On the local-global conjecture for integral Apollonian gaskets. With an appendix by Péter P. Varjú (Q2247931)

Let \(\mathcal G\) be an Apollonian gasket and \(b(C)\) be the curvature of \(C\in {\mathcal G}\), that is, the reciprocal of its radius (the bounding circle has negative orientation). In 1937, \textit{F. Soddy} [Nature 139, 77--79 (1937; Zbl 0015.36704)] proved that there exists integral gaskets \(\mathcal G\), that is, gaskets \(\mathcal G\) for which \(b(C)\in \mathbb{Z}\) for all \(C\in {\mathcal G}\). Let \({\mathcal B} ={\mathcal B_G} := \{ b(C) : C\in {\mathcal G}\}\). A gasket \(\mathcal G\) is called primitive if \(\gcd({\mathcal B})=1\). An integer \(n\) is called admissible if \(n\in {\mathcal B}\pmod q\) for all \(q\geq 1\). In 2003, \textit{R. L. Graham} et al. [J. Number Theory 100, No. 1, 1--45 (2003; Zbl 1026.11058)] posed the following conjecture: \textit{Local-Global Conjecture.} Fix a primitive, integral Apollonian gasket \(\mathcal G\). Then every sufficiently large admissible number is the curvature of a circle in \(\mathcal G\). In this paper, the authors proves the following nice result. Theorem. Fix a primitive, integral Apollonian gasket \(\mathcal G\). Almost every admissible number (in the sense of density) is the curvature of a circle in \(\mathcal G\). Quantitatively, the number of exceptions up to \(N\) is bounded by \(O(N^{1-\eta})\), where \(\eta\) is an effectively computable positive constant. On pages 639--648 an Appendix: Spectral gap for the Apollonian group by Péter P. Varjú is given.