$\newcommand{\Ex}{\mathbb E}$ I'm reading an argument in the proof of Proposition 3.8. in the paper Nonlinear self-stabilizing processes - I Existence, invariant probability, propagation of chaos.

Let $X_0$ and $X_0^{\prime}$ be two independent real-valued random variables, having the same distribution. We consider two independent $1$-dimensional Brownian motions $B$ and $B'$. Let $X$ and $X'$ be solutions of $$ X_t=X_0+B_t-\frac{1}{2} \int_0^t b\left(s, X_s\right) \mathrm{d} s, $$ and $$ X_t^{\prime}=X_0^{\prime}+B_t^{\prime}-\frac{1}{2} \int_0^t b\left(s, X_s^{\prime}\right) \mathrm{d} s, $$

where $b:\mathbb R_{\ge 0} \times \mathbb R \to \mathbb R$ is regular enough. Let $$ Y_t := X_t-X_t^{\prime} \quad \text{and} \quad \mu_n(t) := \Ex \left ( \left|Y_t\right|^n \right), \quad n \geqslant 2 . $$

Then $Y$ is a semi-martingale with decomposition $$ Y_t=Y_0+B_t-B_t^{\prime}-\frac{1}{2} \int_0^t\left(b\left(s, X_s\right)-b\left(s, X_s^{\prime}\right)\right) \mathrm{d} s . $$

We apply the Itô's formula and take the expectation and the derivative. We obtain $$ \mu_{2 n}^{\prime}(t)=n\left\{2(2 n-1) \mu_{2 n-2}(t) - \Ex \left[Y_t^{2 n-1}\left(b\left(t, X_t\right)-b\left(t, X_t^{\prime}\right)\right)\right]\right\}. \tag{1}\label{1} $$

My understanding By Itô's lemma, $$ \begin{align*} & (Y_t)^{2n} - (Y_0)^{2n} \\ = & 2n \int_0^t (Y_s)^{2n-1} \mathrm{d} Y_s + \frac{2n(2n-1)}{2} \int_0^t (Y_s)^{2n-2} \mathrm{d} \langle Y \rangle _s. \end{align*} $$

We have $$ \begin{align*} \mathrm{d} Y_s &= \mathrm{d} B_s - \mathrm{d} B'_s- \frac{\left(b\left(s, X_s\right)-b\left(s, X_s^{\prime}\right)\right) }{2} \, \mathrm{d} s, \\ \mathrm{d} \langle Y \rangle_s &= \mathrm{d} \langle B - B' \rangle _s = 2 \, \mathrm{d} s. \end{align*} $$

Hence $$ \begin{align*} \Ex[(Y_t)^{2n}] - \Ex [(Y_0)^{2n}] &= -n \int_0^t \Ex [ (Y_s)^{2n-1} \left(b\left(s, X_s\right)-b\left(s, X_s^{\prime}\right)\right)] \, \mathrm{d} s \\ & \qquad + 2n(2n-1) \int_0^t \Ex [ (Y_s)^{2n-2} ] \, \mathrm{d} s. \end{align*} $$

So $$ \begin{align*} \mu_{2n} (t) - \mu_{2n}(0) &= -n \int_0^t \Ex [ (Y_s)^{2n-1} \left(b\left(s, X_s\right)-b\left(s, X_s^{\prime}\right)\right)] \, \mathrm{d} s \\ & \qquad + 2n(2n-1) \int_0^t \mu_{2n-2} (s) \,\mathrm{d} s. \tag{2}\label{2} \end{align*} $$

Could you please explain how to go from $\ref{2}$ to $\ref{1}$?


1 Answer 1


I believe you meant lemma 3.8. Here at the final step of your calculation, you just apply Lebesgue-differentiation theorem since those quantities are continuous and the integral is Lebesgue. That will give you the $\mu'(0)$.

So to get the $\mu'(t_{0})$, I would apply Ito's formula with starting points $t=t_{0}$ as opposed to $t=0$. So the integrals will be $\int_{t_{0}}^{t_{0}+t}$. Meaning apply Ito's lemma separately to $t_{0}$ and to $t_{0}+t$ and then take their difference.

  • $\begingroup$ You're right! I got it. Thank you so much for your help! $\endgroup$
    – Akira
    Feb 28 at 14:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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