🔢 mathematics

Slice and Partition Rank Criteria for Polynomial Zero-Avoidance

This paper establishes new bounds for polynomial zero-avoidance and higher-degree Erdős–Ginzburg–Ziv constants over finite vector spaces by effectively applying the support-entropy method and partition rank techniques to derive explicit entropy gaps and exponential bounds, including a novel result for the fourth elementary symmetric polynomial over $\mathbb{F}_5^n$ .

Original authors: Simone Costa, Stefano Della Fiore, Mattia Fontana

Published 2026-08-03

📖 7 min read🧠 Deep dive

CC BY 4.0

Original authors: Simone Costa, Stefano Della Fiore, Mattia Fontana

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Great Zero-Hunt in a Digital Maze

Imagine you are playing a high-stakes game of hide-and-seek inside a giant, multi-dimensional maze. But this isn't a normal maze; it's built out of numbers from a finite world, like a video game grid where the numbers wrap around after a certain point (like how a clock resets after 12). In this world, mathematicians study "zero-sum" games. The goal is simple: find a group of numbers that, when you mix them together in a specific recipe, the result is exactly zero.

The classic version of this game is the Erdős–Ginzburg–Ziv problem. It asks: "How many numbers do I need to pick from a bag to guarantee that I can find a small group of them that adds up to zero?" It's like asking, "How many people do I need in a room to guarantee that three of them have birthdays that add up to a multiple of 365?"

But this paper dives into a much trickier, higher-level version of the game. Instead of just adding numbers, the "recipe" is a complex polynomial equation (a fancy algebraic formula involving multiplication and addition). The question becomes: "How many numbers do we need to guarantee that a specific group of them makes this complex formula equal to zero?"

To solve this, the authors use two powerful mathematical "flashlights" called slice rank and partition rank. Think of these as special ways of looking at a giant 3D (or even 10D) block of data.

Slice Rank is like slicing a loaf of bread. If you can describe a complex shape by stacking up simple, flat slices, you can measure the size of the shape by counting the slices. If the shape is "thin" enough (has a low slice rank), it means the shape is small, and you can prove that a large group of numbers must contain a zero-sum group.
Partition Rank is like sorting a messy pile of toys into boxes based on which toys are identical. It helps mathematicians handle the rule that all the numbers in our group must be different from each other. This is crucial because in the real world, you cannot just pick the same number twice to repeat the game.

The authors are trying to find the "magic number"—the minimum size of a group needed to force a zero-sum solution. If they can prove this number is smaller than the total size of the alphabet (the set of available numbers), they have found a non-trivial, efficient way to solve the puzzle.

The Paper's Discovery: Sharper Flashlights and New Tricks

In this paper, Simone Costa, Stefano Della Fiore, and Mattia Fontana take these mathematical flashlights and polish them until they shine much brighter than before. They tackle two main challenges: making the "slice rank" flashlight more precise and using the "partition rank" flashlight to handle the tricky rule of "all distinct numbers."

1. Sharpening the Slice Rank Flashlight (The "Entropy" Gap)
First, the authors look at a specific polynomial called the "quadratic elementary symmetric polynomial" (basically, $xy + yz + zx$) over fields of characteristic three (a world where numbers wrap around after 3).
Previously, mathematicians knew that the "slice rank" method worked, but they couldn't always calculate exactly how much smaller the solution set was compared to the total alphabet. It was like knowing a box is smaller than a room, but not knowing by how much.
The authors developed a new "dual certificate." Imagine trying to prove a room is too small for a party. Instead of just counting people, they found a specific mathematical "witness" (a certificate) that proves, with a clear margin, that the room is too cramped.

The Result: They proved that for this specific polynomial, the maximum size of a group that avoids a zero-sum solution is strictly smaller than the total number of available digits.
The Numbers: For a field with $3^k$ elements, they found a new, tighter bound. For example, when $k=2$ (a field of 9 elements), the exponential base of the bound is approximately 8.311, which is strictly less than 9. This is an improvement over previous estimates that were slightly looser. They provided a single, clean formula that works for all sizes of these fields, avoiding the need to solve a new, messy puzzle for every single field size.

2. The "Distinctness" Puzzle (Partition Rank)
The second, more difficult part of the paper deals with the rule that all numbers in the group must be different.
If you just use the standard "slice rank" method, it does not care if you pick the same number twice. It is like a game where you can repeat the same card over and over. The authors needed a way to force the players to pick unique cards.
They used a clever trick involving "contractions." Imagine you have a complex equation with variables $x_1, x_2, x_3$ . If you force $x_1$ to equal $x_2$ , the equation simplifies (contracts). The authors realized that the problem of finding "all distinct" solutions could be broken down into a sum of these simpler, "contracted" problems.

The Strategy: They used a mathematical tool called the "partition lattice" (a way of organizing how variables can be equal or different) to split the big problem into many smaller, manageable pieces.
The Breakthrough: They applied this to a restricted alphabet: the "multiplicative torus." This is a fancy way of saying they only looked at numbers that are not zero. By doing this, they could use a sharper version of the slice rank method.
The Result: They successfully transferred these results back to the full space (including zeros) using a technique called "support stratification" (grouping numbers by how many zeros they have).
The Big Win for $q=5$ : The most significant new finding is for the field with 5 elements ( $F_5$ $F_{5}$ ). They studied the polynomial $e_4$ $e_{4}$ (which involves multiplying four numbers at a time).
- Before this paper, for the case of 5 elements and degree 4, the best known bound was "trivial" (meaning it did not actually prove that a solution must exist within a reasonable group size).
- The authors proved a non-trivial exponential bound. They showed that the maximum size of a group avoiding a zero-sum solution is at most roughly $480 \times (4.9556902)^n$ .
- Crucially, the base 4.9556902 is strictly less than 5. This proves that for large groups of numbers in this specific setting, you are guaranteed to find a zero-sum solution, and the group size needed is significantly smaller than the total number of possible combinations.

What They Did Not Do
It is important to note what this paper does not claim.

They did not solve the problem for every possible polynomial or every field size. Their new, tightest bounds are specifically for the quadratic case in characteristic three and the degree-4 case in characteristic five.
They did not claim to have found the absolute smallest possible number (the exact "Erdős–Ginzburg–Ziv constant"). They found an upper bound—a guarantee that the answer is at most this number. The true answer might be even smaller.
For the field of 3 elements ( $F_3$ ), they noted that their new method still gives a trivial result (the base is 3, which is not smaller than the alphabet size). They explicitly state that it remains an open question whether a non-trivial bound exists for $F_3$ using this specific approach.

In Summary
This paper is a masterclass in refining mathematical tools. By creating a precise "certificate" to measure the size of solution sets and by inventing a way to break down the "all distinct" rule into simpler pieces, the authors have tightened the net on these zero-sum problems. They proved that for specific, complex algebraic games played with numbers, the "safe zone" (where you can avoid a zero-sum) is smaller than we thought, and for the first time, they provided a concrete, non-trivial guarantee for the difficult case of 5-element fields. They did not just say "it's possible"; they gave a specific, calculable limit on how big the group can get before the zero-sum becomes unavoidable.

Technical Summary: Slice and Partition Rank Criteria for Polynomial Zero-Avoidance

Problem Statement
This paper investigates polynomial zero-avoidance over finite vector spaces $\mathbb{F}_q^n$ , specifically focusing on the higher-degree Erdős–Ginzburg–Ziv (EGZ) constant. The classical EGZ constant asks for the minimum length of a sequence in a finite abelian group that guarantees a zero-sum subsequence of a prescribed length. The higher-degree version, introduced by Caro and Schmitt, replaces the linear sum with an elementary symmetric polynomial $e_m$ . The authors seek to determine the least integer $\ell$ such that every sequence of length $\ell$ in $\mathbb{F}_q^n$ contains a subsequence of length $t$ on which $e_m$ vanishes coordinatewise.

The paper addresses two primary difficulties in applying tensor rank methods to this problem:

Support Entropy Gaps: Establishing explicit, quantitative entropy gaps for the coefficient support of the detecting tensor, particularly when the support admits no probability distribution with uniform marginals.
Pairwise Distinctness: Handling the constraint that variables in the subsequence must be pairwise distinct, which introduces partial diagonals that standard slice-rank arguments do not automatically eliminate.

Methodology
The authors employ two main tools: slice rank and partition rank.

