This must be the most damning quote of 2015 in chemistry land: "Computational predictions vary vastly, and it is not clear that any significant accurate information that was not already apparent from experiment could have been garnered from computations". The authors are Erik Plata and Dan Singleton (DOI). It is thanks to Arthur Winters news&views contribution in the June issue of Nature Chemistry (DOI) that this blog has become aware of their article (sorry!). It has already been making some waves at C&EN, Computational Chemistry Highlights and In The Pipeline. So what was it all about?

Plata and Singleton experimentally reinvestigated the kinetics and thermodynamics of a typical Morita Baylis-Hillman reaction (MBH reaction) and then compared the results with different computational methods. On a simple level the MBH reaction is a simple Aldol reaction / elimination but as Plata and Singleton explain (and Wikipedia can confirm) researchers like McQuade and Aggerwal made much more of it, for example introducing proton-transfer effects.

Practical complications: the model reaction chosen is an equilibrium reaction, reaction intermediates can be protonated and their precise concentration may not be known. In the end it all boiled down to the generation of an experimental free-energy profile with a total of 5 activation barriers. Main conclusion: The "simple" MBH reaction mechanism is the most viable. Surprisingly the computational profiles did not even get close. It was like a weather forecast for next week with temperatures not more precise than between -5° and 30 °C. Calculated equilibrium constants were off by 10 to 25 orders of magnitudes. Activation energies varied by 50 kcal/mole. In several predictions the overall reaction was supposedly endothermic. So what went wrong? Plata and Singleton are not very specific except when they state that entropy effects are not part of the problem. In their view many different computational predictions can be made for any given reaction and only the absurd ones are dismissed.

Arthur Winters in his commentary notes that "experimental chemists needn`t fear being replaced by computational overlords in the immediate future" and asks the question if "abstinence (is) the best approach for computing reaction mechanisms that are challenging for current theory". The solution would be to be more careful. In any event "the terror of having ones work discredited in the footnotes of a future Singleton paper is sure to cause some computational chemists to wake in the occasional cold sweat in the early hours"