Decoding the surface code: the classical side of error correction

Short answer: Quantum error correction produces a stream of parity measurements called syndromes, and a classical decoder has to work out from them where errors occurred. Papers on this say decoding must run in real time, within about one microsecond per round on superconducting machines, to stop errors backing up. One 2025 paper reports a minimum-weight perfect matching decoder, sparse blossom, that meets that rate for distance-17 surface code circuits on a single core in simulation. These are research benchmarks, not a statement about any deployed system.

Why real-time decoding is necessary

Quantum devices suffer from high error rates, rendering them ineffective for practical applications without protection. Quantum Error Correction (QEC) protects quantum information by encoding logical qubits using data qubits and parity qubits. The data qubits collectively store the quantum information, while parity qubits are measured periodically to produce a syndrome. This syndrome is then decoded by a classical decoder to identify the location and type of errors.

To prevent errors from accumulating and causing a logical error, decoders must accurately identify errors in real-time. A 2023 paper says software decoders are slow, which it says necessitates hardware solutions. Ideally, a real-time decoder must match the performance of the Minimum-Weight Perfect Matching (MWPM) decoder. However, the complexity of the underlying Blossom algorithm has historically limited state-of-the-art real-time decoders. Previous approaches often relied on lookup tables, which the paper describes as not scalable, or approximate decoding, which significantly increases logical error rates. A 2024 paper puts the constraint this way: decoding must happen in real time, within syndrome data within the timeframe dictated by the physical hardware, about 1 microsecond on superconducting machines, to prevent a backlog of errors.

The role of the Minimum-Weight Perfect Matching decoder

The 2024 paper calls MWPM an accurate decoding algorithm for the surface code and is the most widely used decoder for several important families of quantum error correcting codes. It serves as the benchmark for accuracy in real-time decoding. The paper says recent research has demonstrated real-time implementations of MWPM (RT-MWPM) for a code distance of up to 9. However, beyond distance 9, the number of flipped parity bits in the syndrome, referred to as the Hamming weight of the syndrome, exceeds the capabilities of existing RT-MWPM decoders.

The 2024 paper proposes adaptive predecoding to address this. This technique converts high Hamming weight syndromes into low Hamming weight syndromes, which are then accurately decoded by the RT-MWPM decoder. An effective predecoder must balance both accuracy and speed, as any erroneous decoding by the predecoder contributes to the overall logical error rate. The authors say their goal is to enable larger-distance real-time MWPM decoders.

Algorithmic optimizations for speed

To achieve the necessary real-time performance, specific algorithmic variants have been developed. One such approach is "sparse blossom," a fast implementation of the MWPM decoder. Sparse blossom is, in the authors' words, a variant of the blossom algorithm that directly solves the decoding problem relevant to quantum error correction. It avoids the need for all-to-all Dijkstra searches, which are common amongst standard MWPM decoder implementations.

For 0.1% circuit-level depolarising noise, sparse blossom processes syndrome data in both X and Z bases of distance-17 surface code circuits in less than one microsecond per round of syndrome extraction on a single core. The authors say this matches the rate at which syndrome data is generated by superconducting quantum computers. The implementation is open-source and has been released in version 2 of the PyMatching library.

What this does not tell you

These are research results on decoding software and algorithms, reported by the papers' authors. We have not verified the benchmarks, and this article describes no deployed machine. Decoding at the speed of a real quantum computer is one requirement among many for fault tolerance.

FAQ

What does a decoder do?

It takes the syndrome, the periodically measured parity qubits, and works out the location and type of errors so they can be corrected.

Why must decoding be real time?

A 2024 paper says error decoding must happen within about 1 microsecond on superconducting machines, to prevent a backlog of errors.

What is MWPM?

Minimum-Weight Perfect Matching, an accurate decoding algorithm for the surface code and, according to the sparse blossom paper, the most widely used decoder for several important families of quantum error-correcting codes.

What did the sparse blossom paper report?

For 0.1% circuit-level depolarising noise, it processes syndrome data for distance-17 surface code circuits in under one microsecond per round on a single core, which the authors say matches the rate at which superconducting quantum computers generate syndrome data.

What is the limit for existing real-time MWPM decoders?

A 2024 paper says real-time implementations have been shown up to distance 9, and proposes adaptive predecoding to go beyond that.