Could a quantum chip’s defects become its thermometers?

How do you measure temperature inside an insulating barrier that forms part of a superconducting quantum circuit? An external thermometer measures its own surroundings, not necessarily the microscopic environment responsible for errors in the circuit. Thermal bottlenecks and unwanted excitation can leave those environments different.

Asghar Ullah and colleagues propose using a defect inside the barrier itself. Their theoretical study, available as an arXiv preprint and not presented as a completed thermometry experiment, estimates that a carefully selected defect could deliver a temperature-fit precision of 0.58% at 20 millikelvin—twenty thousandths of a kelvin.

The important qualification is that this is a projection for one defect under specified measurement conditions, not a demonstrated accuracy for a working thermometer.

Reading temperature from shape

Some microscopic defects behave as two-level systems: they have a lower-energy state and a higher-energy state. In thermal equilibrium, the probability of occupying the higher state depends on the energy separation and temperature. A warmer defect is more likely to be excited.

The proposal changes that energy separation by mechanically straining the device with a piezoelectric actuator. Spectroscopy measures the defect’s transition frequency, providing its energy scale. Repeated measurements then build a curve showing how its excited-state population changes across the sweep.

The population would be read through a superconducting qubit using a controlled exchange of excitation, called a swap. The measured signal contains both an unknown background offset and an unknown contrast: the apparatus may report a shifted, weakened version of the true population response.

Rather than calibrating those two quantities beforehand, the researchers fit them together with temperature. Offset moves the curve vertically; contrast stretches or compresses it. Temperature changes its shape. With enough distinct energy settings, those effects can in principle be separated.

Here, “calibration-free” has a specific meaning. It removes the need for prior calibration of readout offset and contrast, not the need to characterize the apparatus. The protocol still requires spectroscopy, controlled readout and measurements of how quickly the defect relaxes between shots.

Why a broad sweep matters

A central result concerns identifiability: whether different unknown quantities produce distinguishable measurements. Over a very narrow energy interval, a temperature change can look almost like a change in background or contrast.

The authors use Fisher-information analysis, a statistical method for estimating how tightly measurements can constrain parameters, to quantify this problem. In their narrow-sweep regime, usable temperature information scales as the fourth power of sweep width. Halving that width therefore leaves roughly one-sixteenth as much information under otherwise comparable conditions.

Low transition energies matter too. At 20 millikelvin, a defect with a large energy separation is almost always in its ground state, leaving little thermal signal to measure.

For their main example, the researchers use parameters of a previously reported defect whose minimum transition frequency is 1.3 gigahertz. Sweeping it to 2.69 gigahertz would require an approximately eight-volt excursion of the strain actuator along one branch of its response.

Their budget projects a temperature uncertainty of about 0.116 millikelvin at 20 millikelvin, or 0.58%, for an acquisition lasting approximately 7.6 hours. Simulated measurements support the statistical fitting method. The final projection also includes residual frequency wandering, which competes with the benefit of collecting more data. An additional readout-related bias allowance of up to about 0.22% is listed separately.

What remains unproved

The example defect was inferred through its interaction with another defect, rather than directly read out in the proposed low-frequency experiment. Its relaxation time was not measured. Frequency-wandering and relaxation assumptions are extrapolated from higher-frequency defects and need testing in the intended operating band.

Readout offset and contrast must also remain sufficiently stable across the sweep. The proposed fixed-frequency readout qubit avoids one changing-background problem, but requires a calibrated parametric drive. Suitable, readable defects may be uncommon.

A curve that fits the thermal model is not proof of equilibrium at sub-percent accuracy. In the authors’ illustrative hot-component model, a detectable mismatch appears only after the fitted temperature is already biased by about 13%. Instrumental drift can produce similar mismatches.

Nor does one defect establish a common temperature for every defect in a junction. Comparisons across defects, devices and cooldowns remain missing. This article is based on the supplied paper text, including its supplement. Its strongest conclusion is a testable sensing proposal: imperfections might become local thermometers, provided experiments establish what their fitted temperatures actually represent.