1
$\begingroup$

Consider the stochastic ODE $$ dX = F(X)dt + dB $$ where $B$ is Brownian motion. If the drift $F$ is locally Lipschitz, then the solution exists and is unique over $[0,T]$ where $T$ is an "almost surely positive" random time.

My question is, can we find a measure 1 set and a deterministic constant $C$, such that on this measure 1 set the random time $T$>$C$? This seems stronger than saying "$T$ is almost surely positive" random time.

$\endgroup$
1
$\begingroup$

In general, the answer to your question is no. For instance, consider $$ d X(t) = X(t)^2 dt + \sigma d B(t) \;, \quad X(0) = x_0 \;. $$ In this case, the drift $F(x) = x^2$ is locally Lipschitz since its derivative is locally bounded. When $\sigma =0$, we obtain the ODE $$ \dot{x}(t) = x(t)^2 \;, \quad x(0) = x_0 $$ with solution $$ x(t) = \frac{x_0}{1-x_0 t} $$ which is well-defined for $0 \le t < 1/x_0$. Thus, for any $C>0$ and $x_0 < 1/C$, the ODE solution exists and is unique over $[0,C]$.

The situation is different for the SDE: in any time interval, there is a nonzero (but possibly tiny) probability that $X(t)$ reaches a state where the drift dominates. From this point forward, with high probability, the SDE solution evolves like the ODE solution but with a random initial condition. Hence, the solution is defined up to a random time.

This phenomenon is illustrated in the figure below with $x_0 = 1/10$ and $C=10$. In red is the ODE solution for comparison. In different shades of grey are ten independent realizations of $X(t)$.

enter image description here

$\endgroup$

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

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