Citation
Thin p-clad InGaAs single quantum well lasers

Material Information

Title:
Thin p-clad InGaAs single quantum well lasers
Creator:
Wu, Chih-Hung, 1960-
Publication Date:
Language:
English
Physical Description:
viii, 176 leaves : ill. ; 29 cm.

Subjects

Subjects / Keywords:
Diodes ( jstor )
Laser power ( jstor )
Lasers ( jstor )
Lasing ( jstor )
Light refraction ( jstor )
Quantum wells ( jstor )
Sensors ( jstor )
Stripes ( jstor )
Threshold currents ( jstor )
Wavelengths ( jstor )
Dissertations, Academic -- Electrical and Computer Engineering -- UF
Electrical and Computer Engineering thesis, Ph. D
Quantum wells ( lcsh )
Semiconductor lasers ( lcsh )
Genre:
bibliography ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1996.
Bibliography:
Includes bibliographical references (leaves 169-175).
General Note:
Typescript.
General Note:
Vita.
Statement of Responsibility:
by Chih-Hung Wu.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
The University of Florida George A. Smathers Libraries respect the intellectual property rights of others and do not claim any copyright interest in this item. This item may be protected by copyright but is made available here under a claim of fair use (17 U.S.C. §107) for non-profit research and educational purposes. Users of this work have responsibility for determining copyright status prior to reusing, publishing or reproducing this item for purposes other than what is allowed by fair use or other copyright exemptions. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder. The Smathers Libraries would like to learn more about this item and invite individuals or organizations to contact the RDS coordinator (ufdissertations@uflib.ufl.edu) with any additional information they can provide.
Resource Identifier:
023282624 ( ALEPH )
35146238 ( OCLC )

Downloads

This item has the following downloads:


Full Text










THIN P-CLAD InGaAs SINGLE QUANTUM WELL LASERS


By

CHIH-HUNG WU













A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA


1996















TO MY FAMILY














ACKNOWLEDGMENTS


I wish to express my sincere gratitude to Dr. Peter S. Zory, my advisor, for accepting and supporting me for graduate study. His guidance and constant source of encouragement have proved to be essential in the successful completion of this work. I attribute the knowledge I have acquired in the field of quantum well lasers to his academic leadership and expertise. His commitment to hard work and discipline has inspired me through the course of this study.

I would like to thank the members of my supervisory committee, Dr. Gijs Bosman, Dr. Arnost Neugroschel, Dr. Ramakant Srivastava and Dr. Robert M. Park, for giving me helpful suggestions and critiques. I am grateful to Dr. Gary Evans of Southern Methodist University for providing the Modeig program, of great use in laser structure modal loss computations. I want to express my greatest appreciation to Dr. Mark A. Emanuel of Lawrence Livermore National Lab. for providing the MOCVD grown InGaAs/GaAs SQW laser material used in this study. I would also like to thank Dr. David P. Bour of Xerox PARC for supplying GaInP/AlGaInP material during the experiments. My appreciation is especially extended to the Institute of Nuclear Energy Research (INER) in Taiwan for supporting and encouraging me to complete my Ph.D. study.








I have to recognize other individuals who have provided invaluable assistance. I wish to thank Mr. James Chamblee and Mr. Tim Vanght of the Microelectronics Laboratory for all the technical assistance they provided over the years. In addition, I thank all my colleagues in the Photonics Research Laboratory: Horng-Jye Luo, Young-Soh Park, Michael Grove, Chi-Lin Young, Craig Largent, Chia-Fu Hsu and Jeong-Seok 0 for their friendship and inspiration.

Last, I want to thank my parents for their love and for teaching me how important education is. I am deeply indebted to my mother-in-law for supporting and helping taking care of my son during my studying abroad. I would like to express greatest appreciation to my wife Yu-Fen Shen and my sons PeiShen Wu and Conan S. Wu for their love and understanding. I would also like to thank my sisters and my wife's sisters for their encouragement and support.














TABLE OF CONTENTS


page

ACKNOWLEDGMENTS ........................................ iii

ABSTRACT ............................................... vii

CHAPTERS

1 INTRODUCTION .................................... 1

2 CONTACT REFLECTIVITY EFFECTS ON THIN P-CLAD
InGaAs SINGLE QUANTUM WELL LASERS ............... 8

2.1 Introduction ................................ 8
2.2 Laser Material and Theoretical
Calculations ................................ 9
2.3 Experimental Results and Discussions ....... 19

3 THIN P-CLAD RIDGE GUIDED LASERS ................. 37
3.1 Introduction ............................... 37
3.2 Theoretical Calculations ................... 38
3.3 Device Fabrication ......................... 40
3.4 Pulse Current Characteristics ............... 41
3.5 CW Current Characteristics .................. 43
3.6 Ridge Height Dependent Diode Laser
Performance ................................ 45

4 DUAL WAVELENGTH DIODE LASERS AND THERMAL
RESISTANCE IMPROVEMENT ......................... 59

4.1 Introduction ............................... 59
4.2 Fabrication of Dual Wavelength
Diode Lasers ............................... 61
4.3 Dual Wavelength Laser Experimental
Results and Discussion ..................... 62
4.4 Thermal Resistance Improvements ............. 65

5 P -GaAs CAP LAYER THICKNESS EFFECTS ............ 91

5.1 Introduction ............................... 91
5.2 Laser Structure and Theoretical
Calculations ............................... 92
5.3 Long Time Lasing Delay in Gain-Guided
Lasers ..................................... 94
5.4 Long Time Lasing Delay and Q-Switching







Lasing Delay in Narrow Stripe, 300 nm RidgeHeight Lasers .............................. 105

6 SURFACE SENSITIVE LASER DIODES .................. 134

6.1 Introduction ............................... 134
6.2 Device Structure and Theoretical
Calculations ............................... 135
6.3 Device Fabrications ........................ 142
6.4 Device Characterizations .................... 143

7 SUMMARY and RECOMMENDATION ...................... 160

7.1 Summary .................................... 160
7.2 Recommendation for Future Study ............ 163



REFERENCES ............................................. 169

BIOGRAPHICAL SKETCH .................................... 176














Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy


THIN P-CLAD InGaAs SINGLE QUANTUM WELL LASERS


By

CHIH-HUNG WU

May 1996


Chairman: Peter Zory

Major Department: Electrical and Computer Engineering

The surface sensitive aspects of InGaAs single quantum well, semiconductor diode lasers with thin p-type cladding layers (thin p-clad lasers) are studied. Experiments using different thickness quantum wells (QW), contact layers (pt-cap layers), contact layer metallurgies and laser geometries are described and modeled.

Lasers fabricated with a QW thickness of 10 nm, an AlGaAs p-clad thickness of 250 nm, a GaAs p -cap layer thickness of 100 nm and a nickel contact show significantly higher threshold currents and lower slope efficiencies than lasers fabricated with gold contacts. For a range of cavity lengths, the lasing wavelength of the nickel contact lasers is about 50 nm shorter than the gold contact lasers. This phenomenon is exploited to demonstrate side by side lasers on


vii







the same chip with operating wavelengths of 960 and 910 nm. In a 5 gm-stripe, low-ridge configuration, the gold contact lasers operate continuously with single spatial mode output power comparable to those of conventional, high-ridge, thick p-clad lasers.

When the QW thickness is reduced to 8 nm and the GaAs p cap layer thickness is increased from 100 nm to 200 nm, new types of effects are observed. For example, when lasers with 50 micron stripe widths and gold contacts are operated, they show microsecond-long lasing delays. This long delay is attributed to the time it takes for the active region to heat to the point where net mode gain exceeds mirror loss. Net mode gain increase is due to an increase in mode overlap with the QW material gain as well as a decrease in mode overlap with the lossy gold contact layer. When the p -cap layer is decreased below 170 nm, laser performance is significantly improved and no lasing delay obtained.

By combining 100 nm and 200 nm p -cap layer structures into one laser and removing the gold layer from the 200 nm section, laser output power at fixed current becomes dependent on the type of material placed on the 200 nm section. Experiments using these "surface sensitive" diode lasers are described and their possible use in sensor applications discussed.


viii













CHAPTER 1

INTRODUCTION








Semiconductor material growth technology continues to improve and very high quality epitaxial, multi-layered structures are now available. The improved materials are being used in the fabrication of various types of semiconductor devices including semiconductor diode lasers. The features of small physical dimension, high efficiency, high speed and low cost have made semiconductor diode lasers important and widely-used devices in optical fiber communication and optical memory systems.

In most diode laser applications, it is usually desirable that the devices have low threshold current density (J h) and high differential quantum efficiency (id) . To fulfill these two requirements, diode laser structures must be designed to have both optical field and carrier confinement. Normally, the active region of diode lasers is grown and confined in a pair of cladding layers which have higher energy bandgaps and smaller refractive index than those of the active region. Additionally, the thickness of the








cladding layer is usually greater than 1.0 pm to reduce the optical loss from the contact layer and substrate.

For continuous wave (cw) or rapidly pulsed operation, the removal of heat generated in the active region is crucial to the reliability of diode lasers. To obtain highly reliable cw operation and reduce the temperature sensitivity of threshold current, the thermal impedance of the diode laser structure must be as small as possible. Theoretical work showed that the thermal resistance of diode lasers can be reduced by decreasing the thickness of the cladding layers [Joyc75]. Experimental study of double heterostructure (DH) lasers with -0.12 gm active layer thickness showed that Jth increases and Td decreases dramatically when the p-type AlGaAs cladding layer thickness is less than 0.8 pm [Case75]. Similar results were also theoretically obtained in [Butl75], where the active region gain at threshold (Gth) increased significantly as the p-AlGaAs cladding layer thickness was reduced to less than -1.0 gm due to the increase of optical loss.

Despite the increase of Gth caused by the optical loss, several reports have been published using this "thin cladding layer" diode laser structure. Since the cladding layer thickness is thin, the separation distance between the active region and p-contact layer or substrate is reduced. As a result, the interaction between the tail of lasing modes and either contact layer or corrugated-substrate could be very







strong to provide grating coupled output [Zory75] or distributed feedback (DFB) action [Scif75]. In recent years, several publications have been issued relating the utilization of thin p-cladding layer diode laser structures with the fabrication of surface emission [Maco87, Mott89] and edge emission [Shan88] DFB diode lasers. Additionally, in order to reduce the optical loss of diode lasers with thin pcladding layers, shiny Au-contacts were found to be essential [Luo90].

In previous studies of thin p-clad diode lasers, the device structures were mostly concentrated on DH diode lasers. The effects of decreasing p-cladding layer thickness on the performance of quantum well (QW) lasers are relatively unknown. In order to study these effects, single quantum well (SQW) InGaAs lasers with thin p-cladding layers (250 nm) were used in this study. The diode laser performance is found to be greatly related to the reflectivity of the contact metal. As mentioned before, there could be strong interaction between the contact metal and the tail of the lasing mode when the p-cladding layer thickness is thin enough. This interaction not only changes the modal loss [Wu94b] and lasing threshold conditions but also shifts the lasing wavelength of the QW lasers [Wu94a]. In addition to the study of contact reflectivity effects, the device fabrication technique of this thin p-clad diode structure is also explored.







Diode lasers with single spatial mode operation are useful for increasing the coupling efficiency in optical fiber communication applications. Conventionally, "ridgeguided" diode lasers with thick p-cladding layer (>1.0 gm) are fabricated to achieve this single spatial mode operation output. By using such a thick p-cladding layer diode laser structure, a deep etching step has to be performed to remove certain amounts of the outside stripe material to obtain a sufficient refractive index step between the stripe and outside stripe region for maintaining the ridge-guided property. This removal process can be obtained by either wet chemical etching or using more sophisticated RIE (reactive ion etching) techniques. Because the ridge height is critical for diode lasers to operate in the index-guided regime, a very thin "etch-stop" layer is usually grown inside the diode laser structure to make the device fabrication process somewhat less complicated. On the contrary, by using the thin p-clad laser structure, one does not need to do any etching and the fabrication process of index-guided diode lasers is greatly simplified. Single spatial mode lasing to high power levels comparable to that of the conventional thick p-clad diode lasers was obtained [Wu95].

Multi-wavelength emission diode lasers are very attractive for the application of wavelength division multiplexed communication systems. Several approaches have been reported for fabricating monolithic multi-wavelength emission diode lasers, such as changing active layer







composition [Saka82, Boua82], grating period variations [Dutt86, Aiki76], stripe width related modal loss control [Toku86] and material desorption technique [Eple90]. The associated techniques for most of these approaches are either low yield and non-reliable or complicated and time consuming. Previously, we have mentioned that contact reflectivity has a great effect on the thin p-clad SQW diode lasers [Wu94b]. One of the important characteristics is that, by selecting shiny or less shiny contact metal, one can control the lasing wavelength of diode laser on either the first or the second energy level of a single quantum well active layer. Based on this concept, the process of fabricating dual wavelength emitting diode lasers should be much simpler and more reliable than the other approaches as stated above.

Diode lasers with epi-side up bonding have less diffraction noise and better lasing beam quality than diode lasers with epi-side down bonding. In the epi-side up bonding configuration, however, thermal resistance reduction becomes an important problem. Since the heat sink is far away (-100 Jim) from the active region, heat generated inside the active region flowing through the cap layer becomes an effective heat dissipation path for diode laser operation. Theoretical study [Joyc75] has shown that the thermal resistance of diode lasers can be reduced effectively by depositing a thick Au layer as the heat spreader on the top laser. Based on this concept, a thick Au plating technique was developed and







applied to the devices fabricated to show the reduction of thermal resistance and performance improvements.

In designing thin p-clad InGaAs lasers, the p -GaAs cap layer thickness has to be determined with care in order to avoid the possible increase of modal loss from the cap layer. This is because the refractive index value of the cap layer is very close to that of the quantum well region at the lasing wavelength. In this case, due to the short separation between the quantum well and the p -GaAs cap layer, a thick cap layer allows the lasing mode to penetrate through the thin p-cladding layer and couple into the p-cap layer forming a twin-guide laser structure [Suem75] . As a result, more modal loss will be generated and poor diode laser performance obtained. Additionally, due to the smaller optical confinement of lasing mode caused by the thick (200 nm) p -cap layer, microsecond-long lasing delays are observed on 50-im stripe width lasers [Wu96a].

Although thick p -cap layer, thin p-clad lasers appear not to be useful, the insertion of a thick p -cap section into a thin pt-cap layer laser may have practical applications. The combination of one or more 100 nm p -cap layer sections as the electron pumped sections and a 200 nm p -cap layer section as the surface sensitive section in one laser structure makes a "surface sensitive" diode laser [Wu96b] which may prove useful in sensor applications.

This dissertation is organized as follows: Chapter 2 shows the details of contact reflectivity effects on the thin








p-clad diode lasers. Both the theoretical calculations of QW modal gain, modal loss and experimental results are described. Chapter 3 presents the theoretical calculations and the experimental results of ridge-guided thin p-clad diode lasers. The comparison between thin p-clad and thick pclad diode lasers are also included. Chapter 4 states the fabrication process and the experimental results of dual wavelength diode lasers made from a thin p-clad laser structure. In addition, the details of a thick Au plating heat spreader technique developed to show diode laser thermal resistance improvements are described. Chapter 5 outlines the influence of pt-GaAs cap layer thickness on the thin p-clad diode laser performance. Chapter 6 describes the details of theoretical and experimental work on surface sensitive laser diodes (SSLD) based on the thin p-clad laser structure with both 100 nm and 200 nm pt-cap layers. Finally, the conclusion and recommendations for future study are presented in Chapter 7.














CHAPTER 2

CONTACT REFLECTIVITY EFFECTS ON THIN P-CLAD InGaAs SINGLE QUANTUM WELL LASERS






2.1 Introduction



In this chapter we show how the performance of shiny contact, InGaAs single quantum well (SQW) lasers is changed when the contact metal is changed from shiny gold to less shiny nickel. In addition to the expected difference in threshold current and slope efficiency, operating wavelength differences of more than 50 nm are observed [Wu94] for cavity lengths between 250 and 700 microns. At shorter (gold) and longer (nickel) cavity lengths, large shifts in operating wavelength are observed. In Section 2.2, the laser material used is described. Additionally, theoretical calculations of the quantum well laser modal gain and the modal loss induced by the different contact metal are outlined. In Section 2.3, the experimental results of thin p-clad diode lasers with shiny gold (Au) and less shiny nickel (Ni) fabricated are shown and compared with the theoretical calculations. Also presented are the changes of diode laser performance under various annealing time and constant annealing temperature.







Diode lasers with and without facet coating are life-time tested to evaluate the reliability of these thin p-clad lasers.



2.2 Laser Material and Theoretical Calculations



Figure 2.1 shows the diode laser structure used in this study which consists of a 250 nm n -GaAs buffer layer grown on

a n+-GaAs substrate, a 70 nm AlGaixAs graded layer(x=0.050.6), a 1400 nm n-type Al0.6Ga0.4As cladding layer, a 200 nm graded AlYGa1_YAs (undoped) barrier layer(y=0.3-0.6), an active layer of 10 nm InGa1_zAS (z-0.15) undoped strained quantum well centered in a pair of 7 nm GaAs bounding layers, a 200

nm graded AlYGa1jAs (undoped) barrier layer(y=0.6-0.3), a 250 nm p-Al 0.6Ga0.4As cladding layer, a 25 nm AlGal-,As graded layer(x=0.6-0.05) and a 100 nm p+-GaAs contact layer. As can be seen the p-cladding layer is only 250 nm and is much thinner than those of conventional laser devices (>1.0 pm).

Since the total thickness of the epitaxial layers above the quantum well in this structure is only -600 nm, the type of metallization used on the p+-contact layer is expected to be important in determining the mode loss coefficient ci [Luo90]. To understand the effect of p-contact metal on the diode laser loss coefficient (M), the a1 values for the laser structure shown in Figure 2.1 have to be calculated.







To do the computations of x, the lasing wavelength is determined first. In computing the lasing wavelength of the laser structure, the energy gap shift due to the presence of biaxial strain induced by the lattice mismatched-material InGaAs/GaAs has to be calculated. The strained effects can be devided into two components: hydrostatic component and shear component [Van89]. Under the hydrostatic strain effect both the conduction band edge and the valence band edge will move up by AEc and AE, eV as referenced from the corresponding bulk material band edge. The energy gap change AE; due to hydrostatic strain effect is obtained from the following equations (Van891:



AEg = AEc - AEv (2.2.1)


AEc = ac * (2.2.2)


AEv = av * - (2.2.3) A2 (as - ao C1
E xx + Cyy + Ezz = 2 * ) * (1) - m)(2.2.4) a. C11

where

a,, av: hydrostatic deformation potential for the conduction band and valence band
Af)
A : fractional volume change due to strain effect
Q

C., 8_Y, Ez : diagonal components of the strain tensors







as, ao: lattice parameter for the barrier (GaAs) and quantum

well (In.Ga1_xAs)

C1I, C12: elastic stiffness coefficient



Since the exact value of either the hydrostatic deformation potential or the elastic stiffness coefficients of the

ternary material InGa _,As has not yet been determined, in this study we assume that both parameters can be expressed as the combination of those of InAs and GaAs. In other words, ac,

av, Cn and C22 for the In.Ga _xAs can be expressed as: a, = (1 - x) * ai(GaAs) + x * a1(InAs) i=c, v (2.2.5) C= (1 - x) * Cij(GaAs) + x * Cj(InAs) j=l, 2 (2.2.6)



Under the shear component of strain effects the valence band edge will become degenerated and split into heavy-hole band edge and light-hole band edge. The separation between heavyhole band edge and light-hole band edge is derived from the equations as shown below:




S = 2* 8es * (1- --) (2.2 .7)
A
&s = b * a-ao *(1 + 2 * 2) (2.2.8)

a. C0

b = (1 - x) * b(GaAs) + x * b(InAs) (2.2.9) A = (1 - x) * A(GaAs) + x * A(InAs) (2.2.10) where

S: bandedge split energy between heavy hole and light hole








&,,: heavy hole splitting energy b: uniaxial deformation potential A: split orbit energy



The lattice constant of InGa _xAs and GaAs can be expressed as ao(x)=(5.6533+0.405*x) A and a.=5.6533 A, respectively [Corz93]. As referenced from data shown in Van de walle [Van89], one can calculate the total energy gap shift caused by the hydrostatic and shear components of strain by substituting the related parameters into equation (2.2.1) to equation (2.2.10). Figure 2.2 shows the schematic diagram of bandgap lineup for In0. 15Ga0.85As/GaAs quantum well structure, where strain-induced bandedge split energy S between the heavy hole and the light hole, and the heavy hole energy bandgap Eg(HH) are computed as S=63.9 meV and Eg(HH)=1253.3 meV. A conduction band offset ratio of 0.55 is also assumed [Cole93] to compute the conduction band energy barrier, heavy hole valence band energy barrier and light hole valence band energy barrier as 93.9 meV, 76.8 meV and 12.9 meV, respectively. The energy state distribution of the InGaAs/GaAs quantum well structure is calculated by solving the Schrodinger equation [Bast88], in which a square well structure with finite barrier height is assumed. The effective masses used for the electron, heavy hole and light hole in the InGaAs quantum well are 0.0599 m., 0.4395 m. and 0.0847 m. respectively [Pan88], where m. is the electron rest mass. For the conduction band, the energy states inside the








100 A In0.15Ga0.85As quantum well are calculated as Eci=22.9 meV and Ec2=62.9 meV (referenced from the conduction bandedge). For the valence band, there are four heavy hole energy states: Evhl=6 meV, EVh2=23.8 meV, Eh3=52.2 meV and Evh4=86.8 meV, and one light hole energy state: E,11=82.5 meV (referenced from the heavy hole band edge) inside the quantum well. According to the results of the computations, a lasing wavelength of -960 nm is obtained for this InGaAs SQW diode laser and used for the following computations.

The modal loss induced by both shiny Au and less shiny Ni contacts are calculated from Modeig program [Deme89]. The refractive indices of Au and Ni at this wavelength are interpolated from [Weas87] as 0.091-i6.03 and 2.75-i4.87, respectively. The corresponding optical constants of each layer of the diode laser structure are obtained [Taka78], [Adac85]. Only the fundamental mode is considered from the calculated results in which the highest effective refractive index value of the diode structure is found. Figure 2.3 shows the dependence of the calculated mode loss aX on the thickness of the p-cladding layer when Au and Ni p-contact metals are used. From this plot, we can see as the p-clad layer thickness is reduced below 400 nm, the optical mode loss for the Ni contact lasers increases rapidly whereas that for the Au contact remains relatively flat. In other words, Au contacts are shiny compared to Ni contacts. At 250 nm pcladding layer thickness, the calculated mode loss of diode








laser with Au and Ni metal contact is a,(Au)=3 cm-' and ci(Ni)=70 cm-1, respectively. Since the threshold mode gain Gth is proportional to the modal loss of diode lasers, the larger modal loss induced by the Ni contact indicates that Ni contact lasers should have higher Gth than those of Au contact lasers. The magnitude of the increase in Gth required for lasing when using Ni rather than Au is given by [Chin88]


1 1
Gth = rgh = a1 + - ln - (2.2.11) L R

where r is the optical confinement factor, gt, is the threshold gain in the quantum well, L is the cavity length of diode laser and R is the mode reflectivity at the facets.

As stated earlier, for L in the range from 250 to 700 gm, the Ni and Au thin p-clad devices differ in operating wavelength by more than 50 nm. In order to understand this effect, it is necessary to calculate the spectral gain function g(hv) [Chin88] as well as threshold spectral gain gth for these two different optical mode loss devices.

Since the energy state distributions in the QW structure have been determined, the carrier density required for obtaining a given quasi-Fermi level can be calculated. In the computations, a parabolic subband model is used for conduction band, heavy hole valence band and light hole valence band. For the conduction band, the electron concentration is denoted [Corz93] as










N = 2f p2D(E - E:) * f, * dE
E. (2.2.12)
me * k8 1* T I ln l + exp[-(Ecn - Efc)/ kB*, T]}
n h , * L Z n



where

p 2D(E-E,) : density of states in the quantum well fc: Fermi-Dirac distribution for electrons in the conduction

band

m,: effective mass of electron in the conduction band KB: Boltzmann's constant E:n: quantized energy level of electron in the conduction band Efc: quasi-Fermi level in the conduction band Lz: quantum well thickness h: reduced Planck constant



For the valence band, the hole concentration is the summation of light hole concentration and heavy hole concentration, and can be expressed as [Corz93]



P=P h+PI and




Ph mh k B T {ln[l + exp( k] * (2.2.13)
= m1 2 *k Lz nhk *
mikaTEvn -EfV

P, = 2 k{ln[1 + exp( 1 (2.2.14)
p *2 Lz n kB *T









where mh , m1: effective mass of heavy hole and light hole; nh, n,: quantum state number of heavy hole and light hole in the valence band; EVh, Evn : quantum state energy for heavy hole and light hole in the valence band; Efv: quasi-Fermi level in the valence band.

By using equations (2.2.12), (2.2.13), (2.2.14) and the quantum state energy inside the quantum well, one can get the relationship between the quasi-Fermi level EfC, EfV and the corresponding carrier density in the conduction band and valence band.

In semiconductor the optical gain caused by the photoninduced transitions of electrons from conduction band to valence band is defined as the fractional increase in photons per unit length



1 dOF
g = -=WcV -W.c (2.2.15) (D dz

and can be written as[Corz93]


1 Ke2h nf 2
g(hv) = 2h - IMT!2Pred(fc - fV) (2.2.16)
hv �0cm02n n

where F is the photon flux (unit; s-cm-2); WCv, Wv+C: total transition rate from conduction band to valence band and from valence band to conduction band, respectively; ng, H : group refraction index and refraction index of the crystal; IMTI2:

transition matrix element; rred: reduced density of state; fc,







f,: Fermi-Dirac distribution for electrons in the conduction band and valence band. In the calculations, we assume the difference between n-q and H can be neglected. Considering the TE polarized transition, k-selection rule and ignoring the forbidden transition, the transition matrix element can be expressed as I 2TEn = ST 1M12 < FV nz 2
M T r. ' n z SE 2T E
VrvnzM < F rnz > rvnz IM[ (2.2.17)


where


STE = STE= 1 (1 + COS2 0"n) for e-hh transition STE = T 1 (5 - 3 COS2 en) for e-lh transition cos2 0nz = AErn, + AEvnz
hv - Eg


Premi = i=h for e-hh transition;
-2h2 Lz

1 1 1 = - + - i=l for e-lh transition. mr,i mc mi


(2.2.18) (2.2.19)


(2.2.20) (2.2.21) (2.2.22)


AErz: energy difference between the electron energy state and

the conduction bandedge E,.

AEvnz: energy difference between the heavy hole energy state

and the heavy hole bandedge or between the light hole

energy state and the light hole bandedge.



For the InGa1_xAs quantum well, 1M12 term in equation (2.2.17) can be written as [Corz93]








IMI2 = m (28.8 - 6.6 * x) (2.2.23)
2

By ignoring intraband relaxation [Garb93], the total optical gain is then calculated from equation (2.2.16) to equation (2.2.23) as the summation of optical gain due to ehh transitions and e-lh transitions. Figure 2.4 shows the dependence of spectra gain on photon energy for the diode laser structure shown in Figure 2.1, where the injected carrier density is assumed at N=2.5x1018 cm-3 and 5.5xi018 cm-3. It is evident in Figure 2.4 that the photon energies at peak gain for the two carrier densities are considerably different. At 2.5xi0'8 cm-3 injection carrier density, the peak gain occurs at an energy corresponding to the band edge transition energy from n=l conduction band electrons to n=l heavy holes, E11-1286 meV. When the injection carrier density increases to 5.1x1018 cm-3, the peak gain is the same at El. and E22 (the band edge transition energy from n=2 conduction band electrons to n=2 heavy holes, E22 -1364 meV) . Above 5.1x10'8 cm-3, the gain peak occurs at E22, The gth values shown in Figure 2.4 are calculated from equation (2.2.11) using the cL values from Figure 2.3, a confinement factor F=0.02, mode reflectivity R=0.31 [Lee9l] and cavity length L=380 gm. It is clear that the increase in gth required by the extra mode loss due to the Ni contact can cause a large shift in lasing wavelength.








2.3. Experimental Results and Discussion



In order to fabricate wide-stripe lasers, 50 Jim stripes on 500 gm centers are defined on the pt-GaAs contact layer using standard photolithography techniques. Mesas and current blocking oxide layer are formed using a pulsed anodic oxidation technique [Grov94]. Usually the oxidation time is

-6 minutes (with anodic oxidation current density J-100 mA/cm, repetition rate= 50 Hz and pulse width= 700 Jsec) to remove the p -GaAs cap layer and part of the p-cladding layer to reduce the diode current spreading effect. After the oxidation process, wafers are cleaned and covered with new photoresist to protect the epilayer surface for the following wafer thinning process. This wafer surface protection step is important for making good adhesion shiny contact for the diode lasers. The wafers are thinned to about 100 pLm, a metallurgy of Ge(20 nm)/Au(40 nm)/Ni(20 nm)/Au(15 nm) is evaporated sequentially on the substrate side and annealed at 430 C for 5 minutes to obtain the n-type ohmic contact. A non-alloyed Au(80 nm)/Ni(50 nm)/Au(50 nm) or Ni(80 nm)/Au(50 nm) is evaporated as the shiny or less shiny p-contact respectively. Then, another non-alloyed Ni(25 nm)/Au(50 nm) is deposited on the subtract side to improve the metal contact conductivity. Diode lasers with 500 Lm wide and cavity lengths ranging from 120 9m to 1200 pm are cleaved, soldered on copper blocks by using either indium soldering or







sliver epoxy (l part Ag epoxy + 1 part thinner and qued at 150 C for -5 minutes) and characterized.

The measured emission spectrum of Au and Ni lasers with L=380 pm is shown in Figure 2.5. The difference in operating wavelength is -52 nm, in good agreement with the predicted difference of -55 nm shown in Figure 2.4. The corresponding pulsed output power versus input current characteristics are shown in Figure 2.6. Threshold current and total slope efficiency for these two lasers are Ith(Au)=60 mA, T1(Au)=0.76 mW/mA and Ith(Ni)=240 mA, Tjr(Ni)=0.41 mW/mA respectively. In addition, from measured inverse slope efficiency versus cavity length plots, the internal mode loss for Au-contact lasers and Ni-contact lasers are a1(Au)=8 cm- and ai(Ni)=51 cm-1 respectively. These are to be compared with the calculated values from Figure 2.3, a1(Au)=3 cm-1 and a,(Ni)=70 cm-. The difference of 5 cm- in ai(Au) and 19 cm- in a1(Ni) can be explained if the complex refractive index of the Au, n(Au), and the complex refractive index of the Ni, n(Ni), used in our experiments are somewhat different from the values interpolated from [Weas87]. For example, if n(Au) is changed from 0.091-i6.03 to 0.16-i4.08, Cci(Au) changes from 3 cm-1 to measured value of 8 cm'. On the other hand, if n(Ni) is changed from 2.75-i4.87 to 2.30-i5.25, aZ(Ni) changes from 70 cm-1 to the measured value of 51 cm-1. We believe that







variations in both n(Au) and n(Ni) of these magnitudes are possible since it is known that refractive index values can be strongly dependent on the details of the deposition process.

