xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 001670391
007 cr mnu|||uuuuu
008 051123s2005 flu sbm s000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0001263
Zantye, Parshuram B.
Processing, reliability and integration issues in chemical mechanical planarization
h [electronic resource] /
by Parshuram B. Zantye.
[Tampa, Fla.] :
b University of South Florida,
Thesis (Ph.D.)--University of South Florida, 2005.
Includes bibliographical references.
Text (Electronic thesis) in PDF format.
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
Title from PDF of title page.
Document formatted into pages; contains 231 pages.
ABSTRACT: Global planarization is one of the major demands of the semiconductor industry. Chemical mechanical polishing (CMP) is the planarization method of choice use to achieve the required stringent tolerances essential for successful fabrication of next generation Integrated Circuits (IC). The predominant reason for CMP defects is the shear and normal stresses during polishing to which the material is subjected. Understanding the process of CMP and factor that contribute to overall stress addition during polishing requires an approach that encompasses all the four major categories of variables, namely: a) machine parameters, b) material properties, c) polishing pad characteristics, and d) polishing slurry performance.In this research, we studied the utilized in-situ technique involving acoustic emission (AE) signal monitoring and coefficient of friction (COF) monitoring using a CETRTM Bench Top CMP Tester to evaluate the impact of variation in machine parameters on the CMP process. The mechanical and tribological properties of different candidate materials have been evaluated bring potential challenges in their integration to the fore. The study also involves destructive and non destructive testing of polishing pads performed for characterization and optimization of polishing pad architecture. Finally, the investigation concludes proposing novel nanoparticle CMP slurry which has a predominant chemical component in its polishing mechanism.
Adviser: Prof. Ashok Kumar.
Chemical mechanical polishing (cmp).
x Mechanical Engineering
t USF Electronic Theses and Dissertations.
Processing, Reliability And Integration Issu es In Chemical Mech anical Planarization By Parshuram B. Zantye A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Mechanical Engineering Department of Mechanical Engineering College of Engineering University of South Florida Major Professor: Ashok Kumar, Ph.D. Rajiv Dubey, Ph.D. Muhammad Rahman, Ph.D. Tapas Das, Ph.D. Julie Harmon, Ph.D. John Bumgarner, Ph.D. Yaw Obeng, Ph.D. Date of Approval: July 15 2005 Keywords: chemical mechanical polishing (cmp), tribology, polishing pad, slurry, damascene, metrology, end point, de lamination, ultrasound, reliability Copyright 2005, Parshuram B. Zantye
DEDICATION My entire work in the field of Chemical Mechanical Planarization including this dissertation is dedicated to my Parents, Mrs. Vrinda B. Zantye and Mr. Balkrishna P. Zantye.
ACKNOWLEDGMENTS I would like to generously thank my advisor Prof. Ashok Kumar for giving me an opportunity to pursue research under his able guidance and with his support throughout during my MastersÂ’ as well as Doctoral studies at Un iversity of South Florida. I would like to thank the Chai rman of Department of Mechanical Engineering, Prof. Rajiv Dubey for his constant encouragemen t and assistance throughout my tenure as a graduate student, and also for serving my di ssertation committee. A special thanks goes to Dr. Yaw Obeng (Texas Instruments, Dallas TX) for his invaluable help in terms of sharing technical know how all along the course of my rese arch during his tenure earlier at psiloQuest Inc, Orlando, FL and later at Texas Instruments, Inc. I wish to thank the Chair of my dissertation committee, Prof. S unil Saigal for serving on my committee and giving me invaluable suggestions for this pa st year. I would like to highly acknowledge the help of Dr. Michael Kovac, Director, Nanomater ials and Nanomanufacturing Research Center (NNRC) and all the team of engineers, technicians and support staff working under his able leadership at the NNRC. Special thanks also go to all my committee members: Dr. Mohammed Rahman, Dr Tapas Das, Dr. Julie Harmon, and Dr. John Bumgarner. Also, I wish to thank Dr. Sergei Ostapenko, who for logistic reasons could not serve on my committee but was act ively involved with my research. The financial support for this research cam e from NSF CAREER Grant #9983535 (Ashok Kumar) and NSF GOALI Grant #DMII 0218141. Part of this research was also supported by USF-Agere High-Tech Corridor Grant #2112142LO, International Sematech Grant # 2112-139 and psiloQuest, Inc. Florida High Tech Corridor Grant. I would also like to
thank Dr. Ashok Das from Applied Materials, Dr. Jeffery Lee from Intel Corporation, Dr. M. Sanganeria from Novellus Systems, a nd Dr. Jiro Yota from Skyworks Inc., for providing valuable inputs, mate rials and supplies. Finally, my parents, family, friends, former Post Doctoral fellows (Dr. Arun K. Sikder and Dr. Arun Kumar), and colleagues, especially S. Mudhivarthi, also deserve a ve ry special thank you note at this time for constantly supporting me throughout the thick and thin of my life.
i TABLE OF CONTENTS LIST OF TABLES vii LIST OF FIGURES viii ABSTRACT xvii CHAPTER 1: INTRODUCTION 1 1.1 Generalized Semi conductor Fabrication Modules 1 1.2 Increase in Device Density 2 1.3 Scaling and Time Delay 3 1.4 Need for Planarization 9 1.5 Shallow Trench Isolation 10 1.6 Damascene Process 12 1.7 Different Planari zation Techniques 13 1.7.1 Doped Glass Reflow 14 1.7.2 Spin Etch Planarization (SEP) 16 1.7.3 Spin on Deposition (SOD) 17 1.7.4 Reactive Ion Etch and Etch Back (RIE+EB) 19 1.7.5 Chemical Mechanical Planarization (CMP) 20 1.8 General Chemical Mechanical Planarization Applications 21 1.9 Overview 26
ii CHAPTER 2: BACKGROUND OF CMP 27 2.1 Evolution of CMP 27 2.1.1 History of CMP 27 2.1.2 Road Map of CMP Process 28 2.2 CMP: A Multi-stage Process 31 2.2.1 Slurry Mixing 33 2.2.2 Slurry Distribution 34 2.2.3 Working of the CMP Process 35 2.2.4 CMP Polisher Consideration 36 2.2.5 First Generation CMP Polisher 37 2.2.6 Second Generation CMP Polisher 38 18.104.22.168 Multi-wafer Per Platen Polisher 38 22.214.171.124 Sequential Rotational Systems 39 2.2.7 Third Generation Polishers 40 126.96.36.199 Sequential Linear Polishers 40 188.8.131.52 Orbital Polishers 41 184.108.40.206 Rotary Inverted 42 220.127.116.11 Pad Feed Polishers 43 2.3 Physics of CMP Process 44 2.4 Parameters Governing CMP Process 46 2.5 Research Objectives 48
iii CHAPTER 3: METROLOGY OF CMP 49 3.1 Need for Effective CMP Metrology 49 3.2 CMP Tester 50 3.3 Coefficient of Friction (COF) a nd Material Removal Rate (MRR) 52 3.4 Importance of Slurry Flow 55 3.5 CMP End Point Detection (EPD) 58 3.6 In-situ Process Monitoring 63 3.7 Delamination during CMP 72 3.8 Other CMP Defects 76 CHAPTER 4: PROPERTIES OF INTERCONNECT MATERIALS 77 4.1 Need for Evaluation of Material Properties 77 4.2 Effect of Annealing on Copper 78 4.3 Significance of Interfacial Reli ability in Damascene Structure 81 4.4 Effect of Anneali ng on Cu-TaN Interface 82 4.5 Techniques of Interfacial Evaluation of Thin Films 83 4.5.1 Four Point Bend Technique 84 4.5.2 Nanoscratch Testing 86 18.104.22.168 Failure Domain Analysis of Nanoscratch Tests 89 22.214.171.124.1 SEM/Auger Spectroscopy Analysis 89 126.96.36.199.2 FIB Cross section and SEM Analysis 92 4.6 Quantitative Evaluation of Thin Film Adhesion Energy 93
iv 4.7 Qualitative Evaluation of Thin Film Adhesion Energy 95 4.8 Low Dielectric Constant Ma terials (low k materials) 96 4.8.1 CVD Based Low k Materials 98 4.8.2 Spin On Low k Materials 99 4.8.3 Potentially Feasible Low k Materials 100 4.9 Mechanical Characteriza tion of Low k Materials 103 4.10 CMP of Low k Materials 107 4.10.1 Doped and Undoped Ceramic Material Removal 108 4.10.2 Polymeric Material Removal 109 4.10.3 CMP Process Conditions 110 4.10.4 Tribological Propert ies of Low k Materials 112 4.10.5 Variation of Material Rem oval Rate for Low k Materials 116 4.10.6 Surface Characterization of Low k Materials 118 4.10.7 Findings of Low k Materials Evaluation 120 CHAPTER 5: INVESTIGAT ION OF NON UNIFORMITIES IN POLISHING PAD 123 5.1 Significance of Polishing Pad 123 5.2 Method of Mapping and Isolation of Pad Non-Uniformities 124 5.3 Ultrasound Testing System 126 5.4 Evaluation of Mechanical and Trib ological Properties of Polishing Pad 128 5.5 SEM Evaluation of Polishing Pad 130 5.6 Isolation of Polishing Pad Coupons 131 5.7 Isolation of 6 Inch Polishing Pad Coupons 133
v 5.8 Pad Dynamic Mechanical Analysis 134 5.9 Frictional Characteristics of Polishing Pad Regions 137 5.10 Summary of Investigations of Polishing Pad Non uniformities 143 CHAPTER 6: INVESTIGATION AND OPTIMIZATION OF APPLICATION SPECIFIC CMP PADS 144 6.1 Application Specific Pad (ASP) for CMP 144 6.2 Scheme for Optimization of ASP Properties 146 6.3 Experimental Techniques for ASP Property Evaluation 147 6.4 Evaluation of ASP Surface Micromechanical Properties 149 6.5 Evaluation of ASP Surface Micromechanical Properties 153 6.6 Evaluation of Static Trib ological Properties of ASP 157 6.7 CMP (Dynamic Tribological Property) Evaluation of ASP 158 6.8 Optimization of ASP Total Thickness 163 6.9 Summary of ASP Optimizat ion and Characterization 165 CHAPTER 7: INDENTIFICATION AND MODIFICATION OF CMP SLURRY 167 7.1 Background of CMP Slurry 167 7.2 Effect of CMP Slurry in Global Planarization 167 7.3 Effect of CMP Slurry on Removal Rate 170 7.3.1 Effect of Slurry Chemistry on Removal Rate 171 7.3.2 Effect of Slurry Particle Size Hardness and Concentration 172 7.4 Effect of Temperature on Slurry Performance 174
vi 7.5 Novel Nanoparticle Based Slurry 177 7.5.1 Composition of Novel Nanoparticle Based Cu Slurry 178 7.5.2 Mechanism of Polishing of Novel Cu Slurry 179 7.5.3 Polishing Performance of N ovel Nanoparticle Cu Slurry 182 7.6 Summary of Investigation a nd Modification of CMP Slurry 185 CHAPTER 8: SUMMARY 187 8.1 Summary of Research 187 8.2 Major Findings and Contributions 189 8.3 Future Trends in CMP and Po tential Areas for Investigation 191 REFERENCES 193 ABOUT THE AUTHOR End Page
vii LIST OF TABLES Table 1.1 Interconnection Delay (RC) in Silicon VLSI Chip 6 Table 1.2 Advantages of Chemical Mechanical Planarization 23 Table 1.3 Disadvantages of Chemical Mechanical Planarization 24 Table 1.4 Applications of Chemical Mechanical Planarization 25 Table 2.1 Interconnect In ternational Technology Road map for Semiconductors 30 Table 3.1 Effect of Slurry Flow rate on COF and AE 56 Table 3.2 STI-CMP Experimental Details 66 Table 4.1 Details of the Samples Undergoing Two Stage Annealing 79 Table 4.2 Details of the Samples S ubjected to Interfacial Studies 83 Table 4.3 Relative Concentr ation of the Elements (a t.%) on Scratch Surface 91 Table 4.4 Details of the Differe nt Low k Materials Evaluated 105 Table 4.5 Results of Nanoindent ation of the ILD materials 106 Table 4.6 ILD CMP Expe rimental Details 111 Table 4.7 Surface Roughness (Rrms) of the ILD Samples Before and After CMP 120 Table 5.1 Details of the CMP Experi ments Performed on the Polishing Pad evaluated by UST 129 Table 6.1 Experimental Details of CM P Process Used for ASP Evaluation 149 Table6.2 Multivariate Correlations between Various Independent Results and PECVD Coating Time 162
viii LIST OF FIGURES Figure 1.1 Scanning Electron Micrographs of Cross-section of the Structures Fabricated by BEOL Technology: (a) BEOL structure of 0.5 m CMOS Logic Device and (b) Stacked Contacts and Vias 2 Figure 1.2 Trends Over the Years in Logic and Memory Devices 3 Figure 1.3 Chronology of Key Interc onnect Technology Introduction Through the Years. LM Denotes Levels of Metallization 4 Figure 1.4 Variation of RC Time Dela y with Minimum Feature Size 7 Figure 1.5 Predicted Future Trends in IC Interconnect Technology, (Courtesy: Jeffery Lee, Intel Corporation) 8 Figure 1.6 Chart Showing Decrease in Inte rmediate Interconnect Wiring Pitch for Future Generation IC, (Courtesy: Jeffe ry Lee, Intel Corporation) 8 Figure 1.7 Schematic of a) Non-planarized and b) Planarized MLM structure 10 Figure 1.8 Schematic of a Processes Sequence of Direct STI CMP without Reverse Moat 11 Figure 1.9 Comparison between Subtrac tive Etch (Conventional Approach) and the Damascene Approach for Metallization 13 Figure 1.10 Schematic Showing Degrees of Surface Global and Local Planarity 14 Figure 1.11 Schematic Showing BPSG Void Formation after Reflow 15
ix Figure 1.12 Schematic of SEP Chamber Show ing a Cut-away View of the Process Pot, Four Chambers and Chuck. The Chemical Dispense Arm, Drain Lines and Exhaust Ports also are Indicated 17 Figure 1.13 Schematic showing Spin on De position with Partial planarization 18 Figure 1.14 Schematic Showing Smootheni ng and Partial Planarization using Reactive Ion Etching with Etch Back 20 Figure 1.15 Scanning Electron Micrograph Showing Cross-section of Structure Planarized by SOD RIE EB 21 Fig. 2.1 Flow Chart of the Isolated Industrial CMP Process 31 Fig. 2.2 Schematic of the Slurry Mi xing at a Centralized Location 33 Fig. 2.3 Schematic of CMP Slurry Distribution System 34 Fig. 2.4 Schematic of Wafer Planarization by CMP Process 35 Fig. 2.5 Photograph of Speefam (Novellus) Multi Wafer per Platen Polisher 38 Fig. 2.6 Applied Materials Inc., Sequent ial Rotational CMP Polisher (Courtesy: Ashok Das, Applied Materials, Inc., Santa Clara, CA) 39 Fig. 2.7 Photograph of Sequential Li near Polishing System 41 Fig. 2.8 Schematic of Rotary CMP Polisher 42 Fig. 2.9 a) Set up of Multi Wafer Rotary Inverted CMP Polisher and b) Polishing Action and End Point detection 43 Fig. 2.10 Schematic of a Web-type Polisher 44 Fig. 2.11 Schematic of the Force Field on the Wafer and the Pad during CMP 45 Fig. 2.12 Non-uniformity in Removal Rate with in a Wafer 46
x Fig. 2.13 Parameters Governing the CMP Dynamics 47 Fig. 2.14 Schematic Showing Objectives of the Current Research 48 Fig. 3.1 Photograph of the CETR CP-4 CMP Tester 51 Fig. 3.2 a) Variation of COF with RPM (p laten velocity) for Different Values of Down Pressure (PSI), b) Variation of COF with PSI (Down Pressure) with Different Values of Platen Velocity (RPM) 53 Fig. 3.3 a) Variation of Average Remova l Rate with RPM at Different Down Pressure (PSI) for the Evaluated Se t of Samples, b) Variation of Average Removal Rate with PSI (Down Pressure) at Different Platen Velocity (RPM) for the Evaluated Set of Samples 54 Fig. 3.4 Average Material Removal Rates Plotted with RPM PSI. Linear Relation Indicates that Polishing Follows Preston's Equation 55 Fig. 3.5 Schematic of the Positions of Slurry Feeding on the Pad during Polishing for Feeding Position Optimization; Distance 0Â–1.4 = 15 mm, 0Â–2.5 = 30 mm, 0Â–3.6 = 45 mm, 0Â–7.9 = 45 mm and 0Â–8 = 25 mm 56 Fig. 3.6 Average Removal Rate with th e Slurry Feeding Position on the Lower Platen Position 57 Fig. 3.7 a) Variation of COF at 2 PSI Do wn Force and Variable RPM in Slurry Cu1; b) Variation of COF at 2 PSI Down Force and Variable RPM in Slurry Cu2; c) Variation of COF at 2 PSI Down Force and Variable RPM in Slurry Ta; and d) Vari ation of COF at 2 PSI Down Force and Variable RPM in Slurry CuÂ–Ta 59
xi Fig. 3.8 a) Variation of AE and COF for Different Materials Polished in a Commercial Cu Slurry, b) Variation of COF fo r 3 Materials Polished in Three Different Slurries 60 Fig.3.9 Cu and TaN Polishing Results (Head 40 RPM, Table 40 RPM, and Pressure 7 PSI) 61 Fig. 3.10 Schematic Illustration of Dishing and Erosion 63 Fig. 3.11 SEM of FIB Cross Section of STI Structure 64 Fig. 3.12 3D AFM Image of the Evaluate d STI Structure with Step Height Measurement 65 Fig. 3.13 In-situ Variation of COF during CMP Split in to Different Stages 68 Fig. 3.14 Ex-situ 3D AFM Image of the Evaluated STI Structure with Reduced Step Height before Termination of Stage A 68 Fig. 3.15 Ex-situ 3D AFM image of the Evaluated STI Structure after Process End Point just after Beginning of Stage B 69 Fig. 3.16 In-situ data for Repeatability of 2 Trial Runs Shown to Demonstrate Repeatability 71 Fig. 3.17 a) Picture of Sample Coupons ACL, BCL, and TC after Polishing, b) Variation of AE signal of Three Different Samples 73 Fig. 3.18 SEM Micrographs Showing of Delamination of Different ACL, BCL Samples and No Delamination for TC Sample 74 Fig. 3.19 a) Variation of AE Signal during Polishing in the Time Interval of 230 Â– 250 s for Three Different Samples. Peaks are Seen for Sample ACL006, b) Sample ACL006 and (c) Sample A006r 75
xii Fig. 4.1 Challenges Surrounding Implem entation of Novel Low k Materials Cu in a Damascene Structure (Courtesy: Jeffery Lee, Intel Corporation) 81 Fig. 4.2 Schematic of Four Point Bend Test (University of California, Santa Barabara) Set Up 84 Fig. 4.3 Variation of Load (lbf) Vs Displacement (Microns) during a Four Point Bend Test 85 Fig. 4.4 Schematic of the Nanoscratch Testing Setup 88 Fig. 4.5 Raw Data Showing the Variat ion of a) Acoustic Emission (AE), b) Frictional Force (Ff) with Time during a Scratch Test. (Sample 1 at Linearly Increasing Downward Force of 2 gm/sec Linear Velocity of 0.2 mm/sec) 88 Fig. 4.6 Surface SEM of the Nanoscrat ch at Region a) at the Initial Stages of Microscratch, and b) Upon Completion of Delamination 90 Fig. 4.7 FIB Cross-section and SEM Micr ograph of the Thin Film Obtained a Region of Onset of Delamination 92 Fig. 4.8 Variation of Interfacial Adhesion Energy for the Different Samples Evaluated by Four Point Bend Technique 93 Fig. 4.9 Elemental Dispersion Spectrosc opy Analysis of the Failed Interface after Four Point Bend Analysis 95 Fig. 4.10 Variation of Critical Load for Different Evaluated Samples obtained from the Nanoscratch Test 96 Fig. 4.11 Simplified Classification of Low k Dielectrics 102
xiii Fig. 4.12 Typical Curve Showing the Loading and Unloading as a Function of Indenter Penetration Depth 105 Fig. 4.13 Variation of the Normalized Hardness with Dielectric Constant of the Candidate ILD Materials 107 Fig. 4.14 Material Removal Schematic during Ceramic ILD CMP 108 Fig. 4.15 Schematic showing the Action of ILD Slurry on the Polymer Surface during CMP 110 Fig. 4.16 Variation of COF with Down Pre ssure (PSI) and Linear Velocity (m/s) for a) SiO2U, b) SiOF, c)Polyimide, and d) BCB 113 Fig. 4.17 Variation of MRR with Pressure (P) and Platen RPM (Linear Velocity V) for a) SiO2 (U), b) SiOF, c) Polyimide, d) BCB 117 Fig. 4.18 Surface Morphology of the Candidate ILD Samples Before and After CMP for: a) SiO2 (U), b) SiOF, c) Polyimide and d) BCB 121 Fig. 5.1 Schematic of the Construction of the Ultrasound Testing Equipment 125 Fig. 5.2 Cross-section Scanning Elect ron Micrograph of the Evaluated Commercial Polyurethane Polis hing pad with a Sub-pad 130 Fig. 5.3 Ultrasonic Transmission Maps of the Quarter IC1000/SubaIV pad a) Before and b) After punching of th e 1Â” coupons. Three coupons were punched at the High-Intensity (whi te) Area and Four Coupons at LowIntensity (Black) Area. The Effect of Pad Acoustic Homogenization after Punching is Illustrated in Respective Histograms (b, d) 132 Fig. 5.4 a) Pad Mapped Before Punching 6-inch Coupons, b) Area of Pad Remapped after Replacing the Punc hed HT Coupon, c) Area of Pad
xiv Remapped after Replacing the P unched LT Coupon (All the Values have been Normalized over the Entire Area). 135 Fig. 5.5 a) Variation of Storage Modulus Vs Temperature, b) Variation of loss modulus Vs Temperature c) Variation of tan Vs Temperature of Samples Tested from Low and High Intensity region of the Pad 136 Fig. 5.6 Ultrasound Transmission Scan of Pad with Sub-pad 138 Fig. 5.7 Ultrasound Transmission Scan of Pad without Sub-pad 138 Fig. 5.8 Variation of COF for a) Pad 1 and b) Pad 2 during the CMP Process for the Various Values of PSI and RPM 142 Fig. 6.1 Cross-sectional SEM microgra ph of a) PsiloQuestÂ’s ASP and b) Commercial Polyurethane Polis hing pad with a sub pad 150 Fig.6.2 Loading and Unloading Behavior Exhibited by the PsiloQuestÂ’s ASP coated for 10 minutes (Pad 1). Note the non-linearity at the Specific Sites of the Loading and Unlo ading Curve Represent Plastic Deformation under Indentation. 151 Fig.6.3 The Variation of Depth of First N on-linearity of in the Loading Curve (Xa) as a Function of PECVD-TEOS Coating Time for Different Pads Evaluated 152 Fig.6.4 Variation of Elastic Modulus and Hardness with TEOS Coating Time Evaluated by Nanoindentation for Diffe rent Evaluated Polishing Pads 153 Fig. 6.5 X Ray Photoelectron Spectrosc opy (XPS) Data for Polishing Pads with Different TEOS Surface Coating Time 155
xv Fig.6.6 Variation in Surface Silanol / Si licates in PECVD-TEOS Coating as a Function of TEOS Deposition Time 156 Fig. 6.7 Variation of Static Wear Rate a nd Static Coefficient of Friction for Pads with Different PECVD TEOS Surface Coating Time 157 Fig. 6.8 Variation of the Average Value of COF (taken for First 60 Seconds) for the Different Candidate Pads during the Polishing Runs at 2 PSI Down Force and 150 RPM Platen Rotation 158 Fig.6.9 Variation of the Average Value of MRR for the Different Candidate Pads during the Polishing Runs at 2 PSI Down Force and 150 RPM Platen Rotation 159 Fig.6.10 Variation of COF for Different Total ASP Thickness for an Optimized Surface Coating Time of 40 Minutes 164 Fig. 6.11 Variation of MRR for Different Total ASP Thickness for an Optimized Surface Coating Time of 40 Minutes 165 Fig. 7.1 Prime Criteria for Slurry Design 168 Fig. 7.2 Schematic of Microscale a nd Nanoscale Phenomena during CMP 170 Fig. 7.3 Variation of Rate of Surface Layer Formation in Cu with Different Slurry Chemistry 171 Fig. 7.4 Transient Electrochemical Chr onoamperometry Measurements of W 171 Fig. 7.5 Variation of Removal Rate with Particle Size and Concentration 173 Fig. 7.6 Removal Rate of Silica with Diffe rent Particle Size and Concentration 174 Fig. 7.7 Variation of COF and MRR with Increase in Slurry Temperature 176
xvi Fig. 7.8 Variation of ln of MRR in m/ s with Inverse of Temperature (1/T) Showing an Arrhenius Relationship 176 Fig. 7.9 Schematic Showing Mechan ism of Polishing Action of Novel Nanoparticle CMP slurry 181 Fig. 7.10 Variation of MRR for Different Trials Using Nanoparticle Slurry without the Surface Catalyst 182 Fig. 7.11 Variation of MRR for Different CMP Trials for Nanoparticle Slurry with Surface Catalyst 183 Fig. 7.12 Variation of COF for Differen t CMP Trials Using Novel Nanoparticle Based CMP slurry 183 Fig. 7.13 Variation of COF with Time M easured In-situ for Polishing of EP Cu Samples Under 3 PSI Down Pressure and 100 RPM Platen Rotation Using a) Novel Nanoparticle Cu Sl urry, and b) Commercial Cu Slurry 184
xvii PROCESSING, RELIABILITY AND INTE GRATION ISSUES IN CHEMICAL MECHANICAL PLANARIZATION Parshuram B. Zantye ABSTRACT Global planarization is one of th e major demands of the semiconductor industry. Chemical mechanical polishing (CMP ) is the planarization method of choice use to achieve the required stringent tolerances essential for successful fabrication of next generation Integrated Circuits (IC). The pre dominant reason for CMP defects is the shear and normal stresses during polishing to whic h the material is s ubjected. Understanding the process of CMP and factor that contribute to overall stress addition during polishing requires an approach that encompasses all the four major categories of variables, namely: a) machine parameters, b) material propertie s, c) polishing pad char acteristics, and d) polishing slurry performance. In this research, we studied th e utilized in-situ technique involving acoustic emission (AE) signal monitoring and coefficient of friction (COF) monitoring using a CETRTM Bench Top CMP Tester to eval uate the impact of variation in machine parameters on the CMP process. The mechanical and tribological properties of different candidate materials have been evaluated bring potential challenges in their integration to the fore. The study also involve s destructive and non de structive testing of polishing pads performed for characteri zation and optimization of polishing pad
xviii architecture. Finally, the investigation concludes proposing novel nanoparticle CMP slurry which has a predominant chemical co mponent in its polishing mechanism. It was found that the decrease in the mechanical shear and normal loadi ng by: a) operating the process in the low stress regime, b) using potential materials that are mechanically stronger, c) using polishing pa ds with lesser variation in specific gravity and with a surface that is has its mechanical properties fine tuned to those of the wafer, and d) deploying polishing slurry with a significant chemical component mechanical removal, are some of the approaches that can be em ployed to meet the future challenges of the CMP process and reduce the defect associated with it.
1 CHAPTER ONE INTRODUCTION 1.1 Generalized Semiconductor Fabrication Processes The relentless competitor and customer driven demand for increased circuit density, functionality and versat ility has led to evolutionary and revolutionary advances in the Â“front endÂ” of the chip manufacturing line where the devices are fabricated, and the Â“back endÂ” where these devices are appropriately wired within the integrated circuit (IC) . Chip interconnections, or Â“interc onnects,Â” serve as local and global wiring, connecting circuit elements and distributing power . To incorporate and accommodate the improvements such as decreased feature size, increased device speed and more intricate designs, research in the Â‘back end of the lineÂ’ (BEOL) processes has become equally important as the development of the Â‘front end of lineÂ’ (FEOL) processes to reduce gate oxide thickness and channel length. Fig. 1.1 (a and b) shows the multilevel interconnect structure which is fabricated us ing the BEOL processes. The current viable technologies and future trends in scali ng bipolar and Complimentary Metal Oxide Semiconductor (CMOS) devices fabrication a nd FEOL technologies have been discussed at length by Taur et al. .
2 Fig.1.1 Scanning Electron Micrographs of Crosssection of the Structures Fabricated by BEOL Technology: (a) BEOL Structure of 0.5 m CMOS Logic Device and (b) Stacked Contacts and Vias . 1.2 Increase in Device Density Over the last 20 years, circuit de nsity has increased by a factor of approximately 104 (Fig. 1.2), while cost has constantly decreased [e.g., the historical 27% per year decline in price per bit for dynamic random access memories (DRAMs)] . The trend is expected to continue in the future even as 45 nm processes are set for production in 2007 . While recent path break ing innovations in the field of lithography and patterning  have brought about progr essive device scaling, the development of a planar back-end-of-line approach, which inco rporates the use of chemicalÂ–mechanical polishing to planarize inter-level dielectrics a nd metal stud levels, represents a significant advance in BEOL processes. Innovation in BEOL technology is required in each
3 technology generation (Fig. 1.3), si nce only part of the density increase could be achieved with improvements in lithography (Fig. 1.3). Th e evolution and progressive improvement in the BEOL technology and processes along with the future trends ha ve been elaborately discussed by Ryan et al . Fig. 1.2 Trends Over the Years in Logic and Memory Devices . 1.3 Scaling and Time Delay At the outset, the CMOS device structure had multiple isolated devices connected by single level of interconnect wiring. Scaling down of the device was very effective in achieving the goals of increased device density, functi onal complexity and performance. However, scaling down of the devices became less profitable, and speed and complexity were dependant on the charac teristics of interconne cts that wired the devices . With the single level metalliz ation scheme the total area occupied by the wiring on the chip significantly increased with the increase in the active density of devices on the chip. Keyes  cited an exampl e of a bipolar chip w ith a gate count of
4 1500 gates and a chip area of 0.29 cm2, fabricated using a single level metal with a pitch of 6.5 m. The total wiring area occupied by the metal was 0.26 cm2, which was about 90% of the surface ar ea of the chip! Fig. 1.3 Chronology of Key Interconnect T echnology Introduction Through the Years. LM Denotes Levels of Metallization .
5 The total time taken by the voltage at one end of the metal line to reach to 63% of the total value of the step input applied at the other end is known as the interconnect delay and this is due to resi stance of the interconnect wiring metal ( R ) and the interlayer dielectric capacitance ( C ) . The RC (Resistance X Capacitance) delay can be expressed as shown the following equation: 2 2 2 01 4 2 T P L RC (1.1) where is the resistivity of the wiring material, k is the dielectric constant of the ILD, 0 is the permittivity of vacuum, L is the length of the interconnect line, P is the interconnect wiring pitch, T is the thickness of the line. It can be seen from (Eq. (1)) that there is an increase in the RC delay with the decrease in interconnect wiring pitch. Hence, in order to decrease the RC delay: 1) Cu has replaced Al as interconnect wiring materials due to its lower resistivity, 2) several novel low k materials are being explored, and 3) multilevel metallization scheme of wiring is being implemented. As Cu cannot be effectively etched due its inability to form non toxic volatile by-pr oducts and due to its property of diffusion in neighbor ing materials, present day ML M structures are fabricated using the damascene process. Table 1 calculates the simple RC time constants calculated for a few metals of given Rs (sheet resistance) and 1 mm length on 1 m thick SiO2 . The increasing in the levels of the metallization lines means that packing density need not keep pace w ith the device density and the minimum metal line feature does not have to scale with the same pace as the gate width. The foremost reason behind the implementation of multilevel metallization schemes is the reduction in the length of the metal lines, which in turn reduces the RC delay sizably (Eq. (1)).
6 Table 1.1 Interconnection Delay (RC) in Silicon VLSI chip Metal Bulk Resistivity ( -cm) Poly crystalline film resistivity ( -cm) Film Thickness (Ao) Rs (/square) Delay a (ps /mm) Poly-Si CoSi2 MoSi2 TaSi2 TiSi2 W Al Cu 10 ~35 45 13 5.65 2.65 1.67 ~1000 15 ~100 55 15 8-10 2.7 2.0 5000 2500 2500 2500 2500 2500 2500 2500 20 0.6 4 2.2 0.6 0.32-0.4 0.11 0.08 690 21 138 76 21 11-14 4 3 aDelay = RC = 34.5 Rs (ps/mm) for 1mm length conductor on 1-m thick SiO2
7 In places where metal wiring length ca nnot be reduced, routing can be done at the upper levels without reducing the metal line width, thus reducing the RC delay due to the higher surface area. It must be noted that (Eq. (1)) takes in to account only the line to ground capacitance and does not take in to a ccount the capacitance between adjacent metal lines. The line-to-line capacitance is negligible for wide isolated lines but is significantly large in any sub 3 m interconn ect regime. In sub 0.5 m the line to line capacitance dominates, there by increasing the RC time delay significantly with scaling. As seen from Fig. 1.4, there is a dramatic increase in RC time delay in sub 0.5 m feature size interconnect lines. Starting with two levels of metallization, the levels of metallization have increased up to 8 by 2001  The future trends in the levels of metallization can be seen in Fig. 1.5. Fig. 1.4 Variation of RC Time Dela y with Minimum Feature Size .
8 Fig. 1.5 Predicted Future Trends in IC In terconnect Technology, (C ourtesy: Jeffery Lee, Intel Corporation) [13-14] Fig. 1.6 Chart Showing Decrease in Intermed iate Interconnect Wiring Pitch for Future Generation IC, (Courtesy: Jeffery Le e, Intel Corporation) [13-14]. The design and layout of interconnect lines is done using the numerous analytical and numerical techni ques available. Various techniques have been proposed to investigate the time domain and pulse propaga tion characteristics of parallel coupled lossless and lossy lines used to model the interconnect lines in the high speed USLI
9 circuits [15, 16]. These techniques include method of characteris tics with necessary modifications to incorporate frequency depe ndant losses [16-18] and congruent modeling techniques where an attempt is made to model the interconnect systems in terms of lumped and distributed circuit elements in computer aided design programs such as SPICE and CADENCE [15, 16]. It is widely accepted that the minimum feature size of the devices on the chip also implies the de crease in the intermediate pitch of the interconnect wring that connects thes e active devices (Fig. 1.6) . 1.4 Need for Planarization With the decreasing intermediate wi ring pitch, non-planarized surface topography results in several processing difficulties. The irregular surf ace causes a hindrance in conformal coating of the photor esist and efficient pa ttern transfer with contact lithography. The anomalies in the surf ace cause the variation of the thickness in fine line widths (sub 0.5 m) depending upon photo resist thickness. Effectively planarized surface has enormous amount of be nefits such as: 1) higher photolithography and dry etch yields, 2) elimination of step coverage concerns, 3) minimization of prior level defects, 4) elimination of contact interruption, undesired contacts and electromigration effects, 5) reduction of hi gh contact resistance and inhomogeneous metallization layer thickness, and 6) limitati on in the stacking height of metallization layers. Fig. 1.7 (a and b) shows a compar ison between planarized and non-planarized surface topography.
10 Fig. 1.7 Schematic of a) Non-planarized and b) Planarized MLM structure  1.5 Shallow Trench Isolation Shallow trench isolation (STI) has become a key technology for device isolation in recent times [ 20, 21]. The importance and the need for shallow trench isolation have been discussed by Wolf  The method comprises of making a shallow trench on a silicon wafer, depositing SiO2 thereon, and then planarizing with a chemical mechanical polishing (CMP) process. The meth od can separate elements within a much narrower area, and shows much better performa nce than the conventi onal local oxidation of silicon (LOCOS) method, which cause s bird's beak structures . The details of fabrication of STI stru ctures have been elaborately given discussed Jeong et al . Until now, a complicat ed reverse moat etch process had to be used in the absence of sufficiently selective slurries for SiO to SiN polishing. Using an etch process, the high-density moat regions can be reduced to an acceptable level, and
11 therefore the chip or wafer level polishing uni formity can be greatly enhanced. If direct CMP without the reverse moat etch process wa s applied with conven tional low selectivity slurries, damage might occur to active regions in the case of excessive CMP, whereas, in the case of insufficient CMP, nitride residues might remain in the active regions after the nitride strip process due to oxide residues [20-25]. The schematic representation of the STI structure fabrication reported by Kim and Seo is shown in Fig. 1.8 . Fig. 1.8 Schematic of a Processes Sequence of Direct STI CMP without Reverse Moat 
12 The process of fabrication of STI st ructures is still under considerable research [29, 30]. One of the main areas of interest is development of silica and ceriabased high selectivity slurries (HSS)  with a high polishing se lectivity for silicon oxide and silicon nitride [25, 26]. There is considerable re search currently underway in the STIÂ–CMP aspects such as effective a nd in situ end point detection [20-25], reproducibility , defect analysis [27, 28] pattern density effects , etc. The STI CMP process has also been extensively modeled [31-33]. 1.6 Damascene Process In the conventional metallization techni que as seen in Fig. 1.9, in the conventional metallization technique, the meta l deposited on top of the dielectric is positively patterned with photoresist. The metal is then etched out and dielectric material is deposited on top of the metal using proce sses such spin coating or chemical vapor deposition (CVD) . The dielectric is then planarized and subsequently to make a multilevel metallization structure, more diel ectric is deposited on top of the planar dielectric and the process is repeated. In cas e of the damascene process, the dielectric is negatively patterned, and then etched to form a pattern that is then filled with metal. A seed layer of metal is deposited using physical vapor deposition (PVD). Depending upon the metal, a barrier layer of metal is deposit ed before the seed layer deposition . The metal is then electroplated on top of the seed layer. The excessive metal is polished off and planarized using the CMP process. For the purpose of making multilevel metallization structures, dielectric is then spin coated or CVD deposited and entire procedure is repeated.
13 Fig. 1.9 Comparison between Subtractiv e Etch (Conventional Approach) and the Damascene Approach for Metallization 1.7 Different Planarization techniques The different degrees of global and local surface planarity  can be seen from Fig. 1.10. Techniques such a sp in on deposition (SOD), reflow of boron phosphorous silicate glass (BPSG), spin etch planarization (SEP), reactive ion etching and etch back (RIE EB), SOD + EB have been discussed in this section. These are the prominent of several competing technologies presently being used to achieve local and global planarization.
14 Fig. 1.10 Schematic Showing Degrees of Surface Global and Local Planarity  1.7.1 Doped Glass Reflow Synthesis of low pressure chem ical vapor deposited (LPCVD) boron and phosphorous doped silicon oxide was one of the fi rst planarization tec hniques in the IC industry used to fabricate the first layer of dielectric (pre metal dielectric) due to its excellent planarization and getteri ng properties [38-42]. By doping SiO2 with boron and phosphorous, the film boro-phosphateÂ–silicate gl ass (BPSG) has better smoothing of step corners and it can be made to re flow at high te mperature (850Â–959 C).
15 Kobayashi and co-workers [38-42] ha ve given the details of formation of doped BPSG using n-type lightly doped Si wafe rs. Dielectric glass layers were deposited on the wafers in a (LP-CVD) reactor equipped with Si(OC2H5)4, B(OCH3)3 and PH3 gas sources and O2 and N2 carrier gases. As the reflow ch aracteristics are mainly controlled by viscosity, which in turn is a function of glass chemical bonding [41, 42] and structure , less viscous, non-crystallized glasses are ideally used for reflow and planarization. These glasses are therefore deposited by LP CVD technique, as they are amorphous, more fluid, have low connectivity a nd have a released structure. Even though, LPCVD highly boron-contai ning glasses with low polarizability are favorable for the device planarization in DRAMs and static random access memory (SRAM) cells, these glasses can be used only for the first level of ILD. This is due to the fact that even the low temper ature reflow glasses would me lt the metal once deposited as the standard temperature of reflow far exceeds melting point of aluminum. Moreover, high temperatures are unsuitable for other me tals due to diffusion and electro-migration issues. Also, due to void formation (Fig. 1. 11) during reflow, and very high thermal budget, the process of doped glass reflow is not a very widely implemented process of planarization. Fig. 1.11 Schematic Showing BPSG Vo id Formation after Reflow 
161.7.2 Spin Etch Planarization The process of CMP gained increasing prominence due to controlled chemical etching of some metals like Cu was not a very feasible task. However, spin etch planarization, a process developed by Levert et al. at SEZ America Inc.  is based on the principles of controlled chemical etching of metals. During SEP, the wafer is suspended horizontally on a nitrogen cush ion above a rotating chuck (Fig. 1.12). The substrate is held in place laterally with lo cking pins on the wafer edge. As the chuck and wafer are spun, wet etch chemistries are disp ensed onto the wafer. A planar final surface is achieved by using an appropr iate etching solution and the spinning of the wafer while removing the excess Cu. Deionized water and ni trogen are then applied onto the wafer to achieve rapid cleaning and dry-in /dry-out-processing. Results show that the etch rates can be as high as 14,000 /min. 200 mm electropl ated wafers can be planarized with appropriate chemistries and processing parameters . As there is no contact of any extern al body with the wafer surface, there is no possibility of typical CMP def ects like micro scratches, delamination, peel off, etc. There is reduced instance of dielectric dishing and erosion of metal lines and with in wafer nonuniformity is kept as low as 9.2%. Even though this process has some distinct advantages over CMP, this being a totally new process, is yet to be applied in the industry. The process is expected to increase the cost of ownership (CoO), has not been demonstrated on any other materials such as ceramics and insulators. The pattern dependence and etch anisotropy are yet to be further investigated CMP may be still needed after SEP process to remove pattern dependent bumps on the su rface of the wafer. Efficient end point detection mechanisms, in addition to the opti cal end point detecti on mentioned by Levert
17 et al. have to be developed for the proce ss. Therefore, for implementation of the SEP further studies, characterization and optimization is necessary. Fig. 1.12 Schematic of SEP Chamber Showing a Cut-away View of the Process Pot, Four Chambers and Chuck. The Chemical Dispense Arm, Drain Lines and Exhaust Ports also are Indicated . 1.7.3 Spin on Deposition (SOD) Porous lowk dielectrics, different glasses such as OSG, TEOS used as dielectric materials and polymeric ILD are t ypically deposited using spin on technique. The precursor solution for the material to be deposited is prepared mostly at room temperature by mixing the base catalyst and suitable organic additives. The wafer surface is pretreated to promote effective sol sp reading, followed by dripping the sol on the spinning wafer. Small amount of sol is dripped on wafers that are then rinsed, spun dried, baked and later cured.
18 SOD demonstrates excellent gap fi lling capabilities but shows very poor global planarization. Spin on deposited hydrogen silsesquioxane (HSQ) (dielectric constant k = 3.0), has been reported to be successfu lly integrated into devices with five levels of Al interconnect [44, 45] and sili con di oxide formed on surface of silicon using silicic acid solution by spin technolo gy , has shown relatively good local planarization  and is know n to have a positive impact on the global planarization of the ILD achieved by CMP. Fig. 1.13 Schematic showing Spin on Deposition with Partial planarization  Numerous defects are known to arise in the spin on deposited materials. There is non-homogeneity in the valu e of the dielectric constant of these materials with the exposure to plasma in subsequent processi ng . The spin-on materials also have a tendency to absorb moisture and then release it in the air during the thermal processes. This induces undue stresses in the SOD film s there by causing defect s such as cracking, shrinking, peel off, degradation, contam ination of interconnects and poor thermal
19 stability . For this purpose, techniques such as laser curing need to be implemented to prevent stresses from building into the dielectri c film . Thus in spite of the fact that SOD materials show excellent local planar ization, blanket SOD materials are not implemented in the industry. SOD materials are implemented only as layers sandwiched between two oxide layers. The schematic of SOD and partial/local planarization can be seen in Fig. 1.13. 1.7.4 Reactive Ion Etch and Etch Back (RIE + EB) A competing technology for SOD oxid e planarization and reflow is the reactive ion etch and etch back (RIE + EB ). The technique of reactive ion etching, conventionally used to pattern the thin film on a substrate in this case is used for planarization. The pattern is spin coated with photoresist. The resists fills the trenches and vias of the pattern leaving the hills and mounts on the pattern e xposed to the reactive species in the plasma. Typically RIE + EB is used to etch SiO2 and other dielectrics. Although wet etching is well developed for etching SiO2, it has inherent limitations due to undercutting of the mask materials, esp ecially for sub micron pattern sizes. A dry etching technique, like RIE + EB, on the other hand, can generate anisotropic etch profiles and for this reason has come into fa vor. The mechanism of material removal is more due to chemical reaction than du e to physical sputte ring, although the two mechanisms are synergistic; i.e. the bombardment catalyzes the surface chemical reactions. This leads to anis otropic etching due to the directional nature of the bombardment catalyzed surface chemical r eactions [51-54], as well as by physical sputtering. In general, the rate controlling mechanism of etching by the RIE process may
20 be due to physical effects (as in sputtering wi th inert ions), or chemical phenomena in the sense that the ion bombardment enhances surf ace chemical reactions with the reactants yielding highly volatile reaction products. Fig. 1.14 Schematic Showing Smoothening and Partial Planarization using Reactive Ion Etching with Etch Back  Due to the inadequacies of different planarization techniques, the combination of the two techniques has been used in orde r to compliment each other, with some degree of success. SOD with Etch back has proved partic ularly useful in this respect. As the spin on deposited glass has the ability to fill vo ids and gaps permanently, the technique was developed along with the development of the Reactive Ion Etch and Etch Back (RIE + EB) technique. With the emergence of the new spin on polymeric lowk dielectrics  and other novel spin on materials, technique s like SOD and EB have been pursued with some degree of success in achievement of local planarization on the su rface of the wafer.
21 The SOD materials are used to fill the trenches and vias and then RIE process is used to etch back or sacrifice the materials on the higher regions. Subsequently the same material might be deposited using spin on or CVD pro cess to get considerable degree of local planarization. This kind of process is prev alent in gap filling of memory devices (Fig. 1.14). Even though the usage of both SOD a nd RIE EB processes together tend to overcome the drawbacks of each of the proces ses, the extensive optimization is required for the two processes to work in tandem there by giving good surface planarity. Fig. 1.15 shows the cross-section of a device structur e planarized using SOD RIE EB process. Fig. 1.15 Scanning Electron Micrograph Showi ng Cross-section of Structure Planarized by SOD RIE EB 
221.7.5 Chemical Mechanical Planarization Presently, CMP is the only technique that can offer excellent local and global planarity on the surface of the wafer. CMP has known to yield local planarization of features as far as 30 m apart as well as excellent global planarization. The plasma enhanced chemical vapor deposited oxides ha ve limited capability of gap filling and are restricted in their gap fill ing ability below patterns having 0.3-m feature size. Highdensity plasma deposited oxides have acceptabl e gap filling capabilities; however, they produce variation in surface topog raphy or local as well as global level. Even though spin on deposited (SOD) doped and undoped oxides and polymeric materials have acceptable ability for gap filling, CMP is the only tech nique that produces exce llent local and global planarity of these materials. The advantag es and disadvantages of the CMP technique have been listed in Table 1.2 and 1.3 respec tively. The details and various aspects of CMP are discussed subsequen tly in different sections of this dissertation. 1.8 General Applications of CMP The process of CMP was initiall y developed and implemented for planarization of SiO2 which is used as interlayer di electric in multilevel metallization scheme. The initial developmental focus of CMP was oxide planarization . Tungsten is used as an interconnect plug to the sour ce, drain, and gates of transistors in Si microprocessor chips. Initially Ti and Ti N barrier layers are deposited, followed by chemical vapor deposition of W to fill the contact vias.
23 Table 1.2 Advantages of Chemi cal Mechanical Planarization Benefits Remarks Planarization Achieves Global Planarization Planarize different materials Wide range of wafer surfaces can be planarized. Planarize multi-material surfaces Useful for planarizing multiple materials during the same polish step. Reduce severe topography Reduces severe topography to allow fabrication with tighter design rules an add itional interconnection levels Alternative method of metal patterning Provides an alternate means of patterning metal, eliminating the need to plasma etch, difficult to etch metals and alloys. Improved metal step coverage Improves metal step covera ge due to reduction in topography. Increased IC reliability Contributes to increasing IC reliability, speed, yield (lower defect dens ity) of sub 0.5 m and circuits. Reduce defects CMP is a subtractive process and can remove surface defects. No hazardous gases Does a not use hazardous gas common in dry etch process.
24 Table 1.3 Disadvantages of Chem ical Mechanical Planarization Disadvantages Remarks New technology CMP is a new technol ogy for wafer planarization. There is relatively poor control over the process variables with narrow process latitude. New defects New types of def ects from CMP can affect die yield. These defects become more critical for sub0.25 m feature sizes. Need for additional process development CMP requires additional process development for process control and metrology. An example is the endpoint of CMP is difficu lt to control for desired thickness. Cost of ownership is high CMP is e xpensive to operate because of costly equipment and consumables. CMP processes materials require high maintenance and frequent replacements of chemicals and parts. Going ahead from achieving local and global planarization of SiO2, removal of excessive tungsten from the horizontal surfaces on the wafer pattern proved to be an asset for subsequent Al metallization [56-58]. Hence CMP was developed with a twofold approach of: 1) planarizing oxide and 2) removing the via fill metal from the horizontal surfaces. The major applications of CMP are given in Table 1.4.
25 Table 1.4 Applications of Chem ical Mechanical Polishing  Materials Application Al Interconnections Cu Interconnections Ta Diffusion Barrier/Adhesion Ti Diffusion Barrier/Adhesion TiN, TiNxCy Diffusion Barrier/Adhesion W Interconnection e-Emitter Metal Cu-Alloys Interconnections AlAlloys Interconnections Polysilicon Gate/Interconnect SiO2 ILD BPSG ILD PSG ILD Polymers ILD Si3N4 or SiOxNy Passivation Layer, Hard Dielectric Aerogels ILD ITO Flat Panel High K Dielectrics Packaging High Tc Superconductors Interconnections /Packaging Optoelectronic Materials Optoelectronics Plastics, Ceramics Packaging Other Silicon on Insulator (SOI) A dvanced Device /Circuits Along with its successful implementati on for the achievement of the abovementioned objectives, CMP has now extended to 1) polishing of different metals like Al, Cu, Pt, Au, Ti, Ta, etc., 2) polishing of different insulators like SiO2, Si3N4, various lowk dielectrics, doped and undoped oxides of silic on, 3) polysilicon, 4) ceramics like SiC,
26 TiN, TaN, etc., 5) multichip modules, 6) packag ing, 7) optoelectronic components, 8) flat panel displays, 8) microelectromechanical systems (MEMS), and 9) magnetic recording heads and CD read write drives . 1.9 Overview This dissertation begins with the need for device scaling and implementation of novel materials in the present day se miconductor industry. The importance of planarization and the various available planarization t echniques with emphasis on CMP have been discussed in Chapter 1. Chapter 2 gives an overview of the CMP process in general and gives the backgr ound of the process and equipm ent used to carry out the process. The different types of equipment us ed for CMP process and innovations there in have been discussed in this section. The CMP findings of the st udies involving in-situ metrology of CMP and detection of process end point, slurry sele ctivity, as well CMP defects using the CETRTM CMP tester have been discu ssed in Chapter 3. Chapter 4 elaborates the various issues surrounding integration of Cu and novel low dielectric constant materials in the ne xt generation damascene structures. The investigation of specific gravity non uniformities and the use of application specific pads with novel pad architecture have been discussed in Chapte r 5 and 6 respectively. The investigation of CMP slurry and synthesis of novel nanoparticle based Cu CMP slurry has been dealt with Chapter 7. Finally, Chapter 8 summari zes the research, highlights significant contributions and gives an idea of the future directions in which CMP is heading.
27 CHAPTER TWO BACKGROUND OF CMP 3.1 Evolution of chemical mechanical polishing By definition, chemical mechanical polishing is a process whereby a chemical reaction increases the mechanical removal rate of a material. CMP is mostly used for material removal by polishing the Â“hillsÂ” on th e wafer and Â“flatteningÂ” the thin film. The chemical reaction between the slurry and wafe r is tailored to enhance material removal and bring about quicker planar ization of the thin film. 2.1.1 History of CMP The modern day application of CMP process in the semiconductor industry was for polishing the surface raw silicon wafe rs to achieve a global flatness over raw silicon wafers. After sawing, the single crysta l silicon rod and removing the mechanically damaged surfaces, the wafer needs to be flattened globally and a uniform scratch free surface needs to be made available for fabric ation of semiconductor devices. The idea of using colloidal silica, then made by Monsanto, instead of standard abrasives, was developed by Bob Walsh in 1961  and thus the first wafers polished using CMP were commercially available in the early 1960s [59, 60]. Before its implementation in polishing raw single crystal silicon wafers, th e process of CMP was traditionally used in
28 glass polishing. One of the most wide spread application of CMP outside the semiconductor industry is optical lens polishi ng. In fact, the first machinery used by Monsanto was very similar to the commercial machine used in the optical industry. The first semiconductor CMP machine was an i nnovation of the optical lens polishing machine. The proper polishing abrasives in pres ence of the slurry chemicals were used to achieve a superior degree of precision and flatness to meet the demands of the semiconductor industry. By supplementing m echanical polishing with high hardness abrasives such as silica in an alkaline medi um, there are significant gains in material removal and reduction the process time. A further improvement to the CMP pr ocess was made at IBM in the late seventies and early 80 s. The new process wa s faster than the previous silica-based polishing method and resulted in ultra flat, ultra smooth surface to meet the stringent requirements of the IC industry . The slu rry was later tailored to reduce defects and surface non-planarity introduced by the etching and deposition processes. The IBM process was then applied for trench isolation by the late 1980s in Japan for various logic and DRAM devi ces. There was wide spread industrial implementation of the different variants of the CMP by companies such as NEC, National Semiconductor, Hitachi, etc. This led to th e introduction of the fi rst commercial polisher designed specifically for CMP by Cybeg in Japan in 1988. Later, International SEMATECH identified CMP as a technology critical for the fu ture of IC manufacturing and launched a project to develop competitive, advanced CMP tools in the US .
292.1.2 Road Map of CMP Process The semiconductor industry has effec tively adapted its CMP technology for the 300 mm wafer . Beyond the adopti on of copper interconnects, several technologies are necessary to continue th e shrinkage of device dimension and the increase of packing density in ULSI manufacturing. The use of ultra lowk materials as interlayer dielectrics has been at the forefront for decreasing the Â“ C Â” of the Â“ RC delayÂ”. However, polishing of ultra lowk dielectrics which are soft mechanically and weak is a daunting task in itself. The single and dual da mascene structures comprising of ultra lowk porous or polymeric material s are more prone to buckling and crushing failures. It can be seen from Table 2.1, that these mate rials have significantl y lower hardness and Young's modulus as compared to silicon di oxide which has a dielectric consta nt of about 3.2Â–4.0 available in the market today. Acco rding to the ITRS ro admap  materials with dielectric constant 2.2 will be integrat ed in the IC by year 2007 (Table 2.1). The difference between the polishing rates of copper and the lowk materials available will significantly affect post-CMP surface planarity New processes must be developed to address the problems associated with this non-uniform polishing phenomenon as well as the complexity of the materials structures. Also, there needs to be a marked improvement in slurry selectivity for accurate end point detection when the constituent layers of the damascene structure namely, metal, hard mask, cap layer, barrier layer and dielectric are polished . Furthermore, it is necessary to explore the niche of the CMP process in shallow trench isolation and other applications such as backside polishing and the fabrication of micro-electro-mechanical sy stems (MEMS). The CMP process there must be integrated horizontally and vertically to achieve high thorough put and performance.
30 Table 2.1 Interconnect Inte rnational Technology Roadmap for Semiconductors  In MEMS applications horizontal in tegration ensures reliability and good performance of a specific CMP process run. The development of new polishing pads and pad architecture, novel slurries new metrology techniques, etc. comes with in the scope of horizontal CMP integration. Vertical inte gration ensures success of every successive CMP operation. This includes the integrat ion of the upstream processes such as Cu/barrier deposition and etch ing and downstream processes such as ILD deposition and
31 lithography. This opens a wide scope for research for further optimization and development of Cu CMP process. 3.2 CMP: A Multi-stage Process Fig. 2.1 Flow Chart of the Is olated Industrial CMP Process The recent developments in the semiconductor industry described previously imply that CMP is fast becoming the estab lished technology for planarizing metal and interlayer dielectrics of multilevel sub-0.5 m devices. Along with the rapid growth of CMP and its application for polis hing various materials, have come a variety of slurries, different pads, complex process recipes, mo re complex slurry mixing and distribution systems and an increase in the volume of wast ewater. Polishing of di fferent materials and customized needs of the various semiconduc tor industries have given birth of more complex CMP equipment with different proc ess dynamics such as linear, orbital and fixed head machines. Reliable filtration and wa ste distribution are al so required to avoid
32 hazardous environmental implic ations, as the increase in the number of CMP process steps have given rise to a large amount of di sposable slurry waste. The schematic of the industrial CMP process (isolated from the othe r process in the fabr ication line) is as shown in Fig. 2.1. As seen from Fig. 2.1, the slurry is mixed in a tow at a central location from where it is distributed to the various CMP machin es in the fabrication. The slurry that not used for feeding any machines is then return ed back to the tow for recirculation. The slurry that is used for CMP process is later disposed off. The wafer which is dry from the previous process is loaded in the CMP pr ocess equipment where it undergoes polishing and then cleaning. It is then dried in the CMP cleaning sta tion (which may be integrated with the polisher). Most machines follow this dr y in dry out methodology. 2.2.1 Slurry Mixing CMP slurry feeder equipment is compos ed of a stock solution unit, a mixing and circulation unit, and the CMP equipment. The thick slurry s upplied from the stock solution unit is diluted to a fixed density in the mixing and circul ation unit using ultra pure water. The mixed and diluted solution is then supplied to the CMP equipment. All of the equipments are manufactured in a clean room with high degree of cleanness, under strict quality checking. Consis tent construction method is used from the manufacturing of the slurry feeder unit to local piping, wiring, cleaning and th e trial run adjustment of the system. The system thus needs to be of hi gh quality and stability. A tolerance of about 1% is maintained in meeting the recommended slurry parameters. The schematic of the slurry mixing is shown in Fig. 2.2. In this section, the various physical aspects of the
33 CMP process are discussed in detail in order to give a better unders tanding of the general working of the industrial CMP process and the va rious parameters associ ated with it . Fig. 2.2 Schematic of CMP Slurry Distribution System 2.2.2 Slurry Distribution Fig. 2.3 gives a simplified physi cal representation of a semiconductor fabrication line in which, the slurry distri bution to the various CMP tools is shown. As seen from the diagram, the slurry is mixe d and blended at a centralized location from which it is distributed to the various machines through the distribution lines. The distribution loop shown in the schematic ensure s that the Â“goodÂ” slurry, which is unused for the process, is delivered back the mixi ng chamber or the tow, in order to prevent slurry wastage. Although CMP slurries are com posed of very fine particles up to 200 nm (0.2 m), Â“largeÂ” particles of 1Â–3 m and greater are often present in slurries at the point of dispense. Such particles can be formed as a result of agglomeration or the presence of foreign material. Metal CMP slurries, in particular, are prone to formation of aggregates. The agglomerated slurry particles often cau se numerous defects in the wafer during CMP. Microscratching is the most promin ent defect which occurs mainly due the agglomerated particles present in the slurry delivered to the machin e as well as such particles embedded in the pads . This ma kes the continuous mixing of the slurry in
34 the tow absolutely imperative. In order to prevent the particle agglomeration  during the distribution stage, dispense filters are installed on each of the machine which filter out the agglomerated particles before slurry delivery. Care must be taken that the filtration of the agglomerated particles does not change the particles dist ribution and concentration of the slurry. Most often, a seri es of filters are used in or der to minimize the drop in the slurry pressure and flow. The slurry after being used for the actual CMP process in the tool is then disposed off using appropria te methods and environmental damage is restricted . Fig. 2.3 Schematic of CMP Slu rry Distribution System  2.2.3 Working of the CMP Process Current semiconductor fabrication technology for logic and memory devices requires CMP to achieve the required multilevel interconnections densities. Indeed, each silicon wafer can be exposed to 15 or more CMP steps before final device assembly. A schematic diagram of the CMP process is shown in Fig. 2.4. During CMP, the wafer is
35 pressed face down against a rotating polishing pad, while a chemically and physically (abrasive) active slurry planarizes the wafer. As wafer size grows, devices sizes shrink and process requirements grow more stringent within die/wafer uni formity and removal rate increase becomes a greater concern. Di fferent CMP processes attempt to achieve a balance between removal rate and global/lo cal planarization through a combination of solution chemistry, speed, applied pressure and pad properties [56, 65]. Often a change in slurry or operating conditions lead to conflicting performance. Fig. 2.4 Schematic of Wafer Planarization by CMP Process
362.2.4 CMP Polisher considerations The key issues affecting industry use of CMP during the semiconductor chip manufacturing are the high cost of ownershi p (CoO), the lack of industry wide CMP technology and less than thorough understandi ng of the knowledge of the underlying science behind the CMP process . Improve ment in CoO has been brought about in the fast few years due to: 1) hi gher raw throughput, 2) in situ film thickness metrology, 3) dry in dry out configuration ensuring low de fectivity , and 4) process equipment, implementation and integration s upport from CMP vendors . The first generation CMP tools base d on rotational platen had low throughput values of about 10Â–18 wafers/h . The second generation tools emphasized on evolutionary improvements while the third ge neration equipment desi gns were modified to stay in production for long period of tim e by giving them adaptability to future technology modifications. The throughput of th e machine can be enhanced by increasing the removal rate and improving the wafer ha ndling. However, for effective CMP of materials, the increase in removal rate by increasing the down force should not be brought about by compromising on the defectivit y (like increase in wafer to wafer and within wafer non-uniformity, delamination, dishi ng, erosion, etc.) [70, 71] of the wafer. For this reason sometimes, the throughput is compromised to polish the wafers at lower down force there by increasi ng the polishing time . 2.2.5 First Generation CMP Polisher The first generation CMP polishers use a single robot system to move the wafer and hold it on the carrier. The polisher is comprised of two rotating platens; one
37 covered with a hard pad for bulk material re moval and other with re latively soft pad for buffing. The wafer is presse d face down on the pad by the carrier and the slurry is deposited close to the center of the pad, fr om where the centrifugal force spreads it all over. The mechanical propertie s, surface morphology, structure, absorbency, etc. strongly affect the slurry distribution and polishing [ 73]. The polishing platen on these tools is around 22 in. in diameter which is more than 7.5 times the size of a 200 nm wafer . The actual slurry utilization of these processors is poor and the pH of the slurry changes during use as there is an absence of any sl urry reprocessing unit . The amount of slurry that is actually used for processing at the interface is function of pad properties , pad conditioner , pad topography  and slurry viscosity . Thus, the pads must be conditioned to: 1) bring the pad back to flat, 2) remove materials from pores, and 3) rebuild the nap. Simple manipulation of the machine parameters is sufficient to increase the material removal rate in these polishers. However, issues such as platen wobble need to be taken care off in order to deliver CMP wafers in the acceptable range. The schematic of the first generation polishers is similar to that shown in Fig. 2.5. 2.2.6 Second Generation CMP Polishers The second generation polishers like are basically made of rotating carrier and platen designs but have numerous ch anges to improve the raw throughput. The second generation polishers can be classified br oadly in two distinct types: 1) single large (22 in.) platen polishing numerous wafer c oncurrently on the same pad, and 2) single wafer per platen, multi-platen systems .
3188.8.131.52 Multi-wafer Per Platen Polishers There is a natural incr ease in the throughput due to the increase in the number of polishing heads per platen. However, this approach presents several challenges. The most severe issue is the quantity of the wafer put at risk at one time. If one wafer breaks, the pieces can damage several wafers at one tim e. The more subtle issue is that of load balancing. As long as all the carriers on the platen are loaded with wafers, polishing can be consistent. However in ce rtain cases, like application specific integrated circuit (ASIC) fabrication where in just one or two wa fers need to be polished at one time, this issue is of considerable importance. The schematic of multi wafer per platen polisher is shown in Fig. 2.6. Fig. 2.5 Photograph of Speefam (Novellus) Multi Wafer per Platen Polisher  184.108.40.206 Sequential Rotational Systems Another approach usually adopted to a void the risk of damaging more number of wafers due to pad anomalies is sequential po lishing of wafers on di fferent platens. This
39 approach involves multi-step po lishing wherein the first platen is used for bulk material removal without particular regard to superficia l surface defects; the s econd platen is used for global planarization while the third platen is used to for a fine buff to get a defect free mirror like surface. However, s ynchronization of all these pro cesses is an issue that the process engineer needs to tackle with this approach. This implies the process speed is limited by the slowest process. This is especi ally a problem when polishing metals such as copper as the corrosion might results due to the wafer stay ing wet in the slurry for a longer duration of time [79, 80] Also, small damage to the pad can result in damage to all wafers in the sequential polishing run and it is sometimes hard to determine the damage on the pad and hence, all pads need to be changed. In case of tool failure, all the processes need to be stopped unt il repairs and tool utiliza tion is limited due to tool inflexibility. Fig. 2.7 shows an illustration of sequential rotational CMP polisher. Fig. 2.6 Applied Materials Inc., Sequentia l Rotational CMP Polisher (Courtesy: Ashok Das, Applied Materials, In c., Santa Clara, CA) 
402.2.7 Third Generation CMP Polishers The third generation polishers have a seri es of evolutionary and revolutionary systems built in them or integrated in them as modules. Dry in dry out feature is one of the most prominent enhancements in this kind of polishers. This considerably reduces the wafer defects especially in metal CMP as corrosion of the metal is drastically reduced as a result of cleaning after the polishing step. The addition of in situ metrology modules such as motor current detection , sensor array for integrated steering , thin film reflectivity  in situ optical end point detection method  ha ve markedly improved CMP process performance and reduced def ectivity. There are several new end point detection and other metrology modules such ar e integration of acoustic emission sensor , force sensor  and Cu radioactivity detection [8 7] that are candidates for implementation in the third generation polishers. 220.127.116.11 Sequential Linear Polishers The sequential linear third generation polishers are generally used in CMP for STI and rarely in ILD structures. The polishe rs have a moving belt on which the wafer is pressed device side down and rota ted slowly about th e carrier axis. The belt which is held in tension between rollers moves rapidly [ 88]. This type of polisher can achieve high removal rate owing to high be lt speed and can achieve faster planarization as for STI application where large amount of material needs to be rem oved from a relatively lower pattern density structure. The low down for ce and high relative velo city polishing regime limits the damage to the film . The linear polishers require new set and architecture of polishing pads which comprise of single polyurethane belt wit hout foam or felt (sub-pad).
