⇦Chinese Physics Letters, 2019, Vol. 36, No. 10, Article code 104201 III–V/Si Hybrid Laser Array with DBR on Si Waveguide * Yan-Ping Li (李艳平)1**, Li-Jun Yuan (袁丽君)2, Li Tao (陶利)1, Wei-Xi Chen (陈娓兮)1, Bao-Jun Wang (王宝军)2, Jiao-Qing Pan (潘教青)2** Affiliations 1Institute of Condensed Matter and Material Physics, School of Physics, Peking University, Beijing 100871 2The Key Laboratory of Semiconductor Materials Science, Institute of Semiconductor, Chinese Academy of Sciences, Beijing 100083 Received 3 July 2019, online 21 September 2019 *Supported by the National Basic Research Program of China under Grant No 2013CB632105, the National High Technology Research and Development Program of China under Grant No 2012AA012203, and the National Natural Science Foundation of China under Grant Nos 61404003 and 11174018.
**Corresponding author. Email: liyanping@pku.edu.cn; jqpan@semi.ac.cn Citation Text: Li Y P, Yuan L J, Tao L, Chen W X and Wang B J et al 2019 Chin. Phys. Lett. 36 104201    Abstract We report an eight-channel silicon evanescent laser array operating at continuous wave under room temperature conditions using the selective-area metal bonding technique. The laser array is realized by evanescently coupling the optical gain of InGaAsP multi-quantum wells to the silicon waveguides of varying widths and patterned with distributed Bragg reflector gratings. The lasers have emission peak wavelengths in a range of 1537–1543 nm with a wavelength spacing of about 1.0 nm. The thermal impedances $Z_{\rm T}$ of these hybrid lasers are evidently lower than those DFB counterparts DOI:10.1088/0256-307X/36/10/104201 PACS:42.55.Px, 42.60.By, 42.82.Fv © 2019 Chinese Physics Society Article Text Silicon integrated photonics rapidly expands for supporting ever increasing data rates in internet, parallel many-core computations and wireless communications etc.[1–5] For silicon integrated photonic circuits, the most important active component should be silicon light source, which is the greatest challenge due to the poor spontaneous emission ability of silicon stemming from its indirect energy bandgap.[4,5] Fortunately, many exciting achievements have been made in the past few years.[5–11] Among them, bonding III–V compound semiconductors to silicon platform and evanescently coupling the optical gain of III–V part to silicon waveguide is a low-cost and practical approach.[12–17] Based on the existing hybrid integration technology, the recent research focus shifts to the multi-wavelength and wavelength tunable silicon lasers for wavelength division multiplexing (WDM), which is the largest advantage of silicon optical interconnects. In a multi wavelength laser array, each channel must operate with a single longitudinal mode and uniform channel spacing. The distributed Bragg reflector (DBR) or distributed feedback (DFB) laser is stable and highly reliable single-mode laser,[18–21] and by changing the grating period in each channel, the different lasing wavelength can be realized. Unfortunately, according to the gating equation, $2n_{\rm eff}{\it \Lambda}=\lambda$, a 1 nm deviation of the grating period (${\it \Lambda} $) will cause a wavelength ($\lambda $) error about 5–8 nm due to the high effective refractive index ($n_{\rm eff}$), thus the reported four channel arrayed silicon lasers defined a wide wavelength spacing of 20 nm.[22] By careful use of electron-beam lithography (EBL), the grating with different periods can narrow the wavelength spacing to 5 nm,[23,24] whereas EBL suffers from high costs and low yields, and thus is not CMOS compatible. Could the channel wavelength spacing be as narrow as $\sim $1 nm similar to the ITU standard by ordinary technique? Here, we fix the grating period on silicon passive waveguide but vary the width of silicon waveguide to realize the wavelength varying laser array. By computations and experiments,[25,26] the variation ($\sim $10%) in the ridge width leads to a slight effective index difference (1‰) and then to a lasing wavelength difference ($\sim $1 nm, nonlinearly). This is a simple, low cost and highly-stable method suitable for mass production,[25,26] and these width-varying grating waveguides can be fabricated simply by use of standard photolithography and holographic lithography. Recently, we reported an InGaAsP-Si distributed feedback (DFB) evanescent laser array.[27] However, the surface corrugated gratings bonded directly to the III–V gain region cause higher thermal impedance, the rough surface of the sidewall of the DFB gratings usually causes a scattering loss and then a higher lasing threshold. Meanwhile, DBR lasers putting the grating at the ends of the gain region can overcome the both shortcomings. This work reports an 8-$\lambda$ DBR InGaAsP-Si hybrid laser array based on selective area metal bonding to overcome the disadvantage of DFB ones. The bonding step follows both the whole silicon components and III–V ones are finished separately, avoiding the crossover of the silicon and III–V process flow. This 8-$\lambda$ laser array can support an $8 \times 25$ GHz silicon photonic transceiver. A transverse cross section of the InGaAsP-Si hybrid laser is shown in Fig. 1(a). A buried ridge (BRS) InGaAsP quantum well epitaxial structure is flip-chip bonded onto a patterned SOI wafer with the SAMB method. The III–V material has the same epitaxial structure as in our previous work.[27] In the SOI region, the gratings of 244 nm period are formed on an undoped SOI substrate by conventional holographic exposure combined with the optical photolithography[28] and HBr/He/O2 based inductively coupled plasma (ICP) etching. Eight silicon waveguides of different widths (2.0, 2.1, 2.2, 2.3, 2.4, 2.6, 2.8, 3.0 µm) are defined simultaneously on the photolithographic mask in a periodic unit. The silicon waveguides are fabricated by lithography and subsequent second ICP dry etching (the height of silicon waveguides is 0.5 µm). Then metal lift-off technology is adopted to selectively deposit the bonding metal on each side of the silicon waveguide. The bonding metal is indium, which is widely used in metal bonding and packing due to its low melting point and good fluidness. Finally, the III–V wafer is diced into bars with a length of 300 µm and flip-chip bonded onto the SOI wafer by a bonder with alignment accuracy $\pm 0.5\,µ$m. The bonding process is carried out at 170$^{\circ}\!$C for 5 min under a pressure of about 2 MPa. The 45$^{\circ}$-angular view of the hybrid laser is shown in Fig. 1(b). The effective refractive indexes of the unetched and the etched regions are estimated using the finite differential in time domain (FDTD) solutions (Lumerical Solution Inc., Canada), resulting in a coupling constant $\kappa$ of $\sim $67 cm$^{-1}$. The lengths of the front and back gratings are 150 and 300 µm, resulting in power reflectivities of 48.3% and 93.1% for the front and back mirrors, respectively.
cpl-36-10-104201-fig1.png
Fig. 1. (a) The transverse cross section of the hybrid laser. (b) Scanning electronic microscopic image of 45$^{\circ}$ angle view of the fabricated laser. Inset: the SEM image of the DBR silicon waveguide.
The chip with the fabricated laser is mounted on a thermo electric cooler which allows the operating temperature of the laser varying from 0 to 70$^{\circ}\!$C. The laser output from the waveguide facet of the front mirror is collected by a lensed multi-mode fiber and then characterized using a spectrum analyzer or an optical power meter. Figure 2(a) shows the measured continuous wave output power from one facet as a function of injected current at operating temperatures ranging from 10 to 35$^{\circ}\!$C. As can be seen, at room temperature, the threshold current is 20 mA with an output power of 42 µW at 70 mA. The coupling efficiency from waveguide to fiber is about 30%. Figure 2(b) shows the measured lasing spectrum of the silicon evanescent laser driven slightly above the threshold of 20 mA at room temperature. The spectrum was measured with an Agilent 86142B spectrum analyzer. The used wavelength resolution is 0.3 nm, thus the measured linewidth is slightly larger. The lasing wavelength is around 1544 nm with a side-mode suppression ratio (SMSR) of $\sim $25 dB. The lower SMSR and larger spectral linewidth are due to the lower quality factor $Q$ of the cavity, that is, the lower reflectivity of the front and back regions of DBR, which can be improved by increasing the length of DBR regions or the etching depth of the gratings. Furthermore, the lasing wavelength is shifted to longer wavelength at higher injection currents due to device heating. The wavelength shift is about 0.018 nm/mA.
cpl-36-10-104201-fig2.png
Fig. 2. (a) Curves of light output power versus injection current for stage temperature of 10–35$^{\circ}\!$C. (b) The lasing spectrum with 30 dB side mode extinction ratio at 20 mA at room temperature.
cpl-36-10-104201-fig3.png
Fig. 3. (a) Lasing wavelength versus electrical dissipated power of DBR laser. (b) Lasing wavelength versus stage temperature of DBR laser. (c) Lasing wavelength versus electrical dissipated power of DFB laser. (d) Lasing wavelength versus stage temperature of DFB laser.
