⚛️ high-energy theory

Inequivalent modular tensor categories with identical $S$ , $T$ and $W$

This paper constructs two braided-inequivalent Dijkgraaf–Witten modular tensor categories for $(\mathbb{Z}/p\mathbb{Z})^3$ ( $p \ge 5$ ) that share identical modular data $(S, T)$ , the Whitehead-link matrix $W$ , and all Reshetikhin–Turaev invariants for framed oriented links with at most two components, thereby proving that these invariants are insufficient to determine a modular tensor category up to ribbon equivalence.

Original authors: Ran Luo, Jiahua Tian

Published 2026-09-29

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Ran Luo, Jiahua Tian

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

In the vast landscape of modern mathematics, there exists a field dedicated to understanding the hidden rules that govern how things can be combined, twisted, and transformed. Imagine a universe where objects do not just sit side by side, but interact in ways that depend on the order in which they are brought together, creating a rich tapestry of possibilities. This is the world of modular tensor categories, a mathematical framework that describes these interactions with extreme precision. These structures are not merely abstract games; they serve as the blueprint for certain types of quantum physics and the mathematics behind knot theory, helping scientists understand how particles might behave in exotic states of matter. For decades, mathematicians have tried to find a set of standard measurements that could uniquely identify every possible version of these categories. They knew that two specific sets of data, which describe how objects braid around one another and how they twist, were not enough to tell them apart. A third measurement, involving a specific three-dimensional knot known as a Whitehead link, was introduced to see if it would provide the missing piece of the puzzle. The question remained: if two different categories shared these three sets of data, would they necessarily be the same thing, just dressed differently?

A team of researchers has now answered this question with a definitive no. By constructing two distinct mathematical worlds that look identical under every standard test, they have proven that these three measurements are insufficient to distinguish between different types of modular tensor categories. The scientists built these examples using a specific type of symmetry group based on prime numbers, creating a family of categories that are fundamentally different from one another yet share the exact same numerical fingerprints. They demonstrated that for any prime number greater than or equal to five, one can create a pair of these categories that produce identical results for every possible knot made of one or two loops. In the language of the field, this means that if you were to probe these mathematical worlds with any knot that has only one or two strands, you would get the exact same answer, regardless of which category you were actually looking at.

The researchers achieved this by carefully tuning the internal rules of their categories. They kept the rules governing how objects interact in pairs and small groups constant, while subtly altering the rules that only appear when three objects interact simultaneously. This hidden alteration is invisible to any test involving just one or two strands, but it fundamentally changes the nature of the category. To prove that these categories are indeed different, the team had to look beyond simple knots. They constructed a specific, complex three-dimensional shape formed by linking three loops together in a way that is famously interdependent, known as the Borromean rings. When they calculated a specific value associated with this shape, the results for their two categories were different. This single value acted as a unique signature, revealing that the categories were not equivalent, despite their identical behavior in all simpler scenarios.

The significance of this finding is that it forces mathematicians to expand their toolkit. It shows that the standard measurements used to classify these structures, which include the data from one and two-component knots, are not enough to capture the full picture. To truly understand and categorize these mathematical objects, one must look at more complex interactions involving three or more components. The work provides a concrete example of two structures that are indistinguishable by their most common properties but are revealed to be distinct when examined through a more intricate lens. This discovery closes a long-standing open question in the field, confirming that the combined data of the S, T, and W matrices does not determine a modular tensor category up to equivalence. Instead, it reveals a deeper layer of complexity that requires new invariants to fully describe.

The construction of these counterexamples relied on a precise manipulation of mathematical functions called cocycles, which dictate how the objects in the category associate with one another. The researchers fixed the part of these functions that controls interactions between two objects, ensuring that any test involving one or two strands would yield the same result. They then varied the part of the function that governs the interaction of three objects, a change that remains hidden until one probes the system with a three-component link. By doing this, they created a family of categories where the difference between them is invisible to all standard two-strand probes but becomes starkly apparent when a three-strand probe is used. This approach allowed them to generate an infinite number of such pairs, one for each prime number greater than or equal to five, demonstrating that this phenomenon is not a rare anomaly but a widespread feature of these mathematical structures.

In the end, the paper establishes that the universe of modular tensor categories is richer and more subtle than previously thought. The ability to distinguish between these categories requires looking at the whole picture, including the complex ways in which three or more components can intertwine. The researchers have shown that relying solely on the data from simpler knots is like trying to identify a person by only looking at their shadow; while the shadow might look the same for two different people, the people themselves are distinct. By calculating a specific partition function on a closed three-dimensional manifold, the authors provided a concrete method to separate these categories, proving that the braided equivalence classes are determined by a parameter that can take on multiple values, specifically (p minus 1) divided by 2 distinct classes for a given prime p. This result resolves the question of whether the standard modular data is sufficient, confirming that it is not, and points the way toward a more complete understanding of these intricate mathematical worlds.

Technical Summary: Inequivalent Modular Tensor Categories with Identical S, T, and W

Problem Statement
The central question addressed in this work is whether the modular data $(S, T)$ , augmented by the Whitehead-link matrix $W$ , uniquely determine a modular tensor category (MTC) up to ribbon equivalence. Previous results established that $(S, T)$ alone are insufficient, with counterexamples arising from Drinfeld centers of groups $G = \mathbb{Z}_q \rtimes \mathbb{Z}_p$ . Subsequent work introduced the $W$ -matrix as a normalized invariant of colored framed oriented Whitehead links, demonstrating that it distinguishes certain examples where $(S, T)$ fail. However, it remained an open question whether the combined data $(S, T, W)$ suffices for a complete classification of MTCs up to ribbon equivalence.

