Let $X_t$ be such that $X_t$ is bounded continuous martingale adapted to the filtration $\mathcal{F}_t$ such that

$$\Bbb{E} \bigg[\int_0^t e^{X_s} \, d\langle X\rangle_s\bigg] = 0$$

and $X_0=0$. Does it follow that $X_t$ is identically zero?

remark: This question came from the following $$\forall \lambda >0 ; e^{\lambda X_t} \text{ is a martingale} \Rightarrow X_t \equiv 0$$

I see two ways to solve this:

0) $\Bbb{E}[e^{\lambda (X_t - X_s)}\vert \mathcal {F}_s] = 1$ therefore

$$\Bbb{E}[e^{\lambda (X_t - X_s)} - 1\vert \mathcal {F}_s] = 0\\ \Bbb{E}\bigg[\frac{e^{\lambda (X_t - X_s)}-1}{\lambda}\vert \mathcal {F}_s\bigg] = 0\\ \lim_{\lambda\to 0}\Bbb{E}\bigg[\frac{e^{\lambda (X_t - X_s)}-1}{\lambda}\vert \mathcal {F}_s\bigg] = 0\\ \Bbb{E}\bigg[\frac{\lim_{\lambda\to 0}e^{\lambda (X_t - X_s)}-1}{\lambda}\vert \mathcal {F}_s\bigg] = 0 \\ \Bbb{E}[X_t - X_s\vert \mathcal {F}_s] = 0$$

therefore $X_t$ is a martingale.

A) Since the exponential transform (extend it to complex you get characteristic functions, or just consider the laplace transform) allows one to recover the distribution, we get that $$\Bbb{E}[e^{\lambda X_t}] = 1 = \Bbb{E}[e^{\lambda 0}] $$

B) by Itô

$$e^{X_t} = 1 + \int_0^t e^{X_s} \, dX_s + \int_0^t e^{X_s}\, d\langle X\rangle_s $$ take expectations

$$\Bbb{E}[e^{X_t}] = 1 + \int_0^t \Bbb{E}[e^{X_s}]\, d\langle X\rangle_s $$

So basically I am asking how to proceed along the lines of $B)$

  • $\begingroup$ What about $X_t := 1$? (Or have you forgotten mentioning that $X_0=0$?) $\endgroup$ – saz Jul 5 '15 at 19:51
  • $\begingroup$ Yes $X_0 = 0$, thank you $\endgroup$ – Conrado Costa Jul 5 '15 at 19:59

Recall that for any function $f \geq 0$ and any measure $\mu$ we have

$$\int f \, d\mu = 0 \implies f=0 \quad \mu\text{-a.s.} $$

Therefore, as $(\langle X \rangle_t)_{t \geq 0}$ is an increasing process,

$$\mathbb{E} \left( \int_0^t e^{X_s} \, d\langle X \rangle_s \right)=0$$


$$\int_0^t e^{X_s} \, d \langle X \rangle_s =0$$

$\mathbb{P}$-almost surely. Since $e^x$ is strictly positive and $(\langle X \rangle_t)_{t \geq 0}$ an increasing continuous process, this already implies

$$\langle X \rangle_t = \text{const} = \langle X \rangle_0 = 0$$

almost surely. Indeed: Suppose that it is not constant. Then we can pick $t_0 >0$ such that $\langle X \rangle_{t_0}(\omega) > \langle X \rangle_{0}(\omega)$. By the very definition of the Riemann-Stieltjes integral and the strict positivity of $e^x$, this implies

$$\int_{0}^{t_0} e^{X_s(\omega)} \, d\langle X \rangle_s(\omega)>0$$

in contradiction to our assumption.


$$\mathbb{E}(X_t^2) = \mathbb{E}(\langle X \rangle_t) = 0$$

which implies $X_t = 0$ almost surely.


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.