Stochastic hyperbolic systems, small perturbations and pathwise approximation

Publication date 15 December 2020

Published in Osaka Journal of Mathematics (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1310.7559, https://projecteuclid.org/euclid.ojm/1602230465

large deviation principle pathwise approximation small perturbation stochastic hyperbolic system support theorem

Initial value problems for first-order hyperbolic equations (35L03) Large deviations (60F10) Stochastic calculus of variations and the Malliavin calculus (60H07) Stochastic partial differential equations (aspects of stochastic analysis) (60H15)

Abstract: This paper is devoted to the study of hyperbolic systems of linear partial differential equations perturbed by a Brownian motion. The existence and uniqueness of solutions are proved by an energy method. The specific features of this class of stochastic partial differential equations are highlighted and the comparison with standard existence results for SPDEs is discussed. The small perturbations problem is studied and a large deviation principle is stated. A pathwise approximation result, similar to the stochastic differential equations case, is established, with an application to a support theorem.

Review text

The author addresses several problems concerning systems of linear hyperbolic equations driven by a one-dimensional Brownian motion \(w\). Let \(s\in\mathbb R\) and let \(d,d^\prime\) be positive integers. First, the existence and uniqueness of solutions of the equation \[ \frac{\partial u(t,x)}{\partial t}=a_t(x,D)u(t,x)\circ\frac{dw(t)}{dt}+b_t(x,D)u(t,x)+f(t,x)\circ\frac{dw(t)}{dt}+g(t,x),\quad u(0,\cdot)=u_0(\cdot) \] on \(\mathbb R^d\) that are bounded in \(H^s\) and \(\gamma\)-Hölder continuous in \(H^{s-2}\) for every \(\gamma<\frac{1}{4}\) are proved. Here \(u_0\in(H^s)^{d^\prime}\), \(f\) and \(g\) have paths in \(C^0([0,T],(H^s)^{d^\prime})\), paths of \(f\) are bounded in \(L^4(\Omega,H^{s+1})\), paths of \(g+\langle f,w\rangle\) are bounded in \(L^4(\Omega,H^s)\), \(a_t(x,D),b_t(x,D)\) are a bounded family of pseudo-differential operators satisfying \[ |D^\alpha_\xi D^\beta_xc(x,\xi)|\le C(\alpha,\beta)(1+|\xi|))^{1-|\alpha|} \] for all multiindices \(\alpha,\beta\) and every \(c\in\{a_t,b_t:\,t\in[0,T]\}\), \(t\mapsto a_t,b_t\) are continuous in \(C^\infty(\mathbb R^{2d})\) and \[ |D^\alpha_\xi D^\beta_xc(x,\xi)|\le C(\alpha,\beta)(1+|\xi|))^{-|\alpha|} \] holds for all multiindices \(\alpha,\beta\) and every \(c\in\{A_t,B_t,L_t:\,t\in[0,T]\}\), where \[ A_t=a_t+a^*t,\quad B_t=b_t+b^*_t,\quad L_t=A_ta_t+a^*_tA_t. \] Second, it is shown that solutions of \[ du^\varepsilon=\sqrt\varepsilon a_t(x,D)u^\varepsilon\circ dw+b_t(x,D)u^\varepsilon\,dt,\quad u^\varepsilon(0)=u_0\in H^s \] converge (in probability) uniformly in \(H^{s-2}\), as \(\varepsilon\to 0+\), to the solution of \[ du=b_t(x,D)u,\quad u=u_0 \] and the laws of \(u^\varepsilon\) satisfy a large deviation principle in \(C_{u_0}([0,T],H^{s-2})\) with the good rate function \[ I_u(\phi)=\inf\left\{\frac{1}{2}\int_0^T|\dot h(t)|\,dt:\Psi(h)=\phi\right\}, \] where \(\Phi:C_0([0,T],\mathbb R)\to C_{u_0}([0,T],H^{s-2})\) is given by \[ \Phi(h)(t)=u_0+\int_0^ta(x,D)\Phi(h)(\tau)\dot h(\tau)\,d\tau+\int_0^tb(x,D)\Phi(h)(\tau)\,d\tau. \] Third, if one considers solutions of the equations \[ \frac{\partial u^n}{\partial t}=a_t(x,D)u^n(t)\dot w^n(t)+b_t(x,D)u^n+f(t,x)\dot w^n(t)+g(t,x),\quad u^n(0,\cdot)=u_0(\cdot), \] where \(w^n\) is the polygonal approximation of \(w\) on the partition \(\{iT/n:0\le i\le n\}\) then \[ \lim_{n\to\infty}\mathbb E\,\sup_{t\in[0,T]}|u^n(t)-u(t)|^2_{H^{s-2}}=0 \] and the topological support of the law of \(u\) in \(C^0([0,T];(H^{s-2})^{d^\prime})\) coincides with the closures of the sets \[ \{v(t,\phi):\,\phi\in H^\infty\},\quad\{v(t,\phi):\,\phi\in H_p^\infty\} \] in \(C^0([0,T];(H^{s-2})^{d^\prime})\), where \(H^\infty\) and \(H^\infty_p\) denote the sets of real infinitely and piecewise infinitely differentiable functions, respectively, starting from the origin and \(v\) solves the equation \[ dv(t)=a_t(x,D)v(t)\dot\phi(t)\,dt+b(x,D)v(t)\,dt,\quad v(0)=u_0. \] The author subsequently applies the above Wong-Zakai approximation result to a problem of existence of a stochastic flow of diffeomorphisms for a class of backward stochastic hyperbolic equations. Finally, the scalar equation \[ du=a_t(x,D)u(t)\circ dw(t)+b_t(x,D)u(t)\,dt,\quad u(0)=u_0\in H^s \] is considered where, in addition, the principal symbols of \(a_t\) and \(b_t\) are imaginary, and \(s>d/2\). Then \(u\) is defined for every \((t,x)\), \(u(t)\in\mathbb D^{1,2}(H^{s-2})\), \(D_\theta u(t)=0\) for \(\theta>t\) and \[ D_\theta u(t)=a_\theta(x,D)u(\theta)+\int_\theta^ta_\theta(x,D)D_\theta u(\tau)\circ dw(\tau)+\int_\theta^tb_\theta(x,D)D_\theta u(\tau)\,d\tau. \] Here \(D\) stands for the Malliavin derivative.

Recommendations

Cites work

Cited in

(11)

This page was built for publication: Stochastic hyperbolic systems, small perturbations and pathwise approximation

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2215992)