A new method for choosing the computational cell in stochastic reaction-diffusion systems (Q694697)

The authors study reaction-diffusion systems from the point of view of stochastic analysis. The problem is to choose the cell size for the stochastic simulation. A measure based on the spectrum of the covariance matrix is applied to show that for a rather wide class of first-order reaction networks this measure converges and the re-normalized measure stabilizes as the volume of the cells tends to zero. This provides a criterion for the maximal volume used for the numerical solution. A new criterion for all orders of reactions is suggested; it is based on the sensitivity of the network. The results are illustrated by some practical examples.