The rapid evolution of photonic integrated circuit (PIC) technologies, as represented by the current flourish of silicon photonics [1], is a basis for modern information technologies. State-of-the-art PICs are being used for an expanding array of applications [2], such as photonics-based artificial neural networks [3] with laser light inputs. Introducing quantum light into PICs [4,5] will allow for advanced PIC-based optical information processing, such as linear optical quantum computation [6].

To that end, it is vital to use near-ideal quantum light sources [7]: e.g., it is required that single-photon sources (SPSs) provide single photons with near-unity efficiency, indistinguishability, and purity. Even after the long-term development of diverse SPS technologies [8–10], only a few solid-state materials, including InAs/GaAs semiconductor quantum dots (QDs) [11–15], have currently been proven to potentially fulfill these demanding requirements [7]. As such, the combination of such SPSs with existing high-end PIC platforms is an immediate route for the realization of large-scale quantum PICs.

In this context, the hybrid integration of SPSs into PICs is very promising. In previous demonstrations [16–18], photonic structures for the SPSs and the waveguides have been jointly processed on single wafers made, e.g., by conventional wafer bonding [16]. Such a joint-fabrication process of hybridized material platforms could hinder the highly optimized fabrication of each element: indeed, experimental waveguide coupling efficiencies of so-far demonstrated hybrid SPSs have been limited to around 10% to 40%. The necessary complicated process flows will also hamper the straightforward use of existing PIC platforms that are in general fabricated in specially customized facilities [such as complementally metal-oxide (CMOS) process foundries]. Very recently, the hybrid integration of QD SPSs on silicon waveguides by using a micro manipulator has been reported [19], succeeding in the independent preparation of the SPS and the waveguide for high-quality fabrication. However, the demonstrated waveguide coupling efficiency is still far less than the unity.

In this Letter, we propose and demonstrate an alternative approach based on transfer printing [20–24], by which SPSs and PICs can be prepared independently and integrated easily in a pick-and-place manner. The SPS is based on a QD in a nanocavity and is placed above a waveguide buried in glass, supporting near-perfect theoretical coupling of single-photon emission into the waveguide. Transfer printing largely simplifies the required three-dimensional integration of the optical elements, allowing for the demonstration of QD-based SPSs with high experimental coupling efficiencies as well as the dense integration of two SPSs into a waveguide.

Figure 1 shows the basic flow of the proposed hybrid integration process (see Supplement 1 for more details). First, we prepare a QD wafer [Fig. 1(a)] for the fabrication of nanocavity-based air-bridge SPSs [Fig. 1(b)]. We also use another wafer [Fig. 1(d)] to separately fabricate wire waveguides buried in glass cladding [Fig. 1(e)], which, if necessary, can be prepared by CMOS process foundries [25]. Then, we use transfer printing to pick up a suitable SPS from the processed QD substrate using a transparent rubber stamp [Fig. 1(c)]. We transfer the SPS onto the waveguide under an optical microscope: the SPS can be released on the waveguide by slowly peeling the stamp off [Fig. 1(f)]. A schematic of the completed waveguide-coupled SPS is shown in Fig. 1(g). This structure enables near-unity coupling of QD emission into the waveguide, as discussed later. By pre-selecting suitable SPS candidates prior to the transfer (with appropriate emitter linewidths, positions, and wavelengths), as well as by co-integrating post-tuning functionalities (such as heaters and stressors), the transfer-printing approach may solve major difficulties for incorporating multiple solid-state SPSs, which are often inherently random in their spatial locations and emission wavelengths [7,26].

 

Fig. 1. Process begins with the preparation of SPSs on (a) a QD wafer, by patterning (b) an array of PhC nanobeam cavities. (c) An airbridge SPS is picked up by attaching and quickly peeling off the transparent rubber stamp. Meanwhile, we use (d) another wafer for preparing (e) glass-cladded wire waveguides. (f) The picked-up SPS is transferred by placing it above the waveguide and slowly releasing the stamp. (g) Final structure.

Download Full Size | PDF

Our design of the SPS structure allows near-unity coupling of QD radiation into the waveguide. Figures 2(a) and 2(b), respectively, show top-view and cross-sectional schematics of the waveguide-coupled SPS. The nanocavity is based on a one-dimensional photonic crystal (PhC) with local lattice deformation [27–29]. Here we utilize the fundamental cavity mode, which resonates at around 900 nm and possesses a high $Q$-factor of $5.4\times {10}^6$ and a small mode volume of $V=8.6\times {10}^{-3}{\mathrm{\mu m}}^{-3}$ when located solely on a glass clad without the waveguide. When introducing the waveguide underneath [30], the cavity $Q$-factor exhibits an exponential dependence on the cavity–waveguide distance, $d$, as shown in Fig. 2(c). The sharp reduction in $Q$ stems predominantly from the coupling of light into the waveguide, characterized with a cavity–waveguide coupling efficiency $\eta$ of over 99% for $250\mathrm{nm}<d<450\mathrm{nm}$. Further reduction of $d$ leads to a degradation of $\eta$, as the waveguide becomes too close to the cavity and starts to scatter cavity photons into free space. In the design presented here, we tuned the cavity/waveguide parameters to minimize the free space scattering for a given $d$ (see Supplement 1).

