Superinductor built in commercial silicon chip boosts quantum dot readout

Engineers have packed a superinductor—an electronic component whose impedance exceeds the superconducting resistance quantum of roughly 6.45 kilo-ohms—into a silicon integrated circuit made in a commercial chip factory, and used it to build one of the most sensitive single-electron sensors yet demonstrated on such a platform. The device, described in Nature Electronics, relies on titanium nitride films that are native to GlobalFoundries’ 22-nanometre fully depleted silicon-on-insulator process, rather than the bespoke superconducting structures that superinductors have historically required.

Superinductors are prized in metrology, sensing and quantum computing because their enormous impedance lets circuit designers isolate delicate quantum signals from lossy environments. Until now, however, they have generally been built from high-inductance structures such as Josephson junction arrays, superconducting nanowires or twisted two-dimensional materials, all of which demand specialised fabrication that sits awkwardly alongside conventional chip manufacturing. The new work shows that the kinetic inductance of a titanium nitride thin film already present in the foundry’s toolkit—around 1 nanohenry per square—can be harnessed instead, opening a route to superconducting inductive elements that emerge from an ordinary complementary metal–oxide–semiconductor process line.

The researchers did not stop at demonstrating the inductor in isolation. They co-integrated a resonator containing the superinductor with a silicon quantum dot formed within the same integrated circuit, creating a radio-frequency single-electron transistor, or rfSET. This type of sensor is widely used in semiconductor-based quantum computers, where it monitors the charge state of a quantum dot to read out the spin of an electron qubit.

The reported sensor occupies a footprint of just 18 square micrometres, small enough that many such sensors could be arrayed across a chip alongside the qubits they monitor. The low parasitics of the monolithic design, combined with the high impedance of the superinductor-based resonator and its nonlinear response, yielded a power signal-to-noise ratio of 4,500 in an integration time of 4 nanoseconds. That figure represents the maximum achievable signal, corresponding to a full linewidth shift of a Coulomb peak caused by a nearby quantum dot, according to the authors’ definition of signal-to-noise ratio in their methods section.

A radio-frequency signal generated by a Rohde & Schwarz source passes through a directional coupler and into attenuated copper–nickel coaxial lines that thermalise at each stage of a dilution refrigerator. At the mixing chamber, a second directional coupler routes the signal to the device, and the reflected signal travels back up through superconducting niobium–titanium cables with no attenuators between the mixing chamber and the 4-kelvin stage, where a low-noise amplifier boosts it. Further room-temperature amplification feeds an I–Q demodulator, whose differential outputs are digitised by an analogue-to-digital converter sampling at 250 megahertz.

To quantify sensitivity, the team recorded time traces at the top of a Coulomb blockade peak and at the background level away from the peak, producing two clusters of points in the in-phase and quadrature plane of the demodulated signal. Fitting two-dimensional Gaussians to those clusters gave the on- and off-state positions and noise widths, from which the signal-to-noise ratio followed. The minimum integration time—defined as the integration time at which the signal-to-noise ratio falls to 1, below which the signal can no longer be distinguished from background noise—was extracted by extrapolating a linear regression of signal-to-noise ratio against integration time, using boxcar averaging with varying window sizes to extend the effective integration window.

Circuit simulation played a central role in the design. The team modelled the poly-resistor-based kinetic inductor as an ideal inductor whose value was estimated from the geometry of the integrated structure using the Mattis–Bardeen formula in the low-frequency limit, supplemented by parasitic elements: a series resistance representing contact resistance to the thin film, estimated from d.c. measurements of previously fabricated superconducting structures, and a shunt capacitance representing coupling between the thin film and the chip substrate, extracted from the foundry’s conventional-temperature model of the poly-resistor structure. Simulations of the resulting circuit predicted a change in the magnitude of the reflection coefficient larger than 3 decibels between the two resistance states, a contrast large enough for reliable discrimination.

The authors state that the next step is to integrate the resonator with multigate transistors to demonstrate high-speed, high-fidelity spin read-out well below the microsecond timescale.

Subject of Research: A nonlinear superinductor fabricated in a commercial 22-nanometre silicon CMOS process and integrated with a quantum dot to form a compact radio-frequency single-electron transistor sensor.

Article Title: A nonlinear superinductor in a deep-submicrometre integrated circuit

Article References: Swift, T. H., Gomez-Saiz, A., Olivieri, F., Aizpurua-Iraola, G., Suárez-Rodríguez, M., Kirkman, J., Noah, G. M., de Kruijf, M., von Horstig, F.-E., Morton, J. J. L., & Gonzalez-Zalba, M. F. (2026). A nonlinear superinductor in a deep-submicrometre integrated circuit. Nature Electronics. https://doi.org/10.1038/s41928-026-01728-8

Image Credits: AI Generated

DOI: 10.1038/s41928-026-01728-8

Keywords: superinductor, titanium nitride, CMOS, kinetic inductance, radio-frequency single-electron transistor, quantum dot, spin qubit readout, GlobalFoundries 22-nm FD-SOI, quantum computing, parametric amplifiers, kinetic-inductance detectors, cryoelectronics