Divisibility properties of generalized Vandermonde determinants
As a generalization of the classical Vandermonde determinant \(V({\mathbf x})=\det(x_i^j)\), with \({\mathbf x} = (x_0,\dots,x_{n-1})\) an \(n\)-tuple of indeterminates and \(j=0,\dots,n-1\), one defines \(V_{{\mathbf a}}({\mathbf x})=\det(x_i^{a_j})\), where \({\mathbf a} = (a_0,\dots,a_{n-1})\) is any increasing \(n\)-tuple of nonnegative integers. Let \(x_0,\dots,x_{n-1}\) be pairwise distinct 2-integral odd rational numbers, all in the same residue class mod 4. When investigating the Kubota-Leopoldt 2-adic \(L\)-function, one of the authors of the present article asked whether \[ \text{ord}_2 V_{\mathbf a}({\mathbf x}) = \text{ord}_2 V({\mathbf x}) + \text{ ord}_2 V({\mathbf a}) - \text{ ord}_2 V(\mathbf s), \tag{*} \] where \({\mathbf s} = (0,\dots,n-1)\). The authors' main result is that this identity indeed holds if \(n\) satisfies a certain condition. They in fact prove this result more generally under the assumption that all the \(x_i\) lie in the same residue class mod \(2^{c+1}\), where \(c\geq 1\) is fixed. The condition on \(n\) requires that the blocks of identical digits of \(n-1\) in base 2 are not too large. Anyway, for a fixed \(c\) the equation (*) holds for infinitely many \(n\), and for a fixed \(n\geq 2\) it holds for any \(\mathbf a\) and \(\mathbf x\) if \(c\) is large enough. On the other hand, the authors find infinitely many counterexamples to (*) (for \(c=1\)). They also provide a method for verifying (*) by a computer when their main result fails to give an answer.
This page was built for publication: Divisibility properties of generalized Vandermonde determinants
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4429666)