The local-global principle for integral Soddy sphere packings (Q2289405)

Let \(\mathcal P\) be the set of spheres in an integral Soddy sphere packing, and let \(\mathcal B\subset\mathbb Z\) be the set of bends (sometimes called curvature) for the spheres in \(\mathcal P\). Let the metric be suitably scaled so that the elements of \(\mathcal B\) have no common factor. Let \[ \mathcal A=\{n\in\mathbb Z: \hbox{for any }q>1,\text{ there exists }m\in \mathcal B\text{ such that }n\equiv m \pmod q\}. \] It is clear that \(\mathcal B\subset \mathcal A\). In this paper, the author shows that, for any fixed Soddy sphere packing, there exists an effectively computable upper bound for \(\max\{\mathcal A\setminus \mathcal B\}\). That is, if \(n\in \mathcal A\) is sufficiently large, then \(n\in \mathcal B\). More generally, given a set \(\mathcal B\subset\mathbb Z^k\), one can similarly define the larger set \(\mathcal A\). We say \(\mathcal B\) satisfies a local to global principal if \(\mathcal A\setminus \mathcal B\) is empty, finite, or (as in this case) bounded above. For the case where \(\mathcal B\) is the set of integer solutions to a polynomial equation and \(\mathcal A\setminus \mathcal B=\emptyset\), this is known as the Hasse principal. For the case where \(\mathcal B\) is the set of numbers represented by a quadratic form, the set \(\mathcal A\) is the set of \textit{admissible} numbers and the problem is related to Hilbert's 11th problem. The situation at hand has some of the flavors of both problems. There is an underlying Diophantine equation, which is a quadratic form \(Q\) with signature \((4,1)\) that is roughly derived from an analog of Descartes' theorem. Its integer solutions can be described as an orbit under the action of a discrete orthogonal group \(O_Q\) acting on \(\mathbb H^4\). One is interested in those solutions that lie in an orbit under the action of a subgroup \(\Gamma \leq O_Q\). As with the Diophantine problem, finding all solutions in the \(\Gamma\)-orbit modulo \(q\) is a finite problem for any fixed \(q\). The group \(\Gamma\) is thin, which means its limit set in \(\mathbb S^3\cong \partial\mathbb H^4\) has measure zero. In contrast, the limit set of \(O_Q\) is all of \(\mathbb S^3\). One is therefore interested in only a very thin subset of solutions to the Diophantine equation. On the other hand, these solutions are then projected onto \(\mathbb Z\) to give \(\mathcal B\). The nature of this projection depends on the choice of integer Soddy packing. The author notes that this is the first local to global phenomenon demonstrated for a thin group. In particular, the analogous problem for bends of circles in an integer Apollonian circle packing remains open (see \textit{J. Bourgain} and \textit{A. Kontorovich} [Invent. Math. 196, No. 3, 589--650 (2014; Zbl 1301.11046)]). The set of admissible numbers depends on the packing/projection, and is either all positive integers equivalent to \(0\) or \(1\) modulo \(3\), or is all positive integers equivalent to \(0\) or \(2\) modulo \(3\). In an example provided by the author, many (early) admissible numbers are not represented.