A high-efficiency InGaAsP Mach-Zehnder modulator is integrated with hydrogen-free deuterated silicon nitride (SiN:D) waveguide circuits on a Si substrate. A thin InP-based layer provides a high optical confinement factor of around 50% and enables easy optical coupling to the SiN:D waveguides, which are fabricated by a low-temperature backend process. The fabricated Mach-Zehnder modulator with a 300-μm-long phase shifter shows a VπL of 0.4 Vcm, insertion loss of ~4.5 dB, and an error-free operation for 40-Gbit/s non-return-to-zero signal.
© 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
To support the sustainable growth of future optical fiber network systems, it is critically important to increase the transmission capacity of the optical transceivers while reducing their cost, size, and power consumption. A Mach-Zehnder modulator (MZM) is one of the key components for large-capacity optical fiber links, because it enables advanced modulation formats for modulating both the phase and intensity of optical signals. A promising technology for fabricating the low-cost MZMs is Si photonics. On the Si platform, compact and high performance Si and silicon nitride (SiN) waveguide circuits can be integrated with Si MZMs on large, low-cost Si wafers. Typically, high-speed Si MZMs have been demonstrated with a carrier-depletion p-n diode structure [1–3]. However, their VπLs are over 1 Vcm because of the poor carrier plasma effect of Si. Thus, although they have already reached their theoretical performance limit, their large power consumption still limits the reductions of the size and power consumption of the integrated optical transceivers.
To overcome the issue, an integration of III-V semiconductors on the Si platform is a promising solution. InP-based materials provide large electron-induced refractive index changes due to their large carrier plasma and band filling effects [4,5]. In addition, the electro-optical effects, including the Franz-keldysh (F-K) effect, quantum confined Stark effect (QCSE), and Pockels effect, also contribute to increasing the refractive index change further. In previous work, the InP-based carrier-depletion MZMs were integrated with the Si waveguide circuits [6,7]. Their phase shifters have vertical p-n diode structures, whose III-V layer thicknesses were around 2-μm. They showed very high modulation efficiency (VπL: 0.2~0.4 Vcm) using the III-V semiconductor multiple quantum well (MQW) layers. However, the large mismatch in the effective refractive indices between the thick III-V and ~220-nm-thick Si waveguide layers made it difficult to couple the III-V phase shifter to Si photonic circuits. The previous work avoided this problem by increasing the Si waveguide thickness to ~500 nm for effective refractive index matching [6,7], but the thick Si waveguide was not compatible with the widely developed thin Si waveguide circuits. Furthermore, the difference in the effective refractive indices becomes a more serious problem for SiN waveguides with a moderate refractive index, which have the potential to provide low-loss and high-performance waveguide circuits [8–10].
As a solution, we reduce the III-V layer thickness to reduce the effective refractive index. The thin (membrane) III-V layer provides efficient optical coupling to both the Si and SiN waveguide circuits [11–13]. In addition, the membrane layer is also suitable for improving the modulation efficiency due to its high optical confinement factor [12–14]. In this work, we used a 200-nm-thick III-V layer to demonstrate a membrane carrier-depletion type InGaAsP MZM integrated with low-loss hydrogen-free SiN waveguide circuits. The low-loss SiN waveguide is fabricated with the deuterated silane gas source at temperatures less than 300°C. With an optimized overlap between the optical mode field and the carrier distribution of the lateral p-n diode, the fabricated device shows small VπL (0.4 Vcm) and low insertion loss (4.5 dB). In addition, the fabricated device shows error-free operation for 40-Gbit/s non-return-to-zero (NRZ) signal.
Figure 1(a) and 1(b) show a top view of the MZM and cross sectional view of the phase shifter. The asymmetric MZM is designed with the SiN waveguides, and the difference in arm length was 178 μm. As shown in Fig. 1(b), the MZM has a 100-nm-thick InGaAsP core buried in the 200-nm-thick InP layer. The photoluminescence peak wavelength of the InGaAsP film is around 1.3 μm for operation in the C band. The InGaAsP core is mainly doped with donors, whose concentration is 1 x 1017 /cm3. The donor and acceptor regions with carrier concentrations of ~1018 /cm3 are formed in the membrane InP layer to form the lateral p-n diode for the carrier-depletion MZM. The 50-nm-thick InGaAs layers with carrier concentrations over 1 x 1019 /cm3 are formed above the InP layer for the ohmic contacts to the metal electrodes. By applying voltages to the electrodes, the lateral p-n diode is depleted. The phase of the propagated light is modulated by the carrier plasma effect, band-filling effect , and F-K effect. Figure 1(c) shows the calculated mode field pattern of the MZM. Thanks to the membrane structure, the optical confinement factor is around 50%, which is much higher than that of the previously reported thick III-V MZM on Si. Notably, this value is slightly smaller than that of the conventional Si carrier-depletion MZM. However, it is large enough to overcome the efficiency limit of the Si MZM, because the carrier-induced refractive index change of the n-type InGaAsP is much larger than that of Si . In addition, the F-K effect additionally contributes to increasing the refractive index change with reverse biases applied [12,13]. As a result, the membrane InGaAsP modulator can provide much higher modulation efficiency than that of conventional Si modulators. In this work, the phase shifter length was set to 300 μm, which is around ten times shorter than the lengths of the conventional Si carrier-depletion MZMs.
The membrane phase shifter has a smaller effective refractive index than those of the thick III-V semiconductor phase shifters. Therefore, it can be easily coupled to the SiN waveguide with an InP inverse taper. In this work, we used a 300-μm-long InP inverse taper for optical coupling between a 500-nm-wide InP nanowire waveguide and a 1.1-μm-width single-mode SiN waveguide . The width of the taper tip is around 100 nm. The narrow taper can be easily patterned because the membrane film provides a small aspect ratio of the taper tip cross section. The optical mode field of the 500-nm-wide InP waveguide is widened at the 1.5-μm-wide InP waveguide by using a second inverse taper with a length of 80 µm. Then, the 1.5-μm-wide InP waveguide is butt coupled to the InGaAsP core of the membrane phase shifter with the coupling loss of around 0.15 dB. Notably, although we set the taper length relatively long (300 μm) for large fabrication tolerance, it can be reduced further. Figure 1(d) shows the calculated relationship between the coupling efficiency and InP inverse taper length. Here, the width of InP taper tip was set to 100 nm. The calculated results indicate that the required taper length is less than only around 10 μm.
Figure 2 shows the fabrication procedure for the device. First, an InP/InGaAsP film is grown on an InP substrate. Here, the donors are doped in the InGaAsP layer. Then, the InP wafer is bonded to the SiO2 film on the Si wafer. After the InP substrate is removed [Fig. 2(a)], the n-InGaAsP film is patterned to form the core of the phase shifter [Fig. 2(b)]. Next, the InGaAsP core is buried in a thin intrinsic InP layer by a regrowth process at 600°C. The membrane film, which is thinner than the critical thickness of the InP layer on the Si substrate, enables us to perform on-Si regrowth process without degrading of the quality of the regrown film [12,15]. After that, a 50-nm-thick intrinsic InGaAs film is also regrown [Fig. 2(c)]. This is followed by the formation of the donor and acceptor regions by Si ion implantation and Zn thermal diffusion [Fig. 2(d)] [12,14]. The masks for the ion-implantation and Zn diffusions are photo-resist and SiO2, respectively. Then, the InGaAs and InP layers are patterned by wet etching for the contact [Fig. 2(e), (f)], followed by SiO2 deposition. The deposited SiO2 film is polished by chemical mechanical polishing. Finally, the metal electrodes [Fig. 2(g)] and SiN waveguides [Fig. 2(h)] are fabricated in the backend process.
One of the key processes is the low-temperature backend process for the low-loss SiN waveguide. In the backend process, the SiN films are typically formed by using a plasma-enhanced chemical vapor deposition (PECVD) method. However, the typical gas source of silane (SiH4) generates the large amount of hydrogen atoms, and thus N-H bonds form in the deposited SiN films. The large optical absorption of the N-H bonds in the C-band becomes a critical issue. Alternatively, we used hydrogen-free gas sources (SiD4 and N2) in an electron cyclotron resonance (ECR) PECVD method . With this technique, there are almost no N-H bonds in the deposited SiN:D film, so the deposited film is almost completely transparent in the entire C-band. Figure 3(a) shows the measured transmittance of a ~4-mm-long SiN:D waveguide fabricated at a temperature less than 300°C. As a reference, the transmittance of a typical 1.8-cm-long SiN waveguide, fabricated by using SiH4 gas, is also shown . The N-H bond absorption at the wavelength of around 1530 nm clearly disappears in the spectrum of the SiN:D waveguide. Typically, the propagation loss of an SiN:D waveguide is 0.55 dB/cm . By using the SiN:D waveguide, low-loss waveguide circuits can be integrated by the backend process without causing thermal damage to the III-V semiconductor layer. Figure 3(b) shows a cross-sectional scanning electron microscope (SEM) image of the membrane phase shifter of the MZM. The 600-nm-wide InGaAsP core is buried in the 200-nm-thick InP layer. The membrane III-V layer is bonded to the SiO2 film on the Si substrate.
4. Measured results and discussions
First, we measured the transmission spectrum of the fabricated MZM with DC voltages applied to the two arms. Here, the voltage at one arm was changed, while that at the other arm was fixed at 0 V. Figure 4(a) shows the measured transmission spectrum. The transmittance was normalized by that of the reference SiN waveguide. The peak of the spectrum shifted due to the optical phase shift. From the peak shifts, the VπL was estimated to be around 0.4 Vcm. This value is almost three times smaller than those of conventional Si carrier depletion type MZMs. Moreover, it is comparable to that of a vertical p-n diode MZM with a III-V semiconductor MQW layer (λg = 1.36 μm) , though our device has a bulk InGaAsP film (λg = 1.3 μm). Although the QCSE in the MQW layer is stronger than the F-K effect in the bulk layer, the high optical confinement factor in this work helps to improve the modulation efficiency. Figure 4(b) shows the wavelength dependence of the measured VπL. We evaluated the free-spectrum range and peak wavelength shift in wavelengths ranging from 1530 to 1570 nm. Then, we estimated the VπLs by the linear fitting of the phase shift for DC voltages from 0 to −4 V. Although the F-K effect contributes to the phase shift, the wavelength dependence of the measured VπL is small due to the large detuning of the material bandgap.
The measured insertion loss at DC bias of 0 V was around 4.5 dB, which is almost optimal for obtaining a small loss-efficiency product of the phase shifter. The key design parameter is p-gap, which is defined as the distance between the edge of the InGaAsP core and mask for the Zn diffusion, as shown in Fig. 5(a) . The p-gap determines the position of the p-n junction in the InGaAsP core. In this work, we used the p-gap of 0.4 μm. Positioning the p-n junction close to the center of the InGaAsP core is beneficial for high modulation efficiency, but doing so increases the concentration of the Zn atoms in the InGaAsP core. Thus, the optical loss increases due to the carrier-induced absorption in the p-type InGaAsP region. This is a fundamental tradeoff between VπL and insertion loss of the proposed MZM. In this work, we tuned the position of the p-n junction by changing the p-gap. Figure 5(b) shows the measured relationships between the measured VπL and insertion loss and the p-gap of the fabricated test patterns. With increasing p-gap, the p-n junction is positioned farther from the center of the InGaAsP core; therefore, VπL increases. However, the reduction of the insertion loss almost saturates for the p-gap of 0.4 μm, because the mode overlap to the p-type region becomes small. These results indicate that the smallest loss-efficiency product is obtained by using the p-gap of 0.4 μm. Here, the insertion loss of the MZM includes the losses of two MMIs, phase shifters, InP-SiN coupling, and InP-InGaAsP butt coupling. The measured insertion loss would be increased due to the fabrication errors. For example, the loss of two MMIs (2.8 dB) was very large because our lithography process has not been optimized yet.
