Source-linked AI summary
Non-Abelian sheaf quantum LDPC codes: good and magical
Zimu Li, Fuchuan Wei, Zhengyi Han, Zi-Wen Liu
TL;DR
Non-Abelian qLDPC codes raise questions about achievable encoding rate and distance, and about extending sheaf-code constructions beyond the Pauli stabilizer setting. This work gauges sheaf codes using cup products and establishes good parameters, long-range magic in an almost-good family, and encoded magic-state preparation.
Problem
Existing gauging constructions leave rigorous parameter guarantees unresolved, while gauging couples logical cohomology classes and makes code dimension and distance require explicit analysis.
Source: §1 · p. 2Method
The paper gauges sheaf codes via cup-product constraints on logical cohomology classes and proves error protection using the Knill–Laflamme condition with expansion and cleaning.
Source: §1 · pp. 2–3Results
For every integer r ≥2, non-Abelian qLDPC families attain [[N, Θ(N), Θ(N)]] parameters, and an almost-good family has long-range magic in every family of code states.
Source: §1 · p. 3Takeaways & Limitations
The framework provides an algebraic basis for studying non-Abelian phases beyond geometric locality and the connections between non-Abelian structure, error correction, and quantum magic.
Source: §5 · p. 41Assessing practical advantages requires efficient decoders and fault-tolerant measurement circuits, as well as quantifying their resource costs.
Source: §5 · p. 41
Abstract
from arXiv · showhide
Non-Abelian quantum codes connect quantum error correction, phases of matter, and computational resources. In this work, we develop a general framework for constructing non-Abelian quantum low-density parity-check (qLDPC) codes by gauging sheaf codes via cup products and use it to obtain families with constant encoding rate and linear distance. We resolve the coupled logical constraints using explicit representatives to characterize the full gauged code space. We provide a fundamental treatment of code distance based on the general Knill--Laflamme condition and combine expansion with cleaning to establish protection against arbitrary low-weight errors. We further construct an almost-good family whose entire code space exhibits long-range magic. Gauging and ungauging also enable logical Clifford measurements that prepare encoded magic states. These results extend good qLDPC codes beyond the Pauli stabilizer setting and provide a concrete foundation for exploring non-Abelian phases beyond geometric locality and pursuing the no low-energy trivial magic conjecture.
1 Introduction
The introduction frames open questions about non-Abelian qLDPC parameters and gauging, then presents a construction with constant encoding rate and linear distance. It also develops distance and magic results, including a protocol for encoded magic-state preparation.
1 Introduction
- Non-Abelian codes extend the stabilizer setting and connect non-Abelian phases with magic and non-Clifford computation.Prior work cited in the introduction describes these codes as a route to studying non-Abelian quantum phases on general connectivity [19] [20] [21].Source: §1 · p. 2
- Prior gauging constructions leave rigorous parameter guarantees unresolved, in part because gauging couples logical cohomology classes and changes the code dimension.The familiar minimum-weight X- and Z-type logical-representative characterization of distance also does not apply directly.Source: §1 · p. 2
- Good non-Abelian qLDPC codes are expected to contribute to the NLTM conjecture, which concerns low-energy states inaccessible from arbitrary stabilizer states by shallow circuits.The paper studies whether magic can persist throughout a low-energy sector.Source: §1 · p. 2
- Gauging good sheaf codes via cup products yields constant encoding rate and linear distance, with coupled logical constraints and protection across the resulting code space addressed explicitly.The distance framework uses the general Knill–Laflamme condition, expansion, and cleaning to handle arbitrary low-weight errors.Source: §1 · pp. 2–3
- For every integer r ≥2, the paper gives non-Abelian qLDPC families with parameters [[N, Θ(N), Θ(N)]] and stabilizer generators formed from Pauli and C_{r−1}Z operators.This is the stated informal construction theorem.Source: §1 · p. 3
- The construction characterizes gauged code states by cup-product constraints on logical cohomology classes and an explicit orthonormal basis.The code dimension is the number of tuples of logical cohomology classes satisfying those constraints; polarized representatives are used to estimate it [27].Source: §1 · p. 3
- An almost-good family with constant encoding rate and almost-good distance has long-range magic in every family of code states.The introduction presents this as a dimension-obstruction result [23].Source: §1 · p. 3
- Gauging and ungauging measurements prepare encoded magic states by measuring logical Clifford operators on three code blocks with a dual sheaf construction.The approach also extends to logical multi-controlled-Z measurements using good qLTCs with transversal multi-controlled-Z gates.Source: §1 · p. 3
2 Preliminaries
This section introduces HGP and sheaf codes as CSS-code building blocks, along with cubical complexes, local coefficients, and sheaved coboundaries used to realize them. It also defines cup and cap products and sets out finite-field qudit operators and the Knill–Laflamme and cleaning tools used later.
2.1 HGP codes and sheaf codes
- Codes over Fq of characteristic 2 group s physical qubits at each site, a necessary condition for product-expanding classical codes used in nearly optimal sheaf-code constructions.Source: §2.1 · p. 3
- CSS logical operators are represented by cohomology classes for X type and homology classes for Z type, with code dimension equal to the dimension of either space.Source: §2.1 · p. 4
- Sheaf codes extract three consecutive terms of a cochain complex on a cell complex, with local coefficient data generalizing scalar coefficients and sparse incidence ensuring LDPC checks.Local codes provide a concrete way to realize the sheaf coefficients.Source: §2.1 · pp. 5–6
2.2 Cup and cap products on cubical complexes
- A graph lift by a finite abelian group yields a 1D double cover, and its t-dimensional product construction supplies cubical complexes whose squares are basic building blocks.The cubical complexes support the cup and cap products central to the non-Abelian-code construction.Source: §2.2 · pp. 6–8
- Local generator or parity-check matrices of classical codes provide a concrete implementation of the sheaf coefficients on the resulting complex.Source: §2.1 · p. 6, §2.2 · p. 8
- A one-dimensional sheaved coboundary operator on a graph defines a classical Tanner code.Source: §2.2 · p. 9
- Cup and cap products combine cubes in specified ways, vanish on disjoint cubes, and the cup product depends on factor order.The products are bilinear; cup products are associative and obey a Leibniz rule, while cap products satisfy a discrete Stokes identity.Source: §2.2 · pp. 9–10
- For sheaf codes, cup products combine the supporting cubes first and then combine local coefficients by tensor products.When the sheaves come from local codes, their tensor product is determined by the Khatri–Rao product of the local generator or parity-check matrices.Source: §2.2 · p. 9
2.3 Quantum gates over Fq and the Knill–Laflamme condition
- In Fq, Pauli operators are indexed by field elements, so an Fq-basis of a code alone does not generate all its X- or Z-type Pauli operators.The passage emphasizes that the corresponding operators must be included when describing the Fq Pauli operators.Source: §2.3 · pp. 10–11
- The finite-field gate preliminaries define CZ using the trace to Fp, extend it to multi-controlled-Z gates including CCZ, and define CNOT by addition in Fq.Source: §2.3 · pp. 11–12
- Theorem 2.9 (Knill–Laflamme condition) considers Q ⊆ (Cq)⊗n of distance d and integer w ≥ 0; d ≥ 2w + 1 is one equivalent condition, and Pw contains all Pauli strings of weight at most w.Source: §2.3 · p. 12
- For an Fq-linear CSS code, any region M with |M| < d can be cleaned from X- and Z-type logical representatives [49, 50].The resulting representatives have supports disjoint from M.Source: §2.3 · p. 12
3 Gauging HGP codes with scalar coefficients
This section develops gauging for HGP codes with scalar coefficients as a warm-up to the general sheaf-code construction, recovering gauged toric codes and quantum doubles while characterizing the resulting code space and parameters. It also counts gauged tuples of cohomological classes in LG to bound gauged-code dimension and establish good coding rate for gauged sheaf codes.
