# Techniques: The domain wall color code

## What the paper contributes (one paragraph)
The paper introduces the domain wall (DW) color code, a Clifford-deformed family of 2D colour codes tailored for biased Pauli noise. Applying Hadamard rotations to half of the qubits creates domain walls that permute anyon types on crossing, confining anyon propagation to quasi-1D domains under strong bias. The dense X3Z3 instance attains a 50% code-capacity threshold at infinite bias (decoding reduces to repetition codes) and matches the noise-tailored XZZX surface code threshold at every finite bias, while remaining more resource-efficient and retaining transversal Clifford gates. For practical decoding the authors adapt an approximate maximum-likelihood tensor-network decoder for optimal thresholds and a scalable matching-based restriction decoder whose thresholds rise monotonically with bias (12.6% depolarising to ~41.7% at high bias). They also show the XZZX surface and XYZ colour codes are instances of the DW framework, and give a materialised-symmetry proof of the 50% threshold.

## Techniques offered
- **Domain wall colour code (biased-noise-tailored quantum error-correcting code)** - a Clifford-deformed colour-code family that introduces domain walls confining anyons to 1D domains, giving high biased-noise thresholds with transversal Clifford gates and low overhead.
  - guarantee: 50% code-capacity threshold at infinite bias (rigorously proven), thresholds matching the XZZX surface code at all finite bias, and sub-threshold logical failure scaling exp(-d) with open boundaries and exp(-d^2) on co-prime periodic lattices.
  - quote: "We introduce the domain wall color code, a new variant of the quantum error-correcting color code that exhibits exceptionally high code-capacity error thresholds for qubits subject to biased noise."
- **Matching-based restriction decoder** - a scalable MWPM-based restriction decoder adapted to the X3Z3 code that splits the colour-code syndrome onto restricted (surface-code-like) graphs and exploits noise bias in the edge weights.
  - guarantee: efficient decoding with thresholds that monotonically increase with bias (12.6% at depolarising noise, up to 41.7% at eta=30000, and 50% at infinite bias), below the optimal maximum-likelihood thresholds.
  - quote: "we adopted a matching-based restriction decoder of Ref.~\cite{PhysRevA.89.012317,PhysRevA.89.012317} to the \xz{} code."
- **Approximate maximum-likelihood tensor-network decoder** - an adapted approximate maximum-likelihood (tensor-network) decoder used to compute the optimal code-capacity thresholds of the DW codes under biased noise.
  - guarantee: near-optimal (maximum-likelihood) threshold estimates whose accuracy improves with bond dimension chi, at run-time exponential in chi.
  - quote: "For decoding, we adapt the approximate \emph{maximum-likelihood decoder}~\cite{PhysRevX.9.041031,PhysRevA.90.032326}"
- **Unified domain wall framework for noise-tailored topological codes** - a construction that reformulates existing noise-tailored codes (XZZX surface, XYZ colour) as instances of DW codes parameterised by domain-wall density and orientation.
  - guarantee: none (unifying reformulation; no new bound certified beyond placing prior codes in one framework).
  - quote: "Lastly, we show that various noise-tailored topological QECC studied so far---such as the XZZX surface~\cite{XZZX} and the XYZ color~\cite{miguel2022cellular} codes---can be formulated as instances of DW codes."

## Where it could apply
- Target primitive(s): none (the codes protect logical qubits and are decoder-agnostic to the encoded computation rather than tied to VQE/QAOA/QPE/sampling).
- Target application group(s): error_correction; platforms_tooling.
- Code family / hardware assumptions: 2D topological colour codes (6.6.6 hexagonal and 4.8.8 lattices) and surface codes via the DW framework; biased single-qubit Pauli noise (dephasing-dominated), as in superconducting qubits, trapped ions, quantum dots, and cat qubits.

## Caveats
Results are code-capacity only: stabiliser measurements are assumed perfect, so no fault-tolerant (circuit-level) thresholds are established, and optimised scalable decoders are left to future work. The restriction-decoder thresholds sit noticeably below the optimal maximum-likelihood values. Lattice surgery, magic state distillation, and code switching are noted only as compatible standard techniques, not contributed here.