Quantum Error Correction Faces Another Hurdle
Building a working quantum computer is challenging because its basic components, qubits, are highly sensitive to environmental disturbances that compromise computation. Whereas classical bits can only undergo bit-flip errors that change 0 to 1 or vice versa, qubits also suffer from so-called phase errors that degrade the fundamental quantum interference effects essential for quantum computation. Joining several good, but not perfect, physical qubits into a logical qubit makes quantum error correction possible (see Research News: Cracking the Challenge of Quantum Error Correction). But that strategy can fail if too many qubits become faulty at the same time. In one leading hardware platform, superconducting circuits, such correlated qubit errors are typically triggered every few tens of seconds when ionizing radiation from the environment deposits energy into the chip hosting the circuits. Now Vladislav Kurilovich, Gabrielle Roberts, and their colleagues at Google Quantum AI in California have shown that superconducting qubits designed to be less sensitive to radiation-induced bit-flip errors are nonetheless affected by phase errors, highlighting the need for improved protection strategies [1].
Quantum-error-correction schemes assume that errors affect each qubit randomly and independently. But in practice, solid-state qubits share a single substrate, which is typically silicon or sapphire for superconducting qubits. The substrate itself can be a source of errors. For instance, defects in the material can interact with a nearby qubit, causing a loss of quantum coherence. Currently, researchers are investigating various superconducting materials and fabrication processes to minimize this decoherence mechanism, which generally plays the main role in limiting the qubits’ coherence times [2]. Those times range from tens of microseconds to a few milliseconds, much longer than the tens to hundreds of nanoseconds needed to perform a quantum gate operation. This large difference in timescales is one ingredient that enables error correction.
Another, more dangerous error source comes from ionizing radiation (Fig. 1). When radiation from cosmic rays or radioactive decays passes through the substrate, it can generate many energetic lattice vibrations known as phonons, which quickly reach multiple qubits [3, 4]. There, the phonons break apart Cooper pairs—electron pairs that give rise to superconductivity when they condense at low temperature, opening a superconducting gap in the electronic energy spectrum. The broken Cooper pairs then act as quasiparticles and are a source of decoherence. When they pass through a Josephson junction—the circuit element underpinning all superconducting qubits—they can exchange energy with the qubit, causing a bit flip. One way to limit this quasiparticle passage is called gap engineering, in which the two sides of the junction are intentionally made slightly different [5]. That works because when a superconducting gap is lower in energy on one side than on the other, a quasiparticle in the lower-gap side might not have enough energy to reach the other side.
The Google team has tackled a more subtle effect of radiation-driven errors. In addition to causing bit flips, the quasiparticles also slightly change the qubits’ energy levels, leading to phase errors. But unlike for bit flips, quasiparticle passage through the junction is not a necessary condition for phase errors. In the absence of gap engineering and for typical “transmon” superconducting qubits, roughly half the energy-level change is expected to be induced by quasiparticles passing through the junction. The remainder comes from quasiparticles staying on one side [6]. The latter quasiparticles slightly suppress the superconducting gap, thus reducing the energy difference between the qubits’ two states. The Google team studied the key question of whether this gap-suppression process still causes errors in gap-engineered qubits. To answer that question, the researchers modified a measurement scheme that they had used previously to identify bit-flip errors [4]. Rather than checking if a qubit in the 1 state had flipped to 0 after a certain time, they prepared the qubit in a quantum superposition of 0 and 1, undid the superposition after a certain time, and then measured the qubit to check for an error.
The researchers repeatedly checked for bit-flip and phase errors in several tens of superconducting qubits on a chip. They found that they could not only identify “bursts” of errors spreading across several qubits but also characterize the bursts’ spatial extent and duration. Remarkably, compared with the bit-flip errors, the phase ones persisted tens of times longer and involved about 1.5 times as many qubits. For the bit-flip errors, the qubits were mostly protected by gap engineering, with residual errors arising from energetic quasiparticles that quickly cooled down. In contrast, the phase errors were mitigated only after the slow process of the quasiparticles recombining into Cooper pairs. As expected, during the error bursts, the qubits’ transition frequencies were lower, and the number of errors detected as part of error correction rose. Moreover, intentionally changing the qubits’ frequencies also boosted the error number.
This work unambiguously identifies quasiparticle-induced phase errors as one culprit for the nonzero error rate detected in qubits supposedly protected by gap engineering [7]. The results also represent another step in the decades-long quest to tame these unwanted effects [6, 8]. For superconducting qubits, it means that additional error-mitigation strategies besides gap engineering must be implemented. Researchers have already demonstrated two: traps that keep quasiparticles away from Josephson junctions [9]; and phonon down-conversion, in which the phonons’ energy is lowered below the pair-breaking threshold so that they do not harm the qubits [10]. Semiconductor-based qubits and other solid-state platforms could also benefit from these efforts. In addition to hardware-level approaches, the Google team has also demonstrated improvements in a standard error-correction scheme that make it more resilient to correlated phase errors. Devising even more robust correction schemes would advance the entire field of practical quantum computation.
References
- V. D. Kurilovich et al., “Correlated phase error bursts in a gap-engineered superconducting qubit array,” Phys. Rev. X 16, 021025 (2026).
- I. Siddiqi, “Engineering high-coherence superconducting qubits,” Nat. Rev. Mater. 6, 875 (2021).
- C. D. Wilen et al., “Correlated charge noise and relaxation errors in superconducting qubits,” Nature 594, 369 (2021).
- M. McEwen et al., “Resolving catastrophic error bursts from cosmic rays in large arrays of superconducting qubits,” Nat. Phys. 18, 107 (2021).
- G. Marchegiani et al., “Quasiparticles in superconducting qubits with asymmetric junctions,” PRX Quantum 3, 040338 (2022); M. McEwen et al., “Resisting high-energy impact events through gap engineering in superconducting qubit arrays,” Phys. Rev. Lett. 133, 240601 (2024).
- G. Catelani et al., “Relaxation and frequency shifts induced by quasiparticles in superconducting qubits,” Phys. Rev. B 84, 064517 (2011).
- Google Quantum AI and Collaborators, “Quantum error correction below the surface code threshold,” Nature 638, 920 (2024).
- J. Aumentado et al., “Quasiparticle poisoning in superconducting quantum computers,” Phys. Today 76, 34 (2023).
- R.-P. Riwar et al., “Normal-metal quasiparticle traps for superconducting qubits,” Phys. Rev. B 94, 104516 (2016); X. Pan et al., “Engineering superconducting qubits to reduce quasiparticles and charge noise,” Nat. Commun. 13, 7196 (2022).
- F. Henriques et al., “Phonon traps reduce the quasiparticle density in superconducting circuits,” Appl. Phys. Lett. 115, 212601 (2019); V. Iaia et al., “Phonon downconversion to suppress correlated errors in superconducting qubits,” Nat. Commun. 13, 6425 (2022).