Subsystem Codes with Constrained Check Weight

• The paper establishes tight bounds on code parameters and provides explicit constructions for subsystem codes with low check weights, balancing error correction, rate, and distance.
• Subsystem codes are defined by splitting logical qubits into protected and gauge parts, using bounded weight gauge operators to ensure efficient error detection and hardware compatibility.
• Decoding strategies such as analytical methods, MWPM, and neural-network decoders offer practical solutions for error recovery in constrained-weight quantum codes.

Subsystem codes with constrained check weight constitute a central theme in quantum error correction, targeting practical implementations where measurement of large-weight operators is infeasible due to hardware limitations. These codes utilize the subsystem framework to encode quantum information while ensuring all gauge (check) operators have bounded weight. Recent developments have established tight bounds on achievable code parameters under check-weight constraints, constructed explicit families of subsystem codes with asymptotically good rate and distance, and explored decoding strategies compatible with the subsystem structure. This article comprehensively reviews the structures, limitation theorems, construction methods, decoding algorithms, and the practical trade-offs characterizing subsystem codes under check-weight constraints.

1. Formal Definitions and Algebraic Framework

Subsystem codes extend stabilizer codes by splitting logical degrees of freedom into protected (logical qubits) and unprotected (gauge qubits), with error correction performed via measurement of gauge operators ("checks"). For CSS-type subsystem codes on $n$ physical qubits, two classical codes $Q_X,Q_Z\subseteq\mathbb{F}_q^n$ specify the code, generating gauge groups via $X$-type and $Z$-type operators corresponding to vectors in $Q_X^\perp$ and $Q_Z^\perp$ respectively. The stabilizer subgroup consists of commuting gauge operators, and logical operators are cosets built from $Q_X, Q_Z$ and their duals (Golowich et al., 8 Oct 2025). A code is said to have check-weight constraint $w$ if every generator of the gauge group acts nontrivially on at most $w$ qubits. Codes with constant or slowly growing $w$ are termed quantum LDPC subsystem codes.

2. Bounds on Distance and Rate under Check-Weight Constraints

Sharp limitations follow from the restriction to low-weight checks. For weight-two CSS subsystem codes, the fundamental trade-off is

$Q_X,Q_Z\subseteq\mathbb{F}_q^n$0

where $Q_X,Q_Z\subseteq\mathbb{F}_q^n$1 is the minimum weight of any dressed logical operator and $Q_X,Q_Z\subseteq\mathbb{F}_q^n$2 the number of logical qubits (Wang et al., 21 Jan 2026). This is proven via matrix characterizations: in a subsystem code specified by an $Q_X,Q_Z\subseteq\mathbb{F}_q^n$3 binary matrix $Q_X,Q_Z\subseteq\mathbb{F}_q^n$4, the X- and Z-distances correspond to column and row distances of $Q_X,Q_Z\subseteq\mathbb{F}_q^n$5, with the product bounded by the number of occupied entries. Extensions to check weights $Q_X,Q_Z\subseteq\mathbb{F}_q^n$6 systematically enlarge the achievable region (higher $Q_X,Q_Z\subseteq\mathbb{F}_q^n$7 for fixed $Q_X,Q_Z\subseteq\mathbb{F}_q^n$8). These bounds hold without assuming geometric locality, applying equally to generic qLDPC subsystem codes.

3. Explicit Code Families and Constructions

3.1 Weight-2 Codes: Bacon–Shor and Generalized Matrix Codes

Bacon–Shor codes on an $Q_X,Q_Z\subseteq\mathbb{F}_q^n$9 grid use $X$0 and $X$1 checks of weight two along rows and columns, achieving $X$2, $X$3, $X$4, saturating the $X$5 and $X$6 bounds (Bravyi, 2010, Wang et al., 21 Jan 2026). More generally, matrix constructions assign qubits to the nonzero entries of a binary matrix $X$7, using row and column pairs for gauge checks, with code parameters $X$8 at weight $X$9.

3.2 Codes with Weight-three and Four Checks

Subsystem surface codes utilize three-qubit gauge checks ($Z$0 or $Z$1) on triangles of the lattice, leading to codes with topological order and encoding logical qubits via punctures or holes while maintaining locality and efficient matching or RG decoders (1207.1443). Concatenated many-hypercube codes ("subsystem $Z$2" codes) achieve constant check weight 4 irrespective of the concatenation level, physical qubits $Z$3, logical qubits $Z$4, and distance $Z$5 (Nakai et al., 6 Oct 2025). These utilize gauge checks acting on lines in each hypercube direction.

