🔢 mathematics

On the Waring Problem for Matrices over Finite Fields

The paper proves that for any finite field $\mathbb{F}_q$ with $q \neq 2$ and any positive integer $n$ satisfying $q^n > (k-1)^4$ , every $n \times n$ matrix over $\mathbb{F}_q$ can be expressed as the sum of two $k$ -th powers.

Original authors: Simion Breaz

Published 2026-07-17

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Simion Breaz

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 a world where numbers aren't just for counting, but for building structures. In the realm of mathematics, there's a famous puzzle called "Waring's Problem." Think of it like a game where you have a specific type of building block (a number) and you want to see if you can build any other number by stacking up a certain number of those blocks. For example, can you make any number by adding up three perfect cubes? Mathematicians have been solving this for regular numbers for a long time. But this paper takes that game and moves it into a stranger, more colorful universe: the world of "matrices" over "finite fields."

To understand this, picture a matrix not as a boring grid of numbers, but as a complex machine or a magical transformation that can shuffle, stretch, or rotate a set of objects. A "finite field" is like a universe with a limited number of elements—imagine a clock that only has 5 hours, or a deck of cards with only 10 specific cards. In this tiny, closed universe, you can still do math, but the rules are different; once you go past the limit, you wrap around. The question this paper tackles is: In these tiny, finite universes, can we always build any complex machine (matrix) by adding together just two "k-th powers"? A "k-th power" here is like taking a machine and running it through itself $k$ times. The goal is to see if, no matter how complicated the machine is, we can always find two simpler machines (that have been powered up $k$ times) that, when added together, recreate the original complicated one.

This isn't just a game for mathematicians; it helps us understand the fundamental building blocks of algebra and how information can be broken down and reconstructed in digital systems. If we know that any machine can be built from just two powered-up parts, it tells us something profound about the flexibility and structure of these mathematical worlds.

The Paper's Big Discovery

In this paper, the author, Simion Breaz, acts like a master architect trying to prove that in most of these tiny, finite universes, you never need more than two "super-charged" machines to build any other machine. The paper focuses on a specific condition: the size of the universe (the number of elements, $q$ ) and the size of the machine (the number of rows and columns, $n$ ).

The main finding is a powerful guarantee. The author proves that if the universe is big enough—specifically, if the universe has a cardinality $q \neq 2$ and the total number of elements in the universe raised to the power of the machine's size ( $q^n$ ) is greater than $(k-1)^4$ —then every $n \times n$ matrix over that field can be expressed as the sum of exactly two $k$ -th powers. There is one small exception: the paper explicitly rules out the universe where $q=2$ (the smallest possible field with just two elements) for this specific general rule, noting that the case for $q=2$ was already investigated and solved differently in other work.

The proof is a clever construction. The author doesn't just say "it works"; they show how to find these two special machines. They use a strategy involving "companion matrices," which are special, standard forms of machines that are easy to analyze. The logic goes like this:

First, they show that if the field is large enough, you can always find a "primitive" machine that, when powered up $k$ times, creates a unique, non-repeating pattern.
Then, they use a "trace" (a specific number calculated from the machine that acts like a fingerprint) to match the pieces.
Finally, they demonstrate that you can split any non-special machine into two parts: one part that is a $k$ -th power, and another part that is also a $k$ -th power, provided the universe is big enough to hold the necessary variety of patterns.

The paper is very confident in its results. It doesn't just suggest or simulate; it provides a rigorous mathematical proof. The author establishes a clear boundary: if $q \neq 2$ and $q^n > (k-1)^4$ , the statement is true. This is a "weak version" of a famous conjecture by a mathematician named Larsen. Larsen's original conjecture predicted that a relationship involving $q^{n^2}$ would be sufficient to guarantee that every matrix is a sum of two $k$ -th powers. This paper confirms that a slightly simpler relationship ( $q^n$ ) is sufficient to guarantee the result for non-scalar fields, offering a significant step toward understanding the full conjecture.

The author also tackles the tricky case of "scalar matrices" (machines that just scale everything by the same amount). They prove that even these special, uniform machines can be built from two $k$ -th powers under the same condition.

So, what does this mean for our story? It means that in almost any finite mathematical universe that isn't the tiniest possible one, the "Waring Problem" for matrices is solved with a very low number: two. You don't need a pile of ten or twenty powered-up machines to build a complex one; you only need two. The paper draws a line in the sand: as long as your universe is big enough relative to how many times you power up your machines ( $k$ ), the construction is always possible. It's a definitive "yes" for a vast range of mathematical scenarios, turning a difficult puzzle into a solved recipe for building matrices.

Technical Summary: On the Waring Problem for Matrices over Finite Fields

Problem Statement
This paper addresses the Waring problem for matrices over finite fields. Specifically, given a finite field $\mathbb{F}_q$ and a positive integer $k$ , the problem seeks to determine conditions under which every $n \times n$ matrix over $\mathbb{F}_q$ can be expressed as a sum of two $k$ -th powers. While previous literature has established that matrices are sums of a bounded number of $k$ -th powers (e.g., seven) under certain conditions, and has provided lower bounds for $q$ involving high-degree polynomials in $k$ , this work aims to establish explicit, tighter conditions on $q$ and $n$ that guarantee every matrix is a sum of exactly two $k$ -th powers. The paper specifically targets the case where $q \neq 2$ , noting that the case $q=2$ has been investigated separately in prior works.

Methodology
The author employs a combination of algebraic number theory, finite field theory, and matrix canonical forms to construct the required decompositions. The methodology proceeds through several key steps:

Companion Matrices and $k$ -th Powers: The paper utilizes the property that if a polynomial $P$ is irreducible and its associated companion matrix $C_P$ is a $k$ -th power, then $C_P$ serves as a building block for matrix decompositions. Lemma 2.2 establishes that if the orbit of an element $a \in \mathbb{F}_{q^n}$ under the Frobenius automorphism yields distinct values for $a^k$ , then the companion matrix $C_{P_{a^k}}$ is a $k$ -th power in $M_n(\mathbb{F}_q)$ .
Existence of Irreducible Polynomials with Prescribed Traces: To ensure the flexibility needed for decomposition, the author relies on Proposition 2.4. This proposition uses the Weil bound for the number of solutions to trace equations and estimates on the number of irreducible polynomials (via the Möbius function) to prove that, under specific inequalities involving $k, q,$ and $n$ , there exist irreducible polynomials of degree $n$ with an arbitrary prescribed trace.
Additive Decomposition via Canonical Forms: The core structural argument relies on Lemma 2.6, which demonstrates that any non-scalar matrix $A$ over a field $\mathbb{F} \neq \mathbb{F}_2$ can be decomposed (up to similarity) into the sum of two matrices: a non-derogatory matrix $D$ of size $(n-1) \times (n-1)$ and a non-derogatory matrix $E$ of size $n \times n$ , subject to a trace constraint.
Construction of $k$ -th Powers: By selecting $D$ and $E$ such that they are $k$ -th powers (using the results from steps 1 and 2) and satisfying the trace condition $t = \text{Tr}(A) - \text{Tr}(D) - \text{Tr}(E)$ , the author constructs the decomposition $A \sim D' + E'$ .
Scalar Matrices: For scalar matrices, the paper treats the problem within the algebra $\mathbb{F}_q[C]$ generated by a companion matrix with an irreducible characteristic polynomial, applying results from [17] to show that non-zero scalars are sums of two $k$ -th powers under the derived bounds.

Key Results
The paper establishes the following main theorems and corollaries:

Theorem 2.10: If $n$ and $k$ are positive integers satisfying the condition:
$k < \frac{q^{n/2} - \omega(n)q + q - 1}{q - 1}$
where $\omega(n)$ is the number of prime divisors of $n$ , then every non-scalar $n \times n$ matrix over $\mathbb{F}_q$ ( $q \neq 2$ ) is a sum of two $k$ -th powers.
Corollary 2.14 (Main Result): If $q \neq 2$ and the inequality $q^n > (k-1)^4$ holds, then every $n \times n$ matrix over $\mathbb{F}_q$ is a sum of two $k$ -th powers. This result covers both non-scalar and scalar matrices.
Corollary 2.15: For $n \geq 4$ , if $k \leq q$ , every $n \times n$ matrix is a sum of two $k$ -th powers.

Significance and Contribution
The paper claims to provide a "weak version" of a conjecture by Larsen, which posits that for every $k$ , there exists a constant $C_k$ such that if $q^{n^2} > C_k$ , every matrix is a sum of two $k$ -th powers. While the paper does not prove the full conjecture, it improves upon existing results in the following ways:

Explicit Bounds: Previous results (e.g., [9], [10]) provided bounds for $q$ as polynomial expressions in $k$ involving radicals (e.g., $C_k \leq k^{16}$ or $k^3 - 3k^2 + 3k$ ). This paper provides a bound of the form $q^n > (k-1)^4$ , which captures a relationship where the field size and dimension interact more directly, rather than just a lower bound on $q$ for a fixed $n$ .
Two Powers: The results specifically target the decomposition into two powers, which is a stronger condition than the previously established bounds for sums of seven or more powers.
Generality: The results apply to all $n \times n$ matrices (including scalar matrices) over finite fields with cardinality $q \neq 2$ , provided the dimension and field size satisfy the derived inequality.

The author notes that the case $q=2$ was previously investigated in [10], where it was shown that for odd $k$ , all matrices over $\mathbb{F}_2$ are sums of two $k$ -th powers. The current work focuses on extending these findings to general finite fields with $q \neq 2$ .

More like this