🔢 mathematics

Cappell-Shaneson knot pairs with the same Alexander polynomial

This paper constructs an infinite family of new Cappell-Shaneson knot pairs and provides specific examples of such pairs that share the same Alexander polynomial yet remain topologically inequivalent.

Original authors: Hisaaki Endo, Kazunori Iwaki, Andrei Pajitnov

Published 2026-04-02

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Hisaaki Endo, Kazunori Iwaki, Andrei Pajitnov

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 detective trying to solve a mystery about shapes in a very high-dimensional world. This paper is about a specific type of puzzle involving "knots," but not the kind you tie in your shoelaces. These are mathematical knots made by looping a sphere inside a larger sphere.

Here is the story of what the authors, Endo, IwakI, and Pajitnov, discovered, explained in plain English.

1. The Mystery: The "Ghost" Twins

In the world of knots, there is a famous rule: usually, if you take a knot out of the space around it (its "exterior"), the shape of that empty space tells you exactly what the knot looks like. It's like looking at the shadow of a statue; usually, the shadow is unique enough that you can tell exactly what the statue is.

However, mathematicians found a weird exception. Sometimes, you can have two different knots that cast the exact same shadow.

The Knots: They are distinct; you can't turn one into the other without cutting it.
The Shadow (Exterior): If you look at the empty space surrounding them, the spaces are identical. They are "ghost twins."

For a long time, mathematicians knew these twins existed, but they were rare and hard to find. The big question was: How many of these twin pairs are there, and can we find new ones?

2. The Recipe: The Cappell-Shaneson Machine

Back in 1976, two mathematicians named Cappell and Shaneson built a "machine" to create these ghost twins.

The Input: You feed the machine a specific type of number grid (a matrix).
The Output: The machine spits out a pair of knots.
The Magic: No matter how you tweak the machine, the two knots it produces will always have the same "shadow" (diffeomorphic exterior), but they will be different knots.

The authors of this paper decided to run this machine with a new set of ingredients to see what happens.

3. The New Discovery: The "Fingerprint" Problem

For decades, mathematicians thought they could tell these twin knots apart using a specific mathematical tool called the Alexander Polynomial. Think of the Alexander Polynomial as a fingerprint.

The old rule was: "If two knots have the same shadow, they might be twins, but if they have different fingerprints, they are definitely different."
The authors found a loophole. They discovered pairs of knots that:
1. Have the same shadow (they are Cappell-Shaneson twins).
2. Have the exact same fingerprint (the same Alexander Polynomial).

The Analogy: Imagine you have two identical twins. They have the same face (shadow). For a long time, we thought their DNA (fingerprint) would be different enough to tell them apart. But these authors found twins who have the same face and the same DNA. They are indistinguishable by the tools we used to use!

4. The Scale: Thousands of Twins

The authors didn't just find one or two of these super-twins. They found thousands.

They proved there are over 10,000 different pairs of knots that look the same, have the same shadow, and have the same fingerprint.
They created an infinite family of these knots, meaning you can keep generating new ones forever.

5. How They Did It: The "Ideal Class" Map

To find these twins, the authors had to translate the knot problem into a pure math problem involving numbers and rings (a branch of algebra).

They realized that every knot pair corresponds to a specific "class" of numbers.
They used a tool called an Ideal Class Monoid (imagine a giant filing cabinet of number groups) to sort these knots.
By organizing the files in this cabinet, they could see exactly where the "duplicates" were hiding. They found that for certain number patterns, the filing cabinet had multiple entries that looked identical on the outside but were actually different inside.

6. Why This Matters

This paper changes how we understand the universe of knots.

Before: We thought the "fingerprint" (Alexander Polynomial) was a strong enough tool to tell most knots apart.
Now: We know that for these specific high-dimensional knots, the fingerprint is useless. You need a much more powerful tool (the "Ideal Class Monoid") to tell them apart.

