Growth of Low Work Function Materials Using Aerosol-Assisted Metalorganic Chemical Vapor Deposition

GROWTH OF LOW WORK FUNCTION MATERIALS USING AEROSOL-ASSISTED METALORGANIC CHEMICAL VAPOR DEPOSITION

YONG SUN WON

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

UNIVERSITY OF FLORIDA

2006

This document is dedicated to my deceased father.

ACKNOWLEDGEMENTS

The author appreciates Dr. Olga Kryliouk and Dr. Lisa McElwee-White for their

service as committee members and precious academic advice. He also thanks Dr. Tim

Anderson, his committee chair, for the encouragement and support all the time in every

situation. The author wants to express his gratitude to his colleagues too, Young Seok

Kim and Byung Jin Chun. It was a great pleasure and relief to have them besides the

author throughout his study. So many people helped the author as well; Dr. Chatu

Sirimanne and Seth Dumbree (from the Chemistry department) for the synthesis of ZrC

precursors, Laurel Reitfort (from the Chemistry department) for the synthesis of LaB6

precursors, Dr. Omar Bchir, Hiral Ajmera, and Andrew Heitsch (from the Dr. Anderson’s

group) for the LaB6 MOCVD growth, Dr. Venu Varanasi (from the Dr. Anderson’s group) for the ThermoCalc simulation, Dr. Adrian Roitberg (from the QTP group) for the

Gaussian simulation, Rob Holoboff (A&N Corporation), James Hinnant and Dennis

Vince (the staffs of Chemical Engineering department) for the construction of the

MOCVD system. Finally, the author would like to give all the credit for his achievements to his family.

iii

TABLE OF CONTENTS

ACKNOWLEDGMENTS ...... iii

LIST OF TABLES...... vii

LIST OF FIGURES ...... ix

ABSTRACT...... xii

CHAPTER

1 INTRODUCTION ...... 1

1.1 Field Emitter Arrays (FEAs) ...... 2 1.1.1 Gated Emitter (Spindt Emitter) ...... 2 1.1.2 Fowler-Nordheim Equation...... 2 1.1.3 Tip Failure ...... 3 1.1.4 Applications...... 4 1.2 Emitter Materials ...... 5 1.2.1 General Requirements for Emitter Tip Materials...... 5 1.2.2 Advantages of ZrC and LaB6 as Emitter Tip Materials ...... 5 1.2.3 Gated Si and Mo FEAs...... 5 1.2.4 Crystal Structure of ZrC ...... 7 1.2.5 Crystal Structure of LaB6 ...... 9 1.3 Equilibrium Thermodynamics of the Zr-C System ...... 10 1.3.1 Cystal Chemistry of Transition Metal Carbides and Nitrides ...... 10 1.3.2 Analysis of Themodynamic Properties and Phase Stability in the Zr-C System...... 13 1.3.3 Cohesive Properties and Vibrational Entropy of 3d Transition Metal Compounds ...... 18 1.4 Computational Thermochemistry ...... 21 1.4.1 Density Functional Theory and Effective Core Potentials ...... 22 1.4.2 Ensemble Properties and Statistical Themodynamics...... 23 1.5 Film Growth...... 25 1.5.1 Physical Methods...... 26 1.5.2 Chemical Methods...... 27 1.5.3 Aerosol-Assisted MOCVD (AA-MOCVD)...... 28 1.6 Thin Films and Bulk Materials for FEAs ...... 29 1.7 Summary...... 31

2 EQULIBRIUM ANALYSIS OF ZIRCONIUM CARBIDE CVD GROWTH ...... 39

2.1 Introduction...... 39 2.2 Calculations and Themochemical Properties...... 39 2.3 Results and Discussion ...... 40 2.3.1 Phase Change as a Function of Temperature and Pressure ...... 41 2.3.2 Phase Change as a Function of Temperature and the Inlet H/Zr Ratio...... 42 2.3.3 Phase Change as a Function of Temperature and the Inlet C/Zr Ratio ...... 44 2.3.4 The Addition of Chlorine to the System...... 45 2.4 Summary...... 47

3 STUDY OF PRECURSOR DECOMPOSITION USING COMPUTATIONAL THERMOCHEMISTRY ...... 53

3.1 Introduction...... 53 3.2 Experimental Methods...... 53 3.3 Decomposition of Tetraneopentyl and Tetrabenzyl Zirconium Precursors for the CVD of Zirconium Carbide ...... 54 3.3.1 Comparison of Decomposition Behaviors of ZrNp4 and ZrBn4...... 55 3.3.2 Initial Stage of the Decomposition of ZrNp4...... 56 3.3.3 Isobutene Cleavage...... 58 3.3.4 Summary...... 60 3.4 Decomposition of Alkyl- and Arylimido Precursors for CVD of Tungsten Nitride ...... 61 3.4.1 Overview on the Tungsten Nitride MOCVD Growth ...... 61 3.4.2 NMR Kinetics of Acetonitrile Exchange in 2 ...... 63 3.4.3 Optimized Geometries...... 64 3.4.4 Dissociation of Acetonitrile...... 65 3.4.5 Cleavage of W-Cl Bonds...... 67 3.4.6 Bond Dissociation Energies for W-N(imido) and N(imido)-C in Complexes 1-3 ...... 69 3.4.7 Interpretation of Positive Ion EI MS Data...... 72 3.4.8 Feasibility of Alkyl- and Arylimido Precursors for Tungsten Nitride ALD ...... 73 3.4.9 Summary...... 74

4 GROWTH OF ZrC THIN FILMS BY AEROSOL-ASSISTED MOCVD ...... 89

4.1 Introduction...... 89 4.2 AA-MOCVD System Description and Growth Procedure...... 90 4.3 Precursor Synthesis...... 93 4.3.1 Precursor Candidates...... 93 4.3.2 General ...... 93 4.3.3 Tetraneopentyl Zirconium (ZrNp4) ...... 93 4.3.4 Trineopentyl Zirconium Monochloride (ZrNp3Cl) ...... 94 4.4 Optimization of Growth Conditions ...... 94 4.4.1 Comparison with Equilibrium Analysis ...... 94

4.4.2 Summary of the Suggested Growth Conditions ...... 97 4.5 Film Charactrization ...... 97 4.5.1 General ...... 98 4.5.2 Structural Analysis ...... 98 4.5.3 Chemical Composition and Phase Determination...... 99 4.5.4 Surface Morphology and Film Thickness ...... 102 4.5.5 Magnetron Sputtering Growth for Comparison ...... 103 4.6 Summary...... 104

5 CONCLUSIONS AND RECOMMENDATIONS FOR FUTURE WORK...... 114

5.1 Conclusions...... 114 5.2 Recommendations for Future Work ...... 116

APPENDIX

A EXPLORATORY RESULTS ON THE GROWTH OF LaB6 THIN FILMS BY AEROSOL-ASSISTED MOCVD...... 118

B CHARACTERIZATION TECHNIQUES ...... 124

LIST OF REFERENCES...... 129

BIOGRAPHICAL SKETCH ...... 136

LIST OF TABLES

Table page

1-1. Physical properties of ZrC and LaB6 related to the field emitter application...... 38

1-2. Metal lattice and interstitial sites for Hagg’s 'Simple' crystal structures...... 38

2-1. ZrC films composition at different deposition conditions (P = 300 Torr and inlet C/Zr = 3856)...... 52

3-1. Bond lengths (Å) and bond angles (°) for ZrBn4...... 85

3-2. Atomic charges and Wiberg indices for ZrR4 complexes from NBO analysis ...... 85

3-3. Energeticies of possible pathways for ZrNp4 decomposition...... 85

3-4. Calculated bond lengths (Å) and bond angles (°) for complexes 1-3...... 86

3-5. Calculated bond lengths (Å) and bond angles (°) for complexes 1a-3a...... 86

3-6. Reaction enthalpies for CH3CN dissociation from complexes 1-3 ...... 86

3-7. Refined activation enthalpies and entropies for the first step σ-bond metathesis of 1a-3a...... 86

3-8. Bond dissociation enthalpies (∆Ho, kcal/mol) for the N1-C and W-N1 bonds in 1a-3a...... 87

3-9. Comparison of deposition behavior for 1a-3a...... 87

3-10. Bond dissociation enthalpy (∆Ho, kcal/mol) for the N1-C and W-N1 bonds in σ- bond metathesis products ...... 87

3-11. Summary of relative abundances for positive ion EI mass spectra of tungsten imido complexes, 2 and 3...... 87

3-12. Calculated dipole moments of 1a-3a and their derivatives via σ-bond metathesis..88

4-1. Bulk composition of ZrC films grown at various growth conditions...... 112

4-2. Summary of the suggested growth conditions for ZrC MOCVD growth using ZrNp4 ...... 113

vii

4-3. Composition with Ar+ sputtering time of ZrC film deposited on Si(111) substrate at 500 oC in He carrier and annealed at 860 oC for 30 min following a long period (> 3 months) of air exposure...... 113

A.1. Summary of the CVD results for the precursors evaluated for LaB6 deposition.....123

viii

LIST OF FIGURES

Figure page

1-1. Spindt type field emitter cathode array...... 32

1-2. Tip failure ...... 32

1-3. Schematic of an ion bombardment and nanoprotrusion model of FEA failure...... 32

1-4. Hierarchy of applications...... 33

1-5. Fabrication process (a) – (d) of Si gated field emitter...... 33

1-6. Process sequence of a Mo metal field emitter using isotropic silicon etching and oxidation...... 33

1-7. Rock-salt lattice structure of ZrC ...... 34

1-8. CsCl lattice structure of LaB6 ...... 34

1-9. A unit cell of the LaB6 (100) surface...... 34

1-10. Octahedral interstitial sites in fcc, bcc, and hcp structures and trigonal prism interstitial sites in the simple hexagonal structure...... 35

1-11. The entropy-related effective force constant kS for hcp Ti [111], various Ti carbides, hcp Zr [111] and the Zr carbides, as a function of the atomic fraction of carbon, xC, in the compound ...... 35

1-12. FIB micromachining of individual emitter apertures for small arrays ...... 36

1-13. Calculated Gibbs energy vs. temperature plots for LaB6 deposition reactions ...... 36

1-14. Schematic and image of ultrasonic nebulizing system...... 37

1-15. Transfer mold fabrication process ...... 37

1-16. SEM micrographs of gated transfer mold LaB6 FEAs, using sputtering for LaB6 deposition ...... 38

2-1. The reproduced Zr-C phase diagram...... 48

2-2. CVD phase diagram showing equilibrium phases as a function of temperature and pressure for three inlet H/Zr atom ratios [inlet C/Zr = 20]...... 48

2-3. CVD phase diagram showing deposited phases as a function of temperature and the inlet H/Zr ratio for four pressures [inlet C/Zr = 3856]...... 49

2-4. Deposition efficiency as a function of growth temperature. The insert shows 5 ZrZrC/Zrinlet [C/Zr = 3856, H/Zr = 10 , P = 300 Torr]...... 49

2-5. CVD phase diagram showing deposited phases as a function of the inlet C/Zr ratio and temperature at 0.1 and 1.0 atm ...... 50

2-6. CVD phase diagram showing deposited phases as a function of the inlet H/Zr ratio and temperature for two values of inlet Cl/Zr ratio [P = 300 Torr, C/Zr = 3856]...... 51

2-7. CVD phase diagram showing deposited phases as a function of the inlet H/Zr and Cl/Zr ratios [P = 300 Torr, T = 500 oC, C/Zr = 3856]...... 51

3-1. Optimized geometries for ZrNp4 (left) and ZrBn4 (right)...... 75

3-2. Initial decomposition step for ZrNp4 ...... 76

3-3. Energetics of the reaction scheme (3-3) and computed structures of 2b-t (a transition state for the second γ-hydrogen abstraction), 2b, and 3b...... 76

3-4. Optimized geometries for complexes 1-3...... 77

3-5. Optimized geometries for complexes 1a-3a...... 77

3-6. Assumed vibrational mode leading to CH3CN dissociation from 1...... 77

3-7. Standard Gibbs energy change (∆G°) vs. temperature for CH3CN dissociation from complexes 1-3 ...... 78

3-8. Plot of ln(k/T) vs. 1/T for acetonitrile exchange in complex 2 ...... 78

3-9. Calculated transition states (center) and products (right) for σ-bond metathesis of 1a-3a with hydrogen ...... 79

3-10. Calculated transition states and intermediates for σ-bond metathesis (solid arrows) and reductive elimination pathways (dotted arrows) ...... 80

3-11. Conjugation of N lone pair to metal d and through the phenyl ring as observed in the HOMO-18 ...... 81

3-12. Atomic charges (underlined) from NPA and Wiberg indices for 1-3 ...... 81

3-13. Geometries of the products of W-N1 and N1-C bond dissociation from 2a...... 82

3-14. Comparison of nitrogen content in the films grown from 1a-3a (AES) ...... 82

i + + 3-15. Two possible decomposition pathways of [Cl3W(N Pr)] or [Cl3W(NC3H5)] ...... 83

i + 3-16. Energetics of two possible pathways for the decomposition of (a) [Cl3W(N Pr)] + and (b) [Cl3W(NC3H5)] ...... 84

3-17. Energetics of two possible pathways for the decomposition of (a) 2a and (b) 3a ...84

4-1. Process flow diagram for the AA-MOCVD system...... 105

4-2. Images of the AA-MOCVD system ...... 106

4-3. Comparison of computed deposition phase diagram for ZrC deposition at 300 Torr by MOCVD using a ZrNp4 solution (0.0177 M in benzonitrile) in either H2 or He carrier gas...... 107

4-4. EDS spectrum of the ZrC film deposited in a helium environment at 500 oC...... 107

4-5. AES spectra and depth profiles of a ZrC film deposited on Si (111) at 400 oC with o H2 carrier gas followed by annealing in H2 at T = 860 C for 30 min...... 109

4-6. AES survey data for the ZrC films deposited on Si (111) substrates at 500 oC ...... 109

4-7. AES survey data with Ar+ sputtering time, for the ZrC film deposited on Si (111) substrates at 500 oC in He carrier and annealed at 860 oC for 30 min following a long period (> 3 months) of air exposure...... 110

4-8. XPS data for films deposited at 500 °C in He carrier; (a) C 1s peak and (b) Zr + 3d3/2, 3d5/2 peaks. Both spectra were measured after 3 keV Ar sputtering for 30

min. XPS depth profiling for the film deposited at 500 °C in H2 carrier (c)...... 111

4-9. AFM surface images (1 µm × 1µm) of the films deposited at: (a) T = 400 oC, (b) T = 500 oC, and (c) T = 600 oC ...... 111

4-10. X-SEM micrographs of ZrC films deposited in a helium environment at (a) 700 oC and (b) 600 oC ...... 111

4-11. XRD & GIXD (Grazed Incident X-ray Diffraction) spectra of a ZrC film deposited on Si (111) at room temperature by magnetron sputtering using argon; o as deposited and annealed H2 at T = 860 C for 30 min...... 112

A.1. Representative examples of La precursor compounds tested in this study...... 121

A.2. EI mass spectrum of precursor 1...... 122

A.3. Derivatized borohydride complexes from 3...... 122

A.4. Promising precursor candidates for LaB6 CVD growth...... 122

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

GROWTH OF LOW WORK FUNCTION MATERIALS USING AEROSOL-ASSISTED METALORGANIC CHEMICAL VAPOR DEPOSITION

Yong Sun Won

August 2006

Chair: Timothy James Anderson Major Department: Chemical Engineering

Low work function materials such as ZrC and LaB6, are promising candidates for

cathode or coating materials in field emitter arrays. An aerosol delivery system was used

to transport metalorganic precursors ZrNp4 and ZrBn4 to the growth zone of a cold-walled,

rf-heated, horizontal chemical vapor deposition reactor.

Equilibrium analysis of the Zr-C-H system identified growth conditions where

stoichiometric ZrC could be grown without carbon codeposition. A window exists that

includes a certain minimum amount of hydrogen carrier and an upper bound in

temperature that varies with the system pressure and inlet C/Zr atom ratio. ZrC films

were grown using tetraneopentyl zirconium (ZrNp4) as the precursor and the composition of the films as estimated by AES measurements was in qualitative agreement with the equilibrium expectations. XPS and AES spectra confirmed the existence of carbide or oxycarbide phases, while XRD results indicated all films were amorphous. Attempts to grow ZrC films using tetrabenzyl zirconium (ZrBn4) and LaB6 films using the precursors

xii

La(fod)3, La(thd)3, borohydride complex, tris(amido) complex were unsuccessful.

Computational thermochemistry studies of the reaction pathways between ZrBn4 and

ZrNp4 suggest that facile Zr-C bond cleavage occurs in ZrBn4 but not in ZrNp4. In addition, likely gas phase decomposition pathways for ZrNp4 were explored by DFT.

The results of these calculations show that the intermediate, Zr(H2C=CH2) should be an important surface reactant due to its empty coordination site.

As another example to demonstrate the general applicability of computational

thermochemistry to study decomposition mechanisms of metalorganic precursors,

possible reaction pathways for growth of WNx films from the isopropylimido complex

i Cl4(CH3CN)W(N Pr) (1), the phenylimido complex Cl4(CH3CN)W(NPh) (2), and the

allylimido complex Cl4(CH3CN)W(NC3H5) (3) were studied and compared to experimental results. The dissociation of the acetonitrile ligand (CH3CN) from 1-3

proved to be facile in the temperature range typically used for growth. The transition

states of σ-bond metathesis of the coordinatively unsaturated complexes 1a-3a with

hydrogen demonstrated the probable pathway for the subsequent chlorine cleavage.

Finally, bond dissociation energies for N(imido)-C and W-N(imido) bonds were

calculated to explain trends in nitrogen incorporation in films grown from 1-3, along with

qualitative natural bonding orbital analysis.

xiii CHAPTER 1 INTRODUCTION

Electron source technology based on thermionic emission has been used for more

than 30 years. Although these devices have demonstrated high reliability, there is a

strong demand for decreasing system response time, greater compactness, higher current

densities, and improved power efficiency, especially in military applications. Field

emission array (FEA) devices may be a viable alternative solution to these requirements.

Gated field emitter arrays can provide instant device turn-on, higher overall anode

current, more efficient frequency modulation, space/weight savings, lower operating

temperature, and lower power consumption [Sch95, Slu97, Oh98, Mat00]. Challenges,

however, still remain with regard to the reliability of FEA devices, such as the need to

maximize the current density and minimize sudden cathode failure due to vacuum arcing

[Goo94, Sch95, Cha98, Cha99, Cha01].

One of the methods used for increasing current density at relatively low operational voltage without cathode failure is fabrication of cathodes from low work function and hard materials or coating of those materials on conventional cathodes, such as silicon and molybdenum. It has been reported that certain materials, such as some carbides, borides, nitrides, and carbon-based thin films, have low work functions [Aiz94, Tra00]. By integrating these materials into a high brightness electron source, it may be possible to increase the emission current by a factor of 100 or more as compared to a device fabricated from silicon or molybdenum [Mac94, Mac98, Mac99, Kan00, Spi00,Yat00].

1 2

1.1 Field Emitter Arrays (FEAs)

1.1.1 Gated Emitter (Spindt Emitter)

The cathode is an addressable planar source of electrons. A number of different

techniques and materials have been developed for field emission applications, such as

gated tips (Spindt emitter), edge emitters, volcanoes, and diamond films. The most

commonly used structure is the gated emitter.

The gated emitter consists of a sharp tip, usually metal, which is fabricated to be

centered on an aperture in a metal electrode that forms the gate (see Figure 1-1). When,

in a vacuum, a sufficient voltage is applied between the tip and the gate, the electric field will cause electrons to emit from the tip surface towards the gate. Some electrons strike the gate, but others pass through the aperture into the vacuum. The voltage that is required to extract the electrons depends on a number of parameters, including the distance of the tip to the edge of the gate, the sharpness of the tip, the properties of the tip material, and the surface properties of the tip [Sch95]. Generally, the electric field necessary for electron emission to occur is reported to be 107 V/cm [Slu97]. The reciprocal dependency on the feature size implies that the smaller the gate aperture, the lower the voltage that is required to stimulate electron emission. Lower voltage is desirable for lower power and lower beam divergence [Slu97].

1.1.2 Fowler-Nordheim Equation

With the development of Sommerfeld’s theory of metals and quantum mechanics,

Fowler and Nordheim formulated the theory for the emission of electrons from metal surfaces under the action of intense electric fields and found that the current density J varies with the applied voltage V as

J = A()βV 2 exp(− Bφ 3 / 2 βV ) (1-1) where φ is the work function, A is a function of φ, B is a constant, and β is a geometrical factor dependent on the electrode configuration such that F = βV, where F is the magnitude of the electric field at the emitting surface. β is referred to as the field-voltage proportionality factor.

From the Fowler-Nordheim equation, it is clear that the electron emission current depends exponentially on both the electronic state (as modified by adsorbed gas species and so forth) and the geometry of the emitting surface, through φ and β, respectively. In other words, brighter emission is achieved with lower work function and greater β factor

[Sch95]. But a lower work function material coating on a tip contributes to a decrease of

β factor because it is proportional to 1/r, where r is the tip radius. Therefore, a process should be developed to uniformly coat the tip without significantly changing the tip radius. Among available techniques, atomic layer deposition (ALD) is considered as a promising one.

1.1.3 Tip Failure

It has been found that FEAs can be operated stably for hundreds or thousands hours and suddenly, for no apparent reason, suffer a voltage breakdown event that results in destruction of the cathode. An example in which primarily melting and vaporization of the gate film has occurred, exposing the underlying SiO2 film, and vaporizing the emitter tip, is shown in Figure 1-2 (a) [Sch95, Pol01].

It is believed that abrupt failure at a low current level is due to ion bombardment creating sharp nanoprotrusions, which result in intense FE current directed at the gate.

As illustrated in Figure 1-3, some ions formed by electron beam ionization are

4 accelerated and impact the field emitter surface. Rarely, the impact creates a sharp nanoprotrusion, which results in a large increase in the local field. At point A (see Figure

1-3), the local field is large enough to cause intense emission, which, in extreme cases, may initiate a vacuum arc. At point C, the substrate field is low enough that the enhanced emission is easily tolerated. The worst case is impact at point B, where the local field is high enough to cause intense emission. The intense emission is directed and focused at the gate, causing heating, desorption and ionization of gate surface contaminants, intense ion bombardment of the emitter, in a regenerative sequence of events that may culminate in a vacuum arc [Cha98, Cha99, Cha01].

Based on presented factors, there is an active search for FEA materials that may have better overall performance, stability, and life than Mo or Si. Especially, Mo FEAs demand an extremely high vacuum because of their vulnerability to ion bombardment, which attributes to their tip crystallinity ranging from polycrystalline to amorphous.

Even if it does not result in the whole cathode failure, ion bombardment makes the tip dull as shown in Figure 1-2 (b). This results in the reduction of β factor, proportional to the reciprocal of tip radius, and ultimately leads to long-term emission current degradation.

1.1.4 Applications

High current electron sources have a variety of applications (see Figure 1-4), especially in defense-related areas. Moreover, they have tremendous potential in propulsion applications of satellites and spaceships, although there are other technical challenges in these applications because emitter arrays are operated in a gas discharge or ambient plasma environment.

1.2 Emitter Materials

1.2.1 General Requirements for Emitter Tip Materials

The low work function and the high hardness of a cathode material are essential to guarantee the FEAs’ performance and reliability, as discussed in Sections 1.1.2 and 1.1.3.

The high melting temperature is another indicator of the hardness of the material. Of course, a high electrical conductivity is also a basic requirement. Table 1-1 lists several important physical properties of ZrC and LaB6, which are the subject of this study, in comparison with conventionally used silicon and molybdenum.

1.2.2 Advantages of ZrC and LaB6 as Emitter Tip Materials

ZrC and LaB6’s melting temperatures are among the highest of any material, along with their low work functions. These properties translate into robust field emitters, capable of withstanding high fields and/or high temperature without field build-up or cathode failure due to vacuum arcing. The advantages of ZrC and LaB6 for FEAs are summarized as follows [Cha99].

Physical passivation of the emitter surface: minimizes geometrical instabilities and formation of sharp nanoprotrusions under ion bombardment.

Chemical passivation: greatly reduces changes in work function caused by adsorption or desorption of contaminants. The physical and chemical passivation favors stable operation at higher current levels, at higher residual pressures, and/or for longer periods.

Reduced work function, therefore reduced gate voltage and increased beam conductance, and thus increased basic FE current limit.

1.2.3 Gated Si and Mo FEAs

Silicon and molybdenum are commonly employed in forming gated FEAs, based on their physical properties shown in Table 1-1 and their compatibility with procedures involved in the microfabrication process. A typical fabrication process for a gated Si

FEA is shown in Figure 1-5 [Mat00]. First, thermally grown oxide on silicon (100) is patterned into a 1.2 µm-diameter disk (a). Using the SiO2 disk as a mask, the outline of the emitter tip is formed by reactive ion etching (RIE) using SF6 and O2 gas mixture (b).

The sample is then again thermally oxidized at 1050 °C to sharpen the tip apex. After the oxidation for sharpening, a niobium layer (250 nm) is evaporated as the gate electrode

(c). Finally, the SiO2 disk and the niobium layer on the Si tip are removed by a lift-off technique with buffered hydrofluoric acid (BHF) (d).

The fabrication process for a gated Mo FEA using isotropic silicon etching and oxidation is shown in Figure 1-6 [Cha98]. Boron-doped (100)-oriented silicon is used as a starting material. A cathode electrode is formed by phosphorus diffusion. A 56 nm- thickness silicon nitride film is deposited using low pressure chemical vapor deposition

(LPCVD). Then, the nitride film is patterned into 2.25 µm-diameter disks using a conventional mask aligner and a reactive ion etcher (a). 400 nm of the silicon wafer is isotropically etched by RIE (b) for easier oxidation under the nitride disks. The exposed surface is then thermally oxidized to form a 820 nm-thick gate oxide layer (c). The nitride disk is removed in phosphoric acid. The exposed silicon is isotropically etched by

RIE using SF6 to form a gate hole (d). In these steps, the gate hole diameter is reduced from 2.25 to 0.57 µm. Thereafter, the conventional Spindt process is applied.

Gate metal is evaporated onto the wafer by an e-gun evaporator at a 90o angle with respect to the substrate (vertical) to avoid bridging between the gate and the cathode (e).

Aluminum is deposited using e-gun evaporation at a grazing angle to form a parting layer

(f). Then, the molybdenum deposition for the tip formation is performed (g). The parting

7 layer is selectively removed using KOH solution (h). As a result, the fabrication of gated

Mo FEAs is completed.

1.2.4 Crystal Structure of ZrC

ZrC has the covalent rock-salt (NaCl) structure as shown in Figure 1-7 with a room temperature lattice constant of 4.697 Å. Most of the transition metal monocarbides are stable in the NaCl structure, which can be viewed as the metal atoms occupying an fcc sublattice and the carbon occupying the octahedral interstitial sites. At 100% site occupancy, the stoichiometry of the carbide is MC1.0, although this situation is rarely realized.

This structure can be rationalized using Engel-Brewer theory of metals [Eng39,

Bre68], which states that the structure adopted by a metal or alloy depends on the s-p electron count [Eng39]. With increasing the s-p electron count, the metal structure progresses from bcc to hcp to fcc across the transition series. The Group IV (Zr:

4s24p64d25s2) and V metal carbides behave similarly. The stoichiometric metal carbide,

MC, forms in the NaCl structure rather than a hexagonal structure because the incompletely filled bands of the host metals can accommodate a high ratio of sp-electron- rich carbon to metal.

The nature of bonding in the monocarbides is a matter of some debate, although all agree that the simple ionic model (M+C- or M-C+) is not consistent with the properties of the carbides. Ionic materials will not typically slip on the {111} planes due to the strong repulsive (Coulombic) interactions across the shear plane in the half-glide position; instead, they will slip on the {110} or {100} planes [Oya92]. But plastic deformation occurs particularly via a mechanism of dislocation glide along {111} planes in transient metal carbides. Also, the exact direction of electron transfer is the subject of some

8 controversy. X-ray photoelectron spectroscopy (XPS) and electronegativity considerations suggest simple M→C electron donation, but the XPS data is questionable due to the possibility for back-donation or screening effects [Tot71]. Furthermore, simple M→C donation would result in ionic compounds such as the alkali and alkaline earth metal carbides. These materials, however, are resistors with low optical transparency or reflectivity, and readily hydrolyze to the metal oxide and hydrocarbon.

In contrast, the transition metal carbides are conductors with a shiny metallic and colored appearance and are hydrolytically stable. Band occupation suggests C→M electron transfer. Carbon appears to combine its sp electrons with the metal spd bands. Such a donation scheme may be used to explain, if crudely, the trend in melting point maxima, which occurs in Group VI for the metals, Group V for the carbides, and Group IV for the analogous mononitrides: the maxima may be associated with the half-filled d shell

[Oya92].

Group IV-VI carbides of the transition metals have extremely high melting temperature and are therefore referred to collectively as the 'refractory carbides'. In addition to their stability at high temperature, these compounds are extremely hard, finding industrial use in cutting tools and wear-resistant parts. Their hardness is retained to very high temperatures. In addition, they have low chemical reactivity – they are reactive only with concentrated acids or bases in the presence of oxidizing agents at room temperature, and retain good corrosion resistance to high temperature. The refractory carbides are strong, with Young’s modulus values – a measure of elastic deformation resistance – rivaling those of SiC at room temperature. In addition, they have good thermal shock resistance and good thermal conductivity, permitting heat to be drawn

9 away from the working surface of the tool. This gives them a benefit over other refractory materials, which do not conduct heat so well. The metal carbides share many characteristics with the metals themselves, having a plastic deformation like the fcc metals which, while lowering the high-temperature hardness, protects parts fabricated from the carbides from catastrophic failure in response to stresses.

1.2.5 Crystal Structure of LaB6

̀̀LaB6 has a rather complex crystal structure. Six boron atoms in each B6 octahedron are arranged as units in a body centered cubic (CsCl) lattice with the metal atoms; each octahedron is surrounded by eight metal atoms, and vice versa (see Figure 1-

8). The cage of octahedra linked together in all six directions constructs a firm but open framework and it is known that the metal atoms can be removed up to one quarter by evaporation without the collapse of the framework. The hardness of LaB6 is explained by this rigid skeleton of boron cage [Tho64].

The atypical structure of LaB6 also gives explanation on the low work function and high electrical conductivity of LaB6. The B6 octahedron requires 20 electrons by the octet rule, but six boron atoms have only 18. The hexaboride of a divalent metal should not thus be a conductor (e.g. pure alkaline earth hexaborides), but LaB6, with one free electron per lanthanum atom (confirmed by Hall effect measurements) is an excellent conductor. In other words, lanthanum atoms (5s25p65d16s2) transfer two 6s2 electrons to the boron cage and the remnant one 5d1 electron becomes free [Tho64]. The itinerant electron makes the La atoms charged positively and B6 octahedra negatively. Thus, the low work function of LaB6 (100) surface is reasonable in terms of the electropositive layer formation for the La-terminated surface (see Figure 1-9) [Mar98].

The LaB6 (100) surface has been shown to adsorb oxygen dissociatively at room temperature, which results in an increase in the work function by 1.4 to 1.6 eV. On the other hand, after exposing LaB6 to air, its high emission characteristics are readily regenerated by vacuum annealing at 1650 °C [Mar98]. The lattice constant is 4.156 Å

[She81] and the boron atoms are situated in + (x, ½, ½) where the coordinate x denotes the distance from the cell face, 0.88 Å for LaB6.

1.3 Equilibrium Thermodynamics of the Zr-C System

1.3.1 Crystal Chemistry of Transition Metal Carbides and Nitrides

Transition metal carbides and nitrides have structures where metal atoms are arranged in the close-packed or nearly close-packed way with nonmetal atoms (carbon and nitrogen) inserted into interstitial sites. There are no apparent localized interactions between nonmetal atoms for most structure. Borides, however, do clearly exhibit localized interactions between boron atoms in their structures as discussed in Section

1.2.4, including chains, layers, or three-dimensional networks (e.g. boron cage in LaB6).

Carbides and nitrides are featured with the metal-nonmetal interaction and the geometry of the interstitial site. Carbon and nitrogen are generally placed in either an octahedral interstitial site or in the center of a trigonal prism. Figure 1-10 demonstrates the types of interstitial sites in fcc, bcc, hcp, and simple hexagonal structures. The whole structure can be easily reconstructed by a structural unit (coordination polyhedron) composed of the interstitial atom and its nearest metal neighbors [Tot71].

The interstitial model is the easiest to visualize and the most well-known. Metal atoms in carbides and nitrides often form fcc, hcp, or simple hexagonal substructures; filling all the octahedral sites in fcc generates the B1 (NaCl) structure. The metal structure is also explained by the sequential ordering of the close-packed metal-atom

11 layers with the usual designations: –ABCABC– for fcc, -AB,AB- for hcp, and –AA– for simple hexagonal. Carbide and nitride structures have the simple notations –AXBXCX",

AXBX'CX" – for B1, –AXBX,AXBX– for L3', and –AX,AX– for the WC type, where X represents a complete or partial filling of the interstitial sites (octahedral or trigonal prism) created by two adjacent metal layers. X' and X" differ from X layer by a lateral shift [Tot71].

According to Hagg’s empirical rule presented in 1931, the structure of transition- metal carbides, nitrides, borides, and hydrides is determined by the radius ratio

r = rX / rMetal , where rX and rMetal are the radius of the interstitial and transition metal atom, respectively. If r < 0.59 (for ZrC, r = 0.483), the metal atoms have very simple structures: A1 (fcc), A2 (bcc), A3 (hcp), or simple hexagonal. If r > 0.59, the transition metal and interstitial nonmetal atoms form complex structures. When r < 0.59, the nonmetal atoms are accommodated in the largest interstitial sites of the relatively simple metal host structure. To provide sufficient bonding between the metal and nonmetal atoms, the interstitial sites should be smaller than the interstitial atom. On the other hand, too much smaller interstitial site than the interstitial atom (nonmetal atom) results in an unstable structure because of the expanded metal host lattice by the presence of a relatively larger interstitial atom and the weaker metal-metal interactions [Tot71].

Table 1-2 listed the types of interstitial sites in simple metal structure (see also

Figure 1-10). The tetrahedral interstitial sites of the simple metal structures are too small to accommodate carbon and nitrogen, and only the octahedral and trigonal prism interstitial sites are thus occupied. The B1 (NaCl) structure is very common among monocarbides and mononitrides, where all the octahedral interstitial sites in a fcc metal

12 lattice are occupied by carbon or nitrogen. A random occupation of half of the interstitial sites in the hcp metal lattice results in the L3' structure, which is common among the

Me2C and Me2N structures. Occupation of the trigonal prism interstitial sites in the simple hexagonal structure produces the WC structure [Tot71].

According to Andrews and Hughes [Tot71], there is a remarkable relationship between the crystal structure of the pure metal and of its carbides and nitrides. The crystal structure of the metal element changes upon forming its carbide and nitride. The metal having an hcp structure will not form a carbide or nitride in which the metal atoms are hexagonally arranged. The metal having a fcc structure will not form a carbide and nitride in which the metal atoms are cubically arranged. The metal having the bcc structure and no close-packed allotropes will form carbides and nitrides in which the metal atoms are both hexagonally and cubically arranged. These relationships are not only based on size relationships, but on the stabilizing effect of the interstitial in a particular coordination polyhedron and metal-metal interactions. For example, Zr has bcc and hcp metal structures and thus Zr atoms are cubically, not hexagonally, arranged in its carbide and nitride [Tot71].

There are two types of interstitial sites in the bcc structure. The tetrahedral site situated at (½, ¼, 0) and equivalent sites are the larger ones, and a distorted octahedral site at (0, 0, ½) and (½, ½, 0) and equivalent sites are the smaller ones. Since these sites are hard to be occupied by the interstitial nonmetal atoms, carbon and nitrogen, the metal atoms form a fcc or hcp substructure to provide larger octahedral site. The small size of the distorted octahedral site in bcc also accounts for the limited solubility of carbon and nitrogen in many IV- to VI- group elements [Tot71].

1.3.2 Analysis of Thermodynamic Properties and Phase Stability in the Zr-C System

Equilibrium analysis is a very useful tool to anticipate the deposition behavior and thus estimate near optimal growth conditions. While the thermodynamic properties of gas phase species are well established experimentally, the solid properties are not always well known and thus depend on theoretical approaches to suggest values. Throughout this section and the following section, the methodologies used to obtain the thermodynamic properties in the Zr-C system and evaluate its phase stability will be discussed and summarized. There is general agreement on the existence of five stable/metastable condensed phases in the Zr-C system [Gui95]. These phases are the following (see also Figure 1-10):

1) Zr-rich solid solution, hcp (hP2) Zr, α phase; METASTABLE

2) Zr-rich solid solution, bcc (cI2) Zr, β phase; METASTABLE

3) (cF8) NaCl-type structure non-stoichiometric carbide, ZrCx, with x ≤ 1, γ phase; STABLE

4) Hexagonal (hP4) graphite modification of carbon

5) Liquid phase.

Gibbs Energy of Interstitial Phases (α, β, γ).

The Gibbs energy of the interstitial phases, α, β, and γ, is described using a two- sublattice model (see Section 1.3.1, which describes the 2 sublattices in these phases). Zr atoms are assumed to occupy the first sublattice whereas C atoms and vacant interstitial sites (Va) are assumed to substitute for each other on the second one [Gui95].

Accordingly, the hcp (α), bcc (β) and ZrCx (γ) phases of the Zr-C system are represented by the two-sublattice model (Zr)1(C, Va)c (c = the number of interstitial sites per metallic

φ atom, i.e., carbon occupancy), and the Gibbs energy Gm (φ = phase, m = mixture of two sublattices) per mole of formula unit is represented by the expression

φ 0 φ 0 φ E φ Gm = yC GZr:C + yVa GZr:Va + cRT(yC ln yC + yVa ln yVa )+ Gm (1-2)

The variable, yi (i = C,Va), which has been referred to as the site fraction of i , measures the fraction of the available sites that are occupied by the component i . The

0 φ 0 φ quantity GZr:Va is the Gibbs energy of Zr with the structure φ(φ = α, β,γ ) and GZr:C represents the Gibbs energy of a state where all the interstitial sites are filled with C atoms. All 0Gφ values are referred to the enthalpy H of reference states recommended by the Scientific Group Thermodata Europe organization. The stable element reference

(SER) state is defined as the stable state of the elements at 298.15 K and 105 Pa.

The last term in Eq. (1-2), which represents the excess Gibbs energy of the φ phase, is described by using the subregular approximation of the Redlich-Kister phenomenological power series, e.g., a first order model:

E φ 0 φ 1 φ Gm = yC yVa [ LZr:C,Va + LZr:C,Va (yC − yVa )] (1-3)

In Eqs. (1-2) and (1-3), the comma separates components that interact in the same sublattice and the colon separates components in different sublattices. The composition- independent parameters Lφ in Eq. (1-3) account phenomenologically for the interaction between C atoms and vacant interstitial sites. They are allowed to vary with temperature according to

0 φ 0 φ φ φ φ 2 LZr:C = AZr:C,Va + BZr:C,VaT + CZr:C,VaT lnT + DZr:C,VaT (1-4)

1 φ 1 φ φ φ φ 2 LZr:C = AZr:C,Va + BZr:C,VaT + CZr:C,VaT lnT + DZr:C,VaT (1-5)

In the treatment of the interstitial phase γ, the five coefficients in Eqs. (1-4) and (1-

5) were estimated from experimental data. The remaining interstitial phases are treated

1 φ φ using the regular solution approximation, by setting the parameters L , CZr:C,Va and

φ DZr:C,Va equal to zero.

Stoichiometric ZrC (cF8) (γ Phase).

The γ carbide phase has c = 1 and it is written as (Zr)1(C, Va)1 according to the

0 γ two-sublattice model. In this case, Eq. (1-2) contains the parameter GZr:Va , the Gibbs

0 γ energy of fcc Zr, and GZr:C , the Gibbs energy of the stoichiometric (cF8) NaCl-type

0 γ SER structure carbide ZrC. Information about the function, GZr:Va − H Zr , is taken from

0 γ Refs. [Gui87, Sau88], whereas the quantity, GZr:C , is referred to the enthalpy of the elements Zr and C in their reference states as

0 γ SER SER 0 γ GZr:C − H Zr − H C = ∆ Gm (T ) (1-6)

0 γ and the temperature-dependent function ∆ Gm

0 γ 2 3 −1 −3 ∆ Gm = a + bT + cT lnT + dT + eT + fT + gT . (1-7)

All parameters describing the Gibbs energy of formation of ZrC were determined by analyzing experimental data on the thermochemical properties and equilibrium boundaries of the non-stoichiometric ZrCx carbide by Guillermet [Gui95].

For example, values of the enthalpy of formation of the Zr carbides have been obtained from calorimetric measurements and from studies on the vaporization behavior of the ZrCx phase. The calorimetric values are obtained by combining measurements of the so-called heat of combustion of ZrCx, i.e. the enthalpy change of the reaction

ZrC x ()s + (x +1)O2 (g)→ ZrO2 (s)+ xCO2 (g) with values for the enthalpy of formation of ZrO2(s) and CO2(g) according to the reactions

Zr()hcp + O2 (g)→ ZrO2 (s) and

C()graph + O2 (g)→ CO2 (g) at 298.15 K and 105 Pa.

Interstitial Solution Based on hcp Zr and the ZrC0.5 (hP3) (α Phase).

The arrangement of metallic atoms in hcp Zr yields one octahedral interstitial site per atom. By adopting the approximation used in modeling the solubility of C and N in the hcp phase of Ti and Co, it is assumed that two neighboring interstitial sites in the c direction are never simultaneously occupied. This phase is thus treated with the two-

0 α sublattice model (Zr)1(C, Va)0.5. The parameter GZr:Va represents the Gibbs energy of

0 α hcp Zr. Its function is taken from the study by Guillermet [Gui87]. GZr:C represents the

Gibbs energy of the ZrC0.5 carbide based on hcp Zr, which is metastable in the Zr-C system. This quantity is referred to the enthalpy of the elements as in Eq. (1-6), with the

0 α corresponding ∆ Gm (T ) function represented as in Eq. (1-7) [Gui95].

Interstitial Solution Based on bcc Zr and the ZrC3 (β Phase).

For the interstitial solution of C in bcc Zr, call the β phase, the two-sublattice form

0 β is (Zr)1(C, Va)3. In this case, Eq. (1-2) involves the parameter, GZr:Va , i.e. the Gibbs

0 β energy of bcc Zr [Gui87] and GZr:C , which refers to the carbide in which all octahedral interstitial sites of the bcc structure of Zr are filled with C atoms. This carbide has the

0 β formula ZrC3 and is metastable in the Zr-C system. GZr:C is referred to the enthalpy of

0 β the elements in their reference states (Eq. (1-6)), and the ∆ Gm (T ) function is described as in Eq. (1-7) [Gui95].

Graphite.

0 graph SER The Gibbs energy of graphite is described by using GC − H C function assessed by Gustafson [Gut86].

Liquid Phase.

The Gibbs energy of the liquid phase is described by adopting a substitutional solution model as follows for a binary system:

liq 0 liq 0 liq E liq Gm = xC GC + xZr GZr + RT(xC ln xC + xZr ln xZr )+ Gm (1-8)

where xi (i = Zr,C) is the atomic fraction of the component i in the liquid phase. The

0 liq 0 liq quantities GC and GZr are taken from [Gus86] and [Gui87], respectively, and the

E liq excess Gibbs energy term Gm is treated by using the Redlich-Kister power series, i.e.

E liq 0 liq 1 liq 2 liq 2 Gm = xC xZr [ LC,Va + LC,Zr (xC − xZr )+ LC,Zr (xC − xZr ) ] (1-9)

In Eq. (1-9), 0 Lliq is allowed to vary linearly with temperature, but 1Lliq and 2 Lliq are treated as constants [Gui95].

Lacking direct measurements, the parameter aφ , bφ , cφ , d φ , eφ , f φ and g φ of

Eq. (1-7) (with φ = α, β ), which describe the temperature dependence of Gm for the metastable ZrC0.5 and ZrC3 carbides, are determined by estimation, as explained in

Section 1.5.3.

1.3.3 Cohesive Properties and Vibrational Entropy of 3d Transition Metal Compounds

For a single harmonic oscillator, the frequency ω is related to the mass M and the force constant k by ω = k / M .

d 2 x k − kx = m where x = Dsin(ct + e), c 2 = , and 2πν = c (1-10) dt 2 m

In a real solid, a spectrum of vibrational frequencies exists. The entropy at high temperature measures the logarithmically averaged frequency. Therefore, an entropy-

2 S ˆ related effective force constant is defined by k S = M eff (k B Θ D ()T / h) , which contains information on the strength of the average interatomic forces in the compound.

1/()1+x 1/(1+x) M eff ( = ()M Zr ()M C , for ZrCx) is the logarithmically averaged mass of the

S compound, and Θ D is an entropy Debye temperature, i.e., the Θ value that gives the

S experimental vibrational entropy Svib (T ) if Θ D is inserted in the Debye-model

expression S D for the entropy,

S Svib ()T = S D (Θ D ()T /T ). (1-11)

0 φ Svib (T ) is then directly used to evaluate the temperature dependency of ∆ Gm in

(1-7), given that non-vibrational and non-harmonic contributions are negligible. It is

S generally true at low temperature (T < Θ D ). Therefore. if kS is available, the

0 φ temperature dependency of ∆ Gm can be calculated by the following relationship;

S 0 φ kS → Θ D → Svib (T )→ ∆ Gm (T ).

Despite the lack of experimental data especially for metastable ZrC phases (α and

β), the evaluation of thermodynamic properties of these phases by Guillermet, in fact,

relied on remarkably accurate empirical relations between ne and ES [Gui95]. The

2 / 3 characteristic energy ES is related to kS and the volume per atom, Ω ; ES = kS Ω .

The average number of valence electrons per atom, ne , is calculated by

ne = (3nM + 2nX )/ 5 for a compound M 3 X 2 as an example. Here nM and n X are the number of valence electrons for atoms M and X .

0 S kS and θ S(Stable Value for Θ D) for Zr Carbides.

By applying interpolation and extrapolation procedures based on the assumption of

a smooth variation of the cohesive properties with ne , it has been possible to derive kS

S (and Θ D ) values for various stable and metastable compounds. Figure 1-11 plots kS vs.

the carbon atomic fraction, xC , in the carbides of Zr and Ti. This figure shows the kS

value extracted from the entropy of hcp Ti [111] and of TiC, and the kS values for the metastable carbides Ti3C2, Ti7C3, Ti5C2, Ti3C and Ti23C6 predicted by Guillermet and

Grimvall [Gui92]. From Ti carbide data, it is seen that ne (or xC ) and kS (or

2 / 3 ES = kS Ω ) have a linear relation and thus gives confidence it will also hold for the Zr carbides [Gui95].

Thus, once kS values are available (for example, values of Ti [111] and TiC from

experiments) vs. ne (or xC ), the unknown kS values for other related metastable compounds such as Ti3C2, Ti7C3, Ti5C2, Ti3C and Ti23C6 can be predicted.

The information about Zr compounds, however, is meager. The kS values for hcp

-1 0 Zr [111] ( kS = ~ 160 N/m) and ZrC (cF8) ( kS = 429 Nm , corresponding to θ S = 675

K) are plotted in Figure 1-11. Lacking additional information, the kS values of ZrC0.5

and ZrC3 are estimated by assuming that the variation in kS with xC for Zr carbides is similar to that shown by the results on Ti carbides. That variation is indicated in Figure

-1 -1 1-16 using a broken line, which gave kS = 429 Nm for ZrC0.5 and kS = 429 Nm for

0 ZrC3. From these results, the corresponding θ S values are obtained, i.e. 505 and 998 K, respectively [Gui95].

Temperature Dependence of θS and Entropy Functions for Zr Carbides.

S The temperature dependence of Θ D (T ) for the Zr carbides is estimated by applying

S a scaling relation. Values of θ D for ZrC0.5 and ZrC3 at various T are derived by

0 S 0 0 combining their θ S values with an estimate of their Θ D (T )/θ S vs. T/θ S function based on assuming that this function is the same as evaluated for ZrC (cF8) from the entropy and neglecting electric or magnetic contributions.

S The Θ D (T ) functions are used to calculate Svib (T ) values at various temperatures.

Then non-vibrational contributions are neglected and those values are treated as approximations of the total entropy of Zr carbides and used in evaluating the temperature dependent part of ∆0Gφ ()T (φ = α, β ). The coefficients, bφ , cφ , d φ , eφ , f φ and g φ , are obtained in this way by Guillermet [Gui95]. The following (Kubaschewski

polynomial as an example) gives the idea of how the coefficients in CP , H , S ,G are related with each other.

m C = m + m T + 5 + m T 2 (1-12) P 3 4 T 2 6

m m m H = m + m T + 4 T 2 + 5 + 6 T 3 (1-13) 1 3 2 T 3

m T 2 S = m + m lnT + m T − 5 + m (1-14) 2 3 4 2T 2 6 2

T 2 m m G = m − m T + m T ()1− lnT − m − 5 T − 6 T 3 (1-15) 1 2 3 4 2 2 6

Enthalpy of Formation of Metastable Zr Carbides.

Another useful finding of recent work on the cohesive properties of 3d transition metal carbides and nitrides is that the standard enthalpy ∆0 H (298.15) of formation at

298.15 K, expressed per mole of atoms in the compound varies smoothly with ne

[Gui95]. By using this fact, it is possible to derive values of ∆0 H for various stable and metastable carbides and nitrides of Sc, Ti, V, Cr, Mn, Fe, Co, and Ni. Since the complex, metastable carbides of metals from the 4d transition series have not been studied, the

0 enthalpy of formation of ZrC0.5 and ZrC3 are estimated by combining the ∆ H (298.15)

0 value for ZrC (cF8) with the assumption that the ∆ H (298.15) vs. xC function for Zr carbides is similar to that obtained for Ti carbides [Gui92]. By searching for the best fit to the set of input data, it is possible to determine the aφ coefficient of the polynomial in

Eq. (1-7) for each carbide.

1.4 Computational Thermochemistry

The thermodynamic properties implemented in a conventional thermodynamic database such as the ThermoCalc database [Sun85, Bar89, Din91, Kub96, NIST99] are not available for most metal organic precursors and their fragments (especially for a novel precursor). Fortunately, most metalorganic precursors are not thermodynamically stable at typical deposition temperatures and thus it is often appropriate to input the precursor stoichiometry but not include the metalorganics in the species list.

Computational thermochemistry, however, can be used to estimate thermodynamic

22 properties, as well as reaction kinetic parameters. The approach used in this study is based on the density functional theory (DFT) and statistical thermodynamics.

1.4.1 Density Functional Theory and Effective Core Potentials

The theoretical basis of density functional theory (DFT) and effective core potential

(ECP) basis set are not discussed here [Cra02]. Instead, the rationale for their selection will be justified. The major advantage of DFT is its computational efficiency. A rule of thumb for the scaling behavior of computational model chemistries is as follows; HF ≈

N4, MP2 ≈ N5, CISD ≈ N6, MP4 ≈ N7, CISDT ≈ N8 [Cra02]. The formal scaling behavior of DFT has been noted to be in principle no worse than N3, where N is the number of basis functions. Thus DFT and especially hybrid DFT (e.g. B3LYP used in this study) are more efficient, showing mean unsigned errors almost equal in quality to the much more computationally expensive multilevel correlated methods. The utility of

DFT in computing bond strengths between transition metals and hydrides, methyl groups, and methylene groups has also been demonstrated. Because of non-dynamical correlation problems associated with the partially filled metal d orbitals, such binding energies are usually poorly predicted by MO theory methods (such as HF). Moreover, it is known empirically that DFT is generally more robust in dealing with open-shell systems where HF methods show high spin contamination. Thus DFT is more appropriate for the calculations involving transition states or radicals [Cra02].

