We present two effective approaches to improve the responsivity of high speed waveguide-based Ge photodetectors integrated on a 0.25μm silicon-on-insulator (SOI) platform. The main cause of poor responsivity is identified as metal absorption from the top contact to Ge. By optimizing Ge thickness and offsetting the contact window, we have demonstrated that the responsivity can be improved from 0.6A/W to 0.95A/W at 1550nm with 36GHz 3dB bandwidth. We also demonstrate that a wider device with double offset contacts can achieve 1.05A/W responsivity at 1550nm and 20GHz 3dB bandwidth.
©2011 Optical Society of America
Silicon photonics [1–3] has been identified as a key technology for next generation chip-level interconnects for super-computing applications . The Ge-based photodetector, serving as the optic to electric converter, is one of the key components of silicon photonic technology. Significant progress has been made in the development of these devices on this platform [5–13], including photodetectors integrated on submicron waveguides [6–11] and large core 3μm waveguides [12,13]. Particular attention has been devoted to submicron waveguide based devices because of their small footprint, a key aspect for application to chip-level optical interconnects where device size is a major concern .
Photodetector responsivity and 3dB modulation bandwidth are two key parameters in photodetector performance. In our previous work , we demonstrated a very compact waveguide Ge photodetector integrated on the 0.25μm silicon-on-insulator (SOI) platform with 0.56A/W responsivity and 25GHz speed. It has also been identified that the relatively low responsivity is caused by metal absorption from the Ti/Al contact on top of the Ge waveguide, and that improvement in the responsivity of the device is possible by re-design of the device structure. In this paper, we present effective approaches to improve the responsivity without sacrificing high speed performance. Two approaches have been employed: increasing the Ge thickness and offsetting the metal contact window. The re-designed Ge photodetector has demonstrated significant improvement in responsivity. We demonstrate that by using double offset contacts, we can improve the responsivity from 0.56A/W to 1.05A/W for a device with an active area of 6X10μm2. A greater than 20GHz 3dB bandwidth has been achieved for this type of device. We also demonstrate that the responsivity can be improved from 0.56A/W to 0.95A/W for a device with an active area of 1.6X10μm2. A greater than 36GHz 3dB bandwidth has been achieved with this smaller-area device.
2. Responsivity improvement
A three-dimensional schematic view of the reported Ge photodetector is shown in Fig. 1(a) . Figure 1(b) illustrates the schematic cross-sectional view of the device. The hollow metal slab shown in Fig. 1(a) and the dashed metal contact via shown in Fig. 1(b) indicate that there can be double offset metal contacts on top of the Ge film for wide devices while narrow devices can only have one metal contact on top of the Ge. The structure utilizes the evanescent coupling scheme. Similar to our previous work reported in Ref , the intrinsic Ge absorption layer is integrated on top of the silicon waveguide to form a vertical oriented p-i-n structure. The device fabrication begins with a 0.25μm thick SOI wafer with 3μm thick buried oxide layer. The silicon wire waveguide was formed by etching a 200nm thick Si ridge with a width of 0.5μm. About 50nm of silicon slab was left intentionally to create an electrical path for contact purposes. The waveguide width in the Ge section is tapered in order to realize different detector width designs. After silicon waveguide fabrication, the wafer was implanted with boron in the silicon waveguide surface and then heavily implanted in the contact areas to form p-type ohmic contacts as illustrated in Fig. 1(b). The wafers went through a rapid-thermal-annealing (RTA) process at 1050°C to activate the dopants. The Ge layer was selectively grown on top of the Si waveguide with a 100nm thick Ge buffer layer using low-temperature (400°C) growth followed by thick Ge growth at high-temperature (670°C). The Ge waveguide width is controlled by the epitaxial growth window opening. After chemical-mechanical-polishing (CMP) steps, a final Ge thickness of 0.9µm was obtained which was 0.1µm thicker than the design target. The wafers then underwent a post-growth-annealing step to reduce the threading dislocations in the Ge film. Phosphorus was implanted on top of the Ge film with a dosage of 1e15 cm−2 to form an n-type ohmic contact area, the activated doping concentration was about 2e19 cm −3 from SRP (spreading resistance profiling) measurement after RTA annealing. The metal contacts for both p and n sides were formed by depositing and patterning a Ti/Al metal stack on top of the doped areas.
From previous work , we learned that the metal on top of the Ge film plays a significant role in light absorption which can degrade the device responsivity. In order to reduce the light absorption by metal, the thickness of Ge material was increased to reduce the overlap of the optical mode with the metal contact. Also, the n-type metal contacts on Ge were offset from the Ge waveguide center to further reduce the overlap with the optical mode. As illustrated in the schematic layouts shown in Fig. 1(a) and Fig. 1(b), two types of devices were designed according to Ge width. Type A devices with a narrow Ge width of 1.6µm and a length of 10µm were designed with one offset metal contact on top of Ge. Type B devices were designed with a Ge width of 6µm and length of 10µm with double metal contacts offset on top of Ge. The dual metal contacts help to reduce the contact resistance by increasing the contact area. For type A devices, the final Ge thickness was carefully chosen to ensure the device speed is dominated by transit time instead of the RC time constant.
Metal contact windows were 0.6µm wide. Phosphorus implantation and n-contacts were designed to be 0.3µm away from the Ge waveguide edge to minimize the electric field at the Ge side wall interfaces. Scanning electron microscopy (SEM) cross-sectional images of the fabricated type A and B detectors are shown Fig. 1(c), and Fig. 1(d), respectively.
Measurement results from the above two types of photodetector are presented here. The dark current was measured to be 2.6nA for the type A detector and 11nA for the type B detector at −1V reverse bias voltage, corresponding to dark current densities of 49mA/cm2 and 22mA/cm2 respectively. Figure 2(a) shows the measured photocurrent and dark current over the voltage range of −2V to 1.0V. Contributions to the dark current mainly come from two sources: Ge bulk material dislocations and surface defects. The dark current densities of the two devices are measured and calculated as 32 mA/cm2 and 11mA/cm2 at −0.5V. From an Arrhenius plot analysis of the dark current , we know that at lower bias voltage the dark current arises mainly from bulk Ge material threading dislocation defects, therefore we can conclude that the large dimension Ge detector has a superior bulk Ge material quality from the selective area growth, while the narrow device has higher defect density in terms of bulk material and surface defects. The dark current can be further reduced by optimizing the Ge growth and annealing conditions, and by improving the passivation of the sidewalls.
The photocurrents of the devices under luminescence were measured using a lensed fiber pair with 3μm spot size. Passive waveguide propagation loss and coupling loss were measured and calculated to enable the responsivity calculation. In order to help the passive alignment and to monitor the power reaching the Ge detector, waveguide taps were attached to each waveguide before the Ge detector section with a tap ratio of 1% (−20dB). Taking into account the input facet coupling loss, Si waveguide propagation loss before Ge waveguide detector and tap coupling loss, the power reaching the Ge detector was calculated. With ~1mW laser power, the optical powers reaching the Ge detector was calculated to be about 60µw for both type A and type B devices. The responsivities of type A and B detectors were then estimated to be 0.95A/W and 1.05A/W at 1550nm at −1V respectively. BeamPROP simulation revealed that about 53% of the light is absorbed by Ge in a detector structure with 0.5µm Ge thickness, 1.6µm Ge width, 10µm Ge length and 1µm wide TiAl contact metal stacks on top of the Ge layer. Light absorption in Ge will increase to 69% if the Ge thickness is increased from 0.5µm to 0.9µm with the same structure. Reducing TiAl metal contact width to 0.6µm increased light absorption in the Ge to 76%. Offseting the 0.6µm metal contact toward the edge of Ge waveguide without modifying the geometry of Ge film further increased light absorption in the Ge to 81%. Figure 2 illustrates the simulated field distribution of 6 µm wide Ge detectors along the device longitudinal direction. The thickness of Ge material is 0.9 µm in simulation. The light is incident from the bottom silicon waveguide and then evanescently coupled back and forth between the top Ge layer and the bottom silicon waveguide. The monitor value plots shown on the right side of Fig. 2 reveal the total power (green line), power in Ge layer (blue line) and power in Si layer (red line) versus the propagation distance. It is found out that only 67% light was absorbed by Ge material when metal contact was placed on top of Ge wave guide as shown in Fig. 2(a), and almost 100% of the light was absorbed by Ge with two offset metal contacts as illustrated in Fig. 2(b).
The responsivity spectra of the two types of device over the range of 1520 nm to 1620 nm are shown in Fig. 3(b) for −1V bias case. A flat responsivity was measured up to 1580 nm wavelength for both types of detectors. The type A device (1.6μm x 10µm) achieves an average responsivity of 0.95A/W over the range of 1520nm to 1570nm, and type B (6μm x 10µm) achieves a responsivity of 1.05A/W at 1550nm and reaches 1.19A/W at 1530nm. The responsivity at 1620nm was measured as 0.4A/W and 0.6A/W at −1V for type A and B detectors respectively.
3. High speed performance
The frequency responses of the reported devices were measured by a vector network analyzer (VNA). A high-speed RF signal from the VNA was applied to an external high-speed modulator with a bandwidth of about 40GHz. A reverse voltage bias was applied to the device through a bias-tee. The modulated light at 1550nm was then coupled to the device and the electrical output was measured through a high speed RF probe. The system, including RF cable, bias-tee, and modulator was calibrated and its response was factored out from the high-speed results. Figure 4(a) shows the normalized optical response of a type A detector with an active area of 1.6µm x 10µm. The measured 3dB bandwidths of the device are 22GHz and 36GHz at biases of 0V and −1V, respectively. The device at −1V bias is fast enough to detect a 40Gbs/s optical signal. The measured resistance and capacitance of the device is about 260Ω and 8.5fF, respectively. The RC delay limited frequency is about 60GHz considering the cable impedance of 50Ω. The transit time limited performance is estimated as being as large as 41GHz based on the equation ttransit = 0.44 tGe /vsat, where vsat is the saturation drift velocity in Ge, and tGe is the thickness of the intrinsic Ge film. The measured 3dB bandwidth of 36GHz is close to the calculated transit time limited speed of 41GHz, indicating that type A device speed is limited by the transit time. The high speed performance of the device can be further increased by carefully designing a thinner Ge layer without sacrificing much responsivity.
In the case of the dual contact type B detector, the series resistance is measured to be 169 Ω and the capacitance is measured to be 32fF, so its RC delay limited frequency is about 22GHz. The normalized frequency response of the device with an active area of 6µm x 10µm is shown in Fig. 4(a). The measured 3dB bandwidth of the type B detector is 20GHz, which confirms that this device is RC delay limited. For the dual contact detector, high speed performance can be improved by optimizing device active area to reduce RC delay time constant. Simulation reveals that with a more careful design of the metal width and active area, the 3dB bandwidth can be improved above 40GHz with very limited responsivity reduction.
The eye-diagram measurement used a similar experimental set up. A pseudorandom binary sequence (PRBS) signal with (223-1) pattern length at a 12.5 Gb/s transmission rate was applied to the device. The PRBS signal was amplified by a commercial modulator driver. The signal was combined with DC bias using a bias Tee and applied to the commercial modulator. The modulated light signal was amplified by an EDFA and fed into a digital communication analyzer with an optical module. A typical optical eye-diagram for the 1.6μm x 10μm type A detector at 12.5Gb/s transmission rate is shown in Fig. 4(b) for 1550nm wavelength. A clear eye opening is observed. Higher transmission rates are possible given the device 3dB bandwidth of 36GHz, which suggests that it can be operated at > 40Gb/s. However, 12.5Gb/s is the maximum capability of the pattern generator available to us. Nevertheless, the 3dB bandwidth and eye-diagram measurements confirm that the reported device is capable of high speed operation.
In conclusion, we have demonstrated compact, low-dark current, high responsivity, and high-speed Ge p-i-n photodetectors integrated on 0.25μm thick SOI waveguides. An evanescent-coupled vertical p-i-n structure is used for high performance and straightforward fabrication. One demonstrated device has a very compact active area of only 16µm2, a 3dB bandwidth of over 36GHz, a responsivity of 0.95A/W over the wavelength range of 1520nm to 1550nm, and a dark current of 4.6nA at −1V reverse bias. Another detector with a different design achieved a higher responsivity of 1.05A/W at 1550nm with an optical bandwidth of 20GHz and a dark current of 11nA. The fabrication process used to fabricate this device is fully compatible with CMOS technology developed for microelectronic circuits. The device can be readily integrated with a trans-impedance amplifier (TIA) to form a high-speed, high performance receiver.
