betaFeedback

WorksPhysics and Astronomy

Calibration-free quantum thermometry with strain-swept two-level defects

Abstract

Superconducting qubits operate within a well-defined temperature range, making local temperature monitoring important. Here we show that intrinsic two-level defects in Josephson-junction barriers can act as local thermometers by exploiting the shape of their population response, without requiring calibration of the readout offset or contrast. Marginalizing over these unknown parameters imposes an identifiability requirement absent from single-parameter thermometric bounds: for narrow sweeps, usable information scales as the fourth power of the sweep width. Hence, a narrow sweep becomes locally indistinguishable from a change in offset or contrast. We illustrate this principle with a strain-swept two-level defect in a Josephson junction. For a reported defect with tunneling energy Δ_{0}/h=\SI{1.3}{\giga\hertz}, swept along one branch of its hyperbola to \SI{2.69}{\giga\hertz} (an asymmetry excursion of \SI{2.36}{\giga\hertz}, or \SI{8}{\volt} of piezo voltage), the analysis projects a single-defect fit precision of \SI{0.58}{\percent} at \SI{20}{\milli\kelvin}. The resulting response enables temperature estimation without prior calibration of the readout offset and contrast, while the same sweep overdetermines the Boltzmann form and identifies significant deviations from thermality. These results establish intrinsic junction defects as a resource for calibration-free local thermometry and highlight identifiability as a key design criterion for primary temperature sensing.

Introduction—Superconducting quantum processors are designed around error budgets in which an effective temperature often characterizes residual qubit excitation. Beyond computation, these architectures serve as versatile platforms for high-precision quantum metrology [1]. For the device, however, the relevant quantity is the local temperature in the junction barrier, where microscopic two-level systems (TLSs) contribute to dielectric loss and decoherence [2, 3, 4, 5, 6]. External primary thermometers, including Coulomb blockade (CBT) [7, 8, 9], shot-noise thermometry [10], and 3^3He melting curve [11], are located outside the device and do not directly measure this local temperature. Thermal boundary resistance and electron–phonon decoupling can further produce a mismatch between the measured bath temperature and the local device temperature, with the latter described by Pep=ΣΩ(Te5−Tph5)P_{\mathrm{ep}} = \Sigma\Omega(T_{\mathrm{e}}^5 - T_{\mathrm{ph}}^5) [9]. This mismatch is evident in practice: the excited-state population of a transmon has been observed to saturate at an effective floor Teff≈30T_{\mathrm{eff}} \approx30 mK to 45 mK even when the mixing-chamber bath sits at 15 mK [12]. Driven two-level defects can also be pushed into out-of-equilibrium or population-inverted states [13].

On-chip secondary thermometers, such as TLS microresonators [14], infer temperature from a frequency shift whose amplitude depends on the TLS filling factor and zero-temperature loss tangent. They require calibration against a resistive standard and recalibration of the zero-temperature resonance f0f_0 after each cooldown, as frequency shifts can be comparable to the TLS-induced signal [14]. Similarly, thermometry via single fluctuators in qubit readout dots [15] or defects inside superconducting barriers [16, 17, 18] extract population ratios e−E/kBTe^{-E/k_{\mathrm{B}}T}, but without an independent measure of the energy EE, they yield only a temperature ratio instead of an absolute temperature. Qubit-population thermometry avoids amplitude calibration, but at a fixed frequency, it gives only one effective temperature and cannot confirm thermal equilibrium with the bath [12, 19]. This limitation is evident from the saturation observed below 35 mK [12]. Similarly, transmon population thermometry tracks the cryostat temperature only over a limited range of about 100 mK to 200 mK [20]. However, qubit coherence depends on microscopic processes within the barrier, making the local temperature near defects particularly relevant to understanding qubit errors. This raises a question: can the intrinsic properties of defects already present in superconducting qubits be exploited to monitor their local operational temperature, turning an otherwise unavoidable source of error into a useful sensing resource?

In this Letter, we show that mechanically tuning a defect’s asymmetry energy [21, 22] allows its population to be used for thermometry without calibrating the readout amplitude. The temperature is extracted from the shape of the response, avoiding the amplitude calibration required in conventional TLS thermometry [14]. We derive the operating point from the quantum Fisher information (QFI): in the low-temperature regime considered here, thermometric sensitivity becomes negligible at the splittings typical of superconducting qubits, so the minimum splitting Δ0\Delta_0 of the defect must be a few kBTk_{\mathrm{B}}T, about 2 GHz or less at 20 mK. This exponential degradation of equilibrium thermometry for probes with a lower-bounded gap is known [23] and identifies the low-energy defect regime as the relevant operating window, rather than the larger splittings of standard transmons. Finally, sweeping the energy axis overdetermines the Boltzmann form, so the fit residual reveals significant departures from thermal equilibrium. Such departures are otherwise difficult to detect when the steady state is non-Gibbsian, as when a TLS ensemble is driven through population inversion, where a single-point ratio would still return a finite temperature [13].

Hamiltonian and symmetry—In the localized double-well basis ∣L⟩,∣R⟩|L\rangle, |R\rangle, the system Hamiltonian is

H=HTLS+Hq+Hint+HB+HSB,(1)H = H_{\mathrm{TLS}} + H_q + H_{\mathrm{int}} + H_B + H_{SB}, \tag*{(1)}

where the Hamiltonians describing the defect, readout qubit, their coupling, and environment are given by [5, 24]

HTLS=12[ϵ(ϵs)τz+Δ0τx],ϵ=ϵ0+2γSϵs,(2)H_{\mathrm{TLS}} = \frac{1}{2}\left[\epsilon(\epsilon_s)\tau_z + \Delta_0\tau_x\right], \qquad\epsilon= \epsilon_0 + 2\gamma_S\epsilon_s, \tag*{(2)}
Hq=12ℏωqΣz,Hint=ℏg0τz⊗Σx,(3)H_q = \frac{1}{2}\hbar\omega_q\Sigma_z, \qquad H_{\mathrm{int}} = \hbar g_0\tau_z\otimes\Sigma_x, \tag*{(3)}
HSB=τz⊗B^,B^=∑kλk(bk+bk†).(4)H_{SB} = \tau_z\otimes\hat{B}, \qquad\hat{B} = \sum_k\lambda_k(b_k+b_k^\dagger). \tag*{(4)}

Here, ϵs=δL/L\epsilon_s = \delta L/L denotes the applied strain, γS\gamma_S is the deformation potential, ϵ\epsilon and Δ0\Delta_0 denote energies, and ωq\omega_q and g0g_0 denote angular frequencies. We follow the superconducting-circuit convention [5], where ϵ\epsilon and Δ0\Delta_0 denote the asymmetry and tunneling energy, respectively, unlike the original tunneling-model notation [2, 3]. An electric field enters equivalently through ϵ→ϵ+2p⋅E\epsilon\to\epsilon+ 2\mathbf{p}\cdot\mathbf{E} [5], but a dc bias applied to a transmon does not generate the corresponding field inside the junction barrier because the potential difference is compensated by Cooper-pair tunneling [25]. Thus, barrier defects are effectively strain-tuned in the present protocol [22, 25]. The qubit coupling, the strain drive, and the defect baths, phonons [3] and photons [5], all act through τz\tau_z, as they perturb the well asymmetry. Quasiparticles affect the junction differently: by altering the phase across the junction, they primarily degrade the readout qubit [15, 26, 27].

Diagonalizing Eq. (2) with cos⁡θ=ϵ/E\cos\theta= \epsilon/E, sin⁡θ=Δ0/E\sin\theta= \Delta_0/E, and E=(ϵ2+Δ02)1/2E = (\epsilon^2+\Delta_0^2)^{1/2} gives HTLS=Eσz/2H_{\mathrm{TLS}} = E\sigma_z/2. The coupling operator τz\tau_z in terms of u≡sin⁡2θ=(Δ0/E)2u \equiv\sin^2\theta= (\Delta_0/E)^2 becomes

τz=1−u σz−u σx.(5)\tau_z = \sqrt{1-u}\,\sigma_z - \sqrt{u}\,\sigma_x. \tag*{(5)}

The parameter uu sets the transverse and longitudinal coupling weights. The transverse component drives relaxation, Γ1=uS(E/ℏ)\Gamma_1 = uS(E/\hbar), while the longitudinal component drives pure dephasing, Γϕ=(1−u)S(ω→0)\Gamma_\phi= (1-u)S(\omega\to0). For a smooth bath, Γ1∝Δ02/E2\Gamma_1 \propto\Delta_0^2/E^2 [28], recovering the Δ02\Delta_0^2 dependence of the one-phonon tunneling model [3]. The defect noise is similarly divided between the frequencies ±E\pm E and zero frequency, with weights uu and 1−u1-u, respectively [5]. The qubit–TLS swap element is g=g0ug = g_0\sqrt{u}, giving the sin⁡θ\sin\theta dependence of the measured avoided-crossing size [4].

Three symmetry properties follow, each checked against exact diagonalization. First, τxτzτx=−τz\tau_x\tau_z\tau_x=-\tau_z gives HTLS(−ϵ)=τxHTLS(ϵ)τxH_{\mathrm{TLS}}(-\epsilon)=\tau_xH_{\mathrm{TLS}}(\epsilon)\tau_x, so the spectrum and all τz\tau_z-coupled rates are even in ϵ\epsilon. Thus, Γ1(ϵ)\Gamma_1(\epsilon) is symmetric about the degeneracy point [28]. An odd component in ϵ\epsilon provides a useful diagnostic. It cannot arise from a bath coupled only through τz\tau_z, and may indicate a symmetry-breaking mechanism, such as a strain-tuned partner defect. Experimentally, the symmetry of Γ1(ϵ)\Gamma_1(\epsilon) has been used to rule out near-resonant partner defects as the main cause of relaxation. When a partner defect is tuned through its own symmetry point, the response instead shows an S-shaped curve [29, 30]. Second, at ϵ=0\epsilon=0 one has {HTLS,τz}=0\{H_{\mathrm{TLS}},\tau_z\}=0. The operator that couples longitudinally to the baths then has strictly vanishing diagonal matrix elements, so first-order pure dephasing is forbidden by symmetry; a quadratic contribution of order δϵ2/2Δ0\delta\epsilon^2/2\Delta_0 remains. Third, within the rotating-wave approximation the Hamiltonian conserves σz+Σz\sigma_z+\Sigma_z, and in the one-excitation manifold

λ±=±12δ2+4ℏ2g02u,δ=E−ℏωq,(6)\lambda_{\pm} = \pm\frac{1}{2}\sqrt{\delta^2+4\hbar^2g_0^2u}, \qquad\delta= E-\hbar\omega_q, \tag*{(6)}

so the avoided-crossing gap is 2ℏg0u2\hbar g_0\sqrt{u} which is largest at u=1u=1 [31]. This reproduces the splitting S=Smax⁡cos⁡ηsin⁡θS=S_{\max}\cos\eta\sin\theta measured in phase-qubit spectroscopy [4], with sin⁡θ=u\sin\theta=\sqrt{u}, and the same u\sqrt{u} dressing of the bare coupling reported in transmon defect spectroscopy, g=g′Δ0/Δ02+ϵ2g=g'\Delta_0/\sqrt{\Delta_0^2+\epsilon^2} [22]. The suppression of dephasing at ϵ=0\epsilon=0 is also reflected in the data: defects with coherence times of several microseconds lie near their symmetry point, whereas typical coherence times are only 50 ns to 100 ns [22]. Direct strain-sweep measurements similarly show that most defects exhibit strongly suppressed pure dephasing at their degeneracy points [28]. The degeneracy point is thus simultaneously the point free of first-order dephasing, the point of strongest readout, and, because Δ0→kBT\Delta_0 \to k_BT for the defect considered here, the point of the sweep closest to the optimum of thermal sensitivity. These three optima coincide, which makes the measurement practical. The coincidence is in uu at fixed g0g_0; across different defects, coherence can instead be optimal at intermediate coupling [32], which is a different parameter axis from that traversed by a single-defect strain sweep.

Principle of the measurement—Swap spectroscopy maps the TLS population onto the qubit readout [30, 32, 33]. The raw signal is affine in the excited-state occupation

S(ϵ)=a+bpe(E(ϵ)),pe=[1+eE/kBT]−1,(7)S(\epsilon)=a+b p_e(E(\epsilon)), \qquad p_e=\left[1+e^{E/k_BT}\right]^{-1}, \tag*{(7)}

where aa and bb account for readout infidelity, contrast, qubit relaxation during the swap, and the reduced swap visibility associated with the coupling g0ug_0\sqrt{u}. The key prerequisite is that aa and bb remain fixed during the sweep while EE changes. This

is not automatic: a readout qubit tuned across the band acquires a frequency-dependent thermal population, which can contribute a background comparable to the TLS signal unless the qubit and defect share a temperature. We therefore adopt a fixed-frequency readout with a parametric swap, which keeps aa independent of EE [31].

Temperature controls the functional shape of the response S(E)S(E). So the hyperbolic strain dependence directly yields E/hE/h in frequency units, while its minimum fixes Δ0\Delta_0 [21]. Multiple strain points allow the three unknown parameters (a,b,T)(a,b,T) to be extracted simultaneously from the shape of the response, so the temperature inference requires only hh, kBk_{\mathrm{B}}, a frequency scale, and a measured population ratio, without an independently calibrated thermometric standard. As illustrated in Figure 1, changing TT uniquely alters the response curvature, whereas variations in offset aa or contrast bb merely translate or rescale the trace. This extraction of temperature from response shape mirrors the operating principle of normal-metal–insulator–superconductor primary junctions [7], but generates its energy axis locally inside the junction barrier.

TLS response for several temperatures

Figure 1. TLS response S(E)S(E) for several temperatures. Dashed curves show affine transformations of a reference response, illustrating that unknown offset and contrast cannot reproduce the temperature-dependent curvature.

However, this parameter identifiability strictly requires sufficient spectral leverage. Narrow or high-energy sweeps diminish the curvature, causing temperature changes to become degenerate with linear offsets and contrast shifts. More generally, for any tunable probe read through an unknown affine readout, the temperature precision in the narrow-sweep regime (ΔE≪kBT)(\Delta E \ll k_{\mathrm{B}}T) and in the readout-floor regime (a≫pe)(a \gg p_e) obeys

σTT≃720a(1−a) eEmin⁡/kBTbNtot(kBTΔE)2.(8)\frac{\sigma_T}{T} \simeq\frac{\sqrt{720a(1-a)}\,e^{E_{\min}/k_{\mathrm{B}}T}}{b\sqrt{N_{\mathrm{tot}}}}\left(\frac{k_{\mathrm{B}}T}{\Delta E}\right)^2 . \tag*{(8)}

When pep_e instead dominates the readout noise, the exponential becomes eEmin⁡/2kBTe^{E_{\min}/2k_{\mathrm{B}}T}. The (ΔE/kBT)−2(\Delta E/k_{\mathrm{B}}T)^{-2} scaling, a Δx−4\Delta x^{-4} dependence of the variance, is the narrow-sweep asymptote (Δx≪1)(\Delta x \ll1). It does not limit our budget sweep, which has Δx≈3.3\Delta x \approx3.3 and pe(Δ0)=4.2%>ap_e(\Delta_0)=4.2\%>a, and whose 0.58% precision is computed numerically (§ 2C of SM [31]). The same structure applies to other platforms: for silicon-germanium charge fluctuators, narrowing the energy window from the accessible 22.4 μ\mueV to 2 μ\mueV raises the bound from

about 2% to above 100% (§ 2A of SM [31]). Thus, sufficient spectral leverage is required not only to obtain high precision, but also to separate temperature from the unknown offset and contrast. Figure 2 maps this information loss across the accessible sweep parameter space, demonstrating how insufficient spectral leverage leads to severe bb–TT parameter correlation and matrix ill-conditioning. Here, the defect’s internal energy is tuned directly within the junction barrier. The residuals of the overdetermined single-temperature fit reveal significant deviations from thermal equilibrium. For a precision-optimized sweep, however, these residuals alone cannot distinguish genuine non-thermality from low-level readout-contrast drift, necessitating the spectroscopic readout protocol detailed below.

Four heat maps showing effective Fisher information, smallest Fisher eigenvalue, condition number, and $b$–$T$ correlation as functions of $x_{\min}$ and $\Delta x$

Figure 2. The effective Fisher information T2FT∣abT^2F_{T|ab}, smallest Fisher eigenvalue λmin⁡(F)\lambda_{\min}(F), condition number κ(F)\kappa(F), and bb–TT correlation are shown as functions of xmin⁡=Emin⁡/kBTx_{\min}=E_{\min}/k_{\mathrm{B}}T and Δx=(Emax⁡−Emin⁡)/kBT\Delta x=(E_{\max}-E_{\min})/k_{\mathrm{B}}T. The marked points indicate the four architecture windows and their frequency ranges: 1, 4–8 GHz (transmon); 2, 2–5 GHz (transmon); 3, 0.5–2.0 GHz (fluxonium); and 4, 0.5–1.6 GHz (low-ff probe).

**Sensitivity and operating point—**For a diagonal two-level steady state, the QFI is [23, 34]

FQ(T)=1T2x24cosh⁡2(x/2),wherex=EkBT.(9)\mathcal{F}_{Q}(T)=\frac{1}{T^2}\frac{x^2}{4\cosh^2(x/2)}, \qquad\text{where}\quad x=\frac{E}{k_{\mathrm{B}}T}. \tag*{(9)}

The population readout saturates this bound so that the quantum and classical Cramér–Rao bounds (QCRB and CRB) coincide. For a transmon probe, the two-level form is an approximation, and retaining higher anharmonic levels can increase the extractable information [35]. A barrier defect, by contrast, is a genuine double-well TLS with intrawell excitations at much higher energies, making the two-level description valid in the present temperature range. That same analysis places the useful transmon window at 50 mK to 500 mK, well above our operating point and without the identifiability constraint.

Equation (9) further assumes that the defect equilibrates to the bare Gibbs state of HTLSH_{\mathrm{TLS}}. For a system coupled to a bath, the stationary state is the mean-force Gibbs state [36]. Evaluated to second order in the coupling implied by the measured symmetry-point lifetimes, the resulting bias in the recovered temperature is below 10−310^{-3} mK, which is two to three orders of magnitude below the systematic floor, and it increases with the transverse weight uu [31]. The dimensionless quantity T2FQT^2\mathcal{F}_Q peaks at x∗=2.399x^*=2.399, determined by ex=(x+2)/(x−2)e^x=(x+2)/(x-2) [34], which at 20 mK corresponds to E/h=E/h= 1.00 GHz [Fig. 3(a)]. Including the reset cost, τshot∼5/Γ1\tau_{\mathrm{shot}} \sim5/\Gamma_1 with Γ1∝uE3coth⁡(E/2kBT)\Gamma_1 \propto uE^3\coth(E/2k_{\mathrm{B}}T) for a Debye bath, the information per unit time scales as x5coth⁡(x/2)/cosh⁡2(x/2)x^5\coth(x/2)/\cosh^2(x/2) and peaks at x∗≃5.0x^\ast\simeq5.0, corresponding to E/h=2.08 GHzE/h = 2.08\,\mathrm{GHz} [Fig. 3(b)]. This criterion compares defects at their symmetry points, where u=1u = 1, and therefore sets the condition Δ0≲5kBT/h\Delta_0 \lesssim5k_{\mathrm{B}}T/h. Along a single sweep at fixed Δ0\Delta_0, u=(Δ0/E)2u = (\Delta_0/E)^2 gives Γ1∝Δ02Ecoth⁡(x/2)\Gamma_1 \propto\Delta_0^2 E\coth(x/2) and moves the per-time optimum to x≈3.0x \approx3.0. Optimizing FQ/τshotF_{\mathrm{Q}}/\tau_{\mathrm{shot}} is the criterion used for nonequilibrium probes [37] and whenever estimation time is scarce [23]. Symmetry-point lifetimes of 0.4 μs0.4\,\mu\mathrm{s} to 6 μs6\,\mu\mathrm{s} have been measured for barrier defects [28, 32] with Δ0/h\Delta_0/h between about 66 and 13 GHz13\,\mathrm{GHz}. Rescaling the two bracketing lifetimes, 0.44 μs0.44\,\mu\mathrm{s} at 6.2 GHz6.2\,\mathrm{GHz} and 3.2 μs3.2\,\mu\mathrm{s} at 7.34 GHz7.34\,\mathrm{GHz}, with the same Γ1\Gamma_1 scaling gives T1T_1 between 23 μs23\,\mu\mathrm{s} and 0.53 ms0.53\,\mathrm{ms} over E/h=1.3 GHzE/h = 1.3\,\mathrm{GHz} to 2.69 GHz2.69\,\mathrm{GHz}, so τshot\tau_{\mathrm{shot}} lies between 0.11 ms0.11\,\mathrm{ms} and 2.6 ms2.6\,\mathrm{ms}; the maximum thermal-TLS phonon rate quoted at 20 mK20\,\mathrm{mK} is likewise only ≈1.9 kHz\approx1.9\,\mathrm{kHz} [17]. The assumed shot duration, τshot=0.30 ms\tau_{\mathrm{shot}} = 0.30\,\mathrm{ms}, averaged over the sweep, lies within this window. The reset cost therefore determines the sweep duration, and the optimum τ∗\tau^\ast derived below is the time required to acquire the shots in that sweep. Operating at 5 GHz5\,\mathrm{GHz} costs a factor of 496496 in shots, and at 10 GHz10\,\mathrm{GHz} a factor of 2×1072 \times10^7. Unlike a pure-dephasing probe, whose disturbance arises from coupling and decoupling [37], the swap removes one quantum from the defect per shot. The reset interval τshot\tau_{\mathrm{shot}} therefore limits the rate of measurement-induced disturbance, making the sweep slow by construction.

Temperature dependence of the quantum Fisher information

Figure 3. Temperature dependence of the quantum Fisher information. (a) FQ(T)F_{\mathrm{Q}}(T) for four splittings E/hE/h. (b) Normalized information per shot and per unit time. The per-shot Fisher information scales as x2/cosh⁡2(x/2)x^2/\cosh^2(x/2) and is optimized at x∗=2.40x^\ast= 2.40. Including the reset cost τshot≃5/Γ1\tau_{\mathrm{shot}} \simeq5/\Gamma_1, with Γ1∝uE3coth⁡(x/2)\Gamma_1 \propto uE^3\coth(x/2), the information accumulated per unit time scales as x5coth⁡(x/2)/cosh⁡2(x/2)x^5\coth(x/2)/\cosh^2(x/2), shifting the optimum to x∗≃5.0x^\ast\simeq5.0. At T=20 mKT = 20\,\mathrm{mK} this corresponds to E/h≃2.08 GHzE/h \simeq2.08\,\mathrm{GHz}. (c) Optimal splitting versus temperature for both criteria, compared with the transmon band.

This temperature-identifiability constraint determines which architecture is viable. Propagating Eq. (7) through a Fisher analysis in which aa and bb are marginalized gives temperature uncertainty σT=108 mK\sigma_T = 108\,\mathrm{mK} for a 4 GHz4\,\mathrm{GHz} to 8 GHz8\,\mathrm{GHz} sweep at N=106N = 10^6 shots per point, compared with 0.048 mK0.048\,\mathrm{mK} for 0.4 GHz0.4\,\mathrm{GHz} to 2.0 GHz2.0\,\mathrm{GHz}. The failure is one of rank: at 5 GHz5\,\mathrm{GHz} and 20 mK20\,\mathrm{mK}, pe≃6×10−6p_e \simeq6 \times10^{-6} and is nearly flat across any accessible sweep, so (a,b,T)(a,b,T) become degenerate.

The same conclusion is reached from the transmon side: for barrier defects at ETLS/h≈5.5 GHzE_{\mathrm{TLS}}/h \approx5.5\,\mathrm{GHz} and T=15 mKT = 15\,\mathrm{mK}, the Boltzmann factor is estimated at ∼10−8\sim10^{-8}, and the defects are taken to exhibit no thermal dynamics at all [16]. Primary operation therefore requires the low-frequency band, and the only way to reach it with a readable defect is to strain-tune toward the hyperbola minimum, which is therefore an enabling requirement. The constraint is the defect-specific manifestation of a general result. For an equilibrium probe whose gap is bounded below, the temperature uncertainty diverges exponentially as T→0T \to0, with the low-temperature scaling σT∼(T2/Emin⁡)eEmin⁡/2kBT\sigma_T \sim(T^2/E_{\min})e^{E_{\min}/2k_{\mathrm{B}}T}. Avoiding this exponential suppression requires the relevant energy gap to scale with temperature, such that the dimensionless ratio E(T)/kBTE(T)/k_{\mathrm{B}}T remains finite [23]. A transmon fixed near 5 GHz5\,\mathrm{GHz} is precisely a gap bounded below, and the tracking curve of Figure 4(b) realizes that escape. What the present analysis adds is the concrete hardware consequence: the failure is a loss of parameter identifiability, and the bound excludes the un-strained standard transmon outright unless the defect’s tunneling minimum Δ0\Delta_0 is properly matched to kBTk_{\mathrm{B}}T and swept over a sufficient strain range.

Precision bounds at 20 mK

Figure 4. Precision bounds at 20 mK. (a) Lower bound with a,ba,b known, attainable bound with a,ba,b marginalized, the systematic floor, and their quadrature sum. The CRB with a,ba,b known crosses the systematic floor at N≃1.2×104N \simeq1.2 \times10^{4}, whereas the marginalized CRB crosses it at a total budget of Ntot∗≈9.27×107N_{\mathrm{tot}}^{*} \approx9.27 \times10^{7} shots (N∗≈8.42×106N^{*} \approx8.42 \times10^{6} shots per strain point). The latter defines the crossover relevant to the attainable bound. (b) σT\sigma_T for a fixed sweep and for one that tracks the optimal energy ratio as TT varies. (c) Relative uncertainty against CBT and PLTS-2000, reaching 0.58%.

Precision bounds—Figure 4 collects the precision bounds and systematic limits. Energy measurements on a Gibbs state give an exponential-family likelihood, so the single-parameter bound is saturable at finite shot count [23]. The asymptotic bound strictly applies to the marginalized three-parameter model, so we checked it by Monte Carlo simulation with a binomial maximum-likelihood estimator (MLE) started from randomized values. At N∗N^\ast and 20 mK20\,\mathrm{mK} the estimator spread is 0.0817 mK0.0817\,\mathrm{mK} against a CRB of 0.0820 mK0.0820\,\mathrm{mK}, the bias is 0.001 mK0.001\,\mathrm{mK}, and the empirical coverage is 68.8%68.8\% at 1σ1\sigma and 95.5%95.5\% at 2σ2\sigma [Sec. S4 of SM] [31]. The cooldown provides a reasonably narrow prior range for the temperature, allowing us to use local estimation theory to optimize the measurement around the expected operating point. If the temperature is not known within such a range, a global approach [38] is needed to avoid bias from the local approximation. The bound with aa and bb known is nonetheless unattainable in practice, since they are unknown; marginalizing over these nuisance parameters increases uncertainty. Under real-time in-situ spectroscopic tracking, linear first-order frequency drift is continuously compensated. The remaining contribution is the tracking residual, the part of the frequency wander within each dwell that tracking leaves uncorrected, which enters the estimate at first order. We take this residual equal to the untracked intra-dwell wander, the conservative case η=1\eta= 1 of Sec. S8 of SM [31], which contributes 0.082 mK at the optimal sweep duration. It changes from sweep to sweep and is therefore a random term. The rate is fixed by the measured D=2.5 MHz/hD = 2.5\,\mathrm{MHz}/\sqrt{\mathrm{h}}, defined through σ(t)=2Dt\sigma(t) = 2D\sqrt{t} and correspond to an rms wander of 5.0 MHz after one hour [16]. It has since been measured independently for eight defects in different devices, using the same convention. The measured value is (2.2±0.1) MHz/h(2.2 \pm0.1)\,\mathrm{MHz}/\sqrt{\mathrm{h}} [39]. The slow spectral instability of barrier defects has also been reported [17, 40]. Sub-second adaptive spectroscopy has revealed a faster component of frequency wander, characterized by Ornstein–Uhlenbeck jitter with a correlation time of 32 s and a spectral diffusivity of D≈0.9 MHz2s−1D \approx0.9\,\mathrm{MHz}^{2}\mathrm{s}^{-1}, alongside telegraphic switching on a timescale of 2.2 s. These dynamics have been observed in independently fabricated devices in two laboratories [41]. Both are transmon-band measurements which are treated on the same footing as DD, and are shorter than the 45.8 min spent at each strain point by factors of 86 and 1.2×1031.2 \times10^{3}. They therefore smear the energy axis and enter only at second order in the fast rms wander σf=5.5 MHz\sigma_f = 5.5\,\mathrm{MHz}, contributing 7×10−4 mK7 \times10^{-4}\,\mathrm{mK} [31]. The statistical uncertainty and the tracking residual are equal at N∗≈8.42×106N^{*} \approx8.42 \times10^{6} shots per strain point (9.27×1079.27 \times10^{7} in total) Figure 4(a), where each contributes 0.082 mK at the optimal sweep duration τ∗=7.63 h\tau^{*} = 7.63\,\mathrm{h}. Combining all contributions in quadrature yields an attainable total uncertainty of σT=0.116 mK\sigma_T = 0.116\,\mathrm{mK}, corresponding to a relative precision of σT/T=0.116/20=0.58%\sigma_T/T = 0.116/20 = 0.58\% at T=20 mKT = 20\,\mathrm{mK}. This 0.58% figure represents the projected single-defect fit precision under in-situ tracking, establishing the protocol’s attainable floor while leaving inter-defect and cross-device reproducibility to be validated experimentally [31]. A probe whose minimum splitting follows Δ0=x∗kBT\Delta_0 = x^{*}k_{\mathrm{B}}T would avoid the exponential loss of a fixed band [dashed curve in Figure 4(b)]; the reported defect cannot do this at 20 mK, where its sweep starts at x=3.12x = 3.12. For its fixed sweep, Figure 4(c) shows a minimum of 0.50% near 28 mK and 0.74% at 50 mK; the uncertainty exceeds 1% only below about 14 mK and above about 62 mK. The budget describes one defect, with Δ0/h=1.3 GHz\Delta_0/h = 1.3\,\mathrm{GHz} setting the lower energy scale, and the local slope ϵ/h\epsilon/h converts frequency wander into temperature uncertainty point by point across the sweep [31]. A tracking residual smaller than the untracked wander improves the precision as η1/2\eta^{1/2} at the cost of a run longer by η−1\eta^{-1}: a single frequency per dwell gives η=1/2\eta= 1/\sqrt{2} and 0.44% at τ∗=13.1 h\tau^{*} = 13.1\,\mathrm{h} [31].

**Conclusion—**We proposed that a strain-swept TLS can serve as a candidate primary thermometer located inside the Josephson barrier. With a projected single-defect fit precision of 0.58% (0.116 mK) at 20 mK for a reported defect with Δ0/h=1.3 GHz\Delta_0/h = 1.3\,\mathrm{GHz} swept to 2.69 GHz, useful operation constrains the hyperbola minimum rather than the upper end of the sweep. The sweep must also span enough energy to make the response identifiable, which for this defect requires an asymmetry excursion of 2.36 GHz, or 8 V along one branch at the measured tuning rate of 295 MHz V−1^{-1} (16 V if both branches are swept). Our analysis shows that this requirement is architectural. At 20 mK, the transmon band is exponentially penalized, whereas low-frequency probes can reach the sub-percent regime before accounting for the systematic floor. This advantage combines precision with local sensing: the thermometer probes microscopic degrees of freedom inside the barrier that directly contribute to qubit decoherence, allowing local temperature changes to be tracked within the qubit’s operating range. The quoted 0.58 % is the single-defect fit precision under in-situ spectroscopic tracking and excludes the readout-contrast allowance, at most 0.22 %, of Table S2. Experimental validation should therefore rely on comparison with an independent primary thermometer and on the consistency of the swept response with a single temperature. The residual-based test detects a 2ω2\omega admixture only when it has already biased the inferred temperature by approximately 13% at the operating shot count. The low-frequency TLS temperature affects qubit errors through spectral diffusion, dephasing, and TLS-induced excitation. Applying this temperature to the 5 GHz TLSs that dominate dielectric loss requires assuming a common local temperature across the defect population, which our protocol does not test. Three experimental issues remain. First, the tracking residual and the spectral diffusivity in the sub-transmon band set the budget and remain to be measured. Second, the readout-linearity assumption underlying Eq. (7) requires direct validation, particularly when the defect passes near the readout resonator. Third, the reported defect couples to its readout qubit at only 3 MHz, so the swap readout needs a readout-qubit lifetime of about 12 µs or longer. These tests determine whether the projected precision can be achieved in a fully implemented in-barrier thermometer. Thus, the goal is a sufficiently precise thermometer at the location where the dominant qubit errors originate.

ContributionσT\sigma_T (mK)σT/T\sigma_T/T
Random uncertainty budget (quadrature sum)
Intra-dwell spectral smearing, propagated (7.63 h)0.08240.412 %
Counting statistics (8.42 × 10610^{6} per point)0.08200.410 %
Hyperbola fit (Δ0\Delta_0)0.00600.030 %
EE spectroscopy0.00020.001 %
Fast bounded wander, refitted (σf=5.5\sigma_f = 5.5 MHz)≤0.0007\le0.0007≤0.004\le0.004 %
Total random uncertainty (σrand\sigma_{\mathrm{rand}}, in quadrature)0.1164 mK0.582 %
Systematic bias budget (one-sided / linear bound)
Mean-force correction≤+0.0012\le+0.0012≤+0.006\le+0.006 %
Total primary systematic bias (ΔTsys\Delta T_{\mathrm{sys}})[0, +0.0012][0, +0.006 %]
Additional readout allowance: ∣κ∣≤0.011\lvert\kappa\rvert\le0.011
Contrast drift (T1,q≳12T_{1,q} \gtrsim12 μs at g0/2π=3g_0/2\pi= 3 MHz)≤0.043\le0.043 mK≤0.215\le0.215 %
Combined systematic bias bound (∑∣bias∣\sum\lvert\mathrm{bias}\rvert)≤0.044\le0.044≤0.22\le0.22 %
Total Expanded Uncertainty Budget
Total uncertainty (0.1164 mK random + biases)0.1164 (rand); ≤0.044\le0.044 (sys)0.582 % (rand); ≤0.22\le0.22 % (sys)

Table S2. Accuracy budget at 20 mK, evaluated at the sweep duration τ∗=7.63\tau^* = 7.63 h (Ntot≈9.27×107N_{\mathrm{tot}} \approx9.27 \times10^7 total sweep shots, N∗=8.42×106N^* = 8.42 \times10^6 per strain point, τshot=0.30\tau_{\mathrm{shot}} = 0.30 ms averaged over the sweep, D=2.5D = 2.5 MHz h−1/2^{-1/2}), for the reported defect with Δ0/h=1.3\Delta_0/h = 1.3 GHz swept to 2.69 GHz, which requires an asymmetry excursion of 2.36 GHz, or 8 V along one branch. The smearing term is the tracking residual at η=1\eta= 1, propagated point by point with Eq. (53).

