI for LUCITA ( you are welcome to do it on gitlab ). To use the .GAS SHELLS for 6 electrons in (4, 2, 2, 2) orbitals for a case with 4 irreps:

Code: Select all

.GAS SHELLS
 1
  6 6 / 4 2 2 2
  

A full CI corresponds to one GAS space with min and max number of electrons equal to the total number of electrons. It is also easy to freeze core orbitals in the CI by excluding the core electrons and corresponding core orbitals from the GAS space.

Probably the easiest way to find the correct number of orbitals in each symmetry is to perform a test calculation which stops after reading the molecule input. You could also use a two step procedure: first do the HF calculation, read the info about total number of orbitals and distribution over symmetries from there, write the .GAS SHELLS input, do the GAS-CI/LUCITA starting from the saved converged SCF orbitals.

[ststober]

Thank you Hans Jørgen for your reply. I have just tried Peter's approach and it works very well (thanks Peter!), but I will be sure to try your suggestion if I have scaling issues with larger active spaces.