Next, we measured the eye diagrams of the fabricated device. Figure 6(a) shows the measured eye diagrams at 28, 40, 56, and 64 Gbit/s. Here, the peak-to-peak voltage and DC voltages were 4 and −4 V (reverse bias), respectively. These RF signals were applied to one arm while the DC voltage at the other arm was fixed at 0 V. The wavelength of the input light was 1560.9 nm to obtain the largest extinction ratio at DC bias of −4 V. The input RF signals was NRZ signals, whose pattern length was pseudo-random bit sequence (PRBS) 231-1. Thanks to the high modulation efficiency, the eyes clearly opened even though the phase shifter length was only 300 μm. We measured the extinction ratio while changing the input data rate from 10 to 64 Gbit/s. Figure 6(b) shows the measured relationship between the extinction ratio and input data rate. The extinction ratio at high data rates would be limited by the RC bandwidth, which depends on the capacitance of the p-n junction . Nevertheless, the extinction ratio is larger than 9 dB up to the data rate of 40 Gbit/s.
Finally, we measured the back-to-back bit error rate (BER) of the fabricated device. Figure 7(a) shows the experimental setup. Continuous-wave light, whose wavelength was 1556.9 nm, was input from the tunable laser diode (TLD) to the fabricated chip in the transverse electric mode. The wavelength was set to minimize the detectable power of the receiver. The electrical signals were input from the pulse pattern generator (PPG) then fed into the linear amplifier. The DC voltage of −4 V was also applied through the bias tee. Here, the RF signals were applied to one arm of the MZM. The output light from the fabricated chip was amplified by the erbium doped fiber amplifier (EDFA) and then attenuated by the variable optical attenuator (VOA). The attenuated optical signals were input to the photodetector/trans-impedance-amplifier (PD/TIA) module. The detected electrical signals were fed into the BER tester. Figure 7(b) shows the measured BER curves for 25- and 40-Gbit/s NRZ signals (PRBS 27-1). Up to 40 Gbit/s, error-free operation (BER < 10−12) was achieved. The minimum detectable power at 40 Gbit/s is higher than that at 25 Gbit/s, because the extinction ratio of the modulated signals is reduced with increasing symbol rate, as shown in Fig. 6(b). Here, the minimum detectable power is limited by the receiver with the PD/TIA.
We demonstrated a high-efficiency MZM on a Si platform using a 300-μm-long membrane InP-based phase shifters and low-loss deuterated SiN waveguides. The small VπL of 0.4 Vcm was achieved with the insertion loss of 4.5 dB by optimizing the overlap between the distributed carriers and optical mode field. In addition, the fabricated device showed error-free operation for 40-Gbit/s NRZ signal. These results shows that the proposed device has the potential to further reduce the cost, size, and power consumption of optical transceivers.
