An explicit estimate for the lattice point discrepancy of ellipsoids of rotation (Q2466365)

One of the key issues when it comes to count the integer points that lie in big dilated domain is to give a good estimate of the lattice discrepancy, \textit{i.e.} the difference between the number of points and the volume of the domain. For a domain \(\mathcal{K}\) of the real 3-dimensional space, let \(P_\mathcal{K}(x)\) denote \[ P_\mathcal{K}(x) = \# (\mathbb{Z}^3 \cap \sqrt{x} \mathcal{K}) - \text{vol} (\mathcal{K}) x^{3/2}. \] The authors consider the case when \(\mathcal{K}\) is an ellipsoid \(\mathcal{E}\) of revolution, say of equation \((x_1^2+ x_2^2)/a + a^2 x_3^2 \leq 1 \). They show that, under some technical assumption, \[ |P_\mathcal{K}(x) | \leq 1237 a^{1/8} x^{11/16} \left( \log (100x) + \log |a| \right) + \text{ explicit terms of lower order}. \] The main term improves all the general purpose estimates and enhances \textit{F. Chamizo}'s approach for bodies of revolution [Acta Arith. 85, No. 3, 265--277 (1998; Zbl 0919.11061)]. Note also that for \(a=1\) (\(\mathcal{E}\) is then a ball) the best known upper bound is \(O(x^{21/32 + \varepsilon})\) according to \textit{D. R. Heath-Brown} [in Number theory in progress. Volume 2: Elementary and analytic number theory. Berlin: de Gruyter, 883--892 (1999; Zbl 0929.11040)].