A note on a conjecture of Gao and Zhuang for groups of order \(27\) (Q6972666)

The small Davenport constant \(d(G)\) of a finite group \(G\) is the maximal length of a sequence over \(G\) that contains no nontrivial product-one subsequence. A conjecture by \textit{J.~Zhuang} and \textit{W.~Gao} [Eur. J. Comb. 26, No. 7, 1053--1059 (2005; Zbl~1077.20044)] asserts that \(E(G) = d(G) + |G|\), where \(E(G)\) is the Erdős-Ginzburg-Ziv constant of \(G\), defined as the smallest integer \(N\) such that every sequence of length at least \(N\) over \(G\) contains a product-one subsequence of length \(\exp(G)\).\N\NIn this paper, the authors confirm the Zhuang-Gao conjecture for the non-abelian group \(G\) of order \(27\) and exponent \(3\). Specifically, they prove that \(d(G) = 6\) and \(E(G) = 33\). The group in question is the Heisenberg group of order \(3^3\), which is non-abelian, of class \(2\) and has center and commutator subgroup of order \(3\).\N\NThe proof is based on an intricate structural analysis of product-one-free sequences, their behavior under the central product condition and a detailed classification of \(z\)-classes.\N\NThe reviewer notes that the argument might be significantly simplified by employing the well-known Baer trick [\textit{I. M. Isaacs}, Finite group theory. Providence, RI: American Mathematical Society (2008; Zbl 1169.20001), Lemma 4.37]. Let \(p\) be an odd prime and let \((G, \cdot)\) be a finite \(p\)-group of nilpotence class at most \(2\). Then, the operation\N\[\Nx \circ y = x\cdot y\cdot [x,y]^{\frac{1}{2}}\N\]\N(where, for \(g \in G\), the element \(h = g^{1/2}\) denotes the unique element in \(G\) such that \(h^2 = g\)) defines a new group structure on \(G\) such that: (a) \((G, \circ)\) is an abelian group; and (b) for all \(g \in G\), the order of \(g\) is the same in \((G, \cdot)\) and in \((G, \circ)\).\N\NThanks to the Baer trick, the computation of \(d(G)\) reduces to the abelian group \((G, \circ)\). For any finite sequence \((g_1, \dots, g_n)\) in \(G\), one has \(g_1 \cdots g_n = 1\) in \((G, \cdot)\) if and only if \(g_1 \circ \cdots \circ g_n = 1\) in \((G, \circ)\). This follows by induction and the fact that \(x \circ y = 1\) if and only if \(xy = 1\). Hence, a sequence over \(G\) is product-one-free in \((G, \cdot)\) if and only if it is so in \((G, \circ)\) and therefore \(d(G) = d\big((G, \circ)\big)\). The problem is thus reduced to the abelian case, since by point (b) of Lemma 4.37 [loc. cit.], the order of any element in \((G, \cdot)\) coincides with its order in \((G, \circ)\). If \((G, \circ) \simeq \bigoplus_{i=1}^{r} C_{p^{n_i}}\), then it is well known that\N\[\Nd\big((G, \circ)\big) = \sum_{i=1}^{r} \big(p^{n_i} - 1\big)\N\]\N[\textit{J. E. Olson}, J. Number Theory 1, 8--10 (1969; Zbl 0169.02003)] and that \(E(G) = d(G) + |G|\) [\textit{W. D. Gao}, J. Number Theory 58, No. 1, 100--103 (1996; Zbl 0892.11005)].\N\NIn particular, if \(G\) is a \(p\)-group of class \(2\), order \(p^r\) and exponent \(p\), then \(d(G) = r(p - 1)\) and \(E(G) = d(G) + |G|\), which provides a positive answer to the question posed at the end of Section 1 of the paper.