Projects in Scientific Computing

Download the PDF version of this article as it appeared in Projects in Scientific Computing, 2004.

Convection in Giant Planets


Image of convection in Jupiter's interior.

Click Image to Enlarge

Scientists know that much as Earth generates its magnetic field from the geodynamo — turbulent, roiling motion of fluids in the planet’s interior, giant planets like Jupiter and Saturn also have a magnetic field produced by turbulent fluid convection in the interior. For these giant planets, furthermore, this interior fluid motion is also related to zones of differential rotation on the planet’s surface. Although scientists have studied these phenomena for decades, we still lack detailed understanding.

Gary Glatzmaier of the University of California, Santa Cruz developed the first computational model of Earth’s geodynamo that evolves on its own — self consistently. In 1995, running at PSC, this model produced the first simulation of convection and magnetic-field generation in the Earth’s fluid core, and the first simulated magnetic-field reversal, a phenomenon that has happened many times over Earth’s geological history.

In recent work with LeMieux, Glatzmaier and graduate student Martha Evonuk simulated rotating convection in Jupiter’s interior. This snapshot from their simulation represents temperature on a slice through the planet (brightness corresponding to temperature). It shows small-scale turbulent convection within a large-scale spiral pattern. “Longitudinal flow near the outer boundary is oppositely directed to that near the inner boundary,” notes Glatzmaier, “so the spiral continues to wind up.”

Understanding Metalloenzymes

		proteolytica aminopeptidase.

Click Image to Enlarge

A wide variety of proteins in humans and other mammals use metal ions within the body, such as zinc or manganese, to catalyze biochemical reactions. These metalloenzymes perform many different roles, including DNA repair, hormone regulation and tissue repair. Some of them are involved in carcinogenesis, and some cancer drugs now in use inhibit the action of these enzymes as a way to retard or reverse the growth of cancerous cells. More detailed understanding of how these metalloenzymes function can translate into the design of better chemotherapeutic drugs.

The research group led by Michael Klein at the University of Pennsylvania carries out wide-ranging studies in materials science and biochemistry, and metalloenzymes have been the focus of recent PSC computations with LeMieux. The active sites of these enzymes — where the catalytic reaction takes place — contain one or more metal ions, held in place by nearby amino acids. During a reaction, many changes take place in the electronic structure within and near the active site, which can be explored only with quantum-mechanical simulations. These demanding calculations have become possible only recently, with the advent of terascale-class systems such as LeMieux.


This graphic depicts simulation results from an enzyme called Aeromonas proteolytica aminopeptidase (AAP). Involved in the final stages of protein manufacture, AAP is representative of other similar metalloenzymes. The ball-and-stick model represents its active-site structure (H-white, O-red, N-blue, C-green, Zn-lavender). The catalytic action at the bimetal core (two Zincs) depends on the exact positioning of amino acids closest to the metals (first shell amino acids) and their hydrogen-bonding pattern with nearby (second shell) amino acids.

The cross-sectional plane maps electron density as color contours (purple increasing through blue, green, yellow, red). This plane showcases the strong hydrogen bond between a first and second shell amino acid (upper right center, contours that touch) thought to be important in the fine- tuning of events at the core. Electronic data from these calculations is of great value in experimental efforts to design new therapeutic drugs.

Water’s Magic Number


Water. Where would we be without it? You might think science would know all there is to know about this ubiquitous substance, without which our planet couldn't sustain life. But it's not so simple. School children know the basic formula, but H2O holds many mysteries, not least among them how multiple H2Os join together to form molecular clusters.

Image of water's magic
                  number cluster.

Click Image to Enlarge

One of the more enduring mysteries has been the structure of water’s “magic number cluster.” Mass spectrometry shows that a cluster of 21 water molecules — with one extra proton (H+) — is much more stable than clusters with either 20 or 22 water molecules. “There’s something imparting special stability,” says Ken Jordan, professor and chair of the University of Pittsburgh Department of Chemistry, “and that’s often associated with a special geometrical arrangement.”

Studies over the past 30 years have postulated a dodecahedron, a cage of 20 water molecules, with an H2O in the middle. But where’s the extra proton? Does it go with the central H2O or on the cluster surface? This question has persisted. Using LeMieux, Jordan did calculations to complement laboratory teams at Yale and the University of Georgia. Their collaborative findings appear to settle the question.

