THE ANNUAL DEIXIS2010 DEPARTMENT OF ENERGY COMPUTATIONAL SCIENCE GRADUATE FELLOWSHIP SCIENCE GRADUATE OF ENERGY COMPUTATIONAL DEPARTMENT

DETONATION DETERMINATION Alejandro Rodriguez’s Practicum Provides an ”Explosive” Surprise

PAGE 5 THE ANNUAL DEIXIS TABLE OF CONTENTS 13 20 DEIXIS, The DOE CSGF Annual is published by the Krell Institute. The Krell Institute administers the Department of Energy Computational Science Graduate Fellowship program for the DOE under contract DE-FG02-97ER25308.

For additional information about the Department of Energy Computational Science Graduate 5 Fellowship program, the Krell Institute or topics 17 22 covered in this publication, please contact:

Editor, DEIXIS The Krell Institute 1609 Golden Aspen Drive, Suite 101 Ames, IA 50010 OUR NEW LOOK (515) 956-3696 www.krellinst.org/csgf 4 20 26 Since 2002, DEIXIS, the journal of the Department of Energy Practicum Profiles Alumni Profiles Winning Essays Copyright 2010 by the Krell Institute. Computational Science Graduate Fellowship (DOE CSGF), has The Pause That Refreshes Graduates Notch Achievements Encouraging Communication All rights reserved. connected fellows, attracted applicants, and informed academics and policy-makers about computational science and one of the 5 Alejandro Rodriguez 20 Michael Wu 26 Anubhav Jain, Winner nation’s top education programs. Now DEIXIS has a new look and Detonation Determination Researcher Tunes Why Don't Batteries DEIXIS (ΔΕΙΞΙΣ) transliterated from classical Greek into the Roman alphabet, (pronounced da¯ksis) means a display, a greater focus on fellows and alumni. Into Networks Improve Like Transistors? mode or process of proof; the process of showing, proving or The first section features fellows’ practicum experiences. 9 James Martin demonstrating. DEIXIS can also refer to the workings of an individual’s keen intellect, or to the means by which such The cover story highlights Alejandro Rodriguez, whose practicum Dealing With Dimensionality 22 Judith Hill 28 Milo Lin, Honorable Mention individuals, e.g. DOE CSGF fellows, are identified. produced surprising discoveries about explosive azides. In May, Speaking Many Languages Under the Hood Alex added to his CV with a Proceedings of the National Academies 13 Sarah Richardson DEIXIS is an annual publication of the Department of Energy Computational Science Graduate Fellowship (DOE of Sciences paper outlining a novel method to calculate the Practicum Sends Fellow 24 Jon Wilkening 30 Scott Clark, Honorable Mention CSGF) program that highlights the DOE CSGF fellows and Casimir force. Back to the Bench Wilkening Wades Solving Genomic Jigsaws alumni. The DOE CSGF is funded by the Office of Science Next DEIXIS revisits three alumni, including Jon Wilkening, Into Fluid Situations and the National Nuclear Security Administration’s Office of Defense Programs. a competitive swimmer whose interest in fluid mechanics is more 17 Jack Deslippe than academic. Shedding Light on 32 This issue also incorporates winners in an essay competition that Nanoscale Interactions Howes Scholars encourages fellows and alumni to write for lay readers. This year’s Living Up to a Legacy top piece, by fellow Anubhav Jain, is of interest to anyone whose laptop battery expired at the wrong time. Finally, DEIXIS reviews 10 years of the Frederick A. Howes Scholar in Computational Science award. Past recipients say the 24 35 honor set a standard for their careers — a standard 2010 honoree Directory Julianne Chung is sure to exceed. Editor Copy Editor The redesigned DEIXIS is accompanied by the launch of 35 Fellows Shelly Olsan Ron Winther deixismagazine.org, an on-line counterpart where you’ll find articles Senior Design emphasizing computational science at some of the multi-program 38 Alumni Science Writer julsdesign, inc. DOE national laboratories. Thomas R. O’Donnell We welcome your comments.

DEIXIS 10 DOE CSGF ANNUAL P3 practicum profiles THE PAUSE THAT REFRESHES THE PRACTICUM IS A VITAL PART OF THE DOE CSGF — A PART THAT HAS BEEN KNOWN TO CHANGE LIVES

The Department of Energy Computational DETONATION DETERMINATION Science Graduate Fellowship (DOE CSGF)

supports the nation’s ALEJANDRO RODRIGUEZ Massachusetts Institute of Technology brightest science and Lawrence Livermore National Laboratory engineering students, allowing them to concentrate on learning AS IS OFTEN THE CASE in science, things didn’t go quite as planned during Alejandro Rodriguez’s practicum at Lawrence Livermore National Laboratory. and research. The work But, as it frequently is the case, too, the results were interesting anyway. Rodriguez is a recipient of a Department of Energy Computational Science Graduate of more than 200 DOE Fellowship and a doctoral candidate in condensed matter theory at the Massachusetts CSGF alumni has helped Institute of Technology (MIT). His summer 2008 practicum assignment was to help decipher the electronic structure and atomic-level chemistry when a shock hits cuprate the United States remain azide. Azides are a class of difficult-to-handle primary explosives often used to ignite more competitive in a powerful charges in mining and other applications. Lead azide, for instance, is a common Apart of car airbags. global economy. Defining the structure and chemistry could help researchers understand why heavy-metal azides are so sensitive and release such large amounts of energy. ~~~~~ “Few things are clear about the azides,” says Livermore Staff Scientist Evan Reed, who oversaw Rodriguez during the practicum. Experiments are difficult because the materials are dangerous and detonation occurs in mere picoseconds. “You can’t study reactions on that time scale with any existing experimental tools,” Left to right: THE YOUNG SCIENTISTS in the Department of Energy Reed adds. With the simulations he and Rodriguez worked on, “We’re getting the Sarah Richardson, Computational Science Graduate Fellowship devote years to their doctoral first-ever glimpse at what happens in the process of detonating these materials.” Alejandro Rodriguez, research. For at least three months, however, they break their laser-like It isn’t easy. Such simulations require modeling a cell of dozens of atoms and, more Jack Deslippe and focus to try something different. specifically, the degrees of freedom for hundreds of electrons. “You’re looking at a very, James Martin in Fellows travel to DOE national laboratories, usually during the summer, very, very large system,” Rodriguez says. Using density functional theory (DFT), a common Washington, D.C., to serve practicums. They work under lab researchers on projects that electronic structure method, may work, but modeling even a tiny time scale at the annual DOE rarely bear more than a tangential relation to their doctoral studies. The like picoseconds or nanoseconds increases the difficulty dramatically, he adds. CSGF conference. experience lets fellows apply what they know while learning new things; to Rodriguez was charged with modeling cuprate azide’s inert properties and molecular test their abilities while adding new ones; and to keep one foot in academia structure using DFT software coupled with Multiscale Shock Technique (MSST), an ab while dipping a toe into the world of the national laboratory system. initio multiscale method Reed helped develop. Rodriguez was able to predict the material’s They return to their universities with new perspectives, new energy ground state energy and other characteristics — and discovered that in detonation it and sometimes even new career paths. It’s the pause that refreshes quickly develops metallic properties. and recalibrates.

P4 DEIXIS 10 DOE CSGF ANNUAL DEIXIS 10 DOE CSGF ANNUAL P5 practicum profiles

Rodriguez was able to predict the material’s ground state energy and other characteristics — and discovered that in detonation it quickly develops metallic properties. ~~~~~

“That was a surprise to us and it will electromagnetism. In recent years he’s the Casimir forces on one hand,” says several methods he, McCauley, be a surprise to many others,” Rodriguez focused on electromagnetic field Rodriguez’s advisor, Steven Johnson. Joannopoulos and Johnson developed is says. “That’s a good thing. It means we fluctuations at zero temperature — That’s changed since Johnson, fellow based on the finite-difference time-domain and other people didn’t understand the otherwise known as the Casimir effect MIT professor John Joannopoulos, physics (FDTD) scheme. As the name implies, the properties as much as we thought we did.” or Casimir force. graduate student Alexander McCauley method calculates equations for electric Dutch physicist Hendrik Casimir and Rodriguez developed algorithms that and magnetic fields as they evolve in NOT SO EASY AFTER ALL predicted the force’s existence in 1948. for the first time efficiently and accurately time, independent of the frequency This visualization shows In short, Reed and Rodriguez believed The classic characterization of the force compute the Casimir force between of electromagnetic conduction. reaction products cuprate azide would be a relatively easy target describes two plates in a vacuum, facing objects with complex geometries. Now The method discretizes the partial during detonation for analysis, a belief that appears incorrect. each other just micrometers apart. Quantum “If you want to calculate some crazy shape differential equations used to calculate the of nitromethane. “It was basically metallic at a very electromagnetic field fluctuations, Casimir or some complicated periodic structure … Maxwell Green’s function at points around Some of the molecules, shown in silver, are early stage — sufficiently early that we said, should push the two plates together. we can do it.” the complex bodies the researchers want responsible for couldn’t really get very far with the shock The effect typically manifests itself Rodriguez “has really been a huge to model. The equations are solved in electronic conduction. simulation,” Reed says. As a result, the only on a tiny scale, but it’s become part of this. He really was the one who successive time steps as the electric and researchers turned their attention to more important with the rise of pushed this work to completion,” he adds. magnetic fields respond to current pulses hydrogen azide, also known as hydrazoic microelectromechanical systems Johnson had a crude proof of concept he at each point. Totaling the electric field acid, another explosive they believed (MEMS), microscopic machines that developed as a postdoctoral researcher, results provides the Casimir force, with would be unlikely to have metallic often include moving parts. but Rodriguez “was the one who actually accuracy depending on computer power properties, since it consists of only “Because the Casimir force is usually turned it into a practical method.” and time available to run the problem. opposite results. You would predict the correspondence between calculating the hydrogen and nitrogen. attractive, these objects can stick together The new algorithms have predicted plates would decrease the force between the quantum electromagnetics generating the That assumption also appears wrong, and, therefore, fail,” Rodriguez says. TRACKING THROUGH TIME some surprising Casimir force quirks. In two blocks,” Rodriguez says. Researchers are Casimir force in a vacuum and calculating Reed says. Simulations indicate that Accurate calculations could help develop Calculating the Casimir force relies one case the researchers calculated the scrambling to understand the results. the classical electromagnetic behavior of hydrazoic acid, under detonation, changes ways to cancel — or harness — the force. heavily, Rodriguez says, on the ability to force between two metal blocks when two The idea to apply FDTD to Casimir larger bodies in a conducting fluid. from an insulating to a strongly metallic But portraying the Casimir effect calculate the Maxwell Green’s function, parallel plates are placed above and below force calculations arose from another He and his colleagues propose placing state then back to insulating in a mere mathematically is demanding. which characterizes electromagnetic them. As the plates move closer to the concept Rodriguez helped develop: A centimeter-scale metal models in a 10 picoseconds. “Five years ago you could count the response to electronic sources around the blocks, the Casimir force between the “Casimir analog computer” capable of conducting fluid such as saltwater and “This is an unexpected result and geometries for which you could calculate surface of an object of interest. One of blocks decreases. At a critical point, calculating the force without supercomputers then bombarding them with microwaves. basically an extension of what Alex did however, the force begins to increase. or tiny plates in vacuum chambers. How the bodies respond should represent during his time here,” Reed says. When the plates actually contact the In essence, the computer enlarges the how the Casimir force acts on similar Cuprate azide’s unexpected behavior blocks, the force is bigger than when the micrometer-sized structures in a standard bodies at micrometer scale. kept Rodriguez from achieving his plates were nowhere near. Casimir force experiment to centimeter The experiment could allow researchers practicum goal of simulating a shock in the “If you apply the intuition people had scale. Because Maxwell’s equations are to repeatedly calculate the Casimir force on material. Even so, “it was a good experience before, you would predict completely the scale-invariant, there’s a mathematical complex bodies such as capsules and cubes, both for Evan Reed and for me. We found the azides we studied became metallic much too early, so studying these systems using the multiscale method would be difficult.” Rodriguez says learning DFT Calculations indicate it may be possible to use the interplay between the repulsive Casimir force and methods complements his usual research gravity to create stable nanostructure configurations levitated above dielectric slabs. This plot shows the into fluctuation-induced interaction in stable equilibrium center-surface (Lc) and surface-surface (hc) separation between either a Teflon (red) or nanophotonic media — how photonic silicon (blue) nanoparticle and a semi-infinite gold slab. It shows that decreasing R acts to increase the materials respond to light and other surface-surface separation and decrease center-surface separation. The gray areas depict regions in which the Lc or hc can be made equal by an appropriate choice of radii.

P6 DEIXIS 10 DOE CSGF ANNUAL DEIXIS 10 DOE CSGF ANNUAL P7 practicum profiles

DEALING WITH Using mathematical methods DIMENSIONALITY he helped develop, Alejandro Rodriguez has calculated Casimir forces in these and other complex structures. JAMES MARTIN University of Texas Sandia National Laboratories – California

FOR JAMES MARTIN, the hard work of observing something in the field or Martin is applying running a laboratory experiment often starts when data are in. His research focuses on finding the conditions that led to the observation or output — and on how trustworthy the uncertainty quantification findings are. to seismic wave Martin, a Department of Energy Computational Science Graduate Fellowship recipient, focuses on large-scale statistical inverse problems, which track back to infer what parameters propagation. How influenced a result. these waves reflect off “That’s the classical description for inverse problems,” says Martin, a doctoral candidate Rodriguez says. “You could probably build anything to do with experiments, so this is epiphany came his sophomore year, when in computational and applied mathematics at the University of Texas. Inversion is a many different structures easily and run sort of an ambitious goal for me,” he says. he took his first physics course. underground strata of Flong-standing — and challenging — mathematical process, he adds. “The thing that’s new many, many experiments. The sampling “I remember coming home to my and interesting is … we’re trying to quantify the uncertainty in the inversion.” rate is much faster.” ROUNDABOUT ROUTE mom every day and saying, ‘Why didn’t I earth, rock and fluids Uncertainty quantification (UQ) is a hot area in computational science. More powerful Johnson gives Rodriguez and McCauley It’s just the latest ambitious goal in study this before?’” Rodriguez adds. He helps geologists computers and improved algorithms have made simulation an important tool for discovery. much of the credit for describing the a circuitous life that began in Cuba. told his school counselor he aimed to be But if researchers are to rely on simulations, they also must understand the uncertainty computer. When the idea came up, Rodriguez’s stepfather, a physics professor, valedictorian, and achieved his goal while understand just what’s inherent in them. UQ puts a number on the degree to which a simulation can be trusted. “Alejandro got very excited and started was fired for refusing to identify the taking multiple advanced placement “All large-scale, complex models have uncertain parameters that go into them,” says underground, such talking to everyone. He came back to me writers of a letter opposing the Communist courses. He chose MIT largely because Omar Ghattas, Martin’s doctoral advisor at Texas. Climate models, for instance, can’t a week later and had this whole idea really government. His mother also lost her his idol, Richard Feynman, had been an as oil, gas and precisely calculate the effects of clouds and ice sheets. fleshed out.” teaching job, so the family fled to Mexico undergraduate there. “How does one come up with the uncertain parameters in these models?” Ghattas asks. Rodriguez’s work in Casimir force with a plan to emigrate to the United Rodriguez expects to graduate in mineral deposits. “Typically, it’s through the process of an inverse problem: given measurements you try to has created enormous possibilities, States. When the agent hired to get them spring 2010 and has accepted a postdoctoral ~~~~~ infer what the parameters should have been.” Johnson says. “Every time I turn around, into the country legally raised his fee, fellowship at Harvard University to begin in Martin is applying uncertainty quantification to seismic wave propagation. How these Alejandro is calculating another structure 13-year-old Alejandro and his mother June. Besides possibly testing the Casimir waves reflect off underground strata of earth, rock and fluids helps geologists understand or is in another collaboration,” he adds. As crossed the border illegally and later analog computer, he’s interested in just what’s underground, such as oil, gas and mineral deposits. of early 2010, Rodriguez had contributed received political asylum. Fortunately, his applying numerical techniques the group Making predictions of underground structures from seismic readings is a difficult, to 25 papers — with lead authorship on stepfather was able to emigrate legally prior has developed to interactions in finite- classical inverse problem. Martin and Ghattas want to take it a step further by quantifying 11 — in just three years of graduate to Alejandro and his mother's arrival in the temperature cases, in which one object is the uncertainty behind those predictions. school. Impressively, four papers appeared U.S. They settled in the Miami area and warmer than another. So far Martin is testing his methods on “toy” applications using synthetic seismic data, in Physical Review Letters and for two of became citizens. Reed says the practicum gave Rodriguez but he has big plans for his doctoral project. Meanwhile, Martin’s research took on an added those Rodriguez was first author. Rodriguez says his stepfather’s experience in quantum materials techniques, aspect during his 2008 practicum. For part of the summer he again used seismic data to test Some experimental scientists have physics background had an indirect complementing his doctoral training in his methods. expressed interest in building the Casimir influence on his career path. In middle quantum electromagnetism. “He may be The problem he confronted was the notorious “curse of dimensionality”: As more analog computer. Rodriguez himself is and high schools he was more concerned well-positioned to do some innovative dimensions are added to a problem — and more parameters that can influence the outcome interested. “I have never in my life had about fitting in than about grades. The work sort of on the boundaries of those — the size of the problem grows exponentially. fields,” Reed says. “I’m hoping during his The result is “a lot more potentially interesting space to explore” to find a solution, postdoc he’ll think about how to bridge Martin says. this gap.”

