Is science sometimes in danger of getting tunnel vision? Recently published ebook author, Ian Miller, looks at other possible theories arising from data that we think we understand. Can looking problems in a different light give scientists a different perspective?

Share this |

Share to Facebook Share to Twitter Share to Linked More...

Latest Posts

Archive for February, 2019
In my last post, I promised to expose one of my failures, so let us start. If nothing else, it might be a lesson for those beginning their adventure in research. It all started properly with the start of my PhD. I had previously done some research while still an undergraduate, without a lot of success and I think that was because my supervisor was using me to try out some "wild ideas". In a sense, that is fair enough, but I think that it would have been better if he had tried them himself. The point is, it takes skill to make something wild work, and by definition, the beginner is short on skill. Anyway, for my PhD I was given (at the beginning of December – NZ is out of phase with other places) quite a respectable project, and I did the necessary background reading, then, with a few days before Christmas there was little point in starting so I did general reading. The day before we were all to break for Christmas, I presented my supervisor with the answer: he was scooped in the latest JACS. After Christmas, he gave me two new projects, then went on summer holidays. These were awful. For example, one involved measuring the rate of some reactions. The problem here was there was one literature example and the rate was zero at seriously elevated temperatures. The Head of Department saw my disconsolate appearance, and he suggested (to give me something to do, I think) that I find a project for myself.

I tried the question, does cyclopropane conjugate? This might seem partly lazy, but (a) it was a real question, judging by the JACS papers at the time (which also made it easier to write the thesis introduction) and secondly, prior to the PhD I had done some work on the Hammett equation (my first paper!). It was known that functional groups adjacent to a cyclopropane ring were found not to behave the same as the same groups adjacent to, say, cyclohexyl or isopropyl. There were two possible explanations. The first was that electrons in the ring bonds behaved similarly to olefinic electrons, and could delocalize into the adjacent function if it were unsaturated, but by not as much. The second was that they did not delocalize, but "somehow the strain was responsible". The problem with the strain explanation was the "somehow". The advantage of the delocalization proposition was that it was easily understood, at least in principle.

As I saw it, the Hammett equation was interesting because some substituents had two values of the σ"constant": if delocalization to the site was interrupted by something that did not permit delocalized electrons to cross then it had one value, and if there was no such interruption, a different one. At the time, a lot of work had been put into establishing these constants; here, I thought, at last was something that was worth using this equation on. In fact, others thought so too. There were two papers, both relating to the various substituted trans-2-phenycyclopropanecarboxylic acids. One measured dissociation constants in water and found no sign of delocalization in that the ρvalue was slightly less than for phenylpropionic acid. (A vinyl group as in cinnamic acid maintains an enhanced value of ρ, while methylene groups attenuate it.) Measurement of the dissociation constants in 50% ethanol led to the opposite conclusion in that the ρvalue for the cyclopropyl group had a mild enhancement over the propionic acid. 

So that proves delocalization? Hold on! What I argued was that when the ρvalue was measuring inductive routes, it was experimentally observed that the ρvalue was attenuated in a regular fashion by the number of intervening methylene groups. I argued that the cyclopropyl group offered two routes, so it should be higher. Despite the fact it took my supervisor an awfully long time to write up my work, I believe the significance of this second route was still announced for the first time. As for the σvalues, the ρvalue, which measures the slope of the line and thus the degree of difference a substituent makes, had too low a value to reach a conclusion. Also, the values were really too erratic, especially those in water, and I felt the chances were the samples were not properly soluble. My plan was to make the amines, and measure the dissociation constants and the rates of reaction with a nucleophilic substitution. My supervisor was not impressed, and I felt he did not want to go into "risky" territory, but he did not have a better project, so off I went.

Thus far, only the unsubstituted amine had been made. That required synthesis of new compounds. There were two ways of going about this: the long way to the acids or through one of a few newer methods, but ones that had not been used for these syntheses. Only one of the newer methods worked to give the unsubstituted acid. I was able to purchase some starting materials that would give me substituted acids if the reaction worked, but they did not turn up until the day before I was to set off on my first post doc. They had to come by ship from the US, and somehow they got sent to Thailand and stored in a shed until much later they were located and forwarded. Thanks, shipping company! So, back to the long route. The main route employed to get to amines was the Curtius rearrangement on the acids, which were known, and each amine involved a 13-step synthesis.

So eventually I got my amines, then a weakness in the plan emerged. The only substituent I could make that would have an adequate variation in its σvaluewas the p-nitro substituent. (The problem was not in making different conjugative electron-withdrawing substituents, but rather having literature records of their behaviour. I needed σconstant variation with functional groups like aminesand they were not available.) I made the hydrochloride of that easily enough (I made hydrochlorides because they are much easier to purify and they are stable.) It was only when I came to measure the dissociation constants that I had a very unpleasant surprise: even at the dilution required to measure UV spectra, the p-nitro substituted amine became unstable, and the solution slowly turned brown. I could have made a rapid determination but I decided not to as I could not guarantee the significance of the answers. What particularly bothered me was the possibility of observer bias. The results would be too error prone. Of course I made all the measurements I could, and got neat graphs, etc, but I had a problem in that I was not answering the major question. The ρvalue of both the dissociation constants and the nucleophilic reactions were more or less where I expected, based on the attenuation effect of methylenes, but the comparison was not as firm in terms of reference compounds.

My problem was clear. I had tackled a big project and failed. I had made new compounds and measured some dissociation constants and reaction rates, so I would probably get a paper, but it was not one to set the world on fire. And it most certainly was not one to help get an academic job in the future. What to do? More in the next post.
Posted by Ian Miller on Feb 9, 2019 10:43 PM GMT