🔢 mathematics

On the existence of linear rank-metric intersecting codes

This paper establishes new parameter restrictions and geometric characterizations for linear rank-metric intersecting codes using $q$ -systems, proving that the previously open upper bound $n \le 2m-3$ is attainable only for specific parameters and resolving the non-existence of $[6,3,3]_{q^5/q}$ codes for any prime power $q$ .

Original authors: Martino Borello, Olga Polverino, Ferdinando Zullo

Published 2026-04-03

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Martino Borello, Olga Polverino, Ferdinando Zullo

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 organizing a massive, high-stakes game of Secret Handshakes.

In this game, every player holds a secret "code" (a list of numbers). The rules of the game are strict:

The Intersection Rule: If you take any two players' secret codes and look at them side-by-side, they must share at least one number in the same spot. They can't be completely different; they have to overlap.
The Rank-Metric Twist: This isn't just a simple list of numbers. These codes live in a complex, multi-dimensional world (like a grid that exists in multiple layers of reality). The "distance" between codes isn't just about how many numbers differ, but how many entire layers of information differ.

This paper is about figuring out the limits of this game. Specifically, the authors are trying to answer: How long can these secret codes be before the rules break down?

The Characters in Our Story

To make this understandable, let's use a few metaphors:

The Codes: Think of these as unique fingerprints. Each fingerprint is a pattern of ridges.
The "Rank-Metric" World: Imagine these fingerprints aren't just on a flat piece of paper, but are carved into 3D blocks of glass. The "rank" is how many layers of glass you have to slice through to see the pattern.
The "Intersecting" Rule: Two fingerprints are "intersecting" if, when you stack them, their ridges touch at least once.
The "Dual" Subspace: Think of this as the shadow cast by the fingerprint. If the fingerprint is the object, the dual is its shadow on the wall. The authors realized that to understand the fingerprint, you have to study the shape of its shadow.
The "Scattered" Subspace: This is the most important character. Imagine a sprinkler watering a garden. A "scattered" pattern is one where the water drops are spread out so perfectly that no two drops ever land in the same small puddle. They are maximally spread out, avoiding clumps.

The Big Discovery: The "Shadow" Rule

The authors (Borello, Polverino, and Zullo) took a deep look at these codes and their "shadows" (dual subspaces). They found a surprising secret:

For a code to follow the "Intersection Rule," its shadow must be incredibly "evasive."

It's like a ninja. The shadow must be so good at hiding that it never gets "caught" by certain geometric traps. If the shadow is too "clumpy" (not evasive), the original code breaks the rules.

This "evasive" property is the key that unlocks the math. It's like realizing that to build a house that never collapses, the foundation must be made of a specific, rare type of rock.

The Main Findings (The "Aha!" Moments)

1. The Length Limit (The Ceiling)
Previously, mathematicians knew there was a ceiling on how long these codes could be. They knew the length ( $n$ ) couldn't exceed a certain number related to the complexity of the world ( $m$ ).

Old Guess: The ceiling was $2m - 3$ .
New Discovery: The authors proved this ceiling is real, but it's very hard to reach. You can only hit this maximum length if the code is very small (dimension $k=3$ ) and the world is large enough ( $m \ge 6$ ).
The Analogy: It's like saying, "You can only build a skyscraper that touches the clouds if you use a very specific, narrow design and the ground is very solid." For most other designs, the building must be shorter.

2. The "Scattered" Connection
They found that the only way to build a code that hits this maximum length is if its "shadow" is a Scattered Subspace.

Translation: To build the perfect code, you need to arrange your "water drops" (the shadow) so perfectly that they are maximally spread out.
The Good News: We know how to build these perfect "scattered" patterns when the world size ( $m$ ) is an even number. So, for even numbers, these perfect codes exist.
The Bad News: When the world size is an odd number, we don't know how to build these patterns yet. So, for odd numbers, the existence of these perfect codes is still a mystery.

3. Solving a Specific Mystery
There was one specific case that had been bothering mathematicians: A code with length 6, dimension 3, in a world of size 5 ( $[6, 3, 3]_{q^5/q}$ ).

The Verdict: The authors proved with a clever combinatorial argument (counting lines and points like a detective) that this code cannot exist. It's impossible to arrange the fingerprints to satisfy the rules in this specific scenario. They closed the door on this open problem.

Why Should You Care?

You might think, "Who cares about secret handshakes in 3D glass blocks?"

Well, these codes are the backbone of future technology:

Secure Communication: They are used in "Secret Sharing" schemes (splitting a password among friends so only a group can unlock it).
Network Coding: They help data travel efficiently through the internet without getting lost.
Quantum Security: As we move toward quantum computers, these specific types of codes are being studied to protect our data from future hackers.

The Takeaway

This paper is like a map for treasure hunters.

Before: They knew the treasure (the perfect code) was somewhere in a huge region, but the map was fuzzy.
Now: The authors drew a precise line. They said, "The treasure only exists in this specific valley (when $k=3$ and $m$ is even). If you look in the mountains (odd $m$ ) or the plains (other dimensions), you won't find it."
The Mystery Remains: They also found a specific spot (the $m=5$ case) where they proved the treasure definitely isn't there, solving a puzzle that had been open for a while.

In short: They used the geometry of shadows to prove exactly where these magical, overlapping codes can and cannot exist, bringing us one step closer to building more secure and efficient communication systems.

1. Problem Statement

The paper investigates the existence and structural constraints of linear rank-metric intersecting codes.

Definition: A rank-metric code $C \subseteq \mathbb{F}_{q^m}^n$ is intersecting if the rank supports of any two distinct non-zero codewords have a non-trivial intersection.
Context: While intersecting codes are well-studied in the Hamming metric (where they coincide with minimal codes in the binary case), their rank-metric analogue exhibits distinct behaviors.
Open Problem: Previous work [BBMS25] established that for a non-degenerate $[n, k, d]_{q^m/q}$ rank-metric intersecting code, the parameters must satisfy $2k - 1 \le n \le 2m - 3$ . While the lower bound was proven tight, the tightness of the upper bound ( $n = 2m - 3$ ) remained an open question. Specifically, it was unknown for which dimensions $k$ and field extension degrees $m$ this maximum length is achievable, and whether codes exist for lengths in the gap between $2m - 2k + 2$ and $2m - 3$ .

2. Methodology

The authors employ a geometric approach based on the correspondence between rank-metric codes and $q$ -systems (subspaces of $\mathbb{F}_{q^m}^k$ over $\mathbb{F}_q$ ).

Geometric Translation: A rank-metric code $C$ corresponds to an $\mathbb{F}_q$ -subspace $U \subseteq \mathbb{F}_{q^m}^k$ . The code is intersecting if and only if its associated system $U$ is not 2-spannable (i.e., $U$ cannot be generated by the sum of its intersections with two hyperplanes).
Dual Space Analysis: The authors analyze the dual subspace $U^{\perp'}$ (defined via a trace bilinear form). They prove that if $C$ is intersecting, $U^{\perp'}$ must satisfy strong evasiveness properties.
Evasive Subspaces: A subspace is $(h, r)$ -evasive if its intersection with any $h$ -dimensional $\mathbb{F}_{q^m}$ -subspace has dimension at most $r$ . The authors connect the "not 2-spannable" condition to $U^{\perp'}$ being a specific type of evasive subspace.
Combinatorial Counting: For specific small cases (e.g., $n=6, m=5$ ), the authors use projective geometry (linear sets in $PG(2, q^m)$ ) and double-counting arguments on the weights of points and lines to derive contradictions.

3. Key Contributions and Results

A. New Parameter Bounds

The authors derive a refined upper bound on the length $n$ of rank-metric intersecting codes:
$n \le 2m - \left\lfloor \frac{k+4}{2} \right\rfloor$
This improves upon the previous loose bound of $n \le 2m - 3$ for general $k$ .

B. Characterization of Extremal Codes ( $n = 2m - 3$ )

The paper completely characterizes when the maximum length $n = 2m - 3$ is attainable:

Existence Condition: A non-degenerate $[2m-3, k, d]_{q^m/q}$ $[2 m - 3, k, d]_{q^{m} / q}$ rank-metric intersecting code exists if and only if:
1. $k = 3$ ;
2. $m \ge 6$ ;
3. The dual system $U^{\perp'}$ is a scattered subspace of dimension $m+3$ in $\mathbb{F}_{q^m}^3$ .
Minimum Distance: In this extremal case, the minimum distance is fixed at $d = m - 1$ .
Implication: The existence of such codes is reduced to the existence of scattered subspaces of dimension $m+3$ in $\mathbb{F}_{q^m}^3$ .

C. Existence Results

Even $m$ : Using known constructions of maximum scattered subspaces (which exist for dimension $3m/2$ when $m$ is even), the authors prove that for every even $m \ge 6$ , a rank-metric intersecting code with parameters $[2m-3, 3, m-1]_{q^m/q}$ exists. This confirms the tightness of the upper bound for $k=3$ and even $m$ .
Odd $m$ : The existence remains open for odd $m \ge 7$ , as the construction of scattered subspaces of dimension $m+3$ in $\mathbb{F}_{q^m}^3$ is not fully resolved for these parameters.

D. Non-Existence of Specific Cases

The authors resolve a specific open problem from [BBMS25] regarding the case $[6, 3, 3]_{q^5/q}$ :

Result: They prove that no $[6, 3, 3]_{q^5/q}$ rank-metric intersecting code exists for any prime power $q$ .
Method: By analyzing the associated linear set in $PG(2, q^5)$ , they show that the combinatorial constraints on the number of lines of specific weights lead to a contradiction (specifically, a negative number of lines) for both $q=2$ and $q > 2$ .

4. Significance

Theoretical Advancement: The paper bridges the gap between rank-metric coding theory and finite geometry, specifically utilizing the theory of scattered subspaces and evasiveness. It establishes that the structure of intersecting codes is fundamentally tied to the geometry of their duals.
Resolution of Open Problems: It settles the tightness of the upper bound $n \le 2m-3$ , showing it is only achievable for $k=3$ (and not for higher dimensions). It also definitively answers the non-existence question for the $[6, 3, 3]_{q^5/q}$ case.
Future Directions: The work highlights that the existence of these codes for odd $m$ and for lengths $n = 2m-4$ (even for even $m$ ) depends on unsolved problems regarding the existence of scattered subspaces in specific dimensions, guiding future research in finite geometry.

Summary Table of Main Results

Parameter/Case	Result	Condition
General Bound	$n \le 2m - \lfloor \frac{k+4}{2} \rfloor$	$k \ge 3$
Max Length ( $n=2m-3$ )	Exists iff $k=3, m \ge 6$	Dual is scattered subspace of dim $m+3$
Existence (Even $m$ )	Exists	For all even $m \ge 6$
Existence (Odd $m$ )	Open	Depends on scattered subspace constructions
Specific Case	Non-existent	$[6, 3, 3]_{q^5/q}$ for any $q$