2
$\begingroup$

I was wondering if anything could be said at all about the well-psedness of the following time-inhomogeneous singular diffusion SDE:

\begin{align}d X_t&=\sigma(X_t,t ) d W_t , \qquad t\geq 0, \, X_0 \in \mathbb R \\ \sigma(x,t )&=t^{-\alpha} b(x) \qquad \, \alpha>0, \end{align}

with $b$ as nice as required (say bounded, or Lipschitz continuous and sublinearly growing).

I can imagine for existence one needs at least $\alpha <1/2$ so that $\sigma \in L^2(\mathbb R, [0,T])$. Perhaps something could be achieved by time reversal?

The following paper may be relvant, but I am not entirely sure of its applicability

https://projecteuclid.org/journals/annals-of-probability/volume-46/issue-3/Strong-solutions-to-stochastic-differential-equations-with-rough-coefficients/10.1214/17-AOP1208.pdf

$\endgroup$

1 Answer 1

4
$\begingroup$

It seems natural to view the Itô process on a different time scale where it becomes a Itô diffusion, and in turn, invoke existence/uniqueness theory for Itô diffusion.

In particular, under the (deterministic) time change given by $\tau(t) = (t-2 \alpha t )^{\frac{1}{1-2 \alpha}}$ (so that $\tau'(t) = \tau(t)^{2 \alpha}$ and $\tau(0)=0$), the time-changed process is (in law) an Itô diffusion that satisfies $$ d \tilde{X}_t = b(\tilde{X}_t) d B_t $$ where $\tilde{X}_t = X_{\tau(t)}$ and $B_t=\int_0^{\tau(t)} s^{-\alpha} dW_s$ is again a standard Brownian motion. Note that this time change only makes sense for $\alpha \in [0,1/2)$.

Sufficient conditions for existence and uniqueness of the time-changed process can be found in, e.g., Section 1.2 of the monograph

Cherny, Alexander S.; Engelbert, Hans-Jürgen, Singular stochastic differential equations., Lecture Notes in Mathematics 1858. Berlin: Springer (ISBN 3-540-24007-1/pbk). viii, 128 p. (2005). ZBL1071.60003.

For more about the correspondence between an Itô diffusion and Itô process via time change, see

Øksendal, Bernt, When is a stochastic integral a time change of a diffusion?, J. Theor. Probab. 3, No. 2, 207-226 (1990). ZBL0698.60046.

$\endgroup$
1
  • $\begingroup$ That's crystal clear, thanks $\endgroup$
    – Mr_3_7
    Commented Jul 19, 2022 at 13:57

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .