How Quantum Error Correction Protects Fragile Information

quantum error correction

Qubits are fragile, and you may wonder how anything useful survives that. Quantum error correction spreads a piece of information across many physical qubits, so mistakes can be found and fixed without reading the data directly.

For the wider picture, start with our guide to quantum computing, then come back here for the error-correction details.

Item Value
Logical error per cycle, distance-7 code 0.143% ± 0.003
Qubits in that distance-7 code 101 qubits
Error suppression (Λ) per code-distance increase of 2 2.14 ± 0.02
Logical memory lifetime versus best physical qubit 2.4 ± 0.3 times longer
Average real-time decoder latency, distance 5 63 microseconds
Error-correction cycle time 1.1 microseconds

These figures come from the record of the Nature paper, published on Feb 27 2025. Results like these can change as new experiments appear, so check the source before you rely on them.

Why quantum information is so fragile

Google’s research team wrote in its December 9, 2024 post that even state-of-the-art quantum devices will typically experience at least one failure in every thousand operations. The same post says most applications need billions if not trillions of operations to run reliably.

Classical computers also lose bits now and then, and they cope by copying data so a failed bit is easy to spot. A Physics World report dated 20 Mar 2023 points out that quantum information cannot be copied, so quantum machines cannot be corrected the same way.

The same report says physicists believe an error rate of about one in a million is needed for practical large-scale machines. At that date, error-correction technology could only reach about one in a thousand.

How a logical qubit spreads information

Google’s February 22, 2023 post starts with a classical analogy. Bob wants to send Alice a single bit, so instead of one bit he sends three bits: “111”. If one bit flips, Alice takes a majority vote and still reads the intended message.

A quantum code has to go further, for two reasons given in that post:

  • It must correct bit flips, which take a qubit from 0 to 1, and also phase flips, an error unique to quantum states.
  • Checking a qubit’s state would destroy its superposition, so errors must be detected without measuring the states directly.

The surface code meets both demands. According to Google’s post on Willow, the encoded state is stored on an array of data qubits, while measure qubits check for errors on their neighboring data qubits.

The cycle is simple to describe. Data qubits start in a known state, repeated parity checks highlight errors, and at the end the data qubits are measured and the results are decoded into an error-corrected logical measurement.

quantum error correction Connected devices
A hand holds a smartphone displaying a teal interface in front of an open laptop.

What is the error-correction threshold?

A logical qubit is built by combining many physical qubits. The paper’s abstract says the logical error rate is then suppressed exponentially as more qubits are added, but only if the physical error rate is below a critical threshold.

The reason is a trade-off, explained in the Google Research post of December 2024. Making the lattice bigger also introduces more opportunities for error. The critical error rate that divides the two cases, below which error correction turns from harmful to helpful, is called the threshold.

Passing that line took a long time. The same post says quantum computers had not crossed it after almost 30 years of advances in fabrication, calibration and qubit design.

An earlier step came in 2023. Google reported that it had scaled a surface code from 17 to 49 physical qubits and shown, for the first time, that a bigger code lowered the logical error rate, as described in its 2023 announcement.

What the Willow experiments measured

The 2025 Nature paper reports surface code memories on a superconducting processor called Willow. Here is what the sources say, as of the Nature publication on Feb 27 2025:

  • Hardware: Willow is a 105-qubit processor, and average qubit lifetimes improved from about 20 microseconds to 68 microseconds, give or take 13, according to Google’s post.
  • Scaling: stepping the lattice from 3×3 to 5×5 to 7×7 suppressed the encoded error rate by a factor of two, or 2.14 more precisely, per the same post.
  • Beyond breakeven: the distance-7 memory outlasted its best physical qubit by a factor of 2.4 ± 0.3, says the paper record.
  • Speed: a real-time decoder averaged 63 microseconds at distance 5, up to a million cycles, with a cycle time of 1.1 microseconds, per the same record.

In its December 9, 2024 post, Google described Willow as the first quantum processor where error-corrected qubits get exponentially better as they get bigger. That is the team’s own claim about its device, so read it as a reported result, not a settled ranking.

What still stands in the way

The same paper ran repetition codes up to distance 29. It found that logical performance was limited by rare correlated error events, which struck roughly once every hour, as reported on the paper record.

The authors conclude that the device performance, if scaled, could realize the operational requirements of large-scale fault-tolerant quantum algorithms. The key word is “if”: a 101-qubit memory is a demonstration of one logical qubit, not a finished machine.

Error correction is the bridge between today’s noisy devices and the long computations behind ideas such as quantum advantage. Our guide to shor’s algorithm covers one famous example of a long computation that depends on that reliability.

Frequently asked questions

What is a logical qubit?

Show answer

A logical qubit combines multiple physical qubits so that information is encoded across them. In a surface code, each group is a square lattice of qubits that represents a single encoded qubit, according to Google’s December 2024 post.

Why can’t quantum computers just copy their data?

Show answer

Physics World’s 20 Mar 2023 report states that quantum information cannot be copied. That is why quantum codes use indirect checks on measure qubits instead of duplicates.

What does “below threshold” mean?

Show answer

It means the physical error rate is low enough that adding qubits makes the logical qubit better instead of worse. The Nature abstract says the exponential suppression only occurs below a critical threshold.

Does this mean practical quantum computers are here?

Show answer

Not yet. As of the Nature paper published on Feb 27 2025, the authors describe results that, if scaled, could meet the requirements of large-scale fault-tolerant algorithms. Scaling is still the open task.