Citation
Evolution of interfacial phases and their effects on ohmic contacts to n-GaAs in Ni-Ge-Ti metallizations

Material Information

Title:
Evolution of interfacial phases and their effects on ohmic contacts to n-GaAs in Ni-Ge-Ti metallizations
Creator:
Kim, Tae-Jin, 1965-
Publication Date:
Language:
English
Physical Description:
vii, 168 leaves : ill. ; 29 cm.

Subjects

Subjects / Keywords:
Annealing ( jstor )
Atoms ( jstor )
Contact resistance ( jstor )
Diffraction patterns ( jstor )
Doping ( jstor )
Electrical phases ( jstor )
Electrons ( jstor )
Phase diagrams ( jstor )
Regrowth ( jstor )
Semiconductors ( jstor )
Dissertations, Academic -- Materials Science and Engineering -- UF ( lcsh )
Materials Science and Engineering thesis, Ph. D ( lcsh )
Genre:
bibliography ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1996.
Bibliography:
Includes bibliographical references (leaves 158-167).
General Note:
Typescript.
General Note:
Vita.
Statement of Responsibility:
by Tae-Jin Kim.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
The University of Florida George A. Smathers Libraries respect the intellectual property rights of others and do not claim any copyright interest in this item. This item may be protected by copyright but is made available here under a claim of fair use (17 U.S.C. §107) for non-profit research and educational purposes. Users of this work have responsibility for determining copyright status prior to reusing, publishing or reproducing this item for purposes other than what is allowed by fair use or other copyright exemptions. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder. The Smathers Libraries would like to learn more about this item and invite individuals or organizations to contact the RDS coordinator (ufdissertations@uflib.ufl.edu) with any additional information they can provide.
Resource Identifier:
026345267 ( ALEPH )
36676994 ( OCLC )

Downloads

This item has the following downloads:


Full Text










EVOLUTION OF INTERFACIAL PHASES AND THEIR EFFECTS ON OHMIC CONTACTS TO n-GaAs IN Ni-Ge-Ti METALLIZATIONS




















By
TAE-JIN KIM





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

UNIVERSITY OF FLORIDA


1996





























Copyright 1996

by

Tae-Jin Kim

















ACKNOWLEDGEMENTS

With all my sincere respect and love, I must thank my parents for their love,

sacrifice and support. Without what they gave me, I could not have come this far. Now, I dedicate this dissertation to them.

I would like to express my gratitude for my wife, Jin Hee, who has patiently endured this challenging time in her life without losing her bright smile.

I must also thank Dr. Paul Holloway for his guidance. I have had a wonderful

opportunity to learn how to approach academic issues. But more importantly, for the first time I could discover the example of the professor and scientist I want to be.

Many people helped me with the work presented in this dissertation. Maggie did all of the SIMS analysis and Eric did the AES analysis. Wish helped me collect the TEM data and Sohn helped me analyze the TEM diffraction pattern. I appreciate the opportunity to work with Dr. Kenik at the Oak Ridge National Laboratory. I have enjoyed my companionship with our group member, Troy, Sean, Joe, Jonathan, Philip, Jeff; all of them helped me whenever I asked their help and I thank them. Also there are several other unnamed students and people who deserve my gratitude in our department. Even though I started the work in this dissertation myself, it would have been impossible to finish without all the people who helped me.








TABLE OF CONTENTS

ACKNOWLEDGEMENTS iii

ABSTRACT vi

CHAPTERS

1. INTRODUCTION 1

2. REVIEW OF LITERATURE 6

Electrical Properties of GaAs Surface and Interfaces 6

Ex situ Contact Schemes 19
Au/GaAs Metallizations 19 Ni/GaAs Metallizations 22 AuGe/GaAs Metallizations 25 NiGe/GaAs and Other Bielements/GaAs Metallizations 27 AuNiGe/GaAs Metallizations 31 Hybrid Contact Metallizations 35

In situ Contact Schemes 36

Current Status of Contact Metallizations 38

3. EXPERIMENTAL PROCEDURE 42

GaAs Substrate Preparation 42 Metal Evaporation and In Situ Annealing 43 Post-Evaporation Annealing 46 Electrical Measurements 46 Analytical Characterizations 47
Auger Electron Spectroscopy (AES) 48 Secondary Ion Mass Spectroscopy (SIMS) 48 Transmission Electron Microscopy (TEM) 49

4. RESULTS 53

Summary of Prepared Samples 53 Electrical Measurements 54 Elemental Depth Profiles 58 Transmission Electron Microscopy 63








5. DISCUSSION


Evolution of Ni2.4GaAs 126 Mechanism for the Formation of Ohmic Contacts 136 6. CONCLUSIONS 149 APPENDICES

A HEAT OF FORMATION 152 B CALCULATION OF PHASE DIAGRAM 155 REFERENCES 158 BIOGRAPHICAL SKETCH 168














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 EVOLUTION OF INTERFACIAL PHAES AND THEIR EFFECTS ON OHMIC
CONTACTS TO n-GaAs IN Ni-Ge-Ti METALLIZATIONS By

Tae-Jin Kim

December 1996

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


Electrical and metallurgical properties of Ti/Ge/Ni metallizations used to form ohmic contacts to n-type (100) GaAs have been studied with current-voltage (I-V) measurements, secondary ion mass spectroscopy, Auger electron spectroscopy, and transmission electron microscopy with energy dispersive analysis of x-rays. The purpose of this study was to investigate the evolution of Ni-Ga-As phases and their effects on the formation of ohmic contacts.

Thin films of Ti, Ge and Ni deposited by electron beam to thickness of 30 - 60nm, 25 -75 nm and 65nm, were heat treated in vacuum at 500'C either as all three deposited layer at the same time or the Ni film was annealed in situ to react with GaAs, followed by in situ deposition of Ge and Ti and ex situ annealing in vacuum at 5000C. For the case of Ni deposition followed by in situ annealing at 3000C, a 120 - 130 nm








thick hexagonal ternary Ni2.4GaAs phase was produced. After deposition of Ge / Ti layers on this Ni2.4GaAs and vacuum annealing at 5000C for 5 minutes, the electrical behavior switched from rectifying to ohmic. Simultaneously with development of ohmic behavior, the Ni2.4GaAs phase was transformed into NiI.3As and Ni2.1Ga phases, which were decomposed by the formation of Ni-Ge compounds and NiTi. The decomposition of the Nil.3As resulted in solid phase epitaxial regrowth of GaAs doped with Ge at Ga sites and the Ge concentrations were measured up to mid- 1020cm3. This entire evolution was towards the equilibrium phases in the Ni-Ga-Ge-As-Ti system. The thickness of the Ti and Ge layers, and microstructure of the regrown GaAs were critical factors controlling development of ohmic contacts. Regrown GaAs was characterized by various defects and planar interfacial morphology for the -30nm regrowth range. A model describing the level of Ge incorporated on Ga sites to act as donor was developed based on control of the Fermi level of the GaAs at the regrowth interface with the Ni-based phases. This model quantitatively predicts that the conduction band will shift below the Fermi level and lead to tunneling transport for ohmic contacts.
















CHAPTER 1
INTRODUCTION


GaAs is the best known and most widely used III-V semiconductor. The most advantageous properties of GaAs over Si are higher electron mobility, lower voltage operation, semi-insulating substrates, monolithic integration of optical and electronic functions and radiation hardness [Scl88]. Utilizing the higher electron mobility of GaAs, Integrated Circuits (ICs) such as Monolithic Microwave Integrated Circuits (MMIC) are fabricated for high-frequency devices. The direct band gap of GaAs makes it possible for use in infrared-emitting diodes and solid-state lasers. Also GaAs has played a key role in epitaxial growth of ternary or quaternary compounds for diverse optical applications [Hai89].

Regardless of application, operation of all GaAs devices requires two types of

electrical contacts, rectifying (or Schottky) and ohmic contacts. For instance, ohmic contact is required for the drain and source and rectifying contact for the gate in a Metal Semiconductor Field Effect Transistor (MESFET) structure. To guarantee successful device operation and performance, the contact should have a low resistance, reliability with time and reproducibility as well as adaptability to the fabrication








2
process. In current GaAs fabrication technology, ohmic contacts are formed mostly by a metallization step using a lift-off process with the Au-Ge-Ni system developed in 1967 for n-type GaAs [Bra67].

The issues associated with contact formation become of critical importance in order to fully utilize the best performance of the devices, especially the high frequency and high power characteristics [Chr95]. In addition to the device performance, reliability is also strongly affected by the electrical contacts as the level of circuit integration approaches Ultra Large Scale ICs (ULSI) and oscillation frequency of GaAs ICs approach several hundreds of GHz [And90]. In a typical MESFET structure, the channel length between source and drain is -1 tm. For a higher speed operation and higher integration, the channel length is being reduced; consequently, the spatial tolerance for contact metallization becomes more stringent. In a high power switch where current density approaches 104 A/cm2, a high resistance ohmic contact ( > 10-3 gcm2) can cause excessive Joule heating leading to failure of the devices [Bar92].

It has been reported that current Au-Ge-Ni metallization is unlikely to be used for future ULSI and other applications because of its severe interdiffusion and metal/semiconductor reactions often resulting in degradation of the performance and failure of the devices [Woo86, And9O]. Problems associated with Au-Ge-Ni metallization such as irregular morphology [Gup90, Ba189], thermal stability [Wil88, Hug92] and interdiffusion [Kol88] have been addressed by several authors. Tremendous efforts have been made to solve the problems of the Au-Ge-Ni system









[Rid75, Pio83, She92] and they have resulted in improvements non-alloyed and Aufree contact schemes [Meh89, Mur89, Tan92].

Generally, the electrical properties of metal contacts to semiconductors depend on composition, structure and morphology of the interface, which are determined by interfacial reactions. The interfacial reactions are unavoidable because semiconductors and metals have enormously different properties, for example, lack of lattice match and different electronic characteristics. Consequently, a reaction takes place at the interface. The reaction are very complicated to characterize because several elements are involved with many processing variables. This might explain why ohmic contact technology had begun as an art rather than as a science, and why the complete mechanism of forming ohmic contacts in Au-Ge-Ni system is not yet understood even though Au-Ge-Ni metallizations provide practically acceptable low contact resistances Rid75, Woo86].

While Au-Ge-Ni metallization has been the subject of extensive investigations

throughout the 1970s and 1980s, simpler metallizations such as single element/GaAs and bi-elements/GaAs were studied in the 1980s as an effort to better understand the complicated Au-Ge-Ni system. As a consequence, the model of solid-phase regrowth has been developed [San88, Hol9l, Li92]. In the solid-phase regrowth, a metal element, such as Ni or Pd, reacts with GaAs, producing an interfacial phase such as NixGaAs or PdyGaAs. When these interfacial phases are decomposed, the electrical contact property switches from rectifying to ohmic. As for the current transport mechanism in conjunction with the formation of ohmic contacts, the solid-phase









regrowth assumes the creation of degenerately doped n'-GaAs leading to tunneling across the metal/GaAs contact. Even though some guidelines for the solid-phase regrowth has been established, several key issues in the mechanism still need to be better studied and clarified. One such issue is the evolution of interfacial phases, which are believed to control or dramatically affect the contact property.

The purpose of this study was to investigate the evolution of interfacial phases

consisting of Ni, Ga and As and their effects on the formation of ohmic contacts. Ni has been used a standard element in ohmic contact metallizations for GaAs devices, and it forms various compounds with GaAs. Particularly, special attention was given to a ternary hexagonal phase, NiGaAs, since previous studies showed that the contact properties were dependent on how this ternary phase evolved [San88]. Ge was selected as an n-type dopant. Ti was selected to cause the continued evolution of the interfacial phases by reacting with Ni causing it to be removed from the Ni-Ga-As phases. Ni and Ti have a strong affinity and therefore form Ni-Ti compounds. Ti has been used to play the role of diffusion barrier as an element or as a constituent in the compounds conventionally. In this study, however, Ti was used to react with Ni rather than to remain inert. This type of use of Ti in forming ohmic contacts has not been studied before. In order to focus on the evolution of the interfacial phases, in situ annealing was adopted, which will be described in Chapter 3.

For the dissertation organizations, Chapter 2 begins with a review of the literature regarding the unique surface/interface properties of GaAs. Then, various metallizations with different complexity are reviewed. During this review, the most








important models of the mechanisms for the formation of ohmic contacts to GaAs are discussed. In Chapter 3, the experimental approach and procedures are described. These include sample treatment, evaporation, in situ annealing and post-evaporation annealing along with the analytical techniques such as transmission electron microscopy (TEM), secondary ion mass spectroscopy (SIMS) and Auger electron spectroscopy (AES) to characterize the samples. In Chapter 4, results of the metallurgical study and electrical measurements are described and correlated to each other. In Chapter 5, all the results are reviewed and discussed and a model for the formation of ohmic contact is proposed. Finally, conclusions from this study are summarized in Chapter 6.















CHAPTER 2
REVIEW OF THE LITERATURE


Today's semiconductor devices have been developed based on the understanding of electrical properties of semiconductor surfaces and interfaces. There is a need for such a understanding since semiconductor interfaces control the transport of current across the electrical junctions. The understanding was made possible largely due to the comprehensive surface science work on III-V surface and interfaces. This literature review begins with a description of one important electrical property of the GaAs surface, Fermi-level pinning, which has been investigated for several decades. Knowledge of Fermi-level pinning phenomena in GaAs is a central issue because the mechanism of Schottky barrier formation is strongly affected by the Fermi level pinning.



Electrical Properties of GaAs Surfaces and Interfaces



The rectifying properties of Schottky contacts were first described by using a simple approach [Mot38, Sch40]. Ideal Schottky contacts form as a result of charge transfer between metal and semiconductor to align the Fermi level across the interface. Figure 2-1 illustrates this process.








Figure 2-1 (a) shows the energy band diagram of a metal and a n-semiconductor before contact (unless otherwise mentioned, the following discussion always refers to n-type semiconductors). As shown, the work function of the metal 4)m is larger than the work function of semiconductor (D., As a result of the potential difference, (m - (D., electrons from the semiconductor flow into the metal until the Fermi levels on each side align at one level when metal and semiconductor are joined. The surface of semiconductor becomes depleted of free electrons, resulting in a positively charged space region. Because of the low densities of free carriers in non-degenerately doped semiconductors, a large length of space charge region is established. The length of this space charge region is given by the depletion length W. In GaAs, W ranges from -10 A to -1000A depending on the doping concentration at room temperature [Sze8 1]. On the metal surface, penetration depth of the induced is determined by the Thomas-Fermi screening length. Because of higher density of states per unit volume at the Fermi level, typical Thomas-Fermi screening length is less than -0.5A. Consequently, the space charge layer lies entirely on the more weakly screened semiconductor side of the interface.

An energy barrier (Db is defined as the energy distance between the Fermi level and the edge of the majority-carrier band (in this case, the conduction band minimum) and determines the Schottky Barrier Height (SBH), Ob. This is the barrier that electrons must surmount or tunnel through in order to conduct current across the metal/semiconductor contact. The barrier Ob of an ideal interface is shown (Fig. 2-1 (b)) to be equal to: b b= ((Dm - 0.) + (4). - X) = D. - X, (2-1) where X is the electron affinity of semiconductor as defined in Fig. 2-1 (b).









The resulting band bending due to the space charge region is described by the spatial variation of electrostatic potential, V(x), V(x) =(1/q).(Ew -Ev(x)) = (1/q).(Ecb-Ec(x)) (2-2) where subscripts b refer to bulk (i.e., x= oo), and V and C refer to the valence and conduction bands. In thermal equilibrium, band bending at the surface (x=O) is uniquely determined by:

qV(x=O) = Evb -Ev. =Ecb - Ec. (2-3) where the subscript s refers to the surface (i.e., x=O). With the assumptions of semiinfinite boundary conditions and no interface states, the potential as a function of x can be obtained by solving Poisson's equation: d2V(x)/dXi = -p(x)IsS (2-4) where Fs is the dielectric constant of the semiconductor and p(x) is the spatial distribution of charge. Provided that there are no free carriers in the depletion region of n-type GaAs and complete ionization, i.e., p(x) = ND, solving Poisson's equation yields V(x) = Vs +Em(x - x2/2W) (2-5) where Em is the electric field at x=O given by (qNDW)/ss, W is the depletion length, (2ssVbi/qND)"2, and Vb, is the built-in potential defined by Ec - Ec,. Here, Vs is the surface band bending at x=O. Equation (2-5) shows that the band edge is a parabolic function of x from x =0 to x =W (Fig. 2-1).

Along with 'Db, another important figure of merit in the formation of an ohmic contact is the specific contact resistance given by: Pc = {(dJ/dV)v=0O'l[!cm2 (2-6)









where J is current density and V is the applied bias across the contact. In order to obtain low resistance ohmic contacts, the barrier height Db should be small for thermionic emission of current over the barrier, or the depletion region, W in Fig. 2-1 (b), should be narrowed for tunneling of carriers through the barrier. Practically, ohmic contacts to GaAs devices must be tunneling contacts because the Fermi level is pinned and therefore ctb can not easily be reduced.

Calculating the contact resistance, Pc, requires details of the current density, J, which can be estimated from

J = no q � V P((Ib) (2-7) where n is the concentration of free electrons, q is the charge of an electron, P(cDb) is the tunneling probability, and v is the thermal velocity of electrons. While P(QDb) is a complicated function of the detailed band structure of the semiconductor, an approximate form of the solution for equation (2-7) was used to define three mechanisms of current transport across the metal/semiconductor contacts [Bar92].

For thermionic emission over the Schottky barrier, the solution for pc is given by pc = (k/qA*T) . exp(Qb/kT) (2-8) where A" is the Richardson constant, k is the Boltzmann constant and T is the absolute temperature. As can be seen in equation (2-8), the contact resistance is independent of the doping concentration and hence the depletion width.

In the case of tunneling, the contact resistance is approximated by

Pc cc exp((Db/Eoo) (2-9)









where E00 = (h/47t) � (ND/m*&,)"2. The contact resistance is lowered for ohmic contacts when ND is mad large.

In the third category, a current density is obtained from a mixed form of tunneling and thermionic emission, and the contact resistance is given by pc oc exp(,/Eoo) / coth(Eoo/kT) (2-10) where the temperature dependence of thermionic emission is contained in the coth(Eoo/kT) term. This mode of transport is called Thermionic (assisted) Field Emission (TFE) in which the carriers are thermally activated above Fermi level to a region where they can tunnel efficiently through a narrower portion of the barrier. Because coth(z) approaches unity for large z, equation (2-10) has been used empirically to predict ohmic or non-ohmic behavior of the interface for a given doping concentration:

Eoo/kT >> 1 => coth(Eoo/kT) =1 Ohmic by tunneling

E0o/kT = I => coth(Eoo/kT) > 1 Non-linear Eoo/kT << 1 => coth(Eoo/kT) >>1 Non-ohmic.

For z =2.7, the error in setting coth(z) =1 is less than 1%. Therefore, it is possible to predict the doping concentration required to form a tunneling contact by setting E0o/kT >> 2.7.

Then Eoo= 18.5 x 1012 - (ND/mrc,) 2> 2.7.kT eV (2-11) where mr =m*/rm, s, = &A.o and ND is the carrier concentration (cm3). For n-type GaAs at 300K, equation (2-11) predicts ND > 1.2 x 10'9cm3. This approximation suggests the effective doping concentration should be greater than 10 19cm-3 at room temperature to form tunneling a ohmic contact to n-GaAs.








The Shottky barrier height is obviously an important parameter in these formulations, and it plays a key role in the determination of current transport mechanisms. According to the ideal Schottky barrier in equation (2-1), the equilibrium barrier height (Db depends on the work function of the metal. In reality, however, experimentally measured barrier heights showed a very weak dependence on the metal work function, especially for group IV elements and weakly ionic III-V compound semiconductors such as GaAs [Bar47, Cow65, Spi79].

This deviation from ideal Shottky behavior was first attributed to the existence of

intrinsic or extrinsic surface states by J. Bardeen in 1947 [Bar47]. When surface states are present within the bandgap of n-type semiconductors, they are occupied by electrons. Occupation is controlled by the Fermi level which is constant throughout the crystal. Consequently, these surface states may pin the Fermi level resulting in band bending as shown in Fig. 2-2. Now the barrier height 'Fb is not given by Equation 2-1 but must be modified by the details of Ess, the energy levels of the surface states. For instance, in Fig. 2-2, the barrier height can be described as (Db = Eg - Ess [Bar47]. As shown, the barrier height is nearly independent of the metal work function and is called Fermi level pinning. Fermi level pinning has been observed not only in GaAs, but also in Si and Ge [Mye47].

Kurtin et al. proposed intrinsic mechanisms as a plausible origin of the surface states, and suggested a correlation between a tendency to form surface states and types of chemical bonding in the bulk. Based on barrier heights of the various metals on Si, GaSe and SiO2 related to the electronegativities of the metals, it was suggested that the termination of periodic atomic arrangement at the surface of covalently bonded








semiconductors gave rise to surface states [Kur69]. However, it has been shown experimentally [Gud76, Spi76, Tan84] and theoretically [Sch90] that intrinsic states in the bandgap for the (110) surface of GaAs do not play a role in Fermi-level pinning.

A large body of experimental data were analyzed and resulted in the Unified Defect Model. Spicer et al. discovered that clean, cleaved (110) surfaces of II-V compounds with an exception of GaP did not exhibit intrinsic states, but the surfaces were strongly perturbed by coverages of monolayer or less resulting in surface states and Fermi-level pinning [Spi79, Spi80]. They proposed that the pinning resulted from defects created by the interactions between the adatoms (metals and oxygen) and semiconductors. Later, the AsGa antisite defect was suggested to be controlling the pinning level, either at 0.5eV or 0.7eV above the valence band maximum (VBM) [Spi88]. According to this model, the Fermi level varied from 0.5eV to 0.7eV above the VBM depending on the concentration of the antisite defects. The heat of condensation was proposed to result in atom displacement in one version of this mechanism but limited data do not support this mechanism. One difficulty in understanding of the UDM is that the defect states can not be directly detected nor quantified.

The validity of two pinning level in the UDM model was argued based on theoretical calculation [Zur83]. Later it was found that conditions of metallization determined one of two pinning values [Cao89]. Unique to the UDM is that Fermi level pinning is a consequence of extrinsic effects, not intrinsic properties of the semiconductors.

The Metal Induced Gap (MIG) model has been developed as another explanation for the Fermi level pinning [Hei65]. In the MIG model, surface states are an intrinsic property







13
of the semiconductor, and continuum levels are predicted in the energy range between the VBM of the semiconductor and the Fermi level, which is a central contradiction to the descrete levels described by UDM. The MIG are formed in the semiconductor at the initial interface due to intimate contact with the electrons of the metals. By such an intimate contact, the tails of metallic wave functions decaying into the semiconductor side provide the interface states within the band gap of the semiconductor. Heine first pointed out that, at the metal/semiconductor interface, the tails of the metal wavefunctions into the semiconductor are derived from the virtual gap states of the complex band structure of the semiconductor [Hei651. When specific boundary conditions are satisfied, the "virtual" gap states which are solutions of Schrodinger's equation for semiconductor band structure with complex wave vectors become "real" surface states with a decay distance into the semiconductor when specific boundary conditions are satisfied [Lou76, Ter84, Mon93].

In the M[G model, the character of the surface states changes across the band gap from predominantly donor- to predominantly acceptor-like closer to the VBM and the CBM (conduction band minimum), respectively, since the surface states are derived from the band structure of the semiconductor. The energy at which the contributions from both bands are equal in magnitude is called the branch point, and the branch point is located at the mid-bandgap when the effective mass of electrons and holes are of equal value [Mon93]. Again this is an intrinsic property of semiconductor band structure. The Charge-Neutrality Level (CNL) is defined as the energy where the Fermi level coincides with the branch point and this is a critical energy level which determines barrier height.








14
According to the MIG model, charge transfer occurs across the interface depending on the electronegativity difference between metals and semiconductors and determines the final level of Fermi-level pinning. For example, when the electronegativity of a metal is the same as the semiconductor, the final Fermi level pinning is located at the branch point of the semiconductor and there is no charge transfer. The final pinning position of the Fermi level should be above or below the branch point of the semiconductor when the metal exhibits a smaller or a larger eletronegativity than the semiconductor. Consequently, the MIG model predicts a dependency of the barrier height on the electrical properties of metals as follows [Mon88]:

tbm= 0=1 + Sx(Xm - Xs) (2-12) where 'D,,j is the charge neutral level, S, is a parameter defined as ODb /Xm, Xm and Xs are the electronegativities of metal and semiconductor respectively.

In the recent development of the MIG [Mon88], it was shown from experimental data [Cao87, 89] that the position of Fermi level pinning was a function of metal layer coverage; the Fermi level approached its final position as the surface was saturated from submonolayer coverage (isolated adatoms) to a continuous metallic film. The dependency of Fermi level position as a function of metal coverage, which can't be explained by the UDM, was explained in the MIG model by using the idea of metal-induced surface states at submonolayer coverage and by the continuum of metal-induced gap states at several monolayer coverage.

In the beginning, the idea of MIGs had begun as a simple model lacking of microscopic knowledge of real interface structure and developed largely by theoretical calculations.








Now it appears that the idea is gaining more attention with further refinement and supporting experimental data [Mc188, Bur91].

Chemical reactivity between semiconductors and metals was correlated to explain the deviation from the ideal Schottky barrier by the fact that chemical reactions affect the electrical properties of the interfaces. Heat of formation was used as a measure of the reactivity between metals and Si [And75]. It was found that the barrier heights (4b) between transition metals and Si exhibited a linear relation with the heat of formations (-AHf). Later, Freeouf adopted a different approach to correlate the barrier heights of the metals/Si contacts to chemical reactions [Fre80]. He proposed that Ob had a relationship with (Isilicide (work function of silicides) instead of the metal work functions. Even though this idea was developed to describe metal/Si contacts, it was later applied to various compound semiconductors and developed as the Effective Work Function (EWF) Model [Fre81]. In the EWF model, Ob,, (Schottky barrier height with a metal) is given by Ot= Oefr - X (2-13) instead of

Ob (mct.I - X

where X is the electron affinity of the semiconductor. 4Deff is mainly due to the work function of the anion, OAno.. The EWF model suggests that the Fermi level at the surface (or interface) is not fixed by surface states but rather is related to the work functions of microclusters of one or more interface phases resulting from either oxygen contamination (oxidation) or metal-semiconductor reactions which occur during metallization. The EWF









model is a refinement of the original Schottky description since it does not introduce the concept of surface states.

Another chemical argument proposed is that chemical reactions at the interface on a microscopic scale modify the ideal Schottky barrier via local charge transfer and creation of extrinsic interface states as a result of the interfacial reaction [Bri78, Bri90]. In this argument, discrete levels of defects, native or extrinsic, were used to account for the deviation from the ideal Schottky behavior with chemical reactivity. This approach does not seem to be a main theory in that it does not propose a microscopic mechanism for Fermi level pinning.

Relatively recently, Fermi level pinning was attributed to amphoteric native defects in semiconductors by Walukiewicz [Wal87, 88, 89]. A remarkable correlation exists between the Fermi-level position (EFs) at metal-semiconductor interfaces and the Fermi level (ERI) in heavily irradiated III-V compound and column IV semiconductors [Wal88]. Table 2-1 shows the range of Fermi-level pinning positions deduced from the Schottky Barrier Heights for metal-semiconductor contacts and the Fermi level stabilization energy in heavily irradiated III-V compounds and column IV elemental semiconductors.


Table 2-1. Fermi level stabilization energy in irradiated semiconductors (ERI) and
at metal-semiconductor interfaces (EFs). EB represents charge-neutrality level from MIG model. All energies are with respect to the valence-band
edges. [Wa188, and references therein].
ER, (eV) EFS (eV) EB (eV)
Si 0.4 0.3 -0.4 0.36 Ge 0.07 0.16 0.18
GaAs 0.5-0.7 0.5-0.7 0.5









From the similarity of the values listed in table 2-1, the same mechanism as for En in heavily irradiated materials was proposed for the EFS in the metal-semiconductor contacts.

According to his reports, there is a Fermi-level stabilization energy (EFq) in covalent or weakly ionic semiconductors including GaAs, which is independent of the type of doping and the doping level. Therefore, this property is regarded as an intrinsic property of the semiconductors. As a consequence of this intrinsic property, native defects such as vacancies or substitutional dopants exhibit an amphoteric character depending on their energy level relative to the Fermi-level stabilization energy. For example, a Ga vacancy is a stable acceptor in n-type GaAs, but it transforms to a donor complex (Asc , + VA) in ptype materials. This behavior results from a large electronic contribution to the defect reactions. For example, the formation energy of a Ga vacancy is lowered from -4eV to

-0.2eV as the Fermi level varies from the Valence Band Maxima (VBM) to the Conduction Band Maxima (CBM) under As-rich condition [Zha91 ]. Such a transformation continues until the Fermi level reaches the stabilized position, EF. In such a case, introduction of further electrically active species does not affect the stabilized Fermi-level. In the case of GaAs, the stabilized Fermi-level is located between E, +0.5 eV to E, +0.7 as can be seen in table 2-1.

The amphoteric native defects model is unique in that the behavior of native defects are responsible for the Fermi level pinning, which appears to be similar to the UDM model but the behavior of defects are controlled by the stabilized Fermi level EFR, which is an intrinsic properties of semiconductors, similar to the MIG model.








For many years, all the models mentioned above were examined and compared.

Nonetheless, the main idea of whether intrinsic or extrinsic effects play a primary role in the Fermi level pinning still remains controversial [San85, Spi85, Ter88, Spi93]. It is now generally accepted that a single theory can not explain the Fermi level pinning and a special effort is being made to build general consensus for the main idea [Spi93].

Meanwhile, appropriate descriptions as to the Fermi level pinning tend to be more specific, depending on the actual structure of the interface [Fre90, Bri90] and even on semiconductor growth method [Vit93]. In fact, it is readily and clearly seen that there is a range of-0.3eV in the Fermi-level pinning, 0.5 - 0.8eV above the VBM by reviewing the extensive experimental data collected to date. This variation might imply that several mechanisms are simultaneously playing roles and even interacting with each other in the determination of SBH. In the latter case, understanding of the Fermi-level pinning would be more complex than expected.

With the plethora of models and inability to critically discriminate between them, there is not a general model which enables us to predict the barrier height for practical applications. In GaAs technology, we can not rely on controlling the Schottky barrier height by selecting metal elements. This situation led to the need to produce a tunneling ohmic contact scheme by controlling barrier width rather than barrier height.









Ex situ Contact Schemes



Au/GaAs Metallizations



Au is frequently used as a contact metal for GaAs as well as a base metal in

semiconductor processing because of its ease of deposition and etching, high ductility and conductivity. Au begins to react with GaAs at -250�C, but significant reaction occurs at

-400'C with large change in the Schottky barrier heights. Typically, Au/GaAs diodes exhibited a -0.9eV barrier height as evaporated and less than 0.7 eV or ohmic behavior as heat treated [Gyu7l, Leu85, Ho192]. The interfacial reactions between Au and GaAs upon annealing can be generally expressed as follows: Au +GaAs => Au-Ga + As (gas)

which represents the formation of Ga-rich Au solid solutions or Au-Ga compounds with As loss in the case of an open system. In the case of a closed system, the As might exist in the form of a precipitate. The Au-Ga compounds can be one or a combination of the following compounds; Au7Ga2, Au3Ga, Au2Ga, AuGa, or AuGa2 upon cooling [Kum76, VanS0, Yos83, Lin86, Kim90].

Unique to the Au/GaAs reactions is the formation of elongated pyramidal pits bounded by { 111 } planes, aligned in [110] directions of GaAs, which are noticeable when annealed above -3500C [Pez86, Bau86]. In the pyramidal pits, pure Au and/or Ga-rich Au solutions were observed. These contents in the pits were separated from the GaAs substrate by an intermediate layer of Au-Ga compounds which are reaction products of









the Au/GaAs interfacial reactions. The pyramidal reactions pits were thought to be formed through solid-state dissolution of GaAs up to 4500C although a liquid state reaction increased their size and governs overall morphology when annealed above the melting temperature (-500'C) of the Au-rich solid solutions. In addition to the reaction pits, agglomerated regions (Ga-rich Au solution) which cover the underlying pits were observed [Gyu7l, Kim90].

The reaction between Au and GaAs proceeded in a highly inhomogeneous manner,

resulting in formation of monocrystalline gold, a hexagonal Au-Ga phase, and precipitates of Au2Ga in the matrix of GaAs with certain crystallographic orientations [Bau86]. The crystallographic orientational relationships between the reaction products and the parent Au and GaAs were studied and described by the degree of misfit at the interface; Au/AuGa layer and the Au-Ga/GaAs interfaces are formed in such a way that lattice misfits at the interfaces are minimized [Yos83]. The inhomogeneity of the interfacial reaction can result from different diffusivity of Au depending on the crystallographic orientation of GaAs and the manner that Au penetrates the native oxide to react with the substrate GaAs, which will vary depending on the thickness and distribution of the native oxide.

The effects of phase structure and morphology on the electrical properties of Au-Ga compounds/GaAs were studied by depositing Au-Ga phases with various Ga concentrations on GaAs [Leu85]. They could not attribute the variation of electrical properties to the presence of specific phase because a phase mixture with nonuniform distribution was present at the interface. However, it was observed in general that the formation of the Au-Ga compounds as a result of Au/GaAs interfacial reactions reduced








21
the SBH (Schottky Barrier Height) below that of pure Au. This result was confirmed by depositing and annealing chemically inert AuGa2 on GaAs [Lin86]. This study showed that chemical reactions were minimal at the AuGa2/GaAs interface after annealing up to 500'C because the AuGa2 phase is an equilibrium phase.

Chemical reactions between Au/GaAs were reviewed to understand the variation in the electrical properties of the Au/GaAs system by studying the details of the pyramidal reaction pits resulting from the chemical reactions [Ho192]. It was shown that Au/GaAs interfacial reactions vary depending on the conditions of the initial interface; surprisingly, the interface with a native oxide exhibited the most pronounced formation of reaction pits, suggesting that the stability of the interface is critically dependent on the surface stoichiometry. Also it was observed that the reactions pits underwent Ostwald ripening phenomenon, i.e., some pits disappear and some pits grow as the reaction proceeds during isothermal annealing. A segregation of Si, which was a dopant in the substrate GaAs, was detected in the reactions pits by SIMS. Because the appearance of reactions pits indicates the dissolution of GaAs, the disappearance of the pits indicates the regrowth of GaAs, suggesting the possibility of dopant incorporation into the regrown GaAs as the reason for the change in electrical behavior from rectifying to ohmic behavior. The change from Schottky behavior to ohmic behavior was correlated to the formation of Au crystallites after annealing above 3600C [Lil85]. The ohmic behavior was attributed to the leakage currents at the periphery of the contacts. But they could not provide a reasonable mechanism for current transport mechanism associated with their observation, therefore, this result could not be regarded as a mechanism for ohmic contacts.








22
In general, Au dissociates GaAs inhomogeneously producing Au-Ga compounds upon cooling, resulting in reduced barrier heights. The morphologies and details of the reaction products in Au/GaAs appear to be dependent on the relative concentration of Ga in Au-Ga compounds possibly because the variation in the Ga concentration dictates different equilibrium liquidous temperatures in Au-Ga-As ternary phase. The low melting temperature of the Au-Ga compounds is known to be a possible cause for poor morphology in the vicinity of device processing temperatures -4000C. In terms of the effects of Au metallization on the electrical properties, Au has been known to enhance Ga out-diffusion, thereby increasing the chance that external doping elements occupy Ga sites leading to the formation of n+-layer. However, solid-phase regrowth model that will be discussed in this review, and the results of this study clearly showed that Ge doping was the consequence of solid-phase epitaxial regrowth following interfacial reactions not a consequence of Ga out-diffusion.



Ni/GaAs Metallizations



Ni reacts uniformly with GaAs except where the native oxide-hydrocarbon interfacial layer remained intact [San87]. Contamination at the Ni/GaAs interface affects diffusion and compound formation [Sol87]. Solid-state reactions between Ni and GaAs produced a hexagonal ternary phase after annealing at 100'C - 400'C, for times of 5 minutes to 5 hours:


xNi + GaAs = NixGaAs.









The ternary phase designated as NixGaAs was epitaxial with the GaAs substrate and analyzed by diffraction analysis in TEM and XRD [Oga8Ob, Che83, Lav86, San86, Che88, Gui89]. The composition x varied between 2 and 4 according to those authors. The quantification of AES data yielded Ni2GaAs [Oga8Ob, Lav86, So187] while EDS and/or TEM measurement suggested Ni3GaAs [San86, Lin88]. The variation in the composition ranging from 2 to 4 was observed to accompany different co/ao ratios in the same NiAstype hexagonal structure [Che88, Gue89]. An experiment on bulk material to determine the Ni-Ga-As ternary phase diagram showed that there were five ternary phases with broad homogeneity ranges extending toward the binary phases [Gue89]. They all exhibited hexagonal NiAs-type symmetry and were unstable in contact with GaAs.

It was proposed that Ni3GaAs would adopt a B8 structure since Ni-Ga and Ni-As binary systems exhibit B8 structures with lattice parameters similar to Ni3GaAs and the symmetry of NiAs-type structures. The lattice parameters of NixGaAs were intermediate between those of Ni3.55Ga2.0 and NiAs [San86, 87]. NiGaAs is characterized by a hexagonal unit cell with lattice parameters ao-4/&, co-5A, similar to the Ni-As and Ni-Ga systems (NiAs: ao = 3.619, c. = 5.034A and Ni3Ga2: a = 4.0, c. = 4.983A) [San86, Che88]. Bright-field TEM images of NixGaAs showed a columnar structure with microtwins and diffraction patterns from TEM and XRD exhibited twin variants with respect to the GaAs matrix as well as several epitaxial relationships with the substrate GaAs [Gui89].

NixGaAs was more often observed in thin film reaction and is thought not to be an equilibrium phase. Due to the epitaxial relationship of NixGaAs with, the nucleation









barrier for formation of Ni.GaAs is much lower in thin films than that of the equilibrium phases such as NiAs and NiGa [Lav86]. Also, the formation of NixGaAs might be favored kinetics since Ni has been found to be the diffusing species in the reactions with GaAs. Contrary to Ni, Ga and As were relatively immobile at temperatures below -300'C and the formation of NixGaAs was triggered by in-diffusion of Ni [San87b, Lin88, Che88].

Ogawa showed that Ni2GaAs separated into binary NiGa and NiAs by annealing at

500'C for 5 minutes and suggested that Ni2GaAs was not stable in the presence of excess GaAs [Oga80]. In another study, Ni2GaAs was found to be stable up to 6000C, but the stability was dependent on the orientation of the substrate. Ni2GaAs substrate was more stable on (111) versus (001) GaAs [Lav86, Gui89]. On the bulk materials, the phases after annealing at 600'C for 1 hour were NiGa and NiAs [Gui89]. This indicates that there are only tie lines connecting GaAs and NiGa and NiAs binaries in their experimentally determined Ni-Ga-As ternary phase diagram [Gui89]. After annealing 5000C, 5 minutes, NiAs or As-rich phases were found near the GaAs and f-NiGa near the surface [Oga8Ob]. In this study, the role of Ni was explained to be a highly reactive element which eliminated the contamination layer such as an oxide layer, and thus assists the dissociation of GaAs. Decomposition proceeded through NiAs precipitation in a matrix of Ni2GaAs at temperatures higher than 3500C. After annealing at 6000C, 13-NiGa and NiAs grains with an average size of 0.7 gam were observed. The P3-NiGa showed an epitaxial relationship with GaAs substrate. On the contrary, NiAs grains were randomly oriented [Lav86, Gui89]. In another investigation, NiAs and NiGa formed after annealing at 6000C for 1 hour was epitaxial on the GaAs substrate [San87b].








25
The Schottky barrier height was increased when the Ni2GaAs was formed from 0.76eV to 0.83eV, which was similar to the AuGeNi with Ni diffusing. The barrier height was reduced by decomposition of the Ni2GaAs [Lav86]. Sheet also increased at the temperatures where the ternary phases was formed (200 - 300'C) and decreased at the temperatures of the decomposition of the ternary phase (above 4000C) [Gui89]. All the phases from the Ni/GaAs reactions were highly textured with epitaxial relationships to the substrate (except cubic NiGa) [Gui89]. The intermediate NixGaAs phases were isostructural with the final, equilibrium binary phases.