Heavy elements in the periodic table add a purely technical hurdle in that such elements have large numbers of electrons, and there is thus a concomitant requirement to use a large number of basis functions. The solution to this problem is to replace the electrons with analytical functions that would reasonably, accurately, and much more efficiently represent the combined nuclear-electron core to the remaining valence

23 electrons. Such functions are referred to as effective core potentials (ECPs) or pseudopotentials. Among them, the most widely used pseudopotentials are those of Hay and Wadt (sometimes also called the Los Alamos National Laboratory (or LanL) ECPs.

In this study, LanL2DZ ECPs will be used for transition metals [Cra02].

1.4.2 Ensemble Properties and Statistical Thermodynamics

Statistical thermodynamics is the basis of the procedure to convert single-molecule potential energies calculated from electronic-structure calculations to ensemble thermodynamic properties. The first step is to obtain the electronic energy (Eelec) for the stationary point on the Born-Oppenheimer PES (Potential Energy Surface) and the corresponding local minimum geometry by quantum mechanical calculations. Secondly, within the harmonic oscillator approximation, all molecular vibrational frequencies and the zero-point vibrational energy (ZPVE) are defined by the sum of all hω/2 energies over all molecular vibrations. Because the lowest vibrational energy level for any bound vibration is not zero even at 0 K, the internal energy at 0 K for a molecule is defined as

mod es 1 E0 = Eelec + ∑ hωi . (1-16) i 2 where h is Plank constant.

Then, statistical thermodynamics is introduced. Just as there is a fundamental function that characterizes the microscopic system in quantum mechanics, i.e., the wave function, similarly in statistical thermodynamics there is a fundamental function having equivalent status, and this is called the partition function. For the canonical ensemble, the partition function, q(V,T), is related by the following established thermodynamic definitions and assuming that the ensemble is an ideal gas, all of which are true and can be easily applied to obtain other thermodynamic properties.

⎛ ⎛ ∂ ln q ⎞ ⎞ S = R⎜ln()q()V ,T e + T⎜ ⎟ ⎟ (1-17) ⎝ ⎝ ∂T ⎠V ⎠

2 ⎛ ∂ ln q ⎞ E = Nk BT ⎜ ⎟ (1-18) ⎝ ∂T ⎠V where N is Avogadro’s number, and kB is Boltzmann constant.

The ensemble partition function, q(V,T), is composed of the contributions from translational, electronic, rotational, and vibrational molecular motions. In other words, it is the product of the molecular partition functions as

q()V ,T = qt qe qr qv . (1-19)

Equations used in the evaluation of each contribution are summarized below.

Contribution from Translation.

3 / 2 ⎛ 2πmk BT ⎞ k BT qt = ⎜ ⎟ (1-20) ⎝ h 2 ⎠ P

3 3 S = R()ln q + 5/ 2 , E = RT , C = R (1-21) t t t 2 t 2 where m is molecular mass, and R is gas constant.

Contribution from Electronic Motion.

−ε0 / RT −ε1 / RT −ε 2 / RT qe = ω 0e + ω1e + ω 2e +L (1-22) where ω is the degeneracy of the energy level, and εn is the energy of the n level.

Here, the first excitation energy is assumed to be much greater than kBT.

Therefore, the first and higher excited states are assumed to be inaccessible at any temperature. Further, the energy of the ground state is set to zero. These assumptions simplify the electronic partition function to:

qe = ω 0 , Se = R lnω 0 , Ee = Ce = 0 (1-23)

Contribution from Rotation.

1 ⎛ T ⎞ q = ⎜ ⎟ for a linear molecule (1-24) r ⎜ 2 2 ⎟ σ r ⎝ h /8π Ik B ⎠

S r = R()ln qr +1 , Er = RT , Cr = R (1-25)

Contribution from Vibration.

Each of the 3natoms – 6 ( or 3natoms – 5 for linear molecules) modes has a

characteristic vibrational temperature, Θv,K = hν K / k B .

e −Θv,K / 2T qv,K = for a given vibrational mode (1-26) 1− e −Θv,K / 2T

e−Θv,K / 2T qv,K = the overall vibrational partition function (1-27) ∏ −Θv,K / 2T K 1− e

Θ /T ⎛ v,K −Θv,K / T ⎞ SV = R ⎜ − ln(1− e )⎟ (1-28) ∑⎜ Θv,K / T ⎟ K ⎝ e −1 ⎠

⎛ 1 1 ⎞ EV = R Θv,K ⎜ + ⎟ (1-29) ∑ Θv,K / T K ⎝ 2 e −1⎠

2 ⎛ Θv,K /T ⎞ CV = R e ⎜ ⎟ (1-30) ∑ ⎜ Θv,K / T ⎟ K ⎝ e −1⎠

The details are given elsewhere [Och99, Och00, Cra02].

1.5 Film Growth

The performance of promising emitter materials (ZrC and LaB6 in this study) has generally been verified by testing bulk single emitters [Mac94, Ota96, Cha98, Cha99,

Mac99, Mac00, Cha01]. Single emitter fabrication starts with single crystal material grown for example by floating zone crystal growth. Then, the emitter tip is formed

26 through electrochemical etching (for example, in a 10% perchloric / 90% acetic acid solution for ZrC).

The coating or thick film deposition of emitter materials for the practical application of FEAs is accomplished by both physical and chemical methods. The former includes e-beam evaporation, pulsed laser deposition (PLD), and magnetron sputtering, while the latter includes chemical vapor deposition (CVD), metal organic

CVD (MOCVD), and laser-induced solution deposition (LISD). Generally, physical methods give better control of the deposition behavior and the stoichiometry of the deposited film. On the other hand, chemical methods have advantages in relatively moderate operating conditions (low temperature and high pressure) and uniform coating coverage of small features.

1.5.1 Physical Methods

Fabrication of ZrC-tip FEAs has been tried using physical vapor deposition (PVD) of emitter cones of ZrC [Mac99, Cha01]. Starting with a Si wafer with a 1 µm thick SiO2 layer, a gate layer was formed from a 150 to 250 nm Mo film deposited on the oxide using an e-beam heated PVD source. Processing proceeded by using a focused ion beam

(FIB) system to micromachine holes in the Mo and SiO2 layers. A buffered oxide etch was used to complete the formation of the blank array as shown in Figure 1-12. Then, a lift-off layer of Al2O3 was deposited at a shallow angle while rotating the blank substrate.

Cone deposition followed with e-beam evaporation of a stoichiometric, zone-refined, crystalline carbide target. The deposition was normal to the substrate and again progressed while the substrate was rotated. This carbide evaporation continued until the

27 emitter aperture had been closed. Finally wet chemical etching of the sacrificial lift-off layer completed the carbide emitter array.

The work function reduction effect of a ZrC or LaB6 coating on various substrates

(Si, Mo, W) has been investigated primarily with physical growth techniques, such as pulsed laser ablation [D’Al00], by evaporation [Tes93, Mac95, Mac98, Mac99, Cha01,

Spi00], and RF magnetron sputter deposition [Wal95].

Physical methods usually require high operating temperature (for example, 3250 K in evaporation) and UHV (for example, 10-9 Torr range in evaporation) [Cha01, Mac95], while PLD is performed in low deposition temperature.

1.5.2 Chemical Methods

Compared to physical methods, a few reports of ZrC and LaB6 deposition using chemical methods (especially, for LaB6) have appeared [Smi93, Par94, Gir98, Hea94,

Khe98, Spe94].

Although atmospheric halide CVD using zirconium tetrachloride (ZrCl4), methane

(CH4), and hydrogen precursors gave reasonable quality of ZrC film, growth requires a relatively high temperature (1000 to 2000 °C), thus limiting the list of suitable substrates

[Duc85, Gla99, Bla91]. Alternatively, the single source metal organic tetraneopentyl zirconium (Zr[CH2C(CH3)3]4, or ZrNp4) has been successfully used to grow thin films of

ZrC in the temperature range 300 to 750 °C. The films, however, were not stoichiometric, but contained excess carbon in an approximate Zr/C ratio 1:2 to 5.

Moreover, metal organic chemical vapor deposition (MOCVD) of ZrC thin films from

-4 ZrNp4 has been reported only at relatively low pressure (< 10 Torr) because ZrNp4 is less-volatile [Smi93, Par94, Gir98, Hea94].

CVD of high quality polycrystalline LaB6 thin films has been achieved by vacuum copyrolysis of the boron hydride clusters (BnHm species) with LaCl3 at 800 to 900 °C

[Khe98]. Thermodynamic calculations (see Figure 1-13) on CVD of LaB6 [Pes00] show that a boron halide like BBr3 is a good candidate for a boron source with its low toxicity and non-explosive behavior, but these boron halides still pose safety concerns. As plotted in Figure 1-13, the Gibbs energy of the reaction with BBr3 (reaction 3) is negative at for T > 1300 K (~ 1000 °C). Although this equilibrium temperature limit is higher compared to use of boranes (reactions 4 to 6), safety concerns exist with using the boranes. CVD of metal boride films using metal borane cluster compounds has been reported [Spe94], however, most attempts using a lanthanum single source precursor has been unsuccessful because of the insolubility and especially non-volatility of the prepared precursors.

Selective area laser-induced solution deposition (LISD) of LaB6 has been reported recently [Zho00]. The mechanism of LISD clearly resembles that of CVD, but deposition from solution is compatible with thin film formation on thermally sensitive substrates because of the large thermal sink of the solvent/solute mixture.

1.5.3 Aerosol-Assisted MOCVD (AA-MOCVD)

The low volatilities of Zr and La metal organic precursors pose a limit on their usage in CVD growth, as mentioned in Section 1.5.2. Thus, a different approach,

'aerosol-assisted MOCVD (AA-MOCVD)', will be adapted in this study to lift the volatility limit. The details are explained in Section 4.2.

Aerosol deposition, including spray deposition, begins with generation of aerosol from a liquid precursor. That is then directed towards a heated substrate where it

29 undergoes decomposition and/or chemical reaction near the vicinity of the heated substrate to produce stable and solid adherent coatings with good adhesion [Kor99].

Aerosol methods bridge many superb features of gas-phase and liquid-phase techniques and have the following advantages: (a) a low cost and low reservoir temperature, (b) simple and flexible equipment without the use of sophisticated reactor or vacuum system,

(c) universal precursors without the limit of low volatility, and (d) easy and precise composition control [Kor99]. With the proper solvent available, the precursor solution can be ultrasonically nebulized to produce micrometer and sub-micrometer size droplets with a narrow size distribution, which is an aerosol. The aerosols may be generated by pneumatic atomizers and ultrasonic atomizers, as illustrated in Figure 1-14 [Bch03a].

1.6 Thin Films and Bulk Materials for FEAs

ZrC coating on Si FEAs has been successfully applied with an apparent 35% decrease in the surface work function (4.82 → 3.15 eV) [Tes93, Mac95, Mac98, Cha99], far below the work function of ZrC itself (4.0 eV). In the case of LaB6, it has been reported that the measured thermionic surface work function is approximately 2.8 eV for the magnetron sputtered LaB6 coating on tungsten and molybdenum substrates [Wal95].

ZrC and LaB6, however, have sufficiently low work functions and mechanical hardness to serve as a tip material, although coatings of these materials have a reported work function reduction, compared to the bulk material. In addition, thick film growth is somewhat easier than thin film growth. A practical obstacle is that similar etch processes as in the fabrication of Si FEAs (see Figure 1-5) are not available for ZrC and LaB6.

Therefore, a totally different approach such as the 'Transfer Mold Technique' might be applied as illustrated in Figure 1-15 [Bye98, Nak03]. In this process, a Si (100) substrate is anisotropically etched through a thermally oxidized SiO2 mask using 30%

KOH aqueous solution to make pyramid holes with very sharp corners, i.e., 'Mold' (a). In contrast to the conventional fabrication method of Si FEAs using Si anisotropic etching, the etching process is automatically stopped. Thus, it is easy to fabricate multiple molds uniformly and reproducibly. The widths of the openings usually range from 1.0 to 3.0

µm. After removing the SiO2 mask, the Si molds are thermally oxidized to form an emitter-to-gate high quality insulator layer. The resistivity of the thermally oxidized SiO2 layer is twice or three times as high as that of the deposited SiO2 insulators. During the thermal SiO2 layer growth, the SiO2 layer shapes on sidewalls of the molds become convex. Subsequently, an emitter material such as ZrC or LaB6 is deposited on the SiO2 layer (b). Then, the emitter material is bonded to a glass substrate having an Al rear surface electrode by applying DC power at several hundred Volts (electrostatic bonding)

(c). Next, an anisotropic Si etchant (tetramethyl ammonium hydroxide (TMAH) solution) and a SiO2 etchant (buffered HF solution) are used to remove the Si mold substrate and the SiO2 layer (d). Thus, the emitter array is transferred from the Si substrate to the glass substrate. Since emitter tips are intrinsically sharpened in the process, there is no need to sharpen them after their formation [Bye96, Nak02].

To make a gated emitter array, the following processes are required. Gate and resist layers are coated on the SiO2 layer, which is not removed in the gate fabrication process (e). The thin resist layer is dry etched in an oxygen plasma to reveal only the tips of the coated emitters (f). Finally, wet etchings of both the gate and the oxide (g and h) result in sub-micron openings in the gate center on the emitters, as shown in Figure 1-16

[Nak02].

The transfer mold emitter fabrication technique has been developed to solve problems resulting from conventional approaches. For example, the popular technique of coating low work function emitter materials on the conventional emitters such as Mo and

Si emitters causes reduction of emitter tip radii and non-uniformity of emitter shapes, and has difficulty in emitter material evaluation. In the transfer mold fabrication, the typical emitter tip radii are typically less than 10 nm because of the tip sharpening effect produced by Si mold surface oxidation and subsequent etching.

1.7 Summary

The desirable properties of ZrC and LaB6 for application on FEAs were reviewed.

Most reports on fabricated devices use the films grown by physical methods. The low volatility of Zr and La metalorganic precursors has been an obstacle for CVD growth of

ZrC and LaB6 films. An aerosol-assisted technique has been selected for this work to remove the volatility limit. As preliminary work to the deposition studies, both an equilibrium analysis and computational reaction mechanistic studies were performed.

The thermodynamic properties of the Zr carbide and the liquid phase were reviewed while an equilibrium analysis to suggest growth parameters of ZrC by MOCVD growth is provided in Chapter 2. Computational thermochemistry based on DFT is useful to probe decomposition pathways of metalorganic precursors by extracting both thermodynamic and reaction kinetic parameters quantitatively. Calculations for both ZrC and WN precursors are presented in Chapter 3.

(a)

(b) (c)

Figure 1-1. Spindt type field emitter cathode array [Sch95]. (a) Schematic diagram of an array. (b) Scanning electron micrograph of an array. (c) Scanning electron micrograph of a single emitter in the array shown above.

( a ) ( b )

Figure 1-2. Tip failure. (a) Arcing damage to tips and gate electrodes [Pol01]. (b) Dulling of emitter tip due to ion bombardment [Pol01].

Figure 1-3. Schematic of an ion bombardment and nanoprotrusion model of FEA failure [Cha99]

High > 10 A/cm2 Modulation Emission Gated Amplifiers Radar Communications Electronic Warfare

Space-based Applications (Thrusters, Tethers, Discharge, etc.)

Electron Sources X-ray Sources Electron Beam Lithography Flat Panel Displays

Mass Spectrometers Scanning Electron Microscopes (SEMs) 2 Chemical Analysis Spectrometers < 10 A/cm Low or No Modulation

Figure 1-4. Hierarchy of applications [Jen01]

SiO 2 RIE Wet oxidation Nb BHF etching Nb deposition Lift-off

(a) (b) (c) (d)

Figure 1-5. Fabrication process (a) – (d) of Si gated field emitter [Mat00]

Figure 1-6. Process sequence of a Mo metal field emitter using isotropic silicon etching and oxidation [Cha98]

4.697 Å Zr

Figure 1-7. Rock-salt lattice structure of ZrC

4.156 Ǻ

Figure 1-8. CsCl lattice structure of LaB6

Figure 1-9. A unit cell of the LaB6 (100) surface [Mar98]. La atoms are shown by large spheres and B atoms by small spheres. The first subsurface plane of B atoms is shifted by 0.88 Å with respect to the outermost plane of La atoms.

fcc (ZrC-γ) bcc (ZrC -β) 3

hcp (ZrC0.5-α, half-filled) Simple hexagonal

Figure 1-10. Octahedral interstitial sites in fcc, bcc, and hcp structures and trigonal prism interstitial sites in the simple hexagonal structure

Figure 1-11. The entropy-related effective force constant kS for hcp Ti [111], various Ti carbides, hcp Zr [111] and the Zr carbides, as a function of the atomic fraction

of carbon, xC , in the compound. The numbers refer to (1) Ti23C6, (2) Ti3C, (3) Ti5C2, (4) Ti7C3, (5) Ti2C, (6) Ti3C2, and (7) TiC [Gui95].

Mo gate layer

Oxide insulator

(a) Si substrate (b)

Figure 1-12. FIB micromachining of individual emitter apertures for small arrays. (a) Blank array after FIB processing [Mac99, Cha01]. (b) Finished blank array after buffered oxide etch.

1500 1 2 1000 3

500 6 5 ) l 4

mo 0 kJ/ ( T -500 G ∆

-1000

-1500

-2000 0 500 1000 1500 2000 Temperature (K)

1. LaCl (g) + 6BCl (g) + 10.5H (g) = LaB (s) + 21HCl(g) 3 3 2 6 2. LaBr3(g) + 6BBr3(g) + 10.5H2(g) = LaB6(s) + 21HBr(g) 3. LaCl3(g) + 6BBr3(g) + 10.5H2(g) = LaB6(s) + 3HCl(g) + 18HBr(g) 4. LaCl3(g) + 3B2H6(g) = LaB6(s) + 3HCl(g) + 7.5H2(g) 5. LaCl3(g) + 1.2B5H9(g) = LaB6(s) + 3HCl(g) + 3.9H2(g) 6. LaCl3(g) + 0.6B10H14(g) = LaB6(s) + 3HCl(g) + 2.7H2(g)

Figure 1-13. Calculated Gibbs energy vs. temperature plots for LaB6 deposition reactions [Pes00].

Precursor aerosol and carrier gas to CVD reactor

Precursor aerosol

Vibrating quartz plate Dissolved precursor 1/16" from syringe pump Capillary tubing Carrier gas Cable to power to nebulizer supply

Figure 1-14. Schematic and image of ultrasonic nebulizing system [Bch03a].

Resist

Mold Anisotropic Gate etching (a) Gate formation Conduction Layer (e) Emitter Emitter SiO material (b) 2 deposition Resist etching Electrode (f)

Glass

Electrostatic bonding Gate (c) etching (g)

Si SiO2 etching etching (d) (h)

Figure 1-15. Transfer mold fabrication process [Nak02].

Figure 1-16. SEM micrographs of gated transfer mold LaB6 FEAs, using sputtering for LaB6 deposition [Nak02].

Table 1-1. Physical properties of ZrC and LaB6 related to the field emitter application

Micro-hardness Melting point Resistivity Work function Materials Color (N/mm2) (°C) (Ω-cm) (eV) ZrC Gray 25500 3400 50×10-6 4.0 -6 LaB6 Purple 25600-26000 2770 20× 10 2.74 Si Dark gray 12400 1410 0.1 4.52 Mo Silvery white 2600 2620 5× 10-6 4.6 Data are taken from references [Kie52, Tho64, Ord89, Aiz94, Kun96, CER97, Tra00].

Table 1-2. Metal lattice and interstitial sites for Hagg’s 'Simple' crystal structures

Metal- Minimum Maximum Metal atomic Interstitial atomic Metal Structure Interstitial site nonmetal r for r for positions positions neighbors occupancy occupancy 0 0 0, 0 ½ ½ ½, ½ ½ ½ ½, ½ 0 0 fcc Octahedral 6 0.41 0.59 0 ½, ½ ½ 0 0 ½ 0, 0 0 ½

±¼ ¼ ¼, ±¼ ¾ ¼ Tetrahedral 4 0.23 - ¾ ¼ ¾, ¾ ¾ ¾

hcp 0 0 0, ⅓ ⅔ ½ Octahedral ⅔ ⅓ ¼, ⅔ ⅓ ¾ 6 0.41 0.59

0 0 ⅜, ⅓ ⅔ ⅞ Tetrahedral 4 0.23 - 0 0 ⅝, ⅓ ⅔ ⅛ ±0 ½ ¼, ±½ 0 ¼ bcc 0 0 0, ½ ½ ½ Tetrahedral ±½ ¼ 0, ±0 ¼ ½ 4 0.29 - ±¼ 0 ½, ±¼ ½ 0 Simple hexagonal 0 0 0 Trigonal prism ±⅓ ⅔ ½ 6 0.53 0.59 c/a = 1

CHAPTER 2 EQUILIBRIUM ANALYSIS OF ZIRCONIUM CARBIDE CVD GROWTH

2.1 Introduction

To anticipate the effect of operational parameters on the compositions of the films, chemical equilibrium calculations were carried out using the ThermoCalc software to evaluate the CVD phase diagram [Tot71, Gui89, Gui95]. The equilibrium phases and their compositions were computed at 300 Torr as a function of deposition conditions

(growth temperature, inlet molar ratios, and carrier gas) with a constant basis of tetraneopentyl zirconium (ZrNp4) solution (0.0177 M in benzonitrile) and thus fixed C/Zr inlet atom ratio to C/Zr = 3856.

Equilibrium conditions, however, are not often encountered during CVD as a result of mass transfer or reaction kinetic limitations. If mass transfer limited growth conditions exist and the gas phase mass transfer coefficients are not significantly different, then trends in the intensive properties at the growth interface approach the equilibrium ones.

Thus the phase constitution and compositions can be qualitatively predicted by an equilibrium analysis, although the extensive quantities such as growth rate are not.

2.2 Calculations and Thermochemical Properties

The thermodynamic properties for gas species were excerpted from the existing

ThermoCalc database [Sun85, Bar89, Din91, Kub96, NIST99], while the Zr-C database assessed by Guillermet was used to define solid and liquid phases [Gui89, Gui95]. As mention in Section 1.3.2, the following five stable/metastable condensed phases were considered in his analysis; hcp (α) phase with formula (Zr)1(C, Va)0.5, bcc (β) phase with

39 40

formula (Zr)1(C, Va)3, fcc (γ) phase with formula (Zr)1(C, Va)1, graphite, and liquid solution. All gas phase species, provided by the ThermoCalc database, were included in the calculation. In this calculation, the only species with partial pressure ≥ 10-8 atm were

H2, CH4, HCl, ZrCl4, C2H6.

The Compound Energy Model (CEM) was used to describe the solution thermodynamics of the first three interstitial phases, while the liquid phase was treated as a random solution using the substitutional model for the excess Gibbs energy. The CEM model parameter estimation was based on the regular behavior of the vibrational entropy and other cohesive properties that had been established in previous studies on transition metal compounds [Gui89, Gui95]. The details are presented in Section 1.5.

Thermochemical equilibrium diagrams were computed in this work using the

ThermoCalc computational software package [Sun85]. The results were generalized by selecting atomic ratios as independent variables normalized by the inlet Zr atomic content. The calculation basis for conventional MOCVD using ZrNp4 was Zr : C : H = 1

: 20 : 44 in addition to variable carrier H2. For aerosol assisted MOCVD, the basis changed considerably to Zr : C : H = 1 : 3856 : 2784 plus carrier H2 as a result of the inclusion of the organic solvent benzonitrile (PhCN) at 0.0177 mol/l [Bch03a]. Droplets of the precursor/solvent are transported in and sprayed onto the surface of a heated substrate using a feed rate of solution of 0.067 sccm and H2 carrier flow rate of 1.5 slm.

2.3 Results and Discussion

Figure 2-1 reproduces the Zr-C phase diagram proposed by Guillermet [Gui89,

Gui95] to verify the database and to assist in the discussion of the results. The extent of

the ZrCx homogeneity range is evident in this phase diagram. It is noted that all data points in the figures in this chapter are calculation data points, not experimental ones.

2.3.1 Phase Change as a Function of Temperature and Pressure

The temperature-pressure deposition phase diagram is shown in Figure 2-2 for three values of H/Zr atom ratio (103, 104, and 105) at a fixed inlet C/Zr = 20 (i.e., only

ZrNp4 and H2 sources). It is clear that single phase ZrC deposition is thermodynamically favorable at lower temperature and the extent of the single phase region increases with higher pressure and higher H/Zr ratio. The carbon resides at equilibrium in ZrC and gas phase organics (predominantly CH4), and under some conditions as graphite. Since the formation of ZrC from graphite and Zr is thermodynamically favorable in the temperature range of this study, single phase ZrC formation is possible with excess carbon in the system. The key is to identify conditions that retain the excess carbon in volatile organic species. Considering only competition between graphite and CH4, a higher pressure and higher H/Zr ratio drive reaction (2-1) towards methane formation at fixed C/Zr. As the temperature increases, the negative entropy change of reaction (2-1) becomes more important and the equilibrium shifts to favor graphite formation.

C (graphite) + 2 H2 ↔ CH4 (2-1)

Considering that the typical growth temperature of LP-MOCVD is in the range 300 to 750 °C, carbon codeposition in the films is predicted in the range of conditions shown in Figure 2-2. Reaction kinetic limitations, however, are likely at low temperature. It is reported that a temperature of ~ 300 °C is required to initiate α-hydrogen abstraction

(reaction 2-2), which is an important step for neopentane elimination leading to Zr = C

42 bond formation [Wu96]. Consistent with this kinetic limitation, growth of ZrC using

ZrNp4 in AA-MOCVD at less than 400 °C did not result in film growth.

ZrNp4 → Np2Zr = C + CMe4 (2-2) (t-Bu)

For comparison, Girolami et al. reported deposition of amorphous TiC with (Ti :

C) = (1 : 0.93), at as low as 150 °C from tetraneopentyl titanium (TiNp4) using low pressure MOCVD [Gir87]. At temperature higher than 300 °C, the Ti/C ratio of the film was found to be about 1 : 2.5, indicating that carbon codeposition occurred [Smi93]. This suggests that the activation energy of α-hydrogen abstraction for ZrNp4 is much higher than for TiNp4.

2.3.2 Phase Change as a Function of Temperature and the Inlet H/Zr Ratio

To increase the temperature and deposit single phase ZrC, a combination of increasing the reactor pressure and H/Zr ratio is needed. The deposition rate must also be increased, which will likely occur at high temperature. Also use of the aerosol delivery system will provide higher precursors partial pressure to the growth surface. This of course is at the expense of the Zr/C ratio. Figure 2-3 shows the equilibrium T-H/Zr deposition diagram for a fixed C/Zr = 3856 at four reactor pressures. The CVD phase diagram shows that ZrC (FCC) deposition is possible at higher temperature by increasing the pressure and maintaining a minimum amount of hydrogen to prevent codeposition of elemental carbon.

To validate the results of the calculations and compare them to experimental data, a set of films was deposited by AA-MOCVD at the conditions summarized in Table 2-1.

The detailed description of the deposition conditions and film properties will be presented in Chapter 4, but the phase constitution is of interest here. The reactor pressure (P = 300

Torr) and inlet C/Zr ratio (= 3856) were maintained constant while the temperature was varied in the range from 400 to 600 °C. A solution of ZrNp4 dissolved in benzonitrile

(0.0177 mol/l) was used as a precursor and delivered at the rate of 0.067 sccm. In three experiments the temperature was varied, while the H2 flow rate was decreased by a factor of 2 in the fourth growth run.

Chemical compositions of the films were estimated based on AES measurements

(using elemental sensitivity data embedded in the equipment), as shown in Table 2-1.

The measured ratio of C/Zr remains was close to unity throughout the growth temperature range 400 to 600 °C at the higher H2 carrier flow rate (circle symbol in

Figure 2-3). These operating conditions are below the phase transition line (dotted line with P = 300 Torr in Figure 2-3) and single phase ZrC should form. Upon decreasing the

H2 carrier flow rate by half (square symbol in Figure 2-3), the C/Zr ratio increases significantly to 1.25, suggesting the possibility of carbon codeposition, although this operating condition lies inside ZrC 1-phase deposition region.

Oxygen was always present in the films, while N was detected in the films with a composition C/Zr > 1. This suggests the cyano group (-CN) from the PhCN solvent possibly dissociatively absorbs on the surface to provide C(ads) or N(ads) for incorporation into the film. As the temperature increases, however, hydrogen is able to react with the CN radical to form hydrogen cyanide (HCN), a stable gas phase species. It is interesting that no N was detected in films (b) and (c), for which the mean C/Zr < 1.

Perhaps the excess C in films (a) and (d) is related the decomposition of CN, providing

44 both reactive C and N. X-ray diffraction characterization of the films did not detect any crystalline phases. It seems likely then that the film is amorphous and upon exposure to air quickly draws oxygen. It is also noted that the oxygen content increases with growth temperature along with a decrease in the C/Zr ratio. The excess Zr may provide additional adsorption/reaction rates.

Figure 2-4 shows the relative distribution of C in the gas and condensed phases as a function of growth temperature, as predicted by the chemical equilibrium analysis. The stoichiometry of the ZrC phase (C/Zr = 1) is practically unchanged and almost all the Zr is in the solid. The graphitic carbon content in the solid phase increases rapidly as the temperature increases above ~ 660 °C. Based on the calculated phase diagram and experimental results, a certain amount of hydrogen is needed to prevent codeposition of elemental carbon over a wide range of temperature, pressure, and inlet C/Zr ratio. The apparent proposed growth temperature at P = 300 Torr and H/Zr ratio = 1.16×105 appears to be approximately T = 500 °C. A higher temperature is not preferred since gas phase parasitic reactions increases with the temperature increase, which are unfavorable for film deposition as discussed in Chapter 4.

2.3.3 Phase Change as a Function of Temperature and the Inlet C/Zr Ratio

In the deposition of metal carbides, a more reactive hydrocarbon is sometimes used

(e.g., propane). Figure 2-5 shows the effect in equilibrium of changing the inlet carbon content for two different reactor pressures. As expected the addition of another carbon source lowers the 2-phase transition temperature. If intermediate species are not reactive, propane or another reactive hydrocarbon may be needed to deposit stoichiometric ZrC.

For example, propane can be mixed in the carrier gas stream up to 2 vol% at 500 °C and

0.1 atm, 14 vol% at 500 °C and 1 atm. But for ZrNp4 the inlet carbon content is sufficient for the growth of ZrC in AA-MOCVD based on experimental data. It seems that either an alternative solvent (less carbon containing) should be sought or the solvent amount should be reduced. In the calculation shown in Figure 2-5, the C/Zr was changed by adding C3H8 to a constant amount of ZrNp4 and benzonitrile (C/Zr = 3856). The H/Zr was then determined by the mole balance at a given C/Zr ratio calculated from the volume fraction of propane in the carrier with a fixed flow rate = 1.5 slm.

2.3.4 The Addition of Chlorine to the System

The growth rate of ZrC using ZrNp4 is very low. It is suspected that the physisorption on the surface of the ZrNp4 molecule is not favorable because the tetrahedral symmetry between the center Zr and four nearest carbon atoms results in no permanent dipole moment. The moiety after α-hydrogen abstraction, Np2Zr=CH(t-bu), also forms a structure close to the planar three-fold symmetry, so that its dipole moment is small as well. By substituting Cl for one of the neopentyl (Np) ligands, the dipole moments of the precursor (ZrNp3Cl) and its moiety after α-hydrogen abstraction

(NpClZr=CH(t-bu)) are anticipated to be increased by a factor of three or four. Chlorine is sometimes used as a transporting agent (e.g. transport of Zr as ZrCl4) accompanied by a reducing agent (e.g. H2) or added to the system to increase the reaction reversibility or improve the film purity. Thus the addition of Cl into the system may improve both the kinetics as well as the equilibrium conversion. To test this latter hypothesis, equilibrium calculations were performed.

Figure 2-6 shows the T-H/Zr CVD phase diagram for Zr-C-H-Cl system for two values of Cl/Zr at a fixed value of C/Zr=3856. The same C/Zr ratio was retained because

46 the loss of only four carbons by the replacement of a Np ligand with Cl was negligible compared to the carbon supplied by the solvent. When the inlet Cl/Zr ratio is less than 4

(see Figure 2-6a), the CVD phase diagram is almost identical to the one displayed in

Figure 2-3 for Zr-C-H system, except that ZrC(FCC) + ZrCl4(s) + graphite + gas and

ZrC(FCC) + ZrCl4(s) + gas phase fields with ZrCl4 (s) appear at low temperature (< 300

°C). In contrast to the low Cl addition result, the phase fields Graphite + gas and Gas appear as the inlet Cl/Zr ratio becomes greater than 4 (see Figure 2-6b), since there is sufficient Cl in the system to form ZrCl4(g), which is more stable than ZrC. Thus, the introduction of chlorine to the reactor (e.g. chlorinated precursor) to a large extent (Cl/Zr

≥ 4) reduces the extent of the desired 2-phase ZrC(FCC) + gas field.

This reduction occurs by shifting the lower boundary of the 2-phase region to higher temperature, which retains the possibility of high deposition temperature. A little

Cl introduction (C/Zr < 4) may, however, decrease a kinetic barrier to increase the rate of deposition without greatly impacting the equilibrium boundaries.

The addition of HCl or a chlorine-based solvent as the Cl source is expected to reduce the equilibrium growth rates. The effects of hydrogen and chlorine addition are almost independent of each other since they are competing for the same elements and do not deposit in the film (H competes with Zr for C and Cl competes with C for Zr). Figure

2-7 illustrates this more clearly, which shows the CVD phase diagram as a function of the inlet composition (Cl/Zr and H/Zr) at fixed T, P and C/Zr. As indicated by the dotted arrows, the vertical phase transition line moves to the right and horizontal curve (y abscissa = 4) moves upward independently, as the temperature increases. The shift to higher H/Zr for the vertical phase transition line is explained by the tendency of graphite

formation over the formation of CH4 and other hydrocarbons, while the shift to higher

Cl/Zr of horizontal phase transition line is attributed to the tendency of non-Zr containing gas species, such as HCl, over ZrCl4 (g) formation. It is noted that the horizontal phase transition curve is not linear. In the atmospheric halide CVD mentioned in Section 1.5.2, the inlet Cl/Zr ratio is 4, so that the condition is not in the complete-etching zone as shown in Figure 2-7. ZrC growth without carbon codeposition is thus possible by increasing the hydrogen carrier amount or reducing the inlet C/Zr ratio through adjustment of CH4 flow rate, as depicted in Figure 2-7 and Figure 2-5, respectively.

2.4 Summary

A chemical equilibrium study was performed to investigate the effect of operation parameters on the constitution in ZrC films grown by chemical vapor deposition (CVD).

The equilibrium analysis of the Zr-C-H system demonstrated that ZrC (FCC) deposition is favorable and that a certain minimum amount of hydrogen is needed to prevent codeposition of elemental carbon over a wide range of temperature, pressure, and inlet

C/Zr ratio. The results of the equilibrium analysis were compared to the phase constitution of films grown in conventional high vacuum metal organic CVD (< 10-4

Torr) reactor. Only carbon-rich ZrC films were grown and demonstrated the possibility of an aerosol-assisted CVD approach to stoichiometric ZrC film growth.

Figure 2-1. The reproduced Zr-C phase diagram [Gui89, Gui95].

Figure 2-2. CVD phase diagram showing equilibrium phases as a function of temperature and pressure for three inlet H/Zr atom ratios [inlet C/Zr = 20].

Figure 2-3. CVD phase diagram showing deposited phases as a function of temperature and the inlet H/Zr ratio for four pressures [inlet C/Zr = 3856]. The circle and square symbols represent the growth condition of four films grown by AA- MOCVD.

Figure 2-4. Deposition efficiency as a function of growth temperature. Csolid/Zrinlet denotes the ratio of total solid carbon (ZrC + graphite) to inlet Zr, Cgas/Zrinlet denotes the ratio of total carbon in gas phase. The insert shows ZrZrC/Zrinlet [C/Zr = 3856, H/Zr = 105, P = 300 Torr].

Figure 2-5. CVD phase diagram showing deposited phases as a function of the inlet C/Zr ratio and temperature at 0.1 and 1.0 atm

Figure 2-6. CVD phase diagram showing deposited phases as a function of the inlet H/Zr ratio and temperature for two values of inlet Cl/Zr ratio [P = 300 Torr, C/Zr = 3856].

Figure 2-7. CVD phase diagram showing deposited phases as a function of the inlet H/Zr and Cl/Zr ratios [P = 300 Torr, T = 500 oC, C/Zr = 3856].

Table 2-1. ZrC films composition at different deposition conditions (P = 300 Torr and inlet C/Zr = 3856).

Entry (a) (b) (c) (d) Growth temperature (°C) 400 500 600 500

H2 Carrier flow rate (slm) 1.5 1.5 1.5 0.75 Zr (at. %)a 39.2 44.4 44.0 38.7 C (at. %)a 42.3 42.3 41.0 48.2 C/Zr 1.08 0.95 0.93 1.25 O (at. %)a 11.1 13.4 15.0 9.9 N (at. %)a 7.4 - - 3.2 aValues measured by AES

CHAPTER 3 STUDY OF PRECURSOR DECOMPOSITION USING COMPUTATIONAL THERMOCHEMISTRY

3.1 Introduction

The study of CVD film growth often follows a time consuming and costly linear approach: investigate sequentially the effect of each process variable (e.g., growth temperature, pressure, inlet precursor concentration) on film properties assessed by ex situ measurement techniques. The details of reactions occurring during growth are rarely studied because of their complexity and the lack of proper in situ tools. This linear approach could pose challenges especially when a novel precursor is used since little preliminary information is available.

Quantum mechanical calculations are useful to provide thermochemical and kinetic properties of possible molecules involved. Using statistical methods to average over large systems can provide estimates of thermodynamic quantities such as reaction enthalpy. Additionally, activation energies can be obtained and most probable reaction pathways located. Through this theoretical experiment, the precursor design strategy can be explored prior to experiment or in conjunction with nominal preliminary screening of precursors. A better understanding of probable growth reactions will assist in the design of experiments.

3.2 Experimental Methods

All calculations were carried out using the GAUSSIAN 03 program, along with the

B3LYP DFT method and split basis set (LanL2DZ for transition metals and 6-311G(d) or

53 54

6-31G(d) for other elements) [Lee92, Bec93, Gau04]. Full geometry optimization was carried out for all species. The transition states (TS) was optimized using the Berny

Algorithm as implemented in the GAUSSIAN 03 program. Harmonic vibrational frequencies were calculated for each structure, and used as the basis for computing enthalpy and Gibbs energy for thermochemical analysis. Atomic charges and Wiberg bond indices were evaluated using natural bond orbital (NBO) analysis to be used for the qualitative evaluation of bond strength. GaussView, MolDen and gOpenMol programs were used for the visualization of the results [Laa92, Ber97, Sch00].

3.3 Decomposition of Tetraneopentyl and Tetrabenzyl Zirconium Precursors for the CVD of Zirconium Carbide

Although thermodynamic analysis provides a general basis for determining the optimum operating condition, the real deposition behavior, on many occasions, does not follow what thermodynamics has proposed. As a clear example in our study, an analogous series of CVD experiments using tetrabenzyl zirconium (ZrBn4) failed to produce any Zr-containing films despite the thermodynamic expectation of at least carbon-rich ZrC film deposition. Accordingly, there have been no reports on ZrC film growth using ZrBn4 so far. Thus, the study of possible precursor decomposition mechanisms is of value in understanding the decomposition behavior.

In this study, computational thermochemistry was utilized as a tool for the theoretical study of decomposition pathways of ZrNp4 (tetraneopentyl zirconium) and

ZrBn4.

3.3.1 Comparison of Decomposition Behaviors of ZrNp4 and ZrBn4

According to the thermodynamic analysis (see Chapter 2), ZrBn4 is expected to produce at least carbon-rich ZrC deposition like ZrNp4 does. Qualitative analysis was carried out to explore the bonding differences between ZrNp4 and ZrBn4.

Figure 3-1 shows optimized geometries for ZrNp4 and ZrBn4. The average bond distances of Zr-C1 and C1-C2, and the bond angle Zr-C1-C2 were compared to data from the single crystal X-ray structural determination for ZrBn4 (see Table 3-1) [Ted98]. The calculated values are in very good agreement with experimental data, especially for bond distances. The discrepancy in the Zr-C1-C2 bond angle is attributed to crystal packing forces, which are not present in gas phase calculations. Since no crystallographic data have been reported for ZrNp4, no comparison could be made for that complex.

To compare the difference between the two precursors with respect to the Zr-C1 and C1-C2 bonds, a natural bonding orbital (NBO) analysis was employed. Table 3-2 contains atomic charges on Zr, C1, and C2 calculated via natural population analysis

(NPA) and Wiberg bond indices. It is noticeable that the Wiberg bond index of the Zr-

C1 bond of ZrNp4 is higher than that of ZrBn4 and the Coulombic interaction between Zr and C1 is also stronger in ZrNp4. Due to the resonance stabilization of the benzyl radical

(PhCH2·) that would be formed upon homolysis, the Zr-C1 bond in ZrBn4 is expected to be weak. Facile Zr-C1 bond cleavage is consistent with the lack of growth from ZrBn4, since deposition of ZrC requires incorporation of carbon into the film. In contrast, the neopentyl groups of ZrNp4 will have higher Zr-C bond dissociation energies due to the formation of the less stable primary (neopentyl) radical upon homolysis. Thus, other decomposition processes are more likely to compete with Zr-C homolysis in ZrNp4.

Neopentyl groups are known to undergo C-H activation at the α and γ positions as well as

C-C bond cleavage within the ligand [Gir87, Wu96, Che9, Wu99]. These processes within the neopentyl ligand would provide pathways for carbon incorporation into the films. Quantitative comparison will be discussed in the following section.

3.3.2 Initial Stage of the Decomposition of ZrNp4

Experimental and theoretical studies on the thermolysis of tetraneopentyl titanium

(TiNp4) proposed two possible reactions – α-hydrogen abstraction and γ-hydrogen abstraction – as the initial step in the decomposition of TiNp4 and the preference is for α- hydrogen abstraction over γ-hydrogen abstraction.[Wu96, Che97, Wu99].

α-hydrogen abstraction Η TiNp4 Np2Ti = + CMe4 (Neopentane) (t-Bu) (3-1) γ-hydrogen abstraction TiNp4 Np2Ti + CMe4 (3-2)

By analogy, the same initial decomposition reactions are examined for ZrNp4 as shown in Figure 3-2. Geometries of transition complexes (1a-t, 1b-t) and products (1a,

1b) were consistent with ones obtained by Wu et al. using molecular orbital theory

(HF/3-21G) instead of DFT [Wu99] as used in this study. To determine the preference, the energetics of the possible pathways in the initial step of ZrNp4/ZrBn4 decomposition were listed and compared in Table 3-3. Apparently, α- hydrogen abstraction (entry 1) and γ-hydrogen abstraction (entry 2) are much more favorable than the homolysis of

o,‡ neopentyl (entry 5). For comparison, the ∆H 298 of the homolysis of benzyl from tetrabenzyl zirconium (ZrBn4) is 45.6 kcal/mol, which has the small difference of about 5 kcal/mol in activation energy from α- or γ-hydrogen abstraction from ZrNp4 (entry 1 or

2). It indicates the homolysis of benzyl from ZrBn4 readily occurs in the typical temperature range of MOCVD, 400 to 800 °C. Moreover, it explains the reason that there have been no reports on ZrC film growth using ZrBn4 so far and our trials of ZrC film growth using ZrBn4 also failed.

In terms of equilibrium thermodynamics, it is noted that α-and γ-hydrogen abstractions (entry 3 and 4) are both endothermic and γ-hydrogen abstraction is more exothermic because its product (1b) still maintains the original stable tetrahedral configuration at zirconium by forming a ring at one side. Also in terms of kinetics, γ- hydrogen abstraction is more favorable, having lower activation energy. These results are consistent with the theoretical and experimental results reported by Wu et al. [Wu99].

o,‡ The calculated difference in ∆G 298 between the two hydrogen abstractions from ZrNp4 is only about 1.7 kcal/mol, corresponding to 17 times faster γ-hydrogen abstraction over

α-hydrogen abstraction at 298 K or 3.0 to 3.5 times faster at 400 to 500 °C, according to

TST kinetics. For comparison, Wu et al. determined that γ-hydrogen abstraction is 4.9 times faster at the same temperature range (400 to 500 °C), through an experimental study using MS and deuterium labeled ZrNp4 [Wu99]. Therefore, one of the two pathways is not overwhelming the other but instead they are competing, even if γ- hydrogen abstraction is a little more favorable.

Wu et al. also reported computational analysis results using molecular orbital

o,‡ theory to support their experimental study [Wu99]. However, the difference of ∆G 298 values of α- and γ-hydrogen abstractions using HF/3-21G was 5.9 kcal/mol in their calculation. Although an adjustment on the basis of comparison with the results obtained

‡ for ZrMe4 using HF/6-31G and MP2/HW3 lowered the difference of ∆E values of α-

58 and γ-hydrogen abstractions by 1.4 kcal/mol, their calculation still indicated the dominant preference of γ-hydrogen abstraction compared to that of this study by three orders of magnitude (~ 2000 times faster) at 298 K. The calculation using HF method with the

o,‡ same basis set of 6-31G(d) in this calculation also generated a large difference in ∆G 298 values (3.1 kcal/mol), corresponding to the preference of γ-hydrogen abstraction by two orders of magnitude (~ 180 times faster), as shown in Table 3-3 (entries 7 and 8).

It is worthwhile to compare the activation energy of γ-hydrogen abstraction of

ZrNp4 to the activation of α-hydrogen abstraction of TiNp4. It was confirmed theoretically [Wu96] and experimentally [Che97] that α-hydrogen abstraction is more favorable in the initial decomposition of TiNp4 due to its larger release of steric interactions. The activation energy of the α-hydrogen abstraction of TiNp4, experimentally determined in solution, was 21.5 ± 1.4 kcal/mol [Che97] and it is much lower compared to 37.4 kcal/mol, theoretically determined for the γ-hydrogen abstraction of ZrNp4, as shown in Table 3-3. This large difference, even with the practical limit of theoretical calculation, is consistent with the reported CVD growth of TiC using TiNp4 at temperatures as low as 150 °C, while 400 °C is the minimum deposition temperature in our AA-MOCVD of ZrC thin films using ZrNp4.

As a summary, calculations suggest that γ-hydrogen abstraction is preferred over α- hydrogen abstraction in the initial decomposition of ZrNp4, but they are competing especially at CVD temperature (400 to 800 °C).

3.3.3 Isobutene Cleavage

The isobutene formation via a ring opening from the γ-hydrogen abstraction intermediate was suggested by Wu et al. [Wu99]. Along with the information of the

59 preference of γ-hydrogen abstraction in the previous section, the following decomposition pathway can be suggested.

γ-hydrogen abstraction γ-hydrogen abstraction ZrNp 4 Np2Zr C(CH 3)2 - CMe4 1b - CMe4

(CH3 )2C Zr C(CH 3)2 (CH 3)2C Zr=CH 2 (3-3) 2b - H2CCMe2 3b

Figure 3-3 shows the structures of 2b-t (transition state for the second γ-hydrogen abstraction), 2b, and 3b. It is noted that the tetrahedral structure at zirconium is still maintained in 2b. For the formation of isobutene via a ring opening, the effort to locate the corresponding transition state failed, but it is expected that the transition structure is close to the reaction product [Wu99]. Accordingly, the structure of 3b shows an empty coordination site, which likely isobutene used to occupy. The activation energy of 1b →

o 2b and ∆H 298 of 2b → 3b, as shown in Figure 3-4, indicates these reactions are accessible in our typical growth temperature range, 400 to 600 °C. The structure of 3b, tetrahedral with an empty coordination site, suggests the participation of this molecule into the heterogeneous deposition process using its empty coordination site. The second isobutene formation via the other ring opening from 3b to form Zr(H2C=CH2) requires

o less energy (∆H 298 = 27.2 kcal/mol) compared to the first isobutene formation (2b → 3b) due to the stable triangular structure in Zr(H2C=CH2), as shown in Figure 3-4. Thus, 3b or Zr(H2C=CH2) might be the last product when the reaction occurs in the gas phase in the typical growth temperature range of MOCVD (400 to 800 °C).

It is noted that the activation energies for the main mechanistic steps – α-hydrogen abstraction (40.6 kcal/mol in Table 3-3), γ-hydrogen abstraction (37.4 kcal/mol in Table

3-3 and 39.2 kcal/mol in Figure 3-3), isobutene cleavage (41.1 kcal/mol in Figure 3-3) are all in the narrow range, 37 to 42 kcal/mol. These results suggest that these reactions should be competitive in the temperature range used for MOCVD.

The decomposition of ZrNp4 in the gas phase stops at 3b or Zr(H2C=CH2). It is noted that three (3b) or two (Zr(H2C=CH2)) carbons surround zirconium. Thus, if they participate in heterogeneous reactions for film growth, C/Zr ratio in the deposited film will likely be greater than the stoichiometric value of ZrC without considering the H2 reducing effect. As an illustration, 3b has an empty coordination site that is potentially usable for the adsorption on the surface. Then, isobutene will be possibly cleaved off from the ring side to leave behind Zr(H2C=CH2) (ad) to probably result in C/Zr ~ 2 in the deposited film. Further study is required for the details of the heterogeneous reaction mechanism for film growth. As presented so far, computational thermochemistry is a very useful tool for elucidating the gas phase decomposition mechanism of a precursor

(especially, a novel precursor) in the MOCVD growth.

3.3.4 Summary

Qualitative and quantitative analyses using computational thermochemistry suggested that Zr-C bond cleavage should be more facile in tetrabenzyl zirconium

(ZrBn4) than in ZrNp4, consistent with growth of ZrC from ZrNp4 but not ZrBn4. In the initial stage of ZrNp4 decomposition, the preference of γ-hydrogen abstraction over α- hydrogen abstraction was confirmed, but they were competing. The subsequent isobutene cleavage via a ring opening was also examined by computational

thermochemistry. The facile gas phase decomposition of ZrNp4 ends at intermediates containing two or three nearest carbons to zirconium, so that C/Zr ratio higher than its stoichiometric value (1:1) in the deposited film in several MOCVD growth reports using

ZrNp4 seems to be consistent. As a summary, the computational thermochemistry was very useful to predict the gas phase decomposition of an unknown precursor in the CVD growth.

3.4 Decomposition of Alkyl- and Arylimido Precursors for CVD of Tungsten Nitride

Computational thermochemistry, as shown in previous sections, is useful for predicting gas phase decomposition pathways of metal organic precursors and interpreting experimental observations. As another case study, specially designed alkyl- and arylimido precursors for CVD of tungsten nitride, whose experimental results are available [Bch03a, Bch03b, Bch04a, Bch04b, Bch05], were selected for the computational analysis.

