Skip to main content

Analytical Chemistry

Cooling Crystals is Great. Except When It Isn’t.

If you’ve ever been around an X-ray crystallography setup, one of the constants is a tube directing a blast of chilly vapor at the crystal that’s mounted for analysis. It’s usually a stream of cold nitrogen gas, often set up as a blast of the cold stuff surrounded by a second concentric layer of dry room-temperature nitrogen, so that you don’t immediately start growing frost all over your sample. Cooling the crystal is nearly universal in the X-ray diffraction business, for several good reasons. It preserves the crystal from deterioration, for one thing, both from being exposed to ambient conditions and from being blasted by an X-ray beam. The X-ray beam will slowly demolish your crystal – or if you’re using a synchrotron source, they will quickly demolish it – and the key is to extract good reflection data before the damage gets too severe.

And even more importantly, cooling slows down the molecular/atomic motions as much as possible, giving you tighter data. It’s not like you can’t get a decent X-ray at room temperature (you certainly can) but keeping all the thermally wiggling parts (getting technical here) from wiggling quite so much can be a real help. Those drawings of your molecule that software will give you from X-ray data are “thermal ellipsoids” for each atom, and they look a lot better as peas than they do as watermelons or serving platters.

But the benefits of such cooling are not universal. Here’s a good example from the recent metal-organic framework literature, and it describes an effect that I’ve encountered a bit myself. (I’ve actually had occasion to synthesize some MOFs, and I can tell you that from a bench chemistry perspective that although it can be a frustrating field, making all those brilliant multicolored crystals is a great time). The thing about MOFs, though, is that they are, by definition, frameworks of organic ligands held together by linkages to metal atoms, and the spaces in between are wide open – and full of solvent.

That’s a problem, because that solvent is pretty much disordered, except in some special cases where you can see (for example) benzene molecules lined up in there, which sends you toward the very interesting and still-evolving world of getting small molecule guests in their for crystallographic studies themselves. But in all MOF crystal studies, you have to deal with the X-ray diffraction noise from all that disordered solvent. There are software tools that can help, but they have to be applied judiciously, because you can throw away or obscure useful information while you’re trying to deal with the slop. And crystallographers, a famously rigorous bunch, would rather not massage the data any more than absolutely necessary, and good for them. (A nice thing about a full crystallographic report is that you can see exactly what had to be done to get the structure, and decide if you agree with those decisions or not).

The new paper referenced above points out in detail what some crystallographers had been noticing empirically: some MOF crystals can actually look better at room temperature than they do under cryo conditions. The Yaghi group (very well-known in the field) reports some zirconium MOFs, of which they have prepared a good many over the years, with typically large unit cells and high porosity. Collecting reflections at 100K gave a notably worse data set than a room-temperature run, with weaker high-angle data and worse signal/noise overall (you really need those high-angle spots for good resolution). Ask a crystallographer which of those two data sets shown they’d be more enthusiastic about working with, and you’ll get a definite and possibly profane answer.

What seems to be going on is interaction between the solvent molecules and the framework. As you cool things down, these become more fixed and less fluid/exchangeable, and they deform the framework at random places throughout the crystal lattice. That pretty much destroys your chances of getting clean reflection data. Interestingly, taking a cooled crystal and warming it back to RT gives improved data, but not back to what it was initially – some of those solvent interactions apparently stick, which means that you in turn are stuck.

This is not going to be the case every time, because there are several factors at work simultaneously. The amount of solvent in the pores (and the basic layout of the MOF lattice) is one variable, and the degree of solvent interaction with the framework and metal atoms is another big one. That in turn will vary by what solvent you have in there, and the degree to which it coordinates with the framework (and with itself). But it’s definitely something to consider – the Yaghi paper shows this effect in several different MOFs, to varying degrees depending on framework and solvent. If you’ve made a new one and it looks crappy at 100K then, try fishing out another crystal and just mount the thing and collect at room temperature like it was 1952 again. You might be surprised!

22 comments on “Cooling Crystals is Great. Except When It Isn’t.”

  1. Anon says:

    JACS ? really??

