Skip to main content

Analytical Chemistry

Covalent Organic Frameworks

Let us pause to consider the weirdness of diamond. Not because diamonds are rare – they’re not, at least compared to many other minerals and gemstones. But diamond itself has very unusual physical properties, and that comes down to its structure. As is well known to chemists, it’s a three-dimensional lattice of bonded sp3 carbons, which sort of stretches ones idea of what a molecule is. Hold that thought, though: can you name another crystalline organic substance that’s assembled from a framework of covalent bonds?

I’ll bet not. In fact, until very recently there have been no examples at all of any single crystals of a substance like that (other than diamond itself). Update: I missed the azodioxy-linked ones reported by the Wuest group at Montréal!) You can get terrific crystals of metal-organic frameworks, where organic spacer groups are held together at the junction points by coordination to metal ions – I was able to spend some time a couple of years ago growing such things, and it’s a blast. But a crystal of a sheer covalent-bond-framework material, that’s something else again. People have made such substances, but they almost always come out as stuff that can only be characterized by powder X-ray diffraction at best, and amorphous gorp at worst.

The problem is assembling the framework. If you picture the process down on the molecular level, you have to have each unit come in, find its proper place, form bonds between the appropriate neighbors, and wait for something to come bond with it in turn. If you have a unit come in and dangle off the surface a bit, then the next building block that comes in might bond with it instead of with the growing surface, and now both of those have to come down together into the right place. And what’s surely going on, even more than that, is the start of untold millions of tiny frameworks all over the bulk solution, nucleation sites everywhere, with zero chance of them every fitting together into something larger in any kind of ordered fashion.

There are only a few ways out of this situation, as far as anyone knows. One is the diamond-forming process, where the final product is a thermodynamic sink. You can just take less orderly forms of carbon and heat and squeeze them under extreme conditions until the bonds rearrange into the lowest-energy structure. It helps if you let geology do the work for you. Diamonds are believed to be distributed near the surface by kimberlite eruptions, most of which seem to have taken place a very long time ago (and a good thing, too), and the diamonds themselves – along with many other other interesting igneous minerals – are apparently formed at great depths in the Earth’s mantle. This is not going to be a general process for covalent organic frameworks.

The other main way you can imagine making them would be if you could have some sort of reversible bond formation involved. That way, you could in theory adjust the rate of crystal growth by changing the conditions and thus let things come along in a more ordered way. That’s the secret to growing good crystals of anything, actually – slow and careful, with plenty of chances for mistakes to be corrected down at the surface. And now it’s been done with covalent organic materials.

Here’s a new paper from a large multicenter team (involving the Yaghi group at Berkeley who are very experienced in the related metal-organic framework field). They take the tetra-amine building block at right and the dialdehyde to form the large open framework shown, held together by imine bonds, but if you just mix those two together, you’re going to get a terrible mess. The key is to add aniline (about 15 equivalents) during the crystal-growing process. This forms a reversible imine with the free aldehyde groups, and slows the crystal formation down enough to allow things to settle into a reasonable form. They have very nice resolution of crystals of the parent and its hydrate, with water molecules filling the empty spaces. You’d look at that structure and say “That’s a lot of room”, but that’s a schematic: the real structure is interlaced and interpenetrated (sevenfold, actually), so there’s not a lot of empty space in the crystal itself. The same thing happens very often with metal-organic frameworks; it can be tricky to grow crystals with huge open spaces in them, although that depends completely on the geometries involved. Interestingly, if you look at this structure, it’s basically a huge diamondoid lattice: what you’re looking at is a whopper adamantane, the repeating diamond unit, since the branch points are also tetrahedral. (Actual diamond doesn’t have the room to have an interpenetrated lattice). Update: the nitroso/azodioxy framework reported earlier are also from a tetraphenylmethane scaffold, and form a similar diamondoid lattice, fourfold interpenetrated.

If you switch the functional groups, with a tetrabenzaldehyde-methane and the 1,4-diamine, you get crystals as well under the aniline conditions. In fact, they are almost identical in every way – the only way you can tell one from the other, apparently, is through single-crystal X-ray. The team went on to form other covalent organic framework crystals with different building blocks, so the slow-it-down-with-aniline conditions seem to be reasonably general for imine frameworks. No doubt you could mess around with more or less reactive anilines or other amines to refine the conditions for a wide range of these things. Crystal growing, as has been mentioned around here before, is a brutally empirical art.



And simultaneously, here’s a paper from Northwestern/Cornell/Argonne showing macrocrystalline two-dimensional covalent organic frameworks for the first time. You get things like this in the metal-organic framework field as well. They can look like three-dimensional crystals, but if they’re put into a slightly stronger solvent than the one used to form them or given any mechanical stress, they delaminate into vast numbers of sheets. Watching a clear crystal turn into a pile of what looks like Danish pastry or a spilled-over deck of cards under the microscope is an odd experience, I can report. The strong bonds are in two dimensions, and the sheets are able to just slide over each other: instead of diamond, the exact carbon equivalent is graphite, with sheets of graphene. And as with the three-dimensional cases, graphite has been the only organic example known that’s not an amorphous/microcrystalline material.

