Several recent studies suggest quantum error correction faces a tougher problem than expected: non-Markovian errors, where qubits retain memory of past interactions, may be undermining standard assumptions behind fault tolerance.
A Johns Hopkins-led analysis found a noise model including non-Markovian effects matched an IBM quantum computer accessed via the cloud seven times better than commonly used error models, while separate work tied hard-to-correct errors in Google’s Willow experiment to the same issue.
New measurements by Ioan Pop’s team also uncovered long-lived defects in superconducting qubits’ environments that can act as memories, directly violating the long-standing assumption that errors at different times are uncorrelated.
Researchers say this does not prove quantum error correction will fail, but it shifts the burden onto hardware makers to show correlated noise stays weak enough; IBM said the cited study used older hardware and newer devices are designed differently.
The broader outlook remains open because newer QLDPC error-correction codes, improved device diagnostics and classical control tweaks may adapt to correlated noise, though no one yet knows how much memory effect practical quantum computers can tolerate.