Here, Martin Zanni explains two dimensional (2D) infrared spectroscopy, the role of lasers, and current applications, such as studying the kinetics of protein aggregation in diabetes.
Martin Zanni and his group in the Department of Chemistry at the University of Wisconsin-Madison (Madison, Wisconsin) are specialists in a new class of infrared spectroscopy: two-dimensional (2D) infrared spectroscopy. To collect a 2D IR spectrum, one uses a series of infrared femtosecond laser pulses to pump and then probe the response of the system. Using this technique, it is possible to probe the structures and dynamics of molecules. In this interview, Zanni explains the technique and how it is enabled by specialized laser methods. He also discusses current applications of the technique, such as solar cell research and the study of the kinetics of protein aggregation in type 2 diabetes.
What is 2D IR spectroscopy?
Zanni: Everyone knows that infrared spectroscopy is one of the most used analytical and research techniques in the world. It is often the first tool used to assess the chemical composition of a substance or the success of a reaction, because functional groups have characteristic vibrational frequencies. Students learn to interpret Fourier transform infrared (FT-IR) spectra in their undergraduate organic chemistry class. While frequencies are useful, vibrational motions of molecules contain an enormous wealth of information that goes far beyond what can be measured in an FT-IR spectrum. Vibrational coupling tells us if two molecules are bound to one another. Vibrational dynamics reveal solvation. Vibrational energy flow provides bond proximity. Vibrational transition dipoles give bond angles. Two-dimensional IR spectroscopy provides the means to extract these quantities, which are otherwise not available from an FT-IR spectrum.
Two-dimensional IR spectroscopy is the multidimensional analog of FT-IR spectroscopy. An FT-IR spectrometer measures the vibrational spectrum of a sample by probing it with an infrared pulse. In 2D IR spectroscopy, we first provide an infrared pump pulse that vibrationally excites the molecules in the sample, and then we probe the sample again, some time later. By varying the frequency of the pump pulse, we can create a 2D spectrum that correlates the vibrations that were pumped with those that were probed. The correlation occurs because the molecules are vibrationally coupled. Vibrational coupling provides information about structure, binding, and dynamics.
As an example, consider the FT-IR spectrum shown in Figure 1, which is collected for a dilute mixture of two molecules. The FT-IR spectrum contains three absorption bands. We know from the frequency range that these bands must correspond to functional groups with triple-bond character, but which functional group belongs to which molecule? The measured 2D IR spectrum for the same mixture is also shown. Notice that each of the absorption bands in the FT-IR spectrum produces a pair of peaks along the diagonal in the 2D IR spectrum. A diagonal slice along a 2D IR spectrum contains essentially the same information as an FT-IR spectrum. In addition, there are pairs of cross peaks between peaks 2 and 3, but not between peaks 1 and 2 or between 1 and 3. Thus, peaks 2 and 3 are vibrationally coupled, but neither is coupled to peak 1. Therefore, from the peak pattern we can quickly and confidently assign peak 1 to one molecule and peaks 2 and 3 to the other. In fact, the solution is a mixture of W(CO)6 (peak 1) with a rhodium metal dicarbonyl (peaks 2 and 3). The peaks in the rhodium dicarbonyl exhibit very strong cross peaks because they share a common metal atom, but do not couple to the modes of W(CO)6 because the two molecules do not bind to one another. Thus, the cross peaks show us connectivity between absorption bands, and this information in turn teaches us about structure. Of course, mixtures like this can be disentangled using just FT-IR spectroscopy, such as by peak fitting with a spectral library or changing the concentration. But in many cases, the additional information from 2D IR spectra provides information not easily obtainable by other means. I give a few examples below in answers to other questions.
