2
$\begingroup$

Let $(G,\circ)$ be a Polish group, with identity $e$.
Let $\Omega$ be the set of continuous functions $\omega:\mathbb{R} \to G$ such that $\omega(0)=e$.
For each $t \in \mathbb{R}$, define the projection $\pi_t:\Omega \to G$ by $\pi_t(\omega)=\omega(t)$.
Let $\mathcal{F}:=\sigma(\pi_t : t \in \mathbb{R})$. (Equivalently, $\mathcal{F}$ is the Borel $\sigma$-algebra generated by the compact-open topology on $\Omega$.)
For each $t \in \mathbb{R}$, define the $t$-shift map $\theta^t:\Omega \to \Omega$ by $\theta^t(\omega)(s)=\omega(s+t) \circ \omega(t)^{-1}$.
Let $\mathbb{P}$ be a probability measure on $(\Omega,\mathcal{F})$ that is ergodic with respect to the dynamical system $(\theta^t)_{t \geq 0}$. (The main case of interest to me is when $(G,\circ)=(\mathbb{R},+)$ and $\mathbb{P}$ is the "Wiener measure", meaning that the stochastic processes $(\pi_t)_{t \geq 0}$ and $(\pi_{-t})_{t \geq 0}$ are independent Wiener processes under $\mathbb{P}$.)

Does there exist a $\mathbb{P}$-null set $A \subset \Omega$ such that the set

$$\Omega_A \ := \ \{ \omega \in \Omega \, : \, \forall \ n \in \mathbb{N}, \ \exists \ t \geq n \ \ \textrm{s.t.} \ \theta^t(\omega) \in A \}$$

is not a $\mathbb{P}$-null set? Does there exist a $\mathbb{P}$-null set $A \subset \Omega$ such that $\Omega_A$ is a $\mathbb{P}$-full set, or even such that $\Omega_A$ is the whole of $\Omega$?

As I say, I am primarily interested in the case of the Wiener space described above. Many thanks.

$\endgroup$
5
  • 2
    $\begingroup$ What about Brownian motion on the circle; $A$ is the event that you're at one particular point of the circle? $\endgroup$ Commented Feb 21, 2014 at 17:33
  • $\begingroup$ @AnthonyQuas I don't see how "the event that you're at one particular point of the circle" is a well-defined subset of $\Omega$. $\endgroup$ Commented Feb 22, 2014 at 13:48
  • $\begingroup$ However, I think I have just managed to answer the main part of my question: Let $\mathbb{P}$ be the Wiener measure, and let $A$ be the set of all $\omega \in \Omega$ for which there exists $\varepsilon>0$ such that $\omega(t) \neq 0$ for all $t \in (0,\varepsilon)$. Then $A$ is a $\mathbb{P}$-null set, and $\Omega_A$ is a $\mathbb{P}$-full set. $\endgroup$ Commented Feb 22, 2014 at 13:52
  • $\begingroup$ @JulianNewman: your example is of the same taste as Anthony Quas' one. "The event that you're at one particular point of the circle" means you fix $x\in\mathbb{S}^1$, (the "particular point") and then consider $A=\{\omega\in\Omega | \omega(0)=x\}$. It is very useful in probability to learn how to be able to translate between short sentences and full statements of this kind. $\endgroup$ Commented Feb 22, 2014 at 16:33
  • $\begingroup$ @BenoîtKloeckner: I agree that this would have been the obvious interpretation of "the event that you're at one particular point of the circle", if I had defined $\theta^t$ to be the "horizontal shift" on $C(\mathbb{R},G)$; however, I defined $\theta^t$ to be the "diagonal shift" on $C_0(\mathbb{R},G)$. Nonetheless, after reading Anthony's post, I did wonder whether for some fixed $\tau > 0$ and $x \in \mathbb{R}$ (or $x \in \mathbb{S}^1$) the set $A:=\{\omega \in \Omega | \omega(\tau)=x\}$ would work. For some reason I thought that it wouldn't - but I now realise that it probably does. $\endgroup$ Commented Feb 22, 2014 at 19:03

0

You must log in to answer this question.

Browse other questions tagged .