CMOS compatible infrared waveguide Si photodiodes are made responsive from 1100 to 1750 nm by Si+ implantation and annealing. This article compares diodes fabricated using two annealing temperatures, 300 and 475 °C. 0.25-mm-long diodes annealed to 300 °C have a response to 1539 nm radiation of 0.1 A W-1 at a reverse bias of 5 V and 1.2 A W-1 at 20 V. 3-mm-long diodes processed to 475 °C exhibited two states, L1 and L2, with photo responses of 0.3 ±0.1 A W-1 at 5 V and 0.7 ±0.2 A W-1 at 20 V for the L1 state and 0.5 ±0.2 A W-1 at 5 V and 4 to 20 A W-1 at 20 V for the L2 state. The diodes can be switched between L1 and L2. The bandwidths vary from 10 to 20 GHz. These diodes will generate electrical power from the incident radiation with efficiencies from 4 to 10 %.
© 2007 Optical Society of America
In 1959 Fan and Ramdas  reported that radiation damaged Si contains mid-bandgap states that produce optical absorption and photocurrents for wavelengths as large as 4 µm, 0.3 eV, as shown in Fig. 1. This infrared photocurrent was thought to be of little practical value until Knights et al. [2,3] in 2003 reported that diodes formed in ion-implant-damaged Si rib waveguides exhibit photo response of 0.008 A W-1 at 1550 nm with an electrical bandwidth of 2 MHz. The diodes consisted of a long Si optical waveguide, ~1 cm, doped on either side to form a PIN diode. Later work  using submicrometer waveguides reported photo response from 0.05 to 0.8 A W-1 with bandwidths between 10 and 20 GHz. Ion implantation induces a variety of crystal defects which anneal out of the crystal below 100 °C [5,6] forming specific atomic arrangements including divacancies, V2, and interstitial clusters. These specific atomic arrangements could be considered as nano-crystal structures or defect molecules imbedded in a single crystal Si matrix. As reported by Fan et al. , V2 are believed to be responsible for photo response at 1550 nm. Since they anneal out of Si between 150 and 300°C [7,8], the devices reported by Knights et al. were not heated above 300 °C during processing after ion implantation.
This article reports on submicrometer strip waveguides heated to 475 °C during fabrication after ion implantation, which affects optical loss, quantum efficiency and defect characteristics within the silicon. At these temperatures, only the interstitial clusters, In, are electronically active, where n is the number of interstitials in the cluster. The distribution of n is unknown, but is believed to be between 2 and 200 [8–12] with the most likely value of 8 . For waveguide photodiodes reported here the optical absorption decreases from ~100 db cm-1 for devices processed to 300 °C to 8 or 18 dB cm-1 for devices processed to 475°C, while the quantum efficiency at 1550 nm increases from 0.2 to between 0.5 and 0.8 A W-1 for a reverse bias voltage of 5 V.
Devices processed to 475 °C, as required for a typical CMOS metallization process, can be in one of two stable optically-active states, arbitrarily labeled L1 and L2. After device fabrication the interstitials are in state L1 with a waveguide photodiode optical absorption of 8 dB cm-1. If the diode is forward biased for several minutes the In change their character into the L2 configuration with absorption of 18 dB cm-1. The L2 state can be transformed back to the L1 by annealing the diode to 250 °C for 10 s in air. In have been cycled between these two states without any measurable character change. A 3 mm long L1 state diode will absorb 40 % of the light passing through it and gives a photo response of 0.3 ±0.1 A W-1 at 5 V to 0.7 ±0.2 A W-1 at 20 V. The same diode in the L2 state will absorb 70 % of the input light and has a photo response 0.5±0.2 A W-1 at 5V and 4 to 20 A W-1 at 20 V. The bandwidth measurements of shorter 0.25 mm devices vary between 10 and 20 GHz. The shorter devices absorb 5 % and 10 % of the incoming light respectively for the L1 and L2 states. The diodes are stable in either state for days to weeks, however, the L2 state will slowly transform into a modified L2 state, L2m, over a period of minutes to hours when operated at reverse bias voltages>10 V.
These diodes will generate electrical power when illuminated with 1550 nm light. A measured overall optical to electrical power efficiency of 4 to 10 % with an open circuit voltage of -0.67 V for 1 mW of 1539 nm radiation has been achieved. Diodes in L1 state and those only processed to 300 °C containing V2 have open circuit voltages of -0.63 and -0.52 V respectively.
Diagrams and a micrograph of the photodiodes are shown in Fig. 2. The central region of the waveguide has a cross section of 0.52×0.22 µm with 50-nm-thick Si wings connecting the waveguide to electrical contacts 900 nm away. The thin Si wings have little effect on the mode of the waveguide, which has the strong optical confinement of a strip waveguide. The p+ and n+ regions are doped to 1×1019 cm-3 and the p and n region are doped to 1×1018 cm-3.
The diodes were fabricated in Smart-cut  silicon-on-insulator (SOI) substrates using standard CMOS processing. The initial Si thickness of 0.5 µm was thinned to 0.22 µm through oxidation and SiO2 removal in buffered HF. A final wet oxidation performed at 900°C for 40 min coated the surface with 90 nm of SiO2. The SiO2 and Si layers were dry etched to form the Si structure shown in Fig. 2. Ion implantation doping with 5×1013 cm-2 at 20 keV P and 5×1013 at 7 keV B ions was used to form the p and n regions. The p+ and n+ regions were formed with an additional ion implantation of 5×1015 cm-2 at 20 keV of P and 5×1015 cm-2 of B ions respectively. All these ion-implanted dopants were activated at the same time with a 7-s 1000°C anneal in 3% O2 and Ar. Of these implants only the 5×1015 cm-2 P ions will amorphize the silicon layer so that the n+ region is polycrystalline after annealing. The diodes were then implanted with 1×1013 cm-2, 190-keV Si+ using photo resist to restrict the Si implant to the rib region of the waveguide. The resist was removed in an O2-N2 plasma at ~300°C for 60 s. Electrical contact to the diodes processed to 300 °C was made by a lift-off process of 300 nm of Ti-Al. Devices processed to higher temperatures were coated with ~1.5 um of LPCVD SiO2 for 145 min at 430 °C. Contact holes were then dry etched into the deposited SiO2 and the wafer was coated with TiN-W, this deposition required 5 min at 450 °C and 2 min 475 °C. The wafers where then chemical-mechanical-polished to form TiN-W plugs after which TiNAl contacts are sputtered deposited.
3. Static Characterization
Even with the same initial Si+ implantation, the photodiode properties can vary by an order of magnitude depending upon the processing temperature and bias voltage, as shown in Fig. 3. Diodes processed to 300 °C exhibited optical absorption of 80 to 100 dB cm-1 at 1550 nm . This large absorption allows for half of the light to be absorbed in 0.37-mm-long diode. Diodes processed to 475 °C exhibit two stable In states, arbitrarily labelled L1 and L2. After fabrication In are in the L1 state, which is characterized by absorption coefficients of 8 dB cm-1 and a photo response nearly independent of bias voltage, varying by ±7 % between 1 and 15 V. If the diode is forward biased with a current density >300 mA cm-1 for ≥ 6 min, the L1 states are transformed into the L2 states, which is characterized by an absorption coefficient of 18 dB cm-1, and a photo response that is more dependent upon bias voltage varying by a factor of 3 between 1 and 15 V. Figure 3(b) shows the leakage currents for the same set of diodes, displayed in Fig. 3(a). The leakage currents are scaled to the lengths where half of the incoming light is absorbed in the diode. The length of the diode for each process is listed in the legend.
The leakage and photocurrents for a 3-mm-long diode in the L2 state are shown in Fig. 4(a). The ratio of photocurrent generated by 1 mW of 1539 nm radiation to leakage current is >1×106 at 1 V and is comparable to commercial 1550-nm InGaAs PIN photodiodes. Although several groups have reported two photon absorption and photocurrents proportional to the light intensity squared , all the diodes reported here, implanted and otherwise, exhibit a photocurrent linear with light intensity. This linearity is demonstrated in Fig. 4(b) for three orders of magnitude of optical power.
These diodes can generate significant electrical power from the infrared optical input. Figure 5 shows the normalized diode current as a function of bias voltage. Electrical power is generated for the portion of the curves where the current is >0 and the voltage is <0. The efficency of electrical power generation to input optical power is between 4 and 10 % for 3-mm-long photodiodes in the L2 state. The open circuit voltages and fill factors are shown in Table 1 for several processing variations and for an InGaAs photodiode. Similar subband gap radiation results have been reported for Si solar cells ion damaged and annealed to 500 °C .
4. Frequency and Transient Response
The frequency and transient responses of 0.25-mm-long diodes are shown in Fig. 6.
Ideally the minimum transient time response is limited by the 3 ps required for the light pulse to transit the length of the diode, 0.25 mm, and by the ~5 ps for the charge carriers to transit the diode width, 0.52 µm, at saturated velocity ~1×107 cm s-1. However, for our device geometry the response is limited by the contact pad capacitance of ~0.15 pF loaded by a 50 Ω termination, giving a time constant of ~7 ps and a maximum bandwidth of 20 GHz. The high frequency response of these short devices is at the expense of limited optical absorption of the incoming light, 5 % for the L1 and 10 % for the L2 states. The longer 3 mm device, which absorbs 40 % of the incoming light for the L1 state and 70 % for the L2 state, has a time constant of ~80 ps and a bandwidth of ~2 GHz.
The frequency response for the same diode in the L1 and L2 states was measured directly with a network analyzer and by a Fourier transform of two transient responses, one from 0 to 150 ps, part of which is shown in Fig. 6(b), and the other from 0 to 30 ns. The frequency response derived from the transient measurements for the L1 and L2 states were made to coincide with network analyzer measurements at 10 GHz. Although the response derived from the transient measurements and the network analyzer agree within experimental error for the diode in the L1 state, significant deviation between the two techniques below 2 GHz exists for the L2 state. The cause of this divergence is explained in the next section.
Typical transient responses for a diode in the L1 and L2 states are shown in Fig. 6(b). The sampling scope, electrical probes, and cabling used to obtain this data, have a system transient response full-width-at-half-maximum, FWHM, of ~12 ps giving a frequency response of ~10 GHz. In spite of the comparatively low frequency response, the Fourier transformed signals, are above the noise for frequencies up to 100GHz.
Although the transient response of the diode in the L1 state appears nearly equivalent to the 50 GHz InGaAs photodiode, the same diode in the L2 state exhibits substantial “after pulse current” for times >30 ps, which will significantly reduce the bandwidth by increasing the low frequency, <2 GHz, response. We have found an additional modification of the process used to create the L2 state, which reduces the “after pulse current” making the normalized transient response shown in Fig. 6(b) to nearly coincide with the curve of the diode in the L1 state while still exhibiting the higher photo response. This will be discussed in the next section.
5. Activation of L1 and L2 States
As the diode changes from state L1 to L2, caused by a forward current >300 mA mm-1, both the optical absorption and photo response increase, as displayed in Figs. 7 and 8. To the accuracy of our measurements the increase in photocurrent for bias voltage less than 10 V is the result of an increase in photo absorption and the quantum efficiency remains constant. However above 10 V, the diodes in the L2 state exhibit quantum efficiencies exceeding those in the L1 state. As shown in Fig. 7(a), the increase in photocurrent for voltages <10 V saturates in ~2 s of forward bias time, while it takes ~6 min to saturate the photocurrent for larger bias voltages. Once fully transformed into the L2 state the diode will remain there for days to weeks. However, if illuminated while biased above 10 V for several minutes, the photo response decreases for bias voltages >10 V and increases slightly for bias voltages <10 V, as shown in Fig. 7(b). The loss in photo response for bias voltages >10 V is accompanied with a reduction of the “after pulse current” such that the normalized pulse response, as in Fig. 6(b), for diodes in the L1 and L2 states nearly overlay. The state with the reduced “after pulse current” is labeled L2m.
The discrepancy between the frequency responses determined by the network analyzer and by the Fourier transform of the transient response, in Fig. 6(a), is the result of “after pulse current.” During transient-response measurements the average photocurrent is 1 to 10 uA, which does not significantly reduce the “after pulse current” even when biased to 20 V. With the network analyzer the diode is continuously illuminated, with photocurrents of 0.5 to 20 mA at 20 V bias. This continuous illumination reduces the “after pulse current”. Thus the transient measurements exhibit higher low frequency response, being in the L2 state, than the same measurements made with the network analyzer where the diode is in the L2m state.
