Support-sensitive bounds for shortest zero-sum subsequences

Technical Summary: Support-Sensitive Bounds for Shortest Zero-Sum Subsequences

Problem Statement
The paper addresses a localized question within zero-sum theory over finite abelian groups. While the classical Davenport constant determines the minimum length a sequence must have to guarantee the existence of any zero-sum subsequence, this work investigates how the structure of a sequence—specifically the number of distinct elements it contains (its support size)—constrains the length of its shortest nonempty zero-sum subsequence.

Let $G$ be a finite abelian group of order $n$, and let $S$ be a sequence over $G$. The authors define $MZ(S)$ as the minimum length of a nonempty zero-sum subsequence of $S$. The central problem is to bound $MZ(S)$ in terms of $n$ and $t = |\text{supp}(S)|$, the size of the support of $S$.

Methodology
The paper employs a combination of elementary combinatorial arguments and deep structural theorems specific to cyclic groups:

1. Elementary Support Bounds (General Abelian Groups):
   The authors utilize the elementary lower bound for the size of the set of subsequence sums of a zero-sumfree sequence, $|\Sigma(P)| \ge |P|$. By removing specific occurrences of support elements from a sequence of length $n$ and analyzing the intersection of the sumset of the remaining sequence with the set of negated removed elements, they derive a contradiction if the shortest zero-sum subsequence is too long relative to the support size.

2. Structural Refinement (Cyclic Groups):
   For cyclic groups $C_n$, the authors integrate the Savchev–Chen structure theorem. This theorem states that sufficiently long zero-sumfree sequences in $C_n$ can be transformed (via multiplication by a unit) such that the sum of the least positive representatives of their terms is strictly less than $n$. This allows for a sharper estimation of the sum of terms in a zero-sumfree subsequence, leading to improved bounds when the shortest zero-sum subsequence is "long" (specifically, when $MZ(S) - 1 > n/2$).

3. Arithmetic Translation:
   The results are translated into the language of algebraic number theory using the transfer principle from Krull monoids to block monoids over the class group. This involves mapping prime ideals to their ideal classes in the class group $\text{Cl}(K)$.

Key Contributions and Results

• General Support-Sensitive Bound:
  The primary result (Theorem 3.1) establishes that for any finite abelian group $G$ of order $n$ and any sequence $S$ of length $n$:
  $$MZ(S) \le n - |\text{supp}(S)| + 1$$
  This bound is shown to hold for any sequence of length at least $n$ (Corollary 3.2). The proof relies on the observation that a larger support leaves less "room" for all zero-sums to be long.

• Cyclic Refinement:
  For cyclic groups $C_n$, the authors prove a sharper estimate (Theorem 4.1) under the condition that the shortest zero-sum subsequence is relatively long ($MZ(S) - 1 > n/2$):
  $$MZ(S) \le n - \binom{|\text{supp}(S)|}{2}$$
  This result combines the general support bound with the Savchev–Chen structure theorem.

• Exact Extremal Case for Support Size 3:
  As a specific consequence of the cyclic refinement, the paper determines the exact value for sequences with support size 3 over $C_n$ (for $n \ge 5$):
  $$MZ(S) \le n - 3$$
  This bound is proven to be sharp, witnessed by the sequence $g^{[n-2]}(2g)(3g)$ where $g$ is a generator of $C_n$.

• Arithmetic Application:
  The authors apply these bounds to the factorization of products of prime ideals in a number field $K$ with class group of order $h$.

  • General Case: If $m \ge h$ prime ideals occupy $t$ distinct ideal classes, there exists a subproduct of at most $h - t + 1$ ideals that is principal and generated by an irreducible element (Proposition 5.1).
  • Cyclic Case: If the class group is cyclic, a refined bound $B_h(t)$ is provided (Proposition 5.2), which is the minimum of the linear bound and the cyclic refinement. Specifically, for $h \ge 5$ and $t=3$, a principal irreducible subproduct of length at most $h-3$ is guaranteed.

Significance and Claims
The paper positions itself as providing "support-sensitive" bounds that serve as a sequence-level companion to existing estimates on subsequence sums (such as those by Freeze and Smith). The authors claim that their results offer a more granular understanding of zero-sum phenomena by explicitly linking the length of the shortest zero-sum subsequence to the diversity of elements in the sequence.

The significance of the arithmetic application lies in its restriction to prime ideals. The authors note that for arbitrary integral ideals, a principal factor of a product does not necessarily arise from a subproduct of the original ideals; thus, the restriction to primes is essential for the direct ideal-factorization argument. The paper concludes by identifying future directions, specifically the determination of the function $f_{C_n}(t)$ for larger fixed $t$ and the investigation of stability questions regarding when equality is attained in the general bounds.