Study of Homogeneous Thermal Decomposition of Triethylgallium, Triethylaluminum with Ammonia and Tungsten Dimethylhydraz...

MISSING IMAGE

Material Information

Title:
Study of Homogeneous Thermal Decomposition of Triethylgallium, Triethylaluminum with Ammonia and Tungsten Dimethylhydrazido Complex using in Situ Raman Spectroscopy and Computational Chemistry
Physical Description:
1 online resource (212 p.)
Language:
english
Creator:
Lee,Jooyoung
Publisher:
University of Florida
Place of Publication:
Gainesville, Fla.
Publication Date:

Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Chemical Engineering
Committee Chair:
Anderson, Timothy J
Committee Members:
Hagelin-Weaver, Helena E
Merz, Kenneth Malcolm
Omenetto, Nicolo

Subjects

Subjects / Keywords:
aluminum -- cvd -- dft -- diffusion -- dimethylhydrazido -- fem -- gan -- kinetics -- raman -- spectroscopy -- triethylaluminum -- triethylgallium -- tungsten -- wnxcy
Chemical Engineering -- Dissertations, Academic -- UF
Genre:
Chemical Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract:
The gas phase thermal decomposition pathways for the metal-organic precursor of TEGa ((C2H5)3Ga, Triethylgallium) were followed using in situ Raman spectroscopy measurement in an up-flow, cold-wall CVD (Chemical Vapor Deposition) reactor. Raman shift measurements shown at 490cm-1, 517cm-1, 537cm-1, and 555cm-1 were assigned to the vibrational frequencies between gallium and the ??carbon of (Et)3Ga, (DEGa)2, (Et)GaH?Ga(Et)2 and (Et)GaH?GaH2, respectively. DFT (Density Functional Theory) calculations were in good agreement with experimental observations. Identification of intermediates by Raman spectroscopy confirmed that both ??hydride elimination and homolysis reactions are present in the reactor during thermal decomposition. The mass transport simulation using the finite element method supports the presence of both reactions. In addition to the peak assignments, DFT calculations using B3LYP/LanL2DZ (Becke, three-parameter, Lee-Yang-Par/Los Alamos National Laboratory 2-Double-Zeta) model chemistry were performed for screening 17 feasible routes of 34 possible reactions. Thermal decomposition pathways of triethylaluminum ((C2H5)3Al; TEAl) were investigated. Raman scattering experiments were performed on TEAl neat and in the presence of ammonia in the in situ reactor. Raman shifts were observed for the decomposition products TEAl:NH3, DEAlH, TEAl:NH3 TS, H2N-AlH-NH-AlH2, H2Al-NH2, MEAlH2, MEAlH-AlH2 and DEAl-AlH2. DFT calculations using the B3LYP/LanL2DZ level of theory were carried out to optimize the geometry of each intermediate and likely transition structures to estimate activation energies. The gas-phase decomposition pathways of the tungsten dimethylhydrazido complexes Cl4(RCN)W(NNMe2) (1a: R = CH3; 1b: R = Ph), precursors for single source deposition of WNxCy, were investigated using a combination of Raman scattering experiments and DFT calculations. DFT calculations (B3LYP/LanL2DZ) were used to estimate Raman active frequencies and explore the reaction surface. Dimethylamine and methylmethyleneimine, products from N-N cleavage of the hydrazido ligand, were observed under deposition conditions and identified by comparison with previously reported Raman shifts and calculated frequencies. Combining experimental thermal decomposition studies by Raman spectroscopy with DFT calculations and FEM (Finite Element Method) reactor modeling is a powerful approach for quantitative CVD kinetics analysis.
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Bibliography:
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility:
by Jooyoung Lee.
Thesis:
Thesis (Ph.D.)--University of Florida, 2011.
Local:
Adviser: Anderson, Timothy J.

Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Classification:
lcc - LD1780 2011
System ID:
UFE0043313:00001


This item is only available as the following downloads:


Full Text

PAGE 1

1 STUDY OF HOMOGENEOUS THERMAL DECOMPOSITION OF TRIETHYLGALLIUM, TRIETHYLALUMINUM WITH AMMONIA AND TUNGSTEN DIMETHYLHYDRAZIDO COMPLEX USING IN SITU RAMAN SPECTROSCOPY AND COMPUTATIONAL CHEMISTRY By JOOYOUNG LEE A DISSERTATION PRESENTED TO THE GRADUATESCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2011

PAGE 2

2 2011Jooyoung Lee

PAGE 3

3 To my parents, wife and son

PAGE 4

4 ACKNOWLEDGMENTS First of all, Id like to thank God. Whatever I do and however I do, He has been with me. I have been pleased to work with Dr. Anderson including all group members. Dr. Anderson always has kept his eyes on my research and has encouraged the progress Due to his belief in my studies, I coul d finish my doctoral research. I have enjoyed discussion with group members and that always gives me good ideas and breakthroughs. The morning coffee with them refreshes my thinking and gives me energy to work efficiently. I feel five years in Gainesville is very short. My wife, Hankyung Seong, always has given encouragement and energy to my research and Id like to dedicated this dissertation to her. My only son, Geon, has been everything to my family and I hope five years of Gainesville life would be his blessings from the heavenly God.

PAGE 5

5 TABLE OF CONTENTS ACKNOWLEDGMENTS .................................................................................................. 4 page LIST OF TABLES ............................................................................................................ 8 LIST OF FIGURES ........................................................................................................ 10 LIST OF ABBREVIATIONS ........................................................................................... 15 ABSTRACT ................................................................................................................... 18 INTRODUCTION ........................................................................................................... 20 APPLICATIONS OF RAMAN SCATTERING TO GAS PHASE DECOMPOSITION REACTIONS ........................................................................................................... 24 Theoretical Background of Raman Scattering ........................................................ 25 Classical Theory of Raman Scattering ............................................................. 26 Selection Rule .................................................................................................. 27 Raman Scattering Intensity and Cross section ................................................. 28 Signal to Noise ................................................................................................. 30 CVD Reactor with Raman Spectrometer ................................................................ 31 Reactor Setup .................................................................................................. 31 Characteristics of Spectroscopic System ......................................................... 32 Applications of Raman Scattering ........................................................................... 34 Gas Phase ........................................................................................................ 34 Temperature d etermination ........................................................................ 34 Concentration m easurement ...................................................................... 39 Liquid Phase ..................................................................................................... 40 Solid Phase ...................................................................................................... 42 Summary ................................................................................................................ 47 COMPUTATIONAL CALCULATIO NS ........................................................................... 59 Vibrational Frequency ............................................................................................. 61 Classical Approach ........................................................................................... 61 Quantum Mechanical Approach ....................................................................... 62 Potential Energy Surfaces ...................................................................................... 65 Thermodynamic Properties ..................................................................................... 66 Contribution from Translation ........................................................................... 66 Contribution from Electronic Motion ................................................................. 67 Contribution from Rotational Motion ................................................................. 67 Contribution from Vibration ............................................................................... 68 Kinetic Parameters ................................................................................................. 69

PAGE 6

6 Theoretical Raman Intensity and Absolute Differential Raman Cross section ........ 72 The Selection of Basis Set ...................................................................................... 74 REACTOR MODELING AND PARAMETER ESTIMATION .......................................... 83 Mathematical Description of CVD Reactor .............................................................. 84 Parameter Estimation ............................................................................................. 86 Solving Inverse Problem Using Optimization Algorithm .......................................... 91 Summary ................................................................................................................ 93 INVESTIGATION OF THE THERMAL DECOMPOSITION OF TRIETHYLGALLIUM USING IN SITU RAMAN SPECTROSCOPY AND DFT CALCULATIONS ........... 103 Overview ............................................................................................................... 103 Experimental and Theoretical Methods ................................................................. 104 Results and D iscussion ......................................................................................... 109 TEGa Thermal Decomposition Experiments Using in situ Raman Spectroscopy 113 The Comparison of DFT Calculation with Experimental Results ........................... 115 Concluding Remarks ............................................................................................. 116 HOMOGENEOUS THERMAL DECOMPOSITION STUDIES OF TRIETHYLALUMINUM (TEAL): EFFECT OF NH3 ................................................ 127 Overview ............................................................................................................... 127 E xperimental and Theor etical Methods ................................................................. 128 Results and Discussion ......................................................................................... 1 30 Homogeneous TEAl Thermal Decomposition Experiments ............................ 130 DFT Calculations ............................................................................................ 134 The Comparison of Experimental Results with DFT Calculations .................. 136 Concluding Remarks ............................................................................................. 137 EXPERIMENTAL AND COMPUTATIONAL STUDIES OF THE HOMOGENEOUS THERMAL DECOMPOSITION OF THE TUNGSTEN DIMETHYLHYDRAZIDO COMPLEX CL4(CH3CN)W(NNME2) DURING DEPOSITION OF WNXCY THIN FILMS ................................................................................................................... 145 Overview ............................................................................................................... 145 Experimental and Theoretical Methods ................................................................. 146 Results and Discussion ......................................................................................... 148 Kinetics of Acetonitrile Dissociation from 1a ................................................... 148 Ion Cyclotron Resonance Experiments .......................................................... 148 In situ Raman Experiments ............................................................................ 149 DFT Calculations ............................................................................................ 151 Comparison of Experimental Results with DFT calculations .......................... 153 Concluding Remarks ............................................................................................. 158 DESIGN SIMULATION AND SETUP OF ULTRA HIGH VACUUM SYSTEM FOR SURFACE SCIENCE AND I TS APPLICATION .................................................... 171

PAGE 7

7 Overview ............................................................................................................... 171 Characterization of Surface Kinetics in UHV ......................................................... 172 Temperature Programmed Desorption (TPD) ................................................ 172 Low Electron Energy Diffraction (LEED) ........................................................ 173 Auger Electron Spectroscopy (AES) .............................................................. 174 Attenuated Total Reflection Using FTIR Spectroscopy (ATR FTIR) ............... 175 Current state of ATR FTIR technique ...................................................... 175 FT IR (Fourier Transform Infrared) spectroscopy .................................... 176 Applications of ATR FTIR technique to thin film research ........................ 177 UHV System Design and Sample Probe Simulation Using CFD Software ........... 182 UHV System Design ....................................................................................... 182 Sample Probe S imulation Using CFD Software ............................................. 183 CONCLUSIONS AND RECOMMENDATIONS ........................................................... 198 LIST OF REFERENCES ............................................................................................. 201 BIOGRAPHICAL SKETCH .......................................................................................... 212

PAGE 8

8 LIST OF TABLES Table page 2 1 Absolute differential Raman scattering cross section of the vibrational rotational band of gas phase nitrogen of Q branch ............................................ 49 2 2 Absolute differential Raman scattering cross section of widely used liquids using the laser wavelength of 514.5nm .............................................................. 50 2 3 Selected principal plasma lines in the argon ion laser ........................................ 51 2 4 Vales of rotational constant 0B and centrifugal distortion 0D for nitrogen and oxygen from Butcher et al .[21 ] ............................................................................ 53 2 5 Vales of and /eeB for various simple diatomic molecules ............................ 53 2 6 Raman shifts for single wall carbon nanotubes in two kinds of organic solvents .............................................................................................................. 57 2 7 Observed Raman shift of CuInSe2 nanopowder and proposed assignment ....... 57 2 8 Penetration depth according to laser wavelength in Si, Ge, Si0.78Ge0.22, 4H 6H SiC and CdS ................................................................................................. 58 3 1 Rotational symmetry numbers for molecular point groups[62] ............................ 78 4 1 Dimensionless groups in modeling equations .................................................... 95 4 2 Comparison of experimental and kinetic theory derived diffusivities for H2gas pairs and collision diameters ........................................................................ 95 4 3 Deviations of heat capacity between experimental and calculated values ......... 97 4 4 A factor and activation energy value of homolysis hydride elimination reaction before and after optimization .............................................................. 101 5 1 Experimental and calculated rate parameters for first ethy l group dissociation 118 5 2 Reported relative Raman cross section for group II, III metal organic sources 119 5 3 Computed enthalpies, entropies and Gibbs free energies for selected gas phase decomposition reactions ........................................................................ 121 5 4 Calculated and corrected Raman active Ga C stretching frequencies only for the symmetric motions ...................................................................................... 126 6 1 Calculated and corrected Raman active stretching frequencies ....................... 143

PAGE 9

9 7 1 Experimental and computationally optimized bond lengths () and bond angles () for complex 1a .................................................................................. 164 7 2 Calculated bond dissociation enthalpy ( H) and Gibbs energy change ( G) for the W N1 and N1N2 bonds ........................................................................ 166 7 3 Calculated and corrected Raman active stretching frequencies ....................... 170 8 1 The thermal pr operties used in CFD simulation................................................ 192 8 2 The temperature difference between both ends of Si wafer according to each design ............................................................................................................... 195

PAGE 10

10 LIST OF FIGURES Figure page 2 1 Comparison of three most wi dely used spectroscopic methods ......................... 48 2 2 Jablonski energy level diagram for typical IR and Raman scattering. The line thickness is roughly proportional to the signal strength ...................................... 48 2 3 A graphical representation of 1 steradian ........................................................... 49 2 4 Raman scattering results of pure liquid benzonitri le in the range of 500 ~ 1150 cm1 ............................................................................................................ 50 2 5 Schematic drawing of the chemical vapor deposition reactor for in situ R aman spectroscopic measurements ................................................................ 51 2 6 A schematic diagram of a Raman spectroscopic system ................................... 52 2 7 Conventional schematic diagram of photomultiplier tube ................................... 52 2 8 Conventional diagram of a CCD ......................................................................... 53 2 9 Temperature determination process ................................................................... 54 2 10 Overlapped rotational Raman spectra of nitrogen and ammonia, and the rotat ional Raman spectrum of ammonia ............................................................. 55 2 11 Raman scattering measurement for single wall carbon nanotubes .................... 56 2 12 Raman spectrum of CuInSe2 nanopowder measured at room temperature ....... 57 3 1 Potential energy diagram of simple harmonic oscillator ...................................... 76 3 2 Potential energy diagram for harmonic and anharmonic oscillator ..................... 76 3 3 Morse potential. (DeDo) means zeropoint energy ............................................. 77 3 4 Example of potential energy surface with two variables. Figure was taken from Ref. [79] ...................................................................................................... 77 3 5 Relative energy diagram according to the change of rotation angle of ethyl group in triethylgallium ........................................................................................ 78 3 6 The nature of simple reaction including transitionstate A ................................. 79 3 7 The first hydride elimination process of triethylgallium with geometry optimized transition state using B3LYP/LanL2DZ level of chemistry .................. 79

PAGE 11

11 3 8 Comparison of experimental and simulated results ............................................ 80 3 8 Continued ........................................................................................................... 82 4 1 Experimental and calculated heat capacity ........................................................ 96 4 2 Simulated temperature contours (unit is shown as C) and streamlines and concentration contours in the reactor ................................................................. 98 4 3 Radial triethylgallium temperature and concentration profi le at different axial positions ............................................................................................................. 99 4 4 Optimization process using genetic algorithm followed by simplex method for solving inverse problem .................................................................................... 100 4 4 Continued ......................................................................................................... 101 4 5 Concentration and temperature profi le of gas phase triethylgallium ................. 102 5 1 Schematic of the experimental reactor for in situ R aman spectroscopic measurements .................................................................................................. 118 5 2 Recorded Ga C3 vibrational measurement of 490cm1line ............................... 119 5 3 hydride elimination with Wiberg indices ......................................................................... 120 5 4 hydride eliminat ion transition state structures ....................... 122 5 5 Calculated energetics of the two major thermal decomposition pathways of TEGa with reacti on enthalpies [kcal/mol] listed ................................................ 122 5 6 17 screened reaction pathways out of 34 with reaction enthalpies [kcal/mol] listed ................................................................................................................. 123 5 7 Raman spectrum of gas phase TEGa. Different horizontal line means different height measurement along the centerline ........................................... 123 5 8 Simulated and experimental concentration profile of TEGa in the reactor ........ 124 5 9 Optimized values of rate parameters of two main reactions for activation energy and A factor (pre exponential factor) .................................................... 125 6 1 (Color online) Schematic drawing of the CVD reactor interfaced with in situ Raman spectrometer ........................................................................................ 139 6 2 (Color online) R aman spectrum of gas phase TEAl ......................................... 139 6 3 (Color online) Raman spectrum of gas phase TEAl with ammonia .................. 140

PAGE 12

12 6 4 (Color online) Measured and simulated temperature and ammonia concentration profile (with and without reaction) in CVD reactor ...................... 140 6 5 Wiberg indices tracking of C2H6 elimination from TEAl:NH3 adduct .................. 141 6 6 (Color online) Relative energy diagram of C2H6 elimination from TEAl:NH3 adduct (Energies in parenthesis are from Re f.[124]) ........................................ 141 6 7 (Color online) Relative energy diagram of C2H6 elimination in the case of excess ammonia (Energies in parenthesis are from Ref. [124] ) ....................... 142 6 8 (Color online) Simulated tetramer and hexane formation from H2Al NH2 with enthalpy change and free energy change (underlined) at 298K ....................... 142 6 9 (Color online) Comparison of Raman spectra at 2840cm1 for two cases, without ammonia and wi th ammonia ................................................................ 144 6 10 Proposed overall decomposition mechanism from neat TEAl and TEAl with ammonia experiment (Et = ethyl group; C2H5) ................................................. 144 7 1 Schematic of the nebulizer assisted experimental reactor for in situ R aman spectroscopic measurements ........................................................................... 159 7 2 Plot of ln(k/T) vs. 1/T(K) for acetonitrile exchange in complex 1a ..................... 160 7 3 Raman spectra of gas phase dimethylhydrazido complex 1a in benzonitrile as a function of distance below the heated susceptor ........................................... 161 7 4 Relative species concentration and gas phase temperature profiles along the reactor centerline for benzonitrile (BN), 1b and 2 ............................................ 162 7 5 Peak deconvolution results ............................................................................... 163 7 6 Liquid Raman spectra of 1a in benzonitrile solution (red) compared to pur e liquid benzonitrile (black) .................................................................................. 164 7 7 Computationally optimized geometry of 1a ....................................................... 165 7 8 Products of cleavage of the W N1 and N1 N2 bonds of complex 2. Gibbs energy values ( G) are in kcal/mol ................................................................. 165 7 9 Limiting resonance structures of complexes 1a and 1b .................................... 166 7 10 Wiberg bond indices for 1a and 1b ................................................................... 166 7 11 Gibbs energy change ( G) for initial ligand substitution and dissociation reactions of complex 1. All energy values are in kcal/mol at 298 K ................. 167

PAGE 13

13 7 12 Production of dimethylaminyl radical ( 3 ) after W N1 cleavage to form 5 Gibbs energy values are in kcal/mol ................................................................. 167 7 13 Known reactions of dimethylaminyl radicals and their calculated (B3LYP/LanL2DZ ) reaction energies at 298 K .................................................. 168 7 14 Suggested evolution of an atomized droplet as i t approaches a heated susceptor .......................................................................................................... 168 7 1 5 Detected Raman spectrum profile assigned to phenyl breathing mode of benzonitrile (1003 cm1) .................................................................................... 169 8 1 Desorption rate of argon from tungsten for various values of the incident ion energy. (taken from Ref. [163]) ......................................................................... 185 8 2 A multiple reflection ATR system (F igure was taken from Ref.[182]) ................ 185 8 3 V arious types of ATR FTIR system .................................................................. 186 8 4 Fluoridebased etching conditions, leading to hydrideterminated flat and porous silicon surfaces. (Fi gure was taken from Ref. [183]) ............................. 186 8 5 Internal reflection spectra of HF treated Si(111) surfaces (Fi gure was taken from Ref. [184]) ................................................................................................. 187 8 6 Ge H vibrations observed by ATR FTIR. Surfaces prepared by etching 10% HF solution. (Fi gure was taken from Ref. [170]) ............................................... 187 8 7 [2+2] cyclo additions on Si surface possible models f or adsorption of ethylene and FTIR spectra for cis and trans 1,2 dideuterioethylene ............... 188 8 8 Si H IR vibrational modes for the initial surface and after various precursor exposure doses of TDEAHf at 25 and 250 oC .................................................. 188 8 9 Calculated vibrational spectra for the species present in the H transfer process and for surface species i n the Si H abstraction process ..................... 189 8 10 A time series of IR absorbance spectra of c H and Si H stretching vibrational modes. (Fi gure was taken from Ref. [178]) ...................................................... 190 8 11 Saturatio n infrared spectra of ED/Ge, PD/ED/Ge, ED/PD/ED/Ge and PD/ED/PD/Ge, s chematic illustration of PD on Ge an d second layer reaction 190 8 12 Proposed decomposition mechanism for Ti(O iPr)2(dpm)2 on a Pt surface. (Fi gure was taken from Ref. [179]) ................................................................... 191 8 13 Four chamber UHV system and process flow diagram ..................................... 191

PAGE 14

14 8 14 Perspective view of CAD drawing for deposition chamber, ATR FTIR chamber and analysis chamber ........................................................................ 192 8 15 Proposed design of sample probe for perspective view CFD simulation profile and beveled edge Si wafer ................................................................... 193 8 16 Temperature profile simulations at 1000 K for the modified designs ................ 194 8 17 Quadrangular grid mesh for Fig. 816E ............................................................ 195 8 18 CAD drawing of sample probe for perspective, fr ont, side, and top view .......... 196 8 19 CAD drawing of designed UHV system ............................................................ 197

PAGE 15

15 LIST OF ABBREVIATION S AACVD Aerosol assisted Chemical Vapor Deposition uses aerosol produced by nebulizer or high pressure spraying system as low vapor pressure reactant delivery method. Ab initio Ab initio is Latin word and it means from the beginning In quantum chemistry, ab initio method means basic and fundamental laws of nature that doesn t count on additional assumptions or models. AES Auger Electron Spectroscopy ATR Attenuated Total Refle ction Basis set A set of MO functions that is used to create the molecular orbitals. Typically these functions are combined with atomic orbitals. There are several categories according to type of fundamental function such as minimal basis set (STO nG; n= 3, 4, 6), split valence basis set, and Pople basis set (321G, 6 311G, 6311+G*). CCD Charge Coupled Device CFD Computational Fluid Dynamics CH Chalcopyrite CVD Chemical Vapor Deposition. CVD is a chemical process that is used to deposit high purity, high performance solid film materials. DFT Density functional theory DFT is a quantum chemical modeling method used in physics and chemistry to simulate the electronic structure of N body systems. DMABEE 4 (dimethylamino)benzoate FEM Finite Element Method is a numerical technique to find solutions of partial differential equations. It is also known as FEA (Finite Element Analysis). FEM can provide good method to find solutions for complicated domains, especially for changing or moving domains. FIB Focused ion beam FT IR Fourier Transform Infrared FWHH Full width at half height HEMT High Electron Mobility Transistor

PAGE 16

16 HF Hartree Fock method is an approximate technique for determining the ground state wave function and ground state energy of a quantum many body system .Hartree Fock method assumes that N body wave function can be simplified by a single Slater determinant IRC Internal reaction coordinate is an simplified one dimensional diagram that represents reaction progress along with reaction process. IRE Internal Reflection Element LED Light Emitting Diode LEED Low Electron Energy Diffraction LOD Limit of Detection is the lower limit of detection of species LOQ Limit of Quantitation is the smallest concentration of a measur ing unit that can be reliably measured by an analytical procedure. LO mode Longitudinal mode is a particular electromagnetic field pattern of radiation or crystal vibration measured along with the propagation direction. MOCVD Metalorganic Chemical Vapor Deposition is one of CVD technique using metalorganic precursors. MP M ller Plesset Perturbation theory is a post HartreeFock ab initio method. It is improved by adding electron correlation to HF method. NA Numerical aperture NBO Natural Bond Orbital The orbital which is formed from natural hybrid orbital ( definition by IUPAC) Nd:YAG Neodymium doped yttrium aluminum garnet is a lasing medium for a solid state laser. The triply ionized Nd replaces yttrium in the crystal of the YAG as dopant. Typically dopant concentration of Nd is around 1 at. %. NIR Near Infrared NMR Nuclear Magnetic Resonance OFHC Oxygen Free High Conductivity PCM Polarized Continuum Model

PAGE 17

17 PES Potential Energy S urface PL Photoluminescence PMT Photo Multiplier Tube is a sensitive photon detector multiplying the number of photons as much as 108. QMS Quadr u pole Mass Spectroscopy is a method of mass analy sis using four highly parallel circular rods to detect mass to charge ratio. SCF Self consistent Field is another term for the Hartree Fock method. TDEAHf Tetrakis(diethylamino) hafnium ; Hf[N(C2H5)2]4 TDMAHf Tetrakis(dimethylamino) hafnium ; Hf[N(CH3)2]4 TEAl Triethylaluminum ; (C2H5)3Al TEGa Triethylgallium ; (C2H5)3Ga TEIn Triethyl i ndium ; (C2H5)3In TMAl Trimethyl a luminum ; (CH3)3Al TO mode Transverse modeis a particular electromagnetic field pattern of radiation or crystal vibration measured in a plane perpendicular to the propagation direction. TPD Temperature Programmed Desorption TPRS Temperature Programmed Reaction Spectroscopy UHV Ultra High Vacuum means the pressure< 1010 Torr. XPS X ray Photoelectron Spec troscopy

PAGE 18

18 Abstract of Dissertation Presented to the GraduateSchool of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy STUDY OF HOMOGENEOUS THERMAL DECOMPOSITION KINETICS OF TRIETHYLGALLIUM, TRIETHYLALUMINUM WITH AMMONIA AND TUNGSTEN DIMETHYLHYDRAZIDO COMPLEX IN CVD REACTOR USING IN SITU RAMAN SPECTROSCOPY AND COMPUTATIONAL CHEMISTRY By Jooyoung Lee August 2011 Chair: Timothy J. Anderson Major: Chemical Engineering The gas phase thermal decomposition pathways for the metal organic precursor of TEG a ( (C2H5)3Ga, T riethylgallium ) were followed using in situ Raman spectroscopy in an upflow, cold wall CVD ( C hemical V apor D eposition) reactor. Raman shift located at 490cm1, 517cm1, 537cm1, and 555cm1 were assigned to the vibrational frequencies between gallium and the carbon of (Et)3Ga, (DEGa)2, (Et)GaH Ga (Et)2and (Et)GaH Ga H2, respectively DFT (Density Functional Theory) calculations were in good agreement with experimental observations. Identification of intermediates by Raman spectroscopy confirmed that both hydride elimination and homolysis homogeneous thermal decomposition reactions occur under the reactor conditions T he mass transport simulation using the finite element method supports the presence of both reactions. In addition to the peak assignments, DFT calculations using the B3LYP/LanL2DZ (Becke, threeparameter, LeeYang Par/Los Alamos National Laboratory 2 D oubleZ eta) model che mistry were performed to screen 17 envisioned routes from 34 reactions

PAGE 19

19 Homogeneous t hermal decomposition pathways for triethylaluminum ((C2H5)3Al; TEAl) were investigated. Raman scattering experiments were performed on TEAl neat and in the presence of ammonia in the reactor Raman shifts were observed for the decomposition products TEAl:NH3, DEAlH, TEAl:NH3 TS, H2N AlHNHAlH2, H2Al NH2, MEAlH2, MEAlH AlH2 and DEAl AlH2. DFT calculations using the B3LYP/LanL2DZ level of theory were carried out to optimi ze the geometry of each intermediate and likely transition structures to estimate activation energies. The gas phase decomposition pathways of the tungsten dimethylhydrazido complex es Cl4(RCN)W(NNMe2) (1a: R=CH3; 1b : R=Ph), precursor s for single source deposition of WNxCy, were investigated using a combination of Raman scattering experiments and DFT calculations. DFT calculations (B3LYP/LanL2DZ) were used to estimate Raman active frequencies and explore the reaction surface. Dimethyl amine and methylmethyleneimine, products from N N cleavage of the hydrazido ligand, were observed under deposition conditions and identified by comparison with previously reported Raman shifts and calculated frequencies Combining experimental thermal decom position studies by Raman spectroscopy with DFT calculations and FEM (Finite Element Me thod) reactor model ing is a powerful approach for quantitative CVD kinetics analysis.

PAGE 20

20 CHAPTER 1 INTRODUCTION When the semiconductor industry bloomed Gordon Moore predicted a long term trend i n computing hardware, so calledMoores law [1]. Moores law suggests that the number of transistors that can be placed onto an integrated circuit will double every two years. Alt hough various semiconductor materials and devices were devised, new applications motivate the search for new semiconductor material s. Recent applications include light emitting dev ices, power electronics, and Cu metalgates ICs (Integrated Circuits) These potential new applications have spurred the development of compound semiconductors formed by combining elements from group III and group V or ones from group II and group IV. Amon g st a large spectrum of applications for compound semiconductor s, optoelectronic compound semiconductor devices that covers blue and green LEDs (Light Emitting Diodes) and UV (Ultraviolet) light detectors aretwo of the most important area s. After group IIInitride s semiconductor s proved their superiority over II VI semiconductor (e.g. ZnSe ) they have beenunder intensive research and development The bandgaps of InN, GaN, and AlN are 0.7[2] 3.4 [3] and 6.2 eV [4] respectively, which span the UV and the full visible ranges. Th is suggests that one can select bandgap energy when alloyed by manipulating the composition of In, Ga, and Al ( e.g AlxGayIn1 x yN) Moreover, group III nitrides have good thermal conductivity, chemical stability and high mechanical strength[5] Based on the outstanding physical properties of group III nitrides, a range of devic es has been developed, including LEDsdetectors, lasers and transistors.

PAGE 21

21 Amongst a large number of film growth technologies, metalorganic chemical vapor deposition (MOCVD hereafter) is the most widely used technique for many materials H owever, most MOCVD film growth results have been empirical, focused on film quality with variation in reaction conditions such as the III/V ratio and growth temperature. This is a long and expensive route to optimizing growth conditions. Supplementing experimental results with modeling would be more efficient, but kinetics are necessary. Unfortunately t he MOCVD reaction mechanism and associated rate parameter ( reaction order, r ate constant, and activation energy) are not well known. There are a few examples, however, in the literature[6 9] It should be noted that several obstacles, for example, homogeneity of reactions, physical sampling of r eactant, non intrusive temperature measurement, and wall deposition in flow reactor smust be addressed In this study, MOCVD kinetics for triethylgallium (TEGa), triethylaluminum (TEAl) and the dimethylhydrazido tungsten complex are investigated using in situ Raman spectroscopy. Raman spectroscopy is a nondestructive probe similar to infrared spectroscopy. I t can detect react ants, reaction intermediates and products during film growth and gather information from Raman signal to estimate temperature. Alt hough Raman spectr oscopy is less sensitive than infrared, it can detect an induced dipole of a molecule and measure local temperature remotely with high accuracy. This study consists of eight chapters including this chapter. In C hapter 2, the characteristics of Raman scattering and its relationship with CVD are described. Details of gas phase temperature measurement sin the reactor and concentration measurement s for gas phase chemical species as well as Raman applications for gas, liquid, and solid phase research with experimental results are also described. Chapter

PAGE 22

22 3 gives an overview of the procedures and techniques for applying computational chemistry to interpret Raman spectra. In this chapter, practical quantum chemical calculations used for Raman cross section, calculation of vibrational frequency, thermodynamic properties that contain molecular motions such as rotational, translational, vibrational and electronic motions, and potential energy surfaces are covered. In the next two chapters, two group III metal alkyls that used in chemical vapor deposition of group III V compound semiconductors. In C hapter 4, CVD reactor modeling and parameter s estimation are carried out. Computational chemistry with an appropriate method and basis set is used t o obtain fundamental reaction or thermodynamic parameter s for reaction intermediates Data from Raman scattering experiment are optimized by simplex and genetic algorithm s. In C hapter 5, the behavior of triethylgallium (Ga(C2H5)3) in CVD reactor i s examined. Both homolysis and hydride elimination are considered and the activation energy andpre exponential factor are estimated for the first GaC bond breakage of triethylgallium The measured and simulated parameters are optimized using a FEM reactor model. In C hapter 6, the gas phase thermal decomposition of triethylaluminum (Al(C2H5)3) with and without ammonia i n CVD r eactoris studied. In the next chapter, the thermal decomposition of a low volatility organometallic compound is examined. The tungs ten dimethylhydrazido complex (Cl4(CH3CN)W(NNMe2) 1 a hereafter)is of interest for deposition of WN, which has potential as a diffusion barrier in Cu metallization[10] To investigate the behavior and decomposition mechanism of 1a experiments and quantum chemical calculations are performed.

