WorksPhysics and Astronomy
Two-tone spectroscopy of high-frequency quantum circuits with a Josephson emitter
Abstract
We perform two-tone spectroscopy on quantum circuits, where high-frequency radiation is generated by a voltage-biased superconductor-normal-superconductor Josephson junction and detection is carried out by an ancillary microwave resonator. We implement this protocol on two different systems, a transmon qubit and a resonator. We demonstrate that this two-tone Josephson spectroscopy operates well into the millimeter-wave band, reaching frequencies larger than 80 GHz, and is well-suited for probing highly coherent quantum systems.
Introduction
Circuit quantum electrodynamics (cQED) is based on microwave measurements and control of superconducting circuits [1, 2, 3, 4]. In practice however, the accessible frequencies are typically smaller than 30 GHz, due to the limited range of commercially available instruments and components and the difficulty to route high-frequency signals in a cryostat. Increasing this upper limit could help to probe quantum systems in hybrid architectures [5] as well as elucidate decoherence mechanisms caused by the environment at high frequency [6]. It would also open alternative routes towards the development of high-frequency qubits that can operate at higher temperatures [7]. In order to take a step forward in this direction, one can exploit the ac Josephson effect [8, 9] to generate high-frequency radiation directly at the cold-stage level. When biased at voltage , a Josephson junction indeed radiates photons at frequency , thereby working as an on-chip tunable-microwave source [10, 11, 12, 13, 14, 15, 16, 17, 18]. This phenomenon enables broadband spectroscopy that extends from a few tens of MHz up to the THz regime. Its principle relies on power conservation and is quantumly described by the theory of dynamical Coulomb blockade [19, 20]. Each time a photon of energy is absorbed and dissipated in the junction’s electromagnetic environment, a Cooper pair of energy inelastically flows across the junction. Consequently, when matches a resonance of the target system, one can detect a current that is directly proportional to the photon emission rate .
Absorption Josephson spectroscopy has been employed for characterizing macroscopic systems, such as ensembles of cobalt atoms [21] or microwave resonators [22–27], as well as elementary quantum objects, including Cooper-pair boxes [30, 31] [28, 29] and Andreev bound states [32, 33, 34, 35]. This spectroscopy technique is, however, challenging as it is sensitive to the whole electromagnetic environment of the junction and relies on low-frequency measurements of small dc currents. Even more crucially, this approach is not suited to probe highly coherent quantum systems like superconducting qubits, as its detection sensitivity decreases dramatically for long coherence time. Indeed, the dissipation rate of photon per coherence time interval induces an upper bound for . A typical would thus amount in detecting a current of a few fA, 4 orders of magnitude below the resolution of the best power-absorption Josephson spectrometer [27].
In this work, we develop another paradigm for Josephson spectroscopy, inspired by cQED, where the detection is realized using an ancillary system (see Figure 1a). First, we utilize a superconductor-normal-superconductor (SNS) Josephson junction as an on-chip microwave emitter (red). We couple such an SNS emitter to a quantum object of interest with frequency (green). This system is in turn dispersively coupled to a microwave cavity that acts as a detector (blue) resonating at a readily accessible frequency . When driven at resonance , the quantum object gets excited with photons, which induces a frequency shift of the cavity that can be detected using sensitive and fast radiofrequency techniques. Following this approach, our goal is to perform two-tone Josephson spectroscopy at high frequency. As a proof of concept, we implement this strategy to probe two coherent though very different quantum objects, a superconducting transmon qubit that resonates at 6 GHz, and electromagnetic modes of a resonator up to 83 GHz.
Figure 1. (a) Principle of two-tone Josephson spectroscopy. A Josephson emitter (red) emits photons at frequency onto the system of interest (green) that resonates at frequency . A microwave resonator (blue) with frequency is dispersively coupled to the system of interest and serves as a detector. When the system of interest is excited, the resonator’s frequency shifts, which is detected by a probe tone. (b) Current-voltage characteristic of an SNS Josephson junction measured at 7 mK. The gray region is not accessible as the current falls below the retrapping current and the junction jumps to the zero-voltage state. Inset: scanning electron microscope image of the device. The normal part is a Ti(5 nm)-Au(15 nm) bilayer and the superconducting part is a 55 nm-thick layer of Nb. (c) Josephson power as a function of both frequency and voltage across the emitter. The Josephson radiation is here emitted by an Al(100 nm)-Hf(7 nm)-Al(100 nm) junction with nA and measured using a spectrum analyzer by integrating the power in a 2 MHz bandwidth around . Cuts corresponding to the blue (), orange () and green () dashed lines are shown in the top left inset. Bottom right inset: simplified version of the setup, where a 4–8 GHz band-pass filter protects the junction from the 1.5 K noise emitted by the HEMT.
SNS Josephson junction as a microwave emitter
Our Josephson emitter consists of two superconducting electrodes connected by a normal metallic wire, forming an SNS Josephson junction (see micrograph in Figure 1b and details of fabrication in Appendix A). We conduct measurements either on Nb-Au-Nb or Al-Hf-Al structures. In the latter case, although the Hf wire might actually become superconducting at the base temperature of our experiments [36], it essentially behaves like a normal metal above its critical current at finite voltage.
We initially characterize our devices through dc electrical measurements and extract their current-voltage characteristic (see Figure 1b and Appendix C). Depending on the thickness, width, length and composition of the normal wire connecting the two superconductors, the critical current can vary from a few hundreds nA to a few tens of . When biased above , the junction switches and acquires a finite normal resistance , ranging from about 10 to a few hundreds of Ohms depending on the devices. Due to Joule heating, retrapping occurs at a current significantly smaller than . The thermalization is mostly ensured by the normal metal wings, which act as thermal baths and maintain the emitters temperature below a few hundreds of mK. In most of our designs, the product is around 100 , which is consistent with results of other experiments with similar geometries [37, 38]. This suggests that our SNS junctions possess a minigap GHz and are likely to exhibit additional dissipation beyond this frequency [39].
The operation of SNS microwave sources relies on the ac Josephson effect, as they carry, when voltage-biased, an oscillating current at the Josephson frequency . The voltage-to-frequency conversion ratio is thus given by the inverse of the magnetic flux quantum GHz/. Consequently, Josephson junctions can emit microwave photons ranging from a few GHz up to a few hundreds of GHz. This upper bound is set by the value of the superconducting gap, which can reach GHz, 1.4 THz and 1.8 THz for Al, Nb and NbN respectively, the three most commonly used materials in superconducting circuits. Beyond this limit, the amplitude of the microwave signal is expected to decrease slowly [40, 41]. In the characteristic of Figure 1b, the ac Josephson current is averaged out and thus invisible. To detect it, we measure the power spectral density emitted by the junction in the 4-8 GHz band by routing the signal with a bias tee to a cryogenic HEMT amplifier anchored at 4K. In the power spectra presented in Figure 1c, we observe an emission peak whose frequency coincides with , demonstrating that this signal is indeed Josephson emission.
The peak has a Gaussian shape with a standard deviation MHz, which we attribute to a 10 nV RMS noise on the bias voltage across the junction of both thermal and technical origin [42]. Such a thin linewidth, obtained thanks to significant filtering and thermalization of the bias lines (details in Appendix B), is comparable to the best performances achieved so far [43, 44, 45] [46]. Additionally, we observe a frequency dependence of the signal amplitude within the 4 – 8 GHz bandwidth of our detection chain, with a large decrease of the signal around 7.3 GHz, probably due to a parasitic resonance on the junction chip or in our setup. This illustrates how challenging it is to efficiently route the signal emitted by the junction in the desired direction and to determine it quantitatively. By integrating the signal from Figure 1 over the full linewidth and comparing it to the noise floor added by the HEMT amplifier, we find an emission power of dBm. This microwave power corresponds to the emission of about photons per second, i.e. of the order of magnitude of the read-out pulses used in circuit-QED experiments. Note that the power emitted by the SNS junction strongly depends on the environment it is embedded into and might change for each experiment. At larger voltage, we observe an additional broadband emission due to shot noise (see Appendix C), whose impact is negligible in the following experiments.
These dc and ac measurements therefore show that SNS junctions function properly as microwave sources, at least in the GHz range. Interestingly, SNS microwave sources offer several advantages compared to more standard tunnel Josephson junctions. First, their characteristic is mostly linear, with remaining relatively small. Their impedance thus weakly changes with and, as such, does not significantly affect other components to which it is coupled. This aspect is crucial for applications like two-tone spectroscopy, where a sensitive microwave cavity interacts with the junction (see Appendix § 1.13). Second, one can mention the relative ease of their fabrication process compared to tunnel junctions when using niobium as a superconductor [47–49]. Finally, the low backaction from the environment on their characteristics make them simpler to use as their resistance does not significantly vary depending on the objects to which they are coupled. In the following, we use these on-chip SNS microwave emitters to perform two-tone spectroscopy of quantum circuits.
Two-tone Josephson spectroscopy of a transmon qubit
As a first proof of concept, we employ two-tone Josephson spectroscopy on a well-known quantum system: a transmon qubit [50]. While its frequency may be relatively low, it allows us to validate our technique by comparing it to measurements obtained using conventional microwave instrumentation. The transmon is capacitively coupled to a microwave resonator of frequency for dispersive readout (see Figure 2a). The qubit and cavity being off-resonantly coupled with a large enough detuning, they can be described by the Hamiltonian , where and are the ladder operators of the cavity, and the ground and excited state of the qubit, and is the dispersive frequency shift. The cavity frequency is thus conditioned on the state of the qubit, which provides the basis for performing its two-tone spectroscopy. Details on the design and fabrication of this circuit, which was initially created for microwave photon counting are given in Refs. [51, 52].
Figure 2. Two-tone Josephson spectroscopy of a transmon qubit. (a) Schematic of the experimental setup. An Al-Hf-Al SNS Josephson emitter (red), mounted in its own sample holder, emits photons at frequency . These photons are routed via an SMA cable to another sample holder towards the drive line of a transmon qubit (green), which is dispersively coupled to a microwave resonator (blue). By probing the resonator close to its resonant frequency , one can detect the excitation of the qubit. Though not represented, a GHz bandpass filter is also present in the setup. (b) Reflection coefficient , normalized by its value in absence of the Josephson drive, as a function of the emitted Josephson frequency . Experimental data are shown as blue dots and fits as solid lines. The latter corresponds to the sum of two Gaussians of standard deviation centered at and , with a relative weight given by the finite population of the Fock state in the resonator. Fit parameters: GHz MHz, MHz, , MHz.
We first characterize the transmon circuit using standard microwave instruments and thus access to its characteristic frequencies and coherence times (see Appendix E). We then turn to two-tone Josephson spectroscopy using our on-chip SNS microwave emitter ( nA, see Figure 5 in Appendix C). For this experiment, the emitter and the transmon are separated on two independent chips, each mounted in its own sample-holder, connected to each other by an SMA cable (see Figure 2a). We emit Josephson radiation onto the transmon while probing the resonator via reflectometry close to its resonant frequency GHz. As we vary , we observe a clear suppression of the reflection coefficient at the qubit frequency GHz, demonstrating
Figure 5. Dc and ac characterization of the Al(100 nm)-Hf(7 nm)-Al(100 nm) Josephson junction. (a) Current-voltage characteristic . The switching current is 385 nA and its normal resistance is 75 Ω. (b) Excess noise emitted by the junction at 5 GHz, in units of the HEMT noise background . The red arrow indicates the AC Josephson effect peak.
that it gets excited by the SNS microwave source (see Figure 2b).
The transition linewidth with is much larger than the intrinsic qubit decay rate. This width is set by the voltage noise across the SNS emitter and is similar to the one measured in Figure 1c. On top of that, the resonant dip displays an asymmetry, as a second transition can be excited at frequency , with . The qubit frequency indeed depends on the number of photons in the cavity due to the dispersive coupling. As a consequence, we observe a transition between states and , owing to a finite population of the Fock state . We therefore demonstrate that our on-chip SNS microwave emitter can be used to detect an elementary quantum system, a transmon qubit that resonates around , using two-tone Josephson spectroscopy.
Two-tone Josephson spectroscopy of high-frequency modes
To further test this alternative detection scheme, we now probe the electromagnetic modes of a resonator. The advantage of such a system is that it has a large number of regularly spaced modes, which is well suited to explore two-tone Josephson spectroscopy at high frequency and over a large range. The resonator, shown in Figure 3a-Figure 3b, is made out of niobium and resonates at its harmonic frequencies , where is the mode index. It is inductively coupled to an rf SQUID, which consists of an SNS Josephson junction enclosed in a superconducting loop threaded by a magnetic flux . This introduces some nonlinearity into the circuit, which results in cross-Kerr interaction between the resonant modes. The system can thus be described by the Hamiltonian , where are the cross-Kerr frequencies [53]. Therefore, when mode gets excited with 1 photon, the fundamental mode’s frequency is shifted by . This provides the basis to perform a two-tone Josephson spectroscopy of the high-frequency modes, while the fundamental mode of the resonator is used as a detector (see Figure 3c).
Figure 3. Two-tone Josephson spectroscopy of high-frequency modes. (a) Schematic of the circuit, and (b) illustration of the device layout (with niobium in white and silicon in black). The resonator (blue) is inductively coupled at its end, both to a superconducting loop closed by an SNS junction (rf SQUID in pink) and to an SNS Josephson emitter (red). Both Josephson junctions are Nb-Au-Nb structures. A flux line, visible at the bottom, is used to adjust the magnetic flux in the rf SQUID loop. (c) Diagram illustrating the principle of two-tone spectroscopy based on the cross-Kerr effect between electromagnetic modes within the same resonator. The fundamental mode is continuously monitored with a probe tone , while the -index mode is irradiated at frequency . At resonance , mode gets excited with photons. Consequently, the fundamental mode frequency shifts to and a phase shift of the reflected signal is observed. (d) Phase shift of the probe tone as a function of . Orange vertical lines serve as markers for the expected frequencies of the resonator’s harmonics. Close to each line, a resonance corresponding to an electromagnetic mode of the resonator is observed. Inset: detailed view around a resonance close to 83 GHz, interpreted as the mode of the resonator. The signal-to-noise ratio was here optimized by carefully selecting the probe tone’s frequency and the magnetic flux . (e) Detailed view of the resonance for mode , measured at (solid line) and (dashed line). The observed sign change of the dispersive coupling confirms that this resonance corresponds to a mode of the resonator.
As before, we use a voltage-biased SNS Josephson junction as a microwave emitter. It is positioned on the same chip as the resonator, with which it is inductively coupled. We emit Josephson radiation while probing the fundamental mode of the resonator at frequency . Figure 3d shows the phase shift of the reflected signal as a function of the Josephson frequency . The spectrum exhibits regularly spaced resonances at frequencies , with a mean relative deviation of about between measurement and theory. Note that no fitting parameter is used as is extracted from one-tone spectroscopy measurement of the cavity (see Appendix § 1.12). These resonances thus correspond to the higher harmonic modes of the resonator. Other resonances appear in our signal, but are not clearly identified at this stage. We believe they might be geometric resonances of the circuit that would couple to the resonator. Additionally, a slowly varying background is noticeable. It is attributed to a flux change within the loop terminating the resonator, induced by the dc current carried by the Josephson emitter.
The measured linewidth of the different transitions, which is roughly constant and about , is much larger than the intrinsic decay rate of the modes that we estimate to be based on the measured quality factor of the fundamental mode. It is again limited by the Josephson emission linewidth, which is here almost 2 orders of magnitude larger than previously observed. This larger linewidth is due to a larger voltage noise across the SNS emitter, owing to the different filtering setup needed for this experiment (see Appendix § 1.7 and Appendix J). Within this architecture, we are able to detect the modes up to , corresponding to . We do not detect modes , which we attribute to dissipation in the emitter junction shunting the high-frequency Josephson radiation. In essence, the normal resistance of the junction and its coupling inductance induce a cutoff frequency at above which the signal decreases as , until it is too low to be detectable within a reasonable averaging time. Although it limits our sensitivity at high frequencies, this cutoff is less drastic than the LC cutoff of tunnel junction-based spectrometers, where the sensitivity is suppressed as [32]. This cutoff frequency can be increased either by opting for an emitter with higher normal resistance or by reducing the junction’s length to increase the size of its minigap. This problem is however unlikely to arise when probing a more nonlinear system, such as Andreev bound states, where the coupling with the resonator should be much larger (see section F).
Going further, we adjust the flux threading the rf SQUID loop. The primary consequence is the modulation of the fundamental mode’s frequency , with a measured magnitude of . Consequently, the higher harmonic modes are also expected to modulate with . However in practice, this modulation is not observable in the two-tone Josephson spectrum, as it is much smaller than our resolution, limited by the Josephson emission width. Nonetheless, changing the flux from to strongly alters the spectrum, as illustrated in Figure 3e for resonance around . Instead of a peak, a resonant dip is observed at , demonstrating an inverted frequency shift. This change can be understood as a reversal of the Josephson inductance of the rf SQUID when varying the phase drop across the junction [53] (see Appendix § 1), which further confirms the identification of the harmonic modes of the resonator.
Conclusion and perspectives
These measurements demonstrate that SNS Josephson junctions behave as on-chip broadband microwave emitters, capable of performing two-tone spectroscopy on superconducting quantum circuits at unprecedented frequencies. With this innovative technique, we were able to probe both a transmon qubit resonating around 6 GHz and the electromagnetic modes of a resonator up to 83 GHz. Two-tone Josephson spectroscopy is here well adapted to detect these highly coherent systems, with lifetimes µs and µs respectively. All indications suggest that this technique should operate at higher frequencies, albeit necessitating some adjustments in both the emitter properties and its coupling with the system under study. Conducting the same type of spectroscopy using commercial components, like frequency doublers, presents significant challenges as it would require routing high-frequency signals from room temperature to the cryostat. Two-tone Josephson spectroscopy thus provides a simple and easy to implement method to characterize high-frequency quantum circuits. Greater flexibility can also be achieved by using a flip-chip apapproach, in which the Josephson emitter and the system to be probed are placed on two separate chips that are brought close together [27].
More fundamentally, two-tone Josephson spectroscopy could be instrumental in detecting fermionic or bosonic excitations in mesoscopic systems, such as Andreev bound states in superconducting quantum dots [54, 55] or topological Weyl band structures in multiterminal Josephson junctions [56–58]. It could also prove valuable for studying quasiparticle generation [59] and recombination in exotic 2D superconductors [60–63]. Another promising axis is to explore the quantum nature of Josephson radiation [43, 44, 45] [64, 65, 66, 67, 68, 69]. This could lead to the development of Josephson photonics devices, such as quantum-limited amplifiers [64] or single-photon detectors [66, 70], operating at high frequencies. More generally, the architectures that we developed could be used to probe the high-frequency environment of standard superconducting qubits and how it induces decoherence on them. Finally, an exciting research direction involves sharpening the emitted Josephson frequency, achievable either by reducing voltage noise via even better filtering and thermalization or by employing clever injection locking techniques [17, 18] [71, 72, 73]. This advancement would enable time-domain measurement and control at unprecedented frequencies, thereby opening up a realm for high-frequency cQED.
Acknowledgments
We first want to emphasize the invaluable help of the late F. Portier and P. Jacques from the Nanoelectronics Group on low-noise electronics. We acknowledge valuable discussions with the Quantronics Group, in particular with C. Urbina for sharing the initial idea of two-tone Josephson spectroscopy. Special thanks to Ç. Girit, J.-L. Smirr, and Z. Leghtas on microwave measurements. Gratitude is extended to the SPEC of CEA-Saclay for their help on nanofabrication, and to D. Roux and R. Mohammedi from LPMC for their technical support. L.B. acknowledges support of the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (Grant Agreement No. 947707). J.D.P. acknowledges support of Agence Nationale de la Recherche through Grant No. ANR-20-CE47-0003. This work has been supported by the French ANR-22-PETQ-0003 Grant under the France 2030 plan.
Sample fabrication
Resonator fabrication begins with sputtering 150 nm of Nb onto an oxidized high-resistivity silicon wafer, which is subsequently diced into chips. For processing a single chip, a layer of S1813 resist is spun on top of the Nb. Following this, the negative of the resonator pattern is defined using a laser writer and developed in MF319. Subsequently, reactive ion etching is utilized to etch away the Nb through the resist mask. The resist is then removed by immersing the chip in acetone maintained at for 1 hour, followed by sonication in an ultrasonic bath.
Our SNS junctions are fabricated via metallic deposition on the same chip using a suspended resist mask. For this, a bilayer of MAA EL6 resist is spun at 4000 rpm, followed by a layer of PMMA A6 at a slower speed of 3000 rpm. This results in a bilayer with a thin bottom layer (210 nm) and a very thick top layer (520 nm). The SNS junction mask is defined using e-beam lithography. The bilayer is then developed in a 1:3 MIBK/IPA solution for 1 minute, which provides a suspended mask. The chip is then loaded into a metal evaporator. Before metal deposition, we perform an ion-milling step using an argon plasma to remove the layer of Nb oxide at the chip surface, which could hinder proper electrical contact between the SNS junction and the voltage leads. Initially, 5 nm of Ti is evaporated to serve as a sticking layer for subsequent metals. This is followed by a 15 nm Au deposition. The sample is then tilted by 22°, and 60 nm of Nb is evaporated. Nb is deposited almost everywhere over the gold layer but not at the constriction level, where it is blocked by the mask of PMMA due to the angle, thus preventing from shorting the SNS junction. The pressure inside the loadlock is maintained below a few mbar throughout the process to ensure a clean SN interface. Finally, the resist is lifted off after evaporation in a hot acetone bath.
Measurement setups
In Figure 4, we represent the three implementations of the experiments presented in this paper.

Figure 4. Schematics of the experimental setups for (a) Josephson emission measurements, (b) two-tone Josephson spectroscopy of a transmon qubit and (c) of a resonator.
The setup in Figure 4a corresponds to one used for the measurement of the Josephson emission, i.e., the experimental data presented in Figure 1c. The power emitted by the junction is directed, via a bias tee, towards a power spectrum analyzer (PSA) through a ZX60-83LN12 and a room-temperature amplifier as well as a 4 − 8 GHz HEMT amplifier from LNF anchored at 4 K. The junction is protected from the noise emitted by the HEMT by means of three 4 − 8 GHz cryogenic isolators, a 12 GHz low-pass filter from K&L, and a 4 − 8 GHz bandpass filter from Micro-Tronics. This setup also allows for the measurement of the current-voltage characteristic of the Josephson emitter. For this, a voltage is applied with a Yokogawa GS200 source through a voltage divider, and both voltage and current are measured using SP1004 voltage amplifiers from Basel Precision Instrument and a Yokogawa DL350 oscilloscope. The current is determined by measuring the voltage across a resistor in series with the Josephson junction, whose resistance value (97 Ω) was measured during the run. For all dc lines, we use Thermocoax coaxial cables, RC low-pass filters and filters, both at 10 mK, to protect the junctions from the noise. Detailed references of most components can be found in Ref. [42].
In Figure 4b, we show the setup used for the two-tone Josephson spectroscopy of a transmon qubit (Figure 2). The dc part of the setup is similar to before, but the rf part is changed significantly. This time the rf port of
the bias tee directs the Josephson emission towards the readout port of a transmon qubit through a 4 − 8 GHz bandpass filter, a 12 GHz lowpass filter and a circulator. The same port is also used for cavity spectroscopy by sending a microwave signal from a ZNB20 vector network analyzer (VNA) through a −60 dB attenuator, a 12 GHz low-pass filter, and then through the −20 dB port of a directional coupler. The latter allows the VNA signal to be combined with the one emitted by the junction, enabling two-tone spectroscopy of the transmon qubit. Once the microwave signal from the VNA is reflected off the cavity, it is redirected back to the VNA, passing through three 4–8 GHz isolators, a 12 GHz low-pass filter, and a similar amplification chain as before.
The final setup, depicted in Figure 4c, corresponds to the two-tone Josephson spectroscopy experiment of the resonator shown in Figure 3. Although not fully represented, the dc part is similar to the previous two setups, with a few minor modifications. Additionally, this setup includes a flux line that allows a current to flow close to the superconducting loop terminating the resonator, enabling us to adjust the magnetic flux in the loop. The spectroscopy of the fundamental mode of the resonator is performed in a standard manner, with a VNA connected to an attenuated input line and an output line with a similar amplification chain as before.
Characterization of SNS junctions
The current-voltage characteristic of the Al-Hf-Al junction, used for the experiments described in Figure 1 and Figure 2, is shown in Figure 5a. Unlike our Nb-Au-Nb junctions, it was not designed for an efficient evacuation of Joule power. Consequently, it exhibits a significantly lower retrapping current compared to its switching current. In order to characterize the Al-Hf-Al junction as a microwave source, we measured the emitted power in a narrow band centered at 5 GHz as a function of the bias voltage (see Figure 5b). The narrow peak close to 10 V, indicated by a red arrow, corresponds to the AC Josephson effect. The continuously increasing background corresponds to incoherent noise, due to shot noise and thermal noise, and remains lower than the AC Josephson emission in the voltage bias range of interest.
Calibration of the Josephson frequency
In order to properly calibrate the frequency of the signal emitted by the SNS junction, it is crucial to accurately know the voltage across its terminals. As we apply a voltage to the bias line of the junction, we simultaneously measure using a low-noise amplifier. As depicted in Figure 6, these two voltages are related by the equation , which is consistent with the RCSJ model [74]. Here is a constant that depends on the attenuation of our line, is the voltage offset between the room-temperature bias source and the sample extracted from the I(V) characteristic, using the supercurrent branch as a reference, and is a constant extracted from the fit. This relation allows us to determine the actual value of , and thus the Josephson frequency , without constantly monitoring it when performing two-tone Josephson spectroscopy.

Figure 6. Voltage across the Nb-Au-Nb SNS junction measured as a function of the bias voltage . This allows us to calibrate the Josephson frequency , which is displayed on the right y axis. The gray region is not accessible as the biasing is unstable due to retrapping of the junction.
Standard rf measurements of the transmon qubit
The complete design of the circuit investigated in Section III, which mostly consists in a transmon qubit capacitively coupled to a microwave resonator, is described in Refs. [51, 52]. We characterized this circuit-QED architecture using standard microwave instruments. Figure 7a shows the transmon qubit spectrum. The extracted transmon frequency GHz matches the value obtained from two-tone Josephson spectroscopy. Note however that the measured linewidth is here about 1 order of magnitude smaller than the one obtained via two-tone Josephson spectroscopy, the latter being limited by the radiation linewidth of the Josephson emitter. Finally, we also observe a second transition at associated to the finite population of Fock state in the cavity.

Figure 7. Spectral and time-domain measurements of the transmon qubit. (a) Two-tone spectroscopy of the transmon qubit. Here the excitation tone is provided by a commercial microwave source. (b) relaxation curve: qubit population as a function of delay time after a pulse, which was calibrated from the Rabi oscillations. The solid line is a fit to an exponential decay that yields s. (c) Ramsey oscillations: qubit population as a function of delay time between two pulses detuned at MHz. The fitting (solid line) yields s.