Figure 1: Experimental FT-IR and 2D IR spectra for a mixture of W(CO)6 and a rhodium dicarbonyl (RDC). For each peak in the FT-IR spectrum, the 2D IR spectrum exhibits a pair of diagonal peaks. The cross peaks in the 2D IR spectrum reveal that the two higher frequency peaks are coupled to one another, which is because peaks 2 and 3 are from a rhodium dicarbonyl (RDC) whereas peak 1 is from W(CO)6. W(CO)6 and RDC do not have cross peaks between them because the mixture is too dilute. (Data collected by Tianqi Zhang.)
Why did you start pursuing this technique?
Zanni: Peter Hamm, Manho Lim, and Robin Hochstrasser published the first 2D IR spectrum (1) in 1998 just as I was finishing my PhD. I was thinking ahead to my postdoctoral research and was looking for an emerging research direction that had a lot of potential. I chose well. The first spectra in 1998 were rough, but with technological improvements and a better understanding of vibrations, the field has blossomed. 2D IR spectroscopy is now being used in scientific fields as diverse as materials, biophysics, and nanotechnology, as well as becoming a useful analytical tool.
What types of information can be obtained with 2D spectroscopy that cannot be obtained with one-dimensional IR?
Zanni: With FT-IR spectroscopy you are pretty much limited to absorption frequencies and pattern recognition of the fingerprint region. Two-dimensional IR spectroscopy provides information about connectivity through vibrational couplings and environment through dynamics.
Perhaps the best way to answer this question is to provide some examples. We have studied the transmembrane domain of the M2 proton channel from the influenza virus. The M2 channel is the binding site of amantadine, which is an anti-flu drug that has been used for 40 years. By studying the 2D lineshapes, which provide information about hydration, we were able to determine the residues that line the pore and observed a previously unknown structural change upon channel gating. We were able to determine hydration by creating a photon echo pulse sequence that measured the amount of Gaussian vs. Lorentzian lineshape that contributes to the absorption band. It turns out that Gaussian lineshapes are very sensitive to hydration but Lorentzians are not. It is very difficult to rigorously extract these components from an FT-IR spectrum. Another example is the aggregation of the human islet amyloidpolypeptide (amylin). This peptide self-assembles into long fibers that are associated with type 2 diabetes. As a result, there is an enormous interest in understanding and inhibiting this fiber growth. By monitoring the coupling between strands, we were able to map the mechanism by which these proteins assemble, in what I believe is still the most detailed mechanism for any of the 20 human diseases caused by amyloids (2).
What were the main challenges you had to overcome to make 2D IR spectroscopy work?
Zanni: The main challenge to implementing 2D IR spectroscopy is generating the pulse sequences. In many ways, 2D IR spectroscopy is analogous to 2D nuclear magnetic resonance (NMR) spectroscopy. In 2D NMR, one uses a sequence of radio frequency pulses to measure the coupling between nuclear spins. In NMR, it is quite simple to generate the pulse sequences because radio frequency technology has been around for decades. For 2D IR, most often four laser pulses are required to generate a spectrum. As you might imagine, overlapping four laser beams in space and time is very challenging even for experts in ultrafast spectroscopy, especially considering that mid-IR laser beams are invisible to the naked eye.
My research group made two contributions that helped solve this technical barrier. First, we showed that one does not need four separate beams, but that two would suffice. Using two beams also eliminates a whole bunch of other difficulties associated with producing the highest resolution spectra with the proper phasing so that positive peaks point up and negative peaks point down. Second, we invented a way of computer programming the laser pulses, so that pulse sequences can be generated with ease and on the fly, without having to rearrange optics. We can now collect spectra in seconds that used to take hours. Our approach is now being used by many research groups across the world. It simplifies the spectrometer to such an extent that my former postdoctoral researcher Dr. Chris Middleton and I have started a company together to commercialize pulse shapers and 2D IR spectrometers, which is called PhaseTech Spectroscopy, Inc.
How exactly does mid-IR pulse shaping work?