In conjunction with the potential effects of Ni/GaAs interfacial reactions on the

electrical properties of Ni-containing ohmic contact metallizations, Ni improved greatly the uniformity of the reactions. The Ni-Ga and Ni-As compounds are characterized by high melting temperatures relative to compounds in Au/GaAs reactions, which is believed to lead to improved thermal stability of the contacts. Decomposition of NiAs phase containing Ge was found to be responsible for regrowth of GaAs doped with Ge at Ga sites in this study.



AuGe/GaAs Metallizations



Adding Ge to Au modifies the metallurgical reactions with GaAs because 88wt% Au with 12wt% Ge forms a eutectic composition with a melting temperature of 3630C. However, even adding a small amount of Ge (-0.6wt%) drastically changed the morphology of the reaction products(Au7Ga2, Au3Ga) and lowered compound formation








temperatures with increasing Ge concentration [Kim86]. In addition to Au-Ga compounds, a Au-Ge-As phase was found to cover the contact surface at -4000C, and to a lesser extent above -4000C. Regrown GaAs was also observed with the disappearance of the Au-Ge-As phase, which indicates local melting and solidification [Kim90]. After annealing above -4000C, melting of multi-phases initiated by the Au-Ge eutectic phase and/or Au-Ga compounds following recrystallization were observed, which generally resulted in the formation of ohmic contacts with specific contact resistance of -10' 2cm2.

It was suggested in the process that the Au-Ge liquid phase dissolves Ga and As into the melt, and upon cooling, recrystallized or regrown GaAs would incorporate Ge forming heavily doped GaAs. Ge was suggested to be critical to initiation of the melting process by formation of Au-Ge or Au-Ge-As eutectic compositions, as well as formation of the n'layer leading to ohmic contacts [Ili83]. After annealing at 450'C - 500'C which is well above the melting temperatures of Au-Ge and Au-Ga compounds, Au-Ge, Au-Ga compounds and epitaxially regrown GaAs were found to be major reaction products, all of which were governed predominantly by liquid phase reactions [Kim90]. The formation of ohmic contacts was attributed to incorporation Ge during the solidification of GaAs.

In-diffusion of Ge and out-diffusion of Ga were observed after annealing below the Au-Ge eutectic temperature as well as above the temperature. It was reported that a dramatic reduction in the barrier height from 0.77eV to -0.4eV occurred even after annealing below the Au-Ge eutectic temperature without major reactions [Ili83, 87]. These experiments were correlated to the formation of a heavily doped n' GaAs layer and the formation of ohmic contacts with Au-Ge metallizations [Gyu7l, 1i83, Ku186]. Again,








the formation of n'-GaAs layer based on Ga out-diffusion is believed to be unreasonable as will be obvious by the results of this study.

For the samples annealed above or below the Au-Ge eutectic temperature, the reduced barrier height was attributed to the formation of a gradually disordered interface rather than the formation of heavily doped n' layer, or formation of graded heterostructure was suggested to account for the formation of ohmic contacts from the idea that the interface would be heavily restructured after complicated alloying reactions above the Au-Ge eutectic temperature [Kir87].

In summary, AuGe/GaAs exhibited a lower contact resistance than Au/GaAs. The overall reaction morphology, such as the formation of pyramidal reaction pits, was still observed with some extra phases due to the presence of Ge. Similar to the case of Au/GaAs reactions, Au-Ge eutectic contacts also lacked a uniform morphology.



NiGe/GaAs and Other Bielements/GaAs Metallizations



As reported in the chapter 1, marginal thermal stability, irregular morphology, and

degradation of contact resistance with time in AuGeNi metallizations has been correlated with the presence of Au. Au-free metallization was pursued to determine if this would eliminate the problems and resulted in simpler bielemental metallization such as NiGe [Tan92], PdSi [Wan88], and PdGe [Tsu9 I]. The metallurgical reactions in Au-free metallization are often described as sintering reactions or solid-phase reactions emphasizing that the overall reactions are completed through solid-state diffusion.








Without Ni, very little interdiffusion between Ge and GaAs was observed after sintering at 4500C, 30 minutes [And78]. This study showed that the presence of Ni overlayer on top of Ge greatly enhanced interdiffusion of Ge into GaAs resulting in the formation of ohmic contacts. This supports the model of a heavily doped n' layer as a mechanism in forming ohmic contacts.

Solid-phase epitaxy of Ge was investigated to form non-alloyed contacts to n-GaAs by using Ge/Pd/GaAs metallization [Mar87]. Pd and Ge reacted to form PdGe and excess Ge was transported through the Pd layer to grow epitaxially on the GaAs substrate. As to the mechanism for the formation of ohmic contacts, tunneling through Ge/GaAs interface and formation of n+-GaAs was suggested. However, the study failed to provide information regarding type or concentration of free carriers in the Ge layer nor experimental evidence for the existence of n+-GaAs layer. To clarify the mechanism for ohmic contacts, backside SIMS analysis was used to measure Ge incorporation into the GaAs [Pal90]. This study showed that the concentration of Ge (-1 x 10'9cm-3) at the contact with GaAs was correlated to the onset of ohmic behavior. Based on their metallurgical study and Ge detection, dissolution of GaAs leading to the formation of Pd4GaAs and following regrowth of GaAs was suggested as the primary mechanism for formation of ohmic contacts. Even though this backside SIMS could not provide reliable quantitative data regarding the extent of regrowth, this result supports the Ge incorporation into GaAs, resulting in the formation of n' layer.

The characteristics of Si/Pd/GaAs metallization were investigated to compare to those of Ge/Pd/GaAs metallization [Wan88]. It was shown that the epitaxial Ge layer on the








GaAs substrate or low barrier heterojunction was not responsible for ohmic behavior. They suggested a regrowth mechanism in which Pd reacted with GaAs, and produced a very thin Pd4GaAs. However, excess Si can react with Pd4GaAs to produce the following:

2Si + Pd4GaAs(Si) => 2Pd2Si +GaAs(Si).

A final product of this reaction sequences is regrown GaAs doped with incorporated Si. Si was suggested to have diffused into the ternary phase before decomposition and occupied Ga sites in regrowth, creating an ne-layer doped - 2 x 10'9cm'3. From this study, the ohmic behavior in Si/Pd/GaAs and Ge/Pd/GaAs were attributed to the formation of n+ GaAs. This solid-phase regrowth mechanism was further tested in Ni/Si/GaAs, which clearly showed that the regrowth could take place due to the relative thermodynamic stabilities between phases [San88]. Another study using Ni/Si, however, did not support the regrowth model based on the formation and decomposition of NixGaAs phase because they could not detect NixGaAs phase even after annealing at 3000C [Tak92].

The concept of the regrowth mechanism was applied to explain ohmic behavior for Ni/Ge/GaAs metallizations [Tan88]. First, Ni reacted with GaAs to form NixGaAs, then decomposition of the ternary phase was driven by the lower free energy of formation of NiGe to result in:

Ni.GaAs + xGe => xNiGe+ GaAs (regrown).

Again Ge was suggested to be incorporated into regrown GaAs, forming n -GaAs. There was a minimal amount of Ge (29 - 38 at. %) in Ni/Ge thickness ratio to form ohmic contact. The regrown GaAs was characterized by a high density of stacking faults,








microtwins, and precipitates. The regrown GaAs was postulated to contain a high concentration of dopant (e.g. Ge with a concentration of low-1019cm'3). However, no experimental data was presented to support the incorporation of Ge (1019 - 1020cm3) nor the formation of n'-GaAs.

An important unresolved issue for the regrowth mechanism is the specific mechanism of incorporation and site selection of doping elements into regrown GaAs [San88]. One speculation concerning Ge transport is a higher diffusivity and/or solubility of Ga in the metallization layers than As. Thus, prior to regrowth of GaAs a greater number of Ga atoms diffuse from the ternary phases into the outer layers of the metallization. When regrowth takes place resulting from decomposition of PdyGaAs or NixGaAs, regrown GaAs is deficient in Ga relative to As. Thus Ge could occupy Ga sites in regrown GaAs. This speculation also assumes Ge in-diffusion to the NixGaAs or PdyGaAs phases before regrowth, which will be shown to be not correct by our study.

A Ga vacancy-dependent diffusion model for ohmic contacts was suggested as another attempt to explain how Ge atoms are transported and occupy Ga sites [Gup85]. This model postulates grain-boundary diffusion in Au layer to generate Ga vacancies to account for the formation of ohmic contacts. This model is based solely on the role of Au, which enhances out-diffusion of Ga. Thus this model is not capable of explaining how another metallizations without Au still can form ohmic contacts. They also justified their model based on the fact that contact resistance is high after either low or high temperature annealing, but low for intermediate temperatures. However, this behavior is universally observed regardless of the presence of Au.








31
The results of our study clearly showed that neither the site selection nor the transport of Ge was so simple as suggested by the diffusion model or speculations. The site selection and transport of Ge are determined by evolution of interfacial phases followed by regrowth of GaAs.



AuGeNi/GaAs Metallizations



AuGeNi metallization was first introduced in 1967 by Braslau et al.to form ohmic

contacts to GaAs-based microwave devices [Bra67]. This method is still used most often to form ohmic contacts to n-GaAs. At first, the effect of Ni in the AuGe eutectic composition (Au 88wt%, Ge 12wt%) was to prevent the AuGe melt from "balling up" due to surface tension. Therefore, the three elements were evaporated so as to form a bilayer structure, i.e., Ni/AuGe/GaAs. However, it has been shown by several investigators that the sequence of each layer did not affect the final metallurgical properties or electrical properties [Chr79, Mar83]. As a second postulate, it was suggested that Ni improved the uniformity of surface morphology by improving the wetting of liquid Au-Ge to GaAs [Rob75]. The high reactivity of Ni was correlated to dissociation of GaAs and the following reaction was hypothesized:

Au + Ni + GaAs => Au-Ga + Ni-As.

where Ni was thought to be a catalyst [Oga8O]. During the reaction, Ge was trapped in the Ni-As compounds near the GaAs interface. Dissociation of GaAs by Ni was proposed to occur through solid-phase reaction which should lead to uniform alloying behavior in








the Ni/AuGe/GaAs [Oga8O]. A major difference from the dissociation of GaAs by Au versus Ni/Au was that As was retained by the stable Ni-As compounds. The ability of Ni to penetrate thin native interfacial oxides also led to a uniform reaction.

The Au-Ge eutectic was not thought to play a major role in the morphology by forming a liquid phase at 363�C since the Ge concentration for the Au-Ge eutectic was expected not to reach nor maintain because Ge was gettered by Ni via solid-state diffusion [1i84, Re189]. Again, this could improve the surface morphology compared to AuGe/GaAs. It was suggested that Ni changed reaction between AuNiGe and GaAs by initiating the reaction with GaAs, resulting in a smoother interfacial morphology [Shi87]. Near the AuGe eutectic temperature, Ni and Ge accumulated at the interface with GaAs, while Ga from GaAs dissociation built up at the surface [Rob75]. The barrier height, ctbn, increased from -0.7eV to -0.9eV as Ni accumulated at the interface [Rob75, Lav86].

When annealed above the eutectic temperature at 400 to 6001C, the surface

morphology became nonuniform by formation of rectangular, submicron reaction [Chr79, Oga8O, Hei82, Pro87, Buh91]. A higher concentration of Ni was found in the pits versus in the surrounding matrix, and postulated to control the specific contact resistance [Chr79]. Ni-Ge clusters were found on the surface and their distribution was directly correlated to the formation of ohmic contact [Hei82]. The surface morphology and contact resistance were direct functions of the Ni/Ge ratio. The higher the Ni concentration led to better surface uniformity, while higher Ge concentration led to lower contact resistance [Pat86, Kov9O, Buh9l, Chu94].








33
Ni is always found to diffuse toward the interface and to react with GaAs. Ge tended to follows Ni distributions. Au was a rapid diffusing species, and often formed Au-Ga compounds upon cooling [Rob75, Witt77, Chr79, Oga8O, Hei82, Kua83, Mar83, Ili84, Re189, Kim90]. A transmission electron microscopy (TEM) study showed that elemental distributions were correlated to the formation of some compounds. Generally, annealing above 4000C for a few minutes produces the following compounds; NiGe with a small concentrations of Ga and As, NiAs with Ge and Ga, and AuGa [Kua83, Shi87]. Spatially, the NiAs phase was in direct contact with GaAs resulting in a smooth interfacial morphology. A good ohmic contact with low contact resistance was attributed to the Ni2GeAs phase being in contacting with GaAs. Diffusion of Ge from Ni2GeAs into GaAs to form n' layer was proposed [Kua83]. NiAs phases containing small concentrations of Ge were suggested as tunneling ohmic contacts and the contact resistance was a function of the NiAs(Ge) coverage at the interface [Shi87]. Besides NiAs phases, AuGa compounds were found near the free surface [Shi87, Re189]. During the initial stage of annealing, the high concentration of Au resulted in AuGa phases at the interface with irregular morphology, which generally increased the contact resistance. However, at higher annealing temperature, more NiAs phase with a highly oriented epitaxial relation was found at the interface [Kim90]. The epitaxy was suggested to result from a close lattice matching between NiAs and GaAs. It was noted that better ohmic behavior was produced when the phase contacting GaAs was changed from AuGa to NiAs(Ge) phase.

The metallurgical reactions between AuGeNi/GaAs were different from those between AuGe/GaAs due to the high reactivity of Ni with GaAs. Dissociated Ga and As formed







34
compounds with Ni, and uniform interfacial and surface morphologies are observed. As a common feature, Ge was found at the interface with GaAs for alloyed metallizations with and without Ni and led to formation of ohmic contacts.

Even though the metallurgical reactions in AuGeNi/GaAs are different from

AuGe/GaAs, the mechanism for formation of ohmic contact appears to be dissociation of GaAs followed by alloying with the metallization elements. Upon cooling, Ge was postulated to be incorporated into GaAs by occupying Ga sites creating a heavily doped GaAs layer with _ 1019cm3 Ge. This concentration of Ge leads to an n' layer through which electrons could tunnel [Bra81]. This mechanism is broadly known as the "doping" model, and it is widely accepted to n-GaAs to account for the formation of ohmic contacts. Backside SIMS measurements in an alloyed AuGeNi system were used to resolve Ge doping [Bru89, Sch90]. Diffusion of Ge and Ni into GaAs was observed. However, the major reduction in contact resistance was poorly correlated the observed Ge [Bru90]. Cross-sectional TEM showed that the NiGexAsy phase was in contact with the substrate. Based on these result, Bruce et al. suggested an increased n -doping level was less important to the reduction of the contact resistance than the formation of Ni-Ge-As phase [Bru90]. It appears that these measurements are not clear data to support Ge incorporation into GaAs because of the lack of high spatial resolution and quantification.








Hybrid Contact Metallizations



Many variants to AuGeNi systems have been developed in order to improve thermal stability and to prevent interdiffusion. For high-temperature applications of microelectronics, the solid-state devices are required to function reliably for long time in hot environments (>3000C). And the interdiffusion is considered to be particularly important in the case of a shallow junction.

W60N4o has been added to the AuGeNi system to prevent the reaction between Au and GaAs and the solid-phase reaction of the Ni/Ge layer could form ohmic contacts [Ko188]. A Cr layer was inserted between Au and Ge to prevent the formation of Au-Ge eutectic [Wi188]. Similarly, WSi2 layer acted as a diffusion barrier between Au and Ge to prevent both Ga and As out-diffusion as well as Au-Ge eutectic formation [Gup90]. These metallizations formed ohmic contacts with contact resistances of 10-5 - 10-6 f2cm2, and exhibited smooth morphology above -600'C by preventing interdiffusion between elements.

Indium was added to metallizations to study the formation of graded GaxInl.,As heterostructure in addition to the formation of n+-GaAs. Indium films of 50 - 200 thickness were added between the Ni and Ge layers and shown to reduce contact resistance by forming GaIn..As (x-0.4) [Oku94]. TEM images of this study showed that the metal/GaAs interface was covered by regrown GaAs and GaxIn1.As. Compared to Ni/Ge bimetallization, the contact resistance was reduced from -1.2 Qmm to --0.3 Qmm in this study, which was attributed to the formation of n+-regrown GaAs and GaxInlxAs.











In Situ Contact Schemes



Contrary to the ex situ contact scheme, in situ contact schemes refer to the processes which do not utilize alloying reactions or solid-phase reactions to incorporate doping elements into the surface of the GaAs substrate. Instead, very heavy doping is accomplished in situ during the growth of GaAs, or heterojucntions are formed to lower the barrier height between metals and GaAs. Heterojunctions are defined as junctions between two different semiconductors. In principle, they may be either gradual or abrupt junctions. A gradual junction is one in which two bulk crystals are joined by continuously varying composition, e.g., GaxIn1 .As/GaAs where x can vary from 1 to 0. An abrupt junction is one in which there is a sharply defined interface between two homogeneous semiconductors, e.g., Ge/GaAs. In situ contact schemes are described as nonalloyed schemes because the ohmic contacts are formed without post-deposition heat treatment. Nonalloyed schemes were studied and developed in efforts to reduce the drawbacks of conventional metallization schemes, such as poor morphology, reproducibility, severe interdiffusion and reactions with GaAs, and poor thermal stability.

Molecular Beam Epitaxy (MBE) is frequently used to grow heavily doped n -GaAs where n' can be as high as -mid-1019cM-3. The reason MBE can produce n-type GaAs layers with a carrier concentration higher then most conventional growth techniques is that the incorporation of the dopant is not limited by solubility or thermodynamic equilibrium conditions, but may be controlled by surface kinetics. Thereby, MBE can be used to








37
incorporate ten times more carrier than the concentration obtained in bulk doping. Sn was selected because it is known to be a less-compensated amphoteric dopant versus Si or Ge. A free electron concentration as high as 6 x 1019cm3doped-GaAs could be achieved with Sn [Bar78, Dil79]. In cases where high-1019cm'3dopant concentration in GaAs were achieved, specific contact resistance as low as 2 x 10-691 cm2 were obtained simply by depositing metals without heat treatment. Even -1 x 1020cm-3 Si doping was reported and yielded -1.3 x 10-6 K cm2 with in situ metallization [Kir85].

Formation of ohmic contacts was attempted through creation of low barriers between metal/Ge layers in conjunction with favorable conduction band alignment at the Ge/GaAs heterojunction. The localized potential barriers of -0.45eV at the metal/Ge interface and less than 100 meV at the Ge/GaAs heterojunction are more favorable for ohmic behavior at room temperature than a single -0.8eV barrier at the metal/GaAs interface. The Ge layer must be heavily doped, up to -1020cm3, often with As while GaAs was doped with Ge or Si on the order of 1017 .101Scm3. Low contact resistances, typically -106fcm2 [Dev80] and even as low as -10"72cm2 [Sta81 ] were measured with smooth interfacial morphology. The Ge/GaAs heterojunction was studied because Ge has nearly equal lattice parameters (within 0.5%), compatible crystal structure, and the thermal expansion coefficient of Ge (6.6 x 10-6/C) matches well with that of GaAs (6.0 x 10"6/oC). Most importantly, the conduction band discontinuity at the Ge/GaAs interface, AEc, is - 60meV which will exhibit a negligibly small tunneling resistance [Sta81, Ba183].

Another heterojunction scheme is to Gal.ln.nAs at the interface. This scheme is based on the fact that Fermi level pinning occurs at or in the conduction band on InAs [Woo81 ].







38
Because the conduction band discontinuity between InAs and GaAs establishes an barrier, a graded Gal..InxAs between InAs and GaAs is necessary to remove the abrupt discontinuity in the conduction band. In this scheme, tunneling is not required and low resistance contacts can be made for a wide range of doping without need of alloying to form n' surface layers.

InAs and Gal.InxAs layers were grown in situ on GaAs typically by MBE process and produced contact resistances of-10"7fgcm2 [Kum89, Meh89]. This was at least one order of magnitude lower than the contact resistance achieved by ex situ contact schemes, which are typically, - 10"60cm2.

Even though InAs-based contact or in situ heavy doping scheme provided an extremely low contact resistance, these schemes are not practical because of the incompatibility of MBE with semiconductor processing and the incapability of the InAs-based structures to endure high temperature processing.



Current Status of Contact Metallizations



In pursuit of lower resistance ohmic contacts, efforts to improve AuGeNi

metallizations are being made [Chr95, Hao96]. In terms of understanding for formation of ohmic contacts, the solid-phase regrowth model is the most recent development and is being applied to explain not only formation but also degradation of ohmic contacts [Wan95]. Important refinements on the solid-phase Regrowth model have been discussed by Holloway et al [Hol91, Li92, Lam92, Kim96]. They clearly demonstrated that








formation of ohmic contacts coincides with regrowth of GaAs, and contact resistances were a function of morphology of regrown GaAs [Li92]. The results they reported, the measurement of free carrier density (1019 - 1020cm'3) in regrown GaAs, is only the direct data to date for formation of n' layer in regrown GaAs [Li92]. Metallurgical concentrations of Ge up to -1020cm-3 as a function of the extent of regrowth is reported for the first time in this dissertation. Therefore, it is believed that Ge incorporation during the regrowth creates n' surface layer, resulting in tunneling ohmic contacts. Equally important to these data, microscopic reaction paths through which Ge could occupy Ga sites in regrown GaAs were clarified for the first time [Kim96]. Our results are believed to be the starting point to more precisely evaluate current metallizations.

It should be acknowledged that experiments by other researchers using simpler system such as Ni/GaAs [Oga8O, Lav86, Che88, Gui89] and thermodynamic interpretation [Bey84, Moh95] are also important contributions to the understanding of ohmic contacts.










Vacuum level

Om


Metal


Vacuum level


Ev


n-Semiconductor


(a) Before contact


Vacuum level


Metal


Vacuum Is level n-Efs

n-Semiconductor


x=0


(b) After contact


Figure 2-1.


Energy band diagram for metal and n-semiconductor in electrical and thermal equilibrium. (Dm, (D, are the work function of the metal and semiconductor, respectively.
















n-Semiconductor


Vacuum levels


Ess lee


"-Vacu level ,x Ec
- Efs Eg


Figure 2-2.


Energy band diagram for a Schottky barrier formed by a metal and n-semiconductor with surface states. Ess is the surface state energy levels within the bandgap. (m, (D. are the work function of the metal and n-semiconductor, respectively. This diagram assumes acceptor-like surface states.


m


Efm


A

O1

_ J


Metal















CHAPTER 3
EXPERIMENTAL PROCEDURE


The majority of the experimental work for this dissertation was carried out at the

Department of Materials Science and Engineering at the University of Florida, Gainesville, Florida. The samples were prepared by using an electron-beam evaporator in the MICROFABRITECH, University of Florida, and characterized by using analytical instruments in the Major Analytical Instrumentation Center. High spatial resolution EDS analysis was carried out in the metal/ceramic division at the Oak Ridge National Laboratory, Oak Ridge, Tennessee under the SHaRE program.



GaAs Substrate Treatment



All the GaAs wafers used in this study were n-type GaAs grown and bulk doped with Si by LEC (Liquid Encapsulated Czochralski) in Sumitomo Ltd., Japan. The wafers were doped with two different levels, 2 x 1016cm"3 and 2 x 1018cm'3 as measured by the Hall technique at room temperature. The wafers were cut on the { 1001 � 0.50 plane and had average thicknesses of 400 - 500 ptm. In order to be mounted onto the sample holder,









GaAs sections were cleaved into rectangular-shaped I to 2 cm small pieces along <110> planes.

The GaAs pieces were cleaned in a ultrasonic bath of TCE, acetone, and methanol followed by N2 blow drying. Prior to evaporation of metal films onto the surface, the GaAs substrate was cleaned in HCL:H20 (1:1 by volume) for 60 seconds to reduce the native oxide and then mounted on the sample holder. The sample holder was made of Ta sheet and a ceramic block. After the mounting, the holder was immediately loaded into the vacuum chamber of the electron-beam evaporator. The chamber was promptly pumped down. To anneal samples without breaking vacuum, a resistive wire radiation heater was located behind the sample holder as illustrated in Figure 3-1.



Metal Evaporation and In situ Annealing



Samples with in situ annealing.



After the loading the sample, the chamber was pumped by a Varian M-6 diffusion

pump system with a liquid nitrogen trap. When the base pressure reached - 1 x 10-6 Torr or less, the sample holder was radiatively heated using a regulated DC-power supply which was adjusted to generate -15A (ampere) DC current to the Ta radiation heater. When the temperature reached -61 0C in 1 minute, the samples were maintained at the temperature for 15 seconds to desorb residual oxides on the GaAs. After the -15 second hold, the power to the Ta heater was turned off and the sample was allowed to cool down







44
to room temperature under vacuum. The temperature of the sample was determined by a

0.125mm diameter K-type thermocouple embedded between the sample and the Ta substrate holder. This heating was also thought to cause outgassing of any volatile residual impurities on the heater and the sample holder.

When the base pressure reached mid-10-7 Torr or lower. First, 650A of Ni film was evaporated onto the GaAs substrate. During Ni evaporation, the base pressure increased to less than 1 x 10- Torr. General evaporation conditions were electron beam current

-45mA with acceleration voltage of 8kV, and deposition rate of -3,A/sec. The thickness of Ni layer was monitored by a crystal oscillation monitor, QXM-500 purchased from Kert J. Lesker Company.

After Ni evaporation, an annealing step was carried out using the radiation heater

without breaking vacuum. The base pressure before the annealing was typically 4 - 7 x 10"7 Torr. The current to the heater was manually adjusted by monitoring the temperature of the sample. The sample temperature reached 3000C in one and a half minutes and stabilized at 300 � 40C with a heater current of 7 ampere. During annealing, the base pressure rose for the first -1 minute to high 10-7 Torr, then dropped back to the initial pressure. Annealing step caused the Ni to react with GaAs to form an interfacial layer, Ni2.4GaAs ternary phase. Deposition of 650A of Ni followed by an annealing at 3000C for 15 minutes without breaking the vacuum was the standard process for this study and will be referred to in situ annealing. In addition to 3000C, some 650A Ni films were in situ annealed at 4000C for 15 minutes. In situ annealing without a specification of temperature automatically indicates 3000C for 15 minutes. Since the primary purpose of this study was








to investigate the evolution of interfacial phases consisting of Ni and GaAs and their effects on ohmic contacts, the Ni2.4GaAs phases were produced as described above.

After the in situ anneal, samples were allowed to cool to -60'C and a Ge layer was evaporated at a rate of-3ksec. During the Ge evaporation, the sample temperature increased to -80'C. The pressure during Ge evaporation was less than 8 - 9 x 10-7 Torr. The thickness of Ge layer was varied from 250 to 750A.

Evaporation of Ge was followed by Ti evaporation onto the Ge layer at a deposition rate of -4A/sec. Unique to Ti evaporation was that the base pressure did not increase during the evaporation due to gettering pumping by the Ti film. The sample temperature before Ti evaporation was -60 -70'C and increased to -100�C when evaporation was completed. The thickness of Ti layer was varied from 250 to 750A.

The overall structure with the in situ anneal will be designated as GaAs / Ni24GaAs / Ge / Ti. Some in situ annealed samples were prepared only with Ge or Ti layers.



Samples without in situ annealing. In addition to samples with in situ annealing, a number of samples were prepared without the in situ annealing for comparison. The entire procedure was the same except that the first 650A Ni was not in situ annealed before the Ge and Ti evaporation. Thus the layer structure without the in situ anneal were GaAs / Ni / Ge / Ti where the thickness of Ge and Ti were varied from 300A to 900K.

In the case of the samples with in situ annealing, it took approximately one hour to prepare from the Ni deposition to the Ti deposition. And for the samples without the in situ annealing, it took only -15 minutes. Throughout the evaporation, two types of









shadow masks, circular shape of 0.5mm diameter made of stainless steel and lines of

-320prm x 4000 ptm made of Ta, were used over the substrate.



Post-Evaporation Annealing



After the evaporation was completed, samples were exposed to atmosphere and cut into two piece. One was kept for an as-deposited sample and the other piece loaded into the electron-beam chamber, and annealed in vacuum to avoid oxidation of the Ti. When the background pressure was 1 - 2 x 10-6 Torr, the sample was annealed at 500'C for 5 minutes by the radiation heater. The standard annealing temperature in this study, 5000C, was reached within 2 minutes and stabilized at 500'C � 5. During the heating the background pressure reaches < 3 x 10-6 Torr for the first 1 min and gradually decreased back to 1 - 2 x 106 Torr during the remaining time. Some samples were annealed at 530, 550, and 600'C for 20 and 35 minutes.



Electrical Measurements



As-deposited samples and vacuum annealed samples were electrically characterized by their current-voltage (I-V) characteristics. These measurements were conducted at room temperature. For moderately doped semiconductors, the I-V characteristics is controlled by thermionic emission and the Schottky barrier height. Generally, I-V data was obtained by measuring the current flow between two front surface dot contacts. The resulting








current above breakdown can be extrapolated back to zero, which represent the reverse bias breakdown voltage for one of the two back-to-back Schottky diodes formed by the metal-semiconductor contacts. Changes in this breakdown voltage may be related to changes in barrier height, Jt. For ohmic contacts, the linearity was typically on the lmA/IV scales.

The I-V characteristics were measured and recorded using an Tektronix 177 curve tracer, and by an automated system which consists of an IBM PC with IEEE-4888 communications, a Hewlett-Packard 6112A DC power supply, and a Hewlett-Packard 3478A multimeter. Some of the measurements on the Tektronics curve tracer were instantly recorded with Polaroid photos. Typically more than tow contacts were measured per sample and the curves shown are average, among the many contact dots or lines.



Analytical Characterizations



Auger Electron Microscopy (AES) and Secondary Ion Mass Spectroscopy (SIMS) were used to obtain compositional information. Transmission Electron Microscopy (TEM) was used to investigate the microstructure and phases of the samples. These data were correlated with electrical properties of the samples.









Auger Electron Spectroscopy (AES)



The AES technique has been used extensively to investigate the interdiffusion in a variety of materials [Smi94]. Since the information depth is determined by the escape depths of the Auger electrons, which are typically 0.5 - 1.Onm, destructive depth profiling by ion sputtering along with simultaneous measurement of Auger peak intensities is used to determine the distribution of elements with respect to depth. AES was used in this study was to measure interdifflusion within the entire layer structure. Since the sensitivity of AES is limited to - 1 at %, the data collected by AES was cross-checked with the data collected by SIMS, which has a much lower sensitivity.

A Perkin-Elmer PHI 660 Scanning Auger Electron Spectrometer was used to obtain depth profiles in this study. Typical electron beam parameters were 5 KeV primary electron beam energy with a 30nA beam current. For depth profile, 5 KeV Ar ion gun with rastered current 25nA was used for sputtering.



Secondary Ion Mass Spectroscopy (SIMS)



Secondary ion mass is an appropriate technique to determine elemental distributions because of its high sensitivity at least an order of magnitude better than AES or electron microprobe analyses [Reu81]. The depth resolution of SIMS is dependent upon several variables such as surface roughness and thickness nonuniformity, ion beam mixing, knockon effects, and preferential sputtering [Wil89]. Since the sputtering yield of each








49
constituent is different, a slightly roughened surface will be produced, and the depth/time is different for different layers. Also, the primary ions mix the elements and degrade the resolution. Nevertheless, valuable information with high sensitivity was obtained from SIMS which complement the data from cross-sectional TEM and AES depth profiles.

SIMS data were collected with a Perkin-Elmer 6600 system with a 5 KeV Cs' primary ion beam and positive secondary ion detection ( CsX+ cluster ion spectrometry, where X is the impurity element of interest). The primary ion beam current was 40 nA and the raster size was 400 x 400 jtm2 with 55% gating (detected area: 220 x 220 Pm2).



Transmission Electron Microscopy (TEM)



Unlike SIMS and AES, TEM requires special and extensive sample preparation. First, samples were sliced into strips -250jtm thick by using a dicing saw. The strips were bonded face to face with M-bonding followed by curing in an oven maintained at 1 10�C for 1 hour. Bonded strip samples were mechanically ground down to 15 - 20 gtm, then mounted on the 3mm diameter Cu ring for mechanical support. Finally, the sample on the Cu ring was further thinned using a Gatan ion miller until a -0.2mm diameter hole was visible. Typical ion milling conditions were a 4kV, 1 mA beam at a milling angle of 13 17'. Sample preparation is a critical step for successful TEM analysis, and extreme caution was essential.

In TEM, the whole area of interest is simultaneously illuminated by the electron beam contrary to STEM (Scanning Transmission Electron Microscopy) where a high density







50
electron is required in a small probe. In TEM, the conventional image is obtained using an aperture in the back focal plane which allows only one diffracted or transmitted electron beam to form the image. Contrast in the image may result from either diffraction or phase. A bright field image is formed if the directly transmitted beam is selected and a dark field image if a diffracted beam is selected.

A diffraction pattern is formed on the back focal plane of the objective lens in the transmission electron microscope. Hence, if the diffraction pattern is focused onto the back focal plane of the objective lens and the objective aperture is removed, the diffraction pattern will be visible on the viewing screen.

In this study, most TEM data were collected using a JEOL 200CX for analytical or a JEOL 4000FX for high resolution analysis. In the JEOL 200CX, electrons emitted from tungsten filament were accelerated to 200keV. The majority of the bright field images and diffraction patterns were taken along the [110] zone axis of GaAs. However, some images and diffraction patterns were taken along the [ 111] zone axis of GaAs for comparison. In the JEOL 4000FX, a LaB6 filament emitted electrons which were accelerated to 400keV. In some samples, energy dispersive analysis of fluorescent X-rays (EDX) were used. The spatial resolution of the EDX analysis was 600A.

High spatial resolution X-ray microanalysis was performed in a Philips EM400IFEG

analytical electron microscope equipped with a field emission gun and EDAX 9100 energy dispersive X-ray spectrometry (EDX). The voltage for the EM400 was 100kV and the probe size was -1.5nm in the STEM mode and -5 -10nm in the TEM mode. Composition measurements were performed in the STEM mode with a -1.5nm diameter, -0.67nA








probe. A low-background, cooled holder (-1300C) was used to minimize contamination under the focused probe. The "in-hole" spectrum from the sample was corrected for the normal "in-hole" counts associated with the fluorescence from the entire specimen area resulting from uncollimated radiation. The convergent beam electron diffraction pattern was used for precise location of the interphase boundary by reflections from both phases. The primary beam position was repeatedly checked during EDX acquisition in order to minimize any specimen drift. A standardless quantification program, NEDQNT2, was used to convert the observed peak intensities to composition for elements with atomic number Z > 10 [Lor94].





























Tl WTa wire for radiation heating


Shadow maskGaAs substrate . evaporating flux

-electronbeam .



tungstet-'lament C heart


Fig. 3-1. Simplified schematics for electron beam evaporation of contact metals.


Ta sample holder \,















CHAPTER 4
RESULTS


Summary of Prepared Samples



TEM data revealed that a 650A thick Ni film in situ annealed (300'C for 15 minutes) produced 1200 - 1300A of a Ni2.4GaAs layer, which will be discussed below. Samples were prepared as detailed in table 4-1 through 4-3, GaAs substrate without a specified doping concentration automatically indicates -2 x 10'8cm-3. All the samples in table 4-1 through 4-3 were as deposited i.e., before vacuum annealing at 500'C. As-deposited samples were all rectifying as will be described in the electrical measurements.


Table 4-1. Samples containing 650A Ni film in situ annealed at 300'C, 15 minutes.
Layered structures Ni2.4GaAs (A) T Ge (A) Ti( ) GaAs/Ni2.4GaAs /Ge 1200 - 1300 250 0 1200-1300 330 0 1200- 1300 500 0 GaAs/Ni24GaAs/It 1200- 1300 0 600 GaAs/Ni2.4GaAs/Ge/It 1200 - 1300 250 300 1200-1300 330 200
GaAs(1016)/Ni2.4GaAs/Ge/Ti 1200- 1300 250 300 1200-1300 500 300 1200-1300 750 300 1 1200- 1300 250 900
GaAs(1016) indicates -2 x 1016cm-3.








As stated in Chapter 2, Ni/GaAs reaction was reported to produce Ni-As and Ni-Ga binaries at temperatures above 4000C [Oga8O, Lav86, Gue89]. Data from the present study showed that a 650X Ni film in situ annealed at 4000C, for 15 minutes produced a combination of Ni.GaAs and NiAs phases. The thickness of the reaction layer was 1400 1500A from TEM measurements.


Table 4-2. Samples containing 650A Ni film in situ annealed at 400'C, 15 minutes.
Layered structure (A) NiGaAs+ NiAs Ge Ti GaAs(1016)/Ni.GaAs+NiAs/Ge/Ti 1400-1500 250 300 1400-1500 750 300 1400-1500 500 600 1400-1500 250 900


Samples without in situ annealing are shown in Table 4-3.


Table 4-3. Samples without Ni film in situ annealed.
Layered structure Ni (A) Ge (A) Ti(A)
GaAs / Ni / Ge / Ti 650 250 300 650 250 600 650 250 900


Electrical Measurements



All the I-V curves presented here were measured between two circular dots with 0.5

mm diameter which were the most representative of the corresponding structures.







55
GaAs/Ni. Figure 4-1 is typical example of the I-V curves measured for GaAs / 650A Ni. As-deposited sample exhibited well defined rectifying I-V characteristics with breakdown voltage of -2.OV on GaAs doped with 2 x 1018cm-3 Si. After in situ annealing at 300'C, for 15 minutes, rectifying behavior was still measured but with a barrier reduced to -1.2V. Since the in situ annealing produced the Ni2.4GaAs phase, (see below), the I-V characteristics represents the contacts between this phase and GaAs, i.e., GaAs/ Ni2.4GaAs.



GaAs/ Ni2.aGaAs /Ge. Figure 4-2 is a typical example of I-V data from GaAs I

Ni2.4GaAs / 250A Ge samples. The data were also typical for GaAs / Ni2.4GaAs / 330k Ge samples. As-deposited and vacuum annealed contacts. (i.e., annealed for 2.5 and 5 minutes after Ge evaporation) exhibited rectifying behavior, although the vacuum annealed contacts showed a reduced barrier. Contacts of GaAs / Ni2.4GaAs / 500/ Ge exhibited a linear I-V characteristics as shown in Fig. 4-3, after vacuum annealing at 500'C for 5 minutes, indicating the transition from rectifying to ohmic was a function of Ge thickness.



GaAs / Ni2.4GaAs / Ti. The I-V characteristics from as-deposited and vacuum

annealed sample of GaAs / Ni2.4GaAs / 600A Ti showed rectifying barriers (figure 4-4) similar to those from the GaAs / Ni2.4GaAs / 250k Ge. However, the annealed contacts exhibited an increased barrier.








GaAs / Ni2.4GaAs / Ge / Ti. Figure 4-5 shows the I-V characteristics of GaAs / NiI4GaAs / 250A Ge / 300k Ti and GaAs / Ni24GaAs / 330A Ge / 200A Ti. These structures exhibited ohmic behavior after 5000C for 5 and 20 minute vacuum annealing. Since the vacuum annealed structures of GaAs / Ni2.4GaAs / 250k Ge and GaAs / Ni2.4GaAs / 330, Ge did not show linear I-V, this indicates that 200 - 300k of Ti is necessary to observe ohmic behavior after vacuum annealing.



GaAs / Ni / Ge / Ti. Figure 4-6 show the I-V data measured from GaAs / 650k Ni / 250k Ge / 300A Ti (where the nomenclature indicates that Ni film was not in situ annealed before evaporation of Ge and Ti). The I-V characteristics before and after vacuum annealing were rectifying. Since this structure did not show linear I-V while GaAs / Ni2.4GaAs / 250k Ge / 300k Ti did, in situ annealing of 650A Ni played a critical role in the formation of ohmic I-V characteristics. To determine the effects of Ti thickness, Ti was increased from 300k to 600k or 900k. The I-V data in figure 4-7 shows that as-deposited and 500'C, 5 minute vacuum annealed exhibited little change in the I-V curves. A noticeable reduction in the Schottky barrier height was measured for 20 and 35 minute annealing. Figure 4-8 shows the results from GaAs / 650k Ni / 250k Ge / 900k Ti. There was no significant change in I-V characteristics between 5 minute and 20 minute annealing, unlike for the samples shown in figure 4-7. As can be seen in these results, the increase of Ti thickness with a fixed thickness of Ge layer did not cause the change from rectifying to ohmic behavior.









