On the rank of Hankel matrices over finite fields (Q2669189): Difference between revisions

Given a field \(F,\) any \((n+1)\)-tuple \(x= (x_{0}, x_{1}, \cdots, x_{n})\in F^{n+1}\) and nonnegative integer numbers \(p, q\) satisfying \(p+q \leq n,\) generate the structured matrix \(H_{p,q}(x)= (x_{i+j})_{0\leq i \leq p, 0\leq j \leq q } \in F^{p+1, q+1}\) that is known in the literature as a Hankel matrix. These matrices appear in many fields of mathematics, as the theory of orthogonal polynomials (as Gram matrices in the monomial basis for inner products associated with probability measures supported on the real line), in the theory of total positivity and in control theory, among others. When \(F\) is a finite field with \(|F|=s\), an interesting problem is to find the number of Hankel matrices of a given rank. Indeed, if \(r, m, n\) are nonnegative integer numbers satisfying \(r\leq m\), \(r\leq n\), then the number of \((m+n+1)\)-tuples \(x\in F^{m+n+1}\) such that rank \(H_{m,n} (x)\leq r\) is \(s^{2r}.\) This result has been stated in an equivalent way and proved in [\textit{D. E. Daykin}, J. Reine Angew. Math. 203, 47--54 (1960; Zbl 0104.01304)]. In the present paper, the authors obtain the following generalization of the above result. Let \(F\) be a finite field and \(|F|=s\) and consider nonnegative integer numbers \(r, m, n\) such that \(r\leq m, r\leq n\) and fix a \(k\)-tuple \(a= (a_{0}, a_{1}, \cdots, a_{k-1}).\) Then the number of \(m+n+1\)-tuples \(x\in F^{m+n+1}\) such that \((x_{0}, x_{1}, \cdots, x_{k-1})=a\) and rank\((H_{m,n} (x) )\leq r\) is \(s^{2r-k}.\) This means that if one fixes the first \(k\) entries of \(H_{p,q}(x),\) then \(s^{2r-k}\) is the number of ways to choose the remaining \(p+q- k+1\) entries for some \(k\leq r \leq \min\{p,q\},\) in order to rank\((H_{p,q}(x))\leq r.\) This reflects the fact that the first \(k\) entries have no effect on the rank of the matrix. As a straightforward consequence, if \(r=m=n, \) then \(s^{2n-k}\) is the number of \((2n+1)\)-tuples \(x\in F^{2n+1}\) such that \((x_{0}, x_{1}, \cdots, x_{k-1})=a\) and \(H_{n,n}(x)\) is a singular matrix. Notice that in the framework of random matrices, in [\textit{B. Anzis} et al., Ann. Comb. 22, No. 3, 447--489 (2018; Zbl 1402.05214)], the authors deduce an expression of the probability of a certain Hankel matrix of size \((n+1)\times (n+1)\) over a finite field to be singular.