🔢 mathematics

Single conjugacy classes of isometries in orthogonal groups over local fields

This paper characterizes all isometries in a quadratic space over a non-archimedean local field of characteristic not 2 for which conjugacy within the general linear group implies conjugacy within the orthogonal group, thereby generalizing a previous result by Mil.

Original authors: Fei Xu, Bo Zhang

Published 2026-02-25

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Fei Xu, Bo Zhang

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 master architect working with a very special set of building blocks. These blocks aren't just wood or plastic; they are mathematical shapes called vectors, and they live in a space with a special rulebook called a quadratic space.

In this space, there is a specific type of movement called an isometry (let's call it a "dance move"). This dance move shuffles the blocks around, but it has a strict rule: it must preserve the distance between any two blocks. If two blocks were 5 steps apart before the dance, they must be 5 steps apart after.

Now, imagine you have two different dance moves, Move A and Move B.

If you look at them using a "General Lens" (the General Linear Group, or GL), they might look exactly the same. They shuffle the blocks in the exact same pattern.
But, if you look at them through a "Strict Rulebook Lens" (the Orthogonal Group, or O), they might be different. One might twist a block in a way that the other doesn't, even though the final positions of the blocks are identical.

The Big Question

The authors, Fei Xu and Bo Zhang, are asking a very specific question:

"When can we be absolutely sure that if two dance moves look the same to the General Lens, they are also the same to the Strict Rulebook Lens?"

In math-speak, they are looking for "Single Conjugacy Classes." This means: "Is there only one way to perform this specific dance move that respects the rules?"

The Analogy: The Shuffling Deck

Think of your vector space as a deck of cards.

GL (General Linear): You can shuffle the deck any way you want, as long as the cards end up in the same order. You can stretch the deck, squish it, or twist it, as long as the final sequence matches.
O (Orthogonal Group): You can only shuffle the deck by rotating it or flipping it over. You cannot stretch or squish it. The "shape" of the deck must remain perfect.

The paper asks: If I give you a specific shuffle (Move A), and I tell you that any other shuffle (Move B) that produces the same final card order must be achievable just by rotating or flipping, what does Move A have to look like?

The "Recipe" for a Unique Dance

The authors discovered that for these dance moves to be "unique" (meaning there's no hidden twist hiding in the shadows), the move must follow a very specific recipe. They broke the dance move down into its "ingredients" (mathematically, the characteristic polynomial).

They found that a dance move is unique if and only if it satisfies one of three conditions:

The "Simple" Case: The dance move is very simple. It doesn't have many complex "twists" (mathematically, the number of odd-dimensional "special" parts is 0 or 1, and there are no "reciprocal" pairs).
- Analogy: Imagine a dance where everyone just spins in place or swaps two partners. It's so simple there's no room for a secret hidden twist.
The "One Special Twist" Case: The dance has exactly one complex, non-symmetric ingredient (a specific type of polynomial factor), and it appears in a very specific, odd-sized block.
- Analogy: Imagine a dance with one very complicated solo routine. As long as that routine is done in a specific, odd-numbered formation, the rules force it to be unique.
The "Balanced" Case: The dance has one complex ingredient AND one simple ingredient, but the simple ingredient must be "perfectly balanced."
- Analogy: Imagine a dance with a solo and a group formation. The group formation must be either a single person or a perfect "hyperbolic" pair (like two people mirroring each other perfectly). If the group is messy or unbalanced, a secret twist could hide inside it.

Why Does This Matter?

You might wonder, "Who cares if a math dance move has a secret twist?"

The paper mentions that this was originally motivated by knot theory (studying knotted strings). In knot theory, understanding when two shapes are truly the same (congruent) versus when they just look the same from a distance is crucial.

If you are trying to untie a knot, you need to know if two different-looking knots are actually the same knot just viewed from a different angle.
This paper provides a checklist. If a knot's "dance move" fits the checklist, you know for a fact that there is no hidden complexity. If it doesn't fit, you know there might be a "ghost" version of that knot that looks identical but is actually different.

The "Local Field" Twist

The paper specifically works in non-archimedean local fields.

Analogy: Think of this as a specific type of universe where the rules of distance are a bit weird (like a digital world where you can only measure in whole numbers or specific steps, rather than a smooth, continuous line).
The authors took a known rule that worked for "simple" dances (semi-simple isometries) and proved it works for all dances, even the messy, complicated ones, in this specific digital-like universe.

The Bottom Line

Xu and Zhang have solved a puzzle that mathematicians had been staring at for decades. They figured out exactly which mathematical "dance moves" are so rigid that they cannot hide any secrets.

If a move fits their three conditions, you can be 100% sure: If it looks the same as another move, it is the same move. No tricks, no hidden twists, no "ghost" versions. It's a complete map of when uniqueness is guaranteed in the world of geometric symmetries.

1. Problem Statement

The paper addresses a fundamental question in the classification of isometries within orthogonal groups over non-archimedean local fields (of characteristic $\neq 2$ ).

Context: Let $(V, \langle, \rangle)$ be a non-degenerate quadratic space over a field $k$ . Let $O(V)$ be the orthogonal group and $GL(V)$ the general linear group.
The Question: Determine all isometries $\sigma \in O(V)$ such that any isometry $\tau \in O(V)$ which is conjugate to $\sigma$ in $GL(V)$ is also conjugate to $\sigma$ in $O(V)$ .
Background: Milnor [3] previously proved that if $\sigma$ and $\tau$ have the same irreducible minimal polynomial, they are conjugate in $O(V)$ if they are conjugate in $GL(V)$. This paper seeks to extend this result to arbitrary isometries (which may have reducible characteristic polynomials and non-semi-simple structures) over non-archimedean local fields.

2. Methodology

The authors employ a combination of structural decomposition, hermitian form theory, and local field invariants to solve the problem.

A. Structural Decomposition

The characteristic polynomial $f(x)$ of an isometry $\sigma$ is decomposed into irreducible factors:
$f(x) = (x+1)^{m_+}(x-1)^{m_-} \prod_{i=1}^{m_1} p_i(x)^{e_i} \prod_{j=1}^{m_2} (\tilde{p}_j(x)\tilde{p}_j^*(x))^{\tilde{e}_j}$
where:

$p_i(x)$ are irreducible polynomials satisfying $p_i(x) = x^{\deg(p_i)}p_i(x^{-1})$ (reciprocal).
$\tilde{p}_j(x)$ are irreducible polynomials where $\tilde{p}_j(x) \neq \tilde{p}_j^*(x)$ (non-reciprocal).
The space $V$ decomposes orthogonally into subspaces $V_{q(x), \sigma}$ corresponding to these factors.

B. Classification via Hermitian Spaces

For the non-reciprocal case ( $\tilde{p}_j \neq \tilde{p}_j^*$ ), the authors utilize the theory of hermitian spaces over the algebra $E = k[x]/(\tilde{p}(x)\tilde{p}^*(x))$ .

They establish that the classification of conjugacy classes in this context reduces to the classification of hermitian forms over $E$ .
Key Lemma (2.6 & 2.7): They prove that for a non-reciprocal factor, the quadratic space structure induces a unique hermitian inner product, and the sequence of isomorphism classes of these hermitian spaces forms a complete invariant for the pair $(V, \sigma)$ .

C. Realization of Invariants

A critical technical step involves proving that any sequence of hermitian spaces (satisfying certain rank conditions) can be realized as the invariants of an actual quadratic space with an isometry.

Proposition 3.2 & 3.3: The authors construct explicit non-degenerate quadratic spaces and isometries corresponding to given hermitian forms (for non-reciprocal cases) and symmetric bilinear forms (for reciprocal cases $x \pm 1$ ). This ensures that the "missing" conjugacy classes in $O(V)$ can actually be constructed to demonstrate non-conjugacy.

D. Local Field Invariants

Since the field $k$ is a non-archimedean local field, the classification of quadratic and hermitian spaces relies on specific invariants:

Determinant: The class in $k^\times / (k^\times)^2$ .
Hasse Symbol: A local invariant distinguishing non-isomorphic forms of the same dimension and determinant.
The authors use the fact that over local fields, there exist non-isomorphic forms with the same determinant but different Hasse symbols (or different determinants) to construct counterexamples.

