Citation
Impact of the growth kinetics on deep level defect production in GaN films grown by molecular beam epitaxy

Material Information

Title:
Impact of the growth kinetics on deep level defect production in GaN films grown by molecular beam epitaxy
Creator:
Liu, Hsing-Long
Publication Date:
Language:
English
Physical Description:
vi, 108 leaves : ill. ; 29 cm.

Subjects

Subjects / Keywords:
Adatoms ( jstor )
Atoms ( jstor )
Emission spectra ( jstor )
Luminescence ( jstor )
Nitrogen ( jstor )
Sapphire ( jstor )
Shutters ( jstor )
Spectral bands ( jstor )
Specular reflection ( jstor )
Surface temperature ( jstor )
Dissertations, Academic -- Materials Science and Engineering -- UF ( lcsh )
Materials Science and Engineering thesis, Ph. D ( lcsh )
Genre:
bibliography ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1996.
Bibliography:
Includes bibliographical references (leaves 102-107).
General Note:
Typescript.
General Note:
Vita.
Statement of Responsibility:
by Hsing-Long Liu.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright [name of dissertation author]. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Resource Identifier:
026354467 ( ALEPH )
36714284 ( OCLC )

Downloads

This item has the following downloads:


Full Text










IMPACT OF THE GROWTH KINETICS ON DEEP LEVEL DEFECT PRODUCTION
IN GaN FILMS GROWN BY MOLECULAR BEAM EPITAXY













By


HSING-LONG LIU


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





UNIVERSITY OF FLORIDA

1996


.';:i.:;i:P '" ,.
:i.i., .....:::,i:. :ii .:. iiM i:: ...


















ACKNOWLEDGMENTS


I would like to express my greatest appreciation to Prof. Robert M. Park, my

supervisory committee chairman, for his constant professional guidance, support and

encouragement during the research and the writing of this dissertation. I would also like

to thank Prof. Paul H. Holloway, Prof. Joseph H. Simmons, Prof Kevin S. Jones and

Prof. Sheng S. Li who kindly served on my dissertation committee.

I extend special thanks to George Kim and Paul Precure, my fellow graduate

students, whose assistance in this research is gratefully acknowledged, especially thanks to

George for providing AFM images of the sapphire substrate morphology. Special thanks

also go to Lynn Calhoun for sharing his knowledge of RHEED intensity monitoring.

Thanks also to George, Lynn, Paul and Brent Gila for their friendship.

I would also like to acknowledge Dr. Matthias Ludwig of Dr. Hummel's group

for his assistance with the PL measurements and for helpful discussions.

I would like to thank my parents, especially my father who passed away in 1993,

for their prayers and love. And finally, I thank my Savior, Jesus Christ.













TABLE OF CONTENTS


Page



ACKN OW LEDGM ENTS ....................................................... ................................ ii

A B ST R A C T .................. ......... ......... .................................................... ............. v

CHAPTERS

1. IN TR O D U CTIO N ..................................................................................... ......

M otivation and Objectives ................................................ .......................... 1
B background ................................................................................... ............. 5
Theoretical Models of Deep-Level Defects in GaN ....................... ........ ... 5
Experimental Observations of Deep Levels................... ...................7
Migration Enhanced Epitaxy (MEE) with reference to GaAs.................... 11
Migration Kinetics During MEE With Reference to GaAs ......................... 12
MOCVD Growth of GaN Using an Alternate Source Exposure
M ethod ................................... ........................... ....... ..... ...... 14
D issertation O utline............................................................................................ 15

2. MBE SYSTEM AND CHARACTERIZATION TECHNIQUES ........................ 19

Introduction .................................................. .................. ..................... ...... 19
M olecular Beam Epitaxy................................... ........................... ..... .. 19
Varian GEN II MBE System Configuration...................................................... 20
Substrate Temperature Calibration........................................................ 22
In-situ A nalyzer.................................................... ................................ 22
Ex-vacuo RHEED Specular Reflection Beam Intensity Monitoring
S y stem .................................................... ................ ............................ 2 3
N itrogen Free-Radical Source....................................................................... 24








iii











3. GROWTH BY CONVENTIONAL MBE AND CHARACTERIZATION
OF GaN / SAPPHIRE FILM S..................................................... ...... ........ ..32

Ex-vacuo Substrate Preparation........... ............................ ................................ 32
In-situ Substrate Preparation................................................................. ....34
Conventional MBE Growth of GaN on Sapphire........... ............ .............34
Characterization of Conventionally Grown GaN Films............................................ 36
SEM Analysis........................................................................ 36
Hall-effect Measurements and PL Analysis................................ ....... 36

4. INVESTIGATION OF A NOVEL GROWTH MODE........................................ 51

Introduction ................................................................................. ............. 51
AEEE Growth Procedures ................................................. ........................ 52
AEEE Growth Temperature Optimization...................................................... 52
Nitrogen Delay Time Optimization.................. .......................... 54
Ga Delay Time Optimization ................................................................ .... 54
Activation Energy of Ga Migration under N-Free Conditions ........................... 56
Impact of Impinging RHEED Beam on Ga Migration...................................... ... 57

5. INVESTIGATION OF GaN GROWTH KINETICS AND THEIR
CORRELATION WITH DEEP-LEVEL DEFECTS ................................... 81

Introduction ........................................ .... ................................ .......................... 81
M orphological Considerations........................................................................... 81
Speculation Concerning the Source of Deep-Level Defects Responsible
for the Y ellow -Band Em mission ....................................................................... 84
Influence of Ga and N Exposure Times and Delay Times.................................. 85
M odel for Ga Adatom M igration.......................... ......... ............................... 87
Ga Exposure Stage.................................................. .......................... 88
First Stage of Recovery ............................................... ...................... 90
Second Stage of Recovery ................................................. .......... ..... 91

6. CONCLUSIONS AND RECOMMENDATION ................................................ 99

REFEREN CE LIST............................................................................................... 102

BIOGRAPHICAL SKETCH ....................................................................................... 108







iv
NA' F H M.












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


IMPACT OF THE GROWTH KINETICS ON DEEP LEVEL DEFECT PRODUCTION
IN GaN FILMS GROWN BY MOLECULAR BEAM

By

Hsing-Long Liu

December, 1996

Chairman: Robert M. Park
Major Department: Materials Science and Engineering


As is presently understood, the luminescence behavior of GaN epitaxial material is

dictated by the presence of both extended defects (dislocations) and also native point

defects. Near-band-edge luminescence at around 365 nm (ascribed to excitons bound to

dislocations) is typically observed in addition to a, so-called, yellow band luminescence

centered around 550 nm (2.3 eV) whose origin is the subject of speculation.

The specific objective of this research was to establish a relationship between

growth kinetics and the production of deep level defects in GaN epitaxial films by

correlating morphological changes monitored in-situ and in real-time during epitaxial

growth with yellow band emission intensity variations recorded post-growth and ex-situ.

The role of Ga adatom migration on growing GaN surfaces was found to be

critical with regard to the production of deep level defects and it was found that a

nitrogen-free surface condition was optimum to promote efficient Ga adatom migration as








was the provision of a sufficient time delay following Ga flux exposure prior to subsequent

N flux exposure.

The critical kinetic processes, such as Ga migration, were monitored in real-time

during epitaxial growth using reflection high energy electron diffraction (RHEED)

specular reflection beam intensity monitoring. Based on this analysis and the noted

correlations between surface morphological changes and the yellow band luminescence

intensity variations, it is speculated that Ga vacancies ( associated with insufficient Ga

adatom migration) are responsible, in association with dislocations, for the deep level

emission.

An activation energy of 1.46 0.25 eV has been determined for Ga adatom

migration on N-free GaN surfaces and a model is presented that accounts for the kinetic

processes taking place during both Ga adsorption and migration phases of growth under

N-free (no N-flux arriving) conditions






























vi


*** ***,.*












CHAPTER 1
INTRODUCTION


Motivation and Objectives


Gallium nitride (GaN) has a direct band-gap of about 3.4 eV at room temperature

(RT) and when alloyed with aluminum nitride (AIN) and indium nitride (InN) has the

potential to be applied to the fabrication of light-emitting diodes (LEDs) and laser diodes

(LDs) operating in the blue-green and ultraviolet (UV) wavelength regions of the optical

spectrum.

World-wide interest in GaN was intensified by the recent realization of GaN-based

blue LEDs, blue LDs, and solar-blind photodetectors. The high-output GaN-based blue

LEDs are in high demand for outdoor giant displays, desktop flat-panel displays, and

traffic lights, as well as devices which emit white light by combining red, green and blue

emitters. Moreover, the blue LD will have far-reaching implications for space and under-

sea communications, as well as high-density optical data storage. Such short-wavelength

blue emitters will provide the small spot size needed to increase the storage capacity of

optical storage systems over that of the current near-IR emitter-based technologies and

provide faster data-transfer rates for computer and multimedia use, thus meeting the

growing demands of the computer industry. GaN also is a promising semiconducting

material having potential applications in electronic devices operating at high temperature,

high frequency, and high power.

Despite some impressive technological achievements of the last few years,

particularly by Nakamura el al. [1] of Nichia Chemical Industries in Japan (demonstrating

the first electrically-pumped blue laser based on the InGaN/GaN multiple quantum well







system), much work remains to be done to realize a low-threshold long-lived LD suitable

for commercial application. Further progress toward the development of stable,
continuous wave (CW) operation of III-V nitride LDs will require a thorough

understanding of how impurities and defects influence the quantum efficiency of these

materials. For instance, it will be important to understand the origin of deep level

defects in GaN which are known to cause yellow band luminescence, suppressing the

more desirable band-edge luminescence. The yellow emission band is commonly

observed in the photoluminescence (PL) spectra recorded from nominally undoped GaN
epilayers regardless of the substrates used (a-plane sapphire [2,3] c-plane sapphire

[2,4,5,6], silicon carbide (a-SiC) [7,8], Si (111) [9], and bulk GaN [10,11]). It is also

observed regardless of the crystal growth technique used (low-pressure metalorganic

chemical vapor deposition (LPMOCVD) [8], metalorganic chemical vapor deposition

(MOCVD) [4,11,12], metal-organic vapor phase epitaxy (MOVPE) [2,3,5,9], MBE [6,8,

10, 13,14] or high pressure methods [10]). Ramsteiner et al.. observed electronic Raman

scattering in resonance with the yellow luminescence transition and concluded that the

deep-level concentrations must be rather large to make Raman resonance observable [15].

There is little understanding of the deep-level defects in n-type GaN, however,

sources of the deep-level defects have been proposed, such as Ga vacancies (VGa)

[16,17,18,19,20], antisite defects (NGa and GaN) [21,22,23] and Ga interstitials (Gai)

[22,23,24]. All of these theories are based on the assumption of thermodynamic
equilibrium, but GaN films are currently grown under conditions that deviate significantly

from equilibrium. In addition, these calculations use atomic structure factors calculated

for neutral, spherical atoms. They do not take into account the effects of the charge

redistribution due to chemical bonding (ionic bonding) and the stress field induced by high

densities of dislocations and strong inhomogeneities present in the films. CL images, in

fact, have shown strong inhomogeneities [25]. A hypothesis has recently been proposed

that states that the deep-level broadband emission arises directly from the atomic structure








of dislocations or is associated with the clustering of native point defects (vacancies,
interstitials) at dislocations due to the lattice distortions induced by the dislocations
[25,26].

The work described in this dissertation was inspired by a desire to minimize the

concentration of deep level defects occurring in GaN epitaxial films through an

understanding of the factors affecting their production.

The highest quality GaN films grown today are produced by conventional

MOCVD and MBE at substrate temperatures in the neighborhood of 1000 C and 700 C,

respectively. Such high growth temperatures could result in Ga rich material due to

desorption of the N adatoms. At low substrate temperatures, however, the diffusion

rate of Ga atoms is slow compared to the growth rate of the GaN film, which can result in

Ga accumulating into small Ga droplets on the growing surface which can impede the

migration of nitrogen atoms [27]. Ponce et al.. reported that Ga droplets on the order of

10-100 p.m in size were found in the GaN substrates obtained from high pressure growth

methods [11]. These Ga droplets are very common even in MBE-grown GaN films,

particularly in films grown at low growth temperatures. In fact, Crawford et al. [28]

recently reported that excess surface Ga decreases the GaN formation rate when the

substrate temperature is too low or the Ga flux is too high in the MBE growth case.

A systematic study of Ga migration and incorporation at some optimum growth

temperature therefore seems important at this juncture with a view to controlling the deep-

level defects in GaN.

Ga-migration has been investigated on GaAs surfaces. For instance, Horikoshi et

al. [29] employed an alternate element supply technique using As and group m species to

grow AIGaAs/GaAs quantum well structures. These authors reported excellent PL

characteristics and smooth interfaces for films grown at a relative low growth temperature

and attributed these results to a dramatic increase in the Ga atom migration rate under As-

free conditions. Ewing and Greene [30] reported that growth on the Ga-terminated







(111) A face of GaAs was slow and produced films which had a tendency to form planar

surfaces. On the other hand, growth on the As-terminated (T I 1) B face was fast and

film surfaces were rough. The {111} planes of cubic GaAs are similar to the (0001) (or

basal) plane of wurtizite GaN grown on, for instance, the (0001) plane of sapphire. The

GaN basal plane has a polar configuration with two atomic subplanes, each consisting of

either the cationic or the anionic component of the binary compound. Therefore, an

alternating supply of Ga and N atoms may, in principle, be used to grow (0001) GaN

epitaxially, possibly at growth temperatures lower than those used for MBE-growth.

Such an alternate element exposure growth method has not, however, been reported

previously for MBE-GaN.

In the presently described work, a growth technique which we call, alternate

element exposure epitaxy (AEEE), is introduced in which the Ga and N fluxes do not

impinge on the surface at the same time. During AEEE, the Ga adatoms are assumed to

migrate rapidly on the growing surface under N-free conditions and by virtue of forming a

Ga-covered surface and, hence, a chemically active surface, the N sticking coefficient is

expected to be enhanced as well. In addition, the insertion of a Ga delay time ( the time

allowed to elapse between closing the Ga shutter and opening the N shutter ) should

allow Ga adatoms to migrate to step or kink sites and, hence, to be incorporated into the

GaN film. In this manner, point defects, such as Ga vacancies and Ga interstitials, which

are thought to be associated with the deep level emission in GaN may be minimized.

The specific objective of this research was to establish a relationship between

growth kinetics and the production of deep level defects in GaN epitaxial films by

correlating morphological changes monitored in-situ and in real-time during alternate

element exposure epitaxy with yellow band emission intensity variations recorded post-

growth and ex-situ.







Background


Theoretical Models of Deep-Level Defects in GaN

There are three main theoretical models which attempt to explain the origin of the

deep-level emissions observed from unintentionally doped n-type GaN. All of the models

predict that the deep-levels are related to native defects and impurities. Applying a tight-

binding Hamiltonian model, Jenkins and Dow [21] suggested that a Ga vacancy (VGa) can

act as either an electron or a hole trap with a p-like level in the gap, and the antisite defect

(Gaw) can also act as either an electron or a hole trap. Jenkins and Dow also suggested

that an N vacancy (VN) results in a shallow donor state with a p-like level in the

conduction band, and its s-like deep level occupied in the gap. This will be discussed later.
By performing first-principles calculations under the assumption of thermodynamic

equilibrium,Van de Walle and co-workers [16-19] predicted that nitrogen vacancies have

a rather high formation energy (-4 eV) in n-type, as-grown GaN; however, n-type

behavior has traditionally been attributed to nitrogen vacancies. Van de Walle claimed

that self-interstitials (Gai and Ni) and antisite defects (NGa and Gan) also have high

formation energies as shown in Fig. 1.1; therefore, they are also very unlikely to occur in

significant concentrations. However, these authors added that the quality of the GaN

epilayers can depend strongly on the concentration of line defects such as dislocations and

stacking faults, etc. The first-principles total-energy equation for a native defect in a

charge state q is as follows:

Ef(q) = E' (q) no, P Ga njv -P- qE------ (1.1)
where nG, and nf are the number of Ga and N atoms, p G. and /ny are the Ga and N

chemical potentials, E- is the Fermi energy, and E't is the total energy for a given charge

state. The chemical potentials for Ga and N are not independent, since both species are in

equilibrium with GaN. If boundary conditions such as p/w < pm, and p Ga < PGafbk) are







not satisfied, the system would be thermodynamically unstable and would segregate into

different components. Figure 1.1 shows that a Ga vacancy may have a low enough

formation energy to occur in high concentrations. Van de Walle et al. suggested that Ga

vacancies may play a role in the yellow band luminescence.

Boguslauski el al. [22,23,24] used an ab initio molecular dynamics approach in
which the computations were carried out for the case of the perfect crystal using quantum

molecular dynamics (MD) and the local density approximation (LDA) in supercells

containing 72 atoms. Due to the wide-gap of nitrides, the formation energies of defects

depend strongly on the position of the Fermi level. It is important to point out that

because of the inaccuracy due to the use of LDA theory which seriously underestimates

band-gap and overestimates the cohesive energy, an uncertainty on the order of 0.5 eV for

the formation energy of the deep levels may occur. This is especially true if the growth

process involves local excitations due to a plasma source or due to specific kinetic

conditions.

In n-type GaN, the high formation energy of VN (4.8 eV) would indicate that the N

vacancy cannot be formed in sufficient concentrations to account for the n-type character

of the as-grown material. However, since the formation of a Ga vacancy creates N

dangling bonds, its quasi-triplet level is located about 0.4 eV above the valence band edge.

VGa can trap both electrons and holes because the quasi-triplet is populated by 3 electrons

in the neutral charge state. In spite of the proximity to the valence band, the quasi-triple

wave functions are very localized. The formation energy of neutral VG. may appear

large at 8 eV; however, since VG, can release or accept three electrons, its formation

energy depends very strongly on the position of the Fermi level. In particular, in n-type

GaN films the formation energy of VG -3 may become very small. The thermodynamic

levels for Ga vacancies are at 0.5, 0.8, and 1.3 eV above the top of the valence band

maximum, respectively. Boguslauski [23] concluded that the small formation energy

defects in n-type GaN are Vac, GaN and NL







Experimental Observations of Deep-Levels.

Ogino and Aoki [31] investigated the mechanism causing yellow band

luminescence in GaN. A yellow band emission was observed in microcrystals

synthesized from Ga and NH3 by direct reaction, but did not appear in needle-like

crystals grown by sublimation-recrystallization. These authors proposed a mechanism

for the zero-phonon transition assigning the yellow-band emission to a radiative transition
from a shallow donor (25 meV below the conduction band minimum) to a deep acceptor

(860 meV above the valence band maximum). The deep acceptor was assigned to a

complex consisting of a Ga vacancy and a carbon atom. Ramsteiner et al. [15] recently

confirmed that shallow donors are involved in the yellow luminescence by studying the

electronic Raman scattering in resonance with yellow luminescence transitions in a 1 .m

thick GaN film grown on GaAs (001) by conventional MBE. No specific deep localized

state was assigned by Ramsteiner el al.

Tansley and Eagan [20] compared the experimental data on the location of the

levels associated with native point defects in the group-III nitrides with the theoretical

estimates. These authors found strong evidence for the existence of a triplet of donorlike

states associated with N vacancies and suggested that the three electron states associated

with N vacancies in GaN are located at about 30 meV, 100 meV, and 400 meV below the

conduction band minimum as shown in Fig. 1.2 (groups A, B, and C respectively). A

deep level has been observed with thermal activation energies between 0.8 and 1.1 eV

below the conduction-band edge as shown in Fig 1.2 as group D. This deep level has

been attributed to Ga vacancies [16] and N antisite defects [21]. Both Tansley and

Eagan [20] and Van de Walle et al. [16] proposed that group D is the obvious candidate

for the yellow band luminescence due to transitions between the deep-level state and the

valence band. Zhang et al. [32] observed a dramatically reduced deep-level

luminescence in a highly Ge doped (n-1020 cm'3) GaN film and suggested that the deep-







level luminescence emission may be related to Ga vacancies since Ge acts as a donor by

occupying Ga sites. Glaser et al.. [33] found an energy level for the deep-donor state
from optically detected magnetic resonance (ODMR) spectroscopy studies to be -1 eV

below the conduction-band minimum in GaN epitaxial layers grown by OMCVD, thus

agreeing with the group D energy level theory. Other deep traps (group E of Fig 1.2)

which are observed as a band-tail absorption feature at -1 eV above the valence band

maximum are believed to be related to Gas [21] or VGa [23].

Perlin et al. [22] investigated the pressure dependence of the energy of the yellow

band luminescence for n-type GaN films at hydrostatic pressures up to 30 GPa and

showed that the energy of the luminescence band changes linearly with pressure up to

about 18 GPa. Thus, at the highest pressure, the yellow band luminescence becomes blue.

Similarly, Shan et al. [34] studied the PL emission as a function of applied hydrostatic

pressure using the diamond-anvil-cell technique on a 4.2pm thick single-crystal GaN film

grown by MOCVD. The exciton emission lines were found to shift almost linearly

toward higher energy with increasing pressure and the yellow emission band showed a

similar blue shift behavior under applied pressure. Along with the conclusions of Ogino

and Aoki [31] and Ramsteiner et al [15], the report of Perlin et al.. supports group E as

a candidate for yellow band luminescence due to transitions between shallow donors and

deep acceptors. However, Shan et al. [34] attributed the yellow band emission to

transitions involving s-like levels associated with group D defects and the top of the

valence band maximum.

With a view to determining the distribution of the defects responsible for the 2.3

eV yellow band luminescence, AsifKhan et al. [35] performed low temperature PL

analysis on a 9.3-pm-thick LPMOCVD grown GaN/sapphire sample from the front and

back sides; the same yellow band emission intensity was found in both cases. Thus, these

authors concluded that the defects that were responsible for the yellow band emission had

a uniform distribution in the film. However, Cowan et al. [14] used








Cathodoluminescence (CL) to provide an analysis of traps as a function of depth in their
GaN films. For low accelerating voltage (V,. ), the electron probe had little penetration

so data were generated from the film surface. At higher Vacc, the electron beam

penetrated the film-substrate interface, giving a profile of yellow emission versus depth.

They observed that the intensity of the deep-level emission was higher at depths close to

the film-substrate interface than at depths just beneath the film surface. Shan el al. [36]

reported the results of time-resolved studies on the radiative decay on a 7 pm, thick

MOCVD-grown GaN film in which they found that the stronger the yellow emission, the

shorter the decay times for free-excitons and for bound-excitons. These authors

suggested that the fast decay behavior of the PL intensity indicated that the capture of

excitons and trapping of carriers at deep level defects through nonradiative combinations

dominate the decay of exciton population.

On the other hand, Gotz et al. [37] and Hacke et al. [38] used a transient

capacitance method to analyze traps occurring in unintentionally doped n-type GaN. By

virtue of the sensitive measuring technique (deep-level transient spectroscopy [DLTS],)

Gotz et al. detected two deep-levels having thermal activation energies for electron

emission to the conduction band of 0.16 eV and 0.44 eV. However, Hacke et al. found

three majority-carrier traps occurring at different energies below the conduction band

having activation energies, 0.26, 0.58, and 0.66 eV, respectively. The detected

activation energy levels mentioned above belong to group B and group A of Fig. 1-2.

Although DLTS is a sensitive spectroscopic tool for the characterization of electronic

levels in the bandgap of semiconductors, conventional DLTS is of limited use in wide

bandgap materials as the accessible range of level energies in the gap is restricted to about

1 eV. Optical deep-level transient spectroscopy (O-DLTS), also called photoemission

capacitance transient spectroscopy, can overcome this limitation by using monochromatic

light for carrier emission so that levels in the entire bandgap become accessible for

characterization. Using O-DLTS, Gotz et al. detected four new deep levels with








optical threshold energies for electron photoemission of 0.87 eV, 0.97 eV, 1.25 eV, and
1.45 eV which are relatively fitted to group D of Fig. 1.2. Very recently, Sanchez et al.
[39] investigated yellow band and other deep level emissions in undoped MOVPE GaN

samples using photocapacitance and photoinduced current transient spectroscopy

(PICTS). Photocapacitance reveals in all samples two specific signatures at photon

energies of I eV and 2.5 eV below the conduction band and the capacitance step at 1 eV

is only seen after photoionization at an energy above 2.5 eV. In addition, Sanchez et al.

observed the correlation between PICTS amplitude and the PL intensity of the yellow

emission, and they concluded that both transitions are linked to the same trap which is at

2.5 eV below the conduction band. It is assigned to group E as shown in Fig. 1.2.

Lin et al. [40] investigated the intensity of low temperature PL as a function of

annealing temperature in GaN films grown by MBE. These authors found a reduction of

the donor-bound exciton (I2) peak intensity and an enhancement of the yellow-band

emission at high annealing temperatures (900* C). These authors annealed their samples

for 30 minutes in a nitrogen ambient, and attributed their observations to the formation of

intrinsic point defects such as Ga vacancies and antisite defects.


Migration Enhanced Epitaxy With Reference to GaAs

Migration enhanced epitaxy (MEE) is a modified version of conventional MBE

in which the constituent element fluxes are supplied separately rather than simultaneously

to the substrate [28,40,41,42,43].

The basic advantage of MEE over MBE is that rapid migration of adatoms on the

growing surface is ensured even at relatively-low growth temperatures, adatom migration

being very important with regard to the growth of high quality epitaxial layers. During

conventional MBE growth of GaAs and AlAs, for instance, the group III atoms on As-

stabilized surfaces migrate very slowly especially at low growth temperatures due to their








reaction with As atoms. Utilizing reflection high-energy electron diffraction (RHEED)

analysis, Neave et al. [44] determined the activation energy of Ga surface diffusion to be,
ED = 1.3 eV, on an As-stabilized (2x4) surface. For the diffusion of Al during the growth

of AlAs, a higher ED value of 1.6 eV was estimated from the ratio of the cohesive energies

of AlAs and GaAs by Joyce et al. [45]. These large activation energies, apparently

caused by the formation of GaAs and AlAs molecules, control the growth process in

conventional MBE. It is believed that such large activation energies may make the

surface group-HI adatoms almost immobile when the growth temperature is lower than

4000 C, and this is the reason why high-quality GaAs and AlAs films have not been grown

at temperature lower than 4000 C using conventional MBE. Nagata and Tanaka [46]

used a shadowing method to determine the Ga surface diffusion distance on (001) oriented

GaAs under both Ga- and As-stabilized surface conditions. These authors concluded that

at temperatures less than 550 C, the diffusion length of Ga under Ga-stabilized surface

conditions was 1900A, which is about 10 times larger than the diffusion length under As-

stabilized conditions. In conventional MBE, the growth of GaAs and AlAs films is

usually under As over pressure conditions, which means that the surface is As-stabilized.

For this situation, migrating species on the growing surface may move as Ga-As and Al-

As molecules rather than isolated Ga and Al atoms since impinging atoms immediately

react with arsenic to form Ga-As and Al-As molecules. The large activation energy of

surface diffusion as reported in ref. 43 makes these molecules migrate extremely slowly.

On the other hand, when Ga and Al atoms arrive on an As-free growing surface, such as

in the case of MEE growth, migration takes place in atomic rather than molecular form.

Therefore, surface migration is effectively enhanced under Ga- or Al-stabilized

conditions.







Migration Kinetics During MEE with Reference to GaAs


To illustrate how the migration of surface adatoms affects the quality of

epitaxial layers, the kinetic energy of migrant atoms versus temperature is presented in

Fig. 1.3 [29]. As shown in Fig. 1.3, Em represents the minimum energy necessary for

surface adatoms to move to the neighboring site, Ek and E, are the potential energies at

kinks and steps, respectively, and Eo denotes the kinetic energy of the migrant atom.

Because Em >> Eo = kT (where k is the Boltzmann constant and T is the absolute

temperature), the migration of surface adatoms during conventional MBE is usually very

slow. The large value of E. is caused by the fact that the growing surface has a high

density of small islands of GaAs, and E. includes the decomposition energy of GaAs

molecules at the outermost region of the islands. In contrast, excess Ga adatoms are

expected to migrate actively during MEE growth in the absence of As, thus, Em in the

MEE case is not much greater than Eo even at relatively low temperatures. The Ga

atoms diffuse until they find more stable lattice sites at kinks and ledges on the growing

surface where Ga-As bonds are formed. Thus, the energy relationships in the MEE case

can be expressed as Eo, E << Ek, E,. The migration mechanisms occurring during

growth of GaAs by MEE are discussed below.

The first case is the migration of Ga atoms on an As-atom plane. When only Ga

atoms are supplied to the As-plane, each surface Ga atom shares two sp2 bonds with As

atoms in the underlying As-plane which leaves one unsatisfied sp2 bond perpendicular to

the surface [29]. Because of the strong acceptorlike nature of this unsatisfied sp2 bond,

the sp2 bonds between surface Ga and the underlying As atoms could be unstable. Thus,

it can be expected that the surface Ga atoms are very mobile compared to GaAs

molecules.

The second case is the migration of Ga atoms on a Ga-atomic plane, especially

when about two monolayers (ML) of Ga atoms are supplied during a Ga exposure.


.,.'INC .1 -. ....:.








Impinging Ga atoms will form metallic bonds with underlying Ga atoms. In metallic

bonds the electron distribution is wide, which makes the migration of Ga atoms bonded to
Ga atoms very easy. This characteristic holds even below the melting point of metals. Ga

has a particularly low melting temperature; therefore, Ga atoms on a Ga-atomic plane

move freely, even at low temperature, until they reach active sites such as locally exposed

As-islands and other ledges and kinks. This characteristic is important for the MEE

process at low temperatures because it is expected that Ga atoms supplied on either As-

or Ga-atomic planes tend to form a two-dimensional atomic sheet rather than a three-

dimensional cluster.

The third case is the migration of As molecules on a Ga-atomic plane. It is well

known that the Ga-atomic plane surface is very active chemically because of the

unsatisfied sp2 bonds perpendicular to the surface. Thus, the impinging As molecules at

these sites will decompose into As atoms which are incorporated into the epitaxial layer to

form stable sp3 bonds with underlying Ga atoms and neighboring As atoms [29,47].

Therefore, migration of As atoms on a Ga-atomic plane could be slow; on the other hand,

isolated group V atoms on a Ga-atomic plane cannot form sp3 bonds, and in this case they

may migrate easily.

The fourth possible case of diffusion concerns As molecules on an As-atomic

plane. An As surface is less chemically active than a Ga surface because the As atoms of

an As-atomic plane stabilize themselves by forming p' or sp' bonds with each pair of

adjacent As atoms [29,47]. The arriving As molecules are weakly adsorbed on the As-

plane and migrate actively along the surface. As a matter of fact, Foxon and Joyce [48]

reported a surface diffusion activation energy of As4 molecules on As-planes to be 0.25

eV.

However, the growth of more ionic bonded materials such as Group III nitrides

may be different from that of mostly covalent bonded materials such as Si and GaAs,

because of the larger electronegativity difference between the components of the II-V







nitrides [49,50]. Lester el al. [51] suggested that Fermi level pinning at the crystal

surface during epitaxial growth may not occur in the case of n-type GaN. In addition,

almost all as-grown unintentionally-doped GaN films are n-type [52]. Any attempt to

increase the V/II ratio results in a rough growth front. As a matter of fact, unlike GaAs

in which the growth rate is controlled by the impinging Ga flux alone, GaN growth is

controlled by both the Ga flux and the nitrogen flux [53].


MOCVD Growth of GaN Using An Alternate Source Exposure Method.

There have been three reported attempts to grow GaN films by MOCVD using

alternate source exposure. Hwang et al. [54] alternately exposed a sapphire substrate to

trimethylgallium (TMG) and N+H plasmas via a rotating suscepter, but most of the

resulting GaN films were polycrystalline. Sumakeris et al [55] alternately exposed an a -

SiC substrate to triethylgallium (TEG) and ammonia. The film thickness was measured

by ellipsometry to study the effect of the exposure time to the reactant species on the

amount of material deposited per cycle. A self-limited growth was found to occur at a

substrate temperature of 1200 C and was limited to about 67 % of a monolayer. The

films grown at this temperature were amorphous. The crystallinity improved with

increasing temperature but the chemically self-limiting growth characteristic disappeared.

Khan et al. [56] used switched MOCVD to supply TEG and NH3 to a sapphire substrate

at 4500 900 C at a pressure of 76 Torr. The growth rate was about 0.75 monolayer

per cycle with a 10% margin of error. The band-edge PL signal intensity and linewidth

for the alternate source exposure-grown sample grown at 900 C were comparable to

those observed from samples grown by conventional MOCVD at 10400 C.










:'" :." .,. .- .i!







Dissertation Outline

This dissertation is subdivided into six chapters. Chapter 1 includes the

motivation and objectives for this work and a review of the deep-level defects which result

in yellow band luminescence in GaN.

Chapter 2 details the growth techniques employed in this work and the in-situ

analytical techniques used to characterize the growth are also detailed.

Chapter 3 describes the substrate preparation and the conventional MBE growth

procedures used to synthesize GaN films on c-plane sapphire substrates, and reports PL,

Hall-effect and SEM analyses of these films.

Chapter 4 presents the novel AEEE growth method used in this investigation and

growth parameters are correlated with the observed RHEED specular reflection beam

intensity variations. In this chapter, PL and SEM data are reported as functions of the Ga

and N delay times.

Chapter 5 presents a discussion of the GaN growth kinetics and their correlation

with deep level defects. A model is presented to explain the experimental observations.

Finally, Chapter 6 summarizes the principal conclusions drawn from this

investigation and presents suggestions for future work in this area.

























5

(o
0


0
-U
'U
U?


u 1 2 3
E (eV)









Figure 1.1 The predicted formation energies of native deep-level defects
in GaN as a function of the position of the Fermi level, EF
(after Ref. 20).





.9' .











conduction
band-edge


0



0.5-



1.0_



S1.5-



S2.0.




2.5.



3.0_



3.4


Figure 1.2 Native defect levels in as-grown GaN summarized from the
literature. AB,C (VN), D (NG VG.), and E (GaN ,VJ.


A


valence
band-edge


I I




















Em


Es Ek


Figure 1-3 Potential energy diagram for surface migration.
Em represents the minimum energy necessary for
surface adatoms to move to the steps and kinks, Ek
and E, are the potential energies at kinks and steps,
respectively. Eo is the kinetic energy of migrant
atom.












CHAPTER 2
MBE SYSTEM AND CHARACTERIZATION TECHNIQUES


Introduction

All of the samples in this work were grown in a Varian GEN II MBE system.

This system was previously employed to grow zinc blende-GaN films [59,60]. All

samples were characterized in-situ using RHEED analysis as well as ex-vacuo using other

analytical techniques including scanning electron microscopy (SEM), Hall-effect

measurements and photoluminescence (PL) measurements. All samples were grown and

characterized by the author with the exception of the PL analysis which was carried out by

Dr. Matthias Ludwig in Dr. Hummel's research laboratory at the University of Florida.

The MBE growth conditions employed, as well as important considerations and

procedures pertaining to the growth, and the characterization techniques mentioned above

are outlined in the following chapter.


Molecular Beam Epitaxy

Molecular beam epitaxy has become a powerful growth technique particularly

suitable for manufacturing advanced electronic and photonic devices. In a MBE system,

constituent elements and dopants are supplied in the form of molecular or atomic beams,

either thermally generated or formed in a plasma. The beams are directed toward a

suitably heated single-crystal substrate on which epitaxial growth occurs. Since the

molecular or atomic beams can be individually shuttered, complex superlattice structures

can be fabricated using this technique with precise control down to the atomic-level.







Varian GEN II MBE System Configuration

The Varian GEN II MBE system at the University of Florida consists of three

connected ultra-high vacuum (UHV) chambers; an entry/exit (EE) chamber, a buffer

chamber, and the growth chamber. Each UHV chamber is of stainless steel

construction and has an independent pumping system. The chambers are isolated by gate
valves and all system components are bakeable to 2000 C.

The entry/exit chamber allows samples to be loaded into and out of the system

while maintaining the buffer and growth chambers under ultrahigh vacuum conditions.

To reduce contamination, the EE chamber opens into a clean room where the ex-vacuo

cleaning of substrates take places while the opened chamber is vented using dry nitrogen

gas. Two molecular-sieve and liquid nitrogen cooled sorption pumps are used to

evacuate the chamber, in addition to a CTI-100 cryogenic pump which produces ultra-

high vacuum conditions. After entry into the EE chamber, the substrates are heated to

150 C using a pair of quartz halogen lamps in order to desorb water vapor from the

substrate surface and the sample holder.

A transfer trolley facilitates sample movement between chambers. A magnetic

force is employed to transfer the trolley from the EE chamber to the buffer chamber once

the valve separating the two chambers is opened. The buffer chamber is evacuated by a

Vaclon pump. A heater station inside the buffer chamber is employed to evaporate any

impurities whose vapor pressures are too low at 1500 C (the EE bake-out temperature).

The temperature used for buffer bake- out is 4500 C. There are two magnetically-

coupled transfer arms, one for transferring a substrate to the buffer chamber heater

station, the other for the substrate mounting block, which is attached to the continuous

azimuth rotation (CAR) substrate holder, inside the growth chamber. The pressure

difference between the buffer and the growth chamber is minimized before the gate valve

between the chambers is opened.







The growth chamber is illustrated schematically in Fig. 2. 1. Different types of

pumps, a CTI-200 cryogenic pump, a Vaclon pump, and a titanium sublimation pump, are
employed to maintain the UHV environment required for growth. When attached to the

CAR, the substrate can be rotated (1800) from the transfer position to the growth

position and vice-versa, and rotated (3600) in its own plane for RHEED analysis. Heater
filaments for elevating the substrate temperature are mounted on a pyrolitic boron nitride

(PBN) plate inside the substrate mounting block. A W-Re thermocouple is isolated from

the heater filaments by a PBN tube in a radiation cavity which is located in the center of

the substrate mounting block. The water-cooled substrate mounting block can heat the

substrate to 8000 C by radiation through the PBN plate which gives uniform heat

distribution for the entire substrate holder, while the thermocouple serves to monitor and

control the substrate temperature. In addition, the rotation speed of the CAR substrate

holder can be controlled which further improves the substrate temperature uniformity

during epitaxial growth.

Conventional effusion cells are used to supply the atomic and molecular beams,

with the exception of the nitrogen source. There are eight furnace ports to house

different source materials. For the conventional effusion cells, each solid or liquid source

material is placed into an ultraclean PBN crucible. The liquid sources are mounted on

upward-facing ports, while the solid sources are mounted on downward-facing ports. A

W-Re thermocouple is held by spring force tightly against the bottom of each crucible.

To maintain the high purity of the source materials, the crucibles are idled at relatively

high temperature in the range, 200 C to 820 C, depending on the source. Cryoshrouds

surround the effusion cells and the growth chamber itself to dissipate excess heat. A

beam flux monitor (nude ion gauge) is utilized to measure the beam flux from the effusion

cells. The beam flux monitor is mounted on the CAR 180 away from the substrate

position in order that beam fluxes can be monitored without exposing the substrate to the

fluxes. Control over the beam fluxes is achieved using pneumatically controlled shutters








in front of each source orifice. For practical purposes, the beams can be considered as

unidirectional so that the insertion of a mechanical shutter will stop a beam from reaching

the substrate and allow different crystal compositions to be grown. The precise control

of the opening and closing of the shutters plays an important role in this research.


Substrate Temperature Calibration

The actual temperature at the substrate surface is different from that monitored by

the W-Re thermocouple, since the thermocouple is not physically attached to the

substrate. In this work, Omega Temperature Indicating Liquids (OTIL) which have a

time response to temperature changes on the order of milliseconds and an accuracy of+

1 % were used to calibrate the surface temperature. By applying a small amount of OTIL

on a test sapphire surface which was mounted on a Mo block, it was easy to pinpoint at

which monitored temperature the OTIL turned black, an irreversible change, indicating

that the substrate had reached a particular temperature. In this fashion, a calibration was

obtained and a plot of actual sapphire substrate surface temperatures versus monitored

temperatures is presented in Fig. 2.2.


In Situ Analysis


Reflection High Energy Electron Diffraction (RHEED) was utilized to perform

real-time in-situ analysis of the substrate and epilayer surfaces. The RHEED electron

beam energy was in the range of 0 10 keV and the electron beam was incident at a

glancing angle of 1-2 to the substrate surface. Since monochromatic electron beams are

sensitive to atomic order in solids and are ideally suited for the study of crystalline

surfaces, this in-situ surface analyzer was invaluable in providing information about the

surface structure, orientation, and the degree of surface roughness during epitaxial
(,' **.!!,.'f







growth. RHEED is particularly useful in studying structure changes as functions of

temperature or time during epitaxy.


Ex-vacuo RHEED Specular Reflection Beam Intensity Monitoring System

In addition to observing diffraction patterns, the RHEED specular reflection beam

intensity was monitored in this work and the monitoring system employed during growth

consisted of a CCD camera, video monitor, Si-photodiode, bias circuit, and chart-

recorder as shown schematically in Fig. 2.3. The RHEED specular beam intensity was

captured by the CCD camera and displayed on the monitor and the Si-photodiode placed

over the spot on the monitor. The bias circuit placed a potential across the photodiode

causing output (voltage) to be linearly related to the input (light intensity). The chart

recorder logged the output voltage of the photodiode.

By monitoring the variation of the RHEED specular reflection beam intensity

during growth, the migration process and surface conditions on the growing surface could

be evaluated. Because the specular reflection beam intensity is sensitive to the degree of

surface roughness, a decrease of the RHEED specular reflection beam intensity indicates

an increased surface roughness, such as might be associated with island formation for

instance. To illustrate this point a typical RHEED specular reflection beam intensity trace

recorded in this work during alternate element (Ga and N) exposure growth is presented

in Fig. 2.4. As can be seen from the figure, the specular reflection beam intensity is a

strong function of the status of the surface in terms of species present and the particular

kinetic processes taking place. The interpretation of such data will be discussed at length

in the following chapters.







Nitrogen Free-Radical Source

In this work, an Oxford Applied Research (OAR) radio frequency (rf) plasma

discharge source as shown schematically in Fig 2.5 was used to generate nitrogen free-

radicals. The same type of source has been successfully used to supply atomic nitrogen

for p-type doping of ZnSe [57,58] and to grow zinc blende-GaN [59,60]. The source

operates at 13.56 MHz and has a maximum output power of 500 W. The radical source

is mounted on the MBE growth chamber, replacing a conventional effusion cell. A

distance of 15 cm separates the end of the radical source from the substrate.

High-purity molecular nitrogen gas (99.9995%) was introduced through a

precision UHV leak valve into a PBN discharge tube capped with a PBN aperture. The

PBN aperture had 37, 0.3 mm diameter holes to release reactive nitrogen species to the

substrate surface. The gas line between the nitrogen cylinder and the UHV leak valve

was pumped by two liquid nitrogen cooled sorption pumps before opening the valve on

the cylinder to ensure the purity of gas that was introduced into the growth chamber.

In this work a nitrogen plasma was maintained using a forward power of 200W

and a nitrogen chamber pressure of 1 x 10 -5 Torr. Under these conditions, an optical

emission spectrum was recorded from the plasma using a wavelength spectrometer/OMA

apparatus. The emission spectrum shown in Fig. 2.6 is indicative of the presence of both

ground state nitrogen atoms (1' Positive System of N2 transitions) and excited nitrogen

atoms (direct transitions) in the plasma and therefore in the flux emanating from the

source.











High pure N2

\


rf plasma discharge
chamber


Electron
Gun
(RHEED)


Ion
Gauge


Liquid Nitrogen Phosphor
Cooled Shrouds Screen
Residual gas
Analyzer



Figure 2.1 Schematic of the MBE growth chamber used to grow
GaN/sapphire films by conventional growth and by
AEEE techniques.




SLU


I I I 1 1 I
750 800 850 900 950 1000


Monitored temperature (o C)


Figure 2.2


Plot of sapphire substrate surface temperature vs. monitored
temperature obtained using Omega Temperature Indicating
Liquids (OTIL).


750


700


650


600


550


500 -


450 -


400 -


350


700


1050


1


-


S .. .. ... .... ....













Typical RHEED pattern
on the monitor


Bias





Photodiode



Video monitor


Mask


The bright spot of
the RHEED specular
reflection beam
(intensity of which
was monitored as a
function of time).


CCD camera


AIR


RHEED
screen


Typical RHEED pattern


MBE growth
chamber


UHV


Figure 2.3. Experimental set-up used to monitor the RHEED specular
reflection beam intensity as a function of time during
growth.











N delay time
I
i ------

IN
Ig
N shutter open I


Ga delay time


Ga shutter closed


Time (sec.) 1


Figure 2.4 A typical specular reflection beam intensity trace recorded during
one cycle of alternate element (Ga + N) exposure growth.


N shutter closed










PBN Discharge Tube


Water-Cooled
RF Coil


Gas Inlet T

Window Water Feed Gas Feed
RF Shield
PBN Beam Exi
aperture disk
Conflat Flange


Figure 2.5 Schematic of rf plasma discharge source used to generate nitrogen free radicals (N).


13.56 MHz
500 Watts
RF Generator


Optical
Detector


N























2 >1














I I I I I I
500 550 600 650 700 750 800 850

Wavelength (nm)




Figure 2.6 Optical emission spectrum of nitrogen plasma generated
from a 190 W, 13.56 MHz, rf plasma discharge with a nitrogen
background pressure of 2 x 10-' Torr.













CHAPTER 3
GROWTH BY CONVENTIONAL MBE AND
CHARACTERIZATION OF GaN / SAPPHIRE FILMS


Ex-vacuo Substrate Preparation


All of the c-plane sapphire (A1203) substrates used in this work were obtained from

Union Carbide Crystal Products in Washougal, WA. The specifications of these substrates

are listed in Table 3.1.




Table 3.1 Specifications of Union Carbide c-plane sapphire substrates


Wafer Size 10 xll mm

Wafer Orientation
Surface normal to the (0001) or c-axis to within 3

Edges longer edges normal to ( 1 2 1 0 ) or a-axis

Wafer Thickness 500 microns

Surface Finish one side polished

X-ray Rocking Curve linewidth -12 arc seconds

Lattice Parameter
a-axis a=4.758 A
c-axis c=12.991A

Thermal Expansion Coefficients
a-axis 7.5 x 107/K
c-axis 8.5 x 10 /K


P. I'









Ex-vacuo substrate preparation procedures were performed in the clean room.

The A1203 (0001) substrates were degreased by successive rinses in trichloroethane,

acetone and methanol and then etched in H2S04 : H3P04 (3:1) at 160 C for 15 minutes.

The etching was used to eliminate any surface contaminants and mechanical damage due

to polishing. Atomic force microscopy (AFM) analysis showed that the surface

morphology of the etched sapphire substrate was smoother than that of the un-etched

substrate as shown in Fig. 3.1 (rms roughness = 0.398 nm compared to as-received

sapphire, rms roughness = 2.458 nm). The wafers were then rinsed in de-ionized water

and blown dry using dry nitrogen. The backside of the substrates were metallized with

Mo using electron-beam (E-beam) evaporation. In the vacuum (< 5 xl0"7 Torr)

environment of the E-beam evaporation system, a high-intensity beam of electrons was

focused on a high-purity Mo source target. With the shutter closed, the source target

was outgassed for 5 minutes. The shutter was then opened and Mo was deposited at a

rate of-3 A/sec. as measured by a quartz crystal monitor. The resonant frequency of the

crystal shifted in proportion to the thickness of the deposited film. By monitoring the

shift in resonant frequency of the crystal, the deposition rate could be measured with an

accuracy of better than 1 A/sec. The total thickness of the Mo metallization was 0.25

gm. The metallization on the backside of substrate greatly improved the thermal

uniformity across the substrate during growth. Just before loading into the EE chamber,

the metallized substrates were re-rinsed in methanol, de-ionized water, and strain-free

mounted on a hollow molybdenum block which was loaded on a trolley. A stream of dry

nitrogen was used to remove particles from the substrate surfaces and the trolley. The

trolley was then introduced into the EE chamber which was then pumped down.







In-situ Substrate Preparation

All substrates were preheated at 150 C for one hour in the EE chamber and at

450 C for 20 minutes in the buffer chamber to desorb impurities such as water vapor on

the substrates and substrate holders. Before carrying out the preheating procedures in
both the EE and the buffer chambers, it was important to make sure that the background

pressure in both chambers was lower than 10'8 Torr. After transfer to the growth

chamber, the substrate was heated slowly (-30 C/min. ) to 750 C; this temperature was

held for 60 minutes for thermal cleaning. All substrate temperatures quoted in this

dissertation are the actual sapphire surface temperatures calibrated using the OTIL method

described in the previous chapter. The RHEED system was used to monitor the surface

morphology and crystallinity through the thermal cleaning and nitridation processes. The

thermal cleaning resulted in a (1 x 1) RHEED pattern with Kikuchi lines indicating a clean

A1203 (0001) surface. A 9 keV electron beam was oriented along (a) [1 TO 0 ] ((=0)

and (b) [1 2 1 0] (4=30) as shown in Fig. 3.2 (a,b). Then, the thermally cleaned

substrate was exposed to the nitrogen free-radical beam (at a forward power of 250 W)

while at the same temperature (7500 C) for 20 minutes to form a thin AIN layer whose

RHEED pattern is shown in Fig. 3.3.


Conventional MBE Growth of GaN on Sapphire

With both Ga and N shutters closed, the substrate temperature was lowered to

500" C at which temperature a thin GaN buffer layer was deposited. A Ga effusion cell

temperature of 7800 C was used to provide a low Ga flux (J= 7.6 x 1013 cm -2 s' ) in

order to grow the buffer layer. Growth was initiated by simultaneously opening both Ga

and N shutters for 10 minutes to grow a -300 A thick GaN buffer layer. An elongated

spotty RHEED pattern was observed during the growth of this layer. The basal plane of

the GaN overlayer was rotated 300 with respect to that of the substrate during growth; the








orientation relationship between the single crystal GaN film and the substrate was [0001]

GaN // [0001] A1203, [1T00] GaN //[1210] A1203 and [21-0] GaN //[ 100] A1203.

Immediately following buffer-layer deposition, the substrate temperature was ramped to

750 C, 700 "C, 640 C or 600 C in 5 minutes, during which time both source shutters

were kept closed. A streaky RHEED pattern developed during the ramp in each case.

Also during the ramp, the Ga cell temperature was increased to 820" C to provide a flux

ofJ = 2.3 x 1014 cm 2 s1. When the appropriate substrate temperature was reached,

both the Ga and N shutters were opened simultaneously, producing a growth rate of 0.1-

0.2 pm/hr depending on the growth temperature. The growth parameters employed are

summarized in Table 3-2. The surface morphology during growth was very sensitive to

the Ga/N beam flux ratio as well as the growth temperature. For a fixed N flux (1 x 10"

Torr), the RHEED pattern changed from streaky to spotty at 750 C when the Ga beam

equivalent pressure (BEP) was lower than 5 x 108 Torr ( J=l.8 x 1014 cm-'s"). However,

the same Ga flux resulted in a milky surface at 600 C. Hence, a lower Ga BEP of 2 x

10-' Torr (J=7.6 x 1013 cmf2s1) was used for conventional growth at 6000 C, and 6 x 108

Torr (J=2.3 x 1014 cm2s"') was used for growth temperatures of 640" C 700" C and 750"

C. A streaky (lx1) RHEED pattern, as shown in Fig. 3.4, was maintained throughout

growth for all conventionally-grown samples. In order to provide a uniform thermal

distribution, the CAR was rotated at a speed of 1 rpm for all conventional growths. The

total growth time in each case was 4 hours and about 0.8 gm of GaN was deposited. For

the sample grown at 6000 C, a 0.4 pm thick film was deposited since Ga droplets

prevented it from growing thicker, as suggested in references 28 and 29. In each case,

growth was terminated by simultaneously closing both shutters and cooling the sample to

room temperature in 20 minutes in a nitrogen environment. The thermal expansion

coefficient of sapphire is about 26-34% [ 34% in the case of c I (GaN) // c // (sapphire)

and 26 % in the case of c I (GaN) // c (sapphire)] larger than that of GaN. Therefore,


'








the epitaxial GaN film suffered compressive stress in the plane parallel to the c-face when

the sample was cooled, hence a slow cooling rate was necessary.


Table 3.2 Growth parameters used for both conventionally-grown and AEEE-grown
films.

Nitrogen source parameters

RF forward power 200 W
Nitrogen beam flux 1 x 10-5 Torr
Plasma emission intensity 80 100 mV

Ga source condition

Ga beam flux for epilayer growth 6.0 x 10f" Torr (J=2.3 x 1014 cm-21)
Ga beam flux for buffer layer growth 2 x 10-' Torr (J=7.6 x 1013 cm-2S)

Growth parameters

Buffer layer
Growth temperature 500 C
Buffer layer growth time 10 min.

Conventionally-grown epilayer
Growth temperature 6000 C, 640 "C, 700 "C, 7500 C.
Growth rate 0.1 0.2 pm/hr
Film thickness 0.4 and 0.8 pm


Characterization of Conventionally-Grown GaN Films


SEM Analysis


The surface morphology of MBE grown samples was examined by SEM. The

surface morphology of the sample grown at 7500 C showed a typical micrometer range

rough surface as shown in Fig. 3.5. The surface morphology of the sample grown at








6000 C was smoother in most areas, however, -15 pm diameter Ga droplets were

distributed over the surface with a density around, 6 x 104 cm"2, as shown in Fig. 3.6.

The sample grown at 6400 C had a similar surface morphology, however, the Ga droplet

size was smaller (< 5 pm ) in this case. In order to observe what was under the Ga

droplets, samples were etched in (1:10:100) HF : NHO3: H20 before re-performing SEM.

It was found that crack-like voids were observable beneath the Ga droplets as shown in

Fig. 3.7.


Hall-Effect Measurement and Photoluminescence Analysis


The Van der Pauw method was employed to measure carrier concentrations and

mobilities at room temperature using Hall-effect analysis. Small pieces of indium were

pressed in place at the four corners of square-shaped samples to make ohmic contact (the

I-V characteristic was linear) to the GaN. All of the unintentionally doped GaN films

showed 1018 cm"3 n-type carrier concentrations except the sample grown at 6000 C which

had a 3 x 10'9 cm-3 carrier concentration. The carrier mobility was 35 cm2 V1 s" for the

sample grown at 6400 C and was in the single digit range for those samples grown at 6000

C and 7000 C.

In the PL analysis, the samples were excited by the 325 nm wavelength output of a

25 mW HeCd laser. The power density at the sample was about 30 mWmm-2. The

luminescence signal was dispersed by a 0.3 m single-grating monochromator and detected

by a photomultiplier tube. Room-temperature and low-temperature PL spectra were

recorded from the samples grown at 700, 640, and 600 C and these spectra are

presented in Fig. 3.8-3.13. As can be seen from these figures, the yellow band emission

was observed in all of the spectra in addition to the near-band-edge emission.




37



A detailed discussion on luminescence peak assignments (both near-band-edge and

deep level emission) is presented in Chapter 5 based on the results detailed in this chapter

and also Chapter 4.









(a) rms roughness = 2.458 nm


SI




B '' \ ./
"' ..... \ /


rms roughness = 0.398 nm


I
I
r
~~J
Ucl


Figure 3.1


Atomic force microscopy tapping mode images showing the
morphology of c-plane sapphire
(a) as-received, rms roughness = 2.458 nm
(b) etched in H, SO, : HPO, at 1600 C for
15 minutes, rms roughness = 0.398 nm.





























(b) (a)




Figure 3.2 RHEED patterns recorded from thermally cleaned
sapphire (0001) surface, (a) electron beam oriented
along [1TOO] (0 = 0 ), (b) [1210] (0 = 30 0)





























(b) (a)







Figure 3.3 RHEED pattern recorded from a thin AIN layer formed
during nitridation at 7500 C after 20 minutes exposure
to the N-flux (a) [2110] AIN // [1100] sapphire,
(b) [1100] //[1210] sapphire.



















































Figure 3.4


RHEED patterns recorded from a conventionally
grown GaN film, (a) [2110] GaN// 2110]AIN//
[1100] sapphire, (b) [1100] GaN // [1100] AIN//
[1210] sapphire.


................. ......... ...~m~r.~..
............................................nr









































Figure 3.5 SEM micrograph showing the surface morphology of a
GaN film grown by conventional MBE at 750 C.





















(a)


(b)


Figure 3.6 SEM micrographs of the surfaces of GaN films grown by MBE
at a substrate temperature of 600 C (a), and 6400 C (b).
Note the presence of Ga droplets.











































Figure 3.7 SEM micrograph showing the GaN morphology beneath a Ga droplet
after removal of the Ga droplet by wet etching. The GaN was grown at
6000 C.









5000

4500

4000

3500

3000

2500

2000

1500

1000

500

0-


3:


50 400 450 500 550 600 650
Wavelength (nm)

Figure 3.8 Low-emperature PL spectrum recorded from a GaN/sapphire
film grown at 700 C by conventional MBE.


700














4000


3500


3000


2500


2000


1500


1000


500


0


350


400 450 500 550 600 650


700


Wavelength (nm)



Figure 3.9 Room-temperature PL spectrum recorded from a GaN/sapphire
film grown at 7000 C by conventional MBE.













250000 -




200000




2 150000




S 100000




50000




0
350


400 450 500 550 600 650 700
Wavelength (nm)


Figure 3.10 Low-temperature PL spectrum recorded from a GaN/sapphire
film grown at 6400 C by conventional MBE









250000 I I I .. .




200000




150000


364 nm

100000




50000




0
350 400 450 500 550 600 650 700
Wavelength (nm)





Figure 3.11 Room-temperature PL spectrum recorded from a GaN/ sapphire
film grown at 6400 C by conventional MBE.












250000 -




200000




Z 150000
a













0
35100000
50000

-3



350


400 450 500 550 600 650


Wavelength (nm)


Figure 3.12


Low-temperature PL spectrum recorded from a
GaN/sapphire film grown by conventional MBE
at 6000 C.


700












7000

6300

5600

4900

4200

3500

2800

2100

1400

700

50
350


400 450 500 550 600 650


Wavelength (nm)



Figure 3.13 Room-temperature PL spectrum recorded from a GaN/sapphire
film grown by conventional MBE at 6000 C.


700












CHAPTER 4
INVESTIGATION OF A NOVEL GROWTH MODE


Introduction

The PL spectra recorded from all of the conventionally grown GaN/sapphire films

showed near band-edge emission at around 365 nm and a yellow band emission centered

around 550 nm. Decreasing the growth temperature improved the efficiency of the near

band-edge emission but the deep-level defect-related yellow band emission was still

present. Furthermore, decreasing the growth temperature resulted in the formation of Ga

droplets on the growing surface and the yellow band emission became dominant in the PL

spectra. This yellow band emission has been commonly reported in the literature

regardless of the growth techniques and the substrates used [2-14]. With a view to

investigating the factors that lead to deep level generation and consequently to the yellow

band emission in the PL spectra recorded from GaN epitaxial films, a systematic study was

conducted in which the surface morphology was examined by means of the real-time

monitoring of the RHEED specular reflection beam intensity during critical phases of

growth. In particular, a new growth mode was studied in which the substrate was

separately exposure to Ga atoms and N atoms (as in migration enhanced epitaxy [MEE])

and, in addition, time delays were introduced following the Ga and N exposures which we

term the Ga delay and the N delay, respectively. The time delays, in which a variety of

kinetic processes might take place depending upon the element in question, distinguish this

growth mode from MEE and we have termed the novel growth mode "alternate element

exposure epitaxy (AEEE)". This work was carried out as follows; first, a systematic

study was conducted to determine the optimum AEEE growth temperature, second, a








systematic search was made for an optimum N delay time at the optimum growth

temperature, and, finally, a systematic investigation was conducted to find the optimum

Ga delay time, also at the optimum growth temperature.


AEEE Growth Procedures


The ex-vacuo and in-situ substrate preparations were the same as those used for

the preparation of conventionally-grown samples as mentioned in the previous chapter.

After nitridation of the sapphire substrate at 750 C, the substrate temperature was

lowered to 500 0 C, both Ga and N shutters being closed. A 300 A thick buffer layer was

deposited at this temperature by opening the Ga and N shutters simultaneously for 10

minutes. During growth of the buffer, the Ga effusion cell temperature was kept at 7800

C, corresponding to a low Ga flux ofJ = 7.6 x 10 13 cm"2 s-i. After the buffer layer was

deposited, the substrate temperature was increased to 600 OC with both shutters being

closed. During the same substrate temperature ramping period, the Ga effusion cell

temperature was gradually increased to 820 C, corresponding to a Ga flux of J = 2.3 x

10 14cm-2 1.


AEEE Growth Temperature Optimization

In order to determine the optimum substrate temperature for AEEE growth, the same

growth temperatures used in conventional growth were investigated here. Each growth

cycle included opening the N shutter for 10 sec., simultaneously closing the N and opening

the Ga shutter (Ga exposure time, 5 8 sec., depending on the growth temperature),

closing the Ga shutter and allowing the RHEED specular reflection beam intensity to

recovery to a maximum before opening the N shutter again (Ga delay time, 15 30 sec.,

depending on growth temperature). The delay time is defined as being the time elapsed

between closing one shutter and opening the other shutter as indicated in Fig. 2.4 (page








28). The growth conditions for one cycle of the shutter sequence are summarized in

Table 4.1. The recovery of the RHEED specular reflection beam intensity during the Ga

delay was found to be a function of the growth temperature. At higher temperatures the

beam intensity recovered faster but to a lower level of intensity as shown in Fig. 4.1.

Films were not deposited at 7500 C by AEEE because the RHEED pattern did not remain

streaky and the specular reflection beam intensity decreased quickly after repeated cycles.

The specular reflection beam intensity was also found to be a function of the Ga exposure

time. For instance, an 8 sec. Ga exposure time was optimum to have the RHEED

intensity recover to a maximum when the substrate temperature was 700 C, whereas, 7

sec and 5 sec were the optimum Ga exposure times for substrate temperatures of 6400 C

and 600 C, respectively. Failure to control the optimum Ga exposure time results in a

loss of RHEED signal intensity following several cycles. Each of the AEEE samples was

grown using a total of 2000 cycles. The growth was terminated by the last N exposure

and the substrate was cooled to room temperature in 20 minutes under a N2 overpressure

in the growth chamber with both Ga and N shutters closed.


Table 4.1 AEEE growth conditions for one cycle of the shutter sequence at different
growth temperatures.

Growth N shutter N delay time Ga shutter Ga delay Remark
temperature open (sec.) (sec.) open (sec.) time (sec.)
(sample #)
700 C 10 0 8 15 Maximum
(7147) RHEED
intensity-low
640 C 10 0 7 20 Maximum
(7155) RHEED
intensity-
moderate
600 C 10 0 5 30 Maximum
(7165) RHEED
intensity-
high










PL Analysis

Low-temperature PL spectra were recorded from the samples grown at 7000 C,

6400 C, and 6000 C and the spectra are presented in Figs. 4.2, 4.3, and 4.4, respectively.

These spectra all indicate a dominant near band-edge emission, but a low intensity yellow

band emission is still exhibited in samples grown at 700 C and 6400 C. In addition, a

weak peak related to N vacancies [20,37] appeared at 365.5 nm for the sample grown at

700 C and at 385 nm for the sample grown at 6400 C, respectively. On the other hand,

as can be seen from Fig. 4.4, the yellow band emission was absent from the PL spectrum

recorded from the sample grown at 6000 C.


Nitrogen Delay Time Optimization.


The basic advantages of the AEEE technique in the context of GaN are the low

growth temperature and the provision of a chemically active Ga-covered surface. Both

of these factors can greatly enhance the sticking coefficient of N during GaN epitaxy. In

this investigation, the shutter operating sequence was the same as for the sample grown at

6000 C in the previous section except that N delay times, namely, 0 sec., 15 sec., and 30

sec., were inserted between closing the N shutter and opening the Ga shutter as shown

from Fig. 4.5. As can be seen from Fig. 4.5, the RHEED specular reflection beam

intensity recovered only moderately during the N delay phase.

Shutter sequences employed for the N delay experiments are shown in Table 4.2.

It is worth pointing out that during the N shutter open time, the RHEED specular

reflection beam intensity decreased initially and then recovered moderately to a turning

point beyond which the intensity again dropped (see Fig 4.5) and the RHEED pattern

became spotty. The turning point occurred after an exposure time of 10 seconds and thus

a 10 sec. N exposure was employed in this work.


. "f".








Table 4.2 Shutter sequences for the N-delay optimization experiments.
(Growth temperature was 6000 C)

Run N shutter N delay time Ga shutter Ga delay Remark
Number open (sec.) (sec.) open (sec.) time (sec.)

7160 10 0 5 30 Maximum
RHEED
intensity-

7167 10 15 5 30 Maximum
RHEED
intensity-
moderate

7156 10 30 5 30 Maximum
RHEED
intensity-
moderate


PL Analysis

The low-temperature PL spectra recorded from samples grown using N delays of 0

sec, 15 sec. and 30 sec. are shown in Figs. 4.6, 4.7, and 4.8. As can be seen from these

spectra, as the N delay time is increased, the near band-edge emission intensity decreases

and additional peaks appear in the spectra. Consequently, in future growth runs a 0 sec.

N delay was employed.


Ga-delay Time Optimization

In the Ga-delay optimization set of experiments, the N-delay was fixed at 0 sec. based on

our previous experience and the previously determined optimum substrate temperature of

6000 C was used throughout. As indicated in Fig. 4.9, various Ga-delay times were

employed including 0 sec, 15 sec, 30 sec. and 50 sec. As seen in the figure, these delay








times corresponded to various RHEED specular reflection beam intensities. Shutter

sequences employed in the Ga-delay experiments are summarized in Table 4.3.


Table 4.3 Shutter sequences for the Ga-delay optimization experiments.
(Growth temperature was 6000 C)

Run N shutter N delay time Ga shutter Ga delay Remark
Number open (sec.) (sec.) open (sec.) time (sec.)
7161 10 0 5 0

7164 10 0 5 15

7160 10 0 5 30 Highest
maximum
RHEED
S__intensity
7168 10 0 5 50


SEM and PL Analyses

SEM analysis was performed on samples grown with Ga delay times of 0 sec., 15

sec, 30 sec, and 50 sec. No Ga droplets were found on the surfaces of these samples as

can be seen from Figs. 4.10 (a)-(d).

Low- and room-temperature PL spectra recorded from the 0 sec, 15 sec, 30 sec,

50 sec. Ga delay time samples are presented in Figs. 4.11- 4.18. The influence of the Ga

delay time on the PL properties will be discussed in detail in Chapter 5.


Activation Energy of Ga Migration Under N-free Conditions


As illustrated in Fig. 4.19, the RHEED specular reflection beam intensity recovers

upon closing the Ga shutter (during the Ga-delay phase) in two distinct stages. In stage

1, the recovery is fast and the RHEED intensity varies almost linearly with time. However,

beyond stage 1, the recovery appears to be more sluggish in stage 2 in which the







relationship is distinctly non-linear. By assuming monatomic steps with terraces and edges

on the growing surface [62-66], it is believed that the fast stage (stage 1) of the RHEED

specular reflection beam intensity recovery is dominated by Ga adatom migration to step

sites [44,45, 67-71]. With a view to determining the activation energy for the process,

the time to complete stage I recovery was measured as a function of growth temperature

and the migration times recorded in the range, 5600 C to 640 C, are presented in Table

4.4. The data displays an Arrhenius behavior as shown in Fig 4.20. This is the same

approach as that used by Foxon and Joyce [48] who determined the activation energy of

As4 migration on a GaAs surface in the absence of a Ga flux. The activation energy of
Ga migration under N-free conditions was determined from Fig. 4.20 to be 1.46 0.25 eV



Impact of Impinging RHEED Beam on Ga Migration

It is essential to understand the role, if any, that the impinging RHEED electron

beam has on the kinetic behavior of Ga adatoms. In order to investigate this possibility,

the electron gun shutter was opened continuously in the first instance during an operating

cycle and the RHEED specular reflection beam intensity was recorded as shown in Fig.

4.21 (solid line). Immediately following this cycle, the chart recorder was rewound back

to the start position of the previous cycle and this time the RHEED intensity was sampled

by only opening the RHEED gun shutter for 1 sec. every 5 secs. In this case, the trace

recorded is shown in Fig 4.21 as a dashed line. As can be seen from the figure, the

sampled intensities correspond to those recorded during continuous electron beam

exposure and, therefore, it is assumed that Ga migration, for instance, was not electron


beam assisted.







Table 4.4 Ga migration times estimated from the RHEED intensity curves at various
substrate temperatures.

Migration time (sec.) Substrate temperature ( C)

35 -40 560

25 -30 580

13 -17 600

10-13 620

6 -9 640
























7000 C


15 20 30


Ga delay time (sec.)


Figure 4.1


RHEED specular reflection beam intensity as a function of Ga delay
time (time allowed to elapse between closing the Ga and opening the N
shutter) recorded at different substrate temperatures.









50000

T = 27.2K
45000

40000 356.5nm

35000

30000

S25000

A 20000

15000

10000 365.5 nm

5000


350 400 450 500 550 600 650

Wavelength (nm)





Figure 4.2 Low-temperature PL spectrum recorded from a GaN/sapphire
film grown by the AEEE technique at 7000 C.


wV








400000 .

360000

320000 358nm

280000

240000

200000

160000

120000

80000

40000 385 nm


350 400 450 500 550 600 650

Wavelength (nm)





Figure 4.3 Low-temperature PL spectrum recorded from a GaN /
sapphire film grown by the AEEE technique at 6400 C.


700








250000 I II |I I I I .

358nm 24.5K


200000




150000




100000




50000





350 400 450 500 550 600 650 700
Wavelength (nm)



Figure 4.4 Low-temperature PL spectrum recorded from a GaN / sapphire
film grown by the AEEE technique at 6000 C.













N shutter o;
for 10 sec.


I


pen






I
I
I
I
I
I
I


N shutter closed


Ga s
ope


Ga shutter close


I I
I I
I I
I I
I I
I I
I I
I I
I I
I I
I I
I I
I I

I I
I I
I I
I I
I I

I I
I I
SN delay time
(0sec., 15 sec. and 30 sec.)'
I I
I I


hutter
n for 5 sec.





I
I
I
-,q
I
I
I

I
I
I
I
I
I
I
I
I
I

I
I
I


15 sec.
delay


)U sec.
delay


Ga delay time


It
II
I 1
It
I

I
I
I
I
I
I

I
I














Ga delay used for all
Kerinments.


Time (sec)


Figure 4.5 RHEED specular reflection beam intensity recorded as a function of time
during a cycle of AEEE growth. The circles represent the beam intensities
at the three N delay times used in the N delay optimization set of experiments.
Growth temperature was 600" C.









250000 .
358nm


200000




Z 150000







0
S100000




50000





350 400




Figure 4.6


S 450 500 550 600 650


700


Wavelength (nm)


Low-temperature PL spectrum recorded from a GaN / sapphire
film grown by the AEEE technique with a 0 sec. N delay
at 6000 C.













80000 I I I .

72000 -23.5 K
358nm

64000

56000

S48000

.- 40000

, 32000

24000

16000

8000

0
350 400 450 500 550 600 650 700

Wavelength (nm)


Figure 4.7 Low-temperature PL spectrum recorded from a GaN / sapphire
film grown by the AEEE technique with a 15 sec. N delay time
at 6000 C.













70000

63000

56000

49000

42000

35000

28000

21000

14000

7000


0 -" i I I I I I i
350 400 450 500 550 600 650 700
Wavelength (nm)




Figure 4.8 Low-temperature PL spectrum recorded from a GaN / sapphire
film grown by the AEEE technique with 30 sec N delay time
at 6000 C















N shutter open
for 10 sec.


N shutter closed


I I
I I

|I Ga'~shutter
* pa for 5 sec.
I
I
I
I
I
I
I
I I
I I
I II
I II
I
at II


II
I
I
I
I
I
.
I
I
I
II

II
I
I
I


Ga shutter closed


15 sec. Ga
delay


Ga delay


Time (sec)


Figure 4.9 RHEED specular reflection beam intensity recorded as a function of time during
a cycle of AEEE for the 0 sec. N delay case. The circles represent the beam
intensities at the four Ga delay times used in the Ga delay optimization set of
experiments. Growth temperature was 600 C.


30 sec
Ga delay


50 sfc
Gas lay
t
1
I
I



I
I
I


t
t
I
I
I

I
I
I
I












































(b)


Figure 4.10 SEM micrographs of GaN film surfaces grown by
the AEEE technique using various Ga-delay times
(T..n. = 6000 C) (a) Ga delay time = 0 sec.
(b) Ga delay time = 15 sec.





















(c)
















(d)



Figure 4.10 Continued.
(c) Ga delay time = 30 sec.
(d) Ga delay time = 50 sec.













250000 .1 I I I I .

24.5K

200000




^ 150000




S 100000




50000 -358nm




0 L '
350 400 450 500 550 600 650 700

Wavelength (nm)


Figure 4.11 Low-temperature PL spectrum recorded from a GaN /
sapphire film grown by the AEEE technique with a
0 sec.Ga delay time.











250000




200000




r- 150000




s 100000




50000




0
35


' I I I I I I I I I I I I I j I I I I I I I I

24.5K



















-358nm

S I I I i I i I


;0


400


450


500


550


600


650


700


Wavelength (nm)




Figure 4.12 Low-temperature PL spectrum recorded from a GaN /
sapphire film grown by the AEEE technique with a 15 sec.
Ga delay time.









250000 -

358nm


200000




150000




100000




50000




0 1 A. L
350 40(




Figure 4.13


S 450 500 550 600 650


700


Wavelength (nm)


Low-temperature PL spectrum recorded from a GaN / sapphire
film grown by the AEEE technique with a 30 sec. Ga delay
time.










250000




200000




S150000


'5-

b4 100000




50000


0 1 L '.i. I I I I I I I I I I I 1
350 400 450 500 550 600 650 700

Wavelength (nm)






Figure 4.14 Low-temperature PL spectrum recorded from a GaN /
sapphire film grown by the AEEE technique with a 50 sec.
Ga delay time.












250000 i I I .
RI


200000




150000




100000


364nm

50000




0
350 400 450 500 550
Wavelength (nm)





Figure 4.15 Room-temperature PL spectrum recorded from a GaN /
sapphire film grown by the AEEE technique with a 0 sec.
Ga delay time.


600











250000




200000




2 150000

1s 0
4>
0u
S 100000


50000


0 L I i i i i I i I i I I i i I i
350 400 450 500 550

Wavelength (nm)





Figure 4.16 Room-temperature PL spectrum recorded from a Ga /
sapphire film grown by the AEEE technique with a 15 sec.
Ga delay time.


600












250000




200000




150000




100000




50000




0
350


400 450 500 550
Wavelength (nm)


Figure 4.17 Room-temperature PL spectrum recorded from
a GaN/ sapphire film grown by the AEEE
technique with a 30 sec Ga delay time.


600












250000




200000


150000




100000


50000




50
350


600


400 450 500 550
Wavelength (nm)




Figure 4.18 Room-temperature PL spectrum recorded from
a GaN / Sapphire film grown by the AEEE
technique with a 50 sec. Ga delay time.




















CI

Ga migration time ,'





i /r
sgstage 2














closed)



Figure 4.19 RHEED intensity recorded during a Ga-delay phase showing the
two distinct stages of recovery. Ga adatom migration times were
determined for a variety of substrate temperatures by noting the
time required for completion of stage 1 recovery as indicated at
each temperature.
S3










I5









two distinct stages of recovery. Ga adatom migration times were

time required for completion of stage I recovery as indicated at
each temperature.


















ED= 1.46 0.25 eV


I I I I I I


1.10


1.12


1.14


1.16


1.18


1.20


10'/ T (0 K)



Figure 4.20 Plot of In(Ga migration time) versus reciprocal
substrate temperature.


100


10 -


1.08


1.22








N shutter open

for 10 sec.


J I


N shutter closed


.Ga shutter open

:for 5 sec.


Ga shutter closed


I








.. ... .. .. .... .. .



RHEED shutter closed


time (sec.)


RHEED shutter

open


Figure 4.21 RHEED specular reflection beam intensity recorded as a function of time
during one cycle. The solid line represents the RHEED intensity trace when
the RHEED electron beam is continuously impinging on the surface, while
the broken line represents the intensity when the RHEED electron beam
shutter is periodically opened (-1 sec.) and closed (-5 sec.).


*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*

*
*



*













CHAPTER 5
INVESTIGATION OF GaN GROWTH KINETICS AND THEIR
CORRELATION WITH DEEP-LEVEL DEFECTS


Introduction


In this chapter, the AEEE growth mechanisms will be compared with the

mechanisms of conventional MBE growth, specifically, for a-GaN growth on sapphire.

First, a discussion of the growth mechanisms will be given based on the growth

parameters and the SEM characterization results which were presented in Chapters 3 and

4. Second, speculation will be made about the origin of the deep-level defects that are

believed to cause the 2.3 eV yellow band luminescence observed in all of the PL spectra

recorded from conventionally-grown a-GaN films, this speculation being based on the PL

spectra discussed in Chapter 3 and on the theoretical and experimental results concerning

deep-level defects discussed in Chapter 1. Finally, the optimum AEEE shuttering

sequence which contributes to the elimination of the yellow band emission will be

discussed. This discussion includes the presentation of a model concerning Ga adatom

migration which was developed as a result of the real-time RHEED intensity observations

during the Ga exposure and Ga delay phases.


Morphological Considerations


The surface morphology of conventional MBE-grown GaN films deposited on

sapphire is very sensitive to the growth parameters as illustrated by the SEM micrographs

shown in Figs. 3.5 and 3.6. Also, the yellow band emission appears in the PL spectra

recorded from all of the conventionally-grown samples in this study regardless of growth


I








temperature (Figs. 3.8-3.13). The substrate temperature and Ga/N beam flux ratio play

important roles in determining the GaN surface morphology. Since the nitrogen re-

evaporation rate is extremely high, particularly at the high temperatures required for

conventional growth, the residence time of nitrogen atoms is much shorter than that of Ga

atoms. Also, the incorporation of Ga is limited by the lack of N atoms on the growing

surface, especially at high temperatures. This situation is very different from the MBE

growth of GaAs, for instance, in which the growth rate is controlled by the Ga flux only.

Experimental evidence concerning the MBE growth of GaN indicates that the sticking

coefficient of Ga is only about 0.25 and attempts to increase the Ga flux or lower the

growth temperature lead to Ga-rich surfaces which are extremely rough [72]. Crawford

et al. [28] have reported that Ga accumulation on the growing surface actually decreases

the GaN growth rate when the substrate temperature is too low or the Ga flux is too high.

Consequently, it could be expected that the incorporation of Ga and N are correlated in

the MBE growth of GaN [53].

In this work, a strong dependence of the surface morphology of conventionally-

grown GaN films on the growth temperature has been confirmed. The surface coverage

of a GaN film grown at 7500 C is about 90 % as can be seen from Fig. 3.5. As

mentioned above, the desorption rate of N is much higher than that of Ga, hence, there are

fewer incorporation sites for the Ga adatoms. The observed micro-scale discontinuity of

the film will limit the surface migration length of Ga adatoms even at such a high growth

temperature, producing films having localized Ga-rich and Ga-deficient areas. Reducing

the growth temperature to 6400 C and to 6000C decreases the N desorption rate, however,

the migration rates of Ga adatoms and Ga-N molecules are also greatly reduced. The

surface morphologies of the films grown at 6400 C and 6000 C indicate the presence of Ga

droplets as can be seen from Fig. 3.6. The average size of the Ga droplets is about 15

pm for the sample grown at 6000 C and about 5 upm for the sample grown at 6400 C. It is

noteworthy that the Ga flux used for the sample grown at 600 C was actually three times








lower than that used for the sample grown at 640 C and yet the droplet size is larger for

the sample grown at 6000 C. This can be explained by assuming that the Ga desorption

rate is higher at 6400 C than it is at 6000 C and that Ga adatom diffusion is extremely

slow at 6000 C during conventional MBE growth.

As reviewed in Chapter 1 [29,45], with reference to GaAs growth kinetics, the Ga

surface diffusion length at 5500 C under Ga-stabilized conditions is about 10 times larger

than that under As-stabilized conditions. The migrating species on the growing surface

when GaAs is grown with an As overpressure are believed to be Ga-As molecules rather

than isolated Ga atoms, since impinging Ga-atoms immediately react with impinging As

molecules to form Ga-As molecules. As in conventional MBE-growth of GaAs which is

normally carried out using an As-overpressure, conventional MBE-growth of GaN is

normally carried out using an N overpressure. In this case, the impinging Ga-atoms have

a high probability of immediately reacting with the impinging N-atoms to form Ga-N

molecules; thus, surface migration takes place as Ga-N molecules (or GaN clusters)

rather than isolated Ga-atoms. The large activation energy for migration of the Ga-N

molecule on the GaN growing surface is believed to result in extremely slow migration.

The short diffusion length of Ga-adatoms under these conditions enhances the formation

of Ga-aggregates which become sinks for additional Ga adatoms. This phenomenon is

believed to have caused the slow GaN growth rates observed in references 27 and 28.

In contrast, samples grown at 6000 C using the AEEE growth technique had

surface morphologies devoid of Ga droplets as can be seen from the SEM micrographs

shown in Figs. 4. 10 (a)-(d). It is speculated that in this case rapid Ga adatom migration

takes place in the absence ofan N atom flux and thus Ga clustering is avoided.







Speculation Concerning the Source of Deep-Level Defects
Responsible for the Yellow Band Emission

The yellow band emission intensity has a growth temperature dependence which is

evidenced by comparing the PL spectra recorded from the AEEE samples grown at 6000C

with those of samples grown at higher temperatures (Figs. 4.2-4.4). Yellow band

emission is, in fact, completely absent from the spectra recorded from samples grown at a

substrate temperatures of 600" C. It is postulated that higher growth temperatures result

in the desorption of Ga during the Ga delay phase producing Ga-vacancies which when

associated with dislocations result in the yellow band emission. Dislocation densities have

been estimated to be in the range, 108 01o0 cm-2, in typical GaN films grown on c-plane

sapphire [25,26,54,61]. The deposition of Ga on GaN (0001) surfaces followed by

thermal desorption of the Ga has been reported by Khan et al. [73]. In that study, Ga

was deposited at 6200 C and was found to evaporate at temperature around 670 C which

supports the Ga re-evaporation theory. Also, Schoonmaker et al. have reported that

GaN decomposition can be enhanced via catalysis by metal Ga [74]. Furthermore, as

reported in Chapter 4, it was found in this work that during the Ga delay phase, the

RHEED specular reflection beam intensity recovered faster at growth temperatures of

700" C and 640" C but recovered to a lower level of intensity than was the case for a

sample grown at 6000 C as shown in Fig. 4.1. Also, for a Ga exposure time at 7000 C of

5 sec. instead of an optimum 8 sec, the RHEED specular reflection beam intensity was

found to gradually decrease as the shuttering sequence progressed. It is believed that

Ga desorption under N-free conditions is more rapid at higher temperatures and Ga

vacancy generation is promoted. Ga vacancies were predicted as the most likely deep-

level candidate in n-type unintentionally doped GaN in the literature review section

(Chapter 1). Similar observations concerning Ga re-evaporation at relatively low

temperatures have been made with regard to MBE-GaAs. Arthur [75] reported that Ga

deposited on a GaAs (111), Ga-rich surface, has a sticking coefficient of unity below








477C. However, the Ga sticking coefficient is much less than unity unless an As flux is

present when the substrate temperature is in the range, 500- 6000 C.

Strong yellow band emission is evident in the PL spectra recorded from GaN films

grown by conventional MBE as indicated in Figs. 3.8-3.13. Also, as shown in Fig. 3.6, a

high density of Ga droplets is observed on the surfaces of conventionally-grown films. It

is speculated that Ga vacancies are again involved in this case since the material providing

the luminescence will be Ga-deficient due to the participation of Ga adatoms in Ga droplet

formation.

As mentioned previously, radiative transitions in GaN epilayers are thought to be

associated with dislocations [25,26,61], in particular, screw dislocations which are

uncharged in hexagonal GaN [61]. The near-band-edge emission has previously been

ascribed to the recombination of electrons with holes trapped at screw dislocations [61],

as indicated in Fig 5.1. Based on the results of this present study, it is speculated that the

yellow band emission is due to electrons recombining with holes trapped at Ga vacancy /

screw dislocation complexes as illustrated in Fig. 5.1. As the concentration of Ga

vacancies increases, the probability of holes being trapped at VG. / dislocation complexes

will increase and correspondingly so will the intensity of the yellow band emission.


Influence of the Ga and N Exposure Times and Delay Times.

When an N atom flux impinges on a Ga-covered surface while the Ga shutter is

closed, the N atoms will incorporate into the appropriate lattice-sites easily because the

surface Ga atoms have one dangling sp2 bond. As mentioned in Chapter 4, a 10 sec. N

exposure time was required for the RHEED specular reflection beam intensity to decrease

to its lowest level and then to recover to a saturation level beyond which the intensity

again dropped. For N exposure times longer or shorter than 10 sec, the RHEED

specular reflection beam intensity did not recover to the same level as it did in the case of








the 10 sec N-exposure time during subsequent cycles. Consequently, a 10 sec. N-

exposure time was thought to be optimum for the substrate temperature range employed

and such an N exposure time was used throughout this work.

As can be seen from Figs. 4.2 and 4.3, the PL spectra recorded from AEEE-

grown GaN deposited at 7000 C and 6400 C show weak peaks in the range, 365-385 nm.

Such emission is believed to be related to N vacancies [20,37] which provide the group B

level shown in Fig 1.2. Similar emission is also observed in the PL spectra ( Figs. 4.7

and 4.8) recorded from samples grown at 6000 C using N-delay times of 15 sec. and 30

sec., however, this emission is absent from the spectra recorded from the samples grown

with a 0 sec. N-delay time (Figs. 4.11-14). Such evidence suggests that under Ga-free

conditions (no Ga exposure), the N-adatoms have a short residence time and, hence, the

optimum N-delay time was selected as 0 sec. in this work. In general, an N atom

arriving at a chemically-active Ga-covered surface, which has one unsatisfied sp2 bond

perpendicular to the surface per Ga atom will tend to form a stable sp3 bond with

underlying Ga and neighboring N atoms. This will make N incorporation homogeneous

on a Ga-covered surface (similar to As2 on a Ga stabilized surface [76]). However, N

has a very high vapor pressure and so excess N atoms may be only weakly adsorbed on

the surface. Therefore, N desorption is proposed to be the dominated mechanism

contributing to the moderate recovery of the RHEED specular beam intensity during the

N delay phase, as shown in Fig. 4.5.

A primary goal of the AEEE growth mode was to promote the migration of Ga-

adatoms on N-covered surfaces under N-free conditions (no N exposure) at temperatures

lower than those required for conventional MBE growth. The purpose of the Ga-delays

was to eliminate the Ga clusters which are thought to form on Ga-exposed surfaces. Isu

et al. [68] used a scanning reflection electron microscopy (SREM) technique to observe

the formation of Ga-droplets on Ga-exposed, As-covered surfaces. These authors also

observed that the Ga droplets shrank in size when the Ga supply was interrupted.








The density of available incorporation sites for impinging Ga atoms is 1.1347 x

1015 cm2 for GaN. The Ga-exposure time in this work (T.btne = 6000 C) was based on

the maximization of the RHEED specular reflection beam intensity level. For a Ga

effusion cell temperature of 830 C, corresponding to a Ga flux of 2.27 x 1014 cm-2 s", a

5 sec. exposure provided about 1.135 x 1015 Ga-atoms cm'2. When the Ga shutter was

opened, the RHEED specular reflection beam intensity decreased to its lowest level in

-2.5 sec. and then recovered to a saturated level in an addition 2.5 sec. It was found

that the RHEED specular reflection beam intensity recovered to its highest level for a Ga

exposure time of 5 sec. A shorter Ga exposure time resulted in a lower intensity level for

the recovered signal while a longer than 5 sec. Ga exposure resulted in a longer recovery

period. The growth rate of all AEEE-grown GaN films regardless of the Ga-delay time

was measured to be -63% of a monolayer per cycle for a substrate temperature of 6000 C.

In MOCVD "atomic layer epitaxy "of GaN, both Khan et al. and Sumakeris et al.

reported about 2/3 of a monolayer per cycle growth although the films grown by

Sumakeris et al. were amorphous [55,56].


Model for Ga Adatom Migration

As discussed in Chapter 4 and illustrated in Fig. 4.20, an activation energy for Ga

migration on an N-free surface of 1.46 0.25 eV was determined in this work. To the

best of the author's knowledge, this is the first time that the activation energy of Ga

migration on GaN surfaces has been reported. Compared to the activation energy of Ga

diffusion on GaAs, given as, Ed = 1.3 eV, by Neave et al. [44], the barrier for Ga

migrating on a GaN surface is higher. It must be point out, however, that the arsenic

flux was on during the Ga delay in the GaAs case since Neave et al. tried to simulate

conventional MBE growth. Ed is expected to be smaller than 1.3 eV for Ga-adatom

migration under As-free conditions.









The adatom migration rate at site i is taken to be in an Arrhenius expression:

R' = Ro' exp (-E/k T)--------------------------- -----------(5.1)

where Ro' is a prefactor which is related to the vibration frequency of site i surface atom,

T is the substrate temperature, kg is the Boltzmann constant and E is the activation

barrier for migration which depends on the local configuration of site i. E includes a

substrate barrier E, and a barrier E, contribution from each nearest-neighbor bond formed

parallel to the substrate. The homonuclear bond energy of Ga is 1.71 eV [78] and the

Ga-N bond energy is determined to be 3.1 eV using Pauling's approximation [79]. The

larger activation energy of Ga migration on GaN compared with Ga on GaAs is

understandable since GaN has a larger E. (ionic bonding) than the E. covalentt bonding)

of GaAs.

A model has been developed to explain the RHEED specular reflection beam

intensity behavior observed during Ga exposure and subsequent Ga-delay time phases of

growth. The discussion of the model will be presented in terms of three stages, namely,

the Ga exposure stage, the first stage of recovery and the second stage of recovery. These

three stages are illustrated in Fig. 5.2 with regard to a typical RHEED intensity trace

recorded in this work during a cycle of AEEE growth.


Ga Exposure Stage:

As can be seen from Fig. 5.2, the specular reflection beam intensity initially

decreases, goes through a minimum and then increases again during the Ga exposure.

What is not illustrated in this figure is the fact that for Ga exposures longer than 5

seconds, the specular reflection beam intensity saturates at the intensity level

corresponding to a 5 second Ga exposure.







An attempt is made to explain the "Ga exposure stage" behavior with reference to
the illustrations presented in Fig. 5.3. As mentioned in the previously section, dislocation
densities in the range, 108- 1010 cm'2, have been reported for typical GaN films grown on
c-plane sapphire [25,26,54,61] and most of these dislocations are pure edge type, existing
on {1 1 0 0} planes. However, screw, or mixed type, dislocations are also present and
account for -30% of the total dislocation density [79]. The magnitude of the Burger's
vector is equal to the lattice parameter, c, for screw dislocations (5.18 A in GaN) and

a/3 [ 1 2 0] for edge dislocations ( a is equal to 3.19 A in GaN). Therefore, because the
dislocation nucleation energy increases linearly with k. b2, where b is the magnitude of the
Burger's vector and k is a constant, edge dislocations account for a large fraction of the
total dislocation density because of their lower formation energy. The edge dislocations
which formed on (1 1 0 0 } during the coalescence of 3D-islands in the early stage of
GaN / sapphire growth will terminate on (0001) surfaces. These points of termination
disrupt the orderly array of surface atoms as seen from Fig. 5.3 (a).
Consideration of mechanical equilibrium among the interfacial tensions in the case
of heterogeneous nucleation yields Young's Equation Ir cos 0 = sv Ys as shown in
Fig.5.4 (b), where f, s, v and 0 represent film, N-covered GaN surfaces, vapor and contact
angle, respectively. AG* (energy barrier to the nucleation) will depend on the surface
energy of the cluster (If) and contact angle between the cluster and the surface ( ) for

the heterogeneous nucleation case and can be represented as:
AG* = 16 x (1r)' f(0) / 3 (AGv)2 -- ------------------------------(5.2)

where f (0) = (2 + cos 8)(1-cos 0)2 / 4, and AGv is the free energy change per unit volume

accompanying of the of formation of small clusters. Lattice mismatch between film and
substrate will increase the overall energy barrier (AG* ) because the strain energy is
involved in AGv. However, AG* will decrease when the surface defects are presented.

When a surface having a large dislocation density [as shown in Fig. 5.3 (a)] is
exposed to the Ga flux, Ga adatoms will nucleate on the termination sites of dislocations







as seen in Fig. 5.3 (b). Ga aggregates having radii smaller than r* will shrink and

disappear, lowering the free energy. At the same time, because of their lower energy
barrier ( AG* ) to the nucleation process, nuclei having radii larger than r* will grow to

supercritical dimensions by the further addition of atoms as shown in Figs. 5.3 (c)-(d) and
Fig. 5.4 (a). After 2-2.5 sec. Ga exposure ( circle c and d on the RHEED intensity
variation curve in Fig 5.3), islands several atomic-layers high are formed and the RHEED

specular reflection beam intensity drops rapidly to the lowest level, indicating a rough

surface. The RHEED specular reflection beam intensity recovers after a 2.5 sec. Ga

exposure and it is proposed that the small Ga islands merge to form larger Ga clusters,

flattening the cluster surfaces which in turn decreases the surface roughness.

For 5 sec. Ga exposures and longer (circle e on the RHEED intensity variation

curve in Fig. 5.3), the RHEED intensity starts to saturate. For this condition, it is

proposed that Young's equation for partial wetting conditions, Tr > (7Yv Tf. ) > Tf, is

balanced at a critical contact angle. Further Ga deposition will be limited by increasing

f(0) [Fig. 5.4 (c)] which increases the energy barrier for the nucleation process (AG*).


First Stage of Recovery:

As can be seen from Fig. 5.2, following closure of the Ga shutter, which prevents

further Ga exposure, there is a fast recovery phase which lasts for about 15 seconds (for a

substrate of 6000 C). During this first stage of recovery, it is expected that Ga adatoms

migrate fairly rapidly (in the absence of N atoms) on the terraces to the steps which

results in a smoothing of the growth front as illustrated in Figs. 5.5 and 5.6. Large Ga

clusters are unlikely to move freely on the terraces, however, the Ga adatoms can detach
themselves from the edges of the clusters and migrate quickly to the step (kink or ledge)

sites. Also at this stage, the probability of Ga finding incorporation sites is high and the








mean free path for Ga adatom migration is large. It must be pointed out that the Ga

adatoms have equal chance to either migrate to the upper or the lower terraces.


Second Stage of Recovery:

As can be seen from Fig. 5.2, for Ga-delay times in excess of 15 seconds (at

6000C), a second recovery stage begins in which recovery (surface smoothing) is not as

rapid as in the previous stage. During this stage, it is expected that a rearrangement of

the terraces takes place which impacts longer range order as shown in Fig. 5.5 [67,69,70].

For Ga delays longer than 30 sec. the RHEED intensity drops slightly. It is speculated

that step bunching occurs in these cases in which the terrace width locally decreases [80].

As can be seen from Figs. 4.11-14 (low-temperature PL spectra recorded from

AEEE-grown samples with different Ga-delay times, namely, 0 sec, 15 sec, 30 sec and 50

sec.), the near-band-edge emission intensity is about an order of magnitude higher from

the 30 sec. delay sample compared to the 0 sec. or 15 sec. delay samples. Also, weak

deep level emission centered around 450 nm is present at room temperature from the 0

sec. and 15 sec. delay samples but not from the 30 sec. delay sample as shown in Figs.

4.15 and 4.16. These results considered together with the model suggest that the

achievement of long-range order (as indicated by the observation of the second stage

recovery of the RHEED intensity) produces the highest quality epilayers. However, as

shown in Figs. 4.14 and 4.18, the near-band-edge emission intensity is again lower from

the 50 sec. delay sample (compared with the 30 sec. delay sample) and so it appears that

epilayer quality is degraded when step bunching is involved.













near-band-
edge emission
(365 nm) <


Hole trap level provided
by screw dislocations
(after ref. [61])


0,4


..


I U U


7/
AI


S_e


-.~ Yellow band
>emission
(-550 nm)
K


2.3 eV


Range of hole trap
levels associated with
Ga vacancy / screw
dislocation complexes


Fig. 5.1 In defective GaN epilayers, efficient radiative transitions are associated which screw
dislocations that are either isolated or are in association with native point defects.


m


Alk
MW 6-


I I


I I


r










N shutter open
for 10 sec.






Ga shutter
open for 5 sec.


N shutter closed


Ga shutter closed


I I
I I
I I
I I


Ia I
I I
Ga'=
I


exposure
stage
I I
I I

I I

I I
I I

I I
I I

I I
I I
I I
I I


Ist stage
of recovery/
/


7,
/I
'I

//
/ I

-* -


I
1
I
I
I
I
I
I
I

I
I
I


r


time (sec.)




Figure 5.2 A typical trace of the RHEED specular reflection beam intensity
recorded during one cycle of AEEE illustrating the three stages
of the process discussed in the model presentation.








N covered surface


edge dislocation
line


-rrr p--a


I I


\9 '


(a)


(b)


Ga shutter open
for 5 sec.







n(a)


0 2 I
0 2 3


closed


Time (sec.)


Figure 5.3 Ga cluster nucleation and evolution during the Ga exposure phase (in the absence of nitrogen)
as suggested from the RHEED intensity observations.


Ga shutter cl



I /
| ,

/
I /
I -
I /
I
/
I
/


(b)




(c)


w AL

(d)






(e)


II~------- ------------


VWl


. l


\ -- V


AM WL


& 6


At, .0%




Full Text
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID E18HL6GQJ_AAGWWJ INGEST_TIME 2013-04-08T22:26:34Z PACKAGE AA00013538_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES


xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID EEOCH8QKJ_USQDO3 INGEST_TIME 2013-03-12T14:08:16Z PACKAGE AA00013538_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES



PAGE 1

IMP ACT OF THE GROWTH KINETICS ON DEEP LEVEL DEFECT PRODUCTION IN GaN FILMS GROWN BY MOLECULAR BEAM EPITAXY By HSING-LONG LIU A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 1996

PAGE 2

ACKNOWLEDGMENTS I would like to express my greatest appreciation to Prof Robert M Park my supervisory committee chairman for his constant professional guidance support and encouragement during the research and the writing of this dissertation I would also like to thank Prof Paul H Holloway Prof Joseph H Simmons Prof Kevin S Jones and Prof Sheng S Li who kindly served on my dissertation committee I extend special thanks to George Kim and Paul Precure my fellow graduate students whose assistance in this research is gratefully acknowledged especially thanks to George for providing AFM images of the sapphire substrate morphology Special thanks also go to Lynn Calhoun for sharing his knowledge ofRHEED intensity monitoring Thanks also to George Lynn Paul and Brent Gila for their friendship I would also like to acknowledge Dr. Matthias Ludwig of Dr. Hummel s group for his assistance with the PL measurements and for helpful discussions . I would like to thank my parents especially my father who passed away in 1993 for their prayers and love And finally I thank my Savior Jesus Christ. ii

PAGE 3

TABLE OF CONTENTS ACKNOWLEDGMENTS ................ . ..... .. ........................................... .... .. . .. . ... ... .. ii ABSTRACT ................. .............. ........... .......... .. .. .. ......... . ... . ................................. v CHAPTERS 1 INTRODUCTION ............................................. .. .. ....... ... ....... .. ... ............. ...... 1 Motivation and Objectives ... .. .. .... .. . .. ...... .. .... ........ . ..... . . ..... .. ... .. . .... ..... ...... 1 Background ........................................... .. .. .. .... .. ..... .... ....................................... 5 Theoretical Models of Deep-Level Defects in GaN .................................. ...... 5 Experimental Observations of Deep Levels ........................... .. . ...... .... .. ... ..... 7 Migration Enhanced Epitaxy (MEE) with reference to GaAs ........................ 11 Migration Kinetics During MEE With Reference to GaAs ............................ 12 MOCVD Growth ofGaN Using an Alternate Source Exposure Method ....... ............................................................................................ 14 Dissertation Outline .. ..... .................................................................................... 15 2. MBE SYSTEM AND CHARACTERIZATION TECHNIQUES .......... .. .... ... . 19 Introduction ... ...... ... ....... .................. ... ... ........... ..... . ... ................................. 19 Molecular Beam Epitaxy ... .. ........... .. ....... .... ... ..................... ..... .. ......... .. .. ... . ... 19 Varian GEN II MBE System Configuration .................................................. 20 Substrate Temperature Calibration ........... ...... ....... .................................. ... 22 In-situ Analyzer ..... .. .... .. . .. ... ..................................... .................. .. ........ . ... 22 Ex-vacuo RHEED Specular Reflection Beam Intensity Monitoring System ....................... .. .. .. ... ... ..... .. .. ... .......... .. .................. .. .. .. ........... 23 Nitrogen Free-Radical Source ...... .. ..... .. ... ............ ....................................... 24 iii

PAGE 4

3 GROWTH BY CONVENTIONAL MBE AND CHARACTERIZATION OF GaN / SAPPHIRE FILMS ............................................................................ 32 Ex-vacuo Substrate Preparation ........................................................ . ... ............. 32 In-situ Substrate Preparation ............ .......... .. ..... ... . .......................... .. . .... . .... 34 Conventional MBE Growth ofGaN on Sapphire .. ... .. .. ...... ......... .. ................. 34 Characterization of Conventionally Grown GaN Films .... . ........ ... .................... 3 6 SEM Analysis ........ ................... .... ....... ........ .... . . .. . . ...... .... ... ............ 3 6 Hall-effect Measurements and PL Analysis .... .. . .. . ...... ... ............. . ...... 36 4 INVESTIGATION OF A NOVEL GROWTH MODE .... . . .. ............... . .. ......... 51 Introduction . .. .... ... . ....... .. .. .... .. .... ......... ... ................ ... ................... ....... .. ..... 51 AEEE Growth Procedures ... ... ... ... ... .. ... ........ .......... ... .. . .... . ... ....... .. ...... .. 52 AEEE Growth Temperature Optimization .... .. ...... ...... ... ........................ . ...... 52 Nitrogen Delay Time Optimization ........ .............. .. ..................................... ...... 54 Ga Delay Time Optimization .. . ......... ...... .............. ..... . .... . ........ ... ... ... ..... 54 Activation Energy of Ga Migration under N-Free Conditions ...... ....... ............. 56 Impact of Impinging RHEED Beam on Ga Migration ............... ......................... 57 5 INVESTIGATION OF GaN GROWTH KINETICS AND THEIR CORRELATION WITH DEEP-LEVEL DEFECTS ..... .. . ... ...... .............. . .. .. 81 Introduction .......... .............................. ............ ..................................... .......... 81 Morphological Considerations .......... .. ............. ........ . .. ................ .. ... . . .......... 81 Speculation Concerning the Source of Deep-Level Defects Responsible for the Yellow-Band Emission ......................... .. ..... ....... ... ... . . . .... ..... .... ... 84 Influence of Ga and N Exposure Times and Delay Times ............... . . ... .... ...... .. . 85 Model for Ga Adatom Migration ........ ..... .. .................... ................................... 87 Ga Exposure Stage ... .. ... .. .. .... .... .... ... .. ... ... ..................... ...................... .. 88 First Stage of Recovery ...... .. ..... .... . .. ..... .. .. .. .. .. . .......................... . ....... 90 Second Stage ofRecovery .. .... ... . .. ......................... ... ...... . . . .... .......... 91 6 CONCLUSIONS AND RECOMMENDATION ............... .... . .. . . ........ . .. .... ... 99 REFERENCE LIST .............. ... .... ..... ... .. ... ... .. ............... ..... .......................... ... ...... 102 BIOGRAPHICAL SKETCH ......... .................................... .. .. .. ......... ................. ..... 108 iv

PAGE 5

Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosoph y IMP ACT OF THE GROWTH KINETICS ON DEEP LEVEL DEFECT PRODUCTION IN GaN FILMS GROWN BY MOLECULAR BEAM By Hsing-Long Liu December 1996 Chairman : Robert M Park Major Department: Materials Science and Engineering As is presently understood the luminescence behavior of GaN epitaxial material is dictated by the presence of both extended defects (dislocations) and also native point defects Near-band-edge luminescence at around 365 nm (ascribed to excitons bound to dislocations) is typically observed in addition to a so-called yellow band luminescence centered around 550 nm (2 3 eV) whose origin is the subject of speculation The specific objective of this research was to establish a relationship between growth kinetics and the production of deep level defects in GaN epitaxial films by correlating morphological changes monitored in-situ and in real-time during epitaxial growth with yellow band emission intensity variations recorded post-growth and e xs itu The role of Ga adatom migration on growing GaN surfaces was found to be critical with regard to the production of deep level defects and it was found that a nitrogen-free surface condition was optimum to promote efficient Ga adatom migration as V

PAGE 6

was the provision of a sufficient time delay following Ga flux exposure prior to subsequent N flux exposure The critical kinetic processes such as Ga migration were monitored in real-time during epitaxial growth using reflection high energy electron diffraction (RHEED) specular reflection beam intensity monitoring Based on this analysis and the noted correlations between surface morphological changes and the yellow band luminescence intensity variations it is speculated that Ga vacancies ( associated with insufficient Ga adatom migration) are responsible in association with dislocations for the deep level enuss1on An activation energy of 1.46 0 25 eV has been determined for Ga adatom migration on N-free GaN surfaces and a model is presented that accounts for the kinetic processes taking place during both Ga adsorption and migration phases of growth under N-free (no N-flux arriving) conditions vi

PAGE 7

CHAPTER 1 INTRODUCTION Motivation and Objectives Gallium nitride ( GaN) has a direct band-gap of about 3 4 e V at room temperature (RT) and when alloyed with aluminum nitride (AIN) and indium nitride (InN) has the potential to be applied to the fabrication oflight-emitting diodes (LEDs) and laser diodes (LDs) operating in the blue-green and ultraviolet (UV) wavelength regions of the optical spectrum World-wide interest in GaN was intensified by the recent realization of GaN-based blue LEDs, blue LDs, and solar-blind photodetectors The high-output GaN-based blue LEDs are in high demand for outdoor giant displays desktop flat-panel displays and traffic lights, as well as devices which emit white light by combining red green and blue emitters Moreover, the blue LD will have far-reaching implications for space and under sea communications as well as high-density optical data storage Such short-wavelength blue emitters will provide the small spot size needed to increase the storage capacity of optical storage systems over that of the current near-IR emitter-based technologies and provide faster data-transfer rates for computer and multimedia use thus meeting the growing demands of the computer industry GaN also is a promising semiconducting material having potential applications in electronic devices operating at high temperature high frequency, and high power. Despite some impressive technological achievements of the last few years particularly by Nakamura et al. [I] ofNichia Chemical Industries in Japan (demonstrating the first electrically-pumped blue laser based on the InGaN/GaN multiple quantum well 1

PAGE 8

2 system) much work remains to be done to realize a low-threshold long-lived LD suitable for commercial application Further progress toward the development of stable continuous wave (CW) operation oflll-V nitride LDs will require a thorough understanding of how impurities and defects influence the quantum efficiency of these materials For instance it will be important to understand the origin of deep level defects in GaN which are known to cause yellow band luminescence suppressing the more desirable band-edge luminescence The yellow emission band is commonly observed in the photoluminescence (PL) spectra recorded from nominally undoped GaN epilayers regardless of the substrates used (a-plane sapphire (2 3] c-plane sapphire [2 4 5 6] silicon carbide (a-SiC) [7 8] Si (111) [9] and bulk GaN [10 11]) It is also observed regardless of the crystal growth technique used (low-pressure metal organic chemical vapor deposition (LPMOCVD) [8] metalorganic chemical vapor deposition (MOCVD) [4 11 12] metal-organic vapor phase epitaxy (MOVPE) [2 3 5 9] MBE [6 8 10 13 14] or high pressure methods [10]) Ramsteiner et al. observed electronic Raman scattering in resonance with the yellow luminescence transition and concluded that the deep-level concentrations must be rather large to make Raman resonance observable [15]. There is little understanding of the deep-level defects inn-type GaN however sources of the deep-level defects have been proposed such as Ga vacancies (V Ga) [16 17 18 19 20] antisite defects (NGa and GaN) [21 22 23] and Ga interstitials (Ga1) [22 23 24]. All of these theories are based on the assumption of thermodynamic equilibrium but GaN films are currently grown under conditions that deviate significantly from equilibrium In addition these calculations use atomic structure factors calculated for neutral spherical atoms They do not take into account the effects of the charge redistribution due to chemical bonding (ionic bonding) and the stress field induced by high densities of dislocations and strong inhomogeneities present in the films CL images in fact have shown strong inhomogeneities [25]. A hypothesis has recently been proposed that states that the deep-level broadband emission arises directly from the atomic structure

PAGE 9

3 of dislocations or is associated with the clustering of native point defects ( vacancie s, interstitials) at dislocations due to the lattice distortions induced by the dislocation s [25 26]. The work described in this dissertation was inspired by a desire to minimize t he concentration of deep level defects occurring in GaN epitaxial films through an understanding of the factors affecting their production The highest quality GaN films grown today are produced by conventional MOCVD and MBE at substrate temperatures in the neighborhood of 1000 C and 700 C respectively Such high growth temperatures could result in Ga rich material due to desorption of the N adatoms At low substrate temperatures however the diffusion rate of Ga atoms is slow compared to the growth rate of the GaN film which can result in Ga accumulating into small Ga droplets on the growing surface which can impede the migration of nitrogen atoms [27] Ponce et al. reported that Ga droplets on the order of 10-100 m in size were found in the GaN substrates obtained from high pressure growth methods [ 11] These Ga droplets are very common even in MBE-grown GaN films particularly in films grown at low growth temperatures In fact Crawford et al. [28] recently reported that excess surface Ga decreases the GaN formation rate when the substrate temperature is too low or the Ga flux is too high in the MBE growth case A systematic study of Ga migration and incorporation at some optimum growth temperature therefore seems important at this juncture with a view to controlling the deep level defects in GaN Ga-migration has been investigated on GaAs surfaces For instance Horikoshi e t al [29] employed an alternate element supply technique using As and group III species to grow AIGaAs/GaAs quantum well structures These authors reported excellent PL characteristics and smooth interfaces for films grown at a relative low growth temperature and attributed these results to a dramatic increase in the Ga atom migration rate under As free conditions Ewing and Greene [30] reported that growth on the Ga-terminated

PAGE 10

4 ( 111) A face of GaAs was slow and produced films which had a tendency to form planar surfaces On the other hand growth on the As-terminated (1 1 I) B face was fast and film surfaces were rough The { 111} planes of cubic GaAs are similar to the (0001) ( or basal) plane of wurtizite GaN grown on for instance the (0001) plane of sapphire The GaN basal plane has a polar configuration with two atomic subplanes each consisting of either the cationic or the anionic component of the binary compound Therefore an alternating supply of Ga and N atoms may in principle be used to grow (0001) GaN epitaxially possibly at growth temperatures lower than those used for MBE-growth Such an alternate element exposure growth method has not however been reported previously for MBE-GaN. In the presently described work a growth technique which we call alternate element exposure epitaxy (AEEE) is introduced in which the Ga and N fluxes do not impinge on the surface at the same time. During AEEE the Ga adatoms are assumed to migrate rapidly on the growing surface under N-free conditions and by virtue of forming a Ga-covered surface and hence a chemically active surface the N sticking coefficient is expected to be enhanced as well In addition the insertion of a Ga delay time ( the time allowed to elapse between closing the Ga shutter and opening the N shutter ) should allow Ga adatoms to migrate to step or kink sites and hence to be incorporated into the GaN film In this manner point defects such as Ga vacancies and Ga interstitials which are thought to be associated with the deep level emission in GaN may be minimized The specific objective of this research was to establish a relationship between growth kinetics and the production of deep level defects in GaN epitaxial films by correlating morphological changes monitored in-situ and in real-time during alternate element exposure epitaxy with yellow band emission intensity variations recorded post growth and ex-situ

PAGE 11

5 Background Theoretical Models of Deep-Level Defects in GaN There are three main theoretical models which attempt to explain the origin of the deep-level emissions observed from unintentionally doped n-type GaN All of the models predict that the deep-levels are related to native defects and impurities Applying a tight binding Hamiltonian model Jenkins and Dow [21] suggested that a Ga vacancy (V Ga) can act as either an electron or a hole trap with a p-like level in the gap and the antisite defect (GaN) can also act as either an electron or a hole trap Jenkins and Dow also suggested that an N vacancy (VN) results in a shallow donor state with a p-like level in the conduction band and its s-like deep level occupied in the gap This will be discussed later. By performing first-principles calculations under the assumption of thermodynamic equilibrium Van de Walle and co-workers [16-19] predicted that nitrogen vacancies have a rather high formation energy (~4 eV) inn-type as-grown GaN; however n-type behavior has traditionally been attributed to nitrogen vacancies Van de Walle claimed that self-interstitials (Gai and Ni) and antisite defects (NGa and GaN) also have high formation energies as shown in Fig 1 1 ; therefore they are also very unlikely to occur in significant concentrations However these authors added that the quality of the GaN epilayers can depend strongly on the concentration of line defects such as dislocations and stacking faults etc The first-principles total-energy equation for a native defect in a charge state q is as follows : Ff ( q) = Et o t ( q) n Ga Ga n N N qEF ----------------------------( 1. 1) where nGa and n N are the number of Ga and N atoms Ga and N are the Ga and N chemical potentials EF is the Fermi energy and got is the total energy for a given charge state The chemical potentials for Ga and N are not independent since both species are in equilibrium with GaN. If boundary conditions such as N < N2 and Ga< a(bulk) are

PAGE 12

6 not satisfied the system would be thermodynamically unstable and would segregate into different components Figure 1. 1 shows that a Ga vacancy may have a low enough formation energy to occur in high concentrations Van de Walle et al suggested that Ga vacancies may play a role in the yellow band luminescence Boguslauski et al. (22 23 24] used an ab initio molecular dynamics approach in which the computations were carried out for the case of the perfect crystal using quantum molecular dynamics (MD) and the local density approximation (LDA) in supercells containing 72 atoms Due to the wide-gap of nitrides the formation energies of defects depend strongly on the position of the Fermi level. It is important to point out that because of the inaccuracy due to the use ofLDA theory which seriously underestimates band-gap and overestimates the cohesive energy an uncertainty on the order of0 5 eV for the formation energy of the deep levels may occur. This is especially true if the growth process involves local excitations due to a plasma source or due to specific kinetic conditions Inn-type GaN the high formation energy ofV N (4 8 eV) would indicate that the N vacancy cannot be formed in sufficient concentrations to account for the n-type character of the as-grown material. However since the formation of a Ga vacancy creates N dangling bonds its quasi-triplet level is located about 0.4 eV above the valence band edge V Ga can trap both electrons and holes because the quasi-triplet is populated by 3 electrons in the neutral charge state. In spite of the proximity to the valence band the quasi-triple wave functions are very localized The formation energy of neutral V Ga may appear large at 8 e V ; however since V Ga can release or accept three electrons its formation energy depends very strongly on the position of the Fermi level. In particular inn-type GaN films the formation energy of V Ga 3 may become very small. The thermodynamic levels for Ga vacancies are at 0 5 0 8 and 1 3 eV above the top of the valence band maximum respectively Boguslauski (23] concluded that the small formation energy defects in n-type GaN are V Ga GaN and N1.

PAGE 13

7 Experimental Observations ofDeep-Levels Ogino and Aoki [3 1] investigated the mechanism causing yellow band luminescence in GaN A yellow band emission was observed in microcrystals synthesized from Ga and NH 3 by direct reaction but did not appear in needle-like crystals grown by sublimation-recrystallization These authors proposed a mechanism for the zero-phonon transition assigning the yellow-band emission to a radiative transition from a shallow donor (25 meV below the conduction band minimum) to a deep acceptor (860 me V above the valence band maximum) The deep acceptor was assigned to a complex consisting of a Ga vacancy and a carbon atom Ram steiner et al. [ 15] recently confirmed that shallow donors are involved in the yellow luminescence by studying the electronic Raman scattering in resonance with yellow luminescence transitions in a 1 m thick GaN film grown on GaAs (001) by conventional :MBE No specific deep localized state was assigned by Ramsteiner et al. Tansley and Eagan [20] compared the experimental data on the location of the levels associated with native point defects in the group-III nitrides with the theoretical estimates These authors found strong evidence for the existence of a triplet of donorlike states associated with N vacancies and suggested that the three electron states associated with N vacancies in GaN are located at about 30 me V 100 me V and 400 me V below the conduction band minimum as shown in Fig. 1 2 (groups A, B and C respectively) A deep level has been observed with thermal activation energies between 0 8 and 1 1 eV below the conduction-band edge as shown in Fig 1 2 as group D This deep level has been attributed to Ga vacancies [ 16] and N anti site defects [21]. Both Tansley and Eagan [20] and Van de Walle et al [16] proposed that group Dis the obvious candidate for the yellow band luminescence due to transitions between the deep-level state and the valence band Zhang et al. [32] observed a dramatically reduced deep-level luminescence in a highly Ge doped (n ~ 10 20 cm3 ) GaN film and suggested that the deep

PAGE 14

8 level luminescence emission may be related to Ga vacancies since Ge acts as a donor by occupying Ga sites. Glaser et al .. [33] found an energy level for the deep-donor state from optically detected magnetic resonance (ODMR) spectroscopy studies to be~ 1 eV below the conduction-band minimum in GaN epitaxial layers grown by OMCVD thus agreeing with the group D energy level theory Other deep traps (group E of Fig 1 2) which are observed as a band-tail absorption feature at ~ 1 e V above the valence band maximum are believed to be related to GaN [21] or V Ga [23]. Perlin et al. [22] investigated the pressure dependence of the energy of the yellow band luminescence for n-type GaN films at hydrostatic pressures up to 30 GPa and showed that the energy of the luminescence band changes linearly with pressure up to about 18 GPa Thus at the highest pressure the yellow band luminescence becomes blue Similarly Shan et al [34] studied the PL emission as a function of applied hydrostatic pressure using the diamond-anvil-cell technique on a 4 2m thick single-crystal GaN film grown by MOC VD. The exciton emission lines were found to shift almost linearly toward higher energy with increasing pressure and the yellow emission band showed a similar blue shift behavior under applied pressure Along with the conclusions of Ogino and Aoki [31] and Ramsteiner et al. [ 15] the report of Perlin et al. supports group E as a candidate for yellow band luminescence due to transitions between shallow donors and deep acceptors However Shan et al. [34] attributed the yellow band emission to transitions involving s-like levels associated with group D defects and the top of the valence band maximum With a view to determining the distribution of the defects responsible for the 2 3 eV yellow band luminescence AsifKhan et al [35] performed low temperature PL analysis on a 9 3-m-thick LPMOCVD grown GaN/sapphire sample from the front and back sides ; the same yellow band emission intensity was found in both cases Thus these authors concluded that the defects that were responsible for the yellow band emission had a uniform distribution in the film However Cowan et al. [ 14] used

PAGE 15

9 Cathodoluminescence (CL) to provide an analysis of traps as a function of depth in their GaN films For low accelerating voltage (Ya cc ) the electron probe had little penetration so data were generated from the film surface At higher Ya cc, the electron beam penetrated the film-substrate interface giving a profile of yellow emission versus depth They observed that the intensity of the deep-level emission was higher at depths close to the film-substrate interface than at depths just beneath the film surface Shan et al. [3 6] reported the results of time-resolved studies on the radiative decay on a 7 thick MOCVD-grown GaN film in which they found that the stronger the yellow emission the shorter the decay times for free-excitons and for bound-excitons. These authors suggested that the fast decay behavior of the PL intensity indicated that the capture of excitons and trapping of carriers at deep level defects through nonradiative combinations dominate the decay of exciton population On the other hand Gotz et al. [37] and Hacke et al. [38] used a transient capacitance method to analyze traps occurring in unintentionally doped n-type GaN By virtue of the sensitive measuring technique (deep-level transient spectroscopy [DLTS] ) Gotz et al. detected two deep-levels having thermal activation energies for electron emission to the conduction band of0 16 eY and 0.44 eY However Hacke et al. found three majority-carrier traps occurring at different energies below the conduction band having activation energies 0 26 0 58 and 0 66 eY respectively. The detected activation energy levels mentioned above belong to group B and group A of Fig 1-2 Although DLTS is a sensitive spectroscopic tool for the characterization of electronic levels in the bandgap of semiconductors conventional DL TS is of limited use in wide bandgap materials as the accessible range of level energies in the gap is restricted to about I eY Optical deep-level transient spectroscopy (O-DLTS) also called photoemission capacitance transient spectroscopy can overcome this limitation by using monochromatic light for carrier emission so that levels in the entire bandgap become accessible for characterization Using O-DLTS Gotz et al. detected four new deep levels with

PAGE 16

10 optical threshold energies for electron photoemission of0 87 eV 0 97 eV 1.25 eV and 1.45 eV which are relatively fitted to group D of Fig 1.2 Very recently Sanchez et al. [39] investigated yellow band and other deep level emissions in undoped MOVPE GaN samples using photocapacitance and photoinduced current transient spectroscopy (PICTS) Photocapacitance reveals in all samples two specific signatures at photon energies of 1 eV and 2 5 eV below the conduction band and the capacitance step at 1 eV is only seen after photoionization at an energy above 2 5 eV In addition Sanchez et al. observed the correlation between PICTS amplitude and the PL intensity of the yellow emission and they concluded that both transitions are linked to the same trap which is at 2 5 eV below the conduction band It is assigned to group E as shown in Fig 1 2 Lin e t al. [ 40] investigated the intensity of low temperature PL as a function of annealing temperature in GaN films grown by MBE These authors found a reduction of the donor-bound exciton (h) peak intensity and an enhancement of the yellow-band emission at high annealing temperatures (900 C) These authors annealed their samples for 30 minutes in a nitrogen ambient and attributed their observations to the formation of intrinsic point defects such as Ga vacancies and antisite defects Migration Enhanced Epitaxy With Reference to GaAs Migration enhanced epitaxy (MEE) is a modified version of conventional MBE in which the constituent element fluxes are supplied separately rather than simultaneously to the substrate [28 40 41 42 43]. The basic advantage of MEE over MBE is that rapid migration of adatoms on the growing surface is ensured even at relatively-low growth temperatures adatom migration being very important with regard to the growth of high quality epitaxial layers During conventional MBE growth of GaAs and AIAs for instance the group III atoms on As stabilized surfaces migrate very slowly especially at low growth temperatures due to their

PAGE 17

11 reaction with As atoms Utilizing reflection high-energy electron diffraction (RHEED) analysis, Neave et al. [ 44] determined the activation energy of Ga surface diffusion to be Eo = 1 3 eV on an As-stabilized (2x4) surface For the diffusion of Al during the growth of AlAs a higher E 0 value of 1 6 e V was estimated from the ratio of the cohesive energies of AlAs and GaAs by Joyce et al. [ 45]. These large activation energies apparently caused by the formation of GaAs and AlAs molecules control the growth process in conventional MBE. It is believed that such large activation energies may make the surface group-III adatoms almost immobile when the growth temperature is lower than 400 C and this is the reason why high-quality GaAs and AlAs films have not been grown at temperature lower than 400 C using conventional MBE. Nagata and Tanaka [ 46] used a shadowing method to determine the Ga surface diffusion distance on (00 I) oriented GaAs under both Gaand As-stabilized surface conditions These authors concluded that at temperatures less than 550 C the diffusion length of Ga under Ga-stabilized surface conditions was 1900A, which is about IO times larger than the diffusion length under As stabilized conditions In conventional MBE the growth of GaAs and AlAs films is usually under As over pressure conditions which means that the surface is As-stabilized For this situation migrating species on the growing surface may move as Ga-As and Al As molecules rather than isolated Ga and Al atoms since impinging atoms immediately react with arsenic to form Ga-As and Al-As molecules The large activation energy of surface diffusion as reported in ref 43 makes these molecules migrate extremely slowly On the other hand when Ga and Al atoms arrive on an As-free growing surface such as in the case of MEE growth migration takes place in atomic rather than molecular form Therefore surface migration is effectively enhanced under Gaor Al-stabilized conditions

PAGE 18

12 Migration Kinetics During MEE with Reference to GaAs To illustrate how the migration of surface adatoms affects the quality of epitaxial layers the kinetic energy of migrant atoms versus temperature is presented in Fig 1 3 (29]. As shown in Fig 1 3 Em represents the minimum energy necessary for surface adatoms to move to the neighboring site Ek and Es are the potential energies at kinks and steps respectively and E 0 denotes the kinetic energy of the migrant atom Because Em>> Eo = kT (where k is the Boltzmann constant and Tis the absolute temperature) the migration of surface adatoms during conventional MBE is usually very slow The large value of Em is caused by the fact that the growing surface has a high density of small islands of GaAs and Em includes the decomposition energy of GaAs molecules at the outermost region of the islands In contrast excess Ga adatoms are expected to migrate actively during MEE growth in the absence of As thus Em in the MEE case is not much greater than E 0 even at relatively low temperatures The Ga atoms diffuse until they find more stable lattice sites at kinks and ledges on the growing surface where Ga-As bonds are formed Thus the energy relationships in the MEE case can be expressed as E 0 Em <
PAGE 19

1 3 Impinging Ga atoms will form metallic bonds with underlying Ga atoms In metallic bonds the electron distribution is wide which makes the migration of Ga atoms bonded to Ga atoms very easy This characteristic holds even below the melting point of me t als Ga has a particularly low melting temperature ; therefore Ga atoms on a Ga-atomic plane move freely even at low temperature until they reach active sites such as locally e x posed As-islands and other ledges and kinks This characteristic is important for the MEE process at low temperatures because it is expected that Ga atoms supplied on ei t her As or Ga-atomic planes tend to form a two-dimensional atomic sheet rather than a three dimensional cluster. The third case is the migration of As molecules on a Ga-atomic plane It is well known that the Ga-atomic plane surface is very active chemically because of the unsatisfied sp 2 bonds perpendicular to the surface Thus the impinging As molecules at these sites will decompose into As atoms which are incorporated into the epitaxial layer to form stable sp 3 bonds with underlying Ga atoms and neighboring As atoms [29 47]. Therefore migration of As atoms on a Ga-atomic plane could be slow ; on the other hand isolated group V atoms on a Ga-atomic plane cannot form sp 3 bonds and in this case the y may migrate easily. The fourth possible case of diffusion concerns As molecules on an As-atomic plane An As surface is less chemically active than a Ga surface because the As atoms of an As-atomic plane stabilize themselves by forming p 3 or sp 3 bonds with each pair of adjacent As atoms [29 47]. The arriving As molecules are weakly adsorbed on the As plane and migrate actively along the surface As a matter of fact Foxon and Joyce (48] reported a surface diffusion activation energy of As 4 molecules on As-planes to be 0 25 eV However the growth of more ionic bonded materials such as Group III nitrides may be different from that of mostly covalent bonded materials such as Si and GaAs because of the larger electronegativity difference between the components of the IIIV

PAGE 20

14 nitrides [ 49 50]. Lester et al. [51] suggested that Fermi level pinning at the crystal surface during epitaxial growth may not occur in the case of n-type GaN In addition almost all as-grown unintentionally-doped GaN films are n-type [52]. Any attempt to increase the V /III ratio results in a rough growth front. As a matter of fact unlike GaAs in which the growth rate is controlled by the impinging Ga flux alone GaN growth is controlled by both the Ga flux and the nitrogen flux [53]. MOCVD Growth ofGaN Using An Alternate Source Exposure Method. There have been three reported attempts to grow GaN films by MOCVD using alternate source exposure Hwang et al. [54] alternately exposed a sapphire substrate to trimethylgallium (TMG) and N+H plasmas via a rotating suscepter but most of the resulting GaN films were polycrystalline Sumakeris et al. [55] alternately exposed an a SiC substrate to triethylgallium (TEG) and ammonia The film thickness was measured by ellipsometry to study the effect of the exposure time to the reactant species on the amount of material deposited per cycle. A self-limited growth was found to occur at a substrate temperature of 120 C and was limited to about 67 % of a monolayer. The films grown at this temperature were amorphous The crystallinity improved with increasing temperature but the chemically self-limiting growth characteristic disappeared Khan et al.. [56] used switched MOCVD to supply TEG and NH 3 to a sapphire substrate at 4 5 0 900 C at a pressure of 7 6 Torr. The growth rate was about O 7 5 monolayer per cycle with a 10% margin of error. The band-edge PL signal intensity and linewidth for the alternate source exposure-grown sample grown at 900 C were comparable to those observed from samples grown by conventional MOCVD at 1040 C.

PAGE 21

15 Dissertation Outline This dissertation is subdivided into six chapters. Chapter I includes the motivation and objectives for this work and a review of the deep-level defects which result in yellow band luminescence in GaN Chapter 2 details the growth techniques employed in this work and the in-situ analytical techniques used to characterize the growth are also detailed Chapter 3 describes the substrate preparation and the conventional MBE growth procedures used to synthesize GaN films on c-plane sapphire substrates, and reports PL Hall-effect and SEM analyses of these films Chapter 4 presents the novel AEEE growth method used in this investigation and growth parameters are correlated with the observed RHEED specular reflection beam intensity variations In this chapter, PL and SEM data are reported as functions of the Ga and N delay times Chapter 5 presents a discussion of the GaN growth kinetics and their correlation with deep level defects A model is presented to explain the experimental observations Finally, Chapter 6 summarizes the principal conclusions drawn from this investigation and presents suggestions for future work in this area

PAGE 22

10 >
PAGE 23

0 0.5 1.0 > 1.5 Cl) 2.5 3.0 3.4 17 ---------------------A B C D E conduction band edge ~--------------------valence band-edge Figure 1.2 Native defect levels in as-grown GaN summarized from the literature. A,B,C (VN), D (NGa V Ga), and E (GaN ,VaJ

PAGE 24

18 Figure 1-3 Potential energy diagram for surface migration Em represents the minimum energy necessary for surface adatoms to move to the steps and kinks and E s are the potential energies at kinks and steps respectively E 0 is the kinetic energy of migrant atom

PAGE 25

CHAPTER2 :MBE SYSTEM AND CHARACTERIZATION TECHNIQUES Introduction All of the samples in this work were grown in a Varian GEN II :MBE system This system was previously employed to grow zinc blende-GaN films [59 60] All samples were characterized in-situ using RHEED analysis as well as e x-vacuo using other analytical techniques including scanning electron microscopy (SEM) Hall-effect measurements and photoluminescence (PL) measurements All samples were grown and characterized by the author with the exception of the PL analysis which was carried out by Dr. Matthias Ludwig in Dr. Hummel s research laboratory at the University of Florida. The :MBE growth conditions employed as well as important considerations and procedures pertaining to the growth and the characterization techniques mentioned above are outlined in the following chapter. Molecular Beam Epitaxy Molecular beam epitaxy has become a powerful growth technique particularly suitable for manufacturing advanced electronic and photonic devices In a :MBE system constituent elements and dopants are supplied in the form of molecular or atomic beams either thermally generated or formed in a plasma The beams are directed toward a suitably heated single-crystal substrate on which epitaxial growth occurs Since the molecular or atomic beams can be individually shuttered complex superlattice structures can be fabricated using this technique with precise control down to the atomic-level. 19

PAGE 26

20 Varian GEN II MBE System Configuration The Varian GEN II MBE system at the University of Florida consists of three connected ultra-high vacuum (UHV) chambers ; an entry/exit (EE) chamber a buffer chamber and the growth chamber. Each UHV chamber is of stainless steel construction and has an independent pumping system The chambers are isolated by gate valves and all system components are bakeable to 200 C The entry/exit chamber allows samples to be loaded into and out of the system while maintaining the buffer and growth chambers under ultrahigh vacuum conditions. To reduce contamination the EE chamber opens into a clean room where the ex-vacuo cleaning of substrates take places while the opened chamber is vented using dry nitrogen gas Two molecular-sieve and liquid nitrogen cooled sorption pumps are used to evacuate the chamber in addition to a CTI-100 cryogenic pump which produces ultrahigh vacuum conditions After entry into the EE chamber the substrates are heated to 150 C using a pair of quartz halogen lamps in order to desorb water vapor from the substrate surface and the sample holder. A transfer trolley facilitates sample movement between chambers A magnetic force is employed to transfer the trolley from the EE chamber to the buffer chamber once the valve separating the two chambers is opened The buffer chamber is evacuated by a Vaclon pump A heater station inside the buffer chamber is employed to evaporate any impurities whose vapor pressures are too low at 150 C (the EE bake-out temperature) The temperature used for buffer bakeout is 450 C. There are two magnetically coupled transfer arms one for transferring a substrate to the buffer chamber heater station the other for the substrate mounting block which is attached to the continuous azimuth rotation (CAR) substrate holder inside the growth chamber. The pressure difference between the buffer and the growth chamber is minimized before the gate valve between the chambers is opened.