1. D. Marris-Morini, L. Vivien, J. M. Fédéli, E. Cassan, P. Lyan, and S. Laval, “Low loss and high speed silicon optical modulator based on a lateral carrier depletion structure,” Opt. Express 16(1), 334–339 (2008). [CrossRef] [PubMed]
2. N. N. Feng, S. Liao, D. Feng, P. Dong, D. Zheng, H. Liang, R. Shafiiha, G. Li, J. E. Cunningham, A. V. Krishnamoorthy, and M. Asghari, “High speed carrier-depletion modulators with 1.4V-cm V(π)L integrated on 0.25microm silicon-on-insulator waveguides,” Opt. Express 18(8), 7994–7999 (2010). [CrossRef] [PubMed]
3. D. J. Thomson, F. Y. Gardes, J. M. Fedeli, S. Zlatanovic, Y. Hu, B. Kuo, E. Myslivets, N. Alic, S. Radic, G. Z. Mashanovich, and G. T. Reed, “50-Gb/s Silicon Optical Modulator,” IEEE Photonics Technol. Lett. 24(4), 234–236 (2012). [CrossRef]
4. B. R. Bennett, R. A. Soref, and J. A. Del Alamo, “Carrier-induced change in refractive index of InP, GaAs and InGaAsP,” IEEE J. Quant. Elec. 26(1), 113–122 (1990). [CrossRef]
5. T. Hiraki, T. Aihara, K. Hasebe, K. Takeda, T. Fujii, T. Kakitsuka, T. Tsuchizawa, H. Fukuda, and S. Matsuo, “Heterogeneously integrated III–V/Si MOS capacitor Mach–Zehnder modulator,” Nat. Photonics 11(8), 482–485 (2017). [CrossRef]
7. H. Chen, Y. Kuo, and J. E. Bowers, “A Hybrid Silicon–AlGaInAs Phase Modulator,” IEEE Photonics Technol. Lett. 20(23), 1920–1922 (2008). [CrossRef]
8. C. R. Doerr, L. Chen, L. L. Buhl, and Y.-K. Chen, “Eight-Channel SiO /Si N /Si/Ge CWDM Receiver,” IEEE Photonics Technol. Lett. 23(17), 1201–1203 (2011). [CrossRef]
9. Y. Fan, R. M. Oldenbeuving, C. G. Roeloffzen, M. Hoekman, D. Geskus, R. G. Heideman, and K.-J. Boller, “290 Hz Intrinsic Linewidth from an Integrated Optical Chip-based Widely Tunable InP-Si3N4 Hybrid Laser,” Proc. CLEO 2017, JTh5C.9 (2017). [CrossRef]
10. T. Hiraki, T. Aihara, H. Nishi, and T. Tsuchizawa, “Deuterated SiN/SiON Waveguides on Si Platform and Their Application to C-Band WDM Filters,” IEEE Photonics J. 9(5), 2500207 (2017). [CrossRef]
11. T. Aihara, T. Hiraki, K. Takeda, K. Hasebe, T. Fujii, T. Tsuchizawa, T. Kakitsuka, and S. Matsuo, “Lateral current injection membrane buried heterostructure lasers integrated on 200-nm-thick Si waveguide,” Proc. OFC 2018, W3F.4 (2018). [CrossRef]
12. T. Hiraki, T. Aihara, T. Fujii, K. Takeda, T. Tsuchizawa, T. Kakitsuka, H. Fukuda, and S. Matsuo, “Heterogeneously Integrated Mach-Zehnder Modulator Using Membrane InGaAsP Phase Shifter and Hydrogen-free SiN Waveguide on Si Platform,” Proc. ECOC 2018, Mo3C.1 (2018). [CrossRef]
13. T. Aihara, T. Hiraki, T. Fujii, K. Takeda, T. Tsuchizawa, T. Kakitsuka, H. Fukuda, and S. Matsuo, “56-Gbit/s Operations of Mach-Zehnder Modulators Using 300-µm-long Membrane InGaAsP Phase Shifters and SiN waveguides on Si,” Proc. OFC 2019, M4A.3 (2019). [CrossRef]
14. S. Matsuo, T. Fujii, K. Hasebe, K. Takeda, T. Sato, and T. Kakitsuka, “Directly modulated buried heterostructure DFB laser on SiO2/Si substrate fabricated by regrowth of InP using bonded active layer,” Opt. Express 22(10), 12139–12147 (2014). [CrossRef] [PubMed]
15. T. Fujii, K. Takeda, N.-P. Diamantopoulos, E. Kanno, K. Hasebe, H. Nishi, R. Nakao, T. Kakitsuka, and S. Matsuo, “Heterogeneously Integrated Membrane Lasers on Si Substrate for Low Operating Energy Optical Links,” IEEE J. Sel. Quant. Elec. 24(1), 1500408 (2018). [CrossRef]