5
$\begingroup$

Let suppose the we have a "big" system. By "big" I mean a system that have a certain number of atoms that the software is not capable to treat with the default initial parameters.

Question: Is there a recommended way to proceed in order to get the geometry optimized (or the self-consistence field calculation completed) with production parameters?
(By production parameters I mean good for scientific conclusions)

Some possible examples of actions that came to my mind:

  • increase the geometry steps number (straightforward but not always optimal);
  • increase the number of self-consistence field steps (straightforward but not always optimal);
  • decrease the accuracy and after getting and optimized geometry, increase the accuracy;
  • start with a low level basis set, use the optimized geometry as input with a higher basis set;
  • use a simple method (like molecular mechanics) to optimize the geometry, then move to the desired method;
  • etc.
$\endgroup$
6
  • 4
    $\begingroup$ I'd avoid molecular mechanics. If you want to do an initial optimization, I'd consider something like a smaller basis set or DFTB method like GFN2 / xtb for an initial optimization first. For molecules, B97-3c work fairly well as a second-stage optimization. $\endgroup$ Dec 20, 2021 at 5:37
  • 1
    $\begingroup$ @GeoffHutchison in my experience MMFF94 does better than semi-empirical or small basis set for conformer generation. I dont worry about accurate bonds and angles in the initial step, just finding the lowest energy conformer. For actual geometry of a given conformer... I dont use forcefields either. For a big system where we arent after an ideal gas geom opt... I am interested in the answer to this quite abit $\endgroup$
    – B. Kelly
    Dec 20, 2021 at 12:48
  • 4
    $\begingroup$ We benchmarked this. Force fields do very poorly at finding the lowest energy conformers because they just don't get the non-covalent interactions right. In my group, we use GFN2 / xtb which has fairly high correlation with DFT energies. $\endgroup$ Dec 20, 2021 at 14:22
  • 1
    $\begingroup$ What is the difference between a Pearson R2 and a Spearman coefficient? (nvm, monotonic). it is a good paper, but I have not found small basis DFT to be reliable for ranking. Perhaps amines are a unique bunch. $\endgroup$
    – B. Kelly
    Dec 20, 2021 at 18:22
  • 3
    $\begingroup$ Geoff, I was going to drop an additional reference to the strikingly titled "Sobering Assessment" paper that I remember reading a few years ago with great gusto... until I just now realized that you were actually an author on that one as well. :) $\endgroup$
    – Antimon
    Dec 20, 2021 at 20:58

1 Answer 1

9
$\begingroup$

Suggestions for Molecular-Like Systems (not periodic unit cells)

I'll make some general suggestions for molecular / isolated systems, not crystals. (In the case of periodic systems, some methods aren't available or are different.)

You mentioned increasing the number of self-consistent steps. If your SCF calculation doesn't converge, there are a variety of separate questions / methods to try.

The main problem is that getting to a final equilibrium geometry with a DFT or high-level wavefunction method can be slow. For larger systems, there are many degrees of freedom and often the potential energy surface is fairly flat, especially with dihedral angles.

So the goal is to get close to a converged geometry with a faster method, then to switch eventually to your method of choice.

Years ago, the advice would be to use a force field to "pre-optimize" a system, since these are intended to get quality geometries. I no longer advise this unless you have a specialized force field (e.g., biomolecule force fields work well for biomolecules).

In particular in two benchmark papers, we found that force field methods weren't very good:

We found that the MMFF94-optimized geometries had large forces when you switched to semiempirical or DFT methods. We also found that when ranking different conformer poses by energy, the force field had poor correlation with ab initio methods.

Instead, we find much better results with the approximate GFN2 / xtb method, or if possible with methods like ANI-2x, with the caveat that GFN2 works for all elements.

Our most recent workflow is something like this:

  1. Generate an initial geometry (e.g. in Avogadro, RDKit, etc.)

  2. Optimize the geometry using xtb which is fast and often "forgiving" (e.g., easier to converge SCF).

  3. Optimize the geometry using B97-3c or similar GGA method (e.g. r$^2$SCAN-3c looks good too)

  4. Optimize the geometry in your favorite method with dispersion correction (e.g., $\omega$B97X-D4 or $\omega$B97M-V for example, triple-zeta basis set).

If step 4 still takes a long time or fails to converge, we'll often optimize with def2-SVP first (often with looser convergence criteria) and then switch to the larger basis set.

It's a process - in part because getting the electrostatic and non-covalent interactions right is hard. Some ML-based methods may help with this in the future, but beyond ANI, they are not widely available.

$\endgroup$
5
  • $\begingroup$ Can you continue this answer with periodic unit cells ? $\endgroup$
    – Elie H
    Dec 20, 2021 at 22:47
  • 1
    $\begingroup$ I just want to add that this four-step workflow (actually, 5-step since I often do the def2-SVP pre-opt) is almost exactly what I would recommend and is precisely what I have done many times before. One notable exception here is for metal-containing systems, for which xTB can be much less reliable depending on the type of materials being studied (e.g. strange bond dissociations are not uncommon). $\endgroup$ Dec 21, 2021 at 8:18
  • $\begingroup$ @AndrewRosen What do you suggest for metal-containing systems instead? $\endgroup$
    – ksousa
    Dec 21, 2021 at 13:36
  • 2
    $\begingroup$ It seems as if GFN2 has low barriers for bond breaking - this can happen with organic molecules too (ring rearrangements). If it happens, you can add bond constraints. $\endgroup$ Dec 21, 2021 at 14:01
  • 1
    $\begingroup$ @ElieH - I can't help much with periodic unit cells because I have much less experience and the issues vary. I would hope someone else could help there... $\endgroup$ Dec 21, 2021 at 21:53

You must log in to answer this question.

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