3.4.1 Overview on the Tungsten Nitride MOCVD Growth

Although Al-based metallization schemes have historically been used in integrated circuits (ICs), the more recent use of copper is motivated by its higher electrical conductivity and increased resistance to electromigration. The high diffusivity of copper

–5 2 in silicon (Dcu ~ 2 × 10 cm /sec @ 500 °C) [Bro89, Ist00] coupled with its low solubility, however, leads to extensive redistribution of Cu and its accumulation at extended defects, which degrades device performance and process yields. To overcome the problem of copper contamination, thin films acting as diffusion barriers are used

[Nic78, Ist002]. Besides having good resistance to diffusion, barrier thin films should be structurally and thermally stable, have good adhesion to both copper and dielectric layers, be non-reactive to copper, have low resistivity and have enhanced resistance to thermal

and mechanical stresses. Tungsten nitride (WNx) is one of the promising candidates for barrier material based on the above criteria [Nak87, Uek96, Nak97, Iva99, Gal99, Kal00,

Sha01]. Recently, growth of WNx and tungsten carbonitride (WNxCy) thin films has been reported using a series of related single source precursors: the isopropylimido complex

i Cl4(CH3CN)W(N Pr) (1) [Bch03a], the phenylimido complex Cl4(CH3CN)W(NPh) (2)

[Bch03b], and the allylimido complex Cl4(CH3CN)W(NC3H5) (3) [Bch05]. Analyses of mass spectra of 1-3 obtained via positive ion electron-impact (EI) and negative ion electron-capture chemical ionization (NCI) provided insight into probable precursor decomposition pathways, while the activation energies for film growth were correlated with the N-C bond dissociation energies within the imido ligand. Properties of the deposited films were discussed in the context of probable decomposition pathways postulated from the mass spectra. On the basis that no ions containing the acetonitrile ligand were detected by mass spectrometry in either mode, it was postulated that facile dissociation of the acetonitrile ligand (CH3CN) was the first step in the deposition. The pathway for loss of chloride from the precursor was still unclear, although some information could be obtained from the RGA (Residual Gas Analyzer) data. The ultimate fate of the chlorides was HCl formation; no trace of Cl2 was detected in the gas phase and

Cl was not present in the films to the limit of detection by AES (Auger Electron

Spectroscopy).

Computational thermochemistry is an alternative tool for investigation of the decomposition mechanism of metal organic precursors when direct experimental measurements are not available. In this study, density functional theory (DFT) calculations are presented to analyze bonding in the alkyl- and arylimido complexes 1-3.

In addition, a computational study of the CH3CN cleavage step by means of statistical thermodynamics is presented. Cleavage of the W-Cl bonds by σ-bond metathesis with

H2 has also been examined computationally and a pathway based on calculated transition state structures is suggested.

3.4.2 NMR Kinetics of Acetonitrile Exchange in 2

The sample for the exchange study was prepared in the a box by dissolving

1 complex 2 and an equivalent amount of acetonitrile in CDCl3. The H NMR spectrum of this sample at -20 ºC displayed the signals for 2 [δ (ppm) 7.14 (hp, 1H), 2.58 (s, 3H) 1.70

(d, 6H)] together with free acetonitrile (2.11 ppm) in a ratio of 1:3. The exchange of 2 with acetonitrile in the solution was monitored by 1H NMR in the temperature range -54 to 34 ºC. The exchange rate k (see Figure 3-8) was determined by line-shape analysis in the temperature range -6 to 34 ºC. A plot of ln(k/T) vs. 1/T afforded the activation enthalpy (18.52 ± 0.14 kcal/mol) and entropy (15.8 ± 0.5 cal/mol·K) for the exchange of acetonitrile by 2.

The NMR spectra were recorded on a Varian Inova at 500 MHz for proton, equipped with a 5 mm indirect detection probe, with z-axis gradients. The variable temperature spectra were recorded on automation. To achieve temperature stability, for each temperature step of 2 ºC, a preacquisition delay of 1500 s was followed by shimming on the lock level. The spectra were collected in 16 transients, with an acquisition time of 5 seconds. No relaxation delay and no apodization were used. The actual temperature was measured by running a standard of methanol under the same conditions. The simulation of the spectra with exchange was done using gNMR.

3.4.3 Optimized Geometries

The optimized geometries for complexes 1-3 are, as predicted, very similar (see

Figure 3-4 and Table 3-4). All three complexes exhibit octahedral coordination at the tungsten center. Their N-W-Cl bond angles are greater than the ideal 90°, as expected for an octahedral complex with a single multiply bonded ligand [Nug88]. The W-N bond lengths and the bond angles about the tungsten center are in reasonable accord with experimental values obtained for related complexes by X-ray diffraction [Bra87, Gge88,

Bch05] and standard literature values [Orp89]. Minor differences in bond lengths and angles can be attributed, at least in part, to crystal packing forces that are not present in gas phase calculations.

The geometry optimizations of complexes 1-3 revealed additional local minima associated with rotation of the imido substitutent. As an example, two local mimina were found for the phenylimido complex: one has the plane of the phenyl ring eclipsed with two of the chlorines, while in the other the plane of the phenyl ring bisects the Cl-W-Cl angle. The energy difference between these two conformations was only about 0.077 kcal/mol. The lack of an electronic barrier for rotation of a planar ligand is characteristic of octahedral trans-ML4L'L" complexes, in which the four-fold local symmetry renders the dxz and dyz orbitals degenerate in the ML4L' fragment.

In preparation for assessment of the chemistry of 1-3 following loss of the labile

i acetonitrile ligand, their coordinatively unsaturated derivatives Cl4W(N Pr) (1a),

Cl4W(NPh) (2a), and Cl4W(NC3H5) (3a) were also subjected to computational geometry optimization (see Figure 3-5 and Table 3-5). Due to the lowered electron density around the metal upon loss of the CH3CN ligand, the W-Cl and W-N bond lengths in 1a-3a are

65 shorter than in the coordinatively saturated 1-3. Increased N-W-Cl bond angles are a result of distortion toward a square pyramidal geometry with the strong trans influence imido ligand in the apical position.

Note that in all six complexes (1-3 and 1a-3a), the N(imido)-C bond lengths follow the trend expected on the basis of the hybridization at carbon. For the phenyl compounds

1 and 1a, the ipso carbon is sp2 hybridized, leading to a prediction that the N(imido)–C

(sp2) bond length will be shorter than the N(imido)–C (sp3) bond lengths in the isopropyl- and allylimido complexes 2-3 and 2a-3a. Experimental N(imido)-C bond lengths obtained by X-ray crystallographic structure determination follow the same trend. A search of the Cambridge Structural Database [All02] yielded a mean N(imido)–C (sp2) bond length of 1.40Å for 29 tungsten phenylimido complexes. A similar analysis for alkylimido complexes afforded a mean N(imido)–C (sp3) bond length of 1.46 Å for 14 tungsten tert-butylimido complexes, while the less sterically hindered ethylimido complexes averaged an N-C bond length of 1.43 Å (6 structures).

3.4.4 Dissociation of Acetonitrile

It was previously postulated initial dissociation of the CH3CN ligand from 1-3 during film growth on the basis of the strong trans influence of imido ligands [Nug88,

Vau95], which has been correlated to trans effects in the dissociative reactions of imido complexes [Hog05]. This decomposition mechanism for 1-3 has been previously discussed in the context of their mass spectra, in which the highest m/z values in both positive and negative modes correspond to ions from which CH3CN has been lost

[Bch03a, Bch03b, Bch05]. Calculations of the frequencies for the W-N (nitrile) stretching mode shown in Figure 3-6 yielded values of 190 cm-1 for 1, 194 cm-1 for 2 and

191 cm-1 for 3. The frequency of 190 cm-1 is equivalent to the vibrational temperature

(Θv) of 274 K, supporting the hypothesis that acetonitrile dissociation from 1-3 would be kinetically facile at the typical film deposition temperature in the range 450 to 750 °C.

Assuming that CH3CN dissociation occurs via a simple stretching mode, statistical thermodynamics was used to estimate the values of ∆Hf° and ∆Gf° of complexes 1-3, 1a-

3a, and CH3CN allowing ∆H° and ∆G° for the cleavage reaction to be calculated. A plot of ∆G° vs. temperature and ∆Ho for dissociation of acetonitrile from 1-3 are shown in

o o Figure 3-7 and Table 3-10, respectively. Here, ∆Hf (1a) + ∆Hf° (CH3CN) - ∆Hf (1) can be regarded as approximately equal to the activation energy for the CH3CN ligand cleavage because the transition state for this endothermic reaction should be product-like in character. Note that at temperatures within the film growth range, ∆G° is negative, indicating that acetonitrile loss is thermodynamically favorable, as well as kinetically accessible. These calculations are consistent with dissociation of CH3CN as the first step in film growth from 1-3 at temperature above 450 oC.

To obtain experimental values for the activation energy of CH3CN dissociation

1 from complex 2, H NMR kinetics were studied in CDCl3 with excess CH3CN. Upon lowering the temperature to -20 °C, both bound and free acetonitrile could be detected in the 1H NMR spectra of the solution. As the temperature was raised, the bound and free acetonitrile signals coalesced. The exchange rate k was determined by lineshape analysis in the temperature range -6 to 34 °C . A plot of ln(k/T) vs. 1/T yielded an activation energy of 18.52 ± 0.14 kcal/mol and an entropy of activation of 15.8 ± 0.5 cal/mol·K for the exchange of acetonitrile, as shown in Figure 3-8. Since the first order kinetics of the process establish a dissociative mechanism for the exchange process, those values

‡ ‡ correspond to ∆H and ∆S for loss of CH3CN from isopropylimido complex 2. Given

67 that the experimental values were obtained in solution while the calculated values are for a gas phase process, the agreement between them is reasonable.

3.4.5 Cleavage of W-Cl Bonds

According to the RGA (Residual Gas Analyzer) data obtained from the reactor effluent during depositions of films from imido complexes 1 and 2 [Bch03, Bch04a], HCl was the only chlorine-containing decomposition product. Given the high oxidation state of the tungsten center in 1-2, it is noteworthy that the reductive elimination product Cl2 was not observed. In addition, chlorine was not present in the deposited WNx films within the detection limits of AES (Auger Electron Spectroscopy). Thus, the mechanistic pathways from imido complexes 1-3 to HCl are of interest and have been investigated computationally.

Conversion of the chloride ligands to HCl under film deposition conditions suggests involvement of the H2 carrier gas in cleavage of the W-Cl bonds. The most common pathways for reaction of H2 with organometallic complexes include the following: 1) oxidative addition of H2/reductive elimination, 2) coordination of H2 followed by transfer of an acidic proton, and 3) σ-bond metathesis. Oxidative addition is precluded by the d0 electron count of 1-3 and 1a-3a and as expected, local minima associated with formation of the dihydride could not be found. Although precoordination of H2 as a σ-complex has been reported to lie in a shallow well in prior DFT studies of

0 electron poor d species [Fol92], such a minimum could be located for approach of H2 to

1a-3a.

0 For d transition metal complexes, the preferred reaction pathway with H2 is σ- bond metathesis, in which ligand exchange occurs via a 4-center transition state [Fol92,

Cun93]. Observation of σ-bond metathesis is generally confined to those systems where oxidative addition is not a viable pathway. Figure 3-9 shows the transition states for σ- bond metathesis of 1a-3a with H2 and local minimum geometries for the metal hydride products. Each transition state depicted in Figure 3-9 exhibited the requisite single negative force constant with the corresponding vibrational mode leading to formation of the metal hydride and HCl. The reactions are endothermic, with activation energies of approximately 37 kcal/mol, which will be accessible in the reported growth temperature range of 450 to 700 °C. Although σ-bond metathesis of a metal chloride with H2 to yield a metal hydride and HCl has not been reported, under the high temperatures and H2 flux of the CVD reaction conditions, this reaction appears to be a viable pathway to the experimentally observed chloride-free films and HCl byproduct. Table 3-7 summarized refined thermodynamic quantities for the first σ-bond metathesis of 1a-3a. In the calculation of refined quantities, vibrational temperatures corresponding to frequencies less than 100 wavenumbers were separated out because they are attributed to internal rotations. It is noted that the maximum rotational frequency in DFT calculation is around a few tens of wavenumbers (say 50 or so). The values of ∆Ho,‡ are not terribly high, as mentioned above, but negatively high ∆So,‡ values may cause huge entropic effects at high temperature.

Figure 3-10 depicts possible σ-bond metathesis intermediates for successive substitution of chlorides by hydrides and possible pathways for reductive elimination of

HCl from the hydride complexes. As substitution can lead to either the trans (MI-2A) or

i cis (MI-2B) isomer of H2Cl2W(N Pr) (see Figure 3-10), several reaction manifolds were considered. The activation energy for the second or third σ-bond metathesis was just as

69 much as for the first σ-bond metathesis (around 38 ~ 40 kcal/mol), while the activation energy for the fourth step was increased up to 53 kcal/mol. This value, however, was still competitive with the calculated activation energy for reductive elimination of HCl (ca. 74 kcal/mol).

3.4.6 Bond Dissociation Energies for W-N(imido) and N(imido)-C in Complexes 1-3

Growth of WNx from 1-3 requires not only dissociation of the acetonitrile and chloride ligands (vide supra), but also cleavage of the N-C bond in the imido ligand. It has been previously demonstrated that the apparent growth kinetics and composition of

WNx films grown from 1-3 can be correlated with the bond dissociation energies of the

N(imido)-C bond in the precursor [Bch05]. In an effort to understand the effects of bonding within the imido moiety on precursor decomposition under film growth conditions, a computational investigation was carried out.

As seen in Tables 3-4 and 3-5, bonding throughout the imido moiety reflects the difference of 1, which exhibits conjugation between the phenyl group and the imido nitrogen, from 2 and 3, which bear saturated carbons at the C-N1 linkage. In 1, donation of the p-type nitrogen lone pair into an empty metal d orbital is attenuated by conjugation of the N p-orbital into the phenyl ring (see Figure 3-11). This effect weakens the W-N1 bonding of 1 with respect to that in 2 and 3, for which such conjugation into the alkyl group is not possible. The result can be seen in the longer W-N1 bond in 1 (1.737 Å) in comparison to those of 2 and 3 (1.724 Å). A related effect can be seen in the N1-C bond lengths, where the N1-C(sp2) length in phenylimido complex 1 (1.380 Å) is, as expected, shorter than the N1-C(sp3) distances in alkylimido complexes 2 and 3 (1.441 and 1.444

Å).

Additional information on bonding of the imido ligands of 1-3 was obtained using natural bonding orbital (NBO) analysis. Figure 3-12 depicts the atomic charge of each atom calculated from natural population analysis (NPA) and the Wiberg bond index for bonds of interest. As expected, the calculated bond order of the N-C(phenyl) bond of 1 is higher than those of the N-C(alkyl) bonds of 2 and 3. The Wiberg bond indices for the

W-N1 bonds of 1-3 also support the relative bond strengths inferred from bond distances, with the longer W-N1 bond of 1 exhibiting a significantly lower bond index than those of

2 and 3.

To make a quantitative comparison of the N1-C bond energies in 1-3 and 1a-3a, the energetics of the reaction depicted in Figure 3-13 were determined by using statistical thermodynamics (see Section 1.4.2) to obtain values for ∆Hf° and ∆Gf° of complexes 1a-

3a, 1a'-3a', 1a″-3a″, NWCl4 and WCl4(tetrahedral), allowing ∆H° and ∆G for the cleavage reaction to be calculated. Acetonitrile-free complexes 1a-3a were used for this calculation because the lability of the acetonitrile ligand suggests its dissociation to be the first step of the decomposition of 1-3 under film deposition conditions (vide supra). The homolytic cleavage of the N(imido)-C bonds required for calculating the bond dissociation energies resulted in open shell products that were treated as doublets. In the cases of W-N1 cleavage, the products were treated as triplets. The basis set superposition error (BSSE) was corrected, even though it was less than 5% compared to the overall

∆H°  value in each calculation. Table 3-8 summarizes the estimated bond dissociation enthalpy values.

Although the calculated N1-C bond lengths were not significantly different for isopropylimido complex 2a and allylimido complex 3a, the trends in calculated N1-C

71 bond dissociation enthalpies (BDE) for 1a-3a do differ as a result of the relative stabilities of the organic radicals formed upon homolysis. Note that the metal-containing fragment (Cl4WN) is the same in all three N1-C cleavages. These differences in BDE for

1a-3a parallel to the trend in the reported N-C bond dissociation energies for the

i corresponding amines R-NH2 (R = Ph, Pr, allyl) [Ben76, Luo94]. For both the amines and 1a-3a, the N-Ph BDE is ca. 20 kcal/mol higher than that of N- iPr. The N-allyl bond is weaker than the isopropyl, although the calculated 2a-3a comparison affords a larger difference than found in the amines (15.7 vs. 10.7 kcal/mol). The relationship between the estimated N-C BDE for 1a-3a and apparent activation energies for deposition of WNx deposited from these precursors [Bch05] suggests that computational estimation of bond dissociation energies could be of value in screening of CVD precursors.

The computational results can be qualitatively correlated to the deposition behavior of MOCVD growth using 1a-3a (see Table 3-9 and Figure 3-14). The relatively low W-

N bond strength of phenylimido complex 1a with respect to alkylimido complexes 2a and

3a suggests that the phenylimido ligand is more likely to dissociate from the metal intact.

This loss of the PhN moiety would result in lower nitrogen content in films from 1a as seen in Figure 3-14 and lower N/W ratio accordingly as seen in Table 3-9. The calculated N1-C bond dissociation energies for 1a-3a shown in Table 3-8, have a similar trend found in experimentally determined activation energies for film growth from 1a-3a shown in Table 3-9; 1a > 2a > 3a. It suggests that N1-C bond cleavage is the rate- determining step in film growth from 1a-3a. Then, the low growth rate in film growth from 3a as shown in Table 3-9 is possibly attributed to N1-C bond cleavage in the gas phase, consuming the precursor before it reaches the substrate surface. Table 3-10

72 demonstrates the possibility of this gas phase cleavage of 3a more clearly. Hydride product via σ-bond metathesis from 3a, H4W(NC3H5), has much more reduced N1-C bond dissociation energy compared to that of 3a itself, by about 15 kcal/mol. It suggests the gas phase N1-C bond cleavage from 3a is viable following σ-bond metathesis.

3.4.7 Interpretation of Positive Ion EI MS Data

The tungsten nitride or tungsten carbonitride films deposited using tungsten imido precursors 1-3 generally had low N/W ratios (max. 0.05 even with 2 and 3), as shown in

Table 3-9, compared to the stoichiometric covalent compound W2N. For the phenylimido precursor, the facile loss of the PhN moiety is manifested in the very low

N/W ratio (max. 0.01). Another explanation, however, may be needed for the still low

N/W ratio with 2 and 3. The positive ion EI MS data of 2 and 3, as shown in Table 3-11, is the starting point of the new explanation [Bch03b, Bch05]. The relative abundance of

+ + + [Cl3WNH] to that of [Cl3W] shows a different behavior between 2 and 3; [Cl3WNH] >

+ + + [Cl3W] for 2 and [Cl3WNH] < [Cl3W] for 3. Both ions are clearly from

i + + [Cl3W(N Pr)] and/or [Cl3W(NC3H5)] .

Based on this information, two possible pathways were constructed using transition complex search, as shown in Figure 3-15. In the first pathway, β-hydrogen is transferred

+ to nitrogen and propylene or propadiene is formed with [Cl3WNH] (see Figure 3-15 (a1) and (b1)). The other pathway is α-hydrogen is shifted to nitrogen and subsequent W-N

+ bond cleavage generates [Cl3W] (see Figure 3-15 (a2) and (b2)). The energetics of these

+ + two pathways for the decomposition of [Cl3W(NiPr)] and [Cl3W(NC3H5)] , as shown in

Figure 3-16, demonstrate clearly the relative preference between the two pathways for

+ each case, consistent with the MS data. Small ∆ values, 4.5 kcal/mol for [Cl3W(NiPr)]

+ and 1.8 kcal/mol for [Cl3W(NC3H5)] , suggest two pathways are competing and explains

+ + the recognizable abundances of both [Cl3WNH] and [Cl3W] fragments in MS data.

Now, assuming those two pathways are also valid for neutral molecules, 2a and 3a, similar energetic diagrams can be constructed as shown in Figure 3-17. Interestingly, the formation of tungsten tetrachloride (WCl4) is favorable for both cases, probably due to its

o,‡ stable tetrahedral structure. The activation energy (∆H ) for WCl4 formation is 60.4 kcal/mol for 2a and 56.3 kcal/mol for 3a. Considering the activation energy for the fourth σ-bond metathesis of 2a is 52.4 kcal/mol as shown in Figure 3-10, WCl4 formation from 2a and/or 3a is a viable pathway in the gas phase and this N-deficient moiety formation explains the still low N/W ratio in the film deposited using 2 and 3. Chlorine cleavage from WCl4 is achieved homogeneously or possibly heterogeneously given that the σ-bond metathesis product (for example, H4W(NiPr)) generates surface hydrides during heterogeneous growth reaction. As also noted in Section 3.4.4, σ-bond metathesis plays an important role, not only removing chlorines homogeneously but also generating surface hydrides. These hydrides possibly participate in the removal of surface bound chlorides and cyano groups (-CN) on the surface. Further study is needed to better understand the heterogeneous reaction schemes.

3.4.8 Feasibility of Alkyl- and Arylimido Precursors for Tungsten Nitride ALD

WCl4 formation, at first look, seems to be unfavorable because it causes nitrogen deficiency in the film. But if ammonia as a second precursor and nitrogen carrier instead of hydrogen are used, the reaction pathways may be altered. Interestingly, a similar

i pathway for WH4 formation via α shift of hydrogen from H4W(N Pr) or H4W(NC3H5) was not possible; that is no such transition complexes were located, although WH4 also

74 has a tetrahedral structure. In other words, once chlorines are replaced by hydrogens via

σ-bond metathesis, the much stronger W-N bond as shown in Table 3-10, hampers the α- shift of hydrogen from the α-carbon to nitrogen. Thus, the use of nitrogen carrier drives only WCl4 formation via α-shift of hydrogen, blocking σ-bond metathesis. Then, ammonia can be a separate source of nitrogen, as well as surface hydrides for the removal of chlorines [Wid00]. This hypothesis is confirmed by the report that films grown using

2 with NH3 had increased nitrogen levels and decreased carbon and oxygen levels relative to films grown without NH3 over the entire deposition temperature range (450 to

700 °C) [Bch04].

This customized reaction scheme also poses the possibility of using each precursor

1-3 as an ALD (Atomic Layer Deposition) precursor, i.e., bare W structure from WCl4 with ammonia as the nitrogen source (two step ALD scheme). In addition, tungsten imido precursors 1-3, their moieties 1a-3a, and their hydride derivatives via σ-bond metathesis have large dipole moments as shown in Table 3-12. Considering the driving force for physisorption is the interaction between surface dipole and molecular dipole, strong physisorption should be induced [Lut93]. Moreover, because of the lack of an electronic barrier for rotation of a π-bonded ligand, the two components of dipole moment except a vertical one, cancel on average and physisorption would have the directionality, favorable for the layer-by-layer stacking in ALD growth.

3.4.9 Summary

Computational thermochemistry based on density functional theory (DFT) was employed to evaluate the design strategy of alkyl- and arylimido precursors (the phenylimido complex Cl4(CH3CN)W(NPh) (1), the isopropylimido complex

i Cl4(CH3CN)W(N Pr) (2), and the allylimido complex Cl4(CH3CN)W(NC3H5) (3)) for the

MOCVD growth of tungsten nitride and tungsten carbonitride (WNxCy) thin films. The impact of the imido substituent on properties of the deposited film and gas phase homogeneous decompositions of precursor molecules in the earlier stage of growth were also elucidated. Plots of calculated Gibbs energy vs. temperature verified that dissociation of the acetonitrile ligand (CH3CN) should be favorable for 1-3 in the temperature range used for film growth (> 450 °C). A computational search for transition states for chlorine removal by H2 was consistent with W-Cl cleavage via homogeneous

σ-bond metathesis with hydrogen. Natural bonding orbital (NBO) analysis and bond energy calculations indicated that 1 has the strongest N-C bond in the imido ligand and a slightly weaker W-N bond, consistent with W-N cleavage and concomitant low nitrogen content in films deposited from 1. Moreover, it was also shown that 2 and 3 have relatively stronger W-N bonds than 1, matching with the results of positive ion electron- impact (EI) and negative ion electron-capture chemical ionization (NCI) processes and corresponding to the increased nitrogen content in the films from 2 and 3.

C2 C1 C1 C2

Zr Zr

Figure 3-1. Optimized geometries for ZrNp4 (left) and ZrBn4 (right). Hydrogen atoms are omitted for clarity.

(a)

+ CMe4 1a-t 1a

(b)

ZrNp4

+ CMe4 1b-t 1b

Figure 3-2. Initial decomposition step for ZrNp4; (a) α-hydrogen abstraction and (b) γ- hydrogen abstraction.

Zr(H2C=CH2) +i-butene 78.6

27.2‡

3b+i-butene 51.4 2b-t 1b-t 42.3 41.1‡ 37.4 39.2‡ ‡ 2b+CMe4 37.4 10.3 1b+CMe4 ZrNp4 3.1 0.0

Zr(H C=CH ) 2b-t 2b 3b 2 2 Figure 3-3. Energetics of the reaction scheme (3-3) and computed structures of 2b-t (a transition state for the second γ-hydrogen abstraction), 2b, and 3b. Values are in kcal/mol.

N1 W

Cl N2

1 23

Figure 3-4. Optimized geometries for complexes 1-3

C N1 Cl W

1a 2a 3a

Figure 3-5. Optimized geometries for complexes 1a-3a

+ Stretching

Figure 3-6. Assumed vibrational mode leading to CH3CN dissociation from 1

-5

-10 1 -15

(kcal/mol) -20 o

G 2

∆ -25 3 -30

-35

-40 0 100 200 300 400 500 600 700 Deposition Temperature (oC)

Figure 3-7. Standard Gibbs energy change (∆G°) vs. temperature for CH3CN dissociation from complexes 1-3

EA = 15.8 kcal/mol 0

-1

ln(k1/T) -2

-3

-4 0.0033 0.0034 0.0035 0.0036 0.0037 0.0038 0.0039 1/T

Figure 3-8. Plot of ln(k/T) vs. 1/T for acetonitrile exchange in complex 2

+ H2 -HCl

36.8 -22.1 kcal/mol 1a TS-1

+ H2 -HCl

37.2 -23.3 2a MI-1 TS-2

+ H2 - HCl

36.2 -22.3 3a TS-3 Figure 3-9. Calculated transition states (center) and products (right) for σ-bond metathesis of 1a-3a with hydrogen

54.9 -22.2

MI-1 RT-1 RI-2

38.4 39.0

MT-1 trans MT-1 cis

-16.5 -26.1

MI-2 trans MI-2 cis Part (a)

RT-2 trans RT-2 cis

50.8 59.5

MI-2 trans MI-2 cis

-11.0 32.6 44.8 -10.6

RI-3 MT-2 trans MT-2 cis RI-3

-19.2 -22.3

MI-3 Part (b)

-4.3 74.4

RT-3 RI-4

52.6 MI-3 -6.3

MT-3 MI-4 Part (c)

Figure 3-10. Calculated transition states and intermediates for σ-bond metathesis (solid arrows) and reductive elimination pathways (dotted arrows).

