Solutions of nonlinear impulsive Itô functional-differential equations: stability by the linear approximation (Q389324)

The author considers the equation \[ dx(t)=(Vx)(t)\,dZ(t) \text{ for } t\geq 0 \tag{1} \] with the conditions \[ x(\mu_j)=A_jx(\mu_j-0) \text{ for } j=1,2,\dots \text{ at times } 0=\mu_0<\mu_1<\mu_2<\dots\uparrow\infty, \] where \(V\) is a nonlinear Volterra operator mapping progressively measurable \(n\)-dimensional càdlàg processes to \((n\times m)\)-matrix-valued progressively measurable processes locally integrable with respect to \(Z=(t,\mathcal B_2(t),\dots,\mathcal B_m(t))^T\) (the latter space being denoted by \(L^n(Z)\)), \(\mathcal B_2,\dots,\mathcal B_m\) are independent Brownian motions and \(A_j\) are real numbers. The norm \(\|x\|^p_{M^p_\gamma}=\sup_{t\geq 0}\,\{\mathbb E\,|\gamma(t)x(t)|^p\}\), where \(\gamma\) is a positive function, is introduced. It is assumed that \(V0=0\), hence the equation in question has a trivial solution, and a notion of local \(M^p_\gamma\)-stability of the trivial solution is introduced in the standard manner (i.e., if initial conditions converge in \(L^p\) to zero then the corresponding solutions converge to the trivial solution in the norm of \(M^p_\gamma\)). The linearized equation \[ dx(t)=[(V_1x)(t)+f(t)]\,dZ(t) \text{ for } t\geq 0 \] is then considered with the same conditions \[ x(\mu_j)=A_jx(\mu_j-0) \text{ for } j=1,2,\dots, \] where \(f\in L^n(Z)\), \(V_1=V-F\) is assumed to be linear with respect to the field of real \(\mathcal F_0\)-measurable random variables and \(F\) is a nonlinear Volterra operator satisfying \(F0=0\). The notion of \((M^p_\gamma,B)\)-admissibility is introduced for a linear normed subspace \(B\) of \(L^n(Z)\) and it is proved that for such an admissibility pair \((M^p_\gamma,B)\) and \(F\) satisfying a locally-Lipschitz-like property as a mapping from \(M^p_\gamma\) to \(B\), the original equation (1) is \(M^p_\gamma\)-stable. This result is further discussed and refined, and the \((M^p_\gamma,B)\)-admissibility condition and the applicability of the \(M^p_\gamma\)-stability result are demonstrated on an example.