Going further, we have probed the qubit quantum coherence by performing operations in the time domain. The strong coupling between qubit and cavity allows one to perform single-shot measurements of the qubit state. We could thus measure the probability of the qubit to be in the excited state as a function of time. Using Rabi oscillations, one can calibrate and -pulses. Thus, by applying a -pulse that prepares the qubit in state , we could measure after a variable delay an exponential decay trace with an energy relaxation time (Figure 7a). Finally, we have measured the dephasing for an equal qubit superposition , by applying two -pulses detuned at and separated by a varying delay time. Figure 7b shows the corresponding Ramsey oscillations, which indicate coherent precession of the qubit state in the Bloch sphere around the -axis. From the decay, we could extract a dephasing time . This value being much smaller than , our qubit coherence is dominated by pure dephasing with a characteristic time .
Coupling schemes
One key aspect of two-tone Josephson spectroscopy is to probe high-frequency systems while using a detector with a resonant frequency that can easily be accessed with standard instruments and components (). One thus needs to engineer the proper coupling that will result in a measurable frequency shift of the detector. In section III, the coupling between the transmon qubit and the detector is in the dispersive regime, with a detuning larger than the coupling strength yet small enough to allow the exchange of virtual photons. Following Ref. [50], the frequency shift reads
with the qubit charging energy, the coupling strength and the detuning. Although increasing the transmon qubit frequency would be a whole new work in itself, one can estimate the dispersive coupling in the regime of very large detuning. Considering, for instance, an SNS-based transmon qubit resonating at 100 GHz (designed with , , and ), we estimate , which would translate into a large-enough signal to measure using a high-Q resonator.
More crucially, two-tone Josephson spectroscopy is more adapted to coupling in the adiabatic regime, where the detuning between the system and the detector is so large that virtual processes are strongly suppressed. Still, exciting the system can result in a measurable frequency shift of the detector, which does not depend on the detuning. The adiabatic regime indeed accounts for the renormalization of the resonator frequency by the effective capacitance or inductance of the system. For instance, in the case of a Josephson weak link hosting Andreev bound states, one should achieve large couplings by exploiting the variation of the Josephson inductance in the adiabatic regime, as described in Refs. [?, ?]. Using typical parameters, we estimate a frequency shift of several 100 kHz when exciting the Andreev transition.
Complementary two-tone Josephson spectroscopy measurements
Figure 8 shows the one-tone spectroscopy of the resonator used in Sec. IV, measured at low power and around 5 GHz. This measurement allows us to extract the fundamental frequency . The two-tone spectroscopy measurements are performed by probing the cavity response close to its resonance frequency (as indicated by the red arrow), which translates into a large phase response when a higher-frequency mode is excited.

Figure 8. One-tone spectroscopy of the resonator used in Section IV measured either via the magnitude (left) or the phase (right) of the reflected signal.
In Figure 9a, we present two-tone Josephson spectroscopy measurements performed in an additional sample similar to the one of Figure 3. The mutual inductance between the resonator and the superconducting loop was however designed to be smaller, which reduces the Kerr effect. A dip in the amplitude of the reflected signal indicates the position of the mode of the microwave resonator. As we sweep the bias voltage, the Josephson frequency happens to match the frequency of a higher mode. As the latter is populated by microwave photons, the resonance shifts by a few kHz, which allows the detection of high-frequency modes.
Figure 9. (a) Two-tone Josephson spectroscopy for a sample similar to the one of Figure 3, with a smaller coupling between the resonator and the superconducting loop. Reflection coefficient as a function of the bias voltage . The frequency is referred to the resonance frequency of the cavity in the absence of bias voltage. The green arrows indicate the modes , 3, 4, and 5 that we detect in this experiment. (b) Two-tone Josephson spectroscopy for a sample with a much larger emitter-detector coupling. In this sample, we detect many more modes than the harmonic resonances of the cavity, which we attribute to hybridization of the resonator and the emitter junction circuit lines.
In Figure 9b, we present measurements performed on another device, where the coupling between the Josephson emitter and the resonator was made much stronger. We detect many more resonances than in the measurements presented in the main text; however, they do not correspond to the bare resonances of the resonator. We believe in this case the spectroscopy is too invasive, because the modes of the resonator are perturbed by the emitter due to the strong coupling.
Two-tone Josephson spectroscopy with a tunnel Josephson junction
One of the reasons we opt for an SNS junction over an SIS tunnel junction as the Josephson emitter is due to the SNS junction’s relatively constant impedance when a voltage is applied across its terminals, compared to the tunnel junction. In Figure 10, we show measurements of a two-tone Josephson spectroscopy, similar to Figure 3d, but conducted with an SIS junction as the emitter. We observed significant signal variations with numerous structures resembling resonances. However, they do not necessarily signify mode detections but are sometimes simply due to changes of the resonance frequency of the resonator fundamental mode () caused by impedance variations of the SIS junction. Consequently, conducting two-tone Josephson spectroscopy with this type of junction is challenging.
Figure 10. Two-tone Josephson spectroscopy performed using a tunnel Josephson junction as an emitter and a 7 GHz resonator as a detector. We interpret the features at 100 and 200 GHz as excitation of quasiparticles.
Sign of the resonator frequency shift
In Figure 3e of the main text, we observe that the dispersive coupling between modes and changes sign depending on whether the normalized flux in the superconducting loop, , is 0 or . This behavior is actually consistent with the expected electromagnetic modes of the resonator.
We can understand this by modeling our device as a resonator of length and characteristic impedance , which is open at one end and terminated by an inductance at the other, as shown in Figure 3a. The fundamental mode then has a resonance frequency given approximately by
where is the total inductance of the resonator, and is the bare resonance frequency, with being the propagation speed of microwave photons. The contribution of the superconducting loop to the inductance is expressed as
where is the mutual inductance between the two, is the geometrical inductance of the loop, and is the Josephson inductance of the SNS junction. A slight change in Josephson inductance results in a frequency shift
which can be either positive or negative, depending on the sign of .