Acknowledgments— The authors acknowledge the ERC Proof of Concept grant AURORA (grant no. 101293210).

References

End Matter

Predicted configurations—Figure 5(a) maps the achievable relative uncertainty over tunneling energy and temperature, with the black contour marking 0.58%. At 20 mK and N=106N = 10^{6} shots per point, before the tracking residual is included, the statistics-limited σT/T\sigma_T/T is 54% for a transmon swept over 4 GHz to 8 GHz, 4.4% over 2 GHz to 5 GHz, 0.28% for a fluxonium over 0.5 GHz to 2.0 GHz, 0.39% for a dedicated low-frequency probe over 0.5 GHz to 1.6 GHz, and 1.2% for the reported defect over 1.3 GHz to 2.69 GHz [Figure 5(b)]. The reported defect is 0.58% as its shorter reset time allows more shots in the same run. These separate the architectures by several orders of magnitude: only probes operating in the sub-transmon band can reach the sub-percent statistics-limited regime, selecting fluxonium [42] or a purpose-built low-frequency probe. TLS-resonator thermometry faces a similar frequency constraint, with devices operating at 2 GHz usable only above 50 mK [14]. The band is not hypothetical: a granular-aluminum fluxonium at f01=1.2 GHzf_{01} = 1.2\,\mathrm{GHz} lies right at the 1.3 GHz tunneling minimum scale of our recommended defect and is already coupled to a TLS environment [13]. There, the defects are weakly and collectively coupled (g≳2π×1.2g \gtrsim2\pi\times1.2, with no resolved avoided crossings), so the device does not meet the additional requirement for the single-defect swap readout described by Eq. (6).

Three-panel uncertainty and sweep-duration plots

Figure 5. (a) Relative uncertainty of the full budget over (Δ0,T)(\Delta_0,T) for a sweep that starts at the hyperbola minimum and reaches an asymmetry excursion of 2.36 GHz (8 V along one branch); contours mark 0.58%, 1% and 5%, and the dot marks the reported defect at 20 mK. (b) Statistics-limited σT/T\sigma_T/T at N=106N=10^6 shots per point for five sweep windows: transmon, 4–8 and 2 GHz to 5 GHz; fluxonium, 0.5 GHz to 2.0 GHz; a dedicated low-frequency probe, 0.5 GHz to 1.6 GHz; and the reported defect, 1.3 GHz to 2.69 GHz. (c) Sweep-duration optimum τ∗=7.63\tau^*=7.63 h set by the competition between counting statistics (∝τ−1/2\propto\tau^{-1/2}) and the intra-dwell tracking residual (∝τ1/2\propto\tau^{1/2}), whose variance grows as τ/(Npts−1)\tau/(N_{\mathrm{pts}}-1).

Comparison with other in situ approaches—Nho et al. [43] obtain kBTq=hfq/ln⁡(Γ10/Γ01)k_{\mathrm{B}}T_q = h f_q/\ln(\Gamma_{10}/\Gamma_{01}) of a gap-engineered, offset-charge-sensitive 3D transmon, and resolve the substrate returning from 90 mK after a quasiparticle burst [44, 45, 46] with a 6 ms time constant. The regimes are complementary. Their qubit frequencies, fq=3.57 GHzf_q = 3.57\,\mathrm{GHz} to 4.78 GHz, are optimal according to Eq. (9) at 90 mK, but lie 9–50 times away from the optimum at their 25 mK base temperature. Their fixed-frequency population ratio resolves millisecond-scale dynamics, whereas our method uses a static frequency sweep. More importantly, a single population ratio gives a temperature estimate but cannot establish whether the populations follow a Boltzmann distribution. Our sweep tests this by fitting the Boltzmann form across multiple frequencies, with the fit residual revealing deviations from thermal equilibrium. The source explicitly states that its steady-state value of approximately 50 mK does not represent the sample temperature, being dominated by measurement error [43]. A fast secondary probe and a slow primary one answer different questions; the natural use is to calibrate the former against the latter. The same holds for the co-fabricated Dayem-bridge thermometer [47], which reads the chip temperature from the critical current without microwave calibration but remains secondary because of the material dependence of Ic(T)I_c(T).

Remaining assumptions—Three assumptions remain. First, strain is assumed to leave Δ0\Delta_0 unchanged. Although strain-induced changes in the tunneling amplitude are generally considered negligible [30], anticorrelated fluctuations of ε\varepsilon and Δ0\Delta_0 have been proposed to explain the coupling dependence of TLS coherence [32]. The residuals from the hyperbola fit can reveal deviations from this assumption. Second, a defect remains trackable over a multi-hour sweep. Telegraphic switching [17, 29] occurs on τsw=2.2 s\tau_{\mathrm{sw}} = 2.2\,\mathrm{s} and diffusive jitter on τc=32 s\tau_c = 32\,\mathrm{s} [41]. Both were measured in the transmon band, as was DD, and both are much faster than the hours-scale jump observed for the same reference. Since both timescales are short compared with the 45.8 min per strain

MethodTypeσT/T\sigma_T/TIn situWiringτres\tau_{\mathrm{res}}
CBT [8, 11]primary0.35%a^{a}nodcs
Shot noise [10]primary±15%b\pm15\%^{b}norfs
N-I-S junction [7]primary6–11%chipdcs
3^{3}He curve [11]primary0.5%nocells
RuO2_2/Cernox [48]secondary≳10%c\gtrsim10\%^{c}nodcms
TLS resonator [14]secondarycalib.d^{d}chipnonems
Qubit ratio [43]secondary∼10%\sim10\%chipnonems
Qubit population [20]secondary∼9%e\sim9\%^{e}chipnones
This workcandidate primary0.58%f^{f}barrierpiezoh

Table 1. Comparison with sub-kelvin thermometry. “In situ” denotes location inside the device under test; τres\tau_{\mathrm{res}} is the temporal resolution. Quoted uncertainties are those reported at or near 20 mK where available; note that the 1% figure often quoted for CBT applies above 250 mK [11].

provides an additional test of thermality. Without state selection, a ±50\pm50 MHz excursion would increase the total uncertainty from 0.58% to 0.65% [31]. A slow shift during the run can bias individual measurements without changing the overall estimate. The fit residual may reveal the shift, but cannot distinguish it from instrumental distortions over the operating band [31]. The residual quantifies this vulnerability as shown in Figure 6. At N∗N^* shots per point, the precision sweep detects a 200 mK hot admixture of weight wmin⁡≈2.7%w_{\min}\approx2.7\% at the 5σ5\sigma threshold (χ2=44.0\chi^2=44.0). Extending the sweep to 3.22 GHz, requiring an asymmetry excursion of 2.95 GHz (10 V along one branch), lowers this to 0.97%. Detection comes late: at wmin⁡w_{\min}, the fitted temperature is already biased by 2.6 mK (13%) for the precision sweep and by 1.3 mK for the extended sweep. An admixture of just 0.12% biases the temperature by 1σ1\sigma,

Residual chi-squared versus hot-admixture weight

Figure 6. Residual χ2\chi^2 versus the weight ww of a 200 mK hot admixture at N∗N^* shots per point, for the precision sweep (1.3 GHz to 2.69 GHz) and an extended sweep (1.3 GHz to 3.22 GHz). The dashed line marks the 5σ5\sigma threshold (χ2=44.0\chi^2=44.0); the dotted lines mark wmin⁡≈2.7%w_{\min}\approx2.7\% and 0.97%.

point, their effects average out within each point and bias the fit only through the curvature of pep_e [31]. Relative to the 0.30 ms duty cycle, the switching is resolved shot by shot, because a 100 MHz excursion leaves the swap band. The data can therefore be separated by state, with each branch fitted independently to obtain two temperatures whose agreement

while χ2≈0.17\chi^2 \approx0.17. The residual therefore reveals large deviations from thermality but cannot confirm thermality at the quoted precision. Moreover, the residual shape of a hot admixture is strongly correlated with that of readout-contrast drift over both sweeps (∣cos⁡θ∣=0.97|\cos\theta| = 0.97 and 0.930.93), so the residual detects a deviation without identifying its cause [31]. In an 82-defect survey, about 5% shifted spontaneously; the rest remained stable [32]. Third, and most restrictively, is that a suitable defect can be found. Transmon defect spectroscopy finds a junction-barrier density of 15 GHz−1^{-1} per qubit, but only 3% of detected defects show avoided crossings above 5 MHz [22]. Including the stability fraction and the log-uniform Δ0\Delta_0 distribution gives 0.03–0.05 qualifying defects per qubit. A 90% chance of finding at least one usable defect therefore requires about 45–70 junctions, which can be accommodated on a single multi-qubit chip. Junction area is also important: in-barrier defects couple to their qubit at a few to tens of megahertz (3 MHz for the reported defect), favoring the larger junctions used in strain-tuning experiments [31].

The main restriction is that reported defects swept through their symmetry point have Δ0/h\Delta_0/h between 6 GHz and 13 GHz [30, 32], well above the required band. This apparent clustering is partly a selection effect, since the symmetry point is resolvable only when it lies within the available tuning range [21]. A barrier defect with Δ0/h=1.3\Delta_0/h = 1.3 GHz has been strain-tuned through its symmetry point, which lies at a piezo voltage of −14.05-14.05 V, demonstrating that low-Δ0\Delta_0 defects exist [30]. At 20 mK it gives pe=4%p_e = 4\%, compared with 6×10−66 \times10^{-6} at 5 GHz. Its asymmetry tunes at 295 MHz V−1^{-1}, so the sweep to 2.69 GHz needs 8 V along one branch, from −14.05-14.05 V to about −6-6 V, a range covered by the reported measurements. A symmetric two-branch sweep would need −22-22 V to −6-6 V, within the −25-25 V to +7+7 V span of the reported measurements (Sec. S11 of SM [31]). The defect was never observed directly. It couples to the qubit at only 3.0 MHz, and its parameters were inferred through a strongly coupled partner defect in the qubit band (longitudinal coupling −428-428 MHz); its T1T_1 was not measured [30]. At 20 mK, the partner, near 6 GHz, remains in its ground state (thermal population ≈e−15\approx e^{-15}) and only shifts the hyperbola statically, which the fit absorbs. The weak coupling increases the readout-qubit lifetime required to meet the contrast allowance from 0.58 μ\mus at 60 MHz to about 12 μ\mus (Sec. S12 of SM [31]). Direct swap readout of this defect at its symmetry point remains the key experimental step.

Units and conventions

Throughout, we quote energies as frequencies, E≡E/hE \equiv E/h in GHz, and use kB/h=20.8366 GHz K−1k_B/h = 20.8366\,\mathrm{GHz}\,\mathrm{K}^{-1}, so that

x≡EkBT=E[GHz]20.8366 GHz K−1×T[K].(10)x \equiv\frac{E}{k_B T} = \frac{E[\mathrm{GHz}]}{20.8366\,\mathrm{GHz}\,\mathrm{K}^{-1} \times T[\mathrm{K}]}. \tag*{(10)}

At the reference temperature T0=20 mKT_0 = 20\,\mathrm{mK} this gives kBT0/h=0.416 73 GHzk_B T_0/h = 0.416\,73\,\mathrm{GHz} and kBT0/ℏ=2.6184×109 rad s−1k_B T_0/\hbar= 2.6184 \times10^9\,\mathrm{rad}\,\mathrm{s}^{-1}. We henceforth use scaled units in which all frequencies are measured in units of kBT/ℏk_B T/\hbar, such that β=1\beta= 1 and the operating point is x∗=2.3994x^* = 2.3994.

The two-level Hamiltonian, the internal coordinate uu, and the rotation to the energy eigenbasis are as in the main text,

τz=1−u σz−u σx,u=(Δ0E)2,(11)\tau_z = \sqrt{1-u}\,\sigma_z - \sqrt{u}\,\sigma_x,\qquad u = \left(\frac{\Delta_0}{E}\right)^2, \tag*{(11)}

with E=ε2+Δ02E = \sqrt{\varepsilon^2 + \Delta_0^2}.

Estimator and Fisher information analysis

The raw signal is defined as

S(ε)=a+b pe(E(ε)),(12)S(\varepsilon) = a + b\,p_e(E(\varepsilon)), \tag*{(12)}

where

pe(E)=11+eE/kBT.(13)p_e(E) = \frac{1}{1+e^{E/k_B T}}. \tag*{(13)}

The excited-state occupation probability of the TLS and the signal parameters aa and bb are unknown. For NN binomial shots at each of a set of splittings {Ei}\{E_i\}, the three-parameter Fisher information matrix for (a,b,T)(a,b,T) is

F=∑iNμi(1−μi)gigiT,(14)F = \sum_i \frac{N}{\mu_i(1-\mu_i)}g_i g_i^{\mathsf{T}}, \tag*{(14)}

where

gi=∇(a,b,T)μi=(1,pe,i,b pe,i(1−pe,i)xiT),(15)g_i = \nabla_{(a,b,T)}\mu_i = \left(1,p_{e,i},b\,p_{e,i}(1-p_{e,i})\frac{x_i}{T}\right), \tag*{(15)}

and

μi=a+b pe,i,xi=EikBT.(16)\mu_i = a + b\,p_{e,i},\qquad x_i = \frac{E_i}{k_B T}. \tag*{(16)}

After eliminating the nuisance parameters (a,b)(a,b), the effective Fisher information for temperature is given by the Schur complement

FT∣ab=qTW[I−P(PTWP)−1PTW]q,(17)F_{T|ab} = \mathbf{q}^{\mathsf{T}} W \left[I-P(P^{\mathsf{T}}WP)^{-1}P^{\mathsf{T}}W\right]\mathbf{q}, \tag*{(17)}

where

Pi=(1,pe,i),qi=b pe,i(1−pe,i)xiT,W=diag⁡(wi).(18)P_i = (1,p_{e,i}),\qquad q_i = b\,p_{e,i}(1-p_{e,i})\frac{x_i}{T},\qquad W = \operatorname{diag}(w_i). \tag*{(18)}

with

wi=Nμi(1−μi).(19)w_i = \frac{N}{\mu_i(1-\mu_i)}. \tag*{(19)}

Thus, temperature is estimable only from the component of the temperature response that is WW-orthogonal to both an additive offset and a rescaling of the population curve. The corresponding single parameter CRB for m=1m=1 repetitions is σT=(F−1)TT\sigma_T=\sqrt{(F^{-1})_{TT}}. Unless stated otherwise, we use a=0.02a=0.02, b=0.95b=0.95, 11 strain points, and N=106N=10^6 shots per point.

Application to semiconductor quantum-dot charge fluctuators

To show that the Schur-reduced criterion in Eq. (17) applies outside superconducting architectures, we evaluate it for silicon-germanium two-level charge fluctuators [15]. In such devices, pulse heating is inferred from changes in occupation bias, such that

B=e−ΔE/kBT.(20)B=e^{-\Delta E/k_{\mathrm{B}}T}. \tag*{(20)}

However, uncalibrated charge-sensing offset aa and contrast bb complicate absolute thermometry, forcing studies restricted to fixed or narrow tuning ranges to report relative temperature ratios R=T/T0R=T/T_0. The mathematical cause is captured by Eq. (17): in the narrow-sweep regime (ΔE≪kBT\Delta E \ll k_{\mathrm{B}}T), projecting out (a,b)(a,b) causes the effective Fisher information to collapse as FT∣ab∝(Δx)4F_{T|ab}\propto(\Delta x)^4, where Δx=ΔE/kBT\Delta x=\Delta E/k_{\mathrm{B}}T and drive the relative uncertainty as σT/T∝(Δx)−2\sigma_T/T\propto(\Delta x)^{-2}.

Using gate-lever-arm spectroscopy (λP≈5.6 μeV mV−1\lambda_P\approx5.6\,\mu\mathrm{eV}\,\mathrm{mV}^{-1}), an asymmetry tuning range of ΔEsweep≈22.4 μeV\Delta E_{\mathrm{sweep}}\approx22.4\,\mu\mathrm{eV} (ΔE/h≈5.42 GHz\Delta E/h\approx5.42\,\mathrm{GHz}) is experimentally accessible. At T=40 mKT=40\,\mathrm{mK} (kBT≈3.45 μeVk_{\mathrm{B}}T\approx3.45\,\mu\mathrm{eV}), this corresponds to a wide sweep span of ΔE/kBT≈6.5\Delta E/k_{\mathrm{B}}T\approx6.5. For an illustrative single-shot charge-sensing benchmark (a=0.05a=0.05, b=0.90b=0.90, T=40 mKT=40\,\mathrm{mK}, M=25M=25 points, N=3600N=3600 shots/point) [15], exact evaluation of FT∣abF_{T|ab} for a window starting at the symmetry point yields a statistical lower bound of σT/T=1.8%\sigma_T/T=1.8\%. Restricting the window to 0 μeV0\,\mu\mathrm{eV} to 2.0 μeV2.0\,\mu\mathrm{eV} (ΔE/h<0.48 GHz\Delta E/h<0.48\,\mathrm{GHz}) lowers FT∣abF_{T|ab} by about four orders of magnitude and gives σT/T≈190%\sigma_T/T\approx190\%. The same 2 μeV2\,\mu\mathrm{eV} window started at 2 μeV2\,\mu\mathrm{eV} gives 80%80\%, so the narrow-window penalty depends on where the window sits. The published measurement already spans the wide window, so the narrow-window figures illustrate the criterion and do not describe that experiment. Equation (S8) thus sets the spectral window a calibration-free measurement needs on this platform.