As depicted here, water’s magic number cluster is a dodecahedron with the proton on the surface. The molecules are bound together by hydrogen bonds (dotted lines). One H2O (purple) is in the center of the cage, and the excess proton is associated with an H3O+ ion on the surface (blue). This finding, reported in Science (May 21), has stirred much debate. “There’s still a question about how fast the proton can move around the surface,” says Jordan. “It’s not a finished story.”

Recipes for Metallic Glass


A new kind of steel that’s actually a glass is three times as strong as normal steel, resists rust and is nonmagnetic. It may someday be used in cars and buildings, and its magnetic properties offer the prospect of submarines and ships with hulls safe from mines that rely on magnetic detonation.

Research teams at the University of Virginia and Oak Ridge National Laboratory this summer announced such a steel — called a metallic glass or amorphous metal. While their work relied on costly, time-consuming trial-and-error methods of melting and casting alloys, Carnegie Mellon physicist Michael Widom has collaborated with PSC physicist Yang Wang to create a series of recipes for mixing and cooking elements to create many other metallic glasses.

Image of normal metal. Image of metallic glass.

Click Images to Enlarge

As depicted in the two contrasting graphics, in a normal metal, atoms of two different elements (red and blue) arrange themselves in a regular pattern, a crystalline lattice structure. Widom's computations show, however, that adding a small amount of a large atom, yttrium (yellow), disrupts the crystalline lattice, allowing the irregular metallic glass structure to form.

With support from the Defense Advanced Research Projects Agency, Widom and Wang have used computational methods to investigate the composition and properties of various metallic glasses. Widom's computations confirmed the structure of the new amorphous steel and produced recipes for 2,000 other iron-based alloys.

Pipelines to the Stars

Astronomy faces a data avalanche. Breakthroughs in telescope, detector, and computer technology have allowed astronomical surveys to produce terabytes upon terabytes of data about the sky around us. This vast data spans the range of the electromagnetic spectrum — from x-rays through ultraviolet and visible wavelengths to the cosmic microwave background. The next decade will bring several new surveys, with even more data. How can astronomers and astrophysicists keep up with analyzing and drawing insight from this massive trove of information?

Image of normal metal.
Click Images to Enlarge

M78 Nebula in the Orion constellation, from the Sloan Digital Sky Survey. Hot young stars in the nebula’s center illuminate the surrounding gas. Further out, dark clouds of dust prevent much of the scattered light from reaching us, creating a complex pattern of light and shadow. (Image courtesy of SDSCC)


The National Virtual Observatory (NVO) is a major new initiative in astrophysics, sponsored through an NSF Information Technology Research project, together with related international efforts. The objective is to interlace this data across the electromagnetic spectrum and to provide tools to explore and extract from it usefully. Only by collating data from multiple sources can science realize the full potential of this information.

PSC astrophysicist Jeff Gardner collaborates with Andrew Connolly, University of Pittsburgh, Roy Williams, Caltech, and others on a major initiative to combine NVO with the NSF TeraGrid. Still in preliminary stages, this project aims to create an NVO Testbed, through which the massive data can be made available to supercomputing. Ultimately, the researchers expect to test representative astrophysics applications with NVO data and to encourage new ways to enlist supercomputing to extract knowledge from this invaluable data.

Signals for Cell Growth


A family of proteins known as Src tyrosine kinases plays a key role in receiving and passing on a flurry of biochemical messages that regulate the growth of cells. They are active or inactive depending on a biochemical switching mechanism that changes the protein structure. Things can go badly awry, however, when this protein becomes overactive. Mutated versions of Src that send an unwavering signal for cell growth are often found in human cancers.

Scientists are searching for drugs to stop these rogue kinase proteins, and with that underlying objective, biophysicists Benoit Roux and Nilesh Banavali of Weill Medical College of Cornell University are using LeMieux to study Src. While the structure is known, this static picture doesn't show how flexible regions of the protein move and change as the protein switches from the inactive to active state.

Image of activation 
		loop of Src. Image of activation 
		loop of Src. Image of activation 
		loop of Src. Image of activation 
		loop of Src. Image of activation 
		loop of Src.
Click Images to Enlarge

Roux's research team uses molecular dynamics, a computational approach that tracks the protein's movement atom-by-atom over time, and he is looking in particular at the changes associated with opening the activation loop of Src. In this sequence of frames from a simulation of the Src kinase domain over a period of 14 nanoseconds, the activation loop (red) — with an associated tyrosine amino acid — moves from a closed (inactive) to open (active) state.