INTERCONNECTS

by Kathleen M. Wong -- 1926 was a very good year for science. Physicists Werner Heisenberg, Paul Dirac, and Erwin Schrödinger independently came up with mathematical ways to predict the locations of the electrons whirling about an atom’s nucleus. The discovery was a critical one for chemists. The positions of electrons, and the ways they can be shared with other atoms, determine how atoms link into molecules and how molecules react with other molecules to form new substances. Describing these interactions is particularly important for organic chemists, who study the behavior of molecules containing carbon. Organic molecules can link into astonishingly intricate structures that can include multiple rings, side chains, lattices, and polymers. In nature, organic molecules form the basis of hormones and steroids, pheromones and poisons. In industry, they are synthesized into medicines and plastics, flavorings, pesticides, and much more. Despite the help provided by quantum mechanics, many fine-grained details about organic chemistry still remain a black box. In 1929, Dirac wrote that with the discovery of quantum mechanics, "the underlying physical laws necessary for the mathematical theory of a large part of physics and the whole of chemistry are thus completely known, and the difficulty is only that the exact application of these laws leads to equations much too complicated to be soluble." But solving such cumbersome equations is finally within reach. The computational obstacles presented by quantum mechanics equations, says professor of chemistry Kendall Houk of the University of California at Los Angeles, are tailor made for the number-crunching talents of supercomputers. For nearly two decades, he has depended on NCSA’s supercomputing resources to analyze the intricacies of organic reactions. Quantum computations Most chemists work amid shelves filled with bottled solutions and bubbling beakers. But Houk and his 30-strong research group labor largely in cyberspace. "NCSA’s resources are extremely valuable to us. Almost every one of our projects has some really big calculations in it," Houk says. "In principle, you could study anything locally, but you might get answers that were so inaccurate they were not useful. With more powerful resources, you can get a better, more accurate answer." After doing rough approximate calculations on local workgroup computers, Houk and his colleagues have routinely turned to NCSA resources over the years. Today, they rely on the center's IBM p690 to conduct their more realistic but time-consuming computations. The results have grown progressively more rewarding. "Over the last century, we have pretty much been explaining things after the fact, but we're getting to the point where we’re able to predict things of interest to explore in experiments," Houk says. Computing speeds, in other words, are finally catching up to real-time chemistry. Most of the Houk group’s efforts are aimed at building computer models of what happens in the proverbial beaker. Though visions of molecules swinging this way and that might come to mind, their simulations are typically far more abstract. "One molecule has to crash into another and disrupt its otherwise happy, stable structure. It takes energy to break existing electron bonds and make others. Quantum mechanics can tell us what those energies are quantitatively, so we can compare competing reactions," Houk says. The more energy it takes to create a given structure, the more slowly the reaction is likely to proceed--and the less likely that structure is to be formed. Computational chemistry offers many advantages beyond not requiring lab coats and goggles. Researchers can quantify the relative amount of energy it takes to produce different reaction products; study any transition products formed along the way; and even learn larger lessons about organic chemistry theory, such as why a particular catalyst works with some compounds but not others. Transition structures of an intramolecular aldol cyclization catalyzed by the amino acid phenylalanine. In the preferred pathway (green), the molecule forms the transition structure with the lowest energy barrier (right), yielding more of one stereoisomer than the other. Predictions without palmistry Houk and his group often collaborate with experimental chemists--the traditional lab-coat kind--on interesting problems they’ve come across in their work. One recent opportunity was provided by Benjamin List of the Max Planck Institute in Muelheim, Germany. List studies the production of chiral molecules--molecules that, like our two hands, are mirror images of one another. Such stereoisomers can have distinct chemical or bioactive properties, so chiral catalysts, which favor the production of one stereoisomer over than another, can be very valuable. List challenged the Houk group to predict the relative quantities of four different stereoisomers that could be produced by the amino acid catalysts proline and phenylalanine. Houk doctoral student Sami Bahmanyar took on the problem. Bahmanyar took three months to calculate the energies required to produce all possible transition states, products, and pathways--a process called mapping out the potential energy surface--and reported her relative yield predictions to List. When compared to the experimental results, her computed forecasts were astonishingly accurate. "We were able to predict the major products in all cases. The only time we were not exactly right, it was a difference of less than three percent," Houk says. "This should give people some confidence that this method is appropriate for studying these kinds of reactions." Postdoctoral researcher Fernando Clemente used similar quantum mechanical calculations to explain the catalytic properties of proline and pheylalanine. Proline serves as a better catalyst in some cases, while phenylalanine is more effective in other situations. He and his fellow postdoc Christophe Alleman and graduate student Paul Cheong made a variety of predictions about these catalysts that should prove valuable to chemists. At present, it takes far longer to churn out predictions than experiments. But the difference may not stand for long; faster computers should mean shorter calculation times with every passing year. "At some point it will be possible to use theory to screen a lot of potential catalysts," Houk predicts, "to run through a list of 1,000 in a few weeks and pick out a few promising candidates for intensive experimental study." Genie in a bottle In another recent study, Houk graduate student Kelli Khuong helped Boston College chemistry professor Marc Snapper explain the apparently capricious behavior of a group of reactions. Snapper’s experiments were conducted with variations of groups connecting an alkene to the highly reactive molecule cyclobutadiene. In about half of the reactions, the cyclobutadiene reacts with the alkene to which it is tethered. The result is the desired Diels-Alder adduct, which contains a new cyclohexene ring. In the other reactions, very similar versions of cyclobutadiene tend to form dimers--the chemical equivalent of conjoined twins--instead. Snapper wanted to know what caused the differences. Khuong restudied all the reactions computationally, exploring the reaction pathways and determining the lowest energy--and most likely--pathways for each reaction. It’s a feat that’s nearly impossible in real life, because cyclobutadiene reacts far too quickly. But with a computer simulation, says Houk, "we can essentially put in a bottle something that cannot ever be observed by experimenting, inspect it, and determine what it is about the geometry that makes some of these very easy to achieve, why some are distorted and have atoms crashing into each other and reacting, and why some have high energies and are not easily achievable." Understanding why chemicals with some structural variations react, while others don’t, is the kind of theoretical information that is transferable to other problems. Multi-ring carbon molecules are found in many natural products, and learning to efficiently synthesize variations on these chemicals could someday add drugs to our medicine chests or new flavorings to foods. And for theoretical chemists like Houk, shining a light on the black boxes of organic chemistry is what it's all about. This research is supported by the National Science Foundation and the National Institute of General Medical Sciences. Team Members: Christophe Alleman Sami Bahmanyar Paul H.-Y. Cheong Fernando Clemente Kendall Houk Kelli Khuong Benjamin List Marc Snapper