This reaction looks simple but is deceptively complex. To recapitulate: tolyl thiolate (X=Na) reacts with the dichlorobutenone to give two substitution products in a 81:19 ratio, a result that Singleton and Bogle argue arises from a statistical distribution of dynamic trajectories bifurcating out of a single transition state favouring 2 over 3. On the grounds (presumably) that the presence of both the cation X (=Na+) and H-bonded solvent (ethanol) are uninfluential, neither species was explicitly included in the transition state model used to derive the dynamics. I speculated whether in fact the spatial distribution of counterions and solvent (set up by explicit hydrogen bonds and O…Na+ interactions) might in fact be perturbed from un-influential randomness by co-ordination to the carbonyl group present in the system. I also raised the issue of what the origin of the electronic effects leading to the major product might be.
In this post I try to delve deeper into both these issues. In the earlier model, I focused on possible coordination models around that carbonyl, using two Na+ cations (on the premise that such coordination has precedent in crystal structures). This model did (correctly) predict this major product, and we are now discussing what the origins of the minor product may be (it is a measure of how far computational modelling has come that we are nowadays increasingly concerned with these minor outcomes). Here I move to a more stochiometric model using just one Na+ assisted with four solvent molecules (modelled here with just water). This results in an overall charge of zero on the whole system, which avoids having to create what could be regarded as artificially charged models resulting from omission of the counterion. Three possible arrangements of these additional units are shown below, selected for the following reasons:
We might ask why stop at just these four? Surely one should sample all reasonable explicit models that might have a significant Boltzmann population in the real reaction? That is certainly desirable (but a much larger computational project); here I am just using these models for the purpose of understanding a little better what might be going on.
Model (a)
This is optimised using the same level as before (B3LYP/6-31+G(d,p)/SCRF=ethanol) and reveals that the Na+ cation ends up with coordination just from solvent, and not from the aryl face. The chlorine labeled green in the diagram above ends up being evicted, and its trajectory then leads it (slowly) towards the Na+ cation in a reaction that is fully concerted (no enolate anion intermediates along the route).
The IRC for this model has the following intriguing features:
The next task is to see how stable the above effects are to the disposition of the Na+ and solvent molecules. Model (b) shows the same behaviour; the chlorine atom is evicted via stereoelectronic control, rather than simply heading off towards the Na+ atom (i.e. electrostatic control).
Model (c) also demonstrates how the stereoelectronic alignments dominate over stabilisation of the forming chloride anion. This time, the chloride is evicted into a region not occupied by either solvent molecules or the Na+ ion, the charge being stabilised only by the continuum solvent field.
Model (c) was also subjected to a robustness test of the actual wavefunction. The original method was based on B3LYP/6-31+G(d,p)/SCRF=ethanol. Accordingly, (c) was re-computed using ωB97XD/6-311+G(d,p)/SCRF=ethanol. The DFT functional is a more modern one that includes the effects of dispersion attractions, and the basis set is of triple rather than double-ζ quality. The essential features are unchanged.
Model (d) tests whether perturbing the electronic environment has more effect than changing the explicit surroundings.
I will conclude by summarising the above. The formation of the dominant product 2 seems to be the result of stereoelectronic control at the transition state. This outcome seems to be pretty robust to the transition state model constructed, namely whether one (or two) Na+ counter-ions are included in the model, and indeed their position, as well as the inclusion of up to four explicit solvent molecules. This robustness even extends to an electronic perturbation resulting from replacing a C-H bond by a C-F bond. Thus constructing a selection of physically realistic models with neutral charge and solvent has not resulted in locating an explicit transition state which (in terms of its free energy) might account for the formation of the minor product 3.
Another test which might be envisaged would be to take e.g. model (a) and subject it to molecular dynamics to show that the outcome, in which ~20% of the trajectories lead to 3, is itself robust towards addition of counter-ion and solvent to the original model.
These values do seem to be very basis set dependent. Thus using B3LYP/6-311+G(d,p), the σC-Cl(green) to σ*C-S value is 58 and σC-Cl(red) to σ*C-S is 18. The trend however occurs across basis sets.
In an earlier post, I discussed a phenomenon known as the "anomeric effect" exhibited by…
In the mid to late 1990s as the Web developed, it was becoming more obvious…
I have written a few times about the so-called "anomeric effect", which relates to stereoelectronic…
The recent release of the DataCite Data Citation corpus, which has the stated aim of…
Following on from my template exploration of the Wilkinson hydrogenation catalyst, I now repeat this…
In the late 1980s, as I recollected here the equipment needed for real time molecular…
View Comments
I noted in a footnote above that the NBO E(2) energies seem sensitive to the basis set used. I also noted that if ωB97XD/6-311+G(d,p)/SCRF=ethanol replaces B3LYP/6-31+G(d,p)/SCRF=ethanol for model (c) the essential features were unchanged.
But the quantatitive aspects do look different:
For example, the gradient norm at IRC +2 takes a different form for B3LYP/6-31+G(d,p) compared with ωB97XD/6-311+G(d,p). The derivative of these gradients also shows differences. This suggests that the latter is perhaps closer to forming a transient intermediate along the path than the former and also hints that the forces acting on the atoms along this path may in turn be quite sensitive to the method used. Another smaller, but intriguing difference is that the gradients for B3LYP/6-31+G(d,p) are quite "bumpy", whereas ωB97XD/6-311+G(d,p) is much smoother.
We may speculate whether these quantitative differences will impact upon the molecular dynamics calculated using either of these methods, or whether the dynamics turns out to be quite robust to basis set and DFT method.
Oh, you might ask why I use ωB97XD in preference to B3LYP. Well, calibrations on relative energies of a collection of small molecules compared to CCSD(T) calculations suggest that the former matches much better than the latter. For systems where weak interactions between assemblies of molecules matter, ωB97XD does seem to come up with quite realistic simulations. And I believe that dynamics simulations can be sensitive to the potential used.
PS: For proper comparison, the below is the IRC gradient norm computed at ωB97XD/6-31+G(d,p) and ωB97XD/6-311+G(d,p) (no additional counterion or solvent). This shows the the big difference is the functional and not the quality of the basis set, or indeed the presence of counter-ion and solvent.