Superconducting qubits lose coherence through coupling to two-level systems (TLSs) — atomic-scale defects in the materials surrounding the qubit that absorb and re-emit microwave photons incoherently. Identifying which material hosts the dominant TLSs has been the central materials challenge in the field. Candidates include the substrate surface, the metal-substrate interface, the metal surface oxide, and bulk dielectrics.
Hedrick et al. (arXiv:2603.13183) isolate the dominant loss mechanism in aluminum resonators: the 2.7 nm native aluminum oxide (AlOx) that forms spontaneously when aluminum is exposed to air. By varying material properties and correlating microwave loss with surface characterization, they show that TLSs in this thin oxide layer primarily limit the lifetime of superconducting aluminum resonators.
HF treatment strips the oxide and temporarily improves coherence — confirming the oxide as the loss source. But the oxide regrows rapidly in ambient conditions, and the regrown oxide is no better than the original. The improvement is transient. This means that cleaning the surface isn't a solution; preventing oxide formation (or replacing aluminum at the critical interface) is required for permanent improvement.
The quantitative result connects to the broader ecosystem: TLSs in the aluminum oxide interface account for approximately 27% of the relaxation rate in state-of-the-art tantalum-on-silicon qubits that incorporate aluminum Josephson junctions. Even when the qubit's capacitor pads are made of cleaner materials (tantalum, niobium), the junction itself must contain aluminum (the only material with reliable oxide tunnel barriers), and that aluminum's native oxide degrades the entire qubit. The best material for the tunnel barrier is also the worst material for coherence.