Figure 3-11. Conjugation of N lone pair to metal d and through the phenyl ring as observed in the HOMO-18. Contours are drawn at ± 0.015 a.u.

-0.1884 1.4248 -0.2081 -0.3793 1.4666 -0.6364 1.9876 -0.1999 -0.1802 1.0105 1.3281 0.9990 0.10618 -0.0352 -0.2761 1.1243 0.9763 1.0090 -0.3666 -0.3487 -0.3407 1.8864 2.0412 2.0079

0.9277 0.9225 0.9227

-0.2632 -0.2673 -0.2636

1 2 3

Figure 3-12. Atomic charges (underlined) from NPA and Wiberg indices for 1-3. Hydrogens are omitted for clarity.

N1-C Cleavage +

W W-N1 Cleavage +

Figure 3-13. Geometries of the products of W-N1 and N1-C bond dissociation from 2a

%) 2a

8 nc. (at.

6 en Co 3a 4 Nitrog 1a

0 300 400 500 600 700 800

Deposition Temperature (°C)

Figure 3-14. Comparison of nitrogen content in the films grown from 1a-3a (AES) [Bch05].

β-hydrogen abstraction

+ [Cl3WNH] + Propylene

(a1) i + [Cl3W(N Pr)] Transition

α-shift + [Cl3W] + HNC(CH3)2

(a2) i + [Cl3W(N Pr)] Transition Intermediate

+ [Cl3WNH] + Propadiene

(b1) + [Cl3W(NC3H5)] Transition

+ [Cl3W] + HNCHCHCH2

(b2) + [Cl3W(NC3H5)] Transition Intermediate

i + Figure 3-15. Two possible decomposition pathways of [Cl3W(N Pr)] or + i + [Cl3W(NC3H5)] ; (a1) β-hydrogen abstraction from [Cl3W(N Pr)] , (a2) α- i + + shift from [Cl3W(N Pr)] , (b1) β-hydrogen abstraction from [Cl3W(NC3H5)] , + (b2) α-shift from [Cl3W(NC3H5)] .

+ (a) ∆ = 4.5 [Cl3W] + HNC(CH3)2 82.8 87.3

59.0 + [Cl3WNH] + Propylene 41.2 23.1 i + [Cl3W(N Pr)]

86.2 ∆ = 1.8 + (b) [Cl3W] + HNCHCHCH2 84.4 55.4 + [Cl3WNH] + Propadiene 45.0

22.5 + [Cl3W(NC3H5)]

Figure 3-16. Energetics of two possible pathways for the decomposition of (a) i + + [Cl3W(N Pr)] and (b) [Cl3W(NC3H5)] . Electronic energy has the unit of kcal/mol.

(a) 83.6 64.5 Cl4W + HNC(CH3)2 38.8 Cl4WNH + Propylene 33.4 i Cl4 W(N Pr), 2a 31.3 89.1 (b) 60.1

Cl4WNH + Propadiene 37.1 35.8 Cl4W + HNCHCHCH2 Cl W(NC H ), 3a 4 3 5 27.7

Figure 3-17. Energetics of two possible pathways for the decomposition of (a) 2a and (b) 3a. Electronic energy has the unit of kcal/mol.

Table 3-1. Bond lengths (Å) and bond angles (°) for ZrBn4

Experimentala,b Calculateda Zr-C1 2.255(3) 2.288 C1-C2 (Å) 1.458(7) 1.480 Zr-C1-C2 (o) 92.5(26) 97.94 aAverage value for the four equivalent alkyl groups. bData are taken from Tedesco et al. [Ted98].

a Table 3-2. Atomic charges and Wiberg indices for ZrR4 complexes from NBO analysis.

Atomic charges Wiberg indices

ZrNp4 ZrBn4 ZrNp4 ZrBn4 Zr 1.9586 1.8890 0.7236 0.6515 C1 -0.9347 -0.8711 0.9926 1.0977 C2 0.0423 -0.0505 aAverage value for the four equivalent alkyl groups.

Table 3-3. Energetics of possible pathways for ZrNp4 decomposition

o o ∆H 298 or ∆G 298 or o,‡ o,‡ Entry Reactions ∆H 298 ∆G 298 Note (kcal/mol) (kcal/mol) ‡ ‡ 1 ZrNp4 → 1a-t 40.6 41.9 α-H abs., DFT ‡ ‡ 2 ZrNp4 → 1b-t 37.4 40.2 γ-H abs., DFT

3 ZrNp4 → 1a + CMe4 30.1 18.0 DFT

4 ZrNp4 → 1b + CMe4 3.1 -7.9 DFT

5 ZrNp4 → Np3Zr• + •Np 57.1 42.6 Homolysis, DFT ‡ ‡ 6 ZrNp4 → 1a-t 60.3 61.6 α-H abs., HF ‡ ‡ 7 ZrNp4 → 1b-t 55.5 58.5 γ-H abs., HF

Table 3-4. Calculated bond lengths (Å) and bond angles (°) for complexes 1-3

1 2 3 Calc. Calc. Calc. Exp.b W-Cla 2.380 2.377 2.375 2.333 W-N1 1.738 1.715 1.716 1.687 (9) W-N2 2.262 2.347 2.342 2.308 (8) C-N1 1.363 1.428 1.425 1.508 (17) Cl-W-N1a 95.9 98.1 98.1 97.4 W-N1-C 180.0 179.6 177.7 167 (2) aAverage value for four equivalent chlorides. bExperimental values from the X-ray crystal structure of 3 [Bch05].

Table 3-5. Calculated bond lengths (Å) and bond angles (°) for complexes 1a-3a

1a 2a 3a W-Cla 2.344 2.346 2.345 W-N1 1.716 1.702 1.703 C-N1 1.372 1.440 1.436 Cl-W-N1a 102.5 102.6 102.7 W-N1-C 180.0 179.0 178.3 aAverage value for the four equivalent chlorides.

Table 3-6. Reaction enthalpies for CH3CN dissociation from complexes 1-3

o o ∆Hf (1a) + ∆Hf° (CH3CN) - ∆Hf (1) 10.4 kcal/mol o o ∆Hf (2a) + ∆Hf° (CH3CN) - ∆Hf (2) 10.2 kcal/mol o o ∆Hf (3a) + ∆Hf° (CH3CN) - ∆Hf (3) 11.0 kcal/mol

Table 3-7. Refined activation enthalpies and entropies for the first step σ-bond metathesis of 1a-3a

∆Ho,‡ ∆So,‡ 298 298 (kcal/mol) (cal/mol·K) 1a 36.2 -29.1 2a 37.2 -29.3 3a 36.2 -29.4

Table 3-8. Bond dissociation enthalpies (∆Ho, kcal/mol) for the N1-C and W-N1 bonds in 1a-3a

N1-C W-N1 1a 121.3 80.0 2a 98.4 88.2 3a 82.7 86.4

Table 3-9. Comparison of deposition behavior for 1a-3a.

Growth temp. Growth rate N/W E (eV) (°C) (Å/min) a atom ratio 1a 475 – 750 2 – 21 1.41 ± 0.28 0.01 – 0.03 2a 450 – 700 10 – 27 0.84 ± 0.23 0.05 – 0.16 3a 450 – 650 5 – 10 0.15 ± 0.13 0.05 – 0.17 Data are taken from Bchir et al. [Bch05].

Table 3-10. Bond dissociation enthalpy (∆Ho, kcal/mol) for the N1-C and W-N1 bonds in σ-bond metathesis products

N1-C W-N1

H4W(NPh) 102.2 99.7 i H4W(N Ph) 82.8 106.9

H4W(NC3H5) 67.9 106.3

Table 3-11. Summary of relative abundances for positive ion EI mass spectra of tungsten imido complexes, 2 and 3

Abundancea EI fragments m/z Isopropyl (2) Allyl (3) i + [Cl3W(N Pr)] + 348 / 346 100 100 / [Cl3W(NC3H5)] + [Cl4W] 326 26 34 + [Cl3WNH] 306 78 12 + [Cl3W] 291 30 58 + [CH3CN] 41 24 95 aRelative abundances were adjusted by summing the observed intensities for the predicted peaks of each mass envelope and normalizing the largest sum to 100%. Data are taken from Bchir et al. [Bch03b, Bch05].

Table 3-12. Calculated dipole moments of 1a-3a and their derivatives via σ-bond metathesis

Dipole moment Moieties (Debye)a Cl W(NPh) 6.29 Phenyl 4 H4W(NPh) 2.85 i Cl4W(N Pr) 5.48 Isopropyl i H4W(N Pr) 2.40

Cl4W(NC3H5) 5.34 Allyl H4W(NC3H5) 2.80 aFrom Mulliken population analysis

CHAPTER 4 GROWTH OF ZrC THIN FILMS BY AEROSOL-ASSISTED MOCVD

4.1 Introduction

The equilibrium analysis presented in the Chapter 2 suggests a limited range of growth conditions exists that yields stoichiometric ZrC. This range expands with increasing pressure at constant H/Zr and C/Zr and with increasing hydrogen to prevent codeposition of elemental carbon. Furthermore, the deposition is possible only at a relatively low temperature (see Figure 2.2). Thus a good precursor should show a low decomposition temperature and high vapor pressure. Although metalorganic precursors typically have a low thermal decomposition temperature, the low volatility (i.e., vapor pressure Pi°(T)) of most single source precursors like tetraneopentyl zirconium (ZrNp4) limits their partial pressure (yiP=xiPi°(T) for an ideal gas/solution where xi and yi are gas and solution mole precursor fractions and P is the total pressure). Thus a high pure component vapor pressure and high mole fraction in the solvent, if used, favor efficient transport.

To deliver reactant at supersaturation conditions, aerosol-assisted metalorganic chemical vapor deposition (AA-MOCVD) is used in this study. The system operates near atmospheric pressure and the use of aerosol delivery can provide a precursor partial pressure to the growth surface in excess of that at the source condition. A solution of either ZrNp4 or tetrabenzyl zirconium (ZrBn4) in benzonitrile (PhCN) was nebulized as an aerosol and transported to the growth surface by a carrier gas. Films were grown at different values of temperature, total pressure, H/Zr inlet molar ratio, and carrier gas (H2,

89 90

He, or N2). The film composition was characterized by AES and the existence of Zr-C chemical bonds was confirmed by XPS. In addition, XRD and AFM were used to evaluate the crystallinity and surface morphology of the deposited films, respectively.

The film thickness was measured by X-SEM.

4.2 AA-MOCVD System Description and Growth Procedure

The system used for growth of ZrC in this study is termed AA-MOCVD and a reactor system was design and built for this research project. The PFD (Process Flow

Diagram) of the system is shown in Figure 4-1. The hydrogen gas cylinders were stored outside the building. The connection points are labeled in Figure 4-1 as 1 and 2, respectively. The colored lines were unused for this application.

In this study, films were grown from single source precursor solutions at temperatures in the range 400 to 600 °C by AA-MOCVD on p-type B-doped Si(100) and

Si(111) substrates with resistivity in the range 1 to 2 Ω·cm. The growth was performed in a horizontal, cold-wall quartz reactor with a graphite susceptor, equipped with a radiofrequency (RF) heater and load-lock chamber as shown in Figure 4-2. Precursor solutions were subjected to ultrasonic nebulization, which resulted in small droplets with a controlled size, as schematically shown in Figure 1-14 in Section 1.5.3. Hydrogen, nitrogen, or helium (99.999% purity) was used as a carrier gas to transport the droplets, generally at a reactor pressure of 300 Torr and a flow rate of 1.5 slm.

Before loading the Si substrates (approximately, 1 × 1 cm2 pieces) into the reactor, they were first cleaned and degreased in boiling TCE (trichloroethylene), acetone, and methanol successively for 5 min each. Then, any surface oxide on the Si was etched for

2 min in a commercial buffered oxide etchant and dipped in boiling water for 30 sec before they were dried with a commercial air-jet spray.

After substrate preparation, they were next placed on a quartz tray (see Figure 4-2), which was physically resting on the fork of the load-arm in the load-lock chamber. Then, the load-lock chamber was evacuated using an Edwards Model XDS5 dry vacuum pump to minimize possible oxygen contamination. Pressurizing with nitrogen and re- evacuating the load-lock chamber were repeated several times to reduce the background impurity level. Since the load-lock chamber was isolated from the reactor by an 8” manual gate valve, the pressure of the load-lock chamber was next equalized to the system pressure before the gate valve was opened. The reactor chamber was pre-baked in hydrogen at 860 °C for 20 min to remove any unwanted hydrocarbon or residual moisture on the reactor surfaces. The system pressure was then stabilized with at the operating pressure, typically in the range 300 to 350 Torr. The reactor was pumped using a Leybold TRIVAC wet rotary vane pump. The quartz tray was loaded using the mechanical fork of the load-arm. The arm was moved by a computer-controlled (IMS terminal) motorized feed-through and was lowered at a pre-determined position to set the quartz tray on the graphite susceptor. Finally, the load-arm was retracted into the load- lock chamber and the system was isolated again by closing the gate valve. The system was thus prepared for growth.

Note that susceptor and thermocouple assemblies were custom-designed to fit into the system, as shown in Figure 4-2. The graphite susceptor was heated by RF induction heating and the temperature was measured with a type K thermocouple (chromel-alumel,

0 to 1260 °C) assembly, which was custom-designed to allow the thermocouple bead to fit inside the graphite susceptor from the rear. The digital readout connected to the

92 thermocouple indicates the operating temperature, controlled manually with by adjusting the RF power.

The flow rate of the carrier gas was also computer-controlled using a mass flow controller (MFC) from Chunma Data Systems and their compatible software, Flow

Manager. A MKS model 651 self-tuning/digital PID controller was used for pressure control combined with a MKS type 253B throttle valve and a MKS model 121 Baratron capacitance transducer for pressure readout. To cope with the possible back-flow of pump oil and to prevent the solvent (PhCN) from reaching and absorbing into the O-ring inside the throttle valve, a zeolite regenerable trap was installed just before the throttle valve. Swelling of O-ring caused the throttle valve to close and thus total failure of pressure control. Routine maintenance included replacement of the O-ring every 25 hr

(10 runs) of operation. The carrier gas cylinders were stored outside of the building for safety.

The sample experienced a temperature ramp before and after the growth. The first ramp was used in the sample baking step and the second one in the film annealing (if necessary) step. They were performed at 860 °C, for 15 and 30 min, respectively. The sample baking in hydrogen is expected to reduce surface oxides on the substrate surfaces, although AES sputter profiles suggested a presence of surface oxide.

The ultrasonic nebulizing system, mentioned in Section 1.3.3, has three auxiliary units. A Fisher Scientific syringe infusion pump was used to meter the precursor solution

(precursor/solvent). The feed rate of the solution used was 4 to 5 ml/hr and the total amount of solution was typically 10 ml. An Upchurch Scientific U-605 back pressure regulator was used to isolate the syringe contents from evaporation into the system’s

93 vacuum line. The Cetac Technologies nebulizer unit was used to generate an aerosol from the precursor solution. The nebulizer contains a piezoelectric quartz transducer, which vibrates at a frequency of 1.44 MHz and is finely controlled by their autotune power supply. This high frequency vibration resulted in the ejection of liquid droplets from the transducer plate to form an aerosol with a particle diameter of about 2.5 µm. To minimize condensation of the aerosol on the surface of stainless metal structure, the carrier gas line from the nebulizing system to the reactor and the nebulizing chamber were moderately heated (approximately, 50 °C) with heating tapes, controlled by an

Autotransformer. The nebulizing chamber is a custom-design.

4.3 Precursor Synthesis

4.3.1 Precursor Candidates

ZrNp4 was synthesized by Dr. McElwee-White’s group at University of Florida and tested in this study. ZrBn4 was purchased from Sigma-Aldrich Corp (min. 95% purity).

A third single source precursor, ZrNp3Cl, was also synthesized but testing this precursor is beyond the scope of this dissertation.

4.3.2 General

Standard Schlenk and glovebox techniques were employed in the synthesis of

ZrNp4. Neopentylmagnesium chloride (Me3CCH2MgCl) was prepared according to a literature procedure [Hug97]. Solvents were purchased from Fisher and passed through an M. Braun MB-SP solvent purification system prior to use. ZrBn4 was purchased and used without further purification.

4.3.3 Tetraneopentyl Zirconium (ZrNp4)

Tetraneopentyl zirconium was prepared by the following modification of a literature procedure [Hug97]. A slurry of ZrCl4 (0.629 g, 2.70 mmol) and ether (60 ml)

was cooled to 0 ºC and treated with 3 equiv. of Me3CCH2MgCl (8.10 ml, 1.0 M in ether) added drop wise over a period of 30 min. The reaction mixture was stirred at 20 ºC for

18 hrs to obtain a pale brown cloudy solution.

Ether was removed under vacuum to afford a pale yellow solid. The solid was extracted with hexanes (70 ml), the solution portion was collected by filtration through a

1 cm pad of Celite, and the volatiles were removed under reduced pressure to afford a pale brown solid. The solid was sublimed (75 °C, 0.02 mmHg) to produce tetraneopentyl zirconium as an off-white solid. Yield: 0.523 g, 70% based on neopentyl magnesium chloride. The product was identified by comparison to literature data [Hug97].

4.3.4 Trineopentyl Zirconium Monochloride (ZrNp3Cl)

A 10 mmol amount of zirconium tetrachloride was slurried in dry, degassed diethyl ether and cooled to 0 °C. A 30 mmol amount of tetraneopentyl zirconium was dissolved in 50 ml ether and added drop wise to the suspension over a one hour period. The reaction then proceeded at 0 °C for a period of 18 hrs leaving a bright yellow solution.

Volatiles were removed via reduced pressure and purity was assessed by 1H-NMR. The resulting solid was stored in the dark at -30 °C [Hug97].

4.4 Optimization of Growth Conditions

4.4.1 Comparison with Equilibrium Analysis

A series of ZrC films was grown at variable conditions to probe the efficacy of the equilibrium predictions presented in Chapter 2. The following is the summary of the conditions varied:

Temperature: 400 to 600 °C Pressure: 300 to 500 Torr Precursor concentration and infusion rate: 0.0177 M of ZrNp4 or ZrBn4 solution in PhCN at 4 ml/hr Carrier gas: H2, N2, or He

Flow rate of carrier gas: 0.75 to 1.5 slm

The base growth conditions (entry c in Table 4-1) were T = 500 °C, P = 300 Torr, and H/Zr ratio = 1.16×105 (corresponding to the flow rate of 1.5 slm) as indicated by the dashed vertical line in Figure 4-3. Table 4-1 compares the constitution of the film grown at these base conditions to that of films grown by changing one growth condition: T =

400 °C, T = 600 °C, P = 500 Torr, and H/Zr ratio = 5.95×104, as well as one grown with

He replacing H2 as the carrier gas. It is noted that the transport properties of these two carrier gases are similar and thus should develop similar thermal, mass, and hydrodynamic boundary layer characteristics. The film grown at the lower H/Zr ratio

(entry e) is predicted to be in the 1-phase (ZrC) solid region, while the replacement of H2 by He should push the system well into the 2-phase (ZrC + graphite) solid region. All films grown at the conditions listed in Table 4-1 gave carbide or oxycarbide peak in AES spectra or Zr 3d3/2, 3d5/2 peak separation in the XPS data. Phase determination from AES or XPS spectra is discussed in Section 4.5.3.

AES experimental data (using elemental sensitivity data embedded in the equipment) on the carbon content in the films deposited using pure H2 carrier gas were generally consistent with the equilibrium predictions shown in Figure 4-3. The growth conditions labeled as circles in Figure 4-3 lie well within the predicted single solid ZrC phase field (i.e., positioned below the solid curve in Figure 4-3). The carbon content of the films (see Table 4-1, entries b, d and f, post-annealed in H2 to reduce weakly bound oxygen) was nearly constant (41.0 to 42.3 at. %) throughout the growth temperature range (400 to 600 °C). By decreasing the H2 amount in half (see Figure 4-3, triangle and

Table 4-1, entry e), however, the measured C/Zr ratio increases along with the possibility

96 of elemental carbon codeposition, even though the system is in the single solid phase region.

Film growth with He carrier gas gave an unexpected result. Based on the thermodynamic analysis, the deposition conditions denoted as the square in Figure 4-3 and entry g in Table 4-1 should yield a higher C/Zr ratio in the film, in other words, apparent carbon codeposition with ZrC. The comparison of He with H2 carrier gas under equivalent conditions [see Table 4-1, entries c and g], however, showed a very similar

C/Zr ratio, which possibly indicated the formation of ZrC-rich solid phase. Post growth annealing of the film grown at the same condition (entry h in Table 4-1), however, revealed apparent carbon codeposition with ZrC, which is the same result as in the case of low H/Zr ratio (entry e in Table 4-1). In contrast to films grown in He at 500 °C, trials with He at a higher temperature (i.e. 600 and 700 °C) failed to produce any carbide or oxycarbide peak in AES spectra or Zr 3d3/2, 3d5/2 peak separation in the XPS data, although the individual presence of zirconium, carbon, and oxygen in the films was confirmed by EDS measurement (see Figure 4-4). Nitrogen was also tried as a carrier gas but it resulted in minimal deposition. The lower thermal conductivity of nitrogen compared to hydrogen or helium would increase the thermal boundary layer thickness, allowing parasitic gas phase reactions to be more likely.

In an attempt to reduce the oxygen content of the films, post growth annealing in

H2 at 860 °C for 30 min was explored. A reduction of oxygen content in the films was observed by annealing in H2, as shown by comparison of entries a and b, c and d, or g and h in Table 4-1. It was also shown that nitrogen incorporation into the films is related to the type of carrier gas and the growth temperature. As noted above, the carrier gas can

97 affect the thermal and mass diffusivities and thus the mass transfer rates and thermal profiles, including the surface temperature. It is noted that incorporation of benzonitrile fragments in WNx films grown from solutions of precursor complexes in PhCN solution had previously been observed [Bch04a].

The equilibrium calculations suggest that higher pressure at constant H/Zr and C/Zr molar ratios is more favorable for single phase ZrC growth. The experimental trial with a higher operating pressure of 500 Torr [see Table 4-1, entry i] produced films with compositions similar to those obtained with P = 300 Torr [see Table 4-1, entries b, d and f] at the same H/Zr ratio. The growth rates, however, were less for films grown at higher pressure or higher temperature, consistent with the increased role of parasitic gas phase reactions. Use of hydrogen as the carrier gas resulted in no deposition at higher temperature (i.e. 700 °C). It is suggested that the reactivity of H2 produced a greater extent of gas phase reactions compared to He carrier, which produced Zr-containing films at T = 700 °C. The films grown at T = 700 oC with He carrier, however, failed to show carbide or oxycarbide peaks in the AES spectra.

4.4.2 Summary of the Suggested Growth Conditions

Based on the calculated phase diagram and experimental results, the summary of the suggested growth conditions is listed in Table 4-2.

4.5 Film Characterization

The film structure was examined by X-ray Diffraction (XRD) using a Philips APD

3720, operating with Cu-K radiation. The film composition was determined by Auger

Electron Spectroscopy (AES) using a Perkin-Elmer PHI 660 Scanning Auger Multiprobe, while X-ray Photoelectron Spectroscopy (XPS) data were collected using a Perkin-Elmer

PHI 5100 ESCA System with standard Mg anode X-ray source. The film thickness was

98 estimated by cross-sectional Scanning Electron Microscopy (X-SEM) with Energy

Dispersion Spectroscopy (EDS) on a JEOL JSM-6400. The surface morphology was investigated by Atomic Force Microscopy (AFM) using a Digital Instruments Nanoscope

III. Detailed descriptions of the characterization techniques are given in the Appendix.

4.5.1 General

The films were smooth (AFM) and homogeneous (AES depth profile, see Figure 4- (a) 5c), and exhibited good adherence as evidenced by the lack of film peeling from the Si substrates. EDS indicated the presence of zirconium, carbon, and oxygen for all deposited films (see Figure 4-4). The color of the films was generally grayish black but Automatic peak search option used sometimes a yellowish or bluish deposit was obtained.

4.5.2 Structural Analysis

XRD measurements generally did not detect crystalline ZrC, Zr oxycarbides

(having the same fcc structure as ZrC with a negligible difference in lattice constant), or

ZrO2, regardless of the growth temperature.

In this study, a rather high flow rate of H2 carrier (1.5 slm) was chosen to supply a sufficient amount of hydrogen to prevent carbon codeposition in the deposited films. The carrier and aerosols were transported through ¼" nominal O.D. tubing, with the exit only

2" from the front of the graphite susceptor, as shown in Figure 4-2. The linear velocity inside the tubing is estimated to be ~1.0 m/sec. At this velocity, the substrate could be wetted by the aerosol to permit heterogeneous decomposition on the surface. In other words, there may not be enough time for the collision-induced stripping of the hydrocarbon ligands from the ZrNp4 molecules before they reach the surface. Surface reactions without sufficient preliminary gas phase decompositions could result in amorphous growth due to the size effect of ZrNp4 and elemental carbon inclusion.

The solvent PhCN participating in the deposition reactions could be an alternative explanation for the observance of amorphous growth. As shown in Table 4-1, nitrogen incorporation is likely to disappear above 600 °C; more specifically, between 500 to 600 oC. This temperature range is coincidentally matched with the crystallization temperature range of WCxNy films grown using aryl- and alkylimido precursors [Bch03a, Bch03b,

Bch05], where the same PhCN solvent was used. Moreover, the sharp decrease of nitrogen content in the deposited WCxNy films shows a maximum (Figure 3-14) with decomposition temperature that is consistent with the content observed in Table 4.1. A possible mechanism begins with CN radicals, produced from the solvent decomposition, that dissociatively adsorb on the surface, reacts with surface hydrogen radicals to form hydrogen cyanide (HCN), which finally desorbs as a stable gas phase species [Bch04a].