Samples using 1016cm-3 doped GaAs. To investigate the effects of doping

concentration, GaAs / Ni2.4GaAs / Ge / Ti structures were prepared using 2 x1016cm3 doped GaAs. As-deposited sample of GaAs / Ni2.4GaAs / 250k Ge / 300A Ti exhibited rectifying contacts with a breakdown voltage of-25V. Note the breakdown voltage was only -2V for GaAs doped 2 x 1018cm-3. Because of its extremely low current level, it appears that no current flows on the scale of figure 4-9. After vacuum annealing at 500C, 5 minutes, higher levels of current were measured although the contact still rectifying contacts versus 2 xlO'cm"3 doped GaAs which exhibited ohmic behavior. The effects of increasing the Ge thickness from 250 to 750A are shown in the figure 4-9. With 750A Ge, the currents increased consistent with a reduced the barrier height. With 900A Ti, less current was measured. All the sample on the figure 4-9 were rectifying. These results suggest that the amount of Ge is an critical factor for improving the I-V characteristics.



Samples in situ annealed at 4000C. Since no structures on 2 x 1016 doped GaAs exhibited linear I-V characteristics with an in situ Ni anneal at 300'C for 15 minutes, samples with an in situ Ni anneal at 4001C for 15 minutes were studied. The phases produced by the reaction between Ni and GaAs at 400'C were NixGaAs + NiAs compounds. The I-V characteristics from the several samples are shown in figure 4-10. Linear I-V characteristics were measured with 500A or 750A Ge, 300A Ti.








58
Summary of electrical measurements. Linear I-V characteristics were not observed unless Ge was present, therefore, Ge was essential for ohmic contacts. In the structures comprised of 1200 -13 OOA Ni2.4GaAs on 2 x 1018cm3 GaAs, the transition from rectifying to ohmic behavior required Ge films thicker than -500A or 200 - 300A thick Ti layer with 250A. Ge. In situ annealing also helped form ohmic contacts. For metallization on the 2 x 1016 cm"3 GaAs, minimum thickness of Ge layer to form ohmic contacts was 750k.



Elemental Depth Profiles



GaAs / Ni2.4GaAs / Ge. The SIMS depth profile of the GaAs / Ni2.4GaAs / 250A Ge as-deposited and annealed are shown in figure 4-11. As pointed out in Chapter 3, the CsX+ signals were measured but are simply denoted as X (e.g. CsGe' is designated as Ge). In this profile, Ge and a layer of Ni2.4GaAs (see below for composition analysis) are distinguished. In the Ni2.4CaAs layer, the Ni, Ga and As signals are all stable, consistent with formation of a ternary compound. A strong peak of carbon was always detected at the Gel Ni24GaAs interface, as shown in the figure 4-11. This carbon is thought to originate from adsorption of diffusion pump oil during the in situ anneal. The carbon layer was present before and after vacuum annealing, therefore it served as a marker layer indicating the original interface between Ge and the Ni2.4GaAs. Oxygen was also normally detected at the interface but is but not plotted in the figures (for clarity).

The SIMS profile after vacuum annealing at 500'C, 5 minutes shows that interdiffusion took place. Using the carbon marker, Ge diffused into the Ni2.4GaAs region and Ni and







59
Ga diffused beyond the carbon layer towards the free surface. The concentration of As is low in the outer surface region of the original Ni24GaAs layer. Since the Ni between the carbon marker and the surface had to be supplied from the initial Ni24GaAs layer, at least some portions of the Ni24GaAs had to decompose which would allow release of Ga and As and could result in the regrowth of GaAs. The Ni24GaAs decomposition and GaAs regrowth was confirmed in TEM study (see below). Figure 4-12 shows AES depth profiles of GaAs / Ni2.4GaAs / 500A Ge. The profile after the vacuum annealing at 5000C, figure 4-12 (b), looks similar to that in figure 4-11.



GaAs / Ni2.4GaAs / Ti. Figure 4-13 shows the results of depth profiles of GaAs / Ni2.4GaAs / 600A Ti as deposited and vacuum annealed. In the as-deposited profile, figure 4-13 (a), the Ti signal peak at the interface between Ti and the Ni24GaAs probably results from matrix effects on secondary ion sputtering yield [Wil89]. The depth profile after a vacuum anneal at 5000C, 5 minutes showed severe interdifflsion of the elements, figure 4-13 (b). Based on the carbon marker layer, it is clear that Ti diffused into and reacted with Ni24GaAs layer.



GaAs / Ni2.4GaAs / Ge / Ti. Figure 4-14 shows the SIMS depth profile of the GaAs / Ni24GaAs / 250A Ge / 300A Ti. In the as-deposited profile (figure 4-14 (a)), Ti, Ge and the Ni24GaAs layer are distinguished with carbon peak at the Ge/ Ni24GaAs interface. In figure 4-14 (a), the Ti signal was detected beyond the Ge/ Ni2.4GaAs interface, suggesting that Ti is a faster diffusing species in this metallization.








Figure 4-14 (b) is the SIMS profile after vacuum annealing at 5000C, for 5 minutes,

with the carbon marker still evident. The metal layers reacted to form two different region at 500'C, 5 minutes, a Ti-Ni-As layer to the left of the carbon marker and Ni-Ga-Ge layer to the right of the marker layer. Arsenic in the top layer results from depletion of As in the previously Ni2.4GaAs. The majority of Ge diffused into the bottom layer where Ni and Ga are the matrix elements. A small portion of Ti diffused beyond the carbon layer, while almost the same Ni signal was detected in the top layer beyond the marker. This suggest that Ni diffused into the Ti layer while Ti did not diffuse significantly into the Ni2.4GaAs region. Since Ni out-diffusion into the Ti layer, a portion of the Ni2.4GaAs layer was decomposed to supply the Ni in the top layer. Decomposition resulted in regrowth of GaAs from release of Ga and As, as will be seen in TEM study.

In the depth profile of an annealed sample of GaAs / Ni24GaAs / 250A Ge (figure 4-11

(b)), only a slight out-diffusion of Ni from the Ni2.4GaAs layer was observed and this structure did not form linear I-V. In the case of adding 300A Ti or an additional 250A Ge (total 500A Ge) onto GaAs / Ni2.4GaAs / 250A Ge, the depth profile clearly showed that a larger amount of Ni diffused into the Ti layer or Ge layer over the marker and these structures formed linear I-V. From these results, it is postulated that decomposition of Ni2.4GaAs layer was initiated or driven further by the Ni-Ti reaction, resulting in the switch from rectifying to ohmic behavior. A similar structure of GaAs / Ni2.4GaAs / 600 Ti did not exhibited a linear I-V characteristics but decomposition of Ni2.4GaAs occurred by a Ni-Ti reaction similar to figure 4-13 (b). This indicates that Ge is a crucial element for the formation of ohmic contacts.










GaAs / Ni / Ge / Ti. Figure 4-15 (a) shows the depth profile of the as deposited three metal films (Ti, Ge and Ni) on the GaAs substrate without in situ annealing. A relatively strong peak carbon was detected at the Ni/GaAs interface (not shown in the figure).

Figure 4-15 (b) is the profile after vacuum annealing at 5000C, for 5 minutes. The carbon peak was detected at 6 minutes of sputtering time as shown. It is noted that the overall distributions of the elements look very similar to the figure 4-14 (b) in that the entire metal films consist of two layers, one to the left of the carbon layer and the other to the right of the carbon layer.

Despite the similar chemistry of the layer to that of vacuum annealed GaAs / Ni2.4GaAs / 250A Ge / 300A Ti (figure 4-14 (b)), the structure of figure 4-15 (b) did not form linear I-V characteristics. Since the overall chemistry of figure 4-14 (b) and 4-15 (b) appears to be nearly the same, the difference in the electrical property probably results from changes in the microstructure.

Since GaAs / 650A Ni / 250A Ge / 300Ak Ti did not form a linear I-V, a similar structure with 600A of Ti film (rather than 300A of Ti) was prepared and vacuum annealed. The depth profile (figure 4-16) and I-V data (figure 4-7) of this sample were similar to those from the sample with 300A of Ti after vacuum annealing at 5000C, for 5 minutes. Therefore more extensive Ni-Ti reactions failed to produce ohmic contacts after vacuum annealing for 5 minutes.

As annealing time increased to 20 and 35 minutes, a slight in-diffusion of Ti and outdiffusion of Ni were noticed while Ge remains unchanged as shown figure 4-17 (a) and








(b). Arsenic depletion at the interface with the top metal layer became more obvious in the figures. Since a reduction of the I-V was measured after 20 minute annealing, the slight further Ni-Ti reactions appears to be significant. With a 900A Ti layer, similar elemental distributions were observed after annealing for 5 minutes and 20 minutes as shown in figure 4-18 (b) and 4-19 (b). In this case, there was no noticeable reduction in the I-V characteristics.



GaAs samples doped 1016cm-3. The I-V data in figure 4-9 show that GaAs (1016) /

Ni2.4GaAs / 750A Ge / 300A Ti did not exhibit linear I-V behavior after vacuum annealing at 5000C, for 5 minutes, while the same layer structure in situ annealed at 4000C, for 15 minutes exhibited linear I-V (figure 4-10). Both samples were characterized by SIMS with depth profile from (1016) / Ni2.AGaAs / 750A Ge / 300A Ti shown in figure 4-20, while those from GaAs (1016) / NixGaAs + NiAs / 750A Ge / 300A Ti are shown figure 421. Note that Ni2.4GaAs is the nomenclature for the reaction products by in situ annealing of 650A Ni film at 3000C for 15 minutes and NixGaAs + NiAs for the reaction products by in situ annealing of 650A Ni film at 400'C for 15 minutes.

The depth profile in figure 4-21 (b) is very similar to that shown in figure 4-20 (b), but only the in situ annealed at 400'C, for 15 minutes resulted in ohmic contacts after vacuum annealing at 5000C for 5 minutes. Thus, the elemental depth profiles are not sufficient to account for the differences in the electrical behaviors of the two structures.








Transmission Electron Microscopy



Several structures were analyzed by TEM based on the results from the electrical measurements and the SIMS and AES depth profiles. Bright-field images, diffraction pattern and EDS spectra will be described. All the bright-field images were taken along the <110> zone axis of GaAs, unless otherwise specified.



GaAs / Ni2.4GaAs / Ge. Figure 4-22 (a) is a bright-field image of a GaAs / Ni2.4GaAs / 250A Ge as-deposited cross-section, where the Ni2.4GaAs designates the reaction product by the in situ annealing of 650A Ni film at 3000C for 15 minutes. In this particular brightfield image, the GaAs substrate was ion milled faster than the metal layers, showing a large ion milling hole. The diffiraction pattern from the entire structure is shown in figure 4-22

(b) and a schematic of the diffraction pattern is shown in (c). The Ni2.4GaAs phase was identified as a NiAs-type hexagonal ternary by its diffraction pattern, which was identical to the pattern of figure 2 in the reference by T. Sand et al [San86].

The diffraction pattern in figure 4-22 (b) was taken along the < 110> zone axis of

GaAs, so the { 111 }, {220} and (200} spots from GaAs are clearly visible (see figure 422(c)). The four small satellite peaks around each { 111 }G,A, spot indicate that the ternary phase is twinned. The streaky diffiraction spots from the ternary phase suggests that it is under some strain to accommodate twinned sections differently oriented to each other. In fact, the bright-field image in figure 4-22 (a) is a twinned image of Ni24GaAs. The brightfield image shows microtwinned sections with an average size of-IOOA or less. The









microtwins have the shape of elongated rhomboids parallel to each other and perpendicular to the GaAs interface. The contrast in the Ni2.4GaAs layer arises from diffraction, not compositional contrast.

The Ge layer on the Ni2.4GaAs layer was -250A thick. The morphology of the Ge layer is an island structure rather than a smooth layer (figure 4-22 (a)). The results of EDS analysis on the Ni2.4GaAs of figure 4-22 (a) is shown figure 4-23. K a lines of Ni, Ga, and As are displayed respectively in the spectrum. The Cu peak originated from the Cu rings used for the TEM sample support. This spectrum represents a typical composition from the Ni2.4GaAs phase. The intensity ratio of Ga and As signal in figure 423 is similar to that from the GaAs substrate in figure 4-24, supporting the 1:1 ratio of Ga and As in the ternary phase.

The as-deposited GaAs / Ni2.aGaAs / 250A Ge sample was vacuum annealed at 5001C, for 2.5 minutes and figure 4-25 shows the results of this annealing. In the bright field image (figure 4-25 (a)), the entire metal structure consists of two layers. The first layer is only -200A thick and the second layer is the ternary Ni2.4GaAs. At the boundary of these two layers, a bright, sparsely discontinuous line is observed. This line is thought to be the carbon marker produced during the in situ annealing and the following vacuum annealing. This carbon marker was consistently observed in all the structures in situ annealed. The identification of the carbon marker was more obvious when the sample was annealed for an additional 2.5 minutes (5 minutes total). By comparing the elemental depth profiles (figure 4-11) where the carbon divides the entire metal layer into two regions, its location








agrees with the TEM result and again played the role of marker between the outer metal layer and the Ni24CaAs layer.

There are clearly noticeable changes between the two bright field images of Ni2.4GaAs layer before (figure 4-22(a)) and after the 2.5 minute annealing (figure 4-25 (a)). The first difference would be the disappearance of the microsized twinned sections in annealed samples. Instead, the internal structure of the Ni2.4GaAs in the figure 4-25 (a) shows equiaxial grains. One of strongly diffracting (0112) type spots (indicated by an pointer in figure 4-25 (b)) was selected to form a dark field image and the dark field image is shown in figure 4-25 (c). This image clearly shows the formation of equiaxial grains with a diameter of -400A or larger (the grains with white contrast). These data indicate that the microsize twinned sections coalescenced into a lager, equiaxial grains.

The second difference of annealing is that the interfacial morphology between

Ni24GaAs and GaAs becomes non-planar upon annealing. The thickness of the Ni24GaAs layer was -1250A before and -1000A after annealing, indicating that an average of

-250A of Ni2,4CaAs was decomposed. Simultaneously, another phase began to form inside the ternary phase at the interface with GaAs. This phase will be shown to be NiAs containing Ga and Ge as minor elements and one of the grains is marked by arrows in figure 4-25(a) and figure 4-25 (c). Their average size was -500,. Note that the I-V characteristics of this sample was rectifying, similar to the as-deposited condition.

This same structure was annealed for an additional 2.5 minute at 5000C (5 minutes total) and figure 4-26 (a) shows a bright-field image of cross-section. The NiAs phase grew to - 1500,& parallel to the interface and -IOOOA vertically, while the average







66
thickness of the Ni24GaAs layer remained almost the same as for 2.5 minutes annealing. A large size of NiAs grain is indicated by an arrow in figure 4-26(a). In some grains of NiAs phase, TEM showed that microcrystals existed, probably with different orientations. Figure 4-26 (b) is a high resolution images taken in the sample shown in figure 4-26 (a). Note that the outer metal layer and top portion of the bottom layer (previously Ni2.4GaAs) were ion milled away in this area. The micrograph in figure 4-26 (b) shows the Moire fringes near the boundary between the NiAs and surrounding metal layer, suggesting that they are two dissimilar phases with different d-spacings, and are overlapped in this cross section.

To determine whether the NiAs was a new or simply a recrystallized, the structure of figure 4-26 (a) was analyzed by EDS and the results are shown in figure 4-27. Figure 427 (a) illustrates schematically the points where the EDS data were collected, and (b) through (f) are the spectra. Location (b) is an EDS spectrum from the GaAs substrate, similar to data in figure 4-24. At the interface between the Ni.3As and the GaAs, the Ga Kct signal decreased and the Ni Kot increased (figure 4-27 (c)). At point (d) and (e), inside the NiAs, the Ni signal was stronger, the Ga signal further decreased and a weak peak from Ge signal was detected as shown in figure 4-27 (d). The Ge signal was slightly stronger at position (e) versus position (d). Position (f) was at the same lateral location as

(e) but outside the NiAs phase and the spectrum is the same as that from Ni24GaAs, as shown by comparing figure 4-27 (f) with figure 4-23. This EDS results confirmed the formation of NiAs phase with minor concentration of Ge and Ga.








To better understand the EDS results, some characteristic features of the NiAs-type

unit cells were considered as shown in figure 4-28. The unit cell of NiAs is hexagonal, Ni atoms occupy (000, 00'2) sites, and As atoms occupy (1/3 2/3 1/4, 2/3 1/3 3/4) sites while two additional sites (2/3 1/3 1/4, 1/3 2/3 3/4) are vacant, as shown in figure 4-28 (a). When the two vacant sites are half occupied by Ni atoms, it becomes Ni3Ga2 (figure 4-28

(b)). If the two vacant sites are fully occupied by Ni atoms, it becomes Ni2In known as "filled-up" NiAs-type structure (figure 4-28 (c)) [Gue89]. Since the covalent radii of Ga atoms (1.25A) is bigger than that of As atom (1.2 A), the unit cell of NiAs structure expand when Ga atoms substitutes for As atoms, thereby increasing the concentration of Ni atoms in the unit cells. This is how the transition from NiAs to Ni3Ga2 occurs.

From these structural features, it was easily shown that the NiAs phase identified by

our EDS results was formed from Ni24GaAs by substituting the Ga atoms with As and Ge atoms, as will be obvious. I-V characteristics of this structure exhibited a noticeably reduced barrier, with breakdown voltage being reduced from -1.5V to 0.8V in figure 4-2.

Figure 4-29 (a) shows an as-deposited structure of GaAs / Ni24GaAs / 500A Ge. Below the Ge layer, -80A of a reacted layer (figure 4-29 (a)) is visible, suggesting Ge reacts with Ni2.4GaAs at temperatures as low as 1 00�C. Since interdiffusion of Ni and Ge over the carbon layer is clearly seen in the AES depth profile of the structure, figure (4-11

(a)), this reaction layer is believed to consist of Ni and Ge. Below this reaction layer, Ni24GaAs with a sharp interface with GaAs is seen.

Figure 4-29 (b) show the entire metal layer and GaAs substrate after vacuum annealing of the sample in figure 4-29 (a) at 500'C for 5 minutes. About -300A of Ni2.AGaAs was









decomposed. Coalescence of the microsize twinned sections was again observed and evolution of phases was similar to the GaAs / Ni24GaAs / 250A Ge sample. A noticeable difference between 500A Ge and 250A Ge was the extremely sharp interface between the contact metal layer and GaAs resulting from the disappearance of protrusion of NiAs phase towards GaAs. The grains of NiAs phase can be noticed by their dark/bright contrast. The I-V data from this structure was linear as shown in figure 4-3.



GaAs / Ni2.4GaAs / Ge / Ti. All the results from cross-sectional TEM analysis and elemental depth profile showed that both of GaAs / Ni2.4GaAs / 250A Ge / 300A Ti and GaAs / Ni2.4GaAs / 330A Ge / 200A Ti had the same metallurgical phases and linear I-V characteristics (figure 4-5).

Figure 4-30 shows a cross-sectional image of as-deposited GaAs / Ni2.4GaAs / 330A Ge / 200A Ti. In the bright-field image, the Ge layer was amorphous. In the corresponding diffraction pattern (not shown) a weak ring pattern with d-spacing of

-2.5A, in addition to spots from GaAs and Ni2AGaAs, was observed from the randomly oriented polycrystalline (100) Ti grains. The Ni2AGaAs layer was 1200A - 1300A thick, as usual.

The results of a high resolution EDS analysis on the Ni2.4GaAs phase are shown in figure 4-31 (a) while figure 4-31 (b) shows schematically which points were analyzed. It should be noted that OA on the x-axis in figure 4-31 (a) indicates the interface between a protrusion of Ni2.AGaAs into the GaAs, and the distances are relative to this point. In figure 4-30, several protrusion are visible and one of them is indicated by an arrow. The








69
distances are approximate values. EDS spectra were collected at the distances along this line shown in table 4-4. The concentrations in table 4-4 were determined from raw EDS spectra as described in Chapter 3.



Table 4-4. Concentration of elements plotted in figure 4-31 (a)
Distance(A) Ni (at %) Ga (at %) As (at %) Ni/Ga Ni/As
10 1.7 49.9 48.2 0.0 0.0 50 7.3 48.0 44.5 0.2 0.2
100 55.0 23.2 21.9 2.4 2.5 200 55.8 23.2 20.9 2.4 2.7 300 56.2 24.2 19.1 2.3 2.9 400 56.5 24.2 19.1 2.3 2.9


The first two data points at 10 and 50A showed dramatically different values from the expected stoichiometry of Ni24GaAs. Their values were almost the stoichiometry of GaAs. It is probable that those two points were from an area where a thin spike of Ni2.4GaAs protruded into the GaAs substrate. This kind of protrusion can also be seen in figure 1 of the reference by Sands et al [San87].

The EDS data collected outside of the protrusion revealed the composition of Ni24GaAs (see table 4-4). The Ni/As ratios were higher than those of Ni/Ga. The average concentrations of each element in as-deposited Ni2.4GaAs were determined as Ni2.75Gal.16As from table 4-4 and summarized in table 4-4, which has been and will continue to repeat as Ni24GaAs. This result agrees with previous EDS data showing the composition of Ni.GaAs was slightly Ga-rich [San87J.








Table 4-5. Composition of Ni2.4GaAs as-deposited INi Ga Ge As Atomic % 55.9 23.7 0 20.3


Figure 4-32 is a cross-sectional bright field image after 5000C for 2.5 minute. The two layers structure separated by the carbon marker reported in figure 4-26 to 4-29 is clearly evident. The top metal layer was -550A thick and the bottom layer in contact with the GaAs substrate was -110 A thick. As a result of Ge in-diffusion, the bottom layer contained Ge. Internal structure in the top metal layer was revealed by the bright field image in figure 4-32. Some precipitates have formed in the matrix which consists mostly of Ti and Ni, based on the corresponding depth profiles. To investigate the top metal layer, several diffraction patterns were taken along the <110> and <112> zone axis of GaAs from a sample annealed for 5 minutes at 5000C. Only GaAs and NiAs-structure type diffraction spots were detected along the <110> axis, but a strong extra spot was found along the <112> zone axis. The d-spacing of this spot was -2.31A and it corresponded to the (002) planes of NiTi. Thus, the matrix of the top metal layer consists of polycrystalline NiTi grains.

Inside the bottom metal layer near the interface GaAs, several bright grains were

distinguished from their dark surrounding as indicated by arrows in figure 4-32. These grains are thought to be NiAs phase containing Ga and Ge similar to those in figure 4-26

(a). This phase was also detected in samples annealed for an additional 2.5 minutes at 5000C.

Decomposition of Ni2.4GaAs and formation of NiAs phases were more clearly

demonstrated in the cross-section bright field images from GaAs / Ni24GaAs / 250A Ge /








71
300A Ti samples annealed for 5 minutes at 500'C (see figure 4-33). The top metal layer contained Ni, Ti and As, and was -600A thick, while the bottom (previously Ni2.4GaAs) layer was -900A thick. On average -300A of GaAs regrew as a result of the decomposition of the Ni2.4GaAs and NiAs phases. Because the interface with GaAs is now uniform (figure 4-33), versus undulating (figure 4-26), the NiAs phase was under decomposition along with Ni2.4GaAs. Undecomposed NiAs phase grains (bright contrast, indicated by arrows in figure 4-33) located near the interface with the GaAs can be distinguished from the surrounding grains by bright/dark contrast.

The composition of regrown GaAs resulting from the decomposition of NiAs phase and Ni2.4GaAs were investigated by using high resolution EDS. Figure 4-34 (a) and table 4-5 shows the atomic concentration versus distance from the interface between the NiAs phase and regrown GaAs. The interface is at "0" distance with the GaAs substrate being at negative distance. As indicated above, -300A of Ni2.4GaAs was decomposed resulting in the regrowth of GaAs, therefore the location at -300A would be the approximate location of the original interface between GaAs and Ni2.4GaAs before any decomposition.



Table 4-6. Atomic concentrations of elements in figure 4-34 (a)
Distance(A) Ni (at %) Ga (at %) Ge (at. %) As (at %)
-300 0.2 48.5 0.0 51.2 -200 0.2 52.3 0.1 47.1 -100 0.4 52.9 0.2 46.1
-50 3.0 51.5 0.0 45.0 -25 7.4 47.2 0.5 44.5
0 27.0 29.3 2.2 41.2
50 41.7 16.6 3.8 37.6 100 48.1 12.8 3.5 35.3 300 49.8 11.8 4.5 33.5











The exact composition of NiAs was determined to be Ni1.3Gao.4GeO.1As (table 4-7). Using the data acquired at IOOA into the Ni1.3As grain were used. The average vertical size of Ni1.3As grains was -400A or less. Two sets of data was collected at IOOA, then averaged to yield the following composition:



Table 4-7. Composition of Ni1.3As (Ni1.3AsGao.4Geo.i) INi Ga Ge As Atomic % 47.3 13.0 4.0 35.7


This composition confirmed that the result of EDS analysis shown in figure 4-27. It is noted that the As concentration actually increased from 20.25% to 35.63% from the data in table 4-4. This results shows that formation of the Ni1.3As phase involved redistribution of the major three elements, Ni, Ga and As.

Figure 4-34 (b) and table 4-6 shows the results across the interface between Nil 1Ga and regrown GaAs.



Table 4-8. Atomic concentrations of elements in figure 4-32 (b)
Distance(A) Ni (at %) Ga (at %) Ge (at. %) As (at %)
-300 0.2 52.8 0.0 46.6 -200 0.3 52.6 0.0 48.8 -100 2.6 52.0 0.2 44.9
-50 2.2 52.0 0.2 45.0 50 53.5 27.4 11.8 6.9 100 55.1 26.5 11.8 6.2 300 55.2 26.6 11.4 6.4







73
It can be noticed that the overall composition is very uniform over 300A of distance, from 50A to 300A. By averaging the data taken at 100 - 300A, the composition of Ni2.IGa formed from Ni24GaAs after annealing at 500'C, 5 minutes was determined to be Ni2.oGaGe0.4Aso.2 and listed in table 4-7.



Table 4-9. Composition of Ni2.,Ga (Ni2.iGaGeo4Aso.2) I Ni Ga Ge As Atomic % 55.2 26.6 11.62 6.3


By comparing the composition of Ni2.4GaAs as deposited (table 4-4) to that of the table 47, the As concentration significantly decreased from 20.3 % to 6.3 % while the Ga concentration slightly increased (from 23.7% to 26.6%). The Ni concentration was not changed and -11% of Ge was incorporated. The result of table 4-7 suggests that Ni2.4GaAs transformed into NiUAs phase containing Ga and Ge, and Ni2.1Ga phase containing Ge and As, as a result of the annealing at 500'C, for 5 minutes.

As shown in table 4-5, low concentration of Ge were detected in the regrown GaAs. Because the size of probe was - 15Ak and the measurements was taken only 25A away from the interface, there is a chance of secondary x-ray fluorescence from Ge in the NiL3As grain because of the proximity to the probe beam. Also, the Ge concentrations were near the detection limit for EDS analysis.

To verify the existence of Ge in the regrown GaAs, several EDS spectra were

measured from -300A to -50A in regrown GaAs where the beam broadening can be excluded. These spectra were summed to increase the ratio of Ge signal to the








background nose and the regrown GaAs is shown in table 4-8 to contain 0.34 % Ge, which is -1.5 x 102cm'3.



Table 4-10. Composition of Regrown GaAs calculated from summation spectra I Ni Ga Ge As Atomic % 2.6 46.2 0.34 50.7


The accuracy of this procedure was checked by calculating the Minimum Mass Fraction (MMF) for Ge, which indicates the statistical limit of detectability i.e. the statistical error range for Ge detection under these conditions.

MMF for Ge = 3 .(13) "2. ((Ge in wt. %) / (counts of Ge signal)}

where 13 is background noise in counts (count per seconds x total acquisition time). All the data from the summation spectra were plugged into the above equation and yielded

0.08 % as the MMF for Ge. Since the detected Ge concentration (0.34 at %) is four times larger than the error, the detected Ge is a true signal.

The Ge concentration within -50A of the interface was also studied with some

difficulty. Because the probing point was so close to the interface, it was likely that the probing beam illuminated the metal layer containing Ge. For these reasons, EDS analysis within -50A range were carried out with the EDAX 9100 system in the image mode with

1 or 2 x 106 magnification. The locations of the probed points were confirmed by observing diffraction patterns as described in Chapter 3. During the acquisition of EDS spectra, the acquisition was stopped and checked every 15 second during the total of 100 seconds to guarantee that the probing location was constant.







75
On average, 0.5 - 1.3 atomic % of Ge was detected, and this corresponds to a doping concentration of 2.2 - 5.8 x 1020 cm3. It is not clear to what extent the beam broadening effect contributed to these values. However, since 0.34 at. % (1.5 x 1020 cm"3) of Ge was measured in regrown GaAs at > 50A, it is reasonable to conclude that the maximum Ge concentration incorporated during the regrowth does not exceed -5.0 x 1020cm-3.

As an efforts to further investigate the Nil.3As phase, several diffraction patterns were taken from the samples annealed for 5 minutes at 5000C. Figure 4-35 (a) is a dark field image formed from the diffraction spot indicated by the arrow in figure 4-35 (b), which is the diffraction pattern from figure 4-35 (a). The d-spacing of the diffraction spot used to form the image of figure 4-35 (a) was 1.95 - 2.0A. Figure 4-35 (c) is the corresponding bright-field image of figure 4-35 (a) and the same grain is also indicated by an arrow in the figure.

Figure 4-35 (d) shows diffraction patterns simulated using the lattice parameters of NiAs and Ni3Ga2 along [110] zone axis of GaAs (GaAs: a. = 5.64, NiAs: a. = 3.61, c. =

5.03 A, and Ni3Ga2: a = 4.0, c. = 4.98A ) [San86, Che88]. In figure 4-35 (d), three different phases contributed to the entire patterns: GaAs (0), NiAs (x), and Ni3Ga2 (E-1). Figure 4-35 (e) is a diffraction pattern taken along along [110] zone axis of GaAs, where Ni1.3As (a = 3.78, c. = 5.053A) and Ni2.1Ga (a. = 3.95, c. = 4.98A) were identified. By comparing (d) and (e), it was found that the diffraction spot indicated by an arrow in figure 4-35 (b) and (e) was the same spot indicated in the simulated pattern (d). This diffraction spot corresponded to (0112) plane of the NiAs phase with d-spacing of 1.98A. This analysis provided confirming data that Ni2.4GaAs tended to transform into NiAs and







76
Ni3Ga2 upon annealing at 500'C for 5 minutes, and the Nil.3As detected in this study was indeed the NiAs phase containing Ga and Ge.

Figure 4-36 is a high resolution image taken from a sample after the vacuum annealing at 500'C for 5 minutes in an area such as shown figure 4-35. The dashed line in the figure indicates the original GaAs/Ni2.4GaAs interface before regrowth. In the regrown GaAs, stacking faults and precipitates (see arrow) were observed. Along the interface, Moire fringe and contrast are evident, which are believed to be the trace of the decomposed Nil3As phase. The dark/bright contrast in the regrown GaAs is believed to be strain contrast caused by accommodation the defects. One of the precipitates was identified and analyzed by high resolution EDS. The composition of the precipitate was Ni-9 at. %, Ga47 % and As- 44 %. The regrown GaAs also contained 2.6 at. % Ni (table 4-8). In figure 4-35 (a), not only the Nil.3As grains but also precipitates in the regrown GaAs were bright when the diffraction spot was used to form the dark field image. They were typically smaller than 1 OOA. Therefore, some of the precipitate were Ni1.3As binaries. In the metal layer, bright grains (marked A in figure 4-36) were Ni*3As and the dark contrast area was Ni2.1Ga(Ge,As) region.



GaAs / Ni / Ge / Ti. Figure 4-37 shows a bright field image of an as-deposited GaAs / 650A Ni / 250A Ge / 300A Ti. Note that this is a structure without in situ annealing. As shown, the Ni film has columnar structure and diffraction analysis revealed that the Ni film was polycrystalline (111) oriented grains. The Ge layer was amorphous and the Ti layer developed small polycrystalline (100) oriented grains. Between the Ni and Ge layers, an







77
extra layer with a thickness of 1 50A was detected. In this bright-field image, interfaces between metals were very smooth and the corresponding depth profile, figure 4-15 (a) showed interdiffusion between the Ge and Ni layers. Thus, this extra layer appears to be a reaction product of Ni and Ge formed during deposition or the TEM sample preparation at 110�C.

Figure 4-37 (b) shows the annealed structure of figure 4-37 (a) and the entire contact metal consisted of two layers. The top metal layer was about 650A thick and the bottom layer was 750 - 850A thick. Between the two layers, a sparse discontinuous bright line was still visible. In the depth profile of this structure (figure 4-15 (b)) a strong peak of carbon was detected between two layers of constant composition.

From the corresponding SIMS depth profile of chemical composition in figure 4-15 (b) and the diffraction pattern from the entire structure of figure 4-37 (b), it is believed that the bottom layer is similar to the bottom metal layer of the annealed GaAs / Ni24GaAs / 250A Ge / 300. Ti shown in figure 4-35, i.e. a mixture of ternary NiGaGe(As) and NiAs phases. In the bright field image (figure 4-37 (b)), bright grains near the interface with GaAs are distinguishable from the dark surrounding. In our experiment, NiAs grains could be quickly identified by EDS spectra by an As signal is much stronger than that from Ga, such as the spectra (e) of figure 4-27. Our EDS investigation on the structure shown in figure 4-37 (b) revealed that these bright grains were the NiAs phase (see arrow in figure 4-37). However, EDS spectra from the grains in figure 4-37 (b) were not so conclusive as those in figure 4-27 (e), i.e. the As signal was only slightly stronger than those from Ga, suggesting that the grains were in an early stage of the NiAs phase. This







78
was confirmed from our diffraction pattern analysis which did not record diffraction spots of NiAs phase. Also their density was lower and their size was smaller. One crystal is indicated by an arrow.



GaAs doped 1016 Cm3. It was reported above that GaAs (1016) / Ni24GaAs / 750A Ge / 300A Ti did not form linear I-V after vacuum annealing at 5000C, for 5 minutes, while GaAs (1016) / NixGaAs+NiAs / 750k Ge / 300. Ti formed linear I-V after vacuum annealing at 5000C, for 5 minutes. Recall that NixGaAs + NiAs represents the reaction products formed by in situ annealing of 650A Ni film at 400'C, for 15 minutes. Even though the two structures exhibited dramatically different I-V characteristics, the depth profiles did not provide good clues for the difference in the I-V behavior. Thus, these two structures were analyzed by TEM.

Figure 4-38 shows a bright field image of GaAs (1016) / Ni2.4GaAs / 750k Ge / 300k Ti after annealing at 5000C, 5 minutes, and three layers can be distinguished. The first layer was -600A thick (marked A in figure 4-38), the second layer was uniform and

-700A thick (marked B), while the third 400k thick layer was not a well defined layer but appears to be a region where various sizes of fragments of metal phase were scattered in a matrix of GaAs. Twins developed along the I 111 } planes of GaAs were visible along with the metal fragments. This sample exhibited strongly rectifying I-V characteristics.

The results of electron diffraction analysis are tabulated in table 4-11. Several

diffraction patterns were taken along <110> and <111> zone axis of GaAs. The major reaction products were Ni-rich Ni-Ge compounds (A region in figure 4-38) and Ni-Ga-Ge








ternary phase (B region in figure 4-38). The corresponding depth profile from this

structure (figure 4-20(b)) supports this conclusion.

The reaction product formed by the in situ annealing of 650A Ni films at 4000C, 15
minutes were also analyzed by diffraction along the <110> zone axis of GaAs.


Table 4-11. Summary of diffraction patterns recorded and analyzed to identify
reaction products shown in figure 4-38.
GaAs (1016) / Ni2.4GaAs / 750A Ge / 300A Ti after annealing at 5000C, 5 minutes.
Diffraction type Measured d-spacing ( A) Possible phase, d-spacing (A),
(Intensity) intensity*, index (hkl)
Spot (strong) 2.85-2.89 Ni5Ge3 2.863 80 (221) Ni4GaGe2 2.807 80 (203) Ni2GaGe 2.807 50 (130) Spot (strong) 2.390 Ni5Ge3 2.307 20 (222) Ni2Ge 2.345 60 (112) Ni2GaGe 2.339 40 (043) Spot (strong) 1.977 NiGe 1.997 100 (121) NiGe 2.048 80 (211) NisGe3 1.951 95 (602) Ni5Ge3 2.014 100 (313) NisGe3 2.022 95 (203) Ni5Ge3 2.102 60 (403) Ni5Ge2 2.009 100 (115) Ni2Ge 2.042 100 (211) Ni2Ge 1.919 85 (020) Ni4GaGe2 1.957 100 (220) Ni4GaGe2 2.012 100 (206) Ni2GaGe 2.016 60 (060) Spot (strong) 1.699 NiGe 1.687 70 (301) Ni2Ge 1.661 20 (301) Spot (strong) 2.42 Ni4GaGe2 2.424 70 (212) with ring (weak) Ni4GaGe2 2.498 70 (006) Ring (weak) 1.43 Ni2Ge 1.442 20 (123)
"The intensity value is relative the strongest diffraction peak designated as 100 in
the JCPDS file.







80
Diffraction spots from NiAs phase were detected and the measurements yielded a. -3.69A and c. -5.06A. A weak ring pattern was overlapped on (0001) type spot from NiAs phase, suggesting that the NiAs phase consisted of several grains. The corresponding SIMS depth profile (figure 4-21 (a)) showed that the intensity of Ga signal was slightly lower near the GaAs substrate, supporting the formation of NiAs phase near the GaAs substrate. A different set of diffraction spots indicated formation of another NiAs-type phase with a,, -4.01A and Co -4.99k, which is believed to be Ni.GaAs from the SIMS depth profile in figure 4-21 (a).

Considering these results, the reaction products formed by the in situ annealing of 650A of Ni films at 4000C, 15 minutes is a mixture of NixGaAs and NiAs compounds. Our results agrees with the report by Lahav et al. where a Ni2-,GaAs.x matrix with NiAs precipitates were observed as a result of annealing at 350 - 550'C [Lav86].

Figure 4-39 is a bright-field cross section image of the structure of GaAs (1016)!

NixGaAs + NiAs / 750A Ge / 300k Ti after annealing at 5000C for 5 minutes. Again, the entire metal layer consisted of three layers. The top 400k layer and the second 400k. layer showed rough interface between them, while the 500 - 700k third metal layer had a rough interface with GaAs. The GaAs near the interface with the third metal layer contained twins which extended -500 into GaAs, and associated strain contrast was clearly visible. The important difference with the image of figure 4-39 versus 4-38 is that there are not as many as precipitates in the GaAs near the interface. This structure exhibited ohmic behavior and the differences in the microstructure is thought to be the reason.







81
Analysis on diffraction patterns taken along < 110> zone axis from this structure

revealed that NiGe and Ni2Ge were reaction products possibly with Ni-Ga-Ge ternaries. Therefore, 1400 -1500A thick NixGaAs + NiAs layer was converted into Ni-Ge compounds by reaction with 750A Ge and 300A Ti layers.






















0.02


0.01


0.00


-0.01


-n n -


-s, ~


o As-deposited
* 3000C, 15 minutes


.. .. . .. . . .. . ...... . . .
90
0. 0

.00
........... .. .. .. .. .
04
0*
.. . . . . . .. . .. . . . . . . . . .
Ca


I I I I 1 I
-4.0 -2.0 0.0 2.0


4.0


Voltage (V)







Figure 4-1. Current-voltage characteristics for GaAs/650A, Ni as deposited and in situ annealed at 3000C for 15 minutes in vacuum.





















0.04


0.02


0.00


-0.02


-0.04


o As-deposited
* 5000C, 5 minutes
0
* 0
*0
: ~ :0 �






. . . . . . O !......... .......... ......... .. .......... . .
CSo
0 �
0
0


- I I - .I 0
-4.0 -2.0 0.0 2.0 4.0


Voltage (V)






Figure 4-2. Current-voltage characteristics for GaAs/Ni24GaAs/250A Ge both as deposited and ex situ vacuum annealed at 500'C for 5 minutes.


w



















o As-deposited
* 5000C, 5 minutes
0
0
0
.. . . . . . .................. -'.......... O . ..........
0 0:
o �0

0 :0
0 *
o
o
0


.0 -2.0


I .0I I
0.0 2.0 4.0


Voltage (V)






Figure 4-3. Current-voltage characteristics for GaAs/Ni2.4GaAs/500A. Ge both as deposited and ex situ vacuum annealed at 5001C for 5 minutes.