Readout floor and parameter identifiability

Two distinct mechanisms limit the temperature information at large splitting. First, when the background dominates the excited-state population, a≫pea\gg p_e, the binomial variance approaches μ(1−μ)≃a(1−a)\mu(1-\mu)\simeq a(1-a). Even when aa and bb are known, the Fisher information for temperature TT then scales as

FTT∝e−2x,(21)F_{TT}\propto e^{-2x}, \tag*{(21)}

reflecting the readout floor set by the background level. Second, when aa and bb are estimated jointly with TT, marginalising (a,b)(a,b) introduces an additional identifiability penalty, with the corresponding information vanishing as (Δx)4(\Delta x)^4 for narrow sweeps. This loss is distinct from the readout floor: it arises because temperature changes become increasingly indistinguishable from an offset and a rescaling of the population curve. The resulting loss of parameter identifiability is the key limitation addressed here.

Asymptotic behavior

To expose the origin of the large-gap suppression, we write x=x0+sx=x_0+s. In the finite-floor regime, where μ≃a\mu\simeq a, the temperature sensitivity is proportional to se−sse^{-s} after factoring out the overall scale e−x0e^{-x_0}. Since x0e−sx_0e^{-s} lies in the nuisance subspace spanned by {1,e−s}\{1,e^{-s}\}, the offset contribution is removed by the Schur-complement projection, giving

FT∣ab=b2Ntota(1−a)T2e−2x0∥Π{1,e−s}⊥[se−s]∥2,(22)F_{T|ab}=\frac{b^2N_{\mathrm{tot}}}{a(1-a)T^2}e^{-2x_0}\left\|\Pi_{\{1,e^{-s}\}^{\perp}}\left[se^{-s}\right]\right\|^2, \tag*{(22)}

where the projection is taken with respect to the shot-weighted inner product. Thus, the remaining shape factor depends only on the sweep width Δx\Delta x and shot distribution. For a narrow sweep, let s∈[0,L]s\in[0,L] with L=ΔxL=\Delta x. Using 1−e−s∈span⁡{1,e−s}1-e^{-s}\in\operatorname{span}\{1,e^{-s}\}, one obtains

se−s−(1−e−s)=−s22+O(s3).(23)se^{-s}-(1-e^{-s})=-\frac{s^2}{2}+O(s^3). \tag*{(23)}

The mean-square residual of s2s^2 after projection onto span⁡{1,s}\operatorname{span}\{1,s\} over [0,L][0,L] is L4/180L^4/180, yielding

G(Δx)=(Δx)4720+O((Δx)6),(24)G(\Delta x) = \frac{(\Delta x)^4}{720} + O\left((\Delta x)^6\right), \tag*{(24)}

and hence

σTT≃720a(1−a)ex0bNtot(Δx)2.(25)\frac{\sigma_T}{T} \simeq\frac{\sqrt{720a(1-a)}e^{x_0}}{b\sqrt{N_{\mathrm{tot}}}(\Delta x)^2}. \tag*{(25)}

We validate these asymptotic scaling behaviors in Figure S1. For narrow energy sweeps (Δx≪1\Delta x \ll1), the normalized temperature information collapses onto the universal scaling curve G(Δx)=(Δx)4/720G(\Delta x) = (\Delta x)^4/720 across different base frequencies x0x_0, which coincide identically in Figure S1. The Δx4\Delta x^4 scaling law represents the narrow-sweep asymptote and serves as an analytical benchmark; the full 0.58% figure is evaluated numerically across the entire sweep (pe=4.2%p_e = 4.2\%–0.16%0.16\%, Δx≈3.3\Delta x \approx3.3). Furthermore, evaluating T2FT∣abT^2F_{T|ab} across non-zero readout offsets aa confirms a clear scaling crossover [Figure S1]: while an ideal readout (a=0a = 0) decays as ∝e−x0\propto e^{-x_0}, any finite readout floor a>0a > 0 accelerates information loss to ∝e−2x0\propto e^{-2x_0} once x0≳−ln⁡ax_0 \gtrsim-\ln a.

Numerical verification of the asymptotic Fisher-information scaling

Figure S1. Numerical verification of the asymptotic Fisher-information scaling. (a) The shape factor G(Δx)G(\Delta x) versus sweep width Δx\Delta x for several values of x0x_0, together with the narrow-sweep prediction G(Δx)=(Δx)4/720G(\Delta x) = (\Delta x)^4/720. (b) The effective temperature information T2FT∣abT^2F_{T|ab} versus x0x_0 for different readout offsets aa, illustrating the crossover between the e−x0e^{-x_0} and e−2x0e^{-2x_0} regimes.

Single-parameter Fisher information

For a known signal contrast and offset, the corresponding single-parameter QFI is

FQ=x24T2cosh⁡2(x/2).(26)\mathcal{F}_{Q} = \frac{x^2}{4T^2\cosh^2(x/2)}. \tag*{(26)}

All quantities quoted in the main text are obtained using Eq. (14), with xshot∗=2.3994x_{\mathrm{shot}}^* = 2.3994, the root of ex=(x+2)/(x−2)e^x = (x + 2)/(x - 2), and xtime∗=5.0x_{\mathrm{time}}^* = 5.0 when the reset cost τshot∝1/uE3\tau_{\mathrm{shot}} \propto1/uE^3 is included. The corresponding shot penalties are 496×496\times at 5 GHz and 2.0×1072.0 \times10^7 at 10 GHz, relative to the optimum at 20 mK. More explicitly, retaining the thermal occupation factor,

Γ1∝uE3coth⁡(x2),(27)\Gamma_1 \propto uE^3\coth\left(\frac{x}{2}\right), \tag*{(27)}

so that the Fisher information accumulated per unit time scales as

FQτshot∝x5coth⁡(x/2)cosh⁡2(x/2).(28)\frac{\mathcal{F}_{Q}}{\tau_{\mathrm{shot}}} \propto x^5\frac{\coth(x/2)}{\cosh^2(x/2)}. \tag*{(28)}

Maximizing this expression gives xtime∗=5.0x_{\mathrm{time}}^* = 5.0. At T=20 mKT = 20\,\mathrm{mK}, this corresponds to an optimal splitting of approximately E∗/h≃2.08 GHzE^*/h \simeq2.08\,\mathrm{GHz}. Figure S2 shows the corresponding precision σT\sigma_T as the upper sweep limit is increased. The uncertainty remains small when the sweep is confined to the low-frequency band but increases rapidly as the sweep is extended toward the transmon band, demonstrating the loss of thermometric sensitivity and parameter identifiability at high TLS energies.

Temperature uncertainty as a function of the upper sweep limit

Figure S2. Temperature uncertainty σT\sigma_T as a function of the upper sweep limit Emax⁡/hE_{\max}/h. The low-frequency operating range and the transmon-band range are highlighted for comparison.

Tangent-space geometry

To evaluate parameter identifiability during narrow energy sweeps, we formulate a tangent-space geometric framework. We represent parameter sensitivities as data-space response vectors vθ=∂μ/∂θ\mathbf{v}_{\theta} = \partial\boldsymbol{\mu}/\partial\theta across sweep points EiE_i. Systematic distortions span the subspace Ssys=span⁡{va,vb,vΔ,vT1,q}\mathcal{S}_{\mathrm{sys}} = \operatorname{span}\{\mathbf{v}_a, \mathbf{v}_b, \mathbf{v}_{\Delta}, \mathbf{v}_{T_{1,q}}\}, where vT1,q=∂T1,qμ\mathbf{v}_{T_{1,q}} = \partial_{T_{1,q}}\boldsymbol{\mu} accounts for readout qubit relaxation drift.

Under the weighted inner product ⟨x,y⟩W=xTWy\langle\mathbf{x},\mathbf{y}\rangle_W = \mathbf{x}^{\mathsf{T}}W\mathbf{y} with noise variance weights Wii=[μi(1−μi)]−1W_{ii} = [\mu_i(1-\mu_i)]^{-1}, the weighted principal angle θ(vT,Ssys)\theta(\mathbf{v}_T,\mathcal{S}_{\mathrm{sys}}) between the thermal response vT\mathbf{v}_T and the systematics subspace is defined by

cos⁡θ=max⁡u∈Ssys∣⟨vT,u⟩W∣∥vT∥W∥u∥W.(29)\cos\theta= \max_{\mathbf{u}\in\mathcal{S}_{\mathrm{sys}}} \frac{\left|\langle\mathbf{v}_T,\mathbf{u}\rangle_W\right|}{\|\mathbf{v}_T\|_W\|\mathbf{u}\|_W}. \tag*{(29)}

For the reported defect, swept upward from its minimum at 1.3 GHz, Figure S3 shows the angle between the thermal response and the offset–contrast subspace falling from 5.0∘5.0^\circ at ΔE/h=1.39\Delta E/h = 1.39 GHz to 1.1∘1.1^\circ at 0.5 GHz and 0.43∘0.43^\circ at 0.3 GHz. Adding the contrast-drift direction vT1,q\mathbf{v}_{T_{1,q}} (g0/2π=3.0g_0/2\pi= 3.0 MHz, T1,q=20T_{1,q} = 20 μ\mus) reduces the angle to 0.05∘0.05^\circ at the full sweep width. This factor of about 100 reproduces the inflation found when T1,qT_{1,q} is fitted as a fourth parameter (Sec. A), so contrast drift must be controlled experimentally. Adding vΔ\mathbf{v}_{\Delta} makes the thermal response nearly degenerate with the systematics subspace (θ<0.01∘\theta< 0.01^\circ), so Δ0\Delta_0 must come from the spectroscopic hyperbola fit and not from the population fit, as in the protocol. A joint fit of sweeps taken at two readout amplitudes, with separate (ak,bk)(a_k,b_k) and a shared TT, does not lift this degeneracy: its Fisher information is the sum of the two single-sweep Schur complements, which only doubles the shot count. Breaking the degeneracy requires a parameter shared across the sweeps, such as the offset removed by the interleaved reference of § 8.

Weighted principal angle between the thermal response and the systematics subspace versus sweep width

Figure S3. Weighted principal angle between the thermal response and the systematics subspace versus sweep width, for the reported defect swept upward from Δ0/h=1.3\Delta_0/h = 1.3 GHz, for the offset–contrast subspace and with the contrast-drift direction added (g0/2π=3.0g_0/2\pi= 3.0 MHz, T1,q=20 μsT_{1,q} = 20\,\mu\mathrm{s}).

Finite-sample estimator performance and empirical coverage

To validate that our parameter estimation routine saturates the analytical CRB under finite shot counts, we evaluate its statistical performance via Monte Carlo simulations. For each strain-controlled energy point EiE_i (i=1,…,Mi = 1,\ldots,M), NN independent measurement shots yield ki∼Binomial⁡(N,pi)k_i \sim\operatorname{Binomial}(N,p_i) excited-state counts, corresponding to observed fractions yi=ki/Ny_i = k_i/N. The physical excitation model is given by

pi(a,b,T)=a+b1+exp⁡(EikBT),(30)p_i(a,b,T) = a + \frac{b}{1+\exp\left(\frac{E_i}{k_{\mathrm{B}}T}\right)}, \tag*{(30)}

where TT is the target operating temperature, bb is the readout visibility, and aa is the un-reset residual background offset.

To prevent estimation bias caused by low shot counts or low thermal occupation, the parameter vector θ=[a,b,T]T\boldsymbol{\theta} = [a,b,T]^{\mathsf{T}} is extracted directly using a binomial maximum-likelihood estimator (MLE). The optimal parameters minimize the negative log-likelihood function:

θ^=arg min⁡a,b,T[−∑i=1M(kiln⁡pi(a,b,T)+(N−ki)ln⁡(1−pi(a,b,T)))].(31)\hat{\boldsymbol{\theta}} = \underset{a,b,T}{\operatorname{arg\,min}}\left[-\sum_{i=1}^{M}\left(k_i\ln p_i(a,b,T) + (N-k_i)\ln\left(1-p_i(a,b,T)\right)\right)\right]. \tag*{(31)}

Each fit starts from randomized values a∈[0.005,0.05]a \in[0.005,0.05], b∈[0.8,1.0]b \in[0.8,1.0], and TT within ±20%\pm20\% of the true value inside the box a∈[0,0.1]a \in[0,0.1], b∈[0.5,1.0]b \in[0.5,1.0], and T∈[1,200]T \in[1,200] mK. We perform NMC=2000N_{\mathrm{MC}}=2000 Monte Carlo trials on the headline sweep (1.3 GHz to 2.69 GHz, 11 points) at Ttrue∈{10,20,30,50}T_{\mathrm{true}} \in\{10,20,30,50\} mK and N∈{104,105,106}N \in\{10^4,10^5,10^6\} shots per point, together with the operating point N∗=8.42×106N^* = 8.42 \times10^6 at 20 mK. Estimator performance is evaluated by the sample standard deviation

Std⁡(T^)=⟨(T^−⟨T^⟩)2⟩(32)\operatorname{Std}(\hat{T}) = \sqrt{\left\langle\left(\hat{T}-\langle\hat{T}\rangle\right)^2\right\rangle} \tag*{(32)}

and the bias Bias⁡(T^)=⟨T^⟩−Ttrue\operatorname{Bias}(\hat{T}) = \langle\hat{T}\rangle- T_{\mathrm{true}}. At the operating point the spread is 0.0817 mK against a CRB of 0.0820 mK, the bias is +0.001 mK, and the coverage is 68.8% at 1σ1\sigma and 95.5% at 2σ2\sigma (Table S1). Wherever the box constraints are inactive, the spread agrees with the CRB within the Monte Carlo uncertainty of about 1.6%. Where they are active (last column), they act as prior information: the spread falls below the CRB and the coverage exceeds its nominal value. This occurs at low shot counts and at 10 mK, where Δ0>5kBT/h\Delta_0 > 5k_{\mathrm{B}}T/h places the defect outside its design window.

TtrueT_{\mathrm{true}} (mK)NN (shots/point)CRB (mK)Std(T^\hat{T}) (mK)Bias(T^\hat{T}) (mK)Cov (1σ1\sigma)Cov (2σ2\sigma)Box active
10.010410^{4}17.96772.8977−0.2829-0.2829100.0%100.0%82%
10.010510^{5}5.68190.7423+0.4317+0.4317100.0%100.0%79%
10.010610^{6}1.79680.5520+0.3987+0.3987100.0%100.0%58%
20.010410^{4}2.38081.6470+0.8552+0.855282.2%95.5%48%
20.010510^{5}0.75290.5698+0.1864+0.186482.5%97.2%30%
20.010610^{6}0.23810.2310+0.0155+0.015566.2%97.3%8%
20.08.42×1068.42 \times10^{6}0.08200.0817+0.0011+0.001168.8%95.5%0%
30.010410^{4}2.37121.8538+0.6117+0.611784.2%96.4%36%
30.010510^{5}0.74980.6751+0.0517+0.051771.4%97.4%10%
30.010610^{6}0.23710.2331+0.0022+0.002269.3%95.8%0%
50.010410^{4}5.55104.0898+0.4966+0.496678.5%100.0%31%
50.010510^{5}1.75541.7514+0.1110+0.111068.8%95.1%0%
50.010610^{6}0.55510.5553+0.0031+0.003168.2%95.5%0%

Table S1. Finite-sample Monte Carlo performance of the binomial MLE on the headline sweep. Theoretical marginalized CRB, empirical standard deviation Std(T^\hat{T}), bias Bias(T^\hat{T}), empirical coverage at 68.3% (1σ1\sigma) and 95.4% (2σ2\sigma), and the fraction of fits with an active box constraint, over NMC=2000N_{\mathrm{MC}} = 2000 trials.

Measurement sequence and the readout offset

Equation (S5) assumes a constant offset aa throughout the strain sweep. This is a property of the measurement sequence and must be specified before evaluating the bound.

The swap in Eq. (7) of the main text requires the readout qubit to be resonant with the defect. A naive implementation, therefore, tunes fqf_q together with EE across the sweep. The qubit’s residual excitation then contributes to aa, but this excitation itself varies strongly with EE: at 20 mK, an equilibrium qubit has pe=4.23%p_e = 4.23\% at 1.3 GHz and pe=0.157%p_e = 0.157\% at 2.69 GHz. At the experimentally observed Teff≈45T_{\mathrm{eff}} \approx45 mK floor [12], the corresponding values are 20.0% and 5.37%. Thus, aa cannot be assumed constant during the sweep.

The resulting bias has a simple origin. If the qubit and defect are at the same temperature, the qubit excitation has the same EE dependence as the defect population and is absorbed into bb, leaving the temperature estimate unbiased. The problem arises when their temperatures differ. For a defect at 20 mK, fitting a constant aa gives 23.0 mK for a qubit at 25 mK (+15%) and 36.7 mK for a qubit at 45 mK (+83%). The naive sequence therefore fails in precisely the regime of interest and biases the estimate toward the qubit temperature. Two alternative sequences remove this problem, and the protocol requires one of them.

(A) Interleaved reference. At each strain point, alternate the swap measurement SiS_i with an otherwise identical measurement without the swap RiR_i. Their difference

Di=Si−Ri=b pe,i,(33)D_i = S_i - R_i = b\,p_{e,i}, \tag*{(33)}

removes the unknown offset aa and leaves a two-parameter fit for (b,T)(b,T). The additional reference shots double the binomial variance, but the reduced number of fitted parameters can improve precision. For a residual a=1%a = 1\% and active qubit reset, the statistical error is 0.51 times that of the three-parameter fit at equal shot count, giving a total uncertainty of 0.42% at τ∗=3.9\tau^{*} = 3.9 h. Without active reset, thermal excitation of the qubit at Teff=45T_{\mathrm{eff}} = 45 mK increases the reference noise, raising the ratio to 1.35 and the total uncertainty to 0.67% at τ∗=10.3\tau^{*} = 10.3 h. Active reset is therefore part of the measurement sequence.

(B) Fixed-frequency readout with a parametric swap. Keep the readout qubit at a fixed frequency and drive the swap at ∣fq−E∣\lvert f_q - E\rvert. Since the qubit frequency remains fixed during the TLS sweep, aa stays independent of EE, allowing the three-parameter estimator of Eq. (14) to be used. The main additional requirement is calibration of the parametric drive. For fq≈5f_q \approx5 GHz, the drive lies at 2.3 GHz to 3.7 GHz, outside the sweep band. The sideband rate is at most 0.58gg, set by the maximum of the Bessel factor J1J_1, resulting in a longer swap (§ 14).

We use sequence (B) for the main error budget because it gives the more cautious estimate. Sequence (A) provides a direct check of the offset aa and should be used as a control for either implementation.

Power of the residual

To show that the residual can distinguish a non-thermal state from a thermal one, we quantify the statistical power of the test. As a simple non-equilibrium model, we consider a mixture of a bath at temperature TT and a hot component at temperature ThT_h,

pe=(1−w)e−E/(kBT)+we−E/(kBTh)1+(1−w)e−E/(kBT)+we−E/(kBTh)(34)p_e = \frac{(1-w)e^{-E/(k_{\mathrm{B}}T)} + we^{-E/(k_{\mathrm{B}}T_h)}}{1 + (1-w)e^{-E/(k_{\mathrm{B}}T)} + we^{-E/(k_{\mathrm{B}}T_h)}} \tag*{(34)}

where ww is the weight of the hot component. We take Th=200T_h = 200 mK, generate the corresponding population data, and fit them with the constant-(a,b)(a,b) thermal model. We then evaluate the resulting χ2\chi^2 against the 5σ5\sigma threshold,

χ2=∑i(Si−Sifit)2σi2,(35)\chi^2 = \sum_i \frac{\left(S_i - S_i^{\mathrm{fit}}\right)^2}{\sigma_i^2}, \tag*{(35)}

where SiS_i is the measured signal, SifitS_i^{\mathrm{fit}} is the best-fit thermal model, and σi\sigma_i is the statistical uncertainty at strain point ii. For k=8k = 8 degrees of freedom, the 5σ5\sigma detection threshold is χ2≃44.0\chi^2 \simeq44.0 (one-sided χ2≃45.4\chi^2 \simeq45.4). At N∗=8.42×106N^{*} = 8.42 \times10^{6} shots per strain point, this threshold is used here. For the precision sweep, 1.3 GHz to 2.69 GHz, a hot component is nearly constant over the sweep and is largely absorbed into the offset aa: w=5%w = 5\% gives χ2=95\chi^2 = 95, and the minimum detectable admixture is wmin⁡=2.7%w_{\min} = 2.7\%.

At N=106N = 10^{6} the same admixture gives χ2=11\chi^{2} = 11, and no admixture up to 40% reaches the threshold. Extending the sweep to 1.3 GHz to 3.22 GHz (Vsym=20V_{\mathrm{sym}} = 20 V, or 10 V along one branch) gives χ2=441\chi^{2} = 441 at w=5%w = 5\% and wmin⁡=0.97%w_{\min} = 0.97\%. The power therefore grows linearly with the shot count, since χ2∝N\chi^{2} \propto N for a fixed distortion. Detection comes only after the distortion has biased the temperature. At wmin⁡w_{\min} the fitted temperature is 22.6 mK on the precision sweep (+2.6 mK, 13%) and 21.3 mK on the extended sweep, whereas an admixture of w=0.12%w = 0.12\% (0.085% on the extended sweep) biases TT by 1σ1\sigma, 0.116 mK, with χ2=0.17\chi^{2} = 0.17 (0.46). The residual test therefore detects gross non-thermality but cannot certify thermality at the level of the quoted precision.

The residual can also respond to instrumental effects, such as readout contrast drift or a strain dependence of Δ0\Delta_{0}. These effects cannot be separated by the residual alone. We compare residual shapes, the components of each distortion left after projecting out the (a,b,T)(a,b,T) directions, with the weighted inner product. On the precision sweep the residual shape of a hot admixture correlates with that of a linear contrast drift at ∣cos⁡θ∣=0.97|\cos\theta| = 0.97 and with that of a strain-dependent Δ0\Delta_{0} at 0.78, and contrast drift and Δ0\Delta_{0} correlate at 0.89. Extending the sweep to 3.22 GHz changes these to 0.93, 0.76 and 0.92, so the extension does not separate them. Therefore, a nonzero residual indicates a departure from the assumed thermal model. At the same time, the additional controls described in § 14 and the hyperbola-minimum test are needed to identify its origin.

What the residual does not test

A zero residual shows that the defect is consistent with a Gibbs state, but it does not identify the bath that sets its temperature. For example, a defect equilibrated with a confined phonon mode or a photon environment above the mixing chamber can still produce a Boltzmann population, a vanishing residual, and a well-defined temperature different from the cryostat temperature. Thus, the overdetermination of § 2 provides a test of consistency with thermality rather than a determination of the bath responsible for the temperature.

For the intended application, this distinction is not necessarily a problem: the error budget concerns the temperature of the degrees of freedom that limit qubit coherence, which is precisely what the defect probes. The ambiguity matters when comparing the defect temperature with an external thermometer.

The origin of the temperature can be tested by varying the bath while keeping the thermometer fixed. Sweeping the mixing-chamber temperature TMXCT_{\mathrm{MXC}} across the operational range of 20–100 mK against a trusted reference thermometer (such as a co-fabricated metallic resistor or CBT array), we require unit-slope tracking

dTTLSdTMXC=1,(36)\frac{\mathrm{d}T_{\mathrm{TLS}}}{\mathrm{d}T_{\mathrm{MXC}}} = 1, \tag*{(36)}

for a defect that remains thermally coupled to the mixing chamber. A local nonequilibrium source would generally produce a slope different from unity, whereas a static temperature offset would preserve the unit slope. Given the projected single-defect fit precision of 0.58% (σT=0.116\sigma_T = 0.116 mK at 20 mK), this test provides sub-millikelvin sensitivity, resolving static thermal offsets down to ≈0.1\approx0.1 mK across the sub-100 mK regime. Comparing multiple defects in the same and different junctions provides an additional consistency check and can help localize a local thermal load. These are validation tests but not calibrations. A co-fabricated CBT array measures the electron temperature of a metal film on the chip surface, not the temperature inside the junction barrier. Therefore, disagreement between the two does not by itself determine whether the in-barrier thermometer is biased or the barrier is genuinely hotter. Until the unit-slope test in Eq. (36) is demonstrated, the appropriate description is a candidate primary thermometer.

Defect yield

A usable defect must satisfy four requirements: (i) it must be reachable within the operating band, (ii) it must show a resolvable avoided crossing, (iii) it must remain stable throughout the sweep, and (iv) its hyperbola minimum must be sufficiently low. For the first three requirements, transmon defect spectroscopy reports a junction-barrier defect density of 15 GHz−1^{-1} per qubit and an avoided-crossing fraction of about 3% for splittings above 5 MHz [22].

The defect jump statistics are evaluated based on the 82-defect survey [32], where thermal cycling above 20 K was used across 8 cooldowns to reset the TLS distribution, with approximately 5% of defects exhibiting spontaneous spectral instability. Assuming a nominal observation window of τobs=24\tau_{\mathrm{obs}} = 24 h per cooldown cycle, the corresponding jump rate is γjump≈2.1×10−3\gamma_{\mathrm{jump}} \approx2.1 \times10^{-3} h−1^{-1}, yielding a defect survival probability of Psurv=e−γjumpτ∗≈98.4%P_{\mathrm{surv}} = e^{-\gamma_{\mathrm{jump}}\tau^{*}} \approx98.4\% over the optimal τ∗=7.63\tau^{*} = 7.63 h precision sweep. Across a plausible monitoring range of τobs∈[12,48]\tau_{\mathrm{obs}} \in[12,48] h, the survival probability remains high (Psurv∈[96.8%,99.2%]P_{\mathrm{surv}} \in[96.8\%,99.2\%]), confirming that defect jump instabilities contribute negligibly to the overall uncertainty budget.

The sweep-band condition and the requirement on Δ0\Delta_{0} are not independent; instead, both follow from the same joint distribution P(ε,Δ0)P(\varepsilon,\Delta_{0}). We therefore avoid factorizing these two conditions and integrate directly over the accessible region of the (ε,Δ0)(\varepsilon,\Delta_{0}) plane. With

E=ε2+Δ02,andP(Δ0)∝1Δ0(37)E = \sqrt{\varepsilon^{2} + \Delta_{0}^{2}}, \qquad\text{and} \qquad P(\Delta_{0}) \propto\frac{1}{\Delta_{0}} \tag*{(37)}

the normalized distribution over a population spanning

R=Δ0,max⁡popΔ0,min⁡pop(38)R = \frac{\Delta_{0,\max}^{\mathrm{pop}}}{\Delta_{0,\min}^{\mathrm{pop}}} \tag*{(38)}

is

p(Δ0)=1Δ0ln⁡R.(39)p(\Delta_0) = \frac{1}{\Delta_0 \ln R}. \tag*{(39)}

The accessible asymmetry-energy interval at fixed Δ0\Delta_0 for a sweep from Emin⁡E_{\min} to Emax⁡E_{\max} is

εmin⁡(Δ0)=max⁡(Emin⁡2−Δ02,0),εmax⁡(Δ0)=Emax⁡2−Δ02.(40)\begin{aligned} \varepsilon_{\min}(\Delta_0) &= \sqrt{\max(E_{\min}^2-\Delta_0^2,0)},\\ \varepsilon_{\max}(\Delta_0) &= \sqrt{E_{\max}^2-\Delta_0^2}. \tag*{(40)} \end{aligned}

The expected number of qualifying defects per qubit is obtained by integrating the joint (E,Δ0)(E,\Delta_0) distribution over the accessible region of the strain sweep

λ(R)=ρTLSfACfst∫Emin⁡Emax⁡dE∫Δ0,min⁡opmin⁡(Δ0,max⁡op,E)dΔ0Δ0ln⁡R,(41)\lambda(R)=\rho_{\mathrm{TLS}}f_{\mathrm{AC}}f_{\mathrm{st}}\int_{E_{\min}}^{E_{\max}}dE\int_{\Delta_{0,\min}^{\mathrm{op}}}^{\min(\Delta_{0,\max}^{\mathrm{op}},E)}\frac{d\Delta_0}{\Delta_0\ln R}, \tag*{(41)}

where RR is the assumed population span of Δ0\Delta_0. This form avoids treating the sweep-band and Δ0\Delta_0 requirements as independent filters.

For the headline operating range E/h=1.3E/h = 1.3 GHz to 2.692.69 GHz and Δ0/h=1.0\Delta_0/h = 1.0 GHz to 1.51.5 GHz, the resulting yields are re-integrated over the updated energy bounds, giving λ≈0.051\lambda\approx0.051 for a two-decade population span and λ≈0.034\lambda\approx0.034 for a three-decade span. For a narrow Δ0/h=1.2\Delta_0/h = 1.2 GHz to 1.41.4 GHz window centered on the headline defect, the corresponding yields are λ≈0.020\lambda\approx0.020 and 0.0100.010, respectively. These values should be regarded as extrapolations because both the defect density and the avoided-crossing statistics are measured primarily in the transmon band and are being applied here to the lower-frequency operating range. Moreover, the quoted 3 % avoided-crossing fraction was measured for u≪1u \ll1; since g=g0ug = g_0\sqrt{u} is maximal near the symmetry point, applying this fraction to the present operating band is conservative and likely underestimates the usable-defect yield. For a chip with Nq=10N_q = 10 qubits, the Poisson relation P≥1=1−e−NqλP_{\geq1} = 1-e^{-N_q\lambda} with λ=0.034\lambda= 0.034–0.0510.051 gives P≥1=29%P_{\geq1} = 29\%–40%40\% for at least one usable defect and a 90% chance needs 45 to 70 junctions. The avoided-crossing criterion of 5 MHz corresponds to g≳2π×2.5g \gtrsim2\pi\times2.5 MHz, consistent with the 3.0 MHz coupling of the reported defect.

In-situ spectroscopic tracking and intra-dwell smearing budget

Slow spectral wander of the defect transition frequency is continuously tracked and subtracted using in-situ adaptive spectroscopy during the sweep. The remaining contribution is the tracking residual, which is the spectral variation during each dwell that tracking does not correct. It is evaluated by propagating the residual point by point through the estimator using Eq. (53).

We use the diffusivity measured for barrier and near-junction defects by Klimov et al. [16]. They report D=2.5D = 2.5 MHz h−1/2^{-1/2}, with the rms frequency wander given by

σ(t)=2Dt.(42)\sigma(t) = 2D\sqrt{t}. \tag*{(42)}

This gives an rms wander of 5.0 MHz after one hour, rather than 2.5 MHz. The factor of two in Eq. (42) is important because omitting it changes the inferred floor by 2.13×2.13\times.

The same convention has since been tested independently. Reference [39] tracks eight defects over 24 h in flux-tunable transmons with different geometries and fabrication, and obtains D=(2.2±0.1)D = (2.2 \pm0.1) MHz h−1/2^{-1/2} using the same Eq. (42) convention. This agrees with the (2.5±0.1)(2.5 \pm0.1) MHz h−1/2^{-1/2} reported in Ref. [16] however, we retain the larger value as the more conservative estimate. The 20 min timing resolution of Ref. [39] probes the slow component but cannot resolve the faster dynamics discussed below.

Sub-second adaptive spectroscopy has since resolved this faster component [41]. The reported Ornstein–Uhlenbeck correlation time is τc=32\tau_c = 32 s, so the frequency wander saturates at long times. For comparison with Eq. (42), the corresponding short-lag diffusion constant is

Ds=3.47×10−3 MHz2s−1,(43)D_s = 3.47 \times10^{-3}\,\mathrm{MHz}^2\mathrm{s}^{-1}, \tag*{(43)}

while Ref. [41] reports D=0.93D = 0.93 MHz2^2/s for its short-time dynamics. The two measurements are therefore complementary: the 20 min resolution of Ref. [39] cannot resolve a 32 s correlation time, whereas the ∼103\sim10^3 s records of Ref. [41] do not probe hour-scale drift.

Reference [41] defines DD through the short-lag mean-squared displacement 2Dt2Dt, so the stationary rms wander of the fast Ornstein–Uhlenbeck (OU) component is

σf=Dτc=5.5 MHz,(44)\sigma_f = \sqrt{D\tau_c} = 5.5\ \mathrm{MHz}, \tag*{(44)}

with a robustness range of 4.0 MHz to 5.9 MHz. This is comparable to the 4.37 MHz slow-diffusion wander accumulated over one dwell, τ∗/10=0.763\tau^*/10 = 0.763 h. Both the correlation time τc=32\tau_c = 32 s and the telegraphic switching time τsw=2.2\tau_{\mathrm{sw}} = 2.2 s are much shorter than the 45.8 min spent at each strain point, so the fast fluctuations average within each point. To second order in the frequency wander,

⟨pe(x+δ)⟩=pe(x)+12σx2pe′′(x),(45)\langle p_e(x+\delta)\rangle= p_e(x) + \frac{1}{2}\sigma_x^2 p_e^{\prime\prime}(x), \tag*{(45)}

where

σx=εEσfkBT/h′,δT=12σx2pe′′(x∗)(∂pe∂T)x∗−1.(46)\sigma_x = \frac{\varepsilon}{E}\frac{\sigma_f}{k_B T/h'}, \qquad \delta T = \frac{1}{2}\sigma_x^2 p_e^{\prime\prime}(x^*)\left(\frac{\partial p_e}{\partial T}\right)_{x^*}^{-1}. \tag*{(46)}

However, this estimate is not sufficient because σx\sigma_x and pe′′(x)p_e^{\prime\prime}(x) vary across the sweep. We therefore evaluate the second-order correction at each of the eleven strain points and refit the resulting distorted sweep with the same three-parameter estimator. This gives

δT=4.94×10−4 mK,(47)\delta T = 4.94 \times10^{-4}\ \mathrm{mK}, \tag*{(47)}

rising to 6.02 × 10−410^{-4} mK at the upper 4.2 MHz value of the robustness range. The refitted value is used in Table S2. It is still 135× below the residual smearing floor and changes the total uncertainty only in the fourth decimal place. Even an ungated telegraphic excursion of ±50\pm50 MHz would contribute only 0.038 mK, changing the total from 0.581 % to 0.612 %. Thus, the residual intra-dwell smearing floor remains the dominant systematic contribution under in-situ tracking.

