<%BANNER%>

Atomic Layer Deposition of GaN and TaN

Permanent Link: http://ufdc.ufl.edu/UFE0025153/00001

Material Information

Title: Atomic Layer Deposition of GaN and TaN
Physical Description: 1 online resource (129 p.)
Language: english
Creator: Kim, Oh
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2009

Subjects

Subjects / Keywords: air, atomic, cu, diffusion, gan, led, p, tan, taoxny
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: Atomic Layer Deposition of GaN for LED applications and TaN for diffusion barrier applications in Cu metallization were demonstrated in this work. GaN films were grown on Si(100) by ALD using GaCl3 and NH3 as the precursors. ALD process windows for self-saturation limit were identified. In this limit, both the (0002) and (10 1) orientations of GaN were evident, which was attributed to competition between vertical and lateral growth. Also, Cl incorporation was not detected in ALD-GaN grown within the ALD window. To better understand the lack of detectable Cl incorporation in ALD-GaN, chemisorptions of GaCl3 and NH3 onto surface active sites were simulated using computational chemistry. Relatively high energy barrier for GaCl3 adsorption decreased the extent of molecular trapping of a Cl complex on the growing surface. Furthermore, a low energy barrier for desorption of HCl facilitates cleavage of Cl from the growing surface. The ability of growing p-type doped GaN is important for LED application. Mg-doped GaN films were grown on sapphire substrate by HVPE using GaCl3, NH3, and bis-cyclopentadienyl magnesium (Cp2Mg). A systematic study of the growth temperature, Cp2Mg flow rate, and N/Ga ratio produced the highest hole concentration of 3.2 times 1017 cm-3. At this growth condition, the PL emission at 3.12 eV was obtained with 0.14 eV of FWHM. TaN films were grown on Si(100) by ALD using tert-butylimido tris(dimethylamido) tantalum (TBTDET) and NH3. This material is commonly used as a barrier material in Cu metallization schemes. The effect of air exposure on TaN diffusion barriers prior to Cu deposition was studied by comparing Cu/TaN/Si stacks grown with and without air exposure as a function of annealing temperature. Interestingly, it was found that the barrier failure temperature of air exposed films is higher than those vacuum transferred. The TaOxNy interfacial layer appears to have a beneficial role in preventing Cu transport.
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 Oh Kim.
Thesis: Thesis (Ph.D.)--University of Florida, 2009.
Local: Adviser: Anderson, Timothy J.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2011-12-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2009
System ID: UFE0025153:00001

Permanent Link: http://ufdc.ufl.edu/UFE0025153/00001

Material Information

Title: Atomic Layer Deposition of GaN and TaN
Physical Description: 1 online resource (129 p.)
Language: english
Creator: Kim, Oh
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2009

Subjects

Subjects / Keywords: air, atomic, cu, diffusion, gan, led, p, tan, taoxny
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: Atomic Layer Deposition of GaN for LED applications and TaN for diffusion barrier applications in Cu metallization were demonstrated in this work. GaN films were grown on Si(100) by ALD using GaCl3 and NH3 as the precursors. ALD process windows for self-saturation limit were identified. In this limit, both the (0002) and (10 1) orientations of GaN were evident, which was attributed to competition between vertical and lateral growth. Also, Cl incorporation was not detected in ALD-GaN grown within the ALD window. To better understand the lack of detectable Cl incorporation in ALD-GaN, chemisorptions of GaCl3 and NH3 onto surface active sites were simulated using computational chemistry. Relatively high energy barrier for GaCl3 adsorption decreased the extent of molecular trapping of a Cl complex on the growing surface. Furthermore, a low energy barrier for desorption of HCl facilitates cleavage of Cl from the growing surface. The ability of growing p-type doped GaN is important for LED application. Mg-doped GaN films were grown on sapphire substrate by HVPE using GaCl3, NH3, and bis-cyclopentadienyl magnesium (Cp2Mg). A systematic study of the growth temperature, Cp2Mg flow rate, and N/Ga ratio produced the highest hole concentration of 3.2 times 1017 cm-3. At this growth condition, the PL emission at 3.12 eV was obtained with 0.14 eV of FWHM. TaN films were grown on Si(100) by ALD using tert-butylimido tris(dimethylamido) tantalum (TBTDET) and NH3. This material is commonly used as a barrier material in Cu metallization schemes. The effect of air exposure on TaN diffusion barriers prior to Cu deposition was studied by comparing Cu/TaN/Si stacks grown with and without air exposure as a function of annealing temperature. Interestingly, it was found that the barrier failure temperature of air exposed films is higher than those vacuum transferred. The TaOxNy interfacial layer appears to have a beneficial role in preventing Cu transport.
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 Oh Kim.
Thesis: Thesis (Ph.D.)--University of Florida, 2009.
Local: Adviser: Anderson, Timothy J.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2011-12-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2009
System ID: UFE0025153:00001


This item has the following downloads:


Full Text

PAGE 1

1 ATOMIC LAYER DEPOSITION OF GaN AND TaN By OH HYUN KIM A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVER SITY OF FLORIDA 2009

PAGE 2

2 2009 Oh Hyun Km

PAGE 3

3 To my parents Ukdong Kim and Kyungja Ko

PAGE 4

4 ACKNOWLEDGMENTS I thank the chair of my advisory committee, Dr. Timothy J. Anderson, for his support, education, and ex cellent guidance during my Ph D program in University of Florida. He taught me how to solve scientific and engineering black box problem as a PhD. He also inspired me to utilize creative thinking when I faced dead end s in my research. My advisory commi ttee members, Dr. Lisa McElwee White, Dr. Valentin Craciun, and Dr. Fan Ren gave me valuable advice on chemistry, material science, and device fabrication respectively My fellow Ph.D. student, Dojun Kim was an excellent collaborator and mentor. Our d is cussions helped my research in many ways. I also learned how to manage a research project from him. I appreciate t he love from his family which made me feel at home I also thank my colleagues, Jrgen Koller, Jooyoung Lee, Vaibhav Chaudhari, David Wood Revelli, fo r their help. My best friend, J. Patrick McKinney, gave me exceptional friendship a s my second family. I am grateful to m y fiance, Hojung Kim She supported me with love and care during my Ph D course. A nd, her solid trust in me allowed me to concentrate on my work. I am thankful to m y two lovely sisters, Ansoon Kim and Sookheun Kim, for their care in many ways. Also, they were my role model s and excellent mentor s Last but the most, I would like to g ive my sincerest thank s to m y parents, Ukdong Kim and Kyoungja Ko Without them, I would not have had the chance to pursue the P h D T he ir unconditional love and support is at the root of this accomplishment.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ ............... 4 LIST OF TABLES ................................ ................................ ................................ ........................... 8 LIST OF FIGURES ................................ ................................ ................................ ......................... 9 ABSTRACT ................................ ................................ ................................ ................................ ... 12 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .................. 14 1.1 Problem Statement for Atomic Layer Deposition of GaN ................................ ........... 14 1.1.1 Challenges for Solid State Lighting ................................ ................................ 14 1.1.2 Hypothesis ................................ ................................ ................................ ........ 15 1.2 Problem Statemen t for Atomic Layer Deposition of TaN ................................ ........... 16 1.2.1 Diffusion Barrier for Cu Metallization ................................ ............................ 16 1.2.2 Limitation of PVD ................................ ................................ ........................... 17 1.2.3 Hypothesis ................................ ................................ ................................ ........ 18 2 LITERATURE REVIEW ................................ ................................ ................................ ....... 23 2.1 Development of GaN ................................ ................................ ................................ ... 23 2.2 Application of GaN in LED ................................ ................................ ......................... 25 2.3 Atomic Layer Deposition of GaN ................................ ................................ ................ 27 2.4 Doping of GaN ................................ ................................ ................................ ............. 29 2.4.1 n type Doping of GaN ................................ ................................ ..................... 29 2.4.2 p type Doping of GaN ................................ ................................ ..................... 30 2.5 Challenges in Interconnects ................................ ................................ ......................... 33 2.5.1 RC Delay in Interconnects ................................ ................................ ............... 33 2.5.2 Electro migration in Intercon nects ................................ ................................ .. 34 2.5.3 Diffusion of Cu into Si ................................ ................................ ..................... 35 2.6 TaN as Diffusion Barrier ................................ ................................ ............................. 36 2.6.1 CVD of TaN ................................ ................................ ................................ ..... 36 2.6.2 ALD of TaN ................................ ................................ ................................ ..... 39 2.7 Cu Metallization ................................ ................................ ................................ ........... 42 2.7.1 Processing of Cu Interconnect ................................ ................................ ......... 42 2.7.2 Microstructure of Cu Thin Film for Various Deposition Methods .................. 43 2.8 Barrier Failure Mechanism ................................ ................................ .......................... 44 3 EXPERIMENTAL SETUP ................................ ................................ ................................ .... 53 3.1 GaN ALD Reactor Design and Setup ................................ ................................ .......... 53 3.2 Chemical Equilibrium Analysis for Growth Temperature ................................ ........... 55

PAGE 6

6 4 ATOMIC LAYER DEPOSITION OF GaN USING GaCl 3 and NH 3 ................................ .... 61 4.1 Introduction ................................ ................................ ................................ .................. 61 4.2 Experimental Procedures ................................ ................................ ............................. 63 4.3 Results and Discussion ................................ ................................ ................................ 64 4.3.1 Confirmation of Self saturation ALD Growth ................................ ................. 64 4.3.2 Process Temperature Window ................................ ................................ ......... 65 4. 3.3 Growth Characteristic of GaN ALD ................................ ................................ 66 4.3.4 Film Microstructure ................................ ................................ ......................... 68 4.3.5 Chemical Bonding State ................................ ................................ .................. 69 4.4 Conclusions ................................ ................................ ................................ ....................... 70 5 COMPUTATIONAL STUDY FOR ATOMIC LAYER DEPOSITION OF GaN USING GaCl 3 AND NH 3 ................................ ................................ ................................ ..................... 78 5.1 Introduction ................................ ................................ ................................ .................. 78 5.2 Computational Details ................................ ................................ ................................ 79 5.3 Results and Discussion ................................ ................................ ................................ 80 5.3.1 Chemisorption of GaCl 3 ................................ ................................ ................... 80 5.3.2 Chemisorption of NH 3 ................................ ................................ ..................... 82 5.4 Conclusions ................................ ................................ ................................ .................. 83 6 p TYPE DOPING OF GaN USING Cp 2 Mg AS A DOPANT ................................ ............... 88 6.1 Introduction ................................ ................................ ................................ .................. 88 6.2 Experimental Details ................................ ................................ ................................ .... 89 6.3 Results and Discussion ................................ ................................ ................................ 90 6.3.1 Film Microstructure ................................ ................................ ......................... 90 6.3.2 Film Resistivity ................................ ................................ ................................ 90 6.3.3 Hole Properties ................................ ................................ ................................ 91 6.3.4 Optical Properties ................................ ................................ ............................. 92 6.4 Conclusions ................................ ................................ ................................ .................. 94 7 EFFECT OF AIR EXPOSURE ON Cu DIFFUSION BARRIER PROPERTIES OF TaN THIN FILMS GROWN BY ATOMIC LAYER DEPOSITION ................................ ............ 99 7.1 Introduction ................................ ................................ ................................ .................. 99 7.2 Experimental Details ................................ ................................ ................................ .. 101 7.3 Results and Discussion ................................ ................................ .............................. 103 7.3.1 ALD of TaN Thin Films ................................ ................................ ................ 103 7.3.1.1 Film Growth ................................ ................................ .................... 103 7.3.1.2 Barrier Tests ................................ ................................ .................... 104 7.3.2 in situ (IN) Set vs. ex situ (EX) Set ................................ ............................... 105 7.3.2.1 Microstructure of IN and EX Sets ................................ .................. 105 7.3.2.2 Barrier Tests ................................ ................................ .................... 105 7.4 Conclusions ................................ ................................ ................................ ................ 108 8 FUTURE WORK AND SUGGESTIONS ................................ ................................ ............ 116

PAGE 7

7 8.1 Concerns of ALD GaN for Device Application ................................ ............................. 116 8.2 ALD Doping of GaN ................................ ................................ ................................ ...... 116 8.3 Devic e Fabrication of Nanorod based LED ................................ ................................ ... 117 8.4 Control of TaO x N y for Better Diffusion Barrier Properties ................................ ............ 118 LIST OF REFERENCES ................................ ................................ ................................ ............. 120 BIOGRAPHICAL SKETCH ................................ ................................ ................................ ....... 129

PAGE 8

8 LIST OF TABLES Table page 2 1 Acceptor binding energies for C, Mg, Be, Zn and Cd in various III V semiconductors. Taken from Ref [76] ................................ ................................ ............... 47 2 2 Microstructure of Cu films. Taken from Ref [112]. ................................ .......................... 47 3 1 Considered species in Ga Cl system. Taken from Sub94 database. ................................ .. 57 5 1 Standard enthalpy H 298 ) and Gibbs energ y difference at 298K, 1 atm G 298 G 923 K, 0.013 atm ) for chemisorption of GaCl 3 ............... 85 5 2 emisorption of NH 3 ...... 85 7 1 Barrier failure temperatures determined by different barrier testing methods for Cu/TaN/Si stacks of vacuum transferred and air exposed TaN. ................................ ...... 109

PAGE 9

9 LIST OF FIGURES Figure page 1 1 Schematic of novel design LED using nanorod. ................................ ................................ 21 1 2 Schematic of plana r LED structure. ................................ ................................ ................... 21 1 3 SEM images of Cu/Si stacks with and without using TaN diffusion barrier (A) Cu/Si stack after annealing at 500 C, 30 min under N 2 (B) Cu/TaN/Si stack after annealing at 500 C, 30 min under N 2 ................................ ................................ .............. 22 2 1 Experimental configuration of a typical m i n GaN LED. Taken from Ref. [46]. ............ 48 2 2 The struc ture of the high brightness green light emitting diode with 20 thick In 0.43 Ga 0.57 N quantum well as active layer. Taken from Ref. [53]. ................................ ... 48 2 3 Top view of a blue LED fabricated by Nichia with out plastic capsule. Taken from Ref [7]. ................................ ................................ ................................ ............................... 49 2 4 Effect of feature size scaling on total capacitance. Taken from Ref [80]. ......................... 49 2 5 Damaged Al interconnect line showing the formation of hillock and void due to electro migration. Taken from Ref [81]. ................................ ................................ ............ 50 2 6 Electro migration lifetime for aluminum interconnects at different current densities. Taken from Ref [82]. ................................ ................................ ................................ ......... 50 2 7 Band diagram of Cu impurity in silicon. A) p type Si, B) n type Si. [77] ........................ 51 2 8 Hole fill evolution for electroplated Cu with super filling additives. Taken from Ref [100]. ................................ ................................ ................................ ................................ .. 51 2 9 Evolution of hole filling with different deposition conditions varying degrees of conformali ty. Taken from Ref [101]. ................................ ................................ ................. 52 3 1 Schematic of overall GaN ALD system. ................................ ................................ ............ 58 3 2 Schematic of GaN ALD reactor chamber. ................................ ................................ ......... 58 3 3 Control program of GaN ALD process. ................................ ................................ ............. 59 3 4 (A) Equilibrium mole fractions of gas species in the Ga Cl Inert system as a function of te mperature at P=1 atm. (B) Molar ratio of GaCl 3 to Ga 2 Cl 6 in the Ga Cl Inert system as a function of temperature at P=1 atm. ................................ ............................... 60 4 1 Dependence of ALD GaN growth rate on GaCl 3 exposure time. A) T growth = 550 C. B) T growth = 650 C. These films are grown on Si(100) with NH 3 exposure time of 10 s. ................................ ................................ ................................ ................................ ......... 72

PAGE 10

10 4 2 Dependence of the growth rate on temperature. Exposure times were set as GaCl 3 = 7 s, NH 3 = 10 s, and N 2 purge = 30 s. ................................ ................................ ................... 73 4 3 Dependence of film thickness (XRR) on cycle number. Growth conditions were set as GaCl 3 exposure time = 9 s, NH 3 exposure time = 10 s, and N 2 purg e time = 30 s at T growth = 650 C. A) 5 cycles. B) 10 cycles. C) 15 cycles. D) 150 cycles. ......................... 73 4 4 Surface morphology of the GaN films deposited with 9 s of GaCl 3 exposure time at 650 C. A) 5 cycles. B) 10 cycles. C) 15 cycles. D) 150 cycles. ................................ ....... 74 4 5 GIXD patterns for GaN films grown on Si(100) at 650 C. A) ALD GaN films. B) CVD GaN films. ................................ ................................ ................................ ................ 75 4 6 Surface morphology of the 60 nm thick GaN films deposited at 650 C under A) ALD growth condition B) CVD growth condition. ................................ ........................... 76 4 7 XPS spectra of GaN film grown by ALD at various GaCl 3 exposure times. A) gallium 2 p B) nitrogen 1 s C) oxygen 1 s D) chlorine 2 p ................................ ................ 77 5 1 Cluster model used in the calculation to represent GaN GaNH 2 surface. ....................... 86 5 2 H 298 G 298 ) and 923 K, 0.013 G 923 K, 0.013 atm ) for the chemisorption of GaCl 3 representing first half reaction. The unit is kcal/mol. ................................ ................................ ........................... 87 5 3 H 298 G 298 ) and 923 K, 0.013 G 923 K, 0.013 atm ) for the chemisorption of NH 3 representing second half reaction. The unit is kcal/mol. ................................ ................................ ........................... 87 6 1 Crystallinity of Mg doped GaN films as a function of Mg incorporation. ........................ 96 6 2 Resistivities of Mg doped GaN films grown as a function of Mg incorporation. ............. 96 6 3 Hole mobilities of Mg doped GaN films as a function of Mg incorporation. ................... 97 6 4 Hole c oncentrations of Mg doped GaN films as a function of Mg incorporation. ............ 97 6 5 PL spectra of Mg doped GaN films as a function of Mg incorporation. A) undoped, B) Cp 2 Mg: 5 sccm, C) Cp 2 Mg: 10 scc m, D) Cp 2 Mg: 15 sccm, E) Cp 2 Mg: 20 sccm, and F) Cp 2 Mg: 25 sccm. ................................ ................................ ................................ .... 98 7 1 A) Growth rate of ALD TaN as a function of TBTDET exposure time, which is deposited at 250 C, and 10 s of purge and NH 3 exposure time. B) Thickness of ALD TaN films as a function of growth temperature deposited with 9 s of TBTDET exposure time and 10 s of purge and NH 3 exposure time. Taken from Ref [165]. .......... 110

PAGE 11

11 7 2 A) XRD patterns of Cu/TaN/Si annealed at 500 C for 30 min. B) Cross sectional TEM image of Cu/TaN/Si structure. TaN was grown by ALD for 15 cycles at 9 s of TBTDET exposure time at 300 C. Taken from Ref [165]. ................................ ............ 111 7 3 SEM images of Si surface after annealing Cu/TaN/Si at 500 C for 30 min followed by the removal of Cu and TaN, and Secco etching for A) 10 cycles (16 ) TaN film. B) 15 cycles (38 ) TaN film. Taken from Ref [165]. ................................ .................... 112 7 4 Cross sectional TEM image of as grown Cu/TaN/Si structure of A) vacuum transferred TaN. B) air exposed TaN. ................................ ................................ .............. 112 7 5 XRD patterns of annealed Cu/TaN/Si structure of A) vacuum transferred TaN. B) air exposed TaN at elevated annealing temperatures for 30 min. ................................ ......... 113 7 6 Sheet resistances of the Cu/TaN/Si structures for vacuum transferred and air exposed TaN as a function of the annealing temperature. ................................ ............................. 114 7 7 SEM image of etch pits observed on Si surface of Cu/TaN/Si for vacuum transferred TaN at annealing temperat ure of A) 585 C, B) 600 C, C) 615 C, D) 630 C, and E) 650 C. Si surface of Cu/TaN/Si for air exposed TaN at annealing temperature of F) 585 C, G) 600 C, H) 615 C, I) 630 C, and J) 650 C. ................................ ........... 114 7 8 A) Barrier failure temperatures as a function of TaN barrier thickness. B) Plot of ln(L 2 /T 2 ) vs 1/ k T for Cu/TaN/Si stacks of vacuum transferred and air exposed TaN. .... 115

PAGE 12

12 Abstract of Dissertation Presented to t he Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy ATOMIC LAYER DEPOSITION OF GaN AND TaN By Oh Hyun Kim December 2009 Chair: Timothy J. Anderson Major: Chemi cal Engineering A tomic Layer Deposition of GaN for LED applications and TaN for diffusion barrier applications in Cu metallization were demonstrated in this work. GaN films were grown on Si(100) by ALD using GaCl 3 and NH 3 as the precursors. ALD process window s for self saturation limit were identified In this limit, both the (00 0 2) and (10 1 ) orientations of GaN were evident, which was attributed to competition between vertical and lateral growth. Also, Cl incorporation was not d etected in ALD GaN grown within the ALD window To better understand the lack of detectable Cl incorporation in ALD GaN, chemisorptions of GaCl 3 and NH 3 onto surface active sites were simulated using computational chemistry Relatively high energy barrie r for GaCl 3 adsorption decreased the extent of molecular trapping of a Cl complex on the growing surface. Furthermore, a low energy barrier for desorption of HCl facilitates cleavage of Cl from the growing surface. The ability of growing p type dop ed Ga N is important for LED application. Mg doped GaN films were grown on sapphire substrate by HVPE using GaCl 3 NH 3 and bis cyclopentadienyl magnesium (Cp 2 Mg) A systematic study of the growth temperature, Cp 2 Mg

PAGE 13

13 flow rate, and N/Ga ratio produced the highe st hole concentration of 3.2 10 17 cm 3 At this growth condition, the PL emission at 3.12 eV was obtained with 0.14 eV of FWHM. TaN films w e re grown on Si(100) by ALD using tert butylimido tris(dimethylamido) tantalum (TBTDET) and NH 3 This material is commonly used as a barrier material in Cu metallization schemes. The effect of air exposure on TaN diffusion barriers prior to Cu deposition was studied by comparing Cu/TaN/Si stack s grown with and without air exposure as a function of annealing temper ature. Interestingly, it was found that the barrier failure temperature of air exposed films is higher than those vacuum transferred. The TaO x N y interfacial layer appears to have a beneficial role in preventing Cu transport.