Methodology and Construction
The authors construct a family of counterexamples using Dijkgraaf–Witten modular categories associated with the finite abelian group $V = (\mathbb{Z}/p\mathbb{Z})^3$ for primes $p \geq 5$ . The construction relies on the classification of 3-cocycles on $V$ . The cohomology group $H^3(V, U(1))$ decomposes into a symmetric part (quadratic) and an alternating part.

Cocycle Definition: The authors define a family of 3-cocycles $\omega_u = \kappa \cdot \tau_u$ parameterized by $u \in (\mathbb{Z}/p\mathbb{Z})^\times$ .
- $\kappa$ is a normalized 3-cocycle derived from a quadratic cohomological invariant, constructed using integer representatives and carry functions to ensure normalization.
- $\tau_u$ is an alternating 3-cocycle defined via the determinant: $\tau_u(a, b, c) = \zeta^{\frac{u}{6}\det(a,b,c)}$ , where $\zeta = e^{2\pi i/p}$ .
- The modular categories are defined as the Drinfeld centers $C_u = \mathcal{Z}(\text{Vec}_{V}^{\omega_u})$ , equipped with canonical ribbon structures.
Matching Invariants (Part A): To prove that distinct categories share identical invariants, the authors analyze the restriction of these categories to subgroups $P \subseteq V$ of rank at most 2.
- They demonstrate that the alternating factor $\tau_u$ vanishes when restricted to any subspace of dimension $\leq 2$ (since the determinant of three vectors in a 2D space is zero).
- Consequently, the restricted cocycles $\omega_u|_P$ and $\omega_v|_P$ coincide with the quadratic part $\kappa|_P$ .
- By establishing a bijection $\Lambda_{uv}$ between the simple objects of $C_u$ and $C_v$ that preserves fluxes and projective characters, they show that the Reshetikhin–Turaev (RT) invariants for any framed oriented link with at most two components are identical under this bijection.
- This equality extends to the matrices $S$ , $T$ , and $W$ , as these are derived from 1- and 2-component link evaluations.
Obstruction to Equivalence (Part B): To prove the categories are not ribbon equivalent, the authors analyze the structure of braided equivalences.
- Any braided equivalence $\Phi: C_u \to C_v$ must preserve the "electric subcategory" (the subcategory of invertible objects), which is isomorphic to $\text{Rep}(V)$ .
- This equivalence induces a monoidal equivalence between the underlying pointed categories $\text{Vec}_{V}^{\omega_u}$ and $\text{Vec}_{V}^{\omega_v}$ .
- Using cohomological invariants—specifically the quadratic invariant $q$ and the alternating invariant $\text{Alt}$ —the authors derive constraints on the automorphism $M \in GL_3(\mathbb{F}_p)$ induced by the equivalence.
- The quadratic invariant forces $\det(M) = \pm 1$ , while the alternating invariant forces $v \det(M) = u$ .
- These conditions imply that a braided equivalence exists if and only if $u = \pm v$ . Thus, for $u \neq \pm v$ , the categories are braided inequivalent (and consequently ribbon inequivalent).
Distinguishing via 3-Manifold Invariants: The authors further distinguish these categories using an uncolored partition function on a specific closed oriented 3-manifold $M_p$ obtained by surgery on the Borromean link with coefficient $+p$ .
- The calculation involves a finite sum over holonomies that explicitly depends on the alternating term $u \det(a, b, c)$ .
- The resulting partition function $Z_u(M_p)$ separates all $(p-1)/2$ equivalence classes, distinguishing $u$ from $v$ unless $u = \pm v$ .

Key Results

Main Theorem: For every prime $p \geq 5$ $p \geq 5$ , there exists a family of Dijkgraaf–Witten modular categories $\{C_u\}_{u \in \mathbb{F}_p^\times}$ ${C_{u}}_{u \in F_{p}^{\times}}$ associated with $(\mathbb{Z}/p\mathbb{Z})^3$ $(Z / p Z)^{3}$ such that:
1. There exists a bijection of simple objects under which the matrices $S$ , $T$ , and $W$ (and all RT invariants for links with $\leq 2$ components) are identical for any pair $C_u, C_v$ .
2. $C_u$ and $C_v$ are braided equivalent if and only if $u = \pm v$ .
Corollary: For $p=5$ , the categories $C_1$ and $C_2$ (twisted Drinfeld doubles of order 125) share identical $S, T, W$ data but are not ribbon equivalent.

Significance
The paper provides a negative answer to the question of whether $(S, T, W)$ determine an MTC up to ribbon equivalence. It demonstrates that the combined data of colored framed links with at most two components is insufficient to distinguish all modular tensor categories. The construction highlights a specific mechanism: fixing the quadratic cohomological component (which governs rank-2 restrictions and 1- and 2-component link invariants) while varying the alternating component (which is invisible to such probes but detectable by the full braided structure or 3-component invariants).

The work concludes that a complete classification of modular tensor categories up to ribbon equivalence requires invariants beyond those of colored framed links with at most two components. The authors provide an explicit analytic evaluation of an uncolored partition function on a 3-manifold that serves as such a distinguishing invariant for this family.

More like this