41 The concurrent polishing pad conditioning is obta ined by means of a novel polishing pad design where polishing pads have to be mounted in a cylindri cal configuration and not on a the conventional flat surface configuration . A special polishing pad conditioner is provided to refurbish the polishing pad . With more and more publication of data showing improved CMP performance at low down force and high linear velocity, this type of polisher is finding increasing accep tance in the semiconductor processing industry. Fig. 2.8 shows a photograp h of a sequential linear polisher. Fig. 2.7 Photograph of Sequential Linear Polishing System  18.104.22.168 Orbital Polishers Several CMP tool concepts have been developed based on orbital motion. Some orbit the carrier with rotating the carrier [91-93] while others orbit the platen while rotating the carrier . Some of the polishers also i nvolve arbitrary non rotational motion on a fixed polishing pad. In these types of polishers the fundamental principle of relative motion between the wafer and pad to remove the material is used, however,
42 unlike the first generation polishers, slurry is delivered directly at the pad area used for polishing thus improving slurry utilization efficiency. The schematic of the orbital polisher is shown in Fig. 2.9. The preferred mode of operation of orbital polishers is low down force with high relative velocity. With the recent popular ity of the contact retaining ring method for polishing, the pad remains compressed at the edges of the wafer and reduces the area of the die lost due to edge exclusion . The plan arization capabilities of these tools are known to be better than the first and second generation polishers. These of machines are also known for their small dow n times due to the rapid change individual polishing heads. Fig. 2.8 Schematic of Rotary CMP Polisher  22.214.171.124 Rotary Inverted Recently, Nikon Inc. has developed high-precision CMP systems applying proprietary technology based on its long e xperience in lens polishing and optical measurement . The system's special face-up polishing uses small pads applied at very
43 low pressure and high-speed rotation. The comp act polishing pad as se en in (Fig. 2.10 a) enables high-speed rotation. Its light-weight, less pad deformation feature allows superior planarity. These features are especially advantageous ultra lowk polishing process. Through the compact polishing pad the slurry is supplied onto polishing area of the wafer efficiently. This enables less slurry consumption compared to the conventional polishing equipment (Fig. 2.10 b). Face-up polishing at the polishing station enables continuous optical end-point measurement. It helps impr ove S/N ratio while minimizing slurry effect for end-point detection. Mountin g of wafer and pad are also convenient with this system. In spite of some initial excitement about th is system, it is yet to have a proven track record in the semiconductor fabrication environment. Fig. 2.9 a) Set up of Multi Wafer Rotary Inverted CMP Polisher and b) Polishing Action and End Point detection  126.96.36.199 Pad feed Polishers The pad feed polishers are based on a r ecently developed pad type that is held in rolls. These polishing pads are fed to the wafer polishing tables, the wafer is polished,
44 pad is conditioned and then the pad moves further (Fig. 2.11). This methodology is especially useful for pads that have very repeatable first polish performance and their characteristics either degrade or change w ith subsequent polishing runs. This polisher unlike others does not have to be turned o ff for changing the pad th ere by maximizing the equipment utilization time [ 97, 98]. This technology is still in its nascent stage and various industrial giants continue to devel op it even further in order to make a positive impact on the CMP polisher market. Fig. 2.10 Schematic of a Web-type Polisher [98, 99] 3.3 Physics of CMP Process In 1927, the Preston equation was devel oped (Eq. (2.1)) , for modeling the mechanical effects of pressure and velocity in the CMP process: R = KPV (2.1)
45 where R denotes the polish rate, P is the applied downward pressure, V is the linear velocity of the wafer relativ e to the polishing pad, and K is a proportionality constant, called the Preston coefficient. As seen in Fig. 2.12 rcc is the linear distance between the centers of the wafer carrier and the platen, which is mostly assumed to be constant, rH (H stands for head) is the positional vector at any point Q on the wafer from the center of the carrier and this varies with the position of the point Q rth is the positiona l vector of any point Q on the wafer from the center of the platen (this distance varies with the location of Q ). VQ is the velocity of any point Q on the wafer. VT and VH are the linear velocities of the table (platen) and wafer head. T, H are the rotational veloci ties of the platen and wafer carrier. Assuming that T H velocity VQ will vary from point to point over the wafer. The variation in velocity will call cause changes in the removal rate across the wafer in accordance with the Preston's equation. rth= rcc+ rH (2.2) VQ= VT+ VH= ( T rT)+( H+ rH), where VQ=( T rcc)=[ rH( T H)] (2.3) Fig. 2.11 Schematic of the Force Field on the Wafer and the Pad during CMP [37, 70] However, if T is set equal to then linear velocity will be independent of the location of the wafer for VQ = [ T rcc]. This with the VQ maintained as shown before, the velocity of all points on the wafer will be the same and then there will be no change
46 in the removal rate of the material. This fo rce field analysis is taken in to account to fundamentally design any CMP process [37, 70] In any event, many times, the process engineers are still confronted with the problem of wafer to wafer and within wafer nonuniformity (WIWNU and WTWNU) (Fig. 2.13) . Fig. 2.12 Non-uniformity in Removal Rate with in a Wafer  3.4 Parameters Governing CMP process The physical forces acting upon the wafe r govern only a part of the entire tribo-chemical phenomenon of CMP. The pro cess of CMP is governed by various input variables that act on a micro as well as nanoscale to produce desirable or undesirable output parameters. The interplay and inte raction of different micro and nanoscale parameters takes place sequentially or simu ltaneously, predominantly at the pad wafer interface, during the entire CMP polishing run. Though machine parameters when input in the PrestonÂ’s equation may appear to be the dominant in any CMP output, the entire governing dynamics of industrial scale CMP is much more complex. Fig. 2.14 is an attempt to correlate the various para meters in the CMP governing dynamics.
47 Fig. 2.13 Parameters Govern ing the CMP Dynamics 
483.5 Research Objectives The broader objective of this research is to understand the process of CMP which has been more of an art that scie nce. The specific objectives, milestones and subjects of investigation al ong with the expected outcome are shown in Fig. 2.15. Fig. 2.14 Schematic Showing Objectives of the Current Research Study of CMP Process Investigation and Modification of CMP Pads 1. Evaluation of Pad material properties 2. Evaluation of pad non uniformity 3. By-passing Pad conditioning 4. Making Pad specific to Polishing Application 5. Tribological Evaluation of Pad 6. Determination of Polishing Pad Architecture Effect of Machine Parameters 1. Variation of down force 2. Variation of Linear Velocity (Platen Speed and Carrier speed) 3. Variation of Slurry flow 4. Defect Detection and Process Optimization Investigation and Modification of CMP slurry 1. Evaluation of Slurry Frictional Performance 2. Slurry selectivity 3. Removal Rate 4. Overall slurry performance 5. Introduction of surface catalyst Material Aspect of CMP 1. Material Inspection 2. Mechanical Properties 3. Interfacial Adhesion 4. Pre-CMP material Characterization 5. Material Tribological Properties Contribution towards CMP Pr ocess Module Development Investigation and Modification of Chemical Mechanical Planarization Consumables
CHAPTER THREE METROLOGY OF CMP 3.1 Need for Effective CMP Metrology The CMP process combines mechanical and chemical removal mechanisms in a synergistic effect. This synergy has been th e subject of many studies, but focus in the past has been primarily on mechanical eff ects due to the difficulty of identifying the reaction mechanisms of the ch emical effect. However, mechanical effects alone cannot provide the type of polishing necessary for IC manufacturing. Chemical effects contribute to the increased global planarity and redu ced micro roughness required for successful IC fabrication. As discussed in the earlier section, the fundame ntal basis for designing any CMP process module, the force field analysis of the waferÂ–padÂ–slurry abrasion system is made. The variations in the machine parameters to obtain optimal results are the first adjustments made to refine the CMP process. Until recently, slurry flow and slurry flow rate was not given much importance variation of machine parameters , however, with the ever-growing demands for enhanced yiel d and low defects, and also with the knowledge of the heat transfer behavior of the slurry , the slurry flow is also brought in the CMP process control equation. This se ction discusses the broa der impact of these machine parameters on the CMP process. A be tter understanding on th e effect of machine parameters on the CMP process can be obtained by performi ng repetitive CMP
50 experiments on a prototype CMP tester in whic h process data is monitored in situ. We have used the CETR bench top. for studying th e CMP process in detail. The details of the CMP tester are described be low and also can be obtaine d in literatu re . 3.2 CMP Tester Most of the CMP processes during th is research were performed on a CMP tester (CETR, Inc, CA) with variety of pro cess parameters. A photograph of the tester is shown in Fig. 3.1. The lower platen can hold a pad up to 6Â” in diameter. The upper carriage can hold two inch wafers or sa mple coupons up to 1.5Â” X 1.5Â”. The upper carriage can rotate about its own axis a nd oscillate radially with the pad during planarization process. A strain gauge force sensor (0-200 N) can record both vertical and friction forces and hence co-efficient of fric tion (COF) is monitored during the process. The system is also equipped with a hi gh frequency acoustic emission sensor. AEAnalysis is an extremely powerful technology that can be deployed within a wide range of usable applications of nondestructive testing: metal pre ssure vessels, piping systems, reactors, and similar. All solid materials have certain elasticity. They become strained or compressed under external forces and spring b ack when released. The higher the force and, thus, the elastic deformation, the higher is the elastic energy. If the elastic limit is exceeded a fracture occurs immediately if it is a brittle material, or after a certain plastic deformation. If the elastically strained mate rial contains a defect, e.g. a welded joint defect, a nonmetallic inclusion, incompletely welded gas bubble or similar, cracks may occur at heavily stressed spots, rapidly rela xing the material by a fa st dislocation. This rapid release of elastic energy is what we ca ll an AE event. It produces an elastic wave
51 that propagates and can be de tected by appropriate sensors and analyzed. The impact at its origin is a wideband movement (up to so me MHz). The frequency of AE testing of metallic objects is in the range of ultr asound, usually between 100 and 300 kHz. The acoustic emission sensor employed in this test er has a frequency range between 0.5 to 5 KHz. The AE sensor, in conjunction with COF, has been used to dete ct the delamination, endpoint, and debris during polishi ng (described later) during th e course of this research. Fig. 3.1 Photograph of the CETR CP-4 CMP Tester The state-of-the-art CMP tester is a testing tool and thus the following assumption are made when different pads, slur ries and materials we re evaluated: 1)Due to the lack of uniformity on th e surface of the coupon, an aver age of the material removal rate measured at different points from the cen ter to the edge was assumed to be the MRR Pad with Lower Platen Slurr y Flow Slider Assembly Force Sensor Upper Carriage Slurry Outlet
52 of the polishing run, 2)During the in-situ detection of MRR using COF and AE, the non uniformity of the sample surface brought about gradual tapering of the signals even for blanket samples. The end point was assumed to have occurred when more than 70 % of the material was removed from the surf ace and further material removal brought insignificant change in the AE and COF signals. 3.3 Coefficient of Friction (COF) a nd Material Removal Rate (MRR) To understand the variation of COF and MRR, which are the primary output variables of any given CMP process, the pol ishing tests on the CETR tribometer. The samples used in this case were 1 in. 1 in. PECVD SiO2 using Klebesol 1501 (Rodel Inc., DE) colloidal s ilica slurry (pH 10Â–11) on an IC 1000/IV pad with linear velocity 5 mm/s and a radial distance of 50 2.5 mm. The down force used was 4 PSI and the platen rotation was 150 RPM. Influence of machine parameters such as down force, relative velocity, slurry flow on the acoustic emission (AE), coefficient of friction (COF) and material removal rate (MRR) was observe d. For removal rate calculations, thickness of oxide was measured at nine points us ing the ellipsometer. The wear rate was calculated by re-measuring the samp le after polishing at nine points. COF is an important tribological prope rty of films and pad as it gives an estimate of the surface shear which directly affects the MRR. The COF was recorded during all the tests. Fig. 3.2 (a, b) show s the COF versus RPM and PSI, respectively, during polishing. With higher RPM, COF decr eases, whereas, decrease of COF is very small with the increase of PSI.
53 Fig. 3.2 a)Variation of COF with RPM (plate n velocity) for Different Values of Down Pressure (PSI), b) Variation of COF with PSI (Down Pressure ) with Different Values of Platen Velocity (RPM) Fig. 3.3 a, b shows the variati on of MRR as a function of RPM and PSI, respectively. Experiments on two sets of sa mples show the same trend. Removal rate increases with both RPM and PSI. Removal rate decreases slightly at platen rotation 250 RPM. This may be due to inadequate slu rry flow under the sample at higher platen rotation. As COF decreases with increasing RPM and PSI, a lower COF may be related to the higher removal rate in CMP process. Fig. 3.4 shows the removal rate versus RPM PSI and the linear relation indicates that pol ishing of oxide follows Preston's equation .
54 Fig. 3.3 a) Variation of Average Removal Ra te with RPM at Different Down Pressure (PSI) for the Evaluated Set of Samples, b) Variation of Average Removal Rate with PSI (Down Pressure) at Different Platen Veloc ity (RPM) for the Evaluated Set of Samples
55 Fig. 3.4 Average Material Removal Rates Plotted with RPM PSI. Linear Relation Indicates that Polishing Fo llows Preston's Equation 3.4 Importance of Slurry Flow It is important to optimize the effect of slurry flow rate on the COF and AE signal. Results of the optimization experiment s are summarized in Table 3.1. It can be seen from Table 3.1 that COF decreases s lightly while no significant change can be noticed in AE signal. Decrease of COF may be attributed to the higher slurry flow rate during polishing. Therefore, the data suggests that flow rate may not affect the AE signal. The flow pattern of the slurry on the pad affects the polishing rate as well as the WIWNU. Due to the rotation of the lower platen the flow pattern will be different as we feed the slurry at different positions of the pad. Fig. 3.5 shows the different positions of slurry feeding on the platen. If the slurry ca nnot reach uniformly at the pad-film contact points material will not be removed uniformly. It can be seen from Fig. 3.6 that center
56 position and position Â“8Â” (very near to the center) are two better positions to feed slurry while platen is moving clockwise. Table 3.1 Effect of Slurry Flow rate on COF and AE Run # Slurry flow (ml/min)CO F AE signal (arbitrary unit) 1 35 0.39770.4013 2 75 0.39490.4533 3 100 0.39320.4184 4 155 0.39110.4133 5 195 0.38880.4189 Fig. 3.5 Schematic of the Positions of Sl urry Feeding on the Pad during Polishing for Feeding Position Optimization; Dist ance 0Â–1.4 = 15 mm, 0Â–2.5 = 30 mm, 0Â–3.6 = 45 mm, 0Â–7.9 = 45 mm and 0Â–8 = 25 mm The friction generated during CMP br ings about over all increase in the temperature at the wafer pad interface. Certain CMP processes such are silicon polishing are exothermic. Hence, there is a natural incr ease in the temperatur e at the interface. The increase in temperature changes the reaction ki netics of the slurry with the wafer, mostly
57 increasing the removal rate. However, the incr ease in removal rate due to the increased chemical action of the slurry at elevated temperature does not always translate in to greater removal rate during the CMP process . The increase in temperature makes the viscoelastic polyurethane pa d softer, there by reducing th e removal rate due to the reduction in hardness . Hence, an optimum slurry flow must be maintained during the process and should be changed if necessary in order to strike a balance with optimum temperature for enhanced slurry action and non-degradation of the pad . Novel CMP process developers have adopted a new recipe to change the slurry flow during the CMP process for optimization of slurry utility a nd maintaining the temperature during the CMP process [104,105]. Fig. 3.6 Average Removal Rate with the Sl urry Feeding Position on the Lower Platen Position
583.5 CMP End Point Detection (EPD) The difference in removal rate of one material as compared to another in a given slurry yields the measure of the selectivity of that slurry for those two materials. Higher the slurry selectivity, the more effectiv e is the end point dete ction for a particular CMP process step as there is a marked cha nge in the tribological properties of the material being polished and th e under layer. Selectivity is a very important criterion in designing any slurry. For STI, slurry used needs to act on the oxide that is being planarized and not act on the underlying nitride. For polishing Cu, the slurry needs to act selectively on Cu and spare the barrier laye r Ta and underlying layers of silica or lowk dielectric material. Typically slurry sele ctivity of 10Â–25 has been reported for Cu polishing . The selectivity could be considerably improved up to 1000 by introducing particle free slu rry [106, 107]. The reduction in mechanical component due to lack of abrasive implies that majority of material removal takes place due to solution chemistry which can be made highly selective. Slurries with high selectivity facilitate easy end point detection as the tribological properties of the material say Cu being polished in a highly selectivity slurry are markedly different from the properties of the barrier layer Ta or underlying silica layer when polished in the same slurry. The difference in tribological prope rties can be monitored in s itu using techniques such as motor current or force and acoustic emission sensor. The variation in coefficient of fric tion and acoustic emission for polishing of blanket Cu, Ta and ultra lowk dielectric ( k = 2.2) has been studied. The candidate materials have been polished in the form of 1 in. 1 in. coupons on the bench top CMP tester (mentioned in Section 3.2)  to evaluate the selectivity of the slurry. The
59 slurries evaluated were: 1) Cu selective alum ina particle slurry (Cu1), 2) Cu selective particle less slurry (Cu2), 3) Ta selective slurry with colloidal abrasives (slurry Ta), and 4) non-selective slurry (slurry CuÂ–Ta). Fig. 3.7 aÂ–d shows the vari ation of COF and AE at 2 PSI and different platen sp eeds. It can be seen from the figure that the value of COF for a particular material for one polishing condition is unique and hence monitoring the value of COF can give an estimate of the end point of the process. Fig. 3.7 a) Variation of COF at 2 PSI Down Force and Variable RPM in Slurry Cu1, b) Variation of COF at 2 PSI Down Force and Vari able RPM in Slurry Cu2, c) Variation of COF at 2 PSI Down Force and Variable RPM in Slurry Ta, and d) Variation of COF at 2 PSI Down Force and Variable RPM in Slurry CuÂ–Ta [85, 86] Just as the COF is a strong function of the down pressure, platen and carrier velocity, and the choice of the slurry and pad, the acoustic emission during CMP is already affected by the choice of the aforementioned CMP vari ables. The variation of
60 AE and COF for different materials polished in a commercial Cu slurry is shown in Fig. 3.8 a. It must also be noted here that th e same materials displa ys different COF with polished using different slurri es. This can be seen from Fig. 3.8 b where different materials, namely Cu, SiLKTM (ILD material) and Ta are polished in commercial slurries named here are Â“Cu2Â”, Â“TaÂ” and Â“Cu-TaÂ”. The sl urry Cu2 is selective to Cu while Ta is selective to Tantalum. The slurry Cu-Ta is a non se lective slurry. Fig. 3.8 a) Variation of AE and COF for Di fferent Materials Polis hed in a Commercial Cu Slurry, b) Variation of COF for 3 Mate rials Polished in Thr ee Different Slurries 00.511.522.5 Cu Ta Lowk B Lowk C SiC SiO2 COF/AE AE COF 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Slurry Cu 2Slurry TaSlurry Cu-TaCOF Cu SiLK Ta a b
61 Addition of specialized ch emicals that can act as catalysts in the chemical interaction between slurry and the polished ma terial there by increasing the rate of the chemical reaction with the material being removed considerably is a common approach for improving the selectivity of the slurry. Tetra-methyl ammonium hydrate (TMAH) can be added to Cu slurries to considerab ly decrease silica polis hing rate [108, 109]. Phosphoric acid added to alumina and collo idal silica TaN slurry has also shown accelerated chemical reaction with TaN. Fig. 3.9 shows the increase in the polishing rate of TaN with addition of phosphor ic acid in alumina and colloi dal silica slurry . The effect of surface catalyst in the slurry has been investigated and details have been reported in later in this dissertation. Fig. 3.9 Cu and TaN Polishing Results (H ead 40 RPM, Table 40 RPM, and Pressure 7 PSI) 
62 Among the several different models pr oposed for friction and interfacial heat dissipation, pad-wafer interac tions during CMP tend fall in the molecular friction model at a nanoscale. The molecular friction model essentially postulates that when two different surfaces come in contact, the imbalanced Van der Vaal forces of the surface molecules results in adhesion or Â“stictionÂ” between the two interfaces. It must be mentioned here that if one of the surface is significantly rougher than the other (as is the case with CMP when pad is significantly rough er than the polished wafer), macro-scale parameters such as Â“asperity lock-inÂ” also govern the friction phenomen a. In the event of an abrasive containing fluid present at the pad wafer interface, the overall energy required to overcome the combined effect of the Van der VaalÂ’s interaction of pad, slurry and abrasive materials with the polished wafer. Hence, the choice of the polished material, polishing pad, and CMP slurry will directly influence the friction force and hence the COF at the pad wafer interface during polis hing. This interaction of the different aforementioned materials will also impact the strain to which the material subjected to during polishing at the given pol ishing conditions. The strain and the strain relaxation which is the root cause of the elastic wave se nsed by the AE sensor will thus change for different process recipes and parameters. U pon complete removal of polished material, the exposure of the material underneath will change one of the variables (the polished material) in the interactions taking place at the interface of pad and wafer. This will in turn affect the COF and AE values measured in-situ. This principle could be used for effective CMP end point detection. The time taken for complete material removal can give an reasonable estimate of the material re moval rate during CMP. If the polishing is
63 continued for sustained period of time, then the time taken for the complete material removal of the buried layers can give an idea of the selectivity of the slurry. 3.6 In-situ Process Monitoring and Slurry Selectivity The previous section described in-s itu monitoring of CMP process of predominantly blanket samples. However, in reality, wafers that undergo CMP have an intricate pattern based upon the design la yout. This gives many new paradigms to the CMP process monitoring. However, the knowledge gained from in-situ process monitoring of blanket samples can be used for studying the CMP of STI structures. The variation in pattern density in the topography of the wafe r can accentuate a special type of defect known as Â‘dishingÂ’ [111, 112]. A schematic illustration of dishing and erosion defects is shown in Fig. 3.10. This dishing effect is characterized by high polishing rates in localized regions where the pattern is si gnificantly different from its surrounding. The formation of the trough shaped dish has been attributed to exce ssive over polishing in these areas. Efficient end point detection (E PD) helps prevent these and numerous other CMP defects. Fig. 3.10 Schematic Illustration of Dishing and Erosion
64 The overview of the STI process has already been given in Chapter 1. However, before the process monitoring of STI CMP is discussed, it is important to understand the background and layout of the evaluated structures that would undergo CMP. In STI structures, the field oxide is em bedded in the Si wafer in order to clearly separate the device active areas. This allows smaller device pitches and higher packing density. Fig. 3.11 shows an SEM microgra ph of the Focused Ion Beam (FIB) cross section of an STI structure. Fig. 3.11 SEM Cross-section of STI Structure The process of the fabrication of ST I structures (besides minor possible process variations) is relatively well characteri zed and is already discussed in literature . Some of the teething challenges in ST I process are: 1) effective trench oxide filling, 2) thermal stability and wet etch re silience of the oxide, 3) with in die non uniformity after CMP of field oxide, 4) in complete polishing of the oxide, 5) oxide dishing and nitride erosion, etc. Most of th ese STI process challenges can be resolved Trench Oxide Needed Step Height Reduction of field Oxide Potential Active Device Area Si3N4 Layer
65 using effective process EPD. It can be argue d that during the polishi ng of STI structures, the COF will change when the nitride layer is exposed after removal of field oxide. In this study, we used the variation of COF during the CMP process for in-situ EPD. CMP polishing run was carried out for an extended period of time to determine the oxide and nitride MRR and oxide to nitride se lectivity of the slu rry. The surface of the wafer was characterized ex-situ using Atom ic Force Microscopy (AFM) before, in the middle, and after process end point. The evaluation of sample surface morphology was done using AFM in the tapping mode. The surface morphology before polishing showed 3.5 m X 3 m pedestals with 2 m device area (Fig. 3. 12). The AFM characterization was also performed after partial removal of oxide during CMP as well as the complete removal of field oxide. The process parameters specifically studied were: 1) variation of COF with time, and 2) variation of material removal rate (MRR) for each material under different polishing conditions. The deta ils of the experiments are shown in Table 3.2. Fig. 3.12 3D AFM Image of the Evaluated ST I Structure with Step Height Measurement Vertical Distance: 375.50 nm Horizontal Distance: 2.969
66 The COF, which is the ratio of the shear force (Fs) to the normal force (FN) ( N sF F COF ), is measured in situ during the proc ess. When a patterned wafer (in this case STI) is pressed face dow n on the surface of the polishi ng pad under a particular value of down pressure, the pressure is dist ributed uniformly over the entire topography. The density of the pattern thus strongly a ffects the local pressure on the surface of the wafer. The step height reduction takes place at a brisk rate during the initial stages of the CMP process due to the higher local pressure. As the material removal from the surface of the wafer takes place primarily due to shea r, higher shear force generated due to the aforementioned mechanism is the cause of higher values of COF after carrier touch down. The shear force also depends upon the chemically modified surface layer that results due to the interaction of the surface and the slurry. Table 3.2 CMP Experimental Details # Parameter Conditions 1 Down Pressure 4 PSI for EPD experiments, 8 PSI for Slurry Selectivity 2 Platen Rotation 200 RPM Platen Rotation, 195 Carrier Rotation 3 Slider 35mm 5mm (positio n) @ velocity of 5 mm/ sec 4 Slurry ILD 1300 (pH~10.5) with colloidal silica abrasives 5 Pad Rodel, Inc. IC1000 Suba IV A4 perforated 6 Time of polishing Until Process End Point 7 Polishing Specimen 1Â” X 1Â” coupon
67 Fig.3.13 In situ Variation of COF during CMP Split in to Different Stages In case of STI patterns being polishe d by ILD slurry designed to planarize SiO2, the planarization chemical mechanism i nvolves: 1) penetration of water in the material surface in presence of an alkalin e medium, 2) surface layer dissolution, 3) removal of hydrolyzed material by particle impact, 4) Partial redeposition of dissolved products, 5) reaction of surfactant on the abrasi ve on material remova l . Any change in the material removal mechanism will cause th e change in the nature of interactions at the pad wafer interface. There is a drastic change in the sl urry-surface interaction when the Si3N4 layer is exposed after the removal of field oxide In comparison with SiO2, the hydroxylation of the Si3N4 surface is markedly reduced. Also, silicon nitride being significantly harder than SiO2, the silica particles tend to roll over the surface due to the slurry flow and show lesser te ndency for indentation. This causes lesser shear for a given down force, which in turn implies that COF values when Si3N4 is polished by silica ILD slurry are lesser as compared to SiO2. However the COF measured during Si3N4 polishing 0100200300400500 Time sec -1.5 -1.0 -0.5 0 0.5 1.0 1.5 COF-Ff 0.3117 58 163 del 105 Stage A Oxide Polishing Stage B Nitride Polishing Stage C Nitride RemovedOxide Removal Rate: ~ 400nm/min Nitride Removal Rate: ~ 80 nm/min Slurry Selectivity: 5:1 Process Conditions: 8 PSI/ 200 RPM (platen) -195 RPM (carrier) 0100200300400500 Time sec -1.5 -1.0 -0.5 0 0.5 1.0 1.5 COF-Ff 0.3117 58 163 del 105 Stage A Oxide Polishing Stage B Nitride Polishing Stage C Nitride RemovedOxide Removal Rate: ~ 400nm/min Nitride Removal Rate: ~ 80 nm/min Slurry Selectivity: 5:1 Process Conditions: 8 PSI/ 200 RPM (platen) -195 RPM (carrier)
68 is the combined COF of the Si3N4 pads and surrounding trench SiO2. Hence, though there is a difference in COF value when Si3N4 is being polished, the magnitude of the COF is not markedly different (Fig. 3.13). The COF value again changes when the Si3N4 is completely removed and field oxide is po lished. The CMP experiments were first performed on 1Â”X 1Â” STI coupons at a down force of 8 PSI and 200 RPM platen/ 195 RPM carrier rotation for 500 seconds. During the stage A, the COF continuously goes down after the initial high at touch down. Another CMP process run was pe rformed and stopped before completion of Stage A. Based upon the previous discussion, it can be thus hypothesi zed that progressive decrease of COF during the first stage corresponds to SiO2 material removal. As COF is a characteristic of particular material fo r the given set of polishing conditions and consumables, after complete removal of SiO2, and exposure on the ni tride regions, there is a change in the COF. This transition can be seen to reach after 58 seconds for the polishing run reported in Fig. 3. 13. For exposure of Si3N4, the step height of the oxide has to be removed, along the surrounding excess field oxide. The height reduction was approximated to 400 nm and it implied that SiO2 material removal rate (MRR) during the process run was approximately to 400 nm and that implied that SiO2 materi al removal rate (MRR) during the process run was 400 nm/ min. Further polishing of sample coupon showed that CMP of Si3N4 produced 80 nm/min MRR. Thus the oxide to ni tride selectivity of the slurry at the given process condition was found to be around 5:1. This selectivity ratio is in agreement with the industry standards .
69 Fig. 3.14 Ex-situ 3D AFM Image of the Ev aluated STI Structure with Reduced Step Height before Termination of Stage A (ref. Fig. 3.13) Fig. 3.15 Ex-situ 3D AFM Image of the Eval uated STI Structure af ter Process End Point just after Beginning of Stage B (ref. Fig. 3.15)
70 Another polishing run was performed till the process end point was reached (ref Fig. 3.13) and COF transition was observed The sample coupon polished just until the beginning of Stage B was characterized ex-situ using AFM as shown in Fig. 3.14. It can be seen from Fig. 3.15 that, the nitride regi ons on the surface of th e samples are exposed. This shows that the CMP process has reached its end point due to the complete removal of the field oxide. The Â“spikesÂ” on the su rface of the AFM 3D image are due to the particle adhesion. The partic les adhered to the surface due to non availability of an efficient post CMP set up. The AFM section an alysis shows a planar surface with and surface variations can be pr imarily attributed to the surface roughness. The evaluated COF during the entire stage B is a combina tion of the COF of nitride and oxide. With progressive decrease of nitride layer during CMP, the value of COF tends to increase till nitride is completely removed and oxide is exposed. To eval uate the repeatability of the hypothesis, CMP runs were perf ormed at different values of down pressure and platen rotation. The data of the two of such trials performed at a down pressure of 4 PSI, 200 RPM platen rotation and 195 RPM ca rrier rotation is shown in Fig. 3.16. It can be seen from the figure that the CMP process end point occurs at 200 sec. for both process trials. Also the variation of COF follows a similar trend for both the trials. Thus it can be inferred that for the given pr ocess conditions and parameters the process end point is reached at almost a similar time with in ma rgin of acceptable error. The spikes in the COF data could be attributed to the noise dur ing data collection. In order to filter the noise, obtained sharp transitions of end point, and study the Â“eventsÂ” such as microscratching, dishing, erosion etc., duri ng the process, the CO F data needs to be further filtered using specialized signal pr ocessing filters. As the COF is data is non
71 stationary and even random to some extent (depending upon process defects or Â“eventsÂ”), the COF signal has to be filtered in time as well as frequency domain. Wavelet based techniques have been found appropriate for sharp end point detect ion from using COF . Fig. 3.16 In-situ data for 2 Trial Runs Shown to Demonstrate repeatability Thus, it can be concluded from this study that real time monitoring of the CMP process can give valuable informa tion which otherwise would need ex-situ metrology for evaluation. The values of CO F being the Â“signature Â” of a particular pattern, material, pad, slurry and polishing cond itions can be used as a fingerprint of a particular STI CMP process run. The work could be further extended in evaluating different pattern densities and different High Selectivity Slurries for STI CMP.