The during the transition from the L1 to the L2 states the light transmitted through the photodiode decreases and the photo response increases. This transition as a function of forward bias time is shown in Fig. 8(a). Dividing the photocurrent by the light transmitted through the diode, as in Fig. 8(b), we obtain a curve related to the transition from the L1 to the L2 state as a function of time. Since this curve does not fit an exponential function, it is likely that the L1–L2 transition requires the exchange of several electrons or holes [8,16,17].
In this article we compared the optical and electrical properties of waveguide photodiodes that had been implanted with Si ions and then annealed to 300 and 475 °C. These two annealing temperatures generate different infrared-active structures, V2 for devices annealed to 300 °C and In for devices annealed to 475 °C. V2 have four well defined charged states, +1, 0, -1, and -2, and the Fermi energies where the charge transitions occur are well characterized [5,6,18]. In on the other hand consist of interstitial clusters of various sizes. These clusters have 7 well-defined energy levels in the bandgap [17,19,20]. Since the transition rate between these energy levels is not exponential as expected for a point-like defect, it is believed that several electrons or holes are involved in these transitions [8,16,17]. We speculate that these clusters, of ~10 atoms, can be thought of as nano-crystals or defect molecules in a Si matrix where the transitions between energy levels represent atomic rearrangements of the cluster or phase change, after losing or gaining several electrons . The transitions between L1 and L2 states, described in this article, could be the result of a phase change of interstitial nano-crystals or a reconfiguration of a defect molecule.
The transition from the L2 state to the L2m state is not accompanied by a change in the optical absorption, as is the case for the transition from the L1 to L2 state. This result makes it more likely that the enhanced >10-V-bias photo response of the L2 compared to the L2m is the result of charging in the SiO2 cladding or the Si-SiO2 interface. Such charging could cause an increase in the local electric field in the diode, enhance avalanche carrier multiplication and result in a significant “after pulse current.” During photodiode operation with milliampers of photocurrent, much of this charge is neutralized transforming the photodiode into the L2m state with less avalanche gain and less “after pulse current.”
The potential for the photodiodes reported here to still be operational after even higher processing temperatures is likely with Knights’ et al. report that ion-implantation-induced infrared absorption in silicon continues for annealing temperatures >800 °C . Si diodes have been made sensitive to infrared radiation using techniques other than ion implantation, Carey et al.  using pulsed lasers and Raissi et al.  with porous Si, reported sensitivity beyond 1100 nm.
Table 2 summarizes the differences between waveguide photodiodes processed to a maximum temperature of 300 °C and 475 °C after Si+ ion implantation. The lower temperature processed diodes have considerably higher optical absorption 80 to 100 dB cm-1, with a lower quantum efficency and higher leakage current than the diodes processed to the higher temperature. 475 °C processed diodes exhibit two stable states, L1 and L2. The diode can be switched between these two states. Forward biasing the diode will place it in the L2 state and heating the diode to 250 °C will transform it back to the L1 state. The quantum efficiencies are nearly equal for both states. At a high bias voltage, 25 V, the quantum efficiency of the L2 state can exceed 20 A W-1 but after a few minutes at this voltage and with several milliampers of photocurrent the quantum efficiency decreases to the more stable L2m state with a value of 6 to 10 A W-1. The bandwidth for all processing temperatures is between 10 and 20 GHz for 0.25-mm-long photodiodes. Longer diodes have greater photo response, but reduced bandwidth because of increased pad capacitance and optical transit time through the diode.
The all-Si waveguide photodiodes reported here have demonstrated CMOS processing compatibility, photo response from 1100 to 1750 nm with quantum efficiencies between 0.6 and 0.8 A W-1 at 1550 nm, and bandwidth between 10 and 20 GHz. These results are comparable with commercial InGaAs photodiodes, the primary high-frequency photo detector used with optical fiber communication.
The authors are grateful to 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. The Lincoln Laboratory portion of 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.Y. Fan and A.K. Ramdas,“Infrared Absorption and Photoconductivity in Irradiated Silicon,” J. Appl. Phys. 30, 1127–1134 (1959). [CrossRef]
2. 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).
3. 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 (2005). [CrossRef]
4. M. W. Geis, S. J. Spector, M. E. Grein, R.T. Schulein, J. U. Yoon, D. M. Lennon, S. Denault, 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]
5. M. Moll and Ph.D. Thesis, University of Hamburg, DESY-Thesis-1999-040, Dec. 1999, “Radiation Damage in Silicon Particle Detectors,” http://mmoll.web.cern.ch/mmoll/publist/publist.htm.
