Atomically-thin quantum dots integrated with lithium niobate photonic chips

The electro-optic, acousto-optic and nonlinear properties of lithium niobate make it a highly versatile material platform for integrated quantum photonic circuits. A prerequisite for quantum technology applications is the ability to efficiently integrate single photon sources, and to guide the generated photons through ad-hoc circuits. Here we report the integration of quantum dots in monolayer WSe2 into a Ti in-diffused lithium niobate directional coupler. We investigate the coupling of individual quantum dots to the waveguide mode, their spatial overlap, and the overall efficiency of the hybrid-integrated photonic circuit.


Introduction
Solid-state quantum emitters have emerged in the past few years as prominent systems for quantum technology. Single-photon sources based on semiconductor quantum dots [1], defects [2] and impurities [3] have reached extremely high purity [4] and efficiency [1], making them competitive for future quantum applications such as quantum communication [5] or linear optical quantum computing [6]. Additionally, spin-active quantum emitters provide an excellent platform for quantum networking, enabling efficient interfacing between photons to share quantum states over long distances, and spins, which can store and process quantum information locally [7,8].
When implementing either purely optical or hybrid spin-photon schemes, a large number of optical components are required to construct the required photonic quantum gates. For increasing number of qubits, approaches based on bulk optics become quickly impractical, due to the size and the inherent instability of these individual components. Only photonic integration can provide the stability, repeatability and robustness required for high-visibility quantum interference and scalability [9]. For these reasons a large research effort in the past decade has been dedicated to the development of integrated quantum photonics circuits with embedded quantum emitters [10][11][12][13]. However, several challenges are still outstanding, such as total internal reflection in bulk high-index materials that severely limits optical collection efficiency and the in situ manipulation of the phase in interferometer arms. Furthermore, materials such as diamond, while featuring excellent spin properties, are notoriously hard to grow and fabricate, posing a serious challenge to the development of a large-scale integrated platform.
In this respect, quantum dots in two-dimensional layered materials such as transition metal dichalcogenide (TMDs) exhibit promising properties. Bulk TMDs form layered structures with weak interlayer van-der-Waals interactions which can be micro-mechanically exfoliated to obtain monolayer crystals [14]. One of the most prominent examples is WSe 2 which, at the monolayer level, is a direct band-gap semiconductor featuring bright localized excitons [15] or bi-excitons [16] at cryogenic temperatures. Once exfoliated, the TMD material can be stacked to create layer-by-layer van der Waals heterostructures with designer functionality [17,18] or placed onto arbitrary substrates [19], including complex photonic circuits, using a variety of transfer methods [20]. As quantum emission is embedded in the single atomic layer, optical extraction is not limited by total internal reflection, an enormous challenge for bulk materials such as GaAs, diamond or SiC. In addition, the monolayer can be transferred to any photonic structure fabricated in any material platform, removing any need for lattice matching or complex bonding procedures. Finally, the quantum emitter can be deterministically created at a pre-determined position by strain engineering [21][22][23].
An attractive photonic platform to integrate quantum emitters is lithium niobate. Lithium niobate is the industry standard for the photonic telecommunication industry, possessing the ideal properties for integrated photonics, including low-loss transmission in a broad wavelength range, from UV to mid-IR, and ultrafast switching capability. Crystalline lithium niobate uniquely combines nonlinear, piezoelectric, photorefractive and electro-optic effects in the same platform [24], making it an attractive system for quantum technology. Low-loss waveguides and other passive components, such as beam-splitters and ring resonators [25], have been demonstrated, both by Ti in-diffusion or ridge structures [26]. Its electro-optic properties have been exploited to build fast electrically-controlled phase modulators at gigahertz operating speeds [27,28], with recent work reporting 210 Gbit s −1 devices [29]. Integration with superconducting nanowire detectors, which offer single photon detection with the performance required for quantum photonics application, has already been demonstrated [30].
Here we demonstrate the coupling of quantum dots in monolayer WSe 2 to an integrated lithium niobate directional coupler, a fundamental building block for quantum interference experiments. Our work shows for the first time, that monolayers can be pre-selected, based on their properties, and transferred onto a photonic circuit with sufficient accuracy as to couple individual emitters to the photonic mode of the structure. While the titanium-diffused waveguides used in our experiment only allow a limited coupling efficiency, our work is a first step towards the development of a promising platform for quantum photonics. The integration of 2D materials hosting quantum emitters with lithium niobate photonic circuitry holds the promise to realise a fully-integrated architecture embedding low-loss passive components, ultrafast modulators [29] and electrically-tunable optical resonators.