PAGE 14

14 CHAPTER 1 INTRODUCTION 1.1 Pr oblem Statement for Atomic Layer Deposition of GaN 1.1.1 Challenges for Solid State Lighting In candescent and fuel based lamps typically convert the less than 5% of the input energy into the visible light. Fluorescent lamps have higher conversion efficiency ge nerally 20% Solid state lighting however, has the potential to surpass the luminous efficiency limitations of other forms of lighting, which leads to significant energy savings. It also presents many advantages over traditional light sources including longer life time, improved robustness, smaller size, and fast switching. There are however, some issues in achieving higher efficiency A key issue limi ting higher efficiency in solid state lighting is the lack of a suitable substrate. The most commer cialized solid state lighting is In x Ga 1 x N multi quantum well (MQW) structure. In x Ga 1 x N LEDs are currently deposited on the lattice mismatched substrate sapphire o r SiC, which resulting t hreading dislocation densities in the range of 5 10 8 to 5 10 9 cm 2 Cathodoluminescence studies show that threading dislocations in GaN film are non radiative defects, thus limiting the conversion efficiency [1, 2] Therefore, a more closely lattice mismatched substrate is required to reduce threading dislocation density. Another important issue in current solid state lighting is the decrease in quantum efficiency with increasing device current. Auger recombination was reported to play a role in this phenomenon [3, 4] A nother report, however associates the reduced efficiency with the escape of electrons from MQW region caused by polarization field It is additionally explained by the decreased mobility of the heavy mass hole in the MQW region [5, 6] This review leads to the observat ion that the most important issues in advancing solid state lighting are identifying suitable substrate and increasing the quantum efficiency.

PAGE 15

15 1.1.2 Hypothesis The broad application of GaN based LEDs for solid state lighting has been limited due to t heir high cost that in pa r t is a result of their low efficiency. Cost can be lowered by increasing the integrated emission per area, which can be achieved by increasing the quantum efficiency and the active area. It is proposed to use nano rod structure to significantly increase the active surface area relative to current planar structure. As shown in Fig. 1 1 the envisioned double hetero junction (DH) LED structure is proposed using the array of nano rods. InGaN active layer deposited on vertically ali gned n type GaN nano rod s needs to be followed by the growth of p type GaN on InGaN active layer to complete the DH structure. In this structure the emission will occur predominantly from the top side of each nano rod through a transparent p contact unl i ke the c onventional planar structure in Fig. 1 2 T his design has the potential to give much higher emission flux per unit substrate area the value of which depends on the nano rod aspect ratio (AR) and density. There are, however, several challenges to realize this LED structure First of all, the active layer needs to be deposited on high AR n type GaN nano rod s at high density and uniformity, particularly along the wall of nano rod. Deposition of smooth and uniform Ga x In 1 x N active layer on n typ e GaN is essential to achieve high external quantum efficiency [7] Secondly, it is necessary to co ntrol the p type doping level of GaN The d oping level is known to be directly related to the performance of LED, especially output wavelength [8] I n th is LED design each nano rod and the deposited layer are an emission unit and uniformity among the nano rods must also be controlled. Given the goal of demonstrating device structure, t he present wo rk is focused on developing a process for the deposition of GaN and Ga x In 1 x N on GaN nano rod to achieve the required control Atomic layer deposition (ALD) was used to deposit a conformal active layer

PAGE 16

16 along the wall of nano rod. ALD takes the advantage of the binding energy difference between chemisorption and physisorption, which can yield self saturation surface adsorption. Unlike chemical vapor deposition ( CVD ) and physical vapor deposition ( PVD ) ALD is layer by layer deposition technique. This ch aracteristic of ALD results in conformal films and precise thickness. Films grown by ALD are well known to yield precise control of thickness. In addition, the growth temperature can be decreased in ALD once surface reaction times are longer. Lower growt h temperature decreases the N vacancy concentration which is known to limit p type doping level of GaN [8] 1.2 Problem Statement for Atomic Layer Deposition of TaN 1.2.1 Diffusion Barrier for Cu Metallization Transport of Cu into Si and low dielectrics lead to detrimental effects on device feature. To prevent diffusion of Cu into Si and dielectric films, a thin film is first deposited that act s as a diffusion barrier between Cu and the Si or dielectric material There are several requirements to produce effective diffusion barrier s for Cu metallization. The diffusion barrier must prevent the diffusion of Cu during independent device processing steps that can be at elevated temperature. Secondly, the diffusion barrier must show good adhesion to Cu, dielectric, and etch stop layers. Good adhesion is essential to maintain device integrity during chemical me chanical polishing ( CMP ) processing which is currently used with dual damascene process. In addition, the diffusion barrier is required to have low electrical resistivity (< 400 cm) as well as good mechanical stability. Besides these material require ments, there are certain process constraints for deposition of the diffusion barrier film Current minimum feature size dictate the growth o f ultra thin film s with thickness in the range of few nano meter. I t is also necessary for the deposition process to achieve excellent conformality and step coverage Additionally, low deposition temperature is preferable to decrease the thermal budget.

PAGE 17

17 Figure 1 3 shows the integrity of Cu and Si stack with and without using a TaN diffusion barrier. TaN was used a s the diffusion barrier. As shown in Fig. 1 3 B significant Cu i ncorporated into Si without diffusion barrier in place With using barrier, there is no inter diffusion of Cu into Si (Fig. 1 3 A) Currently, TaN/Ta bilayer is employed as diffusion barrier /adhesion promoter in the IC industry. TaN shows good diffusion barrier properties and good adhesion to SiO 2 but not to Cu. Ta, on other hand, shows good adhesion to Cu and TaN film. Ta also promotes the growth of Cu (111) during PVD Cu seed layer deposi tion. It is well known that Cu (111) has higher electro migration resistance as compared to other crystal orientations. 1.2.2 Limitation of PVD S caling of the diffusion barrier film is an important challenge given the shrinking dimensions of interconnects l ines. Based on the International Technology Roadmap of Semiconductors (ITRS ), the thickness of diffusion barrier needs to be scaled from 33 by 20 10 to 19 by 201 5 [9] The scaling of barrier film is required with two reasons. First, the rela tive volume of the barrier film occup ying the line or via increases as Cu line and via dimensions shrink, which results in the decreased available volume for Cu This could increase both the RC delay and the current density in Cu lines and vias. Secondly, t he thinner the barrier is, the lower is the overall electrical resistance between via and Cu lin e. To meet the constraints of future scaling the ITRS calls for an improved deposition method. PVD is currently employed to deposit diffusion barrier film in the industry. It is, however believed for PVD to be incapable of depositing ultra thin films with good conformality and step coverage. The PVD process is a line of sight process and the directional nature of the deposition process results in poor sidewa ll coverage inside the trenches and vias. To solve this sidewall coverage problem, i onized PVD (I PVD) is introduced. I PVD employs high density plasma to

PAGE 18

18 ionize the metal ions during deposition. These ions are then perpendicularly directed to the wafer s urface via magnets to obtain better sidewall coverage [10] W hile I PVD leads to improved sidewall coverage of diffusion barrier film, future generations o f diffusion barrier films require deposition techniques that have a significantly better step coverage than I PVD. 1.2.3 Hypot hesis Since PVD has a serious limitation s in achieving conformal coverage on high aspect ratio contacts and via holes c hemical v apor d eposition (CVD) has received recent attention as a deposition method for TaN If the deposition is performed in the surf ace reaction limited regime and the temperature is controlled, the deposition rate should be uniform at all surfaces. There were two different approaches to CVD of TaN depending on the metal precursor One approach is to use a Ta halide s such as TaCl 5 or TaBr 5 [11, 12] TaN CVD using a halide source however, require s a high deposition temperature to obtain low film resistivity and a high deposition temperature is not preferable for b ack e nd of l ine (BEOL) processing. In addition, Cl or B r incorporation in the diffusion barrier film is known to reduce the adhesion strength and enhance electro migration [13, 14] These are major concerns for long term device reliability The other approach for TaN CVD is using metalorganic (MO) Ta sources. The growth temperature can be signific antly lowered by using MO sources compared to halide sources. Also, halide incorporation in the film is not an issue. Although the ITRS road m ap calls for CVD of barriers ALD will likely emerge as the dominant growth method for diffusion barrier deposi tion in the longer term. ALD shows superior conformality and accurate thickness control compared to other deposition methods. As the barrier thickness decreases standard CVD will approach its usability limit due to difficulties in deposition rate and the thickness control The highly conformal, ultra thin barriers by ALD

PAGE 19

19 will be required to minimize the impact of barrier films on the resistance per unit length in Cu interconnects [15] There have been reports on TaN ALD using halide source (TaCl 5 TaBr 5 ) and NH 3 [16, 17] The deposited films were mostly polycrystalline Ta 3 N 5 Unfortunately, the grain boundaries in polycrystalline film are fast diffusion pathways for Cu. H igh resistivity (> 10 4 cm) in these films was also an issue. As an alternate approach, the MO source such as penta kis(dimethylamido) tantalum (PDMAT) [18] and tert butylimido tris(diethylamido) tantalum (TBTDET) ha ve been used as a precursor for TaN ALD [19, 20] A fcc NaCl type nanocrystalline structure of TaN was obtained by using PDMAT as the Ta source [18] Thermal ALD TaN films deposited with TBTDET showed the lower material density and higher resistivity than plasma enhanced (PE) ALD TaN. Low resistivity of PE ALD TaN was attributed to formation of a Ta C bond s [17] In this study, TaN thin films were grown with ALD for conformal and ultra thin diffusion barrier films. TBTDET was used as a Ta source due to no halide incorporation. To meet the tar get barrier thickness in the ITRS road map, t he effective minimum thickness of ALD TaN barrier to block Cu diffusion was determined by annealing of various ultra thin ALD TaN films at 500 C 30 min under N 2 In addition to the thin film growth study of A LD TaN using TBTDET the effect of air exposure on the diffusion barrier quality was also investigated. Several factors including impurities, thickness, stress, and defects have been suggested to affect the barrier performance [21 24] Generally, O, C, and H are the main impurities in ultra thin films deposited by ALD because they are prepared from using MO precursors. These impurities ar e known to change the st ructure and properties of TaN thin films [17, 25, 26] There is a report of oxygen positively influencing the performance of TiN, presumably by oxygen stuffing in

PAGE 20

20 diffusion pathway [27] To understand the effect of air exposure on the diffusion barrier property, specifically thermal stability Cu/TaN/Si stacks were prepared with and without exposure to air after deposition of the TaN/Si. Both of stacks were annealed at the same temperature, and compared in different categories of thermal stability test.

PAGE 21

21 Figure 1 1. Schematic of novel design LED using nanorod. Figure 1 2. Schematic of planar LED structure.

PAGE 22

22 Figure 1 3. SEM images of Cu/Si stacks with and without using TaN diffusion barrier (A) Cu/Si st ack after annealing at 500 C, 30 min under N 2 (B) Cu/TaN/Si stack after annealing at 500 C, 30 min under N 2

PAGE 23

23 CHAPTER 2 LITERATURE REVIEW 2.1 Development of GaN GaN was first synthesized in 1928 and it was proclaimed n [28] GaN has been an attractive material for high temperature operating device b y its chemical stability at elevated temperatures combined with its wide band gap. GaN crystal s were first grown in 1938 by flowing NH 3 through gallium at high temperatures [29] The photoluminescence of GaN grown by the same method was measured for the first time in 1959 [30] The first GaN film was deposited on sapphire substrate by CVD in 1969 [31] Due to its large background n type carrier concentration, the lack of suitable substrate, and difficulty of p type doping, however, little attention was paid to GaN for use as a semiconductor. GaN regained interest largely due to its potential as a blue light emitte r. Pr evious reports claimed to obtain high quality GaN with room temperature background carrier concentration as low as n ~ 4 10 16 cm 3 Improvement in the structural quality of GaN was reported in 1983 by using low temperature (LT) grown AlN buffer la yer [32] Others reported improved surface morphology as well as electrical a nd optical properties of GaN using LT AlN buffer layer in 1986 [33] To fabricate GaN based devices, electronic properties such as carrier concentration and mobility must be controlled. The doping of wide ba nd gap materials is known to be difficult in part due to the high probability of forming native defects Intrinsic GaN grows unintentionally in n type with a direct wide (3.4 eV) bandgap. The long held belief is that the native n type nature of GaN is du e to the formation of N vacanc ies which exhibit donor behavior. High conductivity p type GaN was not achieved until 1989. It was observed that p type conductivity drastically improved after low energy electron beam irradiation (LEEBI). The first

PAGE 24

24 UV LED was fabricated by u tilizing LEEBI treated Mg doped p type GaN and intrinsic n type GaN in 1989 [34] Nakamura later established that a simple annealing in an inert or vacuum environment wi ll improve p type conductivity similar to the LEEBI treatment [35] The reason for the improvement in p type conductivity caused by LEEBI or annealing was not clear then, but it was later discovered that an Mg H neutral specie s prevents Mg activation [36] GaN based Field Effect Transistors (FET) and Heterojunction Bipolar Transistors (HBT) were also fabricated in the early 1990s, enabled by improvements in crystal quality and conductivity [37] The threading dislocation density of these transistors, however, was about 10 9 to 10 10 c m 2 or about 10 6 times higher than typical semiconductors. Although devices could be fabricated with the high density of dislocations, their long term reliability was questioned. In 1994, a significant reduction of the dislocation density was achieved by adopting lateral epitaxial overgrowth that used either a SiO 2 or Si 3 N 4 mask [38] Beca use threading dislocations tend to form parallel to the growth axis, vertical blocking of the threading dislocations improved crystal quality drastically (from 10 9 10 10 to 10 4 10 5 cm 2 ). GaN based devices began to be fabricated more routinely beginnin g in 1994. flow reactor design significantly improved crystal quality [39] In this process, inert or hydrogen gas is injected vertically onto the substrate to modulate local concentrations of precursors by reducing the boundary layer t hickness. Following these improvements Nakamura reported an InGaN based 1 cd LED in 1994, a 10 cd LED in 1995 a laser diode in 1996 and continuous wave (CW) lasing at room temperature in 1997 [39 43] Nichia Corporation reported an estimated 10,000 hr lifetime of blue laser diodes in 1998 [44] Research related to GaN has continued as a result of the major breakthroughs made during the 80s and 90s. The lack of a suitable substrate material continues to create problems

PAGE 25

25 achieved due to cracking and bowing problems. Also the heat generated from LED devices due to the low thermal conductivity of the sapphire substrate is still problematic Undoub tedly, the integration of GaN technology with Si would be another important step for GaN based device and lower unit device fabrication cost. 2.2 Application of GaN in LE D After the first GaN based LED was reported in 1971, there were many efforts to fabricate similar structures [45] These devices, however, were not conventional p n junction LEDs due to the difficulty of p type doping of GaN. Rather, these were metal insulator n type semiconductor (MIN) structure s where heavily compensated in sulating GaN was grown on uncompensated n type GaN. As shown in Fig. 2 1, a side contact to the n type GaN was fabricated due to the difficult y of chemical etching of GaN [46] In this MIN structure light emission was caused by the impact of hot electrons with ionized luminescence centers in the compensated GaN region [47] A blue green GaN numeric display was demonstrated by selective deposition of metallic electrodes with MIN electroluminescent diode [48] As mentioned earlier, the first p n junction GaN based LED was fabricated by u tilizing LEEBI treated Mg doped p type GaN and intrinsic n type GaN [49] 3 m thick undoped n type GaN was deposited using trimethylgallium (TMG) and NH 3 on an AlN buffer layer. 0.5 m thick p type GaN was deposited on the n type GaN followed by LEEBI t reatment. The LED showed near band edge UV emission (~370 nm) of photo electroluminescence (PL) spectrum. The blue p n junction LED was developed by co flow metal organic chemical vapor deposition (MOCVD) using TMG, NH 3 SiH 4 and Cp 2 Mg [50] The peak wavelength and FWHM from PL were 430 nm and 55 nm, respectively. The output power was 42 W and the external quantum efficiency was 0.18%. To improve the output power and quantum efficiency, a

PAGE 26

26 Si:In 0.2 Ga 0.8 N active layer was grown between n and p type GaN [51] The output power was improved to 125 W, and the external quantum efficiency was 0.22%. A h ighly efficient blue LED with 5.4% quantum efficiency was fabricated by an InGaN/AlGaN double heterostructure where Zn and Si were co doped into the active layer [52] The output was as high as 3 mW, and the peak wave length was 450 nm. The violet emission from In x Ga 1 x N ( x <2) band to band recombination was insufficient for display application s Thus, Zn and Si were incorporated to form donor and acceptor levels 20 30 meV below the conduction band, and 0.4 0.5 eV above the valence ba nd, respectively. This enhanced the radiative recombination, and reduce d the energy difference between electron and hole states. Green and yellow LEDs were fabricated by utilizing single quantum well (SQW) structure s with an increas ed indium fraction in the InGaN active layer. The active layer was deposited between 50 nm th ick Si doped In 0.05 Ga 0.95 N and 100 nm thick Mg doped Al 0.1 Ga 0.9 N layer as shown in Fig. 2 2. The peak wavelength of blue, green, and yellow LEDs were 450, 525 and 590 nm respectively. The output power and external quantum efficiency were 4 mW and 7.3% fo r blue LED s 1.5 mW and 2.1% for green LED s and 0.5 mW and 1.2% for yellow LED s The conventional green GaP LED shows a peak wavelength of 555 nm with an output of 0.04 mW, and the green AlInGaP LED shows a peak wavelength of 570 nm with output of 0.4 mW [53] Thus, InGaN SQW LEDs have much higher output power than conventional green LEDs. The w hite LED was introduced by Nakamura et al. an d is now being commercialized by Nichia. It is composed of a blue LED which is packaged with yttrium aluminum garnet (YAG) phosphor to convert blue light into white as shown in Fig. 2 3 [7] This white LED shows an external quantum efficiency of 3.5%, and 1 mW output. White LED s are an attractive alternative

PAGE 27

27 to the conventional light bulb due to their longer lifetime and better conversion efficiency of electric power to visible light. LED s have many potential application s from light bulb s to display panel s LED flat displays consist ing of thousands of LED are suitable for indoor and outdoor applicatio ns due to the high output intensity of the LED [54] The commercial Toshiba module is 128 mm 256 mm, and contains 2048 units of LED The m ore modern 1.28 m 2.56 m panel contains 204 800 LED units. The indoor application model from Sharp contains 614 400 LED units in a single panel. The market for LED using GaN and other III V compounds continues to g row. After 10.8% growth in 2008, the LED market is expected to grow 2.9% in 2009 due to its application in LCD TV backlight unit even though the overall semiconductor marke t is set to decline 9.4% [55] The number of applications of GaN based LED s is expec ted to keep growing, and more r esearch in this field is required 2.3 Atomic Layer Deposition of GaN Though ALD of GaN has the potential to be used in many optoelectronic applications, v ery little work has been reported on ALD of GaN. The first report on ALD of GaN by Khan et al. employed triethylgallium (TEG) as a Ga source and NH 3 as a N source [56] GaN film was deposited on a thin AlN buffer layer on a sapphire substrate which was grown at 550 C using triethylaluminum (TEA) and NH 3 The growth temperatures ranged from 450 to 900 C. The s elf saturation limit for the exposure time of TEG was found at the growth temperature of 900 C. Single crystal GaN was grown at 450 C which is well below 700 C, required for conventi onal MO CVD. The band edge photoluminescence (PL) intensity and line width of ALD film grown at 900 C were as good as MO CVD grown film at 1040 C. Even though good optical quality GaN was grown at lower temperature s the ALD characteristics such as pro cess temperature window were not investigated. Moreover, the self saturation limit was

PAGE 28

28 found at a relatively high temperature (900 C), which is believed to be out of the process temperature window. Karam et al. reported ALD GaN using Trimethylgallium (TM G) as a Ga source and NH 3 as a N source [57] This work employed unique rotating susceptor ALD reactor Similar to previous GaN report s GaN is deposited on a thin AlN buffer layer on a sapphire substrate. Single crystal GaN was grown at 550 C and the PL spectra of this GaN film show ed a strong emission peak of 3.42 eV. Though good optical and structural quality films were grown at relatively low temperature, ALD characteristics were not investigated. It is known that t he relatively unstabl e MO sources show a narrow process temperature window in ALD of most materials. To resolve this issue, Ga chloride sources such as GaCl and GaCl 3 have been used with NH 3 for the ALD growth of GaN. The ALD growth of GaN using GaCl as Ga source was report ed by Koukitu et al [58] The ALD GaN film was deposited on GaN buffer layer, which is grown on a GaAs substrate by MO vapor phase epitaxy (MOVPE). GaCl was obtained by the reaction of metallic Ga and HCl at 780 C ALD characteristics such as self s aturation limit and process temperature window were investigated in this study. The s elf saturation limit for GaCl exposure time was found over 400 C of growth temperature. Within the bounds of the self saturation limit ALD GaN is shown to have a cubic structure with [001] direction perpendicular to GaAs(001) substrate. Tsuchiya et al reported ALD of GaN using GaCl 3 as Ga source due to its relatively high vapor pressure for [59] A t hin GaAs buffer layer grown on GaAs was used as a substrate. The y failed to find the self saturation limit of GaCl 3 exposure time. From X ray diffraction (XRD) measurement s it was shown that single crystal GaN was grown on GaAs(111). From reflection high energy electron diffraction (RHEED), GaN was found to have the hexagonal structure,

PAGE 29

29 specifically, wurtzite structure ALD of GaN using chloride Ga sources such as GaCl and GaCl 3 showed ALD characteristics, and fairly good structural quality. However, Cl incorporation in the film, which is deleterious to device perfo rmance, was not considered even though this is one of the most important issue s with using halide metal source s 2.4 Doping of GaN 2.4.1 n type Doping of GaN There ha ve been many reports of achiev ing n type doping of GaN by introducing group IV elements such as S i and O The n type doping of GaN using SiH 4 as a dopant was reported by various groups [60, 61] The n type GaN was grown by MOCVD using TMG as a Ga source, NH 3 as a N source and SiH 4 as an n type dopant A GaN buffer layer deposited on c sapphire was used as a substrate. The carrier concentration determined by room temperature Hall measurement was found to increase up to 2 10 19 cm 3 by increasing the flow rate of SiH 4 whereas the mobility of the carrier decr eased from 600 to 100 cm 2 V 1 s 1 under the same increase. The activation energy of n GaN from SiH 4 was reported between 12 and 17 meV [62] The ratio of free electron to incorporated Si atom s was in the range of 0.5 0.1. Si 2 H 6 was employed as an alternat ive n type dopant due to its higher reactivity [63] The n type GaN layer was grown by MOCVD using TMG as a Ga source and NH 3 as a N source. The carrier concentration was increased up to 4 10 19 cm 3 with increasing Si concentration. From the room temperature Hall measurement, t he efficiency of incorporation of carriers is kept constant with the slope of 1 for carrier concentrations from 1 10 17 to 4 10 19 cm 3 The carrier mobility d ecreases from 450 to 100 cm 2 V 1 s 1 as carrier concentration increases The resultant carrier concentration at a given Si concentration was found to be independent of growth temperature in the range of 980 1040 C

PAGE 30

30 Oxygen is of particular interest as a possible alternative n type dopant due to its position next to N in the periodic table and its suspected role as a background impurity in as grown GaN. Several groups have shown O to act as a donor for GaN [64 67] To employ O as an n type dopant, O 16 ions were implanted to MOCVD grown GaN. T h e resulting carrier concentration w as ~1 10 17 cm 3 and the mobility was 100 cm 2 V 1 s 1 The activation energy was ~29 meV, and the activation efficiency was poor. 2.4.2 p ty pe Doping of GaN There ha ve been many efforts aimed at p type doping of GaN by introducing group II and IV elements. The most investigated dopant for p type GaN is Mg. Most of reports on Mg doping in GaN employed MOCVD of GaN on buffer layered sapphire with the co flow of a Mg source. However it is known that one of the biggest roadblocks to achieve p type doping of GaN is the compensation of electrons by Mg or Mg containing complexes. To overcome this compensation issue, there have been many efforts to activate Mg acceptors. As mentioned in C hapter 2.1, Amano et al. showed the first p type conductivity by activating Mg doped GaN with LEEBI After LEEBI activation, the hole concentration was increased to 2 10 16 cm 3 and the mobility was 8 cm 2 V 1 s 1 A lso the high resistivity due to compensation wa s dramatically cm. Nakamura et al. confirmed the work of Amano et al. using the same activation method. Hole concentration s as high as 3 10 8 cm 3 and mobilit ies as large as 15 cm 2 V 1 s 1 were obtained after LEEBI treatment even though the initial hole concentration was 2 10 15 cm 3 Nakamura et al. and Gtz et al. demonstrated that the activation of Mg acceptors can also be achieved by post growth thermal annealing in the temperature range of 600 775 C [35, 68] The p type GaN grown by co flow MOCVD was annealed by rapid thermal annealing (RTA) for

PAGE 31

31 5 min. After thermal annealing, th e hole concentration was increased up to 2.3 10 18 cm 3 And the mobility and resistivity were decrea sed to 10 cm 2 V 1 s 1 cm, respectively. Brandt et al. investigated the effect of post growth annealing with hydrogenation using H plasma above 500 C [69] The hole concentration of p GaN activated by post growth annealing was reduced by one order of magnitude after the hydrogenation process. Based on this observation, it was suggested that H plays a crucial role in passivating the Mg acceptor, and the H passivation is mitigated by thermal annealing. For this re ason Neugebauer et al. theoretically investigated Mg H complexes and concluded that H does not form a bond to Mg atom, but prefers the anti bonding site of one of the N neighbors [70] Thus, the Mg H complex has a modified H N bond, and the dissociation barrier for this bond was estimated to be 1.5 eV This H N bond compensates the electrons, and this compensation can be reduced by breaking Mg H or H N bond with a thermal annealing process. It was reported by Myoung et al. that molecular beam epitaxy (MBE) grown GaN doped with Mg is p type without a thermal activation process [71] At room temperature, Mg doped GaN grown by MBE showed the free hole concentration above 10 18 cm 3 with a mobility of ~10 cm 2 V 1 s 1 without post g rowth thermal annealing. This is consistent with results of previous reports on compensation by H complexes. H is absent during MBE process es and this prevents the formation of H complexes in the p GaN film. Thus, the relatively high hole concentration of p GaN was obtained without thermal annealing activation. Previously mentioned p GaN activated by either LEEBI or thermal annealing showed a binding energy of the Mg acceptor in the range of 230 260 meV. Th is binding energy was found to be temperature dependent. Though Mg has been a reasonable p type dopant for GaN, there is still the question of whether any other acce ptors with smaller binding energies exist, and whether they are suitable

PAGE 32

32 dopants that can be incorporated at high concentration. Zn, C d, and Be also belong to group II like Mg, and can substitute for Ga in GaN as a p type dopant. Bergman et al. reported the p type doping of GaN using Zn and Cd as dopants, and showed the identity of the active recombination centers from transient lumine scent spectra and time resolved luminescence measurements [72] The dominating recombination centers created by Zn and Cd doping in GaN were identified as substitutional acceptors, Zn Ga and Cd Ga The binding energies of Zn and Cd acceptors were measured as 0.34 and 0.55 eV, resp ectively. These results indicate that the acceptor levels by Zn and Cd incorporation in GaN are deeper than Mg acceptor level, and therefore less efficient in creating free holes at room temperature. Neugebauer and Van de Walle reported a comprehensive i nvestigation of acceptor impurities in GaN based on first principle total energy calculations [73] From this calculation, none of the candidates (Li, Na, K, Be, Zn and Ca) exhibited better characte ristics than Mg. Only Be is expected to be a potential alternative dopant though it may have solubility limitation s and compensation issue s by Be interstitial donors. Brandt et al. reported the growth of p type GaN by GaN MBE with co doping of Be and O [74] Be is supplied by evaporat ion from an effusion cell and O is supplied by H 2 O in the background pressure A concentration of 5 10 20 cm 3 Be and O atoms was incorporated into GaN film at 650 C. Hall measurements showed p type conductivity with a hole concentration o f 10 18 cm 3 and a hole mobility of 150 cm 2 V 1 s 1 Zopler et al. achieved p type doping of GaN by co implanting Ca with P at concentrations of 10 19 cm 3 [75] The measured binding energy of holes was 169 meV for Ca The measured hole mobility was 7 cm 2 V 1 s 1 These properties s howed that Ca can be used as a p dopan t for GaN The possible acceptors for various III V materials are summarized in Table 2.1 with binding energies [76]

PAGE 33

33 2.5 Challenges in I nterconnects 2.5.1 RC Delay in Interconnects There have been demand s on increase d transmission speed as the individual feature size s of device s in integrated circuit s (IC) shrinks One of the biggest roadblocks to achiev ing maximum signal transmission in emergi ng chip and system architecture is to minimize the resistance times capacitance ( RC ) time delay. RC delay has replaced intrinsic device delay as the speed limiter in sub quarter micron device nodes, primarily due to the higher RC time constant in the narr ower, more closely spaced interconnect lines. The time delay in an interconnect circuit depends on two factors ; the resistance of the conductor and the capacitance of the dielectric material. For a MOS circuit, the RC delay is defined in terms of the cir cuit response [77] ( 2 1) where the he output voltage of the circuit at time t is V o ut (t), maximum output voltage is V out (max), the resistance of the metal line is R and the total capacitance of the circuit is C The resistance of the interconnect line is defined in terms of its length L width W thickness T m and resistivity ( 2 2) The capacitance between the silicon substrate and the metal line (metal to ground), C (metal ground) is given by Equation 2 3. ( 2 3)

PAGE 34

34 where and T d are the dielectric constant of the dielectr ic material and thickness of the dielectric line respectively, and 0 is the permittivity of free space. The capacitance between two metal lines ( C(metal metal) ) separated by distance D is given by Equation 2 4 [78] ( 2 4) The total capacitance, C can then be calculated from Equations 2 3 and 2 4. ( 2 5) K is a constant that depends on the frin ging field effect. The RC delay is calculated from Equations 2 2 and 2 5. ( 2 6) As the critical device dimension shrinks, C(metal metal) dominates the overall capacitance C [79] The RC delay can then be written as Equ ation 2 7. (2 7) For local interconnects the shrinking of the transistor leads to a corresponding decrease in L W and D Therefore, the actual delay in the local interconnect does not increase with device scaling. However the chip area increases with each successive technology node. This leads to an increase in the length of global interconnects and a corresponding increase in the RC time delay as shown in Fig. 2 4 [80] 2.5.2 Electro migration in Interconnects Electro migration is the phenomenon caused by the movement of metal atoms to the dir ection of electron flow in high current density interconnects. Metal atoms in interconnects are

PAGE 35

35 known to have higher mobility under higher operating temperature which is popular in current microprocessors. When high energy electrons transfer their moment um to these mobile metal atoms, there is a movement of metal atoms in the direction of electron flow. This atom movement leads to formation of voids at the cat hode and hillocks at the anode as shown in Fig 2 5 [81] Electro migration, which causes failure of the circuit, is one of the most impo rtant concerns for the reliability of interconnect s As the interconnect dimensions shrink down the current density in interconnects increases. Thus, the scaling of interconnects requires higher electr o migration resistance in interconnect materials. For Al based metallization, e lectro migration becomes a major concern Figure 2 6 shows that the mean time to failure for Al interconnect s dramatical ly decreases if the current density in the interconnect exceeds 3 MA/cm 2 [82] 2.5.3 Diffusion of Cu into Si To solve primary challenges in metallization such as RC delay and electro migration, Cu has been used as a substitute for Al. The resistivity of pure Cu ranges from 1.7 to 2.0 cm, whic h is much lower than Al. This lowered resistivity attributes to the decrease of RC dealy. Also, Cu is known to have higher resistance to electro migration. For these reasons, Cu is currently used as interconnect material in semiconductor industries. Ho wever, there are some issues with the use of Cu as a interconnect material. The primary issue is the diffusion of Cu into Si. Cu is known to diffuse extremely fast in Si. The diffusion coefficient of Cu in Si at 500 C is 2.0 10 5 cm 2 /sec [83] Furthermore, Cu is mobile in Si even after annealing. And, Cu can travel significant distances within Si wafer d even at room temperature. The interaction of Cu with Si depends on the doping of the Si substrate. I n n type Si, Cu forms neutral or negatively

PAGE 36

36 charged precipitates. There is no barrier or electrostatic repulsion between Cu precipitates and interstitial Cu since interstitial Cu is positively charged in Si. Thus, Cu easily forms Cu x Si precipitates in n type silicon. This Cu x Si can act as recombination centers for minority carriers and significantly lower minority carrier lifetime s [83] In p type Si, Cu forms positively charged precipitates [84] There is electrostatic repulsion between the positively charged Cu precipitates and positively charged interstitial Cu, which acts as a barrier in nucleation and precipitation of copper in p type Si. If ther e is no native oxide on the surface of p type Si wafer, Cu out diffuses from the bulk to the surface. Since segregated Cu on the surface can be easily etched by a H 2 O 2 and HF mixture, it is inferred that the Cu does not form Cu 3 Si precipitates. For n typ e Si, out diffusion occurs only if Si is heated to 400 C. The primary reason why Cu is extremely undesirable in Si is that it forms deep level traps in the band gap of Si as shown in Fig. 2 7. If Cu exists in the triplet state, it forms three acceptor le vels in the Si band gap at 0.24, 0.37 and 0.52 eV [77] Since Cu diffuses extremely fast in Si, detrimental effects of Cu can spread throughout the device. Cu is also detrimental if it diffuses into the dielectric film. Cu in dielectrics decreases the breakdown voltage of the dielectric and also increases the parasitic capacitance. Thus, benefits from low dielectric film are negated. 2.6 TaN as Diffusion Barrier 2.6.1 CVD of TaN TaN diffusion barrier s ha ve been deposited by CVD to avoid the side wall deposition problem s of PVD. In CVD growth of TaN, both halide and metal organic Ta sources have been employed Hal ide sources such as TaCl 5 TaF 5 and Ta Br 5 have been investigated as CVD precursors for TaN. CVD TaN films using TaCl 5 and NH 3 were grown at the low deposition temperature of 350 C [11] However, the film was not suitable for diffusion barrier application s because of its high resistivity (> 10,000 cm) and high Cl contamination (~4.5 at. %).

PAGE 37

37 Moreover, Cl contamination is difficult to reduce s ince the Ta Cl bond is very strong. CVD TaN films using TaF 5 and NH 3 were reported to have relatively low resistivity of 1650 cm at the growth temperature of 350 C [11] However, the film has considerable F contamination ( ~ 4.0 at. %) and poor step coverage. The poor ste p coverage can be explained by considering the high growth rate of deposition. Because a high growth temperature is required to obtain suitable diffusion barrier properties, thermal CVD using TaCl 5 or TaF 5 has not been investigated further TaBr 5 has be en studied as a precursor for CVD of TaN using H 2 and NH 3 as c o reactants. It is believed that halogen contamination is lowered by smaller diffusivity of TaBr 5 into the film due to larger atom size of Br than Cl and F. Moreover, lower deposition temperat ure is required for the reaction of TaBr 5 and NH 3 due to lower bonding energy of Ta Br in TaBr 5 Chen et al reported CVD deposition of TaN films at growth temperature s of 3 50 to 500 C using TaBr 5 H 2 and NH 3 [12] The films were N rich with N:Ta ratio s varying from 1.75:1 to 1.87:1 for growth temperature s between 350 and 500 C The N rich film deposited with TaBr 5 indicates that the film is possibly Ta 3 N 5 instead of TaN. The Br content in the film decreased with i ncrease d deposition temperature. The Br content was highest (~5 at. %) at the growth condition of 350 C, while it was be low the detection limit of AES for deposition at the growth temperature of 500 C. Film resistivity also showed a strong dependence on deposition temperature with the lowest resistivity of 5040 cm at 500 C. To evaluate the diffusion barrier performan ce of TaN films, Kaloyeros et al. employed extensive barrier tests on Ta N films deposited at 425 C [85] The Cu/TaN/Si stack was annealed between 450 and 650 C to determine the barrier failure temperature. From post anneal XRD measurement it was confirmed that TaN barrier film successfully prevent e d bulk diffusion of Cu up to 650 C a nnealing temperature

PAGE 38

38 The other approach for TaN CVD is to employ m etal o rganic Ta sources such as Ta(NMe 2 ) 5 ( pentakis(dimethylamido)tantalum, PDM AT) Ta(NEt 2 ) 5 ( pentakis(diethylamido) tantalum, PDEAT ) and (NEt 2 ) 3 Ta=NB u ( tertbutylimidotris (diethylamido) tantalum, TBTDET). For t hese metal organic Ta sources detrimental halide residues are not an issue. There was, however, high C incorporation in MOCVD TaN films which lead to high electronic resistivity up to ~10 6 cm in the case of the amorphous C phase. TaN from PDEAT showed high electronic resistivity up to 6 10 5 cm with high C content (~30 at %) under sin gle source deposition [86] The addition of NH 3 as an ad ditional nitrogen source, however, decrease s the electronic resistivity down to 70 00 cm while lower ing C content from 30 to 1 at %. To obtain a reasonable range of resistivity (<1000 cm), t he growth temperature should be higher than 600 C Also, TaN from PDEAT and NH 3 exhibited better Cu diffusion barrier performance compared to single source deposition using only PDEAT. However, the step coverage of the film decreased from 80% to 56% with addition of NH 3 which is ascribed to switching the growth condition to the mass transfer limited regi me TaN deposited with PDMAT showed the insulating Ta 3 N 5 phase resulting in high resistivity (>10 6 cm) [87] The film also had the high C incorporation even with addition of NH3 TaN using TBTDET as Ta source showed a relatively low resistivity (920 cm) at 650 C [88] The deposition rate was almo st independent of the growth temperature (450 to 650 C), indicating growth in the mass transfer controlled regime. From XPS measurement, TaN film has 10 at. % of C incorporation A c omparison study of diffusion barrier performance between PVD TaN and CVD TaN from TBTDET showed the barrier failure temperature of 650 C for PVD and 600 C f or CVD. This exhibits that the PVD TaN possesses better diffusion barrier propert ies than CVD TaN due to its higher film density.

PAGE 39

39 2.6.2 ALD of TaN ALD is a self limiting growt h method consisting of alternative dosing of chemical species to control the target material in layer by layer manner while CVD delivers all required reactants to the reactor simultaneously. Each precursor is exposed to the substrate sequentially which r esults in uniform layer adsorption of precursor on the substrate by chemisorptions. The most important characteristic of ALD is obtained from this step, which is self saturation. The self saturation limits further adsorption of precursors by passivating the available adsorption sites after the saturation coverage Due to this self saturation nature, ALD films show much better conformality and uniformity than CVD films. These characteristics of ALD films are essential for diffusion barrier films to minim interconnect. Also, relatively lower growth temperature s make ALD a preferred deposition method for diffusion barrier s [26] For ALD of TaN, two types of Ta precursors such as metal halide and met al organic were employed TaCl 5 is the most commonly used Ta halide precursor. The first report on ALD of TaN employed TaCl 5 as a Ta source w ith N H 3 as a N source [89, 90] The film growth rate wa s 0.22 0.24 / cycle and the temperature window was in the range of 300 500 C. Th e films deposited with the TaCl 5 and NH 3 we re N rich and polycrystalline. From XRD measurement s the films deposited at 400 C were tet ragonal Ta 3 N 5 phase. The Cl incorporation in the f ilm was less than 5 at. %. Cl impurity level in the film decreased with increase d deposition temperature. The resistivity of the film varie d from 0.5 10 6 to 200 10 6 cm. The high resistivity of the film came from the presence of the non conducting Ta 3 N 5 phase. There has been a report o f p lasma assisted ALD (PAALD) of TaN films using TaCl 5 and H 2 + N 2 plasma [91] N content in the film varied as Ta 2 N, TaN, and Ta 3 N 5 with the partial pressure of N 2 in the plasma. The ALD proces s window was not found over the entire growth

PAGE 40

40 temperature range. The resistivity of TaN films was much lower compared to thermal ALD films, vary ing between 350 and 400 cm. Cl content in the f ilm was less than 3 at. %, and H incorporation was as high a s 11 at. % Similar to thermal ALD of TaN using TaCl 5 PAALD also showed a decrease in Cl level with the increase of deposition temperature. Overall, PAALD showed good electronic quality films suitable for diffusion barrier application s However, resulti ng film s w ere not in the amorphous phase, and the Cl incorporation issue was not completely solved. To solve the halogen incorporation issue of halide precursor m etal orga nic precursors such as TBTDET and P DMAT have been used in ALD of TaN. As mentioned in the previous section, CVD TaN films using these metal organic precursors showed high C impurities and high resistivity. The ALD process is expected to lower C impurity levels at even lower deposition temperature s TBTDET was reported to be employed a s a Ta source with NH 3 as a N source for thermal ALD of TaN [92] Deposited TaN films were in amorphous phase between 150 and 300 C of growth temperature The films deposited we r e Ta rich and the N/ Ta ratio was 0.73. As expected, C impurity level s of deposited films w ere less than 5 at. %. The resistivity of the film was very high (> 10 6 cm at 260 C) despite high Ta content in the film. The r esistivity of the film decreased down to 15000 cm w hen the deposition temperature wa s increased to 450 C. Choi et al. reported the temperature window between 200 and 250 C for the ALD proces s using TBTDET and NH 3 [93] PAALD of TaN films using TBTDET with H radical has been reported [94] The H radical was used as a reducing agent for TBTDET. Unlike thermal ALD, TBTDET was employed as a single source precursor. N atoms attached to Ta by single bonds we re removed by H radicals,

PAGE 41

41 whi le double bonded N atoms to Ta remain, leading to stoichimetric TaN. R esulting TaN films we re Ta rich with N/ Ta ratio of 0.6 The C impurity level, however, was very high as 20 at. %. The films were polycrystalline and had low resistivity of 1000 cm. Overall, the plasma process ing improved the electronic propert ies of the films The deposited film however, was in the p olycrystalline phase which is not preferred for an effective diffusion barrier Also, diffusion barrier performance of ALD TaN from TBTDET was not investigated. PDMAT was another metal organic Ta source for T aN ALD Thermal ALD of TaN has been reported using P DMAT and NH 3 [95 98] The ALD t emperature window was found between 225 and 300 C wit h a constant growth rate of 0.3 /cycle The composition and microstructure of deposited films showed strong dependence on substrate The films deposited on Si, SiO 2 and Cu w ere N rich with N / Ta ratio of 2 at 275 C growth temperature However, the film s deposited on Ta we re Ta rich with N / Ta ratio of 0.5 The C level in the film was as low as 2 at. %. The films deposited we re in nanocrystalline phase with average grain size of 4 nm. The diffusion barrier performance was evaluated in the scope of ther mal stability. Cu diffusion was prevented at 750 C of annealing temperature for 30 min by 10 nm thick TaN diffusion barrier. Even though ALD TaN film from PDMAT showed suitable diffusion barrier properties, the electronic resistivity of the film was too high to be measured by standard four point probe To solve the high resistivity issue, PAALD of TaN using PDMAT has been investigated. Kim et al. investigated the effect of plasma on the deposition of TaN x film [99] The ALD temperature window was found between 225 and 300 C of growth temperature. The de posited film is Ta rich, with N/ Ta ratio in the range 0.7 to 0.8. The C impurity level in the film varied from 2 to 14 at. % with N 2 flow rate. The resistivity of the film was as low as 350 cm at 250 C of deposition temperature. Diffusion barrier testing s how ed that Cu diffusion was

PAGE 42

42 prevented at 750 C of annealing temperature for 30 min by 3 nm thick TaN diffusion barrier. Overall, PAALD of TaN from PDMAT provides improved diffusion ba rrier properties in the scope of electronic resistivity as well as minimum thickness to prevent Cu diffusion. To summarize, both halide and metal organic Ta precursors have been used to deposit ALD TaN films. The biggest challenge for halide Ta precurso r s was the incorporation of detrimental halogen atom s in the film at lower growth temperature s Moreover, deposited film was in the insulating Ta 3 N 5 phase instead of the conducting TaN phase The use of metal organic Ta precursors could solve halogen inc orporation issue s However, other challenges such as the C incorporation and high film resistivity by Ta 3 N 5 phase were brought up at low growth temperature Thus, plasma processing was introduced to obtain low resistivity TaN film with a favorable N / Ta r atio. 2.7 Cu Metallization 2.7.1 Processing of Cu Interconnect As mentioned earlier, Cu has been employed as an interconnect material to solve RC delay and electro migration problems. The most prevailing process for Cu interconnect integrity is dual damascene pro cessing. The notation of dual damascene is used to describe a patterning process of metallization. For Al metallization, Al is deposited first, followed by patterning with lithography and etching steps. The SiO 2 dielectric was then deposited into the Al pattern. On other hand, for Cu metallization, the scheme of process is reversed. The dielectric film is deposited first, and then the lines and vias for the metals are patterned into dielectric using lithography and etching steps followed by the deposit ion of Cu into lines and vias. Overburdens are then removed by planarization process es such as CMP Several key process elements have been developed to apply dual damascene process to Cu metallization successfully. T o remove overburden Cu effectively, th e CMP process has been

PAGE 43

43 developed with high polishing rates, no corrosion, and high selectivity. In addition the lithography and reactive ion etching processes have been optimized for the dual damascene patterning at required scales Most of all a prope r Cu electroplating process contributes to the improved reliability of Cu interconnects from dual damascene processing. Cu electroplating has the super filling behavior by utilizing inhibitor additives [100] The super filling behavior is shown in Fig. 2 8. This super filling behavior leads to void free and seam free deposit of Cu into lines and vias As shown in Fig. 2 9, void and seam are present for sub conformal and conformal deposition, respectively. These void s and seam s are fast diffusion pathways for Cu electro migration. Thus, the fast diffusion pathways for Cu electro migration can be eliminate d by super filling. This super filling behavior can be explained by activity and kinetics of a plating inhi bitor at the top surface [101] The inhibitor has relatively low a ctivities in the hole bottom and bottom side walls which enhances the filling of hole bottom and bottom side walls. On the other hand, the inhibitor is active at the top surface, resulting in suppressed deposition rate at the top surface. This behavior o f inhibitor leaves the hole open for filling until the hole is completely filled, which results in super filling of hole 2.7.2 Microstructure of Cu Thin Film for Various Deposition Methods The microstructure of Cu thin films has attracted much interest rece ntly because the micro structural attributes such as grain size, grain size distribution, and crystallographic texture influence mechanical and physical properties of films In particular, these micro structural attributes are directly related to the anis otropic property variations in Cu. A lso the grain orientation of Cu is considered to influence the reliability of Cu. For example, the elastic modulus of Cu in ( 111 ) direction is 2.9 times large r than ( 100 ) direction [102] Additionally, the

PAGE 44

44 reaction rates for oxide and silicide formation of Cu detrimental to device performance, are faster along (100) grain surfaces than along (111) grai n surfaces [103, 104] Al so, Cu(111) grains generate (511) planes by twinning. The formation of twins is known to have beneficial effect s for Cu thin film because twin boundaries are low diffusion pathways for electro migration. Table 2 2 shows the grain size, ratio of (111)/(10 0), and percent (111) versus (100) twins for Cu films grown by different deposition methods: PVD, CVD, and electroplating Clearly the grain size and twin volume fraction s are shown to be considerably higher for the electro plated Cu films. It is known t hat twins are developed during grain growth, and are prevalent in large electroplated Cu grains. In contrast, PVD and CVD films show much fewer twins. This is believed to be originated from t he small grain sizes of PVD films and the effect of grain bound aries in these films to relieve strain. As strain energy increases in a growing grain, twins are formed to relieve the pressure. In summary, (111) orientation of Cu is preferred for better reliability due to low reaction rate for oxide and silicide formati on. Also, the t winning effect is beneficial for Cu thin film due to low er diffusion pathways for electro migration. It is demonstrated that PVD Cu has the most (111) orientation which is preferred for better reliability of Cu interconnect. It is also sh own that electroplated Cu has the most twinning of (111) orientation for better electro migration resistance. 2.8 Barrier Failure Mechanism To identify a viable diffusion barrier for Cu metallization, the fundamental understanding of diffusion failure mec hanism must be established. The dominating mass transport phenomenon in Cu/barrier integrity is the inter diffusion phenomena of Cu. It is well known that the placement of chemically different atoms in proximity leads to atomic migration to lower the

PAGE 45

45 ove rall free energy and establish equilibrium. This a tomic migration is typically driven by concentration gradient s and accelerated by thermal energy [105] Atomic migration result s in a diffusion flux, with the net flow of atoms characterized by a diffusion coefficient D This is so called 8 where C is the atomic concentration, J the atomic flux per unit area per second, and x is the distance [106] (2 8) The temperature dependence of the diffusion coefficient D is i n the form of an Arrhenius relationship as shown in Equation 2 9 where D o is the diffusion coefficient at 0 K, Q is the is the temperature in Kelvin [107] D iffusion parameters such as D o and Q are strongly dependent on the thermodynamic and kinetic properties of the defects responsible for diffusion. (2 9) In the case of the Cu/ barrier integrity, thre e mechanisms typically dominate diffusion barrier failure: (a) diffusion of Cu through bulk defects in the barrier (b) diffusion of Cu along grain boundaries in the barrier and (c) loss of barrier integrity due to a metallurgical or chemical reaction with the Cu and/or substrate [108, 109] For the failure by Cu diffusion through bulk defect s of barrier, t wo types of bulk defects c ont ribute to the diffusion ; vacancies and dislocations. Diffusion through vacancies, dislocation s and grain boundaries has significantly different rates [110] The lattice diffusion rates are proportional to the absolute melting temperature, T m of host material as shown in Equation 2 10 where A is proportionality constant dependent on many factors such as lattice structure and type of material. (2 10)

PAGE 46

46 It is found that d iffusion due to ato m vacancy exchange has the smallest A thus corresponding to the slowest diffusion. Dissociated dislocations demonstrate intermediate A corresponding to intermediate diffusion rate s Diffusion along grain boundaries exhibit s the largest A and thus ha s th e highest diffusion rate. Although Equation 2 10 is an empirical relationship, the reasonable guideline can be deduced in the selection of appropriate diffusion barrier materials, which is high melting temperature materials. Also, the dependency of diffu sion rate s on the type of diffusion pathway demonstrates that the formation of grain boundaries in diffusion barrier needs to be prevented. Based on the discussion above, the microstructure of diffusion barrier plays the critical role in the barrier perfo rmance In general, there are four different types of film microstructure: (a) single crystal, (b) polycrystalline, (c) nano crystalline (i.e. polycrystalline with grain size below 5 nm), and (d) amorphous [111] Among these microstructures, the p olycrystalline barrier tends to show the poorest barrier perf ormance due to grain boundaries throughout entire film thickness, providing an effective Cu diffusion pathway. On the other hand, a s ingle crystal barrier is the most ideal for a diffusion barrier due to the lack of diffusion pathways such as grain bounda ries, dislocations, and vacancies. It is, however, unlikely to deposit single crystal diffusion barrier due to the constraints o n material s and process es such as lattice mismatch between barrier and substrate, and thermal budget limitation. Therefore, th e diffusion barrier in amorphous phase is the most efficient for barrier application s due to fewer grain boundaries than poly or nano crystalline as a result of large grain size.

PAGE 47

47 Table 2 1. Acceptor binding energies for C, Mg, Be, Zn and Cd in variou s III V semiconductors. Taken from Ref [76] Binding Energy (meV) GaAs InP GaP GaN C 27 41.3 54 230 Be 28 41 55 250 Mg 28.8 48 59. 9 250 Zn 30.7 46.4 69.7 340 Cd 34.4 57 102.2 550 Table 2 2. Microstructure of Cu films. Taken from Ref [112] PVD Cu CVD Cu Electroplated Cu Grain size ( m) 0.4 0.9 3.8 (111)/(100) ratio 59/2 13/16 4/1 (111) twins 11% 15% 26% Randomness (111) textur ed Random Weak (111)

PAGE 48

48 Figure 2 1. Experimental configuration of a typical m i n GaN LED. Taken from Ref. [46] Figure 2 2. The structure of the high brightness green light emitting diode with 20 thick In 0.43 Ga 0.57 N quantum well as active layer. Taken from Ref. [ 53]

PAGE 49

49 Figure 2 3. Top view of a blue LED fabricated by Nichia without plastic capsule. Taken from Ref [7] Figure 2 4. Effect of feature size scaling on total capacitance. Taken from Ref [80]

PAGE 50

50 Figure 2 5. Damaged Al interconnect line showing the formation of hillock and void due to electro migration Taken from Ref [81] Figure 2 6. Electro migration lifetime for aluminum interconnects at dif ferent current densities. Taken from Ref [82]

PAGE 51

51 Figure 2 7. Band diagram of Cu impurity in silicon. A) p type Si, B) n type Si. [77] Figure 2 8. Hole fill evolution for electroplated Cu with super filling additives. Taken from Ref [100]

PAGE 52

52 Figure 2 9. Evolution of hole filling with different deposition conditions varying degrees of conformality. Taken from Ref [101]

PAGE 53

53 CHAPTER 3 EXPERIMENTAL SETUP 3.1 GaN ALD Reactor Desig n and Setup Figure 3 1 shows the overall process flow diagram (PFD) of the GaN ALD system. The bubbler, containing GaCl 3 is immersed in a constant temperature bath and heated to a pre determined t emperature. The heat bath temperature is set to a temperature to provide a sufficient vapor pressure (> 0.1 Torr for ALD) to ensure reasonable adsorption time. The N 2 carrier gas flow rate is adjusted to ensure saturation of the carrier gas with GaCl 3 fo r reproducible delivery rate. GaCl 3 was delivered to the reactor or the bypass line depending on the position of the pneumatic valves located upstream of the reactor. The gas lines between the bubbler outlet and the reactor entrance are heated ~ 10 C hig her than the bubbler temperature to prevent condensation of the precursor during delivery. The N 2 stream serves not only as a carrier gas but also as a sweeping (purging) gas in the purge step of ALD process Cylinders of N 2 and NH 3 are contained inside a gas cabinet. All carrier, purge N 2 and NH 3 gases are metered to the reactor or bypass line by mass flow controllers (UNIT 8100) paired with a flow control knob (FM304V). Figure 3 2 shows the schematic of GaN ALD reactor used in this study The reactor body is made of quartz glass. The graphite susceptor and RF coil are located in the growth zone. Each reactant and purge gas is delivered to the reactor through one of the multiple concentric quartz tub es The substrate Si or sapphire is placed on qua rtz tray. The substrate with quartz tray is loaded into the reactor using magnetic robotic arm that is operated and controlled by an Intelligent Motion System LI2. The main purpose of magnetic robotic arm is to load or unload the sample through the load lock system. Load lock system is commonly used to minimize impurities. Once the substrate is loaded into the reactor it is placed on the graphite susceptor. A

PAGE 54

54 RF generator (LEPEL T) heats up the graphite susceptor to indirectly heat the substrate to th e desired temperature. In ALD i t is important to control the flow sequence of the reactants and purge gases. T he gas manifold is equipped with three separate switching mode pneumatic valves ( Swagelok SS 4 BYV 35 C ) Each pneumatic valve controls the on and off action of the gas flow into the reactor or the bypass line, resulting in the sequential flow of reactants and purge gas Sequential exposure of each reactant and purge gas is managed by the control program encoded with National Instrument La b V IEW soft ware. A screen shot of the control program for GaN ALD process is shown in Fig 3 3 T he panel shows a typical cycle of GaN ALD process consisted with four steps. Each column with three of bright and dark circular bullets represents each cycle and the r ow corresponds to a particular pneumatic valve. The first s tep is the reactant 1 (GaCl 3 ) pulse (R1), where only GaCl 3 is dosed into the reactor, while purge N 2 and NH 3 are bypass ed to vent The second step is N 2 purge (P), where N 2 purge gas is dosed int o the reactor, and GaCl 3 and NH 3 are bypassed to vent In the third step reactant 2 (NH 3 ) is pulsed to the reactor (R2), whereas GaCl 3 and N 2 purge gas are bypass ed to the vent. After NH 3 pulse step, another N2 purge step follows which has exactly same configuration as second step (P). The duration time of each s tep and total number of cycles are c ontrolled by the values in the pulse time and total cycle tabs, respectively. The signals from this program control the on and off actions of solenoid valve s via data acquisition card (NI PCI 6514) and switching mode power supply (WRB24SX U). Then the solenoid valves control the flow of compressed air to the actuators of pneumatic valves, which control the direction of the gas either to the reactor or vent.

PAGE 55

55 3.2 Chemical Equilibrium Analysis for Growth Temperature The Ga source used in this study, GaCl 3 is a liquid at room temperature. There are several reactants, however, that can occur including disproportiona ion between GaCl 3 and GaCl. As a preliminary stu dy, the equilibrium concentration was calculated as a function of temperature. To investigate dominant gas species in growth zone, a chemical equilibrium analysis was performed The gas species in the growth zone were investigated for GaCl 3 exposure step of GaN ALD process. For the thermodynamic analysis, the constant Ga/Cl molar raio of 1/3 was applied to the Ga Cl Inert (N 2 ) system Species that were included in a complex equilibrium analysis of subsystem in the Ga C l Inert system are listed in Table 3 1. For Ga containing gas species, Ga, Ga 2 GaCl, GaCl 2 Ga 2 Cl 2 Ga 2 Cl 4 Ga 2 Cl 6 and GaCl 3 were considered. For gas phase of Cl species, Cl and Cl 2 were considered. The condensed phases were liquid and solid phases of GaCl 3 and Ga. Most data for well identified species in the Ga Cl system we re obtained from the Thermo Calc SUB94 database. They have passed the self consistency test as well as 2 nd and 3 rd law verifications with experimental data. The base conditions of molar flow rates in the source zon e for GaN growth by ALD were as follows. T emperature wa s varied through the range 3 00 to 1500 K The i nlet mole fraction of Ga is set to 0. 025, Cl for 0. 075, and He for 0.900 He ambient gas was used instead of N ambient gas to eliminate the formation o f GaN during the calculation although N carrier gas was employed in this study. An inert species in the gas phase primarily contributes to the system entropy. Figure 3 4 A shows the calculated equilibrium mole fractions of gas phase species in the Ga Cl I nert system as a function of growth zone temperature. The Cl/Ga atomic ratio was fixed to 3 and the temperature was varied in the range 3 0 0 to 1 5 00 K. Mole fraction of gas wa s kept to 1 in overall temperature range. Thus, there wa s no condensed phase ev ident from the calculation

PAGE 56

56 under this condition. A single Ga containing gas species in the growth zone wa s predicted to exist in the gas phase over most of temperature. GaCl 3 present s predominantly in the gas phase over most of temperature Figure 3 4 B shows the ratio of GaCl 3 to Ga 2 Cl 6 increases as temperature increases. From this result, the presence of GaCl 3 becomes more dominant at higher temperature. It is most often desirable to operate under condition where self limiting adsorption occupies all surface sites. Thus, GaCl 3 would be more suitable than Ga 2 Cl 6 Then, th e higher growth temperature is preferred.

PAGE 57

57 Table 3 1. C onsidered species in Ga Cl system. Taken from Sub94 database. Ga containing species (Gas phase) Ga, Ga 2 GaCl, GaCl 2 Ga 2 Cl 2 Ga 2 Cl 4 Ga 2 Cl 6 GaCl 3 Cl species (Gas phase) Cl, Cl 2 Condensed phase GaCl 3 (l), Ga(l), GaCl 3 (s), Ga(s)

PAGE 58

58 Figure 3 1. Schematic of overall GaN ALD system. Figure 3 2 Schematic of GaN ALD reactor chamber.

PAGE 59

59 Figure 3 3. Co nt rol program of GaN ALD process.

PAGE 60

60 Figure 3 4 (A) Equilibrium m ole fractions of gas species in the Ga Cl Inert system as a function of temperature at P=1 atm. (B) Molar ratio of GaCl 3 to Ga 2 Cl 6 in the Ga Cl Inert syste m as a function of temperature at P=1 atm.

PAGE 61

61 CHAPTER 4 ATOMIC LAYER DEPOSITION OF GaN USING GaCl 3 and NH 3 4.1 Introduction The g roup III nitrides have found wide application in both electronic and optoelectronic devices. In particular, solid solutions of the direct band gap nitrides can access the entire visible range of the spectrum as well as the near UV [113] Furthermore, the relatively high thermal stability of GaN and AlN has made these nitrides useful in power electronics. Sapphire is commonly used as substrate for growth of GaN, but the low cost, high crystalline quality, and large area of Si make it an attractive substrate choice. In practice the nitrides are typically deposited by metal organic chemical vapor deposition (MOCVD), and for GaN layers, a deposition temperature in the range 900 to 1100 C is needed to produce device quality material. Although the rate of NH 3 cracking is greater at higher temperature, the nitrogen partial pressure in equilibrium with the compound is also higher, which leads to a high concentration of nitrogen vacancies in the GaN film. This higher vacancy concentration is reported to be the reason for the high ba ckground carrier concentration in MOCVD grown GaN [114] Thus, a lower growth temperature is preferre d to reduce the background carrier concentration [57] In contrast, the thermal stability of InN is considerably less than that of GaN and AlN, and thus growth is typically limited to a temper ature below 800 C. Also, a lower growth temperature is preferred for growth of In x Ga 1 x N. Lowering the growth temperature in MOCVD, however, does present difficulties. At lower growth temperature the surface mobility of adsorbed species is reduced, whi ch produces films with poorer structural quality and hence optical properties. As example, it has been reported that In x Ga 1 x N grown at 500 C shows ~1000 times weaker photoluminescence than material grown at 800 C [115] Another issue at lower growth temperature is a decreased extent of NH 3 thermal dec omposition. Thus both poor

PAGE 62

62 mobility of surface species and the low availability of reactive nitrogen make growth at low temperature challenging. Atomic l ayer d eposition (ALD) is a potential solution to these growth issues. Since sequential delivery of t he reactants in a self limiting adsorption mode affords increased residence time, both the extent of species surface transport and chemical reaction are increased. This allows for improved structural quality as well as increased levels of reactive nitroge n. Thus ALD films of acceptable quality can often be grown at lower temperature than by CVD. As example, ALD growth of device quality GaAs, GaP, and InGaP ha s been demonstrated at a temperature as low as ~200 C [116] Of course, ALD gives control of the film thic kness at the atomic layer and shows improved reproducibility of alloy composition [117 ] Potential drawbacks of ALD are the lower growth rate and the fact that impurities can more effectively compete for adsorption sites, and thus increase background concentration of impurities Unfortunately, very little work has been reported on ALD of GaN. Early studies of ALD growth of GaN used a Ga metal organic source ( trimethylgallium ( TMG ) and triethylgallium ( TEG ) ) and NH 3 [37, 57, 118] Films were grown on sapphire or AlN, and a region of self limiting adsorption was shown with Ga source exposure time. The growth temperature window, however, was not investigated in these studies. The relatively unstable MO sources show a narrow process temperature window for most materials. T o resolve this issue, Ga chloride sources (GaCl and GaCl 3 ) have been used with NH 3 in ALD growth mode on GaAs or GaN/GaAs [59, 119] In these studies the self limiting adsorption region was evident with GaCl exposure time, and a temperature window was found when using GaCl. Films showed either hexagonal or cubic GaN(00 0 2) preferred orientation. The self limiting adsorption windows, however, were not established for use of GaCl 3 Also, Cl incorporation in the film, which is

PAGE 63

63 del eterious to device performance, was not considered. This study conclusively demonstrates ALD of GaN on Si(100) for the first time using GaCl 3 and NH 3 sources. It is suggested that GaCl 3 is a superior source because of its thermal stability in an H 2 free a dsorption step, which increases the width of the temperature window. GaCl 3 is also sufficiently volatile (1 mm Hg at 48 C) to allow easy delivery. The microstructure, Cl content, and chemical bonding states were examined for films grown within the ALD p rocess window. 4.2 Experimental Procedures In this study, one ALD cycle consisted of four steps of sequential gas supplies: (1) GaCl 3 /N 2 carrier, (2) N 2 purge, (3) NH 3 /N 2 carrier, and (4) N 2 purge. The ALD growth mechanism of GaN using GaCl 3 and NH 3 is an ticipated to be similar to that of TiN using TiCl 4 and NH 3 [117] During the first st ep, input GaCl 3 adsorbs on the surface active sites to form GaCl x ( x <3). Any remaining GaCl 3 or physisorbed by products are purged from the reactor in the second step. Next NH 3 reacts with adsorbed GaCl x to form chemisorbed GaNH 3 x and x HCl as a physisor bed by product. The r emaining NH 3 and physisorbed HCl are finally purged upon exposure to the N 2 carrier in the last step. ALD of GaN on Si(100) was performed in a custom horizontal quartz reactor equipped with a load lock system as described in C hapter 3. Pneumatically actuate d valves, controlled with a Lab V IEW based algorithm were sequenced to deliver each reactant and purge gas to the reactor or bypass line. The Si(100) surface was degreased (trichloroethylene, acetone, and methanol), followed by a buffered oxide etch (2 min). GaCl 3 (99.999%, Sigma Aldrich) was contained in a bubbler maintained at 70 C and delivered to the reactor in N 2 (99.999%, Praxair) carrier gas. NH 3 (99.9999%, Matheson Tri Gas) was used as the reactive nitrogen source and N 2 as the purge gas. GaN films were deposited in the temperature range 375 to 750 C at a nominal

PAGE 64

64 growth pressure of 1 Torr. One ALD cycle consisted of sequential GaCl 3 supply (50 sccm N 2 ) N 2 purge (30 s, 500 sccm ), NH 3 supply (500 sccm ), and N 2 purge (30 s, 500 sccm ). The film thickness was measured by cross sectional s canning e lectron m icroscop y (SEM) and X ray r eflectometry (XRR) using a JEOL JSM system, respectively. Grazing i ncident a ngle X ray d iffraction (GIXD) us ing a Philips MRD K radiation (40kV and 40mA) was used to characterize the film microstructure. GIXD was performed from 30 to 40 2 at 1 tilt. Atomic f orce m icroscop y (AFM) (Digital Instruments Dimension 3100) was used to examine the film surface morphology. To investigate chemical bonding states, X ray p hotoelectron s pectroscopy (XPS) measurements were performed with a PerkinElmer PHI ESCA system. Monochromatic Mg K ionizing radiation (1254 eV) was used with the X ra y source operating at 300 W (15 kV and 20 mA). To remove surface contamination, as grown films were sputtered with Ar + ions for 10 min prior to XPS measurement. 4.3 Results and Discussion 4.3.1 Confirmation of Self saturation ALD Growth To verify that gr owth occurred in ALD mode, a set of experiments was performed to identify a growth window in temperature and GaCl 3 exposure time that produced self limiting growth. To this end, a series of films was grown (150 cycles) in which the exposure time was varie d in the range 2 to 17 s at a fixed NH 3 exposure time of 10 s and N 2 purge time of 30 s. Fig ures 4 1 A and B show the growth rate (film thickness per cycle) as a function of GaCl 3 exposure time at growth temperature of 550 and 650 C, respectively. The r esults in Fig. 4 1A suggest that the growth rate at 550 C remains constant at 2.0 +/ 0.4 /cycle in the range 4 to 11 s of GaCl 3 exposure time. The measurements at the higher growth temperature, 650 C, show a similar growth plateau at 2.1 +/ 0.5 /cyc le, but as expected a narrower exposure time window

PAGE 65

65 (GaCl 3 exposure time in the range 5 to 9 s) as shown in Fig. 4 1B The constant thickness/cycle regime v ersu s reactant exposure is a feature of ALD and indicates that GaCl 3 is incorporated by self limit ing adsorption. The deposition rate in the saturation growth regime was 1.6 to 2.6 /cycle. As subsequently presented the microstructure of ALD GaN in this process window is a mixture of (00 02) and ( 10 1) oriented GaN. Thus, the thi ckness of 1 ML of ALD GaN is calculated as 5.58 Therefore, the deposition rate in the saturation growth regime is 29 to 47% of 1 ML/cycle It is commonly reported that the saturation coverage of the metal containing precursor for most metal nitride AL D reactions is less than a single monolayer [120] An often cited reason is that steric hindrance of adsorbed metal precursors prevents monolayer coverage [117] The results in Fig. 4 1 show that the growth rate linearly increases with GaCl 3 exposure time for exposure time greater than the window upper limit T he rate of increase is greater at 550 C than 650 C An increase in growth rate with exposure time is usually ascribed to continued chemisorption beyond the saturation limit at longer time. The low er increase in growth rate at highe r temperature suggests that a n additional second reaction occurs which may be desorption of a surface Ga species to more volatile GaCl by the reaction with residual HCl. The results of these experiments also show the commonly observed decrease in growth rate at metal exposure times less than that required for operation in the plateau region and are attributed to insufficient exposure time to reach saturation. It is noted that the rate of change in the growth rate at short exposure time was similar for bo th growth temperatures. 4.3.2 Process Temperature Window To further substantiate that growth occurs in ALD mode using this chemistry, a second set of experiments was performed to establish the extent of the temperature range yielding

PAGE 66

66 saturation limited gr owth. In these experiments depositions were made at the self saturation growth condition: GaCl 3 (7 s, 50 sccm), N 2 purge (30 s, 500 sccm), and NH 3 (10 s, 500 sccm) while varying the growth temperature in the range 375 750 C for runs of 150 cycles. As shown in Fig. 4 2 the growth rate increased with increasing growth temperature below 500 C, but then remained constant (2.1 +/ 0.2 /cycle) above this temperature, which is consistent with operating in the ALD process temperature window. Since precurso r adsorption onto a surface active site and reaction between adsorbed precursor and surface species are thermally activated processes, the growth temperature is expected to be an important factor [117] At temperature below the window limit there is insufficient thermal energy available to promote chemisorption at full coverage in the ava ilable exposure time. Within the ALD growth temperature window there is sufficient thermal energy to reach the saturation condition but insufficient for additional reactions to occur (e.g., precursor decomposition and desorption) that preclude saturation An important feature of using a metal halide precursor is its high thermal stability, and thus the growth rate remains constant even at a relatively high growth temperature. It is noted that GaCl 3 does not react with N 2 in the temperature range of this study and also will not disproportionate since there is no reducing species available (e.g., H 2 ). Thus this Ga source seems well suited for ALD growth of GaN. 4.3.3 Growth Characteristic of GaN ALD Since the first layer to deposit requires adsorption on a Si or SiO x /Si surface, its growth characteristics are expected to be different from those layers subsequently grown on the GaN surface. Furthermore, GaN is not well lattice matched to Si so that residual strain may persist for additional layers to furth er modify the growth characteristics of layers in the near Si/GaN interfacial region. Kinetic modeling and experimental results during the initial stages and the following stabilized stage have been reported [121 123] To better understand the early stages of

PAGE 67

67 growth, a series of films was grown at identical conditions except with variable cycle number. The dependency of film thicknes s on the number of cycles was then examined by XRR. The ALD growth conditions were GaCl 3 exposure time of 7 s at 650 C with the other steps at the base conditions, while the cycle number was varied in the range 5 to 200. The results are shown in Fig. 4 3 and indicate that the growth rate of ALD GaN gradually increased toward a constant value of 2.1 /cycle below 15 cycles, and this transition period is termed the transient stage. The growth rate becomes constant (2.1 +/ 0.2 /cycle) above 15 cycles, ex hibiting a linear dependency of film thickness on the cycle number. This is expected to occur because the adsorption characteristics of GaCl 3 on bare Si or SiO x /Si are different than those on N terminated GaN. During the first few cycles, self limiting a dsorption likely does not occur. However once the surface is fully covered and presumably a few layers thick, relaxed GaN becomes the growth surface, which allows self limiting adsorption at this specific growth temperature During the transient stage, it is possible that the adsorption of reactants and preferential growth on as deposited film compete, leading to three dimensional (3D) growth. For example, studies on ALD of high materials have reported 3D growth during the transient stage [124] To probe this possibility for ALD GaN, AFM images of the surface of films deposited in the transient stage were taken. Fig ure 4 4 depicts the surface morp hology of ALD GaN films grown for 5, 10, and 15 cycles. Figures 4 4 A and B show that the surface roughness, R q (root mean square roughness) is relatively high for films grown with 5 and 10 cycles (R q =0.84 nm for 5 cycles, R q =0.72 nm for 10 cycles), consi stent with 3D growth. The surface of the sample grown for 15 cycles ( Fig. 4 4 C) however, is substantially smoother (R q =0.55 nm). Note that the linear region begins at 15 cycles and is considered the point at which complete coverage of GaN on Si

PAGE 68

68 substra te is obtained. This conclusion is supported by noting that the surface roughness of the sample grown for 150 cycles ( Fig. 4 4 D) is close to that of 15 cycles. (R q =0.58 nm) 4.3.4 Film Microstructure Given the differences in reaction time and delivery se quence, differences in microstructure between ALD and CVD films might be expected. To explore this possibility, GIXD patterns of GaN film deposited in ALD mode were compared to those deposited outside the ALD process window at higher deposition rate and t hose deposited in CVD mode. GIXD patterns for ALD GaN films in Fig. 4 5 A reveals the change in the crystalline orientation for two different GaCl 3 exposure times. The lower GIXD pattern in Fig. 4 5 A shows the result for the film grown in ALD mode with GaCl 3 exposure time of 7 s while the upper GIXD pattern shows the result for a film grown in CVD like mode using an exposure time of 25 s The GIXD patterns of both films indicate that their diffraction patterns are well matched with h GaN(0002) and h Ga N(10 1) plane s peak positions of 34.633 and 36.994 2 respectively (JCPDS 02 1078). It is observed, however, that the h GaN(10 1) and h GaN(0002) peak intensities are similar for ALD film, while the h GaN(0002) peak is dominant for the CVD like film. The dominance of the h GaN (0002) peak becomes more significant in the GIXD pattern of films grown in full CVD mode ( Fig. 4 5 B) The significant presence of the h GaN(10 1) reflection in the self saturation region is believed to be a result of the lower growth rate. As shown in Fig. 4 1 the growth rate of the ALD film is ~1/3 of that grown in the CVD like region. When the growth rate is low, the step spacing is small and the growth mode is two dimensional (2D) step flow. The tendency for 2D step flow growth is stro nger for strained films as is the case here. In 2D step flow growth mode, the lateral growth orientation is preferred with the suppression of 3D growth. [125] Thus, the formation of the (10 1) plane is fou nd in ALD films. As shown in Fig. 4 6, comparison of

PAGE 69

69 the surface morphologies of ALD and CVD is consistent with ALD GaN growth in 2D step flow mode and CVD GaN in 3D growth mode. 4.3.5 Chemical Bonding State It is of interest to examine the composition of the film as well as the bonding states of the elements. XPS was used to determine the bonding states for a set of films grown in ALD mode as well as in the undersaturated and CVD like regimes. Films were prepared using GaCl 3 exposure times in the range 2 to 25 s and at base conditions for the other flow rates and exposure times at 650 C and 1 Torr. Fig ure 4 7 shows the evolution of XPS spectra for Ga 2 p N 1 s O 1 s and Cl 2 p with varying GaCl 3 exposure time. Fig ure 4 7 A shows the Ga 2 p spectra, whi ch were deconvoluted using a Gaussian Lorentzian function with background subtraction. The major Ga 2 p peak is at a binding energy (BE) of 1117.9 eV, which is close to the reported value for GaN [126] The Ga 2 p spectra differ by the presence of a low energy shoulder, which has a BE of 1116.9 eV that is attributed to Ga 2 O 3 For GaCl 3 exposure times shorter than the minimum one for self saturation (<9 s), the presence of Ga 2 O 3 in the film is evident from the Ga 2 p peak at the lower BE of 1116.9 eV, which shows a maximum area of 1 5%. As GaCl 3 exposure time increases, the Ga 2 p BE shifted from lower BE (mixture of Ga 2 O 3 and GaN) to higher BE (GaN only). The presence of Ga 2 O 3 in the film in the self saturation region is in agreement with the O 1 s spectra ( Fig. 4 7 C) showing a BE o f 530.8 eV, which corresponds to Ga 2 O 3 As GaCl 3 exposure time increases, the O 1 s peak intensity decreases and becomes negligible at 25 s of GaCl 3 exposure time. The results represented in these Ga 2 p and O 1 s spectra suggest that the Ga 2 O 3 is present i n film grown in the ALD self saturation region, but disappears for films deposited in the CVD like

PAGE 70

70 region. This trend is opposite that for the Cl Fig ure 4 7 D shows the XPS spectra of Cl 2 p which increases with GaCl 3 exposure time. For GaCl 3 exposure ti mes to 9 s (self saturation region), the Cl content is below the detection limit of XPS measurement (~1 at. %). As the GaCl 3 exposure time increases, the Cl 2 p peak intensity increases to an estimated 1.1 at. % for the 25 s GaCl 3 exposure time. As previo usly mentioned, it is reported that the oligomer (Cl 2 GaNH 2 ) n with n =1 3 is generated from gas phase reaction of GaCl 3 and NH 3 at 650 C [127] It is possible that this adduct forms at long exposure times from the reaction of the NH 3 with a small amount of residual GaCl 3 remaining in the reactor (e.g., adsorbed on the wall). Thu s a small amount of the oligomer adsorbs on the surface and is not fully eliminated by the ammonia flow. To test this possibility a full CVD film was grown, i.e., all reactants flowing simultaneously. Since the oligomer should be produced under these co nditions, high levels of Cl should be evident in the CVD film. The full CVD XPS results shown in Fig. 4 7 D indicate relatively high Cl content (1.6 at. %) consistent with the long GaCl 3 exposure time result. It is known that Cl incorporation can lead to bandgap narrowing and contact material corrosion in optoelectronic semiconductors [128] Thus, it is necessary to minimize Cl contamination. It has been shown that Cl lev els can be lowered by including H 2 in the purge steps [129, 130] The observed Cl contamination, however, is significant only if the GaCl 3 exposure time was longer than self saturation limit (above ~10 s) presumably as a result of adduct formation between GaCl 3 and NH 3 in the gas phase and on the surface. Thus, as long as GaCl 3 is below the self saturation limit, the Cl impurity is negligible within the limits of detection by XPS (~1 at. %) as shown in Fig. 4 7. 4.4 Conclusions ALD of GaN was demonstrated on Si(100) using GaCl 3 and NH 3 T he growth rate of GaN film was observed to be constant at 2.0 to 2.1 +/ 0.5 /cycle (29 to 47 % of 1 ML of h

PAGE 71

71 GaN/cycle) in the ALD self saturation limit. The ALD process window was found for GaCl 3 exposure time to be between 4 and 11 s at T growth =550 C, and 5 and 9 s at T growth =650 C. The high stability of GaCl 3 yielded a relatively wide ALD temperature window in the range 500 to at least 750 C for the specific switching schedule examined. The GIXD results indicated that ALD film contained two differe nt orientations, GaN(10 1) and GaN(0 0 02), in contrast to only GaN (0 0 02) for a CVD like film. For most device applications this is not acceptable and thus Si surface pretreatments for buffer layers should be explored to permit single crystal growth. The XPS results showed that ALD GaN films contained Ga 2 O 3 than CVD GaN films. The Cl content correlated inversely with the oxygen level. When GaCl 3 exposure time is less than that required to give self saturation growth, films were ch l or ine free. F ilms deposited in CVD like region showed 1.1 at. % Cl as did full CVD films.

PAGE 72

72 Figure 4 1. Dependence of ALD GaN growth rate on GaCl 3 exposure time A ) T growth = 550 C B) T growth = 650 C. These films are grown on Si(100) with NH 3 exposure time of 10 s.

PAGE 73

73 Fig ure 4 2. Dependence of the growth rate on temperature. Exposure times were set as GaCl 3 = 7 s, NH 3 = 10 s, and N 2 purge = 30 s. Fig ure 4 3 Dependence of film thickness (XRR) on cycle number. Growth conditions were set as GaCl 3 exposure time = 9 s, NH 3 exposure time = 10 s, and N 2 purge time = 30 s at T growth = 650 C. A) 5 cycles. B) 10 cycles. C) 15 cycles. D) 150 cycles.

PAGE 74

74 Figure 4 4. Surface morphology of the GaN films deposited with 9 s of GaCl 3 exposu re time at 650 C. A) 5 cycles. B) 10 cycles. C) 15 cycles. D) 150 cycles.

PAGE 75

75 Fig ure 4 5. GIXD patterns for GaN films grown on Si(100) at 650 C. A ) ALD GaN films. B ) CVD GaN films

PAGE 76

76 Figure 4 6. Surface morphology of the 60 nm thick GaN films deposited at 650 C under A) ALD growth condition B) CVD growth condition.

PAGE 77

77 Figure 4 7. XPS spectra of GaN film grown by ALD at various GaCl 3 exposure times. A) gallium 2 p B) nitrogen 1 s C) oxygen 1 s D) chlorine 2 p (A) ( B ) ( C ) ( D )

PAGE 78

78 CHAPTER 5 COMPUTATIONAL STUDY FOR ATOMIC LAYER DEPOSITION OF GaN USING GaCl 3 AND NH 3 5.1 Introduction GaN has gained of a considerable interest to its potential application s to optoelectronic device s operating in the visible and near UV range s of the sp ectrum [113] Metalorganic chemical vapor deposition (MOCVD), hydride vapor phase epitaxy (HVPE), and molecular beam epitaxy (MBE) are typical techniqu es used for deposition of GaN [31, 131, 132] In these deposition techniques, a high growth temperature generally in the range of 900 to 1100 C is necessary to obtain d evice quality material. At high temperature, however, the nitrogen vacancy concentration increases which is known to increase the background carrier concentration [114] Thus, a lower growth temperature is desired to decrease the b ackground carrier concentration. However, the structural quality is generally poorer. ALD is a promising deposition techniq ue to address this growth issue. The self limiting nature of the surface reaction in ALD leads to increased surface reaction residence time to increase surface transport and chemical reaction [13 3] This improves the structural quality of ALD films, and allo ws lower growth temperature, and typical ly required to obtain acceptable quality [117] The previous chapter reported the growth of GaN thin films at low temperature by ALD using GaCl 3 and NH 3 [134] It was known that in the self saturation limit Cl free GaN was grown in the temperature range 550 to 650 C Unlike Cl free ALD GaN film, relatively high Cl content (1.6 at. %) was obtained in the CVD film, i.e., all reactants flowing simultaneously. The reason f or higher Cl incorporation in the CVD GaN film was suggested by the formation of the oligomer (Cl 2 GaNH 2 ) n with n =1 to 3 f rom gas phase reaction of GaCl 3 and NH 3 [127] Since GaCl 3 and NH 3 are not directly mixed in ALD, there is n o Cl incorporation in the ALD G aN

PAGE 79

79 film. Cl incorporation in the film can lead to contact material corrosion and bandgap lowering [128] Thus, it is very important to better understand Cl incorporation in ALD GaN film as well as high Cl levels in CVD GaN film. This study aims to establish fundamental description of self limiting surface reactions during ALD processes and to provide insight into Cl incorporation in the film. Atomistic mechanisms were investigated for the reactions of GaCl 3 and NH 3 with appropriate surface active sites. The a ssociated thermochemistry was also studied to estimate the energy barrier along reaction pathway s 5.2 Computational Details Surface reactions in ALD of GaN using GaCl 3 and NH 3 were described into two half reactions: the reaction between GaCl 3 and the growing surface during the GaCl 3 supply step, and the reaction between NH 3 and growing surface during the NH 3 supply step. The g rowing surface was treated as ending w ith the surface active site: NH 2 at the initiation of the GaCl 3 supply step, Cl at the beginning of the NH 3 supply step. The t wo half reactions are described as: GaN Ga NH 2 + GaCl 3 (g) GaN Ga NH Ga Cl 2 + HCl(g) 5.1 (a) GaN Ga Cl 2 + 2NH 3 (g) GaN Ga 2 NH 2 + 2HCl(g) 5.2 (b) Surface active sites are denoted by asterisks. Reaction 5.1(a) was modeled as a cluster consist ing of GaN Ga NH 2 a s depicted in Fig. 5 1. To reduce the computational budget, the tetrahedron formed by the adsorbed Ga species and gro wing (GaN) 3 was truncated with H atoms. To simulate the growing surface, atoms other than surface active sites were treated using the freeze option in the calculation. The B3LYP density functional theory (DFT) method using Gaussian 03 split basis sets was employed for all calculations associated with reaction mechanisms of 5.1 (a) and (b) [135, 136] The LanL2DZ basis set was used for Ga atom, and 6 311G(d) was used for H, N, and Cl [137]

PAGE 80

80 Molecular geometries were optimized by energy minimization to stationary points in the potential energy surface (PES). Transition states were optimized using the Berny algorithm. Natural bonding orbital (NBO) analysis was used to determine the weakest bond for each molecule. Frequency calculation s w ere performed to calculate the zero point energies and thermal corrections for reactants, intermediates, and products at growth temperature and pressure. 5.3 Results and Discussion 5.3.1 Chemisorption of GaCl 3 The proposed mechanism f or the first half reaction 5.1(a) the chemisorption of GaCl 3 (g), is shown in Fig. 5 2 along with reaction energetics ( Gibbs energ y and enthalp y) at both standard ( T =273 K, P =1 atm) and reported growth condition s ( T growth =923 K, P growth =0.013 atm) [134] The resulting changes in enthalpy and Gibbs energy at standard and growth condition are summarized in Table 5 1. As shown in Fig. 5 2, GaCl 3 (g) is molecularly adsorbed on a Ga NH 2 surface active site by electron lone pair donation from a surface NH 2 group to an empty p orbital of a Ga atom in GaCl 3 (g) to form the dative bond. The Ga atom in GaCl 3 (g) is in an electron deficient state due to the strong electro negativity of Cl. The e lectron deficient Ga atom in GaCl 3 (g) precursor is preferred to accept lone pair electron from surface NH 2 group, and this makes the dative bond formation easier. The resulting adsorption energy is relatively low as 2 1 kcal/mol, and the Gibbs ene rgy change is + 21.8 kcal/mol at the growth condition. The next step involves the production of physisorbed HCl(a) on the surface resulting from the recombination of Cl atom in GaCl 3 (g) precursor with one of H atom s in the surface NH 2 group. The propos ed transition state between chemisorbed GaCl 3 (a) and physisorbed HCl(a) is shown in Fig. 5 2 The formation of physisorbed HCl(a) is activated with respect to the chemisorbed GaCl 3 (g) state with an energy of 3.6 kcal/mol and the Gibbs energy of 9.1 kcal/m ol

PAGE 81

81 at the reported growth condition. The Gibbs energy change for the transition state to the chemisorbed GaCl 3 (a) state is almost independent of temperature because those two states are entropically similar due to no gas species involved. The entropy of the physisorbed HCl(a) state is higher than chemisorbed GaCl 3 (a) and transition state, which explain s the Gibbs energy stabilization with increased temperature In the final step, the physisorbed HCl(a) is removed from the surface during the N 2 purge s tep in ALD process, to yield the Ga NH Ga Cl 2 surface group and HCl(g). As shown in Fig. 5 2, The Gibbs energy of desorbed HCl(g) state is lower than the physisorbed HCl(a) state since there is an increase of entropy by the formation of gaseous HCl. Thu s, the Gibbs energy change is negative and becomes more stable with increasing temperature. After the reaction, the as grown GaN surface is saturated by a monolayer of Ga atoms and t he surface is converted from NH 2 terminated to Cl terminated surface. This Cl terminated surface becomes the initial surface for the second half reaction 5.1 (b) The GaCl 3 precursor has a planar structure after geometry optimization as shown in the upper half part of Fig. 5 2. This implies that one Ga 4 s and two Ga 4 p orbitals form 3 sp 2 hybrid orbitals bonded to Cl ligands. After this orbital hybridization, Ga has one empty 4 p orbital to form the dative bond by lone pair electron donation from NH 2 surface group. A similar dative bond formation by lone pair electron donation is reported in the ALD of Al 2 O 3 from AlCl 3 and H 2 O [138] If the energy state of the adsorbed GaCl 3 (a ) complex is lower than dissociated products, a significant fraction of the precursor molecule will be molecularly trapped in the adsorbed complex resulting in incorporation of Cl atoms in the film. As shown in Fig 5 2, however, the energy state of GaCl 3 (a) dative bonded complex is higher than dissociated products (Gibbs

PAGE 82

82 energy difference) Thus, there is a thermodynamic driving force for most of adsorbed precursor molecules to directly dissociate to p roducts instead of molecularly trapped in the adsorbe d complex state This products low Cl incorporation in ALD is consistent with the XPS results given in C hapter 4 [134] 5.3.2 Chemisorption of NH 3 The proposed mechanism of reaction 5.1(b) the chemisorptions of NH 3 (g), is schematically shown in Fig. 5 3 with computed reaction energetics. The changes in enthalpy and Gibbs energy are listed in Table 5 2 at standard a nd grow th condition. In this reaction, the Ga Cl 2 surface species is produced from reaction 5.1(a). The chlorine terminated species is e xposed to NH 3 (g) molecularly adsorb s on the Ga Cl 2 surface site by dative bond formation between the N and surface Ga sim ilar to that in the first half reaction. The electron lone pair of N in NH 3 (g) precursor is donated t o the empty 4 p orbital of the Ga atom in chemisorbed species, Ga Cl 2 *. As listed in Table 5 2, the adsorption enthalpy of NH 3 onto a Ga Cl 2 surface site is 4. 2 kcal/mol, and the Gibbs energy change is + 14 .2 kcal/mol at the growth condition. The transfer of first H atom in NH 3 (a) to a Cl in the Ga Cl 2 surface group to yield p hysisorbed HCl(a) is shown in Fig. 5 3 along with the transition state betwee n the NH 3 (a) adsorption state and HCl(a) physisorption state. Similar to the energetic s of transition state in reaction 5.1(a) the formation of HCl(a) is activated with respect to the chemisorbed NH 3 (a) state, showing an activation energy of 3.8 kcal/ mol The p hysisorbed HCl(a) is removed from the surface during N 2 purge step, and results in Ga ClNH 2 surface group and desorbed HCl(g). The Gibbs energy in this step is stabilized at increased temperat ure due to increased entropy. The re maining Cl atom in Ga ClNH 2 surface group is to be removed from the surface as HCl(g) by the similar reaction with NH 3 (g). After second half reaction, The Cl terminated Ga surface is now saturated with NH 2 The resulting

PAGE 83

83 Ga NH 2 surface is the initial surface for the next cycle of ALD, i.e., G aCl 3 (g) precursor exposure. After completion of reactions 5.1(a) and (b) the surface is saturated with a fractioned molecular layer of GaN, and maintains the same surface active site, NH 2 *. As shown in Table 5 2, the enthalpy for the chemisorption of NH 3 (g) is 4.2 kcal/mol, which is l ower value than the chemisorption of GaCl 3 (g). The lower enthalpy for the NH 3 chemisorption leads to the higher probability for surface Cl atom to be desorbed as HCl(g), resulting in lower Cl in corporation in the film. In addition, the stabilized Gibbs energy of HCl(g) desorption state at increased temperature drives lower the Cl incorporation at growth temperature. 5.4 Conclusions In this study, a detailed atomistic reaction mechanism was pro posed for ALD of GaN using GaCl 3 and NH 3 The chemisorption processes of GaCl 3 and NH 3 onto the growing surface were investigated by considering the adsorption and desorption state s along with the intermediated transition state. GaCl 3 and NH 3 were molecu larly adsorbed on the growing surface by electron lone pair donation from N to Ga The surface active sites were converted from NH 2 terminated to Cl terminated after GaCl 3 chemisorption, and from Cl to NH 2 terminated after NH 3 chemisorption. HCl was the desorbed product in both chemisorptions process. The PES of both chemisorption processes were presented with Gibbs energies at standard and growth condition. The PES of GaCl 3 chemisorption shows no molecular trapping of Cl incorporating adsorbed comp lex due to the higher energetic level of GaCl 3 adsorption state. This results in lowered Cl incorporation in the growing surface. Furthermore, PES of NH 3 chemisorption exhibits a negative energy barrier ( 4.2 kcal/mol). NH 3 chemisorption leads to desorp tion of Cl atoms from the growing surface as HCl products. Thus, Cl incorporation in the film can be much lowered by

PAGE 84

84 the negative energy barrier for NH 3 chemisorption. These results provide a rationale for Cl free ALD, GaN film s grown

PAGE 85

85 Table 5 1. Standard enthalpy H 298 ) and Gibbs energ y difference at 298K, 1 atm G 298 G 923 K, 0.013 atm ) for chemisorption of GaCl 3 GaCl 3 (a) Chemisorption State Transition State HCl(a) Phsysorption State HCl(g) Desorption State 298 (kc al/mol) 2.1 + 1.9 +1.2 1.6 298 (kcal/mol) +13.2 +19.6 +10.4 3.5 923 K, 0.013 atm (kcal/mol) +21.8 +30.9 +11.3 8.7 Table 5 2. H 298 ) and Gibbs energy difference at 298K, 1 atm G 298 G 923 K, 0.013 atm ) for chemisorption of NH 3 NH 3 (a) Chemisorption State Transition State HCl(a) Phsysorption State HCl(g) Desorption State 298 (kcal/mol) 4.2 0.4 2.0 6.4 298 (kcal/mol) +5.1 +9.6 +3.2 6.2 923 K, 0.013 atm (kcal/mol) +14.2 +21.5 +11.9 6.3

PAGE 86

86 Figure 5 1. Cluster model used in the calculation to represent GaN GaNH 2 surface.

PAGE 87

87 Figure 5 H 298 G 298 ) and 923 K, 0.013 G 923 K, 0.013 atm ) for the chemi sorption of GaCl 3 representing first half reaction. The unit is kcal/mol. Figure 5 H 298 G 298 ) and 923 K, 0.013 G 923 K, 0.013 atm ) for the chemisorption of NH 3 representing second half re action. The unit is kcal/mol.

PAGE 88

88 CHAPTER 6 p TYPE DOPING OF GaN USING Cp 2 Mg AS A DOPANT 6.1 Introduction There ha s been many efforts to develop group III nitride based optoelectronic devices for blue light emitting and laser diode applications over past dec ades since they cover the wide band gap energy range 2.0 to 6.2 eV at room temperature [113] One of difficult challenges to develop III nitride device s was to achieve low resistivity p type GaN To activate dopant, p ost growth N 2 ambient annealing or LEEBI treatment was employed [34, 35] Mg is the most widely used p type dop ant by both MO CVD and HVPE. For M O CVD growth, TMG and NH 3 were employed as Ga and N source, respectively. The commonly used Mg precursor in p type doping of GaN for MO CVD is b is cyclopentadienyl magnesium (Cp 2 Mg). MO CVD of Mg doped GaN is known to be very sensitive to the growth paramet ers including TMG and Cp 2 Mg flow rates N / Ga ratio, growth rate, growth pressure, and carrier gas selection Many groups have reported the optimization of growth parameters to achieve high hole concentration and low resistivity The best result to date shows a hole concentration as low 10 18 cm 3 cm [139 142] HVPE of Mg doped GaN was performed to reduce C contamination in the film [143] GaC l and NH 3 were employed as Ga and N source, respectively. Mg was supplied with HCl, and films were grown with AlN buffer layer. The highest measured hole concentration was 1.5 10 18 cm 3 at the room temperature. The structural quality, however, was not investigated in this study. In this study, p type GaN was grown using Cp 2 Mg as the dopant precursor. HVPE was employed in the growth of Mg doped GaN. For the first time, GaCl 3 was used as the Ga source with NH 3 for p type GaN HVPE GaCl 3 allows easy delivery due to its high vapor pressure (1

PAGE 89

89 mm Hg at 48 C) The s tructural quality, electrical properties, and optical characteristics were determined on the Mg doped GaN films grown as a function of Mg source incorporation. The results could be explaine d by introducing the effect of impurities and compensating 6.2 Experimental Details The GaN films were deposited on c plane sapphire substrate s using a c ustom horizontal H VPE system Sapphire substrates were degreased by trichloroethylene, acetone, and methanol. GaCl 3 (99.999% Rohm and Haas ) w as used as Ga source, contained in a bubbler maintained at 75 C Cp 2 Mg ( 99.999%, Epichem ) was used as Mg dopant source, contained in a bubbler maintained at 25 C These metal precursors were delivered to the r eactor by N 2 (99.999%, Airgas South) carrier gas. NH 3 (99.9999%, Matheson Tri Gas) was used as N source. Sapphire substrates were thermally treated in NH 3 at 750 C for 30 min before the growth. To initiate epi growth, thin GaN buffer layer was grown at 5 50 C for 5 min. Then, the growth temperature was raised to 700 C and 450 nm thick Mg doped GaN was grown with V/III ratio of 175. The flow rate of GaCl 3 and NH 3 was kept at 50 sccm (sccm denotes cubic centimeters per minute at STP) and 500 sccm, res pectively. The flow rate of Cp 2 Mg was varied from 5 to 30 sccm, according to the molar flow range 0.1 to 0.6 mol/min. To activate dopants, post growth annealing was performed under N 2 at 750 C for 30 min. The film thickness was measured by cross sect ional s canning e lectron m icroscop y (SEM) using a JEOL JSM 6335F SEM. The microstructure of the film was investigated by d ouble crystal X ray diffraction (DCXRD) DCXRD was performed with Cu K radiation (40kV and 40mA) t o obtain rocking curve of on axis GaN (0002) operating. Carrier concentration, mobility, and resistivity of the film were measured by room temperature Hall measurements (van der Pauw method) using Lakeshore 7507. Hall measurement was performed with the s weeping of magnetic field in the range of 10 mG to 10

PAGE 90

9 0 kG. The o ptical propert ies of Mg doped GaN film s w ere characterized by room temperature photoluminescence (PL) measurement. 6.3 Results and Discussion 6.3.1 Film Microstructure DCXRD was used to investigate the structure properties of Mg doped GaN layer. It is known that full width half maximum (FWHM) of the on axis (0002) rocking curve is related to the screw and mixed type dislocations [144] Figure 6 1 shows the FWHM of on axis rocking cu rve as a function of Cp 2 Mg flow rate for 450 nm Mg doped GaN film. The FWHM of undoped 450 nm GaN is 157 arcsec. In general, the FWHM increases as the incorporation of Mg increases. The increase of FWHM is not obvious when the Cp 2 Mg flow rate is belo w 15 sccm. However, the FWHM dramatically increases with the incorporation of Mg when the Cp 2 Mg flow rate is above 20 sccm. As mentioned above, the increase of FWHM ind icates th e increase of the dislocation density Thus, it is revealed that the disloca tion density of Mg doped GaN film i ncreases with Mg incorporation. The dislocation of Mg doped GaN is believed to result from the difference of atomic size between Ga and Mg, where the atomic radius of Ga and Mg are 136 and 145 pm, respectively [145] 6.3.2 Film Resistivity For the film resistivity, film thickness measured from SEM was employed in electrical measurements using van der Pauw method. The ohmic contacts were fabricated using In Zn alloy, which were annealed at 500 C for 1 min under N 2 in order to decrease contact resistance of the interface. Resistivities of Mg doped GaN as a function of Mg incorporation are shown in Fig. 6 2. The resistivity of undoped GaN film was 890 cm As shown in Fig. 6 2, the resistivity of Mg doped GaN film increases with the Mg incorporation even though there is a slight

PAGE 91

91 decrease at the Cp 2 Mg flow rate of 10 sccm. The high resistivity of highly Mg doped GaN can be explained by H passivation I t is known that Mg acceptors are passivated by H atom s which generated from decomposed NH 3 and Cp 2 Mg during growth [146 148] Post annealing was performed to activate Mg acceptors by cracking Mg H complex. However, the employed annealing condition was not enough to activate Mg acceptor completely. Also, there was another reason of high resistivity which is self compensation by deep donor. This will be discussed in C hapter 6.3.3 6.3.3 Hole Properties Hole concentration and mobility were measured by room temperature Hall measurement using In Zn alloy electrode. Figure 6 3 shows the hole mobility of Mg doped GaN film as a function of Mg incorporation. The hole mobility of undoped GaN was 510 cm 3 V 1 s 1 There was slight increase of the hole mobility up to 1340 cm 3 V 1 s 1 when Cp 2 Mg flow rate was 5 sccm. Then, the hole concentration decreases as Cp 2 Mg flow rate increases, and dropped down to 4.75 cm 3 V 1 s 1 when Cp 2 Mg flow rate was 30 sccm. As shown, Mg doped GaN has the lower hole mobility than undoped GaN, and gets much lower as increase of Mg doping level. The low hole mobility in the Mg doped GaN is not only attributed to the scattering by the neutral or ionized Mg impurities and native defects, but also to the inter band scattering between heavy hole and light hole bands [141] The mobility limitation effect by these scattering s becomes more severe as Mg incorporation increases since Mg impurities, defects, and heavy light holes increase with Mg incorporation Figure 6 4 shows the hole concentration of Mg doped GaN film as a function of Mg incorporation. Undoped GaN film showed n type conduction, and the background carrier concentrati on was 5.1 10 16 cm 3 The GaN film showed p type conduction with the addition of

PAGE 92

92 Cp 2 Mg. The hole concentration increases with Mg incorporation when Cp 2 Mg flow rate is below 15 sccm. The maximum hole concentration is 3.2 10 17 cm 3 at 15 sccm of Cp 2 Mg flow rate. When Cp 2 Mg flow rate is above 15 sccm, the hole concentration begins to decrease with Mg incorporation, and showed the minimum value of 5.6 10 16 cm 3 at 30 sccm of Cp 2 Mg flow rate. The resulting low hole concentration at high Mg incorporat ion indicates the existence of self compensation caused by the association of Mg Ga acceptor and the nearest neighbor N vacancy at high Mg doping level. According to the defect energy diagram and first principle cal culat ion, one of the most dominating defe ct s in highly doped p GaN is N vacancy, V N [149] N vacancy exists as V N 3+ charge state instead of V N 1+ due to a l ower energy of formation and large relaxation by neighboring Ga atoms (moving outward by ~15% of bond leng th) [150] This V N 3+ is mobile, and tends to associate wit h oppositely charged Mg Ga acceptors. The association of V N 3+ and Mg Ga 1 forms a nearest neighbor complex, (V N Mg Ga ) 2+ which is energetically favored than (V N Mg Ga ) 0 [149] This (V N Mg Ga ) 2+ complex acts as compensating donor for Mg Ga acceptor, and the concentration s of V N 3+ and (V N Mg Ga ) 2+ increase at high Mg incorporation Thus, there is more self compensation of Mg Ga acceptor by (V N Mg Ga ) 2+ donor at high Mg incorporation, and this results in low hole concentration at high Mg incorporation. This self compensation effect is well established for the high gap II VI compounds [151] I n Ag doped ZnS, the Ag Zn acce ptors associate with S vacancies, forming V S Ag Zn donor centers. This V S Ag Zn donor compensates Ag Zn acceptor, and limit s the hole concentration at high Ag incorporati on 6.3. 4 Optical Properties As mentioned, the hole concentration decreases at high Mg incorporation due to the self compensation of Mg Ga acceptor by (V N Mg Ga ) 2+ complex donor. To confirm this self

PAGE 93

93 compensation at high Mg incorporation, PL was performed on samples after carrier activation. PL spectra were calibrated to the excitatio n peak location. Figure 6 5 shows the room temperature PL spectra of Mg doped GaN films at different levels of Mg incorporation. For undoped GaN, the near band gap edge emission at 3.4 1 eV (364 nm) is dominant peak. With the increase of Mg incorporatio n, emission at 3.12 eV (398 nm) gradually becomes the dominant peak while the emission at 3.41 eV decreases. The peak intensity of t he emission at 3.12 eV increases with Mg incorporation, and shows the highest value at 15 sccm of Cp 2 Mg flow rate At high er Mg incorporation, the dominant peak becomes the emission at 2.83 eV (438 nm). The peak positions for various Mg incorporation levels are attributed to different types of donor acceptor (D A) pair recombination. The ionization energy of Mg Ga 1 acce ptor is known to be in the range of 0.2 2 0.2 5 eV [7] And, the ionization energy of the V N 3+ dono r, which naturally exists in GaN film, is ~0.03 eV [8] When Mg is fairly doped in GaN, D A reco mbination by Mg Ga 1 acceptor and V N 3+ donor or conduction band becomes dominant, leading to E g shift by 0.2 2 to 0.2 8 eV. This E g shift results in the emission peak in the range 3.1 3 to 3.1 9 eV which is very close to the position of the dominant emission peak at 3.12 eV when Cp 2 Mg flow rate is in the range 5 to 15 sccm in Fig. 6 5 Thus, it is stated that D A recombination by Mg Ga acceptor and V N donor or conduction band is responsible for the emission at 3.12 eV (398 nm). And, Mg Ga acceptor level is hig hest when Cp 2 Mg flow rate is 15 sccm. As mentioned previously, (V N Mg Ga ) 2+ compensating donor tends to be formed by the association of Mg Ga acceptor and the nearest neighbor V N in highly Mg doped GaN. The ionization energy of (V N Mg Ga ) 2+ donor i s re ported to be in the range 0.3 to 0.4 eV [152] And, the ionization energy of Mg Ga acceptor is in the range 0.22 to 0.25 eV. Thus, D A recombination

PAGE 94

94 by Mg Ga acceptor and (V N Mg Ga ) 2+ donor leads to E g shift by 0.52 to 0.65 eV. This E g shift results in the emission peak in the range 2.76 to 2.89 eV, which is same as the peak position of dominant emission peak at 2.83 eV for highly Mg doped GaN in Fig. 6 5. Thus, the emission at 2.83 eV for highly Mg doped GaN film is attributed to D A recombination by self compensation of Mg Ga acceptor with (V N Mg Ga ) 2+ donor. And, the self compensation becomes more severe as the increase of Mg incorporation. In summary, PL spectra exhibit two different types of D A recombination which are responsi ble for two different emission peaks (2.83 and 3.12 eV). D A recombination by Mg Ga acceptor and V N donor or conduction band corresponds to the emission of 3.12 eV. And, D A recombination by the self compensation of Mg Ga acceptor with (V N Mg Ga ) 2+ donor co rresponds to the emission of 2.83 eV. The evolution of 3.12 eV emission peaks demonstrates that MgGa acceptor level is highest at 15 sccm of Cp 2 Mg flow rate. This is well supported by the result of Hall measurement, which shows the highest hole concentra tion at 15 sccm of Cp 2 Mg flow rate. In addition, the appearance of a self compensation emission peak (2.83 eV) at very high Mg incorporation confirms the formation of (V N Mg Ga ) 2+ donor in highly Mg doped GaN. Also, the evolution of self compensation emis sion peaks exhibits the more compensation of Mg acceptor in highly Mg doped GaN This confirms the result of Hall measurement, which shows the decrease of hole concentration at very high Mg incorporation. Results from PL measurements such as emission typ es and positions are in good agreements with results from Mg doped GaN using TMG and Cp 2 Mg [140, 145, 153] 6.4 Conclusions p type GaN was grown by low pressure H VPE system. GaCl 3 NH 3 and Cp 2 Mg were employed as Ga, N, and Mg source, respectively. Structural, electrical, and optical properties were investigated with the variation of Mg incorporation. The defect density of Mg doped GaN

PAGE 95

95 film increased with Mg incorporation due to the large difference of the atomic size between Ga and Mg. Also, the resistivity increased with Mg incorporation due to H passivation and self compensation. Scattering effect decreased the while the hole mobility decreased. A maximum hole concentration was 3.2 10 17 cm 3 at 15 sccm of Cp 2 Mg flow rate. The hole concentration decrease d in highly Mg doped GaN due to self compensation of Mg acceptor. From the results of the room temperature PL, the emission peak at 2.83 eV in highly doped GaN confirmed the self compensation of Mg acceptor Also, the evolution of 3.12 eV emission peaks supported the optimum condition for highest hole concentration, which is at 15 sccm of Cp 2 Mg flow rate.

PAGE 96

96 Figure 6 1. Crystallinity of Mg doped GaN films as a function of Mg incorporation. F igure 6 2. Resistivities of Mg doped GaN films as a function of Mg incorporation.

PAGE 97

97 Figure 6 3. Hole mobilities of Mg doped GaN films as a function of Mg incorporation. Figure 6 4. Hole concentrations of Mg doped GaN films as a function of Mg incorp oration.

PAGE 98

98 Figure 6 5. PL spectra of Mg doped GaN films as a function of Mg incorporation. A) undoped, B) Cp 2 Mg: 5 sccm, C) Cp 2 Mg: 10 sccm, D) Cp 2 Mg: 15 sccm, E) Cp 2 Mg: 20 sccm, and F) Cp 2 Mg: 25 sccm.

PAGE 99

99 CHAPTER 7 EFFECT OF AIR EXPOSURE ON Cu DIFFUSION BARRIER PROPERTIES OF TaN THIN FILMS GROWN BY ATOMIC LAYER DEPOSITION 7.1 Introduction As device size decreases on integrated circuit (ICs), aluminum and SiO 2 based interconnect metallization is being replaced with Cu and low dielectrics. Low resistivity of Cu and low dielectric constant of low materials enable faster electric signal by minimizing RC (Resistance & Capacitance) time delay. Cu also has excellent resistance to electromigration, which is common problem of aluminum based metalliz ation. [111, 154] However, Cu diffuses into Si and SiO 2 very rapidly (D Cu ~ 2 10 5 cm 2 /s at 500 C), and causes many problems in device performance such as increase in contact resistance, barrier height change, leakage in p n junction, and destruction of electric connections. [155, 156] Therefore, an appropriate Cu diffusion barrier is required between Cu and its underlying layer to block interdiffusion of Cu. Among various transition metals and their nitrides, Ta/ TaN is known as optimal Cu di ffusion barrier due to its highest thermal stability and good adhesion with Cu. Currently, a bi layer structure of Ta/TaN deposited by physical vapor deposition (PVD) is being employed in commercial applications. However, PVD has an inherent limitation i n the deposition of sub 100 nm nodes and high aspect ratio structures because of directional nature of depositing flux and high sticking probability. [117] Accordingly, atomic layer deposition (ALD) is receiving extensive attention on the deposition of Cu diffusion barrier due to its unique features such as excellent conformality on high a spect ratio structure d and uniformity over large area s [157] In ALD, the gaseous precursors are alternately supplied to the substrate, and reactor is purged with an inert gas between precursor pulses. Under properly chosen growth conditions, the film growth proceeds via self saturation surface reactions. This self saturation reaction enables the

PAGE 100

100 film thickness to be c ontrolled easily, and the deposited films have excellent conformality and large are a uniformity. There have been reports on TaN ALD using halide source s (TaCl 5 TaBr 5 ) and NH 3 The films were mostly polycrystalline Ta 3 N 5 phase which has too many grain boundaries, in other word, fastest diffusion pathways for Cu. H igh resistivity (> 10 4 cm) was also an issue for diffusion barrier. [16, 17] As an alternate approach, metalorganic (MO) source s such as pentakis (dimethylamido) tantalum (PDMAT) [18] and tert butylimido tris(diethylamido) tantal um (TBTDET) [19, 20] had bee n used as a precursor for TaN ALD. A fcc NaCl type nanocrystalline structure of TaN was obtained by using PDMAT as Ta source. [18] Thermal ALD TaN films deposited with TBTDET showed lower density and higher resistivity than plasma enhanced (PE) ALD TaN. Low resistivity of PE ALD TaN was somewhat attributed to formation of Ta C bond s [17] It is well known that there are several factors such as impurities, thickness, stress, and defects that affect the properties of diffusion barrier s [21 24] General ly, O, C, and H are the main impurities in ultrathin films deposited by ALD because they are prepared using metalorganic precursors. These impurities are known to change the structures and properties of TaN thin films [17, 25, 117] There is a report on the oxygen effect on the performance of nitride barrier in the case of oxygen stuffing process for TiN barrier. [27] In this study, TaN thin films were deposited by the ALD method from TBTDET and NH 3 In contrast to previous studi es on ALD TaN using TBTDET and NH 3 [19, 20] the process temperature window, which is an essential ALD feature for precise thickness control, was examined from 150 to 400 C. As a barrier candidate, it is necessary to investigate thermal behavior of TaN films and corresponding stack of Cu/TaN/Si. Therefore, the barrier

PAGE 101

101 performance of TaN thin films was examined with the stack of Cu/TaN/Si, which was prepared by ALD of TaN followed by CVD of Cu. To understand the effect of air exposure on the thermal stability of diffusion barrier, Cu/TaN/Si stacks were prepared with and without an oxide interfacial layer between Cu and TaN. Both stacks were annealed at the same temperature, and compared using different thermal stability test s 7.2 Experimental Details ALD of TaN was performed on sta ndard p type Si(100) substrate in a custom built cold wall vertical ALD reactor TBTDET (pr epared by Dr. McElwee White group) and NH 3 were used as precursors. TBTDET contained in a bubbler was heated to 85 C and pulsed into the reactor by a carrier gas of N 2 with the aid of a solenoid valve. NH 3 (99.9999 %, Matheson Tri Gas) was introduced intermittently into the reactor using pneumatically actuated valve. Pure N 2 (99.999 %, Airgas) was used as the purging gas, and sweeps TBTDET and NH 3 lines as well as the reactor during the purge step. TaN films were deposited from 200 to 400 C with a typical growth pressure of 1 torr. One ALD cycle consists of four stages: TBTDET supply (step 1), N 2 purge (step 2), NH 3 supply (step 3) and N 2 purge (step4). The gas flow sequence was as follows: TBTDET supply for 6 to 12 s with 50 sccm, N 2 purge for 30 s with 500 sccm, NH 3 supply for 10 s with 200 sccm, and N 2 purge for 30 sec with 500 sccm. To investigate the diffusion barrier performance of TaN films for Cu metall ization, a Cu layer of 100 nm thickness was deposited onto the ALD TaN layer by chemical vapo r deposition (CVD) using (tmvs) Cu I (hfac) as a single source precursor, where tmvs = trimethylvinylsilane and hfac = hexafluoroacetylacetonate. (tmvs)Cu I (hfac) in t he bubbler was heated to 45 C, and delivered to the reactor with the flow rate of 50 sccm by the carrier gas of N 2 H 2 was used as the ambient gas, and supplied into the reactor with the flow rate of 300 sccm. The growth temperature was set to 200 C. To investigate the effect of exposure of the TaN barrier film to

PAGE 102

102 air before deposition of the Cu layer in the stacks, the TaN/Si samples were either directly transferred to the Cu CVD chamber under vacuum or first exposed to room temperature air before loa ding into the chamber. The stack of Cu 100 nm/non air exposed TaN/Si will be designated as in situ group (IN). In comparison, the stack of Cu 100 nm/air exposed TaN/Si will be designated as ex situ group (EX). Both of EX and IN sets of Cu/TaN/Si stacks were then annealed in the Cu CVD chamber at elevated temperatures for 30 min as a diffusion barrier test. Annealing was under N 2 atmosphere in order to prevent Cu from oxidation. After annealing, the quality of diffusion barrier was evaluated by checkin g Cu x Si formation using XRD, four point probe, and the etch pit test. In the etch pit test, Cu was first chemically etched away by diluted HNO 3 solution with HNO 3 :H 2 O=1:20. TaN was etched by mixture of NH 4 OH, H 2 O 2 and H 2 O with a ratio of 1:1:4. Then, t he Si surface was etched with the Seeco etch solution (a mixture of 0.15 M K 2 Cr 2 O 7 and HF with a ratio of 1:2 for 5 s. After etching, the appearance of inverse pyramidal shaped etch pit s was observed by scanning electron microscope (SEM). To me asure th e thickness of the film for the growth rate, X ray reflectometry (XRR) was ray diffraction (XRD) was performed in a Philips APD 3720, operating with Cu K radiation (40kV and 20mA). To measure sheet resistance of annealed samples, 4 point probe by Alessi Industries four point probe was used To observe the appearance of etch pit, SEM was used by JEOL JSM 6335F. To observe the change of interfacial layer of Cu/TaN/Si stack through annealing, transmission electron microscope (TEM) was used by JEOL TEM 2010F, equipped with an X ray energy dispersive spectrometer (EDS). For TEM specimen preparation, focused ion beam (FIB) was used by FEI Strata DB 235.

PAGE 103

103 7. 3 Results and Discussion 7.3.1 ALD of TaN Thin Films 7.3.1.1 Film Growth The effect of the exposure time on the deposition rate of TaN thin films was examined with varying TBTDET exposure time. The TaN film thickness was measured by XRR. The exposure tim e of TBTDET was changed from 6 to 12 s with fixed NH 3 exposure time of 10 s and purge time of 10 s. Figure 7 1 A shows the growth rate in one ALD cycle as a function of exposure time of TBTDET at T growth = 250 C. Figure 7 1 A suggests that the growth ra te increases with the exposure time of TBTDET, and remains constant at 0.26 nm/cycle from 4 to 11 s of exposure ti me indicating it reached the surface saturation region, which is one of the unique features in ALD. The growth rate increased upto 0.35 nm/c ycle above 11 s of exposure time, which indicates the CVD like region. The deposition rate in saturation region was 0.26 nm/cycle, which is 53.6% of lattice constant (0.49 nm) of h TaN phase. The reason of sub monolayer deposition per cycle in ALD as bee n considered to the incomplete saturation coverage of metal containing precursor due to the steric hindrance of adsorbed metal precursors. [117] The thickness of each ALD TaN was measured as a function of growth temperature to verify the process temperature window. All depositions were performed at the self saturation growth conditions ob tained from TBTDET exposure time results above. TBTDET (9 s, 20 sccm), N 2 purge (10 s, 200 sccm), and NH 3 (10 s, 20 sccm) exposure time and flow rate were fixed with changing the growth temperature varied from 200 to 400 C for 80 cycles. Figure 7 1 B sho ws the thickness as a function of growth temperature. The film thickness increased with increasing growth temperature below 200 C. The precursor adsorption on to the surface active site and the reaction between adsorbed precursor and surface species are thermally activated processes. Therefore, the surface saturation is kinetically limited to the thermal energy, which

PAGE 104

104 leads to higher deposition rate at higher temperature. The film thickness remained constant (20 +/ 1 nm) from 200 to 300 C, suggesting the ALD process temperature window. When the growth temperature is high enough to give sufficient thermal energy for the precursor adsorption and surface reactions, the growth rate remains constant as growth temperature goes up. Above 300 C, the film t hickness further increased with increasing growth temperature, indicating the disturbance of self limiting growth. This behavior can be explained by the thermal decomposition of TBTDET, causing a CVD like process. Thermal decomposition of TBTDET has been reported to occur above 300 C [20] 7.3.1.2 Barrier Tests To verify the minimum thickness of ALD TaN that could block the diffusion of Cu ALD TaN layers of 0.7 to 10 nm were tested. The test structures were formed by depositing the various thicknesses of TaN barrier films on Si substrate and covering them with a Cu layer of 100 nm thickness grown by sputter system The samples were then thermally annealed under N 2 at 500 C for 30 min. XRD measurement was used to identify the formation of Cu x Si from annealed samples which reveals the failure of diffusion barrier As shown in Fig. 7 2 A barriers of less than 1.6 nm thickness showed the diffraction peak of Cu x Si, suggesting that the diffusion barrier failed. O ther films thicker than 3.8 nm showed the peaks assigned to TaN and Cu, but not Cu x Si peaks. To confirm the minimum barrier thickness obtained from XRD, cross sectional TEM was us ed on an annealed stack of Cu/3.8 nm TaN/Si. As shown in Fig. 7 2 B there is no diffusion of Cu through TaN and Si after annealing with 3.8 nm thickness of ALD TaN barrier. Both Cu/TaN and TaN/Si interface can be clearly observed without any intermixing between layers, which indicates that a 3.8 nm barrier film blocks the Cu diffusion.

PAGE 105

105 The etch pit test was used to determine the formation of Cu x Si on the Si surface more sensitively. As shown in Fig. 7 3, etch pits are observed on the Si surface with 1.6 nm thickness barrier as a result of barrier failure. On the other hand, 3.8 nm thickness barrier exhibits no etch pit, indicating the critical thickness of ALD TaN barrier is between 1.6 and 3.8 nm. 7.3.2 in situ (IN) Set vs. ex situ (EX) Set 7.3.2.1 M icrostructure of IN and EX Sets The microstructures of as grown Cu/non air exposed TaN/Si (IN) and Cu/air exposed TaN/Si (EX) samples were compared from cross sectional TEM images. As shown in Fig. 7 4 B, there is an interlayer with thickness of about 0.5 to 1 nm between Cu and TaN for EX sample. There is, however, no interlayer found at the interface of Cu/TaN for IN sample (Fig. 7 4 A). The interlayer at the Cu/TaN interface for EX is believed to be TaO x N y which is formed by surface oxidation of TaN t hrough air exposure before loading into the chamber. There is a previous report that TaO x N y forms as a result of the surface oxidation of TaN [158] 7.3.2.2 Barrier Tests To compare the barrier properties of IN and EX sets, the test structures have the same thickness of TaN (4 nm) and Cu (100 nm). Cu/TaN/Si structures were thermally annealed under N 2 at elevated temperature f rom 400 to 650 C for 30 min. XRD measurement s sheet resistance measurement s and etch pit test s were performed on the annealed samples to determine barrier failure temperatures of TaN diffusion barrier for IN and EX sets. XRD measurements were used to identify the formation of Cu x Si on the Si surface due to barrier failure. The TaN barrier with IN set is observed to fail at 630 C (Fig. 7 5 A), while the EX set fails at 650 C (Fig. 7 5 B). The formation of Cu x Si is determined when new reflections ap pear in the XRD pattern near 44.00 to 46.00 2 As the annealing temperature is increased Cu XRD peak intensities generally increase due to the recrystallization of Cu upon annealing.

PAGE 106

106 Cu peak intensities decreas e as Cu x Si peaks appear due to the consum ption of Cu in the formation of silicide [159] In case of the IN set, the Ta 5 Si 3 peak start s to appear at 42.03 2 (JCPDS 06 0594) when Cu x Si peaks are observed. According to the sheet resistance measurement, barriers for the IN and EX sets were stable up to 600 C (Fig. 7 6). Up to annealing at 550 C, the sheet resistances of IN and EX sets decreased gradually a s a result of the grain growth of Cu. This Cu grain growth is shown by the sharpening of Cu peaks in XRD patterns as the annealing temperature is elevated (Fig. 7 5). The EX set generally shows the higher sheet resistance up to 550 C because of the TaO x N y interlayer. After annealing at 615 C, the sheet resistance of the IN set increases more than one order of magnitude. The sheet resistance of the EX set, however, increases markedly after annealing at 630 C. From this result, the barrier failure temp eratures are determined as 615 C for IN set and 630 C for EX set, respectively. The increase of sheet resistance for barrier failure can be explained by the intermixing of Cu and Si through the barrier, and formation of more highly resistive Cu x Si. Fr om XRD and sheet resistance measurements, the change in the sheet resistances is likely to be detectable for lower annealing temperature than the Cu x Si peak detection in XRD. The sheet resistance measurements were done by 4 point probe, which needs smalle r volume of Cu x Si compound than XRD measurement to detect the failure. In order to determine the formation of Cu x Si on Si substrate more sensitively, the etch pit test was used. In the etch pit test, defects caused by the formation of Cu x Si on Si substra te are shown as inverse pyramidal shape etch pit after etching by Secco etchant. As shown in Fig. 7 7, the first etch pits were observed after annealing at 600 C for IN set (Fig. 7 7B), and 630 C for EX set (Fig. 7 7I), respectively. As the annealing t emperature goes up, the number of etch pit s per unit area

PAGE 107

107 increases. Barrier failure temperatures from etch pit tests were lower than those from XRD and sheet resistance measurements. Better sensitivity of the etch pit test for barrier failure temperatur e determination has been previously reported [160, 161] Barrier failure temperatures for IN and EX sets are shown in Table 7 1 with different barrier test methods. Etch pit test is found to be the most sensitive barrier test, while XRD measurement is the least sensitive. EX s et shows a higher barrier failure temperature than IN set over every barrier test method. The TaO x N y interfacial layer formed by air exposure of TaN appears to have a beneficial role in preventing Cu diffusion. To understand the beneficial role of TaO x N y interfacial layer, the failure mechanism of TaN barrier was investigated. It was approximated that the barrier thickness is associated to the average diffusion length at the temperature when the barrier fails. Barrier failure temperatures for different b arrier thickness stacks of IN and EX sets were shown in Fig. 7 8A. The barrier failure temperatures were determined by 4 point probe measurement. In this study, thermal annealing was performed by ramping the annealing temperature at a rate of 1 C/s. T he activation energy (E a ) for diffusion was determined by plotting ln(L 2 /T 2 ) over 1/ k T for each barrier thickness [162] Fig ure 7 8B shows the plot of ln(L 2 /T 2 ) and 1/ k T from L and T data of Fig. 7 8 A From the slope of the plot, E a of diffusion for each barrier was obtained as 1.08 +/ 0.05 and 1.26 +/ 0.02 eV for the IN and EX sets, respectively. The activation energy for both the IN and EX sets are close to previously reported value s of Ea for Cu diffusion along grain boundaries of TaN (1.3 eV) [162, 163] Thus, TaN diffusion barrier failure is mainly due to the diffusion of C u along grain boundaries of TaN, which are generated by the recrystallization of TaN. The difference of E a for IN and EX, however, is 0.2 eV. And, the barrier failure temperature of IN is lower than EX up to 30 C as shown in Table 7 1. The difference ca n be explained by Ta 5 Si 3 formation in failure sample of

PAGE 108

108 IN. As shown in Figs. 7 5 A and B barrier failure sample of IN shows the Ta 5 Si 3 peak, while no Ta 5 Si 3 peak is detected in the EX sample. Ta 5 Si 3 has been previously reported to be found in ternary pha se of Si Ta N because TaN is thermodynamically unstable with respect to Si at 900 K from ternary phase diagram at 900 K [164] Thus, Ta 5 Si 3 in failure IN sample is formed by interfacial reaction of TaN and Si. Ta 5 Si 3 has a b igger grain size than TaN, which gives faster diffusion pathway to Cu along grain boundaries. It is concluded that barrier failure for IN is due to the diffusion of Cu along grain boundaries of TaN and Ta 5 Si 3 formed by interfacial reaction of TaN and Si. T he barrier failure for the EX sample is due to diffusion of Cu along grain boundaries of TaN only. Grain boundaries of TaN and Ta 5 Si 3 for IN sample provide faster diffusion pathways than that of TaN only for EX sample. Thus, IN sample shows lower failu re temperature than EX sample. 7.4 Conclusions TaN films were grown by thermal ALD using TBTDET and NH 3 ALD characteristics such as process windows were shown, and the critical thickness of TaN barrier was determined as 3.8 nm from XRD and TEM measuremen ts. The effect of air exposure on Cu barrier properties of TaN film was investigated by comparison of air exposed TaN to vacuum transferred TaN with stacks of Cu/TaN/Si. XRD, sheet resistance measurements, and the etch pit test were used to determine bar rier failure temperatures of air exposed and vacuum tra nsferred Cu/TaN/Si stacks. It wa s found that air exposed TaN has higher barrier failure temperature consistent with a faster diffusion pathway of grain boundaries of TaN and Ta 5 Si 3 in vacuum transferr ed stacks. TaO x N y interfacial layer appears to have a beneficial role in preventing Cu transport by suppressing the formation of Ta 5 Si 3

PAGE 109

109 Table 7 1. Barrier failure temperatures determined by different barrier testing methods for Cu/TaN/Si stacks of vacu um transferred and air exposed TaN. Cu/TaN/Si T failure by XRD ( C) T failure by R s ( C) T failure by etch pit ( C) Vacuum transferred TaN 630 615 600 Air exposed TaN 650 630 630

PAGE 110

110 Figure 7 1 A) Growth rate of ALD TaN as a function of TBTD ET exposure time Film was deposited at 250 C and 10 s of purge and NH 3 exposure time. B) Thickness of ALD TaN films as a function of growth temperature Film was deposited with 9 s of TBTDET exposure time and 10 s of purge and NH 3 exposure time. Taken from Ref [165]

PAGE 111

111 Fig ure 7 2. A) XRD patterns of Cu/TaN/Si annealed at 500 C for 30 min B) Cross sectional TEM image of Cu/TaN/Si structure. TaN was grown by ALD for 15 cycles at 9 s of TB TDET exposure time a t 300 C Taken from Ref [165]

PAGE 112

112 Fig ure 7 3. SEM images of Si surface after annealing Cu/TaN/Si at 500 C for 30 min followed by the removal of Cu and TaN, and Secco etchin g for A) 10 cycles (16 ) TaN film. B ) 15 cycles (38 ) TaN film. Taken from Ref [165] Figure 7 4. Cross sectional TEM image of as grown Cu/TaN/Si structure of A) vacuum transferred TaN. B) ai r exposed TaN.

PAGE 113

113 Figure 7 5. XRD patterns of annealed Cu/TaN/Si structure of A) vacuum transferred TaN. B) air exposed TaN at elevated annealing temperatures for 30 min.

PAGE 114

114 Figure 7 6. Sheet resistances of the Cu/TaN/Si structures for vacuum tran sferred and air exposed TaN as a function of the annealing temperature. Figure 7 7. SEM image of etch pits observed on Si surface of Cu/TaN/Si for vacuum transferred TaN at annealing temperature of A) 585 C B) 600 C C) 615 C D) 630 C and E) 650 C Si surface of Cu/TaN/Si for air exposed TaN at annealing temperature of F) 585 C G) 600 C H) 615 C I) 630 C and J) 650 C EX IN

PAGE 115

115 Figure 7 8. A) Barrier failure temperatures as a function of TaN barrier thickness. B) Plot of ln(L 2 / T 2 ) vs 1/ k T for Cu/TaN/Si stacks of vacuum transferred and air exposed TaN.

PAGE 116

116 CHAPTER 8 FUTURE WORK AND SUGGESTIONS 8.1 Concerns of ALD GaN for Device Application This work demonstrated the growth of GaN on Si with ALD for the first time ALD charact eristics such as self saturation window and temperature window were revealed F ilm properties such as microstructure, surface morphology, and Cl incorporation were also investigated. As a pioneering research on GaN ALD on Si, this work covers many possib le issues in the application to optoelectronic devices. T he t emperature window makes the deposition process much more controllable. The lack of Cl incorporation under ALD process window solves one of the most detrimental drawbacks of halide ALD. However more device related issues such as optical and electric al properties were not considered here. To be employed in LED or laser diode s the PL emission wavelength and intensity needs to be considered. Also, carrier properties such as concentration and m obility need to be studied using Hall measurements. In addition, the crystallinity of the film needs to be improved. As shown previously, the ALD GaN on Si showed two growth orientations which are not preferred for device application. Introducing buffer layers or sapphire substrate c ould result in better crystallinity. 8.2 ALD Doping of GaN For optoelectronic device application s p type do p ing of Ga N was performed in this work. The m ost commonly used p type dopant, Mg, was employed for HVP E of GaN The optimized growth condition for obtain ing the highest hole concentration, which was 3.2 10 17 cm 3 at 15 sccm of Cp 2 Mg flow rate PL spectra showed the emission peak at 3.12 eV with high intensity. The electrical and optical properties of Mg doped GaN in this work seem to be practical for LED

PAGE 117

117 application. However, HVPE of p GaN is not the most suitable for the application to novel design LED which is proposed in this work. As mentioned in the introduction chapter, p GaN wa s deposited on n GaN nanorod, and filled the space between nanorods in a novel design LED structure. For better performance of this LED, p GaN needs to be deposited on high aspect ratio n GaN nanorod s with good conformality and step coverage. It is known that ALD is a superior deposition for excellent conformality, step coverage, and uniformity. Therefore, p type doping of GaN using ALD method needs to be developed for better performance of nanorod based LED. The biggest difference between HVPE and ALD doping is the flow sequence of precursors. For HVPE, both GaCl 3 and Cp 2 Mg were supplied to the growth zone with NH 3 In ALD, the co flow of GaCl 3 and Cp 2 Mg need s to be supplied to the growth zone separating from NH 3 with a purge step in between. For the better doping level controllability, GaCl 3 and Cp 2 Mg can be supplied to the growth zone through separate exposure steps. Various sequences for the exposure of GaCl 3 Cp 2 Mg, and NH 3 can be tested to obtain optimum doping level or optical property. Also, the control of the Cp 2 Mg exposure step would make ALD doping extendable to the delta doping of GaN which is a promising doping methods to achieve MQW GaN. 8.3 Device Fabrication of Nanorod based LED As mentioned above, improvements in the optical, electrical, and stru ctural quality of ALD GaN are required to achieve nanorod based LED structure s Also, ALD doping of GaN needs to be developed for the better performance of the LED s Even if these materials and processing requirements were satisfied, there is still a roa dblock to achieve the proposed LED structure, which is the device fabrication issue. The basic unit of nanorod based LED can be fabricated by the deposition process es such as nanorod growth and p GaN deposition. However, there are more device components to be fabricated.

PAGE 118

118 First of all, the insulating oxide needs to be deposited on the surface space in between nanorods. If there is an electric al connection between the n GaN nanorod s the device could not convert the electric gain to light emission. In o rder to prevent this failure, each n GaN nanorod needs to be electrically insulated by insulating oxide. Thus, a proper deposition process and material for this insulating oxide are required to be developed. Secondly, the top side of each nanorod needs to expose the interface between p GaN and n GaN nanorod or active layer, if necessary. After the deposition of p GaN, the top side of each nanorod becomes p GaN ended. The emission light, however, is coming from the interface of p n GaN. Thus, it is nec essary to expose the interface of p n GaN to the top side of nanorod. The most promising fabrication technique for this process is CMP. To prevent the collapse of the vertically aligned nanorods, the space between nanorods needs to be filled by p GaN. In addition, p and n contacts need to be fabricated on p and n GaN sides, respectively. Particularly, the p contact has to be fabricated on the limited area of p GaN to maximize t he light emitting surface. The transparent top surface also needs to be fabricated with proper material and process. 8.4 Control of TaO x N y for Better Diffusion Barrier Properties In this work, the effect of air exposure to ALD TaN diffusion barrier properties was studied The air exposed TaN exhibited the formation of TaO x N y in the interface of TaN and Cu. And, it was revealed by various barrier tests that TaO x N y interfacial layer showed a beneficial effect on the thermal stability of TaN diffusion barrier. On the other hand, TaO x N y contributed to the increase of electri cal resistance. In this work, TaO x N y is formed by the exposure of TaN to ambient air. If the thickness and composition ratio of TaO x N y are controlled by introducing separate O source into current ALD process the optimum thickness and composition ratio c ould be found to achieve the best thermal stability with minimum electrical res is tance increase effect.

PAGE 119

119 Then, TaO x N y c ould be employed as the next promising diffusion barrier material for Cu metallization.

PAGE 120

120 LIST OF REFERENCES [1] S. J. Rosner, E. C. Carr, M. J. Ludowise, G. Girolami, H. I. Erikson, Appl. Phys. Lett. 70 (1997) 420 422. [2] J. H. You, H. T. Johnson, J. Appl. Phys. 101 (2007) 023516 6. [3] Y. C. Shen, G. O. Mueller, S. Watanabe, N. F. Gardner, A. Munkholm, M. R. Krames, Appl. Phys. Lett. 91 (2007) 141101 3. [4] G. Bai, S. Wittenbrock, V. Ochoa, R. Villasol, C. Chiang, T. Marieb, D. Gardner, C. Mu, D. Fraser, M. Bohr, in Materials Research Society Symposium Proceedings Vol. 403, 1996, p. 501 506. [5] J. Xie, X. Ni, Q. Fan, R. Shimada, U. Ozgur, H. Morkoc, Appl. Phys. Lett. 93 (2008) 121107 3. [6] R. Leutenecker, B. Froschle, U. CaoMinh, P. Ramm, Thin Solid Films 270 (1995) 621 626. [7] O. Ambacher, J. Phys. D: Appl. Phys. 31 (1998) 2653 2710. [8] C. G. Van de Wa lle, J. Neugebauer, J. Appl. Phys. 95 (2004) 3851 3879. [9] Interconnect, International Technology Roadmap for Semiconductors ( http://www.itrs.net) 2007. [10] S. M. Rossnagel, J. Hopwood, Appl. Phys. Lett. 63 (1993) 3 285 3287. [11] J. Hillman, J. Hautala, S. Caliendo, in Proceedings of Advanced Metallization Conference Vol. 313, San Diego, CA, 2000. [12] X. Chen, G. G. Peterson, C. Goldberg, G. Nuesca, H. L. Frisch, A. E. Kaloyeros, B. Arkles, J. Sullivan, J. Mater. R es. 14 (1999) 2043 2052. [13] A. Sherman, J. Electrochem. Soc. 137 (1990) 1892 1897. [14] N. Yokoyama, K. Hinode, Y. Homma, J. Electrochem. Soc. 138 (1991) 190 195. [15] P. Kapur, J. P. McVittie, K. C. Saraswat, IEEE Transactions on Electron Devices 49 (20 02) 590 597. [16] L. Hiltunen, M. Leskela, M. Makela, L. Niinisto, E. Nykanen, P. Soininen, Thin Solid Films 166 (1988) 149. [17] T. Aaltonen, P. Alen, M. Ritala, M. Leskela, Chem. Vapor Depos. 9 (2003) 45 49. [18] Y. Y. Wu, A. Kohn, M. Eizenberg, J. Appl Phys. 95 (2004) 6167.

PAGE 121

121 [19] J. Park, H. Park, S. Kang, J. Electrochemical Soc. 149 (2002) C28. [20] O. van der Straten, Y. Zhu, K. Dunn, E. T. Eisenbraun, A. E. Kaloyeros, J. Mater. Res. 19 (2004) 447 453. [21] L. Clevengern, A. Mutscheller, J. Harper, C. Cabral Jr, K. Bark, J. Appl. Phys. 72 (1992) 4918. [22] L. Clevengern, N. Bojarczuk, K. Holloway, J. Harper, C. Cabral Jr., R. Schad, F. Cardone, L. Stolt, J. Appl. Phys. 73 (1993) 300. [23] J. Kwak, H. Baik, J. Kim, S. Lee, Appl. Phys. Lett. 71 (1997) 24 51. [24] M. Wang, Y. Lin, M. Chen, J. Electrochem. Soc. 145 (1998) 2538. [25] M. Juppo, M. Ritala, M. Leskela, J. Electrochem. Soc. 147 (2000) 3377 3381. [26] H. Kim, J. Vac. Sci. Technol. B 21 (2003) 2232 2261. [27] H. Chang, L. Chang, F. Chen, J. Kai, E. Tzou, J. Fu, Z. Xu, J. Egermeier, F. Chen, Acta Mater. 46 (1998) 2431. [28] W. C. Johnson, J. B. Parsons, J. Phys. Chem. 36 (1932) 2588 2594. [29] R. Juza, H. Hahn, Anorg. Allg. Chem. 239 (1938) 282 287. [30] H. G. Grimmeiss, H. Koelmans, Z. Naturforsch. 14 (1959) 264. [31] H. P. Maruska, J. J. Tietjen, Appl. Phys. Lett. 15 (1969) 327 329. [32] S. Yoshida, S. Misawa, S. Gonda, Appl. Phys. Lett. 42 (1983) 427 429. [33] H. Amano, N. Sawaki, I. Akasaki, Y. Toyoda, Appl. Phys. Lett. 48 (1986) 353 355. [34] H. Amano, M. Kito, K. Hiramatsu, I. Akasaki, Jpn. J. Appl. Phys. 28 (1989) L2112. [35] S. Nakamura, T. Mukai, M. Senoh, N. Iwasa, Jpn. J. Appl. Phys. 31 (1991) L139. [36] W. Gtz, N. M. Johnson, J. Walker, D. P. Bour, H. Amano, I. Akasaki, Appl. Phys. Lett. 6 7 (1995) 2666 2668. [37] F. Wang, C. F. Lyons, K. B. Nguyen, S. A. Bell, H. J. Levinson, C. Y. Yang, Us Patent 6200907, 2001 [38] Y. Kato, S. Kitamura, K. Hiramatsu, N. Sawaki, J. Cryst. Growth 144 (1994) 133 140. [39] S. Nakamura, T. Mukai, M. Senoh, Appl Phys. Lett. 64 (1994) 1687 1689.

PAGE 122

122 [40] S. Nakamura, M. Senoh, N. Iwasa, S. i. Nagahama, Appl. Phys. Lett. 67 (1995) 1868 1870. [41] S. Nakamura, M. Senoh, N. Iwasa, S. i. Nagahama, T. Yamada, T. Mukai, Jpn. J. Appl. Phys. 34 (1995) L1332. [42] S. Nakamura M. Seno, S. i. Nagahama, N. Iwasa, T. Yamada, T. Matsushita, H. Kiyoku, Y. Sugimoto, Jpn. J. Appl. Phys. 35 (1996) L217. [43] Y. Shimooka, T. Iijima, S. Nakamura, K. Suguro, 36 (1997) 1589 1592. [44] S. Nakamura, M. Senoh, S. i. Nagahama, N. Iwasa, T. Y amada, T. Matsushita, H. Kiyoku, Y. Sugimoto, T. Kozaki, H. Umemoto, M. Sano, K. Chocho, Appl. Phys. Lett. 72 (1998) 211 213. [45] J. Pankove, E. Miller, J. Berkeyheiser, RCA Rev. 32 (1971) 383. [46] H. P. Maruska, D. A. Stevenson, J. I. Pankove, Appl. Phy s. Lett. 22 (1973) 303 305. [47] J. I. Pankove, M. A. Lampert, Phys. Rev. Lett. 33 (1974) 361. [48] J. Pankove, RCA Rev. 34 (1973) 336. [49] H. Amano, M. Kito, K. Hiramatsu, I. Akasaki, Jpn. J. Appl. Phys. 28 (1989). [50] S. Nakamura, T. Mukai, M. Senoh, Jpn. J. Appl. Phys. 30 (1991) L1998. [51] S. Nakamura, T. Mukai, M. Senoh, Jpn. J. Appl. Phys. 32 (1993) L16. [52] S. Nakamura, J. Vac. Sci. Technol. A 13 (1995) 705 710. [53] M. G. Craford, in IEEE Circuits Devices V ol. 8, 1992, p. 25 29. [54] M. Meyer, in Semicond. Comp. Vol. 2, 1996, p. 32. [55] in Industry News PennWell Corporation, Tulsa, OK, 2009. [56] M. A. Khan, R. A. Skogman, J. M. Van Hove, D. T. Olson, J. N. Kuznia, Appl. Phys. Lett. 60 (1992) 1366 1368. [ 57] N. H. Karam, T. Parodos, P. Colter, D. McNulty, W. Rowland, J. Schetzina, N. El Masry, S. M. Bedair, Appl. Phys. Lett. 67 (1995) 94 96. [58] A. Koukitu, J. N. Kuznia, T. Taki, H. Seki, Jpn. J. Appl. Phys. 38 (1999) 4980. [59] H. Tsuchiya, M. Akamatsu, M. Ishida, F. Hasegawa, Jpn. J. Appl. Phys. 35 (1996) L748. [60] N. Koide, H. Kato, M. Sassa, S. Yamasaki, K. Manabe, M. Hashimoto, H. Amano, K. Hiramatsu, I. Akasaki, J. Cryst. Growth 115 (1991) 639 642.

PAGE 123

123 [61] S. Nakayama, T. Mukai, M. Senoh, Jpn. J. Appl. Phys. 31 (1992) 2883. [62] W. Gtz, N. M. Johnson, C. Chen, H. Liu, C. Kuo, W. Imler, Appl. Phys. Lett. 68 (1996) 3144 3146. [63] L. B. Rowland, K. Doverspike, Appl. Phys. Lett. 66 (1995) 1495. [64] B. C. Chung, M. Gershenzon, J. Appl. Phys. 72 (1992) 651 659. [65] H. Sato, T. Minami, E. Yamada, M. Ishii, S. Takata, J. Appl. Phys. 75 (1994) 1405 1409. [66] J. C. Zolper, R. G. Wilson, Appl. Phys. Lett. 68 (1996) 1945. [67] A. Edwards, M. Rao, B. Molnar, A. Wickenden, O. W. Holland, P. H. Chi, J. Elec. Mat. 26 (1997) 334. [68] W. Gtz, N. M. Johnson, J. Walker, D. P. Bour, R. A. Street, Appl. Phys. Lett. 68 (1996) 667 669. [69] M. S. Brandt, N. M. Johnson, R. J. Molnar, R. Singh, T. D. Moustakas, Appl. Phys. Lett. 64 (1994) 2264 2266. [70] J. Neugebauer, C. G Van de Walle, Phys. Rev. Lett. 75 (1995) 4452. [71] J. M. Myoung, K. H. Shim, C. Kim, O. Gluschenkov, K. Kim, S. Kim, D. A. Turnbull, S. G. Bishop, Appl. Phys. Lett. 69 (1996) 2722 2724. [72] P. Bergman, G. Ying, B. Monemar, P. O. Holtz, J. Appl. Phys. 6 1 (1987) 4589 4592. [73] J. Neugebauer, C. G. Van de Walle, J. Appl. Phys. 85 (1999) 3003. [74] O. Brandt, H. Yang, H. Kostial, K. H. Ploog, Appl. Phys. Lett. 69 (1996) 2707 2709. [75] J. C. Zolper, R. G. Wilson, S. J. Pearton, R. A. Stall, Appl. Phys. Let t. 68 (1996) 1945 1947. [76] Semiconductors. Physics of Group IV Elements and III V Compounds, Landolt Brnstein; O. Madelung, Ed.; Springer: Berlin, 1982; Vol. 17. [77] S. P. Murarka, I. V. Verner, R. J. Gutmann, Cooper Fundamental Mechanisms for Microele ctronic Applications; John Wiley and Sons: New York, NY, 2000. [78] R. I. M. Dang, N. Shigyo, IEEE Electron Devices Lett. EDL 2 (1981) 196. [79] S. P. Jeng, R. H. Havemann, M. C. Chang, Mat. Res. Soc. Symp. Proc. 337 (1994) 24. [80] V. G. Oklobdzija, IBM J Res. Dev. 47 (2003) 567.

PAGE 124

124 [81] A. Scorzoni, I. De Munari, M. Impronta, R. Balboni, N. Kelaidis, http://www.diei.unipg.it/RICERCA/www_em.htm 1998. [82] M. Sakimoto, T. Fujii, H. Yamaguchi, K. Eg uchi, in Proceedings of IEEE Reliability Physics Symposium Las Vegas, NV, 1995, p. 333. [83] A. A. Istratov, H. Hedemann, M. Seibt, O. F. Vyvenko, W. Schroter, T. Heiser, C. Flink, H. Hieslmair, E. R. Weber, J. Electrochem. Soc. 145 (1998) 3889 3898. [84] C. Flink, H. Feick, S. A. McHugo, W. Seifert, H. Hieslmair, T. Heiser, A. A. Istratov, E. R. Weber, Phys. Rev. Lett. 85 (2000) 4900. [85] A. E. Kaloyeros, X. Chen, T. Stark, K. Kumar, S. C. Seo, G. G. Peterson, H. L. Frisch, B. Arkles, J. Sullivan, J. Ele ctrochem. Soc. 146 (1999) 170 176. [86] S. L. Cho, K. B. Kim, S. H. Min, H. K. Shin, S. D. Kimd, J. Electrochem. Soc. 146 (1999) 3724 3730. [87] R. Fix, R. G. Gordon, D. M. Hoffman, Chem. Mater. 5 (1993) 614 619. [88] M. H. Tsai, S. C. Sun, H. T. Chiu, C. E. Tsai, S. H. Chuang, Appl. Phys. Lett. 67 (1995) 1128 1130. [89] L. Hiltunen, M. Leskel, M. Mkel, L. Niinist, E. Nyknen, P. Soininen, Thin Solid Films 166 (1988) 149 154. [90] M. Ritala, P. Kalsi, D. Riihela, K. Kukli, M. Leskela, J. Jokinen, Chem. Mater. 11 (1999) 1712 1718. [91] H. Kim, A. J. Kellock, S. M. Rossnagel, J. Appl. Phys. 92 (2002) 7080 7085. [92] E. Eisenbraun, O. van der Straten, Y. Zhu, K. Dovidenko, A. E. Kaloyeros, in IEEE International Interconnect Technology Conference Burlingame CA, 2001, p. 207. [93] K. I. Choi, B. H. Kim, S. W. Lee, J. M. Lee, in Interconnect Technology Conference, 2003. Proceedings of the IEEE 2003 International Burlingame, CA, 2003, p. 129. [94] J. S. Park, H. S. Park, S. W. Kang, J. Electrochem. Soc. 149 ( 2002) C28 C32. [95] Y. Y. Wu, A. Kohn, M. Eizenberg, J. Appl. Phys. 95 (2004) 6167 6174. [96] A. Furuya, H. Tsuda, S. Ogawa, J. Vac. Sci. Technol. B 23 (2005) 979 983. [97] A. Furuya, N. Ohtsuka, N. Ohashi, S. Kondo, S. Ogawa, J. Vac. Sci. Technol. A 24 (2 006) 103 105. [98] W. J. Maeng, S. J. Park, H. Kim, J. Vac. Sci. Technol. B 24 (2006) 2276 2281.

PAGE 125

125 [99] H. Kim, C. Detavenier, O. van der Straten, S. M. Rossnagel, A. J. Kellock, D. G. Park, J. Appl. Phys. 98 (2005) 014308 8. [100] S. Subbanna, P. Agnello, E Crabbe, R. Schulz, S. Wu, K. Tallman, M. J. Saccamango, S. Greco, V. McGahay, A. J. Allen, B. Chen, T. Cotler, E. Eld, J. Lasky, H. Ng, A. Ray, J. Snare, L. Su, D. Sunderland, J. Sun, B. Davari, in IEEE International Electron Devices Conference San Fran cisco, CA, 1996, p. 275. [101] C. K. Hu, K. Y. Lee, C. Uzoh, K. Chan, S. M. rossnagel, in Proceedings of Materials Research Society Symposium San Francisco, CA, 1998, p. 507. [102] R. W. Herztberg, Deformation and Fracture Mechanics of Engineering Materia ls; 3rd ed.; Wiley & Sons: New York, 1989. [103] F. W. Young, J. V. Cathcart, A. T. Gwathmey, Atca Metal. 4 (1956) 145. [104] C. A. Chang, J. Appl. Phys. 67 (1990) 566 569. [105] P. S. Ho, Thin Solid Films 96 (1982) 301 316. [106] S. Kirkalday, D. J. Young Diffusion in the Condensated State; Institute of Metal: London, 1987. [107] D. Gupta, in Encyclopedia of Applied Physics Vol. 5, John Wiley & Sons, 1993, p. 75. [108] K. C. Park, K. B. Kim, I. Raaijmakers, K. Ngan, J. Appl. Phys. 80 (1996) 5674 5681. [1 09] M. Stavrev, D. Fischer, F. Praessler, C. Wenzel, K. Drescher, J. Vac. Sci. Technol. A 17 (1999) 993 1001. [110] N. A. Gjostein, Diffusion in Metals; Am. Soc. Met.: Westerville, OH, 1973. [111] A. E. Kaloyeros, E. Eisenbraun, Ann. Rev. Mat. Sci. 30 (200 0) 363 385. [112] R. Rosenberg, D. C. Edelstein, C. K. Hu, K. P. Rodbell, Annu. Rev. Mater. Sci. 30 (2000) 229 262. [113] S. Strite, H. Morkoc, J. Vac. Sci. Technol. B 10 (1992) 1237 1266. [114] T. C. Shen, G. B. Gao, H. Morkoc, J. Vac. Sci. Technol. B 10 (1992) 2113 2132. [115] N. Yoshimoto, T. Matsuoka, T. Sassaki, A. Katsui, Appl. Phys. Lett. 59 (1991) 2251 2253. [116] B. T. McDermott, K. G. Reid, N. A. El Masry, S. M. Bedair, W. M. Duncan, X. Yin, F. H. Pollak, Appl. Phys. Lett. 56 (1990) 1172 1174. [11 7] H. Kim, J. Vac. Sci. Technol. B 21 (2003) 2231 2261.

PAGE 126

126 [118] Y. Tsai, J. Gong, T. Lin, H. Lin, Y. Chen, K. Lin, Appl. Surf. Sci. 252 (2006) 3454. [119] A. Koukitu, Y. Kumagai, T. Taki, H. Seki, Jpn. J. Appl. Phys. 38 (1999) 4980 4982. [120] R. Puurunen, J Appl. Phys. 97 (2005) 121301. [121] J. Lim, H. Park, S. Kang, J. Appl. Phys. 88 (2000) 6327. [122] S. b. Kang, H. s. Lim, Y. s. Chae, I. s. Jeon, G. h. Choi, US Patent 6287965, 2001 [123] G. Beyer, A. Satta, J. Schuhmacher, K. Maex, W. Besling, O. Kilpel a, H. Sprey, G. Tempel, 64 (2002) 233 245. [124] M. L. Green, M. Ho, B. Busch, G. D. Wilk, T. Sorsch, J. Appl. Phys. 92 (2002) 7168. [125] G. Stringfellow, Organometallic Vapor Phase Epitaxy; Second ed.; Academic Press: London, 1999. [126] R. Pantel, G. A uvert, G. Mascarin, Microelectron. Eng. 37 8 (1997) 49 57. [127] A. Kovacs, C. Esterhuysen, G. Frenking, 11 (2005) 1813 1825. [128] D. King, X. Du, A. Cavanagh, A. Weimer, Nanotechnology 19 (2008) 445401. [129] J. Uhm, H. Jeon, Jpn. J. Appl. Phys., Part 1 40 (2001) 4657 4660. [130] S. Krukowski, P. Kempisty, P. Strak, J. Cryst. Growth 310 (2008) 1391. [131] S. Nakamura, Y. Harada, M. Seno, Appl. Phys. Lett. 58 (1991) 2021 2023. [132] R. C. Powell, N. E. Lee, Appl. Phys. Lett. 60 (1992) 2505. [133] M. Rital a, M. Leskel, In Handbook of Thin Film Materials; Nalwa, H. S., Ed.; Academic Press: San Diego, 2001; Vol. 1, p 2. [134] O. H. Kim, D. Kim, T. Anderson, J. Vac. Sci. Technol. A 27 (2009) 923 928. [135] M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scus eria, M. A. Robb, J. R. Cheeseman, J. J. A. Montgomery, T. Vreven, K. N. Kudin, J. C. Burant, J. M. Millam, S. S. Iyengar, J. Tomasi, V. Barone, B. Mennucci, M. Cossi, G. Scalmani, N. Rega, G. A. Petersson, H. Nakatsuji, M. Hada, M. Ehara, K. Toyota, R. Fu kuda, 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. Ochter ski, P. Y. Ayala, K. Morokuma, G. A. Voth, P. Salvador, J. J. Dannenberg, V. G. Zakrzewski, S. Dapprich, A. D. Daniels, M. C. Strain, O. Farkas, D. K. Malick, A. D. Rabuck, K. Raghavachari, J. B. Foresman, J. V. Ortiz, Q. Cui, A. G. Baboul, S. Clifford, J. Cioslowski, B. B. Stefanov, G. Liu, A. Liashenko, P. Piskorz, I. Komaromi, R. L. Martin, D. J. Fox, T. Keith, M. A. Al Laham,

PAGE 127

127 C. Y. Peng, A. Nanayakkara, M. Challacombe, P. M. W. Gill, B. Johnson, W. Chen, M. W. Wong, C. Gonzalez, J. A. Pople Gaussian 03, Revision B.04 (2004) [136] A. D. Becke, J. Chem. Phys. 98 (1993) 1372 1377. [137] P. J. Stephens, F. J. Devlin, C. F. Chabalowski, M. J. Frisch, J. Phys. Chem. 98 (1994) 11623 11627. [138] A. Heyman, C. B. Musgrave, J. Phys. Chem. B 108 (2004) 5718 5725 [139] P. Kozodoy, S. Keller, S. P. DenBaars, U. K. Mishra, J. Cryst. Growth 195 (1998) 265 269. [140] H. Tokunaga, I. Waki, A. Yamaguchi, N. Akutsu, K. Matsumoto, J. Cryst. Growth 189 190 (1998) 519 522. [141] K. S. Kim, G. M. Yang, H. J. Lee, Solid Stat e Electron. 43 (1999) 1807 1812. [142] B. Schineller, A. Guttzeit, P. H. Lim, M. Schwambera, K. Heime, O. Schn, M. Heuken, J. Cryst. Growth 195 (1998) 274 279. [143] A. Usikov, O. Kovalenkov, V. Soukhoveev, V. Ivantsov, A. Syrkin, V. Dmitriev, A. Yu. Niki forov, S. G. Sundaresan, S. J. Jeliazkov, A. V. Davydov, Phys. Status Solidi C 5 (2008) 1829 1831. [144] B. Heying, X. H. Wu, S. Keller, Y. Li, D. Kapolnek, B. P. Keller, S. P. DenBaars, J. S. Speck, Appl. Phys. Lett. 68 (1996) 643 645. [145] C. R. Lee, J. Y. Leem, S. K. Noh, S. E. Park, J. I. Lee, C. S. Kim, S. J. Son, K. Y. Leem, J. Cryst. Growth 193 (1998) 300 304. [146] S. Fischer, C. Wetzel, E. E. Haller, B. K. Meyer, Appl. Phys. Lett. 67 (1995) 1298 1300. [147] S. J. Pearton, J. W. Lee, C. Yuan, Appl. Phys. Lett. 68 (1996) 2690 2692. [148] D. J. Chadi, Appl. Phys. Lett. 71 (1997) 2970 2971. [149] I. Rogozin, A. Georgobiani, M. Kotlyarevsky, Inorg. Mater. 44 (2008) 1208 1213. [150] S. Limpijumnong, C. G. Van de Walle, Phys. Rev. B 69 (2004) 035207. [151 ] J. W. Allen, Semicon. Sci. Technol. 10 (1995) 1049 1064. [152] U. Kaufmann, M. Kunzer, M. Maier, H. Obloh, A. Ramakrishnan, B. Santic, P. Schlotter, Appl. Phys. Lett. 72 (1998) 1326 1328. [153] Y. Qi, C. Musante, K. Lau, L. Smith, R. Odedra, R. Kanjolia, J. Electron. Mater. 30 (2001) 1382 1386.

PAGE 128

128 [154] E. Kolawa, P. J. Pokela, J. S. Reid, J. S. Chen, M. A. Nicolet, Appl. Surf. Sci. 53 (1991) 373 376. [155] A. A. Istratov, C. Flink, H. Hieslmair, E. R. Weber, T. Heiser, 81 (1998) 1243 1246. [156] M. Wittmer Appl. Phys. Lett. 36 (1980) 456 458. [157] T. Suntola, 216 (1992) 84 89. [158] A. Ibidunni, R. Masaitis, R. Opila, A. Davenport, H. Isaacs, J. Taylor, Surf. Interface Anal. 20 (1993) 559. [159] D. J. Kim, Y. B. Jung, M. B. Lee, Y. H. Lee, J. H. Lee, J. H. Lee, Thin Solid Films 372 (2000) 276. [160] K. C. Park, K. B. Kim, J. Electrochem. Soc. 142 (1995) 3109. [161] J. H. Koo, J. W. Lee, T. H. Doh, Y. D. Kim, Y. D. Kim, H. T. Jeon, J. Vac. Sci. Technol. A 19 (2001) 2831. [162] H. Kim, C. Lavoie, M. Copel, V. Narayanan, D. G. Park, S. M. Rossnagel, J. Appl. Phys. 95 (2004) 5848. [163] T. Oku, E. Kawakami, M. Uekubo, K. Takahiro, S. Yamaguchi, M. Murakami, Appl. Surf. Sci. 99 (1996) 265 272. [164] K. H. Min, K. C. Chun, K. B. Kim, J. Vac. Sci. Technol. B 14 ( 1996) 3263 3269. [165] K. C. Kim, in Department of Chemical Engineering University of Florida, Gainesville, FL, 2006.

PAGE 129

129 BIOGRAPHICAL SKETCH Oh Hyun Kim was born in Daejon, South Korea in 1978. He is son of Ukdong Kim and Kyoungja Ko, and brother of Ansoon Kim and Sookheun Kim. He received Bachelor of Science degree from Seoul National University, South Korea in 2004. During his undergraduate program he completed m ilitary service in U.S. Army as aircraft fuel specialist between 1999 and 2001. In 2004, he joined the PhD program in Chemical Engineering Department at University of Florida. Upon graduation, he plans to work for Intel Corporation in Portland, Oregon as a senior process engineer; his main role is to develop the next generatio n epitaxial process for strained silicon/germanium.