Finding a Transition State

Transcript

Hi, I'm Corin, CEO and co-founder of Rowan, and in this video, we're going to show how to find a transition state in Rowan. In this video, we'll first draw out the reactants, perform a scan to locate a structure that's near the transition state, and then of course, do a transition state search to find the actual transition state and look at the frequencies to confirm that we have indeed found the correct one.

The reaction we're going to be studying is this concerted nucleophilic aromatic substitutions from a 2018 paper by Eugene Kwan and coworkers. And this paper—not super relevant for this video—essentially argues that most SnAr reactions, most nucleophilic aromatic substitutions, in fact, don't proceed through the textbook Meisenheimer intermediate, but instead proceed through a concerted one-step mechanism with no intermediate. And they argue that essentially this only happens in cases where you have extremely electron withdrawing groups that stabilize the anionic intermediate, and then for most reactions of interest in medicinal chemistry, synthetic organic chemistry, etc., that this intermediate is just too unstable and that the reaction happens in a single step. So let's see if we can find that one step mechanism today in Rowan.

We'll start here from the home page of Rowan that shows you all the different calculations you can run. And here we'll click on "New Calculation." And since this is a nucleophilic aromatic substitution, we'll want to start with an aromatic ring. So we'll go to "Add From Library," we'll go to rings, and we'll add this phenyl ring, this benzene ring. And from here, whoops, we'll click on this, turn it into a nitrogen. We'll delete this hydrogen, so we've got ourselves now a pyridine ring. Shift + E gives us a CO2Me, a carboxyl group, sort of a little ester, and then we'll add a bromine here to serve as our electrophile. So this will be a lovely aryl halide that we can use for our nucleophilic aromatic substitution. So we'll say save, we'll say "Ar–Br." You know, maybe if we're gonna look at other ones, we would want to have a better nomenclature than that, but this will be just fine for right now. And we'll say optimize plus frequencies. Say "Submit calculation."

So this has already started running. We expect that this will hopefully be a very fast calculation. Usually things take a sec just to get going. This will be using the AIMNet2 neural network potential, which is a fast and quite accurate neural network potential. It's one of the most efficient methods on Rowan. And it should give very good results for a first pass here. And now here is our optimized structure. So if we play the optimization through, we can see the big changes that we rotate the CO2Me to be in plane, which makes sense. So now we have some conjugation here. And if we look at our frequencies, you know they're all positive. We have no imaginary frequencies, just like we expect. So this is indeed truly a ground state. And this is a lovely reactant complex for our reaction.

So now let's say resubmit, resubmit this now as a scan and now let's edit this. So this is our, you know, this is sort of the ground state of our reactant and now we need a nucleophile. So let's add a nucleophile from the library and for this one let's look at CN-, so cyanide, this is a really good nucleophile and yeah well we'll just try it right here. So we've just sort of, whoops, we only want one. We sort of tossed a cyanide up here. It's sort of hovering above the ring. Maybe this is a little too far away so we can hold Z and drag it down sort of a little more on close. It's a bit tough to do this in three dimensions because you never can quite tell what you're doing in every dimension. But I think this is getting to be about the right spot. Yeah, so now it's hovering over the top of the ring. And this looks like a decent place to start scanning. So let's say save. And now we're gonna say "Ar–Br + CN- scan". So notice we have now a couple of things that demand our attention. So these red boxes here tell us essentially that this combination of parameters does not make sense. So this combination of electrons, charge, and multiplicity is not possible. And it's telling us "I think you've made a mistake." This is either a radical or it has a non-zero charge. In this case, we have added a cyanide to a neutral molecule. So it's right, and our charge should be minus 1. And now this makes sense. We also are going to be doing a scan here, and the scan needs to make sense as well. So we haven't actually specified what we're scanning yet. So let's click. We're going to take this, the nucleophilic carbon here, carbon 18, the carbon on this cyanide. So sorry, let's click into the box. We'll click carbon 18, which auto-populates there. We'll click carbon 5. And now this is saying that right now these are 4 Å apart. OK, so. Let's say our start value, let's say 4 exactly, because we like round numbers. We'll say 1.5 Å is where we should stop the scan, which is about the length of a C–C bond. So that's a very reasonable place to stop the scan. And then we'll say we want, let's say 30 steps. So now we'll say "Submit scan."

Okay, so what this is going to do is it's going to run a constrained optimization with this bond distance frozen at 4 Ã…. And then what's going to happen is it's going to bring them a step closer together. So the 4 Ã… optimization will be the first step of the scan. Then it will bring them a little bit closer together and do another optimization. And it will keep doing that until we are all the way to 1.5 Ã…. So essentially we're scanning along the reaction coordinate or along the specific degree of freedom which we think corresponds pretty closely to the reaction coordinate. And we're going to watch essentially the reaction happen in the computer, which is pretty cool. So on the x-axis here we have, you know, the C18 to C5 bond distance. So from 4 Ã… to 3.91 Ã… to 3.82 Ã…. We can see that these are all pretty low in energy. You know, the cyanide is really pretty far away. It's sort of bouncing around a little bit right now. you know, if some sort of odd and strange things are happening. But in general, we expect that the energy will start climbing at some point once we actually start reacting.

