Quantum Group Codes

• Quantum group codes are quantum error-correcting codes defined by finite groups that control parity checks, logical operators, and error sectors.
• They employ diverse constructions including group-algebra, coset-action, and representation-theoretic methods, yielding explicit examples like [[72,8,9]]_2 and tailored transversal gate operations.
• Practical insights include optimized decoding techniques and trade-offs between locality, symmetry, transversality, and decodability for enhanced quantum fault tolerance.

Searching arXiv for recent and foundational papers on “quantum group codes” and closely related constructions. Search query: "quantum group codes" Quantum group codes are quantum error-correcting codes whose defining structure is supplied by finite groups, group algebras, group actions, or unitary representations. In recent literature, the term is used for several distinct constructions: CSS quantum LDPC codes built from group-algebra blocks and coset actions, stabilizer codes obtained from left ideals in semisimple group algebras, representation-theoretic codes defined as isotypic components or symmetry-invariant subspaces, and group-valued-qudit generalizations of CSS and quantum double models. The same phrase is also used in one line of work for structured classical codes over classical and classical-quantum channels rather than for quantum error correction, so the terminology is not uniform (Lin et al., 2023, Aydin et al., 15 Jun 2026, Sales-Cabrera et al., 8 Dec 2025, Kubischta et al., 2024, McDonough et al., 23 Feb 2026, Pang et al., 2024).

1. Terminological scope and conceptual unification

Recent usage clusters around a small number of algebraic mechanisms. The common feature is that group structure controls either the parity checks, the logical operator algebra, or the error sectors detected by the code. What varies is the object on which the group acts: coordinates, group-algebra basis elements, coset spaces, representation spaces, or group-valued qudits.

Usage of “quantum group codes”	Core object	Representative papers
Quantum LDPC from group algebra or coset actions	commuting left/right actions, two-block CSS checks	(Lin et al., 2023, Aydin et al., 15 Jun 2026)
Stabilizer codes from group rings and left ideals	$\mathbb{F}_q[G]$, Wedderburn blocks, dual-containing ideals	(Sales-Cabrera et al., 8 Dec 2025, Sales-Cabrera et al., 2024, Abdukhalikov et al., 2024, 0711.3983)
Representation-theoretic or symmetry-based codes	isotypic projectors, twisted unitary $t$-groups, invariant sectors	(Kubischta et al., 2024, Bradshaw et al., 8 Dec 2025, Kubischta et al., 18 Nov 2025)
Group-valued CSS and quantum doubles	$G$-valued qudits, word constraints, $D(G)$ models	(McDonough et al., 23 Feb 2026, Manjunath et al., 5 Mar 2026)
Structured classical codes for cq channels	abelian group codes on channel inputs	(Pang et al., 2024)

A persistent misconception is to treat all of these as a single construction. The literature instead supports a broader classification: some quantum group codes are stabilizer LDPC codes, some are nonadditive codes defined by representation theory, and some are topological models on group-valued degrees of freedom. A related but distinct ambiguity is that “quantum group codes” may refer to codes used on classical-quantum channels, not to quantum error-correcting codes at all (Pang et al., 2024).

2. Group-algebra and group-ring constructions

One major branch begins with classical group codes, namely left ideals in a group algebra. In the dihedral and generalized quaternion setting, the group algebra is decomposed by Wedderburn–Artin theory into field and matrix blocks, and duality is computed blockwise. This yields systematic constructions of Hermitian self-orthogonal and dual-containing codes, then quantum stabilizer codes via Hermitian or Euclidean CSS. In this language, length is $|G|$, dimensions are rank sums over simple blocks, and explicit dual maps can be written for each $2\times 2$ matrix component (Sales-Cabrera et al., 8 Dec 2025, Sales-Cabrera et al., 2024).

A parallel group-ring formulation uses generator matrices $M=\sigma(a)$ obtained from Hurley’s embedding $\sigma:\mathbb{F}_q[G]\to M_n(\mathbb{F}_q)$. In that framework, Euclidean self-orthogonality is equivalent to $a a^T=0$, Hermitian self-orthogonality over $\mathbb{F}_{p^2}$ is equivalent to $t$0, and symplectic self-orthogonality can be expressed either as $t$1 or, for two generators, as $t$2. These criteria produce explicit quantum codes including $t$3, $t$4, $t$5, and $t$6 (Abdukhalikov et al., 2024).

