OSA / IPR/PS 2010 a326_1.pdf PWD2.pdf
Super-Long Cavity, Monolithically Integrated 1-GHz Hybrid Mode-Locked InP Laser for All-Optical Sampling Stanley Cheung1, Jong-Hwa Baek1, Francisco M. Soares1, Ryan P. Scott1, Xiaoping Zhou1, Nicolas K. Fontaine1, M. Shearn2, A. Scherer2, Douglas M. Baney3, and S. J. B. Yoo1 1 Department of Electrical and Computer Engineering, University of California, Davis, 95616 Division of Engineering and Applied Science, California Institute of Technology, Pasadena, CA 91125 3 Measurements Research Laboratory, Agilent Technologies, 5301 Stevens Creek Blvd., Santa Clara, CA 95051 2
Abstract: A 1-GHz hybrid mode-locked monolithic semiconductor laser on an InP platform is demonstrated. Monolithic integration of the 4.1 cm cavity mode-locked laser with active quantum well and passive waveguide can be achieved with 400 nm wide and 3μm deeply etched mirrors. ©2010 Optical Society of America OCIS codes: (140.4050) Mode-locked lasers; (250.5960) Semiconductor lasers
1. Introduction High repetition rate (10-50GHz) semiconductor mode-locked lasers (MLL) provide excellent performance in terms of stable output and high optical power and they are commonly investigated for optical frequency comb generation and optical code division multiple access (O-CDMA) networks [1, 2]. However, there have been few studies of monolithically integrated low frequency (< 2-GHz) semiconductor MLLs useful for RF-Photonics and numerous other applications where low-cost electronics and high-speed (short pulses) optics can be leveraged. Photonic ADCs can benefit from short pulse and low repetition rate lasers on a monolithic platform by achieving information quantization in the electrical domain while performing sampling in the optical domain, which entails interfacing low repetition rate lasers with fast all-optical switches with photo-detection [3]. Using many such ADCs in parallel on a monolithic platform provides advantages in cost, power, and equivalent-number-of-bits (ENOBs) compared to very fast electronics ADCs. This paper discusses a 1-GHz hybrid mode-locked (HML) InP laser that can be monolithically integrated with Mach-Zehnder (MZI) modulators that serve as an ultra-compact sampling module. The focus of this paper is to realize an ultra-compact, monolithically integrated high speed sampling module that is compatible with deeply-etched mirrors (DEM) which form the 1-GHz HML laser cavity. 2. Device Design and Fabrication (a)
(b) L = 41195.13 μm
taper to 5 μm WG width at facet is 1.5 at long sections to μm to filter reduce cavity high-order loss modes (c) SOA MLL 1GHz LSOA=4000.0 μm, LABS=150 μm SA MLL 1GHz LSOA=4000.0 μm, LABS=50 μm
(d)
(e) Gain section contact
Absorber contact
Optical Waveguide 1st Epitaxial Growth
Definition of Active Region
Regrowth
Fe-InP Fe-InP (f)
Waveguide Core Metallization
WG width at SOA and SA is 2.5 μm
Selective Area Regrowth of Fe-Doped InP
Waveguide Definition and Dry-Etching
Fig. 1 (a) Schematic of 1-GHz HML laser, and (b) a photo of the fabricated device, and (c) a close-up of the active section in part (a), (d) fabrication procedure (e) SEM pictures of the 1-GHz HML (f) SEM picture of the optical waveguide cross section.
The monolithically integrated 1-GHz HML requires a cavity length of ~41mm, which is difficult to realize with low loss. Our integration process and approach is described in [4, 5] and utilizes the same method in fabricating arrayed waveguide gratings (AWGs), amplitude and phase modulators, and MZIs, providing a InP material platform capable of integrating many functional components on a single monolithic chip. Fig. 1(a) shows a total laser cavity length of ~41mm, a saturable absorber (SA) length of 150 μm, and a SOA length of 4mm. The active regions (SA and SOA) have a cross section dimension of 2.5 μm (width) and 0.5 μm (height). Cavity losses for a straight waveguide of the mentioned dimensions were measured to have a loss of ~2dB/cm, therefore the waveguide is adiabatically tapered to a 5 μm width at the long passive regions to reduce cavity losses down to ~0.69dB/cm. The waveguide at the output is tapered to a 5 μm width to filter out high-order modes.
OSA / IPR/PS 2010 a326_1.pdf PWD2.pdf
3. Experimental Setup Fig.2 (a) shows the experimental arrangement used in measuring continuous wave (CW) and hybrid mode-locked operation of the laser. A 1-GHz drive signal from a RF synthesizer was used in combination with a reverse DC bias applied to the SA via a bias-tee and high frequency ground-signal-ground (GSG) microwave probe. The reverse bias passively mode-locks the laser, while the RF drive signal enables active mode-locking. DC probe tips were used for current injection into the gain region. The laser was mounted on an aluminum plate with indium solder and temperature controlled with a thermo-electric cooler (TEC). Light from the output of the right facet was coupled using a lensed fiber and then the electrical power spectrum, electrical time domain, optical spectrum and high resolution spectrum were measured with a 10-GHz Agilent light-wave receiver (11982A) and a 26-GHz RF spectrum analyzer, an 80-GHz sampling oscilloscope, an optical spectrum analyzer (OSA) and a swept heterodyne detector.
Isolator Lensed Fiber
1GHz
Sampling Oscilloscope
Trigger
1 0.8 0.6
T=3C T=5C T=6C T=7C T=8C
0.4 0.2 0 150
+20 dBm
200 250 Current (mA)
85 80 75 70 65 60 55 50 45 40 35
RF=1.0425GHz (20dBm) Solid Line: T=7°C Dashed Line: T=4°C
Power (mW)
SA Bias
Swept Heterodyne Detector
Power (mW)
DC Current
Optical Spectrum Analyzer
RF Spectrum Analyzer
Pulse Width (ps)
Reverse Bias Voltage
Single sided fiber coupled output power (mW)
4. 1-GHz Hybrid Mode-Locked Laser Performance In order to characterize the laser, the saturable absorber was initially unbiased to investigate continuous wave (CW) operation as shown by the LI curve in Fig. 2(b). It can be seen that the maximum laser output power, threshold, and quantum differential efficiency at T = 3°C are 1mW, 240mA, and 22%, respectively. The above threshold characteristic temperature (T0) is approximately 45 K and can be due to Auger recombination and inter-valence band absorption effects typical of InGaAsP/InP quantum well lasers. I=376mA I=386mA I=400mA I=395mA I=445mA
2 1.5 1 0.5 0
0.4 0.3 0.2 0.1 00 0.5 1 1.5 2 Time (ns)
0 0.5 1 1.5 2 Time (ns)
5.8 6 6.2 6.4 6.6 6.8 7 Reverse Bias Across Saturable Absorber (V)
300
1ns
1
36ps
0.5
0.5
1
1.5
Time (ns)
2
1 -20
(c) ~1GHz
-60
1.2
1.4
1.6
(d)
-40 -60
0.5
1
1.5
Frequency (GHz)
2
-100 1.034 1.036 1.038 1.04 1.042 1.044
Frequency (GHz)
100 GHz
0 5 GHz
0 -300 -200 -100
1.8
-80
-80
-100 0
-5
Time (ns)
-20 -40
0.1
0.05
0 0
(e)
0.15
1.5
0
100
0.15
0.15
(f) 1GHz 70MHz
0.1 0.05 0
200
-4
Frequency (GHz)
Power (mW)
0
0.2
(b)
Power (mW)
1 0.5
2
Power (mW)
RF = 1.04GHz, 20dBm Injection Current = 395mA Reverse Bias = -6.88V T = 4 C
1.5
0
Power (dBm)
(a)
Power (dBm)
Power (mW)
2
Power (mW)
Fig.2 (a) Experimental arrangement (b) Single sided continuous wave LI curves for the 1-GHz HML laser (c) Pulse width as a function of saturable absorber reverse bias at several gain injection currents at T = 4°C (dashed line) and T = 7°C (solid line).
0
4
Frequency (GHz)
(g)
0.1 0.05 0 -15
0
15
Frequency (GHz)
Fig. 3 (a) Time domain measurement of minimum pulse (36ps) at T = 4°C. (b) Close up of the pulse with trailing satellite pulse. (c) RF spectrum showing 1-GHz spacing. (d) Close up of RF spectrum peak centered at 1.038 GHz with FWHM ~20MHz. (e) Relative optical spectrum of HML from heterodyne detection where 0-GHz corresponds to 1578 nm. (f) Close up of optical modes with ~1-GHz FSR and an average FWHM of 70MHz for each mode. (e) Expanded close up of optical modes.
