Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Let $B_t$ be a Brownian motion and $M_t=\max_{0\leq s\leq t}B_s$. Show that: $$(M_t-B_t)^4-6t(M_t-B_t)^2+3t^2$$ is a martingale for $t\geq0$.

share|cite|improve this question
up vote 0 down vote accepted

I would go this way using the fact that (see my comment below) :

$(M_t-B_t)$ and $|B_t|$ are processes equal in law.

I would use Itô's lemma with $f(x,t)=x^2-6t.x+3t^2$ and show (if true) that there is no drift.

The delicate part is then to justify the fact that if you have 2 processes equal in law and if one of them is a martingale then the other one is also a martingale.

Comment (edited thank's to did's comment):

For the fact that $(M_t-B_t)$ and $|B_t|$ are equal in law you can have a look at Karatzas and Shreve's book for a proof.

Best regards

share|cite|improve this answer
Suggestion: The processes $M-B=(M_t-B_t)_{t\geqslant0}$ and $|B|=(|B_t|)_{t\geqslant0}$ are equal in law. – Did Dec 17 '12 at 13:16
If we apply Ito's formula to $f(|B_t|,t)$, we have: $$f(|B_t|,t)=F(|B_0|,0)+\int_0^t12s|B_s|\text{d}s+\int_0^t4|B_s|^3-12s|B_s| \text{d} |B_s|+\\ +\frac{1}{2}\int_0^t12|B_s|^2-12s\text{d}{\left<|B|,|B|\right>}_s$$ How can we proceed? – Nick Papadopoulos Dec 17 '12 at 18:03
@Nick : Apply Itô not to $|B_t|$ but to $B_t^2$. I edit my post accordingly. – TheBridge Dec 17 '12 at 21:15

Observe that

$$(M_t-B_t)^4 = \left(\max_{0 \leq r \leq t} (B_r-B_t)\right)^4 = \max_{0 \leq r \leq t} ((B_r-B_t)^4) \\ = \max_{0 \leq r \leq s} ((B_r-B_s)^4) + \max_{0 \leq r \leq t-s} \left(((B_{s+r}-B_s)-(B_t-B_s))\right)^4$$


$$\max_{0 \leq r \leq t-s} ((B_{s+r}-B_s)-(B_t-B_s))^4 = \max_{0 \leq r \leq t-s} (W_r-W_{t-s})^4 = (M_{t-s}^\ast-W_{t-s})^4$$

where $W_{r} := B_{s+r}-B_s$ (is again Brownian motion), $M_t^\ast := \max_{0 \leq r \leq t} W_r$. From reflection principle we know that

$$M_{t-s}^\ast - W_{t-s} \sim \sqrt{\frac{2}{\pi \cdot (t-s)}} \cdot \exp \left(- \frac{x^2}{2(t-s)} \right) \cdot 1_{(0,\infty)}(x)$$

hence $$\mathbb{E}((M_{t-s}^\ast-W_{t-s})^4)=3(t-s)^2$$


$$\mathbb{E}((M_t-B_t)^4|\mathcal{F}_s) = (M_s-B_s)^4 + \mathbb{E} \left( \max_{0 \leq r \leq t-s} (B_{s+r}-B_t)^4 |\mathcal{F}_s \right) \\ =(M_s-B_s)^4 + \underbrace{\mathbb{E} \left( \max_{0 \leq r \leq t-s} (B_{s+r}-B_t)^4 \right)}_{\mathbb{E}((M_{t-s}^\ast-W_{t-s})^4)} =(M_s-B_s)^4 +3 \cdot (t-s)^2$$

Similar proof works for $(M_t-B_t)^2$. Adding it all up finishs the proof.

share|cite|improve this answer
@ saz : I am not sure to understand how you can switch expectation with max operator in your proof, could you please give some details ? Best regards – TheBridge Dec 17 '12 at 14:26
sorry I still don't get it, take 2 positive random variables, we usually don't have $E[X_1\vee X_2]=E[X_1]\vee E[ X_2]$ this is what I dont get in your proof even with the argument of monotone convergence. Best regards. – TheBridge Dec 17 '12 at 15:12
@TheBridge I thought that it would be simply monotone convergence for conditional expectations - but I see your point. I'll check whether I can fix it somehow... – saz Dec 17 '12 at 15:39
@TheBridge I corrected it. Hopefully it's fine now. – saz Dec 17 '12 at 16:01
I am convinced now. ps : the fact that processes $(M_t-B_t)$ and $|B_t|$ have the same law as also comes from reflection principle I think. – TheBridge Dec 17 '12 at 16:18

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.