Hybrid integration of silicon photonic devices on lithium niobate for optomechanical wavelength conversion
Igor Marinković, Maxwell Drimmer, Bas Hensen, Simon Gröblacher
HHybrid integration of silicon photonic devices on lithium niobatefor optomechanical wavelength conversion
Igor Marinkovi´c, ∗ Maxwell Drimmer, ∗ Bas Hensen, and Simon Gr¨oblacher † Kavli Institute of Nanoscience, Department of Quantum Nanoscience,Delft University of Technology, 2628CJ Delft, The Netherlands
The rapid development of quantum information processors has accelerated the demand for technologies thatcan connect them to a quantum network. One promising approach is to use mechanical resonators as an in-termediary between microwave and optical fields. Signals from a superconducting, topological, or spin qubitprocessor can then be converted coherently to optical states at telecom wavelengths. However, current devicesbuilt from homogeneous structures su ff er from added noise and small conversion e ffi ciency. Combining ad-vantageous properties of di ff erent materials into a heterogeneous design should allow for superior quantumtransduction devices – so far these hybrid approaches have however been hampered by complex fabrication pro-cedures. Here we present a novel integration method based on previous pick-and-place ideas, that can combineindependently fabricated device components of di ff erent materials into a single device. The method allows forprecision alignment by continuous optical monitoring during the process. Using our method, we assemble ahybrid silicon-lithium niobate device with state-of-the-art wavelength conversion characteristics. Hybrid photonic devices have attracted significant attentionfor their potential in both classical and quantum informationprocessing [1–4]. While individual materials rarely possessall desired properties, the combination of several allows forsuperior designs needed for the realization of photonic cir-cuits that include light generation, guiding, modulation, anddetection. For example, the integration of silicon photonic cir-cuits with single photon sources [5], two dimensional materi-als [6], and classical light sources [7], has been demonstrated,promising new capabilities beyond what is achievable with ahomogeneous approach. This optimization and combinationof several desired properties often comes at the expense of sig-nificantly more complex fabrication procedures, complicatingthe development of more advanced hybrid photonic devices.Di ff erent material systems typically react di ff erently to chem-icals or etching procedures, leading to incompatibilities in thefabrication process. Here we present a novel approach to thefabrication of hybrid devices, based on a pick-and-place pro-cedure, which is agnostic to the photonic material and com-patible across a large range of di ff erent platforms. Addition-ally, this technique can use in-situ alignment to achieve ac-curate positioning without a complex imaging system. Wedemonstrate the capabilities of our new method by combininga silicon photonic crystal cavity with a piezoelectric lithiumniobate transducer and experimentally demonstrate state-of-the-art microwave-to-optics wavelength conversion.Coherent conversion of quantum states between opticaland microwave frequencies through a quantum transducer hasbecome an attractive candidate to connect superconductingquantum processors. Though well-suited for local manipu-lation, the low frequency of superconducting circuits makesprocessors in distant cryostats di ffi cult to connect [8, 9]. Aquantum transducer can solve this issue by converting quan-tum information into the optical domain, where it is protectedagainst room temperature thermal noise [10, 11]. Low-loss ∗ These authors contributed equally to this work. † [email protected] optical fibers can then be used to transmit information overlarge distances, creating a network of connected quantum pro-cessors [12, 13].In particular, electro-optomechanical devices have emergedas an leading platform for realizing quantum transducers [14–25]. In such a device, an optomechanical interaction is usedto transfer quantum excitations between optical and mechan-ical modes [4, 26], while a resonant electromechanical (of-ten piezoelectric) drive can be used for e ffi cient conversionbetween mechanical and microwave modes [27]. The ma-nipulation of a quantum state of the mechanical resonatorat the single phonon level has been demonstrated with bothoptomechanical [28–30] and piezoelectric interactions [31–33]. Most recently, photons from a microwave qubit havebeen converted into telecom photons [25] using a hybrid Alu-minum Nitride-on-Silicon-on-Insulator platform. While thesefirst proof-of-principle experiments are highly encouraging,further improvements to the e ffi ciency and fidelity will requireeven stronger piezoelectric materials.State-of-the-art quantum optomechanical experiments pri-marily use silicon due to its high refractive index and photoe-lasticity, which enable large optomechanical coupling rates,as well as its small optical absorption. On the other hand,some of the most promising candidates for facilitating ef-ficient coherent interactions between microwaves and me-chanics are highly piezoelectric materials like LiNbO andAlN [19, 23, 25]. Therefore, combining silicon photonicswith these piezoelectric materials into a single hybrid electro-optomechanical device naturally emerges as an attractive ap-proach to establish a coherent link between microwave andoptical modes [34]. Single material (or homogeneous) ap-proaches have been investigated [21, 23], but so far su ff erfrom a lack of a material that fulfills requirements for bothstrong piezoelectric and optomechanical coupling. Hybrid de-vices o ff er better capabilities, however, this often comes at theexpense of a more complex nanofabrication procedure.We expand on earlier demonstrations of pick-and-placeassemblies [5] by developing a novel technique that al-lows a greatly simplified approach to making hybrid electro- a r X i v : . [ c ond - m a t . m e s - h a ll ] O c t optomechanical devices. Working with only a simple microp-ositioning stage, a microscope, and a digital camera, we uti-lize a tapered optical fiber to transfer a silicon photonic deviceonto a piezoelectric chip (cf. Fig. 1). Replacing commonlyused tungsten tips with a tapered fiber enables one to use thephotonic cavity’s evanescent field as a high precision positionsensor. The simplicity and flexibility of the technique makesit well suited for the development of novel devices, previouslyonly possible through di ffi cult and lengthy fabrication proce-dures, as well as proof-of-principle experiments with almostany combination of materials. Unlike wafer scale bonding ap-proaches, pick-and-place techniques enable a straightforwardapproach to further integration of electro-optomechanical de-vices with other quantum technologies without significantlychanging fabrication procedures. This material-independenttechnique is useful for rapid prototyping and integrating newmaterial combinations in hybrid photonic circuits. It is anespecially attractive approach for coupling cavities to singlephoton sources, as measurement of coupling during the place-ment can enable optimal positioning of the cavity. We illus-trate the capabilities of our new technique by demonstratinga silicon photonic crystal cavity combined with a LiNbO electro-mechanical system, previously a di ffi cult to realizematerial combination. We further experimentally characterizethe device and demonstrate its potential for quantum transduc-tion tasks. (a)(d) (b)(e) (c)(f) ͢ ↑ ↑ ͢ ͢ ͢ a s ͢ ͢ a s ͢ FIG. 1.
