I would like to get help in solving an Ito stochastic equation: $dY_t=-dW_t \, (Y_t^2+1)$

The process $W_t$ is the standard Brownian motion. Is it possible to get a path solution solution in terms of Brownian motion integrals?

  • $\begingroup$ How do you know that there exists a solution to the SDE? The coefficient does not even satisfy a linear growth condition. $\endgroup$ – saz Apr 27 '14 at 10:37
  • $\begingroup$ I agree that the coefficient does not satisfy the Lipschitz global condition, however I have simulated numerically the equation and at least in short finite time it does not diverge. Hence, I conclude that it is solvable at least until a finite explosion time. $\endgroup$ – user48672 Apr 28 '14 at 16:20

A strong solution of $\mathrm dY_t=\sigma(Y_t)\mathrm dW_t$ defined for every $t$ small enough is $Y_t=B_{\tau(t)}$ where $B$ is a Brownian motion and $\tau(t)=\inf\{s\mid A(s)\gt t\}$ with $$ A(t)=\int_0^t\frac{\mathrm ds}{\sigma(B_s)^2}. $$ In particular, $Y_t$ is defined for every $t$ such that $\tau(t)$ is finite hence $Y_t$ is defined for every $t$ if and only if $A_\infty=\sup\limits_{t\gt0}A(t)$ is infinite.

In the present case, $\sigma(x)=1+x^2$ hence $\sigma(x)^2\leqslant4$ on $|x|\leqslant1$ and $$ A(t)\geqslant\frac14\,\mathrm{Leb}\{s\leqslant t\mid|B_s|\leqslant1\}. $$ The Brownian motion $B$ is recurrent hence $A_\infty=\infty$ almost surely, QED.

Recall that explosions of solutions of stochastic differential equations can occur due to drift terms, for example the solutions of $\mathrm dX=X^2\mathrm dW+X^3\mathrm dt$ are given, for $t$ small enough, by $$ \frac1{X_t}=\frac1{X_0}-W_t, $$ hence $X$ is defined only up to the almost surely finite random time $T=\inf\{t\mid W_t=1/X_0\}$.

  • $\begingroup$ Thanks a lot for the feedback! $\endgroup$ – user48672 Jun 20 '14 at 14:50

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.