{"id":"44e47d8f-d641-4664-a553-6a446273fa29","arxiv_id":"2507.14108","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A valley excitation in a singlet-triplet qubit can act as a colocated, dispersive charge-noise sensor with estimated sub-millisecond response times.","lead":"This paper proposes a new way to measure charge noise in silicon spin qubits while they are running, by using an extra valley energy state of the electrons as a built-in sensor coupled to a microwave resonator. If it works, operators could continuously monitor and correct slow drifts in the qubit's exchange coupling, a major source of error in these devices.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Internal inconsistency in valley T1: text claims >10 ms but cites 10–100 kHz relaxation rate; if the shorter lifetime applies, the 13 µs and 1 ms measurement claims fail.","rationale":"I conditioned on the T1 contradiction after checking the strongest claim, namely the 13 µs measurement time and concurrent qubit operation. The numerical simulation (Fig. 2) and the SNR derivation Eq. (30) are internally plausible: the stochastic master equation results match the predicted t_m, and the spin-coherence bound O(θ^2)=4×10^-4 is derived cleanly in the no-decay subspace. However, all of this assumes the HV degree of freedom stays in the singly-excited valley subspace. The paper's own text gives two incompatible lifetimes for the same reference. If the 10–100 kHz rate is correct, T1 is on the order of the integration time, so the probe can decay mid-measurement and the stated 'can exceed 10 ms' support disappears. The reader's weakest assumption pointed to the same lifetime, but framed it mainly as a transfer issue; my concern is stronger because the manuscript contradicts itself on the quoted value. A conditional verdict remains appropriate: the issue is checkable by looking up Ref. [24] and rerunning the SNR with decay. If the reference supports 10 ms, the concern evaporates; if it supports 10–100 µs, the headline must be revised. I do not see a more load-bearing problem: the numerical SNR calibration is a genuine independent check, and the δg/g≈δJ/J approximation is a secondary quantitative worry that would change times by a factor rather than invalidate the protocol.","tokens_in":17267,"tokens_out":15437,"duration_ms":589259,"concrete_test":"Retrieve the actual valley relaxation time reported in Ref. [24] (N. E. Penthorn et al., Phys. Rev. Appl. 14, 054015 (2020)) and resolve the 10 ms vs 10–100 kHz discrepancy. Then add the corresponding T1 Lindblad term, e.g., Γρ = (1/T1)D[|v−⟩⟨v+|](ρ), to the master equation Eq. (16) and recompute the integrated SNR at t=13 µs and at t=1 ms using Table I parameters. If the SNR at 13 µs drops below 1, or if the signal is frequently interrupted by decay before t_m, the sub-millisecond sensing claim is not supported as stated.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Central claim depends on the spectator valley state remaining coherent during the measurement window. The paper is internally contradictory about this lifetime: Sec. I says 'valley relaxation times in silicon can exceed 10 ms [24]' and Sec. III B repeats '~10 ms', but Sec. II C states 'valley relaxation rate∼10–100kHz [24]'. A 10–100 kHz rate means T1≈10–100 µs, i.e., 100–1000× shorter than 10 ms. If the short value is the one reported in Ref. [24], the headline t_shot_m=13 µs calculation (Eq. (30)) is invalid: a probe with T1=10 µs decays with probability 73% before the integration is complete, and even with T1=100 µs the decay probability is 12%; continuous monitoring over 1 µs–1 ms as claimed is impossible without frequent repumping, which interrupts the concurrent-operation premise. This is not merely a transfer assumption from a different device; it is an unresolved conflict in the manuscript's own quoted numbers for the same reference. Since Ref. [24] is a single-electron measurement, using its value in an ST qubit also needs justification, but the first step is to determine which of the two quoted numbers is correct.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proposes a charge-noise sensor based on a silicon singlet-triplet (ST) qubit in which one electron is prepared in an excited valley state. The valley excitation acts as a spectator two-level system coupled to a resonator through the exchange interaction, so that charge-noise-induced fluctuations in exchange and valley splitting produce a measurable dispersive shift. The authors derive the effective Hamiltonian, compute the signal-to-noise ratio of a homodyne readout, estimate measurement times from device parameters (13 micro-seconds with a quantum-limited amplifier, 9 ms with a cryogenic amplifier, and sub-ms with optimized resonator parameters), show that the probe perturbs the spin qubit only at O(theta^2), and validate the SNR formula with a stochastic master-equation simulation. The central claim is that sub-millisecond, in-situ, colocated charge-noise detection is feasible while the qubit continues to operate.","tokens_in":17572,"tokens_out":23833,"duration_ms":226244,"significance":"If the underlying assumptions hold, the proposal is a conceptually attractive route to real-time charge-noise monitoring for exchange-based silicon qubits, with the sensor colocated on the same pair of electrons and no additional sensor dot. The paper contains a complete analytic derivation of the probe Hamiltonian, a transparent SNR analysis, and a numerical simulation, which are strengths. The parameter estimates are grounded in experimental references rather than fitted to produce a small measurement time. However, the significance currently depends on two load-bearing points that need clarification: the valley relaxation time in the proposed operating configuration, and the consistency between the simulated and the analytic signal (delta_chi/chi).","major_comments":[{"comment":"The manuscript is internally contradictory about the valley relaxation time. Section I states 'valley relaxation times in silicon can exceed 10 ms [24]' and Sec. III B repeats '~10 ms', but Sec. II C states 'valley relaxation rate~10–100kHz [24]', which corresponds to T1 = 10–100 micro-seconds. If the 10–100 kHz rate is the correct one for this system, the 13 micro-second measurement time is invalid because the spectator valley state decays with probability 73% (for T1 = 10 micro-seconds) or 12% (for T1 = 100 micro-seconds) during the integration, and continuous monitoring over 1 micro-second to 1 ms as claimed is impossible without frequent repumping, which would interrupt concurrent operation. The authors must resolve this discrepancy and either justify T1 > 10 ms in the pulsed ST configuration or revise the measurement-time claims to account for valley decay and repumping overhead.","section":"Sec. I, Sec. II C, Sec. III B"},{"comment":"The simulation that is presented as numerical confirmation of Eq. (30) uses delta_chi/chi = 0.13, but Eq. (36) with the paper's own inputs (delta_J/J = 3% and delta_Delta/Delta_rv ~ 0.5%) gives delta_chi/chi ~ 6%. The simulated signal is therefore about a factor of 2.2 larger than the realistic estimate, and the simulation does not validate the 13 micro-second measurement time for the parameters of Table I. The authors should either rerun the simulation with delta_chi/chi consistent with Eq. (36), or provide a clear justification for why 0.13 is an appropriate fluctuation amplitude in the simulation.","section":"Sec. IV A and Fig. 2, with Eq. (36)"},{"comment":"The cryogenic-amplifier measurement time is numerically inconsistent. With Table I (T = 4 K, omega_r = 2*pi*5 GHz), k_B T/(hbar omega_r) ~ 16.7, so Eq. (38) gives t_cryo ~ 7 * 16.7 * t_shot ~ 117 * 13 micro-seconds ~ 1.5 ms, not 9 ms. In addition, the derivation in Eq. (A5) uses an incorrect expression for the variance of integrated white noise: the integral of sinc^2(pi f t_m) over the bandwidth 1/t_m does not equal Si(2*pi)/pi ~ 20/9, and the variance should be S0 t_m rather than (20/9) S0 t_m. This affects the quantitative claims for the non-quantum-limited amplifier path and should be corrected.","section":"Sec. III B, Eq. (38), and Appendix A, Eq. (A5)"}],"minor_comments":[{"comment":"The prefactor in the SNR expression appears to be printed as 2 epsilon_d / (sqrt(kappa) (kappa^2/4 + chi^2)); the correct prefactor that yields Eq. (30) is 2 epsilon_d sqrt(kappa) / (kappa^2/4 + chi^2). As written, solving Eq. (29) would give a measurement time proportional to kappa rather than to 1/kappa.","section":"Eq. (29)"},{"comment":"The symbol 'd' in delta_Delta ~ delta_epsilon * d/l is undefined, and the numerical estimate delta_Delta ~ 0.1 micro-eV from |nabla_F Delta_1/2| = 0.2 e*nm, delta_epsilon = 20 micro-V, and l = 100 nm appears to give 0.04 micro-eV rather than 0.1 micro-eV. Please clarify the formula and check the arithmetic.","section":"Sec. III B, Eq. (35)"},{"comment":"The statement 'partial J / partial epsilon = partial J / partial V ~ 0.5 MHz/micro-eV' uses inconsistent units; if V is a gate voltage, the sensitivity should include the lever arm, and if the quantity is partial J / partial epsilon, the units should be MHz/micro-eV consistently.","section":"Sec. III B"},{"comment":"The legend label 'S SV Tz' is cryptic; the figure should clearly distinguish the spin purity, the spin-valley purity, and the expectation value of T_z.","section":"Fig. 2(d)"},{"comment":"Equation (30) is an equality defining t_shot, not an inequality; the text describing 'the inequality (30) is nonlinear in t_m' is inaccurate.","section":"Sec. III B"},{"comment":"The expression for delta_chi_max^+/- should be checked for sign conventions; as written, the denominator can vanish, and the text would benefit from an explicit statement of the monotonic domain in terms of chi_0 and kappa.","section":"Sec. IV A"}],"recommendation":"major_revision","confidential_remarks":"The valley T1 contradiction is the most serious issue because it directly threatens the headline measurement time. The simulation mismatch and the cryo-amplifier numerical error are fixable but still need correction before publication. I would also encourage the authors to clarify the derivation of delta_chi/chi in the simulation, since a reader cannot reproduce the 0.13 value from Eq. (36)."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is a genuine advance over the group's earlier single-electron valley-spectator proposal, and the core physics holds up on careful reading. But the paper contradicts itself on the very number the headline measurement times depend on — the valley relaxation time — and that has to be resolved before the 13 µs claim means anything.\n\nWhat's new: Ref. [23] needed a Ge-doped well and electric-dipole coupling; this paper gets the same spectator-valley idea working with exchange coupling in a standard Si/SiGe singlet-triplet qubit. Real step, not a rehash. The derivation through the Fermi-Hubbard model, the Schrieffer-Wolff reductions, and the hybridized-valley frame is careful. The SNR analysis is standard dispersive-readout physics, and the stochastic master-equation simulation independently checks the analytic SNR. The O(θ²) bound on spin disturbance is clean. The predicted measurement time is assembled from independently measured inputs, not fitted.\n\nWhere it gets soft. The valley lifetime is quoted three ways, and two contradict each other. Sec. I says 'valley relaxation times in silicon can exceed 10 ms [24]'; Sec. III B repeats ~10 ms; Sec. II C says the 'valley relaxation rate ∼10–100 kHz [24]', i.e. T1 = 10–100 µs, about three orders shorter. Same reference, both claims. If the short value is right, a 13 µs integration carries a 12–73% chance the probe decays mid-measurement, and the 1 ms continuous-monitoring window disappears. Load-bearing, not cosmetic. Separate from that, the transfer of a single-electron-device T1 to a pulsed ST qubit needs justification.\n\nThe reader's worry that the simulation's δχ/χ = 0.13 is double the Eq. (36) estimate doesn't survive the numbers: with Table I, Δ_rv = β0 − ω_r ≈ −150 MHz, so δΔ/Δ_rv ≈ 16% dominates and Eq. (36) gives roughly 13–20%. Simulation is consistent.\n\nMinor items: δg/g = δJ/J is asserted, not derived — fine in the exponential-J regime, but say so; the δμ ≈ 2–6 µeV range is looser than Eq. (32) with A0 = 1 µeV gives; the caption says both 100 and 1000 realizations.