3. Key Contributions and Results

The main result is Theorem 1.2, which provides necessary and sufficient conditions for an isometry $\sigma$ to have a "single conjugacy class" (i.e., $GL(V)$-conjugacy implies $O(V)$ -conjugacy).

Let $m_0$ be the count of odd-dimensional blocks associated with eigenvalues $\pm 1$ :
$m_0 = \sharp\{2l+1 : V^{(2l+1)}_{x+1, \sigma} \neq 0\} + \sharp\{2l+1 : V^{(2l+1)}_{x-1, \sigma} \neq 0\}$
Let $m_1$ be the number of distinct reciprocal irreducible factors $p_i(x)$ in the characteristic polynomial.

Theorem 1.2: $\sigma$ satisfies the single conjugacy property if and only if one of the following holds:

Case (i): $m_0 \le 1$ $m_{0} \leq 1$ and $m_1 = 0$ $m_{1} = 0$ .
- Interpretation: The isometry has very few "odd" blocks for eigenvalues $\pm 1$ and no reciprocal irreducible factors.
Case (ii): $m_0 = 0$ $m_{0} = 0$ , $m_1 = 1$ $m_{1} = 1$ , and the component $V_{p_1(x), \sigma}$ $V_{p_{1} (x), σ}$ consists of a single odd power block $V^{(2l_1+1)}_{p_1(x), \sigma}$ $V_{p_{1} (x), σ}^{(2 l_{1} + 1)}$ .
- Interpretation: There is exactly one reciprocal factor, and it appears with an odd exponent in a single block.
Case (iii): $m_0 = m_1 = 1$ $m_{0} = m_{1} = 1$ , the reciprocal component is a single odd block, AND the unique non-zero quadratic space associated with the $\pm 1$ $\pm 1$ eigenvalues (defined via the bilinear form $b^{(2l_0+1)}_\sigma$ $b_{σ}^{(2 l_{0} + 1)}$ ) is either 1-dimensional or a hyperbolic plane.
- Interpretation: If there is a "problematic" odd block for $\pm 1$ , the geometry of that block must be trivial (1D) or hyperbolic to ensure uniqueness. If it is a higher-dimensional anisotropic space, distinct $O(V)$ classes exist within the same $GL(V)$ class.

4. Proof Strategy for Necessity

To prove the conditions are necessary, the authors construct counterexamples. If a condition fails (e.g., $m_1 \ge 1$ with even blocks, or $m_0 \ge 2$ ), they:

Identify the relevant hermitian or quadratic subspaces.
Use the properties of local fields to find a form with the same dimension and determinant but a different Hasse symbol (or different determinant).
Construct a new isometry $\tau$ using these modified forms.
Show that $\tau$ is conjugate to $\sigma$ in $GL(V)$ (same characteristic polynomial and module structure) but not in $O(V)$ (different invariants).

5. Significance

Generalization: This work significantly extends Milnor's 1970s result (which was limited to semi-simple isometries with irreducible minimal polynomials) to the general case of arbitrary isometries over non-archimedean local fields.
Completeness: It provides a complete classification of "single conjugacy classes," resolving Question 1.1 posed in the introduction.
Methodological Unity: The paper unifies the treatment of reciprocal and non-reciprocal factors by systematically applying hermitian form theory and module decomposition, offering a robust framework for future studies on isometries in algebraic groups.
Applications: The results are motivated by knot cobordism theory (referencing Milnor and Levine), suggesting potential applications in low-dimensional topology and the study of algebraic invariants of knots and links over local fields.

In summary, Xu and Zhang have determined exactly when the conjugacy class of an isometry in the general linear group "collapses" to a single class in the orthogonal group, revealing that this uniqueness is fragile and depends critically on the parity of the block sizes and the specific geometric nature (dimension and type) of the subspaces associated with eigenvalues $\pm 1$ .