P8 DEIXIS 10 DOE CSGF ANNUAL DEIXIS 10 DOE CSGF ANNUAL P9 practicum profiles

With seismic wave propagation, Martin is trying to get at the stiffness of earth and rock layers. ~~~~~

“You can think of all the Earth models factors influencing seismic wave propagation about the probability a hypothesis is true, Marzouk’s favorite, polynomial chaos Martin and Marzouk presented their doctoral research. The subject is seismic that are like the true one and as you up the and other phenomena. The most naïve way then repeatedly update the probability expansions — to quantify uncertainty in results in separate talks at the March 2009 data, but the methods he uses could be dimensionality, the proportion of models to attack such problems is to attempt to distribution as more data come in. the approximation. Society for Industrial and Applied applicable to a variety of inverse problems, like yours becomes smaller and smaller,” capture all potentially relevant factors — “What we came up with … was that Besides using seismic data, Martin Mathematics Conference on Computational including electromagnetic wave scattering he adds. To find the correct one, “you have sacrificing computational tractability in the interesting modes — the things we and Marzouk spent much of the summer Science and Engineering, and Marzouk and image reconstruction. to explore a huge proportion of the space the process. really care about — are the ones that testing the technique on “deconvolution” visited Texas in summer 2009. The two With seismic wave propagation, Martin or have an intelligent way to go after the change the most” as the Bayesian problems. They applied a mathematical hope to generate a paper on the method is trying to get at the stiffness of earth and most interesting space.” KNOCKING DOWN distribution is updated, Martin says. formula or kernel to “blur” a randomly as soon as they can extend it to certain rock layers. In a deterministic inverse The curse of dimensionality means DIMENSIONS Marzouk adds, “There are certain generated input signal. They then used nonlinear problems. approach, scientists calculate the set of even the most powerful computers can’t “That’s not the inherent dimensionality modes that aren’t going to change much information from the blurred signal Having Martin for the summer “was stiffness parameters that best fit observations. handle many such problems. And seismic of the problem,” continues Marzouk, who because your data … have little to say about to seek the original. just a fantastic experience,” Marzouk adds. “You capture the model that best wave propagation is a notoriously high- now is a professor at the Massachusetts these modes. … Those are dimensions we Martin and Marzouk found their He and Ghattas, Martin’s advisor at Texas, accounts for your data but it’s very unlikely dimensional inverse problem. Institute of Technology. “There are a lot of want to take out of the problem. technique worked well on linear problems. hadn’t collaborated before; now they’re the true model looks exactly like that During his practicum, Martin worked things you don’t need, and with limited “We’re trying to expose the intrinsic Deconvolution allowed them to test working together on a project “and James best-fit possibility,” Martin says. “If you on ways to construct a “low-rank subspace” data and noise there are things you are low dimensionality of the problem, where nonlinear problems — an area in which is a key part of the mix.” want to have some kind of fault tolerance — to reduce dimensionality, making the never going to be able to infer.” The question that low dimension combines the properties the two continue to collaborate. “Hopefully we’ll have a longer … you need to know the possible ways it uncertainty quantification problem easier is “How do you distinguish the things you of the prior information, the data you “We know how to approximate the relationship,” Marzouk continues. “This can differ from the best-fit scenarios. to solve. are going to be able to infer from the things actually observe and the physics of the model in a handful of dimensions,” Marzouk problem is of interest to both of us. We’re There may be features that are completely Youssef Marzouk, who supervised you aren’t going to be able to infer or don’t forward model.” says. “We haven’t yet mated James’ work going at it from complementary directions undefined in the best fit but are not Martin during the practicum at Sandia need to infer?” With a tractable problem, the with adaptive approximations of the and this core of dimensionality is relevant unlikely, and you would want to know National Laboratories in Livermore, Martin and Marzouk use Bayesian researchers then use other techniques — forward model, but that will provide an to both.” about them.” Those features may include California, puts it another way. Researchers inference to attack inverse problems. like Martin’s specialty, the Markov chain additional level of payoff.” So relevant, in fact, that Martin’s oil, gas or mineral deposits. must confront an enormous number of Bayesian methods make a prediction Monte Carlo (MCMC) method, or practicum experience had an impact on his

Prior Parameter Distribution with Realizations Posterior Parameter Distribution with Realizations 10 10

9 9

8 8 These images depict marginalized probability distributions for uncertain seismic model parameters. The 7 7 gray areas represent the probability density for the parameters; the darker the gray, the more likely it is that the parameter curve passes through that point. The blue curve represents the ground truth parameters 6 6 used to generate the synthetic data. The other colored curves are representative samples from the prior

5 5 and posterior distributions. The prior distribution describes researchers’ knowledge of the parameter values before considering any measurement data. The prior in this case assigns high probability to parameters with 4 4 an average value of 5 and without spikes or sharp boundaries between layers. The researchers use Bayesian Parameter Value (µ ) Parameter Value Parameter Value (µ ) Parameter Value methods to incorporate measurement data and arrive at the posterior distribution. The difference between 3 3 prior and posterior distributions shows what researchers have learned about the parameters by including the measurements. It shows an improvement in knowledge about parameters down to a depth of about 2 2 0.4, since the probability is darker and concentrated in a narrow band at these depths. Parameters are more 1 1 uncertain at depths from 0.7 to 1.0, which are farther from the surface and therefore more difficult to infer.

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Depth Depth

P10 DEIXIS 10 DOE CSGF ANNUAL DEIXIS 10 DOE CSGF ANNUAL P11 practicum profiles

PRACTICUM SENDS FELLOW BACK TO THE BENCH

PARAMETERS AND PROPOSALS Although his goal is to avoid explicitly Statistical inverse problems generate reducing dimensions of the problem, the SARAH RICHARDSON a probability distribution for potential technique Martin worked on at Sandia Johns Hopkins University School of Medicine parameter choices. Scientists get a did come into play in his doctoral project. Lawrence Berkeley National Laboratory spectrum of answers. His solution requires computing the “Many different choices of parameters Hessian of the forward model as a way of can give you the same observation,” Ghattas tuning the proposal distribution. “That is, says, “so instead of picking just one we try to traditionally, a very expensive thing” Attenuators are SARAH RICHARDSON STRUCK an unusual deal with Adam Arkin statistically estimate and assign probabilities.” computationally, he adds. before completing a practicum in his group at Lawrence Berkeley National Laboratory in regulatory sequences Martin adds: “They can say both the Martin found the linear dimension- summer 2008. things you really do know, with this degree reduction problem he tackled over the that halt gene Richardson, a Department of Energy Computational Science Graduate Fellowship of certainty, (and) the things you don’t know.” summer is closely related to finding the recipient, spent much of her high school and college years getting her hands wet in biology The statistical inverse approach is low-rank subspace for the Hessian. He transcription in bacteria labs at Johns Hopkins University School of Medicine. But since becoming a doctoral a relatively new research area, Ghattas reordered some terms in his algorithm and other prokaryotes. candidate in human genetics and molecular biology at the same school, she’s mostly says. “The big goal is to apply it to to incorporate it. worked with computers, devising software for gene design and for tracking synthetic yeast large-scale problems.” Ghattas says that shows “It was the If they succeed, the genome research. The MCMC method Martin uses to ideal sort of practicum.” Martin “got Richardson missed the smells and sensations of the lab. “You can get the computer to attack statistical inverse problems samples exposed to a new environment, a new researchers will have a tell you whatever you want, but you can’t prove it until you go to the bench,” she says. “You a proposed probability distribution. The scientist in Youssef, and a different class standardized tool that Sreally have to keep both of them together.” Martin found the linear distribution is updated as each sample is of techniques, but the ideas and techniques So she asked Arkin, head of Berkeley Lab’s Synthetic Biology, Physical Biosciences tested against the true distribution. The he developed there have direct application” could provide Division, if she could split her summer between computational work and time in the wet lab. dimension-reduction samples are chosen at random, but in an to his doctoral project. He agreed, and Richardson’s eagerness to return to pipetting and running gels became a fine-grained gene problem he tackled over “intelligent” way, based on derivative Martin hopes to graduate in 2011 and running joke. She gratefully took on even the most mundane tasks — like washing test tubes. information from the problem. soon will take a big step toward that goal: expression control. “It was kind of pathetic,” Richardson laughs. the summer is closely This is where Martin and Ghattas are outlining a proposed research subject. Pathetic or not, she made important contributions to Arkin’s research into an RNA- ~~~~~ related to finding the making some unusual tweaks. “The novel That’s where the big thinking comes in. based transcription attenuator found in Staphylococcus aureus. Attenuators are regulatory thing we’re trying to do is tune the proposal “I’d like to do the seismic inverse sequences that halt gene transcription in bacteria and other prokaryotes. If they succeed, low-rank subspace distribution to reflect the actual distribution,” problem with real seismic data on the the researchers will have a standardized tool that could provide fine-grained gene making the algorithm more efficient, entire globe,” Martin says. “That would expression control. for the Hessian. He Martin says. “Hopefully, this will allow be a monumental number of parameters. “This RNA-regulated attenuator provides an opportunity to, for one, engineer it for reordered some terms us to effectively draw samples from the If we can do it, it would be really fascinating better function so it really is an off switch and has a large dynamic range,” Arkin says. Since true distribution.” scientific research. We can describe it’s an antisense-mediated circuit — one that blocks transcription — engineered attenuators in his algorithm to If the proposal distribution can be what’s underground with some certainty.” might be built into a family of parts operating similarly but orthogonally — without tuned correctly, it will effectively result One thing truly is certain: Martin’s interfering with other functions — in cells, Arkin says. incorporate it. in a low-rank subspace. future is bright. Richardson describes attenuators as having two pieces, like a lock and key. Arkin’s ~~~~~ “This kind of proposal distribution group wants to modify them so a specific lock works only with a particular key. Two or more should be a good approximation of what’s could be inserted in the same transcript, working together much like a logic circuit. really there,” Martin adds. “If that’s the “Think of it as a uniform framework for scalable control of gene expression, making it case then the number of samples for more like electronics than it is now,” Arkin adds. MCMC to converge and reach a solution Richardson’s practicum focused on improving and augmenting existing computer shouldn’t vary with dimension.” code to generate new attenuators orthogonal to “wild type” S. aureus. Under postdoctoral researcher Julius Lucks and graduate student Stanley Qi, she learned the Python programming language and wrote an interface between the group’s existing code and the

P12 DEIXIS 10 DOE CSGF ANNUAL DEIXIS 10 DOE CSGF ANNUAL P13 practicum profiles

Vienna RNA Package — software It’s easy to understand why. designed to predict and compare RNA Richardson’s ambitious research and busy secondary structures. The tweaks made class work at Johns Hopkins would be it possible to interchange Vienna with enough to keep two grad students busy. mFold, another RNA secondary structure The Baltimore resident was in high prediction package, and to compare the school when she began working with Johns two programs’ predictions. Hopkins researcher Jef Boeke, a professor The group’s code starts with the of molecular biology and genetics and now wild-type lock and key and then mutates it, director of the medical school’s High Richardson says. She devised an algorithm Throughput Biology Center. Richardson She’s creating to choose a large set of mutually orthogonal stayed on through the summers as she earned or nonorthogonal attenuators from the her undergraduate degree at the University algorithms to break the mutants. The tool could help the group of Maryland, College Park. Her senior target genome into achieve its goal of composability — the thesis is based on work at Johns Hopkins. potential to chain or stack attenuators to In her last year of college Richardson oligonucelotides — control gene expression. worked on what became GeneDesign, a DNA chunks 60 base With a list of predicted orthogons, Web-based program that automates Richardson headed to the wet lab. construction of genes from DNA segments. pairs long. “It went as you can expect a wet lab to For instance, GeneDesign can start with a go, which is not well,” she says. It took her sequence of amino acids comprising a “We want to make a stable platform Richardson’s algorithms help choose ~~~~~ the rest of the summer to troubleshoot a protein and provide the sequence for a — a chassis for biology. Then it’s up to designs and synthesis strategies most likely Clumps of a methicillin-resistant cloning technique and devise a lab protocol. gene that encodes that protein. GeneDesign others what they make of it” by adding to work in the lab. Her codes also tag genes strain of Staphylococcus aureus, also lets users choose specific codons — genes for specific functions, such as drug for easier tracking and predict which DNA magnified 2,381 times under a GOING FOR THE GLYPHS trios of bases that code for particular amino or biofuel metabolism, Richardson adds. alterations are feasible and which would scanning electron microscope. Another of Richardson’s projects had acids — for specific organisms. The program Most minimal genome research has conflict with design goals, Bader says. Adam Arkin’s group at Lawrence Berkeley National Laboratory a more immediate impact. The group had also inserts restriction sites — places focused on bacteria, Bader says, but yeast is studying an RNA-based used a sequence-editing program called where enzymes can cut the DNA strand. is a better choice for building and testing a “CONTROL Z” FOR transcription attenuator found GENOME DESIGN ApE to store plasmid data, and a system of Since the program was published in network to operate in eukaryotic systems. in this common bacterium as a glyphs to visually track myriad plasmids in 2006, with Richardson as the lead author, “The special twist for our project is instead Although most researchers in way to control gene expression. lab notebooks and experiments. thousands of researchers from around the of trying to come up with it on our own, Richardson’s group work on computers, all When Richardson came to Berkeley, world have accessed it on the Web. we will build a genome and the organism have wet lab counterparts — in Richardson’s Source: Janice Haney Carr, U.S. Centers for Disease Control Qi was assembling the glyphs and linking Despite her GeneDesign experience, will do it” — automatically deleting case, it’s graduate student Jessica Dymond the two programs by hand. “Whenever I Richardson loved lab work and didn’t want unnecessary genes, Bader says. in Boeke’s group — who puts the synthetic see someone doing something by hand, I to become “the computer person.” Boeke, “Anything that survives on the other yeast DNA into cells. say ‘Time to automate,’ ” Richardson adds. however, suggested she study with Joel Bader, end is a potential answer to what is the The process is imperfect, so researchers Using Python, she wrote a program that a biomedical engineering professor who minimal genome,” Richardson says. She’s need a way to roll back the genome design reads the ApE file and generates glyphs would help her learn computational biology. creating algorithms to break the target to an earlier version if there are problems. automatically for database entry. Richardson has focused on automating genome into oligonucelotides — DNA Richardson helped design a Genome The program is general enough that and accelerating the Bader lab’s central chunks 60 base pairs long. Commercial Revision Control System — software that other groups have expressed interest in it, project: creation of a synthetic genome labs synthesize the oligos and Johns allows researchers to revert the genome Richardson says. She planned to continue based on the yeast Saccharomyces Hopkins undergraduate students build to an earlier state. The program allows working on it, but months later still hadn’t cerevisiae — one with all unnecessary them into bigger pieces for testing. “It’s researchers to “check out” DNA segments, found the time. genes eliminated. like splicing a movie together,” Bader says. change them, and check them in to a

P14 DEIXIS 10 DOE CSGF ANNUAL DEIXIS 10 DOE CSGF ANNUAL P15 practicum profiles

IS THERE A “NOT A DOCTOR” IN THE HOUSE? SHEDDING LIGHT ON NANOSCALE INTERACTIONS

Sarah Richardson’s e-mail JACK DESLIPPE address starts with “notadoctor.” University of California, Berkeley It has a lot to do with her repository for project leaders to review. That certainly was the case at Oak Ridge National Laboratory unusual position as a non-medical The software also automates tasks such as Berkeley, Arkin says. His group still uses student in a medical school. She and updating genome regions affected by a software Richardson helped create and other human genetics doctoral new sequence. much of it is essentially unchanged since candidates at Johns Hopkins University Bader’s group is assisted by what her departure. Researchers even improved JACK DESLIPPE’S MANY HOBBIES include managing a Web School of Medicine are required to Richardson calls “our mighty undergraduate her lab protocol so much it’s now automated. page on his family’s genealogy. He posted the information — gathered by some Quebec take two years of classes alongside army” — Johns Hopkins students in a “We invited her to join the group if Deslippes — hoping others would fill in missing connections. future physicians. “Build-a-Genome” course she helped she wanted to, and invited her for a postdoc,” “It’s interesting, because my name is not very common,” says Deslippe, a Windsor, “Johns Hopkins has the second- best med school in the country and create. Students learn lab procedures and Arkin says, only half joking. Richardson Working with Fernando Ontario native and doctoral candidate in computational condensed matter theory at the assemble DNA the researchers can study was “bright, articulate, worked with a University of California, Berkeley. “You can almost find everyone in the world named students are keenly aware of this and Reboredo, a researcher proud to be here,” Richardson says. It and piece together into the yeast genome. purpose and was just an incredibly nice Deslippe. It’s actually a tractable problem, unlike a lot of others.” also makes some of them leery of the The arrangement works well, person with no attitude whatsoever.” in the lab’s Materials Deslippe knows something about tractable problems and filling in the blanks. The graduate students, especially since Richardson says. Researchers get large Richardson hopes to graduate in 2010, Department of Energy Computational Science Graduate Fellow develops methods and their priorities are slightly different. DNA libraries at lower cost. The students but the exact date is uncertain, partly Science and designs simulations that help elucidate how nanomaterials interact with light at the “We’re more interested in earn academic credit and learn about because Johns Hopkins requires its human Technology Division, he Jelectronic level. The problems are difficult, but the answers could lead to new uses for these mechanisms and ‘whys’ and nice molecular biology. “They’re typically very genetics graduate students to take two exotic materials, such as capturing solar energy or generating laser beams. long philosophical discussions,” motivated to do the work, and they get star years of medical school (see sidebar). calculated quasiparticle Experimentalists can observe how materials absorb or emit light, but describing the Richardson says. recommendations from their professors,” After earning her degree, Richardson underlying physical mechanisms requires complex theoretical input. Deslippe’s research When Richardson found a T-shirt Richardson adds. hopes for a position combining her devotion and exciton energies steps in to provide missing connections, predicting the electronic, quasiparticle and optical saying “NOT A DOCTOR” and Richardson wrote software for the to both lab and computer work. One properties from first principles. featuring a classic image of a surgeon in poly-vinyl carbazole with a red slash across it, she figured class and lectures on computer-assisted possibility is metagenomics — the analysis For example, a photon hitting a nanomaterial can create an exciton — an excited it was a perfect way to poke fun at genome design each semester. In fall 2009 of all the genetics in a particular environment. (PVK), a polymer with material state composed of an electron in a promoted condition and a hole left in the lower- she — and Bader — attended every class In Richardson’s ideal post, that energy state. Working with advisor Steven Louie, Deslippe has run simulations showing the students’ contrasting perspectives. potential uses in “Dr. Boeke asked me where I got it and participated in building oligos as environment would be the human body. that in nanosystems these excitons act in ways that are very different from their behaviors and I told him he couldn’t have one, Richardson wrote a database program Only about 10 percent of all the genetic solid-state lighting in bulk semiconductors. obviously,” she jokes. The shirt to track output. data humans carry with them come from In those materials an electron and hole normally attract each other because of their became famous, faculty and family Despite her reluctance to become “the the body, Richardson says. The rest is and solar cells. opposite charges; but the attraction is reduced, or screened, by the other charges and members began calling Richardson computer person,” Bader says Richardson from the bacteria, fungi, yeasts and other decreases monotonically over distance, Louie says. Deslippe’s calculations found that at the Not A Doctor, and she changed is just that. The DOE CSGF “has given her entities that “make up the organism that ~~~~~ certain length scales in one-dimensional nanosystems, like carbon nanotubes, “the her e-mail address. great training because she’s been able to walks around and calls itself a graduate attraction becomes much stronger than you would expect due to screening of all the other Despite the humor, Richardson really learn the fundamentals of what she’s student,” she adds. charges of the system in a peculiar way,” he adds. “Instead of reducing the attraction it knows she’s lucky to receive a doing,” compared to many computational “They’re going to need tools for that actually enhances the attraction of the opposite charges.” first-class medical education. “One of the reasons I love this program is biologists who pick up computer skills on and they’re going to need models,” That’s just one contribution Deslippe has made to the research group, Louie says. because they let us sit through those the job. Richardson’s rigorous courses in Richardson says, the words rushing from Besides the physics work, “He really is a whiz kid in terms of making our computer codes med school classes — and it wasn’t computer science and applied mathematics her in anticipation. “As a human geneticist, run much faster and more efficiently.” Working with other researchers, Deslippe has fun and games all the time.” means she better understands the work the human metagenome would be an improved by several orders of magnitude the speed and efficiency of the group’s suite of Of course, the e-mail name and can contribute at a higher level, awesome thing to study.” codes to calculate the excited state properties of materials. holds only until Richardson earns her he adds. Those skills came to bear on a “bigger” problem during Deslippe’s DOE CSGF doctoral degee. Her sister, who also is practicum at Tennessee’s Oak Ridge National Laboratory (ORNL) in summer 2008. a graduate student, has demanded she surrender it. “I think I’ll change it to ‘notarealdoctor,’” Richardson jokes.

P16 DEIXIS 10 DOE CSGF ANNUAL DEIXIS 10 DOE CSGF ANNUAL P17 practicum profiles

Two views of a model of a poly-vinyl carbazole (PVK) unit

cell (C42H33N3). PVK is one of the largest systems studied in the ab initio GW-Bethe Salpeter approach. Each contains time, but those with a long lifetime behave nanomaterials it could only model small three side groups of nearly independent molecules. almost like an independent particle, but structures. Thanks in part to Deslippe’s with modified properties — a quasiparticle. work, “We’re now able to look at “We single out one electron being nanostructures and periodic systems excited and look at its behavior or look at with unit cells that have hundreds of the hole left behind and see how it behaves,” atoms. That’s a major advance.” Louie adds. “Because it’s a many-body Deslippe and Louie plan to extend system the electron behaves in a different their methods to higher-order optical and way than if you were to have a noninteracting electronic effects — calculating the effects Working with Fernando Reboredo, a on using a Green’s function to calculate an Tiago and Louie, are continuing to study THREE STEPS TRACK system of particles.” of multiple photon absorption and the researcher in the lab’s Materials Science excited electron’s energy as it evolves. PVK’s excited state spectrum. Such MANY-ELECTRON EFFECTS The last step is solving the Bethe-Salpeter coupling of the exciton with lattice and Technology Division, he calculated Deslippe worked with Tiago and polymers have an unusual quality: the Louie’s group developed the GW equation to include the electron-hole vibrations to get a more complete picture quasiparticle and exciton energies in Reboredo to combine aspects of that lowest exciton state has the electron and method in the 1980s to calculate the band interaction. “We single out the two particles of a material’s absorption characteristics. poly-vinyl carbazole (PVK), a polymer method with the Green’s function-Bethe hole present on the same side group, while gap and quasiparticle energies of bulk and treat them more accurately. We ask the “You can explore different systems with potential uses in solid-state lighting Salpeter equation (GW-BSE) approach higher energy states are charge transfer systems like silicon from first principles. question of how the electron and hole interact and even create artificial, yet-to-be-made and solar cells. that the Louie group developed and that excitons with the electron and hole separated Now the researchers are combining it with with each other and all the other electrons in systems,” Louie says. “Then hopefully you PVK is known in the laboratory for Deslippe focuses on in his doctoral research. from each other by an increasing distance. the Bethe-Salpeter Equation to compute the system,” Deslippe says. can tell the experimentalists what to make.” its propensity to generate light under an The researchers also are comparing their “When you pull the electron and hole the spectroscopic properties of a variety of The important part, Louie adds, “is Deslippe also is working to make the electric current, much like a light-emitting results with those obtained with the apart, you get valid excitonic states,” systems, including nanostructures. we can do all the steps in the calculation code even more efficient and user-friendly diode. But Reboredo focuses on the quantum Monte Carlo method, which Deslippe says. “I think that’s a good sign Researchers have long known that without any empirical input. We do it from for other researchers. “Frankly speaking, we material’s light-absorbing qualities. It Reboredo has used extensively. for these applications we’re looking at.” including such effects as the electron first principles.” have one of the most efficient computer codes assimilates light at high energies — in The GW-BSE method essentially Another Oak Ridge group, meanwhile, self-energy and excitons is important for Deslippe and Louie, working with for these kinds of first-principles studies.” the ultraviolet and violet region — and is calculates the Green’s function for two is planning experiments to test the correct results, and some have tried to others, have applied the GW-BSE technique Meanwhile, Deslippe, Reboredo, relatively inexpensive, making it a good particles instead of one, characterizing simulation results. account for them in DFT calculations. But to a number of systems. In one paper they Louie and others are composing a paper photovoltaic candidate. the photo-excited state and properties Just as importantly, Deslippe’s these attempts don’t describe the many-body describe calculations of excitonic effects describing PVK’s physical properties “If you’re going to use it in a solar cell of the exciton. practicum demonstrated that the GW-BSE effects well, Deslippe says. In essence, in graphene. Because of its semi-metallic regarding photovoltaic and solid-state it has to absorb light at the frequency of As he did at Berkeley, Deslippe also method could compute energies of such electrons throughout the rest of the system nature, graphene might be expected to light applications. the solar spectrum, but you also have to find optimized the code to run more efficiently large molecules. That means researchers move around to counter the new field or show little electron-hole interaction since “It’s great to get to work with people a way to separate the electron and hole to on an increasing number of computer could calculate polymers’ absorption charge, modifying the interaction. the freely moving electrons around them who are experts in related subjects, but not get the current going,” Deslippe says. processors and to calculate larger molecules. spectra, band edges, work functions and The GW-BSE method accounts for the would have screened the effect. the type of subject you would get” working In particular, he helped implement exciton binding energies, perhaps leading many-electron effects with a computationally But the calculations found a surprisingly in the research group at Berkeley, Deslippe BIG MOLECULE, BIG PROBLEM real-space truncation — a way to limit to improved polymers for both lighting efficient three-step process. large excitonic effect in graphene — a shift of says. He also learned by getting to know The molecule’s large size, at around calculations of the polymer’s electrostatic and solar cells. First is a DFT calculation to arrive at the prominent peak in the optical absorption researchers throughout the lab, from 100 atoms, makes it a computational interaction with neighboring polymer Deslippe’s practicum broadened his the system’s ground state properties — spectra of 600 milli-electron volts (meV), theoreticians to experimentalists. challenge. Reboredo and postdoctoral molecules in a supercell simulation. horizons, Louie says, by giving him the before electrons are excited. “In order to from 5.2 eV to 4.6 eV. “Bound exciton states Deslippe is getting set to graduate. researcher Murilo Tiago had worked on All the changes paid off: The PVK chance to study a class of complex really design a system or nanostructure can even be found in metallic nanotubes, After that, he’s unsure whether he’ll work calculating the electronic properties of system is one of the largest ever calculated materials different from the graphene one has to determine where the atoms where one would expect the interaction to in a national laboratory, academic or PVK and similar polymers for months. using the ab initio GW-BSE method, and nanotube systems he researches for are,” Louie explains. “What we do as a first be very weak due to screening,” Deslippe corporate setting. But he’s sure of one Tiago’s code is based on a real space density Deslippe says. his dissertation. He also was exposed step is to determine the ground-state says. That raises hopes that graphene thing: “I’m definitely trying to do functional theory (DFT) approach — The calculations showed that excitonic to different mathematical methods for geometry” and accompanying properties. could also be tunable for use in tiny research in some environment.” calculating the charge density of the effects are extremely important, and calculating a material’s optical properties. Next, the GW approach calculates the opto-electronic devices. material on a computational grid — and Deslippe and Reboredo, working with attributes and interactions of a single Deslippe’s contributions have been excited electron as it evolves. Particles vital to achieving such results, Louie says. added to the system “meld” with it over When the group first began studying

P18 DEIXIS 10 DOE CSGF ANNUAL DEIXIS 10 DOE CSGF ANNUAL P19 alumni profiles

ALUMNI PROFILES GRADUATES NOTCH ACHIEVEMENTS

Researcher Tunes Into Networks

MICHAEL WU has the training of a statistician, an Wu hopes to get at what motivates superusers and how they Riding the revolution appeals to Wu. Though his 2008 doctoral applied mathematician and a computational neuroscientist, but emerge, and at what predicts who will become one. “Why do they work is in computational neuroscience, he’s drawn to the complex, what he does may have just as much in common with psychology do what they do? … Aside from the virtual psychological benefit fascinating interactions that drive social media. The abundance of and sociology. (of helping others), they don’t seem to get a lot.” data and the fast pace of the business world, he says, lets him test As principal scientist of analytics at Lithium Technologies Inc. The goal is similar for lurkers: understanding what keeps hypotheses and apply solutions more quickly than in a laboratory or min Emeryville, Calif., Wu studies on-line communities and the them engaged without participating, how to bring them into academic setting. interactions within them. Lithium provides software and services the conversation, and even how to coax them into becoming “We focus on real-world problems that are very practical and in that let companies use these and other social media to serve and superusers themselves. some ways messier — there usually are a lot of constraints and we must understand customers. If done right, Wu says, social media create Turning both kinds of users into business assets involves a lot constantly make trade-offs between time to market and the quality of As principal scientist of analytics at user communities that supplement a firm’s product support, provide of psychology, Wu says. “Incentives, rewards, ranking systems, the solution, or between efficiency and accuracy of the solution,” Wu Lithium Technologies, Michael Wu instant product feedback and turn customers into advocates. gaming dynamics and serendipity all play important roles in managing adds. Yet, “In the business world I feel I can actually make a difference uses anonymized data to generate “A lot of companies are afraid of this social revolution because and maintaining a successful community.” faster, and not just when I’m 80 years old.” thousands of social graphs like these in some ways they lose control of their customers or their brand. Such communities could revolutionize business, Wu says, to understand the interactions in on-line user communities and how People can say whatever they want,” adds Wu, a Department of by giving customers an effective channel to voice opinions, they evolve. The lines represent Energy Computational Science Graduate Fellow at the University and giving companies an efficient way to listen. That makes interactions between pairs of users, of California, Berkeley, from 2002 to 2006. But social media also companies more responsive to consumer complaints and needs. which are represented by the dots. The Vertex: Size=Betweenness Centrality Color=Closeness Centrality help companies boost sales through word-of-mouth advertising, size and color of the dots represent and product innovation through idea exchange. a user’s importance and influence. Lithium, whose clients include AT&T, MTV Networks and Barnes & Noble bookstores, has a wealth of customer conversation data from hundreds of communities, Wu says. That gives him plenty of fodder for detecting trends and testing approaches. Vertex: Size=Betweenness Centrality Color=Closeness Centrality Vertex: Size=Betweenness Centrality Color=Closeness Centrality One of Wu’s research projects focused on the role of customer networks in word-of-mouth marketing. Using anonymized data from several Lithium communities, he developed an agent-based modeling technique to simulate the random diffusion of information across the customer network. Among other things, Wu found that targeting specific influential users made word-of-mouth programs more effective and less expensive; 80 percent of the potential value came from targeting less than 5 percent of the potential audience. Wu’s current research concentrates on understanding two classes of users at opposite ends of the spectrum: superusers, who contribute heavily to forums and interact regularly with others, and lurkers, who visit forums for information but never contribute.

P20 DEIXIS 10 DOE CSGF ANNUAL DEIXIS 10 DOE CSGF ANNUAL P21 alumni profiles

Speaking Many Languages

WHEN STRANGERS ASK WHAT she does, “A lot of researchers in the field of optimization work on Judith Hill has an initial straightforward answer: She’s a discrete problems. The work I do is in continuous problems,” computational scientist. Hill says. Solutions could be any number in a range rather than These visualizations show the solution of an inverse problem characterizing the spread of a “They may say, ‘What’s that?’ Then I say I’m part engineer, distinct separated values. “The solution space we’re looking simulated airborne contaminant plume in the Greater Los Angeles basin. The inverse problem was part mathematician and part computer scientist all rolled up in can be a lot larger than many traditional optimization calculated based on measurements synthesized by solving the convection-diffusion equation using winto one,” adds Hill, a researcher in the Computational Math problems, thus requiring large computers to become the target initial conditions and recording measurements on a 21 by 21 by 21 uniform array of Group at Oak Ridge National Laboratory in Tennessee. computationally tractable.” sensors. Compare the predicted concentrations from solving the inverse problem at zero minutes elapsed time and after 120 minutes with the target concentrations at the same times. The combination appeals to Hill, who held a Department Inverse problems drew her attention, Hill says, because of Energy Computational Science Graduate Fellowship from they offer a rare glimpse at an end use for her work. “In applied 1999 through 2003. “I view myself as a communicator, so I math, the direct use of our work in real-world applications can like sitting at the intersection of several disciplines,” she be many years away, because we’re working on new and says. “Applied mathematics allows me to communicate with unproven methods and techniques.” mathematicians, computer scientists and domain specialists. In one case, Hill and her colleagues calculated a test I can speak a lot of different languages and bring in new problem tracking the spread of an airborne contaminant in ideas they may not have thought of.” the Los Angeles area to its source. Such models are important Hill’s first and perhaps best “language” is fluid dynamics. for planning emergency responses, monitoring public health, Her doctoral research at Carnegie Mellon University laid the assessing hazards and other functions. foundation for modeling how red blood cells are deformed Hill has returned to fluid dynamics since moving to Oak and damaged as they flow through an implanted pump Ridge in 2008, but with a different tack. Instead of focusing designed to assist a failing heart. on finite elements methods, a classical numerical technique She expanded her mathematical vocabulary in the that was her main approach, she’s working within the Applied Math and Applications Department at Sandia MADNESS framework. National Laboratories’ New Mexico location from 2005 to Developed by researchers from Oak Ridge and the 2008. She focused on solving inverse problems, a class of University of Colorado, MADNESS stands for Multiresolution optimization problems, that were constrained by a physical Adaptive Numerical Scientific Simulation. It relies on model represented by a set of partial differential equations multiwavelets, a method treating regions of space with Hill is in a good place to test her approaches at Oak disciplines, at Oak Ridge Hill has opportunities to fulfill (PDEs). In inverse problems the value of some model parameter different degrees of accuracy, to calculate things like fluid Ridge, home to powerful machines like Jaguar, rated in 2009 her favorite role as scientific liaison. is inferred from observations or measured data the flow and molecular electronic structure. The method as the world’s fastest computer. That, the lab’s growth and “I view the role of a computational scientist as a parameter produces. approximates some problems more accurately than the mountainous eastern Tennessee region all attracted her, builder putting together the many pieces required by other techniques, Hill says. she says. computer simulations to get science done. These pieces The other major element in Hill’s research is tuning A laboratory environment also offers a balance of things include developing new physical models, improving algorithms to run well on massively parallel computers. It’s an she loves most about research, Hill adds. “I really enjoy numerical techniques and ultimately developing the especially challenging job these days, she says, as designers mentoring students and postdocs, and within the lab we have software for these simulations,” she says. “Most of all, I move from merely adding processors to creating hybrid the opportunity to do that.” And unlike a university, where enjoy working with my colleagues in other disciplines to machines with combinations of chip architectures. professors often are walled-off from colleagues in other solve problems.”

P22 DEIXIS 10 DOE CSGF ANNUAL DEIXIS 10 DOE CSGF ANNUAL P23 alumni profiles

This two-dimensional simulation models a rubber wheel with an inner radius of 8 mm and outer radius of 10 mm, spinning at 955 Hz and pressed 0.1 mm onto a frictionless plane. The image shows a bifurcated deformation state, exhibiting standing waves, and first principal stress contours (N/mm2).

Wilkening Wades Into Fluid Situations

IT’S NATURAL that fluids fascinate Jon Wilkening. Time-periodic solutions are useful for modeling systems Wilkening and David M. Ambrose of Drexel University another PDE related to Navier-Stokes, the adjoint PDE, that As a competitive swimmer, he’s spent a lot of time struggling that reset themselves, such as ocean waves, Wilkening says. used similar methods to tackle an seemingly unrelated problem: tells you how the drag will change if you change your stroke,” through the world’s most ubiquitous liquid. “If you can set the system going so that it later comes back to computing time-periodic vortex sheets with surface tension. Wilkening says. “Solving the adjoint PDE tells you the best “I think about fluid mechanics a lot while I’m swimming,” where it started, the solution will repeat itself forever.” These A vortex sheet, as they describe it in a paper accepted for way to swim faster.” says Wilkening, a Department of Energy Computational Science solutions “tell you a lot about the system, going beyond what publication in the Proceedings of the National Academy of In their paper, Ambrose and Wilkening use similar adjoint iGraduate Fellow at the University of California, Berkeley, from you can get from studying traveling waves, for example.” Sciences in 2010, is the interface between two fluids as they techniques to compute families of smooth, symmetric 1997 through 2001 and the 2003 Frederick A. Howes Scholar That makes time-periodic solutions good for attacking some shear past each other. Prior to their research, it was not “breathers” — oscillatory nonlinear waves — that alternate in Computational Science. Nonetheless, “It’s hard to figure interesting problems, Wilkening adds. In one proposed project, known whether time-periodic solutions exist. between a flat state of maximum kinetic energy and a rest out how to turn that into something you can use to go faster.” he and UC-Berkeley Civil Engineering Professor Sanjay Govindjee “It’s a really interesting application for the algorithm I state with all the energy stored as potential energy in the Fluid mechanics are a theme in Wilkening’s work, but his take on tires. have for finding periodic solutions,” adds Wilkening, who interface. In some cases the interface overturns before interests range widely, as befits an academic who goes “When a tire is rolling, it’s a system that is inherently received a 2010 National Science Foundation CAREER award returning to the initial, flat configuration, the paper says. wherever interesting problems lead him. He’s researched periodic,” Wilkening says, resetting itself with each rotation. for a proposal entitled ”Optimization and Continuation Wilkening has settled into academia, a place he loves for lubrication theory, microchip failure, stress singularities in Methods are available to model smooth tires effectively, but Methods in Fluid Mechanics.” “What we’re bringing to the the opportunities to work with students and pursue interesting solid mechanics, shape optimization and other subjects as a “as soon as you put a tread on it, they don’t know what to do table are numerical methods that can answer some very research, but it wasn’t a goal. “I thought I would be an engineer postdoctoral fellow at New York University’s Courant Institute next. They don’t know how to model it.” difficult analysis questions.” in high school. It wasn’t until college, when I started taking and, since 2005, as an assistant professor of mathematics Wilkening and Govindjee believe their time-periodic The key, he says, is finding an adjoint system that allows advanced physics and math courses, that I realized one could back at UC-Berkeley. methods can capture the dynamic behavior of tires between computation of the gradient of an objective function quickly actually be a professor.” Many of these applications tie into Wilkening’s current each touchdown of regularly spaced treads onto a surface. If and efficiently. For example, a swimmer may want to minimize Now, “I enjoy answering the hard questions you can’t interest in time-periodic solutions of partial differential they’re successful, their research could enable faster, cheaper his drag — the objective function — as he goes through the answer without simulation,” particularly those involving equations (PDEs). prototyping to make tires more safe and durable. water. The first step would be to write the formula for drag in complicated physics. Plus, “I’ve had a knack for coding my terms of the solution of the Navier-Stokes equations. “There’s whole life. It’s good to do things you’re good at.”

This image shows a time-periodic solution of the Benjamin-Ono equation over one spatial and temporal period. The Benjamin-Ono This snapshot from an animation shows a time-periodic solution of equation is a nonlinear, nonlocal, dispersive a vortex sheet with surface tension, with the interface exhibiting a equation that models internal waves in a deep “breather” pattern of oscillatory nonlinear waves. The interface turns stratified fluid (such as the ocean) in certain over before returning to its initial flat state. The particles are added asymptotic limits. Ambrose and Wilkening used for visualization and are color coded by pressure. To see the full their numerical method to compute families of animation, go to http://math.berkeley.edu/~wilken/vtxs3.html. time-periodic solutions connecting pairs of traveling waves through bifurcation.

P24 DEIXIS 10 DOE CSGF ANNUAL DEIXIS 10 DOE CSGF ANNUAL P25 essay contest winners WINNING ESSAYS ENCOURAGING COMMUNICATION THROUGH ANNUAL WRITING CONTEST

WHY DON’T BATTERIES WINNER IMPROVE LIKE TRANSISTORS? points to the problem of simultaneous optimization: new materials must pass (And what a billion transistors can do about it) through several difficult, often-overlapping by Anubhav Jain hoops to become viable products. Like a potential new headache medicine that soothes pain more quickly but may also AWARDING SOMETIMES I TRY to do too much 8 percent a year), but consumer electronics increase the risk of stroke, a new battery at once. Recently, as I devoured my lunch producers typically offset these gains by material might boost operating time but COMMUNICATION while reading e-mail on my laptop, I introducing power-hungry processors or degrade with use or burst into flames Each point in this graphic represents one material, with the point attempted to open a water bottle with using smaller battery packs, thereby keeping when overheated. When promising new size representing the material’s stability. The graph shows there is a my free hand. I succeeded — only to the total battery life constant. While these materials are found, years or even decades tradeoff between the voltage of a potential new battery material and The DOE CSGF subsequently shower the laptop in spring 8 percent gains might be noticeable to a can be spent engineering cures to their its safety. High-voltage materials have better energy densities but launched an annual water. It would never display e-mail again. materials engineer, they don’t hold a candle undesirable side effects.” generally are less safe. Experimental data, represented by the three points shown, suggest this trend but are unable to confirm it. The essay contest in 2005 Still, I took comfort that my laptop had to Moore’s law, which states that the density of Simultaneous optimization is plot also shows that safety varies widely by chemistry. Oxides, which to give current and lived a full existence, reaching the ripe old transistors (the individual logic components time-consuming because each data point are the best-studied cathode materials, are the least safe for a given former fellows an Sage of three years. Most laptops currently of a computer processor) on CPUs doubles can take days, weeks, or even months to voltage. Up-and-coming chemistries like silicates demonstrate better opportunity to write on the market could easily compute circles every two years. collect. However, computational materials safety, suggesting they are fertile territory for more investigation. about their work with a around my old friend. Nearly every major To better understand this contrast, we science can break Eagar’s law by using broader, non-technical component of modern laptops — the CPU, can examine “Bringing New Materials to theoretical calculations to model material audience in mind. The RAM, hard drive, and even the tiny onboard Market,” a February 1995 Technology properties. Computational materials competition encourages camera — becomes outdated in a few years. Review article. Written by Thomas Eagar, science is now so advanced it can predict better communication of I emphasize nearly everything because former head of the Department of Materials many of a material’s characteristics nearly Functional Theory calculations, some performance, cost and toxicity. Because computational science the battery is a holdout; since the 1990s, Science and Engineering at MIT, it as accurately as lab measurements. of the most accurate techniques for each candidate passes through several and engineering and laptop batteries have comfortably thrown in demonstrates that any new materials It also presents several advantages over modeling materials, to create the world’s optimization hurdles before time-consuming its value to society to the towel after about four hours of operating technology, whether Teflon or titanium, experimental lab work. For one, computers largest computational screening endeavor. (and expensive) lab work even begins, non-expert audiences. time. Slate.com journalist Farhad Manjoo takes about 18 years to progress from lab can work around the clock without coffee Though the details are abstruse, such commercialization can come much quicker In addition to jokes, “What we know about batteries today to consumer use. Eighteen years is an eon breaks (unlike scientists). And, because calculations achieve their accuracy by using than materials discovered without this recognition and a cash is pretty much what we knew about batteries in the world of technology, equivalent to the transistors, RAM and magnetic hard drives the principles of quantum mechanics to pre-screening advantage. prize, the winners back when ENIAC [the first computer, time it took to leap from creating the first have greatly improved, one dollar’s worth predict a desired material property. In the high-throughput battery cathode receive the opportunity created in 1946] was invented.” Although IBM PC to having Google in 10 languages. of computing power today is equivalent to Using our method, we’ve screened more project, we have predicted several new and to work with a Manjoo’s statement is technically incorrect, Transistors skirt “Eagar’s 18-year law” hundreds of thousands of dollars worth from than 20,000 materials as we search for a promising materials for use in next-generation professional science his point remains clear: batteries simply because their basic materials largely stay 20 years ago. Nowadays, even a moderate lightweight, compact battery cathode that batteries. These materials were recently writer to critique and don’t improve at the same pace as other the same while their design improves. Like research budget can purchase hundreds quickly absorbs lithium during discharge made and tested in the lab and exhibit copy-edit their essays. computer components. My goal is to turn a musician who learns a difficult passage of powerful computers to work like a and releases it safely when charged. Such a good performance. The latest winning essays this problem on its head by using the by first playing it slowly, then gradually virtual army of lab hands testing endless material could revolutionize electric vehicles Meanwhile, our computing clusters are published here. stunning improvements made to CPUs, increasing the tempo, computer engineers combinations of materials chemistries. or grid-level power storage. (And it will containing a billion transistors are busy For more information RAM, and magnetic storage to rapidly have designed today’s miniscule transistors Called high-throughput computational allow you to watch that second DVD on doing a whole lot at once, computing on the essay contest, visit advance battery science. by first building large ones, then making screening, this new technique has the long plane rides!) Each potential battery thousands of additional chemical www.krellinst.org/csgf. But first things first: Materials them progressively smaller. potential to discover and optimize new material was evaluated as a whole — not combinations in search of the next researchers would argue that battery So why do materials-driven technologies materials in record-breaking time. My only for its ability to store a lot of energy breakthrough battery material. Fortunately, storage capacity steadily improves (about like batteries stagnate in the lab? Eagar colleagues and I have used Density but also for its potential safety, stability, they aren’t juggling bottles of water.

P26 DEIXIS 10 DOE CSGF ANNUAL DEIXIS 10 DOE CSGF ANNUAL P27 essay contest winners

HONORABLE MENTION UNDER THE HOOD Predicting Proteins with Computers by Milo Lin

WE ARE TAUGHT that cells are the The particular sequence of every possible folds for the protein. Even if the ARE PROTEIN STRUCTURES COMPUTABLE? reaching the tens of thousands even for building blocks of life, so the words protein is encoded by a corresponding protein could try each of those folds in a This complexity means computational small proteins (with water), simulations “biological complexity” usually conjure segment of DNA, which is stored in the trillionth of a second, it would take about simulation is the only theoretical tool for using massively parallel computing have up images of bacteria colonies or the nucleus of each cell in our body. When a 20 billion years for it to find its native fold understanding proteins at the molecular only recently been able to span one human brain. Yet if we were to think particular protein is needed, its DNA — longer than the age of the universe. level. In the most common approach, millionth of a second. about cells as more than just black boxes, blueprint is copied and the code translated Instead, proteins actually fold in called molecular dynamics (MD), the MD simulations already demonstrate we would see that the mystery of life starts into the matching amino acid sequence. milliseconds to seconds. simulation is broken into many small time remarkable accuracy in predicting native within them. When the cell is magnified a The linear chain is then molded in three- There are simple things that self- steps. The technique calculates the forces folds for rudimentary proteins. Insights hundred million times, life is revealed to dimensional space into the structure that organize (freezing water) and complicated on the protein at each time step to simulate from such simulations promise to increase Wbe the net result of molecular machines at will carry out the protein’s particular things that external machinery reliably the folding of the 3-D structure. The alongside the exponential growth of work — machines called proteins. Proteins function. Called the native fold, this make (airplanes). There are also many strengths of the many different forces computing power, because computation are the cleaners, builders, motors, structure is stabilized by chemical forces examples of complicated phenomena that are obtained from experimental data. holds the only promise of atomic-scale messengers and transporters of the cell. both within the protein and between the non-reliably self-organize (the weather). The protein and surrounding water resolution. We are nearing the time when Just as a car runs as a result of cooperation protein and its surrounding environment Proteins are unique because they are molecules have uneven distributions of simulations can finally catch up with between many smaller components, the (mostly water). complicated structures that reliably positive and negative charges. Since like experiments in spanning proteins’ entire cell’s vitality arises from the cooperative self-organize. In fact, they fold astronomically charges repel and opposite charges attract, folding times. When that day arrives, we actions of proteins. So how proteins are PROTEINS: AMAZING MOLECULAR MACHINES faster than random search. This seemingly one of the key forces acting on the protein will finally be able to watch as proteins manufactured into such a rich array of It’s unclear just how a protein attains impossible engineering feat at the comes from adding up all the attraction transform themselves from inert chains moving parts is a basic biological question. its native fold. Life requires proteins to be molecular level is the scaffolding that and repulsion forces between every two into active machines. Of course, unlike with a car, we can’t functionally reliable, so each protein allows life to be simultaneously complex atoms in the simulation. This is the Ever since microscopes first showed simply pop the hood and look. But because sequence has evolved to have a unique and efficient. calculation bottleneck in MD simulations, cells revving with vitality, the inner we know the physics governing proteins, shape that is more stable than all other since the number of forces to be computed workings of cellular machinery have been The energy landscape, which it’s theoretically possible to predict their possible folds combined. scales as the square of the number of hidden from view. Now, we are finally plots the energy of a protein structures with computers. This could aid Because native folds are highly complex atoms. With the total number of atoms starting to peer under the hood of life. as a function of its structure, in designing proteins for use as medicines, and irregular, often resembling a tangled helps represent the protein nano-machines and specialized materials. ball of yarn, their stability is one of the folding process. Starting from a For example, current chemotherapy drugs most awesome examples of how natural random shape, or conformation, cannot be tailored to exactly fit the markers selection can lead to creations that are the protein will change until of cancerous cells, leading the drug to also both fantastical and successful. It therefore it reaches the lowest energy (i.e., native) fold, which is disrupt non-cancerous cells. If protein came as a shock when American biochemist at the bottom of the funnel drugs can be designed to specifically fit the Christian Anfinsen discovered in the 1950s in the energy landscape. markers, chemotherapy could be drastically that proteins fold by themselves without the more effective with fewer side effects. aid of any cellular machinery. Even more amazingly, proteins can Four of these subunit proteins combine to form THE BASICS OF PROTEIN SYNTHESIS find their native folds immediately; in fact, hemoglobin, which transports oxygen and carbon All proteins are chains of simple as American molecular biologist Cyrus dioxide through the bloodstream. Both are molecules called amino acids. Though Levinthal noted in 1968, proteins seem to attached to the iron atom (shown in orange) in there are just 20 different natural amino fold trillions of times faster than expected! the heme group. acids, the number of possible protein Consider the example of a protein 100 amino sequences increases exponentially with acids long. If each amino acid in the chain the length of the protein. Each sequence can be in one of two positions (far fewer corresponds to a structure and each than the actual number of possible structure corresponds to a function. orientations), then there would be 2100

