Simulating the Electromagnetic World
MUCH has been written in Science & Technology Review and
other publications about the remarkable accomplishments of simulation as a full partner with theory and experiment, where they form a threefold foundation for scientific advancement. The steady progression of high-performance computing at Lawrence Livermore has helped ensure our leadership in simulating experiments, often of phenomena for which experimental testing is not an alternative. For example, supercomputer simulations supported by more economical small-scale experiments have replaced underground nuclear testing to help maintain a safe, secure, and reliable U.S. nuclear stockpile.
In the biological sciences, advanced simulations are replacing some aspects of clinical laboratory research to help reveal the mechanisms of pathogenicity. Physical science simulations are allowing us to test ideas in their early stages in place of building expensive hardware prototypes and then iterating experimentally. In chemical research, simulations are revealing how molecules bond to each other and the pathways to new nanomaterials.
Electromagnetic phenomena are ubiquitous throughout the Laboratory’s mission, spanning problems from optical regimes to classic microwave and radio-frequency research to the static fields associated with fixed magnets. The accuracy needed for simulations of these problems poses extreme challenges. As described in the article A Code to Model Electromagnetic Phenomena, a Livermore code called EMSolve allows us to thoroughly model the electric and magnetic fields of devices ranging from tiny integrated circuits to entire buildings. EMSolve has become increasingly more accurate and efficient over the past decade, thanks to the combined efforts of the Engineering Directorate’s Dan White, who is its chief designer; fellow simulation experts; and graduate students.
A premier code, EMSolve has influenced the direction of electromagnetic simulations in government, academia, and industry. It not only simulates complex environments that other codes can’t but also offers a platform to couple other kinds of physics that bring intractable problems within reach of our supercomputers. The code is unusually flexible in applying the underlying equations that govern its calculations, allowing the user to perform what-if scenarios across timescales ranging from billionths of a second to tens of seconds.
One reason EMSolve works so effectively is that it was “born parallel.” The code was specifically developed to run on the parallel supercomputers located at Livermore and other national laboratories and major research centers. A relatively small number of people are expert at writing complex parallel codes. White and his codevelopers understand the architecture of parallel computers and know how to use the tens of thousands of microprocessors powering these machines. This talent typifies the Laboratory’s unique contributions to our nation’s technological capability arising from the synergy between collocated high-performance computing specialists, computer scientists, mathematicians, physicists, and computational engineers.
EMSolve developers are currently extending the code to allow it to attack a broader class of problems. They are coupling EMSolve with thermal, structural, and hydrodynamics codes as well as developing the capability to determine electromagnetic properties from a quantum-mechanical perspective. Today, joint projects between computational engineers and computer scientists are focusing on how to adapt the code to the next generation of supercomputers, which will have 10 to 100 times more microprocessors than today’s versions.
Our goal is to pose realistic questions concerning complex situations and then simulate the results with the accuracy needed to make meaningful decisions. We will face future challenges for simulating electromagnetic fields, such as those expected in new generations of integrated electronic and optical circuits. Detecting radio signals propagated in caves, tunnels, and “urban canyons” will also be important. In addition, EMSolve will be used to analyze the complex electromagnetic environment of giant lasers such as the National Ignition Facility. We also expect EMSolve to be an extremely powerful tool for quickly interpreting radar signals bounced off high-speed aircraft. Because of the dedicated work of EMSolve developers and their colleagues, we?are steadily moving toward making that goal a reality.