The principal eigenvalue for jump processes (Q1586079)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | The principal eigenvalue for jump processes |
scientific article |
Statements
The principal eigenvalue for jump processes (English)
0 references
13 June 2001
0 references
The \(q\)-pair \((q(x), q(x,dy))\) is considered on a general measurable space \((E,\mathcal E)\). Denote \(r(x)=q(x)-q(x,E)\). Suppose that the \(q\)-pair is reversible w.r.t. a probability \(\pi\), i.e. \(\pi_q(x,dy)=\pi(dx)q(x,dy)\) is a symmetric measure on \(\mathcal E\times\mathcal E\). Let \[ D(f)=\frac{1}{2}\int \pi_q(dx,dy)[f(y)-f(x)]^2+\int \pi_r(dx)f(x)^2, \] where \(\pi_r(dx)=r(x)\pi(dx)\). The principal eigenvalue studied in the paper is defined by \[ \lambda_0=\inf\{D(f): f\in{\mathcal D}(D), \|f\|=1\}, \] where \(\|\cdot\|\) is the norm in \(L^2(\pi)\), and \({\mathcal D}(D)\) is a completion w.r.t. \(\|\cdot\|_D=(\|\cdot\|+D(\cdot))^{1/2}\) of the space \({\mathcal D}_0=\{f\in L^2(\pi): f\) vanishes out of some \(E_n\}\), \(E_n=\{x\in E: q(x)\leq n\}\) (e.g. \({\mathcal D}(D)={\mathcal D}_0=L^2(\pi)\) in case of the bounded \(q\)-pair). Define the operator \[ \Omega f(x)=\int q(x,dy)f(y)-q(x)f(x). \] The main result of the paper is the following variational formula for the lower bound of \(\lambda_0\): \[ \lambda_0\geq \sup_{0<g\in{\mathcal E}}\pi\text{-ess} \inf (-\Omega g/g). \] For the Markov chains the bound is often sharp. The condition \(g>0\) can be replaced by \(g>0\) \(\pi\)-a.s.
0 references
principal eigenvalue
0 references
jump process
0 references
variational formula for Dirichlet form
0 references
0 references