P28 DEIXIS 10 DOE CSGF ANNUAL DEIXIS 10 DOE CSGF ANNUAL P29 essay contest winners

Local Alignments for soi: Escherichia coli str. K12 substr. DH10B, test seqs (1-1/1)

HONORABLE MENTION SOLVING GENOMIC JIGSAWS Bacillus subtilis subsp. subtilis str. 168 Piecing Together DNA With Computing Power by Scott Clark

This map of local alignments shows what parts of which genes line up and where, illustrating relative WHEN I WAS A CHILD I loved Metagenomics deciphers all the genetic areas of conservation between two chunks of sequence data. The comparison shows which areas jigsaw puzzles. Big, small, skylines, fantasy material of all the organisms in a are conserved, which match up, and where shifts, insertions and deletions are, helping scientists landscapes — whatever I could get my community, like all of a person’s intestinal understand how distantly two sequences are related on an evolutionary time line. hands on. I would sit for hours at the tract bacteria. Assembling all the genetic dining room table, meticulously building information of viruses and organisms like fantastic worlds piece by piece. bacteria into their complete genomes lets Today, I teach computers to us deduce vital information about their reassemble puzzles. It’s called composition and relation to each other. This Herculean task seems impossible Another key trick is the ability to What is certain is that supercomputers computational metagenomic assembly, This can lead to new understanding at first, and normally it would be — for parallelize. Modern supercomputers can will play a vital role. By training them Wbut the idea is the same. The puzzles of their interactions and — ultimately — a human. The sheer quantity of the process many thousands of instructions correctly, we can grasp this otherwise metagenomics poses aren’t traditional; vaccines to fight the viruses. information to be analyzed can be on the simultaneously — in parallel. Solving intractable task and gain a depth of each piece is a small chunk of genetic These metagenomic puzzles differ same order of magnitude as the printed puzzles was always easier with a friend; understanding about the world around information — the DNA that comprises in many other ways from the puzzles I works in the Library of Congress. It would similarly, these supercomputers employ us. Using some of the largest computers the universal building blocks of life. solved as a child. They are simpler in some be impossible for a single person to even thousands of processors to assemble pieces in the world lets us solve some of the respects; instead of an entire picture one look at all of the pieces of this puzzle, let while sharing information as they go and smallest and most complex puzzles in only needs to assemble all of the pieces alone begin to assemble it. allowing for useful information to be nature. This knowledge can be used to InClub and Score matches across all alignments. in a line. Unfortunately, the difficulties All is not lost, however, because it’s discovered in a reasonable time. develop vaccines and better understand Global filter of sigma > 7.0 then local filter of CDF. Total genes compared: 20. Reference Gene: Escherichia coli str. K12 substr. DH10B quickly outpace this apparent simplification. possible to train a computer to help, if you The field of metagenomics is still the composition of and interactions The number of pieces can easily reach the can develop a good algorithm — a specific in its infancy, but scientists are sampling between living things, from the microscopic Residues: 401-500 millions, if not billions, for a single assembly. set of rules — for it to follow. As a child, more microbial communities all the time, to humans. Not only is this a vital step Also, each piece represents only an extremely you quickly learn basic “tricks” that allow from mineshafts to human intestinal in understanding the biological systems Cons. Ident. Start(m) and End(c) minuscule fraction of the fully assembled you to solve puzzles more quickly and tracts, and beginning to publish results. around us, it also gives me an excuse to Residues: 501-600 picture, which is usually unknown. efficiently: find the corner pieces first, It’s still unknown what applications this keep playing with puzzles. In metagenomics scientists also are build the border, put all similar pieces research may have or the extent and Cons. Ident. Start(m) and End(c) trying to assemble many pictures concurrently, together, build related parts together. ramifications of its impact. Residues: and many of the pieces may be missing while In much the same way, computational 601-700 others may have multiple duplicates. “Sky scientists can develop an analogous set of Cons. Ident. Start(m) and End(c) pieces” — repetitive patches of blue and rules that will allow a computer to attack Local Alignments for soi: Escherichia coli str. K12 substr. DH10B vs Bacillus subtilis subsp. subtilis str. 168

Residues: clouds — are the bane of any puzzler. this problem more efficiently: find known

Frequency of matches across all other genes. 701-800 Nature provides an equally challenging gene pieces, find overlapping information, Cons. Ident. Start(m) and End(c)

Score = blue and InClub = green and further in shades of red analogue, as many of the genetic pieces build from unique “seed” pieces.

Sequence compared against, red text indicates at least ratio=0.5 InSetInClub are virtually or actually indistinguishable. Sequence from 401 to 800 of 1343. There also can be long stretches of intentionally repeated information. Imagine taking every puzzle in a store and mixing them while making the pieces These bar graphs illustrate alignments of DNA and amino acid smaller, and then removing the boxes with sequences from multiple organisms. They show areas of high the pictures along with a bunch of the This shows how the computer algorithm finds areas of conservation, in which many organisms share particular sequences. pieces. Then try to rebuild all the puzzles conservation. The top line shows where there are matches Areas that produce high values on the bar graph are highly to a level of accuracy that would allow you between one sequence and an offset of another. The lower conserved across many sequences. That makes them likely to to gain information from the resulting line shows when the density of these matches is larger than a represent important genetic information worthy of more study. pictures. That is metagenomic assembly. chosen cutoff value. This area of high density is a potential match between the two sequences and is marked off in the graph.

P30 DEIXIS 10 DOE CSGF ANNUAL DEIXIS 10 DOE CSGF ANNUAL P31 howes scholars scholars

The Frederick A. Howes Scholar in Computational Science award was established in 2001 to honor the late Frederick Anthony Howes, who was a champion for computational science education. ACCOLADE’S NAMESAKE WAS HOWES AWARD FELLOWSHIP’S ADVOCATE BRINGS RECOGNITION, AFFIRMATION AND RESPONSIBILITY

JEFF HITTINGER first whose research focuses on methods for professor at Wellesley College, where she’s Frederick Howes was widely admired heard of Frederick Howes a few months computational plasma physics. researching computational design and for his integrity, fairness, intelligence, humor after starting a post-doctoral research So Hittinger understood the analysis of drugs and biomolecules. and compassion, says Margaret Wright, then a post at Lawrence Livermore National implications a few weeks later, when he “To me, that’s really important and it researcher at Bell Laboratories and today a professor at the Laboratory (LLNL) in 2000. His was named one of two DOE CSGF alumni felt like that was validated” when she received Courant Institute of Mathematical Sciences. His death was coworkers encouraged him to attend a to receive the first Frederick A. Howes the award, she adds. “a shattering blow” to friends and colleagues, and “the March 2001 applied mathematics and Scholar in Computational Science awards. For Mayya Tokman, a fellow from idea arose semi-spontaneously that we should do computational methods conference at the Now in its tenth year, the award recognizes 1996 through 2000, joining Hittinger as something concrete to honor his memory.” An informal committee soon decided that the nearby University of California, Berkeley, recent doctoral graduates of the DOE CSGF one of two initial Howes Scholars “just Department of Energy Computational Science Graduate that Howes’ colleagues organized in not just for their technical achievements, but gave me a boost of confidence” and Fellowship (DOE CSGF) was the most appropriate avenue his memory. also for outstanding leadership, integrity and “reassured me that what I’m doing is for honoring Howes. His portfolio as manager of DOE’s Howes, manager of the Department character – qualities that brought Howes valued.” She’s now an assistant professor Applied Mathematical Sciences Program included the of Energy’s Applied Mathematical Sciences wide admiration. of applied mathematics at the University fellowship, which supports doctoral students who train JProgram, died unexpectedly on Dec. 4, 1999, The 13 Howes Scholars have gone of California, Merced, where she studies in applied mathematics, computer science and an at age 51. Every speaker at the conference, on to careers in academia, at national methods for modeling complex, multiscale application discipline like biology, physics, engineering from academia and the national laboratory laboratories and in private industry. At systems like astrophysical plasmas and or materials science. system, had received support through least one has founded a business. They’re biological cells. “Fred was a great champion of the CSGF, even when it was threatened” with budget cuts, Wright says, and he 2010 WINNER the program. advancing the science in applied Jaydeep Bardhan, a fellow from deserves credit for much of its early success. JULIANNE CHUNG “It just blew my mind,” says Hittinger, mathematics, chemistry, computational 2002 through 2006, was a postdoctoral The organizers wrote to those who had known Howes, Julianne Chung has a Department of Energy Computational fluid dynamics, computer vision and researcher in Argonne National Wright says, asking for contributions to the newly created been selected as the Science Graduate Fellowship recipient artificial intelligence, geoscience, Laboratory’s Mathematics and Computer Frederick A. Howes Scholar in Computational Science 2010 Howes Scholar in from 1996 through 2000 and now a geography, drug design and other areas. Science Division when he was named a award. Those donations, plus a generous sum from Howes’ Computational Science. computational scientist at LLNL. “Here For each, the award affirmed their Howes Scholar for 2007. “People were family, built the foundation for an endowment. Others Dr. Chung was a fellow you have all these top people in applied accomplishments, both technical and coming down the hall, saying, ‘Not only have since contributed as well. from 2006-2009. She mathematics and scientific computing personal. “There’s more to being a are we proud of you, but just so you know, DOE CSGF recipients who have completed or plan to received her Ph.D. getting up and giving talks in memory of scientist, in some sense, than just the Fred was one of the best guys around,’” complete requirements for their doctoral degree in a from Emory University Fred and sharing memories of Fred.” His science,” says Mala Radhakrishnan, a says Bardhan, now an assistant professor calendar year, both with fellowship support or after and is currently a NSF reaction: “Wow, this guy is really something.” fellow from 2004 through 2007 and the in physiology and molecular biophysics receiving support for the maximum number of years, are Mathematical Sciences “I very quickly became aware of who 2008 Howes Scholar. Science also is at Rush University Medical Center in eligible for that year’s Howes award. Postdoctoral Research Each fall the Krell Institute, which manages the Fellow at the University he was and how important he was to working with others with integrity, Chicago. “It became personal at that point.” program for the DOE, solicits nominations from department of Maryland. researchers in the field,” adds Hittinger, says Radhakrishnan, now an assistant chairs, advisors and fellowship coordinators at universities qualified fellows attended. Nominations must include the names of at least two people who are familiar with the HOWES SCHOLARS candidate’s accomplishments. Krell contacts the references to seek endorsements, 2009 winner David Potere (left) and completed nominations — including the nomination 2008 winner Mala Radhakrishnan (left) letter — are sent to a review committee drawn from the 2007 winners Jaydeep Bardhan (left) national laboratory, academic and DOE CSGF alumni and Kristen Grauman (right) communities. One or two scholars are chosen to receive a 2006 winners Matthew Wolinsky cash award and an engraved crystal memento at the annual (left) and Kevin Chu (right) DOE CSGF conference in Washington, D.C. 2005 winners Ryan Elliott (left) 2009 2008 2007 2006 2005 and Judith Hill (left)

P32 DEIXIS 10 DOE CSGF ANNUAL howes scholars

20 fellows directory GRADUATING CLASS It became even more personal when Howes’ Scholar lecture. The talks he’d heard from previous widow, Mary Hall, and son, Michael, made a rare recipients — some of them good friends — were 10 appearance at the fellows’ conference to participate all top-notch and highly detailed. in the presentation to Bardhan and Kristen In the end, Bardhan says, his talk not only Grauman, a fellow from 2001 through 2005 and described his research, but also addressed the now an assistant professor of computer science responsibilities fellows have when they “leave the Mark Berrill Ashlee Ford Ariella Sasson at the University of Texas at Austin. “It definitely cradle” and move into the field. Colorado State University University of Illinois at Rutgers University makes you think that, ‘All right, you have some “We’re in an environment where people don’t Electrical & Computer Urbana-Champaign Computational Biology & Engineering Chemical Engineering Molecular Biophysics living up to do,’” Bardhan says. Grauman, whose always understand computational science very well research focuses on computer vision and machine or don’t always know what high-performance Advisor: Jorge Rocca Advisor: Richard Braatz Advisor: Anirvan Sengupta Practicum: Los Alamos National Laboratory Practicum: Brookhaven National Laboratory Practicum: Sandia National Laboratories – learning, says the honor “was completely unexpected. computing can do,” Bardhan says. It’s CSGF Contact: [email protected] Contact: [email protected] New Mexico It meant a lot and it was a special occasion for me to graduates’ job to enlarge the community, he adds. Contact: [email protected] come to the meeting.” Many Howes Scholars see the award as a At the conference, Howes Scholars also deliver commission to do just that. At Wellesley, “It’s Robin Friedman Arnab Bhattacharyya Massachusetts Institute presentations describing their research. For many especially important … for students to see Massachusetts Institute of Technology Michael Sekora recipients, the talk marks a transition from graduate professors being recognized for research,” of Technology Computational & Princeton University school to careers in the field. It also could be their Radhakrishnan says. Her award demonstrates that Computer Science Systems Biology Continuum Mechanics, PDE, Numerical Analysis last opportunity to describe their research to an even a smaller institution can contribute to science, Advisor: Madhu Sudan Advisor: Christopher Burge audience with widely varying interests. that professors strive to excel at both teaching and Practicum: Sandia National Laboratories – Practicum: Lawrence Berkeley National Laboratory Advisor: James Stone California Contact: [email protected] Practicum: Lawrence Berkeley National Laboratory “Science has become extremely specialized and research and that women can thrive in technical Contact: [email protected] Contact: [email protected] people sit in a corner and work on their problems,” fields. “That’s a huge thing,” Radhakrishnan adds. Tokman says, but because fellows work in a range of Hittinger believes the award — and the DOE application areas, conference speakers must make CSGF itself — have gained luster over the years as Tal Danino Brian Levine their research understandable to a cross-section of alumni have moved into top research and teaching University of California – Cornell University Benjamin Smith scientists. She says that made her 2001 talk both posts, and that they’re only going to become more San Diego Transportation Systems Harvard University challenging and enjoyable. prestigious. While that’s boosted his career, Dynamics of Systems Biology Experimental High Energy Physics Advisor: Linda Nozick “You have a room full of smart people … and Hittinger says the expectations accompanying Advisor: Jeff Hasty Practicum: Sandia National Laboratories – Advisor: Masahiro Morii you have the opportunity to describe what you do,” the award may have had a greater impact. Practicum: Lawrence Berkeley National Laboratory New Mexico Practicum: Lawrence Berkeley National Laboratory Contact: [email protected] Contact: [email protected] Contact: [email protected] says Tokman, whose lecture on modeling astrophysical “Maybe it’s just the way I interpret it, but when plasmas included a “Star Trek” reference. you’re given an award like the Howes Scholar, it’s Bardhan, whose research focuses on not just in recognition of what you’ve accomplished simulation of large biological molecules, says he but also of your potential,” Hittinger adds. “You Benjamin Sonday Jack Deslippe Carolyn Phillps Princeton University felt some pressure to present a first-class Howes have something you need to live up to.” University of California, Berkeley University of Michigan Applied & Computational Physics Applied Physics Mathematics

Advisor: Steven Louie Advisor: Sharon Glotzer Advisor: Yannis Kevrekidis Practicum: Oak Ridge National Laboratory Practicum: Sandia National Laboratories – Practicum: Sandia National Laboratories – Contact: [email protected] New Mexico California Contact: [email protected] Contact: [email protected]

John Evans University of Texas Alejandro Rodriguez 2004 winner Collin Wick (right) Computational & Applied Massachusetts Institute Mathematics of Technology 2003 winners Oliver Fringer (third Condensed Matter Theory from left) and Jon Wilkening Advisor: Thomas Hughes Practicum: Sandia National Laboratories – Advisor: Steven G. Johnson (blue shirt) New Mexico Practicum: Lawrence Livermore National Laboratory 2001 winners Mayya Tokman (second Contact: [email protected] Contact: [email protected] from left) and Jeffrey Hittinger 2004 2003 2001 (second from right)