Zanni: Femtosecond pulse shaping was invented about 20 years ago to manipulate visible laser light. There are a number of different ways of doing it in the visible, and several research groups tried to extend those methods into the mid-IR, but with limited success. One of the original ways, but possibly the most under appreciated, was by Professor Warren Warren. He used an acousto-optic modulator that filtered the spectrum of the laser light. He tailored the sound wave in the modulator with an arbitrary waveform card and thus could computer program the pulse shapes. We mimicked his approach, but used an acousto-optic modulator made of germanium so that it would work directly in the mid-IR.
Why is pulse shaping better than the other potential solutions — hole-burning and four-wave mixing?
Zanni: One of the neat things about 2D IR via pulse shaping is that you can program the pulse sequence to do whatever you want. Our pulse shapers can collect spectra using either a hole-burning or a four-wave mixing approach. In fact, hole burning done with our pulse shaper is better than with etalons (the traditional way of creating the narrow pulses), because you can use a Gaussian- instead of a Lorentzian-shaped pump pulse (which has better resolution). For any pulse sequence we can also use phase cycling, which was not previously possible, which enables a host of new capabilities including the elimination of mechanical chopping, and thus decreases data collection time by a factor of two. Regarding four-wave mixing, the original way of using four-wave mixing involved having all four pulses have independent laser beams. A setup like that could use our pulse shaper — in fact, we do something similar when collecting 3D IR spectra — but that beam geometry has all the problems associated with high resolution and phasing that I mentioned above. Since inventing our pulse shaping method, we have disassembled all of the four-wave mixing setups in my research group.
What limitations does the pulse-shaping method have?
Zanni: The pump-probe beam geometry that the pulse shaper utilizes has been criticized for being less sensitive than a four-beam mixing geometry, but to my knowledge no one has done a quantitative comparison (I wish that I had done one before disassembling our four-wave mixing setup; rebuilding one would take months).
Polarized laser pulses are a very powerful way of probing molecular structure because they can measure the relative orientations between functional groups or can be used to eliminate diagonal peaks from the 2D IR spectra (which sometimes obscure the weaker cross peaks). The standard pulse shaper can do some, but not all polarizations. My research group has also built a polarization pulse shaper that can create any polarization sequence that one wants, although that style of pulse shaper is not yet available commercially through PhaseTech. Perhaps the biggest drawback is the throughput efficiency of the pulse shaper, which is about 25%. In principle this can be 40 or 50%, and improvements are being made, but it turns out that efficiency is a minor drawback. In the last few years it has become straightforward to generate 25–40 µJ of mid-IR commercial laser sources. To put that quantity in perspective, the first 2D IR experiments in 1998 were done with 1 µJ. In my opinion, the benefits of pulse shaping far outweigh the limitations. I am convinced that the productivity of my research group has surged in the past five years due to mid-IR pulse shaping.
One area where you are applying the technique of 2D IR spectroscopy is to the study of the kinetics of protein aggregation, specifically in the study of amyloid proteins involved in type 2 diabetes. What have you been able to see that you could not have seen with other methods?
Zanni: We published a paper in Nature Chemistry last year that I think exemplifies the types of information that we can obtain with our approach better than other approaches (3). Amyloid fiber formation is an extremely difficult problem for X-ray crystallography and NMR spectroscopy because solving the problem requires information about both structure and kinetics. In this Nature Chemistry paper, we studied a model peptide inhibitor from rodents. The rodent peptide does not aggregate and so it was used to design a drug that was approved by the US Food and Drug Administration. We thought that the rodent peptide would inhibit amylin aggregation by breaking up the C-terminal beta-sheet of the fibers, and would be otherwise inert. Instead, it prevented the N-terminal sheet from forming and ultimately itself templated into beta-sheet fibers onto the side of the human fibers. This rodent peptide had never before been observed to form amyloid fibers. In fact, neither of the usual methods for studying amyloid structures, TEM and Tht fluorescence, revealed any structural changes. Thus, not only were the results surprising, but it exemplified the information content available from 2D IR that is not easily obtained with other methods.
How did your method enable you to see what you saw?
Zanni: We were able to make these novel insights because of three capabilities made possible by 2D IR spectroscopy. First, just as with FT-IR spectroscopy, we can study aggregates, membrane proteins, and other systems that are not easily amenable to X-ray crystallography or NMR spectroscopy. Second, our pulse-shaping technology collects data so quickly that we can monitor kinetics on the fly. That enables us to monitor the real-time aggregation of these proteins. Third, we get good structural information. In our amyloid studies we usually also use isotope labeling, in which case we can obtain residue level structural information on a kinetically evolving system of an aggregate. In this regard there are few, if any, other comparable techniques.
What conclusions or insights are you gleaning from these results?
Zanni: One satisfying result from our work was a collaboration that was formed with James Nowick, who is an organic chemist at the University of California-Irvine. Nowick designs amyloid inhibitors. He saw my talk, and based on the mechanism, he and I designed a series of inhibitors together. They worked as we intended and we are now writing a manuscript on the topic. It illustrates that, for amyloid formation, the final fiber structure is less important for designing inhibitors than are intermediates.
Another field to which you have applied this method is solar cell research — specifically the study of charge injection, which is a key step in the conversion of solar to electrical energy in dye-sensitized nanocrystalline thin films. Why was this a good problem to tackle using your method?
Zanni: This problem is a good one for our method for similar reasons as for amyloid aggregation described above: NMR and most other standard structural tools cannot be applied to semiconductor interfaces like dyes on semiconductors. Thousands of dyes have been studied as potential next-generation solar cell materials, but the structure and orientation for nearly all of these dyes on the TiO2 is unknown. Two-dimensional IR spectroscopy has the added benefit that it is also an ultrafast technique, and so we also used it to time-resolve the injection of electrons from the dyes into the TiO2. For a model compound that we started with, we discovered that the molecule adopted two different conformations on the TiO2 surface and that one of those conformations had an electron injection time that was at least 10 times faster than the other. To my knowledge, no one had so definitively shown multiple conformations and certainly no one had resolved different injection rates on the same sample. The general thinking is that fast electron injection leads to higher efficiency. Thus, if one were trying to optimize a solar cell, one would presumably try to maximize the number of molecules that bind in the preferential conformation.
Your pulse-shaping approach has been extended to 2D visible spectroscopy. What is the status of that work?
Zanni: Soon after collecting our first 2D IR spectrum using a mid-IR pulse shaper we realized that a similar approach could be used to collect 2D visible spectra as well. My research group did not own a visible pulse shaper, so I called a good friend of mine, Niels Damrauer at the University of Colorado. Niels is an expert in pulse shaping and so within a couple of months he had written the computer code to generate the visible pulse sequences and we got our spectra. As I stated above, visible pulse shaping has been around for 20 years so there are many researchers worldwide with the equipment already in their labs to perform experiments like these. There are quite a few groups that have now mimicked our work in this regard and are looking at very interesting materials and biological systems.
Martin Zanni, PhD, is the Meloche-Bascom Professor of chemistry at the University of Wisconsin-Madison. Direct correspondence to: firstname.lastname@example.org
What are the next steps in your work on this method?
Zanni: We have just performed experiments that I think are a new technological and intellectual milestone in 2D visible spectroscopy. I think that it will enable many new insights into polymer and molecular photovoltaics. But it isn't published yet, so you'll just have to wait a few months.
(1) P. Hamm, M.H. Lim, and R.H. Hochstrasser, J. Phys. Chem. B 102(31), 6123–6138, DOI: 10.1021/jp9813286 (1998).
(2) S.-H. Shim, R. Gupta, Y.L. Ling, et al. Proc. Natl. Acad. Sci. U. S. A. 106(16), 6614–6619, DOI: 10.1073/pnas.0805957106 (2009).
(3) C.T. Middleton, P. Marek, P. Cao, et al., Nat. Chem. 4(5), 355–360, DOI: 10.1038/NCHEM.1293 (2012).