We report a waveguide photodetector utilizing a hybrid waveguide structure consisting of AlGaInAs quantum wells bonded to a silicon waveguide. The light in the hybrid waveguide is absorbed by the AlGaInAs quantum wells under reverse bias. The photodetector has a fiber coupled responsivity of 0.31 A/W with an internal quantum efficiency of 90 % over the 1.5 μm wavelength range. This photodetector structure can be integrated with silicon evanescent lasers for power monitors or integrated with silicon evanescent amplifiers for preamplified receivers.
©2007 Optical Society of America
Silicon is an important optical material because it is transparent at the 1.3 and 1.55 μm telecommunication wavelengths and because of the maturity of silicon processing in the CMOS electronics industry, resulting in potentially low cost and large scale manufacturing capability. Recently, significant research in silicon photonics has been focused on realizing individual components of photonic integrated circuits, including active photonic devices such as lasers [1, 2, 3], modulators [4, 5], and photodetectors  as well as passive waveguide devices. Photodetectors are one of the important components that convert optical signals into the electrical domain for further signal processing, and data manipulation. A germanium waveguide photodetector (WPD) has been demonstrated using selective growth on a silicon-on-insulator platform [7, 8, 9], and a SiGe WPD has been investigated to reduce the lattice mismatch experienced by Ge photodetectors  in the wavelength regime of 1.3 μm or 1.5 μm. A silicon waveguide photodetector has also been demonstrated using implantation to increase photoresponse beyond 1100 nm . These developments are promising due to their processing compatibility with standard CMOS materials. However, their dark current densities are typically higher than conventional III–V photodetectors primarily due to dislocations from the growth on a silicon substrate. In addition, their absorption is typically lower at wavelengths beyond 1550 nm, leading to lower responsivity at longer wavelengths. Recently we demonstrated lasers , and amplifiers  on the hybrid silicon evanescent active device platform. These devices consist of III-V quantum wells bonded to silicon waveguides forming a hybrid waveguide. In this paper, we report a hybrid silicon evanescent waveguide photodetector based on the same platform. The device operates with a responsivity of 1.1 A/W, a quantum efficiency of 90 % covering a wavelength range up to 1600 nm, and dark current of less than 100 nA at a reverse bias of 2 V.
2. Device structure and fabrication
The hybrid silicon evanescent photodetector is comprised of AlGaInAs quantum wells bonded to a silicon waveguide as shown in Fig. 1. As light propagates through the hybrid waveguide, it is absorbed in the III–V region generating electron hole pairs. When the device is under reverse bias, the carriers are swept away as shown with the three arrows in Fig 1(a). The input to the photodetector is a passive silicon waveguide. At the junction of the hybrid waveguide and the passive silicon waveguide, the III–V region is tilted by 7° to reduce the reflection at the waveguide transition [Fig. 1(b)].
The silicon waveguide is formed on the (100) surface of an undoped silicon-on-insulator (SOI) substrate with a 1 μm thick buried oxide using standard projection photolithography and Cl2/Ar/HBr- based plasma reactive ion etching. The silicon waveguide was fabricated with a final height of 0.69 μm, width of 2 μm, and slab thickness of 0.19 μm.
The III–V epitaxial structure, including absorbing quantum well layers, is grown on an InP substrate and its specific information can be found in Ref. 12. The photodetector active absorbing region consists of eight compressively strained quantum wells (0.85 %), and nine tensile strained barriers (-0.55 %). The total thickness of the undoped quantum well region is 0.146 μm. This III–V structure is then transferred to the patterned silicon wafer through low temperature oxygen plasma assisted wafer bonding with 300 °C annealing temperature under vacuum for 12 hours. The specific bonding process is described in Ref. 12.
After removal of the InP substrate with a mixture of HCl/H2O, 12 μm wide mesas are formed by dry-etching the p-type layers using a CH4/H/Ar-based plasma reactive ion etch. Subsequent wet-etching of the quantum well layers to the n-type layers is performed using H3PO4/H2O2. Ni/Au/Ge/Ni/Au alloy contacts are deposited onto the exposed n-type InP layer 10 μm away from the center of the silicon waveguide. Four micron wide Pd/Ti/Pd/Au p-contacts are then deposited on the center of the mesas of the absorber region. After proton implantation on the two sides of the p-type mesa, Ti/Au p-probe pads are deposited. A 450 nm thick amorphous SiNx dielectric layer is deposited by plasma enhanced chemical vapor deposition (PECVD) for electrical isolation between the p-probe pad and the n-type InP layer. The III–V mesa region on the silicon input and output waveguide is then dry etched using the same process used during the p-mesa definition, exposing the passive input and output silicon waveguides. The passive silicon waveguides are exposed at the last step to minimize damage, such as added surface roughness, caused by the III–V processing. The sample is diced with a silicon facet angle of 7°. After the facets are polished, an antireflection coating of Ta2O5 (~5 % reflectivity) is deposited on the silicon waveguide facets. The final III-V absorbing region length in the hybrid photodetector is 400 μm. A scanning electron micrograph (SEM) image of the final fabricated hybrid photodetector and a close view of the junction at the device input are shown in Figs. 2(a) and 2(b) respectively. The silicon confinement factor is calculated to be 63 % with the fabricated device dimensions while the quantum well confinement is calculated to be 4 %.
3. Experiment and results
The device is mounted on a temperature controlled stage set to 15 °C. The photodetector responsivity is measured by launching the light into the silicon waveguide through a lensed fiber and measuring the generated photocurrent with a Keithley 2400 source meter while placing the device under reverse bias. The angle between the fiber and the normal to the facet is ∼25° to maximize the light coupling from the laser source to the 7° angled silicon waveguide facet. The coupling efficiency from the fiber to the input silicon waveguide is estimated to be -5.5 dB by measuring insertion loss of a passive silicon waveguide of the same dimensions. The input polarization is controlled by a polarization controller.
Figure 3(a) shows the measured TE responsivity on the first y-axis with an input power of 0.2 mW. The measured TE responsivity at 1550 nm is 0.31 to 0.32 A/W, and is roughly constant over a range of bias conditions from 0.5V to 3V. Figure 3(a) also shows the calculated quantum efficiency of the photodetector using the estimated -5.5 dB fiber coupling loss. At a reverse bias of 3V the quantum efficiency is ∼ 90 % at 1550 nm. Even though the mode overlap between the silicon rib waveguide and the fundamental hybrid waveguide mode is 63 %, the quantum efficiency is higher than this because the 12 μm wide absorbing layer can collect other higher order modes excited from the input silicon waveguide mode. The coupling from the fundamental mode of the silicon waveguide to each mode of the hybrid waveguide is calculated using FIMMWAVE  and the simulation shows ∼95 % of the input mode is coupled to modes of the hybrid waveguide. Table 1 summarizes the calculated coupling efficiency of four different modes of the detector waveguide. The calculation of quantum efficiency does not take into account the scattering loss. The reflection at the edge of the III–V region is calculated to be approximately 10-6. The refection at the interface is small because the fundamental mode of the silicon waveguide doesn’t experience much of the discontinuity of the III–V layers when it is coupled to the hybrid waveguide. The transverse electric (TE) material absorption coefficient is estimated to be 1594 cm-1 at zero bias by measuring the output power from a silicon output waveguide and using the calculated III–V confinement factors (ΓIII–V) of the fundamental and higher order modes as summarized in Table 1.
The TE spectral response is shown in Fig. 3(b). The edge of the spectral response is red-shifted with a higher reverse bias since the applied electric field increases the absorption at longer wavelength . The transverse magnetic (TM) responsivity is measured to be 0.23 A/W at a wavelength of 1550 nm, which is typically lower than TE responsivity because of the lower TM absorption coefficient of the compressively strained quantum wells.
Figure 4 shows the saturation characteristics of the device at an input wavelength of 1550 nm. The x-axis of the graph shows the coupled input power taking into account the 5.5 dB coupling loss. The laser source is amplified through an erbium doped fiber amplifier (EDFA) with a 50 mW maximum available power out of the lensed fiber. This results in a maximum 14 mW of power coupled into the device. At a lower bias the output current generally saturates faster due to carrier screening effects . This can be improved with shallower quantum wells, but we are interested in photodetectors that can be integrated with lasers and amplifiers without any additional fabrication steps. Given this motivation, the same quantum well design was used for both the photodetector and the laser structure outlined in Ref. 12. The 1-dB saturation input power is 1.8 mW and 8.8 mW for 0 V and 1 V reverse bias, respectively. No output current saturation is observed beyond a reverse bias of 4 V for the available 14 mW of fiber coupled power.
The I–V curve of the device is shown in Fig. 5. The dark current is typically 50 nA to 200 nA with a bias range of -1V to -4 V, and breakdown occurs when the reverse bias exceeds 16 V. The exponential increase of dark current as reverse bias is increased (inset Fig 5) indicates that the dark current is likely dominated by band-to-band tunneling. The diode ideality factor (n) under small forward bias (<0.5 V) is measured to be 2, indicating that the recombination current in the quantum well region is dominant. The 11 ohm series resistance beyond diode turn-on (0.8 V) is due to the thin n-layer and the contact resistances.
The device capacitance was measured using a C–V meter with different reverse biases and the results are shown in Fig. 6(a). The capacitance is 7.5 pF under zero bias and decreases down to 5.3 pF as the reverse bias increases. This large capacitance is mainly due to the large III-V mesa size (12 μm x 400 μm). The capacitance of the III-V mesa is calculated to be 3.8 pF with zero bias ignoring the air fringe capacitance. Moreover, two p-probe pads contribute an additional capacitance of 2.95 pF from a 450 nm thick SiNx layer (ε=7.5) sandwiched between the p-probe pad and the n-layers. Figure 6(b) shows measured capacitance with different device lengths. The capacitance is linear with device length since the mesa area and the number of p-probe pads also increase linearly.
The frequency response of the device was measured by a network component analyzer with a 50 Ω termination. The bandwidth of the device is 470 MHz at a reverse bias of 4 V. The measured bandwidth agrees with a RC limited bandwidth of 482 MHz calculated from the measured series resistance and capacitance of the device. The transit time limited bandwidth is 148 GHz, and the frequency response is currently RC limited by the large capacitance from the III–V mesa and the p-probe pads. The capacitance of the mesa can be reduced by reducing the width and length of the III–V mesa. The mesa capacitance can also be reduced by modifying the proton implant profile such that it extends through the top InGaAs p contact layer of the mesa . Moreover, the p-pad capacitance can be minimized by changing the 450 nm thick SiNx insulation layer to a several micron thick benzocyclobutene (BCB, ε=2.6) layer . The RC limited bandwidth and the quantum efficiency calculated with different III–V mesa dimensions is represented in the top and the bottom graph of Fig. 8 respectively. An estimated material absorption coefficient of 1594 cm-1 and a calculated III–V confinement factor of 4 % is used for the quantum efficiency calculation. The calculation shows that a bandwidth of 10 GHz with a quantum efficiency of 60 % is achievable with a 4 μm wide and ∼130 μm long III–V mesa. The figure also indicates that the parasitic capacitance, such as a residual pad capacitance, has a larger effect on the bandwidth of shorter devices and should be kept less than 0.1 pF to achieve a bandwidth greater than 10 GHz. Moreover, traveling wave electrodes can be used to achieve even higher bandwidths, overcoming the tradeoff between the bandwidth and the quantum efficiency .
A hybrid silicon evanescent waveguide photodetector has been demonstrated. The photodetector has a fiber coupled responsivity of 0.31 A/W, with an internal chip responsivity of 1.1 A/W, which is the relevant number for a photonic integrated circuit. The internal quantum efficiency is 90 % and spectral response of the photodetector covers the entire 1.5 μm wavelength range. The device bandwidth is currently RC limited by the large capacitance from the large device size and the SiNx insulation layer of p probe pads, and is measured to be 467 MHz at a reverse bias of 4 V. Calculations show that 10 GHz bandwidth with a quantum efficiency of 60 % can be achieved with smaller III–V mesa dimensions and BCB probe pads. This approach will enable photonic integrated circuits including lasers , amplifiers  and photodetectors with high responsivity and low dark current to be fabricated on a silicon photonics platform. In addition, this device can be used as a power monitor or a pre-amplified receiver combined with a silicon evanescent amplifier without adding any additional fabrication steps or modifications of the III–V epitaxial structure.