These new materials remain as individual two-dimensional sheets, though, at least in some cases (other tend to stack up on each other when imaged on a surface). The team prepares them by forming boronate esters, and adjusting things by very slow addition of the building blocks involved. The first reaction goes through a large number of nucleation sites, as mentioned earlier, and if you scale that up or keep throwing in monomers you just get more tiny particles. But if you take that initial solution and switch to slow addition, you can grow the existing nano-sized seed crystals (which are about 30 nm wide) rather than forming new ones. This took the average size up to about 400 nm, and further tweaking of the conditions on these allowed even larger particles to be produced. These are now large enough to be well-defined by electron microscopy images, and optical transient absorption experiments showed them to be far better behaved than any such materials ever tried before, with a thousandfold increase in signal/noise.

So both of these papers report completely new materials, which will have properties that remain to be explored. As with the metal-organic frameworks and the two-dimensional inorganic materials, the structural, optical, and electronic behavior of these things is an open book, and getting a handle on these things could lead to exotic materials that do things that no one has ever been able to accomplish.

26 comments on “Covalent Organic Frameworks”

  1. Haftime says:

    Worth flagging single crystals of COFs were first reported five years ago, by James Wuest and co-workers. Yaghi’s route to larger crystals of imine based COFs is nice though.

    1. Derek Lowe says:

      I did not know that! Updating the post now – thanks.

  2. Project Osprey says:

    “Can you name another crystalline organic substance that’s assembled from a framework of covalent bonds?”


    1. Derek Lowe says:

      Hmm. Not sure if a 1-D chain material should qualify. I should have been more picky in my definition! Plus, have you ever seen single-crystal X-ray work on celluose? (I haven’t, but I haven’t looked that hard, admittedly)

      1. Anon says:

        I’m going to put it out there – diamond is not an organic substance.

    2. fourtytwo says:

      Silicon carbide?

  3. Anon says:

    “Can you name another crystalline organic substance that’s assembled from a framework of covalent bonds?”


    1. Derek Lowe says:

      I don’t quite count the metal coordination as a covalent bond, though.

  4. Chris Phoenix says:

    This reminds me of R8Si8O12, polyoctrahedral silsesquioxane, and the fun polymers that can be made with it. My impression is that POSS crystals are a lot easier (relatively) to make than the crystals described here, and I wonder if there’s inspiration to be had there.

    1. David says:

      Crystals of discrete POSS compounds are comparatively simple to prepare (depending on your substituents), but crystals of anything polymeric are significantly more challenging, given the potential for unordered networks!

  5. Uncle Al says:

    2D sheets are potentially wonderful – thin selective separation membranes (neat or within a matrix), diffusion barriers within multi-ply polymers, optical coatings and effects. Incorporate unpaired spins, chirality. Langmuir-Blodgett built circuitry (at least capacitors) and metamaterials. Biocompatiblity interface buffers. Superconductivity is low-dimensional.

    “There’s Plenty of Room at the Bottom.” Richard Feynman.

  6. KN says:

    Can we consider graphene a 2D crystal?

  7. Jon says:

    What is on the surface of diamonds? at some point the C-C bonds stop. What replaces them on the surface, is it a layer of OH’s?

    1. Derek Lowe says:

      You know, I’ve wondered the same thing. Anyone have any empirical evidence? Does diamond terminate in C-H groups, C-OH, or what?

      1. Barry says:

        but what happens if I cleave a diamond (homolyse all those C-C bonds) submerged in tetrafluoroethylene? Do I get graft-copolymerized teflon/hydrophobic surface?

  8. Anonyman says:

    There is a third option aside from “thermodynamic sink” crystallization or dynamic covalent chemistry: Covalent assembly in the crystal, usually by photochemistry: You form a layered single crystal of at least trifunctional photoreactive monomers and then irradiate, hopefully in a single-crystal to single-crystal transformation, so you get out an ordered material. For 2D-systems, this had been done using anthracenes (10.1038/nchem.2007, 10.1038/nchem.2008) and stilbene derivatives (10.1021/jacs.6b11857) as polymerizable units.

    Disclaimer: I’m peripherally associated with some of the cited work

  9. fajensen says:

    I think there is another way than “brute force on a tectonic time-scale” to make diamonds, by chemical vapour deposition (CVD). They have been at it since the 1950’s at least and now it seems that there is a working industrial process for it.

    One can now buy oneself fairly large wafers of pure diamond commercially from a few manufacturers. The wafers are often for x-ray transparent vacuum windows, optical elements for lasers or heat-spreaders for really hot-running integrated circuits of the classified kind.

