We demonstrate silicon ridge waveguide photo-detectors capable of sub-bandgap light absorption and avalanche multiplication. The proposed waveguide photo-detectors contain highly doped PN junction, where a strong electric field can generate the photon-assisted tunneling current for sub-bandgap light incidence and amplify the generated photo-current by the avalanche multiplication effect. The voltage-dependent sub-bandgap absorption coefficient and multiplication gain are experimentally evaluated for various doping configurations to find optimal photo-response with low dark currents. As a result, our representative silicon waveguide photo-detector gives sub-bandgap responsivities of ~10 and ~2 A/W under the applied reverse bias voltage of −8.3 V for near-infrared wavelengths of 1.31 and 1.52 μm, respectively. The voltage-dependent frequency photo-response is also demonstrated with theoretical verification.
© 2017 Optical Society of America
Silicon (Si) photonics technologies have attracted considerable attention because of their CMOS-compatible fabrication processes to enable chip-scale integration of photonic and electronic circuits for high-performance information processing systems. The bulk Si is transparent for the near infrared (NIR) wavelength range where the photon energy is below its material bandgap energy of ~1.1 eV (corresponding wavelength of ~1.1 μm). The optical transparency of Si at the NIR region is suitable for guiding light without absorption losses, but also makes light detection very inefficient. For this reason, heterogeneous integration with other semiconductor materials with smaller bandgaps, such as germanium (Ge), has been envisioned as a solution to establish efficient photo-responsive optoelectronic devices at the NIR region in the Si photonics platforms . A rapid progress over the past decade in this field has realized highly responsive and high-speed Ge-on-Si photodetectors (PDs) [2, 3]. However, the monolithic integration of such Ge-based PDs together with other Si electronic circuits requires selective growth techniques while maintaining the stringent CMOS process standards .
There have been growing interests in developing purely Si-based sub-bandgap PDs as an effort to build in-line power monitors , infrared imaging cells , and integrated infrared photo-detectors [7–20]. One way to implement such Si-based NIR PDs is to form deep-level defects in Si lattice by implantation of various ions or dopants such as Au , Si , and Ar  with special post-annealing or -melting processes. Among them, Si ion implantation techniques have been dominantly used by means of forming defects into the intrinsic region of the PIN junction. Such approaches have achieved large responsivities of 0.5~10 A/W for the operation wavelength (λ) near 1.55 μm [9, 10]. In general, however, they typically require relatively large bias voltages ranging from 15 to 35 V to obtain a high carrier multiplication gain and reach a saturation drift velocity [11, 12]. One can also realize Si sub-bandgap PDs based on other defect-mediated absorptions taking advantages of, for example, grain boundaries of poly-silicon , dopant impurities , and surface states at Si-SiO2 interfaces [5, 15]. Zhu et al. have recently demonstrated periodically interleaved PN junctions embedded in narrow Si waveguides capable of enhanced NIR photo-detection using the surface states absorption (SSA) of the guided modes for λ~1.55 μm . The responsivity of the SSA-assisted PDs can be further enhanced up to 2.33 A/W by exploiting the avalanche effects with a multiplication gain of ~284 . More recently, high speed avalanche photo-detection relying on shallow-level impurities has been demonstrated with a responsivity of 0.3 A/W and 3-dB bandwidth of 15 GHz for medium infrared wavelengths longer than λ > 2 μm .
Alternative ways without relying on the intermediate energy levels from the extrinsic defects have also been recently investigated, and they rely on two-photon absorption , internal photo-emission effects , and impurity-assisted Franz-Keldysh absorption (FKA) effects [19, 20]. Zhou et al. have reported on Si core-shell nanowire PDs containing radial PN junctions for λ~1.31 μm through impurity-assisted FKA effects . More recently, they have demonstrated resonance-enhanced sub-bandgap photo-detection for λ~1.31 μm capable of near-unity internal quantum efficiencies .
