arxiv:1310.5276v2 [physics.optics] 11 aug 2014
TRANSCRIPT
![Page 1: arXiv:1310.5276v2 [physics.optics] 11 Aug 2014](https://reader034.vdocuments.us/reader034/viewer/2022042415/625f39ae19ede3109a14e6ee/html5/thumbnails/1.jpg)
Bidirectional and efficient conversion between microwave and optical light
R. W. Andrews,1, 2 R. W. Peterson,1, 2 T. P. Purdy,1, 2 K. Cicak,3
R. W. Simmonds,3 C. A. Regal,1, 2 and K. W. Lehnert1, 2, 3
1JILA, University of Colorado and NIST, Boulder, Colorado 80309, USA2Department of Physics, University of Colorado, Boulder, Colorado 80309, USA
3National Institute of Standards and Technology (NIST), Boulder, Colorado 80305, USA
Converting low-frequency electrical signals into much higher frequency optical signals has enabledmodern communications networks to leverage both the strengths of microfabricated electrical circuitsand optical fiber transmission, allowing information networks to grow in size and complexity. Amicrowave-to-optical converter in a quantum information network could provide similar gains bylinking quantum processors via low-loss optical fibers and enabling a large-scale quantum network.However, no current technology can convert low-frequency microwave signals into high-frequencyoptical signals while preserving their fragile quantum state. For this demanding application, aconverter must provide a near-unitary transformation between different frequencies; that is, theideal transformation is reversible, coherent, and lossless. Here we demonstrate a converter thatreversibly, coherently, and efficiently links the microwave and optical portions of the electromagneticspectrum. We use our converter to transfer classical signals between microwave and optical lightwith conversion efficiencies of âŒ10%, and achieve performance sufficient to transfer quantum statesif the device were further precooled from its current 4 kelvin operating temperature to below 40millikelvin. The converter uses a mechanically compliant membrane to interface optical light withsuperconducting microwave circuitry, and this unique combination of technologies may provide away to link distant nodes of a quantum information network.
Introduction
Modern communication networks manipulate informa-tion at several gigahertz with microprocessors and dis-tribute information at hundreds of terahertz via opti-cal fibers. A similar frequency dichotomy is develop-ing in quantum information processing. Superconduct-ing qubits operating at several gigahertz have recentlyemerged as promising high-fidelity and intrinsically scal-able quantum processors [1â3]. Conversely, optical fre-quencies provide access to low-loss transmission [4] andlong-lived quantum-compatible storage [5, 6]. Convert-ing information between gigahertz-frequency âmicrowavelightâ that can be deftly manipulated and terahertz-frequency âoptical lightâ that can be efficiently dis-tributed will enable small-scale quantum systems [7â9] tobe combined into larger, fully-functional quantum net-works [10, 11]. But no current technology can trans-form information between these vastly different frequen-cies while preserving the fragile quantum state of the in-formation. For this demanding application, a frequencyconverter must provide a near-unitary transformation be-tween microwave light and optical light; that is, the idealtransformation is reversible, coherent, and lossless.
Certain nonlinear materials provide a link between mi-crowave and optical light, and these are commonly usedin electro-optic modulators (EOMs) for just this pur-pose. While EOMs might be capable of reversible fre-quency conversion [12, 13], such conversion has not yetbeen demonstrated, and even optimized EOMs [14, 15]have predicted photon number efficiencies of only a few10â4 [12, 13, 15]. Other intermediate objects that inter-act with both microwave and optical light can be used tocreate the nonlinearity necessary for frequency conver-
sion. Proposed intermediaries include clouds of ultracoldatoms [16, 17], ensembles of spins [18, 19], and mechan-ical resonators [20â22]. All converters face the challengeof integrating optical light with the cryogenic tempera-tures needed for low-noise microwave signals and super-conducting circuitry. Here, we demonstrate a cryogenicconverter that incorporates a mechanical resonator withoptical light and superconducting circuitry, and use it toreversibly transform classical signals between microwaveand optical frequencies.
Early experiments used microwave light [23, 24] andoptical light [25] to manipulate mechanical resonatorsand study the interaction between light and a vibrat-ing mass [26, 27]. The fields of electromechanics andoptomechanics have since evolved at a remarkably com-mensurate pace. Both microwave and optical light havebeen separately used to cool a mechanical resonator toits quantum ground state of motion [28, 29]. This sameinteraction allows the mechanical resonator to serve asan information storage medium [30, 31], and opens upthe possibility of high-fidelity frequency conversion [32â36]. By combining the technologies of electromechanicsand optomechanics, we simultaneously couple a mechan-ical resonator to both a microwave circuit and an opticalcavity. This simultaneous coupling allows information inthe microwave domain to stream through the mechanicalresonator and emerge as optical light, and vice-versa. Weshow that the transformation is reversible and coherent,and further demonstrate a photon number efficiency of0.086±.007 and a transfer bandwidth of 30 kHz. Further-more, the performance is sufficient for nearly noiselessfrequency conversion if we can further precool the de-vice from its current 4 kelvin operating temperature todilution refrigerator temperatures below 40 millikelvin.
arX
iv:1
310.
5276
v2 [
phys
ics.
optic
s] 1
1 A
ug 2
014
![Page 2: arXiv:1310.5276v2 [physics.optics] 11 Aug 2014](https://reader034.vdocuments.us/reader034/viewer/2022042415/625f39ae19ede3109a14e6ee/html5/thumbnails/2.jpg)
2
Cavity electro-opto-mechanics
Our converter consists of two electromagnetic res-onators, one at an optical frequency and one at a mi-crowave frequency, that share a mechanical resonator.The mechanical resonator is formed by a thin silicon ni-tride membrane that is free to vibrate (Fig. 1). The opti-cal frequency resonator consists of a Fabry-Perot cavity,and as the membrane vibrates it moves along the opti-cal intensity standing wave and modulates the resonantfrequency of the optical cavity [37, 38]. The membraneis partially coated with a thin layer of niobium (whichsuperconducts at temperatures below 9 kelvin), and thiselectrically conductive portion is part of a capacitor inan inductor-capacitor circuit that forms the microwaveresonator (K.C., in prep.)[39, 40]. As the membrane vi-brates, it modulates the capacitance of the microwavecircuit, and thus its resonant frequency. Even thoughthe electromagnetic resonators are at vastly different fre-quencies (7 GHz and 282 THz), the coupling mechanismis equivalent: A nanometer of membrane motion shiftsthe microwave resonant frequency by approximately 4MHz, and shifts the optical resonant frequency by ap-proximately 40 MHz, giving coupling constants of Ge â 4MHz/nm and Go â 40 MHz/nm.
During the experiment, a strong pump tone is ap-plied below the resonant frequency of each electromag-netic resonator. The pumps enhance the electrome-chanical and optomechanical interaction, and the me-chanical resonator exchanges information with the mi-crowave and optical resonators at rates ge = Gexzp
âne
and go = Goxzpâno, respectively, where xzp is the zero-
point motion of the mechanical resonator and ne (no) isthe number of photons in the microwave (optical) res-onator induced by the microwave (optical) pump. Theexpressions for ge and go take on this simple form inthe resolved-sideband limit (defined as 4Ïm Îșe, Îșo,where Ïm is the frequency of the vibrational mode ofthe mechanical resonator and Îșe and Îșo are the energydecay rates of the microwave and optical resonators, re-spectively); however, these coupling rates can always beindependently adjusted in-situ by changing the strengthof the pumps and altering ne and no [31]. This coherentexchange of information between electromagnetic and vi-brational modes is capable of quantum state preservingfrequency conversion [41, 42].
A full description of the system includes the inputsand outputs of the microwave, optical, and mechanicalresonators. Of all the energy leaving the microwave (op-tical) resonator, only a fraction ηe (ηo) exits into thepropagating mode that we collect. Some energy is ab-sorbed in the resonators themselves, and the optical res-onator emits light into a particular spatial mode thatdoes not perfectly match the spatial mode of the inci-dent light field. The fraction of light we collect can beexpressed as ηe = Îșe,ext/Îșe and ηo = ΔÎșo,ext/Îșo, whereÎșe,ext (Îșo,ext) is the rate at which energy leaves the mi-crowave (optical) resonator into propagating fields, and
FIG. 1. Layout and operation of microwave-to-optical con-verter (a) A stoichiometric silicon nitride (Si3N4) membrane(light blue) that has been partially covered with niobium(dark blue) interacts with an inductor-capacitor (LC) circuitthat forms the microwave resonator and a Fabry-Perot cav-ity that forms the optical resonator (mode shown in red).Propagating light fields are coupled to the microwave res-onator with an inductive coupler, and to the optical resonatorwith a slightly transmissive input mirror. (b) The membraneis suspended within a silicon frame, and the microwave cir-cuitry is lithographically patterned on a separate silicon sub-strate. The two silicon chips are brought together so that theniobium-covered portion of the membrane comes to within500 nanometers of the microwave circuitry, thus forming theelectromechanical system. The electromechanical system isthen placed inside the optical resonator. The entire struc-ture is made to be cryogenically compatible. (c) A frequencydomain representation of the conversion process. A strong mi-crowave pump (blue arrow) is applied below the microwaveresonance (response shown as a black curve) with detuningâe. Likewise, a strong optical pump (red arrow) is appliedbelow the optical resonance (response shown as a black curve)with detuning âo. This allows a signal (green arrow) to beup- or downconverted in frequency.
Δ is the optical mode matching. If Îșe ge and Îșo go,the electromagnetic resonators couple energy and infor-mation in freely propagating microwave (optical) modesto a vibrational mode of the mechanical resonator at arate Îe (Îo), which has the simple form Îe = 4g2
e/Îșe
(Îo = 4g2o/Îșo) in the resolved-sideband limit [43].
During upconversion, an injected microwave field en-ters the converter at a frequency Ï above the microwavepump, enters and exits the mechanical resonator as de-termined by coupling rates Îe and Îo, and emerges as anoutgoing optical field at a frequency Ï above the opticalpump. A frequency-domain representation of the processis shown in Fig. 1c. Converter performance is character-ized by how efficiently the input microwave field, bin(Ï),is transformed into an output optical field, aout(Ï), andvice versa. The ratio aout(Ï)/bin(Ï) ⥠Soe(Ï) is one of
![Page 3: arXiv:1310.5276v2 [physics.optics] 11 Aug 2014](https://reader034.vdocuments.us/reader034/viewer/2022042415/625f39ae19ede3109a14e6ee/html5/thumbnails/3.jpg)
3
four scattering parameters that characterize the two-portnetwork formed by the converter (Fig. 2a). The fieldsbin(Ï) and aout(Ï) have units of (number · sec)1/2, andso the apparent photon number efficiency for upconver-sion is given by |aout(Ï)/bin(Ï)|2 = |Soe(Ï)|2.
During the experiment, the converter is integrated intoa larger network. To predict and measure converter per-formance, we need to specify which components are partof the converter, and which are part of the measurementnetwork. We choose to define the converter as all thecomponents between the inductive coupler of the mi-crowave resonator and the input mirror of the opticalresonator (Fig. 2). Converter performance then includesinternal losses in the microwave, optical, and mechanicalresonators and imperfect optical mode matching, but ex-cludes losses and gains in other components that are usedin our measurement. With this definition, the converteris a stand-alone component that can be readily integratedinto other networks. We can predict converter efficiencyusing a Hamiltonian that includes radiation pressure cou-pling [44] to generate Heisenberg-Langevin equations ofmotion (see S.I.). This analysis predicts
Soe(Ï) =
âÎeÎo
âı (Ï â Ïm) + (Îe + Îo + Îșm) /2ĂâAηeηo
(1)
where Îșm is the intrinsic mechanical damping and A =AeAo is conversion gain, with Ae = 1 + (Îșe/4Ïm)2 andAo = 1 + (Îșo/4Ïm)2. The reverse process, Seo, is math-ematically identical. The bandwidth of the conversionis set by the total mechanical damping Îe + Îo + Îșm.Conversion efficiency is greatest on mechanical resonance(Ï = Ïm) and when the coupling rates are matched(Îe = Îo) and exceed any intrinsic mechanical damp-ing Îșm; in this case the maximum conversion efficiencyis ηeηo.
Conversion gain A becomes appreciable when 4Ïm <Îșe, Îșo (and also depends on detunings âe and âo; seeS.I.). In this parameter regime, the converter begins toact like a linear, phase-insensitive amplifier, and as suchcan have an apparent efficiency greater than unity at thecost of adding noise [45]. While amplification might bebeneficial for some applications, ideal frequency conver-sion requires unit gain (Ae = Ao = 1) so that conversionadds as little noise as possible.
Reversible Frequency Conversion
By injecting a microwave or optical field and moni-toring the outgoing fields, we measure the S-parametersthat determine the number efficiency of the converter(Fig. 2b). We study conversion using three different vi-brational modes of the mechanical resonator, which areshown as membrane-displacement plots in Fig. 3a. Con-verter performance using the Ïm/2Ï = 560 kHz vibra-tional mode is shown in Fig. 3c. By sweeping the fre-quency of the injected microwave signal, we confirm that
7 GHz 1064 nm
PD
SSBmod
LR
I
SW1
SW2
Networkanalyzer in
out
SSBmod
Microwave Port
b
Converter
loss α
SeeSeo
SoeSoo
a Optical Port
bin(Ï) aout(Ï)gain ÎŽ
gain ÎČ loss Îłbout(Ï) ain(Ï)
FIG. 2. Measurement network. (a) Converter S-parametersare defined between the ports of the microwave and opti-cal resonators. To isolate the converter performance fromthe gains and losses of other components used in our mea-surement, we measure the forward and reverse transmission,αSoeÎŽ and ÎłSeoÎČ, and the off-resonant reflections, αÎČ and γΎ.This gives SoeSeo = (αSoeÎŽ)(ÎłSeoÎČ)/(αÎČ)(γΎ), a quantity in-sensitive to imperfect measurement of the separate gains andlosses α, ÎČ, Îł, ÎŽ. The off-resonant reflections are measured bytuning the injected signal away from the resonatorâs centralfrequency, where the converter acts as a near-perfect mirror.(b) A vector network analyzer generates a signal near themechanical resonant frequency, Ïm. This signal is mixed toa higher frequency using either the microwave pump or theoptical pump as a carrier via single-sideband (SSB) modula-tion. After interacting with the converter, the outgoing fieldsare separated using circulators, and detected with homodynedetection (for microwave frequencies) or direct detection witha photodiode (for optical frequencies). The detected signalis returned to the vector network analyzer, where it is refer-enced to the original signal to obtain a magnitude and phaseshift.
Soe has the expected Lorentzian lineshape of Eqn. 1, witha peak at Ï = Ïm. Conversion efficiency for the reverseprocess, Seo, is the same to within our measurement er-ror, giving initial indication that the conversion processis reversible. We also monitor the reflected microwavepower as See during upconversion. When conversion ef-ficiency peaks at Ï = Ïm, there is a corresponding dipin reflected microwave power, as seen in Fig. 3c, fur-ther demonstrating that some of the injected microwavesignal is indeed being converted. In addition to the mag-nitudes of the converted and reflected light, the measure-ments also yield phase information. We find that, up to afrequency-dependent phase shift, the phase is preservedduring conversion, indicating the conversion between thetwo frequency regimes is coherent.
![Page 4: arXiv:1310.5276v2 [physics.optics] 11 Aug 2014](https://reader034.vdocuments.us/reader034/viewer/2022042415/625f39ae19ede3109a14e6ee/html5/thumbnails/4.jpg)
4
FIG. 3. Reversibility and efficiency of conversion. (a) Contour plot of simulated membrane displacement, normalized to unity,for the three vibrational modes used for frequency conversion. Outlines of the optical mode (dashed circle) and niobiummetallization (dashed bowtie) are shown. (b) Apparent transfer efficiency (|SoeSeo|) and reflected microwave power (|See|2) asa function of damping ratio for the three vibrational modes. Measurement error is smaller than the plotted points. The blacklines are expected values given independently measured system parameters. Gray regions express the uncertainty of our systemparameters (see S.I.). (c) S-parameters for upconverted (Soe) and downconverted (Seo) signals, and reflected microwave signal(See) for the Ïm/2Ï = 560 kHz vibrational mode with Îo/2Ï = 2800 Hz and Îe/2Ï = 1300 Hz. Frequency is relative to themicrowave and optical pump frequencies. Electromagnetic resonators are centered at 7.1 GHz and 282 THz, with power decayrates Îșe/2Ï = 1.6 MHz and Îșo/2Ï = 1.65 MHz and efficiencies ηe = 0.76 and ηo = 0.11 (0.23 from internal resonator loss and0.47 from optical modematching). Pumps are red-detuned with âe â âÏm and âo/2Ï = â730 kHz (see S.I.).
While it is difficult to independently and accuratelycalibrate the quantities Soe and Seo, we can readily cal-ibrate their product, as diagrammed in Fig. 2a. Withthis method, we can determine SoeSeo to about 6% (lim-ited by a combination of standing waves in the microwavetransmission lines and sideband folding during direct de-tection; see S.I.). We find this calibration procedure ap-pealing because it makes equivalent use of both the for-ward and reverse transmission of the converter, demon-strating and emphasizing its reversible nature. The peak
value of the apparent conversion efficiency, plotted as|SoeSeo|, is shown in Fig. 3b along with the reflectedmicrowave power, |See|2, at the frequency of peak con-version efficiency.
The converter performance for all three vibrationalmodes is correctly predicted by linear optomechanicaltheory with independently measured system parameters(see S.I.). By changing the microwave pump power whileholding the optical pump power fixed, we vary the ratioÎe/Îo and observe that maximum conversion efficiency
![Page 5: arXiv:1310.5276v2 [physics.optics] 11 Aug 2014](https://reader034.vdocuments.us/reader034/viewer/2022042415/625f39ae19ede3109a14e6ee/html5/thumbnails/5.jpg)
5
occurs near Îe = Îo, as seen in Fig. 3b, confirming theimpedance matching condition predicted by Eqn. 1. Fur-thermore, we can explore the effect of conversion gain byusing different vibrational modes of the mechanical res-onator for frequency conversion. For the Ïm/2Ï = 380kHz vibrational mode, Ïm/Îșe â Ïm/Îșo â 0.25 and aphoton number gain of A = 3 is expected. This resultsin an apparent efficiency of 0.25, as seen in the peakvalue of |SoeSeo| for this vibrational mode. For conver-sion with the higher frequency Ïm/2Ï = 1.24 MHz vi-brational mode, Ïm/Îșe â Ïm/Îșo â 0.8 and the converterapproaches the resolved-sideband limit. This reduces thephoton number gain to A = 1.4. With this lower gaincomes a lower apparent efficiency, but the absorption ofmicrowave power occurs nearer the peak in conversionefficiency, indicating more ideal conversion (see S.I.).
The converter reaches nearly unit internal efficiencyas Îe,Îo Îșm. For the Ïm/2Ï = 560 kHz vibrationalmode, the converter reaches a matched (Îe = Îo) band-width of 30 kHz, overwhelming the intrinsic mechanicalloss by nearly four orders of magnitude; i.e. only abouttwo parts in 104 of the converted signal are lost throughthe mechanical decay channel. For this vibrational mode,we expect the maximum apparent conversion efficiencyto be |SoeSeo| = Aηeηo = 0.169 ± 0.012, which corre-sponds well with the measured maximum of 0.178±0.011.The efficiency ηeηo does not include conversion gain A,and is limited by absorption and scattering of light inthe electromagnetic resonators (Îșe,ext/Îșe = 0.76 andÎșo,ext/Îșo = 0.23), and imperfect optical mode-matching(Δ = 0.47). These factors combine to give an expectedefficiency ηeηo = 0.082 ± 0.006. This corresponds wellwith the measured maximum apparent efficiency dividedby the expected gain of 0.086± 0.007.
Vibrational noise
Any extraneous vibrations of the mechanical resonatorappear as noise on the converted signal, and can bespecified with a signal to noise ratio (SNR), defined asSNR = Sin/nvib, where Sin is the spectral density of theinput signal, and nvib is the spectral density of vibra-tional noise added during conversion as referenced to theinput of the converter [43]. In units of quanta of addednoise (a quantum corresponding to one photon per sec-ond in a one hertz bandwidth), upconversion adds
nvib =1
Aeηe
(Îșmnenv
Îe+ (Ae â 1) + (Ao â 1)
)(2)
for matched coupling rates, where nenv âĄ1/(e~Ïm/kBTenv â 1), kB is Boltzmannâs constant,and Tenv is the temperature of the mechanical res-onatorâs environment. Any conversion gain addsvibrational noise given by the last two terms in Eqn. 2.Mechanical decoherence Îșmnenv also adds vibrationalnoise, but the effect of mechanical decoherence canbe reduced by increasing Îe and Îo. Only with unit
SNR
(Lin
ear U
nits
)
Đe/2 (Hz)
Added vibrational noise
(photons/secâHz)
0 200 400 6001
15
30
45
60
75
90
1.1Ă105
7400
3700
2500
1800
1500
1200
560 561 kHz
-125
-105
PSD
(dBm
/Hz)
-115Signal
Vibrational noise
FIG. 4. Optically-detected signal to noise ratio (SNR). SNRfor an upconverted microwave signal as a function of couplingrate Îe; injected signal power is kept constant and Îo â Îe. Afit to optomechanical theory is shown (line), and the extractedadded vibrational noise (nvib) is shown on the right verticalaxis. Inset: power spectral density (PSD) of an upconvertedmicrowave signal.
gain (ηe Ae â 1 and ηe Ao â 1) and negligiblemechanical decoherence (ηeÎe Îșmnenv) do vibrationsof the mechanical resonator âfreeze outâ. In additionto low vibrational noise, noiseless frequency conversionrequires unit conversion efficiency (ηe = ηo = 1).
To characterize noise added by vibrations of the me-chanical resonator, we inject a fixed-frequency signal intothe microwave port of the converter and measure thespectral density of the output optical field. We mea-sure vibrational noise added during upconversion (as op-posed to downconversion) because near quantum-limitedphotodetection allows us to quickly characterize noise atoptical frequencies. The microwave signal is injected at afrequency Ïm above the microwave pump with matchedcoupling rates (Îe â Îo). As shown in the inset of Fig.4, the converted signal power appears on a noise back-ground that is dominated by thermal fluctuations filteredby the mechanical resonatorâs susceptibility. As we in-crease the coupling rates Îe and Îo (while keeping theinjected microwave signal power constant), we observean increased SNR (Fig. 4). Fitting the observed SNR tolinear optomechanical theory for the spectral density ofthe injected microwave signal Sin allows us to estimatenvib, the added vibrational noise.
Vibrational noise is non-zero during 4 kelvin operationbecause the mechanical resonator is still âhotâ and drivenby thermal fluctuations. While this effect can be sub-stantially reduced by increasing Îe and Îo, colder tem-peratures are needed to reduce added vibrational noise tothe single quantum level. Thermalizing the vibrationalmodes of the mechanical resonator to a dilution refrig-erator temperature of 40 millikelvin with the achievedÎe = Îo = 2Ï Ă 15 kHz would ideally result in nvib â 1.Spurious mechanical modes in the silicon substrate andmirrors that are another potential source of noise are alsoexpected to be suppressed at colder temperatures [46].
![Page 6: arXiv:1310.5276v2 [physics.optics] 11 Aug 2014](https://reader034.vdocuments.us/reader034/viewer/2022042415/625f39ae19ede3109a14e6ee/html5/thumbnails/6.jpg)
6
Furthermore, temperatures of Tenv < (2Ï Â·7 GHz)~/kB â300 millikelvin are required to eliminate Johnson noiseat microwave frequencies that corrupts weak (quantum)signals.
Discussion
We have demonstrated a converter that reversibly, co-herently, and efficiently connects the microwave and op-tical portions of the electromagnetic spectrum. Threefeatures we have observedâthe reversibility of the con-version, its phase-preserving nature, and the absorptionof injected signal power during conversionâprovide firmevidence that state transfer is occurring, and that wehave accessed the beam-splitter Hamiltonian [42] that isfundamentally capable of noiseless state transfer. Theability to coherently exchange information between mi-crowave and optical light opens new possibilities forquantum information, in particular entanglement be-tween microwave and optical photons [35, 47â49].
While the converter functions well, some modestchanges can significantly improve performance. Remov-ing internal loss in the electromagnetic resonators willimprove ηe and ηo, which currently limit conversion effi-ciency. Lower temperatures will improve the microwaveresonator, which at 4 kelvin suffers from thermally ex-cited quasiparticles that limit ηe. Constructing the op-
tical resonator with asymmetric mirror coatings will im-prove ηo by reducing the amount of light lost throughthe second mirror. In addition, more careful placementof the electromechanical system inside the optical res-onator can decrease scattered light and improve opticalmode matching [38]. Added vibrational noise can be re-duced by using mechanical resonators with higher qualityfactors (Q ⥠Ïm/Îșm). In particular, the type of siliconnitride membrane we use is capable of extremely highquality factors of Q > 107 [39, 46, 50], about two ordersof magnitude greater than the quality factors used in thiswork. With our modular converter design, improvementsto any single converter component can be carried out in-dependently. This inherent flexibility will allow us toexplore the materials and methods necessary to fully in-tegrate optical light with superconducting qubits at mil-likelvin temperatures.
Acknowledgments
This work was supported by the DARPA QuASARprogram and the National Science Foundation undergrant number 1125844. We would like to thank D.R.Schmidt for sharing his knowledge of fabrication tech-niques, J.N. Ullom for lending us equipment, and P.-L.Yu, J.D. Teufel, and J. Kerckhoff for discussions. C.A.R.thanks the Clare Boothe Luce foundation for support.
[1] John Clarke and Frank K. Wilhelm, âSuperconductingquantum bits,â Nature 453, 1031â1042 (2008).
[2] R. J. Schoelkopf and S. M. Girvin, âWiring up quantumsystems,â Nature 451, 664â669 (2008).
[3] M. H. Devoret and R. J. Schoelkopf, âSuperconductingcircuits for quantum information: An outlook,â Science339, 1169â1174 (2013).
[4] Jeremy L. OâBrien, Akira Furusawa, and Jelena Vuck-ovic, âPhotonic quantum technologies,â Nature Photon-ics 3, 687â695 (2009).
[5] Iulia Buluta, Sahel Ashhab, and Franco Nori, âNaturaland artificial atoms for quantum computation,â Reportson Progress in Physics 74, 104401 (2011).
[6] C. Langer, R. Ozeri, J. D. Jost, J. Chiaverini, B. De-Marco, A. Ben-Kish, R. B. Blakestad, J. Britton, D. B.Hume, W. M. Itano, D. Leibfried, R. Reichle, T. Rosen-band, T. Schaetz, P. O. Schmidt, and D. J. Wineland,âLong-lived qubit memory using atomic ions,â Phys. Rev.Lett. 95, 060502 (2005).
[7] Stephan Ritter, Christian Nolleke, Carolin Hahn, An-dreas Reiserer, Andreas Neuzner, Manuel Uphoff, Mar-tin Mcke, Eden Figueroa, Joerg Bochmann, and Ger-hard Rempe, âAn elementary quantum network of singleatoms in optical cavities,â Nature 484, 195â200 (2012).
[8] M. D. Reed, L. DiCarlo, S. E. Nigg, L. Sun, L. Frun-zio, S. M. Girvin, and R. J. Schoelkopf, âRealization ofthree-qubit quantum error correction with superconduct-ing circuits,â Nature 482, 382â385 (2012).
[9] Erik Lucero, R. Barends, Y. Chen, J. Kelly,M. Mariantoni, A. Megrant, P. OâMalley, D. Sank,A. Vainsencher, J. Wenner, T. White, Y. Yin, A. N. Cle-land, and John M. Martinis, âComputing prime factorswith a Josephson phase qubit quantum processor,â Na-ture Physics 8, 719â723 (2012).
[10] H. J. Kimble, âThe quantum internet,â Nature 453,1023â1030 (2008).
[11] T. D. Ladd, F. Jelezko, R. Laflamme, Y. Nakamura,C. Monroe, and J. L. OBrien, âQuantum computers,âNature 464, 45â53 (2010).
[12] Mankei Tsang, âCavity quantum electro-optics,â Phys.Rev. A 81, 063837 (2010).
[13] Mankei Tsang, âCavity quantum electro-optics. II. input-output relations between traveling optical and microwavefields,â Phys. Rev. A 84, 043845 (2011).
[14] D.A. Cohen, M. Hossein-Zadeh, and A. F. J. Levi,âMicrophotonic modulator for microwave receiver,â Elec-tronics Letters 37, 300â301 (2001).
[15] Vladimir S. Ilchenko, Anatoliy A. Savchenkov, An-drey B. Matsko, and Lute Maleki, âWhispering-gallery-mode electro-optic modulator and photonic microwavereceiver,â J. Opt. Soc. Am. B 20, 333â342 (2003).
[16] M. Hafezi, Z. Kim, S. L. Rolston, L. A. Orozco, B. L. Lev,and J. M. Taylor, âAtomic interface between microwaveand optical photons,â Phys. Rev. A 85, 020302 (2012).
[17] J. Verdu, H. Zoubi, Ch. Koller, J. Majer, H. Ritsch, andJ. Schmiedmayer, âStrong magnetic coupling of an ultra-cold gas to a superconducting waveguide cavity,â Phys.
![Page 7: arXiv:1310.5276v2 [physics.optics] 11 Aug 2014](https://reader034.vdocuments.us/reader034/viewer/2022042415/625f39ae19ede3109a14e6ee/html5/thumbnails/7.jpg)
7
Rev. Lett. 103, 043603 (2009).[18] Atac Imamoglu, âCavity QED based on collective mag-
netic dipole coupling: Spin ensembles as hybrid two-levelsystems,â Phys. Rev. Lett. 102, 083602 (2009).
[19] D. Marcos, M. Wubs, J. M. Taylor, R. Aguado, M. D.Lukin, and A. S. SĂžrensen, âCoupling nitrogen-vacancycenters in diamond to superconducting flux qubits,âPhys. Rev. Lett. 105, 210501 (2010).
[20] Amir H Safavi-Naeini and Oskar Painter, âProposal foran optomechanical traveling wave phononphoton trans-lator,â New Journal of Physics 13, 013017 (2011).
[21] C. A. Regal and K. W. Lehnert, âFrom cavity electrome-chanics to cavity optomechanics,â Journal of Physics:Conference Series 264, 012025 (2011).
[22] Joerg Bochmann, Amit Vainsencher, David D.Awschalom, and Andrew N. Cleland, âNanome-chanical coupling between microwave and opticalphotons,â Nature Physics , â (2013).
[23] Vladimir Braginsky, A. B. Manukin, and M. Y.Tikhonov, âInvestigation of dissipative ponderomotice ef-fects of electromagnetic radiation,â Journal of Experi-mental and Theoretical Physics 31, 829 (1970).
[24] A. Gozzini, F. Maccarrone, F. Mango, I. Longo, andS. Barbarino, âLight-pressure bistability at microwavefrequencies,â J. Opt. Soc. Am. B 2, 1841â1845 (1985).
[25] A. Dorsel, J. D. McCullen, P. Meystre, E. Vignes, andH. Walther, âOptical bistability and mirror confinementinduced by radiation pressure,â Phys. Rev. Lett. 51,1550â1553 (1983).
[26] Vladimir Braginsky and A. B. Manukin, âPonderomotiveeffects of electromagnetic radiation,â Journal of Experi-mental and Theoretical Physics 25, 653 (1967).
[27] Carlton M. Caves, âQuantum-mechanical radiation-pressure fluctuations in an interferometer,â Phys. Rev.Lett. 45, 75â79 (1980).
[28] J. D. Teufel, T. Donner, Dale Li, J. W. Harlow, M. S.Allman, K. Cicak, A. J. Sirois, J. D. Whittaker, K. W.Lehnert, and R. W. Simmonds, âSideband cooling ofmicromechanical motion to the quantum ground state,âNature 475, 359â363 (2011).
[29] Jasper Chan, T. P. Mayer Alegre, Amir H. Safavi-Naeini, Jeff T. Hill, Alex Krause, Simon Groblacher,Markus Aspelmeyer, and Oskar Painter, âLaser coolingof a nanomechanical oscillator into its quantum groundstate,â Nature 478, 89â92 (2011).
[30] E. Verhagen, S. Deleglise, S. Weis, A. Schliesser, andT. J. Kippenberg, âQuantum-coherent coupling of a me-chanical oscillator to an optical cavity mode,â Nature482, 63â67 (2012).
[31] T. A. Palomaki, J. W. Harlow, J. D. Teufel, R. W. Sim-monds, and K. W. Lehnert, âCoherent state transferbetween itinerant microwave fields and a mechanical os-cillator,â Nature 495, 210â214 (2013).
[32] L. Tian and Hailin Wang, âOptical wavelength conver-sion of quantum states with optomechanics,â Phys. Rev.A 82, 053806 (2010).
[33] Ying-Dan Wang and Aashish A. Clerk, âUsing interfer-ence for high fidelity quantum state transfer in optome-chanics,â Phys. Rev. Lett. 108, 153603 (2012).
[34] Lin Tian, âAdiabatic state conversion and pulse trans-mission in optomechanical systems,â Phys. Rev. Lett.108, 153604 (2012).
[35] Sh. Barzanjeh, M. Abdi, G. J. Milburn, P. Tombesi, andD. Vitali, âReversible optical-to-microwave quantum in-
terface,â Phys. Rev. Lett. 109, 130503 (2012).[36] S. A. McGee, D. Meiser, C. A. Regal, K. W. Lehnert,
and M. J. Holland, âMechanical resonators for storageand transfer of electrical and optical quantum states,âPhys. Rev. A 87, 053818 (2013).
[37] J. D. Thompson, B. M. Zwickl, A. M. Jayich, FlorianMarquardt, S. M. Girvin, and J. G. E. Harris, âStrongdispersive coupling of a high-finesse cavity to a microme-chanical membrane,â Nature 452, 72â75 (2008).
[38] T. P. Purdy, R. W. Peterson, and C. A. Regal, âObser-vation of radiation pressure shot noise on a macroscopicobject,â Science 339, 801â804 (2013).
[39] P.-L. Yu, T. P. Purdy, and C. A. Regal, âControl ofmaterial damping in high-Qmembrane microresonators,âPhys. Rev. Lett. 108, 083603 (2012).
[40] T. Bagci, A. Simonsen, S. Schmid, L. G. Villanueva,E. Zeuthen, J. Appel, J. M. Taylor, A. SĂžrensen, K. Us-ami, A. Schliesser, and E. S. Polzik, âOptical detectionof radio waves through a nanomechanical transducer,âArXiv e-prints (2013), arXiv:1307.3467 [physics.optics].
[41] U. Akram, N. N Kiesel, M. Aspelmeyer, and G. J. Mil-burn, âSingle-photon opto-mechanics in the strong cou-pling regime,â New Journal of Physics 12, 083030 (2010).
[42] Jing Zhang, Kunchi Peng, and Samuel L. Braunstein,âQuantum-state transfer from light to macroscopic oscil-lators,â Phys. Rev. A 68, 013808 (2003).
[43] Jeff T. Hill, Amir H. Safavi-Naeini, Jasper Chan, andOskar Painter, âCoherent optical wavelength conversionvia cavity optomechanics,â Nature Communications 3,1196â (2012).
[44] C. K. Law, âInteraction between a moving mirror andradiation pressure: A Hamiltonian formulation,â Phys.Rev. A 51, 2537â2541 (1995).
[45] Carlton M. Caves, âQuantum limits on noise in linearamplifiers,â Phys. Rev. D 26, 1817â1839 (1982).
[46] T. P. Purdy, R. W. Peterson, P.-L. Yu, and C. A. Regal,âCavity optomechanics with Si3N4 membranes at cryo-genic temperatures,â New Journal of Physics 14, 115021(2012).
[47] Ying-Dan Wang and Aashish A. Clerk, âReservoir-engineered entanglement in optomechanical systems,âPhys. Rev. Lett. 110, 253601 (2013).
[48] Lin Tian, âRobust photon entanglement via quantum in-terference in optomechanical interfaces,â Phys. Rev. Lett.110, 233602 (2013).
[49] M. C. Kuzyk, S. J. van Enk, and H. Wang,âGenerating robust optical entanglement in weak cou-pling optomechanical systems,â ArXiv e-prints (2013),arXiv:1307.2844 [quant-ph].
[50] B. M. Zwickl, W. E. Shanks, A. M. Jayich, C. Yang,A. C. Bleszynski Jayich, J. D. Thompson, and J. G. E.Harris, âHigh quality mechanical and optical proper-ties of commercial silicon nitride membranes,â AppliedPhysics Letters 92, 103125 (2008).
[51] D. F. Walls and G. J. Milburn, Quantum Optics, 2nd ed.(Springer-Verlag, Berlin, 2008).
[52] M. Aspelmeyer, T. J. Kippenberg, and F. Mar-quardt, âCavity Optomechanics,â ArXiv e-prints (2013),arXiv:1303.0733 [cond-mat.mes-hall].
[53] Joseph Kerckhoff, Reed W. Andrews, H. S. Ku,William F. Kindel, Katarina Cicak, Raymond W. Sim-monds, and K. W. Lehnert, âTunable coupling to a me-chanical oscillator circuit using a coherent feedback net-work,â Phys. Rev. X 3, 021013 (2013).
![Page 8: arXiv:1310.5276v2 [physics.optics] 11 Aug 2014](https://reader034.vdocuments.us/reader034/viewer/2022042415/625f39ae19ede3109a14e6ee/html5/thumbnails/8.jpg)
8
[54] A. A. Clerk, M. H. Devoret, S. M. Girvin, Florian Mar-quardt, and R. J. Schoelkopf, âIntroduction to quantumnoise, measurement, and amplification,â Rev. Mod. Phys.82, 1155â1208 (2010).
[55] A. Schliesser, Riviere R., G. Anetsberger, O. Arcizet,and T. J. Kippenberg, âResolved-sideband cooling of amicromechanical oscillator,â Nature Physics 4, 415â419(2008).
[56] Christina J. Hood, H. J. Kimble, and Jun Ye, âCharac-terization of high-finesse mirrors: Loss, phase shifts, andmode structure in an optical cavity,â Phys. Rev. A 64,033804 (2001).
[57] S. Weis, Riviere R., S. Deleglise, E. Gavartin, G., O. Ar-cizet, A. Schliesser, and T. J. Kippenberg, âOptome-chanically induced transpency,â Science 330, 1520â1523(2010).
![Page 9: arXiv:1310.5276v2 [physics.optics] 11 Aug 2014](https://reader034.vdocuments.us/reader034/viewer/2022042415/625f39ae19ede3109a14e6ee/html5/thumbnails/9.jpg)
9
SUPPLEMENTARY INFORMATION
1. Electro-opto-mechanical theory
Our converter is described by a Hamiltonian that includes radiation pressure coupling between electromagnetic andmechanical modes [44] :
H = ~Ïoaâ a+ ~Ïeb
â b+ ~Ïmcâ c+ ~Go(aâ a)x+ ~Ge(bâ b)x (S1)
Here, a, b, and c denote annihilation operators for the optical, electrical and mechanical modes, respectively. Fre-quencies Ïi describe the mode resonant frequencies, with subscripts i = o, e,m denoting optical, electrical, andmechanical. The couplings from the optical and microwave resonators to the mechanical mode are denoted Go and
Ge. Finally, x is the mechanical mode position operator x = xzp(c+ câ ) with xzp =â
~2mÏm
.
Using this Hamiltonian, we calculate the Heisenberg-Langevin equations of motion [51] and linearize them about astrong, coherent pump field [52]. We model the optical resonator as having three loss ports, with total linewidth (full-width half-maximum)1 Îșo = Îșo,L+Îșo,R+Îșo,int, comprising the left mirror (the port that is used during measurement ofthe converter), right mirror, and internal loss. The microwave resonator has two ports with linewidth Îșe = Îșe,R+Îșo,int,and the mechanical resonator has a single port with linewidth Îșm. The resulting system of first-order, linear differentialequations can be concisely described as a state space model (see, e.g., [53]).
a(t) = Aa(t) +Bain(t) (S2)
aout(t) = Ca(t) +Dain(t) (S3)
where a = (a, b, c, aâ , bâ , câ ) is a vector of the resonator modes that represents the internal state of the converter, andain (aout) is a vector of input (output) fields. See section S 5 for the form of the matrices A, B, C, and D, and thevectors ain and aout. The state-space model can be reduced to a transfer function2, Î(Ï), that relates input fields ain
to output fields aout.
aout(Ï) = Î(Ï)ain(Ï) (S4)
Î(Ï) = C(âiÏI âA)â1B +D (S5)
This formalism reflects our measurement, where we can apply fields (i.e. modify ain) and measure fields (i.e. measurecomponents of aout). Also, many software packages exist to efficiently solve state space and transfer function models.The S-parameter Soe(Ï) corresponds to the transfer function element Îaout,L,bin,R(Ï) = aout,L(Ï)/bin,R(Ï).3 In theweak-coupling (Îșe, Îșo Îe,Îo) limit (which our experiment satisfies),
Soe(Ï) = Îaout,L,bin,R(Ï) = â
âÎșe,R
Îșe
Îșo,L
ÎșoΔ
âÎeÎo
âi(Ï â ÏâČm) + (Îe + Îo + Îșm)/2
âA exp(iÏ) (S6)
Here, ÏâČm = Ïm + Ωe + Ωo, where Ωe and Ωo are the electro- and optomechanical frequency shifts. Îe and Îo are theelectro- and optomechanical damping rates (see section S 5). The gain term A and the associated phase shift Ï aregiven by
A = Ae ĂAo =
[(âe â Ïm)2 + (Îșe/2)2
â4âeÏm
]Ă[
(âo â Ïm)2 + (Îșo/2)2
â4âoÏm
](S7)
Ï = arctan
(Îșe/2
âe + Ïm
)+ arctan
(Îșo/2
âo + Ïm
)(S8)
For âe = âo = âÏm, A simplifies to
[1 +
(Îșe
4Ïm
)2] [
1 +(Îșo
4Ïm
)2], giving it a simple interpretation in terms of the
resolved-sideband parameter. We also define the efficiency of the electrical and optical resonators in the main text asηe =
Îșe,R
Îșe, ηo = Δ
Îșo,L
Îșo, where Δ is the optical mode matching. In the main text, we omit a constant phase shift.
1 In the main text, we use use Îșo,L â Îșo,ext and Îșe,R â Îșe,ext. 2 We use f(Ï) =â«âââ f(t)eâiÏtdt, which is equivalent to the bi-
lateral Laplace transform f(s) =â«âââ f(t)eâstdt with sâ iÏ.
3 We use a to denote the operator expectation value ăaă.
![Page 10: arXiv:1310.5276v2 [physics.optics] 11 Aug 2014](https://reader034.vdocuments.us/reader034/viewer/2022042415/625f39ae19ede3109a14e6ee/html5/thumbnails/10.jpg)
10
Reflection measurements are described generically by the following transfer function element:
Îaout,L,ain,L(Ï) = â[1â Îșo,L
âi(Ï + âo) + Îșo/2
]âi(Ï â Ïm â Ωe â ΩoB) + (Îe + ÎoB + Îșm)/2
âi(Ï â Ïm â Ωe â Ωo) + (Îe + Îo + Îșm)/2(S9)
where B = (1â iÎșo,L/2âo
)/(1â Îșo,L
âi(Ï+âo)+Îșo/2) is a resonant enhancement factor of order unity. Equivalent expressions
for other electromagnetic output fields can be obtained via substitution of appropriate parameters.Another path through the systemârelevant for signal-to-noise measurementsâis from the mechanical bath input
to an optical output:
Îaout,L,cin(Ï) = âiâΔÎșo,L
Îșo
âÎoÎșm
âi(Ï â ÏâČm) + (Îe + Îo + Îșm)/2
âAo exp
[i arctan
(Îșo/2
âo + Ï
)](S10)
Vibrational noise added during conversion can be quantified with the spectral density of the output fields4, S(Ï),
2ÏS(Ï)ÎŽ(Ï â ÏâČ) =âš
[aout(ÏâČ)]â aTout(Ï)
â©= Î?(ÏâČ)
âš[ain(ÏâČ)]
â aTin(Ï)
â©ÎT (Ï) (S11)
More concisely, S(Ï) = Î?(Ï)ÎŁÎT (Ï), where ÎŁ ⥠(2ÏÎŽ(Ï â ÏâČ))â1âš
[ain(ÏâČ)]â aTin(Ï)
â©. Components of the spectral
density matrix have units of number/sec·Hz. The covariance matrix Σ contains information (see section S 5) about the
bath temperature and quantum noise (e.g.âš
[cin(ÏâČ)]â cin(Ï)
â©= 2ÏÎŽ(Ï â ÏâČ)nenv and
âšcin(Ï) [cin(ÏâČ)]
â â©
= 2ÏÎŽ(Ï âÏâČ)(nenv + 1), where nenv is the mechanical thermal bath phonon occupation [54]).
Using this spectral density, we can write a signal to noise ratio (SNR) for upconversion:
SNR = Sin(Ï)
(|Îaout,L,bin,R(Ï)|2 2ÏÎŽ(Ï â ÏâČ)
ăaâ out,Laout,Lă
)(S12)
where Sin is the spectral density of the input microwave field and the second term is noise at the output of the
converter referred to the input. If we assume all fields in ain except cin and câ in are that of vacuum, and additionallyuse âe = âo = âÏm and Îe = Îo, the expression for SNR simplifies to:
SNR = AeÎșe,R
Îșe
[Îșmnenv
Îe+
(Îșe
4Ïm
)2
+
(Îșo
4Ïm
)2]â1
Ă Sin(Ï) (S13)
= Aeηe
[Îșmnenv
Îe+ (Ae â 1) + (Ao â 1)
]â1
Ă Sin(Ï) (S14)
In this case the added vibrational noise, nvib, is found through SNR = Sin/nvib. This expression accurately describesthe measured SNR in Fig. 4 of the main text, even though input microwave fields are not vacuum but a thermalstate, because the mechanical bath is the dominant source of noise (nvib nth,e; see section S 5).
2. Measurement Network
Fig. S1 provides component-level details of the network used to measure performance of the converter.
a. Measurement of converter
For an S-parameter measurement such as Soe(Ï) = aout,L(Ï)/bin,R(Ï), we generate an upper sideband on the
microwave pump (with amplitude and phase given by bin,R(Ï); bâ in,R(Ï) describes the lower sideband) and measure
the upper sideband on the outgoing optical pump (which yields the amplitude and phase of aout,L(Ï)). In general,the conversion process also produces a non-zero lower sideband; that is, transfer function element that connects
4 We use the notational convention [aout(Ï)]â = aâ out(âÏ), as in[54].
![Page 11: arXiv:1310.5276v2 [physics.optics] 11 Aug 2014](https://reader034.vdocuments.us/reader034/viewer/2022042415/625f39ae19ede3109a14e6ee/html5/thumbnails/11.jpg)
11
-90°0°
L RQ
I
87 MHz
-90°0°
L RQ
I
SW1
7 GHz
SpectrumAnalyzer
PBS
PBS λ/2
PD
50:50 BS
to lock
λ/4
EOM-86 MHzAOM
-87 MHzAOM
200 MHzAOM
Nd:YAG1064 nm
Optical SSBMicrowave SSB
Network Analyzerin Q
out in I
PDDetection
SW2
SW3
RF (1 MHz)Microwave (7 GHz)Optical (282 THz)
4K77K
Cryostat
λ/2
-20 dB
PD
PBS
Converter
Îșe,int
Îșe,R
Îșo,LÎșo,R
Îșo,int
ÎșmFaradayRotator
-20 dB
LR
QI
FIG. S1. Measurement network. A network analyzer (green shade) generates a swept signal near the mechanical resonantfrequency of interest. This signal is mixed to a higher frequency using the microwave or optical pump as a reference viasingle-sideband (SSB) modulation. The microwave SSB modulation (blue) is made directly on the pump; the optical SSBmodulation (pink) is first made on an intermediate radio-frequency signal that drives an acousto-optic modulator (AOM). Anystray lower sideband modulation is suppressed at least 20 dB (10 dB) below the upper sideband modulation for the microwave(optical) carrier. The converter (purple) either reflects or transmits/converts the signal originally produced by the networkanalyzer. Reflected and transmitted signal fields are demodulated and detected (yellow) using homodyne (direct) detection formicrowave (optical) fields. SNR data is collected by modulating the microwave pump with a signal at frequency Ï = ÏâČm andanalyzing the converted optical signal on a spectrum analyzer. As drawn, the network measures Soe. Throwing switches SW1and SW2 measures Seo. Throwing SW3 allows signal to noise measurements of the Soe path. Input/output coupling rates ofthe converter are noted.
Ï Ï
âe âo
7 GHz 282 THz
Uppersideband
Lowersideband
FIG. S2. Frequency-domain representation of the conversion process. A microwave pump (blue arrow) is applied below themicrowave resonance (response shown as black curve) with detuning âe. Likewise, an optical pump (red arrow) is appliedbelow the optical resonance (response shown as a black curve) with detuning âo. This allows a signal (green arrow) to beconverted. The conversion also produces a lower sideband when the optical resonator response does not completely suppressit. Suppression occurs when 4Ïm Îșo.
![Page 12: arXiv:1310.5276v2 [physics.optics] 11 Aug 2014](https://reader034.vdocuments.us/reader034/viewer/2022042415/625f39ae19ede3109a14e6ee/html5/thumbnails/12.jpg)
12
bin,R(Ï) to aâ out,L(Ï) is nonzero (see Fig. S2). In the resolved-sideband limit, the electromagnetic resonators greatly
suppress the lower sideband. In our experiment, with a resolved sideband parameter of Ïm/Îșe â Ïm/Îșo â 0.5,we must account for the lower sideband. During downconversion (Seo) and microwave reflection (See), we collectboth in-phase and quadrature data in homodyne detection. No information is lost during measurement, and we canreconstruct the amount of power present in the upper- and lower- sidebands of the reflected microwave pump. Duringupconversion (Soe), the reflected optical pump and sidebands are focused onto a photodetector, and we only detectamplitude modulation of the pump. In order to assign a measured amplitude modulation an equivalent upper-sidebandamplitude, we need to know the relationship between the upper- and lower- sidebands.
The photocurrent we measure is proportional to the incident photon flux (in units of number/sec). The photonflux is
f(t) âĄâšdâ (t)d(t)
â©(S15)
where d(t) is the annihilation operator for the light field incident on the detector and has units ofâ
number/sec. Duringan upconversion measurement, the mode-matched portion of the optical pump, given by
âΔαpe
âiÏpt, is injected intothe converter. The injected pump induces an intracavity field
âΔαpe
âiÏptâÎșo,L/(âiâo + Îșo/2). Mechanical motionmodulates this intracavity field and generates sidebands at a frequency Ïm from the pump frequency. A mechanicaldisplacement x0sin(Ïmt+ Ί) results in a total intracavity field of [55]
a(t) =âΔαpe
âiÏpt
âÎșo,L
âiâo + Îșo/2
(1 +
Gox0
2
(eâi(Ïmt+Ί)
âi(âo + Ïm) + Îșo/2â ei(Ïmt+Ί)
âi(âo â Ïm) + Îșo/2
)). (S16)
This results in an outgoing optical field that we detect of
d(t) =âΔαpe
âiÏpt
(Îșo,L
âiâo + Îșo/2â 1
)+âΔαpe
âiÏptÎșo,L
âiâo + Îșo/2
Gox0
2
(eâi(Ïmt+Ί)
âi(âo + Ïm) + Îșo/2â ei(Ïmt+Ί)
âi(âo â Ïm) + Îșo/2
). (S17)
This expression can be interpreted as consisting of the reflected pump field at frequency Ïp along with an uppersideband at frequency Ïp + Ïm and a lower sideband at frequency Ïp â Ïm. The lower sideband is negligible inthe resolved sideband regime (Ïm > Îșo) and when âo â âÏm. The non-modematched light in the pump fielddoes not beat with the sidebands on the photodetector, and appears as a zero-frequency photocurrent. DefiningΞ = Arg[Îșo,L/(âiâo + Îșo/2) â 1], Ï = Arg[Îșo,L/(âiâo + Îșo/2)], and neglecting the zero-frequency (DC) portion ofthe photocurrent,
f(t) = Geâi(Ïmt+Ί)
(1â e2i(ΞâÏ)
(âi(âo + Ïm) + Îșo/2
i(âo â Ïm) + Îșo/2
))+ c.c. (S18)
G = Δ |αp|2âŁâŁâŁâŁ Îșo,L
âiâo + Îșo/2â 1
âŁâŁâŁâŁ âŁâŁâŁâŁ Îșo,L
âiâo + Îșo/2
âŁâŁâŁâŁ Gox0
2
(eâi(ΞâÏ)
âi(âo + Ïm) + Îșo/2
)In a measurement where we do not detect the lower sideband at frequency ÏpâÏm, we would detect a photocurrent
proportional to fssb(t).
fssb(t)
f(t)=
1
1â e2i(ΞâÏ)(âi(âo+Ïm)+Îșo/2i(âoâÏm)+Îșo/2
) (S19)
This ratio allows us to convert between the measured photon flux f(t) and an equivalent upper-sideband photon fluxfssb(t).
As a note, the pump is mildly attenuated by the optical cavity (the factor Îșo,L/(âiâo + Îșo/2) â 1 in Eqn. S17).This changes the effective gain seen by the sidebands during photodetection. During calibration (discussed in themain text), we account for this effect.
![Page 13: arXiv:1310.5276v2 [physics.optics] 11 Aug 2014](https://reader034.vdocuments.us/reader034/viewer/2022042415/625f39ae19ede3109a14e6ee/html5/thumbnails/13.jpg)
13
TABLE S1. Converter parameters. Sets of values apply to the three mechanical modes of interest.
Îșe/2Ï (kHz) Îșe,R/2Ï (kHz) Îșe,int/2Ï (kHz) âe/2Ï (kHz) Gexzp/2Ï (Hz)
1590±30 1220±30 370±20 -520, -650, -1350±50 2.1±0.1, 2.7±0.1, 0.76±0.04Îșo/2Ï (kHz) Îșo,L/2Ï (kHz) Îșo,int/2Ï (kHz) âo/2Ï (kHz) Goxzp/2Ï (Hz)
1650±50 380±70 850±120 -730±40 7±1 , 23±4, 15±3Îșo,R/2Ï (kHz) Ïm/2Ï (kHz) Îșm/2Ï (Hz)
420±80 380, 560, 1240 0.42±0.05, 4.0±0.5, 4.2±0.5
b. Optical network
The optical network utilizes two laser beams: A pump tone (Fig. S1, red line) and a locking tone (Fig. S1, greenline), both of which are derived from a single 1064 nm Nd:YAG laser (Coherent, Inc. Mephisto). Both beams aredouble-passed through a 200 MHz AOM that is used to lock the laser frequency to the optical resonator. The lockingtone provides feedback to the 200 MHz AOM via the Pound-Drever-Hall (PDH) stabilization technique. The lockingtone operates at low powers and near optical cavity resonance to minimize unwanted optomechanical effects. Thepump tone is detuned 1 MHz red from the locking tone (accomplished by passing each through different âŒ80 MHzAOMs with a relative detuning of 1 MHz) and occupies a polarization mode orthogonal to the locking beam, resultingin a detuning of âo/2Ï = â730 kHz. The birefringent splitting of the cavity is 270 kHz. The AOM used to detune thepump from the locking tone is also used to write the SSB signal used in S-parameter measurements onto the pump. AFaraday rotator, polarizing beamsplitters (PBSs), and a half-wave plate inject the tones into orthogonal polarizationmodes of the optical cavity, and separate the reflected light for detection with photodetectors. Transmitted light isalso detected and used to measure optomechanical parameters.
c. Optical calibration
A summary of optomechanical parameters is in Table S1. In general, calibration is performed as described inprevious work [46]. The optical coupling Goxzp is extracted from a measurement of mechanical damping, using thethermal motion of the membrane at temperature Tenv as a known displacement standard. This differs from theelectromechanical calibration of Gexzp in that this method does not require separate knowledge of the intracavitypower.
We directly measure the output coupling rate Îșo,L from peak transmission and reflection measurements of thecavity in both directions [56]. This measurement has five unknown quantities: the mirror transmissions TL and TR,the internal loss L, and the mode matching factors ΔL, and ΔR. (Here, ΔL is the mode matching Δ used in the rest ofthe text.) Peak power measurements of the cavity in the forward direction (left port input) are Pin,f, Pr,f, and Pt,f,indicating input, reflection, and transmission. We can write the cavity transmission and reflection as
Pt,f
ΔLPin,f= 4TLTR
(1
TL + TR + L
)(S20)
ΔLPr,f â (1â ΔL)Pin,f
ΔLPin,f=
(L+ TR â TL
L+ TR + TL
)2
(S21)
We can eliminate the modematching ΔL by dividing Eqn. S20 by Eqn. S21:
Pt,f
Pr,f â Pin,f=
4TLTR
(L+ TR â TL)2 â (L+ TR + TL)2(S22)
This expression now contains only directly-measurable optical powers. This and an equivalent expression for probingthe cavity from the opposite side, combined with a measurement of the cavity finesse F = 2Ï
TL+TR+L , specify TL, TR,
and L uniquely. We can convert to the loss rate of interest via Îșo,L = ÎșoTL
TL+TR+L , likewise for Îșo,R, Îșo,int. Once theseare determined, we can solve for the modematching ΔL and ΔR.
![Page 14: arXiv:1310.5276v2 [physics.optics] 11 Aug 2014](https://reader034.vdocuments.us/reader034/viewer/2022042415/625f39ae19ede3109a14e6ee/html5/thumbnails/14.jpg)
14
FIG. S3. (a) Contour plot of membrane displacement for three different vibrational modes. (b) Mechanical damping for thesevibrational modes as a function of microwave pump power. The electromechanical coupling rates, Gexzp, are found by fittingthe data to linear optomechanical theory.
d. Microwave network
The electrical network utilizes a microwave signal generator (Agilent Technologies) operating at âŒ7 GHz. Thissource provides the microwave pump tone, and provides the local oscillator for two quadrature-IF mixers (MarkiMicrowave) that are used to modulate and demodulate the pump tone with the signal used in S-parameter mea-surements. The microwave cavity is intrinsically stable (environmental fluctuations change the microwave resonantfrequency by much less than its linewidth), and no stabilization technique is needed.
e. Microwave calibration
Parameters of the microwave resonator, detailed in Table S1, are found by sweeping a signal over a broad rangeof frequencies that includes the central frequency of the microwave resonator. Fitting the reflected signal allowsus to extract the external coupling rate (Îșe,R) and the total cavity linewidth (Îșe). The electromechanical couplingrate is measured by monitoring the total mechanical damping as a function of the strength of the microwave pump.Mechanical damping is inferred from the mechanically dressed response of the microwave resonator, sometimes referredto as optomechanically induced transparency [57]. Mechanical damping rate as a function of microwave pump powerfor the three different mechanical modes used in this work is shown in Fig. S3.
![Page 15: arXiv:1310.5276v2 [physics.optics] 11 Aug 2014](https://reader034.vdocuments.us/reader034/viewer/2022042415/625f39ae19ede3109a14e6ee/html5/thumbnails/15.jpg)
15
Soe Seo
Mag
nitu
dePh
ase
450 kHz 500 550 600 650
0.1
10
450 kHz 500 550 600 650
0.1
10
Mag
nitu
de
Soe
a
b
250 kHz 350 450 550 650 750 8500.01
0.1
1
0.01
0.1
1
FIG. S4. (a) S-parameters for upconversion (Soe) and downconversion (Seo) for the Ïm/2Ï = 560 kHz vibrational mode with 32kHz of transfer bandwidth. A Lorentzian fit (blue) is included. (b) Upconversion over a wide range of frequencies; conversionis visible for three vibrational modes of the mechanical resonator at Ïm/2Ï = 380 kHz, 560 kHz, and 760 kHz.
3. Additional data
The converter reaches a matched (Îe = Îo) bandwidth of 32 kHz, as shown in Fig. S4a. With large bandwidths,measurable conversion occurs across a wide range of frequencies. In particular, the deviation from a Lorentzianfrequency dependence during conversion seen in Fig. S4a can be partially explained by the presence of nearbymechanical modes shown in Fig. S4b. The optomechanical theory we use is only appropriate when the vibrationalmodes of the mechanical resonator are spectrally resolved.
4. Compatibility of superconducting circuitry and optical frequency light
The large circulating optical powers (âŒ100 mW) necessary for converter operation could adversely affect cryogenicperformance in two ways. First, the silicon nitride membrane that forms the mechanical resonator can absorb a smallfraction of the optical photons that pass through it, resulting in localized heating that possibly increases mechanicaldecoherence. Based on previous work, we expect that the mechanical resonator does not absorb enough opticalpower to significantly increase the âŒ4 kelvin bath temperature seen by the mechanical resonator [38]; furthermore,membrane heating would affect the niobium metallization on the membrane, resulting in additional observed loss inthe microwave circuitry (particularly evident if local temperatures exceeded 9 kelvin, the superconducting transitiontemperature of niobium). Second, stray light absorbed by the niobium can directly break Cooper pairs and createadditional quasiparticles and loss with no need for bulk heating. We have observed this behavior. As shown in Fig.S5, circulating optical powers of 10 mW greatly increased internal loss in the microwave resonator for early testsof this type of microwave-to-optical converter. For the device used in this work, additional shielding was added toprevent stray light in the optical resonator from reaching the niobium microwave circuitry. With this shielding, themaximum optomechanical coupling (Îo) can reach a similar value to the maximum electromechanical coupling (Îe)for the Ïm/2Ï = 560 kHz and Ïm/2Ï = 380 kHz vibrational modes.
5. Appendix
In the expressions presented here, ge = Gexzp and go = Goxzp are electro- and optomechanical couplings in Hz.The electrical pump produces a classical intracavity field described by ÎČ, where |ÎČ| = âne and Arg[ÎČ] = Ξp. Likewise,
![Page 16: arXiv:1310.5276v2 [physics.optics] 11 Aug 2014](https://reader034.vdocuments.us/reader034/viewer/2022042415/625f39ae19ede3109a14e6ee/html5/thumbnails/16.jpg)
16
0 20 40 60 80 1000.2
1
2
Power through Si3N4 membrane (mW)
Mic
row
ave
reso
nato
r in
tern
al lo
ss ra
te (M
Hz)
Initial
Intermediate
Current device
FIG. S5. Internal microwave resonator loss rate, Îșe,int, as a function of optical power. Initial converters only functioned withlow (<10 mW) optical powers.
the optical pump produces a classical intracavity field described by α, where |α| =âno and Arg[α] = Ïp. We also
use nth,m = nenv to describe the mechanical bath phonon occupation, and use nth,e (nth,o) to describe the averagephoton occupation of the electrical (optical) input fields at temperature.
Ωe = g2eÎČ
2
[âe + Ï
(âe + Ï)2 + (Îșe/2)2+
âe â Ï(âe â Ï)2 + (Îșe/2)2
](S23)
Ωo = g2oα
2
[âo + Ï
(âo + Ï)2 + (Îșo/2)2+
âo â Ï(âo â Ï)2 + (Îșo/2)2
](S24)
Îe = g2eÎČ
2
[Îșe
(âe + Ï)2 + (Îșe/2)2â Îșe
(âe â Ï)2 + (Îșe/2)2
](S25)
Îo = g2oα
2
[Îșo
(âo + Ï)2 + (Îșo/2)2â Îșo
(âo â Ï)2 + (Îșo/2)2
](S26)
A =
iâo â Îșo
2 0 âieiÏpgoα 0 0 âieiÏpgoα
0 iâe â Îșe
2 âieiΞpgeÎČ 0 0 âieiΞpgeÎČ
âieâiÏpgoα âieâiΞpgeÎČ âÎșm
2 â iÏm âieiÏpgoα âieiΞpgeÎČ 0
0 0 ieâiÏpgoα âiâo â Îșo
2 0 ieâiÏpgoα
0 0 ieâiΞpgeÎČ 0 âiâe â Îșe
2 ieâiΞpgeÎČ
ieâiÏpgoα ieâiΞpgeÎČ 0 ieiÏpgoα ieiΞpgeÎČ iÏm â Îșm
2
(S27)
B =
(M 0
0 M
)(S28)
M =
âÎșo,L
âÎșo,R
âÎșo,int 0 0 0
0 0 0âÎșe,R
âÎșe,int 0
0 0 0 0 0âÎșm
(S29)
![Page 17: arXiv:1310.5276v2 [physics.optics] 11 Aug 2014](https://reader034.vdocuments.us/reader034/viewer/2022042415/625f39ae19ede3109a14e6ee/html5/thumbnails/17.jpg)
17
C =
âÎșo,L 0 0 0 0 0âÎșo,R 0 0 0 0 0
0âÎșe,R 0 0 0 0
0 0 0âÎșo,L 0 0
0 0 0âÎșo,R 0 0
0 0 0 0âÎșe,R 0
(S30)
D =
â1 0 0 0 0 0 0 0 0 0 0 0
0 â1 0 0 0 0 0 0 0 0 0 0
0 0 0 â1 0 0 0 0 0 0 0 0
0 0 0 0 0 0 â1 0 0 0 0 0
0 0 0 0 0 0 0 â1 0 0 0 0
0 0 0 0 0 0 0 0 0 â1 0 0
(S31)
ÎŁ =
(N 0
0 N + I
)(S32)
N =
nth,o 0 0 0 0 0
0 nth,o 0 0 0 0
0 0 nth,o 0 0 0
0 0 0 nth,e 0 0
0 0 0 0 nth,e 0
0 0 0 0 0 nth,m
(S33)
ain =(ain,L ain,R ain,int bin,R bin,int cin aâ in,L aâ in,R aâ in,int bâ in,R bâ in,int câ in
)T(S34)
aout =(aout,L aout,R bout,R aâ out,L aâ out,R bâ out,R
)T(S35)