🔢 mathematics

Positivity preservers over finite fields

This paper resolves an algebraic version of Schoenberg's theorem by characterizing entrywise matrix transforms that preserve positive definiteness over finite fields, proving that for dimensions three and higher, these preservers are precisely the positive multiples of field automorphisms, while also providing results for dimension two under specific conditions.

Original authors: Dominique Guillot, Himanshu Gupta, Prateek Kumar Vishwakarma, Chi Hoi Yip

Published 2026-02-05

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Dominique Guillot, Himanshu Gupta, Prateek Kumar Vishwakarma, Chi Hoi Yip

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 have a giant grid of numbers, like a spreadsheet or a chessboard. In the world of mathematics, some of these grids are special; they are called "positive definite." Think of a positive definite grid as a perfectly balanced structure. If you push on it from any angle, it springs back in a predictable, stable way.

Now, imagine you have a magic function (a rule) that changes every single number in that grid individually. For example, the rule might say "square every number" or "take the cube root." The big question mathematicians have asked for decades is: Which magic rules keep the grid balanced? If you start with a stable grid and apply your rule, does it stay stable?

For a long time, mathematicians only knew the answer for grids made of real numbers (like 1, 2, 3.5, etc.). They found that the rules that work are very specific: they are like adding up different powers of numbers (like $x + x^2 + x^3$ ) with only positive weights. This was a famous discovery by a man named Schoenberg in 1942.

The New Discovery: The Finite Field Puzzle

In this paper, the authors ask a much harder question: What if the numbers in our grid aren't real numbers, but come from a finite field?

Think of a finite field as a "clock" system with a limited number of hours. Instead of counting 1, 2, 3... forever, you wrap around. If you have a clock with 7 hours, the numbers are just 0, 1, 2, 3, 4, 5, 6. Once you hit 6, the next number is 0 again. In this world, "positive" doesn't mean "greater than zero" in the usual sense; it means the number is a "perfect square" on this clock (like how 4 is a square because $2 \times 2 = 4$ ).

The authors wanted to find out: What rules preserve the "stability" of these finite grids?

The Surprising Answer

The authors found a result that is surprisingly simple, yet very different from the real-number world.

For Large Grids (3x3 or bigger):
If your grid is at least 3x3, the only rules that work are field automorphisms multiplied by a positive number.
- The Analogy: Imagine your finite field is a language with a specific alphabet. An "automorphism" is like a secret code that rearranges the letters of the alphabet in a very specific, consistent way (like a Caesar cipher, but more complex).
- The paper proves that if you want to keep a large grid stable, you can only use these specific "secret codes" (and maybe multiply the result by a positive number). You cannot use random rules like "add 1" or "square the number" unless that squaring happens to be one of these secret codes.
- This is a huge surprise because in the real-number world, there are infinitely many rules that work. In this finite clock world, the list of working rules is extremely short and rigid.
For Small Grids (2x2):
The 2x2 case is much trickier, like trying to balance a pencil on its tip.
- Even Clocks (e.g., 2, 4, 8 hours): The rules are "bijective monomials." This means you can multiply by a number and raise it to a power, as long as the power doesn't repeat numbers.
- Odd Clocks (e.g., 3, 5, 7 hours):
  - If the clock size is 3, 7, 11, etc. (numbers that leave a remainder of 3 when divided by 4), the rules are the same as the large grids: only the secret codes work.
  - If the clock size is 5, 9, 13, etc. (numbers that leave a remainder of 1 when divided by 4), the answer depends on the shape of the clock. If the clock size is a perfect square (like 9 or 25), the authors solved it completely: again, only the secret codes work.
  - The Unsolved Mystery: If the clock size is 13, 17, 29, etc. (numbers that are not perfect squares and leave a remainder of 1), the authors couldn't solve it completely. They suspect the answer is the same (only secret codes work), but the mathematical tools they used (which involve looking at patterns in graphs) hit a wall because the patterns in these specific clocks are too messy to analyze yet.

