© WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Computational Chemistry: Playing Darts?
This essay written by three authors who have contributed so much to the visibility and success of computational chemistry is both useful and wonderful to read. I do understand, though, the point of view of the most severe referee. A good part of what is discussed should have been acquired during undergraduate or graduate studies. However it is not. That is an unfortunate fact. The word "stability" is probably one of the most mistreated concept by practicing experimentalists and theoretical chemists alike. How many times have I heard: can you calculate this molecule for me? I want to know why it is stable. On the other hand, there will be a minimum (and thus a "stable" structure) for most combination of atoms. I do remember a colleague with a very dry sense of irony who suggested to throw darts at a Periodic Table, collect the resulting selected set of atoms and optimize a structure. In most cases, a minimum will be found! Not the best way to predict a new molecule, we all agree. To correct this frequent misuse of the meaning of "stability", this paper would do a much more efficient job than to send back most of us to undergraduate classes.
I also agree that the HOMO-LUMO gap criterion is biased towards organic and inorganic molecules but it is still a powerful criterion. Let us use it and teach it, together with its limitations.
About accuracy for structural parameters and for energies. The values suggested by J. A. Pople look so reasonable and are used by a large part of the community, indeed. How can they be ignored? Apparently, they are. One still sees plenty of results in submitted papers and in posters with data given with abnormal accuracy, which as the authors point out translates to a few inches in the distance between Ithaca and New York. Of course, it looks even sillier in SI units (which the authors should have been using under IUPAC guidelines) when the distance is 235.714*1.609344 = 379.344912 km, including all nonsignificant digits.
The problem of highly charged species is not obvious. A charged species exists with its counterion. Ignoring the counterion may lead to wrong conclusions on "stability" or reactivity even though calculations of the isolated ion have given and will still give useful information. The real system (the ion, its counterion and the solvent) is still out of reach to the computational chemist in many cases. That should not stop the computational chemists from carrying out the study of charged species after selection of the most reasonable model.
The authors have not elaborated much on the need to compromise on the quality of calculations to treat very large systems. This is done currently with great success. In other words, what is better: an accurate calculation on a model far from the reality or a less accurate calculation on a model closer to the reality? “Les deux, mon général” as we say in French. The key point of doing a calculation is to know why we do it and what is the question to be answered. This is where theoretical and experimental chemists communicate and exchange. Otherwise carrying out calculations could be as useful as to "pee into a violin" (My apologies for a slightly crude expression but this widely used French expression really conveys the correct sense of hopelessness). With this in mind, I entirely follow the authors in their recommendations for improving the "accuracy" of our language.
Institut Charles Gerhardt Montpellier (France)