Massachusets Institute of Technology, 1994, 243 p. Cellular automata (CA) are fully discrete, spatially-distributed dynamical systems which can serve as an alternative framework for mathematical descriptions of physical systems. Furthermore, they constitute intrinsically parallel models of computation which can be e fficiently realized with special-purpose cellular automata machines.
The basic objective of this thesis is to determine techniques for using CA to model physical phenomena and to develop the associated mathematics. Results may take the form of simulations and calculations as well as proofs, and applications are suggested throughout.
We begin by describing the structure, origins, and modeling categories of CA. A general method for incorporating dissipation in a reversible CA rule is suggested by a model of a lattice gas in the presence of an external potential well. Statistical forces are generated by coupling the gas to a low temperature heat bath. The equilibrium state of the coupled system is analyzed using the principle of maximum entropy. Continuous symmetries are important in eld theory, whereas CA describe discrete fi elds. However, a novel CA rule for relativistic dffi usion based on a random walk shows
how Lorentz invariance can arise in a lattice model. Simple CA models based on the dynamics of abstract atoms are often capable of capturing the universal behaviors of complex systems. Consequently, parallel lattice Monte Carlo simulations of abstract polymers were devised to respect the steric constraints on polymer dynamics. The resulting double space algorithm is very effi cient and correctly captures the static and dynamic scaling behavior characteristic of all polymers. Random numbers are important in stochastic computer simulations; for example, those that use the Metropolis algorithm. A technique for tuning random bits is presented to enable effi cient utilization
of randomness, especially in CA machines. Interesting areas for future CA research include network simulation, long-range forces, and the dynamics of solids. Basic elements of a calculus for CA are proposed including a discrete representation of one-forms and an analog of integration. Eventually, it may be the case that physicists will be able to formulate cellular automata rules in a manner analogous to how they now derive di erential equations.