    1. Derek Lowe says:

      You were expecting Acta Cryst.?

      1. OM says:

        Being an associate editor *absolutely* has nothing to do with it!

        1. Rhenium says:

          After having looked at the paper, yes this is really thin work. Pushing the boundaries of the MPU is JACS worthy I guess.

  2. me says:

    Question related to the author list of this Yaghi paper. One of the authors is Sultan A. Alshmimri, a faculty member King Abdulaziz City for Science and Technology, KACST (not to be confused with King Abdulaziz University).

    KACST (which appears to be something like a national lab) is also listed under the acknowledgements section as a source of funding.

    I’ve heard it’s not uncommon for high-profile US professors to be offered research money under the stipulation that there are Saudi coauthors on any resulting papers.

    That isn’t necessarily what is going on in this Yaghi paper (Alshmimri seems to have a legitimate PhD from UManchester, although his thesis is fairly bare-bones and I can’t find any publications other than an RSC advances article), but it does look suspiciously similar to cases I have heard about.

    What are the ethics of this? Personally, it leaves a bad taste in my mouth but I don’t see an enormous amount of harm being done in exchange for a bit of extra funding for legitimate science.

    1. BK says:

      Most crystallographers want coauthorship for their work instead of a thank you footnote/acknowledgement. What’s the difference between a person asking for coauthorship who has done some works versus a semi-private entity funding your research?

      1. me says:

        What’s the difference between those things? They are totally separate, I would think.

        Where are you seeing that Alshmimri is a crystallographer? His background seems to be catalyst design for polyethylene.

  3. TroyBoy says:

    Freezing crystals doesn’t really prevent radiation damage from the incidental beam. Freezing slows the process of free radicals recombining with other parts of the crystal and deteriorating crystal order and data that you collect.

    “Interestingly, taking a cooled crystal and warming it back to RT gives improved data, but not back to what it was initially – some of those solvent interactions apparently stick, which means that you in turn are stuck.”

    One interpretation is that after they cooled the crystal–collecting a data set and generated free radicals–and then warmed the crystal, which lets all those free radicals loose. So I’m not sure that the hypothesis that sticky solvent interactions at different temperatures explain the differences in diffraction.

    At least for protein crystallography, freezing crystals is still the way to go. Mounting crystals in humidity controlled capillary tubes, collecting at most 5 frames of data, and then having to merge multiple data sets together is a nightmare.

    1. Sofia says:

      I doubt radicals will last that long for your freeze-thaw-ROS theory. The timescales between temp changes and radical interactions are likely orders of magnitude different. So I’m less inclined to believe there are ‘temperature hidden’ radicals than there are just general damage and retained solvent ordering which result in lower grade diffraction data.

  4. sdfds says:

    A nice paper on radiation damage in macro molecular crystals that is relevant here:

    I can’t read past the paywall, but these conclusions don’t appear to be surprising at first look. We’ve long known that freezing increases diffraction pattern mosaicity significantly (presumably due to freezing induced disorder) in biomacromolecular crystals; that would be my first explanation for the reduced intensities of higher resolution reflections (assuming all else is equal). I’m not quite sure how you could eliminate the effect of “tertiary damage”.

    It doesn’t seem (from the abstract) that they did the types of experiments that have been done to work out radiation damage in biomacromolecular crystals. I wonder why “disordered guest–framework interactions” are really different than what we see in biomacromolecular crystals with (generally) high solvent content? I also don’t see what the tradeoff in damage/data quality is in the abstract either (what is the radiation dose tolerance you buy with the lower temp.).

    1. anon says:

      Radiation damage is not a problem for these Zr MOF crystals. They diffract like absolute bonkers compared with the best protein crystal ever made. However, the spatial resolution you’re looking for is much higher, so the mosaicity becomes a more serious problem.

      1. anotheranon says:

        It’s all ’bout the Z, Z, Z

      2. sdfds says:

        If radiation damage isn’t an issue why do they cryo cool them?