0.02 0.01


0.00


-0.01


-0.02-4





















0.02 0.01 0.00


-0.01
4


-0.02


o As-deposited
* 5000C, 5 minutes
0
* 0 0*


00
... ... .. ... ... .. ... ... .. ............. .... .....


0o
* 0
o


-2.0


I
-1.0


0.0


1.0


2.0


Voltage (V)






Figure 4-4. Current-voltage characteristics for GaAs/Ni2.4GaAs/500A Ti both as deposited and ex situ vacuum annealed at 500'C for 5 minutes.




















0.02 0.01


0.00-


-0.01


-0.02


-2.0


o As-deposited
* 5000C, 5 minutes
* 5000C, 20 minutes

G
0
.0
00 0'
00


0 0

*0
* ....


I
-1.0


0.0


1.0


2.0


Voltage (V)






Figure 4-5. Current-voltage characteristics for GaAs/Ni2.4GaAs/250A Ge/300A Ti both as deposited and ex situ vacuum annealed at 5000C for 5 minutes.


















o As-deposited
* 5001C, 5 minutes



* 04 000

o.......... ...............

so


I I


' I


-1.0 0.0


1.0 2.0


Voltage (V)






Figure 4-6. Current-voltage characteristics for GaAs/650A Ni/250A Ge/300A Ti both as deposited and ex situ vacuum annealed at 5000C for 5 minutes.


0.02 0.01


0.00


-0.01


-0.02


-2.0

















o As-deposited
* 5000C, 5 minutes
o 5000C, 20 minutes E8 500OC, 35 minutes





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


I - 1
-2.0 -1.0


I I
0.0 1.0


Voltage (V)






Figure 4-7. Current-voltage characteristics for GaAs/650A Ni/250A Ge/600A Ti both as deposited and ex situ vacuum annealed at 5000C for 5 minutes.


0.02 0.01


0.00


-0.01


-0.02


2.0





















0.02 0.01 0.00


-0.01


-0.02


o As-deposited
* 5000C, 5 minutes @5000C, 20 minutes








o?.................... .........
0�6





Be


I I I I 1 I
-2.0 -1.0 0.0 1.0


2.0


Voltage (V)






Figure 4-8. Current-voltage characteristics for GaAs/650A Ni/250A Ge/900A Ti both as deposited and ex situ vacuum annealed at 500'C for 5 minutes.















* As-deposited
GaAs (1016) / Ni24GaAs / 250A Ge / 300ATi
El GaAs (1016) / Ni.4GaAs / 750A Ge / 300ATi 0 GaAs (10 16) / Ni2 4GaAs / 250A Ge / 900AT1
0.0002 ,1


0.0001 0.0000


-0.0001


-0.0002


-4.0 -2.0 0.0 2.0 4.0 Voltage (V)



Figure 4-9. Current-voltage characteristics after ex situ vacuum annealing of various thicknesses of Ge and Ti at 5000C for 5 minutes. The GaAs substrate doped to 2 x 1016cm-3 and a 650X Ni layer was in situ annealed at 300'C for 15 minutes to produce Ni24GaAs layer.














GaAs (10') / NixGaAs + NiAs / 250AGe / 300A Ti GaAs (1016) / NixGaAs + NiAs / 500AGe / 300A Ti GaAs (1016) / NixGaAs + NiAs / 750AGe / 300A Ti GaAs (1016) / NixGaAs + NiAs / 25 AGe / 900A Ti
0.04- i i
I I
0.02 - -----E 0.00 ......

-0.02- - - - , , - -v - v - - --I I ~ I



-0.02 -- -- 1-I



-4.0 -2.0 0.0 2.0 4.0
Voltage (V)


Figure 4-10. Current-voltage characteristics after ex situ vacuum annealing of various thicknesses of Ge and Ti at 500'C for 5 minutes. The GaAs substrate doped to 2 x 1016cm-3 and a 650A Ni layer was in situ annealed at 400'C for 15 minutes to produce NixGaAs plus NiAs layer.










1 e+5 1 e+4

1 e+3 le+2 1e+1 1 e+O




I e+5

,..1 e + 4


1 e+3
C
le+2 le+1 -


0 5


10 15 20


Sputter time (min)


Figure 4-11. SIMS depth profile for GaAs/Ni24GaAs/250A Ge. (a) as deposited
(b) ex situ vacuum annealed at 500'C for 5 minutes.


0 5 10 15 20
sputter time (min)









(a) As-deposited


40000


30000


20000 10000-


40000


0 10 20 30 40
Sputter time (min) (b) 500 C, 5 minutes


30000 ... . ri.


20000

Ga
10000 Ge Ga

0 -"c Is F- ' -"H .....

0 10 20 30
sputter time (min)


Figure 4-12. AES depth profile for GaAs/Ni2.4GaAs/500A Ge. (a) as deposited
(b) ex situ vacuum annealed at 500'C for 5 minutes.


Ge:


IGa




Full Text

PAGE 1

EVOLUTION OF INTERFACIAL PHASES AND THEIR EFFECTS ON OHMIC CONTACTS TO n-GaAs IN Ni-Ge-Ti METALLIZATIONS By TAE-JIN KIM A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 1996

PAGE 2

Copyright 1996 by Tae-Jin Kim

PAGE 3

ACKNOWLEDGEMENTS With all my sincere respect and love, I must thank my parents for their love, sacrifice and support. Without what they gave me, I could not have come this far. Now, I dedicate this dissertation to them. I would like to express my gratitude for my wife, Jin Hee, who has patiently endured this challenging time in her life without losing her bright smile. I must also thank Dr. Paul Holloway for his guidance. I have had a wonderful opportunity to learn how to approach academic issues. But more importantly, for the first time I could discover the example of the professor and scientist I want to be. Many people helped me with the work presented in this dissertation. Maggie did all of the SIMS analysis and Eric did the AES analysis. Wish helped me collect the TEM data and Sohn helped me analyze the TEM diffraction pattern. I appreciate the opportunity to work with Dr. Kenik at the Oak Ridge National Laboratory. I have enjoyed my companionship with our group member, Troy, Sean, Joe, Jonathan, Philip, Jeff; all of them helped me whenever I asked their help and I thank them. Also there are several other unnamed students and people who deserve my gratitude in our department. Even though I started the work in this dissertation myself, it would have been impossible to finish without all the people who helped me.

PAGE 4

TABLE OF CONTENTS ACKNOWLEDGEMENTS iii ABSTRACT vi CHAPTERS 1 . INTRODUCTION 1 2. REVIEW OF LITERATURE 6 Electrical Properties of GaAs Surface and Interfaces 6 Ex situ Contact Schemes 19 Au/GaAs Metallizations 19 Ni/GaAs Metallizations 22 AuGe/GaAs Metallizations 25 NiGe/GaAs and Other Bielements/GaAs Metallizations 27 AuNiGe/GaAs Metallizations 3 1 Hybrid Contact Metallizations 35 In situ Contact Schemes 36 Current Status of Contact Metallizations 38 3 . EXPERIMENTAL PROCEDURE 42 GaAs Substrate Preparation 42 Metal Evaporation and In Situ Annealing 43 Post-Evaporation Annealing 46 Electrical Measurements 46 Analytical Characterizations 47 Auger Electron Spectroscopy (AES) 48 Secondary Ion Mass Spectroscopy (SIMS) 48 Transmission Electron Microscopy (TEM) 49 4. RESULTS 53 Summary of Prepared Samples 53 Electrical Measurements 54 Elemental Depth Profiles 58 Transmission Electron Microscopy 63 IV

PAGE 5

5. DISCUSSION 126 Evolution of Ni 2 . 4 GaAs 126 Mechanism for the Formation of Ohmic Contacts 136 6. CONCLUSIONS 149 APPENDICES A HEAT OF FORMATION 1 52 B CALCULATION OF PHASE DIAGRAM 1 5 5 REFERENCES 158 BIOGRAPHICAL SKETCH 168 V

PAGE 6

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 EVOLUTION OF EMTERFACIAL PHAES AND THEIR EFFECTS ON OHMIC CONTACTS TO n-GaAs IN Ni-Ge-Ti METALLIZATIONS By Tae-Jin Kim December 1996 Chairman: Paul H. Holloway Major Department: Materials Science and Engineering Electrical and metallurgical properties of Ti/Ge/Ni metallizations used to form ohmic contacts to n-type (100) GaAs have been studied with current-voltage (I-V) measurements, secondary ion mass spectroscopy, Auger electron spectroscopy, and transmission electron microscopy with energy dispersive analysis of x-rays. The purpose of this study was to investigate the evolution of Ni-Ga-As phases and their effects on the formation of ohmic contacts. Thin films of Ti, Ge and Ni deposited by electron beam to thickness of 30 60nm, 25 -75 nm and 65nm, were heat treated in vacuum at 500°C either as all three deposited layer at the same time or the Ni film was annealed in situ to react with GaAs, followed by in situ deposition of Ge and Ti and ex situ annealing in vacuum at 500°C. For the case of Ni deposition followed by in situ annealing at 300“C, a 120 130 nm VI

PAGE 7

thick hexagonal ternary Ni2.4GaAs phase was produced. After deposition of Ge / Ti layers on this Ni2.4GaAs and vacuum annealing at 500 °C for 5 minutes, the electrical behavior switched from rectifying to ohmic. Simultaneously with development of ohmic behavior, the Ni2.4GaAs phase was transformed into Nii.3As and Ni^iGa phases, which were decomposed by the formation of Ni-Ge compounds and NiTi. The decomposition of the Nii.sAs resulted in solid phase epitaxial regrowth of GaAs doped with Ge at Ga sites and the Ge concentrations were measured up to midlO^^cm'^. This entire evolution was towards the equilibrium phases in the Ni-Ga-Ge-As-Ti system. The thickness of the Ti and Ge layers, and microstructure of the regrown GaAs were critical factors controlling development of ohmic contacts. Regrown GaAs was characterized by various defects and planar interfacial morphology for the ~ 30 nm regrowth range. A model describing the level of Ge incorporated on Ga sites to act as donor was developed based on control of the Fermi level of the GaAs at the regrowth interface with the Ni-based phases. This model quantitatively predicts that the conduction band will shift below the Fermi level and lead to tunneling transport for ohmic contacts. vii

PAGE 8

CHAPTER 1 INTRODUCTION GaAs is the best known and most widely used III-V semiconductor. The most advantageous properties of GaAs over Si are higher electron mobility, lower voltage operation, semi-insulating substrates, monolithic integration of optical and electronic functions and radiation hardness [Scl88], Utilizing the higher electron mobility of GaAs, Integrated Circuits (ICs) such as Monolithic Microwave Integrated Circuits (MMIC) are fabricated for high-frequency devices. The direct band gap of GaAs makes it possible for use in infrared-emitting diodes and solid-state lasers. Also GaAs has played a key role in epitaxial growth of ternary or quaternary compounds for diverse optical applications [Hai89]. Regardless of application, operation of all GaAs devices requires two types of electrical contacts, rectifying (or Schottky) and ohmic contacts. For instance, ohmic contact is required for the drain and source and rectifying contact for the gate in a Metal Semiconductor Field Effect Transistor (MESFET) structure. To guarantee successful device operation and performance, the contact should have a low resistance, reliability with time and reproducibility as well as adaptability to the fabrication 1

PAGE 9

2 process. In current GaAs fabrication technology, ohmic contacts are formed mostly by a metallization step using a lift-off process with the Au-Ge-Ni system developed in 1967 for n-type GaAs [Bra67], The issues associated with contact formation become of critical importance in order to fully utilize the best performance of the devices, especially the high frequency and high power characteristics [Chr95], In addition to the device performance, reliability is also strongly affected by the electrical contacts as the level of circuit integration approaches Ultra Large Scale ICs (ULSI) and oscillation frequency of GaAs ICs approach several hundreds of GHz [And90], In a typical MESFET structure, the channel length between source and drain is ~1 pm. For a higher speed operation and higher integration, the channel length is being reduced; consequently, the spatial tolerance for contact metallization becomes more stringent. In a high power switch where current density approaches lO" A/cm^, a high resistance ohmic contact ( > 10'^ Qcm^) can cause excessive Joule heating leading to failure of the devices [Bar92]. It has been reported that current Au-Ge-Ni metallization is unlikely to be used for ftjture ULSI and other applications because of its severe interdiffusion and metal/semiconductor reactions often resulting in degradation of the performance and failure of the devices [Woo86, And90]. Problems associated with Au-Ge-Ni metallization such as irregular morphology [Gup90, Bal89], thermal stability [Wil88, Hug92] and interdiffusion [Kol88] have been addressed by several authors. Tremendous efforts have been made to solve the problems of the Au-Ge-Ni system

PAGE 10

3 [Rid75, Pio83, She92] and they have resulted in improvements non-alloyed and Aufree contact schemes [Meh89, Mur89, Tan92], Generally, the electrical properties of metal contacts to semiconductors depend on composition, structure and morphology of the interface, which are determined by interfacial reactions. The interfacial reactions are unavoidable because semiconductors and metals have enormously different properties, for example, lack of lattice match and different electronic characteristics. Consequently, a reaction takes place at the interface. The reaction are very complicated to characterize because several elements are involved with many processing variables. This might explain why ohmic contact technology had begun as an art rather than as a science, and why the complete mechanism of forming ohmic contacts in Au-Ge-Ni system is not yet understood even though Au-Ge-Ni metallizations provide practically acceptable low contact resistances [Rid75, W0086]. While Au-Ge-Ni metallization has been the subject of extensive investigations throughout the 1970s and 1980s, simpler metallizations such as single element/GaAs and bi-elements/GaAs were studied in the 1980s as an effort to better understand the complicated Au-Ge-Ni system. As a consequence, the model of solid-phase regrowth has been developed [San88, Hol91, Li92]. In the solid-phase regrowth, a metal element, such as Ni or Pd, reacts with GaAs, producing an interfacial phase such as NixGaAs or PdyGaAs. When these interfacial phases are decomposed, the electrical contact property switches from rectifying to ohmic. As for the current transport mechanism in conjunction with the formation of ohmic contacts, the solid-phase

PAGE 11

regrowth assumes the creation of degenerately doped n"^-GaAs leading to tunneling across the metal/GaAs contact. Even though some guidelines for the solid-phase regroAvth has been established, several key issues in the mechanism still need to be better studied and clarified. One such issue is the evolution of interfacial phases, which are believed to control or dramatically affect the contact property. The purpose of this study was to investigate the evolution of interfacial phases consisting of Ni, Ga and As and their effects on the formation of ohmic contacts. Ni has been used a standard element in ohmic contact metallizations for GaAs devices, and it forms various compounds with GaAs. Particularly, special attention was given to a ternary hexagonal phase, NixGaAs, since previous studies showed that the contact properties were dependent on how this ternary phase evolved [San88]. Ge was selected as an n-type dopant. Ti was selected to cause the continued evolution of the interfacial phases by reacting with Ni causing it to be removed from the Ni-Ga-As phases. Ni and Ti have a strong affinity and therefore form Ni-Ti compounds. Ti has been used to play the role of diffiasion barrier as an element or as a constituent in the compounds conventionally. In this study, however, Ti was used to react with Ni rather than to remain inert. This type of use of Ti in forming ohmic contacts has not been studied before. In order to focus on the evolution of the interfacial phases, in situ annealing was adopted, which will be described in Chapter 3. For the dissertation organizations. Chapter 2 begins with a review of the literature regarding the unique surface/interface properties of GaAs. Then, various metallizations with different complexity are reviewed. During this review, the most

PAGE 12

important models of the mechanisms for the formation of ohmic contacts to GaAs are discussed. In Chapter 3, the experimental approach and procedures are described. These include sample treatment, evaporation, in situ annealing and post-evaporation annealing along with the analytical techniques such as transmission electron microscopy (TEM), secondary ion mass spectroscopy (SIMS) and Auger electron spectroscopy (AES) to characterize the samples. In Chapter 4, results of the metallurgical study and electrical measurements are described and correlated to each other. In Chapter 5, all the results are reviewed and discussed and a model for the formation of ohmic contact is proposed. Finally, conclusions from this study are summarized in Chapter 6.

PAGE 13

CHAPTER 2 REVIEW OF THE LITERATURE TodayÂ’s semiconductor devices have been developed based on the understanding of electrical properties of semiconductor surfaces and interfaces. There is a need for such a understanding since semiconductor interfaces control the transport of current across the electrical junctions. The understanding was made possible largely due to the comprehensive surface science work on III-V surface and interfaces. This literature review begins with a description of one important electrical property of the GaAs surface, Fermi-level pinning, which has been investigated for several decades. Knowledge of Fermi-level pinning phenomena in GaAs is a central issue because the mechanism of Schottky barrier formation is strongly affected by the Fermi level pinning. Electrical Properties of GaAs Surfaces and Interfaces The rectifying properties of Schottky contacts were first described by using a simple approach [Mot38, Sch40]. Ideal Schottky contacts form as a result of charge transfer between metal and semiconductor to align the Fermi level across the interface. Figure 2-1 illustrates this process. 6

PAGE 14

7 Figure 21(a) shows the energy band diagram of a metal and a n-semiconductor before contact (unless otherwise mentioned, the following discussion always refers to n-type semiconductors). As shown, the work function of the metal
PAGE 15

The resulting band bending due to the space charge region is described by the spatial variation of electrostatic potential, V(x), V(x) =(l/q)-(Evb -Ev(x)) = (l/q).(Ecb-Ec(x)) (2-2) where subscripts b refer to bulk (i.e , x=oo), and V and C refer to the valence and conduction bands. In thermal equilibrium, band bending at the surface (x=0) is uniquely determined by: qVs(x=0) = Evb -Evs =Ecb Ecs (2-3) where the subscript s refers to the surface (i.e., x=0). With the assumptions of semiinfinite boundary conditions and no interface states, the potential as a function of x can be obtained by solving PoissonÂ’s equation: d^V(x)/dx^ = -p(x)/ss (2-4) where 8s is the dielectric constant of the semiconductor and p(x) is the spatial distribution of charge. Provided that there are no free carriers in the depletion region of n-type GaAs and complete ionization, i.e., p(x) = No, solving PoissonÂ’s equation yields V(x) = Vs+E4x-x^/2W) (2-5) where Em is the electric field at x=0 given by (qNoW)/ss, W is the depletion length, (28sVbi/qND)*^, and Vbi is the built-in potential defined by Ecb Ecs. Here, Vs is the surface band bending at x=0. Equation (2-5) shows that the band edge is a parabolic function of x from x =0 to x =W (Fig. 2-1). Along with Ob, another important figure of merit in the formation of an ohmic contact is the specific contact resistance given by: = {(<^/JV)v=o}'f^cm^ (2-6)

PAGE 16

where J is current density and V is the applied bias across the contact. In order to obtain low resistance ohmic contacts, the barrier height Ob should be small for thermionic emission of current over the barrier, or the depletion region, W in Fig. 2-1 (b), should be narrowed for tunneling of carriers through the barrier. Practically, ohmic contacts to GaAs devices must be tunneling contacts because the Fermi level is pinned and therefore Ob can not easily be reduced. Calculating the contact resistance, pc, requires details of the current density, J, which can be estimated from J = n« q • V P(Ob) (2-7) where n is the concentration of free electrons, q is the charge of an electron, P(Ob) is the tunneling probability, and v is the thermal velocity of electrons. While P(Ob) is a complicated function of the detailed band structure of the semiconductor, an approximate form of the solution for equation (2-7) was used to define three mechanisms of current transport across the metal/semiconductor contacts [Bar92]. For thermionic emission over the Schottky barrier, the solution for pc is given by Pc = (^/qA*T) • exp(Ob/A:T) (2-8) where A is the Richardson constant, k is the Boltzmann constant and T is the absolute temperature. As can be seen in equation (2-8), the contact resistance is independent of the doping concentration and hence the depletion width. In the case of tunneling, the contact resistance is approximated by Pc oc exp((Db/Eoo) (2-9)

PAGE 17

10 where Eoo (h/47i) • (ND/m*8s)^^^. The contact resistance is lowered for ohmic contacts when Nd is mad large. In the third category, a current density is obtained from a mixed form of tunneling and thermionic emission, and the contact resistance is given by Pc oc exp(Ob/Eoo) / coth(Eoo/AT) (2-10) where the temperature dependence of thermionic emission is contained in the coth(Eoo//tT) term. This mode of transport is called Thermionic (assisted) Field Emission (TFE) in which the carriers are thermally activated above Fermi level to a region where they can tunnel efficiently through a narrower portion of the barrier. Because coth(z) approaches unity for large z, equation (2-10) has been used empirically to predict ohmic or non-ohmic behavior of the interface for a given doping concentration: Eoo/^T » 1 => coth(Eoo/A^T) = 1 Ohmic by tunneling Eoo/A:T =1 => coth(Eoo/A:T) > 1 Non-linear Eoo/^T « 1 => coth(Eoo/A:T) »1 Non-ohmic. For z =2.7, the error in setting coth(z) =1 is less than 1%. Therefore, it is possible to predict the doping concentration required to form a tunneling contact by setting EJkJ » 2.7. Then Eoo= 18.5 x 10‘‘^ •(ND/m^r)‘^^> 2.1 •kl eV (2-11) where mr =m /nie, 8r = zjzo and Nd is the carrier concentration (cm’^). For n-type GaAs at 300K, equation (2-1 1) predicts Nd > 1.2 x lO'^cm'^. This approximation suggests the effective doping concentration should be greater than 10‘^cm'^ at room temperature to form tunneling a ohmic contact to n-GaAs.

PAGE 18

11 The Shottky barrier height is obviously an important parameter in these formulations, and it plays a key role in the determination of current transport mechanisms. According to the ideal Schottky barrier in equation (2-1), the equilibrium barrier height b depends on the work function of the metal. In reality, however, experimentally measured barrier heights showed a very weak dependence on the metal work function, especially for group IV elements and weakly ionic III-V compound semiconductors such as GaAs [Bar47, Cow65, Spi79]. This deviation from ideal Shottky behavior was first attributed to the existence of intrinsic or extrinsic surface states by J. Bardeen in 1947 [Bar47]. When surface states are present within the bandgap of n-type semiconductors, they are occupied by electrons. Occupation is controlled by the Fermi level which is constant throughout the crystal. Consequently, these surface states may pin the Fermi level resulting in band bending as shown in Fig. 2-2. Now the barrier height Ob is not given by Equation 2-1 but must be modified by the details of Ess, the energy levels of the surface states. For instance, in Fig. 2-2, the barrier height can be described as Ob = Eg Ess [Bar47]. As shown, the barrier height is nearly independent of the metal work function and is called Fermi level pinning. Fermi level pinning has been observed not only in GaAs, but also in Si and Ge [Mye47]. Kurtin et al. proposed intrinsic mechanisms as a plausible origin of the surface states, and suggested a correlation between a tendency to form surface states and types of chemical bonding in the bulk. Based on barrier heights of the various metals on Si, GaSe and Si 02 related to the electronegativities of the metals, it was suggested that the termination of periodic atomic arrangement at the surface of covalently bonded

PAGE 19

12 semiconductors gave rise to surface states [Kur69], However, it has been shown experimentally [Gud76, Spi76, Tan84] and theoretically [Sch90] that intrinsic states in the bandgap for the (110) surface of GaAs do not play a role in Fermi-level pinning. A large body of experimental data were analyzed and resulted in the Unified Defect Model. Spicer et al. discovered that clean, cleaved (110) surfaces of IIIV compounds with an exception of GaP did not exhibit intrinsic states, but the surfaces were strongly perturbed by coverages of monolayer or less resulting in surface states and Fermi-level pinning [Spi79, SpiSO], They proposed that the pinning resulted from defects created by the interactions between the adatoms (metals and oxygen) and semiconductors. Later, the Asca antisite defect was suggested to be controlling the pinning level, either at 0.5 eV or 0.7eV above the valence band maximum (VBM) [Spi88]. According to this model, the Fermi level varied from 0.5eV to 0.7eV above the VBM depending on the concentration of the antisite defects. The heat of condensation was proposed to result in atom displacement in one version of this mechanism but limited data do not support this mechanism. One difficulty in understanding of the UDM is that the defect states can not be directly detected nor quantified. The validity of two pinmng level in the UDM model was argued based on theoretical calculation [Zur83]. Later it was found that conditions of metallization determined one of two pinning values [Cao89]. Unique to the UDM is that Fermi level pinning is a consequence of extrinsic effects, not intrinsic properties of the semiconductors. The Metal Induced Gap (MIG) model has been developed as another explanation for the Fermi level pinning [Hei65]. In the MIG model, surface states are an intrinsic property

PAGE 20

13 of the semiconductor, and continuum levels are predicted in the energy range between the VBM of the semiconductor and the Fermi level, which is a central contradiction to the descrete levels described by UDM. The MIG are formed in the semiconductor at the initial interface due to intimate contact with the electrons of the metals. By such an intimate contact, the tails of metallic wave functions decaying into the semiconductor side provide the interface states within the band gap of the semiconductor. Heine first pointed out that, at the metal/semiconductor interface, the tails of the metal wavefunctions into the semiconductor are derived from the virtual gap states of the complex band structure of the semiconductor [Hei65]. When specific boundary conditions are satisfied, the “virtual” gap states which are solutions of Schrodinger’s equation for semiconductor band structure with complex wave vectors become “real” surface states with a decay distance into the semiconductor when specific boundary conditions are satisfied [Lou76, Ter84, Mon93]. In the MIG model, the character of the surface states changes across the band gap from predominantly donorto predominantly acceptor-like closer to the VBM and the CBM (conduction band minimum), respectively, since the surface states are derived from the band structure of the semiconductor. The energy at which the contributions from both bands are equal in magnitude is called the branch point, and the branch point is located at the mid-bandgap when the effective mass of electrons and holes are of equal value [Mon93]. Again this is an intrinsic property of semiconductor band structure. The Charge-Neutrality Level (CNL) is defined as the energy where the Fermi level coincides with the branch point and this is a critical energy level which determines barrier height.

PAGE 21

14 According to the MIG model, charge transfer occurs across the interface depending on the electronegativity difference between metals and semiconductors and determines the final level of Fermi-level pinning. For example, when the electronegativity of a metal is the same as the semiconductor, the final Fermi level pinning is located at the branch point of the semiconductor and there is no charge transfer. The final pinning position of the Fermi level should be above or below the branch point of the semiconductor when the metal exhibits a smaller or a larger eletronegativity than the semiconductor. Consequently, the MIG model predicts a dependency of the barrier height on the electrical properties of metals as follows [Mon88]: Obn=Oe„i + Sx(x„-Xs) (2-12) where Ocni is the charge neutral level, Sx is a parameter defined as O'lm, Xm and Xs are the electronegativities of metal and semiconductor respectively. In the recent development of the MIG [Mon88], it was shown from experimental data [Cao87, 89] that the position of Fermi level pinning was a function of metal layer coverage; the Fermi level approached its final position as the surface was saturated from submonolayer coverage (isolated adatoms) to a continuous metallic film. The dependency of Fermi level position as a function of metal coverage, which can’t be explained by the UDM, was explained in the MIG model by using the idea of metal-induced surface states at submonolayer coverage and by the continuum of metal-induced gap states at several monolayer coverage. In the beginning, the idea of MIGs had begun as a simple model lacking of microscopic knowledge of real interface structure and developed largely by theoretical calculations.

PAGE 22

15 Now it appears that the idea is gaining more attention with further refinement and supporting experimental data [Mcl88, Bur91], Chemical reactivity between semiconductors and metals was correlated to explain the deviation from the ideal Schottky barrier by the fact that chemical reactions affect the electrical properties of the interfaces. Heat of formation was used as a measure of the reactivity between metals and Si [And75], It was found that the barrier heights (Ob) between transition metals and Si exhibited a linear relation with the heat of formations (-AHf). Later, Freeouf adopted a different approach to correlate the barrier heights of the metals/Si contacts to chemical reactions [Fre80], He proposed that Ob had a relationship with Osiiicide (work function of silicides) instead of the metal work functions. Even though this idea was developed to describe metal/Si contacts, it was later applied to various compound semiconductors and developed as the Effective Work Function (EWF) Model [Fre81]. In the EWF model, Ob„ (Schottky barrier height with a metal) is given by Obn=Oeff-X (2-13) instead of Obn ~ Ometal “ X where x is the electron affinity of the semiconductor. Oefr is mainly due to the work function of the anion, Anion. The EWF model suggests that the Fermi level at the surface (or interface) is not fixed by surface states but rather is related to the work functions of microclusters of one or more interface phases resulting from either oxygen contamination (oxidation) or metal-semiconductor reactions which occur during metallization. The EWF

PAGE 23

16 model is a refinement of the original Schottky description since it does not introduce the concept of surface states. Another chemical argument proposed is that chemical reactions at the interface on a microscopic scale modify the ideal Schottky barrier via local charge transfer and creation of extrinsic interface states as a result of the interfacial reaction [Bri78, Bri90], In this argument, discrete levels of defects, native or extrinsic, were used to account for the deviation from the ideal Schottky behavior with chemical reactivity. This approach does not seem to be a main theory in that it does not propose a microscopic mechanism for Fermi level pinning. Relatively recently, Fermi level pinning was attributed to amphoteric native defects in semiconductors by Walukiewicz [Wal87, 88, 89]. A remarkable correlation exists between the Ferrm-level position (Eps) at metal-semiconductor interfaces and the Fermi level (Epi) in heavily irradiated III-V compound and column IV semiconductors [Wal88]. Table 2-1 shows the range of Fermi-level pinning positions deduced from the Schottky Barrier Heights for metal-semiconductor contacts and the Fermi level stabilization energy in heavily irradiated III-V compounds and column IV elemental semiconductors. Table 2-1. Fermi level stabilization energy in irradiated semiconductors (En) and at metal-semiconductor interfaces (Eps). Eb represents charge-neutrality level from MIG model. All energies are with respect to the valence-band Ep, (eV) Eps (eV) Eb (eV) Si 0.4 0.3 -0.4 0.36 Ge 0.07 0.16 0.18 GaAs 0.5 -0.7 0.5 -0.7 0.5

PAGE 24

17 From the similarity of the values listed in table 2-1, the same mechanism as for Epi in heavily irradiated materials was proposed for the Eps in the metal-semiconductor contacts. According to his reports, there is a Fermi-level stabilization energy (Epi) in covalent or weakly ionic semiconductors including GaAs, which is independent of the type of doping and the doping level. Therefore, this property is regarded as an intrinsic property of the semiconductors. As a consequence of this intrinsic property, native defects such as vacancies or substitutional dopants exhibit an amphoteric character depending on their energy level relative to the Fermi-level stabilization energy. For example, a Ga vacancy is a stable acceptor in n-type GaAs, but it transforms to a donor complex (Asoa + Vas) in ptype materials. This behavior results from a large electronic contribution to the defect reactions. For example, the formation energy of a Ga vacancy is lowered from ~4eV to ~0.2eV as the Fermi level varies from the Valence Band Maxima (VBM) to the Conduction Band Maxima (CBM) under As-rich condition [Zha91]. Such a transformation continues until the Fermi level reaches the stabilized position, Epi. In such a case, introduction of further electrically active species does not affect the stabilized Fermi-level. In the case of GaAs, the stabilized Fermi-level is located between Ev +0.5 eV to Ev +0.7 as can be seen in table 2-1. The amphoteric native defects model is unique in that the behavior of native defects are responsible for the Fermi level pinning, which appears to be similar to the UDM model but the behavior of defects are controlled by the stabilized Fermi level En, which is an intrinsic properties of semiconductors, similar to the MIG model.

PAGE 25

18 For many years, all the models mentioned above were examined and compared. Nonetheless, the main idea of whether intrinsic or extrinsic effects play a primary role in the Fermi level pinning still remains controversial [San85, Spi85, Ter88, Spi93], It is now generally accepted that a single theory can not explain the Fermi level pinning and a special effort is being made to build general consensus for the main idea [Spi93], Meanwhile, appropriate descriptions as to the Fermi level pinning tend to be more specific, depending on the actual structure of the interface [Fre90, Bri90] and even on semiconductor growth method [Vit93], In fact, it is readily and clearly seen that there is a range of ~0.3eV in the Fermi-level pinning, 0.5 0.8eV above the VBM by reviewing the extensive experimental data collected to date. This variation might imply that several mechanisms are simultaneously playing roles and even interacting with each other in the determination of SBH. In the latter case, understanding of the Fermi-level pinning would be more complex than expected. With the plethora of models and inability to critically discriminate between them, there is not a general model which enables us to predict the barrier height for practical applications. In GaAs technology, we can not rely on controlling the Schottky barrier height by selecting metal elements. This situation led to the need to produce a tunneling ohmic contact scheme by controlling barrier width rather than barrier height.

PAGE 26

Ex situ Contact Schemes 19 Au/GaAs Metallizations Au is frequently used as a contact metal for GaAs as well as a base metal in semiconductor processing because of its ease of deposition and etching, high ductility and conductivity. Au begins to react with GaAs at ~250°C, but significant reaction occurs at -400 C with large change in the Schottky barrier heights. Typically, Au/GaAs diodes exhibited a ~0.9eV barrier height as evaporated and less than 0.7 eV or ohmic behavior as heat treated [Gyu71, Leu85, Hol92]. The interfacial reactions between Au and GaAs upon annealing can be generally expressed as follows; Au +GaAs => Au-Ga + As (gas) which represents the formation of Ga-rich Au solid solutions or Au-Ga compounds with As loss in the case of an open system. In the case of a closed system, the As might exist in the form of a precipitate. The Au-Ga compounds can be one or a combination of the following compounds; Au?Ga 2 , AusGa, Au 2 Ga, AuGa, or AuGa 2 upon cooling [Kum76, VanSO, Yos83, Lin86, Kim90]. Unique to the Au/GaAs reactions is the formation of elongated pyramidal pits bounded by {111} planes, aligned in [1 10] directions of GaAs, which are noticeable when annealed above ~350°C [Pez86, Bau86]. In the pyramidal pits, pure Au and/or Ga-rich Au solutions were observed. These contents in the pits were separated from the GaAs substrate by an intermediate layer of Au-Ga compounds which are reaction products of

PAGE 27

20 the Au/GaAs interfacial reactions. The pyramidal reactions pits were thought to be formed through solid-state dissolution of GaAs up to 450°C although a liquid state reaction increased their size and governs overall morphology when annealed above the melting temperature (~500°C) of the Au-rich solid solutions. In addition to the reaction pits, agglomerated regions (Ga-rich Au solution) which cover the underlying pits were observed [Gyu71, Kim90]. The reaction between Au and GaAs proceeded in a highly inhomogeneous manner, resulting in formation of monocrystalline gold, a hexagonal Au-Ga phase, and precipitates of Au 2 Ga in the matrix of GaAs with certain crystallographic orientations [Bau86]. The crystallographic orientational relationships between the reaction products and the parent Au and GaAs were studied and described by the degree of misfit at the interface; Au/ AuGa layer and the Au-Ga/GaAs interfaces are formed in such a way that lattice misfits at the interfaces are minimized [Yos83]. The inhomogeneity of the interfacial reaction can result from different difilisivity of Au depending on the crystallographic orientation of GaAs and the manner that Au penetrates the native oxide to react with the substrate GaAs, which will vary depending on the thickness and distribution of the native oxide. The effects of phase structure and morphology on the electrical properties of Au-Ga compounds/GaAs were studied by depositing Au-Ga phases with various Ga concentrations on GaAs [Leu85]. They could not attribute the variation of electrical properties to the presence of specific phase because a phase mixture with nonuniform distribution was present at the interface. However, it was observed in general that the formation of the Au-Ga compounds as a result of Au/GaAs interfacial reactions reduced

PAGE 28

21 the SBH (Schottky Barrier Height) below that of pure Au. This result was confirmed by depositing and annealing chemically inert AuGa 2 on GaAs [Lin86], This study showed that chemical reactions were minimal at the AuGa 2 /GaAs interface after annealing up to 500°C because the AuGa 2 phase is an equilibrium phase. Chemical reactions between Au/GaAs were reviewed to understand the variation in the electrical properties of the Au/GaAs system by studying the details of the pyramidal reaction pits resulting from the chemical reactions [Hol92], It was shown that Au/GaAs interfacial reactions vary depending on the conditions of the initial interface; surprisingly, the interface with a native oxide exhibited the most pronounced formation of reaction pits, suggesting that the stability of the interface is critically dependent on the surface stoichiometry. Also it was observed that the reactions pits underwent Ostwald ripening phenomenon, i.e., some pits disappear and some pits grow as the reaction proceeds during isothermal annealing. A segregation of Si, which was a dopant in the substrate GaAs, was detected in the reactions pits by SIMS. Because the appearance of reactions pits indicates the dissolution of GaAs, the disappearance of the pits indicates the regrowth of GaAs, suggesting the possibility of dopant incorporation into the regrown GaAs as the reason for the change in electrical behavior from rectifying to ohmic behavior. The change from Schottky behavior to ohmic behavior was correlated to the formation of Au crystallites after annealing above 360°C [Lil85]. The ohmic behavior was attributed to the leakage currents at the periphery of the contacts. But they could not provide a reasonable mechanism for current transport mechanism associated with their observation, therefore, this result could not be regarded as a mechanism for ohmic contacts.

PAGE 29

22 In general, Au dissociates GaAs inhomogeneously producing Au-Ga compounds upon cooling, resulting in reduced barrier heights. The morphologies and details of the reaction products in Au/GaAs appear to be dependent on the relative concentration of Ga in Au-Ga compounds possibly because the variation in the Ga concentration dictates different equilibrium liquidous temperatures in Au-Ga-As ternary phase. The low melting temperature of the Au-Ga compounds is known to be a possible cause for poor morphology in the vicinity of device processing temperatures ~400°C. In terms of the effects of Au metallization on the electrical properties, Au has been known to enhance Ga out-diffusion, thereby increasing the chance that external doping elements occupy Ga sites leading to the formation of n^-layer. However, solid-phase regrowth model that will be discussed in this review, and the results of this study clearly showed that Ge doping was the consequence of solid-phase epitaxial regrowth following interfacial reactions not a consequence of Ga out-diffusion. Ni/GaAs Metallizations Ni reacts uniformly with GaAs except where the native oxide-hydrocarbon interfacial layer remained intact [San87]. Contamination at the Ni/GaAs interface affects diffusion and compound formation [Sol87]. Solid-state reactions between Ni and GaAs produced a hexagonal ternary phase after annealing at 100°C 400®C, for times of 5 minutes to 5 hours: xNi + GaAs = NixGaAs.