So I've jumped forward about two minutes just to fast forward through some of the scan running time, which isn't... terribly interesting to watch from a video perspective. But this hasn't taken too long. The whole thing has been running for about four minutes. So there's no deceptive video editing occurring here. And what we see is that now the bond is actually starting to form here. So the C18 to C5 distance is about 2.5 Å. So it's definitely not a bond yet, but it's a lot more like a bond than a 4 Å distance is. And if we now hit this play button here the steps that have happened. You know, we can see that down here, when the cyanide is pretty far away, you know, we're sort of just bouncing around over the ring trying to find a stable, I guess, anion–π place to be, but it's, you know, there's not much happening.

Now we're starting to really rise in energy as we actually start reacting. And this is sort of the classical potential energy surface that we expect from a physical organic chemistry class. And so something interesting that we can look at here is, you know, also looking perhaps at this C–Br bond distance. So it, you know, it's bounced around, oh no, that's an angle, sorry about that. So the C–Br bond distance, it's bounced around a little bit down here, sort of as we jitter through the ring. Oh, that just went away as it refreshed. But we do see it now start to increase pretty starkly, which implies that we're essentially breaking the C–Br bond as the C–C bond is formed. So sort of exactly what we expect from a concerted reaction is one bond forms, the other breaks. And so now, you know, if we look at these later steps, now the reaction's really happening. You know, now we're actually just from a distance geometry point of view, we're actually predicting the bond is forming here now. You know, we see this reaction sort of rotating and the bromine coming out of plane and the cyanide coming into plane. And this looks very much like sort of a... potential energy surface. So, you know, we can leave this to keep running and we'll see, you know, the product inexorably form and the bromine be ejected. Oh, there we go. That's pretty formed.

What we now want to do is sort of zoom in and find where we can run a transition state search from. So these points are constrained ground-state optimizations. So we're still finding minima, sort of places where the forces are zero, except for the degree of freedom that we've frozen. This isn't actually what we want for a transition state. So this looks a bit like a transition state, but a transition state is a first-order saddle point on the potential energy surface. So we need actually a totally different optimization algorithm so that we can optimize to a saddle point where we have an imaginary frequency, as opposed to these sort of like constrained ground states. And so if we look here, you know, we can see just the energy, you know, this is sort of the highest energy point. Yep, it's this one.

So let's resubmit from this geometry now. And we'll resubmit this again as a calculation. We'll say "Ar–Br + CN-." But instead of a scan, we're submitting this as a transition state. And so what's really important here is we will deselect optimize, and we'll say optimize parentheses ts for transition state. And so this is now just a different optimization. We're using a different algorithm. We're looking for a different thing. And while ground state optimizations are pretty robust, like you can put something in and you'll usually find a ground state, transition state optimizations, you have to be a lot closer to the transition state for it to converge because it's very difficult to do these saddle point optimizations. There's a lot of spurious saddle points. There's actually some interesting mathematics behind this, but the long and short of it is you want to do a scan first so that you know you're pretty close to where the transition state is before you just start running these optimizations. So we'll say "Submit calculation."

Again, this starts running very quickly. And we'll just give this a sec. Let's go back to our scan, which has since finished. And we can now run the whole thing through, maybe from a little zoom out a little bit and get a different point of view. And what we can really see is just this sort of, you know, almost like a video of the reaction happening, which is pretty cool. So. As we go here, you know, we bring the cyanide in and then we can step through with our arrow keys through here. We do actually, we form one brond, we break the other and just like clockwork, the bromide flies off. And what we see is that the products are much, much more stable than the starting materials, which makes sense because this is indeed a reaction that goes forwards.

Okay, so our transition state optimization took... What was this? 24 steps. And what we should see successfully from a transition state optimization is a single imaginary frequency. So this frequency is actually imaginary. It's an imaginary number because we have sort of an inverted degree of freedom when we do the harmonic approximation, like we have a negative spring constant. We usually write it just as a negative frequency because that's a bit easier to think about. What we want to do is we actually want to visualize what's happening here. We can even tune up the amplitude a little bit to see, and what we expect is that the imaginary frequency corresponds to what's happening in the transition state. So this is showing which atoms are moving. sort of in the saddle point, and what we see is actually, you know, the carbon and the cyanide are getting closer together, and the carbon and the bromine are getting farther apart. And so this is exactly what we expect from an SnAr transition state. And what this shows us, this check that we have only one imaginary frequency, and there's the right imaginary frequency, this shows us that we found the correct transition state.

Okay, so that's how to find a transition state in Rowan. If we want to do this in more depth, so if we wanted to do this perhaps a little bit more rigorously for publication, we could then sort of take this and correct this transition state geometry with a single point energy, perhaps with DFT. So something that's slower, but more reliable and more accurate. This gives us the geometry of the transition state. And then if we want to actually convert this to an energy barrier, so if we want to make a potential energy surface, we would take this number here, the electronic energy, subtract the energy of the starting material, and then subtract the energy of cyanide. So we'd also want to run a calculation on CN minus by itself. And then the barrier to the reaction would be energy of transition state minus energy of aerobromide minus energy of cyanide. And that would give us the barrier to the reaction, and we could predict the rate with the Arrhenius equation.