    The way I understand CVD is, they set up a suitable diamond-like substrate (perhaps even diamond) in a vacuum oven and then one can, with a lot of skill, effort and process control, slowly feed an organic gas into the process, which breaks down at the substrate surface, deposit it’s C content on it and those C atoms arranges themselves Just Right to slowly form a solid disk of diamond.

    1. Barry says:

      I’m gone speculate that in the CVD route to diamond, initial C-C bond formation is reversible and there are a bunch of C-H bonds still intact, only subsequently annealing into the all-C lattice, rather than going over the high hump of free zero-valent mono-atomic carbon

  10. Lloyd T J Evans says:

    Doesn’t beta-carbon nitride count as an organic framework? I know it’s hard to grow large crystals of that, but the bonds within the crystals are all covalent. The closest direct analogue of diamond though is boron nitride, which has an almost identical structure. But that probably doesn’t fall under the umbrella of organic, since it contains no carbon.

    I do however remember a presentation I saw around 15 years ago from a group trying to use boron nitride as a hydrogen storage material. The idea being to progressively hydrogenate it, first forming a (HN=BH)n polymer, then a (H2N-BH2)n polymer and finally NH3BH3 monomer. These steps could all be reversed to get the hydrogen back out, albeit at different temperature and pressure conditions.

  11. Anon says:

    I remember trying to make a flavin analog during my undergrad project, and got out really nice crystals that wouldn’t dissolve to get any mass spec signal. So instead I did some X-ray diffraction studies, and found that the triple ring did not close properly, but instead formed a completely covalent 3D lattice similar to the one shown. I didn’t get full marks for that project because I didn’t get the final product. 🙁

  12. Anonymous says:

    Veering off from the diamond lattices just a little bit … Previously posted to Pipeline (Woodward 2.0, 16-April-2017): “RBW’s Idea Notebooks … were recovered from his office, after his death. Pages and pages of diamond lattices and diamond lattice defects and surface defects.”

    People have made buckyball dimers (think 2+2 cycloaddition between two C60s or double cyclopropanation or other fusion). Has anyone extended it further to form (C60)n = crystalline? I don’t remember the paper or method, but I suspect that organickers went after the simple solubles and tossed out the high MW, hard to characterize, possibly crystalline insolubles.

    (Come to think of it, how many organickers mixed together formaldehyde and phenol and threw out the insoluble crud before Leo Baekeland made something useful out of it — lots of money! — in 1907?)

    There is also Q-carbon (wikipedia) made by laser pulse – quenching of carbon. It is non-crystalline. It is harder than diamond. It is conductive. It is ferromagnetic. It is weird. Also see the linked page for comments. (I am trying to add a link to my posting name. If it doesn’t work, google Q-carbon, discovermagazine, 2015)

    One of Pipeline’s newer contributors has been thinking about diamond and synthetic diamond for a long time and used to post ideas to sci.chem. He had proposed a solution phase approach to diamond synthesis.

    Regarding crystal growth and aggregation, there are models other than diffusion limited aggregation (DLA). I was working on a project that needed to know about materials aggregation. At one meeting, research colleagues explained that they had assumed DLA. I interjected that most of the literature I had read about these materials rejected DLA and argued cluster-cluster or other mechanisms of aggregation prevailing. Minor disagreement, except for one more thing: they weren’t even measuring the actual aggregation! They were making single measurements of different materials and assuming that they were all aggregating the same way. Not likely. But the PI published anyway.

  13. Barry says:

    ” A mixture of H- and O-termination is the typical surface chemistry for
    naturally occurring diamonds 3 due to the presence of water vapour (H2O) in the Earth’s atmosphere,
    and chemical vapour deposition (CVD) diamonds are typically H-terminated (unless the
    gas precursor is specifically set-up to favour O-termination). 4,5 H-termination is vital for stabilising
    the surface structure and preventing spontaneous reconstruction or unwanted phase changes, which
    could also affect the bulk structure. The earliest method of H-terminating diamond is described
    by Emanuel in 1873 (although without his knowing), and involves the mechanical polishing of
    diamond surfaces using olive oil. 6 Bandis et al. demonstrated the H-terminated nature of polished
    diamond in their 1995 paper using low energy electron diffraction (LEED), and concluded that
    olive oil does in fact act as a hydrogen source for the diamond surface during polishing.”

  14. Senor Chemist says:

    Oh, big ben is at it again with his sarin gas adducts of proteins ( amd god knows what else ). I hope you machissmos make a lot if money.

  15. NoniMausa says:

    *reads about kimberlite eruptions*

    Nope nope nope, let’s not do that.

  16. Scott says:

    “Diamonds are believed to be distributed near the surface by kimberlite eruptions, most of which seem to have taken place a very long time ago (and a good thing, too)”

    Probable eruption characteristics like Mt St Helens, but with little to no warning?!? Yeah, let’s NOT.

Comments are closed.