A Multiplicative Property for Zero-Sums I

TL;DR

This paper proves a multiplicative property for the structure of extremal zero-sum sequences in groups of the form C_n ⊕ C_n, extending known results from prime to composite n and simplifying the full characterization of such sequences.

Contribution

It establishes a multiplicative property that reduces the characterization of extremal sequences for composite n to the prime case, extending previous results and conjectures.

Findings

Proves the multiplicative property for the structure of zero-sum sequences in C_{mn} ⊕ C_{mn}.

Unconditionally characterizes extremal sequences for many composite n cases.

Extends the structure conjecture from prime to composite groups, covering many new cases.

Abstract

Let $G=C_n\oplus C_n$ and let $k\in [0,n-1]$. We study the structure of sequences of terms from $G$ with maximal length $|S|=2n-2+k$ that fail to contain a nontrivial zero-sum subsequence of length at most $2n-1-k$. For $k\leq 1$, this is the inverse question for the Davenport Constant. For $k=n-1$, this is the inverse question for the $\eta(G)$ invariant concerning short zero-sum subsequences. The structure in both these cases (known respectively as Property B and Property C) was established in a two-step process: first verifying the multiplicative property that, if the structural description holds when $n=n_1$ and $n=n_2$, then it holds when $n=n_1n_2$, and then resolving the case $n$ prime separately. When $n$ is prime, the structural characterization for $k\in [2,\frac{2n+1}{3}]$ was recently established, showing $S$ must have the form $S=e_1^{[n-1]}\boldsymbol{\cdot}e_2^{[n -1]}\boldsymbol{\cdot} (e_1+e_2)^{[k]}$ for some basis $(e_1,e_2)$ for $G$. It was conjectured that this also holds for $k\in [2,n-2]$ (when $n$ is prime). In this paper, we extend this conjecture by dropping the restriction that $n$ be prime and establish the following multiplicative result. Suppose $k=k_mn+k_n$ with $k_m\in [0,m-1]$ and $k_n\in [0,n-1]$. If the conjectured structure holds for $k_m$ in $C_m\oplus C_m$ and for $k_n$ in $C_n\oplus C_n$, then it holds for $k$ in $C_{mn}\oplus C_{mn}$. This reduces the full characterization question for $n$ and $k$ to the prime case. Combined with known results, this unconditionally establishes the structure for extremal sequences in $G=C_{n}\oplus C_{n}$ in many cases, including when $n$ is only divisible by primes at most $7$, when $n\geq 2$ is a prime power and $k\leq \frac{2n+1}{3}$, or when $n$ is composite and $k=n-d-1$ or $n-2d+1$ for a proper, nontrivial divisor $d\mid n$.