Skip to content
Quantum.world

The Signal

IBM says million-round repetition code can run in one circuit

IBM says its quantum platform can execute a million-round repetition-code experiment in a single circuit. The loop instruction and calibration used for the workload are experimental and have restricted support.

IBM says million-round repetition code can run in one circuit

A repetition-code experiment with 1,000,000 rounds can now run end to end in a single circuit on IBM Quantum Platform, IBM says. The circuit contains a million layers of mid-circuit measurement and reset, according to a company post dated November 11.

A repetition code is a simple form of error correction; IBM's documentation notes that it corrects bit-flip errors but is not a complete error-correction code. The company presented the long-running workload among the developer tools highlighted at its Quantum Developer Conference.

Repeated checks

In repetition-code experiments, researchers measure two-qubit parity checks across a chain of qubits over many rounds. Those results can then be decoded to study processor performance at low error rates over long computations, according to the company's calibration guide.

IBM says the ability to execute the workload end to end lets researchers develop and stress-test error-correction workflows on real hardware. The company points to the needs of long-duration error-corrected computations when logical error rates are very low.

An experimental loop

The experiment uses a for_loop instruction to repeat a block of operations a fixed number of times. The block can include measurements during the circuit and does not have to be unrolled into an enormous circuit, the company says.

The associated alternative calibration, qec_experimental, replaces parts of a backend's default calibration with differently tuned parameters. Both the loop instruction and the calibration are experimental and may change. The loop is currently supported only in Executor and client-side Sampler jobs.

The calibration might not be available on every backend and might change or be removed without notice, the calibration guide says. The documentation also specifies that the loop must have a countable, integer number of iterations and that measurements must use the default classified type.

The circuit example

A code example in the conference post illustrates the pattern with 100 rounds. IBM says the same circuit scales to the million-round workload. It sets up five data qubits alongside syndrome ancillas and registers for measurement results.

The example uses the experimental calibration and the MidCircuitMeasure instruction. Each round writes into the same measurement bits, so the calibration guide says the results must be copied into separate bits to prevent subsequent rounds from overwriting them.

The example then measures the data qubits for final readout. It prepares an Executor program with 10,000 shots and submits the circuit to a selected backend.

What the code tests

Separately, an IBM tutorial describes a bit-flip repetition code as a way to illustrate error correction on a quantum computer. It explains that stabilizer measurements can provide information about some errors without transforming the encoded quantum information.

That tutorial uses a different memory experiment: three data qubits encode a protected state, while two ancillas are used for stabilizer measurements. The tutorial describes comparisons with individual unencoded qubits and with a version that detects and corrects errors only after the final measurement.

Its memory experiment tests a single type of error and technically demonstrates protection of classical information, the tutorial says. It is presented separately from the million-round circuit example.

Sources

  1. IBM Quantum Computing Blog
  2. IBM Quantum Documentation
  3. IBM Quantum Documentation

Companies in this story

Related stories

Explore

More articles