Lattice points in algebraic cross-polytopes and simplices (Q724947)

A classical problem is to find, for a given set \(P\subset\mathbb{R}^d\), good estimates for the cardinality \(|tP\cap\mathbb{Z}^d|\), where \(t\) is a real \(>1\) and \(tP=\{ tx:\, x\in P\}\). \textit{E. Ehrhart} [C. R. Acad. Sci., Paris 254, 616--618 (1962; Zbl 0100.27601); J. Reine Angew. Math. 226, 1--29 (1967; Zbl 0155.37503)] showed that if \(P\) is a polytope with all vertices in \(\mathbb{Z}^d\) then there is a polynomial \(p(t)\in\mathbb{Q}[t]\) of degree \(d\) such that \(|tP\cap\mathbb{Z}^d|=p(t)\) for every positive integer \(t\). Of course the leading term of \(p(t)\) is \(\lambda (P)t^d\), where \(\lambda (P)\) is the \(d\)-dimensional volume of \(P\). Ehrhart's equality is in general false if \(t\) is not integral. In the paper under review, the author considers certain polytopes whose vertices have algebraic coordinates and allows \(t\) to be real. More precisely, let \(a_1,\ldots ,a_d\) (\(d\geq 2\)) be positive real algebraic numbers such that \(a_1^{-1},\ldots ,a_d^{-1}\) are linearly independent over \(\mathbb{Q}\). Consider the cross-polytope and simplex \[ \begin{aligned} C&=\{ x\in\mathbb{R}^d:\, a_1^{-1}|x_1|+\cdots +a_d^{-1}|x_d|\leq 1\},\\ S&=\{ x\in\mathbb{R}^d:\, x_1,\ldots ,x_d\geq 0,\, a_1^{-1}x_1+\cdots +a_d^{-1}x_d\leq 1\}.\end{aligned} \] The author shows that \[ |tC\cap\mathbb{Z}^d|=p(t)+O(t^{r(d)+\varepsilon}),\;\;|tS\cap\mathbb{Z}^d|=q(t)+O(t^{r(d)+\varepsilon})\;\;\text{ as } t\to\infty \tag{\(*\)} \] for any \(\varepsilon >0\), where both \(p(t)\) and \(q(t)\) are polynomials of degree \(d\) with rational coefficients, explicitly given in the paper, and where \(r(d)=(d-1)(d-2)/(2d-3)\). As the author points out, the two assertions in (\(*\)) are equivalent and it suffices to prove the first. The author makes essential use of \textit{W. M. Schmidt}'s results on the simultaneous approximation of algebraic numbers by rationals [Acta Math. 125, 189--201 (1970; Zbl 0205.06702)]; as a consequence the implied constants in the error terms in (\(*\)) are not effectively computable from the method of proof. In fact, the author derives a Poisson summation formula type expression for \(|tC\cap\mathbb{Z}^d|\), by combining Fourier analysis with Schmidt's results, and from this expression he derives (\(*\)) where again Schmidt's results are needed. The author obtains the following by-product, which is of interest in itself. Here, \(\|\cdot\|\) denotes the distance to the nearest integer. For \(d\geq 1\), let \(\gamma_d\) be the infimum of all reals \(\gamma\) having the following property: if \(\alpha_1,\ldots,\alpha_d\) are any algebraic reals such that \(1,\alpha_1,\ldots ,\alpha_d\) are linearly independent over \(\mathbb{Q}\), then \[ \sum_{m=1}^M\frac{1}{\| m\alpha_1\|\cdots\| m\alpha_d\|}=O(M^{\gamma}) \;\;\text{ as } M\to\infty . \] Using Schmidt's results, the author shows that \(\gamma_d\leq 2-1/d\). In fact, from the Poisson type expression for \(|tP\cap\mathbb{Z}^d|\) mentioned above, the author deduces formulas like in (\(*\)) with \((\gamma_{d-1}-1)(d-1)/\gamma_{d-1}\) instead of \(r(d)\), and then deduces (\(*\)) by inserting his upper bound for \(\gamma_{d-1}\).