A fiber-integrated nanobeam single photon source emitting at telecom wavelengths

Single photons are ideal carriers of quantum information because they can propagate over long distances in optical fibers with extremely low loss.¹ But applications such as quantum communication² and photonic quantum computing³ require high coupling efficiency to the optical fiber mode in order to ensure that the quantum signal faithfully transmits to the receiver or the detector. Most single photon sources emit into free space,^4–19 and coupling these sources to fibers necessitates bulky optics that requires extremely precise optical alignment. Additionally, this coupling approach is lossy due to imperfect mode-matching. Single photon sources equipped with an efficient coupling scheme into optical fibers could alleviate these problems and would thus be easier to use and integrate with other photonic devices, such as modulators and phase shifters.

Several previous studies have reported fiber-coupled single photon sources using either semiconductor quantum dots^20–29 or nitrogen-vacancy centers in diamond.^30–34 The majority of these studies control the position of a fiber taper to come into contact with a photonic waveguide on a chip. However, these systems require constant re-alignment and are sensitive to vibrations and temperature fluctuations because the fiber taper and the photonic waveguide move independently. Other studies have directly attached the single photon emitters to a fiber taper or a cleaved facet of the fiber.^26–33 However, some of them do not have engineered structures that enable effective mode-matching of the single photons into the fiber mode. Besides, these sources emit at wavelengths that are outside the telecom bandwidth where fibers exhibit minimal propagation losses.

In this work, we realize an alignment-free fiber-coupled single photon source at telecom wavelengths. We employed quantum dots emitting near 1300 nm as single photon sources. To guide the single photon emission into a well-defined mode, we fabricated a single mode nanobeam with a photonic crystal mirror around the quantum dots. Using a tungsten probe installed in a focused ion beam system, we transferred the photonic crystal nanobeam to a tapered optical fiber. The fabricated device achieved a brightness of 1.4% and a single photon purity of 83%. Since we integrate the single photon source into the fiber taper, we can perform pumping, collection, and detection within the optical fiber system, which alleviates the need for complex optical alignment and minimizes the fluctuation of coupling efficiency from mechanical vibrations.

Figures 1(a) and 1(b) depict a schematic of the fiber-integrated single photon source. The structure consists of a photonic crystal nanobeam, which contains the InAs/InP quantum dots in the middle, attached on top of a tapered fiber. A photonic crystal mirror composed of an array of etched holes on one end of the beam directs the quantum dot emission in one direction. We smoothly taper one end of the nanobeam to adiabatically transfer the nanobeam-guided single photons to the underlying fiber taper. The gradual fiber taper then transforms the optical mode of the photon into the mode of the bare optical fiber.

An important figure of merit for the device's structure is the brightness, defined as the ratio of the number of single photons collected into the fiber to the number of pump pulses. The brightness is given by the equation B = qβη, where q is the quantum efficiency of the quantum dot, β is the single mode coupling efficiency to the nanobeam mode, and η is the coupling efficiency to the fiber mode. We optimized the brightness of the designed structure using finite-difference time-domain simulation (supplementary material Note 1). The single photon source has several design parameters: the lattice constant a, hole radius r of the photonic crystal mirror, nanobeam thickness t, width b, taper length L_taper, fiber taper tip diameter D, and taper angle θ_taper. Figure 1(c) shows the brightness as a function of two parameters: the width of the nanobeam (b) and the length of the taper (L_taper) with all other parameters fixed (supplementary material Note 2 provides the values used for the remaining parameters). Since the quantum efficiency q is usually close to unity for epitaxially grown quantum dots,³⁵ we assumed q to be unity. The brightness of the device reaches up to 90% for L_taper ≥ 15 μm and b ∼ 380 nm (supplementary material Note 3 explains why the brightness dips at b = 340 nm and b = 460 nm). The brightness of the single photons increases as L_taper gets longer because the smooth taper of the nanobeam improves the adiabatic mode transfer. From the optimization, we selected b = 400 nm and L_taper = 15 μm. Other design parameters for the fabrication include a lattice constant of a = 350 nm, a hole radius of r = 98 nm, a thickness of t = 280 nm, a fiber taper diameter of D = 500 nm, and a fiber taper angle of 1° (supplementary material Note 2). Figure 1(d) shows the electric field intensity profile from the side view, which depicts the smooth adiabatic coupling between the nanobeam and the fiber taper. From the simulation of Fig. 1(d), we achieved a fiber-coupled brightness of 88% at b = 400 nm and L_taper = 15 μm.

To fabricate the designed structure, we used electron beam lithography and wet chemical etching to pattern an initial wafer containing InAs/InP quantum dots (density of ∼10 μm⁻²) that were grown by molecular beam epitaxy.³⁶ The wafer consisted of a 280 nm thick InP slab that featured an InAs quantum dot layer at the center, grown on top of a 2 μm thick AlInAs sacrificial layer. We deposited a 220 nm silicon nitride film as an etching mask using plasma-enhanced chemical vapor deposition, followed by electron beam lithography and fluorine-based reactive ion etching to produce the nanobeam pattern on the silicon nitride. Chlorine-based reactive ion etching transferred this pattern onto the InP layer. Then, we removed the sacrificial layer using selective wet etching to make the freestanding nanobeam. We added a rectangular pad a few microns in size on the nontapered side of the nanobeam to help the transfer process [Fig. 2(b)].

We fabricated the fiber tapers from single mode optical fibers using dynamic chemical etching.³⁷ In this process, we stripped one end of the optical fiber from its coating and dipped it into a 50% hydrofluoric acid (HF) solution for approximately 45 min. As the fiber tip etched away, we slowly pulled the fiber out of the solution using a motorized stage. We were able to control the taper angle depending on the HF concentration and the pull-out speed. In this manner, we fabricated fiber tapers with a taper angle of 1° and a tip diameter of approximately 100 nm [Fig. 2(a)].

We transferred the nanobeam onto the fiber taper using a tungsten probe installed in a focused ion beam system.³⁸ After placing the probe onto the pad of the nanobeam, we welded the probe tip to the device by depositing silicon dioxide. Then, the ion beam removed the remaining materials tethering the nanobeam to the substrate, allowing us to pick up the structure from the chip [Fig. 2(c)] and place it onto the desired region of the fiber taper. In order to ensure attachment, we deposited additional silicon dioxide to “glue” the nanobeam and fiber together. Then, we disconnected the pad from the nanobeam by further ion beam etching. Figure 2(d) shows a nanobeam transferred onto a tapered fiber. As a final step, to improve the mechanical stability of the setup, we coated the entire system with a 50 nm thick aluminum oxide layer using atomic layer deposition.

In order to characterize the properties of the fabricated structure, we constructed an all-fiber photoluminescence measurement setup (Fig. 3, the detailed setup figure and explanation are in supplementary material Note 4). We performed all measurements inside a closed-cycle cryostat that cooled the sample to 4 K. We can cool the sample without a direct thermal contact to a coldfinger because we fill the cryostat (attocube attoDry 1000) with helium gas (16 Torr) for the heat transfer. We excited the quantum dots with a Ti:sapphire laser operated at 780 nm in both continuous-wave and pulsed modes with a repetition rate of 76 MHz. In order to measure the spectrum, we spectrally filtered the single photons collected by the fiber using a monochromator and measured the photons using InGaAs array detectors. When we observed the individual peaks of quantum dot emission intensity, we utilized a tunable fiber filter and measured the photons using superconducting nanowire single photon detectors.

Figure 4(a) shows the photoluminescence spectrum obtained by pumping with a 780 nm continuous-wave laser. We observe 20–30 quantum dot lines in the spectrum. The number of quantum dot lines we observed is consistent with the number of quantum dots in the nanobeam, estimated based on the quantum dot density of 10 μm⁻² and the area of the nanobeam except the photonic crystal region (∼3.5 μm²). Each quantum dot has a different wavelength and position, which creates variations in their coupling efficiency. We isolated a particularly bright and narrow quantum dot line at 1376 nm for detailed analysis. Magnifying the region near 1376 nm [yellow stripe in Fig. 4(a)], we observe a narrow peak that corresponds to emission from a single quantum dot [Fig. 4(b)].

To estimate the fiber-coupled brightness, we measured the single photon count rate as a function of the pump power using the Ti:sapphire laser operated at 780 nm in the pulsed mode with a repetition rate of 76 MHz [Fig. 5(a)]. The pump power dependence of the count rate shows clear saturation behavior. We fit the data to the equation $IP=I0+Imax1−e−PPsat$⁠, where I₀ is the dark count rate, I_max is the maximum emission count rate, and P and P_sat are the pump power and the saturation power, respectively. From the fit, we determined a maximum count rate of I_max = 84 kilo-counts per second (kcps).

To verify that the emission originates from a single quantum dot, we measured second-order autocorrelation. We used a 50/50 fiber beamsplitter to divide the photoluminescence signal into two fibers and measured them using independent single photon detectors. We obtained the coincidence counts between the two detectors as a function of photon arrival time delay using a time-correlated single photon counter. Figures 5(b) and 5(c) show the histograms, which represent the second-order correlation g⁽²⁾(τ). To calculate g⁽²⁾(0), we normalized the counts at the center within a period (13.16 ns) to the averaged coincidence counts of the nearest three peaks on each side. We also measured the background counts caused by the dark count of the detectors and subtracted them from the coincidence counts.³⁹ The background coincidence counts due to the dark count are relatively small (∼1 during the integration time of 10 min), and we mark them with gray color in Figs. 5(b) and 5(c) (they can barely be seen). When we set the pump power to 250 nW, which generates a count rate that is half of the saturation rate, g⁽²⁾(0) is 0.09 [Fig. 5(b)], which shows highly suppressed two-photon emission. When we increased the pump power to 1.2 μW to saturate the quantum dot, g⁽²⁾(0) becomes 0.17 due to the background emission by other dots [Fig. 5(c)].⁴⁰ 