\n\nFor whom: spin-qubit groups doing noise metrology and closed-loop control. It deserves a serious referee; the idea is good, the main flaw is fixable, but the paper should come back with the valley-lifetime numbers reconciled and the device transfer argued.","headline":"A real extension of the valley-spectator idea to ST qubits, with sound mechanics, but the paper's own valley-lifetime numbers contradict each other and the headline measurement speeds depend on the longer one.","tokens_in":18111,"tokens_out":19116,"would_cite":true,"duration_ms":193034,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["85.35.Gv","03.67.Lx","72.70.+m"],"model":"deepseek-v4-flash","headline":"This paper claims that the valley degree of freedom of a silicon singlet-triplet spin qubit can serve as a colocated spectator charge-noise sensor, reaching unity SNR in about 13 microseconds through a resonator while preserving spin…","keywords":["charge noise sensing","singlet-triplet qubit","valley state","dispersive readout","exchange interaction","spin qubit","silicon quantum dots","real-time feedback"],"falsifier":"Build the proposed Si/SiGe double-dot resonator setup with Table I parameters, initialize one electron in the excited valley state, and apply a calibrated detuning step of $\\delta\\epsilon\\approx20\\,\\mu\\mathrm{V}$ (corresponding to $\\delta\\mu\\approx4\\,\\mu\\mathrm{eV}$) at a known time; the paper predicts the homodyne current reaches unity SNR within about 13 microseconds with a quantum-limited amplifier. If the detection time is far longer, or if the probe response decays on a sub-millisecond timescale, the valley-coherence assumption or the coupling estimates fail, and measuring the in-situ valley relaxation time under pulsed two-electron operation would settle which.","tokens_in":17065,"feed_emoji":"📡","tokens_out":15689,"duration_ms":149770,"temperature":0.7,"pith_summary":"This paper proposes turning the valley degree of freedom of a silicon singlet-triplet spin qubit into a continuous, in-situ charge noise sensor. One of the two electrons is loaded into an excited valley state; because the exchange interaction couples the valley states and is itself sensitive to charge-noise-induced voltage fluctuations, the valley acts as a colocated spectator whose resonator response tracks the exchange in real time. The authors show by signal-to-noise analysis that a charge noise event can be detected with unity SNR in about 13 microseconds under realistic device parameters with a quantum-limited amplifier, and in the sub-millisecond range with appropriately engineered resonators even without one. They further show that the probe perturbs the spin evolution only at order $\\theta^2$ (about $4\\times10^{-4}$ for their parameters), so it can run concurrently with qubit logic gates and feed corrections that counteract slow exchange drift.","feed_headline":"Valley state turns a spin qubit into a 13-microsecond noise sensor","feed_subtitle":"A spectator valley electron monitors exchange drift in real time without disturbing the qubit spin.","key_machinery":"The central object is the spin-dependent hybridized valley (HV) state: a rotation $U_{SV}(\\theta)=\\exp(-i\\theta\\,\\sigma_z\\tau_y/2)$ with $\\theta=\\tan^{-1}(J_c/\\Delta)$ diagonalizes the singly-excited-valley Hamiltonian into decoupled spin ($\\sigma$) and valley ($T$) Pauli operators, giving $H_{SV}=(J_s/2)\\sigma_z+(\\beta/2)T_z$ with $\\beta=\\sqrt{\\Delta^2+J_c^2}$. This HV degree of freedom plays the role of a spectator qubit living on the same two electrons as the singlet-triplet qubit: it couples transversally to a resonator probe field with strength $g=g_0\\cos\\theta_0$, producing a dispersive shift $\\chi=g^2/\\Delta_{rv}$ that charge noise modulates through $\\delta\\chi=\\chi(2\\delta g/g-\\delta\\Delta/\\Delta_{rv})$. The measurement-speed result rests on the signal-to-noise formula $t_m^{\\mathrm{shot}}=(\\kappa^2/4+\\chi_0^2)/(4\\kappa\\,\\delta\\chi_{\\min}^2\\langle n\\rangle)$, which gives 13 microseconds for the parameters of Table I, together with the numerically verified claim that measurement-induced collapse of the HV state leaves the spin pure to order $\\theta^2$.","core_discovery":"At the paper's core is a model Hamiltonian for the singly-excited-valley subspace of a two-electron singlet-triplet qubit. After a Schrieffer-Wolff reduction and a rotation $U_{SV}(\\theta)=\\exp(-i\\theta\\,\\sigma_z\\tau_y/2)$ with $\\theta=\\tan^{-1}(J_c/\\Delta)$, the spin