Thread by @realscientists

Now that we've (hopefully) recovered from our quick tour of quantum mechanics we'll now look at molecular dynamics (MD) 🧵

At the end of my last thread I mentioned that we often use QM methods when we want high accuracy, but these methods tend to get very computationally expensive as our system size increases.

https://twitter.com/realscientists/status/1381576241678057472?s=20

When studying larger systems the exact QM description of a molecule often isn't necessary and how its dynamics evolve over longer time frames becomes a more relevant.

(this may not always be the case!)

It's useful to think of MD as "zooming out" from our previous QM description. Since we are now at larger time and length scales we can use a classical description of how particles move (we're no longer wavy!)

We often describe MD simulations as a type of “computational microscope” that allow us to investigate processes at temporal and spatial scales that are often hard to probe experimentally 💻🔬

So how do we get to the simulation part? Firstly, we need to describe how all of our now classical particles interact with each other. We do this using a potential energy function (PEF).

Confusingly, within the field the PEF is often called a "force field" - think Star Wars but with (more?) computers.

Ok, so what does this function look like? I know I may have lost your trust in the previous QM thread but this equation (I hope) is a little easier to digest. We can split it up into bonded and non-bonded terms:

Not too scary, right? If we dig a little deeper the terms look something like this (again, don't panic):

All this is saying is that the bonded terms (U_bonded) are made up of contributions from the bonds, angles, and torsions. The non-bonded parts (U_non-bonded) are described by Van der Waals forces and electrostatics.

See, not terrifying right?

Quick explainer:

1) a torsional angle is the angle created between two intersecting planes, where one contains atoms i, j, k whilst the other contains atoms j, k, l.

2) Van der Waals forces are weak electrostatic forces that attract neutral molecules to one another.

Great! Now that we have our PEF let's use it. From this function we can derive the forces present on each atom in the system. Using Newton's equations of motion we can work out where each atom will be after a small increment of time.

If we keep repeating this process we can eventually make a "movie" of how our system evolves over time (i.e. a simulation!).

Depending on if your MD simulation explodes or not dictates the genre: horror film or romantic comedy.

Ok, there are a few other small things we need to do to make sure our simulation behaves. Firstly, since we are simulating our system in a finite box what do we do if our molecule hit one of the walls?

In reality these walls wouldn't exist so we enforce something called periodic boundary conditions (PBC). Here, we surround our system with images of itself. When our central system passes through a wall it appears on the opposite side of the box. PBC saved the day!

The last couple of things we need to include are a thermostat and barostat. These control the temperature and pressure of our system and make sure we stay close to our target 🎯

Leaving out a few minor details, that's pretty much all we need! These types of simulations are known as atomistic for somewhat obvious reasons and are used a lot in the computational chemistry field.

I use atomistic simulations with large biological molecules called proteins. We can use all of the things described above to simulate those and answer some pretty interesting scientific questions!