An \(L^2\)-identity and pinned distance problem (Q1734000)

Let $d\geq 2$. Given a set $E\subseteq \mathbb{R}^{d}$, we define its distance set by \[ \Delta (E):=\left\{\left|x-y\right|: x,y\in E\right\}. \] Intuitively, it is clear that for sets $E\subseteq\mathbb{R}^{d}$ which are sufficiently large, the distance set $\Delta(E)$ must be of full Hausdorff dimension. We let \[ r_{d}:=\sup\left\{r>0: \exists E\subseteq\mathbb{R}^{d}\text{ with }\dim_{H}(E)=r\text{ and }\dim_{H}(\Delta(E))<1\right\}.\] In other words, $r_{d}$ is the smallest number $r\geq 0$ with the property that \[ E\subseteq\mathbb{R}^{d},\, \dim_{H}(E)>r\,\mathbb{R}\rightarrow\, \dim_{H}(\Delta(E))=1. \] The \textit{Falconer distance conjecture} of \textit{K. J. Falconer} [Mathematika 32, 206--212 (1985; Zbl 0605.28005)] proposes that $r_{d}=\frac{d}{2}$ and remains unproven, despite receiving attention from several mathematician. Currently the best known estimate is $r_{d}\leq \frac{d}{2}+\frac{1}{3}$, due to \textit{T. Wolff} [Int. Math. Res. Not. 1999, No. 10, 547--567 (1999; Zbl 0930.42006)] for $d=2$ and \textit{M. B. Erdoğan} [Int. Math. Res. Not. 2005, No. 23, 1411--1425 (2005; Zbl 1129.42353)] for $d\geq 3$. \par The article under review considers a stronger version of the Falconer distance conjecture, namely the \textit{pinned distance problem}, which is formulated as follows: Given a set $E\subseteq \mathbb{R}^{d}$, define its \textit{pinned distance set} with respect to a point $x\in\mathbb{R}^{d}$ by \[ \Delta_{x}(E):=\left\{\left|x-y\right|: y\in E\right\}. \] Determine the quantity \[ s_{d}:=\sup\left\{s>0: \text{$\exists E\subseteq\mathbb{R}^{d}\text{ with }\dim_{H}(E)=s\text{ and }\dim_{H}(\Delta_{x}(E))<1\text{ for all }x\in E$}\right\}, \] that is, determine the smallest number $s\geq 0$ with the property \[ E\subseteq\mathbb{R}^{d},\, \dim_{H}(E)>s\, \mathbb{R}\rightarrow\, \exists x\in E\,\text{such that}\,\dim_{H}(\Delta_{x}(E))=1. \] Since $\Delta_{x}(E)\subseteq \Delta(E)$ for every $x\in E$, we have $r_{d}\leq s_{d}$. To approach the pinned distance problem, we may consider, for a given set $E\subseteq \mathbb{R}^{d}$, the set \[ T_{E}:=\left\{x\in \mathbb{R}^{d}: \left|\Delta_{x}(E)\right|=0\right\} \] and seek upper estimates of its Hausdorff dimension. Here $|\cdot|$ denotes the Lebesgue measure. \textit{Y. Peres} and \textit{W. Schlag} [Duke Math. J. 102, No. 2, 193--251 (2000; Zbl 0961.42007)] prove the bound \[ \dim_{H}(T_{E})\leq d+1-\dim_{H}(E), \] for every $E\subseteq\mathbb{R}^{d}$. This implies $s_{d}\leq \frac{d}{2}+\frac{1}{2}$. In the article under review, the author improves on Peres and Schlag's inequality, proving \[ \dim_{H}(T_{E})\leq \inf\left\{s\geq 0: \dim_{H}(E)+\beta(s)>d\right\}, \] where $\beta(s)=\frac{d+2s-2}{4}$ for $s\in\left[\frac{d}{2},\frac{d}{2}+1\right]$. This implies the improved bound $s_{d}\leq \frac{d}{2}+\frac{1}{3}$. \par The proof works by reducing the pinned distance problem, to a question of existence of measures satisfying a certain integral condition. The new key tool introduced in this paper is an integral identity for Schwartz functions $f$ of $\mathbb{R}^{d}$, namely \[ \int_{0}^{\infty}\left|\omega_{t}*f(x)\right|^{2}t^{d-1}dt=\int_{0}^{\infty}\left|\widehat{\omega_{r}}*f(x)\right|^{2}r^{d-1}dr, \] where $\omega_{t}$ denotes the normalised surface measure on $tS^{d-1}$ and $\widehat{\cdot}$ denotes the Fourier transform. This identity is proved using a group action argument.