1. Introduction

There has been much work towards improving infrared photodetectors for telecommunication and spectroscopy, such as indium gallium arsenide (InGaAs) detectors, mercury cadmium telluride detectors, quantum-confined material detectors, etc. However, the performance of these detectors is degraded by factors including difficulty of material growth, requirement of cooling, low quantum efficiencies, or large dark current [1]. On the other hand, silicon (Si) photodetectors can have high detection efficiency, low dark currents, and excellent timing resolution, even at room temperature [2]. To use Si detectors beyond ∼ 1.1 μm, much recent work has focused on parametric frequency processes to upconvert infrared signal above the Si bandgap, which has allowed efficient single-photon-level near-infrared (NIR)-photodetections in Si avalanche photodiodes [3, 4, 5, 6]. It was shown that periodically poled lithium niobate (PPLN) waveguide-based upconversion detectors can achieve a detection efficiency higher than 35% with excess noise as low as 10³ counts s⁻¹ [6], outperforming InGaAs detectors, though at the loss of spectral bandwidth. Furthermore, by tuning the pump wavelength and using the phase-matching acceptance bandwidth acting as the frequency selective element, upconversion spectrometers were demonstrated [7, 8, 9, 10].

However, PPLN approaches thus far have required bulk-optics setups. Here, we demonstrate an upconversion Si detector assisted by a coupled gallium phosphide (GaP) planar photonic crystal (PPC) nanocavity; the combined device size is nearly that of the Si photodetector itself. In a resonantly pumped optical cavity, the photon flux in the nth-order upconverted mode is proportional to (Q/V_mode)ⁿ [11], where Q and V_mode are the quality factor and mode volume of the resonant mode. The ultrahigh Q/V_mode in PPC cavities therefore results in a strong improvement of the upconversion efficiency. Already, high efficiency second harmonic generation (SHG) and sum frequency generation (SFG) have been demonstrated in PPC cavities fabricated in various semiconductors [11, 12, 13]. However, in previous demonstrations, the infrared pump and the upconverted signal needed to be coupled with the same optical system [12, 13], which is technically challenging and inefficient because of the large frequency separation. Integrating the PPC upconversion cavity directly on the photodetector greatly simplifies the optical design and improves the collection efficiency of the upconverted mode to nearly 50 %, as there are no on-between optical components. In this approach, the photonic crystal acts as a vertically scattering grating for the upconverted mode that redirects it from the slab mode toward the detector. Hence, combining PPC cavities as efficient upconverters with mature Si detector technology could provide a new architecture of compact and efficient upconversion photodetectors.

2. Device schematic and fabrication

Figure 1(a) illustrates the design of the proposed upconversion detector. A PPC cavity is positioned over a Si photodetector; an air gap or a low-index spacing layer serves to maintain total internal reflection confinement in the PPC cavity. Due to the strongly confined resonant modes in the PPC cavity, the coupled NIR signal is efficiently upconverted into sub-micron wavelength, which in turn illuminates the Si detector and generates photocurrent. As a result of the energy and momentum conservations in the nonlinear process, the upconverted signal scatters out of the photonic crystal efficiently by coupling to the Γ (k_x = k_y = 0) point of the photonic crystal pattern [12]. Thus, the photonic crystal effectively acts as a grating for the upconverted light so that 50% is directed towards the Si detector absorber.

 

Fig. 1 (a) Schematic of the PPC cavity-assisted upconversion Si detector; (b) Photograph of a fabricated upconversion detector; (c) Optical microscope image of the air-suspended PPC cavities on the polymer spacing layer; (d) SEM image of the employed L3 PPC cavity.

Download Full Size | PDF

We fabricated the upconversion detector on a commercial Si detector (Thorlabs, DET100A), as shown in Fig. 1(b). Before the integration of PPC cavities, a 2 μm thick polymer spacing layer (Microposit photoresist, S1813) with an array of air-holes (8 μm×8 μm) was placed onto the detector panel using a micro-transfer technique for suspending the PPC cavities [14]. Here, the air-hole array in the polymer layer was defined with optical lithography and chemical developing. The PPC cavities were fabricated in a 138 nm thick GaP membrane, which was grown on top of a 1 μm thick aluminum gallium phosphide (AlGaP) on a thick GaP wafer by molecular beam epitaxy. We then fabricated air-suspeded PPC cavities on this GaP wafer using a combination of electron beam lithography, chlorine-based dry etching, and wet chemcial undercutting. Thousands of such cavities can be fabricated on a single chip in one run. Trenches were added around the PPC edges to allow easy separation of cavities from the GaP wafer. We designed the PPC cavities with a linear three hole defect (L3) with the lattice constant of a = 500 nm and radii of air-holes of r = 0.24a, which yields resonant modes in the telecom band. To improve the coupling efficiency of the PPC cavities, air-holes around the defect were optimized by adding small perturbation holes [15], as shown in the scanning electron microscope (SEM) image in Fig. 1(d). We transferred PPC cavities using a polydimethylsiloxane (PDMS) stamping technique [16]. First, the GaP chip is pressed against a flexible PDMS film, which separates the PPC membranes from the chip while preserving their arrangement. The adhesion between the membranes and PDMS is weak enough so that the cavities can then be picked up by a tungsten probe and subsequently positioned over the air-holes of the spacing layer on the detector, using a micro-manipulator setup. This process works with near-unity yield; as shown in Fig. 1(c), several PPC cavities were placed onto the air-holes precisely.

