🔢 mathematics

Evaluation codes from linear systems of conics

This paper investigates the even characteristic case of a generalization of the Datta-Johnsen evaluation code, which is constructed by evaluating a low-dimensional linear system of symmetric polynomials on points with pairwise distinct coordinates in an affine space over a finite field.

Original authors: Barbara Gatti, Gábor Korchmáros, Gioia Schulte

Published 2026-05-13

📖 4 min read🧠 Deep dive

CC BY 4.0

Original authors: Barbara Gatti, Gábor Korchmáros, Gioia Schulte

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

Imagine you are a librarian trying to organize a massive, chaotic collection of books. You want to create a special "code" (a secret language) to store information efficiently. In the world of mathematics, this is done using evaluation codes. Think of these codes as a way to turn a list of numbers (a message) into a pattern of dots on a grid, where the pattern is created by drawing specific shapes (polynomials) over a finite field (a world with a limited number of points, like a pixelated screen).

This paper is about refining a specific type of code called the Datta-Johnsen code. Here is the story of what the authors did, explained simply:

1. The Setup: Symmetric Patterns

Usually, when you write a code, you might use any shape you want. But this paper focuses on symmetric polynomials.

The Analogy: Imagine you have two variables, $X$ and $Y$ . A "symmetric" rule is one where it doesn't matter if you swap them. If you have a rule like "Add the two numbers," it's symmetric because $3+5$ is the same as $5+3$ .
The authors look at a specific set of points in a 2D grid (the affine plane) where the coordinates are all different from each other. They call these "distinguished points."

2. The Problem: Odd vs. Even

In a previous study, mathematicians figured out how these codes worked when the size of the grid ( $q$ ) was an odd number (like 3, 5, 7). In that world, there was a clear "outside" to a parabola (a U-shaped curve), and the code worked by looking at points outside that curve.

However, this paper tackles the even case (where $q$ is a power of 2, like 2, 4, 8, 16).

The Twist: In an even-numbered world, the concept of "outside a parabola" disappears. It's like trying to find the "outside" of a circle in a world where circles don't exist the same way. The old rules don't apply.

3. The New Map: The "Trace" Parabolas

The authors had to invent a new way to map the points.

The Metaphor: Instead of looking for points outside a single shape, they realized the points they care about are covered by a family of parabolas.
Imagine a set of U-shaped curves, each defined by a specific rule involving a "trace" (a mathematical sum of powers). The authors proved that if you take all these specific parabolas, they perfectly cover the set of points they need, with every point being covered exactly once.
They call this new set of points $\Delta$ . It's their new "playground" for the code.

4. The Challenge: Counting the Intersections

To know how good the code is, they needed to know: "If I draw a random conic section (a circle, ellipse, parabola, or hyperbola) on this grid, how many points of $\Delta$ will it hit?"

The Difficulty: In the odd world, this was easy. In the even world, it's like trying to predict how many fish a net will catch in a stormy ocean. The shapes behave differently.
The Solution: The authors used advanced geometry (algebraic curves) to count these intersections. They found that for most shapes, the number of points hit is within a predictable range. However, there are a few "exceptional" shapes that hit way more or way fewer points.

5. The Result: Better Codes

Using this new understanding of the "even" world, they built two specific types of codes:

Code 1 (The 3-Dimensional Code): They created a code with 3 "degrees of freedom." They proved that the "minimum distance" (a measure of how much error the code can fix) is very high. In fact, they showed that for a grid size of 8, this code is nearly perfect, matching the best possible theoretical limit.
Code 2 (The 4-Dimensional Code): They built a slightly larger code with 4 degrees of freedom. They calculated the exact "weight distribution," which is like a report card showing exactly how many errors different messages can handle.

Summary

Think of the paper as a guidebook for a new territory.

Previous Map: Worked for odd-numbered grids.
New Territory: Even-numbered grids (powers of 2).
New Discovery: The "playground" isn't the outside of a single curve, but a collection of specific parabolas.
The Payoff: By understanding this new landscape, the authors constructed stronger, more efficient error-correcting codes that can handle more mistakes than before, specifically for these even-sized grids.

They didn't just guess; they used deep geometry to prove exactly how many points these shapes would catch, ensuring the codes are mathematically sound and optimal.

Technical Summary: Evaluation Codes from Linear Systems of Conics

Problem Statement
This paper investigates the parameters of evaluation codes derived from linear systems of symmetric polynomials over finite fields $\mathbb{F}_q$ of even characteristic ( $q=2^h$ ). Specifically, the authors focus on the "reduced generalized Datta-Johnsen code," where codewords are formed by evaluating polynomials from a chosen linear system $V$ on the set of "distinguished points" in the affine plane $AG(2, \mathbb{F}_q)$ . Distinguished points are those with pairwise distinct coordinates. The evaluation set is reduced to representatives of equivalence classes under coordinate permutation, effectively identifying the set of points with the unramified points of the quotient variety $\mathbb{F}_q^2 / \text{Sym}_2$ .

