Rational approximation on quadrics: a simplex lemma and its consequences (Q2327755)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Rational approximation on quadrics: a simplex lemma and its consequences
scientific article

    Statements

    Rational approximation on quadrics: a simplex lemma and its consequences (English)
    0 references
    0 references
    0 references
    15 October 2019
    0 references
    The authors prove several results on the approximation of points on a rational quadric \(X\) in \(\mathbb{P}^n (\mathbb{R})\), i.e., the zero locus of a rational quadratic form \(Q\) in \(n+1\) variables, by rational points also on \(X\), i.e. by points from the set \(X(\mathbb{Q})\). In this setting, define the Diophantine exponent of a point \(x\) by \[ \beta(x) = \inf \{\beta0 : \exists c 0 : \forall v \in X(\mathbb{Q}): \operatorname{dist}(x,v) \ge c H(v)^{-\beta}\}, \] where \(\operatorname{dist}(x,v)\) denotes the projective distance between \(x\) and \(v\), i.e., the absolute value of the sine of the angle between them as vectors, and \(H(v)\) denotes the projective, naive height of \(v\), i.e., the maximum modulus of the coordinates, when these are chosen to be coprime integers. A totally isotropic subspace of \(\mathbb{R}^{n+1}\) is a linear subspace on which \(Q\) vanishes. The maximal dimension of such a space is called the \(\mathbb{Q}\)-rank of \(X\), \(\mathrm{rk}_{\mathbb Q} X\). If \(\mathrm{rk}_{\mathbb Q} X 0\) and \(X\) is non-singular, \(X(\mathbb Q) \neq \emptyset\), and it is well-known that \(\beta(x) \ge 1\) for all \(x \in X\). Conversely, \textit{L. Fishman} et al. [Trans. Am. Math. Soc. 370, No. 1, 577--599 (2018; Zbl 1422.11149)]] have shown that for Lebesgue almost every \(x \in X\), \(\beta(x) \le 1\). In the present paper, it is first shown that if \(Y \subseteq X\) is a smooth submanifold with \(\dim Y \ge\mathrm{rk}_{\mathbb Q} X\), the latter property also holds for Lebesgue almost every \(x \in Y\). Subsequently, the set \(W_{\beta}\) of points \(x \in X\) for which \(\beta(x) \ge \beta\). Extending a result of \textit{L. Fishman} et al. [Sel. Math., New Ser. 24, No. 5, 3875--3888 (2018; Zbl 1464.11068)] an upper bound on the Hausdorff dimension of \(Y \cap W_{\beta}\) is obtained, whenever \(Y \subseteq X\) is a suitably nice subset, such as a smooth submanifold or a sufficiently regular fractal subset. Finally, the badly approximable points are studied, i.e., the set of points \[ \mathrm{BA}_X = \{X \in X: \exists c 0 : \forall v \in X(\mathbb Q): \mathrm{dist}(x,v) \ge c H(v)^{-1}\}. \] Extending results of \textit{L. Fishman} et al., [``Intrinsic Diophantine approximation on quadric hypersurfaces'', Preprint, \url{arXiv:1405.7650}], it is shown that this set is winning for a certain variant of the Schmidt game. As a consequence, it intersects any \(C^1\) submanifold \(Y \subseteq X\) of dimension at least \(\mathrm{rk}_{\mathbb Q} X\) in a set of maximal Hausdorff dimension. All three main results depend on a version of the so-called Simplex Lemma, which is a higher dimensional generalisation of the trivial fact that two different rational numbers, \(p/q\) and \(p'/q'\) with \(0 q, q' \le N\) have Euclidean distance at least \(1/N^2\). This is extended nontrivially to the present setting and used several times.
    0 references
    intrinsic Diophantine approximation
    0 references
    quadratic forms
    0 references
    Schmidt games
    0 references
    Euclidean lattices
    0 references
    0 references
    0 references

    Identifiers

    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references