I am a bit new to computational chemistry, but I've been looking into a reaction of a molecule that contains C, H, and F reacting with AgF in a polar solvent. I am looking into one pathway that involves the coordination of $\ce{Ag+}$ to the molecule, but I am unsure of what basis set to use. I tried the 6-311+G** basis set, but it doesn't seem to be compatible with silver. I also just tried to do a calculation with a mixed basis set 6-311+G**(C,H,F) and LANL2DZ(Ag) but the geometry did not converge for the given CPU space. I was wondering whether this combination of basis sets is appropriate for this analysis or would a different one work better?

  • $\begingroup$ Why would a fluorocarbon react? $\endgroup$
    – Mithoron
    Commented Aug 27, 2022 at 12:26
  • 2
    $\begingroup$ @Mithoron While I agree with reframing sort of arguments often, I don't think that detail really matters for this question. The question is really about a basis set compatible with C H F and Ag. $\endgroup$ Commented Aug 27, 2022 at 22:08
  • 2
    $\begingroup$ What about def2 basis sets? They are available for almost all elements, and you also have ECPs for the heavy elements. $\endgroup$
    – S R Maiti
    Commented Aug 29, 2022 at 21:12

2 Answers 2


Firstly, 6-311+G** for C,H,F is usually good enough for DFT calculations (assuming you are using DFT). If F elements are the only possible elements which can be or become anion F- in your studied system, there is a more computationally cost-effective scheme: using 6-311G** for C,H and 6-311+G* for F. Proper diffuse functions can often save computational cost and avoid possible convergence issues in SCF or CP-SCF.

Secondly, LANL2DZ for Ag is somewhat small/insufficient, especially for computational resource nowadays. LANL2TZ or def2-TZVP would be better. If your system contains <50 atoms, using LANL2TZ(f) for Ag is also a good choice, especially if your Ag is involved in catalysis.

Thirdly, "the geometry did not converge for the given CPU space" is hard to tell why. For example, a wrong syntax of LANL2 ECP written in the input file may lead to strange behavior of geometry optimization. So you may want to post your input file (no need to show Cartesian coordinates) and we can check if there is anything inappropriate.

  • 1
    $\begingroup$ Thank you! I finally got a geometry that doesn't have a carbon atom sitting inside a silver atom. $\endgroup$
    – M.L
    Commented Aug 30, 2022 at 1:56

This is a question which is constantly popping up on the forum. The Pople basis sets are outdated. Modern basis set families, like the Karlsruhe def2 series, span (almost) the whole periodic table, include relativistic effects which are missing in the Pople sets, and are defined in a systematic hierarchy of balanced accuracy:

  • def2-SV is a split-valence basis like 6-31G
  • def2-SVP is a polarized split-valence basis like 6-31G**; def2-SV(P) omits polarization functions on hydrogen like 6-31G*
  • def2-TZVP is a triple-zeta valence polarized basis
  • def2-QZVP is a quadruple-zeta valence polarized basis

The sets are consistent for all elements and are available for H to Rn. Diffuse functions are available with the D suffix (e.g. def2-SVP -> def2-SVPD); for post-Hartree-Fock calculations one should add additional polarization functions (def2-TZVPP or def2-QZVPP).

The basic def2 sets are also included in Gaussian.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .