Qubit Noise Sensing via Induced Photon Loss in a Superconducting Cavity

arXiv:2603.06848 · quant-ph · Submitted 2026-03-06 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Quantum Radio. Generated commentary on the latest quantum physics and condensed matter papers.

Kai: Today's paper: "Qubit Noise Sensing via Induced Photon Loss in a Superconducting Cavity".

Mira: Qubit noise sensing via induced photon loss in a superconducting cavity demonstrates a novel method for measuring qubit frequency noise by converting it into measurable photon loss in a coupled…

Kai: First, who's behind it and why it matters.

Title and authors: Kai: Now we get into the main substance of the paper, looking at what they actually accomplished with this protocol 'Qubit Noise Sensing via Induced Photon Loss in a Superconducting Cavity'. We're going to discuss what this technique actually means for measuring qubit noise.

Mira: This section details how they set up the experiment, specifically how they initialize a single photon in the cavity and then employ a sequence of repeated mid-circuit qubit measurements during a variable wait time <ref:2603.06848#pg2>.

Lev: That sequence of measurements is where the magic happens; it's what drives the energy exchange that creates the dressed dephasing channel they are trying to probe <ref:2603.06848#pg1>.

Kai: And after that, they perform a final photon population measurement which acts as a cavity reset, setting up the whole cycle for data collection <ref:2603.06848#pg2>.

Mira: The critical step is separating the noise-induced loss from the intrinsic cavity decay by post-selecting on all qubit measurement outcomes <ref:2603.06848#pg2>.

Lev: That post-selection is what allows them to discard runs where a cavity excitation was transferred to the qubit via a dressed dephasing event and subsequently detected by another measurement <ref:2603.06848#pg2>.

Kai: By doing that, they get the conditional photon survival probability, P g1(t), which is modeled by Equation (two) in the paper <ref:2603.06848#pg1>.

Mira: That equation is where they show how to mathematically model the effect of both intrinsic loss and noise, because kappa = kappa dd + kappa zero represents the total loss rate <ref:2603.06848#pg2>.

Lev: So they are using a mathematical framework to map the physical process—the dressed dephasing—onto a measurable change in photon survival probability <ref:2603.06848#pg1>.

Kai: And then, they use this comparison between two decay curves—one without post-selection and one with it—to isolate the kappa dd term directly <ref:2603.06848#pg1>.

Mira: That isolation is what lets them finally extract the dressed dephasing rate, kappa dd, by comparing the two decay curves in a way that cleanly separates them from each other <ref:2603.06848#pg1>.

Lev: And once they have that kappa dd, they can then directly relate it back to the frequency noise power spectral density S delta omega through Equation (one) in the paper <ref:2603.06848#pg1>.

Kai: So, the summary is that they devise a specific measurement sequence and use post-selection to mathematically isolate the noise component from the cavity's natural loss rate <ref:2603.06848#pg1>.

Mira: That isolation allows them to find that kappa dd and then use Equation (one) to quantify it in terms of the qubit frequency noise PSD at that detuning frequency <ref:2603.06848#pg1>.

Lev: It really shows how a physical process, dressed dephasing, can be mapped onto a quantifiable spectral density measurement, which is very valuable for us designing robust protocols <ref:2603.06848#pg1>.

Kai: So, this whole process boils down to using the cavity as a transducer to convert qubit frequency noise into measurable photon loss <ref:2603.06848#pg0>.

Mira: That transduction mechanism is what makes this technique novel compared to just measuring the qubit directly, opening that higher-frequency spectral window for our analysis <ref:2603.06848#pg1>.

The paper's summary: Kai: Now let's talk about the actual enhancements they propose within this approach, what they suggest to make it even better than what they have done so far with 'Qubit Noise Sensing via Induced Photon Loss in a Superconducting Cavity'. We want to focus on the practical refinements they suggest.

Mira: The main improvement lies in the detection probability xi, which is defined by Equation (three), where xi = kappa dd/kappa times e-kappa T m - e- T m / (one-e-kappa T m) <ref:2603.06848#pg2>.

Lev: That detection probability is key because it tells us how often we can actually resolve that dressed dephasing event, and if xi is small, the signal becomes very hard to see <ref:2603.06848#pg1>.

Kai: The authors point out that in the limit where T m one/, nearly every dressed dephasing event is detected, which means the signal becomes resolvable and reliable <ref:2603.06848#pg1>.

Mira: Furthermore, they show that when they inject controlled noise matching the cavity-qubit detuning, the cavity lifetime reduces significantly, from eleven point three ms down to three point two ms <ref:2603.06848#pg2>.