3.3 High-distance, Sublinear-weight Families via Product Codes

Recent constructions provide quantum codes with linear rate and distance and sublinear check weight by taking tensor products of Reed–Solomon codes ("homological product codes") (Golowich et al., 8 Oct 2025). For two-product codes on $Z$6 qudits, the check weight scales as $Z$7, with code parameters $Z$8 and transversal $Z$9 gates. Three-product codes lower weight further to $Q_X^\perp$0 at cost of larger alphabet size and slightly lower distance. Alphabet reduction via code concatenation achieves $Q_X^\perp$1 with polylogarithmic blowup in other parameters.

4. Decoding Algorithms for Subsystem LDPC Codes

Subsystem codes admit specialized decoding procedures exploiting gauge structure. For product code families, decoding is performed via a multivariate generalization of Prony's method, which reconstructs functions from partial access to their Fourier transform. Classical syndrome measurement yields partial evaluations, and quantum decoding proceeds by measuring the relevant gauge syndromes followed by classical recovery and Pauli corrections (Golowich et al., 8 Oct 2025). For many-hypercube codes, block-MAP decoding (dynamic programming over the error decomposition) achieves optimal logical error rates for moderate code sizes, while neural-network-based decoders are essential for larger instances or high gauge qubit count (Nakai et al., 6 Oct 2025). Subsystem surface codes employ minimum-weight matching or renormalization group decoding on syndrome histories, with efficient algorithms and thresholds approaching 1% (1207.1443).

5. Trade-offs, Hardware Implementation, and Practical Implications

Imposing low check-weight is essential for noisy, limited-connectivity hardware architectures (ion traps, atom arrays, photonic circuits). Weight-2 and weight-3 subsystem codes offer direct hardware compatibility but are limited to $Q_X^\perp$2, necessitating higher-weight ($Q_X^\perp$3 or $Q_X^\perp$4) checks for scalable, high-distance codes. Subsystem codes decouple the syndrome extraction circuit from stabilizer weight, mitigate error propagation (one error per ancilla fault), and permit single-shot fault tolerance under local testability (Golowich et al., 8 Oct 2025). Constrained-weight subsystem codes exhibit clear performance trade-offs compared to stabilizer codes: constant check weight generally demands more gauge qubits and lowers the effective error threshold, but enables hardware simplicity and parallel syndrome acquisition (Nakai et al., 6 Oct 2025). Recent work delineates the achievable region for $Q_X^\perp$5 via linear programming, demonstrating near-optimality of explicit construction up to boundary effects (Wang et al., 21 Jan 2026).

6. Open Problems and Directions

The threshold on check-weight $Q_X^\perp$6 beyond which asymptotically good (distance $Q_X^\perp$7) subsystem codes exist remains precisely undetermined. Theoretical and numerical analyses for small $Q_X^\perp$8 are tight, but for larger weights systematic constructions with optimal threshold and code parameters, especially in the qubit setting ($Q_X^\perp$9), are under active investigation. Extensions to algebraic and geometric code families utilizing subsystem product structures are promising. The interplay between subsystem code gauge group design, fault-tolerant logical gate support (transversal $Q_Z^\perp$0 gates), and hardware-efficient decoding warrants further study (Golowich et al., 8 Oct 2025).

7. Summary Table: Core Subsystem Code Families Under Check-Weight Constraints

Code Family	Gauge Check Weight $Q_Z^\perp$1	Rate ($Q_Z^\perp$2)	Distance $Q_Z^\perp$3	Decoders
Bacon–Shor/matrix	2	$Q_Z^\perp$4	$Q_Z^\perp$5	Analytical, MWPM
Subsystem surface	3	$Q_Z^\perp$6	$Q_Z^\perp$7	MWPM, RG
Many-hypercube	4	$Q_Z^\perp$8	$Q_Z^\perp$9	Block-MAP, NN
Homological product	$Q_X, Q_Z$0	$Q_X, Q_Z$1	$Q_X, Q_Z$2	Prony, classical

Subsystem codes with constrained check weight form a foundational component of practical quantum error correction, balancing code rate, distance, fault-tolerance, and hardware compatibility. Analytical bounds, explicit constructions, and advanced decoding schemes jointly define the landscape of quantum LDPC subsystem codes, with ongoing research refining their optimality under physical constraints.