and valley degrees of freedom decouple: $H_{SV}=(J_s/2)\\sigma_z+(\\beta/2)T_z$ with $\\beta=\\sqrt{\\Delta^2+J_c^2}$, so the excited valley acts as a spectator whose energy responds to the exchange coupling $J$ and the valley-splitting difference $\\Delta$. Coupling a resonator to the detuning gate produces a dispersive shift $\\chi=g^2/\\Delta_{rv}$ that charge noise modulates by $\\delta\\chi=\\chi(2\\delta g/g-\\delta\\Delta/\\Delta_{rv})$, so a homodyne measurement of the reflected field gives a continuous, classical readout of exchange fluctuations. For the realistic parameters of Table I (charge noise amplitude $A_0=1.0\\,\\mu\\mathrm{eV}$, resonator at 5 GHz with quality factor 2500, valley splitting difference $\\Delta=20\\,\\mu\\mathrm{eV}$), the SNR analysis gives a unity-SNR detection time of 13 microseconds with a quantum-limited amplifier, and 9 ms without one, reduced to about 90 microseconds if stray capacitance is cut tenfold. The protocol preserves spin coherence at the $O(\\theta^2)$ level, with $\\theta^2=4\\times10^{-4}$ for Table I parameters, works in standard Si/SiGe heterostructures without the Ge doping that earlier versions of the idea required, and the valley state can be repumped after decay.","pith_inferences":["A natural extension the authors sketch but leave undeveloped is closed-loop exchange stabilization: because the probe supplies a continuous estimate of $\\delta J$ whose uncertainty shrinks as $\\sqrt{t_m/t}$, the same homodyne signal could drive an integrator that cancels slow drift in real time, effectively turning the sensor into an active noise-cancellation loop for two-qubit gates.","The hybridized-valley machinery is not tied to Si/SiGe: the derivation carries over to Si-MOS devices (which the paper notes) and, in principle, to any two-electron system with an exchange-tunable spectator doublet, such as an orbital degree of freedom, so the protocol may generalize beyond silicon.","Because the sensor is colocated with the qubit, the apparatus could double as an exchange-noise spectrometer: recording the time series of $\\delta J$ yields the noise power spectral density exactly where it limits gate fidelity, replacing transport-based noise measurements that rely on imperfect spatial correlations."],"forward_implications":["A charge-noise event that shifts the exchange by about 3 percent (a 13 percent change in the dispersive shift) is detected with unity SNR in roughly 13 microseconds under Table I parameters with a quantum-limited amplifier, fast enough to track slow drift during operation.","Spin coherence is preserved at the $O(\\theta^2)$ level ($\\theta^2=4\\times10^{-4}$ for Table I), so the probe can run concurrently with qubit logic gates without appreciable disturbance.","The probe signal is nonmonotonic in $\\delta\\chi$, so the readout is single-valued only inside a bounded domain; larger shifts can still be resolved by applying a detuning correction and checking whether the baseline is restored.","A valley decay event is distinguishable from an exchange fluctuation because the signal drops to zero and does not respond to detuning correction; the valley state can then be repumped through the avoided crossings.","Without a quantum-limited amplifier the measurement time is about 9 ms, but engineering the resonator (for example, reducing stray capacitance tenfold) brings it to roughly 90 microseconds, so sub-millisecond sensing remains accessible to simpler readout chains."],"supporting_citations":[{"why":"The prior single-electron spectator-valley proposal that this work extends; it also supplies the valley-splitting field sensitivity of about 0.2 e·nm used to estimate δΔ.","marker":"[23]"},{"why":"The measured valley relaxation rate of 10–100 kHz in silicon, the basis for assuming the excited valley survives the roughly 10 ms measurement window.","marker":"[24]"},{"why":"The Fermi–Hubbard model including spin, valley, and doubly occupied states from which the probe Hamiltonian is derived.","marker":"[25]"},{"why":"The argument that each electron samples different alloy disorder, making the two dots' valley-splitting fluctuations uncorrelated in the estimate of δΔ.","marker":"[34, 46]"},{"why":"Defines the resonator