P34 DEIXIS 10 DOE CSGF ANNUAL DEIXIS 10 DOE CSGF ANNUAL P35 fellows directory

TH YEAR RD YEAR ST YEAR 4FELLOWS 3FELLOWS 1FELLOWS Gregory Crosswhite Alex Perkins Carl Boettiger Kathleen King Anne Warlaumont Mark Maienschein-Cline Mary Benage Irene Kaplow University of Washington University of California – Davis University of California – Davis Cornell University University of Memphis University of Chicago Georgia Institute of Technology Stanford University Physics Theoretical Ecology Biology – Ecology and Evolution Applied Operations Research Computational Development Physical Chemistry Geophysics Computational Biology Advisor: Dave Bacon Advisor: Alan Hastings Advisor: Alan Hastings Advisor: John Muckstadt Psycholinguistics Advisor: Aaron Dinner Advisor: Josef Dufek Contact: [email protected] Practicum: Lawrence Livermore Practicum: Oak Ridge National Laboratory Practicum: Lawrence Berkeley Practicum: Argonne National Laboratory Advisor: David Kimbrough Oller Contact: [email protected] Contact:[email protected] National Laboratory Contact: [email protected] National Laboratory Contact: [email protected] Practicum: Argonne National Laboratory Miles Lopes Contact: [email protected] Contact: [email protected] Contact: [email protected] Noah Reddell Seth Davidovits University of California, Berkeley Matthew Reuter Eric Liu University of Washington Princeton University Machine Learning Hal Finkel Northwestern University Eric Chi Massachusetts Institute of Technology Computational Plasma Modeling Applied Physics Advisor: Martin Wainwright Yale University Theoretical Chemistry Rice University Computational Fluid Mechanics for Fusion Energy Advisor: Nathaniel Fisch Contact: [email protected] Physics Advisor: Mark Ratner Bioinformatics/Statistics Advisor: David Darmofal Advisor: Uri Shumlak Contact: [email protected] Advisor: Richard Easther Practicum: Oak Ridge National Laboratory Advisor: David Scott Practicum: Lawrence Berkeley Contact: [email protected] Peter Maginot Practicum: Princeton Plasma Contact: [email protected] Practicums: Lawrence Berkeley National National Laboratory Leslie Dewan Texas A&M University Physics Laboratory Laboratory and Sandia National Contact: [email protected] ND YEAR Troy Ruths Massachusetts Institute of Technology Nuclear Engineering Contact: [email protected] Sarah Richardson Laboratories – California FELLOWS Rice University Nuclear Waste Management Advisor: Jim Morel Johns Hopkins University School of Medicine Contact: [email protected] Brian Lockwood 2 Bioinformatics Advisor: Linn Hobbs Contact: [email protected] Steven Hamilton Human Genetics and Molecular Biology University of Wyoming Advisor: Luay Nakhleh Contact: [email protected] Emory University Advisor: Joel Bader Scott Clark Computational Fluid Dynamics Edward Baskerville Contact: [email protected] Devin Matthews Computational Mathematics Practicum: Lawrence Berkeley Cornell University Advisor: Dimitri Mavriplis University of Michigan Carmeline Dsilva University of Texas – Austin Advisor: Michele Benzi National Laboratory Applied Mathematics Practicum: Argonne National Laboratory Ecology Samuel Skillman Princeton University Chemistry Practicum: Los Alamos National Laboratory Contact: [email protected] Advisor: Peter Frazuer Contact: [email protected] Advisor: Mercedes Pascual University of Colorado – Boulder Chemical Engineering Advisor: John Stanton Contact: [email protected] Practicums: Los Alamos National Contact: [email protected] Astrophysics Advisor: Stanislav Shvartsman Contact:[email protected] Danilo Scepanovic Laboratory and Lawrence Berkeley Douglas Mason Advisor: Jack Burns Contact: [email protected] Joshua Hykes Harvard/Massachusetts Institute of Technology National Laboratory Harvard University Sanjeeb Bose Contact: [email protected] Scot Miller North Carolina State University Signal Processing/Cardiovascular Modeling Contact: [email protected] Physics Stanford University Christopher Eldred Harvard University Nuclear Engineering Advisor: Richard Cohen Advisor: Eric Heller Computational Fluid Dynamics Hayes Stripling University of Utah Atmospheric Sciences Advisor: Yousry Azmy Practicum: Sandia National Laboratories – Curtis Hamman Practicum: Lawrence Berkeley Advisor: Parviz Moin Texas A&M University Climate Modeling Advisor: Steven Wofsy Practicum: Oak Ridge National Laboratory New Mexico Stanford University National Laboratory Contact: [email protected] Nuclear Engineering/Uncertainty Advisor: Thomas Reichler Contact: [email protected] Contact: [email protected] Contact: [email protected] Flow Physics and Computer Engineering Contact: [email protected] Quantification Contact: [email protected] Advisor: Parviz Moin Kurt Brorsen Advisor: Marvin Adams Kenley Pelzer Milo Lin Paul Sutter Practicum: Sandia National Laboratories – Matthew Norman Iowa State University Practicum: Lawrence Livermore Thomas Fai University of Chicago California Institute of Technology University of Illinois at Urbana-Champaign California North Carolina State University Physical Chemistry National Laboratory New York University Theoretical Physical Chemistry Physics Cosmology Contact: [email protected] Atmospheric Sciences Advisor: Mark Gordon Contact: [email protected] Applied Mathematics Advisor: David Mazziotti Advisor: Ahmed Zewail Advisor: Paul Ricker Advisor: Frederick Semazzi Contact: [email protected] Advisor: Charles Peskin Contact: [email protected] Practicum: Brookhaven National Laboratory Practicum: Oak Ridge National Laboratory Ying Hu Practicum: Oak Ridge National Laboratory Travis Trahan Contact: [email protected] Contact: [email protected] Contact: [email protected] Rice University Contact: [email protected] Jeffrey Donatelli University of Michigan Amanda Peters Biomedical Engineering University of California, Berkeley Nuclear Engineering Aleah Fox Harvard University Paul Loriaux Cameron Talischi Advisor: Rebekah Drezek Britton Olson Applied Mathematics Advisor: Edward Larsen University of Pennsylvania Applied Physics University of California – San Diego University of Illinois at Urbana-Champaign Practicum: Lawrence Berkeley Stanford University Advisor: James Sethian Practicum: Argonne National Laboratory Genomics and Computational Biology Advisor: Efthimios Kaxiras Computational Biology Computational Mechanics National Laboratory Computational Fluid Dynamics Contact: [email protected] Contact: [email protected] Advisor: Carlo Maley Contact: [email protected] Advisor: Alexander Hoffmann Advisor: Glaucio Paulino Contact: [email protected] Advisor: Sanjiva Lele Contact: [email protected] Practicum: Argonne National Laboratory Practicum: Sandia National Laboratories – Practicums: Lawrence Berkeley National Piotr Fidkowski Sean Vitousek Christopher Quinn Contact: [email protected] New Mexico Anubhav Jain Laboratory and Lawrence Livermore Massachusetts Institute of Technology Stanford University Charles Frogner University of Illinois at Urbana-Champaign Contact: [email protected] Massachusetts Institute of Technology National Laboratory Structural/Computational Engineering Environmental Fluid Mechanics Massachusetts Institute of Technology Electrical and Computer Engineering James Martin Materials Science and Engineering Contact: [email protected] Advisor: Raul Radovitzky and Hydrology Computational Biology Advisor: Todd Coleman University of Texas John Ziegler Advisor: Gerbrand Cedar Practicum: Argonne National Laboratory Advisor: Oliver Fringer Advisor: Tomaso Poggio Contact: [email protected] Computational and Applied Mathematics California Institute of Technology Practicum: Lawrence Berkeley Cyrus Omar Contact: [email protected] Contact: [email protected] Contact: [email protected] Advisor: Omar Ghattas Aeronautics National Laboratory Carnegie Mellon University Aaron Sisto Practicum: Sandia National Laboratories – Advisor: Dale Pullin Contact: [email protected] Neural Computation Virgil Griffith Norman Yao Evan Gawlik Purdue University California Practicum: Oak Ridge National Laboratory Advisor: Brent Doiron California Institute of Technology Harvard University Stanford University Computational Materials Science Contact: [email protected] Contact: [email protected] Armen Kherlopian Practicum: Los Alamos National Laboratory Theoretical Neuroscience Condensed Matter Physics Applied Mathematics Advisor: Xiulin Ruan Cornell University Contact: [email protected] Advisor: Christof Koch Advisor: Mikhail Lukin Contact: [email protected] Contact: [email protected] Geoffrey Oxberry Computational and Systems Biology Practicum: Lawrence Berkeley Contact: [email protected] Massachusetts Institute of Technology Advisor: David Christini Claire Ralph National Laboratory Christopher Ivey Edgar Solomonik Chemical Kinetics/Transport Phenomena Practicum: Princeton Plasma Cornell University Contact: [email protected] Stanford University University of California, Berkeley Advisor: William Green Physics Laboratory Theoretical Chemistry Flow Physics and Computer Science Practicum: Sandia National Laboratories – Contact: [email protected] Advisor: Garnet Chan Tobin Isaac Computational Engineering Advisor: James Demmel California Practicum: Sandia National Laboratories – University of Texas Advisor: Parviz Moin Contact: [email protected] Contact: [email protected] Jeffrey Kilpatrick New Mexico Advisor: Omar Ghattas Contact: [email protected] Rice University Contact: [email protected] Practicum: Los Alamos Zachary Ulissi Computer Science National Laboratory Massachusetts Institute of Technology Advisor: Luay Nakhleh Brenda Rubenstein Contact: [email protected] Interfacial Physics, Fluid Dynamics Practicum: Pacific Northwest Columbia University and Catalysis National Laboratory Theoretical Chemistry Contact: [email protected] Contact: [email protected] Advisor: David Reichman Practicum: Los Alamos National Laboratory Contact: [email protected]