In the previous section, it is mentioned that transition E1, and E22 have the same peak spectral gain at a carrier density of 5.1x108 cm-3. This implies that lasers requiring this carrier density to lase should operate simultaneously at two wavelengths about 50 nm apart [Zory86]. It is clear from equation (2.2.11) that it should be possible, in principle, to change cavity length L to achieve this condition for both Au and Ni contact lasers (L>380 gm for Ni lasers and L<380 Jm for Au lasers). In order to check this prediction, the emission wavelengths of both laser types are measured as a function of L. As shown in Figure 2.7, the emission wavelength jumps occur for the Ni lasers at L-900 pm and at L-200 gm for the Au lasers. Similar wavelength jumps have been observed for thick p-clad 7.5 nm InGaAs SQW lasers when their stripe widths are reduced from 50 to 15 microns (Shie89] or thick p-clad 15 nm InGaAs SQW lasers with 25 Jm stripe widths when their active region temperature is increased due to current [Beer9la] . Using the above L values in equation (2.2.11) along with the measured values of a, discussed above and R=0.31, we find that Gth=66.5 cm-1 for 200 Jm long Au contact lasers and Gth =64 cm- for 900 Jm long Ni contact lasers, in close agreement as expected. The blue shift of the spectral peaks accompanied with the decrease of








cavity length shown by the dashed lines in Figure 2.7 is mainly due to band-filling modified by bandgap renormalization [Chin88].

So far we have shown that the reflectivity of p-contact metal is important in determining the performance of thin pclad diode lasers, we are interested in understanding the effects of annealing p-contact metal on the thin p-clad diode laser. To check this point, thin p-clad diode lasers with shiny Au contact are annealed at 320 C for different annealing time: 30, 60, 120, 180 and 300 seconds. The variations of threshold current and total slope efficiency as a function of annealing time are shown in Figure 2.8. As can be seen from this plot, threshold current increases rapidly as the annealing time increases upto 120 seconds and then becomes saturated when diodes are annealed further longer. Additionally, the total slope efficiency decreases gradually from 0.8 mW/mA(0 second) to -0.5 mW/mA(120 seconds) and remains constant as annealing time increases further. The deterioration of diode laser performance can be explained by the decrease of contact reflectivity caused by the reactions of Au and GaAs at 320 C. This interaction will make the Au contact become less shiny. As the annealing time increases further, the interactions may saturate and cause the contact reflectivity keep constant. Consequently, even at low temperature (320 C), the annealing effect can decrease the contact reflectivity and diode laser performance. To have







good thin p-clad diode laser behavior, the shiny Au contact should not be annealed.

To further evaluate shiny contact thin p-clad diode lasers, samples without and with facet coatings are CW lifetime tested at 20 C with output power 50 mW/facet (w/o coating) and 100 mW (w/ coating), respectively, where both samples are packaged with p-side up configuration. The coated sample is prepared by depositing a thickness of X/4n Al203 on the emitting facet as the anti-reflection (AR) coating layer and A1203(X/4n)/Al(100 nm)/Al203(X/4n) on the reflective facet as the high reflector (HR) where n is the refractive index of the active region at the lasing wavelength. Figure 2.9 and Figure 2.10 show the output power P versus input current I characteristics and the corresponding variations of threshold current and total slope efficiency of diode laser without facet coating after t hours lifetime test. Similar results of the coated sample are shown in Figure 2.11 and Figure 2.12. For the uncoated sample, threshold current increases from the initial value of 42 mA to -56 mA and the total slope efficiency almost keeps constant at 0.65 mW/mA for the first 500 hours lifetime test and gradually decreases to -0.57 mW/mA after 1057 hours lifetime test. For the coated sample, threshold current increases quicker than that of uncoated sample from the initial value of 40 mA to -65 mA. In addition, the slope efficiency decreases faster than that of the uncoated sample from 0.62 mW/mA to -0.41 mW/mA after 1050








hours lifetime test. The possible reason for the big difference of Ith and TI between both diode lasers could be due to the cavity length difference where the uncoated sample is 600 gm long while that of the coated sample is only 450 gm. Since diode laser with longer cavity length is believed to have better lifetime performance than that of diode laser with shorter cavity length, especially for the high output power condition. From the results of lifetime test, we have successfully demonstrated that thin p-clad diode lasers can "live" for long time CW operation. Further applications from this thin p-clad laser structure could be possible as the fabrication process is improved and optimized.



In summary, we have demonstrated that the reflectivity of p-contact has a significant effect on the performance of thin p-clad diode lasers. Decreasing the reflectivity of the p-contact metal increases optical mode loss, which increases threshold current, decreases slope efficiency and shifts the emission wavelength. The more than 50 nm wavelength difference between the thin p-clad diode lasers with different p-contact metallurgy can be explained quantitatively by superimposing the threshold gain required for lasing in each case with the corresponding spectral gain curve calculated using standard QW laser theory.











































Figure 2.1 Thin p-clad InGaAs single quantum well diode
structure used in the contact reflectivity effect
study.


p+-GaAs contact layer: 100 nm p-AlzGal.zAs ( z=0.6 - 0.05 ): 25 nm p-A10.6Ga0.4As cladding layer: 250 rm AlxGal.xAs ( x=0.3 - 0.6 ): 200 nm

GaAs: 7 nm
InyGal.yAs SQW ( y-0.15): 10 nm

GaAs: 7 nm

AlxGal.xAs ( x=0.6 - 0.3 ): 200 nm n-Al0.6Ga0.4As cladding layer: 1400 nm n-AlzGal.zAs ( z=0.05-0.6 ): 25 rum

n-GaAs substrate







BULK MATERIAL:


I Eg(bulk)=1193.6 miev


GaAs


STRAIN EFFECT:


.82.6 mev Shear Deformation
-Hydrostatic

Bulk Ino.,Gaos5As . on o Hh
12.7 mev 35.6 M. 63.9 mev





BAND LINEUP UNDER STRAIN EFFECT: ( Assume conduction band energy offset ratio = 0.55)


93.9 mev


1424 meV


1253.3 meV


t12.9 mev


GaAs


In0.15G%.85As


GaAs


Figure 2.2 Schematic diagram of bandgap lineup between the
strained InGaAs quantum well and GaAs barrier layers of
the diode laser structure shown in Figure 2.1.



















P0 P0


250 200 150 100


50


0


100 200 300 400


p-clad thickness (nm)









Figure 2.3 The calculated relationship between the optical
mode loss and p-clad thickness for the InGaAs single
quantum well (SQW) laser structure shown in Figure 2.1.


500













I



0






'ml'
U


1250


1300 1350 1400


1450


Energy hv (meV)











Figure 2.4 The dependence of spectral gain on photon energy
for a 100 A In0. 15Ga0. 85As/GaAs structure at two carrier
densities. The two dashed lines are the active layer threshold gains for thin (0.25 gm) p-clad InGaAs SQW
lasers with Ni or Au contact metal.






































860 900 940 980

Wavelength (nm)









Figure 2.5 Measured lasing wavelengths of thin (0.25 Lm) pclad InGaAs SQW diode lasers with different p contact
metals.









120 * , , * * , , i , i , =


100 InGaAs SQW LD

4A - Au-contact
00 , Ni-contact


S60

o 40

*w=5) pgm
20 ., Lf380 pm


0
0 200 400 600 800 1000 Input current (mA)










Figure 2.6 Measured pulsed output power versus input current
characteristics of thin p-clad InGaAs SQW diode lasers
with Au contact or Ni contact.

















0




I


1.5


1A5


1.4


135 1.3 125


0 10 20 30 40 50 60 70 80

Inverse of cavity length 1/L (cm"')










Figure 2.7 Spectral position of peak lasing energy as a
function of reciprocal cavity length of thin p-clad InGaAs SQW diode lasers using Au or Ni as the p-type contact. (Each mark represents an individual laser.)


. . . . I . . . . I . . . . I . . . . I. . . .I . . . . I . . . I

InGaAs SQW LD


0 : Aucontact


4 : Ni-contact




-8-- - ....
-I i00



+ :0
0

11111i. oili fii m l I II oim i l 1 111 11


*w=50 gLm









200


150 100



50



0


0 60 120 180 240 300


Thin p-clad InGaAs SQW LI


S :Ith A :TlS

S I I

,IAg

- . . I .. . A. . .. .- . . . . . .


360


Time (second)










Figure 2.8 The variations of threshold current (Ith) and
total slope efficiency (7h) of thin p-clad diode lasers as a function of p-contact metal annealing time, where
the annealing temperature is set at constant 320 C.


2



1.5



1



0.5


0









. . '50 * * * I t I . . . .

4Thin p-clad InGaAs SQW LD
40
[4 t= 0 hr

529 hr
30 652 hr 1057 hr

o20


S10.
* L=600 gm, w=50 gm o w/o facet coating


0 50 100 150 200 250 Input current (mA)











Figure 2.9 Measured CW output power versus input current
characteristics of Au-contact thin p-clad InGaAs SQW
diode laser after t=O, 529 and 1057 hrs lifetime test, where the output power is set constant CW 50 mW/facet
during the lifetime test.









200 150


100



50


0 200 400 600


' ' . I ' I ' I I I ' 5 I S

* L=600 Prm, w=50 iim
w/o facet coating








Pon@* .~ . ......


800 1000 120


Time (hr)










Figure 2.10 The variations of threshold current and total
slope efficiency of Au-contact, thin p-clad InGaAs SQW diode laser as a function of lifetime test period when
operated with constant output power CW 50 mW/facet.


0.8


.6.


0.4w

0.24 0 0











Thin p-clad InGaAs tf Ohr SQWLD
60 192 hr

517hr
0 w/ AR,HR
40 facet coating 1050 hr w=50 m /
L=450 im

20


0 r

0 50 100 150 200

Input current (mA)












Figure 2.11 Measured CW output power versus input current
characteristics of Au-contact thin p-clad InGaAs SQW diode laser after t=O, 192, 517 and 1050 hr. lifetime
test, where sample is with AR and HR facet coating and
output power is set constant CW 100 mW during the
lifetime test.










200


150


100



50


f' I


U . . . . . . . . . . . .. . . . . . .
0 200 400 600 800 1000 12(
Time (hr)


' ' I ' I * ' ' I * ' ' I ' I ' '













* L=450 4.m, w=50 gm
w/facet coating


Figure 2.12 The variations of threshold current and total
slope efficiency of Au-contact thin p-clad InGaAs SQW diode laser as a function of CW 100 mW constant output
power lifetime test period, where sample is with AR
and HR facet coating.


0.8 0.6


0.4 0.2


0
)0














CHAPTER 3

THIN P-CLAD RIDGE-GUIDED DIODE LASERS







3.1 Introduction



In chapter 2 we have demonstrated the important effects of p-metal contact reflectivity on the thin p-clad diode laser performance. In this chapter, a new and simple method of fabricating ridge-guided diode laser from a thin p-clad structure is proposed and demonstrated. The major advantage of using a thin p-clad diode laser structure instead of the conventional thick p-clad structure for fabricating ridgeguided diode lasers is that only a small part of outside stripe material is required to remove for providing a strong waveguiding. This will make the device fabrication process become simple and economical. Moreover, we have found that these low-ridge, thin p-clad lasers can operate in a single lateral mode with cw performance characteristics similar to those reported for InGaAs SQW lasers fabricated from conventional epitaxial material in a high-ridge configuration [Fisc87, Bour90, Take90]. In other words, thin p-clad diode laser with low ridge height structure can provide strong








waveguiding for diode lasers operating with low threshold current (Ith) and high slope efficiency (T) . This is totally different from those of thick p-clad diode laser structure in which the threshold current of an InGaAs single quantum well ( SQW ) diode laser is very high ( > 100 mA ) when fabricated in either a low-ridge [Shie89] or oxide-defined [Beer9lb], narrow-stripe configuration.

In Section 3.2, the main design principle responsible for low threshold, single lateral mode lasing is discussed. In Section 3.3, the procedure for fabricating ridged-guided, thin p-clad lasers using pulsed anodic oxidation [Grov94] is outlined. The results of pulsed characterization measurements on these lasers are presented in Section 3.4 and compared with measured values reported for thick p-clad lasers [Shie89, Beer9lb]. In Section 3.5, the cw performance characteristics of 5 micron wide, low-ridge, thin p-clad lasers are presented and compared with those of reported thick p-clad, high-ridge waveguide lasers [Fisc87, Bour90, Take90]. In Section 3.6, the dependence of diode laser performance on the ridge height is presented and comparisons are made between thin and thick p-clad ridge-guided lasers with various ridge heights.



3.2 Theoretical Calculations



The thin p-clad diode laser structure used in this study is the same as that stated in previous chapter ( as shown in







Figure 2.1 ) . Since the thickness of the p-clad layer is only 250 nm, the behavior of ridge waveguide lasers fabricated from this material should be quite different from that of ridgeguide lasers with the same ridge height fabricated from conventional material with thick p-clad layers ( > 1000 nm ). The parameter normally used in designing ridgeguide lasers is An, the difference in effective refractive index ( lateral index step ) between the ridge region and the outside-ridge region [Agra84]. Consequently, it is of interest to calculate An as a function of ridge height ( the amount of epitaxial material removed in defining the ridge ) for thick and thin p-clad configurations. In the calculations, it is assumed that the p+-GaAs contact layer, the p-AlxGal-xAs graded-layer and part of the p-clad layer of the outside-ridge region are removed and replaced with a native oxide, the whole structure being covered with Au. A lasing wavelength of 960 nm, an oxide refractive index of 1.8 and a Au refractive index of 0.091-i6.03 are also assumed in the computations. The optical constants of each layer of the laser structure are the same as those used in our previous work on thin p-clad InGaAs SQW lasers [Wu94b] . Figure 3.1 shows the calculated An of thin ( 250 nm ) p-clad diode lasers as a function of ridge height when various oxide values are used as the current blocking layer. It is interesting to note that relatively large values of An are obtained with the removal of small amounts of material. For example, a An value of greater than 3x10-3 is







obtained when 130 nm of epitaxial material is removed and replaced by 130 nm of native oxide. If a conventional thick p-clad layer is used instead of a thin p-clad layer ( 1300 nm vs 250 nm ), a An value of only ~Ix10-6 is obtained for the same removal/replacement process. In order to achieve a An of 3x10-3 in the 1300 nm thick material, an order of magnitude more material ( -1200 nm ) must be removed from the laser structure creating a "high-ridge" device. From a An point of view therefore, a low-ridge, thin p-clad device is equivalent to a high-ridge, thick p-clad device. If An is the key parameter which determines device performance, then lowridge, thin p-clad lasers should not show the high threshold effect as ridge width is narrowed [Shie89, Beer9l]. In addition, they should be capable of operating in a single spatial mode to high cw power levels if the ridge width is sufficiently narrow [Fisc87, Bour90, Take90]. These two speculations are addressed in Sections 3.4 and 3.5 respectively.



3.3 Device Fabrications



Thin p-clad lasers with 130 nm ridge height and four different ridge widths ( w=5 jm, 10 gm, 25 pm and 50 pm ) are fabricated using a pulsed anodic oxidation technique [Grov94]. In this technique, photoresist is used to define the ridge widths on the laser wafer which is then used as one







of the electrodes in an electrolyte composed of 40 parts ethylene glycol, 20 parts water (H20) and one part phosphoric acid (H3PO4) . Running 700 gsec wide current pulses through the electrolyte at 50 Hz repetition rate converted 130 nm of epitaxial material into 100 nm of native oxide in about 3.5 minutes (anodic oxidation current density -100 mA/cm2) . The laser material is then thinned to about 100 gm and a Ge/Au/Ni/Au contact metallurgy deposited onto the n-side by electron beam evaporation. After a standard high temperature anneal, a Au/Ni/Au contact metallurgy is electron beam evaporated onto the p-side and 450 gm wide bars cleaved from the material. The cleaved bars are then scribed into 500 pm wide chips and soldered, substrate-side-down, onto indiumcoated copper blocks. It is important to note that the p-side contact was not annealed. If it were annealed, the p-contact would not be "shiny" and laser thresholds would be very high as shown in chapter 2.



3.4 Pulsed Current Characteristics



As discussed in Section 3.2, low ridges in thin p-clad material produce relatively big An values ( An - 3x10-3 when using 250 nm-thick p-clad material with 130 nm ridges ) . To see if this An is sufficient to avoid the high threshold effect at narrow ridge widths reported in references 1 and 2, the pulsed threshold current ( Ith ) and laser emission energy of the 130 nm-ridge, thin p-clad lasers are measured and








compared with the data reported for thick p-clad devices [Shie89, Beer91] . In Figure 3.2, the thick p-clad lasers are

(a), 200 nm-ridge lasers [Shie89] with 400 Am cavity length and (b), oxide-defined stripe devices [Beer9lb] with 510 pm cavity length. As can be seen, Ith for both laser types increases as the stripe width is narrowed, the increase being very large for the oxide-defined stripe lasers. Moreover, there is a lasing energy jump [Wu94] associated with the increase of threshold current for both thick p-clad laser types. In contrast, as the stripe width of the 130 nm-ridge, thin p-clad laser is decreased from 50 pm to 5 Jim, I th decreases from 50 mA to 10 mA and the lasing energy is essentially constant. It appears that the lateral index step An - 3x10-3 is sufficiently big in the thin p-clad, 130 nm ridgeguides to create a low loss lateral waveguide. If An is too small, the lateral dimension of the lasing mode spreads out into the unpumped region causing a large mode loss. This forces the required carrier density for lasing to be very large leading to carrier-induced antiguidance and additional mode loss [Shie89, Beer9lb].

Since a low loss lateral waveguide appears to exist in the thin p-clad, 130 nm ridgeguide lasers, their lateral farfields should also be quite different from those observed in references [Shie89, Beer9lb]. To check this point, we have to measure the lateral far-field distributions of thin p-clad diode lasers by using the setup shown in Figure 3.3. In the measurements, diode lasers are kept at room temperature and








by rotating the detector to obtain the far-field distributions. As shown in Figure 3.4, a single-lobe lateral far-field pattern is obtained for lasers with 5 Jim, 10 pm and 25 gm ridge widths at I-1.1 I th. The single-lobe far-field pattern of the 5 gm ridge-width diode laser is about 5 times narrower than the single-lobe pattern reported in (Shie89] and totally different from the double-lobe pattern reported in [Beer9lb]. As the drive current increases to -1.5 1 h, both 5 pLm and 10 gm lasers remain single-lobed with no significant change in pattern width. The lateral far field pattern for the 25 Jim laser however, changes from single-lobed to doublelobed as the current increases to 1.5 I Since similar
th"*

behavior is observed in the far field during cw operation of the 25 Jim device, the change can probably be attributed to spatial-hole burning [Garr87] rather than thermal waveguiding [Bour90]. At I=1.5 ITh' the 50 pm stripe-width laser still remains double-lobed.



3.5 CW Current Characteristics


As discussed in Section 3.4, a An value of - 3x10-3 is

sufficient to avoid the high threshold effect at narrow ridge widths. It is interesting to ask if this An is also sufficient to allow single lateral mode operation to high cw power levels when the ridge width is 5 p1m. To check this








point, the cw lateral far-field patterns of several 5 jm wide, 130 nm-ridge, thin p-clad diode lasers are measured at four drive currents: 40 mA, 80 mA, 120 mA and 160 mA ( see Figure 3.5 ) . Typically one observes that the full width at half maximum (FWHM) of the patterns ( - 100 ) increases by less than 10 as the current is increased from 40 mA to 120 mA. Between 120 mA and 160 mA, the FWHM increases by about 30. In order to understand the cause of this broadening, the lateral far field patterns are also measured using short pulse excitation. In this case, no broadening is observed over the entire current range leading us to believe that the slight cw broadening observed up to 120 mA is due to a narrowing of the near field caused by thermal index guiding [Bour90]. While the relatively large increase of the broadening above 120 mA could be due to the onset of higher mode oscillation, the lack of any significant beam steering tends to rule against this possibility [Guth94]. In any event, the addition of a heat spreading layer should be very effective in reducing this broadening effect since the p-cladding layer is very thin. The measured FWHM value of the transverse far-field pattern is - 420, independent of current

The measured cw output power versus drive current ( P-I

characteristic is plotted in the inset of Figure 3.5. As shown, the slope efficiency stays constant up to about 120 mA ( - 70 mW of cw total output power ) and then begins to drop at higher currents due to heating. Threshold current and total differential quantum efficiency are 10 mA and 49 %,








respectively. The threshold current of 10 mA is comparable to that reported for conventional SQW InGaAs high-ridge lasers [Fisc87, Bour90, Take90] although the differential quantum efficiency is - 10 % smaller. In order to see if this result could be improved on, thin p-clad diode lasers are fabricated with a 180 nm-ridge height and, according to Figure 3.1, a An of close to 5x10-3. In this case a cw total differential quantum efficiency of 61 % (cavity length L=600 pm and stripe width w=5 pm) and threshold current of 9 mA are measured. Figure 3.6 shows the measured CW P-I characteristics of the 5-pgm stripe, 180 nm ridge lasers when different cavity length L are considered. As can be seen from this plot, the 180 nm ridge lasers show comparable P-I performance with those reported for the conventional ridge-guided lasers and indicate strong waveguiding in these thin p-clad, narrow stripe lasers.



3.6 Ridge Height Dependent Diode Laser Performance



It is mentioned in Section 3.2, that high ridge structure has to be performed for thick p-clad narrow stripe ridge-guided diode lasers to obtain sufficient index step in operating at low threshold, single spatial mode regime. In addition, CW measurements of thin p-clad ridge-guided lasers have indicated that diode laser performance greatly depends on the ridge height of the device. Since we are interested in








understanding the dependence of narrow stripe laser behaviors on the ridge height, thin p-clad (250 nm) lasers as well as thick p-clad (1300 nm) lasers with 5 gm stripe width and various ridge heights are fabricated and characterized. Both types of samples are prepared by using similar process steps as described in Section 3.3. Thick p-clad samples are anodic oxidized to get right heights: -230 nm, 530�30 nm, 1000�50 nm, 1150�50 nm, 1220�30 nm and 1350�50 nm respectively. On the other hand, thin p-clad samples are oxidized to have ridge heights: 130�5 nm, 180�5 nm and 260�5 nm.

Figure 3.7 shows the measured variations of pulsed P-I characteristics of thick p-clad (1300 nm) 5-11m stripe diode lasers at different ridge height. Figure 3.8 shows the variations of threshold current and slope efficiency as a function of ridge height. It is noticed from Figure 3.7, that P-I behavior of diode laser is not linear until the associated ridge height exceeds 1000 nm for this thick p-clad (1300 nm) laser structure. This is due to the occurrence of higher order modes which generate when the ridge waveguiding is not strong enough to support only single mode operation [Thom80]. The non-linear P-I performance also explains the weak waveguiding inside the low-ridge thick p-clad diode laser and agrees well with the theoretical predictions of Section 3.2. Even for the 1000�50 nm ridge height laser, the non-linear P-I characteristics still occur as the injection current increases up to -100 mA. From Figure 3.8, it is observed that threshold current (Ilh) of thick p-clad ridge








guided laser decreases as the ridge height increases to -1000 nm and becomes saturated when the ridge height increases further in the measured range. In this study, the minimum I,, value obtained for the 5-gm stripe-width, 500 pm long, thick p-clad diode laser is -10 mA. As shown in Figure 3.8, for the samples without significant non-linear P-I performance, the slope efficiency TI increases and reaches the maximum value at ridge height -1150 nm, whereas %, shows a decrease as ridge height increases from -1150 nm to -1350 nm.

For the thin p-clad samples, however, non-linear P-I characteristics are not observed in the measured range as shown in Figure 3.9. This is due to the strong waveguiding from the low ridge structure. In addition, from the corresponding It., 11 vs ridge height plot as shown in Figure 3.10, threshold current shows a small value decrease from Ith

-11 mA to Ith -10 mA as the ridge height increases from -130 nm to 260 nm. This is dramatically different from those of thick p-clad lasers, where -5 time decrease of Ith occurs as the ridge height increases from low ridge (-230 nm) to high ridge (-1000 nm) . Additionally, the 71, value of thin p-clad ridge-guided laser increases as the ridge height increases and reaches a maximum point then decreases as the ridge height becomes higher. The decrease of slope efficiency is due to the contraction of lateral mode caused by the strong index-guiding which could make the mode profile narrower than








the gain profile resulting in a lower slope efficiency [Agra84]. In this thin p-clad laser structure, the 180 nm ridge-height laser shows the best performance of slope efficiency.



In summary, we have demonstrated that the lateral refractive index step generated in low-ridge, thin p-clad diode lasers is sufficient to provide low loss lateral waveguiding. The threshold current of 130 nm-ridge diode lasers decreases from 50 mA to 10 mA when the ridge width is decreased from 50 gm to 5 gm whereas the lasing energy remains essentially constant. In addition, the 5 gm stripe devices were shown to be capable of stable, single lateral mode cw lasing with less than 10 % broadening up to total output power levels of about 70 mW. On the contrary, for the thick p-clad (1300 nm) diode laser, deep etching for obtaining ridge height larger than 1000 nm is required to provide the sufficient waveguiding effects for narrow stripe lasers operating with stable single mode output. By the optimization of diode laser structure, single spatial mode operation with linear P-I profiles to much higher cw powers should be possible. This feature with the combinations of the flexibility of fabricating different diffraction grating types (nickel, gold metal grating or dielectric material grating) in thin p-cladding material ( no regrowth required ) could lead to the development of higher performance gaincoupled DFB lasers [Luo92].









anodic oxide p-clad layer In0.15GaO.85As quantum well n-clad layer -


CD,


3.0r

2.5 2.0 1.5 1.0 0.5

0.02
125


175 225 275 325
Ridge height (nm)


(b)
Figure 3.1 (a) Schematic diagram of narrow stripe thin p-clad
InGaAs SQW ridge-guided laser. (b) Calculated lateral index step An of laser structure in (a) as a function
of ridge-height for various current blocking oxide
values. It is assumed that the ridge guide structure is
covered with Au.


stripe region


ridge
height


375















-4.1


600 500 400 300 200 100
0


1A5 -B - - (a), thick p-clad
-A- -(b), thick p-clad
140 -*... (c), thin p-clad
135
130 .............. a ............-- ----- ---4 125 - --A
120 :,,,, I ... . , . . I . .. . . , I. ,,
0 10 20 30 40 50 60
Stripe width (gm)









Figure 3.2 The dependence of threshold current and laser
emission energy of InGaAs SQW diode lasers on the
stripe width for (a) thick p-clad, 200 nm-ridge, L=400
gm [Shie87], (b) thick p- clad, oxide-defined stripe, L=510 gm [Beer9l], and (c) thin p-clad, 130 nm-ridge,
L=450 Jim [this work].


p' PU * .... ... . .5 U ' ill � I ' ' I '" ' ' 6- -A- -(b), thickp-clad
S----- (c), thin p-clad



- ...- ...... -..............













Detector and slit


Figure 3.3 Schematic diagram of far-field measurement setup
used in this study.






















*


N




0
z


0.8 0.6


0.4 0.2


0


-10 -5 0 5 10 15
Lateral Angle (degree)












Figure 3.4 The measured lateral far-field intensity
distribution of 130 nm-ridge, thin p-clad InGaAs SQW
diode lasers as a function of ridge width.


' a a ' I a a ' i I . . . I * i * * ' I * S a a


Stripe width w=50~m

-w=5 Oim. � "'





: i ai ',.~ l~w
* 10% duty cycle =*th,5


, I I 25C
* , I ", ,. a . I. . . . I . . . . I . . . I . n , ,





































-20
Lateral


0 20
angle (degree)


Figure 3.5 CW lateral far-field patterns of a 5 Jim wide,
130-nm ridge, thin p-clad InGaAs SQW laser with
uncoated facets at a heat sink temperature of 25 oC.
The corresponding total output power versus drive current characteristic is shown in the inset plot.























04


30 25 20 15 10

5


0


0 20 40 60 80 100 120


Input current (mA)


Figure 3.6 The variations of CW P-I characteristics of 5-Jm
stripe, thin p-clad InGaAs SQW diode lasers with 180 nm
ridge height and different cavity length.
















06


100


80 60 40 20


0


0 20 40 60 80 100 120 140
Input current (mA)










Figure 3.7 The variations of pulsed P-I characteristics of
5-gm stripe thick p-clad InGaAs SQW diode lasers with
different ridge height.



























800


1200


1


100


Ridge height (nm)










Figure 3.8 The measured dependence of threshold current Ith
and slope efficiency 7s of 5-Lm stripe thick p-clad
InGaAs SQW diode lasers on the ridge height.


400


. . I I F I I I F I ' S 1

Thick p-clad InGaAs SQW LD

* p-clad=1300 rum
L=500 jim, w=5 jim









, , I , , , I , , , I ,*


1.0

0.8 Q 0.6

0.4

0.2

0
600










7 0 .. . . . . , , , , , , , ' , , *' '

60 Thin p-clad InGaAs SQW LD

~50 ridge height / [~] /
1 +30+_5 nm // .'

'40 180�5nm /
----- ---260+5nm
30- / -"'
0
20 7

10 /7*w=5 Lm, L=500 gm


0 30 60 90 120 150
Input current (mA)










Figure 3.9 The variations of P-I characteristics as a
function of ridge height for thin p-clad 5-jim stripe
diode lasers.




























200


25



20 i15



"10

W

-4 5


250


1 I ' I I I I I I I

Thin p-clad InGaAs SQW LD





4






I , , , I a , , , I , a a , I , a , ,


Ridge height (nm)


Figure 3.10 The dependence of Ith and Th on the ridge height
of 5-gm stripe thin p-clad InGaAs SQW diode lasers.


r


100


0.6




0.4




0.2


300


150















CHAPTER 4

DUAL WAVELENGTH DIODE LASERS AND THERMAL RESISTANCE IMPROVEMENT







4.1 Introduction



Diode laser emitting simultaneously at multi-wavelength is very attractive for the applications to wavelength division multiplexed communication systems. Several reports have been issued about the fabrication of a monolithic multiwavelength emission diode laser. The first approach is to use different active layer composition. In this approach, either selectively etching [Saka82] or etch-and-regrowth technique [Boua82] has to be accurately performed in order to obtain reliable diode laser quality. The second approach is to utilize complex grating technique where different lasing wavelengths are obtained by either changing the stripe width to vary the index of refraction [Dutt86] or different periods of DFB grating [Aiki761. The third approach is to control the modal loss through the change of stripe width of diode laser to obtain lasing on either the first or the second quantized energy levels of a single quantum well active layer [Toku86].








In this method, high temperature and time-consumed material disordering process is used and make this method become hard to control. In addition, more complicated method of using laser-induced desorption to selectively decrease the single quantum well thickness has been developed to fabricate multiple wavelength emission diode lasers [Eple90. Previously, we have demonstrated that contact reflectivity has a great effect on the thin p-clad SQW diode lasers [Wu94b]. One of the important characteristics is that, by selecting shiny or less shiny contact metal, one can control the modal loss of diode laser and make the laser lase on either the first or the second energy level of a single quantum well active layer. Based on this concept, the process of fabricating the dual wavelength emitting diode lasers should be much simpler and more reliable than the other approaches as stated above. In this chapter, a twin stripe diode laser structure with dual wavelength lasing is proposed and demonstrated.

Recently, the influence of heat spreader on the laser performance has been reported on AlInP red light emission material system and shows significant CW P-I performance improvements (Unge93]. However, the dependence of thermal resistance on the heat spreader thickness and the associated effects on the InGaAs/GaAs material system are seldom reported. In this chapter, a thick Au plating technique is developed and applied to the devices fabricated to show the









reduction of thermal resistance and performance improvements of InGaAs/GaAs diode lasers.

