SOI CMOS compatible Si waveguide photodetectors are made responsive from 1100 to 1750 nm by Si+ implantation and annealing. Photodiodes have a bandwidth of >35 GHz, an internal quantum efficiency of 0.5 to 10 AW-1, and leakage currents of 0.5 nA to 0.5 μA. Phototransistors have an optical response of 50 AW-1 with a bandwidth of 0.2 GHz. These properties are related to carrier mobilities in the implanted Si waveguide. These detectors exhibit low optical absorption requiring lengths from <0.3 mm to 3 mm to absorb 50% of the incoming light. However, the high bandwidth, high quantum efficiency, low leakage current, and potentially high fabrication yields, make these devices very competitive when compared to other detector technologies.
©2009 Optical Society of America
High frequency, <1 GHz, optical infrared detectors, sensitive between 1100 and 1700 nm, are usually fabricated in semiconducting InGaAs. Although this semiconductor has the advantage of bandgap tailoring, it is incompatible with Si photo-electronic integrated circuit, IC, fabrication, which has led to the development of other IC compatible technologies. Six of these technologies are bonding InGaAs  or Ge  detectors onto fabricated Si IC wafers, using epitaxial Ge as a direct replacement for InGaAs [3–5], using Schottky-metal-silicide phototransistors  porous Si Schottky diodes , and generating mid-band-gap states in Si by ion implantation resulting in absorption and detection of the infrared radiation [8–10]. Bonding of InGaAs diodes to Si wafers is at present not commonly available and expensive. Although Ge-Si alloys are commonly used with IC device fabrication, detectors require 100%Ge and special processing, which can be difficult to accommodate in an IC fabrication line. Implanted Si detectors, while being compatible with IC fabrication, suffer from low optical absorption requiring comparatively long <0.3 to 3 mm waveguide photodetectors to absorb 50% of 1550 nm radiation. Comparison of Ge and implanted Si infrared detectors has been made in an other publication . This article discusses the response, bandwidth, and limitations of infrared implanted Si pin photodiodes and nin and pip phototransistors using infrared radiation between 1560 and 1530 nm.
2. Si Infrared Detectors
In 1959 Fan  reported that radiation damaged Si would produce a photocurrent when illuminated with sub bandgap radiation. Knight [8,13] and others [9–11] have used this result to make waveguide optical detectors. Two crystal defects are believed responsible for this photocurrent, divacancies and interstitial clusters. While the divacancies were found to be efficient in absorbing light, ~100 dB cm-1, they have a lower quantum efficiency than the interstitial clusters. The divacancies are known to anneal out between 100 and 300°C while the clusters are stable to >600°C [10,11,14]. Since the devices reported here are given a 475°C anneal after ion implantation, interstitial clusters are thought to be the photoactive defect. Unlike most other detectors formed in Ge or InGaAs where the optical absorption is >1000 dB cm-1, the Si detectors exhibit absorption from 8 to >100 dB cm-1 and therefore a significant fraction of the light can pass through the detector without absorption.
The Si detectors were fabricated on silicon-on-insulator (SOI) substrates using standard IC processing, which is detailed elsewhere [10,11]. For the devices reported here, shown in Fig. 1, the detectors consist of Si waveguides 520 nm wide and 220 nm thick, which are electrically contacted by p- or n-doped Si wings. Two variations of the detectors were fabricated, structure-A and -B, which are detailed in Fig. 1(b). The Si waveguides are made optically active by ion implantation with 1013 or 1014 cm-2 190 keV Si+. After implantation the detectors experience several thermal cycles, the highest being 475 °C for 2 min.
4. Si Waveguide Photodiodes
The Si waveguide diodes not ion implanted during fabrication have optical absorption ~3 dB cm-1 at 1550 nm and will generate a small photocurrent [10,11,15]. Once implanted, the diodes can be in one of two optically absorbing states. After fabrication the diodes of structure-A are in the “non-activated” L1 state with optical absorption of 8 to 10 dB cm-1. If the diodes are forward biased at a current density of >0.2 A cm-1 for a few seconds they will be in the “activated” L2 state with absorption of 18 to 20 dB cm-1.
