Help with CDFT-CI

Questions
#1

Dear all,
I need some help to setup a constrained DFT calculation: In particular, I want a DFT-single point calculation of Ar+ where all 3p-orbitals are degenerate (one is virtual and therefore has considerably higher energy in DFT) and according to my understanding, the CDFT-CI is exactly meant for this.

From the manual I understand it like this: I have three configurations which should contribute equally much, so the cdft-section should have the form

$cdft
STATE_1_CONSTRAINT_VALUE_X
0.333   1   1  TYPE1_X
-----
STATE_1_CONSTRAINT_VALUE_X
0.333  1   1    TYPE1_X
-----
STATE_1_CONSTRAINT_VALUE_X
0.333  1    1    TYPE1_X
$end

First: Is the coefficient 1/3 or should it be 1/sqrt(3)? Than, I have as first and last atom the ‘1’.

Each state is specified with the CONSTRAINT_VALUE and the corresponding weights on sets of atoms whose average value should be the constraint value

I am here irritated by the type and the constraint_value. What do I put there to specify the correct state? I’d think that I should tell it somehow that the ground, first and second excited states are relevant here?

Many thanks in advance for all kind of help
Tobias

#2

Hi Tobias,
There are a few misconceptions in your question.

First, your virtual orbital is higher in energy because your wavefunction is single-reference, not because it is DFT per se. (A Hartree-Fock calculations would likely further destabilize that virtual orbital.)

Regarding the CDFT constraints they are linear, so for mixing 3 constraints equally you would use 1/3 not 1/sqrt(3). However, the constraints in CDFT are based on counting electrons on atoms in fragments, whereas your system has only one atom and therefore cannot have more than one fragment.

I suggest that you try ROHF or ROKS instead. (UNRESTRICTED=FALSE in $rem)

#3

Hi @jherbert,
thank you very much for the clarification. I thought that this method would help me to circumvent the problems of the single-reference character (I agree; it is not per se a problem of DFT, but related to the single-reference character).

I’ll have a closer look at the ROKS. Thank you very much for that reference!
Have a nice day
Tobias

#4

Another option is EOM-IP-CCSD starting from closed-shell neutral Ar. Multireference wavefunctions are available also including several types of RAS and CAS.

#5

Thank you for these further suggestions.
Unfortunately, I must stick with DFT. But thank you very much anyway.