I know that because $W_t$ is a martingale, $$E\left[\int_{0}^{T} W_t dW_t\right] = 0$$ then what should the value for this equation be: $$E\left[\int_{0}^{T} W_t^{n}dW_t\right]?$$ $n$ is the power of $W_t$, a constant.

  • $\begingroup$ So, what does the $n$ mean in your notation? Is it a $n$-dimensional Brownian motion? $\endgroup$
    – Olorun
    Commented Dec 10, 2015 at 2:19
  • $\begingroup$ @Olorun It's just a constant n $\endgroup$ Commented Dec 10, 2015 at 2:21
  • $\begingroup$ @Olorun Its a power for $W_t$ $\endgroup$ Commented Dec 10, 2015 at 2:26
  • 2
    $\begingroup$ Note that $$\int_0^T W_t \, dW_t$$ does NOT equal $0$. The expectation of the randm variable is zero, but not the random variable itself. $\endgroup$
    – saz
    Commented Dec 10, 2015 at 6:41
  • 1
    $\begingroup$ It turns out that the expectation of "essentially" any function integrated against $dW_t$ (in the sense of Ito) is zero. $\endgroup$
    – Ian
    Commented Dec 10, 2015 at 23:46

2 Answers 2


If $f: (0,\infty) \times \Omega \to \mathbb{R}$ is a progressively measurable function such that

$$\mathbb{E} \left( \int_0^T f(t)^2 \, dt \right) < \infty \quad \text{for all $T>0$}, \tag{1}$$


$$M_T := \int_0^T f(t) \, dW_t$$

defines a martingale. In particular, we have $\mathbb{E}(M_T) = \mathbb{E}(M_0)=0$, i.e.

$$\mathbb{E} \left( \int_0^T f(t) \, dW_t \right) = 0.$$

Since the Brownian motion $(W_t)_{t \geq 0}$ is progressively measurable (because of the continuous sample paths), it suffices to check that $f(t,\omega) := W_t(\omega)^n$ satisfies $(1)$ in order to conclude

$$\mathbb{E} \left( \int_0^T W_t^n \, dW_t \right)=0.$$

To check $(1)$ we have to use that each $W_t$ has moments of arbitrary order. One possibility is using the scaling property, i.e. $W_t \stackrel{d}{\sim} t W_1$, and Tonelli's theorem:

$$\begin{align*} \mathbb{E} \left( \int_0^T f(t)^2 \, dt \right) &= \mathbb{E} \left( \int_0^T W_t^{2n} \, dt \right) \\ &= \int_0^T \mathbb{E}(W_t^{2n}) \, dt \\ &= \mathbb{E}(W_1^{2n}) \int_0^T t^{n} \, dt < \infty. \end{align*}$$


Another way to look at it is writing in differential form: $$ dX_t = 0 dt + W_t dW_t$$ It's easy to see $X_t$ has a zero drift term (i.e. $dt$ term is zero) and hence is a local martingale. Same argument holds if $W_t$ is raised to some power.

The expectation of both of your integrals will be zero.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .