Heterogeneous catalysis is known to be very important:

Approximately 35% of the world's GDP is influenced by catalysis. The production of 90% of chemicals (by volume) is assisted by solid catalysts.

  • What considerations go into a model of these types of reactions? For example, how would one generally go about modeling surface catalysis (e.g. software/algorithms)?

  • Besides software what considerations should be taken into account that are different from homogeneous catalysis?


2 Answers 2


I did some work during my rotation in the Brooks group on surfaces that could act as abiogenesis sources (self-sustaining chemical reactions that become living organisms).

Key considerations for Heterogeneous catalysis are:

  • The timescales of your reaction, in our case it was a 24hr high pressure/temperature reaction so we required a driving force to have an initiation event. This isn't uncommon, but generally you want a surface that reacts faster than that.

  • Crystal cuts, also called Miller Indices, generally have to be investigated thoroughly. Essentially the angle of a cut through a unit cell is entirely up to chance, so a 001 crystal surface may react differently than a 010 or 110 etc. As a result you have to do some research of the common fracture angles for your surface and generate them as unique catalyst surfaces.

  • Solvation almost always has to be explicit to get a realistic reaction behavior. But oftentimes you'll add some type of force or implicit solvent on top to contain the top of your surface. This is especially important with the periodic boundary conditions that are typically used, so that your molecules don't interact with the bottom of the crystal which is normally buried.

  • Additionally how deep you layer your surface can matter. Many crystals at the atomic level can "breathe" up to 3 or 4 layers deep (some iron minerals can react even farther) so you'll want to have at least enough layers for the full reaction space, and some below. However, you can freeze portions of your surface to avoid additional calculation expenses.

For software we used VASP and TBDFT with periodic boundary conditions with explicit waters. So those are the most familiar to me and the ones I would recommend

For input files I found Avogadro has the most user friendly for generating input files for VASP and TBDFT, however VESTA is another freely available package for opening CIF files and generating computation inputs.

  • $\begingroup$ Interest, what do you mean by "breathing". Perhaps another example of a consideration is adsorption (which I guess could be considered a reaction), although I'm not sure how that would be quantified. Also for timescale, are you doing some kind of meta-dynamics? That is not typically done in homogenous catalysis. $\endgroup$
    – Cody Aldaz
    May 5, 2020 at 22:00
  • 1
    $\begingroup$ Yeah "breathing" is like the surface layers rearranging to allow absorption in a lower layer one might assume is inaccessible but since molecules are moving around in the crystal lattice can be accessed. Yeah since we were looking at an environmental reaction we were doing a dynamic simulation, for a non-dynamic you would simply have a molecule reacting as desired and measure the energy surface $\endgroup$ May 5, 2020 at 22:34

To add some further information to Raz's answer, it is also possible to talk about the nanoparticle case. With computing power getting stronger and codes becoming more efficient it is becoming more and more possible to explore almost physical systems. Nanoparticles for example can be modeled directly based on either experimental observations (TEM, XRD, XPS results) or based on surface energies and wulff constructions.

For these calculations one might go about the following process to create the nanoparticle model.

  • Calculate surface energies of all relevant surface terminations, possibly even modeling changes due to coverage effects
  • Perform a wulff construction to get a good initial guess at the structure
  • Explore deviations in the wulff construction, edges and corners are not modeled directly and their effect becomes stronger and stronger as the particle gets smaller. DFT will tend to not explore massive particles, so this might be a big factor.

It may also be possible to use some type of mixed model for large nanoparticles where part of the nanoparticle is calculated at the DFT theory level and at some distance the DFT is replaced by a force field. This allows for the modeling of corner sites for example which would not be reasonable to model using a pure DFT model.

It is also worthwhile to remember that while solvation is typically needed for calculations in aqueous environments (with a large impact on the surface), implicit models can perform well in non-hydrogen bonding solvation environments. Optionally, you can just restrict yourself to collaborating with people who only work in a perfect vacuum.

  • $\begingroup$ I'm a bit surprised that I'm the only that gave this a +1. I'll have to investigate further what's going on. On thing I see is that you are #10 most active on the site but #34 in voting. Maybe it's a case of "what goes around comes around" :). $\endgroup$ Sep 7, 2020 at 19:46
  • $\begingroup$ It doesn't appear to be such a disparity when filtering to look at the most recent quarter. I have a feeling it will be much more difficult to catch up on voting than reputation. I also try to avoid upvoting all answers (preferring only , I find voting to be more abused than reputation on SE in general. That probably reflects in that stat as well. $\endgroup$ Sep 7, 2020 at 22:28

You must log in to answer this question.

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