PAGE 30

23 The ternary phase designated as NixGaAs was epitaxial with the GaAs substrate and analyzed by diffraction analysis in TEM and XRD [OgaSOb, Che83, Lav86, San86, Che88, Gui89], The composition x varied between 2 and 4 according to those authors. The quantification of AES data yielded N^GaAs [Oga80b, Lav86, Sol87] while EDS and/or TEM measurement suggested NisGaAs [San86, Lin88], The variation in the composition ranging from 2 to 4 was observed to accompany different cJ&o ratios in the same NiAstype hexagonal structure [Che88, Gue89], An experiment on bulk material to determine the Ni-Ga-As ternary phase diagram showed that there were five ternary phases with broad homogeneity ranges extending toward the binary phases [Gue89], They all exhibited hexagonal NiAs-type symmetry and were unstable in contact with GaAs. It was proposed that NisGaAs would adopt a B8 structure since Ni-Ga and Ni-As binary systems exhibit B8 structures with lattice parameters similar to NisGaAs and the symmetry of NiAs-type structures. The lattice parameters of NixGaAs were intermediate between those of Ni 3 . 55 Ga 2 .o and NiAs [San86, 87]. NixGaAs is characterized by a hexagonal unit cell with lattice parameters ao~4A, Co~sA, similar to the Ni-As and Ni-Ga systems (NiAs: a<> = 3.619, Co = 5.034A and Ni 3 Ga 2 : a« == 4.0, Co = 4.983A) [San86, Che88]. Bright-field TEM images of NixGaAs showed a columnar structure with microtwins and diffraction patterns from TEM and XRD exhibited twin variants with respect to the GaAs matrix as well as several epitaxial relationships with the substrate GaAs [Gui89]. NixGaAs was more often observed in thin film reaction and is thought not to be an equilibrium phase. Due to the epitaxial relationship of NixGaAs with, the nucleation

PAGE 31

24 barrier for formation of NixGaAs is much lower in thin films than that of the equilibrium phases such as NiAs and NiGa [Lav86], Also, the formation of NixGaAs might be favored kinetics since Ni has been found to be the diffusing species in the reactions with GaAs. Contrary to Ni, Ga and As were relatively immobile at temperatures below ~300°C and the formation of NixGaAs was triggered by in-diffusion of Ni [San87b, Lin88, Che88], Ogawa showed that Ni 2 GaAs separated into binary NiGa and NiAs by annealing at 500°C for 5 minutes and suggested that Ni 2 GaAs was not stable in the presence of excess GaAs [Oga80], In another study, Ni 2 GaAs was found to be stable up to 600°C, but the stability was dependent on the orientation of the substrate. Ni 2 GaAs substrate was more stable on (1 1 1) versus (001) GaAs [Lav86, Gui89], On the bulk materials, the phases after annealing at 600°C for 1 hour were NiGa and NiAs [Gui89], This indicates that there are only tie lines connecting GaAs and NiGa and NiAs binaries in their experimentally determined Ni-Ga-As ternary phase diagram [Gui89], After annealing 500°C, 5 minutes, NiAs or As-rich phases were found near the GaAs and P-NiGa near the surface [Oga80b], In this study, the role of Ni was explained to be a highly reactive element which eliminated the contamination layer such as an oxide layer, and thus assists the dissociation of GaAs. Decomposition proceeded through NiAs precipitation in a matrix ofNi 2 GaAs at temperatures higher than 350°C. After annealing at 600“C, p-NiGa and NiAs grains with an average size of 0.7 |im were observed. The P-NiGa showed an epitaxial relationship with GaAs substrate. On the contrary, NiAs grains were randomly oriented [Lav86, Gui89]. In another investigation, NiAs and NiGa formed after annealing at 600°C for 1 hour was epitaxial on the GaAs substrate [San87b].

PAGE 32

25 The Schottky barrier height was increased when the Ni 2 GaAs was formed from 0.76eV to 0.83eV, which was similar to the AuGeNi with Ni diffusing. The barrier height was reduced by decomposition of the Ni 2 GaAs [Lav86], Sheet also increased at the temperatures where the ternary phases was formed (200 300°C) and decreased at the temperatures of the decomposition of the ternary phase (above 400°C) [Gui89], All the phases fi'om the Ni/GaAs reactions were highly textured with epitaxial relationships to the substrate (except cubic NiGa) [Gui89], The intermediate NixGaAs phases were isostructural with the final, equilibrium binary phases. In conjunction with the potential effects of Ni/GaAs interfacial reactions on the electrical properties of Ni-containing ohmic contact metallizations, Ni improved greatly the uniformity of the reactions. The Ni-Ga and Ni-As compounds are characterized by high melting temperatures relative to compounds in Au/GaAs reactions, which is believed to lead to improved thermal stability of the contacts. Decomposition ofNiAs phase containing Ge was found to be responsible for regrowth of GaAs doped with Ge at Ga sites in this study. AuGe/GaAs Metallizations Adding Ge to Au modifies the metallurgical reactions with GaAs because 88wt% Au with 12wt% Ge forms a eutectic composition with a melting temperature of 363°C. However, even adding a small amount of Ge (~0.6wt%) drastically changed the morphology of the reaction products(Au 7 Ga 2 , AusGa) and lowered compound formation

PAGE 33

26 temperatures with increasing Ge concentration [Kim86], In addition to Au-Ga compounds, a Au-Ge-As phase was found to cover the contact surface at ~400°C, and to a lesser extent above ~400°C. Regrown GaAs was also observed with the disappearance of the Au-Ge-As phase, which indicates local melting and solidification [Kim90], After annealing above ~400°C, melting of multi-phases initiated by the Au-Ge eutectic phase and/or Au-Ga compounds following recrystallization were observed, which generally resulted in the formation of ohmic contacts with specific contact resistance of ~10’’ Qcm^. It was suggested in the process that the Au-Ge liquid phase dissolves Ga and As into the melt, and upon cooling, recrystallized or regrown GaAs would incorporate Ge forming heavily doped GaAs. Ge was suggested to be critical to initiation of the melting process by formation of Au-Ge or Au-Ge-As eutectic compositions, as well as formation of the n"^layer leading to ohmic contacts [Ili83], After annealing at 450“C 500°C which is well above the melting temperatures of Au-Ge and Au-Ga compounds, Au-Ge, Au-Ga compounds and epitaxially regrown GaAs were found to be major reaction products, all of which were governed predominantly by liquid phase reactions [Kim90], The formation of ohmic contacts was attributed to incorporation Ge during the solidification of GaAs. In-diffusion of Ge and out-diffusion of Ga were observed after annealing below the Au-Ge eutectic temperature as well as above the temperature. It was reported that a dramatic reduction in the barrier height from 0.77eV to ~0.4eV occurred even after annealing below the Au-Ge eutectic temperature without major reactions [Ili83, 87]. These experiments were correlated to the formation of a heavily doped n"^ GaAs layer and the formation of ohmic contacts with Au-Ge metallizations [Gyu71, Ili83, Kul86]. Again,

PAGE 34

27 the formation of n^-GaAs layer based on Ga out-diffusion is believed to be unreasonable as will be obvious by the results of this study. For the samples annealed above or below the Au-Ge eutectic temperature, the reduced barrier height was attributed to the formation of a gradually disordered interface rather than the formation of heavily doped n^ layer, or formation of graded heterostructure was suggested to account for the formation of ohmic contacts from the idea that the interface would be heavily restructured after complicated alloying reactions above the Au-Ge eutectic temperature [Kir87], In summary, AuGe/GaAs exhibited a lower contact resistance than Au/GaAs. The overall reaction morphology, such as the formation of pyramidal reaction pits, was still observed with some extra phases due to the presence of Ge. Similar to the case of Au/GaAs reactions, Au-Ge eutectic contacts also lacked a uniform morphology. NiGe/GaAs and Other Bielements/GaAs Metallizations As reported in the chapter 1, marginal thermal stability, irregular morphology, and degradation of contact resistance with time in AuGeNi metallizations has been correlated with the presence of Au. Au-free metallization was pursued to determine if this would eliminate the problems and resulted in simpler bielemental metallization such as NiGe [Tan92], PdSi [Wan88], and PdGe [Tsu91]. The metallurgical reactions in Au-free metallization are often described as sintering reactions or solid-phase reactions emphasizing that the overall reactions are completed through solid-state diffusion.

PAGE 35

28 Without Ni, very little interdiffiision between Ge and GaAs was observed after sintering at 450°C, 30 minutes [And78], This study showed that the presence of Ni overlayer on top of Ge greatly enhanced interdiffusion of Ge into GaAs resulting in the formation of ohmic contacts. This supports the model of a heavily doped n^ layer as a mechanism in forming ohmic contacts. Solid-phase epitaxy of Ge was investigated to form non-alloyed contacts to n-GaAs by using Ge/Pd/GaAs metallization [Mar87]. Pd and Ge reacted to form PdGe and excess Ge was transported through the Pd layer to grow epitaxially on the GaAs substrate. As to the mechanism for the formation of ohmic contacts, tunneling through Ge/GaAs interface and formation of n^-GaAs was suggested. However, the study failed to provide information regarding type or concentration of free carriers in the Ge layer nor experimental evidence for the existence of n^-GaAs layer. To clarify the mechanism for ohmic contacts, backside SEMS analysis was used to measure Ge incorporation into the GaAs [Pal90]. This study showed that the concentration of Ge (~1 x lO'^cm’^) at the contact with GaAs was correlated to the onset of ohmic behavior. Based on their metallurgical study and Ge detection, dissolution of GaAs leading to the formation of Pd 4 GaAs and following regrowth of GaAs was suggested as the primary mechanism for formation of ohmic contacts. Even though this backside SIMS could not provide reliable quantitative data regarding the extent of regrowth, this result supports the Ge incorporation into GaAs, resulting in the formation of n"^ layer. The characteristics of Si/Pd/GaAs metallization were investigated to compare to those of Ge/Pd/GaAs metallization [Wan88]. It was shown that the epitaxial Ge layer on the

PAGE 36

29 GaAs substrate or low barrier heterojunction was not responsible for ohmic behavior. They suggested a regrowth mechanism in which Pd reacted with GaAs, and produced a very thin Pd 4 GaAs. However, excess Si can react with Pd 4 GaAs to produce the following: 2Si + Pd 4 GaAs(Si) => 2Pd2Si +GaAs(Si). A final product of this reaction sequences is regrown GaAs doped with incorporated Si. Si was suggested to have diffused into the ternary phase before decomposition and occupied Ga sites in regrowth, creating an n^-layer doped ~ 2 x 10‘^cm^ From this study, the ohmic behavior in Si/Pd/GaAs and Ge/Pd/GaAs were attributed to the formation of n^ GaAs. This solid-phase regrowth mechanism was further tested in Ni/Si/GaAs, which clearly showed that the regrowth could take place due to the relative thermodynamic stabilities between phases [San88]. Another study using Ni/Si, however, did not support the regrowth model based on the formation and decomposition of NixGaAs phase because they could not detect Ni^GaAs phase even after annealing at 300°C [Tak92]. The concept of the regrowth mechanism was applied to explain ohmic behavior for Ni/Ge/GaAs metallizations [Tan88]. First, Ni reacted with GaAs to form NixGaAs, then decomposition of the ternary phase was driven by the lower free energy of formation of NiGe to result in: NixGaAs + xGe => xNiGe+ GaAs (regrown). Again Ge was suggested to be incorporated into regrown GaAs, forming n"-GaAs. There was a minimal amount of Ge (29 38 at. %) in Ni/Ge thickness ratio to form ohmic contact. The regrown GaAs was characterized by a high density of stacking faults.

PAGE 37

30 microtwins, and precipitates. The regrown GaAs was postulated to contain a high concentration of dopant (e g. Ge with a concentration of low-lO'^cm’^). However, no experimental data was presented to support the incorporation of Ge (lO'® lO^Vm'^) nor the formation of n"^-GaAs. An important unresolved issue for the regroAvth mechanism is the specific mechanism of incorporation and site selection of doping elements into regrown GaAs [San88]. One speculation concerning Ge transport is a higher diffusivity and/or solubility of Ga in the metallization layers than As. Thus, prior to regrowth of GaAs a greater number of Ga atoms diffuse fi-om the ternary phases into the outer layers of the metallization. When regrowth takes place resulting from decomposition of PdyGaAs or NixGaAs, regrown GaAs is deficient in Ga relative to As. Thus Ge could occupy Ga sites in regrown GaAs. This speculation also assumes Ge in-diffusion to the NkGaAs or PdyGaAs phases before regrowth, which wilt be shown to be not correct by our study. A Ga vacancy-dependent diffusion model for ohmic contacts was suggested as another attempt to explain how Ge atoms are transported and occupy Ga sites [Gup85]. This model postulates grain-boundary diffusion in Au layer to generate Ga vacancies to account for the formation of ohmic contacts. This model is based solely on the role of Au, which enhances out-diffusion of Ga. Thus this model is not capable of explaining how another metallizations without Au still can form ohmic contacts. They also justified their model based on the fact that contact resistance is high after either low or high temperature annealing, but low for intermediate temperatures. However, this behavior is universally observed regardless of the presence of Au.

PAGE 38

31 The results of our study clearly showed that neither the site selection nor the transport of Ge was so simple as suggested by the diffusion model or speculations. The site selection and transport of Ge are determined by evolution of interfacial phases followed by regrowth of GaAs. AuGeNi/GaAs Metallizations AuGeNi metallization was first introduced in 1967 by Braslau et al io form ohmic contacts to GaAs-based microwave devices [Bra67], This method is still used most often to form ohmic contacts to n-GaAs. At first, the effect of Ni in the AuGe eutectic composition (Au 88wt%, Ge 12wt%) was to prevent the AuGe melt from “balling up” due to surface tension. Therefore, the three elements were evaporated so as to form a bilayer structure, i.e., Ni/AuGe/GaAs. However, it has been shown by several investigators that the sequence of each layer did not affect the final metallurgical properties or electrical properties [Chr79, Mar83]. As a second postulate, it was suggested that Ni improved the uniformity of surface morphology by improving the wetting of liquid Au-Ge to GaAs [Rob75]. The high reactivity of Ni was correlated to dissociation of GaAs and the following reaction was hypothesized: Au + Ni + GaAs => Au-Ga + Ni-As. where Ni was thought to be a catalyst [Oga80]. During the reaction, Ge was trapped in the Ni-As compounds near the GaAs interface. Dissociation of GaAs by Ni was proposed to occur through solid-phase reaction which should lead to uniform alloying behavior in

PAGE 39

32 the Ni/AuGe/GaAs [OgaSO], A major difference from the dissociation of GaAs by Au versus Ni/Au was that As was retained by the stable Ni-As compounds. The ability of Ni to penetrate thin native interfacial oxides also led to a uniform reaction. The Au-Ge eutectic was not thought to play a major role in the morphology by forming a liquid phase at 363°C since the Ge concentration for the Au-Ge eutectic was expected not to reach nor maintain because Ge was gettered by Ni via solid-state diffusion [IH84, Rel89]. Again, this could improve the surface morphology compared to AuGe/GaAs. It was suggested that Ni changed reaction between AuNiGe and GaAs by initiating the reaction with GaAs, resulting in a smoother interfacial morphology [Shi87]. Near the AuGe eutectic temperature, Ni and Ge accumulated at the interface with GaAs, while Ga from GaAs dissociation built up at the surface [Rob75]. The barrier height, Obn, increased from ~0.7eV to ~0.9eV as Ni accumulated at the interface [Rob75, Lav86]. When annealed above the eutectic temperature at 400 to 600°C, the surface morphology became nonuniform by formation of rectangular, submicron reaction [Chr79, Oga80, Hei82, Pro87, Buh91]. A higher concentration of Ni was found in the pits versus in the surrounding matrix, and postulated to control the specific contact resistance [Chr79]. Ni-Ge clusters were found on the surface and their distribution was directly correlated to the formation of ohmic contact [Hei82]. The surface morphology and contact resistance were direct functions of the Ni/Ge ratio. The higher the Ni concentration led to better surface uniformity, while higher Ge concentration led to lower contact resistance [Pat86, Kov90, Buh91, Chu94].

PAGE 40

33 Ni is always found to diflRise toward the interface and to react with GaAs. Ge tended to follows Ni distributions. Au was a rapid diffusing species, and often formed Au-Ga compounds upon cooling [Rob75, Witt77, Chr79, OgaSO, Hei82, Kua83, Mar83, Ili84, Rel89, Kim90]. A transmission electron microscopy (TEM) study showed that elemental distributions were correlated to the formation of some compounds. Generally, annealing above 400®C for a few minutes produces the following compounds; NiGe with a small concentrations of Ga and As, NiAs with Ge and Ga, and AuGa [Kua83, Shi87]. Spatially, the NiAs phase was in direct contact with GaAs resulting in a smooth interfacial morphology. A good ohmic contact with low contact resistance was attributed to the Ni 2 GeAs phase being in contacting with GaAs. Diffusion of Ge from Ni 2 GeAs into GaAs to form n layer was proposed [Kua83]. NiAs phases containing small concentrations of Ge were suggested as tunneling ohmic contacts and the contact resistance was a function of the NiAs(Ge) coverage at the interface [Shi87]. Besides NiAs phases, AuGa compounds were found near the free surface [Shi87, Rel89]. During the initial stage of annealing, the high concentration of Au resulted in AuGa phases at the interface with irregular morphology, which generally increased the contact resistance. However, at higher annealing temperature, more NiAs phase with a highly oriented epitaxial relation was found at the interface [Kim90]. The epitaxy was suggested to result from a close lattice matching between NiAs and GaAs. It was noted that better ohmic behavior was produced when the phase contacting GaAs was changed from AuGa to NiAs(Ge) phase. The metallurgical reactions between AuGeNi/GaAs were different from those between AuGe/GaAs due to the high reactivity of Ni with GaAs. Dissociated Ga and As formed

PAGE 41

34 compounds with Ni, and uniform interfacial and surface morphologies are observed. As a common feature, Ge was found at the interface with GaAs for alloyed metallizations with and without Ni and led to formation of ohmic contacts, Even though the metallurgical reactions in AuGeNi/GaAs are different from AuGe/GaAs, the mechanism for formation of ohmic contact appears to be dissociation of GaAs followed by alloying with the metallization elements. Upon cooling, Ge was postulated to be incorporated into GaAs by occupying Ga sites creating a heavily doped GaAs layer with ~10*^cm'^ Ge. This concentration of Ge leads to an n^ layer through which electrons could tunnel [BraSl], This mechanism is broadly known as the “doping” model, and it is widely accepted to n-GaAs to account for the formation of ohmic contacts. Backside SIMS measurements in an alloyed AuGeNi system were used to resolve Ge doping [Bru89, Sch90]. Diffusion of Ge and Ni into GaAs was observed. However, the major reduction in contact resistance was poorly correlated the observed Ge [Bru90]. Cross-sectional TEM showed that the NiGe^rAs^ phase was in contact with the substrate. Based on these result, Bruce et al. suggested an increased n -doping level was less important to the reduction of the contact resistance than the formation of Ni-Ge-As phase [Bru90]. It appears that these measurements are not clear data to support Ge incorporation into GaAs because of the lack of high spatial resolution and quantification.

PAGE 42

Hybrid Contact Metallizations 35 Many variants to AuGeNi systems have been developed in order to improve thermal stability and to prevent interdifflision. For high-temperature applications of microelectronics, the solid-state devices are required to function reliably for long time in hot environments (>300®C). And the interdifflision is considered to be particularly important in the case of a shallow junction. W 60 N 40 has been added to the AuGeNi system to prevent the reaction between Au and GaAs and the solid-phase reaction of the Ni/Ge layer could form ohmic contacts [K0I88]. A Cr layer was inserted between Au and Ge to prevent the formation of Au-Ge eutectic [Wil88], Similarly, \VSi 2 layer acted as a diffusion barrier between Au and Ge to prevent both Ga and As out-diffusion as well as Au-Ge eutectic formation [Gup90], These metallizations formed ohmic contacts with contact resistances of 10'^ 10’® Qcm^, and exhibited smooth morphology above ~600°C by preventing interdifflision between elements. Indium was added to metallizations to study the formation of graded GaxIni.xAs heterostructure in addition to the formation of n^-GaAs. Indium films of 50 200 A thickness were added between the Ni and Ge layers and shown to reduce contact resistance by forming GaxIn^xAs (x~0.4) [Oku94], TEM images of this study showed that the metal/GaAs interface was covered by regrown GaAs and GaxIn^xAs. Compared to Ni/Ge bimetallization, the contact resistance was reduced from -1.2 Qmm to -0.3 Qmm in this study, which was attributed to the formation of n^-regrown GaAs and GaxInj.xAs.

PAGE 43

36 In Situ Contact Schemes Contrary to the ex situ contact scheme, in situ contact schemes refer to the processes which do not utilize alloying reactions or solid-phase reactions to incorporate doping elements into the surface of the GaAs substrate. Instead, very heavy doping is accomplished in situ during the growth of GaAs, or heterojucntions are formed to lower the barrier height between metals and GaAs. Heterojunctions are defined as junctions between two different semiconductors. In principle, they may be either gradual or abrupt junctions. A gradual junction is one in which two bulk crystals are joined by continuously varying composition, e.g., GaxIni.xAs/GaAs where x can vary from 1 to 0. An abrupt junction is one in which there is a sharply defined interface between two homogeneous semiconductors, e.g., Ge/GaAs. In situ contact schemes are described as nonalloyed schemes because the ohmic contacts are formed without post-deposition heat treatment. Nonalloyed schemes were studied and developed in efforts to reduce the drawbacks of conventional metallization schemes, such as poor morphology, reproducibility, severe interdiffusion and reactions with GaAs, and poor thermal stability. Molecular Beam Epitaxy (MBE) is frequently used to grow heavily doped n^-GaAs where n^ can be as high as ~mid-10*^cm'^. The reason MBE can produce n-type GaAs layers with a carrier concentration higher then most conventional growth techniques is that the incorporation of the dopant is not limited by solubility or thermodynamic equilibrium conditions, but may be controlled by surface kinetics. Thereby, MBE can be used to

PAGE 44

37 incorporate ten times more carrier than the concentration obtained in bulk doping. Sn was selected because it is known to be a less-compensated amphoteric dopant versus Si or Ge. A free electron concentration as high as 6 x lO'^cm'^doped-GaAs could be achieved with Sn [Bar78, Dil79], In cases where high-10‘^cm'^dopant concentration in GaAs were achieved, specific contact resistance as low as 2 x lO'^Q cm^ were obtained simply by depositing metals without heat treatment. Even ~1 x lO^^’cm'^ Si doping was reported and yielded -1.3 x lO"^ Q cm^ with in situ metallization [Kir85]. Formation of ohmic contacts was attempted through creation of low barriers between metal/Ge layers in conjunction with favorable conduction band alignment at the Ge/GaAs heterojunction. The localized potential barriers of ~0.45eV at the metal/Ge interface and less than 100 meV at the Ge/GaAs heterojunction are more favorable for ohmic behavior at room temperature than a single ~0.8eV barrier at the metal/GaAs interface. The Ge layer must be heavily doped, up to -lO^'^cm'^, often with As while GaAs was doped with Ge or Si on the order of 10*’ ~10**cm’^. Low contact resistances, typically -lO'^Qcm’ [Dev80] and even as low as -lO'^Hcm’ [Sta81] were measured with smooth interfacial morphology. The Ge/GaAs heterojunction was studied because Ge has nearly equal lattice parameters (within 0.5%), compatible crystal structure, and the thermal expansion coefficient of Ge (6.6 x 10'*/°C) matches well with that of GaAs (6.0 x 10'V°C). Most importantly, the conduction band discontinuity at the Ge/GaAs interface, AEc, is 60meV which will exhibit a negligibly small tunneling resistance [Sta81, Bal83]. Another heterojunction scheme is to GaiJnxAs at the interface. This scheme is based on the fact that Fermi level pinning occurs at or in the conduction band on InAs [Woo81].

PAGE 45

38 Because the conduction band discontinuity between InAs and GaAs establishes an barrier, a graded Gai.xInxAs between InAs and GaAs is necessary to remove the abrupt discontinuity in the conduction band. In this scheme, tunneling is not required and low resistance contacts can be made for a wide range of doping without need of alloying to form n^ surface layers. InAs and Ga^xInxAs layers were grown in situ on GaAs typically by MBE process and produced contact resistances of -lOÂ’^ficm^ [Kum89, Meh89]. This was at least one order of magnitude lower than the contact resistance achieved by ex situ contact schemes, which are typically, -lO'^Qcm^. Even though InAs-based contact or in situ heavy doping scheme provided an extremely low contact resistance, these schemes are not practical because of the incompatibility of MBE with semiconductor processing and the incapability of the InAs-based structures to endure high temperature processing. Current Status of Contact Metallizations In pursuit of lower resistance ohmic contacts, efforts to improve AuGeNi metallizations are being made [Chr95, Hao96]. In terms of understanding for formation of ohmic contacts, the solid-phase regrowth model is the most recent development and is being applied to explain not only formation but also degradation of ohmic contacts [Wan95]. Important refinements on the solid-phase Regrowth model have been discussed by Holloway et al [Hol91, Li92, Lam92, Kim96]. They clearly demonstrated that

PAGE 46

39 formation of ohmic contacts coincides with regrowth of GaAs, and contact resistances were a function of morphology of regrown GaAs [Li92], The results they reported, the measurement of free carrier density (10‘^ 10^“cm'^) in regrown GaAs, is only the direct data to date for formation of n"^ layer in regrown GaAs [Li92], Metallurgical concentrations of Ge up to -lO^’^cm*^ as a function of the extent of regrowth is reported for the first time in this dissertation. Therefore, it is believed that Ge incorporation during the regrowth creates n^ surface layer, resulting in tunneling ohmic contacts. Equally important to these data, microscopic reaction paths through which Ge could occupy Ga sites in regrown GaAs were clarified for the first time [Kim96]. Our results are believed to be the starting point to more precisely evaluate current metallizations. It should be acknowledged that experiments by other researchers using simpler system such as Ni/GaAs [OgaSO, Lav86, Che88, Gui89] and thermodynamic interpretation [Bey84, Moh95] are also important contributions to the understanding of ohmic contacts.

PAGE 47

40 (a) Before contact (b) After contact Figure 2-1. Energy band diagram for metal and n-semiconductor in electrical and thermal equilibrium. are the work function of the metal and semiconductor, respectively.

PAGE 48

41 Metal n-Semiconductor Figure 2-2. Energy band diagram for a Schottky barrier formed by a metal and n-semiconductor with surface states. Egg is the surface state energy levels within the bandgap. are the work function of the metal and n-semiconductor, respectively. This diagram assumes acceptor-like surface states.

PAGE 49

OiAPTER 3 EXPERIMENTAL PROCEDURE The majority of the experimental work for this dissertation was carried out at the Department of Materials Science and Engineering at the University of Florida, Gainesville, Florida. The samples were prepared by using an electron-beam evaporator in the MICROFABRITECH, University of Florida, and characterized by using analytical instruments in the Major Analytical Instrumentation Center. High spatial resolution EDS analysis was carried out in the metal/ceramic division at the Oak Ridge National Laboratory, Oak Ridge, Tennessee under the SHaRE program. GaAs Substrate Treatment All the GaAs wafers used in this study were n-type GaAs grown and bulk doped with Si by LEC (Liquid Encapsulated Czochralski) in Sumitomo Ltd., Japan. The wafers were doped with two different levels, 2 x lO'Am’^ and 2 x 10‘W^ as measured by the Hall technique at room temperature. The wafers were cut on the { 100} ± 0.5° plane and had average thicknesses of 400 500 pm. In order to be mounted onto the sample holder, 42

PAGE 50

43 GaAs sections were cleaved into rectangular-shaped 1 to 2 cm small pieces along <1 10> planes. The GaAs pieces were cleaned in a ultrasonic bath of TCE, acetone, and methanol followed by N 2 blow drying. Prior to evaporation of metal films onto the surface, the GaAs substrate was cleaned in HChHzO (1 : 1 by volume) for 60 seconds to reduce the native oxide and then mounted on the sample holder. The sample holder was made of Ta sheet and a ceramic block. After the mounting, the holder was immediately loaded into the vacuum chamber of the electron-beam evaporator. The chamber was promptly pumped down. To anneal samples without breaking vacuum, a resistive wire radiation heater was located behind the sample holder as illustrated in Figure 3-1 . Metal Evaporation and In situ Annealing Samples with in situ annealing After the loading the sample, the chamber was pumped by a Varian M-6 diffusion pump system with a liquid nitrogen trap. When the base pressure reached ~ 1 x lO"® Torr or less, the sample holder was radiatively heated using a regulated DC-power supply which was adjusted to generate ~15A (ampere) DC current to the Ta radiation heater. When the temperature reached ~6 1 0°C in 1 minute, the samples were maintained at the temperature for 15 seconds to desorb residual oxides on the GaAs. After the ~15 second hold, the power to the Ta heater was turned off and the sample was allowed to cool down

PAGE 51

44 to room temperature under vacuum. The temperature of the sample was determined by a 0. 125mm diameter K-type thermocouple embedded between the sample and the Ta substrate holder. This heating was also thought to cause outgassing of any volatile residual impurities on the heater and the sample holder. When the base pressure reached mid1 O'”' Torr or lower. First, 65 oA of Ni film was evaporated onto the GaAs substrate. During Ni evaporation, the base pressure increased to less than 1x10^ Torr. General evaporation conditions were electron beam current ~45mA with acceleration voltage of 8kV, and deposition rate of ~3 A/sec. The thickness of Ni layer was monitored by a crystal oscillation monitor, QXM-500 purchased from Kert J. Lesker Company. After Ni evaporation, an annealing step was carried out using the radiation heater without breaking vacuum. The base pressure before the annealing was typically 4 7 x 10 Torr. The current to the heater was manually adjusted by monitoring the temperature of the sample. The sample temperature reached 300°C in one and a half minutes and stabilized at 300± 4°C with a heater current of 7 ampere. During annealing, the base pressure rose for the first ~1 minute to high lO'^ Torr, then dropped back to the initial pressure. Annealing step caused the Ni to react with GaAs to form an interfacial layer, Ni 2 . 4 GaAs ternary phase. Deposition of 65oA of Ni followed by an annealing at 300°C for 15 minutes without breaking the vacuum was the standard process for this study and will be referred to in situ annealing. In addition to 300°C, some 650A Ni films were in situ annealed at 400°C for 15 minutes. In situ annealing without a specification of temperature automatically indicates 300“C for 15 minutes. Since the primary purpose of this study was

PAGE 52

to investigate the evolution of interfacial phases consisting ofNi and GaAs and their effects on ohmic contacts, the Nb^GaAs phases were produced as described above. After the in situ anneal, samples were allowed to cool to ~60°C and a Ge layer was evaporated at a rate of -sA/sec. During the Ge evaporation, the sample temperature increased to ~80°C. The pressure during Ge evaporation was less than 8 9 x lO'^ Torr. The thickness of Ge layer was varied from 250 to 750 A. Evaporation of Ge was followed by Ti evaporation onto the Ge layer at a deposition rate of ~4A/sec. Unique to Ti evaporation was that the base pressure did not increase during the evaporation due to gettering pumping by the Ti film. The sample temperature before Ti evaporation was ~60 -70°C and increased to ~100°C when evaporation was completed. The thickness of Ti layer was varied from 250 to 75oA. The overall structure with the in situ anneal will be designated as GaAs / Nb. 4 GaAs / Ge / Ti. Some in situ annealed samples were prepared only with Ge or Ti layers. Samples without in situ annealing. In addition to samples with in situ annealing, a number of samples were prepared without the in situ annealing for comparison. The entire procedure was the same except that the first 65oA Ni was not in situ annealed before the Ge and Ti evaporation. Thus the layer structure without the in situ anneal were GaAs / Ni / Ge / Ti where the thickness of Ge and Ti were varied from 3 00 A to 900 A. In the case of the samples with in situ annealing, it took approximately one hour to prepare from the Ni deposition to the Ti deposition. And for the samples without the in situ annealing, it took only ~15 minutes. Throughout the evaporation, two types of

PAGE 53

shadow masks, circular shape of 0.5mm diameter made of stainless steel and lines of 46 ~320|im X 4000 pm made of Ta, were used over the substrate. Post-Evaporation Annealing After the evaporation was completed, samples were exposed to atmosphere and cut into two piece. One was kept for an as-deposited sample and the other piece loaded into the electron-beam chamber, and annealed in vacuum to avoid oxidation of the Ti. When the background pressure was 1-2x10'* Torr, the sample was annealed at 500°C for 5 minutes by the radiation heater. The standard annealing temperature in this study, 500°C, was reached within 2 minutes and stabilized at 500”C ± 5. During the heating the background pressure reaches < 3 x 10"* Torr for the first 1 min and gradually decreased back to 1 2 X 1 O'* Torr during the remaining time. Some samples were annealed at 530, 550, and 600°C for 20 and 35 minutes. Electrical Measurements As-deposited samples and vacuum annealed samples were electrically characterized by their current-voltage (I-V) characteristics. These measurements were conducted at room temperature. For moderately doped semiconductors, the I-V characteristics is controlled by thermionic emission and the Schottky barrier height. Generally, I-V data was obtained by measuring the current flow between two front surface dot contacts. The resulting

PAGE 54

47 current above breakdown can be extrapolated back to zero, which represent the reverse bias breakdown voltage for one of the two back-to-back Schottky diodes formed by the metal-semiconductor contacts. Changes in this breakdown voltage may be related to changes in barrier height, d>bn. For ohmic contacts, the linearity was typically on the 1mA/ IV scales. The I-V characteristics were measured and recorded using an Tektronix 177 curve tracer, and by an automated system which consists of an IBM PC with IEEE-4888 communications, a Hewlett-Packard 61 12A DC power supply, and a Hewlett-Packard 3478 A multimeter. Some of the measurements on the Tektronics curve tracer were instantly recorded with Polaroid photos. Typically more than tow contacts were measured per sample and the curves shown are average, among the many contact dots or lines. Analytical Characterizations Auger Electron Microscopy (AES) and Secondary Ion Mass Spectroscopy (SIMS) were used to obtain compositional information. Transmission Electron Microscopy (TEM) was used to investigate the microstructure and phases of the samples. These data were correlated with electrical properties of the samples.

PAGE 55

Auger Electron Spectroscopy (AES) 48 The AES technique has been used extensively to investigate the interdiffiision in a variety of materials [Smi94], Since the information depth is determined by the escape depths of the Auger electrons, which are typically 0.5 l.Onm, destructive depth profiling by ion sputtering along with simultaneous measurement of Auger peak intensities is used to determine the distribution of elements with respect to depth. AES was used in this study was to measure interdiflfusion within the entire layer structure. Since the sensitivity of AES is limited to ~ 1 at %, the data collected by AES was cross-checked with the data collected by SEMS, which has a much lower sensitivity. A Perkin-Elmer PHI 660 Scanning Auger Electron Spectrometer was used to obtain depth profiles in this study. Typical electron beam parameters were 5 KeV primary electron beam energy with a 30nA beam current. For depth profile, 5 KeV Ar ion gun with rastered current 25nA was used for sputtering. Secondary Ion Mass Spectroscopy (SIMS) Secondary ion mass is an appropriate technique to determine elemental distributions because of its high sensitivity at least an order of magnitude better than AES or electron microprobe analyses [Reu81]. The depth resolution of SIMS is dependent upon several variables such as surface roughness and thickness nonuniformity, ion beam mixing, knockon effects, and preferential sputtering [Wil89]. Since the sputtering yield of each

PAGE 56

49 constituent is different, a slightly roughened surface will be produced, and the depth/time is different for different layers. Also, the primary ions mix the elements and degrade the resolution. Nevertheless, valuable information with high sensitivity was obtained from SIMS which complement the data from cross-sectional TEM and AES depth profiles. SIMS data were collected with a Perkin-Elmer 6600 system with a 5 KeV Cs^ primary ion beam and positive secondary ion detection ( CsX^ cluster ion spectrometry, where X is the impurity element of interest). The primary ion beam current was 40 nA and the raster size was 400 x 400 |am^ with 55% gating (detected area: 220 x 220 |im^). Transmission Electron Microscopy (TEM) Unlike SIMS and AES, TEM requires special and extensive sample preparation. First, samples were sliced into strips ~250pm thick by using a dicing saw. The strips were bonded face to face with M-bonding followed by curing in an oven maintained at 1 1 0°C for 1 hour. Bonded strip samples were mechanically ground down to 15-20 pm, then mounted on the 3mm diameter Cu ring for mechanical support. Finally, the sample on the Cu ring was further thinned using a Gatan ion miller until a ~0.2mm diameter hole was visible. Typical ion milling conditions were a 4k V, 1mA beam at a milling angle of 13 17 Sample preparation is a critical step for successful TEM analysis, and extreme caution was essential. In TEM, the whole area of interest is simultaneously illuminated by the electron beam contrary to STEM (Scanning Transmission Electron Microscopy) where a high density

PAGE 57

50 electron is required in a small probe. In TEM, the conventional image is obtained using an aperture in the back focal plane which allows only one diffracted or transmitted electron beam to form the image. Contrast in the image may result from either diffraction or phase. A bright field image is formed if the directly transmitted beam is selected and a dark field image if a diffracted beam is selected. A diffraction pattern is formed on the back focal plane of the objective lens in the transmission electron microscope. Hence, if the diffraction pattern is focused onto the back focal plane of the objective lens and the objective aperture is removed, the diffraction pattern will be visible on the viewing screen. In this study, most TEM data were collected using a JEOL 200CX for analytical or a JEOL 4000FX for high resolution analysis. In the JEOL 200CX, electrons emitted from tungsten filament were accelerated to 200keV. The majority of the bright field images and diffraction patterns were taken along the [110] zone axis of GaAs. However, some images and diffraction patterns were taken along the [ 1 1 1 ] zone axis of GaAs for comparison. In the JEOL 4000FX, a LaBe filament emitted electrons which were accelerated to 400keV. In some samples, energy dispersive analysis of fluorescent X-rays (EDX) were used. The spatial resolution of the EDX analysis was 600 A. High spatial resolution X-ray microanalysis was performed in a Philips EM400/FEG analytical electron microscope equipped with a field emission gun and ED AX 9100 energy dispersive X-ray spectrometry (EDX). The voltage for the EM400 was lOOkV and the probe size was ~1.5nm in the STEM mode and ~5 -lOnm in the TEM mode. Composition measurements were performed in the STEM mode with a ~1.5nm diameter, ~0.67nA

PAGE 58

51 probe. A low-background, cooled holder (-130°C) was used to minimize contamination under the focused probe. The “in-hole” spectrum from the sample was corrected for the normal in-hole counts associated with the fluorescence from the entire specimen area resulting from uncollimated radiation. The convergent beam electron diffraction pattern was used for precise location of the interphase boundary by reflections from both phases. The primary beam position was repeatedly checked during EDX acquisition in order to minimize any specimen drift. A standardless quantification program, NEDQNT2, was used to convert the observed peak intensities to composition for elements with atomic number Z > 10 [Lor94].

PAGE 59

52 Ta sample holder ULQJ* -Ta wire for radiation heating Fig. 3-1. Simplified schematics for electron beam evaporation of contact metals.

PAGE 60

CHAPTER 4 RESULTS Summary of Prepared Samples TEM data revealed that a 650A thick Ni film in situ annealed (300°C for 15 minutes) produced 1200 1300A of a Ni 2 . 4 GaAs layer, which will be discussed below. Samples were prepared as detailed in table 4-1 through 4-3, GaAs substrate without a specified doping concentration automatically indicates ~2 x 10**cm’^. All the samples in table 4-1 through 4-3 were as deposited i.e., before vacuum annealing at 500“C. As-deposited samples were all rectifying as will be described in the electrical measurements. Layered structures Ni2.4GaAs (A) Ge(A) Ti(A) GaAs/Ni2.4GaAs /Ge 1200 1300 250 0 1200 1300 330 0 1200 1300 500 0 GaAs/Ni2.4GaAs/It 1200 1300 0 600 GaAs/Ni2.4GaAs/Ge/It 12001300 250 300 1200 1300 330 200 GaAs( 1 0*®)/Ni2.4GaAs/Ge/Ti 1200 1300 250 300 1200 1300 500 300 12001300 750 300 1200 1300 3 250 900 53

PAGE 61

As stated in Chapter 2, Ni/GaAs reaction was reported to produce Ni-As and Ni-Ga binaries at temperatures above 400°C [OgaSO, Lav86, Gue89], Data from the present study showed that a 65oA Ni film in situ annealed at 400°C, for 15 minutes produced a combination of NixGaAs and NiAs phases. The thickness of the reaction layer was 1400 150oA from TEM measurements. Table 4-2. Samples containing 65oA Ni film in situ annealed at 400°C, 15 minutes. Layered structure (A) NixGaAs+ NiAs Ge Ti Ga As( 1 0 *")/NixGa As+Ni As/Ge/Ti 14001500 250 300 1400 1500 750 300 14001500 500 600 14001500 250 900 Samples without in situ annealing are shown in Table 4-3. Table 4-3. Samples without Ni film in situ annealed. Layered structure Ni(A) Ge(A) Ti(A) GaAs / Ni / Ge / Ti 650 250 300 650 250 600 650 250 900 Electrical Measurements All the I-V curves presented here were measured between two circular dots with 0.5 mm diameter which were the most representative of the corresponding structures.

PAGE 62

55 GaAs/Ni. Figure 4-1 is typical example of the I-V curves measured for GaAs / 650A Ni. As-deposited sample exhibited well defined rectifying I-V characteristics with breakdown voltage of -2.0V on GaAs doped with 2 x 10'*cm'^ Si. After in situ annealing at 300°C, for 15 minutes, rectifying behavior was still measured but with a barrier reduced to ~1 ,2V. Since the in situ annealing produced the Ni 2 . 4 GaAs phase, (see below), the I-V characteristics represents the contacts between this phase and GaAs, i.e., GaAs/ Ni2,4GaAs. GaAs/ Ni 2 . 4 GaAs /Ge. Figure 4-2 is a typical example of I-V data from GaAs / Ni 2 , 4 GaAs / 250A Ge samples. The data were also typical for GaAs / Ni 2 . 4 GaAs / 33oA Ge samples. As-deposited and vacuum annealed contacts, (i.e., annealed for 2.5 and 5 minutes after Ge evaporation) exhibited rectifying behavior, although the vacuum annealed contacts showed a reduced barrier. Contacts of GaAs / Ni 2 . 4 GaAs / 500A Ge exhibited a linear I-V characteristics as shown in Fig. 4-3, after vacuum annealing at 500°C for 5 minutes, indicating the transition from rectifying to ohmic was a function of Ge thickness. GaAs / Ni 2 . 4 GaAs / Ti. The I-V characteristics from as-deposited and vacuum annealed sample of GaAs / Ni 2 . 4 GaAs / 60oA Ti showed rectifying barriers (figure 4-4) similar to those from the GaAs / Ni 2 . 4 GaAs / 25oA Ge. However, the annealed contacts exhibited an increased barrier.

PAGE 63

56 GaAs / Ni2.4GaAs / Ge / Ti. Figure 4-5 shows the I-V characteristics of GaAs / Ni2.4GaAs / 250A Ge / 300A Ti and GaAs / Ni2.4GaAs / 33oA Ge / 20oA Ti. These structures exhibited ohmic behavior after 500°C for 5 and 20 minute vacuum annealing. Since the vacuum annealed structures of GaAs / Ni2.4GaAs / 25oA Ge and GaAs / Ni2.4GaAs / 33oA Ge did not show linear I-V, this indicates that 200 300A of Ti is necessary to observe ohmic behavior after vacuum annealing. GaAs / Ni / Ge / Ti. Figure 4-6 show the I-V data measured from GaAs / 65oA Ni / 25oA Ge / 30oA Ti (where the nomenclature indicates that Ni film was not in situ annealed before evaporation of Ge and Ti). The I-V characteristics before and after vacuum annealing were rectifying. Since this structure did not show linear I-V while GaAs / Ni2.4GaAs / 250A Ge / 300A Ti did, in situ annealing of 65oA Ni played a critical role in the formation of ohmic I-V characteristics. To determine the effects of Ti thickness, Ti was increased from 30oA to 600A or 900A, The I-V data in figure 4-7 shows that as-deposited and 500°C, 5 minute vacuum annealed exhibited little change in the I-V curves. A noticeable reduction in the Schottky barrier height was measured for 20 and 35 minute annealing. Figure 4-8 shows the results from GaAs / 650A Ni / 25oA Ge / 90oA Ti. There was no significant change in I-V characteristics between 5 minute and 20 minute annealing, unlike for the samples shown in figure 4 7 . As can be seen in these results, the increase of Ti thickness with a fixed thickness of Ge layer did not cause the change from rectifying to ohmic behavior.

PAGE 64

57 Samples using 10*®cm'^ doped GaAs. To investigate the effects of doping concentration, GaAs / Ni 2 . 4 GaAs / Ge / Ti structures were prepared using 2 xlO^^cm’^ doped GaAs. As-deposited sample of GaAs / Ni 2 . 4 GaAs / 25oA Ge / 300A Ti exhibited rectifying contacts with a breakdown voltage of ~25V. Note the breakdown voltage was only ~2V for GaAs doped 2 x lO'^cm’^. Because of its extremely low current level, it appears that no current flows on the scale of figure 4-9. After vacuum annealing at 500°C, 5 minutes, higher levels of current were measured although the contact still rectifying contacts versus 2 xlO^Vm'^ doped GaAs which exhibited ohmic behavior. The effects of increasing the Ge thickness from 250 to 75oA are shown in the figure 4-9. With 75oA Ge, the currents increased consistent with a reduced the barrier height. With 90oA Ti, less current was measured. All the sample on the figure 4-9 were rectifying. These results suggest that the amount of Ge is an critical factor for improving the I-V characteristics. Samples in situ annealed at 400“C. Since no structures on 2 x 10^® doped GaAs exhibited linear I-V characteristics with an in situ Ni anneal at 300°C for 15 minutes, samples with an in situ Ni anneal at 400°C for 15 minutes were studied. The phases produced by the reaction between Ni and GaAs at 400°C were NixGaAs + NiAs compounds. The I-V characteristics from the several samples are shown in figure 4-10. Linear I-V characteristics were measured with 50oA or 75oA Ge, 30oA Ti.

PAGE 65

58 Summary of electrical measurements. Linear I-V characteristics were not observed unless Ge was present, therefore, Ge was essential for ohmic contacts. In the structures comprised of 1200 -130oA Ni 2 . 4 GaAs on 2 x 10^*cm ^ GaAs, the transition from rectifying to ohmic behavior required Ge films thicker than ~50oA or 200 300A thick Ti layer with 25oA Ge. In situ annealing also helped form ohmic contacts. For metallization on the 2 x 10 cm GaAs, minimum thickness of Ge layer to form ohmic contacts was 75oA. Elemental Depth Profiles GaAs / Ni2.4GaAs / Ge. The SIMS depth profile of the GaAs / Niz 4 GaAs / 25oA Ge as-deposited and annealed are shown in figure 4-11. As pointed out in Chapter 3, the CsX signals were measured but are simply denoted as X (e.g. CsGe'' is designated as Ge). In this profile, Ge and a layer of Ni 2 . 4 GaAs (see below for composition analysis) are distinguished. In the Ni 2 . 4 GaAs layer, the Ni, Ga and As signals are all stable, consistent with formation of a ternary compound. A strong peak of carbon was always detected at the Ge/ Ni 2 . 4 GaAs interface, as shown in the figure 4-11. This carbon is thought to originate from adsorption of diffusion pump oil during the in situ anneal. The carbon layer was present before and after vacuum annealing, therefore it served as a marker layer indicating the original interface between Ge and the Ni 2 . 4 GaAs. Oxygen was also normally detected at the interface but is but not plotted in the figures (for clarity). The SIMS profile after vacuum annealing at 500°C, 5 minutes shows that interdiffiision took place. Using the carbon marker, Ge diffused into the Ni 2 , 4 GaAs region and Ni and

PAGE 66

59 Ga diffused beyond the carbon layer towards the free surface. The concentration of As is low in the outer surface region of the original Ni 2 , 4 GaAs layer. Since the Ni between the carbon marker and the surface had to be supplied from the initial Ni 2 . 4 GaAs layer, at least some portions of the Ni 2 . 4 GaAs had to decompose which would allow release of Ga and As and could result in the regrowth of GaAs. The Ni 2 . 4 GaAs decomposition and GaAs regrowth was confirmed in TEM study (see below). Figure 4-12 shows AES depth profiles of GaAs / Ni 2 . 4 GaAs / 500A Ge. The profile after the vacuum annealing at 500°C, figure 4-12 (b), looks similar to that in figure 4-11. GaAs / Ni2.4GaAs / Ti. Figure 4-13 shows the results of depth profiles of GaAs / Ni 2 . 4 GaAs / 600A Ti as deposited and vacuum annealed. In the as-deposited profile, figure 4-13 (a), the Ti signal peak at the interface between Ti and the Ni 2 . 4 GaAs probably results from matrix effects on secondary ion sputtering yield [Wil89]. The depth profile after a vacuum anneal at 500°C, 5 minutes showed severe interdiffusion of the elements, figure 4-13 (b). Based on the carbon marker layer, it is clear that Ti diffused into and reacted with Ni 2 . 4 GaAs layer. GaAs / Ni2.4GaAs / Ge / Ti. Figure 4-14 shows the SIMS depth profile of the GaAs / Ni 2 . 4 GaAs / 25oA Ge / 30oA Ti. In the as-deposited profile (figure 4-14 (a)), Ti, Ge and the Ni 2 . 4 GaAs layer are distinguished with carbon peak at the Ge/ Ni 2 . 4 GaAs interface. In figure 4-14 (a), the Ti signal was detected beyond the Ge/ Ni 2 . 4 GaAs interface, suggesting that Ti is a faster diffusing species in this metallization.

PAGE 67

60 Figure 4-14 (b) is the SIMS profile after vacuum annealing at 500 °C, for 5 minutes, with the carbon marker still evident. The metal layers reacted to form two different region at 500 “C, 5 minutes, a Ti-Ni-As layer to the left of the carbon marker and Ni-Ga-Ge layer to the right of the marker layer. Arsenic in the top layer results from depletion of As in the previously Ni2,4GaAs. The majority of Ge diffused into the bottom layer where Ni and Ga are the matrix elements. A small portion of Ti diffused beyond the carbon layer, while almost the same Ni signal was detected in the top layer beyond the marker. This suggest that Ni diffused into the Ti layer while Ti did not diffuse significantly into the Ni2.4GaAs region. Since Ni out-diffusion into the Ti layer, a portion of the Ni2.4GaAs layer was decomposed to supply the Ni in the top layer. Decomposition resulted in regrowth of GaAs from release of Ga and As, as will be seen in TEM study. In the depth profile of an annealed sample of GaAs / Ni2.4GaAs / 250 A Ge (figure 4-1 1 (b)), only a slight out-diffusion of Ni from the Ni2.4GaAs layer was observed and this structure did not form linear I-V. In the case of adding 300A Ti or an additional 25oA Ge (total 50oA Ge) onto GaAs / Ni2.4GaAs / 250A Ge, the depth profile clearly showed that a larger amount of Ni diffused into the Ti layer or Ge layer over the marker and these structures formed linear I-V. From these results, it is postulated that decomposition of Ni2.4GaAs layer was initiated or driven fiarther by the Ni-Ti reaction, resulting in the switch from rectifying to ohmic behavior. A similar structure of GaAs / Ni2.4GaAs / 600A Ti did not exhibited a linear I-V characteristics but decomposition of Ni2.4GaAs occurred by a Ni-Ti reaction similar to figure 4-13 (b). This indicates that Ge is a crucial element for the formation of ohmic contacts.

PAGE 68

61 GaAs / Ni / Ge / Ti. Figure 4-15 (a) shows the depth profile of the as deposited three metal films (Ti, Ge and Ni) on the GaAs substrate without in situ annealing. A relatively strong peak carbon was detected at the Ni/GaAs interface (not shown in the figure). Figure 4-15 (b) is the profile after vacuum annealing at 500°C, for 5 minutes. The carbon peak was detected at 6 minutes of sputtering time as shown. It is noted that the overall distributions of the elements look very similar to the figure 4-14 (b) in that the entire metal films consist of two layers, one to the left of the carbon layer and the other to the right of the carbon layer. Despite the similar chemistry of the layer to that of vacuum annealed GaAs / Ni 2 . 4 GaAs / 25oA Ge / 30oA Ti (figure 4-14 (b)), the structure of figure 4-15 (b) did not form linear I-V characteristics. Since the overall chemistry of figure 4-14 (b) and 4-15 (b) appears to be nearly the same, the difference in the electrical property probably results from changes in the microstructure. Since GaAs / 65oA Ni / 25oA Ge / 300A Ti did not form a linear I-V, a similar structure with 60oA of Ti film (rather than 300A of Ti) was prepared and vacuum annealed. The depth profile (figure 4-16) and I-V data (figure 4-7) of this sample were similar to those from the sample with 300A of Ti after vacuum annealing at 500“C, for 5 minutes. Therefore more extensive Ni-Ti reactions failed to produce ohmic contacts after vacuum annealing for 5 minutes. As annealing time increased to 20 and 35 minutes, a slight in-diffusion of Ti and outdifflision ofNi were noticed while Ge remains unchanged as shown figure 4-17 (a) and

PAGE 69

62 (b). Arsenic depletion at the interface with the top metal layer became more obvious in the figures. Since a reduction of the I-V was measured after 20 minute annealing, the slight further Ni-Ti reactions appears to be significant. With a 90oA Ti layer, similar elemental distributions were observed after annealing for 5 minutes and 20 minutes as shown in figure 4-18 (b) and 4-19 (b). In this case, there was no noticeable reduction in the I-V characteristics. GaAs samples doped lO’^cm'^. The I-V data in figure 4-9 show that GaAs (10*^) / Ni 2 . 4 GaAs / 750A Ge / 30oA Ti did not exhibit linear I-V behavior after vacuum annealing at 500°C, for 5 minutes, while the same layer structure in situ annealed at 400°C, for 15 minutes exhibited linear I-V (figure 4-10). Both samples were characterized by SIMS with depth profile from (10*^) / Ni 2 . 4 GaAs / 75oA Ge / 300A Ti shown in figure 4-20, while those from GaAs (10‘") / Ni^GaAs + NiAs / 75oA Ge / 300A Ti are shown figure 421. Note that Ni 2 , 4 GaAs is the nomenclature for the reaction products by in situ annealing of 65oA Ni film at 300“C for 15 minutes and NixGaAs + NiAs for the reaction products by in situ annealing of 650A Ni film at 400‘’C for 15 minutes. The depth profile in figure 4-21 (b) is very similar to that shown in figure 4-20 (b), but only the in situ annealed at 400°C, for 1 5 minutes resulted in ohmic contacts after vacuum annealing at 500°C for 5 minutes. Thus, the elemental depth profiles are not sufficient to account for the differences in the electrical behaviors of the two structures.

PAGE 70

Transmission Electron Microscopy 63 Several structures were analyzed by TEM based on the results from the electrical measurements and the SIMS and AES depth profiles, Bright-field images, diffraction pattern and EDS spectra will be described. All the bright-field images were taken along the <1 10> zone axis of GaAs, unless otherwise specified. GaAs / Ni2.4GaAs / Ge. Figure 4-22 (a) is a bright-field image of a GaAs / Ni 2 . 4 GaAs / 25oA Ge as-deposited cross-section, where the Ni 2 . 4 GaAs designates the reaction product by the in situ annealing of 65oA Ni film at 300“C for 15 minutes. In this particular brightfield image, the GaAs substrate was ion milled faster than the metal layers, showing a large ion milling hole. The diffraction pattern from the entire structure is shown in figure 4-22 (b) and a schematic of the diffraction pattern is shown in (c). The Ni 2 . 4 GaAs phase was identified as a NiAs-type hexagonal ternary by its diffraction pattern, which was identical to the pattern of figure 2 in the reference by T. Sand et al [San86]. The diffraction pattern in figure 4-22 (b) was taken along the <1 10> zone axis of GaAs, so the {111}, {220} and {200} spots from GaAs are clearly visible (see figure 422(c)). The four small satellite peaks around each { 1 1 1 }oaA8 spot indicate that the ternary phase is twinned. The streaky diffraction spots from the ternary phase suggests that it is under some strain to accommodate twinned sections differently oriented to each other. In fact, the bright-field image in figure 4-22 (a) is a twinned image of Ni 2 . 4 GaAs. The brightfield image shows microtwinned sections with an average size of -100 A or less. The

PAGE 71

microtwins have the shape of elongated rhomboids parallel to each other and perpendicular to the GaAs interface. The contrast in the Ni 2 . 4 GaAs layer arises from diffraction, not compositional contrast. The Ge layer on the Ni 2 . 4 GaAs layer was ~25oA thick. The morphology of the Ge layer is an island structure rather than a smooth layer (figure 4-22 (a)). The results of EDS analysis on the Ni 2 . 4 GaAs of figure 4-22 (a) is shown figure 4-23. K a lines of Ni, Ga, and As are displayed respectively in the spectrum. The Cu peak originated from the Cu rings used for the TEM sample support. This spectrum represents a typical composition from the Ni 2 . 4 GaAs phase. The intensity ratio of Ga and As signal in figure 423 is similar to that from the GaAs substrate in figure 4-24, supporting the 1 : 1 ratio of Ga and As in the ternary phase. The as-deposited GaAs / Ni 2 . 4 GaAs / 25oA Ge sample was vacuum annealed at 500°C, for 2.5 minutes and figure 4-25 shows the results of this annealing. In the bright field image (figure 4-25 (a)), the entire metal structure consists of two layers. The first layer is only -200 A thick and the second layer is the ternary Ni 2 . 4 GaAs. At the boundary of these two layers, a bright, sparsely discontinuous line is observed. This line is thought to be the carbon marker produced during the in situ annealing and the following vacuum annealing. This carbon marker was consistently observed in all the structures in situ annealed. The identification of the carbon marker was more obvious when the sample was annealed for an additional 2.5 minutes (5 minutes total). By comparing the elemental depth profiles (figure 4-11) where the carbon divides the entire metal layer into two regions, its location

PAGE 72

65 agrees with the TEM result and again played the role of marker between the outer metal layer and the Ni 2 . 4 GaAs layer. There are clearly noticeable changes between the two bright field images of Ni 2 . 4 GaAs layer before (figure 4-22(a)) and after the 2.5 minute annealing (figure 4-25 (a)). The first difference would be the disappearance of the microsized twinned sections in annealed samples. Instead, the internal structure of the Ni 2 . 4 GaAs in the figure 4-25 (a) shows equiaxial grains. One of strongly diffracting (0ll2) type spots (indicated by an pointer in figure 4-25 (b)) was selected to form a dark field image and the dark field image is shown in figure 4-25 (c). This image clearly shows the formation of equiaxial grains with a diameter of ~400A or larger (the grains with white contrast). These data indicate that the microsize twinned sections coalescenced into a lager, equiaxial grains. The second difference of annealing is that the interfacial morphology between Ni 2 . 4 GaAs and GaAs becomes non-planar upon annealing. The thickness of the Ni 2 . 4 GaAs layer was ~125oA before and ~1000A after annealing, indicating that an average of ~25oA of Ni 2 . 4 GaAs was decomposed. Simultaneously, another phase began to form inside the ternary phase at the interface with GaAs. This phase will be shown to be NiAs containing Ga and Ge as minor elements and one of the grains is marked by arrows in figure 4-25(a) and figure 4-25 (c). Their average size was ~50oA. Note that the I-V characteristics of this sample was rectifying, similar to the as-deposited condition. This same structure was annealed for an additional 2.5 minute at 500“C (5 minutes total) and figure 4-26 (a) shows a bright-field image of cross-section. The NiAs phase grew to ~150oA parallel to the interface and -lOOoA vertically, while the average

PAGE 73

thickness of the Ni 2 . 4 GaAs layer remained almost the same as for 2.5 minutes annealing. A large size of NiAs grain is indicated by an arrow in figure 4-26(a). In some grains of NiAs phase, TEM showed that microcrystals existed, probably with different orientations. Figure 4-26 (b) is a high resolution images taken in the sample shown in figure 4-26 (a). Note that the outer metal layer and top portion of the bottom layer (previously Ni 2 . 4 GaAs) were ion milled away in this area. The micrograph in figure 4-26 (b) shows the Moire fringes near the boundary between the NiAs and surrounding metal layer, suggesting that they are two dissimilar phases with different ^/-spacings, and are overlapped in this cross section. To determine whether the NiAs was a new or simply a recrystallized, the structure of figure 4-26 (a) was analyzed by EDS and the results are shown in figure 4-27. Figure 427 (a) illustrates schematically the points where the EDS data were collected, and (b) through (f) are the spectra. Location (b) is an EDS spectrum from the GaAs substrate, similar to data in figure 4-24. At the interface between the Nii.sAs and the GaAs, the Ga Ka signal decreased and the Ni Ka increased (figure 4-27 (c)). At point (d) and (e), inside the NiAs, the Ni signal was stronger, the Ga signal further decreased and a weak peak from Ge signal was detected as shown in figure 4-27 (d). The Ge signal was slightly stronger at position (e) versus position (d). Position (f) was at the same lateral location as (e) but outside the NiAs phase and the spectrum is the same as that fi'om Ni 2 . 4 GaAs, as shown by comparing figure 4-27 (f) with figure 4-23 . This EDS results confirmed the formation of NiAs phase with minor concentration of Ge and Ga.

PAGE 74

67 To better understand the EDS results, some characteristic features of the NiAs-type unit cells were considered as shown in figure 4-28. The unit cell of NiAs is hexagonal, Ni atoms occupy (000, OO'/a) sites, and As atoms occupy (1/3 2/3 1/4, 2/3 1/3 3/4) sites while two additional sites (2/3 1/3 1/4, 1/3 2/3 3/4) are vacant, as shown in figure 4-28 (a). When the two vacant sites are half occupied by Ni atoms, it becomes Ni 3 Ga 2 (figure 4-28 (b)). If the two vacant sites are fully occupied by Ni atoms, it becomes Ni 2 ln known as “filled-up” NiAs-type structure (figure 4-28 (c)) [Gue89], Since the covalent radii of Ga atoms (1.25 A) is bigger than that of As atom (1.21 A), the unit cell of NiAs structure expand when Ga atoms substitutes for As atoms, thereby increasing the concentration of Ni atoms in the unit cells. This is how the transition from NiAs to Ni 3 Ga 2 occurs. From these structural features, it was easily shown that the NiAs phase identified by our EDS results was formed from Ni 2 . 4 GaAs by substituting the Ga atoms with As and Ge atoms, as will be obvious. I-V characteristics of this structure exhibited a noticeably reduced barrier, with breakdown voltage being reduced from -1.5V to 0.8V in figure 4-2. Figure 4-29 (a) shows an as-deposited structure of GaAs / Ni 2 . 4 GaAs / 500A Ge. Below the Ge layer, ~8oA of a reacted layer (figure 4-29 (a)) is visible, suggesting Ge reacts with Ni 2 . 4 GaAs at temperatures as low as lOOX. Since interdifiusion ofNi and Ge over the carbon layer is clearly seen in the AES depth profile of the structure, figure (4-1 1 (a)), this reaction layer is believed to consist of Ni and Ge. Below this reaction layer, Ni 2 . 4 GaAs with a sharp interface with GaAs is seen. Figure 4-29 (b) show the entire metal layer and GaAs substrate after vacuum annealing of the sample in figure 4-29 (a) at 500X for 5 minutes. About ~30oA of Ni 2 . 4 GaAs was

PAGE 75

68 decomposed. Coalescence of the microsize twinned sections was again observed and evolution of phases was similar to the GaAs / Ni2.4GaAs / 25oA Ge sample. A noticeable difference between 500A Ge and 250A Ge was the extremely sharp interface between the contact metal layer and GaAs resulting from the disappearance of protrusion of NiAs phase towards GaAs. The grains of NiAs phase can be noticed by their dark/bright contrast. The I-V data from this structure was linear as shown in figure 4-3. GaAs / Ni2.4GaAs / Ge / Ti. All the results from cross-sectional TEM analysis and elemental depth profile showed that both of GaAs / Ni2.4GaAs / 250A Ge / 300A Ti and GaAs / Ni2.4GaAs / 33oA Ge / 200A Ti had the same metallurgical phases and linear I-V characteristics (figure 4-5). Figure 4-30 shows a cross-sectional image of as-deposited GaAs / Ni2.4GaAs / 33oA Ge / 200 A Ti. In the bright-field image, the Ge layer was amorphous. In the corresponding diffraction pattern (not shown) a weak ring pattern with i/-spacing of ~2.5A, in addition to spots from GaAs and Ni2.4GaAs, was observed from the randomly oriented polycrystalline (100) Ti grains. The Niz4GaAs layer was 1200A 1300A thick, as usual. The results of a high resolution EDS analysis on the Ni2.4GaAs phase are shown in figure 4-3 1(a) while figure 4-3 1 (b) shows schematically which points were analyzed. It should be noted that oA on the x-axis in figure 4-3 1 (a) indicates the interface between a protrusion of Ni2.4GaAs into the GaAs, and the distances are relative to this point. In figure 4-30, several protrusion are visible and one of them is indicated by an arrow. The

PAGE 76

69 distances are approximate values. EDS spectra were collected at the distances along this line shown in table 4-4. The concentrations in table 4-4 were determined from raw EDS spectra as described in Chapter 3. Table 4-4. Concentration of elements plotted in figure 4-3 1 (a) Distance(A) Ni (at %) Ga (at %) As (at %) Ni/Ga Ni/As 10 1.7 49.9 48.2 0.0 0.0 50 7.3 48.0 44.5 0.2 0.2 100 55.0 23.2 21.9 2.4 2.5 200 55.8 23.2 20.9 2.4 2.7 300 56.2 24.2 19.1 2.3 2.9 400 56.5 24.2 19.1 2.3 2.9 The first two data points at 10 and 50 A showed dramatically different values from the expected stoichiometry of Ni 2 . 4 GaAs. Their values were almost the stoichiometry of GaAs. It is probable that those two points were from an area where a thin spike of Ni 2 . 4 GaAs protruded into the GaAs substrate. This kind of protrusion can also be seen in figure 1 of the reference by Sands et al [San87]. The EDS data collected outside of the protrusion revealed the composition of Ni 2 . 4 GaAs (see table 4-4). The Ni/As ratios were higher than those of Ni/Ga. The average concentrations of each element in as-deposited Ni 2 . 4 GaAs were determined as Ni 2 . 75 Ga 1 .i 6 As from table 4-4 and summarized in table 4-4, which has been and will continue to repeat as Ni 2 . 4 GaAs. This result agrees with previous EDS data showing the composition of NkGaAs was slightly Ga-rich [San87].

PAGE 77

70 Table 4-5, Composition of Nb^GaAs as-deposited Ni Ga Ge As Atomic % 55.9 23.7 0 20.3 Figure 4-32 is a cross-sectional bright field image after 500°C for 2.5 minute. The two layers structure separated by the carbon marker reported in figure 4-26 to 4-29 is clearly evident. The top metal layer was ~55oA thick and the bottom layer in contact with the GaAs substrate was ~1 lOoA thick. As a result of Ge in-diffusion, the bottom layer contained Ge. Internal structure in the top metal layer was revealed by the bright field image in figure 4-32. Some precipitates have formed in the matrix which consists mostly of Ti and Ni, based on the corresponding depth profiles. To investigate the top metal layer, several diffraction patterns were taken along the <1 10> and <1 12> zone axis of GaAs from a sample annealed for 5 minutes at 500“C. Only GaAs and NiAs-structure type diffraction spots were detected along the <1 10> axis, but a strong extra spot was found along the <1 12> zone axis. The csf-spacing of this spot was ~2.3 1 A and it corresponded to the (002) planes of NiTi. Thus, the matrix of the top metal layer consists of polycrystalline NiTi grains. Inside the bottom metal layer near the interface GaAs, several bright grains were distinguished from their dark surrounding as indicated by arrows in figure 4-32. These grains are thought to be NiAs phase containing Ga and Ge similar to those in figure 4-26 (a). This phase was also detected in samples annealed for an additional 2.5 minutes at 500°C. Decomposition of Ni 2 . 4 GaAs and formation of NiAs phases were more clearly demonstrated in the cross-section bright field images from GaAs / Ni 2 . 4 GaAs / 250 A Ge /

PAGE 78

71 300A Ti samples annealed for 5 minutes at 500°C (see figure 4-33). The top metal layer contained Ni, Ti and As, and was ~600A thick, while the bottom (previously Ni 2 . 4 GaAs) layer was -900 A thick. On average -3 00 A of GaAs regrew as a result of the decomposition of the Ni 2 . 4 GaAs and NiAs phases. Because the interface with GaAs is now uniform (figure 4-33), versus undulating (figure 4-26), the NiAs phase was under decomposition along with Ni 2 . 4 GaAs. Undecomposed NiAs phase grains (bright contrast, indicated by arrows in figure 4-33) located near the interface with the GaAs can be distinguished from the surrounding grains by bright/dark contrast. The composition of regrown GaAs resulting from the decomposition of NiAs phase and Ni 2 . 4 GaAs were investigated by using high resolution EDS. Figure 4-34 (a) and table 4-5 shows the atomic concentration versus distance from the interface between the NiAs phase and regrown GaAs. The interface is at “0” distance with the GaAs substrate being at negative distance. As indicated above, ~300A of Ni 2 . 4 GaAs was decomposed resulting in the regrowth of GaAs, therefore the location at -30oA would be the approximate location of the original interface between GaAs and Ni 2 . 4 GaAs before any decomposition. Table 4-6. Atomic concentrations of elements in figure 4-34 (a) Distance(A) Ni (at %) Ga (at %) Ge (at. %) As (at %) -300 0.2 48.5 0.0 51.2 -200 0.2 52.3 0.1 47.1 -100 0.4 52.9 0.2 46.1 -50 3.0 51.5 0.0 45.0 -25 7.4 47.2 0.5 44.5 0 27.0 29.3 2.2 41.2 50 41.7 16.6 3.8 37.6 100 48.1 12.8 3.5 35.3 300 49.8 11.8 4.5 33.5

PAGE 79

72 The exact composition of NiAs was determined to be Ni 1 . 3 Gao. 4 Geo. 1 As (table 4-7). Using the data acquired at lOoA into the Nii.sAs grain were used. The average vertical size of Nii.sAs grains was ~400A or less. Two sets of data was collected at lOoA, then averaged to yield the following composition: Table 4-7. Composition of Nii.sAs (Nii. 3 AsGao. 4 Geo.i) Ni Ga Ge As Atomic % 47.3 13.0 4.0 35.7 This composition confirmed that the result of EDS analysis shown in figure 4-27. It is noted that the As concentration actually increased from 20.25% to 35.63% from the data in table 4-4. This results shows that formation of the Nii.sAs phase involved redistribution of the major three elements, Ni, Ga and As. Figure 4-34 (b) and table 4-6 shows the results across the interface between iGa and regrown GaAs. Table 4-8 Atomic concentrations of elements in figure 4-32 (b) Distance(A) Ni (at %) Ga (at %) Ge (at. %) As (at %) -300 0.2 52.8 0.0 46.6 -200 0.3 52.6 0.0 48.8 -100 2.6 52.0 0.2 44.9 -50 2.2 52.0 0.2 45.0 50 53.5 27.4 11.8 6.9 100 55.1 26.5 11.8 6.2 300 55.2 26.6 11.4 6.4

PAGE 80

It can be noticed that the overall composition is very uniform over 300A of distance, from 5oA to 30oA. By averaging the data taken at 100 300A, the composition of Ni^Ga formed from Ni 2 , 4 GaAs after annealing at 500°C, 5 minutes was determined to be Ni2.oGaGeo.4Aso.2 and listed in table 4-7. Ni Ga ja vi>i2.lvra\jeo.4-fVSo.2; Ge As Atomic % 55.2 26.6 11.62 6.3 By comparing the composition of Ni 2 . 4 GaAs as deposited (table 4-4) to that of the table 47, the As concentration significantly decreased from 20.3 % to 6.3 % while the Ga concentration slightly increased (from 23.7% to 26.6%). The Ni concentration was not changed and -11% of Ge was incorporated. The result of table 4-7 suggests that Ni 2 . 4 GaAs transformed into Nii.sAs phase containing Ga and Ge , and Ni 2 .iGa phase containing Ge and As, as a result of the annealing at 500°C, for 5 minutes. As shown in table 4-5, low concentration of Ge were detected in the regrown GaAs. Because the size of probe was IsA and the measurements was taken only 25 A away from the interface, there is a chance of secondary x-ray fluorescence from Ge in the Nii.sAs grain because of the proximity to the probe beam. Also, the Ge concentrations were near the detection limit for EDS analysis. To verify the existence of Ge in the regrown GaAs, several EDS spectra were measured from -300A to -5oA in regrown GaAs where the beam broadening can be excluded. These spectra were summed to increase the ratio of Ge signal to the

PAGE 81

74 background nose and the regrown GaAs is shown in table 4-8 to contain 0.34 % Ge, which is -1.5 X 10^°cm‘^. T able 4-10. Composition of Regrown GaAs calculated from summation spe ctra Ni Ga Ge As Atomic % 2.6 46.2 0.34 50.7 The accuracy of this procedure was checked by calculating the Minimum Mass Fraction (MMF) for Ge, which indicates the statistical limit of detectability i.e. the statistical error range for Ge detection under these conditions. MMF for Ge = 3 {(Ge in wt. %) / (counts of Ge signal)} where P is background noise in counts (count per seconds x total acquisition time). All the data from the summation spectra were plugged into the above equation and yielded 0.08 % as the MMF for Ge. Since the detected Ge concentration (0.34 at %) is four times larger than the error, the detected Ge is a true signal. The Ge concentration within -5oA of the interface was also studied with some difficulty. Because the probing point was so close to the interface, it was likely that the probing beam illuminated the metal layer containing Ge. For these reasons, EDS analysis within -50 A range were carried out with the ED AX 9100 system in the image mode with 1 or 2 X 10^ magnification. The locations of the probed points were confirmed by observing diffraction patterns as described in Chapter 3. During the acquisition of EDS spectra, the acquisition was stopped and checked every 1 5 second during the total of 1 00 seconds to guarantee that the probing location was constant.

PAGE 82

On average, 0.5 1,3 atomic % of Ge was detected, and this corresponds to a doping concentration of 2.2 5.8 x 10^” cm'^. It is not clear to what extent the beam broadening effect contributed to these values. However, since 0.34 at. % (1.5 x 10^** cm'^) of Ge was measured in regrown GaAs at > 5oA, it is reasonable to conclude that the maximum Ge concentration incorporated during the regrowth does not exceed ~5.0 x 10^°cm^ As an efforts to further investigate the Nii.sAs phase, several diffraction patterns were taken from the samples annealed for 5 minutes at 500®C. Figure 4-35 (a) is a dark field image formed from the diffraction spot indicated by the arrow in figure 4-35 (b), which is the diffraction pattern from figure 4-35 (a). The c/-spacing of the diffraction spot used to form the image of figure 4-35 (a) was 1.95 2.oA. Figure 4-35 (c) is the corresponding bright-field image of figure 4-35 (a) and the same grain is also indicated by an arrow in the figure. Figure 4-35 (d) shows diffraction patterns simulated using the lattice parameters of NiAs and NisGa 2 along [110] zone axis of GaAs (GaAs: a« = 5.64, NiAs; ao = 3.61, c„ = 5.03A, and Ni3Ga2: a„ = 4.0, Co = 4.98A ) [San86, Che88]. In figure 4-35 (d), three different phases contributed to the entire patterns: GaAs (•), NiAs (x), and NisGa 2 (). Figure 4-35 (e) is a diffraction pattern taken along along [110] zone axis of GaAs, where Ni,.3As (a« = 3.78, Co = 5. 053 A) and Ni^iGa (ao = 3.95, Co = 4.98A) were identified. By comparing (d) and (e), it was found that the diffraction spot indicated by an arrow in figure 4-35 (b) and (e) was the same spot indicated in the simulated pattern (d). This diffraction spot corresponded to (01l2) plane of the NiAs phase with J-spacing of 1.98 A. This analysis provided confirming data that Ni 2 . 4 GaAs tended to transform into NiAs and

PAGE 83

76 Ni 3 Ga 2 upon annealing at 500°C for 5 minutes, indeed the NiAs phase containing Ga and Ge. and the Nii. 3 As detected in this study was Figure 4-36 is a high resolution image taken from a sample after the vacuum annealing at 500 C for 5 minutes in an area such as shown figure 4-35. The dashed line in the figure indicates the original GaAs/Ni 2 . 4 GaAs interface before regrowth. In the regrown GaAs, stacking faults and precipitates (see arrow) were observed. Along the interface. Moire fringe and contrast are evident, which are believed to be the trace of the decomposed Nil 3 As phase. The dark/bright contrast in the regrown GaAs is believed to be strain contrast caused by accommodation the defects. One of the precipitates was identified and analyzed by high resolution EDS. The composition of the precipitate was Ni-9 at. %, Ga47 % and As44 %. The regrown GaAs also contained 2.6 at. % Ni (table 4-8). In figure 4-35 (a), not only the Nii^As grains but also precipitates in the regrown GaAs were bright when the diffraction spot was used to form the dark field image. They were typically smaller than 100 A. Therefore, some of the precipitate were Nii. 3 As binaries. In the metal layer, bright grains (marked A in figure 4-36) were Nii. 3 As and the dark contrast area was Ni 2 .iGa(Ge,As) region. GaAs / Ni / Ge / Ti. Figure 4-37 shows a bright field image of an as-deposited GaAs / 65oA Ni / 25oA Ge / 300A Ti. Note that this is a structure without in situ annealing. As shown, the Ni film has columnar structure and diffraction analysis revealed that the Ni film was polycrystalline (111) oriented grains. The Ge layer was amorphous and the Ti layer developed small polycrystalline (100) oriented grains. Between the Ni and Ge layers, an

PAGE 84

extra layer with a thickness of ~15oA was detected. In this bright-field image, interfaces between metals were very smooth and the corresponding depth profile, figure 4-15 (a) showed interdiffusion between the Ge and Ni layers. Thus, this extra layer appears to be a reaction product of Ni and Ge formed during deposition or the TEM sample preparation at 110“C. Figure 4-37 (b) shows the annealed structure of figure 4-37 (a) and the entire contact metal consisted of two layers. The top metal layer was about 650A thick and the bottom layer was 750 85oA thick. Between the two layers, a sparse discontinuous bright line was still visible. In the depth profile of this structure (figure 4-15 (b)) a strong peak of carbon was detected between two layers of constant composition. From the corresponding SIMS depth profile of chemical composition in figure 4-15 (b) and the diffraction pattern from the entire structure of figure 4-37 (b), it is believed that the bottom layer is similar to the bottom metal layer of the annealed GaAs / Ni 2 . 4 GaAs / 25oA Ge / 30oA Ti shown in figure 4-35, i.e. a mixture of ternary NiGaGe(As) and NiAs phases. In the bright field image (figure 4-37 (b)), bright grains near the interface with GaAs are distinguishable from the dark surrounding. In our experiment, NiAs grains could be quickly identified by EDS spectra by an As signal is much stronger than that from Ga, such as the spectra (e) of figure 4-27. Our EDS investigation on the structure shown in figure 4-37 (b) revealed that these bright grains were the NiAs phase (see arrow in figure 4-37). However, EDS spectra from the grains in figure 4-37 (b) were not so conclusive as those in figure 4-27 (e), i.e. the As signal was only slightly stronger than those from Ga, suggesting that the grains were in an early stage of the NiAs phase. This

PAGE 85

was confirmed from our diffraction pattern analysis which did not record diffraction spots of NiAs phase. Also their density was lower and their size was smaller. One crystal is indicated by an arrow. GaAs doped 10‘® cm'^. It was reported above that GaAs (10*®) / Ni 2 . 4 GaAs / 75oA Ge / 3 00 A Ti did not form linear I-V after vacuum annealing at 500“C, for 5 minutes, while GaAs (10 ) / NixGaAs+NiAs / 750A Ge / 300A Ti formed linear I-V after vacuum annealing at 500“C, for 5 minutes. Recall that NixGaAs + NiAs represents the reaction products formed by in situ annealing of 65oA Ni film at 400°C, for 15 minutes. Even though the two structures exhibited dramatically different I-V characteristics, the depth profiles did not provide good clues for the difference in the I-V behavior. Thus, these two structures were analyzed by TEM, Figure 4-38 shows a bright field image of GaAs (10*®) / Ni 2 . 4 GaAs / 75oA Ge / 30oA Ti after annealing at 500°C, 5 minutes, and three layers can be distinguished. The first layer was -600 A thick (marked A in figure 4-38), the second layer was uniform and ~70oA thick (marked B), while the third 400A thick layer was not a well defined layer but appears to be a region where various sizes of fragments of metal phase were scattered in a matrix of GaAs. Twins developed along the { 1 1 1 } planes of GaAs were visible along with the metal fragments. This sample exhibited strongly rectifying I-V characteristics. The results of electron diffraction analysis are tabulated in table 4-11. Several diffraction patterns were taken along <1 10> and <1 1 1> zone axis of GaAs. The major reaction products were Ni-rich Ni-Ge compounds (A region in figure 4-38) and Ni-Ga-Ge

PAGE 86

79 ternary phase (B region in figure 4-38). The corresponding depth profile from this structure (figure 4-20(b)) supports this conclusion. The reaction product formed by the in situ annealing of 65oA Ni films at 400°C, 15 minutes were also analyzed by diffraction along the <1 10> zone axis of GaAs. Table 4-11. Summary of diffraction patterns recorded and analyzed to identify reaction products shown in fi^re 4-38. GaAs (10 ) / Ni 2 . 4 GaAs / 75oA Ge / 300A Ti after annealing at 500°C, 5 minutes. Diffraction type Measured d-spacing ( A ) Possible phase , c/-spacing (A), Spot (strong) 2.85-2.89 iiiiciisuy , NisGes inaex yn 2.863 m) 80 (221) Ni4GaGe2 2.807 80 (203) Ni2GaGe 2.807 50 (130) Spot (strong) 2.390 NisGes 2.307 20 (222) Ni2Ge 2.345 60 (112) Ni2GaGe 2.339 40 (043) Spot (strong) 1.977 NiGe 1.997 100 (121) NiGe 2.048 80 (211) NisGes 1.951 95 (602) NisGes 2.014 100 (313) NisGes 2.022 95 (203) NisGes 2.102 60 (403) Ni5Ge2 2.009 100 (115) Ni2Ge 2.042 100 (211) Ni2Ge 1.919 85 (020) Ni4GaGe2 1.957 100 (220) Ni4GaGe2 2.012 100 (206) Ni2GaGe 2.016 60 (060) Spot (strong) 1.699 NiGe 1.687 70 (301) Ni2Ge 1.661 20 (301) Spot (strong) 2.42 Ni4GaGe2 2.424 70 (212) with ring (weak) NuGaGe2 2.498 70 o o Ring (weak) 1.43 Ni2Ge 1.442 20 (123) is relative the strongest diffraction peak designated as 100 in the JCPDS file.

PAGE 87

Diffraction spots from NiAs phase were detected and the measurements yielded a« ~3.69A and Co -5.06 A. A weak ring pattern was overlapped on (0001) type spot from NiAs phase, suggesting that the NiAs phase consisted of several grains. The corresponding SIMS depth profile (figure 4-21 (a)) showed that the intensity of Ga signal was slightly lower near the GaAs substrate, supporting the formation of NiAs phase near the GaAs substrate. A different set of diffraction spots indicated formation of another NiAs-type phase with a« ~4.0lA and c„ -4. 99 A, which is believed to be NixGaAs from the SIMS depth profile in figure 4-2 1(a). Considering these results, the reaction products formed by the in situ annealing of 65 oA of Ni films at 400°C, 15 minutes is a mixture of NixGaAs and NiAs compounds. Our results agrees with the report by Lahav et al. where a Ni 2 -xGaAsi.x matrix with NiAs precipitates were observed as a result of annealing at 350 550°C [Lav86]. Figure 4-39 is a bright-field cross section image of the structure of GaAs (10*®) / NixGaAs + NiAs / 750A Ge / 30oA Ti after annealing at 500°C for 5 minutes. Again, the entire metal layer consisted of three layers. The top 400A layer and the second 40oA. layer showed rough interface between them, while the 500 700A third metal layer had a rough interface with GaAs. The GaAs near the interface with the third metal layer contained twins which extended ~500A into GaAs, and associated strain contrast was clearly visible. The important difference with the image of figure 4-39 versus 4-38 is that there are not as many as precipitates in the GaAs near the interface. This structure exhibited ohmic behavior and the differences in the microstructure is thought to be the reason.

PAGE 88

81 Analysis on diffraction patterns taken along <1 10> zone axis from this structure revealed that NiGe and Ni 2 Ge were reaction products possibly with Ni-Ga-Ge ternaries. Therefore, 1400 -ISOOA thick NixGaAs + NiAs layer was converted into Ni-Ge compounds by reaction with 75oA Ge and 300A Ti layers.

PAGE 89

0.02 o As-deposited 300°C, 15 minutes 0.01 < S 0.00 o 0.01 0.02 — -4.0 -2.0 0.0 2.0 4.0 Voltage (V) Figure 4-1. Current-voltage characteristics for GaAs/650A Ni as deposited and in situ annealed at 300°C for 15 minutes in vacuum.

PAGE 90

83 o As-deposited Figure 4-2. Current-voltage characteristics for GaAs/Ni 2 4GaAs/250A Ge both as deposited and ex situ vacuum annealed at 500°C for 5 minutes.

PAGE 91

84 o As-deposited Figure 4-3. Current-voltage characteristics for GaAs/Ni 2 4GaAs/500A Ge both as deposited and ex situ vacuum annealed at 500°C for 5 minutes.

PAGE 92

85 o As-deposited Figure 4-4. Current-voltage characteristics for GaAs/Ni 2 4GaAs/500A Ti both as deposited and ex situ vacuum annealed at 500°C for 5 minutes.

PAGE 93

86 o As-deposited • 500°C, 5 minutes Figure 4-5. Currentvoltage characteristics for GaAs/Ni 2 4GaAs/250A Ge/300A Ti both as deposited and ex situ vacuum annealed at 500°C for 5 minutes.

PAGE 94

87 o As-deposited Figure 4-6. Current-voltage characteristics for GaAs/650A Ni/250A Ge/300A Ti both as deposited and ex situ vacuum annealed at 500°C for 5 minutes.

PAGE 95

88 O As-deposited 500°C, 5 minutes Figure 4-7. Currentvoltage characteristics for GaAs/650A Ni/250A Ge/600A Ti both as deposited and ex situ vacuum annealed at 500°C for 5 minutes.

PAGE 96

89 o As-deposited • 500°C, 5 minutes Figure 4-8. Current-voltage characteristics for GaAs/650A Ni/250A Ge/900A Ti both as deposited and ex situ vacuum annealed at 500°C for 5 minutes.

PAGE 97

90 • As-deposited 0 GaAs (10^®) / Nij ^GaAs / 250A Ge / 300ATi 16 0 GaAs (10 )/Ni 2 4GaAs/75oAGe/300ATi O GaAs (10^®) / Ni 2 4GaAs / 250A Ge / 900Aji Voltage (V) Figure 4-9. Current-voltage characteristics after ejc situ vacuum annealing of various thicknesses of Ge and Ti at 500°C for 5 minutes. The GaAs substrate doped to 2 X lO^^cm'^ and a 650A Ni layer was in situ annealed at 300 C for 15 minutes to produce Ni 2 4 GaAs layer.

PAGE 98

91 O GaAs (10^®)/ Ni^GaAs + NiAs / 250 AGe / 300A Ti H GaAs (1 0^®) / Ni^GaAs + NiAs / 500 AGe / 300A Ti <> GaAs (10^®)/ Ni^GaAs + NiAs / 750AGe / 300A Ti V GaAs (10^®)/ Ni^GaAs + NiAs / 250 AGe / 900A Ti Voltage (V) Figure 4-10. Current-voltage characteristics after ex situ vacuum annealing of various thicknesses of Ge and Ti at 500°C for 5 minutes. The GaAs substrate doped to 2 X lO'^cm ^ and a 650A Ni layer was in situ annealed at 400°C for 15 minutes to produce Ni GaAs plus NiAs layer.

PAGE 99

92 1 e+5 (a) As-deposited f -vT' 1e+4 n ^ A 'in = ’• As Sle+3 1 cA 1' ' * '(/) \ NiV. S -I 1: \ 1e+1 -= Ji ^ . /I •; : /; . 1e+0 ' 1 ' ' ^ '1 1 lO 1 1 • i" 1 V r ^ ' 0 5 10 15 20 sputter time (min) Figure 4-11. SIMS deoth profile for GaAs/Ni 2 4GaAs/250A Ge. (a) as deposited (b) ex situ vacuum annealed at 500°C for 5 minutes.

PAGE 100

Peak Height Peak Height 93 Figure 4-12. AES depth profile for GaAs/Ni 2 4GaAs/500A Ge. (a) as deposited (b) ex situ vacuum annealed at 500°C for 5 minutes.

PAGE 101

Intensity (cps) Intensit 94 1e+5 1e+4 1e+3 1e+2 1e+1 1e+0 0 5 10 15 20 25 30 Sputter time (min) Figure 4-13. SIMS deoth profile for GaAs/Ni 2 4GaAs/600A Ti. (a) as deposited (b) ex situ vacuum annealed at 500°C for 5 minutes.

PAGE 102

Intensity (cps) Intensity (cps) 95 Figure 4-14. SIMS depth profile for GaAs/Ni 2 4GaAs/250A Ge/300A Ti. (a) as deposited (b) ex situ vacuum annealed at 500°C for 5 minutes.

PAGE 103

Intensity (cps) Intensity (cps) 96 Figure 4-15. SIMS depth profile for GaAs/650A Ni/250A Ge/300A Ti. (a) as deposited (b) ex situ vacuum annealed at 500°C for 5 minutes.

PAGE 104

Intensity (cps) Intensity (cps) 97 Figure 4-16. SIMS depth profile for GaAs/650A Ni/250A Ge/600A Ti. (a) as deposited (b) ex situ vacuum annealed at 500°C for 5 minutes.

PAGE 105

Intensity (cps) Intensity (cps) Figure 4-17. SIMS depth profile for GaAs/650A Ni/250A Ge/600A Ti. (a) ex situ vacuum annealed at 500°C for 20 minutes (b) ex situ vacuum annealed at 500°C for 35 minutes

PAGE 106

Intensity (cps) Intensity (q3s) 99 1e+5 1e+4 1e+3 1e+2 1e+1 1e+0 0 5 10 15 20 25 30 Sputter time (min) 1e+5 1e+4 1e+3 1e+2 1e+1 1e+0 0 5 10 15 20 25 30 Sputter time (min) Figure 4-18. SIMS depth profile for GaAs/650A Ni/250A Ge/900A Ti. (a) as deposited (b) ex situ vacuum annealed at 500°C for 5 minutes.

PAGE 107

Intensity (cps) Intensity (cps) 100 Sputter time(min) Figure 4-19. SIMS depth profile for GaAs/650A Ni/250A Ge/900A Ti. (a) ex situ vacuum annealed at 500°C for 5 minutes. (b) ex situ vacuum annealed at 500°C for 20 minutes.

PAGE 108

Intensity (cps) Intensity (q)s) 101 Figure 4-20. SIMS depth profile for GaAs( 10 i 6 )/Ni 2 4GaAs/750A Ge/300A Ti. (a) as deposited (b) ex situ vacuum annealed at 500°C for 5 minutes.

PAGE 109

Intensity (cps) Intensity (cps) 102 Figure 4-21. SIMS depth profile for GaAs(10i6)/Ni^GaAs+NiAs/750A Ge/300A Ti. (a) as deposited (b) ex situ vacuum annealed at 500°C for 5 minutes. Ni^GaAs+NiAs is the product of 650A Ni in situ annealed at 400°C for 15 minutes.

PAGE 110

f Ion milling hole •-GaAs 500A (b) (c) Figure 4-22. TEM micrographs of as-deposited GaAs/Ni 2 4GaAs/250A Ge. (a) Bright-field image of cross-sectional view. (b) Diffraction pattern from (a), B=[l 10]^^^ (c) Schematics of diffraction pattern in (b).

PAGE 111

104 Ni Figure 4-23. EDS spectra from Ni 2 4 GaAs in as-deposited GaAs/Ni 2 4GaAs/25oA.

PAGE 112

105 keV Figure 4-24. EDS spectra from GaAs in as-deposited GaAs/Ni 2 4GaAs/250A.

PAGE 113

106 ^S[iGa(Ge) -^Ni2 4GaAs(Ge) -GaAs (a) 500A (b) Figure 4-25. TEM micrographs of GaAs/Ni 2 4GaAs/250A Ge ex situ vacuum annealed at 500°C for 2.5 minutes. (a) Bright-field image of cross-sectional view. The arrow indicates a grain imaged in (c). (b) Diffraction pattern from (a), B=[l IOJq^^ (c) Dark field image of (a) formed by the diffraction spot indicated in (b). The arrow indicates the grain indicated by an arrow in (a).

PAGE 114

Figure 4-25. TEM micrographs of GaAs/Ni 2 4GaAs/250A Ge ex situ vacuum annealed at 500°C for 2.5 minutes. (a) Blight-field image of cross-sectional view. The arrow indicates a grain imaged in (c). (b) Diffraction pattern from (a), B=[l (c) Dark field image of (a) formed by the diffraction spot indicated in (b). The arrow indicates the grain indicated by an arrow in (a).

PAGE 115

108 <^sTiGa(Ge) -NiGa(AsGe) -GaAs (a) 500A Figure 4-26. TEM micrographs of GaAs/Ni 2 4GaAs/250A Ge ex situ vacuum annealed at 500°C for 5 minutes. (a) Bright-field image of cross-sectional view. The arrow indicates a NiAs phase grain. (b) High resolution image in cross-sectional view. The arrow indicates a NiAs phase grain.

PAGE 116

109 200A GaAs ' j,' N . Figure 4-26. TEM micrographs of GaAs/Ni 2 4GaAs/250A Ge ex situ vacuum annealed at 500°C for 5 minutes. (a) Bright-field image of cross-sectional view. The arrow indicates a NiAs phase grain. (b) High resolution image in cross-sectional view. The arrow indicates a NiAs phase grain.

PAGE 117

110 (b) (c) lU u keV (d) 10. 0 keV (e) (a) Simplified entire structure to indicate probing points for EDS analysis. (f) koV kev Figure 4-27. Schematics of bright field TEM micrograph and EDS spectra from GaAs/Ni2 4GaAs 250A Ge ex situ vacuum annealed at 500°C, 5 minutes (a) Schematics of the entire structure and location of probed points. (b) (f) EDS spectra measured at the points shown in (a).

PAGE 118

Ill © Ni • As Vacancy (a) NiAs:a — 3.61 A C = 5.03A ^ extra Ni atoms in half occupancy (b) Ni3Ga2:a = 4.00A C = 4.98A ^ extra Ni atoms in fiill occupancy (c) Ni^In Z A Figure 4-28. Unit cells of the crystal structures of (a) NiAs, (b) Ni 3 Ga 2 , and (c) Nijin.

PAGE 119

112 (a) 500A (b) Figure 4-29. TEM microgrphs of bright-field cross-sectional images of GaAs/Ni 2 4GaAs/500A Ge in the (a) as-deposited, (b) ex situ vacuum annealed at 500°C, 5 minute conditions. NiGaGe GaAs

PAGE 120

113 Ni 24 GaAs 500A Figure 4-30. Bright-field cross-section TEM micrograph of as-deposited GaAs/Ni 2 4GaAs/250A Ge/300A Ti. The arrow indicates a protrusion of Ni 2 4 GaAs into the GaAs substrate.

PAGE 121

114 (a) Figure 4-31. High resolution EDS analysis of Ni, Ga and As in Ni 2 4 GaAs from as-deposited GaAs/Ni 2 4 GaAs 250A Ge/300A Ti. (a) Results of EDS analysis, (b) Schematics of approximate location of the EDS probing points. The first two points are from a Ni 2 4 GaAs protrusion into GaAs.

PAGE 122

115 500A Figure 4-32. Bright-field cross-section TEM image of GaAs/Ni2 4GaAs/250A Ge/300A Ti ex situ vacuum annealed at 500°C for 2.5 minutes. The three arrows indicate NiAs grains.

PAGE 123

116 ^ . . NiTi ^aAs 500A Figure 4-33. Bright-field cross-section TEM image of GaAs/Ni 2 4GaAs/250A Ge/300A Ti ex situ vacuum annealed at 500°C for 5 minutes. The two arrows indicate Nij 3AS grains.

PAGE 124

117 Figure 4-34. High Resolution EDS analysis across the interface of regrovm GaAs in GaAs/Ni 2 4GaAs/250A Ge/300A Ti ex situ vacuum annealed at 500°C for 5 minutes. (a) Analysis of the regrown GaAs/Nij 3 AS interface. (b) Analysis of the regrown GaAs/Ni 2 jGa interface.

PAGE 125

118 stance (A) (b) Figure 4-34. High Resolution EDS analysis across the interface of regrown GaAs in GaAs/Ni 2 4GaAs/250A Ge/300A Ti ejc situ vacuum annealed at 500°C for 5 minutes. (a) Analysis of the regrown GaAs/Nij 3 AS interface. (b) Analysis of the regrown GaAs/Ni 2 jGa interface.

PAGE 126

119 (b) Figure 4-35. TEM photomicrograph from GaAs/Ni 2 4GaAs/250A Ge/300A Ti ex situ vacuum annealed at 500°C, 5 minutes. (a) Dark-field cross-section image. The arrow marks a grain shown in (c). This grain is Nij 3 AS phase. (b) Diffraction pattern from (a). B=[ 110 ]q^ 3 . The arrow indicates the diffraction spot used to form the image of (a) (c) Bright-field image of corresponding to the dark-field image shown in (a). The arrow indicates the same grain indicated in (a).

PAGE 127

120 c GaAs 500A Figure 4-35. TEM photomicrograph from GaAs/Ni 2 4 GaAs/ 250A Ge / 300A Ti ex situ vacuum annealed at 500°C, 5 minutes. (a) Dark-field cross-section image. The arrow marks a grain shown in (c). This grain is Nij ^As phase. (b) Diffraction pattern from (a). B=[l The arrow indicates the diffraction spot used to form the image of (a) (c) Bright-field image of corresponding to the dark-field image shown in (a). The arrow indicates the same grain indicated in (a).

PAGE 128

121 • >y< • IX ^ • >6 • fiK xi • ^ • % • K • • : GaAs • x: NiAs >6 • B • : Ni3Ga2 ft M • ^ . 9c • « )P 9c (d) (e) Figure 4-35. Diffraction pattern anlysis for phase identification in GaAs/Ni 2 4GaAs/250 A Ge/300A Ti ex situ vacuum annealed at 500°C, 5 minutes. Beam direction in (d) and (e) is [1 10]^^^ (d) Simulated diffraction patterns of NiAs and Ni 3 Ga 2 (e) Diffraction patterns from Nij 3 AS and Ni 2 jGa phases taken along B = [ 110 ](j^^

PAGE 129

122 looA Figure 4-36. High resolution TEM cross-section image of regrovm GaAs from GaAs/Ni 2 4GaAs/250A Ge/300A Ti ex situ vacuum annealed at 500°C for 5 minutes. The arrow indicates a precipitate formed in regrown GaAs. A indicates grains of Nij 3 AS. The dashed line is the location of original interface between Ni 2 ^GaAs/GaAs (S-GaAs) before regrowth.

PAGE 130

123 (a) (b) Figure 4-37. Bright-field cross-section TEM image of GaAs/650A Ni/250A Ge/300A Ti (a) As-deposited. (b) Ex situ vacuum annealed at 500°C for 5 minutes. The arrow indicates underdeveloped NiAs phase.

PAGE 131

124 Figure 4-38. Bright-field cross-section TEM image of GaAs (10i^)/Ni2 4GaAs/750A Ge/300A Ti ex situ vacuum annealed at 500°C for 5 minutes. A is Ni-Ge compounds. B is Ni^GaGe ternary.

PAGE 132

GaAs 5 ^ Figure 4-39. Bright-field cross-section TEM image of GaAs (10>6)/Ni^GaAs + NiAs/750A Ge/300A Ti ejc situ vacuum annealed at 500°C for 5 minutes.

PAGE 133

CHAPTER 5 DISCUSSION Evolution of Ni2.4GaAs Table 5-1 is a tabulation of the phases observed during evolution of the Ni 2 . 4 GaAs layer formed by in situ annealing of 650A Ni film at 300°C, for 15 minutes. Table 5-1. Phases after annealing at 500°C, 5 minutes. Layered Structure (A) As-deposited Layered Structure After annealing at 500°C, for 5 minutes GaAs Ni 2 . 4 GaAs Ge R-GaAs NiAs, Ni 2 . 4 GaAs(Ge) 1200-1300 250 1200-1300 500 R-GaAs NiAs*, NixGaGe Ni-Ga-Ge GaAs Ni 24 GaAs Ge Ti R-GaAs, Nii.sAs, NiziGa, NiTi R-GaAs, Ni-Ga-Ge, NixGaGe, Ni-Ge 1200-1300 1200-1300 250 750 300 300 GaAs Ni Ge Ti NixGaGe NiTi 650 250 300 NixGaGe indicates Ni-Ga-Ge ternary with NiAs-type hexagonal structure. Ni-Ga-Ge indicates ternary phases consisting of Ni, Ga and Ge (table 4-11). Nii. 3 As is NiAs containing Ga and Ge (table 4-7). NiziGa is NisGa 2 containing Ge and As (table 4-9). Ni-Ge indicates several Ni-Ge compounds (table 4-11). R-GaAs indicates regrown GaAs. 126

PAGE 134

127 Upon annealing Ni 2 . 4 GaAs at 500°C, simple decomposition of Ni 2 . 4 GaAs occurred. Ni 2 . 4 GaAs +2,4Ge = GaAs + 2.4NiGe (5-1 ) To determine whether the reaction (5-1) should proceed to the right, the heat of formation of Ni 2 . 4 GaAs was needed. No experimental or theoretical thermodynamic data for Ni 2 . 4 GaAs was found. Therefore, the following reaction was considered. NiGa +NiAs +2Ge = GaAs + 2NiGe, AHf = -40.9 kJ/mole. (5-2) and NiAs and NiGa should be decomposed by Ge to form NiGe and GaAs (see appendix A for thermodynamic data). Because Ni 2 , 4 GaAs is presumably a metastable phase which eventually separates to NiAs and NiGa, its heat of formation should be less negative than the summation of NiGa and NiAs (-96.8kJ/mole). Because reaction (5-2) is thermodynamically favorable, therefore, reaction (5-1) should proceed to the right under at least some conditions. In the metallization structure GaAs / Ni 2 . 4 GaAs / 250A Ge, approximately 25oA of Nix 4 GaAs was decomposed as a result of reaction (5-1) after annealing at 500°C, for 5 minutes. Simultaneous with reaction (5-1), in some localized area of the interface, the NiuAs phase was formed from Nix 4 GaAs by release of Ga and Ni. When formation of the Nij.sAs phase proceeded, about 4 at. % of Ge diffused into the grains of the Ni^As phase. Upon annealing at 500°C. Ni 2 . 4 GaAs + 0. IGe = Nii.3As(Gao.4Geo.i) + NiGax (5-3) The solubility of Ge in the Nit^As phase was expected to be smaller than in Ni 2 , 4 GaAs (~15%) because the unit cell of Nii.jAs is smaller than Ni 2 . 4 GaAs, provided that the

PAGE 135

128 Nii.sAs maintains the NiAs-type hexagonal structure (see figure 4-28). Recall that the covalent radii of Ga is larger than As, causing the expansion of the unit cells. The fact that GaAs / Ni 2 . 4 GaAs / 250 A Ge did not exhibit linear I-V characteristics after annealing at 500°C for 5 minutes suggests that simple decomposition of Ni 2 . 4 GaAs, such as reaction (5-1) resulting in ~25oA of regrown GaAs, was not sufficient to form ohmic contacts. When 25oA Ge or 200 300A Ti was added to the GaAs / Ni 2 . 4 GaAs / 250 A Ge structure, those additional Ge and Ti layers provided additional potential for decomposition of the Nii.sAs phase (table 4-7) as follows: Nii. 3 As(Gao. 4 Geo.i) + 1.3Ge = 1.3NiGe + GaAs:Ge (5-4) Thus, in case of GaAs / Ni 2 . 4 GaAs / 500 A Ge, approximately 300A of the Nii.sAs phase was decomposed through reaction (5-4) along with the simple decomposition shown by reaction (5-1). The consequence of these decomposition reactions was clear by the phases shown in the bright-field image of figure 4-29 (b), where the very smooth interface with GaAs resulting from decomposition of NiAs phase can be seen. As can be expected in the regrowth of GaAs occurring in reaction (5-4), decomposition of the NiAs phase will allow Ge to occupy Ga sites, forming n^-GaAs. In the case of GaAs / Ni 2 . 4 GaAs / 250 A Ge / 300 A Ti, the simple decomposition of Ni 2 . 4 GaAs took place through a Ni-Ti reaction upon annealing at 500°C for 5 minutes: Ni 2 . 4 GaAs + 2.4Ti = GaAs + 2.4NiTi. (5-5) Reaction (5-5) has more of a driving force towards the right because the heat of formation of the NiTi compound (-67.8kJ/mole) is much more reactive than that of NiGe compound

PAGE 136

129 (-3 1 8kJ/mole), and regrowth of GaAs would result in a larger driving force to the right. While the simple decomposition of Ni 2 . 4 GaAs continued as described in (5-5), the decomposition of Nii.sAs also occurred: Nii. 3 As(Gao. 4 Geo.i) + 1.3Ti = 1.3NiTi +GaAs:Ge (5-6) As a direct consequence of the decomposition of the Nii.sAs phase, ~300A GaAs doped with Geoa regrew, exhibiting the linear I-V shown in figure 4-5. In addition to the formation of the Nii.sAs phase, a Ni-Ga compound was also produced from Ni 2 . 4 GaAs by releasing As and Ni atoms (table 4-9) as follows; Ni 2 . 4 GaAs + 0.4Ge = Ni2.iGa(Aso.2Geo,4) + NiAsx (5-7) Basically, reactions (5-3) and (5-7) indicate the phase separation of Ni 2 . 4 GaAs into Ni-As and Ni-Ga binaries, which has been reported [OgaSO, Lav86, San87, Gue89] Reactions (5-1) through (5-7) describe the evolution of the Ni 2 . 4 GaAs upon annealing at 500C, for 5 minutes as complied in table 5-2 and illustrated in figure 5-1. Table 5-2. Reactions during the evolution of Ni 2 . 4 GaAs upon annealing at 500°C, for 5 minutes. (5-1) Ni 2 . 4 GaAs + 2.4Ge = GaAs + 2.4NiGe (5-2) NiGa +NiAs +2Ge = GaAs + 2NiGe (5-3) Ni 2 . 4 GaAs + 0. IGe = Nii,3As(Gao,4Geo,i) + NiGax (5-4) Nii. 3 As(Gao. 4 Geo.i) + 1.3Ge = 1.3NiGe + GaAs;Ge (5-5) Ni 2 . 4 GaAs + 2.4Ti = GaAs + 2.4NiTi (5-6) Nii. 3 As(Gao. 4 Geo.i) + 1.3Ti = 1.3NiTi +GaAs:Ge (5-7) Ni 2 . 4 GaAs + 0.4Ge Ni 2 .oGa(Aso. 2 Geo. 4 ) + NiAsx As shown in figure 5-1, the formation of Ni 2 .iGa compound released As and Ni (reaction 5-7). And the formation of Nii. 3 As phase released Ga and Ni (reaction 5-3). While such transformations proceeded, Ge diffused into both phases. The released Ga, As

PAGE 137

130 and Ni atoms reach the top metal layer where Ga-Oxygen and Ni-Ti (or Ni-Ge in case of GaAs / Ni 2 . 4 GaAs / 500 A Ge) reactions took place. At 500“C, all three atoms were very mobile [Rob75, Hei82, Che88], These redistributions of elements accounted for the detection of Ga, As and Ni in the top metal layer as shown in the depth profiles (see e.g, figure 4-14). In reaction (5-6), decomposition of the Nii.sAs phase released predominately As over Ga, resulting in an As-rich condition for the regrowth of GaAs. Some additional Ga released during the reaction (5-3) was supplied to the regrowing GaAs in reaction (56). However, as can be seen in figure 5-1, the released Ga atoms must have diffused through the Nii.sAs phase to reach the regrowing GaAs. Thus, it is more likely that Ge occupied Ga sites to be Geoa rather than Ga^ in the regrown GaAs, leading to the formation of n^^-GaAs. Simple decomposition of the Ni 2 . 4 GaAs (reaction (5-5)) is also depicted in figure 5-1. It is not likely that Ge doped GaAs formed by this simple decomposition. In fact, Ni 2 . 4 GaAs was slightly Ga-rich. The evolution summarized in figure 5-1 would correspond to the results of GaAs / Ni 2 . 4 GaAs / Ge / Ti and GaAs / Ni 2 , 4 GaAs/ Ge structures after annealing at 500°C, for 5 minutes as summarized in table 5-1. To better understand the evolution of Ni 2 . 4 GaAs, the Ni-Ga-As ternary phase diagram from Gu’erin et al [Gue89] was considered (figure 5-2). As can be seen, the ternary phase diagram is characterized by lack of tie lines connecting Ni 2 . 4 GaAs and GaAs as discussed in Chapter 2. Note that all the tie lines connected to GaAs are from the Ga-rich Ni-Ga and As-rich Ni-As binaries. The approximate composition of Ni 2 . 4 GaAs identified in our study is marked by the X. The composition of NiuAs and Ni 2 .iGa would be the

PAGE 138

131 point marked A and B in figure 5-2, the concentration of Ga and As, respectively and Ge concentrations (~4%) was not taken into account. It clearly shows that Nii.sAs and Ni 2 ,iGa phases identified in our study were indeed the transformation of Ni 2 . 4 GaAs into Ni-As and Ni-Ga binaries based on the phase diagram. The transformation of Ni 2 . 4 GaAs into Nii.sAs and Ni 2 .iGa phases was followed by the decomposition of these binaries through Ni-Ge and Ni-Ti reaction. The driving force for this decomposition would be the relative thermodynamic stabilities between phases containing these elements shown in reactions (5-2) and (5-5). For example: (5/3)Ni3Ga2 + 2Ge = NisGe 2 + (10/3)Ga AHf = -26.8kJ/mole. (5-8) 2NiGa + Ge = N^Ge + 2Ga AHf = -46.0kJ/mole. (5-9) As shown above, the Ni 3 Ga 2 , Ni 2 .iGa, and NiGa, all phases will decompose to form Ni-Ge compounds with lower heat of formation. It was noticed that Ni-As binaries were characterized by lower heat of formation than Ni-Ga binaries (see appendix A). A consequence of this characteristic would be that the following reaction would move to the left rather than the right:. NiAs + Ge = NiGe + As AHf = +41.5kJ/mole. (5-10) This result implies that reaction (5-4) might not be thermodynamically favorable. To clarify this point and to understand the entire system comprised of Ni-Ge compounds along with Ni-As, Ni-Ga, and GaAs, a quaternary phase diagram was needed. By using the thermodynamic data compiled in appendix A and following the principles explained in appendix B, Ni-Ga-Ge and Ni-As-Ge ternary phase diagrams were calculated and the results are shown in figure 5-3 and 5-4, respectively. It should be emphasized that

PAGE 139

132 the use of these calculated phase diagrams is restricted to description of general trend in phase equilibria in the solid-state since many assumptions and simplification were made to calculate them. The most important characteristic of the Ni-Ga-Ge phase diagram is that there are several possible tie lines between Ni-Ga compounds and Ni-Ge compounds. This implies that pure Ge always reacts with Ni-Ga compounds to produce Ni-Ge compounds. In figure 5-3, two ternary phases, Ni 4 GaGe 2 and Ni 2 GaGe found in JCPDS files are added and marked by C and D, respectively. It is expected that there could be several tie lines connecting Ni-Ge compounds to Ni-Ga compounds through those two ternary compounds in real system. In contrast to the Ni-Ga-Ge system, the Ni-As-Ge system was characterized by the lack of tie lines connecting NiAs to Ni-Ge compounds (see figure 5-4). This characteristic results from the fact NiAs and NiuAsg are thermodynamically stable with respect to pure Ge as follows; NiGe + GeAs = NiAs + 2Ge AHf = -38.6kJ/mole. (5-11) 8NiAs+3NiGe = NiiiAs8 + 3Ge AHf = -91.2kJ/mole. (5-12) The two reactions, (5-1 1) and (5-12), indicate the lack of NiAs-NiGe and NiGe-GeAs tie lines. Two ternary phases found in JCPDS files were marked as E and F, respectively. There could be tie lines connecting NiAs to Ge through these Ni-As-Ge ternary phases in real system. By combining these two calculated Ni-Ga-Ge and Ni-As-Ge diagrams with experimentally determined Ni-Ga-As [Gue89] and Ge-Ga-As [Bey84], a Ni-Ga-Ge-As

PAGE 140

133 quaternary phase diagram was determined and is shown in figure 5-5. The primary use of this quaternary phase diagram was to determine possible tie lines connecting GaAs and NiGe compounds to project the most likely phase equilibria involving Ni-Ge compounds. Since a tie plane and a tie line can not across each other in a quaternary system according to the Gibbs phase rule, a tie plane defined by NiGa-NiAs-Ge and a tie line GaAs-NiGe can not cross. Therefore, a thermodynamic systems consisting of NiGa, NiAs and Ge is thermodynamically unstable, resulting in formation of GaAs and NiGe as follows: NiGa + NiAs +2Ge = GaAs +2NiGe AHf = -4 1 kJ/mole (5-2) The negative heat of formation indicates that the tie line between GaAs-NiGe is stable. As a consequence of this stable GaAs-NiGe tie line, four phases including NiGe and GaAs can define tetrahedrons where one of four phases are located at each of the four corners. For example, GaAs-NiGa-Ni 3 Ge 2 -NiGe and GaAs-NiGa-Ni 2 Ge-NiGe may form tetrahedrons. The presence of Ni-Ga-Ge ternaries in figure 5-3 suggests the possibility of GaAs (Ni-Ga-Ge) NisGe 2 NiGe equilibrium. Contrary to the Ni-Ga compounds, there were no possible tetrahedrons incorporating GaAs, Ni-As binaries and Ni-Ge compounds all together because of the lack of tie lines between NiAs and Ni-Ge compounds. This consequence implies that the Ni-As phases are not thermodynamically stable when Ni-Ge compounds are produced by adding Ge to Ni-Ga-As ternary system. Therefore, in phase equilibria between GaAs and Ni-Ge compounds, Ni-As binaries should not be maintained. This explains the driving force for decomposition of the Nii.sAs phase in the presence of Ni-Ge compounds. In table 5-1, note the different reaction products in GaAs / Ni 2 , 4 GaAs / 25oA Ge / 300A Ti versus GaAs / Ni 2 . 4 GaAs / 75oA Ge / 300A Ti

PAGE 141

134 after anneal at 500°C, for 5 minutes. As listed table 4 11 , all of the possible reaction products in annealed GaAs / Ni2.4GaAs / 750A Ge / 300A Ti were NisGes, NisGe2, Ni2Ge and Ni-Ga-Ge ternary phases, while no Ni-As binaries were detected. This agrees with the reaction products predicted in Ni-Ga-GeAs quaternary phase diagram as Ni-Ge reaction became predominant. When a small of amount of Ti was added to the Ni-Ga-GeAs system, the phase equilibrium involving Ni-Ge compounds was enhanced, since Ti decomposed Ni-Ge compounds to form Ni-Ti compounds because of their lower heat of formation (see appendix A). Released Ge enhanced the Ni-Ge reactions, helping to achieve the phase equilibrium involving Ni-Ge compounds. Analyzing the annealed structure of GaAs / 65oA Ni / 25oA Ge / 300A Ti (without a Ni in situ anneal) helped interpretation of the evolution of Ni2.4GaAs and its effects on the formation of ohmic contacts. As shown in figure 4-35 (b), the thickness of the bottom layer was ~800A, thinner than that of annealed structures of GaAs / Ni2.4GaAs / 25oA Ge / 30oA Ti after 500”C for 5 minutes. The main reason for this thinner bottom layer was that Ni reacted simultaneously with GaAs and Ti because it was not in situ preannealed. Consequently, only some of the 650A Ni film reacted directly with GaAs, resulting in a thinner bottom layer. SIMS data showed that Ge also diffused into the bottom layer, and the bright field image showed dark/bright contrast, indicating the presence of the NiAs phase (figure 4-37 b). However, there was no decomposition of the NiAs phases by the Ni-Ge or Ni-Ti reactions because Ge and Ti were consumed to react with pure Ni during the anneal at 500°C for 5 minutes. Consequently there was no regrown GaAs doped with

PAGE 142

135 Ge. The absence of the regrowth was confirmed by comparing the interfacial morphology and the characteristics of GaAs in figure 4-37 (b) to the characteristics of regrown GaAs in figure 4-36. Regrown GaAs exhibited precipitates containing Ni. Formation of precipitate such as NisGaAs in the Ni-supersaturated GaAs was also reported by Chang et al [Cha94], It is interesting to point out that those precipitates are distributed predominantly within ~15oA region of the interface, and the first ~200A of the regrown GaAs was free of these defects. A similar situation has been observed in recrystallization of amorphized GaAs where the first material to recrystallize was free of defects and the last material to regrow contained more defects, providing nucleation sites for precipitate from the supersaturated Ni solution [Kul80, Shi88]. Planar defects such as twins and stacking faults were also observed in the regrown GaAs in addition to precipitates (see figure 4-36). It was reported that the formation of microtwins during the regrowth of GaAs from the amorphized state involves nonstoichiometry of the Ga7As ratio [Shi88]. Similar nonstoichiometry is expected in our case. Because the source materials for regrown GaAs are Nii.aAs and Ni 2 .iGa where Ga/As ratios are not expected to be 1:1, the regrown GaAs is imperfect in terms of composition as well as structure. As indicated above, the annealed structure without an in situ anneal of Ni film did not exhibit any of these characteristics. As mentioned in the literature review, simple decomposition of NixGaAs [Tan88] or PdxGaAs [Wan88] has been believed as the mechanism by which Ge or Si was incorporated into regrown GaAs. Based on the present study, it is clear that evolution of

PAGE 143

136 Ni 2 . 4 GaAs was not only a simple decomposition described by Ni 2 . 4 GaAs + 2.4Ti (or 2.4Ge) == GaAs + 2.4NiTi (or 2.4NiGe). Besides this simple decomposition, Ni 2 . 4 GaAs transformed into Ni-As and Ni-Ga compounds (Nii.sAs and Ni 2 ,iGa in our experiment). Decomposition of Nii.sAs phase containing Ge is believed to be the route by which Ge incorporation into regrown GaAs on Ga sites to be a n-type donor. The driving force for decomposition of several phases including Nii.sAs and the overall reaction paths are consistent with thermodynamic data and Ni-Ga-Ge-As quaternary phase diagram Mechanism for the Formations of Ohmic Contacts. The structure without Ge (GaAs / Ni 2 . 4 GaAs / 600A Ti ) did not form linear I-V despite the fact that regrowth of GaAs was observed. Also based on our results collected from several different structures, the presence of Ge and its concentration are critical factors in the formation of ohmic contacts. Based on the fact that structures with 250 50oA Ge exhibited linear I-V data with undecomposed Nii.aAs and Ni 2 .iGa containing Ge, formation of Ge/GaAs or GOxGaAsi-x [Bar82] hetero structures are not necessary nor responsible for switching from rectifying to ohmic contacts. Based on TEM data, no Ge epitaxial layers were observed. Therefore, it is concluded that the regrown GaAs must be doped with Ge to a density above that of the substrate, resulting in a n^-region and narrow barrier width in order to allow field-emission or thermionic field emission transport to form an ohmic contact.

PAGE 144

137 It is known that Ni is a deep acceptor in n-type GaAs [Enn8 1, Hiz91], As listed in table 4-10, the Ni concentration in regrown GaAs was 2.6 at. %, much higher than the Ge concentration (0.4 at. %). Since the equilibrium solubility of Ni m GaAs is 10 10 cm' [Mat71, Cha94], the actual Ni concentrations which will act as acceptors are not expected to reach 2.6 at. % and therefore should not fully compensate the Ge donors. Figure 5-6 illustrates the expected electronic band structure after the regrowth with an assumption that the barrier height b is 0.8 eV. In order to fully understate the energy levels shown, the Fermi level in the GaAs substrate was calculated in figure 5-7 [Sch50]. The Fermi level is obtained from the requirement of charge neutrality throughout the solid. [nf = [N/f + [pf (5-13) where [n] is free electron concentration, [Na ] is ionized donor, and [p] is hole concentration. The superscript S indicate the substrate GaAs which has not been decomposed by Ni or regrown and will be designated by S-GaAs hereafter. Because the GaAs substrate was doped with Si on the level of 2 x 10**cm'^, [Na = [Si^]^ ~ [n]^ ~ 2 x 10‘*cm'^ at room temperature. As indicated in equation (5-13) and as shown in figure 5-7, the Fermi level is the energy level where [n] intersects [Nsi^] + [p]. Note that [n] is not linear as the Fermi level approaches the conduction band minima because of degenerate doping [Bla82]. Degenerating doping in GaAs at 300K begins near 1.4eV where [n] exceeds the density of state ( ~6 x 10*’cm'^ ), although noticeable deviation from linear [n] was observed above the conduction band at Ec =1 45eV as shown in figure 5-7. In figure 5-6, all the energy levels were set relative to the valance band maxima at Ev =0.0eV. Therefore, Ec is equal to 1.45eV, the band gap of GaAs. As shown in figure 5-6 and 5-7,

PAGE 145

138 the Fermi level was ~ 1.44eV for [Nd"^] = 2 x 10**cm'\ i.e. very close to the conduction band. With reference to the Fermi level of S-GaAs, the electronic band structure is shown in figure 5-6. The conduction band in the substrate GaAs, Ec^ and the valence band Ev^ would be equal to bands in GaAs before any dissolution or regrowth. The built-in potential Vbi is obtained from Ob (Eg Ef) where Ob = 0.8 is the barrier height between surface pinned Ef and Ec. With respect to Ev* =0 eV, Eg =1 .45 eV and Ef = 1 .44 eV, yielding Vbi = 0.77 eV. Likewise, the difference between Fermi level and the valence band at the interface, i.e. Ef Ev,x=o^ is calculated from 1.45 0.8 = 0.65eV. The superscript R in figure 5-6 indicates regrown GaAs (hereafter R-GaAs). There was no superscript used in the Fermi level because there is only one Fermi level throughout the solid. In the band diagram of figure 5-6, Ef is 1.44 eV with respect to the valence band mainima at x=oo, i.e., in the substrate GaAs. As discussed above, regrown GaAs contained 2.6 at. % of Ni. R-GaAs may also contain Ga vacancies which are well known, stable acceptor in n-GaAs. As a result, a large concentration of compensating acceptors could exist in R-GaAs. The distribution of acceptors would be expected to maximum at the interface, x = 0, and gradually decrease towards the original interface at x = R. Compensating acceptor would result in a Fermi level distribution like the dashed curve indicated by Ef t=oHowever, there would be a limit to the extent of compensation. In figure 5-6, the valence band maximum at the interface, Ev,x=o, is very near to the Epi , the stabilized Fermi energy level of GaAs, which is an important parameter in the

PAGE 146

139 amphoteric native defects model as discussed in Chapter 2 [Wal88, 89], Following the amphoteric native defects model, a maximum shift of ~0.7eV (from the level of 1.44eV to the level of 0.7eV with respect to Ev =0eV) would be expected. Any further attempt to shift the Fermi-level would be counter balanced by the amphoteric nature of native defects which would stabilize the level. Note the Fermi level shift is only with concentration of dopants > 10^*cm'^. Whatever the magnitude of the Fermi level shift, (i .e. compensation) it can not be maintained because change would occur to maintain a constant potential throughout the crystal. Thus, Ge incorporated through evolution of Ni 2 . 4 GaAs will be ionized to Ge^ to maintain a constant Fermi level. In another words, activation of Ge into Ge"^ is controlled by the constant Fermi level. The Fermi level in the R-GaAs is determined by [GeT = [Naf + [nf (5-14) where [Na']^ is the concentration of possible acceptors which are expected to be approximately lO'Vm'^ as discussed already. Because the EDS data confirmed the presence of [Ge] to at least -1.5 x lO^^’cm'^ in the structure which exhibited ohmic behavior, [Ge"^]*^ was taken to be [1.5x1 0^®] / { 1 + 2exp(Ef Eoe)/kT } (5-15) and plotted as function of Fermi level in figure 5-7. Based on this calculation, the Fermi level in the R-GaAs was determined at 1.5 leV where [Ge intersect [n]^, which is well above the conduction band edge of R-GaAs (1 45eV). Thus doping in R-GaAs is degenerate, as shown by bold-dashed line labeled E/ in figure 5-6. This band structure allows tunneling across the metal/GaAs interface. Since [Ge] concentration within 5oA

PAGE 147

140 from the metal/R-GaAs is expected to exceed 1.5 x lO^^cm'^ ( up to 5.8 x lO^^’cm'^) as reported in Chapter 4, the actual conduction band shift in the R-GaAs will be larger than 0.07eV. Therefore, the actual depletion distance will be shorter than 5oA. Even though [Ge] was measured to be 1.5 x 10^® cm'^, the concentration of free electron in the regrown GaAs is expected to be less than 1 .5 x 10^” cm'^ as will be discussed below. Because this range of concentrations is degenerate doping, [n]*^, the concentration of free electrons in the regrown GaAs is given by a Fermi-Dirac distribution as follows: [n]*" = Nc-(2/7i'^) -Fi/2(Ef -E,’'/kT). (5-16) where Nc is density of states at the conduction band edge (~6 x lO^’cm'^ at 300K), and F 1/2 is Fermi-Dirac integral [BlaSl]. For Ef -Ec*^ ~ 0.07eV, Fi/ 2 (Ef -E//kT) is ~3.7 [Bla74]. Using Nc and Fi/ 2 (Ef -EcVkT), [n]^ was calculated to be ~4 x 10'*cm‘^ from equation (5-16). This same value can also be read in figure 5-7. In the degenerate doping range, the periodic potential of the lattice atoms is strongly disturbed through the interactions between lattice atoms and electrons in the conduction bands, resulting in conduction band tailing into the bandgap instead of a well-defined band edge for the density of states [Lou95]. As a consequence of this band-edge tailing, the density of state at the edge of conduction band is not determined by the square root of energy. Such a deviation from conventional square root dependency on energy becomes significant at doping concentrations above ~10^“cm'^, and the effective density of state, Nc, increases by one order of magnitude compared to the density of states for the ~10'*cm'^

PAGE 148

141 doping range [Lou95], Nc is expected be ~ 6 x 10‘*cm'^ with [Ge] = 1.5 x instead of ~6 x 10^’cm'^ for [Si] = 2 x lO'^cm'^, Therefore, [n]*^ in the regrown GaAs, where [Ge] is doped at 1.5 x lO^'^cm'^, is expected to be ~4 x lO^^cm'^. This approximation indicates that the concentration of free carriers (electrons in this case) is less than the doping concentration, which is expected in degenerate doping. From the discussion above, n^-GaAs is expected to form with [n] ~ 4 X 10 ‘^cm’^ with [Ge] = 1.5 x lO^'^cm'^.

PAGE 149

142 (a) as-deposited Ti Ge GaAs Reaction (5-7) Overall reaction :Ni2 4GaAs +Ge +Ti = Nij 4 GaAs + 0.4Ge = R-GaAs:Ge +NiGa+ NiAs + NiTi Ni 2 ,oGa(Aso. 2 Geo. 4 ) + NiAs, Figure 5-1. Schematic of the evolution of the Ni 2 4 GaAs layer. (a) Schematic of as-deposited GaAs/Ni 2 4GaAs/250A Ge/300A Ti. (b) Schematic of the formation of Nij 3 AS, M 2 iGa containing Ge and decomposition of the binaries after ex situ vacuum annealing at 500°C for 5 minutes. The arrows show direction of net diffusion causing, redistribution of elements and ultimately regrowth of Ge doped GaAs. (b) 500®C, 5 minutes GaAs

PAGE 150

143 Ni Figure 5-2. Isotherm section of the solidus portion of the bulk Ni-Ga-As ternary phase diagram (at. %). The general features of this diagram are valid from 298 to -1073K. Only solid line indicates true tie lines. X indicates the approximate composition of Ni 2 4 GaAs in our study. A and B indicates the approximate composition of Nij gAs and Nij jGa compounds, respectively, in our study. From reference by Guerin et al [Gue89].

PAGE 151

144 Ni Figure 5-3. Calculated ternary phase diagram for Ni-Ga-Ge for solid state reactions. C and D indicate the composition of Ni 4 GaGc 2 and Ni 2 GaGe, respectively. Only general features of this diagram (see text) are expected to be valid from 298 to (approximately) 900K. See appendix A and B for an explanation of how this diagram was determined.

PAGE 152

145 Ni Figure 5-4. Calculated phase diagram for Ni-As-Ge for solid state reactions. E and F indicate the compositions of Ni 4 Gc 7 Asg and Ni 3 Gc 2 As 5 , respectively. Only general features of this diagram (see text) are expected to be valid from 298 to (approximately) 900K. See appendix A and B for an explanation of how this diagram was determined.

PAGE 153

146 Ge a:Ni5As2 Figure 5-5. Calculated quartemary phase diagram for Ni-Ga-Ge-As for solid state reactions. The Ni-Ga-Ge and Ni-As-Ge ternary phase diagrams were calculated. The Ni-Ga-As [Gue89] and Ge-Ga-As phase diagrams [Bey84] were experimentally determined. Only general features of this diagram (see text) are expected to be valid from 298 to (approximately) 900K.

PAGE 154

147 Original interface before regrowth Figure 5-6. Schematic of the contact morphology during regrowth of GaAs, with an illustration of the relative E^., and Ef energy levels. See text for explanation. Note Ge^ doping in the regrown GaAs to maintain Fermi level throughout the entire crystal.

PAGE 155

Carrier Con. err? 148 Figure 5-7. Calculated Fermi level vs. carrier concentration at 300K. The substrate GaAs was doped with Si to 2 xlO^^cm"^. See text for explanation.

PAGE 156

CHAPTER 6 CONCLUSIONS Electrical and metallurgical properties of Ti/Ge/Ni metallizations have been studied with CurrentVoltage (I-V) measurement. Secondary Ion Mass Spectroscopy (SIMS), AES (Auger Electron Spectroscopy) and Transmission Electron Microscopy (TEM) with Energy Dispersive X-ray (EDS). The purpose of this study was to identify the evolution of Ni-Ga-As phases caused by Ni-Ge and Ni-Ti reactions and their effects on the formation of ohmic contacts. In situ annealing of Ni/GaAs was adopted to better understand the evolution of the Ni-Ga-As phases. A Ni 2 . 4 GaAs hexagonal ternary phase of 1200 130oA thick resulted from in situ annealing of 65oA Ni films at 300°C, and -1400 A of NixGaAs plus NiAs phase resulted from anneal at 400°C. Using in situ annealing, the evolution of Ni 2 . 4 GaAs during subsequent reactions with Ge, Ti, or Ge/Ti was found to follow a complex route involving both transformation and decomposition. Upon annealing at 500°C, for 5 minutes and longer, Ni 2 . 4 GaAs was decomposed by Ti and Ge to form NiTi and NiGe. Simultaneously with the simple decomposition, Ni 2 . 4 GaAs transformed into Ni^As and N^.iGa binaries with incorporation of Ge from the outer layer. Both of these binaries containing Ge were decomposed during the annealing at 500°C for 5 minutes. 149

PAGE 157

150 In the samples of GaAs / Ni2,4GaAs / 250A Ge / 300A Ti and GaAs / Ni2.4GaAs / 500A Ge, a layer of GaAs -300 A thick GaAs regrew as a result of the decomposition of the binary phases. In samples of GaAs / Ni2.4GaAs / 750A Ge / 300A Ti, Ni-Ga-Ge ternaries plus NisGes, Ni5Ge2, and Ni2Ge binary phases were formed along with ~60oA of regrown GaAs. Simple decomposition ofNi2.4GaAs was a direct consequence of the formation ofNiGe compounds and NiTi with lower heat of formations. The transformation of Ni2.4GaAs into the Nii.sAs and Ni2.iGa binaries was the result of attempting to achieve thermodynamic equilibrium as predicted by the Ni-Ga-As ternary phase diagram. The NiGa-As phase diagram predicts only NiAs and Ni-Ga compounds should exist in equilibrium with GaAs. Thus the Nii^As and Ni2.iGa binary phases detected in this study were formed by releasing Ga and As from Ni2.4GaAs to be NiAs and Ni-Ga binaries. The possibility of Ni-As phases was evaluated using a calculated Ni-Ga-Ge-As quaternary phase diagram where Ni-As binaries were shown to be unstable based on NiGe-GaAs tie lines. The presence of Ti enhanced the phase equilibra predicted in the Ni-Ga-Ge-As system through Ni-Ti reaction. Decomposition of the ternary and binary phases produced 3 00 A 600 A of solid phase epitaxial regrown GaAs. The formation of regrown GaAs where Ge occupied Ga sites was explained by the decomposition of Nii.sAs binaries containing Ge in the Ni-Ga-Ge-As quaternary system. A high resolution EDS analysis confirmed the incorporation of Ge into the regrown GaAs, and Ge concentrations up to the mid-lO^'^cm'^ were measured in the regrown GaAs. From comparison between structures with and without in situ anneals.

PAGE 158

151 regrowth of GaAs from decomposition of the binary phases was the critical factor in the formation of ohmic contacts. Regrown GaAs was characterized by a high density of twins, stacking faults, precipitates, and incorporation of Ni. In some cases, regrowth smoothed the interfacial morphology between the metal/GaAs. Because a high density of precipitates were observed in regrown GaAs which exhibited rectifying contacts, microstructure of the regrown GaAs was found to be critical factor for rectifying behavior. All the metallurgical data, particularly the measured Ge concentrations (1.0 5.0 x 10^” cm’^ ), were correlated to the formation of ohmic contacts, and it was found that the mechanism for the formation of ohmic contacts was the formation of n^-GaAs above the original substrate doping, leading presumably to thermionic field emission transport, behavior. The doping process in the regrown (n^) GaAs was controlled by the Fermilevel throughout the entire solid. With a Ge concentration of -1.5 x lO^^cm'^, the concentration of free electron at the Fermi-level above the edge of the conduction band was approximated to be -4 x lO'^cm'^. This agrees with the observation of ohmic behavior.

PAGE 159

APPENDIX A. HEAT OF FORMATION Table A-I, Heat of formation of various compounds at 300K. Binary systems Compounds AHfor(KJ/mole) at 300K Ni-Ti NiTi -67.8 NiTi2 -80.3 NisTi -138.9 Ni-Ga NisGa -23.5 (-29.4) NisGa2 -35.8 (-38.2) NiGa -38.1 (-37.2) Ni2Ga3 -45.2 (-32.0) NiGa4 -22.2 (-16.1) Ni-Ge NiGe -31.8 (-32.9) NisGe (-24.3) Ni2Ge -110.1(-30.2) Ni3Ge2 (-32.8) NisGe2 (-27.2) Ni-As NiAs -73.3 (-46.7) NiAs2 -90.1* NisAs2 -251.1 NiiiAsg -773.1 Ge-As GeAs (-2.9) GeAs2 (-2.3) Ge-Ga No compounds Ga-As GaAs -74.1 * Stolyarova, enthalpy of formation of C 02 AS and NiAs 2 , p. 121 1, 1982 The values in parenthesis are calculated by using Miedema model [MieSO] Most of data listed in table A-1 came from the reference by Barin [Bar93], For several compounds, experimentally measured values could not be found. 152

PAGE 160

In these cases, a semi-empirical model by Miedema et. al. was used to estimate the heat of formation (or enthalpy) [MieSO], This model considers alloy cohesion by two materials constants for each element. Although this model is semi-empirical, these constants are introduced by consideration of the underlying physical forces thought to play a role in the cohesion of alloys. This model was used to predict several binary systems (such as Ni-Si, Mn-Ni) and shown to be remarkably accurate compared to experimentally measured values [Pre92] Using this model, the heat of formation for Ni-Ga, Ni-Ge and Ni-As binaries were calculated by: AHfo, = {24.6 ./c*).(CaVa^' + CbVb“) • (-AO^ + 9.4(Anws*'y -R/P)} / D where D is given by (n\s'^^^ + Ca and Va are atomic mole fraction and molar volume of atom A, respectively. fiO) is a constant, and O is the work function, and AO is the difference in work fijnction for the two elements. R/P is a constant equal to 1 .9 for Ni-Ga system and 2 for Ni-Ge, and 2.3 for Ni-As. nws is the electron density at the surface of the Wigner-Seitz cell. Table A-2 is a list of data necessary to calculate the heats of formation. Table A-2. Data for calculating AHfor for compounds in Miedema model. Metal Work function(O) (V) n Ilws (d.u)*^^ Vn,""' (cm^) Ni 5.20 1.75 3.5 Ga 4.10 1.31 5.2 Ge 4.55 1.37 4.6 As 4.80 1.44 5.2

PAGE 161

154 The heat of formation of GeAs binary phases was given by AHfor = {2 .y(c’) . (CaVa^ + CbVb^) • [-10.6AO^ + 99. 64(Anws / D where Q/P=9.4 for all systems, P=10.6, and Q=99.64,

PAGE 162

APPENDIX B. CALCULATION OF PHASE DIAGRAM At fixed temperature and pressure, the Gibbs phase rule predicts a maximum of three phases in equilibrium with any portion of the phase diagram. Regions of three-phase equilibrium are found by determining the stable two-phase tie lines, which in turn are established by calculating the Gibbs free energy of reaction at the point where two possible tie lines would cross [Bey84], The Gibbs free energy at a temperature T is given by AGt AH TAS where AH is the change in enthalpy during the reaction at the temperature T and AS the change in entropy. For reactions occurring in the solid-state, the change in heat capacities is for all practical purposes nearly always zero, due to the fact that the heat capacity of the products is approximately equal to the heat capacity of the reactants. This is the well known Neumann-Kopp rule [Kub79]. Following this idea, AGt = AH298 TAS298 for any temperature T. This argument imply that the standard values at T=300K of enthalpy and entropy can be used for thermodynamic calculations at any temperature. The change in enthalpy (or heat of formation) AH is a good measure of the change in free 155

PAGE 163

156 energy AG because the change in entropy AS is usually only about ± O.OOlkJ/deg per mole of atoms during solid state formation of ordered compounds [Pre91], Thus, for all reactions which occur in the solid-state, the Gibbs free energy can be approximated by the standard enthalpy of reaction alone, i.e. AGt ~ AH298 assuming that TAS 298 is small compared to AH298To calculate the Ni-Ga-Ge ternary phase diagram, the heats of formation listed in table A-1 was used. For example, the stable tie line between Ga and NiGe was determined by calculating the following series of reactions; Ni 2 Ga 3 + 2Ge = 2NiGe + 3Ga AHfor — -18.4kJ/mole, Ni 3 Ga 2 + 3Ge = 3NiGe + 2Ga AHfor = -59.6kJ/mole, N^Ga + 3Ge = 3NiGe + Ga AHfor ~ -71 .9kJ/mole. All of these reactions spontaneously proceed to the right. In other words, the tie line between NiGe and Ga is stable. By considering all possible cases, a Ni-Ga-Ge ternary phase diagram was determined and is shown in figure 5-3. Figure 5-4 was calculated using the same principle. The most important factor in determining the accuracy of this phase diagram is set by the quality of the thermodynamic data used for the calculation of tie lines. Since some of necessary data to construct the ternary phase diagrams was unavailable, the data tabulated in appendix A were used. A number of assumptions and simplifications have been made to calculate these diagrams [Bey84]. Use of these simplified ternary phase diagram will be confined to a very general description of trends of the reactions and phases in these systems.

PAGE 164

157 The quaternary Ni-Ga-Ge-As phase diagram was obtained by combining four ternary phase diagrams and is shown in figure 5-5.

PAGE 165

REFERENCES And72 And75 And78 And90 Bal89 Bar47 Bar95 Bar78 Bar82 Bar92 Bar93 Bau86 Ber72 Bey84 Andrew A M., N. Holonyak, Solid-State. Electron., Vol. 15, p.601, 1972. Andrews J.M., J.C. Phillips, Phys. Rev. Letts., Vol. 56, p.35, 1975. Anderson W.T. Jr., A. Christou, J.E. Davey, J. Appl. Phys. Vol. 49, p.2998, 1978. Anderson W.T. , A. Christou, ‘Semiconductor Device Reliablity’, Kluwer, Bonston, p.423, 1990. Ball R.K., Thin Solid Films, Vol. 176, p.55, 1989. Bardeen J., Phy. Rev. Vol. 71, p.717, 1947. BarafFG.A., M. Schluter, Phy. Rev. Letts., Vol. 55, p.l327, 1985. Barnes P.A., A.Y. Cho, Appl. Phys. Lett., Vol. 33, p.651, 1978. Barnett S.A., M.A. Ray, A. Lastras, B. Kramer, J.E. Greene, P.M. Raccah, L.L. Abels, Electronic Letter, Vol. 18, p.891, 1982. Barnes A., SPIE Journal, Vol. 1632, p.98, 1992. Barin I. ‘Thermochemical data of Pure Substance’, VCH, Weinheim, 1993. Bauer C.L., Surface Sci., Vol. 168, p.395, 1986. Berger H.H., Solid-State Electron., Vol. 15, p.l45, 1972. Beyers R., J. Appl. Phys., Vol. 56, p.l47, 1984. 158

PAGE 166

159 Bla74 Bla82 Bra67 BraSl Bri78 Bri90 Bru89 Buh91 Bur91 Cao89 Cha94 Che83 Che88 Chr79 Chr95 Blakemore J.S., ‘Solid State Physics’, Saunders, Philadelphia, 1974. Blakemore J.S., Solid-State Electron., Vol. 25, p. 1067, 1982. Braslau N, J.B. Gunn, J.L. Staples, Solid-State Elec., Vol. 10, p.381, 1967. BraslaueN., J. Vac. Sci. Technol., Vol. 19, p.803, 1981. Brillson L.J., J. Vac. Sci. Technol, Vol. 15, p.l378, 1978. Brillson L.J., S. Chang, J. Shaw, R E. Viturro, Vacuum, Vol. 41, p.1016, 1990. Bruce R.A., W.T. Moore, T. Lester, D.A. Clark, A.J. Springthorpe, Paper presented at Microscopy of Semiconducting Materials Conf., Oxford, 10-13, p. 671, 1989. Buhlmann H.J., M. Llegems, Electrochem. Soc., Vol. 138, p.2795, 1991. Burstein L., J. Bregman, Y. Shapira, J. Appl. Phys. Vol. 69, p.2312, 1991. Cao R., K. Miyano, T. Kendelewicz, I. Lindau, W.E. Spicer, Phy. Rev. B. ,Vol. 39, p. 12655, 1989. J.C.P. Chang, N. Otsuka, E.S. Harmon, M R. Melloch, J.M. Woodall, Appl. Phys. Lett., Vol. 65, p.2801, 1994. Chen L.J., Y.F. Hsieh, Proceedings of the 4T‘ Meeting of the Electron Microscopy Society of America, San Francisco, p.234, 1983. Chen S.H., C.B. Carter, C.J. Palmstrom, J. Mater. Res. Vol. 3, p.l385, 1988. Christou. A., Solid-State Electronics, Vol. 22, p.l41, 1979. Christianson K. A., C. Moglestue, W.T. Anderson, Solid-State Electronics, Vol. 38, p.1623, 1995.

PAGE 167

160 Chu94 Cow65 Dev80 EnnSl FreSO FreSl Fre90 GriSl Gud76 Gue89 Gui89 Gup 8 5 Gup90 Gyu71 Hai89 Chua S.-J., S.H. Lee, Jpn. J. Appl. Phys. Vol. 33, p.66, 1994. Cowley A.M., S.M. Sze, J. Appl. Phys., Vol. 36, p.3212, 1965. Devlin W.J., C.E.C. Wood, R. Stall, L.F. Eastman, Solid-State Electron, Vol. 23, p.823, 1980. Ennen H, U. Kaufmann, J. Schneider, Appl. Phys. Lett., Vol. 38, p.355, 1981. Freeouf J. L., Soild-State Comm., Vol. 33, p.l059, 1980. Freeouf J.L., J.M. Woodall, Appl. Phys. Lett, Vol. 39, p.727, 1981. Freeouf J.L., J.M. Woodall, L.J. Brillson, R.E. Viturro,. Appl. Phys. Lett. Vol. 69, p.56, 1990. Grimaldi M. G, B. M. Paine, M-A. Nicolet, D. K. Sadana, J. Appl. Phys. Vol. 52, p. 4038, 1981. Gudat W., D.E. Eastman, J. Vac. Sci. Technol., Vol. 13., p. 769,1976. Guerin R., A. Guivarc’h, J. Appl. Phys, Vol. 66, p.2122, 1989. Guivarc’h A., R. Guerin, J. Caulet, A. Poudoulec, J. Fontenille, J. Appl. Phys., Vol. 66, p.2129, 1989. Gupta R.P., W.S. Khokle, Solid-State Electron, Vol. 28, p.823, 1985. Gupta R., W.S. Khokle, J. Wuerfl, H.L. Hartnagel, J. Electrochem. Soc., Vol. 137, p.631, 1990. Gyulai J., J.W. Mayer, V. Rodriguez, Y.C. Yu, H.J. Gopen, J. Appl. Phys., Vol. 42, p.3578, 1971. Haigh D., J. Everad, ‘GaAs Technology and Its impact on Circuits and Systems’, Peter Peregrinus Ltd., London, UK, 1989. Hao P.H., L.C. Wang, F. Deng, S.S. Lau, J.Y.Cheng, J. Appl. Phys., Vol.79, p.4211, 1996. Hao96

PAGE 168

161 Hei65 Hei82 Hiz92 Hol91 Hol92 Hug92 IH83 IH84 Ili87 Kim86 Kim90 Kim96 Kir85 Kir87 K0I88 Heine V., Phys. Rev., Vol. A138, p.l689, 1965. Heiblum M., M.I. Nathan, C.A. Chang, Solid-State Electronics, Vol. 25, p.l85, 1982. Hizem N., G. Bremond, L. Mayet, M. Gavand, J. Gregoire, G. Guilot, W. Ulrici, Appl. Surf. Sci., Vol. 50, p.490, 1991. Holloway P.H., L.L-M. Yeh, D.H. Powell, A. Brown, Appl. Phys. Lett., Vol. 59, p.947, 1991. Holloway P.H., C.H. Mueller, Thin Solid Films, Vol. 221, p.254, 1992. Hugon M.C, B. Agius, F. Varniere, J. Appl. Phys. Vol. 72(8), p.3570, 1992. Iliadis A., K.E. Singer, Solid-State Electronics, Vol. 26, p. 7,1983. Iliadis A, K.E. Singer, Solid-State Communications, Vol. 49, p.99, 1984. Ilieadis A., J. Vac, Sci. Technol. B. Vol. 5, p. 1340,1987. Kim T., D.D.L. Chung, J. Vac. Sci. Technol. Vol. B4, p.762, 1986. Kim T., D.D.L. Chung, Phil. Mag. A. Vol. 62, p.283, 1990. Kim T.-J. ‘Evolution of Interfacial Phases and Their Effects on Ohmic Contacts to n-GaAs in Ni-Ge-Ti Metallizations’, Ph.D. dissertation, University of Florida, 1996. Kirchner P.D., T.N. Jackson, G. Dpettit, J.M. Woodall, Appl. Phys. Lett. Vol. 47, p.26, 1985. Kirillov D., Y. Chung, Appl. Phys. Lett. Vol. 51, p.846, 1987. Kolqwa E., C.W. Nieh, W. Flick, J. Molarius, M-A. Nicolet, Mat. Res. Soc. Proc. Vol. 126, p.289, 1988.

PAGE 169

162 Kov90 Kua83 Kub79 Kul80 Kul86 Kum76 Kum89 Kur69 Lam92 Lav86 Li92 Lil85 Leu85 Kovacs B., I. Mojes, R. Veresegyhazy, M. Nemeth-Sally, B.Pecz, Vacuum, Vol. 40, p.l29, 1990. Kuan T.S., P.E. Batson, T.N. Jackson, H. Rupprecht, E.L. Wilkie, J. Appl. Phys. Vol. 54, p.6952, 1983. Kubaschewski O., C.B. Alock, ‘Metallurgical Thermochemistry’, Pergamon Press, Oxford, 1979 Kular S.S., B.J. Paine, M-A. Nicolet, D.D. Sadana, J. Appl. Phys. Vol. 52, p.4083, 1980. Kulkarni A.K., J.T. Lukowski, J. Appl. Phys. Vol. 59, p.2901, 1986. Kumar K., Jpn. J. Appl. Phys., Vol. 18, p.713, 1976. Kumar N.S., J.-I. Chyi, C.K. Peng, H. Morkoc, Appl. Phys. Lett. Vol. 55, p.775, 1989. Kurtin S., T.C. McGill, C.A. Mead, Phys. Rev. Lett., Vol. 22, p.l433, 1969. Lamport W.V., ‘A Study of Aluminum-Germanium-Nickel Ohmic Contact Metallurgical Effects at the Gallium Arsenide Interface’, Ph.D dissertation. University of Florida, 1992. Lahav A., M. Eizenberg, Y. Komem, J. Appl. Phys. Vol. 60, p.991, 1986. Li B., P.H. Holloway, J. Appl. Phys. Vol. 71, p.85, 1992. Liliental-W. Z., N. Newman, W.E. Spicer, R. Gronsky, J. Washburn, E.R. Weber, Materisl Research Society Symposia proceedings, Vol. 54, p.415, 1985. Leung S., T. Yoshiie, C.L. Bauer, A.G. Milnes, J. Electrochem. Soc., Vol. 132, p.898, 1985. Lince J.R., C. Thomas, R.S. Williams, J. Mater. Res. Vol.l, p.537, 1986, Lin85

PAGE 170

163 Lin88 Lor94 Lou76 Lou95 Mar83 Mar87 Mat71 Mcl88 Meh89 Mie80 Mot38 Mon88 Mon93 Moh95 Mur89 Lin J.-C, X.-Y. Zheng, K.-C. Hsieh, Y.A. Chang, Mat. Res. Symp. Proc. Vol. 102, p.233, 1988. Loretto M.H., ‘Electron Beam Analysis of Materials’, 2"“* edition, Chapman&Hall, London, 1994. Louie S.G., M.L. Cohen, Phys. Rev. B, Vol. 13, p.2461, 1976. Lou Y.-S, C.-Y Wu, Solid-State Electronics, Vol. 38, p.l63, 1995. Marlow G.S., M.B. Das, L. Tongson, Solid-St. Electron, Vol. 26, p.259, 1983. Marshall E.D., B. Zhang, L.C. Wang, P.F. Jiao, W.X. Chen, T. Sawada, S.S. Lau, K.L. Kavanagh, T.F. Kuech, J. Appl. Phys., Vol. 62, p.942, 1987. Matare’ A., ‘Defect electronics in Semiconductor’, Wiley , New York, 1971, Chapter 3. McLean A.B., R.H. Williams, J.F. Megilp, J. Vac. Sci. Technol. Vol. B, 6(4), p.1252, 1988. Mehdi I., U.K. Reddy, J. Oh, J.R. East, G.I. Haddad, J. Appl. Phys., Vol. 65, p.868, 1989. Miedema A.R., P.F. de Chatel, F.R. de Boer, Physica lOOB, p.l, 1980. MottN.F., Proc. Camb. Philos. Soc., Vol. 34, p.568, 1938. Monch W, J. Vac. Sci. Technol. B6., p.l270, 1988. Monch W, ‘Semiconductor Surface and Interfaces’, SpringerVerlag, Berlin, 1993, Chapter 6. Mohney S.E., Y.A. Chang, J. Appl. Phys., Vol. 78, p. 1343, 1995. Murakami M., W.H. Price, J.H. Greiner, I D. Feder, J. Appl. Phys. Vol. 65, p.3546, 1989.

PAGE 171

164 Mye47 OgaSO OgaSOb Oku94 Pal90 Pat86 Pez86 Pio83 Pre91 Pre92 Pro89 Rel89 Reu81 Rid75 Rob75 San85 MyerhofM.E., Phys. Rev. Vol. 71, p.727, 1947. OgawaM., J. Appl. Pjys., Vol. 51, p.406, 1980. OgawaM., Thin Solid Films, Vol. 70, p.l81, 1980. Oku T., H. Wakimot, A. Otsuki, M. Murakami, J. Appl. Phys. Vol. 75, p.2522, 1994. Palmstrom C.J., S.A. Schwarz, E. Yablonovitch, J.P. Harbison, C.L. Schwartz, L.T. Florez, T.J. Gmitter, E.D. Marshall, S.S. Lau, J. Appl. Phys. Vol. 67, p.334, 1990. Patrick W., W.S. Mackie, S.P. Beaumont, C.D.W. Wilkinson, Appl. Phys. Lett, Vol. 48, p.986, 1986. Pecz B., E. Jaroli, G. Radnoczi, R. Veresegyhazy, I. Mojzers, Phys. Stat. Sol., Vol. 94, p.507, 1986. Piotrowska A., A. GuivarcÂ’h, G. Pelous, Solid-State Electronics, Vol. 26, p.l79, 1983. Pretorius R., A.M. Vredenberg, F.W. Saris, R. de Reus, J. Appl. Phys., Vol. 70, p.3636, 1991. Pretorius R., T.K. Marais, C.C. Theron, not published, 1992. Procop B., B. Sandow, H. Raidt, L.D. Son, Phys. Status Solidi A. Vol. 104, p.903, 1987. RellingE., A.P. Botha, Appl. Surf. Sci., Vol. 35, p.380, 1989. Reuter W., J.E.E. Baglin, J. Vac. Sci. Technol., Vol. 18, p.282, 1981. Rideout V.L., Solid-State Elec.,Vol. 18, p.541, 1975. Robinson G.Y., Solid-State Elec., Vol. 18, p.331, 1975. Sankey F., R.E. Allen, S.-F. Ren, J.D. Dow, J. Vac. Sci. Technol. Vol. B3, p.1162, 1985.

PAGE 172

165 San86 San87a San87b San88 Sch40 Sch78 Sch78 Sch90 Sch90 Scl88 She92 Shi87 Shi88 Smi94 Sands T., V.G. Keramidas, J. Washburn, R. Gronsky, Appl. Phys. Lett., Vol. 48, p.402, 1986. Sand T., V.G. Keramidas, A.J. Yu, K-M. Yu, R. Gronsky, J. Washburn, J. Mater. Res. Vol. 2, p.262, 1987. Sand T., V.G. Keramidas, K.M. Yu, J. Washburn, K. Krishnan, J. Appl. Phys. Vol. 62, p.2070, 1987. Sands T., E.D. Marshall, L.C. Wang, J. Mater. Res. Vol. 3, p.914, 1988. Schottky W., Physik. Z., Vol. 41, p.570, 1940. SchluterM., Phy. Rev. B, Vol. 17, p.5044, 1978. SchluterM., J. Vac. Sci. Technol. Vol. 15, p.l374, 1978. Schwarz S.A., C.J. Palmstrom, C.L. Schwartz, T. Sands, L.G. Shantharama, J.P. Harbison, L.T. Florez, E.D. Marshall, C.C. Han, S. S. Lau, L.H. Allen, J.W. Mayer, J. Vac. Sci. Technol. A, Vol. 8, p.2079, 1990. Schilfgaarde M.V., N. Newman, Phys. Rev. Letts., Vol. 65, p.2728, 1990. Sclater N., ‘GaAs IC Technology’, Tab books Inc., blue Ridge Summit, PA, 1988. Shen T.C., G.B. Gao, H. Morkoc, J. Vac. Sci. Technol. B, Vol. 10, p.2113, 1992. Shih Y.-C., M. Murakami, E.L. Wilkie, A.C. Callegari, J. Appl. Phys. Vol. 62, p.582, 1987. Shiota I., J. Nishizawa, Y. Oyama, J. Appl. Phys., Vol. 64, p.l 136, 1988. Smith G.C., ‘Surface Analysis by Electron Spectroscopy: Measurement and Interpretation’, Plenum Press, New York, 1994. Sol87 Solomon J.S., S.R. Smith, J. Vac. Sci. Technol. A, Vol. 5, p.l809, 1987.

PAGE 173

166 Spi76 Spi79 SpiSO Spi85 Spi88 Spi93 Sta81 Sze81 Tan92 Tan84 Ter84 Ter88 Tsu91 Van80 Spicer W.E., I. Lindau, P.E. Gregory, C.M. Gamer, P. Pianetta, P. Chye, J. Vac. Sci. Technol, Vol. 13, p.780, 1976. Spicer W.E., P.W. Chye, P.R. Skeath, C.Y. Su, I. Lindau, J. Vac. Sci. Technol., Vol. 16, p.l422, 1979. Spicer W.E., I. Lindau, P. Skeath, C.Y. Su, J. Vac. Sci. Technol, Vol. 17, p.1019, 1980. Spicer W.E., N. Newman, T. Kendelewicz, W.G. Petro, M.D. Williams, C.E. McCants, I. Lindau, J. Vac. Sci. Technol., Vol. B3, p.ll78, 1985. Spicer W.E., Z. Liliental-Weber, N. Newman, T. Kendelewicz, R. Cao, C. Mccants, P. Mahowald, K. Miyano, I. Lindau, J. Vac. Sci. Technol. B. Vol. 6, p.1245, 1988. Spicer W.E., A.M. Green, J. Vac. Sci. Technol., Vol. Bll, p.l347, 1993. Stall R.A., C.E.C. Wood, K. Board, N. Dandekar, L.F. Eastman, J. Devlin, J. Appl. Phy., Vol. 52, p.4062, 1981. Sze S. M., ‘Physics of Semiconductor Devices”, Jon Wiley & Sons, Inc, New York, 1981. Tanahashi K., J. J. Takata, A. Otuki, M. Murakami, J. Appl. Phys. Vol. 71, p.4183, 1992. Tang S.Y., W.N. Mei, G. Xu, J. Vac. Sci. Technol. B2, p.393, 1984. Tersoff J., Phys. Rev. Lett. Vol. 52, p.465, 1984. TersofifJ., ‘Metallization and Metal-Semiconductor Interfaces’, editied by I P. Batra, Plenum Press, New York, 1988. Tsuchimoto J., S. Shikata, H. Hayashi, J. Appl. Phys., Vol. 69, p.6556, 1991. Vandenberg J.M., E. Kinsbron, Thin Solid Films, Vol. 65, p.259, 1980.

PAGE 174

Vit93 Wal87 Wal88 Wal89 Wal93 Wan88 Wan95 Wil88 Wil89 Wit77 W008I Woo83 Zha91 Zur83 Yos83 Vitomirov I.M., A. Raisanen, S. Chang, R.E. Viturro, L.J. Brillson, D.R. Rioux, P.D. Kirchner, G.D. Pettit, J.M. Woodall, J. Electro. Mat Vol. 22, p.lll, 1993. Walukiewicz W., J. Vac. Sci. Technol. Vol. B5, p.l062, 1987. Walukiewicz W., Phy. Rev. B., Vol. 37, p.4760, 1988. Walukiewicz W., Appl. Phys. Lett, Vol. 54, p.2094, 1989. Walukiewicz W., Mat. Res. Soc. Symp. Proc. Vol. 300, p.421, San Francisco, 1993. Wang L.C., B. Zhang, F. Gang, E.D. Marshall, S.S. Lau, T. Sands, T.F. Kuech, J. Mater. Res. Vol. 3, p.922, 1988. WangL.C., J. Appl. Phys., Vol. 77, p.l607, 1995. Wilier J., D. Ristow, W. Kellner, H. Oppolzer J. Electrochem. Soc., Vol. 135, p.l79, 1988. Wilson R.G., ‘Secondary Ion Mass Spectroscopy’, Wiley, New York, 1989. Wittmer M., R. Pretorius, J.W. Mayer, M-A. Nicolet, Solid-State Electro., Vol. 20, p.433, 1977. Woodall J.M., J.L. Freeouf, G.D. Pettit, T. Jackson, P. Kirchner, J. Vac. Sci. Technol., Vol. 19, p.626, 1981. Woodal J.M., ‘Metal Contacts to III-V Semiconductors’, p.223, 1986 Zhang, S B., J.E. Northrup, Phy. Rev. Letts., Vol. 67, p.2339, 1991. Zur A, T.C. McGill, D.L. Smith, Phy. Rev. B, Vol. 28, p.2060, 1983. Yoshiie T., C.L. Bauer, J. Vac. Sci. Technol. A, Vol. 1, p.554, 1983.

PAGE 175

BIOGRAPHICAL SKETCH Tae-Jin Kim was bom on July 22, 1965, in Tae-Gu, South Korea, the second son of Duck-Man Kim and YoungJoo Chung. He has finished his secondary education in Seoul and entered Yonsei University in 1984. In 1988, he received the B.Sc. degree in metallurgical engineering. He studied crystal growth in the graduate school of Yonsei University and received the M.S. degree in 1990. From 1990 to 1991, he finished his military service as a second lieutenant in the Korean Army. In November of 1991, he was hired by Korea Institute of Science and Technology to develop a reflective layer in magneto-optic film. In 1992, he met Jin-Hee Hong and became her husband. In pursuit of higher academic achievement, he left Korea and became a graduate student in the Department of Materials Science and Engineering at the University of Florida in Gainesville, Florida, in fall of 1992. One year later, he joined Dr. HollowayÂ’s group and began studying ohmic contact metallizations to GaAs under the guidance of Dr. Holloway. He devoted the most valuable 4 years not only to learn academic knowledge but also to realize who he is. As a result of his endeavor, he received his Ph.D. in engineering in December 1996. 168

PAGE 176

I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Paul H. Holloway, Chairman Professor of Materials Science and Engineering I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Rolf E. Hummel Professor of Materials Science and Engineering I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Robert M. Park Professor of Materials Science and Engineering I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Kevin S. Jmies Associate/Professor of Materials Science and Engineering

PAGE 177

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