On-Chip Integration of Single Solid-State Quantum Emitters with a SiO$_2$ Photonic Platform

One important building block for future integrated nanophotonic devices is the scalable on-chip interfacing of single photon emitters and quantum memories with single optical modes. Here we present the deterministic integration of a single solid-state qubit, the nitrogen-vacancy (NV) center, with a photonic platform consisting exclusively of SiO$_2$ grown thermally on a Si substrate. The platform stands out by its ultra-low fluorescence and the ability to produce various passive structures such as high-Q microresonators and mode-size converters. By numerical analysis an optimal structure for the efficient coupling of a dipole emitter to the guided mode could be determined. Experimentally, the integration of a preselected NV emitter was performed with an atomic force microscope and the on-chip excitation of the quantum emitter as well as the coupling of single photons to the guided mode of the integrated structure could be demonstrated. Our approach shows the potential of this platform as a robust nanoscale interface of on-chip photonic structures with solid-state qubits.


I. INTRODUCTION
Single solid-state quantum emi ers such as color centers in diamond are promising potential building blocks for future quantum information processing architectures and integrated nanophotonic devices [1]. e most prominent representative is the nitrogen-vacancy (NV) color center which exhibits coherent optical transitions and long-lived nuclear and electron spins, making it a promising solid-state qubit [2][3][4] and single photon source [5], but the ongoing search for new solid-state quantum emi ers recently revealed a variety of new emi ers, e.g. new optical vacancy-impurity defects in diamond [6,7] and hexagonal boron nitride [8].
e most fundamental way of interaction with single emitters is free-space optics, but in order to build scalable quantum architectures, deterministic and e cient on-chip integration of one or more quantum emi ers in combination with passive optical structures such as cavities and couplers is required [9,10]. One fundamental requirement is the ecient collection and routing of single photons. e commonly used platform silicon cannot be used in most cases since its opaque up to about 1.1 µm and a large part of the interesting solid-state emi ers emit at visible wavelengths erefore, in recent years great e orts have been made to develop new platforms for integration of quantum emi ers and to demonstrate the assembly of hybrid nanophotonic systems, for example, based on tapered optical bers [11][12][13][14][15], direct laser wri en [16,17], diamond [18][19][20], other dielectric [21][22][23][24][25] or plasmonic structures [26][27][28].
Each of these platforms poses its own challenges and depending on the selected emi er and speci c application, some are be er suited than others. One common problem is unwanted background uorescence from the passive optical structure, as some emi ers need to be pumped strongly with green or even blue light. Another crucial aspect is low absorption in the material to minimize losses and enable the fabrication of high-Q resonators. A very well-suited material platform is SiO 2 which is thermally grown on silicon, as it offers a broad transparency window from the UV to mid-IR and therefore low uorescence and losses. However, the index of refraction of SiO 2 is rather low (∼ 1.5), so in order to allow for a propagating mode within the SiO 2 , the underlying silicon must to be removed.
In this work we use a free-standing, monolithic SiO 2 photonics platform, where rib-waveguides allow e cient guiding of visible light in thermally grown and undoped SiO 2 similar to the system introduced by Chen et al. [29]. So far, we have been able to show that this system has a very low background uorescence, which is well suited for the integration of single photon emi ers, and in addition high-Q microresonators can be realized [30]. By the numerical analysis of the waveguide structures we present a design that allows an optimal coupling of the guided mode in the waveguide to an external quantum emi er. We also present a 2D tapered section of the waveguide as a mode-size converter to improve the mode overlap with lensed single-mode bers. A er characterizing the fabricated device, we show the functionalization of this nanophotonic platform with a preselected single NV color center hosted within a nanometer-sized diamond [31], and by detecting the single-photon emission from the integrated waveguide, we were able to demonstrate the coupling between the guided fundamental mode of the SiO 2 waveguide and the single quantum emi er. e integrated platform presented here has the advantage of the ultra-low intrinsic uorescence when compared to other material such as Si 3 N 4 or doped optical bers. Its wide transparency window (∼ 0.2 µm − 3.0 µm) is ideal to integrate other solid-state or condensed-phase emi ers, e.g. in a hybrid integrated platform. Finally, fabrication is simple and other passive structures such as high-Q microring resonators [30] or detectors [32] can be integrated.  Figure 1. Waveguide design and functionalization. (a) Illustration of the SiO 2 waveguide structure and the eld pro le (| | 2 ) of the guided TM fundamental optical mode at 700 nm. Also the deterministic positioning process of the diamond-nanocrystal containing a single NV center (the NV crystal structure is shown in the inset) into the inner edge of the integrated SiO 2 rib waveguide with an atomic force microscope (AFM) tip, is shown. (b) Schematic of the assembled device with the single quantum emi er at its desired location within the inner edge of the waveguide, evanescently coupled to the guided mode.
e schematic also points out the underetched freestanding rib waveguide, which allows mode guiding in pure SiO 2 and one possible excitation/detection scheme. (c) Scanning electron microscope image of the fabricated structure. e waveguide is recessed to prevent damage to the facet by dicing during fabrication of the device and also helps prevent damaging the freestanding rib waveguide during the experiments.

A. Waveguide design and functionalization
To minimize background uorescence, the integrated photonic platform was designed to guide the optical mode exclusively within undoped, thermally grown SiO 2 which exhibits ultra-low intrinsic uorescence, even when strongly pumped with a 532 nm laser [30]. To enable this air-clad waveguide, a supporting membrane is required, resulting in the rib-waveguide structure schematically shown in Fig. 1(a). ese waveguides usually support at least two fundamental modes with orthogonal electric elds, one almost purely transversal electric (quasi-TE, herea er referred to as TE) and one mostly transversal magnetic (quasi-TM, herea er referred to as TM). e cross section shows the eld pro le (| | 2 ) of the guided TM fundamental optical mode, which visualizes the evanescent tail of the mode within the inner edge of the rib waveguide. Here we also sketch the positioning of a nanodiamond of some ten nanometers in size hosting a single NV center to the designated position within the inner edge of the rib waveguide. As it is evident from the eld pro le, this is the best accessible spot to place an optical emi er in the evanescent eld of the guided mode. In order to have a signi cant part of the evanescent eld accessible the dimensions of the waveguide should be in the range or smaller than the wavelength of the guided light ( ≈ 700 nm).
A nanoparticle hosting a single NV center can be precharacterized and transferred to the integrated structure by pick-and-place manipulation using a commercial atomic force microscope (AFM; JPK Instruments) as elucidated in [31]. Since the Silicon substrate is opaque to visible light, the positioning and veri cation of placement of the NV center could not be done in-situ, but rather the diamond was placed some micrometers next to the waveguide. A er optically verifying the successful placement of the nanodiamond hosting a single NV center by performing a confocal scan, the nanoparticle is pushed to the inner edge of the rib wave-guide in a subsequent step, using a specially shaped tip (Ad-vancedTEC™NC, NANOSENSORS™).
In Fig. 1(b) the assembled device is illustrated in more detail, also showing the silicon substrate that is removed underneath the waveguide during the fabrication process, which will be presented elsewhere in detail [33]. In the gure we also present a possible experimental con guration in which the deposited nanodiamond emi er is excited by free-space pumping through a microscope objective and the emi ed single photons are evanescently coupled to the guided modes of the waveguide, which will be discussed in more detail in the next section. Fig. 1(c) shows a scanning electron microscope image of the integrated free-standing rib waveguide structure. e waveguide is slightly recessed to prevent damage to the facet during the fabrication process and also proves to be advantageous during the experiments, as it provides some protection for the waveguide's facet.

B. Guided modes and coupling e ciency
To conveniently couple the light in and out of the waveguide chip, we chose silica core single mode lensed bers (S630-HP) with a spot diameter of (2.0 ± 0.5) µm (OZ Optics). Due to the mismatch between the ber's spot size and the dimensions of the mode in the emi er coupling region of the waveguide (sub-micron) we designed a mode-size converter. e inverse tapers usually employed in strip waveguides for mode conversion [34] are not feasible for our rib waveguide structure due to the requirement of the supporting membrane, therefore a two-dimensional (2D) section of the waveguide is tapered both laterally and vertically, as illustrated in Fig. 2(a). is tapered section acts as an adiabatic mode size converter and increases the on-and o -chip coupling eciency with the single mode bers. e fundamental TE and TM eigenmode pro les at a wavelength of = 700 nm before and a er the 2D tapered section are shown in the lower part of Fig. 2(a). e eigenmode solutions are obtained from To experimentally estimate the on-chip coupling eciency, we used a laser with an emission wavelength of 637 nm, which was coupled into the waveguide from one side with the lensed ber and collected the guided light on the other side of the waveguide chip with an objective lens. e lensed ber was mounted with a V-groove ber holder, coarse and ne positioning was performed with a XYZ linear translation stage ( orlabs) and a XYZ piezo positioning stage (PiezoJena) respectively. A er careful optimization of the bers position with respect to the waveguide, the transmi ed light is recorded on a photodetector and its power is compared with the laser power measured directly behind the lensed ber. A ber beam spli er is used to generate a power reference signal in order to take into account uctuations in laser power within the ber. In addition, transmi ed light polarization can be adjusted with an inline ber polarization controller to be mainly TE or TM polarized. e ratio of transmi ed light to incident light can then be used as lower bound for on-chip coupling e ciency, assuming zero losses within the waveguide and tapered sections. Typical overalltransmission values obtained from waveguides with 2D tapers and dimensions comparable to those in Fig. 2(a) were approximately 35 % both for TE and TM incident polarization.
e interaction e ciency of the emi er with the guided modes in the waveguide can be described by the factor, de ned by the fraction of the total emi ed energy which is coupled to the guided mode = wg tot , where wg denotes the emi ers decay rate into the guided modes of the waveguide and the emi er's total decay rate. e factor can be obtained from performing a full 3D simulation of the waveguide structure with a radiating dipole emi er where the output into the guided mode is monitored. To facilitate the problem, we exploit the possibility of calculating the Purcell factor = wg 0 for a point-like dipolar emi er from a simple 2D simulation of the guided modes as described by Barthez et al.  where 0 denotes the undisturbed decay rate of the emi er (in vacuum),̂ is a normalized vector pointing in direction of propagation along the waveguide which is normal to the surface , 0 is the absolute value of the photon momentum in air and u denotes the electric eld components parallel to the dipole orientation of an emi er. e layout for the simulation can be found in the center of Fig. 2(b) where the dipole emi er was placed either vertical or horizontal in respect to the waveguide structure = 10 nm from each side of the inner edge. In order to calculate from a 3D calculation of the same structure is carried out, from which the total emi ed power of the radiating dipolar point source si ing next to the waveguide is obtained as well as the total emi ed power in vacuum. By combining the results of the 2D simulation which gives us and the 3D simulation from which we obtain 0 / tot we can then calculate as: e dimensions where is optimal are found by a parameter scan of the height ℎ and width of the waveguide. In the scan the thickness of the membrane was set constant to a value of = 400 nm where a robust fabrication and handling of the integrated structures was still ensured and the wavelength was set to = 700 nm as this is approximately the peak of the NV centers emission. e factor for the fundamental TE (TM) guided mode coupling to a horizontal (vertical) dipole emi er = 10 nm from the inner edges can be found in Fig. 2(b) le (right). As TE modes become no longer strongly guided and therefore very lossy for smaller dimensions of the waveguide was set to zero here and the area is marked shaded. e nal device was designed to support both the fundamental TE and TM mode due to the random orientation of the NV centers dipole axis, so we chose a width = 500 nm, height ℎ = 800 nm and membrane thicknesses = 400 nm, marked with red circles in Fig. 2(b). ese dimensions were chosen as a trade-o between coupling e ciency and fabrication limitations. e simulated factor for a device with these dimensions is TE = 5.7 % for a horizontally oriented dipole coupling to the TE mode and TM = 3.8 % for a vertical oriented dipole coupling to the TM mode in the waveguide.

III. RESULTS OF THE ASSEMBLED DEVICE
For functionalization, samples with multiple straight waveguides and dimensions as derived in the last section ( = 400 nm, ℎ = 800 nm, = 500 nm in the coupling area) were fabricated and characterized. A waveguide was then selected and functionalized with a preselected single NV center hosted in a nanometer sized diamond, as previously described in detail. A er AFM manipulation, the presence of the NV center is veri ed in an all-confocal con guration where the the continuous wave 532 nm pump laser is focused onto the sample using a NA = 0.9 objective lens (Olympus, MPLAPON60X). e re ected light is spectrally ltered by two 620 nm longpass lters (Omega optics) and spatially by a pinhole. e remaining uorescence is directed either onto two avalanche photo diodes (APDs; Perkin-Elmer) in a Hanbury Brown and Twiss (HBT) con guration to verify the single photon generation, a CCD camera or a spectrometer. In the confocal scan, the NV center can be identi ed as a bright uorescent spot near the waveguide (see upper part of Fig. 3(a)). e lower part of Fig. 3(a) shows the resulting second order correlation, measured with the HBT setup. Fi ing the autocorrelation data to a three-level model: where denotes the bunching amplitude, 1 the antibunching time and 2 the bunching time, we obtained a value of (2) (0) = (0.19 ± 0.01) <0.5 for this confocal excitation and detection scheme. e bright spot approximately 1 µm le of the NV center does not originate from NV uorescence as determined from the photon statistics, it most likely is some broken o part of the AFM tip used to position the NV center. Fortunately, this sca erer does not have a negative in uence on the performance of the device.
To verify the coupling of the waveguide mode to the NV center, we couple the green pump laser to one input port of the waveguide via a lensed ber and use the confocal microscope only to detect the resulting uorescence. By scanning the confocal detection over the sample, the uorescence map in Fig. 3(b) is recorded, which shows that no background uorescence of the waveguide can be detected even with strong pumping ( ≈ 20 mW a er the lensed ber). e autocorrelation data originating from the bright spot reveals a (2) (0) = (0.18 ± 0.02) <0.5, which is comparable to the (2) (0) obtained during confocal excitation of the NV center, verifying that the NV center can also e ciently be excited via the waveguide without increasing background uorescence, thanks to the excellent optical properties of the SiO 2 .
In a third con guration, the NV center is optically pumped via the microscope objective and the NV uorescence directly coupled to the waveguide is detected by o -chip coupling the light from both waveguide ends to single-mode bers. e remaining pump light is ltered out by a single 620 nm longpass lter at the end of each ber before being detected by two APDs. Fig. 3(c) shows the uorescence map detected when scanning the laser over the waveguide and NV. In this con guration the waveguide itself acts as an intrinsic beamspli er. Here the autocorrelation function recorded between the two output arms shows an anti-bunching with (2) (0) = (0.23 ± 0.03) <0.5 clearly indicating the mainly single-photon characteristics of the guided light in the waveguide.
As the NV centers electronic spin state can easily be initialized and read out optically and controlled coherently via a microwave eld, the NV defect can be used e.g. as a spin based magnetic eld sensor or solid-state qubit. To demonstrate the ability to control the waveguide-integrated NV centers electron spin in our assembled device we applied an external microwave (MW) eld by a nearby wire antenna and detected the NVs uorescence change when scanning the MW frequency. In Fig. 4   Electron spin resonance. Continuous wave optically detected magnetic resonance (ODMR) signal of the waveguideintegrated nitrogen-vacancy centers = 0 ↔ ±1 transition under zero external magnetic eld. e solid line is a Lorentzian t to the data revealing a transition frequency of = 2.870 GHz. = 2.870 GHz clearly shows that the integrated NV centers electron spin can be manipulated and read out.

IV. CONCLUSION AND OUTLOOK
In summary, we present on-chip SiO 2 photonic structures with ultra-low uorescence at visible wavelengths that are very well suited for the integration of solid state singlephoton emi ers that require relatively high excitation powers, such as the NV center. e integrated structure allows e ective routing of excitation laser sources and single photons via a freestanding rib-waveguide con guration. e onand o -chip power coupling e ciency to single mode bers can be increased by introducing 2D tapered sections of the waveguide to adiabatically transform the size of the guided mode. By deterministically positioning a nanodiamond with a single NV center in close proximity to the waveguide, we were able to demonstrate single photon generation by either pumping the NV center over the waveguide and detection in free-space, or by pumping over the microscope objective and detection of the photons coupled to the single guided mode of the waveguide. We could also demonstrate the control of the integrated NV centers electron spin by simply combining the integrated system with a microwave antenna. e device and functionalization presented here is not limited to NV centers in diamond, but can also be transfered to other solid-state quantum emi ers in the visible, such as other defect centers in diamond or defects in 2D materials. Furthermore, the integrated device can be extended with other on-chip photonic structures such as high-Q ring resonators [30], directional couplers, on-chip detectors [32], and microwave antennas which allows for the monolithic realization of complex devices consisting of various functionalities, e.g. the optical microintegration of those chips with high functional density together with light sources, detectors, and electronics.
In this way there is no limitation to build up a complex structure with several integrated functionalities. Even the packaging of the chip and pump laser diodes within an optical module can be envisioned. is would be highly a ractive for compact modules for integrated quantum technologies.

V. ACKNOWLEDGMENTS
We thank Günter Kewes and Bernd Sontheimer for useful discussions.
is work was supported by the European Fund for Regional Development of the European Union in the framework of project iMiLQ, administrated by the Investitionsbank Berlin within the Program to Promote Research, Innovation, and Technologies (ProFIT) under grant 10159465.