00001.tif

PDF

Download

Open document

Flip pages

Contact Us

Copy asset link

Request this asset

Transcript (if available)

Content INFORMATION TO USERS This manuscript has been reproduced from the microfilm master. UMI films the text directly from the original or copy submitted. Thus, some thesis and dissertation copies are in typewriter face, while others may be from any type of computer printer. The quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, print bleedthrough, substandard margins, and improper alignment can adversely affect reproduction. In the unlikely event that the author did not send UMI a complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note will indicate the deletion. Oversize materials (e.g., maps, drawings, charts) E ire reproduced by sectioning the original, beginning at the upper left-hand comer and continuing from left to right in equal sections with small overlaps. Each original is also photographed in one exposure and is included in reduced form at the back of the book. Photographs included in the original manuscript have been reproduced xerographically in this copy. Higher quality 6” x 9” black and white photographic prints are available for any photographs or illustrations appearing in this copy for an additional charge. Contact UMI directly to order. UMI A Bell & Howell Information Company 300 North Zeeb Road, Ann Arbor MI 48106-1346 USA 313/761-4700 800/521-0600 1.3 |Lim and 1.55 |Lim InGaAsP-InP Qtiantam Well Lasers by A tul Mathur A Dissertation Presented to the FACULTY OF THE GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY May 1995 Copyright 1995 Atul Mathur UMI Number: 9617119 UMI Microform 9617119 Copyright 1996, by UMI Company. All rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code. UMI 300 North Zeeb Road Ann Arbor, MI 48103 UNIVERSITY OF SOUTHERN CALIFORNIA THE GRADUATE SCHOOL UNIVERSITY PARK LOS ANGELES. CALIFORNIA 90007 This dissertation, written by fjjTUL M ajhor ................................... under the direction of h£f........ Dissertation Committee, and approved by all its members, has been presented to and accepted by The Graduate School, in partial fulfillment of re quirements for the degree of DOCTOR OF PHILOSOPHY Dean of Graduate Studies Date DISSERTATION COMMITTEE Chairperson Acknowledgments I wish to express my gratitude to Prof. P. Daniel Dapkus for his guid ance and support during this work. His vast experience, excellent advice and insights, friendly attitude and ability to motivate were of immense help both during successful and unsuccessful periods of research. Dr. Serge Dubovitsky and Prof. William H. Steier's collaborative work on polarization insensitive amplifiers is part of this thesis. Serge was always willing to discuss issues, help solve problems and exchange ideas during my entire stay at USC. Dr. Steven P. DenBaars of the University of California at Santa Barbara provided a lot of useful information about starting a new MOCVD system and the use of TBAs and TBP. Dr. Vijay Jayaraman and Prof Larry A. Coldren of the University of Cal ifornia at Santa Barbara, Dr. Deborah Crawford of the Jet Propulsion Labora tory and Dr. Hanmin Zhao at USC collaborated with me in the fabrication of DFB and DBR lasers. Free flow of samples and information with all of them was responsible for the success of this project. Dr. Piotr Grodzinski, Dr. Julian S. Osinski and Dr. Yao Zou initiated the Indium Phosphide effort at USC and trained me in crystal growth, processing and characterization techniques. Their helpful attitude and sharing of useful information made it possible to accomplish a lot more than what would oth erwise be possible. Michael MacDougal did some of the doping measurements for this work, helped with reactor maintenance and start-up of the new MOCVD sys tem, was a constant source of information and help and was always a very friendly presence in the group. Table of Contents Chapter 1 Introduction........................................................................ 1 1.1 Optical Fiber System s.......................................................... ............ 1 1.2 InP/InGaAsP Material System................................................... . 2 1.3 Use of Strain in Quantum Wells.................................................... 3 1.4 Distributed Feedback (DFB) Lasers................................ .............5 Chapter 2 Experimental Techniques........................................... 9 2.1 Crystal Growth............................................................................ . 9 2.1.1 MOCVD System for 1.3 pm Materials.............................. 11 2.1.2 MOCVD System for 1.55 pm Materials............... 12 2.2 Material Characterization........................................................... 13 2.2.1 Double Crystal X-Ray Diffraction (DCXRD) ------ 14 2.2.2 Photoluminescence (PL) — 16 2.2.3 Scanning and Transmission Electron Microscopy (SEM & TEM)...............................................................................17 2.2.4 Polaron C-V Measurement............................................ IS 2.3 Device Processing......................................................................... IS 2.4 Laser Testing............................................................................... 20 2.5 Fabrication of DFB Lasers............................................................. 20 Chapter 3 1.3 pm Quantum Well Lasers............................................— 23 3.1 Gain Analysis in Quantum Well Lasers........................... 23 3.2 Materials for 1.3 pm Quantum W ells................................. 34 3.3 Optimization of the Growth of Interfaces in Quantum W ells. 36 3.4 Broad Area Lasers........................................................................ 39 3.5 Discussion of Fabry-Perot Laser Results........................... 45 3.6 Distributed-Feedback (DFB) Lasers...................................... 47 Chapter 4 1.55 pm Quantum Well Lasers...............................................53 4.1 Survey of 1.55 pm Laser Results.................................................... 53 4.2 Growth of Bulk Layers..................................................................... 56 4.3 Growth of Quantum Wells.............................................................58 4.4 Broad Area Lasers............................................................................ 63 4.5 Discussion of Laser Results............................................................ 70 4.6 Distributed Feedback (DFB) and Distributed Bragg Reflector (DBR) Lasers...................................................................................... 76 4.6.1 High Speed DFB Lasers......................................................... 76 4.6.2 Tunable DFB Lasers................................................................ 80 4.6.3 Broad Tuning Range Using Sampled Gratings................. 81 Chapter 5 Polarization Insensitive Devices............................................ 87 5.1 Polarization Insensitive Amplifier Schemes............................... 8 8 5.2 Active Regions Containing Compressive and Tensile Strained Quantum W ells................................................................. 89 5.3 1.3 pm Dual Polarization Laser..................................................... 92 5.4 1.3 pm Polarization Insensitive Amplifier................................... 94 5.5 1.55 pm Dual Polarization Laser................................................... 96 5.6 1.55 pm Dual Polarization Laser Analysis................................... 100 5.7 Conclusions........................................................................................ 108 Chapter 6 Conclusion................................................................................. I l l 6.1 Summary............................................................................................. I l l 6.2 Future Work....................................................................................... 112 v List of Figures Figure 1.1: Effect of biaxial strain (compressive and tensile) on the band-struc- ture of a direct band gap semiconductor................................................................ 4 Figure 2.1: Experimental x-ray diffraction rocking curve of a seven quantum well InGaAsP - InP structure with 0.85 % strain...................................................15 Figure 2.2: Schematic of a processed laser............................................................ 19 Figure 3.1: Effect of compressive and tensile strain on the band structure of bulk material with same unstrained band-gaps................................................... 25 Figure 3.2: Transition element strength in three orthogonal directions in a quantum well can be calculated by multiplying the optical matrix element by the appropriate factor from this figure.................................................................... 28 Figure 3.3: Calculated gain spectra of a 1 % compressive strained quantum well laser at three different injection current densities................................................ 30 Figure 3.4: Calculated peak modal gain of 1.3 pm quantum well lasers as a function of radiative current density for different strains...................................31 Figure 3.5: Transparency current and threshold radiative current density for 1.3 pm single quantum well lasers having different strains.............................32 Figure 3.6: Calculated modal differential gain as a function of modal peak gain for 1 0 0 A single quantum well lasers with different strains emitting at 1.30 p m .................................................................................................................................... 33 Figure 3.7: X-Ray rocking curve of an InGaAsP (-1.2 %) - InP five quantum well structure with 62 A wells and 1130 A period.........................................................35 Figure 3.8: Calculated emission wavelengths as a function of well thickness for unstrained, 0.85 % compressive strained and 1.2 % tensile strained quantum w ells.................................................................................................................................36 Figure 3.9: PL linewidth at 5 K for compressive strained quantum well grown with different growth interruption schem es..........................................................37 Figure 3.10: PL linewidths at 5 K for different well widths of lattice-matched, 0.85 % compressive strained and 1.2 % tensile strained quantum wells grown with optimal growth interruption schem e.............................................................39 Figure 3.11: Laser structure for two quantum well 1.3 pm devices..................40 vi Figure 3.12: Threshold current density as a function of cavity length for lattice- matched, 0.85 % compressive strained and 1.2 % tensile strained two quantum well lasers emitting at 1.30 p m ................................................................................. 41 Figure 3.13: Reciprocal of differential efficiency as a function of cavity length for lattice-matched, 0.85 % compressive strained and 1.2 % tensile strained two quantum well lasers.................................................................................................... 42 Figure 3.14: Modal threshold gain as a function of current density for lattice- matched, 0.85 % compressive strained and 1.2 % tensile strained two quantum well 1.30 pm lasers...................................................................................................... 43 Figure 3.15: Calculated non-radiative current density for lattice-matched, 0.85 % compressive strained and 1.2 % tensile strained two quantum well 1.3 pm lasers................................................................................................................................44 Figure 3.16: Temperature dependence of threshold current for lattice-matched, 0.85 % compressive strained and 1.2 % tensile strained two quantum well 1.3 pm lasers........................................................................................................................ 45 Figure 3.17: Lasing spectra of typical DFB ridge waveguide devices showing single mode operation................................................................................................ 49 Figure 4.1: Calculated emission wavelength as a function of well width for InxGaj.xAs quantum wells with lattice -matched InGaAsP (Ag = 1.22 pm) bar riers..................................................................................................................................55 Figure 4.2: Experimental x-ray rocking curve of an InGaAsP/InP six quantum well structure................................................................................................................62 Figure 4.3: Structure of a typical single quantum well laser............................. 63 Figure 4.4: Threshold current density for different cavity lengths for two quantum well 1.55 pm lasers.....................................................................................64 Figure 4.5: Reciprocal of differential efficiency for different cavity lengths for two quantum well 1.55 pm lasers............................................................................ 65 Figure 4.6: Threshold current density for different cavity lengths for single quantum well 1.55 pm lasers.....................................................................................6 6 Figure 4.7: Reciprocal of differential efficiency for different cavity lengths for single quantum well 1.55 pm lasers......................................................................... 67 Figure 4.8: Threshold modal gain for single quantum well lasers...................6 8 Figure 4.9: Temperature dependence of threshold current for two quantum well lasers...................................................................................................................... 69 Figure 4.10: Temperature dependence of threshold current for strained single quantum well lasers.................................................................................................... 70 Figure 4.11: Transparency current density and differential modal gain as a function of strain for single quantum well lasers................................................. 72 Figure 4.12: Calculated peak modal gain as a function of radiative current den sity for 1.55 pm single quantum well lasers with different amounts of strain in the active region............................................................................................................74 Figure 4.13: Non-radiative current density as a function of peak gain for 1.55 pm single quantum well lasers with different amounts of strain.....................75 Figure 4.14: Non-radiative current density as a function of carrier density for 1.55 pm single quantum well lasers with different amounts of strain 75 Figure 4.15: Schematic of a polyimide buried ridge waveguide laser for high speed operation.......................................................................................................... 78 Figure 4.16:1-V and L-I characteristics of a typical DFB laser under continuous excitation........................................................................................................................ 79 Figure 4.17: Lasing spectra of DFB laser at three different current levels 79 Figure 4.18: Schematic of two segment DFB laser................................................ 80 Figure 4.19: Spectra showing tuning of a two segment DFB laser.................. 81 Figure 4.20: Schematic of sampled grating DBR laser.........................................83 Figure 5.1: Polarization insensitive structure using lattice-matched quantum well and tensile strained barrier............................................................................... 89 Figure 5.2: Schematic of polarization insensitive active region using compres sive and tensile strained quantum w ells.................................................................90 Figure 5.3: Calculated gain spectra of active region containing compressive and tensile strained quantum w ells......................................................................... 91 Figure 5.4: Measured L-I curves of a 1625 pm long device emitting in both TE and TM polarizations.................................................................................................. 93 Figure 5.5: Measured TE and TM spectra of a 1625 pm long device at three dif ferent injection levels.................................................................................................. 93 Figure 5.6: Measured TE and TM spectra of a 753 pm long device at four dif ferent injection levels.................................................................................................. 94 Figure 5.7: Gain characteristics of polarization insensitive amplifier 95 Figure 5.8: Pulsed L-I characteristics at three different temperatures of 1.55 pm lasers with compressive and tensile quantum well active regions...................98 Figure 5.9: Lasing spectra of a 1.55 |im ridge waveguide device under pulsed excitation........................................................................................................................ 99 Figure 5.10: Lasing spectra of a 1.55 pm ridge waveguide device under contin uous bias........................................................................................................................99 Figure 5.11: Calculated TE and TM peak modal gain as a function of the radi ative current density at three different temperatures........................................1 0 0 Figure 5.12: Schematic of various carrier related processes in an active region containing tensile and compressive active regions.............................................1 0 2 Figure 5.13: Calculated TE and TM L-I curves for compressive and tensile strained quantum well lasers with different carrier capture rates for the tensile well (1.5 ps and 1.9 ps)..............................................................................................104 Figure 5.14: Calculated TE and TM L-I curves for compressive and tensile strained quantum well lasers with different carrier capture rates for the tensile well (2 ps and 4 ps).................................................................................................... 105 Figure 5.15: Calculated L-I curves of compressive and tensile strained quan tum well laser for carrier escape time constant of 50 psec from the quantum w e ll............................................................................................................................... 106 Figure 5.16: Calculated L-I curves for escape time constant of 9.5 ps for the compressive w ell....................................................................................................... 107 Figure 5.17: Calculated L-I curves for escape time constant of 10 ps from the compressive w ell....................................................................................................... 107 Figure 6.1: Proposed buried heterostructure designs with p-n-p-n blocking di odes and semi-insulating current blocking layers..............................................113 Figure 6.2: Three segment laser structure with electronic control of lasing po larization...................................................................................................................... 115 Figure 6.3: Principle of wavelength conversion in a saturated optical amplifi er ....................................................................................................................................116 ix List of Tables Table 3.1: Summary of device parameters for two quantum well 1.30 pm la sers ................................................................................................................................ 42 Table 4.1: PL linewidths of quantum wells of three different widths for differ ent growth interruption times in Ino.5 3 Gag.4 7 A s /InP ........................................60 Table 4.2: PL linewidths of quantum wells of three different widths for differ ent interruption times in Ing 5 3 GaQ 4 7 A s/In 0 7 7 GaQ 2 3 AS0 4 9 P0 5 1 .....................61 Table 4.3: PL linewidths of quantum wells of two different widths for different interruption times in In0 .^ G a g 2 \ As0 ,72Po.28 / Ino.77Gao.23-^so.49Po.5i ............ 61 Table 4.4: Device parameters for various 1.55 pm quantum well la sers 6 8 x Atul Mathur Prof. P. Daniel Dapkus 1.3 |iim and 1.55 pm InGaAsP-InP Quantum Well Lasers 1.3 pm and 1.55 pm lasers are important components of optical fiber based communication systems. In this work, low threshold current density InGaAsP/InP strained quantum well lasers emitting at 1.3 pm and 1.55 pm were fabricated by optimizing the design of the laser structure and the crystal growth process. Threshold current densities as low as 93 A /cm 2 for 1.55 pm single quantum well lasers and 187 A /cm 2 for 1.3 pm two quantum well lasers were obtained. The 1.55 pm devices were grown by metalorganic chem ical vapor deposition using tertiarybutylarsine and tertiarybutylphosphine as group V sources instead of the more hazardous arsine and phosphine that are conventionally used. Distributed feedback (DFB) lasers emitting at 1.3 pm and 1.5 pm were fabricated with first order gratings and using a two step crystal growth pro cess. Polyimide buried ridge waveguide devices had 3 dB bandwidths of 9 GHz. Two segment DFB lasers with tuning range of 2.5 nm and distributed Bragg reflector (DBR) lasers containing sampled gratings with tuning range of 57 nm were demonstrated. To reduce the strong polarization selectivity of conventional quantum well active region, a novel active region containing both compressive and ten sile strained quantum wells was proposed. Dual polarization lasers emitting at 1.3 pm and 1.55 pm and polarization insensitive amplifiers at 1.3 pm were demonstrated with such active regions. The amplifiers maintain polarization insensitivity even when the optical intensity in the cavity saturates the gain. A theoretical model based on carrier capture and escape processes in quantum wells was developed to analyze the performance of such devices and to help in the design of appropriate active regions. Chapter 1 Introduction 1.1 Optical Fiber System s Single mode optical fibers that are utilized for long haul telecommunica tions have immense transmission bandwidth (more than 10 THz). Currently only a small fraction of the data carrying capability of these networks is being utilized because of limitations of the components used in these systems [1]. A very important component of these systems is the semiconductor diode laser used as the transmitter and is the focus of this thesis. To fully utilize the bandwidth of the fibers and to ease transmission bot tle-neck in a system with multiple nodes, wavelength-division-multiplexing (WDM) is a promising scheme [2]. A WDM network has a number of closely spaced channels, each of which can carry information simultaneously. The spacing and the number of channels that can be used depend on the charac teristics of the components of the system - sources, detectors, amplifiers and multiplexers and demultiplexers. The characteristics of silica fibers offer two operating wavelengths - at 1.312 pm the dispersion of the fiber is zero while at 1.55 pm the loss in the fiber is minimum (0.16 dB/km). In the 1.55 pm window, erbium doped fiber amplifiers [3] can be used to compensate for fiber attenuation and network 1 distribution losses. Currently most of long distance telecommunications is being set up at 1.55 pm wavelength [4] while shorter distance high speed net works are being built at 1.3 pm [5]. In this thesis, semiconductor diode lasers operating at both 1.3 pm and 1.55 pm have been studied. The goal has been to reduce their threshold cur rents and improve the output efficiency by optimizing the design and the growth of the active region. In applications requiring arrays of sources, heat dissipation is a problem and hence low threshold devices are desired. Also, the temperature sensitivity of these lasers makes it desirable that heat genera tion be minimized. Light traveling through an optical fiber drifts in polarization. This requires components used in fiber systems to be polarization insensitive. Semiconductor amplifiers are generally strongly polarization sensitive [6 ]. In this work w e have designed a novel semiconductor gain medium that is polarization insensitive and used it to fabricate lasers operating in two orthogonal polarizations [7] and polarization insensitive amplifiers [8 ]. 1.2 InP/lnGaAsP Material System Several material systems in direct band gap III-V compound semicon ductors can be used for optoelectronic devices operating at 1.3 pm and 1.55 pm wavelengths. These include InxGa1.xAsyPi_y , A^G ayln^.yAs and In^-Ga^ xASySbj.y alloys epitaxially grown on indium phosphide (InP) substrates. Of these the InxGai.xASyPj.y material system is most commonly used. This alloy is not reactive with oxygen or moisture, unlike AlxGayln^.yAs [9], and that 2 permits easy regrowths for distributed feedback lasers, buried heterostruc ture lasers and integration of a variety of components on the same wafer [1 0 ]. The antimony alloys are also very difficult to grow because antimony has very little surface migration during crystal growth [11]. The InxGa^ASyP^y alloys can be lattice-matched to InP, over a wide range of band gaps, from 0.75 eV to 1.35 eV. This includes the energies corresponding to both 1.3 and 1.55 |im operation. The binary compound InP has the largest band gap and the smallest refractive index in this material system, which makes it suitable for the cladding layers in lasers, amplifiers and other wave-guiding structures. Lattice-matched InxGa^xAsyPj.y alloys are used for the waveguide layers. The gain is provided by quantum wells that can either be lattice-matched to the substrate or biaxially strained. 1.3 Use of Strain in Quantum Weils Quantum well lasers offer a number of advantages over bulk double het erostructure lasers such as lower thresholds, higher modulation bandwidth, higher differential gain and lower linewidth enhancement factor [12]. Meta- lorganic chemical vapor deposition (MOCVD) and molecular beam epitaxy (MBE) are two growth techniques that enable growth of high quality quan tum well structures for lasers and other device applications. Biaxial strain in quantum wells has been shown, both theoretically [13,14] and experimentally [15,16] to improve laser performance further - even lower thresholds, reduced non-radiative recombination, higher differen tial gain and higher efficiencies. Biaxial strain modifies the in-plane band 3 structure near the r point such that the asymmetry in electron and hole mass is reduced and the degeneracy between heavy holes and light holes is removed. This is illustrated in figure 1.1. The InxGa1_ xAsvP1 _ v - InP system allows growth of both compressive and tensile strained quantum wells for operation at 1.3 pm or at 1.55 pm. In this thesis, w e have theoretically and experimentally compared the performance of lattice-matched, compressive and tensile strained lasers at 1.3 pm and 1.55 pm. Compressive Strain Lattice Matched Tensile Strain Figure 1.1: Effect of biaxial strain (compressive and tensile) on the band- structure of a direct band gap semiconductor The use of strain in quantum wells affects the polarization of the gain in the medium. Compressive strained devices favor TE polarization and tensile strained devices favor TM polarization. By incorporating both types of wells in the same gain medium, the inherent polarization selectivity of a semicon E E E 4 ductor gain medium can be altered [7,8]. Theoretical and experimental results obtained from structures utilizing such active media will be discussed in this thesis. 1.4 Distributed Feedback (DFB) Lasers A Fabry-Perot laser does not emit light at a single wavelength. The strongest emission intensity is at the wavelength corresponding to the Fabry- Perot mode closest to the gain peak but there is also some stimulated emis sion at other neighbouring Fabry-Perot modes. When such a laser is modu lated, the power distribution in different Fabry-Perot modes changes considerably resulting in reduction of the available bandwidth in an optical fiber. For WDM and high modulation rate applications it is desirable to have a light source with a much higher mode selectivity than a Fabry-Perot laser. This can be accomplished by having a frequency specific feedback element in the laser cavity like a distributed grating as in a DFB laser or a frequency selective mirror as in a distributed Bragg reflector (DBR) laser. This concept was first utilized by Kogelnik and Shank [17] and is described in detail in ref erences [6 ] and [18]. In a DFB cavity, the Bragg condition for constructive interference between forward and backward propagating waves is A = m X / 2 , where A is the grating period, X is the wavelength of light inside the laser cavity and m is the order of the coupling between the grating and the optical wave. Thus a grating period of 2300 A is required for 1.55 pm operation with a first order grating. First order coupling results in suppression of electromagnetic waves 5 in directions other than the laser cavity and hence leads to DFB lasers with the best performance. The magnitude of the feedback in a DFB cavity is characterized by the coupling coefficient k that depends on the shape, depth and period of the cor rugations [6 ]. Coupling coefficients of about 100 cm' 1 can be obtained with 500 A deep first order grating. The mode discrimination in a DFB laser is gov erned by the product of the coupling coefficient K and cavity length 1 . k ! of 1-2 with anti-reflection coated facets and 3-5 with cleaved facets is sufficient to obtain reasonable mode suppression (> 30 dB) in a DFB laser. In this thesis, DFB and DBR lasers fabricated to have high modulation bandwidth and wide tuning range will be described. References 1. C.A. Brackett, "Dense wavelength division multiplexing networks: principles and applications/' IEEE J. Selected Areas Comm., vol. 8 , no. 6 , pp. 948-964, Aug. 1990'. 2. B.E.A. Saleh and M.C. Teich, Fundamental of Photonics, John Wiley, 1991. 3. A. Bjarklev, Optical Fiber Amplifiers: Design and System Applications, Artech House, 1993. 4. K. Ogawa, "10 G b/s long-haul experimental systems," IEEE LEOS Annual Meeting, paper OC 5.2, Boston, Oct. 1994. 5. K. Wakao and H. Imai, "Ultra-low threshold 1.3 pm strained-layer quantum well lasers for interconnection," IEEE LEOS Annual Meeting, paper SL 12.1, Boston, Oct. 1994. 6 . G.P. Agrawal and N.K. Dutta, Semiconductor Lasers, Van Nostrand Reinhold, 1993. 6 7. A. Mathur and P.D. Dapkus, "Polarization insensitive strained quantum well gain medium for lasers and optical amplifiers," Appl. Phys. Lett., vol. 61, no. 24, pp. 2845-2847, Dec. 1992. 8 . S. Dubovitsky, A. Mathur, W.H. Steier and P.D. Dapkus, "Gain saturation properties of a polarization insensitive semiconductor amplifier implemented with tensile and compressive strain quantum wells," IEEE Photon. Technol. Lett., vol. 6 , no. 2, pp. 176-178, Feb. 1994. 9. C-E. Zah, R. Bhat, B.N. Pathak, F . Favire, W. Lin, M.C. Wang, N.C. Andreadakis, D.M. Hwang, M.A. Koza, T-P. Lee, Z. Wang, D. Darby, D. Flanders and J.J. Hsieh, "High-performance uncooled 1.3 |im A l^ ayln ^ x _ yA s/InP strained-layer quantum-well lasers for subscriber loop applications," IEEE J. Quantum Electron., vol. 30, no. 2, pp. 511-523, Feb. 1994. 10. U. Koren, "Advanced integrated laser designs," IEEE LEOS Annual Meeting, paper IO 5.1, Boston, Oct. 1994. 11. M. Behet, B. Stoll, W. Brysch and K. Heime, "Growth of GaSb and InSb by low-pressure plasma MOVPE," J. Crystal Growth, vol. 124, pp. 377-382, 1992. 12. Y . Arakawa and A. Yariv, "Quantum well lasers - gain, spectra, dynamics," IEEE J. Quantum Electron., vol. 22, no. 9, pp. 1887-1899, Sep. 1986. 13. A.R. Adams, "Band structure engineering for low-threshold high- efficiency semiconductor lasers," Electron. Lett., vol. 22, pp. 249-250, Feb. 1986. 14. E. Yablonovitch and E.O. Kane, "Band structure engineering of semiconductor lasers for optical communications," J. Lightwave Technol., vol. 6 , pp. 1292-1299, Aug. 1988. 15. M. Yamamoto, N. Yamamoto and J. Nakano, "MOVPE growth of strained InAsP/InGaAsP quantum-well structures for low-threshold 1.3-gm lasers," IEEE J. Quantum Electron., vol. 30, no. 2, pp. 554-561, Feb. 1994. 16. P.J.A. Thijs, L.F. Tiemeijer, J.J.M. Binsma and T. van Dongen, "Progress in long-wavelength strained-layer InGaAs(P) quantum-well semiconductor lasers and amplifiers," IEEE J. Quantum Electron., vol. 30, no. 2, pp. 477-499, Feb. 1994. 7 17. H. Kogelnik and C.V. Shank, "Stimulated emission in a periodic structure," Appl. Phys. Lett., vol. 18, pp. 152-154,1971. 18. J. Buus, Single Frequency Semiconductor Lasers, SPIE Press, 1990. 8 Chapter 2 Experimental Techniques In this chapter, the various experimental techniques used in this work will be briefly described. The experimental work can be divided into crystal growth, material characterization, device processing and device testing. AH crystal growth was done by metalorganic chemical vapor deposition (MOCVD) using two different growth systems. Bulk and quantum well epil- ayers grown on indium phosphide (InP) substrates were characterized using double crystal x-ray diffraction (DCXRD), photoluminescence (PL), scanning and transmission electron microscopy (SEM and TEM) and capacitance-volt- age (C-V) profiling. Lasers and amplifiers were fabricated using standard chemical etching, patterning and metallization techniques. Light-current (L-I) characteristics and emission spectra were then measured for various devices. Fabrication of distributed feedback (DFB) lasers will also be discussed. 2.1 Crystal Growth MOCVD, first used by Manasevit [1] for epitaxial crystal growth, is now the dominant technique for both research and production of optoelectronic devices that require phosphorus containing compounds. Compared to liquid phase epitaxy, MOCVD yields very abrupt interfaces that are essential for 9 successful operation of quantum well devices. Molecular Beam Epitaxy (MBE) with phosphorus source is extremely difficult due to high vapor pres sure of P2 and P4 while in MOCVD the use of phosphorus poses no special problem. In this work, two different MOCVD systems have been used for the growth of 1.3 pm and 1.55 pm materials and devices. Both systems have hori zontal gas flow through a 2 cm diameter quartz reaction chamber. The reac tion chamber is designed to have laminar gas flow in the growth zone after some initial turbulence to promote reactant mixing. The substrate is placed on a graphite susceptor that is heated by a halogen lamp. A thermocouple placed into the susceptor measures the temperature and provides closed loop feed back to the heating lamp power supply. This allows reproducible control of the growth temperature. A vent-run manifold is utilized to ensure quick transitions in growth of different compounds. Group III sources and the p-dopant (Indium, Gallium and Zinc) and Group V sources and the n-dopant (Phosphorus, Arsenic and Silicon) are brought into separate manifolds and mixed just prior to entry into the reaction chamber to avoid any pre-reactions. Hydrogen gas, purified by diffusion through a heated palladium membrane, is used as the carrier. Mass flow controllers are used to measure and control gas flow rates. Hydrogen gas flowing through bubblers filled with volatile metal-organic sources gets satu rated with source vapors and transports them to the reactor. 10 The reactants diffuse through a stagnant boundary layer to the hot sur face of the substrate and their pyrolysis in the hot zone results in the libera tion of the elements that are incorporated into the crystal lattice [2 ]. The differences in the two systems that impact the crystal growths done in this work will now be discussed separately. 2.1.1 MOCVD System for 1.3 pm Materials The MOCVD system used for 1.3 pm materials is a home-built system employing two Thomas Swan vent-run manifolds [3]. The sources used in this system are trimethylgallium (TMGa), trimethylindium (TMIn), phos- phine (PH3), arsine (AsH3), 100 ppm disilane in hydrogen and 250 ppm dime- thylzinc in hydrogen. AsH 3 and TMGa have double-dilution systems that allow variation of the effective flow of these sources by two orders of magni tude. All growths, except for the regrowths for distributed feedback (DFB) lasers, were done at 625 °C temperature and atmospheric pressure with a car rier flow rate of 5 slpm. Regrowths for DFB lasers were done at 600 °C to min imize the planarization of DFB grating. Growth of InP and InxGaj.xAsyP].y alloys was done with a gas phase molar ratio of Group V sources to Group III sources of 75-300 for different compositions. This growth system is a relatively simple system with only one source and gas line into the manifold for each element. This requires changes in gas flow at the transition from growth of one composition to another, which takes several seconds to occur and stabilize. During this time, the sample surface must suffer minimal degradation or composition change. Hence to grow 11 quantum well structures with abrupt interfaces, it is important to carefully study and optimize the growth interruption scheme. This will be discussed in more detail in the next chapter. 2.1.2 MOCVD System for 1.55 pm Materials This system is a state-of-the-art commercial 1" wafer MOCVD system built by Thomas Swan. The sources used are TMGa, TMIn, tertiarybutylarsine (TBAs), tertiarybutylphosphine (TBP), dimethylzinc (DMZn) and 100 ppm disilane in hydrogen. All sources, except TMIn have double-dilution arrange ment for mass flow controllers that allows a large range for variation of gas flows. Ultrasonic monitors are used to measure the concentration of each source in the gas phase. Each element has two inputs into the vent-run mani fold, allowing very fast transition between different compositions by merely substituting the gas lines being input to the reactor. The growth conditions used in this system were 640°C temperature, 76 torr pressure and a total carrier flow of 4 slpm. Low pressure operation results in a much higher gas velocity through the quartz tube and that offers the advantages of very fast transitions, a more uniform stagnant layer thickness over a larger area resulting in better thickness and composition uniformity and the ability to do selective area growth on substrates patterned with masks prior to growth [4]. 12 2.2 Material Characterization Structures containing bulk and quantum well layers were grown on commercially available InP substrates and were then characterized by x-ray diffraction and photoluminescence for measurement of composition, by elec tron microscopy for measurement of thickness and by a Polaron electrochem ical C-V system for measurement of background and intentional doping levels. Each of these techniques will be briefly discussed in this section. The lattice-constant of InxGa-^AsyPj.y alloys is given by linear interpo lation of the lattice constants of the four binaries [5]: a(x, y) = xya(InAs) + x(l - y)a(InP) + (1 - x)ya(GaAs) + (1 - x)(l - y)a(GaP) (Eq. 2-1) The four binary lattice constants are: 6.0583 A for InAs, 5.8687 A for InP, 5.6533 A for GaAs and 5.4505 A for GaP [6 ]. The band gap of this alloy is also given by linear interpolation of the band gaps of the binaries and correction terms for non-linearity [7]: Eg (x,y) = xyEg(InAs) + x(l - y)Eg(InP) + (1 - x)yEg (GaAs) +(1 - x)(l - y)Eg(GaP) + x(x - l)[yCIn_Ga(InGaAs) +(l-y)Cin-Ga(InGaP)] + y (y -l)[x C As_P(GaAsP) ( q' ' > +(1 -y)C A s-p(InAsP)] where Cjn.Ga(InGaAs) etc. are the band gap non-linearity terms. The binary values are listed in reference [5] and the non-linearity parameters in reference [71- Thus experimental measurement of lattice constant and band gap of an InxGaj.xASyPj.y layer is required to determine the composition of the alloy. 13 The lattice constant can be measured by double crystal x-ray diffraction and band gap by photoluminescence. 2.2.1 Double Crystal X-Ray Diffraction (DCXRD) DCXRD is used to measure the lattice constant of an epilayer by using Bragg's Law. In a double crystal system, the x-ray beam is first diffracted by a high quality reference sample and then by the sample being measured. The measurement is then free of effects due to wavelength dispersion of the initial x-ray beam. In our system, Cu Koq x-rays are used alongwith GaAs as the first crystal. The (400) reflection is used for most measurements as it is the most intense and the planes are parallel to the surface of the sample. When an InxGa^ASyP^y layer is grown on InP substrate, a typical DCXRD spectrum has two peaks - one from the epilayer and the other from the substrate. The difference in lattice constant between the two layers Aa is then given by: A a= _ s in 0 B _ _ 1 (Eq. 2-3) a sin(9g + A9) where a is the lattice constant of the substrate, 9g is the corresponding Bragg angle and A9 is the angular separation of the two peaks. The layer thickness should be sufficiently large so that no elastic strain is present. This can be ver ified by measuring the diffraction from an asymmetric plane like the 511 fam ily [8 ]. A mismatched epitaxial layer is usually tilted with respect to the substrate, hence two measurements are done with the sample rotated azi- muthally by 180° to obtain the true mismatch [4]. 14 We have also used DCXRD on multiple quantum well samples to mea sure the strain and the period of the structure [4]. A typical spectra is shown in figure 2.1 and contains a number of satellite peaks. The periodicity of the satellites gives the period of the well-barrier structure while the separation of the Oth order peak from the substrate peak measures the strain in the wells. The structure is designed such that critical thickness for double-kink relax ation is not exceeded [9,10]. The sharpness and the number of satellite peaks gives qualitative information on the abruptness of the interfaces and the uni formity of the structure over a large number of periods. InGaAsP (+0.85%) - InP 7 QWs A = 625A 53A Well Substrate 4 10 -2000 -1000 0 Angle (arcsec) 1000 2000 Figure 2.1: Experimental x-ray diffraction rocking curve of a seven quantum well InGaAsP - InP structure with 0.85 % strain 2.2.2 Photoluminescence (PL) In photoluminescence, a sample is excited with an external optical source that disturbs the equilibrium state of carriers. The excess carriers recombine radiatively emitting light which is analyzed as a function of energy. In our case, the excitation source is the 5145 A line of an argon (Ar+) laser. The emission from the sample is spatially dispersed by a scanning spec trometer and then detected by a liquid nitrogen cooled germanium (Ge) detector. The sample can be mounted either at room temperature or at 5 K in a liquid helium cryostat. Room temperature PL spectra of direct band gap bulk InxGa^ASyP^y layers gives an emission peak corresponding to band-to-band transition. This allows the measurement of the band gap of the material. The intensity of the emission indicates the quality of the epilayer. The presence of misfit disloca tions result in a significant decrease in the emission intensity. PL spectra of quantum well samples exhibit peaks corresponding to allowed transitions involving the bound states of the quantum wells. If the composition of the well and barrier layers is known, then the well thickness can be calculated from the PL spectra. For the fabrication of DFB lasers, the active region of the laser was grown in the first crystal growth step. Then room temperature PL emission wavelength from the active region was measured and used to calculate the period of the grating that would be etched on the sample. This procedure enabled us to match the grating pitch to the gain peak of the quantum wells despite any variations that may happen in the crystal growth process. 16 At 5 K, thermal broadening of the emission spectra is negligible and the line width of the emission depends on the abruptness of the interface between the well and the barrier. This measurement has been used in this work to opti mize the growth interruption between well and barrier and will be discussed in the next two chapters in more detail. 2.2.3 Scanning and Transmission Electron Microscopy (SEM & TEM) In a SEM, a sample is excited by an high energy electron beam (10-20 kV) and the secondary electron emission from the sample is then detected to look at surface steps or chemical composition variations. SEM measurement of cross-section of epilayers has been used to measure layer thicknesses that are larger than about 2000 A and to study etching and growth profiles. In a TEM, a very high energy electron beam (120 - 200 kV) is incident on a very thin sample (about 200 A thick) and the transmitted beam is used to image the sample with monolayer resolution. TEM has been used for mea surement of quantum well thicknesses and to qualitatively investigate the abruptness of interfaces between different layers. Samples were prepared and mounted such that the 0 1 1 crystal direction was parallel to the electron beam. After obtaining the diffraction pattern for the 011 zone axis, the sample was tilted to establish two beam condition, where 011 and 200 diffraction spots are most intense [11]. This condition provides the best chemical contrast between layers of different compositions. 17 2.2.4 Polaron C-V Measurement C-V measurement in a Polaron electro-chemical cell has been used to measure both the background carrier density in various layers and the inten tional doping achieved with our p and n dopant sources. The Polaron mea sures the sample capacitance as a function of bias as it is etched in the region of contact with an electrolyte, thus generating a profile of doping versus the etch depth. The electrolyte used was 0.1 molar hydrochloric acid for both InP and IriQ 5 3 Ga0 . 4 7 As layers. Background carrier concentrations as low as 1 X 101 4 cm ' 3 for InP grown with PH3, 1 X 101 5 cm' 3 for InP grown with TBP and 1 X 101 6 cm ' 3 for Ir tQ 5 3 Ga0 47As were obtained, which demonstrates the high purity of the crys tal growth systems and the sources used as well as successful optimization of growth conditions. 2.3 Device Processing The cross-section schematic of a typical laser or amplifier is shown in fig ure 2.2. The stripe width is typically 55 pm for broad area devices fabricated for material evaluation and 3 pm for ridge waveguide devices that support a single lateral optical mode. The processing is done using standard photoli thography, wet etching and metallization procedures [1 2 ] which are briefly described here. The first photolithography mask uses stripe widths 2-3 pm wider than the desired active region widths. The top p+ Ino.5 3 Gao.4 7 As layer is etched 18 using 1:1:10 sulfuric acid-hydrogen peroxide-water. The etchant undercuts the photoresist mask, so the stripe width can be controlled by the etch time. Then the p InP cladding layer is etched all the way down to the waveguide using 3:1 hydrochloric acid-water. A 1000 A thick Si3 N 4 insulator layer is deposited on the sample using plasma assisted chemical vapor deposition. The Si3 N 4 layer is then opened in stripes on the top of the ridge structures by CF4 plasma etching. Au-Zn alloy is electroplated through the Si3 N 4 mask. The sample is then lapped down to 100 pm thickness and the back side is elec troplated with Au-Sn alloy. The p and n contacts are alloyed in hydrogen atmosphere at 400 °C for one minute to get good ohmic contacts. More Au is then electroplated on both sides of the sample. A 200 pm wide Ti-Pt-Au pad is evaporated over the alloyed stripe to create a large area for probing. The sam ple is then cleaved into bars of different lengths. Au contact p+ InGaAs Silicon Nitride Waveguide n-lnP Figure 2-2: Schematic of a processed laser 19 2.4 Laser Testing The laser bars are tested by probing the p contact pad of individual devices with the n side down on a metal mount. Light-current (L-I) character istics are measured using a large area InGa As photodetector. The lasers can be tested under pulsed (typically 100 nsec pulses at 10 kHz repetition rate) exci tation or continuous bias. The 55 pm wide broad area lasers operate only under pulsed condition, as heat dissipation under continuous bias is a prob lem. L-I measurements give the threshold current and the differential effi ciency of the lasers. The temperature of the laser mount can be controlled from 0 °C to 50 °C. The emission spectra of the devices is measured using a spectrometer and a Ge detector. Polarization of the emitted light can be analyzed in either L-I or spectra measurement using a polarizing beam splitter. 2.5 Fabrication of DFB Lasers For fabrication of DFB lasers the crystal growth is done in two steps. In the first step, the n-InP layer and the waveguide containing the quantum wells are grown. The quantum well emission wavelength is then measured by PL and used to calculate the desired pitch for a first order DFB grating. Two different techniques were used for grating fabrication - holographic exposure using a helium-cadmium (He-Cd) laser (done at University of California, Santa Barbara by Dr. Vijay Jayaraman and Prof. Larry A. Coldren) and elec- tron-beam lithography (done at Jet Propulsion Laboratory, Pasadena by Dr. 20 Deborah Crawford). He-Cd holography allows rapid exposure of large sized areas and is hence helpful for the development and optimization of the fabri cation process. Electron beam lithography is a much slower and more expen sive process that can be used only on small areas because of time constraints. However, it is very easy to include quarter wavelength phase-shifts or fabri cate an array of DFB lasers with different elements having different grating pitch using this scheme. The grating was etched into the waveguide 500-750 A deep by wet chemical etcing. The resist used for grating fabrication was removed by oxygen plasma cleaning before the second growth.The heat-up time in the reactor prior to regrowth is kept very short and growth is com menced at 575 °C so as to minimize grating planarization. The p-cladding and contact layers are grown in this step. The device is then processed similar to Fabry Perot lasers as described before. References 1. H.M. Manasevit, "Single-crystal gallium arsenide on insulating substrate," Appl. Phys. Lett., vol. 12, no. 4, pp. 156-159, Feb. 1968. 2. J.P. Hirtz, M. Razeghi, M. Bonnet and J.P. Duchemin, in GalnAsP Alloy Semiconductors, edited by T.P. Pearsall, Wiley, 1982. 3. P . Grodzinski, Ph.D. Dissertation, University of Southern California, 1992. 4. V. Swaminathan and A.T. Macrander, Materials Aspects of GaAs and InP Based Structures, Prentice-Hall, 1991. 5. S. Adachi, "Material parameters of Ini_x GaxASyPi_y and related binaries," J. Appl. Phys.,vol. 53, no. 12, pp. 8775-8792, Dec. 1982. 21 6 . O. Madelung, Semiconductors - Group IV Elements and III-V Compounds, Springer-Verlag, 1991. 7. T. Ishikawa and J.E. Bowers, "Band lineup and in-plane effective mass of InGaAsP or InGaAlAs on InP strained-layer quantum well," IEEE J. Quantum Electron., vol. 30, no. 2, pp. 562-570, Feb. 1994 8 . J. Hornstra and W .J. Bartels, "Determination of the lattice constant of epitaxial layers of III-V compounds," J. Cryst. Growth, vol. 44, pp. 513- 517,1978. 9. J.W. Matthews and A.E. Blakeslee, "Defects in epitaxial multilayers - misfit dislocations," J. Cryst. Growth, vol. 27, pp. 118-125,1974. 10. G.A. Vawter and D.R. Myers, "Useful design relationships for the engineering of thermodynamically stable strained-layer structures," J. Appl. Phys., vol. 65, no. 12, pp. 4769-4773, June 1989. 11. J.W. Edington, Practical Electron Microscopy in Materials Science, N.V. Philips, 1976. 12. J.S. Osinski, Ph.D. Dissertation, University of Southern California, 1992. 22 Chapter 3 1.3 jL im Quantum Well Lasers In this chapter, the theoretical and experimental results obtained with 1.3 pm quantum well lasers will be described. We have developed a simple algo rithm for calculation of gain and spontaneous emission in strained quantum well lasers and have used it to analyze various laser structures for low thresh old operation. The issues involved in the growth of quantum wells in the MOCVD system described in the last chapter will be described. Laser results obtained in this work will be presented, analyzed and compared with other works. The results of distributed feedback lasers containing quantum well active regions will be presented. 3.1 Gain Analysis in Quantum Well Lasers An exact numerical simulation of strained quantum well lasers is extremely computation intensive and requires the knowledge of a large num ber of material parameters that are not well determined. An one-dimensional simulation procedure has been used here that utilizes several assumptions for simplifying the computation while still providing useful information for device design. The procedure will now be briefly described and then the results obtained will be discussed. 23 In these calculations, all the material parameters of InxGai_xASyPi_y are calculated by linear interpolation of the published values for the four binaries (GaAs, InP, InAs and GaP) taken from [1] and [2]. The effect of biaxial strain on the band gap of a material is shown in figure 3.1. Strain has two effects on the band structure - the hydrostatic component moves both the heavy hole and light hole bands equally with repect to the conduction band (compressive strain increases the band gap while tensile strain decreases the band gap) while the shear component causes a splitting of heavy hole and light hole bands at the T point (compressive strain moves heavy hole band above light hole band while tensile strain has the opposite effect). In this figure the three layers are assumed to have the same unstrained band gaps. The quantitative values of these effects are given by the following equations [3]: 5Eh v d - 2 a(l - — )e c,, (Eq. 3-1) 5Es ,l c a r = b(l + 2 ^ ) e (E q 3 .2) where a and b are the hydrostatic and shear deformation potentials, c^ and Ci2 are the elastic coefficients interpolated from the binary values, e is the strain in the layer and SE^y^ and 8 Eshear are the hydrostatic and shear defor mation energies respectively. The net effect on the band gap is given by [3]: E - E - 8 E + ^ s llc a r (Eq. 3-3) e -h h - ^g h y d ^ ^ 8 E Ec-ih ~ E g - 8 Ehyd ^ (Eq. 3-4) where Eg is the relaxed band-gap and Eg.^ and Ec_ u 1 are the electron to heavy hole and light hole energies respectively in the strained layer. Strain removes 24 the degeneracy between heavy holes and light holes at the T point and 8 Es|- ,ear is the measure of the splitting between them.The conduction band offset at all hetero-junctions is assumed to be 39% of the band gap (electron to heavy hole) difference between the two layers. The experimental data as well as cal culations regarding the band offset are summarized in reference [4]. ear No strain Compressive Strain Tensile Strain Figure 3.1: Effect of compressive and tensile strain on the band structure of bulk material with same unstrained band-gaps The effective mass in the valence band of a strained quantum well can be calculated using the k .p method. We have used the analytical solution of Chuang [3] obtained by decoupling the effects of strain and quantum well 25 geometry on the valence band. This gives the following dispersion relation for the valence band in the kx-kz (quantum well is in x-y plane) plane: E = — ■ n2 y, 2 m (ki+kl): 2„, S t : "\ 5C„„ + 3 f j y 2 2m .2 ft. 2 , + 12 ryj 2m (Eq. 3-5) where Yi/ Y 2 and y3 are Luttinger parameters interpolated from binary values and m is the free electron mass. The opposite signs in the equation are for light holes and heavy holes. This anisotropic band equation is numerically approximated near the T point by a parabolic band with two different effec tive masses - one in the x-y plane and the other in the z direction [5]. This approximation simplifies further calculations considerably. The bound levels for electrons, heavy holes and light holes in the quan tum well are calculated using the standard quantum mechanical solution for a bound particle in a finite potential well [6 ]. Only those optical transitions that conserve the quantum number associated with the bound levels are allowed by the selection rules. The carrier population of quantum well levels as well as three dimensional states in the barrier are calculated using standard Fermi-Dirac statistics as a function of quasi-Fermi level position [7]. Electron and hole neutrality is imposed over the entire active region. This gives the position of both electron and hole quasi-Fermi levels as a function of injected carrier density, assuming that carriers in the active region are in steady-state. We have neglected band-gap narrowing with increasing carrier density [8 ] as the parameters involved are not well known and the major effect is in shifting 26 the transition energies and not in the magnitude of gain or spontaneous emis sion. The quantum well gain from each allowed optical transition g(E) is cal culated as a function of energy E as follows [7]: 8(E) = P -(E „ - W - f.) (B <I- 3' 6) C q v I H q I I C / where e is the electron charge, ng is the group index of the optical wave, n is the index of refraction in the crystal, e0 is the permittivity of free space, m 0 is the electron free mass, M f is the dipole matrix element for the transition, pre cl is the reduced density of states for the transition, Eeh is the transition energy, Eg' is the band-gap between the sub-bands of the quantum well and fc and fv are the electron probabilities in the conduction and valence band respectively, given by Fermi-Dirac statistics. The total material gain is then obtained by adding the contribution from each allowed transition. The modal gain is obtained by multiplying the material gain by the optical confinement factor, which takes into account the actual overlap of the electromagnetic field with the active medium [9]. The optical matrix element M is related to conduction band curvature by the following [7]: r \ mn M = I T ■ - 1 (Ee + 2 } mo Es (Eq- 3 -7 ) Vmc 2(Eg+-A ) where mc is the conduction band effective mass, Eg is the direct band gap and A is spin-orbit splitting energy. This optical matrix element M is related to the transition strength by the overlap of the electron orbital momentum vec 27 tor with the electric field vector. The magnitude of the overlap is shown in fig ure 3.2 for different directions in a quantum well. The spontaneous emission from each transition Rgp(E), in units of sec' 1 cm ' 3 energy'1, is given by [7]: iM E>% ™ \ l M. . / p „ l(E.t - e ;)P (E)fta - f . ) (Eq-3-8> £0m0n b where Mave is the average of the transition matrix element MT over all polar izations and popt(E) is the optical mode density. The radiative current density Jr a c j can then be calculated by integrating the spontaneous emission from all possible transitions as follows: = < * - , £ £ R,„(E)dE (Eq. 3-9) n 8 where Lz is the well thickness and Eg' is the energy corresponding to the nth quantized transition. C-HH 0 A C-LH4/3 Quantum well C-HH 1 C-LH 1/3 C-HH 1 C-LH 1/3 Figure 3.2: Transition element strength in three orthogonal directions in a quantum well can be calculated by multiplying the optical matrix element by the appropriate factor from this figure 28 So far the effect of electron-phonon collisions has not been included. These interactions result in broadening of both gain and spontaneous emis sion spectra. This effect has been incorporated into these calculations by con volving the unbroadened spectra with a Lorentzian broadening function L(E) where xin is the interband relaxation time and is usually taken to be 0.1 ps [7]. A detailed discussion which gives a more exact analysis of electron-phonon broadening mechanism is available in reference [1 0 ]. For 1.3 pm - 1.55 pm lasers, non-radiative recombination by Auger pro cesses [9] is a significant part of the total recombination rate. Auger recombi nation is widely taken to depend on the cube of the carrier density [9]. The Auger coefficient has been measured experimentally [11] and calculated theo retically [1 2 ] but enough data or understanding of the process does not exist to obtain Auger recombination rate for different amounts of strain and differ ent well widths for use in the following calculations. Hence in the following calculations only the radiative component of the injection current has been evaluated. An example of calculation of gain spectra is shown in figure 3.3. This calculation is done for a 1% compressive strained 100 A single quantum well laser emitting at about 1.30 pm. The figure shows the TE gain spectra at three different injection current densities. While most of the TE gain comes from electron-heavy hole recombination, the small shoulder seen at 1 . 2 1 pm for the given by: ____ n ' ti n ____ 7 t(E -E ')2 +(fi/Tin)2 (Eq. 3-10) 29 higher injection densities is due to the TE component of electron-light hole recombination. 50 n = 1.5 X 1 0 13 cm n = 4 .0 X 1 0 13 cm n = 7 .5 X 1013 cm 4 0 E o 3 0 c c o 0 " io T3 O 20 10 1 .1 0 1.15 1 .2 0 1.25 1 .3 0 1.35 W avelength (pm) Figure 3.3: Calculated gain spectra of a 1% compressive strained quantum well laser at three different injection current densities From gain-energy plots like this, the peak modal gain can be obtained at different carrier densities. From similarly calculated spontaneous emission, the radiative component of the injection current can also be calculated corre sponding to each carrier density. Thus a plot of peak modal gain as a function of radiative current density can be obtained. To study the effect of strain on gain in a quantum well laser, we have analyzed single well lasers with different strains (lattice-matched and 0.5 %, 1 % and 1.5 % compressive and tensile strained) and 1 0 0 A well widths, all 30 emitting at 1.30 pm. Figure 3.4 shows the calculated peak modal gain as a function of radiative current density for different amounts of strain. 5 0 0% — ■ + 0.5% — + 1.0% — + 1.5% 0.5% 4 0 E o 3 0 c '(0 o > ra ■ o o E ( 0 q 0. - - 1.5% 20 10 -10 0 2 0 4 0 6 0 8 0 1 0 0 R adiative current density (A/cm2) Figure 3.4: Calculated peak modal gain of 1.3 pm quantum well lasers as a function of radiative current density for different strains The plot shows that for low losses (a < 18 cm'1), 1.5 % compressive strained devices will have the lowest radiative current density while for higher losses (a > 18 cm"1), 1.5% tensile strained devices will have lower radiative current. In figure 3.5, the radiative components of transparency current and the threshold current density for a 1 mm long cavity length are plotted for differ ent amounts of strain. Transparency current is reduced by a factor of two from no strain to 1.5 % tensile strain and there is an even larger reduction by 1.5 % compressive strain. The threshold current for 1 mm cavity length is also reduced by both tensile and compressive strains. However, small amounts of 31 compressive strain result in less reduction than equivalent tensile strain due to smaller peak gain at higher losses. At short cavity lengths, tensile strain is expected to offer lower threshold currents than an equal amount of compres sive strain. 8 0 6 0 £ w c 0 ■ o 4 0 D O 0 > 20 4-. ro T ) t o O C 1.30 pm SQW lasers 3 - At tran sp aren cy 3— At th resh o ld (L=1mm) -1.5 - 1.0 -0.5 0.0 0.5 1.0 1.5 Strain in quantum well (%) Figure 3.5: Transparency current and threshold radiative current density for 1.3 pm single quantum well lasers having different strains Figure 3.6 shows the differential gain, defined as the derivative of the peak modal gain with the current density, as a function of the peak modal gain for different amounts of strain. At and above threshold in a laser, the modal gain is pinned at the value of total losses, so this plot offers a compari son of differential gain for quantum wells with different strains operating under the same threshold gain condition. 1.5 % compressive and tensile strained quantum wells have almost equal differential gain but for lower 32 amounts of strain, tensile strained quantum wells have larger differential gain than quantum wells with the same amount of compressive strain. 1.6 1.3 pm SQW Lasers 0 % + 0.5 % + 1.0 % ^ 1.4 E o £ 12 c o O ) m 1 ° (0 - - 1.5 % 0.8 0.6 0 .4 0.2 5 0 5 10 15 20 2 5 Modal peak gain (cm 1) Figure 3.6: Calculated modal differential gain as a function of modal peak gain for 100 A single quantum well lasers with different strains emitting at 1.30 pm We have also calculated the gain curves for different well widths from 25 A to 150 A using 1 % compressive strain in the quantum well and keeping the emission wavelength fixed at 1.30 pm. There is no significant difference in peak modal gain or transparency current for wells that have only one bound electron state. For wide wells (wider than 100 A in this case) that have two bound electron states, the calculated transparency current is about two times higher than that for the thinner wells that have only one bound electron state. To minimize threshold current density in quantum well lasers, the well width 33 should be chosen such that there is only one bound electron state and biaxial strain should be used to increase the separation of heavy hole and light hole levels. Both of these considerations are important for reducing the transpar ency current in a laser. Non-radiative recombination processes are not well understood, though they are generally assumed to have a polynomial dependence on the carrier density [9]. Thus reduction of threshold carrier density by application of strain and appropriate choice of well width also leads to decrease in non-radi ative recombination rate. In rest of this chapter and the next chapter, experi mental results on 1.3 pm and 1.55 pm lasers will be presented to verify the advantages of strained quantum well active regions. 3.2 Materials for 1.3 pm Quantum Wells The waveguide and barrier layer for 1.3 pm lasers was chosen to be Ino.8 2 Gao,i8 -Aso.3 9 Po.6 i/ which is lattice matched to InP and has a band gap of 1.08 eV. Three different compositions were utilized in the quantum well - ^ no.68^ao.32i^so.70^ *0.30 which is lattice-matched to InP, Ino.8 7 Gao.1 3 Aso.5 5 Po.4 5 which has 0.85 % compressive strain and In0. 4 7 Ga0 .5 3 As0 . 7 7 P0 .2 3 which has 1.2 % tensile strain. The strain in these layers was determined by x-ray diffraction measurement of multiple quantum well structures, as described in the last chapter. The experimental x-ray rocking curve of a seven quantum well Ino.8 7 ^a0,i3 Asq 5 5 P0 4 5 - InP structure is shown in figure 2.1 and that of a five quantum well Ino.4 7 Ga0 5 3 Aso.7 7 Po.2 3 ' InP structure in figure 3.7. 34 Substrate ,4 10 InGaAsP (-1.2%) 5 QWs A =1130 A ,3 1 0 ' ,2 10 m=0 i 10 ,o 10 1 10 -2000 -1000 0 1000 2000 Angle (arc sec) Figure 3.7: X-Ray rocking curve of an InGaAsP (-1.2 % ) - InP five quantum well structure with 62 A wells and 1130 A period Figure 3.8 shows the calculated emission wavelengths corresponding to the transition between the first quantized electron state and the first valence band state as a function of well thickness. The well thickness required for 1.3 pm operation is 55 A for compressive and lattice matched wells and 125 A for tensile well. We have chosen compositions that do not require very thin wells as that would require a very stringent control over the growth of interfaces. The larger well width used for tensile strain is required to achieve reasonable separation between light hole and heavy hole quantized levels as quantum confinement has the opposite effect on splitting of light hole and heavy hole bands than that due to tensile strain. 35 1.40 1.35 E 3 . O ) c Q ) 1 0 3 5 1.30 1.25 c o ' c n m E L U + 0.85 % Strain No Strain -1.2 % Strain 1.20 1.15 0 50 100 150 200 Well thickness (A ) Figure 3.8: Calculated emission wavelengths as a function of well thickness for unstrained, 0.85 % compressive strained and 1.2 % tensile strained quantum wells 3.3 Optimization of the Growth of Interfaces in Quantum Wells Lattice-matched Inx Gai_xAs - InP and h^Ga^ASyPj.y - InP quantum wells have been studied extensively and 2 K PL linewidths of 3.2 meV [13] and 3.5 meV [14] have been reported for 100 A wells. Extremely short growth interruptions were found to give the best linewidths because of exchange of P atoms from the surface with As atoms during interruptions [15]. For the In^ G a^ A syjP j.y! - Inx2Gaj.x2 ASy2 P1.y2 type of wells that are used in our quantum wells, PL linewidths of 7.6 meV at 5 K have been reprted for a 85 A compressive strained well operating at 1.55 pm [16]. 36 To investigate the effect of various schemes for switching gas flows at well-barrier interfaces, a 55 A wide 0.85 % compressive strained quantum well was grown with different growth interruptions. In all cases, the same scheme was used at both lower and upper interface of the well. TMIn, PH3 and AsH3 gas flows were changed at each interface. Growth interruptions from 0.5 sec to 20 sec with gas flow of PH3 or PH3 - AsH3 mixture required for the succeeding layer were investigated. The PL linewidths measured at 5 K for these quantum wells are shown in figure 3.9. > < D E ■ g 5 CD c 14 12 10 8 T = 5K Compressive QW - PH3 + AsH3 purge Compressive QW - PH3 purge 6 4 5 10 15 Growth Interruption Time (sec) 20 Figure 3.9: PL linewidth at 5 K for compressive strained quantum well grown with different growth interruption schemes Shorter interruptions were found to give narrower linewidths, indicat ing that exchange of As and P atoms from the surface during an interruption leads to the formation of compositionally graded interfacial layers. However 37 interruptions shorter than 5 sec resulted in change of emission wavelength of the quantum well, implying that the gas flows did not stabilize at the values needed for growth of the new layer. At these interfaces, the gas flow through In, P and As mass flow controllers and through In bubbler is changed and requires some time to stabilize. Also the gas mixture in die reactor tube needs to change to that required for the next layer. Ponging with a mixture of PH3 and AsH3 was found to give narrower linewidths than purging with PH3 alone, indicating that the use of the gas mixture needed for the next layer reduced the formation of transition layers with graded composition. Based on these results, a 5 sec interrupt with PM3 and AsH 3 flow corre sponding to that required for the growth of the next layer was selected as the best compromise between flow stabilization and surface degradation and was then used for all device growths. Figure 3.10 shows the PL linewidths for lat tice-matched, 0.85 % compressive strained and 1.2 '% tensile strained quan tum wells of different thickness grown under this optimized interruption scheme. The best linewidths that we have obtained are 7.9 meV for a 150 A wide lattice matched well, 6.7 meV for a 130 A w ide compressive w ell and 9.4 meV for a 150 A wide tensile well. Of these three quantum wells, the tensile layer has the largest change in composition from the barrier layer and hence has the largest PL linewidth while the compressive well has the smallest change in composition and hence has the narrowest linewidth. 38 > CD E, JO ■ g 5 CD C 25 T = 5 K - i t - Tensile Strained QWs Lattice Matched QWs -•••- Compressive Strained QWs 20 15 10 5 0 40 80 0 20 60 100 140 120 160 Well Thickness (A) Figure 3.10: PL linewidths at 5 K for different well widths of lattice-matched, 0.85 % compressive strained and 1.2 % tensile strained quantum wells grown with optimal growth interruption scheme 3.4 Broad Area Lasers The laser structure used in this work is shown in figure 3.11. It is a single step separate confinement structure with the waveguide thickness optimized to maximize the transverse confinement factor. The following layers were grown on an n+ (100) InP substrate - 0.6 pm of 1 X 1018 cm'3 n doped InP buffer layer, 1900 A of Ino.82Gao.l8-^so.39Po.6l lower waveguide layer, two quantum wells (lattice matched, 0.85 % compressive strain or 1.25 % tensile strain) with a 150 A barrier, 1900 A of Ing^Gao.isAso.3 9 P0 .6 I upper waveguide layer, 1500 A of undoped InP, 1.5 pm of 6 X 1017 cm"3 p InP cladding layer and 39 2000 A of 5 X 1018 cm'3 p doped In0 6 8 Gao.3 2 ^ so.7 oGao.3 0 contact layer. Broad area lasers were fabricated and pulse tested as discussed in chapter 2. 1750 A InGaAsP (1.15 pm) | 2000 A p - InGaAsP (1.38 urn) p - InP InP r 1500 A __L 1.5 pm InGaAsP (1.15 pm) A T InGaAsP (1.15 pm) ' * f i 7cn A v ^ ’ 150 A i»1750 A n - InP 6000 A Two Quantum Wells Figure 3.11: Laser structure for two quantum well 1.3 pm devices The measured threshold current densities (Jth) are plotted in figure 3.12 as a function of cavity length. The lines connecting the data points are obtained from a semi-logarithmic fit given by the following equation [17]: 1 , . 1. „ | / J (Eq. 3-11) a int + ~ l n(^r)-G o[l + ln(— )] L K J0 Threshold current densities as low as 314 A /cm 2 for 2 mm long lattice matched lasers, 187 A /cm 2 for 3 mm long 0.85% compressive strained laser 40 and 221 A /cm 2 for 3 mm long 1.2% tensile strained laser were measured. The lasing polarization is TE for lattice matched and compressive strained lasers and TM for tensile strained lasers. 2500 C\J E < 2000 □ Lattice Matched O Compressive Strained A Tensile Strained w § 1500 Q c CD O s z c n CD L_ -C h- 500 'tr o 500 1000 1500 2000 2500 3000 Cavity Length (pm) Figure 3.12: Threshold current density as a function of cavity length for lattice-matched, 0.85 % compressive strained and 1.2 % tensile strained two quantum well lasers emitting at 1.30 pm The external quantum efficiency of these devices is plotted in figure 3.13 as a function of cavity length. From the slope and the y-intercept of this plot, the internal quantum efficiency ( r | j n t ) and the optical loss were estimated and are listed in table 3.1. From the data of figures 3.12 and 3.13, the modal gain at threshold can be calculated for devices with different cavity lengths and hence different threshold currents. The gain is plotted as a function of injec- 41 tion current in figure 3.14 and the transparency currents (Jtr) are listed in table 3.1. 3.5 £ 3.0 Q ) o it= Ui o 2.5 (0 o 2 Q . O s . 2 0 □ Lattice Matched O Compressive Strained A Tensile Strained 1000 1500 500 2000 Cavity Length (pm) Figure 3.13: Reciprocal of differential efficiency as a function of cavity length for lattice-matched, 0.85 % compressive strained and 1.2 % tensile strained two quantum well lasers Strain Jth,m in (A /cm ) hint (%) Optical loss (cm'1) Jtr per well (A /cm 2) Modal gain per well (cm"1) T0 (K) 0% 322 52 3.0 85 6.5 44 +0.85 % 187 62 4.5 49 7.6 42 -1.2 % 221 56 4.8 81 12.9 40 Table 3.1: Summary of device parameters for two quantum well 1.30 pm lasers 42 50 □ Lattice Matched O Compressive Strained A Tensile Strained 40 2x10 100 1000 2 Current Density (A/cm ) Figure 3.14: Modal threshold gain as a function of current density for lattice- matched, 0.85 % compressive strained and 1.2 % tensile strained two quantum well 1.30 |im lasers For these three structures, we have calculated the peak modal gain as a function of radiative current density using the procedure described in section 3.1. By subtracting the calculated radiative current density from the measured total injection current density, the non-radiative current density can be obtained and is plotted in figure 3.15. The non-radiative current density has been plotted here as a function of peak modal gain, instead of carrier density, so that a comparison can be made for the three devices at the same threshold gain condition. The non-radiative current density for lattice-matched lasers has a much stronger gain dependence than for the two strained devices and that leads to high threshold currents for short cavity length devices. Non- 43 radiative recombination is suppressed for both 0.85 % compressive strain and 1.2 % tensile strain compared to no strain, leading to lower threshold currents and higher internal efficiency. For cavity lengths below 750 pm, 1.2 % tensile strain devices have less non-radiative current than the other two devices. E 4 0 0 Unstrained + 0.85 % Strain - 1.2 % Strain & in c q T 3 3 0 0 200 o 0 > 1 0 0 c o 0 10 5 15 20 2 5 P eak modal gain (cm"1) Figure 3.15: Calculated non-radiative current density for lattice-matched, 0.85 % compressive strained and 1.2 % tensile strained two quantum well 1.3 pm lasers The temperature dependence of the threshold current density of typical devices is plotted in figure 3.16. The characteristic temperature (T0 ) is obtained from an exponential fit of the threshold current density dependence on temperature and is found to be 40 - 44 K for all three devices. This agrees with other published results that T0 is not increased by biaxial strain [12]. The reasons for the temperature dependence of these devices are not yet fully understood. 44 E o «! c Q Q ■ 4— * C 2 3 o o s z w CD T 1000 - 8 7- 6- 5- Tn = 40 K . - D - Tn = 44 K . . - S ' . -A 0 -- A - " ' A " ' -B ■ Q ......t>—" O ' O-"' .S'" .-O '' ..O ' Tn = 4 2 K ..G " .-O - O '" 1 0 0 _ L □ Lattice Matched O Compressive Strained A Tensile Strained 10 20 30 Temperature (°C) 40 50 Figure 3.16: Temperature dependence of threshold current for lattice- matched, 0.85 % compressive strained and 1.2 % tensile strained two quantum well 1.3 pm lasers 3.5 Discussion of Fabry-Perot Laser Results The threshold current densities obtained in this work were lower than any other published results at the time this work was completed. This was achieved in a very simple MOCVD reactor by very careful optimization of the crystal growth process. Since then even lower threshold current densities have been reported in single quantum well 1.3 pm lasers - 8 8 A /cm 2 with InAso.4 3 Po.5 7 wells [18] and 100 A /cm 2 with A^Gaylnj^.yAs wells [19]. 45 Both the 0.85 % compressive strained lasers and the 1.2 % tensile strained lasers have superior performance compared to the lattice-matched lasers. The compressive strained device has the lowest transparency current while the tensile strained device has the highest gain coefficient. The high transparency current of the tensile devices is probably due to the presence of two quantized electron levels. Transparency current obtained here are higher by about 30 % from those achieved in 1.55 |xm lasers in this work (chapter 4) as well as in reference [20] while the modal gain coefficients are lower by about 10 % . Modal gain is pri marily affected by the density of states and is hence only marginally different at the two wavelegths in the same material system. Transparency current on the other hand is higher in 1.3 pm lasers than in 1.55 pm lasers as the smaller energy difference between the well and the barrier in the former case results in more carrier leakage. This agrees with other recent results on 1.3 pm lasers - very low threshold current density in long cavity length single quantum well lasers comparable to those obtained for 1.55 pm lasers has been achieved only with InAsP [18] and AlGalnAs [19] wells where the band offsets result in a larger well depth for electrons while threshold currents in short cavity mul tiple quantum well lasers as low as those obtained for 1.55 pm lasers have been reported [21,22]. The internal efficiency of the lattice-matched devices is 52 % . It increases to 56 % for tensile strained devices and 62 % for compressive strained devices. The internal efficiency represents the fraction of total injected carriers that are captured by the quantum well [23]. Hence it depends on the height of the 46 potential barrier for electrons in the quantum well and the position of the electron quasi-fermi level under lasing condition. The measured increase in internal efficiency for the two strained devices is due to the fermi-level being deeper inside the potential well at lasing threshold than in the lattice-matched devices. The lower efficiency for the tensile strained device compared to the compressive strained device is probably due to the smaller band gap differ ence between the well and the barrier in the former case. The internal efficien cies are lower than those reported for 1.55 pm lasers in the next chapter as the smaller band gap difference between quantum well and barrier at 1.3 pm leads to more carrier leakage out of the wells. The characteristic temperatures obtained here are about 10-12 K lower than those obtained for 1.55 pm lasers in this work (chapter 4). This is also due to the smaller well depth in the conduction band in these devices leading to more electron leakage from the well to the barrier. Application of strain does not lead to any change in the characteristic temperature as the potential well depths change very little between the three devices fabricated here. These results and other work done since then [18,19,21,22] demonstrate that despite the smaller potential well depths, 1.3 pm strained quantum well lasers with performance comparable or only slightly inferior to 1.55 pm devices can be fabricated for use in various optical communication systems. 3.6 Distributed-Feedback (DFB) Lasers The use of biaxial strain in quantum well DFB lasers has several advan tages - lower thresholds [24], higher differential gain, smaller linewidth 47 enhacement factor leading to narrower linewidths [25], higher relaxation oscillation frequency [26] and increase in threshold margin between TE and TM polarizations which in unstrained devices is smaller than in Fabry-Perot lasers [27]. DFB lasers containing the three active regions developed for the Fabry- Perot lasers were fabricated using the procedure described in the last chapter. The gratings were fabricated with helium-cadmium holography at the Uni versity of California in Santa Barbara. Broad area laser threshold current densities as low as 163 A /cm 2 for the 0.85 % compressive strained two quantum well active region and 231 A /cm 2 for the 1.2 % tensile strained two quantum well active region were measured for these DFB lasers. 3 pm wide ridge waveguide DFB lasers were also fabricated and thresh old currents of 12 mA for lattice-matched and 1.2 % tensile strained two quan tum well devices and 10 mA for 0.85 % compressive strained devices were obtained. The coupling coefficient K was measured to be 150 cm'1 from stop band measurement. Single longitudinal mode operation was obtained, with side mode suppression ratio of 15-20 dB at operating currents about 30 to 40 % above the threshold current in compressive strained devices. Typical lasing spectra from all three devices are shown in figure 3.17 and clearly demon strate single wavelength operation of these devices. The lattice-matched and compressive strained devices operated in TE polarization while the tensile strained devices operated in TM polarization. 48 0 ) 'c 13 X 3 3 , * (0 c 0 ) c Lluiitt I I I Lattice matched DFB Compressive strained DFB Tensile strained DFB 1.28 1.29 1.30 1.31 1.32 1.33 1.34 Wavelength (pm) Figure 3.17: Lasing spectra of typical DFB ridge waveguide devices showing single mode operation References 1. O. Madelung, Semiconductors - Group IV Elements and III-V Compounds, Springer-Verlag, 1991. 2. S. A dachi," Material parameters of In^GaxAsyP^y and related binaries," J . Appl. Phys., vol. 53, no. 12, pp. 8775-8792, Dec 1982. 3. S.L. Chuang, "Efficient band-structure calculations of strained quantum wells," Phys. Rev. B, vol. 43, no. 12, pp. 9649-9661, Apr. 1991. 4. T. Ishikawa and J.E. Bowers, "Band lineup and in-plane effective mass of InGaAsP or InGaAlAs on InP strained-layer quantum well," IEEE J . Quantum Electron., vol. 30, no. 2, pp. 562-570, Feb. 1994. 49 5. Z-M. Li, M. Dion, Y . Zou, J. Wang, M. Davies and S.P. McAlister, "An approximate k.p theory for optical gain in strained InGaAsP quantum- well lasers," IEEE J. Quantum Electron., vol. 30, no. 2, pp. 538-546, Feb. 1994. 6. P. W. Atkins, Molecular Quantum Mechanics, Oxford University Press, 1983. 7. S.W. Corzine, R-H Yan and L.A. Coldren, "Optical gain in III-V bulk and quantum well semiconductors," in Quantum well lasers, edited by P.S. Zory, Academic Press, 1993. 8. D. Ahn and S.L. Chuang, "The theory of strained-layer quantum-well lasers with bandgap renormalization," IEEE J. Quantum Electron., vol. 30, no. 2, pp. 350-365, Feb. 1994. 9. G.P. Agrawal and N.K. Dutta, Long-Wavelength Semiconductor Lasers, Van Nostrand Reinhold, 1986. 10. M. Asada, "Intraband relaxation effect on optical spectra," in Quantum Well Lasers, edited by P.S. Zory, Academic Press, 1993. 11. W.W. Lui, Y . Yoshikuni, T . Yamanaka and K. Yokoyama, "A Monte Carlo method for study of Auger recombination effects in semiconductors," J. Appl. Phys., vol. 73, no. 3, pp. 1226-1234, Feb. 1993. 12. Y . Zou, J.S. Osinski, P . Grodzinski, P.D. Dapkus, W. Rideout, W.F. Sharfin, J . Schlafer and F.D. Crawford, "Experimental study of Auger recombination, gain and temperature sensitivity of 1.5 pm compressively strained lasers," IEEE J. Quantum Electron., vol. 29, no. 6, pp. 1565-1575, June 1993. 13. H. Kamei and H. Hayashi, "OMVPE growth of GalnAs/InP and GalnAs/GalnAsP quantum wells," J. Cryst. Growth, vol. 107, pp. 567- 572.1991. 14. PJ.A. Thijs, E.A. Montie and T . van Dongen, "Structures for improved 1.5 pm wavelength lasers grown by LP-OMVPE; InGaAs-InP strained-layer quantum wells a good candidate," J. Cryst. Growth, vol. 107, pp. 731- 740.1991. 50 15. T.Y. Wang, T.H. Reihlen, H.R. Jen and G.B. Stringfellow, "Systematic studies on the effect of growth interruption for GalnAs/InP quantum wells grown by atmospheric pressure organometallic vapor-phase epitaxy," J. Appl. Phys., vol. 66, pp. 5376-5383,1989. 16. P . Grodzinski, Ph.D. Dissertation, University of Southern California, 1992. 17. P.W.A. Mcllroy, A. Kurobe and Y . Uematsu, "Analysis and application of theoretical gain curves to the design of multi-quantum-well lasers," IEEE J. Quantum Electron., vol. 21, pp. 1958-1963, Dec. 1985. 18. M. Yamamoto, N. Yamamoto and J. Nakano, "MOVPE growth of strained InAsP/InGaAsP quantum-well structures for low threshold 1.3 pm lasers," Proc. Intl. Conf. InP and Related Materials, pp. 231-234, 1993. 19. C.E. Zah, R. Bhat, F .J. Favire, M. Koza, T.P. Lee, D. Darby, D.C. Flanders and J.J. Hsieh, "Low threshold 1.3 pm strained-layer A^G ayln^.yAs quantum well lasers," Electron. Lett., vol. 28, no. 25, pp. 2323-2325, Dec. 1992. 20. J.S. Osinski, Ph.D. Dissertation, University of Southern California, 1992. 21. K. Uomi, T. Tsuchiya, M. Komori, A. Oka, K. Shinoda and A. Oishi, "Ultralow-threshold (0.56 mA) 1.35-pm InGaAsP/InP compressive- strained-MQW lasers," paper M 3.2, International Semiconductor Laser Conference, Maui, Sept. 1994. 22. T. Terakado, K. Tsuruoka, T . Ishida, T. Nakamura, K. Fukushima, S. Ae, A. Uda, T . Torikai and T . Uji, "Extremely low thresholds (0.4 mA @ 20°C, 3.0 mA @ 85°C) 1.3 pm strained MQW laser with novel p-substrate buried- heterostructure (RIBPBH) grown by MOCVD using TBA and TBP," paper PD 9, International Semiconductor Laser Conference, Maui, Sept. 1994. 23. P.R. Claisse and G.W.Taylor, "Internal quantum efficiency of laser diodes," Electron. Lett., vol. 28, no. 21, pp. 1991-1992, Oct. 1992. 24. C.E. Zah, R. Bhat, S.G. Menocal, F . Favire, P.S.D. Lin, A.S. Gozdz, N.C. Andreadakis, B. Pathak, M.A. Koza and T.P. Lee, "Low threshold and narrow linewidth 1.5 pm compressive strained multiquantum well distributed feedback laser," Electron. Lett., vol. 27, no. 18, pp. 1628-1630, Aug. 1991. 51 25. F. Kano, Y. Yoshikuni, M. Fukuda and J. Yoshida, "Linewidth enhancement factor of 1.3-pm InGaAsP/InP strained-layer multiple-quantum well DFB lasers,” IEEE Photon. Technol. Lett., vol. 3, no. 10, pp. 877-879, Oct. 1991. 26. M. Kito, M. Ishino, N. Otsuka, N. Hoshino, K. Fujihara, F . Fujito and Y . Matsui, "Low distortion up to 2 GHz in 1.55 pm multiquantum well distributed-feedback laser," Electron. Lett., vol. 28, no. 9, pp. 891-893, Apr. 1992. 27. G.P. Agrawal and N.K. Dutta, "Polarization characteristics of distributed feedback semiconductor lasers," Appl. Phys. Lett., vol. 46, no. 3, pp. 213- 215, Feb. 1985. 52 Chapter 4 1.55 jum Quantum Well Lasers In this chapter, the growth of materials related to 1.55 pm quantum well lasers and the experimental results obtained with these lasers will be dis cussed. The crystal growth for these devices is done in the Thomas Swan MOCVD reactor that was described in chapter 2. Instead of arsine and phos- phine, tertiarybutylarsine (TBAs) and tertiarybutylphosphine (TBP) were used as arsenic and phosphorus sources in this system. The performance of the lasers fabricated with this system is comparable to the best reported in the literature. Distributed feedback (DFB) and distributed Bragg reflector (DBR) lasers fabricated to achieve large modulation bandwidth and wide tuning range are also described in this chapter. 4.1 Survey of 1.55 pm Laser Results Because of the importance of 1.55 pm light sources in long distance opti cal fiber systems, many research groups have been working on various char acteristics of 1.55 pm quantum well lasers. To put this work into perspective, the published threshold current densities for 1.55 pm quantum well lasers will be summarized here. 53 Most of the work done on 1.55 pm quantum well lasers has utilized h ^ G a^ A s quantum wells. The ternary alloy I^ G a ^ A s is much easier to grow and characterize than InxGa1_ xASyP1_ y quaternary alloys. With x = 0.53 the quantum well is lattice-matched to InP, x > 0.53 provides compressive strain and x < 0.53 corresponds to tensile strain. However the well width required to get 1.55 pm emission changes with composition. This is illustrated in figure 4.1, where the calculated emission wavelengths of InxGa^As quantum wells with different strains are plotted as a function of well width. This calculation assumes that the barrier layer is InxGa^AsyPi.y that is lattice-matched to InP and has a band gap of 1.03 eV. This figure shows that to get 1.55 pm emission with 1.0-1.5 % compressive strain, 25-35 A well widths are required. With similar magnitude of tensile strain, it is no longer possible to get 1.55 pm emission. For compressive strained I^ G a ^ A s single quantum well lasers, Tsang et al. [1] at AT&T Bell Laboratories have obtained threshold current density of 170 A /cm 2 with 5 mm cavity length and Thijs et al. [2] at Philips Optoelec tronics Centre have obtained 147 A /cm 2 in a 4.5 mm long device. For tensile strained single quantum well lasers, Zah et al. [3] at Bellcore have obtained 197 A /cm 2 for 4 mm cavity length and Thijs et al. [2] have achieved 92 A /cm 2 with 1.1 cm cavity length, which is the lowest reported threshold current den sity to date. 54 InGaAs / InGaAsP (1.22 um) QWs 2.0 E= 1.5% E E = x : O ) c 0 ) O > (0 5 c o w m E LU 0.5% - 1 ’o % E = - 1.5 % 1.4 0 5 0 1 0 0 1 5 0 200 Well width (A) Figure 4-1: Calculated emission wavelength as a function of well width for I^ G a ^ A s quantum wells with lattice -matched InGaAsP = 1.22 |im) barriers The thin quantum wells required for 1.55 |im emission with compressive strain are undesirable because the requirement on the quality of the well-bar- rier interfaces becomes too stringent, thin wells have higher transparency cur rents and significant difference in well thickness or emission wavelength between compressive strained devices and lattice-matched or tensile strained devices makes any comparison very difficult. The use of InxGa^ASyP^y for compressive strained quantum wells allows independent choice of both well width and emission wavelength. Osinski et al. [4] have earlier used such wells in our laboratory and obtained threshold current densities of 140 A /cm 2 in 3.5 mm long devices with 90 A wide 1.8% compressive strained wells. More 55 recently, Yamamoto et al. [5] of NTT have achieved 98 A /cm 2 for 5 mm cavity length from a 60 A 0.75% compressive strained single quantum well laser. All of these results are obtained by metal organic chemical vapor deposition (MOCVD), using arsine and phosphine as group V sources, except for Tsang et al., who have used chemical beam epitaxy for crystal growth. Holmes et al. [6] at University of California, Santa Barbara have used TBAs and TBP as group V sources in MOCVD and have obtained threshold current densities of 121 A /cm 2 for 3.5 mm cavity length with a 90 A wide 1.6% compressive strained quantum well. An alternative material system for 1.55 |xm lasers is AlxGayIn;i_x_yAs. Bhat et al. [7] at Bellcore have obtained threshold current density of 166 A /cm 2 for a 0.98% compressive strained single quantum well laser. However, the presence of aluminum makes growths and particularly regrowths required for fabrication of buried heterostructure lasers and integrated devices more difficult and hence this material system is not as widely used as the InxGa^xAsyPj.y system. 4.2 Growth of Bulk Layers Surface morphology of InP epilayers grown by MOCVD using trimeth- ylindium (TMIn) and tertiarybutylphosphine (TBP) has been reported to be dependent on growth temperature [8]. Below a certain temperature, incom plete pyrolysis of TBP results in oval shaped defects on sample surface. Above a certain temperature, loss of phosphorus from the solid phase leads to 56 hazy surface. The temperature window for specular surface morphology is dependent on TBP partial pressure, being wider at higher TBP concentrations. In our MOCVD system, InP growth was evaluated by growing 4 |im thick epilayers at 76 torr pressure, carrier flow rate of 4 slpm and vapor phase P /In ratio of 40 and then characterized through an optical microscope. Oval defects were observed on epilayers grown below 625°C and hazy surfaces were observed for films grown above 650°C. Based on these results, the growth temperature was set at 640°C. With the growth pressure at 76 torr and temperature at 640°C, the back ground carrier concentration of the InP epilayer was measured using polaron C-V profiling. The background level was measured to be 4.2 X 101 5 cm ' 3 with P /In ratio of 40 and 1.8 X 101 5 cm' 3 with P/In ratio of 80, both n-type. This is higher than the background level we had previously obtained using phos- phine, which was about 1.0 X 101 4 cm "3 and is probably due to sulfur and sili con impurities in the TBP source, as reported by Watkins et al. [8 ] using magnetoluminescence spectroscopy. In0 5 3 Ga0 . 4 7 As epilayers with specular surface morphology were obtained at the same growth temperature and pressure as above with As/(Ga+In) ratio of 20 in the vapor phase. The background carrier concentra tion in this epilayer was 1.0 X 101 6 cm"3, also n-type. The waveguide and barrier layers for 1.55 |im devices was chosen to be In0 7 7 Ga0 2 3 AS0 4 9 P0 .5 1 which is lattice-matched to InP and has a bandgap of 1.03 eV. This layer was grown with gas phase (As+P)/(Ga+In) ratio of 40. 57 P and n doping of bulk layers was also characterized by polaron C-V profiling. We were able to achieve n-type doping in InP from 4.0 X 101 7 cm"3 to 4.9 X 101 8 cm"3 and p-type doping from 4.0 X 101 7 cm ' 3 to 2.0 X 101 8 cm"3. P doping of In0 5 3 Ga0 . 4 7 As was achieved upto 1 . 0 X 1 0 1 9 cm' 3 at growth temper ature of 640°C. 4.3 Growth of Quantum Wells Optimization of the formation of well-barrier interfaces in this growth system was done independently of the procedure described in last chapter because of the considerable differences between the two growth systems. In this system, each element has two inputs into the reactor, so no change of source flows through bubblers and mass flow controllers is required at a well- barrier interface. The high gas velocity due to low pressure operation results in fast change of gas composition in the reactor following the switching of inputs using the vent-run manifold. These factors permit use of much shorter growth interruptions than those utilized for the 1.3 pm devices. The general growth interruption scheme contains four purge times. Dur ing this period, group III source flow to the reactor is interrupted and group V overpressure is maintained over the hot sample surface. On each interface, there are two time periods in the total interruption time - during the first one, group V flow corresponding to that used in the previous layer growth is maintained through the reactor and during the second part, group V flow cor responding to that required for the next layer is introduced into the reactor. These purge times are called tj and t2 at the bottom interface of the quantum 58 well and t3 and t4 at the top interface of the quantum well. Quantum well structures were grown with different durations for the four periods and then characterized by photoluminescence at 5 K. The full-width-at-half-maxima (FWHM) of the emission from different quantum wells was measured and used as an indicator for the quality of the quantum well. Three different kinds of quantum wells were studied - Ino.5 3 Gao.4 7 As wells with InP barriers, Ino.5 3 Gao.4 7 As wells with In0 . 7 7 Ga0 2 3 AsQ.4 9 P0 .5 1 barri ers and Ino.7 9 Gao.2 1 Aso.7 2 Po.2 8 wells with Ino.7 7 Gao.2 3 Aso.4 9 Po.5 1 barriers. The first two have lattice-matched wells while the third one has 0.85% compres sive strain. The first structure had three quantum wells with widths of 2 0 A, 45 A and 140 A, the second structure had three wells having widths of 30 A, 50 A and 1 2 0 A and the third structure had two wells - 40 A and 1 0 0 A wide. PL linewidths from the first structure are tabulated in table 4.1. The thin nest quantum well is most sensitive to the abruptness of the interface and can hence be used as the indicator for the quality of the interface. We see that the PL linewidth is smallest when there is no growth interruption and increases for longer interruptions. Ino.5 3 Gao.4 7 A s/InP quantum well requires a com plete change in group V composition (from 100% As to 100% P) at the inter face. During growth interruptions, the group V element from the layer on the surface is desorbed, creating interface layers that are compositionally graded and hence have wider PL emission peaks. We also see splitting of the PL peak for the 20 A well for all non-zero interruption times, confirming this observa tion. 59 tl * 2 * 3 * 4 2 0 A 45 A 140 A 0 s 0 s 0 s 0 s 12.7 meV 9.9 meV 8 . 8 meV 0.25 s 0.25 s 0.25 s 0.25 s 14.7 meV 9.2 meV 9.5 meV 0.4 s 0 .1 s 0 .1 s 0.4 s 27.6 meV 7.9 meV 10.2 meV 0.5 s 0.5 s 0.5 s 0.5 s 16.7 meV 8.5 meV 9.4 meV 1 . 0 s 1 . 0 s 1 . 0 s 1 . 0 s 22.5 meV 8 . 6 meV 9.6 meV 2 . 0 s 2 . 0 s 2 . 0 s 2 . 0 s 24.5 meV 15.6 meV 9.3 meV Table 4.1: PL linewidths of quantum wells of three different widths for differ ent growth interruption times in Ino.5 3 Gao.4 7 A s/InP PL linewidths from the second quantum well structure are summarized in table 4.2. For Ino.5 3 Gao.4 7 As/Ino.7 7 Gao.2 3 Aso.4 9 Po.5 1 wells, w e find that a short but non-zero growth interruption leads to the narrowest PL linewidth. Since the change in group V composition across the interface is not as large as in the last case, a very short growth interruption does not cause any signifi cant grading and allows time for the new gas mixture to stabilize in the reac tion tube. Since quantum well thickness to be used in laser structures is about 75 A, w e chose the 0.1 sec TBAs purge over In0 5 3 Ga0 4 7 As and O.lsec TBP purge over Ino.7 7 Gao.2 3 Aso.4 9 Po.5 1 to be the optimum growth interruption for use in device growths. The effect of growth interruptions on the third quantum well structure are summarized in table 4.3. In this structure, w e see that the optimum inter ruption time is longer than that in the last structure. The 0.4 sec interruption, during first half of which the group V flow from the preceding layer is used and during second half the flow for the next layer is used, leads to the small- 60 tl * 2 * 3 t4 30 A 50 A 1 2 0 A 0 s Os 0 s 0 s 24.8 meV 15.9 meV 12.9 meV 0 s 0 s 0 s 0 .1 s 23.4 meV 13.2 meV 12.1 meV 0 .1 s 0 s 0 .1 s 0 s 23.8 meV 11.7 meV 10.2 meV 0 .1 s 0 .1 s 0 .1 s 0 .1 s 26.4 meV 19.0 meV 16.2 meV 0.25 s 0.25 s 0.25 s 0.25 s 33.8 meV 23.0 meV 1 1 . 0 meV Table 4.2: PL linewidths of quantum wells of three different widths for differ ent interruption times in In0 5 3Ga0 4 7 A s/In 0 7 7 Ga0, 2 3 Aso 4 9 P0 5 1 est linewidths and was adopted for use in laser structures. These linewidths are comparable to the best reported [9-11] and indicate that our quantum well - barrier interfaces are extremely abrupt. tl t2 t3 t4 40 A 1 0 0 A 0 s 0 s 0 s 0 s 5.9 meV 5.9 meV 0 .1 s 0 s 0 .1 s 0 s 5.7 meV 5.4 meV 0 s 0 .1 s 0 s 0 .1 s 5.7 meV 5.7 meV 0 .1 s 0 .1 s 0 .1 s 0 .1 s 5.6 meV 6 . 0 meV 0 . 2 s 0 . 2 s 0 . 2 s 0 . 2 s 5.1 meV 5.4 meV 0.3 s 0.3 s 0.3 s 0.3 s 7.3 meV 6 . 6 meV 0.5 s 0.5 s 0.5 s 0.5 s 6.1 meV 6.0 meV 0 . 2 s 0 . 8 s 0 . 2 s 0 . 8 s 8 . 8 meV 6 . 8 meV 0 . 8 s 0 . 2 s 0 . 8 s 0 . 2 s 8 . 6 meV 7.0 meV 1 . 0 s 1 . 0 s 1 . 0 s 1 . 0 s 7.9 meV 6.1 meV Table 4.3: PL linewidths of quantum wells of two different widths for different interruption times in In0 .7 9 Ga0 2 1 As0 . 7 2 P0 .2 8 / Ino.7 7 Gao.2 3 As0.49P0 .5 1 61 The strain in quantum wells was estimated by double crystal x-ray dif fraction of multiple quantum well structures, as described in chapter 2. A typ ical rocking curve of a six quantum well structure is shown in figure 4.2. From TEM measurement of well thickness and peak separations in this x-ray curve, the strain is estimated to be +1.5%. The large number of satellite peaks and their sharpness confirms that the well - barrier interfaces are abrupt and smooth. InP InGaAsP/lnP Six 35A QWs £ = +1.5% 10 m = +1 m = 0 1 0 m = -1 m = m = +4 :m = -11 10 10 -2000 -1000 0 1000 Angle (arcsec) Figure 4.2: Experimental x-ray rocking curve of an InGaAsP/lnP six quantum well structure To estimate the uniformity of grown samples, w e have measured PL peak wavelength for 1.55 pm quantum well structures at different positions over a 12 mm X 12 mm sample. The maximum deviation in the peak wave- 62 length was 26 nm for a 65 A quantum well and 15 nm for a 105 A well. This uniformity is sufficient for the devices that w e will be fabricating. 4.4 Broad Area Lasers A typical single quantum well laser structure is shown in figure 4.3. It is a single step separate confinement structure with the total waveguide thick ness kept at 4500 A in all structures. p + InGaAs 2500 A p InP 1.5 pm InP 1500 A InGaAsP (1.22 pm) 2200 A InGaAsP (1.22 pm) 2200 A n InP 8000 A n+ InP substrate Figure 4.3: Structure of a typical single quantum well laser In the two quantum well devices, 2 0 0 A barriers were used between the wells. The n InP cladding layer doping was graded from 5 X 101 8 cm"3 to 4 X 101 7 cm'3. The p InP cladding layer doping was graded from 4 X 101 7 cm ' 3 to 1 X 101 8 cm"3. The p+ contact layer was doped to 1 X 101 9 cm'3. Quantum wells with five different amounts of strain were utilized. The lattice-matched and 63 0.85% and 1.5% tensile strained wells are ternary alloys while the 0.85% and 1.5% compressive strained wells are quaternary alloys. Broad area lasers were fabricated and pulse tested as discussed in chapter 2 . The measured threshold current densities (Jth) of the two quantum well lasers with zero, + 0.85% and - 0.85% strain are plotted in figure 4.4 as a func tion of cavity length. The lines connecting the data points are obtained from a semilogarithmic fit [12] of the measured data. Threshold current densities of 351 A /cm 2 for 1.5 mm long lattice matched lasers, 191 A /cm 2 for 2.1 mm long 0.85% compressive strained laser and 246 A /cm 2 for 1.5 mm long 0.85% ten sile strained laser were measured. 1200 E o '5 5 c 0) Q • Lattice M atched 2 QW ■ + 0 .8 5 % 2 QW ▲ - 0 .8 5 % 2 QW 1000 8 0 0 c 0) v . 6 0 0 13 o T > O .C W <D S I \ - 4 0 0 200 2000 1000 1 5 0 0 5 0 0 Cavity Length (gm) Figure 4.4: Threshold current density for different cavity lengths for two quantum well 1.55 pm lasers 64 From this plot, reduction of threshold current density is observed for both compressive and tensile strain with respect to the lattice-matched lasers. The reciprocal of the external quantum efficiency of these devices is plot ted in figure 4.5 as a function of cavity length. An increase in efficiency is observed for both strained devices relative to the lattice matched lasers. From the slope and the y-intercept of this plot, the internal quantum efficiency (riint) and the optical loss were estimated and are listed in table 4.4. 2 .4 > o c a > o ic a > a j 2.2 2.0 Q ) -a O Lattice M atched 2 QW + 0 .8 5 % 2 QW - 0 .8 5 % 2 QW ra o o 1 _ Q . 'o O £E 1000 2000 5 0 0 1 5 0 0 Cavity Length (pm) Figure 4.5: Reciprocal of differential efficiency for different cavity lengths for two quantum well 1.55 |im lasers Figure 4.6 shows the measured threshold current densities for + 0.85 % , - 0.85 % , + 1.5 % and - 1.5 % strained single quantum well lasers as a function of cavity length. Single quantum well lasers usually have the lowest threshold 65 current densities and are hence used for comparison of quality of epitaxial growth and quantum well optimization. We have obtained threshold current densities as low as 93 A /cm 2 for a 5 mm long 1.5% compressive strained laser and 101 A /cm 2 for a 4 mm long 1.5% tensile strained laser. These results are among the best published data for 1.55 pm lasers and show that either com pressive or tensile strain can be used to get very low threshold current densi ties. The reciprocal of the differential efficiency of these lasers are plotted as a function of cavity length in figure 4.7. The internal quantum efficiency is esti mated to be as high as 91.4% for the 1.5% compressive strained lasers. 8 0 0 ■ + 0 .8 5 % SQW □ -0 .8 5 % SQW 6 0 0 to c 0 Q 4 0 0 D o ■ o o ■ C ( / ) 0 L. s z 200 3 0 0 0 4 0 0 0 5 0 0 0 1000 2000 Cavity Length (gm) Figure 4.6: Threshold current density for different cavity lengths for single quantum well 1.55 pm lasers 66 4.0 3 .5 3 .0 4 - » C 0 } L _ 2 .5 **- o 2.0 - + 0.85 % SQW • • • - 0.85 % SQW - + 1.5 % SQW - - 1 .5 % SQW m o o u . Q . O 0 ) ec 1000 2000 3 0 0 0 4 0 0 0 5 0 0 0 Cavity Length (pm) Figure 4.7: Reciprocal of differential efficiency for different cavity lengths for single quantum well 1.55 |im lasers From the threshold current density - cavity length data, the modal gain at threshold can be obtained for devices with different cavity lengths and hence different threshold currents. The gain is plotted as a function of current density in figure 4.8 and the values of transparency current Qtr) obtained from here are listed in table 4.4. The lasing polarization is TE for lattice matched and compressive strained lasers and TM for tensile strained lasers. 67 35 - - + 0 .8 5 % Strain - 0 .8 5 % Strain --•• + 1 .5 % Strain 1.5 % Strain 3 0 2 5 CD 20 15 3 .5 4.0 4 .5 5.0 6 .5 5.5 6.0 7 .0 Log of th resh o ld cu rren t density in A/cm 2 Figure 4.8: Threshold modal gain for single quantum well lasers Strain Well thick ness (A) Num ber of wells hint (% ) Optical loss (cm'1) Jtr per well (A /cm 2) Gain coef ficient per well (cm'1) 0 70 2 67 3.4 56 7.5 + 0.85% 115 1 78 4.2 34 7.6 + 0.85% 115 2 80 4.6 36 8.4 + 1.5% 80 1 91 5.7 38 10.4 - 0.85% 100 1 74 3.0 42 10.5 - 0.85% 100 2 78 4.2 45 8.8 -1.5% 130 1 83 5.2 44 14.9 Table 4.4: Device parameters for various 1.55 pm quantum well lasers 68 The temperature dependence of the threshold current for two quantum well lasers is plotted in figure 4.9. The characteristic temperature (T0) obtained from an exponential temperature dependence of the threshold cur rent is found to be 54-55 K for all three devices. This is comparable to other reports in this material system [13-15]. 5 0 0 □ - 0.85% 2 QW s O + 0.85% 2 QW s A Unstrained 2 QWs < E 4 0 0 c d ) w. D o o 3 0 0 .c w 2 200 2 7 0 2 8 0 2 9 0 3 0 0 3 1 0 3 2 0 T em p eratu re (K) Figure 4.9: Temperature dependence of threshold current for two quantum well lasers The temperature dependence of threshold for single quantum well lasers is shown in figure 4.10. T0 is 53 K for 1.5 % tensile strain, 44 K for 0.85 % ten sile strain, 38 K for 1.5 % compressive strain and 28 K for 0.85 % compressive strain. The lower T0 values in single quantum well lasers are because of increased carrier overflow over the shallow electron barriers in the quantum 69 well. The lowest T0 values are obtained when the threshold condition requires quantum well operation at high carrier densities, leading to more electron leakage. 4 0 0 O + 1.5 % SQW □ - 0.85 % SQW < E 0 + 0.85 % SQW + - * c C D V— I— 3 0 0 3 o T3 O n < n CD sz H 200 100 2 7 0 2 8 0 2 9 0 3 0 0 3 1 0 3 2 0 3 3 0 T em perature (K) Figure 4.10: Temperature dependence of threshold current for strained single quantum well lasers 4.5 Discussion of Laser Results The effects of biaxial strain on laser performance can be seen from the data in table 4.4. Enhancement of internal quantum efficiency, reduction of transparency current density and increase in modal gain coefficient are observed with both compressive and tensile strain. This leads to reduction in threshold current and improvement of the differential efficiency for strained quantum well lasers. 70 Internal quantum efficiency of these devices is a measure of carrier con finement in the quantum wells [15]. Hence it depends on the barrier heights and the position of the quasi-fermi levels under lasing condition. The reduc tion of both radiative and non-radiative current in strained quantum well lasers leads to quasi-fermi levels that are deeper into the quantum wells and hence leads to reduction in carrier leakage and higher internal efficiency. Con sequently the internal quantum efficiency increases from 67 % in lattice- matched lasers to 76 % in 0.85 % tensile strained lasers, 79 % in 0.85 % com pressive strained lasers, 83 % in 1.5 % tensile strained lasers and 91 % in 1.5 % compressive strained lasers. The smaller internal efficiency of the tensile strained devices compared to the compressive strained devices having the same magnitude of strain is probably due to the smaller band gap difference between the well and the barrier for tensile quantum wells. In reference [16], higher internal efficiencies are obtained with lattice- matched quantum wells than with compressive strained wells. That data shows a large difference in internal efficiency between single, two and four quantum well devices with the same amount of strain which is not observed in this work. This is probably due to improved carrier capture as the number of wells in the active region is increased. The lack of such a strong depen dence in our results indicates more efficient carrier capture in the quantum wells because of the different waveguide design and that has led to high internal efficiencies even in single quantum well devices. Figure 4.11 shows the dependence of transparency current density on the strain in the quantum well.Transparency current density is reduced from 71 56 A /cm 2 per well in lattice-matched lasers to about 44 A /cm 2 per well in tensile lasers and 36 A /cm 2 per well in compressive lasers. This is almost independent of the amount of strain and is due to the reduction in the in plane mass of the deepest bound hole state around the r point. Increase in strain from 0.85 % to 1.5 % does not lead to any further significant decrease in the hole sub-band effective mass and hence there is no appreciable change in the transparency current. The transparency currents obtained in this work for lattice-matched and compressive strained devices are similar to those reported in reference [16]. _ 60 0 E o < 50 w c < D o 40 c < D 30 3 O 20 > » o c < D C O Q. ( / > c < 0 1 — o T ransparency current □ Modal gain i- 1.5 1.0 0.5 0.0 0.5 1.0 1.5 0.5 g 0.4 S' — t < D D 0.3 ^ o Q. B L 0.2 a Q ) 3' c T 01 > 0.0 Strain (%) Figure 4.11: Transparency current density and differential modal gain as a function of strain for single quantum well lasers 72 From figure 4.11, w e also observe that the differential modal gain is higher for 1.5 % tensile strained devices than for 1.5 % compressive strained devices. There is a significant increase in differential modal gain as the strain is increased from 0 % to 0.85 % and then from 0.85 % to 1.5 % . The enhance ment is more with tensile strain than with compressive strain. Hence for low loss devices, compressive strained devices will have lower threshold currents while for high loss devices, tensile strained lasers will have lower threshold currents. This is in agreement with both the observed and calculated results obtained in chapter 3 for 1.3 |im devices. Non-radiative recombination processes, such as Auger recombination, are a significant part of the total threshold current density in these devices and are known to be reduced significantly by the application of biaxial stain [13]. To estimate the effect of biaxial stain on non-radiative recombination, we have combined the experimental gain plots obtained in the last section with theoretical calculations based on the model described in section 3.1. The procedure described in section 3.1 was used to calculate the gain- radiative current characteristics of single quantum well lasers with the five quantum wells that were fabricated in this work. The results are shown in fig ure 4.12. The calculated radiative current density at different peak modal gains was then subtracted from the experimentally measured current density at the same values of peak modal gain for each of the five strained quantum wells. Non-radiative current density obtained by this procedure is plotted in figure 4.13 as a function of the peak modal gain. That permits direct compari son of devices with the same cavity length, assuming that the internal loss is 73 about the same in lasers with different amounts of strain. We observe a signif icant reduction in the non-radiative current density from lattice-matched devices by the application of biaxial strain. Unlike the transparency current, there is a large decrease in non-radiative recombination as the strain is increased from 0.85 % to 1.5 % . The lowest non-radiative currents are obtained for the 1.5 % tensile strained quantum well lasers. Calculation of Auger recombination rates by Monte Carlo simulations have been reported for lattice-matched and compressive strained quantum wells [18] but not for tensile strained wells. 3 0 2 5 -© - 0 % - 0 - + 0.85 % -A - + 1.5 % - 0.85 % 20 O J 0 2 0 4 0 6 0 8 0 2 Radiative current density (A/cm ) Figure 4.12: Calculated peak modal gain as a function of radiative current density for 1.55 pm single quantum well lasers with different amounts of strain in the active region 74 1000 C V J E o < 0 % + 0.85 % 6 0 0 3 O 4 0 0 0 ) > 200 I C o z 1 o o 5 1 5 20 2 5 Peak gain (cm 1) Figure 4.13: Non-radiative current density as a function of peak gain for 1.55 pm single quantum well lasers with different amounts of strain E < > > * < / ) c 0 " D 4 — « c 0 D O 0 > 0 ' ~ o 0 k_ C o 4 1 0 0 % + 0.85 % 0.85 % 3 1 0 2 1 0 1 1 0 2 3 4 5 6 7 8 9 2 3 4 5 6 7 8 9 ^ 12-2 Carrier density above transparency (X 10 cm ) 1 0 Figure 4.14: Non-radiative current density as a function of carrier density for 1.55 pm single quantum well lasers with different amounts of strain 75 Figure 4.14 shows the non-radiative current density for the five strained quantum well lasers as a function of the carrier density above the transpar- recombination rates but agrees with the Auger rates reported in [13] and [18]. 4.6 Distributed Feedback (DFB) and Distributed Bragg Reflector (DBR) High speed DFB lasers and tunable DFB and DBR lasers emitting at 1.50 pm were fabricated using the compressive strained multiple quantum well active regions reported in references [4] and [9]. The crystal growths were done on the growth system used for the fabrication of 1.3 pm lasers which were reported in the last chapter. 4.6.1 High Speed DFB Lasers Design of a laser for high speed operation involves different issues than the design for low threshold current operation, so these factors will be briefly discussed first. The resonance frequency fr in a semiconductor laser is given where vg is the group velocity in the cavity, g' is the differential gain, S is the photon density and T p is the photon lifetime. High speed operation requires that the device cavity length be optimized as both photon lifetime and differ ential gain decrease with shorter lengths. The optimum cavity length ency condition. The data does not fit the conventional power law [19] used for Lasers by [20]: (Eq. 4-1) 76 becomes shorter with larger number of quantum wells in the active region. The maximum number of wells that can be used is limited by critical thick ness consideration due to strain and by hole trapping in quantum wells lead ing to non-uniform carrier distribution in the wells. The differential gain can be enhanced by the use of strained quantum wells. Carrier transport across the waveguide region also affects the speed of the laser - a thinner waveguide is preferable but that results in higher threshold current due to loss of optical confinement, so an optimization between these two factors is required [21]. The active region chosen for this device was four 1.2 % compressive strained quantum wells with a 2600 A wide waveguide layer, which is about 1000 A thinner than that used for low threshold operation. A high speed device needs to have low capacitance, so that the modula tion bandwidth is not adversely affected by the RC product. The structure used most commonly for this purpose is a buried heterostructure laser with semi-insulating InP current blocking layers [22]. This structure requires multi ple crystal growth steps, so we chose a ridge waveguide laser buried with polyimide because of its easier fabrication process. The schematic of this device is shown in figure 4.15. To reduce the device resistance, skin diffusion of zinc was done into the p + InGaAs layer. Resistance of 1.5 ohms could then be achieved for 2 pm wide ridge waveguide devices. 77 Po lyim ide Metal contact pad — - ^ n+ p - InP Waveguide n -ln P p+ InGaAs Silicon Nitride Figure 4.15: Schematic of a polyimide buried ridge waveguide laser for high speed operation DFB lasers were fabricated with gratings having quarter wavelength phase shifts. Threshold current of 10-15 mA and power output of 10-15 mW per facet was achieved with uncoated devices. The voltage-current (V-I) and power-current (L-I) characteristics of a typical DFB device are shown in figure 4.16. This device has a threshold current of 19 mA and linear L-I curve till 100 mA bias. The lasing spectra of the same device is shown in figure 4.17. Single m ode operation can be seen at all three bias levels. The modulation response of one of these devices was tested by Prof. John Bowers at the University of California, Santa Barbara. 3 dB modulation bandwidth of 9 GHz was measured at 30 mA bias current. Fabry Perot lasers with the same structure had 3 dB bandwidth of 7.5 GHz at 60 mA bias. No furthur increase in modulation bandwidth was obtained at higher bias cur rents. 78 12 3.0 Power Voltage 10 2.5 8 2.0 6 1.5 Q . 4 1.0 2 0.5 0 0.0 0 20 40 80 60 100 0 > (Q (D o -a Current (mA) Figure 4.16: I-V and L-I characteristics of a typical DFB laser under continuous excitation 1.0 0.8 to c 3 • 0.6 J Q L - < a I = 75 mA 0.4 (O c < D I = 50 mA c 0.2 I = 30 mA V. "vl. 0.0 1.490 1.491 1.492 1.493 1.495 1.494 1.496 1.497 Wavelength (pm) Figure 4.17: Lasing spectra of DFB laser at three different current levels 79 4.6.2 Tunable DFB Lasers In a semiconductor laser, change in carrier density leads to change in both the refractive index and the gain coefficient, the two being related by Kramers-Kronig relations. In a DFB laser, this leads to a shift in the Bragg wavelength and hence in the lasing wavelength. A two segment DFB laser has a uniform grating through the device and two independent contacts controlling current injection into two segments. The device is shown schematically in figure 4.18. The Bragg wavelengths of the two segments can then be independently varied by changing the current in the two sections. The lasing wavelength tunes due to two basic mecha nisms - the phase change in each segment and the reflection amplitude of each segment changes with the current in each segment [23]. Tuning range as large as 7.2 nm has been demonstrated with two segment DFBs, however most of this tuning is thermal and not electronic, there are large differences in tuning range of individual devices and suppression of spatial hole burning is very important to get such large tuning range [24]. Figure 4.18: Schematic of two segment DFB laser We have fabricated two segment DFB lasers from the same material as used for high speed DFB lasers and have obtained tuning of upto 25 A. The 80 tuning of a typical device is shown in figure 4.19. Here the current through one segment is kept constant and the current through the other segment is varied while the device is kept on a constant temperature mount. The tuning range is limited by the change in refractive index that is possible when the carrier density is pinned at and above threshold. Larger tuning range can be obtained in DBR lasers as the carrier density in the mirror sections that pro vide wavelength selective feedback is not pinned and hence a larger change in refractive index is possible. w c Z3 s i ■ 3 >. w c © c lb = 95 mA | lf = 20 mA T = 20°C - A !b = 75 mA I lb = 50 mA . ..... j _____ lb = 25 mA J[ lb = 13 mA J L _ . _ I I I I 1.490 1.491 1.492 1.493 1.494 1.495 Wavelength (|jm) Figure 4.19: Spectra showing tuning of a two segment DFB laser 4.6.3 Broad Tuning Range Using Sampled Gratings Tuning range larger than that possible with index change in a DBR laser can be obtained by synthesizing a pair of slightly mismatched mirrors with 81 periodic reflection spectra. The mirrors are realized by sampled gratings, which are conventional gratings with grating elements removed periodically. The periodic modulation of the grating leads to comb-like reflection spec trum. Lasing occurs at the wavelength where reflection maxima from the two mirrors are aligned. A small change in index induced in one mirror segment causes another pair of maxima to align resulting in a large change in the las ing wavelength. This idea was first proposed by Prof. Larry Coldren of the University of California at Santa Barbara (UCSB) [25] and is described theo retically in reference [26]. A DBR laser incorporating the sampled grating mirrors was fabricated in collaboration with Prof. Coldren. The 1.2 % compressive strained four quantum well active region was grown at USC in the first step. Then the active region was patterned into 320 pm stripes by etching down to a thin InP etch-stop layer placed just underneath the quantum wells. The sampled grat ing was then formed at UCSB using holographic and masked lithography fol lowed by wet chemical etching. The mirrors consisted of 3.5 pm bursts of grating at 50 pm period in one mirror and 45 pm in the other. The p cladding and contact layers were then grown at USC by atmospheric pressure MOCVD. The schematic of the structure is shown in figure 4.20. Ridge waveguide lasers were then processed and tested at UCSB. By changing the current in the two mirror sections, tuning range of 57 nm (with mode hops) was observed with side-mode suppression ratio larger than 30 dB. The tuning range can be made continuous by adding a phase adjusting section to the 82 device and using the current in that section to fine tune the wavelength between the discrete mode hops allowed by the sampled grating mirrors. p+ InGaAs Sampled g ra ting m irro r Sampled g ra tin g mirror p-lnP InGaAsP (1.25 pm) n-lnP QW Active Ftegion Figure 4.20: Schematic of sampled grating DBR laser References 1. W.T. Tsang, F.S. Choa, M.C. Wu, Y.K. Chen, A.M. Sergent and P.F. Sciortino, Jr., "Very low threshold single quantum well graded-index separate confinement heterostructure InGaAs/InGaAsP lasers grown by chemical beam epitaxy", Appl. Phys. Lett., vol. 58, no. 23, pp. 2610-2612, June 1991. 2. P.J.A. Thijs, J.J.M. Binsma, L.F. Tiemeijer and T . van Dongen, "Improved performance 1.5 pm wavelength tensile and compressively strained InGaAs-InGaAsP quantum well lasers", Proc. ECOC, vol. 2, pp. 31-38, 1991. 83 3. C.E. Zah, R. Bhat, B. Pathak, C. Caneau, F.J. Favire, N.C. Andreadakis, D.M. Hwang, M.A. Koza, C.Y. Chen and T.P. Lee, "Low threshold 1.5 pm tensile-strained single quantum well lasers", Elect. Lett., vol. 27, no. 16, pp. 1414-1416, Aug 1991. 4. J.S. Osinski, Y . Zou, P. Grodzinski, A. Mathur and P.D. Dapkus, "Low- threshold-current-density 1.5 pm lasers using compressively strained InGaAsP quantum wells", IEEE Photon. Tech. Lett., vol. 4, no. 1, pp. 10- 13, Jan. 1992. 5. N. Yamamoto, K. Yokoyama, T. Yamanaka and M. Yamamoto, "Very low threshold graded-index separate-confinement-heterostructure strained InGaAsP single-quantum-well lasers", Elect. Lett., vol. 30, no. 3, pp. 243- 244, Feb. 1994. 6. A.L. Holmes,Jr., M.E. Heimbuch and S.P. DenBaars, "Strained GalnAsP single-quantum-well lasers grown with tertiarybutylarsine and tertiarybutylphosphine", Appl. Phys. Lett., vol. 63, no. 25, pp. 3417-3419, Dec. 1993. 7. R. Bhat, C.E. Zah, M.A. Koza, D-M.D. Hwang, F.J. Favire and B. Pathak, "OMCVD growth of strained AlxGayIn2_ x.yAs for low threshold 1.3 pm and 1.55 pm quantum well lasers", Proc. Intl. Conf. InP and Related Materials, pp. 453-456, Apr. 1992. 8. S.I. Watkins, M.K. Nissen, G. Haacke and E.M. Handler, "Residual donor and acceptor incorporation in InP grown using trimethylindium and tertiarybutylphosphine", J. Appl. Phys., vol. 72, no. 7, pp. 2797-2801, Oct. 1992. 9. P. Grodzinski, Ph.D. Thesis, University of Southern California, pp. 113-118, 1992. 10. H. Kamei and H. Hayashi, "OMVPE growth of GalnAs/InP and GalnAs/GalnAsP quantum wells", J. Cryst. Growth, vol. 107, pp. 567- 572,1991. 11. P.J.A. Thijs, T. van Dongen, P.I. Kuindersma, J.J.M. Binsma, L.F. Tierneyer, J. M. Lagemaat, D. Moroni and W. Nijman, "High quality InGaAsP-InP for multiple quantum well laser diodes grown by low-pressure OMVPE", J. Cryst. Growth, vol. 93, pp. 863-869,1988. 84 12. P.W.A. Mcllroy, A. Kurobe and Y . Uematsu, "Analysis and application of theoretical gain curves to the design of multi-quantum-well lasers", IEEE J. Quantum Electron., vol. 21, pp. 1958-1963, Dec. 1985. 13. Y. Zou, J.S. Osinski, P. Grodzinski, P.D. Dapkus, W. Rideout, W.F. Sharfin, J. Schlafer and F.D. Crawford, "Experimental study of Auger recombination, gain and temperature sensitivity of 1.5 pm compressively strained lasers", IEEE J. Quantum Electron., vol. 29, no. 6, pp. 1565-1575, June 1993. 14. H. Temkin, T . Tanbun-Ek, R.A. Logan, D.A. Cebula and A.M. Sergent, "High temperature operation of lattice matched and strained InGaAs- InP quantum well lasers", IEEE Photon. Tech. Lett., vol. 3, no. 2, pp. 100- 102, Feb. 1991. 15. C.P. Seltzer, S.D. Perrin and P.C. Spurdens, "Low threshold current, high output power buried heterostructure MQW lasers with strained InGaAsP wells", Electron. Lett., vol. 28, no. 19, pp. 1819-1820, Sep. 1992. 16. J.S. Osinski, Ph.D. Thesis, University of Southern California, pp. 146,1992. 17. P.R. Claisse and G.W.Taylor, "Internal quantum efficiency of laser diodes," Electron. Lett., vol. 28, no. 21, pp. 1991-1992, Oct. 1992. 18. W.W. Lui, T . Yamanaka, Y . Yoshikuni, K. Yokoyama and S. Seki, "Suppression of Auger recombination effects in compressively strained quantum-well lasers," IEEE J. Quantum Electron., vol. 29, no. 6, pp. 1544- 1552, June 1993. 19. G.P. Agrawal and N.K. Dutta, Long-Wavelength Semiconductor Lasers, Van Nostrand Reinhold, 1986. 20. K. Y . Lau and A. Yariv, "Ultra-high speed semiconductor lasers," IEEE J. Quantum Electron., vol. 21, pp. 121-138,1985. 21. R. Nagarajan, M. Ishikawa, T . Fukushima, R.S. Geels and J.E. Bowers, "High speed quantum-well lasers and carrier transport effects," IEEE J. Quantum Electron., vol. 28, no. 10, pp. 1990-2008, Oct. 1992. 85 22. G. Eisentein, U. Koren, A.H. Gnauck, R.S. Tucker, T.L. Koch, P.J. Corvini and B.I. Miller, "Modulation and spectral properties of semi-insulating blocked planar buried-heterostructure distributed feedback lasers," Appl. Phys. Lett., vol. 53, no. 20, pp. 1905-1907, Nov. 1988. 23. M. Kuznetsov, "Theory of wavelength tuning in two-segment distributed feedback lasers," IEEE J. Quantum Electron., vol. 24, no. 9, pp. 1837-1844, Sep. 1988. 24. D.V. Eddols, S.J. Vass, R.M. Ash and C.A. Park, "Two-segment multiquantum well lasers with 7 nm tuning range and narrow linewidth," Electron. Lett., vol. 28, no. 11, pp. 1057-1058, May 1992. 25. L.A. Coldren, U.S. Patent 4 896 325. 26. V. Jayaraman, Z.M. Chuang and L.A. Coldren, "Theory, design and performance of extended tuning range semiconductor lasers with sampled gratings," IEEE J. Quantum Electron., vol. 29, no. 6, pp. 1824- 1834, June 1993. 86 Chapter 5 Polarization Insensitive Devices Semiconductor optical amplifiers can easily be integrated with other semiconductor-based components for multi-channel optical communication systems. The use of multiple quantum well active region in these amplifiers leads to high saturation output power, wide gain bandwidth and fast recov ery time. Their main disadvantage is strong polarization sensitivity due to the orientation of the dipole moments associated with band-to-band transitions. In this chapter, various schemes that can be utilized to decrease the polariza tion sensitivity of semiconductor optical amplifiers will be discussed. Conventional quantum well lasers also strongly favor one polarization - transverse electric (TE) for lattice-matched and compressive strained wells and transverse magnetic (TM) for tensile strained wells. The incorporation of both compressive and tensile strained wells into the same active region allows fabrication of lasers that can operate in both TE and TM polarizations at wavelengths that are very close to each other. Experimental results obtained with this concept and the theoretical analysis that has been done to understand the operation of these novel devices will also be presented in this chapter. 87 5.1 Polarization Insensitive Amplifier Schemes Several schemes have been used to demonstrate polarization insensitive semiconductor optical amplifiers. These different schemes will be discussed here briefly. In a bulk III-V semiconductor active region, the heavy hole and light hole bands are degenerate at the T point leading to gain in both TE and TM polarizations. To remove the polarization selectivity of the waveguide struc ture, the waveguide should have a nearly square cross-section. Fabrication of active regions with sub-micron widths is technologically very difficult. It has been utilized by Kitamura et al. [1] to fabricate polarization insensitive ampli fiers at 1.3 pm. Besides the complicated fabrication procedure, these devices also lack the advantages of quantum well active regions. In a quantum well, tensile strain tends to move the light hole level deeper into the well while quantum size effect does the opposite. Thus a strained quantum well with a small amount of tensile strain (0.1-0.25 % ) can be designed that has degenerate heavy hole and light hole levels. This scheme has been utilized by Joma et al. [2] to fabricate 1.50 pm polarization insensi tive amplifiers and by Bour et al. [3] to fabricate 642 nm dual polarization lasers. However devices with such small amounts of tensile strain have very high threshold currents [4] and hence are not very useful. Researchers at NTT [5,6] have fabricated 1.50 pm polarization insensitive amplifiers by using lattice-matched wells and tensile strained barriers. A schematic of this structure is shown in figure 5.1. The light hole band edge in the barrier is close to the valence band edge in the well, so light holes can not 88 be quantized in the well. This leads to bulk-like TM gain while TE gain comes from the quantized electron to heavy hole transitions in the well. The differ ence in the nature of the two transitions (bulk versus quantum well) makes it difficult to obtain polarization insensitive gain over a wide bandwidth or a large range of injection current. TE e TM ( + 7 E ) r:------- - ▼ Ih band edge hh hh band edge L a ttic e -m a tc h e d T e n sile QW Barrier Figure 5-1: Polarization insensitive structure using lattice-matched quantum well and tensile strained barrier 5.2 Active Regions Containing Compressive and Tensile Strained Quantum Wells In this work, a novel polarization insensitive amplifier scheme was designed that we believe offers several advantages over the schemes described above. This structure contains both compressive and tensile 89 strained quantum wells in the same active region. A schematic of the struc ture is shown in figure 5.2. TE TM TE hh Ih Compressive QW Tensile QW Figure 5.2: Schematic of polarization insensitive active region using compressive and tensile strained quantum wells Electron to heavy hole transitions in the compressive strained quantum well provide most of the TE gain in this structure. Electron to light hole transi tions in the tensile strained well provide all of the TM gain and a small part of the TE gain. The material compositions and the widths of the two wells are chosen such that the transition energy in the two wells is the same. The amount of strain in the two wells and the relative number of wells of each type can be used to get equal TE and TM gain from this structure. Thus this structure has the flexibility to control emission wavelengths and gain in the 90 two polarizations independently. An example of this is shown in figure 5.3 where the calculated gain spectra of a structure similar to that shown in figure 5.2 is shown for three different injection currents. TE and TM gain are equal to about 10 % from 1.46 pm to 1.51 pm. 1 20 1 00 8 0 16 8 0 A/ cm' 6 0 4 0 - ----- TE .......TM 20 1 .4 6 1 .4 7 1 .4 8 1 .4 9 1 .5 0 1.51 1 .5 2 W avelength (pm) Figure 5.3: Calculated gain spectra of active region containing compressive and tensile strained quantum wells TE and TM gain in this structure are coupled by two mechanisms - the light holes in the tensile strained well interact with both TE and TM polariza tions and the finite electron barriers allow electrons to escape out of one type of well and get captured by the other kind of well [7]. The strong interaction between TE and TM polarizations due to these two mechanisms helps to 91 maintain polarization insensitivity even when the gain is saturated by input intensity [8]. 5.3 1.3 |nm Dual Polarization Laser A 1.3 pm laser structure was fabricated utilizing the concept described above. The structure had six quantum wells - three 55 A wells with 0.85 % compressive strain and three 125 A wells with 1.2 % tensile strain, alternately placed with 150 A barriers in between them. Broad area lasers as well as ridge waveguide devices of different cavity lengths were found to lase in both TE and TM polarizations. A typical L-I curve for a 1625 pm long broad area device is shown in figure 5.4. Both TE and TM modes have about the same threshold currents while the TM mode has slightly higher power output. The spectra of a similar device at three dif ferent injection levels is shown in figure 5.5. The lasing wavelength for the two polarizations are very close to each other. The spectra of a device with TE and TM lasing wavelengths separated by about 90 A is shown in figure 5.6 at four different injection levels. The TM mode starts lasing at smaller injection current as the energy of that transition is smaller than that of the TE transition in this device. 92 60 L= 1625 |jm TE TM 50 40 30 C D Q . i _ C D O C L 20 10 0 400 200 600 800 C u rre n t (mA) Figure 5.4: Measured L-I curves of a 1625 pm long device emitting in both TE and TM polarizations L = 1625 M m TM TE I = 450 mA I = 380 mA - I = 350 mA 1.32 1.34 1.36 1.38 1.40 Wavelength (pm) Figure 5.5: Measured TE and TM spectra of a 1625 |im long device at three different injection levels 93 TE TM Six QW Laser L = 753 pm = 400 mA (0 = 350 mA w c o = 300 mA = 275 mA 1.31 1.32 1.33 1.34 1.35 1.36 1.37 1.38 Wavelength (pm) Figure 5.6: Measured TE and TM spectra of a 753 pm long device at four different injection levels Devices with cavity lengths shorter than 500 pm lase in TM polarization only. This dependence of the lasing polarization on cavity length can be explained by the higher gain coefficient of the tensile well and the lower transparency current of the compressive well. As cavity length decreases, the threshold gain for lasing increases and hence the TM polarization which has a higher gain coefficient starts to dominate. 5.4 1.3 pm Polarization Insensitive Amplifier A 980 pm long ridge waveguide device having the same structure as described in section 5.3 was coated with silicon monoxide anti-reflection coat- 94 ings on both facets to fabricate a travelling wave amplifier. The amplifier gain at 1.319 pm was characterized under pulsed bias in a pump-probe setup by Dr. Serge Dubovitsky and Prof. William Steier [8,10]. Gain in TE and TM polarizations was measured with the amplifier pumped by TE and TM light separately, whose intensity was varied from zero to well above gain satura tion. The measurement results at 90 mA bias current are shown in figure 5.7. There are four measured quantities in this plot - TE gain with TE pump ( G t e , t e ) / TM gain with TE pump (GT m x e ) / TE gain with TM pump (GT E TM ) and TM gain with TM pump (G t m T m). The results show that there is less than 1 dB difference in the four gains, even when the amplifier is strongly sat urated. 12 10 8 G . 6 T M ,T M r T E ,T E 4 2 0.1 1 10 O utput pow er (mW) Figure 5.7: Gain characteristics of polarization insensitive amplifier 95 TE gain is saturated equally by TE or TM intensity, implying that cross saturation is as strong as self-saturation. This indicates a strong coupling between the two polarizations in the amplifier. The coupling is partially due to inter-action of light holes in the tensile well and partially from coupling of electrons in the compressive and tensile wells because of carrier leakage from one well and then recapture by the other well. This strong interaction between the two polarizations is essential to fabricate polarization insensitive optical amplifiers that maintain polarization insensitivity even in the gain sat uration regime. 5.5 1.55 pm Dual Polarization Laser A laser structure emitting at 1.55 pm was fabricated with six quantum wells - three 120 A wide 1 % tensile wells and three 75 A wide 1.5 % compres sive wells, alternately placed with 200 A lattice-matched barriers in between the wells. The pulsed L-I characteristics of a typical ridge waveguide device are shown in figure 5.8 at three different operating temperatures. At 5 °C, the las ing polarization is TE while at 40 °C it is TM. At 25 °C, both TE and TM polar izations have similar threshold currents and about equal output power till 600 mA current. At high injection, the TE power saturates while the TM power keeps on increasing linearly. This is in contrast to the 1.3 pm lasers reported in the last section where no such saturation behavior was observed. Similar behavior is seen when the length of the current pulses is varied. For long 96 pulse lengths (about 1 psec), the TE output starts decreasing near the end of the pulse. Lasing spectra of a similar device operating under pulsed injection at three different operating temperatures are shown in figure 5.9. Only TE out put is seen at 15 °C and only TM output is seen at 33 °C. Both TE and TM polarizations with almost equal intensity of the strongest Fabry Perot mode is observed at 28 °C. The strong temperature sensitivity of the lasing polarization makes it difficult to operate these devices under continuous bias and obtain both TE and TM polarizations in the output. A ridge waveguide device lasing into both TE and TM polarizations at 14 °C is shown in figure 5.10. A 4 °C change in operating temperature makes the device lase into a single polarization. Dual polarization in these devices is more critically dependent on condi tions such as device cavity length, operating temperature and bias pulse duration than the 1.3 pm devices. The deeper quantum wells in the 1.55 pm devices would result in less carrier leakage and hence less cross-coupling between the two polarizations. With reduced coupling between the two polarizations, start of lasing in one polarization could inhibit lasing in the other polarization. 