Survivor: Mechanisms (now accepting logo submissions)

gibbsfreepassI read an interesting article in May’s issue of J. Chem. Ed. titled “Can Reaction Mechanisms Be Proven?” by Allen Buskirk and Hediyeh Baradaran of BYU.  Intriguing.  So I pop open the pdf and a Note from the Editor is boxed at the top of the page before the article starts.  It says:

“Can Reaction Mechanisms Be Proven?” generated spirited responses from its reviewers. The reviews were approximately evenly divided, and all were of very high quality. The authors agreed with the editor’s proposal that the reviewers convert their reviews into rebuttals or affirmations of the authors’ position for publication along with the article, which has been revised based on the reviews. Most agreed to such a process and their comments appear here. We hope that publication of this paper and well-reasoned rebuttals such as those provided here will initiate a wide-ranging discussion. JCE will provide an online forum for further discussion of the issue. Our hope is that both faculty and students will contribute their opinions and ideas to this discussion. -JWM

Huh.  You don’t usually hear about that happening too often.  So now I had to read the article.  It’s pretty fascinating, and I encourage you to read it all.  I’ll summarize and give my thoughts below the jump

The paper starts out with a bit of philosophy: how do we know when we “know” something?  Note that textbooks routinely tell us  reaction mechanisms can never be proven.  That is, we always talk about the “proposed” mechanism, and we never leave out that adjective and refer in the definite to “the” mechanism.  The authors conclude this is ultimately a philosophical limitation to knowledge: inductive reasoning cannot actually prove the mechanism of a reaction.  Repeated observations may suggest a mechanism operates a certain way, but 100 observations does not guarantee the 101st observation will definitely be the same.

The only correct way to theorize and “prove” a reaction mechanism – according to this philosophy of knowledge – is to postulate all plausible mechanisms… then run experiments to refute as many as possible.  The last mechanism standing doesn’t even win, though, as it only claims the title of “proposed” mechanism.  This sort of Survivor: Mechanisms mentality implies that “all knowledge is negative.”  We cannot Know (capital K) the mechanism of a reaction, we can only Know every mechanism the reaction does not follow.

The authors criticize this theory of knowledge sometimes referred to as “strong inference.”  We as scientists might think about atoms and bonds and orbitals and electrons and trajectories and charges……. but these are all just pictorial metaphors and mental constructs of the physical attributes of molecules and molecular motion.  Thus, all mechanistic observations are made through these man-made representations.  Does a negative result really negate the hypothesis?  Or is there a flaw in our representation and interpretation of the data?  Furthermore, the “strong inference” model also downplays experiments supporting a proposed mechanism.  They do have merit and deserve to be supporting evidence for a mechanisms – not just a lack of evidence against a mechanism.

As to the history of the elucidation of a mechanism, early mechanistic studies certainly failed to offer conclusive proof of a mechanism – especially the structure and conformation of transition state species.  Steady-state kinetics rely on observations of reactants, intermediates and products.  With no direct observation of transition state species (which have lifetimes on the order of molecular vibrations, or dozens of femtoseconds), definitive conclusions about those species and their formation and subsequent degradation can only be speculative.

However, advances in technology are now allowing the ability to resolve molecular motion to the femtosecond scale.  The field of femtochemistry has allowed observation of some of these exceedingly short lived structures.  Zewail received the 1999 Nobel Prize in Chemistry for his work in femtochemistry utilizing ultra-short laser pulses.  His Nobel lecture is adapted into print here.

So what can we say about mechanisms?  Can we prove them?  The authors conclusion is ultimately a dogmatic one: reinforcing a negative view on the study of reactions and their mechanisms (all we can conclusively do is prove them false) potentially discourages aspiring students from utilizing all possible tools (and very powerful tools, at that) at their disposal.  But they also note that, philosophically speaking, “we can never “prove” a mechanism – or any other scientific theory – absolutely.” (emphasis added).

I think I tend to agree with reviewer David Lewis of Wisconsin-Eau Claire.  In his review, his answer to the title question is: it depends on what your definition of the word “proof” is.  It’s basically semantics.  Do we mean “proof” in the mathematical, Absolute Certainty sense of the word; or do we mean “proof” in the jurisprudence, Beyond a Reasonable Doubt sense of the word?  The answer to the title question changes depending on how you mean “proof.”  A sufficient body of support gleaned from a series of properly constructed experiments can serve to “prove” a proposed mechanism beyond a reasonable doubt … until such time as new knowledge or new technology allows for the construction of more and more thorough experiments to be carried out.

Two interesting case studies: the SN2 and the Mitsunobu reaction.  We all know how the SN2 reaction works, right?  Or do we?  The nucleophile should approach the electrophile from the back face, and – to preserve momentum – the leaving group should continue along the same trajectory as the incoming nucleophile, right?  Well, new evidence from the Wester and Hase labs suggests there might be more to it than originally thought.  Certain leaving groups were travelling significantly slower than expected by conservation of momentum.  The research team concludes that the nucleophile spins the electrophile 360 degC before undergoing SN2 addition.

mitsunobuThe Mitsunobu reaction is a favorite quiz question for first year organic graduate students.  Next time you get asked the Mitsunobu mechanisms, tell your inquisitor, “it depends.”  Really.  The mechanism (click image for larger) changes depending on the nature of the azodicarboxylate, the nature of the phosphine, the pKa of the acidic proton, the phase of the moon, and the record of the Cleveland Indians.  Who’s to say the mechanism of the Mitsunobu reaciton has been “proven?”

I guess my final point would involve the conditions under which a reaction mechanism was “proven.”  A good mechanistic study will survey pH, concentration, electronic effects of the reactant molecules, etc.  But even still, all we can say is this is the postulated mechanism under these conditions.  Or, more melodramatically, this is the postulated mechanism on Earth. Remember, a reaction can theoretically proceed under any number of ionic, radical, diradical, and fragmentation pathways.  The energy surface for a reaction to occur is more like a mountain range.  The pathway requiring the lowest energy is the “accepted” mechanism.  But given sufficient energy and reaction conditions, other mechanistic pathways are attainable.  Given that the possibility for alternative mechanistic pathways exists to me says we cannot “prove” one mechanism as the mechanism.

In fact, this is kinda what the paper is talking about when it discusses eliminating all other possible mechanisms.  We can think of all sorts of crazy mechanisms to get to the product, and each of them will take a different path through our mountanous region of the energy surface.  We probe each mechanism to test its validity, and only the strongest mechanism surives.  This is why I liken this approach to Survivor.

Ya, know, I think we need a logo for Survivor: Mechanisms.  I’m no good with photoshop (or GIMP, for the open-sourced among us). But if you are, and you want to put together a logo for our Survivor island, you can email it to me, and I’ll post them here on the blog.  We’ll also need a name for our Survivor island.  I suggest Gibbs Free Pass (slogan: Minimal Energy Required).

Oh, and what do you think?  Can a mechanism be proven given our advances in technology?  Or is this all just a semantic triviality?

I notice Sabbaitcal Epistles also covered this paper


  1. A very interesting topic. I think another point needs to be made: The “reaction mechanism” we write is really only an abbreviation for the things that really happen. We don’t (usually) take things like atom trajectories or timing into account, or only on a very crude level. So the written mechanism is just shorthand for the main events of the actual mechanism.

    If you compare the energy surface to a mountain range, it might be possible that more than one mechanistic pathway is involved. Given enough energy, a more unfavourable pathway might take place. Looking at the ensemble of molecules, a distribution over all possible reaction mechanisms should be present. We usually only talk about the most favourable one.

    To conclude, I would say that it is impossible to prove a reaction mechanism as long as we don’t define *what* we want to prove. If we make a very clear statement about a mechanism, however, it is possible to gather evidence for (or against) that statement. For instance, I could say: “The SN1 reaction mechanism involves a carbocationic species”, which can be proven, as far as the scientific method allows a “proof”.

    • The SN1 reaction mechanism…

      You mean ‘the lowest energy accepted pathway for the SN1 reaction mechanism…’ 🙂

  2. I can’t discuss the particular mechanism in question but while I was at a Fortune 500 company I was involved in some work that suggested that a tried and tested, documented to the hilt mechanism might be invalid. There were boat loads of data validating the mechanism and then this one observation…serendipitous too. Then over the next year we conducted piles of experiments invalidating the documented and accepted mechanism. It resulted in us inserting one more step. And then another observation…one that many years later we STILL cannot explain..but is repeatable…invalidated our improved mechanism. But, we cannot explain it. It’s the ONLY piece of chemistry that remains unresolved for me..but its proprietary :-(. Mechanisms are consistent representations for the observations…not the truth. The typical “data supports the hypothesis that..”

    By the way…did you see this? Call me 🙂

  3. One important point Theodore Brown pointed out was that in a sense positive information like NMR or crystal structure is actually negative information in that if the NMR turned out differently it would disprove the structure. I would add that it is also negative information in the sense that it disproves other proposed structures that conflict with the data.
    David Lewis’ point that false evidence does not disprove a theory is also interesting. He writes that negative evidence only tells us that either the theory is wrong or the theory behind the test is wrong. I think that he should add that it might also prove that the theory behind the application of the test is wrong. What I mean is, a negative result either means that:
    a) the proposed mechanism is wrong
    b) whichever test is being used is invalid
    c) the test being used is invalid in this case

    • David Lewis’ point that false evidence does not disprove a theory is also interesting. He writes that negative evidence only tells us that either the theory is wrong or the theory behind the test is wrong.

      This is a good point. It’s true for all aspects of science, not just mechanism elucidation. This reaction isn’t working… are the reagents just not compatible, or do I need to change the temperature or concentration?

  4. Yes, the set of articles in J Chem. Ed. are rather interesting.

    But according to my reading NONE of the reviewers agreed with the authors that mechanisms can somehow be definitively established even with sophisticated modern methods. So why did the editor feel compelled to publish the original paper along with the four reviews? In any case it is good to see more discussion of philosophical topics in chemistry.

    The comments on Popper make fascinating reading. Ted Brown, one of the reviewers, is especially dismissive of Popper, I think rather unfarily. He claims that Popper knew no science and was basically into psychology when he wrote his famous work on refutation.

    Overall one has to wonder whether the authors of the actual paper under discussion might have some hidden religious agenda for wanting to go against everybody else and claim that some things can be confirmed contrary to Popper’s view.

    Just to mention a conference, International Society for the Philosophy of Chemistry, annual meeting, Philadelphia, mid-August. You can Google the society website for full details. I cannot recall the exact dates.

    eric scerri
    author of The Periodic Table, Its Story and Its Significance, OUP, 2007.

  5. I also want to suggest that it is not Platt’s “strong inference” that is at the center of this issue but rather Popper’s method of conjecture and refutation, as the authors themselves point out very clearly.

    I was a bit puzzled that in his interesting comment above, azmanam does not even mention Popper but seems to get side-tracked by this ‘strong inference’ idea due to Platt.

    eric scerri

    • Quite true. My original post was much longer. I realized I should probably edit it if I wanted people to read the whole thing 🙂 Thanks for pointing that out

  6. I’m glad that in science, the data doesn’t change based on the philosophical dogma the author chooses to view the world through.

    • The argument for scientific reductionism seems to come down to limiting the range of possibilities through observation such that problems can only be solved deductively. Deductive proof, of course, being the traditional highest standard of certainty. Can mechanisms be proven in this way? The argument from the authors seems to be no, but that we can close enough for their standard of proof. Good for them. Their argument is that heuristic certainty should be lower bound for proof. Ultimately though, any mechanism found in this way can be proven wrong (only if it is wrong, of course), while any mechanism found deductively (hypothetically) cannot be possibly be wrong. Their argument seems to be that if we don’t know, the judgment should be made nonetheless. It’s throwing your hands up in the air and saying, “Well, close enough!” No, it’s not.

  7. cookingwithsolvents says:

    Thanks for the extremely interesting articles. In general, I also agree most with Lewis in the “reasonable doubt” vs “absolute” framing of the problem of what constitutes “proof”. I have a bit of my own to add, though.

    Insisting that you can prove a mechanism implies that we are capable of imagining/deducing all the POSSIBLE routes that A could take to B. I’m not OK with that. Indeed, there is a decent amount of hubris involved in that line of thinking. Our conceptual models of reality are incredibly powerful and predictive but they are still models (which is mentioned in the paper and they STILL make the leap that mechanisms can be proved in an absolute sense).

    I am frequently reminded in my work that most chemical reactions are relatively RARE events among the collisions going on in your flask or in your CVD reactor or wherever. One talk on time-dependent calculations will show you how little we understand about the phonon coupling which prepares a molecule for the transition state. There is the pesky quantum mechanics/observer problem inherent in all of this, too.

    All of this takes nothing away from the practical utility of knowing the most likely course of an SN2 reaction (e.g.). Chemists make and break bonds OK with our limited understanding of what’s really going on. . .it’s the ensemble that matters anyways, right? Certainly the authors are correct that we should strive for further understanding.

    The debate is informative and certainly one every scientist should think about and revisit from time to time. I will probably use these articles for teaching at some point.

  8. delta_star says:

    I’m not a philosopher, but the notion that we can only know negative knowledge contradicts my intuition somewhat. Before I would have asserted we can only know positive knowledge – I know that there are black and white zebras. Are there red zebras? I can’t definitively say there’s no such thing as a red zebra just because every zebra I’ve seen is black and white. I can only grasp what I see…but I guess reaction mechanisms are a different beast from zebras.

    In any case, my view of science is that it’s always been changing, modifying – everything we know is in “models”. Some models work well but fail when we dig deeper. Nevertheless a good model is always a good thing to have, if only for its utility in designing better technology, etc. Regardless of how well a model approaches “truth”, the real societal utility of science comes when we use our understanding to make better cars, faster computers, etc. It might be counter to the dogma of science as the seeking of truth, but it’s the reality of the matter, and I don’t think the two are mutually exclusive. F=ma worked for a long time, and still does a damn good job. Even if we know it’s not really true (e.g. speeds near c), it describes a model of our physical world that we can use to improve our lives. Will we ever get the perfect model? Will we ever know the truth? Will we ever really know a reaction mechanism? Who’s to say we can’t? Given our most current model of physics and what sorts of spectroscopic methods we’ve come up with based on it, mechanisms are sort of hard to get a handle on, and we use a lot of other rationales for characterizing them (kinetics, etc.). But who’s to say we won’t have another quantum revolution?

    • The “perfect model” does not exist – that would be reality itself. Again, this is a philosophical rather than scientific argument.

Leave a Reply

Your email address will not be published. Required fields are marked *