voltage V_rms at the detuning gate and thereby the coupling strength g₀ used in the Hamiltonian.","marker":"[36]"},{"why":"The dispersive-regime elimination that produces the effective Hamiltonian with dispersive shift χ = g²/Δ_rv.","marker":"[40]"},{"why":"The solution for the resonator field amplitudes and the homodyne current formalism used in the SNR calculation.","marker":"[41]"},{"why":"Device data giving the lever arm α = 0.2 eV/V and exchange sensitivity ∂J/∂V ≈ 0.5 MHz/µeV used to convert chemical-potential noise into δJ.","marker":"[45]"},{"why":"Experimental charge noise amplitudes A₀ ≈ 1 µeV that set the noise level in the measurement-time estimates.","marker":"[8, 10–13, 47]"},{"why":"The numerical simulation of the stochastic master equation that confirms the predicted unity-SNR measurement time.","marker":"[52]"}],"fun_headline_variants":["Valley spectator senses charge noise in 13 µs","Spin qubit uses valley sidekick for fast noise sensing","Real-time charge noise monitor via spectator valley state","Microsecond-scale charge noise detection with valley qubit","Charge noise measured in 13 µs without disturbing spin"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the excited valley state stays coherent through the full measurement window: the paper imports a valley relaxation rate of roughly 10–100 kHz measured in a single-electron device and assumes it holds in the pulsed two-electron singlet-triplet configuration, so if valley $T_1$ is actually much shorter the probe signal decays and the sensor fails.","fun_headline_variants_meta":{"raw":{"variants":["Valley spectator senses charge noise in 13 µs","Spin qubit uses valley sidekick for fast noise sensing","Real-time charge noise monitor via spectator valley state","Microsecond-scale charge noise detection with valley qubit","Charge noise measured in 13 µs without disturbing spin"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000469,"raw_usage":{"total_tokens":2414,"prompt_tokens":1103,"completion_tokens":1311,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":719,"completion_tokens_details":{"reasoning_tokens":1233}},"tokens_in":719,"tokens_out":1311,"duration_ms":15774,"temperature":1.0,"reasoning_tokens":1233,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T16:01:57.905811+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Build the proposed Si/SiGe double-dot resonator setup with Table I parameters, initialize one electron in the excited valley state, and apply a calibrated detuning step of $\\delta\\epsilon\\approx20\\,\\mu\\mathrm{V}$ (corresponding to $\\delta\\mu\\approx4\\,\\mu\\mathrm{eV}$) at a known time; the paper predicts the homodyne current reaches unity SNR within about 13 microseconds with a quantum-limited amplifier. If the detection time is far longer, or if the probe response decays on a sub-millisecond timescale, the valley-coherence assumption or the coupling estimates fail, and measuring the in-situ valley relaxation time under pulsed two-electron operation would settle which.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The prior single-electron spectator-valley proposal that this work extends; it also supplies the valley-splitting field sensitivity of about 0.2 e·nm used to estimate δΔ."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The measured valley relaxation rate of 10–100 kHz in silicon, the basis for assuming the excited valley survives the roughly 10 ms measurement window."},{"cited_title":"Tariq and X","cited_arxiv_id":null,"evidence_quote":"The Fermi–Hubbard model including spin, valley, and doubly occupied states from which the probe Hamiltonian is derived."},{"cited_title":"Blais, A","cited_arxiv_id":null,"evidence_quote":"The dispersive-regime elimination that produces the effective Hamiltonian with dispersive shift χ = g²/Δ_rv."},{"cited_title":"Frisk Kockum, L","cited_arxiv_id":null,"evidence_quote":"The solution for the resonator field amplitudes and the homodyne current formalism used in the SNR calculation."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Device data giving the lever arm α = 0.2 eV/V and exchange sensitivity ∂J/∂V ≈ 0.5 MHz/µeV used to convert chemical-potential noise into δJ."}],"review_version":1}