3 Gauging HGP codes with scalar coefficients
- Scalar-coefficient gauging recovers the well-known gauged toric codes and quantum doubles [20].The section presents this construction as a warm-up before the general sheaf-code results.Source: §3 · p. 12
3.1 Parameters of gauged HGP
- In the 2D construction, cup-product dressing yields a non-Abelian stabilizer collection containing both Pauli and Clifford operators.The described local cup products dress X stabilizers, while the Z stabilizers are undressed.Source: §3.1 · pp. 13–14
- The non-Abelian code space is spanned by orthonormal logical states and has dimension equal to the number of gauged cohomological classes, |LG|.The theorem applies to gauged tuples of logical representatives and vectors from the three blocks’ 0-cochain spaces.Source: §3.1 · pp. 14–15
- The construction extends to t-dimensional HGP codes using CtZ gates, with a distance bound depending on HGP distance and 0-cocycle expansion.The bound’s constant depends only on node degree n and cochain-complex dimension t.Source: §3.1 · p. 15
3.2 Gauged toric codes and quantum double
- Canonical logical representatives support counting gauged tuples, and a simple counting argument indicates that general gauged HGP code dimension scales like that of the original CSS code.The section notes that a closed-form dimension formula for general hypergraph products is intricate.Source: §3.2 · pp. 15, 18
- The gauged code dimension is at least q^{Θ(k)}, based on gauged tuples of cohomological classes in LG.Source: §3.2 · p. 18
- The same argument extends to gauged sheaf codes to achieve good coding rate.Source: §3.2 · p. 18
4 Gauging sheaf codes
This section develops gauging sheaf codes to construct non-Abelian qLDPC codes, characterize their code spaces, and bound their distance using expansion and cleaning. It also constructs an almost-good family with long-range magic throughout its code spaces and describes gauging-based measurements for generating logical magic states.
4 Gauging sheaf codes
- The construction needs two-way product-expanding local codes; punctured Reed–Solomon codes provide a nontrivial Khatri–Rao kernel [31], and projective Reed–Solomon codes support good qLTC constructions [28] [29] [30].The text notes that whether the required kernel is nontrivial for random choices remains unclear.Source: §4 · p. 19
4.1 Gauged code space and dimension
- The gauged stabilizers have bounded weight, and each physical qudit participates in a constant number of cup products, yielding a non-Abelian LDPC code.The bounds hold when n, m, m′, and t are constants and q is constant.Source: §4.1 · p. 20
- Up to phase, explicit gauged representatives form an orthonormal basis of the full non-Abelian code space, whose dimension equals |LG|.The result characterizes all code states, not only a subset of states satisfying the stabilizers.Source: §4.1 · pp. 23–24
- The gauged code dimension is at least q^{cN} for a constant c ∈ (0, 1), using polarized logical representatives whose same-direction cup products vanish.For the last code block there are Θ(N) representatives in one direction, while each of the first t blocks has Ω(1) in that direction.Source: §4.1 · p. 25
- The dimension result extends to arithmetic cubical complexes with projective Reed–Solomon local codes used for good qLTCs [28] [29] [30].For these constructions, the code dimension is |LG|, and a zero-factor argument gives the same exponential scaling without an abelian-lift decomposition.Source: §4.1 · p. 25
4.2 Gauged code distance
- Expansion and cleaning establish a Knill–Laflamme distance lower bound for gauged sheaf codes by making the region requiring cleaning smaller than d.The argument verifies the condition for errors within half the stated bounds; the resulting distance bound is given in terms of the expansion parameters, dimension, node degree, and local coefficient dimensions.Source: §4.2 · pp. 26, 29
- The gauged codes have distance Θ(N) for left-right Cayley complexes and good qLTCs, and Θ(N/polylog N) for high-dimensional sheaf codes based on abelian lifts.These bounds are stated for the respective families in Corollary 4.6.Source: §4.2 · p. 29
4.3 Long-range magic throughout the gauged code space