3. Results and discussions

Before the evaluation of the fabricated upconversion detector, we optically characterized the transferred PPC cavities using a confocal microscope with cross-polarized excitation and collection to minimize the reflected background. The employed objective lens has a numerical aperture of 0.55, yielding a ∼1.5 μm focusing spot. Figure 2(a) shows the reflection spectrum from one of the PPC cavities, obtained using a supercontinuum laser excitation source. The two modes at wavelengths of 1472.3 nm and 1495.6 nm correspond to the higher-order and fundamental mode of the L3 cavity, respectively, consistent with finite-difference time-domain (FDTD) simulations. Here, due to the perturbation design, the coupling efficiency of the fundamental mode is greater than that of the higher-order mode [15]. The fundamental mode has a Q factor higher than 4,200, obtained from a Lorentzian fitting, as shown in the inset of Fig. 2(a).

 

Fig. 2 Optical characterization of the employed PPC cavity. (a) Reflection spectra of the PPC cavity, showing resonant peaks with high Q factors, indicated by the inset Lorentzian fitting; (b) Spectrum and far-field radiation (inset) of the cavity-enhanced SHG signal; (c) FDTD simulations of the cavity mode and upconverted mode, from left to right are x- and y-components of the fundamental mode, and z-component of the SHG signal at Γ point.

Download Full Size | PDF

To examine the upconversion process in the PPC cavity, we switched the excitation source to a tunable telecom-band laser. By tuning the pump laser to the fundamental mode at 1495.6 nm, a strong SHG signal at the wavelength of 747.8 nm was measured, as shown in Fig. 2(b). By spatially scanning the device, tuning the pump wavelength, and rotating the incident light polarization using a half wave plate, we observed in the second harmonic signal the expected spatial, spectral and polarization dependences of the cavity fundamental mode for the cavity-enhanced upconversion.

The inset of Fig. 2(b) shows a far-field image of the second harmonic radiation from the PPC cavities, recorded on a Si camera. Due to the conservations of energy and momentum, the SHG signal of the fundamental mode would couple to the air band of the Γ point (k_x = k_y = 0) with transversal magnetic polarization, as demonstrated in the simulated electric fields shown in Fig. 2(c). Hence, the far-field radiation of the SHG signal evolves into a dipole distribution via the coherent interference [11]. With the spectrally and spatially aligned excitation, the measured absolute SHG radiation P_SHG is about 73 pW with an input power P_in of 500 μW through the objective lens. By considering the coupling efficiency η of the resonant mode to the objective lens (about 1%), we estimate an efficiency of cavity-enhanced upconversion process P_SHG/(P_inη)² approaching 290%/W.

We then examined the infrared signal observed on the PPC-integrated Si photodetector. To reduce the background noise, these measurements were performed using a mechanical chopper and lock-in amplifier with 1 MΩ load resistance. With a NIR laser coupled resonantly into the cavity, the photocurrent was then recorded on the Si detector. Figure 3(a) displays the measured lock-in amplifier voltage as a function of the incident optical power, showing a good fit to a quadratic dependence on the input power, which is consistent with the power dependence of the second order upconversion process. We plot in Fig. 3(b) a spatial mapping of the photocurrent as the input laser is scanned across the sample. The sub-micron spot of high photocurrent closely matches the expectation for SHG from the sub-micron nanocavity. In addition, because the far-field radiation of the PPC cavity mode is polarized, the photocurrent should be polarization dependent, as verified in Fig. 3(c): the photocurrent nicely fits a cos⁴(θ) function with respect to the angle θ between the polarization of the incident light and the y-axis of the cavity.

 

Fig. 3 Photodetection performances of the upconversion detector. (a) Power, (b) spatial, and (c) polarization dependences of the incident light for the photoelectric conversion. (d) Photoresponsivity versus the coupled power.

