A study of aluminum-germanium-nickel ohmic contact metallurgical effects at the gallium arsenide interface

MISSING IMAGE

Material Information

Title:
A study of aluminum-germanium-nickel ohmic contact metallurgical effects at the gallium arsenide interface
Physical Description:
vi, 116 leaves : ill. ; 29 cm.
Language:
English
Creator:
Lampert, William V., 1948-
Publication Date:

Subjects

Genre:
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1992.
Bibliography:
Includes bibliographical references (leaves 109-115).
Statement of Responsibility:
by William V. Lampert.
General Note:
Typescript.
General Note:
Vita.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001800861
notis - AJM4619
oclc - 27669827
System ID:
AA00003697:00001

Full Text












A STUDY OF ALUMINUM-GERMANIUM-NICKEL OHMIC CONTACT
METALLURGICAL EFFECTS AT THE GALLIUM ARSENIDE INTERFACE















By

WILLIAM V. LAMPERT


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


1992














ACKNOWLEDGMENTS


The author would like to thank all of the members of his supervisory committee

for both their patience and assistance throughout his research. In particular, he

appreciates the guidance and constant encouragement of his committee chairman and

advisor, Dr. P. H. Holloway, and also that of his committee member and group leader

at the Wright Laboratory Materials Directorate, Dr. T W. Haas. He would also like to

thank Eric Lambers for the assistance with the Auger Electron Spectroscopy, Dr. Baoqi

Li for assistance with the room temperature I-V measurements, Chris Roubelet for the

assistance with the real time heat treatment and I-V measurements, co-workers Dr.

Scott Walck and Dave Tomich for their help with data translation for the computer, and

his other co-workers at the Materials Directorate for their help and agitation to finish.

The author would most like to express his gratitude to his sons, Curt, Thomas,

and Doug and especially to his wife, Nancy for the constant love, support, and sacrifice

they gave without question over the many long years he pursued his education. Special

thanks also go to his parents, Thomas and Edna for their support and instilling in him

the belief that anything was achievable given enough effort.

Acknowledgment is also due the United States Air Force for their Long-Term

Training Program which made it possible for the author to pursue this work.














TABLE OF CONTENTS

ACKNOWLEDGMENTS .................................... ii

ABSTRACT .............................................. v

CHAPTERS

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

2 REVIEW OFLITERATURE ................................. 6

Au/Ge/Ni Ohmic Contacts to GaAs .......................... 6
Other Metals and GaAs .................................. 15
Background for Aluminum Based Ohmic Contacts to GaAs .......... 19

3 EXPERIMENTAL PROCEDURE ............................ 30

Sample Preparation .................................... 30
M etal Deposition ...................................... 32
Electrical Measurement .................................. 37
Heat Treatment ...................................... 39
Real Time I-V Measurement and Heat Treatment ................. 40
Auger Electron Spectroscopy ............................. 41
Thin Film X-ray Diffraction ............................... 42

4 RESULTS ............................................. 44

Sample Preparation .................................... 44
Sample Deposition ..................................... 45
Heat Treatment ........................................ 47
Electrical Measurement .................................. 48
Real Time I-V Measurement and Heat Treatment ................ 56
Summary of Current-Voltage Data ......................... 59
Elemental Depth Profiles ................................. 62
Three Element Contacts ................................ 62
Two Element Metallizations .............................. 69
Summary of Elemental Depth Profiles ....................... 71
Diffraction Analysis of Phase Formation in Al-Ge-Ni Thin Films....... 73
Three Element Metallizations on GaAs ...................... 73
Three Element Metallizations on Sapphire ..................... 80
Two Element Metallization .............................. 80
Summary of Phase Formation in Al-Ge-Ni Thin Films ............ 82







iv
5 DISCUSSION .......................................... 86

Reaction Path Analysis ................................... 87
Mechanism of Doping ................................... 90
General Model of Ohmic Contact Formation .................... 90

6 CONCLUSIONS ......................................... 94


APPENDICES

A GaAs SUBSTRATE AND SAMPLE CLEANING .................. 97
B METALLIZATION SEQUENCE, HEAT TREATMENT,
AND ELECTRICAL CHARACTERIZATION ................... 103

REFERENCES ............................................ 109

BIOGRAPHICAL SKETCH ................................... 116














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


A STUDY OF ALUMINUM-GERMANIUM-NICKEL OHMIC CONTACT
METALLURGICAL EFFECTS AT THE GALLIUM ARSENIDE INTERFACE


By

William V. Lampert

August 1992
Chairman: Paul H. Holloway
Major Department: Materials Science and Engineering

Electrical and metallurgical properties of layered thin film Al-Ge-Ni ohmic

contacts to GaAs have been studied with current-voltage (I-V) measurements, Auger

electron spectroscopy, and X-ray diffraction. The objective of this study was to

identify the interfacial reactions which took place upon heating. These data were

correlated with elemental profiles and alloy formation to rationalize models of formation

of ohmic contacts.

I-V data collected by a transmission line method demonstrate dramatic

differences existed in the time and temperature required to convert Al-Ge-Ni on GaAs

to ohmic behavior. For samples with Ge adjacent to heavily doped GaAs (Si, -1 X
1018 cm-3), heat treatments of 500' C and 425' C required times up to 9 and 20

minutes, respectively, for conversion to ohmic behavior. Similarly, for low doped

substrates (Si, -5 X 1016 cm-3) heat treatment at 500" C required 42 minutes, while for

heat treatment at 425' C theconversion was not complete after 100 minutes. These data

show the time required to convert to ohmic behavior varied inversely with doping

concentration and directly with temperature. For samples with the Ni layer at the







vi
interface of both heavily and low doped GaAs, the time required to convert was 2 to 3

minutes for either 500' C or 425' C. Thus, conversion to ohmic behavior was much

less dependent on doping concentration or temperature with Ni adjacent to GaAs.

Simultaneous with the transition to ohmic behavior, Auger electron

spectroscopy and X-ray diffraction show that interdiffusion took place across the metal-

semiconductor interface and between metal layers. Neither Ga nor As were found in

the Al overlayer, while Ni diffused into the Al and Ge layers and formed both solid

solutions and intermetallic phases (e.g. A13Ni, GeNi, and Ge3Nis). However, Ga and

As did alloy with Ni to form phases such as Ni4GaGe2, NiAs, Ni5As2, NiAs(Ge), and

Ni2Ga3. Only limited phase formation was found between Ge and Al. As a result, the

Ni-Ge, Ni-GaAs and Ge-GaAs reactions control the ohmic contact formation.

Formation of Ni-GaAs compounds followed by degenerative doping of the substrate by

Ge best explain the change from rectifying to ohmic contacts.















CHAPTER 1
INTRODUCTION

The desire to have faster electronic components that were capable of operating at

higher temperatures precipitated this study of Al-based ohmic contacts for the III-V

semiconductor GaAs. Zuleeg et al.1,2 originally proposed the use of the Al-Ge-Ni ohmic

contact system for n-type GaAs because of its potential to operate at higher temperature

and be more radiation hard as compared to the more commonly used Au-Ge-Ni ohmic

contact system. GaAs was of interest because, in comparison with Si, it has higher

electron mobility, greater radiation hardness, can operate at higher temperature, uses less

power, absorbs sunlight more efficiently and can alloy with additional elements to form

materials with continuously variable properties. Possibly most important, GaAs is a

direct bandgap semiconductor and therefore efficiently converts electrical power to light.

Thus the opportunity for integration of opto-electronic devices with electrical integrated

circuits exists for GaAs.

All semiconductor devices require contacts to communicate with the outside

world. Metal/semiconductor contact literature dates back to at least 1874, when Braun3

published a paper on the asymmetric relationship of electrical conduction between metals

and semiconductors. In 1938 Schottky4 and Mott5 independently published the first

acceptable model of metal-semiconductor contacts, explaining the mechanism of barrier

formation based on the barrier height which depended on both the work function of the

metal and the electron affinity of the semiconductor. According to their model, a contact

metal whose work function is less than the electron affinity of the n-type semiconductor

should exhibit ohmic properties. The model works reasonably well for semiconductors

such as Si; however, experimental evidence shows this is not the case for most







2
circumstances with a low work function metal on n-type GaAs. A model, proposed by

Bardeen6 in 1947, attributes deviations from the model of Schottky and Mott to a large

density of surface states. These states are distributed in the forbidden energy gap

between the conduction band and the valance band energy levels, and can trap both

electrons and holes near the surface. For an n-type crystal these states will trap electrons,

producing a negative charge on the surface that shifts the valence and conduction band

edges up in energy relative to the bulk semiconductor. To maintain electronic charge

balance these electronic defects near the surface of GaAs pin the Fermi level over a rather

narrow energy range at the middle of the bandgap, which renders the Schottky

relationship invalid. According to Bardeen's model, this problem will develop when the

population of surface states is large (>1013 cm-2). A number of explanations advanced

more recently have described the origin of the surface states and resultant Fermi level

pinning.7-10 These models were based on either atomic rearrangements near the metal-

semiconductor interface, or on redistribution of valence electron charge density at this

interface. The unified defect model,7 chemical reactivity model,8 and effective work

function model9 are examples of explanations for the Fermi level pinning. Derivations of

these models were based on data from metals deposited on vacuum cleaved (110) GaAs

and air exposed (100) GaAs. The existence of the Fermi level pinning was important for

this study; however, the exact cause of its formation was not critical to the formation of

ohmic contacts.

Under ideal circumstances, the current carrying capacity of an ohmic contact

obeys Ohm's Law; that is, the current flow between the metal and the semiconductor is

linearly proportional to the voltage under both positive and negative biases. Practically,

an ohmic contact is defined as a metal-semiconductor contact that has a negligible contact

resistance relative to the bulk or spreading resistance of the semiconductor. An ohmic

contact will supply the current with a sufficiently small voltage drop such that the device

will perform properly.11 Various metals and metal alloy systems were used as ohmic







3
contacts to GaAs over the years. Alloyed contacts generally form a phase which melts at

low temperature and a dopant for the type of GaAs of interest (n or p type). For a typical

process the metals were deposited onto the GaAs substrate and heated above a eutectic

temperature. A portion of the GaAs dissociates and mixes with the alloy, resulting in a

degenerative doped layer of GaAs beneath the contact because of dopant incorporation

during GaAs regrowth. Charge is transported across the interface by a combined

thermionic and field emission conducting mechanism.11

There is an incredibly large body of literature discussing the many aspects of

ohmic contacts to GaAs. There have also been many different types of ohmic contact

metallizations developed for GaAs. For a comprehensive review of this literature before

1984 see the articles of Rideoutl2 and Piotrowska et al.13 For more recent reviews see

the articles of Sands,14 Murakami,15 and Piotrowska and Kaminskal6 as well as the

reference books and book chapters of Williams,17 Sharma,18 and Robinson.19 In this

dissertation, the literature on Au-Ge-Ni and Al-Ge-Ni ohmic contacts will be reviewed in

detail. The rest of the ohmic contact systems will be referenced only as they specifically

apply to the present study.
As stated earlier, Zuleeg et al.1,2 originally proposed the use of an Al-Ge-Ni

contact system as a substitute for the Au/Ge/Ni contact system, predicting that the Al-

based metallization would provide higher radiation tolerance and improved thermal

stability for higher temperature applications. This contact system has been investigated

much less than the Au-Ge-Ni contact system. Only a relatively small body of literature

exists for the Al-Ge-Ni contact system.20-23 However, the initial ohmic contact

properties were promising both due to the potential for higher temperature applications

and to the uniformity of the metal-semiconductor interface. The existing literature

suggests that the metallurgy of Al-Ge-Ni ohmic contacts differs from Au-Ge-Ni ohmic

contacts. However, the relationship between ohmic behavior and microstructure of the

Al/Ge/Ni/GaAs ohmic contact as a function of thermal processing is unknown. The







4
objective of this dissertation was to study the interfacial reactions between the Al-Ge-Ni

metallization and the surface of clean GaAs, then to correlate this information with phase

diagram data to rationalize alloy formation and elemental profiles. These data were then

used to study any correlation between phase formation and ohmic behavior. A model to

explain these observations based upon a decomposition and regrowth mechanism will

then be postulated. The variables studied were the component sequencing, layer

thickness and thermal processing of the Al, Ge, and Ni thin films to determine their

effects on elemental distributions and phase formations after the conversion from

Schottky to ohmic behavior. The data show that the layer sequence of Ni versus Ge

adjacent to GaAs dramatically affects the kinetics of the reactions between the metals and

semiconductor. While the sequence of metals has been shown to affect the quality of

ohmic contacts in some other systems, it has not previously been reported to affect the

reaction rates which control the conversion from Schottky to ohmic behavior. Electrical

measurements will be presented which show the dramatic effect upon kinetics of first

depositing Ni at the metal/semiconductor interface.

In Chapter 2, Review of the Literature, the general ohmic contact literature and the

Au-Ge-Ni contact system are reviewed first. Also included is a detailed review of the Al-

Ge-Ni contact system, literature which describes ohmic contact formation by a

decomposition and regrowth mechanism, and phase diagram literature related to ohmic

contacts. In Chapter 3, Experimental Procedures, sample preparation, sample

deposition, and thermal processing methodology are described. Details of the analytical

procedures are also included in this chapter. In Chapter 4, Experimental Results,

changes in electrical properties are reported in the first section. Interdiffusion of Al, Ge,

Ni, and the GaAs substrate was studied by Auger electron spectroscopy as reported in the

section entitled Elemental Depth Profiles. The effects of interdiffusion on formation of

both solid solutions and intermetallic phases are reported in the third section, Phase

Formation in Al-Ge-Ni Thin Films. These results are then discussed in Chapter 5,







5
Discussion, and a model to explain the observations is postulated. Finally, conclusions
regarding the effects of metal layer sequence, temperature, time and dopant level upon
formation of ohmic contacts between Al-Ge-Ni and GaAs are summarized in Chapter 6.















CHAPTER 2
REVIEW OF LITERATURE


Various metals and metal alloy systems have been used as ohmic contacts to

GaAs over the years. An alloyed contact generally contained a low melting phase and a

dopant for the GaAs. The alloy was deposited onto the GaAs substrate and then heated

until a portion of the GaAs dissociated and mixed with the alloy resulting in a

degenerative doped layer of GaAs beneath the metal contact. Charge is transported

between the metal and semiconductor by a combined thermionic and field emission

conducting mechanism. While the exact mechanism responsible for the degenerative

doped surface layer is still controversial, there is no doubt that the contact resistance can

be made low enough (- 10-7 Q-cm2) to be functional in nearly all applications.


Au/Ge/Ni Ohmic Contacts to GaAs


The most commonly used metal alloy system for ohmic contacts to n-type GaAs

is the Au-Ge-Ni metallization introduced by Braslau et al.24 They were in search of

contacts which would provide low linear series resistance, were noninjecting in the

presence of high applied fields, could be manufactured and processed in a uniform and

reproducible manner, could form patterns of controllable size and shape, and would be

adaptable to batch fabrication. These contacts were intended for use instead of the

common contacts of the day: Sn, In, Au, Ag, as well as alloys of Sn-Ni, Ni-In, and Au-

In. Tin had been widely used as an alloyed contact for GaAs, but the large contact angle

of liquid Sn on GaAs causes it to form into islands on the GaAs surface which, unless







7
constrained, causes nonuniform morphology. Sn also has a tendency to form conducting

channels in the semiconductor when subjected to high electrical fields, and the electrical

properties depend rather critically on the alloying temperature.25 This led Braslau et al.24

to search for a system not involving Sn. They found that Au and Ge, evaporated from a

eutectic composition, could form ohmic contacts to n-type GaAs. However, like Sn, the

deposition from the Au-Ge eutectic also formed minute droplets on the GaAs surface.

Even though good wetting was attained at the higher temperature at which alloying took

place, the resulting film had an island structure similar to Sn. Braslau et al.24 resolved

this problem by adding a small amount of Ni (2-11% by weight) to their eutectic Au-Ge

evaporant. Though these films were evaporated from a single mixture, they did not

evaporate congruently, which resulted in a layered structure. During heat treatment above

the eutectic temperature the Au and Ge recombined, while the Ni layer remained intact

owing to its low solubility in Au-Ge at this temperature. Braslau et al.24 speculated that

the presence of the solid Ni overlayer held the Au-Ge melt in intimate contact with the

substrate at the alloying temperature, resulting in the contact being uniform over the

surface of the semiconductor after solidification.

To be useful a metallization must form an ohmic contact over a broad range of

doping concentrations in the contact layer of the GaAs. The Au-Ge-Ni contact system

has worked over a wide range of doping concentrations from lightly doped Gunn

oscillators to degenerative doped injection lasers.26 Owing to its practical importance, a

substantial body of literature exists concerning formation and characterization of the Au-

Ge-Ni contact system by various processing and analysis techniques. As stated earlier,

there are a number of review papers as well as the reference books and book chapters

describing ohmic contacts to GaAs.12-19 The Au-Ge-Ni contact system is thought to

behave according to the general model for an ohmic contact to GaAs as described above.

Au-Ge forms a low melting eutectic and Ge is a dopant for the GaAs. The effects of Ni

are still controversial. Ni causes a portion of the GaAs to dissociate with in-diffusion of







8
the Ge resulting in a degeneratively doped layer of GaAs beneath the metal contact

according to one model.27 28 An alternative model suggests a regrowth mechanism of

the highly doped layer during cool down or heat treatment rather than the diffusion

mechanism.29-35 The regrowth mechanism will be discussed later in more detail.

To develop a better understanding of the Al-Ge-Ni contact system, an

understanding of the materials and processing of the Au-Ge-Ni contact system is useful.

Therefore, a more detailed description of the effect of each of the components and their

interrelationships on the complete metal and semiconductor contact system follows. The

well-known high concentration of surface states, present under ordinary processing of

GaAs, pins the Fermi level near midgap. Deposition of most metals onto a cleaned n-

type GaAs surface results in a Schottky contact with a barrier height of -0.8 eV. To

lower the contact resistance, it is necessary either to reduce the barrier height to allow

greater electron transport by thermionic emission or to increase the doping concentration

so that the electron transport by thermionic emission is enhanced by field emission

(tunneling). As stated earlier, Ge is believed to degenerative dope the GaAs so that the

contact conducts through the combination thermal activation assisted tunneling

mechanism. However, Ge is amphoteric in GaAs. Andrews and Holonyak36 reported

that bulk Ge-doped GaAs grown from a Ga rich solution is p-type, while bulk Ge-doped

GaAs grown under conditions of exact stoichiometry or excess As is n-type. Evidence

that Ge formed a degenerative doped n+ layer was provided by Anderson et al.27 and later

confirmed by Heiblum et al.37 Anderson et al. deposited the Ge and Ni onto a p-type Zn

doped GaAs substrate and produced a pn junction with Ni and Ni-Ge contacts. The pn

junction indicated that Ge in-diffused to compensate the p-type substrate.

There are several opinions as to whether Ge could diffuse into GaAs to

accomplish this fact. According to Aina et al.38,39 Ge is a substitutional diffuser with

diffusion properties similar to other substitutional diffusers such as Te, S, and Sn.

Auger depth profiles were used to show the Ge diffused from the metallization toward the







9
metal-semiconductor interface.40 Ge penetrates the GaAs by way of a Ga vacancy

dependent diffusion model according to Gupta and Khokle.41 Dell et al.42 showed that a

critical amount of Ge was required for the Au-Ge-Ni metallization to convert to ohmic

behavior by varying the total amount of each of the metals deposited while maintaining

the overall ratio. The resistance increased dramatically when the total amount of metal

was less than 600A. They used a single source of Au-Ge eutectic with -5 % Ni. They

concluded a sufficient amount of Ge was needed to produce the tunneling. Robinson

suggested the formation of an n+ layer resulted from Ge penetration of the dissociating

GaAs.40 He further concluded that Ge replaced the Ga to dope the GaAs. The role of

Ge was reported to be involved in the dissociation of GaAs as well as serving as the n+

dopant to GaAs by Iliadis and Singer.43 However, Anderson et al.27 found that even Ge

layers grown by liquid phase epitaxy did not significantly penetrate the GaAs without the

presence of a Ni overlayer. Ge combined with the Ni as long as the amount of Ni was

sufficient to consume the Ge. If so, the contact remained uniform; however, if the ratio

of Ge was greater than Ni, the contact layer became laterally nonuniform.44 Thus the

ratio of Ge to the other materials was important. Excess Ge caused out diffusion and

precipitation on the surface during heat treatment.28, 44-46 Crouch et al. reported a

dendritic growth of 1 to 1 Ge to Au precipitated on the surface of Au-Ge contacts heat

treated by rapid thermal annealing (RTA).45

As stated earlier, Braslau et al. originally concluded the Ni only served to cap the

eutectic Au-Ge melt and to hold the Au-Ge in place during solidification and prevent it

from balling up due to surface energy.24 It was later reported that Ni aided the wetting

rather than serving as a cap to hold the Au-Ge in place.40 47 The Auger depth profiles of

Robinson indicated that Ni also diffused to the metal-semiconductor interface during heat

treatment. For samples heat treated below the eutectic temperature of the Au-Ge, Ni was

also found to diffuse rapidly through the separated Au and Ge layers to the GaAs

interface.40 His results demonstrated GaAs decomposition and interdiffusion of Ni, Au,







10
and Ge during alloying. Ogawa found Ni to be a very fast diffuser in both the contacts

and in GaAs.28 48 He further suggested that Ni also served to catalyze the reaction
between Au and the GaAs. Auger depth profiles were used to show out-diffusion of the

Ga and As. Ni and As Auger depth profile peak overlap was used to infer formation of a

NiAs phase or at least the formation of Ni and As grains.28, 40, 49 They suggested that

the Ni increased the solubility of GaAs and enhanced the in-diffusion of Ge so that Ge sat

on substitutional Ga sites and created the n+ layer. Ogawa28 also reported X-ray

diffraction results which indicated that the NiAs was textured during the early stages of

heat treatment and that it incorporated Ge during the high temperature stage of the heat

treatment. He made a case for shorter heat treatment times at higher temperatures (e. g.

500" C), to avoid the deeper penetration of the contact metals such as Ni into the GaAs

substrate. Later Ogawa studied only the Ni-GaAs reaction and identified the ternary

NizGaAs.48 The Ni2GaAs formed at -200" C, but decomposed to NiGa and NiAs above

400'C. Similar observations of a ternary phase decomposition for the Ni-GaAs system

were made by others.50-53 The depth of penetration by the various Au-Ge-Ni

components has been reported to be from 1000A to 3000A using analytical techniques

such as Rutherford Backscattering, RBS, SIMS, backside SIMS, and transmission

electron microscopy (TEM).50, 51, 54-55 The ratio of Ni to the other materials was also

reported to be important. For example, Patrick et al.56 found that the heat treating

temperature required for forming low resistance contacts was lower when the ratio of 5%

Ni to Au-Ge was maintained. As stated above, excess Ni has led to excessive out-

diffusion of Ga, leaving excess Ga vacancies and too little Ge to fill these donor sites.

The noble metal gold was the part of the metallization to which the outside
connections were made. The eutectic of Au and Ge provided a lower melting alloy for

lower temperature heat treatment of the ohmic contact metallization. The Au-Ge eutectic

composition evaporates upon deposition into the two terminal phases. The Auger depth

profiles indicated a separate though overlapping layered structure in the as-deposited







11
contacts.28, 40, 45, 49 Au was found to diffuse to the metal-semiconductor interface and

react with the GaAs during heat treatment. Au is a fast interstitial diffuser (Do=10-3

cm2/sec, Ea=1.2eV) in GaAs.39 Liliental et al. showed Au rich protrusions into the

GaAs.57 At a temperature of 300" C, Au began to react with the GaAs.28 Whether the

heat treatment was above or below the eutectic temperature, GaAs dissociated until one

component reached its saturation limit in the metal layer.26 Dissociation and out-

diffusion of Ga and As had been observed for Au-Ge contacts.43, 58 Similar results

were reported for the Au-Ge-Ni contacts.26,28,40,49,58 The Ga out-diffusion was by

way of grain boundary diffusion according to Gupta and Khokle.41 The studies of

Kulkami and Lai59 also supported a grain boundary diffusion mechanism. A significant

dissociation of the GaAs at the interface and Ga out-diffusion allowed Ga accumulation at

the surface. Robinson characterized the Ga out-diffusion as having a low activation

energy.40 As in the case of excess Ni, excess Au also led to more out-diffusion of Ga

which caused greater decomposition of the interface; if too little Ge was present, the Au

or Ni might compensate as p-type dopants.26, 58 Auger studies showed in-diffusion of

Au and Ni and out-diffusion of Ga, which was correlated with a decrease in resistivity

during aging.58 Auger studies also showed that the solid phase reactions were not

complete at the lower temperatures and continued with time and electrical stress, leading

to long-term degradation to the contact resistance.60 Braslau60 also stated that the 500' C

thermal cycle with short alloying times at temperature led to better aging results. That is,

contacts reached their lowest contact resistance for high temperature heat treatment,

especially for short times.26,45 Robinson suggested this was due to Ge penetration of

the dissociating GaAs when the Au-Ge melted and led to an n+ layer and ohmic

behavior.40 He further stated that this was made possible by the out diffusion of the Ga

allowing the formation of Ge doped, As rich GaAs at the interface. The out-diffusing Ga

and in-diffusing Au formed a Ga-Au layer at the surface. The Au and Ga formed a 13-
AuGa phase at heat treatment temperatures below approximately 550' C and an a-AuGa







12
phase at heat treatment temperatures above approximately 600' C.61, 62 The B-AuGa

phase, with a hexagonal crystal structure, was stable above 275' C and ranged in
composition between about 26.5 and 29.2 at.% Ga.63 The a-AuGa phase was an Au-

rich boundary solid solution with Au composition up to 12.5 at. % at 455' C.59 The

spatial distributions of these constituents of the contact were related to the electrical

properties. Ni and As rich grains at the metal-semiconductor interface have been

correlated with lower resistivity, while resistance degradation was correlated with 8-
AuGa and/or a-AuGa phases at the metal-semiconductor interface.49, 61,62

Since the Au-Ge eutectic source materials did not deposit congruently, there were

studies using separate evaporation of the components as well as studies varying the

deposition sequence and both of which showed that these variations had little or no effect

on the resistivity of the ohmic contacts.28 40, 45,49, 64 For example, Robinson grew

contacts with Ni at the metal-semiconductor interface.40 He suggested that the order of

deposition and whether the Au and Ge were evaporated separately, co-evaporated, or

evaporated from a eutectic did not affect the ohmic contact. Thus it was suggested that

the sequence of deposition had little effect on the contact. The sequence of evaporation

did not seem to matter to the final arrangement of constituents after heat

treatment.38, 37, 40, 45, 47,49, 64 The elemental Auger depth profile distributions at the

metal-semiconductor interface were very similar after the metallization becomes ohmic

and similar Ni-Ge-As phases were present as determined by X-ray diffraction, XRD.28

Likewise, Kuan65 used TEM and EDS to show that the as-deposited samples were not

completely intermixed, but that a portion of the Ge separated from the Au even when the

metals were co-evaporated. The lowest resistivity was achieved for a heat treatment of

410' C for 115 sec and he correlated the phase formation of Ni2GeAs with lower

resistance. Initially NiAs and NiGe formed, and the Ni2GeAs was said to have formed

from these two phases. This phase was also identified by Lahav and EizenbergS0, 51

using Auger and XRD, and by Chen et al.53 using TEM to study lateral diffusion







13
samples. It should be pointed out that selected area diffraction, SAD, patterns are very

similar for Ni2GeAs and NiAs. Prolonged heat treatment caused a decrease in the

amount of Ni2GeAs and an increase of both the resistivity and the amount of Au at the

GaAs interface.

Contrary to the above conclusion about layer sequence, several authors have

reported the sequencing arrangement of the components did affect the resistivity of the
contacts.37, 47,61,62 Heiblum et al. deposited 50 A of Ni first and found a correlation

with lower resistance and smaller, dense clusters of NiGe.37 They proposed a high

resistance layer formed under the metal transmission line measurement (TLM), lines.

The sequencing also was found to affect the amount of the Ni2AsGe formed.66 Ni at the

metal-semiconductor interface was also credited with initially acting as a sink for Ge.

Then during heat treatment, NiGe was a source of Ge and a diffusion barrier against too

great of out-diffusion of Ga and As.66,67 Both Murakami et al. and Shih et al. have

reported that the sequencing of the materials does affect the contact resistance.61, 62

They varied the sequence of the metallization and the amount of Ni at the metal-

semiconductor interface and found that 50 A of Ni at the interface caused the onset of

ohmic behavior at lower temperature and that the spread in the measured resistance versus

temperature was also reduced. This behavior was correlated with the formation of a

NiAs(Ge) phase similar to that identified by Kuan et al.64 However, based on Auger

depth profiles, Murakami et al. suggested that the phase was not Ni2GeAs but rather was

a NiAs phase with a small amount of Ge.61 The diffraction patterns for these two phases

are similar. Liliental et al.57 found a Ni-Ge-As phase but the convergent beam electron

diffraction pattern, CBED, was not that of Ni2AsGe. This was in disagreement with the
results of Kuan et al.64

In summary, it is generally accepted that Ge dopes the GaAs when aided by the

presence of Ni in the Au-Ge-Ni metallization. Ni wets the substrate surface to prevent

the irregular contact surface morphology due to Au-Ge balling up. Au provides the







14
material to which the outside leads are attached and forms Au-Ga phases. The contact

interface morphology is very rough, with deep penetration of Au and Ni. Ni and Ge

diffuse to the metal-semiconductor interface and combine with As to form compounds

which appear to be necessary to produce a low resistivity ohmic contacts. Important

compounds have been identified as either Ni2AsGe or NiAs with some Ge (NiAs(Ge)).
Au and Ga formed compounds, such as 13-AuGa and/or a-AuGa which were postulated

to create the vacancies necessary to allow doping by Ge. The compounds also

contributed to the degradation of the ohmic contact when they diffused to the metal-

semiconductor interface. The recommended heat treatment process was for temperatures

near 500' C and for short periods of time. Contrary to earlier literature, the more recent

literature indicates that the sequence of the materials does influence both the final

resistivity and the scatter in the resistivity data. There is disagreement over the

mechanism by which the heavily doped layer of GaAs(Ge) is formed. However, the

disagreement is best illustrated by data from other ohmic contact metallization systems, as

will be discussed in the next section.

As stated earlier, the Au-Ge-Ni ohmic contact literature is not without

disagreements and controversy. Some of this is probably due to the dependence of the

quality of contacts on metal and semiconductor starting materials, and on processing

procedures for substrate preparation, metal deposition, and sample heat treatment. In

many of the papers the substrate cleaning procedures and the contact doping

concentration levels or type were not reported. The absence of this information

frequently makes comparisons difficult. The same concerns exist when comparing the

Au-Ge-Ni ohmic contact literature with the Al-Ge-Ni ohmic contact literature. However,

the vast amount of work on the Au-Ge-Ni ohmic contact system provides a great

background for the subject of this study, the Al-Ge-Ni contact system, when care is taken

and an effort is made to understand the inherent differences.








Other Metals and GaAs


The Au-Ge-Ni literature can be used to explain many of the results observed in

this study. However, there were a number of observations made in other metal systems

which more adequately describe the mechanism believed to cause the conversion to ohmic

behavior in this study. Several of these other metal ohmic contact systems and theoretical

and experimental studies of single metal-GaAs or ternary systems are discussed below.

A number of papers have suggested an alternative mechanism to Ge diffusion to explain

the formation of ohmic contacts. For example, a series of papers on Pd containing

contacts have reported evidence of Pd4GaAs ternary formation at low temperatures that

resulted in regrowth of a heavily doped GaAs layer after higher temperature heat

treatment.31-34 It was proposed that if a thin, highly doped GaAs layer was formed

during ohmic contact formation, it could have resulted either from diffusion of a dopant

into the GaAs or from solid phase regrowth of GaAs incorporating the dopant.34 A

regrowth mechanism has been used to explain the behavior of Pd-Ge and Pd-Si

metallizations deposited onto GaAs. Both contact metallizations have shown conversion

to ohmic contacts began with a limited low temperature (-100' C) reaction to form and

decompose the ternary Pd4GaAs at the metal/GaAs interface. The thickness of the GaAs

reacting with the Pd was uniform and estimated to be 60-150A.31 Ge/Pd/GaAs contacts

have been shown to be rectifying at this stage, 100-225' C, and a correlation between the

onset of ohmic behavior and heat treatment between 225 and 250' C has been

reported.33 However, RBS spectra before and after the onset of ohmic behavior were

nearly identical.33 Subsequent reaction at higher temperatures (-325' C) between the Ge

overlayer and the intermediate Pd4GaAs phase resulted in formation of PdGe phases and

decomposition of the Pd4GaAs phase and the epitaxial regrowth of a Ge doped n+-GaAs

surface layer, GaAs(Ge).34 It was suggested that any excess Ge (beyond PdGe)

migrated across the PdGe layer and was regrown epitaxially on the GaAs substrate.33







16
The Si-Pd system behaved in an analogous manner, but without a Si epi-layer at the

metal/semiconductor interface.31, 32 Enhanced probability of electron tunneling through

the n+ regrown GaAs layer was proposed as the primary reason for ohmic behavior.31-34

A secondary reason suggested for the Ge-Pd system was the small conduction band

discontinuity between the epitaxial Ge layer and the GaAs substrate.31, 32

Holloway et al.29 demonstrated that pure Au contacts on GaAs converted to

ohmic behavior as the result of segregation of Si dopant in areas where GaAs had

decomposed by reacting with the Au overlayer. SIMS data were used to show

segregation of Si from the metal solid solution to the GaAs surface. Ohmic behavior of

Au-Ge-Ni contacts can also be induced by regrowth of GaAs and dopant incorporation as

shown by Li et al.68

The Ni-GaAs system has also been experimentally treated by several authors.

Using TEM, electron diffraction, EDS, and RBS, Sands et al.52 identified the first phase

formed during heat treatment to 480"C to be NixGaAs, where x is -3. They reported Ni

reacts with GaAs during deposition or during TEM sample prep to form a thin layer of

NixGaAs under the native oxide. Equally important, Sands et al.52 noted the ability of Ni

to penetrate thin native oxides to react with the GaAs in this study. The NixGaAs was

formed during their 220' C heat treatment and decomposed about 400" C to NiGa and

NiAs. Sand et al. also demonstrated that regrowth of GaAs occurred when the NixGaAs

phase decomposed.35

Recently several authors have attempted to use multidimensional phase diagrams

to better understand existing ohmic contact literature and to use the basic thermodynamic

principles to explain metallization processing problems. For example, Tsai and

Williams69 concluded that Au and GaAs will not react unless gas phase As species were

allowed to form. In an open system, if As sublimes or crystallizes, psuedobinary phases

between several intermetallic Au-Ga phases can form from reactions between Au and

GaAs. Because the solubility of Ga in Au decreases to near zero at room temperature







17
GaAs should regrow in all instances during cool down. Instead Au-Ga phases are

observed because the As vapor was not given enough time to diffuse back into the solid

during cool down.69 However, utilizing phase diagrams rather than trial and error

experimental methods to determine phases formed may be more useful for ternary

systems than for the larger multicomponent systems. The number of possible reactions

and the number of stable or metastable phases are large in the ternary systems, but much

larger in the multicomponent systems of most ohmic contact systems. Therefore,

predictions quickly become much more difficult without understanding of the

thermodynamics and phase relations.

Beyers et al.70 and Schmid-Fetzer71 have calculated ternary phase diagrams for a

number of metal-semiconductor systems. Beyers et al.70 proposed a classification

scheme for phase equilibria in seven generic elemental metal-GaAs systems. Their

purpose was to describe how phase diagrams can provide a framework for interpreting

previous results and for identifying suitable materials for stable contacts. They compared

one of their generic ternary phase diagrams to that of Ni and the ternary Ni-Ga-As phases

identified by Ogawa,28 Sands et al.35, 52 and Lahav and Eizenberg.50, 51 Because of

uncertainty between bulk versus thin film results, they indicate that either the ternary

Ni2GaAs is a metastable intermediate phase or a low temperature phase with tie lines to

GaAs. Later Schmid-Fetzer71 calculated the ternary phase diagrams for 19 different

metal-GaAs systems using bulk thermodynamic data and an approximation suggested by

a personal communication with Y.A. Chang. Schmid-Fetzer also pointed out the

difficulties arising from inferring phase diagrams from thin film experiments. For metal-

GaAs, the identification of phases in thin films is more ambiguous due to As loss during

vacuum processing. More seriously, the limited amount of metal may exclude the

occurrence of certain phases, thus turning the film thickness into another degree of

freedom.







18
Guivarc'h et al. studied the solid portion of the bulk Ni-Ga-As ternary phase

diagram.72 Their samples were prepared in closed, small volume containers to prevent

formation of gas phase products. They used XRD of powders and single crystals to

determine that neither Ni nor any of 4 possible ternary phases were in thermodynamic

equilibrium with GaAs below -800' C. They showed instead binaries of NiGa, Ni2Ga3,

and NiAs were in thermodynamic equilibrium at lower temperatures. However, cooling

from above 800' C, several Ni rich phases, including ternary phase compositions in

regions of 3 or 2 phase equilibria, likely preceded the formation of the final equilibrium

products. Based on their results NiGa and NiAs phases would be expected, as seen by

others, after 400' to 500' C heat treatment for sufficiently long times.28, 35,50-52 In a

second paper Guivarc'h and Guerin conducted a similar study on molecular beam epitaxy

(MBE) grown Ni thin films on (100) and (111) GaAs.73 These films were grown

simultaneously in the MBE system by indium bonding different substrates to a Mo

sample holder. After heat treating the samples for 1 hour at temperatures from 200' C to

600' C, a series of ternary phases was formed. However, at the 600' C heat treatment,

the binary phases of NiGa and NiAs again formed along with a ternary of

Ni3Gal.5Aso.5.

Care must be taken when referring to the above calculated phase diagrams. Some

of the conclusions were inferred from chemical ratios rather than from diffraction

experiments. Most of the phases identified by Guivarc'h and Guerin72,73 were based on

a single 001 X-ray diffraction line due to the epitaxial nature of their MBE films. As they

pointed out, there were inherent differences in studies of bulk materials grown in a closed

environment and thin film studies carried out in a vacuum or another open system. It

must also be remembered that interdiffusion can and will take place without formation of

new phases. In addition, phase formation in thin films is often determined by kinetics

rather than thermodynamics. Nucleation and growth, diffusion, and short circuit

diffusion paths through intervening layers are just a few of the likely examples of







19
important kinetic effects. In spite of these cautions, the phase diagrams can be used as

guides to determine likely paths for a reaction to proceed. At the very least,

thermodynamics will determine the reaction direction and the final equilibrium position of

the phases, given enough time at temperature.

The previous sections on Au-Ge-Ni and other metals reviewed the general

literature required to better understand the Al-Ge-Ni literature and the observations

reported in this study. The final section of the Review of Literature will give an in-depth

description of the Al-Ge-Ni ohmic contact system.

Background for Aluminum Based Ohmic Contacts to GaAs



As stated earlier, Zuleeg et al. originally proposed the use of an A-/Ge-Ni

metallization system as an ohmic contact for n-type GaAs.l,2 Zuleeg et al. predicted the

Al based metallization would provide a number of benefits, as follows. It would have

improved thermal stability and therefore reliability, especially in higher temperature

applications since the eutectic temperature of Al-Ge (424'C) is higher than that of Au-Ge

(356'C). Since Al has a lower atomic number than Au, it should have higher radiation

tolerance. It could have lower interconnect resistance because evaporated Al has a lower

interconnect resistance than sputtered Au. Better definition of fine patterns would be

possible by dry etching as used for Al metallization on Si Very Large Scale Integrated,

VLSI, circuitry. According to R. Zuleeg, the metallization should be ohmic for both n-

and p-type GaAs. And finally, Al would be less expensive than Au.

The Al-Ge-Ni ohmic contacts studied by Zuleeg et al.1 2 were prepared by

separate evaporation in sequence of 400A of Ge, 300A of Ni, and 2000A of Al onto

Liquid Encapsulated Czochralski, LEC, semi-insulating GaAs substrates which were ion

implanted with Si. The Si content was in the range of 1017 to 1018 cm-3. The ion

implanted substrates were activated by RTA at 825'C prior to metallization. Samples







20
prepared with a TLM test structure were heat treated in a reducing atmosphere of

hydrogen using a graphite strip heater at 500' C for 1 to 30 minutes. Contact resistance

as low as 1.4x10-6 f-.cm2 were observed. Zuleeg et al. also reported the successful

fabrication of a set of 50 JFET (junction field-effect transistor) devices with this

metallization as the first-level interconnect. They proposed the mechanism causing the

metallization to become ohmic was similar to that commonly attributed to the Au-Ge

system, i.e., heat treating above the eutectic temperature caused the Al-Ge to melt and the

liquid state enhanced Ge diffusion into the GaAs. After solidification, this resulted in a

metal alloy n+ semiconductor contact. The Ni was added to prevent balling up of the Al-

Ge similar the Au-Ge based contacts.24 40 The presence of Ni afforded good wetting

properties of the molten Al-Ge and led to smooth surface texture and contact uniformity.

Zuleeg et al. later reported that their specific contact resistance measurement versus

doping concentration correlated with the theory for the field emission (tunneling)
conduction mode (Rc a (ND)-1/2), rather than the mixed mode conduction model

(Rc a 1/ND) which fits the behavior of the Au based contacts.2 They also reported less

resistance degradation during thermal aging at 250' and 300' C for their Al-Ge-Ni versus

Au-Ge-Ni ohmic contacts.

Liliental-Weber et al.20 reported an investigation of the structure and composition

of A/Ge/Ni ohmic contacts by TEM, SIMS, and depth profiling with Auger electron

spectroscopy. Two types of contacts (A and B) were examined in this study. Type A

contacts, fabricated by Zuleeg, were similar to those mentioned above except the layer

thickness of Ge and Ni was 300A each. The type A contacts were on ion implanted

substrates. Liliental-Weber et al.20 reported that Zuleeg had not previously been able to
fabricate Al-Ge-Ni ohmic contacts on LEC substrates that were not ion implanted. These

contacts were heat treated in a reducing atmosphere of hydrogen on a graphite strip heater

at 500' C for 1 minute. The type B contacts were evaporated from Knudsen cells with

A1203 crucibles, using a different layering sequence, onto semi-insulating LEC GaAs







21
without an n-type contact layer. The type B LEC substrates received no chemical etching

or cleaning prior to the deposition of the contact metallization. The metallization layering

was 50A of Ni on 1000A of Al-30.3 at. % Ge, coevaporated at the eutectic ratio,

followed by 300A of Ni, and finally 500A of Al. The type B contacts were heat treated at

500' C in forming gas for 3.5 minutes. These contacts were not characterized electrically

since they were deposited on undoped GaAs. Both types of samples were prepared as

TEM cross-section samples by gluing two pieces, with thin metal layers adjacent, using

silver epoxy (the epoxy was cured by heating at 90* C for 30 minutes to allow

hardening). Mechanical polishing was used to thin to -50 jim, and then Ar' ion milling

with a low energy ion gun and cold stage was used to make electron transparent

samples.

After heat treating the A-type contacts at 500" C for 1min the contact resistance
was 1.4 X 10-6 Q-cm2. TEM analysis of the A-type metal layers, after heat treatment,

showed an increase in thickness and their interface was more uniform than the as-

deposited samples. Primarily Ni5Ge3 and Al3Ni phases were found in contact with the

GaAs. This indicated that the Ni and Ge layers were being dispersed during the heat

treatment and that Al was diffusing toward the GaAs. A hexagonal ternary Al-Ni-Ge

phase was identified as the top layer and the composition was tentatively determined to be

Al3NiGe4 based on elemental ratios from energy dispersive spectroscopy, EDS, of

fluorescent X-rays. EDS also indicated that Ga and As diffused out into the contact, with

-15% more Ga than As. The Ga and As were believed to be in solid solution since no

phases containing Ga or As were found in the TEM investigation, unless new ternary or

quaternary phases containing Ga or As had formed with lattice structures and parameters

close to that of Ni5Ge3 or Al3Ni. SIMS data showed that the Ge penetrated the substrate.

By comparison with EDS data, it was suggested that the Ge concentration was less than

1%, the detection limit for EDS. For type B samples after heat treatment, there was not

an increase in metal layer thickness, but an amorphous metal/GaAs interfacial oxide layer







22
remained intact. The interface of these samples was very flat. A thin layer of Ge, 70A-

150A thick, was just above the amorphous layer, and the next layer consisted of mostly

Al with embedded Ni5Ge3 and Al3Ni grains. Again, Ga and As were found to have

diffused into the metal by both EDS and Auger depth profiling. In this case the Ga to As

ratio was that of the bulk as determined by EDS. It was noted again that the oxide served

as a reaction barrier for GaAs in the type B sample, and this was suggested to be due to

the Ni layer at the interface. It was also suggested that the oxide may also prevent Ge

diffusion into the GaAs. It was suggested that the type A contacts became ohmic due to

the diffusion of Ge into the substrate and substitution on the Ga sites forming an n+

layering in the GaAs. It was further suggested that lack of phase formation between Al

and Ga or As was responsible for extra flat interfaces found in the Al/Ge/Ni ohmic

contact metallization system.

Since the original efforts of Zuleeg and his various co-authors, only one other

group has studied the Al/Ge/Ni ohmic contact system.21-23 This group has written a

series of papers detailing work with the deposition of Al/Ge/Ni contacts onto both n-type

and p-type GaAs with objectives of understanding the contact system and optimizing its

performance.21,22 Research was conducted to correlate the contact microstructure with

electrical performance. Initially Al/Ge/Ni contacts were deposited onto bulk doped GaAs

that were polished, degreased, and chemically etched prior to deposition of the layering

sequence of Zuleeg et al.,12 which was 400A of Ge deposited first, followed by 300A of

Ni and then 2000A of Al. The metals were deposited by electron beam evaporation

through a shadow mask onto n-type (Si @ ND=1018 cm-3) and p-type (Zn @ NA=1019

cm-3) GaAs, then heat treated at 425'C in forming gas for 1 to 4 minutes. The mask

arrangement allowed only qualitative determination of contact ohmicity. Later, in an

attempt to study factors governing ohmic contact formation and to measure specific

contact resistance as a function of contact layer deposition, additional samples were

prepared using less heavily doped n-type GaAs substrates.23







23
In the as-deposited samples, although evaporated separately, the Ni and Ge layers

were found by CBED and EDS to have alloyed to produce a layer (-600A thick)
consisting of a hexagonal Nil.7Ge (Ni5Ge3) phase, unreacted Ge, and possibly other Ni-

Ge phases as well. It should be pointed out that during sample preparation for TEM

cross-sectioning, the temperature may increase to at least 1000 C during the epoxy cure
which could contribute to the diffusion of the as-deposited samples. The authors state

that a temperature of 1000 C was also expected due to the heat of condensation during

metal deposition. Parallel electron energy loss spectroscopy (PEELS) showed that a thin

(10A-50A) layer between the GaAs substrate and the Ni-Ge layer contained oxygen,

presumably the native oxide on the GaAs surface which was not removed during surface

preparation. Columnar Al grains were observed between the Ni-Ge layer and the ambient
interface.

In the first paper, all the contacts were described as lustrous and uniform prior to

heat treatment, but the contacts on p-type GaAs became matte in appearance while the

contacts to n-type material remained largely lustrous after the 425' C heat treatment. All

contacts deposited onto the heavily doped p-type GaAs became ohmic; however, on the

heavily n-type GaAs contacts heat treated for 1-2 minutes remained rectifying while those

heat treated for 3 and 4 minutes were ohmic as determined by the qualitative current-

voltage (I-V) measurements. Following electrical characterization, planar and cross-

sectional samples were prepared for TEM examination. Large microstructural differences

between p-type and n-type samples were reported in the first paper. A thick amorphous

layer with a web-like structure consisting of Al and Ni was found in the p-type ohmic

contacts. EDS of this layer indicated a ratio of 6:1 for Al:Ni. Contrary to these results,

in the second paper the microstructures were reported to be independent of the type of

doping. For all samples, the contact/semiconductor interface was reported to be
extremely flat at the atomic level, while the metal/ambient interfaces were nonuniform.

Graham et al.21-23 cautioned that the TEM results on the surface were thought to be







24
affected by the etching procedure used to remove the GaAs from the back of the contact

and possibly could account for the discrepancy between the first and second reports.

Using TEM and EDS, no Ge was found in the metal-semiconductor interface region of

the heat treated specimens after they became ohmic. Instead the surface became decorated

by a Ge rich growth. EDS identified this floret-like surface precipitate to be primarily Ge

interspersed on a background of Al and Ni. Below this floret precipitant they found a

nearly uniform continuous layer adjacent to the GaAs. The continuous layer at the
interface was identified as mainly an orthorhombic Al3Ni with excess unreacted Ni. The

interface was observed to be extremely flat and uniform without penetration into the

GaAs but with a thin native oxide layer between the metal and semiconductor, similar to

those observed by Liliental-Weber et al.20 Some As and slightly more Ga were found in

the metal after heat treatment

In order to analyze selected contacts with SIMS since the surfaces were non-

uniform due to the florets, the metallization was removed by a chemical etch to improve

depth resolution. A secondary electron microscope (SEM) examination of adjacent

etched and unetched regions of contact viewed edge-on indicated that their procedure did

not remove any underlying GaAs. It was possible for them to successfully etch away

the metallization only from contacts heat treated for the longest times, a fact ascribed to
the complete formation of the Al3Ni phase. Al, Ni, and Ge all were found to have

diffused into the substrate up to several hundred Angstroms, while Ga and As profiles

remained essentially flat. Ni diffused about the same distance in both n and p substrates,

while the Al profile fell off much more abruptly in p-type material. Similarly, Ge

penetrated much further (500' C) in n-type material, with the diffusion depth being

similar to that reported previously by Liiental-Weber et al.20 An interfacial accumulation

of Zn dopant in the p-type GaAs was also found. They stressed that they did not believe

this to be an artifact of the SIMS measurement. They also believed the Al penetration to







25
be real, although they were unable to quantify the depth since calibration standards were

unavailable.

Quantitative I-V measurements of less heavily doped contacts were reported in a

third paper.23 Contacts were deposited onto semi-insulating GaAs substrates with an n-

type layer (Si @ ND=6-10 x 1017 cm-3) grown by MOCVD (metal-organic chemical

vapor deposition). Utilizing photolithography, mesas of conducting material were etched

in epitaxial films, the metallization deposited, and the TLM pattern was produced by a

lift-off process. These samples were made to quantitatively evaluate specific contact

resistance. However, samples with the same intended metallization ratio as previously

reported were incorrectly made without Ge. For n-type contacts, only the sample with a
200A Ge layer was actually made and measured. For ND @ 1017-1018 cm-3 the specific

contact resistance was found to vary from 1.8x10-3 to 3.6x10-5 f.cm2. A correlation

was reported between electrical data and structural properties of the n-type contact. For

shorter than 3-minute heat treatment times, the contacts were rectifying and the layer

closest to the GaAs was a NiGe phase. Contacts heat treated for longer than 3 minutes

were found to be ohmic and the NiGe phases had disappeared while an Al3Ni phase was

adjacent to the GaAs. They therefore attributed the formation of n-type ohmic contacts to

the formation of this Al3Ni phase at the GaAs interface and speculate that in-diffusion of

the contact components, particularly Ge, contribute to ohmic formation.

For the p-type GaAs with NA @ 3.5x1018 to 1.4x1019 cm-3, all contacts were

ohmic as-deposited, the specific contact resistance did not change appreciably with

alloying, the contact resistance decreased with increasing doping but increased with

increasing Ge layer thickness and with increased heat treatment time. The specific contact

resistance varied from 4.0 x 10-4 to 3.6x10-5 Q.cm2. Graham et al. suggested that for

the p-type contact, almost any metallic system would provide a suitable contact to such a

heavily doped material, but also pointed out that the dopant levels studied were about

those used in common p-type devices. Graham et al.22,23 were unable to explain why







26
the metallization functioned as an ohmic contact. However, they did point out a number

of unresolved dilemmas. For n-type systems it is generally said that Ge diffuses into

GaAs to heavily dope the semiconductor to allow charge transport and ohmic behavior by

a tunneling mechanism resulting from band bending. Band bending obviously would be

counter-productive for p-type systems since the bands would be bent in the wrong

direction, making the contact less ohmic. They propose Al diffusion onto vacant Ga

sites, thus forcing the amphoteric Ge to locate on As sites. In this way, the Ge atoms

were either adding themselves to the existing acceptor population or were introducing

themselves in donor-acceptor pairs that do not alter the original band bending at the

surface. Graham et al. further speculate that the smooth morphology of the interface

probably leads to quite a different transport mechanism than that of the Au-Ge-Ni/GaAs

system whose interface is usually characterized by irregular morphologies and

concomitant high-field regions. They state that the Al/Ge/Ni contact system was

originally intended for use on n-type materials, and that on p-type materials the decrease

in specific contact resistance may be due to the amphoteric nature of Ge. Ge may prefer

the Ga site at the low temperature of heat treatment and thus actually compensate the

Zn.43 They attribute negligible electrical effects to the Ni.

To explain their observations Graham et al. made sometimes contradictory

arguments. They initially proposed that Al diffused into the Ni-Ge layer via grain

boundaries and formed an Al-Ge eutectic during the heat treatment. Recall that in as-

deposited contacts, the Ge and Ni layers were said to combine to form a single reacted

layer during the deposition. The eutectic would then melt and in this molten state the Ge

could more easily penetrate the substrate to allow Ge doping. In this way, the GaAs was

doped heavily enough that the Al-Ge-Ni contact can work in the same manner as the Au-

Ge-Ni contacts. Graham et al. attributed the floret structure on the surface to liquid Al-

Ge eutectic formation during heat treatment Thus, during heat treatment Al diffused into
and reacted with the Ni-Ge layer to form the thin polycrystalline layer of Al3Ni and larger







27
single crystals of the same phase. The Ge released was more than sufficient to form a

eutectic liquid with the remaining Al which "balls up" on the contact surface to produce

the observed florett" structures. In the absence of asperities on the contact surface in the
form of the A13Ni single crystals, Graham et al. believe the balling up would probably be

more severe and result in a less uniform contact surface. They express surprise at the

abrupt, flat, uniform, intact interface at the metal-native oxide-semiconductor interface

considering the slight excess Ga observed in the metal. They further suggested that
uniformity of the continuous layer of the Al3Ni at the interface meant that the liquid Al-Ge

eutectic never came in direct contact with the GaAs. They pointed out that one might

expect to observe contact metal-arsenide phases or a poor lattice image due to disorder

immediately under the native oxide, but that neither of these were seen. They offer as a

possible explanation for this that Al diffused into the GaAs during heat treatment and

occupied the Ga vacancies to create a graded layer of AlGaAs at the interface. The layer

would be very thin, perhaps 100A thick, with a low Al content lattice matched to the

underlying GaAs. They further suggested that Al in-diffusion was made possible by the

dissociation of the GaAs in the presence of the Ni at the GaAs surface as reported by

Ogawa.28 However, they were unable to verify this suggestion and their SIMS data

suggested that such an AlGaAs layer did not exist. They then suggested that an

alternative possibility for the very flat interfacial morphology was the apparently strong

tendency for Al and Ni to react and form a single phase alloy rather than Ni causing GaAs

to dissociate, as happens in Au based contact system. In the third paper Graham et al.

pointed out that the earlier explanation in which the eutectic composition became liquid

and led to ohmic contact formation did not appear to be appropriate since the only

evidence of liquid formation was the dendritic decoration at the surface. Furthermore,

EDS data indicated that the surface precipitates were too Ge rich to be entirely the result

of the eutectic solidification.







28
In summary, both Zuleeg et al.1.2 and Graham et al.21. 22 have been able to

produce ohmic contacts on heavily doped n-type GaAs. Both Liliental-Weber et al.20 and

Graham et al.21, 22 found the contacts to be very uniform and flat at the semiconductor-

metal interface. The electrical data of Zuleeg et al.2 suggested that field emission is the

conduction mechanism and that Al-Ge-Ni contacts were stable at elevated temperatures.

Liliental-Weber et al.20 and Graham et al.21, 22 observed Ge-Ni and Al-Ni compounds at

the semiconductor interface utilizing EDS and TEM. These were identified as NisGe3

and Al3Ni. Both groups determined that a small amount of Ge penetrated the GaAs

substrates, but did not find any compounds containing both metals and semiconductor

components.

Thus the Al-Ge-Ni ohmic contact literature is also subject to disagreements and

controversy. Some of this may be due to the dependence of the contact quality on the

starting materials, both metals and semiconductor, and the processing procedures for

sample preparation, contact deposition, and sample heat treatment Similar to the Au-Ge-

Ni contacts, the reactions occurring during the deposition and heat treatment of the Al-

Ge-Ni contacts are apparently quite complex.

On the basis of the Al-Ge-Ni literature, the present study should expect to find

Al3Ni and Ni5As3 as the primary phases formed upon heat treatment of the ohmic contact

metallizations. Since the heat treatments of the contacts in this study were at 425' or

500" C, the ternary phases of Ni-Ga-As described in the phase diagram literature

(NixGaAs) would not be expected. However, NiGa, Ni2Ga3 and NiAs phases were said

to be equilibrium phases after heat treatment of 400-500' C and therefore should form,

although formation of these phases has not been previously reported in the Al-Ge-Ni

contact literature. The Au-Ge-Ni contact literature has identified the key phases of

NixAs(Ge) (e. g. Ni2AsGe or NiAs with some Ge in solid solution), B-AuGa and/or
a-AuGa, which were postulated to create the vacancies necessary to allow doping by Ge.

Based on the work of Sands et al.35 a NixGaAs phase might also form for Ni adjacent to







29
GaAs, a condition used as part of one of the layering sequence in this study. An

analogous NixAs(Ge) phase for the Al-Ge-Ni system would seem possible, but based on

previous works, an Al-Ga phase seems unlikely. Therefore, if the Al-Ge-Ni contact

system is to achieve ohmic behavior by degenerative doping of GaAs, the reactions

taking place during heat treatment must allow Ge incorporation on Ga sites and in

addition to the Al3Ni and Ni5As3 phases previously identified, NixAs and NixGa phases

should be found. In fact, the regrowth mechanism which has been used to explain the

formation of ohmic contacts for several other metallization systems will be used in this

dissertation to explain a dramatic effect in the conversion to ohmic behavior which results

when Ni is deposited onto the GaAs first. The effects of layering sequence upon the

kinetics of ohmic formation have not been previously reported. Further the equilibrium

phases predicted from the phase diagram literature will be correlated with X-ray

diffraction analysis.














CHAPTER 3
EXPERIMENTAL PROCEDURE


The experimental work for this dissertation was carried out at both the Materials

Directorate of Wright Laboratory, Wright-Patterson Air Force Base, Ohio, and at the

Department of Materials Science and Engineering at the University of Florida,

Gainesville, Florida. The samples were grown and the X-ray diffraction was

accomplished at the Materials Directorate, while the heat treatment, electrical

characterization, and the Auger depth profiling were conducted at the University of

Florida.


Sample Preparation


A number of different types of GaAs substrates were selected for this study to test

the effects of bulk doping versus ion implantation over a range of doping levels, and the

influence of the various growth techniques to produce epitaxial doped layers.

Specifically, contacts were deposited on 1) semi-insulating LEC GaAs substrates with

silicon ion implanted active and contact layers, 2) semi-insulating LEC GaAs substrates

with Si doped active and contact layers grown by vapor phase epitaxy, 3) semi-insulating

LEC GaAs substrates with a Si doped contact layer and an arsenic passivation cap grown

by MBE, 4) conductive horizontal Bridgeman (HB) GaAs substrates bulk doped with Si,

and 5) single crystal sapphire substrates. Substrates were implanted with

2.9 x 1017 cm-3 Si to a depth of 0.34 pm as an active layer and a higher doping density
of 1.0 x 1018 cm-3 Si to a depth of 0.16 gm as a contact layer in undoped semi-insulating







31
substrates with a resistivity of > 107 -.cm. The vapor phase epi-substrates were doped

at 2.7 x 1017 cm-3 Si to a depth of 0.63 pm as an active layer and at 1.6 x 1018 cm-3 Si to

a depth of 0.25 pm as a contact layer, also on undoped semi-insulating LEC GaAs

substrates with a resistivity of > 107 fQcm. Both of these substrates were cut with an

orientation of 2' off (100) toward 110) and polished on both sides. The ion implanted

and vapor phase epi substrates were used both as-received and with a mild etch of 1:1:4 -

NH40H: H202:H20 for 20 seconds followed by a rinse with 17 Mi water and blown

dry with filtered dry nitrogen. The horizontal Bridgeman grown GaAs substrates were

bulk doped with 2.6-3.5 x 1016 cm-3 Si and were cut on axis and polished on both sides.

These substrates were subjected to the same sample cleaning procedures.

The quality of devices grown on GaAs substrates can be dramatically influenced

by the effect of substrate processing, packaging, and preparation. Historically a great

deal of effort has gone into the determination of proper sample preparation and packaging

prior to device fabrication. In order to avoid the complications presented by surface

preparations, some of the substrates used in this study had contact layers grown by MBE

with an As overlayer to protect the surface. The As overlayer oxidizes upon exposure to

air, but it passivates the GaAs surface by preventing it from forming a native oxide.

While the actual removal of the As cap layer took place in the metal deposition chamber, it

was a form of in-situ vacuum sample preparation. Thus, the GaAs substrate surface was

not exposed to the atmosphere prior to the metal deposition.

The MBE substrates used in this study were LEC semi-insulating with a

resistivity of >107 -.cm, were cut on axis (100) 0.5', whole wafer etched, and

polished on the top side only. Prior to MBE growth, these wafers were chemically

cleaned with a 6 minute 1:1:5 NH40H:H202:H20 etch, followed by a rinse with 17 MG

water and blown dry with filtered nitrogen. The substrate was then mechanically

mounted, rather than using the typical indium mounting, and loaded into the MBE

system. After the native oxide was thermally desorbed, a 2.0 im layer of Si doped







32
n-GaAs was grown. This surface was then As capped by cooling the wafer to room

temperature in an overpressure of As. Wafers were grown with Si doping levels at the

mid 1018 cm-3 range and also with the Si doping levels in the low 1016 cm-3 range.

In addition to the various types of GaAs substrates, sapphire was also used as an

inert substrate for the metal depositions. The sapphire substrates were used as-received

and chemically cleaned in the same manner as the GaAs wafers. Since the metals, the

semiconductor, and the interface contamination make this a many component system, the

use of sapphire substrates allowed for potentially simpler determination of phase

formation between the Al, Ge, Ni, Ga, and As components. The inert sapphire substrate

will not react with the contact metals making it easier to determine the phases formed

when all the materials were present.


Metal Deposition


All of the metal depositions were carried out in an all metal-sealed stainless steel

ultrahigh vacuum chamber, which was configured to include a Perkin Elmer model

TBNX 3, 300 liter/sec vac-ion pump with a titanium sublimation pump, a Spectramass

Selectorr Residual Gas Analyzer, an Inficon Leybold Hereaus model 751-001-G1 quartz

crystal oscillator, a Thermionics sample manipulator with a home-made sample heater and

de power supply, and a Thermionics three position electron beam evaporator, model 010-

0030, with both beam steering and rastering options.

Chemically etched samples were loaded into the deposition chamber immediately

after they were dried with nitrogen. The GaAs substrates were loaded into the heater

portion of the sample manipulator while the sapphire was loaded into a position on the

sample manipulator which was not heated. The sample manipulator, with the samples,

was then loaded into the vacuum chamber. The sample manipulator was positioned with

the substrates facing up and away from the electron beam evaporator. The vacuum







33
chamber was pumped down using sorption pumps, and the bake-out was begun at this

time using the sorption pumps. After the pressure reached the mid 10-5 Torr range, the

system was switched to the vac-ion pump and baked at approximately 125' C for 18 to

60 hours (overnight to over-the-weekend). Typical background pressure after the bake

and at least an overnight cool down was 2 to 4 xl0-10 Torr. The titanium sublimation

pump was sublimed several times before the metal deposition was begun to provide

additional pumping capacity.

The As capped MBE grown samples were heated as quickly as possible to

approximately 600' C by turning the dc power supply to full power. When the sample

reached 600' C the power supply was turned off, allowing the sample to cool. Due to the

mass of the sample heater and poor thermal contact, it required 6 to 8 minutes for the

sample to reach 600' C and 30 to 40 minutes to return to room temperature. The sample

temperature was monitored by both an IR pyrometer and a platinel thermocouple attached

to the sample holder. The As capped samples had a matte finish as-deposited, but after

desorption at 600" C the sample had a mirror-like appearance. The reflectivity change

was also used as an indicator of desorption. Only the MBE grown, As capped substrates

received the thermal 600' C treatment.

Electron beam evaporation, in principle, is a simple, efficient method of

deposition of uniform pure thin contact layers onto GaAs and sapphire substrates. The

source material was placed in a crucible liner made of graphite to avoid contamination

from both the evaporator hearth and any previously evaporated materials. Then a beam

of electrons was directed onto the source material. In larger production systems, the

material is generally only melted at the point of impingement of the electron beam and

evaporated from a relatively small melted pool at this position. The rest of the material is

heated only through conduction and cooled at the crucible by cooling water circulating

through the evaporator hearth. Therefore the source material acts as its own container.

In the small system used in this project, the entire source load melted. Prior to proper







34
outgassing, the source materials tended to spit if heated too quickly. If the sample was

too close to the evaporator, small balls of the source material deposited onto the substrate.

Thus it was important to slowly heat and outgas each source prior to deposition. For the

Ge and Ni sources this did not pose a problem after the materials were sufficiently

outgassed. However, for the evaporation of Al it did cause major problems. In addition

to its tendency to spit, the Al also broke its crucible during cool down and solidification.

When the Al was evaporated, it melted and wetted the surface of the crucible walls. The

Al would adhere to the walls of the crucible so well during cool down that it would pull

the walls of the crucible in and break them during the solidification process. Slower

heating and cooling, especially through the melting point of Al, was used in an attempt to

keep the crucible from breaking. This slower cool down procedure is a common practice
used in MBE systems to prevent crucible breakage. This approach proved fruitless in the

electron beam evaporator. Eventually it was determined that simply adding more and

more Al to the broken crucible, until it overflowed both the broken crucible and the

hearth, allowed the remaining Al to act as its own crucible. The Al overflowed its

crucible, made contact with the cold hearth of the evaporator and solidified. This

overflow formed into a bowl-like configuration somewhat like the top of a volcano and

behaved much like a larger crucible allowing evaporation of the Al in a more controlled

manner (Figure 1). Another important key in allowing this procedure to work, was the

addition of the beam steering apparatus to the electron beam evaporator. This allowed the

electron beam to be very accurately placed in the center of the Al source.

Prior to deposition of each metal layer, each source was thoroughly outgassed

and a small amount of material, -100A, was evaporated to clean up the source surface.

During outgassing, the sample was left face up in the deposition chamber to prevent any

undesired deposition. Initially the outgassing of the sources was monitored with the

Residual Gas Analyzer, (RGA). The gases were found to be primarily hydrogen, with

lesser amounts of water, carbon monoxide and carbon dioxide. During the outgassing



































Figure 1. The Al source and crucible after many depositions.


the base pressure usually rose for only a brief period (10-15 sec) into the 10-7 Torr range

then quickly fell to the mid 10-8 to the mid 10-9 Torr range during the actual metal

deposition. Following the outgassing of all three sources, the layering was done

sequentially from a three source electron beam evaporator. Thus only single component

layers of Al, Ge, and Ni in various ratios were possible from the pure source materials.

Typical evaporation rates were approximately 0.5-1.5 A/sec for Ge and Ni and

approximately 1.0-5.0 A/sec for Al. The actual deposition of each layer of metal was

monitored using the quartz crystal oscillator placed symmetrically with the sample above

the evaporator. Each of the source materials in their graphite crucibles were moved into

position under the immobile electron gun and evaporated onto the substrates in the







36
desired amount and sequence. On several occasions the evaporation of Al would climb
as high as 10-20 A/sec for brief periods (typically only a few seconds) due to the
difficulties of electron beam evaporation of Al as mentioned earlier. The choice of the
slower deposition rates was the result of trying to keep the evaporation of Al under
control, and also to keep the deposition chamber background pressure as low as possible.
Just prior to the deposition, a movable Ta mask was positioned over the
substrate. The mask was not in intimate contact with the sample in the up position, since
the sliding may have caused scratching and or contamination of the substrate. However,
when the substrate was turned face down for deposition the substrate laid on the mask.
The mask was used to define the three thin lines on one quarter of the sample, see
Figure 2. The electrical lines, -0.5 mm X -4 mm with spacings of -875 pm and
-2150 pm, were later used to electrically characterize the contact via the TLM test
structure.




I I Electrical Line 1
I ] < Electrical Line 2


I | Electrical Line 3





Metallization





Figure 2. Sample metallization configuration.







37
Once the difficulties of deposition were overcome, the samples were then able to

be made routinely. The deposition sequence of elements and the ratio of elements were

chosen as variables. These were simultaneously deposited on both GaAs and sapphire

substrates. For each different ratio of materials at least four samples were made, two for

each heat treatment temperature. Samples were made for two different purposes. First,

sets of samples were made which were intended to become ohmic contacts. Second, sets

of samples were made to help delineate which materials reacted and which phases

formed. There were two types of samples in the first group, one with Ni as the initial

layer on the substrate, and one with Ge as the initial layer on the substrate. Examples of

these types of samples were deposited onto all of the different types of GaAs substrates

and also onto sapphire substrates. The second group of samples were made of

combinations of only two of the Al-Ge-Ni components grown on both GaAs and

sapphire substrates. These were made in ratios of materials similar to those of the ohmic

contacts and also in equal parts of the two components. A list of the samples, their

description, and electrical results can be found in the Appendices A and B.


Electrical Measurements


Electrical characterization was required to determine if the Al/Ge/Ni metallization

deposited upon GaAs had become ohmic. Transmission line measurement (TLM) current

versus voltage (I-V) measurements74 were chosen to provide this determination because

the complete contact system can be deposited onto one side of the substrate. With the one

sided contacts, the deposition process was accomplished without having to break vacuum

and expose the sample surfaces to atmosphere. The metals were deposited, as stated

earlier, through a Ta shadow mask. Therefore, the choice of the TLM method was also

an indirect means to avoid additional contamination of the semiconductor surface. As

indicated in Figure 3, the I-V measurement was made three times on each sample,







38
measuring between electrical lines 1 and 2, 2 and 3, and 1 and 3. I-V measurements
were made on nearly all of the metallization samples.



lto2 2to3 lto3




\\- Electrical Line 1
I IElectrical Line 2


-- Electrical Line 3




Metallization





Figure 3. I-V probe contact to sample electrical lines.


Most of the I-V measurements were made using a system that consisted of a
Hewlett Packard model 3478A multimeter, a Hewlett Packard model 6112A DC power
supply, and a Hewlett Packard 59501B power supply programmer. It was automated
and controlled through an IBM PC with an IEEE-488 computer interface. This was a
table top system with the measurements made in atmosphere at room temperature. Some
of the measurements made on the as-deposited samples were made on a Tektronics curve
tracer oscilloscope with Polaroid photos taken of the curves. Contact to the metallization
was made using tungsten tipped probes. The I-V measurements were made on the as-
deposited samples first, then the samples were heat treated generally in 2 or 3 minute







39
increments, and the heat treatment was followed by another I-V measurement. This

process was repeated until the contact became ohmic or no further change was observed

in the plots of the I-V measurements. Each I-V measurement consisted of three actual

measurements made across electrical lines 1 to 2, 2 to 3, and 1 to 3.



Heat Treatment


The choice of heat treatment temperatures was influenced by the references of
Zuleeg et al.1 and Graham et al.21-23 Zuleeg et al. heat treated their samples at 500* C

and Graham et al. heat treated their samples at 425' C. The 500* C heat treatment

temperature is also a common temperature recommended for heat treating Au-Ge-Ni

ohmic contacts on GaAs for short periods of time.28. 60 It is high enough to allow

interdiffusion of the metal alloyed contacts but low enough to avoid too great a

decomposition of the GaAs. Thus it was thought that if the metallizations on low doped

GaAs were to become ohmic, the 500' C heat treatment temperature was a likely choice.

Graham et al. did not explain their choice of the 425' C heat treatment temperature. It

was chosen for this effort because it was approximately the eutectic temperature of Al-Ge

and should slow any diffusion or intermixing of the metals by more than a factor of two,

according to the rule of thumb, which states that the diffusion distances are usually a

factor of two smaller for each 50' C decrease in temperature. Also, by using both

temperatures the parameters of both of the other groups could be duplicated, and used to

gain some insight into the existing literature.

Prior to heat treatment, the GaAs and the sapphire samples were each split into
two pieces. After the initial I-V measurement, the part of the GaAs with the electrical

measurement lines and a piece of the sapphire were then heat treated. Both samples were

heat treated in a 3-zone furnace with a 10 cm flat zone with a flowing forming gas
atmosphere, 10 % H2 and 90 % N2. One set of samples was heat treated at 425 C and







40
a second set of samples was heat treated at 500" C. In both cases the heat treatments

were carried out in a similar manner. Several samples were loaded into a quartz boat and

slowly inserted into the center of the 3-zone furnace, the elapsed time for insertion was

about 20 to 30 seconds. The samples were then held at the center for a short time,

generally 2 or 3 minutes. The listed heat treatment times were the elapsed time the sample

was held at the center hot zone of the furnace. No correction was made for either heat up

or cool down time. The samples were then slowly removed from the center of the

furnace, the elapsed time for removal was about 30 to 45 seconds. The samples were

removed from the quartz boat and the next I-V measurement was made immediately. The

quartz boat was stored at the end of the furnace between heat treatments. This process

was repeated until the contact became ohmic or no further change was observed in the

plots of the I-V measurements.


Real Time I-V Measurement and Heat Treatment


Because the two heat treatment temperatures apparently affect the contacts

differently it was decided to use a real time heat treatment and I-V measurement to

investigate the time effects. Two different types of the samples were measured in this

second system, which consisted of a Hewlett Packard curve tracer oscilloscope, a tube

furnace, a combination quartz holder and insertion rod equipped with thermocouple and

electrical leads. Both type of leads were spring loaded to make contact with the TLM

pads 1 and 3 on the sample. The output of the I-V and temperature measurements were

sent to the curve tracer and a digital thermocouple meter. Polaroid photos were used to

record the output of the curve tracer. Two samples were taken from the same 3 inch

GaAs. The initial I-V curve of the as-deposited metallization of both samples was that of

a Schottky barrier. This measurement was made on the same non real time system at

room temperature as the other samples. The sample was then mounted in the real-time







41
system, and the initial I-V measurement was also initially rectifying. The temperature at

the center of the furnace was set at 425' C, but 74' C at the end of the furnace. The

sample was left at the end of the furnace approximately 1 minute, then slowly inserted to

the center of the furnace. The changes were so rapid that there was not enough time to

photograph the curves on the oscilloscope. After the sample was observed to exhibit

ohmic behavior, it was slowly removed from the furnace. After removal from the

furnace, an I-V of the sample was again taken on the non-real time system at room

temperature.


Auger Electron Spectroscopv


Auger depth profiling was the technique chosen to investigate the interdiffusion of

the metals and semiconductor both as-deposited and after the sample had become ohmic.

In addition, Auger survey scans were used to characterize the contamination. The

sensitivity of AES is in the range of 0.1 to 1 atomic percent, therefore the technique is

capable of monitoring the alloy interdiffusion and the penetration of the alloy into the

semiconductor. But, AES is not capable of monitoring the degenerative doping of the

semiconductor by the metal alloy. The goal of the surface analysis was to use the Auger

techniques to characterize the movement of these materials in the samples to complement

the XRD measurements, and to correlate these changes with the electrical properties.

Auger electron spectroscopy was performed in a Perkin-Elmer Physical

Electronics Model 660 Scanning Auger Microscope (SAM). Typical electron beam

parameters were a 5 keV primary electron beam energy with a 30 nA beam current. The

data were collected in the N(E) mode with repetitive scans to increase the signal to noise

ratio, summing 20 scans in an elapsed time of approximately 17 minutes. The electron

beam was rastered during data collection to reduce the possibility of beam-induced

changes in the surface such as contamination by electron beam cracking of the residual







42
gases in the vacuum. The electron spectrometer was a cylindrical mirror analyzer. The

system base pressure was typically in the low 10-10 Torr range. For Auger depth

profiling a 3 keV Ar ion gun with a rastered current density of 25 mA was used for

sputtering. The raster size was varied from 2 mm x 2 mm to 5 mm x 5 mm to control the

sputter rate. The differentially pumped ion gun was operated with an Ar pressure of 15

mPa inside the gun. The pressure during sputtering was in the mid to the high 10-8 Torr

in the analyzing chamber. The electron beam was positioned near the center of the

sputtered crater and rastered over a sufficiently small area to avoid crater edge effects.
The system was also configured with a secondary electron detector so that it would

function as a secondary electron microscope (SEM).


Thin Film X-rav Diffraction


X-ray diffraction was chosen to identify the phase formation resulting from the
interdiffusion of the ohmic contact metals and semiconductor. Glancing angle X-ray

diffraction was performed in a computer controlled Rigaku X-ray diffractometer using a
thin film attachment and Cu Ka radiation from a tube operated at 40 kV and 30 mA. The

X-ray source has a line focus, with divergence of the incident X-rays in the vertical

direction limited by the vertical divergence Soller slit, and narrowed by a 0.6* incident

slit combined with a 5 mm incident height limiting slit. The diffracted X-rays from the

thin film sample pass through the receiving Soller slit with a 1" width limiting receiving

slit and without a scatter slit, are again diffracted by a single crystal graphite

monochromator, and are counted by a scintillation counter. Because the thin films yield

low intensities, the data was collected for very long times, up to nearly 3 full days in

many cases. The metal films of Ni and Ge were relatively thin, 500A and 250A

respectively, with an overlayer of Al of 2000A for most of the ohmic contact samples.

Glancing angle or thin film X-ray diffraction was used for the measurement of thin films.







43
It is generally difficult to obtain a distinct diffraction pattern from thin film samples using

conventional X-ray diffraction techniques because the diffracted intensity of X-rays are

very weak and on relatively high backgrounds. The conventional diffractometer utilizes
8 20 scans where the movement of both sample and detector allows the X-ray

incidence angle to be equal to the take-off angle. The angular resolution and the

diffracted intensity are both a function of the take-off angle. The constant take off angle

maintains constant angular resolution and intensity. In the case of thin film X-ray
diffractometer the incident angle, 0, is held constant to maximize the intensity through

maximizing the interaction volume between the incident X-rays and the diffracting
material over the entire 20 scan. The smaller the incidence angle, the greater the

interaction volume and thus the intensity is relatively greater.

Phase identification was accomplished by comparing standard peak position and
intensities from International Centre for Diffraction Data (ICDD) cards75 with the XRD

data collected from the deposited samples. Peak position was the primary source of

identification, but peak intensity was also used for verification where possible. There are

a number of reasons for the measured intensities to vary from those reported on ICDD

cards. For example, a sample with preferred orientation within a metal layer results in

intensity changes and alters the measured intensity values relative to the values given on

the ICDD cards. Poor crystallinity caused by defect structures, small size, or stress will

also alter the measured peak intensity. Most importantly, for this study, overlapping of

peaks from one or more phases will cause large errors. A number of the identified

phases have peaks which overlap, which is common in systems with a large number of

components. Because of the likely occurrence of several of these sources of error as well
as the single crystal nature of the GaAs and sapphire substrates and the possibility of

epitaxy, the samples were run at two different glancing angles to increase the chance of

detection from oriented thin films and substrates.















CHAPTER 4
RESULTS


Sample Preparation


As stated in Chapter 3, the quality of devices grown on GaAs substrates can be

dramatically influenced by the effects of substrate processing. None of the more exotic

methods, such as ozone treatment, and low energy ion sputtering at elevated temperatures

have been used in this study to clean the substrate prior to deposition. Substrates with

ion implanted active and contact layers, substrates with vapor phase epitaxy doped active

and contact layers, HB grown, bulk-doped substrates, and sapphire substrates were used

as-deposited or cleaned with a mild etch. Though sample prep has evolved in an attempt

to reduce the general contamination levels, even today there is disagreement over which

processes are necessary and whether or not the surface states which result from the

contamination can be removed, reduced, or at least passivated. Therefore, because of

concern over potential complications presented by surface preparations, some of the

substrates used in this study had contact layers grown by MBE with an As overlayer to

protect the surface. As described earlier, the thermal desorption of the As cap proved to

occur quite easily. The sample was simply heated until the surface became mirror like.

In one instance two pieces of the same substrate were mounted in the sample

heater to desorb the As cap for later use in a Hall Measurement to check doping

concentration. When the As cap began to desorb, as indicated by a pressure increase in

the deposition chamber, the As cap of only one of the pieces appeared to coalesce then

disappear from the surface completely. When the surface of this piece appeared







45
mirror-like, the heater was turned off. The second of the two pieces apparently was not

making as good as thermal contact with the sample holder as the first piece. At the time

the sample heater was turned off, the second piece of GaAs was still covered with As as

indicated by a matte surface. Its covering of As also desorbed during the cool down and

its surface became highly reflective. The removal of As from the second piece of GaAs

gave confidence that the desorption process was somewhat forgiving and that the surface

reflectivity could generally be taken as an indicator of its removal.

In addition to reducing the potential effects of surface contamination, the MBE

grown samples also allowed a selection of substrates with various doping concentration

levels for the contact layer. To date the Al-Ge-Ni ohmic contact literature only contains

reports for doping concentration levels in the high-1017 cm-3 range and above. To prove

that the Al-Ge-Ni metallization was a viable candidate to become a generally accepted

ohmic contact material for GaAs, the metallization must convert to ohmic when deposited

on the full range of doping concentration levels, from lightly doped (1016 cm-3) to

heavily doped (1018 cm-3 and above). The MBE grown substrates doped with Si in the

1016 cm-3 range were important samples to prove that the AlGe/Ni ohmic contact system

will work at low concentrations.


Sample Deposition


Once the rate of heating of the source materials was under control, the Al source

material was used as its own crucible, and the electron beam steering apparatus was

added to the deposition system, sample deposition was accomplished routinely in
ultrahigh vacuum. Over 75 samples were deposited, but the first 30 were quite variable

while learning how to control the system, to test the various deposition parameters, and

to determine proper analysis parameters. Approximately 45 samples were made after the

deposition system was under control. Samples were made for two different purposes.







46
Samples made in set I were intended to become ohmic contacts. Samples made in set II

were intended to help delineate which materials reacted and which phases formed. There
were two types of samples in set I. The first, sample set la had Ni deposited first on the
GaAs and sapphire substrates with a layering sequence of ambient/Al/Ni/Ge/Ni on GaAs

or sapphire. The second, sample set Ib had Ge deposited first on the GaAs and sapphire
substrates with a layering sequence of ambient/Al/Ni/Ge on GaAs or sapphire. The
metallizations were therefore a series of diffusion couples. Different phases should form
depending on which components were next to one another in the contact layers, how
much of the component was present, and at which temperature they were heat treated.
The starting sequence of the deposited materials were representative of the starting
positions in a multidimensional phase diagram. These sets of samples were therefore

meant to test the hypothesis that the sequence in which the metals were deposited would
influence the resulting phase formation. While other recipes were also used, most of

these samples were made using a total ratio of material of 2000A of Al, 500A of Ni, and
250A of Ge. The total amount of material was held constant for samples in sets la and
Ib, in this way the reactions if driven far enough would proceed to the same end point,

only their pathways would be different. Thus, samples in set Ia were deposited in the
sequence of Al (2000A)/Ni (450A)/Ge (250A)/Ni (50A) and samples in set Ib were

deposited in the sequence of Al (2000A)/Ni (500A)/Ge (250A) on GaAs and sapphire.
This layer identification scheme will be used throughout the dissertation, i.e. the first

metal layer contacts the environment and the last metal layer contacts the substrate. The

substrate was GaAs unless specified as sapphire. This ratio of materials was a slight

variation of the ratio of materials used in the previous Al/Ge/Ni ohmic contact
literature.1, 2, 20-23 It was selected over the other ratios used due Ge out-diffusion,
which will be described later. These types of metallization schemes were deposited onto
all of the different types of GaAs substrates and also onto sapphire substrates. Set I
consists of approximately 20 of the samples. A complete list of the last 45 samples, their







47
description, heat treatment processing, and electrical characterization can be found in the

Appendices A and B.
Samples in set II consisted of combinations of only two of the three metal

Al/Ge/Ni components grown on GaAs and sapphire. These two element layers were

made in ratios of materials similar to those of the ohmic contacts, i.e. Ni (450A)/Ge

(250A)/Ni (50A), as well as samples with layers of equal thicknesses, i. e. Ni (250A)/Ge

(250A)/Ni (250A)/Ge (250A). Since the metals, the semiconductor, and the interface
contamination make this a many component system, the use of two rather than three

component layers on GaAs or sapphire substrates allowed for potentially simpler

determination of the interdiffusion and phase formation. There were approximately 20

samples made for set II. Again, the complete list and description of these samples is in

the Appendices A and B. Further effects of sample deposition will be discussed in the
following characterization subsections.


Heat Treatment


As stated in the Chapter 3, the choice of heat treatment temperatures was
influenced by the references of Zuleeg et al.1 and Graham et al.21-23 Zuleeg et al. heat

treated their samples at 500' C and Graham et al. heat treated their samples at 425' C.

The 500' C heat treatment temperature is also a common temperature recommended for

heat treating Au/Ge/Ni ohmic contacts on GaAs for short periods of time.28, 60 The heat

treatment was accomplished routinely. A complete list of heat treatment times is in

Appendix B. Further effects of heat treatment will be more thoroughly discussed in the
following characterization subsections.







48
Electrical Measurement


As stated in chapter 3, electrical characterization was required to determine if the

Al/Ge/Ni metallization deposited upon GaAs had become ohmic and the TLM type I-V

contacts were deposited onto one side of the substrate without breaking vacuum.

Figure 4 is an example of an I-V curve of the three TLM measurements made on

the as-deposited sample 47. The contact layer for sample 47 (Si @ 5.0x1016 cm-3) was

grown in the MBE system, As capped, and the As cap was desorbed in the vacuum

chamber immediately before deposition of the metal layers with a sequence of

Al/Ni/Ge/GaAs. The three curves in Figure 4 correspond to the three I-V measurements

made on the TLM electrical lines of the sample.



10 + .. 1 ..a 1 I I. I I I I a I I I I a a i a I I-
1 to 2
Sample 47 ,
As Deposited 2 to 3
o *".'** 1 lto3
10 .. .3.....

,.,.."" I..'.






-10
-5 /" ,,..


-10 At n at I us l III at I to a am t to I .... i n I if. me .I t
-6 4 -2 0 2 4 6
Voltage (Volts)

Figure 4. I-V curves of as-deposited metallization Al/Ni/Ge/GaAs, sample 47.







49
Although the I-V measurements indicated some of the metallizations deposited

onto heavily doped substrates were ohmic as-deposited, all of the as-deposited contacts

on lightly-doped substrates exhibited Schottky behavior. The I-V curves of the contacts

indicate back-to-back Schottky barriers which result from the TLM type of measurement

on two front surface contacts, both of which were rectifying. Thus, one metal contact

was a diode under forward bias while the second metal contact was under reverse bias.

The resulting I-V curve indicates rectification under both voltage biases. After the contact

converts from Schottky to ohmic behavior the contact resistance will be independent of

the bias voltage direction, (Fig. 5). The three separate measurements differ in the length

of the current path through the substrate, and this resistance increases with the spacing

between the metal contact pads.




Slto2
Sample 47
M Heat Treated at 500 C 2to3
o 2 Min Increments to 42 Min
S"'to 3





S-20
"



4 -2 0 2 4
Voltage (Volts)

Figure 5. I-V curves after heat treatment of 42 minutes to become ohmic of
metallization Al/Ni/Ge/GaAs for sample 47.







50
After the initial I-V measurement the samples were heat treated in 2 or 3 minute

increments. Sample 47 was heat treated in 2 minute increments at 500' C. After each 2

minute heat treatment another set of I-V data were collected. The sample underwent 21

of these iterations before it became ohmic (total time of 42 minutes). Figure 5 shows the

I-V data from sample 47 after it became ohmic, as demonstrated by the straight lines. As

in the as-deposited sample, the three separate curves indicate the three separate

measurements made between line 1,2, and 3 on sample 47.

The iterative heat treatment and measurement process was originally chosen

because the time required for the contact to become ohmic was unknown. A series of the

I-V curves versus time can be plotted, as shown in Figure 6 for sample 47. To simplify

the graph, only the data between lines 1 and 2 were plotted from each set of I-V

measurements. Data from the other sets of lines would have produced a similar plot.

Note that the current scale has changed for the as-deposited I-V curve from Figure 4 to

Figure 6, because this sample was highly resistive as-deposited but very conductive after

it became ohmic. Thus, the as-deposited contacts appear to be carrying nearly zero

current. The changes taking place in the I-V measurement of this sample, which result

from the increased heat treatment time, were initially very gradual. Little, if any, change

was observed in the I-V curve in approximately the first 15 minutes of I-V measurement

and heat treatment. In the next 15 minutes of heat treatment the I-V curve did begin to

change gradually and over the last 12 minutes of heat treatment the rate of change in the

I-V measurement increased. As shall be demonstrated, the changes in the series of I-V

curves reflect metallurgical changes taking place in the contact metal layers and at the

metal-semiconductor interface.










AI .. . . .


0




a)

U


-40


4 -2 0 2 4

Voltage (Volts)

Figure 6. Series of I-V curves for sample 47, selected electrical measurements
between lines 1 and 2 with increasing heat treatment time.



The time required for the metallization to become ohmic was found to be

dependent on the deposition sequence, particularly whether Ni or Ge was immediately

adjacent to the metal-semiconductor interface. Sample 43 was taken from the same

wafer and MBE growth as sample 47, which was a 3 inch semi-insulating GaAs wafer.

However, the metallization sequence for sample 43 was Al/Ni/Ge/Ni/GaAs rather than

Al/Ni/Ge/GaAs as for sample 47 The metals were deposited as 50A of Ni, 250A of Ge,

450A of Ni, and 2000A of Al. Therefore sample 43 was identical in all aspects to 47

except 50A of Ni was moved to the first layer of metal on the GaAs, rather than the 250A

of Ge on sample 47. The total amount of Ni in sample 43 remains the same as sample

47, a total thickness of about 500A. For all of the samples discussed in this section

(unless otherwise noted) the same ratio of 250A of Ge, 500A of Ni, and 2000A of Al

was maintained.


42 Min

Sample 47 / 40Min
Heat Treated at 500 C /
2 Min Increments to 42 Min / 30Min




.. AsDep & 12Min
# ?
S,
4/

I

J#
4.
I .. .. I I .I I. .







52
Sample 43, like sample 47, was heat treated in 2 minute increments at 500' C,

but it became ohmic after just one 2 minute heat treatment cycle, rather than the 21 cycles

required for sample 47. Because the sample became ohmic in one iteration, only two I-V

curves are shown in Figure 7. The as-deposited I-V data show that sample 43 was

initially less resistive than sample 47.

The influence of the metal sequence on the time required for conversion to an

ohmic contact has not previously been reported for the Al-Ge-Ni ohmic contact

metallization and represents an important finding from this research. For the Au-Ge-Ni

ohmic contact system, Murakami et al.15 reported that placing Ni at the metal-

semiconductor interface reduced the minimum in the specific contact resistance versus

heat treatment temperature, as well as reduced the scatter in these data. However, they

did not report an effect on the time at temperature required for conversion to ohmic

behavior.





40-
Sample 43 2Min
Heat Treated at 500' C
(?, One 2 Minute Increment .







-40 -

.0 .. .... .. ..... .... .. ,. .. ... ...,, l ... D, .
-40

-4 -2 0 2 4
Voltage (Volts)
Figure 7. Series of I-V curves with increasing heat treatment time for Sample 43,
electrical measurements between lines 1 and 2.









Table 1. Metallization of GaAs.


Sample Dopant/Concentration Interface Heat Treatment I-V Results
Number Metal Temp Time

19 SS 1.X10E18-.16um 300A Ge 425 3+2+2+3+3 Ohmic after 13 Min
2.9X10E17-34um

20 SI 1.0X10E18-.16um 50A Ni 500 1+2 Ohmic after 3 Min
2.9X10E17-.34um

25 SI 1.1X10E18-.16um 251A Ge 425 2+2+2+2 Ohmic after 20 Min
3.0X10E17-34um 3+3+3+3

26 SI l.IX10E18-.16um 51A NI 500 2 Ohmic after 2 Min
3.0X10E17-34um

27 Si 1.1X10E18-.16um 251k Ge 500 3+2+2+2 Ohmic after 9 Min
3.0X10E17-.34um

28 S 1.1X10E18-.16um 52A NI Ramped Real Time Turned Ohmic quickly
3.0X10E17-.34um to 425 at -180 C

29 SI 1.6X10E1-.25um 251A Ge Ramped Real Time Ohmic after ~15 Min
2.7X10E17-.63um -20 Min to 425

30 S 1.6X10E18-.25um 51A NI 500 3 Ohmic after 3 MIn
2.7X10E17-.63um

39 SI ~x10E16-2.0um 52A NI 425 2 Lines 2 Min 2Min & 50 Mi
3rd lane 50 Min 2 ea to 40+10 to SO

43 Si -5x10E16-2.0um 51A Ni 500 2 Ohmic after 2 Min


45 SI -5xl0E16-2.0um 252A Ge 425 2 + 2-= 40 Almost, Not Quite
then +10 to 100

47 SI -Sx10E16-2.0um 254A Ge 500 2 ea to 44 Min Ohmic after 42 MIn


The elapsed time required to form ohmic contacts was also influenced by the heat

treatment temperature. The metallizations which were Schottky as-deposited and became

ohmic after heat treatment are listed in Table 1. The samples made early in this study did

not have the same thickness ratios as samples numbered with 25 or higher (as reported

above). Therefore caution must be used in the comparisons of the time it took for

samples numbered less than 25 to convert from Schottky to ohmic behavior.








54


(W,
0


20 Min 14 Min
Sample 25 //, 11 Min
Heat Treated at 425 / 4 Mi
Electrical Lines 1 to 2 /, /, .
\ // .As Dep




Series of I curves with increasing heat treatment time at 425Cfor


















9 Min
93 Min
4 .. .. I . I I.. .

-2 0 2 4

Voltage (Volts)

Series of I-V curves with increasing heat treatment time at 425' C for
sample 25, electrical measurements between lines 1 and 2.



"9Min
Sample #27 / 7Min "
Heat Treated at 500 C 5 Min
Electrical Lines 1 to 2 y ..).-













-2 -1 0 1 2
Voltage (Volts)

Series of I-V curves with increasing heat treatment time at 500' C for
Sample 27, electrical measurements between lines 1 and 2.


Figure 8.





40


0
U,
o






U
1,,,=

s^


Figure 9.











Samples 25 and 27 were taken from the same 3 inch wafer, a LEC semi-

insulating GaAs wafer with an ion implanted active layer (2.9x1017 cm-3 Si to a depth of

0.34 gm) and an ion implanted contact layer (1.0x1018 cm-3 Si to a depth of 0.16 Ipm).

Both samples also have the same metal deposition sequence of Al/Ni/Ge/GaAs, but

sample 25 was heat treated at 425' C for four 2 minute increments then for four 3 minute

increments, totaling 20 minutes. The I-V profiles of sample 25 are shown in Figure 8.

Sample 27 was heat treated at 500' C for one 3 minute increment and for three 2 minute

increments, totaling 9 minutes. The I-V profiles of sample 27 are shown in Figure 9.

The slope of the I-V plot of sample 25 was approximately 60% of that of sample 27 (i. e.

sample 25 was more resistive). Thus to achieve a similar conversion to ohmic behavior

requires slightly more than a factor of 2 longer time at 425' C than at 500' C. This

clearly suggests that a thermally activated rate limiting mechanism was controlling the

conversion from Schottky to ohmic behavior for the samples which have Ge adjacent to

GaAs.

An effect due to changing from 425 to 500' C was not detected for samples with

Ni adjacent to GaAs. The conversion from Schottky to ohmic behavior generally

occurred in just a few minutes (refer to Table 1) when heat treated at either 425' C or

500' C. However, the real time I-V data reported below show that a temperature

activated process is operative in conversion to ohmic behavior. There is clearly a

difference in the mechanism or rate limiting step with Ni versus Ge adjacent to the GaAs.

The time required for the metallization to convert from Schottky to ohmic
behavior was also dependent on the doping concentration of the contact layer. This effect

was also observed only in samples with Ge adjacent to GaAs. For this comparison it

was not possible to use the same GaAs wafer as the substrate for both samples, since the

contact layer and its doping concentration were an integral part of the substrate. The

series of I-V curves of samples 27 and 47, in Figures 9 and 6 respectively, show the







56
effect of substrate doping. Both samples were heat treated at 500' C and both samples

had approximately 250A of Ge as the first metal layer at the metal-semiconductor

interface. The more heavily doped sample 27, with a Si ion implanted contact layer of
1.0x1018 cm-3, required 9 minutes to convert from Schottky to ohmic behavior. Sample

47, with a Si MBE grown contact layer of 5.0x1016 cm-3 required 42 minutes to make

the same conversion, a difference of nearly 5 times. Since the only major difference in

these samples was the doping concentration of their contact layers, it was reasonable to

attribute the difference in conversion time to the time required to create a Ge degenerative

doped surface layer, probably due to Ge doping. Incorporation of Ge in the contact layer

of the GaAs, as stated earlier, was a key part of the generally accepted mechanism for the

formation of an ohmic contact from metal alloys such as Al-Ge-Ni or Au-Ge-Ni.


Real Time I-V Measurement and Heat Treatment


Because heat treatment temperatures apparently affected contacts differently

depending upon whether Ge or Ni was adjacent to the GaAs, real time I-V measurements

were used to investigate the effects of time. Both of the samples (28 and 29, see the

appendices) used in this experiment became ohmic. The initial I-V curve of the as-

deposited metallization of both samples was that of a Schottky barrier, as measured on

the same system as samples measured iteratively. Both samples for the real time studies

were also taken from the same 3 inch GaAs as samples 25 and 27, discussed above. The

first sample, sample 28, had a layer sequence of Al/Ni/Ge/Ni/GaAs and the as-deposited

I-V data are shown in Figure 10. After these data were collected, the sample was

mounted on a quartz holder for insertion into the heat treatment furnace. The holder was

equipped with thermocouple and electrical leads spring loaded to make contact with the
TLM pads 1 and 3 on the sample.









A-l.


V V--


-40
4


Figure 10.


Sample 28
As Deposited






I'"


, I I I I| i I, I | I, I -I

lto2
S2to3
i l1 to3


.33. mall... .3.... I***l* 1....


-2 0 2

Voltage (Volts)
Room temperature I-V measurement of sample 28, as-deposited.


The I-V measurement made in the real time measurement system was also initially

Schottky. As sample 28 was slowly inserted toward the furnace hot zone the I-V curve

changed very quickly (less than 1 minute), the curve began to change at approximately

110' C. The changes were so rapid that they could not be photographed on the

oscilloscope. As the sample was inserted further into the furnace, the temperature

continued to rise and the I-V curve became linear at 150' C and less resistive as indicated

by greater current at the same voltage (i. e. an increase in the slope of the I-V curve). By

180' C the changes in the I-V curve slowed dramatically, although the change in the

slope of the I-V curve continued to slowly increase to approximately 300-325' C; by

380' C the slope had stopped increasing. The sample was inserted into the center of the

furnace (425' C) and left there for approximately 2 min. Only a slight decrease was

observed in the slope of the I-V curve during this time. As the sample was slowly

removed from the furnace, the slope continued to decrease slowly. The entire experiment







58
occurred in 4 to 5 minutes. The I-V curve at room temperature after heating in the real

time system was very similar to the curve shown in Figure 11, for sample 29.


UI


60 .. i I I i I P I W I I i I i 4-
Sample 29
After Real Time Heat Treatment and I-V 1 to 2
40 / .2to 3


20 --
2, .-
/ .*/

20 ,





40 1-


-60



Figure 11.


'. I I . I . I .. ....1
4 -2 0 2 4
Voltage (Volts)
I-V measurement of sample 29 after the real time I-V measurement and
heat treatment.


The second sample (sample 29) had a layer sequence of Al/Ni/Ge/GaAs. The

initial I-V curve was Schottky as-deposited. This sample was moved toward the hot

zone of the furnace more slowly and the I-V curve changed much more gradually than

sample 28, consistent with the previous results for samples with Ge at the metal-

semiconductor interface. Approximately 7 minutes elapsed before it reached a

temperature of 275' C with little or no change in the I-V curve. At 275' C the breakdown

voltage and the effective resistance decreased slowly. At a temperature of 340' C and an

elapsed time of 10 minutes, the I-V curve became slightly asymmetric. By 350' C and

an elapsed time of 12 minutes, the I-V curve was again symmetric with further reduction







59
in the breakdown voltage and effective resistance. The sample did not become ohmic

until its temperature reached 430' C after approximately 15 minutes. Shortly after the

sample became ohmic, electrical contact to this sample was lost. It was heat treated to a

total of 20 minutes, then removed from the furnace. Of the 20 minutes, insertion of the

sample to the center of the furnace consumed 15 minutes. The changes observed in the

I-V curves are very similar to the series of I-V curves of sample 25, shown in Figure 8.

The room temperature I-V curves for sample 29 after heating are shown in Figure 11.

The results from both samples 28 and 29 were completely consistent with the

results from the samples iteratively heat treated and electrically characterized, even the

observation of the I-V curves becoming asymmetric. In addition, the amount of heat

treatment time required for the samples to become ohmic during the two different types

of heat treatment was equal to within experimental error when the heat up and cool down

for the iterative heat treatments were considered. The real time I-V measurement and heat

treatment do show, however, that the sample with Ni adjacent to GaAs began to exhibit

ohmic behavior as low as 180' C and had reached its lowest resistivity by approximately

325' C.


Summary of Current-Voltage Data


In summary, the electrical characterization has been used to demonstrate dramatic

differences in the amount of heat treatment time and temperature required to convert Al-

Ge-Ni from Schottky to ohmic behavior, depending on whether Ge or Ni was at the

metal-semiconductor interface. The times and temperatures needed to become ohmic

were also found to be dependent on the doping concentration of the contact layer. For

samples with Ge at the metal-semiconductor interface, the time required to convert from

Schottky to ohmic behavior took longer with lower doping concentration and with lower

heat treatment temperature. For samples with Ni at the metal-semiconductor interface, the







60
time required to convert from Schottky to ohmic behavior was much less and about the

same for both high and low doping at either 425 or 500' C. As indicated by the real time

I-V measurement during heat treatment, the samples with Ni at the metal-semiconductor

interface show ohmic behavior at lower temperature, similar to reports by Murakami et

al.15 A possible explanation for the longer times required to convert from Schottky to

ohmic behavior for the low doped substrates with Ge at the metal-semiconductor

interface, may be the time required for enough Ge incorporation into the GaAs to

degenerately dope the contact layer n+. The contacts with Ni at the metal-semiconductor

interface; however, convert from Schottky behavior to ohmic behavior much faster. It

has been reported that Ni dissociates GaAs.27, 28 Thus, a more likely explanation for

degenerative doping for both types of samples is that Ni dissociates GaAs and causes

faster incorporation of Ge. For the samples with Ge at the metal-semiconductor

interface, the rate limiting step may simply be the time required for diffusion of Ni

through the Ge layer to the metal-semiconductor interface. Further details on elemental

profiles and phase formation are given in the next section.

The electrical lines defined by the Ta mask were intended to provide only a

qualitative measurement of electrical performance, i.e. to determine when or if the Al-Ge-

Ni metallizations converted to ohmic behavior. The emphasis of this study was to

describe the metallurgical effects of changing the layering sequence. This study has not

attempted to optimize the elemental ratios or the heat treatment processing to produce the

lowest resistivity ohmic contact. However, it was possible to make an estimate of the

contact resistivity by making several assumptions, all of which have deleterious effects

on the accuracy of the measurement and would lead to overestimating the specific contact

resistance. For example, the probe contact resistance was assumed to be negligible, the

sheet resistance of the interface layer was assumed to be equivalent to the sheet resistance

of the metal contact, and the contact spacings were assumed to be uniform from sample to

sample and equal to the mask spacings; all of these assumptions are well known to cause









Table 2. Specific contact resistance estimate, assuming R = Rsc.


Sample Specific Contact Interface Heat Treatment
Number Resistance Metal Time Te
ohm.cm-2 Min. c
19 2.8 x 10-2 Ge 13 425
20 1.7 x 10-1 Ni 3 500
25 3.9 x 10-1 Ge 20 425
26 2.6 x 10-2 Ni 2 50
27 3.7 x 10-2 Ge 9 500
28 1.1 x 10-1 Ni -5 425
29 7.4 x 10-1 e 20 425
30 2.2x 10-1 Ni 3 500
39 5.0 x 10-2 N 2-50 425
_43 8.8 x 10-2 Ni 2 500
47 2.5 x 10-6 Ge 44 500


inaccuracies in the specific contact resistivity. The calculated values are listed in Table 2.
The calculation of specific resistivity is made from the equation rc = Rc2W2/Rsc
(rc = specific contact resistivity, Rc = contact resistivity, Rsc = sheet resistivity of the
metal contact/semiconductor interface layer, W = length of the contact). Values for Rc

and Rsc/W are taken from plots of the resistance calculated from the I-V curves versus the
contact spacing, Rc/2 is the y-intercept and Rsc/W is the slope. R. E. Williams19 has
succinctly described of the TLM measurement technique and typical measurement errors.
The large range specific contact resistance (10-6 to 10-1) in Table 2 probably indicate
random error in measurement in addition to the systematic errors associated with contact
probe resistance and contact spacing. Minor errors in measurement of R or spacing can
cause large errors (orders of magnitude) in contact resistance. The measurements would
also be improved by using a wire bonded to the contact rather than contact probes and by
more accurate placement of the line spacings.
















Auger depth profiling was used to investigate the interdiffusion of the contact

metals and the semiconductor as-deposited and after heat treating to form an ohmic

contact The Auger survey scans of the as-deposited metals indicated the surface layer of

the metal has the usual contamination from O and C of surfaces exposed to the

atmosphere. The Auger survey scan of sample 20, shown in Figure 12, is typical for all

of the as-deposited metallized samples. These surface contaminants were quickly

sputtered away (-2 minutes) during the Auger depth profiling. The metal films contained

very little or no C or O below the sample surface. A few of the early metallized

substrates were not cleaned and they had C and O at the metal-semiconductor interface.


3000

2000

1000

0

-1000

-2000

-3000


200 400 600 800 1000 1200 1400 1600 1800 2000
Kinetic Energy (eV)


Figure 12. Auger survey scan of sample 27, as-deposited.


Elemental Depth Profiles







63

I I I I I L,' I I I '
"o 24 -
< 22 Sample 27 Ni
As Deposited .,
520- AAl-Ni-Ge on GaAs
18
16
14 Ni
14
12
10 Al Al / G
8 wk -
6 Fe Ge As


[ mw a
0 400 800 1200 1600 2000 2400 2800 3200 3600 4000
Sputter Time (Sec)

Figure 13. Auger depth profile of sample 27, as-deposited, Al/Ni/Ge on GaAs.


Auger depth profiles appear to indicate a slight interdiffusion between Al and Ni,

and a very extensive interdiffusion between Ni and Ge in the as-deposited condition.

This is shown in Figure 13 for sample 27, which consists of Al (2000A)/Ni (500A)/Ge

(250A)/ion implanted GaAs. There appears to be some in-diffusion between Ni and Ge

with GaAs, with little out-diffusion of the Ga or As into the metal layers. The suggestion

from Auger depth profiles of interdiffusion of Ni and Ge agrees with the results of

Liliental-Weber et al.20 and Graham et al.21, 23 who reported interdiffusion and Ge-Ni

phase formation in their electron beam as-deposited samples. However, they also report

out-diffusion of Ga and As into the metals, which was not observed in this study.

There was no observable difference in the Auger depth profiles of as-deposited

contact metallizations with and without the 50A of Ni at the metal-semiconductor

interface. Sample 26 had the Ni at the interfacial layer and only limited interdiffusion of

the Al and Ni is seen in Figure 14. The depth profiles for samples 26 and 27 are very







64


0 22 Sample 26 i
S20 As Deposited
S Al-Ni-Ge-Ni on GaAs
18 -
S16 \

0 Ni;
12
10 J Ga



4 '


0 1000 2000 3000 4000 5000
Sputter Time (Sec)

Figure 14. Auger depth profile of sample 26, as-deposited, Al/Ni/Ge/Ni on GaAs.


similar with respect to interdiffusion as found by comparing data in Figures 14 and 13,

respectively. It was apparent however, that the Ni and Ge profiles overlap slightly more

in sample 26 versus sample 27. This was attributed to Ni being at the metal-

semiconductor interface and the interdiffusion of Ni into the Ge from both directions for

sample 26. There may also be deeper Ni penetration into the GaAs substrate in sample

26. Note that the sputter time for sample 26, shown in Figure 14, is much greater than

for sample 27, shown in Figure 13 because the ion beam raster size was increased for

sample 26 to sputter at a slower rate and increase the depth resolution. The difference in

the sequence of the metals at the substrate interface in the as-deposited contact samples as

well as the claims of interdiffusion cannot be easily discerned from these Auger depth

profiles due to intermixing during the deposition during the sputtering process. The ion

sputtering process can also cause ion mixing and ion knock-on.76 Therefore the

interdiffusion must be corroborated with another analytical technique.







65
Contamination by Fe was detected on many of the samples. The Fe was always

associated with the Ni layers, and is known to be a frequent contaminant of even highly

refined Ni. The Fe concentration was 10 15% of the Ni layer based on quantitation

using Auger peak-to-peak values and handbook sensitivities.77 However, the depth

profiles show that the Fe did not interdiffuse in the metal layers as much as the Al, Ni,

and Ge and therefore probably did not modify diffusion between and phase reaction of

the contact metals with the semiconductor. Fe in the GaAs is a deep level electron trap

which would impede charge transfer required for ohmic conduction and make the contact

more resistive. Because of the lack of diffusion and the fact that ohmic contacts were

observed, the Fe is not believed to have an important effect on the behavior of this contact

metallization system.

Heat treated samples showed varying degrees of further intermixing of metal

layers, the degree of which correlates with temperature and time. Nearly complete

interdiffusion of the metal layer components was observed for samples heat treated at

higher temperature or for longer periods of time at the lower temperature, while less

interdiffusion was observed at lower temperatures for shorter times. For example,

sample 27 was iteratively heat treated at 500" C for 9 minutes total. From the I-V data in

Figure 7, sample 27 converted from Schottky to ohmic behavior between 7 and 9

minutes. The Auger depth profile in Figure 15 shows that the contact metals have

completely interdiffused after 9 minutes at 500' C. The Al appears to have completely

intermixed with the Ni and Ge layers and was present at the metal-semiconductor

interface. While reaction of Al to form an AlGaAs layer was suggested by Graham et

al.22,23 as a possible explanation of the conversion to ohmic behavior, no evidence to

support such a mechanism was found in previous studies or this study. The Ni appears

to have diffused towards and away from the metal-semiconductor toward the surface and

is distributed throughout the metal contact. The Ni and Ge penetration into the GaAs

does not appear to be much different for sample 27 before or after heat treatment. While







66

I24 I I I I I
22 Sample 27
0 O Al-Ni-Ge on GaAs
Heat Treated at 500 C
c 18 3+2+2+2 Min
16
14 -
Ni
12 12
Ni
10 Ga
8 -

C Ge Fe
4 Al Al
0

0 500 1000 1500 2000 2500 3000
Sputter Time (Sec)
Figure 15. Auger depth profile of sample 27, after 500' C heat treatment for 9
minutes total time, Al/Ni/Ge on GaAs.


the Ga and As appear to have interdiffused with the metal layer slightly more than in the

as-deposited samples, no Ga or As were detected at the ambient interface of in the metal

layers. This indicates no out-diffusion occurred for either element. The Ge is

concentrated both at the metal/semiconductor and at the metal/ambient interfaces. This

depth profile was representative of the depth profiles for all of the samples heat treated at

500' C.

Samples heat treated at 425' C exhibited less interdiffusion for time shorter than

15 minutes. The Auger depth profile of sample 19 (Al (2000A)/ Ni (400A)/ Ge (300A)/

GaAs) was heat treated at 425' C for a total time of 13 minutes in increments of 3, 2, 2,

3, and 3 minutes, as shown in Figure 16. Diffusion of Ni into both Al and Ge is

evident, but no Ge was detected at the ambient interface. Interdiffusion of the metal layer

components was observed for samples heat treated at higher temperature or for longer

periods of time at the lower temperature, while less interdiffusion was observed at lower












20 Sample 19
S Al-Ni-Ge on GaAs Ni
Heat Treated at 425 C
3 + 2+2+3+ 3 Min
Ni

< 10- Al A1
I r





0 500 1000 1500 2000
Sputter Time (Sec)

Figure 16. Auger depth profile of sample 19, after 425' C heat treatment for 13
minutes total time, Al/Ni/Ge on GaAs.


temperatures for shorter times. For example, sample 27 was iteratively heat treated at

500' C for 9 minutes total. From the I-V data in Figure 7, sample 27 converted from

Schottky to ohmic behavior between 7 and 9 minutes. The Auger depth profile in Figure

15 shows that the contact metals have completely interdiffused after 9 minutes at 500' C.

The Al appears to have completely intermixed with the Ni and Ge layers and was present

at the metal-semiconductor interface. While reaction of Al to form an AlGaAs layer was

suggested by Graham et al.22, 23 as a possible explanation of the conversion to ohmic

behavior, no evidence to support such a mechanism was found in previous studies or this

study. The Ni appears to have diffused towards and away from the metal-semiconductor

toward the surface and is distributed throughout the metal contact. The Ni and Ge

penetration into the GaAs does not appear to be much different for sample 27 before or

after heat treatment. While the Ga and As appear to have interdiffused with the metal

layer slightly more than in the as-deposited samples, no Ga or As were detected at the







68
ambient interface of in the metal layers. This indicates no out-diffusion occurred for

either element. The Ge is concentrated both at the metal/semiconductor and at the

metal/ambient interfaces. This depth profile was representative of the depth profiles for

all of the samples heat treated at 500' C.

In the early samples, Ge was seen quite distinctly to form dendritic-like florets on

the surface of metal on either GaAs or sapphire substrates. Florets were observed on

samples with a high Ge concentration (e. g. a Ge/Ni ratio of 3:4), but not on samples

with a lower Ge/Ni ratio (e. g. 1:2) and did not depend on whether Ge or Ni was adjacent

to the substrate. An SEM image of sample 20, heat treated at 500' C for 3 minutes

(Figure 17) shows the Ge florets. Precipitates of Ge were observed by Graham et

al.22,23 for Al-Ge-Ni ohmic contact samples with similar Ge to Ni (1:1) ratios, and have

been reported for excess Ge in the Au-Ge-Ni contacts.28 44-46 After Ar+ sputtering to

remove surface contamination, Auger analysis showed that the florets were pure Ge.


Figure 17. SEM micrograph of sample 20 after heat treatment at 500' C for 3 min.







69
Samples with various Al-Ge-Ni sequences deposited on sapphire were also depth

profiled before and after heat treatment. Their Auger depth profiles were very similar to

those of the contacts grown on GaAs. For example, the depth profiles of those samples
heat treated at 500' C indicated the same interdiffusion. As stated previously, even the
Ge out-diffusion and floret growth on the surface was observed for a 3 to 4 ratio of Ge to
Ni.

Two Element Metallizations

To better understand intermixing, samples without the outer 2000A layer of Al
were Auger sputter profiled. The 1 to 2 ratio of Ge to Ni [Ge(250A)/Ni(500A) or
Ni(50A)/Ge(250A)/Ni(450A)] was grown on both GaAs and sapphire. Depth profiles of
as-deposited samples indicated that the metals did not interdiffuse as much when Al was
present The Auger depth profile of sample 34 [Ni(450A)/Ge(250A)/Ni(50A)/sapphire]
shown in Figure 18, and the 50A layer of Ni is easily observed. This was also true for
similar metallizations on GaAs. Auger depth profiles of samples with Ge adjacent to the

substrate indicated the Ni has penetrated the Ge layer, but not nearly as much as indicated
in the depth profiles of contacts with all three elements. This result was consistent with

the difference in electrical characterization results for heat treatment times required to
convert the samples from Schottky to ohmic behavior, which indicated a major difference

with and without Ni at the GaAs interface. The two component system on sapphire show
the Ni/Ge/Ni layers did not completely intermix during deposition.
In another type of the 2-element sample, two of the three metals were deposited in
four alternating layers of 250A each (i. e. a 1:1 ratio) onto both GaAs and sapphire.
Auger depth profiles of the as-deposited samples of Ni and Ge and similar samples of Al
and Ni deposited on both GaAs and sapphire appeared to be layered with some
interdiffusion between the layers. While the three layers of Ni/Ge/Ni for sample 34 were

obvious, as shown Figure 18, the four layer samples of Al/Ge/Al/Ge/GaAs appeared to





























100 200 300 400 500 600 700 800 900

Sputter Time (Sec)


Figure 18. Auger depth profile of sample 34, as-deposited, Ni/Ge/Ni on sapphire.


200 400 600 800 1000


1200


Sputter Time (Sec)


Figure 19. Auger depth profile of Sample 13, as-deposited,
250A each, on GaAs.


AVGe/AVGe,


"0
1-
X

m
Y-

;
3










have rearranged themselves as indicated by the Auger depth profile of sample 13, shown

in Figure 19. The Auger depth profile would suggest that the four layers have rearranged

with Ge diffusing to the substrate and Al diffusing to the ambient during deposition in the

Al/Ge samples. The Auger depth profile of the same metallization deposited onto

sapphire during the same deposition, also exhibited this rearrangement.

The 2-element 4 layer samples on GaAs and sapphire were heat treated at 500' C

for 4 minutes. The Auger depth profiles of the Ni and Ge samples deposited on both

GaAs and sapphire indicate complete interdiffusion. The Auger depth profiles of the Al

and Ni samples deposited on GaAs and sapphire were different, with limited

interdiffusion on GaAs but complete on sapphire. The depth profiles of Al and Ge

deposited on both GaAs and sapphire after heat treatment appear to be dominated by their

original surface morphology.

Summary of Elemental Depth Profiles

Auger survey scans have been used to demonstrate that the surfaces of the

samples exposed to the atmosphere were contaminated with O and C. The Auger survey

scans were also used to identify surface florets on high Ge:Ni ratio samples as being pure

Ge. At Ge:Ni ratios 2, the layers remain uniform; however, if the ratio of Ge:Ni was

~3:4, the contact layer became laterally nonuniform, similar to the Au-Ge-Ni contacts.44

Thus the ratio of Ge to the other materials is important with excess Ge causing out-

diffusion and precipitation on the surface during heat treatment.28, 44-46 Crouch et al.45

have reported a dendritic growth precipitating to the surface which they identify as a 1 to

1 Ge to Au for Au-Ge contacts heated by RTA. Precipitation of Ge rich compounds on

the surface of the sample was shown to be controlled in this study by reducing the ratio

of Ge to Ni.







72
Although the depth profiles of the thicker, complete contact samples must be

interpreted with caution, the Al and Ni layers and the Ni and Ge layers appear to exhibit

some interdiffusion both as-deposited and heat treated. The difference in layering

sequence of the Ni and Ge layers was only evident in the Auger depth profiles without

the thick Al layer due to artifacts causing poor depth resolution in the sputter profiles.

However, intermixing during deposition is in agreement with previous works on both the

Al-Ge-Ni and the Au-Ge-Ni contact systems. For example, as-deposited samples of

Al-Ge-Ni contacts from separately evaporated Ni-Ge layers were found to be intermixed

on a GaAs substrate as determined with convergent beam electron diffraction (CBED) by

Liliental-Weber et al.20 Depth profiles after conversion from Schottky to ohmic electrical

behavior indicate varying degrees of interdiffusion. Interdiffusion was extensive for

samples heat treated at 500' C and limited for heat treatment at 425' C for times <15

minutes. For Au-Ge-Ni contacts, the sequence of evaporation was reported not to affect

the final arrangement of constituents after heat treatment. 28, 37, 40, 47, 49, 63 While a
difference in the mixing between the two different sequences was not obvious from

Auger depth profiles, the electrical results were dramatically affected by the layering

sequence. According to the elemental profiles, the arrangement of metal components

after conversion from Schottky to ohmic behavior ranges from limited interdiffusion to

extensive interdiffusion. These results demonstrate that the electrical measurements are

sensitive to small metallurgical changes, and that complete interdiffusion was not critical

to ohmic contact formation. Even limited metallurgical changes obviously were sufficient

to form the phase or cause the diffusion necessary to achieve ohmic behavior. In fact the

increased resistance at longer times during real time I-V measurement suggest complete
interdiffusion may be detrimental. Further details of phase formation based upon X-ray

diffraction analysis are discussed below.








Diffraction Analysis of Phase Formation in Al-Ge-Ni Thin Films

Three Element Metallization


As stated earlier, X-ray diffraction was chosen to identify the phases formed

during the heat treatment necessary for conversion from Schottky to ohmic electrical

behavior. Glancing angle X-ray diffraction was performed on various samples, such as

those with complete three element or two element metallization on either GaAs or

sapphire. The goal of the XRD analysis was to identify the phases present after

processing and to correlate this information with the interdiffusion results of the Auger

depth profiles and the electrical characterization results.

Both of the different layering sequences for the complete contact metallization

were examined with X-ray diffraction. As-deposited, diffraction peaks were strong from

elemental Al, with much weaker peaks due to GaAs, elemental Ge, and elemental Ni.

Very weak peaks consistent with Al3Ni, GeNi, and Ge3Ni5 were also detected. For

example, an XRD spectra from sample 43 taken at 8' incidence angle is shown in Figure

20. The chances of diffraction from the single crystal substrate is greatly reduced in the

glancing angle mode since the incidence angle is fixed. To increase the probability of

diffraction from the sample, data was collected at two incidence angles (4' and 8').

Common to nearly all samples were broad peaks at the peak positions of GaAs and/or

Ge. It is difficult to conclude on the basis of position alone, whether the diffraction is

from Ge or GaAs, since their lattice parameters are so similar. However, the low

intensity, broadened peaks were also seen in spectra of the two element samples without

Ge. Therefore, the broad low intensity peaks were attributed to strain induced in the

GaAs by the thin metal films in the absence of Ge. Ge may contribute to these peaks

when it is present. Diffraction peaks can be broad if the sample has phases with very

small crystallite sizes, or contain microstresses, disorder, stacking faults, dislocations,

and/or inhomogeneous solid solutions. The width of the diffraction peak will also












6000

5000

4000

3000

2000

1000


nI


r I

- Sample 43
As Deposited









5 6

_ !


V -
20


1+ 3
1 +(2, 3)


3 + (2)


SI I I

Al-1 Al3Ni-5 -
Ge -2 Ge3Ni 6
Ni-3 GeNi-7 -
GaAs-4


3, 2
1 3 1 32 38


1 j I 4


30 40 50 60 70 80 90 100


Diffraction Angle 2 0 (Degrees)


XRD Spectra of Sample 43, as-deposited, Al-Ni/Ge/Ni on GaAs



X-ray Diffraction data for as-deposited Al-Ge-Ni metallization. The
numbers indicate the probability that the phase is present (4 Many of
the peaks, 3 Some of the peaks, 2 Several of the peaks, 1 One or
two of the peaks from the phase detected, No entry No indication of the
phase).


Sample Number 43 43 47 47

Layering Al-Ni-Ge-Ni Al-Ni-Ge-Ni Al-Ni-Ge Al-Ni-Ge
Incidence Angle 8 4 8 4

Al 4 4 4 4
Ge 2 2 2 2
Ni 3 33 3
GaAs 2 2 2 2
Al3Ni 1 2 2 3
GeNi 1 1
Ge3Ni5 2 2 3


Figure 20.



Table 3.


U







75
increase as the film thickness decreases. Thus, the peak broadening of GaAs and/or Ge

peaks seen in Figure 20 can be attributed to microstresses in the GaAs substrate caused

by the metal overlayer, an amorphous or small crystallite size for Ge, the thinness of the

Ge layer, or a combination of these factors.

X-ray diffraction data from as-deposited samples 43 (Al/Ni/Ge/Ni) and 47

(Al/Ni/Ge), are summarized in Table 3. The table utilizes numbers to indicate the

existence of the particular phases, with 4 indicating the highest probability and 1

indicating the lowest probability of the presence of that phase. No entry means that no

evidence for that phase was detected in that particular sample. There were no differences

in the phases identified in the three element as-deposited samples for Ni versus Ge

adjacent to GaAs.

For contact metallizations heat treated at either 500' C or 425' C, XRD data were

collected from a cross-section of samples with both types of layering sequences.

Common to all heat treated samples was the increased formation of the A13Ni phase, with

nearly all of the peaks from this phase being present. Their peak intensities did not

correspond to the those of the ICDD card, as the data were collected in the glancing angle

mode. Also common to nearly all of the heat treated samples were broad peaks at the

peak positions of GaAs and/or Ge. The XRD spectra for sample 43 after heat treatment

for 2 minutes at 500' C is shown in Figure 21. In addition to the Al3Ni and the broad

GaAs and/or Ge peaks, peaks from elemental Al, AlNi3, GeNi, GeNi2, Ge3Ni5, Ga2Ni3,

Ni4GaGe2, and Ni5As2 phases have been identified. For samples with Ni at the metal

semiconductor interface and heat treated at 500' C, the XRD data have strong peaks from

Al3Ni as well as elemental Al and GaAs. The indications of GeNi, Ge3Ni5, Ga2Ni3,

Ga3Ni2, Ni5As2, and Ni4GaGe2 phases were based on peak position, but the intensities

of the peaks from these phases were weak even after collecting data for 69 hours.

Samples heat treated to 500' C with interfacial Ni always converted from Schottky to

ohmic behavior in 2 to 4 minutes, allowing less time for interdiffusion or extensive phase











3000 Oamp l Al I N 4iuaue2
2 Heat Treated 1
20 Ge -2 GeNi 7
0 2500 GaAs 3 GaNi 8 -
S1,4, 6, Ge3Ni5 -4 NisAs2-9
2000- 7, 8,9 _
0. A13Ni -5
1500 5 4 5
S50 265 4 4, 2,3,9
856
1000 35 4 -
7 5 5
5 5 1,8 8
500 8
2, 3
0 14 I I 2 2, 3
20 30 40 50 60 70 80 90 100
Diffraction Angle 20(Degrees)

Figure 21. XRD Spectra of Sample 43, After 500' C Heat Treatment for 2 Minutes.


formation. In addition, peak overlaps from these phases are quite extensive, so they

cannot be unambiguously identified. Few if any diffraction peaks could ever be
identified with iron, presumably because Fe was in solid solution with Ni and XRD from

a separate phase was unlikely. Representative examples of X-ray diffraction data from

samples 31 (Al/Ni/Ge), 43 (Al/Ni/Ge/Ni), and 47 (Al/Ni/Ge) after being heat treated at

500' C are summarized in Table 4. All of the samples listed in this table were heat
treated and electrically characterized by the iterative procedure described earlier.
According to the XRD data, the samples with Ge at the metal semiconductor interface
(samples 31 and 47) heat treated for 50 and 44 minutes, respectively, have many of the

same phases as sample 43 (e. g. A13Ni, GeNi, NiGaGe2). These type of samples exhibit
more NiAs2 and Ni5As2 phases. Although no effort was made to quantify the XRD data
due to the large number of peaks which overlap, the peaks assigned to the Ni4GaGe2
phase were more intense for sample 47 than for sample 31. Despite these similarities,
sample 47 converted from Schottky to ohmic behavior, while sample 31 did not convert







77
to ohmic behavior. With Ni at the interface (sample 43) more Ge3Ni5, GeNi2 phases
were observed (higher percent Ni concentration). The samples with Ni at the interface
have slightly greater indication of Ge-Ni and Ga-Ni phases while those with Ge at the
interface have more indication of Ni-As phases and the NiGe phase (lower percent Ni
concentration).


Table 4.


X-ray diffraction data after heat treatment at 500' C from the complete
contact metallization with Ni at the metal/semiconductor interface. The
numbers indicate the probability that the phase is present (4 Many of
the peaks, 3 Some of the peaks, 2 Several of the peaks, 1 One or
two of the peaks from the phase detected, No entry No indication of
the phase).


Sample Number 43 43 31 31 47 47

Layering Al-Ni-Ge-Ni Al-Ni-Ge-Ni Al-Ni-Ge Al-Ni-Ge Al-Ni-Ge Al-Ni-Ge
Temperature 500 500 500 500 500 50
Time (Min) 2 2 50 50 44 44
Incidence Angle 8 4 8 4 8 4

A 3 3 1 1
Ge 3 3 2 2 1 1
Ni 1
GaAs 3 3 2 2 1
Al3Ni 4 4 4 4 4
A1Ni3 1 1 1 3 3
GeNi 3 3 2 3 3 3
Ge3Ni5 4 3 1 1
GeNi2 3
Ga2Ni3 4 4 3
NiAs(Ge) 1 3 3 2
NiAs2 3__
Ni5As2 1 1 2 3 3 1
Ni4GaGe2 4 3 3 3 3 3


For samples heat treated at 425' C, the X-ray diffraction data are summarized in
Tables 5 and 6 for samples with Ge or Ni at the metal/semiconductor interface,
respectively. All of the samples heat treated at 425' C, similar to those heat treated at







78

Table 5. X-ray diffraction data after heat treatment at 425' C from the complete
contact metallization with Ge at the metal/semiconductor interface.
The numbers are as described in the caption for Table 4.

Sample Number 29 29 45 45

Layering A-Ni-Ge A-Ni-Ge ANi-Ge A-Ni-Ge
Temperature 425 425 425 425
Time (Min) RT 20 RT-20 100 100
Incidence Angle 8 4 4

Al T 1 1 1
Ge 3 3 3 3
GaAs 3 3 3 3
Al3Ni 4 4 4 4
AIGe 3 3 3 3
GeNi 3 3 3 3
GeNi2 1 1
Ga2Ni3 1 1
NiAs2 1 2
Ni5As2 1 1 2 2


500" C, had XRD spectra with indications of Al3Ni and the broad GaAs/Ge peaks.
Samples 29 (Al/Ni/Ge) and 28 (Al/Ni/Ge/Ni) were heat treated and electrically
characterized in real time. The primary difference in the XRD data from these two
samples was, again, that with Ge first more GeNi phase was present, while with Ni first
more Ge3Ni5, GeNi2, and Ga2Ni3 phases were present, similar to samples 43 and 47.
There generally seems to be more Ni-As phases present with Ge first; however, some
was observed with Ni first also. Recall that sample 29 (heated to 425" C)took several
times longer to convert from Schottky to ohmic behavior compared to sample 28 (heated
to 500' C). Samples 45 (Al/Ni/Ge) and 39 (Al/Ni/Ge/Ni) were heat treated and
electrically characterized by the iterative procedure. Sample 45, with Ge first, did not
completely convert to ohmic behavior even after 100 minutes at 425' C, since its
I-V curve was nearly, but not completely linear. Sample 39 converted from Schottky to







79
Table 6. X-ray diffraction data after heat treatment at 425' C from the complete
contact metallization with Ni at the metal/semiconductor interface. The
numbers are as described in the caption for Table 4.

Sample Number 28 28 39 39

Layering Al-Ni-Ge-Ni Al-Ni-Ge-Ni Al-Ni-Ge-Ni A-Ni-Ge-Ni
Temperature 425 425 45 425
Time (Min) RT -5 RT -5 50 50
Incidence Angle 8 4 8

Al 1 1
Ge 2 2 2 2
Ni 1 1
GaAs 3 3 2 2
A13Ni 4 4 4 4
Ge3Ni5 2 2 2 2
GeNi2 2 2
Ga2Ni3 2 2 3 3
NiAs(Ge) 3 3
NiAs2 1 1
Ni5As2 1 1 2 2


ohmic behavior, with two of the electrical lines being converted in 4 minutes, while the

last line required 50 minutes. Each sample was electrically characterized and heat treated

in 2 minute increments up to 40 minutes, then in 10 minute increments to 50 minutes total

time for sample 39 and to 100 minutes total time for sample 45. The primary difference,

as indicated by XRD data, was the presence of AlGe and NiGe for sample 45 with Ge

deposited first, and the presence of Ge3Ni5, Ga2Ni3, and NiAs(Ge) for sample 39 with

Ni deposited first, similar to the samples discussed previously. The XRD data for

samples 45 and 39, which were heat treated several times longer than either sample 29 or

sample 28, also indicate the presence of NiAs and Ni5As2 phases. Sample 39 with Ni

adjacent to GaAs was heat treated much longer than other Ni adjacent to GaAs samples.

While it does exhibit more NiAsx, it nevertheless exhibits Ni-Ga and the higher Ni

concentration Ni-Ge phases.








Three Element Metallization on Sapphire

Because of the complexity of the XRD spectra for the three element contact

metallizations, two-element samples on GaAs or sapphire and the complete three element

metal depositions on sapphire were made to aid in the interpretation of the X-ray data. As

stated previously, the metal on the sapphire substrates were deposited simultaneously

with those on GaAs. XRD data of the complete contact metallizations grown on

sapphire, similar to the contacts grown on GaAs, strongly indicate the Al3Ni phase. In

addition, common to samples with both type of layering sequence, sapphire was detected

along with elemental Al and Ge, while there is weak indication for the formation of

several Ge-Ni and Al-Ge compounds. Again the most prominent phase in the spectra of

both types of samples was Al3Ni. The presence of the elemental Ge phase after the 500'
heat treatment lends credence to its identification on GaAs. Other than the absence of Ga

and As containing phases, the phases from the complete metallization on GaAs and

sapphire were similar. XRD data for representative examples of the complete contact

metallizations on sapphire is summarized in Table 7.

Two Element Metallization

Because of the dominance of the Al3Ni phase in the XRD spectra of the complete

contact metallization, samples with the same Ni:Ge or Ni:Ge:Ni ratio and sequence were

examined with XRD. Samples of Ni (500A)/Ge (250A) and Ni (450A)/Ge (250A)/Ni

(50A) on both GaAs and sapphire substrates were heat treated at either 425' C or

500' C. XRD data for the Ni-Ge and Ni-Ge-Ni samples is summarized in Tables 8 and

9 respectively. Qualitatively, the phases identified on both types of samples grown

on GaAs and heat treated at 425" C and 500' C were similar. Samples with both types

of layering sequence and heat treated at 425' C have less indication of NiAs(Ge), while

samples with both types of layering sequence and heat treated at 500" C have less

indication of elemental Ni. Sample 35, with Ni at the metal/semiconductor interface and










Table 7.


X-ray diffraction data from the three element contact metallization on
sapphire. The numbers are as described in the caption for Table 4.

Sample Number 27 27 31 31

Layering Al-Ni-Ge Al-Ni-Ge Al-Ni-Ge-Ni Al-Ni-Ge-Ni
Temperature 500 500 500 500
Time (Min) 9 9 3 3
Incidence Angle 8 4 8 4

Al 3 3 4 4
Ge 3 3 3
Sapphire 4 4 4 4
AI3Ni 4 4 4 4
AMNi 3 3
Al2Ge 22
GeNi 2 2 2 2
Ge3Ni5 2 2
GeNi2 1 2 2
Gamma-Al203 1 1 3


X-ray diffraction from the two element Ni-Ge metallization on GaAs.
The numbers are as described in the caption for Table 4.

Sample Number 33 33 44 44
Layering Ni-Ge Ni-Ge Ni-Ge Ni-Ge
Temperature 500 500 425 425
Incidence Angle 8 4 8 4

Ge 4 4 3 3
Ni 1 3 1
GaAs 4 4 2 3
Ni4GaGe2 2 1 2 3
NiAs(Ge) 2 2 1 2
Ga3Ni2 1 1 1 1
Ga2Ni3 1 1 1 1
GeNi 3 T3 3 2----
GeNi2 1 1 1 1
Ge3Ni2 1 1 3
Ge3Ni5 1 1 3 1
Ge12Ni19 1 1


Table 8.







82
heat treated at 500' C, has more indication of the Ni4GaGe2 and NiAs(Ge) phases. In
addition, XRD data from all of the samples indicate the same Ge-Ni and Ga-Ni phases as
the complete contact samples. Again intensities of most of these peaks were weak, in
spite of long collection times, due to the thinness of the metal layers. XRD data for these
samples on sapphire substrates also exhibit Ge-Ni phases, again similar to the complete
contact metallization on sapphire.

Table 9. X-ray diffraction data from Ni-Ge-Ni two element metallization on GaAs.
The numbers are as described in the caption for Table 4.


Sample Number 34 34 35 35
Layering Ni-Ge-Ni Ni-Ge-Ni Ni-Ge-Ni Ni-Ge-Ni
Temperature 425 425 500 500
Incidence Angle 8 4 8 4

Ge 3 4 3 3
Ni 2 2 1 1
GaAs 3 4 3 3
Ni4GaGe2 1 3 4 4
NiAs(Ge) 1 1 3 3
Ga3Ni2 1 1 1
Ga2Ni3 1 1 1
GeNi 3 3 3 3
GeNi2 1 2 1
Ge3Ni2 2 1 3
Ge3Ni5s 3 2 4
Ge12Ni19 1 I 1 1


Summary of Phase Formation in Al-Ge-Ni Thin Films


In summary, X-ray diffraction data demonstrated that the films interdiffused
during the deposition process as shown by A13Ni, GeNi and Ge3Ni5 phase formation,
consistent with TEM and EDS results.20-23 In previous studies the phases could have
formed during TEM sample preparation. However, the XRD results in this study
indicated that the phases were formed during deposition, since the samples did not







83
indicated that the phases were formed during deposition, since the samples did not

receive further processing. These phases, in addition to GaAs, elemental Al, Ge, and Ni,

were all detected from as deposited samples, independent of whether Ge or Ni was at the

metal semiconductor interface. For the samples deposited onto sapphire, in

addition to sapphire, elemental Al, Ge, Ni, and Al3Ni were observed. For all heat treated

contacts, the dominant phase observed from deposited layers was Al3Ni. All of the

samples generally had indications of many of the same phases, with only slight

differences due to the layering sequence and heat treatment time. These differences are

summarized in Table 10. It should be pointed out again, that in general, the heat

treatment times for the Ge first samples were much longer than for the Ni first times. For

the complete contact samples with Ge at the metal semiconductor interface and heat

treated at 425' C, there was a stronger indication of the NiGe phase than of the Ge3Ni5

phase, that is, the Ni-Ge phase with higher Ge concentration. For the complete contact

samples with Ni at the metal semiconductor interface and heat treated at 425' C, peaks

from Ge3Ni5, Ga2Ni3, and NiAs(Ge) phases were more numerous, that is, Ni-Ge

phases with higher Ni concentration and Ni-Ga phases. In addition, samples heat treated

for longer time to convert contacts to ohmic behavior also show more evidence for Ni-As

phases. For the complete contact samples heat treated at 500' C, the difference due to

layering sequence was more dramatic, consistent with the fact that with Ni at the

metal/semiconductor interface, less time was required to convert from Schottky to ohmic

behavior. The limited time at temperature may inhibit the amount of phases formed.

Electrical measurements were obviously more sensitive to metallurgical changes than

were XRD measurements. XRD data for the complete contact samples with Ni at the

metal semiconductor interface and heat treated at 500* C did indicate elemental Al and

Al3Ni was present. In addition there were slight indications for GeNi2, Ge3Ni5,

Ga2Ni3, Ga3Ni2 and Ni4GaGe2 phases when data were collected for 69 hours. The







84
samples with Ge at the metal/semiconductor interface and heat treated at 500' C required

longer periods at temperature to convert to ohmic behavior, and generally their XRD data


Table 10. Summary of thin film X-ray diffraction results for complete contact
metallizations on GaAs.


had stronger indication of Ni-As, NiGe, and Ni4GaGe2 phases with weaker indications

of Ge-Ni phases with higher concentrations of Ni. While there are similar phases

indicated for both types of samples, there are some different trends for samples with Ni

adjacent to the GaAs versus samples with Ge adjacent to the GaAs. Ni first samples

always have more Ni-Ge phases with higher concentration of Ni and more Ni-Ga phases


As-Deposited
Elemental Al, Ni, and Ge
Al3Ni and Ni5Ge3 form during deposition


Heat Treated
All samples dominated by Al3Ni
Other common phases: -- NiGe, Ni5Ge3, Ni3Ga2, NiAs(Ge),
NiAs2, Ni5As2, Ni4GaGe2
425' C -- Ge First
Greater indication of NiGe than Ni5Ge3
(e. g. higher Ge concentration Ni-Ge phase)
Longer heat treatment -- more Ni-As phases

425' C -- Ni First
Greater indication of Ni5Ge3, Ni3Ga2, and NiAs(Ge)
(e. g. higher Ni concentration Ni-Ge phases)
Longer heat treatment -- more Ni-As phases
Greater indication of decomposition GaAs substrate

500' C -- Ge First
Less indication of Ni-Ge phases
Greater indication of Ni-As, Ni4GaGe2

500' C -- Ni First
More elemental Al and Ge
Less indication of NiGe, Ni5Ge3, Ni2Ga3, and Ni4GaGe2
Greater indication of decomposition GaAs substrate







85
present. Ge first samples always have the NiGe phase present, that is, the phase with
less Ni. The Ge first samples have Ni-As phases present, but so did the Ni first sample

which was heat treated for a longer period. While the XRD data was taken for only

before and after the formation of ohmic contacts, these phase formations are consistent

with diffusion couple reaction paths for each of the layers. The Ni first samples appear to

be reactive with the GaAs substrate, as indicated by recent work by Chambers and

Loebs.78 XRD results for three element metallizations on sapphire and the two element

samples on both GaAs and sapphire exhibited similar phases when the missing

components were considered.














CHAPTER 5
DISCUSSION


In this discussion section, important new results will be discussed and

summarized. This is followed by discussion of reaction paths to help explain the phases

observed. Next, a mechanism of surface doping during the reaction to form the ohmic

contact is discussed. Finally a comprehensive model for ohmic contacts from Al-Ge-Ni

on n-type GaAs is presented and discussed. Electrical measurements have been used to

demonstrate dramatic differences in the times and temperatures required to convert the

Al/Ni/NiGaAs versus Al/NiGe/Ni/GaAs metallizations from Schottky to ohmic electrical

behavior. For samples with Ge at the metal-semiconductor interface, the times required

to become ohmic behavior were shown to be dependent on both the doping concentration

of the contact layer of the GaAs, and on the heat treatment temperature. The times were

longer for contact layers with low doping concentration and for lower heat treatment

temperature. For samples with Ni at the metal-semiconductor interface, the times

required to convert from Schottky to ohmic behavior at either 425 or 500' C were much

less and did not depend on doping concentration or temperature. Real time I-V

measurements indicated that samples with Ni at the metal-semiconductor interface had an

onset of ohmic behavior as low as 180' C. Conversion to ohmic behavior at such a low

of temperature confirms the speculation of Edwards et al.23 that molten eutectics are not

required for contact formation. The kinetic effect has not been previously reported for

any ohmic contact system.

One objective of this research was to demonstrate that the Al-Ge-Ni metallization

was a viable candidate for ohmic contacts to all ranges of doping in GaAs. Since

Al-Ge-Ni ohmic contacts have only been reported for doping concentration levels in the







87
high-1017 cm-3 and above range,1,2, 20-23 lightly doped (Si @ 1016 cm-3) as well as
heavily doped (Si @ 1018 cm-3 and above) n-type GaAs were tested. In all cases, ohmic

contacts were observed to form. The electrical data shows that Al-Ge-Ni converts to
ohmic behavior on contact layers grown by MBE and VPE in addition to ion implanted

layers. These different substrates were used because of the reported difficulty of

converting Al-Ge-Ni contacts to ohmic behavior on layers other than ion implanted

contact layers.22 This difficulty was not observed in this study. Thus it appears that this

metallization is a good candidate for ohmic contacts for GaAs devices.
Auger survey scans have been used to demonstrate that surfaces which have been

exposed to the atmosphere were contaminated with O and C. The scans were also used
to show that the dendritic precipitates on some of the samples were pure Ge. The cause

of Ge precipitation in the early samples was shown to be a high ratio of Ge to Ni (-1:1).

Precipitation was controlled by reducing the ratio of Ge to Ni (~1:2). Auger depth

profiles were also used to demonstrate that interdiffusion took place during deposition of

the metals and during the heat treatment procedure. This interdiffusion was consistent
with phase changes as analyzed by XRD.


Reaction Path Analysis


Auger depth profiles after heat treatment times and temperatures necessary to

convert from Schottky to ohmic electrical behavior indicated varying degrees of

interdiffusion. This interdiffusion behaved as expected with extensive interdiffusion for

samples heat treated at 500* C or 425' C for longer than 15 min. and limited
interdiffusion for heat treatment at 425' C for shorter times. Unlike the electrical data, the

Auger depth profiles were not dramatically affected by the layering sequence. The effects
of deposition sequence have been studied using Auger depth profiles and RBS for Au-

Ge-Ni, Pd-Ge, and Pd-Si metallizations.28, 33, 37, 40, 45, 47, 49, 64 The conclusions







88
were similar, that is interdiffusion and phase formation were detected, but not all of these

events correlated well with electrical behavior of the contacts. In addition, layer sequence

effects have only been reported for magnitude and variation in resistance and not the time

to form ohmic contacts. The Auger depth profiles illustrate that Ni diffused to the metal-

semiconductor interface during heat treatment, as reported for the Au-Ge-Ni system by
Robinson.40 It was reported that Ni increased the solubility of Ge in GaAs.28, 40, 48

X-ray diffraction data demonstrate that the films interdiffused during the
deposition process consistent with the Auger depth profile data. These data indicate the
formation of AI3Ni, Ge3Ni5, and GeNi phases during deposition, which was were

reported in the literature based upon EDS and TEM data.20-23 A13Ni was the primary
phase observed by XRD spectra for all of the contacts after heat treatment, but all samples
exhibited many of the same phases (see Table 10 for a summary of differences due to the

layering sequence and heat treatment time). Samples with Ni adjacent to GaAs converted

quickly to ohmic behavior and exhibited Ni-Ge, Ni-Ga, Ni-As and Ni-Ga-Ge phases.

These samples exhibited more Ni-Ga and Ni-Ge (with higher concentration of Ni) phases

than samples with Ge adjacent to GaAs. Samples heat treated at 500' C had more

indication of Ni4GaGe2 and NiGe phases than samples heat treated at 425' C. All

samples with Ge adjacent to GaAs converted to ohmic behavior after longer times and

had more NiGe, Ni-As, and NiAs(Ge) phases. These samples exhibited more NiGe

(with higher concentration of Ge) and fewer Ni-Ga phases. The Al-Ni and Ni-Ge phases

identified in this study were also reported in the previous Al-Ge-Ni literature.20-23

However, the Ni-Ga and Ni-As phases identified here were not reported previously. As

pointed out in the literature review, calculated and experimental phase diagrams show that
the only phases in thermodynamic equilibrium with GaAs at temperatures below ~ 800' C

would be NiGa, Ni2Ga3, and NiAs binaries.73 This is consistent with the observation

of Ni-Ga and Ni-As phases.







89
One of the controls of this study was to maintain the same total amount of each
element for both the Ni first and the Ge first samples. Although the reaction paths are

obviously different dependent upon whether Ni or Ge is adjacent to GaAs, the reactions

will proceed toward the same end point and should reach the same end point given

enough time. Since both types of samples, Ni first and Ge first ,were heat treated until

they did become ohmic, the phases present could be expected to be similar if ohmic

contacts require a particular phase or phases to be present independent of reaction

pathways.. Obviously the reaction pathways will be different between these two

conditions. For the two different layer sequences, i. e. Al/Ni/Ge/Ni/GaAs versus

Al/Ni/Ge/GaAs, reactions with the semiconductor will begin as Ni-GaAs or Ge-GaAs.

Ge-GaAs is a stable system as shown below by diffusion calculations and in the MBE

literature79-81 for Ge epitaxially grown on GaAs. Ballingall et al.80 estimated that

interdiffusion of MBE Ge-GaAs heterojunctions to be -10A after 1 hour at 400' C. On

the other hand, Ni has been reported to decompose GaAs quickly and at low

temperatures.78 Therefore, dissociation of GaAs and formation of Ni-Ga and Ni-As

phases is expected to occur with Ni adjacent to GaAs while Ge adjacent to GaAs is

expected to be nonreactive. At the Ni-Ge interfaces the two components will behave as a

diffusion couples, with Ni rich compounds at the Ni side and Ge rich compounds at the

Ge side. If the Ge-GaAs interface is stable, then presumably Ni had to reach the GaAs to

cause ohmic contacts, even when the Ge was first. If this is true, slow conversion to

ohmic contacts may have resulted from both kinetic (time for Ni diffusion through Ge) as

well as thermodynamic factors (formation of NiGex reduces the driving force to form Ni-

Ga and Ni-As phases).








Mechanism of Doping


As discussed above, in-diffusion of Ge to form a heavily doped n+ layer under

the metal contact is the most common explanation for forming ohmic contacts on GaAs.

The depletion width is reduced by the n+ layer and carrier are transported by a tunneling

mechanism. However, this model does not appear to be appropriate for the Al-Ge-Ni

contact system. It is well known that Ge-GaAs is a very stable system. Recently Ge

diffusion in GaAs has been quantified using a concentration dependent diffusion
equation.82 In-diffusion was modeled using a vacancy mechanism and compared to

SIMS profiles of annealed Ge implanted into GaAs. The effective diffusion coefficient

was determined to be Deff = 2 x 10-3 cm2/sec exp [ -2.9eV/kT ] (n/ni)2, where n and ni

are the doping concentration of the contact layer and the intrinsic doping concentration of

the GaAs respectively. Using the times and temperatures required to convert to ohmic

behavior for Ge first samples (20 min/425' C and 9 min/500* C for depositions on

heavily doped substrates), the calculated values are Deff/500 =3.9 x 10-16 cm2/sec and

Deff/425 = 1.5 x 10-17 cm2/sec. Using these diffusivities in an error function diffusion
distance calculation gives values of X500 = 45.8A and X425 = 13A. Similar calculations

of diffusion distances for low doped (5 x 1016 cm-3) substrates yield diffusion distances

of less than 1A. The calculated diffusion distances are less than the depletion width of

~100-150A for heavily doped substrates. Therefore, the commonly used diffusion

argument does not seem appropriate for the Al-Ge-Ni contacts in this study since it does

not appear to provide enough Ge for the n+ layer.


General Model of Ohmic Contact Formation


It is now possible to postulate a model that will explain the conversion of

Al-Ge-Ni contacts to ohmic behavior based on results and previous literature described







91
above. As stated earlier, the contacts with Ni at the metal-semiconductor interface

converted from Schottky to ohmic behavior much faster, therefore, the Ni at the metal-

semiconductor interface must have enhanced the conversion to ohmic contact. It has been

reported that Ni dissociates GaAs.28, 78 It has also been reported that intermediate

phases of GaAs with Pd and Ni form at low temperature,31-35, 83, 84 and conversion to

ohmic behavior has been reported at the onset of their decomposition and concomitant

regrowth of the substrate. For example, Pd4GaAs has been reported to form at

temperatures as low as -100' C. At higher temperature Ge reacted with this ternary

compound resulting in formation of PdGe and regrowth of a GaAs layer.84 This

regrown layer/PdGe contact was ohmic, presumably due to heavy doping of the GaAs

during regrowth. This is an extension of the model for Au-Ge-Ni/GaAs ohmic contacts

when some investigators suggested that GaAs regrew upon cooling the from the Au-

GaAs solid solutions which had formed during heating. Recall that Yeh and Holloway28

showed that decomposition and regrowth of GaAs took place during heat treatment of Au

contacts to Si doped GaAs. In this case, regrowth occurred during isothermal heat

treatments rather than during cool-down. In a continuation of this work Li and

Holloway68 showed that GaAs will regrow during isothermal heating of GaAs covered

by discrete layers of As and Ga under an outer film of either Au or Au-Ge. In addition,

Li and Holloway30 have measured the specific contact resistance of ohmic contacts using

this regrown GaAs technique and found values of 10-5 flcm2. Thus it is certain that

regrowth of GaAs can lead to formation of ohmic contacts. Finally the presence of a

dopant in the Au may not be necessary since Holloway et al.29 demonstrated that dopant

segregation occurred upon reaction of Au with GaAs. They showed that Si dopant from

the contact layer was concentrated in the near-surface regions where GaAs decomposed

in order to react with the Au contacts. They suggested that the concentrations of Si were

sufficient to reduce the depletion distance and allow tunneling transport of charge carriers

across the interface for ohmic behavior.







92
For the Al-Ge-Ni system it is suggested that a similar decomposition and

regrowth mechanism is operable for Al/Ni/Ge/Ni contacts. For these samples, elemental

Ni would dissociate the GaAs to form an intermediate NixGaAs ternary phase at low

temperature. During further heat treatment the layers would continue to interdiffuse. At

longer times, formation of Al-Ni and Ni-Ge phases would remove the Ni from and

destroy the Ni-Ga-As phase causing regrowth of a heavily doped GaAs(Ge) layer under

the Ni5Ge3, NiGe, NixAs, and NiAs(Ge) phases observed by XRD. Admittedly, no Ni-

Ga-As ternary phase was detected in this study; however, the model is equally valid if the

NiAs, NiGa, and Ni4GaGe2 identified in this study serve the same purpose as the

postulated Ni-Ga-As phase. These phases may well be the unidentified binary or ternary

phases shown in the deconvoluted XPS spectra by Chambers and Loebs.78 Their data

also show that Ni adjacent to GaAs is much more reactive than either AINi or Al. The Al-

Ga reaction was very limited and the Al-As reaction did not appear to occur at all. In

other words, with Ni adjacent to GaAs, reaction products are Ni-Ga and Ni-As or

Ni-Ga-As phases which are subsequently reduced by Al-Ni and Ni-Ge phase formation.

It is also possible that segregation of Si from the n-type contact layer might be involved in

the heavy doping of the regrown layer as discussed above for Au contacts.29, 30, 68 In

all instances, an n+ doped GaAs region, sufficient to reduce the depletion distance and

permit tunneling transport of the charge across the metal/semiconductor interface, is

postulated to result in ohmic behavior. This Ni dissociation and GaAs regrowth

mechanism is completely consistent with the observations for samples with Ni adjacent to

the GaAs. The elemental profiles show that diffusion takes place during heat treatment.

The resultant phase formation observed by XRD can be interpreted to be reaction

products which form due to decomposition and regrowth.

It seems likely that a similar mechanism should explain the samples with Ge next

to GaAs. The extra time required to convert to ohmic behavior could be due to the time

required for Ni to diffuse to the metal/semiconductor interface. In addition, the reaction







93
pathway would cause rapid formation of Ni-Al and Ni-Ge phases in this case, which

would reduce Ni-Ga and Ni-As phase formation. However, one observation which

needs explanation is that longer times were required to convert samples with low doped
versus high doped substrates. Thus the rate limiting step must either proceed at a
different rate or proceed along a different path for changes in doping. Both possibilities

exist based on the model and the literature cited above. For example, the data for Ge

diffusion in GaAs shows that the rate depends upon the square of the doping density. If

a similar dependence existed for Ni in Ge the short time for high doping, longer time for

low doping is easily accounted for. This is especially true since with the Al/Ni/Ge/GaAs

sequence, the Ni must first react with Ge before it can reach the GaAs. If Ni-Al and or
Ni-Ge phases lead to decomposition of NiGa and Ni-As or Ni-Ga-As phases, then the

amount of Ni reaching the Ge/GaAs interface would be very limited. Presumably a

regrowth layer approximately the thickness of the depletion width must be regrown to

form a good ohmic contact. Since the depletion width is larger at low doping, more

regrowth and therefore more reaction and therefore longer diffusion times would be

required. A second effect leading to longer time at low doping could result from
segregation of the Si dopant.29, 30, 68 If segregated dopant led to the n+ region and

ohmic behavior, again a larger volume of GaAs would need to react at low versus high

doping densities. For either regrowth and/or dopant segregation, limited transport of Ni

in Ge would require longer times at low versus high doping levels.

Segregation of dopant to the metal/semiconductor interface may be indicated the

results of Graham et al.21-23 Recall that they reported Al-Ge-Ni contacts converted to

ohmic behavior on both n and p-type GaAs. If Ge is incorporated to form an n+ layer,

the metallization on p-GaAs should have resulted in a pn junction. Alternatively, their p-
type Al/Ni/Ge/GaAs contacts may have converted to ohmic behavior due to Zn

segregation to the metal/semiconductor interface during decomposition of the GaAs

substrate. If GaAs regrowth occurred, Ge could be incorporated into the regrowth layer.







94
Presumably this is consistent with their results since their p-GaAs substrates were heavily

doped and Ge incorporation may have been only partially compensated the Zn dopant.

As reported above, there was no dependence upon temperature or doping

concentration for ohmic formation when Ni was adjacent to GaAs. This probably results

because the times of heat treatment were long compared to fast reactions between Ni and

GaAs at the interface. If these samples were heat treated for much shorter periods or at

much lower temperatures, an effect of doping concentration upon heat treatment time

required for conversion to ohmic behavior would be observed.

The different times required for conversion to ohmic behavior based on contact

layer sequence, temperature, and doping of the contact layer are the basis for the building

blocks for a process model of the conversion of Al-Ge-Ni contacts from Schottky to

ohmic behavior. One of the more widely used process model simulators is SUPREM

(Stanford University Erocess Emulator). This model predicts distribution of elements

that result from a given process treatment for a certain time. The process model

simulations for silicon devices are now well developed in SUPREM, and include the one

and two dimensional distributions of oxide layers (SiO2) onto Si resulting from a given

treatment in an oxidation furnace for a specified time and temperature. Process modeling

for Si based devices is much more advanced than for GaAs based devices. Nonetheless,

the work reported in this dissertation is a first crude step in building a process model for

ohmic contact formation in GaAs for the Al-Ge-Ni system. In this work, the distribution

of elements resulting from specified time-temperature treatments have been determined.

More precise knowledge of the distributions of dopant are clearly needed and could

probably be obtained by SIMS profiles of the ohmic contact. This work should be

conducted in future experiments.