0
$\begingroup$

This is a reference request question.

Statement: I am interested in an SDE of the form

\begin{equation}\fbox{1}~~~ {\rm d}X_t = f(X_t)\,{\rm d} t + g(t) \, {\rm d} B_t \end{equation}
Where we assume that $g(t)\in (0,\infty)$ (notice that g(t) is a determistic function), and $f(\cdot)$ satisfies all the conditions required such that $\fbox{1}$ admits a strong continuous solution for all $t\geq 0$.

Result: I am interested in a reference of a result of the type $\mathbb{P}( \text{for some }t>0,\,X_t=x)>0$ for all $x \in \mathbb{R}$, in other words $X_t$ can visit any point with some positive probability.

Remark$^1$: The aforemnetioned conditions imposed on f and g can be changed, the spirit of the question is given an SDE with some necessary regularity conditions the previously described result holds.

Remark$^2$: I had posted this question in math.stachexchange and I had not received an answer after an expired bounty and a passage of two weeks. I have now deleted the post on math.stackexchange.

$\endgroup$
  • $\begingroup$ What kind of additional hypotheses might you hope for? For example I think it is not hard to get such a result if $g$ is uniformly bounded below (away from zero) and $|f|$ is uniformly bounded above. $\endgroup$ – Ian Apr 17 '17 at 21:22
1
$\begingroup$

How about changing the time? Write $d\sigma(t) = (g(t))^2 dt$ and $dY_t = g(t) dB_t$. Then $\tilde{Y}_s = Y_{\sigma^{-1}(s)}$ is a Wiener process (provided that $g$ is nice enough; more precisely, we require $\sigma(t)$ to be finite for all $t > 0$ and unbounded), and if $\tilde{X}_s = X_{\sigma^{-1}(s)}$, then $$d\tilde{X}_s = \tilde{f}(s, \tilde{X}_s) ds + d\tilde{Y}_t,$$ where $\tilde{f}(s, x) = f(x) / (g(\sigma^{-1}(s)))^2$.

Now you should be able to use the Girsanov theorem, provided that $\tilde{f}$ is not too rough.

(This is more a comment than an answer, but I am not privileged enough to add comments).

$\endgroup$

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.