A perturbation result for semi-linear stochastic differential equations in UMD Banach spaces (Q2441323): Difference between revisions

The authors consider the effect of perturbations of the generator \(A\) on the solution to the semi-linear stochastic differential equations : \[ \begin{cases} d U(t) = A U(t) dt + F(t, U(t)) dt + G(t, U(t)) d W_H(t), & t > 0, \\ U(0) = x_0. \end{cases} \tag{1} \] Here, \(A\) is the generator of an analytic \(C_0\)-semigroup \(S\) on a UMD Banach space \(X\), \(H\) is a Hilbert space, \(W_H\) is an \(H\)-cylindrical Brownian motion, \[ G : [0, T] \times X \to {\mathcal L} ( H, X_{ \theta_G}^A), \qquad F : [0, T] \times X \to X_{\theta_F}^A \] for some \(\theta_G > - \frac{1}{2}\), \(\theta_F > - \frac{3}{2} + \frac{1}{\tau}\) with \(\tau \in [1,2]\), and \(X_{ \theta_F}^A\) is the fractional domain space \(D( (- A)^{\theta_F})\). They assume that \(F\) and \(G\) satisfy certain global Lipschitz and linear growth conditions. The perturbed equation is given by \[ \begin{cases} d U^{(0)}(t) = A_0 U^{(0)}(t) dt + P_0 F(t, U^{(0)}(t) ) dt + P_0 G( t, U^{(0)}(t)) d W_H(t), & t> 0, \\ U^{(0)}(0) = P_0 x_0, \end{cases} \tag{2} \] where \(P_0 : X \to X_0\) is a bounded projection with a finite dimensional closed subspace \(X_0\) of \(X\). Let \(R( \lambda; A)\) denote the resolvent of \(A\) in \(\lambda \in {\mathbb C}\), i.e., \(R(\lambda; A)=( \lambda I - A )^{-1}\), and \[ D_{\delta}(A, A_0) := \| R( \lambda_0; A) - i_{X_0} R( \lambda_0; A_0) P_0 \|_{ {\mathcal L}( X_{\delta -1}^A, X)} < \infty \tag{3} \] for some \(\lambda_0\) with \(\mathrm{Re}( \lambda _0)\) sufficiently large and \(\delta \geq 0\). Assume that \(A_0\) is the generator of an analytic \(C_0\)-semigroup \(S_0\) on \(X_0\), and \(\omega \geq 0\), \(\theta \in ( 0, \frac{\pi}{2} ]\) and \(K > 0\) are such that \(A\) and \(A_0\) are both of type \(( \omega, \theta, K)\). Here \(A\) is said to be of type \(( \omega, \theta, K)\) where \(\omega \in {\mathbb R}\), \(\theta \in ( 0, \frac{\pi}{2} ]\) and \(K > 0\), if \(\omega + \Sigma_{ \frac{\pi}{2} + \theta}\) \(\subseteq\) \(\rho(A)\) (the resolvent set), \(( e^{ \omega t} S(t) )_{ t \geq 0}\) is bounded, and \[ | \lambda - \omega | \cdot \| R( \lambda; A) \|_{ {\mathcal L}(X)} \leq K \quad \text{for all} \quad \lambda \in \omega + \Sigma_{ \frac{\pi}{2} + \theta}, \] and \(\Sigma_{\delta}\) \(:=\) \(\{ z \in {\mathbb C} \setminus \{0 \}: | \arg (z) | < \delta \}\). Here is the main result. Theorem. Suppose that there exist \(\delta \in [0,1]\) and \(p \in (2, \infty)\) satisfying \[ 0 \leqslant \delta < \min \left\{ \frac{3}{2} - \frac{1}{\tau} + \theta_F, \frac{1}{2} - \frac{1}{p} + \theta_G \right\} \tag{4} \] such that, for some \(\lambda_0 \in {\mathbb C}\) such that \(\mathrm{Re}( \lambda_0) \geq \omega\), we have \[ D_{\delta}( A, A_0) < \infty. \tag{5} \] Suppose that \(x_0 \in L^p( \Omega, {\mathcal F}_0; X_{\delta}^A)\) and \(y_0 \in L^p( \Omega, {\mathcal F}_0 ; X)\). Then for any \(\alpha \in [0, \frac{1}{2} )\) there exists a unique process \(U^{(0)}\) \(\in\) \(V_c^{\alpha, p}( [0, T] \times \Omega; X_0)\) such that \[ s \mapsto 1_{ [ 0, t]}(s) S_0 ( t-s) \times P_0 G( s, U^{(0)} (s)) \] is \(L^p\)-stochastic integrable for all \(t \in [0, T]\), and for all \(t \in [0, T]\) we have \[ \begin{multlined} U^{(0)} (t) = S_0(t-s) P_0 y_0 + \int_0^T S_0(t-s) P_0 F( s, U^{(0)} (s)) ds \\ + \int_0^T S_0(t-s) P_0 G(s, U^{(0)} (s)) d W_H (s), \quad \text{a.s.} \end{multlined}\tag{6} \] Moreover, \[ \| U - i_{X_0} U^{(0)} \|_{ V_c^{\alpha, p} ( [0,T] \times \Omega; X)} \leqslant \| x_0 - y_0 \|_{ L^p(\Omega; X)} + D_{\delta} ( A, A_0) ( 1 + \| X_0 \|_{ L^p( \Omega; X_{\delta}^A )}), \tag{7} \] where \(V_c^{\alpha, p} ( [a,b] \times \Omega; X)\) denotes the space of adapted continuous processes \(\Phi : [a,b] \times \Omega \to X\) such that \[ \begin{multlined} \| \Phi \|_{V_c^{\alpha, p} ( [a,b] \times \Omega; X)} := \| \Phi \|_{ L^p( \Omega; C([a,b]; X))} \\ + \sup_{ a \leqslant t \leqslant b} \| s \mapsto (t-s)^{- \alpha} \Phi(s) \|_{ L^p( \Omega; \gamma(L^2(a,t), X) )} < \infty. \end{multlined}\tag{8} \] Here, \(\gamma({\mathcal H},X)\) is the Banach space defined as the completion on \({\mathcal H} \otimes X\) with respect to the norm when \({\mathcal H} = L^2(0,T; H)\).