Sympathetic cooling of a radio-frequency LC circuit to its ground state in an optoelectromechanical system

Nicola Malossi,1, 2 Paolo Piergentili,1, 2 Jie Li,3, 4 Enrico Serra,5, 6, 7 Riccardo Natali,1, 2 Giovanni Di Giuseppe,1, 2 and David Vitali1, 2, 8 1Physics Division, School of Science and Technology, University of Camerino, I-62032 Camerino (MC), Italy 2INFN, Sezione di Perugia, via A. Pascoli, I-06123 Perugia, Italy 3Kavli Institute of Nanoscience, Department of Quantum Nanoscience, Delft University of Technology, 2628CJ Delft, The Netherlands 4Zhejiang Province Key Laboratory of Quantum Technology and Device, Department of Physics and State Key Laboratory of Modern Optical Instrumentation, Zhejiang University, Hangzhou 310027, China 5Institute of Materials for Electronics and Magnetism, Nanoscience-Trento-FBK Division, I-38123 Povo, Trento, Italy 6INFN, Trento Institute for Fundamental Physics and Application, I-38123 Povo, Trento, Italy 7Dept. of Microelectronics and Computer Engineering /ECTM/DIMES, Delft University of Technology, Feldmanweg 17, 2628 CT Delft, The Netherlands 8CNR-INO, Largo Enrico Fermi 6, I-50125 Firenze, Italy (Dated: March 23, 2021) We present a complete theory for laser cooling of a macroscopic radio-frequency LC electrical circuit by means of an optoelectromechanical system, consisting of an optical cavity dispersively coupled to a nanome- chanical oscillator, which is in turn capacitively coupled to the LC circuit of interest. The driven optical cavity cools the mechanical resonator which in turn sympathetically cools the LC circuit. We determine the optimal parameter regime where the LC resonator can be cooled down to its quantum ground state, which requires a large optomechanical cooperativity, and a larger electromechanical cooperativity. Moreover, comparable op- tomechanical and electromechanical coupling rates are preferable for reaching the quantum ground state.

I. INTRODUCTION to their quantum ground state, and here we show that this can be achieved by appropriately engineering the interactions Over the past decade, the experimental realization of in a hybrid tripartite optoelectromechanical system. This quantum states of macroscopic objects has made significant result could be considered as a further example of quan- progress in the fields of opto- and electromechanics. These tum manifestation at macroscopic level, involving photons include mechanical ground state cooling [1–5], mechani- with macroscopic wavelengths and typically realised with cal squeezing [6,7], entanglement between mechanical, mi- macroscopically-sized circuit elements. crowave and optical modes [8–12]. Also facilitated by this Laser cooling of an LC circuit via the intermediate cou- progress, hybrid quantum systems [13] provide interesting op- pling to a mechanical resonator has been first proposed in portunities and a variety of novel platforms for new techno- Ref. [15]. Here we extend that analysis, showing that one can logical applications. In particular, optoelectromechanical de- cool down the LC resonator to its quantum ground state, pro- vices has received significant attention, especially in transduc- viding an alternative route to what has been recently demon- ing radio-frequency (rf) and microwave signals to the optical strated through the coupling to a superconducting qubit [35], domain [14–32]. or to an ultra-cryogenic microwave cavity [36]. In this paper However, most of optoelectromechanical systems are using we provide a detailed analysis of the system, by first determin- a GHz microwave resonator. Here, we focus onto the case of ing its optimal working point, and then analysing its stationary a MHz rf resonator, for which operation in the quantum do- state, focusing onto the parameter regime in which the rf LC main is more difficult because, due to the lower resonance fre- resonant circuit can be ground state cooled. From a physical quency, it is normally in a thermally excited state even at ultra- point of view, this occurs when the energy exchange process cryogenic temperatures. Radio-frequency signals in the MHz of the LC circuit with its own thermal reservoir is dominated and kHz regimes are used in a large variety of research field by the exchange process with the much colder reservoir rep- and applications [14], ranging for example from astronomical resented by the mechanical resonator cooled by the driven op- signal detection at long wavelength (Astronomical Plasmas, tical cavity mode. In more intuitive terms, the driven cavity arXiv:2009.04421v2 [quant-ph] 22 Mar 2021 sun activity and exoplanets research) [33] to ultra-low Mag- cools the mechanical resonator which in turn sympathetically netic field Nuclear Magnetic resonance and imaging (SQUID cools the rf resonator [37]. In general, we find that ground coupled to an LC circuit) [34]. Therefore the possibility of op- state cooling of the rf resonator is possible when the optome- erating in a quantum regime at the MHz and even kHz range chanical cooperativity is large, and the electromechanical co- with extremely low noise can be advantageous either for po- operativity is even larger. A preliminary study of the quantum sitioning, timing and sensing (imaging) applications, and for behavior of the same optoelectromechanical system has been more fundamental science applications, such as the sensitive recently shown in Ref. [38], which however focused only on detection of rf-signals of astrophysical nature. the entanglement between the mechanical and the rf resonator. Quantum operation of rf circuits requires cooling them close Ground state cooling and stationary entanglement are gener- 2

where   † † −iωLt iωLt Hˆopt = ~ω(x)a ˆ aˆ + i~E aˆ e − aeˆ , (2) pˆ2 mω2 xˆ2 Hˆ = + 0 , (3) mech 2m 2 φˆ2 qˆ2 Hˆ = + − qˆVˆ . (4) LC 2L 2C(x ˆ) FIG. 1: Schematic description of the system. A metal coated nanomembrane is coupled via radiation pressure to a cavity field, In the optical contribution we consider a specific cavity mode, † and capacitively coupled to an rf resonant circuit via the position- with photon annihilation (creation) operatora ˆ (ˆa ), with the † dependent capacitance Cm(x). The rf resonator is modelled as a usual bosonic commutation relations [ˆa, aˆ ] = 1, which is lumped-element RLC series circuit with an additional tunable capac- driven by a laser of frequency ωL and input power√ P. Conse- itance C0 in parallel with Cm(x), a resistance R, and an inductance L. quently, the driving rate can be written as E = 2κinP/~ωL, The rf-circuit is driven by a DC bias VDC and by the Johnson-Nyquist with κin the cavity amplitude decay rate through the input port. voltage noise δV. The mechanical Hamiltonian corresponds to a resonator with mass m, displacement operatorx ˆ and conjugated momentum pˆ, with commutation relation [x ˆ, pˆ] = i~, which is associ- ally related quantum phenomena, but, as already verified in ated to a given vibrational mode of the metalized membrane optomechanics [39, 40], they are optimised under quite differ- with bare frequency ω0. The dispersive optomechanical cou- ent conditions. pling arises due to the dependence of the cavity mode fre- The paper is organized as follows. In Sec.II, we intro- quency ω(x ˆ) upon the membrane displacementx ˆ, as discussed duce our tripartite optoelectromechanical system and provide in Refs. [41–45]. its Hamiltonian and the corresponding Langevin equations. In The electrical contribution HˆLC refers to the rf resonator, Sec. III we determine the working point of the system and which we will describe here as a lumped-element series RLC derive the linearized equations for the system quantum fluc- circuit (see Fig. 1), whose dynamical variables are given by tuations. In Sec.IV we show how to exactly solve these lin- the concatenated flux φˆ and the total capacitor chargeq ˆ, with earized equations and determine the steady state of the system, the canonical commutation relation [ˆq, φˆ] = i~. We have also while in Sec.V we provide an approximate analytical theory included a driving term associated with the possibility to con- for the steady state occupancy of the rf resonator. In Sec.VI trol the circuit via a voltage bias Vˆ . The electromechanical we describe the results and determine the optimal parameter coupling is capacitive and it arises from the displacement de- regime for laser cooling the LC circuit to its quantum ground pendence of the effective circuit capacitance C(x ˆ). In the case state. Then, in Sec. VII we discuss in detail the challenges of the chosen optoelectromechanical setup based on a metal- one has to face for an unambiguous detection of the stationary ized membrane, such as those of Refs. [15, 19, 20, 24], one state of rf resonator, while Sec. VIII is for concluding remarks. can write

ε0Aeff C(x ˆ) = C0 + Cm(x ˆ) = C0 + , (5) h0 + xˆ II. THE SYSTEM i.e., the effective capacitance is the parallel of a tunable capac- itance C0 with the capacitor formed by the metalized mem- We consider a generic hybrid optoelectromechanical sys- brane together with the electrodes in front of it. As shown in tem, which consists of an optical cavity, a nanomechanical Refs. [20, 24], we can assume a parallel plate model and de- oscillator, and a radiofrequency (rf) resonant circuit. Differ- fine the effective area Aeff of the membrane capacitor; h0 is ent kind of systems and configurations have been already pro- the steady state distance between the membrane and the elec- posed and characterized experimentally [15–22, 24, 25, 27– trodes, in the absence of any bias voltage and cavity laser driv- 31] and the treatment presented here can be applied to all the ing, while ε0 is the vacuum permittivity. cases in which the electromechanical coupling is capacitive, A realistic RLC circuit is always quite involved, with its and the optomechanical coupling is dispersive. Nonetheless, behaviour determined by a number of parasitic capacitances in order to be more specific, we will refer to the configuration and resistances whose values depend upon the specific circuit in which the optomechanical system is the membrane-in-the- implementation. However, the simplified description adopted middle (MIM) one [41–45], i.e., a driven optical Fabry-Perot´ here in terms of the three lumped-element effective quanti- cavity with a thin semitransparent membrane inside. The ties, the inductance L, the resistance R, and the capacitance membrane is metalized [15, 19, 20, 24, 46] and capacitively C(x ˆ), is possible and perfectly suited for our purposes. In fact, coupled via an electrode to an LC resonant circuit formed by a our goal is to laser cool the rf circuit via its quasi-resonant coil and additional capacitors, see Fig.1. The Hamiltonian of interaction with the mechanical resonator, and the dynami- the system can be written in general as the sum of an optical, cal behaviour is essentially determined by the frequency com- mechanical and electrical term, ponents around the rf resonance peak, which is characterised by two easily measurable quantities, the rf-resonant frequency ˆ ˆ ˆ ˆ (0) H = Hopt + Hmech + HLC, (1) ωLC and the width, γLC, of the uncoupled resonator. The two 3

(0) quantities define the rf-circuit quality factor QLC = ωLC/γLC, i.e., the sum of a DC bias and the Johnson-Nyquist voltage which must be large enough, QLC  1, in order to achieve noise operator δVˆ with autocorrelation function [47, 49], an appreciable cooling [15]. A third circuit quantity that 1 can be directly measured is its effective inductance L, which hδVˆ (t)δVˆ (t0) + δVˆ (t0)δVˆ (t)i (14) can be obtained from the low frequency behaviour of the cir- 2 Z ! cuit. Therefore,√ since in a high-Q series RLC circuit one has dω 0 ~ω (0) = R cos ω(t − t )~ω coth , ωLC = 1/ LC and γLC = R/L, once that the value of the cir- 2π 2kBT cuit inductance L has been measured, one can define the other two effective circuit parameters as which again, in the case of a large LC quality factor can be approximated with the Markovian expression hδVˆ (t)δVˆ (t0) + ε0Aeff 1 ˆ 0 ˆ i ' ~ (0) − 0 C(0) = C0 + ≡ , (6) δV(t )δV(t) /2 R ωLC(2¯nLC + 1)δ(t t ), wheren ¯ LC = h (0) i2 (0) h0 ~ω /kBTLC −1 L ωLC [e LC − 1] is the mean thermal rf photon number. We have assumed in general T T because the rf circuit tends R ≡ LγLC. (7) LC , to pick up ambient noise and the effective rf noise temperature The full quantum dynamics of the system and its stationary can be larger than ambient temperature. state can be determined from the Heisenberg-Langevin equa- tions of the system which are obtained from the Hamiltonian of Eq. (1) and by including fluctuation-dissipation processes III. WORKING POINT AND LINEARIZED DYNAMICS OF for the three resonators, which in the frame rotating at the laser THE QUANTUM FLUCTUATIONS frequency ωL, are given by ˙ In order to look for the possibility to reach the quantum xˆ = pˆ/m, (8) regime for the macroscopic rf resonator, we have to evalu- 2 ∂ω † ate the stationary quantum fluctuations around the classical pˆ˙ = −mω xˆ − γm pˆ − ~ (x ˆ)ˆa aˆ 0 ∂xˆ steady state of the system, which is obtained by replacing all qˆ2 ε A the operators in the Heisenberg-Langevin equations Eqs. (8)- − 0 eff + Fˆ, (9) 2 (12) with the corresponding average values, neglecting all 2 [C0(h0 + xˆ) + ε0Aeff] noise terms, and setting all the derivatives to zero. In this way φˆ qˆ˙ = , (10) one defines the working point of the system, which is deter- L mined by the two external drivings, i.e., the laser driving rate ˙ qˆ E V φˆ = − − γLCφˆ + VDC + δVˆ , (11) and the DC bias voltage DC. If stability conditions are sat- C0 + ε0Aeff/(h0 + xˆ) isfied (see Appendix), the steady state is characterized by the p p aˆ˙ = i[ωL − ω(x ˆ)]a ˆ − κaˆ +E+ 2κinaˆin + 2κexaˆex, (12) cavity mode in a coherent state with amplitude αs, the mem- brane in an equilibrium position displaced by xs, the rf circuit where γm is the mechanical damping rate, and κ = κin + κex with no current, and the capacitor with a stationary charge qs. is the total cavity amplitude decay rate, given by the sum of Using the fact that ps = φs = 0, one can express the working the decay rate though the input port κin and the decay rate point parameters in terms of xs only, i.e., through all the other ports κex. The latter optical loss processes are associated with the corresponding input noise operators E αs = , (15) aˆin anda ˆex, which are uncorrelated and whose only nonzero κ + i∆ † 0 0 h i − qs = C(xs)VDC, (16) correlation is aˆj(t)a ˆj (t ) = δ(t t ), j = in, ex. We have included two zero-mean noise terms in the equa- tions: Fˆ(t) is the Langevin force operator which accounts where ∆ = ω(xs) − ωL is the effective cavity mode detuning, for the Brownian motion of the mechanical oscillator, whose and it is the parameter which is actually fixed in an experiment symmetrized correlation function is in general equal to [47, by the cavity stabilisation system. The static membrane dis- 48] placement xs is the solution of the equilibrium condition for the three forces applied to the membrane, i.e., the membrane 1 elastic force, the electrostatic force and the radiation pressure hFˆ(t)Fˆ(t0) + Fˆ(t0)Fˆ(t)i (13) 2 force , ! Z dω ~ω = mγ cos ω(t − t0)~ω coth , 2 m 2 ε0AeffVDC ∂ω 2π 2kBT mω x − − ~ x n , 0 s = 2 ( s) cav (17) 2(h0 + xs) ∂x which, in the case of a large mechanical quality factor 2 2 2 2 Qm = ω0/γm  1 valid here, can be approximated with where ncav = |αs| = E /(κ + ∆ ) is the intracavity photon the Markovian expression [48], hFˆ(t)Fˆ(t0) + Fˆ(t0)Fˆ(t)i/2 ' number. 0 ~ω0/kBT −1 mγm~ω0(2¯nm + 1)δ(t − t ), wheren ¯m = [e − 1] is the The quantum fluctuations dynamics can be described with equilibrium mean thermal phonon number, with kB the Boltz- very good approximation by linearizing the exact Heisenberg- mann constant and T the environmental temperature. We have Langevin equations Eqs. (8)-(12) around the classical station- also rewritten the external bias voltage as Vˆ (t) = VDC + δVˆ (t), ary state defining the system working point, i.e., by keeping 4 only first order terms in such fluctuations. In fact, if sta- bility conditions are satisfied, the system dynamics will not significantly depart from the steady state defined above, and higher order terms in the fluctuation operators can be ne- glected [39, 40]. It is convenient to express these equations in terms of dimensionless fluctuation operators, scaled by the corresponding quantum zero-point fluctuation units, i.e., by redefining s ~ xˆ → xs + xzpfδxˆ = xs + δxˆ, (18) mω0 p pˆ → ps + pzpfδpˆ = ~mω0δpˆ, (19) s ~ qˆ → qs + qzpfδqˆ = qs + δqˆ, (20) Lω(0) LC FIG. 2: Electro-mechanical coupling g versus the DC voltage VDC q and the membrane-electrode distance h . The black dotted line in- φˆ → φ + φ δφˆ = ~Lω(0) δφ,ˆ (21) 0 s zpf LC dicates the value of h0 which is used in the plots of Sec.VI, cor- responding to 2 µm. The other electro-mechanical parameters are: so that the commutation relations are rewritten as [δxˆ, δpˆ] = (0) 6 −10 ω0/2π = ωLC /2π = 1 MHz, Qm = 10 , m = 0.7 × 10 kg, L = 1 [δqˆ, δφˆ] = i. By introducing also the two intracavity quadra- −7 2 mH, Aeff = 1.1 × 10 m . ture fluctuation operators

iθ † −iθ ˆ δaeˆ + δaˆ e δX = √ , (22) while the modified mechanical resonance frequency ωm is 2 given by the expression δaeˆ iθ − δaˆ†e−iθ δYˆ = √ , (23) 2 2 i 2 ~ ∂ ω(x ) V ε0Aeff ω2 ω2 s n − DC . m = 0 + 2 cav 3 (33) where θ = arctan ∆/κ, one gets the following linearized m ∂x m(h0 + xs) Heisenberg-Langevin equations 2 We recall that the system is stable provided that ωm > 0 and δxˆ˙ = ω0δpˆ, (24) the latter expression shows that there is a maximum value for 2 ω VDC, the pull-in voltage, beyond which the effective mechan- δpˆ˙ = − m δxˆ − γ δpˆ + GδXˆ − gδqˆ + ξ,ˆ (25) m ical frequency ωm becomes imaginary and the membrane is ω0 pulled onto the other electrode of the capacitor (see Appendix ˙ (0) ˆ δqˆ = ωLCδφ, (26) A). We also notice that for physically interesting parameter ω2 regimes, the shift x may be not negligible with respect to h ˆ˙ LC ˆ ˆ s 0 δφ = − δqˆ − γLCδφ − gδxˆ + δV, (27) and tends to −h /3 when approaching the pull-in voltage (see ω(0) 0 LC √ Appendix A). As a consequence, due to Eq. (17) and Eq. (31), δXˆ˙ = ∆δYˆ − κδXˆ + 2κXˆ , (28) the coupling g has a nonlinear dependence upon V , and it vac √ DC ˙ never surpasses a maximum value when VDC approaches its δYˆ = −∆δXˆ − κδYˆ + Gδxˆ + 2κYˆvac. (29) maximum value Vpull. This is explicitly shown in Fig.2, where We have introduced the two relevant coupling rates, the op- the electromechanical coupling g is shown versus the elec- tomechanical coupling rate trode distance h0 and VDC. The stationary membrane shift xs is determined by the equilibrium between the mechanical stress, ∂ω(xs) p the electrostatic force and the radiation force. As shown in G = −xzpf 2ncav, (30) ∂x Appendix B, where we provide the explicit expressions for the and the electromechanical coupling rate membrane-in-the-middle case based on the treatment of Ref. [44], the contribution of the radiation force on xs is negligible. ε0AeffVDC Finally we have also introduced rescaled noise operators: i) g = q . (31) 2 (0) the mechanical thermal noise term ξˆ(t) = Fˆ(t)/pzpf, with sym- C(xs)(h0 + xs) mLωLCω0 metrized autocorrelation function (in the high Qm Markovian We notice that both the mechanical and the LC resonance limit) (0) frequencies, ω0 and ωLC respectively, are modified when the cavity is driven and the DC voltage bias is applied, acquiring 1 D 0 0 E 0 ξˆ(t)ξˆ(t ) + ξˆ(t )ξˆ(t) = γm(2¯nm + 1)δ(t − t ); (34) new values: from Eq. (5) and Eq. (6) one has 2 " !#−1 ˆ 2 −1 ε0Aeff ii) the rescaled Nyquist noise operator on the rf circuit δV = ωLC = [LC(xs)] = L C0 + , (32) ˆ h0 + xs δV(t)/φzpf, with symmetrized autocorrelation function (in the 5 high QLC Markovian limit) In the linearized regime, the steady state of the tripar- tite optoelectromechanical system can be fully character- 1 0 0 0 ized because the noise terms are zero-mean quantum Gaus- hδVˆ (t)δVˆ (t )+δVˆ (t )δVˆ (t)i = γLC(2¯nLC +1)δ(t−t ); (35) 2 sian noises, and as a consequence, the steady state of the system is a zero-mean tripartite Gaussian state, fully iii) the two vacuum optical noises determined by its 6 × 6 correlation matrix (CM) V =   i j hu ∞ u ∞ u ∞ u ∞ i / 1 h √  iθ † −iθ ˆi( )ˆ j( ) + ˆ j( )ˆi( ) 2. Xˆvac = √ κin aˆine + aˆ e (36) 2κ in Starting from Eq. (39), this steady state CM can be de- √  iθ † −iθi termined in two equivalent ways, either as an integral (see + κex aˆexe + aˆexe , Eqs. (45)-(46) below), or as solution of a matrix equation (see −i h √  iθ † −iθ Eq. (47) below). Using the Fourier transformsu ˆi(ω) ofu ˆi(t), Yˆvac = √ κin aˆine − aˆ e (37) 2κ in one has √  iθ † −iθi ZZ 0 + κ aˆ e − aˆ e , dωdω − 0 1D 0 0 E ex ex ex V (t)= e it(ω+ω ) uˆ (ω)ˆu (ω )+uˆ (ω )ˆu (ω) . (41) i j (2π)2 2 i j j i which are uncorrelated and possess the same autocorrelation function Then, by Fourier transforming Eq. (39) and the correlation functions of the noises in the Markovian limit, Eqs. (34), (35) 1 D E and (38), one gets Xˆ (t)Xˆ (t0) + Xˆ (t0)Xˆ (t) (38) 2 vac vac vac vac D 0 0 E 1 D E 1 uˆi(ω)ˆu j(ω )+uˆ j(ω )ˆui(ω) 0 0 0 h 0 T i 0 = Yˆvac(t)Yˆvac(t ) + Yˆvac(t )Yˆvac(t) = δ(t − t ). = M(ω)DM(ω ) δ(ω+ω ), (42) 2 2 2 i j where we have defined the 6 × 6 matrix IV. DETERMINATION OF THE STEADY STATE M(ω) = (iω + A)−1 , (43)

The linearized Heisenberg-Langevin Equations in (24)-(29) and the diagonal diffusion matrix can be rewritten in the following compact matrix form  0 0 0 0 0 0  ˙   uˆ(t) = Auˆ(t) + nˆ(t), (39)  0 γ (2¯n + 1) 0 0 0 0   m m   0 0 0 0 0 0  whereu ˆ(t) = [δxˆ(t), δpˆ(t), δqˆ(t), δφˆ(t), δXˆ(t), δYˆ(t)]T is the col- D =   . (44)  0 0 0 γLC(2¯nLC + 1) 0 0  umn vector of fluctuations (the superscript T denotes trans-   √ √  0 0 0 0 κ 0  ˆ ˆ ˆ ˆ T   position),n ˆ(t) = [0, ξ(t), 0, δV(t), 2κXvac(t), 2κYvac(t)] is 0 0 0 0 0 κ the corresponding column vector of noises, and A is the matrix The δ(ω + ω0) factor is a consequence of the stationarity of   0 ω0 0 0 0 0 the noises, and inserting Eq. (42) into Eq. (41), one gets the  2   ωm   − −γm −g 0 G 0  following expression for the stationary correlation matrix  ω0   (0)   0 0 0 ω 0 0  ∞  LC  Z dω A =  ω2  . (40) ∞ †  LC  V = M(ω)DM(ω) , (45)  −g 0 − (0) −γLC 0 0   ω  −∞ 2π  LC   0 0 0 0 −κ ∆    G 0 0 0 −∆ −κ which can be equivalently rewritten as an integral in the time domain as The formal solution of Eq. (39) is Z ∞ V∞ = dtM(t)DM(t)T . (46) Z t 0 uˆ(t) = M(t)ˆu(0) + dsM(s)ˆn(t − s), 0 From the latter expression one can derive an alternative way to get the stationary CM V∞. In fact, when the stability condi- where M(t) = exp{At}. The system is stable and reaches its tions are satisfied [M(∞) = 0], one can verify, by an explicit steady state for t → ∞ when all the eigenvalues of A have integration, that Eq. (46) is equivalent to the following Lya- negative real parts so that M(∞) = 0. Here we will consider punov equation for the steady-state CM, the parameter region where the driven cavity mode cools the mechanical resonator, corresponding to a driving laser red- AV∞ + V∞AT = −D. (47) detuned with respect to the cavity, ∆ > 0 [40]. Within this parameter region, the stability condition is violated only at This is a linear equation for V∞ which can be analytically very large values of the optomechanical coupling G, which solved, but the general exact expression of the matrix elements are detrimental for cooling, correspond to the onset of optical is too cumbersome and will not be reported here. We have bistability [50], and which are of no interest here. adopted this latter method, and the numerical analysis and the 6 plots of Sec.VI are obtained from the numerical solution of exchange process at rate γLC with its own thermal reservoir Eq. (47). withn ¯ LC mean excitations is dominated by the exchange pro- In this paper we are interested only in the stationary state cess with the much colder “polariton” reservoir represented of the rf resonator and in its stationary energy in particular, by the mechanical resonator hybridized with the driven optical which is equal to cavity mode in the regime of efficient sideband cooling. This latter effective reservoir is characterised by an effective decay ~ω(0) h i ~ω(0)   rate γeff = γ + Γ , a nonzero mean number of excitations LC 2 ˆ2 LC ∞ ∞ m m m ULC = hδqˆ i + hδφ i = V33 + V44 (48) eff 2 2 n¯m [see Eq. (49)], and the LC resonator will scatter polaritons ! into the corresponding Stokes and anti-Stokes sidebands with (0) eff 1 ≡ ~ω n¯ + , rates that are respectively given by LC LC 2 2 eff g γm eff ALC . wheren ¯ is the effective mean occupation number of the LC ± = 2 (51) LC  eff 2 oscillator. γm + 4 (ωm ± ωLC) An intuitive explanation of the present expression is the fact that, when comparing with the optomechanical case of V. APPROXIMATE EXPRESSION FOR THE RF Eq. (50), the rate γeff/2 plays the role of the cavity decay rate RESONATOR OCCUPANCY m κ, and the electromechanical coupling g plays the role of the optomechanical coupling G. As a consequence one has an One can adapt standard resolved sideband cooling theory of effective polariton cooling rate optomechanical systems for the derivation of an approximate LC LC expression of the stationary occupancy of the rf resonator. In ΓLC = A− − A+ > 0. (52) the case of a standard optomechanical system, the stationary One can then apply the same arguments used for deriving occupancy of the mechanical resonator, far from the strong Eq. (49) to the present situation, and arrive at the following coupling regime, can be very well approximated as [51–53]: expression for the rf resonator occupancy eff LC eff γmn¯m + Γmn¯c + A+ eff γLCn¯ LC + ΓLCn¯m + A+ n¯m = , (49) n¯ LC = . (53) γm + Γm γLC + ΓLC This is the desired approximation we were looking for. It wheren ¯c ' 0 is the mean excitation of the optical reservoir at zero temperature, Γ = A − A > 0 is the net laser cooling works in the optimal regime for sideband cooling, that is, m − + ∼ ∼ eff ∼ 2 rate, with ∆ ωm ωLC > κ > G as well as γm /2 G /4κ > g. From Eq. (53) one can see that the rf resonator cannot be cooled G2κ/2 more than the mechanical resonator and that therefore at best A eff eff ± = 2 2 (50) one can achieven ¯ ∼ n¯ . The latter condition is achieved κ + (∆ ± ωm) LC m LC LC when ΓLC ∼ A−  A+ , γLC, which is obtained at resonance eff 2 2 2 the scattering rates into the Stokes (A+) and anti-Stokes (A−) ∆ ∼ ωm ∼ ωLC  γm ∼ G /2κ, when 2g κ/G  γLC. sidebands, corresponding respectively to the absorption or Defining the two relevant cooperativities, the optomechani- 2 emission of a mechanical vibrational quantum. Eq. (49) can cal cooperativity Com = G /2κγm and the electromechanical 2 be seen as the result of the balance between the two energy cooperativity Cem = g /γLCγm, the necessary condition to exchange processes involving the mechanical resonator: i) the eff eff achieve simultaneous ground state cooling,n ¯ LC ∼ n¯m < 1, one with rate γm with its thermal reservoir withn ¯m mean ex- can be written as citations; ii) the other one with rate Γm with the effective op- Cem  Com  1. (54) tical reservoir at zero temperature (¯nc ' 0) represented by the driven and decaying cavity, and which is responsible for This latter condition for the cooperatives can be satisfied only cooling. The scattering rate A+ is responsible for the quantum for an LC circuit with a large enough value of its quality factor, back-action limit associated with the quantum fluctuations of so that γLC  g, κ, because the electromechanical coupling g the radiation pressure force. cannot be too large with respect to G2/κ for the validity of In the optoelectromechanical system under study, the rf res- the above expressions. Nonetheless, the results of Sec.VI onator we are interested in is directly coupled to the mechan- based on the exact numerical solution of the Lyapunov equa- ical resonator, which is in turn coupled to the driven optical tion of Eq. (47) show that cooling of the rf resonator close to cavity. In the proposal of Ref. [15] one can laser cool the rf the quantum regime is possible also when the above assump- resonator by driving on the red sideband of the optical cav- tions are not fully satisfied and Eq. (53) is not too accurate. ity as in the usual optomechanical sideband cooling, and then exploiting the resonant electromechanical interaction in order to extend cooling to the rf circuit. This is why one can view VI. RESULTS FOR THE COOLING OF THE LC this process as sympathetic cooling [37] of the LC resonator RESONATOR by means of the laser cooling of the mechanical resonator. An equivalent description of the desired cooling process is Let us now determine the optimal parameter conditions un- the following: the rf resonator is cooled because the energy der which one can cool a MHz rf circuit down to its quantum 7 ground state. We show the main results in Fig.3 and Fig.4, where we apply the exact treatment of Eqs. (47)-(48). Then in Fig.5 we compare these results with the approximate treat- ment of Sec.V and the corresponding analytical prediction of Eq. (53), showing a satisfactory agreement between them. As we have seen above, the most relevant parameters one has to optimise are the optomechanical coupling G, the elec- tromechanical coupling g, the ambient temperatures T and TLC (which here will be taken to be identical for simplicity), and the quality factor of the rf resonant circuit, QLC. The other parameters will be kept fixed and corresponding to typical experimental values for a metalized membrane-in- the-middle configuration [20, 24], that is, laser optical wave- length λ = 1064 nm, membrane effective mass m = 0.7×10−10 kg, membrane intensity reflectivity (in the non-metalized sec- tion) R = 0.4, bare mechanical resonance frequency ω0 = 6 2π × 1 MHz, mechanical quality factor Qm = 10 , opti- FIG. 3: a) and c): Stationary LC circuit occupancyn ¯eff from the so- −3 LC cal cavity length Lc = 8 × 10 m, optical cavity finesse lution of the Lyapunov equation, Eq. (47), as a function of the scaled F = 5 × 104, yielding a total optical cavity amplitude decay electromechanical coupling g/κ, and of the scaled optomechanical eff rate κ = 2π × 374.74 kHz. We have also chosen κin = 0.4κ and coupling G/κ. b) and d):n ¯ LC versus g/κ, at fixed values of the op- a laser driving around the red mechanical sideband, that is, tomechanical coupling rate, corresponding to the horizontal lines in ∆ ' ω0. Finally, we have chosen an equivalent circuit induc- a)-c): G/κ = 0.2 (full yellow lines), G/κ = 0.5 (dashed red lines), tance L = 1 mH, and a membrane capacitor with an effective G/κ = 0.8 (dash-dotted blue lines), G/κ = 1.1 (dotted black lines). area A = 1.1 × 10−7 m2 and distance between the electrodes The upper plots a) and b) refer to the resonance of the bare mechan- eff ical and LC frequencies, ω = ω(0) , while the lower plots c) and d) equal to h = 2 µm. As a consequence, the two coupling rates 0 LC 0 refer to the resonance of the effective mechanical and LC frequen- G and g, and the corresponding system working point, can be cies, ωm = ωLC . We have chosen a quality factor of the rf resonator tuned by varying the two external parameters, the driving laser 4 QLC = 4 × 10 and a temperature T = 10 mK (see text for the other power P which fixes the intracavity photon number ncav, and system parameters). The dash-dotted black horizontal line in b) and the DC voltage bias V . eff DC d) refers ton ¯ LC = 1. In Fig.3 we assume a cryogenic temperature T = TLC = 10 4 mK, and a rf resonator quality factor QLC = 4 × 10 . Then in Fig.3a)-3b) we take identical bare frequencies of the LC that can be applied before the pull-in effect (see Fig.2); when (0) and mechanical resonators, ωLC = ω0, and display the be- we impose the resonance condition ωm = ωLC, since the me- eff haviour of the stationary rf circuit occupancyn ¯ LC predicted chanical frequency ωm decreases due to the applied DC bias eff by Eqs. (47)-(48). In Fig.3a) ¯nLC is plotted versus the scaled [see Eq. (33)], the overall capacitance of the LC circuit must electromechanical and optomechanical coupling rates, g/κ be increased, implying a further limitation to the value of g eff and G/κ respectively, while in Fig.3b) we plot ¯nLC versus [see Eq. (31)]. From Fig.3 we can conclude that ground state g/κ at four fixed values of G/κ corresponding to the horizon- cooling of the LC circuit is achievable for comparable values tal lines with the same color and style in Fig.3a). These plots of the coupling rates G and g, with the latter ones obtained clearly show that an experimentally achievable parameter re- with a voltage bias VDC always far enough from the pull-in gion exists where it is possible to reach the quantum regime voltage (see Fig.2). with an LC resonator occupation number below 1. The more demanding experimental condition assumed in However, the analysis of Sec.V predicts that optimal laser Fig.3 is the one on the rf resonator quality factor, because 2 3 cooling of the LC resonator occurs when the modified me- typical values are in the range 10 . QLC . 10 [20, 35], chanical and rf resonance frequencies of Eqs. (32)-(33) are even though very recently Ref. [36] demonstrated a rf res- (0) 4 resonant, ωm = ωLC, rather than their bare counterparts, ω onator with QLC ∼ 1.7×10 . For this reason in Fig.4 we have LC eff and ω0. Therefore, in Fig.3c)-3d) we display the same plots studied the LC circuit occupancyn ¯ LC also as a function of the of Fig.3a)-3b) but now under the optimal resonance condi- rf resonator quality factor QLC and of temperature T = TLC, tion ωm = ωLC, which can always be obtained by adjusting, at at a fixed value of the optomechanical coupling, G/κ = 0.8 (0) each working point, the value of ωLC by means of the tuning (blue dash-dotted line of Fig.3) and ωLC = ω0 for simplic- capacitance C0 [see Eq. (32)]. As expected, cooling of the LC ity. At each point of the plot, we have chosen the optimal eff resonator is improved, with a wider parameter region where value of the electromechanical coupling g minimisingn ¯ LC. As laser cooling is efficient, even though the qualitative behavior expected, cooling improves for lower temperature and larger is not appreciably modified. QLC, and the dash-dotted line sets the boundary of the “quan- eff By comparing the upper and lower plots in Fig.3, we no- tum region” wheren ¯ LC ≤ 1. tice that the coupling g spans a shorter interval of possible Finally, in Fig.5 we compare the exact numerical result values in the latter resonant case. In fact, for a given choice of of Eq. (47) with the approximate analytical theory developed eff h0 and Aeff, g is already upper limited by the maximum VDC in Sec.V. We show the mean rf photon number ¯nLC versus 8

eff FIG. 4: Stationary LC circuit occupancyn ¯ LC from the solution of the Lyapunov equation, Eq. (47), as a function of the rf resonator quality factor QLC and of temperature T (we have assumed here T = TLC ), for a chosen value of the optomechanical coupling, G/κ = 0.8 FIG. 5: LC resonator photon occupation numbern ¯eff versus g/κ, at (0) LC (dash-dotted blue line of Fig.3), choosing ω0 = ωLC , and by fixing, the same value of the optomechanical coupling, G/κ = 0.8, chosen (0) for any given pair of values of QLC and T, the optimal value of the in Fig.4, at temperature T = TLC = 300K, choosing ω0 = ωLC , and g neff 2 electromechanical coupling minimising ¯ LC . The dash-dotted line for two different values of the quality factor, QLC = 10 (red full and neff 3 refers to the upper bound of the “quantum region”, ¯ LC = 1. dashed upper curves), and QLC = 10 (blue full and dashed lower curves). Full lines refer to the exact numerical solution of Eq. (47), while dashed lines refer to the approximate treatment of Sec.V [see g/κ for the numerical solution (full lines) and for the approx- Eq. (53)]. imate analytical theory of Eq. (53) (dashed lines) at T = 300 K, choosing again ω(0) = ω , and for two different, realis- LC 0 mode in equations from Eq. (24) to Eq. (29), we get 2 tic values of the LC quality factor, QLC = 10 (red upper 3 √ κ − iω  curves), and QLC = 10 (blue lower curves). The approxi- −1 χ (ω)δXˆ(ω)=Gδxˆ(ω) + 2κ Xˆvac(ω)+Yˆvac(ω) (55), mate theory well reproduces the numerical solution for rela- c ∆ −1 ˆ ˆ tively low values of the electromechanical coupling g, up to χm (ω)δxˆ(ω)=GδX(ω) − gδqˆ(ω) + ξ(ω), (56) the value roughly corresponding to the minimum occupancy. χ−1 (ω)δqˆ(ω)=−gδxˆ(ω) + δVˆ (ω), (57) For larger g, the prediction of Eq. (53) increases more than LC the numerical solution, which is somehow expected because the approximated theory is valid as long as g is not larger than where χc(ω), χm(ω), and χLC(ω) are the natural susceptibil- 2 the effective optomechanical decay rate G /2κ. Nonetheless, ities of the cavity, mechanical, and electrical modes, respec- the approximate theory provides a very good estimate of the tively, given by achievable cooling limit as well as of the g-interval where the ∆ minimum rf-photon occupancy could be achieved. χ (ω) = , c ∆2 + (κ − iω)2 ω χ (ω) = 0 , m 2 2 (58) VII. DETECTION OF THE RF RESONATOR STEADY ωm − ω − iγmω STATE ω(0) χ (ω) = LC . LC 2 2 ωLC − ω − iγLCω The effective mean photon number of the rf circuit at the steady state can be measured following two ways: i) measur- The mutual interactions among the three modes lead to the ing directly the rf voltage signal between two points of the cir- modification of their natural susceptibilities. Inserting δXˆ(ω) cuit; ii) measuring the optical output of the cavity and trying and δqˆ(ω) in Eq. (56) into the equation for δxˆ(ω), we obtain to get information about the rf circuit state from it. In both h i−1 √ κ − iω  χeff(ω) δxˆ(ω) = χ (ω)G 2κ Xˆ (ω) +Yˆ (ω) cases these measurements are carried out in the frequency m c ∆ vac vac domain and therefore here we will focus on the solution of + ξˆ(ω) − χ (ω)gδVˆ (ω), (59) the Fourier transform of the Heisenberg-Langevin equations, LC eff Eq. (39). This solution has been already given in compact where χm (ω) is the effective mechanical susceptibility, de- form in Sec.IV, but it will be convenient to re-express it in fined by more physical terms using effective susceptibilities. h eff i−1 −1 2 Solving separately the two quadrature equations for each χm (ω) = χmc(ω) − g χLC(ω), (60) 9 with δVC(ω), or the voltage noise across the circuit inductance L, δVL(ω). Since δVC(ω) = [qzpf/C(xs)]δq(ω), δVL(ω) = −1 −1 2 2 χmc(ω) = χm (ω) − G χc(ω), (61) −(ω φzpf/ωLC)δq(ω), and both measurements will be affected by an imprecision noise originating from the environment and where χmc(ω) is the effective mechanical susceptibility in the the detection apparatus, one can write for the two cases presence of only the optomechanical interaction. Eq. (59) to- 2 gether with δpˆ(ω) = −i(ω/ω0)δxˆ(ω) provides the mechanical qzpf response of the system to external perturbations. S = S (ω) + S C , (71) δVC C x 2 δq imp Following the same approach, for the electrical mode we ( s) 4 2 obtain ω φzpf S S ω S L , δVL = 2 δq( ) + imp (72) h i−1 ω eff ˆ LC χLC(ω) δqˆ(ω) = δV(ω) − χmc(ω)g (62) 4 q2  √ κ − iω   ω zpf L ˆ ˆ ˆ = S δq(ω) + S imp, × χc(ω)G 2κ Xvac(ω) + Yvac(ω) + ξ(ω) , 4 C(x )2 ∆ ωLC s eff where S C and S L are the voltage imprecision noise spec- where χLC(ω) is the effective rf circuit susceptibility, given by imp imp tra for the measurement cases, we have assumed that they are eff −1 −1 2 [χLC(ω)] = χLC(ω) − g χmc(ω). (63) uncorrelated from the other noise terms, and we have used Eq. (6) and Eq. (21) in Eq. (72). These expressions show In the same way Eq. (62) together with δφˆ(ω) = that measured voltage noise spectra can provide the station- −i(ω/ωLC)δqˆ(ω) provides the rf response of the system to ex- ary photon occupancy of the rf resonant circuit provided that ternal perturbations. the spectra are properly calibrated and, above all, that the im- Eq. (48) can be rewritten as precision noise is small enough so not to alter the evaluation of the area below the measured spectrum [see Eqs. (64)-(66)]. hδqˆ2i + hδφˆ2i − 1 n¯eff = , (64) We can express this condition on the imprecision noise LC 2 C,L S imp in more quantitative terms by exploiting the fact that the that is, the effective stationary rf photon number can be ex- shape of the charge spectrum S δq(ω) is determined by the ef- pressed in terms of the dimensionless charge and flux vari- fective LC susceptibility of Eq. (63). The parameter regime ances. In turn, using Eq. (45), these variances are given by the which is optimal for cooling the LC resonator is distinct from integral over the corresponding noise spectra the strong electromechanical regime where the electrical and mechanical modes hybridize yielding two spectrally resolved Z +∞ Z +∞ 2 dω h †i dω peaks. On the contrary, in the cooling regime of interest here, hδqˆ i= M(ω)DM(ω) = Sδq(ω), (65) the effective LC susceptibility χeff is characterized by a single −∞ 2π 33 −∞ 2π LC Z +∞ Z +∞ 2 peak, significantly broadened by the interaction with the op- 2 dω h †i dω ω eff hδφ i= M(ω)DM(ω) = Sδq(ω).(66) tomechanical system, and therefore one can approximate χ 44 2 LC −∞ 2π −∞ 2π ωLC as a standard susceptibility with modified effective frequency ωeff and damping γeff [53, 54], Therefore laser cooling of the rf resonator can be experimen- LC LC tally verified by measuring the charge noise spectrum S (ω), h i2 δq ω(0) which can be explicitly written in terms of the effective sus- eff 2 LC χLC(ω) ' , (73)  2 2  2 ceptibilities defined above as eff 2 eff ωLC − ω + ω γLC 2  2    S (ω) = χeff (ω) g2 χ (ω) S + S + S , (67) δq LC mc rp ξ δV where r where Srp is the radiation pressure noise spectral contribution g2κ2 ωeff ' ω2 + ' ω , (74) LC LC G2 LC ∆2 + κ2 + ω2 S (ω) = G2κ , (68) rp (∆2 + κ2 − ω2)2 + 4κ2ω2 under typical experimental conditions, and eff and Sξ and SδV are, respectively, the Brownian force noise γLC ' γLC + ΓLC, (75) and the voltage noise spectra, which are constant, white, con- tributions due to the Markovian approximation made on the in agreement with the analysis of Sec.V. Therefore the charge ' eff ' Brownian and Johnson-Nyquist noise, noise spectrum S δq(ω) is peaked at ω ωLC ωLC, and, using Eqs. (67)-(70), one can write its maximum value with Sξ = γm(2¯nm + 1), (69) very good approximation as S = γ (2¯n + 1). (70) δV LC LC peak ' 1 S δq S δq(ωLC) = eff 2 {γLC(2¯nLC + 1) (76) (γLC ) The charge noise spectrum can be measured by measuring 2 2  2 2 2  g ω0 G (2ωLC +κ ) + 2 2 2 eff 2 γm(2¯nm + 1) + 2 2 , either the voltage noise across the circuit capacitance C(xs), (ωm−ωLC ) +(ωLC γm ) κ(4ωLC +κ ) 10 where we have approximated also the effective optomechani- cavity mode interacts directly only with the mechanical res- cal susceptibility in the Lorentzian-like form [53, 54] onator, and therefore it detects the dynamics of the rf circuit only indirectly, via its effects on the mechanical motion. As it ω2 is customary in cavity optomechanics [40], the resulting opti- |χ (ω)|2 ' 0 . (77) mc  2 cal output spectra allows a good measurement of the effective ω2 − ω2
2 + ω γeff m m mechanical occupancy, from which however it is hard to ex-

4 4 tract direct information about the steady state of the rf circuit. Due to the peaked structure of S δq(ω), one has ω ' ωLC in Eq. (72), and therefore the calibration factor for the two voltage noise measurements is practically the same, implying that the condition for a faithful, direct, spectral measurement VIII. CONCLUSIONS of the LC resonator photon occupancy reads We have investigated a tripartite optoelectromechanical q2 system formed by an optical cavity, a mechanical oscillator, C,L zpf peak S  S . (78) and a MHz rf resonator which, due to its low resonance fre- imp C(x )2 δq s quency, is normally in a thermally excited state even at ultra-

We also notice that, again due to the peaked form of S δq(ω), cryogenic temperatures. We have derived the optimal condi- one has hδφ2i ' hδq2i [see Eqs. (65)-(66)] and therefore tions for achieving ground state, sympathetic, cooling of the rf resonator, modeled as an LC circuit, by means of its interac- 1 tion with the mechanical resonator cooled by the laser-driven n¯eff ' hδq2i − . (79) LC 2 optical cavity. This requires a large optomechanical coopera- tivity, and an even larger electromechanical cooperativity. Un- If we consider experimentally achievable parameters, en- der these conditions, the LC resonator can be cooled close to abling to approach the quantum regime for the rf circuit, its quantum ground state, as confirmed by the exact numerical eff ' n¯ LC 1, one sees that the condition of Eq. (78) is nontriv- results in the linearized regime around the optimal working ial to satisfy, because its right hand side is of the order of point of the circuit. Manipulating rf resonant circuits at the 10−20 V2/Hz. In fact, in this regime the charge noise spectrum eff quantum level would be extremely useful for the quantum- peak is flattened and broadened because γLC becomes larger limited detection of weak rf signals, such as those employed and larger. Under these conditions the resonance peak height for positioning, timing, and for the sensitive detection of rf becomes comparable to the background noise level and a di- signals of astrophysical nature. rect measurement of the rf photon occupancy becomes hard. Nonetheless, this charge spectrum detection method is cer- tainly able to detect a significant laser cooling of the circuit. One can alternatively use the spectral analysis of the system Acknowledgments stationary state to get an indirect experimental detection of the rf circuit cooling process. In fact, one can always probe the We acknowledge the support of the European Union Hori- linear response of the system by driving the rf circuit with a zon 2020 Programme for Research and Innovation through tunable AC voltage VAC(ω), small enough in order not to mod- the Project No. 732894 (FET Proactive HOT), of the the ify its working point, but at the same time larger than Brow- Project QuaSeRT funded by the QuantERA ERA-NET Co- nian, Johnson-Nyquist and radiation pressure noises. From fund in Quantum Technologies, the support of the University Eq. (62) one has of Camerino UNICAM through the research project FAR2018 and of the INFN through the tHEEOM-RD project. P. Pier- eff δq(ω) ' χLC(ω)VAC(ω), (80) gentili acknowledges support from the European Union’s Horizon 2020 Programme for Research and Innovation under that is, one directly measures the effective susceptibility of the grant agreement No. 722923 (Marie Curie ETN - OMT). eff LC circuit, and in particular its FWHM γLC = γLC + ΓLC [see Eq. (73)]. However, such a measurement provides also an in- direct measurement of the rf photon occupancy in a large pa- Appendix A: Pull-in voltage rameter regime, i.e., when the Johnson-Nyquist spectral con- tribution dominates over the mechanical and radiation pres- sure ones in the charge noise spectrum of Eq. (67). In fact, in As discussed in the main text in Sec. III, soon after Eq. (33), this regime, one has simply [see also Eq. (53)] we cannot apply a too large value of the DC voltage bias VDC due to the pull-in effect of the electrode in front of the metal- γ ized membrane, softening the intrinsic spring constant of the n¯eff ' LC n¯ , (81) LC eff LC membrane mechanical mode. The quantity ω2 of Eq. (33) γLC m must be always positive, and using Eq. (17), one can rewrite that is, the temperature of the rf circuit is scaled down by the the stability condition of Eq. (33) as eff ratio γLC/γLC. 2 0 00 An alternative way to probe the system properties is to de- mω (h0 +3xs)+~ncav[2ω (xs)+ω (xs)(h0 + xs)] 0 > 0, (A1) tect the output of the optical cavity. However, any optical h0 +xs 11

0 00 where ω (xs) and ω (xs) denote respectively the first and sec- sure coupling between the optical mode and the mechanical ond order derivatives of the cavity frequencies with respect to resonator. In fact, the results shown in the main text can x. The denominator h0 + xs is always positive because it is just be applied to a generic geometry of the optoelectromechan- the distance between the two electrodes of the effective plane- ical setup. However, here we provide more details for the parallel capacitor modeling the membrane capacitor, so that membrane-in-the-middle case, based on the treatment of Ref. the stability condition is equivalent to impose the positivity of [44]. One can always express the frequency of a chosen cavity the above numerator. However, it is possible to verify that the mode in the presence of a semi-transparent membrane with in- static radiation pressure frequency shift proportional to ncav is tensity reflectivity R, placed at the static position z0 along the always negligible with respect to that of electrostatic origin cavity axis, as under typical experimental values, and therefore one gets the very simple stability condition c n √ o ω(x) = ω + Θ arcsin R cos [2k(z + x)] , (B1) h0 c 0 x > − . (A2) Lc s 3 Using Eq. (17) without the negligible radiation pressure term, where Lc is the cavity length, k = ωc/c is the wave vector the critical point xs = −h0/3 can be re-expressed as a condi- tion for the maximum applicable voltage, which is given by associated with the chosen cavity mode, and Θ is the over- lap parameter, 0 ≤ Θ ≤ 1, quantifying the transverse overlap s 2 3 between the chosen optical and membrane vibrational modes. 8mω0h0 Vpull = , (A3) The first order derivative determines the optomechanical 27ε0Aeff coupling according to Eq. (30), and it is given by which can be rewritten as a condition on the maximum elec- trical field within the membrane capacitor s s ∂ω 2ωc R ! 2 (x)=−Θ sin[2k(z + x)] . (B2) VDC 8mωm 0 2 = . (A4) ∂x Lc 1−R cos [2k(z0 + x)] h0 max 27Cm(0) The second order derivative instead enters into the expression for the renormalized mechanical frequency of Eq. (33) and it is given by

Appendix B: Explicit expressions in the case of a 2 2√ 2 ∂ ω 4ωc 1−2R+R cos [2k(z0+x)] membrane-in-the-middle setup (x)=−Θ R cos [2k(z0 + x)] 3/2 . (B3) 2 cLc 2 ∂x {1−R cos [2k(z0+x)]} We have not specified in the text the explicit form of the function ω(x), which is responsible for the radiation pres-

[1] A. D. O’Connell, M. Hofheinz, M. Ansmann, R. C Bialczak, M. [8] T. A. Palomaki, J. W. Harlow, J. D. Teufel, R. W. Simmonds, Lenander, E. Lucero, M. Neeley, D. Sank, H. Wang, M. Weides, and K. W. Lehnert, Nature (London) 495, 210 (2013). J. Wenner, J. M. Martinis, A. N. Cleland, Nature (London) 464, [9] C. F. Ockeloen-Korppi, E. Damskagg, J. M. Pirkkalainen, M. 697 (2010). Asjad, A. A. Clerk, F. Massel, M. J. Woolley, and M. A. Sil- [2] J. D. Teufel, T. Donner, D. Li, J. W. Harlow, M. S. Allman, K. lanpa¨a,¨ Nature (London) 556, 478 (2018). Cicak, A. J. Sirois, J. D. Whittaker, K. W. Lehnert, and R. W. [10] R. Riedinger, A. Wallucks, I. Marinkovic,` C. Loschnauer,¨ M. Simmonds, Nature (London) 475, 359 (2011). Aspelmeyer, S. Hong, and S. Groblacher,¨ Nature (London) 556, [3] J. Chan, T. P. Alegre, A. H. Safavi-Naeini, J. T. Hill, A. Krause, 473 (2018). S. Groblacher,¨ M. Aspelmeyer, and O. Painter, Nature (London) [11] S. Barzanjeh, E. S. Redchenko, M. Peruzzo, M. Wulf, D. P. 478, 89 (2011). Lewis, G. Arnold, and J. M. Fink, Nature (London) 570, 480 [4] M. Rossi, D. Mason, J. Chen, Y. Tsaturyan, and A. Schliesser, (2019). Nature (London) 563, 53 (2018). [12] J. Chen, M. Rossi, D. Mason, and A. Schliesser, Nat. Commun. [5] L. Qiu, I. Shomroni, P. Seidler, T. J. Kippenberg, Phys. Rev. 11, 943 (2020). Lett. 124, 173601 (2020). [13] A. A. Clerk, K. W. Lehnert, P. Bertet, J. R. Petta, and Y. Naka- [6] E. E. Wollman, C. U. Lei, A. J. Weinstein, J. Suh, A. Kronwald, mura, Nat. Phys. 16, 257 (2020). F. Marquardt, A. A. Clerk, and K. C. Schwab, Science 349, 952 [14] L. Midolo, A. Schliesser, and A. Fiore, Nature Nanotech. 13, (2015). 11 (2018). [7] J. M. Pirkkalainen, E. Damskagg, M. Brandt, F. Massel, and ¨ [15] J. M. Taylor, A. S. Sørensen, C. M. Marcus, and E. S. Polzik, M. A. Sillanpa¨a,¨ Phys. Rev. Lett. 115, 243601 (2015). Phys. Rev. Lett. 107, 273601 (2011). 12

[16] C. A. Regal, and K. W. Lehnert, Journal of Physics: Conference Adv.Space Res., 65, Issue 2, 856-867 (2020). Series 264, 012025 (2011). [34] M. Sarracanie, C. D. LaPierre, N. Salameh, D.E.J. Waddington, [17] S. Barzanjeh, D. Vitali, P. Tombesi, and G. J. Milburn, Phys. T. Witzel, M.S. Rosen, Sci. Rep. 5, 15177 (2015). Rev. A 84, 042342 (2011); S. Barzanjeh, M. Abdi, G. J. Mil- [35] M. F. Gely, M. Kounalakis, C. Dickel, J. Dalle, R Vatre,´ B. burn, P. Tombesi, and D. Vitali, Phys. Rev. Lett. 109, 130503 Baker, M. D. Jenkins, and G. A. Steele, Science 363, 1072 (2012). (2019). [18] J. Bochmann, A. Vainsencher, D. D. Awschalom, and A. N. [36] I. C. Rodrigues, D. Bothner, G. A. Steele, arXiv:2010.07975 Cleland, Nat. Phys. 9, 712 (2013). [quant-ph]. [19] R. W. Andrews, R. W. Peterson, T. P. Purdy, K. Cicak, R. W. [37] D. Larson, J. Bergquist, J. Bollinger, W. Itano, and D. Simmonds, C. A. Regal, and K. W. Lehnert, Nat. Phys. 10, 321 Wineland, Phys. Rev. Lett. 57, 70 (1986); C. J. Myatt, E. A. (2014). Burt, R. W. Ghrist, E. A. Cornell, and C. E. Wieman, Phys. [20] T. Bagci, A. Simonsen, S. Schmid, L. G. Villanueva, E. Rev. Lett. 78, 586 (1997). Zeuthen, J. Appel, J. M. Taylor, A. Sørensen, K. Usami, A. [38] J. Li and S. Groblacher,¨ New J. Phys. 22, 063041 (2020). Schliesser, and E. S. Polzik, Nature (London) 507, 81 (2014). [39] C. Genes, A. Mari, P. Tombesi, and D. Vitali, Phys. Rev. A 78, [21] K. C. Balram, M. I. Davanc¸o, J. D. Song, and K. Srinivasan, 032316 (2008). Nat. Photonics 10, 346–352 (2016). [40] M. Aspelmeyer, T. J. Kippenberg, and F. Marquardt, Rev. Mod. [22] A. Vainsencher, K. J. Satzinger, G. A. Peairs, and A. N. Cleland, Phys. 86, 1391 (2014). Appl. Phys. Lett. 109, 033107 (2016). [41] J. D. Thompson, B. M. Zwickl, A. M. Jayich, F. Marquardt, S. [23] K. Takeda, K. Nagasaka, A. Noguchi, R. Yamazaki, Y. Naka- M. Girvin, and J. G. E. Harris, Nature (London) 452, 72 (2008). mura, E. Iwase, J. M. Taylor, and K. Usami, Optica 5, 152 [42] A. M. Jayich, J. C. Sankey, B. M. Zwickl, C. Yang, J. D. (2017). Thompson, S. M. Girvin, A. A. Clerk, F. Marquardt, and J. G. [24] I. Moaddel Haghighi, N. Malossi, R. Natali, G. Di Giuseppe, E. Harris, New J. Phys. 10, 095008 (2008). and D. Vitali, Phys. Rev. Applied 9, 034031 (2018). [43] D. J. Wilson, C. A. Regal, S. B. Papp, and H. J. Kimble, Phys. [25] A. P. Higginbotham, P. S. Burns, M. D. Urmey, R. W. Peterson, Rev. Lett. 103, 207204 (2009). N. S. Kampel, B. M. Brubaker, G. Smith, K. W. Lehnert, and [44] C. Biancofiore, M. Karuza, M. Galassi, R. Natali, P. Tombesi, C. A. Regal, Nat. Phys. 14, 1038 (2018). G. Di Giuseppe, and D. Vitali, Phys. Rev. A 84, 033814 (2011). [26] A. Simonsen, S. A. Saarinen, J. D. Sanchez, J. H. Ardenkjær- [45] M. Karuza, C. Molinelli, M. Galassi, C. Biancofiore, R. Natali, Larsen, A. Schliesser, and E. S. Polzik, Opt. Express 27, 18561 P. Tombesi, G. Di Giuseppe, and D. Vitali, New J. Phys. 14, (2019). 095015 (2012). [27] M. Forsch, R. Stockill, A. Wallucks, I. Marinkovic,´ C. Gartner,¨ [46] P.-L. Yu, T. P. Purdy, and C. A. Regal, Phys. Rev. Lett. 108, R. A. Norte, F. van Otten, A. Fiore, K. Srinivasan, and S. 083603 (2012). Groblacher,¨ Nat. Phys. 16, 69–74 (2020). [47] C. W. Gardiner, and P. Zoller, Quantum Noise, Springer, Berlin [28] W. Jiang, C. J. Sarabalis, Y. D. Dahmani, R. N. Patel, F. M. (2004). Mayor, T. P. McKenna, R. Van Laer, and A. H. Safavi-Naeini, [48] V. Giovannetti and D. Vitali, Phys. Rev. A 63, 023812 (2001); Nat. Commun. 11, 1166 (2020). R. Benguria and M. Kac, Phys. Rev. Lett. 46, 1 (1981). [29] J. M. Fink, M. Kalaee, R. Norte, A. Pitanti, O. Painter, Quantum [49] H. Nyquist, Phys. Rev. 32, 110–113 (1928). Sci. Technol. 5, 034011 (2020). [50] A. Dorsel, J. D. McCullen, P. Meystre, E. Vignes, and H. [30] X. Han, W. Fu, C. Zhong, C.-L. Zou, Y. Xu, A. Al Sayem, M. Walther, Phys. Rev. Lett. 51, 1550 (1983). Xu, S. Wang, R. Cheng, L. Jiang, H. X. Tang, Nat. Commun. [51] I. Wilson-Rae, N. Nooshi, W. Zwerger, and T. J. Kippenberg, 11, 3237 (2020). Phys. Rev. Lett. 99, 093901 (2007). [31] G. Arnold, M. Wulf, S. Barzanjeh, E. S. Redchenko, A. Rueda, [52] F. Marquardt, J. P. Chen, A. A. Clerk, and S. M. Girvin, Phys. W. J. Hease, F. Hassani, J. M. Fink, Nat. Commun. 11, 4460 Rev. Lett. 99, 093902 (2007). (2020). [53] C. Genes, D. Vitali, P. Tombesi, S. Gigan, and M. Aspelmeyer, [32] Y. Chu, and S. Groblacher,¨ Appl. Phys. Lett. 117, 150503 Phys. Rev. A 77, 033804 (2008). (2020). [54] A. Dantan, C. Genes, D. Vitali, and M. Pinard. Phys. Rev. A 77, [33] M.J. Bentum, M.K. Verma, R.T. Rajan, A.J. Boonstra, C.J.M. 011804 (2008). Verhoeven, E.K.A. Gill, A.J. van der Veen, H. Falcke, M. Klein Wolt, B. Monna, S. Engelen, J. Rotteveel, L.I. Gurvits,