Earlier group-ring work already emphasized nilpotent or symmetric elements $t$7 satisfying conditions such as $t$8, $t$9, and rank constraints. Those hypotheses produce self-dual or dual-containing classical families, and then quantum codes through standard dual-containing constructions. Reported quantum examples include $G$0, $G$1, and $G$2 (0711.3983).

Within quantum LDPC, the central group-algebra model is the two-block group algebra code. Choose $G$3, let $G$4 be the left regular action matrix of $G$5, and let $G$6 be the right regular action matrix of $G$7. Then

$G$8

and CSS commutation follows from $G$9. These 2BGA codes are “the smallest lifted-product codes,” generalize generalized-bicycle codes from cyclic to arbitrary finite groups, and include square-matrix hypergraph-product codes as special cases. Their length is $D(G)$0, and non-abelian realizations can have odd $D(G)$1, which gives a criterion for an essentially non-abelian code (Lin et al., 2023, Wang et al., 2023).

The same 2BGA literature also established concrete finite-length performance. Enumeration over connected binary 2BGA codes with stabilizer generator weight $D(G)$2, length $D(G)$3 for abelian groups, and $D(G)$4 for non-abelian groups produced examples such as $D(G)$5 and $D(G)$6 (Lin et al., 2023).

3. Coset-action quantum LDPC codes

A recent extension replaces the regular left/right actions of a quotient group by commuting coset actions. Let $D(G)$7 be finite, $D(G)$8, and $D(G)$9 with $|G|$0. The left action $|G|$1 is given by $|G|$2, while the right action $|G|$3 is given by $|G|$4. Their permutation matrices commute, and with

$|G|$5

one defines

$|G|$6

When $|G|$7, this reduces to a 2BGA code on the quotient $|G|$8; for non-normal $|G|$9, it produces genuinely new coset-action families outside the 2BGA class (Aydin et al., 15 Jun 2026).

This “breaking the bicycle frame” construction substantially enlarges the search space at fixed blocklength $2\times 2$0. A computer search found weight-6 codes $2\times 2$1, $2\times 2$2, and $2\times 2$3, and weight-8 codes $2\times 2$4, $2\times 2$5, $2\times 2$6, and $2\times 2$7. The paper also introduced a maximally packed interleaved $2\times 2$8 syndrome-extraction schedule of depth at most $2\times 2$9 per round, where $M=\sigma(a)$0 is the maximum stabilizer weight, and proved that the Tanner-graph thickness is exactly $M=\sigma(a)$1 for $M=\sigma(a)$2 and exactly $M=\sigma(a)$3 for $M=\sigma(a)$4. Under a standard circuit-level noise model and BP-OSD decoding, the weight-6 family achieved thresholds around $M=\sigma(a)$5, and the weight-8 family around $M=\sigma(a)$6, competitive with BB codes (Aydin et al., 15 Jun 2026).

The same work developed a group-theoretic cover formalism. Given a base 2BGA code $M=\sigma(a)$7, a sequence of groups $M=\sigma(a)$8 with normal subgroups $M=\sigma(a)$9 and quotient isomorphisms $\sigma:\mathbb{F}_q[G]\to M_n(\mathbb{F}_q)$0 produces Tanner graphs that are $\sigma:\mathbb{F}_q[G]\to M_n(\mathbb{F}_q)$1-fold covers of the base graph, where $\sigma:\mathbb{F}_q[G]\to M_n(\mathbb{F}_q)$2. This recovers and extends voltage-graph and BB-cover constructions to non-abelian groups (Aydin et al., 15 Jun 2026).

4. Representation-theoretic, symmetry-resolved, and intrinsic codes

A second major lineage treats codes as subspaces selected by representation theory. In a Fourier-transform approach on group algebras of subgroups of an error group, primitive central idempotents and more general transform-domain idempotents generate stabilizer codes, Clifford codes, direct sums of Clifford translates, and broader nonadditive constructions. The projector onto an irrep $\sigma:\mathbb{F}_q[G]\to M_n(\mathbb{F}_q)$3 of a subgroup $\sigma:\mathbb{F}_q[G]\to M_n(\mathbb{F}_q)$4 has the familiar form

$\sigma:\mathbb{F}_q[G]\to M_n(\mathbb{F}_q)$5

which specializes to stabilizer projectors for abelian $\sigma:\mathbb{F}_q[G]\to M_n(\mathbb{F}_q)$6 (Kumar et al., 2012).

Twisted unitary $\sigma:\mathbb{F}_q[G]\to M_n(\mathbb{F}_q)$7-groups yield a more explicitly error-correcting representation-theoretic family. If $\sigma:\mathbb{F}_q[G]\to M_n(\mathbb{F}_q)$8 is a $\sigma:\mathbb{F}_q[G]\to M_n(\mathbb{F}_q)$9-twisted unitary $a a^T=0$0-group, then any $a a^T=0$1-subspace of $a a^T=0$2 is a quantum code with distance $a a^T=0$3, and every transversal operator $a a^T=0$4 acts logically as $a a^T=0$5. This produces nonadditive codes with large finite transversal gate groups. Concrete examples include distance-3 qubit codes from the binary icosahedral group $a a^T=0$6 for odd $a a^T=0$7, and distance-2 qutrit codes from $a a^T=0$8 (Kubischta et al., 2024). For $a a^T=0$9, the criterion becomes purely character-theoretic: $\mathbb{F}_{p^2}$0 so code-finding reduces to irreducible products of characters; this perspective yields many distance-2 free codes with unusual transversal groups (Kubischta et al., 2024).

A broader symmetry-based framework defines the codespace as the $\mathbb{F}_{p^2}$1-symmetric subspace

$\mathbb{F}_{p^2}$2

More generally, one resolves the full isotypic decomposition using

$\mathbb{F}_{p^2}$3

In this language, stabilizer codes and qudit stabilizer codes appear as abelian special cases, while nonabelian examples include a one-logical-qubit code associated to the dihedral group (Bradshaw et al., 8 Dec 2025). A related Lie-algebraic construction organizes Pauli spinors by Cartanions and bi-subalgebra partitions inside $\mathbb{F}_{p^2}$4, producing additive and nonadditive codes from coset-labeled syndrome subspaces (Tsai et al., 2013).

The intrinsic-code program pushes the same idea further. An intrinsic quantum code is a subspace $\mathbb{F}_{p^2}$5 of a unitary group representation $\mathbb{F}_{p^2}$6, with error-detection properties defined directly sector-by-sector in $\mathbb{F}_{p^2}$7. The Schur–Bootstrap theorem states that any $\mathbb{F}_{p^2}$8-equivariant embedding $\mathbb{F}_{p^2}$9 transports those sectorwise Knill–Laflamme relations to the physical realization $t$00. One intrinsic code therefore controls an entire family of realizations. The paper’s central example is an $t$01 irrep-$t$02, $t$03, $t$04 code whose realizations include a four-qubit code, a single spin-$t$05 code, a two-qutrit code, a bosonic two-mode code, rigid-rotor codes, molecular rotational codes, and Landau-level codes (Kubischta et al., 18 Nov 2025).

5. Group-valued qudits, quantum doubles, and group surface codes

A third branch generalizes CSS itself from abelian qudits to $t$06-valued qudits. For a finite group $t$07, a single qudit has basis $t$08, left/right multiplication operators

$t$09

and $t$10-checks given by projectors onto solutions of group word equations. These group-CSS codes reduce to qubit CSS codes when $t$11, but for general $t$12 they generalize Kitaev’s quantum double model. On oriented two-dimensional CW complexes, the Hamiltonian takes the standard commuting-projector form $t$13. For non-Abelian simple $t$14, any $t$15-covariant group-CSS code with suitably upper-bounded $t$16-check weight and lower-bounded $t$17-distance reduces to a CW quantum double. In this setting the codespace with ghost vertices $t$18 is

$t$19

the $t$20-distance is the girth of $t$21, and there are intrinsically non-Abelian families with $t$22, $t$23, and $t$24 (McDonough et al., 23 Feb 2026).