The thermal impedance $Z_{\rm T}$ of the hybrid laser is determined by the peak emission wavelength shifts with the increase of the input electrical power ($P$) and the corresponding temperature ($T$) rise of the device; that is, $Z_{\rm T}=(d\lambda /dT)^{-1}(d\lambda /dP)$. Figure 3(a) shows the plot of wavelength versus dissipated electrical power, and Fig. 3(b) shows the changes in lasing wavelength as a function of stage temperature (controlled using a Peltier element). Using linear fit, we measure the wavelength shift with the increase of dissipated power $\Delta \lambda/\Delta P=9.68$ nm/W, while the wavelength shift with the increase of stage temperature is $\Delta \lambda/\Delta T=0.133$ nm/$^{\circ}\!$C. Dividing these values yields a thermal impedance of 72.8 K/W. This is much better than that of our reported DFB[27] laser of 137.3 K/W, which can be calculated from Figs. 3(c) and 3(d). This thermal impedance can be improved by optimizing the bonding condition or changing the bonding metal. Our thermal impedance value is at a moderate level. The thermal impedance for the adhesive bonding silicon laser is higher, typically 7000 K/W,[29] and the one for a direct bonding laser is about 500 K/W.[30] However, the optimized thermal impedance for the 3D integrated hybrid silicon laser can be as low as 6.2–18.3 K/W.[31] As the light source of silicon optoelectronic integrated chip, the lower thermal impedance is beneficial to the stability of wavelength, and the lower chip operating temperature.
Table 1. Effective refractive index and the lasing wavelength of the eight Si waveguides versus waveguide width.
Ch.1 Ch.2 Ch.3 Ch.4 Ch.5 Ch.6 Ch.7 Ch.8
$w$ (µm) 2.0 2.1 2.2 2.3 2.4 2.6 2.8 3.0
$n_{\rm eff}$ 3.14576 3.14793 3.14983 3.15149 3.15295 3.15536 3.15728 3.15883
$\lambda$ (nm) 1535.1 1536.2 1537.1 1537.9 1538.6 1539.8 1540.8 1541.3
To demonstrate the scalability of this hybrid laser method for a WDM optical system, we achieve an eight-channel laser array by varying the width of silicon waveguide. To guarantee the coupling efficiency, the eight widths of the silicon waveguide are chosen between 2.0 and 3.0 µm, i.e., 2.0, 2.1, 2.2, 2.3, 2.4, 2.6, 2.8 and 3.0 µm. The effective refractive index of each Si waveguide is listed in Table 1 calculated by FDTD. The effective refractive index increases with the silicon waveguide width exponentially, and increases more rapidly below 2.5 µm. The refractive index of passive waveguide as a function of silicon waveguide width can generally be described by the following empirical formula fitted from the computation data $$ n_{\rm eff} =3.1643-0.238e^{-1.269w}, $$ where $n_{\rm eff}$ is the effective refractive index of the passive waveguide, and $w$ is the width of the silicon waveguide. The wavelength spacing is around 0.8 nm estimated by $\Delta \lambda =2(\Delta \eta_{\rm eff}{\it \Lambda})$, where ${\it \Lambda} =244$ nm is the Bragg gratings period. The lasing spectrum of each device driven slightly above its own threshold is shown in Fig. 4(a). The lasing wavelengths are shown in Table 1 and Fig. 4 with an SMSR around 25 dB. There is slight difference between the measured and calculated wavelength values due to the fabrication and calculation errors. The fabrication tolerances of the silicon waveguide width and grating constant are 10 and 0.2 nm, respectively, resulting in a wavelength offset of 1 nm in total. The effective refractive index of the hybrid waveguide increases nonlinearly with the increase of the silicon waveguide width, and the slope of the curve descends gradually,[27] thus the wavelength spacing decreases accordingly as the width spacing of silicon waveguide is fixed. By changing the width of silicon waveguide nonlinearly, the fixed wavelength spacing can be obtained. Furthermore, narrower wavelength spacing can be achieved by reducing the silicon waveguide width interval. Figure 4(b) shows the imaged output facet of the InGaAsP/Si evanescent laser array operating simultaneously. The devices are driven with a common pulsed current 250 mA at a repetition rate of 1 kHz and 0.2% duty cycle. More channel hybrid lasers with different silicon waveguide widths can be fabricated to meet the needs of dense wavelength division multiplexing (DWDM) systems using this approach. Furthermore, if the eight channel outputs are guided into a single silicon nanowire waveguide, a considerably high optical power density can be achieved, which is needed for on-chip silicon nonlinear photonics and microwave photonics.
