Environmentally clean, affordable power from fossil fuel is the goal of research at the National Energy Technology Laboratory, and one of the most promising technologies is coal gasification. Processes to convert coal with 100-percent efficiency into clean-burning natural gas exist. The challenge is to do it in a commercially viable way at industrial scale.
In a program of collaboration between field-test experiments and computer simulation, NETL researchers are working to solve the design problems. NETL consulting engineers Chris Guenther and Madhava Syamlal and U.S. Department of Energy project manager James Longanbach are collaborating with on-site engineers to simulate coal gasification in a plant-sized Kellogg Brown & Root, Inc. fluidized-bed reactor.
SIMULATIONS CAN TEST DESIGN CHANGES AT FAR LESS COST THAN EXPERIMENTS
Using LeMieux, they have carried out an extensive series of 3D simulations with data from the KBR facility. The goal is a realistic computational model, which will make it possible to test design changes at a fraction of the cost of scale-model experiments. The recent simulations represent major progress in this direction. “This study is unique,” says Guenther, “in the scale of our 3D simulations that include both chemistry and heat transfer. We’ve obtained excellent agreement with the experimental data. These computations could not have been attempted in a reasonable amount of time without access to LeMieux.”
MAPPING THE RIDGES AND VALLEYS OF PROTEIN ENERGY
Before it can carry out its biological role, a protein must transform from its newborn form as a droopy string of amino acids into just the right folded shape. How it does this — in seconds or less — is of great interest to molecular biologists.
“We want to understand the folding process,” says Carlos Simmerling of SUNY at Stony Brook, “not just the final structure, but how it gets there and what can go wrong, and how this may be related to disease.” In recent work, Simmerling has used LeMieux to determine the “free-energy landscape” — a map of how energy interactions among the atoms change as molecular structure changes — for several small proteins.
Each landscape represents many inter-related computations. To carry out these studies, Simmerling has implemented the software for a method, called “replica exchange,” that exploits massively parallel systems like LeMieux with high efficiency. As many as 50 simulations at a time, running on up to 32 processors each, communicate with each other periodically to map the landscape’s ridges and valleys.
MORE DETAILED LOOKS AT EARTH'S MAGNETIC FIELD
Why is it that instead of fading away, as magnetic fields eventually do when new energy isn’t added, Earth’s magnetic field is still going strong — guiding ships, planes and Boy Scouts — after billions of years? Einstein is said to have considered it one of the most important unsolved problems in physics.
Many scientists believe the answer lies in the turbulent, convective motion of fluids in the Earth’s molten outer core. Buoyancy forces, it’s thought, that arise from the solid, hot inner core help to drive this fluid motion, which in turn generates new magnetic field.
Gary Glatzmaier of the University of California, Santa Cruz and his colleague Paul Roberts of UCLA developed the first computational model of these geodynamic processes that evolves on its own — self consistently. This model has successfully simulated many features of Earth’s magnetic field, including magnetic-field reversal, a recognized phenomenon that has happened many times over Earth’s history. In recent work with LeMieux, Glatzmaier and graduate student Darcy Ogden have investigated heat from radioactive decay in the inner core as a buoyancy source that may help to drive the geodynamo.
Start with a detailed picture of the oldest light in the universe, then simulate 13 billion years of time. The outcome is a distribution of mass over the history of the universe until now. “We can look at that,” says Princeton University astrophysicist Paul Bode, “and say ‘Does this look like the present-day universe?’”
THE MOST DETAILED COSMOLOGICAL SIMULATION EVER
Jeremiah Ostriker of Princeton and Bode used LeMieux to carry out the most detailed simulation of this type ever done, with two-billion particles to represent cosmic mass. They used data from NASA’s WMAP (Wilkinson Microwave Anisotropy Probe) satellite — a detailed map of the cosmic “microwave background” at 379,000 years after the Big Bang — to initialize the simulation, which took five days of computing time on 420 LeMieux processors.
They are now analyzing the data, work that will produce catalogues of halos — from hundreds to thousands of galaxies in clusters held together by gravity — out to the farthest expanding edges of the universe. These catalogues will allow scientists to compare the simulation’s cosmological model — spatially flat, containing both dark energy and dark matter — with astronomical observations, and they will help in planning new observations.
How long does it take for your e-mail to make the journey to its recipient? Seconds, minutes, or — when traffic is clogged — days? How many crossroad stops along the way? How can networks become faster and use available resources more efficiently?
One of the most powerful tools to answer these questions is packet-level network simulation, which uses computational modeling to track individual parcels of bits as they make their way from sender to receiver across complex webs of routers, switches, nodes and other network components. Because this kind of design-and-analysis tool places great demands on computing time, however, its use normally is limited to relatively small networks of a few hundred components.
NETWORK SIMULATIONS A HUNDRED TIMES FASTER WITH LEMIEUX
Using LeMieux, researchers at Georgia Tech have shown that faster, much expanded simulations are feasible. A team of computer scientists — Richard Fujimoto, George Riley, Kalyan Perumalla and Mostafa Ammar — simulated networks of more than five-million elements and modeled more than 106-million packet transmissions in one second of clock time — a hundred times faster than similar prior simulations. “This offers new capability for engineers and scientists,” says Fujimoto, “to design networks with improved speed, reliability and security.”
NEW TECHNOLOGY ALLOWS SCIENTISTS TO "FEEL" PROTEINS
Two things that don’t go together well at border crossings are fast traffic and high security. Biological cells need to be highly selective in what molecules they allow to pass through their walls, and they accomplish this with membrane proteins, which reside in the wall and, like a border guard, control what passes. How they exercise this control, in many cases, however, remains a mystery, the solution of which requires detailed knowledge of how the proteins interact with the molecules that seek to cross.
To help solve these puzzles, scientists led by biophysicist Klaus Schulten at the University of Illinois at Urbana-Champaign have developed a new research technology called “interactive molecular dynamics.” Using scientific visualization and molecular dynamics simulation — which tracks how atoms in a molecule move — along with computer-game technology, a researcher can, in effect, interact with the protein and “feel” mechanical resistance as an ion, for instance, passes through the channel.
In initial work, they’ve used LeMieux to run the simulations. One family of membrane proteins they have studied extensively is aquaporins, an important family in many organisms, including people. Last year, they answered a big question about aquaporins’ selectivity. In a recent study, they explained how some aquaporins transport sugars in a highly selective manner. In ongoing work, they hope to uncover how aquaporins block ions while, at the same time, allowing water to pass freely.
© Pittsburgh Supercomputing Center.