1
$\begingroup$

Let $(p_t)_{t >0}$ be the Gaussian heat kernel on $\mathbb R^d$ and $(P_t)_{t >0}$ its induced semi-group, i.e., $$ \begin{align} p_t (x) &:= (4\pi t)^{-\frac{d}{2}} e^{-\frac{|x|^2}{4t}}, \quad t>0, x \in \mathbb R^d, \\ P_t f (x) &:= \int_{\mathbb R^d} p_t ( x- y) f(y) \, \mathrm d y. \end{align} $$

Let $\nu$ be a Borel probability measure on $\mathbb R^d$ with finite first moment. We denote by $\ell_\nu$ the probability density function of $\nu$. For $R>0$, let $B_R$ be the open ball centerted at $0$ with radius $R$. Let $B_R^c := \mathbb R^d \setminus B_R$ and $$ I_R := \int_{B_R^c} |x| (P_t \ell_\nu) (x) \, \mathrm d x, \quad R >0. $$

Is there an upper bound of $I_R$ in terms of $R, \nu, t$?

Thank you so much for your elaboration!

$\endgroup$

1 Answer 1

1
$\begingroup$

Let $Z$ be a standard normal random vector in $\mathbb R^d$. Let $Y$ be a random vector in $\mathbb R^d$, independent from $Z$, with distribution $\nu$. Then, for each $s\in(0,1)$, $$I_R=E|\sqrt{2t}\,Z+Y|\,1(|\sqrt{2t}\,Z+Y|\ge R) \\ \le E(\sqrt{2t}\,|Z|+|Y|) \,[1(\sqrt{2t}\,|Z|\ge sR)+1(|Y|\ge(1-s)R)] \\ =\sqrt{2t}\,E|Z|1(|Z|\ge sR/\sqrt{2t}) \\ +\sqrt{2t}\,E|Z|\,P(|Y|\ge(1-s)R) \\ +E|Y|\,P(|Z|\ge sR/\sqrt{2t}) \\ +E|Y|\,1(|Y|\ge(1-s)R). $$

In turn, you can upper-bound the latter four summands using the following applications of Markov's inequality: $$ E|Z|1(|Z|\ge sR/\sqrt{2t})\le \frac{E|Z|^2}{sR/\sqrt{2t}}=\frac n{sR/\sqrt{2t}}, \\ E|Z|\,P(|Y|\ge(1-s)R)\le\sqrt{E|Z|^2}\,\frac{E|Y|}{(1-s)R} =\frac{\sqrt n\,E|Y|}{(1-s)R} \\ P(|Z|\ge sR/\sqrt{2t})\le\frac{E|Z|^2}{(sR/\sqrt{2t})^2} =\frac{n}{(sR/\sqrt{2t})^2} \\ E|Y|\,1(|Y|\ge(1-s)R)\le\frac{E|Y|^2}{(1-s)R}. $$

$\endgroup$

Your Answer

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

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