In Section 4.2 the dual wavelength laser structure and fabrication process are described. The experiment results and discussion of the dual wavelength diode lasers are stated in Section 4.3. The main principle of thermal resistance measurements, details of thick Au plating technique as the heat spreader and the performance improvements of thin p-clad InGaAs/GaAs diode lasers with Au heat spreader are presented in Section 4.4.



4.2 Fabrication of Dual Wavelength Diode Lasers



Figure 4.1 shows the twin stripe diode laser structure used in this study. Basically, the diode laser structure is the same as shown in previous chapters except a twin stripe geometry is used instead of a single stripe. The stripe width of Au contact and Ni contact laser is designed as 50 jim and 20 Jim respectively in order to have close threshold current for both laser types. The separation distance between these two stripe edge is 50 gm with an isolation slot of 20 gm used to make these two diode laser operate independently.

Wafer is first cleaned in boiling TCA, ACE and methanol. Then, photoresist AZ-1350 is spreaded and baked to define a 20 gm window. Since the isolation slot is important to make the twin stripe diode lasers operate independently, the etching depth of the isolation slot has to be larger than 0.6








Jim for this laser structure. The etching depth can be obtained by chemical etching, anodic oxidation or the combination of both methods. In this experiment, the isolation slot is first etched by the chemical solution of 1 part NH4OH + 1 part H202 + 50 parts H20 at room temperature. After the etching step, wafer is cleaned and spreaded with new photoresist to define the two-stripe patterns (w=20 pm and w=50 gm) . Pulsed anodic oxidation is then performed to remove parts of the p+-GaAs cap layer and grow an oxide layer to cover the whole area except the twin stripes. Wafer is then covered with new photoresist and thinned to about 100 pm. A metallurgy of Ge(20 nm)/Au(40 nm)/Ni(30 nm)/Au(50 nm) is evaporated on the substrate side and annealed at 430 C for 5 minutes to get n-type ohmic contact. Photoresist of AZ-1375 is spreaded to define p-contact patterns for the lift-off process. At first, by using lift-off technique, Ni(60 nm)/Au(50 nm) are deposited sequentially to form the Ni-metal contact for the 20 im stripe lasers. Following this step, photoreisst AZ1375 is used again to define Au contact pattern. A metallurgy of Au(80 nm)/Ni(50 nm)/Au(30 nm) is sequentially evaporated and lift-off to get the Au-metal contact for the 50 pm stripe lasers. Finally, diode lasers are cleaved, soldered on the copper blocks and characterized. 4.3 Dual Wavelength Lasers Experiment Results and Discussion


Figure 4.2 shows the top view and the side view of the twin stripe diode laser after being soldered and bonded on









the copper block. The measured current (I) versus voltage (V) characteristics of the twin stripe diode lasers are shown in Figure 4.3(a) and 4.3(b). As can be seen, both devices show the same cutin voltage Vc-l.4 Volts. At I=50 mA, the corresponding voltage value of the Ni contact laser is little bit higher than that of Au contact laser (2.1 Volts versus 2.0 Volts), which could be attributed to the fact that Au contact device has larger contact surface than that of Ni contact laser. Figure 4.3(c) shows the I-V characteristic measured between Au contact electrode and Ni contact electrode. Even at high voltage V=12 volts, the current-flow between these two electrodes is very small (< 20 JiA) and indicates good electric isolation between these two diode lasers.

The measured pulsed P versus I characteristics of the twin stripe diode lasers are plotted in Figure 4.4. At room temperature the slope efficiency 71 and threshold current Ith of the diode laser with cavity length L=350 gm are 11(Au)=0.55


mW/mA per facet, Ilh(Au)=45 mA and Tj,(Ni)=0.18 mW/mA per facet, Ih(Ni)=105 mA respectively. At current I=150 mA, more than 8 mW/facet output power from the Ni contact laser and more than 50 mW/facet from the Au contact laser can be obtained. The PI characteristics are comparable to those data reported before [Aiki76, Boua82, Saka82, Dutt86]. The lasing wavelengths of the twin stripe diode laser at injection current 1-1.5 Ith are shown in Figure 4.5. At II=160 mA and








12=0 mA, only the Ni contact laser can lase with the emission wavelength X,=914 nm. While at I1=0 mA and 12=65 mA only the Au contact laser can lase with the lasing wavelength A2=967 nm. When I1=160 mA and 12=65 mA, both laser can lase with X1=914 nm and X2=967 nm, respectively. Therefore a monolithic diode laser with dual wavelength emission capability has been successfully demonstrated by using contact reflectivity effect on the thin p-clad diode laser for the first time. This method of achieving dual wavelength operation is much simpler and more reliable than the other approaches as stated in Section 4.1.

To further check the lasing characteristics of the twin stripe diode laser, near field emission patterns are observed under three different injection current conditions. Figure 4.6(a) shows the near field pattern of the Ni contact laser at I1=120 mA and 12=0 mA. Figure 4.6(b) is the near field pattern of the Au contact laser at I1=0 mA and 12=60 mA. Under these independent operating conditions, no significant spontaneous emission radiated from the unpumped stripe region is observed and indicates a negligible leakage current between these two devices which is consistent with the results shown in Figure 4.3 (c) . Figure 4.6(c) shows the near field pattern as I1=120 mA and 12=60 mA. No significant light intensity increase is found in each individual diode laser as compared to the results of Figure 4.6(a) and Figure 4.6(b).








This also demonstrates that these two devices can be operated independently.



4.4 Thermal Resistance Improvements



Low thermal resistance are essential for diode lasers to increase the CW operation temperature and maximize CW output power [Jone75]. Thick Au heat spreader is found to play an important role for improving thermal resistance of diode laser when soldered with epi-side up configuration [Joyc75]. To have thick Au deposited on the top of p-contact of diode lasers, Au plating is the quick and economic method. Two important points are found to be crucial for obtaining thick Au when using Au plating technique: (1) the thickness of photoresist (PR) for the patterns defined and (2) the conditions used for Au plating. In this study, -5 Jim thick photoresist is obtained by spreading PR AZ1375 three times at 4000 RPM for 30 seconds each time. In addition, an optimal pulsed Au plating condition (pulse width 700 sec.) is obtained to have thick Au heat spreader. The evolution of the optimal Au plating condition as a function of time is shown in Figure 4.7. It is found that the initial condition is important for obtaining good quality thick Au film, where Au plating current density and repetition rate must be adjusted appropriately to avoid too high values used. Moreover, Au thickness obtained does not only depends on the plating time but also depends on the plating conditions and Au density








inside the plating solution. Normally, the plating conditions shown in Figure 4.7 will give one 4-5 gm Au thickness. By following the Au plating process as shown in Figure 4.7, diode lasers are Au plated, cleaved, soldered and characterized.

Figure 4.8 (a) is the schematic diagram of the Au heat spreader designed for the experiments where the separation between each individual Au heat spreader is important for the diode laser cleaving process. The finished top view of the heat spreader on the p-contact metal is shown in Figure 4.8(b). Figure 4.9 shows the CW P-I characteristics of 5-pm stripe, thin p-clad diode lasers with and without thick Au (-7 pm) heat spreader (HS) . It is interesting to note that the linear P-I performance of sample without heat spreader is less than 100 mW output power, however the sample with heat spreader shows a linear P-I performance up to 150 mW. In addition, the maximum output power Pmax significantly improves from Px(without HS) -150 mW to Pmax(with HS) -220 mW. These

-50% improvements in both linear CW P-I performance and Pma. value apparently signify the important heat dissipation effects provided by the thick Au heat spreader.

To further understand the influence of heat spreader on the diode laser performance, we have to measure diode laser thermal resistance. Figure 4.10 shows the setup used for the thermal resistance measurements. The main principle for this setup is to utilize the temperature dependence of refractive index n within the laser waveguide. The advantage of this








measurement setup is that no preliminary calibration measurement is required and only a null measurement of the exact wavelength of a single Fabry-Perot mode is made for the Rth determination [Paol75] . In this study, diode lasers are driven at a constant current near threshold (I~Ith) with 2 pIsec constant pulse width. The temperature of laser resonator is controlled by varying the input current duty cycle through the change of its repetition rate. The variations of laser temperature can cause lasing wavelength shift. The wavelength shift due to the temperature change from T+AT to T can be expressed as

AXFP=xFP(T+AT)-XP(T)=(dk/dT)AT (4.4.1)


where Akp is the Fabry-Perot mode shift due to the variation of refractive index, XFP(T+AT), Xp (T) are the modes corresponding to the refractive index at temperature T+AT and T respectively.

From the laser oscillation condition, the wavelength X can be written as

X,=2nL/q (4.4.2)

where q is an integer and L is laser cavity length. Therefore, AX P can be expressed as


AX P={d(2nL/q)/dT}AT=(X/n) (dn/dT)AT (4.4.3)

From equation (4.4.3), if we assume the refractive index change of InGalxAs (x-0.15) quantum well is the same as that









of GaAs as (dn/dT)=4x10-4 (Mar 641, then we can get &kP/AT-1.1 A/K. Since the width of a Fabry-Perot mode near threshold can be less than 0.2 A/K, the wavelength of a single longitudinal mode is a sensitive indicator of the temperature of the laser resonator [Pao175]. Figure 4.11(a), (b) and (c) show the wavelength shifts of the Fabry-Perot modes of 50 jLm stripe width diode laser at different duty cycles and heat sink temperature. The wavelength shift of the selected Fabry-Perot mode is compensated by the drop of heat sink temperature ATHs. Since the heat sink temperature change is a linear function of duty cycle [Paol75], one can obtain the temperature change ATc. by measuring ATHS only to certain value (<100 %) of duty cycle and get the temperature rise ATc, from the plot of ATHS versus duty cycle. Figure 4.12 is the result of heat sink temperature change as a function of operation duty cycle for the 50 gm stripe lasers with and without thick Au heat spreader. In this plot, ATc, is obtained by elongating the linear relationship to 100% duty cycle and get the corresponding intercept ATHS value i.e. ATcw. After the temperature rise AT, is determined, thermal resistance Rth of the laser can be calculated from ATcW=RthPcw [Paol75], where Pw is the average supplied electrical power given by the current I times the diode voltage V. Figure 4.13 is a typical plot of I-V and CW P-I characteristics for the thin p-clad SQW 50 jim stripe diode laser. Since the selected operating current for R th measurement is close to the threshold current, the radiative output power is only a small fraction of the input








power, one can get Pcw value directly from Figure 4.13 and calculate Rth. From the measurements, the thermal resistance for the sample with heat spreader and without heat spreader as shown in Figure 4.12 are Rth(w/HS)=70 C/W and Rth(W/O HS)=I02 C/W respectively. Obviously, samples with heat spreader can have less thermal resistance than those without heat spreader.

In order to check the dependence of thermal resistance on the thickness of heat spreader, wide stripe (50 gxm) thin p-clad lasers are deposited with various thickness of Au heat spreader and thermal resistance measured as stated above. Figure 4.14 shows the measured variations of thermal resistance of diode laser as a function of Au plating thickness. It is noted that thermal resistance decreases significantly for the first 7-8 4m Au heat spreader thickness and becomes gradually saturated when Au thickness is thicker than 8 gm.

Since diode laser performance is temperature dependent, threshold current Ith of diode laser can be normally expressed as Ith(T) exp(T/To), where T, is the characteristic temperature of diode laser. We are interested in understanding if there is any relationship between diode laser thermal resistance and its temperature related performance. To check this point, diode lasers with different thermal resistance are measured at various temperature (from 20 C to 80 C) in both pulsed and CW operation. Figure 4.15 shows the variations of laser characteristic temperature T. as function of thermal








resistance when operated in pulse and CW conditions. For both pulsed and CW operations the characteristic temperatures remain relatively flat when thermal resistance is less than 80 C/W. As thermal resistance becomes larger than 80 C/W, the associated T, values start to decrease gradually. Additionally, due to the heat effect from the CW operation, the average characteristic temperature is - 15 C lower than that of pulsed operation. The measured dependence of pulsed and CW differential quantum efficiency (dQE) on the heat sink temperature of diode lasers with different thermal resistance Rth= 44 C/W, 57 C/W and 96 C/W are shown in Figure 4.16. In the pulsed operation, the dQE values show a weak function of heat sink temperature and no significant thermal-resistance dependence is observed. However, in the CW operation, differential quantum efficiency of diode laser shows a great dependence on heat sink temperature as well as the thermal resistance. The dQE value of laser with Rlh=96 C/W decreases significantly as the heat sink temperature is beyond 60 C. This pronounced decrease in differential quantum efficiency could be due to the increase of non-radiative Auger recombination [Dutt83] or carrier leakage over heterobarriers [Good75, Dutt8l] for the high Rth diode laser when operated at high temperature CW condition.



In summary, thin p-clad diode lasers with a two-stripe configuration emitting dual wavelengths through the control of the contact reflectivity are successfully demonstrated for








the first time. The fabrication process of this dual wavelength diode laser is much simpler and more reliable than those approaches used before. Based on this experiment results, one can design a thin p-clad diode laser with several emission energy levels inside the single quantum well and select the suitable contact metal to control the emission wavelength to obtain a monolithic multiple wavelength emission diode laser. Additionally, thermal resistance was found to play an important role for diode lasers operating at high temperature in CW condition. Laser performance can be greatly improved by reducing its thermal resistance through the deposition of thick Au heat spreader. However, both the characteristic temperature T. and differential quantum efficiency dQE show very weak dependence on the thermal resistance when diode laser is in pulsed operation.



















Anodic-oxide
p-cladding layer>N Al0.6Ga0.4As InGaAs SQW
n-cladding layer A10.6Ga0.4As

n+-GaAs
substrate

Ge/Au/Ni/Au n-type ohmic contact


Figure 4.1 Schematic diagram of dual wavelength diode laser
structure used in this study. The details of the epilayer thickness and the associated compositions are the
same as those shown in Figure 2.1.


I2
Ni-contact A-contact


I - I MWOO

















































Figure 4.2 (a) The top view and (b) side view of dual
wavelength InGaAs SQW diode laser after soldered and
bonded on the copper block



















































Figure 4.3 Measured I-V characteristics of dual wavelength
emission InGaAs SQW (a) Au contact and (b) Ni contact
diode lasers. The I-V characteristics between both
lasers are also shown in (c).



















































Figure 4.3 -Continued.





















04


0 100 200 300 400
Input current (mA)


500


Figure 4.4 Measured pulsed output power versus input current
characteristics of dual wavelength InGaAs SQW diode
laser.





















967 nm

*
Ii=0 mA I2=65 mA


967 nm
914 n






800 900 1000
Wavelength (nm)







Figure 4.5 The measured lasing wavelength of the dual
wavelength InGaAs SQW diode laser when operated at
different input current conditions where I and 12 are
the same as those shown in Figure 4.1.


















































Figure 4.6 Measured near-field distributions of dual
wavelength emission InGaAs SQW diode laser when
operated at (a) I1(Ni)=l.l Ith' I2(Au)=0, (b) I1(Ni)=0, 12(Au)=1.1 I2th, and (c) I (Ni)=l.l I1th' I2(Au)=l.l I2th.




















































Figure 4.6 -Continued.
















-Im







00


. . . . I . .. . I. ... ... ... I . . ..













* pulse width= 700 Lsec.


0 10 20 30 40 50 60 Au plating time (min.)


Figure 4.7 The evolution of current density and repetition
rate as a function of plating time for Au plating used
in the deposition of thick Au heat spreader.


500 400 ,

4)

300 r 200 A 100








Au heat spreader


Figure 4.8 (a) Schematic diagram of Au heat spreader
patterns. (b) The outlook of finished Au heat spreader
on the top of thin p-clad InGaAs SQW diode lasers,
where Au heat spreader thickness is -5 gm.











250


$4 200


150


*100
0


0

0 0 E- I


100


200


300


400


Input current (mA)


Figure 4.9 Total CW output power versus input current
characteristics for 5-gm stripe thin p-clad InGaAs SQW
diode lasers with heat spreader and without heat spreader, where heat spreader thickness is -7 Jim.


Thin p-clad InGaAs SQW ILD

* w=5 pLm, L=500 gm w/o facet coating T=25 C, p-side up







solid: with HS
dashed: w/o HS
(HS: Heat Spreader)











0.50m GRATING SPECTROMETER


Intensity attenuator



50x object lens and XYZ controller


-Laser diode


Figure 4.10 Schematic diagram of thermal resistance Rth
measurement setup used in the experiments.


















































Figure 4.11 The variations of the Fabry-Perot mode shifts
in 50 .tm stripe width, 500 gm cavity length, InGaAs SQW diode laser measured by using the setup shown in Figure
4.10, where laser is operated at I-l.l Ith and, (a)
THS=21.8 C, 40 % duty cycle, (b) THS=21.8 C, 20 % duty
cycle and (c) THs=24.8 C, 40 % duty cycle. The vertical
central line is used as the referenced line for the
thermal resistance measurements.


















































Figure 4.11 -Continued.



















rJ2


20 40 60 80


100


Input current duty cycle (%)














Figure 4.12 The measured heat sink temperature (THS) rise as
a function of duty cycle for the 50 pm stripe width
thin p-clad InGaAs SQW diode lasers with and without Au
heat spreader. The dashed lines are the extensions of the linear relationship to obtain the ATCW value from
the intersect points at 100 % duty cycle respectively.










250


200 S150 100



50
0


0


0 50 100 150


3.0




2.0




1.0


". . . ' . ' 0.0 200 250 300 350


Input current (mA)










Figure 4.13 Typical CW P-I and I-V characteristics of 50 pm
stripe width, Au contact, thin p-clad InGaAs SQW diode
laser.





















P9


E-4


160



120 80



40



0


Thickness of Au heat spreader (gim)










Figure 4.14 The measured dependence of thermal resistance
on the Au heat spreader thickness for thin p-clad
InGaAs SQW diode lasers.


* I I ' U I U I I I I




* w=50 pm, L=500 pm
p-side up









, .. l lI..l .. l , a..










1 6 0 . , . . . . , . . * . * . * . * . . . .


140 Thin p-clad InGaAs SQW LD


120 e pulse
W-CW

" 1 100 0
0 O

80 "--A-A --A- �

60
6 w=50 gim, L=500 gm


4 0 .. l ... . .. . . . . . I . . . . I . . . . I . . . .
40 50 60 70 80 90 100 110 Rth (C/W)


Figure 4.15
on the lasers


The dependence of characteristic temperature T, thermal resistance Rth of diode lasers, where are pulsed operation or CW operation.

























----- 44

10 23


10 20 30


40 50 60 70 80 90


Heat sink temperature TIs (C)










Figure 4.16 The variations of differential quantum
efficiency (dQE) as a function of heat sink temperature of thin p-clad InGaAs SQW diode lasers with various Rth,
where lasers operated in pulse and CW are compared.


40 30



20


,--N


4-4


Thin p-clad InGaAs SQW LD






a I

R (C/W)=


j




I -


---"-m ' ti * solid: CW
empty: pulse
,I,I,,IN,,,,tIt,,,,II is


I I M M


\0















CHAPTER 5

p -GaAs CAP LAYER THICKNESS EFFECTS







5.1 Introduction



In previous chapters we have described the influence of contact reflectivity on the thin p-clad diode laser performance and several diode laser fabrication advantages of this interesting structure. However, since the p-cladding layer is only 250 nm thick and, especially, the associated refractive index value at the lasing wavelength for the cap layer is very close to that of quantum well, the p -GaAs cap layer thickness could substantially affect the diode laser behavior [Suem75]. In our continuous work on this thin p-clad (250 nm) laser structure, we have found that both wide stripe and narrow stripe lasers with 200 nm p -GaAs cap layer show a long time delay between the application of an excitation current pulse and the onset of stimulated emission [Wu96a]. In addition, this lasing delay time becomes shorter as the input current increases. Besides the lasing time delay characteristics, threshold current of this thick p-cap (200 nm) laser is abnormally high and show strong dependence on








the period of the excitation current on-time. Moreover, as the cap layer thickness is reduced below 170 nm, diode laser performance improves significantly and no lasing delay is observed for the wide stripe lasers. For the narrow stripe 300 ridge-height lasers, stripe width (w) becomes an important factor in determining diode laser performance. Long time lasing delay behaviors are observed for the 6-11m stripewidth lasers and Q-switching lasing performances are obtained for both 3.5-Im stripe-width and 2.5-m stripe-width lasers. However, when stripe width is reduced to w=l.5 gm, neither long time lasing delay nor Q-switching performance is found. In Section II, diode laser structure used in this study is presented. Also included are the theoretical calculation results of optical confinement factor F and modal loss ai of thin p-clad laser structure. The gain-guided device fabrication process, experimental results and discussions are described in Section III. In Section IV, the details of fabricating process for narrow stripe, 300 nm ridge-height lasers and performance characterized are presented.



5.2 Laser Structure and Theoretical Calculations



Figure 5.1 shows the thin p-clad InGaAs SQW diode laser structure used in this study, where the quantum well thickness is 80 A and p -GaAs cap layer thickness t is 200 nm. As mentioned before, the increase of cap layer thickness could substantially affect diode laser performance, in order




Full Text

PAGE 1

THIN P-CLAD InGaAs SINGLE QUANTUM WELL LASERS By CHIH-HUNG WU A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 1996

PAGE 2

TO MY FAMILY

PAGE 3

ACKNOWLEDGMENTS I wish to express my sincere gratitude to Dr. Peter S. Zory, my advisor, for accepting and supporting me for graduate study. His guidance and constant source of encouragement have proved to be essential in the successful completion of this work. I attribute the knowledge I have acquired in the field of quantum well lasers to his academic leadership and expertise. His commitment to hard work and discipline has inspired me through the course of this study. I would like to thank the members of my supervisory committee. Dr. Gijs Bosman, Dr. Arnost Neugroschel, Dr. Ramakant Srivastava and Dr. Robert M. Park, for giving me helpful suggestions and critiques. I am grateful to Dr. Gary Evans of Southern Methodist University for providing the Modeig program, of great use in laser structure modal loss computations. I want to express my greatest appreciation to Dr. Mark A. Emanuel of Lawrence Livermore National Lab. for providing the MOCVD grown InGaAs/GaAs SQW laser material used in this study. I would also like to thank Dr. David P. Bour of Xerox PARC for supplying GalnP/AlGalnP material during the experiments. My appreciation is especially extended to the Institute of Nuclear Energy Research (INER) in Taiwan for supporting and encouraging me to complete my Ph.D. study. iii

PAGE 4

I have to recognize other individuals who have provided invaluable assistance. I wish to thank Mr. James Chamblee and Mr. Tim Vanght of the Microelectronics Laboratory for all the technical assistance they provided over the years. In addition, I thank all my colleagues in the Photonics Research Laboratory: Horng-Jye Luo, Young-Soh Park, Michael Grove, Chi-Lin Young, Craig Largent, Chia-Fu Hsu and Jeong-Seok 0 for their friendship and inspiration. Last, I want to thank my parents for their love and for teaching me how important education is. I am deeply indebted to my mother-in-law for supporting and helping taking care of my son during my studying abroad. I would like to express greatest appreciation to my wife Yu-Fen Shen and my sons PeiShen Wu and Conan S. Wu for their love and understanding. I would also like to thank my sisters and my wife's sisters for their encouragement and support. iv

PAGE 5

TABLE OF CONTENTS ACKNOWLEDGMENTS iii ABSTRACT vii CHAPTERS 1 INTRODUCTION 1 2 CONTACT REFLECTIVITY EFFECTS ON THIN P-CLAD InGaAs SINGLE QUANTUM WELL LASERS 8 2.1 Introduction 8 2.2 Laser Material and Theoretical Calculations 9 2.3 Experimental Results and Discussions 19 3 THIN P-CLAD RIDGE GUIDED LASERS 37 3.1 Introduction 37 3.2 Theoretical Calculations 38 3.3 Device Fabrication 40 3.4 Pulse Current Characteristics 41 3.5 CW Current Characteristics 43 3.6 Ridge Height Dependent Diode Laser Performance 45 4 DUAL WAVELENGTH DIODE LASERS AND THERMAL RESISTANCE IMPROVEMENT 59 4.1 Introduction 59 4.2 Fabrication of Dual Wavelength Diode Lasers 61 4.3 Dual Wavelength Laser Experimental Results and Discussion 62 4.4 Thermal Resistance Improvements 65 5 P^-GaAs CAP LAYER THICKNESS EFFECTS 91 5.1 Introduction 91 5.2 Laser Structure and Theoretical Calculations 92 5.3 Long Time Lasing Delay in Gain-Guided Lasers 94 5.4 Long Time Lasing Delay and Q-Switching V

PAGE 6

Lasing Delay in Narrow Stripe, 300 nm RidgeHeight Lasers 105 6 SURFACE SENSITIVE LASER DIODES 134 6.1 Introduction 134 6.2 Device Structure and Theoretical Calculations 135 6.3 Device Fabrications 142 6.4 Device Characterizations 143 7 SUMMARY and RECOMMENDATION 160 7 . 1 Summary 160 7.2 Recommendation for Future Study 163 REFERENCES 169 BIOGRAPHICAL SKETCH 176 vi

PAGE 7

Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy THIN P-CLAD InGaAs SINGLE QUANTUM WELL LASERS By CHIH-HUNG WU May 1996 Chairman: Peter Zory Major Department: Electrical and Computer Engineering The surface sensitive aspects of InGaAs single quantum well, semiconductor diode lasers with thin p-type cladding layers (thin p-clad lasers) are studied. Experiments using different thickness quantum wells (QW) , contact layers (p"^-cap layers) , contact layer metallurgies and laser geometries are described and modeled. Lasers fabricated with a QW thickness of 10 nm, an AlGaAs p-clad thickness of 250 nm, a GaAs p^-cap layer thickness of 100 nm and a nickel contact show significantly higher threshold currents and lower slope efficiencies than lasers fabricated with gold contacts. For a range of cavity lengths, the lasing wavelength of the nickel contact lasers is about 50 nm shorter than the gold contact lasers. This phenomenon is exploited to demonstrate side by side lasers on vii

PAGE 8

the same chip with operating wavelengths of 960 and 910 nm. In a 5 flm-stripe, low-ridge configuration, the gold contact lasers operate continuously with single spatial mode output power comparable to those of conventional, high-ridge, thick p-clad lasers. When the QW thickness is reduced to 8 nm and the GaAs p+cap layer thickness is increased from 100 nm to 200 nm, new types of effects are observed. For example, when lasers with 50 micron stripe widths and gold contacts are operated, they show microsecond-long lasing delays. This long delay is attributed to the time it takes for the active region to heat to the point where net mode gain exceeds mirror loss. Net mode gain increase is due to an increase in mode overlap with the QW material gain as well as a decrease in mode overlap with the lossy gold contact layer. When the p'^-cap layer is decreased below 170 nm, laser performance is significantly improved and no lasing delay obtained. By combining 100 nm and 200 nm p*-cap layer structures into one laser and removing the gold layer from the 200 nm section, laser output power at fixed current becomes dependent on the type of material placed on the 200 nm section. Experiments using these "surface sensitive" diode lasers are described and their possible use in sensor applications discussed. viii

PAGE 9

CHAPTER 1 INTRODUCTION Semiconductor material growth technology continues to improve and very high quality epitaxial, multi-layered structures are now available. The improved materials are being used in the fabrication of various types of semiconductor devices including semiconductor diode lasers. The features of small physical dimension, high efficiency, high speed and low cost have made semiconductor diode lasers important and widely-used devices in optical fiber communication and optical memory systems. In most diode laser applications, it is usually desirable that the devices have low threshold current density (J^^) and high differential quantum efficiency (Tl^) . To fulfill these two requirements, diode laser structures must be designed to have both optical field and carrier confinement. Normally, the active region of diode lasers is grown and confined in a pair of cladding layers which have higher energy bandgaps and smaller refractive index than those of the active region. Additionally, the thickness of the 1

PAGE 10

2 cladding layer is usually greater than 1.0 |iin to reduce the optical loss from the contact layer and substrate. For continuous wave (cw) or rapidly pulsed operation, the removal of heat generated in the active region is crucial to the reliability of diode lasers. To obtain highly reliable cw operation and reduce the temperature sensitivity of threshold current, the thermal impedance of the diode laser structure must be as small as possible. Theoretical work showed that the thermal resistance of diode lasers can be reduced by decreasing the thickness of the cladding layers [Joyc75] . Experimental study of double heterostructure (DH) lasers with -0.12 |lm active layer thickness showed that J^^ increases and decreases dramatically when the p-type AlGaAs cladding layer thickness is less than 0.8 (Xm [Case75]. Similar results were also theoretically obtained in [Butl75], where the active region gain at threshold (G^^) increased significantly as the p-AlGaAs cladding layer thickness was reduced to less than ~1.0 |lm due to the increase of optical loss . Despite the increase of G^j, caused by the optical loss, several reports have been published using this "thin cladding layer" diode laser structure. Since the cladding layer thickness is thin, the separation distance between the active region and p-contact layer or substrate is reduced. As a result, the interaction between the tail of lasing modes and either contact layer or corrugated-substrate could be very

PAGE 11

3 strong to provide grating coupled output [Zory75] or distributed feedback (DFB) action [Scif75] . In recent years, several publications have been issued relating the utilization of thin p-cladding layer diode laser structures with the fabrication of surface emission [Maco87, Mott89] and edge emission [Shan88] DFB diode lasers. Additionally, in order to reduce the optical loss of diode lasers with thin pcladding layers, shiny Au-contacts were found to be essential [Luo90] . In previous studies of thin p-clad diode lasers, the device structures were mostly concentrated on DH diode lasers. The effects of decreasing p-cladding layer thickness on the performance of quantum well (QW) lasers are relatively unknown. In order to study these effects, single quantum well (SQW) InGaAs lasers with thin p-cladding layers (250 nm) were used in this study. The diode laser performance is found to be greatly related to the reflectivity of the contact metal. As mentioned before, there could be strong interaction between the contact metal and the tail of the lasing mode when the p-cladding layer thickness is thin enough. This interaction not only changes the modal loss [Wu94b] and lasing threshold conditions but also shifts the lasing wavelength of the QW lasers [Wu94a] . In addition to the study of contact reflectivity effects, the device fabrication technique of this thin p-clad diode structure is also explored .

