Exactly Solvable Birth and Death Processes

Many examples of exactly solvable birth and death processes, a typical stationary Markov chain, are presented together with the explicit expressions of the transition probabilities. They are derived by similarity transforming exactly solvable `matrix’ quantum mechanics, which is recently proposed by Odake and the author. The ($q$-)Askey-scheme of hypergeometric orthogonal polynomials of a discrete variable and their dual polynomials play a central role. The most generic solvable birth/death rates are rational functions of $q^x$ ($x$ being the population) corresponding to the $q$-Racah polynomial.

💡 Research Summary

The paper presents a unified framework for constructing a broad class of exactly solvable birth‑and‑death (BD) processes, which are discrete‑time, one‑dimensional Markov chains describing the stochastic evolution of a population size. The key idea is to exploit a recently introduced “matrix quantum mechanics” (MQM) formalism, in which the transition matrix of the BD process is treated as a similarity‑transformed version of a solvable Hamiltonian. The eigenfunctions of this Hamiltonian are precisely the orthogonal polynomials of the (q‑)Askey scheme—hypergeometric or basic hypergeometric polynomials defined on a discrete variable—and their duals. Because these polynomials satisfy three‑term recurrence relations, they diagonalise the transition matrix, yielding explicit expressions for all time‑dependent transition probabilities.

The authors first review the standard BD formalism: birth rates λ(x) and death rates μ(x) defined on states x = 0,1,…,N give a tridiagonal transition matrix T with entries T_{x,x+1}=λ(x), T_{x,x‑1}=μ(x), and T_{x,x}=1‑λ(x)‑μ(x). In conventional treatments one must either compute matrix powers numerically or solve a system of difference equations, both of which become intractable for large N. By contrast, in the MQM approach one writes T = Φ Λ Φ⁻¹, where Φ_{x,n}=φ_n(x) is the matrix of orthogonal polynomials φ_n(x) (n = 0,…,N) and Λ = diag(λ_0,…,λ_N) contains the eigenvalues. The eigenvalues are simple rational functions of the polynomial degree n, typically of the form λ_n = −(q^{-n}‑1)(1‑ab q^{n+1}) for q‑Racah polynomials, with parameters a, b, q encoding the birth‑death rates.

A series of concrete families is then derived:

1. Krawtchouk BD process – λ(x)=p(N‑x), μ(x)=(1‑p)x. The eigenfunctions are Krawtchouk polynomials; transition probabilities reduce to binomial expressions.
2. Hahn BD process – λ(x)=(x+α)(N‑x), μ(x)=(x+β)(N‑x). Eigenfunctions are Hahn polynomials, leading to {}_3F_2 hypergeometric transition kernels.
3. q‑Hahn and q‑Krawtchouk processes – q‑deformations of the above, with rates expressed as rational functions of q^x. Their kernels involve basic hypergeometric series {}_3φ_2.
4. q‑Racah process – the most general solvable case. Birth and death rates are rational functions of q^x with four independent parameters (a, b, c, d). The eigenfunctions are q‑Racah polynomials, and the transition probabilities are given by a {}_4φ_3 series. All previously listed families appear as special limits of the q‑Racah case.

Having diagonalised T, the time evolution of any initial distribution P_0(x) is obtained analytically. Expanding P_0 in the polynomial basis, P_0(x)=∑{n}c_n φ_n(x) with coefficients c_n determined by the orthogonality weight, the distribution after t steps is simply P_t(x)=∑{n=0}^N c_n λ_n^t φ_n(x). Thus the dynamics reduces to raising the scalar eigenvalues λ_n to the t‑th power, bypassing any matrix multiplication. This yields closed‑form expressions for the full transition matrix P(t)=T^t, which are especially valuable for computing hitting times, stationary distributions, and correlation functions.

The paper emphasizes that q‑Racah polynomials sit at the top of the (q‑)Askey hierarchy; consequently, the q‑Racah BD process provides a universal template from which all lower‑level processes can be obtained by parameter limits. This unification offers both theoretical insight—linking stochastic processes to the algebraic structure of basic hypergeometric orthogonal polynomials—and practical benefits, such as exact likelihood formulas for statistical inference in population genetics, epidemiology, and reliability engineering.

In the concluding discussion the authors note several directions for future work: extending the MQM approach to birth‑death processes with immigration or catastrophes, exploring continuous‑time limits (yielding exactly solvable birth‑death generators), and investigating connections with quantum algebras and integrable systems. Overall, the paper delivers a powerful algebraic toolkit that transforms a traditionally numerical problem into an analytically tractable one, opening new avenues for both applied probability and mathematical physics.