PAGE 23

23 T his chapter also includes a brief review of temperature programmed desorption (TPD), low electron energy diffraction (LEED) Auger electron spectroscopy (AES) and a detailed review of attenuated total reflection using FT IR (Fourier Transform Infrared) To assist design of the UHV system, commercial CFD (Computational Fluid Dynamics) software is used to optimize and verify the temperature distribution of a rotatable sample probe. In chapter 9, this thesis is concluded and recommendations for future studies are given.

PAGE 24

24 CHAPTER 2 A PPLICATIONS OF R AMAN SCATTERING TO GAS PHASE DECOMPO SITION REACTIONS Raman scattering has been studied for over 70 years and its applicability to material characterization and chemical analy sis has been demonstrated for a variety of systems. In particular, t here has been a considerable use of Raman scattering to study gas phase reactions Some advantages of Raman spectroscopy over Infrared spectroscopy include easy preparation of sample, the ability to obtain higher spatial resolution from visibl e light instead of infrared light, and the ability to identify molecules. For the gas phase kinetics research, Raman spectroscopy shows the strength especially in homonuclear species such as N2, O2, H2. These species do not appear in infrared spectroscopy since those are symmetric molecules. Moreover, Raman shows good spatial resolution, although infrared has the resolution no better than 4 cm1. And the visible laser used for Raman can be focused more tightly than infrared beams since it is visible. Ty pically the cross section for Raman scattering and IR (Infrared) absorption are in the order of 1029 and 1018cm2, respectively. R ecent technological advances such as low noise multichannel detectors and efficient spectrographs,have boosted the Raman sign al strength by ~104, thus increasing its sensitivity. Among various techniques that probe molecular vibrations, FTIR is the oldest and most developed technique. FTIR spectra have narrow line widths and rich spectral detail, i.e. a distinguishable fingerprint. Currently, FTIR instrumentation is highly refined due to its widespread use however it has some disadvantages. MidIR light cannot penetrate many common optical material thus it has sampling difficulty. Due to the time co nsuming and/or destructive sample preparation as a KBr pellet, Nujol null, and the

PAGE 25

25 like, application of FTIR can be interrupted. Although NIR (Near Infrared) is not as informative as FTIR, NIR has many advantages, for example, noninvasive measurement, sim plicity of sampling, and remote sampling with fiber optic cable[11] The comparison of Raman scattering with FTIR and NIR is s chematized in Fig. 21. In this chapter, the theory of Raman scattering fo r an experimental view pointis described and its applications to the three phases ( gas, liquid and solid) are discussed for the methodology of following experimental chapters. L astly the CVD reactor system that is interfaced with the Raman spectrometer is described in detail. Theoretical Background of Raman Scattering In 1928, Indian physicist Raman discovered that a very small portion of light is scatt ered off from the molecules as red shifted or blueshifted, and the frequency difference between the incident light (at frequency L) and scattered light (at frequency 0) equals to the molecular vibrational frequency, L = L o (2 1) In other words, if the frequency or wavel ength of the scattered light is analyzed, not only is the incident light wavelength seen (i.e., Rayleigh scattering) but also a small amount of light that has scattered at some different wavelength (i.e., Stokes and Anti Stokes Raman scattering). T h is phenomenon described above is called Raman e ffect and shown in Fig. 22. Rayleigh scattering was named after Lord Rayleigh who described this process from the blue color of the sky. Depending on the vibrational state of the molecule, photons of light c an have either higher or lower frequency by the Raman effect. Based on these principles there has been explosive progress in the field of Raman spectroscopy of materials with the invention of lasers in the 1960 s. Today

PAGE 26

26 Raman scattering is one of the most powerful and widely used tools for materials study in all fields even in archaeology [12] Classi cal Theory of Raman Scattering A ssume a plane of electromagnetic wave is traveling along the positive direction of the z axis. As a consequence of Maxwell s equations, such a wave has only transverse components. And thus the electric field intensity is giv en by 0xx cos()xt (2 2) where is the angular frequency. And it is noted that 2 where is the frequency; that / c where is the wavelength. If the phase is not an issue, Eq 2 2 can be simplified as follows by replacing x with zero. 0xx cos t (2 3) If an electric field, is applied to a molecule, the distribution of electrons is slightly changed and thus the dipole moment , is polarized. When the electromagnetic field is weak is proportional to the electric fi eld intensity and is given by: 0x = cos t (2 4) where is polarizability tensor and has 9 components for threedimensional space. According to the periodic vibration of a molecule, also has a periodic change as follows: 011cos t (2 5) where 0 and 1 are the timeindependent and timedependent term, respectively. Eq. 2 6 results from inserting Eq. 2 5 in to Eq. 2 4 00 0x0 x1 1x1 111 = cos cos() cos() 22 ttt (2 6)

PAGE 27

27 The first term on the right hand side (R.H.S) of Eq. 2 6 reflects Rayleigh scattering that is not affected by the change of polarizability tensor, the second term is for Stokes scattering and the third term expresses Anti Stokes scattering. Selection Rule Suppose a transition between two quantum states of a mol ecule, initial state, i and final state, f The wave functions of the initial and final state are expressedas i and f respectively. When the relevant transition dipole moment if is non zero, transition can be observed since dipole moment is given by: ,iffi (2 7) Other than the form of Eq. (2 5), polarizability can be expanded as 2 0 0 01 () 2ij ij ijij k lm k lm k lmQ QQ Q QQ (2 8) where jQ denotes the normal coordinate of a vibration mode with frequency j By separating the vibrational and rotational motion from the electronic part, a space fixed component of the polarizability tensor xy can be expanded as the form given by: 0()()xy xy fkik fi kkQQ 0()()xy fkkik k kk kQQQ Q (2 9) where j denotes the vibrational wave function of quantum state j To proceed further the following relations are required for harmonic oscillator functions:

PAGE 28

28 0 ()() 1ij fkik ijfor QQ for (2 10) and 0 ()()11 1ij i fkkikkk k i kkkfor QQQbfor bfor (2 11) where /4kbh Eq. 2 10 is for the Rayleigh scattering and the Raman selection rule is 1k while the second and third case of Eq. 2 11 is for the Stokes and anti Stokes Raman scattering, respectively [13] Raman Scattering Intensity and Cross s ection The intensity of a Raman line is closely related to the scattering cross section of the transition between the two energy levels. And various factors such as scattering angle, temperature, and the wavelength of incident light are correlated with the intensity of Raman scattering. Taking into account the thermal population at temperature, T and a 90o scattering angle, one can obtain following relation: 0 rot rotd INI d (2 12) where N is the number of scattered molecules, rot is the rotational scattering cross section, is the solid angle, and (/)rotdd is called the differential cross section [14] The total scattering cross section consists of numerous partial cross sections, for instance, elastic fission and production Therefore, it should be noted that the total cross sectio n for a particular process is the integral over all angles of the differential cross -

PAGE 29

29 section for that specific process. The direct measurement of absolute Raman scattering crosssection is challenging and laborious. To this end using a relative Raman cross section technique has been developed by many rese archers for several decades. As for earlier efforts, Hyatt and his coworkers focused on establishing the absolute Raman crosssection for molecular nitrogen gas [15] and Fenner and his coworkers performed absolute measurement for some simple gases like N2, O2, H2, CO, NH3, CH4, C2H6, and C6H6[16] The scattering cross section of the rotational lines of gas p hase hydrogen can be measured easily, since the optical anisotropy 2 0 and its wavelength dependence are well known. The differential scattering cross section of a Stokes rotational Raman line is expressed in the S branch ( 2 J ). 4 42 0016 3(1)(2) ()7 45 2(23)(21)rot rotd JJ d JJ (2 13) Since the measured Raman intensity of a molecule is directly proportional to the absolute Raman cross section as seen in Eq. 2 12, a useful and appropriate equation for species j versus nitrogen reference can be derived as follows: 2 2()N j j Nd I d J dd I (2 14) where () J is the thermal population of the J th level with a Boltzmann distribution. 0 0(21)exp(()/) () (21)exp(()/)JJ J JNgJFJhckT J N gJFJhckT (2 15) where Jg is the nuclear statistics factor and () FJ is the term value for the J th level. After obtaining a firm base for simple gases, one can measure the relative Raman crosssection of target molecule with ease. For convenience, the results of the

PAGE 30

30 measurements of the absolute differential Raman scattering cross section of the Q branch of nitrogen at 2331cm1 are tabulated in Table 21. It should be noted that the wavelength dependence of Raman scattering cross section is quite large since it has the 4 0()j dependence. The Raman scattering crosssection has units of cm2/sr, where the sr is the SI unit of solid angle. The steradian(sr) is dimensionless and is the unit angle of the cone it is depicted in Fig. 23. Cross sections in liquids are usually higher than those for the same vibration in gases by a factor of 2 to 4 and some reference values are tabulated in Table 22.Usually ( / dd ) is larger for molecules with extended systems, because the electrons in those molecules are more easily polarized whereas molecules with only single C H, C O, and C C bonds have small cross sections. A lso multiple bond stretches have high ( / dd ) values. Signal t o Noise The noise in analytical spectroscopyis defined as the fluctuation in the signal level due to different causes. It is generally associated with the standard deviation, or the root mean square noise. Noise types can be shot and flicker, the former being proportional to the square root of the signal and the last being proportional to the signal. In a typical experiment, background, detector and readout noises add to the signal noise. Therefore, within the assu mption of independent, non correlated noises, the total noise of the observed signal is defined as follows: 2222()ySBdrNoise (2 16) where S B d and r are the standard deviation of signal, the background, dark signal, and readout noise, respectively. Normally the signal to noise ratio for the peak

PAGE 31

31 intensity of a Raman signal is the average peak height, S divided by the total noise a s defined above, i.e., Signal-to-Noise Ratio = yS 2222()SBdrS (2 17) Generally, the limit of detection in quantitative analysis is often considered to be a signal to root mean square noise ratio of 3 for the concentration of the sample. It should be noted that th e signal to noise ratio is calculated relative to the backgr ound, since it is this noise that determines the limit of detection. If a signal peak is below a certain background noise level, it is indistinguishable from the noise. In addition, if a peak is too weak, it cannot accurately be quantif ied Therefore, the Limit o f Detection (LOD) corresponds to the value over which a peak can be identified with a certain statistical confidence factor (usually 3). As an example for both detectable and quantifiable signals, a liquid benzonitrile Raman scattering measurement was performed using a liquid chamber with pure benzonitrile as shown in Fig. 24. From the magnified inset, it can be seen that the peak noise intensity is 15 counts. The peak at 929 cm1 has the intensity of 62 counts with a corrected baseline. In this co ntext, signal to peak to peak noise ratio is 4.1 (=62/15) and the S/N RMS is ~21. CVD Reactor with Raman Spectrometer Reactor Setup A schematic diagram of the chemical vapor deposition reactor used in this study i s shown in Fig. 25 T his CVD reactor is an up flow impinging jet and coldwall reactor which was custom designed to study the gas phase decomposition kinetics of

PAGE 32

32 metalorganic precursors Since quartz is optically transparent, the outermost body is constructed with optical flat quartz that is shap ed as a rect angular parallelepiped. I nside of the parallepiped reactor, there is a cylindrical reactor that is also made of quartz and it has four longitudinal narrow slits to secure beam pathways. Thi s cylindrical arrangement permits a stable flow pattern in the reactor. The inlet tube consists of three parts center, annular and sweeping flows. Using three inlet tubes enables various combinations of reaction kinetic studies. For example, when triethylaluminum is introduced in the center inlet by nitrogen as a carrier gas and ammonia is introduced in the annulus inlet, one can study diffusivity between triethylaluminum and ammonia at specific temperatures, as well as aluminum nitride deposition mechanisms. Each tube is packed with 3mm glass beads confined by stainless steel screen to provide parallel flow stream s A m etalorganic precursor is introduced through the center line via a N2 carrier gas envelops the metal organic gas to limit wall deposition of precursors that interrupts homogeneous reaction To help reach reaction temperatures Boralectric sample heating material provided by tectra GmbH is used for all experiments in this study. To prevent deposition on the heating material, the sample heater is enclosed with a ceramic cap and a NiCr the rmocouple measures the temperature of the sample heater and it is locate dbetween the heater and this ceramic enclosure. Characteristics of Spectroscopic System The MOCVD reactor used in this study is interfaced with a Raman spectroscopic system (Jobin Yv on Ramanor U 1000). The excitation light sources are an argon ion laser that emits 488 nm of wavelength (Innova 90, Coherent Inc.) and the frequency doubled Nd:YAG (Neodymium doped Yttrium Aluminum Garnet) solid state laser

PAGE 33

33 (Verdi 8, Coherent Inc.) for 53 2.08 nm Among numerous plasma lines, the argon ion laser predominantly produces 488 and 514 nm lines and partly selected principal plasma lines from R ef. [17] are tabulated in Table 23. A laser line is focused by a lens that has 110mm focal length and direct s it into the CVD reactor described above. Scattered light from the sample which is perpendicular to the incident lase r line is reflected into the double monochromator. The spectrometer is equipped with two gratings which have a groove density of 1800 lines/mm and its schematic diagram is shown in Fig. 26. Scattered light is dispersed and then observed with either a l iq uid nitrogen cooled CCD ( C harge C oupled D evice) or a PMT (P hoto M ultiplier T ube). PMT s are devised to count photons to achieve high sensitivity and low enough dark noise to avoid overwhelming the signal of interest. Basically PMT uses the photo electron effect that was first discovered out in 1887 by Heinrich Hertz. When the photon strikes a photocathode and thus the energy of this photon exceeds the work function of the photocathode, a photoelectron is ejected from the metal surface. The ejected electr on is accelerated and amplified by striking sequential dynodes under high negative voltage. Conventional schematic diagram of PMT is shown in Fig. 27. In this study, the R94302 PMT model (Hamamatsu Photonics K.K.) that has a 51 mm diameter headon type and has GaAs photocathode is used. A t ypical gain is 2 106 at a temperature 25 C and cathode to anode voltage of 1750 V is the typical value of this study. The PMT used is cooled with cold water to decrease dark counts. The CCD was invented in 1969 at AT&T Bell Labs by Willard Boyle and George E. Smith. As integrated circuit techn ology progressed at a rapid rate for the past a few

PAGE 34

34 decades, CCD gets utilized due to its convenience and high capability to detect spectroscopic signals. CCD is based on the storage and manipulation of electrons and holes in a photosensitive silicon semiconductor. The photosensitive area in CCD which is covered by the circuit mask is constructed of pixel arrays in columnar form The pixels in this column detect Raman shif t as a form of slit image. T all and narrow pixels corresponds to a Raman shift value, in other words, each pixel location is rigidly related to a specific wavelength. When CCD gets Raman shift information, it accumulates photoelectrons in its array of potential wells and the electrons accumulated are converted to a digital value. Each packet of stored electrons is amplified and digitized by a converter. Usually the number of stored electrons in CCD is stated as electrons per count (e-/count). As a mul tichannel detector, the CCD used is cooled with liquid nitrogen that holds 1 liter with a downlooking Dewar and is of the Symphony Spectrum One model provided by Jobin Yvon (Horiba scientific). Conventional CCD unit is shown in Fig. 2 8. Applications of Raman Scattering Gas Phase Temperature d etermination A variety of techniques are available for measuring the local temperature of the sample. It might depend on the sample states ( gas, liquid, and solid phase ) as well as the heat sensitivity or sample stability [18] By using in situ Raman spectroscopy, one can monitor a local temperature of a gas molecule with high accuracy and without thermocouple. There are two basic methods for determining the temperature by Raman scattering and those are the Stokes to anti Stokes ratio method and the Stokes Raman

PAGE 35

35 method. Since t he intensities of the Stokes and anti Stokes lines depend on the population of the ground and excited state of the specific vibrations, the relative occupation of both states can be desc ribed by Boltzmann distribution, 1exp(/)S ASI hkT I (2 18) where T is absolute temperature and 1h is the energy gap between two states. In addition to this, since photon count is the most frequent ly measured quantity in a Raman experiment rather than photon intensity, scattered inte nsity is governed by a third power of the optical frequency (fourth power for photon intensity) [19] and thus the frequency dependence of the Stokes and anti Stokes Raman scattering can be described by: 3 001()SI (2 19a) 3 001()ASI (2 19b) where 0 is the frequency of the exciting laser line. From Eq s. 2 18, 2 19a and 219b, the ratio of the Stokes and anti Stokes Raman intensity is 3 01 1 3 01() exp(/) ()S ASI hkT I (2 20) This method seems relatively easy and convenient, however, the accuracy of temperature determination decreases drastically since the anti Stokes Raman scattering lines are very weak [20] A m ore detailed study on the measurement of uncertainty over this method has been reported by LaPlant and his coworkers [19] Briefly speaking, the absolute temperature error T is expressed as 2 1/rTkThc by using Eq. 2 20, where r is the relative uncertainty and defined as

PAGE 36

36 (/)/(/)rSASSASIIII At low temperatures the relative uncertainty becomes large due to the loss of anti Stokes line intensity At high temperatures, the 2 T term dominates the temperature error, the uncertainty becomes high as well. Secondly, in the Stokes Raman method, the temperature of the gas can be extracted from a measured rotational distribution of the gas, mainly N2 or H2, since a rotational state di stribution of gas is governed by temperature. Usually nitrogen or hydrogen is used as a carrier gas for CVD system s, and this is advantageous for a Raman scattering experiment. Rotational transitions at the ground state of vibration can be correlated wit h temperature of the gas by using Eq s. 2 12 and 213. Combining t hose t wo equations produces: 4 42 0 00exp(()/) 16 ()7 45 exp(()/)B rotS BFJhckT I gNI FJhckT 3(1)(2)(21) 2(23)(21) JJJ JJ (2 21) where 2 0 the anisotropy of the polarizability tensor, J is the rotational quantum number and Sg is the statistical weight factor ( equal to 6 and 3 for even and odd nitrogen lines ), and 0 and rot is the wavenumber shift of the incident light and rotational line, respectively () FJ denotes the rotational term and it becomes: 22 00()(1)(1) FJBJJDJJ (2 22) where 0B and 0D is the rotational constant and the centrifugal distortion constant, respectively. Values of 0B and 0D were taken from Butcher et al [21] and those values are shown in Table 24.

PAGE 37

37 As shown in T able 2 4 the magnitude of 0D is much smaller than the magnitude of 0B where typically 6 00/10 BD [22] With Eq. 2 22, one can calculate the frequency shift () J of the J th rotational Raman line as follows [22, 23] : 3 00()(2)()(46){3(23)(23)} JFJFJBJDJJ (2 23) Therefore, the rotational Raman spectrum is specified by the frequencies of 0() J where 0 is the wavenumber of the incident light. In addition, for taking into consideration of nonideality, a few correction factors are required for Eq. 2 21. In other words, the anisotropy influenced by the centrifugal distortion and spectral response of the spectroscopic system should be considered for real gas molecular behavior. From these considerations, Eq. 2 21 can be rewritten as follows: 4 42 0 00exp(()/) 16 ()7 45 exp(()/)B rotS BFJhckT I gNI FJhckT 03(1)(2)(21) (,)() 2(23)(21) JJJ RJfJ JJ (2 24) where 0(,) RJ is the spectral response as a function of wavenumber and () fJ is the correction term for the anisotropy and is theoretically described as follows [24] : 2 ''4 ()1[(1)(1)]e eB fJ JJJJ (2 25) Values of and /eeB for various simple diatomic molecules are tabulated in Table 25. However, these two functions are needed only at high temperatures and these corrections are not necessary up to approximately 1000K for nitrogen. When taking natural logarithm of Eq. 2 24 and rearranging the result, one can obtain Eq. 2 26b via Eq. 2 26a:

PAGE 38

38 2 400 4 07 (23)8 exp(()/) ()(1)(2)15exp(()/)B Srot BNIIJ FJhckT g JJ FJhckT (2 26a) 0 4 0()/ (23) 1 ln ()(1)(2)B Srot BFJhck IJ C gJJ kT (2 26b) where 2 4 00 07 8 15exp(()/)BNI C FJhckT and this is the constant for a given experimental condition. As an example, the nitrogen rotational Raman spectrum measured at 300 K and linear fit were shown in Fig. 29. As one can see, very good linearity is kept for the entire nitrogen rotational spectrum and one can calculate gas temperature with the unit of Kelvin from the slope of 1/ T When using a mixed gas, for example, nitrogen and ammonia for group III nitride deposition or nitrogen and hy drogen, it should be noted that the rotational Raman spectra of two gases overlap each other. Fig. 210 shows that the rotational Raman spectra of 95 % nitrogen and 5 % ammonia overlap so the gas temperature cannot be measured. Generally, since the interv al between each rotational band for ammonia is 39 or 38 cm1, as shown in Fig 2 10B, one should be careful for temperature measurement under this experimental condition or a peak deconvolution technique can be applied to screen the nitrogen rotational Ra man spectrum. Even though uncertainty in temperature measurement s using Raman spectroscopy over 3000K increases drastically (more than 300K) [25] the temperature measurement technique with Raman scattering is effective in the range of approximately 202230 C and has 7 % accuracy [18]

PAGE 39

39 Concentration m easurement One of the most outstanding advantages of Raman spectroscopy compared to infrared spectroscopy is the capability of measuring sample concentration. Fundamental formulation starts from Eq. 2 27. 0 rot rotd INI d (2 27) T he term in parenthesis is expressed by some experimental variables. 4 0() [1exp(/)]k k k kkd C d hckT (2 2 8 ) where kC is a constant and it includes k th vibrational mode information such as Raman scattering activity. Calculation of absolute Raman cross section described previously at known concentration, allows for us to calculate the constant kC For experimental simplicity, the relative normalized differential Raman cross section is defined as follows: 2 224 0 4 0() (/) 1exp(/) (/)()1exp(/)N kk k NkNdd hckT dd hckT (2 2 9 ) In Eq. 2 2 9 the exponential Boltzmann factor for nitrogen can be omitted at ambient temperature. N ote that the quantity k is closely related to the relative scattering coefficient and the scattering activity. When performing concentration measurement s with Raman spectroscopy, measured Raman apparent intensity and the number de nsity ratio are to be calculated at the region of interest. Using Eq s. 2 27 and 22 9 the relationship between the number density ratio, 2/NkNN and the intensity ratio is described as follows: 222(/) (/)NNN kkkNIdd NIdd

PAGE 40

40 22 24 0 4 01exp(/) () ()1exp(/)NN k k kN kI hckT I hckT (2 30) For a binary system, the mole fraction of component k is given by 2k k kNN x NN 21/1N kN N 22 24 0 4 01exp(/) () 1/1 ()1exp(/)NN k k kN kI hckT I hckT (2 3 1 ) However, the relative normalized Raman cross section has to be measured to use Eq. 2 3 1 A m ore detailed example for gas concentration measurement of groupIII precursor will be discussed in later chapters. Liquid Phase Typically a liquid phase sample is more concentrated than a gas phase sample, consequently a liquid sample can show much stronger Raman scattering intensity. Due to th is distinguishing feature, Raman spectroscopy reveals its suitability to the characterization for liquid phase samples. Th e association between acrylonitrile (CH2=CH CN) and the Ni2+ ion in aqueous solution was studied by Ramanspectroscopy [26] By changing the nickel nitrate hexahydrate concentration in the range 30~70 %(w/w) and temperature in the range 248~338 K as well, Raman shift was observed and enthalpy change was calculated. In addition to above experiment, Al a et al. have reported spectroscopic behavior for various combinations of silver nitrate solutions in four nitriles (e.g. acetonitrile, benzonitrile, ethylcyanoacetate, and acrylonitrile ) to elucidate the relationship between the silver ion and the nitrile [27] An

PAGE 41

41 i nteraction between nitriles and cations was revealed and s o lvation spheres around the silver ion are identified spectroscopically. A n ew so lvation number of acetonitrile was calculated based on the band deconvolution and component fitting in the () CN region. Yin et al [28] measured the Raman spectral intensities of CCl4 in CS2 at concentrations ranging from 90 to 0.05%(v/v) for tracing Raman scattering cross sections of the 218, 314 and 459 cm 1of CCl4 bands A lso from the Raman scattering observations, the Raman scattering cross sections of three CCl4 bands were increase d with d ecreasing concentration of CCl4 over the entire range. Raman spectroscopy is widely used in carbon nanotube research [29 32] Rao et al [33] have measured the Raman spectrum of single walled carbon nanotubes in CS2where the tubes are present in orderedbundles or ropes. From the typical tube diameter estimation formula of 1224/Rcmnmd where R and d is the radial mode frequency and the tube diameter, respectively, modified formula of 1(224/)RRcmnmd Interestingly, the radial mode frequencies for the tubes in solution are found to be approximately 10 cm1higher than those observed for tubes in a rope F or the solvent effect, single walled carbon nanotubes were placed in two types of solvents and liquid phase scattering experiment s were performed with a Raman spectrometer. Two sets of experiments were performed using only sodium dodecyl sulfate (SDS) and a mixed solution of sodium cholate (SC) and SDS in the ratio of 4:1, and the measured Raman sp ectra of two experiments are shown in Fig. 211. Among various measured Raman shifts, peaks compared in Fig. 211 show outstanding difference. Peaks around 270 and 1500 cm1suggest metallic characteristics of carbon nanotube and those around 1600 cm1show s semiconducting

PAGE 42

42 characteristics. Raman measurements reveal that sodium cholate downshifts the metallic band rather than semiconducting features. Other than these features, the measured Raman shift for two cases are tabulated in Table 26. Mitambo and L oppnow recorded Raman spectra of 4(dimethylamino)benzoate (DMABEE) in 18 solvents of varying polarity and hydrogen bond strength to find the molecular mechanism between solvent and solute[34] Raman shifts of the C=O stretch of DMABEE in various solvents show large differences as much as 45cm1. Other than this, Raman spectroscopy can give reaction kinetic information as Jeon et al reported[35] By observing time dependent Raman spectra of Bromo o toluic acid at 25 ~ 55 C, they estimated the rate constant for the product of phthalide in an 80 % ethanol solution. In addition, far after the early trial of polymerization of the styrene monomer to polystyrene[36] a detailed polymerization kinetic study was revealed by Chu et al by observing Raman spectra of three initial temperatures [37] Lee et al performed in situ monitoring of benzylideneaniline formation from benzaldehyde and aniline under chloroform in a glass microfluidic chip and compared experimental signals with calculated data using B3LYP/631G(d) level of chemistry [38] Increasing C=N stretching of benzylideneaniline at 1628 cm1 shows the progress of the imine formation reaction. The variety of applications makes Raman spectroscopy more useful than before and its capability allows researchers to perform experiments more conveniently and accurately. Solid Phase Raman spectroscopy can also provide various kinds of information for a solid phase as well as gas and liquid phase. Especially Raman shift measurement is

PAGE 43

43 especially useful for layered heterostructure device studies Since photoluminescence (PL) resolution is limited by carrier diffusion w hereas Raman spectroscopy resolution is l imited only by laser spot size, vibrations probed by Raman are very sensitive to changes in the lattice. Lattice order, stress and even chemical composition profiling can be measured by observing changes in peak width, position and relative peak strength. Lattices have steady vibrations even at a stable state and phonon modes explain those vibrations. Phonon modes consist of an acoustic mode and an optical mode and both may be transverse (TO) or longitudinal (LO). Lattice strain, order, structure, compo sition, and orientation can be confirmed by phonon mode analysis. Kim and Spitzer have reported Raman shift of a GaxAl1 xAs mixed crystal due to the influence of the surrounding lattice [39] A lso Wetzel and his coworkers showed the carrier dens ity N in n type GaN can be expressed as: 170.7641.110() N (2 3 2 ) where is the observed Raman frequency shift within the interesting range of 193110 Ncm [40] In addition, Kim et al characterized GaN on sapphire after etching by Focu sed Ion Beam (FIB) with changing probe beam current near the etched ring area[41] Even though it is a qualitative study, that research contributes to the field of Raman spectroscopy. Like gas phase temperature measurement, Cui et al derived a noncontact temperature measurement function. It has the form of: 0 0() exp[(/)]1 C T DhckT (2 3 3 ) where 0 is the Raman line position at 0 K, C and D are free parameter s decided by experiments [42] Alt hough it is a purely empirical function, its good linear dependency

PAGE 44

44 shows it can be a good nondestructive technique for measuring device temperature. As an application, Kim et a l has reported contact free measurement of channel temperature of AlGaN/GaN HEMTs (High Ele ctron Mobility Transistors) under different operation modes [43] By plotting the plot of the peak position of the 2 2E phonon mode as a function of the passive heating temperature from 20 to 150C, they provided a device temperature measurement reference so that Raman shift of an active heating device can be compared to that. Other than these applications, Raman spectroscopy can be conveniently used to identify crystalline structures ,and evaluate material quality and composition. As an example of identifying crystalline nanostructures, CuInSe2 synthesized by a low temperature solution based method was characterized by micro Raman spec trometer. The room temperature microRaman spectra were recorded by using a Jobin Yvon Ramanor U 1000 spectrometer equipped with an optical microscope BH 2, manufactured by the Olympus Company. All measurements were performed by using 532.08nm line of a N d:YAG solid state laser with 0.2W intensity. For better signal to noise ratio, 20 scans for each measurement were performed. The backscattered light is collected with the numerical aperture (NA) of the microscope and imaged to the double monochromator eq uipped with two 1800 grooves/mm gratings. The photomultiplier tube (PMT) equipped with a GaAs cathode (R94302, Hamamatsu) was used as a detection unit. For several years, various studies of the CuInSe2 chalcopyrite ( ch ) structure (D2d) for thin film a nd powder characterization were reported[44 46] The vibrational spectrum of CuInSe2 chalcopyrite structure consists of 24 zonecentered vibrational

PAGE 45

45 modes of 1A1 + 2A2 + 3B1 + 4B2 + 7E. Besides two silent modes (2A2), 22 Raman active vibrations (1A1 + 3B1 + 4B2 + 7E) are expected from the CuInSe2 chalcopyrite 15 symmetry are Raman active in a sphalerite structure (Td). The vibrational frequencies for the LO and TO modes with 15 symmetry in sphalerite structure were r eported at 210~217cm1 and 230~235cm1, respectivel y [45, 46] B esides, CuxSey ( e.g CuSe and Cu2Se ) and InxSey ( e.g., InSe2 and In2Se3) structure s can coexist with CuInSe2 structures in the sample. Raman spectra of various CuInSe2 nanopowder samples for this research, measured using the 532.08nm line, are shown inFig 2 12. In order to exploit a Raman spectrum, the main part from 50 to 300cm1 was extracted. Results are tabulated in Table 2 7 Raman frequencies and proposed peak assignment s were described. As described above, two major CuInSe2 sphalerite characteristic peaks and some binary peaks (CuxSey and InxSey) were measured. Although the highest intensity of the feature at 124cm1 from D in Fig. 2 12 was attributed to the B1 mode [47] it does not seem clearly assigned because 124cm1 of B1 mode is from chalcopyrite. Besides the feature at 124cm1, the other four peaks are relatively clear. First of all, features at about 200cm1 can be attributed to Cu3Se2[48] And the peak at 212 and 236cm1 correspond to B2 and E modes of CuInSe2 phase as reported for sphalerite structure (Td) [49] As shown in Fig 2 12, 212cm1feature appears as shoulders, and two 236cm1 peaks were measured only from B and D. However, the possibility that the feature at 236cm1 could be overlapped with that of Sen polymer moleculecannot be rule d out As for the peaks from 256 to 259cm1, there are three possibilities. Witte et al. reported 260cm1 as the

PAGE 46

46 Cu2 xSe phase from a Cu rich Cu(In,Ga)Se2 sample[50] On the other hand, the Raman shift at 260cm1 can be attributed to In2Se3[51] a nd its frequency typically coincides with that of a Se8 ring structure observed from monoclinic and amorphous selenium. In addition to that, Lin e t al [52] r ecently r eported Raman spectra of 259cm1 for Cu2Se structure from the film deposited on copper in a hydrazine/water system. From the analysis of the micro Raman spectra described above, the sample analyzed has a unique structure of sphalerite (Td) and it is confirmed by Raman shift analysis although CuInSe2usually has a chalcopyrite (D2d) structure. When it comes to thin film characterization for semiconductor applications, note that each laser with a different wavelength has a different penetration depth. Moreover, penetration depth also depends on sample compositions. For strong absor ption materials like semiconductors, the Raman signal originates froma volume defined by the penetration depth and thediameter of the laser beam. A shorter laserwavelength penetrates less and provides lattice strain information near the surface.Different penetration depths in silicon andgermanium are tabulated for various wavelengths. When deeper penetration is required, Raman measurement after sputtering is an available procedure. Ramsteiner and his coworkers sputtered the ion impl anted GaAs:Si+ samples and performed Raman observations with various sputtering depth of 100, 150, 200 nm and as grown film as well [53] As shown in Table 28, penetration depth data according to laser wavelength is very limited and one can consult the calculation method. The optical penetration depth can be calculated as follows: 1 2pd (2 3 4 ) where denotes the absorption coefficient [54]

PAGE 47

47 Summary Raman spectroscopy has numerous applications for most research fields even archeology, geology, or art. In the same context, Raman spectroscopy interfaced with CVD reactor can be used for elucidating reaction kinetics by measuring local temperature and species concentrations at precisely focused points inside the react or. The principles of Raman scattering, detectors such as CCD and PMT were described. Moreover, Raman applications for gas, liquid, and solid phase research were explained and exemplified by experiments. In addition, practical information relevant to further research for each phase was stated.

PAGE 48

48 Figure 2 1. Comparison of three most widely used spectroscopic methods Figure 22. Jablonski energy level diagram for typical IR and Raman scattering. The line thickness is roughly proportional to the signal strength. Spectroscopic Method Spectroscopic MethodRaman Scattering Narrow linewidths Water compatible Easy sampling Noninvasive Low frequency mode observable Low sensitivity Interference Enhancement possible FTIR Absorption Narrow linewidths Good fingerprint good libraries Fundamental vibrations Sampling often difficult Strong water absorption NIR Absorption Water compatible Noninvasive Fiber optics, R emote sampling Wide linewidths Calibration complex Poor fingerprint Advantages Disadvantages Spectroscopic Method Spectroscopic MethodRaman Scattering Narrow linewidths Water compatible Easy sampling Noninvasive Low frequency mode observable Low sensitivity Interference Enhancement possible FTIR Absorption Narrow linewidths Good fingerprint good libraries Fundamental vibrations Sampling often difficult Strong water absorption NIR Absorption Water compatible Noninvasive Fiber optics, R emote sampling Wide linewidths Calibration complex Poor fingerprint Advantages Disadvantages

PAGE 49

49 Table 21. Absolute differential Raman scattering cross section of the vibrational rotational band of gas phase nitrogen of Q branch (table is taken from R ef. [14] ) Exciting line 0 (nm) d d (1032 cm2/sr) 14 0(2331) d cm d (1048 cm6/sr) 632.8 21 3 6.4 1 514.5 44 17 43 2 42 2 43.2 0.8 5.1 2 5.0 0.3 4.9 0.3 5.05 0.11 488.0 33 11 43 54 3 55.8 2 3.0 1 4.0 1 5.0 0.3 5.13 0.2 457.9 76 5 73.7 3 5.2 0.4 5.09 0.25 363.8 204 25 5.1 0.6 300.0 970 10.5 Figure 23. A graphical representation of 1 steradian.

PAGE 50

50 Table 22. Absolute differential Raman scattering cross section of widely used liquids using the laser wavelength of 514.5nm (table is taken from R ef. [11] ) Sample d d (103 0 cm2/sr) 14 0(2331) d cm d (1048 cm6/sr) B enzene 992 cm 1 30.6 235 3060 cm 1 45.3 Cyclohexane 802 cm 1 5.2 A ll C H 43 CHCl 3 3032 cm 1 4.4 758 cm 1 3.2 667 cm 1 6.6 364 cm 1 6.3 261 cm 1 7.0 Figure 24. Raman scattering results of pure liquid benzonitrile in the range of 500 ~ 1150 cm1. The peak to peak noise (about 1/5 of the RMS noise) is shown, together with clearly visible drift, which has the character of flicker noise. 40 60 80 100 120 140 900 950 1000 Wavenumber(cm-1) Intensity (a.u.) 0 500 1000 1500 2000 2500 3000 3500 4000 500 600 700 800 900 1000 1100 Wavenumber(cm-1) Intensity (a.u.) 40 60 80 100 120 140 900 950 1000 Wavenumber(cm-1) Intensity (a.u.) 15 counts Drift 40 60 80 100 120 140 900 950 1000 Wavenumber(cm-1) Intensity (a.u.) 0 500 1000 1500 2000 2500 3000 3500 4000 500 600 700 800 900 1000 1100 Wavenumber(cm-1) Intensity (a.u.) 40 60 80 100 120 140 900 950 1000 Wavenumber(cm-1) Intensity (a.u.) 15 counts Drift

PAGE 51

51 Figure 2 5 Schematic drawing of the chemical vapor deposition reactor for in situ Raman spectroscopic measurements Table 2 3 Selected principal plasma lines in the argon ion laser Relative Intensity Wavelength (air) (nm) Wavenumber (vac) (cm1) Shift (cm 1 ) relative to 488.0 nm Shift (cm 1 ) relative to 514.5 nm 5000 487.9860 20486.67 0 960 496.5073 20135.07 351.6 1500 500.9334 19957.16 529.5 1400 506.2036 19749.39 737.3 1000 514.5319 19429.73 1056.9 0 >1750 611.4929 16348.90 4137.8 3080.8 1400 617.2290 16196.96 4289.71 3232.8

PAGE 52

52 Figure 26. A schematic diagram of a Raman spectroscopic system Figure 27. Conventional schematic diagram of photomultiplier tube.

PAGE 53

53 Figure 28. Conventional diagram of a CCD Table 2 4 Vales of rotational constant 0B and centrifugal distortion 0D for nitrogen and oxygen from Butcher et al [21] Molecule 0B (cm 1 ) 0D ( 106 cm1) Nitrogen (N 2 ) 1.98950 0.00002 5.48 0.05 Oxygen (O 2 ) 1.43768 0.00001 4.85 0.01 Table 2 5 Vales of and /eeB for various simple diatomic molecules Molecule a /eeB b H 2 0.38 0.01 1.38310 2 D 2 0.38 0.01 9.762 10 3 N 2 0.45 0.09 8.476 10 4 O 2 0.23 0.07 9.148 10 4 CO 0.27 0.13 8.901 10 4 aTaken from Ref. [55] bSpectroscopic data was taken from Ref. [56]

PAGE 54

54 Figure 2 9. Temperature determination process: ( A) a rotational Raman spectrum taken at 300 K and ( B) linear relationship between LHS and ()/BFJhck in Eq. 2 26b. 0 50 100 150 200 250 300 350 40 60 80 100 120 140 160 180 200 Raman shift (cm-1) Intensity (a.u.) A 0 50 100 150 200 250 300 350 40 60 80 100 120 140 160 180 200 Raman shift (cm-1) Intensity (a.u.) A -37 -36 -35 -34 -33 -32 0 300 600 900 1200 1500 F(J)hc/k LHS B 1 slope T -37 -36 -35 -34 -33 -32 0 300 600 900 1200 1500 F(J)hc/k LHS B 1 slope T

PAGE 55

55 Figure 210. ( A) Overlapped rotational Raman spectra of nitrogen and ammonia, ( B) the rotational Raman spectrum of ammonia. 0 500 1000 1500 2000 2500 3000 3500 4000 30 80 130 180 230 Raman shift (cm-1) Intensity (a.u.) A 0 500 1000 1500 2000 2500 3000 3500 4000 30 80 130 180 230 Raman shift (cm-1) Intensity (a.u.) A 80 180 280 380 480 580 180 230 280 330 380 430 Raman shift (cm-1) Intensity (a.u.) B 80 180 280 380 480 580 180 230 280 330 380 430 Raman shift (cm-1) Intensity (a.u.) B

PAGE 56

56 Figure 211. Raman scattering measurement for single wall carbon nanotubes in ( A) sodium dodecyl sulfate (SDS) and ( B) sodium cholate (SC) and SDS in the ratio of 4:1. 0 200 400 600 800 1000 1200 1400 16001800 2000 100 150 200 250 300 350 400 450 Raman shift [cm-1] Intensity [a.u.] A SDS only SC:SDS = 4:1 276 cm-1271 cm-1 0 200 400 600 800 1000 1200 1400 16001800 2000 100 150 200 250 300 350 400 450 Raman shift [cm-1] Intensity [a.u.] A SDS only SC:SDS = 4:1 276 cm-1271 cm-1 0 1000 2000 3000 4000 5000 6000 7000 8000 1000 1200 1400 1600 1800 2000 2200 Raman shift [cm-1] Intensity [a.u.] B SDS only SC:SDS = 4:1 1545 cm11535 cm1 0 1000 2000 3000 4000 5000 6000 7000 8000 1000 1200 1400 1600 1800 2000 2200 Raman shift [cm-1] Intensity [a.u.] B SDS only SC:SDS = 4:1 1545 cm11535 cm1

PAGE 57

57 Table 2 6 Raman shifts for single wall carbon nanotubes in two kinds of organic solvents Molecule Raman shift [cm 1 ] SDS only 229 238 276 290 1527 1545 1588 1928 SC:SDS=4:1 228 235 271 290 1526 1535 1588 1927 Difference 1 3 5 0 1 10 0 1 Figure 212. Raman spectrum of CuInSe2 nanopowder measured at room temperature Table 2 7 Observed Raman shift of CuInSe2 nanopowder and proposed assignment Entry Raman shift (cm 1 ) Proposed assignment A B C D 1 124 B 1 2 200 200 201 199 Cu 3 Se 2 3 212 212 212 212 15 (LO) 4 237 236 15 (TO) / Se n polymer 5 256 257 258 259 Cu 2 x Se/Cu 2 Se/In 2 Se 3 /Se 8 ring molecule 0 100 200 300 400 500 600 700 800 900 1000 40 90 140 190 240 290 Raman Shift (cm-1) Intensity (a.u.) A B C D 124 200 212 236 256 0 100 200 300 400 500 600 700 800 900 1000 40 90 140 190 240 290 Raman Shift (cm-1) Intensity (a.u.) A B C D 124 200 212 236 256

PAGE 58

58 Table 2 8 Penetration depth according to laser wavelength in Si, Ge, Si0.78Ge0.22, 4H 6H SiC and CdS Laser wavelength (nm) Penetration depth in Si (nm) Penetration depth in Si 0.78 Ge 0.22 (nm) Penetration depth in Ge (nm) Penetration depth in 4H 6H SiC (nm) Penetration depth in CdS (nm) 633 3000 c 32 c 514 762 c 340 a 300 d 19.2 c 8 a ~500 2mme 488 569 c 19 c 100 b 457 313 c 140 a 65 d 18.7 c 8 a 413.1 61 a 12 d 7 a 364 50~100 m e 351 5 a 5 d 5 a 325 ~10 c ~9 c 266 100~250 e 244 ~6 c ~7 c 50~100 e ( Data were taken from a) Ref. [57] b) Ref. [58] c) Ref. [59] d) Ref. [60] and e ) Ref. [61] )

PAGE 59

59 CHAPTER 3 COMPUTATIONAL CALCUL ATIONS Raman spectroscopy can detect various types of molecules such as adduct s, molecules in a transition state as well as the parent metalorganic precursor with its high spatial resolution. H owever, unfortunately Raman observation does not indicate w hat signal scorrespond to. This research involves quantifying Raman signal and calculating various properties, for example, rate constant, activation energy, free energy and so on, so a n external method is required for these calculations. Typically IR measurement is used for organic molecules which has enormous database for assignment, so IR does not undergo this problem. Many theoretical chemists have studied and developed computational chemistry for predicting experimental results for several decades a nd they achieved eyeopening progress with the help of developing computational performance. A molecule s quantum states and its eigenvalues are expressed by wave functions for time independent case as follows: (,)(,)(,) HRrRrERr (3 1) where (,) HRr (,) Rr and E are the electronic wavefunction, the electronic Hamiltonian operator, and the electronic energy, respectively. Vector R is vector notation for the coordinates of the N nuclei and r represents the coordinates of the n electrons. To solve this well known Schr dinger equation, many researchers have studied various approaches Among them, there are four main approaches to calculating molecular properties ab initio methods, semi empirical methods, the density functional method, and the molecular mechanics method [62] Semi empirical methods do not use th e correct molecular Hamiltonian, rather a simple Hamiltonian and

PAGE 60

60 parameters whose values are adjusted to fit experimental results. On the other hand, ab initio calculation uses the correct Hamiltonian and the value of the fundamental physical constants while neglecting experimental data. Latin ab initio means from the beginning and it implies a calculation based on fundamental principles. The goal for a HartreeFock SCF (Self consistent Field) calculation is to find the antisymmetrized product , of oneelectron functions which minimize *Hd where H is the true Hamiltonian. The c lassical ab initio method based on HartreeFock (HF) formally scales as number of electrons to N4 or higher powers. The density functional method d oes not calculate the molecular wavefunction, instead, it calculates the molecular electron probability density and calculates the molecular electronic energy from the density. The properties of a many electron system can be determined by using functionals which is the spatially dependent electron density. DFT methods have a better scaling of N3 or lower and it make calculations on typical CVD precursors more accessible because t he molecular mechanics method handles the molecule as a collection of atoms held together by bonds and molecular energy in terms of force constants. Stric tly speaking, the molecular mechanics method is not a quantum mechanical method since it does not use a molecular Hamiltonian operator or wavefunction. Applications of computational chemistry have expanded rapidly with the growth in computational speed and lower computing cost. These methods have been re ndered effective in studying molecular properties of compounds especially in the areas of combustion and halogenated hydrocarbon chemistry [63] When it comes to the CVD process, the ab initio method and DFT calculations have provided a thermodynamic and

PAGE 61

6 1 kinetic perspective for the development of reaction mechanisms for deposition of Si and related compounds [64 66] Appl ying computational chemistry to the gas phase precursors in CVD reactor s has been investigated less partial ly because of the large number of electrons associated with the second, third, and fourth row elements of the metalorganic materials for compound semiconductors [67] In this chapter, a practical study of computational chemistry will be considered for combining experimental results with theoretical values to secure their validity At first, a vibrational frequency calculation method will be described with two types of approaches based on the harmonic vibration assumption, and then the estimation method for thermodynamic properties and kine tic parameters will be shown in detail. Lastly, a calculation method of theoretical Raman intensity and differential Raman cross section will be described as practical applications. Vibrational Frequency Classical Approach A single particle of mass m attracted toward the origin by a force proportional to the displacement from the origin as follows : Fkx (3 2) where k is the force constant and F is the force on the particle. The force F is the source that the spring is not stretched greatly from its equilibrium position. From Newton s second law, Fma Eq. 3 2 gives 2 2dx kxm dt (3 3)

PAGE 62

62 where t is the time. By solving Eq. 3 3, one can have the solution of sin(2) xAtb (3 4) where A and b are the integration constants, and the vibrational frequency is 1 2 k m (3 5) A photon can be absorbed by a molecule if the frequency of the photons is the same as one of the normal vibrational modes of the molecule. That suggests a molecule that i s in its ground vibrational state c an be excited, so that it vibrate s at a given frequency. Quantum Mechanical Approach As described earlier, a diatomic molecule vibrates like two masses on a spring with a potential energy that depends on the square of the displacement from equilibrium. However, the energy levels are quantized at equally spaced values in the case of simple harmonic oscillator as shown in Fig. 3 1. T h e Schr dinger equation for a harmonic osc illator can be converted by using the classical spring potential 22211 () 22 Vxkxmx (3 6) where / km is the angular frequency. T h e Schr dinger equation with this potential energy of Eq. 3 6 is expressed as 22 22 2()1 ()() 22 dx mxxEx mdx (3 7) Eq. 3 7 can be solved with a differential equation solving method by substituting wavefunction () x for 2/2()xxCe, where / m T his leads to the ground state energy for the quantum harmonic oscillator which is nonzero. This energy is called the

PAGE 63

63 zero point energy and is expressed as 1 2 h or equally 1 2 since 2 and 2 h where h is the Planck s constant and has the value of 6.626 1034m2kg/s. Note that the energy levels of the quantum harmonic oscillator are 1 2nEn n =0, 1, 2, 3, (3 8) This zeropoint energy is the most outstanding difference with the classical approximation. This means, even at the ground state and absolute zero temperature, molecules are not completely at rest. One can make Eq. 3 6 more useful by replacing x with eRR R is the distance between atoms and eR is the equilibrium distance when being in the ground state. 21 ()() 2eVRkRR (3 9) T he simple harmonic oscillator approximation closely matches energies at lower energy levels as shown in Fig. 31, however, it yields unmatched energy levels as energy increases Since a real molecule has anharmonicity to some extent, additional terms should be cons idered[68] 23 12111 222nEnn n higherterms (3 10) where 1 and 2 are anharmonicity constants. These correction terms provide much better match of the calculated energies to the experimentally observed energies. And it is depicted in Fig. 32 as red dotted lines. However, Fig. 3 2 does indicate one obvious deficiency. In reality, at some internuclear distance, which occurs at 4.748 eV fo r the H2 molecule, two atoms dissociate and each

PAGE 64

64 atom is not bound each other anymore. In that state, the bond is broken and they are isolated. A more realistic model which describes this condition is the Morse potential as shown in Fig. 33. The Morse potential energy function is of the form () 2()(1)eRR eVrDe for 2ek D (3 11) where eD is the depth of the potential well and k is the bond force constant so dictates potential well width. The Morse potential shows anharmonicity in its shape and converges to ()eVrD as the bond distance R In addition, the 1 D Schr dinger equation is analytically solvable if the potential well is expressed by the Morse potential with its eigenvalues described by Eq. 3 10. For simplici ty, Eq. 3 10 can be truncated to the secondorder term 2 111 22nEnn (3 10b) where anharmonicity constant 1 can be expressed 2 124eD (3 12) With the energy expression above, the hot band transition can be calculated. The harmonic oscillator approximation works fine with the light molecules because vibrational frequencies are very high, however, when it comes to heavy molecules, it could yield considerable error for metalorganic molec ules. Therefore, it should be noted that the anharmonicity of the potential energy surface on hot band transition can affect frequency shifts.

PAGE 65

65 Potential Energy Surface s The potential energy surface (PES) is deeply related to the BornOppenheimer approximation and is a hypersurface defined by the potential energy of a collection of atoms. T his approximation allows us to separate electronic and nuclear motion as a product form. In addition, the validity of this separation between electronic and n uclear motion is due to the large ratio between electronic and nuclear masses. Since the energy of molecule is a function of the positions of nuclei, the potential energy surface expresses energy of a molecule in terms of its structure. From the minimum energystate, reactant s undergoes various energy gradient, i.e. hills or s addle points, as shown in Fig. 3 4. The potential energy surface has 3N 6 coordinate dimensions, where N is the number of atoms not less than 3. Due to a large number of dimensions, complete potential energy surface for polyatomic molecules are very difficult to visualize. Thus a potential energy surface of a molecule is very important for finding transition state s and stable product s. Various attempts have been done to find more easily find transition states[69] and the internal reaction coordinate (IRC) with relatively a simple molecule due to the reason described above[70] As a one variable example, there are two forms of stable triethylgallium structure, i.e. as reported by Mitzel and his coworkers [71] According to crystallographic data, t hree of the four molecules are found in a propeller like structure with C3 symmetry and only one molecule of the four has two ethyl groups pointing toward each other. Usually ethyl groups connected to metal center s can rotate and it has been known that the rotation has an energy barrier. The p otential energy surface representing the changing of the dihedr al angle of C C Ga C was simula ted and depicted in Fig. 3 5 and appropriate structures were also shown.

PAGE 66

66 Since the maximum relative energy difference is only 1.97 kcal/mol, the ethyl group can rotate without much restriction. In addition, the structure at a rotation angle of 90o and 270o is the initial structure for hydride elimination and it seems located in the relatively stable region to proceed into further reactions. Thermodynamic Properties As one of the most useful applications of quantum chemistry to metalorganics, estimation of thermodynamic properties plays a crucial role. For estimating thermodynamic properties, one should note that there are a few very important approxim ations to be aware of. Firstly all the equations assume particles do not interact with each other ,an ideal gas approximation. Secondly it is assumed that the first and higher excited states are inaccessible. This approximation can introduce some error for systems with low lying electronic excited states. Usuallyfour constituents contribute to one thermodynamic property, i.e. translation, electronic motion, rotational motion, and vibrational motion[72] Although among these four contributions vibrational motion plays the major role, all four components should be co nsidered. Contribution from T ranslation The starting point for a thermodynamic analysis is from considering the partition function for the corresponding component of the total partition function. The usual partition function for the translational contribution is 3 2 22B tmkT qV h (3 13) where Bk is Boltzmann constant and has the value of 1.381 1023m2kg/s2K.

PAGE 67

67 From Eq. 3 13, the contribution to the internal thermal energy due to translation is as follows: 2lntAB Vq ENkT T 3 2 RT (3 14) since ln3 2Vq TT Contribution from Electronic Motion The usual electronic partition function is: 12/ // 12oB BBkT kT kT eoqeee (3 15) where n is the degeneracy of the energy of the n th level, and n is the energy of the n th level. From the assumption of inaccessibility between the first and higher excited states at any temperature, the energy of the ground state is set to zero. This assumption simplifies the electronic partition function so that the electronic components of internal energy and the electronic heat capacity are zero such as 0elecU and 0eC respectively. Contribution from R otational M otion The molecular rotation contribution can be separated into several cases according to the molecule s type, i.e. single atom, linear polyatomic molecule and nonlinear polyatomic molecule. For a single atom case, because rq does not depend on temperature, 1rq For a linear polyatomic molecule, the rotational partition function is as follows:

PAGE 68

68 1r rrT q (3 16) where r is 1 for asymmetric linear molecules and 2 for symmetric linear molecules. Other than these values, rotational symmetry numbers are different according to the molecular point groups as shown in Table 31. 22/8rBhIk and I is the molecular moment of inertia of the linear molecule. From Eq. 3 16, the contribution from rotational motion to the internal energy is: 2lnr r Vq ERT T RT (3 17) Lastly, for the general nonlinear polyatomic molecule, the rotational partition function is: 1/2 3/2 ,,, r r rxryrzT q (3 18) where rx , ry and rz are x, y, and z directional contribution at temperature T respectively. From Eq. 3 13, the contribution to the internal thermal energy due to rotation for the polyatomic linear molecule is as follows: 2lnr r Vq ERT T 3 2 RT (3 19) Contribution from V ibration The contributions to the partition function, internal energy, entropy and heat capacity from vibrational motions are a little more complicated than those of other

PAGE 69

69 motions and are composed of a product or sum of the contributions from each vibrational mode, K. As for the contributions from vibration, only the real frequencies are counted. Imaginary frequencies that were expressed with a minus sign in the quantum ca lculations output and describe imaginary bond such as one in the transition state would not be considered. A molecule with n atoms has 36 n degree of freedom while linear molecule has 35 n and each mode has a characteristic vibrational temperature, ,/vKKBhk A nd the partition function from a given vibrational mode is: ,/2 /21vK vKT vK Te q e (3 20) and the overall vibrational partition function is described by a product of partition functions for each vibrational mode and expressed as follows: ,/2 /21vK vKT v T Ke q e (3 21) From Eq. 3 21, the contribution to internal thermal energy resulting from molecular vibration is: ,, /211 2 1vKv vK T KER e (3 22) One should note that all four contributions are considered for estimating thermodynamic properties. Kinetic Parameters Calculating kinetic parameters, such as rate constant and activation energy for the reaction depends on transitionstate theory. Although the geometry and energy of stable molecules can be specif ied with experimentally to a greater accuracy, transitionstate s cannot be isolated experimentally since its lifetime is very short and it is

PAGE 70

70 unstable[69] T hus elucidating transitionstate s is one of the most advantageous targets for computational chemistry. Among a number of different methods to derive the fundamental equations of transitionstate theory, one can follow the most intuitive way with the exam ple hereafter. First of all, consider the si mple reaction depicted in Fig. 3 6. The product B was produced from the reactant A via transition state A. A r eversible first reaction has rate constants k1 and k1 for forward and reverse reactions, respectiv ely. For the relation between rate constants, molar concentration and equilibrium constant: 1[] [] kA k A kK (3 23) where K is the equilibrium constant between the transitionstate complex and the reactant. From the relationship between Gibbs free energy and the equilibrium constant : ()/ABGGkTKe (3 24) where the difference of free energy, AGG is meant for the activation free energy, i.e. G F rom the relation lnoBGUPVkTQ where Q is the partition function, Eq s. 3 23 and 324 can be combined as follows: [( ln)( ln))/B AABABUPVkTQUPVkTQkTKe ()/()/AB ABUUkTPVPVkT AQ ee Q ()/ABUUkT AQ e Q (3 25)

PAGE 71

71 It can be assume d the changes in PV are negligible, thus from Eq s. 3 23 and 325: ()/ 1ABUUkT AQ kke Q (3 26) By separating the partition function for the reaction coordinate degree of freedom, it can be written ()/ 1ABUUkT B AkT Q ke hQ ,/oGRT BkT e h (3 27) or equally, ,,// 1ooHRTSR BkT kee h (3 28) By using Eq. 3 28, one can calculate rate constant s based on following criteria: AGGG (3 29) F or example, the first hydride elimination of triethylgallium follows reaction as shown in Fig 3 7; Since triethylgal l ium has an ethyl group and it is known to undergo hydride elimination with relatively low energy. The spin polarized hybrid density functional B3LYP with the LanL2DZ calculations using Gaussian 03 program suite [73] were performed. Isolation of hydride elimination during the thermal decomposition process of triethylgallium is difficult, and so information about kinetic parameters of hydride elimination is limiting. However, compared with experimental result of 38.7 kcal/mol reported by Wong and his coworkers [6] and calculated one of 43.6 kcal/mol by Tsu d a and his coworkers [74] for Ea,1, the enthalpy change result of 38.3 kcal/mol in this study

PAGE 72

72 agrees very well with reference data. In addition k1 was calculated to be 30.6 s1 and it has been reported in the range 11.3 to 36.1 s1 by other researchers and they reported Ea,1in the range 22.7 to 47.2 kcal/mol [7, 9, 75] As McDaniel and his coworker reported[76] d ecomposition reaction kinetics of metalorganic precursors are inclined to change according to experimental condition such as carrier gas, reactor temperature, and reactor type. Complete decomposition mechanism of triethlygallium will be discussed more in detail in a later chapter. Theoretical Raman Intensity and Absolute Differential Raman Crosssection The progress of computational chemistry for vibrational Raman spectral predictions allows us to estim ate theoretical Raman intensity and absolute differential Raman cross section among other values Schr tter and Kl ckner reported details of various gases and liquids for analytic calculations [14] They provide an excellent review of experimental data with the Raman scattering model. T he relationship of the absolute differential Raman scattering cross section of a Stokes shifted vibrat ional Raman band j is given by: 042 44 0 ,() 2 (,) 451exp(/)j jj j jdb Sfj d hckT (3 30) where 22/8jjbhc the square of the zero point amplitude, the angular factor 2(,)[2(1)sin]/(1)jjjfj and the Raman activity 22(457)jjjjSga Eq. 3 30 depends on the temperature through the exponential, which is negligible (less than 1%) for the region below 1000 cm1[14] jg in Raman activity denotes the degree of degeneracy of the normal vibration j Typically Raman experiments are set for 90o

PAGE 73

73 scattered angle where is the angle between the direction of observation without polarization and the polarization vector of the linearly polarized exciting radiation and so the depolarization ratio j of the Raman band j is given by: 2 223 454j j jja (3 31) and the angular factor (,) fj approaches unity. From this basis, the theoretical Raman intensity, R jI which simulates the measured Raman spectrum can be calculated according to the formula 4 0() 1exp(/)j R i j jjS If hckT (3 32) where f is an appropriately chosen common scaling factor for all peak intensity [77] Raman scattering activity, jS can be calculated ab initio or by the DFT method [78] Using Eq s. 3 30 and 332, the Raman scattering intensity and the absolute differential Raman scattering cross section can be predicted. For comparison of experimental results with simulated ones the liquid phase Raman experiment for neat benzonitrile was performed. Using the liquid chamber, t he 0.1 W Nd:YAG solid state laser line w as used to excite the neat benzonitrile source and vibrational Raman excitation lines were recorded through the range of 100~4000cm1. Even at a low intensity of the excitation line the measured intensity was sufficiently high and it shows a very good signal to noise (S/N) ratio. The experimental Raman scattering data and various simulated results are compared and shown in Fig. 38 with pure Lorentzian line shapes and a bandwidth (FWHH; Full Width at Half Height) of 5 cm1.

PAGE 74

74 As one can find out, all three HF simulat ion results sho w large discrepanc ies in predicting vibrational frequencies and intensities, the B3LYP method shows relatively good results and the MP2 method shows midperformance. The Selection of Basis Set Choosing a model of a chemical system always involves a trade off between computational cost and accuracy because larger basis set runs longer. Even though larger basis set can describe the locations of the electrons in space more accurately, one cannot ignore computational cost. However, inappropriate basis set cannot give good description of electron state. Therefore, the selection of basis set is very important in computational calculations. At an early stage of computational c hemistry, minimal basis sets, for example STO nG (Slater Type Orbital), where n is an integer, have been used. This approximates Slater orbitals with Gaussian functions. Other than this, one can increase the size of a basis set by adding more functions to each atom, for instance, split valence basis set. Similarly, triplet split valence basis set such like 6311G uses three sizes of contracted functions for each orbital. In this research, t he B3LYP density functional model with LanL2DZ basis set was selected because it shows good performance on electron affinities, excellent performance on bond energy and reasonably good results on vibrational frequencies and geometries of ion and inorganic compounds. B3LYP is one of the hybrid functionals that approximate exchangecorrelation energy functional by constructing a linear combination of the HartreeFock exact exchange functional and any number of exchange and correlation density functionals. It was developed by Hay and Wadt and has been widely used in quantum chemistry, especially in study of compounds that

PAGE 75

75 contains heavy atoms. In addition to that, basis set superposition error was considered using counterpoise(CP) method to correct basis set overlapping especially in the case of ammonia adduct formation.

PAGE 76

76 Figure 31. Potential energy diagram of simple harmonic oscillator Figure 32. Potential energy diagram for harmonic and anharmonic oscillator. R ( ) R ( )

PAGE 77

77 Figure 33. Morse potential. (DeDo) means zeropoint energy. Figure 34. Example of potential energy surface with two variables. Figure was taken from Ref. [79]

PAGE 78

78 Figure 35. Relative energy diagram according to the change of rotation angle of ethyl group in triethylgallium Table 31. Rotational symmetry numbers for molecular point groups [62] Point Group r vC 1 hD 2 nS 2,4,6, n /2 n nC 2,3,4, n n nhC 2,3,4, n n nD 2,3,4, n 2 n nhD 2,3,4, n 2 n T 12 hO 24 0 0.5 1 1.5 2 2.5 0 30 60 90 120 150 180 210 240 270 300 330 360 Rotation Angle Relative Energy (kcal/mol) (o) 0 0.5 1 1.5 2 2.5 0 30 60 90 120 150 180 210 240 270 300 330 360 Rotation Angle Relative Energy (kcal/mol) (o)

PAGE 79

79 Figure 36. The nature of simple reaction including transitionstate A. Figure 37. The first hydride elimination process of triethylgallium with geometry optimized transition state using B3LYP/LanL2DZ level of chemistry. E Reaction Coordinate A AB kk1k1 E Reaction Coordinate A AB kk1k1 +(C2H5)3Ga [(C2H5)3Ga](C2H5)2GaH C2H4k1Ea,1 +(C2H5)3Ga [(C2H5)3Ga](C2H5)2GaH C2H4k1Ea,1

PAGE 80

80 Figure 38. Comparison of experimental and simulated results. ( A) experimental Raman spectra of liquid benzonitrile, ( B) Simulated results using HF/6311G level of chemistry, ( C) HF/LanL2DZ, ( D) HF/SDD, ( E) B3LYP/6 311G, ( F) B3LYP/LanL2DZ, ( G) B3LYP/SDD, ( H) MP2/LanL2DZ, ( I) MP2/SDD Wavenumber (cm1)Intensity ( a.u.) 0 500 1000 1500 2000 2500 3000 3500 4000 A Wavenumber (cm1)Intensity ( a.u.) 0 500 1000 1500 2000 2500 3000 3500 4000 A Wavenumber (cm1)Intensity ( a.u.) 0 500 1000 1500 2000 2500 3000 3500 4000 B Wavenumber (cm1)Intensity ( a.u.) 0 500 1000 1500 2000 2500 3000 3500 4000 B Wavenumber (cm1)Intensity ( a.u.) 0 500 1000 1500 2000 2500 3000 3500 4000 C Wavenumber (cm1)Intensity ( a.u.) 0 500 1000 1500 2000 2500 3000 3500 4000 C

PAGE 81

81 Figure 38 Continued Wavenumber (cm1)Intensity ( a.u.) 0 500 1000 1500 2000 2500 3000 3500 4000 D Wavenumber (cm1)Intensity ( a.u.) 0 500 1000 1500 2000 2500 3000 3500 4000 D Wavenumber (cm1) Intensity ( a.u.) 0 500 1000 1500 2000 2500 3000 3500 4000 E Wavenumber (cm1) Intensity ( a.u.) 0 500 1000 1500 2000 2500 3000 3500 4000 E Intensity ( a.u.) Wavenumber (cm1) 0 500 1000 1500 2000 2500 3000 3500 4000 FIntensity ( a.u.) Wavenumber (cm1) 0 500 1000 1500 2000 2500 3000 3500 4000 F

PAGE 82

82 Figure 38. Continued. Wavenumber (cm1)Intensity ( a.u.) 0 500 1000 1500 2000 2500 3000 3500 4000 G Wavenumber (cm1)Intensity ( a.u.) 0 500 1000 1500 2000 2500 3000 3500 4000 G Wavenumber (cm1)Intensity ( a.u.) 0 500 1000 1500 2000 2500 3000 3500 4000 H Wavenumber (cm1)Intensity ( a.u.) 0 500 1000 1500 2000 2500 3000 3500 4000 H Intensity ( a.u.) Wavenumber (cm1) 0 500 1000 1500 2000 2500 3000 3500 4000 IIntensity ( a.u.) Wavenumber (cm1) 0 500 1000 1500 2000 2500 3000 3500 4000 I

PAGE 83

83 CHAPTER 4 REACTOR MODELING AND PARAMETER ESTIMATION Metalorganic Chemical vapor deposition (MOCVD) reactors have been widely used to deposit thin films in the semiconductor industry. Frequently, the performance of these films depends on their uniformity in thickness and composition. And this uniformity strongly depends on the heat and mass transfer conditions in the reactor, especially the region adjacent to the heated substrate since homogeneous reactions more readily occur in this layer. Reactor modeling is very useful because it can suggest efficient reactor design s, optimize reaction conditions and thus improve device properties. However, factors important in reactor model ling are complex and include buoyancy driven convection, homogeneous and heterogeneous gas phase reactions of precursors, the strong temperature dependence of the physical properties of gases, and parasitic reactions at the coldwall [80, 81] In particular, the limiting factor is most often the lack of reaction mechanism and quantitative rate parameters In MOCVD thermal decomposition reactions can produce hydrocarbon radicals that are responsible for carbon incorporation in films or the formation of undesired compounds consuming the desired metals. To minimize parasitic gas phase reactions and improve film deposition performance, transport processes are critical in this context. The large thermal gradient that is typical condition in a CVD react or can produce buoyancy driven recirculation flow leading to growth rate variation, inconsistent film quality, impurity incorporation and grad ed composition across the heterojun ctions [82] In other words, those secondary flows adversely affect film thickness and composition uniformity by superimposing on the main flow. When it comes to recirculation flow of the CVD reactor used in this research, the upflow reactor geometry has advantage that the cold gas flows up toward

PAGE 84

84 the heated substrate and it makes the flow stable since the buoyancy tends to progress upward. Due to the importance of transport phenomena, it has been studied extensively in experimental and modeling studies [83] In the following, a brief mathematical description of transport phenomena, parameter estimation process using analytical and computational calculations and the procedure for solving the inverse problem using optimization algorithm encoded FORTRAN language are described. Mathematical Description of CVD Reactor Modeling a CVD reactor requires solving the equations for conservation o f momentum, energy and mass In the presence of homogeneous gas phase reactions, the steady state transport processes within CVD reactors are described with the following set of equations [84] Conservation of momentum (vv)p(v(v))Tg (4 1) Conservation of energy vpCTkT (4 2) Conservation of mass for species k -1 k kii i=1vwLwlnkkGTr 1,,1 k (4 3) Continuity equation v0 (4 4) Ideal gas law (equation of state) MP RT (4 5)

PAGE 85

85 where M P R T are the density, molecular weight of the gas, the nominal pressure of the reactor, gas constant, and reactor temperature, respectively. And pC v p k and g are the viscosity, heat capacity of the gas, pressure within the reactor, thermal conductivity and gravity vector, respectively. In addition, kiL kG kw and kr represent the multi component diffusion effect, thermal diffusion effect, the mass fraction and production rate of species k, respectively [84] Based on those governing equations, appropriate boundary conditions must be set to describe reactor conditions correctly, and natural boundary conditions were applied along the reac tor centerline and reactor walls. Fi r st of all, due to the axisymmetry of the reactor, the right half of the reactor was set for the modeling domain. For the momentum calculations, the noslip condition was applied to the walls of the reactor and the sus ceptor surface. Three temperatures at the heater, ambient, and inlet region are given for the energy conservation. The ambient temperature is contributed to consider Newton s law of cooling. Because the reactor walls and susceptor are impenetrable and nonreacting surfaces (chemically inert), mass diffusion flux normal to the surface of component k is zero. More often than not, the e quations in dimensionless form are helpful in analysis of mass transport phenomena[85] S ince these dimensionless equations giveapproximate orders of magnitude, dimensionless groups are sometimes very useful. Typical useful dimensionless groups for describing CVD reactor with typical order of magnitude and physical meaning are tabulated in Table 4 1. For example, the Peclet number can be expressed as follows:

PAGE 86

86 RePr LL Pe (4 6) or 0100 2 100()/RePr ()/pCTTl Pe kTTl (4 7) The c alculated Peclet number value for the CVD reactor used in this study at typical experimental conditions is approximately 99.96 at 375.15 K and this value is almost identical to the results for an infinite Peclet number. At the higher temperature, Peclet number is increased. In other words, heat transport by convection is far more dominant than that by conduction in this system [80] Parameter Estimation To predict the kinetic behavior of gases, various kinetic theories have been studied. A mong widely used theories, elementary kinetic theory characterized by a mean molecular velocity and a mean free path for low density gases describes three useful properties i.e., thermal conductivity( k ), binary diffusivity( ABD ), viscosity( ), at temperature T as follows [86] : 322 1 3()p AB ABAC MM RT k MMNd (4 8) 33 32() 11 32AB AB ABAMM RT D MMpNd (4 9) 322 11 3()AB ABAMM RT MMNd (4 10) where, AM BM pC A N p and d are molecular weight of molecule A and B, heat capacity, Avogadro number, pressure, and molecular diameter, respectively. However, one should note that these relationships show discrepancies and can be corrected by

PAGE 87

87 the rigorous k inetic theory of Chapman and Enskog, which considers in some detail the effect of intermolecular potential energies on the interactions between colliding molecules. By applying ChapmanEnskog theory to binary gas mixture, the kinetic parameters mentioned above can be predicted more accurately since ChapmanEnskog theory takes into account the force field between molecules and their finite diameter [87] In addition the ChapmanEnskog theory treats the molecules as symmetric spheres and assumes that all collisions are bi nary and elastic and that molecular motion during collisions can be described by classical mechanics. 2 2 ,()/ 5.907310ABAB ABkABTMMMM k [J/m ] (4 11) 3 7 2 ,()/ 1.858310ABAB AB ABDABTMMMM D p [m2/s] (4 12) 6 2 ,/() 3.775010ABAB ABDABTMMMM p [Pa s] (4 13) where and s are the collision diameter and collision integrals. In addition the collision integral for thermal conductivity, k is identical to that for viscosity, and values of k are given for the LennardJones intermolecular potential as a function of the dimensionless temperature, / kT [85] T hese collision integrals are interpretedas describing the deviation from the rigid sphere behavior. The quantity in the LennardJones potential is a characteristic energy denoting themaximum energy of attraction between a pair of molecules. For binary mixtures AB is usually approximated by a geometric average of the contributionsfrom the two species as ABAB and AB

PAGE 88

88 is approximated as ) ( 2 1B A AB [88] Relevant collision diameter s, actual and calculated binary diffusivities are tabula ted in Table 42. The LennardJones potential parameters of ( ) and (J/mol) are estimated using the method of Le Bas [84] 1/31.18BV (4 14) 1.15BkT (4 15) where BT and BV are the normal boiling point (K) and the Le Bas volume (cm3/mol). In addition LennardJones parameter is used to calculate the dimensionless collision integrals (1,1)* (1,2)* and (2,2)* and these are calculated empirically by Neufeld and his coworkers In this work following collision integrals are used among 16 collision integrals suggested[89] **0.77322.43787 0.14874 *1.161450.524872.16178k TTTee (4 16) ***0.476351.529963.89411 0.15610 *1.060360.193001.035871.76474D TTTTeee (4 17) where */ TkT A simple semi empirical method which accounts for the energy exchange in polyatomic gases was developed by Eucken [90] and thermal conductivity of a polyatomic gas at low density is described as: 5 4pkCR M 2 2 ,/() 4.177210ABAB ABkABTMMMM [J/m ] (4 18)

PAGE 89

89 where 2/()ABABMMMMM Eq. 4 11 can be replaced with Eq. 4 18 for more accuracy since target system consists of polyatomic molecules. For CVD modeling, heat capacities of metalorganic sources and their reaction intermediates are required. Rotational and vibrational contributions to the heat capacities can be calculated based on quantum chemistry which has been discussed in Chapter 3 in detail As examples, the comparison of calculated results with experimental values of heat capacity at 1 atm for N2, C2H4, C4H10 are shown in Fig 4 1. For these calculations, Gaussian 03 program was used with same level of chemistry (B3LYP/LanL2DZ) for consis tency. Obviously, calculated estimations for heat capacity show good agreement with experimental observations. For N2, although difference between two sets of data increases as temperature rise, the maximum calculat ed deviation from experimental results is at most 0.55 % at 800K. The deviation was calculated bas ed on the follow ing expression: ( ) (%) 100 ExperimentvalueCalculationvalue Deviation Experimentvalue (4 19) Deviations of all available temperatures are tabulated in Table 43. Even though deviation of butane at 300K is 10.21 %, it shows a decreasing trend as temperature increases and thus using the calculation data would not be problematic since the typical temperature range in which butane is produced as an intermediate from parent metalorganic precursors used in this st udy is at l east higher than 400K. As one can see in Table 4 3, the calculated values for nitrogen are underestimated, while those for ethane and butane are overestimated.

PAGE 90

90 For the triethylgallium case, boundary conditions were set as follows: 5 % triethylgallium was introduced through center inlet and pure nitrogen was used as a carrier gas with the velocity of 2.5 cm/s for all three gas inlets. The heater, inlet, and wall temperatures were set at 700, 200, and 22 C, respectively. The heater temperature was measured with a thermocouple underneath the heating material, the inlet temperature with rotational Raman spectrum of nitrogen as described in chapter 2, and the wall t emperature with handheld thermocouple. Fig 4 2 shows temperature and concentration contours in the reactor. One should note that Fig 4 2 shows time independent case, which means sufficient time has been allowed to make stable stream in the reactor. In addition, one can certify any directional profiles such as radial, axial or even diagonal positions. In Fig. 4 2A continuous arrows show velocity vector notation of gas flow and it obviously confirm s there is no eddy or recirculation flow that inhibits good quality film deposition by occurring flow instability. Temperature and concentration profiles versus the radial position are extracted and shown in Fig 4 3 with 10 mm interval s from 50 mm to 110 mm of z altitude. Obviously, the temperature profile decreases as distance from susceptor increases, however, concentration propagation in the direction of reactor wall can be confirmed and the interval between each profile at zero radial position increases and then decreases. One can presume the steady inc rease of triethylgallium concentration at 15 mm through 20 mm is due to the diffusion from the centerline to the reactor wall since pure nitrogen is the only component in the annulus and sweep flows. Due to this phenomenon, axial concentration profilessho w sigmoidal shape.

PAGE 91

91 From these results, one can estimate reaction parameters such as A factor (pre exponential factor) and activation energy of each reaction associated with a specific CVD reaction mechanism For example, for two major initial reactions of triethylgallium, homolysis and hydride elimination of the ethyl group, the A factor swereestimated as 65.0 kcal/mol and 40.3 kcal/mol for the se two reactions. Andactivation energy was estimated using DFT calculations as 4.6 81022s1and 1.95 1013s1, respectively. One can find more details about these calculation results from next chapter The overall process of CVD ,including gas phase hydrodynamics, mass transport, homogeneous and surface mediated chemistry, and physical diffusion and nucleation on t he surface, is so complex that one cannot elucidate all the processes clearly. However, CVD modeling helps to examine each step individually so that the relative importance of each step can be established. Solving Inverse Problem Using Optimization Algori thm From a forementioned initial guess es for A factor and activation energy by DFT calculations, the numerical reactor model can simulate the triethylgallium decomposit ion hydride elimination. For so lving parameter estimation problem, maximum likelihood estimation is used. Introducing an optimization scheme into inverse problem solution process ensures a global minimum As an optimization scheme, genetic algorithm followed by simplex method was impl emented for the parameter estimation. As described in detail elsewhere[91] due to the high nonlinearity of heat, mass and energy transport model, genetic algorithms are chosen to solv ethis optimization problem since it has been shown to be successful for solving highly nonlinear space [92] Eq. 4 20 describes objective function for the rate constants and activation energy.

PAGE 92

92 22 22 4 0 4 101exp(/) () min1/1 ()1exp(/)N NiN j ji i N jijhckTI fhckTI (4 20) where h, c, k T I and j represent wavenumber, Planck constant, speed of light, Boltzmann constant, temperature, scattering intensity, and the relative Raman crosssection of species j, respe ctively. In addition if is the model function. By minimizing the difference between experimental and simulated results, one can set the binary diffusion coefficient and/or relative Raman cross section of species j. Also optimizatio n process of A factor and activation energy can be performed through reparametrization[92] From well known equation of following empirical equation, expaE kA RT (4 21) where k A aE and T represent rate constant at temperature T A factor (pre exponential factor), activation energy of a reaction and reaction temperature, respectively. By reparametrizing the Eq. 4 21 with newly introduced parameters, 1 and 2 1211 expmk TT (4 22) nonlinear relation between A and aE was changed into linear relation between 1 and 2 where 1ln()a mE A RT and 2 aE R And mT denotes the mean temperature of measurement. Optimization processes for A factor and activation energy of homolysis and hydride elimination are shown in Fig. 44. Though homolysis and hydride elimination processes are shown in different panel in Fig. 44, those were simulated in

PAGE 93

93 one system. In other words, each system has influence on another system in the reactor simu lated. In Fig. 4 4 C the goodness of fit is the internal criteri on that de scribes conversion threshold. A t otal of 43 iterations was made and optimized activation energies are 60.0 and 44.0 kcal/mol and A factors are 1.261021s1and 8.91013s1for homolysis and hydride elimination, respectively as tabulated in Table 44. From T able 44, one can find the difference between the values before and after the optimization process T he values of before optimization were calculated from DFT calculati ons are shown to be a good initial guess for reactor modeling. For comparison of optimized values, the concentration and temperature profiles of triethylgallium are depicted in Fig 4 5. The dots with error bars were measured experimentally, and dotted and solid line mean calculated profile of before optimization and after optimization, respectively. As obviously shown above, profilesthat use values after optimization (black solid line) shows better performance than that before optimization case (blue dotted line). Summary In this chapter, the CVD reactor simulation adopting the Galerkin FEM model was described and applied to experimental data for the homogeneous decomposition of tri ethylgallium. The FEM model used in this study was described by five major governing equations, i.e., conservation of momentum, energy, and mass for species k, continuity equation and ideal gas law. By estimating the transport properties thermal conducti vity, binary diffusivi ty, viscosity, heat capacity, as well as othersby DFT calculations, appropriate initial values were estimated. Aforementioned DFT calculations showed only small deviations with experimental data in the wide range of

PAGE 94

94 temperature, whic h indicates that DFT calculations can provide good estimat es for unknown parameters. In addition the inverse problem was resolved by using a genetic algorithm followed by simplex method to estimate rate parameters for triethylgallium thermal decomposition Compared with the pre optimized triethylgallium concentration pro file, the optimized profile shows better accordance with experimental measurements. After the 43 iterations o ptimized activation energies are 60.0 and 44.0 kcal/mol and pre exponential factors are 1.261021 and 8.91013 hydride elimination, respectively

PAGE 95

95 Table 4 1. Dimensionless groups in modeling equations Name Definition Physical meaning Typical order of magnitude Prandtl Pr Momentum diffusivity Thermal diffusivity 0.7 Schmidt Sc D Momentum diffusivity Mass diffusivity 1 ~ 10 Reynolds Re L Momentum flux by convection Momentum flux by diffusion 101 ~ 102 Peclet RePr Pe Thermal flux by convection Thermal flux by diffusion 101 ~ 102 Grashof 3 2gLT Gr Buoyancy force Viscous force 1 ~ 105 Rayleigh Pr RaGr Buoyancy force Viscous force 1 ~ 105 Damk hler (,)ref refRCTL Da C Characteristic time for flow Characteristic time for gas phase reacti on 103 ~ 103 (Table was taken from Ref. [82] ) Table 4 2 Comparison of experimental and kinetic theory derived diffusivities for H2gas pairs and collision diameters Gas Collision diameter () Actual binary diffusivity (m 2 /s) Calculated binary diffusivity (m 2 /s) H2 1.41 1.02 10 4 2.73 10 4 He 1.13 1.13 10 4 2.91 10 4 C 2 H 6 2.22 5.37 10 5 1.20 10 4 C4H10 2.34 3.61 10 5 1.11 10 4 C6H12 3.09 3.19 10 5 7.67 10 5 (Table was taken from Ref. [87] )

PAGE 96

96 Figure 41. Experimental and calculated heat capacity of ( A) nitrogen, ( B) eth e ne and ( C) butane. Blue circles show calculation data and red rectangles experimental data. All data are based on gas phase. Experimental data were taken from Ref. [93] for nitrogen, Ref. [94] for ethane, and Ref. [95] for butane. 28 29 30 31 32 33 100 200 300 400 500 600 700 800 900 1000 Temperature(K) Cp (J/molK) A 28 29 30 31 32 33 100 200 300 400 500 600 700 800 900 1000 Temperature(K) Cp (J/molK) A 0 20 40 60 80 100 100 200 300 400 500 600 700 800 900 1000 Temperature(K) Cp (J/molK) B 0 20 40 60 80 100 100 200 300 400 500 600 700 800 900 1000 Temperature(K) Cp (J/molK) B

PAGE 97

97 Figure 41. Continued. Table 4 3 Deviations of heat capacity between experimental and calculated values Temperature (K) Deviation for N 2 (%) Deviation for C 2 H 4 (%) Deviation for C 4 H 10 (%) 300 0.08 2.91 10.21 400 0.23 2.78 7.08 500 0.36 4.83 600 0.48 700 0.54 800 0.55 900 0.52 0 50 100 150 200 250 100 200 300 400 500 600 700 800 900 1000 Temperature(K) Cp (J/molK) C 0 50 100 150 200 250 100 200 300 400 500 600 700 800 900 1000 Temperature(K) Cp (J/molK) C

PAGE 98

98 Figure 42. Simulated ( A) temperature contours (unit is shown as C ) and streamlines and ( B) concentration contours in the reactor when 5 % triethylgallium with nitrogen carrier gas at 700 C Axial position [mm]60 70 80 90 100 110Radial position [mm]10 30 20 0 40 50 A750 497 375 200 204 208 221 201 749 200 201 110 192 166 Axial position [mm]60 70 80 90 100 110Radial position [mm]10 30 20 0 40 50 B0.026 0.049 0.050 0.010

PAGE 99

99 Figure 43. Radial triethylgallium ( A) temperature and ( B) concentration profile at different axial positions: distances from the heated susceptor are (a) 0.3, (b) 2,(c)4, (d ) 6, (e) 8, and(f) 10 mm. 0 100 200 300 400 500 600 Temperature [ C]A (a) (b) (c) (d) (e) (f) Radial position [mm]0 5 10 15 25 30 20 0 100 200 300 400 500 600 Temperature [ C]A (a) (b) (c) (d) (e) (f) Radial position [mm]0 5 10 15 25 30 20 0 0.01 0.02 0.03 0.04 0.05 0.06 Radial position [mm]Concentration [mol %]B (a) (b) (c) (d) (e) (f)0 5 10 15 25 30 20 0 0.01 0.02 0.03 0.04 0.05 0.06 Radial position [mm]Concentration [mol %]B (a) (b) (c) (d) (e) (f)0 5 10 15 25 30 20

PAGE 100

100 Figure 44. Optimization process using genetic algorithm followed by simplex method for solving inverse problem. ( A) and ( B) denote detailed process of A factor (pre exponential factor) and activation energy of homolysis and hydride elimination, respectively. ( C) describes trace for goodness of fit during the process of ( A) and ( B) Genetic Algorithm Simplex A Genetic Algorithm Simplex Genetic Algorithm Simplex A Genetic Algorithm SimplexB Genetic Algorithm Simplex Genetic Algorithm SimplexB

PAGE 101

101 Figure 44. Continued. Table 4 4 A factor and activation energy value of homolysis and hydride eliminationreaction before and after optimization H omolysis hydride elimination A factor Activation energy A factor Activation energy Before optimization 4.681022 65.0 1.951013 40.3 After optimization 1.261021 60.0 8.91013 44.0 Genetic Algorithm SimplexC Genetic Algorithm Simplex Genetic Algorithm SimplexC

PAGE 102

102 Figure 45. Concentration and temperature profile of gas phase triethylgallium 0 0.01 0.02 0.03 0.04 0.05 0.06 0 2 4 6 8 10 12 14 Distance from susceptor [mm] Mole fraction 0 100 200 300 400 500 600 Reactor temperature [ ] After optimization Calculated gas phase temperature Experiments Experiments Before Optimization 0 0.01 0.02 0.03 0.04 0.05 0.06 0 2 4 6 8 10 12 14 Distance from susceptor [mm] Mole fraction 0 100 200 300 400 500 600 Reactor temperature [ ] After optimization Calculated gas phase temperature Experiments Experiments Before Optimization

PAGE 103

103 CHAPTER 5 INVESTIGATION OF THE THERMAL DECOMPOSITIO N OF TRIETHYLGALLIUM USING IN SITU RAMAN SPECTROSCOPY AND DFT CALCULATIONS Overview Gallium based compound semiconductors have played an important role in many electronic and optoelectronic devices, and MOCVD is often the preferred method of film deposition. The e lectronic s industry has developed several useful Ga related materials and is still experiencing expansion. Amongthe commercially availableGa precursors, TEGa ((C2H5)3Ga, Triethylgallium) and TMGa ((CH3)3Ga, Trimethylgallium) have been most widely used in the CVD of Gabased semiconductor materials such as GaAs, GaSb or GaN [96 98] Although TEGa and TMGa are widely used for MOCVD, there are only few reports on the gas phase thermal decomposition mechanism of TEGa. Most of the research describes thermal decomposition on a surface such as GaAs or Si [6, 98, 99] Those studies gas phase decomposition mechanisms for TEGa only suggest either simple homolysis of the C hydride elimination as the initial step[7, 8, 100] As summarized in Table 5 1, the reported experi mental activation energy for removal of the first ethyl group ranges range from 22.6 to 47.2 kcal/mol, and frequency factor ranges from 8.1 104s1to 4.8 1015 s1[7, 75, 101] To a certain ext ent, the reason why frequency factors show the large deviation is because each experiment was carried out at different conditions, e.g. carrier gas, reactor temperature, and reaction that was taken place. Unfortunately, this wide range in the values of the rate parameters produce s large uncertainties in reactor modeling results. These deviations may in part be attributed to chemical effects on the decomposition reactions arising from selection of carrier gas (i.e., toluene or hydrogen)

PAGE 104

104 and reaction conditions that may favor one mechanism over another. In addition to these values derived from experimental data, the activation energy values 60.0 kcal/mol hydride elimination were calculated using ab initio calculations (restricted HartreeFock MO method) by Tsuda et al [74] DFT calculation results from this work suggest activation energy values of 65.0 kcal/mol for hydride elimination, and pre exponential factors of 4.7 1022 s1 and 1.9 1013 s1, respectively. In this study, the TEGa homogeneous decomposition kineticswere investigated using in situ Raman spectroscopy in a vertical upflow, cold wall CVD reactor T he results of this nonintrusive measurement techniquewere used to estimate rate parameters using adetailed reactor model Theoretically its reaction pathways are going to be validated with DFT calculations were used to assist in assignment of Raman shift s and estimating rates and thermodynamic driving forces for selected reaction pathways Experimental and Theoretical Methods To better understand the decomposition mechanisms of TEGa, this study used experimental CVD reactor shown in Fig 5 1 that is interfaced with a Raman spectrometer (Ramanor U 1000, Jobin Yvon) which includesa double additive monochr omator and uses the 532.08 nm line of Nd:YAG solidstate laser as the light source. As described in detail elsewhere[91, 102] this CVD reactor is an upflow impinging jet reactor designed to produce a stable 2D flow condition while isolating the reactants from the reactor walls to prevent wall deposition. The entire CVD reactor assembly could be translated x y z to allow measurement of temperatur e and composition profile s. Three concentric inlet tubes are incorporated into the design with

PAGE 105

105 the reactant input into the center tube. Each tube is packed with 3mm glass beads to provide an equal velocity, parallel flow inlet boundary condition. TEGa a s a metal organic precursor is introduced through a center line and N2 carrier gas envelops is introduced through the annular and outer inlets to prevent wall deposition of precursors or reaction products. Based on the previous study [91] that developed and validated a steady state, two dimensional mass transport model and performed a CH4 tracer experiment, conditions were used t hat assure no recirculation flow s in the reactor. The measurement of temperature using Raman spectroscopy is well established. In low density flows, several authors have succeeded in measuring local temperatures with oxygen rotational spectrum at 1 atm r anging from 243 to 343 K [103] Because Raman scattering does not disturb the flow it can give relatively accurate results. It has been used to determine temperature in reactive flows [104] flame s[24, 105] and in the atmosphere[23] It was reported that temperature measurements with rotational state distribution using Raman spectroscopy can be accu rate to less than 7 % uncertainty in the range of 202230 oC [18] From t he modified equation for Stokes Rama n scattering intensity of Eq. 5 1, temperature distributions are obtained from linear regression of experimental data. 1 4 2,(23) () ln ()(1)(2)()()oJJJ BJ FJhc C gJJRJfJkT ( 5 1) where 10ln7/30(21)exp[()/]JB JCNCgJFJhckT and this is a constant at a given experimental setup and o is the wavenumber of the incident light. J, h, kB, T, R() and F(J) are rotational quantum number, Planck constant, Boltzmann constant, spectral

PAGE 106

106 response of the spectrometer and the correction to the anisotropy of the polarizability, respectively. If R() and F(J) are known, the temperature is determined from the linear regression method from a set of rotational peaks. With the nitrogen Q branch standard, the relative Raman cross section k for vibration mode k is defined as : 2 224 0 4 0() (/) 1exp(/) (/)()1exp(/)N kk k NkNdd hckT dd hckT ( 5 2) where k and 2N are, respectively, the wavenumbers of the molecular vibration and Q branch of N2 carrier gas. Relative Raman cross section of kth material can be calculated with Eq. 5 2 using Q branch of the N2 vibrational motion of 2N which is found stably with relatively high intensity at 2331 cm1. In Eq. 5 2, the exponential Boltzmann factor for nitrogen can be omitted at ambient temperature. A relationship between the Raman intensity ratio and the number density ratio for two components is necessary to perform the concentration measurement in a detection region. One should note that the quantity k is closely related to the relative scattering coefficient and the scattering activity. When performing concentration measurement with Raman spectroscopy, measured Raman apparent intensity and the number density ratio are to be calculated at the region of interest. Using Eq. 5 2 and equation of thermal population at temperature T, the relationship between the number density ratio, 2/NkNN and the intensity ratio is described as follows: 222(/) (/)NNN kkkNIdd NIdd

PAGE 107

107 22 24 0 4 01exp(/) () ()1exp(/)NN k k kN kI hckT I hckT ( 5 3) For a binary system, the mole fraction of component k is given by 2k k kNN x NN 21/1N kN N 22 24 0 4 01exp(/) () 1/1 ()1exp(/)NN k k kN kI hckT I hckT ( 5 4) However, one should note that the relative normalized Raman cross section has to be measured to use Eq. 5 4. Although the relative Raman cross section for most of the metal organic sources is not known, various methods such as relative Raman intensity meas urement with using circumventive materials and solvents [106] and density functional theory calculation[107] provide different techniques for understanding optical characteristics better than ever. In a preliminary experiment, the relative Raman cross section of TEGa was measured in the reactor at room temperature and atmospheric pr essure. In this experiment, 2.5 cm/s steady flow of 5 mol % TEGa in N2 as a carrier gas in the center line was used. For the annulus and sweep flows, pure N2 gas was delivered with the same flow rate (2.5 cm/s) and sufficient gas flow time before measureme nts was allowed for form a steady state and stable flow pattern. Then the 1.5W Nd:YAG solidstate laser line was used to excite TEGa source and the Ga C3 vibrational Raman excitation line (490cm1 line) was recorded (Fig 5 2).

PAGE 108

108 Although the scattered intensity of TEGa is much weaker than other group II, III metal sources as shown in T able 52 the signal to noise ratio was sufficient to quantify the TEGa concentration from the recorded 490cm1 excitation line. With repeated measurements (10 times) and long integration time (5 s) at room temperature, the relative Raman cross section was estimated to be 2.7. This value is relatively low in compari son with other well known metal organic sources e.g. 17 5 for trimethylgallium, 21.0 for dimethylcadmium By connecting experimental observations with computational calculations, one can identify intermediate Raman signals and provide thermal decomposition mechanisms. All calculations were performed using Gaussian 03 program suite[73] For all bond dissociation, activation energy and frequency calculations, the density functional theory level with the spin polarized hybrid density fu nctional B3LYP together with the LanL2DZ for all elements was used. The reference experimental value of Ga C3 bond length and Ga C C bond angle that was measured with diffractometer from the TEGa single cylindrical crystals of 0.3mm diameter [71] were used for choosi ng model chemistry. After the discreet comparison of a large number of cross combinations of methods (HF, B3LYP, MP2, MP4) and basis sets (631G*, 6 31G**, 6 311G, 6 311G**, SDD, LanL2DZ) as well as the split basis set of B3LYP/(LanL2DZ for Ga, 6 311G for other elements), the B3LYP/LanL2DZ model chemistry was chosen by comparing the experimental Ga C3 average bond length (1.966 ~ 1.996) and the Ga C C bond angle (113.3 ~ 116.1) with the former (1.995) and the latter (115.3) of computational calculation results. According to the crystrallographic data, there are two forms of TEGa structure. Three of the four molecules are found in a propeller like arrangement with C3

PAGE 109

109 symmetry and only one molecule does not exhibit the same symmetry and has two ethyl gr oups pointing at each other. Mitzel et al. [71] also reported that the wide GaC C angles in the TEGa structure resemble those in the B C C structure of solid B(C2H5)3, which was explained by hyperconjugat ion of the boron p orbital into C H orbitals [108] The tight convergence option was used for the TEGa geometry optimization, especially for the Ga C3 vibrational frequency, since this investigation focuses on the precise frequency value among the many kinds of Ga C vibrations of expected intermediates. As described above, calculated values using B3LYP/LanL2DZ model chemistry will be used for validating the frequencies of reaction intermediates. Results and D iscussion DFT C alculations The basic procedure in using DFT calculations to understand gas phase reactions is to first identify possible reaction pathways, perform DFT calculatio ns on the relevant gas phase species to estimate their thermodynamic properties, and then envision transition states for each reaction and perform DFT calculations to estimate rate constants Two major routes have been proposed for the initial stages of t hermal decomposition of TEGa by several researchers for gas phase reaction as well as adsorbed forms on surfaces such as GaAs or Si [6, 8, 98100] One mechanism is homolysis of the C G a bond to yield an ethyl radical hydride elimination, both of which could occur homogeneously or heterogeneously. Specific studies have proposed either just one of the mechanisms [99] or both mechanisms [8, 98, 100] based on the detected decomposition products It is noted that detection of butane or ethylene is decisive evidence to discriminate between hydride elimination, respectively. To the best of this investigators knowledge, however, subsequent reactions such as dimerization of two radicals or

PAGE 110

110 association of two intermediate products, have not been studied eithercomptationally or experimentally. Since these two major mechanisms are not exclusive, this study assumes initially both routes are possible DFT calculations were per formed to represent species in the TEGa decomposition system. Based on the TEGa structure data, the B3LYP level calculation with LanL2DZ basis set was also chosen to investigate the NBO (Natural Bond Orbital) analysis and evaluate the thermodynamic proper ties enthalp y entropy and Gibbs energ y Two initial stages, i.e. hydride elimination, are supported by the Wiberg bond index in the NBO analysis. In the case of homolysis, the Wiberg bond index between Ga and C (0.5976) is much smaller than that between C and C (1.0563), the products from C C breakage such as CH3 from the ethyl group are thus not considered as a thermal decomposition pathway hydride elimination case, it is more complicated since the transition state whi ch includes the imaginary bond between hydrogen connected with C and gallium center should be considered. Additional information on the imaginary ring which consist s of gallium, C, C and hydrogen connected with C was analyzed from Wiberg indices chang e. These changes with TEGa energetics and schematized TEGa structure are shown in Fig 5 3. The four numbers beneath the type of structure indicate Wiberg bond indices for bonds 1 ~ 4 depicted in the schematized TEGa structure. From Fig 5 3, the transiti on of the bond strength for the four membered Ga C C H imaginary ring can be easily investigated. As clearly shown in Fig. 5 3 bonds1 and 3 are increasing in strength and bonds 2 and 4 are becoming hydride elimination. From the Wiberg indices profile s one can suggest hydride elimination of TEGa.

PAGE 111

111 The theoretical enthalp y y y changes of the selected association and di hydride elimination reactions in the thermal decomposition pathways of TEGa at 298 and 900K as well as all the possible reaction pathways leading to formation of the 4membered ring, (MEG)4, consisted of 35 species and 39 reactions including three transition states hydride elimination. This result sare summarized in T able 53 R1 R3 and R5 R7 in T able 53 hydride elimination reactions and homolysis reactions, respectively. T1 through T5 describe transition state energy changes for TEGa, (Et)2GaH and (Et)GaH2 hydride eliminations. One should pay attention to the fact that some reactions are connected with others. For example, R8 R11 R12R14 R15 R18 R19 R20 R25 R26 R27 R28 and R29 R31 respectively, are under competition. Moreover R19 and R23 are indispensable reactions for R21 R24 and R27 R28 respectively. That means if R19 is not available, R21 R24 are not feasible and then R27 R28 will not be favorable automatica lly. As described above, two natural forms of TEGa should be included for three hydride elimination reactions. Each transition structure is constructed from the different TEGa structure. TEGa transition structure 1 (TEGa TS1) comes from t he structure that has two ethyl groups pointing at each other and TEGa transition structure 2 (TEGa TS2) comes from the propeller like arrangement. For the (Et)2GaH transition state structures, two different types ((Et)2GaH TS1 and TS2) were construct ed and these two structures are available from any two TEGa transition structures. Lastly, the (Et)GaH2 transition structure has only one transitional form, and these schemes are depicted in Fig 5 4. As stated in the Table 53 because the first (T1

PAGE 112

112 and hydride elimination steps have almost same energy barriers, initial structure of TEGa would not be very critical. hydride elimination, simple elongations of the ethyl group fr om Ga center constitute homolysis and its energy variation profile shows a smooth curve without a sudden upturn in potential energy. One could confirm the potential energy profile that shows continuous decrease during the elongation of ethyl radical. Bas ed on the potential energy surface calculation for the elongation of the ethyl radical from TEGa molecule, the potential energy of TEGa follows the singlet state for short separation distance and triplet state at long erseparation distance A Morse potential fit and these two potential energy surface calculati ons show very good agreement. A m ore detailed computational calculation was performed for two initial stages of TEGa thermal decomposition using same model chemistry as shown in Fig 5 5 To the best of this investigators knowledge, since there is no previous research which dealt with entire mechanism of TEGa thermal decomposition and proposed evidences, thermodynamic properties of these reactions was used for selection priority. The f results of two main reaction series considering five transition state structures shown in Fig 5 5 and minute investigation of selected decomposition pathways, 17 feasible rea ctions were screened on the basis of reaction enthalpies as shown in Fig 5 6. Many reactions were eliminated due to too high reaction energy and abse nce of source intermediates

PAGE 113

113 TEGa Thermal D ecomposition E xperiments U sing in situ Raman S pectroscopy To in vestigate TEGa thermal decomposition pathways in an upflow, impinging jet cold wall CVD reactor, a set of experiments was performed under strictly controlled conditions. As described earlier, 5% TEGa was introduced by the pure N2 into the reactor centerl ine. Nitrogen was selected as a carrier gas, since it has a large relative Raman cross section and it gives a strong Raman band that is used to normalize the peaks of intermediates. The susceptor heater was set at 750oC to provide sufficient energy for t he thermal decomposition. The relative mole fraction of each gas phase chemical species in the reactor was obtained from the ratio of peak integral of the primary peak [Ga C3] at 490cm1 to that of the N N vibration at 2331cm1. Fig 5 7 shows the Raman s pectrum that was recorded by measuring several centerline points from the susceptor to the inlet. Fig 5 7 shows five ranges of Raman spectr asuggestive of gas phase TEGa and detectable intermediates. The data were acquired in a series of scans over the ful l range of wavenumbers. Four bands at i.e. 517, 537, 555, and 785cm1, may indicate an unknown intermediate vibrations. T here are also observations for expected hydrocarbon products around 3100cm1 Raman shift. Given the large number of possible hydrocarbon products and motions, it is not easy to assign these signals to the exact vibrational motions. Al though spectral lines for Ga Ga vibration are proof of the presence of many di gallium intermediates such as (DEGa)2, (Et)GaH GaH2 and (GaH2)2, those spectral lines are not apparent in the scans From DFT calculations the Ga Ga vibration spectral lines are expected in the region around wavenumber of 190through 200cm1. Because intensities of nitrogen rotational transition lines are periodical and considerably

PAGE 114

114 stronger than those of other signals in the region of ~220cm1, it is difficult to make a ssignements Based the known characteristics of the nitrogen rotational bands, the spectral lines for vibration of Ga Ga are not easy to detect in the m easured spectrum, which overlaps with the rotational transitions of the nitrogen carrier molecules. If more evidence can be supplied from the Raman spectrum, it will not be difficult to assign each spectrum to a certain molecule. However, quantum calculation s have shown thatthey can predict Raman shifts of intermediate molecules with considerable exactness [109] The concentration of TEGa was calculated from measured Raman scattering intensities using Eq. 5 2 and the temperature profile inside the r eactor was calculated with Eq. 5 1. Fig 5 8 represents the measured TEGa concentration at the centerline below the susceptor heated to 750oC. It should be noted that experimental results with error bars agree well with concentration profile which considers both reactions, i.e., hyd ride elimination. As shown in the Fig 5 8, the experimental concentration cannot be explained using only one specific reaction. This suggests both hydride elimination and homolysis reactionsare active during TEGa thermal decomposition. Fig 5 8 also indicates temperature (orange solid line) and concentration (thin black solid line) profile simulated using the 2 D axisymmetric reactor model. This reactor model is described in more detail elsewhere[91, 110] and those references contain its validation procedures. In short, the 2D axisymmetric reactor model was developed to describe the experimentally observed temperature and concentration profiles and it uses the f inite element Galerkin method (FEM) to simulate heat, momentum and mass

PAGE 115

115 transfer mechanisms simultaneously. For optimization of the frequency factor and activation energy from measured data on the disappearance of TEGa, a genetic algorithm followed by si mplex algorithm was implemented. The parameter estimation was performed by minimizing the objective function that is the sum of squares of the differences between experimental data and simulated data for total 8 measurements. More detailed description of optimization scheme is reported elsewhere[91] Fig 5 9 shows thevalues of the frequency factor and activation energy for two main reactions at each step during the optimization. Optimized activation energies are 60.0 and 44.0 kcal/mol and frequency factors are 1.261021s1and 8.91013s1 hydride elimination, respectively. Interestingly, rate parameters of homolysis are hydride elimination are decreased after optimization. The Comparison of DFT Calculation w ith Experimental R esults A value of 474cm1 was produced for the symmetrical Raman active [Ga C3] stretching f rom the B3LY P/LanL2DZ model chemistry Based on the DFT calculations and the experimental results, a scaling factor of 1.0408 was produced. Eleven species including TEGa related to the 17 screened reactions were examined and corrected with scaling factor using same m odel chemistry (see Table 54 ) T his t able shows the calculated and corrected Raman frequencies for each selected species from the stable reactions accompanied by the Ga C symmetric motions. The scaling factor of 1.0408 mentioned above is applied only to the Ga C stretching mode and frequencies from calculation results were corrected. As stated in Table 5 4, four frequencies related to three molecules matched well with experimental Raman shifts. Especially, one of the two corrected frequencies for (DEGa)2 shows exact matching with 517cm1 experimental Raman shift. Moreover, another corrected

PAGE 116

116 frequency of (DEGa)2, frequency from (Et)GaH Ga(Et)2 and one from (Et)GaH GaH2is in good agreement with 555, 537 and 555cm1, respectively. Alt hough there are two candidates for 555cm1, it is not obvious which frequency is more accurate since relative Raman cross sections of two molecules are not known. Moreover, although long absorption time is not mandatory for Raman spectroscopic measurement, in the case of depleting intermediate materials simultaneously such as trimethylindium decomposition reaction[111] the lifetime of these molecules can be critical. Investigation of T able 54 firmly suggests the presence of (DEGa)2, (Et)GaH GaH2, and (Et)GaH Ga(Et)2. According to the presence of (DEGa)2 as a reaction product of the DEGa, presence of the homolysis reaction which make DEGa and MEGa radical can be supported though hydride elimination reactions. Moreover, the presence of (Et)GaH GaH2, and (Et)GaH Ga(Et)2 evidences hydride elimination decomposition. Based on these results, it is suggested that both hydride elimination reactions take place during the thermal decomposition of TEGa in the CVD reactor. Conc lu ding Remarks An analysis of the decomposition kinetics for TEGa with N2 carrier gas in a custom upflow, cold hydride elimination and homolysis reactions are present in the reaction zone of the reactor. This was ev idenced by observation of (DEGa)2, (Et)GaH GaH2, and (Et)GaH Ga(Et)2 using in situ Raman spectroscopic and values of the reaction enthalpy preferences computed by DFT calculations at the B3LYP/LanL2DZ chemist ry level. Raman shifts of 490, 517, 537, and 555cm1 were found and assigned to TEGa, (DEGa)2, (Et)GaH Ga(Et)2, (Et)GaH GaH2, respectively. The concentration profile for TEGa and the temperature distribution

PAGE 117

117 of reactor centerline were measured and these were compared to those from the simulated result s using finite element Galerkin method. Activation energy and frequency factor for both reactions were optimized using genetic and simplex algorithm and optimized rate parameters show better fit with experimental results. The similarity confirmed from ex periment and DFT calculation results is remarkable and suggests this methodology can be considerable for metal organic materials research with in situ Raman spectroscopy, which does not interrupt the flow pattern in the CVD reactor.

PAGE 118

118 Table51 Experiment al and c alculated r ate p arameters for f irst e thyl g roup d issociation T(K) k 0 Ea (kcal/mol) Method Ref 610 ~ 749 543 ~ 1023 573 ~ 653 553 ~ 673 585 600 750 ~ 850 750 ~ 850 400 ~ 900 400 ~ 900 4.8 10 15 8.0107 8.1104 2.4101 0 4.71022 1.9 10 1 3 47.2 22.7 22.9 32.0 37.8 38.7 60.0 43.6 65.3 40.5 Flow system, toluene carrier Flow system, hydrogen atmosphere Static system desorption on GaAs decomposition on GaAs(100) decomposition on GaAs(100) Hartree Fock (Homolysis) Hartree hydride elimination) DFT (homolysis) elimination) [7] [7 5] [75] [9] [ 6] [6] [74] [74] This work This work Figure 5 1. Schematic of the experimental reactor for in situ Raman spectroscopic measurements

PAGE 119

119 Figure 5 2. Recorded Ga C3 vibrational measurement of 490cm1line T able 52 Reported relative Raman cross section for group II, III metal organic sources Precursor DMCd TMGa TMIn DEZn TEGa Relative Raman Cross section 21.0a 17.5b 22.3c 4.2a 2.7d (a : Ref. [109] b : Ref. [112] c : Ref. [113] d: this work) 0 20 40 60 80 100 120 440 460 480 500 520 540 Intensity [a.u.] Raman shift [cm 1 ]

PAGE 120

120 Figure 5 hyd ride elimination with Wiberg indices

PAGE 121

121 T able 53 phase decomposition reactions at 298K and 900K using B3LYP/LanL2DZ model chemistry (E t : ethyl group, C2H5 ) 298K 900K R1 (Et)3Ga = (Et)2GaH + C2H4 T1 (Et)3Ga = (Et)3Ga TS1 T2 (Et)3Ga = (Et)3Ga TS2 R2 (Et)2GaH = (Et)GaH2 + C2H4 T3 (Et)2GaH = (Et)2GaH TS1 T4 (Et)2GaH = (Et)2GaH TS2 R3 (Et)GaH2 = GaH3 + C2H4 T5 (Et)GaH2 = (Et)3Ga TS R4 GaH3 = Ga + 3/2H2 R5 (Et)3Ga = (Et)2Ga + C2H5 R6 (Et)2Ga = (Et)Ga + C2H5 R7 (Et)Ga = Ga + C2H5 R8 2(Et)2GaH = (Et)HGa -Ga(Et)2 + C2H6 R9 2(Et)2GaH = ((Et)GaH)2 + C2H6 + C2H4 R10 2(Et)2GaH = ((Et)GaH)2 + C4H10 R11 2(Et)2GaH = ((Et)2Ga)2 + H2 R12 2(Et)GaH2 = ((Et)GaH)2 + H2 R13 2(Et)GaH2 = (Et)HGa -GaH2 + C2H6 R14 2(Et)GaH2 = (GaH2)2 + C4H10 R15 (Et)GaH2 + (Et)2GaH = ((Et)GaH)2 + C2H6 R16 (Et)GaH2 + (Et)2GaH = (Et)HGa -GaH2 + C4H10 R17 (Et)GaH2 + (Et)2GaH = (Et)HGa -Ga(Et)2 + H2 R18 (Et)GaH2 + (Et)2GaH = (Et)HGa -GaH2 + C2H6 R19 C2H6 = 2CH3 R20 C2H6 = C2H4 + H2 R21 (Et)Ga + 2CH3 = (Et)Ga(CH3)2 R22 (Et)2Ga + CH3 = (Et)2Ga(CH3) R23 C2H5 + CH3 = C3H8 R24 (Et)Ga + CH3 + C2H5 = (Et)2Ga(CH3) R25 2C2H5 = C2H6 + C2H4 R26 2C2H5 = C4H10 R27 (Et)2Ga + C3H8 = (Et)2Ga(CH3) + C2H5 R28 (Et)2Ga + C3H8 = (Et)3Ga + CH3 R29 (Et)2Ga + C4H10 = (Et)2Ga(CH3) + C3H7 R30 (Et)2Ga + C4H10 = (Et)3Ga + C2H5 R31 (Et)2Ga + C4H10 = (Et)2Ga(C3H7) + CH3 R32 2(Et)2Ga = ((Et)2Ga)2 R33 3(Et)Ga = ((Et)Ga)3 R34 4(Et)Ga = ((Et)Ga) 4 25.47 38.60 38.33 26.46 37.18 37.03 26.06 34.88 32.85 65.24 73.02 8.61 -7.06 19.45 0.17 2.86 1.29 -9.63 -15.60 -7.00 -17.02 1.24 -9.32 83.06 34.75 -151.54 -71.72 -80.42 -144.74 -58.52 -77.80 19.84 15.17 8.72 12.56 11.72 -52.16 -152.76 249.20 32.64 -2.16 -3.16 38.90 -0.70 -1.21 28.75 -5.94 34.48 51.81 44.42 29.48 -8.15 36.24 1.55 2.52 -12.56 -1.80 -11.24 -2.66 2.08 -18.06 -2.69 41.04 28.99 -74.05 -36.87 -45.60 -81.29 -17.61 -52.31 17.18 -6.21 4.23 0.50 -4.64 -31.19 -74.97 117.35 15.74 39.24 39.27 14.86 37.39 37.39 17.49 36.65 22.58 49.81 59.78 -0.18 -4.63 8.65 -0.30 2.11 5.03 -9.10 -12.25 -6.21 -17.64 6.62 -8.52 70.83 26.11 -129.48 -60.73 -66.83 -120.51 -53.27 -62.22 8.93 17.02 7.46 12.41 13.11 -42.86 -130.42 214.22 24.62 39.75 39.46 26.85 38.30 38.15 25.30 34.76 33.69 63.22 73.12 8.81 -9.23 18.81 0.42 7.17 2.03 34.19 -17.53 -8.03 28.27 0.84 -10.31 84.98 36.92 -151.27 -71.03 -81.93 -144.15 -60.49 -78.88 21.87 18.71 9.68 15.66 13.18 -47.80 -150.97 244.76 29.47 -0.48 -1.54 -7.58 0.91 0.40 25.79 -6.84 36.07 40.78 43.62 29.26 1.27 34.38 2.42 11.82 -10.52 28.42 -14.82 -3.87 34.71 -17.42 -3.79 43.30 31.60 -71.43 -34.43 -47.08 -78.05 -21.00 -52.95 12.65 6.29 5.67 12.17 -2.54 -21.42 -69.51 105.44 -1.90 40.19 40.84 38.25 37.48 37.79 2.09 40.91 1.23 26.52 33.86 -17.53 -11.60 -12.13 -1.76 -3.47 11.50 8.62 -4.19 -4.55 -2.97 17.32 -6.89 46.01 8.48 -86.98 -40.04 -39.56 -73.90 -41.59 -31.22 10.49 13.04 4.57 4.71 15.46 -28.53 -88.41 149.87

PAGE 122

122 Figure 5 4. hydride elimination transition state structures ((a) TEGa TS1, (b) TEGa TS2, (c) (Et)2GaH TS1, (d) (Et)2GaH TS2, (e) (Et)GaH2 TS) Figure 5 5. Calculated energetics of the two major thermal decomposition pathways of TEGa with reac tion enthalpies [kcal/mol] listed.

PAGE 123

123 Figure 5 6. 17 screened reaction pathways out of 34 with reaction enthalpies [kcal/mol] listed Figure 5 7. Raman spectrum of gas phase TEGa. Different horizontal line means different height measurement along the ce nterline. ((a) 1mm (b) 2mm (c) 4mm (d) 6mm (e) 8mm (f) 10mm (g) 11mm from the heated susceptor).

PAGE 124

124 Figure 5 8. Simulated and experimental concentration profile of TEGa in the reactor 0 100 200 300 400 500 600 0 0.01 0.02 0.03 0.04 0.05 0.06 0 2 4 6 8 10 12 14 Reactor temperature [ ] Mole fraction Distance from susceptor [mm] No reaction Homolysis only Both reactions (before optimization) Calculated gas phase temperature Experiments hydride elimination only Experiments Both reactions (after optimization)

PAGE 125

125 A B Figure 5 9. Optimized values of rate parameters of two main reactions for ( A ) activation energy and ( B ) A factor (pre exponential factor) 34 36 38 40 42 44 46 59.95 60 60.05 60.1 60.15 60.2 0 10 20 30 40 Activation Energy of elimination Activation Energy of Homolysis Number of Generations Genetic Algorithm Simplex 0 1 2 3 4 5 6 7 8 9 10 6 8 10 12 14 16 0 10 20 30 40 A factor of elimination ( 10 13 ) A factor of Homolysis ( 10 20 ) Number of Generations Genetic Algorithm Simplex

PAGE 126

126 Table 54 Calculated and corrected Raman active Ga C stretching frequencies only for the symmetric motions (Lowercase s, m and w attached to the calculated frequencies mean str ong, medium and weak intensity, respectively) Calculated Experimental Scaling Factor Corrected TEGa DEGa MEGa (Et)2GaH (Et)GaH2 (Et)GaH Ga(Et)2 (Et)GaH GaH2 DEGa CH3 (DEGa)2 (MEGa)3 (MEGa) 4 470.8 s 453.9 s 336.4 m 490.8 m 450.5 m 563.1 m 513.4 m 533.0 w 486.0 m 496.8 m 531.0 m 493.0 s 526.4 s 490 537 555 517 555 1.0408 1.0408 1.0408 1.0408 1.0408 1.0408 1.0408 1.0408 1.0408 1.0408 1.0408 1.0408 1.0408 490 472.4 350.1 510.8 468.9 586.1 534.3 554.7 505.8 517.1 552.7 513.1 547.9

PAGE 127

127 CHAPTER 6 HOMOGENEOUS THERMAL DECOMPOSITIO N STUDIES OF TRIETHYLALUMINUM(TEA L) : EFFECT OF NH3 Overview Aluminum nitride is a wide and direct bandgap semiconductor (6.2 eV at 300K) and a refractory ceramic with a high thermal conductivity (285 W/ mK ). In recent years t he properties of aluminum nitride thin films make this material interesting for various applications such as blue and UV photodetectors and light emitting diodes [114, 115] However, most interest is focused toward the properties of its alloys with GaN which can be fabricated AlGaN based devices. Among various deposition technique, metalorganic chemical vapor deposition (MOCVD) provides a useful strategy to grow III V compoun d semiconductor. For AlN film, trialkylaluminum compounds, such as triethylaluminum (TEAl) and trimethylaluminum (TMAl), are the most widely used metalorganic aluminum sources. However, AlN films deposited using TMAl usually contain very high levels of car bon [116] Since TEAl has three ethyl groups and these can be eliminated through c hydride eliminations even under relatively low temperature, TEAl has been focused as one of the good alternatives, such as, triisobutylaluminum (TIBAl), dimethylaluminum hydride (DMAH), triethylamine alane (TEAA) and trimethylamine alane (TMAA ) [116] Ho et al. have deposited films that have comparatively better crystallinit y with TEAl and NH3 at 723 ~ 1173K [117] Although TMAl and TEAl are commonly used for MOCVD, it is very difficult to find studies for the gas phase thermal decomposition, especially, considering the effect of ammonia on the mechanism. Like the other triethyl metal precursors, such as TEGa and TEIn, TEAl h as hydride eliminations and homolysis and Smith and his coworker have reported both reactions using IR and mass spectrometry by

PAGE 128

128 detectinghydrocarbons such as C2H4 and C4H10 hydride eliminations and homolysis, respectively [118] In this study, TEAl decomposition kinetics in a vertical, impinging jet, coldwall CVD reactor were investigated and reaction intermediates were identified. In addition to TEAl study, ammonia effect on TEAl decomposition mechanism was studied using gas phase in situ Raman spectroscopy and DFT calculations by Raman shift corrections. E xperimental and Theoretical Methods To elucid ate homogeneous thermal decomposition mechanisms of TEAl in CVD reactor, the reactor configuration in Fig. 6 1 was used. CVD reactor shown in Fig. 6 1 was interfaced with in situ Raman spectrometer (Ramanor U 1000, Jobin Yvon) that has double additive monochromator for better resolution. 532.08nm line of Nd:YAG solidstate laser was used as the excitation of metalorganic precursor.As described in detail elsewhere[119] this CVD reactor is custom designed cold w all, up flow impinging jet reactor to study the gas phase decomposition kinetics of TEAl. Metalorganic precursor and carrier gas including co reactant such as ammonia can be introduced to CVD reactor through three concentric inlet tubes center, annulus and sweep flows. All three tubes are packed with 3 mm glass beads to supply parallel flow from inlet to susceptor. Moreover, N2 sweep flow envelops metalorganic gas not to make wall deposition (Al2O3). In the case of introducing ammonia with TEAl, to prevent adduct formation of TEAl and ammonia before inlet, TEAl and ammonia were introduced center and annulus tube, respectively. Pure N2 was introduced into the reactor through three inlets at 21 oC as a carrier gas. The gas velocity was set at 2.5 cm/s for all inlets and sufficient residence time was allowed to maintain stable flow. The vapor pressure of TEAl is much lower than trimethylaluminum (TMAl) leading to its

PAGE 129

129 application and massive deposition be limited.However, owing to ethyl group, TEAl has advantage on low carbon contamination on film compositions. The vapor pressure of TEAl was estimated using Eq. 6 1: 2361.2 ()9.00 ()73.8 LogPmmHg TK ( 6 1 ) where T stands for temperature in Kelvin. Bubbler of TEAl was maintained 90 oC and the vapor pressure was calculated 6.91 mmHg (0.9 %) in the current experiment. Usually aluminum alkyls exist dimer form of AlR3 (R=CH3, C2H5, n C3H7, n C4H9, i C4H9, etc) in the liquid phase, however, most of those have monomer form in the gas phase[120] In the case of TEAl, the entropy of vaporization (176.6 J/ mol K ) is almost exactly two times larger than the value of 88 J/ mol K calculated by Troutons rule. This indicates that 1 mole of liquid is converted into 2 moles of vapor in the process of vaporization. T herefore, one can conclude TEAl dimers are present in the liquid, however the vapor consists of monomeric TEAl. Based on this information, gas phase TEAl introduced to CVD reactor is complete monomeric TEAl since the lowest temperature of CVD reactor is around 128 oC. Raman scattering signals were recorded along with the centerline of the reactor and heater temperature was set at 500 oC. Not to interrupt the flow pattern in the reactor, temperature measurement was performed with spectroscopic method. Since this technique does not make any hindrance in the reactor, it can make very accurate results and Raman spectroscopic temperature measurement could be accurate with less than 7% uncertainty in the range of 20~2230oC [18] From the modified equation for Stokes Raman scattering intensity, temperature distributions are obtained from linear regression of experimental data.

PAGE 130

130 To study ammonia effect on thermal decomposition mech anism of TEAl, two types of experiments were performed. First set of experiment used pure TEAl through center inlet and second experiment introduced 5 % ammonia into annulus inlet with nitrogen. Since ammonia was introduced annulus inlet and Raman scatteri ng measurement was obtained at the reactor centerline, one can easily expect ammonia concentration would be increase from inlet to the heated susceptor in an upflow reactor. The spin polarized hybrid density functional B3LYP with the LanL2DZ calculations using Gaussian 03 program [73] were performed to assist experimental results. Using this model chemistry, optimized geometry, activation energy, vibrational frequency were calculated at temperature range of 298~1000 K. Results and Discussion Homogeneous TEAl Thermal D ecomposition E xpe riments A few decades ago, Smith and his coworker have studied the thermal decomposition of TEAl in static systems over the temperature range of 162~192 oC [118] Even though that previous research was performed in static systems and under narrow temperature range, it can have a few similarities in decomposition mechanisms. They have followed the partial pressure of TEAl according to the duration of reaction and i t could make them for having kinetic parameters. From the Arrhenius plot for aluminum ethyl bond decomposition between 162.0 and 192.4 oC, the value for the activation energy, Ea, was 29 kcal/mol, and the A factor was 1.6 108s1. Like other ethyl containing metal compound, such as triethylgallium (TEGa), TEAl also has possibility to hydride elimination and homolysis reaction. hydride elimination and homolysis reaction proceed following manner.

PAGE 131

131 253 25224(CH)Al (CH)AlH + CH ( 6 2) 252 25 24(CH)AlH (CH)AlH + CH ( 6 3) 252324(CH)AlH AlH+ CH ( 6 4) 323 AlH Al + H 2 ( 6 5) 253 252 25(CH)Al (CH)Al + CH ( 6 6) 252 25 25(CH)Al (CH)Al + CH ( 6 7) 25 25(CH)Al Al + CH ( 6 8) Through Eqs. 6 2 to 6 8 TEAl can produce metal aluminum hydride elimination would have energy barrier for transition state to form hydride molecule and homolysis reaction proceeds through elongation of bond between Al and ethy hydride elimination have metal H bond, it would arise distinct signal other than C H bond (~2000 cm1 and ~3000 cm1 for metal H and C H, respectively). However, if it decomposes with co reactant, ammonia in this case, t hose two main mechanisms might be interrupted from co reactant. In other words, any intermediates can react with ammonia and following steps would be changed. 2533252226(CH)Al + NH (CH)Al-NH + CH ( 6 9) 252225 26(CH)Al-NH (CH)Al-NH + CH ( 6 10) 25 26(CH)Al-NH AlN + CH (under 3NH ambient) ( 6 11) Jiang and his coworker have reported AlN thin film and powder with TEAl with ammonia following above reactions, i.e., Eqs. 6 9 to 6 11 by chemical vapor deposition[121] Certainly, Eqs. 6 9 to 6 11 have to be considered adduct formation in

PAGE 132

132 hydride elimination and/or homolysis reaction products can react with ammonia, one can assume following side reactions. 252 325 226(CH)AlH + NH(CH)AlH-NH + CH ( 6 12) 25232226(CH)AlH+ NHHAl-NH+ CH ( 6 13) To study TE Al thermal decomposition pathways in an upflow, cold wall CVD reactor as described before, experiments w ere performed under well controlled conditions. 0.9 % of TEAl was introduced in to the reactor centerli ne with pure 2N N itrogen rotational Raman signals give us temperature information in the reactor and it has a large relative Raman cross section so that gives a strong Raman band that is used to normalize the peaks of intermediates. Sufficient residence time was allowed for stable flow in the reactor under the heater set at 500 oC. Although Kvisle and his coworker measured monomeric IR spectroscopy of TEAl [122] and Yamamoto had dimeric TEAl Raman spectroscopic measurement [123] to the best of my knowledge, no gas phase monomeric TEAl Raman signals have been reported. Since Al C stretch at 648 cm1 has been observed from IR spectroscopy, one can expect Raman signal might be observed at similar region. Unfortunately even the repetitive experiments, signals for aluminum ethyl vibration were not observed. However, a few signals for intermediates have been observed and it would give enough information for determining reaction mechanism. Fig. 6 2 shows Raman spectrum profile from the cold inlet to the heated susceptor for pure gas phase TEAl. From this first set of experiment, it is confirm ed that the observed signals of 600, 1989, 2025, 2580 2835, 2849, 2900, 2918, 2939, 3173 cm1.

PAGE 133

133 Raman shift of 600 cm1 can be assigned to H wagging of Al H in DEAlH (diethylaluminum hydride). This will be specified in detail on following chapter. Unlike 600 cm1, signals at the region over 2800 cm1 are not easy to assign since various C H vibrations have similar IR and/or Raman frequencies.To elucidate ammonia effect on TEAl thermal decomposition mechanisms, 5 % of ammonia was introduced through annulus inlet. Fig. 6 3 shows Raman spectrum observation along the reactor centerline from cold inlet to the heated susceptor. Raman shift at 449 cm1 is from NH3 rotational spectrum and 934, 964, 3219, 3334 cm1 are from NH3 vibrational spectra. However, one can find 452 cm1 shoulder attached to 449 cm1 signal and 1462, 1525, 1639, 2580, 2838 cm1 Raman shifts. Since observed spectra from two different sets of experiments have only small overlap, one hydride elimination or homolysis reaction are consumed right after being produced not to be observed or TE Al decomposition with ammonia are carried out with different routes compared with neat TEAl experiment. From these expectations, decomposition mechanism can be supposed and it would be confirmed and described in detail later. Although there are various co mplexes that can consume ammonia based on Fig. 6 3 experimental ammonia concentration profile in CVD reactor can provide information on mechanism. Fig. 6 4 shows ammonia concentration profile in CVD reactor for two cases, with and without reaction. And tem perature profile that was measured by laser is also shown in the Fig. 6 4 This technique that measures nitrogen rotational state distribution with focused laser does not make any hindrance to flow and

PAGE 134

134 shows high accuracy. Solid line with temperature measurement is simulated profile with FEM reactor modeling described in detail elsewhere[119] Ammonia concentration profile was measured for two cases, i.e., (a) without reaction and (b) with reaction. For case (a), 5 % ammonia was used at 500 oC without TEAl introduction to the reactor, and for case (b), same 5 % ammonia with 0.9 % TEAl were introduced to the reactor and ammonia concentration profile was dramatically changed at around 9 mm from susceptor. And then concentration of case (b) has increasing gap compared with case (a). One should note that ammonia concentration with reaction shows equilibrium at ca. 4 mm from susceptor. Numerous possibility can be supposed for this equilibrium, for instance, TEAl:N H3 adduct can release NH3 at higher temperature. Unexpectedly, this flat concentration region was shown at temperature ramping area. DFT C alculations In TEAl/NH3 systems, TEAl molecule makes a stable Lewis acidbase adduct bound to ammonia, i.e., (C2H5)3Al:NH3[114, 124] The adduct formation is supported by NBO (Natural Bond Orbital) an alysis even though various researchers have considered it fundamental reaction. By tracking Wiberg bond indices of each bond, it can be clearly explained. As described above, B3LYP method with LanL2DZ basis set was used for DFT calculations. From Wiberg indices shown in Fig. 6 5 one can assure that N in ammonia makes strong bond with Al. Moreover Al C bond is weakened and ethyl group with H in ammonia forms new moiety, i.e., C2H6, during reaction coordinate. It should be noted that each bond in four membe red ring, Al C H N has similar bond strength at the transition state. Al N bond strength of product is approximately three times higher than

PAGE 135

135 adduct. Detailed reaction mechanism with optimized geometry for TEAl adduct with ammonia is described in Fig. 6 6 T his energy change shows small differences with results from Ikenagas study even though he used same level of theory. [124] Due to Lewis acidbase interactions between TEAl and ammonia, adduct forms stable complex that has lower energy by 22.8 kcal/mol than reactants. And DEAl NH2 and C2H6 are produced via transition state that has slightly higher energy than initial reactants. Other than 1:1 ratio for TEAl and ammonia, one can suppose the possibility of reacting more ammonia with TEAl. If TEAl molecules are in the presence of excess ammonia, TEAl makes a stable complex with two ammonia molecules without potential energy barrier. This tendency is applied to other trialkyl metal compound with the form of R3M (R= CH3, C2H5, M=Al, Ga, In) [124] Interestingly, overall scheme in Fig. 6 7 does not show much difference with Fig. 6 6 Adduct, transition state and final products have similar relative energy compared with simple ammonia adduct case, i.e., TEAl:NH3.One of the products H2Al NH2 can form dimer or trimmer before deposition of AlN thin films. In this study, only dimer and trimmer were considered to make tetramer and hexane, respectively as shown in Fig. 6 8 There can be numerous possible alternative pathways to produce tetramer and hexane during the course of structural growth. In this study, only a specific route to the desired target is investigated computationally. One can find out detailed review of broad coverage from Timoshkins work [125] As an analogous complex, Hwang et al. have reported the thermal conversion of cyclotrigallazane, [H2GaNH2]3, to cubic GaN at 150

PAGE 136

136 oC [126] and Jegier and his coworkers showed the formation of [HGaNH]n from diamidogallium hydride and characterized IR spectroscopy [127, 128] The Comparison of Experimental R esults with DFT C alculations For assigning frequencies to intermediates, simulated frequencies using the B3LYP/LanL2DZ level of theory and reference data were used. DFT calculations suggest vibrational frequencies for intermediates suggested in the proposed decomposition mechanisms. E xperimental Raman active stretching frequencies with calculated and corrected values for all complexes including values for ammonia are shown in Table 6 1 One should note that three bands at 2835, 2841, 2849 cm1 are mixed in together and it should be dec onvol v ed into each signal. Spectrum was deconvol ved for more precise analysis using PeakFit v 4.12 And one can confirm the ammonia effect on spectrum around 2840 cm1 as shown in Fig. 6 9. Moreover, intermediates produced by hydride elimination such as DEAlH and DEAlAlH2 were not observed at the experiment of TEAl + NH3 system and ammonia related intermediates, i.e., H2Al NH2 and H2N AlHNHAlH2, were observed. From these results, DEAlH and DEAl AlH2 would react with ammonia fast enough not to be detect hydride elimination. Sauls et al. have reported TEAl:NH3 adduct forms [(Et)2AlNH2]3 (i.e., amide six membered ring) and it forms AlN via [EtAlNH]n (i.e., imide six membered ring) [129] However, amide formation was processed under liquid phase with (TEAl)2 at 50 oC. The peak at 1639 cm1 would be assigned to NH2 scissoring mode of H2Al NH2 and this assignment is also supported by computed result and Raman characteristic group frequency.

PAGE 137

137 Certainly one can see the lack of peak B in Fig. 6 9b. Peaks A, B, and C were assigned to C H st retching of MEAlH2, TEAl:NH3 and MEAlH AlH2, respectively, by frequency analysis. From above analysis, proposed decomposition mechanism of TEAl with and without ammonia as shown in Fig. 6 10 and it shows ammonia effect on TEAl decomposition process. Fig. 6 10 shows small overlap in mechanisms, (Et)AlH AlH2 formation and two hydride elimination products. And one should note that homolysis intermediates such as diethylgallium dimer, (Et)2Ga Ga(Et)2 or C4H10, were not detected from both experiments. Possibly relatively low operating temperature could prohibit homolysis reaction that has higher energy barrier. The fact that TEAl + NH3 system did not show homolysis intermediates can be one of evidences that homolysis is not a main process in these systems. When it comes to ammonia effect on TEAl decomposition, ammonia attacks TEAl and (Et)AlH2 directly and makes adducts with them. Although H2N AlH NH AlH2 was observed, four membered ring, [HAlNH]2 or higher membered rings were not detected, unfortunately. However, this observation can show one of stepping stones to AlN film formations. Concluding Remarks A study for ammonia effect on the thermal decomposition kinetics of TEAlwith N2 carrier gas in a custom upflow, cold wall CVD reactor was shown that combination of ammonia with hydride elimination product produce H2Al NH2 that is the another reactant for AlN film formation in the reaction zone of the reactor. By suggesting the presence of TEAl:NH3 adduct DEAlH, TEAl:NH3 TS, H2N AlHNHAlH2, H2Al NH2, MEAlH2, MEAlH AlH2 and DEAl AlH2, ammonia effect on TEAl decomposition mechanism was confirmed using in situ Raman spectrosc opy and DFT calculations at

PAGE 138

138 the B3LYP/LanL2DZ level of chemistry. Raman shifts of 600, 1989, 2025, 2580, 2835, 2849, 2900, 2918, 2939, and 3173 cm1 were found at the experiment of neat TEAl, and 452, 1462, 1525, 1639, 2580, 2853, 2841, and 2849 cm1 were found at the experiment of TEAl with ammonia. In addition to that, four vibrational spectra of 930, 965, 3230, and 3334 cm1 for ammonia wer e observed with high intensity. The temperat ure distribution of reactor centerline w as measured and this was compared to the si mulated results using custom FEM Galerkin method. Those measured and simulated temperature profiles agree very well. This methodology that compares the results from the experiment using in situ Raman spectroscopy with DFT calculations and FEM reactor model suggests considerable way to investigate the thermal decomposition mechanism in the CVD reactor since this method does not inter rupt the flow pattern and reaction pathways.

PAGE 139

139 Figure 6 1. (Color online) Schematic drawing of the CVD reactor interfaced with in situ Raman spectrometer Figure 6 2. (Color online) Raman spectrum of gas phase TEAl. Different horizontal signal means different height measurement along the centerline from the cold inlet (black bottom signals ) to the heated susceptor (red top signals ) 2550 2590 2780 2820 2860 3170 3190 2880 2920 1950 1990 2030 0 50 100 150 200 570 610 Intensity [a.u.] Raman Shift [cm1]

PAGE 140

140 Figure 63 (Color online) Raman spectrum of gas phase TEAl with ammonia Different horizontal signal means dif ferent height measurement along the centerline from the cold inlet (black bottom lines) to the heated susceptor (red top lines) Figure 64 (Color online) Measured and simulated temperature and ammonia concentration profile (with and without reaction) in CVD reactor 920 940 960 1430 1490 1550 1610 0 50 100 150 200 250 430 450 470 Intensity [a.u.] 2560 2590 2750 2850 3330 3340 3200 3250 Raman Shift [cm1]X 0.1

PAGE 141

141 Figure 65 Wiberg indices tracking of C2H6 elimination from TEAl:NH3 adduct Figure 66 (Color online) Relative energy diagram of C2H6 elimination from TEAl:NH3 adduct (Energies in parenthesis are fro m Ref. [124] )

PAGE 142

142 Figure 67 (Color online) Relative energy diagram of C2H6 elimination in the case of excess ammonia (Energies in parenthesis are from Ref. [124] ) Figure 68 (Color online) Simulated tetramer and hexane formation from H2Al NH2 with enthalpy change and free energy change (underlined) at 298K.

PAGE 143

143 T able 6 1 Calculated and c orrected R amana ctive stretching f requencies Molecule Experimental (without NH 3 ) Experimental (with NH 3 ) Calculated Corrected Assignment TEAl:NH 3 452 467 454 Al C str DEAlH 600 616 601 Al H wagging NH 3 930 N H str NH 3 965 N H str TEAl:NH 3 TS 1462 1496 1459 Al H N br H 2 N AlH NH AlH 2 1525 1565 1526 NH str H 2 Al NH 2 1639 1664 1622 NH sci 1989 2025 2580 2580 MEAlH 2 2835 2835 2997 2829 C H str TEAl:NH 3 2841 3012 2843 C H str MEAlH AlH 2 2849 2849 3023 2854 C H str DEAl AlH 2 2900 3077 2905 C H str DEAl AlH 2 2918 3093 2920 C H str MEAlH AlH 2 2939 3106 2932 C H str 3173 NH 3 3230 NH str (Abbreviations for the internal coordinates: vib, vibrational; sci, scissor; br, bridging; str, stretching)

PAGE 144

144 Figure 69. (Color online) Comparison of Raman spectra at 2840cm1 for two cases, (a) without ammonia, (b) with ammonia. Figure 6 10. Proposed overall decomposition mechanism from (a) neat TEAl and (b) TEAl with ammonia experiment (Et = ethyl group; C2H5) 0 50 100 150 200 2720 2760 2800 2840 2880 Intensity [a.u.] Raman shift [cm 1 ] ( a) ( b)A B C 0 50 100 150 200 2720 2760 2800 2840 2880 Intensity [a.u.] Raman shift [cm 1 ] ( a) ( b)A B C

PAGE 145

145 CHAPTER 7 EXPERIMENTAL AND COM PUTATIONAL STUDIES O F THE HOMOGENEOUS THERMAL DECOMPOSITIO N OF THE TUNGSTEN DIMETHYLHYDRAZIDO COMPLEX CL4(CH3CN)W(NNME2) DURING DEPOSITION OF WNXCY THIN FILMS Overview Continual reduction of the average feature size found in integrated circuits and the concomitant changes in via and line resistivity, current density, and RC (Resistive Capacitive) time delays have driven the gradual replacement of Al based metallization to that based on Cu given its lower electrical resistivity and higher resistance to electromigration. Use of Cu metallization requires prevention of Cu migration into the underlying Si and SiO2, which can produce an increase in contact resistance, leaky pn junctions, variations in the barrier height, and contact layerembrittlement [130, 131] To avoid CuSi interaction an amorphous diffusion barrier layer at a thickness well below the via dimension is conformally deposited at low temperature. As these requirements become more demanding, a chemical deposition method will li kely be required. WNxCy films grown by both chemical vapor deposition (CVD) [132, 133] and atomic layer deposition [134] have shown promise as a suitable diffusion barrier material. Tungstenbased barriers have demonstrated improved adhesion to Cu and ease of chemical mechanical planarization as compared to the commonly used TaNx. It was recently reported the preparation of the diorganohydrazido(2) tungsten complexes Cl4(CH3CN)W(NNR2) (R2 = Me2, Ph2, and (CH2)5 ) and Cl4(pyridine)W(NNPh2) [135] as precursors for the CVD of WNxCy films for use as Cu diffusion barriers [10, 136] To probe the mechanism of precursor decomposition under deposition conditions, the gas phase decomposition kinetics of thehydrazido complex

PAGE 146

146 Cl4(CH3CN)W(NNMe2) ( 1a) were investigated using in situ Raman spectroscopy in a vertical upflow, cold wall CVD reactor. Additional experiments involved NMR kinetics and observation of the dissociation of 1a under ion cyclotron resonance conditions. Comparison of the experimental results with literature data and DFT calculations was used for assignment of the observed Raman bands and evaluation of likely decomposition pathways. Experimental and Theoretical Methods The homogeneous thermal decomposition of the tungsten dimethylhydrazido complex 1 a and its benzonitrile derivative Cl4(PhCN)W(NNMe2) (1b ) was studied in a custom d esigned CVD reactor ( Fig. 7 1). The reactor is interfaced with an in situ Raman spectrometer (Ramanor U 1000, Jobin Yvon), which uses the 532.08 nm line of Nd:YAG solid state laser as the light source and includes a double additive monochromator fitted wit h a diffraction grating of 1800 groo ves/mm The Raman system was calibrated against the 546.07 nm emission line of mercury. As described elsewhere [119] this up flow, impinging jet CVD reactor was designed to produce a stable 2D flow pattern while isolating the reactantsfrom thereactor walls to prevent parasitic reactions. Since the entire CVD reactor chamber assembly could be translated in the x y z directions, it was possible to measure gas phase composition and temperature profiles inside the reactor to quantitatively study the gas phase decomposition kinetics of complex 1a. The three concentric inlet tubes shown in Fig. 7 1 are packed with 3 mm glass beads to supply an equal veloci ty and uniform flow inlet boundary condition. Aerosol assisted CVD (AACVD) was carried out for the Raman experiment with 1 a because this complex has low volatility [10] Solid 1 a was dissolved in benzonitrile (PhCN, 0.0174 mol/L), which generates a mixture of 1a and 1b in

PAGE 147

147 solution. The solution is then pumped into a nebulizer from a syringe. A piezoelectric material in the nebulizer vibrates at a frequency of 1.44 MHz, which generates an aerosol of precursor 1 a/1b (hereafter referred to as 1a ")and benzonitrile. The aerosol is introduced into the reactor with N2 (99.999 %, Airgas) carrier at a flow rate of 2.5 cm/s, which transports the mixture of 1a and solvent that was injected at a rate of 1.0 mL/h. For the concentric annular and sweep flows, pure N2 gas was delivered with the same flow velocity (2.5 cm/s) and sufficient gas flow time was allowed for the reactor to reach steady state before measurements were made Based on a previous study [91] that validated a steady state, two dimensional mass transport model and included CH4 tracer experiments, disruptive recirculation flow patterns in the reactor are not anticipated. A 3W Nd:YAG solid state laser line (532.08nm) was used to excite the mixture of 1a, benzonitrile, while the N2 vibrational Raman excitation lines were recorded along the centerline to estimate the temperature profile In addition to bands for gas phase benzonitrile, several other bands were observed. To assist in the assignment of these additional Raman bands as well as to assess possible reaction pathways, DFT calculations were performed using the Gaussian 03 program suite [73] For all bond dissociation, recombination, activation energy, thermodynamic properties, and frequency calculations, the spin polarized hybrid density functional B3LYP was combined with the LanL2DZ basis set for all elements. Crystallographic structure data [135] of complex 1 a collected at 173 K were used to obtain the starting geometry for calculations. NMR kinetic studies of the dissociation of acetonitrile from 1a were conducted on a Varian Inova at 500 MHz. Compound 1a along with a molar equivalent of acetonitrile

PAGE 148

148 was dissolved in toluened8. The 1H spectrum displayed signals for 1a at 0.53 ppm and acetonitril e at 0.83 ppm with a ratio of 1:2.09. Results and Discussion Kinetics of Acetonitrile Dissociation from 1a To estimate the rate of loss of acetonitrile from 1a in the CVD reactor the kinetics of acetonitrile exchange were determined via variable temperature NMR (Nuclear Magnetic Resonance) spectroscopy The exchange rate k was determined by line shape analysis in the temperature interval 50 to 84 C. A plot of ln(k/T) vs. 1/T ( Fig. 7 2) afforded an activation enthalpy of 23.0 0.2 kcal/mol and an activation entropy of 28.8 0.6 cal/mol K, consistent with dissociative exchange. The corresponding Gibbs free energy of activation is 14.4 kcal/mol, indicating thatdissociation of acetonitrile should be facile under CVD conditions. Based on these results it can be assumed that benzonitrile solutions of 1a have been converted almost completely to 1b by exchange of the acetonitrile ligand with solvent before the solution is introduced into the CVD reactor. Thus complex 1b can be co nsidered as the predominant precursor species during deposition of WNxCy from solutions made from 1a Ion Cyclotron Resonance Experiments It was previously demonstrated that mass spectrometry can provide insight into the gas phase dissociation chemistr y of the related tungsten imido complexes Cl4(CH3CN)W(NR) where R = iPr, Ph and allyl [132, 137, 138] as well as the series of tungsten hydrazido complexes Cl4(CH3CN)W(NNR2) [135] However, the chemical ionization mass spectra of 1a were anomalous in that the tungsten nitrido fragment Cl4WN+, wh ich is a marker for the ability of the imido complexes to deposit high quality

PAGE 149

149 films at low temperatures, could not be detected unambiguously in the spectra of 1a despite the fact that 1a is an excellent CVD precursor for WNxCy. To further investigate th e gas phase ion dissociation pathways of 1a and their relevance to the deposition conditions, an ion cyclotron resonance study was carried out. In situ Raman Experiments To investigate the thermal decomposition of hydrazido complex 1a, in situ Raman exper iments were performed using a susceptor set point temperature of 850 C. Fig. 7 3 shows the observed Raman bands along the centerline at seven distances below the heated susceptor (labeled ag in Fig. 7 3). Estimation of the local gas phase t emperature along the reactor centerline was made by analysis of theN2rotational state distribution with accuracy 20 to 30 C in the temperature range of interest [18] .The measured temperature is plotted in Fig. 7 4 as a function of reactor position and the value associated with each spectrum is listed in Fig. 7 3. To minimize the influence of blackbody radiation, Raman signals of the blank reactor were recorded and subtracted for each measurement position. For the results shown in panel A and D each spectrum was deconvol ved using GaussianLorentzian peak shape routine in the commercialsoftware packagePeakFIT ( v4.12 ) Measurem ents on neat benzonitrile provided frequency values for C Hin plane bending coupled with C CN stretching ( 1182 cm1)and the C Hin plane bending ( 1198 cm1) mode s in this region. Other modes detected in this region includeC Hin plane bending W N1 and W N3 stretching with C N C torsion (1193 cm1),C C N stretching for complex 1b (1178 cm1) and W 2 (1189 cm1) with the assignments based on DFT calculations. T he peak deconvolution

PAGE 150

150 results are s hown in Fig. 7 4A and concentration profileswere obtained as shown in Fig. 7 4B In Fig. 7 4B measurements were not possible closer than 3 mm from the heated susceptor in this range. Using the results shown in panel D ( Fig. 7 3), two C were detected. However, since this region also includes signal from the periodic vibration of excited N2, these two vibrations can be extracted from the spectra in panel D by comparison with the gas phase data from authentic benzonitrile. Peak deconvolut ion results show six components: four peaks for the N2 vibrational mode and two for C stretching as shown in Fig. 7 5 A. Examination of the spectra at four different reactor positions (ad) shown in Fig. 7 5 B reveal the four periodic N2 vibrational peaks ( spectrum a) at 2220.5, 2229, 2237, and 2245 cm1. Two additional C detected at 223 4 and 2241 c m1 as shown in Fig. 7 5 B panels b, c, and d. These last two frequencies lie in the reported ranges for CN of benzonitrile as obs erved in gas phase Raman experiments [139141] and the origin of the second band under our experimental conditions is not yet understood. To probe for additional Raman active species that were notdetected in the gas phase experiments, supplementary liquid phase Raman experiments for title compound d issolved in benzonitrile were performed. Using a standard liquid chamber, t he 532.08 nm Nd:YAG solid state laser line at 0.1 W w as used to excite neat benzonitrile and the vibrational Raman excitation lines were recorded in the range 100~4000 cm1. Soluti on phase Raman spectra of 1a (0.0174 mol/L in benzonitrile) and pure benzonitrile were measured and spectra for three selected wave number ranges are shown in Fig. 7 6 It is evident that the higher liquid phase density affords strong scattering intensity to produce

PAGE 151

151 very good signal to noise (S/N) ratio. It is also evident that spectra for the liquid samples show features not be detected in the gas phase experiments. Specifically, the three spectra show peaks at 361 cm1, associated with the W Cl4 vibration al mode, 1133 cm1, assigned to the C N C asymmetric vibrational mode, and 1396 cm1 assigned to the symmetrical umbrella mode of two terminal CH3 groups. Although these Raman shifts were not detected in the gas phase experiments, more detailed experiment s are in progress. DFT C alculations To better understand the experimental results and model some of the possible mechanistic steps in the thermal decomposition of 1a DFT calculations were performed. The initial calculations were a geometry optimization ( B3LYP/LanL2DZ ) using experimentally determined bond lengths and angles for 1a [135] as the starting point as a test of the computational method. Although the results of the NMR study indicate that 1a will have been converted to 1b before it is injected into the reactor, only the substituent of the nitrile ligand is different and the crit ical bond lengths and bond angles of 1a will carry over to 1b The computationally optimized geometry of 1a is summarized in Table 7 1 along with experimental data [135] and previous calculationsusing a split valence basisset [142] For most of the calculated v alues in Table 7 1 the B3LYP/LanL2DZ geometry optimization more closely approximates the experimental solid state structure Since the dissociation of acetonitrile from 1a was observed to be facile near room temperature, further calculations assumed that loss of the nitrile was a rapid first step in decomposition of 1a in the reactor. Further calculations on possible intermediates thus began with Cl4W(NNMe2) ( 2 ), the product of acetonitrile loss from 1a The experimental

PAGE 152

152 observation of dimethylamine and methylmethyleneimine among the decomposition products suggest two possible dissociative reactions of 2 as depicted in Fig. 7 8 B oth dissociation reactions were considered since AACVD from 1a involves high deposition temperature. Values for the enthalpy and Gibbs energy of W N1 and N1 N2 dissociation (Table 7 2 ) were obtained using statistical thermodynamics. Homolysis of W N1 affords open shell products and the spin states were set accordingly during calculations. In the case of N1 N2 homolysis, the products are doublets. The calculated W N1 bond dissociation enthalpy is significantly higher than the N1 N2 dissociation energy over the 298 to 900 K temperature range. These bond strengths can be viewed in terms of the limiting resonance structures A and B ( Fig. 7 9 ). Crystallographic data are consistent with representation B being the major contributor, which is in accord with calculated N1 N2 BDE (Bond Dissociat ion Energy) values that are higher than the experimental value for free 1,1dimethylhydrazine (49.6 kcal/mol) [143] The dissociation of benzonitrile to form Cl4W(NNMe2) (2 ) (Fig. 7 8 ) is postulated to be the first step in thermal decomposition of 1b on the basis of the nitrile dissociation rates for conversion of 1 a to 1 b The assumed parallel between the reactivity of 1a and Cl4(CH3CN)W(NiPr) is consistent with computational asse ssment of the bond strengthbetween Wand N3 using the Wiberg bond index in the natural bond orbital (NBO) analysis of complex 1a As shown in Fig. 7 10 the coordinate covalent bond of the acetonitrile ligand with W has the weakest bond order (0.3310). Dim ethylaminyl radical ( 3 ) [144 146] and 1,1 dimethyldiazene( 5 ) have been previously generated by other methods and their reactions have been reported.

PAGE 153

153 Dimerization of 1,1alkyldiazen es is known to form tetrazene derivatives, which in the case of 5 would afford tetramethyl 2 tetrazene ( 7 ). Since decomposition of 7 affords two equivalents of dimethylaminyl radical 3 ( Fig. 7 12), the critical reactive intermediate from cleavage of either the W N1 or N1 N2 bond of 2 ( Fig. 7 8 ) will be 3 Known reactions of dimethylaminyl radical ( disproportionation, dimerization, dissociation and rearrangement) are summarized in Fig. 7 13. Two dimethylaminyl radicals can yield methylmethyleneimine ( 8 ) and dimethylamine ( 9 ) [147149] or tetramethylhydrazine ( 7 ) by disproportionation or recombination, respectively [145, 150] ( Fig. 7 13 a). Since bimolecular reactions of radicals have low activation en ergies, products of recombination and disproportion are expected to be formed in similar quantities [151] In addition to these reactions, dimethylaminyl radical undergoes dissociation into methylnitrene and methyl radical ( Fig. 7 13b). Subsequent hydrogen shift in methylnitrene affords methyleneimine [152154] ( Fig. 7 13c) Comparison of Experimental Results with DFT calculations The analysis of this study assumes that the decomposition reactions that occur in the gas phase along the probed centerline are for inlet species 1b benzonit ri le, and acetonitrile, and that no surface or particle formation reactions occur. The aerosol assisted delivery of the low volatility precursor complicates the analysis of the experimental results since the precurso r is initially in a liquid solution but quickly evolves to a gas mixture due to the high vapor pressure of the solvent. Furthermore, the precursor is a metastable gas species and a large driving force exists for particle formation. This sequence is depict ed in Fig. 7 1 4 An analysis was performed to answer the question whether the aerosol is volatilized before decomposition occurs.

PAGE 154

154 To better describe the behavior of intermediates in the CVD reactor, the calculations should include the polarizable continuum model (PCM) to model solvation, as well as gas phase calculations if there exists liquid phase in the reactor since simple gas phase simulation is often not a good predictor of condensed phase behavior. To investigate the homogeneity of the introduced precursor, further study was performed. The initial droplet size produced by the nebulizer was first estimated using Langs correlation [155] This model assumes each droplet generated by the ultrasonic nebulizer is a homogeneous sphere. The droplet diameter estimated for the system modeling the evaporation process. The Nusselt number, Nu, is the ratio of convective to conductive heat transfer across the liquidgas interface, and for a liquid droplet in a flowing gas, this is empirically given by the relation 0.50.332.00.6RePr Nu (7 1) where Re is the Reynolds number and Pr is the Prandtl number. Using this equation, the heat transfer coefficient, ch is estimated as: 0.33 0.52.00.6p c ddC hD D (7 2) where D and d are the droplet diameter and liquid thermal conductivity (0.1317 W/mK at 366.5 K) In addition, , and pC denote mean fluid velocity, density, viscosity, and specific heat capacity at constant pressure, respectively. With property values

PAGE 155

155 mentioned above, Re and Pr have the value of 5.45 105 and 3.8 1 respectively. From Eq. 7 2, the heat transfer coefficient is calculated as 9.79 104 J/sm2K A model that describes the coupled heat and mass transfer of an evaporating droplet in a laminar carrier gas stream was next examined. Briefly, the energy balance equation for a benzonitrile droplet gives: c chAT N (7 3) where cN ch A T and denote the solvent evaporation rate, heat transfer coefficient across the liquidgas (9.79 104 J/sm2K ) ,total droplet surface area introduced to the reactor as 6.18 104 m2 temperature difference between droplet surfaceand ambient, and latent heat of vaporization as 3.67 102 J/g, respectively. T he value of the t emperature difference was assumed to be 10 K since the drop sizeis fairly small. If one assumes a larger temperature difference since evaporation process lowers the surface temperature, the solvent evaporation rate would be increased.Eq. 7 3 gives the value of the solvent evaporation rate as 1.65 g/s and applying this model to these reaction conditions gives a drop evaporation time of only 17 ms, which at an average 2.5 cm/s velocity yields a distance at 4.25 104 m. In other words, a drop introduced to the reactor evaporates in 17 ms and only travels 4.25 104 m The droplet diameter is predicted to show a normal distribution with standard deviation, as follows [156] :

PAGE 156

156 0.1 5 0.410.500.184.610 f (7 4) where f is the nebulizer frequency [Hz], is the surface tension [N/m] and is the dynamic viscosity as 1.25 103 Pa diameter, normal distribution curve, which is deviated from the mean value as much as twice the standard deviation, is only 2.3 % among total solvent and statistically it is produced only 1.8 particles per one reactor volume. Al though it can reside 2.0 103 m in the reactor, it can be highly possible to evaporate in the center tube during approaching the inlet. There is experimental evidence to support this conclusion. Frolov et al. [156] tetradecane droplets vaporize in 70 ms at a liquid surface temperature of 293.15 K and a gas temperature of 573.15 K. It is noted that n tetradecane has higher boiling temperature (526 K) than benzonitrile (464 K) and their droplet diameter was approximately 4 times larger than that in this study. As further evidence, the spectral positions of the strongest phenyl ring breathing mode of benzonitrile were carefully measured. No displacement of peak positions along the reactor centerline was detected as shown in Fig. 7 1 5 Based on the modeling results, comparison to experiment using a less volatile solvent, and measurement of the lack of peak shift in the phenyl ring breathing mode, the assumption of homogeneous gas phase reaction in the whole CVD reactor can be substantiated. The calculated vibrational frequencies can be correlated to products postulated in the proposed decomposition mechanisms ( Fig. 7 8 and Fig. 7 13a ). Calculated and

PAGE 157

157 corrected Raman active stretching frequencies with experimental values for all compounds including the gas phase authentic sample of benzonitrile ar e listed in Table 7 3 Assignments for 9 of the observed bands attributed to benzonitrile (462, 762, 1003, 1181, 1197, 223 4 3077, 3128, 3175 cm1) were derived from gas phase benzonitrile Raman spectra and deconvolution data as well as DFT calculations The reported Raman bands of 1240 and 3377 cm1 of gas phase dimethylamine [157] and DFT results were used for assigning 1247 and 3377 cm1 to t he CH3 rocking and N H stretching motions in dimethylamine, respectively. In addition, 1638 cm1 was assigned to C=N stretching of methylmethyleneimine ( 8 ). This assignment is supported by IR data [150] and the Ra man characteristic group frequency [158] Given that dimethylamine ( 9 ) and imine 8 would ultimately result from homolysis of the N1 N2 bond followed by disproportionation of radical 3 observation of these products is consistent with N1 N2 cleavage. The possibi lity of W N1 dissociation, however, cannot be excluded since this reaction can produce the same radical intermediate ( 3 ) through tetramethyltetrazene formation ( Fig. 7 13). The peak at 3426 cm1 is tentatively assigned to the N H stretching mode of HNWCl4 ( 14) based on the predicted gas phase reaction chemistry and the experimentally observed N H vibrational frequencies of W(VI) model compounds with imido (NH) and amido (NH2) ligands [159] The N N cleavage reaction ( Fig. 7 8 ) would produce the nitrogencentered radical NWCl4 ( 4 )as the inorganic product. Subsequent abstraction of a hydrogen atom by 4 would afford the W(VI) parent imido c omplex 14. The NspH bond of 14 would be expected to have a high bond dissociation energy,

PAGE 158

158 making hydrogen abstraction from several species in the reactor energetically favorable[160] Concluding Remarks In the present work, using Raman spectroscopy, NMR kinetics, ion cyclotron resonance and DFT calculations at the B3LYP/LanL2DZ level of theory, the decomposition pathway of the dimethylhydrazido tungsten complex (RCN)Cl4W(NNMe2) ( 1 a ) was investigated. In a custom upflow, cold wall aerosol assisted CVD reactor, identifiable Raman shifts for the starting material and intermediates were detected and these frequencies were assigned using gas phase DFT calculations and literature data. The observed complex Raman spectra were deconvolved using commercial software with high accuracy. Analytical and spectroscopic evidences provided homogeneousness of introduced precursor by performing droplet evaporation analysis. The bond cleavages of both W N1 and N1N2 are possible, although the calculated bond strength of W N1 is larger than that of N1N2. Peaks consistent with methylmethyleneimine ( 8 ), dimethylamine ( 9 ), and HNWCl4 ( 14) were observed in the Raman experiments, consistent with proposed decomposition pathways for 1a The methods used in this study are complementary to each other and seems very promising for kinetics study, thermal decomposition of precursors in chemical and engineering perspectives.

PAGE 159

159 Figure 71. Schematic of the nebulizer assisted experimental reactor for in situ Raman spec troscopic measurements. Vent Vent Ceramic cap Heating element Quartz wall Stainless screen Quartz balls Sweep inlet Annular inlet Center flow Annular flow Sweep flow LASER (532.08 nm) Carrier gas M.O. source Nebulizer r z (N2) (N2) (N2) (1a/benzonitrile) Vent Vent Ceramic cap Heating element Quartz wall Stainless screen Quartz balls Sweep inlet Annular inlet Center flow Annular flow Sweep flow LASER (532.08 nm) Carrier gas M.O. source Nebulizer r z (N2) (N2) (N2) (1a/benzonitrile)

PAGE 160

160 Figur e 7 2 Plot of ln(k/T) vs. 1/T(K) for acetonitrile exchange in complex 1a.

PAGE 161

161 Figure 73. Raman spectra of gas phase dimethylhydrazido complex 1 a in benzonitrile as a function of distance below the heated susceptor along the reactor centerline Spectra in panel E (nitrogen vibrational mode) are marked with reduced scale for concentration reference. Raman shift values for the primary modes are spe cified at the top of each panel. Raman spectra profiles are marked as a function of position from the heated susceptor surface as (a) 1 mm, (b) 3 mm (c) 4 mm, (d) 5 mm, (e) 7 mm, (f) 9 mm and (g) 11.5 mm

PAGE 162

162 Figure 74. Relative species concentrat ion and gas phase temperature profiles along the reactor centerline for benzonitrile (BN), 1 b and 2 using results in panel A, Fig. 7 3.

PAGE 163

163 Figure 75 Peak deconvolution results: (A) results shown in panel D of Fig. 73 N2 vibrational bands (spectrum a) and (B) peaks associated with N2 vibrational modes and with C d).

PAGE 164

164 Figure 7 6 Liquid Raman spectra of 1a in benzonitrile solution (red) compared to pure liquid benzonitrile (black). Note that the red spectra are intensity shifted for clarity. Table 7 1 Experimental and computationally o ptimized b ond l engths () and b ond a ngles () for complex 1a Experime ntal [135] Split Basis Set [142] a This Work b W1 N1 1.769(5) 1.749 1.757 W1 N3 2.224(7) 2.268 2.220 W1 Cl 2.347(16) 2.388 2.429 N1 N2 1.271(8) 1.289 1.283 N2 C1 1.438(7) 1.467 1.478 N1 W1 Cl 95.9(4) 96.8 96.2 N3 W1 Cl 84.07(4) 83.24 83.75 N2 N1 W1 180.0(0) 178.0 179.6 aB3LYP/LanL2DZ for W, 6311G for other elements, bB3LYP/LanL2DZ for all elements

PAGE 165

165 Figure 77 Computationally optimized geometry of 1a. Figure 78 Products of cleavage of the W N1 and N1 N2 bonds of complex 2. Gibbs energy values ( G) are in kcal/mol.

PAGE 166

166 Table 7 2 Calculated b ond d issociation e nthalpy ( H ) and Gibbs e nergy change ( G) for the W N1 and N1N2 b onds BDE(W N1) (kcal/mol) a BDE(N1 N2) (kcal/mol) a o (298 K) 113. 1 88.0 o (900 K) 111. 4 86.9 o (298 K) 101.0 73. 2 o (900 K) 77. 8 43. 7 aB3LYP/LanL2DZ Figure 79 Limiting resonance structures of complexes 1 a and 1b Figure 7 10. Wiberg bond indices for 1a and 1b

PAGE 167

167 Figure 711. Gibbs energy change ( G) for initial ligand substitution and dissociation reactions of complex 1. All energy values are in kcal/mol at 298 K. Figure 712. Production of dimethylaminyl radical ( 3 ) after W N1 cleavage to form 5 Gibbs energy values are in kcal/mol.

PAGE 168

168 Figure 713. Known reactions of dimethylaminyl radicals and their calculated (B3LYP/LanL2DZ) reaction energies at 298 K. Gibbs energy values are in kcal/mol. Figure 7 14. Suggestedevolution of an atomized droplet as it approaches a heated susceptor

PAGE 169

169 Figure 7 1 5 Detected Raman spectrum profile assigned to phenyl breathing mode of benzonitrile (1003 cm1). Raman spectra are shifted for clarity. Black bottom signal is for right above inlet (cold region) and red top signal is for right below the heated susceptor 985 995 1005 1015 Raman shift [cm-1] Intensity [a.u.]

PAGE 170

170 Table 73. Calculated and c orrected Raman a ctive stretching f requencies Molecule Experimental Calculated Corrected Assignment BN 462 466 461 Ph ring stretching BN 762 769 760.5 Ph ring stretching BN 1003 1025 1004.5 Ph ring breathing 1b 1177 1226 1177 C H in plane bend BN 1181 1236 1186.6 C H in plane bend+ C CN vibration 2 1189 1246 1196 CH 3 scissoring + W 1b 1193 1243 1193 C Ph stretching + N3 W N1 stretching + C N C torsion BN 1197 1248 1198 C H in plane bend 9 1247 1279 1228 CH 3 rocking 8 1638 1696 1638 C=N stretching 1708 BN 2233.6 2334.5 2236.5 C 2241 C 2516 2547 2580 2834 CH 3 symmetric stretching ( N(CH 3 ) 2 ; aromatics) BN 3077 3142 3079 C H asymmetric stretching BN 3128 3192 3128.2 C H asymmetric stretching BN 3175 3238 3173 9 3377 NH stretching 14 3426 NH stretching

PAGE 171

171 CHAPTER 8 DESIGN SIMULATION AND SETUP OF ULTRA HIGH VACUUM SYSTEM FO R SURFACE SCIENCE AND ITS APPLICATION Overview The increasing importance of organic and metal organic films in semiconductor technology has led to research efforts which direct toward understanding and controlling hydrocarbon reactions with surfaces such as Si, Al2O3, Ge, etc. Although efforts to experimentally observe and analyze fundamental reaction mechanisms during thin film deposition of organic and met al organic materials on silicon surfaces have been made for decades, single technique could not have satisfied these needs. T o analyze initial step of film deposition, various techniques, for example, XPS (X ray photoelectron spectroscopy), AES (Auger elec tron spectroscopy), LEED (Low energy electron diffraction), TPD (TemperatureProgrammed Desorption) using QMS (Quadru pole mass spectroscopy), ATR FTIR (Attenuated total reflection Fourier transform infrared spectroscopy) have been used for this purposes. Understanding of first step of film deposition is very important in semiconductor field since initial step plays an important role for controll ing the film composition, quality, growth rate, and so on. Among surface analysis techniques mentioned above, ATR FTIR can be used for identifying the functional group on surface and semi quantifying the adsorbates and it would be more powerful when using in situ type in UHV (Ultra high vacuum) chamber. ATR FTIR technique which uses internal reflection element (IRE) allows the vibrational properties of the adsorbed species and this information can enhance understandings of bond between adso rbate and surface by using DFT calculations [161] Even though some molecular vibrations are not IR active, IR technique provides s trong signal compared to

PAGE 172

172 Raman spectroscopy for various vibrational modes, in other words, an extremely reliable and well recognized fingerprinting method. Characterization of S urface K inetics in UHV Temperature Programmed Desorption (TPD) The method of temperature programmed desorption (TPD) is one of the most effective and well known methods for the determination of bond energy in adsorption. And it has been used to study reactions at the gas metal interface [162] TPD offers a way to better understand the growth of metal organic films. To control chemical processes at the atomic level during the growth of a film, information about reactions on the substrate surface such as decomposition, adsorption, desorption, and migration of chemical species is important. Once absorbed, the surface of the sample is heated according to a known temperature ramp to desorb the bonded reactant [163] While flash filament desorption method records the pressure in the system to analyze information on various adsorption parameters, a technique called TPRS (temperature programmed reaction spectroscopy) is of ten used to record the identity and number of species desorbing off of the surface. The difficulty of the analysis of the desorption curves using TPRS relies on if the rate determining step is the desorption of the species or the decomposition of the spec ies on the surface. W hile the flash filament analysis can be applied to the desorption curves in the former case, the basic rate equation for TPD can be given in the latter case, (8 1) w here , , and is the preexponential factor, the surface coverage, the order of kinetics, the temperature at time and the activation energy, respectively. If expndE A dtRT A n T E t

PAGE 173

173 the peak temperature and heating ramp rate are known, it is possible to calculate the pre exponential factor Desorption curves for the surface will be produced by numerical integration of Eq. 8 1. Desorption rate curves for argon on tungsten with various incident ion energy are depicted in Fig. 8 1. What one can determine from the shape of the peaks on the desorption rate curve is whether the activation energy is constant or a function of the surface coverage. However, there should not be any overlapping peaks to determine the dependence of activation energy of surface coverage. The only property that ca n be calculated is the temperature at which the desorption rate is at a maximum since Fig. 8 1 has overlapping peaks. One can find out detailed analysis from Redhead s articles [163, 164] Recently, TPD depends on the use of a mass spectrometer, which can identify different species in the gas phase. More specifically, a mass spectrometer is used to detect the species coming off from the surface of the sample while it is being heated. In addition, mass spect rometry can be used to deduce fragmentation patterns of the precursors as they interact with the substrate surface. Fragmentation data can provide important information for bonding state and effect of the precursor to the surface. Low Electron Energy Diff raction (LEED) LEED (Low Electron Energy Diffraction) is a technique for investigating the crystallography of surfaces and overlayers or films adsorbed on surfaces. LEED uses a beam of low energy electrons, typically 101000 eV [165, 166] The electrons are incident on the surface of the sample and the resulting backscattered and secondary electrons are used to determine the position of the atoms on the surface of the sample. Diffraction of electrons occurs because of the periodic arrangement of atoms in the surface. This periodic arrangement can be conceptualized as parallel rows of atoms A

PAGE 174

174 analogous to grating lines in a diffraction grating. The simplest diffraction measurement is the determination of the surface or overlayer unit mesh size and shape if the electron beam is incident normal to the surface. This can be done by inspection of the diffraction pattern at any energy of the incident beam. The surface structure can be determined by a plot o f intensity versus energy or accelerating voltage, which is referred to as an I V curve. Diffraction is very useful whenever there is a distinct phase relationship between scattering units. The greater the order, the better defined are the diffraction features. Although LEED can be used for quantitative determination of atomic positions in surfaces, it is complicated by the multiple scattering of the electrons since an accurate goniometer and detection scheme are required to measure intensities for which the scattering geometry. LEED is the most powerful, widely used and developed technique for investigating periodic surface structures. It has been used very widely as a method to check surface order [165] Auger Electron Spectroscopy (AES) AES (Auger Electron Spectroscopy) is very similar to LEED in which it depends on the use of electrons to determine the nature of the surface. AES is the most commonly used surface, thin film, or interface compositional analysis technique. AES is used to determine the elemental composition of the surface, and in many cases, the chemical bonding of the atoms in the surface region of solid sample while LEED is used to check the surface structure. Among all the electrons in the beam incident on the surface o f the sample, Auger electrons have the least amount of energy. Due to the lack of energy, Auger electrons only penetrate the surface deep enough to analyze the outermost 2 through 10 atomic layers and this characteristic make AES most useful in identifying thin films grown on the substrate surface. It has very good lateral spatial

PAGE 175

175 resolution which can be as low as 300 relying on the electron gun used. In addition to good lateral spatial resolution, it also has very good depth resolution, as low as 20 depending on the characteristics of the ion beam used for sputtering [165] And it has a good absolute detectability of 100 ppm for most elements under good conditions. It can be combined with ionbeam sputtering to remove material from the surface and to continue to monitor the composition and chem istry of the remaining surface as this surface moves into the sample. Attenuated Total Reflection Using FTIR Spectroscopy (ATR FTIR) Current s tate of ATR FTIR technique Attenuated total reflection is a sampling technique used in conjunction with IR spectr oscopy which enables samples to be examined directly in the solid or liquid phase without further sample preparation[167] ATR uses a property of total internal reflection called the evanescent wave, the phenomenon which was observed by Newton in the early 1700s. Total internal reflection can occur when a ray of light passes from a denser to a rarer medium. In such environments, light bends away from the normal in to the rarer medium, however, from at a certain incidence angle which is called as critical angle, the light is completely reflected back to the denser medium. Fig. 8 2 shows a multiple reflection ATR system. As seen in Fig. 8 2 ATR crystal has some type of beveled edges and infrared beam go through it with one or more reflections. Fig. 8 3 shows various types of ATR FTIR system. For pur suit of diversity of sample phases, IRE has been developed for each specific sample type. Among these types, Fig. 8 3 ( A ) is the most typical one for UHV surface system.

PAGE 176

176 In view of number of reflections, the more, the better. However, there are some constr aints for that. The total number of reflections (N) of a given length (L) and width (W) of an ATR element is given by; (8 2 ) where is the critical incidence angle. As one can see in Eq. 8 2 long and thin crystal can give more reflection numbers, however, long and thin crystal might lose uniformity and have beam focusing problem, respectively. FTIR (Fourier Transform Infrared)spectroscopy Every molecule is in constant motion, producing vibrations like stretching, bending, rocking, wagging and scissoring. A molecule composed of natoms has 3n degrees of freedom, six of those are translations and rotations modes. Those 3n6 degrees of vibrational freedom make different modes. And it can be symmetric or asymmetric stretches and bends and bending mode can be either inplane or out of plane. If the frequency (energy) of incoming IR radiation matches that of the molecular vibration, energy is absorbed by the vibrating bonding. Since every vibrational mode has its own fingerprint region, one can assign each signal to specific vibrational mode, even though some are overlapped. In other words, C O, C S, C H, N H, S O, etc. demonstrate their unique vibration bands in the infrared spectrum. And wavenumber for each mode can be red or blue shifte d depending on its neighboring group. The report from Mui and his coworkers shows good conjunction of ATR FTIR experiment and detailed calculation result for vibrational frequencies of adsorbed molecules including bonding energy calculations. They used A TR FTIR technique and DFT calculations to cotcL N W c

PAGE 177

177 test the applicability of the surface proton affinity using methylamine, dimethylamine, and trimethylamine at the Si(100) 21 and Ge(100) 21 surfaces [168] Applications of ATR FTIR t echnique t o t hin f ilm r esearch As interest in the surface chemistry of group IV semiconductors such as silicon and germanium increases rapidly, technological applications as well as fundamental research is getting more important. Energetic research products for various types of surfaces, for example, flat, porous, and reactive surfaces, have been performed with surface analysis tools. The most widely used singlecrystal silicon wafers of high purity are commercially available and relatively cheap. Si(100) and Si(111) is the most common surface orientations, it becomes coated with a native oxide upon air exposure. For flat silicon surf ace, native oxide can be removed by fluoride ion chemically or thermally under UHV environments. Moreover depending on desired electronic properties, silicon wafers can be doped as ntype or ptype with electrondonors (P, As, Sb) or electronwithdrawer ( B), respectively. Compared with silicon wafers, germanium wafers is approximately 500 times more expensive due to its limited use. Disordered and water soluble GeO2 interface is one of the main obstacles for applications. Ranke reported NH3 adsorption on three orientations of Ge surfaces, Ge(001), (113), and (111) [169] Due to disordered interface, initial adsorption on Ga(111) which does not form reconstruction dimers is much weaker. And even beyond two monolayers, the layer continues to grow with different adsorption enthalpy on the different orientations and only after 4 monolayers, the whole layer rearrange irreversibly to form the usual NH3. ATR FTIR technique has been used widely, especially in surface functionalization research. For analyzi ng surface flatness and makeup of the hydride-

PAGE 178

178 terminated surfaces that offer many advantages, ATR FTIR has been put to use. Rapid and efficient preparation of Si H hydrideterminated flat surfaces has been known and it is outlined in Fig. 8 4 Since the S i(111) H monohydride surface has a very sharp and narrow Si H stretch at 2083.7cm1 as shown in Fig. 8 5 one can confirm smooth surface over nanometer scale distances from ATR FTIR results. Also for preparing GeH surface which was first published only recently in 2000[ 170] the roughness of this surface is related with the broad GeH stretching observed by ATR FTIR at around 2100cm1 with ~50cm1line width As shown in Fig. 8 5, germanium surface was smoothened by etching with a 10% HF solution with time evolution. And Choi and Buriak [170] reported that the H terminated Ge is stable in air for up to 1 h our Other than H terminated silicon and germanium surface researches, chlorinated silicon (SiCl) [171] brominated silicon (Si Br), [172] halogenated silicon (Si X, X=Cl, Br, I), chlorinated germanium (GeCl) [173] etc. have been studied using ATR FTIR for surface functionalization. UHV approach to Si C bond formation is very important for metal organic precursor initialization step with surface. Working under UHV environment (below 1010 Torr) provides good conditions that are as close as at the atomic level. Among various possibilities, [2+2] reactions of carbon containing unsaturated molecules with silicon surface and diels alder like ([4+2]) reactions of dienes with silicon surface have been paid attention and studied over twenty years. Bozack and his coworkers started research about small unsaturated hydrocarbons such as propylene to chemisorbed to Si(100) surfaces at room temperature[174] From ATR FTIR results, two new Si

PAGE 179

179 emaining the Si Fig. 8 7 In addition, the reactivity of cis and trans 1,2 dideuterioethylene by ATR FTIR was examined and compared the stretching frequencies of the observed vibrations with calculated values with Gaussian 94. In this case due to the lack of data for ethylene adsorption to silicon surface with different stereochemistry, DFT calculations can provide definite clues for identifying peaks. Other than the vibrational frequency calculations, various thermodynamic calculations such like adsorption/desorption energy, activation energy, reaction enthalpy, etc. can be performed with satisfactory level [168] Filler and his coworkers adopted ATR FTIR, TPD, and DFT calculations for elucidation of carbon and oxygen coupling in the reaction of formaldehyde on Ge(100) surface [175] In their work, evidence that a species with an electrondeficient carbonyl group, likely a dative bond, initially exists near room temperature with a minority C H dissociation product was elucidated by infrared sp ectroscopic method. And using DFT calculations, formation of a C=O [2+2] cycloaddition product was predicted. Although Konecny predicted that the [4+2] reaction will be 1529 kcal/mol more stable than the [2+2] addition product theoretically [176] this work identified the first catalytic reaction on a group IV semiconductor surface both experimentally and theoretically. In the case of adsorption of more complex molecu les, experimental and theoretical study for mechanisms and processes in hafnium alkylamine precursor adsorption and reaction on silicon oxide and H terminated silicon surfaces has carried out by Kelly and his coworkers [177] From the UHV experiment of TDEAHf (Tetrakis (diethylamino) hafnium) exposure to Si H and Si OH surfaces, they reported conversion

PAGE 180

180 of the higher hydrides to monohydride (Si H) units on the surface through hydrogen abstraction, or change in one or more of possible back bonds made by the local silicon atom. The structures of the Si H IR vibrational modes for the initial surface and after various exposure doses at 25 and 250C are shown in Fig. 8 8 Comparison the spectra in Fig. 8 8 indicates a broad Si H mode, consistent with surface roughness and a variety of Si H, SiH2, and Si H3 surface species and the loss of these converted (or frequency shifted) Si H bonds. By observing ATR FTIR results and matching with DFT calculation results for each reaction step as shown in Fig. 8 9 they proposed adsorption mechanism of TDEAHf on functionalized silicon surface. In Fig. 8 9 ( A ), one can confirm Si H and O H stretching modes of the initial surface at 2207 and 3880cm1 from the first spectrum and O H modes disappear after reaction with TDMAHf via H transfer from OH group. While Si H stretch remains, other modes of the adsorbed precursor appear including the N H stretch of the adsorbed product at 3512cm1. In the same context, after the reaction with TDMAHf via H abstraction, the Si H mode which was shown in the first spectrum (surface group) disappears while the other modes of the adsorbed precursor appear. For another H terminated silicon experiment, Kosuri and his coworkers reported vapor phase adsorption kinetics of 1decene on Si(100) surface[178] Adsorbed 1decene can be detected with ATR FTIR owing to an evanescent electric field, which gathers signal of distant functional group from surface approximately 0.5~5 m. And this distance is enough to analyze thin film surface. Proposing reaction mechanism of 1decene on hydrogenated silicon, they obtained preexponential factor and activation energy with regression of Arrhenius plot experimentally. Reasonable initial and

PAGE 181

181 boundary conditions of coverage were assumed and those were used to solve the rate equations. A time series of infrared absorbance spectra are taken during the Si(100) surface exposure to 1decene as shown in Fig. 8 10. And interesting and outstanding results were reported by Kim and his coworkers. They demonstrated the layer by layer growth via urea coupling reaction to form an ultrathin film on Ge(100) surface between two bifunctional molecules, ethylenediamine and 1,4phynylene diisocyanate at room temperature under vac uum environment [179] Since multilayer functionalization under vacuum condition is not common and this developed dry met hod for multiple layer organic functionalization, this research is more outstanding. Moreover, since gas phase reactants should undergo spontaneous reaction with the terminal functional groups at the surface under vacuum based condition, this reaction should meet severe requirement. By reacting two molecules sequentially, layer by layer growth mechanism was discovered as shown in Fig. 8 11. In addition to that, they observed and reported characteristic infrared peak areas plotted as a function of exposur e. Since one carbonyl group is created per molecule of ED or PD adsorbed beyond the fist layer, the peak areas of each layer are expected to be equal and that was confirmed. Their demonstration of these layer by layer deposition methods could provide door to a strategy to design and produce precisely tailored organic materials. Among various surface analysis tools, the most widely used machine would be XPS and QMS. Due to the difficulties of operating and designing barrier, ATR FTIR is not that common s o far. However, ATR FTIR can give powerful information for surface functional group and surface analysis could be incomplete without it. Cho and his

PAGE 182

182 coworkers have reported thermal decomposition of Ti(O iPr)2(dpm)2 on Pt foil and they used QMS, XPS and A ES[180] They proposed thermal decomposition mechanism with sufficient data of TPD and binding energy as shown in Fig. 8 12 and final species remaining on the Pt surfaces after the decomposition of the Ti compound at temperature above 650K with XPS data. However, they failed to explain the bond between the adsorbate and Pt surface and functional group variation of intermediates. Although ATR FTIR can be used all phases (i.e. solid, liquid and gas) even for the interface[181] this review has focused on gas phase surface rea ction under vacuum. ATR FTIR would have unlimited applications for surface science area owing to its fingerprint capability and high sensitivity. And its utilization will make surface science field more resourceful and exciting one. The unlimited technological applications of ATR FTIR are only a matter of time. One should note its potential. UHV System Design and Sample Probe Simulation Using CFD Software UHV System Design Fig. 8 13 shows UHV system design and process flow diagram interfaced with QMS, LEED/AES and FTIR spectroscopy. This in situ system enables to analyze surface science kinetics under ultra high vacuum atmosphere. Two bubblers that typically contain metalorganic precursor are immersed in the constant temperature bath to provide sufficient vapor pressure. UHV system designed in this work consists of four chambers. Chamber 1 is loadlock to mount the sample on the sample probe and enables to save time to make chamber 2 ultra high vacuum level. Since chamber 1 is small, one can easily lower the chamber pressure to high vacuum level. Chamber 2 is deposition chamber. T wo metalorganic precursors are used to deposit films on beveled edge silicon wafer. And then deposited film is conveyed to chamber 3 to analyze

PAGE 183

183 spectroscopic characteristics. Chamber 3 is for ATR FTIR chamber and deposited film is analyzed using ATR technique. To conserve IR beam source two KBr windows are adopted. Chamber 4 is analysis chamber which will perform main analysis using QMS and LEED/AES including TPD option. Fig. 8 14 depicts perspective view of three chambers of four. For effective movement of sample probe, four chambers are connected linearly. At each chamber, sample can be rotat e in one direction completely to face or align appropriately. Sample Probe Simulation Using CFD Software The most demanding part to design is sample probe. Basically, mass and momentum transfers are not related to this simulation, conservation of energy w as considered with geometry and boundary conditions specific to this sample probe geometry. Since heating and cooling of sample should be done during experiment and temperature distribution is very important for TPD experiment, temperature uniformity would be the key to success or failure. Temperature simulation was performed using F LUENTTM Computational Fluid Dynamics (CFD) commercial software. Heating filament, heater tube and front and rear clamps are made of tungsten, Oxygen Free High Conductivity ("OF HC" hereafter) copper, and 304SS (stainless steel 304), respectively. The thermal properties used in CFD simulations are tabulated in Table 81. After various trials of sample probe design, improved model was proposed and further CFD simulations were per formed to investigate temperature profile mainly focused on Si wafer at the heating material temperature 1000 K as shown in Fig. 8 15. Fig. 8 15 (A) shows perspective view with IR pathway and cross section for simulation.

PAGE 184

184 However, since temperature differe nce 1T2) between Xand Yshown in Fig. 8 15B and C was around 85 K, it requires another modification to increase temperature uniformity. The large temperature difference between both ends can give improper information about desorption temperature d uring TPD experiments. Based on this experience, a few modified designs were simulated as shown in Fig. 8 16. With the simulated results with 1000 K heating material, temperature difference between both ends of beveled Si wafer is tabulated in Table 82 The mesh was generated using the GambitTM software and the unstructured quadrangle grid was employed. After initializing the values with boundary condition, the iteration process converged after 228 iterations. Fig. 8 16E shows the most uniform temperat ure distribution and threedimensional model was prepared using CAD (Computer Aided Design) software as shown in Fig. 8 18. With all these simulation and analysis, prototype of UHV system is schematized as shown in Fig. 8 19.

PAGE 185

185 Figure 81. Desorption rate of argon from tungsten for various values of the incident ion energy. (taken from Ref. [163] ) Figure 8 2 A m ultiple reflection ATR system (F igure was takenfrom Ref. [182] )

PAGE 186

186 Figure 8 3 Various types of ATR FTIR system Figure 8 4 Fl u oride based etching conditions, leading to hydrideterminated flat and porous silicon surfaces. (Figure was taken from Ref. [183 ] ) C B A

PAGE 187

187 Figure 8 5 Internal reflection spectra of HF treated Si(111) surfaces (Figure was taken from Ref. [184] ) Figure 8 6 Ge H vibrations observed by ATR FTIR. Surfaces prepared by etching 10% HF solution. (Figure was taken from Ref. [170] )

PAGE 188

188 Figure 8 7 (a) [2+2] cycloadditions on Si(100) surface with alkenes and alkynes (b) possible models for adsorption of ethylene on Si(100) surface (c) FTIR spectra for cis and trans 1,2 dideuterioethylene. ( Fig. 8 7(A) was taken from Ref. [183] and ( B ) and ( D ) were from Ref. [175] ) Figure 8 8 Si H IR vibrational modes for the initial surface and after various precursor exposure doses of TDEAHf at 25 and 250oC. (Figure was taken from R ef.[177] ) C B A

PAGE 189

189 Figure 89 ( A ) Calculated vibrational spectra for the species present in the H transfer process ( B ) Calculated vibrational spectra for surface species in the Si H abstraction process. (Figure was taken from Ref. [177] ) B A

PAGE 190

190 Figure 8 10. A time series of IR absorbance spectra of c H and Si H stretching vibrational modes. (Figure was taken from Ref. [178] ) Figure 8 11. (A ) Saturation infrared spectra of ED/Ge(100), PD/ED/Ge(100), ED/PD/ED/Ge(100), and PD/ED/PD/Ge(100) for 1st through 4th layer. ( B ) Schematic illustration of PD adsorbed on Ge(100) and second layer reaction with subsequent exposure to ED (Figure was taken from Ref. [ 178] ) B A

PAGE 191

191 Figure 8 12. Proposed decomposition mechanism for Ti(O iPr)2(dpm)2 on a Pt surface. (Figure was taken from Ref. [179] ) Figure 813. Four chamber UHV system and process flow diagram LEED/AES QMS IR detector IR source Ion gun Door Reactant 1 Reactant 2 Carrier N2 Vent MO1 MO2 LEED/AES QMS IR detector IR source Ion gun Door Reactant 1 Reactant 2 Carrier N2 Vent MO1 MO2

PAGE 192

192 Figure 814. Perspective view of CAD drawing for (A) deposition chamber, (B) ATR FTIR chamber and (C) analysis chamber Table 81. The thermal properties used in CFD simulation. Si W OFHC 304SS (Stainless steel) Density [kg/m 3 ] 2329 19300 8978 8030 Cp [J/kg K] 19.789 134.4 381 502.48 Thermal conductivity [W/m K] 149 168 387.6 16.27 A A B B C C

PAGE 193

193 Figure 815. Proposed design of sample probe (A) perspective view (B) CFD simulation profile and (C) beveled edge Si wafer. A

PAGE 194

194 Figure 816. Temperature profile simulations at 1000 K for the modified designs with (A) 6 unit elongated front heating element, (B) 2 unit retracted liquid nitrogen reservoir from design A, (C) 4 unit retracted liquid nitrogen reservoir, (D) 6 unit retracted reservoir, (E) 6 unit elongated heater from design D, (F) 8 unit retracted reservoir, and (G) 6 unit retracted reservoir from Fig. 815 (B). (1 unit = 1.67 mm in real scale)

PAGE 195

195 Figure 817. Quadrangular grid mesh for Fig. 816E Table 82. The temperature difference between both ends of Si wafer according to each design Design No. 1 2 3 4 5 6 7 200 40 33 30 50 22 25

PAGE 196

196 Figure 8 18. CAD drawing of sample probe (Fig. 8 16 (E)). (A) perspective, (B) front, (C) side, and (D) top view A A B B C C D D

PAGE 197

197 Figure 819. CAD drawing of designed UHV system. (A) perspective, (B) front, (C) rear, (D) side, (E) another side, and (F) top view

PAGE 198

198 CHAPTER 9 CONCLUSIONS AND RECO MMENDATIONS In the present study, several issues have been investigated in the homogeneous thermal decomposition of some metalorganic precursors basically using an in situ Raman spectroscopy and computational calculations. T he CVD reactorused in this study is interfaced with an in situ Raman spectrometer (Ramanor U 1000, Jobin Yvon), which uses the 532.08 nm line of Nd:YAG solidstate laser and adopts a double additive monochromator with a diffraction grating of 1800 groo ves/mm DFT calculations were performed using Gaussian 03 program package to calculate molecular optimized geometry, thermodynamic properties, and rate parameters. In triethylgallium thermal decomposition study in the CVD reactor, one can confirm the pres ence both of homolysis and hydride elimination experimentally. DFT calculations were performed to assist vibrational frequency prediction and assignment and thermodynamic properties estimation, etc. In this study, Raman shift measurements shown at 490cm1, 517cm1, 537cm1, and 555cm1 were assigned to the vibrational frequencies between gallium and carbon of (Et)3Ga, (DEGa)2, (Et)GaH Ga (Et)2and (Et)GaH Ga H2, respectively, by DFT calculations Optimized activation energies are 60.0 and 44.0 kcal/mol and frequency factors are 1.261021s1and 8.91013s1for homolysis and hydride elimination, respectively. These results were obtained by using FEM reactor modeling and parameter optimization algorithm. And these results showed very good agreement with experimental observations. The study of metalorganic precursor interaction with ammonia to construct group IIInitride was performed using triethylaluminum. To find out ammonia effect on triethylaluminum, two separate studies were carried out, i.e. triethylaluminum

PAGE 199

199 decomposition with and without ammonia. To find out ammonia effect on triethylaluminum, in situ Raman spectroscopy measurement in an up flow, cold wall CVD reactor was used and the results of Densi ty Functional Theory (DFT) calculations were performed. The observed Raman shifts were assigned based on references and DFT calculations as follows: TEAl:NH3 (452 and 2841 cm1), DEAlH (600 cm1) TEAl:NH3 TS (1462 cm1) H2N AlHNHAlH2(1525 cm1) H2Al NH2(1639 cm1) MEAlH2, (2835 cm1) MEAlH AlH2(2849 and 2939 cm1) and DEAl AlH2(2900 and 2918 cm1) DFT calculations using B3LYP functional combined with LanL2DZ basis set were also carried out to find the optimized geometry of each intermediate and t ransition structure and to calculate activation energies. The decomposition pathways of the tungsten dimethylhydrazido complexes Cl4(RCN)W(NNMe2) (1a: R = CH3; 1b : R = Ph), precursors for single source deposition of WNxCy, were investigated using a combina tion of experiments and calculations. Since complex 1a has low volatility, Raman scattering studies were performed in an aerosol assisted CVD reactor to identify reaction intermediates. Though complex 1a is introduced as an aerosol state to reactor inlet, it can be realized that aerosol vaporizes in very short time enough to be considered as gas phase by droplet evaporation analysis. DFT calculations (B3LYP/LanL2DZ) were used to estimate Raman active frequencies and these results were compared with experimental and literature data. methylmethyleneimine ( 8 ), dimethylamine ( 9 ), and HNWCl4 ( 14), products from N N cleavage of the hydrazido ligand, were observed under deposition conditions a nd identified by comparison with previously reported Raman shifts and calculated frequencies.

PAGE 200

200 Based on the obtained results through this study, a few modifications and future works are recommended. At first, although in situ Raman spectroscopic study has good aspects to investigate reaction kinetics of metalorganic precursors, infrared spectroscopy can still provide useful information. Since infrared spectroscopy is much more sensitive than Raman spectroscopy and has large number of database especially fo r organic materials, it can give very good supplementary information to reaction intermediates analysis. If infrared spectroscopy gives presence and concentration variation criteria for intermediate organic materials, performance of reaction kinetics anal ysis using Raman spectroscopy can be boosted much more. In practical perspective, attaching infrared spectrometer to the reactor used in this study is worth a try. If the reactor wall is changed to KBr or CaF2 unless infrared source beam is absorbed to the wall during radiation, the present reactor scheme and movable reactor setting can still be used for in situ IR for analysis. Secondly, by combining kinetics study and film growth using appropriate substrate, one can reveal direct correlation between fil m quality or characteristics and kinds or quantity of reaction intermediates. At last, further study with two different metalorganic precursors can give more practical results by trying ternary phase study, for instance, gallium and aluminum precursors wit h ammonia to build AlGaN. By performing these, one can find out very useful reaction mechanisms for characteristics of doped film.

PAGE 201

201 LIST OF REFERENCES [1] G. Moore, Electronics, vol. 38, 1965. [2] J. Wu, W. Walukiewicz, K.M. Yu, J.W. Ager, E.E. Haller, H. Lu, W.J. Schaff, Y. Saito, Y. Nanishi, Applied Physics Letters 80 (2002) 3967. [3] B.G. Streetman, S. Banerjee, Solid State Electronic Devices, Prentice Hall, New Jersey, 2000. [4] W.M. Yim, E.J. Stofko, P.J. Zanzucchi, J.I. Pankove, M. Ettenberg, S.L. Gilbert, Journal of Applied Physics 44/1 (1973) 292. [5] O. Ambacher, Journal of Physics D: Applied Physics 31 (1998) 2653. [6] K.C. Wong, B.G. McBurnett, R.D. Culp, A.H. Cowley, J.G. Ekerdt, Surface Science 416 (1998) 480. [7] M.C. Papu ta, S.J.W. Price, Canadian Journal of Chemistry 57 (1979) 3178. [8] Z. Mahmood, I. Hussain, R.E. Linney, D.K. Russel, Journal of Analytical and Applied Pyrolysis 44 (1997) 29. [9] J.A. McCaulley, R.J. Shul, V.M. Donnelly, J. Vac. Sci. Technol. A 9/6 (1991) 2872. [10] H.M. Ajmera, T.J. Anderson, J. Koller, L. McElweeWhite, D.P. Norton, Thin Solid Films 517/21 (2009) 6038. [11] R.L. McCreery, Raman Spectroscopy for Chemical Analysis, A John Wiley & Sons, Inc., New York, 2000. [12] D. Parras, P. Vandenabeele, A.S. chez, M. Montejo, L. Moens, N. Ramos, Journal of Raman Spectroscopy 41/1 (2009) 68. [13] D.A. Long, Raman spectroscopy, McGrow Hill, New York, 1977. [14] H.W. Schr Liquids, Springer, N ew York, 1979. [15] H.A. Hyatt, J.M. Cherlow, W.R. Fenner, S.P.S. Porto, Journal of the optical society of america 63/12 (1973) 1604. [16] W.R. Fenner, H.A. Hyatt, J.M. Kellam, S.P.S. Porto, Journal of the optical society of america 63/1 (1973) 73. [17] N. C. Craig, I.W. Levin, Applied Spectroscopy 33/5 (1979) 475.

PAGE 202

202 [18] P.R.N. Childs, J.R. Greenwood, C.A. Long, Review of Scientific Instruments 71/8 (2000) 2959. [19] F. LaPlant, G. Laurence, D. BenAmotz, Applied Spectroscopy 50/8 (1996) 1034. [20] B.J. Kip, R.J. Meier, Applied Spectroscopy 44/4 (1990) 707. [21] R.J. Butcher, D.V. Willetts, W.J. Jones, Proceedfings of The Royal Society (London) A 324/1557 (1971) 231. [22] J.J. Barrett, J. Opt. Soc. Am. 66 (1976) 801. [23] G. Vaughan, D.P. Wareing, S.J. Pepler, L. Thomas, V. Mitev, Applied Optics 32/15 (1993) 2758. [24] B.S. Samyal, R.S. Hickman, American Institute of Aeronautics and Astronautics, Fluid and Plasma Dynamics Conference, 13th (1980) 8. [25] M.C. Drake, Optics Letters 7/9 (1982) 440. [26] J.M. Ala, H.G.M. Edwards, Journal of Molecular Structure 354/2 (1995) 97. [27] J.M. Ala, H.G.M. Edwards, J. Moore, Spectrochimica Acta Part A: Molecular Spectroscopy 51 (1995) 2039. [28] J. H. Yin, S. Q. Gao, Z.W. Li, Y. N. Yu, G.H. Lu, Y.J. Tian, Journal of Raman Spectroscopy 35 (2004) 1042. [29] P.T. Araujo, A. Jorio, Phys. Stat. Sol. 245/10 (2008) 2201. [30] A. Jorio, R. Saito, J.H. Hafner, C.M. Lieber, M. Hunter, T. McClure, G. Dresselhaus, M.S. Dresselhaus, Physical Review Letters 86/6 (2001) 1118. [31] P .W. C. Silvara Batista, J.E. Butler, and K.J. Ziegler, J. Am. Chem. Soc. 131 /35 (2009) 12721 12728. [32] R.K.W. W. C. Chen, and K.J. Ziegler, ACS Appl. Mater. Interface 1/8 (2009) 1821. [33] A.M. Rao, J. Chen, E. Richter, U. Schlecht, P.C. Eklund, R.C. Ha ddon, U.D. Venkateswaran, Y.K. Kwon, D. Tom?ek, Physical Review Letters 86/17 (2001) 3895. [34] M.M. Mitambo, G.R. Loppnow, Chemical Physics Letters 261 (1996) 691. [35] S. Jeon, J. Woo, J. B. Kyong, J. Choo, Bull. Korean Chem. Soc. 22/11 (2001) 1264. [36] A. Palm, J. Phys. Chem. 55/8 (1951) 1320.

PAGE 203

203 [37] B. Chu, G. Fytas, G. Zalczer, Macromolecules 14 (1981) 395. [38] M. Lee, J. P. Lee, H. Rhee, J. Choo, Y.G. Chai, E.K. Lee, Journal of Raman Spectroscopy 34 (2003) 737. [39] O.K. Kim, W.G. Spitzer, Journal of Applied Physics 50/6 (1979) 4362. [40] C. Wetzel, W. Walukiewicz, E.E. Haller, J. Ager, I. Grzegory, S. Porowski, T. Suski, Physical Review B 53/3 (1996) 1322. [41] H.y. Kim, J. Ahn, J. Kim, Journal of Ceramic Processing Research 9/1 (2008) 10. [42] J.B. C ui, K. Amtmann, J. Ristein, L. Ley, Journal of Applied Physics 83/12 (1998) 7929. [43] J. Kim, J.J.A. Freitas, J. Mittereder, R. Fitch, B.S. Kang, S.J. Pearton, F. Ren, Solid State Electronics 50/3 (2006) 408. [44] S. Roy, P. Guha, S.N. Kunda, H. Hanzawa, S. Chaudhuri, A.K. Pal, Materials chemistry and physics 73 (2002) 24. [45] S. Shirakata, Y. Kannaka, H. Hasegawa, T. Kariya, S. Isomura, Jpn. J. Appl. Phys. 38 (1999) 4997. [46] J.H. Ely, T.R. Ohno, T.E. Furtak, A.J. Nelson, Thin Solid Films 371 (2000) 36. [47] C. Rincon, F.J. Ramirez, J. Appl. Phys 72/9 (1992) 4321. [48] O. Ramdani, J.F. Guillemoles, D. Lincot, P.P. Grand, E. Chassaing, O. Kerrec, E. Rzepka, Thin Solid Films 515 (2007) 5909. [49] S. Shirakata, H. Kubo, C. Hamaguchi, S. Isomura, Jpn. J. Appl. Phys. 36 (1997) 1394. [50] W. Witte, R. Kniese, M. Powalla, Thin Solid Films 517 (2008) 867. [51] J. Weszka, P. Daniel, A. Burian, A.M. Burian, A.T. Nguyen, Journal of NonCrystalline Solids 265 (2000) 98. [52] F. Lin, G.Q. Bian, Z.X. Lei, Z.J. Lu, J Dai, Solid State Sciences 11 (2009) 972. [53] M. Ramsteiner, J. Wagner, P. Koidl, Applied Physics Letters 59/27 (1991) 3616. [54] D.E. Aspnes, S.M. Kelso, R.A. Logan, R. Bhat, J. Appl. Phys 60/2 (1986) 754. [55] C. Asawaroengchai, G.M. Rosenblatt, J. Chem. Phys. 72/4 (1980) 2664. [56] M. Hizushima, The Theory of Rotating Diatomic molecules, Wiley, New York, 1975.

PAGE 204

204 [57] D.E. Aspnes, A.A. Studna, Physical Review B 27/2 (1983) 985. [58] K. Senthil, D. Mangalaraj, S.K. Narayandass, R. Kesavamoorthy, G.L.N. Reddy, Nuclear Instruments and Methods in Physics Research Section B: Beam Interactions with Materials and Atoms 173/4 (2001) 475. [59] H. JobinYvon, in: http://www.horiba.com/fileadmin/uploads/Scientific/Documents/Raman/Semiconductors 01.pdf (Ed.), Horiba JobinYvon. [60] M. Holtz, W.M. Duncan, S. Zollner, R. Liu, Journal of Applied Physics 88/5 (2000) 2523. [61] H. Harima, Microelectronic Engineering 83/1 ( 2006) 126. [62] C.J. Cramer, Essentials of Computational Chemistry Theories and Models, John Wiley & Sons, Chichester, England, 2006. [63] D.R. Burgess, M.R. Zachariah, W. Tsang, P.R. Westmoreland, Progress in Energy and Combustion Science 21/6 (1995) 453. [64] M.D. Allendorf, C.F. Melius, P. Ho, M.R. Zachariah, The Journal of Physical Chemistry 99/41 (1995) 15285. [65] P. Ho, C.F. Melius, The Journal of Physical Chemistry 99/7 (1995) 2166. [66] P. Ho, M.E. Colvin, C.F. Melius, The Journal of Physical Chemi stry A 101/49 (1997) 9470. [67] H. Simka, B.G. Willis, I. Lengyel, K.F. Jensen, Prog. Crystal Growth and Charact. 35 (1998) 117. [68] I.N. Levine, Quantum Chemistry, Prentice Hall, Upper Saddle River, NY, 2006. [69] D.K. Malick, G.A. Petersson, J. John A. Montgomery, Journal of Chemical Physics 108/14 (1998) 5704. [70] M.J.S. Dewar, E.F. Healy, J.J.P. Stewart, J. Chem. Soc., Faraday Trans. 2 80 (1984) 227. [71] N.W. Mitzel, C. Lustig, R.J.F. Berger, N. Runeberg, Angewandte Chemie International Edition 41/14 (2002) 2519. [72] D.A. McQuarrie, Statistical Mechanics, University Science Books, Sausalito, CA, 2000.

PAGE 205

205 [73] M.J. Frisch, G.W. Trucks, H.B. Schlegel, G.E. Scuseria, M.A. Robb, J.R. Cheeseman, J.J.A. Montgomery, T. Vreven, K.N. Kudin, J.C. Burant, J.M. Mil lam, 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. Danie ls, 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, J.A. Pople, 2004. [74] M. Tsuda, S. Oikawa, M. Morishita, M. Mashita, Japanese Journal of Applied Physics 26/5 (1987) L564. [75] V.A. Yablokov, N.V. Yablokova, Russian Chemical Reviews 64/10 (1995) 953. [76] A.H. McDaniel, M.D. Allendorf, Chemistry of Materials 12/2 (2000) 450. [77] G. Keresztury, S. Holly, G. Besenyei, J. Varga, A. Wang, J.R. Durig, Spectrochimica Acta Part A: Molecular Spectroscopy 49/1314 (1992) 2007. [78] D. Michalska, R. Wysokinski, Chemical Physics Letters 403 (2005) 211. [79] H.B. Schlegel, Geometry Optimization 1, Wiley, Chichester, 1998. [80] J. Ouazzani, K. C. Chiu, F. Rosenberger, Journal of Crystal Growth 91/4 (1988) 497. [81] M .E. Coltrin, R.J. Kee, J.A. Miller, J. Electrochem. Soc. 133/6 (1986) 1206. [82] K.F. Jensen, Journal of Crystal Growth 98/12 (1989) 148. [83] D.I. Fotiadis, M. Boekholt, K.F. Jensen, W. Richter, Journal of Crystal Growth 100/3 (1990) 577. [84] W.L. Holstein, Prog. Crystal Growth and Charact. 24 (1992) 111. [85] R.B. Bird, W.E. Stewart, E.N. Lightfoot, Transport phenomena, John Wiley & Sons, New York, 2001. [86] W.M. Deen, Analysis of Transport Phenomena, Oxford University Press, New York, 1998. [87] J.L. Plawsky, Transport phenomena fundamentals, Marcel Dekker, Inc., New York, 2001.

PAGE 206

206 [88] H.A. Jakobsen, Elementary Kinetic Theory of Gases, Springer, Berlin, 2008. [89] P.D. Neufeld, A.R. Janzen, R.A. Aziz, The Journal of Chemical Physics 57/3 (1972) 1100. [90] A. Eucken, Physik. Z. 14 (1913) 324. [91] J.Y. Hwang, C. Park, M. Huang, T. Anderson, Journal of Crystal Growth 279/521530 (2005). [92] T. Y. Park, G.F. Froment, Industrial & Engineering Chemistry Research 40/20 (2001) 4187. [93] R. Span, E.W. Lemm on, R.T. Jacobsen, W. Wagner, A. Yokozeki, Journal of Physical and Chemical Reference Data 29/6 (2000) 1361. [94] J. Smukala, R. Span, W. Wagner, Journal of Physical and Chemical Reference Data 29/5 (2000) 1053. [95] H. Miyamoto, K. Watanabe, International Journal of Thermophysics 22/2 (2001) 459. [96] M.S. Tomar, R. Rutherford, C. New, K.A. Kuenhold, Solar Energy Material & Solar Cells 63 (2000) 437. [97] C.A. Wang, Jounal of Crystal Growth 272 (2004) 664. [98] A.A. Aquino, T.S. Jones, Applied Surface Science 104/105 (1996) 304. [99] R. Lin, T.R. Gow, A.L. Backman, L.A. Cadwell, F. Lee, R.I. Masel, Journal of Vacuum Science and Technology B 7/4 (1989) 725. [100] D.K. Russell, G.P. Mills, J.B. Raynor, A.D. Workman, Chemical Vapor Deposition 4/4 (1998) 61. [1 01] A.S. Grady, R.E. Linney, R.D. Markwellt, D.K. Russell, Journal of Material Chemistry 3/5 (1993) 483. [102] Y.S. Kim, Y.S. Won, H. Hagelin Weaver, N. Omenetto, T.J. Anderson, Journal of Physical Chemistry A 112/42464253 (2008). [103] R.S. Hickman, L.H. Liang, The Review of scientific intruments 43/5 (1972). [104] R.W. Dibble, S.H. Starner, A.R. Masri, R.S. Barlow, Applied Physics B: Photophysics and Laser Chemistry B51 (1990) 39. [105] E.J. Burlbaw, R.L. Armstrong, Applied Optics 22 (1983) 2860.

PAGE 207

207 [106] J.D. Lorentzen, S. Guha, J. Menendez, P. Giannozzi, S. Baroni, Chemical Physics Letters 270 (1997) 129. [107] S.D. Williams, T.J. Johnson, T.P. Gibbons, C.L. Kitchens, Theoretical Chemical Accounts 117 (2006) 283. [108] R. Boese, D. Blser, N. Niederprm, M Nsse, W.A. Brett, P.v.R. Schleyer, M. Bhl, N.J.R.v.E. Hommes, Angewandte Chemie International Edition in English 31/3 (1992). [109] Y.S. Kim, Ph. D. Thesis, University of Florida (2007). [110] C.H. Park, J.Y. Hwang, M. Huang, T.J. Anderson, Thin Solid Films 409 (2002) 88. [111] C. Park, W. Jung, Z. Huang, T.J. Anderson, Journal of Materials Chemistry 12 (2002) 356. [112] M. Huang, PhD, Chemical Engineering, University of Florida, Gainesville, FL, 2003. [113] J.Y. Hwang, Ph. D. Thesis, University of Florida (2004). [114] J.N. Kidder, Jr. H.K. Yun, J. J.W. Rogers, T.P. Pearsall, Chem. Mater. 10 (1998) 777. [115] B. Abdallah, C. Duquenne, M.P. Besland, E. Gautron, P.Y. Jouan, P.Y. Tessier, J. Brault, Y. Cordier, M.A. Djouadi, The European Physical Jo urnal Applied Physics 43 (2008) 309. [116] D.C. Bertolet, H. Liu, J. J.W. Rogers, J. Appl. Phys 75/10 (1994) 5385. [117] K. L. Ho, K.F. Jensen, J. W. Hwang, W.L. Gladfelter, J.F. Evans, Journal of Crystal Growth 107/14 (1991) 376. [118] W.L. Smith, T. War tik, J. Inorg. Nucl. Chem 29 (1967) 629. [119] J. Lee, Y.S. Kim, T. Anderson, ECS Transactions 25/8 (2009) 41. [120] J.E. House, Inorganic chemistry, Academic Press, Oxford, UK, 2008. [121] Z. Jiang, L.V. Interrante, Chem. Mater. 2 (1990) 439. [122] S. Kvi sle, E. Rytter, Spectrochimica Acta 40A/10 (1984) 939. [123] O. Yamamoto, Bulletin of the Chemical Society of Japan 35/4 (1962) 619. [124] M. Ikenaga, K. Nakamura, A. Tachibana, K. Matsumoto, Journal of Crystal Growth 237239 (2002) 936.

PAGE 208

208 [125] A.Y. Timoshk in, Coordination Chemistry Reviews 249 (2005) 2094. [126] J. W. Hwang, S.A. Hanson, D. Britton, J.F. Evans, K.F. Jensen, W.L. Gladfelter, Chem. Mater. 2 (1990) 342. [127] J.A. Jegier, S. McKernan, W.L. Gladfelter, Inorg. Chem. 38 (1999) 2726. [128] A.Y. Ti moshkin, I. Henry F. Schaefer, Inorganic Chemistry 43 (2004) 3080. [129] F.C. Sauls, W.J. Hurley, L.V. Interrante, P.S. Marchetti, G.E. Maciel, Chem. Mater. 7 (1995) 1361. [130] A.E. Kaloyeros, X.M. Chen, T. Stark, K. Kumar, S. Seo, G.G. Peterson, H.L. Frisch, B. Arkles, J. Sullivan, J. Electrochem. Soc. 146/1 (1999) 170. [131] E. Kolawa, P.J. Pokela, J.S. Reid, J.S. Chen, M.A. Nicolet, Applied Surface Science 53 (1991) 373. [132] 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, L. McElwee White, J. Am. Chem. Soc. 127 (2005) 7825. [133] H.M. Ajmera, A.T. Heitsch, T.J. Anderson, C.B. Wilder, L.L. Reitfort, L. McElweeWhite, D.P. Norton, J. Vac. Sci. Technol. B 26 (2008) 1800. [134] S. H. K im, S.S. Oh, H.M. Kim, D. H. Kang, K. B. Kim, W. M. Li, S. Haukka, M. Tuominen, J. Electrochem. Soc. 151/4 (2004) C272. [135] J. Koller, H.M. Ajmera, K.A. Abboud, T.J. Anderson, L. McElweeWhite, Inorg. Chem. 47 (2008) 4457. [136] D. Kim, O.H. Kim, T.J. Anderson, J. Koller, L. McElweeWhite, L.C. Leu, J.M. Tsai, D.P. Norton, J. Vac. Sci. Technol. A 27 (2009) 943. [137] O.J. Bchir, S.W. Johnston, A.C. Cuadra, T.J. Anderson, C.G. Ortiz, B.C. Brooks, D.H. Powell, L. McElwee White, Journal of Crystal Growth 249 (2003) 262. [138] O.J. Bchir, K.M. Green, M.S. Hlad, T.J. Anderson, B.C. Brooks, C.B. Wilder, D.H. Powell, L. McElweeWhite, J. Organomet. Chem. 684/12 (2003) 338. [139] J.H.S. Green, D.J. Harrison, Spectrochim. Acta, Part A 32/6 (1976) 1279. [140] T.G Burova, A.A. Anashkin, Optics & Spectroscopy 102/6 (2007) 825. [141] S. Ishikawa, T. Ebata, N. Mikami, Journal of Chemical Physics 110/19 (1999) 9504. [142] H.M. Ajmera, Ph. D. Dissertation, Chemical Engineering, University of Florida, Gainesville, Fl., 2007.

PAGE 209

209 [143] J.A. Kerr, Chemical Reviews 66 (1966) 465. [144] C.J. Michejda, W.P. Hoss, J. Am. Chem. Soc. 92/21 (1970) 6298. [145] J.R. Roberts, K.U. Ingold, J. Am. Chem. Soc. 95/10 (1973) 3228. [146] F.O. Rice, C.J. Grelecki, J. Am. Chem. Soc. 79/11 (1957) 2679. [147] D. Mackay, A. Waters, J. Chem. Soc. C (1966) 813. [148] H. Gesser, J.T. Mullhaupt, J.E. Griffiths, J. Am. Chem. Soc. 79 (1957) 4834. [149] B.G. Gowenlock, K.E. Thomas, J. Chem. Soc. B (1966) 409. [150] S. Salim, C.K. Lim, L.F. Jensen, Chemistr y of Materials 7/3 (1995) 507. [151] Y.G. Lazarou, P. Papagiannakopoulos, The Journal of Physical Chemistry 97 (1993) 9133. [152] H. Shang, C. Yu, L. Ying, X. Zhao, Chemical Physics Letters 236 (1995) 318. [153] J. Demuynck, D.J. Fox, Y. Yamaguchi, H.F.S. III, J. Am. Chem. Soc. 102/20 (1980) 6204. [154] C. Richards, C. Meredith, S. J. Kim, G.E. Quelch, H.F. Schaefer, Journal of Chemical Physics 100/1 (1994) 481. [155] R.J. Lang, J. Acoust. Soc. Amer. 34/1 (1962) 6. [156] S.M. Frolov, F.S. Frolov, B. Basara, J. Russ. Laser Res. 27/6 (2006) 562. [157] J.R. Durig, M.G. Griffin, P. Groner, The Journal of Physical Chemistry 81/6 (1977) 554. [158] G. Socrates, Infrared and Raman Characteristic Group Frequencies: Tables and Charts, John Wiley & Sons, Inc, New York, 2004. [159] T.E. Glassman, M.G. Vale, R.R. Schrock, Organometallics 10/12 (1991) 4046. [160] D.F. McMillen, D.M. Golden, Ann. Rev. Phys. Chem. 33 (1982) 493. [161] G.T. Wang, C. Mui, C.B. Musgrave, S.F. Bent, The Journal of Physical Chemistry B 103/32 (19 99) 6803. [162] M. Bhattacharya, W.G. Devi, P.S. Mazumdar, Applied Surface Science 218 (2003) 1. [163] P.A. Redhead, Vacuum 12 (1962) 203.

PAGE 210

210 [164] P.A. Redhead, Transactions of the Faraday Society 57 (1961) 641. [165] C.R. Brundle, J. Charles A. Evans, S. Wilson, Encyclopedia of materials characterization : Surfaces, Interfaces, Thin Films, ButterworthHeinemann, Greenwich, CT, 1992. [166] P.W. Palmberg, R.E. DeWames, L.A. Vredevoe, Physical Review Letters 21/10 (1968) 682. [167] S. Ye, T. Ichihara, K. Uos aki, Journal of Electrochemical Society 148 (2001) C421. [168] C. Mui, J.H. Han, G.T. Wang, C.B. Musgrave, S.F. Bent, J. Am. Chem. Soc. 124/15 (2002) 4027. [169] W. Ranke, Surface Science 342/13 (1995) 281. [170] K. Choi, J.M. Buriak, Langmuir 16/20 (2000) 7737. [171] A. Bansal, X. Li, I. Lauermann, N.S. Lewis, S.I. Yi, W.H. Weinberg, J. Am. Chem. Soc. 118/30 (1996) 7225. [172] J. He, S.N. Patitsas, K.F. Preston, R.A. Wolkow, D.D.M. Wayner, Chemical Physics Letters 286/56 (1998) 508. [173] G.W. Cullen, J. A. Amick, D. Gerlich, J. Electrochem. Soc. 109/2 (1962) 124. [174] M.J. Bozack, P.A. Taylor, W.J. Choyke, J.T. Yates Jr, Surface Science 177/1 (1986) L933. [175] M.A. Filler, C.B. Musgrave, S.F. Bent, The Journal of Physical Chemistry C 111/4 (2007) 1739. [176] R. Konecny, D.J. Doren, J. Am. Chem. Soc. 119/45 (1997) 11098. [177] M.J. Kelly, J.H. Han, C.B. Musgrave, G.N. Parsons, Chemistry of Materials 17/21 (2005) 5305. [178] M.R. Kosuri, H. Gerung, Q. Li, S.M. Han, B.C. Bunker, T.M. Mayer, Langmuir 19/22 ( 2003) 9315. [179] A. Kim, M.A. Filler, S. Kim, S.F. Bent, J. Am. Chem. Soc. 127/16 (2005) 6123. [180] Y.S. Cho, S.I. Cho, J.S. Heo, J.C. Kim, S.H. Moon, J. Electrochem. Soc. 153/3 (2006) C195. [181] A.R. Hind, S.K. Bhargava, A. McKinnon, Advances in Colloid and Interface Science 93/13 (2001) 91.

PAGE 211

211 [182] I. PerkinElmer, in: http://las.perkinelmer.com/content/TechnicalInfo/TCH_FTIRATR.pdf (Ed.), PDF, PerkinElmer Life and Analytical Sciences, 2005. [183] J.M. Buriak, Chemical Reviews 102/5 (2002) 1271. [184] G.S. Higashi, Y.J. Chabal, G.W. Trucks, K. Raghavachari, Applied Physics Letters 56/7 (1990) 656.

PAGE 212

212 BIOGRAPHICAL SKETCH Jooyoung Lee was born in 1976 in Seoul, Korea. He entered Seoul National University in March 1995 and received his Bachelor of Science degree in chemical engineering in February 1999. After the Bachelor of Science degree, he started mandatory military se rvice as a tactical driver in Yangyang si, Kangwondo. When he finished 26 months of service, he got married to Hankyung Seong in 2002 and he started graduate school in Seoul National University for process optimization and safety and received his Master of Science degree in chemical engineering in August, 2005. Then Jooyoung Lee decided to pursue his Ph.D. degree in the University of Florida. He started his Ph.D. program in August 2006 at the University of Florida and joined the electronic materials processing group to work with Dr. Tim Anderson a nd he devoted himself to chemical vapor deposition kinetics research for semiconductor processing.