Lev: That reduction in lifetime is a strong signal that the induced loss is indeed due to dressed dephasing, which helps us validate the physical mechanism being modeled <ref:2603.06848#pg2>.

Kai: They also show that a simulation of the system dynamics using kappa dd as the only free parameter reproduces both that lifetime reduction and the transient qubit heating observed during the process <ref:2603.06848#pg2>.

Mira: That simulation result is powerful because it confirms that dressed dephasing accounts for both observable phenomena, which strengthens the underlying assumption of their model <ref:2603.06848#pg2>.

Lev: It’s good to see a simulation that validates the physical process, because we need confidence in our models before we can trust them for running complex error correction simulations <ref:2603.06848#pg2>.

Kai: So the improvement is less about a new hardware component and more about refining the mathematical tools—using post-selection and careful modeling to extract reliable parameters <ref:2603.06848#pg1>.

Mira: Precisely, so instead of just measuring loss, they’ve developed a conditional probability framework to ensure that the noise term is cleanly separated from the cavity's inherent dissipation <ref:2603.06848#pg2>.

The paper's improvements: Kai: We've got to wrap up this discussion on 'Qubit Noise Sensing via Induced Photon Loss in a Superconducting Cavity', summarizing what we learned about its implications and where it leads next. What are the specific practical steps forward?

Mira: Essentially, the paper shows that by using repeated mid-circuit measurements and post-selection, they can successfully convert qubit frequency noise into measurable photon loss in a coupled cavity <ref:2603.06848#pg1>.

Lev: The key result we're focusing on is that without adding external noise, they establish an upper bound of five times ten cubed Hz two/Hz on the qubit frequency-noise power spectral density at five hundred eight MHz <ref:2603.06848#pg1>.

Kai: That specific number gives us a concrete benchmark for how quiet our qubits can be in that high-frequency region, which is very useful for setting performance targets <ref:2603.06848#pg1>.

Mira: The big implication is that this opens access to a higher frequency spectral window than standard qubit-based spectroscopy typically allows, potentially enabling noise characterization during strongly driven operations <ref:2603.06848#pg0>.

Lev: This suggests that future error correction strategies might benefit from incorporating this kind of high-frequency noise sensing capability directly into the control loop or syndrome measurement process <ref:2603.06848#pg1>.

Kai: I think the work provides a solid foundation for understanding how environmental noise couples into the system in a way that's hard to see otherwise <ref:2603.06848#pg1>.

Mira: And by providing these constraints, they give us more information to build better predictive models for coherence degradation under realistic operating conditions <ref:2603.06848#pg1>.

Lev: It’s a step toward making our error correction codes more robust by giving us a better handle on the noise landscape at these higher frequencies <ref:2603.06848#pg1>.

Kai: So, this paper 'Qubit Noise Sensing via Induced Photon Loss in a Superconducting Cavity' offers a sophisticated method for probing qubit noise through cavity loss and sets some useful limits on that noise spectrum when no added noise is present <ref:2603.06848#pg1>.

Conclusion: Kai: So we've covered how they used photon loss in a cavity to measure qubit frequency noise through dressed dephasing in this paper, "Qubit Noise Sensing via Induced Photon Loss in a Superconducting Cavity." We’ve seen the core mechanism and the results.

Mira: Exactly, and the way they separated that intrinsic decay from the noise-induced loss using post-selection is mathematically quite elegant for isolating kappa dd <ref:2603.06848#pg1>.

Lev: From an error correction standpoint, those extracted parameters give us a concrete value for how much noise we can expect in this specific frequency window, which helps us benchmark our code's performance assumptions <ref:2603.06848#pg1>.

Kai: Right, and the validation they did by injecting controlled tones to see that lifetime drop really proved the physical mechanism they were modeling was actually happening in their system <ref:2603.06848#pg2>.

Mira: That experimental validation is crucial because it backs up their Lindblad master equation formalism for describing those dynamics under strong driving <ref:2603.06848#pg1>.

Lev: If we can reliably extract kappa dd this way, it means we might be able to build better real-time noise estimators directly into the quantum control stack for fault tolerance research.

Kai: That's what I mean, because if we can continuously estimate this noise spectrum during a gate operation, we could adapt our pulse shapes dynamically instead of relying on static calibrations <ref:2603.06848#pg1>.

Mira: It opens up that high-frequency spectral window you mentioned earlier, which is something standard qubit spectroscopy simply can't reach, providing a much richer view of the environment <ref:2603.06848#pg1>.