6. M. Moll, E. Fretwurst, M. Kuhnke, and G. Lindstrom, “Relation between microscopic defects and microscopic changes in silicon detector properties after hadron irradiation,” Nucl. Instrum. and Methods B 186, 100–110 (2002). [CrossRef]
7. L. J. Cheng, J.C. Corelli, J.W. Corbett, and G. D. Watkins, “1.8-, 3.3-, 3.9-u bands in irradiated silicon: correlation with the divacancy,” Phys. Rev. 152, 761–774 (1966). [CrossRef]
8. J. L. Benton, S. Libertino, P. Kringhoj, D.J. Eaglesham, and J. M. Poate, “Evolution from point extended defects in ion implanted silicon,” J. Appl. Phys. 82, 120–125 (1997). [CrossRef]
9. C. J. Ortiz, P. Pichler, T. Fuhner, F. Cristiano, B. Colombeau, N. E. B. Cowern, and A. Claverie, “A physically based model for the spatial and temporal evolution of self-interstitial agglomerates in ionimplanted silicon,” J. Appl. Phys. 96, 4866–4877 (2004). [CrossRef]
10. F. Schiettekatte, S. Roorda, R. Poirier, M.O. Fortin, S. Chazel, and R. Heliou, “Direct evidence for 8-interstital-controlled nucleation of extended defects in c-Si,” Appl. Phys. Lett. 77, 4322–4324 (2000). [CrossRef]
11. N. E. B. Cowern, G. Mannino, P. A. Stolk, F. Roozeboom, H. G. Huizing, J. G. M. van Berkum, F. Cristiano, and A. Cleaverie, “Energetics of self-Interstitial clusters in Si,” Phys. Rev. Lett. 82, 4460–4463 (1999). [CrossRef]
12. P. K. Giri, “Photoluminescence signature of silicon interstitial cluster evolution from compact to extended structures in ion-implanted silicon,” Semicond. Sci. Technol. 20, 638–644 (2005). [CrossRef]
13. M. Bruel, “Separation of Silicon wafers by the smart-cut method,” Mater. Res. Innovations 3, 9–13 (1999). [CrossRef]
14. T. K. Liang and H. K. Tsang, “Role of free carriers from two-photon absorption in Raman amplification in silicon-on-insulator waveguides,” Appl. Phys. Lett. 84, 2745–2747 (2004). [CrossRef]
15. M. Ley and Z. T. Kuznick, “Near-IR improvement of Si photovoltaic conversion by a nanoscale modification,” Physica E 14, 255–258 (2002).
16. P. K. Giri and Y. M. Mohapatra, “Evidence of metastability with athermal ionization from defect clusters in ion-damaged silicon,” Phys Rev. B 2, 16561–16565 (2000). [CrossRef]
17. S. Libertino, S. Coffa, and J. L. Benton, “Formation, evolution and annihilation of interstitial clusters in ion implanted Si,” Phys. Rev. B 63, 195206 (2001). [CrossRef]
18. P. Dolgolenko, P. G. Litovchenko, M. D. Varentsov, G. P. Gaidar, and A. P. Litovchenko, “Particularities of the formation of radiation defects in silicon with low and high concentration oxygen,” Phys. Status Solidi B 243, 1842–1852 (2006). [CrossRef]
19. 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]
20. S. Libertino, S. Coffa, C. Spinella, J. L. Benton, and D. Arcifa, “Cluster formation and growth in Si ion implanted c-Si,” Mater. Sci. Eng. B 7, 137–142 (2000). [CrossRef]
21. A. P. Knights and G. H. Hopper, “Effect of ion implantation induced defects on optical attenuation in silicon waveguides,” Electr. Lett. 39, 1648–1649 (2003). [CrossRef]
22. J. E. Carey, C. H. Crouch, M. Shen, and E. Mazur, “Visible and near-infrared responsivity of femtrosecond-laser microstructured silicon photodiodes,” Opt. Lett. 30, 1773–11775 (2005). [CrossRef] [PubMed]
23. F. Raissi and M. M. Far, “Highly sensitive PtSi/porous Si Schottky detectors,” IEEE Sensors J. 2, 476–481 (2002). [CrossRef]