There is, however, an important observation-window limitation. The records of Ref. [41] span only ∼103\sim10^3 s, so a component with a correlation time between that timescale and one hour would remain unresolved and be included in the slow diffusion estimate of Eq. (42). The floor is therefore unchanged, but its uncertainty is not reduced. Moreover, the values of DD, τc\tau_c, and τsw\tau_{\mathrm{sw}} are measured for defects in the transmon band. Applying these timescales to the sub-2 GHz operating band is therefore an extrapolation, just as for DD itself.

The two extrapolations enter the budget differently. The diffusivity DD enters linearly and sets a substantial part of the systematic floor, so measuring it directly in the sub-2 GHz band is the first experimental priority. The timescales enter only through τc,τsw≪τ∗/10\tau_c,\tau_{\mathrm{sw}} \ll\tau^*/10, which currently holds by factors of 86 and 1.2×1031.2 \times10^3, respectively. A defect with τc\tau_c longer than 46 min would instead appear as a slow-shift distortion and could be identified through the hyperbola residual discussed in End Matter. The underlying strain wander projects onto transition-energy fluctuations through the local slope dE/dε=ε/EdE/d\varepsilon= \varepsilon/E of the hyperbolic dispersion. This slope varies across the sweep, from zero at the hyperbola minimum to 0.875 at 2.69 GHz, so we propagate it point by point through Eq. (53) rather than through a single operating-point factor. The systematic diffusion floor thus incorporates a geometric factor ε/E=3/2≈0.866\varepsilon/E = \sqrt{3}/2 \approx0.866, which projects strain-induced noise measured on asymptotic reference defects (ε/E≈1\varepsilon/E \approx1 at 5.5 GHz) onto the local dispersion slope of the budget defect at 1.0 GHz:

dEdε=εε2+Δ02=εE=1−(Δ0E)2=32≈0.866.(48)\frac{dE}{d\varepsilon}=\frac{\varepsilon}{\sqrt{\varepsilon^2+\Delta_0^2}}=\frac{\varepsilon}{E}=\sqrt{1-\left(\frac{\Delta_0}{E}\right)^2}=\frac{\sqrt{3}}{2}\approx0.866. \tag*{(48)}

The resulting energy wander is σE(τ)=(ε/E) h σ(τ)\sigma_E(\tau)=(\varepsilon/E)\,h\,\sigma(\tau). Propagating this fluctuation into temperature error yields the systematic floor:

σTdiff(τ)=(εE)σ(τ)E/h T0.(49)\sigma_T^{\mathrm{diff}}(\tau)=\left(\frac{\varepsilon}{E}\right)\frac{\sigma(\tau)}{E/h}\,T_0. \tag*{(49)}

This local hyperbola slope reduces the systematic temperature floor by a factor of E/ε=2/3≈1.155E/\varepsilon=2/\sqrt{3}\approx1.155 (1/0.866) relative to unprojected frequency diffusion. The τ1/2\tau^{1/2} growth of this systematic floor arises directly from the time-budget allocation: for a total sweep duration τ\tau partitioned across 11 strain points (10 step intervals), the dwell duration at each point scales as τ/10\tau/10. Under slow spectral diffusion, the frequency variance accumulated within a single dwell window grows as σ2(τ/10)=4D2(τ/10)∝τ\sigma^2(\tau/10)=4D^2(\tau/10)\propto\tau. In the absence of frequency tracking, slow spectral diffusion enters at first order in temperature bias via Eq. (49). However, because the in-situ tracking protocol subtracts the mean frequency estimated during each dwell window, the residual intra-dwell offset has zero mean (⟨δ⟩=0\langle\delta\rangle=0). The leading linear term in the expansion therefore cancels out, forcing the residual slow diffusion to enter the systematic temperature floor at second order O(σx2)O(\sigma_x^2) through the exact same curvature expression derived in Eq. (46). While in-situ tracking subtracts inter-dwell drift between strain points, this intra-dwell diffusion remains unmitigated. Thus, counting statistics improve as τ−1/2\tau^{-1/2}, whereas the intra-dwell diffusion contribution grows as τ1/2\tau^{1/2}. The total uncertainty can therefore be written as

σT2(τ)=Aτ+Bτ,(50)\sigma_T^2(\tau)=\frac{A}{\tau}+B\tau, \tag*{(50)}

which has a shallow optimum at

τ∗=τrefσTstat(τref)σTdiff(τref),σTmin⁡=2σTstat(τref)σTdiff(τref).(51)\tau^*=\tau_{\mathrm{ref}}\frac{\sigma_T^{\mathrm{stat}}(\tau_{\mathrm{ref}})}{\sigma_T^{\mathrm{diff}}(\tau_{\mathrm{ref}})},\qquad \sigma_T^{\min}=\sqrt{2\sigma_T^{\mathrm{stat}}(\tau_{\mathrm{ref}})\sigma_T^{\mathrm{diff}}(\tau_{\mathrm{ref}})}. \tag*{(51)}

The resulting precision budget is summarized in Table S2. The five terms above are intrinsic to the protocol and are what the 0.58 % of the main text quotes. The readout-contrast drift is a requirement on the apparatus, so it is listed separately: it enters only to the extent that the readout fails to meet the tolerance derived in § 14. The EiE_i values are treated as spectroscopically resolvable quantities, while Δ0\Delta_0 is obtained from the hyperbola fit; their uncertainties are included separately in Table S2. The hyperbola residuals independently test the assumed strain invariance of Δ0\Delta_0.

Point-by-point propagation of the intra-dwell wander. Equation (S40) evaluates the residual wander at a single operating point. That form is adequate when the sweep is wide and the operating point sits away from the hyperbola minimum, but it fails in two regimes. First, for a defect swept from its own minimum, the local slope ε/E\varepsilon/E vanishes there, so the single-point expression returns a zero floor and the optimum of Eq. (51) is undefined. Second, it omits the amplification of energy errors by the same ill-conditioning that governs the statistical bound, so it underestimates the floor for narrow sweeps. We therefore propagate the wander point by point. Each dwell of length τ/(Npts−1)\tau/(N_{\mathrm{pts}}-1) leaves a residual rms frequency error σν\sigma_\nu, which enters point ii through the local slope

σE,i=(εiEi)h σν(τNpts−1),(52)\sigma_{E,i}=\left(\frac{\varepsilon_i}{E_i}\right)h\,\sigma_\nu\left(\frac{\tau}{N_{\mathrm{pts}}-1}\right), \tag*{(52)}

and shifts the predicted signal by δμi=(∂μi/∂Ei) δEi\delta\mu_i=(\partial\mu_i/\partial E_i)\,\delta E_i with ∂μi/∂Ei=−b pe,i(1−pe,i)/kBT\partial\mu_i/\partial E_i=-b\,p_{e,i}(1-p_{e,i})/k_{\mathrm{B}}T. The weighted least-squares estimator maps such a perturbation onto the parameters through its influence function (JTWJ)−1JTW(J^{\mathsf T}WJ)^{-1}J^{\mathsf T}W, with the Jacobian JJ and the binomial weights WW of Eq. (14), so that

[σTdiff]2=∑i{[(JTWJ)−1JTW]T,i∂μi∂EiσE,i}2.(53)\left[\sigma_T^{\mathrm{diff}}\right]^2=\sum_i\left\{\left[(J^{\mathsf T}WJ)^{-1}J^{\mathsf T}W\right]_{T,i}\frac{\partial\mu_i}{\partial E_i}\sigma_{E,i}\right\}^2. \tag*{(53)}

Equation (S44) reduces to the single-point expression when the slope and the per-point sensitivity are flat across the sweep, and it is finite at the hyperbola minimum because the points away from the minimum carry the information. For a 0.5 GHz defect swept to 1.6 GHz it raises the total from 0.57% to 0.74% and shortens the optimum from 9.18 h to 5.5 h. All numbers quoted in this Supplemental Material use Eq. (53).

The size of the residual is fixed by the tracking protocol, which an experiment must specify and report. A residual ησν\eta\sigma_\nu gives σTmin⁡∝η1/2\sigma_T^{\min} \propto\eta^{1/2} and τ∗∝η−1\tau^* \propto\eta^{-1}: η=1/3\eta= 1/\sqrt{3} (one reference per dwell) gives 0.44% at τ∗=13.1\tau^* = 13.1 h, and η=1/12\eta= 1/\sqrt{12} (references at both ends) gives 0.31% at 26.2 h. Exact removal of the dwell mean, η→0\eta\to0, would leave only the second-order smearing of Eq. (46), and the budget would become statistics-limited with no optimum τ∗\tau^*. Tracking also costs measurement time: spending a fraction ff of the run on it scales the precision by (1−f)−1/4(1-f)^{-1/4}, so f=10%f = 10\% gives 0.60%. Finally, the equal-shot allocation makes the dwell time differ by a factor of 1.9 across the sweep; using the actual per-point dwell in place of τ/10\tau/10 changes the total from 0.581% to 0.579%.

Scope and limitations of the uncertainty budget in Table S2. Every entry is a single-realisation error for one defect swept once: it is a fit precision, of the same kind as the 0.35 % quoted for CBT in Table I of the End Matter, and not a metrological uncertainty. Three contributions that a metrological budget would carry are absent, because none of them can be estimated from the present literature: reproducibility across defects in one junction, across devices, and across cooldowns. For TLSs the first of these is not obviously small, because Δ0\Delta_0, γS\gamma_S, g0g_0 and the local phonon environment all vary by orders of magnitude across the population. A protocol that returns 1 % on one defect and disagrees with the next defect in the same barrier by 10 % is not a 1 % thermometer. Fixing the size of that term requires a measurement, which is why the main text identifies inter-comparison, rather than fit precision as the validation target. The quoted precision is what the protocol can achieve on a single defect. This determines whether the experiment is worth attempting and sets an upper bound on the accuracy ultimately achievable.

The diffusivity in Eq. (42) was measured for defects near 5.5 GHz. Its extrapolation to the sub-2 GHz range required by the protocol remains untested. Measuring it should be the first experimental step. The budget becomes statistics-limited once the rms wander falls below 2.05 MHz/h/\sqrt{\mathrm{h}}, a factor 2.4 below the measured value. Because σT2=A/τ+Bτ\sigma_T^2 = A/\tau+ B\tau with B∝D2B \propto D^2, the optimum obeys

σTmin⁡=2(AB)1/4∝D1/2,τ∗=A/B∝D−1.(54)\sigma_T^{\min} = \sqrt{2}(AB)^{1/4} \propto D^{1/2}, \qquad\tau^* = \sqrt{A/B} \propto D^{-1}. \tag*{(54)}

Thus, the precision depends only weakly on the assumed phonon-diffusion extrapolation, with DD entering the budget. The same optimization gives σTmin⁡∝T11/4\sigma_T^{\min} \propto T_1^{1/4} and τ∗∝T11/2\tau^* \propto T_1^{1/2}, because τshot=m/Γ1∝T1\tau_{\mathrm{shot}} = m/\Gamma_1 \propto T_1. Therefore, increasing T1T_1 by two orders of magnitude worsens the precision by a factor of 1001/4≃3.2100^{1/4} \simeq3.2, while increasing the optimum averaging time by a factor of 1001/2=10100^{1/2} = 10. A factor-100 increase in T1T_1 therefore gives a precision of approximately 1.8 % and pushes τ∗\tau^* to roughly 76 h. Using the independently measured D=2.2 MHz h−1/2D = 2.2\,\mathrm{MHz\,h}^{-1/2} gives a precision of 0.54% at τ∗=8.7\tau^* = 8.7 h. If the diffusivity in the operating band is twice the Klimov value, the precision becomes 0.82% at 3.8 h. At four times that value, it becomes 1.16% at 1.9 h. A suppressed phonon density of states at the operating frequency would therefore degrade protocol throughput via the T11/4T_1^{1/4} scaling derived above, while simultaneously weakening the low-frequency continuum extrapolation. Only a diffusivity an order of magnitude above the transmon-band value would put the protocol outside the range where it competes with the entries of Table I.

Voltage & defect sensitivity analysis

To justify the choice of operational parameters (Vsym=16V_{\mathrm{sym}} = 16 V and Δ0=1.3\Delta_0 = 1.3 GHz), we perform a systematic sensitivity analysis across the symmetric strain voltage VsymV_{\mathrm{sym}} and tunneling splitting Δ0\Delta_0. Table S3 summarizes the resulting relative precision σT/T\sigma_T/T evaluated at T=20T = 20 mK, 30 mK, and 50 mK. Here Vsym=2Δmax⁡/κV_{\mathrm{sym}} = 2\Delta_{\max}/\kappa is the symmetric-equivalent range; a one-branch sweep needs Vsym/2V_{\mathrm{sym}}/2, listed in the fourth column. Each entry quotes the point-by-point propagated budget. The single-operating-point formula is not shown, because it returns an artificially small floor whenever the sweep starts at the hyperbola minimum, where ∂E/∂ε→0\partial E/\partial\varepsilon\to0. The two-dimensional parameter landscape at T=20T = 20 mK is illustrated in Figure S4. Contour levels demonstrate a broad performance valley where relative errors drop below 1%. As shown in Figure S4, increasing VsymV_{\mathrm{sym}} initially yields rapid improvements in precision by widening the available spectral sweep window. Above Vsym≈20V_{\mathrm{sym}} \approx20 V the precision curves flatten, because the tracking residual of the TLS spectral diffusion takes over from counting statistics. The selected configuration, Vsym=16V_{\mathrm{sym}} = 16 V (8 V along one branch) and Δ0=1.3\Delta_0 = 1.3 GHz, lies near the knee of this curve at σT/T=0.58%\sigma_T/T = 0.58\%. A 0.5 GHz defect would do better at the same voltage (0.40%); the 1.3 GHz defect is chosen because it is the one reported.

VsymV_{\mathrm{sym}} (V)DefectSweep range (GHz)One branch (V)20 mK30 mK50 mK
6.3Δ0=0.5\Delta_0 = 0.5 GHz0.5 – 1.063.152.40%4.44%11.75%
6.3Δ0=1.3\Delta_0 = 1.3 GHz1.3 – 1.603.155.68%7.07%13.59%
10.3Δ0=0.5\Delta_0 = 0.5 GHz0.5 – 1.605.150.74%1.12%2.59%
10.3Δ0=1.3\Delta_0 = 1.3 GHz1.3 – 2.005.151.37%1.54%2.74%
16.0Δ0=0.5\Delta_0 = 0.5 GHz0.5 – 2.418.000.40%0.43%0.77%
16.0Δ0=1.3\Delta_0 = 1.3 GHz1.3 – 2.698.000.58%0.50%0.74%
20.0Δ0=0.5\Delta_0 = 0.5 GHz0.5 – 2.9910.000.35%0.31%0.46%
20.0Δ0=1.3\Delta_0 = 1.3 GHz1.3 – 3.2210.000.47%0.33%0.41%

Table S3. Sensitivity of the propagated budget to strain range and defect. Relative precision σT/T\sigma_T/T at three temperatures; each entry reoptimizes the total sweep duration τ∗\tau^*.

Attainable relative temperature precision over tunneling energy and symmetric strain range, and precision against strain range

Figure S4. (a) Attainable σT/T\sigma_T/T at 20 mK over tunneling energy and symmetric strain range, propagated model, with contours at 0.58%, 1%, 2%, and 5%. The circle marks the recommended configuration (Δ0=1.3\Delta_0 = 1.3 GHz, Vsym=16V_{\mathrm{sym}} = 16 V) and the square a 0.5 GHz comparison defect at the same voltage. (b) Precision against strain range for both defects at 20 and 30 mK; dotted curves show the single-point formula at 20 mK for comparison. Both curves flatten above about 20 V, where the systematic floor takes over from counting statistics.

Shot budget and duty cycle

Throughout the analysis, we use N∗≈8.42×106N^* \approx8.42 \times10^6 shots per point at τ∗=7.63\tau^* = 7.63 h and 11 strain points ($N_{\mathrm{tot}}^* \approx 9.27 \times 10^7 total shots). This corresponds to a shot duration

τshot=τ∗Ntot∗=τ∗NptsN∗=3.27 ms,(55)\tau_{\mathrm{shot}} = \frac{\tau^*}{N_{\mathrm{tot}}^*} = \frac{\tau^*}{N_{\mathrm{pts}}N^*} = 3.27\ \mathrm{ms}, \tag*{(55)}

which is independent of τ\tau because N∗∝τ∗N^* \propto\tau^*. This duration is consistent with the estimated defect lifetime in the required frequency range, although it is longer than the microsecond lifetimes measured at higher splittings.

At the symmetry point, reported lifetimes are 0.44 µs to 3.2 µs for defects with Δ0/h=6.2\Delta_0/h = 6.2 GHz to 7.34 GHz [28, 29]. Rescaling the same Γ1∝uE3coth⁡(E/2kBT)\Gamma_1 \propto u E^3 \operatorname{coth}(E/2k_{\mathrm{B}}T) dependence used for the per-unit-time optimum gives τshot=5/Γ1\tau_{\mathrm{shot}} = 5/\Gamma_1 between 0.01 ms and 10.0 ms over 1.3 GHz to 2.69 GHz. Thus the sweep-averaged 0.30 ms lies within the expected range. The reset time therefore contributes to the total sweep time and cannot be treated as a negligible overhead.