97 < 1 ) O 0 3 0 ) Q. < D £ 35 30 25 20 TE TM 0 200 400 600 800 a > o c o C D a. C D 5 £ 35 25 °C 30 25 20 5 0 5 0 0 200 400 600 800 Injection Current (mA) Injection Current (mA) 0 o C O * * - U. a > a C D 5 o a. 40 C 0 2 0 0 4 0 0 6 0 0 8 0 0 Injection C urrent (mA) Figure 5.8: Pulsed L-I characteristics at three different temperatures of 1 |im lasers with compressive and tensile quantum well active regions Pulsed Operation TE 15 “ C TM 28 "C > TM 33 ° C 1 .5 4 5 1 .5 5 0 1 .5 5 5 1 .5 6 0 1 .5 6 5 1 .5 7 0 1 .5 7 5 W avelength (pm) Figure 5.9: Lasing spectra of a 1.55 pm ridge waveguide device under pulsed excitation 18 ” C c 3 14 °C 2 jN ^ iW < n c C D CW Operation TE TM 10 "C 1 .5 5 0 1 .5 6 0 1 .5 7 0 W avelength (pm) 1 .5 8 0 Figure 5.10: Lasing spectra of a 1.55 pm ridge waveguide device under continuous bias 99 5.6 1.55 pm Dual Polarization Laser Analysis To understand the temperature sensitivity of the 1.55 pm dual polariza tion lasers, the TE and TM gain in these structures was calculated at three dif ferent temperatures using the procedure described in section 3.1. Carriers were assumed to be in quasi-equilibrium in the active region, so that the quasi-fermi levels are continuous across the active region. This assumption means that these calculations are valid only at low photon densities, i.e. below and close to lasing threshold only. The result of this calculation is shown in figure 5.11. 60 TE TM E o 4 0 c f O o > " t o " O o E to 0 CL o ° c 20 4-5 °C 100 120 1 40 160 180 200 2 R adiative current density (A/cm ) Figure 5.11: Calculated TE and TM peak modal gain as a function of the radiative current density at three different temperatures 100 The gain curves in this figure show that the TE mode has lower transpar ency current while the TM mode has higher differential gain, leading to a cross-over of the TE and TM gain curves. In a laser, the threshold gain is pinned by the losses, so the device operates in TE mode if losses are below the cross-over point and in TM mode if they are above the cross-over. This cross over point moves to lower magnitudes of gain at higher temperatures. Hence a 1 mm cavity length device lases into both TE and TM polarization around 25 °C but only in TE at lower temperatures and only TM at higher tempera tures. To study the above-lasing behavior in these devices, w e have adapted the rate equations for a semiconductor laser [9] to a system having three lev els for carriers [10] and photons of two different polarizations. Figure 5.12 shows the carriers in this system - Nj is the total number of carriers in the bar rier region, N c is the number of carriers in the compressive well and N t is the number of carriers in the tensile well. The various processes in this system are denoted by their rates - yj is the capture rate into the compressive well, Y 2 is the capture rate for the tensile well, y3 is the escape rate out of the compres sive well, Y 4 is the escape rate out of the tensile well, yc is the total recombina tion rate in the compressive well and yt is the recombination rate in the tensile well. In this analysis, only the electrons are considered as they are more mobile and have smaller potential barriers. 101 InP 74 N-i 73 — n l r f - r n l I f ▼ B a rrie r ▼ 1 71 I s 72 7c Compressive QW 7t Tensile QW Figure 5.12: Schematic of various carrier related processes in an active region containing tensile and compressive active regions. The coupled rate equations for this system can be written as: = - “ (7i + 72)N i + Y 3N c + Y 4N t dt q dN c dt — YjNi (Y3"*"Yc)^c ^te,cPte d N t _ dt — Y2^i (Y4+ Yt)Nt G ^ P ^ Gte tPte dP, dt te _ — (G ^e ttte)^*te Pte^sp,te dP = (Gtm — a tm )Ptm + Ptm^sp,tm (Eq. 5-1) (Eq. 5-2) (Eq. 5-3) (Eq. 5-4) (Eq. 5-5) where I is the current flowing through the active region, q is the electron charge, Pte is the number of TE photons in the cavity, Ptm is the number of TM 102 photons in the cavity, Gtec and Gte t are the TE modal gain in the compressive well and in the tensile well respectively, Gtm is the TM modal gain, a te is the total TE loss in the device (waveguide and facet), a tm is the total TM loss, RS p,te and Rsp/tm are the TE and TM spontaneous emission rates and Pte and (3 tm are the fractions of TE and TM total spontaneous emission that get cou pled into the lasing mode. These equations are solved under steady-state conditions to obtain the TE and TM photon densities as a function of injection current density. Gain and spontaneous emission rates are calculated separately for the tensile and compressive wells used in the 1.55 |im structure described in the last section as a function of carrier density in the well, using the procedure described in section 3.1, and fitted to polynomial form for use in these rate equations. The spontaneous emission coupling factor is taken to be 10'4 for both TE and TM polarizations. The carrier lifetime in both wells is taken to be 1 nsec. In this particular structure, the compressive well has smaller differential gain and lower transparency carrier density than the tensile well. Carrier capture time in 1.55 pm lattice-matched quantum wells has been measured to be between 2-7 psec [11] and 1 psec [12] by different groups. Cal culations done for GaAs/AlGaAs wells show strong dependence on well thickness, barrier thickness and the potential barrier height [13]. Carrier escape time has been calculated to be 4-80 psec [14] and is strongly dependent on the potential barrier height and the carrier density in the well. Hence the following calculations have been done by varying the carrier escape and cap ture rates for the two wells. 103 The results shown in figure 5.13 and 5.14 assume that the escape time constant for both wells is 10 psec, the capture time constant of the compres sive well is 4 psec and the capture time constant of the tensile well is varied from 1.5 psec to 4 psec. If the capture time constant of the tensile well is 4 psec (figure 5.14), the lasing mode is TE and when it is 1.5 psec (figure 5.13), the lasing mode is TM. For the two intermediate time constants lasing is obtained in both polarizations, with TE power being larger for 1.9 psec (figure 5.13) and TM power being larger for 1.5 psec (figure 5.14). 10 10 10 9 10 8 10 7 10 6 10 TM 1/ y2=1.5ps TM 1/ y 2=1.9ps 5 10 - - TE 4 10 0 500 1000 1500 2000 2500 3000 2 Current Density (A/cm ) Figure 5.13: Calculated TE and TM L-I curves for compressive and tensile strained quantum well lasers with different carrier capture rates for the tensile well (1.5 ps and 1.9 ps) 104 E o ( / > c C D ~o c o • * — > o - C Q . 10 10 10 9 10 8 10 7 10 6 10 TE TM 1/ y 2=4ps TE TM 1/ Y2=2ps 5 10 4 10 500 1000 0 1500 2000 2500 3000 Current density (A/cm ) Figure 5.14: Calculated TE and TM L-I curves for compressive and tensile strained quantum well lasers with different carrier capture rates for the tensile well (2 ps and 4 ps) In the next set of calculations, the capture time constant of the compres sive well is taken to be 4 psec and that of the tensile well as 2 psec. The escape time constant from the two wells is taken to be the same and is varied from 5 psec to 50 psec. For short escape times, the laser has strong polarization selec tivity, based on the relative capture times of the two wells. As the escape times become larger, the polarization selectivity decreases. This results in TE and TM polarizations having threshold currents that are closer and power levels that are comparable to each other. The TE and TM L-I curves when the escape time constants are 50 psec is shown in figure 5.15. 105 E o < / ) c Q ) Q c o • « — * o - C CL 10 TE TM 10 10 9 10 8 10 7 1/ Yi=4ps 1/ Y2=2ps 1/'Y3=1/ Y4=50ps 10 6 10 5 10 0 1000 1500 500 2000 2500 3000 Current Density (A/cm ) Figure 5.15: Calculated L-I curves of compressive and tensile strained quantum well laser for carrier escape time constant of 50 psec from the quantum well In figures 5.16 and 5.17, the capture time constants are again taken to be 4 psec and 2 psec for the compressive and the tensile well respectively. The escape time constant for the tensile well is kept fixed at 10 psec while that for the compressive well is 9.5 psec in figure 5.16 and 10 psec in figure 5.17. For 9.5 psec, the TM power is greater than the TE power and for 10 psec the TE power is larger than the TM power. For time constants smaller than 9.5 psec, lasing is in TM polarization only while for time constants larger than 10 psec, lasing is only in TE polarization. Thus w e again observe a very strong depen dence of the lasing polarization on the escape time constant, similar to that seen for the capture time constant in figures 5.13 and 5.14. 106 E o ( 0 c 0 * o c o * - * o - C 0. 10 10 TE TM 10 9 10 8 10 7 10 6 10 5 10 4 10 0 1000 500 1500 2000 2500 3000 Current density (A/cm ) Figure 5.16: Calculated L-I curves for escape time constant of 9.5 ps for the compressive well E o (/) c 0 T 5 C o O - C CL 10 10 10 TM 9 10 8 10 7 10 6 10 5 10 4 10 0 1000 500 1500 2000 2500 3000 Current density (A/cm ) Figure 5.17: Calculated L-I curves for escape time constant of 10 ps from the compressive well 107 These calculations show that polarization selectivity of an active region containing both tensile and compressive strained quantum wells is extremely sensitive to the carrier capture and escape rates for the two wells. If the cap ture rates and the escape rates are of the same order of magnitude, then the condition for dual polarization operation for a laser is very sharp and that requires careful balance of the capture and escape time constants in the two wells, depending on the gain of the two wells. However, if the escape rates are at least an order of magnitude lower than the capture rates, then dual polarization lasing can be expected under a wider range of conditions. In structures with more than one pair of wells, carrier transport effects have to be taken into account and may modify the effective capture or escape rate of the quantum wells. 5.7 Conclusions A novel concept of utilizing compressive and tensile strained quantum wells in the same active region to achieve polarization insensitive gain has been introduced. The choice of strain, well width and the relative number of the two types of wells offers enough flexibility to design polarization insensi tivity over a wide bandwidth at both 1.3 pm and 1.55 pm. Coupling of TE and TM gain through electron-light hole transitions and carrier leakage and re capture helps to maintain polarization insensitivity into gain saturation regime. Active regions containing both compressive and tensile strained quan tum wells were utilized to fabricate dual polarization lasers at 1.3 pm as well 108 as 1.55 Jim lasers. The 1.55 |im devices were theoretically analyzed by calcu lating TE and TM gain under quasi-equilibrium condition and TE and TM light outputs under steady-state condition. References 1. S. Kitamura, K. Komatsu and M. Kitamura, "Polarization-insensitive semiconductor optical amplifier array grown by selective MOVPE,” IEEE Photon. Tech. Lett., vol. 6, no. 2, pp. 173-175, Feb. 1994. 2. M. Joma, H. Horikawa, C.Q. Xu, K. Yamada, Y . Katoh and T. Kamijoh, "Polarization insensitive semiconductor laser amplifiers with tensile strained InGaAsP/InGaAsP multip.'e quantum well structure," Appl. Phys. Lett., vol. 62, no. 2, pp. 121-122, Jan. 1993. 3. D.P. Bour, K.J. Beemink, D.W. Treat, T.L. Paoli and R.L. Thornton, "Dual polarization, single quantum-well AlGalnP laser diode structure," IEEE J. Quantum Electron., vol. 30, no. 12, pp. 2738-2742, Dec. 1994. 4. P.J.A. Thijs, J.J.M. Binsma, L.F. Tiemeijer and T . van Dongen, "Improved performance 1.5 pm wavelength tensile and compressively strained InGaAs-InGaAsP quantum well lasers," Proc. ECOC, vol. 2, pp. 31-38, 1991. 5. K. Magari, M. Okamoto, H. Yasaka, K. Sato, Y. Noguchi and O. Mikami, "Polarization insensitive traveling wave type amplifier using strained multiple quantum well structure," IEEE Photon. Tech. Lett., vol. 2, no. 8, pp. 556-558, Aug. 1990. 6. M. Okamoto, K. Sato, H. Mawatari, F . Kano, K. Magari, Y . Kondo and Y . Itaya, ’TM mode gain enhancement in GalnAs-InP lasers with tensile strained-layer superlattice," IEEE J. Quantum Electron., vol. 27, no. 6, pp. 1463-1469, June 1991. 7. S. Dubovitsky, W.H. Steier, A. Mathur and P.D. Dapkus, "Gain saturation properties of a semiconductor gain medium with tensile and compressive strain quantum wells," IEEE J. Quantum Electron., vol. 30, no. 2, pp. 380-391, Feb. 1994. 109 8. S. Dubovitsky, A. Mathur, W.H. Steier and P.D. Dapkus, "Gain saturation properties of a polarization insensitive semiconductor amplifier implemented with tensile and compressive strain quantum wells," IEEE Photon. Tech. Lett., vol. 6, no. 2, pp. 176-178, Feb. 1994. 9. G.P. Agrawal and N.K. Dutta, Long-wavelength semiconductor lasers, Van Nostrand Reinhold, 1986. 10. S. Dubovitsky, Ph. D. Dissertatio'n, University of Southern California, 1994. 11. S. Weiss, J.M. Wiesenfeld, D.S. Chemla, G. Raybon, G. Sucha, M. Wegener, G. Eisenstein, C.A. Burrus, A.G. Dentai, U. Koren, B.I. Miller, H. Temkin, R.A. Logan and T. Tanbun-Ek, "Carrier capture times in 1.5 pm multiple quantum well optical amplifiers," Appl. Phys. Lett., vol. 60, no. 1, pp. 9- 11, Jan. 1992. 12. B. Deveaud, J. Shah, T.C. Damen and W.T. Tsang, "Capture of electrons and holes in quantum wells," Appl. Phys. Lett., vol. 52, no. 22, pp. 1886- 1888, May 1988. 13. P.W.M. Blom, J.E.M. Haverkort and J.H. Wolter, "Optimization of barrier thickness for efficient carrier capture in graded-index and separate- confinement multiple quantum well lasers," Appl. Phys. Lett., vol. 58, no. 24, pp. 2767-2769, June 1991. 14. S.C. Kan, D. Vassilovski, T.C. Wu and K.Y. Lau, "Quantum capture and escape in quantum-well lasers - implications on direct modulation bandwidth limitations," IEEE Photon. Tech. Lett., vol. 4, no. 5, pp. 428- 431, May 1992. 110 Chapter 6 Conclusion 6.1 Summary In this work, strained quantum wells grown by metal organic chemical vapor deposition (MOCVD) have been utilized to fabricate very low thresh old current density 1.3 pm and 1.55 pm lasers. At both wavelengths, threshold current densities comparable to the best reported in the literature were obtained by careful optimization of the crystal growth process. Tertiarybuty- larsine and tertiarybutylphosphine were used as sources for group V ele ments in crystal growth for part of this work and performance comparable to that obtained with the conventional sources (arsine and phosphine) were obtained. Development of less hazardous sources is important for more extensive commercial use of MOCVD as a crystal growth technique. 1.3 pm and 1.5 pm distributed feedback (DFB) lasers operating in single longitudinal mode were fabricated with first order gratings and two step MOCVD growth. The 1.5 pm DFBs had 3 dB bandwidth of 9 GHz when pro cessed as polyimide buried ridge waveguide devices. Tuning range of 2.5 nm in two segment DFB lasers and 57 nm in distributed bragg reflector (DBR) lasers with sampled grating mirrors were obtained. I ll To remove the polarization sensitivity of conventional quantum well active regions, a novel active region containing compressive and tensile strained quantum wells was proposed. Dual polarization lasers emitting at 1.3 pm and 1.55 pm wavelengths were demonstrated. A polarization insensi tive amplifier at 1.3 pm was fabricated and was found to maintain polariza tion insensitivity in both linear gain and gain saturation regimes. A simple theoretical simulation procedure was developed to analyze the performance of strained quantum well lasers and was used to predict the effect of strain and well width on laser performance. The procedure was extended to calculate gain, spontaneous emission and radiative current in polarization insensitive structures. The above lasing properties of dual polar ization lasers were studied by formulating and solving the steady state rate equations for these novel structures. 6.2 Future Work The quantum well materials developed in this work can readily be applied to several other new and interesting devices. Crystal growth in multi ple steps allows the fabrication of a variety of integrated structures for use in wavelength division multiplexed systems. A few devices that can be easily fabricated based on this work are described here. The low threshold strained quantum well laser material can be used to fabricate buried-heterostructure lasers with low leakage currents. Two differ ent current-blocking structures are possible - one with n-p-n reverse biased 112 junctions in InP [1] and the other with a semi-insulating InP layer [2]. Two such structures are illustrated in figure 6.1. p-lnP ---- n-lnP p-lnP n-lnP p+ InGaAs p-lnP n-lnP a. BH laser with n-p-n-p blocking layers p+ InGaAs p-lnP r-—i Active region ■ ■ InGaP layer b. BH laser w ith si blocking layer Figure 6.1: Proposed buried heterostructure designs with p-n-p-n blocking diodes and semi-insulating current blocking layers These structures would require three MOCVD growth steps for device fabri cation. In the first structure, the n-cladding layer, the active region and part of the p cladding layer are grown in the first step, in the second step the current- blocking layers are grown through a dielectric mask and in the third step the rest of the p-cladding layer and the p+ contact layer are grown. The second structure would require high resistivity InP layer that can be obtained with a semi-insulating dopant such as iron. Ferrocene is available as the iron source on our MOCVD system currently. For this device, the entire laser structure except for the p+ contact layer are grown in the first step. A thin wide band SI InP n-lnP 113 gap layer and a semi-insulating InP layer are then grown selectively through a dielectric mask. The mask is then removed and the p+ contact layer is grown in the third step. This structure would have low threshold currents due to very low leakage current as well as high modulation bandwidth due to low parasitic capacitance. The active region containing tensile and compressive strained quantum wells can be used to fabricate a three segment device whose lasing polariza tion can be electronically controlled. The schematic of the structure is shown in figure 6.2. The long segment contains both compressive and tensile strained quantum wells and is designed to provide both TE and TM gain. The two shorter segments contain tensile and compressive strained wells sepa rately. The quantum wells are placed on the top of the waveguide rather than in the center, so that the etching required for regrowths does not affect the waveguide, thus minimizing scattering losses in the waveguide. When the short section with tensile wells is pumped the device lases in TM polarization and when the section with compressive wells is pumped, the device lases in TE polarization. Another interesting device is an integrated wavelength convertor. Wave length conversion has already been demonstrated in a saturated polarization insensitive amplifier using separate laser sources and bulk optics to couple light in and out of the amplifier [3]. The principle of the process is shown in figure 6.3. One laser source saturates the gain of the amplifier while the signal is impressed on a second source at a different wavelength. The signal modu lates the gain of the amplifier and hence gets transferred to the wavelength of 114 the first source. At the output, the two signals can be separated based on their polarization. The whole process can be integrated on a single InP chip by using an array of distributed feedback lasers (whose lasing wavelengths are different because of different grating periods) as light sources present on the same chip as the polarization insensitive amplifier in which wavelength con version process takes place. O O I O i r 1 1 p-lnP v r i A i < / = _ n-lnP p+ contact Waveguide layers vi A Compressive QWs Compressive & Tensile QWs Tensile QWs Figure 6.2: Three segment laser structure with electronic control of lasing polarization 115 i k Gain Output signal -► Input Power Input signal *1 Figure 6.3: Principle of wavelength conversion in a saturated optical amplifier References 1. T.A. Ohtoshi, K. Yamaguchi and N. Chinone, "Analysis of current leakage in InGaAsP/InP buried heterostructure lasers," IEEE J. Quantum Electron., vol. 25, no. 6, pp. 1369-1375, June 1989. 2. S. Asada, S. Sugou, K. Kasahara and S. Kumashiro, "Analysis of leakage current in buried heterostructure lasers with semiinsulating blocking layers," IEEE J. Quantum Electron., vol. 25, no. 6, pp. 1362-1368, June 1989. 116 3. S. Dubovitsky, P.D. Dapkus, A. Mathur and W.H. Steier, "Wavelength conversion in a quantum well polarization insensitive amplifier," IEEE Photon. Technol. Lett., vol. 6, no. 7, pp. 804-807, July 1994. 117

Linked assets