We propose and demonstrate active resonance wavelength stabilization for silicon microring resonators with an in-resonator defect-state-absorption (DSA)-based photodetector (PD) for optical interconnects. We integrate an electro-optic (EO) tuner and a thermo-optic (TO) tuner on the microring, which are both feedback-controlled following a photocurrent threshold-detection method. Our BF2-ion-implanted DSA-based PIN PD exhibits a cavity-enhanced sub-bandgap responsivity at 1550 nm of 3.3 mA/W upon −2 V, which is 550-fold higher than that exhibited by an unimplanted PIN diode integrated on the same microring. Our experiment reveals active stabilization of the resonance wavelength within a tolerance of 0.07 nm upon a step increment of the stage temperature by 7 °C. Upon temperature modulations between 23 °C and 32 °C and between 18 °C and 23 °C, the actively stabilized resonance exhibits a transmission power fluctuation within 2 dB. We observe open eye diagrams at a data transmission rate of up to 30 Gb/s under the temperature modulations.
© 2015 Optical Society of America
Silicon photonics leveraging complementary metal-oxide-semiconductor (CMOS) technology has now found high-impact technological applications in optical communications and interconnections for data centers and high-performance computing . In many of these demonstrated devices, silicon microring resonators are adopted as one of the constituent functional elements due to their sharp resonances for wavelength selectivity, compact size, accessibility by integrated waveguides and tunability by way of thermo-optical (TO) and electro-optical (EO) mechanisms.
However, one major hurdle to practical utilization of silicon microrings for optical interconnects is that the microring resonance wavelengths are extremely sensitive to not only fabrication-induced imperfections that can be trimmed by post-processing methods , but also dynamic and random variations of the photonic chip operational conditions such as the chip temperature and the laser carrier wavelength. The resulting dynamic misalignment between the resonance and the carrier wavelength can significantly compromise the microring performance.
In order to address the issue, several research groups have recently developed various methods to actively stabilize or lock the silicon microring resonance wavelength [3–5] (see Table 1 in Sec. 5 for performance comparisons). Notably, three methods have been demonstrated including (i) the dithering signal detection method by Padmaraju et al.  using an integrated off-resonator defect-state-absorption (DSA)-based photodetector (PD) and an integrated TO tuner for microring switches/filters, (ii) the balanced homodyne detection method by Cox et al.  using an integrated off-resonator Ge PD and an integrated TO tuner for microring modulators/filters, and (iii) the threshold-detection method by Zheng et al.  using an integrated off-resonator Ge PD and an integrated TO tuner for microring modulators. All of these three methods utilize an off-resonator PD as a monitor and only TO effect to actively realign the resonance. However, the use of an off-resonator PD potentially imposes additional footprints for the monitor, especially for DSA-based PDs that typically require hundreds of μm length , and cannot be readily scaled for monitoring individual microrings in a large-scale-integrated photonic network or a switch fabric [1,6]. Although the use of an integrated TO tuner offers a large resonance wavelength tuning range, it requires a pre-biased voltage even at room temperature (a pre-heated tuner to the maximum expected temperature in the environment) in order to compensate for the temperature rising effect by reducing the biased voltage [4,5].
Previously [7,8], our research group has proposed a method to actively stabilize the resonance wavelength of a silicon microring using an in-resonator DSA-based PD  as a monitor, and using both integrated EO and TO tuners to actively realign the resonance wavelength following a photocurrent threshold-detection method  or a photocurrent slope-detection method . The use of an in-resonator PD offers the key merits of compactness and scalability. Besides, with cavity enhancement of the on-resonance optical power in the microring, the PD length can be shorter than the off-resonator counterpart. The use of both integrated EO and TO tuners enables resonance realignments upon either a temperature rise or fall. It also enables resonance realignments upon a carrier wavelength blueshift or redshift.
In this paper, we report the experimental demonstration of our active resonance wavelength stabilization scheme using a photocurrent threshold-detection method. We introduce defect states into the in-resonator PD by a BF2-ion-implantation process to enhance the responsivity at 1550 nm. We detail the analysis and parameters of our device and scheme.
Figure 1(a) schematically shows our proposed scheme. For an in-resonator PD, the detected photocurrent lineshape follows the drop-port transmission lineshape. Therefore, we can use the photocurrent value to detect the alignment of a resonance with a carrier wavelength λo. A resonance aligns with λo gives a photocurrent Id at a maximum value of Ip. When the resonance wavelength shifts away from λo due to either a temperature rise or fall, Id drops. Once Id falls below a pre-set threshold value of Ith, it triggers the microprocessor to output a voltage (VEO or VTO) to either the EO or TO tuner for realigning the resonance toward λo.
Figure 1(b) schematically shows the top-view of our device structure and the feedback-control circuit. The device comprises a silicon microring resonator coupled with a waveguide crossing. We have adopted this particular configuration as it constitutes a switch element in a switch fabric . We integrate two identically designed PIN diodes along the microring, with one serving as the monitor (i) and the other as the EO tuner (ii), each occupying one quarter of the microring. In order to enhance the sub-bandgap photocurrent generation for the monitor, we implant only the intrinsic waveguide region of the PIN PD with BF2 ions to form defect states (inset i), while that of the EO tuner is left unimplanted (inset ii). The energy band diagram illustrates the DSA in silicon for photons at 1550 nm (0.8 eV). In order to minimize the electrical crosstalk between the two diodes, we electrically isolate the two diodes by integrating in between two leakage blocks in the form of a p--doped wire with a 0.5μm width and a few-μm long (inset iii). The TO tuner comprises an open-ring-shaped p--doped resistor sandwiched by an inner ring-shaped p+-doped region and an outer open-ring-shaped p+-doped region for Ohmic contacts (inset iv).
For a proof-of-concept demonstration, we use an off-the-shelf microprocessor as the basic logic-control unit, with an analog input port to read Id, an analog output port to output VEO, and digital outputs to output VTO through an external D/A converter.
Figure 2 illustrates the working principle. The scheme requires a prior knowledge of a peak photocurrent value Ip at λo under the conditions that the resonance is aligned with λo (black lines in Figs. 2(a) and 2(b)), and that the laser power coupled to the chip is fixed at a certain level. Then we set Ith at λo, corresponding to a red- or blue-shifted resonance at a threshold wavelength shift λth or λth’. Once the resonance wavelength is shifted beyond λth or λth’, Id drops below Ith and triggers the microprocessor to output a voltage step ΔVEO or ΔVTO to the EO or TO tuner, with a total VEO or VTO to realign the resonance wavelength toward λo. During each operation loop of the microprocessor, the resonance realignment can only approach λo by a fixed step ΔλEO/TO.
Here we consider three representative cases to illustrate the realignment processes. In the near-threshold limit that Id is marginally below Ith, the shifted resonance wavelength is marginally beyond λth or λth’ (yellow line in Fig. 2(a), blue line in Fig. 2(b)). If we set ΔVEO or ΔVTO to tune the resonance wavelength by a ΔλEO = (λth − λo) or ΔλTO = (λo − λth’), we can obtain a nearly perfect realignment with λo. Once Id is above Ith, the microprocessor locks VEO and VTO.
In the case that Id is below Ith and Id ≥ (2Ith − Ip), the shifted resonance wavelength is beyond λth or λth’. Given a fixed ΔVEO or ΔVTO that yields ΔλEO/TO, we can realign the resonance toward λo within (λth − λth’) in a single operation loop of the microprocessor. In the case that Id is significantly below Ith and Id < (2Ith − Ip), with the shifted resonance wavelength far beyond λth or λth’ (red line in Fig. 2(a), purple line in Fig. 2(b)), we can still realign the resonance toward λo within (λth − λth’), albeit in multiple operation loops.
The range for resonance realignment is partially limited by the EO tuning range, ΔλEOmax, and the TO tuning range, ΔλTOmax. The resonance red-shifting range that can be realigned by EO tuning is given by Δλred = ΔλEOmax + (λth − λo). Similarly, the resonance blue-shifting range is given by Δλblue = ΔλTOmax + (λo − λth’).
The choice of Ith determines the resonance stabilization range. A higher Ith gives a narrower stabilization range. However, as the EO tuner introduces free-carrier absorption (FCA) loss, the realigned resonance at λo only generates a reduced peak photocurrent value of Ip’, which depends on ΔλEOmax. Therefore, we should set Ith < Ip’.
Here we evaluate as an example the EO tuning scheme assuming the device dimensions following the fabricated devices (see Sec. 3). We use the transfer matrix modeling to calculate the drop-transmission spectrum and the corresponding photocurrent spectrum . We assume a square microring with four 90° arcs of a radius of 13 μm and four straight waveguides with a length of 24 μm. We assume the EO tuner length (LEO) is a quarter of the microring perimeter (Ltotal ≈177 μm). We adopt an effective refractive index neff = 2.65 at 1550 nm, assuming a silicon-on-insulator (SOI) rib waveguide and calculated according to beam-propagation method (BPM). We adopt a silicon waveguide propagation loss of 2 dB/cm and an in-resonator PD responsivity of 10 mA/W [10,11]. We choose the symmetric input- and output-coupling coefficients to be 0.6 in order to obtain a Lorentzian resonance linewidth of 0.42 nm to support a 25Gb/s data transmission.
The EO tuning range is given by Eq. (1), we therefore require a Δneff to be ~8 × 10−3. Assuming Δneff is approximately given by the free-carrier plasma dispersion Δn in silicon according to the Soref’s and Bennett’s equations , we calculate the corresponding injected electron concentration ΔNe and hole concentration ΔNh to be ~2.5 × 1018 cm−3. This gives a FCA loss Δα of ~155 dB/cm.
Figure 3 depicts the modeled drop-transmission and corresponding photocurrent spectra without (black lines) and with (blue lines) carrier injection. Our modeling suggests a 1.6dB drop-port loss and a 72% photocurrent drop upon a ΔλEOmax = 0.8 nm. In order to compensate for a temperature rise of 10 K, the maximum Ith = 72% Ip. By setting Ith = 72% Ip, we obtain a (λth − λo) = 0.14 nm for this particular resonance profile.
Figure 4 schematically depicts the algorithm of the threshold-detection method using both EO and TO tuners. The microprocessor continuously reads in an Id value during every operation loop and compares the Id value with Ith. When Id ≥ Ith, the microprocessor maintains the values of VEO and VTO. When Id < Ith, with VEO = 0 and VTO = 0 initially, the microprocessor outputs a temporary VEO beyond the diode turn-on voltage and reads in Id again within the same operation loop in order to determine whether to turn on the EO or TO tuner. If Id increases, the EO tuner remains on. If Id decreases, the microprocessor turns off the EO tuner and turns on the TO tuner. When Id < Ith, with either VEO ≠ 0 or VTO ≠ 0, the microprocessor continues to tune the voltage without the decision step until Id ≥ Ith.
3. Device fabrication
Figure 5(a) shows the top-view scanning-electron micrograph of the fabricated device. The square microring resonator has the aforementioned dimensions. The silicon waveguide has a width of ~450 nm, a height of ~180 nm and a slab thickness of ~70 nm on a 3μm-thick buried-oxide layer. Both the PIN-based monitor and EO tuner have a length of 40 μm and are separated with an end-to-end distance of 16 μm. The two leakage block structures are integrated in the middle of the separation, with a width of 1 μm and a length of 8 μm (for the one next to the waveguide crossing) and 10 μm. The TO tuner is integrated inside the microring with an edge-to-edge distance of 3.5 μm to the microring waveguide.
We fabricate the microrings following the standard CMOS nanoelectronics fabrication processes using i-line (365 nm) photolithography and reactive ion etching. The doping concentration of the p+ and n+ regions in the PIN is ~1 × 1020 cm−3. The doping concentration of the p- implantation region in the TO tuner is ~1 × 1017 cm−3. After the low-temperature oxide (LTO) deposition and the metal pad formation, we open an implantation window on top of the waveguide region of the PIN-based monitor. We implant BF2 ions with a doping concentration of ~1 × 1018 cm−3 in order to form the defect states. The chip finally undergoes the photoresist stripping for 30 minutes at a temperature of 170 °C, which anneals out some of the defect states. As the annealing temperature is below 300 °C, we believe both the divacancies and interstitial clusters, which contribute to sub-bandgap absorption and photocurrent generation at 1550 nm [10,11], are not totally annealed out.
