On a multi-dimensional Poissonian pair correlation concept and uniform distribution (Q2274064)

For an element $\mathbf{x}=(x_1, \ldots, x_d)$ in $[0, 1)^d$, the authors define $\|\mathbf{x}\|_{\infty}$ to be the maximum of the quantities $\|x_1\|,\ldots \|x_d\|$, where $\|\cdot\|$ is the so-called distance to the nearest integer. A sequence $(\mathbf{x}_n)_n$ is said to have \textit{Poissonian pair correlations} if the pair correlation statistics \[ F_N^{(d)}(s) := \frac{1}{N}\left|\left\{1 \le l \neq m \le N : \|\mathbf{x}_l -\mathbf{x}_m\|_{\infty} \le \frac{s}{N^{1/d}}\right\}\right| \] tends to $(2s)^d$, for every $s\ge0$, as $N \rightarrow \infty$. This a generalization of the case $d=1$. As in the case $d=1$, for a sequence $(\mathbf{X}_n)_n$ of i.i.d. uniform random variables on $[0, 1)^d$, $F_N^{(d)}(s)$ tends to $(2s)^d$ almost surely, as $N$ tends to infinity. The proof is similar to that of the case $d=1$. For a sequence $(\mathbf{x}_n)_n$, the Poissonian pair correlations property implies the uniform distribution. The authors raise the question to know whether this property would be still valid under a different choice of a norm on $[0,1)^d$, for example an $\ell^2$ norm. For an increasing sequence of integer $\mathcal{A}=(a_n)_n$, some upper bound on its \textit{additive energy} $E_{\mathcal{A}_N}$ \; [the number of the quadruples $(a, b, c, d)$ in $\mathcal{A}_N = \{a_n : n\le N\}$ such that $a+b=c+d$] \; implies that for almost all real $\alpha$, the sequence $(\{a_n \alpha\})_n$ has Poissonian pair correlations. The authors consider $d$-dimensional sequences $(\{a_n \boldsymbol{\alpha}\})_n$, where ${u\boldsymbol{\alpha}}=(\{u\alpha_1\},\ldots, \{u\alpha_d\})$ and show that if \[ d \ge 2 \; \text{ and if, for any } C > 1, \text{ one has } E_{\mathcal{A}_N} = O\left(\frac{N^3}{(\log N)^C} \right)\text{ then } \] for almost all $\boldsymbol{\alpha} \in \mathbb{R}^d$, the sequence $(\{a_n \boldsymbol{\alpha}\})_n$ has Poissonian pair correlations. The authors notice that for $d=1$, the above relation is known only for \textit{some} $C$ and is suspected to hold for any $C>1$, a result due to \textit{T. F. Bloom} and \textit{A. Walker} [Isr. J. Math. 235, No. 1, 1--11 (2020; Zbl 1450.11094), Preprint, \url{arXiv:1806.07849}]. The difference comes from some power in the $\gcd$ to be averaged. On the other direction, if $E_{\mathcal{A}_N}$ is not a $O\left(N^3\right)$, then for any $\boldsymbol{\alpha} \in \mathbb{R}^d$, the sequence $(\{a_n \boldsymbol{\alpha}\})_n$ does not have Poissonian pair correlations.