Summary

Think of the authors as cartographers exploring a new continent. They found a valley where thousands of identical-looking houses (knots) stand. For a long time, people thought you could tell the houses apart by their address (the Alexander Polynomial). But these authors proved that in this valley, the addresses are fake! They built a new map (the algebraic classification) that finally allows us to see the true differences between these "ghost twins."

They didn't just find a few; they found a whole city of them, proving that the world of mathematical knots is far stranger and more crowded with look-alikes than we ever imagined.

1. Problem Statement

The paper addresses a classical problem in high-dimensional knot theory: the existence of non-reflexive knot pairs.

Context: An $m$ -knot is a smooth embedding of $S^m$ into $S^{m+2}$ . Two knots are equivalent if there is a self-diffeomorphism of the ambient sphere mapping one to the other.
The Question: Can two inequivalent knots have diffeomorphic (or homeomorphic) complements?
Reflexivity: A knot is reflexive if it is uniquely determined by its complement. If a knot is not reflexive, it forms a "non-reflexive knot pair" with another knot sharing the same complement.
Known Results: For $m=1$ (classical knots), the Gordon-Luecke theorem states knots are reflexive. For $m \ge 2$ , it is known that at most two inequivalent knots can share a diffeomorphic complement. Cappell and Shaneson (1976) provided a construction for such pairs in dimensions $m=3, 4$ and later generalized the method.
The Gap: While Cappell-Shaneson (CS) pairs were known, it was an open question whether distinct CS pairs could share the same Alexander polynomial. Previously, all known CS pairs were distinguished by their Alexander polynomials. This paper seeks to construct infinite families of CS pairs that are inequivalent yet share the exact same Alexander polynomial.

2. Methodology

The authors employ a combination of geometric topology and algebraic number theory to classify and construct these knots.

A. Geometric Construction (Cappell-Shaneson Method)

The construction relies on mapping tori of torus automorphisms:

Matrix Selection: Start with a matrix $A \in SL(n, \mathbb{Z})$ satisfying specific conditions (Definition 2.1), known as a Cappell-Shaneson matrix. These conditions ensure that the resulting surgery yields a homotopy sphere.
Knot Formation: The matrix $A$ induces a map $f_A: T^n \to T^n$ . The mapping torus $M_A$ is fibered over $S^1$ . Performing surgery on the zero-section circle $C \subset M_A$ yields a homotopy sphere $\Sigma^{n+1}$ .
Framing Ambiguity: There are exactly two equivalence classes of framings for the circle $C$ (related to $\pi_1(SO(n))$ ). These two framings produce two distinct $(n-1)$ -knots, denoted $K_0(A)$ and $K_1(A)$ .
The Pair: The unordered pair $\{K_0(A), K_1(A)\}$ is the Cappell-Shaneson knot pair $P(A)$ .
Invariants: The Alexander polynomial of both knots in the pair is the characteristic polynomial of $A$ .

B. Algebraic Classification

The core of the paper is the classification of these knot pairs via matrix equivalence.

$\ast$ -Equivalence: Two matrices $A$ and $B$ define equivalent knot pairs if and only if they are $\ast$ -equivalent (Definition 2.4). This means $A$ is conjugate to $B$ or $B^{-1}$ in $GL(n, \mathbb{Z})$ .
Reduction to Ideal Classes: The authors reduce the classification of $\ast$ $*$ -equivalence classes of matrices to the classification of ideal classes in the ring $\mathbb{Z}[\theta]$ $Z [θ]$ , where $\theta$ $θ$ is a root of the characteristic polynomial $f(x)$ $f (x)$ .
- Using the Latimer-MacDuffee theorem, there is a bijection between conjugacy classes of matrices with characteristic polynomial $f$ and the ideal class monoid $C(\mathbb{Z}[\theta])$ .
- The authors extend this to $\ast$ -equivalence by considering the signed reciprocal polynomial $f^*(x) = (-x)^n f(x^{-1})$ .
Complete Invariant: The ideal class monoid $C(\mathbb{Z}[\theta])$ serves as a complete invariant for the equivalence classes of Cappell-Shaneson knot pairs. If the class group has more than one element, multiple non-equivalent knot pairs exist for the same Alexander polynomial.

