Irregularity of distribution in Wasserstein distance (Q2007445): Difference between revisions

The author considers the non-uniformity of probability measures on the (unit) interval and the circle, by making use of the so-called Wasserstein-\(p\) distance. To be more precise, let \(X\) be equal to \([0,1)\) or to \(\mathbb{R}/\mathbb{Z}\), and let \(\lambda\) be the uniform measure on \(X\). Furthermore, let \(\mathcal{P} (X)\) be the space of probability measures on \(X\). For \(p\in [1,\infty]\) and \(\mu\in\mathcal{P} (X)\), the Wasserstein-\(p\) distance between \(\lambda\) and \(\mu\) is defined as \[ W_p^X (\mu,\lambda):=\| \mu - \lambda \|_{\dot{W}^{-1,p}(X)}, \] where \(\|\cdot\|_{\dot{W}^{-1,p}(X)}\) is a negative Sobolev norm. Now, for a given sequence of points \((x_n)_{n\in\mathbb{N}}\) in \([0,1)\) and \(N\in \mathbb{N}\), define \[ \mu_N:=\frac{1}{N}\sum_{n=1}^N \delta_{x_n}. \] One of the main results in the paper is the following estimate. For \(X\) as above, for every \(p\in [1,\infty]\) there exists a positive constant \(C_p\) such that for any sequence of points \((x_n)_{n\in \mathbb{N}}\) in \(X\) it is true that \[ W_p^X (\mu_N,\lambda)\ge C_p \frac{(\log N)^{\alpha_p}}{N} \] holds for infinitely many \(N\in\mathbb{N}\). Furthermore, this bound is sharp. Here, \(\alpha_p\) is \(1/2\) for finite \(p\), and 1 if \(p=\infty\). Moreover, the author shows an explicit upper bound on non-uniformity, and applies these findings to the equidistribution of quadratic residues in finite fields.