Let’s talk crystals for a few minutes. Those of us (like me) who are familiar with chemistry and biology, but who are not crystallographers themselves, will know the broad outlines of X-ray crystallography, and can appreciate its extension to diffracting electrons instead of X-rays. But there are a lot of odd and subtle things that creep into the topic of crystal structure when you look closely, and the better these techniques get, the closer we’re able to look. Several new papers illustrate these, in various ways.
For one thing, how do crystals form, anyway? At some point several molecules (or individual atoms, if it’s that sort of crystal) have to come together in the orderly arrangement that will end up repeating throughout the whole bulk crystal, and start attracting other members from nearby to join the new gathering. This commentary in Nature has some illustrations of the standard model by which this has been thought to happen (at least in many cases): “classical nucleation theory”, which is pretty much what I just described. That is, a very tiny replica of the bulk crystal forms, maybe only a few molecules wide but having the same intrinsic lineup as will repeat in the bulk crystal, and molecules come in from the surrounding solution and arrange themselves on the newly developing surfaces in turn. And that actually is how many crystals form (although it’s already been obvious for some time that deviations from CNT happen a lot under real-world conditions).
This new paper has some painstaking observations in a metal alloy system (iron and platinum) at an unprecedented level of detail, achieved not via diffraction methods, but via atomic electron tomography. Electron tomography (inferring the shape of very small particles via the scattering of an electron beam) is how cryo-EM protein structures are produced, and with the powerful scattering of the heavier elements you can get such data all the way down to particles only a few atoms wide. This new work watches as disordered nanoparticles of Fe/Pt take on the tetragonal crystalline lattice known for this alloy in the bulk phase, and they’re able to follow both the shapes of the individual atom clusters and their evolution over time. The result is. . .well, it’s pretty complicated down there. The early stage nuclei have all sorts of different irregular shapes, for one thing. The very center of these things have the closest fit to the tetragonal structure, but they’re much more fuzzy and variable just outside that – it’s definitely not the orderly stack-up of classical theory. These nuclei vary over a pretty wide size distribution as well, and they’re neither spherical nor tetragonal themselves, which you’d have expected to be the low-energy shapes. These alloy crystals all end up in the same place (a nice tetragonal lattice), but they sure get there through all sorts of pathways.
So there’s one. A couple of other recent papers take on crystal structures of what are actually rather small and simple molecules, but whose crystal forms have been giving people fits for years. One of this is orthocetamol, a structural isomer of the familiar drug paracetamol (known as acetaminophen in the US, Canada, and Japan). This ortho-substituted isomer has been known since 1876 (discovered even before X-rays were!), and it forms what appear to be nice square-shaped crystals (see at right). But experienced small-molecule crystal types would already have been giving that first photo the fish-eye, because those things, while most certainly crystalline, don’t quite look like they’re full of perfect right angles. Sure enough, as you zoom in, what you find is that they’re actually conglomerates of far smaller crystals, and these individual pieces are too small for conventional X-ray crystallography to deal with. If you try to shoot the larger bunches, you get data that can’t be refined to anything useful.
Enter micro-electron diffraction. That’s rapidly turning into the default answer when someone says “My crystals are too small”, and so it proves here, providing what appears to be the first micro-ED-based structure of a small molecule that has been resistant to X-ray techniques. Gently crushing some of those crystals and scattering the pieces across an EM grid allows for data collection on a whole range of pieces, and many of them are tractable (although the authors note, alarmingly, that even pieces less than a micrometer wide showed evidence of inhomogeneity in the diffraction patterns). The data looked sort of like a tetragonal unit cell (which is what you’d have guessed from square crystals), but never really settled into a good one, thanks to plenty of stacking disorder and twinning even on the nanoscale. Dropping down to monoclinic symmetry (a classic move in such situations) allowed for a direct-methods solution, which that was backed up strongly by X-ray powder diffraction and Rietveld refinement – powder diffraction data are very useful for such microcrystalline substances, as you’d imagine.
What you get are infinite ribbons of nearly flat orthocetamol molecules, hydrogen-bonded between the acid of one and the phenol of another. There are a lot of ways that the arrangement can go subtly off – there’s twinning on the scale of about 150nM, for one thing, and the flatness of the molecules (and lack of the some of the intramolecular interactions seen in the other isomers) give them a chance slide around a bit while still maintaining a solid crystalline form. It’s remarkable that there are so many defects and variations in something that is, at first glance, a perfectly reasonable crystalline solid.
And here’s another paper on the crystal structure of a common drug, hydrochlorothiazide. It has more than one crystal form, and one of them look like the photo at left – a sort of two-winged symmetric form. Separation of these “wings” and individual X-ray analysis shows that (weirdly) they are each made up of a single pure enantiomer, and one of the weird things is that you have to stare at hydrochlorothiazide’s structure for a while before you can even see how it could be chiral at all. The sultam nitrogen is nonplanar, and its ring can adopt two different conformations. That and a bond-rotation of the sulfonamide group give you two mirror-image structures, and there they are as two wings of a crystalline butterfly. In solution, the two enantiomers seem to interconvert pretty rapidly, though, so you always have a racemic mixture when the compound is dissolved. It’s remarkable that no one has ever realized this before.
Having a racemic mixture crystallize out that way is pretty odd, too. Most of the time you get racemic crystals, 1:1 mixtures on a molecular scale of each mirror-image enantiomer. More rarely, you get separate crystals of each enantiomer (as happened, famously, with tartaric acid when Pasteur investigated it and performed the first chiral resolution). But seeing the two enantiomeric crystals form along a common surface like this – that’s weird. There seems to be only one other example known, hydrobenzoin. The authors believe that this happens as the crystals try to deal with the electric dipole field that forms as things start to line up – having the two “wings” like this allows the dipole to cancel out. It’s thought that some crystal formation routes in general actually end up being unstable due to “uncancellable” dipole charges building up.
But then, there are a lot of ways to build a crystal, and if you’d like to explore those, this recent JACS article will send you to the appropriate literature. There’s a lot to the ordered solid state of matter, a lot that we don’t understand, and a lot to explore. . .