P36 DEIXIS 10 DOE CSGF ANNUAL DEIXIS 10 DOE CSGF ANNUAL P37 Ethan Coon Michael Driscoll Matthew Farthing Christopher Gesh Columbia University Boston University University of North Carolina Texas A&M University Applied Mathematics Bioinformatics and Systems Biology Flow and Transport Phenonmena in Porous Media Computational Transport Theory, Nuclear Reactor ALUMNI DIRECTORY Fellowship Years: 2005-2009 Fellowship Years: 2002-2006 Fellowship Years: 1997-2001 Analysis, Nuclear Non-Proliferation Matthew Adams Allison Baker Nawaf Bou-Rabee Jarrod Chapman Current Status: Dataspora, Inc., Current Status: Research Hydraulic Fellowship Years: 1993-1997 University of Washington University of Colorado California Institute of Technology University of California, Berkeley John Costello San Francisco, CA Engineer, USACE ERDC Coastal and Current Status: Pacific Northwest Computational Electromagnetics Iterative Methods for Linear Systems, Parallel Monte Carlo Methods, Numerical Solution Whole Genome Shotgun Assembly; University of Arizona Hydraulics Laboratory National Laboratory Fellowship Years: 2007-2008 Computing, Software for Scientific Computing of Stochastic Differential Equations, Computational Genomics Applied Mathematics Jeffrey Drocco Fellowship Years: 1999-2003 Molecular Dynamics Fellowship Years: 1999-2003 Fellowship Years: 1998-2002 Princeton University Michael Feldmann Matthew Giamporcaro Joshua Adelman Current Status: Lawrence Livermore Fellowship Years: 2002-2006 Current Status: Post-Doctoral Fellow, Current Status: Software Engineer, Microsoft Biophysics and Computation California Institute of Technology Boston University University of California, Berkeley National Laboratory Current Status: Courant Instructor, Computational Genomics Program, Fellowship Years: 2004-2008 Computational Finance Adaptive algorithms, Artificial Neural Networks Biophysics New York University DOE Joint Genome Institute Nathan Crane Current Status: Graduate Student, Fellowship Years: 1999-2002 Fellowship Years: 1998-2000 Fellowship Years: 2005-2009 Devin Balkcom University of Illinois at Urbana-Champaign Princeton Physics Tank Lab Current Status: Executive Vice President, Current Status: Engineering Consultant, Current Status: Postdoctoral Fellow, Carnegie Mellon University Jenelle Bray Eric Charlton Computational Mechanics Quantitative Research, Walleye Trading GCI, Inc. University of Pittsburgh Robotics California Institute of Technology University of Michigan Fellowship Years: 1999-2002 Brian Dumont Software LLC Fellowship Years: 2000-2004 Computational Structural Biology Aerodynamics and Computational Fluid Dynamics Current Status: Sandia National University of Michigan Ahna Girshick Zlatan Aksamija Current Status: Faculty, Dartmouth College Fellowship Years: 2006-2009 Fellowship Years: 1992-1996 Laboratories – New Mexico Aerospace Engineering Krzysztof Fidkowski University of California, Berkeley University of Illinois at Urbana-Champaign Current Status: Postdoctoral Fellow, Current Status: Lockheed Martin Aeronautics Fellowship Years: 1994 Massachusetts Institute of Technology Visual Perception (Computational Modeling Nanostructured Semiconductor Thermoelectrics Michael Barad Stanford University Simbios NIH Center Company, Fort Worth, TX Stephen Cronen-Townsend Current Status: Airflow Sciences Corporation Computational Fluid Dynamics and Psychophysics) Fellowship Years: 2005-2009 University of California, Davis for Biomedical Computation Cornell University Fellowship Years: 2003-2007 Fellowship Years: 2001-2005 Current Status: Computing Innovation Computational Fluid Dynamics Michael Chiu Computational Materials Physics Amanda Duncan Current Status: Assistant Professor, Aerospace Current Status: Researcher, New York University Postdoctoral Fellowship in the NTG Group, Fellowship Years: 2002-2006 J. Dean Brederson Massachusetts Institute of Technology Fellowship Years: 1991-1995 University of Illinois at Urbana-Champaign Engineering, University of Michigan University of Wisconsin-Madison Current Status: NASA Ames Research Center University of Utah Mechanical Engineering Current Status: Drupal Developer, Electrical Engineering Kevin Glass Synergistic Data Display Fellowship Years: 1992-1996 Cronen-Townsend Consulting Fellowship Years: 1991-1995 Stephen Fink University of Oregon Bree Aldridge Jaydeep Bardhan Fellowship Years: 1996 Current Status: Engineering Manager, Teradyne Current Status: Senior Staff Engineer, University of California, San Diego Computational Ecology Massachusetts Institute of Technology Massachusetts Institute of Technology Current Status: Graduate Research Assistant, Robert Cruise Intel Corporation Computer Science Fellowship Years: 1996-2000 Systems Biology/Computational Biology Numerical Methods for Molecular Analysis University of Utah Kevin Chu Indiana University Fellowship Years: 1994-1998 Current Status: Scientist, Molecular Science Fellowship Years: 2002-2006 and Design Massachusetts Institute of Technology Computational Physics Mary Dunlop Current Status: IBM Computing Facility, EMSL, Pacific Northwest Current Status: Postdoctoral Research Fellow, Fellowship Years: 2002-2006 Paul Bunch CS&E, Applied Math, Material Science, Fellowship Years: 1997-2001 California Institute of Technology National Laboratory Harvard School of Public Health Current Status: Assistant Professor, Purdue University Computer Vision, Artificial Intelligence Current Status: Department of Defense Bioengineering, Synthetic Biology, Biofuels, Robert Fischer Molecular and Biophysics and Physiology, Chemical Engineering Fellowship Years: 2002-2005 Dynamical Systems Harvard University Larisa Goldmints Erik Allen Rush University Fellowship Years: 1994-1997 Current Status: Research Scientist/ Aron Cummings Fellowship Years: 2002-2006 Security, Privacy, Mobile Agents, Carnegie Mellon University Massachusetts Institute of Technology Current Status: Merck & Co. Inc. Consultant, Serendipity Research; CEO, Arizona State University Current Status: Postdoctoral Scholar, Software Engineering Structural and Computational Mechanics Molecular Simulation Edward Barragy Galapagos Computing, Inc. Nanoscale Electronics Joint BioEnergy Institute Fellowship Years: 1994-1998 Fellowship Years: 1997-2001 Fellowship Years: 2004-2008 University of Texas Jeffrey Butera Fellowship Years: 2003-2007 Current Status: Quant Current Status: GE & Rensselaer Current Status: CTO, Svaya Nanotechnologies Engineering Mechanics North Carolina State University Julianne Chung Current Status: Postdoctoral Researcher, Lewis Dursi Polytechnic Institute Fellowship Years: 1991-1993 Mathematics Emory University Sandia National Laboratories – California University of Chicago Jasmine Foo Marcelo Alvarez Current Status: Intel Corporation Fellowship Years: 1993-1997 Computational Science, Applied Mathematics, Computational Astrophysics, Large-scale Brown University William Gooding University of Texas Current Status: Administrative Computing, Biomedical Imaging Joseph Czyzyk Simulation, Hydrodynamics, Combustion, Applied Mathematics Purdue University Astrophysics William Barry Hampshire College Fellowship Years: 2006-2009 Northwestern University Magneto Hydro Dynamic Fellowship Years: 2004-2008 Chemical Engineering Fellowship Years: 2001-2005 Carnegie Mellon University Current Status: NSF Mathematical Sciences Industrial Engineering and Management Fellowship Years: 1999-2003 Current Status: Postdoctoral Fellow, Fellowship Years: 1991-1994 Current Status: Postdoctoral Researcher, Computational Mechanics Michael Bybee Postdoctoral Research Fellow, University Fellowship Years: 1991-1994 Current Status: Senior Research Sloan Kettering Institute Kavli Institute for Particle Astrophysics Fellowship Years: 1994-1998 University of Illinois at Urbana-Champaign of Maryland Current Status: Business Intelligence Associate, CITA Kristen Grauman and Cosmology (KIPAC), Stanford University Current Status: Assistant Professor, School Hydrodynamic Simulation of Analyst, Central Michigan University Gregory Ford Massachusetts Institute of Technology of Civil Engineering, Asian Institute Colloidal Suspensions Kristine Cochran Research Corporation Ryan Elliott University of Illinois at Urbana-Champaign Computer Vision, Machine Learning Asohan Amarasingham of Technology Fellowship Years: 2004-2008 University of Illinois at Urbana-Champaign University of Michigan Chemical Engineering Fellowship Years: 2001-2005 Brown University Current Status: Gamma Technologies, Inc. Computational Mechanics, Material Modeling, William Daughton Shape Memory Alloys and Active Materials Fellowship Years: 1993 Current Status: Clare Boothe Luce Assistant Theoretical Neuroscience Paul Bauman Fracture Mechanics Massachusetts Institute of Technology Fellowship Years: 2000-2004 Professor, Deartment of Computer Sciences, Fellowship Years: 1998-2002 University of Texas Brandoch Calef Fellowship Years: 2002-2006 Plasma Physics Current Status: Assistant Professor, Oliver Fringer University of Texas at Austin Current Status: Postdoctoral Fellow, Center Multiscale Modeling, Error Estimation, Automatic University of California, Berkeley Current Status: Engineering Researcher, Fellowship Years: 1992-1996 Department of Aerospace Engineering Stanford University for Molecular and Behavioral Neuroscience, Adaptivity, Uncertainty Quantification Imaging Research Oak Ridge National Laboratory Current Status: Staff Scientist, Los Alamos & Mechanics, University of Minnesota Parallel Coastal Ccean Modeling Corey Graves Rutgers University Fellowship Years: 2003-2007 Fellowship Years: 1996-2000 National Laboratory Fellowship Years: 1997-2001 North Carolina State University Current Status: Postdoctoral Researcher, Current Status: Scientist, Boeing Joshua Coe Thomas Epperly Current Status: Assistant Professor, Pervasive Computing/ Image Processing Kristopher Andersen University of Texas at Austin University of Illinois at Urbana-Champaign Gregory Davidson University of Wisconsin Stanford University Fellowship Years: 1996-2000 University of California, Davis Patrick Canupp Chemical Physics, Electronically Excited States, University of Michigan Component Technology for Current Status: Business Owner, Scholars’ Computational Materials Science Martin Bazant Stanford University Monte Carlo Methodology Nuclear Engineering and Computational Science High-Performance Computing Kenneth Gage Advocate; Assistant Professor, North Fellowship Years: 2001-2005 Harvard University Aeronautical and Astronautical Engineering Fellowship Years: 2001-2002 Fellowship Years: 2002-2006 Fellowship Years: 1991-1995 University of Pittsburgh Carolina A&T State University Current Status: Assistant Professor, Applied mathematics, Fluid Mechanics, Fellowship Years: 1991-1995 Current Status: Los Alamos National Current Status: University of Michigan Current Status: Lawrence Livermore Molecular Imaging, Computational Fluid Department of Physics and Astronomy, Electrochemical Systems Current Status: Chief Aerodynamicist, Laboratory National Laboratory Dynamics Design of Artificial Organs Michael Greminger Northern Arizona University Fellowship Years: 1992-1996 Joe Gibbs Racing Jimena Davis Fellowship Years: 1998-2002 University of Minnesota Current Status: Associate Professor, James Comer North Carolina State University Susanne Essig Current Status: Radiology Resident (Research Mechanical Engineering Matthew Anderson Department of Mechanical Engineering, Christopher Carey North Carolina State University Uncertainty Quantification, PBPK Modeling, Massachusetts Institute of Technology Track), Johns Hopkins Medical Institutions, Fellowship Years: 2002-2005 University of Texas Stanford University University of Wisconsin Computational Fluid Mechanics, Risk Assessment Aeronautics and Astronautics, Baltimore, MD Current Status: Seagate Technology Numerical Relativity, Relativistic Magneto Hydro Plasma Physics Fluid Structure Interaction Fellowship Years: 2004-2008 Computational Turbulence Dynamics, High Performance Computing Kathleen Beutel Fellowship Years: 2005-2009 Fellowship Years: 1991-1995 Current Status: Postdoctoral Fellow, Fellowship Years: 1997-2002 Nouvelle Gebhart Noel Gres Fellowship Years: 2000-2004 University of Minnesota Current Status: Scientific Staff, Current Status: Technology Leader, Modeling & US Environmental Protection Agency, University of New Mexico University of Illinois at Urbana-Champaign Current Status: Post-doctorate Researcher, Computational Chemistry MIT Lincoln Laboratories Simulation Department, Procter & Gamble Research Triangle Park, NC Annette Evangelisti Chemistry Electrical Engineering Louisiana State University Fellowship Years: 2009-2010 University of New Mexico Fellowship Years: 2001-2003 Fellowship Years: 1999-2001 Kent Carlson Gavin Conant Mark DiBattista Computational Molecular Biology Deceased Jordan Atlas Bonnie Beyer Florida State University University of New Mexico Columbia University Fellowship Years: 2001-2005 Boyce Griffith Cornell University University of Illinois at Urbana-Champaign Solidification of Cast Metals Molecular Evolution Computational Fluid Dynamics Current Status: Graduate Student, Sommer Gentry New York University Computational Biology Avionics for Business & Regional Aircraft Fellowship Years: 1991-1995 Fellowship Years: 2000-2004 Fellowship Years: 1992-1994 University of New Mexico Massachusetts Institute of Technology Computational Methods for Problems Fellowship Years: 2005-2009 Fellowship Years: 1991-1995 Current Status: Assistant Research Engineer/ Current Status: Assistant Professor of Animal Optimization in Physiology Current Status: Microsoft Corporation Current Status: Technical Project Manager, Adjunct Assistant Professor, University of Iowa Sciences, University of Missouri-Columbia John Dolbow Matt Fago Fellowship Years: 2001-2005 Fellowship Years: 2000-2004 Rockwell Collins Northwestern University California Institute of Technology Current Status: Assistant Professor, Current Status: Medtronic/American Heart Teresa Bailey Nathan Carstens William Conley Computational Methods for Evolving Computational Structural Mechanics Mathematics, U.S. Naval Academy Association Postdoctoral Fellow, Courant Texas A&M University Mary Biddy Massachusetts Institute of Technology Purdue University Discontinuities and Interfaces Fellowship Years: 2000-2003 Institute, New York University Deterministic Transport Theory University of Wisconsin Computational Simulation of BWR Fuel Nonlinear Mechanics of Fellowship Years: 1997-1999 Current Status: Research Scientist, Charles Gerlach Fellowship Years: 2002-2006 Engineering Fellowship Years: 2001-2004 Nano-Mechanical Systems Current Status: Yoh Family Professor, Colorado Springs, CO Northwestern University Eric Grimme Current Status: Lawrence Livermore Fellowship Years: 2002-2006 Current Status: AREVA Fellowship Years: 2003-2008 Duke University Finite Elements, High Strain Rate Solid University of Illinois at Urbana-Champaign National Laboratory Current Status: British Petroleum Current Status: Doctoral Candidate Michael Falk Mechanics Electrical Engineering Edward Chao Laura Dominick University of California, Santa Barbara Fellowship Years: 1995-1999 Fellowship Years: 1994-1997 Edwin Blosch Princeton University Natalie Cookson Florida Atlantic University Stress Driven Materials Processes Including Current Status: Southwest Research Institute Current Status: Intel Corporation University of Florida Computed Tomography/Radiation Therapy University of California, San Diego Computational Electromagnetics/Electromagnetic Fracture, Deformation and Semiconductor Aerospace Engineering Fellowship Years: 1992-1995 Systems Biodynamics and Computational Biology Performance of Materials Crystal Growth Timothy Germann Fellowship Years: 1991-1994 Current Status: Scientist, TomoTherapy Fellowship Years: 2005-2009 Fellowship Years: 1993-1997 Fellowship Years: 1995-1998 Harvard University Current Status: Technical Lead, CFD-FASTRAN Current Status: Large Military Engines Current Status: Associate Professor of Physical Chemistry Division, Pratt & Whitney Materials Science and Engineering, Johns Fellowship Years: 1992-1995 Hopkins University Current Status: Staff Member, Los Alamos National Laboratory P38 DEIXIS 10 DOE CSGF ANNUAL DEIXIS 10 DOE CSGF ANNUAL P39 alumni directory