While the odd characteristic case was previously resolved by identifying the evaluation set with the external points of a parabola, the even characteristic case presents a distinct geometric challenge: in $AG(2, \mathbb{F}_q)$ with $q$ even, there are no external points to a conic. The paper aims to characterize the intersection sizes between this specific pointset and conics to determine the weight distribution and minimum distance of the resulting codes.

Methodology
The authors employ a combination of finite geometry, algebraic geometry, and coding theory. The core methodology involves:

Geometric Characterization of the Evaluation Set:
The evaluation set $\Delta$ (excluding the origin) is identified as the set of points covered by a specific family of parabolas. For even $q$ , $\Delta$ is shown to be the set of points $(x, ax^2)$ where $x \in \mathbb{F}_q^*$ and the coefficient $a$ has a trace of zero ( $\text{Tr}(a)=0$ ). This set is denoted as $\bar{\Delta} \setminus \{(0,0)\}$ , where $\bar{\Delta}$ is covered by parabolas $Y = aX^2$ with $\text{Tr}(a)=0$ .
Transformation to Algebraic Curves:
To determine the size of the intersection $|\Delta \cap C|$ for a conic $C$ , the authors utilize a quadratic transformation $\phi_2: (x_1, x_2) \to (x_1+x_2, x_1x_2)$ . This maps the problem of counting intersections in the symmetric domain to counting rational points on specific plane algebraic curves defined over $\mathbb{F}_q$ .
- The intersection size is related to the number of rational points $N_q(F)$ on a curve $F$ derived from the conic equation.
- Further quadratic transformations are applied to map $F$ to curves $G$ and subsequently to cubic curves $H$ .
Analysis of Cubic Curves:
The problem is reduced to analyzing the number of rational points on the cubic curve $H$ . The authors distinguish between cases where $H$ is reducible (degenerate) and irreducible.
- For irreducible $H$ , the authors apply the Hasse-Weil bound and results by Waterhouse regarding the spectrum of points on elliptic and rational curves.
- For reducible $H$ , the analysis relies on the decomposition of the curve into linear components (lines) and counting their intersections with the specific pointset.
Coding Theory Application:
The derived intersection sizes are used to calculate the weight distribution of the evaluation codes. The weight of a codeword corresponds to the number of points in $\Delta$ not lying on the curve defined by the polynomial.

Key Contributions and Results

Geometric Description of $\Delta$ : The paper establishes that for even $q$ , the set $\Delta$ is covered exactly once by the family of parabolas $Y = aX^2$ where $\text{Tr}(a)=0$ . This provides the necessary geometric framework to replace the "external points" concept used in odd characteristic.
Intersection Spectrum: The authors determine the possible sizes of the intersection between $\Delta$ $Δ$ and a non-degenerate conic $C$ $C$ .
- General Bound: For a non-degenerate conic $C$ , $|\Delta \cap C|$ is bounded by $\frac{1}{2}(q - 2\sqrt{q} - 2) \leq |\Delta \cap C| \leq \frac{1}{2}(q + 2\sqrt{q} - 1)$ , with specific exceptions.
- Exceptions: The paper identifies specific degenerate cases (e.g., when the conic equation lacks certain terms) where the intersection size can be $0$, $1$, $q-1$ , or $q$ .
Code Parameters:
- Full Conic System: The reduced generalized Datta-Johnsen code arising from the linear system of all conics is determined to be a $[\frac{1}{2}q(q-1), 6, \frac{1}{2}q(q-3)]_q$ code.
- Dimension 3 Subsystem (Construction 1): By selecting a specific 3-dimensional linear system of conics (derived from a net of conics in $PG(2, \mathbb{F}_{q^3})$ invariant under the Frobenius map), the authors construct a code with parameters $[\frac{1}{2}q(q-1), 3, d]_q$ . The minimum distance $d$ is at least $\frac{1}{2}(q^2 - 2q - 2\sqrt{q} + 1)$ . For $q=8$ , exhaustive search confirms $d$ is either 21 or 22, showing the bound is sharp. The dual of a specific instance of this code is a $[28, 25, 3]_8$ code with optimal minimum distance.
- Dimension 4 Subsystem (Construction 2): A 4-dimensional system consisting of all parabolas yields a $[\frac{1}{2}q(q-1), 4, \frac{1}{2}q(q-3)]_q$ code with a fully determined weight distribution.

Significance
The paper successfully extends the theory of Datta-Johnsen codes to the even characteristic case, a scenario where standard geometric interpretations (like external points to a conic) fail. By establishing a rigorous link between the evaluation set and the rational points of specific cubic curves, the authors provide the fundamental parameters (dimension, minimum distance, and weight distribution) for these codes.

The work demonstrates that while the geometric structure differs from the odd characteristic case, the resulting codes maintain strong parameters. Specifically, the construction of a dimension-3 code with a minimum distance approaching the theoretical limit for that dimension highlights the utility of these symmetric polynomial evaluations. The results contribute to the broader understanding of evaluation codes on finite geometries and the interplay between symmetric polynomials, quotient varieties, and algebraic curves over finite fields.