The authors acknowledge funding of this work by DARPA MTO office under UNIC program supervised by Jagdeep Shah (contract agreement with SUN Microsystems HR0011-08-9-0001). The authors greatly acknowledge Dr. Jonathan Luff from Kotura Inc. for helpful discussions and Mr. Chatchai Bushyakanist, Mr. Ky Tran from Kotura Inc. for their support on device measurement.
The views, opinions, and/or findings contained in this article/presentation are those of the author/presenter and should not be interpreted as representing the official views or policies, either expressed or implied, of the Defense Advanced Research Projects Agency or the Department of Defense.
References and links
1. R. A. Soref, “Silicon-based optoelectronics,” Proc. IEEE 81(12), 1687–1706 (1993). [CrossRef]
2. L. C. Kimerling, D. Ahn, A. B. Apsel, M. Beals, D. Carothers, Y.-K. Chen, T. Conway, D. M. Gill, M. Grove, C.-Y. Hong, M. Lipson, J. Liu, J. Michel, D. Pan, S. S. Patel, A. T. Pomerene, M. Rasras, D. K. Sparacin, K.-Y. Tu, A. E. White, and C. W. Wong, “Electronic-photonic integrated circuits on the CMOS platform,” Proc. SPIE 6125, 612502 (2006). [CrossRef]
3. G. T. Reed and A. Knights, Silicon Photonics (Wiley, 93–97, 2004).
4. A. V. Krishnamoorthy, R. Ho, X. Zheng, H. Schwetman, J. Lexau, P. Koka, G. Li, I. Shubin, and J. E. Cunningham, “Computer systems based on silicon photonic interconnects,” Proc. IEEE 97(7), 1337–1361 (2009). [CrossRef]
5. J. 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(1), 011110 (2005). [CrossRef]
6. D. Ahn, C. Y. Hong, J. Liu, W. Giziewicz, M. Beals, L. C. Kimerling, J. Michel, J. Chen, and F. X. Kärtner, “High performance, waveguide integrated Ge photodetectors,” Opt. Express 15(7), 3916–3921 (2007). [CrossRef] [PubMed]
7. T. Yin, R. Cohen, M. M. Morse, G. Sarid, Y. Chetrit, D. Rubin, and M. J. Paniccia, “31 GHz Ge n-i-p waveguide photodetectors on Silicon-on-Insulator substrate,” Opt. Express 15(21), 13965–13971 (2007). [CrossRef] [PubMed]
8. L. Vivien, J. Osmond, J. M. Fédéli, D. Marris-Morini, P. Crozat, J. F. Damlencourt, E. Cassan, Y. Lecunff, and S. Laval, “42 GHz p.i.n germanium photodetector integrated in a silicon-on-insulator waveguide,” Opt. Express 17(8), 6252–6257 (2009). [CrossRef] [PubMed]
10. S. Assefa, F. Xia, S. W. Bedell, Y. Zhang, T. Topuria, P. M. Rice, and Y. A. Vlasov, “CMOS-integrated 40 GHz germanium waveguide photodetector for on-chip optical interconnects,” Optical Fiber Communication Conference (OFC), OMR4 (2009).
11. N.-N. Feng, S. Liao, P. Dong, D. Zheng, H. Liang, C.-C. Kung, R. Shafiiha, D. Feng, G. Li, J. E. Cunningham, A. V. Krishnamoorthy, and M. Asghari, “A compact high-performance germanium photodetector integrated on 0.25μm thick silicon-on-insulator waveguide,” Proc. SPIE 7607, 760704, 760704-6 (2010). [CrossRef]
12. D. Feng, S. Liao, P. Dong, N.-N. Feng, H. Liang, D. Zheng, C.-C. Kung, J. Fong, R. Shafiiha, J. Cunningham, A. V. Krishnamoorthy, and M. Asghari, “High-speed Ge photodetector monolithically integrated with large cross-section silicon-on-insulator waveguide,” Appl. Phys. Lett. 95(26), 261105 (2009). [CrossRef]
13. N.-N. Feng, P. Dong, D. Zheng, S. Liao, H. Liang, R. Shafiiha, D. Feng, G. Li, J. E. Cunningham, A. V. Krishnamoorthy, and M. Asghari, “Vertical p-i-n germanium photodetector with high external responsivity integrated with large core Si waveguides,” Opt. Express 18(1), 96–101 (2010). [CrossRef] [PubMed]