Q: Appropriate method for spin population

Questions
#1

Greetings,

I want to get a spin population of radical cations.

Once I relied on Mulliken spin population,
but some article says, Mulliken spin population is not very reliable.

So, I found another ways in manual; Löwdin and Hirshfeld and NBO.

My molecule has no experimental value to refer to,
so there is no way to determine which method fits better.

My molecule has no experimental value to refer to, so there is no way to determine which method fits better.

I would appreciate it if you could tell me the proper calculation method under general condition.

Thanks

#2

I think Mulliken can be fine provided you’re aware of the fact that the numbers vary strongly with basis set and the most reliable values are probably those obtained in small basis sets lacking diffuse functions, e.g., 6-31G*. We used it here, precisely for spin populations, because it’s easy (and because we wanted to make contact with some previous work):

Hirshfeld charges are likely to be much more stable with respect to basis sets and the algorithm somehow feels more physical (space is weighted and assigned to atoms based on superposition of atomic densities), although it can be proved that these charges give you the least change with respect to free-atom charges. See Section II of this paper for some references:
https://doi.org/10.1021/acs.jpca.0c11356
The only way to “see” the spin per se is to compute something observable like an ESR spectrum, these charges are just models but they can be useful for comparing how the spin charge changes as a function of molecular structure (e.g., for a series of related compounds or the same molecule at different geometries).

By the way, were you able to get spin charges out of Q-Chem?

1 Like
#3

Thank you for your kind reply.

But I can’t understand “By the way, were you able to get spin charges out of Q-Chem?” your question.

As you know, I can get spin density and partial charge from output.

Is there another meaning?

And I have one more question.
Would you do me a favor?

  1. (RI)MP2 with small basis set
  2. (RI)MP2 with large basis set
  3. QCISD with small basis set
  4. w-B97XD with large basis set

Which one is better or worst way for spin density?

#4

with those first three methods, what you are actually get is probably the Hartree-Fock charges. Given the known problems with delocalization of unpaired spins in DFT, this may be a better choice (although hybrid and range-separated functionals suffer less from this problem, as compared to GGAs).

#5

Really helpful, thanks!