Lev: For practical implementation, having an estimator that works in real-time under strong driving would significantly improve our ability to mitigate errors before they accumulate in long sequences.

Kai: It's a solid step toward making our hardware more resilient by giving us better data on the noise landscape during active computations <ref:2603.06848#pg1>.

Mira: I agree, and the paper's focus on those conditional probabilities really underscores how critical it is to model the detection efficiency xi accurately for any practical application <ref:2603.06848#pg1>.

Lev: So, while this method isn't a complete solution, providing these precise noise characterization tools is valuable groundwork for developing those next-generation control algorithms we need.

Kai: Indeed, and the work on "Qubit Noise Sensing via Induced Photon Loss in a Superconducting Cavity" gives us tangible parameters to work with when designing our next generation of quantum hardware <ref:2603.06848#pg1>.

Mira: It sets a high bar for how we can use cavity coupling as a sensing mechanism, showing that this isn't just an academic curiosity but a viable tool for probing system noise <ref:2603.06848#pg1>.

Lev: That's the main point, and it gives us something concrete to aim for when we start looking at how to integrate these sensing capabilities into our error correction frameworks <ref:2603.06848#pg1>.

Nitzan Kahn, *Dror Garti, *Uri Goldblatt, Lalit M. Joshi, Fabien Lafont, Serge Rosenblum

Department of Condensed Matter Physics, Weizmann Institute of Science

quant-ph

Submitted: 2026-03-06

Updated: 2026-10-05

Comments: 14 pages, 10 figures, including supplemental materials

License: http://creativecommons.org/licenses/by-nc-nd/4.0/

Importance score: 82/100

The gist: Qubit noise sensing via induced photon loss in a superconducting cavity demonstrates a novel method for measuring qubit frequency noise by converting it into measurable photon loss in a coupled

Key concepts

Dressed Dephasing
This occurs when qubit frequency noise causes energy exchange between the qubit and the cavity. This interaction creates an additional channel for photon loss that is directly proportional to the frequency noise spectrum at a specific detuning. It's a mechanism where noise drives measurable loss in the cavity.
Photon Survival Probability
This measures the chance of finding a single photon in the cavity after time t. The method uses two probabilities: one for all outcomes and one post-selected on ground state returns. Comparing these helps separate the intrinsic cavity decay from noise-induced loss events.
Frequency Noise Power Spectral Density (PSD)
This describes how much noise exists at different frequencies affecting the qubit's frequency. The paper uses this PSD to calculate the dressed dephasing rate, which is directly proportional to it. This allows researchers to quantify the specific type of noise impacting qubit coherence.
Post-Selection
This involves only analyzing experimental runs where all qubit measurements returned the ground state. By filtering out runs where a cavity excitation was transferred to the qubit and detected, researchers can isolate the signal caused specifically by noise-induced loss events.

Terminology

Summary

Qubit noise sensing via induced photon loss in a superconducting cavity demonstrates a novel method for measuring qubit frequency noise by converting it into measurable photon loss in a coupled high-Q superconducting cavity. This technique is significant because it opens access to a higher-frequency spectral window than standard qubit-based spectroscopy and enables noise characterization during strong driving operations, which are often inaccessible to conventional methods.

The gist: A cavity-based technique for sensing qubit frequency noise through photon loss induced by dressed dephasing places an upper bound on the intrinsic dressed-dephasing rate of (0.3s)−1 at 508 MHz, corresponding to a PSD of Sδω(∆) < 5×103 Hz2/Hz.

How it works

The core concept relies on dressed dephasing, where qubit frequency noise drives energy exchange between the cavity and the qubit, creating an additional photon loss channel whose rate is proportional to the frequency-noise spectral density at the cavity-qubit detuning [30, 31]. The protocol involves several distinct stages:

  1. Initialization of a single photon in the cavity.

  2. A sequence of repeated mid-circuit qubit measurements during a variable wait time.

  3. A final photon population measurement that serves as a cavity reset.

Separating Noise Contributions

The key to this method is separating the noise-induced loss from intrinsic cavity decay using post-selection on all qubit measurement outcomes. The unconditional photon survival probability is modeled as P1⟩(t) = e−κt, where κ = κdd + κ0 represents the total loss rate. Post-selecting on trajectories in which every qubit measurement returns the ground state discards runs where a cavity excitation was transferred to the qubit via a dressed dephasing event and detected by a subsequent measurement. This conditional photon survival probability is given by Eq. (2): Pg1⟩(t) = e−κt e−κt + (1−ξ)(1−e−κt), where ξ is the detection probability of the loss event.