The radical formation rate increases with temperature as does HCN desorption, which may produce the maximum N content. In addition to changing the composition, competitive adsorption by both the solvent and precursor reactive species could possibly hamper the formation of the crystalline texture.

4.5.3 Chemical Composition and Phase Determination

AES survey data confirmed the existence of zirconium and carbon with contamination from oxygen and nitrogen (see Figure 4-5). The source of the nitrogen likely involves the solvent (PhCN) during film growth. The N content decreased with growth temperature and pressure (see Table 4-1). The oxygen and carbon levels show different profiles, suggesting different incorporation mechanisms. The source of the oxygen has not been determined but the high concentration near the surface is likely a result of post-growth handling of the samples in air prior to evaluation by AES. The oxygen depth profile (see Figure 4-5c) below the near surface region shows a flat profile,

100 indicative of background incorporation during the deposition process. Potential sources of oxygen in the film include impurities in the precursors or carrier gas and residual oxygen or moisture in the reactor, although since the reactor was equipped with a load- lock chamber this source should be minimized. The AES oxygen profile gradually increases as the film/substrate interface is approached. This increase may result from reaction of Zr with the SiOx surface layer on the Si prior to deposition (ZrO2 is ~ 21 kcal/g atom O more stable than SiO2) coupled with diffusion during the high temperature anneal.

The boxed regions highlighted in Figure 4-6 show the shape difference of the KLL transitions of the C1 peaks for the as-received surface (see Figure 4-5a) and after Ar+ sputtering for 60 sec (see Figure 4-5b). After sputtering, a new peak appeared around

262 eV, producing a "toothed" profile [Par94]. In addition, the intensity of the Zr1 peak increased after Ar+ sputtering [Par94]. Appearance of an intermediate peak between the elemental carbon and carbide peaks has been reported by Bellucci [Bel05] to indicate the formation of zirconium oxycarbide, in which carbon and oxygen incorporate randomly into the interstitial sites in the zirconium lattice (FCC) [Ber96]. The spectrum of a ZrC target, purchased for magnetron sputtering experiments, after Ar+ sputtering for 20 min is compared in Figure 4-5d. Some oxygen and carbon (a little higher C/Zr ratio = 1.12) resulted from highly rough or porous surface of the ZrC target because it had been used for magnetron sputtering experiments.

Figure 4-6 provides an expansion of the C1 peaks for films grown at 500 °C after 3 keV Ar+ sputtering for 60 sec. There are three peaks observed in (b), as compared to the two present in (a). The C1 peaks observed in (b), which are almost same as the enlarged

101

C1 peak of ZrC target (shown in Figure 4-5d), can be assigned to the carbide, oxycarbide, and elemental carbon phases [Bel05]. The intermediate peak from (a) is much broader compared to the spectrum obtained in (b); and the peak position is somewhere between the reported values for the oxycarbide and the carbide, which could indicate that two peaks (carbide and oxycarbide) overlap in this case.

Figure 4-7 shows how the C1 peak changes with Ar+ sputtering time. The sample deposited at 500 oC in He carrier without annealing (entry g in Table 4-1) was intentionally exposed to air for a long period of time to check the behavior of oxygen diffusion into the film. To the same level of oxygen content or C/Zr ratio as achieved in the growth at the same condition without annealing (entry g in Table 4-1), it took two or three times more sputtering time, as indicated in Table 4-3. Due to the amorphous film structure (less dense), oxygen diffusion from the air progressed deep into the films. The oxycarbide peak or the composite peak of oxycarbide and carbide, as explained in Figure

4-6, increased for the first 120 sec of sputtering time without any recognizable change of elemental carbon peak. The C1 peak showed two major peaks, assigned to elemental carbon and oxycarbide peaks as shown in Figure 4-6a, at a sputtering time of 240 sec.

The final bulk composition obtained at the sputtering times of 240 and 300 sec (see Table

4-3) had lower oxygen content and higher C/Zr ratio as discussed in Section 3.1, compared to the bulk composition of the film grown without annealing at the same condition (entry g in Table 4-1). Nitrogen, however, was not detected after annealing.

This result suggests that the chemical state of nitrogen might be that in a cyanide group

(CN), and not of ZrNx. The cyanide group is expected to be relaxed inside the film after the long period of air exposure before being removed by the annealing with hydrogen.

102

For clarity, the reason that AES composition after Ar+ sputtering for 60 sec was assumed to be the bulk composition of the film is there was generally not much difference in composition between 30 and 60 sec sputtering.

XPS measurements were carried out to confirm the existence of Zr-C bonding

(carbide phase) in the deposited films. Figures 4-8a and 4-8b show spectra for films grown using He as the carrier gas. Deconvolution of the C 1s peak demonstrated that there are two different states, the higher binding energy is associated with elemental carbon, and the other from carbide [Smi93, Ber96]. Observation of the Zr 3d3/2 and 3d5/2 peaks clearly confirmed the existence of the carbide phase. Upon fixing the intensity ratio of 3d3/2 to 3d5/2 to that of elemental zirconium (2/3 : 1), deconvolution of the whole peak indicates that two binding states exist: a higher binding energy oxide and a lower energy carbide [Smi93]. Direct comparison of peak heights for both states, however, shows the oxide is dominant relative to the carbide when a He carrier gas is used. XPS depth profiling of the film deposited in H2 carrier (see Figure 4-8c) suggested the improvement of film quality, i.e. enriched in carbide. Although it is more difficult to deconvolute the Zr peak in Figure 4.8c compared to that in Figure 4-8b, the apex of the peak is clearly shifted towards carbide state. It is also consistent with reduced oxygen content of the corresponding growth condition (entry b) shown in Table 4-1, relative to the other condition with He carrier (entry g). AES or XPS spectra generally showed the existence of elemental carbon against the expectations from the thermodynamic analysis and discussions in Section 4.4.1.

4.5.4 Surface Morphology and Film Thickness

The surface morphologies of the ZrC films grown at different temperatures are shown in Figure 4-9. All films were grown in H2 carrier gas followed by annealing in H2

103 at 860 °C for 30 min. As expected, the low surface diffusion of adatoms at the lowest temperature nucleates a greater number of grains on the surface (see Figure 4-9a). As the temperature increases, the rate of adatom surface migration becomes greater, which results in larger grains as shown in Figure 4-9c [Smi93].

Based on X-SEM measurements (see Figure 4-10) and AES depth profiles (not shown here), the film thickness generally decreased with increasing temperature above the base temperature of 500 °C. This result is consistent with gas phase parasitic reactions occurring to a greater extent at the higher temperature to reduce the reactant concentration at the growing surface. No deposition was observed at 700 °C using hydrogen as the carrier gas. The film thickness at the base growth condition was in the range of 200 to 300 nm to give a relatively low growth rate of about 2 nm/min.

4.5.5 Magnetron Sputtering Growth for Comparison

ZrC films were also grown by magnetron sputtering for comparison. The as-grown films were amorphous as judged by XRD and post-growth annealing leads to formation of ZrO2 crystalline phase, as shown in Figure 4-11. The color of the films was whitish azure. The carbon content in the deposited films was far below the stoichiometric ratio

(1:1) C:Zr; Zr 65.9%, C 7.5%, O 26.6% by AES. It seems that only Zr atoms are sputtered from the ZrC target surface, which leads to Zr metal deposition. ZrO2 is formed from the post-growth oxygen incorporation due to air exposure after film removal from the UHV magnetron sputtering system. As discussed in Chapter 1.3.1, the crystal structure of ZrC consists of a bonded fcc Zr metal structure with C inserted into the interstitial sites and thus the wide range of homogeneity for ZrC. This suggests that the

Zr-C bonding in its crystal structure is not purely covalent. Instead, an average of covalent bonds in the unit cell produces the stoichiometric ratio (1:1) of ZrC according to

104 the octet rule. Thus, when highly excited Ar+ ions impinge on the ZrC target surface, the sputtering yield of Zr is considerably greater than that of C. This higher yield allows Zr to form the fcc metal structure on the surface. At room temperature, however, there is not sufficient energy for insertion of carbon atoms into the Zr metal fcc structure.

Increasing the growth temperature or adding another carbon source (reactive magnetron sputtering in an argon-methane atmosphere) could enhance the carbon incorporation into the films [Bru93, Che05].

4.6 Summary

Thin films of ZrC have been grown on Si (111) substrates in the temperature range from 400 to 600 °C by aerosol-assisted metalorganic chemical vapor deposition (AA-

MOCVD) from tetraneopentyl zirconium (ZrNp4) in benzonitrile (PhCN). Preliminary equilibrium calculations presented in Chapter 2 suggested that a minimal level of H2 was needed to prevent codeposition of solid carbon, and the minimum level decreased with decreased temperature. Films were grown at different values of temperature, pressure, and H/Zr inlet molar ratio, as well as carrier gas (H2, He, or N2), and all films were judged by their XRD patterns to be amorphous. The trends in the measured composition

(AES) of each film were qualitatively consistent with those of the equilibrium calculation, while XPS analyses confirmed formation of Zr-C and Zr-O bonds. It was not possible to grow films in N2 carrier gas.

105

Figure 4-1. Process flow diagram for the AA-MOCVD system

Quartz tray Graphite susceptor

Quartz boat

(a) (b)

Thermocouple (c) (d) ¼" inlet port RF induction coil

106

Quartz transducer Nebulizing chamber

(e) (f) Fork of the load arm Syringe infusion pump

RF Load-lock generator assembly

Nebulizer assembly Reactor assembly

(g) Figure 4-2. Images of the AA-MOCVD system; (a) susceptor assembly, (b) schematic of the susceptor assembly, (c) horizontal cold-wall quartz reactor, (d) schematic of thermocouple assembly, (e) the load-lock chamber, (f) the nebulizer assembly, and (g) the whole system.

107

800 P = 300 Torr C(graph.) + ZrC + Gas 700

(f) C) o 600

He carrier (e) 500 (g), (h) (c), (d) ZrC + Gas Temperature (

400 (a), (b) Base value 300 0 30000 60000 90000 120000 150000

Inlet H/Zr

Figure 4-3. Comparison of computed deposition phase diagram for ZrC deposition at 300 Torr by MOCVD using a ZrNp4 solution (0.0177 M in benzonitrile) in either H2 or He carrier gas (see Table 4-1 for experimental coding of data).

Figure 4-4. EDS spectrum of the ZrC film deposited in a helium environment at 500 °C

108

1500 (a) 1000

500

0 dN(E) Zr1 N1 Zr2 -500 3.7 % 29.9 %

-1000

-1500 O1 C1 18.7 % 47.7 % -2000 50 250 450 650 850 1050 1250 1450 1650 1850 2050

Kinetic Energy (eV)

1500 (b) 1000

500

0 dN(E) 262 eV Zr2 Zr1 N1 -500 39.2 % 7.4 % -1000 O1 11.1 % -1500 C1 42.3 % -2000 50 250 450 650 850 1050 1250 1450 1650 1850 2050 Kinetic Energy (eV)

Si2

O1 Zr2 Area counts N1

0 20406080100 Sputter cycles (0.2 min/cycle)

109

300

200 (d)

100

dN(E) -100

-200 Zr1 O1 Zr2 -300 5.8 % 44.4 %

-400

-500 C1 49.8 % -600 50 250 450 650 850 1050 1250 1450 1650 1850 2050

Kinetic Energy (eV)

Figure 4-5. AES spectra and depth profiles of a ZrC film deposited on Si (111) at 400 °C with H2 carrier gas followed by annealing in H2 at 860 °C for 30 min; (a) as received surface, (b) 3 keV Ar+ sputtered for 60 sec, (c) 3-point depth profile, (d) ZrC target with 3 keV Ar+ sputtered for 20 min.

(a) (b) Elemental carbon Elemental carbon Carbide

Carbide + Oxycarbide Oxycarbide

He carrier H2 carrier

240 250 260 270 280 290 240 250 260 270 280 290

Kinetic Energy (eV) Kinetic Energy (eV)

Figure 4-6. AES survey data for the ZrC films deposited on Si (111) substrates at 500 °C; (a) in He carrier (b) in H2 carrier followed by annealing in H2 at 860 °C for 30 min. Both spectra were taken after 3 keV Ar+ sputtering for 60 sec.

110

400 120 sec 200 60 sec 30 sec 0

dN(E) -200

-400 As received

-600

-800 240 sec

-1000 240 250 260 270 280 290 Kinetic Energy (eV) Figure 4-7. AES survey data with Ar+ sputtering time, for the ZrC film deposited on Si (111) at 500 °C in He carrier and annealed at 860 °C for 30 min following a long period (> 3 months) of air exposure.

(a) Elemental C 284.5 eV

Carbide 282.9 eV

290 289 288 287 286 285 284 283 282 281 280

Binding Energy (eV)

(b) Oxide 185.1 eV, 182.7 eV Carbide 183.5 eV, 181.1 eV

188 187 186 185 184 183 182 181 180 179 178 Binding Energy (eV)

111

30 min 50 min

188 187 186 185 184 183 182 181 180 179 178 Binding Energy (eV) Figure 4-8. XPS data for films deposited at 500 °C in He carrier; (a) C 1s peak and (b) Zr + 3d3/2, 3d5/2 peaks. Both spectra were measured after 3 keV Ar sputtering for

30 min. XPS depth profiling for the film deposited at 500 °C in H2 carrier (c).

(a) (b) (c)

RMS roughness 0.730 nm 1.735 nm 3.631 nm

Figure 4-9. AFM surface images (1 µm × 1 µm) of the films deposited at: (a) T = 400 °C, (b) T = 500 °C, and (c) T = 600 °C.

(a) (b)

1 µm 1 µm

Figure 4-10. X-SEM micrographs of ZrC films deposited in a helium environment at (a) 700 °C and (b) 600 °C.

112

Figure 4-11. XRD and GIXRD (Grazing Incident X-ray Diffraction) spectra of a ZrC film deposited on Si (111) at room temperature by magnetron sputtering using argon; as deposited and annealed in H2 at T = 860 °C for 30 min.

Table 4-1. Bulk composition of ZrC films grown at various growth conditions

Entry a b c d e f g h i Growth 400 400 500 500 500 600 500 500 400 temperature (°C) Growth 300 300 300 300 300 300 300 300 500 pressure (Torr) Carrier gas & H2 H2 H2 H2 H2 H2 He He H2 Flow rate (slm) 1.5 1.5 1.5 1.5 0.75 1.5 1.5 1.5 1.5 Annealing N Y N Y Y Y N Y Y 30 min, 860 °C Zr (at. %)a 34.5 39.2 42.8 44.4 38.7 44.0 45.6 38.9f 44.3 C (at. %)a 42 42.3 32.9 42.3 48.2 41.0 34.6 48.6f 41.8 O (at. %)a 16.9 11.1 18.6 13.4 9.9 15.0 16.1 12.5f 13.9 N (at. %)a 6.6 7.4 5.7 - 3.2 - 3.7 - - C/Zr atomic ratio 1.22 1.08 0.77 0.95 1.25 0.93 0.76 1.25 0.94 ZrO C c ZrO C c Detected solid ZrO C c ZrO C c ZrO C c x y ZrO C c x y ZrO C c ZrO C c ZrO C c x y x y x y , ZrCe, x y , ZrCe, x y x y x y b , Cd , Cd , Cd , Cd , Cd , Cd , Cd phases C C aValues measured by AES after 3 keV Ar+ sputtering for 60 sec. bDetermined from AES spectra (see Figure 4-6). cZirconium oxycarbide. dElemental carbon. eZirconium carbide. fValues measured by AES after 3 keV Ar+ sputtering for 240 sec after annealing at 860 °C for 30 min following a long period of air exposure.

113

Table 4-2. Summary of the suggested growth conditions for ZrC AA-MOCVD growth using ZrNp4

Optimum Parameter Remarks conditions Minimum was set to overcome the T = 500 °C activation energy of the initial Temperature opt (400 to 600 °C) decomposition. Maximum was set to prevent carbon codeposition. The higher, the better according to Pressure 300 Torr equilibrium analysis. Reducing environment is essential to Carrier gas H 2 prevent carbon codeposition. Flow rate of carrier gas 1.5 slm H/Zr ratio = 1.16×105 It depends on the solubility of the Precursor concentration 0.0177 mol/L precursor. This value is the maximum with PhCN as solvent.

Table 4-3. Composition with Ar+ sputtering time of ZrC film deposited on Si(111) at 500 °C in He carrier and annealed at 860 °C for 30 min following a long period (> 3 months) of air exposure

Sputtering time C (at. %) O (at. %) Zr (at. %) C/Zr ratio As received 31.6 27.1 41.3 0.77 30 sec 25.5 26.9 47.6 0.54 60 sec 29.1 23.2 47.7 0.61 120 sec 36.2 19.6 44.1 0.82 180 sec 41.9 15.2 43.0 0.97 240 sec 48.6a 12.5a 38.9a 1.22a 300 sec 47.0 13.9 39.1 1.20 aSame values as in entry h in Table 3-1.

CHAPTER 5 CONCLUSIONS AND RECOMMENDATIONS FOR FUTURE WORK

5.1 Conclusions

Growth of thin films of zirconium carbide (ZrC) as a promising candidate for a cathode or coating material in the application of field emitter arrays (FEAs) from tetraneopentyl zirconium (ZrNp4) by aerosol-assisted metalorganic chemical vapor deposition (AA-MOCVD) has been demonstrated.

Equilibrium analysis of the Zr-C-H-(Cl) system indicated the thermodynamic feasibility of CVD for stoichiometric ZrC growth without carbon codeposition. The analysis was applied specifically to the deposition of ZrC using ZrNp4 as well as novel chloride precursors. For ZrNp4 a key idea was to bypass the limit of low volatility of

ZrNp4 by aerosol transport and counteract the added C from the solvent towards methane formation, C (graphite) + 2 H2 ↔ CH4, by adding H2 and operating at higher pressure and lower temperature. The concept was tested experimentally but the temperature range for stoichiometric ZrC deposition was not sufficiently wide to overcome the activation energy for α- or γ-hydrogen abstraction, the initial step to forming the Zr=C bond by neopentane elimination.

Amorphous zirconium oxycarbide films, as evidenced by XRD spectra, were grown and confirmed by AES and XPS measurements. It was observed that a minimum amount of hydrogen and a maximum temperature were required to grow films with C/Zr

114 115 ratio close to the stoichiometric value (C/Zr = 1). These experimental results were in qualitative agreement with the thermodynamic predictions.

Facile Zr-C bond cleavage due to the resonance stabilization of the benzyl radical

(PhCH2•) is consistent with the lack of growth from ZrBn4. In contrast, qualitative and quantitative analyses demonstrated that the neopentyl groups of ZrNp4 have higher Zr-C bond dissociation energies and other decomposition processes (α- or γ- hydrogen abstraction) are more likely to occur than Zr-C homolysis in ZrNp4. A probable gas phase decomposition pathway of ZrNp4 was examined by the computational thermochemistry. While the preference for γ-hydrogen abstraction of neopentane over α- hydrogen abstraction was confirmed in the initial state of ZrNp4 decomposition, they turned out to be competing. An intermediate, Zr(H2C=CH2), is believed to be important for the ZrC deposition as a surface reactant given its empty coordination site. Thus, C/Zr ratio in the deposited films, however, will likely be greater than the stoichiometric value

(C/Zr = 1) without considering the reducing environment of a hydrogen carrier, which is very effective for removing elemental carbon according to the equilibrium analysis.

Computational thermochemistry proved to be general tool to study the gas phase decomposition mechanism of metalorganic precursors. As a case study, several reaction pathways for growth of WNx films from the isopropylimido complex

i Cl4(CH3CN)W(N Pr) (1), the phenylimido complex Cl4(CH3CN)W(NPh) (2), and the allylimido complex Cl4(CH3CN)W(NC3H5) (3), whose experimental results are available, were analyzed. The calculation results are consistent with facile dissociation of the acetonitrile ligand (CH3CN) from 1-3 in the temperature range used for CVD.

Computational study of reaction of the coordinatively unsaturated complexes 1a-3a with

116

H2 located possible transition states for chloride loss via σ-bond metathesis with hydrogen to yield HCl, the experimentally observed chlorine-containing product in the reactor effluent. Finally, through qualitative and quantitative theoretical analyses for

N(imido)-C and W-N(imido) bonds, nitrogen content in the films grown from 1-3 were linked to the calculated bond dissociation energies.

Several precursors were tested for the LaB6 film growth, but none were successful.

With its complex crystal structure, especially the B6 octahedra placed at eight vertices of the cubic boron cage, LaB6 is extremely difficult to grow. One alternative is to synthesize a precursor that contains the B6 octahedron or an equivalent boron cluster.

5.2 Recommendations for Future Work

Exploratory growth studies using ZrNp3Cl were not successful at growth conditions

o found suitable by thermodynamic analysis and using ZrNp4 (T = 500 C, P = 300 Torr,

H2 flow rate = 1.5 slm). The strong Zr-Cl bonding is suspected to make gas phase decomposition of ZrNp3Cl more difficult than in ZrNp4, and thus this approach does not seem feasible. The strong Zr-Cl bonding, however, proposes a possibility of ZrC ALD using dineopentyl zirconium dichloride (ZrNp2Cl2). HCls are possibly cleaved off on the surface via the reaction between chlorines and surface hydrogens. Zr=C bond is then formed via the γ-hydrogen abstraction of neopentane and the subsequent isobutene cleavage. Further study is required for this postulated scheme.

As presented in the Section 3.5.7 and 3.5.8, the selection of carrier gas for the growth of WNx films using alkyl- and arylimido precursors could tailor gas phase reaction pathways. The use of ammonia would drive WCl4 formation via α-shift of hydrogen instead of the formation of hydride intermediates via σ-bond metathesis in

117

hydrogen environment. A heterogeneous reaction between WCl4 (ad) and NH3 (ad) would then result in higher N content in the deposited films, which has not been achieved with hydrogen only. The ammonia is a good source of nitrogen, as well as surface hydrides for removing chlorines and reducing hydrocarbons on the surface. The increase of nitrogen levels in the deposited WNx films was already verified using (NH3 + H2) carrier gas [Bch04b]. Therefore, the use of only ammonia or (NH3 + N2) carrier gas seems a reasonable next step.

APPENDIX A EXPLORATORY RESULTS ON THE GROWTH OF LaB6 THIN FILMS BY AEROSOL-ASSISTED MOCVD

The feasibility of LaB6 MOCVD film growth was tested using different kinds of precursors; La(fod)3, La(thd)3, borohydride complex and tris(amido) complex. Although all the trials were unsuccessful, the rational for the choice of precursors and growth conditions used in the trials are discussed.

A.1 Precursor Synthesis

Figure A.1 depicts the prepared precursors, compounds 1-6, for the growth of LaB6 films. The β-diketonate complexes 1, La(thd)3 (thd = 2,2,6,6-tetramethyl-3,5- heptanedione) and 2, La(fod)3 (fod = 6,6,7,7,8,8,8-heptafluoro-2,2-dimethyl-3,5- octanedionate) were used as La precursors. They were known to be volatile and previously used in CVD of oxide- and sulfide-containing La materials [Eis65, Spr67].

Because of the absence of boron in themselves (1, 2, and 4-6), a secondary boron precursor (BH3·THF and/or B2H6 in this study) was also admitted in the reactor (see

Table A.1). On the other hand, the presence of La-O bonds in 1 and 2 introduced the possibility of oxygen incorporation into the films. Thus electron impact mass spectra were taken to investigate the decomposition of compound 1. The peak at m/z 505, as shown in Figure A.2, suggests that is the possible to cleave the β-diketonate ligands intact and thus avoid oxygen incorporation. Although it is difficult to connect EI-MS fragmentation patterns directly to homogeneous or heterogeneous decomposition

118 119 pathways during MOCVD (thermal decomposition in the temperature range 300 to 800

°C), this result is promising.

Compounds 4 and 5 (see Figure A.1) are hydridotris(pyrazolyl) borate (Tp) compounds. In these compounds polydentate N-bound ligands are contained in their coordination spheres [Mos89]. There are a couple of disadvantages, however, for their application as CVD precursors. They both have a low solubility in most solvents and chelation of the rings should hinder their dissociation. Although the tris(amido) complex

6 was prepared to promote the solubility and tested, the dissociation problem ceased the synthesis of the compounds 4 and 5.

The borohydride complex 3 [Tit82] was tested as a single source precursor for

LaB6 film growth. The low volatility of 3 was accommodated with the aerosol delivery system in AA-MOCVD in proper solvent (benzonitrile in this study). Complex 3 is also used as a starting material for the preparation of derivatized borohydride complexes (see

Figure A.3). Na contamination was, however, reported in the published procedure and observed in AES film compositional analysis as well, as shown in Table A.1.

All the attempts at LaB6 film growth with compounds 1-6, unfortunately, were not successful. The difficulty of assembling the B6 octahedra at each of the eight vertices of the cubic boron cage seemed to be the largest challenge. Of course, finding a transportable La precursor is another challenge. This works lead me to envision a precursor with the B6 octahedron or an equivalent boron cluster already formed. Figure

- A.4 showed examples of La complexes of the boron hydride cluster B3H8 , which include half of the B6 cage as in the LaB6 lattice. The figure also illustrates La complexes of

120

- 2- borohydride (BH4 ) as single source precursor candidates. A precursor having B6H6 boron cluster, not shown in Figure A.4, was prepared and tried as well.

A.2 AA-MOCVD Growth

The custom-built MOCVD system for LaB6 film growth was a little different than the one used for the ZrC deposition studies in Chapter 4. The reactor is a cold-walled, vertical design that allows for low pressure operation. Due to the low volatility of tested precursors, the aerosol delivery system was also incorporated in the design [Bch05]. The details of the ultrasonic nebulizer are given in Sections 1.5.3 and 4.2. A key issue is selection of a suitable solvent that gives sufficient solubility of the precursor yet avoid decomposition at the deposition temperature and subsequent impurity incorporation into the film.

For the precursors (1, 2, 3, and 6) discussed in Section A.1 were tested and the results are summarized in Table A.1; Compound 6 is a representative of the class of

2- compounds that include compounds 4-6. The precursor containing B6H6 boron cluster was also tested with the prospect of providing a preassembled B6 cage, but the result was not successful as well (not shown in Table A.1).

Table A.1 summarizes the growth experiments based on 8 runs. It is noted that both La(fod)3 and La(thd)3 contain no B and thus a B source is required. A ‘cocktail’ mixture of the La precursor and the B source BH3·THF was prepared with the solvent

(benzonitrile or triglyme). For the precursor 6 also containing no B, B2H6 was used as a boron precursor. B2H6 was also added to the gas stream for some runs in an effort to increase boron content in the films, as shown in Table A.1.

As summarized in Table A.1, the only mixture of precursors/solvent to deposit both

La and B on Si (both (100) and (111) orientations) was La(fod)3 + BH3·THF + PhCN,

121 although there was also substantial contamination from the ligands and possibly solvent, including C, N, O and F. According to AES spectra of the deposited material, this precursor/solvent mixture gave a film with a La:B atom ratio of 1:0.5, which is much less than the stoichiometric 1:6 La:B ratio. The effect of the B2H6 addition as a co-reactant could not be evaluated in that La did not exist in most cases. The Na contamination using compound 3 was attributed to an impurity from the synthesis procedure as mentioned in

Section A.1.

Figure A.1. Representative examples of La precursor compounds tested in this study

122

Figure A.2. EI mass spectrum of precursor 1

Figure A.3. Derivatized borohydride complexes from 3

Figure A.4. Promising precursor candidates for LaB6 CVD growth

123

Table A.1. Summary of the CVD results for the precursors evaluated for LaB6 deposition

Co- Carrier Temp. Precursor Solvent XRD AES reactant gas (°C) BH ·THF + 1, La(thd) 3 - N 700-800 *A B,C,O 3 *Triglyme 2 BH ·THF + 1, La(thd) 3 - H 700 A B,C,O 3 Triglyme 2 BH ·THF + 1, La(thd) 3 B H H 800-900 A B,C,O 3 Triglyme 2 6 2 BH ·THF + La,B,C,O, 2, La(fod) 3 - H 600-1000 A 3 *PhCN 2 F,N BH ·THF + 2, La(fod) 3 B H H 800-900 A B,C,O,N 3 PhCN 2 6 2 C,O,N, 3, La(BH ) ·THF PhCN - 600 A 4 3 Na B,C,O,N, 3, La(BH ) ·THF PhCN B H H 300-700 A 4 3 2 6 2 Na 6, Triglyme B2H6 H2 600 A B,C,O,N La[N(Si(CH3)3)2]3 *A: X-ray amorphous, PhCN: Benzonitrile, Triglyme: Triethylene Glycol Dimethyl Ether

APPENDIX B CHARACTERIZATION TECHNIQUES

B.1 AES (Auger Electron Spectroscopy)

When electrons having high energy (3 to 20 keV) are incident upon a conducting sample, these electrons cause core electrons from atoms contained in the sample to be ejected resulting in a photoelectron and an atom with a core hole. The atom then relaxes via electrons with a lower binding energy dropping into the core hole. The energy thus released can be converted into a characteristic X-ray (see EDS) or emit an electron. This electron is called an Auger electron after Pierre Auger who discovered this relaxation process. After the emission of the Auger electron, the atom is left in a doubly ionized state. The energy of the Auger electron is characteristic of the element that emitted it, and can thus be used to identify the element. The short inelastic mean free path (2 to 6

Ǻ) of Auger electrons in solids ensures the surface sensitivity of AES.

AES is a popular technique for determining the composition of the top few layers of a surface. It cannot detect hydrogen or helium, but is sensitive to all other elements, being most sensitive to the low atomic number elements.

AES must be carried out in UHV (< 10-9 Torr) conditions. A popular method of looking at buried layers with AES is to use the technique in combination with sputter cleaning etching. Normally, when a sample is brought into the UHV environment from air, it will be coated with carbon and oxygen. This material has to be removed (usually by sputtering) before the clean surface can be investigated. Sputtering involves directing a beam of ions (usually Ar+ ions) at between 500 eV and 5 keV at the sample. This

124 125 process cleans the surface, but can also be used to erode away the sample to reveal structure beneath the surface. Obviously this is a destructive technique.

B.2 XPS (X-ray Photoelectron Spectroscopy)

XPS was developed in the mid 1960s by K. Siegbahn and his research group. K.

Siegbahn was awarded the Nobel Prize for Physics in 1981 for his work in XPS. The phenomenon is based on the photoelectric effect outlined by Einstein in 1905 where the concept of the photon was used to describe the ejection of electrons from a surface when photons impinge upon it. For XPS, Al Kα (1486.6 eV) or Mg Kα (1253.6 eV) is often the photon energy of choice. The XPS technique is highly surface specific due to the short range of the photoelectrons that are excited from the solid. The energy of the photoelectrons leaving the sample is often determined using a CHA (Concentric

Hemispherical Analyzer) and this gives a spectrum with a series of photoelectron peaks.

The binding energy of the peaks is characteristic of each element. The peak areas can be used (with appropriate sensitivity factors) to determine the composition of the materials surface. The shape of each peak and the binding energy can be slightly altered by the chemical state of the emitting atom. Hence, XPS can provide chemical bonding information as well. XPS is not sensitive to hydrogen or helium, but can detect all other elements. The Ar+ sputtering process can also be incorporated for surface cleaning or depth profiling.

B.3 SEM (Scanning Electron Microscopy)

SEM is a very widely used technique to study surface topography. A high energy

(typically 10 keV) electron beam is scanned across the surface. The incident electrons cause low energy secondary electrons to be generated, and some escape from the surface.

The secondary electrons emitted from the sample are detected by attracting them onto a

126 phosphor screen. This screen will glow and the intensity of the light is measured with a photomultiplier.

The incident electrons will also cause X-rays to be generated, which is the basis of the EDS technique. Some of the incident electrons may strike an atomic nucleus and bounce back into the vacuum. These electrons are known as backscattered primaries and can be detected with a backscattered electron detector. Backscattered electrons can also give information on the surface topography and on the average atomic number of the area under the electron beam.

B.4 EDS (Energy Dispersive Spectroscopy)

This technique is used in conjunction with SEM and is not a surface science technique. An electron beam strikes the surface of a conducting sample. This causes characteristic X-rays to be emitted from the point of the material. The energy of the X- rays emitted depends on the material under examination. The X-rays are generated in a region about 2 microns in depth, and thus EDS is not a surface science technique.

B.5 AFM (Atomic Force Microscopy)

Atomic force microscopy (AFM) is a scanning probe technique, in which a sharp tip positioned on a cantilever is scanned in a raster-pattern along a surface. Due to forces acting between the surface and tip the cantilever is deflected away from its equilibrium position. It is possible to record the deflection as function of time and thereby form an image of the surface topography with close to atomic resolution.

The distance between the surface and the tip as well as the horizontal movement is controlled by an arrangement of piezoelectric elements controlling the x, y (horizontal) and z (vertical) coordinates of the sample surface. The angle of deflection of the cantilever is measured as a function of time using a laser beam reflecting off the

127 cantilever and into a two-segment photodiode. The topographical data are equal to the sum of the deflection and the z-position of the surface. By using a feedback system the piezoelectric elements can be controlled in the z-direction, so that the deflection of the cantilever is at a minimum. In this case the topographical data is approximately equal to the z-position of the surface only.

B.6 XRD (X-Ray Diffraction) & GIXD (Grazing Incidence X-ray Diffraction)

XRD is a technique in which the pattern produced by the diffraction of X-rays through the closely spaced lattice of atoms in a crystal is recorded and then analyzed to reveal the nature of that lattice. This generally leads to an understanding of the material and molecular structure of a substance. The spacings in the crystal lattice can be determined using Bragg’s law. The electrons that surround the atoms, rather than the atomic nuclei themselves, are the entities which physically interact with the incoming X- ray photons.

Grazing incidence x-ray diffraction (GIXD) is an ideal structural probe of thin films and surfaces, combining the power of conventional XRD for determining bulk structure with refraction effects to probe the interface structure. X-rays have a refractive index of slightly less than 1 in a solid and hence undergo total external reflection for

o angles of incidence α less than a critical angle αc (typically 0.2 ). This totally-reflected beam only penetrates the top 50 layers at the surface. A small fraction of this beam will be diffracted giving a weak diffraction pattern from the surface region alone. For α > αc, a diffraction pattern from the bulk of the film is obtained. Comparison of the two scans shows how the effect of the surface is on the film structure. Because x-rays penetrate solids, GIXD can also be used to study buried interfaces.

128

* Most of information for this section is also accessible in the following web sites:

(1) www.inano.dk [Interdisplinary Nanoscience Center]

(2) www.uksaf.org [Surface Analysis Forum: Surface Analysis Site]

(3) en.wikipedia.org [Wikipedia: The free encyclopedia]

LIST OF REFERENCES

Aiz94 T. Aizawa, Rep. of National Institute for Research in Inorganic Materials 81, 27 (1994)

All02 F.H. Allen, Acta Crystallogr. B58, 380 (2002)

Bar89 I. Barin, Thermochemical Data of Pure Substances (VCH, Weinheim, Germany, 1989).

Bch03a O.J. Bchir, S.W. Johnston, A.C. Cuadra, T.J. Anderson, C.G. Ortiz, B.C. Brooks, D.H. Powell, L. McElwee-White, J. Cryst. Growth 249 (1-2), 262 (2003)

Bch03b O.J. Bchir, K.M. Green, M.S. Hlad, T.J. Anderson, B.C. Brooks, C.B. Wilder, D.H. Powell, and L. McElwee-White, J. Organometallic Chemistry. 684, 338 (2003)

Bch04a O.J. Bchir, K.M. Green, M.S. Hlad, T.J. Anderson, B.C. Brooks, and L. McElwee-White, J. Crystal Growth 261, 280 (2004)

Bch04b O.J. Bchir, K.C. Kim, T.J. Anderson, V. Craciun, B.C. Brooks, and L. McElwee-White, J. Electrochem. Soc. 151 (10), G697 (2004)

Bch05 O.J. Bchir, K.M. Green, H.M. Ajmera, E.A. Zapp, T.J. Anderson, B.C. Brooks, L.L. Reitfort, D.H. Powell, K.A. Abboud, and L. McElwee-White, J. Am. Chem. Soc. 127, 7825 (2005)

Bec93 A.D. Becke, J. Chem. Phys. 98, 1372 (1993)

Bel05 A. Bellucci, D. Gozzi, T. Kimura, T. Noda, and S. Otani, Surf. Coat. Technol. 197 (2-3), 294 (2005)

Ber96 H. Bernt, A.Q. Zeng, H.R. Stock, and P. Mayr, Surf. Coat. Technol. 74-75, 369 (1996)

Ber97 D.L. Bergman, L. Laaksonen, and A. Laaksonen, J. Mol. Graph. Model. 15, 301 (1997)

Bla91 H.T. Blair, D.W. Carroll, and R.B. Matthews, Proceedings of the 8th symposium on Space Nuclear Power Systems (AIP, New York, 1991).

129 130

Bra87 D.C. Bradley, R.J. Errington, M.B. Hursthouse, R.L. Short, B.R. Ashcroft, G.R. Clark, A.J. Nielson, and C.E.F.J. Rickard, Dalton Trans., 2067 (1987)

Bre68 L. Brewer, Science 161, 115 (1968)

Bro89 A. Broniatowski, Phys. Rev. Lett. 62, 3074 (1989)

Bru93 J. Bruckner, T. Mantyla, Surf. Coat. Technol. 59 (1-3), 166 (1993)

Bye96 B.K. Ju, S.J. Kim, J.H. Jung, Y.H. Lee, B.S. Park, Y.J. Baik, S.K. Lim, and M. H. Oh, Mat. Res. Soc. Symp. Proc 424, 399 (1996)

CER97 CERAO incorporate, Technical Publications 7 (2) (1997)

Cha98 F.M. Charbonnier, J. Vac. Sci. Technol. B 16 (2), 880 (1998)

Cha99 F.M. Charbonnier, W.A. Mackie, T. Xie, and P.R. Davis, Ultramicroscopy 79, 73 (1999)

Cha01 F.M. Charbonnier, W.A. Mackie, R.L. Hartman, and T. Xie, J. Vac. Sci. Technol. B 19 (3), 1064 (2001)

Che97 J.W. Cheon, H.D. Lawrence, and G.S. Girolami, J. Am. Chem. Soc. 119, 6814 (1997)

Che05 C.S. Chen, C.P. Liu, C.Y.A. Tsao, Thin Solid Films 479 (1-2), 130 (2005)

Cra02 C.J. Cramer, Essentials of Computational Chemistry (John Wiley & Sons, Chichester, U.K., 2002).

D’Al00 L. D’Alessio, A. Santagata, R. Teghil, M. Zaccagnino, I. Zaccardo, V. Mariotta, D. Ferro, and G. De Maria, Applied Surface Science 168, 284 (2000)

Din91 A.T. Dinsdale, CALPHAD: Comput. Coupling Phase Diagrams Thermochem. 15, 317 (1991)

Duc85 M. Ducarroir, J. Elecrochem. Soc.: Solid-State Science and Technology 132(3), 704 (1985)

Eng39 N. Engel, Ingenioren N101 (1939)

Fol92 E. Folga, and T. Ziegler, Can. J. Chem. 70, 333 (1992)

Gal99 C. Galewski, and T. Seidel, Eur. Semicon. 21 (1), 31 (1999)

131

Gau04 Gaussian 03, Revision C.02, M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuseria, M. A. Robb, J. R. Cheeseman, J. A. Montgomery, Jr., T. Vreven, K. N. Kudin, J. C. Burant, J. M. Millam, S. S. Iyengar, J. Tomasi, V. Barone, B. Mennucci, M. Cossi, G. Scalmani, N. Rega, G. A. Petersson, H. Nakatsuji, M. Hada, M. Ehara, K. Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda, O. Kitao, H. Nakai, M. Klene, X. Li, J. E. Knox, H. P. Hratchian, J. B. Cross, V. Bakken, C. Adamo, J. Jaramillo, R. Gomperts, R. E. Stratmann, O. Yazyev, A. J. Austin, R. Cammi, C. Pomelli, J. W. Ochterski, P. Y. Ayala, K. Morokuma, G. A. Voth, P. Salvador, J. J. Dannenberg, V. G. Zakrzewski, S. Dapprich, A. D. Daniels, M. C. Strain, O. Farkas, D. K. Malick, A. D. Rabuck, K. Raghavachari, J. B. Foresman, J. V. Ortiz, Q. Cui, A. G. Baboul, S. Clifford, J. Cioslowski, B. B. Stefanov, G. Liu, A. Liashenko, P. Piskorz, I. Komaromi, R. L. Martin, D. J. Fox, T. Keith, M. A. Al-Laham, C. Y. Peng, A. Nanayakkara, M. Challacombe, P. M. W. Gill, B. Johnson, W. Chen, M. W. Wong, C. Gonzalez, and J. A. Pople, Gaussian, Inc., Wallingford CT, 2004.

Gge88 A. Gge, K. Dehnicke, and D. Fenske, Z.Naturforsch.(B) 43, 677 (1998)

Gir87 G.S. Girolami, J.A. Jensen, D.M. Pollina, W.S. Williams, A.E. Kaloyeros, and C.M. Allocca, J. Am. Chem. Soc. 109, 1579 (1987)

Gir98 G.S. Girolami, J.A. Jensen, J.E. Gozum, and D.M. Pollina, Mat. Res. Soc. Symp. Proc. 121, 429 (1998)

Gla99 J.A. Glass Jr., N. Palmisiano Jr., and R.E. Welsh, Mat. Res. Soc. Symp. Proc. 555, 185 (1999)

Goo94 W.D. Goodhue, P.M. Mitishin, C.T. Harris, C.O. Bozler, D.D. Rathman, G.D. Johnson, and M.A. Hollis, J. Vac. Sci. Technol. B 12 (2), 693 (1994)

Gui87 A.F. Guillermet, High Temp. High Pressure 19, 119 (1987)

Gui89 A.F. Guillermet, Phys. Rev. B 40, 10582 (1989)

Gui92 A.F. Guillermet and G. Grimvall, J. Phys. Chem. Solids 53, 105 (1992)

Gui95 A.F. Guillermet, J. Alloys and Comp. 217, 69 (1995)

Gut86 P. Gustafson, Carbon 24, 169 (1986)

Hea94 M.D. Healy, D.C. Smith, R.R. Rubiano, R.W. Springer, and J.E. Parmeter, Mat. Res. Soc. Symp. Proc. 327, 127 (1994)

Hog05 G. Hogarth, and I. Richards, Dalton Trans., 760-773 (2005)

Hug97 A.K. Hughes, and A.J. Kingsley, Journal of Organometallic Chemistry 539 (1- 2), 109 (1997)

132

Ist00a A.A. Istratov, and E.R. Weber, J. Electrochem. Soc. 149 (1), G21 (2000)

Ist00b A.A. Istratov, and E.R. Weber, ECS Proceedings 27, 90 (2000)

Iva99 A.R. Ivanova, C.J. Galewski, C.A. Sans, T. Seidel, S. Grunow, K. Kumar, and A.E. Kaloyeros, Mat. Res. Soc. Symp. Proc., Advanced Interconnects and Contacts, 321 (1999)

Jen01 K.L. Jensen, Naval Research Laboratory-Cathode Workshop (2001).

Kal00 A.E. Kaloyeros, and E. Eisenbraun, Annual Review Material Science 30, 363 (2000)

Kan00 D.H. Kang, V.V. Zhirnov, G.J. Wojak, R.C. Sanwald, M. Park, J.J. Hren, and J.J. Cuomo, Mat. Res. Soc. Symp. Proc. 558, 563 (2000)

Kie52 R. Kieffer, Proc. Intern. Symp. Reactive Solids, 1001 (1952)

Khe98 S.S. Kher and J.T. Spencer, J. Phys. Chem. Solids 59 (8), 1343 (1998)

Kor99 T.T. Kodas and M. Hampden-Smith, Aerosol Processing of Materials (Wiley– VCH, New York, 1999), p. 537.

Kub96 O. Kubaschewski, C.B. Alcock, and P.J. Spencer, Materials Thermochemistry (Pergamon Press, Oxford, U.K., 1996).

Kun96 S. Kunii, PHYBE3, Physica B (Amsterdam) 218, 181 (1996)

Laa92 L. Laaksonen, J. Mol. Graph. 10, 33 (1992)

Lee92 W.Y. Lee, J.R. Strife, and R.D. Veltri, J. Am. Ceram. Soc. 75, 2803 (1992)

Mac94 W.A. Mackie, R.L. Hartman, M.A. Anderson, and P.R. Davis, J. Vac. Sci. Technol. B 12 (2), 722 (1994)

Mac95 W.A. Mackie, T. Xie, and P.R. Davis, J. Vac. Sci. Technol. B 13 (6), 2459 (1995)

Mac98 W.A. Mackie, T. Xie, M.R. Matthews, B.P. Routh Jr., and P.R. Davis, J. Vac. Sci. Technol. B 16 (4), 2057 (1998)

Mac99 W.A. Mackie, T. Xie, and P.R. Davis, J. Vac. Sci. Technol. B 17 (2), 613 (1999)

Mac00 W.A. Mackie, T. Xie, K.S. Lee, A.T. McCallum, M.J. Kirby, and P.R. Davis, Mat. Res. Soc. Symp. Proc. 558, 569 (2000)

Mar98 A.A. Marchenko, V.V. Cherepanov, D.T. Taraschchenko, and Z.I. Kazantseva, Surface Science 416, 460 (1998)

133

Mat00 T. Matsukawa, K. Tkunaga, S. Kanemaru, and J. Itoh, Mat. Res. Soc. Symp. Proc. 558, 67 (2000)

Nak87 T. Nakajima, K. Watanabe, and N. Watanabe, J. Electrochem. Soc. 134 (12), 3175 (1987)

Nak97 K. Nakajima, Y. Akasaka, K. Miyano, M. Takahashi, S. Suehiro, and K. Suguro, Appl. Surf. Sci. 117/118, 312 (1997)

Nak02 M. Nakamoto, and K. Fukuda, Applied Surface Science 202, 289 (2002)

Nic78 M.A. Nicolet, Thin Solid Films 52 (3), 415 (1978)

NIS99 NIST-JANAF, Thermochemical Tables, 4th ed. (American Chemical Society and the American Institute of Physics, New York, NY, 1999).

Nug88 W.A. Nugent, and J.M. Mayer, Metal-Ligand Multiple Bonds (John Wiley & Sons, New York, 1988).

Och99 J.W. Ochterski, Vibrational Analysis in Gaussian, Technical Notes in Gaussian.com (1999).

Och00 J.W. Ochterski, Thermochemistry in Gaussian, Technical notes in Gaussian.com (2000).

Oh98 C.W. Oh, C.G. Lee, B.G. Park, and J.D. Lee, J. Vac. Sci. Technol. B 16 (2), 807 (1998)

Ord89 S.S. Ordan'yan, E.N. Gardagina, and S.N. Barabanova, INOMAF, Inorg. Mater. (Engl. Transl.) 25, S1631 (1989)

Orp89 A.G. Orpen, L. Brammer, F.H. Allen, O. Kennard, D.G. Watson, and R. Taylor, J. Chem. Soc.-Dalton Trans., S1 (1989)

Ota96 S. Otani, and Y. Ishizawa, Journal of Allows and Compounds 245, L18 (1996)

Oya92 S.T. Oyama, Catal. Today 15, 179 (1992)

Par94 J.E. Parmeter, D.C. Smith, and M.D. Healy, J. Vac. Sci. Technol. A 12 (4), 2107 (1994)

Pes00 P. Peshev, Journal of Solid State Chemistry 154, 157 (2000)

Pol01 J. Polk, An overview of JPL’s Advanced Propulsion Concepts Research Program (2001).

Sau88 N. Saunders, A.P. Miodownik and A.T. Dinsdale, CALPHAD 12, 351 (1988)

Sch95 P.R. Schwoebel, and I. Brodie, J. Vac. Sci. Technol. B 13 (4), 1391 (1995)

134

Sch00 G. Schaftenaar, and J.H. Noordik, J. Comput.-Aided Mol. Design 14, 123 (2000)

Sha01 M.J. Shaw, S. Grunow, and D.J. Duquette, J. Electron. Mater. 30 (12), 1602 (2001)

She81 A.I. Shelykh, J. Less Common Metals 182, 291 (1981)

Slu97 Marco M.G. Slusarczuk, Mat. Res. Soc. Symp. Proc. 424, 363 (1997)

Smi93 D.C. Smith, R.R. Rubiano, M.D. Healy, and R.W. Springer, Mat. Res. Soc. Symp. Proc. 282, 642 (1993)

Spe94 J.T. Spencer, U.S. Patent 5,277,932 (filed 1994)

Spi00 C. Spindt, C.E. Holland, and P.R. Schwoebel, SPIE Proc. 3955, 151 (2000)

Sun85 B. Sundman, B. Janson, and J.O. Anderson, CALPHAD: Compt. Coupling Phase Diagrams Thermochem. 9, 153 (1985)

Sur05 T. Surek, J. Cryst. Growth, (2005), 275, 292 (2005)

Ted98 C. Tedesco, A. Imirzi, and A. Proto, Acta Cryst. B54, 431 (1998)

Tes93 T.C. Tessner, and P.R. Davis, J. Vac. Sci. Technol. A 11 (1), 1 (1993)

Tho64 R. Thompson, Progress in Boron Chemistry (Oxford, New York, 1964), p. 173.

Tot71 L.E. Toth, Transition Metal Carbides and Nitrides (Academic Press, New York, 1971), p. 29.

Tra00 A. Trampert and K.H. Ploog, Cryst. Res. Technol. 12(2), 793 (2000).

Uek96 M. Uekubo, T. Oku, K. Nii, M. Murakami, K. Takahiro, S. Yamaguchi, T. Nakano, and O. Tomohiro, Thin Solid Films 286, 170 (1996)

Vau95 W.M. Vaughan, K.A. Abboud, J.M. Boncella, Organometallics 14, 1567 (1995)

Wal95 W. Waldhauser, C. Mitterer, J. Laimer and H. Stori, Surface and Coatings Technology 74-75 (2), 890 (1995)

Wid00 Y. Widiaia, and C.B. Musgrave, Surface Science 469, 9 (2000)

Wu96 Y.D. Wu, Z.H. Peng, and Z. Xue, J. Am. Chem. Soc. 118, 9772 (1996)

Wu99 Y.D. Wu, Z.H. Peng, K.W.K. Chan, L. Xiaozhan, A.A. Tuinman and Z. Xue, Organometallics 18, 2081 (1999)

Yat00 J.E. Yater, A. Shih, and D.S. Katzer, Mat. Res. Soc. Symp. Proc. 558, 551 (2000)

135

Zho00 Z.C. Zhong, V. Holmes, P.A. Dowben, and D.J. Sellmyer: Mat. Res. Soc. Symp. Proc. 558, 91 (2000)

BIOGRAPHICAL SKETCH

Yong Sun Won was born in Wonju, Kangwondo, Korea, on February 3, 1971. He received his Bachelor of Science in chemical engineering from Sogang University in

1993. His Master of Science was done on the study of “Electrostatic Interactions in

Concentrated Colloidal Dispersions” in chemical engineering, Pohang University. After receiving his master’s degree, he joined Samsung Petrochemical Co. Ltd. as a research fellow. Until he moved to Samsung ElectroMechanics Co. Ltd. in 1999, he accumulated various experience related to pilot testing, process development and basic/detail engineering through many projects. During his latest career in the company, he was in charge of surface analysis part, managing SEM, FE-SEM, AFM, and other relevant instruments. In 2001, he decided to pursue a doctorate and joined Dr. Timothy

Anderson’s research group in University of Florida. He finished his doctorate study in the summer semester of 2006 in chemical engineering.

136