Following on from last week’s post, getting my PhD did not go quite the way I had envisaged. The night before the oral exam, my fiancée terminated the engagement. Just before the oral exam I had arranged to get a smallpox injection, and I was not prepared for the rapidity of what came next. At the end of the exam I was sweating, and not because of the questions. I went to my car, and my left arm was the nearest to being paralysed and I needed it to change gears. Then I found out my mother was going into hospital for breast cancer (the good news was it worked, and she lived for quite a lot longer and she did not die of cancer.)
So, when I went to North America for a post-doc, and being lonely, I had time to work on a paper on strained molecules, and to explain exactly what I thought was going on. Because this was my first scientific paper, I got it wrong. No, not the work, but the presentation. Unless you were a big name, there were strict page limits on what a “newbie” could expect to get published. What I aimed for was to present a general method for calculating the properties of strained molecules that were dependent on the strain. In the associated post, for those interested, I shall give a description of the science as I see it, but for those that are not so interested in that, basically what I wanted to write was a means not only of accounting for what we already knew, but to present an approach that would make predictions for molecules that were yet to be measured, or even made.
The first draft was too long. It involved four parts: a general discussion of strain and the philosophy of the approach, the calculating method, a new proposed method for estimating the strain in a molecule, then the results of the calculations in terms of the ring bending strain in molecules and the resultant dipole moments. I had to shorten the paper so what I did was to cut out most of the first part, because it was “obvious”. It was obvious to me, but it appears nobody else could see it. What I should have done was to cut out the method for estimating the strain in molecules, and submit that as a separate paper because it could stand on its own and it had merit outside the paper I was writing. As I recall, the peer reviewers let it through without comment.
I wrote a series of further papers, and these showed that the properties of strained systems adjacent to charge or unsaturation were properly explained simply through standard electromagnetic theory, and no special quantum effects were required. More importantly, the predictions of parts of the electromagnetic explanation were opposite to those of the quantum delocalization explanation. Thus while both theories predicted that adjacent positive charge would be stabilized, only the electromagnetic explanation predicted that negative charge would be destabilized, so there was a clear and discrete difference between the two possible explanations. A limited number of spectral transitions could be used to separate these two theories, and where electrons moved towards the strain in the excited state, I predicted a shift that was in the opposite direction to delocalization. Even more importantly, experimental results showed I had calculated the shift almost exactly. Standard theory could not even get the direction of the shift right, so in principle the delocalization theory was falsified.
So, what happened? Two reviews came out, asserting the delocalization theory was right. End of story. How did that happen? First there was a review restricted to spectra. The review simply dismissed the few examples such as those I had identified that gave opposite effects for the two theories on the grounds these compounds were not very important!! Real science in action??? There was no mention of my paper on the subject, although it was possible that when that author submitted his review the paper may not yet have been in print.
The second “authoritative review” was much worse, because for no good reason, it ignored all my work. Worse than that, much later I tried to write a counter review, carrying out a logic analysis on all the data as of then. What I found was that there were about sixty different types of observations that were not in accord with the delocalization theory as generally presented. Now, some of those might have been explained away, as always there are some “special circumstances” but sixty is an awful lot. That review could not get published. Reason: the journals “did not publish logic analyses”! So, what did this “authoritative review” do about such papers? Simple. It ignored those as well. Basically, it found what it thought agreed with what it wanted (and some was debatable) and it ignored what it did not want. That, to me, is not real science.
The clincher, in the “authoritative review”, was that molecular orbital computations proved the cyclopropyl ring did delocalize! Quantum mechanics is obviously right, so this must be right. The problem here, of course, is that quantum mechanics produces equations that cannot be solved for systems like this, so all sorts of approximations have to be made. They prove nothing; they might predict something, but in this case they accounted for what we knew, but how? As an aside, calculations from exactly the same school of computations (I.e., the same programs were used) proved the stunning additional stability of polywater.
Never heard of polywater? That was a blot on science. Tiny amounts of water were collected by distilling them through microfine quartz capilliaries, and the water had a much higher boiling point, about 30% higher density, and a much higher viscosity. (That sample had dissolved silica.) It was shown to have a different infrared spectrum to that of water (later to be found to be the spectrum of sweat – another lowlight!)
To illustrate why I distrust computational chemistry, much later a new form of delocalization was proposed. The first such paper (Dewar, M. J. S. J. Am. Chem. Soc. 1984, 106, 669-682.) argued that if bond bending is simple harmonic then (U)bend = kq2, with k the carbon-carbon bond bending force constant, q the angle of deformation from the tetrahedral angle, in which case the strain energy of cyclopropane should be about 437 kJ/mol. It is actually about 120, so he calculated, using molecular orbital theory, a proposed s aromaticity to correct this, and came up with almost exact agreement with observation. Shortly after, there was a counter (Cremer, D.; Kraka, E. J. Am. Chem. Soc. 1985, 107, 3800-3810.) This paper argued the force constant used was wrong, and the strain should be 313 kJ/mol. Then, using exactly the same variant of molecular orbital theory, he calculated a s aromaticity of 200 kJ/mol, again claiming exact agreement with observation. It is impossible (at least for me) to work out where the difference was in the two sets of computations. (My equation for calculating the strain had no excess strain, and the function was proportional to sinq. There is a big difference in functionality between sinq and q2.)
So, I am a sore loser? I should just go away and rejoin the mainstream? Well, I am not entirely alone in this. One of the best current mathematicians and theoretical physicists, Roger Penrose, has just written a book called “Fashion, Faith, and Fantasy in the New Physics of the Universe “. For him, the success of quantum mechanics makes physicists insensitive to the theory’s conceptual problems and generates an unjustified degree of faith in its so-called “basic principles”. Interestingly enough, he cites chemistry as an example of quantum mechanics impressive record. One of the problems with computational chemistry is the equations are insoluble. In his Nobel lecture, Pople introduced the process of validation, which involved “the optimization of four parameters from 299 experimentally derived energies”. This sets the parameters to be used on similar molecules, and as far as I am concerned is not much better than empirical relationships. What happens if the nature of the molecules change? It appears they re-validate. When Moran et al. (2006 J. Am Chem Soc. 128: 9342-9343.) used some computational programs then commonly in place and available as packages for the general chemist, they found that when they applied them to molecules that had been satisfactorily computed earlier, they got wrong answers. The change in validation now lost the ability to calculate earlier molecules, and, as an aside, these errors were dramatic, thus benzene was no longer planar.
That is why I am less than impressed with modern science. It is too ridden with fashion. The reason? I believe it is due to the funding mechanism. Nobody wants to go back over accepted material. If it were yours, you have trouble getting funding the next time. If it were someone else’s, that person had better not be a peer reviewer. Better to go with the flow. My trouble was, I could not get myself to do that.
At the time of the cyclopropane issue, there was another young chemist who carried out an experiment that came to a similar conclusion that there was no delocalization. As things settled down, he abandoned the area, got into a “hot” area, and eventually became very prestigious. I asked him about that paper once and why he left it alone. “Oh that,” he said, with the air of someone who wished I would go away. “That was just . . . ” and he gave a shrug of disinterest. Even if you are right, it is better career-wise to forget it and get back into the flow. Unfortunately for me, that was not the science I signed up for so long ago.