- An abelian group action on gauged cohomological classes has fixed points in one-to-one correspondence with gauged cohomological classes of the associated HGP codes.The construction uses covering spaces and reduces the dimension-counting task to three constant-size HGP codes.Source: §4.3 · pp. 30–31
- The construction yields an asymptotic family with constant coding rate, almost-good distance, and long-range magic in every code state.The dimension criterion is met by proving that the code dimension is divisible by 3.Source: §4.3 · pp. 29, 31
- For connected G0 and its double cover, translation invariance makes the local coefficient vectors of ζ identical across 0-cubes; Vandermonde column independence makes the specified nonzero expansions nontrivial cocycles.The nontrivial-cocycle conclusion applies when the expansion indices are chosen as rA, sA = m + 1, . . . , n.Source: §4.3 · pp. 32–33
- Induction from a constant-size HGP instance with gauged code dimension divisible by 3 yields an almost-good non-Abelian qLDPC family whose code spaces are filled with long-range magic.The starting gauged dimension is not a power of q.Source: §4.3 · p. 37
- Attaining good distance requires a non-Abelian lift, but the method's starting identity assumes abelian lifts and does not hold for non-Abelian lifts.The authors leave resolving this incompatibility for future work.Source: §4.3 · p. 37
4.4 Gauging measurements and magic state generation
- The standard gauging/ungauging process yields a joint magic state using three sheaf code blocks over a two-dimensional expander.The blocks use the constant scalar-coefficient sheaf and a sheaf generated by product-expanding punctured Reed–Solomon codes.Source: §4.4 · p. 37
- The protocol uses constant-depth CNOT and physical CCZ circuits, measures ancillas, then ungauges with U† and Z-basis measurements to generate magic states in the original code blocks.With good qLTCs and transversal logical multi-controlled-Z of polynomial subrank, the gauging measurement can be implemented in that setting.Source: §4.4 · pp. 38, 39, 41
- For the constant sheaf, gauging measurement projects the two code blocks into a joint logical magic state because the all-ones vector is the unique kernel basis vector.The projection is onto eigenstates of CZ on the two blocks.Source: §4.4 · p. 40
- A single logical CZ projection on |++⟩ yields a magic state with probability 3/4.The stated criterion is that pure stabilizer states have computational-basis support of size 2^k, with all nonzero amplitudes equal in magnitude.Source: §4.4 · p. 41
- For two initialized qubit pairs, the two normalized outputs have 10 and 6 computational-basis terms, occur with probabilities 5/8 and 3/8, and are both magic and entangled across the pairs.The results generalize to more qubit pairs and generic qudits with q = 2s.Source: §4.4 · p. 41
5 Discussion
The discussion frames non-Abelian sheaf codes as a bridge between error correction, quantum magic, and non-Abelian phases, while identifying open steps toward broader physical and computational applications.
5 Discussion
- Cup-product gauging yields non-Abelian qLDPC codes with constant encoding rate and linear distance, with distance bounds developed through Knill–Laflamme, expansion, and cleaning.The framework connects non-Abelian structure, error correction, and quantum magic through the algebraic properties of the underlying codes.Source: §5 · p. 41
- The results provide an algebraic foundation for studying non-Abelian phases beyond geometric locality, while extending long-range magic to wider low-energy sectors requires additional control such as soundness or local testability.Open questions include achieving full-code-space long-range magic with strictly linear distance and establishing robustness under approximate preparation and noise.Source: §5 · p. 41
- Assessing practical advantages for quantum computation requires efficient decoders, fault-tolerant measurement circuits, and quantification of resource costs.Further work could investigate magic-state generation schemes in fault-tolerant protocols and tailor codes to experimental platforms such as neutral atom arrays.Source: §5 · p. 41