Commuting birth-and-death processes (Q2268727)

The finite time behavior of a one-dimensional birth-and-death process is easy to study because its transition matrix \(P\) is tri-diagonal and can be diagonalized using related family of orthogonal polynomials. This enables the computation of the power \(P^n\) whose entries are the probabilities of going from one state to another in \(n\) steps. Unfortunately, one-dimensional theory can not be extended to higher-dimensional birth-and-death processes (having as their state space products of intervals in higher-dimensional integer lattices) even when such processes have only nearest neighbor transitions. In the article, a special case of \(m\)-dimensional process is considered which transition matrix allows the presentation \(P=\sum_{k=1}^m P_k\). Each \(P_k\) describes nearest neighbor transitions in \(k\)-th direction only and all these matrices commute, i.e. \(\forall l,k: P_kP_l=P_lP_k\). This allows to use one-dimensional theory to compute \(P^n\). For example, if \(m=2\) then \( P^n=\sum_{s=0}^n {n \choose s}P_1^s P_2^{n-s}\) and \(P_1^s\), \(P_2^{n-s}\) can be computed using one-dimensional theory. The set of commuting birth-and-death processes decomposes as a union of toric varieties, with the main component being the closure of all processes whose nearest neighbor transition probabilities are positive. An explicit monomial parametrization for this main component is given. The boundary components are explored using primary decomposition.