Tunable open-access microcavities for on-chip cQED C.A. Potts ECE Department, University of Alberta, T6G 2V4, Edmonton, AB, Canada and Department of Physics, University of Alberta, T6G 2E1 Edmonton, AB, Canada A. Melnyk, M.H. Bitarafan, and R.G. DeCorby∗ ECE Department, University of Alberta, T6G 2V4, Edmonton, AB, Canada H. Ramp and J.P Davis† Department of Physics, University of Alberta, T6G 2E1 Edmonton, AB, Canada 6 1 D. Vick 0 2 National Institute for Nanotechnology, T6G 2M9, Edmonton, AB, Canada n L.J. LeBlanc a J Department of Physics, University of Alberta, T6G 2E1 Edmonton, AB, Canada and Canadian Institute for Advanced Research, Toronto, M5G 1Z8 Canada 3 (Dated: January 14, 2016) 1 We report on the development of on-chip microcavities and show their potential as a platform ] l forcavityquantumelectrodynamicsexperiments. Microcavityarrayswereformedbythecontrolled l buckling of SiO /Ta O Bragg mirrors, and exhibit a reflectance-limited finesse of 3500 and mode a 2 2 5 h volumesassmallas35λ3. Weshowthatthecavityresonancecanbethermallytunedintoalignment - with the D2 transition of 87Rb, and outline two methods for providing atom access to the cavity. s Owingtotheirsmallmodevolumeandhighfinesse,thesecavitiesexhibitsingle-atomcooperativities e as high as C =65. A unique feature of the buckled-dome architecture is that the strong-coupling m 1 parameter g /κ is nearly independent of the cavity size. Furthermore, strong coupling should be 0 . achievable with only modest improvements in mirror reflectance, suggesting that these monolithic t a devices could provide a robust and scalable solution to the engineering of light-matter interfaces. m - d The implementation of a distributed quantum net- operations,[7,28]andtoimplementanelementaryquan- n work could enable a global quantum communication tum network. [29] These works place single-atom quan- o system,[1,2]distributedquantumcomputation,[3]distri- tum systems as a leading candidate for use in large-scale c bution of quantum entanglement,[4] and may even pro- quantum networks. [ vide a global time standard by embedding atomic clocks As a result, there is a strong interest in the integra- 1 at quantum nodes.[5] To realize these objectives, coher- tion of alkali atoms into robust, scalable, packaged opti- v ent control of individual quantum states, and their in- 4 cal cavities.[30, 31] Furthermore, it is desirable for these teractions, is required. Since photons - the prototypical 4 optical cavities to have small mode volumes and be tun- 3 distributed qubit - exhibit no interactions with them- able to atomic transitions.[32–34] Here we report the de- 3 selves in vacuum, [6] matter systems must act as inter- velopment of ‘buckled-dome’ Fabry-P´erot microcavities 0 mediariestoperformquantum-stateoperations,[7]serve designedforcQEDapplications,specificallyon-chipcou- . as quantum memories,[8] and interface with quantum 1 plingbetweensinglephotonsandsinglerubidiumatoms. 0 processors.[9] These cavities produce high single-atom cooperativities, 6 Numerous systems have been investigated as quantum can be easily tuned to atomic transitions, and can facili- 1 light-matterinterfacesincludingphononsinoptomechan- tate open-access for incorporation of atoms. : icaldevices,[10,11]rare-earth-iondopedcrystals,[12,13] v The buckled-dome microcavities were fabricated via a i nitrogenvacanciesindiamond,[14–16]quantumdots,[17, X monolithic self-assembly procedure. [35, 36] First, a dis- 18] and alkali gases such as rubidium [9, 19] and cesium. tributedBraggreflector(10.5periodsSiO /Ta O ,start- r [20]Ofparticularinterestaresinglealkali-atomstrapped 2 2 5 a ing and ending with Ta O ) was deposited on a fused withinhigh-finesseopticalcavities,inthecontextofcav- 