PAGE 12

4 Diode lasers with single spatial mode operation are useful for increasing the coupling efficiency in optical fiber communication applications. Conventionally, "ridgeguided" diode lasers with thick p-cladding layer (>1.0 ^.m) are fabricated to achieve this single spatial mode operation output. By using such a thick p-cladding layer diode laser structure, a deep etching step has to be performed to remove certain amounts of the outside stripe material to obtain a sufficient refractive index step between the stripe and outside stripe region for maintaining the ridge-guided property. This removal process can be obtained by either wet chemical etching or using more sophisticated RIE (reactive ion etching) techniques. Because the ridge height is critical for diode lasers to operate in the index-guided regime, a very thin "etch-stop" layer is usually grown inside the diode laser structure to make the device fabrication process somewhat less complicated. On the contrary, by using the thin p-clad laser structure, one does not need to do any etching and the fabrication process of index-guided diode lasers is greatly simplified. Single spatial mode lasing to high power levels comparable to that of the conventional thick p-clad diode lasers was obtained [Wu95] . Multi-wavelength emission diode lasers are very attractive for the application of wavelength division multiplexed communication systems. Several approaches have been reported for fabricating monolithic multi-wavelength emission diode lasers, such as changing active layer

PAGE 13

5 composition [Sak:a82, Boua82], grating period variations [Dutt86, Aiki76], stripe width related modal loss control [Toku86] and material desorption technique [Eple90] . The associated techniques for most of these approaches are either low yield and non-reliable or complicated and time consuming. Previously, we have mentioned that contact reflectivity has a great effect on the thin p-clad SQW diode lasers [Wu94b] . One of the important characteristics is that, by selecting shiny or less shiny contact metal, one can control the lasing wavelength of diode laser on either the first or the second energy level of a single quantum well active layer. Based on this concept, the process of fabricating dual wavelength emitting diode lasers should be much simpler and more reliable than the other approaches as stated above. Diode lasers with epi-side up bonding have less diffraction noise and better lasing beam quality than diode lasers with epi-side down bonding. In the epi-side up bonding configuration, however, thermal resistance reduction becomes an important problem. Since the heat sink is far away (-100 |lm) from the active region, heat generated inside the active region flowing through the cap layer becomes an effective heat dissipation path for diode laser operation. Theoretical study [Joyc75] has shown that the thermal resistance of diode lasers can be reduced effectively by depositing a thick Au layer as the heat spreader on the top laser. Based on this concept, a thick Au plating technique was developed and

PAGE 14

6 applied to the devices fabricated to show the reduction of thermal resistance and performance improvements. In designing thin p-clad InGaAs lasers, the p^-GaAs cap layer thickness has to be determined with care in order to avoid the possible increase of modal loss from the cap layer. This is because the refractive index value of the cap layer is very close to that of the quantum well region at the lasing wavelength. In this case, due to the short separation between the quantum well and the p'^-GaAs cap layer, a thick cap layer allows the lasing mode to penetrate through the thin p-cladding layer and couple into the p-cap layer forming a twin-guide laser structure [Suem75] . As a result, more modal loss will be generated and poor diode laser performance obtained. Additionally, due to the smaller optical confinement of lasing mode caused by the thick (200 nm) p^-cap layer, microsecond-long lasing delays are observed on 50-)lm stripe width lasers [Wu96a] . Although thick p^-cap layer, thin p-clad lasers appear not to be useful, the insertion of a thick p'^-cap section into a thin p'"-cap layer laser may have practical applications. The combination of one or more 100 nm p''-cap layer sections as the electron pumped sections and a 200 nm p'^-cap layer section as the surface sensitive section in one laser structure makes a "surface sensitive" diode laser [Wu96b] which may prove useful in sensor applications. This dissertation is organized as follows: Chapter 2 shows the details of contact reflectivity effects on the thin

PAGE 15

7 p-clad diode lasers. Both the theoretical calculations of QW modal gain, modal loss and experimental results are described. Chapter 3 presents the theoretical calculations and the experimental results of ridge-guided thin p-clad diode lasers. The comparison between thin p-clad and thick pclad diode lasers are also included. Chapter 4 states the fabrication process and the experimental results of dual wavelength diode lasers made from a thin p-clad laser structure. In addition, the details of a thick Au plating heat spreader technique developed to show diode laser thermal resistance improvements are described. Chapter 5 outlines the influence of p*-GaAs cap layer thickness on the thin p-clad diode laser performance. Chapter 6 describes the details of theoretical and experimental work on surface sensitive laser diodes (SSLD) based on the thin p-clad laser structure with both 100 nm and 200 nm p'^-cap layers. Finally, the conclusion and recommendations for future study are presented in Chapter 7 .

PAGE 16

CHAPTER 2 CONTACT REFLECTIVITY EFFECTS ON THIN P-CLAD InGaAs SINGLE QUANTUM WELL LASERS 2 . 1 Introduction In this chapter we show how the performance of shiny contact, InGaAs single quantum well (SQW) lasers is changed when the contact metal is changed from shiny gold to less shiny nickel. In addition to the expected difference in threshold current and slope efficiency, operating wavelength differences of more than 50 nm are observed [Wu94] for cavity lengths between 250 and 700 microns. At shorter (gold) and longer (nickel) cavity lengths, large shifts in operating wavelength are observed. In Section 2.2, the laser material used is described. Additionally, theoretical calculations of the quantum well laser modal gain and the modal loss induced by the different contact metal are outlined. In Section 2.3, the experimental results of thin p-clad diode lasers with shiny gold (Au) and less shiny nickel (Ni) fabricated are shown and compared with the theoretical calculations. Also presented are the changes of diode laser performance under various annealing time and constant annealing temperature. 8

PAGE 17

Diode lasers with and without facet coating are life-time tested to evaluate the reliability of these thin p-clad lasers , 2 . 2 T.aser Matprial and Theoretical Calculations Figure 2.1 shows the diode laser structure used in this study which consists of a 250 nm n*-GaAs buffer layer grown on a n^-GaAs substrate, a 70 nm Al^Ga^.^As graded layer (x=0 . 050.6), a 1400 nm n-type Alo_gGaQ,,As cladding layer, a 200 nm graded Al^Ga^.^As (undoped) barrier layer (y=0 . 3-0 . 6) , an active layer of 10 nm In^Ga^.^As (z~0.15) undoped strained quantum well centered in a pair of 7 nm GaAs bounding layers, a 200 nm graded Al^Ga^.yAs (undoped) barrier layer (y=0 . 6-0 . 3) , a 250 nm p-Alg gGag ,,As cladding layer, a 25 nm Al^Ga^.^^As graded layer (x=0 . 6-0 . 05) and a 100 nm p*-GaAs contact layer. As can be seen the p-cladding layer is only 250 nm and is much thinner than those of conventional laser devices (>1.0 ^m) . Since the total thickness of the epitaxial layers above the quantum well in this structure is only -600 nm, the type of metallization used on the p*-contact layer is expected to be important in determining the mode loss coefficient [Luo90]. To understand the effect of p-contact metal on the diode laser loss coefficient (a^) , the values for the laser structure shown in Figure 2.1 have to be calculated.

PAGE 18

10 To do the computations of a^, the lasing wavelength is determined first. In computing the lasing wavelength of the laser structure, the energy gap shift due to the presence of biaxial strain induced by the lattice mismatched-material InGaAs/GaAs has to be calculated. The strained effects can be devided into two components: hydrostatic component and shear component [Van89] . Under the hydrostatic strain effect both the conduction band edge and the valence band edge will move up by AEj, and AE^ eV as referenced from the corresponding bulk material band edge . The energy gap change AEg due to hydrostatic strain effect is obtained from the following equations [Van8 9] : AEg = AE^ AE„ (2.2.1) AQ AE = a * (2.2.2) Q AE, = a, * ^ (2.2.3) AQ a. a„ C,, TT = £xx + 8yy + 8„ = 2 * °-) * (1 -^) (2.2.4) where a^, a,: hydrostatic deformation potential for the conduction band and valence band AQ : fractional volume change due to strain effect Q ^xx' £yy' ^22= diagonal components of the strain tensors

PAGE 19

11 a^, a„: lattice parameter for the barrier (GaAs) and quantum well (In^Ga^.^As) ^11' ^12elastic stiffness coefficient Since the exact value of either the hydrostatic deformation potential or the elastic stiffness coefficients of the ternary material In^Ga^.^As has not yet been determined, in this study we assume that both parameters can be expressed as the combination of those of InAs and GaAs. In other words, a^, a^, and for the In^^Ga^.^As can be expressed as: a^ = (1 X) * ai{GaAs) + x * a^dnAs) i=c, v (2.2.5) C^^ = (1 X) * C^j(GaAs) + X * C^^dnAs) j = l, 2 (2.2.6) Under the shear component of strain effects the valence band edge will become degenerated and split into heavy-hole band edge and light-hole band edge. The separation between heavyhole band edge and light-hole band edge is derived from the equations as shown below: S = 2 * be^ * a — ) (2.2.7) A a — a C 5e, = b * -5 2. * (1 + 2 * (2.2.8) b = (1 X) * b(GaAs) + x * b(InAs) (2.2.9) A = (1 X) * A(GaAs) + x * A(InAs) (2.2.10) where S: bandedge split energy between heavy hole and light hole

PAGE 20

12 ^e^: heavy hole splitting energy b: uniaxial deformation potential A: split orbit energy The lattice constant of In^Ga^.^As and GaAs can be expressed as a^ (x) = (5 . 6533+0 . 405*x) A and aQ=5.6533 A, respectively [Corz93] . As referenced from data shown in Van de walle [Van89] , one can calculate the total energy gap shift caused by the hydrostatic and shear components of strain by substituting the related parameters into equation (2.2.1) to equation (2.2.10) . Figure 2.2 shows the schematic diagram of bandgap lineup for Ing ^jGag gsAs/GaAs quantum well structure, where strain-induced bandedge split energy S between the heavy hole and the light hole, and the heavy hole energy bandgap Eg(HH) are computed as S=63 . 9 meV and (HH) =1253 . 3 meV. A conduction band offset ratio of 0.55 is also assumed [Cole93] to compute the conduction band energy barrier, heavy hole valence band energy barrier and light hole valence band energy barrier as 93.9 meV, 7 6.8 meV and 12.9 meV, respectively. The energy state distribution of the InGaAs/GaAs quantum well structure is calculated by solving the Schrodinger equation [BastBB] , in which a square well structure with finite barrier height is assumed. The effective masses used for the electron, heavy hole and light hole in the InGaAs quantum well are 0.0599 m„, 0.4395 m^ and 0.0847 m^ respectively [Pan88], where m^ is the electron rest mass. For the conduction band, the energy states inside the

PAGE 21

13 100 A Ino_;^5Gao_85As quantum well are calculated as £^^=22 . 9 meV and £,,2=62.9 meV (referenced from the conduction bandedge) . For the valence band, there are four heavy hole energy states: £^^,^=6 meV, £^^2=23.8 meV, £^^3=52.2 meV and £^(,^=86. 8 meV, and one light hole energy state: E„ii=82.5 meV (referenced from the heavy hole band edge) inside the quantum well. According to the results of the computations, a lasing wavelength of -960 nm is obtained for this InGaAs SQW diode laser and used for the following computations. The modal loss induced by both shiny Au and less shiny Ni contacts are calculated from Modeig program [Deme89] . The refractive indices of Au and Ni at this wavelength are interpolated from [Weas87] as 0.091-16.03 and 2.75-14.87, respectively. The corresponding optical constants of each layer of the diode laser structure are obtained [Taka78], [Adac85] . Only the fundamental mode is considered from the calculated results in which the highest effective refractive index value of the diode structure is found. Figure 2.3 shows the dependence of the calculated mode loss on the thickness of the p-cladding layer when Au and Ni p-contact metals are used. From this plot, we can see as the p-clad layer thickness is reduced below 400 nm, the optical mode loss for the Ni contact lasers increases rapidly whereas that for the Au contact remains relatively flat. In other words, Au contacts are shiny compared to Ni contacts. At 250 nm pcladding layer thickness, the calculated mode loss of diode

PAGE 22

14 laser with Au and Ni metal contact is ai(Au)=3 cm"^ and ai(Ni)=70 cm'\ respectively. Since the threshold mode gain G^^ is proportional to the modal loss of diode lasers, the larger modal loss induced by the Ni contact indicates that Ni contact lasers should have higher G^^^ than those of Au contact lasers. The magnitude of the increase in G^^, required for lasing when using Ni rather than Au is given by [Chin88] Gt. = Tg,, = a, +yln^ (2.2.11) Li K where F is the optical confinement factor, g^^ is the threshold gain in the quantum well, L is the cavity length of diode laser and R is the mode reflectivity at the facets. As stated earlier, for L in the range from 250 to 700 |J.m, the Ni and Au thin p-clad devices differ in operating wavelength by more than 50 nm. In order to understand this effect, it is necessary to calculate the spectral gain function g(hv) [Chin88] as well as threshold spectral gain g^^ for these two different optical mode loss devices. Since the energy state distributions in the QW structure have been determined, the carrier density required for obtaining a given quasi-Fermi level can be calculated. In the computations, a parabolic subband model is used for conduction band, heavy hole valence band and light hole valence band. For the conduction band, the electron concentration is denoted [Cor293] as

PAGE 23

15 N = 2 J p'°(E EJ * f,^ * dE ^' (2.2.12) = " ' X ma + exp[-(E,„ E,j I * T]} where p^^CE-E^): density of states in the quantum well f^: Fermi-Dirac distribution for electrons in the conduction band m^: effective mass of electron in the conduction band Kg: Boltzmann's constant E^„: quantized energy level of electron in the conduction band Ej^: quasi-Fermi level in the conduction band : quantum well thickness h : reduced Planck constant For the valence band, the hole concentration is the summation of light hole concentration and heavy hole concentration, and can be expressed as [Corz93] P=P^+P^ and P. = m*k*T E— E, ' ' Z{ln[l + exp{-^ ^)]} kg * T (2.2.13) P, = m^ * kg * T „ . . .^vn, ~ I { In [1 + exp ( i ) ] } kg * T (2.2.14)

PAGE 24

16 1 where nij, , : effective mass of heavy hole and light hole; n^, n^: quantum state number of heavy hole and light hole in the valence band; E^„^ , E^„^ : quantum state energy for heavy hole and light hole in the valence band; E^^: quasi-Fermi level in the valence band. By using equations (2.2.12), (2.2.13), (2.2.14) and the quantum state energy inside the quantum well, one can get the relationship between the quasi-Fermi level E^^, E^^ and the corresponding carrier density in the conduction band and valence band. In semiconductor the optical gain caused by the photoninduced transitions of electrons from conduction band to valence band is defined as the fractional increase in photons per unit length 1 dO O dz g = -— = W,^, W,^, (2.2.15) and can be written as[Corz93] g(hv) = (— ) M, p,,,(f, fj (2.2.16) hv EgCmo'^n n where F is the photon flux (unit; s'^cm'^) ; W^..^^, W^^^,: total transition rate from conduction band to valence band and from valence band to conduction band, respectively; , n : group refraction index and refraction index of the crystal; |m^|^ : transition matrix element; r^^^ : reduced density of state; f..

PAGE 25

17 f^: Fermi-Dirac distribution for electrons in the conduction band and valence band. In the calculations, we assume the difference between and n can be neglected. Considering the TE polarized transition, k-selection rule and ignoring the forbidden transition, the transition matrix element can be expressed as iM^frvn, = S^r^njMp < F.„Jf,„^ > = SllX (2.2.17) where S" = S'^l = (1 + cos^ 6, ) for e-hh transition (2.2.18) Srf,„ = Sr^,„ = -(5 3 cos^ e„ ) for e-lh transition (2.2.19) cos' e = — ^-^ ^ (2.2.20) hV 1 1 Prort I = — 't — i=h for e-hh transition; (2.2.21) ' 2Kh^ — ^ = — + — i=i for e-lh transition. (2.2.22) AEj.^^ : energy difference between the electron energy state and the conduction bandedge E^,. AE^^^ : energy difference between the heavy hole energy state and the heavy hole bandedge or between the light hole energy state and the light hole bandedge. For the In^Ga^.^As quantum well, |Mp term in equation (2.2.17) can be written as [Corz93]

PAGE 26

t |M|2 = i^^(28.8 6.6 * x) (2.2.23) By ignoring intraband relaxation [Garb93], the total optical gain is then calculated from equation (2.2.16) to equation (2.2.23) as the summation of optical gain due to ehh transitions and e-lh transitions. Figure 2.4 shows the dependence of spectra gain on photon energy for the diode laser structure shown in Figure 2.1, where the injected carrier density is assumed at N=2.5xl0^^ cm"^ and 5.5x10^' cm"^ . It is evident in Figure 2.4 that the photon energies at peak gain for the two carrier densities are considerably different. At 2.5x10^® cm'^ injection carrier density, the peak gain occurs at an energy corresponding to the band edge transition energy from n=l conduction band electrons to n=l heavy holes, E^j^~1286 meV. When the injection carrier density increases to S.lxlO-'-^ cm"^, the peak gain is the same at Ej^^ and (the band edge transition energy from n=2 conduction band electrons to n=2 heavy holes, ~1364 meV) . Above 5.1x10^^ cm'^, the gain peak occurs at E22. The g^^ values shown in Figure 2.4 are calculated from equation (2.2.11) using the values from Figure 2.3, a confinement factor r=0.02, mode reflectivity R=0.31 [Lee91] and cavity length L=380 \Lm. It is clear that the increase in g^^, required by the extra mode loss due to the Ni contact can cause a large shift in lasing wavelength .

PAGE 27

19 1 2.3. Experimental Re^s ults and nisrn.q?;ion In order to fabricate wide-stripe lasers, 50 |lm stripes on 500 |im centers are defined on the p^-GaAs contact layer using standard photolithography techniques. Mesas and current blocking oxide layer are formed using a pulsed anodic oxidation technique [Grov94]. Usually the oxidation time is ~6 minutes (with anodic oxidation current density J~100 mA/cm% repetition rate= 50 Hz and pulse width= 700 Usee) to remove the p^-GaAs cap layer and part of the p-cladding layer to reduce the diode current spreading effect. After the oxidation process, wafers are cleaned and covered with new photoresist to protect the epilayer surface for the following wafer thinning process. This wafer surface protection step is important for making good adhesion shiny contact for the diode lasers. The wafers are thinned to about 100 |im, a metallurgy of Ge(20 nm)/Au(40 nm)/Ni(20 nm)/Au(15 nm) is evaporated sequentially on the substrate side and annealed at 430 C for 5 minutes to obtain the n-type ohmic contact. A non-alloyed Au(80 nm)/Ni(50 nm)/Au(50 nm) or Ni(80 nm)/Au(50 nm) is evaporated as the shiny or less shiny p-contact respectively. Then, another non-alloyed Ni(25 nm)/Au{50 nm) is deposited on the subtract side to improve the metal contact conductivity. Diode lasers with 500 ^Jti wide and cavity lengths ranging from 120 ^m to 1200 \Lm are cleaved, soldered on copper blocks by using either indium soldering or

PAGE 28

. 20 sliver epoxy (1 part Ag epoxy + 1 part thinner and quad at 150 C for ~5 minutes) and characterized. The measured emission spectrum of Au and Ni lasers with L=380 flm is shown in Figure 2.5. The difference in operating wavelength is ~52 nm, in good agreement with the predicted difference of -55 nm shown in Figure 2.4. The corresponding pulsed output power versus input current characteristics are shown in Figure 2.6. Threshold current and total slope efficiency for these two lasers are I^^(Au)=60 mA, 'n3(Au)=0.76 mW/mA and I^^,(Ni)=240 mA, n3(Ni)=0.41 mW/mA respectively. In addition, from measured inverse slope efficiency versus cavity length plots, the internal mode loss for Au-contact lasers and Ni-contact lasers are a^(Au)=8 cm"^ and ai(Ni)=51 cm"^ respectively. These are to be compared with the calculated values from Figure 2.3, (Au) =3 cm'^ and a^(Ni)=70 cm'\ The difference of 5 cm"^ in (Au) and 19 cm"^ in a^(Ni) can be explained if the complex refractive index of the Au, n(Au), and the complex refractive index of the Ni, n(Ni), used in our experiments are somewhat different from the values interpolated from [Weas87] . For example, if n(Au) is changed from 0.091-16.03 to 0.16-14.08, ai(Au) changes from 3 cm'^ to measured value of 8 cm'\ On the other hand, if n(Ni) is changed from 2.75-14.87 to 2.30-15.25, ai(Ni) changes from 70 cm"^ to the measured value of 51 cm"^ . We believe that

PAGE 29

variations in both n(Au) and n(Ni) of these magnitudes are possible since it is known that refractive index values can be strongly dependent on the details of the deposition process . In the previous section, it is mentioned that transition E^i and have the same peak spectral gain at a carrier density of 5.1x10^® cm"\ This implies that lasers requiring this carrier density to lase should operate simultaneously at two wavelengths about 50 nm apart [Zory86] . It is clear from equation (2.2.11) that it should be possible, in principle, to change cavity length L to achieve this condition for both Au and Ni contact lasers (L>380 |J.m for Ni lasers and L<380 \Lm for Au lasers) . In order to check this prediction, the emission wavelengths of both laser types are measured as a function of L. As shown in Figure 2.7, the emission wavelength jumps occur for the Ni lasers at L~900 |Jin and at L~200 |J.m for the Au lasers. Similar wavelength jumps have been observed for thick p-clad 7.5 nm InGaAs SQW lasers when their stripe widths are reduced from 50 to 15 microns [Shie89] or thick p-clad 15 nm InGaAs SQW lasers with 25 |i.m stripe widths when their active region temperature is increased due to current [Beer91a] . Using the above L values in equation (2.2.11) along with the measured values of discussed above and R=0.31, we find that G^f,= 66,5 cm"^ for 200 Hm long Au contact lasers and G^^ =64 cm'^ for 900 \lxa. long Ni contact lasers, in close agreement as expected. The blue shift of the spectral peaks accompanied with the decrease of

PAGE 30

cavity length shown by the dashed lines in Figure 2.7 is mainly due to band-filling modified by bandgap renormalization [Chin88] . So far we have shown that the reflectivity of p-contact metal is important in determining the performance of thin pclad diode lasers, we are interested in understanding the effects of annealing p-contact metal on the thin p-clad diode laser. To check this point, thin p-clad diode lasers with shiny Au contact are annealed at 320 C for different annealing time: 30, 60, 120, 180 and 300 seconds. The variations of threshold current and total slope efficiency as a function of annealing time are shown in Figure 2.8. As can be seen from this plot, threshold current increases rapidly as the annealing time increases upto 120 seconds and then becomes saturated when diodes are annealed further longer. Additionally, the total slope efficiency decreases gradually from 0.8 mW/mA(0 second) to -0.5 mW/mA(120 seconds) and remains constant as annealing time increases further. The deterioration of diode laser performance can be explained by the decrease of contact reflectivity caused by the reactions of Au and GaAs at 320 C. This interaction will make the Au contact become less shiny. As the annealing time increases further, the interactions may saturate and cause the contact reflectivity keep constant. Consequently, even at low temperature (320 C) , the annealing effect can decrease the contact reflectivity and diode laser performance. To have

PAGE 31

23 good thin p-clad diode laser behavior, the shiny Au contact should not be annealed. To further evaluate shiny contact thin p-clad diode lasers, samples without and with facet coatings are CW lifetime tested at 20 C with output power 50 mW/facet (w/o coating) and 100 mW (w/ coating) , respectively, where both samples are packaged with p-side up configuration. The coated sample is prepared by depositing a thickness of X,/4n AljOj on the emitting facet as the anti-reflection (AR) coating layer and AI2O3 (X./4n) /Al (100 nm) /Al^Oj (A./4n) on the reflective facet as the high reflector (HR) where n is the refractive index of the active region at the lasing wavelength. Figure 2.9 and Figure 2 . 10 show the output power P versus input current I characteristics and the corresponding variations of threshold current and total slope efficiency of diode laser without facet coating after t hours lifetime test. Similar results of the coated sample are shown in Figure 2.11 and Figure 2.12. For the uncoated sample, threshold current increases from the initial value of 42 mA to -5 6 mA and the total slope efficiency almost keeps constant at 0.65 mW/mA for the first 500 hours lifetime test and gradually decreases to ~0.57 mW/mA after 1057 hours lifetime test. For the coated sample, threshold current increases quicker than that of uncoated sample from the initial value of 40 mA to ~65 mA. In addition, the slope efficiency decreases faster than that of the uncoated sample from 0.62 mW/mA to -0.41 mW/mA after 1050

PAGE 32

24 hours lifetime test. The possible reason for the big difference of I^j, and between both diode lasers could be due to the cavity length difference where the uncoated sample is 600 |lm long while that of the coated sample is only 450 |im. Since diode laser with longer cavity length is believed to have better lifetime performance than that of diode laser with shorter cavity length, especially for the high output power condition. From the results of lifetime test, we have successfully demonstrated that thin p-clad diode lasers can "live" for long time CW operation. Further applications from this thin p-clad laser structure could be possible as the fabrication process is improved and optimized. In summary, we have demonstrated that the reflectivity of p-contact has a significant effect on the performance of thin p-clad diode lasers. Decreasing the reflectivity of the p-contact metal increases optical mode loss, which increases threshold current, decreases slope efficiency and shifts the emission wavelength. The more than 50 nm wavelength difference between the thin p-clad diode lasers with different p-contact metallurgy can be explained quantitatively by superimposing the threshold gain required for lasing in each case with the corresponding spectral gain curve calculated using standard QW laser theory.

PAGE 33

25 p"*"-GaAs contact layer: 100 nm p-AlzGa^.^As ( z=0.6 0.05 ): 25 nm P-A1q 5Gao 4AS cladding layer: 250 mn Al^Ga^.^As ( x=0.3 0.6 ): 200 nm GaAs: 7 nm InyGaj.yAs SQW ( y~0.15) : 10 nm GaAs: 7 nm Al^Gai.xAs ( x=0.6 0.3 ): 200 nm n-Alg gGag 4AS cladding layer: 1400 nm n-Al2Gai.2As ( z=0.05-0.6 ): 25 nm n-GaAs substrate Figure 2.1 Thin p-clad InGaAs single quantum well diode structure used in the contact reflectivity effect study .

PAGE 34

26 1 BULK MATERIAL: -I Eg=1424 mev I Eg(bulk)=1193.6 mev STRAIN EFFECT: 82.6 mev oi. t» i Shear Deformation f Hydrostatic I°0.15P«0^5A« i I>eformation ^ 12.7 mev 3 5.6 mev ^ Bulk Inn i«^fana.;As 1 iJeiormation ^ Hh | 63.9 mev f 4 ir^4 BAND LINEUP UNDER STRAIN EFFECT: ( Assume conduction band energy offset ratio = 0.55 ) 1 93.9 mev 1424 meV 1253.3 meV i 76.8 mev Hh i GaAs Lh t 12.9 mev J 0.15^^.85^* GaAs Figure 2.2 Schematic diagram of bandgap lineup between the strained InGaAs quantum well and GaAs barrier layers of the diode laser structure shown in Figure 2.1.

PAGE 35

27 250 200 s 150 S 100 i2 50 0 T — I — m — I — I — I — r I ' ' ' ' I ' ' Ni metal contact I Au metal contact \ This work -i / ' L_L. ' 0 100 200 300 400 500 p-clad thickness (nm) Figure 2.3 The calculated relationship between the optical mode loss and p-clad thickness for the InGaAs single quantum well (SQW) laser structure shown in Figure 2.1,

PAGE 36

gure 2.4 The dependence of spectral gain on photon energy for a 100 A Ino.15Gao.85As/GaAs structure at two carrier densities. The two dashed lines are the active layer threshold gains for thin (0.25 |lm) p-clad InGaAs SQW lasers with Ni or Au contact metal .

PAGE 37

29 1 1 1 1 1 1 1 1 1 1 1 • 1 • • 1 • . 1 1 1 1 1 909 nm l\ Ni-contact, L=380 nm 961 nm Au-contact, L=380 )im 1=1-2 Ith 1 1 1 1 1 1 1 I 1 1 1 1 1 1 1 1 1 1 II i 1 1 1 860 900 940 980 Wavelength (nm) Figure 2.5 Measured lasing wavelengths of thin (0.25 Hin) pclad InGaAs SQW diode lasers with different p contact metals .

PAGE 38

30 120 I — ' — ' — ' — I — ' — ' — ' — I — ' — ' — ' — I — ' — ' — ' — I — ' — ' — ' 0 200 400 600 800 1000 Input current (mA) Figure 2.6 Measured pulsed output power versus input current characteristics of thin p-clad InGaAs SQW diode lasers with Au contact or Ni contact.

PAGE 39

31 1.5 ^ 1.4 § 1-35 CA g 1.3 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 i 1 [ III 1 1 1 1 1 1 1 1 1 1 1 : InGaAs SQW LD * w=50 )iin o : Au-contact +. : Ni-contact "o'"" 8 + o o o _ o 1 1 1 1 1 1 1 1 ill 1 1 1 1 1 1 1 1 1 1 1 1 1 1 III ill 1 1 1 1 M 1 1 0 10 20 30 40 50 60 70 80 Inverse of cavity length 1/L (cm"^) Figure 2.7 Spectral position of peak lasing energy as a function of reciprocal cavity length of thin p-clad InGaAs SQW diode lasers using Au or Ni as the p-type contact. (Each mark represents an individual laser.)

PAGE 40

32 200 2 150 100 I IS 0 60 120 180 240 300 360 Time (second) Figure 2.8 The variations of threshold current (1^^,) and total slope efficiency (r\^) of thin p-clad diode lasers as a function of p-contact metal annealing time, where the annealing temperature is set at constant 320 C.

PAGE 41

33 0 50 100 150 200 250 Input current (mA) Figure 2.9 Measured CW output power versus input current characteristics of Au-contact thin p-clad InGaAs SQW diode laser after t=0, 529 and 1057 hrs lifetime test, where the output power is set constant CW 50 mW/facet during the lifetime test.

PAGE 42

34 200 150 2 'o u ' ' ' I 100 0 I I I — I — I — I I I — I I I I I I ' * L=:600 [im, w=50 w/o facet coating I I I J I I 1 I I I I L ' 0.8 0.6 a; 0.4 0.2 5 (A 0 -5 0 200 400 600 800 1000 1200 H Time (hr) Figure 2.10 The variations of threshold current and total slope efficiency of Au-contact, thin p-clad InGaAs SQW diode laser as a function of lifetime test period when operated with constant output power CW 50 mW/facet.

PAGE 43

35 g U o o 60 40 20 0 1 1 1 1 1 1 1 1 1 f— T 1 1 1 1 1 — 1 — 1 1 — Thin p-clad InGaAs SQWLD t= ohr / 192hr _ / / OA i nr w/ AR,HR / facet coating / / / lOSOhr w=50 |im / / / L=450 |im / / / 1 1 1 1 1 1 0 50 100 150 Input current (mA) 200 Figure 2.11 Measured CW output power versus input current characteristics of Au-contact thin p-clad InGaAs SQW diode laser after t=0, 192, 517 and 1050 hr. lifetime test, where sample is with AR and HR facet coating and output power is set constant CW 100 mW during the lifetime test.

PAGE 44

36 O 200 -| — I — I — I — I — I — I — I — I — I I I — I — I — I — r 0 3 50 Jsf^^^-^^ * L=450 ^iin, w=50 |im w/facet coating J I I I I I I I I I I I L J0.8 S 33 0.4 0 200 400 600 800 1000 1200 Time (hr) 0.2 ^ Figure 2.12 The variations of threshold current and total slope efficiency of Au-contact thin p-clad InGaAs SQW diode laser as a function of CW 100 mW constant output power lifetime test period, where sample is with AR and HR facet coating.

