Native-oxide-passivated trilayer junctions for superconducting qubits

TL;DR

The paper presents native-oxide-passivated trilayer (NAPA) Josephson junctions using an Al–AlOx–Nb stack with native Al oxide as a low-loss sidewall passivation and a galvanic via to connect Nb wiring, addressing reproducibility and microwave-loss challenges in trilayer junctions. By avoiding aggressive oxide-removal on the passivation and implementing a via-based contact, the authors demonstrate wafer-scale fabrication of transmon-like qubits with time-averaged $T_1$ up to $30\ \mu\mathrm{s}$ at $f_q \approx 5\ \mathrm{GHz}$ and qubit quality factors near $10^6$, surpassing prior trilayer-based implementations. The work includes detailed junction characterization (RA scaling, subgap leakage, and gap measurements) and configurable double-junction qubits showing low charge dispersion and robust coherence across wafers, with Nb devices achieving higher $Q$ than Al devices on average. Collectively, these results indicate that NAPA junctions enable scalable, low-loss superconducting qubits on industry-standard platforms and open avenues for further optimization of etching, shunt-capacitor processing, and studies of quasiparticle phenomena in Nb–AlOx–Nb systems.

Abstract

Superconducting qubits in today's quantum processing units are typically fabricated with angle-evaporated aluminum--aluminum-oxide--aluminum Josephson junctions. However, there is an urgent need to overcome the limited reproducibility of this approach when scaling up the number of qubits and junctions. Fabrication methods based on subtractive patterning of superconductor--insulator--superconductor trilayers, used for more classical large-scale Josephson junction circuits, could provide the solution but they in turn often suffer from lossy dielectrics incompatible with high qubit coherence. In this work, we utilize native aluminum oxide as a sidewall passivation layer for junctions based on aluminum--aluminum-oxide--niobium trilayers, and use such junctions in qubits. We design the fabrication process such that the few-nanometer-thin native oxide is not exposed to oxide removal steps that could increase its defect density or hinder its ability to prevent shorting between the leads of the junction. With these junctions, we design and fabricate transmon-like qubits and measure time-averaged coherence times up to 30 $μ$s at a qubit frequency of 5 GHz, corresponding to a qubit quality factor of one million. Our process uses subtractive patterning and optical lithography on wafer scale, enabling high throughput in patterning. This approach provides a scalable path toward fabrication of superconducting qubits on industry-standard platforms.

Figures (11)

• Figure 1: (a) Schematic illustration of a NAPA junction and (b) false-colored cross-sectional transmission electron micrograph of a typical fabricated NAPA junction along the dashed black line shown in panel (a). The aluminum base electrode (Al-BE, orange), tunnel barrier (white dashed line) and niobium counter electrode (Nb-CE, light green) are insulated from the Nb wiring layer (Nb2a and Nb2b) by a native oxide passivation layer (red dashed line). The via providing galvanic contact between Nb2 and Nb-CE is on top of the junction.(c) False-color scanning electron micrograph of fabricated Xmon-type shunt capacitor patterned into Nb2, capacitively coupled to a readout resonator (partially visible at the top). (d) Close-up of two NAPA junctions in series, with a small island of Al-BE in between. (e) Further zoomed in close-up of the junction in panel (d).
• Figure 2: (a)-(d) TEM image and EDS analysis spectra on the junction cross-section shown in Fig. \ref{['fig:schematic']}, with the relative concentrations (b) aluminum, (c) niobium, and (d) oxygen. Native oxide layers, indicated by red arrows, are visible as areas with light contrast in panels (a) and (d) and dark contrast in panel (b). Note the absence of oxide between Nb-CE and Nb2b on the left and between Nb2a and Nb2b on the right, indicated by blue arrows, implying a galvanic contact. (e) The resistances of test junctions versus nominal junction size $d$, along with a fit showing the expected scaling with junction area (solid orange line). (f) Low-temperature current-voltage characteristics of a test junction with $d=$ 629 nm measured at $40~$mK, showing a supercurrent branch at zero voltage and a sharp current rise at a voltage 1.6 mV indicated by arrows, corresponding to the sum of the superconducting gaps of Nb and Al, confirming the Josephson behaviour of NAPA junctions.
• Figure 3: (a) Decay of a qubit at frequency $f_{ge} = 5.17$ GHz from the excited state, measured data (points) fit to exponential decay with $T_1=36µs$ (solid line), corresponding to a qubit quality factor $Q = 2 \pi f T_1 \approx$ 1 million. (b) Ramsey and (c) Hahn echo experiments fit to a decaying sinusoid and exponential decay, respectively. (d) $T_1$, $T_2^*$, $T_{2,\mathrm{echo}}$ fluctuate over time when monitored for several hours. (e) Two-tone spectroscopy experiment, with qubit transitions at frequencies $f_{ge}$, $f_{ef}$ and $f_{gf}/2$ indicated by arrows. We extract an anharmonicity of -56MHz.
• Figure 4: (a) Time-averaged qubit quality factors $Q$. Qubits with ground planes and the qubit island (filled markers) patterned into Nb2 produce on average $Q \sim 0.5$ million, while qubits with Al-BE (open markers) produce an average $Q$ of 0.2 million. Symbols indicate devices from different wafers.Inset shows two types of qubit variants with shunt capacitor and ground places patterned into Nb2 (top) or Al-BE (bottom). (b) Qubit frequency versus nominal junctions size $d$. Dashed gray lines are guides to the eye.
• Figure 5: Electrical circuit representation of a double-junction qubit.