12
$\begingroup$

We are all aware of the fact that MD Simulations using force fields can treat very big systems. But they cannot handle "exotic" materials. Why?

$\endgroup$
3
  • 4
    $\begingroup$ You have to first define "exotic". Without defining what an "exotic material" is, the claim may become non-falsifiable - obviously there is a non-zero portion of systems that cannot be handled reliably with usual MD approaches, and you can always modify the definition of "exotic" to include (and only include) those systems. And it's pointless to ask why a non-falsifiable statement holds true. $\endgroup$
    – wzkchem5
    Oct 24, 2021 at 7:07
  • 3
    $\begingroup$ Besides the obvious problem with "what is exotic?" (this is not really a chemical term), it is also worth defining what do you mean by MD: classical force field-based MD? How about ML-MD? AIMD? reactive force fields? $\endgroup$
    – Greg
    Oct 25, 2021 at 3:33
  • $\begingroup$ I think you mean why do force-fields have trouble with some materials. If your FF doesn't work for MD, I doubt using Monte Carlo would fix the issue... $\endgroup$
    – Wesley
    Oct 26, 2021 at 17:20

2 Answers 2

14
$\begingroup$

"But they cannot handle "exotic" materials. Why?"

You have claimed this without a reference! MD in principle can be done for any group of interacting particles, especially if you have an accurate forcefield. The issue is that classical MD with a classical forcefield will probably not do as well for an exotic material like a superconductor, semiconductor, or even just a superalloy containing a transition-metal (e.g. chromium), as it would do for simulating, for example some atmospheric molecules like oxygen, nitrogen and carbon dioxide. So the calculations can be done, but their accuracy will be questionable.

All of the examples of exotic materials that I've mentioned, are either intrinsically quantum mechanical (can you successfully describe superconductors without quantum mechanics?), or contain significant quantum mechanical effects, so the forcefield would have to have quantum mechanical effects included (which is not always easy, or is not very accurate in cases where a forcefield was made easily). Furthermore, all the exotic materials mentioned, have relatively heavy elements, meaning relativistic effects are needed too. Basically, exotic materials quite often involve elements that are not even easy to describe for static electronic structure properties using DFT, so treating their dynamical motion using classical MD (or AIMD involving DFT calculations that are already difficult or inaccurate), will not be so easy either.

Basically, it might be hard to get great accuracy in an MD simulation of an exotic material compared to a system involving just atmospheric molecules (for example), but it's not easy to get great accuracy for exotic materials' static properties like their electronic structure, either.

$\endgroup$
8
$\begingroup$

By molecular dynamics, I assume you mean molecular dynamics using force-fields (a.k.a. Classical Molecular Dynamics). Remember that there is also ab initio MD.

Force fields are parameterised, i.e. they depend on the a set of constant parameters, which have been previously fitted to reproduce experimental data/high level ab initio results. These parameters are necessary to calculate all of the interaction terms between the particles in your system. So, the applicability of force-field based molecular dynamics depends on whether there are accurate parameters that are available for the system that you want to study.

A system which is "exotic" would be unlikely to have accurate parameters developed for it, because so few researchers would have studied that system (otherwise it wouldn't be exotic, would it?). It is also likely that an "exotic" system would have unusual properties which would make it difficult to develop parameters for it (maybe the species is unstable, or its potential energy surface has an odd shape etc.).

In the absence of parameters fitted for that particular system, you would be forced to use general force-field parameters (like UFF for instance). There you would run into the problem of poor transferability—parameters fitted to one system are rarely usable in other unrelated systems.

So, the answer to your question is simple: MD cannot handle exotic systems well because there are no good parameters for those systems, precisely because the system is exotic. MD can be done on any system, provided you can fit parameters.

[In the recent years, there has been a lot of investigation into force field development for exotic molecules using ab initio and machine learning. For example, a PhD researcher I know is working on developing parameters for some short-lived species like $\ce{NO3}$, $\ce{SOF}$ etc. So, a lot of things that were previously exotic are no longer exotic.]

$\endgroup$

You must log in to answer this question.

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