  5. another crystallographer says:

    One the the key points that is that this is a phenomenon observed in MOFs with solvent/guest-filled channels – rather than in those with evacuated (or mostly evacuated) channels – which is generally where you tend to start when you have made a potentially new MOF (unless it collapses when you remove solvent). They show some of their results for the evacuated systems in the ESI, where the differences are minor enough that they could be due to radiation damage (or related effects) over multiple data-collections. An interesting comparison that they don’t appear to have done, would have been to collect data on a fresh crystal (either solvated or evacuated) at low temperature, then take it to the higher temperature being used and collect again, before returning to low temperature. This would have given at least some idea of how much of the difference in their temperature series is due to temperature-induced framework/guest effects, and how much due to radiation damage and related effects.

    1. Anon says:

      Yeah this is actually the first thing they should have done. I’m surprised they got away with it.

  6. Scott says:

    “Those drawings of your molecule that software will give you from X-ray data are “thermal ellipsoids” for each atom, and they look a lot better as peas than they do as watermelons or serving platters.”

    You need to live in a place with better watermelons, Derek. Out west, you can get watermelons in the 35-50lb range from some farmers, and 25-35lbs regularly in the store. A 35lb watermelon is about the size of most serving platters I’ve seen!

    1. Derek Lowe says:

      Hah! I was referencing the usual Tic-Tac shaped ones and the others where motion is constrained, and you get a much flatter ellipsoid. We had good ones down South, too, but I have to say that the ones in the Central Valley are pretty mighty.

  7. Barry says:

    Is anyone else here bothered that the bound solvents are said to be “random”? That’s not how host-guest crystals should work. Is the problem not that the crystals are cooled, but that perhaps they’re cooled too abruptly and should have been annealed?

    1. Derek Lowe says:

      I think the idea is that they’re distributed at more or less random metal atoms throughout the lattice, not in an orderly fashion (as with the small-molecule soaking MOF crystallography work)

      1. Barry says:

        Maybe I’m too fundamentalist on what “random” means. Say “sub-stoichiometric complex” or “incomplete occupancy” and I’m happier. But either would make a flimsy paper sound even more hollow.

  8. Lloyd T J Evans says:

    I can report that solvent interference effects are not just a problem for MOFs, but can be a problem with some smallish molecule organometallic compounds too. I worked with several lithium and potassium amide salt ligands between 2004 and 2008, some of which formed the best crystals when slow-chilled from toluene solution. As is common for this class of organometallic, toluene molecules would integrate themselves into the crystals, with the aromatic ring co-ordinated to the alkali metal ions.

    This didn’t present too much of a problem for the alkali metal salts, since the toluene molecules were fairly well glued on, particularly for the potassium variants. But there was one particular ligand (think KHMDS but with one of the methyl groups on each silicon replaced by an anisole ring, joined at the ortho position) that formed an interesting co-ordination compound by salt metathesis with zirconium tetrachloride, sticking a pair of amides directly to the zirconium.

    The resulting zirconium compound would also crystallise from toluene, and after much frustration, the departmental X-ray crystallographer (the talented prof. Peter Hitchcock) got a single crystal diffraction pattern from it. The frustration was a combination of two factors: First, the sheer size of the unit cell, which had 4 differently orientated molecules of the zirconium complex and 15 molecules of toluene. Second, the toluene molecules were rather reluctant to stay in the unit cell. They stayed in place once the single crystal was mounted in the cold nitrogen stream, but getting it there before the crystal disintegrated from efflorescence was a real challenge.

  9. anon the II says:

    So, I’ve been thinking about this for a few days, but not very intensely. Is this really any different from any other coalescence phenomena like you see in NMR or chromatography where you have two or more different molecular entities interconverting on the same time scale as the experimental sampling? Like DMF going from two peaks to a broad mess to one peak on heating. I’ve also seen it happen in a GC. I didn’t read the article but if this was done on a garden variety in-house diffractometer, then data collection takes a little time (seconds). Maybe on a synchrotron, where the data is collected and the crystal is toast in milliseconds, you’d get different data at low temperature. Maybe you’d separate out that disorder if you could account for it properly. Just a thought.

Comments are closed.