Infinite-dimensional calculus under weak spatial regularity of the processes (Q1661583): Difference between revisions

Let $H$ be a Hilbert space and $X$ be a $H$-valued Itô process, solving some equation \[ dX_t = A X_t dt + B_t dt + C_t dW_t, \quad 0\le t\le T, \] where $W$ is a Hilbert space-valued Wiener process, $B_t$ and $C_t$ are progressively measurable and almost surely $dt$-square integrable processes, and $A$: $D(A)\subset H \to H$ is an unbounded linear operator. \par The purpose of the authors is to derive an Itô formula for $F(t,X_t)$ in that context, which is generally not available since some formal terms entering it can be hardly defined. Their idea here is to focus on a somewhat particular case in which cancellations of divergent parts occur, namely in the sum \[ \partial_tF(t,x) + \langle Ax,D_xF(t,x)\rangle =: G(t,x), \] yielding some regular function $G$ and thereby allowing to derive a convenient Itô formula. \par An example is given by $F(t,x) = F_0(e^{-tA}x) + \int_0^t G(s,e^{(s-t)A}x) ds$. \par To proceed, the authors assume the existence of a Banach space $B$ continuously embedded in $H$ such that: $D(A)\subset B, e^{tA}$ is strongly continuous on $B$, $X_t\in B$, $X_{[0,T]}$ is a.s. relatively compact in $B$. They consider a function $F\in C([0,T]\times H; \mathbb{R})$ having continuous $H$-derivatives $DF,D^2F$ and such that the above sum, defined a.s. on $D(A)$, extends into a continuous function $G$. Then, they establish the following Itô formula (where $Q$ is the nuclear covariance operator of $W$): \[ \begin{multlined} F(t,X_t) - F(t,X_0) = \int_0^t G(s,X_s)ds \\ + \int_0^t \Big[\big\langle B_s,D_xF(s,X_s)\big\rangle + \frac12 Tr\big(C_sQC_s^* D_x^2F(s,X_s)\big) \Big] ds \\ + \int_0^t \big\langle D_xF(s,X_s),C_sdW_s\big\rangle . \end{multlined} \] Then the authors extend that to some particular Banach spaces having a product structure, with the noise in a Hilbert component. They finally show that the Itô formula they have obtained in this way applies to path-dependent functionals $F$, which was their initial motivation, and to the uniqueness of solutions to some path-dependent Kolmogorov equations.