In order to estimate the fiber-coupled brightness of this system, we measured the transmissions of all the components, such as the fiber vacuum feed-through, fiber filter, couplers, and connectors (supplementary material Note 4). We determined the total system detection efficiency, which includes the transmissions of every component, to be 7%. To calculate the brightness, we used the measured a single photon count rate I_max of 84 kcps, a setup detection efficiency T of 7%, a g⁽²⁾(0) value of 0.17, and a repetition rate of the pump laser R of 76 MHz. We excluded the additional counts from the multiphoton emission using the g⁽²⁾(0) value⁴¹ as $ISP=Imax·1−g20=77 kcps.$ The collected single photon at the first fiber was $I=Imax·1−g20/T=1.1 Mcps.$ Therefore, the brightness of our fiber-integrated single photon source (device A) is $B=I/R=1.4%$⁠. We performed similar measurements using a different device (device B) with the same design parameters as device A and obtained a brightness of 1.2%, which shows the consistency of our scheme (supplementary material Note 5). The yield of finding a good single photon source was ∼20%, which was mainly limited by the nanobeam dropping, possible misalignment between the nanobeam and the fiber taper, and fabrication imperfections of the nanobeam and the fiber taper (supplementary material Note 5).

There is a significant discrepancy between the measured and calculated brightness values. In order to understand where the additional loss is coming from, we measured the reflectivity of the fiber-integrated nanobeam with a broadband light source. We found that the amount of back-reflected light was approximately 1%, which means that η is approximately 10% (supplementary material Note 6). Thus, we presume that the contact and alignment between the nanobeam and the fiber taper are not in the best condition (supplementary material Note 7). We also note that the nanobeam sometimes drops from the fiber taper during the cooling procedure (supplementary material Note 8). We monitor the broad photoluminescence emitted from the nanobeam during the cooling procedure and sometimes observe the sudden disappearance of the signal indicating the nanobeam dropping. The nanobeam dropping is possibly due to the different thermal expansion coefficients between InP and SiO₂. To prevent the nanobeam dropping, we added a silicon dioxide and aluminum oxide layer after the nanobeam transfer, which possibly causes additional loss (supplementary material Note 9 discusses the numerical simulation of the fiber coupling efficiency depending on the added silicon dioxide and the aluminum oxide layers).

Our system currently has a brightness of 1.4%. This efficiency is significantly higher than that of previous telecom wavelength sources,²⁶ but we note that the scheme of the previous work enables easy access to many different quantum dots with the same fibers. The other work reported brightness as high as 5.8% for self-aligned quantum emitters,²⁷ but these emitters were not at telecom wavelengths. The current brightness we achieve is still far below the theoretical limit of our structure, which is 88%, indicating significant room for improvement. To improve the brightness of this system, we can promote adhesion between the nanobeam and the fiber taper by either chemical or plasma treatment.^42,43 Besides poor adhesion, there is a possibility that chemical etching of the fiber causes surface roughness, which can give rise to scattering loss. However, it is difficult to characterize the transmission of the etched region of the fiber. As an alternative, we can replace the chemically etched fiber with a flame-pulled fiber taper,⁴⁴ which can achieve a high transmission (>90%).⁴⁵ The scattering loss of the flame-pulled fiber can be easily characterized by measuring its transmission, enabling us to investigate whether the main loss is due to the nanobeam-fiber coupling or the fiber taper itself.

In summary, we have demonstrated a fiber-integrated single photon source at telecom wavelengths. We transfer a nanobeam containing InAs/InP quantum dots to a fiber taper using a tungsten probe installed in a focused ion beam system and obtain a brightness of 1.4% through the fiber. Because of its configuration, this system does not require precise free space optical alignment or careful positioning of the fiber taper with respect to the quantum dots. We could improve the efficiency by surface treatment of the nanobeam and the fiber taper or by employing a flame-pulled fiber, which could potentially elevate the brightness to as high as 88%. Although this approach may not be practical to build a large scale quantum network with multiple quantum dots in parallel, the improved device with high efficiency will pave the way for scalable quantum information processing by employing photon qubits, such as photonic boson sampling,⁴⁶ cluster state generation,⁴⁷ or one-way quantum repeaters.⁴⁸ Employing nanobeam cavities with quantum dots, our system could enable fiber-coupled spin-photon interfaces for applications such as quantum phase switches^49,50 or single photon transistors.⁵¹ 

See supplementary material for detailed finite-difference time-domain simulations provided for the structure optimization. This file also includes the measurement setup details, reflectivity measurement, and single photon measurement data from other devices.

The authors would like to acknowledge support from the U.S. Department of Defense, The Center for Distributed Quantum Information at the University of Maryland and Army Research Laboratory, and the Physics Frontier Center at the Joint Quantum Institute.