We want to design materials from our laptop.
In order to do this, we need to accurately predict the property of the materials that we are designing and propose actionable synthesize route to actually make them in the lab (manually or automatically, I prefer automatically).
In order to accurately predict the property of the materials we need:
(1) Good prediction of the energy of given atomic coordinates
(2) Good sampling of the Potential Energy Surface (PES) to get free energy or statistical behavior of the system.
Similar as Olympics, we want computational chemistry to be:
Better, Faster, Larger
- Better means better accuracy
- Larger means larger systems (more atoms) and longer trajectory (more samples)
In order to achieve this goal. As computational chemist, we need:
(1) Physics: help us be "better" (more accurate). Idead of approximation that can reduce the computational complexity (e.g. the idea of pseudopotential).
(2) Computer science and software enginerring: faster and larger
Chemistry is just our playground. But solving Schrödinger equations faster, more accurate on larger systems, that's the never-ending pursuit of our computational chemist.
I want to end this post by quoting David Hilbert:
"Wir müssen wissen. Wir werden wissen."
"We must know. We will know."
I have chatted with my lab mates about the role of computational chemistry in chemistry in general, and I gained new insights of this topic.
We have experimental results -->
something works nicely -->
try to understand why -->
propose theories (guess) -->
using computation to validate whether the theory is correct -->
make new predictions based on the theory -->
further experimental validation (new loop)
In the procedure above, the most important thing (for me) is from "something works nicely" to "try to understand why", from experimental observation to the understanding of why that happens.
The insight is: our goal is to find a factor that has same trend as experimental findings, not the factor. Since there are too many factors that can affect one thing, it is really hard to determine which factor is the most important one. Sometimes it doesn't make sense to ask such question. A more practical questions would be: "Which factor has positive contribution to the experimental observation?" And this is how theoretician and computational scientist can collaborate with experimentalists. Our job is not to fully replace the experiments in the computer, but to propose new insights and use simple models to show the effect of these insights conclusively.
So for us computational scientist, our job is:
- Come up with new understanding of the chemical system (not necessary the whole picture, only one/few factors will do) --> this is the most important work for us to do, is to generate idea that could be useful to explain the experimental results (or finding factor(s) that have similar trends as experimental results)
- Use computational tools to validate our understanding (whether it is true or not)