cpl-36-10-104201-fig4.png
Fig. 4. (a) Spectra of the eight lasing channels. (b) Infrared microscope image of the lasing array.
Eight-channel InGaAsP-Si DBR array lasers fabricated with this technique can be easily scaled up to more channels by increasing the silicon waveguide width range and reducing the silicon waveguide width interval. Only photolithography and holographic lithography are used for the fabrication of silicon waveguides, which are CMOS technology compatible. These DBR lasers have better thermal properties than those DFB counterparts, which can be easily used in silicon optical interconnects. References Optical interconnects to siliconStrained silicon as a new electro-optic materialAll-optical control of light on a silicon chipRecent progress in lasers on siliconA continuous-wave Raman silicon laserOptical gain and stimulated emission in periodic nanopatterned crystalline siliconStrong quantum-confined Stark effect in germanium quantum-well structures on siliconIII-V/Si hybrid photonic devices by direct fusion bonding13-μm InAs/GaAs quantum-dot lasers monolithically grown on Si substrates using InAlAs/GaAs dislocation filter layersGraphene-silicon hybrid photonic integrated circuitsElectrically pumped hybrid AlGaInAs-silicon evanescent laser1310-nm Hybrid III–V/Si Fabry–Pérot Laser Based on Adhesive BondingA Selective-Area Metal Bonding InGaAsP–Si LaserElectrically Pumped Room-Temperature Pulsed InGaAsP-Si Hybrid Lasers Based on Metal BondingA Buried Ridge Stripe Structure InGaAsP-Si Hybrid LaserA Single Mode Hybrid III–V/Silicon On-Chip Laser Based on Flip-Chip Bonding Technology for Optical InterconnectionA Distributed Bragg Reflector Silicon Evanescent LaserA 16-Channel Distributed-Feedback Laser Array with a Monolithic Integrated Arrayed Waveguide Grating Multiplexer for a Wavelength Division Multiplex-Passive Optical Network System NetworkA hybrid silicon single mode laser with a slotted feedback structureExperimental demonstration of a hybrid III–V-on-silicon microlaser based on resonant grating cavity mirrorsIntegrating silicon photonicsA Compact EADFB Laser Array Module for a Future 100-Gb/s Ethernet TransceiverUltracompact, 160-Gbit/s transmitter optical subassembly based on 40-Gbit/s × 4 monolithically integrated light sourceAnalysis, design, and fabrication of GaAlAs/GaAs DFB lasers with modulated stripe width structure for complete single longitudinal mode oscillationStatic and dynamic characteristics of MQW DFB lasers with varying ridge width4-λ InGaAsP-Si distributed feedback evanescent lasers with varying silicon waveguide widthThe Fabrication of Eight-Channel DFB Laser Array Using Sampled GratingsSilicon-Integrated Hybrid-Cavity 850-nm VCSELs by Adhesive Bonding: Impact of Bonding Interface Thickness on Laser PerformanceLow Threshold Electrically-Pumped Hybrid Silicon Microring LasersHigh-Thermal Performance 3D Hybrid Silicon Lasers
