A generalization of two \(q\)-identities of Andrews (Q5947370)

Let \(m\) and \(r\) be complex numbers whose sum is a nonnegative integer, and let \(n\) be a nonnegative integer at most equal to \(m + r\). Also, let \({{n} \brack k}\) be the Gaussian \(q\)-binomial coefficient. The paper under review establishes the identity \[ \sum _{k=0} ^n q ^{-k(k-1)/2}{{m - 2(n - k) - 2} \brack k} = \sum _{k=0} ^n (-1) ^{(n-k)} q ^{-(n-k)(n-k-1)/2}{{m + r} \brack {n - k}}A _{2k+2+r}, \] where \(A _{2k+2+r} := \sum _{j=0} ^k q ^{j ^2 + (m - 2n - 1)j} {{2k + 2 + r - j - 1} \brack j}\). This, applied to the special cases of: (i) \(m = 2n + 1\) and \(r = 0\); and (ii) \(m = 2n + 1\) and \(r = - 1\) (\(n \in {\mathbb N}\)), yields the two \(q\)-identities found by Andrews and used by him for obtaining a direct proof of D. Krammer's conjecture about the values \(K _{[n/2]} (e ^{2\pi /n})\) [see \textit{G. E. Andrews}, Ganita 43, 69-73 (1992; Zbl 0844.11020)]; these identities relate the sum \(K _n = 1 + 2.\sum _{k=1} ^n (-1) ^k q ^{-k(k-1)/2} {{2k - 1} \brack k}\) and the sums \(D _1,\dots , D _{2n+2}\), where \(D _{\nu +1} (q) = \sum _{k=0} ^{\infty } q ^{k ^2} {{n - k} \brack k}\), \(\forall \nu \in {\mathbb N}\). When \(q\) tends to \(1\), the main result of the paper reduces in the special cases of \(r = 0\) and \(r =-1\) to two identities relating the Fibbonacci numbers \(F _1,\dots, F _{2n+2}\); when \(m = 2n + 1\), the corresponding identities have been obtained by Andrews in the paper referred to.