723.7 Delamination during CMP The next generation lowmaterials are generally por ous in nature and poor adhesion of dielectrics in a stack may resu lt in delamination during the CMP process. Loading forces and rotation rates may also affect the reliability of the structure. Delamination can occur due to either: 1) adhesive failure at the lowand barrier layer interface, 2) failure within the lowmaterial, or3) failure of the cap layer and lowinterface. The process conditions must be adjust ed in such a way that the overall hardness of the pad surface does not cause the aforementioned defects. It has been found that the density and size of abrasives along with the CMP process conditions have a strong effect on delamination during materi al planarization . To understand and demonstrate the phenomenon of delamination, three different types of materials namely: 1) Patter ned Cu samples with low-k A (A) (k~<2), 2) Patterned Cu samples with low-k B (B) (k~2 .0), 3) Patterned Cu samples with SiO2 (TEOS) (k~4.0). Henceforth, these samples will be referred to as ACL, BCL, and TC respectively. Small coupons (2cm X 2cm) of samples were used to polish on either 6Â” or 9.5Â” pad coupon at different ro tation of platen (0.2 to 1.5 m/s) and down force (1-10 PSI). Upper specimen was allowed to oscillate b ack and forth by 5 mm (47.5-52.5 mm) with a speed of 1 mm/s. Polishing slurry was fed into the interface continuousl y with the rate of 55 ml/min. On some occasions, tests were st opped before 300 sec due to high vibration and sample holder rotation. Slider movement for these experiments were 502.5 with a velocity 1 mm/s. Slurry flow was fixed at 55 ml/min. Polishing pad was a perforated IC1000/SubaIV while alumina based Cu select ive slurry was fed c ontinuously at the center of the pad.
73 Fig. 3.18(a) shows the AE signal vari ation while polishing with active upper rotation and Fig. 3.18 (b) shows the optical pict ure of the coupons afte r polishing. As it is seen for the samples polished with non upper rotational machine, highest AE signal is recorded for sample ACL and lowest signal for sample TEOS while a little smaller value is recorded for sample B than sample A. As sample ACL consists of Cu-ultra low-k samples with lowest k values, delamination may be more for this sample. It can be seen from Fig. 3.18 (b) samples get polished more uniformly with an upper rotation whereas samples gets more polished at the leading edge only for no upper rotation. From the samples it could also be seen that all the ACL and BCL samples are delaminating whereas TEOS has no sign of delamination. 050100150200250300350 Time sec 0 1 2 3 4 5 6 AE Volt Fig. 3.17 a) Picture of Sample Coupons ACL, BCL, and TC af ter Polishing, b) Variation of AE signal of Three Different Samples Further analysis of the samples with SEM, shown in Fig. 3.18, reveals that A B TEOS a b
74 during the process Cu either gets peeled off or delaminated. Even delamination of underlying interlayer dielectric films could be seen. In all conditions, and experimental setups AE signal was higher for sample A and B than TEOS. Also, sample ACL produced more AE signal than sample BCL. It is prudent to mention that the only difference among the three samples is the unde rlying ultra low-k ILD film. High AE for ACL and BCL sample may be an indication of delamination of Cu from barrier and ILD interface. As Low-k B has slightly higher k values, which implies better mechanical strength, sample B shows better resistance to the CMP process. Lower magnitude of AE for sample BCL than sample ACL is the reason for this conclusion. ACL 2 ACL 2 ACL6 BCL4 BCL 5 TC 12 Fig. 3.18 SEM Micrographs Showing of Dela mination of Different ACL, BCL Samples and No Delamination for TC Sample Severe delamination or scratching on th e surface could be monitored with the AE sensor, which produces spikes during th e unusual polishing. Two ACL samples were run in the same conditions along with a TEO S sample. The expanded AE signals for all three samples are shown in Fig 3.19(a). Optic al and SEM picture of the sample coupons are also shown in Fig. 3.19 (b) and (c). It can be seen from Fig. 3.19 (a) that several
75 spikes occur for the sample with visible scratch on it. AE signal for TEOS shows very smooth and reduced noise level. 230235240245250 Time sec 0 1 2 3 4 5 6 7 8 9 10 AE Volt Optical (A006) SEM (A006) Optical(A006r) (a) (b) (c) Fig. 3.19 (a) Variation of AE Signal during Polishi ng in the Time Interval of 230 Â– 250 s for Three Different Samples. Peaks are Seen for Sample ACL006, (b) Picture of Sample ACL006 and (c) Picture of Sample A006r As mentioned earlier, the AE signal monitors the elastic acoustic wave in the ultrasound range during the CMP process. The shear generated by the down pressure and platen rotation brings about a stra in in the thin film that is being polished and thus is also responsible for material removal. If the sh ear force is sufficient enough to overcome the interfacial adhesion of the thin film and th e buried layer, the inte rfacial adhesion energy is dissipated in the form of acoustic vibr ation, besides other modes. This excessive emission of acoustic energy dur ing polishing due to interfacial delamination is also sensed by the AE sensor. This results in th e higher magnitude of AE signal in the event of delamination during the CMP process. In -situ monitoring of the AE signal during TEOS A006 A006r
76 CMP can thus give a qualitative estimation of delamination. Filtering of AE sensor data using frequency and time domain filtering (Wav elet based approach) can provide further meaningful information of the CMP process. In-situ isolation and quantification of the phenomenon of delamination may also be help ful in setting up a feed back mechanism which based upon statistical process control pr inciples. Work related to Wavelet based AE signal filtering and analysis has been published in literature . 3.8 Other CMP Defects Besides, delamination and faulty end poi nt detection, several defects plague the process of CMP. The defect of microscrat ching results in the ev ent of indentation or impingement of abrasive being deeper than the modified surface. Excessive surface passivation during polishi ng results in the predominantly mechanical removal of material and reduces the MRR during CMP. The loss in global planarity and poor surface finish that my sometimes result after CMP can have severe implications on the subsequent lithography step. Due to these reasons, it is important to understand the behavior of candidate materials from a CMP perspec tive and understand the challenges surrounding the successful integration of these materials. These and allied issues will be discussed in the subsequent chapter.
77 CHAPTER FOUR PROPERTIES OF INTERCONNECT MATERIALS 4.1 Need for Evaluation of Material Properties As elaborately discussed in Chapter1 there is a general agreement on the choice of Cu as the interconnect material fo r subsequent generations of microprocessors for the time being. However, integration of Cu alone (without taking the low-k dielectric into consideration) in the damascene process also comes with its share of challenges. Copper is typically much softer (microhardness: 80 Kg/mm2; Mohs scale: 2.5) than the conventional materials presently being used (t ungsten and silica) in the interconnect CMP technology. This is in sharp contrast with the slurry abrasives made up of silica (hardness: 1200 kg/mm2; Mohs scale: 6-7) or alumina (hardness: 2000 kg/mm2; Mohs scale: 9). Hence, use of these abrasives resu lts in formation of p its, craters, micro and macro scratches on the surface after CMP . Before developing a CMP process for Cu, one has to evaluate the tribol ogical properties of Cu under different process conditions. Also, certain peculiar traits Cu such as self annealing and varying impact of post electroplating annealing also n eed to be considered before implementation of Cu. Besides Cu, the polishing performance other materials su ch barrier layer Ta or TaN and different novel lowk materials is also important to determine the process CMP process recipe. Several lowmaterials do not meet thermal, mechan ical and electrical requirements for
78 their integration with metal in the damas cene structure. For CMP applications, lowmaterials have to be strong enough to prev ent delamination and mechanical breakdown. Due to all these reasons, it is important to devote some part of studies performed on materials from a CMP to their reliability issues which may surface during or after processing. In this chapter will study the prope rties of different materials that potentially may undergo CMP in the future. For this, we will elaborately disc uss the impact of multistage annealing on the properties that may have implications for Cu processing. Also we will understand the eff ect of decreasing dielectric co nstant and its implication on the mechanical and tribological properties of low k materials. 4.2 Effect of Annealing on Copper Electroplated Cu films undergo a st ructural transition even at room temperature, which involves grain growth a nd texture changes [119, 120]. This so-called self-annealing process takes place over a peri od of hours or days afte r the deposition. Its kinetics depends on various deposition and postdeposition parameters such as electrolyte composition, current density, film thickne ss, seed layer composition, substrate morphology, annealing temperature, etc. [ 121-129]. This phenomenon, which leads to a dramatic drop of the resistivity over time is associated with evolution of the microstructure at room temperature. The se lf-annealing for EP Cu films has become a constraint for reliability and reproducibility of Cu interconnect process, because changes of grain size and hardness have influen ce on electromigration and CMP process [130, 131].
79 Table 4. Details of the Samples Undergoing Two Stage Annealing Sample # Oxide (nm) Barrier (nm) PVD Cu (nm) EP Cu (nm) Aging 1st Anneal 2nd Anneal 1 500 30 150 No No No No 2 500 30 150 500 7days in air 150 C, 60min 250 C, 60min 3 500 30 150 500 7days in air 200 C, 60min 250 C, 60min 4 500 30 150 500 7days in air 250 C, 60min No 5 500 30 150 500 No 150 C, 60min 250 C, 60min 6 500 30 150 500 No 200 C, 60min 250 C, 60min 7 500 30 150 500 No 250 C, 60min No For most reliability tests, knowledge of the thin film constitutive mechanical behavior is required. Mechanic al properties of thin films often differ from those of the bulk materials. Hence there is a need for eval uation of mechanical properties at nanoscale using state-of-the-art ev aluation techniques like Na noindentation [132, 133]. To
80 understand the impact of multi-stage anneal ing on the material properties of Cu, investigations have been focused on the Si/SiO2/TaN/SeedCu/EPCu thin film stack annealed at different stages with different temperatures a nd duration. The details of the evaluated samples can be seen Table 4.1. Grazing incident x-ray diffraction pa ttern showed stronger x-ray reflections form Cu (111) and (220) planes but weaker reflections from ( 200), (311) and (222) planes in all the electroplated Cu samples. This implies that all the Cu films have preferential crystallization al ong the Cu (111) and (220) plan es as compared to (200), (311) and (222) planes. The Cu samples that did not undergo multi-step annealing, had lower hardness and modulus values. Nanoindentation was performed on all the samples using the continuous stiffness measurement (C SM) technique. The elastic modulus varied from 121 to 132 GPa while the hardness va ried from 1 to 1.3 GPa depending on the annealing conditions. The surface mor phology and roughness of Cu films were characterized using atomic force microscopy. The tribological properties of the copper films were measured using the Bench Top CM P (chemical mechanical polishing) tester. Nanoindentation was performed on the samples after CMP and an increase in hardness and modulus was observed. This may be attribut ed to the work hardening of the Cu films during CMP. In addition, the hardness and modulus values increase w ith the increase in temperature of the first-step annealing. Samp les annealed in one stage (sample 4 and 7) show better mechanical properties, whic h may be due to higher (111) oriented crystallites. The Cu seed layer is very sm ooth and has lower COF and AE values during CMP. No distinctive trend is found for the CO F and AE values for annealed Cu samples.
81 Improved mechanical properties on post-CMP sa mples may indicate the work hardening of the Cu surface due to the high shear for ce applied on the sample surface during CMP. The surface roughness of annealed Cu samples in creased with the increase in temperature of the first-step annealing. Annealing at higher temperature may reveal further insight of` the microstructural evolution of electroplated Cu films, wh ich is the key in successful implementation of Cu as an interconnect mate rial. These results have been reported in literature and though these properties do have a bearing on the CMP performance of Cu, this chapter will focus on the interfacial a dhesion of Cu and TaN barrier layer which directly dictates the maximum amount of down force and platen rotation during CMP process. 4.3 Significance of Interfacial Reliability in Damascene Structures Fig. 4.1 Challenges Surroundi ng Implementation of Novel Low k Materials Cu in a Damascene Structure (Courtesy: Jeffery Lee, Intel Corporation)
82 Normally in copper damascene process, lowk materials are protected from exposure to CMP environment. In dual-damas cene Cu-interconnect system CMP is a two step process. In first step Cu is being polis hed with a copper selective slurry and barrier layer, which acts as a protecting layer of lowk system. In second step barrier layer is being removed with barrier selective slurry and a hard inorganic mask/dielectric capping layer is typically used to provide mechanical support and prevent in teraction between the slurry and the lowk materials. Fig. 4.1 outlines some of the significant challenges on the roadmap before successful implementation of damascene and dual damascene structures containing next generation ultra low k materi als. The challenge pertaining to the mechanical integrity of the interface of the cons tituent thin films and the effect of annealing on the same has been elabor ately discussed in this chapter. 4.4 Effect of Annealing on Cu-TaN interface To implement novel materials in th e MLM schemes for next generation interconnect systems, the inte rfacial reliability of materi al thin films needs to be extensively studied. In this direction, we ha ve evaluated the effect of annealing on the metal-barrier interface. To demonstrate the impact of multi-stage annealing on Cu-TaN interface, the aforem entioned stack (Si/SiO2/TaN/SeedCu/EPCu thin film) with different In this research, interfacial reliability of blanket sample s of Cu (interconnect wiring material) /TaN (barrier / adhesion layer) / Tetr aethylorthosilicate (TEO S) (dielectric) has been evaluated. Also, Cu has been annealed at different temperature and duration to study the effect of annealing on th e interface. Even though studi es have revealed that, the problem of delamination is more teething at the low kcap layer interface ; this
83 study aims at demonstrating a proof of concep t for improvement of in terfacial reliability of Cu-barrier interface by treatments such as annealing after self annealing of Cu is mostly complete . The details of the sa mples evaluated during this investigation are shown in Table 4.2. Table 4.2 Details of the Samples Subjected to Interfacial Studies 4.5 Techniques for Interfacial Evaluation of Thin Films Before proceeding to study the imp act of annealing on thin film interface, it is important to understand the techniques used to evaluate the interfacial adhesion and fracture toughness. During this study, the interfacial a dhesion was qualitative evaluated using the four point bend technique. The results of this technique were compared to qualitative evaluation of interfacial adhesion using a novel nanoscratch technique developed by us. Sample Designation Thickness Cu (seed + EP) Aging 1st Anneal (N2 atmosphere) 2nd Anneal (N2 atmosphere) 1 6500 7days in air150oC 60min 250o C, 60min 2 6500 7days in air200oC 60min 250o C, 60min 3 6500 7days in air250oC 60min No 4 11500 7days in air150oC 60min 250o C, 60min 5 11500 7days in air200oC 60min 250o C, 60min 6 11500 7days in air250oC 60min No
844.5.1 Four Point Bend Technique The four point bending techni que is commonly used to evaluate the thin film stacks in microelectronics applications. The technique involves extensive sample preparation and post failure analysis [ 135-138]. The sample preparation procedure includes dicing, curing and notching, which ma y induce permanent damage into soft and weak materials. Due to the extensive human el ement involved in the entire testing setup, there is a high possibility of erroneous results. In order to correlate the thin film interface performance with the CMP results, large statis tical set of samples need to be evaluated for accurate estimation of the interfacial adhesion energy of the thin films in a stack. Fig. 4.2 Schematic of Four Point Bend Test (University of Califor nia, Santa Barabara) Set Up [137, 138] In the test, two 60 mm X 10 mm specimens of with the aforementioned thin films on them were bonded to each other usi ng an epoxy resin. The samples were cut using a diamond coated dicing saw and epoxy was coated manually. The samples with the sandwiched epoxy were then cured for 2 hour s and later notched used a dicing up to a
85 depth approximately equal to one third of the thickness of the substrate .The samples were then gradually loaded in the conventi onal four point bend se t up, the details of which are already published [135-138].The uppe r substrate is notched up to roughly one third of its thickness. The sandwich specimen wa s gradually loaded in the set up which is schematically shown in Fig. 4.2. Fig. 4.3 Variation of Load (lbf) Vs Disp lacement (microns) during a Four Point Bend Test When the critical load is reached, the crack propagates through the substrates and kinks through the interface having the lowest interfacial adhesion energy. This can be clearly seen from the load Vs displacement cu rve obtained from the test as seen in Fig. 4.3. The interfacial adhesion energy (G) of the weakest inte rface is calculated using the critical load (P) of interfacial fracture from the equation 4.1 . 3 2 2 2 216 ) 1 ( 2 h Eb L P l G (4.1)
86 In equation 4.1, E is the YoungÂ’s modulus of Cu, is the PoissonÂ’s ration, l is the length of the entire specimen and L, b, h are shown in the Fig 4.2. The interfacial failure domain analysis was performed using the Elementa l Dispersion Spectroscopy (EDS). The EDS was performed on a 10 lt EDAX Dewar standard EDS detector couples with a Hitachi Scanning Electron Microscopy (SEM). EDS is used to detect the elemental composition of the surface exposed after failure of the sp ecimen due to crack propagation. The results from EDS analysis are thus helpful in de termining the weakest interface in the given specimen. 4.5.2 Nanoscratch Tests The nanoscratch technique has been de veloped by fabricating a modified suspension and set up on the CETR Bench T op CMP tester. This technique mostly involves a sharp diamond tip (in this case Be rkovich indenter tip), which directly processes the surface of material. The diamond tip may have different shapes such as cylindrical, conical, three or four sided pyramidal, rounded etc. The variation of the frictional forces induced is monitored in-situ or ex-situ to analyze th e failure of the thin film being evaluated . A standard thre e-sided Berkovich tip ha s been utilized on a modified setup incorporated in the CMP tester described before. An edge of the pyramid will cut through the material surface to finish a scratch test. The schematic of the proposed scratch testing setup is shown in Fig. 4.4. Th e tip has been will be mounted on a stainless steel holder with a custom de signed housing. The holder has been then fit into a CETR tester with a specialized su spension. A light-duty sensor setup has been used during testing as the magnitude of dow n force applied during nanoscrat ch testing is significantly
87 lower than that applied during CMP. The light duty suspension gives better resolution for monitoring and controlling the down force and frictional force. The variation of AE and COF signals upon film delamination will produ ce an in-situ qualitative metrology of interfacial adhesion. The load at which the onset of delamination is recorded has been termed as the Â“critical loadÂ”. The parametric optimization of nanos cratch testing will include the study of the effect of: 1) down loading, 2) scratc h tip velocity, 3) thin film thickness, 4) thin film m echanical properties (Hardne ss and YoungÂ’s Modulus) on the magnitude of critical load obt ained from the scratch test. The scratch tests were star ted by downward loading (Fz) of 5 gm on the tip and progressively increasing the load at the ra te of 2 gm /sec. The ultimate load used for each test was 105 gm. The linear velocity of the scratch tip was kept constant at 2 mm/ sec. The linear variation of down force, frictional force, lateral force and acoustic emission (AE) sensor signal with time were mo nitored in situ. When critical load of the thin film delamination was reached, the AE sensor signal showed a significant increase in magnitude while there were prominent changes in the gradient of lateral force and frictional force. It was thus concluded, that the area of instability of these aforementioned signals represented the delamination process of the thin film from the underlying film from the stack (in this case, delamination of Cu from TaN). The load at which the AE signal, frictional force (Ff) and lateral force (Fx) again deviate from their trend of variation initially followed during the test, was assumed to the critical load P for the thin film interface. When Ff, Fx, and AE regain stability, th e process of delamination is presumed to be complete.
88 FXFY Berkovich Micro Â–scratch TestingTip Multiple Layers of Thin Films on the substrate Substrate Interfacial delamination Direction of Motion FXFY Berkovich Micro Â–scratch TestingTip Multiple Layers of Thin Films on the substrate Substrate Interfacial delamination Direction of Motion Fig. 4.4 Schematic of the Nanoscratch Testing Setup Fig. 4.5 Raw Data Showing the Variation of a) Acoustic Emission (AE), d) Frictional Force (Ff) with Time during a Scratch Test. (Sam ple 1 at Linearly Increasing Downward Force of 2 gm/sec Linear Velocity of 0.2 mm/sec) 0102030405060 Time sec -0.5 0 0.5 1.0 1.5 2.0 2.5 3.0 AE Volt AE 0.1213 27 45 del 19 0102030405060 Time sec 0 5 10 15 20 25 30 Ff g Ff 20.2563 27 45 del 19 a b
89 The raw data out put and variation of the different parameters such as: 1) variation of AE signal with time, and 2) va riation of monitored dur ing the scratch tests can be seen in Fig 4.5 (a, b). The critical load at which the delamination occurs is a measure of the practical work of adhesion (W ) which can be evaluated using equation 3. In fig. 4.5 (a, b), it can be seen that delami nation started after 27 sec onds of loading at the downward 59 gm. The Cu thin film completely delaminated after 46 sec and the tip, then began its movement on Ta surface. 188.8.131.52 Failure Domain Analysis of Nanoscratch Tests The increase in AE near the end of sc ratch tests reveals that there is some phenomenon occurring which brings about a sign ificant dissipation of energy in form of an elastic wave in the ultrasound fre quency range. Though, we had observed delamination of thin film interface that corr esponded to the increase in AE signals during CMP, there was need conclusively prove that AE signal increase was due to delamination only and other phenomena such as crack initia tion and propagation due to loading, or buckling of the buried layer, or purely cohesi ve failure of the material was not causing the increase in AE signals. For this purpose, failure domain analysis of the nanoscratch tests was performed using SEM/Auger Spectroscopy and FIB/SEM analysis. 184.108.40.206.1 SEM/Auger Spectroscopy Analysis In order to conclusively prove that the increase in AE signal at the end of the scratch test is due to the dela mination of the thin film from the buried layer, the surface SEM of the nanoscratch was performed (Fig. 4.6 a, b ). Two points on the surface of the
90 nanoscratch were chosen as seen from Fig. 4.6. The first point (point 1 in Fig. 4.6 a) was chosen to evaluate the chemical compos ition of the surface before the significant penetration of the scratch tip beneath the film surface. Th e point 2 (Fig. 4.6 b) was chosen at a location where the delamination wa s complete (just after the elevation in the magnitude of the AE signal subsided). Auge r electron spectroscopy (AES) analysis was carried on point 1 and 2 using the as received samples. An Ar back sputtering step was performed and the variation of elemental peak intensity with the respective binding energy was re-plotted. Similarly Auger analys is was performed on poi nt 2 at the end of the scratch, before and after the Ar back sputtering. Table 4.3 shows the percentage elemental composition of points 1 and 2 be fore and after Ar back sputtering. Fig. 4.6 Surface SEM of the Nanoscratch at Region a) at the Initial Stages of Microscratch, and b) Upon Completion of Delamination a b 1 2
91 Table 4.3 Relative Concentration of the Elements (at.%) on Scratch Surface Element C (0.128) N (0.158) O (0.371) Cu (0.574) Ta (0.194) As-received Point 1 42 5 22 31 Point 2 30 8 37 3 22 After 12 s sputter Point 1 19 81 Point 2 15 22 9 14 40 It can be seen from Table 4.3 that when Auger analysis is done on the as received sample at point 1, there is an evidence of metallic Cu along with Carbon and oxygen on the surface. The presence of C c ould be traced on to surface organic contamination (in appropriate handling of the sample outside a clean room environment) which the presence oxygen is due to forma tion of passivating oxide of Cu. Upon back sputtering, the organic impurities are removed and presence of predominantly metallic copper with some oxide formation is seen. Th e Auger analysis performed at point 2 on the as received sample suggest s that there is significant redu ction in the metallic Cu on the surface. The surface is then comprised of predominantly Ta, which is consistent with the fact that TaN is the barrier layer used for Cu in the evaluated thin film stack. The presence of C again on the surface could be at tributed to the impuritie s on the surface. Ar back sputtering of the region denoted at point shows that the exte nt on Ta and N on the surface increases showing that the tip moves on the barrier layer af ter the completion of
92 delamination. The increase in the presence of Cu on the surface could be attributed to the re-deposition of Cu from the surrounding debris on the scratched surface. 220.127.116.11.2 FIB Cross section and SEM Analysis Fig. 4.7 FIB Cross-section and SEM Micrograph of the Thin Film Obtained a Region of Onset of Delamination The Auger analysis proved that the sc ratch tip initially caused cohesive failure in Cu thin film by penetrating in it and th en after delamination, the tip moved on the barrier thin film surface. However, in order to conclusive prove that the increase in AE signal did not result from phenomena such as buckling of the burie d TaN or TEOS film, Pt Layer Cu film TaN TEOS layer Onset of Delamination Si Substrate
93 cross section FIB and SEM analysis was perf ormed at the Cu-TaN interface at a region where delamination began occurring. Fig. 4.7 shows the SEM of the FIB cross section of the thin film stack obtained at the position at which there is on set of delamination. It can be seen from Fig. 4.7 that th ere are no obvious signs of buckling of the buried TEOS layer. Also there does not seem a ny crack initiation of pr opagation in any of the thin film layers below Cu. It can be thus concluded that the elevation in the magnitude of the AE signal is du e to the peeling of the Cu thin film from TaN interface, as shown in Fig. 4.7. There also a reorga nization of the Cu thin film grains due to external crack initiation and propagation under down loading and lateral velocity (nanoscratching), however, this subj ect is a separate investiga tion in itself and is beyond the scope of this research. 4.6 Quantitative Evaluation of Thin Film Adhesion energy Fig. 4.8 Variation of Interfacial Adhesion En ergy for the Different Samples Evaluated by Four Point Bend Technique 0 5 10 15 20 25 30 35 40 45 Adhesion Energy J/m2 S1S2S3S4S5S6 Sample
94 The results of the interfacial adhesion energy (G J/m2) evaluated using the four point technique can be seen in Fig. 4. 8. The sample dimensions were measured and substituted in equation 4.2 (ref. Fig. 4.2) for calculation of the adhesion energy for each of the sample. The average of 4 readings is shown as the adhesion energy for that particular sample in Fig 4.2. It can be seen from Fig. 4.2 that the interfacial adhesion energy does not show correlation with the Cu th in film thickness as the value of G in J/m2 for sample 16 is seen smaller than sample 22 but the values of G for sample 21 is greater than sample 15. Figure 4.8 also shows that th ere is no significant di fference in interfacial adhesion energy of samples that have been annealed at 150oC for 60min and later at 250o C for 60min, and samples which were annealed at 250o C for 60min and did not have a second annealing step. This held true for sa mples irrespective of their Cu thin film thickness. However, it can be seen that samp les 2 and 5 showed markedly higher values of interfacial adhesion energy. These samples were annealed first at 200o C for 60min and the second stage annealing was done at 250o C for 60min. It can be thus concluded that duration of annealing along with temper ature affected the interface. Annealing at a higher temperature as well as extended duration (200o C & 60min stage I and 250o C, 60min stage II), can bring about significant improvement in the interfacial adhesion. Even though optical evaluation of failed inte rface was sufficient to determine that the layer exposed was TaN failure domain analys is was done using EDS to reinforce this conclusion. The sample was subjected to ED S analysis and an elemental composition of the sample constituents was obtained. Alt hough, due to the nature of the technique, constituent elements of all the thin films as well as substrate were obtained as results of EDS analysis, Fig. 4.9 shows the area which s howed distinct presence of tantalum which
95 has been isolated for its distinct importan ce for the purpose of study. Apart from Ta, which existed in form of three different states (this goes on to show that Ta was combined with another element to form a comp ound in the thin film), elemental peaks of Si, O, C (impurities) and N were also observed (not seen in the Fig. 4.8). The element Cu was conspicuous by its absence. Taking all thes e factors in to cons ideration, it can be safely said that that thin film stack failed at the Cu-TaN interface. Fig 4.9 Elemental Dispersion Spectroscopy An alysis of the Failed Interface after Four Point Bend Analysis
964.7 Qualitative Evaluation of Thin Film Adhesion energy The interfacial adhesion was qualitativel y evaluated using nanoscratch testing technique. In sharp contrast to the four point bending techniqu es discussed earlier, nanoscratch testing technique is fast, does not have excessi ve sample preparation time and has less chances of human error that might occur during bonding and notching the specimen before four point bend test. The va riation of critical load for the candidate samples has been shown in Fig. 4.10. When th e samples of a given Cu thin film thickness were compared, the critical load showed a hi gher value for sample that was annealed for 200o C & 60min in stage I and 250o C, 60min in stage II. There is a difference in the magnitude of the critical loading for samp le with 650 nm Cu thickness and 1150 nm of Cu thickness could be attributed to the hi gher work done by the scratch tip to overcome the cohesive forces of a thicker film. The results of the scratch tests on the candidate samples also qualitatively showed the sample that were annealed for a relatively higher temperature and higher duration had an impr oved interfacial adhe sion between Cu and TaN as compared to the other two samples. This is in agreement with quantitative evaluation of interfacial adhesion pe rformed using four point bend test. Fig. 4.10 Variation of Critical Load for Di fferent Evaluated Samples obtained from the Nanoscratch Test 0 20 40 60 80 100 120 Critical Load (gm) S1S2S3S4S5S6 Sample
974.8 Low Dielectric Constant Ma terials (Low k Materials) After evaluation of the Cu -TaN interface, the focu s of study of material properties of potential materials undergoi ng CMP shifts to some candidate low k materials. CVD and spin-on methods ar e main deposition techniques of present generation lowk materials. Several spin-on lowk materials, like hydrogen silsesquioxane (HSQ), spin-on-glass (SOG), SiLK (trade ma rk of Dow Chemical Company), etc. have been studied . Lowk dielectric materials can be categorized as follows: doped oxides (FSG, HSQ, MSQ and HOSP), organics (BCB, SiLK, FLARE and PAE-2), highly fluorinated materials (paryleneAF4, a-CF a nd PTFE), and porous materials (aerogel and xerogel) . In some cases, combinations of these materials (for example, porous organics) are also being explored . Silicon dioxide (SiO2) has a dielectric constant of 4, while air is considered as the perfect insulator with a dielectric constant of 1. Porous materials can therefore achieve lowe r dielectric constants than the constituent materials . Among many lowk candidates, however, onl y a few materials have shown all the required properties needed for integration into high-volume manufacturing processes. Finding polymers with lowk value is a relatively easy task; but finding those with all the required chemical, mechanical, electrical and ther mal properties for use in IC applications is more difficult. Proponents of CVD approaches, most notably carbondoped siloxanes also known as orga no-silicate glasses (OSGs) with k ranging below 2.5, claim the advantage of being able to reuse of existing tool sets and simpler integration due to the SiO2-like structure of CVD siloxanes [ 142]. Alternatively, manufacturers of spin-on materials and spin-on equipments c ontend the better extendibility of future
98 generationsÂ’ lowk dielectrics, especially in the sub-2.5 range when porous lowk materials will likely be used. Toda y, porous versions of many lowk spin-on materials are available for testing, whereas porous CVD lowk materials have yet to be demonstrated . Most spin-on materials are organic polymers, some are inorganic and some are blends of the two. Their k values vary significantly from each other, depending on the material. However, in general the values ar e below 3.0. The semiconductor industry first broached the subject of new dielectrics in 1998 with the 0. 18-m circuitry that is now standard in high-end semiconductor chips like the Pentium 4. At this production node, manufacturers found a relatively easy answer for lowk material in fluorinated silicate glass (FSG), a CVD material created by doping traditional SiO2 with fluorine from silicon tetra fluoride dur ing the deposition . 4.8.1 CVD Based Low k Materials Relative dielectric consta nt for conventional SiO2 films is in the range of 3.9Â–4.6 at the frequency of 1 MHz. The values de pend on the source materials and formation technique. Basically the dielectric constant is defined by the dielectr ic polarization, such as electronic polarization, ionic polarizati on and orientational pol arization. Since the electronic polarization is basically defined by atomic radius, the molecular structure differences are essential. The dielectric c onstant is also increa sed by residual HÂ–OH/or HÂ–OH absorption, due to orientational polariz ation increment. One of the techniques to reduce the electronic polarization is to introduce an atom with a small atomic radius in the SiÂ–O networks. Hydrogen (H), carbon (C) and fluorine (F) atoms are very effective
99 for this purpose. The binding energy of SiÂ–F bon d (129.3 kcal/mol) is higher than that of SiÂ–O bond (88.2 kcal/mol), SiÂ–F bond is therma lly more stable. Therefore fluorination of SiO2 films has been one of the main methods ex amined to reduce the dielectric constant of SiO2 films . Incorporation of fluorine in SiÂ–O network causes a reduction of dielectric constant as a re sult of fluorine being the mo st electronegative and least polarizable element in the periodic table . Additionally, precu rsors for depositing SiOF films are readily available and are inexpensive. Several companies are developing lowk CVD films using a variety of carboncontaining precursors. The resulting organos ilicate glass (OSG) f ilms are also called carbon-doped silicon oxides (SiOC). The organi c groups in OSGs invariably take the form of tetravalent silicon with a wide range of alkyl and alkoxy substitutions. In these films, the siliconÂ–oxygen network seen in glass is occasionally interrupted, in a more or less homogeneous fashion, by the presence of organic functional groups, typically methyl (Â–CH3) groups. Additionally, hydride (Â–H) substitution at silicon can also be present in the network. The film's lower k results due to these changes to the SiO2 network and the reduced density of the OSG film relative to SiO2. In typical CVD lowk films, 10Â–25% of the silicon atoms are substituted with organi c groups. In amorphous OSG structure, the ratio of Si and carbon atom is typically 4:1 . 4.8.2 Spin On Low k Materials The spin on deposition technique is particularly suited when good local planarization and gap fill is required for inter-metal dielectr ic. Both inorganic and organic films can be deposited by spin on methods and final structure of the film could be
100 amorphous or crystalline. The di electric precursors are used in forma of Â“solÂ”. The thin film coating is performed by dispensing a liquid precursor on the center of the substrate which is spun on a spinner. The rotation of the spinner causes uniform dispersion of the solution. The resultant thin film is hard ba ked or soft baked and then sintered at an elevated temperature (300Â–400 C). The sinter ing process ensures fi nal cross-linking of polymer chains and results in mechanically stable thin film. Amorphous dielectric materials such as spin on glasses are co ated using this method. A combination of hydrolysis in which there is a transition fr om SiÂ–OR to SiÂ–R functional groups takes place in presence of moisture and condensati on process in which SiÂ–OÂ–Si structure is formed with elimination of H2O to form the spin deposited film. The viscosity of the films increases sharply after spinning there by helping the film to settle on the substrate. Subsequently the gel is dried by baking or curing step . In order to further re duce the dielectric constant of the ILD, porous films are made using spin on technique. The formation of more or less rigid skeleton before water extraction is important for formation of hi gh porosity materials. The porosity can be induced in the material during the sol gel pr ocess or can be formed by using sacrificial nanoparticles or porogens. The details of various techniques to fabricate lowk dielectrics and methods to induce porosity have been extensively discussed in literature. 4.8.3 Potentially Feasible Low k Materials Various oxide-based lowk materials were deposited using CVD as well as spin-on methods. Preparation technique and pr ecursor gases for different types of doped and undoped oxides studied are briefly discussed in this section. Undoped silicon dioxide
101 (SiO2 U) films were deposited on 8 in. Si wa fer using PECVD met hod in a six-station sequential deposition system. The pr ecursor gases were silane (SiH4), nitrous oxide (N2O) and nitrogen (N2) and the substrate temperatur e was maintained at 400 C. SiOF is a fluorine-doped silicon dioxide film. It was deposited using inductively-coupled high-density plasma ch emical vapor deposition (HDP CVD) method with SiF4, SiH4, O2 and Ar at 400 C. Details of the deposition method have been discussed earlier . SiOF films grown by HDP CVD method has been shown to result in a film that has excellent film quality and gap fill characteristics [146, 147]. The dielectric film SiOC SP is a carbon-doped silicon dioxide, also known as carbon-doped siloxane, or organosil icate glass (OSG). This lowk film was grown on Si substrate using PECVD method at 400 C in a six-station sequentia l deposition system. The precursor used in this process is a cy clic 1,3,5,7-tetramethylcyclotetrasiloxane (TMCTS). TMCTS is prepared through hydrolysis of methyldichlorosila ne to firstly form a linear siloxane polymer that is end capped with trimethylsilyl groups. Another carbon-doped lowk film (SiOC NSP) was deposited using a different and non-standard precursor tetramethylsilane gas, with N2O and N2 using PECVD method at 400 C in a six-station sequential de position system. It is similar to the film SiOC SP. In order to investigate th e difference in CVD deposited lowk with the spin-on lowk we have also used a carbon-doped oxide-based lowk (SiOC SO), deposited by spin-on method. SiOC SO is siloxane polym er-based material and is an organic and inorganic hybrid. As this dielectric fi lm was deposited by a spin-on method, it has flowable and planarizing characteristics.