[1] Miller D A B 2000 IEEE J. Sel. Top. Quantum Electron. 6 1312
[2] Jacobsen R S, Andersen K N, Borel P I, Pedersen J F, Frandsen L H, Hansen O, Kristensen M, Lavrinenko A V, Moulin G, Ou H, Peucheret C, Zsigri B and Bjaklev A 2006 Nature 441 199
[3] Almeida V R, Barrios C A, Panepucci R R and Lipson M 2004 Nature 431 1081
[4] Liang D and Bowers J E 2010 Nat. Photon. 4 511
[5]Li X Y, Yuan T N, Shao S Q Shi Z J, Wang Y, Yu Y D and Yu J Z 2011 Physics 40 28
[6] Rong H, Jones R, Liu A, Cohen O, Hak D, Fang A and Paniccia M 2005 Nature 433 725
[7] Cloutier S G, Kossyrev P A and Xu J 2005 Nat. Mater. 4 887
[8] Kuo Y H, Lee Y K, Ge Y, Ren S, Roth J E, Kamins T I, Miller D A B and Harris J S 2005 Nature 437 1334
[9] Tanabe K, Watanbe K and Arakawa Y 2012 Sci. Rep. 2 349
[10] Tang M, Chen S, Wu J, Jiang Q, Dorogan V G, Benamara M, Mazur Y I, Salamo G J, Seeds A and Liu H 2014 Opt. Express 22 11528
[11] Xiao T H, Yu Y and Li Z Y 2017 Acta Phys. Sin. 66 217802 (in Chinese)
[12] Fang A W, Cohen H Park O, Jones R, Paniccia M J and Bowers J E 2006 Opt. Express 14 9203
[13] Stankovic S, Jones R, Sysak M N, Heck J M, Roelkens G and Thourhout D V 2011 IEEE Photon. Technol. Lett. 23 1781
[14] Hong T, Ran G Z, Chen T, Pan J Q et al 2010 IEEE Photon. Technol. Lett. 22 1141
[15] Chen T, Hong T, Pan J Q, Chen W X, Cheng Y B, Wang Y, Ma X B, Liu W L, Zhao L J, Ran G Z, Wang W and Qin G G 2009 Chin. Phys. Lett. 26 064211
[16] Yuan L J, Tao L, Chen W X, Li Y P, Liang S, Yu H Y, Ran G Z, Pan J Q and Wang W 2015 IEEE Photon. Technol. Lett. 27 352
[17] Wang H L and Zheng W H 2016 Chin. Phys. Lett. 33 124207
[18] Fang A W, Koch B R, Jones R, Lively E, Liang D, Kuo Y H and Bowers J E 2008 IEEE Photon. Technol. Lett. 20 1667
[19] Zhao J Y, Chen X, Zhou N et al 2014 Chin. Phys. Lett. 31 074205
[20] Zhang Y, Qu H, Wang H, Zhang S, Liu L, Ma S and Zheng W 2013 Opt. Express 21 877
[21] Konink Y, Raineri F, Bazin A, Raj R, Roelkens G and Baets R 2013 Opt. Lett. 38 2496
[22] Paniccia M 2010 Nat. Photon. 4 498
[23] Kanazawa S, Fujisawa T, Ohki A et al 2011 IEEE J. Sel. Top. Quantum Electron. 17 1191
[24] Fujisawa T, Itoh T, Kanazawa S, Takahata K, Ueda Y, Iga R, Yamanaka T, Kotoku M and Ishii H 2013 Opt. Express 21 182
[25] Nakano Y and Tada K 1988 IEEE J. Quantum Electron. 24 2017
[26] Hong J, Huang W P, Makino T and Pakulski G 1994 IEE Proc. - Optoelectron. 141 303
[27] Tao L, Yuan L J, Li Y P, Yu H Y, Wang B J, Kan Q, Chen W X, Pan J Q, Ran G Z and Wang W 2014 Opt. Express 22 5448
[28] Zhu H L, Hu X D, Wang H, Hua D H, Liang S, Zhao L J and Wang W 2010 IEEE Photon. Technol. Lett. 22 353
[29] Haglund E P, Kumari S, Haglund E, Gustavsson J S, Baets R G, Roelkens G and Larsson A 2017 IEEE J. Sel. Top. Quantum Electron. 23 1700109
[30] Liang D, Fiorentino M, Srinivasan S, Bowers J E and Beausoleil R G 2011 IEEE J. Sel. Top. Quantum Electron. 17 1528
[31] Song B W, Megalini L, Dwivedi S, Ristic S and Jonathan K J 2017 IEEE Photon. Technol. Lett. 29 1143