It Takes a Community
December 17, 2007
David Levermore
Talk of the Society
James Crowley
As the Council and Board of Trustees once again take up the idea of a SIAM fellows program, I am reminded of our existing vehicle for recognizing outstanding achievement of individuals: the SIAM major awards program. SIAM awards prizes across a broad spectrum. A prize might recognize contributions to a specific area---control theory, say, or by people at a certain career level---young professionals, say.
I mention this program by way of encouraging readers to be alert to notices of soon-to-be-awarded prizes and to submit nominations when appropriate. Five prizes currently have nomination deadlines that fall between December 15 and February 15. We maintain a list of prizes open for nomination at http://
www.siam.org/prizes/nominations.php. A reminder to all: The success of our awards program hinges on the strength of the nominations we receive.
Current Topics at Washington Agencies
This is an unusual time for programs at U.S. funding agencies, where a buzzword in recent months has been "transformative." Proposals submitted to some programs are to be judged on the basis of their ability to show innovation and new ideas that can radically change a field and its impact. New programs at two agencies illustrate this point.
First, through a Broad Agency Announcement, DARPA (the Defense Advanced Research Projects Agency) unveiled a radically different program in the mathematical sciences, soliciting proposals for research that will advance the solution of any of 23 problems. The problems, referred to as challenges, include such topics as "Beyond Convex Optimization. Can linear algebra be replaced by algebraic geometry in a systematic way?" Each of the 23 challenges is defined in a very brief statement. Not all the challenges are directed solely within the discipline. Challenge 23, for example, asks "What are the fundamental laws of biology?" The idea here is to place biology on the same mathematical footing as physics. Ben Mann is managing this new DARPA program; the funding allocated to the program is not clear.
At NSF, Cyber-enabled Discovery and Innovation (CDI), a new foundation-wide multidisciplinary program, does have clearly specified funding (pending ap-propriations by Congress, of course)---$50 million for fiscal year 2008. The funds are to be dispersed among various NSF directorates, with computer science and the mathematical sciences playing prominent roles (CISE, the Directorate for Computer and Information Science and Engineering, for example, is scheduled to receive $20 million of the $50 million).
CDI seeks multidisciplinary research proposals, according to the call for proposals, within or across three areas:
* From Data to Knowledge: generating knowledge from heterogeneous digital data;
* Understanding Complexity in Natural, Built, and Social Systems: deriving fundamental insights on systems comprising multiple interacting elements; and
* Building Virtual Organizations: enhancing discovery and innovation by bringing people and resources together across institutional, geographical, and cultural boundaries.
It's easy to see threads from these areas that weave across the SIAM community: complexity, systems, data-driven understanding. From the initial $50 million in 2008, this NSF program is to grow by similar increments in subsequent years, meaning that additional opportunities may be available in the future. A strong response from the community is essential. As with many such initiatives, it is the quality of the proposals that will define the program.
Complexity and risk assessment are major themes not only of CDI, but across many areas of science and technology. This was made clear in early November at a meeting of the Board on Mathematical Sciences and their Applications, where these topics are the subjects of recently completed or proposed studies.
BMSA, whose current chair is David Levermore, is affiliated with the National Research Council. Under a restructuring (in which the former BMS got an A), BMSA now has many representatives from areas of science and engineering that use mathematics, in addition to researchers from within the discipline. BMSA conducts independent studies for government agencies.
The authors of a recently released BMSA report, co-sponsored by the Federal Reserve Bank of New York, took on the timely topic of risk in financial systems. Titled Understanding Systemic Risk, the report illustrates the role of complexity in risk assessment. Borrowing concepts from ecology, engineering, and dynamical systems, the authors investigated various aspects of dynamics in financial markets and the risks inherent in them.
Complexity, as portrayed in the report, covers a wide variety of phenomena. Included are systems with large numbers of heterogeneous components, as well as systems of any type that have complex dynamics. Such systems can exhibit synchrony, or behavior can be coordinated within a system; all of these systems can be characterized by rapid and large transitions from one state to another. It is the latter context, with the potential for large and catastrophic shifts, especially within the banking system, that is the report's concern.
Another report, Mathematical Research Challenges in Optimization of Complex Systems, this one from the Department of Energy, also targets the themes of complex systems and risk. Based on a workshop chaired by Bruce Hendrickson and Margaret Wright, the report looks at four specific applications of interest to DOE, including the power grid. The systems considered are not traditional systems based on physics models, but rather heterogeneous coupled systems that sometimes interact in nonlinear fashion. One question considered is how an analysis of risk can be incorporated into models of complex systems.