Slice Rank and Dual Certificates: The paper refines the support-entropy method. While previous work established that a non-uniform marginal condition implies a sub-exponential bound, it often lacked explicit numerical gaps. The authors formulate the existence of a uniform-marginal distribution as a finite linear feasibility problem. By using a separating hyperplane, they construct an explicit dual certificate (a set of functions $w_i$ and a margin $\eta$ ) that yields a quantitative entropy gap. This allows for explicit bounds without solving a new linear program for each specific field size.
Partition Rank and Equality Profiles: To handle the "pairwise distinct" constraint, the authors utilize Naslund's partition-rank framework. They expand the distinctness indicator over the lattice of set partitions. This expansion reduces the global problem of finding distinct zeros to a finite sum of slice-rank estimates on contracted local tensors. Each term in the expansion corresponds to an "equality profile" (a partition of the variables).
- Contractions with many blocks are controlled by degree bounds.
- Contractions with few blocks are analyzed by checking if their zero sets consist solely of diagonal points (constant tuples). If so, these profiles can be eliminated or handled via Möbius inversion on the partition lattice.
Restricted Alphabets and Stratification: The authors apply these methods to the multiplicative torus $(\mathbb{F}_q^\times)^n$ . By retaining the restricted alphabet, they derive bounds with exponential bases strictly smaller than the torus size ( $q-1$ ). They then use coordinatewise inversion to map problems involving $e_{p-1}$ to linear equations on the torus. Finally, support stratification lifts these toric bounds to the full space $\mathbb{F}_q^n$ by summing over strata defined by the support of the vectors.

Key Contributions and Results

Explicit Entropy Gaps via Dual Certificates:
The authors provide a method to convert the qualitative condition "no uniform marginal distribution exists" into an explicit quantitative bound. For the quadratic elementary symmetric polynomial $e_2$ over fields of characteristic three ( $\mathbb{F}_{3^k}$ ), they derive a certificate with an optimal normalized margin. This yields a uniform analytic bound for the EGZ constant:
$EGZ(3, \mathbb{F}_{3^k}^n, 2) \le C_k^{n+o(n)}$
where $C_k < 3^k$ is given by a one-dimensional infimum formula, avoiding the need for field-by-field optimization.
Partition-Rank Reduction for Distinct Variables:
The paper formalizes the reduction of the distinct-variable problem to slice-rank estimates on contracted tensors. This allows the authors to handle cases where the standard degree-splitting argument (which requires $t > 2m$ ) fails.
New Bounds for Higher-Degree EGZ Constants:
- Restricted Alphabet: The authors establish nontrivial exponential bounds for zero-sum problems on the multiplicative torus. For $q=9$ (characteristic 3), the bound for cap sets on the torus improves upon viewing $\mathbb{F}_9^n$ as $\mathbb{F}_3^{2n}$ and applying standard cap-set estimates.
- Full Space $\mathbb{F}_5^n$ : The main new result concerns the case $q=p=5$ and $m=4$ . The authors prove:
  $EGZ(5, \mathbb{F}_5^n, 4) \le 480(1 + \vartheta_5)^n + 1$
  where $1 + \vartheta_5 \approx 4.95569 < 5$ .
  Additionally, they bound the size of strongly $e_4$ -free subsets:
  $|A| \le 5(1 + \eta_{5,5})^n$
  where $1 + \eta_{5,5} \approx 4.25636 < 5$ .
  These are the first nontrivial exponential bounds for the higher-degree EGZ problem over $\mathbb{F}_5^n$ associated with the fourth elementary symmetric polynomial.
Comparison with Previous Work:
The paper notes that while their toric-stratification method yields a uniform bound, it is not uniformly sharper than previous computational optimizations for all $q$ . For instance, for $q=9$ , the direct slice-rank analysis of Section 3 yields a slightly better base than the toric-stratification approach. However, for $q=p=5$ , the toric-stratification method provides the first nontrivial bound, whereas previous general estimates were trivial (i.e., the base was equal to $q$ ).

Significance
The paper claims significance in making the support-entropy method "effective" by providing explicit dual certificates that yield quantitative gaps without ad-hoc optimization. Furthermore, it successfully extends partition-rank techniques to handle distinct-variable constraints in higher-degree polynomial zero-avoidance, a setting where standard degree arguments fail. The primary achievement is the derivation of the first nontrivial exponential bounds for the higher-degree EGZ constant over $\mathbb{F}_5^n$ for $e_4$ , demonstrating that the combination of partition rank, coordinatewise inversion, and support stratification can yield results in regimes previously inaccessible to these methods.

The Great Zero-Hunt in a Digital Maze

The Paper's Discovery: Sharper Flashlights and New Tricks

More like this