Calculating the change in Josephson inductance, , due to mode population requires knowledge of the exact current-phase relation (CPR) of the junction. To estimate this change, we use a simplified model where the SNS junction has a sinusoidal CPR, similar to that of a tunnel junction: . When a high-frequency current, , flows through the junction, its critical current is renormalized as , where is a Bessel function of the first kind. The Josephson inductance then becomes
In the limit of small excitation current (), we find
Thus, the sign of the inductance change is determined by the sign of . Near , is positive, meaning the resonator frequency decreases as increases (i.e. ). Conversely, near , the resonator frequency increases at finite (i.e. ).
Finally, knowing that adding a single photon to the mode at frequency is equivalent to setting , the dispersive shift between modes and 0 can be written as . In our experiment, we estimate , –, and . The SNS junction was designed with a critical current of , corresponding to a Josephson inductance of . This critical current is approximately 10 times smaller than that of the emitter junction, due to the normal metal part being twice as long (), and scaling as for SNS junctions in the long-junction limit. The dispersive shift is thus on the order of 1–10 Hz in our system, with its sign dependent on the sign of . Based on the measured phase shift, this indicates that our Josephson emitter populates the mode with approximately thousands of photons.
Bandwidth of the Josephson emission in the high-frequency modes spectroscopy experiment
In the experiment shown in Figure 3, the peaks of the modes that we observe have a width of approximately 500 MHz. This broadening is not related to the lifetime of these modes but rather to the linewidth of the Josephson emission, which is 100 times larger than in the other experiments presented here. This difference is due to the use of a different filtering setup, which, unlike the other experiments, did not include the 4–8 GHz bandpass filter. We observed that the voltage noise across the junction was significantly higher, resulting in an emission linewidth that was increased by a factor of 100. This is illustrated in Figure 11, where the emission linewidth is several hundred MHz, caused by the increased voltage noise across the junction.


Figure 11. Josephson power measured without the GHz passband filter. (a) as a function of frequency for a voltage across the SNS emitter µV. (b) as a function of both and .
References
References
- [1]A. Blais, R.-S. Huang, A. Wallraff, S. M. Girvin, and R. J. Schoelkopf, Cavity quantum electrodynamics for superconducting electrical circuits: An architecture for quantum computation, Phys. Rev. A 69, 062320 (2004).DOI
- [2]A. Wallraff, D. I. Schuster, A. Blais, L. Frunzio, R.-S. Huang, J. Majer, S. Kumar, S. M. Girvin, and R. J. Schoelkopf, Strong coupling of a single photon to a superconducting qubit using circuit quantum electrodynamics, Nature 431, 162 (2004).pubmed.ncbi.nlm.nih.gov/15356625
- [3]R. J. Schoelkopf and S. M. Girvin, Wiring up quantum systems, Nature 451, 664 (2008).DOI
- [4]A. Blais, S. M. Girvin, and W. D. Oliver, Quantum information processing and quantum optics with circuit quantum electrodynamics, Nat. Phys. 16, 247 (2020).DOI
- [5]A. A. Clerk, K. W. Lehnert, P. Bertet, J. R. Petta, and Y. Nakamura, Hybrid quantum systems with circuit quantum electrodynamics, Nat. Phys. 16, 257 (2020).DOI
- [6]S. Huang, B. Lienhard, G. Calusine, A. Vepsäläinen, J. Braumüller, D. K. Kim, A. J. Melville, B. M. Niedzielski, J. L. Yoder, B. Kannan, T. P. Orlando, S. Gustavsson, and W. D. Oliver, Microwave Package Design for Superconducting Quantum Processors, PRX Quantum 2, 1 (2021).DOI
- [7]A. Anferov, S. P. Harvey, F. Wan, J. Simon, and D. I. Schuster, Superconducting Qubits above 20 GHz Operating over 200 mK, PRX Quantum 5, 030347 (2024).arxiv.org/abs/2402.03031
- [8]B. D. Josephson, Coupled superconductors, Rev. Mod. Phys. 36, 216 (1964).DOI
- [9]K. K. Likharev, Dynamics of Josephson Junctions and Circuits (Routledge, London, 1986).
- [10]I. Yanson, V. Svistunov, and I. Dmitrenko, Experimental Observation of the Tunnel Effect for Cooper Pairs with the Emission of Photons, Sov. Phys. JETP 21, 650 (1965).
- [11]D. N. Langenberg, D. J. Scalapino, and B. N. Taylor, Josephson-Type Superconducting Tunnel Junctions as Generators of Microwave and Submillimeter Wave Radiation, Proc. IEEE 54, 560 (1966).DOI
- [12]A. H. Dayem and C. C. Grimes, Microwave emission from superconducting point-contacts, Appl. Phys. Lett. 9, 47 (1966).DOI
- [13]C. Varmazis, J. E. Lukens, and T. F. Finnegan, Broadband generation of tunable Josephson radiation at microwave frequencies, Appl. Phys. Lett. 30, 660 (1977).DOI
- [14]P. Crozat, G. Vernet, and R. Adde, Direct detection of the Josephson current in Nb superconducting microbridges, Appl. Phys. Lett. 32, 506 (1978).DOI
- [15]J. B. Hansen, M. T. Levinsen, P. E. Lindelof, B. Dueholm, J. Mygind, N. F. Pedersen, and O. H. Soerensen, Subharmonic energy gap structure in the Josephson radiation at 35 GHz from a superconducting thin-film microbridge, Appl. Phys. Lett. 35, 28 (1979).DOI
- [16]R. S. Deacon, J. Wiedenmann, E. Boucquillon, F. Domínguez, T. M. Klapwijk, P. Leubner, C. Brüne, E. M. Hankiewicz, S. Tarucha, K. Ishibashi, H. Buhmann, and L. W. Molenkamp, Josephson radiation from gapless andreev bound states in HgTe-based topological junctions, Phys. Rev. X 7, 1 (2017).