Leakage currents were measured to be minimal. The lasing threshold is relatively large due to the material loss of the 41mm InP long cavity and low internal quantum efficiency. Standard Fabry-Perot based waveguide loss measurements show a loss of 2 dB/cm and 0.69 dB/cm for 2.5 μm and 5.0 μm wide waveguides respectively. In order to determine the minimum pulse width, we optimize the gain current injection and absorber reverse bias and compare mode-locking performance at a temperature of T = 4°C and T = 7°C as shown in Fig. 2(c). A 20dBm RF drive signal was held constant at 1.0425-GHz. One can observe that the pulse width decreases monotonically as the reverse bias and current injection are increased. Stronger SA bias decreases the carrier sweep out time so that the
OSA / IPR/PS 2010 a326_1.pdf PWD2.pdf
temporal pulse can be decreased. Stronger injection current is then required because of increased loss at faster absorber recovery times. A minimum pulse width (36ps) can be obtained at T = 4°C with 6.87V of reverse bias and 395 mA of injected current. A close-up of the minimum pulse width obtained at T = 4°C is shown in Fig. 3(a) and (b). The pulse is accompanied by a longer low amplitude satellite pulse which may arise from the coupling between the relaxation oscillation and gain recovery process (1ns) being comparable to the 1-GHz repetition rate. Pulse reshpaing dynamics may also play a role because of the long quantum well region (4000μm). Operation in the mode-locked regime can be identified in the RF spectrum if the heights of the main resonant peak compared to the most pronounced peak at lower frequencies is SNR > 25dB. Fig. 3 (c) and (d) show a carrier-to-noise ratio (CNR) in excess of 50dB at a 10-MHz resolution bandwidth. In addition, stable mode-locking is indicated by the narrow width of the RF peak (~1-MHz at -20dB from the peak). Further proof of stable mode-locking is indicated by the narrow widths of the optical modes (~70-MHz) in the optical spectrum using swept heterodyne detection. 5. 1-GHz Hybrid Mode-Locked Laser Integration with Deeply Etched Mirrors (DEM) Successful integration of the HML laser depends on the bandwidth requirements. Bandwidth constraints also eliminate the use of low-index contrast DBR gratings. The simple approach of using the Fresnel reflection between air and InP (~31%) as a cavity is inadequate for efficient coupling between the laser and waveguide. The proposed approach here is based on photonic crystals which use high index contrast between air and semiconductor. This is realized by fabricating a 1-D air slot which can achieve ~70% reflectivity. The DEM is fabricated by using a focused ion beam and an inductively coupled plasma reactive ion etch (ICP-RIE) machine [6]. Fig. 4(a) shows a few SEM pictures of the fabricated slot mirrors for laser integration. (a)
(b) 3.02μm 400nm 334nm
3.02μm 400nm 334nm 266nm 266nm
(d)
MZI Sampling module
(c)
EA
Input Signal
SOA
1Ghz HML laser
Photonic Crystal Mirrors
Fig. 4 (a,b) SEM picture of the side view of the DEM (c) Schematic and SEM picture of a 1-GHz HML with DEM integrated with a proposed sampling core consisting of electro-absorption (EA) modulators and SOAs (d) 3D-FDTD simulation of DEM mirror reflectivity vs. air slot width at an operating wavelength of 1550nm
A 3D-finite-difference-time-domain (FDTD) simulation was used to study the reflectance and transmittance vs. the air gap spacing of our waveguide as shown in Fig. 4(d). The optical sampling core is highlighted by the MZI structure in Fig. 4 (c). In the absence of any control pulses from the 1-GHz HML, the optical signal in the sampling core propagates through and is destructively cancelled at the sampler output. The 1-GHz HML provides control pulses that create a differential phase shift in the SOA or EA that momentarily causes the interferometer to move from the destructive output state to the through state, thus providing sampling of the input signal. In essence, the MZI modulators serve as a time gating with the temporal pass window set by the presence of the 1-GHz HML pulses, EA phase shifts, and the recovery times of the SOA and EA modulators. 6. Conclusion We have demonstrated a 1-GHz HML on an InP platform. HML operation is evidenced by the RF spectrum, optical spectrum, time domain, and swept heterodyne detection. The DEM mirrors offer a viable solution for on-chip laser integration for an ultra-compact, monolithically integrated sampling module which can benefit photonic ADCs. Acknowledgements The authors would like to thank Tiehui Su for the SEM pictures. References [1] W. Cong, et al., "Demonstration of 160- and 320-Gb/s SPECTS O-CDMA network testbeds," IEEE Photon. Technol. Lett., vol. 18, pp. 1567-1569, August 2006. [2] W. Xu, et al., "Field trial of 3-WDM * 10-OCDMA * 10.71-Gb/s asynchronous WDM/DPSK-OCDMA using hybrid E/D without FEC and optical thresholding," J. Lightw. Technol., vol. 25, pp. 207-215, January 2007.
OSA / IPR/PS 2010 a326_1.pdf PWD2.pdf
[3] [4] [5] [6]
G. C. Valley, "Photonic analog-to-digital converters," Opt. Express, vol. 15, pp. 1955-1982, 2007. J. Chen, et al., "Monolithically integrated InP-based photonic chip development for O-CDMA systems," IEEE J. Sel Topics Quantum Electron., vol. 11, pp. 66-77, January-February 2005. F. M. Soares, et al., "Compact InP-based 16-channel O-CDMA encoder/decoder," in LEOS, 2007, pp. 723724. N. K. Fontaine, et al., "Monolithically integratable colliding pulse modelocked laser source for O-CDMA photonic chip development," OFC/NFOEC 2008. 2008 Optical Fiber Communication Conference/National Fiber Optic Engineers Conference, pp. 705-7, 2008.
This work was supported in part by the DARPA DSO and SPAWAR under OAWG contract HR0011-05-C-0155.