Generalized visibility of lattice points in higher dimensions (Q2167501)

Given a lattice \(\Lambda\) and a fixed point \(\vec{u} \in \Lambda\), we say that \(\vec{n} \in \Lambda\) is \textit{visible} from \(\vec{u}\) if \(\vec{n} \neq \vec{u}\) and there is no other lattice point on the line segment connecting \(\vec{n}\) and \(\vec{u}\). In the typical integer lattice \(\mathbb{Z}^k\) for \(k \geq 2\), classical result of Christopher shows that the proportion of visible lattice points from the origin is \(1/\zeta(k)\). There is a related notion of being \textit{joinly visible} from multiple points. Formally, given a set of watch-points \(\mathcal{W} \subset \mathbb{Z}^k\) such that all points in \(\mathcal{W}\) are visible to all other points in \(\mathcal{W}\), one can look at the proportion of lattice points that are simultaneously visible to all points in \(\mathcal{W}\). This was studied previously by \textit{D. Rearick} [Norske Vid. Selsk. Forhdl. 39, 41--45 (1966; Zbl 0161.04903)]. There is yet another related notion of \textit{visibility along curves}. Given \(\vec{b} = (b_1, \ldots, b_k) \in \mathbb{N}^k\) with \(\gcd(b_1, \ldots, b_k) = 1\) and \(\vec{u} = (u_1, \ldots, u_k) \in \mathbb{Z}^k\), we say that a curve \(f: \mathbb{R} \longrightarrow \mathbb{R}^k\) through the point \(\vec{u}\) is type-\(\vec{b}\) if it is given by \(f(t) = (u_1 + a_1 t^{b_1}, \ldots, u_k + a_k t^{b_k})\) for some \(\vec{a} = (a_1, \ldots, a_k) \in \mathbb{R}^k\) with \(\vec{a} \neq \vec{0}\). Then \(\vec{n}\) is said to be \(\vec{b}\)-visible from \(\vec{u}\) if \(\vec{n} \neq \vec{u}\) and there are no other lattice points on the type-\(\vec{b}\) curve segment connecting \(\vec{n}\) and \(\vec{u}\). This paper combines these two perspectives and considers joint visibility of lattice points in \(\mathbb{Z}^k\) along curves. The authors also consider joint visibility along curves of different types. Their primary result is Theorem 1.1, which quantifies joint visibility along curves. To state it, we give a bit of additional notation. For \(\vec{b} = (b_1, \ldots, b_k) \in \mathbb{N}^k\) with gcd \(1\), define \(s(\vec{b}) = b_1 + \cdots + b_k\). Fix a set of watch-points \(\mathcal{W} \subset \mathbb{Z}^k\) such that every pair of points in \(\mathcal{W}\) are \(\vec{b}\)-visible to each other. Let \(N\) be the size of \(\mathcal{W}\), and suppose that \(1 \leq N < 2^{s(\vec{b})}\). Finally, let \(N_1(x)\) be the number of lattice points in \([1, x]^k \cap \mathbb{Z}^k\) which are simultaneously \(\vec{b}\)-visible from all watch-points in \(\mathcal{W}\). Then they show that \[ N_1(x) = x^k \prod_p \left( 1 - \frac{N}{p^{s(\vec{b})}} \right) + O(x^{k-1} \log^N x + x^{k - \frac{k(s(\vec{b}) - 1)}{N + s(\vec{b}) - 1} + \epsilon}) \] for any \(\epsilon > 0\), where \(p\) runs over all primes and the implicit constant depends on \(k, \vec{b}, \mathcal{W}\), and \(\epsilon\). Restricting to straight lines, their result improves on Rearick's earlier result. Their secondary result is Theorem 1.3, which quantifies joint visibility with curves of different types. Given two watch-points \(\vec{u}_1\) and \(\vec{u}_2\) and two curve-types defined from auxiliary vectors \(\vec{b}_1\) and \(\vec{b}_2\), let \(N_2(x)\) be the number of lattice points in \([1, x]^k \cap \mathbb{Z}^k\) that are simultaneously \(\vec{b}_1\)-visible from \(\vec{u}_1\) and \(\vec{b}_2\)-visible from \(\vec{u}_2\). Then the authors show that \[ N_2(x) = x^k \prod_p \left( 1 - \frac{1}{p^{s(\vec{b}_1)}} - \frac{1}{p^{s(\vec{b}_2)}} \right) + O(x^{k-1} \log^2 x + x^{k - \frac{k(s_* - 1)}{s_* + 1} + \epsilon}), \] where \(s_* := \min(s(\vec{b}_1), s(\vec{b}_2))\). The proof begins with a criterion for \(\vec{b}\)-visibility from \textit{C. Benedetti} et al. [Involve 14, No. 1, 103--118 (2021; Zbl 1468.11027)], but then incorporates classical analytic number theory and combinatorial sieving.