Computational science and engineering is another important research area currently in the spotlight (and not only at SIAM, where SIAG/CS&E is by far the largest of the activity groups). The goal of CS&E---a blend of applied mathematics (modeling and computational methods), computer science (software, tools, and environments), and specific application domains---is to solve challenging real-world problems using appropriate (normally high-end) computing. Multiscale phenomena and processes are characteristics of many of the most challenging problems that arise in CS&E.
Modeling and simulation of magnetically confined fusion is one such challenge. In this process plasma---matter in which the electrons have been separated from the atoms---is confined by an intense magnetic field. The aim is to get the nuclei of heavy isotopes of hydrogen (deuterium and tritium) to fuse, producing helium and releasing usable energy. (There are other ways to confine plasma to this end: Stars like ours use self-gravity, and nuclear weapons use inertia, either from lasers or from a primary nuclear fission reaction in a spherical shell outside the hydrogen fuel.)
A recent report from DOE, Workshop on the Fusion Simulation Project, describes the computational science needs underlying ITER---the International Thermonuclear Experimental Reactor www.iter.org ---a $12 billion magnetically confined fusion experiment currently under construction in Cadaraches, France. The Fusion Simulation Project, which has a proposed annual budget of $25 million over 15 years, will provide predictive capability for plasma physicists designing experiments to be performed in the ITER tokamak facility, which should lead, in turn, to demonstration reactors for the electric power industries in the dozens of participating countries. Each experimental "shot" performed in ITER (a discharge of energy lasting on the order of minutes) will cost approximately $1 million. (This amount reflects both the amortization of the ITER facility over a few decades of experiments and the recurring operating costs and salaries of the physicists.) Fusion plasmas are susceptible to a variety of instabilities that could result in unintended release of the energy stored in the confining magnetic fields in the form of particles, radiation, and heat that could damage the reactor. Given the danger and expense of each experiment, credentials earned in advance by predictive simulations are likely to be required.
The workshop, chaired by Arnold Kritz, a plasma physicist from Lehigh University, and David Keyes, a computational mathematician from Columbia University, considered the algorithmic and software developments, and the computational resources needed to confer on U.S. physicists the credibility to successfully exploit the national investment in ITER, which will exceed $1 billion. The re-entry of the U.S. into the ITER project was itself motivated by simulations, which, though not yet predictive with quantifiable uncertainty, indicated that the notorious instabilities of the plasma torus could likely be controlled. (Keyes, currently SIAM's vice president at large, is the 2007 recipient of the Sidney Fernbach Award, given by IEEE Computer Society for innovative uses of high-performance computing in problem solving.)
David Keyes
Simulation of fusion energy is inherently multiscale. Extrapolating from a na�ve approach of uniformly refined gridding and explicit timestepping techniques, an estimated 12 additional orders of magnitude in computing capacity will be required for adequate resolution of issues arising at ITER scales. Over the next ten years, at which point ITER should be operational, only a few of these 12 orders of magnitude will come from improvements in hardware, with the most optimistic extrapolations of Moore's law for silicon-based microprocessors and for multithreaded parallel computers. Keyes has argued that algorithmic developments, both in modeling and in numerical methods, are also necessary. Historically, for scientific kernels, such as solving Poisson's equation or computing the forces on interacting particles, algorithmic developments and hardware improvements have contributed equally to advances in computational capability.
On the SIAM Journal Front
We continue to improve access to SIAM journals. With the move of electronic delivery of our journals to AIP's Scitation platform, SIAM is able to take advantage of innovations in electronic delivery, such as RSS feeds, while maintaining a high quality of service. Managing our transition to Scitation is David Marshall, who joined the staff six months ago as SIAM publisher. Working closely with the SIAM vice president for publications, Tim Kelley of North Carolina State University, Marshall manages the SIAM journal and book programs.
We are always seeking ways to improve access to papers in our journals. To mention a recent example, we have joined with the IEEE and other partners in developing Scitopia, a gateway to the literature from nonprofit scientific and engineering societies. Scitopia.org searches member electronic science/technology libraries and, according to the developers, "provides relevant results, without the noise of other Internet search engines. More than three million documents, including peer-reviewed journal content and technical conference papers, spanning 150 years of science and technology, can be searched through the site."