Slapping: Pick-and-place assembly with tapered opticalfibers.
A cartoon depiction of the ‘slapping’ procedure. (a) Photoniccrystal nanobeams are patterned in a thin film of silicon (white) re-leased from a thick oxide layer (gray). (b) A tapered optical fiber(cyan) touches the nanobeam and sticks due to van der Waals forces.(c) The thin tether of silicon connecting the nanobeam to its chipis broken through repeated motion with the fiber and the photonicdevice is lifted away. (d) The fiber and nanobeam are brought intoclose proximity with the substrate of a di ff erent chip (yellow). Theoptical and mechanical spectra of the nanobeam can be measured inorder to map the surface and locate an optimal placement location(indicated by the red outline), in principle creating the possibility toalign to various defects, such as spins (blue), in the new material. (e)The nanobeam is then touched down onto the surface. (f) Once thedevice is properly assembled, the fiber is lifted. Device design and fabrication
Several quantum transducers using thin film lithium niobatehave shown promising results [19, 20, 23], due to the largepiezoelectric tensor of the material. However, these homoge-neous approaches failed to achieve the high optomechanicalcoupling necessary for e ffi cient transduction, as LiNbO hasa relatively small refractive index ( n LN ≈ . ff ective mass is added to the mechanicalmode. We design our electromechanical device with 2 fingerpairs and dimensions of 0 . × . × µ m, with a pitch of1.5 µ m. From finite-element simulations, we expect an e ff ec-tive electromechanical coupling coe ffi cient k ff ≈
1% for ourmembrane design.The silicon nanobeam optical cavity on top of the mem-brane is formed by a photonic crystal mirror at each end and atapered defect region in the middle [37], made out of 250 nmthick silicon. Finite-element simulations of the bare siliconphotonic crystal cavity nanobeam show a fundamental reso-nance around 1565 nm with a quality factor exceeding 2 × .When the nanobeam is placed on top of the LiNbO mem-brane its resonance is shifted to around 1595 nm (see Fig. 2cfor details). We further calculate an optomechanical couplingrate of 24 kHz for an optimally positioned nanobeam.We fabricate the lithium niobate devices using two elec-tron beam lithography steps. The first places electrodes ontop of the membrane and the second etches into the lithiumniobate. The silicon nanobeam is fabricated in a single lithog-raphy step, with the photonic crystal cavity connected withsingle narrow bridge to the rest of the chip in order to be ableto pick it up with the fiber during the transfer procedure (seethe Supplementary Information for more details). FIG. 2.
Hybrid piezo-optomechanical devices. (a) A false colorscanning electron microscope (SEM) image of a LiNbO film acous-tic resonator with a silicon photonic crystal cavity placed on top. Thereleased LiNbO is light gray, the silicon nanocavity is blue, the goldelectrodes are yellow, and the lithium niobate marker used for po-sitioning is purple. The silicon cavity features a tapered waveguidethat is used for coupling light from an optical fiber (not shown). (b)Simulated strain component s yy , which is dominant for photoelasticand piezoelectric coupling. Shown is the profile of a mode of thehybrid device, top view (left) and cross-section with the mechanicaldeformation (right). (c) Fundamental resonant optical mode of thesilicon photonic crystal cavity nanobeam, viewed from the top. Theindicated axes correspond to the lithium niobate axes, while the sili-con nanobeam is fabricated from a [100] wafer, with the cavity alongthe [110] direction. Slapping technique
Our approach for making hybrid photonic integrated cir-cuits is based on the ‘pick-and-place’ method, where struc-tures are removed from a ‘donor chip’ made of a desired ma-terial and are transferred to a ‘device chip’ made of a tradi-tional photonic material. In the past, this transfer has been ac-complished using AFM tips [38], polymer stamps [39, 40], ortungsten nano-manipulator probes [5, 41–43]. In contrast, wedeveloped a new method using an optical fiber, which we referto as slapping . One of the major advantages of our approachis that we can achieve high-precision positioning without sac-rificing the ease of operation, because the silicon photonic de-vices can be directly and continuously measured during thetransfer process.Tapered optical fibers are being used to couple light intonanophotonic devices with high e ffi ciency [44–46] and havealso been used to rip away loosely connected photonic crystalcavities from the chip they were fabricated on [47, 48]. Here,we extend this technique by placing the cavities onto a di ff er-ent substrate to create a hybrid photonic device. In our proce- dure (cf. Fig. S3), silicon photonic crystal cavity nanobeamsare attached by a thin tether ( ≈
50 nm) on a donor chip. Thechip is placed on a piezo-controlled motorized stage whereit is viewed from above through a 500x microscope objec-tive using a CCD camera. A tapered fiber is placed above thedonor chip at an angle of a few degrees lower than horizontal,such that the tip is nearly parallel to the surface of the donorchip. The motorized stage is then used to touch the waveg-uide of a photonic crystal cavity using the tapered fiber andvan der Waals forces cause the fiber to adhere to the silicon.At this point, the optical and mechanical spectra of the op-tomechanical nanobeam can be measured with a tunable laserand photodetector using the reflected light in the optical fiber.This measurement can be used to pre-select photonic struc-tures supporting high quality optical and mechanical modes,increasing the yield of the final devices.Once the fiber is stuck to the nanobeam, the stages can bemoved back and forth by a few microns until the tether brakesand the nanobeam is ripped away. The donor chip is then low-ered away from the fiber and the device chip is moved into itsplace. The fiber is brought close to the surface of the devicechip such that the silicon nanobeam and lithium niobate mem-brane are in focus simultaneously. Rough alignment can bedone using the camera and microscope. Using the evanescentfield of the cavity the environment surrounding the nanobeamcan be probed and its position with respect to the substrate canbe determined. To further improve the accuracy of this align-ment we include a thin marker in the LiNbO layer near themechanical resonator (see Fig. 2a). By moving the nanobeamover this marker and monitoring the optical spectrum a sharpreduction in the quality factor of optical modes due to scat-tering in close proximity to the marker can be observed (cf.Fig. S4). Using this strategy, we are able to position the centerof the optical cavity directly over the marker before transfer-ring it to the mechanical resonator, which results in reliablyplacing our nanobeams with an accuracy of less than 100 nm,which is significantly better than what is possible with opticalimaging methods. The angular alignment accuracy is typi-cally (cid:28) ◦ . Additional reduction in the alignment error caneasily be realized through contact sensing and monitoring ofthe mechanical resonance of the cavity.Once positioned, the fiber is then lowered until thenanobeam touches down onto the membrane. As the van derWaals force makes the nanobeam stick to the substrate morestrongly than to the fiber, the fiber can be lifted away fromthe hybrid structure. A finalized device using this methodis shown in Figure 2a. We note that this technique is notmaterial-specific and only uses materials typically used in alaboratory capable of optical characterization. Results
After assembly, we measure the microwave and optical re-flection spectra of our devices, which were all patterned withthe electrodes wired to large pads in a ground-signal-ground(GSG) configuration. We measure the signal in reflection us-ing a coaxial RF GSG probe and a vector network analyzer Ω m /2 π = 1.400 GHzQ = 383 Ω m /2 π = 1.734 GHzQ = 260 Ω m /2 π = 1.854 GHzQ = 632 Ω m /2 π = 2.056 GHzQ = 3400.0-1.0-2.0-3.0-4.0 M i c r o w a v e S ( d B ) Frequency (GHz)
FIG. 3.
Characterization of a piezo-mechanical device.
Elec-tromechanical reflection spectrum of a fabricated lithium niobate de-vice before the slapping procedure. The frequency and quality factorof four significant modes are indicated. After slapping, the mode fre-quencies shift slightly and the mode linewidths change (cf. Fig. 4). (VNA) and the signal is normalized to the open response ofthe probe.For our full procedure, we first characterize the lithium nio-bate resonator before slapping the silicon nanobeam by mea-suring the microwave reflection spectrum. Prominent dipsin the reflection spectrum (see Fig. 3) indicate power trans-fer into the mechanical modes of the device. After slapping,the same measurement reveals that the resonances are slightlyshifted, while the quality factors experience a modest decrease(Fig. 4b, red trace). We observe groups of modes spacedby approximately 330 MHz, which matches the free spectralrange of the A0-like mechanical modes in our simulations.Optical spectroscopy of the hybrid device is performed usinga tunable external-cavity diode laser. The reflection spectrumof the device shows two prominent resonances at 1588 and1617 nm with a linewidth of 6.6 GHz and 11.3 GHz, respec-tively (Fig. 4a).We then proceed to measure microwave-to-optical trans-duction at room temperature. A microwave tone is sweptover the device, exciting the mechanical supermodes of thehybrid nanobeam-and-membrane structure on resonance. Atthe same time, optical laser light detuned to the blue side of thecavity by 1.4 GHz is continuously pumped into the nanobeam,through an optical circulator and a tapered optical fiber. Onresonance, the mechanical motion excited by the microwavesignal will drive the optomechanical Stokes process and cre-ate a sideband resonant with optical cavity. The optical inten-sity resulting from the beat note between the reflected cavityand pump photons can be measured by monitoring the lightreflected from the cavity on a fast photodiode. Key param-eters such as single-photon optomechanical coupling rate g and microwave-to-optical e ffi ciency η eo can be extracted bycomparing the response of the output of the fast photodetectorto the microwave input. Using a VNA yields a transmission-type S measurement of microwave-to-optical transduction(Fig. 4b, blue trace).We select the mechanical mode at 1.7 GHz as our mechan-ical resonance as it has the highest value in our S measure-ment. The input optical power sent into the cavity through the tapered fiber is I in = . µ W, while the overall received opti-cal power at the photodiode of I rec =
804 nW. Following [19],the single-photon optomechanical coupling rate g can be ex-tracted from the S in the sideband resolved regime, wherewe use R load = Ω for the impedance of the VNA ports and R PD = / W as the responsivity of the photodiode. In ourcase the anti-Stokes sideband is not completely suppressed,leading to a slight under-estimation of g . The microwave-to-optical transduction e ffi ciency η µ → o is the product of the singlephoton cooperativity C , the intracavity photon number n cav ,and the microwave and optical external coupling e ffi ciencies( η ext ,µ , η ext , o , respectively): η µ → o = C · n cav · η ext ,µ · η ext , o = g γκ · n cav · γ e γ · κ e κ (1) FIG. 4.
Microwave-to-optical transduction using a slapped Si-on-LiNbO device. (a) Optical reflection spectrum. (b) Full acoustic re-flection ( S , red) and transmission ( S , blue) spectra. (c) Diagramof the experimental setup used to characterize the transduction of thedevice (indicated by the green arrow). PD, photodiode; VNA, vectornetwork analyzer. We find g / π ≈
10 kHz, which matches well with thesimulated value of 13 kHz for a slightly non-optimally placednanobeam. While the large loss rates of the optical and me-chanical resonators of 6 . C o n cav > ffi ciencies result in a photon num-ber conversion e ffi ciency of η µ → o ≈ × − at an opticalpower of I in = µ W in the fiber. This result is on parwith homogeneous state-of-the-art LiNbO devices [19, 23].While [19] exhibits large microwave to mechanics conversion γ e /γ = .
15, it su ff ers from small optomechanical coupling g / π = . g / π ≈
80 kHz,but at the expense of low γ e /γ = . g / π =
10 kHz and γ e /γ = . ∼ π ×
40 MHz.An additional advantage of our design is the better thermallyanchoring, resulting in a reduced susceptibility to optical ab-sorption induced heating.Additional adjustments to our platform can be used to sig-nificantly improve the performance of this new class of de-vices by several orders of magnitude. Coupling between mi-crowave and mechanical modes can be increased by using anon-chip impedance matching circuit and switching to S0-like modes. In our LiNbO membranes these modes su ff ered fromsmall quality factors, which can be overcome by adjustmentsto the fabrication procedure, as similar designs have demon-strated S0-like modes with much larger Q factors [49]. Thiswill enable us to use a smaller membrane (while preservingthe piezo k e f f ) which will increase the optomechanical cou-pling rate. Further improvements in the fabrication will leadto higher optical quality factors and hence better overall ef-ficiency, which has already been demonstrated in a similarstructure [50]. Our new slapping approach is also directly ap-plicable to precision positioning of photonic crystal cavities tospins in a substrate, such as for color centers in diamond [51]and rare-earth ions [52, 53]. ACKNOWLEDGMENTS
We would like to thank Robert Stockill, Moritz Forsch,and Roald van der Kolk for valuable discussions. Thiswork is supported by the Foundation for Fundamental Re-search on Matter (FOM) Projectruimte grant (16PR1054), theEuropean Research Council (ERC StG Strong-Q, 676842),and by the Netherlands Organization for Scientific Research(NWO / OCW), as part of the Frontiers of Nanoscience pro-gram, as well as through Vidi (680-47-541 / [1] O. Benson, Nature , 193 (2011).[2] D. Marpaung, J. Yao, and J. Capmany, Nature Photon. , 80(2019).[3] A. W. Elshaari, W. Pernice, K. Srinivasan, O. Benson, andV. Zwiller, Nature Photon. , 285 (2020).[4] Y. Chu and S. Gr¨oblacher, arXiv:2007.03360 (2020).[5] J.-H. Kim, S. Aghaeimeibodi, C. J. K. Richardson, R. P. Leavitt,D. Englund, and E. Waks, Nano Lett. , 7394 (2017).[6] M. Liu, X. Yin, E. Ulin-Avila, B. Geng, T. Zentgraf, L. Ju,F. Wang, and X. Zhang, Nature , 64 (2011).[7] D. Thomson, A. Zilkie, J. E. Bowers, T. Komljenovic, G. T.Reed, L. Vivien, D. Marris-Morini, E. Cassan, L. Virot, J.-M.F´ed´eli, J.-M. Hartmann, J. H. Schmid, D.-X. Xu, F. Boeuf,P. O’Brien, G. Z. Mashanovich, and M. Nedeljkovic, J. Opt. , 073003 (2016).[8] G. Wendin, Rep. Prog. Phys. , 106001 (2017).[9] P. Magnard, S. Storz, P. Kurpiers, J. Sch¨ar, F. Marxer, J. L¨utolf,J.-C. Besse, M. Gabureac, K. Reuer, A. Akin, B. Royer,A. Blais, and A. Wallra ff , arXiv:2008.01642 (2020).[10] N. J. Lambert, A. Rueda, F. Sedlmeir, and H. G. L. Schwefel,Adv. Quantum Technol. , 1900077 (2020).[11] N. Lauk, N. Sinclair, S. Barzanjeh, J. P. Covey, M. Sa ff man,M. Spiropulu, and C. Simon, Quantum Sci. Technol. , 20501(2020).[12] H. J. Kimble, Nature , 1023 (2008).[13] S. Wehner, D. Elkouss, and R. Hanson, Science , eaam9288(2018).[14] L. Tian and H. Wang, Phys. Rev. A , 053806 (2010).[15] J. Bochmann, A. Vainsencher, D. D. Awschalom, and A. N. Cleland, Nature Phys. , 712 (2013).[16] T. Bagci, A. Simonsen, S. Schmid, L. G. Villanueva,E. Zeuthen, J. Appel, J. Taylor, A. Sørensen, K. Usami,A. Schliesser, and E. Polzik, Nature , 81 (2014).[17] K. C. Balram, M. I. Davanc¸o, J. D. Song, and K. Srinivasan,Nature Photon. , 346 (2016).[18] A. P. Higginbotham, P. S. Burns, M. D. Urmey, R. W. Peterson,N. S. Kampel, B. M. Brubaker, G. Smith, K. W. Lehnert, andC. A. Regal, Nature Phys. , 1038 (2018).[19] L. Shao, M. Yu, S. Maity, N. Sinclair, L. Zheng, C. Chia,A. Shams-Ansari, C. Wang, M. Zhang, K. Lai, and M. Lonˇcar,Optica , 1498 (2019).[20] W. Jiang, R. N. Patel, F. M. Mayor, T. P. McKenna,P. Arrangoiz-Arriola, C. J. Sarabalis, J. D. Witmer, R. V. Laer,and A. H. Safavi-Naeini, Optica , 845 (2019).[21] M. Forsch, R. Stockill, A. Wallucks, I. Marinkovi´c, C. G¨artner,R. A. Norte, F. van Otten, A. Fiore, K. Srinivasan, andS. Gr¨oblacher, Nature Phys. , 69 (2020).[22] X. Han, W. Fu, C. Zhong, C.-L. Zou, Y. Xu, A. A. Sayem,M. Xu, S. Wang, R. Cheng, L. Jiang, and H. X. Tang, NatureCommun. , 3237 (2020).[23] W. Jiang, C. J. Sarabalis, Y. D. Dahmani, R. N. Patel, F. M.Mayor, T. P. McKenna, R. Van Laer, and A. H. Safavi-Naeini,Nature Commun. , 1166 (2020).[24] G. Arnold, M. Wulf, S. Barzanjeh, E. S. Redchenko, A. Rueda,W. J. Hease, F. Hassani, and J. M. Fink, Nature Commun. ,4460 (2020).[25] M. Mirhosseini, A. Sipahigil, M. Kalaee, and O. Painter,arXiv:2004.04838 (2020). [26] M. Aspelmeyer, T. J. Kippenberg, and F. Marquardt, Rev. Mod.Phys. , 1391 (2014).[27] M. Wu, E. Zeuthen, K. C. Balram, and K. Srinivasan, Phys.Rev. Applied , 014027 (2020).[28] J. Chan, T. P. M. Alegre, A. H. Safavi-Naeini, J. T. Hill,A. Krause, S. Gr¨oblacher, M. Aspelmeyer, and O. Painter, Na-ture , 89 (2011).[29] R. Riedinger, S. Hong, R. A. Norte, J. A. Slater, J. Shang, A. G.Krause, V. Anant, M. Aspelmeyer, and S. Gr¨oblacher, Nature , 313 (2016).[30] A. Wallucks, I. Marinkovi´c, B. Hensen, R. Stockill, andS. Gr¨oblacher, Nature Phys. , 772 (2020).[31] Y. Chu, P. Kharel, W. H. Renninger, L. D. Burkhart, L. Frunzio,P. T. Rakich, and R. J. Schoelkopf, Science , 199 (2017).[32] K. J. Satzinger, Y. P. Zhong, H.-S. Chang, G. A. Peairs, A. Bi-enfait, M.-H. Chou, A. Y. Cleland, C. R. Conner, E. Dumur,J. Grebel, I. Gutierrez, B. H. November, R. G. Povey, S. J.Whiteley, D. D. Awschalom, D. I. Schuster, and A. N. Cleland,Nature , 661 (2018).[33] P. Arrangoiz-Arriola, E. A. Wollack, Z. Wang, M. Pechal,W. Jiang, T. P. McKenna, J. D. Witmer, R. V. Laer, and A. H.Safavi-Naeini, Nature , 537 (2019).[34] A. H. Safavi-Naeini, D. V. Thourhout, R. Baets, and R. V. Laer,Optica , 213 (2019).[35] B. A. Auld, Acoustic fields and waves in solids (Wiley Inter-science Publications, 1973).[36] C. Campbell,
Surface Acoustic Wave Devices for Mobile andWireless Communications, Four-Volume Set (Academic press,1998).[37] J. Chan, A. H. Safavi-Naeini, J. T. Hill, S. Meenehan, andO. Painter, App. Phys. Lett. , 081115 (2012).[38] A. W. Schell, G. Kewes, T. Schr¨oder, J. Wolters, T. Aichele,and O. Benson, Rev. Sci. Instr. , 073709 (2011).[39] R. Katsumi, Y. Ota, M. Kakuda, S. Iwamoto, and Y. Arakawa,Optica , 691 (2018).[40] A. M. Dibos, M. Raha, C. M. Phenicie, and J. D. Thompson,Phys. Rev. Lett. , 243601 (2018).[41] S. L. Mouradian, T. Schr¨oder, C. B. Poitras, L. Li, J. Gold-stein, E. H. Chen, M. Walsh, J. Cardenas, M. L. Markham, D. J.Twitchen, M. Lipson, and D. Englund, Phys. Rev. X , 031009(2015).[42] I. E. Zadeh, A. W. Elshaari, K. D. J¨ons, A. Fognini, D. Dalacu,P. J. Poole, M. E. Reimer, and V. Zwiller, Nano Lett. , 2289(2016).[43] N. H. Wan, T.-J. Lu, K. C. Chen, M. P. Walsh, M. E. Trusheim,L. D. Santis, E. A. Bersin, I. B. Harris, S. L. Mouradian, I. R.Christen, E. S. Bielejec, and D. Englund, Nature , 226(2020).[44] S. Gr¨oblacher, J. T. Hill, A. H. Safavi-Naeini, J. Chan, andO. Painter, Appl. Phys. Lett., , 181104 (2013).[45] T. G. Tiecke, K. P. Nayak, J. D. Thompson, T. Peyronel, N. P.de Leon, V. Vuleti´c, and M. D. Lukin, Optica , 70 (2015).[46] M. J. Burek, C. Meuwly, R. E. Evans, M. K. Bhaskar,A. Sipahigil, S. Meesala, B. Machielse, D. D. Sukachev,C. T. Nguyen, J. L. Pacheco, E. Bielejec, M. D. Lukin, andM. Lonˇcar, Phys. Rev. Appl. , 024026 (2017).[47] J. D. Thompson, T. G. Tiecke, N. P. de Leon, J. Feist, A. V.Akimov, M. Gullans, A. S. Zibrov, V. Vuleti´c, and M. D. Lukin,Science , 1202 (2013).[48] L. Magrini, R. A. Norte, R. Riedinger, I. Marinkovi´c, D. Grass,U. Deli´c, S. Gr¨oblacher, S. Hong, and M. Aspelmeyer, Optica , 1597 (2018).[49] S. Gong and G. Piazza, IEEE Trans. Microw. Theory Tech. ,403 (2013). [50] J. D. Witmer, J. T. Hill, and A. H. Safavi-Naeini, Opt. Express , 5876 (2016).[51] B. J. M. Hausmann, B. J. Shields, Q. Quan, Y. Chu, N. P.de Leon, R. Evans, M. J. Burek, A. S. Zibrov, M. Markham,D. J. Twitchen, H. Park, M. D. Lukin, and M. Lon˘car, NanoLett. , 5791 (2013).[52] J. M. Kindem, A. Ruskuc, J. G. Bartholomew, J. Rochman,Y. Q. Huan, and A. Faraon, Nature , 201 (2020).[53] M. Raha, S. Chen, C. M. Phenicie, S. Ourari, A. M. Dibos, andJ. D. Thompson, Nature Commun. , 1 (2020).[54] I. Marinkovi´c, Optomechanical Devices in the QuantumRegime , Ph.D. thesis, Delft University of Technology (2019).[55] R. Wang, S. A. Bhave, and K. Bhattacharjee, in
Proceedings ofthe IEEE International Conference on Micro Electro Mechani-cal Systems (MEMS) (2013) pp. 165–168.[56] C. Wang, M. J. Burek, Z. Lin, H. A. Atikian, V. Venkataraman,I.-C. Huang, P. Stark, and M. Lonˇcar, Opt. Express , 30924(2014).[57] P. Arrangoiz-Arriola, E. A. Wollack, M. Pechal, J. D. Witmer,J. T. Hill, and A. H. Safavi-Naeini, Phys. Rev. X , 031007(2018).[58] M. Itano, F. W. Kern, M. Miyashita, and T. Ohmi, IEEE Trans.Semicond. Manuf. , 258 (1993).[59] L. Cai, A. Mahmoud, and G. Piazza, Opt. Express , 9794(2019). SUPPLEMENTARY INFORMATIONA. Fabrication
We begin with a Lithium Niobate on Silicon (LNOS) 5 ×
10 mm chip, consisting of a 340 nm layer of X-cut, singlecrystal lithium niobate doped with 5 mol% MgO bonded toa 500 µ m thick silicon substrate. Electrodes are defined byfirst patterning a 280 nm thick coating of electron beam re-sist and then evaporating an adhesion layer consisting 5 nmof chromium and 15 nm of platinum under a 40 nm layer ofgold. The evaporation step is followed by a lift-o ff , wherethe chip is placed in a beaker of anisole heated to 80 ◦ C andset in an ultrasonic bath for 5 min. Next, the second electronbeam lithography step defines the shape of the thin film res-onators in a thick 550 nm coating of CSAR62.13 resist. Theexposed lithium niobate is etched through by a large cross-sectional beam of argon ions incident at a 90-degree angleto the surface of the chip. This process etches through thelithium niobate membrane and results in a sidewall angle of75 ◦ . The devices are released using a dry SF reactive ionetch which isotropically etches the silicon through the holesin the lithium niobate film. The resist is not stripped until af-ter the underetching step in order to prevent contamination ofthe etcher. At this point, resist removal is achieved by ashingin oxygen plasma. Finally, we employ an inorganic cleaningstep to remove lithium niobate redeposited during the physi-cal etching. We prepare a mixture similar to RCA-1, a 2:2:1ratio of NH OH [28%] : H O [31%] : H O. The mixture isheated to 85 ◦ C and mixed with a stir bar for 5 minutes beforeour samples are placed in for another 5 minutes. A fabricateddevice is shown in Fig. S1a. The silicon nanobeam fabrica-tion is described in detail in [54], with the liquid HF undercutbeing replaced with vapor HF. A fabricated nanobeam deviceis shown in Fig. S1b.
B. Redeposition of Lithium Niobate
Despite being a widely used material across multiple in-dustries for decades, the nanofabrication of lithium niobateis considered di ffi cult. There is no known chemical etch for FIG. S1.
Fabrication. (a) A SEM image of a lithium niobate res-onator on the device chip before the slapping procedure. (b) A SEMimage of a silicon photonic crystal cavity nanobeam on the donorchip before slapping. The structure is connected to the chip by thetether at the top. lithium niobate that produces desirable devices with accept-able sidewalls and low surface roughness without unwantedcontamination. This leaves as the only viable alternativephysical etching. Inspired by Refs. [55–57], we process ourlithium niobate using argon ion beam etching (also called ionmilling) in an inductively coupled plasma etcher. However,ion milling lithium niobate causes amorphous LiNbO rede-position, creating unwanted features and increasing sidewallroughness as seen in Figs. S2a and S2b. We mitigate thisproblem first by optimizing the plasma etching parameters tominimize redeposition and then by chemically cleaning oursample. The results are shown in Fig. S2c. The chemicalcleaning step is based on RCA-1, which is typically used onsilicon devices [58] and found to also be e ff ective on lithiumniobate [59]. FIG. S2.
Redeposition of Lithium Niobate. (a) A SEM imageof lithium niobate after ion milling, with the profile in false color.LiNbO is orange and the remaining resist is blue. After millingthrough the LiNbO , angular features about 45 nm wide and 60 nmtall are redeposited above the LiNbO layer of our sample, indicatedby Marker 1. Marker 2 points to the resulting rough sidewalls. (b) Acurved structure in a suspended thin film of LiNbO shows the sameissues. (c) The same device after RCA-1 cleaning. The sidewallsappear smooth and no redeposited features are present. FIG. S3.
Slapping Pick-up. (a) An optical microscope image of a photonic crystal cavity nanobeam before slapping. (b) A tapered opticalfiber is placed in contact with the tapered waveguide of the nanobeam, allowing to measure its resonance with high e ffi ciency. (c) The fiber isthen moved around, dragging the nanobeam with it until the tether is broken. (d) With the tether broken, the nanobeam can be moved awayfrom the chip. (e) Finally, the fiber is lifted from the donor chip. The nanobeam is now ready to be slapped onto a new device. C. Slapping Procedure
The slapping technique consists of first picking up and thenslapping down a structure from a donor chip onto a devicechip. The pick-up procedure begins by placing a silicon donorchip under a microscope and locating a photonic crystal cav-ity nanobeam (Fig. S3a). In this image, the device is con-nected to the chip by a thin tether at the bottom (see SEM inFig. S1b). Next, a tapered optical fiber is lowered until the de-vice and the tip of the fiber are simultaneously in focus. Thetapered fiber is then placed on the tapered waveguide of thenanobeam (Fig. S3b). It is easy to distinguish when the fiberis touching the nanobeam, as the fiber changes color on con-tact. At this stage, the optical and mechanical spectra of thenanobeam device can be measured in order to confirm that it isfunctioning as intended. When the fiber is moved to the left orthe right, the nanobeam adheres and pivots around the tether(Fig. S3c). We find that sweeping the fiber left and right by afew microns reliably breaks a 50-100 nm thick tether. Oncethe tether is broken, the nanobeam can be pulled away fromthe donor chip (Fig. S3d) and finally be lifted (Fig. S3e).Next, after picking up the nanobeam, it can be placed backdown wherever desired. Slapping the nanobeam only requiresthat the structure sticks better to the new substrate than it doesto the tapered fiber tip. This is usually true for any material, asthe nanobeam is picked up in such a way that it contacts onlya small fraction of the nanobeam’s surface area. When thenanobeam is slapped down, its entire bottom surface touchesthe substrate and thus it adheres. We found that once slapped,the structures’ position can be further adjusted by pushingthem across the substrate with the tapered fiber after slapping.
D. Increasing Slapping Accuracy with In-Situ Optical Sensing
The accuracy of slapping an arbitrary structure is in princi-ple limited by the optical di ff raction limit and practically bythe operator. However, in our approach we are picking upoptical cavities with an optical fiber. This allows us to sig- nificantly improve the accurate placement by measuring thephotonic crystal cavity during the slapping procedure.In order to demonstrate and characterize the in-situ align-ment procedure of the position of the cavity with respect toa feature on a device chip, we monitor the resonance as weapproach a marker. The marker is simply a 270 nm thick can-tilever, similar to the lithium niobate marker visible in the bot-tom of Fig. 2b. In order to sense the marker, the nanobeam isplaced a few hundred nanometers above the device chip whilethe optical resonance is measured on reflection using the ta-pered fiber. Scattering drastically reduces the quality factor ofthe optical cavity when the marker is near the defect region ofthe photonic crystal. In addition, the resonances are slightlyredshifted close to the marker (see Fig. S4). Both e ff ects areclear, as the redshift and the linewidth of the fundamental op-tical mode increase with proximity to the marker. By fitting O p t i c a l Q u a l i t y F a c t o r Distance from Marker ( µ m) Wavelength (nm) O p t i c a l R e fl e c t i o n ( a r b . u n i t s ) FIG. S4.
In-Situ Sensing.
Measurement of the silicon optomechan-ical crystal nanobeam attached to an optical fiber enables increasedpositioning accuracy with respect to an on-chip marker. The qual-ity factor of the fundamental optical mode is significantly reduced asthe nanobeam approaches the marker. Inset: Corresponding opticalreflection spectrum of the nanobeam at each point. the data to find the position of maximum absorption, we canlocate the center of the photonic crystal cavity directly overthe on-chip marker. Using this technique, we are able to posi-tion a nanobeam with an accuracy of better than 100 nm, wellunder the optical di ff raction limit (100s of nanometers in thiscase). E. Photonic Crystal Design
The cavity is formed by two photonic crystal mirrors ateach end and a tapered region in the middle, designed andfabricated according to [50]. The unit cell periodicity of themirror is a m =
338 nm and hole axes are r =
78 nm and r =
258 nm, while the width of the nanobeam is w =
650 nmand the thickness of the silicon layer is 250 nm. The latticespacing is quadratically tapered to a value a d = . g of various mechanical modesto the photonic cavity, we confirm that the coupling inside thelithium niobate is small and hence can be neglected, as mostof the optical mode is contained inside of the silicon. Therelatively low quality factor of the fabricated cavities can beattributed to poor etching of the silicon. F. g of S0 mode We calculate the optomechanical coupling rate for S0-like modes of our hybrid device (Fig S5), as these modeshave higher piezoelectric coupling. We obtain 25 kHz for a3 .
68 GHz mode.
FIG. S5.
S0-like mode of the device . This mode gives the bestperformance, as most of the device is either coupled to optics or mi-crowaves through high photoelastic / piezoelectric coe ffi cients. G. Estimation of g Here we would like to sketch how we estimate g from ourexperimental data. Light in the photonic cavity can be decom-posed into three fields: the pump and the two optical sidebandsshifted from the pump by +Ω m and − Ω m . Expressions for theamplitudes of these waves ( A , A + and A − ) are derived in [19]: A = κ e A in i ∆ + κ/ A − = − iGA i ( ∆ − Ω d ) + κ/ A + = − iGA i ( ∆ + Ω d ) + κ/ ∆ is the detuning of the laser ( ω ) with respect tothe cavity. G = g P in (cid:126) Ω d γ e /γ , with Ω d and P in being the fre-quency and power of the microwave drive of the mechanicalresonator, respectively, and A in = I in / (cid:126) ω . We drive the me-chanics with microwaves at resonance so Ω d = Ω m . Lightcoming out of the cavity is then given by: A = ( A in − √ κ e A ) e i ω t − √ κ e A − e i ( ω +Ω m ) t − √ κ e A + e i ( ω − Ω m ) t After propagating to the detector with an e ffi ciency η = .
28, we detect the signal at the mechanical frequency withan amplitude U = R PD (cid:126) ω | A | . We then calculate the powerthat the VNA receives as P out = U / (2 R load ), and finally ob-tain S = P out / P in . Using this relationship between g and S we can then numerically calculate g . H. V π of a Hybrid Device We have characterized another device with higher mechan-ical frequency and slightly smaller optical linewidth. This en-ables us to clearly see the modification of the optical reflec-tion spectrum due the presence of an RF drive on the IDTs(see Fig. S6), as done in Ref. [19]. This makes the resonancefrequency of the cavity to oscillate with respect to the laserat the drive frequency. In the cavity’s frame of reference thelaser frequency is modulated, giving rise to sidebands at theRF drive frequency. We use the formula given in [19] S = π R PD I rec V π (S1)to estimate the V π ≈
50 mV of our device. -30 dBm-20 dBm1566.68 1566.72 1566.76 1566.8
Wavelength (nm) R e fl e c t i o n ( a r b . u n i t s ) FIG. S6. V π of a slapped hybrid deviceof a slapped hybrid device