Notes on a problem of H. Cohn (Q1125401)

It is well-known that for the Legendre symbol \((a/p)= f(a)\), \(\sum f(a) f(a+b)\) over \(a\bmod p\) is \(-1\) except for \(b\equiv 0\). The reviewer posed the problem of whether such a correlation can characterize the symbol \((a/p)\), which was answered affirmatively by H. J. Lenstra and others [see Problem 39 in \textit{H. L. Montgomery}, Ten lectures on the interface between number theory and harmonic analysis, Regional Conference Series in Mathematics 84. Providence, RI: AMS (1994; Zbl 0814.11001)]. The author's main generalization uses \(a_i\) defined over \(\mathbb{F}_p\) (with \(a_0= 0\), \(a_1= 1\), \(a_i\neq 0\) for \(i\neq 0\)). Then the assertion \(\sum a_{i+j}/a_i= -1\) (for \(i\in \mathbb{F}_p\), \(j\neq 0\)), implies \(a_i= i^A\) for a constant \(A\). His main tool is the use of the bilinear correlation \(\sum f(a) g(a+b)\) to define subspaces created by ``characters''.