Reset completeness is an EE-dependent contrast

A further constraint is that the defect must be sufficiently reset between shots. The reset probability is

η(E)=1−e−τshotΓ1(E),(56)\eta(E) = 1 - e^{-\tau_{\mathrm{shot}}\Gamma_1(E)}, \tag*{(56)}

where Γ1∝uE3coth⁡(E/2kBT)\Gamma_1 \propto u E^3 \operatorname{coth}(E/2k_{\mathrm{B}}T) varies across the sweep. Therefore, a fixed τshot\tau_{\mathrm{shot}} produces an EE-dependent reset probability, which acts as an additional contrast variation and cannot in general be absorbed into the constant parameter bb.

With a fixed τshot=0.30\tau_{\mathrm{shot}} = 0.30 ms, the shortest reported symmetry-point lifetime, 0.44 µs at 6.2 GHz, gives T1=44T_1 = 44 µs to 23 µs across the budget sweep. The resulting reset probability is η=0.9989→1.0000\eta= 0.9989 \to1.0000, giving a negligible bias of +0.10%. The longest reported lifetime, 3.2 µs at 7.34 GHz, gives T1=0.53T_1 = 0.53 ms to 0.28 ms and η=0.43→0.66\eta= 0.43 \to0.66, producing a much larger bias of +18.6%. Thus the duty cycle cannot simply be fixed in advance. We set the shot duration from the local relaxation rate,

τshot,i=mΓ1(Ei).(57)\tau_{\mathrm{shot},i} = \frac{m}{\Gamma_1(E_i)}. \tag*{(57)}

This makes

ηi=1−e−m(58)\eta_i = 1 - e^{-m} \tag*{(58)}

constant across the sweep, so the reset factor is absorbed into the contrast parameter bb. The baseline analytical protocol considered in the main text adopts m=5m = 5, corresponding to a mean τshot=0.30\tau_{\mathrm{shot}} = 0.30 ms across the sweep and an optimal sweep duration of τ∗=7.63\tau^* = 7.63 h, for which the attainable statistical temperature uncertainty is σT/T=0.58%\sigma_T/T = 0.58\%, corresponding to σT=0.116\sigma_T = 0.116 mK at 20 mK. As a robustness check, we also consider the more restrictive choice m=10m = 10, which doubles the mean per-shot reset time to 0.59 ms. Re-optimizing the total acquisition time gives 0.69% at τ∗=10.8\tau^* = 10.8 h for the lifetime calibration of the budget. Bracketing the defect lifetime with the two reported values gives

short lifetime:0.50%  (τ∗=5.5 h), m=5;0.59%  (7.9 h), m=10,long lifetime:0.93%  (19.4 h), m=5;1.10%  (27.4 h), m=10.(59)\begin{aligned} \text{short lifetime:}\quad&0.50\% \;(\tau^* = 5.5\ \text{h}),\ m = 5;\quad0.59\% \;(7.9\ \text{h}),\ m = 10,\\ \text{long lifetime:}\quad&0.93\% \;(19.4\ \text{h}),\ m = 5;\quad1.10\% \;(27.4\ \text{h}),\ m = 10. \tag*{(59)} \end{aligned}

Doubling the reset overhead therefore costs about 0.1 to 0.2 percentage points.

These values depend on the experimentally observed defect. In particular, Γ1\Gamma_1 is extrapolated from u=1u = 1 and E/h=6.2E/h = 6.2 GHz to the operating band (1.3 GHz to 2.69 GHz), where uE3uE^3 is smaller by factors of 52 to 108. The relevant T1T_1 should therefore be measured at the hyperbola minimum before the sweep, not inferred from the high-frequency data.

The choice m=10m = 10 also makes the protocol less sensitive to changes in T1T_1 during a multi-hour sweep. The reset prescription in Eq. (57) therefore requires an EE-resolved T1T_1 calibration before the thermometry sweep; otherwise, strain-dependent T1T_1 drift produces an EE-dependent reset contrast that cannot be absorbed into the single contrast parameter bb. At m=5m = 5, doubling T1T_1 produces a −1.4%-1.4\% bias, and tripling it gives −3.6%-3.6\%. These biases fall to −0.03%-0.03\% and below 0.25%0.25\% for m=8m = 8 and m=10m = 10, respectively. Increasing mm from 5 to 10 raises the statistical uncertainty only as m1/4m^{1/4}, from 0.58% to 0.69% for the fast branch and from 1.48% to 1.76% for the slow branch.

Strain range and voltage calibration

Reaching Emax⁡E_{\max} from a tunneling minimum at Δ0\Delta_0 requires an asymmetry excursion ε=Emax⁡2−Δ02\varepsilon= \sqrt{E_{\max}^2 - \Delta_0^2}. Along one branch of the hyperbola this needs a voltage excursion ε/r\varepsilon/r, with r=295r = 295 MHz V−1^{-1} the measured tuning rate of the asymmetry of the Δ0/h=1.3\Delta_0/h = 1.3 GHz defect [30]; a symmetric sweep over both branches needs

Vsym=2εr.(60)V_{\mathrm{sym}} = \frac{2\varepsilon}{r}. \tag*{(60)}

For Emax⁡/h=2.69E_{\max}/h = 2.69 GHz, ε/h=2.36\varepsilon/h = 2.36 GHz, so one branch needs 8.0 V and both need 16 V. The symmetry point of this defect lies at −14.05-14.05 V [30]. The one-branch sweep therefore runs from −14.05-14.05 V to about −6-6 V, inside the range of the reported measurements, whereas a symmetric sweep would need −22-22 V to −6-6 V, beyond them. We use the one-branch sweep; sweeping both branches doubles the voltage range and adds the evenness test of the symmetry properties in the main text. The extended sweep to 3.22 GHz needs ε/h=2.95\varepsilon/h = 2.95 GHz, that is 10 V along one branch.

Readout-contrast drift along the sweep

Equation (S5) assumes that the offset aa and contrast bb are constant across the strain sweep. This assumption must therefore be checked. The transverse weight u=(Δ0/E)2u = (\Delta_0/E)^2 decreases from 1 at the hyperbola minimum (Δ0/h=1.3\Delta_0/h = 1.3 GHz) to 0.233 at E/h=2.69E/h = 2.69 GHz. Since g=g0ug = g_0\sqrt{u}, the swap time increases by a factor of 2.07 across the sweep. Relaxation of the readout qubit during the swap can therefore make the readout contrast, and hence the fitted parameter bb, depend on EE. A barrier defect tuned through a readout resonator can also strongly reduce the readout signal [28].

We model a small contrast drift as

b(E)=b0[1+κ(E−Eˉ)Eˉ],(61)b(E) = b_0 \left[1+\frac{\kappa(E-\bar{E})}{\bar{E}}\right], \tag*{(61)}

generate the signal with this EE-dependent contrast, and fit it with the constant-bb three-parameter model. The resulting temperature bias is approximately linear in κ\kappa. We adopt ∣κ∣≤0.011|\kappa| \le0.011 as the contrast-drift allowance, corresponding to a temperature bias of about 0.22% over the full sweep. This allowance is kept separate from the intrinsic 0.58% statistical precision budget and is included only as an apparatus requirement. The numerical results are shown in Table S4.

κ\kappaδT/T\delta T/T
0.0030.059%
0.0110.215%
0.0300.593%

Table S4. Bias in the recovered temperature caused by a linear readout-contrast drift across the 1.3 GHz to 2.69 GHz sweep.

The contrast requirement can be related directly to the readout-qubit lifetime. The swap time is

tswap=π2g,g=g0u.(62)t_{\mathrm{swap}} = \frac{\pi}{2g},\qquad g = g_0\sqrt{u}. \tag*{(62)}

For the coupling g0/2π=3.0g_0/2\pi= 3.0 MHz reported for this defect [30], the swap time increases from 83 ns at the hyperbola minimum to 172 ns at 2.69 GHz (from 4.2 ns to 8.6 ns at the 60 MHz assumed in earlier versions). The visibility contains the factor

b(E)=b0exp⁡[−tswap(E)T1,q].(63)b(E) = b_0 \exp\left[-\frac{t_{\mathrm{swap}}(E)}{T_{1,q}}\right]. \tag*{(63)}

Requiring the adopted allowance ∣κ∣≤0.011|\kappa| \le0.011 corresponds to approximately

T1,q≥11.6 μs.(64)T_{1,q} \ge11.6\,\mu\mathrm{s}. \tag*{(64)}

A parametric swap runs at most at about 0.58 gg, the maximum of the Bessel factor J1J_1, which raises the requirement to about 20 μ\mus. Both values are within reach of modern transmons and fluxonium. The phase qubits of the earlier barrier-defect measurements, with T1,q≈100T_{1,q} \approx100 ns [28], could not read this defect out by swap, because the swap alone lasts 83 ns to 172 ns.

The drift must be controlled, not fitted

The linear model in Eq. (61) is only a convenient way to quantify the tolerance. The physical contrast is nonlinear because tswap∝1/ut_{\mathrm{swap}} \propto1/\sqrt{u}. Using the full visibility

b(E)=b0exp⁡[−tswap(E)T1,q](65)b(E) = b_0 \exp\left[-\frac{t_{\mathrm{swap}}(E)}{T_{1,q}}\right] \tag*{(65)}

gives, at g0/2π=3.0g_0/2\pi= 3.0 MHz, temperature biases of −20%-20\% at T1,q=100T_{1,q} = 100 ns, −2.4%-2.4\% at 1 μ\mus, −0.25%-0.25\% at 10 μ\mus and −0.12%-0.12\% at 20 μ\mus (−1.2%-1.2\%, −0.30%-0.30\% and −0.012%-0.012\% at 100 ns, 0.42 μ\mus and 10 μ\mus for 60 MHz). It requires T1,q≥11.6 μT_{1,q} \ge11.6\,\mus$ for the ∣κ∣≤0.011|\kappa| \le0.011 allowance, the same as the linear model.

Fitting the contrast drift as an additional free parameter is not an effective alternative. Promoting T1,qT_{1,q} to a fourth fit parameter makes the Fisher matrix nearly singular because the contrast variation and thermal population have similar dependence on EE. At T1,q=20 μT_{1,q} = 20\,\mus and g0/2π=3.0g_0/2\pi= 3.0 MHz, the statistical uncertainty increases by a factor of about 100, so the counting term alone would rise from 0.41% to about 40%. The contrast should therefore be controlled experimentally, not fitted.

An in situ check is to repeat the sweep at different readout amplitudes and verify that the extracted temperature is unchanged. This directly tests the readout assumption without calibrating the thermometer’s energy axis.

Mean-force Gibbs correction

Equation (7) of the main text assumes that the defect is in the bare Gibbs state of HTLSH_{\mathrm{TLS}}. However, a system coupled to a bath at finite strength relaxes to the mean-force Gibbs state [36],

τMF=Tr⁡Be−βHTr⁡e−βH,(66)\tau_{\mathrm{MF}} = \frac{\operatorname{Tr}_{B}e^{-\beta H}}{\operatorname{Tr}e^{-\beta H}}, \tag*{(66)}

where

H=HTLS+HB+τz⊗B^.(67)H = H_{\mathrm{TLS}} + H_{B} + \tau_z \otimes\hat{B}. \tag*{(67)}

Equation (S57) differs from e−βHTLS/Ze^{-\beta H_{\mathrm{TLS}}}/Z at finite coupling. Because the protocol is intended to be candidate primary, this correction is important: a coupling-dependent distortion of pe(E)p_e(E) could otherwise introduce a material-dependent parameter.

Second-order result

Expanding Eq. (66) to second order in the system–bath coupling gives

Tr⁡Be−βH=ZBe−βHTLS[1+K+O(λ4)](68)\operatorname{Tr}_{B}e^{-\beta H}=Z_{B}e^{-\beta H_{\mathrm{TLS}}}\left[1+K+O(\lambda^{4})\right] \tag*{(68)}

, with

K=∫0βdτ1∫0τ1dτ2 A~(τ1)A~(τ2)C(τ1−τ2),(69)K=\int_{0}^{\beta}d\tau_{1}\int_{0}^{\tau_{1}}d\tau_{2}\,\widetilde{A}(\tau_{1})\widetilde{A}(\tau_{2})C(\tau_{1}-\tau_{2}), \tag*{(69)}

where

A~(τ)=eτHTLSτze−τHTLS(70)\widetilde{A}(\tau)=e^{\tau H_{\mathrm{TLS}}}\tau_z e^{-\tau H_{\mathrm{TLS}}} \tag*{(70)}

and

C(τ)=⟨B~(τ)B~(0)⟩B=∫0∞dω J(ω)[(n+1)e−ωτ+neωτ].(71)C(\tau)=\langle\widetilde{B}(\tau)\widetilde{B}(0)\rangle_B=\int_{0}^{\infty}d\omega\,J(\omega)\left[(n+1)e^{-\omega\tau}+ne^{\omega\tau}\right]. \tag*{(71)}

In the energy eigenbasis, the diagonal elements are

Kmm=∑k∣Amk∣2I(Em−Ek),(72)K_{mm}=\sum_k\lvert A_{mk}\rvert^2 I(E_m-E_k), \tag*{(72)}

where

I(Δ)=∫0βds (β−s)eΔsC(s)=∫0∞dω J(ω)[(n+1)g(Δ−ω)+ng(Δ+ω)],(73)\begin{aligned} I(\Delta)&=\int_{0}^{\beta}ds\,(\beta-s)e^{\Delta s}C(s)\\ &=\int_{0}^{\infty}d\omega\,J(\omega)\left[(n+1)g(\Delta-\omega)+ng(\Delta+\omega)\right], \tag*{(73)} \end{aligned}

with

g(z)=eβz−1−βzz2,g(0)=β22.(74)g(z)=\frac{e^{\beta z}-1-\beta z}{z^2},\qquad g(0)=\frac{\beta^2}{2}. \tag*{(74)}

Using Eq. (11),

∣Aee∣2=∣Agg∣2=1−u,∣Aeg∣2=u,(75)\lvert A_{ee}\rvert^2=\lvert A_{gg}\rvert^2=1-u,\qquad\lvert A_{eg}\rvert^2=u, \tag*{(75)}

which separates the longitudinal and transverse contributions to the TLS dynamics, as illustrated in Figure S5. The population correction is therefore

Eigenvalue spectrum and the frequency bands each level contributes

Figure S5. Eigenvalue spectrum and the frequency bands each level contributes. (a) Four-level TLS–qubit spectrum versus strain (Δ0/h=1.3 GHz\Delta_0/h = 1.3\,\mathrm{GHz}, g0/2π=3.0 MHzg_0/2\pi= 3.0\,\mathrm{MHz}, ωq/2π=5.0 GHz\omega_q/2\pi= 5.0\,\mathrm{GHz}). (b) Avoided-crossing gap 2g0u2g_0\sqrt{u} and pure dephasing ∝1−u\propto1-u (the Gaussian channel, which is the measured echo form; see text): readout is strongest where dephasing vanishes. (c) Frequencies at which the environment is sampled: the splitting E/hE/h (relaxation), ω→0\omega\to0 (dephasing), and the dressed branches ω±\omega_{\pm} of Eq. (84)

δpe=pe(1−pe)u[I(E)−I(−E)],δTT=u[I(E)−I(−E)]x.(76)\begin{aligned} \delta p_e &= p_e(1-p_e)u\left[I(E)-I(-E)\right],\\ \frac{\delta T}{T} &= \frac{u\left[I(E)-I(-E)\right]}{x}. \tag*{(76)} \end{aligned}

Thus, only the transverse weight uu shifts the populations. We therefore compute only the diagonal elements KmmK_{mm} for the temperature bias.

Reference [36] also shows that the second-order mean-force state can retain coherences in the energy eigenbasis through terms coupling eigenoperators at different Bohr frequencies. Here, these frequencies are 0 and ±E\pm E, so such terms are generally nonzero. They do not enter the present budget because the swap readout of Eq. (7) of the main text measures the population, not the energy-basis coherences. The mean-force correction is therefore controlled by the same internal coordinate uu that already determines the relative weights of relaxation, dephasing, and readout in the main text. It is largest near the degeneracy point, where uu is largest.

Calibration of the bath

We assume a super-Ohmic (Debye) spectrum, J(ω)=Aω3J(\omega)=A\omega^3, consistent with the relaxation rate Γ1∝uE3coth⁡(E/2kBT)\Gamma_1 \propto uE^3\coth(E/2k_{\mathrm{B}}T). We determine AA from the measured symmetry-point lifetimes using the golden-rule relation

Γ1=2πuJ(E)coth⁡(βE2).(77)\Gamma_1=2\pi uJ(E)\coth\left(\frac{\beta E}{2}\right). \tag*{(77)}

The reported lifetimes of 0.44 μ\mus to 3.2 μ\mus at Δ0/h=6.2\Delta_0/h=6.2 GHz to 7.34 GHz [28, 29] correspond to values of the dimensionless coupling

α≡J(E∗)kBT/ℏ,(78)\alpha\equiv\frac{J(E^*)}{k_{\mathrm{B}}T/\hbar}, \tag*{(78)}

defined in the scaled frequency units of § 1, between 4.8×10−84.8\times10^{-8} and 5.8×10−75.8\times10^{-7} at the operating point.

Estimator bias

The observable that matters is not δpe\delta p_e but the bias in TT returned by the estimator of § 2. We therefore evaluate Eq. (76) at each point of the 1.3 GHz to 2.69 GHz sweep and fit the result with the same S=a+bpeS=a+bp_e model. Table S5 reports the outcome as a function of the bath bandwidth.