2 5 silica substrate by reactive magnetron sputtering. Mi- ity quantum electrodynamics (cQED). [21, 22] Within crocavities were defined by the lithographic patterning the past decade, single atoms trapped in cavities have of a thin (∼15 nm) low-adhesion fluorocarbon layer, fol- beenusedtostorequantuminformation,[23–25]produce lowedbythedepositionofasecondBraggreflectoriden- on-demandsinglephotons,[26,27]performquantumgate tical to the initial reflector. Films with low loss and highcompressivestress(∼200MPa)wererealizedbyus- ing high target power (200 W), elevated substrate tem- ∗ [email protected] perature (150◦C), and low chamber pressure (4 mTorr). † [email protected] [37] Optical constants for single films were measured us- 2 ing an ellipsometer. The refractive indices of SiO and 2 Ta O were estimated to be 1.49 and 2.14 respectively, 2 5 at a wavelength of 780 nm. At the same wavelength, extinction coefficients less than 10−6 were estimated for both materials. The layer thicknesses were chosen such that all layers are nominally one quarter wavelength thick (λ /4n) at the D2 transition of 87Rb (λ = 780.24 0 0 nm).[38] Using a transfer matrix formalism,[39] and the optical constants extracted from measurements on sin- gle films, a peak reflectance of 0.9991 is predicted for the Bragg mirrors, corresponding to a reflection-limited cavity finesse of F = 3500. Following the deposition of the top mirror, the sam- ples were heated (400◦C) to induce a loss of adhesion between the two mirrors in the region of the fluorocar- bon layer. The built-in compressive stress drives the re- FIG. 2. Transmission spectrum for a 100 µm diameter buckled-dome microcavity, with an optical cavity height of lease of the top mirror, forming a dome-like buckle. The 2.67 µm. The inset shows the fundamental mode resonance height and exact morphology is dependent on the dome ingreaterdetail,withaLorentzianfitshowingafull-widthat diameter.[36,40]Inthisworkmicrocavitieswithbasedi- half-maximumof27pm,correspondingtoκ=2π×6.7GHz, ametersof100µmto300µm,Figure1,andpeakheights and an image of the fundamental mode. ofδ =2.5µmto10µmwerestudied. Wedescribeinde- tailthe100µmdiametercavities,andshowselectresults for larger devices. for 100 µm diameter cavities. Using the relation [41] Optical resonances were examined by performing κ = πc/2LF, an estimate of the finesse may be made, transmission spectroscopy using a fiber-coupled tunable where the effective cavity length is L=δ+2dp, the geo- diodelaser(NewFocusVelocityTLB6712),focusedonto metric cavity length is δ, and the penetration depth into the microcavity using an objective lens (50× Mitutoyo the mirrors [42] is dp ≈(λ0/2)(nH−nL)−1. For the case Plan APO). Transmitted light was captured using a sec- shown in Figure 2, this yields a finesse of F = 3560, in ond objective lens (100× Mitutoyo Plan APO SL) and good agreement with the predicted value. focusedontoaphotodiode(ThorLabsDET36A).Atypi- The volume of the fundamental Gaussian mode of a cal wavelength sweep is shown in Figure 2, revealing res- Fabry-P´erot cavity can be approximated as [41] Vm = onances associated with the fundamental TEM00 mode (π/4)w02L, where w0 is the mode waist (radius). In the (at 777.80 nm) and three higher-order transverse modes. paraxialapproximationforahalf-symmetricFabry-P´erot A digital camera was used to verify the profiles of these cavity, the mode waist can be approximated as [41] Laguerre-Gaussian modes. The fundamental mode was (cid:114) λ fit to a Lorentzian, revealing linewidths of κ = 2π×2.7 w ≈ (R ×L)1/4, (1) GHz for 300 µm diameter cavities and κ=2π×6.7 GHz 0 π c where R is the radius of curvature of the upper mirror c and L (cid:28) R is assumed. R was determined by fitting c c a circular segment to the profile of the optical cavity, (a) 4 (b) as measured by optical profilometry (ZYGO MetroPro). The 100 µm buckled-dome cavity has a peak height of δ 2 = 2.67 µm, a total cavity length of L = 3.26 µm, and a μm] 0 radius of curvature of Rc = 210 ± 15 µm. Eq. 1 then [ produces a mode waist of w ≈ 2.55 µm, implying that 0 -2 V ≈ 35λ3. This matches well with COMSOL simula- m -40 tions that yield a mode volume of 35.7λ3. Similar values -20 0 20 40 have been reported for other visible-wavelength optical [μm] 20 40 -20 0 [μm] cavities. [33, 43–45] For a Fabry-P´erot device to be considered viable for cQED applications, it should allow for practical tuning FIG.1. (a)Microscopeimageofabuckled-domemicrocavity and stabilization of the resonance conditions. We have withbase diameterof100 µm. White-light interferencerings demonstratethehighdegreeofcylindricalsymmetryattained previously reported thermal tuning of the buckled-dome throughtheself-assemblyprocess. (b)Cross-sectionofafinite microcavities. [46] Here samples were attached to a cop- element model of a 100 µm diameter microcavity (horizontal perheatsinkinavacuumchamberat∼5mTorr,andthe and vertical axes different scales) showing alternating SiO2 temperature was regulated by a proportional-integral- (grey)andTa2O5 (white)layersoftheBraggreflectorsanda derivativecontroller. Transmissionspectroscopywasper- simulated optical mode. formed in one degree intervals, as seen in Figure 3. The 3 peak wavelength of the fundamental mode was tracked as a function of temperature, revealing a temperature dependence of ∆λ/∆T = 0.2346 ± 0.0007 nm/K. This tunability is vital because the stochastic nature of the buckling process produces cavities of varying resonance frequencies that must be drawn into resonance with the desired atomic transition. By heating the chip of micro- cavities, thermal expansion increases the cavity length sufficientlytoaligntheopticalresonancewiththedesired 87Rb transition at 308 K. The cavity wavelength could then,inprincipal,belockedtoanatomictransition. One drawback of the current method of tuning is the inabil- itytotuneindividualcavities,howeverintegratingheater electrodes[47] or electrostatic actuation[31] would allow for individually addressable on-chip microcavities. AnotherrequirementofacQEDdeviceisthatitshould have open-access for injection of atoms into the cavity. The buckling self-assembly process inherently produces closed cavities, but can be modified or supplemented to provideopen-access. Onestrategyistocouplethemicro- cavitiestohollowchannelsbypatterningnarrowstripsof fluorocarbon that buckle along with the domes and in- FIG.4. (a)Microscopeimageofabuckled-domemicrocavity withabasediameterof200µmintersectedbya60µmhollow tersect the microcavity. A representative, 60 µm wide, channel. Inset is a 3D cartoon illustrating how the channel channel is shown in Figure 4a. This channel has a peak intersects the dome, to provide open access to the interior of height of 2.7 µm, providing access for atomic gas injec- theopticalcavity. (b)Transmissionspectraofthefundamen- tion. Previous work has demonstrated that mirror spac- talopticalmodeofdomecavityshownin(a). (c)SEMimage ings as narrow as 110 nm were sufficient for atom injec- ofa30µmx5µmholecutthroughthetopmirrorofadome tion, [42] and recent work has demonstrated vapor cells cavity via FIB milling. The open cavity is visible through the hole, as well as the distinct layers of the Bragg reflector. Thecavityisoutlinedtoimprovevisibility. (d)Transmission spectra and(e)imageofthefundamentalmodeofthecavity shown in (c). with critical dimensions as narrow as 30 nm. [48] In ad- dition, since these hollow channels are formed by two Braggreflectors, theycanalsoactasopticalwaveguides, possiblyusefulforatomtrapping. [49]Asecondstrategy for open access is to use focused ion beam (FIB) milling to remove small portions of the top Bragg reflector. Fig- ure 4c shows an example of an access hole milled into a buckled-dome cavity. We found that the properties of the fundamental mi- crocavity resonance were retained in the case of inter- sectingwaveguides(seeFigure4b)likelybecausetheop- ticalmode-waistissmallcomparedtothedomediameter. Thus, modifications of the dome periphery has minimal impact on the central part of the dome where the fun- damental mode resides. However, we found that FIB milling resulted in microcavities with optical resonances that exhibit non-linear behavior (see Figure 4d), even at FIG. 3. (a) Fits of the experimental resonance peaks for the low input powers. This may be a result of the conduc- fundamental mode of a 100 µm diameter dome, at temper- tive carbon layer applied in order to avoid charging, or atures from 292 K (blue, 776.4 nm) to 310 K (red, 780.6 nm) in 1 K increments. (b) The variation of the fundamen- parasitic gallium implantation, during the FIB milling tal resonance wavelength with temperature. Orange circles process. are experimental data points, and the grey line is a linear fit Finally we consider some pertinent parameters for yielding a thermal tunability of ∆λ/∆T = 0.2346 ± 0.0007 cQED applications: the atom-cavity coupling rate g , 0 nm/K. and the cavity decay rate κ. For a cavity with resonance 4 √ g /κ ∝ D/ D2. That is, to first order the strong cou- 0 pling parameter is independent of the buckled-dome di- ameter. Experimentally we observe a slight increase in the strong coupling parameter with decreasing dome di- ameter, as shown in Figure 5a. This behavior is in con- trasttothetypicaltrendofanimprovingstrongcoupling parameter with increasing cavity length. [41] Theabilitytomaintainaconstantstrongcouplingpa- rameter while decreasing the mode volume is important for another key figure of merit of cQED, C , known as 1 the single-atom cooperativity, which we define following Law et al.[51] as g2 FIG. 5. (a) Strong coupling parameter, g /κ, as a function C = 0. (4) 0 1 κγ ofdomediameter. Fit(greydashedline)derivedusingcavity finesseandopticalprofilometrymeasurementsofdomeheight and radius of curvature. Deviations from the predicted zero- C is of particular interest for single-photon sources as 1 slope behavior are a result of the divergence from the elastic it defines the Purcell factor; the probability of a spon- buckling model, producing a non-linear dependence between taneously emitted photon entering the cavity mode is the radius of curvature and the dome diameter. [40] (b) Ex- given by 2C /(2C +1). It should be noted that both perimentalcooperativityasafunctionofdomediameter. Fit 1 1 κ and γ are defined as the half-width at half-maximum. as described in (a). Error bars, from the statistical uncer- Giventheparametersofthebuckled-domecavities,Eq.4 tainty in κ and measurement uncertainty in R , are smaller c produces C = 65 (Figure 5b). For comparison, cooper- than the symbol size. 1 ativities as high as C =290 and 51 have been reported 1 for fiber-based cavities[52] and macroscopic cavities, [53] frequencyω, andasingleatomlocatedatthemaximaof respectively. the cavity field, g is defined as In conclusion, we have developed on-chip Fabry-P´erot 0 microcavities specifically for cQED applications using (cid:115) 3λ2cγ alkali atoms, with tunable resonance frequencies and g = , (2) 0 4πV open-access for atom inclusion. Despite only moder- m ate atom-photon coupling (g /κ = 0.17), high single- 0 where γ is the half-width at half-maximum linewidth of atomcooperativity(C1 =65)providesstrongmotivation the excited state of the atom (γ = 2π × 3.0 MHz for for further investigation and optimization. The results 87Rb), andV istheopticalmodevolume. Tomaximize presented in this letter were obtained with ten-period m g ,themodevolumemustbeminimized. Amodevolume Braggreflectors, producingreflectionlimitedfinessecav- 0 of 35λ3 results in a coherent atom-cavity coupling rate ities with F = 3560. Increasing the number of periods of g =2π×1.12 GHz, greater than the most optimistic would increase reflectance, and macroscopic cavities us- 0 predictionsformacroscopicFabry-P´erotcavities,[42]and ing the same materials have reported finesses as high as amongthehighestreportedtodateintheliterature. [41] F = 480,000. [42] In order to achieve strong coupling Despitethehighatom-cavitycouplingrate,amorerel- (g0/κ ≥ 1) with our buckled-dome Fabry-P´erot cavities, evant figure of merit for cQED applications is the strong a finesse of F ≥ 21,000 would be required. This is mod- couplingparameterg /κ. [50]Thecavitiesdescribedhere est compared to finesse requirements predicted for other 0 exhibit g /κ ≤ 0.17 placing them in the weakly coupled optical cavity architectures. [29, 41, 54] Furthermore, 0 regime of cQED, which as outlined below could reason- improvement of the finesse to this level would have the ably be improved. Nonetheless, it is informative to con- benefit of increasing C1 as high as 420, making it among sider the dependence of this ratio on the diameter of the the highest cooperativity architectures available. buckled-dome cavity. From above it can be seen that We are grateful for the support of Alberta Innovates Technology Futures (Strategic Chair & iCiNano); the g0 ∝ √L , (3) University of Alberta Faculty of Science; Natural Sci- κ V ences and Engineering Research Council, Canada (RG- m PIN401918&EQPEG458523); theCanadaFoundation where V ∝ R1/2L3/2. Furthermore, it has been shown forInnovation; andtheAlfredP.SloanFoundation. LJL m c previously,[36] that based on elastic buckling mechan- acknowledgesthatthisresearchwasundertaken, inpart, ics L ∝ D, and to first order R ∝ D (where D thankstofundingfromtheCanadaResearchChairspro- c is the buckled-dome diameter). Thus, Eq. 3 predicts gram. 5 [1] H.J. Kimble, Nature 453, 1023 (2008). M.Uphoff,M.Mu¨cke,E.Figueroa,J.BochmannandG. [2] S.J. Van Enk, J.I. Cirac, P. Zoller, H.J. Kimble and H. Rempe, Nature 484, 195 (2012). Mabuchi, J. Mod. Opt. 44, 1727 (1997). [30] G.W. Biedermann, F.M. Benito, K.M. Fortier, D.L. [3] J.I.Cirac,P.Zoller,H.J.KimbleandH.Mabuchi,Phys. Stick,T.K.Loyd,P.D.D.Schwindt,C.Y.Nakakura,R.L. Rev. Lett. 78, 3221 (1997). JareckiJr.andM.G.Blain,Appl.Phys.Lett.97,181110 [4] L.M.Duan,M.D.Lukin,J.I.CiracandP.Zoller,Nature (2010). 414, 413 (2001). [31] C. Derntl, M. Schneider, J. Schalko, A. Bittner, J. [5] P. Ko´m´ar, E.M. Kessler, M. Bishof, L. Jiang, A.S. Schmiedmayer, U. Schmid and M. Trupke, Optics Ex- Sørensen, J. Ye and M.D. Lukin, Nat. Phys. 10, 582 press 22, 18 (2014). (2014). [32] M. Trupke, J. Goldwin, B. Darqui´e, G. Dutier, S. Eriks- [6] D.E. Chang, V. Vuleti´c and M.D. Lukin, Nat. Photon. son, J. Ashmore and E. A. Hinds, Phys. Rev. Lett., 99, 8, 685 (2014). 063601 (2007). [7] A. Reiserer, N. Kalb, G. Rempe and S. Ritter, Nature [33] L.Greuter,S.Starosielec,D.Najer,A.Ludwig,L.Duem- 508, 237 (2014). pelmann, D. Rohner and R.J. Warburton, Appl. Phys. [8] A.I. Lvovsky, B.C. Sanders and W. Tittel, Nat. Photon. Lett. 105, 121105 (2014). 3, 706 (2009). [34] L.C. Flatten, A.A.P. Trichet and J.M. Smith, Laser & [9] D.Petrosyan,G.Bensky,G.Kurizki,I.Mazets,J.Majer Photon. Rev., 17 (2015). and J. Schmiedmayer, Phys. Rev. A 79, 040304 (2009). [35] E.Epp,N.Ponnampalam,W.Newman,B.Drobot,J.N. [10] J.Bochmann,A.Vainsencher,D.D.AwschalomandA.N. McMullin,A.F.MeldrumandR.G.DeCorby,OpticsEx- Cleland, Nat. Phys. 9, 712 (2013). press 18, 24 (2010). [11] T.A. Palomaki, J.D. Teufel, R.W. Simmonds and K.W. [36] T.W. Allen, J. Silverstone, N. Ponnampalam, T. Olsen, Lehnert, Science 342, 710 (2013). A. Meldrum and R.G. DeCorby, Optics Express 19, 20 [12] T. Zhong, J.M. Kindem, E. Miyazono and A. Faraon, (2011). Nat. Commun. 6, 8206 (2015). [37] D.W.Hoffman,JournalofVacuumScienceandTechnol- [13] E.Saglamyurek,J.Jin,V.B.Verma,M.D.Shaw,F.Mar- ogy A 12, 953 (1994). sili,S.W.Nam,D.OblakandW.Tittel,Nat.Photon.9, [38] D.A. Steck, Rubidium 87 D Line Data (2003). 83 (2015). [39] E. Hecht, Optics (Addison-Wesley Publishing, Reading, [14] A. Faraon, P.E. Barclay, C. Santori, K.C. Fu and R.G. MA, 1987), pp. 373-375. Beausoleil, Nat. Photon. 5, 301 (2011). [40] M.H. Bitarafan, H. Ramp, T.W. Allen, C. Potts, X. Ro- [15] L. Li, T. Schro¨der, E.H. Chen, M. Walsh, I. Bayn, J. jas, A.J.R. MacDonald, J.P. Davis and R.G. DeCorby, Goldstein,O.Gaathon,M.E.Trusheim,M.Lu,J.Mower, JOSA B 32, 6 (2015). M. Cotlet, M.L. Markham, D.J. Twitchen and D. En- [41] D.Hunger,T.Steinmetz,Y.Colombe,C.Deutsch,T.W. glund, Nat. Commun. 6, 6173 (2014). Ha¨nsch and J. Reichel, New Journal of Physics 12, [16] G. Burkard and D.D. Awschalom, arXiv:1402.6351 065038 (2010). [17] D. Press, T.D. Ladd, B. Zhang and Y. Yamamoto, Na- [42] C.J. Hood, H.J. Kimble and J. Ye, Physical Review A ture 456, 218 (2008). 64, 033804 (2001). [18] K. De Greve, L. Yu, P.L. McMahon, J.S. Pelc, C.M. [43] K.J. Vahala, Nature 424, 839 (2003). Natarajan, N.Y. Kim, E. Abe, S. Maier, C. Schneider, [44] A. Muller, E.B. Flagg, J.R. Lawall and G.S. Solomon, M. Kamp, S. Ho¨fling, R.H. Hadfield, A. Forchel, M.M. Optics Lett. 35, 2293 (2010). Fejer and Y. Yamamoto, Nature 491, 421 (2012). [45] P.R.Dolan,G.M.Hughes,F.Grazioso,B.R.Pattonand [19] T.Wilk,S.C.Webster,A.KuhnandG.Rempe,Science J.M. Smith, Optics Lett. 35, 21 (2010). 317, 5837 (2007). [46] M.H. Bitarafan, H. Ramp, C. Potts, T.W. Allen, J.P. [20] L.M.Duan,A.KuzmichandH.J.Kimble,Phys.Rev.A Davis and R.G. DeCorby, Optics Lett. 40, 5375 (2015). 67, 032305 (2003). [47] M. Zalalutdinov, K.L. Aubin, R.B. Reichenbach, A.T. [21] H. Walther, B.T.H. Varcoe, B. Englert and T. Becker, Zehnder, B. Houston, J.M. Parpia and Harold G. Craig- Reports on Progress in Physics 69, 5 (2006). head, Appl. Phys. Lett. 83, 3815 (2003). [22] H.MabuchiandA.C.Doherty,Science298,5597(2002). [48] K.A.Whittaker,J.Keaveney,I.G.Hughes,A.Sargsyan, [23] H.P. Specht, C. No¨lleke, A. Reiserer, M. Uphoff, E. D. Sarkisyan and C.S. Adams, Phys. Rev. A 92 052706 Figueroa, S. Ritter and G. Rempe, Nature 473, 190 (2015). (2011). [49] A. Melnyk, C.A.Potts, T.W.Allen and R.G DeCorby, [24] C.No¨lleke,A.Neuzner,A.Reiserer,C.Hahn,G.Rempe submitted (2015). and S. Ritter, Phys. Rev. Lett. 110, 140403 (2013). [50] C.J.Hood,M.S.Chapman,T.W.LynnandH.J.Kimble, [25] N.Kalb,A.Reiserer,S.RitterandG.Rempe,Phys.Rev. Phys. Rev. Lett., 80, 4157 (1998). Lett. 114, 220501 (2015). [51] C.K.LawandH.J.Kimble.J.Mod.Opt.44,2067(1997). [26] J.McKeever,A.Boca,A.D.Boozer,R.Miller,J.R.Buck, [52] Y. Colombe, T. Steinmetz, G. Dubois, F. Linke, D. A. Kuzmich and H.J. Kimble, Science 303, 5666 (2004). Hunger and J. Reichel, Nature 450, 272 (2007). [27] M. Mu¨cke, J. Bochmann, C. Hahn, A. Neuzner, C. [53] J.A.Sauer,K.M.Fortier,M.S.Chang,C.D.Hamley,and No¨lleke,A.Reiserer,G.RempeandS.Ritter,Phys.Rev. M.S. Chapman, Phys. Rev. A, 69, 051804 (2004). A 87, 063805 (2013). [54] H. Mabuchi, J. Ye, and H.J. Kimble, Applied Physics B [28] A. Reiserer, S. Ritter and G. Rempe, Science 342, 6164 68, 6 (1999). (2013). [29] S. Ritter, C. No¨lleke, C. Hahn, A. Reiserer, A. Neuzner,