2

I am trying to prove that the solution of the SDE: $$ Z_t = Y_t + \int_0^t Z_s dX_s $$ is: $$ Z_t = \mathcal{E}(X)_t \left(y_0 + \int_0^t \mathcal{E}(X_s)^{-1}dY_s - \int_0^t \mathcal{E}(X)_s^{-1}d\langle X,Y\rangle_s\right) $$ where $\mathcal{E}(X)_t := \exp(X_t - \frac{1}{2}\langle X\rangle_t)$ is the stochastic exponential, and $y_0$ is the initial value of Y, and the initial value of $X$ is 0.

So far I have shown, by Ito's formula applied to $Y_t \mathcal{E}(X)_t^{-1}$, that: $$ Y_t = \mathcal{E}(X)_t\left( y_0 + \int_0^t \mathcal{E}(X)_s^{-1}dY_s - \int_0^t \mathcal{E}(X)_s^{-1}d\langle X,Y\rangle_s - \int_0^t Y_s\mathcal{E}(X)_s^{-1}dX_s + \int_0^t Y_s \mathcal{E}(X)_s^{-1} d\langle X\rangle_s\right) $$ which feels like it's going in the right direction as it's beginning to resemble the solution, but I don't know how to finish the proof.

Furthermore, I want to prove uniqueness of the solution after this, but I am not sure how I would do this?

user1598
  • 405

1 Answers1

4

Writing \begin{align} E_t&:=\mathcal{E}(X)_t\,,\\ F_t&:=y_0 + \int_0^t \mathcal{E}(X_s)^{-1}dY_s - \int_0^t \mathcal{E}(X)_s^{-1}d\langle X,Y\rangle_s \end{align} Applying Ito to $Z_t=E_tF_t$ gives \begin{align} dZ_t&=E_t\,dF_t+F_t\,dE_t+d\langle E,F\rangle_t\\[3mm] &=dY_t-d\langle X,Y\rangle_t+F_t\,E_t\,dX_t+d\langle X,Y\rangle_t\\[3mm] &=dY_t+Z_\,dX_t\, \end{align} as expected. Here I used \begin{align} dF_t&=E_t^{-1}\,dY_t-E_t^{-1}\,d\langle X,Y\rangle_t\,,\\[3mm] dE_t&=E_t\,dX_t\,,\\[3mm] d\langle E,F\rangle_t&=E_t E_t^{-1}\,d\langle X,Y\rangle_t=d\langle X,Y\rangle_t\,,\\[3mm] d\langle X,\langle X,Y\rangle\rangle_t&=0\,. \end{align} Regarding uniqueness: Let $Z$ and $Z'$ be solutions with $Z_0=Z'_0$. Then $\widetilde{Z}:=Z-Z'$ is a solution of $$ d\widetilde{Z}_t=\widetilde{Z}_t\,dX_t $$ with $\widetilde{Z}_0=0\,.$ Since this equation has a unique strong solution of the form $$ \widetilde{Z}_t=\widetilde{Z}_0\,e^{X_t-\frac{1}{2}\langle X\rangle_t} $$ it follows that $\widetilde{Z}_t\equiv 0$.

Kurt G.
  • 17,136
  • Thanks Kurt. It seems so simple when you write it like that - I get it now! Do you have any ideas on how you would show uniqueness? I know how I would show uniqueness for $Z_t = 1+\int_0^t Z_s dX_s$ (by showing that if $Z,Z'$ are solutions then $Z/Z'=1$ by Ito's lemma) but that method doesn't seem to work here because of the $Y_t$? – user1598 Apr 28 '22 at 15:29
  • 1
    General solutions for linear case : https://math.stackexchange.com/questions/1788853/solution-to-general-linear-sde – TheBridge Apr 28 '22 at 16:51
  • @TheBridge . Thanks for that link. – Kurt G. Apr 28 '22 at 17:01
  • 1
    @Jacob . I was not able to show uniqueness by considering $Z/Z'$ or $(Z-Y)/(Z'-Y)$ but it is probably simpler than one thinks. See edit. – Kurt G. Apr 28 '22 at 17:03