4. Experimental characterization
4.1 Device characterization
Figure 5(b) shows the measured transverse-electric (TE)-polarized drop-transmission spectrum of the microring resonator, and the simultaneously measured photocurrent spectrum from the diode monitor, both measured under a reverse bias voltage of 2 V across the monitor. The transmission spectrum reveals a resonance linewidth of ~0.45 nm, corresponding to a quality factor (Q) of ~3400. The extinction ratio (ER) is ~13 dB. The photocurrent spectrum (black line) shows a consistent resonance lineshape, with an ER of ~20 and a linewidth of ~0.43 nm. The dark current upon −2 V is ~5 nA. As a control, we also measure the photocurrent spectrum from the EO tuner (under −2 V bias), with a dark current of ~0.12 nA, integrated in the same microring resonator. The peak photocurrent value measured from the ion-implanted monitor is ~550-fold higher than that measured from the tuner.
Figure 5(c) shows the measured photocurrent values at resonance wavelength 1546 nm upon −1 V, −2 V and −3 V as a function of the estimated waveguide input power just before the microring. We estimate the waveguide input power using a lensed-fiber-to-chip-to-lensed-fiber setup by adding the fiber-out-coupled power by ~13 dB, assuming the coupling losses at the waveguide input and output facets are equal, and that the waveguide propagation losses before and after the microring are equal (with the microring situated near the middle of the bus waveguide). The linear fits (red lines) show the linearity of the DSA process upon a relatively low waveguide input power in the order of 0.1 mW. The fitted responsivity values at the resonance wavelength of 1546 nm are 2.7 mA/W, 3.3 mA/W and 3.9 mA/W upon −1 V, −2 V, and −3 V, respectively. With further optimization on the implantation and annealing conditions, the responsivity could be further increased to the order of 10 mA/W [10,11].
Figure 6(a) shows the measured drop-transmission and photocurrent spectra upon biasing the monitor at −2 V under different tuner conditions. At VEO = 1.2 V, the spectra (blue lines) are blue-tuned by ~0.58 nm, with a drop-transmission loss of ~0.8 dB, a drop in the Q factor from ~3400 to ~3100 and a corresponding photocurrent drop to ~83% of the original peak photocurrent. At VTO = 2.4 V, the spectra (red lines) are red-tuned by ~0.65 nm, without a noticeable drop-transmission loss or Q factor degradation or photocurrent drop.
Figure 6(b) shows the measured resonance wavelength blue-tuning as a function of VEO (black squares) and the corresponding electrical power consumption (red circles). At VEO = 1.2 V and a power consumption of ~2.0 mW, a blue-tuning of ~0.58 nm indicates an EO tuning efficiency of ~3.4 mW/nm. We note that the blue-tuning is not linear to VEO. Upon 1.45 V, the resonance has a ~0.8nm blue-shift, which can compensate for a ~10K temperature rise. This, however, approaches the saturation of the EO blue-tuning due to the Ohmic heat generation.
Figure 6(c) shows the measured resonance wavelength red-tuning as a function of VTO (black squares) and the corresponding electrical power consumption (red circles). At VTO = 2.4 V and a power consumption of ~6.6 mW, a red-tuning of ~0.65 nm indicates a TO tuning efficiency of ~10.1 mW/nm. The red-tuning is not linear to VTO.
4.2 Active resonance wavelength stabilization upon a temperature rise
In order to demonstrate the stabilization scheme upon an external temperature variation, we attach the fabricated chip on top of a thermal-electric cooler (TEC) using silver paste. We use three high-speed radio-frequency probes to simultaneously measure the photocurrent from the monitor and to apply voltages to the EO and TO tuners. We use a 100MHz microprocessor (mbed LPC1768) as a logic-control unit. Here, we raise the waveguide input power to a few mW, while maintaining a linear photoresponse  in order to enable a sufficiently large photocurrent signal (> 10 mV) for the microprocessor. The photocurrent is read across a 2kΩ resistor into the 12-bit analog-in port of the microprocessor. The EO tuner is connected to the 10-bit analog-out port of the microprocessor, as the EO tuner requires a fine voltage step. The TO tuner is connected to the 4 digital outputs of the microprocessor through a 4-bit external D/A converter (see Fig. 1(a)).
Figure 7 shows the measured drop-transmission intensity (Fig. 7(b)) and photocurrent (Fig. 7(c)) at 1546 nm (initially aligned to a resonance) over time, with a step increment of the TEC stage temperature from the room temperature at 24 °C to 31 °C (Fig. 7(a)). At the room temperature, the optical transmission intensity is ~-22.9 dBm and the photocurrent is ~29 μA. At 33 s, we apply a step-increment voltage to the TEC and gradually raise the temperature. Without the feedback control, the transmission intensity drops by ~5.3 dB, following the blue side of the resonance lineshape (inset of Fig. 7(b)), and the photocurrent drops to ~5 μA. At 100 s, we turn on the feedback control, and the transmission intensity rises to ~-23.6 dBm (~0.7 dB below the initial transmission), with the photocurrent increases to ~25 μA (~85% Ip). Here, we set Ith = 85% Ip, assuming a VEOmax of 1.15 V. The Ith value corresponds to a (λth − λo) = 0.07 nm (inset of Fig. 7(c)). We set ΔVEO = 0.05 V in order to obtain a ΔλEO of 0.06~0.09 nm (see Fig. 6(b)).
We examine the drop-transmission intensity upon various choices of Ith values. The lower the Ith value, the larger the (λth – λo) or (λo - λth’), thus the lower the transmission intensity at λo upon feedback control. Figure 8 shows the measured drop-transmission intensity (Fig. 8(b)) and photocurrent variation (Fig. 8(c)) at 1546 nm over time upon a step raised temperature at 31 °C. The feedback control is switched on and off alternatively, and the Ith values are switched from 90% Ip to 60% Ip in a step of 5% Ip in between intervals of ~50 s without applying the feedback control (Fig. 8(a)). We set ΔVEO = 0.05 V. At the room temperature, we measure a transmission intensity of ~-23 dBm and a corresponding photocurrent of ~30 μA. When the temperature is steady at 31 °C during the intervals without the feedback control, we observe a significantly reduced transmission intensity near ~-28 dBm and a photocurrent value of ~5 μA. With the feedback control turned on during the intervals using various Ith values in decreasing steps, we observe the transmission intensity decreasing in steps from ~-23.7 dBm to ~-25 dBm, while the photocurrent decreasing in steps from ~25 μA to ~16 μA.
We also examine the active resonance wavelength stabilization with modulated external temperature variations. We apply a 5mHz square-wave function (Fig. 9(a) upper curve) and a 5mHz sine-wave function (Fig. 9(b) upper curve) to the TEC, which offer temperature modulation between the room temperature (23 °C) and 32 °C. Upon the square-wave temperature modulations and without the feedback control before ~350 s, the measured drop-transmission intensity at 1546 nm (Fig. 9(a) lower curve) varies between ~-24.5 dBm and ~‑30.5 dBm. With the feedback control (Ith = 80% Ip, ΔVEO = 0.05 V), the measured drop-transmission intensity only varies between ~-24.5 dBm and ~-26.5 dBm. We attribute the ~2dB intensity fluctuation upon the feedback control to the relatively large temperature modulation amplitude close to the range allowed by our ΔλEOmax (~0.58 nm). Similarly, upon the sine-wave temperature modulation and with feedback control, the drop-transmission intensity fluctuates within 2 dB.
4.3 Active resonance wavelength stabilization upon a temperature fall
Figure 10(a) shows the measured drop-transmission intensity (Fig. 10(a) lower curve) at 1546 nm over time, with a step drop of the TEC stage temperature from the room temperature at 23 °C to 18 °C (Fig. 10(a) upper curve). At the room temperature, the optical transmission intensity is ~-24 dBm. At 27 s, we apply a step-drop voltage to the TEC and gradually drop the temperature. Without the feedback control, the transmission intensity drops by ~7 dB, following the red side of the resonance lineshape. At 40 s, we turn on the feedback control, and the transmission intensity rises to ~-24 dBm. Inset shows a zoom-in view of the transition, with a time resolution of ~50 ms limited by our data acquisition system. Here, we set Ith = 85% Ip. We set ΔVTO = 0.3 V in order to obtain a ΔλTO of 0.06 ~0.09 nm (see Fig. 6(c)).
Figure 10(b) shows the measured drop-transmission intensity modulation (Fig. 10(b) lower curve) at 1546 nm upon a square-wave modulated temperature cooling process at 10 mHz between 23 °C and 18 °C. Without the feedback control, the transmission power varies between ~-23.0 dBm and ~-28.4 dBm. With the feedback control (Ith = 80% Ip, ΔVTO = 0.3 V) turned on at 230 s, the transmission power varies between ~-23.0 dBm and ~-25.0 dBm.
4.4 Active resonance wavelength stabilization upon a data transmission
We characterize the stabilization scheme upon a 10mHz square-wave temperature modulation, while transmitting data at a rate ranging from 10 to 30 Gb/s. We use a 40Gb/s modulator to encode the pseudorandom bit sequence (PRBS) signal with a pattern length of 231-1. In order to measure the eye diagrams, we increase the waveguide input power to ~15 mW, while maintaining the dominance of the linear photoresponse.
Figure 11 shows the measured eye-diagrams at 20 Gb/s (Figs. 11(a) and 11(b)) and 30 Gb/s (Figs. 11(c) and 11(d)) at 1546 nm during a period of the temperature modulation between 23.5 °C and 32 °C. Without the feedback control, the measured eye-diagrams show a wide noise band and a large ER variation of ~1 dB - ~2.5 dB (Figs. 11(a) and 11(c)). With the feedback control (Ith = 80% Ip, ΔVEO = 0.05 V), we observe clear open eye diagrams with an ER variation of ~0.3 dB (Figs. 11(b) and 11(d)).
Figure 12 shows the measured eye-diagrams at 10 Gb/s (Figs. 12 (a) and 12(b)) and 30 Gb/s (Figs. 12(c) and 12(d)) at 1546 nm during a period of the temperature modulation between 23.5 °C and 18 °C. Without the feedback control, the measured eye-diagrams show a wide noise band and a large ER variation of ~2 dB - ~3.5 dB (Figs. 12(a) and 12(c)). With the feedback control (Ith = 80% Ip, ΔVTO = 0.3 V), we observe clear open eye diagrams with an ER variation of ~0.3 dB - ~0.6 dB (Figs. 12(b) and 12(d)).
5. Discussion and Summary
Here we benchmark our work with other active resonance wavelength stabilization schemes that have been recently demonstrated, as shown in Table 1.
In terms of the control range, the prior arts demonstrated a temperature range from 5 K (in , Padmaraju et al.) to 55 K (in , Cox et al.) under TO-only tuning schemes. While in this work, our total control range is from 18 °C to 32 °C under both EO and TO tuning. We believe the EO tuning range is difficult to compensate for a temperature rise exceeding ~10 K, given the Ohmic heat generated during carrier injection can saturate the blue-tuning and the FCA can reduce both the transmission and the Q factor of the microring. Whereas in principle, the TO tuning can be extended to compensate for a temperature drop of few tens of degree.
In terms of the additional power consumption, our EO tuner consumes a maximum of ~0.2 mW/K and our TO tuner consumes a maximum of ~1.32 mW/K within the control range. This is comparable with the power consumption of ~0.24 mW/K in , Zheng et al. using only TO tuners. We attribute our relatively large TO-tuner power consumption to the wide edge-to-edge separation of 3.5 μm between the tuner and the microring.
In terms of the temperature change speed limit within which the stabilization schemes can respond to, the literature values are ~1 ms (in , Padmaraju et al. and in , Zheng et al.). Limited by our current data acquisition speed, our minimum measurement time step is only ~50 ms. As we have shown that the system can be stabilized within a measurement time step, we believe our feedback-control scheme can respond to a temperature change within 50 ms.
However, one shortcoming of our scheme is that it is sensitive to the carrier optical power fluctuation. A drop in the carrier power can cause Id to fall below Ith even upon a perfect alignment. Thus, the resonance wavelength tuning cannot return Id above Ith. Besides, the defects in the DSA-based monitor may be annealed out at an elevated temperature over a long period of time, resulting in a decreased responsivity. We are currently studying the DSA-based monitor under various annealing temperature and time conditions.
We remark that in order to apply the proposed scheme to a high-speed silicon microring modulator, an additional PIN/PN diode for carrier injection/depletion needs to be integrated along the remaining part of the microring. Given the limited room available along the microring, the additional diode for modulation could only occupy less than half of the microring. Therefore, although it is possible to implement a high-speed silicon microring modulator with the proposed scheme, we expect a certain performance penalty that may limit its utilization.
In summary, we have proposed and demonstrated an active resonance wavelength stabilization scheme for silicon microring resonators integrated with a DSA-based PD, an EO tuner and a TO tuner. Our in-resonator BF2-ion-implanted PD exhibited a cavity-enhanced sub-bandgap responsivity at 1550 nm of ~3.3 mA/W upon −2 V, which is ~550-fold higher than that exhibited by an unimplanted PIN diode integrated on the same microring. Our proof-of-concept experiment reveals active stabilization of the resonance wavelength within a tolerance of ~0.07 nm upon a step increment of the stage temperature by ~7 °C. Upon temperature modulations between 23 °C and 32 °C and between 18 °C and 23 °C, our feedback-controlled transmission power exhibited a fluctuation within ~2 dB. We have observed open eye diagrams at a data transmission rate of up to 30 Gb/s under the temperature modulations. We therefore envision that such an active resonance wavelength stabilization method can enable practical utilization of silicon microring resonators, and the method can in principle be scaled for large-scale-integrated photonic networks using silicon microrings.
The work is supported in part by the Innovation and Technology Fund of the Hong Kong Special Administrative Region (HKSAR) under project no. ITS/087/13, the Proof-of-Concept Fund (PCF) of The Hong Kong University of Science and Technology (HKUST) under project no. PCF007.12/13, and the General Research Fund (GRF) of the HKSAR under project no. 16208114. The authors gratefully acknowledge the Nanoelectronics Fabrication Facility (NFF) of HKUST for the device fabrication. The authors also gratefully acknowledge the microprocessor sponsored by the ARM Corporation.
References and links
1. P. Dong, Y. K. Chen, T. Gu, L. L. Buhl, D. T. Neilson, and J. H. Sinsky, “Reconfigurable 100Gb/s Silicon Photonic Network-on-Chip,” J. Opt. Commun. Netw. 7(1), A37–A43 (2015). [CrossRef]
3. K. Padmaraju, D. F. Logan, T. Shiraishi, J. J. Ackert, A. P. Knights, and K. Bergman, “Wavelength locking and thermally stabilizing microring resonators using dithering signals,” J. Lightwave Technol. 32(3), 505–512 (2014). [CrossRef]
4. J. A. Cox, A. L. Lentine, D. C. Trotter, and A. L. Starbuck, “Control of integrated micro-resonator wavelength via balanced homodyne locking,” Opt. Express 22(9), 11279–11289 (2014). [PubMed]
5. X. Zheng, E. Chang, P. Amberg, I. Shubin, J. Lexau, F. Liu, H. Thacker, S. S. Djordjevic, S. Lin, Y. Luo, J. Yao, J. H. Lee, K. Raj, R. Ho, J. E. Cunningham, and A. V. Krishnamoorthy, “A high-speed, tunable silicon photonic ring modulator integrated with ultra-efficient active wavelength control,” Opt. Express 22(10), 12628–12633 (2014). [CrossRef] [PubMed]
6. A. W. Poon, X. Luo, F. Xu, and H. Chen, “Cascaded microresonator-based matrix switch for silicon on-chip optical interconnection,” Proc. IEEE 97(7), 1216–1238 (2009). [CrossRef]
7. Y. Li, S. Feng, Y. Zhang, and A. W. Poon, “In-microresonator linear-absorption-based real-time photocurrent-monitoring and tuning with closed-loop control for silicon microresonators,” (US patent application 14/057,679 filed on the 18th, Oct. 2013).
8. Y. Zhang, Y. Li, S. Feng, and A. W. Poon, “Towards adaptively tuned silicon microring resonators for optical networks-on-chip applications,” IEEE J. Sel. Top. Quantum Electron. 20(4), 136 (2014). [CrossRef]
9. 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]
10. D. Logan, K. Murray, J. Ackert, P. Velha, M. Sorel, R. M. De La Rue, P. Jessop, and A. Knights, “Analysis of resonance enhancement in defect-mediated silicon micro-ring photodiodes operating at 1550 nm,” J. Opt. 13(12), 125503 (2011). [CrossRef]
11. 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]
12. R. A. Soref and B. R. Bennett, “Electrooptical effects in silicon,” IEEE J. Quantum Electron. 23(1), 123–129 (1987). [CrossRef]