Central limit theorem for a critical multitype branching process in random environments (Q824986)

Consider a \(p\)-multitype branching process \((Z_n)_{n \geq 0}\) in random environment, defined as follows. Let \(((\rho^{(i)}_n, i \leq p))_{n \geq 1}\) be i.i.d. vectors of probability distributions on \(\mathbb N^p\). At each generation \(n \in \mathbb N\), each particle of type \(i \in [|1,p|]\) in the process creates a number of offspring of types \(1, 2, \ldots, n\) independently of every other particles, according to the distribution \(\rho^{(i)}_n\). For each \(1 \leq i,j \leq p\), \(Z_n(i,j)\) represents the number of children of type \(j\) descending from a single individual of type \(i\) at time \(0\), and we denote by \(M_n(i,j)\) its expectation conditionally on the environment \((\rho^{(i)}_n, i \leq p)_{n \geq 1}\). The sequence \((M_n, n \geq 1)\) is a product of random matrices; we denote by \(\pi_\mu=\lim_{n\to\infty}\frac{1}{n} \mathbb E\left( \left| \log M_n \right| \right)\) its upper Lyapunov exponent. The multitype branching process in random environment \((Z_n)\) is said to be supercritical if \(\pi_\mu > 0\), subcritical if \(\pi_\mu < 0\) and critical if \(\pi_\mu = 0\). In the present article, the authors prove the convergence in law of \(Z_n(i,j)/|M_n|\) as \(n \to \infty\) for a critical branching process in random environment conditioned on surviving up to time \(n\), provided some weak integrability conditions hold. Under more restrictive conditions on the reproduction law, they prove that conditionally on surviving up to time \(n\), \((\log Z_n) / \sqrt{n}\) converges in law to a Rayleigh distribution. The proofs are partly based on tight estimates on the asymptotic behaviour of product of random matrices obtained in [\textit{T. Da C. Pham}, ALEA, Lat. Am. J. Probab. Math. Stat. 15, No. 1, 67--100 (2018; Zbl 1388.60032)].