102 Silica aerogel/xerogel which is known as nanoporous silica, has numerous properties which suggest applications such as low dielectric constant (1.1 < K < 2.5) materials for inter-level dielectrics for the next generation. The advantages of these materials, in addition to the low dielectric constant, include high temperature stability, pores much smaller than mi croelectronic feature sizes, deposition using conventional spin-on and vapor deposition methods, and precurs ors similar to those currently used in the microelectronic industry. SiLK is a ne w polymer from the Dow Chemical Company which does not contain silicon. A valuable f eature of SiLK is its high thermal stability. SiLK films are prepared by spin-coating of di ssolved initial products in organic solvents and curing . Fig. 4.11 Simplified Classificati on of Low k Dielectrics  As the dielectric constant decreases the mechanical properties of the materials deteriorate. This is undesirable for interconn ect materials. A material with optimized
103 mechanical strength is very much required for successful chemical mechanical polishing of these interlayer dielectric materials. Thus to ensure a low-cost packaging solution for lowk devices, issues regarding mechanical prop erties must be understood and considered as carefully as electrical properties [149, 150]. In the case of porous materials of dielectric constant 2, porosity near 65% are utilized These materials are so weak mechanically that they may not be robus t enough for the CMP process. Simplified classification of the lowk materials can be seen in Fig. 4.11. 4.9 Mechanical Characterization of Low k Materials To evaluate the mechanical properties nano-indentation over a small scale has been used extensively in recent years [151-155] This is a depth sensing indentation at low loads and is a well-established tec hnique for the investigation of localized mechanical behavior of materials. The displ acement and load resolution can be as low as 0.02 nm and 50 nN, respectively. A typical load versus displacement curve showing contact depth ( hc), and maximum depth ( ht) after unloading is shown in Fig. 4.12. Hardness and Young's modulus of elasticity are derive d from the experimental indentation data by an analytical method usi ng a number of simplifi cations . Contact depth hc can be calculated by: (4.2) where ht is maximum depth of penetration incl uding elastic deformation of the surface under load, F is the maximum force, and = 0.75 is a geometrical constant associated with the shape of the Ber kovich indenter . Once hc is determined, the projected area
104 A of actual contact can be calculated from the cross-sectional shape of the indenter along its length. S is the stiffness, which can be deri ved experimentally from the following equation: (4.3) where Er is the reduced modulus. Hardness is then calculated from the simple relation: (4.4) The reduced modulus Er is normally defined as: (4.5) Details of the samples al ong with their refractive index, k -values, and density information are shown in Table 4.4. Mechani cal properties of the films were measured using nanoindentation technique using Nano Indenter XP (MTS System Corporation, Oak Ridge, TN). A three-sided Berkovich-shape d diamond indenter is used to indent on the material surface. The load and displacemen t data obtained in the nanoindentation tests were analyzed according to the method of Oliver and Pharr [151, 152]. The continuous stiffness measurement (CSM) technique was used for measuring absolute and depth dependent hardness and modulus values. Valu es were calculated by averaging a number of separate indentations at particular depth specifications Initially the instrument was calibrated with the standard sample (fuse d silica) provided by MTS and other single crystal metal samples.
105 Fig. 4.12 Typical Curve Showing the Loading and Unloading as a Function of Indenter Penetration Depth Table 4.4 Details of the Differe nt Low k Materials Evaluated # Sample Thickness () Grown by Refractive Index 1 SiO2 (U) 3850 PECVD 1.474 ~ 4.0 2 SiOF 1700 HDPCVD 1.430 ~3.7 3 Polyimide 16000 Spin-On 1.920 ~2.9 4 BCB 17000 Spin-On 1.542 ~2.6 The mechanical properties of the candidate samples were evaluated by nanoindentation. The depth of penetration fo r the indenter was fixed at ~50 % of the sample film thickness. The calculation of mechanical properties was performed at 50 % of the indentation depth (i.e. ~25 % of the f ilm thickness). This depth of calculation of the mechanical properties chosen to avoid the su bstrate effects as well as the effect of the surface oxide or other complex that might be pr esent due to the reacti on of the film with
106 ambient air. The values of the mechanical properties i.e. hardness and YoungÂ’s modulus have been tabulated in Table 4.5. The normaliz ed hardness of the dielectric materials in comparison with the hardness of SiO2 is shown in Fig.4.13. It can be seen from Table 4.4 and Fig. 4.13 that there is almost an e xponential decrease in the hardness and YoungÂ’s Modulus of the material with a decrease in the dielectric constant. These reduced mechanical properties adversely impact the ability of the material to withstand down pressure and shear during CMP. The material removal during three body contact abrasion mechanism (most likely) during CMP takes plac e when the particles abrade the surface layer complex formed by the chem ically active slurry . Table 4.5 Results of Nanoindent ation of the ILD materials # Sample Thickness (nm) Indentation Depth (nm) Hardness (Gpa) YoungÂ’s Modulus (Gpa) 1 SiO2 (U) 385 200 6.31.2 68.11.2 2 SiOF 170 100 4.30.7 42.43.4 3 Polyimide 1563 750 0.335.0.03 18.104.22.168 4 BCB 1665 800 0.290.08 3.50.4 The particles suspended in the slurry ar e significantly harder than all the ILD materials except SiO2. In the event of agglomeration of the slurry particle s or due to any other reasons, if the depth of particle indentation is greater than the thickness of the surface layer formed in the slurry, then a materials with lower hardness is more susceptible to deep surface scratches. Usually, a buffing step accompanies every CMP
107 polishing run. However, in the event of a d eep surface microscratch, it is difficult for the buffing step to produce a mirror flat surface after CMP. SiO2 (k~ 4.0) SiOF (k~3.7) Polyimide (k~2.9) BCB (k~2.6) Air (k=1) 0 0.2 0.4 0.6 0.8 1Normalized Hardness 22.214.171.124.61Dielectric Constant Fig. 4.13 Variation of the Normalized Hardness with Dielectric Consta nt of the Candidate ILD Materials 4.10 CMP of Low k Materials Before we discuss the result of the CMP of low k materials, it is necessary examine the theory behind the material rem oval of the candidate dielectric materials. Material removal mechanisms for: 1) ceramics (doped and undoped SiO2) and 2) polymeric (Polyimide and BCB) have been di scussed in this section. The material removal mechanism of the ILD is strictly de pendant on the chemistry of the slurry used. However, for the scope of this study, the ge neric mechanism of removal in alkaline ILD polishing slurry with colloidal silica pa rticles as abrasives has been discussed.
108 4.10.1 Doped and Undoped Cera mic Material Removal SiO2 (or SiOF) + H2O Si (OH) 4 (4.6) Fig 4.14. Material Removal Sc hematic during Ceramic ILD CMP The fumed or colloidal silica abrasives in the slurry and the ceramic dielectric surface are of comparable hardness. Duri ng polishing of dielectrics such as SiO2 or SiOF, the alkaline slurry hydrolyzes the few surface at omic layers into relatively softer Si(OH)4. This material can then be easily removed with the help of slurry abrasive particles in pressure of the polishing pad for given operating CMP conditions . The chemical reaction (unbalanced) indicative of the mate rial surface transformation is shown in equation 2, and schematic of the material removal is shown in Fig. 4.14.
1094.10.2 Doped and Undoped Cera mic Material Removal Fig. 4.15 Schematic showing the Action of ILD Slurry on the Polymer Surface during CMP When the polymeric ILD materials are polished using alkaline slurry with silica abrasive particles, the slurry can e ither 1) effectively react with the surface producing a thicker and weaker polymeric la yer there by yielding hi gh removal rates or 2) passivate the surface and produce low rem oval rates primarily due to mechanical abrasion. The reaction at the surface and pr operties of the surface layer are dependent upon the physical characteristics of the polym er. For the high removal rate polymer CMP process, the material removal mechanism can be divided in to the following parts: 1)
110 adsorption of the slurry on the surface of the polymer if the material is hydrophilic, 2) diffusion of the slurry in the polymer surface, 3) reaction of the alkaline slurry with the polymer surface to create a significantly so fter and weaker surface or breaking of the polymer bonds by the slurry diffused in the polymer thereby creating a thick surface polymer complex, 4) removal of the surface laye r with abrasives in the slurry in presence of the polishing pad and 5) dissolution of the removed material . The schematic of the slurry attack on the polymer su rface is shown in Fig. 4.15. 4.10.3 CMP Process Conditions The experiments for this study were pe rformed using: 1) commercial IC 1000 Suba IV polyurethane pad manufactured by R odel, Inc, 2) Rodel, Inc. Klebesol 1501 commercial polishing slurry. The pad was subj ected to break in for 20 minutes before each set of material polishing runs with flowing DI water using TBW grid-abrade diamond pad conditioners. A conditioning run lasting 20 sec was performed using DI water between every two CMP runs. The experi ments were performed at 1, 3, and 6 PSI down pressure. At each value of down pressu re, CMP runs were performed at three different values of platen ro tation: 1) 42.4 RPM, 2) 148.5 RPM and 3) 254.6 RPM. The values of platen rotation corresponded to an av erage linear velocity of : 1) 0.2 m/s. 2) 0.8 m/s, and 3) 1.2 m/s. This study being comp arative, the upper carri er rotation action of CMP tester was not used. The movement of th e polished sample was restricted to slider oscillation. The details of the experiment s have been tabulated in Table 4.6.
111 Table 4.6: ILD Material CMP Experimental Details # Parameter Conditions 1 Down Pressure Variable (1-6 PSI) 2 Platen Rotation Variable (42. 4-254.6 RPM) or (0.2 m/s-1.2 m/s) 3 Slider Position and Movement 45mm 5mm @ velocity of 5 mm/ sec 4 Slurry Oxide Slurry Rodel, Inc. Klebesol 1501 (pH~10.5) with colloi dal silica abrasive particles. 5 Pad Rodel, Inc. IC1000 S uba IV A4 perforated with specific gravity ~745 6 Time of polishing 120 sec ( Values of AE and COF noted in the first 30 seconds have been plotted) 7 Polishing Specimen 1Â” X 1Â” coupon The process parameters specifically st udied were: 1) variation of COF with time and 2) variation of material removal rate (MRR) for different sets of polishing conditions, for each material. The removal of the material was calculated as shown in equation 4.6. The thickness of the thin film was measured using He-Ne Laser Ellipsometry. Time Thickness Thickness MRRFinal Initial (4.6)
1124.10.4 Tribological Properties of Low k Materials A total of 9 CMP experiments were performed on each of the candidate ILD samples at the pressure and linear velocity conditions mentioned in the experimental section of the paper. The variation of COF for the entire duration of the CMP run (120 sec) was noted. For the purpose of compara tive ILD study, the variation of COF for the first 30 seconds was considered. The variati on of AE signal was also noted in situ. However, off hand there was no distinct tre nd in AE signals. To explore the subtle phenomena occurring during each polishing run, the AE signals need to be filtered using the wavelet based approach  This study does not fall within the scope of this current research. The variation of COF with time for th e different conditions of down pressure and velocity for all the samples are shown in Fi g. 4.16 a-d. It can be seen from Fig. 4.16 that the magnitude of COF (COF= N sF F ; where Fs is the shear force and FN is the normal force during the CMP run) 40 is strongly dependent on 1) the down pressure, 2) platen velocity and 3) material that is being polis hed. Every material for given conditions of pressure and velocity shows a unique value of COF. This principle is used for effective end point detection (EPD) of the CMP proce ss. The change in the value of COF upon complete material removal depends upon the tr ibological properties of the buried layers. This transition can be attributed to the complete material removal. Details of the EPD using COF and AE have already been reporte d . The end point obtained using in-situ data is not sharp. Further signal filtering is currently being carried out for accurate estimation.
113 SiO2 (U)0 0.1 0.2 0.3 0.4 0.5 136 Down pressure (PSI)COF 0.2m/s 0.8m/s 1.2m/s SiOF0 0.1 0.2 0.3 0.4 0.5 0.6 136Down pressure (PSI)COF 0.2m/s 0.8m/s 1.2m/s Polyimide0 0.2 0.4 0.6 0.8 136 Down Pressure PSICOF 0.2 m/s 0.8m/s 1.2m/s BCB 0 0.2 0.4 0.6 0.8 136 Down Pressure PSICOF 0.2m/s 0.8m/s 1.2m/s Fig. 4.16 Variation of COF with Down Pressu re (PSI) and Linear Velocity (m/s) for a) SiO2U, b) SiOF, c)Polyimide, and d) BCB Normally, the values of COF should d ecrease with the increase in platen velocity. This is due to the fact that fricti on is generated at the interface of the pad and wafer is not only due to the in teraction of the high points on the wafer and the slurry but also due to the interaction of ceramic slurry abrasive particles with the pad and wafer. With the increase in their speed due to high pl aten velocity, the depth of indentation of the particles on the surface of the pad decreases. Due to the partial elastic behavior, at shallow depth of indentation, the response of the pad is bris k and there is lesser drag to the particle motion on the surf ace of the polishing pad. He nce, there is lesser dynamic friction at the pad-wafer inte rface . Previous experiment s carried out with carrier a b c d
114 rotation have shown the verifi cation of this phenomenon . However, anomalies can be seen in some cases (ref Fig. 4.16 a-d) due to lack of upper rotation of the sample coupon. The lack of sample rotation causes non uniform material removal. More material is removed from the leading edge as compared to the lagging edge. The variable thickness on the surface of the sample contri butes to the variation of the COF value during the polishing run. Fig 4.16 (a-d) shows an overall trend of decrease in the valu e of COF with an increase in down force. These results are in agreement with those obtained previously on this set up . It must be noted here that the increase in down force brings about a corresponding increase in the shear force (Fs) as shown in equation 4.7 . tan H W Fs (4.7) where W is the downward loading, is the maximum shear stre ss on the given area, H is the hardness of the polishing pad and tan is the ratio of the loss modulus to the storage modulus (which is also a measure of the toughness) of the pad. The applied down force Fz on a viscoelastic polishing pad can be written as shown in equation 4.8: N v zF F F (4.8) where Fv is the force dissipated for pad deform ation due to the viscous nature and FN is the normal force due to the el asticity of the polishing pad. Due to non linearity of the strain displacement reactions in the viscoela stic materials like polyurethane, second order geometric non linearity is observed in the pol ishing pads upon applica tion of down force. This geometric non linear expansion of the pad produces an axial normal tensile force in the polishing pad. The second order normal force (FSN) for a rotating pad with radius r,
115 modulus of rigidity G and uniform angle of twist is expressed as shown in the equation 4.9 . 4 24 1 r G FSN (4.9) The term 4 2r G is nothing but the shear force (Fs) acting on the surface of the pad under the given loading. Thus, the second order norma l force can be expressed in terms of the shear force as shown in equation 4.10. s SNF F4 1 (4.10) Hence, the COF during a polishing run can be ex pressed as a function of elastic recovery and shear force as shown in equation 4.11 s ER sF F F COF4 1 (4.11) where FER is the first order elastic normal r eaction (recovery) component of the normal force. It can be concluded from equation 4. 11 that the increase in shear force with the increase in the down force, brings about a corresponding incr ease in the normal force due to the increase in the shear component and nor mal elastic reaction of the pad. This can explain an overall decrease in the values of COF with the increase in the down pressure. 4.10.5 Variation of Material Removal Rate for Low k Materials The variation of MRR of all the ILD samples under variable pressure and linear velocity conditions is shown in Fig. 4. 17 a-d. An average of thin film thickness readings at the leading edge was used for MRR calculation due to more material removal
116 near it. It can be seen from the Fig. 4.17 that the candidate ILD samples follow the PrestonÂ’s equation. Material removal is str ongly dependent on solution chemistry of the slurry as well as the CMP process conditions To achieve highest defect free MRR, the material passivation or surf ace reaction at the pad wafer interface should more or less match the abrasion of material due to the process conditions. If the material removal is lesser than the surface chemical complex form ation, there is a decrease in the removal rate. In case of the lesser material passivation, partial material removal takes place due to mechanical abrasion. This increases the chan ces of the defects on the surface of the wafer and also reduces the removal rate. It can be seen from Fig. 4.17 that th e MRR for the same process conditions is higher for SiOF as compared to SiO2. The higher PrestonÂ’s coefficient, which is indicative of the dependence of the MRR on th e pressure and velocity conditions during CMP for a given material, is higher for SiOF (1.903) as compared to SiO2 (U) (1.468). Fig. 4.17 c, d shows that the variation of MRR of the polymers (Polyimide and BCB) does not follow the PrestonÂ’s equation accurate ly. Materials in the past, especially ILD, have exhibited two PrestonÂ’s coefficients [ 163, 164]. Further polishi ng investigations are needed at low values of down pressure and li near velocity to ascer tain whether evaluated polymers display this phenomenon of havi ng two different values of PrestonÂ’s coefficients governing their MRR with process conditions.
117 SiO2 U MRR = 1.468PV 0 500 1000 1500 2000 2500 0500100015002000 PSI*RPMMRR (A/min) SiOF MRR = 1.903PV 0 1000 2000 3000 4000 0500100015002000 PSI*RPMMRR (A/min) Polyimide y = 0.3157x + 263.04 0 200 400 600 05001000 PSI*RPMMRR A/mi n BCB MRR = 0.1269PV + 190.77 0 100 200 300 400 05001000 PSI*RPMMRR A/mi n Fig.4.17 Variation of MRR with Pressure (P) and Platen RPM (Linear Velocity V) for a) SiO2 (U), b) SiOF, c) Polyimide, d) BCB a b c d
118 The MRR for SiO2 (U) and SiOF (ceramics) is significantly greater than that of the polymers. The rate at which the surface is softened and directly attacked is more for the ceramics as the slurry is tailor made and commercialized for SiO2. The chemical attack of the slurry on the polymers is relatively subdued due to their chemical composition. Hence, even though the extent of mechanical forces contributing to the material removal is almost the same, MRR is more for SiO2 and SiOF. Both BCB and Polyimide have the tendency to uptake mois ture. BCB moisture uptake (after being fully cured), however, is much less than Polyimide after cure. BCB moisture weight uptake or absorption is < 0.2%, while Polyimide moisture uptake is about 1-2%. Though this data was obtained in air at around 80% relative hum idity, similar behavior is expected to persist when the polymers are in contact with the slurry. Thus, when Polyimide and BCB wafers are subjected to CMP, there may be slightly more Â“bond breakingÂ” due to higher moisture uptake in Polyimide than BCB in addition to the slurry chemical attack. 4.10.6 Surface Characterization of Low k Materials As a part or the comparative study, pre and post CMP AFM was performed on ILD samples that underwent CMP at 3 PSI down force and 148.5 RPM (0.8 m/s linear) platen rotation. The surface morphology a nd roughness before and after CMP were evaluated. The change in Rrms due to CMP for each of the sample at the aforementioned polishing conditions is shown in Table 4.7. The AFM images for the sample before and after CMP can be seen in Fig. 4.18 a-d. It can be seen from Table 4.6 and Fig. 4.18a that the surface roughness of SiO2 (U) decreases significantly upon polishing. As the slurry has been commercialized
119 for SiO2, this factor must have been taken in to account in the slurry design. The sample SiOF shows a very smooth surface before CMP. However, there is an increase in surface roughness after CMP. This may be due to the fact that the surface complex formed during CMP of SiO2 (U) and SiOF may be similar or even the same. Hence after planarization, the ILD surface may display similar propert ies including surface r oughness. Further XPS analysis is however needed to confirm the occurrence of this phenomenon. Also, due to the lesser hardness of the SiOF as compared to SiO2, there might be slightly greater indentation of the ILD surf ace leading to a slightly rougher surface. However, the threshold of surface scratch does not appear to be reached in case of SiOF, despite its lower hardness. Both the polymeric ILD materials show a very smooth surface before CMP. However, the surface roughness of both Polyimide and BCB increases very significantly after CMP (Polyimide has a rougher surface as compared to BCB). Due to the low hardness of the polymers films, numerous scratches caused by slurry abrasives or agglomerates can be seen on the surface of the both the polymers after CMP. There may formation of a passivated surface complex and lack of polymer weakening in presence of the slurry. Thus, the material removal may be primarily due to mechanical abrasion. This explains the significantly lower material re moval of the polymers as compared to the ceramics. The scratches left on the surf ace of Polyimide are far numerous when compared with those occurring on BCB, in spite of the fact that hardness of the Polyimide is slightly greater than BCB. As th e water in the slurry attacks Polyimide more than BCB, the surface layer formed due to the reaction of the Polyimide with the slurry may be softer due to relatively higher wate r content or absorption in comparison with
120 BCB. This explains almost two and half times higher post CMP roughness, more surface scratches and slightly higher removal rate s hown by Polyimide. Details investigation of the wear and removal mechanism of these polym ers needs to be carried out in the future. Table 4.7 Surface Roughness (Rrms) of the ILD samples before and after CMP # Sample Rrms Before CMP (nm) Rrms After CMP (nm) 1 SiO2 (U) 3.42 0.37 2 SiOF 0.18 0.40 3 Polyimide 0.529 8.326 4 BCB 0.657 3.658 4.10.7 Findings of Low k Materials Evaluation There is a decrease in the mechanical properties of the materials with the decrease in their dielectric constant. Th e COF of the material is an individual characteristic which changes with a change in the CMP input parameters. This can be used for accurate EPD, MRR estimation, qualitative estimation of the selectivity of the slurry towards the material etc. There is an overall decrease in the COF with increase in down pressure. However, there is an increase in the removal rate with the increase in pressure. This is due to the fact the norm al force exerted by the pad is also partly comprised of the shear force experienced by the material due to the viscoelastic nature of the pad.
121 Pre-CMP Post-CMP Fig. 4.18 Surface Morphology of the Candidate ILD Samples Before and After CMP for: a) SiO2 (U), b) SiOF, c) Polyimide and d) BCB SiO2 U (a) SiOF (b) Polyimide (c) BCB (d)
122 As there is an increase in the magnitude of the shear force with the increase in down pressure, MRR increases with pressure even though the value of COF decreases. There should be a decrease in COF with in crease in linear velocity. The anomalies observed in the study are due to the non-uniform material removal due to the absence of upper rotation. The sample SiO2 (U) and SiOF follow the Pr estonÂ’s equation accurately while the polymeric samples do not. The possibility of the polymeric samples showing two PrestonÂ’s coefficient needs to be invest igated by carrying out experiments at lower values of pressure and velocity. There is a sizable increase in the surface roughness of the polymeric materials after CMP. Surface of the polymeric materials shows the existence of scratches after CMP which are absent in case of the ceramic samples materials. There is weakening and softening of the polymer in presence of the slu rry but the passivating layer formation and removal is not as effective as demonstrated by the ceramic samples. This explains the relatively higher material removal rate shown by the ceramics. The lower removal rates, high surface roughness, ex istence of microscratches coupled with lower hardness are definitely some of the inte gration challenges that need to be overcome before successful integrati on of polymeric materials.
123 CHAPTER FIVE INVESTIGATION OF NON UNIFORMITIES IN POLISHING PAD 5.1 Significance of Polishing Pad The polyurethane CMP polishing pad form s a very integral part of any CMP process. The choice of the pad depends upon the CMP polishing equipment, nature of material polished and polishing output requ irements. For all practical purposes, commercial polishing pads are of viscoelastic nature and, are mostly used are made up of a matrix of cast polyurethane foam with filler material to control hardness or polyurethane impregnated felts [165, 166]. Th e pad carries the slurry on top of it, executes the polishing action, and transmits th e normal and shear forces for polishing. At the pad wafer interface, the slurry acts on the wafer and forms a compound with the material that is being polished. This co mpound is then removed when by the abrasive particles that act between the asperities of the pad and high points on wafer. The material removed, is then washed away due to the constant slurry flow on the pad. There is ongoing research to inves tigate the dependence of various pad material properties on the CMP process. Some of the findings thus far show that: 1) there is a drop in material removal rate, as a f unction of time due to the varying mechanical response under conditions of crit ical shear;  2) wafer pl anarity is a function of pad stiffness, which is determined by the elastic properties of the pad material [77, 103]. The
124 pad might be directly responsible for seve ral process defects lik e wafer to wafer non uniformity (WTWNU) where there is non homog eneity of polishing when one wafer is compared to another or within wafer non uniformity (WIWNU) where there is non homogeneity of polishing at different areas of the same wafer. Several techniques have been used before for characterization of the CMP polishing pad with Dynamic Mechanical Analysis (DMA) being the most prominent one [167-169]. However, almost all these techniques are de structive. A novel non destru ctive Ultrasound Transmission (UST) developed at USF can be effectively used for evaluation of the CMP pads [170, 171]. This technique works on the principle of ultrasound permeability through absorbing viscoelastic medium. The difference in the ultrasound transmission is used to determine the non uniformity or variation of specific gravity within a single pad (Fig. 5.1). 5.2 Method for Mapping and Isolation of Non Uniformities in Pad The UST technique has been used to study 32Â” and 20Â” diameter commercial polyurethane pads with and w ithout sub-pads. The regions of the pad which showed very high and low amplitude of UST (Â“regions of interestÂ”) were subjected to further evaluation. One inch coupons from the regions of interest were punched and its impact on the pad uniformity was studied. Material from the regions of interest was then evaluated using DMA to estimate their mechanical pr operties. Finally 6 inch coupons from the regions of interest were subj ected to CMP process evaluation to understand the impact of pad non uniformity on its CMP performance. Commercial CMP polyurethane pads were used for this study. The first evaluated pad was a closed cell pol yurethane polishing pad with pores created usually with a blowing agen t. This pad was attached with a soft sub
125 pad which comprised of non woven polyester fi bers impregnated with polyurethane to leave open porosity. This pad sub pad combination henceforth will be referred to as Pad 1. The second type of polishing pad used for the study comprised of the same aforementioned polishing pad without the sub-pad. This pad will henceforth be referred to as Pad 2. Both the pads were 32Â” in diameter had concentric grooves on them (K grooves). Fig. 5.1 Schematic of the Constructi on of the Ultrasound Testing Equipment
126 The combined thickness of pad 1 ( polishing pad + sub pad + 2 pressure sensitive adhesive layers) was measured to be around 110 7 mils. The thickness of Pad 2 (polishing pad + pressure sensitiv e adhesive layer) was 50 4 mils. 5.3 Ultra Sound Testing (UST) System The UST system consists of a flat square table that can accommodate polishing pads as big as 32Â” in diameter. The center of the ta ble has a circular hole that allows the two screws that holds the pad to pass through. One side of the table has a slot which enables the transverse movement of a 3Â” Valpey Fisher piezoceramic transducer along the radial directio n of the polishing pad. The transducer has a hole in the center and has trenches in the sides which help in gene ration of the vacuum which is used to hold the pad on the surface during UST measurement. The piezoceramic transducer emits resonance ultrasound vibrations at 26 KHz (fir st resonant frequency of the piezoceramic transducer), while a 7 mm diameter quartz r od or a pinducer housed in aluminum casing aligned directly above the transducer act s as the receiver. Th e received ultrasound frequency is then converted in to electri cal energy and the raw output is seen on the oscilloscope. The signal from the probe and re ference input are both sent to a lock in amplifier which records the amplitude of the received signal at the same frequency as the emitted signal. Different sections of the polishing pad are scanned be rotating the pad using the mounting screws by a step motor w ith 2 degrees angular steps, and moving the emitting piezoceramic transducer and the ali gned receiver using another step motor at 7mm radial steps with help of a threaded spindle and sc rew. There is a provision for vertical movement of the receiving pinducer w ith the help of vertically positioned spindle
127 and screw. The regions of polishing pad ha ving variations in sp ecific gravity transmit different amplitudes of UST at a same fr equency. The amplitude of the ultrasound permeability with in a pad obtained as a result of UST is normalized to against the average UST amplitude to estimate the compara tive variations in specific gravity in the different regions of the same polishing pad. The output of the measurement is a Doppler diagram in which different colors correspond to different amplitudes of UST with in the pad. The measurements are taken at a distance of 100 m below the pad surface to eliminate the presence of possibl e Â“air pocketsÂ”. It must be understood here that due to the viscoelastic nature of the cross linked polymer material of the polishing pad, all measurements taken are Â“curve fittedÂ” taki ng the effect of me asurement stress and temperature on the material into consideration. The details of the UST set up, measurement techniques, characterization pr ocedures and operation have already been published in literature [170, 171]. The UST experiments were performed on the Â“as receivedÂ” polishing pad with the plastic releas e liner below it. It is assumed that the bottom PSA and the plastic liner are uniform and will have a similar effect on the ultrasound transmission. These effects could then be filtered out when the entire data is normalized against the mean UST. The area showing the highest ultrasound transmission over the entire pad was designated as Â“hi gh transmissionÂ” or Â‘HTÂ’, while the areas showing the lowest ultrasound transmission we re designated as Â“low transmissionÂ” or Â‘LTÂ’. After the UST mapping, 1Â”, 3Â” and 6Â” diameter coupons were punched out of the HT and LT regions for subsequent experime nts. The pad was remapped using the same procedure discussed above without the 1Â” a nd 3Â” coupons being replaced. In case of 6Â” coupons, the coupons were replaced and part ial remapping of the pad was done.
1285.4 Evaluation of Mechanical and Tribological Properties of Polishing Pad Mechanical properties of small sectio ns of the pads from the HT and LT regions were evaluated using DMA to examin e the possible change in structure-property relationship of these regions. It has been show n before, that orientation of the polishing pad grooves or the directional orientation of the pad surface perforati ons directly affect the values of loss and storage modulus of th e pad materials that are evaluated using DMA [167, 169]. The pad displays hi gher mechanical properties when the perforations or grooves are oriented parallel to the length of the coupon used in the DMA. Two coupons (20mm X 10mm) were cut from Pad 1. The gr ooves on the pad were kept parallel to the length of the coupon used in DMA. The plastic release liner is removed as itÂ’s relatively adhesion with PSA was an impediment duri ng flexural loading of the sample during DMA. To facilitate the clam ping of the coupon, even the PSA is scrubbed off to before DMA. The bottom glue layer was scrubbed o ff of the coupon, allowing more efficient clamping in the holder. Data was obtained on a TA Instruments DMA 2980 (New Castle, DE) at temperature increments of 4oC with an isothermal time of 1 min. per increment starting from room temperature (20oC) and going up to 80oC. The flexural mode was used with single cantilever clamp and 3.0 m amplitude. Frequencies ranged from 0.6 Â– 30 Hz for all measurements. The set up and working of the DMA has been elaborately discussed in literature [172-174].
129 Table 5.1 Details of the CMP Experiments Performed on the Polishing Pad evaluated by UST Sr. No Item Description 1 Polishing Sample Materials 1. Patterned HDP 5000o A ILD (2 cm X 2 cm) (HDP 5K) 2. Blanket SiO2 ILD (2 cm X 2 cm) 2 Polishing Pad Pad 1 and Pad 2 (described before) 3 Slurry Cabot SSK 12 (ILD slurry) 4 Slurry Flow rate 52 ml/min 5 Conditioning 1. Break-in 20 min. with 3M diamond conditioner with DI water 2. Conditioning for 2min with D.I. water after ever set of thr ee polishing runs of 90 seconds each 5 Down force 1. 1, 3, 6 PSI for Patterned HDP 5000o A ILD 2. 3 PSI for Blanket SiO2 ILD 6 Platen Rotation/Carrier Rotation (RPM) 1. 100/95, 200/195, 300/295 for Patterned HDP 5000o A ILD 2. 200/195 Blanket SiO2 ILD The polishing experiments were done using the CETR CP4 bench top CMP tester which is a stand-alone bench-top simula tor with instrumented process control. The
130 CMP tester provides real-time measurements of tribological parameters such as coefficient of friction (COF). The CMP pe rformance of the pads was evaluated at different conditions of down force (PS I) and rotation (RPM). Experiments were performed on 1Â” X 1Â” coupons having 5000 high density plasma (HDP) dielectric material using Cabot Microelectronics SSK 12 commercial ILD slurry. The values of down force and bottom platen rotation used were: 3 and 6 PSI, and 100, 200 and 300 RPM respectively. The carrier rotation was ma intained at 5 RPM lower than the platen rotation. The slurry flow was maintained at 40 ml/min. The process conditions during all the experiments have been tabulated as show n in Table 5.1. The details of the CMP tester have been discussed before. 5.5 SEM Evaluation of Polishing Pad Usually, commercial polyurethane pads, either perforated or grooved consist of pores of approximately 30 m in diam eter and account for around 30 % of the volume of the pad. Fig 5.2 shows the cross-section of Pad 1. It can be seen from the figure that there are two layers of pad held together with the help of a pressure sensitive adhesive. The entire stack is attached to the platen with the help of another layer of PSA. In case of Pad 2, there is only one layer of PSA that is needed to attach the pad on the platen, due to the absence of a sub-pad.
131 Fig. 5.2 Cross-section Scanning Electron Micrograph of the Evaluated Commercial Polyurethane Polishing pad with a Sub-pad 5.6 Isolation of Polishing Pad Coupons The results of UST s can over one quarter (90o scan) of the donut shaped Pad 1 are shown in Fig. 5.3 a-d. There are distin ct HT and LT regions on the surface of the pad with UST amplitude variation changing by a factor of two. Th e variation of the ultrasound transmission is normalized. Fig 5.3 b shows the histogram of the distribution of the intensity of the ultrasound transmission ove r the section of the pa d. There is a large variation in ultrasound transmissi on within a quarter of the d onut shaped pad. A set of 1 IC1000 Suba IV PSA 2 PSA 2 (Interface) adhesive)
132 inch coupons were punched out from HT and LT regions. The pad was then remapped and results of the ultrasound scan are shown in Fig. 5.3 c. Dark areas can be seen on the pad due to the presence of an air gap in region from where the coupons were punched out. There is a significant change in the in tensity in ultrasound transmission of the pad without the coupons. Fig. 5.3 Ultrasonic Transmission Maps of th e Quarter IC1000/SubaIV pad a) Before and b) After punching of the 1Â” coupons. Three c oupons were punched at the High-Intensity (white) Area and Four Coupons at Low-Inte nsity (Black) Area. The Effect of Pad Acoustic Homogenization after Punching is Illustra ted in Respective Histograms (b, d)
133 The Â“dark areasÂ” of LT are virtually absent, except from the region from where coupons were punched out. Fig. 5. 3 c shows a more homogenous distribution profile of the ultrasound transmission over the entire pad. The region of the pad which showed high ultrasound transmission and from where the pad coupons were previously punched showed significantly lower ultra s ound transmission, while the regions of the pad which previously showed low ultras ound transmission exhibited an increase in ultrasound transmission. A reduced inhomogeneity in the histogram shown in Fig. 5.3 d can be attributed to the stra in relaxation accompanied with elastic stress release due to punching of coupons. It can be thus said that the built in stresses in the polishing pad bring about a variation in the specific gravity. 5.7 Isolation of 6 Inch Polishing Pad Coupons Fig. 5.4 a shows the ultrasound map of the 360o scan of Pad 1. The variation of the ultra sound permeability through the region was norma lized over a scale from 0.8 to 1.2. Upon identification of the HT and the LT regions of the pad, coupons of 6Â” diameter were punched out with the intention of entrappi ng the entire non-homogeneous region. (It was also chosen to make sure no superfluous effects from cutting the wafer were seen in the interior of the coupon as was seen with the 3 inch and 1 inch coupons.) Fig. 5.4 (b, c) shows the partial scans of the polishing pad when the HT and LT pad coupons were put back in the region from wher e they were isolated. It can be seen from Fig. 5.4 b that there was a reduction in the ultrasound transmission in the HT coupon after punching and remapping. Also, the c oupon showed reduced ultrasound signal transmission in comparison to the surrounding pa d. This indicates that HT region in the
134 full map corresponds to the pad region under compressive stress which is relaxed when coupon was punched. When the 6Â” coupon from the LT area of the pad was punched and remapped after placing it in its original position (ref. Fig 5.4c) the ultra sound transmission over the entire coupon remains appr oximately same as seen in the previous scan (ref Fig. 5.4a). However, there is an increase in the UST amplitude in the center of this coupon suggesting that certain regions corresponding to the LT area of the pad were under tensile stress. After th e coupon was punched out, there was a stress release from these areas there by making them denser. Th is increases the ultrasound permeability due to the localized increase in the pad specifi c gravity. The distinct lower ultrasound transmission seen at the edge of both th e coupons in the scans obtained during remapping (ref Fig. 5.4(b, c)) shows the presence of an air gap at th e boundary of the pad and the coupons. Certain distinct areas of signifi cantly reduced UST amplitude in the coupon shown in Fig. 6b also indicates that there is an air gap underneath the coupon when it is put back in place. This may be due to th e loss of flatness on the bottom of the coupon after stress release. We estimated that 20% variation of the UST permeability between high and low intensity areas corresponds to 10% relative change of the pad density (specific gravity). It must be mentioned here that previously expe riments were performed by isolating 3Â” inch coupons from the pad. Th e results of this expe riment were found to be similar to those reported in Fig.5.4. Th e coupons punched out of the pad to entrap the regions of non-uniformity bring about a genera te change in the stre ss distribution regime in the pad. Hence, the local density varia tion needs to be examined on the Pad 1n its entirety.
135 Fig. 5.4 a) Pad Mapped Before Punching 6inch Coupons, b) Area of Pad Remapped after Replacing the Punched HT Coupon, c) Area of Pad Remapped after Replacing the Punched LT Coupon (All the Values have b een Normalized over the Entire Area) c b HT LT LT HT
1365.8 Pad Dynamic Mechanical Analysis Variation of Storage Modulus 0 20 40 60 80 100 25456585 Temperature (oC)Storage Modulus (E) (Mpa) 30 Hz (Low Amplitude of UST) 30 Hz (High amplitude of UST) 0.6 Hz (High Amplitude of UST) 0.6 Hz (Low Amplitude of UST) Variation of Loss Modulus 0 5 10 15 20 25 30 25456585 Temperature (oC)Loss Modulus (E) (Mpa) 30 Hz (Low Transmission) 30 Hz (High Transmission) 0.6 Hz (Low Transmission) 0.6 Hz (High Transmission) Variation of Tan d0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 305070Temperature (oC)tan d 30 Hz (LowTransmission) 30 Hz (High Transmission) 0.6 Hz (HighTransmission) 0.6 Hz (Low Transmission) b c Fig 5.5 a) Variation of storage modulus Vs te mperature, b) Variation of loss modulus Vs temperature c) Variation of tan Vs temperature of samples tested from low and high intensity region of the pad a
137 In order to evaluate a possible variati on in the viscoelastic properties of the HT and LT, DMA was performed on pieces of pads from the HT and LT coupons that were punched out. Fig 5.5 (a, b, c) shows th e variation of storage modulus, loss modulus and tan with temperature respectively as eval uated by DMA experime nts carried out at 30 Hz and 0.6 Hz frequency (two extreme fr equency values in the data set reported). There is no significant difference between the magnitude and variation of storage modulus, loss modulus and tan for HT and LT samples. It ca n be thus said that though DMA gives an estimate of the bulk viscoelast ic properties of the polishing pad materials (polyurethane), and is not the best technique to characterize the entire polishing pad. 5.9 Frictional Characteristic s Polishing Pad Regions For evaluation of tribological propertie s (which would direct ly correlate with the CMP performance) of the regions of nonhomogeneity of the polishing pad, new samples of Pad 1 and Pad 2 were mapped using the same UST system. The 360o scans of both these samples are shown in Figs. 5.6 and 5. 7. It appeared that there was a lesser area with extreme amplitude (very high or very low) amplitude of ultrasound transmission and hence less variation of specific gravity in Pa d 2 as compared to Pad 1. We propose that the relatively lesser variation of specific gr avity in Pad 2 could be attributed to the absence of an additional layer of the PSA wh ich joins the pad and s ub-pad. The variation of the thickness of the PSA at the interface, in addition to the possible contact stresses can generate variable pressure in the body of the polishing pa d. Due to the viscoelastic nature of the polishing pad, these variable stresses may cause variable deformation in the pad matrix which in turn causes the variation in the specific gravity in pad material.
138 Fig.5.6 Ultrasound Transmission Scan of Pad with Sub-pad Fig. 5.7 Ultra-sound Transmission Scan for Pad without Sub-pad HT LT HT LT
139 It must be mentioned here that ther e was no transient variation of in the density of different polishing pad regions when the same pad was remapped in its entirety after an interval of 24 hours. When HT and LT coupons were punched out, there was a general stress release in the pad matrix The histograms of the pad ultra sound permeability obtained before and after punching coupons out of the pad also support this hypothesis (ref. Fig. 5.3 (b, d)). When 6Â” c oupons are pressed out of the pad, there might not be complete relaxation of the built in stre sses due to the partially viscous nature of polyurethane. However, as can be see from Fig. 5.4 (a, b, c), there is a definite reorganization of the stress topography. Due to the nature of pad fabrication and viscoelastic properties of polyurethane materi al a complete uniform and stress free Pad 1s difficult to fabricate and there will be minor va riations of specific gravity in the polishing pad. However, it is necessary to ensure that specific gravity variations in the pad stay to the absolute unavoidable minimum in order to obtain uniform CMP output The 6 inch diameter HT and LT both Pa d 1 and Pad 2 (ref Figs. 5.6, 5.7) were used to polish 5000 HDP ILD material at the conditions elaborated in the experimental section of this paper to evaluate their CMP performance. Each polishing run was performed for 90 seconds. The COF, whic h is the ratio of the shear force (Fs) to the normal force (FN) ( N sF F COF ), was measured in situ. The average of the values of COF obtained for the first 30 seconds of polishing ha s been reported as th e characteristic COF value of the pad coupon. Fig. 5.8 (a, b) show s the variation of COF with down force and platen rotation for 3, 6 PSI and 100, 200 a nd 300 RPM respectively for Pad 1 and 2. It can be safely said that the HT and LT regi ons show different values of COF during CMP
140 under the various CMP conditions of PSI and RPM. There is a general trend of a decrease in the values of COF values with the increase in platen velocity. It must be noted here that, the values of COF reported are an average of 5 CMP runs with a error of 57%. The COF for a given process condition is due to the combination of shear at the pad asperity heights and shear due to the slurry film. The shear or friction produced by the Â“t wo body abrasionÂ” at the pad wafer interface is higher than the Â“three body abrasionÂ”. With in crease in the platen ve locity, there is an increase in the slurry film thickness on the surf ace of the pad. This in turn increases the part of the down pressure supported by slurry film and decreases the pad asperity contact with the wafer. Due to increase in the exte nt of three body contact abrasion at the pad wafer interface, there is a decrease in COF with increase in platen velocity and slurry film thickness [175, 176]. Figs. 5.8 (a, b) also s hows a decrease in COF for both Pad 1 and 2 with an increase in down force. These resu lts are in agreement with those obtained previously on this set up. Fig. 5.8 a shows the comparative COF va lues for HT and LT regions of Pad 1. At 3 PSI, the LT region shows consistently hi gher values of COF for all the different values of platen rotation. However, at 6 PSI the COF values for HT region are mostly higher. Similar trend can be seen from Fig. 5.8b, where the values of COF for HT region of Pad 2 are consistently higher as comp ared to the corresponding LT region for the given set of machine input parameters. It ha s been shown previously, that there is no change in the mechanical properties (sto rage modulus, loss modulus and toughness) the HT and LT regions (ref Fig. 5.6 a-c). Hence, when the values of COF are compared at the given set of machine conditions, for the di fferent regions of ultrasound permeability, it
141 can be expected that the mechanical response of pad under those conditions should remain consistent for both types of the pad samples. The change in COF can be thus attributed to the change in the shear force experienced on the pad under the same machine conditions. The values of applied down pressure W, tan and H remaining the same, the shear force changes due to the change in shear stress per unit area experienced by the pad. This may be a function of the variable specific gravity of the pad. A comparison of this pad behavior could be made with a spring coil mattress wherein, certain different small regions are kept unde r constant compressive and tensile loading for sustained period of time. After the pressure on all the regions is released, there is, in general, a relaxation of stress in the spring coils of the mattress. However, the equilibrium state of the coils that were form erly under tensile load ing is different from the spring coils form the regions that were previously under compre ssive stress. In the same way, upon stress relaxation due to isolat ion from the parent pad, the different HT and LT regions of the pad show slight variat ion in their respective specific gravity at equilibrium. This difference in the specific gr avity may thus be res ponsible for the slight variation of COF seen during the CMP experi ments (ref Figs. 5.8 a, b). The dependence of the maximum shear stress of the pad surface on its specific gravity needs to be further investigated. The material removal during CMP directly depends upon the shear force exerted on the material surface. In the event of the variation in shear force for given polishing conditions, there is a chance of non uniform material removal. This phenomenon thus has the potential to cau se the WIWNU and WT WNU defects during the CMP process.
142 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 100200300 Platen Rotation (RPM)COF 3 PSI Low Transmission 3 PSI High Transmission 6 PSI Low Transmission 6 PSI High Transmission 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 100200300 Platen Rotation (RPM)COF 3 PSI Low Transmission 3 PSI High Transmission 6 PSI Low Transmission 6 PSI High Transmission a b Fig. 5.8 Variation of COF for a) Pad 1 a nd b) Pad 2 during the CMP process for the Various Values of PSI and RPM
1435.10 Summary of Investigations of Polishing Pad Non-uniformities The spatial variation of the ultr asound permeability in the polishing pad measured by the scanning UST metrology is an excellent indicator of the specific gravity variation in the CMP polishing pad. The most prominent reason for the variation in the specific gravity is the variation of PSA laye r which bonds the pad with the sub-pad. The built in stresses are entrapped in the pad a nd hence, isolation of the Â“highÂ” and Â“lowÂ” ultrasonic transmission regions brings about an overall redistributi on (relaxation) of the stresses in the pad and improves pad uniform ity. The polishing pad must be evaluated as whole in a non-destructive manner, as the mechanical properties of the small padÂ’s portion represent the bulk mechanical propert ies of the pad polyuret hane materials and are not adequate indicators of the pad CM P performance. The pad when subjected to CMP evaluation shows a decrease in COF w ith increase in platen rotation and down pressure due to the mechanical response char acteristics of polymers. There is a variation in the polishing performance of the different isolated regions of non uniformity. This may be translated in to variation in material removal during the CMP and cause the WTWNU and WIWNU defects. Such a variation in the shear force occurs in spite of HT and LT region isolation. Hence, CMP performed with the polishing pads with built in stress and having areas of varying specific gravity will have a direct impact on the process repeatability.
144 CHAPTER SIX INVESTIGATION AND OPTIMIZATION OF APPLICATION SPECIFIC CMP PADS 6.1 Application Specific Pad (ASP) for CMP Due to its crucial role in CMP, th e mechanical, chemical, and physical properties of the polishing pad play a signifi cant role in the material removal and final global planarization of the wafer surface. The properties of the pad are both intrinsic and extrinsic functions of the polymer or the foam from which the pad is made [177-179]. Historically, polyurethane-bas ed pads (e.g., IC1000 / Suba IV stacked pad (Rodel, Newark, DE)) have been used to obtain both good uniformity and efficient topography reduction, due to their unique property of combining high strength with high hardness and modulus, plus high elongation at break [ 77, 103]. With increas ed polishing cycles, slurry particles and polishing debris can be trapped on the surface pores. Also, the pad surface becomes glazed due to: 1) deposition of worn/abr aded material and 2) smoothing effect of the slurry residue, causing a decrease in pad surf ace micro-roughness . This causes a drop in the material remova l rate (MRR) during the CMP process. The conditioning process done typically using di amond grit restores pad surface asperities, removes worn out debris and the reaction pr oducts of the pad and the slurry. However, the process of conditioning accelerates pad w ear, reduces the pad life and modifies pad
145 surface after every run. The modified surface may even affect the CMP resulting wafer uniformity after polishing [181-183]. If the conditioning step is bypassed, there will be far reaching benefits in terms of: 1) operato r convenience, 2) process optimization, 3) process down time and 4) uniformity of process output. It has been shown that the material properties of the polis hing pad directly impact its CMP output [184, 185]. There is a dr op in MRR, as a function of time due to the varying mechanical response under conditions of critical shear . There is also a variation in wafer planarity as a function of pad stiffness, which is determined by the elastic properties of the pad material [ 186, 187]. Also, there can be numerous other process defects such as dishing , eros ion , non planarity etc. due to the mismatch of the pad wafer hardness. The use of a surface coating on the pad is becoming increasingly widespread as an effective means of enhancing and modi fying the surface mechanical properties by Â“surface engineeringÂ” . The mechanical and viscoelastic properties of the pad material are altered if the polishing pad is co ated appropriately. Variation in the thickness of coating can yield a thin f ilm of varying hardness which can be then matched with the hardness of the wafer . The matching of mechanical properties inherently affords polishing rate selectivity and reduc es process induced defects. Taking the aforementioned factors in to account a novel polishing pad using polyolefin  instead of polyur ethane has been developed. This pad is coated with a ceramic on top to tune the mechanical prope rties of pad surface during the CMP process . The coatings lend application speci ficity to the pad by matching its surface hardness to that of the material being polished. The salient features of this pad are: 1) no
146 need for the traditional pad Â‘break-inÂ’ befo re polish, no conditioning / dressing ever, 2) no need to keep pads wet in idle mode, 3) long pad life, 4) high selectivity, 5) ergonomically friendly / easy pad changes due to a redesigned pressure sensitive adhesive and 6) demonstrated pad-to-pad reproducibility. 6.2 Scheme for Optimization of ASP Properties In this research, we have performe d metrology, characterization and prototype CMP performance evaluation of different va riants of ASP for Tungsten polishing. The cross section analysis, surface characte rization, estimation of surface and bulk mechanical properties, evaluation of stat ic and dynamic tribological properties for different candidate polishing pads coated with ceramic material for different time durations has been performed. The pads w ith varying total thickness and fixed optimum ceramic surface coating time were then subjected to the prototype CMP process. The results of the experiments were directly used to finalize: 1) surf ace ceramic coating time and 2) pad thickness, prior to commercialization. The polishing pad used in this research was made up of talc and calcium carbonate (CaCO3) filled EVA-polyethylene blend ther moplastic. Cross linking is done via a free-radical mechanism, catalyzed by cumyl-peroxide, during the thermal extrusion process . The use of metal-oxides or ceramics such as Te traethylorthosilicate (TEOS) (used in this case) to coat th e pad surface makes the pad permanently hydrophilic. The surface modifi cation of the psiloQuestÂ’s application specific pads (ASP) was accomplished through plasma e nhanced vapor deposition (PECVD) from organometallic precursors. The details of the surface coating technique have already been
147 elaborated elsewhere . The surfaces of fi ve different pads (for convenience referred to Pad 1, Pad 2, Pad 3, Pad 4 and Pad 5) were coated with TEOS for five different time durations of 10, 20, 30, 40 and 60 minutes respectively. 6.3 Experimental Techniques for ASP Property Evaluation Scanning Electron Microscopy (SEM) was used to examine the pad surface and cross section, while nanoindentation was used to evaluate the surface micromechanical properties as a function of coati ng time. Nanoindentation technique is widely used for estimation of the mechanical propert ies of thin films on a substrate [194, 195]. The mechanical response to th e indentations, usually by diam ond indenters, is used to evaluate the mechanical properties such as Hardness and YoungÂ’s modulus. In this case, indentation experiments were carried out on the surface modified/grafted pads using a NANOTEST 600. Indentations were performed under ultra low load range. Initial load was 0.1mN, which is a machine parameter. The control parameter was set to Â‘depth controlledÂ’ and each pad was indented for varying depth with a maximum depth of 10000 nm. The results reported are an average of 10 indentations performed for the constant depth of 1600 nm. The surface modification of the polymer pad for different deposition time was characterized using PHI 5400 X -Ray photoelectr on spectrometer. The base pressure used was 10-10 Torr. The spectrometer was calibrated using a metallic gold standard (Au (4f7/2): 84.0 0.1 eV). Nonmonochromatic Mg K X-ray source with energy 1253 eV at a power of 250 W was used for the analys is. Charging shifts produced by the sample were removed by using binding energy scale referenced with re spect to the binding
148 energy of the hydrogen part of adventitious carbon line at 285.0 eV. Peak deconvolution was carried out usi ng PEAK FIT Software. The static coefficient of friction (COF ) of the candidate pads was evaluated using the wear test was run with a Kiedon U 45 static COF tool . The variation of static coefficient of friction and static wear rate was plotted against the surface coating time of the candidate pad. The prototype CM P testing was performed on the CETR CP-4 bench top CMP tester. The Si wafers having 6000 tungsten over buried SiO2 layer were used during the CMP run. The polishing was pe rformed at an optimized set of machine parameters known as Â“best known methodÂ” (BKM ). The details of the CMP experimental conditions can be seen in Table 6.1. The co efficient of friction (COF), acoustic emission (AE) and MRR for a given process run were determined in-situ. To examine the interdependence of the parameters evalua ted during all the aforementioned metrology experiments, a cross tabulation of the differe nt variables was done using the commercial JMP software. Finally, after optimization of pad surface coating time, candidate pads with different total thickness (thin film + polishing pad), starting from 70 mils and going up to 135 mils (intermediate pad thickness in mils were: 75, 85, 95, 110, 115), and having an optimized coating time were subjected to prototype CMP perfor mance evaluation. It must be noted here, that CMP polishing runs were repeated for 8-10 times for a given condition. The values of data for samples w ith very high surface microscratching defects were discarded. The surface microscratching may sometimes result from slurry particle agglomeration or impregnation of abraded chip on the pad. The average values (with an error of 5 %) MRR and COF were reported.
149 Table 6.1 Experimental details of CM P process used for ASP evaluation # Parameter Conditions 1 Down Pressure 2 PSI 2 Platen Rotation 150 RPM 3 Carrier Rotation 145 RPM 3 Slider Position and Movement Statio nary at 40 mm from platen center 4 Slurry Commercial Ca bot W-2000 (pH 2.5) 5 Pad Different psiloQuest ASP with spiral grooves 6 Time of polishing 90 sec ( Values of AE and COF noted in the first 30 seconds have been plotted) 7 Polishing Specimen 1Â” X 1Â” coupon of 6000 W/SiO2/Si (wafer) 6.4 Evaluation of ASP Surface Mi cro-mechanical Properties The surface of the ASP can be permanently modified when the pad asperities and troughs are covered with TEOS. The cro ss section the ASP in comparison with a commercial IC1000 Suba IV polishing pad can be seen in Fig. 6.1 a, b. The commercial polyurethane pads need two layers of pressu re sensitive adhesive (PSA) when a sub pad is attached to the polishing pad to impart better conformability. The variation in the thickness and expansion of the PSA with time and temperature in conjunction with the interfacial contact stresses is known to cause several specifi c gravity variations in the body of the polishing pad .
150 Fig. 6.1 Cross-sectional SEM micrograph of a) PsiloQuestÂ’s ASP and b) Commercial Polyurethane Polishing pad with a sub pad The load vs. displacement data curv es obtained from nanoindentation were analyzed using the Oliver and Pharr Method to estimate the micromechanical properties of the surface TEOS film [155, 156]. The viscoelastic nature of the substrate may directly affect the surface penetration of the nanoi ndentation [198, 199]. Fig. 6.2 shows the Ceramic Coated Polyolefin Foam Condensed Polyolefin PSA IC1000 Suba IV PSA 2 PSA 1 (Interface) a b
151 loading-unloading curve when nanoindentation was performed on Pad 1. During loading, several nodes of non-uniform tip penetrati on were observed with increasing load (Xb: pop in). Conversely during unloading, the tip showed exactly the opposite phenomenon of non uniform withdrawal (Xc: pop out). The depth at wh ich the non-linearity in the loading curve occurs for the first time is marked as Xa. Such non-uniform penetration of the indenter tip into the coatings probably results from the initia tion and/or propagation of the crack in the ceramic coating . It is well known that plastic deformation under indentation, which is analogous to the impr egnation of abrasives on the surface, is a critical attribute of CMP pads. Thus, these Â‘eventsÂ’ in the load-depth curves may be good predictors of pad performance when put in to service. The studies revealed that Xa is a function of PECVD coating time. Fig. 6.2 Loading and Unloading Behavior E xhibited by the PsiloQuestÂ’s ASP coated for 10 minutes (Pad 1). Note the non-linearity at the Specific Sites of the Loading and Unloading Curve Represent Plasti c Deformation under Indentation
152 Fig. 6.3 shows the correlation of Xa on PECVD-TEOS coating time for all the candidate pad samples with 1600 nm deep i ndentations. The data suggest that Xa is related to the thickness of the ceramic coat ed on the surface of the pad foam. The depth of penetration before plastic deformation increases with increase in TEOS thickness, till it reaches a plateau for a coating time of 40 minutes. The mechanical properties (Hardness and YoungÂ’s Modulus) derived from th e loading unloading curve are shown in Fig. 6.4. Fig. 6.3 The Variation of Depth of First Nonlinearity of in the Loading Curve (Xa) as a Function of PECVD-TEOS Coating Ti me for Different Pads Evaluated The increase in pad coating time: 1) will increase the thickness of the surface coating and 2) can possibly improve the de nsity and conformab ility of the surface 40 45 50 55 60 65 70 75 0102030405060 Coating Time (min)Xa (nm)
153 coating. There is an increase in the har dness and YoungÂ’s Modulus of the surface with the increase in PECVD-TEOS depos ition time. This may be primarily due to the fact that as the thickness of the rigid film on the comp liant substrate increases, for a given depth of penetration, nanoindentation pre dominantly tends to yield the actual thin film mechanical properties with decreasing s ubstrate compliance effect. 0 10 20 30 40 50 60 70 80 0102030405060 Coating TimeMPa Elastic Modulus (Mpa) Hardness (Mpa) Fig. 6.4 Variation of Elastic Modulus and Hardness with TEOS Coating Time Evaluated by Nanoindentation for Different Evaluated Polishing Pads 6.5 Evaluation of ASP Surface Mi cro-mechanical Properties The information on the variation in the pad surface chemistry with the change in coating time was obtained using XPS analysis The C (1s) signal resolved into three major peaks: a peak at 285.0 eV represen ts the C-C and C-H bonds and a peak observed at 286.5 eV represents the C-O bond. These re sults agree with the previously published literature . A peak com ponent centered at 289-289.3 eV can be attributed to carbamide [-O-C (NH2) =O] functional group from the residues of the blowing agent
154 used in the pad substrate ma nufacturing process. For the specimens coated for 40 and 45 minutes respectively, an additional peak was observed near 283.6 eV. This peak can be assigned to a C-Si bond. The XPS Si (2p) spectrum was deconvoluted into two major peaks at 102.3 and 103.4 eV, which represent the silicate and Si-O species, respectively. The data indicate that PECVD processes produced surface films rich in silanol, consistent with TEOS films deposited at low process temperatures [202-207]. Fig. 6.5 shows the XPS results for pad samples coated with PECVD-TEOS for the given different time durations. Fig. 6. 6 shows the oxygen to sili con (O/Si) intensity ratio values calculated from the XPS data. This ratio is proportional to the concentration of the silanol in the deposited coatings. Ther e was a decrease in the silanol concentration with the increase in the co ating time up to coating time of 30 minutes; beyond which the silanol concentration again show ed an increase. Furthermore, a small peak is observed at 102.1 eV, which represents Si-N bond for Pads 3, 4 and 5. There is also an abrupt reduction in Si to N ratio, wh ich indicates an increase in ni trogen species on the surface. The XPS observations suggest that as the deposition time increases, substrate (polymer pad) temperature also increases, which causes outgassing of the nitrogen used in foaming the substrate. The nitrogen reacts with the Si species on the pad surface, which causes formation of Si3N4 species in stoichiometric conversion from SiO4 to Si3N4 [208-210]. At higher coating times, the ion bombardment of the foam surface may generate appreciable amounts of C radicals on the su rface of the pad. These ra dicals react with the silicon species to form silicon carbide (SiC), whic h is subsequently incorporated in surface coating of the pad.
155 Fig. 6.5 X-Ray Photoelectron Spectroscopy (X PS) Data for Polishing Pads with Different TEOS Surface Coating Time
156 The incorporation of Si3N4 and SiC species would certainly be expected to make the pad surface stiffer and may contribute towards the improvement in its mechanical properties. The XPS results also show that the PECVD TEOS forms a very complex compound on the surface of the polyolefin foam due to the combination of several elementary reactions The PECVD-TEOS coating pr ocess produces both silica and silicates (SiOX and SiO2) on the foam surface. The change in the surface chemistry of the substrate as a function of deposition time at a fixed TEOS/O2 mixture can be progressively observed from the XPS results. Fig. 6.6 Variation in Surface Silanol / Silicat es in PECVD-TEOS Co ating as a Function of TEOS Deposition Time 0 2 4 6 8 10 12 0102030405060 Coating Time (min)O/Si
1576.6 Evaluation Static Tribological Properties of ASP Fig 6.7 Variation of Static Wear Rate and St atic Coefficient of Friction for Pads with Different PECVD TEOS Surface Coating Time A static wear test was performed to ga uge the impact of the variation of the surface chemistry on the tribological properties of the pad. The variation of pad surface wear rate and the static COF for different pa ds evaluated using the blanket W wafers is shown in Fig. 6.7. There was a general tre nd showing an increase in the static COF and wear rate with the increase in pad surface coating time. Initially there was a plateau observed when COF and Wear ra te for Pads 3 and 4 were evaluated. A special case of pad coated with 35 minutes of PECVD TEOS co ating was evaluated specifically for this experiment. The 35 minutes PECVD TEOS coat ed pad showed a higher wear rate and static COF as compared to Pad 3 and 4. The in crease in the static COF and wear rate for 0.2 0.3 0.4 0.5 0.6 0.7 0.8 1000 1200 1400 1600 1800 2000 010203040506070 Mean Satic COF W-RR / A/minMean Satic COFW-RR / A/minCoating Time / mins.
