I'm modelling a small set of proline analogues - extra functional groups, sulfur in the ring. Using Gaussian / DMACRYS.

--functional B3LYP EmpiricalDispersion=GD3BJ --basis_set 6-311G**


-fun B3LYP EmpiricalDispersion=GD3BJ -bas 6-311G**
  1. Would EmpiricalDispersion=GD3BJ go under -fun, or would it go under --additional_args? (If so, please show how, I can't get this to run correctly, unsure of the syntax)

  2. To model this more accurately, PBEPBE has been suggested, pure DFT rather than hybrid functional, so there is increased accuracy. Is there a sufficient increase in accuracy for the increase in computational time, or is it better to apply dispersion corrections to B3LYP? (Or PBEPBE and corrections?) Is there another level of theory you would suggest trying?

  3. The electrostatic interactions seem to be over-modelled using B3LYP / 6-311G**, is there a basis set / functional combination (without or with corrections) that would not over-estimate the energy of the + / - charge interactions? (I've tried CAM-B3LYP, but doesn't work on my system.)

  • $\begingroup$ It would significantly help if you could indicate how you're running Gaussian. My guess is that the empirical dispersion correction goes under "additional_args" but you seem to be using some program to create the Gaussian input, so it depends on that. $\endgroup$ Apr 20, 2015 at 12:40
  • $\begingroup$ @GeoffHutchison is quite right, it would be interesting to have the Gaussian input file. Pure DFT would cut down a little on the computational cost, if it enhances accuracy is probably something only benchmarks can tell you. $\endgroup$ Apr 20, 2015 at 12:53

1 Answer 1


You seem to be running Gaussian through another program or script. Consequently, I have no idea where the various arguments should go. Conventionally, you make a Gaussian input file with keywords, such as:

%n B3LYP/6-311G** EmpiricalDispersion=GD3BJ Opt

[... molecule goes here ..]

  1. For dispersion-correction (i.e., GD3BJ) you would need an appropriately recent version of Gaussian and use the separate term as indicated. It is not part of the functional designation. (Although some functionals implicitly have dispersion correction, B3LYP does not.)

  2. I don't know a priori that PBEPBE would give more accurate results than B3LYP or any other functional. Your best bet is to either consider results from papers in the field (e.g., molecular crystal prediction or runs with DMACRYS) or testing a few functionals and finding out for yourself. Consider reading this question and answers.

  3. Intermolecular interactions are not likely to be handled well by B3LYP. This is the whole reason for range-separated functionals and dispersion corrections. As to a specific functional suggestion, I highly recommend trying a few and looking at how to choose a functional.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.