betaFeedback

WorksPhysics and Astronomy

Addressing spins at the clock transitions with a frequency- and bandwidth-tunable superconducting resonator

Abstract

Solid-state spin ensembles addressed via superconducting circuits are promising candidates for quantum memory applications, offering multimodal storage capability and second-long coherence times at their clock transition. Implementing practical memory schemes requires dynamic control over both the resonator frequency and bandwidth. In this letter, we report measurements of a superconducting resonator whose frequency can be tuned by passing a DC current through the high-kinetic-inductance thin film, and whose bandwidth can be tuned by parametric coupling to a low-Q buffer resonator. Using this resonator, we address an ensemble of bismuth donors at their clock transition, measuring a Hahn-echo coherence time of 450 ms. We demonstrate RF driving of the bismuth donor hyperfine transitions, as well as dynamic bandwidth control of the resonator.

Despite continuous improvements, the limited coherence time remains a key challenge for superconducting quantum processors, constraining both noisy intermediate-scale quantum (NISQ) algorithms [1] and quantum error correction schemes [2]. A promising solution [3] relies on dedicated quantum memory units capable of parallel storage of multiple qubit states over extended periods. Amongst various platforms explored to this end [7, 8, 9] [4, 5, 6], solid-state spin ensembles stand out for their multimodal storage capability [9, 10, 11, 12, 13] and second-long coherence times [14]. An example of architecture consists in encoding the qubit states in single- or few-photon microwave pulses to be stored in and retrieved from the spin ensemble memory through refocussing sequences [15, 16, 17, 18]. One difficulty of this approach is to achieve efficient absorption of the wave packet in the spin ensemble. Even with the use of high-quality-factor resonators to enhance the spin-photon interaction strength, efficient photon absorption requires relatively large spin density, which leads to broad ensemble spin linewidth and reduced spin coherence time due to dipolar interactions [19].

One possible way to solve this issue is to bias spins at specific magnetic fields where their mean magnetic dipoles vanish, allowing for long coherence times even at large spin concentrations. These so-called clock transitions (CTs), or zero first-order-insensitive (ZEFOZ) points, occur in coupled electron-nuclear spin systems with large hyperfine interactions. Enhanced coherence times at CTs have been observed in donors in silicon [14], molecular spins [20], and rare-earth-ion-doped crystals [21, 22]. Because CTs occur only at specific magnetic fields, they require resonators tuned precisely to the corresponding transition frequency. While this can be achieved via trial and error with fixed-frequency resonators [9], a more general and scalable approach uses resonators whose frequency can be tuned in-situ and dynamically to the CT frequency. Additionally, several quantum memory protocols [18, 23] also require dynamical control of the resonator bandwidth, to avoid maser emission when the spin ensemble is inverted by the control pulses.

Aluminium-based Josephson devices offer many tools to design resonators with tunable characteristics [24], but they are usually incompatible with the application of magnetic fields above ∼10 mT\sim10\,\mathrm{mT} that are necessary to bias the spins. Another mechanism, compatible with large magnetic fields, arises from the quadratic dependence of the kinetic inductance of a superconducting wire, Lk(I)=Lk0+αI2L_k(I) = L_{k0} + \alpha I^2 on the current II passing through it. Field-resilient kinetic inductance (KI)-based frequency-tunable resonators have been demonstrated [25, 26, 27] and used for interfacing spins at ∼200−500 mT\sim200-500\,\mathrm{mT} fields [28]. Here, we design and fabricate a frequency- and bandwidth-tunable resonator based on KI. We demonstrate its suitability as a spin-ensemble quantum memory interface by addressing bismuth donor spins in silicon (Bi:Si) at their CT where we measure extended coherence times, and we perform a spectroscopy of Bi:Si sub-gigahertz transitions by electron nuclear double resonance (ENDOR).

The circuit schematics is shown in Figure 1. The core part of the device is a pair of superconducting microwave resonators (“resonator A” and “resonator B” in Figure 1c and the subcircuits of matching colours in Figure 1a) that are inductively coupled via a small mutual inductance, called the kinetic inductance coupler (KIC). Resonator A is a high quality-factor resonator which inductively couples to the spin ensemble. It consists of a 800 nm-wide inductive wire with large KI (referred to as microwave A) and a capacitor in parallel. Resonator B is a low-Q auxiliary that acts as a tunable microwave buffer for resonator A. Its inductor is 4 μm\mu\mathrm{m} wide. The KIC comprises a pair of 5 μm\mu\mathrm{m}-long, 250 nm-wide nanowires (Figure 1d). An inductor is tunable when it owes a large fraction of its value to kinetic inductance, as ensured by lateral confinement for microwire A and the KIC. Microwire B is much wider, and has next to no tunability, as we confirm shortly thereafter. The resonator frequency fAf_A can be tuned by passing a DC current IAI_A through microwire A. Its bandwidth tuning is achieved via three-wave mixing with resonator B through the KIC [26, 29]. Both tuning processes require low-frequency or DC bias of the nonlinear inductors. To this end we flank the dual resonator on either side with notch filters, centred at fAf_A and fBf_B respectively. These so-called “Bragg mirrors” [30], consisting in successions of quarter-wave-length coplanar waveguides (CPW) of alternating impedances, can maintain high quality factors while permitting DC current and radio-frequency (RF) pump tones to pass through. The separate control of DC currents IAI_A and IBI_B injected into the devices ports implies independent biases of microwire A (IAI_A) and the KIC (IA+IBI_A+I_B), so that the bandwidth and frequency tunabilities are in principle decoupled. Note that the Bragg mirror A contains twelve {high-ZZ, low-ZZ} repetitions, whereas the Bragg mirror B contains only four. As a result, microwave transmission through port A is less than 103 s−110^3\,\mathrm{s}^{-1} and thus negligible, so for both resonators the only relevant microwave coupling are the ones through Bragg mirror B, the rate of which we denote as κcA\kappa_c^A and κcB\kappa_c^B hereinafter. Accordingly, the internal loss rates are denoted by κiA(B)\kappa_i^{A(B)}, and the respective total linewidths κA(B)=κiA(B)+κcA(B)\kappa^{A(B)}=\kappa_i^{A(B)}+\kappa_c^{A(B)}. The whole circuit is patterned on a 50 nm-thick NbTiN film, on which we measured kinetic inductance per square Lkinsq=2.2L_{\mathrm{kin}}^{\mathrm{sq}}=2.2 pH/□\square. The circuit is cooled to 10 mK in a dilution refrigerator, with appropriate attenuation and filtering to suppress thermal noise.

Figure 1. Device layout. (a) Schematic circuit diagram. The lumped element models of resonator A (blue) and resonator B (red) are highlighted in (c). The kinetic inductance coupler (KIC) is shown as an inductive shunt to the ground. (b) Overview micrograph of the circuit substrate before flip chip bonding. The green box outlines the footprint of the Bi-doped silicon die which is subsequently glued on top. The external polarising field BzB_z is applied parallel to the microwire of resonator A, i.e., horizontal in this image. (c) Zoom-in of the resonator region (brown box in (b)). Resonators A and B are marked with dashed boxes. The arrows point to the KIC. (d) Optical and scanning electron micrographs of the KIC nanowire. The bright areas of various shades are NbTiN films with different vortex-trapping hole densities. The darkest regions are the exposed silicon substrate.

We first characterize the circuit by microwave reflectometry through port B. In absence of bias current, we find fA=7.422f_A=7.422 GHz, κcA=9.4×104 s−1\kappa_c^A=9.4\times10^4\,\mathrm{s}^{-1}, κiA=7.5×105 s−1\kappa_i^A=7.5\times10^5\,\mathrm{s}^{-1}; fB=6.605f_B=6.605 GHz, κcB=2.6×107 s−1\kappa_c^B=2.6\times10^7\,\mathrm{s}^{-1}, κiB=5.7×106 s−1\kappa_i^B=5.7\times10^6\,\mathrm{s}^{-1}. The resonator frequency shift Δf\Delta f is then measured as functions of IAI_A and IBI_B (see Figure 2ab). We observe that resonator A can be tuned over an ∼80\sim80 MHz range, whereas resonator B shifts by less than 10 MHz, as designed. All tuning curves fit well to simple Ginzburg-Landau theory [25], from which we can extract the critical current of microwire A, 9.53 mA, and that of the KIC, 5.73 mA. We note the critical currents are proportional to the total cross-section areas, as expected [31].

Figure 2. Resonator tunabilities. (a-b) Resonance frequency shift of the two modes in response to sweeping DC bias current IAI_A (a) or IBI_B (b), under different IBI_B or IAI_A offsets. Solid curves: fits to Ginzburg-Landau theory [25]. (c-d) Amplitude of the microwave reflection S11S_{11} off port B near the resonance frequency of mode B (c) and A (d), as the RF pump frequency f3WMf_{\mathrm{3WM}} is swept across the frequency difference fA−fBf_A-f_B. (e) Dependence of the measured mode A linewidth κiA\kappa_i^A (open circles) on the RF pump amplitude, under various DC bias through IBI_B while IAI_A is held neutral. (f) Ringdown suppression via dynamic control of resonator bandwidth. The green curves of varying shades represent the ringdown amplitude after a resonant microwave pulse, as the 3WM pump pulses are delayed by 0 μs0\ \mu\mathrm{s}, 1 μs1\ \mu\mathrm{s}, ..., 4 μs4\ \mu\mathrm{s}. The pulse sequences are shown in insets with the varying parameters marked in red with arrows. (g) Resonance frequency fAf_A (left axis) and internal loss rate κiA\kappa_i^A (right axis) of mode A as functions of the magnetic field BzB_z applied in-plane, parallel to microwire A.

In addition to frequency tuning, the KIC creates a three-wave mixing (3WM) term (a+a†)(b+b†)2(a+a^\dagger)(b+b^\dagger)^2 in the circuit Hamiltonian [32], where aa (bb) is the annihilation operator of mode A (B), and a†a^\dagger (b†b^\dagger) the creation operator. When a strong drive tone is applied through port B at frequency f3WM=fA−fBf_{\mathrm{3WM}}=f_A-f_B, this term simplifies to a frequency-conversion Hamiltonian g3WM(ab†+a†b)g_{\mathrm{3WM}}(ab^\dagger+a^\dagger b) [33], with g3WMg_{\mathrm{3WM}} being proportional to the pump amplitude P3WM\sqrt{P_{\mathrm{3WM}}} and DC current IA+IBI_A+I_B flowing through the KIC. Since resonator B has a significantly larger decay rate than resonator A, whenever the 3WM is enabled, it translates into an additional loss rate κiA=4g3WM2/κB\kappa_i^A=4g_{\mathrm{3WM}}^2/\kappa_B for resonator A (see Figure 2c and Figure 2d) [34]. The linear dependence of κiA\kappa_i^A on P3WMP_{\mathrm{3WM}} and (IA+IB)2(I_A+I_B)^2 is confirmed in Figure 2e. The bandwidth can also be tuned dynamically. It is for instance possible to suddenly accelerate the resonator field decay, as demonstrated in Figure 2f.

We finally measure the resonator A frequency and internal loss rate as a function of a magnetic field BzB_z applied approximately parallel to the sample surface and to the microwire (Figure 2f). The frequency decreases due to the kinetic inductance dependence on magnetic field.

The loss rate shows an increase around 260 mT, and otherwise remains constant except for a slight increase close to 500 mT. A closer look into the 260 mT feature reveals a narrow peak that we tentatively ascribe to dangling bond spins at the Si/SiO2_2 interface with g=2.0g = 2.0 (so-called Pb0 centers) [35], and a broader peak centred at g=1.922g = 1.922, similar to the one called US0 in [36] and that has been ascribed to a Ti-related paramagnetic defect.

We now demonstrate that the frequency- and bandwidth-tunable resonator can be used to address spins in solids. A 28^{28}Si-enriched silicon chip is bonded on top of microwire A, the surface layer facing the resonator pre-implanted with 209^{209}Bi+^{+} ions. Bismuth is a donor for silicon, which resumes a neutral state at low temperatures. Its spin Hamiltonian is H=(γeS+γnI)Bz+AI⋅SH=(\gamma_{\mathrm{e}}\mathbf{S}+\gamma_{\mathrm{n}}\mathbf{I})B_z+A\mathbf{I}\cdot\mathbf{S}, with S\mathbf{S} (I\mathbf{I}) being the electron (nuclear) spin operator, γe/2π=−28\gamma_{\mathrm{e}}/2\pi=-28 GHz/T (γn/2π=8\gamma_{\mathrm{n}}/2\pi=8 MHz/T) the electron (nuclear) spin gyromagnetic ratio, and A/2π=1.47507A/2\pi=1.47507 GHz the hyperfine coupling constant [40]. At Bz=0B_z=0, the energy states are eigenstates of the total angular momentum operator, F=I+S\mathbf{F}=\mathbf{I}+\mathbf{S}, with eigenvalue F=4F=4 for the ground-state manifold of 9 levels, and F=5F=5 for the excited state manifold of 11 levels. Application of a small magnetic field lifts the degeneracy between levels with different mm values of the FF projection on zz (see Figure 3a). An oscillating magnetic field perpendicular to BzB_z can induce transitions between these levels whenever Δm=±1\Delta m=\pm1. This can occur at microwave frequency for transitions between levels with F=4F=4 and F=5F=5 (electron spin resonance (ESR)–like transitions), or at radio-frequency between levels within the same FF manifold (nuclear magnetic resonance (NMR)–like transitions). Owing to this peculiar energy-level diagram, several CTs can be found in the Bi:Si spectrum [14, 37]. Here we will concentrate primarily on the one at 25.6 mT and 7.3382 GHz [40].

Figure 3. Tracking bismuth spin transitions with a tunable resonator. (a) Calculated bismuth donor energy spectrum as a function of the polarising field BzB_z [37, 38]. The arrows mark the clock transitions characterised in Figure 4, and the relevant spin states are coloured green. (b) The resonator absorption spectrometry, measured by the internal loss rate κiA\kappa_i^A of mode A as its resonance frequency ff is swept across the tuning range, and BzB_z between 0 and 65 mT. The contrast is enhanced and the field-independent background is removed for visibility. The red dashed curves indicate the calculated transition frequencies. (c) The absorption spectrum (open circles) at 2.1 mT. The bismuth transition peaks are indicated by the gray dash lines. Inset: the extracted bismuth transition peak width Γ\Gamma versus the gyromagnetic ratio γ\gamma in units of the Bohr magneton γe\gamma_e. The dashed line represents the homogeneous linewidth corresponding to an Overhauser field δB0=4 μT\delta B_0 = 4\ \mu\mathrm{T} due to the residual 500 ppm of 29Si{}^{29}\mathrm{Si} [30, 39].

We perform the spin spectroscopy, first by measuring the resonator loss rate as a function of frequency fAf_{\mathrm{A}} (by application of IAI_{\mathrm{A}}) and polarising field BzB_z. We subtract the field-independent background and plot the change in internal loss rate ΔκA\Delta\kappa_{\mathrm{A}} in Figure 3(b). BzB_z-dependent resonant losses are observed, and their position matches the ESR-like transitions of Bi:Si. The Bz=2.1B_z=2.1 mT data are plotted in Figure 3c; the Bi:Si resonances have linewidth of ∼300\sim300 kHz, indicative of significantly lower strain than in devices where the metallic resonator was deposited directly on top of the crystal [41, 42, 43], but they are not yet at the homogenous linewidth limit. We remove part of the spectrum around Bz=9B_z=9 mT where aluminium bonding wires were transiting from superconducting to normal states.

The Bi:Si spectrum at the CT is measured by sweeping the resonator frequency fAf_{\mathrm{A}} and recording the amplitude of a Hahn echo. It consists of two resolved peaks of width ∼90\sim90 kHz, corresponding respectively to the transitions ∣4,−1⟩↔∣5,0⟩|4,-1\rangle\leftrightarrow|5,0\rangle and ∣4,0⟩↔∣5,−1⟩|4,0\rangle\leftrightarrow|5,-1\rangle (see Figure 3a). The energy relaxation time is measured as T1=53T_1=53 s by an inversion recovery sequence (see Figure 4a). This is two orders of magnitude shorter than typical non-radiative relaxation times in Bi:Si at 10 mK, indicating that the donor spins are well in the Purcell regime [44] [45]. We then measure the echo amplitude as a function of the Hahn echo delay τ\tau (see Figure 4b). The data are well-fitted by an exponential decay, with a time constant T2=450T_2=450 ms, similar to those already observed at Bi:Si CTs using fixed-frequency resonators [9, 14]. This demonstrates that we can address the Bi:Si donor spins at their CT, using a frequency-tunable resonator. Echo emission can be controlled by dynamically tuning the resonator away from resonance at the time of the echo formation, as already demonstrated in Ref. [46, 47]. Such echo silencing is useful in several quantum memory protocols [15, 23] [48].

Figure 4. Clock transitions and electron nuclear double resonance (ENDOR). The open circles represents the raw data, and the solid curves Lorentzian (a, e) or exponential (b-d) fits. The pulse sequences are shown in insets with the varying parameters marked in red with arrows. (a) Hahn echo spectroscopy at 25.6 mT. T1T_1 (b) and T2T_2 (c) measurement at 7.3382 GHz (dashed grey line in (a)). (d) Echo silencing via resonator shifting. The normalised echo magnitude decays with increasing resonance frequency detuning ΔfA\Delta f_{\mathrm{A}} of mode A during the period of echo. (e) Normalised microwave echo magnitude (open circles) as a function of the disruptive radio frequency pulse frequency fNMRf_{\mathrm{NMR}}. The grey lines indicate the nuclear magnetic resonance (NMR)–like transitions. The ones involving ∣4,1⟩|4,1\rangle, ∣4,0⟩|4,0\rangle, ∣5,0⟩|5,0\rangle, or ∣5,1⟩|5,1\rangle are highlighted with solid lines. Left inset: the extracted peak width Γ\Gamma of the NMR-like transitions versus their gyromagnetic ratio γn\gamma_{\mathrm{n}}. The dashed line represents the homogenous linewidth corresponding to a magnetic noise δB0=4 μ\delta B_0=4\,\muT.

The ability to send sub-gigahertz pulses is harnessed

to drive NMR-like transitions between Bi:Si levels. To that goal, the amplitude of a Hahn echo is measured as a function of the frequency fNMRf_{\mathrm{NMR}} of a radio frequency pulse sent in-between the two echo control pulses. With the magnetic field set to Bz=13.49B_z = 13.49 mT and the resonator to 7422.5 MHz, we primarily probe the ∣4,0⟩↔∣5,1⟩\lvert4,0\rangle\leftrightarrow\lvert5,1\rangle transition, though the ∣4,1⟩↔∣5,0⟩\lvert4,1\rangle\leftrightarrow\lvert5,0\rangle transition also contributes slightly due to spectral overlap. The echo signal diminishes when fNMRf_{\mathrm{NMR}} matches an NMR-like transition involving one of these states. This is evident in Figure 4(e), where the dips on the normalised echo magnitude correspond to the relevant NMR-like transition spectrum, from low to high frequency: ∣5,2⟩↔∣5,1⟩\lvert5,2\rangle\leftrightarrow\lvert5,1\rangle, ∣4,2⟩↔∣4,1⟩\lvert4,2\rangle\leftrightarrow\lvert4,1\rangle, ∣5,1⟩↔∣5,0⟩\lvert5,1\rangle\leftrightarrow\lvert5,0\rangle, ∣4,1⟩↔∣4,0⟩\lvert4,1\rangle\leftrightarrow\lvert4,0\rangle, ∣5,0⟩↔∣5,−1⟩\lvert5,0\rangle\leftrightarrow\lvert5,-1\rangle, ∣4,0⟩↔∣4,−1⟩\lvert4,0\rangle\leftrightarrow\lvert4,-1\rangle. In particular, the ∣4,2⟩↔∣4,1⟩\lvert4,2\rangle\leftrightarrow\lvert4,1\rangle, ∣5,0⟩↔∣5,−1⟩\lvert5,0\rangle\leftrightarrow\lvert5,-1\rangle dips are visibly shallower than the other four, as expected from the lower contribution of levels ∣4,1⟩\lvert4,1\rangle and ∣5,0⟩\lvert5,0\rangle to the echo signal. Other NMR-like transitions have no impact on the echo amplitude.

In conclusion, we have demonstrated a field-resilient resonator with tunable frequency and bandwidth, suitable for addressing paramagnetic impurities. Such device should find applications in microwave quantum memories, as well as nanoscale magnetic resonance spectroscopy [49–51].

Acknowledgments

The authors acknowledge the technical support of P. Sénat, D. Duet, P.-F. Orfila, and S. Delprat, and fruitful discussions within the Quantronics group. The authors acknowledge the support of the AIDAS joint laboratory, of Région Ile-de-France through the DIM SIRTEQ, of the Agence Nationale de la Recherche under the Chaire Industrielle NASNIQ, and under the PEPR Plan Project ROBUSTSUPERQ, and IARPA and Lincoln Labs for providing the Josephson travelling-wave parametric amplifier.

Supplemental materials to “Addressing spins at the clock transitions with a frequency- and bandwidth-tunable superconducting resonator”

Sample

The bottom chip containing the superconducting device is made out of 50-nm-thick NbTiN deposited on Silicon. Our fabrication starts from a 2” Si wafer (high-resistivity grade supplied from Siltronics), that is deoxidised in 5% HF for two minutes before NbTiN sputtering. The device fabrication process comprises three lithography steps, using either a 30 keV Raith electron-beam lithography system or a Heidelberg Instruments Maskless Aligner. Afterwards the pattern is either converted to 30 nm of aluminium hard mask via lift-off, or we use directly the developed photoresist as a soft mask for reactive ion etching to remove the appropriate parts of the NbTiN film or the silicon substrate.

The cover chip is a silicon chip topped by a 800 nm-thick epilayer of 28Si{}^{28}\mathrm{Si}. Bismuth atoms have been implanted in this layer with the following implantation energies and fluences:

Ion Energy (keV)Area Dose (Ions/cm2^2)
23001.585×10121.585 \times10^{12}
20001.2184×10121.2184 \times10^{12}
14008.9772×10118.9772 \times10^{11}
9006.6804×10116.6804 \times10^{11}
5003.1748×10113.1748 \times10^{11}
3601.0517×10111.0517 \times10^{11}
1207.0692×10107.0692 \times10^{10}
Total Dose4.8626×10124.8626 \times10^{12}

Table 1.

We can then estimate the implantation profile using Stopping and Range of Ions in Matter (SRIM) simulation, as shown in Figure S1.

Image not extracted

Figure S1. Stopping and Range of Ions in Matter (SRIM) simulation of the bismuth donor concentration as a function of the distance from the surface.

The two chips have been assembled in a flip-chip bonding machine, glued together by PMMA 950A6 only at the contact pillars of the base chip (elliptical structures at the ends of the green box in Figure 1b). Two 200 μ\mum-deep reservoirs are etched out around the pillars (dark pentagon regions) to contain excess PMMA, preventing it from wetting the entire device region through capillary action.

Measurement setup

The cryogenic and room temperature setups are shown in Figure S2.

Figure S2. Experimental setup at room temperature and at low-temperature. The microwave tone driving the spins is controlled through IQ-mixing as well as with fast switches. Its transmission, or any signal emitted from the spins, is amplified by a series of four amplifiers: a Josephson parametric amplifier provided by the Lincoln lab, a HEMT amplifier, and two mini-circuits amplifiers. The DC currents controlling the KIC and microwire inductances are generated by Yokogawa voltage sources. Both channels are fitted with home-made RC filters (225 Ω225 \ \Omega, 220 pF220 \ \mathrm{pF}). On the buffer side, the dc current is combined with the 3 wave-mixing tone using a bias-tee at room-temperature, and is further recombined with the microwave signal at the mixing-chamber stage. On the other side, the current is recombined at low temperature with the rf drive for nuclear spin manipulation. An Anapico APSUASYN30-4 generates the 4 microwaves tones necessary to the experiments (rf, three-wave mixing, JTWPA pump, and microwave drive tone). A Quantum Machine OPX generates the low-frequency control signals (yellow squares) and perform the acquisition of the low-frequency signals (red squares). The experiment is realized in a Bluefors LD system equipped with a 3D vector magnet. Some unused components are omitted for simplicity.

Dissipation engineering through three-wave mixing using kinetic inductance non-linearity

Kinetic inductance modulation

In this section, we detail the action of the three-wave mixing drive on the coupled resonators system. This drive induces a modulation of the kinetic inductance of the microwire and of the nanowire. The frequencies of resonators AA and BB show a quartic dependence on their biasing current (see Figure 2(a-Figure 2b)), so that the kinetic inductance dependence on current can be modelled in all generality as

Lk(I)=Lk0[1+(II∗)2+α(II∗)4],(1)L_k(I) = L_k^0 \left[1+\left(\frac{I}{I^*}\right)^2+\alpha\left(\frac{I}{I^*}\right)^4\right], \tag*{(1)}

where α≲1\alpha\lesssim1 can be determined by fitting the experimental data. When applying a rf current drive IrfI_{\rm rf} at fixed bias IdcI_{\rm dc}, the inductance modulation thus expresses as

L(Idc+Irf)=Lg+Lk(Idc)+Lk(0)∑i=14ciIrfi,(2)L(I_{\rm dc}+I_{\rm rf})=L_g+L_k(I_{\rm dc})+L_k(0)\sum_{i=1}^{4}c_i I_{\rm rf}^i, \tag*{(2)}

where we have included the geometrical contribution LgL_g and where:

c1/I∗=2IdcI∗+4α(IdcI∗)3,(3)c_1/I^*=2\frac{I_{\rm dc}}{I^*}+4\alpha\left(\frac{I_{\rm dc}}{I^*}\right)^3, \tag*{(3)}
c2/I∗2=1+6α(IdcI∗)2,(4)c_2/I^{*2}=1+6\alpha\left(\frac{I_{\rm dc}}{I^*}\right)^2, \tag*{(4)}
c3/I∗3=4α(IdcI∗),(5)c_3/I^{*3}=4\alpha\left(\frac{I_{\rm dc}}{I^*}\right), \tag*{(5)}
c4/I∗4=α(6)c_4/I^{*4}=\alpha \tag*{(6)}

For small drive currents (Irf≪I∗I_{\rm rf}\ll I^*) and at small finite bias (Idc<I∗I_{\rm dc}<I^*), one can safely approximate the kinetic inductance modulation up to its second order expansion, i.e. discarding all terms in α\alpha.

Circuit Hamiltonian

We now wish to diagonalize the circuit (Figure S3), taking into account the kinetic inductance non-linearity. Denoting CaC_a and CbC_b (LaL_a and LbL_b) the capacitance (inductance) of each resonator, and LcL_c the coupling KIC inductor, the Lagrangian of the circuit is [52]:

Device circuit diagram

Figure S3. Device circuit diagram.

L=−Qa22Ca−Qb22Cb+La2Q˙a2+Lb2Q˙b2+Lc2(Q˙a+Q˙b)2,(7)\mathcal{L}=-\frac{Q_a^2}{2C_a}-\frac{Q_b^2}{2C_b}+\frac{L_a}{2}\dot{Q}_a^2+\frac{L_b}{2}\dot{Q}_b^2+\frac{L_c}{2}\left(\dot{Q}_a+\dot{Q}_b\right)^2, \tag*{(7)}

where QaQ_a and QbQ_b are the charge on each capacitor. We consider that only LaL_a and LcL_c are non-linear inductors, so that their values depend on the currents flowing through them (respectively Q˙a\dot{Q}_a and Q˙a+Q˙c\dot{Q}_a+\dot{Q}_c) as described above.

La=La0+c1aQ˙a+c2aQ˙a2(8)L_a=L_a^0+c_1^a\dot{Q}_a+c_2^a\dot{Q}_a^2 \tag*{(8)}
Lc=Lc0+c1n(Q˙a+Q˙b)+c2n(Q˙a+Q˙b)2.(9)L_c=L_c^0+c_1^n(\dot{Q}_a+\dot{Q}_b)+c_2^n(\dot{Q}_a+\dot{Q}_b)^2. \tag*{(9)}

The Hamiltonian of the circuit is given by:

H=ΦaQ˙a+ΦbQ˙b−L,(10)H=\Phi_a\dot{Q}_a+\Phi_b\dot{Q}_b-\mathcal{L}, \tag*{(10)}

where the phase operators can be evaluated through Φi=∂L∂Q˙i\Phi_i=\frac{\partial\mathcal{L}}{\partial\dot{Q}_i}, yielding:

Φa=(La0+Lc0)Q˙a+Lc0Q˙b+32c1aQ˙a2+32c1n(Q˙a+Q˙b)2+2c2aQ˙a3+2c2n(Q˙a+Q˙b)3(11)\Phi_a=(L_a^0+L_c^0)\dot{Q}_a+L_c^0\dot{Q}_b+\frac{3}{2}c_1^a\dot{Q}_a^2+\frac{3}{2}c_1^n(\dot{Q}_a+\dot{Q}_b)^2+2c_2^a\dot{Q}_a^3+2c_2^n(\dot{Q}_a+\dot{Q}_b)^3 \tag*{(11)}
Φb=(Lb+Lc0)Q˙b+Lc0Q˙a+32c1n(Q˙a+Q˙b)2+2c2n(Q˙a+Q˙b)3.(12)\Phi_b=(L_b+L_c^0)\dot{Q}_b+L_c^0\dot{Q}_a+\frac{3}{2}c_1^n(\dot{Q}_a+\dot{Q}_b)^2+2c_2^n(\dot{Q}_a+\dot{Q}_b)^3. \tag*{(12)}

Expressing the Hamiltonian purely as a function of (Φa,Φb,Qa,Qb)(\Phi_a,\Phi_b,Q_a,Q_b) requires inverting Eq. (12) to express Q˙a/b\dot{Q}_{a/b} as a function of Φa\Phi_a and Φb\Phi_b. We do so by defining the vectors Φ=(Φa,Φb)\boldsymbol{\Phi}=(\Phi_a,\Phi_b) and Q˙=(Q˙a,Q˙b)\dot{\mathbf{Q}}=(\dot{Q}_a,\dot{Q}_b), and express Eq. (12) as:

Φ=ΛQ˙+R,(13)\boldsymbol{\Phi}=\Lambda\dot{\mathbf{Q}}+\mathbf{R}, \tag*{(13)}

where Λ\Lambda is a 2-by-2 matrix independent of Q˙a/b\dot{Q}_{a/b} and RR contains the terms in Q˙a/bk\dot{Q}_{a/b}^{k} with k>1k>1. We now have:

Q˙=Λ−1Φ−Λ−1R.(14)\dot{\mathbf{Q}}=\Lambda^{-1}\boldsymbol{\Phi}-\Lambda^{-1}\mathbf{R}. \tag*{(14)}

One can then express the higher terms contained in RR by substituting them by Eq. (14). This creates a new expression for Q˙\dot{\mathbf{Q}}. Performing this process iteratively enables to express Q˙a/b\dot{Q}_{a/b} as a function of Φa\Phi_a, Φb\Phi_b, and higher order cross-products, and these expressions can then be trimmed at a given order before injection into the Hamiltonian. Cutting-off the resulting expressions at the third order (up to Φa/b3\Phi_{a/b}^3), we find:

H=H0+g1,1ΦaΦb+g2,1Φa2Φb+g1,2ΦaΦb2+g3,0Φa3+g0,3Φb3(15)H=H_0+g_{1,1}\Phi_a\Phi_b+g_{2,1}\Phi_a^2\Phi_b+g_{1,2}\Phi_a\Phi_b^2+g_{3,0}\Phi_a^3+g_{0,3}\Phi_b^3 \tag*{(15)}

where:

H0=Qa22Ca+Qb22Cb+Φa22L~a+Φb22L~b,(16)H_0=\frac{Q_a^2}{2C_a}+\frac{Q_b^2}{2C_b}+\frac{\Phi_a^2}{2\widetilde{L}_a}+\frac{\Phi_b^2}{2\widetilde{L}_b}, \tag*{(16)}
L~a=La0+Lc011+Lc0/Lb0,(17)\widetilde{L}_a=L_a^0+L_c^0\frac{1}{1+L_c^0/L_b^0}, \tag*{(17)}
L~b=Lb+Lc011+Lc0/La,(18)\widetilde{L}_b=L_b+L_c^0\frac{1}{1+L_c^0/L_a}, \tag*{(18)}
g1,2=−32L~ac1a(Lc0La0Lb(1+Lc0/Lab))2−32c1n1(Lb+Lc0)2(La0+Lc0)La0LbL~aL~b(19)g_{1,2}=-\frac{3}{2\widetilde{L}_a}c_1^a\left(\frac{L_c^0}{L_a^0L_b(1+L_c^0/L_{ab})}\right)^2-\frac{3}{2}c_1^n\frac{1}{(L_b+L_c^0)^2(L_a^0+L_c^0)}\frac{L_a^0L_b}{\widetilde{L}_a\widetilde{L}_b} \tag*{(19)}

and others gi,jg_{i,j} are geometrical terms that can be expressed as a function of Lc0L_{\mathrm{c}}^{0}, La0L_{\mathrm{a}}^{0}, LbL_{\mathrm{b}}, c1ac_{1}^{a}, c1nc_{1}^{n}, but do not involve c2ac_{2}^{a} and c2bc_{2}^{b}. We defined Lab=La0LbLa0+LbL_{\mathrm{ab}}=\frac{L_{\mathrm{a}}^{0}L_{\mathrm{b}}}{L_{\mathrm{a}}^{0}+L_{\mathrm{b}}}. We now introduce the annihilation operators a^=(Φa+iZaQa)/2ℏZa\hat{a}=(\Phi_{\mathrm{a}}+iZ_{\mathrm{a}}Q_{\mathrm{a}})/\sqrt{2\hbar Z_{\mathrm{a}}} and b^=(Φb+iZbQb)/2ℏZb\hat{b}=(\Phi_{\mathrm{b}}+iZ_{\mathrm{b}}Q_{\mathrm{b}})/\sqrt{2\hbar Z_{\mathrm{b}}}, where Za=L~a/CaZ_{\mathrm{a}}=\sqrt{\widetilde{L}_{\mathrm{a}}/C_{\mathrm{a}}} and Zb=L~b/CbZ_{\mathrm{b}}=\sqrt{\widetilde{L}_{\mathrm{b}}/C_{\mathrm{b}}} are the impedances of the coupled resonators. The Hamiltonian S15 can be re-expressed as a function of these operators.

Three-wave mixing for dissipation engineering

We now wish to derive the effective three-wave Hamiltonian, and show that it can be used to control the dissipation rate of resonator AA. We thus consider the action of a pump tone of amplitude ξ\xi with frequency ωp∼ωa−ωb\omega_{\mathrm{p}}\sim\omega_{\mathrm{a}}-\omega_{\mathrm{b}} applied on the buffer port, as well as a drive tone applied at ωd∼ωa\omega_{\mathrm{d}}\sim\omega_{\mathrm{a}}. We accomplish this by considering the interacting Hamiltonian Hint=UHU†−iℏUdU†dtH_{\mathrm{int}}=UHU^{\dagger}-i\hbar U\frac{dU^{\dagger}}{dt} calculated using the following evolution operator

U=eiωdta^†a^ei(ωd−ωp)tb^†b^e−ξ~b^†+ξ~∗b^.(20)U=e^{i\omega_{\mathrm{d}}t\hat{a}^{\dagger}\hat{a}}e^{i(\omega_{\mathrm{d}}-\omega_{\mathrm{p}})t\hat{b}^{\dagger}\hat{b}}e^{-\widetilde{\xi}\hat{b}^{\dagger}+\widetilde{\xi}^{*}\hat{b}}. \tag*{(20)}

where we remove the classical displacement due to the application of pump on the buffer flux. The displacement ξ~=ξe−iωpt\widetilde{\xi}=\xi e^{-i\omega_{\mathrm{p}}t} can be related to the rf drive current flowing through the inductor by ξ=L~bIrf/2ℏZb\xi=\widetilde{L}_{\mathrm{b}}I_{\mathrm{rf}}/\sqrt{2\hbar Z_{\mathrm{b}}}. In our experiment, owing to the Bragg mirror, this current can be safely related to the input pump power on the buffer port PinP_{\mathrm{in}} as Pin/Z0\sqrt{P_{\mathrm{in}}/Z_{0}}, since the pump tone is applied far below the Bragg mirror bandpass frequency range.

Keeping only terms of HintH_{\mathrm{int}} which are non zero after a time-average over one drive period, we effectively perform the rotating-wave approximation and we find:

HRWA=ℏδaa^†a^+ℏδbb^†b^+ℏg3WM(a^b^†+a^†b^)(21)H_{\mathrm{RWA}}=\hbar\delta_{\mathrm{a}}\hat{a}^{\dagger}\hat{a}+\hbar\delta_{\mathrm{b}}\hat{b}^{\dagger}\hat{b}+\hbar g_{\mathrm{3WM}}(\hat{a}\hat{b}^{\dagger}+\hat{a}^{\dagger}\hat{b}) \tag*{(21)}

where we have δa=ωa−ωd\delta_{\mathrm{a} }=\omega_{\mathrm{a}}-\omega_{\mathrm{d}}, δb=ωb+ωp−ωd\delta_{\mathrm{b}}=\omega_{\mathrm{b}}+\omega_{\mathrm{p}}-\omega_{\mathrm{d}} and:

ℏg3WM=4ξ(2ℏ)3ZaZb2g1,2=12kIdcIrfI∗2ℏωaωb.(22)\hbar g_{\mathrm{3WM}}=4\xi\sqrt{(2\hbar)^3Z_{\mathrm{a}}Z_{\mathrm{b}}^{2}}g_{1,2}=12k\frac{I_{\mathrm{dc}}I_{\mathrm{rf}}}{I_{*}^{2}}\hbar\sqrt{\omega_{\mathrm{a}}\omega_{\mathrm{b}}}. \tag*{(22)}

with k≈Lk,n0/LaLbk\approx L_{k,n}^{0}/\sqrt{L_{\mathrm{a}}L_{\mathrm{b}}} and where we have neglected the contribution of LaL_{\mathrm{a}} to the third-order non-linearity. We now derive the quantum Langevin equations from Eq. (21), including the dissipation on modes aa and bb, as well as a drive of strength αin\alpha_{\mathrm{in}} applied to mode aa:

a˙=−(κa2+iδa)a−ig3WMb+κacα(23)\dot{a}=-\left(\frac{\kappa_{\mathrm{a}}}{2}+i\delta_{\mathrm{a}}\right)a-ig_{\mathrm{3WM}}b+\sqrt{\kappa_{\mathrm{a}}^{\mathrm{c}}}\alpha \tag*{(23)}
b˙=−(κb2+iδb)b−ig3WMa(24)\dot{b}=-\left(\frac{\kappa_{\mathrm{b}}}{2}+i\delta_{\mathrm{b}}\right)b-ig_{\mathrm{3WM}}a \tag*{(24)}

In the steady-state, whenever δb=0\delta_{\mathrm{b}}=0, we find:

a=2κacαinκa+4g3WM2κb+2iδa,(25)a=\frac{2\sqrt{\kappa_{\mathrm{a}}^{\mathrm{c}}}\alpha_{\mathrm{in}}}{\kappa_{\mathrm{a}}+\frac{4g_{\mathrm{3WM}}^{2}}{\kappa_{\mathrm{b}}}+2i\delta_{\mathrm{a}}}, \tag*{(25)}

where we can identify that through the three-wave mixing interaction, the buffer mode creates an additional decay channel for mode aa of rate 4g3WM2/κb4g_{\mathrm{3WM}}^{2}/\kappa_{\mathrm{b}}.

References

References

  1. [7]T. Zhong, J. M. Kindem, J. G. Bartholomew, J. Rochman, I. Craiciu, E. Miyazono, M. Bettinelli, E. Cavalli, V. Verma, S. W. Nam, F. Marsili, M. D. Shaw, A. D. Beyer, and A. Faraon, Science 357, 1392 (2017).
  2. [8]S. Freer, S. Simmons, A. Laucht, J. T. Muhonen, J. P. Dehollain, R. Kalra, F. A. Mohiyaddin, F. E. Hudson, K. M. Itoh, J. C. McCallum, D. N. Jamieson, A. S. Dzurak, and A. Morello, Quantum Science and Technology 2, 015009 (2017).
  3. [9]V. Ranjan, J. O’Sullivan, E. Albertinale, B. Albanese, T. Chanalière, T. Schenkel, D. Vion, D. Esteve, E. Flurin, J. J. L. Morton, and P. Bertet, Physical Review Letters 125, 210505 (2020).
  4. [10]D. I. Schuster, A. P. Sears, E. Ginosar, L. DiCarlo, L. Frunzio, J. J. L. Morton, H. Wu, G. A. D. Briggs, B. B. Buckley, D. D. Awschalom, and R. J. Schoelkopf, Physical Review Letters 105, 140501 (2010).
  5. [11]C. Grezes, Y. Kubo, B. Julsgaard, T. Umeda, J. Isoya, H. Sumiya, H. Abe, S. Onoda, T. Ohshima, K. Nakamura, I. Diniz, A. Auffeves, V. Jacques, J.-F. Roch, D. Vion, D. Esteve, K. Molmer, and P. Bertet, Comptes Rendus Physique Quantum microwaves / Micro-ondes quantiques, 17, 693 (2016).
  6. [12]S. Probst, A. Bienfait, P. Campagne-Ibarcq, J. J. Pla, B. Albanese, J. F. Da Silva Barbosa, T. Schenkel, D. Vion, D. Esteve, K. Mølmer, J. J. L. Morton, R. Heeres, and P. Bertet, Applied Physics Letters 111, 202604 (2017).
  7. [13]J. O’Sullivan, O. W. Kennedy, K. Debnath, J. Alexander, C. W. Zollitsch, M. Šiménas, A. Hashim, C. N. Thomas, S. Withington, I. Siddiqi, K. Mølmer, and J. J. L. Morton, Physical Review X 12, 041014 (2022), publisher: American Physical Society.
  8. [14]G. Wolfowicz, A. M. Tyryshkin, R. E. George, H. Riemann, N. V. Abrosimov, P. Becker, H.-J. Pohl, M. L. W. Thewalt, S. a. Lyon, and J. J. L. Morton, Nature Nanotechnology 8, 561 (2013).
  9. [15]M. Afzelius, N. Sangouard, G. Johansson, M. U. Staudt, and C. M. Wilson, New Journal of Physics 15, 065008 (2013).
  10. [16]B. Julsgaard and K. Mølmer, Physical Review A 86, 063810 (2012), publisher: American Physical Society.
  11. [17]B. Julsgaard and K. Mølmer, Physical Review A 88, 062324 (2013).
  12. [18]J. Z. Bernád, M. Schilling, Y. Wen, M. M. Müller, T. Calarco, P. Bertet, and F. Motzoi, Journal of Physics B: Atomic, Molecular and Optical Physics 58, 035501 (2025).
  13. [19]G. Wolfowicz, S. Simmons, A. M. Tyryshkin, R. E. George, H. Riemann, N. V. Abrosimov, P. Becker, H.-J. Pohl, S. A. Lyon, M. L. W. Thewalt, and J. J. L. Morton, Physical Review B 86, 245301 (2012).
  14. [20]M. Shiddiq, D. Komijani, Y. Duan, A. Gaita-Ariño, E. Coronado, and S. Hill, Nature 531, 348 (2016).
  15. [21]J. Alexander, G. Dold, O. W. Kennedy, M. Šiménas, J. O’Sullivan, C. W. Zollitsch, S. Welinski, A. Ferrier, E. Laffite-Houssat, T. Lindström, P. Goldner, and J. J. L. Morton, Physical Review B 106, 245416 (2022), publisher: American Physical Society.
  16. [22]A. Tiranov, E. Green, S. Hermans, E. Liu, F. Chiossi, D. Serrano, P. Loiseau, A. M. Kumar, S. Bertaina, A. Faraon, and P. Goldner, Sub-second spin and lifetime-limited optical coherences in $^{171}$Yb$^{3+}$:CaWO$_4$ (2025), arXiv:2504.01592 [quant-ph].arxiv.org/abs/2504.01592
  17. [23]B. Julsgaard, C. Grezes, P. Bertet, and K. Molmer, Physical Review Letters 110, 10.1103/PhysRevLett.110.250503 (2013).
  18. [24]P. Scarlino, J. H. Ungerer, D. J. Van Woerkom, M. Mancini, P. Stano, C. Müller, A. J. Landig, J. V. Koski, C. Reichl, W. Wegscheider, T. Ihn, K. Ensslin, and A. Wallraff, Physical Review X 12, 031004 (2022).
  19. [25]A. J. Annunziata, D. F. Santavicca, L. Frunzio, G. Catelani, M. J. J. Rooks, A. Frydman, and D. E. Prober, Nanotechnology 21, 445202 (2010).
  20. [26]M. R. Vissers, J. Hubmayr, M. Sandberg, S. Chaudhuri, C. Bockstiegel, and J. Gao, Applied Physics Letters 107, 062601 (2015).
  21. [27]Y. Wu, C. Wang, D. Wang, M. Xu, Y. Zhou, and H. X. Tang, Physical Review Applied 24, 024015 (2025).
  22. [28]A. T. Asfaw, A. J. Sigillito, A. M. Tyryshkin, T. Schenkel, and S. A. Lyon, Applied Physics Letters 111, 032601 (2017).
  23. [29]M. Malnou, M. Vissers, J. Wheeler, J. Aumentado, J. Hubmayr, J. Ullom, and J. Gao, PRX Quantum 2, 010302 (2021).
  24. [30]E. Abe, H. Wu, A. Ardavan, and J. J. L. Morton, Applied Physics Letters 98, 251108 (2011).
  25. [31]T. Van Duzer, Principles of Superconductive Devices and Circuits, 2nd ed. (Prentice Hall PTR, 1999).
  26. [32]D. J. Parker, M. Savyt skiy, W. Vine, A. Laucht, T. Duty, A. Morello, A. L. Grimsmo, and J. J. Pla, Physical Review Applied 17, 034064 (2022).
  27. [33]P. K. Tien, Journal of Applied Physics 29, 1347 (1958).
  28. [34]See Supplemental Material.
  29. [35]E. H. Poindexter, P. J. Caplan, B. E. Deal, and R. R. Razouk, Journal of Applied Physics 52, 879 (1981).
  30. [36]A. Bahr, M. Boselli, B. Huard, and A. Bienfait, Applied Physics Letters 124, 114004 (2024).
  31. [37]M. H. Mohammady, G. W. Morley, and T. S. Monteiro, Physical Review Letters 105, 067602 (2010).
  32. [38]G. W. Morley, M. Warner, A. M. Stoneham, P. T. Greenland, J. Van Tol, C. W. M. Kay, and G. Aeppli, Nature Materials 9, 725 (2010).
  33. [39]R. E. George, W. Witzel, H. Riemann, N. V. Abrosimov, N. Nötzel, M. L. W. Thewalt, and J. J. L. Morton, Physical Review Letters 105, 067601 (2010).
  34. [40]The hyperfine coupling constant A, and consequently the CT configurations, are strain-dependent [43] and thus vary between samples [14, 38, 53].
  35. [41]A. Bienfait, J. J. Pla, Y. Kubo, M. Stern, X. Zhou, C. C. Lo, C. D. Weis, T. Schenkel, M. L. W. Thewalt, D. Vion, D. Esteve, B. Julsgaard, K. Mølmer, J. J. L. Morton, and P. Bertet, Nature Nanotechnology 11, 253 (2016).
  36. [42]J. J. Pla, A. Bienfait, G. Pica, J. Mansir, F. A. Mohiyaddin, Z. Zeng, Y. M. Niquet, A. Morello, T. Schenkel, J. J. L. Morton, and P. Bertet, Physical Review Applied 9, 044014 (2018).
  37. [43]V. Ranjan, B. Albanese, E. Albertinale, E. Billaud, F. Flanigan, J. J. Pla, T. Schenkel, D. Vion, D. Esteve, E. Flurin, J. J. L. Morton, Y. M. Niquet, and P. Bertet, Physical Review X 11, 031036 (2021).
  38. [44][44*] A. Bienfait, J. Pla, Y. Kubo, X. Zhou, M. Stern, C.-C. Lo, C. Weis, T. Schenkel, D. Vion, D. Esteve, J. Morton, and

Paper details

vlnc:W_m492upxr
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 11

downloads · 0

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

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents