Roots of unity and unreasonable differentiation (Q2274006)

The authors prove a cute result in elementary number theory which seems to fall out from an investigation about certain questions concerning input-automata on finite groups. They have defined certain groups called \(J\)-groups related to Jacobson radicals of near rings. In order to prove that many metacyclic groups are \(J\)-groups, they needed to establish the \(k-1\) case of the following theorem. For \(k\geq 0\), denote by \(n^{\underline{k}}\), the usual falling factorial \(n(n-1) \cdots (n-k+1)\); here, the notation means \(n^{\underline{0}}=1\). For a polynomial (or even for a formal power series) \(f = \sum_n f_nt^n \in \mathbb{Z}[t]\), one has \(f^{(k)}(t) = \sum_{n \geq k}n^{\underline{k}}f_n t^{n-k}.\) If \(f(t) = \sum_{i=0}^{n-1} t^i\), then evidently \(f^{(0)}(a) = f(a) \equiv 0\bmod n\) whenever \(a\) is an integer satisfying \(a^n \equiv 1\bmod n\). The authors prove the surprising result that such a congruence holds for \(f^{(k)}(a)\) for various other \(k\)'s also. The precise result asserts: Theorem: Let \(k \geq 0, n \geq 1, f = \sum_{i=0}^{n-1}t^i\). Then, \(f^{(k)}(a) = \sum_{i-0}^{n-1} i^{\underline{k}} a^{i-k} \equiv 0\bmod n\) if, and only if, at least one of the following holds: (i) \(k \neq 3, q-1\) for a prime \(q\); or (ii) \(k=3\), and \(4\) does not divide \(n\); or (iii) \(k=q-1\), where \(q\) is a prime so that \((n,q)=1\) and \(a \not\equiv 1\bmod q\).