PAGE 45

1 CHAPTER 3 THIN P-CLAD RIDGE-GUIDED DIODE LASERS 3 . 1 Introduction In chapter 2 we have demonstrated the important effects of p-metal contact reflectivity on the thin p-clad diode laser performance. In this chapter, a new and simple method of fabricating ridge-guided diode laser from a thin p-clad structure is proposed and demonstrated. The major advantage of using a thin p-clad diode laser structure instead of the conventional thick p-clad structure for fabricating ridgeguided diode lasers is that only a small part of outside stripe material is required to remove for providing a strong waveguiding. This will make the device fabrication process become simple and economical. Moreover, we have found that these low-ridge, thin p-clad lasers can operate in a single lateral mode with cw performance characteristics similar to those reported for InGaAs SQW lasers fabricated from conventional epitaxial material in a high-ridge configuration [Fisc87, Bour90, Take90] . In other words, thin p-clad diode laser with low ridge height structure can provide strong i 37

PAGE 46

waveguiding for diode lasers operating with low threshold current (I^^) and high slope efficiency (Tl^) . This is totally different from those of thick p-clad diode laser structure in which the threshold current of an InGaAs single quantum well { SQW ) diode laser is very high ( > 100 mA ) when fabricated in either a low-ridge [Shie89] or oxide-defined [Beer91b], narrow-stripe configuration. In Section 3.2, the main design principle responsible for low threshold, single lateral mode lasing is discussed. In Section 3.3, the procedure for fabricating ridged-guided, thin p-clad lasers using pulsed anodic oxidation [Grov94] is outlined. The results of pulsed characterization measurements on these lasers are presented in Section 3.4 and compared with measured values reported for thick p-clad lasers [Shie89, Beer91b] . In Section 3.5, the cw performance characteristics of 5 micron wide, low-ridge, thin p-clad lasers are presented and compared with those of reported thick p-clad, high-ridge waveguide lasers [Fisc87, Bour90, Take90] . In Section 3.6, the dependence of diode laser performance on the ridge height is presented and comparisons are made between thin and thick p-clad ridge-guided lasers with various ridge heights. 3 . 2 Theoretical Calculations The thin p-clad diode laser structure used in this study is the same as that stated in previous chapter ( as shown in

PAGE 47

39 Figure 2.1 ). Since the thickness of the p-clad layer is only 250 nm, the behavior of ridge waveguide lasers fabricated from this material should be quite different from that of ridgeguide lasers with the same ridge height fabricated from conventional material with thick p-clad layers ( > 1000 nm ) . The parameter normally used in designing ridgeguide lasers is An, the difference in effective refractive index ( lateral index step ) between the ridge region and the outside-ridge region [Agra84] . Consequently, it is of interest to calculate An as a function of ridge height ( the amount of epitaxial material removed in defining the ridge ) for thick and thin p-clad configurations. In the calculations, it is assumed that the p"*"-GaAs contact layer, the p-AlxGai-xAs graded-layer and part of the p-clad layer of the outside-ridge region are removed and replaced with a native oxide, the whole structure being covered with Au . A lasing wavelength of 960 nm, an oxide refractive index of 1.8 and a Au refractive index of 0.0 91-16.03 are also assumed in the computations. The optical constants of each layer of the laser structure are the same as those used in our previous work on thin p-clad InGaAs SQW lasers [Wu94b] . Figure 3.1 shows the calculated An of thin ( 250 nm ) p-clad diode lasers as a function of ridge height when various oxide values are used as the current blocking layer. It is interesting to note that relatively large values of An are obtained with the removal of small amounts of material. For example, a An value of greater than 3x10"^ is

PAGE 48

40 obtained when 130 nm of epitaxial material is removed and replaced by 130 nm of native oxide. If a conventional thick p-clad layer is used instead of a thin p-clad layer ( 1300 nm vs 250 nm ) , a An value of only ~lxlO~6 is obtained for the same removal/replacement process. In order to achieve a An of 3x10"^ in the 1300 nm thick material, an order of magnitude more material ( ~1200 nm ) must be removed from the laser structure creating a "high-ridge" device. From a An point of view therefore, a low-ridge, thin p-clad device is equivalent to a high-ridge, thick p-clad device. If An is the key parameter which determines device performance, then lowridge, thin p-clad lasers should not show the high threshold effect as ridge width is narrowed [Shie89, Beer91]. In addition, they should be capable of operating in a single spatial mode to high cw power levels if the ridge width is sufficiently narrow [Fisc87, Bour90, Take90] . These two speculations are addressed in Sections 3.4 and 3.5 respectively . 3 . 3 Device Fabrications Thin p-clad lasers with 130 nm ridge height and four different ridge widths ( w=5 ^m, 10 |lm, 25 [Lm and 50 |im ) are fabricated using a pulsed anodic oxidation technique [Grov94] . In this technique, photoresist is used to define the ridge widths on the laser wafer which is then used as one

PAGE 49

41 of the electrodes in an electrolyte composed of 40 parts ethylene glycol, 20 parts water (H2O) and one part phosphoric acid (H3PO4) . Running 700 |isec wide current pulses through the electrolyte at 50 Hz repetition rate converted 130 nm of epitaxial material into 100 nm of native oxide in about 3.5 minutes (anodic oxidation current density -100 mA/cm^) . The laser material is then thinned to about 100 |Xm and a Ge/Au/Ni/Au contact metallurgy deposited onto the n-side by electron beam evaporation. After a standard high temperature anneal, a Au/Ni/Au contact metallurgy is electron beam evaporated onto the p-side and 450 \im wide bars cleaved from the material. The cleaved bars are then scribed into 500 |lm wide chips and soldered, substrate-side-down, onto indiumcoated copper blocks. It is important to note that the p-side contact was not annealed. If it were annealed, the p-contact would not be "shiny" and laser thresholds would be very high as shown in chapter 2 . 3 . 4 Pulsed Current Characteristics As discussed in Section 3.2, low ridges in thin p-clad material produce relatively big An values ( An ~ 3x10"^ when using 250 nm-thick p-clad material with 130 nm ridges ) . To see if this An is sufficient to avoid the high threshold effect at narrow ridge widths reported in references 1 and 2, the pulsed threshold current ( I^^ ) and laser emission energy of the 130 nm-ridge, thin p-clad lasers are measured and

PAGE 50

42 compared with the data reported for thick p-clad devices [Shie89, Beer91]. In Figure 3.2, the thick p-clad lasers are (a), 200 nm-ridge lasers [Shie89] with 400 |Xm cavity length and (b) , oxide-defined stripe devices [BeerSlb] with 510 |im cavity length. As can be seen, I^^ for both laser types increases as the stripe width is narrowed, the increase being very large for the oxide-defined stripe lasers. Moreover, there is a lasing energy jump [Wu94] associated with the increase of threshold current for both thick p-clad laser types. In contrast, as the stripe width of the 130 nm-ridge, thin p-clad laser is decreased from 50 \lm to 5 )lm, I^^ decreases from 50 mA to 10 mA and the lasing energy is essentially constant . It appears that the lateral index step An ~ 3x10"-^ is sufficiently big in the thin p-clad, 130 nm ridgeguides to create a low loss lateral waveguide. If An is too small, the lateral dimension of the lasing mode spreads out into the unpumped region causing a large mode loss. This forces the required carrier density for lasing to be very large leading to carrier-induced antiguidance and additional mode loss [Shie89, Beer91b] . Since a low loss lateral waveguide appears to exist in the thin p-clad, 130 nm ridgeguide lasers, their lateral farfields should also be quite different from those observed in references [Shie89, Beer91b] . To check this point, we have to measure the lateral far-field distributions of thin p-clad diode lasers by using the setup shown in Figure 3.3. In the measurements, diode lasers are kept at room temperature and

PAGE 51

43 by rotating the detector to obtain the far-field distributions. As shown in Figure 3.4, a single-lobe lateral far-field pattern is obtained for lasers with 5 |Xin, 10 |Xm and 25 ridge widths at I~l.l I^^. The single-lobe far-field pattern of the 5 |im ridge-width diode laser is about 5 times narrower than the single-lobe pattern reported in [Shie89] and totally different from the double-lobe pattern reported in [Beer91b] . As the drive current increases to -1.5 I^^, both 5 |lm and 10 |lm lasers remain single-lobed with no significant change in pattern width. The lateral far field pattern for the 25 |J.m laser however, changes from single-lobed to doublelobed as the current increases to 1.5 I Since similar th • behavior is observed in the far field during cw operation of the 25 |lm device, the change can probably be attributed to spatial-hole burning [Garr87] rather than thermal waveguiding [Bour90] . At 1=1.5 I^^, the 50 |Xm stripe-width laser still remains double-lobed. 3 . 5 CW Current Characteristics As discussed in Section 3.4, a An value of ~ 3xl0~3 is sufficient to avoid the high threshold effect at narrow ridge widths. It is interesting to ask if this An is also sufficient to allow single lateral mode operation to high cw power levels when the ridge width is 5 ^m. To check this

PAGE 52

44 point, the cw lateral far-field patterns of several 5 |i.m wide, 130 nm-ridge, thin p-clad diode lasers are measured at four drive currents: 40 mA, 80 mA, 120 mA and 160 mA ( see Figure 3.5 ). Typically one observes that the full width at half maximum (FWHM) of the patterns ( ~ 10° > increases by less than 1° as the current is increased from 40 mA to 120 mA. Between 120 mA and 160 mA, the FWHM increases by about 3°. In order to understand the cause of this broadening, the lateral far field patterns are also measured using short pulse excitation. In this case, no broadening is observed over the entire current range leading us to believe that the slight cw broadening observed up to 120 mA is due to a narrowing of the near field caused by thermal index guiding [Bour90] . While the relatively large increase of the broadening above 120 mA could be due to the onset of higher mode oscillation, the lack of any significant beam steering tends to rule against this possibility [Guth94] . In any event, the addition of a heat spreading layer should be very effective in reducing this broadening effect since the p-cladding layer is very thin. The measured FWHM value of the transverse far-field pattern is ~ 42°, independent of current The measured cw output power versus drive current ( P-I ) characteristic is plotted in the inset of Figure 3.5. As shown, the slope efficiency stays constant up to about 120 mA ( ~ 70 mW of cw total output power ) and then begins to drop at higher currents due to heating. Threshold current and total differential quantum efficiency are 10 mA and 49 %,

PAGE 53

45 respectively. The threshold current of 10 mA is comparable to that reported for conventional SQW inGaAs high-ridge lasers [Fisc87, Bour90, Take90] although the differential quantum efficiency is ~ 10 % smaller. In order to see if this result could be improved on, thin p-clad diode lasers are fabricated with a 180 nm-ridge height and, according to Figure 3.1, a An of close to 5xl0~3. In this case a cw total differential quantum efficiency of 61 % (cavity length L=600 \Jua and stripe width w=5 |lm) and threshold current of 9 mA are measured. Figure 3.6 shows the measured CW P-I characteristics of the 5-|im stripe, 180 nm ridge lasers when different cavity length L are considered. As can be seen from this plot, the 180 nm ridge lasers show comparable P-I performance with those reported for the conventional ridge-guided lasers and indicate strong waveguiding in these thin p-clad, narrow stripe lasers. 3 . 6 Ridge Height Dependent Diode La ser Performance It is mentioned in Section 3.2, that high ridge structure has to be performed for thick p-clad narrow stripe ridge-guided diode lasers to obtain sufficient index step in operating at low threshold, single spatial mode regime. In addition, CW measurements of thin p-clad ridge-guided lasers have indicated that diode laser performance greatly depends on the ridge height of the device. Since we are interested in

PAGE 54

46 understanding the dependence of narrow stripe laser behaviors on the ridge height, thin p-clad (250 nm) lasers as well as thick p-clad {1300 nm) lasers with 5 |im stripe width and various ridge heights are fabricated and characterized. Both types of samples are prepared by using similar process steps as described in Section 3.3. Thick p-clad samples are anodic oxidized to get right heights: -230 nm, 530±30 nm, 1000±50 nm, 1150±50 nm, 1220±30 nm and 1350±50 nm respectively. On the other hand, thin p-clad samples are oxidized to have ridge heights: 130±5 nm, 180±5 nm and 260±5 nm. Figure 3,7 shows the measured variations of pulsed P-I characteristics of thick p-clad (1300 nm) 5-\lm stripe diode lasers at different ridge height. Figure 3.8 shows the variations of threshold current and slope efficiency as a function of ridge height. It is noticed from Figure 3.7, that P-I behavior of diode laser is not linear until the associated ridge height exceeds 1000 nm for this thick p-clad (1300 nm) laser structure. This is due to the occurrence of higher order modes which generate when the ridge waveguiding is not strong enough to support only single mode operation [ThomSO] . The non-linear P-I performance also explains the weak waveguiding inside the low-ridge thick p-clad diode laser and agrees well with the theoretical predictions of Section 3.2. Even for the 1000±50 nm ridge height laser, the non-linear P-I characteristics still occur as the injection current increases up to -100 mA. From Figure 3.8, it is observed that threshold current (1^^) of thick p-clad ridge

PAGE 55

47 guided laser decreases as the ridge height increases to -1000 nm and becomes saturated when the ridge height increases further in the measured range. In this study, the minimum 1^^ value obtained for the 5-|im stripe-width, 500 |Xm long, thick p-clad diode laser is ~10 mA. As shown in Figure 3.8, for the samples without significant non-linear P-I performance, the slope efficiency increases and reaches the maximum value at ridge height ~1150 nm, whereas T]^ shows a decrease as ridge height increases from -1150 nm to -1350 nm. For the thin p-clad samples, however, non-linear P-I characteristics are not observed in the measured range as shown in Figure 3.9. This is due to the strong waveguiding from the low ridge structure. In addition, from the corresponding I^^,, Tl^ vs ridge height plot as shown in Figure 3.10, threshold current shows a small value decrease from I,. tn -11 mA to I^j, -10 mA as the ridge height increases from -130 nm to 260 nm. This is dramatically different from those of thick p-clad lasers, where -5 time decrease of 1^^ occurs as the ridge height increases from low ridge (-230 nm) to high ridge (-1000 nm) . Additionally, the value of thin p-clad ridge-guided laser increases as the ridge height increases and reaches a maximum point then decreases as the ridge height becomes higher. The decrease of slope efficiency is due to the contraction of lateral mode caused by the strong index-guiding which could make the mode profile narrower than

PAGE 56

48 the gain profile resulting in a lower slope efficiency [Agra84] . In this thin p-clad laser structure, the 180 nm ridge-height laser shows the best performance of slope efficiency . In summary, we have demonstrated that the lateral refractive index step generated in low-ridge, thin p-clad diode lasers is sufficient to provide low loss lateral waveguiding. The threshold current of 130 nm-ridge diode lasers decreases from 50 mA to 10 mA when the ridge width is decreased from 50 |im to 5 |J.m whereas the lasing energy remains essentially constant. In addition, the 5 \im stripe devices were shown to be capable of stable, single lateral mode cw lasing with less than 10 % broadening up to total output power levels of about 70 mW. On the contrary, for the thick p-clad (1300 nm) diode laser, deep etching for obtaining ridge height larger than 1000 nm is required to provide the sufficient waveguiding effects for narrow stripe lasers operating with stable single mode output . By the optimization of diode laser structure, single spatial mode operation with linear P-I profiles to much higher cw powers should be possible. This feature with the combinations of the flexibility of fabricating different diffraction grating types (nickel, gold metal grating or dielectric material grating) in thin p-cladding material ( no regrowth required ) could lead to the development of higher performance gaincoupled DFB lasers [Luo92] .

PAGE 57

49^ anodic oxide p-clad layer quantiim well n-clad layer — metal \ stripe , region Urn — p+-GaAs P-AIq eGao 4AS ^ 0.25 ^im AlxGa^.xAs: x=0.3-0.6 GaAs GaAs Al^Ga^.xAs :x=0.6-0.3 n-Alo sGag 4AS (a) 4 ^ 3.0 I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I 1 1 Thin p-clad InGaAs SQW LD I I I ' ' I ' ' ' I ' I ' ' I ' I ' ' ' 2 0.0 S 125 175 225 275 325 375 Ridge height (nm) (b) Figure 3.1 (a) Schematic diagram of narrow stripe thin p-clad InGaAs SQW ridge-guided laser, (b) Calculated lateral index step An of laser structure in (a) as a function of ridge-height for various current blocking oxide values. It is assumed that the ridge guide structure is covered with Au .

PAGE 58

50 600 _ 500 ^ 400 5 300 5 200 100 0 145 ^ 140 135 > 130 E^ 125 E120 I I I I I I I I I I I I I I I I I I I I I I I I I I . 1 \ • -fa-B (a), thick p-clad — A — (b), thick p-clad (c), thin p-clad A— Jl-"--^— B • (a), thick p-clad — A(b), thick p-clad (c), thin p-clad 1 •• ^ A 0 10 20 30 40 50 Stripe width (|im) 60 Figure 3.2 The dependence of threshold current and laser emission energy of InGaAs SQW diode lasers on the stripe width for (a) thick p-clad, 200 nm-ridge, L=400 |im [Shie87], (b) thick pclad, oxide-defined stripe, L=510 [Beer91], and (c) thin p-clad, 130 nm-ridge, L=450 [Lm [this work] .

PAGE 59

51 Diode laser DC power supplier ILX LDC-3472 Detector and slit Diode laser Opto-Logic diode laser module & TE coller Angle rotator Temperature controller ILX LDT-5910 Detector bias source & optical current output module Digital meter Amplifier DC power supply I Signal amplifier Figure 3.3 Schematic diagram of far-field measurement setup used in this study.

PAGE 60

52 Stripe width w=50 \im w=25 )j.in w=10 |im w=5 |im * L=450 \im 10% duty cycle 1=1.1 Ith .25C 10 -5 0 5 10 15 Lateral Angle (degree) Figure 3.4 The measured lateral far-field intensity distribution of 130 nm-ridge, thin p-clad InGaAs SQW diode lasers as a function of ridge width.

PAGE 61

53 10 6 « 6 cn ^ 4 0 T — I — I — I — I — |— 1 — I — I — I — I — I — I — I — I — I — I — I — I— r Thin p-clad SQW InGaAs LD 100 a 40 mA iii|iiii|iiii|iiii|iiii|ini|iiii|iiii /ilmiliiiilnnlmilmilinilmi 0 50 100 150 200 Input current (mA) . J I I ' I -20 0 20 Lateral angle (degree) Figure 3.5 CW lateral far-field patterns of a 5 wide, 130-nm ridge, thin p-clad InGaAs SQW laser with uncoated facets at a heat sink temperature of 25 °C . The corresponding total output power versus drive current characteristic is shown in the inset plot.

PAGE 62

54 Input current (mA) Figure 3.6 The variations of CW P-I characteristics of 5-|im stripe, thin p-clad InGaAs SQW diode lasers with 180 nm ridge height and different cavity length.

PAGE 63

55 ^ 100 52 80 g 60 U O 40 ft 20 ® 0 ' ' I ' ' I ' ' ' I ' ' I ' ' I ' ' I ' Thick p-clad InGaAs SQW LD ridge height = 1150±50 run 1000±50 nm 530±30 nm 230±20 nm 0 20 40 60 80 100 120 140 Input current (mA) Figure 3.7 The variations of pulsed P-I characteristics of 5-JXm stripe thick p-clad InGaAs SQW diode lasers with different ridge height.

PAGE 64

56 0 400 800 1200 1600 Ridge height (nm) Figure 3.8 The measured dependence of threshold current 1^^ and slope efficiency of 5-p.m stripe thick p-clad InGaAs SQW diode lasers on the ridge height.

PAGE 65

57 Thinp-cladlnGaAsSQWLD ridge height I I I I I I ' ' ' ' ' I ' 130±5 nm 180±5 nm 260±5 nm w=5 nm, L=500 p.m Q r I ^ ^ I I I I I I I I I I I I I I I 1 I 0 30 60 90 120 150 Input current (mA) Figure 3.9 The variations of P-I characteristics as a function of ridge height for thin p-clad 5-)Xm stripe diode lasers .

PAGE 66

58 25 -I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — r Thin p-clad InGaAs SQW LD rS 20 15 en 9 _i I I I I I I i_ 0.6 0.4 5*H s 0.2 09 100 150 200 250 300 Ridge height (nm) Figure 3.10 The dependence of 1^^, and T]^ on the ridge height of 5-|lm stripe thin p-clad InGaAs SQW diode lasers.

PAGE 67

CHAPTER 4 DUAL WAVELENGTH DIODE LASERS AND THERMAL RESISTANCE IMPROVEMENT 4 . 1 Introduction Diode laser emitting simultaneously at multi-wavelength is very attractive for the applications to wavelength division multiplexed communication systems. Several reports have been issued about the fabrication of a monolithic multiwavelength emission diode laser. The first approach is to use different active layer composition. In this approach, either selectively etching [Saka82] or etch-and-regrowth technique [Boua82] has to be accurately performed in order to obtain reliable diode laser quality. The second approach is to utilize complex grating technique where different lasing wavelengths are obtained by either changing the stripe width to vary the index of refraction [Dutt86] or different periods of DFB grating [Aiki76] . The third approach is to control the modal loss through the change of stripe width of diode laser to obtain lasing on either the first or the second quantized energy levels of a single quantum well active layer [Toku86] . 59

PAGE 68

60 In this method, high temperature and time-consumed material disordering process is used and make this method become hard to control. In addition, more complicated method of using laser-induced desorption to selectively decrease the single quantum well thickness has been developed to fabricate multiple wavelength emission diode lasers [Eple90] . Previously, we have demonstrated that contact reflectivity has a great effect on the thin p-clad SQW diode lasers [Wu94b] . One of the important characteristics is that, by selecting shiny or less shiny contact metal, one can control the modal loss of diode laser and make the laser lase on either the first or the second energy level of a single quantum well active layer. Based on this concept, the process of fabricating the dual wavelength emitting diode lasers should be much simpler and more reliable than the other approaches as stated above. In this chapter, a twin stripe diode laser structure with dual wavelength lasing is proposed and demonstrated. Recently, the influence of heat spreader on the laser performance has been reported on AllnP red light emission material system and shows significant CW P-I performance improvements [Unge93] . However, the dependence of thermal resistance on the heat spreader thickness and the associated effects on the InGaAs/GaAs material system are seldom reported. In this chapter, a thick Au plating technique is developed and applied to the devices fabricated to show the

PAGE 69

61 reduction of thermal resistance and performance improvements of InGaAs/GaAs diode lasers. In Section 4.2 the dual wavelength laser structure and fabrication process are described. The experiment results and discussion of the dual wavelength diode lasers are stated in Section 4.3. The main principle of thermal resistance measurements, details of thick Au plating technique as the heat spreader and the performance improvements of thin p-clad InGaAs/GaAs diode lasers with Au heat spreader are presented in Section 4.4. 4.2 Fabrication of Dual Wave length Diode Lasers Figure 4 . 1 shows the twin stripe diode laser structure used in this study. Basically, the diode laser structure is the same as shown in previous chapters except a twin stripe geometry is used instead of a single stripe. The stripe width of Au contact and Ni contact laser is designed as 50 |Im and 20 |lm respectively in order to have close threshold current for both laser types. The separation distance between these two stripe edge is 50 |J.m with an isolation slot of 20 |Xm used to make these two diode laser operate independently. Wafer is first cleaned in boiling TCA, ACE and methanol. Then, photoresist AZ-1350 is spreaded and baked to define a 20 lim window. Since the isolation slot is important to make the twin stripe diode lasers operate independently, the etching depth of the isolation slot has to be larger than 0.6

PAGE 70

62 )im for this laser structure. The etching depth can be obtained by chemical etching, anodic oxidation or the combination of both methods. In this experiment, the isolation slot is first etched by the chemical solution of 1 part NH4OH + 1 part H^Oj + 50 parts H^O at room temperature. After the etching step, wafer is cleaned and spreaded with new photoresist to define the two-stripe patterns (w=20 |im and w=50 |J.m) . Pulsed anodic oxidation is then performed to remove parts of the p*-GaAs cap layer and grow an oxide layer to cover the whole area except the twin stripes. Wafer is then covered with new photoresist and thinned to about 100 |im. A metallurgy of Ge(20 nm)/Au(40 nm)/Ni(30 nm)/Au(50 nm) is evaporated on the substrate side and annealed at 430 C for 5 minutes to get n-type ohmic contact. Photoresist of AZ-1375 is spreaded to define p-contact patterns for the lift-off process. At first, by using lift-off technique, Ni(60 nm)/Au(50 nm) are deposited sequentially to form the Ni-metal contact for the 20 |lm stripe lasers. Following this step, photoreisst AZ1375 is used again to define Au contact pattern. A metallurgy of Au(80 nm)/Ni(50 nm)/Au(30 nm) is sequentially evaporated and lift-off to get the Au-metal contact for the 50 |J.m stripe lasers. Finally, diode lasers are cleaved, soldered on the copper blocks and characterized. ^ -3 Dua l Wc^vel ength T , a se rs Exp eriment ResTilts and Di.snns.^inn Figure 4.2 shows the top view and the side view of the twin stripe diode laser after being soldered and bonded on

PAGE 71

63 the copper block. The measured current (I) versus voltage (V) characteristics of the twin stripe diode lasers are shown in Figure 4.3(a) and 4.3(b). As can be seen, both devices show the same cutin voltage V^~1.4 Volts. At 1=50 mA, the corresponding voltage value of the Ni contact laser is little bit higher than that of Au contact laser (2.1 Volts versus 2.0 Volts), which could be attributed to the fact that Au contact device has larger contact surface than that of Ni contact laser. Figure 4.3(c) shows the I-V characteristic measured between Au contact electrode and Ni contact electrode. Even at high voltage V=12 volts, the current-flow between these two electrodes is very small (< 20 \LA) and indicates good electric isolation between these two diode lasers . The measured pulsed P versus I characteristics of the twin stripe diode lasers are plotted in Figure 4.4. At room temperature the slope efficiency T)^ and threshold current I^j, of the diode laser with cavity length L=350 \Lm are T|g(Au)=0.55 mW/mA per facet, I^^(Au)=45 mA and 1]^ (Ni) =0 . 18 mW/mA per facet, I^^(Ni)=105 mA respectively. At current 1=150 mA, more than 8 mW/facet output power from the Ni contact laser and more than 50 mW/facet from the Au contact laser can be obtained. The PI characteristics are comparable to those data reported before [Aiki76, Boua82, Saka82, Dutt86] . The lasing wavelengths of the twin stripe diode laser at injection current 1-1.5 I^^, are shown in Figure 4.5. At 1^=160 mA and

PAGE 72

64 l2=0 mA, only the Ni contact laser can lase with the emission wavelength X^=91^ nm. While at li=0 mA and l2=65 rtiA only the Au contact laser can lase with the lasing wavelength X2=967 nm. When 1^=160 mA and l2=65 mA, both laser can lase with X,i=914 nm and A.2=967 nm, respectively. Therefore a monolithic diode laser with dual wavelength emission capability has been successfully demonstrated by using contact reflectivity effect on the thin p-clad diode laser for the first time. This method of achieving dual wavelength operation is much simpler and more reliable than the other approaches as stated in Section 4.1. To further check the lasing characteristics of the twin stripe diode laser, near field emission patterns are observed under three different injection current conditions. Figure 4.6(a) shows the near field pattern of the Ni contact laser at Ii=120 mA and l2=0 mA. Figure 4.6(b) is the near field pattern of the Au contact laser at li=0 mA and l2=60 mA. Under these independent operating conditions, no significant spontaneous emission radiated from the unpumped stripe region is observed and indicates a negligible leakage current between these two devices which is consistent with the results shown in Figure 4.3 (c) . Figure 4.6(c) shows the near field pattern as 1^=120 mA and l2=60 mA. No significant light intensity increase is found in each individual diode laser as compared to the results of Figure 4, 6 (a) and Figure 4.6(b).

PAGE 73

65 This also demonstrates that these two devices can be operated independently. 4 . 4 Thermal Resistance Improvements Low thermal resistance are essential for diode lasers to increase the CW operation temperature and maximize CW output power [Jone75]. Thick Au heat spreader is found to play an important role for improving thermal resistance of diode laser when soldered with epi-side up configuration [Joyc75]. To have thick Au deposited on the top of p-contact of diode lasers, Au plating is the quick and economic method. Two important points are found to be crucial for obtaining thick Au when using Au plating technique: (1) the thickness of photoresist (PR) for the patterns defined and (2) the conditions used for Au plating. In this study, ~5 |ijn thick photoresist is obtained by spreading PR AZ1375 three times at 4000 RPM for 30 seconds each time. In addition, an optimal pulsed Au plating condition (pulse width 700 ^.sec.) is obtained to have thick Au heat spreader. The evolution of the optimal Au plating condition as a function of time is shown in Figure 4.7. It is found that the initial condition is important for obtaining good quality thick Au film, where Au plating current density and repetition rate must be adjusted appropriately to avoid too high values used. Moreover, Au thickness obtained does not only depends on the plating time but also depends on the plating conditions and Au density

PAGE 74

66 inside the plating solution. Normally, the plating conditions shown in Figure 4.7 will give one 4-5 Hm Au thickness. By following the Au plating process as shown in Figure 4.7, diode lasers are Au plated, cleaved, soldered and characterized . Figure 4.8 (a) is the schematic diagram of the Au heat spreader designed for the experiments where the separation between each individual Au heat spreader is important for the diode laser cleaving process. The finished top view of the heat spreader on the p-contact metal is shown in Figure 4.8(b). Figure 4.9 shows the CW P-I characteristics of 5-|J.m stripe, thin p-clad diode lasers with and without thick Au (~7 Hm) heat spreader (HS) . It is interesting to note that the linear P-I performance of sample without heat spreader is less than 100 mW output power, however the sample with heat spreader shows a linear P-I performance up to 150 mW. In addition, the maximum output power F^^^ significantly improves from P„^^ (without HS) -150 mW to P^^Jwith HS) -220 mW. These -50% improvements in both linear CW P-I performance and P * max value apparently signify the important heat dissipation effects provided by the thick Au heat spreader. To further understand the influence of heat spreader on the diode laser performance, we have to measure diode laser thermal resistance. Figure 4.10 shows the setup used for the thermal resistance measurements. The main principle for this setup is to utilize the temperature dependence of refractive index n within the laser waveguide. The advantage of this

PAGE 75

67 measurement setup is that no preliminary calibration measurement is required and only a null measurement of the exact wavelength of a single Fabry-Perot mode is made for the R^^ determination [Paol75] . In this study, diode lasers are driven at a constant current near threshold with 2 |lsec constant pulse width. The temperature of laser resonator is controlled by varying the input current duty cycle through the change of its repetition rate. The variations of laser temperature can cause lasing wavelength shift. The wavelength shift due to the temperature change from T+AT to T can be expressed as AK?=Ke (T+AT) (T) = (dA./dT) AT (4.4.1) where AA^p is the Fabry-Perot mode shift due to the variation of refractive index, X^p(1+AT) , X^g{T) are the modes corresponding to the refractive index at temperature T+AT and T respectively. From the laser oscillation condition, the wavelength X, can be written as ^2nL/q (4.4.2) where q is an integer and L is laser cavity length. Therefore, A^^p can be expressed as AXf,p={d(2nL/q) /dT}AT=(>./n) (dn/dT)AT (4.4.3) From equation (4.4.3), if we assume the refractive index change of In^Ga^.^As (x~0.15) quantum well is the same as that

PAGE 76

68 of GaAs as (dn/dT) =4x10"" [Mar 64], then we can get A\,p/AT~1.1 A/K. Since the width of a Fabry-Perot mode near threshold can be less than 0.2 A/K, the wavelength of a single longitudinal mode is a sensitive indicator of the temperature of the laser resonator [Paol75]. Figure 4.11(a), (b) and (c) show the wavelength shifts of the Fabry-Perot modes of 50 |ijn stripe width diode laser at different duty cycles and heat sink temperature. The wavelength shift of the selected Fabry-Perot mode is compensated by the drop of heat sink temperature AT„g. Since the heat sink temperature change is a linear function of duty cycle [Paol75], one can obtain the temperature change AT^,, by measuring AT^j only to certain value (<100 %) of duty cycle and get the temperature rise AT^,,^ from the plot of AT^g versus duty cycle. Figure 4.12 is the result of heat sink temperature change as a function of operation duty cycle for the 50 |im stripe lasers with and without thick Au heat spreader. In this plot, AT^^^ is obtained by elongating the linear relationship to 100% duty cycle and get the corresponding intercept A^^^^ value i.e. AT^^. After the temperature rise AT^^ is determined, thermal resistance of the laser can be calculated from AT^„=R^,,P^^ [Paol75], where P^ is the average supplied electrical power given by the current I times the diode voltage V. Figure 4.13 is a typical plot of I-V and CW P-I characteristics for the thin p-clad SQW 50 ^Im stripe diode laser. Since the selected operating current for R,^ measurement is close to the threshold current, the radiative output power is only a small fraction of the input

PAGE 77

69 power, one can get P^^ value directly from Figure 4.13 and calculate R^^. From the measurements, the thermal resistance for the sample with heat spreader and without heat spreader as shown in Figure 4.12 are R^„(w/HS)=70 C/W and R^^^^w/o HS)=102 C/W respectively. Obviously, samples with heat spreader can have less thermal resistance than those without heat spreader. In order to check the dependence of thermal resistance on the thickness of heat spreader, wide stripe (50 |Xm) thin p-clad lasers are deposited with various thickness of Au heat spreader and thermal resistance measured as stated above. Figure 4.14 shows the measured variations of thermal resistance of diode laser as a function of Au plating thickness. It is noted that thermal resistance decreases significantly for the first 7~8 |lm Au heat spreader thickness and becomes gradually saturated when Au thickness is thicker than 8 |im. Since diode laser performance is temperature dependent, threshold current I^j, of diode laser can be normally expressed as I^h (T) -exp (T/T^) , where is the characteristic temperature of diode laser. We are interested in understanding if there is any relationship between diode laser thermal resistance and its temperature related performance. To check this point, diode lasers with different thermal resistance are measured at various temperature (from 20 C to 80 C) in both pulsed and CW operation. Figure 4.15 shows the variations of laser characteristic temperature T„ as function of thermal

PAGE 78

70 resistance when operated in pulse and CW conditions. For both pulsed and CW operations the characteristic temperatures remain relatively flat when thermal resistance is less than 80 C/W. As thermal resistance becomes larger than 80 C/W, the associated values start to decrease gradually. Additionally, due to the heat effect from the CW operation, the average characteristic temperature is ~ 15 C lower than that of pulsed operation. The measured dependence of pulsed and CW differential quantum efficiency (dQE) on the heat sink temperature of diode lasers with different thermal resistance R^h= 44 C/W, 57 C/W and 96 C/W are shown in Figure 4.16. In the pulsed operation, the dQE values show a weak function of heat sink temperature and no significant thermal-resistance dependence is observed. However, in the CW operation, differential quantum efficiency of diode laser shows a great dependence on heat sink temperature as well as the thermal resistance. The dQE value of laser with Rth=96 C/W decreases significantly as the heat sink temperature is beyond 60 C. This pronounced decrease in differential quantum efficiency could be due to the increase of non-radiative Auger recombination [Dutt83] or carrier leakage over heterobarriers [Good75, DuttSl] for the high R^,, diode laser when operated at high temperature CW condition. In summary, thin p-clad diode lasers with a two-stripe configuration emitting dual wavelengths through the control of the contact reflectivity are successfully demonstrated for

PAGE 79

71 the first time. The fabrication process of this dual wavelength diode laser is much simpler and more reliable than those approaches used before. Based on this experiment results, one can design a thin p-clad diode laser with several emission energy levels inside the single quantum well and select the suitable contact metal to control the emission wavelength to obtain a monolithic multiple wavelength emission diode laser. Additionally, thermal resistance was found to play an important role for diode lasers operating at high temperature in CW condition. Laser performance can be greatly improved by reducing its thermal resistance through the deposition of thick Au heat spreader. However, both the characteristic temperature T^, and differential quantum efficiency dQE show very weak dependence on the thermal resistance when diode laser is in pulsed operation.

PAGE 80

72 Anodic-oxide p-cladding layer ^0.6^30.4^8 InGaAs SQW n-cladding layer ^.6^30.4^3 n'*'-GaAs substrate Ge/Au/Ni/Au : n-type ohmic contact Au-contact ymy/////m'7//my/////////y////y/////^^^^ Figure 4.1 Schematic diagram of dual wavelength diode laser structure used in this study. The details of the epilayer thickness and the associated compositions are the same as those shown in Figure 2.1,

PAGE 81

73 (a) Figure 4.2 (a) The top view and (b) side view of dual wavelength InGaAs SQW diode laser after soldered and bonded on the copper block

PAGE 82

74 r 1 1 mum miimi mmH — ^ 1 iO tj H 1 i ^ 1 1 V 1 i Immi mam III Ul (a) (b) Figure 4.3 Measured I-V characteristics of dual wavelength emission InGaAs SQW (a) Au contact and (b) Ni contact diode lasers. The I-V characteristics between both lasers are also shown in (c) .

PAGE 83

75 Figure 4.3 -Continued.

PAGE 84

76 Figure 4.4 Measured pulsed output power versus input current characteristics of dual wavelength InGaAs SQW diode laser .

PAGE 85

77 :3 T 1 T ' I I I I 914 run Il=160inA l2=0 mA ^ijai4iiitlti,#*i»lWyil)i'<^tii|li|i|i)>|)^ 967 nm 1 1=0 mA l2=65 mA i 967 nm 914 nm 1 1=160 mA 1 2=65 mA 800 900 1000 Wavelength (nm) Figure 4,5 The measured lasing wavelength of the dual wavelength InGaAs SQW diode laser when operated at different input current conditions where Ij^ and are the same as those shown in Figure 4.1.

PAGE 86

78 (b) Figure 4.6 Measured near-field distributions of dual wavelength emission InGaAs SQW diode laser when operated at (a) (Ni) =1 . 1 I^,,,, l2(Au)=0, (b) Ii(Ni)=0, l2(Au)=l.l and (c) Ii(Ni)=l.l I,,^, l2(Au)=l.l I^,,.

PAGE 87

79 Figure 4 , 6 -Continued .

PAGE 88

3 1 80 g 60 50 30 10 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 — > ! ^ — 1 \ * pulse width= 700 }isec. 1 1 ' 1 ' 1 ' 1 ' ' ' 1 500 400 U 300 o •iH 200 ft 100 0 10 20 30 40 50 60 Au plating time (min.) Figure 4.7 The evolution of current density and repetition rate as a function of plating time for Au plating used in the deposition of thick Au heat spreader.

PAGE 89

81 Au heat spreader (b) Figure 4.8 (a) Schematic diagram of Au heat spreader patterns, (b) The outlook of finished Au heat spreader on the top of thin p-clad InGaAs SQW diode lasers, where Au heat spreader thickness is ~5 ^m.

PAGE 90

82 400 Input current (mA) Figure 4 . 9 Total CW output power versus input current characteristics for 5-|im stripe thin p-clad InGaAs SQW diode lasers with heat spreader and without heat spreader, where heat spreader thickness is ~7 |im.

PAGE 91

83 0.50m GRATING SPECTROMETER Figure 4.10 Schematic diagram of thermal resistance R^^ measurement setup used in the experiments.

PAGE 92

84 (b) Figure 4.11 The variations of the Fabry-Perot mode shifts in 50 |J.m stripe width, 500 cavity length, InGaAs SQW diode laser measured by using the setup shown in Figure 4.10, where laser is operated at I~l.l I^^ and, (a) Ths=21.8 C, 40 % duty cycle, (b) T„s=21.8 C, 20 % duty cycle and (c) Ths=24.8 C, 40 % duty cycle. The vertical central line is used as the referenced line for the thermal resistance measurements.

PAGE 93

85 Figure 4.11 -Continued.

PAGE 94

86 15 1 ' ' ' I ' ' ' I ' ' ' ' ' ' ' ' ' ' 0 20 40 60 80 100 Input current duty cycle (%) Figure 4.12 The measured heat sink temperature (T„g) rise as a function of duty cycle for the 50 \Lm stripe width thin p-clad InGaAs SQW diode lasers with and without Au heat spreader. The dashed lines are the extensions of the linear relationship to obtain the AT^,, value from the intersect points at 100 % duty cycle respectively.

PAGE 95

87 Input current (mA) Figure 4.13 Typical CW P-I and I-V characteristics of 50 ^Im stripe width, Au contact, thin p-clad InGaAs SQW diode laser .

PAGE 96

88 160 cd (A cn u g Vi 120 80 40 0 T — I — I — I — I — I r -I — I — I — r * w=50 ^im, L=500 fim p-side up "4+ -I I I I I I I I I L. 0 4 8 12 16 Thickness of Au heat spreader (|im) Figure 4.14 The measured dependence of thermal resistance on the Au heat spreader thickness for thin p-clad InGaAs SQW diode lasers.

PAGE 97

. 89 160 140 120 100 80 60 40 I I I I I I I I I I I I I I I I t I I I I I I I I I Thin p-clad InGaAs SQW LD T-T -e — pulse -k -CW * w=50 |im, L=500 |im 40 50 60 70 80 90 100 110 (CAV) tn Figure 4.15 The dependence of characteristic temperature on the thermal resistance R^,, of diode lasers, where lasers are pulsed operation or CW operation.

PAGE 98

90 40 30 20 10 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 Thin p-clad InGaAs SQW LD G ©. R th O44 -57 -e— 96 -44 -k -57 -—96 I I I I I * solid: CW . empty: piilse 10 20 30 40 50 60 70 80 90 Heat sink temperature (C) Figure 4.16 The variations of differential quantum efficiency (dQE) as a function of heat sink temperature of thin p-clad InGaAs SQW diode lasers with various R^^, where lasers operated in pulse and CW are compared.

PAGE 99

CHAPTER 5 p*-GaAs CAP LAYER THICKNESS EFFECTS 5 . 1 Introduction In previous chapters we have described the influence of contact reflectivity on the thin p-clad diode laser performance and several diode laser fabrication advantages of this interesting structure. However, since the p-cladding layer is only 250 nm thick and, especially, the associated refractive index value at the lasing wavelength for the cap layer is very close to that of quantum well, the p'^-GaAs cap layer thickness could substantially affect the diode laser behavior [Suem75] . In our continuous work on this thin p-clad (250 nm) laser structure, we have found that both wide stripe and narrow stripe lasers with 200 nm p'"-GaAs cap layer show a long time delay between the application of an excitation current pulse and the onset of stimulated emission [Wu96a] . In addition, this lasing delay time becomes shorter as the input current increases. Besides the lasing time delay characteristics, threshold current of this thick p-cap (200 nm) laser is abnormally high and show strong dependence on 91

PAGE 100

92 the period of the excitation current on-time. Moreover, as the cap layer thickness is reduced below 170 nm, diode laser performance improves significantly and no lasing delay is observed for the wide stripe lasers. For the narrow stripe 300 ridge-height lasers, stripe width (w) becomes an important factor in determining diode laser performance. Long time lasing delay behaviors are observed for the 6-|Jin stripewidth lasers and Q-switching lasing performances are obtained for both 3.5-|J.m stripe-width and 2.5-|lm stripe-width lasers. However, when stripe width is reduced to w=1.5 |Xm, neither long time lasing delay nor Q-switching performance is found. In Section II, diode laser structure used in this study is presented. Also included are the theoretical calculation results of optical confinement factor F and modal loss a.^ of thin p-clad laser structure. The gain-guided device fabrication process, experimental results and discussions are described in Section III. In Section IV, the details of fabricating process for narrow stripe, 300 nm ridge-height lasers and performance characterized are presented. 5 . 2 Laser Structure and Theoretical Calculations Figure 5.1 shows the thin p-clad InGaAs SQW diode laser structure used in this study, where the quantum well thickness is 80 A and p^'-GaAs cap layer thickness t is 200 nm. As mentioned before, the increase of cap layer thickness could substantially affect diode laser performance, in order

PAGE 101

93 to check this point the optical confinement factor and the corresponding modal loss of the laser structure shown in Figure 5.1 are calculated as a function of p^'-GaAs cap layer thickness t. In the computations, a lasing wavelength of 950 nm and shiny Au metal with refractive index of 0.174 i 5.691 [Gray72] used as the p-contact are assumed. The calculated results are shown in Figure 5.2. In this plot, it is noticed that T and remain relatively flat upto t ~100 nm and then changes slowly when t is increased from 100 nm to 140 nm. In addition, when t is increased beyond 140 nm, the modal loss increases quickly and the optical confinement T decreases significantly. At 200 nm cap layer thickness, modal loss ttj^ is increased to ~66 cm"^ and optical confinement factor r becomes as small as ~0 . 4 %. The calculated results indicate that much more material gain is required for thin pclad lasers to reach threshold condition if the p-cap layer thickness is changed from t = 100 nm to t=200 nm. This speculation could also be seen from the calculated near field intensity distributions as shown in Figure 5.3, where the peak intensity in the quantum well decreases quickly with the increase of t (t>150) and becomes smaller than that in the pcap layer when t is thicker than 175 nm. As p-cap thickness is continuously increased to t=200 nm, most of the near field intensity is "coupled" inside the p-cap layer. Consequently,

PAGE 102

94 a very small value of F and a large are obtained as shown in Figure 5.2. 5 . 3 Long Time Lasing Delay in Gain-Guided Lasers To check diode laser performance, gain-guided lasers with stripe width w=50, 25, 10 and 5 are fabricated, where non-alloyed Au and alloyed Ge/Au/Ni/Au metallurgy are used as the p-type and n-type contact respectively. Diode lasers with various cavity length are then soldered on the In coated copper blocks and characterized at room temperature. Figure 5.4 shows the pulsed output power (P) versus input current (I) characteristics of thin p-clad gain-guided diode lasers with 325 |im cavity length and different stripes when measured at 2 fisec pulse width and 1 KHz repetition rate. As can be seen from this plot, a lot of spontaneous emission occurs for all samples before the stimulated emission starts. In addition, threshold currents are abnormally high and slope efficiencies are much smaller (< 0.10 mW/mA) than those of lasers fabricated with normal p-cap layer thickness (100 nm) . From the plot of inverse slope efficiency r\^'^ versus cavity length L, a measured modal loss of ~64 cm'^ is obtained, which is in good agreement with the calculated results. As stated in Section 5.1, these diode lasers show long time delay between the excitation of current pulse and the

PAGE 103

95 start of stimulated emission. To find out this point, lasers with various stripe widths are measured at a constant repetition rate: IK Hz and different pulse widths: 2, 4, 6, 8, 10 and 12 |isec. All the samples show similar lasing delay behaviors and the lasing delay time decreases as the input current increases. Figure 5 . 5 (a) -5 . 5 (d) show the variations of the time response (upper trace) of a 50 \lm stripe, 500 |Xm cavity gain-guided thin p-clad InGaAs SQW laser when a current pulse I (lower trace) with 12 |lsec pulse width and 1 KHz repetition rate is applied. At 1=900 mA, the lasing delay time is -10 |isec . and then decreases to ~5 )lsec . when I is increased to -1100 mA. In addition to the lasing delay, we have found that lasing threshold currents of diode lasers are greatly dependent on the period of input current on-time X. Figure 5 . 6 (a) -5 . 6 (d) show the measured X dependent lasing behaviors (upper trace) of a 50 )lm stripe width diode laser, where the lasing output power is kept constant at -5 mW/facet. Two interesting results are observed from the measurements :( 1 ) the input current I for having the same output power decreases remarkably from I -1600 mA to I -1000 mA as X is increased from 2 [Lsec to 8 |lsec and then remains unchanged as x is beyond 12 |lsec . However, if x is increased upto very high value, due to heating effect, stimulated emission would not occur. (2) The spontaneous emission intensity decreases as X is increased. Figure 5.7 (a) and (b)

PAGE 104

show the schematic diagram of the lasing delay phenomena and the measured dependence of threshold T on the input current for diode lasers with 500 M.m cavity length and various stripe width: 5, 10, 25 and 50 |lm, respectively. For the 50 |J.m stripe width laser, the averaged current decreases significantly from -1000 mA to -720 mA (-28 % reduction) as the threshold input current pulse on-time is increased from 2 ^sec to 12 |lsec. For the 5 |lm stripe width laser, the averaged current decreases from -280 mA to -220 mA (-21 % reduction) when the threshold input current on-time T^t, increases from 2 [Lsec to 12 |lsec. The possible reason for the difference of threshold current improvements between 50 |Xm stripe laser and 5 |J.m stripe laser could be due to the higher current spreading effects from the 5 |lm stripe laser than that of the 50 |lm stripe laser [Yone73] . The calculated results of Figure 5.2 indicate that thin p-clad laser structure with 200 nm p-cap layer shows small optical confinement factor F and large modal loss . To understand what is the dominant cause for the abnormal lasing delay performance in wide stripe gain-guided quantum well diode laser, both thick p-clad and thin p-clad InGaAs SQW gain-guided diode lasers with thin (100 nm) p'^-GaAs cap layer thickness are fabricated and characterized. Moreover, wide stripe thin p-clad InGaAs SQW diode laser with less shiny Ni contact are measured and used as the large modal loss referenced sample for comparison. Table 5.1 shows the

PAGE 105

97 measured results of lasing delay behaviors of various stripe width, gain-guided InGaAs SQW lasers, where sample L978 is thick p-clad (1300 nm) laser with 100 nm p-cap layer, sample L979 and sample L529 are thin p-clad (250 nm) lasers with 100 nm p-cap layer and sample L980 is thin p-clad (250 nm) laser with 200 nm (structure as shown in Figure 5.1). It is noticed from this table that lasing delay occurs for the narrow stripe (<10 |J.m) gain-guided SQW lasers of all three sample types and is independent of the optical confinement factor F and modal loss . However, for the wide stripe lasers, only samples fabricated from laser structure (L980) with small optical confinement F ~ 0.4 % have the unusual lasing delay behaviors. The comparison between sample L980 and L529 also clearly indicates that the decrease of optical confinement induced by the increase of p*-GaAs cap layer thickness in the thin p-clad of laser structure is the main cause of the abnormal lasing delay observed in the wide stripe single quantum well laser. This statement is also indirectly consistent with the experiment results observed in double heterostructure (DH) lasers [Dyme72], where optical confinement factor F is big and no lasing delay obtained. Injection carriers induced refractive index changes of active region have been shown important in determining the lasing delay behaviors of semiconductor lasers [Grun74, Thom74, Nune77]. In addition, this index change was found as

PAGE 106

98 a function of current pulse time and injection current density [Nune77]. For the InGaAs quantum well, carrierinduced refractive index depression larger than 0.5 has also been reported [Shie89] . To see how the refractive index change of active region An^^^^^g could affect diode laser performance, the variations of optical confinement factor F and the associated modal loss are calculated for the laser structure shown in Figure 5.1. Figure 5.8 shows the calculated dependence of optical confinement factor and modal loss on An^^^^yg, where two different p-cap layer thickness t = 100 nm and t=200 nm are considered. For the 200 nm p-cap layer device, both the optical confinement factor and modal loss show strong dependence on An^^^^^g. When An^^^^^^ is changed from An3j,j^yg =-0.2 to An^^j.^^^ = +0.2, the associated optical confinement factor changes from 0.21 % to 0.81 % and modal loss changes from 72 cm'^ to 52 cm"^ . For the 100 nm p-cap layer device, under the same variations of An^^^j^^^ the corresponding optical confinement factor changes from ~2 % to 2.7 % and modal loss shows a relatively small change from ~4 cm"'^ to 2 cm"''. The combined variations of optical confinement factor and modal loss due to the change of An^^^^^g could lead to a dramatic difference in threshold material gain g^^ for the two different laser types to reach threshold conditions. Figure 5.9 is the calculated results of threshold quantum well gain g^^, as a function of refractive index change of active region An^^^^^g for diode lasers with 100 nm, 200 nm p-

PAGE 107

cap layer thickness, where a cavity length L=500 |im and a facet reflectivity of 0.3 are assumed. It is noticed from Figure 5.9 that g^^, changes significantly for the laser with 200 nm p-cap layer from (^"active^ 2.7x10^ cm"^ to gth(An,,,,,,= -0.2) -4.5x10^ cm"^ and g,, (An,,,,,, = +0.2) -0.94x10^ cm'^ when An,,,,,, changes from 0 to -0.2 and +0.2, respectively. In contrast, under the same variations of An,,,,,,, only a small amount of g^^, change from gth ('^"active^ -0.2)= 1400 cm"^ to g^f, (An3^,,,g= +0.2)= 970 cm'^ is found for the 100 nm p-cap layer laser. The dramatic difference of An^^-^i^yg-dependent g^^, variations between the 100 nm p-cap layer laser and 200 nm pcap layer device are consistent with the experimental results shown in Table 1. This is because the variations of g,^ caused by An^^^i^g of 100 nm p-cap laser is so small that 1,^ could remain relatively unchanged during the whole current pulse of measurements and no lasing delay is obtained. However, for the 200 nm p-cap laser, the variation of g^^, induced by An^^^^^ is very big that 1,^, could be changed dramatically at different input current on-time and causes lasing delay behaviors . To explain the unusual lasing delay behaviors of wide stripe SQW lasers, the model proposed by Nunes et al. [Nune77] for the SH diode lasers is adopted, where the refractive index of the active region is primarily affected by three different factors: (1) the injection carriers, (2) the gainguiding confinement and (3) the heating of active region. Factor (1) could results in a negative increase of refractive

PAGE 108

100 index of quantum well while factor (2) and (3) could cause a positive refractive index change. The combination of these three factors could result in a great variation of An^^^j^^^ when input current pulse width T and the injection current density J are changed. Figure 5.10 shows the schematic diagram of the lasing delay model used to explain the experimental results. In this Figure, at some constant current density J^, the refractive index change due to the injection carriers could dominate An^^^^^^ at the initial current on-time T^^ and causes a depression of An^^^^^g, causing lasing mode overlap with quantum well gain to decrease and lasing mode overlap with loss in Au to increase, hence, higher g^,, is required for lasers to start the stimulated emission. If the quantum well gain is not sufficient to overcome total loss, laser could not reach threshold condition. As X continues to increase, An^j.^^^^ could become positive, causing lasing mode overlap with quantum well gain to increase and lasing mode overlap with loss in Au to decrease, and less g^^, is needed to reach threshold condition. Stimulated emission could begin if the laser quantum well gain is large enough to compensate the total loss and lasing delay occurs. Consequently, the lasing threshold of diode laser strongly depends on the current ontime, in other words, the threshold behaviors of lasers are input current on-time dependent. On the other hand, at certain current on-time Xj, An^^^^^^ could be dominated by the

PAGE 109

101 injection carriers at some injection current density and cause a negative value of An^^^^^^, hence higher g^^ is required to reach threshold condition and no stimulated emission is obtained if the material gain is not equal to the total loss. At higher injection current density Jj, the heating effects of active region could dominate the index change of quantum well and An3<,ji^g could be increased in the positive direction. At this point, less g^^ is needed for reaching threshold condition and laser could start lasing at some time point before if the laser material gain is well above the total loss. As a result, lasing delay would become shorter as the input current density increases. In Section 5.2 we have shown that optical confinement factor r as well as modal loss could be significantly improved by decreasing the p*^-GaAs cap layer thickness of thin p-clad laser structure and lead to an improvement potential of diode laser performance. To check this point, wide stripe (50 \im) thin p-clad diode lasers with three different p-cap layer thickness: 170±5 nm, 130±5 nm and 70±5 nm are fabricated with non-alloyed shiny Au and alloyed Ge/Au/Ni/Au as the p-contact and the n-contact . The various p-cap layer thickness is performed by etching away part of the 200 nm p^-GaAs cap layer shown in Figure 5.1 through the utilization of pulsed anodic oxidation technique [Grov94] and MIF 312 developer to strip off the remaining anodic oxide. Lasers with various cavity length are cleaved, soldered on In

PAGE 110

102 coated copper blocks and characterized. Figure 5.11 shows the variations of pulsed threshold currents as a function of cavity length for samples with different p-cap layer thickness when measured at 2 |Xsec current pulse width and 1000 Hz repetition rate. It is interesting to see that threshold currents 1^^ decrease tremendously as p-cap layer thickness is reduced from 200 nm to -130 nm and then slowly decrease when p-cap layer thickness is reduced further. In addition, the increase rate of 1^^ with cavity length is much smaller for the lasers with thin p-cap layer thickness (< 170 nm) than that of devices with 200 nm p-cap layer thickness. The decrease of lasing threshold behavior is well consistent with our predictions as shown in Figure 5.2, where the improvements of optical confinement factor T as well as the modal loss could directly result in a great reduction of diode laser threshold performance as described in equation (2.2.11). Another important feature is that no lasing delay is observed for the wide stripe lasers during the measurements. This again indicates a significant improvement of optical confinement from the decrease of p-cap layer thickness. The decrease of modal loss with the reduction of p-cap layer thickness shown in Figure 5.2 could also be checked from the measurements of laser slope efficiency TJ^ at various laser cavity lengths. Figure 5.12 is the measured modal loss of thin p-clad InGaAs SQW diode lasers as a function p''-GaAs cap layer thickness. The dashed line in this

PAGE 111

103 plot is the computation results for comparison. As can be seen from the measured points of this figure, the modal loss decreases rapidly from =64 cm'^ to a^=3i cm~^ when p-cap thickness is decreased from 200 nm to -170 nm. At ~130 nm pcap layer, the measured modal loss is ~8 cm"^ and changes slowly to ~3 cm'^ as p-cap layer thickness is reduced further to -70 nm. The measured results are in good agreement with the theoretical predictions. 5 . 4 Long Time Lasing Delay and O-Switching Lasing Delay in Narrow stripe. 300 nm Ridge-Height Lasers. It has been reported that lasing delay behaviors of narrow stripe quantum well lasers can be eliminated if there is a built-in guide along the junction plane with an effective refractive index step greater than lO"'' [Prin85] . In addition, strong waveguiding effects obtained by using thin p-clad structure to fabricate narrow stripe (5 \lm) indexguided lasers with performance comparable to those of conventional thick p-clad lasers have been demonstrated [Wu95] . For similar thin p-clad laser structure with 200 nm p-cap layer thickness studied in this chapter we are interested in understanding whether the strong waveguiding effects of narrow stripe lasers could be obtained to eliminate lasing delay by simply removing parts of the outside stripe material. In Section 5.2 and 5.3 we have demonstrated that the variations of p-cap layer thickness could substantially

PAGE 112

104 change the waveguiding of thin p-clad laser structure. Consequently, the behaviors of ridge waveguide lasers fabricated from thin p-clad material with 200 nm p-cap layer should be quite different from that of ridge-guide lasers fabricated from thin p-clad material with 100 nm p-cap layer. The difference in effective refractive index ( lateral index step ) between the ridge region and the out side-ridge region An is normally used in designing ridge-guide lasers [Agra84] . Therefore, it is interesting to calculate An as a function of ridge height ( the amount of epitaxial material removed in defining the ridge ) for thin p-clad configuration. In the computations, it is assumed that the p"'"-GaAs cap layer, the p-AlxGai-xAs graded-layer and part of the p-clad layer of the outside-ridge region are removed and replaced with a native oxide, the whole structure being covered with Au . A lasing wavelength of 950 nm, an oxide refractive index of 1.8 and a Au refractive index of 0.174-15.691 are also assumed in the computations. Figure 5.13 shows the calculated An as a function of ridge height (distance measured from the top of p-cap layer on the stripe region to the top of p-clad layer on the outside stripe region) when various oxide values are used as the current blocking layers. As can be found from this plot that a calculated effective refractive index step An as high as 3.3x10"^ is obtained when the outside stripe p'^GaAs cap layer is totally removed. This An value is -10 times larger than that calculated from thin p-clad laser structure with p-cap layer 100 nm (as shown in Figure 3.1 (a) ) . As a

PAGE 113

105 result, strong waveguiding effects are expected for the narrow stripe lasers when fabricated from this thin p-clad QW laser structure. Besides the An computations, we also calculate the lateral far field intensity distributions of diode lasers with different stripe widths. In the computations, a three layer waveguide structure is assumed and the associated refractive index values are quoted from the calculated effective refractive index of the stripe region and outside stripe region. The calculated results are shown in Figure 5.14, where various stripe width w=l, 2, 3 and 5 |i.m are concerned. It is noticed from this plot that single lateral lobe of far field pattern could be obtained when stripe width w is not wider than 2 jim. At w=3 )Xm, a double lobe pattern is obtained while at w=5 |i.m a triple lobe occurs . To fabricate narrow stripe diode lasers, 6 \lm, 3.5 |lm, 2.5 |lm and 1.5 |lm stripe width on 500 \im centers are defined by using standard photolithography technique. Ridge-guided laser structures are formed by etching away the outside stripe p^-GaAs cap layer material through the chemical solution of 1 NH^OH + 1 H^Oj + 50 H^O for -35" and followed by a anodic oxidation ~4' to grow oxide on the outside stripe region. The resultant ridge height is -300 nm after measured by Dektak. Wafer is then thinned to 100 \lm and deposited with alloyed Ge/Au/Ni/Au and nonalloyed Au as n-type and ptype contact respectively. Lasers with various cavity length

PAGE 114

106 are cleaved, soldered on the In coated copper blocks and characterized . Table 5.2 shows the suminarized results of narrow stripe diode lasers with cavity length L=500 |lm and different stripe width when measured at 2 |lsec . current pulse width and 1000 Hz repetition rate. It is interesting to find that all measured sample types except the 1.5 |lm stripe-width sample show either Q-switching or long time lasing delay performance [Ripp74] . Additionally, threshold currents for the samples with lasing delay behaviors are quite high as compared with the results presented in Chapter 3 and then decreases quickly when stripe width is decreased to w=l . 5 |J.m. The 6 Jim stripewidth sample shows high average threshold current 1^^ ~250 mA and long time lasing delay behaviors similar to those observed in the gain-guided lasers. For the samples with either 2.5 |lm or 3.5 \im stripe-width, a transition from Qswitching to long time lasing delay performance are observed during the measurements . The average threshold currents for both sample types are I^f,(w=2.5 ^im) ~80 mA and Ij^(w=3.5 |lm)~180 mA respectively. Figure 5.15 (a) -(d) shows the evolution of lasing behaviors of the 2.5 |lm stripe-width sample as a function of input current I, where the measured condition is repetition rate: 1000 Hz, pulse width: 2 |i.sec. At 1=80 mA, it is clear to see that stimulated emission starts at the time pulse position corresponding to the ending edge of the input current pulse i.e. Q-switching lasing behavior. As the input current I is increased to 100 mA, both

PAGE 115

107 Q-switching and long time lasing delay occur at the same time. When I is increased further to 120 mA, Q-switching disappears and long time lasing delay dominates diode laser performance. For the 1.5 |lm stripe-width laser, neither Qswitching nor long time lasing delay is observed and threshold current decreases dramatically as compared with the samples with wider stripe width (I^j,(w=1.5 ^m)~25 mA vs I^j, (w>2 . 5 |Xm) >80 mA) . Possible explanation for the totally different lasing behaviors between the various stripe-width samples could be due to the different waveguiding effects caused by the effective refractive index step An. From the calculated results shown in Figure 5.13, at 300 nm ridge height the associated lateral effective refractive index step is An -3.5x10"^. For the 6 |im stripe-width sample the An -3.5x10"^ could be insufficient to provide strong waveguiding effects to compensate the weak transverse waveguiding effects for laser to operate at low threshold condition. For both 3.5 [im stripe-width sample and w=2 . 5 |J.m stripe-width sample, although the lateral waveguiding formed by the refractive index step An is still weak to obtain index guiding, the lateral optical confinement could be somehow improved such that Q-switching performance is observed. Since it has been theoretically studied that total loss of diode laser could decrease faster than the gain and confinement so that stimulated emission can occur at the end of current pulse even though laser can not lase during the current pulse

PAGE 116

108 [Nune77] . We believe the large index step An= 3.5x10'^ in our devices (lasers with stripe with w=2 . 5 and w=3.5 pm) could somehow improve the optical confinement of laser waveguide structure and make stimulated emission occur at the ending edge of current pulse. The significant decrease of threshold current of the 1.5 \im stripe-width laser lead us to believe that the effective lateral refractive index step An= 3.5x10"^ could be sufficient to provide strong waveguiding effects to overcome the drawback of small transverse optical confinement as shown in Section 5.2. Since it is believed that waveguiding effects in narrow stripe, ridge-guided lasers are two dimensional properties (transverse and parallel to the quantum well plane) and the waveguiding effect in each dimension is related to that of the other. In order to check this point, we have used the far-field measurement setup (shown in Figure 3.3) to measure the far-field intensity distributions of all four sample types. Figure 5.16 shows the measured far-field intensity distributions of diode lasers when measured at 1-1.3 1^^, with 10 |i,sec. current pulse, 10 KHz repetition rate and room temperature. As can be seen from this plot that a four-lobe pattern is obtained for the 6 |i.m stripe-width sample. For the 3.5 |i.m stripe-width sample, a triple lobe pattern with an unsymmetric side lobes is observed. For the 2.5 \im stripe-width sample, a symmetric double lobe pattern with a large separation is obtained. When the input current or the current pulse width is increased, each lobe is widen and the separation becomes smaller. For

PAGE 117

109 the 1.5 }i.m stripe-width laser, a single lobe pattern is found during the measurement and remains single lobe pattern when current pulse width is increased. The far-field intensity distribution measurements indicate that An= 3.5x10'^ is sufficient to provide strong lateral waveguiding effects for the 1.5 |J.m stripe-width sample to overcome the weak transverse waveguiding and operate with a single spatial mode output. Figure 5.17 shows the measured room temperature CW PI characteristics of the 1.5 |im stripe-width laser. The threshold current and differential quantum efficiency are I^h -29 mA and r[^= 50 % respectively. The CW transverse (perpendicular to the QW plane) and lateral far-field intensity distributions when measured at 25 C and 1=1.3 I^^ are shown in Figure 5.18 (a) and (b) . It is noticed that both transverse and lateral far field pattern show single lobe behavior despite the transverse waveguiding is very weak. In addition, the FWHM values of both lateral and transverse far field are larger than those obtained before. In summary, we have shown that p-cap layer thickness is important in determining thin p-clad SQW diode laser performance. At 200 nm p-cap layer thickness, lasing delay occur in both wide stripe and narrow stripe gain-guided lasers. As p-cap layer thickness is decreased, diode laser performance could be dramatically improved and no lasing delay are observed in the wide stripe diode lasers. The lasing delay behaviors can be explained by the refractive

PAGE 118

110 index change of quantum well induced by the current pulse time related injected carriers. In addition, at 1.5 ^im stripe width, strong lateral waveguiding effects are sufficient to overcome the deficit of transverse waveguiding and make laser operate in low threshold regime .

PAGE 119

Ill p'^-GaAs cap layer: t nin p-Al^Ga^.zAs: 25 nm (z=0.6-0.05) p-clad layer AIq 5GaQ 4AS: 250 nm p-Al^Ga^.^As: 200 nm (x=0.3-0.6) GaAs: 7 nm InyGa^.yAs, SQW (V'-O. lSr S nm) GaAs: 7 nm n-AljjGa-|^.^j^: 200 nm (x=0.6-0.3) n-clad layer AIq gGaQ 4AS: 1400 nm Figure 5.1 Thin p-clad InGaAs SQW diode laser structure with 80 A quantum well and thick p'^-GaAs cap layer t=200 nm used in the p''-GaAs cap layer thickness effect study.

PAGE 120

112 80 ^ 60 g (A o O 40 20 0 -J— I— I— I— r— r T— r T — r Thin p-clad InGaAs SQW LD * lasing wavelength X = 950 nm •••••v-r-r""""I I 3.0 2.5 2.0 1.5 1.0 0.5 0 0 50 100 150 200 p'*'-GaAs thickness, t (nm) u o S o o Figure 5.2 The calculated dependence of modal loss and optical confinement factor T on the p''-GaAs cap layer thickness t of the laser structure shown in Figure 5.1.

PAGE 121

113 S3 (A a 0) C3 2 2 5 1 0 I I I I I I I I I I I I I I I I I I [ I I I I I I I I I I I I I I I I I I I . Thinp-cladlnGaAsSQWLD p*-GaAs cap layer = 100 run 150 nin J 175 nm 200 nm metal I semiconductor ' ' I ' ' I ' ' ' I ' ' ' I ' ' ' I ' ' ' 0 400 800 1200 1600 Position (nm) Figure 5.3 Calculated near field intensity distributions of thin p-clad InGaAs single quantum well laser structure shown in Figure 5.1 with different p^'-GaAs cap layer thickness: 100 nm, 150 nm, 175 nm and 200 nm, where non-alloyed Au is assumed as the p-contact .

PAGE 122

114 15 0^ S 9 ^^^^ u o O T— I — I I I [ — I — r— I — I — I I I I I I I I I I I I I I Thin p-clad InGaAs SQW LD w=5 |im 10 |im 25 |im 6 3 * w/0.2 p*-GaAs, cavity length L=325|xm 0 T • t • t A A A J I I I I I I I I 50 |j.m 9 I I _ 0 _l I L. 0 0.2 0.4 0.6 0.8 1 1.2 Input current (A) Figure 5.4 Measured pulsed output power P versus input current I characteristics of various stripe width, gain-guided thin p-clad InGaAs SQW diode lasers, where non-alloyed Au and alloyed Ge/Au/Ni/Au are used as the p-contact and n-contact respectively.

PAGE 123

115 (b) Figure 5.5 Measured lasing delay characteristics of wide stripe (50 |lm) thin p-clad InGaAs SQW diode laser at various input currents : (a) 800 mA, (b) 900 mA, (c) 1000 mA and (d)llOO mA, where the horizontal scale is H=2 Jlsec/div., vertical scales are 100 mV/div. (upper trace), and V^=500 mA/div. (lower trace).

PAGE 124

116 Figure 5.5 -Continued.

PAGE 125

117 Figure 5.6 Measured input current pulse width dependent lasing threshold characteristics of wide stripe (50 |im) thin p-clad InGaAs SQW diode laser at pulse widths: (a) 2 iisec, (b)4 ^sec, (c) 6 ^isec and (d) 8 ^isec where the horizontal scale is H=2 |isec/div., vertical scales are V^= 50 mV/div. (upper trace), and V^=500 mA/div. (lower trace)

PAGE 126

118 Figure 5.6 -Continued.

PAGE 127

119 ^ A 3] 3 t below threshold above threshold time 4 (a) 700 600 500 400 300 200 stripe width w = 50 I I I I I I I I I I I I I I I I I I I I I I I I I I I I I Thin p-clad InGaAs SQW LD (b) I ' ' I ' I ' I ' I I I I 1500 1000 g 500 ^ 0 0 2 4 6 8 10 12 14 (|isec.) Figure 5.7 (a) Schematic diagram of lasing delay phenomena, (b) Measured dependence of current on the threshold input current on-time of different stripe width, gainguided, thin p-clad InGaAs SQW diode lasers with 200 nm p*-GaAs cap layer, where the repetition rate is 1000 Hz.

PAGE 128

120 Lasing time delay Sample No. /V Modal loss and Optical confinement factor Gain-guided laser stripe width 50 [im 25 |im 10 ^mi 5 L978 (thick p-clad) Au contact: apl cm'^ r = 2.3 % NO NO NO YES L979 (thin p-clad) Au contact: aj=3 cm'^ r = 2.3 % NO NO NO YES L980 (thin p-clad) Au contact: ap66 cm'^ r = 0.4 % YES YES YES YES L529 (thin p-clad) Ni contact: aj=70 cm"-'^ r = 2.3 % NO e 5.1 Comparison of measured lasing delay for various gain-guided SQW lasers with stripe width: 50, 25, 10 and 5 |lm, where sample L978 is with 1300 nm p-cladding layer and 100 nm p-cap layer; sample L979 and sample L529 are with 250 nm p-cladding layer and 100 nm p-cap layer; sample L980 is with 250 nm p-cladding layer and 200 nm p-cap layer.

PAGE 129

121 Index change of active region Figure 5.8 Calculated variations of optical confinement factor r and modal loss of thin p-clad InGaAs SQW laser as a function of refractive index change of active region An^^.^^^^, where p*-GaAs cap layer thickness =100 nm and 200 nm are considered, respectively.

PAGE 130

122 I' l l — I I ' ' 1 — ' ' " I ' ' ' I ' Thin p-clad InGaAs SQW LD I_l I I L_l I I I I I I I— I l—i 1 1 1 l_I -0.4 -0.2 0 0.2 0.4 Index change of active region Figure 5.9 The calculated dependence of threshold quantum well gain g^^, on the refractive index change of active region An^^^j^^g for thin p-clad InGaAs SQW laser with different p*-GaAs cap layer thickness: 100 nm and 200 nm, where a facet reflectivity of 0.3 and cavity length L=500 |J.m are used in the computations.

PAGE 131

123 < + 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 : ^1 < ^2 < ^3 / III 1 J 1 1 1 1 ill I 1 r time during the / ; current pulse = / ? r ^3 y / 1 1 1 ^""''^ 1 ~ 1 1 * ' ' 1 l-J 1 1 ' I ^ « ' ' ' (Ji < J2) : ' 1 1 0 J J 1 2 Input current density J (a. u.) Figure 5.10 Schematic diagram of the dependence of refractive index change of quantum well on the input current density and current pulse time during the current pulse used to explain the unusual lasing delay behavior of thin p-clad InGaAs SQW lasers with thick pcap (200 nm) layer.

PAGE 132

124 400 300 200 o « 100 0 •" Q— 170±6nm •— V-— 130±5 nm •--•A---70±5 nm — • — 200 nm T-i— T— r -i — I — I — I — I — I — I — I — I — I — I I I p^-GaAs thickness * stripe width w=50 nm. a I _l I I L. 1.6 1.4 1.2 1.0 0.8 0.6 0.4 0.2 200 400 600 800 1000 1200 Cavity length (}im) 2 u Figure 5.11 The variations of threshold current as a function of cavity length for thin p-clad InGaAs SQW diode lasers with 50 |J.m stripe width and different p*GaAs cap layer thickness when measured at IK Hz repetition rate and 2 |lsec pulse width.

PAGE 133

125 I g O 80 60 S 40 O 20 0 ' ' ' I I ' ' ' ' I ' ' ' ' Thin p-clad InGaAs SQW LD experiments calculations * stripe width w= 50 |im _i I i_ _l I I I I I I L. 0 50 100 150 200 250 p*-GaAs thickness (nm) Figure 5.12 The variations of the measured modal loss of thin p-clad InGaAs SQW diode lasers as a function of the p''-GaAs cap layer thickness, where the calculated results as shown in Figure 5.2 are also plotted for comparison .

PAGE 134

126 anodic oxide p-clad layer metal quantum well n-clad layer — 1 stripe region p-Alo.6GaO.4As ^ 0.25 \im p+-GaAs ^ AlxGai.xAs: x=0.3-0.6 GaAs GaAs Al^Gai.xAs :x=0.6-0.3 n-Alo gGag 4AS ridge height (a) 6.0
PAGE 135

127 >» 1.0 g 0.8 .S 0.6 I 0.4 I ^ 0.2 « 0.0 o 2 T 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 T stripe width J I I I I -60 -40 -20 0 20 40 Lateral angle (degree) Figure 5.14 Normalized lateral far-field intensity distribution calculated from the laser structure shown in Figure 5.13 (a) when different stripe widths w=l )Xm, 2 |lm, 3 |Xm and 5 |lm are concerned. It is assumed the ridge height is 300 nm and current blocking oxide on the outside stripe region is 100 nm.

PAGE 136

128 Narrow stripe thin p-clad InGaAs SQW diode lasers with 300 nm ridge height Laser performance stripe width: w 1.5 fj.m 2.5 )im 3.5 |im 6 ^m average threshold current: \^\^ 25 mA 80 mA 170 mA 280 mA La sing delay behavior NO lasing delay Q-switching to long time delay Q-switching to long time delay long time delay Table 5.2 Summary of narrow stripe, thin p-clad SQW laser performance when measured at 2 jlsec. current pulse width and 1 KHz repetition rate, where the diode laser cavity length is 500 )lm.

PAGE 137

129 V„=20 mV/div. V^=50 mA/div. V^=20 mV/div. Vi=50 mA/div. (b) Figure 5.15 The evolution of lasing behavior from Qswitching to long time lasing delay of a 2.5 [Lm stripewidth thin p-clad InGaAs SQW laser with 300 nm ridge height when measured at 2 |J,sec. current pulse width, 1000 Hz repetition rate and input current (a) 1=60 mA, (b) 1=80 mA, (c) 1=100 mA, (d) 1=120 mA. The upper trace and the lower trace are the optical power and input current signal respectively. The horizontal scale is 0.5 |lsec./div.

PAGE 138

130 Figure 5.15 -Continued.

PAGE 139

131 !3 •FN 2 I I r stripe width = 6 |im — 3.5 \Lm 2.5 |im 1.5 ^im O _i I I I i_J. I I -80 -40 0 40 Lateral angle (degree) Figure 5.16 Normalized lateral far-field intensity distributions of thin p-clad InGaAs SQW lasers with 300 nm ridge height and various stripe width when measured at 1=1.3 I^^, 4 |lsec. current pulse width, 1000 Hz repetition rate and room temperature.

PAGE 140

132 Input current (mA) Figure 5.17 Measured room temperature CW output power versus input current characteristics of a 1.5 \lm stripe-width, thin p-clad InGaAs SQW laser with 300 nm ridge height.

PAGE 141

133 0 N •a a I 1.0 0.5 0 1.0 0.5 0 * p-cap= 200 nm w=1.5 |J.m L=500 |im CW (//) I I I I I I I I I I L I I I I I I I I [ I I I I I I I I I I I I I I I Thin p-clad InGaAs SQW LD (+) (a) (b) ' 62° I ' ' I ' I ' ' I ' I 75 -50 -25 0 25 50 75 Angle (degree) Figure 5.18 (a) Lateral and (b) transverse far-field intensity distributions of thin p-clad InGaAs SQW laser with 300 nm ridge height, 1.5 |J.m stripe-width when measured at input current 1 = 1.3 1^^,, CW and room temperature .

PAGE 142

CHAPTER 6 SURFACE SENSITIVE LASER DIODES 6 . 1 Introduction In chapter 5 we have shown that p'^-cap layer thickness has great effects on the thin p-clad laser performance. Especially, when the p*-cap layer thickness is thick (200 nm) , a significant part of the lasing mode could penetrate into the p*-cap layer and make the laser lose the optical confinement and behave abnormally. Therefore, with this thin p-clad laser with thick p'^-cap layer configuration, the interactions between the lasing mode and the thin film material on the top of the p'^-cap layer should be stronger than those of similar lasers with thin p*-cap layer (100 nm) . This is because lasing mode confinement of the thin p-clad laser with 100 nm p^-cap layer is much larger than that of laser with 200 nm p*-cap layer. Additionally, we have also shown that thin p-clad laser performance could be greatly improved when the p^-cap layer thickness is decreased below ~170 nm [Wu96a] . Based on these results, the combinations of using the 100 nm p-cap layer as the electron pumped section 134

PAGE 143

135 and the 200 nm p-cap layer as the surface sensitive section into one laser structure could make it desirable to fabricate a ^^surface sensitive" diode laser [Wu96b] . With this hybrid laser structure, the application of the semiconductor diode laser as a "sensor" could be possible. The laser structure and the main design principle of the "surface sensitive" device are presented in Section II. In Section III, the details of the fabrication process of the SSLD are described. The device characterizations of the SSLD are outlined in Section IV. 6.2 Device Structure and Theoretical Calculations The SSLD structure used in this study is shown in Figure 6.1, where the device is designed by inserting a thick p*-cap section (t^ap nm) into a thin p^-cap layer laser (t^^p ~100 nm) in which the "thick" p*-cap layer is used as the surface sensitive section and the "thin" p^-cap layer as the electronpumped section. A thin film material with a variable absorbing property at the lasing wavelength and t^^^^ thickness is deposited on the top of the sensor section as a sensor medium for the diode laser. The main operation principle of this laser structure to be used as a surface sensitive laser (SSLD) is to use shiny Au as the p-contact for the electronpumped sections and change the absorption property of the deposited film through a chemical reaction between the

PAGE 144

136 absorbing film and the material to be sensed, which could lead to changes of diode laser performance. To check the effects of the absorbing material on the top of the sensor section, we have calculated the modal loss of the sensor section as a function of material loss a^^^^ of the absorbing film at various absorbing film thickness tj^^^. Figure 6.2 shows the calculated dependence of modal loss on C'fiini St various film thickness: tjii„= 10 nm, 50 nm and 100 nm. As can be seen from this plot, the modal loss of the sensor section is a linear function of O.^^^^ and greatly depends on tfiimWhen tfj^„ is less than 50 nm and CCj^in is constant, modal loss increases quickly with tfj^„ and becomes saturated when tfiim is thicker than 100 nm. Besides the effects caused by OL^i^^^, we are interested in checking the influence of the p*-cap layer thickness t^^p on the modal loss of the sensor section. The variations of the modal loss of the sensor section are then calculated at various t^^p and shown in Figure 6.3. In the computations, a tfii„=100 nm and various a^^^^ of 0.2x10* cm"^, l.OxlO" cm"\ 2.5x10^ cm'^ 5x10^ cm'^ are assumed. From this figure, we can find that strong interactions between the lasing mode and the absorbing film only occur when t^^^^ is larger than -150 nm, in addition, a^^^^ has to be larger than 10000 cm"^ to generate significant effects. The peak interactions occur around t =

PAGE 145

180 nm. At constant t^^p= 200 nm, the modal loss is -86 cm' when ttf^i^ is at lO" cm"^ and becomes as high as -190 cm"^ when ttfii^ is changed to 2.5xl0\ In order to understand the threshold behavior of the SSLD device, a Fabry-Perot model shown in Figure 6.4 is used to model the surface sensitive laser diode performance. In this plot, the laser cavity length L is devided into three sections: two pumped sections (section 1 and section 3), L^^ and Lj,* and one unpumped section i.e. sensor section (section 2) L^. The associated real part and imaginary part of the refractive index value of section i (i=l,2,3) is expressed as n^ and n/ respectively. In order to simplify the calculation, we assume a normal incidence case of the light intensity inside the laser cavity. The power reflectivity and power transmission coefficient when light intensity is passing from region i to region j is denoted as R^^ and T^^ respectively, where Rij=Rji and Tij=Tji=l-Ri^ . R^ and R3 are the power reflectivity of the laser facets (R^=R3-0,32 in this case) . and represent the optical gain and modal loss of section i (i=l,2,3) . At first, a lasing mode with power intensity I^ propagating from z=0 in the section 1 of laser cavity is assumed. After the combined effects of optical gain and modal loss in the cavity, part of the power intensity in section 1 is transmitted into section 2. At z=L^ of the section 2, the power intensity is l^. Similarly, part of the

PAGE 146

138 I power intensity in section 2 is transmitted to the section 3 after passing through the sensor section. At z=Lj+L2 the section 3, the power intensity becomes I3. At z=L^+L2+L3 in this section, part of the power intensity is reflected and becomes I4. After propagating L3 cavity length in the negative z direction, part of the power intensity is transmitted into the section 2 and becomes I5 at z=L-L3 in the section 2. At z=L-L3-L2 in section 1, the transmitted power intensity from section 2 is Ig. After propagating the cavity length in section 1, part of the power intensity would be reflected from the laser facet at z=L-L;^-L2-L3 in section 1 and becomes I7. The associated power intensity 1^ (j=l to 7) can be expressed as the following equations. 12 = Ii * e'^'i'^^'^i * (6.2.1) 13 = I2 * e"'2-«2)L2 * (6.2.2) I, = I3 * e'=3~"^''^ * R3 (6.2.3) I5 = I, * e'^^""3'^3 * T32 (6.2.4) = I3 * e'=2-a2)L2 * T21 (6.2.5) 1, = 1, * e'=^-"^"^ * (6.2.6) After a simple manipulation from equation (6.2.1) to equation (6.2.6) and assume T^^=Tj^=T2 (i,j=l,2,3), one can obtain the relationship between I, and as shown in equation (6.2.7) .

PAGE 147

139 _ * g2((Gi-ai)Li+(G2-a2)L2+t=3-«3'L3] * T2'' * * R3 (6.2.7) Equation (6.2.7) can be further simplified, if the pumped section is made of the same t^^p and G^=G^, a^=a^. At the laser oscillating condition, I7 is equal to . Therefore, one can obtain a closed form equation relating the optical gain with a^, L2 and Lp^„p. G, = a, + {[2a,L, + ln(— ^)]/(2Lp,„p)} (6.2.8) where Lp^^^p is the total pumped cavity length; Lpjj„p=L^+L3 In deriving equation (6.2.8), only the transmitted parts of the power intensity at the junction between section 1 and section 2, and the junction between section 2 and section 3 are considered, the reflected parts are ignored. This is because the power reflectivity R^j (i,j = l,2,3) at the junction is very small and can be negligible (this will be shown in the following) . At threshold condition, the optical gain should be equal to threshold mode gain G^^. Consequently, the threshold mode gain G^^, of the SSLD can be expressed by the following equation: Gth = G, = a, + {[2a,L, + In (— ^) ] / (2Lp,„p) } (6.2.9) T2 R1R3

PAGE 148

140 As can be seen from equation (6.2.9), the threshold gain G^^ depends on both the SSLD geometrical parameters (Lp^^^p and L2) and the modal loss in the pumped section as well as the sensor section. By using equation (6.2.9), one can get the G^^ value required for the SSLD with various designed structure. For example, we select the pumped section Lp^^p=850 \Xm, sensor section L^^^^^^=150 [Lm and t^3p=200 nm. The complex effective refractive index values of the sensor section and the pumped section at 950 nm lasing wavelength are calculated as n(sensor)= (n^, n/ ) = (3.32, j0.0027), n(pump)= (n^, n/) = (3.263, jO. 000023), where we assume the sensor section is covered with tj^^^=100 nm film (afj^^=5xl0'' cm'^) and the pumped section with shiny Au . Under this condition, we can estimate the power transmittance T^2=0. 999925 ~1 . Therefore, equation (6.2.9) can be reduced as: Gth = Gi = ai + { [2a2L2 + In (-^) ] / (2Lp^„p) } (6.2.10) By substituting the calculated results of Figure 6.2 into equation (6.2.10), we can obtain the dependence of threshold mode gain G^^ on the variations of O.^^-^^ as shown in Figure 6.5, in which the sensor section L3g„^g^=150 \lm and a modal loss a^=8 cm"^ [Wu94b] in the pumped sections are assumed. In this plot.

PAGE 149

141 we can find that G^^ increases monolithically with CL^^^^ and the increasing rate becomes saturated as t^^^^ is thicker than 100 nm. Figure 6.6 shows the calculated variations of G^^ as a function of t^^p by substituting the results of Figure 6.3 into equation (6.2.10), where l^^^^^^^l^O |lm and tfii„=100 nm are assumed. As can be seen from this plot, at t^3p=200 nm, G^^ could be changed from 34 cm"^ to 8 6 cm"^ when a^i^^ is increased from ajii„=lxlO* cm'^ to (X^^^^ =5x10^ cm'^ . In addition to the calculated effects caused by a^^^^ and t^^p, we are interested in understanding the variations of G^^, at different pumped cavity length Lp^^p. Figure 6.7 shows the dependence of G^^ on the various pumped cavity length Lp^^p, where a t^3p=200 nm of the sensor section and a tfii„=100 nm with different Clfii„ values are considered. From this figure, with OL^^^^ < 10^ cm"^, we can find that threshold mode gain G^j, increases quickly when Lp^,^ is shorter than 400 )lm. However, for larger CXfj^j^^, value (CX^j^^^^ > 2.5x10^ cm'^) , the calculated G^^ increases significantly as ^pump is decreased below 800 }im. At Lp^^p=850 \lm, G^^ could be changed from 28 cm"^ to 87 cm"^ when (Xfii„ is increased from 0.5x10'' cm'^ to 5.0x10" cm'\ This indicates that threshold mode gain required for lasers to reach threshold condition could be very sensitive to the material loss OL^^^^ of the absorbing film. Consequently, it is possible for us to fabricate a

PAGE 150

surface sensitive laser diode (SSLD) by using the proposed device structure as shown in Figure 6.1. 6 . 3 Device Fabrications To fabricate surface sensitive laser diodes, 150 |i.m stripe width on 1000 |lm centers are first defined by using the standard photolithography process as the "sensor stripes". Pulsed anodic oxidation technique is then used to etch away part of the p''-cap layer on the outside sensor stripe region. Usually, the remaining p*-cap layer thickness is -100 nm by controlling the oxidation time at ~2 . 5 minutes. The remaining native oxide is removed by using a MIF-312 developer solution. After the first oxidation step, 100 \Lm stripe width on 500 |J.m centers are defined as the "pumped stripes" with the stripe direction perpendicular to the sensor stripes. The outside pumped stripe region is oxidized by the second anodic oxidation step to remove the remaining p^-GaAs cap layer for reducing the possible current spreading effects. At this step, the sensor section and the pumped section are then generated with the associated p''-cap layer thickness as t^^p ( sensor ) = 200 nm, tj,3p(pump)= ~100 nm. After this step, wafer is cleaned and covered with new photoresist to protect the p^'-cap layer surface during the lapping process. Following the lapping step, a metallurgy of Ge(20 nm)/Au(40 nm)/Ni(25 nm)/Au(50 nm) is evaporated sequentially on the substrate side and annealed at 430 C for ~5' to obtain

PAGE 151

143 a n-type ohmic contact. To deposit the Au contact on the p^cap layer on the pumped stripe, Au-plating technique is used as described in Chapter 4. Before this process, the sensor stripe is protected by new photoresist. Lasers with total cavity length L=1000 |lm (L=Lp^„p+L,^„3„^) , 500 |lm wide is cleaved and soldered on the In-coated copper blocks and characterized. Figure 6.8 (a) and (b) show the top view picture and the schematic diagram of the finished surface sensitive laser diode. As can be seen from the device picture, the sensor section is well defined in the central portion of the device and makes it easy for depositing the absorbing material on this sensor section. 6 . 4 Device Characterizations To check the uniformity of the surface sensitive laser diode, 20 devices are prepared and measured. Figure 6.9 shows the measured pulse P-I characteristics of these 20 SSLDs at 2 |isec . pulse width and 1000 Hz repetition rate. The lasing wavelengths of these 20 samples are ranged from 941 nm to 943 nm, which are ~10 nm shorter than the regular lasers without sensor section. This could be due to band filling effects [Chin88] caused by the fact that more gain are required for these lasers to reach threshold condition. The measured results indicate good uniformity of device performance and also confirm the feasibility of fabricating lasers by using this partially oxidized technique.

PAGE 152

To know if the diode laser performance is sensitive to the material placed on the top of the sensor section, several material types are used for the experiments. Figure 6.10 shows the variations of pulse P-I characteristics for sample #1 measured before and after the sensor section is covered with dye 1 material which has smaller absorption effects on the lasing wavelength. As can be seen from this figure, the threshold current almost remains unchanged. At fixed current, only a small decrease of output power is obtained after applying the dye 1 material. In the second experiment, a dye 2 material with a strong absorption effects on the lasing wavelength is placed on the partial sensor section of sample #2. Figure 6.11 is the variations of the P-I characteristics of this sample measured without and with this dye 2 thin film. It is very interesting to find that this dye 2 material has a significant effects on the laser threshold performance. Before the application of this dye film, threshold current is I^(,= 200 mA. After the application of dye 2 film, 1^^ is increased to 1^^,= 280 mA, i.e. 40% increases. In addition, at fixed current output power changes dramatically, for example, at 1=300 mA output power changes from P (w/o dye 2 film) =37 mW/facet to P (w/dye 2 film)= 11 mW/facet. These results indicate that the diode laser behaviors are very sensitive to this dye 2 material. In the third experiment, sample #3 is measured before and after the dye 2 material is placed on the whole sensor section. As shown in Figure 6.11, the measured P-I characteristics change more significantly than that of

PAGE 153

145 sample #2. Threshold current is almost two times increased from Ith=200 mA to 390 mA after the sensor section is wholly covered with dye 2 material. At fixed current 1=300 mA, laser changes from lasing with output power p=37 mW per facet to non-lasing behaviors. So far we have proved that laser performance is dependent on the type of material placing on the sensor section, the next step is to check the effects from the material to be detected. Since the absorption coefficient of the dye 2 material can be somewhat changed with NHj material, i.e. sensitive to the NHj material, we are interested in understanding the effects of NHj vapor on the laser performance. To check this point, sample #3 with dye 2 film covered on the sensor section is passing through a solution made of 1 part NHj + 200 parts H^O for ~2 seconds and P-I characteristics measured. From the measured results shown in Figure 6.13, it is noticed that threshold current is increased from 1^^=390 mA to 1^^=440 mA and output power (at 1=500 mA) decreased from P=26 mW/facet to P=18 mW/facet after this experiment. The above experiments clearly point out the feasibility of using this unique laser structure as a surface sensitive device. In order to see if diode laser performance could be recovered by removing the material on the sensor sector, the same sample #3 is dipped in Methanol solution for ~30 seconds and P-I performance characterized. The measured results are shown in Figure 6.13. It is interesting to note that diode laser performance is reversible with threshold current

PAGE 154

146 changed from I^f,=440 mA to I^f,=250 mA. As compared with the results measured before placing dye 2 material, there is a -50 mA increase of threshold current. This could be due to the material is not thoroughly removed from the sensor section or other unknown reason. Nevertheless, we have shown the possibility of fabricating a "surface sensitive" laser diode with reversible behaviors from this hybrid laser structure which may become useful as a "sensor" in some environment condition. In summary, a surface sensitive laser diode structure is proposed and fabricated successfully. The basic device operating mechanism discussed in this chapter can be extended to the other material system and could be possible to explore a new application of semiconductor diode lasers for use as a "sensor" .

PAGE 155

147 ^ shiny Au contact absorbing film 1 t cap ~ 100 nm i shiny Au contact p'''-GaAs cap layer p-Al^Ga^.zAs: 25 nm (z=0.6-0.05) p-clad layer Alg gGag 4AS: 250 nm p-Al^Ga^.^As: 200 nm (x=0.3-0.6) GaAs: 7 nm m.\^ '^'<-^'-^'' "^mmmm^^'m^ ' smmi^mm^iiX_ . [n^Gai.vAs, SQW (y~0.15: 8 nm) GaAs: 7 nm n-Al^Ga^.^As: 200 nm (x=0.6-0.3) n-clad layer AIq gGao 4AS: 1400 nm n"^-GaAs substrate ohmic contact : Ge/Au/Ni/Au Figure 6.1 Surface sensitive laser diode structure used in this chapter.

