Diophantine approximation on rational quadrics (Q2571046): Difference between revisions

The Jarnik-Besicovitch theorem asserts that the Hausdorff dimension of the (null) set \[ S_{\alpha}(\mathbb R)=\{x\in \mathbb R: |qx-p|<q^{-\alpha} \text{ for infinitely many } q\in \mathbb N, p\in\mathbb Z\} , \tag{*} \] where \(\alpha>1\), of very well approximable real numbers is given by \(\dim_{\text{H}}(\mathcal S_\alpha(\mathbb R))=2/(\alpha+1)\). It is one of the fundamental results in the theory of metrical Diophantine approximation and generalisations include \(n\)-dimensional Euclidean space, Kleinian groups acting on hyperbolic space and manifolds in Euclidean space. When \(n>1\), there are two natural types of approximation: the simultaneous, in which one considers the distance of a point from the rational point \(\mathbf p/q\), \(q\in \mathbb N\), \(\mathbf p\in \mathbb Z^n\), and the dual or linear form, in which one considers the distance from a rational hyperplane \(\{\mathbf x\in\mathbb R^n\colon \mathbf q\cdot\mathbf x=k\}\), \(\mathbf q\in\mathbb Z^n\), \(k\in \mathbb Z\). The theory for \(\mathbb R^n\) (independent variables) is on the whole now well understood but for manifolds (dependent variables) it is still incomplete. Although there has been dramatic progress for dual approximation, knowledge of the subtler simultaneous case is limited. Indeed the only complete general results hitherto known are for non-degenerate planar curves (\(1\)-dimensional manifolds) and these are very recent [\textit{V.~Beresnevich, H.~Dickinson} and \textit{S.~L.~Velani}, Diophantine approximation on planar curves and the distribution of rational points (with appendix by R.~C.~Vaughan): Sums of two squares near perfect squares. (To appear in Ann. Math.)]. The present paper establishes an extension of the Jarnik-Besicovitch theorem for the general rational quadric \(\mathcal Q\) and so represents a significant advance. Her results imply that the Hausdorff dimension of the set \[ \mathcal{S}_\alpha(\mathcal Q)= \{\mathbf x\in \mathcal{Q}: |q \mathbf x-\mathbf p|_\infty<q^{-\alpha} \text{ for infinitely many } q\in \mathbb N, \mathbf p\in\mathbb Z^n\}, \] where \(\alpha >1/n\), of simultaneously very well approximable vectors on the quadric is \[ \dim_{\text{H}}(\mathcal{S}_\alpha(\mathcal Q) )=\frac{n-1}{\alpha+1}, \] and that the Hausdorff measure at \((n-1)/(\alpha+1)\) is infinite. The proof replaces the set of \(\psi\)-simultaneously approximable vectors on a rational quadric by an equivalent set in a closely related locally symmetric space. A counting result for horoballs, which implies the equidistribution of of rational vectors on rational quadrics, leads via a standard covering argument to the upper bound for the Hausdorff dimension. To establish this as the lower bound, the author makes use of a deep `rigidity' type development of ubiquity in [\textit{V.~Beresnevich, H.~Dickinson} and \textit{S.~L.~Velani}, Measure theoretic laws for limsup sets, Mem. Am. Math. Soc. 179, No. 846, 1--91 (2006; Zbl 1129.11031)] that allows inferences to be drawn about Hausdorff measure and dimension. A dynamic interpretation arises by considering sets of moving geodesic rays. Quadrics with signature \((n,0)\) or \((1,n-1)\) essentially correspond to hyperbolic space and are covered by results concerning Kleinian groups acting on hyperbolic space in the paper just cited. The author points out that her methods use the intrinsic geometry and so cannot at present be used to consider the analogues of badly approxinable numbers and Khintchine's theorem, where a knowledge of the ambient space is required. A wide range of techniques and results are used in the proof but full explanations or references are included, as is a very useful survey of metrical Diophantine approximation results for Euclidean space and manifolds.