arXiv:2610.03677
Exact Recovery for Non-Abelian Surface Codes
Alison Warman, Nathanan Tantivasadakarn, Sakura Schafer-Nameki
Abstract
We study exact recovery for non-Abelian topological surface codes based on the quantum double $D(G)$ of any finite group $G$. We determine an orthogonal error basis, that is comprised of group multiplication and irreducible representation operators. Whenever the errors are supported on closed loops on the lattice or dual lattice, there is a redundancy in the basis, due to stabilizers. To remedy this, we introduce a gauge-fixing that results in a complete and orthogonal error basis on the code space. Assuming as input predetermined, neutral correctable error clusters, we construct charge and flux transfer circuits that move the errors onto ancillas that then get projected out. This is an exact and deterministic recovery protocol, applicable to surface codes for any finite, in particular non-Abelian, group $G$.
AI Summary
The paper addresses explicit recovery for Kitaev quantum-double surface codes $D(G)$ for arbitrary finite groups, including non-Abelian and non-solvable groups. It assumes that neutral, correctable error clusters have already been identified and constructs recovery circuits for errors within those supplied supports. The construction starts from a single-edge orthogonal operator basis $E^{h,\mu}=L^h\sqrt{d_\Gamma}Z^\Gamma_{ab}$, combining group multiplication with irreducible-representation matrix elements. Vertex and plaquette constraints create redundancies on multi-edge supports. A gauge-fixing procedure removes multiplication labels and representation labels on selected edges. For a completed, bulk support of $N$ edges, the resulting family has $|G|^{2N-|\vec i_A|-|\vec i_B|}$ elements. The paper proves completeness and an identity Knill–Laflamme matrix for supports carrying no logical operator, using stabilizer identities, Schur orthogonality, and checks incident on a single remaining error. Recovery first transfers multiplication errors to plaquette ancillas through controlled-holonomy circuits, then transfers representation errors to vertex ancillas through controlled vertex operations and group multiplication. Each free label is transferred to one $|G|$-dimensional ancilla, restoring the encoded state exactly. Related constructions treat ribbon errors using two ancillas per ribbon. The protocol assumes perfect operations and syndrome measurements. Identifying neutral clusters, integrating recovery with a decoder, establishing thresholds for particular noise models, and handling measurement errors remain open. Recovery proceeds sequentially inward through clusters, so circuit depth can grow with cluster size; no size-independent-depth recovery or complete fault-tolerant decoding scheme is established.
Error found: The displayed gauge-fixed family and Knill–Laflamme theorem do not consistently implement the boundary-label reduction that the paper itself identifies, and retaining the displayed G-irrep labels gives nonorthogonal errors on rough boundary edges. For a proper-subgroup boundary K, the proposed unrestricted vertex gauge fixing and full-G charge-transfer derivation are not justified by the available K-only stabilizer invariance.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 22 pages