You only mentioned you need to evaluate excitation energy. If you also need to calculate interaction energy, then diffuse functions will be needed for 2- or 3-zeta basis sets.

About question 1: This is actually what I am most confused, because I thought I was correct based on your blog posts. Could you please correct me if I’m interpreting your texts in a wrong manner? (or wrong translation?) See below:

In your blog post “On the diffuse function and the "month" basis set (//www.umsyar.com/119)”,

1. When is a diffusion function needed?
…
• Adding dispersion is very important and strongly recommended:
Calculate the weak interaction energy (if using a 4-zeta level basis set and not involving anions, then dispersion is not necessary)
…
• Adding dispersion can improve results, and should be done
optimize the structure of anionic or weakly interacting systems;
...

Also in “On the choice of basis sets in quantum chemistry (//www.umsyar.com/336)”,

2. Selection of different task basis sets
2.2 Weak Interactions
…
The greater the proportion of dispersive interactions, the more necessary a diffuse function is required, and the more important the BSSE problem needs to be considered.

Basis set recommendations for weak interaction energy calculations:
Minimum: 6-31+G** or ma-def 2-SVP for DFT
…

Many thanks in advance!

1 This does not constitute a necessity for adding diffusion functions.

2 When diffuse functions are added, LUMO and/or other unoccupied MOs may be quite diffuse, in this case they do not have clear chemical meaning (this phenomenon may be more obvious when HF composition of the employed DFT functional is relatively high).

I’d be really grateful if you could provide some additional feedbacks, basically whether you agree to the following statements.

1. The structure (of which I’m studying the CT state) is a bimolecular interacting complex, which features intermolecular pi-stacking between two aromatic groups, and CT occurs between these two rings.  In this case, even without significantly negatively charged atoms, it is recommended to add diffuse functions.
2. In the case of 1. or in case there are actually highly negatively charged atoms, I just have to accept that the method I have to use to get accurate energy (using diffuse functions) is not going to permit analysis of correct LUMO or LUMO+1.

Thank you very much.

CT excitation doesn't necessarily lead to heavily negatively charged atoms. However, when it is indeed true, adding diffuse functions for those atoms may improve result.

Thank you for the reply,

If you don’t mind, could you help me understand this in a little more detail? Because I thought that in a charge-transfer state where certain atoms (that received charge) are significantly negatively charged, at least those atoms should need to be treated with diffuse functions.

You do not need to add diffuse functions for common CT states. Only Rydberg states, and excitation state calculations for negatively charged system, need diffuse functions.

Dear Prof Lu,

I have a neutral doublet structure (D0) whose first excited state (D1) turns out to be a charge-transfer state, which shows charge-separation (significantly negative charges on specific atoms and positive charges on other atoms). From what I learned, I would need to add diffuse functions to accurately measure the excitation energy.

At the same time, I want to study the D1 state's electronic state, examine molecular orbitals, etc. But I noticed that with diffuse functions used, the alpha-HOMO in the D1 state (alpha-LUMO in the D0 state) is severely broken in shape. Only when I do not add diffuse functions I see excitation to the correctly shaped orbital.

How should I resolve both energy evaluation and accurate wavefunction analysis at the same time?