If the diode is heated to 250°C for a few seconds it will return to the L1 state. The diode appears to be stable in both states and can be switched between states without degradation. Typical photo response of a photodiode is shown in Fig. 2. The internal quantum efficiency for bias voltages <5 V is the same for both L1 and L2 states and the increase in photocurrent for the L2 state is the result of increased optical absorption. Structure-B Si waveguide detectors with thicker Si wings and higher doping levels exhibit the L1 and L2 states but with inferior quantum efficiency and greater absorption.
The waveguide photodiodes, as shown in Fig. 1, have an optical group velocity of ~7×109 cm s-1, which is smaller than the group velocity, 1.3 ×1010 cm s-1., of the transmission line formed by the metal contacts to the diode. Thus the optical group velocity can limit the bandwidth of the detector, and optical absorption and length of the waveguide detectors directly impacts the tradeoff between the optical response and the electrical bandwidth. As the waveguide length is increased a larger fraction of the incoming light is absorbed and converted into an electrical signal. However, the increased propagation time of the light through the longer device reduces the electrical bandwidth. For the case where the optical group velocity limits the bandwidth the normalized AC electrical power P(ω)/P(0) generated by the incoming light modulated at frequency f in Hz is give by:
where ω =2πf, α= 0.23 A, A is to optical absorption in dB cm-1 , ν is the speed of light in the waveguide in cm s-1 , and F is the fraction of light absorbed the by waveguide photodiode. Fig. 3 shows the trade off between the fraction of light absorbed in the detector and the light transient time limited bandwidth, P(ω)/P(0)=0.5, for several absorption coefficients. What fraction of the absorbed light actually generates a photocurrent or is lost to some other mechanism does not affect the bandwidth. A traveling wave photodiode structure would have a larger bandwidth, but only half of the photo response . To absorb half of the incoming light with A=20 dB cm-1 a waveguide-photodiode-length of 1.5 mm is required, which limits the bandwidth to 22 GHz. Although only half of the incoming light is absorbed in the diode, the internal photoresponse of ~0.5 to 10 AW-1 implies an external quantum efficiency of 0.2 to 5-1 AW is achieved. Even if the waveguide length is increased to 5 mm, which will absorb 90%of the incoming light, the transit-time bandwidth is ~8 GHz.
In addition to the propagation-time of light through the detector, the external circuit components can limit the detector’s bandwidth. These include the resistance and capacitance of the Si wings and contact pads. To minimize the effect of these parasitic components the device geometry was changed from structure-A to structure-B, as shown in Fig. 1. This resulted in a reduction of the resistance of the Si wings from 0.5 to 0.18 Ω cm-1 and a reduction of the contact pad size, not shown in Fig.1, which reduced the parasitic capacitance of <0.06 pF. The capacitance decrease resulted in a detector bandwidth increase from ~20 GHz to >50 GHz. The resistance of the Si wing has little effect on the bandwidth of the detectors with their output impedance of 2 kΩ at 15 V (Fig. 2). The change in geometry and doping was made to improve the response of the Si Mach-Zehnder modulators , which are on the same chip and have the same geometry as the detectors. However the thicker Si wings and the higher doping increase the parasitic optical absorption of the detector waveguide changing it from 8 – 10 dB cm-1 to 13 – 15 dB cm-1and reducing the quantum efficiency.
To determine the intrinsic photodiode bandwidth, the frequency response of a 100-μm-long waveguide photodiode was measured. This short diode absorbs only 4.5%of the incoming light but the 100 μm length implies a propagation-time-limited bandwidth of >200 GHz. The frequency response of the diode was determined using a measurement system containing a LiNbO3 Mach-Zehnder optical modulator with a bandwidth of ~34 GHz and a commercial InGaAs photodiode with a bandwidth >50 GHz, used as a reference. The response of the measurement system with the Si diode and the reference diode are shown in Fig. 4. The normalized photodiode response was obtained by taking the ratio between the measured response of the Si and the InGaAs-reference detectors and is shown in Fig. 5. This procedure eliminates the response of the modulator, RF losses in the cables, and the bias tee. However the response of the RF probes used to contact the Si waveguide photodiode cannot be compensated. The Si-diode response agrees with the InGaAs-reference within 3 dB up to 35 GHz. The calculated RC-time-constant bandwidth of the contact pad capacitance and the 50 Ω input impedance of the network analyzer is ~100 GHz. Also a ~90 GHz bandwidth was calculated considering only the transit time of holes and electrons at their saturated velocity of 1×107 cm s-1. The RF power to the optical modulator was increased with operating frequency to compensate for reduced response. Measurements were made at different RF power levels to ensure that the modulator and photodiodes were operating in their linear region.
Increasing the ion implantation dose to 1×1014 cm-2 increases the absorption coefficient to >100 dB cm-1. The higher absorption makes it possible to absorb 50%of the incoming radiation in a device length of <0.3 mm. However, for bias voltages <5V the enhanced absorption is offset by a lower quantum efficiency lowering the overall device performance. These devices become more desirable when operated at higher voltages, >15V, where the quantum efficiency increases. Figure 6 shows the internal quantum efficiency for photodiodes implanted with 1013 and 1014 cm-2 Si+. The trade-off between ion implantation dose, annealing temperatures and device performance, to our knowledge, has not been investigated.
5. Si Waveguide Phototransistors
Instead of forming a diode with n and p doping, a lateral phototransistor can be formed by doping the Si waveguide wings only n or p. At low bias voltages, <5 V, the photodiode can at most generate one photoelectron per absorbed photon. The phototransistors and photoconductors have an inherent gain mechanism generating several electrons per absorbed photon [18,19]. Figure 7 shows the photo response of nin and pip phototransistors implanted with 1013 cm-2 Si+. The nin devices are especially interesting giving a response of ~50 AW-1 with a 5 V bias. These devices exhibit a significantly higher leakage current than the pin diodes. It can be reduced by a factor of ~3 by biasing the carrier wafer either -60 V or 60 V for the nin and pip devices respectively. The optical absorption coefficient of the phototransistors has not been determined but is between 8 and 20 dB cm-1 and is not dependent upon previous biasing history. The data of Fig. 7 were obtained for approximately the same light input for each device.
To first order, the bandwidth and photo gain of pip and nin transistors depend upon the lifetime of the minority carriers in the ion-implanted region . As the minority carriers, holes, in the nin remain in the transistor they will catalyze the transport of electrons across the device, but at the expense of the bandwidth. Figure 8 compares the bandwidth for pin and nin detectors. While the 1-mm-long pin diode has a length-limited bandwidth of 40 GHz, the nin transistor has a carrier-limited bandwidth of ~0.2 GHz. Pip transistors have a lower gain than nin, but a larger bandwidth of ~1 GHz.
Increasing the Si+ implant dose to 1014 cm-2 in the nin and pip transistors increases the optical absorption to >100 dB cm-1 with some reduction in photo response at the same bias voltage. However the photo response of nin devices implanted with 1014 cm-2 Si+ is still superior to the pin photodiode implanted with 1013 cm-2 at the same bias voltages. The carrier-limited bandwidth is the same for nin devices implanted with 1013 and 10 14 cm-2, ~0.2 GHz.
