Multiwfn forum

wham09

Member

Registered: 2025-09-19

Posts: 62

Basis functions for 6-31G* used to fit SMD

Dear Prof. Lu,

I have so far evaluated SMD dissolution free energies of organometallic compounds using M05-2X/genecp (mixed 6-31G*/SDD), which should default to 5D/7F basis functions.

However, I realized that when M05-2X/6-31G* was used to fit the SMD model, the basis functions may have been 6D/10F (am I right?).

Should I actually specify 6D 10F when I use genecp to obtain accurate SMD energies? Or will the difference be negligible? I am currently testing it, but I would like to know if the basis function types actually matter for this type of energy calculation, based on a theoretical reasoning.

wham09

Member

Registered: 2025-09-19

Posts: 62

Re: Basis functions for 6-31G* used to fit SMD

Additional question: For single-point energy calculation using an ordinary functional, if I want to compare the performance of the following Pople-type basis sets (6-31g*, 6-311g**, 6-311+g**) with def2-series (which use 5D 7F), would you say that it is more proper to use 6D 10F uniformly for all Pople-type basis sets?

sobereva

Tian Lu (Multiwfn developer)

From: Athens, Greece

Registered: 2017-09-11

Posts: 2,213

Website

Re: Basis functions for 6-31G* used to fit SMD

Gaussian by default uses 7F for all basis sets. The difference between 5D and 6D in the case of solvation energy calculation is negligible.

sobereva

Tian Lu (Multiwfn developer)

From: Athens, Greece

Registered: 2017-09-11

Posts: 2,213

Website

Re: Basis functions for 6-31G* used to fit SMD

Additional question: For single-point energy calculation using an ordinary functional, if I want to compare the performance of the following Pople-type basis sets (6-31g*, 6-311g**, 6-311+g**) with def2-series (which use 5D 7F), would you say that it is more proper to use 6D 10F uniformly for all Pople-type basis sets?

To completely fairly compare accuracy of different basis sets, in principle, it is suggested to use the basis function form employed by the respective original authors of the basis sets. e.g. 6D for Pople series and 5D for def2-series. For angular moment of F or higher, spherical-harmonic form should always be used (note that even for 6-31G, 7F is used by default in Gaussian).

wham09

Member

Registered: 2025-09-19

Posts: 62

Re: Basis functions for 6-31G* used to fit SMD

Thank you very much.

I have one follow-up question about 6-31g*: When I tested the automatic density fitting of Gaussian 16 (I used PBEPBE/genecp/auto 6D 10F, with 6-31g* for main group elements and LANL2DZ for TM), the converged SCF energy differed significantly from the result of not using /auto. Depending on the structure, the difference reached over 30 kcal/mol. Moreover, a multi-step reaction energy profile changed drastically by using or not using /auto for SP calculations of the reactants and the products.

Is this because Gaussian’s density fitting is generally really bad? Or, is the automatically generated auxiliary basis set for PBE/6-31g* (or LANL2DZ) normally ill-defined? Or maybe the automatic density fitting is somehow sensitive to the type of d and f basis functions?

sobereva

Tian Lu (Multiwfn developer)

From: Athens, Greece

Registered: 2017-09-11

Posts: 2,213

Website

Re: Basis functions for 6-31G* used to fit SMD

Please only focused on how /auto affects relative energies, e.g. reaction energies and barriers. If this change is indeed significant, perhaps /auto in Gaussian is not well compatible with pseudopotential basis sets, it is suggested to use an all-electron basis set for TM instead and then check if the influence by /auto is in a reasonable range.