All-optical superconducting qubit readout

Abstract

The rapid development of superconducting quantum hardware is expected to run into substantial restrictions on scalability because error correction in a cryogenic environment has stringent input-output requirements. Classical data centres rely on fibre-optic interconnects to remove similar networking bottlenecks. In the same spirit, ultracold electro-optic links have been proposed and used to generate qubit control signals, or to replace cryogenic readout electronics. So far, these approaches have suffered from either low efficiency, low bandwidth or additional noise. Here we realize radio-over-fibre qubit readout at millikelvin temperatures. We use one device to simultaneously perform upconversion and downconversion between microwave and optical frequencies and so do not require any active or passive cryogenic microwave equipment. We demonstrate all-optical single-shot readout in a circulator-free readout scheme. Importantly, we do not observe any direct radiation impact on the qubit state, despite the absence of shielding elements. This compatibility between superconducting circuits and telecom-wavelength light is not only a prerequisite to establish modular quantum networks, but it is also relevant for multiplexed readout of superconducting photon detectors and classical superconducting logic.

Keywords: Applied optics; Qubits.

Fig. 1. Comparison of conventional and optical qubit readout set-ups in a dilution refrigerator.

a, Typical microwave in–microwave out set-up consisting of carefully thermalized coaxial cables, attenuators, filters, circulators, a driven parametric amplifier (faded) and a d.c.-biased high-electron-mobility-transistor amplifier, all of which are approximately wavelength sized (centimetres). Note that the components are inserted above the respective temperature stage to make the illustration more compact. b, Reduced microwave in–optics out readout set-up replacing the microwave output components with an optically driven, resonant EO transducer. c, All-optical, optics in–optics out circulator-free qubit readout based on simultaneous microwave downconversion and upconversion of an optical carrier. Here, all cryogenic microwave components are replaced by a single EO transceiver.

Fig. 2. Conventional and optical single-shot readout of a superconducting qubit.

a–c, Sketches of the different readout schemes involving a microwave cavity with bare resonance frequency ω_c dispersively coupled to a transmon qubit (qubit-cavity system in jade) and the EO transducer, consisting of a second microwave cavity (blue-grey) at ω_e = ω_c coupled to an optical whispering gallery mode resonator (light blue). The qubit state was prepared by means of a separate port at ω_q. The EO transducer was operated with an optical pump pulse at ω_p to parametrically enhance the interconversion of microwave ω_e and optical ω_o signals. Conventional microwave readout: a microwave pulse probed the qubit-cavity system and was detected by means of microwave heterodyne detection (a). Optical detection of a microwave readout tone: the microwave pulse reflected from the qubit-cavity system was upconverted to the optical domain and detected with optical heterodyne detection (b). All-optical readout: a modulated optical carrier was converted to the microwave domain to probe the qubit-cavity system. Its reflection was simultaneously converted back to the optical domain and detected with an optical heterodyne set-up (c). d–f, Averaged time traces (d), (e) and (f) of the measured heterodyne signal powers corresponding to the readout schemes shown in a, b and c postselected on successful measurements of the prepared qubit state (g∣g and e∣e) based on 15,000 independent trials. Grey lines show theoretical predictions which are expected to deviate for $∣\mathrm{e}⟩$ before the steady state is reached (see text and Supplementary Information). The shaded areas highlight the difference between both qubit-state responses, for the interval where we extract the weighting functions f = I_e − I_g for the temporal in-phase quadrature integration. The inset in e is a normalized measurement of the optical pump power. For reference, panel f also shows the simulated optical response of the EO transducer without the reflection from the qubit-cavity system, that is, only due to electro-optically induced transparency (EOIT). g–i, Histograms of 15,000 single shots obtained by integrating the weighted in-phase quadrature f(t)I(t) shown in g, h and i with state assignment fidelities ${\mathcal{F}}_{\mathrm{EE}}\text{,}{\mathcal{F}}_{\mathrm{OE}}\text{and}{\mathcal{F}}_{\mathrm{OO}}$, corresponding to the readout schemes a, b and c.

Fig. 3. Qubit coherence for different readout methods.

a, Measured excited state detection probability $P_{\det }\left(∣\mathrm{e}⟩\right)$ after a π pulse for varying measurement delays t using the three different readout methods shown in Fig. 2. b, Measured Ramsey oscillations using two π/2 pulses separated by a variable delay t and detuned by ~2 MHz from the qubit transition for the three readout methods.

Fig. 4. Impact of the optical pump.

a, Measured state assignment fidelities ${\mathcal{F}}_1$ and ${\mathcal{F}}_2$ of two consecutive JPA-assisted microwave measurements (filled circles) and corresponding QND metric ($\mathcal{Q}$, crosses) obtained in the presence of a 2-μs-long optical pump pulse of ~0.14 W applied during the first readout as a function of repetition rate and calculated dissipated optical power (top axis) together with theory (lines and 3σ confidence bands). Approximately $1-{\left(1-2{\eta }_{\mathrm{o}}\right)}^2\approx 69\%$ of the average optical power sent to the sample was dissipated in the device. Empty circles (mostly overlapping with filled circles) denote measurements for which the optical pump was applied also during state preparation. Data are represented as mean ±3σ but error bars are smaller than the marker size. The insets show pulse sequences for the differently triggered measurements. Qubit preparation, readout and optical pump are denoted by ω_q, ω_e and ω_p, respectively. b, Measured qubit coherence times $\left(T_1,T_2^{*}\right)$ when the optical pulse was synchronized with each qubit preparation and readout pulse (empty circles) and for a free-running measurement sequence (filled circles) versus optical pulse repetition rate. Squares indicate the mean of the optical readout results in Fig. 3. The decrease in T₁ and $T_2^{*}$ was accurately modelled with theory (red and blue line with 3σ confidence band), based on the measured thermal occupancy shown in c, the expected quasiparticle distribution and Purcell decay (red line). Data error bars show the two-sided 90% confidence interval of the mean according to a Student t distribution for five measurements (compare with Fig. 3). c, Measured temperature of the mixing chamber plate (yellow dots) and the different microwave modes (dots) together with power law fits as a guide to the eye. Mean values and error bars stem from the respective fits with 3σ confidence bands and corresponding error propagation calculations.