Modular transformations and topological orders in two dimensions

Fangzhou Liu, Zhenghan Wang, Yi‐Zhuang You, Xiao-Gang WenarXiv 2013 · DOI · arXiv

Abstract

The string-net approach by Levin and Wen and the local unitary transformation approach by Chen, Gu and Wen provided ways to systematically label non-chiral topological orders in 2D. In those approaches, different topologically ordered many-body wave functions were characterized by different fixed-point tensors. Though extremely powerful, the resulting fixed-point tensors were mathematical abstractions and thus lacked a physical interpretation. As a result it was hard to judge if two different fixed-point tensors described the same quantum phase or not. We want to improve that approach by giving a more physical description of the topological orders. We find that the non-Abelian Berry's phases, $T$- and $S$-matrices, of the topological protected degenerate ground states on a torus give rise to a more physical description of topological orders. Using the Verlinde conjecture, we can even choose the canonical basis for the $T$- and $S$-matrices. It is conjectured that the $T$ and $S$-matrices form a complete and one-to-one characterization of non-chiral topological orders and can replace the fixed-point tensor description to give us a more physical label for topological orders. As a result, all the topological properties can be obtained from the $T$- and $S$-matrices, such as number of quasiparticle types (from the dimension of $T$ or $S$), the quasiparticle statistics (from the diagonal elements of $T$), the quantum dimensions of quasiparticles (from the first row of $S$), etc.

Introduction

The classification of phases of matter has traditionally relied on Landau’s symmetry-breaking theory. However, the discovery of the Fractional Quantum Hall effect introduced "topological orders"—phases defined by long-range entanglement rather than symmetry. While mathematical frameworks like the Levin-Wen string-net approach have successfully used "fixed-point tensors" to classify these states, these tensors remain abstract. A significant challenge is their "many-to-one" nature: multiple distinct tensors can describe the same physical phase, making it difficult to identify equivalent models. This research proposes a more physical labeling system using $T$ and $S$ matrices derived from modular transformations on a torus.

Methodology: Modular Transformations and the Canonical Basis

To bridge the gap between abstract tensors and physical observables, the authors analyze topologically ordered states on a torus. They focus on two geometric operations: the $T$-transformation (Dehn twist), which involves twisting the torus by 360° along one axis, and the $S$-transformation, which rotates it by 90°.

As these transformations are applied to the degenerate ground states, the system picks up non-Abelian Berry phases, forming the $T$ and $S$ matrices. A key technical contribution is the use of the Verlinde Formula to define a canonical basis, ensuring these matrices provide a unique, one-to-one characterization of the phase.

The derivation of the T-matrix elements through a Dehn twist followed by F-moves and deformations of the string-net wavefunction.

Mathematically, calculating these matrices from fixed-point tensors is shown to be equivalent to finding the "Drinfeld center" of a unitary fusion category. This allows for a rigorous connection between the abstract tensor category description and measurable physical properties.

Results: A Physical Label for Topological Orders

The authors demonstrate that the entire physics of a system's excitations is encoded in the ground state's response to geometric twists. By applying this method to a variety of models—including $\mathbb{Z}_2$ gauge theories (Toric Code), Fibonacci states, and $S_3$ non-Abelian states—they show that $T$ and $S$ matrices can uniquely identify the phase.

Example of canonical basis states for the ground state manifold on a torus, represented as vectors of fixed-point wavefunctions.

From these matrices, one can directly extract critical topological data:

  • Quasiparticle Types: Determined by the dimension of the $T$ or $S$ matrix.
  • Braiding Statistics: Derived from the diagonal elements of the $T$ matrix.
  • Quantum Dimensions: Obtained from the first row of the $S$ matrix.

This approach successfully recovers known modular matrices for standard models, proving its validity and its ability to handle complex string-net states with clean, manageable matrix representations.

Conclusion and Future Directions

The proposed $T$ and $S$ matrix characterization offers a robust, physical alternative to abstract tensors for labeling non-chiral topological orders. Future work aims to extend this framework to chiral states, which currently lack a non-trivial fixed-point description. Additionally, the method could be applied to "messy" numerical wavefunctions from simulations to identify their topological order and to study phase transitions by observing how these matrices change at critical points.

Summary updated

Abstract

The string-net approach by Levin and Wen and the local unitary transformation approach by Chen, Gu and Wen provided ways to systematically label non-chiral topological orders in 2D. In those approaches, different topologically ordered many-body wave functions were characterized by different fixed-point tensors. Though extremely powerful, the resulting fixed-point tensors were mathematical abstractions and thus lacked a physical interpretation. As a result it was hard to judge if two different fixed-point tensors described the same quantum phase or not. We want to improve that approach by giving a more physical description of the topological orders. We find that the non-Abelian Berry's phases, $T$- and $S$-matrices, of the topological protected degenerate ground states on a torus give rise to a more physical description of topological orders. Using the Verlinde conjecture, we can even choose the canonical basis for the $T$- and $S$-matrices. It is conjectured that the $T$ and $S$-matrices form a complete and one-to-one characterization of non-chiral topological orders and can replace the fixed-point tensor description to give us a more physical label for topological orders. As a result, all the topological properties can be obtained from the $T$- and $S$-matrices, such as number of quasiparticle types (from the dimension of $T$ or $S$), the quasiparticle statistics (from the diagonal elements of $T$), the quantum dimensions of quasiparticles (from the first row of $S$), \etc.
BibTeX
@misc{liu2013modular,
  author = {Fangzhou Liu and Zhenghan Wang and Yi‐Zhuang You and Xiao-Gang Wen},
  title = {{Modular transformations and topological orders in two dimensions}},
  year = {2013},
  eprint = {1303.0829},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1303.0829},
}
Figure