Running Dihedral Scans on Rowan

Transcript

Hi, I'm Corin, CEO and co-founder of Rowan, and in this video we're going to look at scans: what they are, how to submit scans through Rowan, and how to analyze the results.

First off, what is a scan? A scan provides us a way to ask how a molecule or system of molecules responds when we change one of its geometric parameters. This could be the distance between two atoms, the angle, or the dihedral angle between groups of atoms. And what a scan does is essentially run a series of constrained optimizations for different values of the coordinate we're scanning along and then shows us how the geometry and energy of the system responds to this perturbation. It's easiest to show with an example, so let's hop right in.

From the home page of Rowan we can go to this “new scan” button and click it, and then we can go ahead and just draw a molecule. So let's start by inputting the SMILES string for butane. We can, as always, also draw the molecules ourselves, upload a file, or load in from an external source. So this is butane, the four-carbon alkane. And to look at a scan, we say “butane scan,” we want to first specify which coordinate we want to scan along. So we can click into this box and then click the four atoms for which we want to study the dihedral. Here we can see we've selected the four carbon atoms of butane and we can see the current value for their dihedral angle both here and here. In both cases we see that this is 295 degrees or so, which is corresponds pretty closely to the gauche geometry of butane. So let's look at all of the different possible values here: we'll scan from 0 to 360 and then we will look at 30 points for this and we'll say go.

This is going to queue a series of constrained optimizations like we talked about. We selected 30 different steps, so at 30 different coordinates we will run constrained optimizations and we're going to be doing this using the AIMNet2 level of theory which is a machine-learned interatomic potential that provides very fast and quite accurate results for studies like this. And what we can see is that this is actually so fast that we're already seeing these optimizations finish. It helps that butane is such a small system. Each of these points represents a different geometry that's been optimized with the constraint that this dihedral indicated is frozen, and we can see how the energy changes as we move along this coordinate. Looking more closely at this plot here, obviously we have the dihedral angle in question here, and right now we're seeing values from 220 to 360 and we'll populate the rest of these values as time goes on, and we're seeing the relative energy of these different constrained geometries. The lowest energy, which currently is this one, is shown as zero, and the highest energy, which is this one, is shown as 4.62. You can see the relative energy here.

This is a really nice way to visualize the torsional preferences of this bond, like you might do in an introductory conformational analysis portion of a class, and what we might see is we've actually just found a new lowest energy geometry over here while I've been talking. This corresponds to a value of about 180 degrees—I'm sure if we had exactly 180 as a scan point that that would be the minimum. And this sort of intuitively makes sense. Here we have the two bulky methyl substituents around the central bond as far away as possible, so they're 180 degrees from each other, that's sort of the fully staggered geometry. And here we can see we have some sort of local minima that are about 0.3 or 0.4 kcals uphill in energy, so this corresponds to the gauche configuration. So these bonds are still interleaved nicely here: we're in a nice staggered conformation, but the two methyls are a little bit closer to one another, and so we pick up these extra destabilizing interactions. Here we can say these barriers here, where we're staggered but the methyls are 120 degrees from each other, those are a good bit higher in energy, about 3 kcal/mol above the minimum. And then this conformation here, where the methyls are fully eclipsed with one another, is very, very unstable—that's almost 5 kcal/mol above the minimum.

So this shows what a scan is: it's pretty nice to click through, it quickly gives you a sense for where this compound will prefer to hang out. Let's resubmit it now with a perturbation and see how it changes. So we can resubmit this here. Again we can say “resubmit as scan,” and instead of butane this time, let's look at 2-fluorobutane. And so to do this we will go to the periodic table function here, select fluorine, put this in here. So this will break the symmetry of the system. We'll say “save,” and now we'll have an extra bulky group here—probably not quite as bulky as the methyl group, but still a noticeable amount larger than hydrogen. And so this will be pretty interesting to look and see how this actually modulates the energetics around this C–C single bond. So we'll submit this scan, just like we did the first, and it will start running.

We can see here, if we want, that we can go to the logfile and actually read details about what's happening, for debugging or just if you want to get a good sense for all of the steps involved in a scan. It really ends up being quite a lot because we're running so many constrained optimizations. And furthermore, Rowan uses something called the “wavefront propagation” method, which came out a couple years ago in the literature, where we essentially scan not just in one direction, but also backwards along the points we've already scanned to make sure we don't discover any new low-energy geometries. And this leads to nice smooth-looking scan curves that are more reproducible and a bit more rigorous than if you just scan over the whole potential-energy surface once. You can sometimes get these artifacts from incomplete relaxation.

So here we're seeing this potential energy surface unfold as the scan continues running, and it's not necessarily obvious how this differs from the previous one. What we can do is now actually select both scans and see them superimposed on the same graph. And so now we have the butane scan that's shown in red—here's our butane molecule when we click here—then we have our 2-fluorobutane scan shown in blue. And what you can see is that actually some of these barriers are about the same, such as this one here, where we have the two methyl groups right next to each other, which is still really high in energy. This one's also pretty similar where we have the two methyls gauche. But this one is much, much higher in energy than it was before. And this now starts to make sense because now we have this, the fluorine group we added, which is sort of bonking into this methyl group here. And so that adds an extra steric destabilization, which raises this by quite a bit, it looks like almost two kcal/mol of extra destabilization that's given by this interaction.