In Defense of (Certain) Pople-Type Basis Sets

22 January 2024, Version 1
This content is a preprint and has not undergone peer review at the time of posting.


A recent study suggests that Gaussian basis sets in the 6-311G family are inappropriate for thermochemical calculations based on density functional theory, emphasizing the need for polarization functions but omitting tests of Pople basis sets containing a full complement thereof. Here, we point out that certain basis sets in the 6-311G category yield error statistics with respect to benchmark calculations that are comparable to def2-TZVP, at about half the cost. More elaborate Pople basis sets can rival the accuracy of def2-QZVPD at 5-10% of the cost. We also clarify the role of integral thresholds in achieving robust convergence in the presence of diffuse basis functions.


basis sets
density functional theory

Supplementary materials

Supporting Information
Additional error statistics and timing data


Comments are not moderated before they are posted, but they can be removed by the site moderators if they are found to be in contravention of our Commenting Policy [opens in a new tab] - please read this policy before you post. Comments should be used for scholarly discussion of the content in question. You can find more information about how to use the commenting feature here [opens in a new tab] .
This site is protected by reCAPTCHA and the Google Privacy Policy [opens in a new tab] and Terms of Service [opens in a new tab] apply.
Comment number 1, Georgi Lazarov Stoychev: Jan 24, 2024, 08:33

This kind of paper really frustrates me. Sure, I would agree with the starting premise: sweeping "bans" of a whole class of basis sets are unwarranted (it is really 6-31G* we would like to see less of). And it seems like the authors would go on to a well-devised analysis. They set up with "Our assessment includes def2-QZVPD, which should lie near the basis-set limit and establishes the inherent accuracy of each functional." OK, so they know there is an inherent error in the functional and will not be fooled by error cancellation, right? No - they then go on to stress how some of the Pople basis sets have a smaller std. dev. than some of the Karlsruhe ones: "In particular, 6-311+G(2df,p) reduces the standard deviation of the errors by 2 kcal/mol, relative to def2-TZVP, for the meta-generalized gradient approximations (mGGAs), a significant improvement that is not adequately reflected in the median absolute errors." The numbers here are: std. dev. | B3LYP | M06-2X | wB97M-V 6-311+G(2df,p) | 9.0 | 5.5 | 5.1 def2-TZVP | 10.0 | 7.6 | 7.2 def2-QZVPD | 9.6 | 8.1 | 7.3 So what they are saying is true, but 6-311+G(2df,p) also gives a smaller std. dev. than def2-QZVPD - which they said was close to the CBS limit. How can they fail to discuss this?! At this point the whole argument loses credibility for me...