We present an integration platform based on quantum well intermixing for multi-section hybrid silicon lasers and electroabsorption modulators. As a demonstration of the technology, we have fabricated discrete sampled grating DBR lasers and sampled grating DBR lasers integrated with InGaAsP/InP electroabsorption modulators. The integrated sampled grating DBR laser-modulators use the as-grown III–V bandgap for optical gain, a 50 nm blue shifted bandgap for the electrabosprtion modulators, and an 80 nm blue shifted bandgap for low loss mirrors. Laser continuous wave operation up to 45 °C is achieved with output power >1.0 mW and threshold current of <50 mA. The modulator bandwidth is >2GHz with 5 dB DC extinction.
© 2008 Optical Society of America
A significant amount of research has been devoted to developing a laser source that is compatible with silicon photonics and CMOS processing [1–3]. Such a device has the potential to take advantage of the high volume manufacturing infrastructure that has been developed in the CMOS microelectronics industry. One of most promising technologies for achieving this goal is a wafer bonded approach where a III–V active region is bonded onto a pre-patterned silicon on insulator (SOI) wafer. This technique uses the III–V material for optical gain, the Si layer for optical waveguiding, and does not require any alignment steps in the bonding process. Using this approach, Fabry-Perot (FP) lasers, ring lasers, photodetectors, modulators, DFB lasers, and amplifiers have been demonstrated [4,5]. However, one of the key drawbacks to this approach is that devices requiring multiple bandgaps, such as multi-section lasers and integrated laser transponders, are difficult to fabricate without the use of multiple bonding steps and different epitaxial structures.
In this work we present an integration approach for hybrid silicon evanescent devices that require multiple III–V bandgaps. Using this method, we have fabricated hybrid silicon multisection sampled grating DBR (SGDBR) lasers and hybrid silicon SGDBR lasers with monolithically integrated electrabsorption modulators. The SGDBR lasers consist of five electrically isolated regions, including a front mirror, rear mirror, gain, phase, and backside absorber . To achieve low optical loss in the mirrors and create an intermediate bandedge for the EA modulator, a quantum well intermixing (QWI) process is used to shift the bandgap of the III–V mirrors, phase, and modulator regions away from the active region. Both laser designs exhibit side mode suppression ratio >30 dB, threshold current of <50 mA, and CW operation up to or >40 °C. The maximum output power for the stand-alone SGDBR laser was >1 mW, with tuning over 13 nm. The modulator bandwidth is >2 GHz with DC extinction ratio >5 dB.
2. Quantum well intermixing
The QWI process in this work is based on implant enhanced intermixing in combination with selective removal of an InP buffer . Details of the QWI process and the as-grown hybrid laser III–V base structure are shown in Figs. 1(a) through 1(e). Intermixing begins with a blanket phosphorous implant that is captured in the patterned InP buffer as shown in Fig. 1(a). A 400 nm thick SiNx dielectric masks sections of the III–V from the implant to preserve the as-grown bandgap. The implant generates vacancies, which are then diffused through the quantum wells and barriers via a rapid thermal anneal (RTA) (Fig. 1(b)). The vacancy diffusion causes atomic interdiffusion between the wells and barriers, modifying their potential profile, and hence bandgap. To stop the intermixing, the InP buffer layer (the source of vacancies) is selectively wet etched as shown in Fig. 1(c). Additional RTA steps are then used to continue bandgap shifting in regions where the InP buffer remains (Figs. 1(d)-1(e)).
There are several parameters that affect the intermixing performance. These include the implant species, dose, energy, and temperature, and the RTA temperature and time. Optimization of many of these parameters for a similar base structure can be found in , which presents a detailed study of this QWI process for InP photonic integrated circuits. In this work, the goal was to adapt a similar process for hybrid silicon evanescent devices. The implant conditions selected include a dose of 5×1014 cm-2, energy of 100 keV, and temperature of 200 °C. This combination of settings has been shown to maximize the number of vacancies generated per implanted P ion in the InP buffer layer without the formation of vacancy complexes, which are detrimental to the intermixing process. Using these implant conditions, the RTA time and temperature was selected based on two objectives. The first was to achieve a large separation between the PL of the optical gain (implant protected) and intermixed regions. A large PL separation between these two regions minimizes the optical loss associated with the Urbach tail of the bandgap in the intermixed mirrors, phase and taper sections. The second objective is to minimize parasitic PL shifts where intermixing is not desired. A shift in the PL wavelength of the gain section for example, affects control over the operating wavelength range of the laser.
Figure 2(a) shows the change in PL of the intermixed and the implant protected quantum well regions as a function of RTA temperature after a 240s anneal. Higher RTA temperatures result in larger PL shifts due to an increase in the vacancy diffusion constant. PL shifts of up to 190 nm can be achieved with a 240s anneal at 775 °C. While such large shifts are attractive for low optical loss, these high temperature conditions have large parasitic PL shifts in regions where intermixing is not desired (>80 nm). Reducing the anneal temperature to 725 °C results in PL shifts of >110 nm with a parasitic PL shift in the implant protected regions of <30 nm. Further reductions of the anneal temperature show a maximum PL shift of 60 nm at 675 °C, which is not sufficient to minimize the Urbach tail loss in mirror, phase, and taper regions. Based on achieving the a large separation between the intermixed and protected regions and minimizing shifts in the laser gain section PL, an RTA temperature of 725 °C was selected.
Figure 2(b) shows the PL shift in the quantum wells as a function of RTA time with a fixed annealing temperature of 725 °C. The initial rapid shift in the PL is related to the strong atomic concentration gradient between the wells and barriers in the as-grown material. However, as the RTA time increases and the driving force in the intermixing process is reduced, the rate of change in the PL peak slows. Eventually, the rate of PL shift in the implanted regions slows enough to become equivalent to the shift in the implant protected regions. At this point the maximum band edge separation between the optical gain and intermixed regions has been achieved, and further RTA is not useful in reducing optical loss associated with the Urbach tail. Based on achieving the maximum bandedge separation between the intermixed and implant protected regions in the SGDBR and SGDBR-EAM devices, an RTA time of 330s was selected for the QWI process. Figure 2(b) also shows the shift in PL as a function of anneal time for the regions where an intermediate bandgap is desired. An RTA shift of 45s was selected before the InP buffer is removed to achieve a 60 nm separation in the PL of the gain and EAM. Similar to the behavior in the implant protected regions, the EAM bandedge continues to shift slightly with continuing RTA after removal of the buffer.
Figure 2(c) shows PL spectra from the final three bandgaps in the SGDBR-EAM. Good uniformity of the PL full width half max (FWHM) can be seen for all three bandgaps, indicating consistent material quality for the modulator, gain, and mirror regions .
3. Hybrid laser fabrication process
The hybrid silicon SGDBR and SGDBR-EAM fabrication process can be separated into three sections: pre-bonding, bonding, and post-bonding.
The prebonding process includes intermixing, gratings, and etching a set of alignment marks through the patterned face of the III–V into the p-InP substrate. A schematic of the III–V after prebonding is complete is shown in Fig. 3. Both SGDBR and SGDBR-EAM devices share a common, 600µm long, backside absorber and are designed to emit light out of opposite sides of the integrated chip. Fabrication begins with the QWI process, where stripes of active (bandgap 1), modulator (bandgap 2) and passive (bandgap 3) regions are patterned. Following QWI, similar long stripes of grating bursts with first order gratings are patterned and etched into the InGaAsP/InP superlattice layers and the n-contact layer. Finally, alignment marks are etched through the p-contact layer into the p-InP substrate and backfilled with SiNx dielectric. The SiNx protects the alignment marks from the etchant used for substrate removal. Separately from the III–V processing, shallow straight waveguides and alignment marks are etched into the SOI wafer. A top down schematic and cross section of the patterned SOI wafer is shown in Fig. 4.
The plasma assisted bonding process used to bond the patterned surface of the III–V to the patterned surface of the SOI wafer is described in . What is unique about using this bonding process in combination with QWI is that since the III–V wafer is patterned before bonding, some rotational alignment must be performed to ensure that the III–V gratings are orthogonal to the silicon waveguides. Ensuring the gratings are perpendicular also ensures that active, modulator, and passive regions are perpendicular to the silicon waveguides. To avoid complex alignment in the horizontal or vertical directions, III–V patterns (gratings, QWI regions) are such that they are constant in the vertical direction, and silicon patterns are constant in the horizontal direction. While the rotational requirement is an added complication, simple manual placement of 1 cm×1 cm InP and SOI samples resulted in misalignment of <0.2 degrees over 4 samples. Improvements in rotational alignment will allow mask tolerances to be tightened and allow scaling to larger sample sizes.
3.3 Post bonding
After bonding, the remaining fabrication steps are similar to that in reference . These steps include substrate removal, mesa etch, n and p contact metal deposition, proton implantation, and probe metallization. Modulators undergo an additional lithography to etch away the n- InP contact layer beneath the p-probe metal to minimize parasitic capacitance. In addition to providing lateral current confinement in the gain region, the proton implantation is used to electrically isolate the various laser sections. Alignment marks on both the silicon and III–V are used during the backend processing to ensure correct alignment to both the silicon waveguides and intermixed III–V regions. After backside processing, the laser waveguides are diced at a 7 degree angle, polished, and AR coated. Figure 5 shows cross sections of active, passive, and modulator regions after processing is complete. A cross section of the completed SGDBR-EAM is shown in Fig. 6.
4. Overview of SGDBR and SGDBR-EAM devices
The laser portion of the stand alone SGDBR and the integrated SGDBR-EAM consists of five electrically isolated sections. Both lasers use a taper to transition the optical mode from the hybrid waveguide with both III–V and Si to a purely silicon waveguide. For the stand alone SGDBR, these sections include a 650 µm long back mirror, 80 µm long phase section, 550 µm long gain region, a 270 µm long front mirror, and 100 µm long taper. The taper design is identical to that used for optical amplifiers in . The back mirror has fourteen 7.6 µm wide grating bursts and a 46.4 µm sampling period. The front mirror has five 5.2 µm wide grating bursts with a sampling period of 52.4 µm. Each section of the laser is separated by a 10 µm wide proton implanted region for electrical isolation. The measured resistance between neighboring sections was >10 kΩ. The design of the integrated SGDBR-EAM uses a 650 µm backside absorber, 760 µm long rear mirror, 80 µm long phase section, 550 µm long gain region, a 780 µm long front mirror, a 200 µm long EAM, and a 100 µm long taper.
Both lasers use a 20 µm wide III–V mesa in the gain, mirror, phase and backside absorber regions and a 4 µm wide mesa in the modulator. The silicon epitiaxial layer beneath the III–V mesa is 0.4 µm thick and waveguide widths vary from 1.0 to 2.5 µm with an etch depth of 0.3 µm. The buried oxide layer is 1 µm thick. The measured loss and kappa of the III–V gratings was 165 cm-1 and 3dB/100 µm, respectively, for a 2 µm wide Si waveguide and 100 nm etch depth (into the III–V).
5. Device performance
5.1 Design A—sampled grating DBR laser
The SGDBR laser continuous wave LI (optical power-current) characteristics are shown in Fig. 7. The laser is mounted on a temperature controlled stage and the optical power is measured using an integrating sphere to minimize fiber coupling uncertainty. For a device with a 1.0 µm wide Si waveguide, CW lasing is achieved up to 30 °C with gain section resistance of 14 ohms, output power up to 1.0 mW, and a threshold current of 48 mA. The low output power is a result of the high loss in the etched gratings. The dips in the LI characteristics shown in Fig. 7 are a result of temperature induced cavity mode hops  as shown by the lasing spectra from the SGDBR as a function of gain current in Fig. 8. As the bias current is increased the temperature of the gain section increases resulting in the cavity modes shifting to longer wavelengths. As the lasing mode moves across the mirror stop band, it eventually comes into competition with a neighboring cavity mode at a lower wavelength and a mode hop occurs.
Overlaid output spectra from the SGDBR with current injection into both front and back mirrors are shown in Fig. 9. Tuning over three cavity modes can be achieved between 1501 and 1514 nm with side mode suppression >30 dB.
The limited tuning range in Fig. 9 comes from several factors, including a small quantum well confinement factor (3%), and large device thermal impedance since the SGDBR is fabricated using silicon on insulator material . The increase in amplified spontaneous emission at lower wavelengths in Fig. 8 is a result of reduced loss in the front mirror when current is injected for tuning. The tuning can be improved by increasing the number of quantum wells and introducing a tuning layer into the III–V base structure .
5.2 Design B – integrated sampled grating DBR laser – EA modulator
The integrated SGDBR-EAM continuous wave LI characteristics are shown in Fig. 10. For a device with a 2.5 µm wide Si waveguide, CW operation is achieved up to 45 °C with output power up to 0.5 mW at 10 °C and 170 mA of gain current. Dips in the LI characteristics are again due to temperature induced cavity mode hops. Although the temperature performance of the SGDBR-EAM is improved compared with the stand alone SGDBR, the difference in maximum operating temperatures (30 °C vs. 45 °C) is within the variation observed for all tested devices. Overlaid output spectra from the SGDBR-EAM with current injection into both front and back mirrors are shown in Fig. 11. Tuning over four supermodes can be achieved at wavelengths of 1524, 1518, 1512, and 1554 nm with side mode suppression >35 dB. For high current injection levels into the front mirror, the supermode at 1554 nm occurs as a result of the repeat mode spacing in the mirror design . The reduction in output power compared to the stand alone SGDBR is due to additional loss in the longer front mirror.