Device design, fabrication, and experimental set-up
The directional coupler was fabricated by photolithographically patterning the waveguide structure (4 µm width) into a Ti thin film (70 nm thickness) on the surface of an X-cut LiNbO 3 wafer. The titanium was diffused into LiNbO 3 in a wet oxygen atmosphere at 1010°C for several hours. After the diffusion, the wafer was diced into a 2.5 cm long chip and the chip end faces were polished to optical grade. To enable stable fiber coupling inside the cryostat, a fiber array was aligned to the waveguides and glued to one facet of the chip using cryogenic compatible epoxy glue. The coupling was found to remain constant after several cryogenic cooling cycles. The waveguide propagation loss is estimated to be 0.5 dB/cm, while the coupling efficiency from fiber to waveguide is estimated to be 3.5 dB. This is confirmed by a total loss from output fiber to input facet of 4 dB, with the reflectivity of the directional coupler measured to be 70%. The facet was confocally imaged with a 760 nm laser placed into one of the fiber output ports to illuminate the input mode. The image (see Fig. 1(a)) shows an elliptical mode with major and minor axes of 10 µm and 6 µm, respectively (1/e 2 ). The numerical aperture (NA) of the waveguide is estimated to be 0.07.
Micromechanical exfoliation was used to obtain monolayer WSe 2 flakes, identified by optical microscopy. We select a monolayer featuring a wrinkle as it has been shown that the associated localized strain induces single photon emitters [21]. Quantum dots in monolayer WSe 2 exhibit an in-plane dipole orientation, so a flake positioned on the input facet of the waveguide maximizes the coupling efficiency. The flake was transferred onto the waveguide using the all-dry viscoelastic stamping procedure [31]. To illuminate the waveguide optical mode position, a HeNe (633nm) laser was sent from the opposite end of the directional coupler. This allowed deterministic transfer of flakes with wrinkles directly onto the optical mode. One example is shown in Fig. 1(b).
Photoluminescence (PL) measurements were performed in a confocal microscope setup with the sample mounted onto a 3-axis nanopositioner and cooled to 4K. The quantum dots were excited off-resonantly, close to saturation with 1 µW of continuous wave 532 nm laser light to maximize PL photon count rate. An objective of numerical aperture 0.82 is used to focus the beam to a diffraction limited spot of around 500nm, sufficiently small to selectively address individual emitters with limited unwanted excitation of nearby emitters. The experimental setup, shown in Fig. 1(c), enables PL to be collected confocally from the WSe 2 flake or from the two waveguide collection ports.

Results a
The confocal a flake transf region (solid w the bilayer re shown), while the areas of s of emission w strain induced To demon spatially excit 3(b)) and coll port which m directional co splitter around were located efficiency) fo waveguide wi and discussio image produce ferred onto the white line) clo gion we find t e the monolay trongly localiz wavelengths. T d by the wrinkl 2. (a) Confocal PL 0-735 nm, measu n while the solid li corresponding to guide mode (mar layer WSe 2 with r bars show the nor nstrate the coup te at the locali lect PL from th match the emis oupler acts as d the peak of th directly at th or this particu ith PL from cen on ed in Fig. 2 An estima collected by a matrix approa interface is co fiber coupling geometry. In niobate wave from the wav fiber ports. Th (Ports 1 & 2) mode center.
We confi modelling a d at the mode c which reduce Taking the ra confocally, w ( Fig. 3(a)), wh simulated col 3). The low 3. (a) Spatially-res sents the normaliz t Ports 1 and 2 a ured at Ports 1 an ved in confocal ge rate.
ate of the coup a waveguide w ach, around 10 ollected into an g losses, appro contrast, a fra guide with nu veguide into si herefore, the e and through t rm these valu dipole with var entre we expec es to 0.12% w atio of the cou we achieve a co hich agrees we lection efficien overall coupli solved PL spectra zed intensity. (c) and confocally wh nd 2 and confocal eometry is replicat ling efficiency with NA = 0.07 0% of emission n objective of n oximately 3% action of 0.7% merical apertu ingle-mode fib expected ratio b the confocal po ues through fi rying displacem ct a collection e when the dipol nts experimen ollection effici ell with FDTD ncies are displa ing efficiency ally filtered at 745 PL spectra meas hen the position o lly when Emitter ted at both output y can be obtain , in the ray opt n from a dipo numerical aper of the total e % of the dipol ure NA = 0.07 bers, we expec between the co ort is approxim inite-difference ment from the efficiency of 0 le is located 1 tally measured iency of 0.11 D and transfer m ayed in Table  of [25][26][27][28]32,33] which feature much larger NA (~0.6). In this case, for an emitter optimally positioned at the center of the optical mode, we estimate an overall efficiency of ~11%, representing a 100x improvement.

Conclusion
In conclusion, we have achieved the integration of quantum emitters in atomically thin WSe 2 onto a Ti in-diffused lithium niobate waveguide with a fiber coupled, on-chip beam-splitter. PL from the quantum dots was measured confocally at the waveguide facet and identical emission was collected through each output port of the device, demonstrating the operation of the on-chip Hanbury-Brown and Twiss interferometer. We expect the integration of twodimensional materials into lithium niobate elements will be a key platform in the development of integrated quantum photonic circuits. In particular, the high refractive index contrast of thin film LNOI waveguides can provide improved optical coupling between the quantum dots and the waveguide and local strain-inducing features (such as nanopillars) can be incorporated on the facet edge. The hybrid quantum emitter-LNOI platform can be extended with in situ strain tuning capability via the piezoelectric effect as well as access to passive and active optical elements, including grating couplers, compact integrated beamsplitters and ring resonators, and efficient reconfigurable interferometers.