In this paper, we experimentally demonstrate sub-bandgap NIR photo-current generation without relying on extrinsic defects in the heavily-doped silicon PN-junction ridge waveguides by employing both field-enhanced FKA and avalanche carrier multiplication effects under the non-resonant condition. The vertical PN junction is formed along the ridge waveguide by CMOS-compatible implantation processes and its longitudinal interaction length can be readily controllable in accordance with the available sub-bandgap absorption coefficient. For sub-bandgap light detection, the depletion region of the PN junction becomes absorptive and generates photo-currents by taking advantages of the optically-assisted carrier tunneling (sub-bandgap FKA effects). Since both the carrier multiplication and FKA exist for sub-bandgap NIR light detection with a large internal electric field near the breakdown voltage of the silicon PN junction, considerably high total photo-responsive gains can be realized for both O-band (1.31 μm) and S-band (1.52 μm) wavelengths at the modest voltage bias conditions (<10 V). As a result, 600 μm-long silicon waveguide PDs at a reverse bias of −8.3 V can support high responsivities of ~10 and ~2 A/W with the photo-to-dark current ratio of ~140 and ~30 for O- and S-band light, respectively, when the input power is ~30 μW. Although the dark currents increase as the reverse bias voltage approaches the PN-junction breakdown voltage, the photocurrent can be clearly detected when the doping concentration is not too high. We also investigate bias-dependent operation speeds for different PD lengths to identify how the operating bandwidths are affected and limited by various factors: carrier transit time, avalanche build-up, and RC time delay.
2. Device fabrication
Typical CMOS-compatible silicon photonics fabrication processes were performed to fabricate the proposed silicon ridge waveguide in-line photo-detectors from the silicon-on-insulator substrate. The cross-sectional schematic view and SEM image of the fabricated structure are shown in Figs. 1(a) and 1(b), respectively. The thickness of the silicon, buried oxide, and top-clad oxide layer were chosen to be 250, 3000, and 1000 nm, respectively. The ridge waveguides were defined by a target etch depth of ~150 nm and a width of ~600 nm. The width of the waveguide was chosen to suppress the SSA, which is dominant in narrower waveguides (3~400 nm width) . For sub-bandgap NIR photo-detection, the P + and N + doped regions were formed at the center of the waveguide by ion implantation into 7° tilted targets and subsequent post-annealing processes at 1050 °C for 20 minutes and at 900 °C for additional 30 minutes, which determine the doping profile in the waveguide region. The ion implantation was performed with 40 keV boron and 80 keV phosphorous for P + and N + doping, respectively. The consecutive long-period annealing processes can sufficiently dissolve the clustering of point defects including the dislocation loops and the extended defects [21, 22]. The P + + and N + + region were heavily doped with a doping density of 1020 cm−3 to obtain good contact resistances. The rapid thermal annealing was also subsequently performed at 1050 °C for 30 sec for the activation of the heavily doped region. The TiN-Al electrodes (not shown here) are formed within the P + + and N + + regions. The heavily doped regions are 1.2 μm apart from the waveguide edges to avoid excessive optical losses.
Since the doping profiles of the silicon PN junction play crucial roles in both the optoelectronic performance of the waveguide PDs and the electric characteristics of the reverse-biased diode, we investigate the effects of various doping concentrations for P + and N + doped regions. We consider three different implantation conditions to control the doping density categorized as low-, medium-, and high-doped conditions. According to the implantation dose information and the corresponding process-level simulation results, the average P + (N + ) doping densities of waveguide core are estimated to be ~1 × 1018 (~2 × 1018), ~3 × 1018 (~5 × 1018), ~4 × 1018 (~6 × 1018) cm−3 for low-, medium-, and high-doped samples, respectively. Although the corresponding depletion region is not exactly located at the center of the ridge waveguide due to the asymmetric P + and N + doping configurations, it rarely affects the spatial overlap between the depletion region and the propagating optical mode profile. The confinement factor Γ of the fundamental TE mode [shown in Figs. 1(c) and 1(d)] within the depletion region is shown in Fig. 1(e). For example, only 0.6% reduction of the confinement factor from the asymmetry takes place in the low-doped sample at −8.3 V. The corresponding carrier density profiles of the low-doped waveguides are shown in Fig. 1(f) for the different applied voltages of 0, −2, −4, −6, and −8 V.
The modal free carrier absorption (FCA) coefficient for the low-, medium-, and high-doped waveguides are 14.9, 34, and 48.8 cm−1 (20.1, 47.3, and 64.9 cm−1) at λ~1.31 (1.52) μm, respectively, for the unbiased condition. The FCA coefficients are measured by the cutback method. Together with the carrier density distribution profile, the Drude-Lorentz model can predict the complex refractive index profile for the waveguide cross-section . The modal FCA coefficient can be estimated from the spatial overlap between the optical mode field profile and the complex refractive index distribution. For simplicity, however, the carrier distributions and the corresponding refractive index profile in the ridge waveguide core region are assumed to be uniform for each P + and N + doped region in our theoretical prediction. The estimated FCA coefficients for the low-, medium-, and high-doped waveguides at the unbiased condition are 15.1, 38.1, and 52.3 cm−1 (20.4, 51.3, and 70.3 cm−1) at λ~1.31 (1.52) μm, respectively, and agree well with the measurement values. Differences in doping densities also result in different breakdown voltages (Vb) of −8.35, −6.85, and −6.00 V for low-, medium-, and high-doped PN-junction waveguides, respectively. The Vb values for each doping scenarios are also estimated by computer simulations, and are found to be nearly identical to the measurement results. This comparison confirms that the simulated structures with uniform doping profiles can reasonably describe the behavior of the fabricated devices. The commercial software packages (ATHENA/ATLAS from Silvaco and FDTD solutions from Lumerical) were used to perform the process- and device-level simulations as well as photonic simulations for this work. We believe that the high FCA losses can be reduced by intentionally overlapping the P + and N + region at the junction interface.
A TE-polarized sub-bandgap NIR light from a single-mode optical fiber was coupled into and out of a planar waveguide device through unoptimized grating couplers with coupling losses of 19 and 7 dB per facet for λ~1.31 and 1.52 μm, respectively. As shown in Fig. 1(c) and 1(d), the guided TE mode profiles at 1.31 and 1.52 μm are well confined within the ridge waveguide. The fractions of the optical field energy contained in the core waveguide region are 0.97 and 0.95 for λ~1.31 and 1.52 μm, respectively. The confinement factors within the depletion region of the low-, medium-, and high-doped waveguides under zero bias voltage are estimated to be 0.154 (0.141), 0.132 (0.121), 0.111 (0.100) at λ~1.31 (1.52) μm, respectively. Figure 1(e) summarizes the voltage- and doping-dependent confinement factor Γ for λ~1.52 μm. The propagation losses of un-doped silicon ridge waveguides are measured to be 8 and 2.3 dB/cm (1.85 and 0.53 cm−1) for the input wavelengths of 1.31 and 1.52 μm, respectively, and these corresponds to the propagation losses (αprop) due to the fabrication imperfections such as sidewall roughness.
3. Operation principles: qualitative approach
The silicon material exhibits direct band-edge absorption at 3.4 eV, where the incident photons can be efficiently absorbed with large absorption coefficient > ~106 cm−1 by means of direct-gap electron-hole pair generation . For smaller photon energies from 3.4 eV to ~1.1 eV, the light detection is associated with inefficient phonon-assisted processes due to the indirect optical absorption. When the incident wavelength approaches near the bandgap wavelength, the light incidence suffers material absorption with absorption coefficient of ~2.7 cm−1 in intrinsic Si material at room temperature . In contrast, the Si material no longer absorbs the sub-bandgap photons (λ > ~1.1 μm). A possible solution here is to apply a strong electric field across the material. The strong electric field and the corresponding spatial gradient of the energy bands facilitate the carrier tunneling process with a narrower effective barrier width compared to the zero-field band-edge states. These enhanced tunneling process can greatly promote the absorption probability for sub-bandgap photons.
The metallurgical PN junctions formed in the semiconducting material can provide a convenient way of greatly increasing and controlling the electric field with reverse voltage bias. It has been experimentally verified that such tunneling-assisted absorption in silicon can be enhanced by increasing an applied electric field in the bulk Si PN junctions . This work reported that the applied electric field of 105 V/cm across the low-doped PN junctions, whose doping densities ranging from 1012 to 1016 cm−3, can double the original absorption coefficient of Si for λ~1.17 μm. In this case, the sub-bandgap light detection is basically associated with the inefficient phonon-related indirect absorption processes, and different from the dopant-mediated sub-bandgap impurity state effects where the high electric field condition is not strictly required for photocurrent generation. In order to further enhance the poor indirect optical absorption, highly-doped Si nanowires were previously considered to involve the direct transitions from the impurity states to energetic electrons as well as quasi-quantum confinement effects . Regarding higher doping concentrations on the order of 1018~1019 cm−3, the tunneling-assisted direct transitions through the impurity states become predominant absorption processes due to the abundant impurity density . As a result, a reasonably high sub-bandgap absorption coefficient of αsub~15 cm−1 for λ = 1310 nm at a low bias voltage of ~2.5 V has been predicted for the epitaxially grown PN junctions with a doping density of 5 × 1018 cm−3 .
To qualitatively understand the light absorption phenomena, we show the calculated energy band diagrams near the silicon PN junction for the low-doped device (P + and N + doping densities of 1 × 1018 and 2 × 1018 cm−3, respectively) under two different bias voltages of 0 (black curves) and −8 V (red curves) in Fig. 2(a). When a reverse bias voltage of −8 V is applied, the potential energy across the depletion region increases from qVbi to q(Vbi + 8), where q is the electron charge and Vbi is the built-in potential. Under a large reverse bias voltage, a strong electric field across the junction leads to shorten the spatial distance between the conduction and valence band. Therefore, the bound electrons in the valence band have a chance to tunnel into the nearest conduction band through the triangular potential barrier containing the potential barrier height Eb and the tunneling width wb [described in Fig. 2(b)]. The reduction of Eb and/or wb for this triangular barrier results in the exponential increase of tunneling probability, which can be approximately written as 
Figure 2(b) shows the effect of the bias condition on the tunneling length of the triangle barrier. When sub-bandgap photons are incident on the junction, the photo-excited electrons can undergo a lower barrier height Eb = Eg-Ephoton (sub-bandgap photon energy Ephoton is smaller than the material bandgap energy Eg) and finite wb even for zero bias condition [lower panel of Fig. 2(b)]. The tunneling length wb can be further reduced by increasing the reverse bias voltage. The photo-excited carriers under large reverse bias [upper panel of Fig. 2(b)] can more easily tunnel through the barrier due to the shorter barrier width wb when compared to the initial barrier length without any external voltage bias. Another way of enhancing the tunneling probability, and thus the photo-current and responsivity, is to increase the doping density and thereby increase the built-in potential Vbi. At a given bias voltage (including the unbiased case), more highly doped PN junctions can form shallower wb due to the stronger internal electric field across the junction and result in higher T. The drawback of this approach (higher doping concentrations) is the inevitable increase of dark currents, and will be further discussed in the experimental result section.
Computer simulations (ATHENA/ATLAS from Silvaco) were performed to qualitatively estimate the trend of tunneling probability T as a function of the reverse bias voltage for each doping condition with different photon wavelengths of λ ~1.31 and ~1.52 μm, as shown in Figs. 3(a) and 3(b), respectively. The numerical simulations involved doping-induced bandgap narrowing effects in silicon. The low-, medium-, and high-doped PN junctions under Vb can enhance T up to 0.24 (8.4 × 10−3), 0.286 (0.015), and 0.367 (0.027) for λ~1.31 μm (1.52 μm), respectively. The sub-bandgap light with a lower photon energy level (λ~1.52 μm) gives more sensitive variation of T than that with a higher photon energy (λ~1.31 μm), when the reverse bias voltage and/or the doping density are varied. As the applied reverse bias voltage increases from −0.5 V to the breakdown voltage Vb for each doping concentrations, T for λ~1.31 μm (1.52 μm) increases by a factor of 162 (2.9 × 108), 45 (2.5 × 106), and 6.5 (420) for low-, medium-, high-doped structures, respectively. These factors correspond to the enhancement of initial tunneling-assisted photo-current before considering the carrier multiplication gain under the large reverse bias voltage. At a typical voltage bias of - 4 V, T for the high-doped device is 40 times larger than the low-doped case for λ~1.52 μm, while the ratio of those for λ~1.31 μm is relatively small (~3).
4. Experimental results and discussions
4.1 Static response
The sub-bandgap guided modes along the silicon ridge waveguide PDs with a device length (L) of 600 μm undergo tunneling-based absorption with a sub-bandgap light absorption coefficient of αsub, which mainly arises from the light absorption within the depletion region and contributes the photocurrent. In addition to the tunneling-based absorption process, there are other optical loss mechanisms along the silicon waveguide related with the modal FCA coefficient αfca and propagation loss coefficient αprop, which do not directly contribute the photocurrent generation. To determine the sub-bandgap absorption within the effective cross-sectional area (depletion region) of the silicon waveguide, we consider the confinement factor Γ defined by the amount of energy contained in the depletion region. As the reverse bias voltage approaches toward the avalanche breakdown voltage (Vb), the photo-generated current is amplified with a multiplication gain (M). Therefore, the responsivity can be expressed as
To determine R and αsub, we firstly measured the dark and photo current for each doped sample for λ~1.31 and 1.52 μm, as shown in Figs. 4(a)-4(c) for the low-, medium-, and high-doped samples. The low-doped devices show much lower dark currents than the other doping concentrations. This is because the tunneling probabilities at the dark condition increase significantly with the doping concentrations. As a result, it is hard to distinguish the dark and photocurrent of the high-doped device at the voltage range above −3 V. The medium-doped device also gives only small difference between the dark and photo currents, particularly for λ~1.52 μm. On the other hand, when the coupled input power is 30 μW, the low-doped device at the bias voltage of −8.3 V can give ~140 and ~30 times larger photo current when compared to the dark current level for λ~1.31 and 1.52 μm, respectively. The photo-current linearly increases with the coupled incident power from ~30 μW to 3 mW.
To rationally estimate the sub-bandgap absorption coefficients from the measured output currents, it is important to distinguish the contributions from direct photocurrent generation through sub-bandgap absorption and subsequent carrier multiplication. Figure 4(d) shows the measured carrier multiplication gain of the photo-generated current for waveguide coupled above-bandgap light at λ~0.975 μm. The above-bandgap light confined within the waveguide core region can minimize the diffusion current originated from the heavily doped slab region far from the depletion layer. The incident power of the above-bandgap light was carefully chosen to make the current density comparable to that for sub-bandgap incident case. And At this wavelength, we assume that the above-bandgap absorption is not a function of the input bias voltage, and that the bias-dependent photo-response is only attributed to the multiplication processes. The device operation stability limits the upper bound of the applied reverse bias to be −8.3, −6.75 and −4.3 V, where the maximum multiplication gain is ~930, ~35, and ~2 for low-, medium-, high-doped samples, respectively. The photo-generated carriers in the doped area of the core waveguide can experience more enhanced impact ionization processes for the above-bandgap light than those in the core depletion region for the sub-bandgap light. Therefore, the actual multiplication gain for the sub-bandgap light is possible to be smaller than that for the above-bandgap light. For simplicity, the following analyses assume that the multiplication gain profiles shown in Fig. 4(d) are maintained for sub-bandgap photon incidence.
In Fig. 4(e), the total photo-responsive gain of the low-doped device for the sub-bandgap light of λ~1.52 μm is decomposed into the sub-bandgap absorption and carrier multiplication gain. Under the maximum stable reverse bias of −8.3 V, the total photo-responsive gain, when compared to the original case without any reverse bias, reaches up to ~106, where the sub-bandgap absorption and multiplication gain become ~1300 and ~930, respectively. As a result, a considerably high responsivity of R~2 A/W, where R/M~2 mA/W, can be realized for λ~1.52 μm.
