Quantum error correction

Quantum error correction is the set of techniques that protect fragile quantum information by encoding one reliable Logical qubit across many noisy physical qubits, continuously detecting and reversing errors faster than they accumulate. It is the essential ingredient that would turn today's error-prone machines into fault-tolerant computers capable of long computations.

#Why physical qubits are noisy

Every physical qubit interacts weakly with its environment, causing Decoherence, and every gate and measurement is slightly imperfect. The best physical qubits as of early 2026 err roughly once per 1000 operations. A cryptanalytic run of Shor's algorithm would involve billions of operations, so an uncorrected machine would fail almost immediately. Classical computers solve reliability with redundancy, but quantum information cannot be copied (the no-cloning theorem) and cannot be measured without disturbance, so correction must be indirect.

#How it works

The trick, introduced in 1995, is to spread one qubit's information across an entangled block of several qubits so that errors can be detected without measuring the encoded data itself (Shor 1995). Auxiliary qubits measure carefully chosen parities, called syndromes, that reveal whether an error occurred and which type it was, while leaving the logical state intact. Because any single-qubit error can be decomposed into bit-flip and phase-flip components, correcting both is sufficient to correct arbitrary small errors. A decoder then infers and reverses the most likely error from the syndrome pattern.

#The threshold theorem

The foundational result is the threshold theorem: if the physical error rate per operation is below a critical value, then adding more layers of encoding suppresses the logical error rate arbitrarily, at a manageable overhead (Aharonov and Ben-Or 1997). Above the threshold, adding qubits makes things worse; below it, scaling up wins. This is why demonstrating operation below threshold is the field's central scientific milestone.

#Surface codes

The leading practical code is the Surface code, which arranges physical qubits on a two-dimensional grid and requires only local, nearest-neighbor measurements, matching the connectivity of real chips (Fowler et al. 2012). Its threshold is relatively forgiving, near 1 percent, which is why it dominates roadmaps. The cost is overhead: reaching cryptographically relevant logical error rates takes on the order of 1000 physical qubits per logical qubit, so headline physical-qubit counts overstate usable capability by roughly three orders of magnitude.

#The 2024 below-threshold result

In December 2024 Google reported surface-code memories at code distances 3, 5, and 7 in which each increase in distance roughly halved the logical error rate, and the encoded qubit outlived the best individual physical qubit (Google Quantum AI 2024). This below-threshold scaling, long a theoretical target, is the property fault tolerance depends on, shown for quantum memory at small scale rather than for full computation.

#Status and relevance

Error correction is advancing but remains far from the scale a cryptographically relevant quantum computer needs, which would require thousands of logical qubits sustained for days. The remaining path, from tens of logical qubits to thousands, is an unsolved, decade-scale engineering program. Its uncertain pace is a major reason the timing of Q-Day cannot be predicted, and why post-quantum cryptography is being deployed now rather than waiting for a machine to appear.