Almost sure stability of linear stochastic differential equations with jumps (Q1849495)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Almost sure stability of linear stochastic differential equations with jumps
scientific article

    Statements

    Almost sure stability of linear stochastic differential equations with jumps (English)
    0 references
    0 references
    1 December 2002
    0 references
    Let \(W\) be a standard \(m\)-dimensional Wiener process and \(N\) a Poisson point process on \(\mathbb{R}_{+}\times Z\) with a finite characteristic measure \(\lambda\). The system of linear stochastic differential equations with jumps \[ dx(t) = Ax(t) dt + \sum^{m}_{i=1} B_{i}x(t) \circ dW_{i}(t) + \int_{Z} C(z)x(t-)N( dt, dz), \;x(0) = x_0,\tag{1} \] is considered. It is supposed that \(x_0\in\mathbb{R}^{d}\), \(x_0\neq 0\), \(A\), \(B_{i}\), \(C(z)\) are \(d\times d\)-matrices, \(C\in L^{d+1}(\lambda)\), and that the matrix \(I+C(z)\) has full rank \(\lambda\)-almost everywhere on \(Z\). By generalizing the procedure used in the diffusion case \(C\equiv 0\), the authors obtain a rather complete description of the long time behaviour of the solution \(x\) to (1), provided that a non-degeneracy condition on the diffusion part \(\sum^{m}_{i=1} \langle y,B_{i}x\rangle^2 \geq \kappa |x|^2|y|^2\) is satisfied for some \(\kappa>0\) and all \(x,y\in \mathbb{R}^{d}\). In particular, it is proved that the process \(\theta(t) = x(t)/|x(t)|\) on the unit sphere \(\mathbb{S}^{d-1}\) is strong Feller, its transition probability has a continuous density which is strictly positive for sufficiently large times, and there exists a unique invariant probability measure \(\mu\) for \(\theta\). Therefore, it is possible to obtain a formula of Furstenberg-Khas'minskii type showing that \(\lim_{t\to\infty}t^{-1}\log|x(t)|= \int_{\mathbb{S} ^{d-1}} Q(\varphi) d\mu(\varphi)\) for an explicitly given function \(Q\) on \(\mathbb{S}^{d-1}\) and all initial conditions \(x_0\). Furthermore, several interesting two-dimensional examples are discussed in detail.
    0 references
    0 references
    invariant measure
    0 references
    Lyapunov exponent
    0 references
    large deviations
    0 references
    Poisson random measure
    0 references

    Identifiers

    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references