INDUSTRY

By Kathleen Ricker, NEESgrid/TRECC Research Editor -- Combining molecular dynamics and quantum mechanics, a Hope College chemistry professor and his student are developing a computational method in order to study how substances dissolve. Solvation is an everyday process, one most people don't give much thought. Sugar dissolves in tea, oil-based paints in turpentine, nail polish in acetone. It's also well known that certain substances, like oil, will not dissolve in other substances, like water. While it seems intuitive that solutes (like salt or sugar) dissolve best in fluid solvents (like water), whether a solute and a solvent are compatible depends on their polarity, that is, how charges are distributed within a molecule. If there is an uneven distribution of positive and negative charges the molecule is polar. Polar solutes will dissolve in polar solvents; hence, sugar will easily dissolve in water, because both substances have some negative and some positive atoms, but because oil is non-polar, consisting of nearly neutral atoms, it will not dissolve sugar. Solvation is involved in the vast majority of chemical reactions; however, this important process is not entirely understood. "A lot of what's important in chemistry is understanding why a reaction happens in a particular way…or why you mix two things together, A and B, and get C, but not D, when D is a perfectly viable option," says Brent Krueger, an assistant professor of chemistry at Hope College in Holland, Michigan. "The way solvation happens, and the way solvent affects a chemical reaction, are really the same thing, so if we can understand one of them, we can understand the other in detail." Solvation on the particle level Solvation occurs when molecules of a solvent surround and stabilize those of a solute. The solvent molecules are always in random thermal motion--the higher the temperature, the faster they move around. As the solvent molecules move, the solute sees a constantly changing environment--sometimes surrounded by six solvent molecules, sometimes seven, sometimes turned this way, sometimes that. All these different environments yield slightly different solute energies such that the solute energy fluctuates rapidly. These solvent-driven energy fluctuations have a crucial effect on the outcome of a chemical reaction. The problem, however, is that there are limits to physical methods used to study chemical interactions, such as optical spectroscopy, a technique that uses light to examine molecular interactions. "Spectroscopy can give you a pretty detailed picture of the timescales of motions in the system, but it can't actually tell you what's moving around in the system," explains Matt Zwier, a 2004 Hope College graduate who has been a student of Krueger's. However, says Zwier, simulating the interaction between solvent and solute provides a way to study these movements that spectroscopy doesn't. "A simulation will allow you to see what's actually moving." Largely using NCSA's Platinum and Titan Linux clusters, Krueger and Zwier are working on perfecting a computational approach that uses a combination of molecular dynamics and quantum mechanics to identify and calculate the movements of solvent molecules and their effect on the excitation energy of solute molecules. Their method is based on an earlier method developed by Ian Mercer, Ian Gould, and David Klug of Imperial College in London. It combines classical mechanics--specifically molecular dynamics (MD)--with quantum mechanics (QM) to calculate the optical properties of a solute-solvent system. Krueger emphasizes that while there exist a number both of computational and experimental methods for studying solvation, "there's not a strong connection between computational and experimental research. So one of the things we're trying to achieve is to connect our computational method very directly with experimental results." The experimental and computational parts of their work are complementary; each allows them to examine details of the interaction that the other might leave obscure. Dissolving pictures The simulation that forms the basis of Krueger and Zwier's work involves a single solute molecule, in this case a dye called oxazine-4. The oxazine-4 molecule is surrounded by about 12,000 methanol molecules, which constitute the solvent. "Basically…we're looking at the fluctuations that occur as the system just sits there at room temperature," says Krueger. "We're not doing anything to the system; it just sits there, with all the oxazine and methanol molecules moving around…all those little fluctuations tell us about how the solute and solvent are interacting." The molecular dynamics component of the simulation, which is applied classical mechanics, involves sampling all the configurations that take place as the system fluctuates and taking periodic snapshots that show the positions of all the atoms at a given point in time. "The classical mechanics part [of the simulation] is valuable to us because it's a very simple treatment, so that we can afford to have a very large system with 12,000 methanol molecules," Krueger explains. "We can treat it for a fairly long time and get millions of different snapshots that show how all the atoms are arranged." However, classical mechanics does not work for elementary particles--in this case, electrons. Therefore, a quantum mechanics component, which helps to predict the behavior of electrons, is used to calculate the excitation energy of the solute for each of the millions of snapshots. "The oxazine molecule is bathed in all these tiny charges from the methanol molecules, and in each snapshot the 80,000 charges from the methanol molecules are going to be a little bit different and so, therefore, is the oxazine molecule," Krueger explains. "When we do the quantum mechanics, it registers both the changes in the oxazine structure and the effects of the methanol solvent through all those little charges." Putting it all together The system that Krueger and Zwier have put together involves four separate parts. Two are standard applications, including AMBER 7, a standard molecular dynamics code, and Gaussian 98, which does the quantum mechanics. What's unique about the code, however, is the "glue" that holds the two methods together--dozens of scripts written by Zwier that automate the collection of the molecular dynamics snapshots, convert them into a Gaussian-compatible format, parse out essential information, and send it to an output file for processing later. Performed once or a few times, these are simple procedures. However, Krueger and Zwier need to perform these tasks millions of times--which, as Krueger says, changes everything. "A lot of people write scripts to make their lives easier, but in this case, these scripts are absolutely necessary to doing the calculation." The fourth component of the code has also been developed by Zwier. Deceptively simple, it seems merely to plot the excitation energy of the oxazine molecules over time. However, this information is the key to the whole experiment: It describes the fluctuations in both the oxazine-4 molecules and the methanol molecules. "It turns out that after you've done both the molecular dynamics and the quantum mechanics, that energy versus time is really the only information you need in order to simulate the results from any kind of optical experiment," says Krueger. He anticipates that the code that Zwier has put together, with some modification, will be able to simulate a broad variety of spectroscopic experiments. Science on a shoestring The work of pulling together a complex framework for a scientific application is often done by advanced graduate students as part of a large research group. However, most of the work on this project has been done by Zwier, who doesn't actually begin his graduate work until Fall of 2004, when he will be entering the doctoral program in the Department of Chemistry at the University of Illinois at Urbana-Champaign. "I always was about 70 or 80 percent sure that I wanted to be in research as a career," says Zwier, "but this project has really solidified that. Furthermore, it indicated what I want to do in research--I've got it narrowed down to computational theoretical work or spectroscopy, as opposed to a broader area like physical chemistry or biochemistry." Krueger says that the development of the code has been a critical component both of his research and of Zwier's undergraduate education. "Many people probably have clusters of 100 CPUs or more, but at a liberal arts college, where our budgets aren't huge, we don't have a lot of local computing resources. NCSA has made it possible for us to do this." This research is supported by Research Corporation, Hope College, and the National Science Foundation's Research Experiences for Undergraduates program. Team members: Brent Krueger Matthew Zwier Courtesy of the National Center for Supercomputing Applications (NCSA) and the Board of Trustees of the University of Illinois