- [17]M. C. Cassidy, A. Bruno, S. Rubbert, M. Irfan, J. Kammhuber, R. N. Schouten, A. R. Akhmerov, and L. P. Kouwenhoven, Demonstration of an ac Josephson junction laser, Science 355, 939 (2017).DOI
- [18]C. Yan, J. Hassel, V. Vesterinen, J. Zhang, J. Ikonen, L. Grönberg, J. Goetz, and M. Möttönen, A low-noise on-chip coherent microwave source, Nat. Electron. 4, 885 (2021).DOI
- [19]M. Devoret, D. Esteve, H. Grabert, and G. Ingold, Effect of the electromagnetic environment on the Coulomb blockade in ultrasmall tunnel junctions, Phys. Rev. Lett. 64, 1824 (1990).pubmed.ncbi.nlm.nih.gov/10041498
- [20]G.-L. Ingold and Y. Nazarov, Charge Tunneling Rates in Ultrasmall Junctions, in Single Charge Tunneling, NATO nen, Tunneling of Cooper pairs across voltage-biased asymmetric single-Cooper-pair transistors, Phys. Rev. B 74, 054504 (2006).
- [30]P.-M. Billangeon, F. Pierre, H. Bouchiat, and R. Deblock, Very High Frequency Spectroscopy and Tuning of a Single-Cooper-Pair Transistor with an On-Chip Generator, Phys. Rev. Lett. 98, 126802 (2007).pubmed.ncbi.nlm.nih.gov/17501148
- [31]J. Basset, H. Bouchiat, and R. Deblock, Emission and Absorption Quantum Noise Measurement with an On-Chip Resonant Circuit, Phys. Rev. Lett. 105, 166801 (2010).DOI
- [32]L. Bretheau, Ç. Ö. Girit, H. Pothier, D. Esteve, and C. Urbina, Exciting Andreev pairs in a superconducting atomic contact, Nature 499, 312 (2013).DOI
- [33]L. Bretheau, Localized Excitations in Superconducting Atomic Contacts : Probing the Andreev Doublet, PhD thesis, Ecole Polytechnique (2013).
- [34]L. Bretheau, Ç. Ö. Girit, C. Urbina, D. Esteve, and H. Pothier, Supercurrent Spectroscopy of Andreev States, Phys. Rev. X 3, 041034 (2013).
- [35]D. J. Van Woerkom, A. Proutski, B. Van Heck, D. Bouman, J. I. Væyrynen, L. I. Glazman, P. Krogstrup, J. Ny gård, L. P. Kouwenhoven, and A. Geresdi, Microwave spectroscopy of spinful Andreev bound states in ballistic semiconductor Josephson junctions, Nat. Phys. 13, 876 (2017).
- [36]V. Y. Safonova, A. V. Gordeeva, A. V. Blagodатkin, D. A. Pimanov, A. A. Yablokov, O. L. Ermolaeva, and A. L. Pankratov, Investigation of Hafnium Thin Films for Design of TES Microcalorimeters, Materials 17, 222 (2024).DOI
- [37]H. Courtois, M. Meschke, J. T. Peltonen, and J. P. Pekola, Origin of hysteresis in a proximity josephson junction, Phys. Rev. Lett. 101, 067002 (2008).DOI
- [38]J. Basset, M. Kuzmanović, P. Virtanen, T. T. Heikkilä, J. Estève, J. Gabelli, C. Strunk, and M. Aprili, Nonadiabatic dynamics in strongly driven diffusive Josephson junctions, Phys. Rev. Res. 1, 1 (2019).DOI
- [39]B. Dassonneville, A. Murani, M. Ferrier, S. Guéron, and H. Bouchiat, Coherence-enhanced phase-dependent dissipation in long SNS Josephson junctions: Revealing Andreev bound state dynamics, Phys. Rev. B 97, 1 (2018).DOI
- [40]D. N. Langenberg, Physical interpretation of the cos φ term and implications for detectors, Rev. Phys. Appliquée 9, 35 (1974).DOI
- [41]U. K. Poulsen, On the cos φ term in Josephson’s expression for the tunneling current, Rev. Phys. Appliquée 9, 41 (1974).DOI
- [42]A. Peugeot, Quantum microwave sources from inelastic Cooper pair tunneling, PhD thesis, Université Paris-Saclay (2020).DOI
- [43]M. Hofheinz, F. Portier, Q. Baudouin, P. Joyez, D. Vion, P. Bertet, P. Roche, and D. Esteve, Bright Side of the Coulomb Blockade, Phys. Rev. Lett. 106, 217005 (2011).DOI
- [44]C. Rolland, A. Peugeot, S. Dambach, M. Westig, B. Kubala, Y. Mukharsky, C. Altimiras, H. Le Sueur, P. Joyez, D. Vion, P. Roche, D. Esteve, J. Ankerhold, and F. Portier, Antibunched Photons Emitted by a dc-Biased Josephson Junction, Phys. Rev. Lett. 122, 186804 (2019).DOI
- [45]A. Peugeot, G. Ménard, S. Dambach, M. Westig, B. Kubala, Y. Mukharsky, C. Altimiras, P. Joyez, D. Vion, P. Roche, D. Esteve, P. Milman, J. Lep-päkangas, G. Johansson, M. Hofheinz, J. Ankerhold, and





