Disappearing polymorphs

Why can a compound made the same way for years suddenly refuse to crystallize into the form the factory depends on?

A plant has run the same crystallisation for years. Same supplier, same solvent, same cooling profile, same operators. Then one batch comes out as a different solid — denser, higher-melting, less soluble — and fails specification. Worse, every batch afterwards comes out that way too, in that plant and soon in others, and the form the process was built on cannot be made again.

The molecules did not change. The recipe did not change. So the thing that changed cannot be a property of the substance at all, and that is the assumption worth giving up: that a compound has a solid, the way it has a molecular weight.

Start with what a crystal is. Not a molecule but an arrangement of molecules — a repeating packing. A given molecule can often be packed in several distinct ways that each repeat consistently, and each such packing is a polymorph, with its own density, melting point, mechanical behaviour and, crucially for a medicine, its own solubility.

Ask which one is correct. Only one can have the lowest free energy at a given temperature and pressure; that one is the stable form and the rest are metastable. And here is the link that makes the failure comprehensible: the more stable form is necessarily the less soluble one, because solubility is the concentration at which solid and solution are in balance, and a lower-energy solid balances against less dissolved material. A tablet designed around a soluble form and delivered as a less soluble one may fail to dissolve fast enough to be absorbed.

So why was the factory ever getting the metastable form? Because thermodynamics names the destination and says nothing about the route. Crystallising requires nucleating — assembling a first ordered cluster large enough to keep growing — and, as the supercooling case in this collection works out in detail, that step has an energy barrier of its own, unrelated to which final phase is more stable. Different packings face different barriers. Nothing says the deepest well is the easiest to fall into.

Ostwald noticed the pattern: a system leaving an unstable state tends to reach not the most stable state available but the nearest one. Treat that as a widely observed tendency rather than a law — it has plenty of exceptions — but it is enough to explain a factory. If the metastable packing nucleates readily and the stable one almost never does, then for as long as the stable one fails to appear anywhere, every batch gives the metastable form, reliably, for years. The process was not robust. It was protected.

Now ask what removes that protection, permanently. A single crystal of the stable form. Seeds bypass nucleation entirely: growth on an existing surface has no barrier to clear, so once a stable-form crystal exists anywhere in a slurry, it grows while the metastable crystals dissolve to feed it, and the batch converts. And crystals are not tidy — they ride on dust, glassware, gloves, a person walking between rooms. Once the stable form has been made once in a building, the building contains seeds, and no batch mixed in it is ever unseeded again.

That, and not any change in chemistry, is what "disappearing" means. Ritonavir is the case everyone cites: an antiretroviral launched in 1996 whose formulation depended on the only form then known, until in 1998 a second, much less soluble form appeared, spread through the facilities, and made the original effectively unobtainable there — the capsule product had to be withdrawn and reformulated. Dunitz and Bernstein, reviewing several such histories, made the point that matters: nothing vanished from nature. The old form remained perfectly real, recoverable where the new seeds had never reached.

Separate the two halves cleanly, because almost every confusion here is that mix-up. Thermodynamics decides which form wins eventually, and it had already decided — the stable form was more stable in 1996 too, silently, unmade. Kinetics decides which form you actually get, and kinetics is the only thing that changed. Nobody discovered a new compound; somebody removed a barrier.

The account is testable in both directions. If seeding is the mechanism, adding a trace of the stable form to a crystallisation that would otherwise give the metastable one must convert it — it does, dependably enough to be a standard manufacturing tool. And material made in a facility the new form has never entered must still give the old form — it did. What would refute it? If a rigorously seed-free laboratory, using verified starting material, also could not obtain the old form, then something in the substance or process had genuinely changed and seeding is not the explanation. And had the newly appearing form turned out to be more soluble than the old, the thermodynamic direction would be backwards and the whole story wrong.

The most common misreading is that the stable form somehow "became" more stable, or that feedstock impurity caused it. Neither. The stable form was always the endpoint; what a seed supplies is not stability but a surface — a kinetic gift, not a thermodynamic one.

One genuine complication. Two forms may be monotropic — one more stable at every temperature up to melting — or enantiotropic, swapping order above some transition point. An enantiotropic pair can be steered by choosing the crystallisation temperature; a monotropic pair cannot, and its metastable form is only ever held by kinetics.

And the honest limit of prediction: which packings a molecule can adopt is now computable with some success, but the free-energy differences involved are small and the ranking is often within the error of the calculation. Knowing a stable form exists on paper is not knowing whether it will ever nucleate, which is why the industrial answer is brute-force screening across hundreds of solvent and cooling conditions — trying to find the disappearing form before it finds you.

Every term the collection defines is gathered in the glossary.