How They Solved It

The authors didn't use calculus or smooth curves (which work for real numbers). Instead, they used a mix of:

Algebra: Treating the numbers like a rigid code.
Graph Theory: They imagined the numbers as dots on a map (called a Paley graph). Two dots are connected if their difference is a "positive" number. They studied how these dots cluster together (cliques).
Number Theory: They used deep results about how numbers behave on these clocks.

The Big Takeaway

The paper resolves a major algebraic puzzle. It shows that in the world of finite fields, the "positivity preservers" are not a flexible family of functions, but a very rigid, small group of functions that are essentially just rearrangements of the field's structure (automorphisms).

It's like discovering that in a specific type of locked room, the only keys that open the door are the original master keys, and no amount of filing or bending a new key will ever work. This is a stark contrast to the real world, where you can build many different keys that fit the lock.

What They Didn't Solve
The authors admit they couldn't fully solve the puzzle for 2x2 grids when the clock size is a specific type of number (congruent to 1 mod 4 but not a perfect square). They leave this as a challenge for future mathematicians.

Technical Summary: Positivity Preservers Over Finite Fields

1. Problem Statement and Context

The paper addresses the classification of entrywise positivity preservers over finite fields, an algebraic analogue of a celebrated result by Schoenberg (1942) concerning real and complex matrices. In the classical setting, Schoenberg's theorem characterizes functions $f: \mathbb{R} \to \mathbb{R}$ that preserve positive semidefiniteness for matrices of arbitrary dimension as absolutely monotone functions (power series with non-negative coefficients).

While significant progress has been made in characterizing preservers for matrices of a fixed dimension over $\mathbb{R}$ and $\mathbb{C}$ , the problem remains largely open. This paper shifts the setting to finite fields $\mathbb{F}_q$ . The authors define a symmetric matrix $A \in M_n(\mathbb{F}_q)$ to be positive definite if all its leading principal minors are non-zero squares in $\mathbb{F}_q$ (denoted $\mathbb{F}_q^+$ ). The central problem is to characterize all functions $f: \mathbb{F}_q \to \mathbb{F}_q$ such that for any positive definite $A$ , the entrywise transformed matrix $f[A] = (f(a_{ij}))$ is also positive definite.

The authors note that standard definitions of positive definiteness (e.g., via quadratic forms or eigenvalues) do not yield useful notions over finite fields; the definition based on leading principal minors, as established by Cooper, Hanna, and Whitlatch, is adopted.

2. Methodology

Unlike previous works on real or complex preservers which rely heavily on analytic techniques, this paper employs algebraic, combinatorial, and number-theoretic arguments. The methodology is structured around three distinct cases based on the characteristic and properties of the finite field $\mathbb{F}_q$ :

Even Characteristic ( $q = 2^k$ ): Here, every non-zero element is a square. The problem reduces to characterizing transformations that preserve non-singularity. The authors utilize properties of permutation polynomials and monomial maps.
Odd Characteristic with $q \equiv 3 \pmod 4$ : In this case, $-1$ is not a square. The authors leverage the quadratic character $\eta$ and properties of Carlitz's theorem regarding functions preserving the quadratic character of differences. They employ indirect algebraic arguments and character sum estimates (Weil's bound) to establish bijectivity and specific structural properties of the preservers.
Odd Characteristic with $q \equiv 1 \pmod 4$ : Here, $-1$ is a square. This case is significantly more challenging. The authors utilize the structure of Paley graphs $P(q)$ $P (q)$ , where vertices are elements of $\mathbb{F}_q$ $F_{q}$ and edges connect elements whose difference is a square.
- They connect positivity preservers to automorphisms of Paley graphs and their subgraphs.
- They utilize the Erdős-Ko-Rado (EKR) theorem for Paley graphs of square order ( $q=r^2$ ) to analyze maximum cliques.
- They apply Lucas' Theorem for binomial coefficients modulo a prime to analyze power functions in the case where $q \equiv 3 \pmod 4$ .