DC extinction characteristics and small signal electrical-optical bandwidth measurements from the integrated EAM are shown in Figs. 12 and 13 respectively for a 200 µm long modulator. The integrated EAMs show extinction >5 dB at -6V reverse bias depending on the wavelength of operation. Shorter wavelengths show more efficient operation than longer wavelengths due to the proximity between the modulator bandedge and the operating wavelength. The bandwidth of the integrated modulators depends largely on the applied DC reverse bias achieving greater than 2 GHz with a 6V bias. The series resistance of the modulator is 40 ohms. The large variation in frequency response is due to carrier diffusion effects as photocurrent is generated outside the 4 um ridge and diffuses towards the center of the structure. It is possible to improve both the bandwidth and extinction performance of the modulator by increasing the number of quantum wells in the base structure and using an undercut III–V quantum well design as in .
We have demonstrated an integration approach based on quantum well intermixing that can be used for devices that require multiple III–V bandgaps in the hybrid silicon evanescent device platform. As a demonstration of the technology, we have fabricated a multi-section SGDBR laser, and an SGDBR laser integrated with an electroabsorption modulator. The stand alone SGDBR laser uses two bandgaps at 1440 nm and 1520 nm for the mirrors and the gain region respectively. The stand alone SDGDBR showed CW operation up to 30°C with a maximum output power of 1.0 mW and side mode suppression ratio >30 dB. The integrated SGDBR-EAM uses three bandedges at 1440 nm, 1470 nm, and 1520 nm for the passive, modulator, and gain regions respectively. The integrated modulator showed DC extinction >5 dB with 6V reverse bias and a 3-dB bandwidth up to 2 GHz. While the performance of the SGDBR and SGDBR-EAM is not as good as the best devices on InP substrates, this represents the first tunable laser on a silicon platform, and the first integration of tunable lasers with external modulators on silicon.
The authors would like to thank Di Liang and Alexander Fang for useful discussions and help with the bonding process. This project was partially supported by DARPA/MTO and ARL under grant W911NF-05-1-0175.
References and links
3. S. Lombardo, “A Room-temperature luminescence from Er3+-implanted semi-insulating polycrystalline silicon,” Appl. Phys. Lett. 63, 1942–1944 (1993). [CrossRef]
4. A. W. Fang, R. Jones, H. Park, O. Cohen, O. Raday, M. J. Paniccia, and J. E. Bowers, “Integrated AlGaInAs-silicon evanescent race track laser and photodetector,” Opt. Express 5, 2315–2322, (2007). [CrossRef]
5. A. W. Fang, E. Lively, Y.-H. Kuo, D. Liang, and J. E. Bowers, “Distributed Feedback Silicon Evanescent Laser,” OFC/NFOEC, postdeadline session PDP15, 2008.
6. V. Jayaraman, Z. Chuang, and L. A. Coldren, “Theory, Design, and Performance of Extended Tuning Range Semiconductor Lasers with Sampled Gratings,” IEEE J. Quantum Electron. 29, 1824–1834 (1993). [CrossRef]
7. E. J. Skogen, J. S. Barton, S. P. Denbaars, and L. A. Coldren, “A Quantum-Well-Intermixing Process for Wavelength-Agile Photonic Integrated Circuits,” J. Sel. Top. Quantum Electron. 8, 863–869 (2002). [CrossRef]
8. E. Skogen, “Quantum Well Intermixing for Wavelength Agile Photonic Integrated Circuits,” PhD. Thesis University of California Santa Barbara, (2003).
9. D. Nie, T. Mei, H. S. Djie, M. K. Chin, X. H. Tang, and Y. X. Wang, “Implementing multiple bandgaps using inductively coupled argon plasma enhanced quantum well intermixing,” J. Vac. Sci. Technol. B 23, 1050–1053 (2005). [CrossRef]
10. H. Park, Y.-H. Kuo, A. W. Fang, R. Jones, O. Cohen, M. J. Pannicia, and J. E. Bowers, “A hybrid AlGaInAs-silicon evanescent preamplifier and photodetector,” Opt. Express 15, 13539–13546 (2007). [CrossRef] [PubMed]
11. M-C Amann and J. Buus, Tunable Laser Diodes (Artech House1998).
12. M. N. Sysak, H. Park, A. W. Fang, J. E. Bowers, R. Jones, O. Cohen, O. Raday, and M. Paniccia, “Experimental and theoretical thermal analysis of a Hybrid Silicon Evanescent Laser,” Opt. Express 15, 15041–15046 (2007). [CrossRef] [PubMed]
13. M-C Amann, S. Illek, C. Schanen, and W. Thulke, “Tunable twin-guide laser: A novel aser diode with improved tuning performance,” Appl. Phys. Lett. 54, 2532–2533 (1989). [CrossRef]
14. Y.-H. Kuo, H.-W. Chen, and J. E. Bowers, “A hybrid silicon evanescent electroabsorption modulator,” OFC 2008, San Diego, CA, (2008).