 

Fig. 2. (a) Top view of the designed PhC nanobeam cavity coupled to the waveguide underneath, displayed together with the electric field distribution ($E_y$ component) of the fundamental cavity mode. (b) Schematic cross section of the waveguide-coupled SPS. The underlying waveguide is also made of GaAs. (c) Simulated cavity $Q$-factors (black), $\eta s$ (red), $\beta s$ (blue), and total single-photon coupling efficiencies $\eta \beta s$ (green), plotted as a function of $d$.

Download Full Size | PDF

The emitter–cavity coupling efficiency, $\beta$, reduces when decreasing cavity $Q$ due to the reduction of the Purcell effect, which scales with $Q/V$. Nevertheless, even when $d=350\mathrm{nm}$ ($Q=3800$), the maximum possible $\beta$ is as large as 99.9%, thanks to the very small $V$. Overall, the theoretical maximum single photon coupling efficiency from the QD into the waveguide, $\eta \beta$, is deduced to be a near-unity value of 99.7% for $d=350\mathrm{nm}$ (see Supplement 1). We note that such near-unity $\eta \beta$ can be obtained even in different material combinations (see Supplement 1). It is also noteworthy that the SPS design can maintain high efficiencies even under the presence of the misalignment between the cavity and waveguide (see Supplement 1), which is inherent to the transfer printing approach.

Figure 3(a) shows an optical microscope image of a fabricated sample. For the waveguide, we fabricated a glass-clad GaAs structure using transfer printing and spin-on-glass coating (see Supplement 1). From the microscope image, a good alignment between the top nanobeam cavity and the underlying waveguide can be seen. Indeed, there is less than 100 nm position deviation between the nanobeam and the waveguide in the $y$ direction (see Supplement 1). This high alignment accuracy was reproducibly obtained in our manual printing process relying on a high-magnification optical microscope. The waveguide is terminated by two exit ports, which are composed of diffraction gratings to direct the single photons into free space [31]. In order to optically characterize this structure, we performed low-temperature photoluminescence (PL) measurements (see Supplement 1). As an initial experiment, we focused a pump laser beam onto the center of the nanocavity and measured a sample image, as displayed in Fig. 3(b). We observed bright light out-coupling from the exit ports, indicating efficient waveguiding of the cavity mode emission.

 

Fig. 3. (a) Visible microscope image of a completed device. (b) Low-temperature (7 K) PL image of the sample. (c) PL spectra taken by pumping the cavity center with an 815-nm continuous wave laser with a power of 20 nW. The red curve is the spectrum measured through the exit port. An offset is added to the curve. The green curve is measured above the cavity by spatially selecting the radiation near the cavity center.

Download Full Size | PDF

Then, with low optical pumping to the cavity center, we measured emission spectra of the radiation from one of the exit ports, as shown in Fig. 3(c). In the upper red curve, an intense cavity mode emission at 902.5 nm, together with cavity-coupled QD emission is clearly seen. The measured cavity $Q$ was 3600, which is 3.6 times smaller than those measured for the nanocavities placed on flat glass. This significant reduction of the $Q$-factor suggests a large experimental cavity–waveguide coupling efficiency (${\eta }_{\exp }$) of 72% [30]. The estimation is based on the assumption that the $Q$-factor reduction here occurs solely by the waveguide coupling (see Supplement 1 for the validity of the assumption and the details of the ${\eta }_{\exp }$ estimation). The lower, green, spectrum in Fig. 3(c) is of the emission measured above the cavity center. This spectrum does not show the cavity peak, implying that the leakage into free space is largely suppressed, and that the emission occurs predominantly into the waveguide. Here, the largest perturbation to the cavity is provided by the waveguide, which drains the predominant cavity photon leakage and in turn suppresses the radiation into free space. We note that a more direct measurement of ${\eta }_{\exp }$ will be possible by measuring the transmission spectra of the waveguide–cavity system [30].

