Future scientific software systems
IEEE Computational Science & Engineering
The Applications There is an enormous array of scientific applications to be solved. As our computers, algorithms, and software improve some of them come into the feasible range, are discovered, and attempted. The size, complexity, and difficulty of applications that we can see now greatly exceed our solution capabilities for the next 50-100 years. Consider animals; each living cell has thousands of different kinds of chemical gizmos (molecules) that make it work. Some are simple, some are
... mple, some are extremely complex. There are more of these in a single cell than there are people in the U.S. And there are more cells in an average animal than there are people in the U.S. I conclude that the accurate simulation of animals is beyond thinking about at this time, it involves simulating the behavior and interactions of perhaps 100 quadrillion (10 16 ) gizmos. Another such application is the ab initio computation of material properties. One starts with models of atoms and their interactions. With Teraflop computers one may be able to handle a few thousand atoms. But, except for a few crystalline materials, there is a complex microstructure with gaps, impurities, and micro-crystals with different sizes and orientations. On a larger scale there is a granular structure that is equally important; the micro-grains coalesce into grains or fibers which then coalesce to form "materials n as we normally think of them. The nature of these structures (and the material properties) depends critically on how the materials are made. It is clear that we cannot, in the 21st Century, simulate all the atoms in an interesting piece of material; we probably will not even be able to simulate all the micro-grains.