Download Full Size | PDF

From the measurement results, for an incident power P_in of 500 μW through the objective lens, we obtained an electrical signal of 20.2 μV on the lock-in amplifier, corresponding to a generated photocurrent I_SHG of 20.2 pA. Since the SHG signal is scattered at the Γ point of the photonic air-band, approximately 50% of it can be assumed to be collected on the Si detector. In combination with the responsivity of the Si detector (approximately 0.5 A/W), and the cavity-enhanced SHG power of 73 pW, the theoretical photocurrent of the upconversion detector should be approximately 0.5 A/W×73 pW×50%≈18.3 pA, which is close to our measurements. In Fig. 3(d), we display the responsivity of the upconversion photodetector defined as I_SHG/(P_inη), which is calculated from the measured electrical signal I_SHG and the power coupled into the PPC cavity (P_inη). Unlike the responsivity of detectors based on linear optical absorption, which is nearly constant for illuminations below saturation, the responsivity of the upconversion detector is a linear function of the incident power due to the nonlinear up-conversion, as shown in the red curve of Fig. 3(d). In our experiment, with 5 μW light coupled into the PPC cavity, the responsivity is about 4.05 × 10⁻⁵ A/W, which is low for an infrared detector. However, for a power of 1 W (which may be the peak power of a pulsed pump field), the responsivity can be as high as 0.81 A/W, which is comparable with that of the currently used InGaAs and germanium detectors. The upconversion detector could also be implemented by employing cavity-enhanced SFG [13]. In a PPC cavity with multiple resonant modes, to detect a weak signal resonant at one of the cavity modes, a strong pump beam could be coupled to another resonant mode to upconvert it via SFG. Similar to SFG-based photodetectors using PPLN [3, 4, 5, 6], the responsivity of the detected light would be determined by the other strong pump light. For a sufficiently strong pump on the order of tens of watts, the responsivity of the detected light could be higher than the current InGaAs detectors. Here, the SFG upconversion detection was not implemented on the fabricated device because a filter on the silicon detector is required to eliminate the SHG signal generated by the strong pump light.

Upconversion processes assisted by nanophotonic devices is a good candidate to realize compact autocorrelators [19]. As a further application of our proposed upconversion detector, we demonstrate here its function as an efficient and compact autocorrelator for ultrafast optical pulse characterization. Figure 4(a) shows the experimental setup. Time-delayed pulses with full width at half maximum of 250 fs, produced by an optical parametric oscillator (OPO) pumped by a mode-locked Ti:sapphire laser (Mira-OPO, Coherent, repetition rate 78 MHz) and passing through an unbalanced Michelson interferometer, were coupled into the PPC cavity via an objective lens. The central wavelength of the OPO output was tuned to spectrally overlap with the cavity resonant mode at 1495.6 nm. We measured the interferometric autocorrelation (IAC) by scanning the delay between the two arms of the interferometer. Figure 4(b) shows the autocorrelation trace of the laser pulse obtained by the upconversion detector. The envelope of the constructive interference shows the features of the IAC, with a peak-to-wing ratio of 8 to 1. In addition, the fringes of the autocorrelation trace wash out in the wings, indicating an instantaneous frequency sweep (chirp) due to the dispersion [20]. We then extracted the pulse duration and chirp from this trace by expressing the complex envelope of the measured pulse as a Gaussian form of $E(t)=\exp \left[-\frac{1}{2}{\left(\frac{t}{{\tau }_p}\right)}^2(1+iA)\right]\exp (-i\omega t)$ and only considering the linear dispersion in the experimental setup. Here, τ_p and A describe the width and chirp of the pulse, and ω is the angular frequency of the pulse. The corresponding IAC can be calculated from

 

Fig. 4 (a) Experimental layout for measuring an ultrafast pulse using the fabricated upconversion detector; (b) Measurement result of the 250-fs ultrafast pulse.

Download Full Size | PDF

By fitting the experimental data into the IAC trace, we found that the pulse width was broadened to about 1.24 ps with a linear chirp parametrized by A = 4.85. This pulse length is consistent with expected pulse dispersion introduced in the optical setup, which includes a 10 m single mode fiber, two fiber collimators, and beam splitters with thickness of 0.1m [20].

4. Conclusions

In conclusion, we have introduced a compact frequency upconversion detector by transferring a PPC cavity onto a commercial Si detector. With the strongly cavity-enhanced SHG, we obtained a responsivity for NIR light as high as 0.81 A/W with the coupled power of 1 W. By side-coupling the PPC cavity with waveguides or tapered fibers, one could expect an improvement in the efficiency η larger than 95% [17, 18], making this detector suitable for telecommunications and optical spectroscopy in the infrared spectrum. SFG-based detections could also be implemented based on the PPC cavity-assisted upconversion detector, which promises a high responsivity of the detected light with a strong pump light. While the demonstrated upconversion detector operates at the telecom band, based on the wide electronic band gap of III–V semiconductors, PPC cavities resonant at even longer wavelength can be fabricated and integrated with commercial detectors, which can extend the detectable wavelength range into mid-infrared. The combination of PPC cavity-based upconverter and photodetector therefore enables a compact and efficient optical detection in challenging spectral regions, and could benefit a range of applications ranging from active imaging to spectroscopy and quantum information technologies.