Contemporary problems in chemistry are in fact problems in information theory
Chemistry is considered to be an exact science. But it isn’t.
Truth is, we haven’t come nearly as far as we pretend to since alchemy split into psychoanalysis and chemistry. Recently I conducted a small experiment. I selected four small organic molecules and presented their structure to well-trained chemists, asking them to assign to each structure one out of four given properties:
- a painkiller,
- a semiconductor,
- a perfume ingredient,
- or a metal surface protection agent.
They failed miserably.
Out of 10 participants 0 were able to assign even a single (!) structure right, among them professors of chemistry.
If you want to get drunk for free, just invite chemists out to a bar and play a little game with them. Let them guess molecular properties — if they fail they owe you one. Honestly, the stakes are low. Right now there is little to no possibility to deduct molecular function from molecular structure. The only way for a chemist to win the getting-drunk-game would be to know beforehand which role has to be assigned to which molecule.
All we know in chemistry today is molecular structure, i.e. we can tell what type of atom is connected to which other type of atom. The whole bunch of interconnected atoms is what we call a molecule. While it required quite an effort only some decades ago to find out the chemical structure of a molecule, we are meanwhile actually pretty good at that. But if we want to know what the molecule will do when exposed to different environments is hard to tell, and there’s only one way to find out: by testing it and memorizing the result.
Chemists found a way to represent molecular structures only in the second half of the 19th century. That led to a consequence as simple as it is obscure: the depiction of matter with structural formulae enabled chemists for the first time ever to see and talk about what they were dealing with. Structural formulae showed chemists how atoms stick together to make molecules, thus enabling them to find ways to rebuild the sought-after examples from nature in the laboratory and the factory, eventually leading to the advent of modern industrial chemistry with all great successes like the famous synthetic Indigo dye to name but one.
The construction of chemical reality
The connection between structural formulae and modern chemistry, or actually the modern world to a large extend, can’t be overestimated. Structural formulae separate chemistry from alchemy and the reason for this is access to information. Access to information separates mythology from ontology, belief from knowledge, a magical world from an enlightened one.
Of all natural sciences chemistry always had the problem of a lack of ontology. While a phenomenon of classical physics is quite “real”, a chemical reaction is not. Hard spheres rolling down an inclined plane, an apple falling onto your head or bodies of different volume can be touched, measured and understood from their very existence, i.e. they can be described and talked about immediately — they have an ontology. That type of immediate description is not possible in chemistry, or to be more precise, it is possible, but it doesn’t help. All you could state immediately would be something like: mixing a white solid substance into a clear liquid and adding another clear liquid will yield upon drying in a white solid substance. What does that help? What you could probably find out with your al-chemical skills is that something happened. But not what, how or why, leaving you unclear about the rules of the game, preventing any true understanding, opening the doors for all strange ideas, mixing with depth psychology eventually leading to semi-magical “explanations” of matter and its behaviour.
Thinking more closely about it, the first task of modern chemistry was actually a task of semiotics, long before a modern theory of signs was defined by Morris (et. al.). It was the task of getting an ontology, to be able to see, imagine and talk about the structure of matter. That ontology seems to be correct. The structures of molecules as we draw them seem to represent the connections of atoms to an extend that is near certainty. But we must not forget, that these structural formulae are manmade, that we have an ontology that is manmade. Still we can’t see matter as it is, we can’t see atoms and electrons in action, still we can only see matter in the way we depict it! Structural formulae are in fact a sophisticated and ingenious language to talk about what we think is happening in matter during a chemical reaction.
Don’t think about things, think about function
As much as structural formulae provide access to information on structure (hey, they are called structural formulae after all) as little do they provide access to function. To some extend can we detect patterns in them that seem to be necessary for certain chemical functionality, e.g. COOH-groups for acidic behaviour, or certain electronic structures that give rough estimates on the substance’s colour. But the understanding of chemical functionality or properties is too rough to give us tools to really predict what will be. Being able to predict is a core requirement for an exact science, and the apparent failure of chemistry in prediction is why it largely remained to be an art form. We should at least doubt whether to name chemistry exact.
This is a serious issue. In the 19th century physics was declared to be nearly finished. People like Max Planck were told that “eager young men shouldn’t flock to physics, because everything has been invented”. Then thermodynamics led the transition to electrodynamics and quantum theory, opening a whole new era of unforeseen technical progress. Using mathematics as a language to describe physics changed the ability to apply it, by moving from an ontology of things to an ontology of function. Chemistry has not made that step; it is stuck in the ontology of things: there are only structural formulae, no functional formulae.
In physics the application of mathematics was a step of abstraction: from the things themselves to their function represented in mathematical ways. Doing the same for chemistry would mean to abstract from structural formulae to a new language that we could call functional formulae. But aren’t structural formulae already an abstraction? I think it would be highly impractical to invent a second language for chemistry. Maybe it is doable and maybe the result would be great. But I think there is a much more easy way: we should progress in the same way like 19th century physicists did, i.e. by applying statistics to the phenomena we care about.
After all it is pretty straight forward. The crucial point is to understand that contemporary problems in chemistry are in fact problems in information theory. We do know the structure of matter, and we do know the phenomenon, i.e. the function molecules exhibit, e.g. a certain smell. What we don’t know is where the phenomenon has its origin in the structure, as structural formulae don’t provide access to function. A thorough statistical investigation should match molecular patterns to molecular phenomena and reveal many facets of molecules we don’t understand so far.
In the case of chemistry we do not need to move to a second abstraction layer, we need to explore the language we already invented for all meaning it bears. We need to move from a constructed theory of reality to a true theory of signs. We need semantics for chemistry. That is the great task we are faced with today.