Extracting Noise Parameters

The dressed-dephasing rate, κdd, is directly proportional to the frequency noise power spectral density (PSD) at the cavity-qubit detuning frequency ∆: κdd = 4g2∆2/Sδω(∆) (Equation 1). By comparing two decay curves—one discarding all qubit measurement outcomes and one post-selected on ground state returns—the dressed-dephasing rate is extracted. The detection probability ξ is defined by Eq. (3): ξ = κdd/κ · e−κTm −e−ΓTm / (1−e−κTm). In the limit Tm ≪ 1/Γ, nearly every dressed-dephasing event is detected, and the signal becomes resolvable.

Validation and Results

The protocol was validated by injecting controlled frequency noise into the transmon using two noise-broadened microwave tones whose difference frequency matches the cavity-qubit detuning ∆. When this injected noise induces dressed dephasing, the cavity lifetime is reduced from 11.3 ms to 3.2 ms (Fig. 2a). The extracted intrinsic loss rate κ0 = (9.8 ± 0.5 ms)−1 indicates that the protocol correctly separates noise-induced loss from intrinsic cavity decay, and the corresponding detection probability is ξ = 0.63, since Tm≪1/Γ in this regime. The protocol places an upper bound on the intrinsic dressed-dephasing rate of (0.3s)−1 at 508 MHz, corresponding to Sδω(∆) < 5×103 Hz2/Hz when no added noise is present. This approach extends noise spectroscopy beyond the bandwidth of standard qubit-based methods, which are typically limited to frequencies below a few hundred megahertz [32].

System Setup and Modeling

The experimental setup couples a transmon qubit (resonance frequency ωq/2π = 3.800 GHz) dispersively to a high-Q niobium cavity (resonance frequency ωc/2π = 4.308 GHz). The system dynamics are modeled using the Lindblad master equation formalism, incorporating collapse operators for cavity decay, transmon relaxation, pure dephasing, and dressed dephasing. Bayesian inference is employed to extract κdd and κ0 by fitting both unconditional and post-selected decay curves simultaneously. The results confirm that the extracted noise PSD scales linearly with the injected noise PSD at high power levels, consistent with Equation (1). At low noise power, the extracted values plateau, defining a detection floor where the signal is no longer resolvable. This method opens access to a high-frequency spectral window that is difficult to probe with conventional qubit-based methods and is directly relevant to strongly driven qubit operation.

Improvements for AI systems

Here are the specific improvements to AI systems that could be made by leveraging the techniques described in this scientific paper:

The core improvement lies in developing AI/ML architectures that can operate directly on or as a feedback mechanism for superconducting quantum processors, specifically addressing noise characterization and signal extraction under strong driving.

  1. A new class of Noise-Aware Quantum Control Algorithms:

  2. An improved AI system for Real-Time Noise Spectroscopy (RTNS):

  3. A robust Qubit Performance Predictive Model:

The improved AI systems can do the following specific things:

  1. A new class of Noise-Aware Quantum Control Algorithms can perform real-time, adaptive gate operations on superconducting qubits by dynamically adjusting control pulses based on the instantaneous qubit frequency noise spectrum, rather than using pre-calibrated fixed pulses.

  2. An improved AI system for Real-Time Noise Spectroscopy (RTNS) can automatically employ the mid-circuit measurement and post-selection protocol described in Section S4 to continuously estimate the qubit frequency noise power spectral density, allowing for the characterization of noise during high-speed or strongly driven operations that would otherwise be impossible with standard methods.

  3. A robust Qubit Performance Predictive Model can use the extracted parameters (like the dressed dephasing rate, κdd) to predict long-term coherence degradation and gate fidelity under varying environmental noise conditions, enabling proactive error mitigation strategies in quantum computing hardware.

Abstract

Characterizing noise in superconducting qubits is essential for improving coherence and gate performance. Conventional noise-sensing methods typically use the qubit itself as the sensor, which limits both accessible bandwidth and applicability during driven operation. Here, we measure qubit frequency noise by detecting the photon loss it induces in a coupled high-Q superconducting cavity. We use repeated mid-circuit qubit measurements with post-selection to separate this induced loss from intrinsic cavity decay. We validate the protocol using injected noise and show that the extracted loss scales as expected with the applied noise strength. We place an upper bound of 0.70 times10 cubed, rad 2/s on the intrinsic qubit frequency-noise power spectral density at the cavity-qubit detuning of 508 MHz. The protocol opens access to a higher-frequency spectral window than standard qubit-based spectroscopy and may enable noise characterization during strong driving.

Sources

Related papers