Lattices with few distances (Q1176400): Difference between revisions

The authors consider the Erdős number \(E\) of a lattice \(\Lambda\) in \(n\)-dimensional Euclidean space, \(n\geq 3\), which is defined as follows. For \(x>0\) let \(P(x)\) be the number of values taken in \([0,x]\) by the quadratic form corresponding to \(\Lambda\), and set \(E=F d^{1/n}\) where \(d=\hbox{det} \Lambda\) and \(F\) = limit of \(P(x)/x\) as \(x\to\infty\). Let \(n\) be fixed. The main result of the paper says that \(E\) becomes minimal (and so \(\Lambda\) has ``fewest distances'') if and only if \(\Lambda\) is, up to a scalar factor, an even lattice with minimal \(d(=1,2,3\) or 4 depending on \(n\mod 8)\). The key to the proof lies in the observation that, when \(\Lambda\) is integral, \(E\) depends only on which numbers are represented by the genus of \(\Lambda\). For \(n=3\) this is based on a difficult result about the exceptions (to the local-global principle) for which the authors refer to \textit{W. Duke} and \textit{R. Schulze-Pillot} [Invent. Math. 99, No.1, 49-57 (1990; Zbl 0692.10020)] and, in particular, to an argument quoted from a letter of Schulze-Pillot. The paper then proceeds with the local computations. There are also tables of lattices with relatively small \(E\) for \(n=3\) and 4.