Quantum LDPC Codes
Quantum low-density parity-check codes use sparse checks to encode many logical qubits per physical qubit, promising far lower overhead than the surface code.
Sparse checks, high rate
A quantum low-density parity-check (LDPC) code is a CSS code whose stabilizer checks each touch only a bounded number of qubits, and each qubit participates in only a bounded number of checks, regardless of how large the code grows. The surface code is technically an LDPC code, but the term usually refers to codes with a much better rate: many logical qubits and large distance for a given number of physical qubits, which the surface code cannot achieve.
Good codes exist
A landmark line of work culminating around 2021 to 2022 proved the existence of good quantum LDPC codes: families where both the encoding rate k/n and the relative distance d/n stay bounded away from zero as n grows, with checks of bounded weight. This means the number of logical qubits and the protection both scale linearly with the number of physical qubits, an asymptotic efficiency the surface code, with its d^2 qubits per logical qubit, fundamentally lacks.
- Bounded-weight checks; each qubit in only a few checks.
- High rate: many logical qubits per physical qubit.
- Good families keep both rate and relative distance constant as they grow.
- Potentially large reduction in fault-tolerance overhead.
The catch: connectivity and decoding
The price is nonlocal checks. Unlike the surface code's flat, nearest-neighbor layout, good LDPC codes require qubits that interact across the chip, which is hard to build in planar hardware and is a major engineering challenge. Decoding is also harder: syndromes are not simple string endpoints, so matching does not apply, and general decoders such as belief propagation with post-processing are used instead, with real-time speed still an open problem.
Despite these hurdles, quantum LDPC codes are among the most active research directions because they attack the dominant cost of fault tolerance, the overhead, at its asymptotic root, potentially reducing the physical qubits needed for a large computation by a substantial factor.