The sub-bandgap absorption coefficients αsub are evaluated by using Eq. (2) and plotted in Figs. 5(a) and 5(b) for different operation wavelengths. As expected in the previous section, the sub-bandgap absorption is sensitively affected by the doping and bias conditions. The low-, medium-, and high-doped PN junctions at the bias voltage of −8.3, −6.75, and −4.3 V give αsub of 2.7 (0.19), 3.9 (0.57), and 2 (0.56) for λ~1.31 μm (1.52 μm), respectively. Note that the maximum bias voltages for low- and medium-doped devices are only slightly different from the actual breakdown voltages Vb of −8.35 and −6.85 V, respectively, while the maximum reverse bias voltage for the high-doped device case (−4.3 V) is quite different from its Vb of −6 V. The rapid increase in dark tunneling current limits the bias range of the high-doped device up to −4.3 V, where the dark current overwhelms the net-photo-current. As the applied voltages varies from 0 V to the reverse bias voltage of −8.3, −6.75, and −4.3 V for the low-, medium-, and high-doped devices, the values of αsub at λ~1.31 μm (1.52 μm) increase by a factor of 82.7 (538.9), 19.3 (60.7), and 5.3 (21.6), respectively. The bias-dependent enhancement of αsub seems to be smaller than the variation of the tunneling probability (T) predicted in the previous section. This is because there is unavoidable sub-bandgap light absorption associated with the defect states arising from the Si-SiO2 surfaces and dislocation loops. Such defect-mediated absorption is irrespective of the voltage bias condition and thus important only at the lower bias range where the tunneling-assisted and defect-mediated absorption are comparable to each other. When the reverse bias increases, however, the tunneling-assisted photo-current generation increases significantly and becomes much larger than the purely defect-mediated absorption.
Figures 5(c) and 5(d) show the measured PD responsivities (R) as a function of the applied reverse bias voltage for λ~1.31 and ~1.52 μm, respectively. It is interesting to observe that the responsivity levels are comparable to each other, particularly for λ ~1.31 μm as shown in Fig. 5(c), even though the values of αsub vary significantly with respect to the doping concentrations as indicated in Figs. 5(a) and 5(b). This discrepancy mainly comes from the fact that the responsivity not only depends on αsub but also other parameters such as Γ and M. Although highly-doped PDs with higher internal electric fields at the depletion region can provide better sub-bandgap absorption values (αsub) when compared to the low-doped PDs, the lower doped devices have wider depletion regions and therefore provide better confinement factor (Γ). Since the lower doped PDs also have smaller FCA coefficients, the overall responsivity levels for different doping concentrations can be similar to each other.
Since our PDs employ the carrier multiplication effects as well, R can be significantly enhanced at near Vb, except for the high-doped PDs due to their too large dark currents. We point out that more dramatic responsivity enhancement takes place in the low-doped PDs than the medium-doped PDs. The bias-dependent absorption enhancement (αsub) and carrier multiplication gain (M) can together provide extremely large overall photo-current enhancement. When the reverse bias voltage approaches the breakdown voltage, the responsivity R for the low-doped PDs reaches up to ~10 and ~2 A/W at λ~1.31 and 1.52 μm, respectively.
4.2 Dynamic response
We also demonstrate the frequency response of our low-doped silicon waveguide PDs obtained by impulse response measurements using a femtosecond laser with 600 fs pulse width and a 40 GHz bandwidth sampling oscilloscope (Keysight 86100B/86109A). The measured temporal response curves are Fourier-transformed to obtain the frequency response curves. The frequency responses from the external instruments and cables are de-embedded to consider only the integrated PD devices. The resultant frequency photo-responses of the 600 μm long PD are shown in Fig. 6(a) for different applied bias voltages. The 3-dB bandwidths (f3dB) for three different active lengths of 200, 400, and 600 μm are plotted in Fig. 6(b). The maximum achievable f3dB are 28 and 18 GHz at −6 V for 200 and 400 μm long devices, respectively, and 14 GHz at −7 V for a 600 μm long PD.
To study the effects of device size and applied bias on the frequency response, we numerically calculated the RC- (resistance-capacitance), transit-, and avalanche build-up-limited f3dB with different reverse bias conditions as shown in Fig. 6(c). The applied reverse bias voltage can reduce both the capacitance (decrease in RC delay time) and carrier transit time (increase in carrier drift velocity) across the PN junctions. As a result, the RC- and transit-limited f3dB increase with the reverse bias voltage. In contrast, the avalanche multiplication processes, accelerated by applying the reverse bias, can significantly impair the carrier build-up time and therefore downgrade the build-up-limited f3dB. Around the voltage ranging from −6 to −8 V, the build-up-limited f3dB becomes comparable to the RC- and transit-limited f3dB. For the 600 μm long PD, the RC-limited f3dB are predominant factor determining the total f3dB, as shown in the blue curve of the Fig. 6(d). The total f3dB for 200 and 400 μm long PDs mainly depend on the transit-limited f3dB and build-up-limited f3dB at the low bias range (from 0 to −6 V) and high bias range (from −6 to −8 V), respectively.
The predicted peak f3dB (estimated to be 26 and 19 GHz) for the 400 and 600 μm long PDs are positioned at −6 and −7 V, respectively, which agree well with the measured ones. The predictions give the peak f3dB of 39 GHz for the 200 μm long PD whereas the measured f3dB is only 28 GHz. We believe such deviations between the predicted and measured values of f3dB are originated from the parasitic impedance effects, which can affect the PD’s actual operation speed.
We have demonstrated efficient sub-bandgap NIR photo-detection in silicon PN-doped waveguides. The photon-assisted tunneling process at the PN-junction depletion region has been described by a simplified theoretical model to qualitatively explain the effects of various parameters (doping density, incident wavelength, applied bias voltage) on the sub-bandgap photon absorption process. With sufficient reverse bias voltages, enhanced sub-bandgap photon absorption and carrier multiplication result in a large photo-responsive gain capable of realizing efficient NIR photo-detection at the silicon rib waveguide-based photodetector. Our approaches can open up an efficient way to build precise in-line power monitors due to their predictable and controllable intrinsic absorption processes. We believe that the performance of the suggested PD structures can be further improved by optimizing the doping profiles and employing microring-type resonant cavities.
This research was supported by the National Research Civil-Military Technology Cooperation Program and the Center for Integrated Smart Sensors funded by the Ministry of Science, ICT and Future Planning as Global Frontier Project (CISS-2012M3A6A6054191).
References and links
1. J. Michel, J. Liu, and L. C. Kimerling, “High-performance Ge-on-Si photodetectors,” Nat. Photonics 4(8), 527–534 (2010). [CrossRef]
3. L. Vivien, A. Polzer, D. Marris-Morini, J. Osmond, J. M. Hartmann, P. Crozat, E. Cassan, C. Kopp, H. Zimmermann, and J. M. Fédéli, “Zero-bias 40Gbit/s germanium waveguide photodetector on silicon,” Opt. Express 20(2), 1096–1101 (2012). [CrossRef] [PubMed]
4. G. Masini, S. Sahni, G. Capellini, J. Witzens, and C. Gunn, “High-speed near infrared optical receivers based on Ge waveguide photodetectors integrated in a CMOS process,” Adv. Opt. Technol. 2008, 196572 (2008). [CrossRef]
5. H. Zhu, L. Zhou, X. Sun, Y. Zhou, X. Li, and J. Chen, “On-chip optical power monitor using periodically interleaved PN Junctions integrated on a silicon waveguide,” IEEE J. Sel. Top. Quantum Electron. 20(4), 1–8 (2014).
6. J. P. Mailoa, A. J. Akey, C. B. Simmons, D. Hutchinson, J. Mathews, J. T. Sullivan, D. Recht, M. T. Winkler, J. S. Williams, J. M. Warrender, P. D. Persans, M. J. Aziz, and T. Buonassisi, “Room-temperature sub-band gap optoelectronic response of hyperdoped silicon,” Nat. Commun. 5(3011), 3011 (2014). [PubMed]
7. A. P. Knights, J. D. Bradley, S. H. Gou, and P. E. Jessop, “Silicon-on-insulator waveguide photodetector with self-ion-implantation-engineered-enhanced infrared response,” J. Vac. Sci. Technol. A 24(3), 783–786 (2006). [CrossRef]
8. S. Park, K. Yamada, T. Tsuchizawa, T. Watanabe, H. Nishi, H. Shinojima, and S. Itabashi, “All-silicon and in-line integration of variable optical attenuators and photodetectors based on submicrometer rib waveguides,” Opt. Express 18(15), 15303–15310 (2010). [CrossRef] [PubMed]
9. M. W. Geis, S. J. Spector, M. E. Grein, R. J. Schulein, J. U. Yoon, D. M. Lennon, C. M. Wynn, S. T. Palmacci, F. Gan, F. X. Käertner, and T. M. Lyszczarz, “All silicon infrared photodiodes: photo response and effects of processing temperature,” Opt. Express 15(25), 16886–16895 (2007). [CrossRef] [PubMed]
10. M. W. Geis, S. J. Spector, M. E. Grein, J. U. Yoon, D. M. Lennon, and T. M. Lyszczarz, “Silicon waveguide infrared photodiodes with >35 GHz bandwidth and phototransistors with 50 AW-1 response,” Opt. Express 17(7), 5193–5204 (2009). [CrossRef] [PubMed]
11. R. R. Grote, K. Padmaraju, B. Souhan, J. B. Driscoll, K. Bergman, and R. Osgood Jr., “10 Gb/s error-free operation of all-silicon ion-implanted-waveguide photodiodes at 1.55,” IEEE Photonics Technol. Lett. 25(1), 67–70 (2013). [CrossRef]
12. J. J. Ackert, A. S. Karar, D. J. Paez, P. E. Jessop, J. C. Cartledge, and A. P. Knights, “10 Gbps silicon waveguide-integrated infrared avalanche photodiode,” Opt. Express 21(17), 19530–19537 (2013). [CrossRef] [PubMed]
14. Y. Li, S. Feng, Y. Zhang, and A. W. Poon, “Sub-bandgap linear-absorption-based photodetectors in avalanche mode in PN-diode-integrated silicon microring resonators,” Opt. Lett. 38(23), 5200–5203 (2013). [CrossRef] [PubMed]
15. H. Zhu, L. Zhou, Y. Zhou, Q. Wu, X. Li, and J. Chen, “All-silicon waveguide avalanche photodetectors with ultrahigh gain-bandwidth product and low breakdown voltage,” IEEE J. Sel. Top. Quantum Electron. 20(6), 3803006 (2014).
16. J. J. Ackert, D. J. Thomson, L. Shen, A. C. Peacock, P. E. Jessop, G. T. Reed, G. Z. Mashanovich, and A. P. Knights, “High-speed detection at two micrometres with monolithic silicon photodiodes,” Nat. Photonics 9(6), 393–396 (2015). [CrossRef]
17. H. Chen and A. W. Poon, “Two-photon absorption photocurrent in pin diode embedded silicon microdisk resonators,” Appl. Phys. Lett. 96(19), 191106 (2010). [CrossRef]
20. Y.-H. Liu, Y. Zhou, and Y.-H. Lo, “High efficiency silicon 1310 nm detector without defect states or heteroepitaxy,” Appl. Phys. Lett. 103(4), 041119 (2013). [CrossRef]
21. N. A. Sobolev, A. M. Emel’yanov, E. I. Shek, and V. I. Vdovin, “Extended structural defects and their influence on the electroluminescence in efficient Si light-emitting diodes,” Physica B 340-342, 1031–1035 (2003). [CrossRef]
22. M. A. Lourenço, M. Milosavljević, G. Shao, R. M. Gwilliam, and K. P. Homewood, “Boron engineered dislocation loops for efficient room temperature silicon light emitting diodes,” Thin Solid Films 504(1), 36–40 (2006). [CrossRef]
23. R. A. Soref and B. R. Bennett, “Electrooptical effects in silicon,” IEEE J. Quantum Electron. 23(1), 123–129 (1987). [CrossRef]
24. M. A. Green, “Self-consistent optical parameters of intrinsic silicon at 300K including temperature coefficients,” Sol. Energy Mater. Sol. Cells 92(11), 1305–1310 (2008). [CrossRef]
25. M. A. Green and M. J. Keevers, “Optical properties of intrinsic silicon at 300 K,” Prog. Photovolt. Res. Appl. 3(3), 189–192 (1995). [CrossRef]
26. P. Wendland and M. Chester, “Electric field effects on indirect optical transitions in silicon,” Phys. Rev. 140(4A), A1384–A1390 (1965). [CrossRef]
27. S. S. Li, Semiconductor Physical Electronics (Springer, 2012).