Almost periodicity and the remainder in the ellipsoid problem (Q5954602)

Let \(Q(\vec x)\) denote an \(m\)-dimensional positive definite quadratic form and \(\vec a\in \mathbb{R}^m\). Let \(\Delta_{\lambda,a}(X)\) be the error term in the asymptotic representation of \[ \sum_{\vec x\in \mathbb{Z}^m: Q(\vec x-\vec a)\leq X} \Biggl( 1- \frac{Q(\vec x-\vec a)}{X} \Biggr)^\lambda. \] If \(\lambda=0\), then \(\Delta_{0,a}(X)\) is the error term in the ellipsoid problem, which describes the number of lattice points in the ellipsoid \(\vec a+ \sqrt{X}E\), where \(E= \{\vec x\in \mathbb{R}^m\mid Q(\vec x)\leq 1\}\). One is also interested in results for \(\lambda>0\), because then the error term is somewhat smoothed out and almost periodicity can be applied. The two main results are proved: Let \(m\geq 3\), \((m-3)/2< \lambda< (m-1)/2\). (1) If \(E\) is rational and \(\vec a\) arbitrary, then \(t^{\lambda-(m-1)/2} \Delta_{\lambda,a} (t^2)\) is \(B^1\)-almost periodic. (2) Moreover, if also \(\vec a\) is rational, then this function is \(B^2\)-almost periodic (in the sense of Besicovitch). The second result holds also for \(\vec a\) arbitrary, provided that \(m\geq 8\).