The fabrication of superconducting circuits requires multiple deposition, etch and cleaning steps, each possibly introducing material property changes and microscopic defects. In thiswork, we specifically investigate the process of argon milling, a potentially coherence limiting step, using niobium and aluminum superconducting resonators as a proxy for surface-limited behavior of qubits. We find that niobium microwave resonators exhibit an order of magnitude decrease in quality-factors after surface argon milling, while aluminum resonators are resilient to the same process. Extensive analysis of the niobium surface shows no change in the suboxide composition due to argon milling, while two-tone spectroscopy measurements reveal an increase in two-level system electrical dipole moments, indicating a structurally altered niobium oxide hosting larger two-level system defects. However, a short dry etch can fully recover the argon milling induced losses on niobium, offering a potential route towards state-of-the-art overlap Josephson junction qubits with niobium circuitry.
As the superconducting qubit platform matures towards ever-larger scales in the race towards a practical quantum computer, limitations due to qubit inhomogeneity through lack of processcontrol become apparent. To benefit from the advanced process control in industry-scale CMOS fabrication facilities, different processing methods will be required. In particular, the double-angle evaporation and lift-off techniques used for current, state-of-the art superconducting qubits are generally incompatible with modern day manufacturable processes. Here, we demonstrate a fully CMOS compatible qubit fabrication method, and show results from overlap Josephson junction devices with long coherence and relaxation times, on par with the state-of-the-art. We experimentally verify that Argon milling – the critical step during junction fabrication – and a subtractive etch process nevertheless result in qubits with average qubit energy relaxation times T1 reaching 70 μs, with maximum values exceeding 100 μs. Furthermore, we show that our results are still limited by surface losses and not, crucially, by junction losses. The presented fabrication process therefore heralds an important milestone towards a manufacturable 300 mm CMOS process for high-coherence superconducting qubits and has the potential to advance the scaling of superconducting device architectures.