Apparent misunderstanding of the solution to an SDE

111 Views Asked by At

I came across the following question in a past paper for an SDE course I am following: Let $(B_t)_{t\geq 0}$ b on the filtered probability space $(\Omega,\mathcal F,(\mathcal F_t)_{t\geq 0},\mathbb P)$. Consider the SDE $$dX_t=-\frac{1}{2}X_tdt+\sqrt{1-X_t^2}dB_t,\quad\quad X_0=0. \tag{1}\label{eq1}$$ The question as given has two parts:
a) Find a solution of the form $f(B_t)$ and investigate the properties $f$ should have.
b) Show that a solution exists for $t<\tau:=\inf\{s>0:B_s\notin[-\frac{\pi}{2},\frac{-\pi}{2}]\}$.

Using It$\hat{\text{o}}$'s formula it is easy to derive that we require $f''(x)=-f(x)$ and $f'(x)=\sqrt{1-f(x)^2}$ for all real $x$. The only real valued function I'm aware of satisfying these properties is $\sin$, so we set $f(x)=\sin(x)$ on $\mathbb R$. As we derived $f$ via It$\hat{\text{o}}$'s formula we know then that $X_t=\sin(B_t)$ satisfies \eqref{eq1} (technically we should check it but that is easy to do). I believe this is a satisfactory answer to part a) (please inform me if it is not).

I am not quite sure what is being asked in part b), however. As far as I am aware we can say that $X_t$ is a solution to the SDE \eqref{eq1} on an interval $I:=[0,T]$ if it satisfies the integral form of the SDE for every $t\in I$, which we know $\sin(B_t)$ does, and if it is continuous and $(\mathcal F_t)$ adapted (which $\sin(B_t)$ is), and if $$\int_0^T\frac{1}{2}dt+\int_0^T\sqrt{1-X_t^2}dt<\infty\qquad \text{a.s}.$$ Now as far as I am aware $X_t=\sin(B_t)$ satisfies all 3 of the above listed properties for any $T>0$, so I would just have concluded that SDE \eqref{eq1} has a solution no matter the value of $t$. Question b) however makes me think something funky happens outside of $[-\frac{\pi}{2},\frac{-\pi}{2}]$, which I presume has something to do with $\sin$ being monotone increasing on that interval. Other than that I am clueless as to how to answer b) and I am worried that it implies I have a deep misunderstanding of the subject. Any comments on the question and my thought processes would be welcome.

1

There are 1 best solutions below

2
On BEST ANSWER

No, there is no deep misunderstanding. You are simply missing the fact that the identity

$$\cos(x) = \sqrt{1-\sin(x)^2}$$

holds if, and only if, $\cos(x) \geq 0$. In particular,

$$\cos(B_t) = \sqrt{1-X_t^2} = \sqrt{1-\sin(B_t^2)}$$

fails to hold when the Brownian motion exits the interval $[-\pi/2,\pi/2]$.