The higher photo response and limited bandwidth of the nin compared to the pin devices can be understood by comparing the mobility of carriers in the ion-implanted waveguides, as shown in Fig. 10. The carrier mobility, μ cm2 V-1 s-1, was estimated by using the Si carrier wafer, as shown in Fig. 1(a), as a gate and measuring the current across the phototransistor as a function of gate bias, using equation :
Where LSi is the width of the Si waveguide, 520 nm, LSiO2 is the thickness of the buried oxide, 3 μm, LL is length of the waveguide transistor, 1 to 3 mm, VSi is the voltage across the waveguide, 0.1 to 1 V, ε is the dielectric constant of the buried oxide 3.9, εo is the permittivity of vacuum 8.85×10-14 F cm-1, and dISi/dVSub, A V-1, is the slope of the current across the waveguide, ISi, as a function the substrate voltage, VSub.
Device performance is determined by the bulk carrier mobilities. Since we have only measured the carrier mobility at the Si-SiO2 interface, it is assumed for this discussion that the carrier surface mobility is an indication of the Si waveguide bulk properties. The mobility of electrons for the unimplanted Si is ~100 cm2 V-1 s-1. When implanted with Si~ to 1013 cm-2 the electron mobility is reduced and is a strong function of carrier density, approaching the mobility of unimplanted Si at higher carrier densities. Since the hole mobility is orders of magnitude below that of electrons, when a photon generates a hole-electron pair in the nin transistor the hole will remain in the implanted region lowering the barrier between the n-doped regions. This results in a photocurrent gain. This gain is proportional to the lifetime of the holes in the implanted region, which in turn depends upon the recombination time of holes and electrons and the time for the holes to diffuse out of the waveguide . Although ion implantation damage is known to enhance hole-electron recombination, annealing above 400°C removes many of these recombination damage sites . Neglecting recombination, the phototransistor gain is proportional to the ratio of electron and hole diffusion constants , which is proportional to the ratio of the electron and hole mobilities. This enhanced gain is at the expense of bandwidth, since the transistor cannot respond faster than the lifetime of the holes. For pip phototransistors the electrons having a higher mobility remain in the implanted region for a shorter time than the holes resulting in less photo gain. Therefore pip devices have a higher bandwidth ~1 GHz, but a lower response. The 1014 cm-2 implant reduces the electron mobility and the resulting nin devices have a lower photo gain at the same bias voltage, but the hole mobility remains constant and the bandwidth is still ~0.2 GHz, the same for nin devices implanted to 1013 cm-2.
The unusual versatility of these Si photodetectors opens the way for several on-chip feedback structures. Figure 11 shows a potential use for the high-photo-response nin transistors, setting the operating point of a Si ring resonator digital modulator . A common problem with these modulators is keeping the Si-ring resonance matched to the incoming light wavelength despite wavelength drift in the light source and thermal fluctuations that shift the ring resonance. Figure 11 shows a feedback circuit, based on nin phototransistors that can automatically tune the resonator such that the average optical power is equal in the two output ports. A pair of nin phototransistors in a push pull circuit can generate ~35 mW of electrical power for each 1mW of absorbed light. Assuming that the ring is heated to ~50°C with ~1.5 mW of heater power to maintain it at resonance, then a minimum of 150 μW of optical power is required to maintain the ring at resonance. For an input power level of 1 mW each detector should be ~100 μm long to absorb sufficient optical power.
The feedback circuit is designed to compensate for slow drift of the input wavelength or resonant wavelength. The loop bandwidth is limited by the millisecond response time of the heater. At RF frequencies above the loop bandwidth the circuit functions as an optical modulator but requires that the average value of the modulated light be ~50%of full transmission for time periods longer than the heater thermal response time. Similar electro-optic feedback structures have been proposed and fabricated, but they have required additional amplifier gain making for a more complicated fabrication procedure . To our knowledge, this is the first report of on-chip feedback using only the gain of an optical detector.
In sumary, Si waveguide photodetectors activated by ion implantattion can be fabricated as phototransistors with photoresponse of 50 AW-1 and a bandwidth of ~0.2 GHz or photodiodes with response to incoming light of 0.5 - 10 AW-1, bandwidth of >35 GHz, and comparatively low leakage currents, 0.5 nA to 5 μA. The photodiodes implanted with 1013 cm-2 Si+ will operate with a response of ~0.5 AW-1 at zero bias voltage . While the pin photodiodes implanted to 1013 cm-2 exhibit two photo-electric states , L1 and L2, no other devices reported here have these variable photo properties. The properties of the pin, nin and pip devices are summarized in Table 1 below.
All these photodetector design variations are obtained using standard IC fabrication technology, and do not require wafer bonding or epitaxial Ge. The lower absorption of optical radiation requires more chip area than the other photodetectors formed from Ge, InGaAs, metal silicides, or porous Si Schottky diodes. However these Si detectors have the advantages of lower leakage current, higher bandwidth, comparable photoresponse, ease of fabrication, and potentially higher yield. In addition M. E. Grein  has reported implanted Si photodiodes with third order nonlinearity, IP3, of 28 dBm, which is approaching the IP3 of InP-InGaAs photodiodes. All of these properties and the general availibility of Si IC fabrication make these implanted Si infrared detectors practical devices in Si opto-electric integrated circuits.
The authors are grateful to F. X Kärtner, J, Orcutt P. Juodawlkis, N. Spellmeyer, D. Caplin, and R. Drangmeister, for helpful discussions and to D. A. Shibles, D. Castro, J. DeCaprio, D. C. Holohan, K. Keenan, J. Jarmalowicz, I. Poore, S. B. Roy, R. Crocker, M. M. Wood, F. O’Donnell, J. Knecht, K. Krohn, S. Cann, and M. Marchant for expert technical assistance. This work was sponsored in part by the EPIC Program of the Defense Advanced Research Projects Agency and in part by the Department of the Air Force under Air Force Contract FA8721-05-C-0002. Opinions, interpretations, conclusions, and recommendations are those of the authors and do not necessarily represent the view of the United States Government.
References and links
1. H. Park, A. W. Fang, R. Jones, O. Cohen, O. Raday, M. N. Sysak1, M. J. Paniccia, and J. E. Bowers, “A hybrid AlGaInAs-silicon evanescent waveguide photodetector,” Opt. Express 15, 6044–6052 (2007). [CrossRef] [PubMed]
2. L. Chen, P. Dong, and M. Lipson., “High performance germanium photodetectors integrated on submicron silicon waveguides by low temperature wafer bonding,” Opt. Express 16, 11513–11518 (2008). [CrossRef] [PubMed]
3. T. Yin, R. Cohen, M. M. Morse, G. Sarid, Y. Chetrit, D. Rubin, and M. J. Paniccia, “31GHz Ge n-i-p waveguide photodetectors on silicon-on-insulator substrate,” Opt. Express 15, 13965–13971 (2007). [CrossRef] [PubMed]
4. D. Ahn, C. 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, 3916–3921 (2007). [CrossRef] [PubMed]
5. M. Kim, O. O. Olubuyide, J. U. Yoon, and J. L. Hoyt, “Selective Epitaxial Growth of Ge-on-Si for Photodiode Applications,” ECS Transactions 16, 837–847 (2008). [CrossRef]
6. S. Zhu, G. Q. Lo, M. B. Yu, and D. L. Kwong, “Low-cost and high-gain silicide Schottky-barrier collector phototransistor integrated on Si waveguide for infrared detection,” Appl. Phys. Lett. 93, 071108 (2008). [CrossRef]
7. F. Raissi and M. M. Far, “Highly sensitive PtSi/porous Si Schottky detectors,” IEEE Sensors J. 2, 476–481 (2002). [CrossRef]
8. A. Knights, A. House, R. MacNaughton, and F. Hopper, “Optical power monitoring function compatible with single chip integration on silicon-on-insulator,” Conference on Optical Fiber Communication, Technical Digest Series 86, 705–706 (2003).