We thank K. Callegari, and G. Zeng for assistance in device fabrication and sample preparation; B. Kim for taking SEM images; K. Gan, and J. Shin for high speed measurements and helpful discussions; Mike Haney, Jag Shah and Wayne Chang for supporting this research through DARPA contract W911NF-04-9-0001.
References and links
4. A. Liu, L. Liao, D. Rubin, H. Nguyen, B. Ciftcioglu, Y. Chetrit, N. Izhaky, and M. J. Paniccia, “High-speed optical modulation based on carrier depletion in a silicon waveguide,” Opt. Express 15, 660–668 (2007). [CrossRef] [PubMed]
6. J. F. Liu, D. D. Cannon, K. Wada, Y. Ishikawa, S. Jongthammanurak, D. T. Danielson, J. Michel, and L. C. Kimerling, “Tensile strained Ge p-i-n photodetectors on Si platform for C and L band telecommunications,” Appl. Phys. Lett. 87, 011110 (2005). [CrossRef]
7. J. F. Liu, “Waveguide-Integrated Ge p-i-n Photodetectors on SOI Platform,” Proc. Group IV Photonics Conference, ThA2, 2006. [CrossRef]
8. P. C. D. Hobbs, R. B. Laibowitz, F. R. Libsch, N. C. LaBianca, and P. P. Chiniwalla, “Waveguide integrated MIM tunnel junction NIR detectors,” Proc. SPIE 6477, 647705, (2007).
9. C. Gunn, G. Masini, J. Witzens, and G. Capellini, “CMOS Photonics Using Germanium Photodetectors,” ECS Transactions 3, 17–24 (2006). [CrossRef]
10. A. O. Splett, et. al. “Integrated optoelectronic waveguide detectors in SiGe for optical communications,” Proc. SPIE 2550, 224–234, (1995). [CrossRef]
11. M. W. Geis, S. J. Spector, M. E. Grein, R. T. Schulein, J. U. Yoon, D. M. Lennon, S. Deneault, F. Gan, F. X. Kaertner, and T. M. Lyszczarz, “CMOS-Compatible All-Si High-Speed Waveguide Photodiodes With High Responsivity in Near-Infrared Communication Band,” IEEE Photon. Technol. Lett. 19, 152–154 (2007). [CrossRef]
13. H. Park, A. W. Fang, O. Cohen, R. Jones, M. J. Paniccia, and J. E. Bowers, “An Electrically Pumped AlGaInAs-Silicon Evanescent Amplifier,” IEEE Photon. Technol. Lett. 19, 230–232 (2007). [CrossRef]
14. Integrated Optics Software FIMMWAVE 4.5, Photon Design, Oxford, U.K. [Online]. Available: http://www.photond.com
15. D. A. B. Miller, D. S. Chemla, T. C. Damen, A. C. Gossard, W. Wiegmann, T. H. Wood, and C. A. Burrus, “Electric Field Dependence of Optical Absorption near the Bandgap of Quantum Well Structures,” Phys. Rev. B 32, 1043–1060 (1985). [CrossRef]
16. T. H. Wood, J. Z. Pastalan, C. A. Burrus, B. C. Johnson, B. I. Miller, J. L. Demiguel, U. Koren, and M. G. Young. “Electric Field Screening by Photogenerated Holes in MQWs: A New Mechanism for Absorption Saturation,” Appl. Phys. Lett. 57, 1081–1083 (1990). [CrossRef]
17. J. W. Raring, E. J. Skogen, L. A. Johansson, and L. A. Coldren, “Enhanced frequency response in buried ridge quantum well intermixed SGDBR laser modulators,” Proc. CLEO, CWC1, (2004).
18. A. Tauke-Pedretti, M. Dummer, J. S. Barton, M. N. Sysak, J. W. Raring, and L. A. Coldren, “High Saturation Power and High Gain Integrated Photoreceivers,” IEEE Photon. Technol. Lett. 17, 2167–2169 (2005). [CrossRef]
19. K. S. Giboney, M. J. W. Rodwell, and J. E. Bowers, “Travelling-Wave Photodetector Design and Measurements,” IEEE J. of Sel. Top. Quantum Electron. 2, 622–628, (1996). [CrossRef]