Group surface codes specialize this picture to planar patches with rough left/right and smooth top/bottom boundaries, and are equivalent to quantum double models $t$25 with specific boundary conditions. A planar $t$26-surface code has code dimension $t$27 and a logical basis $t$28. Boundary-supported operators

$t$29

implement left and right logical multiplication, while global automorphism gates

$t$30

are transversal. For suitably chosen groups, arbitrary reversible classical gates can be implemented transversally in the group surface code. In the $t$31 example, one obtains a logical $t$32 and an outer automorphism acting as $t$33; for $t$34, transversal gates appear deep in the Clifford hierarchy (Manjunath et al., 5 Mar 2026).

Code switching is implemented by extension and splitting maps for knit products $t$35. In semidirect-product cases, sliding one patch through another realizes controlled conjugation. This unifies recent sliding constructions and magic-state-preparation protocols, and the spacetime tensor-network description makes explicit contact with topological gauge theories. The paper argues that these models bypass Bravyi–König restrictions because they are non-Pauli topological models and because key operations use code switching, boundaries, and domain walls rather than locality-preserving logical gates of a fixed two-dimensional Pauli stabilizer code (Manjunath et al., 5 Mar 2026).

6. Transversal non-Clifford logic, decoding, and algorithmic criteria

A recent quasi-group-code framework isolates a family of quantum CSS codes explicitly designed for non-Clifford logic. Starting from classical quasi-$t$36 codes and a compatible $t$37-module section, one obtains quantum group codes with transversal multi-control-$t$38 gates that are addressable and parallelizable. A lifting procedure from classical AG codes via class field theory then produces good quantum group codes over $t$39 with improved decoding complexity and logical gate scheduling. Two asymptotic families are emphasized: regular codes with parameters $t$40, supporting fully addressable transversal $t$41, and free codes with parameters $t$42, supporting a transversal global $t$43 and orbit-wise addressable $t$44. The decoder runs in quasi-quadratic time with a linear decoding radius, in contrast to cubic-time decoders for previous quantum AG codes, and this implies an almost linear reduction in the time complexity of state-of-the-art magic-state distillation protocols (Gasnier et al., 25 Jun 2026).

Tensor-product group-algebra constructions give a complementary route to constant-depth entangling gates. The copy-cup formalism determines when balanced or hypergraph-product-type quantum codes admit constant-depth $t$45 and $t$46 gates. Up to check weight $t$47, the pre-orientation conditions reduce to a perfect matching problem in graph theory. The conditions are fully determined for 2- and 3-copy-cup gates, including odd check weights. The paper shows that bivariate bicycle codes do not have the pre-orientation for either copy-cup gate, that abelian weight-4 group algebra codes satisfying the non-associative 3-copy-cup conditions necessarily have distance $t$48, and that the symmetric 3-copy-cup conditions are compatible with higher distances and automatically imply the 2-copy-cup conditions (Tiew et al., 26 Feb 2026).

The same practical emphasis appears in quantum Margulis codes, where 2BGA codes are reformulated on a left-right Cayley complex and instantiated on $t$49. These codes have length $t$50, check degree $t$51, and girth growing as $t$52. Under depolarizing noise and BP-OSD-10 decoding, the family with $t$53 and $t$54 exhibited a threshold around $t$55, with examples $t$56, $t$57, and $t$58 at blocklengths $t$59, $t$60, and $t$61 (Pacenti et al., 2024).

Across these strands, the main trade-off is now explicit. Lower-weight or more constrained constructions generally improve thresholds and schedule depth, while richer group actions or higher-weight tensor-product structures typically enlarge the logical gate set, rate, or addressability. This suggests that “quantum group codes” are best understood not as a single family, but as an algebraic design space in which group structure mediates between locality, symmetry, transversality, and decodability.