9. Y. Liu, C. W. Chow, W. Y. Cheung, and H. K. Tsang, “In-line channel power monitor based on helium ion implantation in silicon-on-insulator waveguides.” IEEE Photon Technol. Lett. 18, 1882–1884 (2006). [CrossRef]
10. M. W. Geis, S. J. Spector, M. E. Grein, R.T. Schulein, J. U. Yoon, D. M. Lennon, S. Denault, F. Gan, F. X. Kärtner, 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]
11. M. W. Geis, S. J. Spector, M. E. Grein, R. T. Schulein, J. U. Yoon, D. M. Lennon, C. M. Wynn, S. T. Palmacci, F. Gan, F. X. Kärtner, and T. M. Lyszczarz, “All silicon infrared photodiodes: photo response and effects of processing temperature,” Opt. Express 15, 16886–16895 (2007). [CrossRef] [PubMed]
12. H.Y. Fan and A. K. Ramdas, “Infrared Absorption and Photoconductivity in Irradiated Silicon,” J. Appl. Phys. 30, 1127–1134 (1959). [CrossRef]
13. A.P. Knights, J. D. Bradley, S.H. Gou, and P. E. Jessop, “Silicon-on-insulator waveguide photodiode with self-ion-implantation-engineered-enhanced infrared response,” J. Vac. Sci. Technol. A 24, 783–786 (2006). [CrossRef]
14. S. Libertino, S. Coffa, J. L. Benton, K. Halliburton, and D. J. Eaglesham, “Formation, evolution and annihilation of interstitial clusters in ion implanted Si,” Nucl. Instrum. Methods B 148, 247–251 (1999). [CrossRef]
16. K. S. Giboney, M. J. W. Rodwell, and J. E. Bowers, “Traveling-wave photodetectors,” IEEE Photon. Technol. Lett. 4, 1363–1365 (1992). [CrossRef]
17. S. J. Spector, M. W. Geis, G.-R. Zhou, M. E. Grein, F. Gan, M.A. Popovi, J. U. Yoon, D. M. Lennon, E. P. Ippen, F. X. Kärtner, and T. M. Lyszczarz, “CMOS-compatible dual-output silicon modulator for analog signal processing,” Opt. Express 16, 11027–11031 (2008). [CrossRef] [PubMed]
18. A. Rose, “Performance of photoconductors,” Pro. IRE 43, 1850–1869 (1955). [CrossRef]
19. H Benenking, “Gain and bandwidth of fast near-infrared photodetectors: a comparison of diodes , phototransistors, and photoconductive devices,” IEEE Trans Elec. Dev. ED-29, 1420–1431 (1982). [CrossRef]
20. S. M. Sze, “MOSFET” in Physics of Semiconductor Devices, second edition, (John Wiley & Sons, 1981) Chapter 8, pp. 438–446.
21. K. V. Madhu, S. R. Kulkarni, M. Ravindra, and R. R. Damle, “Analysis of generation and annihilation of deep level defects in a silicon-irradiated bipolar junction transistor,” Semicond. Sci. Technol. 22, 963–969 (2007). [CrossRef]
22. S. M. Sze, “Bipolar Transistors” in Physics of Semiconductor Devices, second edition, (John Wiley & Sons, 1981) Chapter 3, pp. 133–147.
24. P.E. Jessop, L. K. Rowe, S. M. McFaul, A. P. Knights, N. G. Tarr, and A. Tam, “Study of the monolithic integration of sub-bandgap detection, signal amplification and optical attenuation of a silicon photonic chip,” J. Mater Sci.: Meter. Electron. 20S456–S459 (2009)., [CrossRef]
25. M. E. Grein, private communicaton.
26. D. A. Tulchinsky, J. B. Boos, D. Park, P. G. Goetz, W. S. Rabinovich, and K. J. Williams, “High-current photodetectors as efficient, linear, and high-power RF output stages,” J. Lightwave Technol. 26, 408–416 (2008). [CrossRef]