C. Computational Approach

The authors use SageMath and number-theoretic criteria (Kummer-Dedekind) to:

Identify specific families of Cappell-Shaneson polynomials for degrees $n=4, 5, 6, 7$ .
Compute the order of the ideal class groups for these polynomials.
Explicitly construct matrices corresponding to non-trivial ideal classes (non-invertible ideals or ideals not equivalent to the identity) to generate the "new" knot pairs.

3. Key Contributions

Existence of Indistinguishable Pairs: The paper proves the existence of inequivalent Cappell-Shaneson knot pairs that share the same Alexander polynomial. This resolves a significant gap in the classification of non-reflexive knots.
Quantification of Non-Uniqueness: The authors show that for certain polynomials, the number of non-equivalent pairs is large.
- Example: They identify more than 10,000 Cappell-Shaneson knot pairs with the Alexander polynomial $x^5 - 99x^4 + 197x^3 - 197x^2 + 100x - 1$ .
New Infinite Families: They construct infinite families of such pairs for dimensions $n=4, 5, 6, 7$ $n = 4, 5, 6, 7$ .
- For $n=4$ , they provide a family parameterized by $l \le 0$ where the Alexander polynomial is fixed, but the knot pairs are distinct.
- Similar infinite families are constructed for $n=5, 6, 7$ .
Algebraic Invariant: They establish the ideal class monoid of $\mathbb{Z}[\theta]$ as a complete algebraic invariant for classifying Cappell-Shaneson knot pairs, providing a concrete method to enumerate all such pairs for a given Alexander polynomial.
Explicit Matrix Construction: The paper provides explicit matrix formulas (e.g., Theorems 4.1–4.4) for matrices that generate these new knot pairs, distinguishing them from the standard companion matrices.

4. Key Results

Theorem 2.6: Establishes that Cappell-Shaneson knot pairs $P(A)$ and $P(B)$ are equivalent if and only if $A$ and $B$ are $\ast$ -equivalent.
Theorem 3.8: Establishes the bijection between $\ast$ -equivalence classes of matrices (with characteristic polynomial $f$ or $f^*$ ) and the ideal class monoid $C(\mathbb{Z}[\theta])$ .
Theorem 4.1 (n=4): For any $l \le 0$ , two specific matrices (one companion, one derived from a non-invertible ideal) generate inequivalent knot pairs with the same Alexander polynomial.
Theorems 4.2, 4.3, 4.4: Generalize the result to $n=5, 6, 7$ , providing explicit matrix families for each dimension that yield non-equivalent pairs with identical Alexander polynomials.
Computational Finding: For the polynomial $x^4 - 8x^3 + 14x^2 - 9x + 1$ , the ideal class group has order 2, yielding exactly two inequivalent pairs. For $x^4 - 25x^3 + \dots$ , the group is $C_4 \times C_2$ , yielding 8 pairs.

5. Significance

Refinement of Knot Invariants: The work demonstrates that the Alexander polynomial is insufficient to classify Cappell-Shaneson knots in higher dimensions. It forces the use of more sophisticated invariants (like the ideal class monoid) to distinguish knots.
Resolution of Open Questions: It answers the question of whether CS pairs with identical Alexander polynomials exist, confirming they do and are abundant.
Algorithmic Classification: By linking knot theory to the arithmetic of number fields (ideal class groups), the paper provides a computational framework to enumerate all possible non-reflexive knot pairs for a given dimension and polynomial.
New Examples: The construction of over 10,000 examples for a single polynomial highlights the richness and complexity of the space of high-dimensional knots, suggesting that "generic" Alexander polynomials may correspond to vast families of distinct knots.

In summary, this paper bridges geometric topology and algebraic number theory to show that the landscape of non-reflexive knots is far more complex than previously thought, with the Alexander polynomial failing to distinguish between thousands of distinct knot pairs in certain dimensions.