The book chapter I mentioned above has been published, see:
Tian Lu,* Qinxue Chen, Partial Charges, Chapter 6 of Exploring Chemical Concepts Through Theory and Computation, pp. 161-187. WILEY-VCH (2024). DOI: 10.1002/9783527843435.ch6

Dear Prof Tian Lu,

Thank you so much for your reply that's clarified a lot for me !!

Thanks a lot

Have a good week-end

My best

Alessio

Dear Alessio Macorano,

Usually QTAIM charge is not recommended, not only its evaluation is expensive, but also its result is often contradict with chemical intuition. For example, at B3LYP/def2-TZVP level, AIM charge of C and H in CH3NO2 is 0.233 and 0.080, which is not in line with electronegativity relationship between C and H. Another example, it is known that CO has dipole moment almost zero, however the AIM charge of C is as high as 1.201.

Because atomic charge is a quite simple model and nonobservable quantity, there does not exist a perfect atomic charge calculation method, the best choice is highly dependent of the use of atomic charges.

I recently finished writing a chapter in a monograph (Chapter 6: Partial charges), in which I overviewed almost all methods in calculating atomic charges, this book will be published by Wiley soon. At the end of this chapter, I give the following suggestion:

1.    For studying charge distribution characteristics of molecular systems, ADCH is a good choice, HI, MBIS and NPA can also be considered simultaneously. Although there are some known issues in Mulliken method, it is still useful for rough discussions and comparisons as it is the cheapest and the most widely supported method by calculation programs; however, the use of diffuse functions must be avoided in this case.
2.    In the case of employing partial charges in MD simulations based on classical force fields, for rigid molecules, CHELPG is generally satisfactory; for flexible molecules, RESP and RESP2 methods are more suitable; for solid surfaces and porous systems, REPEAT is our most recommended method; for dense solids, MBIS and DDEC6 may be preferential choices. To quickly generate partial charges of a very large organic system, EEM based on suitable parameters and MMFF94 are very useful.
3.    For predicting reaction sites and discussing reactivity, Hirshfeld charge is robust and highly recommended.

Best

Tian

Dear Prof Tian Lu,

I'm wondering if I could ask you one question about the population analysis, because i'd like to obtain the partial charges.

Is the QTAIM very reliable respect to the NPA, Mulliken and NBO analysis ?
moreover considering one molecules and trying different population analysis methods, how to know whcih could be the best ?

Thank a lot

My best

Alessio Macorano