Skip to main content

Analytical Chemistry

AFM Marches On

We must be getting close to the future that people have been foreseeing, because here’s a whole review on the topic of using atomic force microscopy to elucidate structures. It’s from the IBM Zürich group that has pioneered so much work in this area. Over just the time I’ve been writing this blog, that idea has gone from “Maybe someday” to “In some well-defined test cases” to “A few real-world examples known”, and now perhaps to “Hey, this might actually start working”.

Actually, it’s not just AFM. There’s also scanning tunneling microscopy (STM) and Kelvin probe force microscopy (KPFM), among other even more exotic ideas, and together you can get information that simply cannot be obtained by any other methods. Here they are in that order:

AFM is done typically by picking up a single CO molecule (from the carbon end) with the tip of the metal probe and scanning with the oxygen end across an adsorbed molecule (at liquid helium temperatures) at a constant height, about 4 Angstroms over the molecule itself. Often you go back and scan at several different tip heights. There are many other possible tips – xenon atoms, other diatomic molecules, halogen atoms, and so on. The cantilever holding the tip is oscillating (over about half an Angstrom) at a high frequency, and changes in this, induced by interactions with the molecule under study, are what get measured out (along with deflections in the angle that the CO molecule makes with the surface).  The mechanism for these changes is just flat-out Pauli repulsion for the most part – electrons not wanting to occupy the same volume of space and getting out of each other’s way – although van der Waals and electrostatic forces can show up as well.

KPFM is pretty similar, except the readout is the voltage dependence of the position of the tip relative to the molecule, rather than changes in frequency of oscillation. You pick a voltage that gives you good signal/noise and scan around as with AFM. In practice, that gives you more of a pure electrostatic read on the surface, and it’s very useful for determining charge states and electron density in the sample. For example, fluorine is very close to the size of hydrogen, but they look dramatically different in KPFM because of the mass of electrons around the F atoms. Interpretation of the data, though, is not very straightforward, and often seems to involve matching up the experimental results to various calculations to try to see what’s going on.

And finally, in STM you apply a voltage to the tip and measure the changes in electric current between the tip and the sample as you scan, often by having the tip height change to keep a constant current level. At these distances, electrons are tunneling right across the gap, so this is sheer quantum mechanics all the way. If you pick the appropriate voltage, electrons can be temporarily attached to the lowest unoccupied molecular orbital (LUMO) of the test molecule (or detached from the highest-occupied, the HOMO, if you flip the voltage around), and the resulting images are absolutely terrifying for someone like me who came up thinking of these as mathematical abstractions.

At right is an example of these, with a CO tip used to show the bonds in the graphene-like molecule and then (in STM mode), its LUMO. As you can see, the resolution on these experiments is getting pretty good, and allows for an unequivocal assignment of the structure. Of course, an X-ray structure determination would almost certainly nail that down, too, but for that you have to have a purified sample and a good crystal. One interesting thing about AFM-type techniques is that you can scatter a mixed sample across the surface and take a look at the various things that might be in it, one molecule at a time.

At right is an illustration of something else that can really only be done under AFM conditions: actual atomic-level manipulation of the sample. Individual atoms can be pushed, twanged, or (as in this case) removed entirely. The result can be highly reactive intermediates (radicals, arynes, etc.) that can be studied directly at the liquid-helium temperatures involved. There are also opportunities to do on-surface reaction by heating followed by cooling down to AFM conditions, through photochemistry, etc. You can also study the interaction of individual molecules with each other

Now, it has to be said that there are not very many labs that are capable of these things at the moment, but that’s changing with the advent of more commercial apparatus and the spread of experienced workers in the field. What’s lacking? As the authors note, there’s a real need for techniques that might be able to read out functional groups and elements more directly. An even bigger challenge is that all the most spectacular results to date have been on planar or near-planar molecules, since these yield the most (and the most interpretable) data. You’re running your quantum fingers over the surface of things here, so the inside of a folded-up conformation is just not going to be available. And finally, there need to be improvements in the time it takes to get a measurement. It’s been stuck at roughly ten-minutes-per-molecule for a long time, but better time resolution would allow for a lot of new things to be picked up (and for better characterization of mixtures, and so on).

Update: here’s a Nature News article on the current developments in the field.

