Integral zeros of quadratic polynomials avoiding sublattices (Q7026249)

Let \(Q(\textbf{x})=Q(x_1,\ldots,x_n)=\sum_{1 \le i,j\le n} f_{ij}x_ix_j+\sum_{i=1}^n l_ix_i+t\) be a quadratic polynomial in \(n\ge 2\) variables, where \(f_{ij},l_i,t\) are integers. For a given vector \(\textbf{z}=(z_1,\ldots,z_n)^\top\in\mathbb{Z}^n\), let \(|\textbf{z}|=\max_{1\le i \le n}|z_i|\) and \(\|\textbf{z}\|=(z_1^2+\cdots+z_n^2)^{1/2}\). The present work establishes the bounds on the least integer solution to \(Q(\textbf{z})=0\) in terms of \(|Q|=\max\{|f_{ij}|,|l_i|,|t|\}\), with the additional restriction that \(\textbf{z}\) avoids a union of finitely many sublattices of \(\mathbb{Z}^n\). \par\N\NSpecifically, it is shown (in Theorem 1.1) that given a rank-\(k\) sublattice \(\Lambda\) of \(\mathbb{Z}^n\), where \(2\le k \le n\), a finite collection of proper sublattices \(\Omega_1,\ldots,\Omega_m\subset \Lambda\), each of finite index in \(\Lambda\), and letting \(\Omega = \bigcap_{i=1}^m\Omega_i\), for a regular integral quadratic polynomial \(Q\) in \(n\) variables such that there exists \(\textbf{z}\in \Lambda \setminus \left( \bigcup_{i=1}^m\Omega_i \right)\) satisfying \(Q(\textbf{z})=0\), then there exists such a \(\textbf{z}\) satisfying\N\[\N|\textbf{z}| \ll\begin{cases} \det(\Omega)(2408(\det(\Omega))^2|Q|)^{860n^2(\det(\Omega))^2|Q|}& \text{if }k=2, \\\N\det(\Omega)^{2\rho(k)+1}|Q|^{\rho(k)} & \text{if }k\ge 3, \end{cases}\N\]\Nwhere \(\rho(k)\) equals \(2100\), \(84\), or \(5k+19+74/(k-4)\), for \(k=3, 4\), or \(\ge 5\), and the implied constants depend only on \(n\) and \(k\) and are effectively computable. The proof uses the Hermite inequality for the lattice bases and a result of \textit{R. Dietmann} [Proc. Lond. Math. Soc. (3) 86, No. 3, 545--582 (2003; Zbl 1036.11009)]. \par\N\NThe remaining results in the paper concern the distribution of angles among the vectors in \(\mathbb{Z}^n\). \textit{K. Yamamoto} proved in [Int. J. Number Theory 21, No. 1, 113--132 (2025; Zbl 1565.11081)] that for all \(n \ge 2\), \(n \ne 3\), the set of attainable angles\N\[\N\Theta_n = \{\angle(\textbf{z}, \textbf{w}) : \textbf{z}, \textbf{w} \in \mathbb{Z}^n\}\N\]\Ncoincides with the set\N\[\N\Theta_n(\textbf{z}) = \{\angle(\textbf{z}, \textbf{w}) : \textbf{w} \in \mathbb{Z}^n\}\N\]\Nfor any nonzero \(\textbf{z} \in \mathbb{Z}^n\), where \(\angle(\textbf{z}, \textbf{w}) = \arccos\dfrac{\textbf{z}^\top\textbf{w}}{\|\textbf{z}\|\|\textbf{w}\|}\). In Corollaries 3.3 and 3.4 it is shown that for a given \(\textbf{0}\not=\textbf{z}\in \mathbb{Z}^n\), \(n\ge 5\), and a fixed angle \(\theta \in \Theta_n\), where \(\tan^2 \theta =q/p \in \mathbb{Q}\), there exists \(\textbf{w}\in \mathbb{Z}^n\) such that \(\angle(\textbf{z},\textbf{w})=\theta\) and\N\[\N|\textbf{w}|\ll(2(p+q)\|\textbf{z}\|^2)^{(n-1)/2}.\N\]\NIf the vector \(\textbf{w}\) avoids a finite union of proper sublattices of finite index in \(\mathbb{Z}^n\), then combining this condition with Theorem 1.1 demonstrates that there exists such a vector \(\textbf{w}\) satisfying\N\[\N|\textbf{w}|\ll \det(\Omega)^{2\rho(n)+1}(2(p+q)\|\textbf{z}\|^2)^{\rho(n)}.\N\]