John Guidi Asegun Henry Nickolas Jovanovic Yury Krongauz David Markowitz Nathaniel Morgan Christopher Oehmen Rick Propp University of Maryland Massachusetts Institute of Technology Yale University Northwestern University Princeton University Georgia Institute of Technology University of Memphis/University of Tennessee, HSC University of California, Berkeley Computer Science Renewable Energy, Atomistic Level Heat Preconditioned Iterative Solution Techniques in Theoretical and Applied Mechanics Computational Neurobiology Computational Fluid Dynamics High Performance Computing in Computational Methods for Flow Through Fellowship Years: 1994-1997 Transfer, First Principles Electronic Boundary Element Analysis Fellowship Years: 1993-1996 Fellowship Years: 2005-2009 Fellowship Years: 2002-2005 Computational Biology Porous Media Current Status: High School Structure Calculations Fellowship Years: 1992-1994 Current Status: BlackRock (financial Current Status: Los Alamos National Fellowship Years: 1999-2003 Fellowship Years: 1993-1996 Mathematics Teacher Fellowship Years: 2005-2009 Current Status: Founding Associate Professor institution), New York City Marcus Martin Laboratory Current Status: Senior Research Scientist, Current Status: Senior Software Current Status: Postdoctoral Researcher, of Systems Engineering, University of University of Minnesota Computational Biology and Bioinformatics Engineer, Workday Brian Gunney Oak Ridge National Laboratory Arkansas at Little Rock Eric Lee Monte Carlo Molecular Simulation, Primarily James Morrow Group, Pacific Northwest National Laboratory University of Michigan Rutgers University Focused on Algorithm Development Carnegie Mellon University Alejandro Quezada CFD, Multi-Physics Simulations, Adaptive Judith Hill Yan Karklin Mechanical Engineering Fellowship Years: 1997-1999 Sensor-Based Control of Robotic Systems Steven Parker University of California, Berkeley Mesh Refinement, Parallel Computing Carnegie Mellon University Carnegie Mellon University Fellowship Years: 1999-2003 Current Status: Director, Useful Bias Incorporated Fellowship Years: 1992-1995 University of Utah Geophysics Fellowship Years: 1993-1996 Computational Fluid Dynamics, Computational Neuroscience Current Status: Engineer, Northrop Current Status: Principal Member of Technical Computational Science Fellowship Years: 1997 Current Status: Lawrence Livermore PDE-Constrained optimization Fellowship Years: 2002-2006 Grumman Corp. Daniel Martin Staff, Sandia National Laboratories – Fellowship Years: 1994-1997 National Laboratory Fellowship Years: 1999-2003 Current Status: Postdoctoral Researcher University of California, Berkeley New Mexico Current Status: Research Assistant Professor Catherine Quist Current Status: Computational Mathematics, Miler Lee Adaptive Mesh Refinement Algorithm and of Computer Science, University of Utah Cornell University Aric Hagberg Oak Ridge National Laboratory Richard Katz University of Pennsylvania Software Development Sarah Moussa Bioinformatics University of Arizona Columbia University Computational Biology Fellowship Years: 1993-1996 University of California, Berkeley Joel Parriott Fellowship Years: 2000-2004 Applied Mathematics Charles Hindman Geodynamics, Coupled Fluid--Solid Dynamics Fellowship Years: 2005-2009 Current Status: Lawrence Berkeley Machine Learning University of Michigan Current Status: Postdoctoral Fellow, University Fellowship Years: 1992-1994 University of Colorado Fellowship Years: 2001-2005 Current Status: Postdoctoral Researcher, National Laboratory Fellowship Years: 2003-2005 Elliptical Galaxies, Computational Fluid of Michigan Cancer Center Current Status: Staff Member, Los Alamos Aerospace Engineering Current Status: Academic Fellow, Department University of Pennsylvania Current Status: Google, Inc. Dynamics, Parallel Computing National Laboratory Fellowship Years: 1999-2003 of Earth Science, University of Oxford Randall McDermott Fellowship Years: 1992-1996 Mala Radhakrishnan Current Status: Air Force Research Laboratory, Seung Lee University of Utah Michael Mysinger Current Status: Program Examiner, Office of Massachusetts Institute of Technology Glenn Hammond Space Vehicles Directorate Benjamin Keen Massachusetts Institute of Technology Numerical Methods for Large-Eddy Simulation Stanford University Management and Budget, Executive Office Computational Drug and Biomolecular Design University of Illinois at Urbana-Champaign University of Michigan Computational Molecular Biology of Turbulent Reacting Flows Chemical Engineering and Quantum Chemistry of the President and Analysis Multiphase Flow and Multicomponent Jeffrey Hittinger Conservation Laws in Complex Geometries Fellowship Years: 2001-2005 Fellowship Years: 2001-2005 Fellowship Years: 1996-2000 Fellowship Years: 2004-2007 Biogeochemical Transport, University of Michigan Fellowship Years: 2000-2004 Current Status: Management Consultant, Current Status: NRC Postdoctoral Researcher, Current Status: Arqule, Inc. Ian Parrish Current Status: Assistant Professor of Parallel Computation Computational Plasma Physics Current Status: IDA Center for Boston Consulting Group Seoul Office NIST, Gaithersburg, MD Princeton University Chemistry, Wellesley College Fellowship Years: 1999-2003 Fellowship Years: 1996-2000 Computing Sciences Heather Netzloff Computational Astrophysics Current Status: Scientist III, Pacific Current Status: Staff Member, Center for Jack Lemmon Matthew McGrath Iowa State University Fellowship Years: 2004-2007 Emma Rainey Northwest National Laboratory Applied Scientific Computing, Lawrence Peter Kekenes-Huskey Georgia Institute of Technology University of Minnesota Quantum/Theoretical/Computational Chemistry Current Status: Chandra Postdoctoral Fellow, California Institute of Technology Livermore National Laboratory California Institute of Technology Mechanical Engineering Monte Carlo Simulations of Fellowship Years: 2000-2004 University of California, Berkeley Planetary Sciences Jeff Hammond Computational Chemistry and Biology Fellowship Years: 1991-1994 Nucleation Phenomena Fellowship Years: 2003-2006 University of Chicago Gordon Hogenson Fellowship Years: 2004-2007 Current Status: Medtronic, Inc. Fellowship Years: 2004-2007 Elijah Newren Tod Pascal Current Status: Arete Associates Computational Chemistry on Supercomputers, University of Washington Current Status: National Science Foundation University of Utah California Institute of Technology Asynchronous Programming Models Physical Chemistry Jeremy Kepner Mary Ann Leung International Research Fellow, Computational Biofluid Dynamics Physical Chemistry Nathan Rau Fellowship Years: 2005-2009 Fellowship Years: 1993-1996 Princeton University University of Washington Helsinki, Finland Fellowship Years: 2001-2005 Fellowship Years: 2003-2007 University of Illinois at Urbana-Champaign Current Status: Director’s Postdoctoral Fellow, Current Status: Technical Writer, Microsoft Computational Cosmology Computational Physical Chemistry Current Status: Postdoctoral Researcher, Civil Engineering Argonne Leadership Computing Facility Fellowship Years: 1993-1996 Fellowship Years: 2001-2005 Richard McLaughlin Sandia National Laboratories – New Mexico Virginia Pasour Fellowship Years: 2000-2001 Daniel Horner Current Status: Senior Technical Staff, Current Status: Program Manager, Princeton University North Carolina State University Current Status: Civil Engineer, Jeff Haney University of California, Berkeley MIT Lincoln Lab Krell Institute Fluid Dynamics Pauline Ng Physical/Biological Modeling, Modeling of Hanson Professional Services Texas A&M University Breakup processes, Quantum Molecular Dynamics Fellowship Years: 1991-1994 University of Washington Epidemiological Dynamics Physical Oceanography Fellowship Years: 2000-2004 David Ketcheson Jeremy Lewi Current Status: Professor of Mathematics, Computational Biology Fellowship Years: 1998-1999 Clifton Richardson Fellowship Years: 1993-1996 Current Status: Research Analyst, Advanced University of Washington Georgia Institute of Technology University of North Carolina, Chapel Hill Fellowship Years: 2000-2002 Current Status: Program Manager, Cornell University Current Status: Computer Programmer, Technology and Systems Analysis Division, Applied Mathematics Neuroengineering Current Status: Informatics Scientist I, Illumina Biomathematics, Army Research Office Physics Dynacon, Inc. Center for Naval Analysis Fellowship Years: 2006-2009 Fellowship Years: 2005-2009 Matthew McNenly Fellowship Years: 1991-1995 Current Status: Assistant Professor, KAUST University of Michigan Diem-Phuong Nguyen Christina Payne Heath Hanshaw William Humphrey Benjamin Lewis Rarefied Gas Dynamics University of Utah Vanderbilt University Christopher Rinderspacher University of Michigan University of Illinois at Urbana-Champaign Sven Khatri Massachusetts Institute of Technology Fellowship Years: 2001-2005 CFD Simulations (Combustion and Reaction) Molecular Dynamics Simulations University of Georgia High Energy Density Physics Physics California Institute of Technology Computational Biology Current Status: Staff, Lawrence Livermore Fellowship Years: 1999-2003 Fellowship Years: 2003-2007 Inverse Design, Quantum Chemistry Fellowship Years: 2001-2005 Fellowship Years: 1992-1994 Electrical Engineering Fellowship Years: 2002-2006 National Laboratory Current Status: Staff, University of Utah Current Status: Process Engineer, URS Fellowship Years: 2001-2005 Current Status: Sandia National Current Status: NumeriX LLC Fellowship Years: 1993-1996 Current Status: Graduate Student, MIT Washington Division Current Status: Army Research Laboratory Laboratories – New Mexico Current Status: Contractor, Honeywell Lisa Mesaros Debra Nielsen Jason Hunt Lars Liden University of Michigan Colorado State University Chris Penland John Rittner Rellen Hardtke University of Michigan Benjamin Kirk Boston University Aerospace Engineering and Scientific Computing Civil Engineering Duke University Northwestern University University of Wisconsin Aerospace Engineering and Scientific Computing University of Texas Biology Fellowship Years: 1991-1995 Fellowship Years: 1992-1996 Computational and Statistical Modeling of Grain Boundary Segregation Particle Astrophysics (Neutrinos from Fellowship Years: 1999-2003 Aerospace Engineering Fellowship Years: 1994-1998 Current Status: Business Manager, Fluent, Inc. Pharmacokinetic/Pharmacodynamic Systems Fellowship Years: 1991-1995 Gamma-Ray Bursts), Gender & Science Current Status: General Dynamics, Advanced Fellowship Years: 2001-2004 Current Status: Chief Technical Officer, Oaz Nir for Biopharma Current Status: Chicago Board Fellowship Years: 1998-2002 Information Systems Current Status: NASA JSC TeachTown, LLC; Software Technology Richard Mills Massachusetts Institute of Technology Fellowship Years: 1993-1997 Options Exchange Current Status: Assistant Professor, Manager, University of Washington College of William and Mary Computational Biology Current Status: Expert Modeler, University of Wisconsin-River Falls E. McKay Hyde Bonnie Kirkpatrick Scientific Computing Fellowship Years: 2006-2009 Pharmacometrics - Modeling & Simulation, Courtney Roby California Institute of Technology University of California, Berkeley Alex Lindblad Fellowship Years: 2001-2004 Novartis Institutes for Biomedical Research University of Colorado Kristi Harris Efficient, High-Order Integral Equation Methods Computer Science University of Washington Current Status: Computational Scientist, Joyce Noah History of Science in the Ancient World University of Maryland, Baltimore County in Computational Electromagnetics Fellowship Years: 2004-2008 Explicit Transient Dynamic Finite Element Oak Ridge National Laboratory Stanford University James Phillips Fellowship Years: 2002-2003 Theoretical Solid State Physics and Acoustics Software Development Theoretical Chemistry University of Illinois at Urbana-Champaign Current Status: Graduate Student, Fellowship Years: 2006-2010 Fellowship Years: 1999-2002 Kevin Kohlstedt Fellowship Years: 2002-2006 Julian Mintseris Fellowship Years: 2001-2003 Parallel Molecular Dynamics Simulation of Large Stanford University Current Status: Analyst, Department Current Status: Assistant Professor, Northwestern University Current Status: Senior Member of Boston University Current Status: Graduate Student, Biomolecular Systems of Defense Computational and Applied Mathematics, Coulomb Interactions in Soft Materials Technical Staff, Sandia National Computational Biology Stanford University Fellowship Years: 1995-1999 David Rogers Rice University Fellowship Years: 2005-2009 Laboratories – California Fellowship Years: 2001-2005 Current Status: Senior Research Programmer, University of Cincinnati Owen Hehmeyer Current Status: Research Fellow, Chemical Current Status: Postdoctoral Fellow, Peter Norgaard University of Illinois Computational Physical Chemistry Princeton University Eugene Ingerman Engineering, University of Michigan Tasha Lopez Harvard Medical School Princeton University Fellowship Years: 2006-2009 Chemical Engineering University of California, Berkeley University of California, Los Angeles Computational Plasma Dynamics Todd Postma Current Status: Postdoctoral Research Fellow, Fellowship Years: 2002-2006 Applied Mathematics / Numerical Methods Justin Koo Chemical Engineering Erik Monsen Fellowship Years: 2005-2009 University of California, Berkeley Sandia National Laboratories – New Mexico Current Status: ExxonMobil Upstream Fellowship Years: 1997-2001 University of Michigan Fellowship Years: 2000-2001 Stanford University Nuclear Engineering, Computational Neutronics Research Corporation Current Status: Senior Scientist, Electric Propulsion Modeling and Simulation Current Status: Sales Specialist, Entrepreneurship, Organization Development, Catherine Norman Fellowship Years: 1994-1998 David Ropp General Electric Fellowship Years: 2000-2004 IBM Cognos Now! and Change Northwestern University Current Status: Director of Engineering, Totality University of Arizona Eric Held Current Status: Engineer, Air Force Research Fellowship Years: 1991-1993 Computational Fluid Dynamics Adaptive Radar Array Processing University of Wisconsin Ahmed Ismail Laboratory, Edwards AFB, CA Christie Lundy Current Status: Senior Research Fellow, Fellowship Years: 2000-2004 David Potere Fellowship Years: 1992-1995 Plasma/Fusion Theory Massachusetts Institute of Technology Missouri University of Science and Technology Max Planck Institute of Economics, Current Status: Research Analyst, Center for Princeton University Current Status: Senior Scientist, SAIC in Fellowship Years: 1995-1999 Molecular Simulations and Multiscale Modeling Michael Kowalok Physics Jena, Germany Naval Analyses Demography / Remote Sensing Arlington, VA Current Status: Associate Professor, Physics Fellowship Years: 2000-2004 University of Wisconsin Fellowship Years: 1991-1994 Fellowship Years: 2004-2008 Department, Utah State University Current Status: Junior Professor, Mechanical Monte Carlo Methods for Radiation Therapy Current Status: Missouri State Government Brian Moore Gregory Novak Current Status: Consultant, Robin Rosenfeld Engineering, RWTH Aachen University Treatment Planning North Carolina State University University of California, Santa Cruz Boston Consulting Group Scripps Research Institute Fellowship Years: 2000-2004 William Marganski Computational Simulation of Nuclear & Theoretical Astrophysics Computational Biophysics Amber Jackson Current Status: Medical Physicist, Waukesha Boston University Thermal-Hydraulic Processes in Boiling Fellowship Years: 2002-2006 Fellowship Years: 1996-1997 University of North Carolina Memorial Hospital, Waukesha, WI Computational Biology, Imaging, Modeling Water Nuclear Reactors Current Status: Postdoctoral Fellow, Current Status: ActiveSight Applied Mathematics Fellowship Years: 1998-2002 Fellowship Years: 1992-1995 Princeton University Fellowship Years: 2004-2008 Current Status: Research Scientist, Systems Current Status: Manager, Nuclear & Biology Department, Harvard Medical School Thermal-Hydraulic Methods, Global P40 DEIXIS 10 DOE CSGF ANNUAL Nuclear Fuel DEIXIS 10 DOE CSGF ANNUAL P41 alumni directory

Mark Rudner Scott Stanley Michael Veilleux Eric Williford Massachusetts Institute of Technology University of California, San Diego Cornell University Florida State University Theoretical Condensed Matter Physics Fluid Mechanics, Turbulence Modelling, Data Computational Fracture Mechanics Meteorology Fellowship Years: 2003-2007 Handling for Large Datasets, Databases, Fellowship Years: 2004-2008 Fellowship Years: 1993-1996 Current Status: Postdoctoral Fellow, Search Engines Current Status: Weather Predict, Inc. Harvard University Fellowship Years: 1994 Rajesh Venkataramani Current Status: Hewlett-Packard Massachusetts Institute of Technology Michael Wolf David Schmidt Chemical Engineering University of Illinois at Urbana-Champaign University of Illinois at Urbana-Champaign Samuel Stechmann Fellowship Years: 1995-1999 Parallel and Combinatorial Communications New York University Current Status: Goldman Sachs Scientific Computing Fellowship Years: 2002-2006 Applied Math, Atmospheric Science Fellowship Years: 2003-2007 Current Status: Epic Systems Fellowship Years: 2003-2007 Stephen Vinay III Current Status: Postdoctoral Researcher, Current Status: Postdoctoral Researcher, Carnegie Mellon University Sandia National Laboratories – Samuel Schofield University of California, Los Angeles Application of Smoothed Particle New Mexico University of Arizona Hydrodynamics to Problems in Computational Fluid Dynamics, Hydrodynamic James Strzelec Fluid Mechanics Matthew Wolinsky Stability, Interface Methods Stanford University Fellowship Years: 1998-2000 Duke University Fellowship Years: 2001-2005 Computational Mathematics Current Status: Manager, Noise Testing & Computational Geoscience Current Status: Scientist, T-5, Los Alamos Fellowship Years: 1992-1994 Analysis, Bettis Atomic Power Laboratory Fellowship Years: 2001-2005 National Laboratory Current Status: Research Scientist, Shell Rajeev Surati Joshua Waterfall International Exploration and Production Christopher Schroeder Massachusetts Institute of Technology Cornell University University of California, San Diego Electrical Engineering and Computer Science Molecular Biology Allan Wollaber Theoretical Particle Physics, Lattice Fellowship Years: 1995-1997 Fellowship Years: 2002-2006 University of Michigan Gauge Theory Current Status: Scalable Current Status: Postdoctoral Researcher, Nuclear Engineering Fellowship Years: 2005-2009 Display Technologies Cornell University Fellowship Years: 2004-2008 Current Status: Postdoctoral Researcher Current Status: Los Alamos Laura Swiler Philip Weeber National Laboratory Robert Sedgewick Carnegie Mellon University University of North Carolina University of California, Santa Barbara Reliability Analysis, Prognostics, Interest Rate Derivative Consulting Brandon Wood Computational Biology Network Vulnerability Analysis, Fellowship Years: 1994-1996 Massachusetts Institute of Technology Fellowship Years: 2000-2003 Combinatorial Optimization Current Status: Chatham Financial Computational Materials Science Current Status: Research Associate, Fellowship Years: 1992-1994 Fellowship Years: 2003-2007 Carnegie Mellon University Current Status: Principal Member of Adam Weller Current Status: NSF IRPF Postdoctoral Technical Staff, Sandia National Princeton University fellow, JNCASR, Bangalore, India Marc Serre Laboratory – New Mexico Chemical Engineering University of North Carolina Fellowship Years: 2001-2002 Lee Worden Environmental Stochastic Modeling Shilpa Talwar Princeton University and Mapping Stanford University Gregory Whiffen Applied Mathematics Fellowship Years: 1996-1999 Array Signal Processing Cornell University Fellowship Years: 1998-2002 Current Status: Assistant Professor, Fellowship Years: 1992-1994 Deep Space Trajectory and Mission Design, Current Status: S. V. Ciriacy-Wantrup University of North Carolina Current Status: Senior Research Scientist, Low-Thrust Mission Design, and Nonlinear Postdoctoral Fellow, Environmental Intel Corporation Optimal Control Studies, Policy and Management, Jason Sese Fellowship Years: 1991-1995 University of California, Berkeley Stanford University Brian Taylor Current Status: Senior Engineer, Outer Hydrogen Storage on Carbon Nanotubes University of Illinois at Urbana-Champaign Planets Mission Design Group, NASA Jet Michael Wu Fellowship Years: 2003-2005 Engineering Mechanics Propulsion Laboratory, Pasadena, CA University of California, Berkeley Current Status: Chemical Engineer, Fellowship Years: 2003-2007 Social Analytics, Graph & Social Network Environmental Consulting Company Collin Wick Analysis, Predictive Modeling, High Dim Mayya Tokman University of Minnesota Data Visualization Elsie Simpson Pierce California Institute of Technology Computational Chemistry Fellowship Years: 2002-2006 University of Illinois at Urbana-Champaign Numerical Methods, Scientific Computing Fellowship Years: 2000-2003 Current Status: Principal Scientist of Nuclear Engineering Fellowship Years: 1996-2000 Current Status: Assistant Professor, Analytics, Lithium Technologies Fellowship Years: 1991-1993 Current Status: Assistant Professor, Louisiana Tech University Current Status: Computer Scientist, University of California, Merced Pete Wyckoff Lawrence Livermore National Laboratory James Wiggs Massachusetts Institute of Technology William Triffo University of Washington Parallel Architectures and Amoolya Singh Rice University Physical Chemistry Distributed Networks University of California, Berkeley Biophysical Imaging, 3D Electron Microscopy Fellowship Years: 1991-1994 Fellowship Years: 1992-1995 Dynamics and Evolution of Stress Fellowship Years: 2003-2007 Current Status: Novum Millennium Organization Current Status: Research Scientist, Ohio Response Networks Current Status: Finishing Ph.D. Supercomputer Center Fellowship Years: 2002-2006 Stefan Wild Current Status: Computational and Mario Trujillo Cornell University Charles Zeeb Life Sciences Postdoctoral Fellow, University of Illinois at Urbana-Champaign Operations Research Colorado State University Emory University Two-Phase Flow, Computational Fluid Fellowship Years: 2005-2008 Mechanical Engineering Mechanics, and Atomization Phenomena Current Status: Director’s Postdoctoral Fellowship Years: 1993-1997 Melinda Sirman Fellowship Years: 1997-2000 Fellow, Mathematics & Computer Science Deceased University of Texas Current Status: Penn State University Division, Argonne National Laboratory Engineering Mechanics Etay Ziv Fellowship Years: 1994-1996 Obioma Uche Jon Wilkening Columbia University Princeton University University of California, Berkeley Computational Biology Steven Smith Molecular Simulation, Statistical Mechanics Numerical Analysis, Computational Physics, Fellowship Years: 2004-2008 North Carolina State University Fellowship Years: 2002-2006 PDE, Scientific Computing Chemical Engineering Current Status: Postdoctoral Fellow, Sandia Fellowship Years: 1997-2001 Scott Zoldi Fellowship Years: 1992-1994 National Laboratories – California Current Status: Assistant Professor, Duke University Current Status: Invista University of California, Berkeley Analytical Modeling Anton Van der Ven Fellowship Years: 1996-1998 Eric Sorin Massachusetts Institute of Technology Glenn Williams Current Status: Senior Director - Analytic Stanford University First Principles Modeling of Thermodynamic University of North Carolina Science, Fair Isaac Corporation Simulational Studies of Biomolecular and Kinetic Properties of Solids Applied and Computational Mathematics, Assembly and Conformational Dynamics Fellowship Years: 1996-2000 Computational Biology, John ZuHone Fellowship Years: 2002-2004 Current Status: Assistant Professor, Environmental Modeling University of Chicago Current Status: Assistant Professor of Department of Materials Science, Fellowship Years: 1993-1996 Astrophysics Computational & Physical Chemistry, University of Michigan Current Status: Assistant Professor, Fellowship Years: 2004-2008 California State University, Long Beach Department of Mathematics and Current Status: Harvard-Smithsonian Statistics, Old Dominion University Center for Astrophysics

P42 DEIXIS 10 DOE CSGF ANNUAL The Krell Institute Funded by: 1609 Golden Aspen Drive, Suite 101 Office of Science and Ames, IA 50010 National Nuclear Security (515) 956-3696 Administration’s Office of www.krellinst.org/csgf Defense Programs