PAGE 27

21 The growth chamber is illustrated schematically in Fig 2 1 Different types of pumps a CTI-200 cryogenic pump, a Vaclon pump and a titanium sublimation pump are employed to maintain the UHV environment required for growth When attached to the CAR, the substrate can be rotated (I 80) from the transfer position to the growth position and vice-versa and rotated (360) in its own plane for RHEED analysis Heater filaments for elevating the substrate temperature are mounted on a pyrolitic boron nitride (PBN) plate inside the substrate mounting block. AW-Re thermocouple is isolated from the heater filaments by a PBN tube in a radiation cavity which is located in the center of the substrate mounting block. The water-cooled substrate mounting block can heat the substrate to 800 C by radiation through the PBN plate which gives uniform heat distribution for the entire substrate holder while the thermocouple serves to monitor and control the substrate temperature In addition the rotation speed of the CAR substrate holder can be controlled which further improves the substrate temperature uniformity during epitaxial growth Conventional effusion cells are used to supply the atomic and molecular beams with the exception of the nitrogen source There are eight furnace ports to house different source materials For the conventional effusion cells each solid or liquid source material is placed into an ultraclean PBN crucible The liquid sources are mounted on upward-facing ports while the solid sources are mounted on downward-facing ports A W-Re thermocouple is held by spring force tightly against the bottom of each crucible To maintain the high purity of the source materials the crucibles are idled at relatively high temperature in the range 200 C to 820 C depending on the source Cryoshrouds surround the effusion cells and the growth chamber itself to dissipate excess heat. A beam flux monitor (nude ion gauge) is utilized to measure the beam flux from the effusion cells The beam flux monitor is mounted on the CAR 180 away from the substrate position in order that beam fluxes can be monitored without exposing the substrate to the fluxes Control over the beam fluxes is achieved using pneumatically controlled shutters

PAGE 28

22 in front of each source orifice For practical purposes the beams can be considered as unidirectional so that the insertion of a mechanical shutter will stop a beam from reaching the substrate and allow different crystal compositions to be grown The precise control of the opening and closing of the shutters plays an important role in this research Substrate Temperature Calibration The actual temperature at the substrate surface is different from that monitored by the W-Re thermocouple since the thermocouple is not physically attached to the substrate In this work Omega Temperature Indicating Liquids (OTIL) which have a time response to temperature changes on the order of milliseconds and an accuracy of 1 % were used to calibrate the surface temperature By applying a small amount of OTIL on a test sapphire surface which was mounted on a Mo block it was easy to pinpoint at which monitored temperature the OTIL turned black an irreversible change indicating that the substrate had reached a particular temperature. In this fashion a calibration was obtained and a plot of actual sapphire substrate surface temperatures versus monitored temperatures is presented in Fig 2 2 In Situ Analysis Reflection High Energy Electron Diffraction (RHEED) was utilized to perform real-time in-situ analysis of the substrate and epilayer surfaces The RHEED electron beam energy was in the range of O 10 ke V and the electron beam was incident at a glancing angle of 1-2 to the substrate surface. Since monochromatic electron beams are sensitive to atomic order in solids and are ideally suited for the study of crystalline surfaces this in-situ surface analyzer was invaluable in providing information about the surface structure orientation and the degree of surface roughness during epitaxial

PAGE 29

23 growth RHEED is particularly useful in studying structure changes as functions of temperature or time during epitaxy Ex-vacuo RHEED Specular Reflection Beam Intensity Monitoring System In addition to observing diffraction patterns the RHEED specular reflection beam intensity was monitored in this work and the monitoring system employed during growth consisted of a CCD camera video monitor Si-photodiode bias circuit and chart recorder as shown schematically in Fig 2 3 The RHEED specular beam intensity was captured by the CCD camera and displayed on the monitor and the Si-photodiode placed over the spot on the monitor. The bias circuit placed a potential across the photodiode causing output (voltage) to be linearly related to the input (light intensity) The chart recorder logged the output voltage of the photodiode By monitoring the variation of the RHEED specular reflection beam intensity during growth the migration process and surface conditions on the growing surface could be evaluated Because the specular reflection beam intensity is sensitive to the degree of surface roughness, a decrease of the RHEED specular reflection beam intensity indicates an increased surface roughness such as might be associated with island formation for instance To illustrate this point a typical RHEED specular reflection beam intensity trace recorded in this work during alternate element (Ga and N) exposure growth is presented in Fig 2.4 As can be seen from the figure the specular reflection beam intensity is a strong function of the status of the surface in terms of species present and the particular kinetic processes taking place The interpretation of such data will be discussed at length in the following chapters

PAGE 30

24 Nitrogen Free-Radical Source In this work an Oxford Applied Research (OAR) radio frequency (rf) plasma discharge source as shown schematically in Fig 2 5 was used to generate nitrogen free radicals The same type of source has been successfully used to supply atomic nitrogen for p-type doping of ZnSe [57 58] and to grow zinc blende-GaN [59 60]. The source operates at 13 56 J\1Hz and has a maximum output power of 500 W. The radical source is mounted on the l\IBE growth chamber replacing a conventional effusion cell A distance of 15 cm separates the end of the radical source from the substrate High-purity molecular nitrogen gas (99.9995%) was introduced through a precision UHV leak valve into a PBN discharge tube capped with a PBN aperture The PBN aperture had 37 0.3 mm diameter holes to release reactive nitrogen species to the substrate surface The gas line between the nitrogen cylinder and the UHV leak valve was pumped by two liquid nitrogen cooled sorption pumps before opening the valve on the cylinder to ensure the purity of gas that was introduced into the growth chamber. In this work a nitrogen plasma was maintained using a forward power of 200W and a nitrogen chamber pressure of 1 x 10 -s Torr. Under these conditions an optical emission spectrum was recorded from the plasma using a wavelength spectrometer/OMA apparatus The emission spectrum shown in Fig 2.6 is indicative of the presence of both ground state nitrogen atoms (1 s t Positive System of N 2 transitions) and excited nitrogen atoms ( direct transitions) in the plasma and therefore in the flux emanating from the source.

PAGE 31

High pure N2 "' Liquid Nitrogen Cooled Shrouds rf plasma discharge chamber 25 Electron Gun (RHEED) I I I I I I Ion Gauge UIN substrate & ubsbheater I I I I ; 'Ion : Gauge I I I I I I I I : [ ':'::;:::::::::::::;::::: : :::::;:::::: : :: j I y Phosphor Screen Residual gas Analyzer Figure 2.1 Schematic of the MBE growth chamber used to grow GaN/sapphire films by conventional growth and by ABEE techniques.

PAGE 32

700 ,-.... u 0 -.._,,, Q) 650 Q) 8 600 Q) -+-Q) u 550 :::::s r/l Q) b 500 r/l ..0 :::::s r/l Q) 450 ..d 0.. 0.. C/) 400 LV 700 750 800 850 900 950 1000 Monitored temperature C) Figure 2 2 Plot of sapphire substrate surface temperature vs monitored temperature obtained using Omega Temperature Indicating Liquids (OTIL). 1050

PAGE 33

Bias circu ~ Photodiode Video monitor Typical RHEED pattern Chart recorder 0 0 27 Mask Typical RHEED pattern on the monitor The bright spot of the RHEED specular reflection beam (intensity of which was monitored as a function of time) CCD camera RHEED ....,__..,.__ __ .,...... screen AIR MBEgrowth chamber UHV Figure 2 3 Experimental set-up used to monitor the RHEED specular reflection beam intensity as a function of time during growth

PAGE 34

28 N delay time I I Iii'! I I N shutter op~n J l __.,_ ____._: 4 ___ G_a_d_el-ay_t_im_e ____ N shutter closed I Ga shutter I I ,open I Ga shutter closed Time (sec.), ________ _____. Figure 2.4 A typical specular reflection beam intensity trace recorded during one cycle of alternate element (Ga+ N) exposure growth

PAGE 35

Optical Detector \ Wmdow Gas Inlet l/RF Matching Unit Water Feed 13.56 MHz 500 Watts RF Generator \RF Shield Conflat Flange PBN Discharge Tube Gas Feed Water-Cooled RF Coil PBNBeamExi aperture disk Figure 2 5 Schematic of rf plasma discharge source used to generate nitrogen free radicals (N) N I.C

PAGE 36

30 1st Positive System of N 2 (Recombination of (4s) N atoms) I< >I 500 550 600 650 700 Atomic Nitrogen (Direct relaxation) 750 800 Wavelength (mn) Figure 2 6 Optical emission spectrum of nitrogen plasma generated 850 from a 190 W 13 56 MHz rf plasma discharge with a nitrogen background pressure of 2 x 105 Torr

PAGE 37

CHAPTER3 GROWTH BY CONVENTIONAL !vffiE AND CHARACTERIZATION OF GaN / SAPPHIRE FILMS Ex-vacuo Substrate Preparation All of the c-plane sapphire (A1 2 0 3 ) substrates used in this work were obtained from Union Carbide Crystal Products in Washougal WA. The specifications of these substrates are listed in Table 3 .1. Table 3 1 Specifications of Union Carbide c-plane sapphire substrates Wafer Size 10 xll mm Wafer Orientation Surface normal to the (0001) or c-axis to within 3 Edges longer edges normal to ( 1 2 1 0) or a-axis Wafer Thickness 500 microns Surface Finish one side polished X-ray Rocking Curve linewidth ~ 12 arc seconds Lattice Parameter a-axts a=4 758 A c-axts c=l2.991A Thermal Expansion Coefficients a-axts 7 5 X 106 /K c-axts 8 5 X 106 /K 31

PAGE 38

32 Ex-vacuo substrate preparation procedures were performed in the clean room The Ah03 (0001) substrates were degreased by successive rinses in trichloroethane acetone and methanol and then etched in H 2 S0 4 : H 3 P0 4 (3 : 1) at 160 C for 15 minutes The etching was used to eliminate any surface contaminants and mechanical damage due to polishing Atomic force microscopy (AFM) analysis showed that the surface morphology of the etched sapphire substrate was smoother than that of the un-etched substrate as shown in Fig 3 1 (rms roughness= 0 398 nm compared to as-received sapphire rms roughness= 2.458 nm) The wafers were then rinsed in de-ionized water and blown dry using dry nitrogen The backside of the substrates were metallized with Mo using electron-beam (E-beam) evaporation In the vacuum(< 5 xl07 Torr) environment of the E-beam evaporation system a high-intensity beam of electrons was focused on a high-purity Mo source target. With the shutter closed the source target was outgassed for 5 minutes The shutter was then opened and Mo was deposited at a rate of ~3 A/sec as measured by a quartz crystal monitor. The resonant frequency of the crystal shifted in proportion to the thickness of the deposited film By monitoring the shift in resonant frequency of the crystal the deposition rate could be measured with an accuracy of better than 1 A/sec The total thickness of the Mo metallization was 0 25 m. The metallization on the backside of substrate greatly improved the thermal uniformity across the substrate during growth Just before loading into the EE chamber the metallized substrates were re-rinsed in methanol de-ionized water and strain-free mounted on a hollow molybdenum block which was loaded on a trolley A stream of dry nitrogen was used to remove particles from the substrate surfaces and the trolley The trolley was then introduced into the EE chamber which was then pumped down

PAGE 39

33 In-situ Substrate Preparation All substrates were preheated at 150 C for one hour in the EE chamber and at 450 C for 20 minutes in the buffer chamber to desorb impurities such as water vapor on the substrates and substrate holders Before carrying out the preheating procedures in both the EE and the buffer chambers it was important to make sure that the background pressure in both chambers was lower than 108 Torr After transfer to the growth chamber the substrate was heated slowly (~30 C/min ) to 750 C ; this temperature was held for 60 minutes for thermal cleaning All substrate temperatures quoted in this dissertation are the actual sapphire surface temperatures calibrated using the OTIL method described in the previous chapter. The RHEED system was used to monitor the surface morphology and crystallinity through the thermal cleaning and nitridation processes The thermal cleaning resulted in a ( 1 x 1) RHEED pattern with Kikuchi lines indicating a clean Al 2 0 3 (0001) surface A 9 keV electron beam was oriented along (a) [l To O] (=0) and (b) [l 2 1 O] (=30) as shown in Fig 3 2 (a b) Then the thermally cleaned substrate was exposed to the nitrogen free-radical beam (at a forward power of 250 W) while at the same temperature (750 C) for 20 minutes to form a thin AIN layer whose RHEED pattern is shown in Fig 3 3 Conventional MBE Growth of GaN on Sapphire With both Ga and N shutters closed the substrate temperature was lowered to 500 Cat which temperature a thin GaN buffer layer was deposited A Ga effusion cell temperature of 780 C was used to provide a low Ga flux (J= 7 6 x 10 13 cm 2 s 1 ) in order to grow the buffer layer Growth was initiated by simultaneously opening both Ga and N shutters for 10 minutes to grow a ~ 300 A thick GaN buffer layer An elongated spotty RHEED pattern was observed during the growth of this layer. The basal plan e o f the GaN overlayer was rotated 30 with respect to that of the s ubstrate durin g g ro wt h ; th e

PAGE 40

34 orientation relationship between the single crystal GaN film and the substrate was (0001] GaN // [0001] AhO3 [1100] GaN //[1210] AhO3 and [2110] GaN // [ 1100] AhO 3 Immediately following buffer-layer deposition the substrate temperature was ramped to 750 700 640 C or 600 C in 5 minutes during which time both source shutters were kept closed A streaky RHEED pattern developed during the ramp in each case Also during the ramp the Ga cell temperature was increased to 820 C to provide a flux of J = 2 3 x 10 14 cm 2 1 When the appropriate substrate temperature was reached both the Ga and N shutters were opened simultaneously producing a growth rate of 0 10 2 m/hr depending on the growth temperature The growth parameters employed are summarized in Table 3-2. The surface morphology during growth was very sensitive to the Ga/N beam flux ratio as well as the growth temperature For a fixed N flux (1 x 10 5 Torr) the RHEED pattern changed from streaky to spotty at 750 C when the Ga beam equivalent pressure (BEP) was lower than 5 x 10 8 Torr ( J=l.8 x 10 14 cm 2 s" 1 ) However the same Ga flux resulted in a milky surface at 600 C. Hence a lower Ga BEP of 2 x 108 Torr (J=7 6 x 10 1 3 cm2 s" 1 ) was used for conventional growth at 600 C and 6 x 10 8 Torr (J=2 3 x 10 1 4 cm 2 1 ) was used for growth temperatures of 640 C 700 C and 750 C A streaky (lxl) RHEED pattern as shown in Fig 3 4 was maintained throughout growth for all conventionally-grown samples In order to provide a uniform thermal distribution the CAR was rotated at a speed of 1 rpm for all conventional growths The total growth time in each case was 4 hours and about 0 8 m of GaN was deposited For the sample grown at 600 C a 0.4 m thick film was deposited since Ga droplets prevented it from growing thicker as suggested in references 28 and 29 In each case growth was terminated by simultaneously closing both shutters and cooling the sample to room temperature in 20 minutes in a nitrogen environment. The thermal expansion coefficient of sapphire is about 26-34% [ 34% in the case of c 1(GaN) // c 11 (sapphire) and 26 % in the case of c 1( GaN) // c 1(sapphire)] larger than that of GaN. Therefore

PAGE 41

35 the epitaxial GaN film suffered compressive stress in the plane parallel to the c-face when the sample was cooled, hence a slow cooling rate was necessary Table 3 2 Growth parameters used for both conventionally-grown and AEEE-gro w n films. Nitrogen source parameters RF forward power 200W Nitrogen beam flux 1 x 10" 5 Torr Plasma emission intensity 80 ~ 100 mV Ga source condition Ga beam flux for epilayer growth 6 0 x 108 Torr (J=2 3 x 10 14 cm2 f 1 ) Ga beam flux for buffer layer growth 2 x 108 Torr (J=7 6 x 10 13 cm2 f 1 ) Growth parameters Buffer layer Growth temperature 500 C Buffer layer growth time 10 min Conventionally-grown epilayer Growth temperature 600 C 640 700 750 C. Growth rate 0 1 0.2 m/hr Film thickness 0.4 and 0 8 m Characterization of Conventionally-Grown GaN Films SEM Analysis The surface morphology of MBE grown samples was examined by SEM The surface morphology of the sample grown at 750 C showed a typical micrometer range rough surface as shown in Fig 3. 5 The surface morphology of the sample grown at

PAGE 42

36 600 C was smoother in most areas however ~ 15 m diameter Ga droplets were dist r ibuted over the surface with a density around 6 x 10 4 cm2 as shown in Fig 3 6 The sample grown at 640 C had a similar surface morphology however the Ga droplet size was smaller ( < 5 m ) in this case. In order to observe what was under the Ga droplets samples were etched in (1:10 : 100) HF : NHO 3 : H 2 O before re-performing SEM It was found that crack-like voids were observable beneath the Ga droplets as shown in Fig 3 7 Hall-Effect Measurement and Photoluminescence Analysis The Van der Pauw method was employed to measure carrier concentrations and mobilities at room temperature using Hall-effect analysis Small pieces of indium were pressed in place at the four comers of square-shaped samples to make ohmic contact (the 1-V characteristic was linear) to the GaN All of the unintentionally doped GaN films showed 10 18 cm 3 n-type carrier concentrations except the sample grown at 600 C which had a 3 x 10 1 9 cm3 carrier concentration The carrier mobility was 3 5 cm 2 v1 f 1 for the sample grown at 640 C and was in the single digit range for those samples grown at 600 C and 700 C In the PL analysis the samples were excited by the 3 25 run wavelength output of a 25 mW HeCd laser The power density at the sample was about 30 mwrrun2 The luminescence signal was dispersed by a 0 3 m single-grating monochromator and detected by a photomultiplier tube Room-temperature and low-temperature PL spectra were recorded from the samples grown at 700 640 and 600 C and these spectra are presented in Fig 3 8-3 13 As can be seen from these figures the yellow band emission was observed in all of the spectra in addition to the near-band-edge emission

PAGE 43

37 A detailed discussion on luminescence peak assignments (both near-band-edge and deep level emission) is presented in Chapter 5 based on the results detailed in this chapter and also Chapter 4

PAGE 44

r \ \ \ \ \ l ,. 38 / / .. ... ... \ ,. /" :,.. ; _\ / ~ / ./ / < (b) rms roughness= 0.398 run / / I \ \ \ \ \ / . .. Figure 3 1 Atomic force microscopy tapping mode images showing the morphology of c-plane sapphire (a) as-received rms roughness= 2.458 nm (b) etched in SO 4 : H 3 PO 4 at 160 C for 15 minutes rms roughness= 0 398 nm

PAGE 45

39 (b) (a) Figure 3 2 RHEED patterns recorded from thermally cleaned sapphire (0001) surface (a) ttlectron beam oriented along [lT00] ( = 0 (b) [1210] (= 30

PAGE 46

40 (b) (a) Figure 3 3 RHEED pattern recorded from a thin AIN layer formed during nitridation aL750 C after _20 minutes exposure to the_N-flux (a) [211 OJ AIN II [1100] sapphire (b) [1100] II [1110] sapphire

PAGE 47

41 ---------"' ""''" .. . . .,.,. ---m .. ---(b) (a) Figure 3 .4 RHEED patterns recorded from a conventionally gr.9wn GaN film, (a) [~110] GaN // [JTT0]AIN // [1100] sapphire (b) [ll00] GaN // [1100] AlN // [1210] sapphire

PAGE 48

42 0 0 0 0 0 0 0 0 : : : : : : : : : : ::: ::~ -:::: ::: : ::: : : :::-: : 0 0 0 0 ,-. 0 0 ~0 0 OH 0 0 0 0 I 00 0, 00 0 : O 0 0 0 0 0 0 0 ::::::;:::::::::::::::::: : : :: ::::::::::::::::::::::::::::: ~ :::: : ;: :. ::: : Figure 3 5 SEM micrograph showing the sutface morphology of a GaN film grown by conventional MBE at 750 C

PAGE 49

43 (a) := -.. .. . . ... . -.. .. . .. .. . .. (b) Figure 3 6 SEM micrographs of the surfaces of GaN films grown by MBE at a substrate temperature of 600 C (a) and 640 C (b ) Note the presence of Ga droplets

PAGE 50

44 Figure 3 7 SEM micrograph showing the GaN morphology beneath a Ga droplet after removal of the Ga droplet by wet etching The GaN was grown at 600 C

PAGE 51

,___ Cl') 0.. u ___,, 0 Cl') i::::: Q) 45 5000 366 3nm T = 26 IK 4500 4000 3500 3000 2500 2000 1500 1000 500 0 350 400 450 500 550 600 650 700 Wavelength (run) Figure 3 8 Low-emperature PL spectrum recorded from a GaN/sapphire film grown at 700 C by conventional MBE.

PAGE 52

,-..._ rfl 0.. u __., >--. rfl (I) 46 4000 ---,-------------------3500 362 nm 3000 2500 2000 "' 1500 ghost peaks 1000 I 500 o--~--~-~--~----~-~ 350 400 450 500 550 600 650 Wavelength (run) Figure 3 9 Room-temperature PL spectrum recorded from a GaN/sapphire film grown at 700 C by conventional MBE. 700

PAGE 53

,,-._ (I) 0.. C) '._/ >-. -+-(I) t:: (1) -+-t:: 47 22K 200000 357 150000 100000 50000 0 LL._~J_.J.......J_..J.......J....l..J......J......L....J..d::::::t:::I::::G:::c::;::~::i::1:::c::i::::r::::r=:::b=a:1:=~ 350 400 450 500 550 600 650 700 Wavelength (nm) Figure 3 10 Low-temperature PL spectrum recorded from a GaN / sapphire film grown at 640 C by conventional MBE

PAGE 54

,,-.__ cri 0.. Q '-._/ ;;;..... cri s::::: (l) s::::: 48 200000 150000 364nm 100000 50000 0 ~...L...J.---1::::i::::t:=:I=:J:::::::CI...l.......J...._L....L...l......J.___,L._L...L..l......J----L......L....L....l..._L....L....L....1._L..L...L....J...._L.J 350 400 450 500 550 600 650 700 Wavelength (nm) Figure 3 11 Room-temperature PL spectrum recorded from a GaN/ sapphire film grown at 640 C by conventional MBE

PAGE 55

,,.-.., (/} 0.. <.) ..._,, >-. +,j (/} Q) +,j 49 24 5K 200000 150000 100000 50000 359nm 0 I.L.L.....J.......J....::t::i::::I=:I::::t::::I::::t:=i::::::c:::1.......L_..L._..L......L---'--L.L...J.......J___J_..L.l___L__..L.....L....,L_L.L....L.....;LL..J 350 400 450 500 550 600 650 Wavelength (run) Figure 3 12 Low-temperature PL spectrum recorded from a GaN/sapphire film grown by conventional MBE at 600 C 700

PAGE 56

,,,-., rfJ 0-c u '-_,/ ;;...... rfJ = (l) = 7000 6300 5600 4900 4200 3500 2800 2100 1400 700 0 350 400 450 50 500 550 Wavelength (run) RT second order harmonic Laser peak 600 650 Figure 3 13 Room-temperature PL spectrum recorded from a GaN/sapphire film grown by conventional MBE at 600 C 700

PAGE 57

CHAPTER4 INVESTIGATION OF A NOVEL GROWTH MODE Introduction The PL spectra recorded from all of the conventionally grown GaN/sapphire films showed near band-edge emission at around 365 nm and a yellow band emission centered around 550 nm Decreasing the growth temperature improved the efficiency of the near band-edge emission but the deep-level defect-related yellow band emission was still present. Furthermore decreasing the growth temperature resulted in the formation of Ga droplets on the growing surface and the yellow band emission became dominant in the PL spectra This yellow band emission has been commonly reported in the literature regardless of the growth techniques and the substrates used [2-14] With a view to investigating the factors that lead to deep level generation and consequently to the yellow band emission in the PL spectra recorded from GaN epitaxial films a systematic study was conducted in which the surface morphology was examined by means of the real-time monitoring of the RHEED specular reflection beam intensity during critical phases of growth In particular a new growth mode was studied in which the substrate was separately exposure to Ga atoms and N atoms ( as in migration enhanced epitaxy [J\IBE]) and in addition time delays were introduced following the Ga and N exposures which we term the Ga delay and the N delay respectively The time delays in which a variety of kinetic processes might take place depending upon the element in question distinguish this growth mode from MEE and we have termed the novel growth mode alternate element exposure epitaxy (AEEE) ". This work was carried out as follows ; first a systematic study was conducted to determine the optimum AEEE growth temperature second a 5 1

PAGE 58

52 systematic search was made for an optimum N delay time at the optimum growth temperature and finally a systematic investigation was conducted to find the optimum Ga delay time also at the optimum growth temperature AEEE Growth Procedures The ex-vacuo and in-situ substrate preparations were the same as those used for the preparation of conventionally-grown samples as mentioned in the previous chapter. After nitridation of the sapphire substrate at 750 C the substrate temperature was lowered to 500 C both Ga and N shutters being closed A 300 A thick buffer layer was deposited at this temperature by opening the Ga and N shutters simultaneously for 10 minutes During growth of the buffer the Ga effusion cell temperature was kept at 780 C corresponding to a low Ga flux of J = 7 6 x 10 1 3 cm2 f 1 After the buffer layer was deposited the substrate temperature was increased to 600 C with both shutters being closed During the same substrate temperature ramping period the Ga effusion cell temperature was gradually increased to 820 C corresponding to a Ga flux of J = 2 3 x 10 14 cm2 s1 AEEE Growth Temperature Optimization In order to determine the optimum substrate temperature for AEEE growth the same growth temperatures used in conventional growth were investigated here Each growth cycle included opening the N shutter for 10 sec ., simultaneously closing the N and opening the Ga shutter (Ga exposure time 5 8 sec ., depending on the growth temperature) closing the Ga shutter and allowing the RHEED specular reflection beam intensity to recovery to a maximum before opening the N shutter again (Ga delay time 15 30 sec. depending on growth temperature) The delay time is defined as being the time elapsed between closing one shutter and opening the other shutter as indicated in Fig 2.4 (page

PAGE 59

53 28). The growth conditions for one cycle of the shutter sequence are summarized in Table 4 1 The recovery of the RHEED specular reflection beam intensity during the Ga delay was found to be a function of the growth temperature At higher temperatures the beam intensity recovered faster but to a lower level of intensity as shown in Fig 4 1 Films were not deposited at 750 C by AEEE because the RHEED pattern did not remain streaky and the specular reflection beam intensity decreased quickly after repeated cycles The specular reflection beam intensity was also found to be a function of the Ga exposure time. For instance, an 8 sec. Ga exposure time was optimum to have the RHEED intensity recover to a maximum when the substrate temperature was 700 C whereas 7 sec and 5 sec were the optimum Ga exposure times for substrate temperatures of 640 C and 600 C, respectively. Failure to control the optimum Ga exposure time results in a loss ofRHEED signal intensity following several cycles. Each of the AEEE samples was grown using a total of 2000 cycles. The growth was terminated by the last N exposure and the substrate was cooled to room temperature in 20 minutes under a N2 overpressure in the growth chamber with both Ga and N shutters closed Table 4.1 AEEE growth conditions for one cycle of the shutter sequence at different growth temperatures Growth N shutter N delay time Ga shutter Ga delay Remark temperature open (sec ) (sec.) open (sec ) time (sec ) (sample#) 700 C 10 0 8 15 Maximum (7147) RHEED intensity-low 640 C 10 0 7 20 Maximum (7155) RHEED intensitymoderate 600 C 10 0 5 30 Maximum (7165) RHEED intensityhigh

PAGE 60

54 PL Analysis Low-temperature PL spectra were recorded from the samples grown at 700 C 640 C and 600 C and the spectra are presented in Figs 4 2 4.3 and 4.4, respectively. These spectra all indicate a dominant near band-edge emission but a low intensity yellow band emission is still exhibited in samples grown at 700 C and 640 C. In addition a weak peak related to N vacancies [20 37] appeared at 365 5 nm for the sample grown at 700 C and at 385 nm for the sample grown at 640 C respectively On the other hand as can be seen from Fig 4.4 the yellow band emission was absent from the PL spectrum recorded from the sample grown at 600 C. Nitrogen Delay Time Optimization The basic advantages of the AEEE technique in the context of GaN are the low growth temperature and the provision of a chemically active Ga-covered surface. Both of these factors can greatly enhance the sticking coefficient ofN during GaN epitaxy In this investigation the shutter operating sequence was the same as for the sample grown at 600 C in the previous section except that N delay times namely 0 sec ., 15 sec ., and 30 sec. were inserted between closing the N shutter and opening the Ga shutter as shown from Fig 4.5 As can be seen from Fig 4 5 the RHEED specular reflection beam intensity recovered only moderately during the N delay phase Shutter sequences employed for the N delay experiments are shown in Table 4 2 It is worth pointing out that during the N shutter open time the RHEED specular reflection beam intensity decreased initially and then recovered moderately to a turning point beyond which the intensity again dropped (see Fig 4 5) and the RHEED pattern became spotty The turning point occurred after an exposure time of IO seconds and thus a IO sec N exposure was employed in this work.

PAGE 61

55 Table 4 2 Shutter sequences for the N-delay optimization experiments ( Growth temperature was 600 C) Run N shutter N delay time Ga shutter Ga delay Number open (sec ) (sec ) open (sec ) time (sec.) 7160 10 0 5 30 7167 10 15 5 30 7156 10 30 5 30 PL Analysis Remark Maximum RHEED intensityhigh Maximum RHEED intensitymoderate Maximum RHEED intensitymoderate The low-temperature PL spectra recorded from samples grown using N delays of 0 sec, 15 sec and 30 sec are shown in Figs 4 6 4 7 and 4 8 As can be seen from these spectra, as the N delay time is increased, the near band-edge emission intensity decreases and additional peaks appear in the spectra Consequently in future growth runs a O sec N delay was employed Ga-delay Time Optimization In the Ga-delay optimization set of experiments, the N-delay was fixed at O sec based on our previous experience and the previously determined optimum substrate temperature of 600 C was used throughout. As indicated in Fig 4 9 various Ga-delay times were employed including O sec 15 sec 30 sec and 50 sec As seen in the figure these delay

PAGE 62

56 times corresponded to various RHEED specular reflection beam intensities Shutter sequences employed in the Ga-delay experiments are summarized in Table 4 3 Table 4 3 Shutter sequences for the Ga-delay optimization experiments ( Growth temperature was 600 C) Run N shutter N delay time Ga shutter Ga delay Number open (sec.) (sec ) open (sec ) time (sec ) 7161 10 0 5 0 7164 10 0 5 15 7160 10 0 5 30 7168 10 0 5 50 SEM and PL Analyses Remark Highest ma:x1mum RHEED intensity SEM analysis was performed on samples grown with Ga delay times of 0 sec ., 15 sec 30 sec and 50 sec No Ga droplets were found on the surfaces of these samples as can be seen from Figs 4 10 (a)-(d). Lowand room-temperature PL spectra recorded from the 0 sec 15 sec 30 sec 50 sec Ga delay time samples are presented in Figs 4 114 18. The influence of the Ga delay time on the PL properties will be discussed in detail in Chapter 5. Activation Energy of Ga Migration Under N-free Conditions As illustrated in Fig 4 19 the RHEED specular reflection beam intensity recovers upon closing the Ga shutter ( during the Ga-delay phase) in two distinct stages In stage 1 the recovery is fast and the RHEED intensity varies almost linearly with time However beyond stage 1 the recovery appears to be more sluggish in stage 2 in which the

PAGE 63

57 relationship is distinctly non-linear By assuming monatomic steps with terraces and edge s on the growing surface [62-66] it is believed that the fast stage (stage 1) of the RHEED specular reflection beam intensity recovery is dominated by Ga adatom migration to st ep sites [44 45 67-71]. With a view to determining the activation energy for the process the time to complete stage 1 recovery was measured as a function of growth temperature and the migration times recorded in the range 560 C to 640 C are presented in Table 4.4 The data displays an Arrhenius behavior as shown in Fig 4 20 This is the same approach as that used by Foxon and Joyce [ 48] who determined the activation energy of As 4 migration on a GaAs surface in the absence of a Ga flux The activation energy of Ga migration under N-free conditions was determined from Fig 4 20 to be 1.46 0 25 eV Impact of Impinging RHEED Beam on Ga Migration It is essential to understand the role if any that the impinging RHEED electron beam has on the kinetic behavior of Ga adatoms In order to investigate this possibility the electron gun shutter was opened continuously in the first instance during an operating cycle and the RHEED specular reflection beam intensity was recorded as shown in Fig 4 21 (solid line) Immediately following this cycle the chart recorder was rewound back to the start position of the previous cycle and this time the RHEED intensity was sampled by only opening the RHEED gun shutter for I sec every 5 secs In this case the trace recorded is shown in Fig 4 21 as a dashed line As can be seen from the figure the sampled intensities correspond to those recorded during continuous electron beam exposure and therefore it is assumed that Ga migration for instance was not electron beam assisted

PAGE 64

58 Table 4.4 Ga migration times estimated from the RH.BED intensity curves at various substrate temperatures. Migration time (sec ) Substrate temperature C) 35 ~ 40 560 25 ~30 580 13 ~17 600 10 ~13 620 6~9 640

PAGE 65

59 6 C 640 C 700 C 10 15 20 30 Ga delay time (sec.) Figure 4.1 RHEED specular reflection beam intensity as a function of Ga delay time (time allowed to elapse between closing the Ga and opening the N shutter) recorded at different substrate temperatures

PAGE 66

,,-.._ V) 0., u -.._, c V) d (1) 50000 T = 27 2K 45000 40000 35000 30000 25000 20000 15000 10000 365 5 nm 5000 0 350 400 450 500 550 600 650 Wavelength ( nm) F i gure 4 2 Low-temperature PL spectrum recorded from a GaN/sapph i re film grown by the AEEE technique at 700 C

PAGE 67

,-._ r/l 0.. u '._/ >, .-4 r/l d Q) d 61 400000 360000 320000 358nm 280000 240000 200000 160000 120000 80000 40000 385 nm 0 350 400 450 500 550 600 650 Wavelength (nm) Figure 4 3 Low-temperature PL spectrum recorded from a GaN / sapphire film grown by the AEEE technique at 640 C. 22K 700

PAGE 68

.,-._ ti'} 0.. Q '-" >-i ti'} c (l) c 62 358nm 24.5K 200000 150000 100000 50000 0 i......i..;-. ........................ ....i.....i.....i......i.....i.....i......i............,._....i.....,1-..1,......!..-oj__,__....i........J ............ ....i.......l--'--....i.......li-..l.......i..............i...... 350 400 450 500 550 600 650 700 Wavelength (nm) Figure 4.4 Low-temperature PL spectrum recorded from a GaN / sapphire film grown by the AEEE technique at 600 C.

PAGE 69

N shutter open for 10 sec J 63 N shutter closed Ga shutter open for 5 sec ~-------G_a_s_h_u_tt_er_c_lo_s_e __ ------1n1------------,N delay time (0sec ., 15 sec and 30 sec ), 0 sec. delay 15 sec. delay 30 sec delay Time (sec) Ga delay ti me 30 sec. Ga delay used for all three experiments ,, I Figure 4.5 RHEED specular reflection beam intensity recorded as a function of time during a cycle of AEEE growth The circles represent the beam intensities at the three N delay times used in the N delay optimization set of experiments Growth temperature was 600 C.

PAGE 70

,,-.._ en 0... C) '-' -0 en s:::: (1) ..... s:::: 64 358nm 24.5K 200000 150000 100000 50000 0 L..1.,.L..::a.i..,_.._.___.__.__.__.__ .......... ....i..... .......... ....i.....L......L..-l-.L......L.._.__L......L.._.__i......&.._.__L......L.._.__L......L.._.__..........._, 350 400 450 500 550 600 650 700 Wavelength (nm) Figure 4 6 Low-temperature PL spectrum recorded from a GaN / sapphire film grown by the AEEE technique with a O sec N delay at 600 C

PAGE 71

,,-._ rfl 0.. u -._,. ;;>-. rfl i::: (I) i::: 65 80000 72000 23 5 K 64000 56000 48000 40000 32000 24000 16000 8000 0 350 400 450 500 550 600 650 7 00 Wavelength ( nm ) Figure 4 7 Low-temperature PL spectrum recorded from a Ga N / sapph ir e film grown by the ABEE technique with a 1 5 sec N dela y ti me at 600 C

PAGE 72

70000 63000 56000 49000 ,,,-..._ 42000 Cl) 0.. C) ....._,, :>-. 35000 ..... Cl) (l) ..... 28000 21000 14000 7000 0 350 66 23.5 K 400 450 500 550 600 650 700 Wavelength (run) Figure 4 8 Low-temperature PL spectrum recorded from a GaN / sapphire film grown by the AEEE technique with 30 sec N delay time at 600 C

PAGE 73

N shutter open for 10 sec 67 J N shutter closed I .,____ ____ Ga:shutter open for 5 sec 0 sec Ga delay Ga sh utter closed 15 sec Ga delay 30 sec Ga delay Time (sec) 50 s~c Ga delay I Figure 4.9 RHEED specular reflection beam intensity recorded as a function of time during a cycle of AEEE for the O sec N delay case The circles represent the beam intensities at the four Ga delay times used in the Ga delay optimization set of experiments Growth temperature was 600 C

PAGE 74

68 (a) (b) Figure 4.10 SEM micrographs ofGaN film surfaces grown by the AEEE technique using various Ga-delay times (Tsubstrate = 600 C) (a) Ga delay time= 0 sec (b) Ga delay time= 15 sec.

PAGE 75

69 (c) (d) Figure 4.10 Continued (c) Ga delay time= 30 sec. (d) Ga delay time= 50 sec

PAGE 76

,,,-__ r/l 0.. u -.._,, .... r/l
PAGE 77

,,-.._ C/1 0.. Q '-' -+-' C/1 = (1) -+-' = 71 24 5K 200000 150000 100000 50000 358nm 0 L...L..1:::=---a.i......i......&,,_..i......1,....1.....i......i.--i-..L......1--i-...&.......l--"-...I...-L......L.....I..-.I-..L.....L....L......L.....I....J........l.--i-...L......J--"-...L......I 350 400 450 500 550 600 650 Wavelength (nm) Figure 4 12 Low-temperature PL spectrum recorded from a GaN / sapphire film grown by the ABEE technique with a 15 sec Ga delay time 700

PAGE 78

,-._ en 0. Q '-' >-i ....,_ en s:= (I) ....,_ s:= 72 358nm 24.5K 200000 150000 100000 50000 0 L..L..J..:::::.......i......i........___.i......i......1..-.,__._....1..-.,__._.....1..-i......i........1..-i......i.........___i......i.........___i......i.........___i......i.........___L...J.........___J........L..... 350 400 450 500 550 600 650 700 Wavelength (nm) Figure 4 13 Low-temperature PL spectrum recorded from a GaN / sapphire film grown by the ABEE technique with a 30 sec. Ga delay time.

PAGE 79

Cl) 0.. u -.._,, >-. Cl) i::::: (I) i::::: 7 3 24 5 K 200000 150000 100000 50000 358nm 0 L..L.1,;:;==---.,1-.1.,_.i......a......__i...........__.....i......__.._._ ___ ................... L......L.....a......L......L..-'--l...-l...-'--L......L.....I..-.I........L......J 350 400 450 500 550 600 650 700 Wavelength (nm) Figure 4 14 Low-temperature PL spectrum recorded from a GaN / sapphire film grown by the AEEE technique with a 50 sec Ga delay time

PAGE 80

,,-.._ rJ} 0.. u "-' >. rJ} i::: Q) i::: 74 RT 200000 150000 100000 364run 50000 0 ----.....__.__ ........................... ___.__.......___.,__._..........._..__.____.__......__..___.___.__....__..___.___.__ .............. ____._--.J 350 400 450 500 550 Wavelength (run) Figure 4.15 Room-temperature PL spectrum recorded from a GaN / sapphire film grown by the ABEE technique with a O sec Ga delay time 600

PAGE 81

rf) 0.. u "-" >-. -+-' .... rf) s:== Q) -+-' s:== 75 RT 200000 150000 364nm 100000 50000 0 ,:;;,_..........,___,,__.__._____.____.___.___.___.____.__ .............................. ........___.____.__......__......__......__...____.....__....__ ........... 350 400 450 500 550 Wavelength (run) Figure 4 16 Room-temperature PL spectrum recorded from a Ga / sapphire film grown by the AEEE technique with a 15 sec Ga delay time 600

PAGE 82

,,-.._ VJ 0.. Q '-' >-. VJ Cl) 76 RT 364run 200000 150000 100000 50000 0 ~~~L...~=:t::::::l:::::I::~~~~~c=~==:J 350 400 450 500 550 Wavelength (run) Figure 4 17 Room-temperature PL spectrum recorded from a GaN/ sapphire film grown by the AEEE technique with a 30 sec Ga delay time 600

PAGE 83

,,--.. r.r:, 0.. u '-' >-. ....... r.r:, = Q) ....... = 77 RT 200000 150000 100000 50000 365nm 0 L6__~~,!==t;;;=t==:I::::I:::i=:~=r=;==~=i======:J 350 400 450 500 550 Wavelength (nm) Figure 4 18 Room-temperature PL spectrum recorded from a GaN / Sapphire film grown by the ABEE technique with a 50 sec Ga delay time 600

PAGE 84

Ga migration tim~ 0 (Ga shutter closed) 78 I I I I stage 2 Time (sec.) Figure 4.19 RHEED intensity recorded during a Ga-delay phase showing the two distinct stages of recovery. Ga adatom migration times were determined for a variety of substrate temperatures by noting the time required for completion of stage I recovery as indicated at each temperature.

PAGE 85

,,-.._ u (I) en (I) s +,j i::::: 0 bb e 0 ....__, i::::: 7 9 100 -----.----------------------. 10 1.08 E n = 1.46 0 .2 5 eV 1.10 1.12 1.14 1.16 1.18 1 20 10 3 / T K ) Figure 4 20 Plot of ln ( Ga m i grat i on t i me ) v er s u s r ec i procal substrate temperature 1.22

PAGE 86

J 0 U') Q) ..... Q) ..0 0 ..... u Q) Q) a ;:::s u Q) 0.. U') Q N shutter open for 10 sec. Ga shutter open for 5 sec RHEED shutter closert--------' 80 N shutter closed Ga shutter closed ,. ,. . . ,, .... I L.__ __ RHEED shutter time (sec.) open Figure 4.21 RHEED specular reflection beam intensity recorded as a function of time during one cycle The solid line represents the RHEED intensity trace when the RHEED electron beam is continuously impinging on the surface, while the broken line represents the intensity when the RHEED electron beam shutter is periodically opened (~l sec ) and closed (~5 sec )

PAGE 87

CHAPTERS INVESTIGATION OF GaN GROWTH KINETICS AND THEIR CORRELATION WITH DEEP-LEVEL DEFECTS Introduction In this chapter the AEEE growth mechanisms will be compared with the mechanisms of conventional MBE growth specifically for a-GaN growth on sapphire First a discussion of the growth mechanisms will be given based on the growth parameters and the SEM characterization results which were presented in Chapters 3 and 4 Second speculation will be made about the origin of the deep-level defects that are believed to cause the 2 3 eV yellow band luminescence observed in all of the PL spectra recorded from conventionally-grown a-GaN films this speculation being based on the PL spectra discussed in Chapter 3 and on the theoretical and experimental results concerning deep-level defects discussed in Chapter 1 Finally the optimum AEEE shuttering sequence which contributes to the elimination of the yellow band emission will be discussed This discussion includes the presentation of a model concerning Ga adatom migration which was developed as a result of the real-time RHEED intensity observations during the Ga exposure and Ga delay phases Morphological Considerations The surface morphology of conventional MBE-grown GaN films deposited on sapphire is very sensitive to the growth parameters as illustrated by the SEM micrographs shown in Figs 3 5 and 3 6 Also the yellow band emission appears in the PL spectra recorded from all of the conventionally-grown samples in this study regardless of growth 8 1

PAGE 88

82 temperature (Figs 3 8-3 13) The substrate temperature and Ga/N beam flux ratio play important roles in determining the GaN surface morphology. Since the nitrogen reevaporation rate is extremely high particularly at the high temperatures required for conventional growth the residence time of nitrogen atoms is much shorter than that of Ga atoms Also the incorporation of Ga is limited by the lack of N atoms on the growing surface especially at high temperatures This situation is very different from the MBE growth of GaAs for instance in which the growth rate is controlled by the Ga flux only Experimental evidence concerning the MBE growth of GaN indicates that the sticking coefficient of Ga is only about 0 25 and attempts to increase the Ga flux or lower the growth temperature lead to Ga-rich surfaces which are extremely rough [72]. Crawford et al. [28] have reported that Ga accumulation on the growing surface actually decreases the GaN growth rate when the substrate temperature is too low or the Ga flux is too high Consequently it could be expected that the incorporation of Ga and N are correlated in the MBE growth ofGaN [53]. In this work a strong dependence of the surface morphology of conventionally grown GaN films on the growth temperature has been confirmed. The surface coverage of a GaN film grown at 750 C is about 90 % as can be seen from Fig. 3 5 As mentioned above the desorption rate of N is much higher than that of Ga hence there are fewer incorporation sites for the Ga adatoms. The observed micro-scale discontinuity of the film will limit the surface migration length of Ga adatoms even at such a high growth temperature producing films having localized Ga-rich and Ga-deficient areas Reducing the growth temperature to 640 C and to 600C decreases the N desorption rate, however the migration rates of Ga adatoms and Ga-N molecules are also greatly reduced The surface morphologies of the films grown at 640 C and 600 C indicate the presence of Ga droplets as can be seen from Fig. 3 6 The average size of the Ga droplets is about 15 m for the sample grown at 600 C and about 5 m for the sample grown at 640 C It is noteworthy that the Ga flux used for the sample grown at 600 C was actually three times

PAGE 89

83 lower than that used for the sample grown at 640 C and yet the droplet s ize i s l a r ger for the sample grown at 600 C This can be explained by assuming that the Ga de s orp t ion rate is higher at 640 C than it is at 600 C and that Ga adatom diffusion is e xt remel y slow at 600 C during conventional MBE growth As reviewed in Chapter 1 [29 45] with reference to GaAs growth kinetic s the Ga surface diffusion length at 550 C under Ga-stabilized conditions is about 10 times larger than that under As-stabilized conditions The migrating species on the growing surface when GaAs is grown with an As overpressure are believed to be Ga-As molecules rather than isolated Ga atoms since impinging Ga-atoms immediately react with impinging As molecules to form Ga-As molecules As in conventional MBE-growth of GaAs which is normally carried out using an As-overpressure conventional MBE-growth of GaN is normally carried out using an N overpressure In this case the impinging Ga-atoms have a high probability of immediately reacting with the impinging N-atoms to form Ga-N molecules ; thus surface migration takes place as Ga-N molecules (or GaN clusters) rather than isolated Ga-atoms The large activation energy for migration of the Ga-N molecule on the GaN growing surface is believed to result in extremely slow migration The short diffusion length of Ga-adatoms under these conditions enhances the formation of Ga-aggregates which become sinks for additional Ga adatoms This phenomenon is believed to have caused the slow GaN growth rates observed in references 27 and 28 In contrast samples grown at 600 C using the AEEE growth technique had surface morphologies devoid of Ga droplets as can be seen from the SEM micro graphs shown in Figs 4 10 (a)-(d). It is speculated that in this case rapid Ga adatom migration takes place in the absence of an N atom flux and thus Ga clustering is avoided

PAGE 90

84 Speculation Concerning the Source of Deep-Level Defects Responsible for the Yell ow Band Emission The yellow band emission intensity has a growth temperature dependence which is evidenced by comparing the PL spectra recorded from the AEEE samples grown at 600C with those of samples grown at higher temperatures (Figs 4 2-4.4) Yellow band emission is in fact completely absent from the spectra recorded from samples grown at a substrate temperatures of 600 C. It is postulated that higher growth temperatures result in the desorption of Ga during the Ga delay phase producing Ga-vacancies which when associated with dislocations result in the yellow band emission Dislocation densities have been estimated to be in the range 10 8 10 10 cm2 in typical GaN films grown on c-plane sapphire [25 26 54 61]. The deposition of Ga on GaN (0001) surfaces followed by thermal desorption of the Ga has been reported by Khan et al. [73]. In that study Ga was deposited at 620 C and was found to evaporate at temperature around 670 C which supports the Ga re-evaporation theory Also Schoonmaker et al have reported that GaN decomposition can be enhanced via catalysis by metal Ga [74]. Furthermore as reported in Chapter 4 it was found in this work that during the Ga delay phase the RHEED specular reflection beam intensity recovered faster at growth temperatures of 700 C and 640 C but recovered to a lower level of intensity than was the case for a sample grown at 600 C as shown in Fig 4 1 Also for a Ga exposure time at 700 C of 5 sec instead of an optimum 8 sec the RHEED specular reflection beam intensity was found to gradually decrease as the shuttering sequence progressed It is believed that Ga desorption under N-free conditions is more rapid at higher temperatures and Ga vacancy generation is promoted Ga vacancies were predicted as the most likely deeplevel candidate in n-type unintentionally doped GaN in the literature review section (Chapter I) Similar observations concerning Ga re-evaporation at relatively low temperatures have been made with regard to 11BE-GaAs Arthur [75] reported that Ga deposited on a GaAs (111) Ga-rich surface has a sticking coefficient of unity below

PAGE 91

85 477C However the Ga sticking coefficient is much less than unity unless an As flu x i s present when the substrate temperature is in the range 500600 C. Strong yellow band emission is evident in the PL spectra recorded from GaN films grown by conventional MBE as indicated in Figs 3 8-3 13 Also as shown in Fig 3 6 a high density of Ga droplets is observed on the surfaces of conventionally-grown films It is speculated that Ga vacancies are again involved in this case since the material providing the luminescence will be Ga-deficient due to the participation of Ga adatoms in Ga droplet formation As mentioned previously radiative transitions in GaN epilayers are thought to be associated with dislocations [25 26 61 ] in particular screw dislocations which are uncharged in hexagonal GaN [61]. The near-band-edge emission has previously been ascribed to the recombination of electrons with holes trapped at screw dislocations [ 61 ] as indicated in Fig 5 1. Based on the results of this present study it is speculated that the yellow band emission is due to electrons recombining with holes trapped at Ga vacancy / screw dislocation complexes as illustrated in Fig 5 1 As the concentration of Ga vacancies increases the probability of holes being trapped at V Ga / dislocation comple x es will increase and correspondingly so will the intensity of the yellow band emission Influence of the Ga and N Exposure Times and Delay Times When an N atom flux impinges on a Ga-covered surface while the Ga shutter is closed the N atoms will incorporate into the appropriate lattice-sites easily because the surface Ga atoms have one dangling sp 2 bond As mentioned in Chapter 4 a 10 sec N exposure time was required for the RHEED specular reflection beam intensity to decrease to its lowest level and then to recover to a saturation level beyond which the intensity again dropped For N exposure times longer or shorter than 10 sec the RHEED specular reflection beam intensity did not recover to the same level as it did in the case o f

PAGE 92

86 the 10 sec N-exposure time during subsequent cycles. Consequently a 10 sec exposure time was thought to be optimum for the substrate temperature range employed and such an N exposure time was used throughout this work. As can be seen from Figs 4.2 and 4.3 the PL spectra recorded from AEEE grown GaN deposited at 700 C and 640 C show weak peaks in the range 365-385 nm Such emission is believed to be related to N vacancies [20 37] which provide the group B level shown in Fig 1.2 Similar emission is also observed in the PL spectra ( Figs 4 7 and 4 8) recorded from samples grown at 600 C using N-delay times of 15 sec and 30 sec. however this emission is absent from the spectra recorded from the samples grown with a 0 sec N-delay time (Figs 4.11-14). Such evidence suggests that under Ga-free conditions (no Ga exposure) the N-adatoms have a short residence time and hence the optimum N-delay time was selected as 0 sec in this work. In general an N atom arriving at a chemically-active Ga-covered surface which has one unsatisfied sp 2 bond perpendicular to the surface per Ga atom will tend to form a stable sp 3 bond with underlying Ga and neighboring N atoms This will make N incorporation homogeneous on a Ga-covered surface (similar to As 2 on a Ga stabilized surface [76]) However N has a very high vapor pressure and so excess N atoms may be only weakly adsorbed on the surface Therefore N desorption is proposed to be the dominated mechanism contributing to the moderate recovery of the RHEED specular beam intensity during the N delay phase as shown in Fig 4 5 A primary goal of the AEEE growth mode was to promote the migration of Ga adatoms on N-covered surfaces under N-free conditions (no N exposure) at temperatures lower than those required for conventional :MBE growth The purpose of the Ga-delays was to eliminate the Ga clusters which are thought to form on Ga-exposed surfaces Isu et al. [68] used a scanning reflection electron microscopy (SREM) technique to observe the formation of Ga-droplets on Ga-exposed As-covered surfaces These authors also observed that the Ga droplets shrank in size when the Ga supply was interrupted

PAGE 93

87 The density of available incorporation sites for impinging Ga atoms is 1 134 7 x 10 15 cm2 for GaN. The Ga-exposure time in this work (Tsub s trat e = 600 C) was based on the maximization of the RHEED specular reflection beam intensity level. For a Ga effusion cell temperature of 830 C, corresponding to a Ga flux of 2 27 x 10 14 cm2 f1, a 5 sec exposure provided about 1. 135 x 10 15 Ga-atoms cm2 When the Ga shutter was opened the RHEED specular reflection beam intensity decreased to its lowest level in ~2 5 sec and then recovered to a saturated level in an addition 2 5 sec It was found that the RHEED specular reflection beam intensity recovered to its highest level for a Ga exposure time of 5 sec A shorter Ga exposure time resulted in a lower intensity level for the recovered signal while a longer than 5 sec Ga exposure resulted in a longer recovery period. The growth rate of all ABEE-grown GaN films regardless of the Ga-delay time was measured to be ~63% of a monolayer per cycle for a substrate temperature of 600 C In MOC VD atomic layer epitaxy "of GaN both Khan et al. and Sumakeris et al. reported about 2/3 of a mono layer per cycle growth although the films grown by Sumakeris et al. were amorphous [55 56]. Model for Ga Adatom Migration As discussed in Chapter 4 and illustrated in Fig 4 20 an activation energy for Ga migration on an N-free surface of 1.46 0 25 eV was determined in this work. To the best of the author's knowledge this is the first time that the activation energy of Ga migration on GaN surfaces has been reported Compared to the activation energy of Ga diffusion on GaAs given as Ed= 1.3 eV, by Neave et al [44] the barrier for Ga migrating on a GaN surface is higher It must be point out however that the arsenic flux was on during the Ga delay in the GaAs case since Neave et al. tried to simulate conventional MBE growth Ed is expected to be smaller than 1.3 eV for Ga-adatom migration under As-free conditions

PAGE 94

88 The adatom migration rate at site i is taken to be in an Arrhenius expression : R i = Ro i e x p (-E I k B T)--------------------15.1) where R oi is a prefactor which is related to the vibration frequency of site i surface atom T is the substrate temperature kB is the Boltzmann constant and E is the activation barrier for migration which depends on the local configuration of site i E includes a substrate barrier Es and a barrier E 0 contribution from each nearest-neighbor bond formed parallel to the substrate. The homonuclear bond energy of Ga is 1 71 e V [78] and the Ga-N bond energy is determined to be 3 1 eV using Pauling s approximation [79]. The larger activation energy of Ga migration on GaN compared with Ga on GaAs is understandable since GaN has a larger E s (ionic bonding) than the E s ( covalent bonding ) ofGaAs A model has been developed to explain the RHEED specular reflection beam intensity behavior observed during Ga exposure and subsequent Ga-delay time phases of growth The discussion of the model will be presented in terms of three stages namely the Ga exposure stage the first stage of recovery and the second stage of recovery. These three stages are illustrated in Fig 5.2 with regard to a typical RHEED intensity trace recorded in this work during a cycle of AEEE growth Ga Exposure Stage : As can be seen from Fig 5 2 the specular reflection beam intensity initially decreases goes through a minimum and then increases again during the Ga exposure What is not illustrated in this figure is the fact that for Ga exposures longer than 5 seconds the specular reflection beam intensity saturates at the intensity level corresponding to a 5 second Ga exposure

PAGE 95

89 An attempt is made to explain the Ga exposure stage behavior with reference t o the illustrations presented in Fig. 5 3 As mentioned in the previously section dislocation densities in the range I 0 8 IO 10 cm 2 have been reported for typical GaN films gro w n on c-plane sapphire [25 26 54 61] and most of these dislocations are pure edge type e xi sting on { 1 I 0 0} planes However screw or mixed type dislocations are also present and account for ~30% of the total dislocation density [79]. The magnitude of the Burger s vector is equal to the lattice parameter c for screw dislocations (5 18 A in GaN) and a/3 [ 1 I 2 0] for edge dislocations ( a is equal to 3 .19 A in GaN). Therefore because the dislocation nucleation energy increases linearly with k. b 2 where b is the magnitude of the Burger's vector and k is a constant edge dislocations account for a large fraction of the total dislocation density because of their lower formation energy. The edge dislocations which formed on { I 1 0 0 } during the coalescence of 3D-islands in the early stage of GaN / sapphire growth will terminate on (0001) surfaces These points of termination disrupt the orderly array of surface atoms as seen from Fig 5 3 (a) Consideration of mechanical equilibrium among the interfacial tensions in the case of heterogeneous nucleation yields Young's Equation Y. r cos 0 = 'Ys v Y rs as shown in Fig 5.4 (b) where f, s v and 0 represent film N-covered GaN surfaces vapor and contact angle respectively ~G* ( energy barrier to the nucleation) will depend on the surface energy of the cluster (1 r ) and contact angle between the cluster and the surface ( 0) for the heterogeneous nucleation case and can be represented as : ~G* = 16 7t (( r ) 3 f (0) / 3 (~G v ) 2 --------------------------------------------------------(5 2) where f (0) = (2 + cos 0)( 1-cos 0)2 / 4 and ~G v is the free energy change per unit volume accompanying of the of formation of small clusters Lattice mismatch between film and substrate will increase the overall energy barrier (~G* ) because the strain energy is involved in ~G v. However ~G* will decrease when the surface defects are presented When a surface having a large dislocation density [as shown in Fig 5 3 ( a)] is exposed to the Ga flux Ga adatoms will nucleate on the termination sites of dislocations

PAGE 96

90 as seen in Fig 5 3 (b) Ga aggregates having radii smaller than r* will shrink and disappear lowering the free energy At the same time, because of their lower energy barrier ( ~G* ) to the nucleation process, nuclei having radii larger than r* will grow to supercritical dimensions by the further addition of atoms as shown in Figs 5 3 ( c )-( d) and Fig. 5.4 (a) After 2-2 5 sec Ga exposure ( circle c and don the RHEED intensity variation curve in Fig 5 3) islands several atomic-layers high are formed and the RHEED specular reflection beam intensity drops rapidly to the lowest level indicating a rough surface The RHEED specular reflection beam intensity recovers after a 2 5 sec Ga exposure and it is proposed that the small Ga islands merge to form larger Ga clusters flattening the cluster surfaces which in turn decreases the surface roughness For 5 sec Ga exposures and longer ( circle e on the RHEED intensity variation curve in Fig 5 3) the RHEED intensity starts to saturate For this condition it is proposed that Young s equation for partial wetting conditions "( f > (Y s v Y rs ) > "l f, is balanced at a critical contact angle Further Ga deposition will be limited by increasing f (0) [Fig 5 .4 ( c)] which increases the energy barrier for the nucleation process (~G*) First Stage of Recovery : As can be seen from Fig 5 2 following closure of the Ga shutter which prevents further Ga exposure there is a fast recovery phase which lasts for about 15 seconds (for a substrate of 600 C) During this first stage of recovery it is expected that Ga adatoms migrate fairly rapidly (in the absence ofN atoms) on the terraces to the steps which results in a smoothing of the growth front as illustrated in Figs 5 5 and 5 6 Large Ga clusters are unlikely to move freely on the terraces however the Ga adatoms can detach themselves from the edges of the clusters and migrate quickly to the step (kink or ledge) sites Also at this stage the probability of Ga finding incorporation sites is high and the

PAGE 97

91 mean free path for Ga adatom migration is large It must be pointed out that the Ga adatoms have equal chance to either migrate to the upper or the lower terraces Second Stage ofRecovery : As can be seen from Fig 5 2, for Ga-delay times in excess of 15 seconds ( at 600C) a second recovery stage begins in which recovery (surface smoothing) is not as rapid as in the previous stage During this stage it is expected that a rearrangement of the terraces takes place which impacts longer range order as shown in Fig 5 5 [67 69 70]. For Ga delays longer than 30 sec the RHEED intensity drops slightly. It is speculated that step bunching occurs in these cases in which the terrace width locally decreases [80]. As can be seen from Figs 4 11-14 (low-temperature PL spectra recorded from AEEE-grown samples with different Ga-delay times namely 0 sec 15 sec 30 sec and 50 sec ) the near-band-edge emission intensity is about an order of magnitude higher from the 30 sec delay sample compared to the 0 sec or 15 sec. delay samples Also weak deep level emission centered around 450 nm is present at room temperature from the 0 sec and 15 sec delay samples but not from the 30 sec delay sample as shown in Figs 4 15 and 4 16 These results considered together with the model suggest that the achievement oflong-range order (as indicated by the observation of the second stage recovery of the RHEED intensity) produces the highest quality epilayers. However as shown in Figs 4 14 and 4.18 the near-band-edge emission intensity is again lower from the 50 sec delay sample (compared with the 30 sec delay sample) and so it appears that epilayer quality is degraded when step bunching is involved

PAGE 98

near-bandedge emission (365nm) < ~ Hole trap level provided by screw dislocations ( after ref. [ 61 ]) ---------Yellow band ~ emission (~550 nm) 2.3 eV --,------,----,..,...._ -,---,----,Range of ho 1 e trap / Aw /// levels associated with Ga vacancy / screw dislocation complexes Fig. 5.1 In defective GaN epilayers efficient radiative transitions are associated which screw dislocations that are either isolated or are in association with native point defects

PAGE 99

N shutter open for 10 sec. I Ga shutter open for 5 sec. 93 N shutter closed _____ ___.I L._I ______ G_a_s_h_u_tt_er_cl_o_s_ed ____ I Ga I ~XpOSl}re 1 stage I 1 1st stage 1 of recover 0 15 time (sec ) 2nd stage of recovery Figure 5 2 A typical trace of the RHEED specular reflection beam intensity recorded during one cycle of AEEE illustrating the three stages of the process discussed in the model presentation 30

PAGE 100

(a) (b) (c) (e) N covered surf ace I I d d l l I e ge 1s ocatton I \ line I I J 0 Ga shutter open for 5 sec. (b) (c) 0 (d) 2 3 5 Ga shutter closed / / / / / / / / / / / / Time (sec.) Figure 5 3 Ga cluster nucleation and evolution during the Ga exposure phase (in the absence of nitrogen) as suggested from the RHEED intensity observations

PAGE 101

(a) (b) (c) 1 0.5 f (0) (-) 95 radius of cluster Y substrate fs Q O ._ =--__________________ _. 0 0 Figure 5.4 (a) Free energy change per unit volume as a function of cluster size (b) Young's equation film-vapor interface coming to some equilibrium contact angle with the substrate surface ( c) Variation of the efficiency factor for nucleation by solid surface f(0) with the contact angle 0

PAGE 102

Electron beam direction 1r2101 Figure 5 5 Ga cluster distribution on the N-covered terraces following a 5 sec Ga exposure. (at T s ub = 600 C). I I I ,/ / Time (sec.)

PAGE 103

Electron beam direction (l0T0) I I I I I I I I Ga open 5 sec ,,-ED intensity 15 sec after Ga shutter closed Time (sec ) F ure 5 6 Ga adatom distribution after about 16 sec (at T ,,,,, 600 0 C) durin g ig Ga delay stage (both Ga and N shutters closed)

PAGE 104

b eam direction Electron ll2101 (1010) \ === 600 0 C) during b t 3 o sec (at T s ub fter a ou d) G datom distnbutton a d N shutters close Figure 5 7 G: ;elay stage (both Ga an

PAGE 105

CHAPTER6 CONCLUSIONS AND RECO:rvtMENDATIONS Yell ow band luminescence is commonly observed in the PL spectra recorded from GaN films regardless of the crystal growth technique and the substrate employed This yellow band emission is associated with radiative transitions involving deep-level defects in GaN In this research the impact of the growth kinetics on the production of these deep centers has been systematically studied Several GaN films were grown by conventional MBE using a variety of growth parameters as reference samples Ga droplets were observed on the surfaces of conventionally-grown samples particularly those grown at low temperatures The yellow band emission appeared in the PL spectra recorded from all of the conventionally grown samples In contrast Ga droplets were not observed on the surfaces of samples grown by the alternate element exposure method even in the case of samples grown at 600C In addition the yellow band emission was suppressed in the PL spectra recorded from the AEEE-grown samples It is concluded that the role of Ga adatom migration on growing GaN surfaces is critical with regard to the production of deep level defects It is essential in fact to promote Ga adatom migration with a view to minimizing the concentration of the deep centers responsible for the yellow band luminescence. It is also imperative as shown in this work to minimize the probability of N desorption occurring from growing GaN surfaces as this process generates other levels in the GaN energy-gap ( due to N vacancies) which themselves provide for radiative transitions 99

PAGE 106

100 An additional conclusion from this work is that further optical finality improvement can be realized by the provision of an optimum delay time following a Ga exposure (prior to a subsequent N exposure) which promotes further Ga adatom migration (in the absence ofN atoms and N 2 molecules) and also allows time for terrace re-arrangement (growth front smoothing) to occur. Figure 6 1 summarizes the impact that enhanced Ga migration has on the quality of GaN epilayers It is recommend that the novel AEEE-growth mode be applied to the growth of InGaN and AlGaN which are difficult to grow using conventional MBE because oflarge growth temperature differences for InN GaN and AlN It is also recommend that the ABEE-growth mode be applied to doping of 111-V nitride materials

PAGE 107

101 7000 6000 second order RT ,-._ harmonic (/:J 0.. 5000 Laser peak u 367 nm '-,/ ;;>-.. 4000 GaN epilayer grown by ... (/:J conventional MBE = (1) 3000 ... at 600 C i::: 2000 1000 0 250000 400 500 600 700 RT 200000 ,-._ (/:J 0.. u 150000 '-,/ ;;>-.. ... GaN epilayer grown by (/:J i::: 100000 migration enhanced (1) 364nm ... = epitaxy (MEE) at 600 C 50000 0 400 500 600 250000 364nm RT 200000 >-,-._ (/:J 0.. u 150000 GaN epilayer grown by '-,/ ;;>-.. ... migration enhanced (/:J epitaxy at 600 C with = 100000 >-(1) ... inclusion of optimum = Ga-delay times 50000 -~ I 0 u I 400 500 600 Wavelength (nm) Figure 6 1 Room temperature PL spectra recorded from variously-grown GaN epilayers

PAGE 108

LIST OF REFERENCES 1. S Nakamura M Senoh S Nagahama N Iwasa T. Yamada T Matsushita H Kiyoku and Y. Sugimoto Jpn J. Appl.Phys ., 35 (lB) L74 [1996]. 2 K Doverspike L.B Rowland D K. Gaskill S C Binari J.A. Freitas W Qian and M Skowronski Inst. Phys Conf Ser. No 141 101 [1995]. 3 P Kung X Zhang D Walker A. Sazier J. Piotrowski and A. Rogaiski Appl. Phys Lett 67 (25) 3792 [1995]. 4. W Rieger T. Metzger H Angerer R. Dimitrov 0 Ambacher and M. Stuzmann Appl. Phys Lett 68 (7) 970 [ 1996]. 5 T. Saaki and T. Matsuoka J. Appl. Phys 77 ( 1 ) 192 [ 1994]. 6 N Grandjean J. Massies and M Lerou x, Appl. Phys Lett. 69 (14) 2071 [1996]. 7 T W. Weeks M D Bremser K.S Ailey E.Carison W.G Perry and R F Davis Appl. Phys. Lett. 67 (3) 401 [1995]. 8 I P Nikitina and V A. Dmitriev Proceeding of the 21st International Symposum of Compound Semiconductors 431 [1994] edited by H Goronkin and V Mishra 9 A. Ohtani K S Stevens and R Beresford Appl. Phys Lett. 65 (1) 61 [1994]. 10 A. Gassmann T Suski N Newman C Kisielowski E Jones E.R Wwber Z L. Weber D M Rubin H.I. Helava I. Grzegory M Bockowski J. Jun and S Porowski J. Appl. Phys 80 (4) 2195 [1996]. 11 F A. Ponce D P Bour W Gotz and N.M Johnson Appl. Phys Lett. 68 (7) 917 [ 1995]. 12 S Nakamura T. Mukai M Senoh and S. Nagahama J. Appl. Phys 74 (6) 3911 [ 1993]. 13 C. Wetzel S Fischer J. Kruger and E E Haller Appl.Phys Lett. 68 (18) 2556 [1996]. 102

PAGE 109

103 14 C T. Cowan K.P O'Donnell S E Hooper and C.T. Foxon Appl. Phys Lett. 68 (3), 355 [1995]. 15 M Ramsteiner J. Menniger 0. Brandt H Yang and K H Ploog Appl.Phys Lett. 69 (9), 1276 [1996]. 16 C.G Van de Walle and J. Neugebauer Mat. Res Soc Symp Proc 378 467 [1995] edited by S Ashok J. Chevallier I Akasaki N M Johnson and B L. Sopori 17 J. Neugebauer and C.G Van de Walle Proceedings of the 22nd International Conference on the Physics of Semiconductors 3 2327 [ 1995] edited by D J Lockwood 18. J. Neugebauer and C.G Van de Walle Phys Rev B 50 8067 [1994]. 19 J. Neugebauer and C G Van de Walle Mat. Res Soc Symp Proc 339 687 [1994] edited by C.H Carter Jr ., G Gildenblat, S Nakamura R J. Nemananich 20 T.L. Tansley and R.J. Egan, Phys Rev B, 45 10942 [1991]. 21. D W Jenkins and J.D Dow, Phys Rev. B 39 3317 [1987]. 22 P Perlin, T. Suski H Teisseyre M Leszczynski I Grzegory J. Jun S Porowski P Boguslawski, J. Bernholc J.C. Chervin, A. Polian and T.D Moustakas Phys Rev Lett. 75 296 [1995]. 23 P Boguslawski E. Briggs T.A. White M G Wensell and J. Bernholc Mat. Res Soc Proc Vol. 339 693 [1994] edited by C.H. Carter Jr. G Gildenblat S Nakamura R.J Nemanich. 24 P Boguslawski E L. Briggs and J. Bernholc Phys Rev. B 51 17255 [1995]. 25 F.A. Ponce D P Bour W Gotz and P J. Wright Appl. Phys. Lett. 68 (1), 57 [1996]. 26 Z. L. Weber, H Sohn N Newman and J. Washburn J. Vac Sci. Technol. B 13 (4), 1578 [1995]. 27. C.R. Eddy Jr. and T D. Moustakas, J. Appl. Phys 73 (1), 448 [1993]. 28 D. Crawford, R Held A.M Johnston A.M Dabiran and P I Cohn http://nsr mij mrs org, MRS Internet J. Nitride Semicond Res 1 12 [ 1996].

PAGE 110

104 29 Y. Horikoshi M Kawashima and H Yamaguchi Jpn. J. Appl.Phys 27 169 [1987]. 30 R E Ewing and P E. Greene J. Electrochem Soc 11 1266 [1964]. 31. T. Ogino and M Aoki, Jpn J. Appl. Phys 19 2395 [1980]. 32 X Zhang P Kung A Saxier D Walker T.C Wang and M Razeghi Appl. Phys Lett. 67 (12) 1745 [1995]. 33 E R. Glaser T.A. Kennedy K. Doverspike L.B. Rowland D.K. Gaskill M Asif Khan D T Olson and J.N. Kuznia Phys Rev B 51 13326 [1995]. 34 W Shan, T.J. Schmidt R.J. Hauenstien and J.J. Song Appl. Phys Lett. 66 (25) 3492 [ 1995]. 35 M. Asif Khan D T. Olson J.N Kuznia W E. Carlos and J.A. Freitas Jr. J. Appl. Phys 74 (9) 5901 [1993]. 36 W Shan X.C Xie J.J. Song and B Goldenberg Appl. Phys Lett. 67 (17) 2512 [1995]. 37 W Gotz N.M Johnson R A. Street H Amano and I Akasaki Mat. Res Soc Symp Proc 378 491 [1995] edited by S Ashok J. Chevallier I. Akasaki N M Johnson, B L. Aopori 38 P Hacke T Detchprohm K. Hiramatsu N Sawaki K Tadatomo and K Miyake J Appl. Phys 76 (1) 304 [1994]. 39 F.J. Sanchez D Basak M A. Sanchez-Garcia E Calleja E Munoz I Izpura F Calle J.M J. Tijero B Beaumont P Lorenzini P Gibart T.S Cheng C.T. Foxon and J.W Orton http://nsr rnij mrs org MRS Internet J. Nitride Sernicond, Res 1 7 [ 1996]. 40 M E Lin B N Sverdlov and H Morkoc Appl. Phys Lett 63 (26) 3625 [1993]. 41. S D Kim H Lee and J.S Harris Jr ., J. Electrochem Soc 142 (5) 1667 [1995]. 42. Y Horikoshi and M Kawaahima J. Cryst. Growth 95 17 [1989]. 43 Y Horikoshi M .. Kawashima and H Yamaguchi Jpn. J. Appl. Phys 25 L868870 [1986] 44 J.H Neave P J. Dobson B A. Joyce and J. Zhang Appl. Phys Lett 47 (2), 100 [1985].

PAGE 111

105 45 B A. Joyce P J. Dobson J.H Neave and J. Zhang Proc 2nd Conf on Modulated Semiconductor Structures Kyoto Surface Science 174 1 [1986]. 46 S Nagata and T. Tanaka J. Appl. Phys 48 (3) 940 [1977]. 47 R J Field and S K Ghandhi J. Cryst. Growth 74 551 [1986]. 48 C.T. Foxon and B A. Joyce Surface Science 50 434 [1975]. 49 S Kurtin T. C McGill and C.A. Mead Phys Rev Lett 22 (26 ), 1433 [1 9 69 ]. 50 J.S Foresi and T.D Moustakas Appl. Phys Lett 62 ( 22) 2859 [1993]. 51 S D Lester F A. Ponce M Craford and D A. Steigerwald Appl.Phys Let t. 66 (10) 1249 [1994]. 52 S. Strite and H. Morkoc J. Vac Sci Technol. B 10 (4) 1237 [ 1 992]. 53 R.C Powell N E Lee Y.W Kim and J.E Greene J. Appl. Phys 73 ( 1 ), 1 89 [1993]. 54 C.Y Hwang P Lu W E Mayo Y. Lu and H Liu Mat. Res Soc Symp Proc 326 347 [1994] edited by G. Gumbs S Luryi B Weiss G W Wicks 55 J. Sumakeris Z Sitar K S Ailey-Trent K.L More and R F. Davis Thin Solid Films 225 244 [ 1993]. 56 M Asif Khan R A. Skogman J.M Van Hove D T. Olson and J.N Kuz ni a Appl. Phys. Lett. 60 (11) 1366 [1992]. 57 R M Park M B Troffer C.M Rouleau J.M DePuydt and M A. Haase Appl. Phys Lett. 57 (20) 2127 [1990]. 58 M A. Haase J. Qiu J.M DePuydt and H Cheng Appl. Phys Let t. 59 (1 1 ), 1 2 7 2 [1992]. 59 J.G Kim A.C. Frenckel H Liu and R.M Park Appl. Phys Lett. 65 (1 ), 91 [1994]. 60 H Liu A.C Frenkel J.G Kim and R.M Park J. Appl. Phys 74 (1 0 ), 61 2 4 [1993]. 61. Y.G Shreter and Y T. Rebane International Conference Extended Defects i n Semiconductors-96 ", Gien France Sept. [ 1996]

PAGE 112

106 62 Y. Tokura H. Saito and T. Fukui J. Cryst. Growth 94 46 [1989]. 63 K Pond AC Gossard A Lorke and P M Petroff, Materials Science and Eng i neer i ng B 30 121 (1995]. 64 A Madhukar J. Cryst. Growth 163 149 (1996]. 65 S.A Chalmers J.Y. Tsao and AC Gossard Appl. Phys Lett. 61 ( 6) 645 [1992]. 66 J.P A v an der Wagt and J.S Harris Jr. J. Cryst. Growth 127 1025 [1993]. 67 J.H Neave B A Joyce P.J Dobson and N Norton App l. Phys A 31 1 [1983]. 68 T. Isu Y. Morishita S Goto Y. Normura and Y. Katayama J Vac Sci Tecnol. A, 12 ( 4) 1176 (1994]. 69 P B. Dobson B A Joyce J.H Neave and J. Zhang J. Cryst. Growth 81 1 [1987]. 70 X M Zhang D W Pashley I. Kamiya J H Neave and B A Joyce J. Cryst. Growth 147 234 [1995]. 71. B F Lewis F J. Grunthaner A Madhukar T.C Lee and R. Fernandez J. Vac. Sci Technol. B 3 1317 (1985] reprinted i n Molecular Beam Epitaxy AIP Press Woodbury NY 295 (1994], edited by A Cho 72 S St r ite J. Ruan Z Li A Salvador H Chen D J. Smith W J. Choyke and H Morkoc J. Vac Sc i Technol. B 9 ( 4) 1924 [1991]. 73 M AKhan J.N Kuznia D T. Olson and R. Kaplan J. Appl. Phys ., 73 3108 [ 1993]. 74 R C Schoonmaker A Buhl and J. Lemley J. Phys Chem 69 3455 [1965]. 75 J R. Arthur J r. J. Appl. Phys 39 4032 (1968] reprinted in Molecular Beam Epitaxy AIP Press Woodbury NY 51 (1994] edited by A Cho 76 F Brones and A Ruiz J. Cryst. Growth 111 194 (1991]. 77. R. T. Sanderson J. Am Chem Soc 105 2259 [1983]. 78 L. Pauling The Chemical Bond Cornell University Press Ithaca NY [1967]. 79 D. Kapolnek X H Wu B Heying S Keller B P Keller U K Mishra S P Denbaars and J S Speck Appl. Phys Lett. 67 (11) 1541 (1995].

PAGE 113

107 80 T. Ide, A. Yamashita and T. Mizutani Surface Science 287/288 1013 [1993].

PAGE 114

BIOGRAPIDCAL SKETCH Hsing-Long Liu received his Bachelor oflndustrial Education from National Taiwan Normal University Taipei Taiwan in June 1981 After graduating he taught at San Chung Technology School. He then continued his education at Auburn University Alabama where he earned a Master of Science degree in materials science in August 1990 His M.S thesis topic was Synthesis of SiC / SiC Ceramic Composites by CVD ". He then proceeded to the Department of Materials Science and Engineering at the University of Florida where he has been investigating the growth of GaN by molecular beam epitaxy working toward his Ph D 108

PAGE 115

I certify that I have read this study and that in my opinion it conform s t o acceptable standards of scholarly presentation and is fully adequate in scope and qual i t y, as a dissertation for the degree of Doctor of Philosophy ~e-v-~-PoJ:: Robert M. Park Chairman Professor of Materials Science and Engineering I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy Pa ../;V' //4i4i.-Y Pat'.tltt. Holl~way /' Professor of Materials Science and engineering I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy Science and Engineering I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy . ,/'/ /25~ Kevin S Jon ?s' Associate Professor of Materials Science and Engineering

PAGE 116

I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philos ~ .... L Sh g S Lt Professor of Electrical and Computer Engineering This dissertation was submitted to the Graduate Faculty of the College of Engineering and to the Graduate School and was accepted as partial fulfillment of the requirements for the degree of Doctor of Philosophy. December 1996 I.Q ~ Winfred M Phillips Dean College ofEngineering Karen A. Holbrook Dean Graduate School

PAGE 117

UNIVERSITY OF FLORIDA II I II IIIIII Ill Ill lllll lllll II IIIIII IIII IIII IIII IIII IIII II IIIII I I 3 1262 08554 4418