Cellular Automata/Excitable media

An introduction to excitable media Excitable media are nonlinear dynamic systems known for exhibiting complex behavior that can be observed as pattern formation. They are usually defined by a reaction-diffusion differential equation. The diffusion part provides stability and propagation of information, the reactive part provides interesting local dynamics. A common example of excitable media are prey-predator systems. Such systems are described by a system of differential equations, one function for each of the observed protagonists. We will discuss two different approaches to modeling excitable media. Discretization of differential equations and modeling with cellular automata. Boundary conditions There are different ways to define boundary conditions for the reaction-diffusion equation. Dirichlet boundary conditions The value of the function at the boundary is given explicitly . Cyclic boundaries If the initial condition is supposed to be periodic in space, cyclic boundary conditions can be used. Zero-flux boundary conditions If zero-flux is expected at the boundary than the component of the functions first derivative normal to the boundary is zero at the boundary. This can be achieved by reflecting function values from the inside over the boundary to the outside. Discretization of differential equations using the explicit FTCS method The traditional method to simulate excitable media is discretization and numerical computation of the governing PDE. First the FTCS (forward-time centered-space method) discretization method is presented. Explicit methods are the simplest and the equations are similar to a cellular automaton, but are inadequate because of stability and convergence problems. Single PDE We will first observe a single PDE describing a single function. One-dimensional problem In the one dimensional case the space vector becomes a single variable . The nabla operator becomes . The partial differential equation is discretized. thumb|300px|right|Forward-time centered-space method Each finite element at time is calculated from three neighboring elements at time (see figure at the right). where the diffusion number is Stability The FTCS method is stable if Boundary conditions For periodic boundaries present values at the left boundary can be used to compute the future values at the right boundary and the other way round. If there is zero-flux at the boundaries than values outside the boundary are reflections of values inside . Two-dimensional problem In the two-dimensional case, the space vector becomes a variable pair . The nabla operator becomes . The partial differential equation is discretized using the forward-time centered-space method. thumb|300px|right|Forward-time centered-space method Each finite element at time is calculated from five neighboring elements at time (see figure at the right). where the diffusion numbers are Stability The FTCS method is stable if and Boundary conditions The same ideas as in the one dimensional case can be used for two dimensions. System of PDE A system of PDE describes two functions that interact with each other (prey-predator). The interaction is local, which means, the dispersion part can be computed separately for each equation, and than the reaction part is added to the result. Other PDE discretization methods Modeling with cellular automata Greenberg-Hastings Model Conclusions References Joe D. Hoffman, Numerical Methods for Engineers and Scientists Toffoli T., Margolus N., Cellular Automata Machines: A New Environment for Modeling, The MIT Press (1987), Cambridge, Massachusetts Toffoli T., Cellular automata as an alternative to Differential equations, in Modeling Physics, Physica 10D, (1984) http://www.jweimar.de/paper-abstracts.html Robert Fisch, Janko Gravner, David Griffeath, Threshold-Range Scaling of Excitable Cellular Automata Robert Fisch, Janko Gravner, David Griffeath, Metastability in the Greenberg-Hastings Model Marcus R. Garvie Finite difference schemes for reaction-diffusion equations modeling predator-prey interactions in MATLAB Stephen Wolfram, A New Kind of Science, Wolfram Media, (2002) Pattern formation Cellular automata interesting collection of books http://www.schatten.info/info/ca/ca.html#CAandDG http://psoup.math.wisc.edu/kitchen.html Primordial soup kitchen] research papers Greenberg-Hastings Models Dynamic Formation of a 3-strand Spiral Belousov-Zhabotinsky oscillating chemical reaction Belousov-Zhabotinsky oscillating chemical reaction continued Living Spirals: Dictyostelium discoideum Zoologisches Institut München - CA modeling of the wave equation Lattice-Gas Cellular Automata