Markoff triples and strong approximation (Q512334)

Consider the affine cubic curve \[ X\,:\, \Phi(x_1,x_2,x_3)=x_1^2+x_2^2+x_3^2 - 3x_1x_2x_3=0. \] A triple \((x_1,x_2,x_3)\) of \textit{natural} numbers is called a \textit{Markof triple} if \((x_1,x_2,x_3)\in\mathbb{A}^3\) is a point of \(X\). The set of Markoff triples is denoted by \(\mathcal{M}\). It is known that all non-zero points of \(X(\mathbb Z)\) are of the form \((\varepsilon_1x_1,\varepsilon_2x_2,\varepsilon_3x_3)\), where \(\varepsilon_j=\pm 1\) (\(j=1,2,3\)), \(\varepsilon_1\varepsilon_2\varepsilon_3=1\) and \((x_1,x_2,x_3)\in\mathcal{M}\), and all triples of \(\mathcal{M}\) are obtained from the triple \(\mathfrak{a}=(1,1,1)\) by repeated applications of the coordinates and the so called \textit{Vieta} involutions \(R_1,R_2,R_3\). Here, \(R_1(x_1,x_2,x_3)=(3x_2x_3-x_1,x_2,x_3)\) and \(R_2,R_3\) are analogously defined. Thus \(\mathcal{M}=\Gamma\cdot\mathfrak{a}\), where \(\Gamma\) is the non-linear group of affine morphisms of \(\mathbb{A}^3\) generated by permutations and the \(R_j\)'s. A natural number is a \textit{Markoff number} if it appears as a coordinate of some Markoff triple. The so-called \textit{fundamental strong approximation conjecture} is the following: For \(p\) a prime, the group \(\Gamma\) acts on \(X(p):=X(\mathbb Z/p\mathbb Z)\) with two orbits: \(\{0\}\) and \(X^*(p)=X(p)\setminus\{ 0 \}\). This conjecture implies, among other things, that the reduction \(\bmod\,p\) from \(\mathcal{M}\) to \(X^*(p)\) is onto. The conjecture is supported by following two theorems, due to the authors: Theorem 1. For \(\varepsilon>0\) and \(p\) large, there is an orbit \(\mathcal{C}(p)\) of \(X^*(p)\), for which \[ |X^*(p)\setminus\mathcal{C}(p)|\leq p^{\varepsilon} \] and all \(\Gamma\)-orbits \(\mathcal{D}(p)\) in \(X^*(p)\) satisfy \(|\mathcal{D}(p)|\gg (\log p)^{1/3}\). Note that, since \(|X^*(p)|\sim p^2\), Theorem 1 guarantees the existence of a ``giant'' orbit \(\mathcal{C}(p)\). Theorem 2. For \(\varepsilon>0\), the number of primes \(p\leq x\), for which the conjecture fails is \(O_{\varepsilon}(x^{\varepsilon})\). In the present paper, a brief overview of the proofs of the above two theorems is given. Another result is also stated, concerning the \textit{Markoff numbers}. By such a number we understand any natural number which appears as a coordinate of a Markoff triple. The result roughly says that \textit{almost all Markoff numbers are composite}, where the meaning of ``almost'' is somewhat technical and is explained in the paper.