Title Page
 Table of Contents
 Anodic oxidation of semiconduc...
 Lasers fabricated with pulsed anodic...
 Analysis of experimental resul...
 Summary and conclusions
 Biographical sketch

Title: On pulsed anodic oxidation and its use in fabricating diode lasers
Full Citation
Permanent Link: http://ufdc.ufl.edu/UF00082370/00001
 Material Information
Title: On pulsed anodic oxidation and its use in fabricating diode lasers
Physical Description: viii, 121 leaves : ill., photos (some col.) ; 29 cm.
Language: English
Creator: Grove, Michael J., 1959-
Publication Date: 1994
Subject: Diodes, Semiconductor   ( lcsh )
Heterostructures   ( lcsh )
Semiconductor lasers   ( lcsh )
Electrical Engineering thesis Ph.D
Dissertations, Academic -- Electrical Engineering -- UF
Genre: bibliography   ( marcgt )
non-fiction   ( marcgt )
Thesis: Thesis (Ph. D.)--University of Florida, 1994.
Bibliography: Includes bibliographical references (leaves 117-120).
Statement of Responsibility: by Michael J. Grove.
General Note: Typescript.
General Note: Vita.
 Record Information
Bibliographic ID: UF00082370
Volume ID: VID00001
Source Institution: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: aleph - 001975596
oclc - 31798905
notis - AKF2426

Table of Contents
    Title Page
        Page i
        Page ii
        Page iii
        Page iv
    Table of Contents
        Page v
        Page vi
        Page vii
        Page viii
        Page 1
        Page 2
        Page 3
        Page 4
        Page 5
        Page 6
        Page 7
        Page 8
        Page 9
        Page 10
        Page 11
    Anodic oxidation of semiconductors
        Page 12
        Page 13
        Page 14
        Page 15
        Page 16
        Page 17
        Page 18
        Page 19
        Page 20
        Page 21
        Page 22
        Page 23
        Page 24
        Page 25
        Page 26
        Page 27
        Page 28
        Page 29
        Page 30
        Page 31
        Page 32
        Page 33
        Page 34
        Page 35
    Lasers fabricated with pulsed anodic oxidation
        Page 36
        Page 37
        Page 38
        Page 39
        Page 40
        Page 41
        Page 42
        Page 43
        Page 44
        Page 45
        Page 46
        Page 47
        Page 48
        Page 49
        Page 50
        Page 51
        Page 52
        Page 53
        Page 54
        Page 55
        Page 56
        Page 57
        Page 58
        Page 59
        Page 60
        Page 61
        Page 62
        Page 63
        Page 64
        Page 65
        Page 66
        Page 67
        Page 68
        Page 69
        Page 70
        Page 71
        Page 72
        Page 73
        Page 74
        Page 75
        Page 76
        Page 77
        Page 78
        Page 79
    Analysis of experimental results
        Page 80
        Page 81
        Page 82
        Page 83
        Page 84
        Page 85
        Page 86
        Page 87
        Page 88
        Page 89
        Page 90
        Page 91
        Page 92
        Page 93
        Page 94
        Page 95
        Page 96
        Page 97
        Page 98
        Page 99
        Page 100
        Page 101
        Page 102
        Page 103
        Page 104
        Page 105
        Page 106
        Page 107
        Page 108
        Page 109
        Page 110
        Page 111
        Page 112
    Summary and conclusions
        Page 113
        Page 114
        Page 115
        Page 116
        Page 117
        Page 118
        Page 119
        Page 120
    Biographical sketch
        Page 121
        Page 122
        Page 123
Full Text










If this work is judged to be a success in some small fashion, I would be

remiss not to thank all the people who have helped me throughout my time at

the University of Florida.

First and foremost, I would like to thank our Maker for the many gifts

he has bestowed on my family and myself. While it may no longer be

politically correct to acknowledge His/Her existence, it is incomprehensible to

me that men and women of science can observe the boundless intricacies and

symmetries of the physical world and still not acknowledge a prime creative

force behind it all.

On a more pragmatic side, I would like to express my deepest thanks to

my advisor, Dr. Peter S. Zory. His technical excellence and real world

experience are a combination that made working for him both enlightening

and rewarding.

I would also like to thank the members of my committee; Dr. Gijs

Bosman, Dr. Sheng S. Li, Dr. Frederik A. Lindholm, and Dr. Robert M. Park.

In addition to their guidance on this dissertation, I have had the privilege of

taking a class from each of these professors. Their contribution to my

understanding of the field of semiconductor devices has been significant.

Special thanks also go to Dr. Mark Orazem for. his helpful discussions on


Although too numerous to mention, I would like to thank all the

professors that I have taken classes from, both here and at West Point. I

would like to mention two by name, Dr. Luis Figueroa and Dr. Ramakant

Srivastava. Their enthusiasm for science and concern for their students merit

special thanks.

I would like to thank the departmental -staff, particularly Mr. Bob

McClain, Mr. James Chamblee, Mr. Allan Herrlinger, Mr. Tim Vaught, Ms.

Gretta Sbrocco, and Ms. Sharon Williams. I deeply appreciate all the things

that are done every day to make a department run and allow us to do our


I would like to thank all the students that I have worked with or

studied with while at the university. Particular thanks go to Doug Hudson,

Hyoun Soo Kim, Sang Kook Han, Young Soh Park, Chi-Lin Young, Chih-Hung

Wu, Chris Hussell, and Craig Largent.

I would like to acknowledge the Department of the Army. Although

they did not provide technical guidance, they did pay the bills and for that I

am sincerely grateful. I hope my years of future service will justify the

confidence that they have placed in me.

Last, but certainly not least, I want to thank my family. Special

thanks go to my mother and father for showing me the importance of a good

education. Thanks go to my wife Linda and my sons, Michael and Bobby.

Often the hours spent on this project were hours I should have spent as a

husband and a father. Through their sacrifice and patience they are

responsible for whatever success this work contains.



ACKNOWLEDGEMENTS.................................................. ...............iii

A B STRA CT ...................................................................................................... vii


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

2.ANODIC OXIDATION OF SEMICONDUCTORS........................ 12
2.1. Introduction.................................. ...................12
2.2. Anodic Oxidation................................................................ 13
2.3. Pulsed Anodic Oxidation.......................... ............. 22
2.3.1. Theory of Pulsed Anodic Oxidation................... 22
2.3.2. Results in Bulk and Diode Laser Material.......... 27

3.1. Introduction......................................... ................... 36
3.2. Device Fabrication ......................................... .......... ... 37
3.3. Ridge Evolution.................................. ............... ............... 43
3.4. Threshold Current .......................................... .......... ... 51
3.5. Slope Efficiency .................................................... .......... .... 57
3.6. Near Fields ....................................................................... 65
3.7. Far Fields .......................................................... .............. 71
3.8. Lifetime Testing .................................................... .......... 75

4. ANALYSIS OF EXPERIMENTAL RESULTS................................. 80
4.1. Introduction............................................. ................... 80
4.2. Current Spreading ................................................ .......... ... 81
4.3. Current Leakage ............................................................ 87
4.3.1. Leakage Through the Oxide ................................88
4.3.2. Leakage Across the Barriers................................ 90
4.4. Beam Astigmatism............................................. ............. 92
4.4.1. Gain and Index Guiding .................................... 93
4.4.2.The Astigmatism Factor ...................................... 99


5. SUMMARY AND CONCLUSIONS................................................ 113

REFEREN CES ................................................................................................ 117

BIOGRAPHICAL SKETCH ........................................................................... 121

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



Michael J. Grove

April, 1994

Chairman: Peter S. Zory
Major Department: Electrical Engineering

Experimental and theoretical results on pulsed anodic oxidation of

semiconductor materials are presented. Special emphasis is placed on using

the pulsed anodic process to fabricate diode lasers. The work began with the

exploration of anodic oxidation on p and n type GaAs in a Glycol:Water:Acid

electrolyte. It was discovered that the characteristics of the anodization

process change when the DC current is replaced .with a pulsed current.

Higher current densities can be used with pulsed currents leading to faster

oxide growth rates without loss of oxide thickness uniformity. The use of

pulsed current also allows for the combination of etching and oxidation in a

single processing step. Using this "simultaneous" etch/oxidation process,

ridge guide lasers can be fabricated with arbitrary ridge heights and fixed

oxide thicknesses. SEM analysis of pulse- anodized devices suggests that the

ridge sidewall profiles can be influenced by varying the current density.

Qualitative electrochemical arguments are given for the origin of the

difference between pulsed and DC anodization.

Shallow ridge lasers were fabricated with pulsed anodic oxides.

AlGaAs and AlGaInP lasers fabricated with pulsed anodic oxidation

performed as well as or better than similar lasers fabricated with deposited

oxides. Narrow stripe, ridge guide lasers were fabricated with A1GaAs

material. Ridge guide lasers of different ridge heights were fabricated to

examine the transition of gain guiding to index guiding. Deep ridge devices

were fabricated that operated on a single spatial mode up to the catastrophic

damage level for the cleaved mirror facets. Data showed that threshold

current for the devices did not decrease once a certain ridge height was

exceeded. On the other hand, slope efficiency of the lasers continued to

increase beyond this ridge height. Both of these results are consistent with

theory. The transition from gain guiding to index guiding is shown to depend

on ridge height and gain induced index anti-guiding.

A number of attributes of the fabricated lasers were measured. Data on

power versus current, current versus voltage, near field profiles, far field

profiles, and lifetime performance was collected. The data demonstrate that

pulsed anodic oxidation is a simple, fast process for fabricating high quality

diode lasers.


The latter half of the twentieth century has witnessed an explosion in

the field of microelectronics. The discovery and ongoing refinements in Field-

Effect and Metal-Oxide-Semiconductor transistor technology for switching

and amplification applications has permeated into every facet of modern life.

Recent advances in heteroepitaxy have expanded the materials available for

use in these devices to include compound semiconductors, most predominantly

those composed from column III and column V elements.

A small but ever growing subset of these devices are semiconductor

diode lasers. Small, highly efficient, and in some cases inexpensive, these

devices are capable of providing spectrally coherent, high power optical

radiation from a semiconductor chip with typical dimensions of 500 microns

by 500 microns by 100 microns. Because of their attractive performance

characteristics, diode lasers are being utilized in more systems and

subsystems. Compact disk players, utilizing diode lasers, have literally

revolutionized the commercial music industry in the last five years. Long haul

trunk telephone transmissions are almost entirely the domain of fiber optic

technology with diode laser transmitters. High power lasers, for such varied

uses as optical radars, laser fusion, and machining/cutting operations, are in

use or are coming into use because they incorporate highly efficient laser

diodes as pump sources.

The earliest diode lasers were simply p/n homojunctions made from

direct bandgap semiconductor material [Nath62]. By forward biasing the

device, electrons were injected across the junction from the n-side while holes

were injected from the p-side. In the area near the junction where there were

excess free electrons and free holes, carriers radiatively recombine and

produce a photon with an energy in the vicinity of the band gap ( Figure 1-1

(a)). Further, if the semiconductor is cleaved, a 30 percent reflective mirror

will be formed and a resonant cavity is made. Photons traveling in this

resonant cavity can stimulate further radiative recombinations and optical

gain is achieved. If the gain exceeds the loss, the device will lase.

These early diode lasers suffered from two major deficiencies. First,

once the carriers were injected across the junction, they could diffuse away

from the junction. This distance, depending on the material, could be as much

as two to three microns. A second problem was that the photons traveling in

the resonator were not guided. As the light traveled between the cleaved

mirrors it would diverge, and would be absorbed in the regions away from the

active layer, the volume with excess free holes and free electrons. This

divergence also reduced the overlap between the light and the active layer,

which reduced the number of stimulated events. Because of these factors it

was very difficult to achieve room temperature lasing.

Fortunately, both of these problems were solved with the invention of

double heterostructure devices [Kroe63]. By using heteroepitaxy, a crystal

grower could grow an active layer from a material with a higher index of

refraction and a smaller bandgap ( Figure 1-1 (b)). The smaller bandgap

material provides for band off-sets at the conduction and valence band edges.

These offsets act as barriers to confine the carriers in the active region. In this

manner, the active region can now be determined by the thickness of the

epilayer and not the diffusion length of the carriers. The higher index of the

active layer material forms a dielectric waveguide that confines most of the

optical wave in the active layer. Since most of the injected carriers and most of

the optical wave are in the active region, stimulated recombination rates are

high and therefore so is the optical gain.

A schematic of a heterostructure diode laser, showing the key

epitaxially grown layers is depicted in Figure 1-2. The carriers are injected

into the active layer from the p-type and n-type cladding layers. A portion of

the optical wave travels back and forth in the longitudinal direction, reflected

by the cleaved mirrors. As outlined above, the use of heterostructures

effectively solves the optical and carrier confinement problems in the

transverse dimension, but these confinement issues must also be addressed in

the lateral dimension. An obvious solution would be to grow a heterojunction

in the lateral dimension as well. Indeed, this can be done with complicated

and expensive etch and regrowth techniques.



A c
E p

________ _________I__ ______ F.

0*- Active Layer -

(a) homojunction



p -side


Material II

Material I
(Active Layer)

Material II

(b) heterojunction

Figure 1-1. Simplified energy band diagrams for diode lasers. (a) for
a homojunction device, (b) for a heterojunction device


- Q J QD Q D_- u-

Contact .

/ Oxide

SActive Layer

... Near Field


.) ~ *C

---- Longitudinal

Figure 1-2. Schematic of a heterostructure diode laser.

A far simpler approach is to use a stripe geometry [Dyme67], where an

as grown epitaxial structure is processed so the injected electrical current

flows only in a defined stripe. Stripe definition may be accomplished by

applying a non-conductive layer onto the device, acting to limit the flow of

current in the lateral dimension. This choice of how to confine the beam in the

lateral dimension results in two broad classifications of diode lasers. If the

lateral optical confinement is provided by a real index difference, such as the

etch and regrowth technique, then the laser is called index guided. In devices

where the beam is defined by the current injection path, the beam diverges as

it travels through the laser. The photons that are in the area with free holes

and free electrons can produce further stimulated photons. Those photons

away from the pumped region will be absorbed. In this manner only the

photons in the pumped region will be amplified and contribute to the wave

propagating down the laser. Because the lateral beam dimension is defined by

the width of the positive gain region, they are called gain guided lasers. A

schematic comparing gain and index guiding is shown in Figure 1-3.

The quality of a laser diode is based on its performance characteristics.

However, different applications can require different performance

characteristics. Diode lasers for transmitters in fiber optic communications

systems require high modulation rates, diode lasers for optical recording

applications require precise output beam control, and diode lasers for optical

pumping of larger lasers require high output powers. In addition to these

application specific requirements, all diode lasers must be sufficiently reliable

direction of
wave propagation

optical wave
constant phase

region of absorption

region of gain

region of absorption

direction of
wave propagation

optical wave
--- constant phase

region of low index

region of high index

region of low index

Figure 1-3. Schematic of different guiding mechanisms in diode
lasers,(a) gain guiding, (b) index guiding.


to satisfy the requirements of the application. If the techniques used in

processing the laser material are complex, the probability is high that the

fabricated lasers will have reliability problems, adding to device cost. As

discussed below, we have discovered a new semiconductor wafer processing

technique which is very simple to implement. The goal of this work is to

explore this new technique and determine its usefulness in fabricating high

quality gain and index guided diode lasers.

In fabricating stripe geometry semiconductor lasers, deposited oxides,

such as SiO2 or A1203, are often used to define the openings for current

injection. In depositing these oxides, relatively sophisticated equipment must

be used with tight specifications if both good adhesion to and etching away-

from the semiconductor surface are to be achieved. An alternative to

deposited oxides are native oxides. These native oxides can be formed in a

variety of ways. A great deal of work has recently been done using a wet

oxidation technique to obtain a native oxide on selected III-V materials

[Kish92]. Previous work has shown that an oxide can be formed through an

electrochemical reaction [Hase76]. This process, called anodic oxidation, has

been refined by myself and coworkers by using a pulsed voltage source to drive

the anodic reaction. I have discovered that by using this pulsed anodic

oxidation process I can fabricate high quality gain and index guided laser

diodes. This dissertation is a systematic study of the fabrication and

performance characteristics of lasers fabricated with pulsed anodic oxidation.

In Chapter 2, I will explain the fundamental concepts of the pulsed

anodic process. I will give a brief review of anodic oxidation along with some

first principles ofelectrochemical reactions. I will then discuss how the use of

a pulsed voltage source changes the anodic process. By using a pulse, I have

demonstrated that one can use much higher current densities and still

maintain a uniform oxide. The higher current densities allow for much faster

growth rates. Another phenomenon that arises with the pulsed process is

what I call a traveling oxide layer. By using the pulsed process, etch and

oxidation steps can be combined to fabricate ridge structures for index guided

lasers. Chapter 2 contains the details of the pulsed process as well the results

of experiments I have conducted where I examined the effects of varying a

number of the pulsed parameters. Finally, the chapter contains a list of

materials that I was successfully able to oxidize using this technique.

In Chapter 3, performance characteristics of a number of lasers

fabricated with pulsed anodic oxides will be discussed. Threshold current and

slope efficiency will be analyzed in some detail. From these measurements

the quality of the material used can be determined and the effect of stripe

width, oxide thickness, ridge height, and other factors will be analyzed. The

analysis of the near field, the optical intensity distribution on the laser facet,

and the far field, the divergence of the optical field away from the laser will be

presented. The relationship between the near and the far field will be used to

determine the amount of gain and/or index guiding for a given laser. I have

explored the physical characteristics of the oxide layer and will report the

findings in this section. Finally, I will report the results of lifetime tests on

these devices that have been operated in a continuous mode for more than

1500 hours. Analysis of this data will give insight as to how the oxide layer

withstands the stresses of operation.

In Chapter 4, I will present a theoretical background to explain the

performance of the devices tested in Chapter 3. Since.diode lasers are current

injection devices, lateral current spreading can be expected to effect both the

threshold current and the slope efficiency. I will explain current spreading

and present a model as to how current spreading is reduced in devices

fabricated with pulsed anodic oxides. In this chapter, I will also explain types

of carrier leakage and how it can reduce slope efficiency and increase

threshold current. One form of carrier leakage occurs when the forward bias

causes current to flow through the oxide layer and not into the desired lasing

area (see Figure 1-2). Another form of carrier leakage occurs if carriers

injected into the active layer overcome the potential barriers of the

heterojunctions defining the active layer and recombine in the cladding layers

(see Figure 1-1 (b)).

As discussed earlier, using the traveling oxide phenomenon I can

fabricate ridge structures with the pulsed anodic process. Because the optical

wave extends into the cladding, this ridge provides a real index difference in

the lateral dimension. The amount of index guiding is a complicated function

of the index of the materials, the ridge height, and the wavelength of the

optical radiation. This is further complicated by the presence of the oxide


which may introduce a stress that further changes the index. All of these

effects will be addressed in Chapter 4 and compared to experimental data

reported in Chapter 3.

Finally, in Chapter 5, I will summarize this work and draw conclusions

about the pulsed anodic process. I will also point out future research areas

that are important and exciting but beyond the scope of this work. These

research areas include but are not limited to use of pulsed anodic oxides to

fabricate lasers in materials for wider bandgap materials such as GaP for red

lasers and ZnSe for blue lasers


2.1 Introduction

Growth of oxides on III-V materials is more complex than that on

silicon because the competition between two or more elements can cause

highly nonuniform chemical composition. For this and other reasons,

thermally grown oxides on III-V materials have been far less successful than

thermally grown SiO2. Deposited oxides are available but have the

disadvantages outlined earlier in Chapter 1. A third alternative is a native

oxide. Due to the low bond strength at the surface of a semiconductor and the

abundance of oxygen in the atmosphere, a thin layer of native oxide always

forms on a semiconductor unless extraordinary means are taken to prevent it.

By incorporating the semiconductor as the anode in an electrochemical cell,

we can continue this oxidation reaction to increase the thickness of the oxide.

The method described for forming oxide in this manner is called anodic

oxidation. Normally the applied voltage or current used in electrochemical

reactions is kept constant (DC) [Newm91]. I have found that the anodic

oxidation process has different characteristics when a pulsed voltage source is

used in lieu of a constant source. These characteristics are interesting, and I

believe beneficial, when used to fabricate diode lasers. In this chapter, I will

describe anodic oxidation and provide a very brief review of the work in this

area. After explaining what anodic oxidation is, I will examine the parameters

of pulsed anodic oxidation and show how these factors affect the formed oxide.

I will report the results of experiments I have conducted where I examined the

effects of varying a number of the pulsed parameters. Finally, the chapter

contains a list of materials that I was successfully able to oxidize using this


2.2 Anodic Oxidation

Simply stated, anodic oxidation is an electrochemical reaction where

the oxygen in an electrolytic solution reacts with the anode of an

electrochemical cell to form an oxide on the anode. The most basic elements

of an electrochemical cell are two electrodes (an anode and cathode), an

electrolytic solution, a voltage source, and a circuit to connect them all in

series. To describe an electrochemical reaction we must know three things

about the system: first, what is the initial state of the system before a voltage

is applied; second, when a voltage is applied, how will the ions that are in the

electrolytic solution be transported to the electrodes, and third, when the ions

are in the vicinity of the electrode, what reactions will occur between the

electrode and the ions. This situation is very similar to the study of

semiconductor devices in which we wish to know the equilibrium band

diagrams, the drift and diffusion of carriers under an applied bias, and what

the carriers do at junction boundaries. There are a large number of analogies

between electrochemistry and the analysis of semiconductor devices.

Unfortunately, the terminology and symbols used are not among those things

the two disciplines have in common. For this reason, I will use the

terminology and symbols prevalent in the electrochemical literature and

attempt to point out the analogous concepts in the study of semiconductors.

The initial state of the system is obtained with the concept of

electrochemical potentials, gi, and thermodynamic equilibrium. Like most

forms of potential, absolute electrochemical potential has no meaning, only

the difference in potential is important. We will adopt Guggenheim's

definition for the difference in electrochemical potential of an ion as follows; if

an ion exists in two adjacent phases, a and P, then gt ji is the work done in

transferring reversibly, at constant temperature and volume, a mole of species

i from phase P to phase a [Newm91]. It is convenient to further decompose the

electrochemical potential into a chemical potential and an electrostatic


gi = gi + RTln (cifi) + ziFQ (2.1)

where, zi is the charge number of species i, and F is Faraday's constant of

96,487 C/equivilant, Q is the electrostatic potential in volts, i is a reference

term in J/mol, ci is the molar concentration of species i in mol/liter, fi is the

molar activity coefficient of species i, R is the universal gas constant, and T is

the absolute temperature. The first two terms in Eq. (2.1) represent the

chemical potential. The third term in the equation represents the

electrostatic potential and is written in such a manner to separate the long

range forces associated with coulombic interactions from the shorter range

forces related to concentration variations. This equation and its constituent

variables point out some of the differences in electrochemical and

semiconductor terminology. Electrochemists like to deal in joules and moles,

while electrical engineers like to deal in electron volts and carrier densities. A

more familiar quantity in semiconductor terminology is the fermi level, Ep

This can be related to the electrochemical potential of electrons in the

conduction band by the following equation:

e- = LE + C (2.2)

where L is Avogadro's number in inverse moles, and C is an arbitrary

constant. The electrochemical system will be in equilibrium if for each phase a

and p, ~4i = g for each species i. This is analogous to setting up a band

diagram for p-n junctions by lining up their Fermi levels. This equilibrium

situation will establish the initial state of the system.

After the initial state has been determined, the transport of the ionic

species to the working electrode must be established. The transport of an

individual ionic species can be obtained from the equation for the flux of

species i:

Ri = -ziuiFciV + (-DiVci) + ci(


where Ni is the flux density of species i in mol/cm2-s, ui is the mobility of

species i in cm2-mol/J-s, Di is the diffusion coefficient of species i in cm2/s, and

v is the electrolyte velocity in cm/s. This equation is readily seen to be

composed of a drift term, a diffusion term, and a term due to the possible

velocity of the electrolyte. Once the flux of a species i is known, you can

calculate the total current density by summing over the contributions of each


F = F zi (2.4)

where J is the current density in A/cm2. Transport is important because even

if the conditions for a desired reaction at the electrode are favorable it

obviously can only occur at the rate which the reacting species can be

transported to the electrode.

The final aspect of the system is the reactions at the electrode, or

electrode kinetics. To analyze the kinetics of an electrode reaction, a model of

the semiconductor (electrode) and electrolyte interface must be chosen. The

model I have chosen is depicted in Figure 2-1. The semiconductor acts as an

electrode and there is a space charge layer (SCL) near the semiconductor

surface. Ions in the solution will be physically adsorbed on the electrode

surface due to dipole interactions and van der Waals forces. The plane

passing through the center of the adsorbed ions is called the inner Helmholtz

plane (IHP), the plane of closest approach for non-adsorbed ions is called the

"1 I

Semiconductor I
Charge Layer






a,,- -- n,--L-',--


Region of
Diffuse Charge

Figure 2-1. Model for semiconductor-electrolyte interface (not to scale).

Bulk Electrolyte


outer Helmholtz plane (OHP) [Oraz86]. The region of diffuse charge is a

layer of accumulated ions that balance the effects of charge at the surface of

the semiconductor, and the adsorbed ions. This layer of diffuse charge may be

10-20 A deep. Next, there is a diffusion layer in the electrolyte. In this layer

electrical neutrality holds but the concentrations of the ions may be a

function of position. Finally, there is the bulk electrolyte. In this region

charge neutrality holds and the ion concentrations are no longer a function of

position. It is clear, even from this simplistic model, that the current density

across these interfaces, J, will be a complicated function of potential offsets

and concentration distributions. As an electrostatic potential is applied to the

system a current density will begin to flow. For most electrochemical systems

the current density can be described by the following equation [Newm91]:

= Jo exp ( ) exp -FRT ) (2.5)

where Jo is the exchange current density in A/cm2, aa and ac are the anodic

and cathodic transfer coefficients, and rl, is the surface overpotential in volts

and is defined as the departure from the equilibrium potential. Finally, the

reaction rate of the anodic oxidation can be expressed as:

r =nF rf-rb (2.6)

where r is the net reaction rate in mol/cm2-s, rfis the forward reaction rate, rb

is the backward reaction rate, and n is the number of electrons transferred in

the electrode reaction. To summarize, the equations presented allow the

calculation of the reaction rates either theoretically or experimentally.

Theoretically, we establish the equilibrium potential of the system from Eq.

(2.1), calculate the species flux from Eq. (2.3), calculate the current density

from Eq. (2.4), and calculate the reaction rate from Eq. (2.6). In practice these

quantities can be very difficult to calculate, so it is common to establish the

parameters of Eq. (2.5) experimentally then use the current density to

calculate the reaction rate from Eq. (2.6).

The above treatment assumes that you understand the reactions that

are occurring at the electrodes, referred to as electrode kinetics. A number of

processes that occur at the electrode are illustrated in Figure 2-2 [Park92]. In

Figure 2-2 (a) negatively charged ions drift through the oxide under the

influence of the electric field and in Figure 2-2 (b) the field causes drift of the

positively charged ions. The exact mechanism of this drift is unknown, but it

has been postulated that the III/V ions drift interstitially and the oxygen ions

drift via vacancies [Bree79]. In Figure 2-2 (c) excess hydrogen in the

electrolyte can react with the oxygen and dissolve the oxide. Finally, in Figure

2-2 (d) the oxide is formed, both at the semiconductor/oxide interface and at

the oxide/electrolyte interface. The most difficult part of electrode kinetics

involves predicting the reactions that occur between the electrolyte and the

electrodes. Fortunately, anodic oxidation of GaAs has been examined

previously, investigated as early as 1963 [Reve63]. Since then, several


0 Oxygen ion
* Hydrogen ion
0 Column II/V ion


+--- Semiconductor -- Oxide --- Electrolyte

Figure 2-2. Schematic of processes that occur at the electrode surface,
(a) inward drift of negative ions, (b) outward drift of positive ions, (c)
oxide dissolution, (d) oxide growth.

researchers have oxidized GaAs and AlGaAs, the primary difference being the

composition of the electrolyte. An outstanding review of the history of anodic

oxidation on GaAs has recently been published [Huds93]. Most of the

researchers agree that the reactions at the anode are:

2Ga3++ + 3(OH') Ga203 + 3H (2.7)

2As3+ + 3(OH' As203 + 3H+

and the reactions at the cathode are:

6H+ + 6e" -4 3H2 (2.8)

6H20 + 6e" 6(OH'J + 3H2

As previously stated, one of the key elements of the electrochemical

process is the electrolyte. We have chosen to work with Glycol:Water: Acid

electrolyte [Hase76]. This electrolyte was chosen because it is safer than

many of the electrolytes normally used in anodization of GaAs [Huds93], and

works well over a broad range of pHs and glycol to water ratios. The purpose

of the acid is to increase the conductivity of the electrolyte. The purpose of the

glycol is to change the relative transport parameters of various ions in the

solution. For example, the OH- anion participates in oxidation so we want its

concentration to be high at the semiconductor surface. However, the (H2PO4)

ion acts as an etchant so its concentration should be low at the semiconductor.

Since the OH- ion is smaller it will be transported faster through the viscous

solution and greater oxidation and less etching will occur. The purpose of the

water in the electrolyte is to provide OH- ions for oxidation.

2.3 Pulsed Anodic Oxidation

In the previously cited work, a constant voltage was used to drive the

anodic reaction. In our work, the applied voltage is pulsed. We have found

that a pulsed source provides uniform oxides about ten times faster than a

constant source, allows etch and oxidation steps for ridge structures of

arbitrary ridge height to be combined, and provides a method for real-time

electrical monitoring of the oxide thickness.

2.3.1 Theory of Pulsed Anodic Oxidation

One of the advantageous of using pulsed electrolysis is the

enhancement of the mass transfer characteristics during the time of the pulse

[Visw78]. Using DC currents you can increase the mass transfer by

increasing the current but eventually secondary and perhaps unwanted

reactions will occur, such as hydrogen evolution at the cathode or etching of

the oxide at the anode. This would indicate that higher currents can be used

with the pulsed method meaning faster oxidation times.

A second advantage is one of uniformity. I have found that one can

obtain uniform films with the pulsed oxidation process about ten times faster

than with the constant voltage process. Using a 0.5 cm2 sample of p-type

GaAs and a constant current density of 2.6 mA/cm2 (this is 2.6 times higher

than that suggested as optimal by Hasegawa and Hartnagel [Hase76]) it took

40 minutes to grow 1000 A of oxide. Using a pulsed voltage at a current

density of 120 mA/cm2, 1000 A of oxide was grown in only five minutes. There

was no difference in the uniformity of the two oxides. The oxidation was

repeated on a sample with constant current density of 46 mA/cm2. The

growth was rapid but severely nonuniform, in addition, significant bubbling

was observed at the cathode which indicates hydrogen evolution.

To examine the causes of the nonuiniformities, it is instructive to

examine Eq. (2.3). The first term of this equation is the drift term from the

local electric field. The second term of this equation arises from any

concentration variation with position. The third term of the equation

represents flux from convective flow of the electrolyte. All three of these terms

have time and spatial dependence. Experimentally, we have determined that

for pulsewidths equal to 700 p.s the spatial variations of the sum of the flux

components are small. This was further verified by performing a pulsed

anodization where the pulsewidth was increased for a fixed repetition rate.

For high applied voltages, as the pulsewidth increased so did the

nonuniformity of the grown oxide, with the oxide near the edges being

significantly thicker. To determine which of the flux components contributes

to the spatial nonuniformities ,for longer pulsewidths, requires extensive

modeling of each of the flux components and is an important development left

for future work.

Although unrelated to the uniformity issue, certain characteristics of

mass transfer do change in the pulsed domain. To examine this more

quantitatively, the continuity equation for a species i can be written:

a = V.i + Ri (2.9)

where Ri is the reaction rate of the species in the electrolyte. If we neglect

convection and drift terms from Eq. (2.3) the flux of the species is given by the

diffusion term only. If we further assume that the species reacts only at the

electrode then Ri is zero. With these assumptions the continuity equation


c = Dd (2.10)
at dx2

where D is the diffusion coefficient and x is the dimension normal to the

electrode surface. Eq. (2.10) is recognized as a form of the heat equation and

can be solved using Fourier series. The boundary conditions can be expressed

as [Cheh71]:

c(x, t) I=o,t>o = Co (2.11)

x t > O, x =

where x=0 is the outer edge of the diffusion layer and x=8 is the inner edge of

the diffusion layer. The solutions are an infinite series of exponentials, but if

we define tr as the period of the pulse and t1 as the pulse duration and we let

the product of (~2DtT)/ 482) become very small the asymptotic limit of

(Jp) lim (Jdc) lim (2.12)

is obtained. The 'lim' indicates the limiting current density that can be

obtained. In an electrochemical system the current density from a reaction is

limited by the rate at which the reacting species can be supplied to the

electrode, this maximum value is called the limiting current density. From

Eq. (2.12) we can see that as the duty cycle gets smaller, Jp gets larger.

However, the anodic oxidation rate is proportional to the current density

times the pulse duration. Since the dc current is on continually and the

pulsed current is only on for the pulse duration, as the duty cycle gets smaller,

the oxidation rate gets smaller. Therefore, the increase in limiting current

density is offset by the decrease in pulse duration and the oxidation rates of

the two should be the same. The problem with this argument is that the

anodic oxidation was not done at the limiting current density. The current

density for both the pulsed and dc cases were kept below the limiting value to

get optimum uniformity.

The actual current density will have a component due to the reaction at

the electrode but it will also have a capacitive component due to the charging

of the various layers in Figure 2-1. The current density equation, Eq. (2.5),

must be modified to include this charging term:

J =C + JoLexp(- RT )F- exp(F RT (2.13)

where C is the double layer capacity in F/cm2. The electrode displays edge

effects much like the edges of a finite parallel plate capacitor. The capacitance

term acts to damp out these edge effects by redistributing charge in the layers

shown in Figure 2-1. After the capacitive effects have decayed the edge

effects will start to become more pronounced. For the pulse case, the charge

gets redistributed in the layers every pulse period. As will be detailed later in

Figure 2-4 these charging effects last approximately 300 is. After they decay

out, then the primary source of current comes from reactions at the electrode.

This theory was verified experimentally. When the pulsewidth was increased

beyond 700 as, nonuniformity at the edges became more pronounced. When

the pulse width is decreased under 700 us, the dominant current component

is capacitive and the oxidation rate decreases. As mentioned before, a more

exact treatment of the interrelationship between current density, pulsewidth,

growth rate, and uniformity would require a three dimensional, time

dependent solution of Laplace's equation and the flux equation and will be left

for future work. In summary, one of the advantages of pulsed anodization is

not that it is faster, but that it is faster while still maintaining uniform


2.3.2 Results on Bulk and Laser Diode Material

To determine the useful process window for pulsed anodic oxidation, a

number of anodizations were done on both bulk and laser material. The

experimental set-up to perform the anodizations consisted of two electrodes

immersed in solution, a pulsed voltage source, and two resistors and is shown

in Figure 2-3. The solution was a GWA mixture of ethylene glycol, deionized

water, and phosphoric acid (40:20:1). A semiconductor wafer served as the

anode and a platinized titanium grid as the cathode. The voltage source was a

HP214A pulse generator. The solution, electrodes, and pulse generator were

connected in series with a small (10 Q) resistor and a large variable resistance

to control the current density. The voltage across the small resistor was read

on an oscilloscope to monitor the time dependence of the current in the circuit.

As explained earlier, the pulse width, repetition rate, and pulse amplitude

were varied to obtain the optimal settings for uniform oxide growth. Good

results on GaAs were obtained with a pulse width of 700 as, a repetition rate

of 50 Hz, a pulse amplitude of 120 V, and a current density of 120 mA/cm2.

As expected, current flow in the circuit decreases as the oxide becomes

thicker. However, the magnitude of the leading edge of the voltage pulse, as

read across the small resistor, does not change with time. Looking at a

representative oscilloscope trace, Figure 2-4, shows that at both an initial

time, to, and a final time, tf, the leading edge of the pulse remains constant,

whereas the trailing edge of the pulse decreases with time corresponding to

the growth of the oxide. The pulse shape can be explained by modeling the

Voltage Pulser

10 Ohm


Variable Resistor

Platinized Titanium Cathode

GWA Electrolyte

Semiconductor Anode

Figure 2-3. Schematic of test set-up for anodic oxidation.

Positive Ground

to 1.2

L ....................... .. ..... .. ....... .. . . . .. ........... .
....... ... ...... ... .... .... ....... .................... ........ ......... .............

S 0.6

o 0.4 ___

0.2 ..............
0.2 -

0 200 400 600 800 1000
Time microsecondss)

Figure 2-4. A representative oscilloscope trace during anodization. The
pulse indicates the current flowing in the system.

solution/oxide/ semiconductor system as a Schottky barrier diode which

transitions to an MOS capacitor as the oxide thickens. At the beginning of

each voltage pulse, the capacitance of the oxide acts to short out the oxide

resistance. After the capacitor charges, the current decreases to a steady state

value within the pulse. As the oxide thickens, the resistance increases,

therefore the steady state current decreases from pulse to pulse. As long as

the voltage pulse width is long enough for the capacitor to achieve a quasi-

steady state, this performance is invariably repeated. The oxide growth

terminates when the oxide resistance is large enough to render the available

voltage pulse insufficient to drive the anodic reaction. When this happens, the

trailing edge of the pulse no longer decreases and the shape looks like the

pulse in Figure 2-4 at tf. This pulse shape variation during oxide growth

provides for convenient, real-time monitoring of the oxide thickness. If the

initial voltage amplitude is not large enough to begin the reaction, no oxide

will form and the pulse will remain rectangular.

For my initial work, pulsed anodic oxides were formed in n-type and p-

type GaAs. The samples were substrate material and therefore relatively

highly doped. I attempted to oxidize semi-insulating GaAs and could not do so

even with current densities as high as 200mA/cm2. We believe that this is due

to the lack of charge carriers in the semiconductor and the barriers at the

interfaces, both of which are affected by doping levels. Typical oxidation times

were from 5 to 10 minutes with the pulse parameters described above. The

deposited oxides ranged from 1000-1500 A, with ridge heights from 2000-3000

A (measured from top of cap to bottom of oxide) depending on material and

oxidation parameters. In addition to GaAs, we were able to anodize p-type

A1GaAs, p-type Gao.sIno.5P, n-type InP, and p-type GaSb [Grov94]. Since our

procedure is not strongly dependent on material composition, no major

changes in pulse parameters were needed.

The oxide formed on p-type GaAs was analyzed for composition and

capacitance. XPS measurements showed that the composition of the oxide was

Ga203 and As203. CV measurements showed a capacitance of approximately

48 pF for a sample with an area of 6.25 x 10-4 cm2 and a thickness of 1000 A.

Using these figures a relative permittivity of 8.7 is obtained. These values

agree with published results for earlier oxides [Mein78]. For diode lasers, a

more important parameter than capacitance is breakdown voltage. To

measure this, a section of laser material was cleaved into two pieces, one piece

was anodized and another piece had CVD SiO2 deposited on it. The two

samples were processed as lasers except no opening in the oxide was created,

that is, the structure was metal, oxide, semiconductor, and metal. The only

way for current to flow in these samples was through the oxide. The anodic

oxide and the CVD oxide both began to pass current at about 10 volts (a

breakdown field of about 1 x 106 V/cm). Since typical operating voltages of

diode lasers are less than five volts, the anodic oxide possesses acceptable

current blocking characteristics, and is as good as the CVD SiO2 which is

currently used today. I-V characteristics of a typical diode laser fabricated

with pulsed anodic oxidation is shown in Figure 2-5. Recent work with


E mm. I.


pulsed anodization of aluminium metals has been suggested that the use of

pulsed anodization may actually improve the breakdown strength of the oxide

[Park92]. Since our work is with semiconductors these results may not be

true for our oxides but the results are interesting nevertheless.

One of our goals in this research has been to develop a processing

method for fabricating stripe geometry semiconductor diode lasers that is

simpler and more reproducible than the deposited oxide technique. For

validity of comparison, we cleaved one piece of AlGaAs laser material into two

samples and processed them into lasers. On one sample we deposited Si02

via chemical vapor deposition (CVD) and subsequently defined stripes

through the oxide with standard photolithography and buffered hydrofluoric

acid. Selective anodic oxidation on the other sample was done using

photoresist to mask the stripe areas and the process described above to form

the oxide. The steps to fabricate a laser with pulsed anodic oxidation are

shown in Figure 2-6. The details of the experiment. have been analyzed in

detail elsewhere [Grov94], [Huds93]. The basic result was that the lasers

fabricated with anodic oxides had performance characteristics as good or

better than lasers fabricated with CVD oxides. A1GaInP and GaInAs lasers

have also been fabricated with anodic oxides.

Researchers have raised questions about the stability of anodic oxides

to subsequent processing [Burt92]. It is true that these native oxides are

easily dissolved in basic solutions such as photoresist developer, but we have

developed a technique that uses a metal ion free photoresist developer

Semiconductor Laser

Flow (a)

,A A

A A A1



p+ cap
O -



(b) Metal
[ % % % % %


Figure 2-6. Steps for fabricating a laser with anodic oxidation, (a) clean
sample, (b) define stripe with photoresist, (c) anodize area not covered
with photoresist, (d) strip photoresist and evaporate metal contacts.


(Shipley Microposit MF 319) which does not rapidly dissolve the oxide and

allows further processing on top of the oxide including lift-offs. This process is

described in detail in Chapter 3. The anodic oxides are also stable in

deionized water, TCE, acetone, and methanol.

The pulsed process also allows the combination of the oxidation and

etching steps in the formation of ridge structures. The oxide growth rate will

become negligibly small when the potential drop across the oxide is so great

that the remaining voltage can no longer drive the anodic reaction. This is

true for both constant and pulsed processes. However, when the pulse is off,

the solution acts as mild etchant for the oxide. Certain oxides can be dissolved

in acids or bases, these oxides are called amphoteric and are formed from

elements between metals and non-metals [Mort75]. Representative reactions

are as follows:

A1203 (s) + 20H (aq) + 3H20 2A1 (OH) 4 (aq) (2.14)

A1203 (s) + 6H (aq) 2A13 + 3H20

When the pulse shape has stabilized (tf in Figure 2-4), a quasi-steady

state is achieved in which the oxide dissolution rate at the oxide/solution

interface equals the oxide formation rate at the oxide/semiconductor interface.

In this manner, an oxide film can be grown, and a ridge of arbitrary height

defined simultaneously. The work using pulsed anodic oxidation to form

lasers with varying ridge heights will be described in detail in Chapters 3

and 4.


3.1 Introduction

In Chapter 1 I discussed some of the factors that affect the

performance of diode laser performance. Lateral current spreading, current

leakage, and differences in the real and imaginary index guiding components

all impact on the laser characteristics. In this chapter, I will discuss the

experiments that I have conducted and report the data I have compiled. I

have made a purposeful effort in this chapter not to analyze the results,

merely to report them. A detailed analysis of the data is reserved for Chapter


To minimize the effects of different laser designs and growth variations,

I did the bulk of my experiments using material from one AlxGal.xAs/GaAs

wafer. In this chapter, I will describe that wafer in detail and the processing

steps that I used to fabricate the various lasers. One of the experiments I

conducted was to make lasers of varying ridge heights and then determine the

effect the ridge height had on threshold current, slope efficiency, near field,

and far field of these lasers. The descriptions and results of these experiments

will also be included in this chapter. To reduce the effects of thermal

gradients and other transients, most of the tests were done with a pulsed

drive voltage. Some applications for lasers require continuous wave (CW) or

near CW performance. Therefore, to evaluate pulsed anodic oxidation

processing technology in its final form, I have conducted several CW lifetime

tests. These results are also included in this chapter.

3.2 Device Fabrication

The material selected for use was a Single Quantum Well Graded Index

Separate Confinement Heterostructure (SQW-GRINSCH) structure grown by

Metal Organic Chemical Vapor Deposition (MOCVD) at Lawrence Livermore

National Laboratories. For convenience, I will refer to the sample as L473. A

schematic of the AlxGal_-As/GaAs epilayer structure is shown in Figure 3-1.

The top layer is a high conductivity cap to facilitate ohmic contacts. The first

graded p layer is a barrier reduction layer. The purpose of the barrier

reduction layer is to reduce the potential drops caused by abrupt

discontinuities in the conduction and valence band edges that occur by

growing GaAs on Alo.6Ga0.4As.

The 1.4 micron layer ofAl0.6Ga0.4As serves as an optical confinement

layer due to its lower index of refraction, approximately 3.27 at a lasing

wavelength of 801 nm. In contrast, Al0.3Ga0.7As has a refractive index of

3.45. The active layer A10.08Gao.92As has an index of approximately 3.64.

The indices were calculated as a function of wavelength and Al fraction from a

paper by Adachi [Adac89]. It has been pointed out that when materials are

incorporated as part of a quantum well structure, the optical properties will be

500A GaAs: p ++
50 A GaAs: p
250 A Ao.05Gao.95As-Alo.6Gao.4As: p (graded)

1.4 p.m Alo.6Gao.4As: p

0.2 pm Alo.6Gao.4As-Alo.3Gao.7As: p- (graded)

100 A Alo.08Gao.92As SQW

0.2 pm Alo.3Gao.7As-Alo.6Gao.4As n- (graded)

1.4 pm Al0.6Gao.4As: n

500 A AIo.6Gao.4As-Alo.05Gao.95As: n (graded)

0.25 pm GaAs: n+

GaAs Substrate: n+

Figure 3-1. Schematic of Laser Structure, L473.


different than the bulk values [Kahe86], but the difference in values for Alx

Gal-xAs is no more than two percent. The second graded p region provides

one side of the potential well for confining the carriers in the quantum well

active layer. The n layers perform the same functions as the corresponding

layers in the p region.

Tosee how pulsed anodic oxides affect laser performance, L473 was

cleaved into several pieces and each piece was processed into lasers of various

ridge heights. The ridge heights were not strictly selected, instead oxidation

times of 3,10, 20, 60, and 90 minutes were chosen. The following illustrates

the rationale for the selected times. As explained in Chapter 2,from earlier

work with Alx Gal-xAs samples, a semiconductor consumption rate of about

200-220 A per minute was established. Based on this rate, the 3 minute

sample would consume only the high conductivity cap; the 10 minute sample

was the approximate time in which the terminal thickness forms; the 20

minute sample would consume a portion of the A10.6Ga0.4As optical cladding

layer so it should begin to effect the index guiding properties of the laser; the

60 minute sample would have a ridge height of approximately 1.2 microns

which is compatible with conventional ridge guide devices; and the 90 minute

sample should come very close to the active layer so we could observe the

effects of carriers in the active layer interacting with the anodic oxide.

The oxidation process was the same as that described in Chapter 2.

The mask used to define the laser stripes had feature widths of 5, 50, and 100

microns on 500 micron centers. The 5 micron stripe lasers were used for this

work, the 50 and 100 micron lasers were used for other experiments. A stripe

width of five microns was selected for a number of reasons. First, I wanted to

obtain single spatial mode performance so the stripe had to be small.

Secondly, from my background in the military, I know that the core diameter

of the standard military single mode fiber is 6 microns. By using a waveguide

on the order of five microns I should obtain a laser that can provide significant

coupling into a standard military single mode fiber (non-military single mode

telecommunication fibers also have cores on the order of 6 microns but I am

unaware of an accepted standard for these fibers).

The oxidation was performed on the samples with the parameters listed

in Table 3-1. I tried to keep the initial current density the same for all samples

in an effort to minimize other variables. However, when I anodized samples 4

and 5, anomalous behavior occurred. As explained in Chapter 2, as the oxide

grows, the trailing edge of the current pulse monotonically decreases until

the terminal thickness is achieved. For samples 4 and 5, this behavior was

observed for the first 35 minutes of oxidation. After this time, the trailing

edge of the current pulse started to increase. I believe this occurs because the

oxidation rate is a function of the chemical composition and the doping. As

the higher doped material is consumed, the oxidation rate changes and the

rate of oxide growth is now slower than the rate of the oxide being etched.

This was visually verified during the growth; the oxide that had been a light

blue color (approximately 1500 A) was now turning purple or dark brown

(approximately 500-800 A). To counter this effect, I decreased the external

Table 3-1 Anodization Parameters

Surface ren Pulsewidth Repetition Duration
Sample Density
Area (cm2) (mA/cm2 (s) Rate (Hz) (minutes)

1 .97 89.7 700 50 3
2 1.04 86.3 700 50 10
3 .95 91.6 700 50 20
4 1.13 86.0 700 50 60
5 .99 87.9 700 50 90

resistance, and thereby increased the current density during the anodization

of sample 4 and 5. The values of current density for samples 4 and 5 in Table

3-1 represent the initial levels. For these samples, I increased the current so

that I obtained a blue oxide at the end of the anodization duration.

The samples were processed into laser bars after anodization. First,

the samples were lapped to a thickness of 100 microns. The lapping is

primarily done to ensure a high quality cleave on the final devices. After

lapping, the samples were cleaned and n-side metallization was done. The

following metallurgy was deposited via e-beam evaporation; 200A of Ge, 400 A

of Au, 500 A of Ni, and 1500 A of Au. The samples were then annealed in

forming gas at 4000 C for 4 minutes. The annealing is done to facilitate

formation of ohmic contacts. The next step is to perform lift-off

photolithography. The ability to do a metal lift-off is one of the advantages of

pulsed anodic oxides. It has been reported [Burt92] that anodic oxides are not

stable to subsequent processing. It has been demonstrated and reported

[Grov93] that for pulsed anodic oxides this is not the case. In the lift-off

photolithography I spun on 1400-23 photoresist and then exposed it through a

mask with 100 micron ridges on 500 micron centers. After exposure, the

photoresist was developed in Shipley Microposit MF 319 developer to provide

the ridges for lift-off. The MF 319 is metal ion free and as such does not

dissolve the anodic oxide. After lift-off photolithography, the p side contact of

1000 A Ni, 1000 A was deposited. Metal covering the isolation regions was

lifted off with acetone and methanol, thereby electrically isolating the

individual lasers. It should be noted that this lift-off step is not required if the

sample will be cleaved into individual laser chips; the lift-off is done so that

each individual laser can be tested in bar form (see Figure 3-2,). The ability to

perform subsequent processing steps after oxidation demonstrates the

stability of pulsed anodic oxides.

After metallization was completed, the lasers were cleaved into cavity

lengths of 500 microns using standard scribe and roll techniques. Some bars

were further cleaved into 500 micron by 500 micron chips for mounting on

small copper blocks. These blocks serve as heat sinks and allow for more

robust handling of the devices. With still further processing, these blocks can

be used to mount the chips for CW testing. This will be addressed in

Paragraph 3.8.

3.3 Ridge Evolution

The index guiding properties of ridge lasers will be influenced by the

profile of the sidewalls of the ridge. Therefore, an inspection of the sidewalls

of samples 1-5 was made to establish the relationship between anodization

parameters and sidewall profile.

The samples were measured with a Dektak II profiler to determine the

height of the ridges and the thickness of the oxide. The Dektak provides a

vertical resolution of 5 A, however, I took all my measurements to the nearest

10 A. The samples were first profiled in metalized bar form. This

measurement provided the thickness of the metal and the partial ridge. After

this measurement, the bars were rinsed with Shipley 606 developer which

Individual Lasers

Exposed Anodic Oxide Stripes

Figure 3-2 Use of lift-off to obtain individually addressable lasers
on a laser bar.

stripped the oxide from the exposed areas (see Figure 3-2). A second profile

was then taken to measure the distance from the unoxidized semiconductor to

the top of the metal. The oxide thickness and ridge height can be determined

by combining the results from the two measurements. By taking the ridge

height and dividing by the duration of the anodization, an average etching

rate can be found.

The results are summarized in Table 3-2. This table illustrates the

growth tendencies explained in Chapter 2, and the anomalous behavior

mentioned in paragraph 3.2. As the anodization time increases, the oxide

growth rate increases until a terminal thickness is achieved. If the

dissolution rate of the oxide equals the growth rate, then a quasi steady state

exists and a fixed oxide thickness travels down through the material. For

samples 4 and 5 the etch rate is faster than the growth rate. Due to the

doping and composition of the lower layers the growth rate of the oxide

declined while the oxide dissolution rate remained the same, this is evident by

the lower overall etch rate. The thicker oxide layer for sample 5 is explained

because I increased the current density near the end of the anodization,

thereby increasing the oxide formation rate.

An examination of the sidewall evolution was done by analyzing

scanning electron micrographs (SEM) of the samples. The features of samples

1 and 2 were too small to get any effective information from SEM analysis.

Sample 3 is shown in Figure 3-3. (a). In the SEMs, the black layer is the

oxide, the lighter layer on top of the oxide is the p-side metallization, and the

Table 3-2. Oxide Thicknesses and Ridge Heights

Oxide Thickness Ridge Height Etch Rate (A/
Sample () () min)

1 780 1530 510
2 1270 2770 277
3 1530 4290 215
4 870 11,770 196
5 1040 15,970 177



Figure 3-3. SEMs of the facets of laser bars from wafer L473. (a) Sample
number 3. (b) Sample number 4.

gray material below the oxide is semiconductor. The SEM for this sample

shows that there is evidence of undercutting. The .sidewall is smooth and

inclined at about a 600 angle. The oxide away from the ridge is uniform in

thickness but not flat. A SEM of sample 4 is shown in Figure 3-3.(b). This

sample also shows evidence of undercutting. The slope of the sidewall starts

at an angle of about 600 and then gently smooths out. The thickness of the

oxide is very uniform and smooth, both on the sidewalls and away from the

ridge. A magnification of the left hand side of Figure 3-3.(b) is shown in

Figure 3-4. The SEM for sample 5 is shown in Figure 3-5.(a). The undercut

for this sample is significantly worse and corners have started to form at the

base of the ridge. In addition, the sidewalls are not as smooth as with

sample 4.

I believe these results can be explained by looking at Figure 3-5.(b).

This is a laser made from a very similar wafer. In anodizing this sample, the

applied voltage remained constant throughout the entire anodization period of

40 minutes. You can see that this sample has very rough sidewalls and away

from the ridge the oxide is rough resembling that of sample 3 (Figure 3-3.(a)).

From these results, I conclude that in order to obtain a smooth oxide surface,

the current density must be varied to account for differences in layer doping

and composition. This statement is supported by other experiments

investigating pulsed anodic oxidation [Huds93]. In this work the investigator

continually varied the current density to maintain a constant color of oxide.

While a degree of accuracy and reproducibility is lost by changing the current


Figure 3-4. A SEM of sample 4 with a greater magnification.

--- -----~'''--~P~~"~~`~- ~ '~41


;z -

* : ~i4

Figure 3-5. SEMs of the facets of laser bars. (a) L473 sample
number 5. (b) A 5 micron ridge from a different wafer.


5 M O R G


during anodization, SEMs of these samples showed smooth sidewalls for

ridge heights as large as 3 microns.

3.4 Threshold Current

As stated before, a key parameter in the performance of any laser is its

threshold current, Ith. The threshold current can be defined as the amount of

current required to make the mode gain in a device compensate the losses.

To get an insight as to what factors influence mode gain and mode loss they

must be more accurately defined. An accepted definition for the gain of an

optical wave propagating in the z direction can be written:

g = (3.1)
g (dz

where g is the gain in cm1, and 0 is the photon flux in cmnls1. In general,

the gain will be a function of the photon frequency, oa For quantum well

structures, the following expression can be derived for spectral gain [Zory93].

g(h) = ( )CIMT 2red(Eeh- Eg)(fc- -f (3.2)

where MT is the transition matrix, pred is the reduced density of states, Eeh is

the transition energy from electron to hole states, f. represents the quasi-

fermi level in the conduction band, and f, represents the quasi-fermi level in

the valence band. Strictly speaking, C is not a constant, but the dependence

on co is small enough that for now it will be ignored. When a forward bias is

applied to a laser diode, carriers will be injected into the active layer, a carrier

density is established in the active layer, and the separation of the quasi-fermi

levels will be increased. As shown in Eq. (3.2), as fc f is increased, the gain

will also increase. This spectral gain will have a peak value, gp, at some

wavelength, and normally the laser will operate at this wavelength. One

final correction is required, the available peak gain will only be used if there is

an optical field passing through the region with gain. To correct for this, the

peak material gain is multiplied by a confinement factor, F. This confinement

factor is defined as the fraction of the total optical field that is propagating in

the gain region. To obtain the mode gain, Gm, the peak gain is then multiplied

by the confinement factor.

The previous paragraph developed expressions for the mode gain and

the mode loss. To predict threshold current density, a relationship must be

established that couples gain with current density. When injecting carriers

into a direct bandgap region of high quality material, radiative

recombinations are the dominant recombination mechanism. An expression

for the radiative recombination rate can be derived as follows [Zory93]:

Rs (h)= m ')C2lMavg 2Pred (Eeh- Eg) Popt (ho)) f (1- f) (3.3)

where Mavg is an average transition matrix, Popt is the density of optical states

that a photon can be emitted into, and C2 is a factor not strongly dependent on

co. By integrating Rsp over all frequencies and multiplying by the electronic

charge, q, the radiative current density is obtained. A non-radiative current

density component may also be present, but for well designed, well grown

AlGaAs lasers, Auger recombination, SRH recombinations, and leakage

current can all be ignored [Reis87]. This radiative current is the link to the

gain. For each applied voltage there is a given injected carrier density, for

each carrier density there is a peak gain and a current density. The collection

of the gp, J pairs establish the relationship between the gain and the current

density. This analysis is often called the g-J method. The threshold current

condition can now be defined more rigorously using these formulas.

After the optical field makes one round trip in the laser cavity the

losses, Gth, can be written:

Gt = [a + 11( )] (3.4)
th = [i RoRb

where, ci is the internal mode loss, L is the cavity length of the laser, and Ro

and Rb are the power reflectivities of the output and back mirrors respectively.

The internal mode loss can be broken down further:

ai = as + af + (1 I) aw (3.5)

ase is defined as the scattering losses in the waveguide, afc represents the

losses from free carrier absorption, and awg represents the losses from areas

outside the active layer. The free carrier absorption will be greatest in the

active layer, so it is multiplied by the confinement factor. The cavity lengths

and facet reflectivities were the same for all the lasers tested, therefore, from

Eq. (3.4) the differences in Gth will be caused be differences in ai. If the

relationship between g and J does not change significantly from samples 1

thru 5, then the samples with the lowest ai will have the lowest threshold


The threshold of the fabricated lasers was measured by two methods,

some were measured in bar form, and some were tested in chip form mounted

on copper blocks. The lasers were tested in a pulsed mode with a short

pulsewidth, 0.5 ms, and a low repetition rate, 1 kHz, to reduce the effects of

heating and other transients. The voltage pulse source was an HP214A

pulser. The output of the laser was collected by a large area PIN 10D Si

photodetector. The test system has a current transformer that reads the

current flowing through the system. The optical power and the current are

sampled by a boxcar average and the output is used to create a data file of

optical power, P, versus current, I. A typical P-I curve of a laser is shown in

Figure 3.6. The threshold is determined from the current intercept of the

straight line where the optical power suddenly increases. The threshold for

Figure 3.6 is about 14 mA. The results of the threshold measurements are

summarized in Table 3-3. Slope efficiency is also included in the table and

will be discussed later in the chapter.

The results of the measurements show a decrease in threshold current

with increasing ridge height. this is to be expected since an increase in ridge

height will increase the amount of index guiding in the plane of the active







0 10 20 30 40
Current (mA)

50 60 70

Figure 3-6. A typical P-I curve for sample 5.

Table 3-3. Threshold Current and Slope Efficiency

Ith TIs
Ridge Average Standard Average Ts Standard
Sample Height (A) Ith (mA) Deviation (W/A) Deviation
(mA) (W/A)
1 1530 38.4 2.4 N/A N/A
2 2770 28.5 1.3 N/A N/A
3 4290 27.1 3.1 N/A N/A
4 11,770 13.8 2.1 .592 .051
5 15,970 14.5 1.5 .740 .101

layer. By increasing the guiding, r will be increased, and less of the wave will

propagate in the unpumped regions of guide. Mathematically, the third term

in Eq. (3.5) will be decreased. The second term of that equation will be

increased but since the relative contributions of the free carrier absorption is

much smaller than band to band absorption, the net result should be a

reduction in ai and a reduction in threshold. It is interesting to note that no

decrease in threshold current is observed from sample 4 to sample 5, this is in

agreement with the theory of the transition from gain to index guiding

[Agra84]. In his theory, Agrawal predicts that once the difference in the mode

index under the ridge and the mode index away from the ridge, AnL, exceeds a

value of approximately 5 x 10-3 the device is in the index guided regime. No

reduction in threshold current will be realized for AnL values greater than 5 x

10-3. From the data for samples 1-5, it appears that this transition has

already taken place for values of ridge heights 11,700 A. A graph of threshold

current versus ridge height is shown in Figure 3.7.

3.5 Slope Efficiency

Another key parameter in laser performance is the slope efficiency, ris.

Slope efficiency is defined as the slope of the P-I curve above threshold and is

illustrated in Figure 3-6. An expression for the slope efficiency can be


rl P q s hv
I [i.. hvLJ = 7[1D][Va (3.6)

10 I -
0 0.5 1 1.5 2

Ridge Height (microns)

Figure 3-7. Threshold current versus ridge height for samples 1-5.
The stripe dimensions were 5 x 500 microns. The lasers were pulse
tested at room temperature.

where ID is the differential quantum efficiency, Vc is the characteristic

voltage and hv is the bandgap energy. We can further define TID as:

1TD = Tilo (3.7)

where Tl0 is the output efficiency defined as the output photon rate divided by

the photon generation rate, and Tli is the internal quantum efficiency.

The slope efficiencies for the fabricated lasers are listed in Table 3-3.

No values for samples 1-3 are listed. This is because in the gain guided

regime, the lasers do not posses a linear P-I characteristic. A typical P-I

curve for a gain guided laser is shown in Figure 3-8. This data was taken from

a laser made from sample 1. The kinks in the slope may be a caused by self-

focusing, index antiguiding, onset of higher order modes, carrier leakage

through the oxide, or a combination of any of the above [Thom83]. This

nonlinear P-I behavior is one of the most unfavorable characteristics of gain

guided lasers. The P-I curves of index guided lasers usually do not exhibit

kinks, but can display a non-linear rollover at high-voltages. High voltage

rollover for index guided lasers can be caused by carrier leakage through the

oxide. To see if the index guided lasers from sample 4 and 5 displayed this

rollover behavior P-I curves were taken on these samples up to facet blow out.

Facet blow out is caused by catastrophical optical damage, COD. Non-

radiative recombinations occur at recombinations sites at the cleaved surface

which produce phonons and local facet heating. This facet heating produces a

0 ------
0 20 40 60 80 100 120

Current (mA)

Figure 3-8. A typical P-I curve for a narrow stripe gain guided
laser. This was a 5 x 500 micron laser fabricated from sample 1.
The test was under pulsed conditions at room temperature.

local bandgap reduction, which in turn increases optical absorption. The

added absorption increases the number of carriers to participate in non-

radiative recombinations. This cycle of absorption and bandgap reduction

continues until the facet temperature is so large that the facet blows out. The

reason I bring up COD is that it represents the upper limit for the P-I curve.

If a laser's slope efficiency is linear up to COD then it is truly linear. If the

laser shows no rollover up to COD then the oxide is as good as it has to be.

Figure 3-9 shows lasers from sample 4 and sample 5 up to COD. The figure

shows that sample 5 is linear up to COD, and sample 4 is linear up to 150 mA,

80 mW. One reason for the rollover of sample 4 could be because the oxide of

sample 4 is slightly thinner than that of sample 5. Another reason for the

rollover is the difference in slope efficiency. Stronger index guiding in sample

5 will lead to a better overlap of the optical field and the gain distribution.

The increased overlap will lead to a greater slope efficiency. For similar I-V

characteristics a large slope efficiency will allow COD to be reached with a

smaller voltage. This will be discussed in detail in Chapter 4.

Another important feature of the slope efficiency is that it provides a

way of a obtaining the injection efficiency and mode loss for a wide area laser.

From Eq. (3.6) and Eq. (3.7) we can write the slope efficiency:

T11 = 'lo1iVc (3.8)

it can also be shown that the output efficiency can be written [Lim91]:






0 50 100 150 200 250

300 350

Current (mA)

Figure 3-9. A comparison of slope efficiencies for lasers from samples 4
and 5 up to COD.


1 1
2L R0
r1o = 1 1 (3.9)
ai + -ln Rb

combining the results of Eq. (3.8) and Eq. (3.9) we can write an expression for

the inverse of the slope efficiency as:

1 2 ai RoRb
= + (3.10)
Ts Vcln J TiVln 1

for well behaved lasers, a plot of the inverse of r1, versus cavity length will

produce a straight line. The intercept of that line when L =0 will yield Tli.

Once you have solved for ri, the slope of the line will yield ai. Measurements

were done on wide stripe lasers from sample 1, with cavity lengths of 500, 750,

1000, and 1250 microns. The results are shown in Figure 3-10. Using this

data the calculated values of li and ai are approximately 1.0 and 3.6 cm-1

respectively. For regions where the slope efficiency is linear, you can show

[Lim91] that the internal quantum efficiency is the same as carrier injection

efficiency, 'rin. For these lasers the internal efficiency, and therefore the

injection efficiency, is approximately 1. The conclusion from this data is that

L473 is a very good wafer. The losses are very low, which is indicative of high

quality growth. The injection efficiency is approximately 1, indicating there

is no leakage of carriers out of the carrier confinement region.






1.3 I
400 600 800 1000 1200 1400

Cavity Length (microns)

Figure 3-10. Plot of inverse slope efficiency versus cavity length for
sample 1.

I performed one additional experiment regarding Tls. To measure the

effect that the anodic oxide had on T1s, I took a piece of sample 3 and stripped

the oxide prior to metallization. In this manner I fabricated a laser that did

not have any high conductivity material outside the stripe region but also did

not have any oxide. Figure 3-11 shows the P-I curves of two 5 pm stripe

lasers from sample 3. The only difference in the two lasers is that one has

pulsed anodic oxide and one has no oxide. Several features can be pointed out

on this figure. First, the threshold of the stripped laser is significantly higher,

approximately 90 mA versus 30 mA. Second, the turn on of the stripped laser

is very soft, that is, the transition from low output, spontaneous emission, to

high output, stimulated emission, is very gradual. Third, the slope efficiency

of the stripped laser is comparable to that of the oxide laser at higher

currents. I believe this is because the stripped laser never operates in the

fundamental mode. It turns on with multiple modes just as the non-stripped

laser supports higher modes at larger currents. The reasons for these results

are a combination of current spreading, current leakage, and mode guidance.

All of these factors will be discussed in Chapter 4.

3.6 Near Fields

The near field of a laser is the spatial distribution of the optical

intensity on the output facet of a laser. It is important because below

threshold, the near field gives a good approximation of the carrier

distribution. This carrier distribution can be used to infer the amount of



50 100 150


Current (mA)

Figure 3-11. A comparison of a laser with anodic oxide and one with
the anodic oxide removed. The lasers were from sample 3 and had
stripe widths of 5 microns.

current spreading and the gain distribution. Above threshold, the near field

provides a measure of the optical mode guiding mechanism.

The experimental set-up used to measure the near fields is shown in

Figure 3-12 (a). The output facet of the device was focused on a CCD

television camera with a 50X microscope objective. The camera had a gamma

factor of one (the gamma factor is the relationship between input intensity

and the voltage output of the camera). The output of the camera was

connected to a video analyzer. One output of the analyzer was connected to a

monitor so the near field could be observed directly. The analyzer allows you

to select and scan a video line. The result of this scan is to produce a voltage

pair corresponding to the position and intensity along the line. To take a near

field measurement, the device current was selected and the x-y-z positioner

used to focus the facet on the camera. The scan line on the analyzer was

superimposed on the near field image and the line was scanned to produce a

data file of position versus intensity. According to the equipment manual, the

linearity of all the pieces of equipment was very'high. However, to test the

linearity of the system, a reference near field was taken, and then a second

near field was taken with a neutral density filter inserted between the

objective and the camera. The results showed that the two profiles were

displaced by equal amounts verifying the linearity of the system. The output

of the video scanning system was a time-ordered pair of voltages as mentioned

above. Since we are only interested in relative near field intensity, the

absolute value of the corresponding voltage is not important. This is not the

CCD Camera


CCD Camera

Figure 3-12. Schematic of systems to measure (a) near fields and (b)
far fields.





case for the position measurement. To calibrate the position voltage, the facet

of a 5 micron laser from sample 5 was back illuminated. The image was then

focused on the camera and the voltage difference from one edge of the ridge to

the other was measured. Using Dektak and SEM measurements, the actual

distance was measured and a calibration factor of 88.1 microns/volt was

obtained. To avoid calibrating each measurement, the objective and the

camera were both fixed on the x-y-z positioner. Because the distance from lens

to camera was fixed, whenever the facet was focused, the distance from facet

to lens must also be fixed and the same calibration factor can be used.

The results of each of the measured near fields was plotted along with

two theoretical distributions. If the gain distribution is proportional to x2,

then a gaussian near field will be the result [Cook75]. If the gain distribution

is proportional to cosh'2(2x/s), where s is the stripe width, then a near field of

the form cosh-2U(2x/s) will result, where u is a complex parameter that

describes the distribution of the gain [Thom80]. A short review of these

derivations and the advantages of each will be included in Chapter 4. To see

how these distributions compare, a plot of the near field with fitted

distributions is shown in Figure 3-13. The cosh2u(2x/s) distribution provides

a better fit for gain guided devices. However, for the deeper ridges the u

parameter becomes large and the two distributions merge.

I tried to measure enough lasers to give a representative analysis of the

near field behavior. Since the current will affect the near field, I had to

measure the near fields at different current levels. I chose four lasers for each






-20 -15 -10 -5 0 5 10 15 20

Position (microns)



| 0.8 ....

S0.2 --.---
S0.2 ------ -^


-20 -15 -10 -5

0 5 10 15 20

Position (microns)

Figure 3-13. Comparison of near fields and theoretical distributions
for sample 1. (a).... near field, cosh-2u(2x/s), (b).... near field,

sample with three current levels per tested device making a total of 60 near

field measurements. Unfortunately, repeated handling of these devices may

lead to damaging of the facets. The near field will be significantly affected if

the facet is scratched or gets dirt on it. Therefore, if the laser had a

significantly asymmetric and irregular near field it was not considered when

calculating the average. The results of the measurements are shown in Table

3-4. At higher currents, the gain guided devices no longer had single lobed

near fields and so the full width at half maximum, FWHM, has no meaning.

For this reason, near fields with multiple lobes were not reported in Table 3-

4. The parameter ul,used in the cosh-2u(2x/s) fit, is also included. Above

threshold, the FWHM of the near fields for samples 1, 2, and, 3 got smaller

with increasing current. This behavior continued until the near fields were no

longer single lobed. The FWHM of the near field decreased above threshold

for samples 4 and 5 but then remained constant with increasing current. A

representative narrowing of samples 1 thru 3 is shown in Figure 3-14. (a),

representative narrowing of samples 4 and 5 are shown in Figure 3-14. (b).

3.7 Far Fields

The far field of a laser is the spatial distribution of the optical intensity

away from the facet. Because the amplitude and phase of the optical field are

functions of position, the optical intensity distribution at the facet may be very

different from the distribution several centimeters away. However, once the

Table 3-4. Measured Near Fields

Near Field
Sample Current (mA) ( ul Parameter
FWHM (gm)

1 below threshold 17.5 .15
2 below threshold 15.4 .15
3 below threshold 13.2 .2
4 below threshold 5.4 1.0
5 below threshold 4.0 1.5






-20 -15 -10 -5

0 5 10 15 20

Position (microns)





-10 -5 0 5

Position (microns)

Figure 3-14. Representative near fields at various currents, (a) Sample 1
in the gain guided regime, (b) Sample 5 in the index guided regime.


optical field has traveled a certain distance the angular distribution of the

intensity will no longer change. This minimum distance is given by [Hent88]:

r > (3.11)

where d is the source size and X is the wavelength. For the lasers under study,

this distance is about 30 microns, the actual measurement distance was

several centimeters. The experimental set-up for the far fields is shown in

Figure 3-12. (b). The laser illuminates a light diffusion screen and the image

of the back of that screen is focused onto a CCD camera. If the diffusion

screen is not used, the resulting image will be the laser spot and not the far

field. Because the far field is so much larger than the near field, the 50X

objective is not required and a small lens attached to the camera provides

adequate magnification. Recording the outputs of the system is similar to the

near field setup. The output is a time-ordered voltage pair corresponding to

an intensity and an angle. As with the near field, the intensity is an arbitrary

measure and no calibration is required. To calibrate the angle measurement,

I measured the near field with an alternate system. In this measurement, a

small area detector is fixed at a distance several centimeters from the laser.

The output of the detector is read by a lock-in amplifier because the optical

signal is small off the beam axis. The laser is mounted on a special calibrated

rotational stage. As the beam is manually rotated through a series of angles

the intensity at each angle is recorded. The plot of relative intensity versus

angle produces a calibrated far field. This same laser, at the same current

level, was then tested on the automated setup and a calibration factor of 20.3

degrees per volt was obtained.

The results of each of the measured far fields was plotted along with a

theoretical gaussian distribution. The near field and far field are fourier

transform pairs; therefore, if the near field is a gaussian, the far field will be

gaussian as well. Representative far fields are shown in Figure 3-15. Far

field measurements were done on all lasers with well-behaved near fields.

The lone exception was the far field of sample 3 that had the anodic oxide

stripped. The far field of this laser was taken to verify that it was operating in

higher order modes just above threshold. The far field was multilobed at all

currents indicating that without the oxide, the laser never operated in a

fundamental mode.

Above threshold, the FWHM of the far fields for samples 1, 2, and 3

got larger with increasing current. This behavior continued until the far fields

were no longer single lobed indicating the onset of higher order modes. The

FWHM of the far fields for samples 4 and 5 remained the same above

threshold. Gain guided behavior of samples 1 thru 3 is shown in Figure 3-15.

(a), index guided behavior of samples 4 and 5 are shown in Figure 3-15 (b).

3.8 Lifetime Testing

Lifetime tests were conducted to study how lasers fabricated with

pulsed anodic oxides perform under CW conditions. In order to conduct the

CW tests, the lasers had to be processed further. Small copper block heat






0.2 .................... i----------

-10 -5


1 ....................|................






15 -10

Angle (Degrees)

-5 0 5 10 15

Angle (Degrees)

Figure 3-15. Representative far fields at various currents, (a) Sample 1
in the gain guided regime (b) Sample 5 in the index guided regime

sinks had thin layers of Ni and Au evaporated on them followed by a thicker

layer (3 microns) of In. The blocks were coated with flux to facilitate the In

bonding. Lasers from sample 2 were cleaved into 500 x 500 micron chips and

mounted p-side down on the copper blocks. For CW operation, the chips must

be mounted epi-side down to allow for adequate heat transfer away from the

active layer. The samples were flash -bonded and then cleaned to remove the

excess flux. Au covered contact rails were mounted on the copper blocks and

the n-side of the chips were wedge bonded to these rails.

The method of lifetime testing I selected was to choose a reference

optical power and then measure the current required to maintain that power

over time. To begin the test, 50 mw /facet was chosen as the reference power

level. A laser was mounted on a TE temperature controller and connected to a

CW current source. The current was set at the level required to maintain an

output power of 50 mW/facet. The TE controller was set to 180 C and the test

was begun. At regular time intervals the power was checked and the current

was adjusted to maintain 50 mW/facet. The laser stabilized after an initial

burn-in period of about 80 hours. The performance was so good that 500

hours into the test, the output power was increased to 100mW/facet. The

laser operated in the CW mode for more than 1200 hours. A plot of current

versus time is shown in Figure 3-16.(a). It should be noted that these devices

were tested without facet coatings; therefore, it is possible that the slight







0 200 400 600 800
Time (hours)






0 200 400 600 800

1000 1200

1000 1200

Time (hours)

Figure 3-16. Lifetime performance of lasers in CW mode. (a) Current
versus time for a constant level of output power. (b) Slope efficiency
versus time. Test were done at 180 C on 50 x 500 micron devices from
sample 2.

100 mW/facet

50 mW/facet

..................... .......................................... ..................... ...........................................

..................... .................... ......................................................................................

........................................... ........................................... .................... .....................


degradation observed is due to facet problems. If the oxide is breaking down,

then there should be a sizeable drop in slope efficiency over time. As shown in

Figure 3-16.(b) no such decrease is observed.


4.1 Introduction

In Chapter 3, I presented experimental data on the performance of

lasers fabricated with pulsed anodic oxides. In this chapter, I will analyze the

results and establish the effects that anodic oxides have on laser performance.

The two main areas that the oxide will affect are carriers transported from the

contacts to the active layer, and optical waveguiding in the device. Two

aspects of carrier transport will be examined; current spreading and current

leakage. The primary waveguiding issue that will be addressed is how the

ridge height affects the transition from a primarily gain guided structure to a

primarily index guided structure.

One factor that will clearly affect the performance of pulsed anodic

lasers is lateral current spreading. The top epilayer in most lasers has a high

conductivity p+ cap to facilitate metal contact. In the anodic process, this cap

is consumed everywhere except under the stripe. In stripe geometries, this

cap is not etched and allows for the lateral spreading of current outside the

boundaries of the stripe. The injected carrier density decreases as the

current spreads and the threshold current increases correspondingly. Lateral

current spreading should become smaller for the lasers fabricated with

greater ridge heights. Near fields will be analyzed to see how ridge height

effects this current spreading.

Another factor impacting on current flow is leakage current. For the

purposes of this work, I will examine leakage current from two sources. One

source of leakage is injected carriers that escape the confinement barriers of

the active layer. The second source is carriers that penetrate through the

oxide and are injected into the active layer away from the stripe region.

Measured slope efficiencies will be used to analyze how the anodic oxide

effects current leakage.

The third major area of analysis is optical wave guiding. As discussed

in Chapter 1, the guiding mechanism of a laser impacts on virtually all of its

performance characteristics. In this chapter, I will take the results of the

measured near and far fields and use them to establish at what ridge heights

the lasers transition from gain guided to index guided performance.

4.2 Current Spreading

In order to understand the behavior of carrier transport across

heterointerfaces, a simplifying assumption is often made that the problem is

one dimensional. While this approach can provide valuable insight into the

physics of a problem, it is a situation that is rarely realized in actual devices.

Consider the geometry of lasers fabricated from sample 3, shown in Figure 4-

1. Clearly, the transport of carriers into the active layer is the major focus,

but because of the three dimensional nature of the structure, there will also

be lateral current spreading in the layers parallel to the heterointerfaces.

Contact Metal

Cap Layer


Figure 4-1. Schematic of current flow in lasers made from sample 3. Io is
the current flowing parallel to the junction. It is the total current, and It
=2Io + Ie.

There is less area for lateral spreading as the ridge height increases, thereby

decreasing the effective conductivity of the layers.

Quantitatively this problem has been treated for both stripe geometries

[Yone73], and ridge structures [Beis91]. Referring to Figure 4-1, we can write

the total current as:

It = Ie + 2I (4.1)

where It is the total current, Ie is the current under the stripe, and Io is the

current flowing in the y direction on one side of the stripe. The current across

a junction between y and y + dy can be written

-dI = LJ,(exp((Q )-1))dy (4.2)

where L is the laser cavity length, J. is the saturation current density, Vy is

the junction voltage at y, n is a constant, and the other variables have their

normal meanings. The voltage drop in the cladding layer -dVy can be written

from Ohm's law:

-dV, = PLC Idy (4.3)

where Pcl is the resistivity in the cladding layer, and dci is the thickness of the

cladding layer. Combining Eq. (4.2) and Eq. (4.3) we can obtain a differential

equation for Iy:

d I LJexpqVy (4.4)
dy2 S nkT)Ldc1
The solution of this equation is:

Iy = (4.5)
S yqpcIo 1 +
2nkTLdl 1

now the difference in the current across the junction at y and y + Ay becomes:

AIj (y) = Ay (4.6)

using Eq. (4.1) and Eq. (4.6) you can solve for Io:

[1+( sqpcl )I] 1
2LnkTdcl)It -1
I0 (4.7)

where s is the stripe width of the laser. Finally, if you define a current density:

J Idy (4.8)

where L is the laser cavity length. When L is normalized to 1, you can use Eq.

(4.6) to write the current density across the junction at y, Jy(y) as:

J (y) = 2 (4.9)

It has already been shown, [Grov93] and [Huds93], that using anodic oxides

significantly reduces current spreading when compared to stripe geometries.

This is because the high conductivity cap is always consumed with anodic

oxidation. When comparing ridge structures of. varying heights, any

differences in lateral spreading will be the result of the remaining clad and

confinement layers.

To see how well this theory predicts lateral current spreading, the

predicted current density distribution was compared to the measured near

field of a laser. Below threshold, the observed near field results from the

spontaneous recombination of injected carriers. Therefore, the near field

should give a fairly accurate representation of the-carrier distribution. Figure

4-2 shows a comparison of current spreading and the near fields below

threshold. Figure 4-2 (a) was calculated using Eq. (4.9) with a current value

of 10 mA. The current density was calculated and normalized for samples 1,

3, and 5. Figure 4-2 (b) is a plot of the near fields for the same samples.

There is good qualitative agreement between the two graphs. The differences

arise because Jy(y) is actually the current density outside the stripe region

while the near field plot is from the center of the stripe. Also, the model does

not consider the effects of carrier diffusion. Carrier diffusion will act to

smooth the carrier distribution from areas under the stripe to areas away





0 5 10 15 20 25

Position (microns)






0 5 10 15 20

Position (microns)

Figure 4-2. A comparison of current density and optical distribution,
(a) theoretical current density versus position for a current of 10 mA, (b)
near fields of three lasers below threshold.

from the stripe. A third source of uncertainty is the actual stripe width of the

laser. In Figure 3-4 and Figure 3-5, SEMs of the deep ridge laser show a

slight amount of undercutting by the oxide. If the stripe width is changed, the

corresponding amount of spreading will also change. Eq. (4.9) was evaluated

for stripe widths of 5, 4.5, 4, and 3.5 microns. The changes in Jy(y) were very

small indicating that the current spreading outside the stripe is not very

sensitive to stripe width. This undercutting will affect the shape of the near

field because the total current density distribution is Jy(y) on the left of the

stripe, plus the stripe width, plus Jy(y) on the right hand side of the stripe.

The important point of this section can be summarized as follows.

Since the high conductivity cap is consumed in the anodization process, anodic

lasers have significantly less current spreading than stripe geometry lasers.

However, current spreading still occurs in ridge lasers of varying ridge

heights. Increased current spreading means that for a given drive current,

the current density in the active layer is smaller; a larger area is being

pumped and a greater current is needed to reach Jth. Therefore, at least from

a current spreading perspective, the larger the ridge height the smaller the

threshold current.

4.3 Current Leakage

In an ideal laser, any carriers that are injected across the ohmic

contacts will eventually occupy energy states in the active layer and

participate in radiative recombination. Any mechanisms that prevent carriers

from doing that will increase the threshold and decrease the slope efficiency

of the laser. Current leakage is a mechanism that will reduce the number of

carriers that participate in radiative recombination. 'I will analyze two types

of current leakage, leakage through the oxide and leakage across the carrier

confinement barriers.

4.3.1 Leakage Through the Oxide

Referring to Figure 4-1, you can see that the metal contact of a laser

extends beyond the stripe width. As long as the resistance of the oxide is large

compared to the resistance in the stripe region, current will flow down the

path defined by the oxide and into the active layer. No matter how good the

oxide layer is as an insulator, when the voltage gets sufficiently large, it can

drive a fraction of the carriers through the oxide layer. These carriers may be

injected into the active layer but they may be far from the stripe region, and

therefore not contribute to the lasing mode. The slope efficiency of the device

is reduced when this occurs. This can be observed as a rollover in the P-I

characteristic at higher voltages. Leakage through the oxide should not

increase Ith because it only occurs at higher voltages. If the laser has a

relatively good I-V characteristic, then the voltages required to obtain Ith

should be low.

As mentioned in Chapter 3, the slope efficiency, I-V characteristic, and

leakage through the oxide are linked for narrow stripe lasers. If the device

has a good I-V characteristic, then only moderate voltage levels are required

to produce large currents. If the device has good slope efficiency then a large

current will produce high optical powers. When the optical power becomes

large enough, COD occurs; therefore, the maximum voltage that an oxide

must be able to withstand is intrinsically related to the I-V characteristics and

slope efficiency of the device. This behavior is illustrated in Figure 3-9. Since

the P-I for sample number 5 is linear until COD is reached, clearly no leakage

through the oxide occurs. In the same figure, the slope efficiency for sample

4 rolls over at higher voltages. A possible explanation for this data would be

that the carriers are leaking through the oxide, but making this assumption

on the basis of the P-I alone would be incorrect. In Chapter 2, I reported that

tests on anodic oxides showed that the oxide would block current up to

approximately 10 V. To accurately see if the rollover of slope efficiency is

caused by leakage through the oxide, you must also analyze the I-V


A typical I-V characteristic for a laser diode is shown in Figure 2-5.

Significant current does not flow until the voltage exceeds the bandgap

voltage, Eg/q. Once the bandgap voltage is exceeded, the slope of the I-V is the

resistance of the laser. For the laser shown in Figure 2-5, the resistance is

approximately 8.5 Q. For all the lasers in this study, the resistance above

threshold varied from 5 to 15 Q The rollover for sample 4 begins at about 100

mA; even for the worst case resistance, this would represent a voltage of 0.1

times 15 plus the bandgap voltage of approximately 1.55 V for a total of 3.05V.

This is well below the voltage found to drive current through the oxide.

Care must be taken when analyzing P-I data because there are other

factors that can cause rollover. Thermal effects can cause rollover but these

are usually important only at high duty cycle operation [Hata93]. Other

causes of rollover are carrier induced guiding and antiguiding [Kirk77] and

evolution of higher order modes. Further evidence that rollover is not always

caused by leakage through the oxide is seen in Figure 3-8. For this gain

guided laser, the slope efficiency starts linear, begins to rollover, then starts to

increase again. The slope efficiency would continue to decrease with higher

voltage, if this rollover was caused by leakage through the oxide. This laser

begins to rollover at 60 mA. Using the worst case scenario, this would

correspond to a voltage of 2.45 V, well below the 10 V where leakage through

the oxide is significant.

The primary difference in all the samples tested was the ridge height.

As stated several times before, this difference will affect the guidance

mechanism of the laser. An analysis of carrier induced guiding, onset of

higher order modes, and other factors that affect the slope efficiency will be

addressed in the following sections on guiding. In summary, current leakage

through the oxide is definitely not a problem for sample 5 and is not likely to

be a problem for the other samples.

4.3.2 Leakage Across the Barriers

In the SQW-GRINSCH structure, electrons from the n-side and holes

from the p-side are injected into the potential well. If the structure is poorly

designed, injected carriers will escape the well and become minority carriers

in the confinement or cladding layers. Possible causes for carriers escaping

the well could be insufficient band offsets in the confinement layers, operation

at high temperatures, or high mode losses that would require very high

carrier densities to reach threshold. If there is leakage across the barriers,

the interface between the anodic oxide and the semiconductor could

theoretically be important. If free electrons were in the p-side, the presence of

a large number of interface states at the semiconductor-oxide interface could

influence carrier transport. For shallow ridges, the distance from the active

layer to the oxide interface is greater than 1 micron. Because the cladding

layers have doping densities on the order of 2 x 1017 or higher, the

recombination rate of the electrons on the p-side would be high and the

diffusion length would be small. In this case, the states at the oxide

semiconductor interface would have little effect. The worst case would be for

deep ridges where the distance from the active layer to the oxide interface

may be 500A. However, even in this case, I think that recombination at the

oxide should not be significant. From the arguments in the section on current

spreading, most of the carriers will be confined to the region under the stripe;

therefore, if an injected electron did escape the potential well it would be

under the stripe, not near the oxide and again the condition of the interface

becomes unimportant. Therefore, I think that the effect of the oxide

interface on carrier transport will be negligible.

I conducted several experiments in an attempt to show that the

condition of the oxide semiconductor interface does not significantly impact

laser performance. In one experiment I cleaved off a piece of sample 2 after

anodization and prior to metallization. This piece was again cleaved into two

smaller pieces A and B. It is widely accepted that anodic oxides grown on p

type GaAs have a large number of interface states [Mein78],[Kohn78]. It has

also been reported that low temperature annealing greatly reduces the

density of interface states [Hase75]. Therefore, I annealed piece A for 1 hour

at 3000 C in forming gas. After this anneal, the two samples were processed

into lasers and their P-I characteristics were measured. No differences in the

threshold currents or slope efficiencies were found for the two devices.

In a second experiment, lasers from sample 5 were cleaved to different

cavity lengths and tested. The inverse slope efficiency was plotted against

cavity length and a plot similar to Figure 3-10 was generated. This data

demonstrates that even for deep ridges the injection efficiency is 1. Since the

injection efficiency of the laser did not decrease at greater ridge heights, and

reduction of surface states did not affect the laser performance characteristics

we can say, at least for this structure, that the oxide semiconductor interface

does not impact on laser performance. Rigorous proof that this is true for all

cases will be left for future efforts.

4.4 Beam Astigmatism

Two requirements for every laser are a gain medium and a resonant

cavity. For semiconductor lasers, the resonant cavity is formed by a dielectric

waveguide bounded by two mirrors at the cleaved facets. In general, a

dielectric waveguide is formed by a boundary of two materials with different

University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - - mvs