The electron-molecule scattering problem is inherently computer-intensive. Although computers have been getting faster at a rapid pace for several decades now, there is no immediate prospect of a single processor being powerful enough to solve such problems accurately for all molecules of interest. One problem is that the most accurate techniques for solving the collision problem have poor scaling: That is, the required work increases much faster than the size of the molecule. For example, changing a scale parameter such as the number of electrons by a factor of 2 may change the required work (or memory, or disk space) by a factor of 8 (cubic scaling), 16 (quartic scaling), or more. The graph below shows how rapidly work increases in the comparatively mild case of cubic scaling.
Because we want to solve the electron-molecule collision problem for molecules of practical interest, including fluorocarbon etchants used in semiconductor processing, we are driven to the largest computers available, which aggregate many individual processors into a massively parallel processor (MPP).
It's true that programming for an MPP is not as easy as programming for a single processor; programming efficiently, so that the resulting program is scalable to dozens or hundreds of processors, is even more difficult. (Indeed, failure to design for scalability--which encompasses division of data as well as work--is probably the most common, and worst, mistake made in developing parallel programs.) However, the payoff from successful parallelization is high: One can tackle problems that cannot be tackled any other way.
Today our largest calculations are conducted primarily on the 256-processor HP Exemplar X-Class operated by Caltech's Center for Advanced Computing Research. However, a parallel computer need not be a single large box; it can also be a collection of smaller boxes connected by a fast network. Through an equipment grant from Intel, we have built just such a cluster for dedicated use in our electron collision calculations. You can read more about that cluster here.