The lattice point discrepancy of a torus in \(\mathbb R^3\) (Q949878)

Consider a compact body \(B\) in the three-dimensional Euclidean space. We are interested in the number of lattice points in a linearly dilated body \(tB\), \(t>1\). The lattice point discrepancy is defined by \[ P_B(t)=\#(tB\cap\mathbb Z^3)-\text{vol}(B)t^3. \] If the boundary of \(B\) is sufficiently smooth and of bounded nonzero Gaussian curvature, then a classical result of E. Hlawka is the estimate \[ P_B(t)\ll t^{3/2}. \] If \(B=R\) is a body rotation with respect to one coordinate axis, F. Chamizo even proved \[ P_B(t)\ll t^{11/8+\varepsilon}. \] A new situation arises, if the boundary contains points with Gaussian curvature zero. In this case our knowledge about the size of the lattice discrepancy is very fragmentary. Here the author makes an interesting contribution by considering the number of lattice points in the torus \(B=T=T_{a,b}\), defined by \[ \begin{pmatrix} x \\ y \\ z \end{pmatrix} =\begin{pmatrix} (a+b\cos \alpha)\cos \beta\\ (a+ b\cos \alpha )\sin \beta\\ b\sin \alpha\end{pmatrix} , \] \(0\leqq \alpha\), \(\beta < 2\pi\), \(a>b>0\). The torus \(T_{a,b}\) contains two circles \(\alpha=\pi/2\), \(3\pi/2\) on his boundary, on which the Gaussian curvature is zero throughout. It is interesting that these two circles yield the main contribution to the lattice point discrepancy of \(tT=tT_{a,b}\). The result is \[ P_T(t)=F_{a,b}(t)t^{3/2}+O(t^{11/8+\varepsilon}). \] \(F_{a,b}(t)\) is a certain periodic function given explicitly by an absolutely convergent Fourier series.