Extinction versus exponential growth in a supercritical super-Wright-Fisher diffusion. (Q2574580)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Extinction versus exponential growth in a supercritical super-Wright-Fisher diffusion.
scientific article

    Statements

    Extinction versus exponential growth in a supercritical super-Wright-Fisher diffusion. (English)
    0 references
    0 references
    0 references
    29 November 2005
    0 references
    Let us denote by \(A\) the closure of the operator \(\frac 12 x(1-x) {d^ 2\over dx^ 2}\) in the space \(C([0,1])\) of continuous functions on \([0,1]\), and consider a semilinear Cauchy problem \[ \partial _ {t}u = Au + \gamma u(1-u), \quad u(0,\cdot ) = f,\tag{1} \] where \(\gamma >0\) and \(f\) is a bounded nonnegative measurable function on \([0,1]\). It is known that there exists a unique mild solution \(u\) to (1) given by \(u(t,\cdot ) = \mathbb U_ {t}f\), \((\mathbb U_ {t})\) being the log-Laplace semigroup of the \((A,\gamma ,\gamma )\)-superprocess \(\mathbb X\). That is, \(\mathbb X\) is the (unique in law) Markov process with values in the space of finite measures on \([0,1]\) and with continuous trajectories satisfying \(\mathbf E_ \mu \exp (-\langle \mathbb X_ {t},f \rangle ) = \exp (-\langle \mu ,\mathbb U_ {t}f\rangle )\) for all \(t\geq 0\), all bounded nonnegative measurable functions \(f\) and all finite measures \(\mu \). The problem (1) is studied by means of the process \(\mathbb X\). Suppose \(\mathbb X_ 0 = \mu \). For example, it is shown that there exist nonnegative random variables \(W_ 0\), \(W_ 1\) and \(W_ {(0,1)}\) (dependent on \(\mu \)) such that \(\lim _ {t\to \infty } e^ {-\gamma t}\langle \mathbb X_ {t},\mathbf 1_ {\{r\}}\rangle = W_ {r}\) almost surely, \(r=0,1\), and \(\lim _ {t\to \infty } e^ {-( \gamma -1)t}\langle \mathbb X_ {t},6x(1-x)\rangle = W_ {(0,1)}\) almost surely. The following relations hold almost surely: \(W_ {r}=0\) if and only if \(\mathbb X_ {t}(\{r\})=0\) for all sufficiently large \(t\)'s; \(W_ {(0,1)}=0\) if and only if \(\mathbb X_ {t} ((0,1)) = 0\) for all sufficiently large \(t\)'s. Further, if \(\gamma \leq 1\), then \(W_ {(0,1)}=0\) almost surely, while \(\mathbf E_ \mu W_ {(0,1)} = 6\int _ {[0,1]} x(1-x)\,d\mu (x)\) for \(\gamma > 1\). These results are used to describe the long-time behaviour of \((\mathbb U_ {t})\). In particular, it is proved that \((\mathbb U_ {t})\) has exactly five fixed points for \(\gamma >1\) and exactly four fixed point for \(\gamma \leq 1\). These fixed points are twice continuously differentiable on \([0,1]\) and are stationary solutions to (1).
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    semilinear Cauchy problem
    0 references
    log-Laplace semigroup
    0 references
    weighted superprocess
    0 references