Prime and almost prime integral points on principal homogeneous spaces (Q617888)

Let integers \(n\geq 2\) and \(m\not=0\) be given, and let \(S\) be the set of \(n\times n\) integer matrices of determinant \(m\). Let \(f(x)\in\mathbb{Q}[x_{ij}]\) be any polynomial in the entries of \(x\in S\) which is integral for all such \(x\). Then the paper is concerned with showing that \(f(x)\) assumes infinitely many almost-prime values \(P_r\), for a suitable \(r\), as \(x\) runs over \(S\). For this it is natural to assume that \(f\) has no fixed prime divisor on \(S\) (i.e., that \(f\) is ``weakly primitive'', in the language of the paper). Moreover it is also natural to assume that \(f\) splits into a specified number \(t\) of irreducible factors \(f_1,\ldots,f_t\), over \(\mathbb{Q}[x_{ij}]/(\det(x_{ij})-m)\). One can obtain results of the type described above by fixing certain of the entries \(x_{ij}\), but that would produce a rather sparse set of almost-primes. The first result of the paper is that the proportion of matrices \(x\in S\) with entries of modulus at most \(T\), such that each of the irreducible factors \(f_j\) takes a prime value at \(x\), is of order at most \((\log T)^{-t}\), just as one might conjecture. In contrast, if \(f\) is weakly primitive, and has \(t\) distinct factors over \(\overline{\mathbb{Q}}[x_{ij}]/(\det(x_{ij})-m)\), then the proportion of matrices \(x\) above for which \(f(x)\) is a \(P_r\) almost-prime is at least of order \((\log T)^{-t}\). Here \(r\) can be any integer greater than \(18tn^3\eta_n\deg(f)\), where \(\eta_n=n-1\) for odd \(n\) and \(=n/(n-1)\) for even \(n\). In the case in which \[ f(x)=\prod_{i,j}x_{ij} \] it is shown that \(f\) is weakly primitive if and only if \(2^{n-1}\mid m\). However if this condition is met, and if \(n\geq 3\), then one can make all the entries of \(x\) prime, using machinery related to the proof of Vinogradov's 3-primes theorem. The line of attack for the above theorems may be extended to more general homogeneous spaces. A general sieve machinery is set up, in which the level of distribution information comes from counting matrices (lattice points) uniformly in appropriate orbits. This is the core of the paper, and uses methods taken from \textit{A. Gorodnik} and \textit{A. Nevo} [The ergodic theory of lattice subgroups. Annals of Mathematics Studies 172. Princeton, NJ: Princeton University Press (2009; Zbl 1186.37004)].