I have asked a question in chemistry, and I found someone already asked question, but no answer yet. So I post again with the hope of suggestions. I have an initial compound A (such as Fe3O4) and I know a final product B (such as FeO). I would like to know intermediate compounds in the reaction from A to B, and maybe activation energy. Is there any computational method/technique to predict intermediate products of a reaction?
$\begingroup$
$\endgroup$
6
-
1$\begingroup$ Do you know all initial compounds and all final products, or just one initial compound and one final product? And, do you know the crystal phases of the initial compound(s) and final product(s)? $\endgroup$– wzkchem5Jul 1, 2022 at 7:24
-
$\begingroup$ Yes, I have. For example, Fe3O4 + C --> FeO. $\endgroup$– Binh ThienJul 2, 2022 at 14:41
-
$\begingroup$ So this is a solid-solid reaction, which is way more complex than if at least one of the reactants isn't solid... Do you know the indices of the exposed crystal surfaces of the reactants? A crystal with the (100) face exposed may have a completely different reaction mechanism with the same crystal with the (110) face exposed, despite that the two crystals have identical chemical composition and bulk structure. Even worse, the mechanism may also depend on what kinds of defects are present on the surface. $\endgroup$– wzkchem5Jul 2, 2022 at 18:48
-
$\begingroup$ @wzkchem5 yes, it is so complicated. Do you have any suggestions? $\endgroup$– Binh ThienJul 3, 2022 at 13:54
-
1$\begingroup$ If you did characterize the surfaces and defects of all reactants, then one may now hope to make non-trivial predictions of the mechanism. But even this is hard: the clean surfaces of the reactants may soon be covered by the solid product, and it is non-trivial to determine how the reaction proceeds further (does the product layer crack open to expose the unreacted reactant? Does the reactant diffuse through the product layer? etc.). I actually believe that there is as yet no general protocol for proposing the mechanism of solid-solid reactions, but maybe I'm poorly informed. $\endgroup$– wzkchem5Jul 3, 2022 at 18:35
|
Show 1 more comment
1 Answer
$\begingroup$
$\endgroup$
3
My answer is focused on molecules, and not solids (like your question suggests). I also assume boring organic molecules, with not so many atoms, and one clear reaction path.
There's two relatively user-friendly ways to find a reaction path. For both of these methods, you first need to have a good knowledge of both the reactants and the products - meaning well-optimized, local minimum geometries.
Method 1 You kinda guess the transition state - for boring organic molecules, this can be done somewhat easily with some experience. Then, you start from your guess, and you optimize to find a saddle point. Most electronic structure codes can do this quite easily. You know that you have reached a transition state when you have one imaginary frequency vibration, and that mode is sort of along the expected reaction path. Then, having these three points, you can find algorithms that interpolate a reaction path between the structures: I've had good experience with the QTS keyword in Gaussian.. Note two things about this implementation in particular: there's an option to start from an approximate transition state too, so you don't really need to optimize it fist, and, in principle, you can even avoid having the two local minima (although finding them is the easiest of all of this).
Method 2 If you have no idea about the transition state, there's some promising methods such as the nudged elastic band method implemented in NWChem. These methods (called minimum energy path optimizations) are probably the closest you can get to what you have in mind. I don't have good experience with these methods, however, because if you have a reaction that's not overly trivial, there can be thousands of competing reaction paths, and it's very hard to find them automatically. Still, these methods are out there, and for simple systems, they do work.
So if your reaction is really simple, you can get automatic reaction paths probably; if it has a normal complexity, then you will have more luck guessing at least the transition state, and you can use much better algorithms from there.
Note: If you only need the activation energy, you don't need the full path, only the initial state and the saddle point.
-
1$\begingroup$ +1. Just one comment: these methods basically only cover elementary reactions. For non-elementary reactions, one usually has to first guess the intermediates, and then guess all transition states between adjacent stationary points on the whole reaction profile. It is usually the involvement of an unknown number of intermediates that makes mechanistic proposals difficult. While it's possible to discover intermediates via QST2/QST3 or NEB, this is prone to failure and is usually less efficient than first guessing the structures of the intermediates. $\endgroup$– wzkchem5Jul 1, 2022 at 20:26
-
$\begingroup$ It seems to be more complicated than what NEB or QST can do. $\endgroup$ Jul 2, 2022 at 14:17
-
2$\begingroup$ +1, I wanted to add that there is an excellent software autodE (github.com/duartegroup/autodE) which automates a lot of the mentioned procedures. If you give it beginning and ending structures, it attempts to build all possible pathways, intermediates, TS etc. between them. $\endgroup$ Jul 31, 2022 at 16:43