A superconducting resonator routes strong microwaves without a bias

A microwave isolator lets signals travel one way while suppressing signals coming back. Superconducting quantum chips use such components to protect sensitive circuits and route control tones, but conventional ferrite devices need magnets and take up space. Could a patterned superconducting film do a narrower version of this job without a separate bias or pump?

Brendan MacKay, Abdul Mohamed, Andrea Alù and Shabir Barzanjeh report a two-port device built around one nonlinear niobium nitride (NbN) resonator. Their arXiv preprint describes a resonator and a partially reflecting path on the same 10-nanometre film, patterned in one lithography step and without Josephson junctions. The key is not a one-way material: it is that a sufficiently strong signal changes the resonator differently depending on which port it enters.

A resonance that moves with the signal

The resonator is a meandering strip whose kinetic inductance changes with current. At low power, waves entering from either direction have the same transmission, as reciprocity requires for this device’s linear response. But the resonator couples more strongly to port 1 than to port 2. A wave of equal power entering from port 1 therefore stores more energy in it and shifts its resonance farther downward in frequency.

That shift alone does not explain the large directional contrast. The device also uses a Fano resonance: a narrow resonant route interferes with a broad direct route through the feed line. At one frequency the two transmitted contributions nearly cancel, producing a deep minimum; nearby, they add. The researchers choose an operating point near that minimum. A forward signal shifts the resonance so its transmission moves toward the high-transmission side, while the weaker reverse-direction response stays near cancellation. The signal itself supplies the energy for this nonlinear effect, so no auxiliary drive is needed. This is not a violation of reciprocity for a fixed, linear device: the response changes with signal power.

In measurements near 10 GHz, the largest reported contrast was 53.0 decibels: forward transmission was −4.6 dB while reverse transmission was −57.6 dB. Contrast above 20 dB extended across 9.7 MHz when forward transmission remained above −6 dB. These are peak and operating-window results for strong tones, not a blanket isolation specification at every frequency or power.

The pulse test exposes a bottleneck

The authors also sent pulses lasting from 2 microseconds to 1 millisecond. In their bonded-port measurement configuration, the device achieved about 33 dB contrast during 2-microsecond and 5-microsecond pulses, and maintained similar contrast over a 1-millisecond pulse. However, the forward response took roughly 0.2–1.6 microseconds to switch at tested settings. A model using only the prompt kinetic-inductance nonlinearity predicted a much faster 10–60 nanoseconds. Tracking the transmission minimum showed that most of its shift developed over 0.6–3 microseconds. The authors suggest local heating or nonequilibrium quasiparticles as possible causes, but do not establish which mechanism is responsible.

The device also has a design trade-off. Its measured high-power contrast exceeded a 45.8 dB ceiling estimated from the low-power response because the transmission minimum deepened under drive. The minimal nonlinear model did not explain that full deepening; the paper discusses saturating internal loss or movement toward a special interference condition as possibilities, without assigning a cause. The coherent loss channel that helps create the deep minimum is inferred from fitting the measurements; a slot-line mode at an asymmetric bridge is proposed as a likely origin, not proven.

A router, not a universal isolator

The diode works only above a power threshold and only when signals arrive from one port at a time. The authors note that a weak reverse signal present alongside a strong forward signal is not blocked to leading order. It therefore cannot direct thermal noise or replace ferrite isolators that protect qubits. Its present operating power is also far above typical qubit-drive and readout levels, although the authors argue that a narrower resonator or stronger nonlinearity might lower the threshold.

This preprint demonstrates a potentially useful component for strong pump or control tones, with no magnetic field, auxiliary pump or modulation. The remaining tests are practical as well as physical: identify the delayed nonlinearity, improve switching speed, and measure a weak reverse probe while the strong forward tone is on. Until then, the result is best understood as a compact, power-selective microwave router rather than a general one-way valve for quantum circuits.