3. Key Contributions and Results

The paper provides a complete characterization of positivity preservers for matrices of dimension $n \ge 3$ and a near-complete characterization for $n=2$ .

Main Theorems

Let $f: \mathbb{F}_q \to \mathbb{F}_q$ .

Theorem A (Even Characteristic, $q=2^k$ ):
- For $n \ge 3$ , $f$ preserves positivity on $M_n(\mathbb{F}_q)$ if and only if $f$ is a non-zero multiple of a field automorphism (i.e., $f(x) = c x^{2^\ell}$ ).
- For $n=2$ , $f$ preserves positivity if and only if $f$ is a bijective monomial ( $f(x) = cx^n$ with $\gcd(n, q-1)=1$ ).
Theorem B (Odd Characteristic, $q \equiv 3 \pmod 4$ ):
- For $n \ge 2$ , $f$ preserves positivity if and only if $f(0)=0$ and $\eta(f(a)-f(b)) = \eta(a-b)$ for all $a,b$ .
- This is equivalent to $f$ being a positive multiple of a field automorphism ( $f(x) = c x^{p^\ell}$ with $c \in \mathbb{F}_q^+$ ).
Theorem C (Odd Characteristic, $q \equiv 1 \pmod 4$ ):
- For $n \ge 3$ , the characterization is identical to Theorem B: $f$ must be a positive multiple of a field automorphism.
- For $n=2$ , the characterization holds for all $q$ except when $q \equiv 1 \pmod 4$ and $q$ is not a perfect square.
- If $q = r^2$ (a perfect square), the $n=2$ case is also fully characterized as positive multiples of field automorphisms. This relies on the specific structure of cliques in Paley graphs of square order.
Theorem D (Sign Preservers):
- The paper also characterizes sign preservers (functions mapping positive definite matrices to positive definite and non-positive definite to non-positive definite).
- For $n \ge 3$ , these are precisely the positive multiples of field automorphisms.
- For $n=2$ , they are bijective monomials (even $q$ ) or positive multiples of field automorphisms (odd $q$ ).

Corollary 1.4

A surprising consequence of the main results is that for any fixed $n \ge 3$ , the positivity preservers on $M_n(\mathbb{F}_q)$ are precisely the positive multiples of the field automorphisms of $\mathbb{F}_q$ . This implies that such functions preserve the "sign" (quadratic character) of all minors of any matrix.

4. Significance and Claims

The authors claim that their work resolves an algebraic version of Schoenberg's theorem in the finite field setting. The significance lies in:

Completeness for Fixed Dimension: Unlike the real/complex fixed-dimension problem which remains largely open, the authors provide a complete classification for $n \ge 3$ over finite fields.
Novel Connections: The proofs unearth new connections between:
- Positivity preservers and field automorphisms.
- Positivity preservers and automorphisms of Paley graphs.
- The structure of cliques in Paley graphs (specifically using the EKR theorem for square orders).
Dimensional Sensitivity: The results highlight a stark difference between $n=2$ and $n \ge 3$ . While $n \ge 3$ forces the preserver to be a field automorphism (up to a scalar), the $n=2$ case in even characteristic allows for a broader class of bijective monomials.
Open Problem: The authors explicitly state that the case where $q \equiv 1 \pmod 4$ and $q$ is not a square remains unresolved for $n=2$ . They note that estimating the clique number of Paley graphs in this specific case is notoriously difficult, which hinders their current techniques. They pose the question of whether injectivity on $\mathbb{F}_q^+$ is a necessary condition in this unresolved case.

The paper concludes by noting that while it focuses on positive definiteness, the analogous problem for positive semidefiniteness over finite fields remains an open area for future investigation, as the standard notions of semidefiniteness do not translate directly to finite fields.