PAGE 156

148 Figure 6.2 The calculated dependence of modal loss of the sensor section (t^3p=200 nm) on the material loss OL^^ of the absorbing film shown in Figure 6.1, where different film thickness are considered.

PAGE 157

149 500 400 300 (A ^ 200 IO § 100 T — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — r Thin p-clad InGaAs SQW SSLD a,,., = 5.0x10'* cm' ^ film w/o oxide 2.5x10'* cm'^ l.OxlO"* cm'^ 0.2x10'* cm'^ ^ 0 100 120 140 160 180 200 220 p'*^-cap layer thickness, t (nm) * * csqp Figure 6.3 The calculated variations of modal loss of the sensor section (shown in Figure 6.1) as a function of p''-cap layer thickness t^^p, where a absorbing film thickness tjj^^„=100 nm and various values of material loss OCj-^^^ are assumed.

PAGE 158

150 Total cavity length: L=L]^+L2+L3 1^32' T32 z=0 Figure 6.4 The schematic diagram of Fabry-Perot model used to derive the threshold behaviors of surface sensitive laser diode.

PAGE 159

151 g 200 O 150 O g 2 o 100 2 50 0 0 -I — I — I — I — I — I — I — I — I — I — T"* — ' — ' — I — ' — ' ' I Thin p-clad InGaAs SQW SSLD Vilm • 10 nm A 50 nm O 100 nm L pump 850 ^m, L sensor = 150 |im _l I I I I I l_ _l I I I I I I I I I l_ 6 8 10 12 Material loss, a_ (xlO^ cm"^) ' film Figure 6.5 The calculated threshold mode gain G^h of SSLD as a function of the material loss Ctj^^^ of the absorbing film. It is assumed that total pump section Lpj^p=850 jJjn are covered with gold and sensor section L3gng„j.=150 jJjn placed with different film thickness.

PAGE 160

152 6 120 ^ 100 a tifl 80 o g 2 CD u 60 40 20 -1 — I — I — I — I — I — I — I — I — I — I — I — I — I — I — l—i — I — r— I — I— r— r Thin p-clad InGaAs SQW SSLD * L =850 |im pump ^ L =150 um sensor ^ **tf, =100 nm 2.5x10'* cm'-^ 1.0x10^ cm"-^ 0.2x10'* cm'-^ I I I I I C 100 120 140 160 180 200 220 p*-cap layer thickness, t^^ (nm) Figure 6.6 The calculated variations of threshold mode gain G^h of SSLD (laser structure shown in Figure 6.1) as a function of p^-cap layer thickness. It is assumed that total pump section Lp,j_,p=850 }i.m are covered with gold and sensor section Lsensor^-'-^'^ M-"^ placed with various CL^^^^ values of absorbing film.

PAGE 161

153 Figure 6.7 The calculated dependence of threshold mode gain G^h of SSLD (laser structure shown in Figure 6.1) as a function of total pump cavity length Lp^^p. It is assumed that total pump section Lp^^p=850 |lm are covered with gold and sensor section L3g„3^^=150 jlm placed with 100 nm absorbing film with various Ct^^^^.

PAGE 162

154 ( Figure 6.8 Top view of the finished surface sensitive laser diode (SSLD) (a) picture (b) schematic diagram.

PAGE 163

155 100 I I I ' ' I ' ' ' ' I ' ' o a O a o 80 60 40 20 0 T — r-i — r— I — r~r I ' ' ' ' I I I I I Thin p<;lad InGaAs SQWSSLD * w/o absorbing film L = 850 urn pump L = 150 [im sensor J I ' I ' I 0 100 200 300 400 500 600 Input current (mA) Figure 6.9 Measured pulse P-I characteristics of 20 SSLD samples when operated at 2 microseconds pulse width and 1000 Hz repetition rate, room temperature.

PAGE 164

156 150 200 250 300 350 400 450 500 Input current (mA) Figure 6.10 Measured pulse P-I characteristics of SSLD sample #1 before and after placing dye 1 material on the sensor section.

PAGE 165

157 0 100 200 300 400 500 600 Input current (mA) Figure 6.11 The variations of measured P-I characteristics of SSLD sample #2 before and after placing with dye 2 material on part of the sensor section.

PAGE 166

158 g o O 70 60 50 t40 30 It 20 10 0 0 t I I I I I I I I I I I I I I I I I I ' ' ' ' I ' ' ' ' . Thin p-clad InGaAs SQWSSLD w/ dye 2 film (wholely covered) w/o dye2 film ^ 100 200 300 400 500 600 Input current (mA) Figure 6.12 The measured variations of P-I characteristics of SSLD sample #3 before and after placing dye 2 material on the whole sensor section.

PAGE 167

159 0 100 200 300 400 500 600 Input current (mA) Figure 6.13 The measured variations of P-I characteristics of SSLD sample #3 with dye 2 material on the sensor section after being processed in different conditions: (1) exposed to NH3 solution ~2 seconds (2) after step (1) then rinsed in Methanol to remove dye 2 film on the sensor section.

PAGE 168

CHAPTER 7 SUMMARY AND RECOMMENDATION 7 . 1 Summary The p-contact reflectivity effects on the device performance has been demonstrated in thin (250 nm) p-clad, 10 nm InGaAs quantum well lasers. Decreasing the reflectivity of the p-contact metal increases optical mode loss, which increases threshold current, decreases slope efficiency and shifts the emission wavelength. The more than 50 nm wavelength difference between the thin p-clad diode lasers with different p-contact metallurgy can be explained quantitatively by superimposing the threshold gain required for lasing in each case with the corresponding spectral gain curve calculated using standard QW laser theory. To avoid the extra mode loss and obtain a well-behaved thin p-clad diode laser, the p-contact metal can not be annealed. The CW life time tests show that 50 |lm stripe-width, Au-contact, thin pclad diode lasers with epi-side up package can live more than 1000 hrs . when continuously operate with 100 mW output power at room temperature. 160

PAGE 169

Besides the contact reflectivity effects on the diode laser performance, one of the most important feature of the thin p-clad laser structure is that the lateral refractive index step generated in low-ridge diode lasers is sufficient to provide low loss and strong lateral waveguiding. As a result, neither the unusual threshold behaviors nor the wavelength jumps are observed for the shallow ridge, thin pclad laser when stripe width is changed from 50 Hm to 5 |im. In addition, the 5 |J.m stripe devices are shown to be capable of stable, single lateral mode CW lasing with less than 10 % broadening up to total output power levels of about 70 mW. On the contrary, for the thick p-clad (1300 nm) diode laser, deep etching for obtaining ridge height larger than 1000 nm is required to provide the sufficient waveguiding effects for narrow stripe lasers operating with stable single mode output . By controlling the p-contact reflectivity, a two-stripe thin p-clad diode laser emitting dual wavelengths was successfully demonstrated. The fabrication process of this dual wavelength diode laser is much simpler and more reliable than those approaches studied before. Based on this experiment results, one can design a thin p-clad diode laser with several emission energy levels inside the single quantum well and select the suitable contact metal to control the emission wavelength to obtain a monolithic multiple wavelength emission diode laser. Additionally, thermal resistance is found to play an important role for diode

PAGE 170

162 lasers operating at high temperature in CW regime. The maximum output power of CW laser performance can be greatly improved by reducing its thermal resistance through the deposition of thick Au layer as the heat spreader. In addition to the contact reflectivity, another important parameter in determining thin p-clad diode laser performance is the p*-GaAs cap layer thickness. The increasing of p-cap thickness could result in the decrease of optical confinement of lasing mode and induce large modal loss. For the thin p-clad (250 nm) laser structure used in this study, long time lasing delay and abnormal high threshold behaviors are observed for both wide stripe and narrow stripe gainguided lasers when p-cap layer thickness is as thick as 200 nm. The lasing delay in the wide stripe lasers can be eliminated and diode laser performance improved dramatically when p-cap layer thickness is decreased below 170 nm. This long delay is attributed to the time it takes for the active region to heat to the point where net mode gain exceeds mirror loss by the increase of overlap between the transverse mode profile (perpendicular to the QW plane) and the QW material gain as well as the decrease of the absorbing loss in the gold contact layer. In contrast to the strong lateral waveguiding effects obtained in narrow stripe, shallow-ridge, thin p-clad lasers fabricated with 100 nm p-cap layer, long time lasing delay and Q-switching lasing behaviors are observed for narrow stripe lasers with 300 nm ridge height and an effective refractive index step of 3.5x10"^ when the

PAGE 171

163 stripe width is w=6 |lm and w=3.5, 2.5 |lm respectively. At 1.5 Hm stripe width, strong lateral waveguiding effects are obtained and become sufficient to somehow improve the transverse waveguiding which makes laser operate in a low threshold current regime with CW differential quantum efficiency 50 % . By combining 100 nm and 200 nm p*-cap layer structures into one laser and removing the gold layer from the 200 nm section, laser output power at fixed current becomes dependent on the type of material placed on the 200 nm section. Experiments of using these "surface sensitive laser diodes" to the possible application for use as a "sensor" have been successfully demonstrated. 7 . 2 Recommendation for Future Study Several areas are needed to do the further study, most important among them are : { 1 ) The improvement of adhesion between metal contact and p''GaAs cap layer. The adhesion quality between Au contact and p-cap layer is very crucial in determining thin p-clad lasers performance. The adhesion quality not only depends on the clean process before the deposition of p-metal but also depends on the metal itself. For the GaAs material, Au shows poor adhesion property as compared to the other metal such as

PAGE 172

154 nickel (Ni) and chromium (Cr) . However, these materials also show very low reflectivity at the lasing wavelength and become lossy metal for thin p-clad diode lasers. On the other hand, the modal loss induced by the low reflective metal could be dependent on the metal thickness. In the calculated results shown in Figure 7.1, we can see that modal loss for both metal type is in linear proportion to the metal thickness upto 100 nm metal thickness and then becomes gradually saturated when metal thickness is beyond 200 nm. If one can control the metal thickness below 20 A, the resultant modal loss could be less than 10 cm'S which could be acceptable in most diode laser performance. The important issue for this method is how to control the uniformity of metal thickness during the deposition process to avoid possible increase of scattering loss due to the nonunif ormity of metal thickness. Another method to improve the adhesion property is to use Au-plating technique instead of Au evaporation process. Both methods need to be further studied to show the uniformity of laser performance. (2) Optimize and extend the surface sensitive device operation mechanism to the other semiconductor laser material system Thin p-clad diode laser performance has been shown greatly dependent on the p-cap layer thickness. Based on these results, a hybrid laser diode structure with surface

PAGE 173

165 sensitive capability has been successfully demonstrated in GaAs/InGaAs material system by selectively controlling the pcap layer thickness used as either the electron-pumped section or the sensor section. However, several parameters need to be optimized to obtain even better device performance and reliability, such as the optimal cavity length for either pumped section Lp^p or sensor section L^^^^^^, facet coating on both facets to protect lasers from being damaged by the working environment. On the other hand, the calculated results shown in Figure 7.2 indicate that it could be possible to design a SSLD with thin native oxide grown on the top of sensor section. With this extra oxide grown process, both the experimental and theoretical work need to be further studied. In addition, by using similar device operating mechanism combined with the semiconductor quantum well characteristics, the operating mechanism of SSLD can be extended to other semiconductor material systems with suitable operating wavelength for use as a sensor. (3) Distributed Feedback (DFB) Diode Laser: Gain coupling DFB lasers with better performance have been theoretically studied and realized. The usual way of realizing gain coupling DFB diode lasers is to grow an absorbing layer close to the active region and followed by the grating formation procedure and the regrowth process, which makes the whole fabrication process complicated. By

PAGE 174

166 using either the contact reflectivity effects or controlling the p-cap layer thickness, one can control the modal loss of thin p-clad diode laser. As a result, the fabrication process of gain-coupled DFB lasers should become much simpler than those reported before. In addition, from the optimization of diode laser structure, single spatial mode operation with linear P-I profiles to much higher CW powers should be possible. This feature combined with the flexibility of fabricating different diffraction grating types in thin pclad material { no regrowth required ) could lead to the development of higher performance gain-coupled DFB lasers.

PAGE 175

167 120 I T — I — I — I — I — I — I — I — I — r a (A (A O O 100 80 60 40 20 0 Thin p-clad InGaAs SQW LD 0 Cr Ni * X^960 nm I I I I I I 100 200 300 400 500 Thickness (A) Figure 7.1 Calculated dependence of modal loss of thin pclad InGaAs SQW laser as a function of different metal thickness: chromium (Cr) and nickel (Ni) .

PAGE 176

168 120 g cn 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 Thin p-clad InGaAs SQW LD * Lossy film =10 nm p-clad=250 nm t =25nm OK I I I I I I ' ' I I ' 0 100 120 140 160 180 200 220 240 p*-GaAs thickness, t (nm) ^ cap Figure 7.2 Calculated variations of modal loss as a function of p-cap layer thickness when two different lossy films with 10 nm thickness are covered on the top of p-cap layer, where the refractive indexes of the oxide and the lossy film are 1.8 and 1.5 respectively.

PAGE 177

REFERENCES [Adac85] S. Adachi, " GaAs, AlAs, and AlxGai-xAs : material parameters for use in research and device applications," J. Appl. Phys., vol. 58, pp. R1-R29, 1985. [Adam73] M. J. Adams, S. Griidorfer, G. H. B. Thompson, C. F. L. Davies, and D. Mistry, "Time delay and Qswitching in homostructure and heterostructure injection lasers," IEEE J. Quantum Electron., vol. QE-9, pp. 328-337, 1973. [Agra84] Govind P. Agrawal "Lateral analysis of quasi-indexguided injection lasers: transition from gain to index guiding," IEEE J. Lightwave Technol . vol. LT2, pp. 537-543, 1984. [Aiki76] K. Aiki, M. Nakamura, and J. Umeda, "Frequency multiplexing light-source with monolithically integrated distributed-f eedback diode lasers," Appl. Phys. Lett., vol. 29, pp. 506-508, 1976. [Bahf90] A. Behfar-Rad, J. R. Shealy, S. R. Chinn, and S. S. Wong, "Effect of cladding layer thickness on the performance of GaAs-AlGaAs graded index separate confinement heterostructure single quantum-well lasers," IEEE J. Quantum Electron., vol.26, pp. 14761480, 1990. [Bast88] Gerald Bastard ed., "Wave mechanics applied to semiconductor heterostructures, " Les Ulis Cedex, France :Halsted Press, 1988, Chapter 1. [Beer91a] K. J. Beernink, J. J. Alwan, and J. J. Coleman, "Wavelength switching in narrow oxide stripe InGaAsGaAs-AlGaAs strained-layer quantum well heterostructure lasers," Appl. Phys. Lett. vol. 58, pp. 2076-2078, 1991. [Beer91b] K. J. Beernink, J. J. Alwan, and J. J. Coleman, "Antiguiding in narrow stripe gain-guided InGaAsGaAs strained-layer lasers," J. Appl. Phys., vol.69, pp. 56-60, 1991. 169

PAGE 178

170 [Boua82] N. Bouadma, J. C. Bouley, and J. Riou, "Dual wavelength (GaAs)Al laser," Electron. Lett. vol. 18, pp871-873, 1982. [Bour90] D. P. Bour, N. A. Dinkel, D. B Gilbert, K. B. Fabian and M. G. Harvey, "980 nm diode laser for pumping Er^''"-doped fiber amplifiers," IEEE Photon. Technol. Lett., vol. 2, pp. 153-155, 1990. [Butl75] J.K Butler, H. Kressel and I Ladany, "Internal optical loss in very thin CW hetero junction laser diodes," IEEE, Quantum Electronics, vol. QE-11, pp. 402-408, 1975. [Case75] H. C. Casey Jr. and M. B. Panish, "Influence of AlxGal-xAs layer thickness on threshold current density and differential quantum efficiency for GaAs-AlxGal-xAs DH lasers," J. Appl. Phys . , vol. 46, pp. 1393-1395, 1975. [Chin88] S. R. Chinn, P. S. Zory and A. R. Reisinger, "A model for GRINSCH SQW diode lasers," IEEE Quantum Electron, vol. 24 pp. 2191-2214, 1988. [Cole93] James J. Coleman "Strained layer quantum well heterostructure lasers," in Quantum Well lasers, P. S. Zory ed., New York: Academic, 1993. [Corz93] Scott W. Corzine, Ran-Hong Yan and Larry A. Coldren "Optical gain in III-V Bulk and Quantum Well Semiconductors," in Quantum Well lasers, P. S. Zory ed.. New York: Academic, 1993. [Deme89] P. L. Demers . et.al. Modeig/II version 1.1, Dallas, TX, Southern Methodist University: Sep. 1989. [Dutt81] N. K. Dutta, "Calculated temperature dependence of threshold current of GaAs-Al^Ga^^.^As double heterostructure lasers," J. Appl. Phys., vol. 52, pp. 70-73, 1981. [Dutt83] N. K. Dutta, "Calculation of Auger rates in quantum well structure and its application to InGaAsP quantum well lasers," J. Appl. Phys., vol. 54, pp. 1236-1245, 1983. [Dutt86] N. K. Dutta, T. Cella, J. L. Zilko, D. A. Ackerman, A. B. Piccirilli, and L. I. Greene, "InGaAsP closely spaced dual wavelength laser," Appl. Phys. Lett., vol. 48, pp. 1725-1726, 1986.

PAGE 179

171 [Dyme72] J. C. Dyment, J. E. Ripper, and J. P. Lee, "Measurement and interpretation of long spontaneous lifetime in double heterostructure lasers," J. Appl. Physic, vol. 43, pp. 452-457, 1972. [Eple90] John E. Epler, D. W. Treat, S. E. Nelson, and Thomas L. Paoli, "Mulltiple-wavelength diode laser superarray," IEEE Quantum Electronics, vol. QE-2 6, pp. 663-668, 1990. [Fisc87] S, E. Fischer, D. Fekete, G. B. Feak, and J. M. Ballantyne, "Ridge waveguide injection laser with GalnAs strained-layer quantum well (X;=l Jim), " Appl. Phys. Lett. vol. 50, pp. 714-716, 1987. [Garb93] D. Z. Garbuzov and V. B. Khalfin, "Single quantum well InGaAsP and AlGaAs lasers: A study of some peculiarities," in Quantum Well lasers, P. S. Zory d.. New York : Academic, 1993. [Garr87] B. Garrett, and J. E. A. Whiteaway, "Self stabilization of the fundamental lateral mode in index-guided semiconductor lasers," lEE Proc, vol. 134, pt. J, No. 1, pp. 11-15, 1987. [Good75] A. R. Goodwin, J. R. Peters, M. Pion, G. H. B. Thompson, and J. E. A. Whiteaway, "Threshold temperature characteristics of double heterostructure Ga^.^Al^As lasers," J. Appl. Phys., vol. 46, pp. 3126-3131, 1975. [Gray72] D. E. Gray ed., American Institute of Physics Handbook, McGraw-Hill, 3rd edition, 1972. [Grov94] M. J. Grove, D. A. Hudson, P. S. Zory, R. J. Dalby, C. M. Harding, and A. Rosenberg, "Pulsed anodic oxides for III-V semiconductor device fabrication, " J. Appl. Phys., vol.76, pp. 587-589,1994. tGrud73] S. Griidorfer and M. J. Adams, "Theoretical considerations of time delays in semiconductor lasers," IEEE J. Quantum Electron., vol. QE-9, pp. 814-819, 1973. [Grud74] S. Griidorfer and M. J. Adams, "New theory of internal Q switching in semiconductor lasers," Electron. Lett., vol. 10, pp. 354-356, 1974. [Guth94] James Guthrie, G. L. Tan, M. Ohkubo, T. Fukushima, Y. Ikegami, T. Ijichi, M. Irikawa, R. S. Mand, and J. M. Xu, "Beam instability in 980-nm power lasers: experiment and analysis," IEEE Photon. Technol. Lett., vol. 6, pp. 1409-1411, 1994.

PAGE 180

172 [Jone75] P. Garel-Jones and John C. Dyment, "Calculations of the continuous-wave lasing range and light-output power for double-heterostructure lasers," IEEE J. Quantum Electron . , vol. QE-11, pp. 408-413, 1975. [Joyc75] W. B. Joyce and R. W. Dixon, "Thermal resistance of heterostructure lasers," J. Appl . Phys . , vol 46, pp , 854-862, 1975. [Lee91] Johnson Lee, C. Shieh,and M. 0. Vassell, "Variation of threshold current with cavity length in strainedlayer InGaAs/GaAs quantum well lasers, " J. Appl. Phys., vol. 69, pp. 1882-1891, 1991. [Luo90] H. J. Luo and P. S. Zory, "Distributed feedback coupling coefficient in diode lasers with metalized gratings," IEEE Photon. Technol . Lett., vol. 2, pp. 614-616, 1990. [Luo92] Y. Luo, H. L. Cao, M. Dobashi, H. Hosomatsu, Y. Nakano, and K. Tada, "Gain-coupled distributed feedback semiconductor lasers with an absorptive conduction-type inverted grating," IEEE Photon. Technol. Lett., vol. 4, pp. 692-695, 1992. [Maco87] S. H. Macomber, J. S. Mott, R. J. Noll, G. M. Gallatin, E. J. Gratrix, S. L. O'Dwyer, and S. A. Lambert, "Surface emitting distributed feedback semiconductor laser," Appl. Phys. Lett. Vol. 51, pp , 472-474, 1987. [Marp64] D. T. F. Marple, "Refractive index of GaAs," J. Appl. Phys., vol. 35, pp . 1241-1242, 1964. [Maws86] L. J. Mawst, M. E. Givens, C. A. Zmudzinski, M. A. Emanuel, and J. J. Coleman, "Near-and far-field observations of transient behavior in pulsed graded barrier quantum well lasers," J. Appl. Phys., vol. 60, pp. 2613-2615, 1986. [Mott89] J. S. Mott and S. H. Macomber, "Two-dimensional surface distributed feedback laser arrays," IEEE, Photo. Technol. Lett. vol. 1, pp. 202-204, 1989. [Nune77] F. D. Nunes, N. B. Patel and J. E. Ripper, "A theory on long time delay and internal Q switching in GaAs junction lasers," IEEE J. Quantum Electron., vol. QE-13, pp. 675-681, 1977. [Pan88] S. H. Pan, H. Shen, Z. Hang, F. H. Pollak, W. Zhuang, Q. Xu, A. P. Roth, R. A. Masut, C. Lacelle, and D. Morris, "Photoref lectance study of narrow-

PAGE 181

173 well strained-layer InxGai-x^s /GaAs coupled multiple-quantum-well structures," Phys . Rev. B, vol. 38, pp. 3375-3382, 1988. [Paol75] Thomas L. Paoli, "A new technique for measuring the thermal impedance of junction lasers," IEEE J. Quantum Electronics, vol. QE-11, pp. 498-503, 1975. [Pate87] N. B. Patel, T. J. S. Mattos, F. C. Prince and A, S. Nunes, "Narrow stripe graded barrier single quantum well lasers-threshold current considerations," IEEE J. Quantum Electron., vol. QE23, pp. 988-991, 1987. [PrinBS] F. C. Prince, N. B. Patel, D. Kasemsel and C. S. Hong, "Long delay time for lasing in very narrow graded barrier single-quantum-well lasers," Electron. Lett., vol. 19, pp. 435-437, 1983. [Prin85] F. C. Prince, T. J. S. Mattos, N. B. Patel, D, Kasemsel and C. S. Hong, "Waveguiding, spectral, and threshold properties of a stripe geometry single quantum well lasers," IEEE J. Quantum Electron., vol. QE-21, pp. 634-639, 1985. [Rast91] A. Rast and A. Zach, "Room temperature cw operation of 1= 1.55mm InGaAsP/InP ITG-DFB-BCRW lasers with contact surface grating," Electron. Lett., vol. 27, pp. 108-109, 1991. [Ripp74] J. E. Ripper and J. A. Rossi, "Delay and Q switching in semiconductor lasers-still an operation question," IEEE J. Quantum Electron., vol. QE-IG, pp. 435-441, 1974. [Sak;a82] S. Sakai, T. Aoki, and M. Umeno, "InGaAsP/InP dual wavelength lasers," Electron. Lett. vol. 18, pp. 1718, 1982. [Scif75] D. R. Scifres, R. D. Burnham, and W. Streifer, "Output coupling and distributed feedback utilizing substrate corrugations in double-heterostructure GaAs lasers," Appl . Phys. Lett., vol. 27, pp. 295297, 1975. [Scif76] D. R. Scifres, W. Streifer, and R. D. Burnham, "Leaky wave room-temperature double heterostructure GaAs:GaAlAs diode laser," Appl. Phys. Lett., vol. 29, pp. 23-25, 1976. [Shan88] Y. Shani, A. Katzir, M. Tacke and H. M. Preier, "Metal clad Pbi-xSnxSe/Pbi-x-yEuySnxSe distributed

PAGE 182

174 feedback lasers," IEEE J. Quantum Electron., vol. QE-24, pp. 2135-2137, 1988. [Shie89] C, Shieh, J, Mantz, H. Lee, D. Ackley, and R. Engelmann, "Anomalous dependence of threshold current on the stripe width in gain-guided strainedlayer InGaAs/GaAs quantum well lasers," Appl . Phys . Lett., vol. 54, pp. 2521-2523, 1989. [Suem75] Y. Suematsu, M. Yamada, and K. Hayashi, "Integrated twin-guide AlGaAs laser with multiheterostructure, " IEEE J. Quantum Electron., vol. QE-11, pp. 457-460, 1975. [Taka78] T. Takagi, "Refractive index of Gai-xInxAs prepared by vapor-phase epitaxy," Japanese J. Appl. Phys., vol. 17, pp. 1813-1817, 1978. [Take90] Tatsuya Takeshita, Masanobu Okayasu, and Shingo Uehara, "High-power operation in 0.98-}im strainedlayer InGaAs-GaAs single-quantum-well ridge waveguidelasers, " IEEE Photon. Technol . Lett., vol. 2, pp. 849-851, 1990. [ThomSO] G.H.B. Thompson ed., "Physics of semiconductor laser devices," New York: Wiley, 1980. [Thom74] G. H. B. Thompson, P. R. Selway, G. D. Henshall and J. E. A. Whiteaway, "Role of optical guiding in critical-temperature behavior, delays and Q switching in single heterostructure GaAs/ (GaAl) As lasers," Electron. Lett., vol. 19, pp. 354-355, 1974. [Tien70] P. K. Tien and R. Ulrich, "Theory of prism-film coupler and thin-film light guide," J. Opt. Soc. Am., vol. 60, pp. 1325-1337, 1970. [Toku87] Y. Tokuda, Y. Abe. T. Matsui, N. Tsukada, and T. Nakayama, "Dual-wavelength emission from a twinstripe single quantum well laser," Appl. Phys. Lett., vo. 52, pp. 1664-1666, 1987. [Unge93] Peter Unger, Gian-Luca Bona, Roland Germann, Peter Roentgen, and David J. Webb, "Low-threshold strained GalnP quantum-well ridge lasers with AlGaAs cladding layers," IEEE J. Quantum Electron., vol. 2 9, pp. 1880-1883, 1993. [Van89] C. G. Van de walle, "Band lineups and deformation potentials in the model-solid theory," Phys. Rev. B, vol. 39, PP1871-1883, 1989.

PAGE 183

175 [Weas87] R.C. Weast, ed., CRC Handbook of Chemistry and Physics, Florida:CRC Press, 68th edition, 1987. [Wu94a] C. H. Wu, P. S. Zory, and M. A. Emanuel, "Large wavelength shifts in thin p-clad InGaAs QW lasers," lEEE/LEOS annual meeting, Boston, Oct. 1994. [Wu94b] C. H. Wu, P. S. Zory, and M. A. Emanuel, "Contact reflectivity effects on thin p-clad InGaAs single quantum-well lasers," IEEE Photo. Technol . Lett., vol.6, pp. 1427-1429, 1994. [Wu95] C. H. Wu, P. S. Zory, and M. A. Emanuel, "Characterization of thin p-clad InGaAs singlequantum-well lasers," IEEE Photo. Technol. Lett., vol.7, pp. 718-720, 1995. [Wu96a] C. H. Wu, P. S. Zory, and M. A. Emanuel, "p^-GaAs cap layer thickness effects on thin p-clad InGaAs single-quantum-well lasers," in preparation. [Wu96b] C. H. Wu, P. S. Zory, and M. A. Emanuel, "InGaAs single-quantum-well surface sensitive diode lasers," in preparation. [Yone73] H. Yonezu, I. Sakuma, K. Kobayashi, T. Kamejima, M. Ueno, and Y. Nannichi, "A GaAs-Al^Ga^.^As double heterostructure planar stripe laser," Japanese J. Appl. Phys., vol. 12, pp. 1585-1592, 1973. [Zory75] P. Zory and L. D. Comerford, "Grating-coupled double-heterostructure AlGaAs diode lasers," IEEE, Quantum Electronics, vol. QE-11, pp. 451-457, 1975. [Zory86] P. S. Zory, A. R. Reisinger R. G. Water, L. J. Mawst, C. A. Zmudzinski, M, A. Emanuel, M. E. Givens, and J. J. Coleman, "Anomalous temperature dependence of threshold for thin quantum well AlGaAs diode lasers" Appl. Phys. Lett. vol. 4 9, pp. 16-18, 1986.

PAGE 184

BIOGRAPHICAL SKETCH Chih-Hung Wu was born in Taiwan in January 1960. He received his B.S. degree in electrical engineering from National Cheng-Kung University, Tainan, Taiwan, in June 1982. After his graduation, he was a communication lieutenant in army for two year of military service. In 1984 he became a graduate student in Institute of Electrical and Computer Engineering of National Cheng-Kung University and received his M.S. degree in 1986 with the thesis titled " The studies of dopants incorporation in GaAs grown by pressure metal organic vapor phase epitaxy." Then he joined the Institute of Nuclear Energy Research (INER) as a assistant researcher in 1986 and was involved in designing and developing the fabrication of high power single heterostructure (SH) diode laser arrays. In 1992, he won a scholarship from INER to start his Ph.D. study in USA. Since August 1993 has been a graduate research assistant at the Photonics Research Laboratory in the Department of Electrical and Computer Engineering of the University of Florida, where he is engaged in research on thin p-clad quantum well lasers and developing fabricating process for CW high power laser array. His research interests include semiconductor material growth and device fabrication process development. 176

PAGE 185

I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Peter S. Zory, Chai/tmai Professor of ElectricaU/ and Computer Engineering I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Philosophy. Gi js/Bosman Professor of Electrical Computer Engineering and I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Arnost Neugroschel Professor of Electrical and Computer Engineering I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Ramakant Srivastava Professor of Electrical and Computer Engineering

PAGE 186

I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Robert M. Park Professor of Materials Science and Engineering This dissertation was submitted to the Graduate Faculty of the College of Engineering and to the Graduate School and was accepted as partial fulfillment of the requirements for the degree of Doctor of Philosophy. May 1996 Winfred M. Phillips Dean, College of Engineering Karen A. Holbrook Dean, Graduate School