158 the 35 minutes coated pad correlates very well with the previous XPS observation of decrease in O/Si ratio or incr ease in Si content on the surface with coating time. Also, the further increase in static COF and W wear ra te is due to combination on Si with Carbon and Nitrogen on the surface to SiC and Si3N4 (as verified by XPS). 6.7 CMP (Dynamic Tribological Properties)Evaluation of ASP The sample-to-sample variation of dynamic COF during prototype CMP experiments for different candidate pads is show n in Fig. 6.8. There does not appear to be a definitive trend in the vari ation of COF (COF of 5 CMP r uns is reported here without averaging) with increasing coating time. Fig. 6.8 Variation of the Average Value of COF (taken for First 60 Seconds) for the Different Candidate Pads dur ing the Polishing Runs at 2 PSI Down Force and 150 RPM Platen Rotation 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 010203040506070 TEOS Deposition TimeCOF
159 The variation of average MRR with coating time for different evaluated polishing pads is shown in Fig. 6.9. It can be seen from the figure, that there was a general decrease in MRR with increase in coating time from Pad 1 to 3. The subsequent increase in coating time brings about an increase in MRR for Pad 4 but saturation is reached at this level. It must be noted here that the pad MRR is significantly lower than accepted standards for W polishing because of: 1) use of prototype CMP experimental set up, 2) no slider oscillation, 3) relatively lo wer values of down force and platen rotation chosen for CMP evaluation. Fig. 6.9 Variation of the Average Value of MRR for the Different Candidate Pads during the Polishing Runs at 2 PSI Down Force and 150 RPM Platen Rotation 60 70 80 90 100 110 120 010203040506070 Coating Time (min)MRR (nm/min )
160 In order to find the interdependence of the different evaluated pad metrology parameters namely: 1) coating time, 2) mean static COF, 3) mean static wear rate, 4) mean dynamic COF and 5) mean MRR, multivar iate cross correlation was done using the JMP TM software. The correlation coefficients of the different evaluated materials have been shown in Table 6.2. It can be seen from the table that there is a direct dependence of static surface properties such as wear rate and static CO F with the pad coating time. However, there was little or no direct co rrelation of pad surface coating time duration with the CMP output parameters of dynamic COF and MRR. Thus, any trend in the variation of dynamic COF and MRR with the pa d coating time can only be considered as the secondary effects of polis hing. There is a very good correlation between the static COF and static wear rate of the pad surface, but there is very little correlation between the dynamic COF and MRR. The static tribol ogical property evaluation, which is non destructive, was done in orde r to correlate it with the CMP performance of the polishing pad. However due to their inadequate co rrelation with the dynamic tribological properties, the static tribologi cal properties cannot be consid ered as good indicators of pad CMP performance. The static COF is the ratio of the late ral frictional force to the normal reaction when two surfaces come in contact, while the dynamic COF during CMP the ratio of the shear force (Fs) to the normal force (FN) ( N sF F COF )56. The increase in down force brings about a corresponding increase in the shear force (Fs) as shown in Eqn. 6.157 tan H W Fs (6.1)
161 where W is the downward loading, is the maximum shear st ress on the given pad area, H is the hardness of the polishing pad and tan is the ratio of the loss modulus to the storage modulus (which is also a measur e of the toughness) of the pad. The dynamic shear is a function of the loading, pad har dness, stiffness and the pad asperity contact with the wafer. The normal reaction upon lo ading depends upon the storage modulus (elastic component) of the polishing pa d. The MRR during CMP on the other hand partially depends upon the surface shear and is also governed by factors such as slurry film thickness, abrasive size, pad complia nce, surface reaction kinetics and heat dissipation from the interface. Hence, dynamic COF and MRR show very weak correlation between themselves. The increase in pad surface micromechanical properties with the increasing in coati ng time however causes an increase in the wear of the W samples. The static wear of W is only due to the lateral frictional force between W and pad surface. Thus a very good correlation between pa d coating time with static friction and wear is seen. The static friction and static wear also have a good correlation due to the very same reason. As the surface film coat ing time increases, the thickne ss and conformability of the film on the surface increases. This may bri ng about an increase in the stiffness of the pad at and near the surface. This may decrease the MRR due to decrease in pad compliance and asperity contact under the app lied load as seen from the MRR data for Pad 1, 2 and 3 (ref. Fig. 6.9). Though the substr ate is maintained at room temperature, the electrode used to strike the plasma in the PECVD chamber operates at an elevated temperature. Due to heat dissi pation by radiation, there is an increase in the substrate temperature which becomes pronounced wh en coating time exceeds 30 minutes. This
162 increase in substrate temper ature may strain or even break the bonds between the atoms in the cross-linked polyolefin foam. The return of the polishing pad to the room temperature after surface coating may not re -establish all the bonds and the foam may not regain its original properties. This hypothe sis is supported by the evaluation of the bulk mechanical properties using Dynamic Mechanic al Analysis of similar ASP surface coated for different times reported elsewhere . The reported data show that there is a general trend showing a decrea se in storage modulus and an increase in the loss modulus of the pad with increase in coating time. Hence, even as the surface mechanical properties of the pad increase with the coating time in excess of 30 minutes the overall compliance of the polyolefin substrate increases the pad asperity contact with the wafer, thereby improving the MRR for Pad 4 and 5. The O/Si ratio on the surface appears to vary in synchronization with the MRR. However, there does not seem to be any direct correlation between the chemical properties of the surface and MRR during CMP. Table 6.2 Multivariate Correlations between Various Independent Results and PECVD Coating Time Coating Time / min Mean Static COF W-RR / A/min Mean (Dynamic COF) W CMP-RR / A/min Coating Time / min 1.0000 0.9168 0.9668 0.0495 -0.9735 Mean (Dynamic COF) 0.0495 0.2897 -0.2022 1.0000 -0.1584 Mean Static COF 0.9168 1.0000 0.8140 0.2897 -0.9558 W-RR / A/min 0.9668 0.8140 1.0000 -0.2022 -0.9155 W CMP-RR / A/min -0.9735 -0.9558 -0.9155 -0.1584 1.0000
163 Though the MRR was found to be highest for Pad 1, nanoindentation studies suggested that the initial surf ace penetration for the onset of plastic deformation for Pad 2 was the lowest (ref. Fig. 6.3). Therefore, th is pad was found to be most susceptible to wear and damage caused by the impregnation of slurry abrasive particles. The value of Xa seemed to reach a plateau for Pad 4 and 5. Thus a TEOS surface coating time of 40 minutes was decided to be optim um from a MRR perspective. 6.8 Optimization of ASP Total Thickness It is difficult to theoretically quantify the appropriate thickness of the pad foam for which a surface TEOS coating of 40 min. would prove op timum in terms of CMP performance, especially MRR. Hence, polishing pads of different total thickness (film + foam) ranging from 75 to 135 mils were subjected to CMP performance evaluation on the CETR bench top CMP tester The variation of COF for different pad coating thickness can be seen in Fig. 6.10. Th e COF shows an increase with the increase in pad thickness. The combined mechanical pr operties of the polyolefin pad comprise of a combination of mechanical properties of the hard ceramic surface and compliant polyolefin substrate. The eff ective contribution to the mechanical properties of the polyolefin substrate in comparison with the surface increases with increase in the pad thickness for the same coating time. The thicker foam adds pad compliance and increases the extent of viscous deformation for a given down force. This results in the decrease in the normal reaction of the polishing for the give n set of process parameters. As discussed earlier, the COF being the ratio of the shear force to normal reaction, the decrease in the normal reaction with the increase in the visc ous nature of the polishing pad brings about
164 progressive increase in the pad COF with increase in total thickness. The variation of MRR for different pad thickness is shown in Fig. 6.11. There is an increase in MRR up to pad thickness of 95 mils; however there is a sharp decrease in MRR for pad thickness of 110 mils. The further increase in pad thic kness does not seem to impact the MRR significantly as further increase in the viscous na ture of the pad in not translated in to the compliance of the surface due to the inherent stiffness of the ceramic coating. Thus the thickness of pad (95 mils) showing the ma ximum removal rate was chosen as the optimum pad thickness from the MRR perspect ive. Finally, it must be noted that considerations besides MRR were taken in to account before the polishing pad was actually commercialized and these findings were interpreted in context of other investigations dedicated towards studyi ng other CMP output va riables such as nonuniformity, dishing, erosion etc. Fig. 6.10 Variation of COF for Different To tal ASP Thickness for an Optimized Surface Coating Time of 40 Minutes 135 115 110 95 85 75 70 0 0.3 0.6 0.9 1.2 6080100120140 Thickness (mils) COF
165 70 75 85 95 110 115135 0 20 40 60 80 100 120 60708090100110120130140 Pad Total Thickness (mils)MRR (nm/min) Fig. 6.11 Variation of MRR for Different Total ASP Thickness for an Optimized Surface Coating Time of 40 Minutes 6.9 Summary of ASP Optimization and Characterization Polyolefin could be effectively used to replace the traditional polyurethane in CMP polishing pads. Polyolefin shows excelle nt adhesion when coat ed with a ceramic surface layer using PECVD to fine tune pad wafer hardness without the sub pad. Due to inherent resistance of polyolefin to slurry chemical and with the added chemical protection and hydrophilicity of TEOS, no c onditioning is required during entire pad lifetime. An elaborate pad metrology matrix ha s been developed to evaluate the polishing pad before being put in service. There does not seem to be any correla tion with pad static
166 tribological properties and dynamic tribological properties. The quality and durability of the surface can be judged by the surface pe netration of the nanoindentation tip. The materials removal rate being one of the vital CMP output variables, the pad surface TEOS coating time can be finalized to 40 minut es taking in to account the occurrence of plastic deformation of the surface under incide nt loading. The pad thickness for enhanced MRR was experimentally finali zed to 95 mils. When interpreted with the findings of the experiments performed to optimize other CMP ou tput variables such as dishing, erosion, non-uniformity, this methodology of pad metrol ogy would be useful for characterization and evaluation of different pads with novel architecture before their commercialization.
167 CHAPTER SEVEN INVESTIGATION AND MODI FICATION OF CMP SLURRY 7.1 Background of CMP Slurry CMP is a process that is influenced to a great extent by numerous slurry parameters such as pH, solution chemistry, charge type, concentration and size of abrasives, complexing agents, oxidizers, buffering agents, surfactants, corrosion inhibitors, etc. [56, 106, 211, 212]. The speci fic and proprietary na ture of the slurry manufacture makes it difficult to elucidate the ex act effects of slurry on the particular thin films that are polished in it. The slurry inter actions at the pad wafer interface are probably therefore, the least understood mechanisms in entire semiconductor fabrication process technology . An ideal CMP slurry should be able to achieve high removal rate, excellent global planarization, should prevent corrosion (in case of metal, especially Cu), good surface finish, low defectivity and high se lectivity. The typical design criteria for slurry are given in Fig. 7.1. These criteria have been broadly identified after survey of literature [111-117]. 7.2 Effect of CMP Slurry in Global Planarization As discussed in the previous section, the global planarization as a result of CMP process is one of the key outputs of the pr ocess. As suggested in Fig. 7.1 the slurry,
168 plays a key role in achieving gl obal planarization. To achieve the requisite level of global planarization without compromising on the removal ra te and producing a defect free wafer surface needs optimization of the slurry parameters. Fig. 7.1 Prime Criteria for Slurry Design 
169 The slurry parameters must be so optimized that the mechanical removal of the material is minimized as slurry de pending upon excessive mechanical removal produces high frictional forces and can thus damage the surface topography. Following are the some of the salient issues that must be considered before specific slurry design: 1) To minimize the frictional forces, the removal rate needs to be compromised and thus the process runs for a longer period of time, 2) also variation in local polishing pressure leads to variable removal rates within the wafer, which seriously compromises global planarization , 3)Excessive chemical et ching adversely affects surface planarity and induces defects on the surface such as corrosi on . Thus, the key to a good polishing step is achievement of synergy between chem ical etching and mechanical planarization with minimization of both the phenomena. Fo r this purpose there is a need for the formation of a passivation layer at the interface of the wafer and pad as seen in Fig. 7.2., 4) The passivation layer has to be thinner that the difference in the height between high and low regions in order to avoid within wafer non-uniformity . In case of Cu polishing, the formation of the passivation la yer is accelerated by oxi dizers such as H2O2, potassium ferricynate, ferric chloride, and ferric iodate and corrosion damage to the surface is prevented by corrosion inhibitors such as benzotriaz ole (BTA) . For tungsten, there is rapid formation of surf ace passivation layers due to the use of peroxygen compounds and stabilizing agents . The purpose of passivation layer in case of silica polishing is to soften the surf ace which is inherently hard. Maintenance of alkaline pH in most cases is sufficient  for appropriate passivation. During Ta polishing, formation of stable Ta2O5 helps in uniform removal of material from the surface .
170 To avoid numerous surface defects, the time to achieve the formation of thin passivation layer should be minimized. All these asp ects were taken in to account before the evaluation and design of novel slurries. Fig. 7.2 Schematic of Microscale an d Nanoscale Phenomena during CMP  7.3 Effect of Slurry on Removal Rate The chemical action of the slurry chem icals on the material, the mechanical abrasion of the particles on the polished materi als, interplay of the different complexing agents, oxidizers and corrosion inhibitors. Chemicals such are oxidizers and corrosion inhibitors vastly affect the reaction rate of th e slurry with similar particle nature, size and distributions. Fig. 7.3 shows the variation of the reaction rate of the different slurry components on Cu when the reaction kineti cs were studied using electrochemical
171 chronoamperometry [224, 225]. From the electr ochemical data, it can be seen from the figure that the surface rate ki netics reaches about 60 /s when Cu is immersed in DI H2O. The reaction rate increases to around 120 /s when 5% H2O2 is added. However upon addition of 10 mM BTA, the reacti on rate came down considerably. Fig. 7.3 Variation of Rate of Surface Laye r Formation in Cu with Different Slurry Chemistry  7.3.1 Effect of Slurry Chemistry on Removal Rate Fig. 7.4 Transient Electrochemical Chr onoamperometry Measurements of W  The surface reaction is not the only contributing factor for achievement of high removal rate during CMP. The time scale at which the passivation layer is formed
172 before the average time of successive particle interaction with the wafer for abrasion is also important to produce a defect free, fast CMP process. Fig. 7.4 shows the electrochemical chronoamperometry (potenti ostatic) analysis of Tungsten by first keeping the samples at cathodic potential to avoid surface oxidation and then Â“anodizingÂ” them. The generation of current, which corresp onds to surface reaction rate, is monitored on a millisecond scale as shown in Fig. 7.4. It can be seen form the figure, that Tungsten surface quickly passivates which is conducive for mechanical removal of the material. 7.3.2 Effect of Slurry Particle Size, Hardness and Concentration The generalized materials removal rate (MRR) for oxide has been modeling in the literature and be expresse d as shown in Eq.7.1 : MRR= n Volremoved (7.1) The variable n is number of active abrasives taki ng part in the process and Volremoved is the volume of material removed by each ab rasive. To estimate the total volume of material removed, it is necessary to estimate the total area of the padÂ–wafer and waferabrasive contact. The area of activ e abrasive contact is given by: A = x (7.2) where A is the area of contact x is diameter of abrasive and is the depth of indentation on the passivating film made by the abrasive particle .
173 If one assumes elastic contact betw een the particles and the surface, the indentation depth as a function of particle size is given by: (7.3) where is the particle size, K is the particle fill factor at the surface and E is the Young's modulus of the surface layer [101, 226, 227]. Th is equation assumes that the particles are much harder than the surface layer. (7.1), (7.2 ) and (7.3) show that the area of contact and indentation depth increase with increase in particle size and hardness. It is thus implied that as particle size and hardness increases the removal rate incr eases. The increase in particle concentration will increase the number of active particles, there by causing more number of indentations to th e passivating film and increasing the removal rate. Fig. 7.5 indicates the increase in removal rate of tungsten with increase in particle size and concentration. The details of the experiments can be obtained in the relevant literature . Fig. 7.5 Variation of Removal Rate w ith Particle Size and Concentration 
174 Increase in particle size or hardness al so gives rise to surface defects such as micro-scratches that cause fatal long-term device failure. Bigger and harder particles would cause deeper micro-scratches, which w ill be very difficult to eliminate even by the final buffing CMP step. The increase in particle concentration translat es in to increase in removal rate only up to a certain extent. As seen in Fig. 7.6 shown by Singh and Bajaj  and Mahajan et al. , the removal rate of the silica increases with increase in particle size and concentration at low particle concentration, however after a particular threshold for every given particle size the mechanism of removal changes and there is considerable decrease in removal rate with increase in particle concentration. For the purpose of this experiment, spherical monosized particles were used in slurry of pH 10. Change in material removal mechanism is e xpected to be the reason of this phenomenon . Fig. 7.6 Removal Rate of Silica with Differe nt Particle Size and Concentration  7.4 Effect of Temperature on CMP Slurry Performance We have studied the impact of slurry temperature on the CMP performance of the slurry. For this study, we maintained the slurry in a range of temperature from 18.5oC
175 to 30oC. Electroplated Copper blanket films of 15000 thick were polished using the CETRTM bench top CMP tester using a 6Â” diamet er polishing pad coupon attached to the bottom platen and a 1Â” X 1Â” sample coupon faced down onto the pad. Cabot iCue 5001 slurry and IC1000/Suba IV polyuret hane perforated pads were used to polish the copper samples. Polishing conditions were maintained at 3 PSI, 100 RPM bottom platen rotation, 95 RPM carrier rotation, slurry flow rate 50 ml/min. The slurry was maintained at the predetermined temperature (both above and be low room temperature) for the duration of the polishing experiment. Removal rates we re calculated from the in-situ endpoint detection ability of the machine, which shows a change of the COF at the complete removal of the thin film. The time for removal was noted, which gives the removal rate information. The variation of the removal rate along with COF with change in slurry temperature is plotted in Fig.7. 7. An overall increase in the removal rate with temperature can be observed from the figure. This increase of the removal rate can be attributed to changes in both mechanical and chemical nature of the polishing process. The change of the mechanical component of the process has two reasons: Firstly, as a result of polishing pad softening (change in m echanical properties of the pad due to the increase in temperature) the area of contact increases [ 228], thus increasing the number of abrasive particles coming in contact with the wafer surface. This is supported by the COF increase with increase in slurry temper ature as shown in Fig. 7.7. These results are in agreement with the results from previous investiga tions by other authors [228-230]. Secondly, a decrease of the viscosity of the slurry occurs with increasing temperature, which increases the friction at the interface and hence increases shear resulting in higher removal rates . Thus, increase in the re moval rate and friction with temperature can
176 be attributed to: 1) increase in reaction kinetics at the padwafer interface, 2) softening of polishing pad there by in creasing the contact area and he nce shear, 3) decrease in the viscosity of the slurry due increase in temperature . Fig. 7.7 Variation of COF and MRR w ith Increase in Slurry Temperature Fig. 7.8 Variation of ln of MRR in m/s w ith Inverse of Temperat ure (1/T) Showing an Arrhenius Relationship
177 It must also be mentioned that th e increase in MRR during CMP with increase in the slurry temperature is governed by an A rrhenius type relationship. The variation of ln of removal rate with temperature (1/T) is a straight line with a negative slope (Fig. 7.8). The slope of this curve can be used to determine the activation energy of the Cuslurry system. If the increase in reaction kinetics can be achieved at room temperature with the help of the chemical action of slurry additives, then there can be an enhancement in slurry performance without the correspondi ng increase in thermal overheads. For this purpose, we have a proposed a novel slurry design (chemistry) which used a surface catalyst to enhance the slurry performance of Cu slurry The details of this novel nanoparticle based slurry are discu ssed in the subsequent section. 7.5 Novel Nanoparticle based Cu slurry Compared to the other materials involved in the IC fabrication, (eg Silicon, Tungsten, Silica, etc.), Cu is much softer as compared to the abrasive particles used in the chemical active solution used (CMP Slurry) during the its polishing process. Controlling Cu polishing to meet the stringent demands of the semiconductor industry and producing Cu wiring without: 1) corrosion, 2) micro/ nano scratching, 3) over polish, pattern damage, trough formation, etc, 4) material dela mination, etc. is of paramount importance to make more sophisticated and cost effective chips of tomorrow. Due to the inherent soft nature of Cu, there are severe restrictions in obtaining the desired removal rate by purely mechanical means (increase in down pressure and platen rotation). There is a need to chemical enhance the performance of a CMP pr ocess step without significant increase in the applied surface shear to th e wafer. The proposed slurry promises to be helpful in
178 achieving the aforementioned demands of the semiconductor industry within the strict tolerances prescribed. Thus, this new technol ogy will facilitate in th e achievement of the widely published objectives of the semiconductor industry. Additionally this proposed slurry does not require any special conditions for operation. These and other advantages will significantly bring down the cost of ownership (CoO) of the CMP operations. The universal benefits of this technology include : 1) decrease in th e IC cost, 2) wide implementation of the IC, 3) more versatil ity and faster operati on the next current electronic products. 7.5.1 Composition of Novel Nanoparticle Based Cu Slurry The proposed slurry consists of colloi dal suspension of nanoparticle abrasives derived from Tetraethylorthosil icate (TEOS), its derivatives and any materials modified from TEOS, in a chemically active medium. Th e base solution of the slurry consists of deionized (DI) water, buffering agents lik e organic and inorganic buffers, cleansing agents, surface modified catalysts, and surface reagents. The chemically active medium is additionally comprised of acids such as HCl, H2SO4, HNO3, and other inorganic acids in different concentrations, an ti corrosive compounds like Be nzotriazole (BTA) ranging from 0.0001 to 4 % by volume, oxidizing agents such as H2O2, KMnO4 or other oxidizing agents ranging from 1 % to 8 % by volume. The details of the chemical composit ion of the slurry are as follows[233: 1)Base Solution: DI water, 2) Buffer: Inorganic Buffer (HCl, HNO3, H2SO4) pH: 5-7, 3) Nanoparticles Particles: TEOS based, 4) Size distribution: 80-120 nm, 5) Shape: Spherical, 6) Dispersion: Colloidal Suspension, 7) Oxidizing Agent: H2O2, KMnO4, (1-5
179 vol%), 8) Surfactants: TWEEN X 100 (0.01 to 1 vol %), 9) Concentration: 2-10 wt %, 11) Chelating Agents: Organic Acids and others 12) Corrosion Inhibitor: BTA (0.01 to 1 vol %) 7.5.2 Mechanism of Polishing of Novel Nanoparticle Cu Slurry The proposed slurry consists of simila r and variable sized TEOS nanoparticle abrasives ranging from 10 nm to 200 nm. The pa rticle concentration of the nanoparticles in the slurry solution ranges from 2 to 12 % by weight. The primary objective to the nanoparticle slurry is to preci sely control the desired CMP output variables and prevent defects such as: 1) high surface roughness, 2) micro/ nano scratching, 3) dishing, erosion or pattern damage, 4) delamination, 5) surf ace corrosion, pits and craters. The slurry solution is made of DI water, glycene and or ganic acids like acetic acid. The slurry buffer contains acids solutions, sp ecifically 25 to 75 vol. % HNO3 to optimize the reaction over the Cu surface. Other inorganic acids such as HCl and H2SO4 have also been explored to optimize the surface reactions. The pH of the slur ry can be maintained within the range of 2.0 to 10.0. The surface reactions are carried out in presence of catalysts such a TiO2, ZnO, SnO2, other metal oxides, ceramics, and elements and compounds derived and modified from them. Ionic and Anionic surf actants specifically TWEEN X100 have been used in the range of 0.001 to 4 vol. % to eliminate surface corrosion and impart hydrophobicity to the surface. To trigger a nd enhance the accele rated surface oxidation for effective material abrasion, oxidizing agents, specifically H2O2 and others such as KMnO4, NH4MnO4, Cr (NO3) 2 etc. have been used in the range of 1 to 8 vol. %. The slurry employs compounds such as KOH, KCl, and other organic and inorganic salts in
180 isolation and/or in combination for gene ration of active ions for maintaining and balancing the slurry pH. The aforementioned organic and inor ganic salts also serve the purpose of generating counter ions to prevent the diffusion of K+ ions in Cu due to their competitive reactions. The competitive reactions are also employed on a stand-alone basis or in series to minimize the heat genera ted as a result of the chemical action of the slurry on the Cu surface. This prevents the adverse process effects that may be caused due to pad surface heating and viscoelastic ma terial reflow which inevitably leads to pad compliance and loss in planarity. The mechan ism for polishing action of the CMP slurry has been illustrated in Fig. 7.9.
181 Fig. 7.9 Schematic Showing Mechanism of Polishing Action of N ovel Nanoparticle CMP slurry Cu surface layer Complex DI H2O Rinse Material Removal/End of CMP Step Physi-adsorbed Hydrophobic Surface Layer Clean Cu Surface Material Removed by 2/3 Body Contact Combined Particle Abrasion Formation of a Physi-adsorbed Hydrophobic surface layer by Complexing Agent binding O and N sites on the surface Surface Cleaning and Surface layer formed by Organic Solvent+ Complexing Agent Removal of remaining particles and surface hydrophobic layer yielding clean Cu surface by DI H2O rinse
1827.5.3 Polishing Performance of Nove l Nanoparticle Cu Slurry The CMP experiments using the nove l slurry were performed using Electroplated Copper blanket films of 15000 thick were polished using the CETRTM bench top CMP tester using a 6Â” diameter polishing pad coupon attach ed to the bottom platen and a 1Â” X 1Â” sample coupon faced down onto the pad. IC1000/Suba IV polyurethane perforated pads were used to polish the copper samples. Polishing conditions were maintained at 3 PSI, 100 RP M bottom platen rotation, 95 RPM carrier rotation, slurry flow rate 50 ml/min (polishing conditions were maintained similar to previous slurry temperature an alysis). Fig. 7.10 shows the variation of MRR for EP Cu for different trials performed without the addition of metal oxide surface catalysts. It can be seen from the figure that the MRR perfor mance is well below the industrial standards with the highest MRR obtained being 55 nm/min. Fig. 7.10 Variation of MRR for Different Tr ials Using Nanoparticle Slurry without the Surface Catalyst 0 10 20 30 40 50 60 12345 TrialMRR nm/min
183 Upon addition of the surface catalyst, there was a dramatic improvement in the slurry performance. Fi g. 7.11 shows the MRR during CMP for different trials performed with newly developed slurry with the metal oxide ceramic surface catalyst. There is at least a 100 % improvement in th e MRR upon addition of the surface catalyst. Fig. 7.11 Variation of MRR for Different CM P Trials for Nanoparticle Slurry with Surface Catalyst Fig. 7.12 Variation of COF for Different CMP Trials Using Novel Nanoparticle Based CMP slurry 0 20 40 60 80 100 120 140 12345 TrialMRR nm/mi n 0 0.05 0.1 0.15 0.2 0.25 12345 TrialCOF
184 Fig. 7.13 Variation of COF with Time Measur ed In-situ for Polishing of EP Cu Samples Under 3 PSI Down Pressure and 100 RPM Plat en Rotation Using a) Novel Nanoparticle Cu Slurry, and b) Commercial Cu Slurry Another salient feature of this slurry is that there rela tively low amount of shear generated during polishing. The MRR mechanism being predominantly chemical 050100150200250 Time sec -0.3 -0.2 -0.1 -0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 AE Volt 0 0.05 0.10 0.15 0.20 0.25 0.30 0.35 COF-Ff AE -0.0146 COF-Ff 0.1960 a b
185 ensures that the contribution of shear remains at its bare minimum. Also, due to the competing reactions of Cu2+ ions and K+ ions, there is effective exchange of heat at the pad wafer interface. This does not cause the polishing pad to soften and then excessively comply with the wafer topography. The e nhancement of MRR without excessive pad compliance will have benefits in terms of enhanced global planarization, extended pad life. Fig. 7.12 shows the variation of COF fo r different trials of the slurry for which removal rate have been reported in Fig. 7.11. There is a significant different between the COF values obtained when the developed slurry when compared to an industrial slurry when used in polishing under the same conditions as seen from Fig. 7.13 a, b. 7.6 Summary of Investigation and Modification of CMP Slurry The CMP slurry has a delicate balance of different chemicals which significantly affect the CMP process output. In case of metal polishing the reaction occurring at the pad wafer interface are highl y dominated by the chemical composition of the CMP slurry. The most significant attributes of the CMP slurry that affect CMP output are: 1) pH, 2) particle shap e, 3) particle hardness, 4) particular concentration and distribution. The increase in temperature of slurry signi ficantly increases the reaction kinetics, there by increasing th e removal rate. However the adverse effects of slurry heating are: 1) decreased pad wear life, and2 inhibition of global planarization. For this purpose we have leveraged the phenomenon of surface catalysis in the CMP slurry which enables the slurry to predominantly em ploy chemical means for material removal. There is a lower COF observed during the ope ration of this slurry which means lower surface shear during CMP. This property of the slurry is beneficial as lower surface shear
186 automatically translates in: 1) lower pad wear 2) reduced defects. It is predicted that predominantly chemically acting slurries will be used in the future as Cu and low k ILD materials are incorporated.
187 CHAPTER EIGHT SUMMARY 8.1 Summary of the Research Chemical Mechanical Polishing being the process of choice for achieving local and global planarization on the surface of the wafer, needs to be extensively studied in order to deal with certa in teething challenges presente d by it. In situ CMP process monitoring is one of the methods by CMP pr ocess can be analyzed and controlled. Monitoring the COF during the CMP process ca n give indications of the process end point and slurry selectivity in situ. The ac oustic emission signal can detect delamination during the process. These two signals if em ployed in conjunction together with a feed back loop can ensure significant reduction in CMP process defectivity along with a reduction in cost of ownership. The proce ss CMP needs to be studied by taking in to account the properties of the materials that und ergo CMP, the polishing pads used for it, at the same time by evaluating the slurry that enables preferential removal of materials. The mechanical and tribological properti es of materials significantly impact their CMP performance. Certain CMP attributes such as wear mechan ism, defectivity are based upon the mechanical properties of th e materials undergoing CMP. There is evidence that pre CMP treatment of thin films undergoing CMP can allow higher freedom for developing CMP process fo r some materials such as Cu.
188 Previous reports of evaluation of CMP polishing pads were based upon evaluation of materials that constitute the pol ishing pad. The ultra sound testing, which is a non destructive polishing pad evaluation tec hnique is explored to evaluate the non uniformity in the pad specific gravity. Howe ver, the polishing pad a whole may contain certain non uniformities in their specific gr avity which may not be observed if small coupons of the polishing pads are evaluated. When attempts were made to isolate the regions of non uniformity of the polishing pad, the coupons showed bulk properties of constituent materials. Thus, it can be said th at the polishing pad as a consumable needs to be evaluated as whole without breaking its integrity Surface modification of the polishing pad is a feasible and encouraging direction for development of novel polishing pads. When coated with a ceramic, the polishing pad surface remains insula ted from the slurry chemical attack. Therefore, there in no need for conditioning of the polishing pa d during operation. The static tribological properties and the mechanical properties do no correlate very well w ith the pad polishing performance. Hence, the CMP evaluation of the polishing pad cannot be predicted using these statically evaluated pad propertie s. The increase in polishing pad surface mechanical properties with increase in polis hing pad coating time br ings about a decrease in material removal rate. The reason bei ng this counter intuitive phenomenon is: 1) decrease in pad surface complian ce, 2) increase in slurry film thickness on the surface of the pad due to decrease in the pad surface r oughness. The impact of pad foam thickness and slurry film thickness can be gauged by experimentally evaluating polishing pads with different total thickness.
