4
$\begingroup$

It's well known that the Ornstein Uhlenbeck semigroup defined by $$ P_tf(x)=\int_{\mathbb{R}}f\left(xe^{-t}+\sqrt{1-e^{-2t}}z\right)\frac{e^{-z^2/2}}{\sqrt{2\pi}}\,dz $$ is not strongly continuous on the space $C(\mathbb{R})$ of continuous functions on $\mathbb{R}$ with the supremum norm. I was wondering if the Ornstein Uhlenbeck semigroup might be strongly continuous if we consider strong continuity over a Banach space of functions bounded with respect to a weighted supremum instead, i.e. $$ \lVert f\rVert_k=\sup_{x\in\mathbb{R}}\frac{|f(x)|}{1+|x|^k}, $$ for some positive integer $k$

$\endgroup$
  • 1
    $\begingroup$ For what it's worth, the O.-U. semigroup is strongly continuous on $C_0(\Bbb R)$, the space of continuous functions on $\Bbb R$ that tends to $0$ at infinity, with the uniform norm. $\endgroup$ – John Dawkins Oct 11 '16 at 16:50
3
$\begingroup$

Nope. To see this, let $X_t(x)$ denote the OU process at time $t>0$ with initial condition $x$; set $f(x) = g(x) (1+|x|^k)$ where $g \in C_b(\mathbb{R})$; and consider: \begin{align*} \frac{P_t f(x) - f(x)}{1+|x|^k} &= \frac{E\{ g(X_t(x)) (1+|X(t)|^k) - g(x) (1+|x|^k) \}}{1+|x|^k} \\ &= \frac{E\{ g(X_t(x)) (1+|X_t(x)|^k) - g(x) (1+|x|^k) \}}{1+|x|^k} \\ &= E\{ g(X_t(x)) - g(x) \} + \frac{E\{ g(X_t(x)) (1+|X_t(x)|^k - 1-|x|^k) \}}{ 1+|x|^k} \end{align*} The second term is nicely uniformly continuous. However, the first term is not always so nice: when $x$ is very large then $X_t(x) \approx e^{-t} x$ (the effect of the noise is negligible) and the first term behaves like $g(e^{-t} x) - g(x)$, which does not in general converge to zero uniformly in $x$ as $t \to 0$. Take, e.g., $g(x) = \cos(x^2)$.

$\endgroup$
  • $\begingroup$ This answer gives a family of continuous functions whose weighted norm does not vanish at infinity because these functions grow at the rate of $1+|x|^k$ with a pre-factor that can be any function in $C_b(\mathbb{R})$. This reduces the original problem to whether or not the OU semigroup is strongly continuous on the space $C_b(\mathbb{R})$ with the supremum norm, and the example provided shows that it is not. $\endgroup$ – Nawaf Bou-Rabee Oct 11 '16 at 3:22
  • $\begingroup$ Thanks for you answer! Indeed, I agree with the statement you wrote. I'm just surprised I have seen (even classical) works where strong continuity of the OU semigroup in such norms is asserted, and stated as an obvious result. $\endgroup$ – RadonNikodym Oct 11 '16 at 14:35
  • $\begingroup$ FYI: Daprato, Giuseppe, and Alessandra Lunardi. "On the Ornstein-Uhlenbeck operator in spaces of continuous functions." Journal of Functional Analysis 131.1 (1995): 94-114. From the abstract: We study the realization of the Ornstein-Uhlenbeck operator A in the space of the uniformly continuous and bounded functions in Rn. We prove that it generates a semigroup which is neither strongly continuous nor analytic, but enjoys nice smoothing properties. $\endgroup$ – Nawaf Bou-Rabee Oct 11 '16 at 22:58

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.