14 comments on “AFM Marches On”

  1. Barry says:

    “all the most spectacular results to date have been on planar or near-planar molecules, since these yield the most (and the most interpretable) data. You’re running your quantum fingers over the surface of things here, so the inside of a folded-up conformation is just not going to be available.”

    Some of us wonder if a 3-D protein sitting on a (copper?) plane will adopt a conformation that accommodates the substrate, potentially different from the conformation in solution. I’d expect that few proteins will retain biologically-relevant conformations if all the structural waters are pumped off?

  2. a says:

    I want to see someone give these guys a substrate where they DON’T know the structure on the surface (or told it’s one of 10 “closely related structures”) and they get the right answer.

    I may be cynical, but these pix look wwaaaay too good-cartoony to me

    1. Imaging guy says:

      I completely agree with “a”. Super-resolution microscopy and electron tomography also face the same problem when you have structures already determined from x ray crystallography or electron microscopy. You must prove your method with specimens blinded to the operator.

    2. Dumbledore says:

      These measurements are already done. An example would be where researchers have looked at completely unknown structures and form a database, i.e. asphaltene molecules. As far as I know, there are many groups finishing up works where they tackle different unknown compounds. So this is not far fetched.

  3. Tim says:

    I’m sorry to mention a trivial formatting thing, but links on these pages tend to have a strike-through. In the first paragraph, “here’s a whole review” is struck through but “test cases” is not. Looking at the source of the page, the first has class=”broken_link” on it. It is not a broken link at the moment. It’s trivial, because it works and I can read it, but it’s a bit odd and distracting, so if there were an easy fix, that would be great.

    1. Mark Thorson says:

      This is an objection which has been noted several times before. Don’t expect it to be fixed.

      1. Derek Lowe says:

        Looks like it may have been fixed – let me know if it shows up again, though.

    2. Derek Lowe says:

      Still not sure why this happens! I will make another run at fixing it, though. There seems to be no reason for it to come out like that, what with it being a valid link and all. . .I’ve just pestered the behind-the-scenes folks at Science about this.

      1. Some idiot says:

        Interesting… the links were ok when you first sent the blog post out (ok, within about 15 minutes of the AFP blog post arriving on the site, at any rate…)

        1. Some idiot says:

          I forgot to add that the text was not “strike through” at the time either (just normal “link text”).

  4. Chris Phoenix says:

    Also check out LEAP / APT. On paper, it looks really good: You take a tiny speck of material, and use high voltage and/or laser to make one atom at a time jump off. The electric field makes it fly away predictably, so you can tell where it was with near-atomic precision (~0.3 nm) – and the time of flight tells you what atom it was. (I have no connection to them; just found them with a web search.)

    Practical and theoretical difficulties to analyzing biological samples are still being overcome (e.g. the sample has to be a special tiny shape, so you have to transfer it from FIB to LEAP in cryo UHV conditions).

    It wouldn’t surprise me if this became quite useful for biomolecules-in-solution in the next 10-20 years.

  5. Piotr Nowak says:

    HOMO and LUMO (or any other molecular orbital) are indeed mathematical constructs. They are not observables; one can pick a different set of molecular orbitals to achieve the same electronic properties, so how can one observe them?

    The phenomenon observed here was explained by Pham and Gordon here:

    “The origin of the misinterpretation of STM images as orbitals can be understood as follows. When a bias voltage is applied to a tip positioned very close to a material surface, electron tunneling can occur. The variation of the tunneling current can be visualized as the STM image. The tunneling current at a particular space point is proportional to the density of states (DOS) of the material at that point, also called the local DOS. The DOS at a particular energy level results from the coupling of the electronic state at that energy level to perturbation sources (e.g., electronic states of adjacent molecules, external fields). The DOS at an energy level of an unperturbed state (e.g., an isolated molecule) is small and proportional to the degeneracy order at that energy level. In the STM technique, a local potential applied to the tip serves as the perturbation; therefore, the DOS is nonzero. In many molecular systems, a high DOS occurs at about the Fermi level of energy, which is usually between the first ionization and electron affinity energies. These energy levels can be roughly approximated by the HOMO and LUMO energies. Unfortunately, several authors have made the incorrect leap to conclude that they are actually observing the HOMO and LUMO themselves.”

    1. Some idiot says:

      Interesting… thanks for the clarification!!! 🙂

      1. Barry says:


Comments are closed.