Computational Science Graduate Fellowships
By Jon Bashor

Four graduate students who are part of a DOE fellowship program took a detour from their main areas of research this summer to spend three months honing their computational science skills with scientists at Berkeley Lab.

As participants in DOE’s Computational Science Graduate Fellowship program, Catherine Grasso, Ben Keen, Heather Netzloff and Catherine Norman are required to do a practicum at a national laboratory. They all chose Berkeley Lab, though for different reasons.

For example, Ben Keen attended a talk by Phil Colella, leader of NERSC’s Applied Numerical Algorithms Group, at the University of Michigan, where he is studying mathematics and scientific computation. Keen looked at both Lawrence Berkeley and Lawrence Livermore as candidate sites for his practicum, and settled on Berkeley Lab. He has spent the summer working with Colella’s group working on the Chombo library, an adaptive mesh refinement (AMR) program developed by the group.

AMR uses grid boxes of varying sizes to focus a computer’s power on the most interesting part of a problem, such as a moving flame front or mixing of gases. Keen’s interested in expanding the capability of AMR codes to efficiently handle embedded boundaries, so that problems with fixed objects, such as air flowing over an aircraft wing, can be addressed. This summer, he implemented an algorithm using the new EBChombo framework for embedded boundaries and AMR, and then testing the application. His code was one of the first flow calculation using the EBChombo extension to the Chombo toolkit.

“It’s been very useful,” Keen said. “I’m definitely going to take away the EB Chombo package. There are a lot of AMR codes out there, but having a chance to sit down and work with the developer of this code is a great opportunity.

“It’s educational to work in this environment because you acquire a point of view on what’s important and how to address a problem,” said Keen, whose university research is in the area of developing tools to study moving boundaries.

Catherine Norman, who is studying applied math at Northwestern University, also spent the summer here working on AMR with Ann Almgren of the Center for Computational Sciences and Engineering. Her worked involved adding level set methods to an existing AMR library for studying fluid dynamics problems. Berkeley Lab’s Mathematics Department is a leader in level sets research and that reputation attracted Norman’s attention.

“It’s been a very good experience at the Lab, seeing that there are communities doing research outside the universities,” she said, adding that she expects to continue working with her newfound colleagues her after she returns to school.. Here you have people of all different backgrounds working together – bringing people together like that is something that makes the Lab unique.”

One of the high points of her summer was helping a friend who lives here and is a certified pyrotechnician set up the fireworks for the Fourth of July celebration in Yountville.

Heather Netzloff, who was born in North Dakota and studies physical chemistry at the Iowa State University, said she found the East Bay quite a change of scenery. "Berkeley is definitely not Iowa."

Netzloff's advisor, Mark Gordon, is a collaborator of Teresa Head-Gordon in the Physical Biosciences Division, and Netzloff spent the summer in Head-Gordon's lab. She parallelized several methods of the Ewald summation technique to treat long-range forces when doing molecular dyanamics simulations.

"It was very useful. Even though it was not exactly my thesis work, it is related in many ways. The practicum also gave me a chance to meet more people that I can talk to about my research--networking is always good," she said. "I will also bring some good ideas back to Iowa from Berkeley."

Her research in Iowa covers a variety of problems in quantum/computational chemistry. She is currently coupling the effective fragment potential method for solvation with molecular dynamics in GAMESS, a quantum chemistry calculation program supported by her group, to study the condensed phase and effects of solvation in chemical reactions. Another project looks at the the theoretical study of high energy density materials, a topic of interest to NASA.

Catherine Grasso's practicum brought her west from Cornell University, where she studies computational biology, and landed her in Stephen Holbrook's lab, where she helped look for RNA genes in yeast. Unlike other genes, RNA genes have no start or stop delimiters, which makes them hard to find. But RNA genes are of great interest because many scientists now believe that the most basic forms of life started as RNA, with proteins and DNA evolving from them. Because life evolves through reappropriation of proteins and RNAs, rather than through building them from scratch, Grasso said, RNAs have continued to play a significant role in cell function.

"We're just starting to get a picture of how extensively RNA is involved in all living things. Recent attempts to find RNA genes, such as those undertaken in the Holbrook Lab, suggest that there are many more than previously imagined." she said.

Grasso's work focuses on using what is known about RNA folding to develop machine learning techniques to search for patterns in the Yeast Genome which have characteristics of patterns found in known RNA genes. By developing general computer codes to search for those patterns, scientists can search for RNA genes in various organisms. "It's a cool problem from a computational biology perspective, because it combines the problem of searching for meaningful motifs in a very large genome, with the problem of accurately modeling the process by which RNA folds." Grasso said.

About the Computational Science Graduate Fellowship Program

The Department of Energy established its Computational Science Graduate Fellowship program 10 years ago to recruit and support graduate students pursuing studies in areas important to DOE’s mission.

Students are selected through a very competitive process and are supported for four years in graduate school. As part of the program, they are required to spend a three-month practicum at a national laboratory. The Lab’s Computing Sciences organization is looking to increase the number of students who do their practicum here and staff have been actively recruiting them. Lab scientists interested in learning more about the participants can view summaries of their research areas or contact Jon Bashor at JBashor@lbl.gov.

The Computational Science Graduate Fellowship program is administered for DOE by the Krell Institute in Ames, Iowa. For more information, go to the Web site.

DOE CSGF@LBNL Home