Turning Qubit Defects Into Free Thermometers — No Calibration Required

Intrinsic two-level defects in Josephson junctions can measure temperature locally at 20 mK with 0.58% precision, no calibration needed — if you sweep hard enough.

Imagine you have a bathroom scale with no zero mark and no unit markings — just a needle that moves. You can't weigh yourself with a single reading, because you don't know where zero is or what a kilogram looks like on the dial. But if you step on and off repeatedly while adding known reference weights, the *shape* of how the needle moves tells you everything: both the zero offset AND the scale factor AND your actual weight. That's what this paper does with quantum defects and temperature. The committed claim: two-level system (TLS) defects that already live inside Josephson junction barriers — previously treated as noise sources to be suppressed — can serve as calibration-free primary thermometers at millikelvin temperatures, provided you sweep their energy splitting over a wide enough range using mechanical strain. No external calibration of readout offset or contrast is needed. The Boltzmann population curve's *shape* does the work. The key theoretical insight is an identifiability threshold that previous single-parameter thermometric bounds missed entirely. When you marginalize over unknown offset and contrast (the parameters you'd normally calibrate away), usable Fisher information about temperature scales as the *fourth power* of the sweep width. A narrow sweep is locally indistinguishable from a shift in your offset or contrast — you literally cannot tell temperature from a calibration drift. This quartic scaling means there's a sharp transition: below a critical sweep width, you get almost nothing; above it, information floods in. For the concrete demonstration, the authors analyze a reported TLS defect with tunneling energy Δ₀/h = 1.3 GHz, swept along one branch of its hyperbolic energy dispersion to 2.69 GHz (an asymmetry excursion of 2.36 GHz, achieved via 8 V of piezo voltage). At 20 mK, their Fisher-information analysis projects a single-defect fit precision of 0.58%. This is a theoretical projection, not an experimental measurement — they are analyzing the information content of the response curve, not reporting a thermometer they built and tested. The architecture is Bayesian parameter estimation applied to the Boltzmann population of a strain-tuned TLS. The sweep traces out the characteristic hyperbolic energy splitting of an asymmetric double-well defect, and the population data is fit to a three-parameter model (temperature, offset, contrast). The key hardware lever is piezo-driven mechanical strain, which tunes the TLS asymmetry energy over a wide range without changing the tunneling matrix element. This is what converts a nuisance defect into a useful sensor. The validation is entirely analytical — Fisher information bounds and projected fit precision from simulated response curves using parameters from existing experimental literature. No actual thermometry experiment was performed. The authors are honest about this: they establish the theoretical framework and identifiability conditions, using published defect parameters to demonstrate feasibility. The same sweep that extracts temperature also overdetermines the Boltzmann functional form, enabling a built-in self-consistency check for deviations from thermal equilibrium. The practical value proposition is real but early. Superconducting qubit processors need local temperature monitoring, and TLS defects are already everywhere in junction barriers — they are the most abundant unwanted feature of the technology. Repurposing them as sensors is elegant. But the gap between 'Fisher information says 0.58% precision is achievable' and 'we built a working thermometer' remains wide. The next milestone is an actual experimental demonstration showing this precision is achievable in the presence of real noise, decoherence, and multi-defect confusion.