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

Is a continuous supermartingale with vanishing drift already a martingale? In my concrete problem, I have a continuous nonnegative local martingale $ (X_t) $ on $ \left[0, T\right] $ which is bounded from below by 0 (and hence a supermartingale) and which in addition satisfies $ sup_{t \in \left[0,T\right]} \mathbb{E} \left[ X_t^p \right] < + \infty$, for some $p > 1$. When I apply Ito, I find

$dX_t = 0 dt + H dW_t$,

where W is a Brownian motion and H is some predictable process. Now, I am wondering if $(X_t)$ is in fact a true martingale (the reason is that I would like to apply the Doob inequality for martingales in $L^p$).

Thanks a lot for your help! Simon

share|cite|improve this question
What is the drift term of a (general?) supermartingale? What does it mean that it is vanishing? Maybe the following fact is useful: if $X_t$ is a supermartingale and $\mathsf E X_t$ is constant in time, then $X_t$ is a martingale. Perhaps, some integrability condition are needed though - if this is something you are looking for, I'll fish them out. – Ilya Jan 9 '13 at 13:36
there is a standard example, which is either the reciprocal of a bessel 3 or of a squared bessel 3 in which routine ito formula calculations show that the drift term is 0, but the process is a strict local martingale, and a supermartingale (by virtue of positivity), but not a martingale. I'm told Chung & William's book works it out in great detail. – mike Jan 9 '13 at 14:35

Take $dX_t= f(t)dt+ dW_t$ such that :
-$F(t)=\int_0^t f(s)ds$ is strictly decreasing (determisitic)
-$f(t)\to 0$ as $t\to \infty$

Then $X_t= F(t)+ W_t$ and as the sum of a martingale term (the Brownian motion) plus a deterministic supermartingale $F(t)$, $X$ it is a supermaringale with a drift term that goes to 0, and $X$ is not be a martingale (unless F is null).

For example $f(t)=-\frac{1}{(1+t)^2}$ gives $F(t)=\frac{1}{1+t}-1$ seems to work fine.

Best regards

share|cite|improve this answer

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.