On the representation of integers by the Lorentzian quadratic form (Q1376564)

Let \(s(n,k,r)\) and \(r(n,k,t)\) denote the number of integral solutions of the equation \[ x^2_1+ x^2_2+ \cdots+ x^2_n-x^2_{n+1} =k \] satisfying the side conditions \(x^2_1+ \cdots+ x^2_{n+1} \leq r^2\) and \(|x_{n+1} |\leq t\), respectively. Here, \(n\geq 2\) and \(k\) is any nonzero integer which is not a square in case \(n=2\). The aim of this work is to show that \[ r (n,k,t) \sim c(n,k) t^{n-1} \text{ as }t \to\infty \quad \text{and} \quad s(n,k,r) \sim d(n,k)r^{n-1} \text{ as } r\to \infty \tag{1} \] where \(c(n,k)\), \(d(n,k)\) are explicitly known constants involving the density \(\delta (n,k)\), which is a product \(\prod_p \delta_p(n,k)\) of local densities \(\delta_p (n,k)\). All \(\delta_p(n,k)\) with \(p\geq 3\) were already known. The authors determine \(\delta_2 (n,k)\) and give an explicit formula for \(\delta(n,k)\) which reveals some surprising facts. Formula (1) gives an asymptotic formula for the mean value of the number of representations of an integer as the sum of \(n\) squares along the second order arithmetic progression \(m^2+k\) \((|m|\leq t\), \(t\to\infty)\). The proof of (1) for negative \(k\) uses the hyperbolic lattice point theorem. Known lower estimates for the least positive eigenvalue of the Laplacian yield an error term in case \(k<0\). For positive \(k\), the proof is based on known results on counting integer points and on Siegel's main theorem.