We performed further detailed optical characterization of the fabricated sample. Figure 4(a) shows a color plot of the temperature-dependent emission spectra measured through the exit port. We observed an enhancement of the QD emission near the cavity resonance [13,14], suggesting that the increase of $\beta$ is due to the Purcell effect. We also performed time-resolved PL measurements under the QD-cavity resonance, as shown in Fig. 4(b). A rapid decay of the emission with a rate of $3.9{\mathrm{ns}}^{-1}$ is observed, which is 3.9 times faster than that of unprocessed QDs (as shown by the gray curve). The estimated emitter–cavity coupling efficiency from these measurements (${\beta }_{\exp }$) is 87% (see Supplement 1). Given ${\eta }_{\exp }=72\%$, the total single-photon coupling efficiency, ${\eta }_{\exp }{\beta }_{\exp }$, is estimated to be 63%. We primarily attribute this non-ideal ${\eta }_{\exp }{\beta }_{\exp }$ to the fabrication imperfection of the nanocavity (degrading ${\eta }_{\exp }$) and the non-maximum Purcell effect on the QD presumably due to its position deviation from the field maximum (degrading ${\beta }_{\exp }$). We estimated that, by increasing the intrinsic cavity $Q$ (without the waveguide coupling) to 50,000, the maximum possible $\eta \beta$ will reach 98.4% (see Supplement 1).

 

Fig. 4. (a) Color plot of temperature-dependent emission spectra measured through the exit port. (b) Time-resolved PL spectra measured by time-correlated single-photon counting. The red curve was taken at 46 K corresponding to the condition that the QD peak is resonant with the cavity mode. The black curve shows the fitting result, and the gray curve is measured for a typical QD embedded in an unprocessed area of the sample. (c) Measured second-order coherence function.

Download Full Size | PDF

Next we performed second-order correlation measurements based on a Hanbury Brown–Twiss setup when the QD is slightly detuned from the cavity resonance by 0.43 nm. Figure 4(c) shows a measured intensity correlation histogram, exhibiting a clear anti-bunching with a zero delay time value of the second-order coherence function of $g^{(2)}[0]=0.23$, demonstrating single-photon generation from the QD. We consider that the relatively large $g^{(2)}[0]$ value originates primarily from the intrusion of the background cavity emission supplied by other off-resonant QDs embedded in the cavity and possibly from multiple pump-and-emission processes in the QD within an excitation pulse.

Finally, we extended our method to integrate multiple SPSs into a waveguide, which will be required for the realization of large-scale quantum PICs. For this demonstration, we used the same waveguide platform and were able to integrate two different SPSs by repeating the transfer printing process. A microscope image of the completed sample is shown in Fig. 5(a). The nanocavities are designed to resonate at different wavelengths, such that any disturbance on the transport of single photons by the other cavity is largely suppressed (see Supplement 1). We characterized the optical performance of the two SPSs using PL measurements through the waveguide exit ports, as shown in Figs. 5(b)–5(g). For both SPSs, strong QD emission peaks are observed [Figs. 5(b) and 5(c)]. These peaks show fast radiative decay rates, as can be seen in time-resolved PL spectra for each emission line [Figs. 5(d) and 5(e)], indicating their Purcell-enhanced emission into the cavity modes. From these results, we deduced high ${\eta }_{\exp }{\beta }_{\exp }$ efficiencies of 71% and 54% for the left and right SPSs, respectively. In addition, we confirmed that the QD emission peaks exhibit strong anti-bunching, as shown in Figs. 5(f) and 5(g), suggesting the successful integration of two Purcell-enhanced, efficient SPSs into the individual waveguide.

 

Fig. 5. (a) Microscope image of the completed sample. Two nanobeam cavities (cavity $L$ and $R$) are integrated onto the single waveguide. (b)–(g) Emission properties of the QDs embedded in the (b), (d), (f) left and (c), (e), (g) right cavities, respectively. (b) and (c) PL spectra measured through the exit port at 3 K. The labels, QD-A and QD-B, indicate the investigated QD emission peaks. The black solid lines indicate the cavity peak positions. (d) and (e) Time-resolved PL spectra. The black solid lines show fitting results. (f) and (g) Measured second-order coherence functions.

Download Full Size | PDF

In summary, we have demonstrated the transfer-printing-based integration of QD SPSs into wire waveguides. We have shown that our strategy allows near-unity total single-photon coupling efficiencies into the waveguide, while largely relaxing the difficulties in the hybrid integration of SPSs into PICs. Experimentally, we demonstrated high ${\eta }_{\exp }{\beta }_{\exp }$ efficiencies of up to $\sim 70\%$, which we expect will be further improved by elaborating the nanofabrication. We believe that our approach will be a key technology for the fusions between the state-of-the-art SPSs and modern PICs, irrespective of material choice.