Bath band ω/2π<\omega/2\pi<strongest (α=5.8×10−7\alpha= 5.8 \times10^{-7})weakest (α=4.8×10−8\alpha= 4.8 \times10^{-8})
1.7 GHz2.6×10−62.6 \times10^{-6}2.2×10−72.2 \times10^{-7}
3.3 GHz2.3×10−52.3 \times10^{-5}1.9×10−61.9 \times10^{-6}
6.7 GHz2.0×10−42.0 \times10^{-4}1.6×10−51.6 \times10^{-5}
13.3 GHz1.2×10−31.2 \times10^{-3}1.0×10−41.0 \times10^{-4}

Table S5. Bias in the recovered temperature from the mean-force correction, after fitting the full sweep with the three-parameter estimator, for the strongest and weakest coupling consistent with the measured symmetry-point lifetimes. The systematic floor for comparison is 0.082 mK.

Even at the widest band and the most strongly coupled defect, the bias is 68×\times below the tracking residual; at the weakest coupling it is 820 times below the 0.082 mK residual. The same code reproduces the values of the earlier 0.5 GHz to 1.6 GHz table to within a factor of 1.5.

Numerical validation

Equation (S67) is checked against an independent, non-perturbative calculation: exact diagonalisation of HH for a discretised bath, followed by construction of the exact thermal state and partial trace over the bath. Because the physical α\alpha is far too small to resolve against double precision, the coupling is amplified by a factor ss and the quadratic scaling is verified. Two truncation errors must be controlled.

Fock truncation. The thermal occupation n(ω)=(eϵω−1)−1n(\omega)=(e^{\epsilon\omega}-1)^{-1} reaches 19.5 at ω/ωref=0.05\omega/\omega_{\mathrm{ref}}=0.05, so an infrared mode needs ∼120\sim120 Fock states. Restricting the exact diagonalisation to ω/ωref∈[1,8]\omega/\omega_{\mathrm{ref}}\in[1,8], where n≤0.38n\leq0.38, the ratio of the exact result to Eq. (76) evaluated on the same discrete modes is 0.99716, 0.99776, 0.99792, 0.99797 for Fock cutoffs M=4,5,6,7M=4,5,6,7. The second-order expression is thus correct to 0.2%.

Mode discretisation. A coarse logarithmic mode grid overestimates the continuum. With the same ω\omega band, the ratio of the discrete-mode evaluation of Eq. (76) to the continuum quadrature converges from 1.42 at N=5N=5 to 1.0000 at N=2000N=2000; all quoted numbers use the continuum quadrature.

Finally, the bias is carried by modes resonant with the splitting, not by the infrared tail: modes with ω>2ωref\omega>2\omega_{\mathrm{ref}} carry 97.9% of the effect. This is the same statement as the uu weighting in Eq. (76), seen from the bath side.

The band cutoff is a modelling choice

The bias in Table S5 grows with the upper band edge, indicating that the choice of cutoff becomes increasingly important. The numerical ratios for successive factor-of-two widenings are 8.7, 8.6 and 6.2, corresponding to effective exponents of 3.3, 3.0 and 2.7, which approach the expected quadratic scaling from above. Indeed, for ω≫E,kBT/ℏ\omega\gg E,k_{\mathrm{B}}T/\hbar,

g(E−ω)−g(−E−ω)≃2βEω2.(79)\mathcal{g}(E-\omega)-\mathcal{g}(-E-\omega)\simeq\frac{2\beta E}{\omega^2}. \tag*{(79)}

so with J(ω)=Aω3J(\omega) = A\omega^3 the integrand scales as 2AβEω2A\beta E\omega, giving

I(E)−I(−E)∝AβEΛ2.(80)I(E) - I(-E) \propto A\beta E\Lambda^2. \tag*{(80)}

Thus, the residual bias scales as Λ2\Lambda^2 rather than Λ4\Lambda^4. Extending J(ω)=Aω3J(\omega) = A\omega^3 with the above AA to the Debye frequency is unphysical. The resulting Debye–Waller exponent reaches ∼130\sim130, driving Δ0\Delta_0 to zero. The resolution is standard tunneling-model physics: the high-frequency phonons are already absorbed into the measured Δ0\Delta_0 while the hyperbola fit reabsorbs them. We confirmed this directly: widening the bath band in the exact diagonalization moved the recovered Δ0\Delta_0 from −19%-19\% to −29%-29\%. In contrast, the recovered temperature barely moved, because the fit absorbs a dressed Δ0\Delta_0 by construction. What Table S5 reports is the residual, non-absorbable distortion of the shape of pE(E)p_E(E).

The Debye-frequency check is consistent with this quadratic scaling: scaling the 13.3 GHz entry of Table S5 by (8900/13.3)2(8900/13.3)^2 gives ∼0.5\sim0.5 K, which already lies outside perturbation theory. This is the one place where the analysis rests on a modeling assumption instead of a bound.

There is a second reason to be careful with the cutoff. The relaxation data used to fix AA are measured at 6 GHz to 13 GHz, where the phonon spectrum of the junction is demonstrably not smooth: the expected Γ1∝Δ02/E2\Gamma_1 \propto\Delta_0^2/E^2 scaling is obscured by structure in the noise spectral density, and three of five defects share a common relaxation maximum near 7.4 GHz, consistent with coupling to the same confined phonon mode [28]. That interpretation turns on the junction being comparable to a phonon wavelength, which at 7.4 GHz it is (λ≈0.9 μm\lambda\approx0.9\,\mu\mathrm{m} for longitudinal phonons in Al, against a ∼1 μm\sim1\,\mu\mathrm{m} junction). Over the 1.3 GHz to 2.69 GHz sweep the wavelength is 4.9 μ\mum to 2.4 μ\mum, two to five times the junction, so the continuum J(ω)=Aω3J(\omega) = A\omega^3 is being extrapolated from a band where it visibly fails into a band where the junction cannot confine a mode at all, or where phononic environment modification strongly alters the local acoustic density of states [49]. A defect-specific measurement of the phonon spectral density in the sub-2 GHz band would remove both assumptions at once, and is named in the main text as the second quantity an experiment must fix.

Symmetry checks

Figure S5 shows the four-level TLS–qubit spectrum and the strain-dependent quantities used throughout the analysis. The three symmetry statements of the main text are verified against exact diagonalisation: (i) the unitary equivalence

HTLS(−ε)=τxHTLS(ε)τx,(81)H_{\mathrm{TLS}}(-\varepsilon) = \tau_x H_{\mathrm{TLS}}(\varepsilon)\tau_x, \tag*{(81)}

which gives an even spectrum in ε\varepsilon; (ii) at ε=0\varepsilon= 0,

{HTLS,τz}=0,(82)\{H_{\mathrm{TLS}},\tau_z\} = 0, \tag*{(82)}

and the diagonal elements of τz\tau_z in the energy eigenbasis vanish which forbids pure dephasing by symmetry; and (iii) the avoided-crossing gap in the four-level TLS–qubit spectrum is

Δgap=2g0u,u=(Δ0E)2,(83)\Delta_{\mathrm{gap}} = 2g_0\sqrt{u}, \qquad u = \left(\frac{\Delta_0}{E}\right)^2, \tag*{(83)}

which agrees with the exact numerical spectrum at the tested strain points (see Figure S5).

The same parameter uu determines the transverse TLS–qubit coupling, g=g0ug = g_0\sqrt{u}, while the longitudinal contribution scales as 1−u1-u. Thus, near the hyperbola minimum, u≃1u \simeq1 and both the transverse coupling and the mean-force temperature bias are maximal, whereas away from the minimum the transverse coupling and associated bias decrease and the longitudinal contribution increases. Figure S5(c) shows the corresponding TLS and hybridized frequencies

ω±=E/ℏ+ωq2±12(E/ℏ−ωq)2+4g02u,(84)\omega_{\pm} = \frac{E/\hbar+ \omega_q}{2} \pm\frac{1}{2}\sqrt{(E/\hbar- \omega_q)^2 + 4g_0^2u}, \tag*{(84)}

together with the E/ℏE/\hbar and ω→0\omega\to0 channels relevant to the relaxation, dephasing, and mean-force analyses.

References

  1. [1]A. Kolosovtov, N. Gusarov, V. Slepnev, M. Perelstein, V. Sevriuk, A. Gubaydullin, and V. Vinokur, npj Quantum Inf. 12, 146 (2026).
  2. [2]P. W. Anderson, B. I. Halperin, and C. M. Varma, Philos. Mag.-Theor. Exp. Appl. Phys. 25, 1 (1972).
  3. [3]W. A. Phillips, J. Low Temp. Phys. 7, 351 (1972).
  4. [4]J. M. Martinis, K. B. Cooper, R. McDermott, M. Steffen, M. Ansmann, K. D. Osborn, K. Cicak, S. Oh, D. P. Pappas, R. W. Simmons, and C. C. Yu, Phys. Rev. Lett. 95, 210503 (2005).
  5. [5]C. Müller, J. H. Cole, and J. Lisenfeld, Rep. Prog. in Phys. 82, 124501 (2019).
  6. [6]S. E. de Graaf et al., Sci. Adv. 6, eabc5055 (2020).
  7. [7]A. V. Feshchenko, L. Casparis, I. M. Khaymovich, D. Maradan, O.-P. Saira, M. Palma, M. Meschke, J. P. Pekola, and D. M. Zumbühl, Phys. Rev. Appl. 4, 034001 (2015).
  8. [8]J. P. Pekola, K. P. Hirvi, J. P. Kauppinen, and M. A. Paalanen, Phys. Rev. Lett. 73, 2903 (1994).
  9. [9]D. I. Bradley et al., Nat. Commun. 7, 10455 (2016).
  10. [10]E. Praks, L. Wang, Y.-C. Chang, R. P. Loreto, M. Prunnila, J. T. Peltonen, and J. P. Pekola, Electronic primary thermometry – experimental comparison of the coulomb blockade and shot noise thermometer (2026), arXiv:2606.04147 [cond-mat.mes-hall].
  11. [11]M. Meschke, J. Engert, D. Heyer, and J. P. Pekola, Int. J. Thermophys. 32, 1378 (2011).
  12. [12]X. Y. Jin, A. Kamal, A. P. Sears, T. Gudmundsen, D. Hover, J. Miloshi, R. Slattery, F. Yan, J. Yoder, T. P. Orlando, S. Gustaysson, and W. D. Oliver, Phys. Rev. Lett. 114, 240501 (2015).
  13. [13]M. Spiecker et al., Nat. Phys. 19, 1320 (2023).
  14. [14]J. Wheeler, M. R. Vissers, M. Malnou, J. Hubmayr, J. N. Ullom, and J. Gao, App. Phys. Lett. 117, 192601 (2020).
  15. [15]F. Ye, L. S. Dhami, and J. M. Nichol, Phys. Rev. Appl. 25, 024045 (2026).
  16. [16]P. V. Klimov, J. Kelly, Z. Chen, M. Neeley, A. Megrant, B. Burkett, R. Barends, K. Arya, B. Chiaro, Y. Chen, A. Dunsworth, A. Fowler, B. Foxen, C. Gidney, M. Giustina, R. Graff, T. Huang, E. Jeffrey, E. Lucero, J. Y. Mutus, O. Naaman, C. Neill, C. Quintana, P. Roushan, D. Sank, A. Vainsencher, J. Wenner, T. C. White, S. Boixo, R. Babbush, V. N. Smelyanskiy, H. Neven, and J. M. Martinis, Phys. Rev. Lett. 121, 090502 (2018).
  17. [17]S. Schlör, J. Lisenfeld, C. Müller, A. Bilmes, A. Schneider, D. P. Pappas, A. V. Ustinov, and M. Weides, Phys. Rev. Lett. 123, 190502 (2019).
  18. [18]J. Lisenfeld, C. Müller, J. H. Cole, P. Bushev, A. Lukashenko, A. Shnirman, and A. V. Ustinov, Phys. Rev. Lett. 105, 230504 (2010).
  19. [19]A. Kulikov, R. Navarathna, and A. Fedorov, Phys. Rev. Lett. 124, 240501 (2020).
  20. [20]D. S. Lvov, S. A. Lemziakov, E. Ankerhold, J. T. Peltonen, and J. P. Pekola, Phys. Rev. Appl. 23, 054079 (2025).
  21. [21]G. J. Grabovskij, T. Peichl, J. Lisenfeld, G. Weiss, and A. V. Ustinov, Science 338, 232 (2012).
  22. [22]J. Lisenfeld, A. Bilmes, A. Megrant, R. Barends, J. Kelly, P. Klimov, G. Weiss, J. M. Martinis, and A. V. Ustinov, npj Quantum Inf. 5, 105 (2019).
  23. [23]M. Mehboudi, A. Sanpera, and L. A. Correa, J. Phys. A: Math. Theo. 52, 303001 (2019).
  24. [24]A. J. Leggett, S. Chakravarty, A. T. Dorsey, M. P. A. Fisher, A. Garg, and W. Zwger, Rev. Mod. Phys. 59, 1 (1987).
  25. [25]A. Bilmes, S. Volosheniuk, J. D. Brehm, A. V. Ustinov, and J. Lisenfeld, npj Quantum Inf. 7, 27 (2021).
  26. [26]G. Catelani and D. M. Basco, SciPost Phys., Phys. 013 (2019).
  27. [27]L. Glazman and G. Catelani, SciPost Phys. Lect. Notes , 31 (2021).
  28. [28]J. Lisenfeld, A. Bilmes, S. Matityahu, S. Zanker, M. Marthaler, M. Schechter, G. Schön, A. Shnirman, G. Weiss, and A. V. Ustinov, Sci. Rep. 6, 23786 (2016).
  29. [29]C. Müller, J. Lisenfeld, A. Shnirman, and S. Poletto, Phys. Rev. B 92, 035442 (2015).
  30. [30]J. Lisenfeld, G. J. Grabovskij, C. Müller, J. H. Cole, G. Weiss, and A. V. Ustinov, Nat. Commun. 6, 6182 (2015).
  31. [31]Supplemental material, supplemental material for units and estimator conventions, spectral-diffusion budget, shot budget, and mean-force analysis.
  32. [32]Y. Shalibo, Y. Rofe, D. Shwa, F. Zeides, M. Neeley, J. M. Martinis, and N. Katz, Phys. Rev. Lett. 105, 177001 (2010).
  33. [33]K. B. Cooper, M. Steffen, R. McDermott, R. W. Simmonds, S. Oh, D. A. Hite, D. P. Pappas, and J. M. Martinis, Phys. Rev. Lett. 93, 180401 (2004).
  34. [34]L. A. Correa, M. Mehbourdi, G. Adesso, and A. Sanpera, Phys. Rev. Lett. 114, 220405 (2015).
  35. [35]A. Mandarino, M. G. A. Paris, and C. Benedetti, Thermometry with multilevel transmon probes (2026), arXiv:2606.30302 [quant-ph].arxiv.org/abs/2606.30302
  36. [36]J. D. Cresser and J. Anders, Phys. Rev. Lett. 127, 250601 (2021).
  37. [37]F. Albarelli, M. G. A. Paris, B. Vacchini, and A. Smirne, Phys. Rev. A 108, 062421 (2023).
  38. [38]J. Rubio, J. Anders, and L. A. Correa, Phys. Rev. Lett. 127, 190402 (2021).
  39. [39]S. Weeden, D. Harrison, S. Patel, M. Snyder, E. Blackwell, G. Spahn, S. Abdullah, Y. Takeda, B. Plourde, J. Martinis, and R. McDermott, Phys. Rev. Appl. 25, 044050 (2026).
  40. [40]J. J. Burnett, A. Bengtsson, M. Scigliuzzo, D. Niepce, M. Kudra, P. Delsing, and J. Bylander, npj Quantum Inf. 5, 54 (2019).
  41. [41]F. Berritta, D. Pahl, L. Pahl, W. P. Banner, G. Cutter, J. A. Krzywda, S. Weeden, S. Patel, P. Buttles, S. Eilhart, M. Gingras, B. M. Niedzielski, R. McDermott, M. E. Schwartz, K. Serniak, M. Hays, J. A. Grover, and W. D. Oliver, Adaptive spectroscopy of fast two-level-system dynamics in superconducting qubits (2026), arXiv:2608.02086 [quant-ph].arxiv.org/abs/2608.02086
  42. [42]U. Vool et al., Phys. Rev. Lett. 113, 247001 (2014).
  43. [43]H. Nho, T. Connolly, P. D. Kurilovich, S. Diamond, C. G. L. Böttcher, L. I. Glazman, and M. H. Devoret, Phys. Rev. Lett. 136, 050601 (2026).
  44. [44]A. P. Vepsäläinen et al., Nature 584, 551 (2020).
  45. [45]T. Thorbeck, A. Eddins, E. Lauer, D. T. McClure, and M. Carroll, PRX Quantum 4, 020356 (2023).
  46. [46]M. McEwen et al., Nat. Phys. 18, 107 (2022).
  47. [47]E. O. Lachman, D. P. Pappas, J. M. Marshall, and J. Y. Mutus, Localized thermometry via dayem bridges integrated on superconducting qubit chips (2026), arXiv:2607.06670 [quant-ph].arxiv.org/abs/2607.06670
  48. [48]C. J. Yeager, S. S. Courts, and W. E. Davenport, AIP Conf. Proc. 613, 1644 (2002).
  49. [49]M. Chen, J. C. Owens, H. Putterman, M. Schäfer, and O. Painter, Sci. Adv. 10, eado6240 (2024).

Paper details

vlnc:W_sanvcrpc
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 7

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents