Large deviations for a symmetric branching random walk on a multidimensional lattice (Q2446224)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Large deviations for a symmetric branching random walk on a multidimensional lattice
scientific article

    Statements

    Large deviations for a symmetric branching random walk on a multidimensional lattice (English)
    0 references
    16 April 2014
    0 references
    Consider a branching random walk on \(\mathbb{Z}^d\) with one type of particles and without extinction, constructed as follows: Let the underlying random walk be symmetric and irreducible with generator \[ ({\mathfrak L}f)(x)= \sum_{z\neq 0} (f(x+ z)- f(x))\,q(z),\quad \sum_{z\neq 0} q(z)=-q(0)= 1. \] Assume the intensity of branching to be of the form \[ V(x)= \sum^N_{l=1} V_l\delta(x- x_l),\;V_l\geq 0. \] When a particle is stopped according to this density at position \(x\), it is replaced by two particles starting at \(x\) independent random walks with generator \({\mathfrak L}\). Assuming that \[ \sum_{z\neq 0} q(z)(e^{(\lambda,z)}- 1)< 0, \] the authors obtain explicit asymptotic expressions, as \(t\to\infty\), for the transition probability \(p(t,x,y)\) of the random walk and for the corresponding Green function (resolvent of \({\mathfrak L}\)), \[ G_\mu(x,y)= \int^\infty_0 e^{-\mu t}p(t,x,y)\,dt,\;\mu\geq 0. \] These results are used to derive properties of the first-moment semigroup of the number of particles: There exist \(\beta_{\mathrm{cr}}\geq 0\) and \(\beta_1> \beta_{\mathrm{cr}}\) such that, for every \(\beta\in(\beta_{\mathrm{cr}},\beta_1)\), the discrete spectrum of the generator of this semigroup consists of one simple eigenvalue \(\lambda_0(\beta)\) with eigenfunction \(\psi_0(x,\beta)\). The authors derive the asymptotic behavior of \(\lambda_0(\beta)\), as \(\beta\to \beta_{\mathrm{cr}}\), and of \(\psi_0(x,\beta)\), as \(|x|\to \infty\). Denoting by \(m_1(t,y)\) the expectation of the number of particles at time \(t\) at position \(y\), if the process was started with one particle at position \(0\), an explicit representation of the population front \(\Gamma_t= \{y: m_1(t,y)\leq f(t)\}\), \(f(t)= o(e^{\varepsilon t})\), is given with logarithmic accuracy.
    0 references
    symmetric random walk
    0 references
    branching random walk
    0 references
    population front
    0 references
    large large deviations
    0 references

    Identifiers