189 The advent of Cu and ultra low k dielect rics, it is crucial to perform CMP in a low force and low linear velocity regime (low stress). The slurry which is a combination of chemically active agents as well as m echanically acting abrasives is crucial in maintaining the optimum and delicate balance of chemical and mechanical component of CMP. The slurry chemical attack and temperat ure determines the react ion kinetics at the pad wafer interface. The increase in meta l oxidation and dissolution enhances the material removal rate during CMP. This e nhancement in the oxidation and dissolution at the pad wafer interface can also be done usi ng a surface catalyst such as metal oxides in the slurry. The use of surface catalysts to enhance slurry performance increases the chemical component during the CMP process. Hence enhanced material removal rate can be obtained at lower surface shear. Finally, understanding the process aspect s, material properties, polishing pad behavior and slurry performance gives a holisti c perspective of the pr ocess of CMP. This could be treated a contribution towards transf orming CMP in to more of a science from the present stage, where in it is more of an art. 8.2 Major Findings and Contributions The process of CMP has been extensivel y investigated during this research. As mentioned before, the process of CMP being a complex interplay of different variables, it is very important to isolate th e each of them and understand their impact on the process output. This research has been an effort to isolate some of the variables and study their specific characteristics and gauge their impact on CMP. The process of understanding CMP be gan my monitoring the polishing runs
190 in-situ. Though there are some techniques avai lable to monitor the process end point, insitu monitoring of polishing behavior of diffe rent materials and consumables has not been extensively done before. Thus findings in th is area identify a range in which CMP outputs such as MRR, End Point, slurry sele ctivity for a combination of polishing pad, slurry, material, as well defects such as dela mination, at a given po lishing condition. This in-situ monitored data also served as a starti ng point for further investigations in area of heuristic training of the CMP apparatus to en gineer a feed back mechanism for accurate process control. This research also helped elucidate wear mechanis ms of novel materials such as cross linked polymers and doped oxides used as ILD. The findings of the research were then extended in three different directions namely: 1) Pre-CMP material evaluation and improvement: The significance of annealing over the interface of the thin films in the damascene structure was investigated and correlation of the interfacial adhesion energy with standard process data available was made. 2) Investigation of Polishing Pads: N ovel application specifi c polishing pads that do not require conditioning was evaluated for the first time ever during this research. The contributions were made to the architecture of these polishing pads which end up being a commercial product.Though evalua tion of application specific pads was done on 6 inch coupons, some of the findings on commercial po lishing pads suggested that polishing pad needs to be evaluated as a consumable on the whole. If the integrity of the polishing pad was broken, the properties eval uated are those of the material that constitutes the polishing pad and do not give an idea of the pad CMP performance in its entirety. The findings of the experiments on novel pads were always implemented with this in mind. 3) Novel Slurry Formulation: The CMP consum able market is booming and expected to
191 soar. With newer and softer materials being implemented in the IC, CMP slurry with a predominant chemical component was formulat ed during this resear ch. The approach of using a surface catalyst to enhance CMP slurry performance was used for the first time in this research. The research thus makes a scientific a nd technical contribution to the field of CMP, and adds its two cents in transforming the process of CMP from being more of an art to being more of a science. 8.3 Future Trends in CMP and Potential Areas for Investigation There are several innovations and modificat ions such as slurry free approach, low down force polishing, abrasive less and nano particle slurry approach, etc. that are being carried out in the CM P process. Process such as reverse electroplating and combination of different planarization process, are also trying to compete with CMP for achievement of effective global and local planarization of the wafer. The process of chemical mechani cal polishing is also finding increasing application in the field of giant magneto re sistive (GMR) and colossal magneto resistive (CMR) disc drives for polishi ng successive layers of thin films (Co and Ni). Special emphasis is laid on successful endpoint detection and selec tivity of the slurry. CMP is used to polish multi-level thin film structure of the drives. The field of Microelectromechanical systems (MEMS) is also increasing adapting the process of CMP. As MEMS st ructures implement smaller and smaller features as constituents, the planarity of the thin films becomes an important issue. For optical MEMS applications the mirror like smoo th surface of the thin films is of utmost
192 importance for reliable and repeatable functi oning of the device, for example, optical features. Metal and highk insulators have replaced the dummy gates, which were used to preserve self-alignment of gate electrode. These metal and insulator layers need to be planarized with utmost control, where CM P has to play a significant role. Deposition techniques for such metal and highk insulator films are yet to be defined and CMP of films deposited by such atomic layer depositions need to be investig ated to get optimum polishing performance. Also, CMP of noble metals, which are used to make gate electrodes in p-channel devices need to be investigated to achieve optimum removal performance.
193 REFERENCES 1. J.G. Ryan, R.M. Geffken, N.R. Poulin and J.R. Paraszczak, IBM J. Res. Dev. 39, 371 (1995). 2. C.L. Borst, W.N. Gill and R.J. Gutmann, ChemicalÂ–Mechanical Polishing of Low Dielectric Constant Polymers and Organosil icate Glasses: Fundamental Mechanisms and Application to IC Interconnect Technology, Kluwer Academic Publishers, Boston (2002). 3. Y. Taur, Y.-J. Mii, D.J. Frank, H.-S. Wong, D.A. Buchanan, S.J. Wind, S.A. Rishton, G.A. Sai-Halasz and E.J. No wak, IBM J. Res. Dev. 39 245 (1995). 4. Internet web site: http://www.intel.com /labs/features/si11032.htm, Intel, 2002. 5. J. Givens, S. Geissler, O. Ca in, W. Clark, C. Koburger and J. Lee, Proceedings of the 11th International VLSI Multilevel Interc onnection Conference, 43, (1994). 6. P. B. Zantye, A. Kumar and A. K. Sikder, Mat. Sci. Eng. R 45, 3-6, 89, (2004). 7. A.K. Bates, M. Rothschild, T.M. Bloo mstein, T.H. Fedynyshyn, R.R. Kunz, V. Liberman and M. Switkes, IBM J. Res. Dev. 45, 605 (2001). 8. F. Bernard, Micro. Eng. 61Â–62, 11 (2002). 9. B. Michel, A. Bernard, A. Bietsch, E. Delamarche, M. Geissler, D. Juncker, H. Kind, J.-P. Renault, H. Rothuizen, H. Schmid P. Schmidt-Winkel, R. Stutz and H. Wolf, IBM J. Res. Dev. 45, 5, 697 (2001).
194 10. S. Wolf and R. Tauber, Silicon Processing for VLSI Era, McGraw-Hill, New York (1991). 11. R.W. Keyes, Proc. IEEE 69, 267 (1981). 12. S.P. Murarka, Metallization: Theory and Practice for VLSI and ULSI, ButterworthHeinemann, Boston (1993). 13. J.A. Lee, M. Moinpour, H.-C. Liou and T. Abell, Proceedings of Materials Research Society San Francisco, CA (2003), p. F7.4. 14. International Technology Road map for Semiconductors, 2001. 15. K. Shenai, VLSI Metallization: Physic s and Technologies, Artech House Inc. (1991). 16. V.K. Tripathi and R.J. Bucolo, IEEE Trans. Electron Dev. 34, 3, 650 (1987). 17. G. Leibmann, Br. J. Al. Phys. 1, 4, 92 (1950). 18. R.F. Harrington and C. Wei, IEEE Trans. Microwave Theory Tech., 32, 705 (1984) 19. P. Wolters, Internet web site: http:// www.peterwolters.com/cmp/cmpmultilevel.htm, (2003). 20. W.-S. Lee, S.-Y. Kim, Y.-J. Seo and J.-K. L ee, J. Mater. Sci.: Mater. Electron. 12, 1, 63 (2001). 21. W.J. Patrick, W.L. Guthrie, C.L. Standley and P.M. Schiable, J. Electrochem. Soc. 138, 555 (1991). 22. S. Wolf, Silicon Processing for VLSI Era II, Process Integration 45, McGraw-Hill, New York (1991). 23. P. Sallagoity, F. Gaillard, M. Rivoire, M. Paoli and M. Haond, Microelectron. Reliab. 38, 271 (1998).
195 24. S.-Y. Jeong, S.-Y. Kim and Y.-J. Seo, Microelectron. Eng. 66, 480 (2003). 25. S.-Y. Kim, C.-I. Kim, E.-G. Chang, Y.-J. Seo, T.-H. Kim and W.-S. Lee, Proceedings of 4th IUMRS-ICEM-98 Ch eju Korea, 24Â–27 August 1998, 468 (1998) 26. S.-Y. Kim and Y.-J. Seo, Microelectron. Eng. 60, 2-4, 357 (2002). 27. Y.-J. Seco, S.-Y. Kim and W.S. Lee, Mech. Eng. 65, 4, 371 (2003). 28. Y.-J. Seo, G.-U. Kim and W.S. Lee, Microelectron. Eng. 71, 209 (2004). 29. C. Chen, J.-Y. Wu, W. Lur, US Patent 6,169,012 (2001). 30. C. Chen, J.-Y. Wu, W. Lur, US Patent 6,486,040 (2002). 31. C. Chen, J.-Y. Wu, W. Lur, US Patent 6,448,159 (2002). 32. D.G. Thakurta, D.W. Schwendeman, R.J. Gu tmann, S. Shankar, L. Jiang and W.N. Gill, Thin Solid Films 414, 78 (2002). 33. C. Zhou, L. Shan, J. Robert Hight, S. H. Ng and S. Danyluk, Wear 253, 3-4, 430 (2002). 34. X. Li, T.K.S. Wong, Rusli and D. Yang, Diamond Relat. Mater. 12, 3-7, 963 (2003). 35. G. Schindler, G. Steinlesberger, M. E ngelhardt and W. Steinhgl, Solid State Electron. 47, 7, 1233 (2003). 36. R.F. Schnabel, D. Dobuzinsky, J. Gambino, K.P. Muller, F. Wang, D.C. Perng and H. Palm, Microelectron. Eng. 37Â–38, 59 (1997). 37. L. Shon-Roy, Integrated Circuit E ngineering Corporation Handbook, 1999. 38. K. Kobayashi, Mater. Sci. Eng. 98, 2-15, 181 (2003). 39. K. Kobayashi, J. Phys. Chem. Solids 59, 9, 1671 (1998). 40. K. Kobayashi and I. Mizushima, Mater. Sci. Eng. B 39, 224 (1996).
196 41. K. Kobayashi and H. Sasaki, Mater. Res. Bull. 33, 5, 811 (1998). 42. K. Kobayashi, J. Non-Cryst. Solids 217, 2-3, 323 (1997). 43. J. Levert, S. Mukherjee and D. De bear, Semicon, Japan (1999). 44. J.-H. Zhao, I. Malik, T. Ryan, E.T. Ogaw a, P.S. Ho, W.-Y. Shih, A.J. McKerrow and K.J. Taylor, Appl. Phys. Lett. 74, 7, 944 (1999). 45. M.P. Petkov, M.H. Weber, K.G. Lynn, K.P. Rodbell and S.A. Cohen, J. Appl. Phys. 86, 6, 3104 (1999). 46. A. Kuntman, R. Yenidnya, A. Ka gz and H. Kuntman, Mi croelectron. J. 30, 127 (1999). 47. A. Maddalon, K. Barla, E. Denis, E. Lous, E. Perrin, S. Lis, C. Lair and E. Dehan, Microelectron. Eng. 50, 1-4, 33 (2000). 48. L. Trabzon and O.O. Awadelkarim, Mi croelectron. Eng. 65, 4, 463 (2003). 49. C.T. Chua, Y.-P. Lee, M.S. Zhou, L. Chan, US Patent 6,121,130 (2000). 50. A.E. Braun, Semicond. Int. 11 May 2003. 51. M. Khler, Etching in Microsystem Technology, W iley, Weinheim (1999). 52. M. Esashi, Microsyst. Technol. 1, 2 (1994). 53. K. Suzuki, S. Hiraoka and S. Nish imatsu, J. Appl. Phys. 64, 3697 (1988). 54. D.M. Knotter, J. Am. Chem. Soc. 122, 18, 4345 (2000). 55. AT&T Bell Laboratory, Solid St ate Technol. 37, 12, 26 (1994). 56. J.M. Steigerwald, S.P. Murarka, and R. J. Gutmann, Chemical Mechanical Planarization of Microelectronic Ma terials, Wiley, New York (1997). 57. D.J. Stein, J.L. Cecchi and D.L. Heth erington, J. Mater. Res. 14, 3695 (1999).
197 58. H. Landis, P. Burke, W. Cote, W. Hill, C. Hoffman, C. Kaanta, C. Koburger, W. Lange, M. Leach and S. Luce, Thin Solid Films 220, 1 (1992). 59. Internet web site: http://www.memc.com/MEMC.nsf/0/6D86FAED316FE14686256AE80058166D/$ File/Ultraflat_wafers.pdf?OpenElement, MEMC, 2003. 60. Internet web site: http://www.komsil.co.j p/en/office/hist.htm, Komatsu, (2003). 61. Internet web site: http://www.research.ibm.com, IBM, (2004). 62. Internet web site: http://www.sematech.org/corporate/index.htm, Sematech Inc., (2004). 63. Internet web site: http://www.mitsubishi.or.jp/e/monitor/mon itor_old/97version/monitor9802/wafer.html, Mitsubishi, (2004). 64. A.K. Sikder, P. Zantye, S. Thagella, A. Kumar, B. Michael Vi nogradov, N.V. Gitis, Proceedings of 8th CMPMIC Conference, 120, 2003. 65. S.H. Li, R. Miller, Chemical Mechanical Polishing in Silicon Processing, vol. 63, Academic Press, New York, 2000, 66. J. Hernandez, P. Wrschka, Y. Hsu, T.-S Kuan, G.S. Oehrlein, H.J. Sun, D.A. Hansen, J. King and M.A. Fury, J. Electrochem. Soc. 146, 4647 (1999). 67. G.B. Basim, J. Colloid Interface Sci. 256, 1, 137 (2002). 68. Internet web site: http://www.pall.co m/catalogs/microelec/cmp, Pall, 2003. 69. Internet web site: http://www.amat.c ome, Applied Materials Inc., 2003. 70. W.J. Patrick, W.L. Guthrie, C.L. Standley and P.M. Schiable, J. Electrochem. Soc. 138, 555 (1991).
198 71. W.L. Guthrie, W.J. Patrick, E. Levine, H.C. Jones, E.A. Mehter, T.F. Houghton, G.T. Chiu and M.A. Fury, IBM J. Res. Dev. 36, 845 (1992). 72. Internet web site: http://www.sea.rl .ac.uk/oldsea/oldpubs/Ale/desc.htm, Semiconductor Equipment Assessment, 2002. 73. D.G. Thakurta, C.L. Borst, D.W. Schwendeman, R.J. Gutmann and W.N. Gill, Thin Solid Films 366, 1-3, 181 (2000). 74. W. Li, D.W. Shin, M. Tomozawa and S.P. Muraka, Thin Solid Films 270, 601 (1995). 75. T.F.A. Bibby, J.A. Adams, K. Holland, G.A. Krulik and P. Parikh, Thin Solid Films 308Â–309, 1-4, 538 (1977). 76. B.J. Hooper, G. Byrne and S. Galligan, J. Mater. Process. Technol. 123, 1, 107 (2002). 77. H. Lu, B. Fookes, Y. Obeng, S. Machinski and K.A. Richardson, Mater. Charact. 49, 1, 35 (2002). 78. B. Mullany, J. Mater. Process. Technol. 132, 1Â–3, 10, 28 (2003). 79. Internet web site: http://www.sfamipec. com/product/polish_sing.html, Speed Fam, 2004. 80. T.-H. Tsai and S.-C Yen, Appl. Surf. Sci. 210, 3-4, 190 (2003). 81. S.-Y. Kim, C.-J. Park and Y.-J. Seo, Microelectron. Eng. 66, 1-4, 472 (2003). 82. J.M. Boyd, H.E. Litvak, US Patent 6,612,902 (2003). 83. C.-H. Lu, M.-Y. Li, US Patent 6,307,628 (2002). 84. N. Saka, J. Nam, H.L. Oh, US Patent 6,476,921 (2001). 85. A.K. Sikder, private communication, 2003.
199 86. A.K. Sikder, private communication, 2002. 87. F.-S. Lee, C.-C. Chen, J.-C. Lee, H.-C Huang, US Patent 6,503,124 (2003). 88. T. Shih, C.-H. Yu, US Patent 6,682,396 (2004). 89. S.R. Roy, US Patent 6,547,652 (2003). 90. E.C. Wilson, H.A. Anderson, G. Appel, US Patent 6,149,512 (2000). 91. Internet web site: http://www.lamrc. com, Lam Research Inc., 2001. 92. N. Shendon, US Patent 5,899,800 (1999). 93. N. Shendon, D.R. Smith, Patent 5,582,534 (1996). 94. T. Laursen and M. Grief, J. Elect ron. Mater., 31, 10, 1509, (2002). 95. L.P. Lai, J.L. Tucker, R.J. Lujan, US Patent 6,224,472 (2001). 96. Internet web site: http://www.nikon.co .jp/main/eng/products/cmp_e.htm, Nikon Corporation, 2003. 97. Infenion Corporation, http://www.infeni on.com/, Infenion Corporation, 2003. 98. O. Kuhn, H. Drummer, I. Dresden and F. Stanke, Proceedings of the 2nd European AEC/APC Conference, 308, (2001). 99. F. Preston, J. Soc. Glass Tech. 11, 214 (1927). 100. C. Rogers, L. Racz, J. Coppeta, A. Philipossian and F. Kaufman, J. Electrochem. Soc. 147, 5, 1903 (2000). 101. R.K. Singh and R. Bajaj, MRS Bull. 27, 10, 743 (2002). 102. A.K. Sikder, F. Giglio, J. Wood, A. Kumar and J.M. Anthony, J. Electron. Mater. 30, 1522 (2001). 103. H. Lu, H. Lu, Y. Obeng and K.A. Richardson, Mater. Charact. 49, 2, 177 (2002). 104. B.-S. Chen, J.S. Steckenrider, B.L. Mueller, US Patent 6,464,562 (2002).
200 105. J.-C. Twu, Y.-H. Chen, T. Shih, S.-M. Jang, US Patent 6,589,872 (2003). 106. S. Kondo, N. Sakuma, Y. Homma, Y. Go to, N. Ohashi, H. Yamaguchi and N. Owada, J. Electrochem. Soc. 147, 3907 (2000). 107. M. Hanazono, J. Amanokura and Y. Kamigata, MRS Bull. 27, 10, 722 (2002). 108. W.R. Morrison, K.P. Hunt, US Patent 5,938,505 (1999). 109. S.-Y. Kim, K.-J. Lee and Y.-J. Seo, Mi croelectron. Eng. 66, 1-4, 463 (2003). 110. S.-Y. Kim, J.-H. Lim, C.-H. Yu, N.-H Kim and E.-G. Chang, KIEEME Trans. Electr. Electron. Mater. 4, 2, 1 (2003). 111. J. Reynolds, A. L. Swecker, A. J. Stro jwas, A. Levy and B. Bell, Solid state technology 39, (1998) 112. T. Park, T. Tugbawa and D. Boning, Pr oc. CMP-MIC, Santa Clara, CA, March, 196, (2000). 113. M. Oliver, Ed., Chemical Mechanical Pl anarization of Semiconductor Materials, Springer, NY, 2004. 114. L.M. Cook, J. Non-Cryst. Solids 120, 152 (1990). 115. David Evans, Sharp Labs of America, Camas, WA, Private Communication, (2005). 116. T. K. Das, R. Ganesan, A. K. Sikde r, A. Kumar, IEEE Transactions on Semiconductor Manufacturing, Accepted fo r publication (U.S. Patent pending). 117. A. K. Sikder and A. Kumar, M. Vinogra dov and N. V. Gitis, Delamination studies on Cu-low-K stack, to be submitted
201 118. R. Ganesan, T. K. Das, A.K., Sikde r, A. Kumar, IEEE Transactions on Semiconductor Manufacturing 16, 4, 677 ( 2003);(U.S. Patent # 60/521,357 filed on April 8th 2004, Approval pending). 119. S. Lagrange, S.H. Brongersma, M. Judelewicz ,, A. Saerens, I. Ve rvoort, E. Richard, R. Palmans and K. Maex, Microelectron. Eng. 50, 449-457 (2000). 120. J. M. Harper, C. Cabral, Jr ., P.C. Andricacos, L. Gignac, I.C. Noyan, K. P. Rodbell and C.K. Hu, J. Appl. Phys. 86, 2516-2525 (1999). 121. A. Furuya, Y. Oshita and A. Ogura, J. Vac. Sci. Technol. A, 18, 2854-2857 (2000). 122. C. Lingk, M.E. Gross, and W.L. Brow n, J. Appl. Phys. 87, 2232-2236 (2000). 123. S.H. Brongersma, E. Richard, I. Vervoort, H. Bender, W. Vander vorst, S. Lagrange, G. Beyer and K. Maex, J. Appl. Phys. 86, 3642-3645 (1999). 124. L. Vanasupa, Y-C. Joo, P.R. Besser a nd S. Pramanick, J. Appl. Phys., 85, 25832590 (1999). 125. C-Y. Yang, J.S. Jeng and J.S. Chen, Thin Solid Films 420-421, 398-402 (2002). 126. M.T. Perez-Prado and J.J. Vlassa k, Scripta Mater. 47, 817-823 (2002). 127. H. Lee, S.S. Wong and S.D. Lopa tin, J. Appl. Phys. 93, 3796-3804 (2003). 128. K. Ueno, T. Ritzdorf and S. Grace, J. Appl. Phys., 86 4930-4935 (1999). 129. C. Lingk, and M.E. Gross, J. Appl. Phys., 84, 5547-5553 (1998). 130. W.H. The, L.T. Koh, S.M. Chen, J. Xie, C.Y. Li and P.D. Foo, Microelectron. J., 32, 579-585 (2001). 131. J. Proost, T. Hirato, T. Fu ruhara, K. Maex and J.-P. Celis, J. Appl. Phys. 87, 27922802 (2000). 132. T.-H. Fang and W.-J. Chang, Micr oelectron. Eng. 65, 231-238 (2003).
202 133. S.H. Kang, Y.S. Obeng, M.A. Decker, M. Oh, S.M. Merchant, S.K. Karthikeyan, C.S. Seet and A.S. Oates, J. Electron. Mater., 30, 1506-1512 (2001). 134. J. Ye, N. Kojima, K. Ueoka, J. Shimanuki, T. Nasuno, and S. Ogawa, J. Appl. Phys. 95, 3704, (2004). 135. P.G. Charalambides, J. Lund, A. G. Evan s and R. M. McMeeking, J. Appl. Mech. 56, 77, (1989). 136. P.G. Charalambides and A.G. Evans, J. Am. Ceram. Soc. 72, 746, (1989). 137. R. H. Dauskardt, M. Lane, Q. Ma and N. Krishna, Engg Fract Mech 61, 1411, (1998). 138. A. A. Volinksy, N. R. Moody and W. W. Gerberich, Acta mater. 50, 441, (2002). 139. N. X. Randall and R. Consiglio, Rev. Sci. Instrum. 71, 2796 (2000). 140. T Homma, Mater. Sci. Eng. R23, 243 (1998). 141. D.W. Gidley, W.E. Frieze, T.L. Dull, J. Sun, A.F. Yee, C.V. Nguyen and D.Y. Yoon, Appl. Phys. Lett. 76, 10, 1282 (2000). 142. L. Peters, Semicond. Int., 108, (2000). 143. L. Pauling, The Nature of the Chemical B ond (third ed.), Cornell University Press, Ithaca, NY (1960). 144. J.W. Yi, Y.H. Lee and B. Farouk, Thin Solid Films 374, 103 (2000). 145. R.K. Laxman, N.H. Hendricks, B. Arkles and T.A. Tabler, Semicond. Int. (2000). 146. A.K. Sikder, A. Kumar, S. Thagella and J. Yota, J. Mater. Res., 19, 4, 996 (2004). 147. E. Korczynski, Solid State Technol. 39, 4, 63 (1996). 148. M.R. Baklanov, M. Muroyama, M. Judelewi cz, E. Kondoh, H. Li, J. Waeterloos, S. Vanhaelemeersch and K. Maex, J. Vac. Sci. Technol. B17, 5, 2136 (1999).
203 149. X. Xiao, N. Hata, K. Yamada and T. Kikkawa, Rev. Sci. Instrum. 74, 10, 4539 (2003). 150. D. Shamiryan, T. Abell, F. Lacopi and K. Maex, Mater. Today 7, 1, 34 (2004). 151. M.R. McGurk and T.F. Page, J. Mater. Res. 14, 2283, (1999) 152. C. Taylor, R. Prince, A.P. Malshe, L. Ries ter, G.J. Salamo and S.O. Cho, Proc. SPIE Int. Soc. Opt. Eng. 4936, 424, (2002). 153. J.-H. Ahn and D. Kwon, Ma ter. Sci. Eng. A285, 172, (2000). 154. N.X. Randall and A. Harris, Wear 245, 1-2, 196, (2000). 155. G.M. Pharr, Mater. Sci. Eng. A253, 151, (1998) 156. W.C. Oliver and G.M. Pharr, J. Mater. Res. 7, 1564, (1992). 157. S. Thagella, A. K. Sikder, and A. Kuma r, J. Electrochem. Soc. 151, 3, 205 (2004) 158. K. Ludema, Friction, Wear, Lubrication: A Text Book in Tribology, p. 90, CRC Press, Boca Raton, (1996). 159. P. B. Zantye, A. Kumar, S. Ostapenko, A. Sikder, M. Dallas, Submitted to the J. Vac. Sci Tech. A, (2005). 160. A.K. Sikder, I.M. Irfan, A. Kumar and J.M. Anthony J. Electron. Mater. 30, 12, 1527, (2001). 161. Maria Ronay and Charles L. Rohn, Tribol. Lett. 14, 3, 205 (2003). 162. Maria Ronay, J. Electrochem. Soc.151, 12, G 847 (2004). 163. D.O. Ouma, Modeling of Chemical M echanical Polishing for Dielectric Planarization, Ph.D. Thesis, Massachuse tts Institute of Technology, USA, (1999). 164. K. Qin, B. Moudgil and C.W. Park, Thin Solid Films 446, 2, 277 (2004). 165. C. Hepburn, Polyurethane Elastomers, A pplied Science Publishers, London, (1982).
204 166. K.L. Johnson, Contact Mechanics, Cambridge University Press, Cambridge (1985). 167. M. Moinpour, A. Tregub, A. Oehler and K. Cadien, MRS Bull. 27, 10, 766, (2002). 168. P. B. Zantye, S. Mudhivarthi, A.K. Sikder, A. Kumar and Y. Obeng, Mater. Res. Soc. Symp. Proc. 816, K 5.6 (2004). 169. A. K. Sikder, I. M. Irfan, Ashok Kumar, A. Belyaev, S. Ostapenko, M. Calves, J. P. Harmon and J. M. Anthony, Mat. Res. Soc. Sym. Proc., 670, M 1.8, (2001) 170. D. G. Totzke, Ultrasound diagnostic of ch emical mechanical planarization pads, MS Thesis, University of South Florida, Tampa, FL, May 2000. 171. D. G. Totzke, A. Belyaev, W. Moreno, S. Ostapenko, I. Tarasov, W. Easter, A. Maury, and A. Crevasse, AIP Conf. Proc. 550, 1, 259 (2001). 172. I. Lack, I. Krupa, M. Stach, A. Ku ma, J. Jur iov and I. Chodk, Polym. Test. 19, 7, 755 (2000). 173. L. Woo, M. T. K. Ling and S. P. We stphal, Thermochim. Acta 272, 171 (1996). 174. F. Podczeck and S. M. Almeida, Eur. J. Pharm. Sci. 16, 3, 209 (2002). 175. A. Philipossian and S. Olsen, J. Electrochem. Soc 151, 6, G 436 (2004). 176. F. Podczeck and S. M. Almeida, Eur. J. Pharm. Sci. 16, 3, 209, (2002). 177. D. Castillo-Mejia, J. Kelchner and S. B eaudoin, J. Electrochem. Soc. 151, G 271, (2004). 178. C.-Y. Chen, C.-C. Yu, S.-H. Shen and M. Ho, J. Electrochem. Soc. 147, 3922, (2000). 179. Y.-R. Jeng, P.-Y. Huang and W.-C. Pan, J. Electrochem. Soc. 150. G630, (2003). 180. S. Machinski, Wear characterization of polyurethane chemical mechanical polishing pads, MS Thesis, University of Central Florida, FL, Summer, (2001).
205 181. A. S. Lawing, Proceedings of the Mat. Res. Soc. Symp. Proc., Vol. 732E, I 5.3.1. (2002), 182. A.S. Lawing, Proceedings of the Sevent h International Chemical Mechanical, Planarization for VLSI/ULSI Interconnecti on Conference (CMP-MIC), San Jose CA, Feb. 27-March 1, 310,(2002). 183. A.S. Lawing, Proceedings of the Fifth International Symposium on Chemical Mechanical Planarization, Phila delphia, PA, May 12-17, (2002). 184. R. K. Singh, S.-M. Lee, Low Stress Po lishing of Copper / Low-K Dielectric Structures, Proceedings of the 8th CMP-IMC Conference, Marina Beach Marriott Hotel, Marina Del Rey, CA, February 19-21, 129, (2003). 185. R. Bajaj, M. Desai, R. Jairath, M. Stell and R. Tolles, Mater. Res. Soc. Symp. Proc. 337, p. 637, (1994). 186. K. Nakamura, S. Kishii and Y. Arimoto Jpn. J. Appl. Phys. 36, 1525, (1997). 187. T. Gan, Modeling of chemical mechanical polishing for shallow trench isolation. MS thesis, Massachusetts Institute of Technology, Cambridge (MA), May 2000 188. S.-H. Chang, Microelectronic E ngineering, 77, 1, 76, (2005). 189. R. Saxena, D. G. Thakurta, R. J. Gutmann and W. N. Gill Thin Solid Films, 449, 12, 192, (2004). 190. D.S. Rickerby, S.J. Bull, Engineering w ith surface coatings: The role of coating microstructure, Surf. Coat. & Techno logy, Volume 39/40, (1989), pp. 315. 191. A. G. Sirota, Polyolefins: modification of structure and properties, Israel Program for Scientific Translatio ns, Jerusalem, (1971).
206 192. Y.S. Obeng, E. M. Yokley, Method of Altering the Surface Properties of a Polishing Pad and Specific Application Ther eof, US Patent # 6579604, Issued on June 17, (2003). 193. I. Chodk, A. Romanov, M. Rtzsch, G. Haudel, Acta Poly. 38, 12, 672, (2003). 194. J.-H. Ahn and D. Kwon, Mat. Sci. Eng., A 285, 172, (2000). 195. A.C. Fischer-Cripps. Nanoindentation, Springer-Verlag, New York, (2002). 196. Internet website: http://www.kett.com/files/br94ai.pdf, Kett USA, (2002). 197. P. B. Zantye, A. Kumar, W. Dallas, S. Osteapenko, A. K. Sikder, J. Vac. Sci Technol. A, In Review (2005). 198. W.D. Nix, Metallurgical Tran sactions A, 20A, 2217, (1989). 199. A.K. Bhattacharya and W.D. Nix, Intl. Journal of Solids and Structures, 24, 1287, (1988). 200. T.F. Page, W.C. Oliver and C.J. McHa rgue, J. Mater. Res., 7, 450, (1992). 201. G. Beamson, D. Briggs, High Resolution XPS of Organic Polymers, The Scienta ESCA300 Database, John Wiley & Sons NY, 1992. 202. Y. Obeng, Proceedings of the 204th Electrochemical Societ y Meeting, Orlando, FL, Editors: S. Seal, R. L. Opila, K. Sundaram, and R. K. Singh, pp. 156. 203. S. Deshpande, S. Kuiry, M. Klimov, Y. Ob eng, S. Seal, J. Electrochem. Soc. 151, G788, (2004); S. Deshpande, S. Dakshinamurt hy, S.C. Kuiry, R. Vaidyanathan, Y.S. Obeng and S. Seal, Thin Solid Films 483,1-2 261, (2005). 204. S. C. Kuiry, S. Seal, E. Megan and B. Heinman, Thin Solid Films, 423, 243, (2003). 205. S. C. Kuiry, S. Seal, W. Fei, B. Heinman, J. Ramsdell, S. V. Babu, Y. Li and B. Wood, J of Electrochemical So ciety, 150, 1, C 36, (2003).
207 206. D. Tamboli, S. Seal, V. Desai and A. Maury, Journal of Vacuum Science and Technology A 17, 4, 1079, (1999). 207. S. C. Kuiry, Y. Obeng, S. Deshpande, S. Seal, Symposium M1, Sixth International Symposium on Chemical Mechanical Polishing as a part of the 204th Meeting of The Electrochemical Society in Orlando, FL from October 1216, pp. 52, (2003). 208. F. Fracassi, R. dÂ’Agostino, P.Favio, J. Electrochem Soc. 139, 9, 2636, (1992). 209. Alexandra Fidalgo and Laura M. Ilharco, Journal of Non-Crystalline Solids, 283, 13, 144, (2001). 210. M. S. Hegde, R. Carracciolo, K. S. Hatt on, and J. B. Wachtman, Jr., Appl. Surf. Sci., 37, 16, (1989). 211. J. S. Steckenrider, B.L. Muelle r, US Patent 6,533,832 (2003). 212. T.-H. Tsai and S.-C. Appl. Surf. Sci. 210, 3-4. 190, (2003). 213. G. Bahar Basim and B. Moudgil, K ONA Power Technol. Jpn. 21, 178, (2003). 214. M. Bielmann, U. Mahajan and R.K. Singh, Electrochem. Solid-State Lett. 2, 401, (1999). 215. U. Mahajan, S.-M. Lee and R.K. Si ngh, Mater. Res. Soc. Symp. Proc 613, E 1.7 (2000). 216. H. Li, M. VanHaneham, J. Quanci, Mate r. Res. Soc. Symp. Proc. 767, F 5.3, (2003). 217. F.B. Kaufman, D.B. Thompson, R.E. Broa die, M.A. Jaso, W.L. Guthrie, D.J. Pearson and M.B. Small, J. Electrochem. Soc. 138, 11, 3460 (1991). 218. R.K. Singh, S.-M. Lee, K.-S. Choi, G.B. Basim, W. Choi, Z. Chen and B.M. Moudgil, MRS Bull. 27, 10, 752 (2002).
208 219. T.-H. Tsai and S.-C. Yen, Appl Surf. Sci. 210, 3-4, 190 (2003. 220. T.-H. Tsai, Y.F. Wu and S.C. Yen, Appl. Surf. Sci. 214, 1-4, 120 (2003). 221. A. Misra, US Patent 6,530,967 (2003). 222. A. Vijayakumar, T. Du, K.B. Sundaram and V. Desai, Polishing, Microelectron. Eng. 70, 1, 93 (2003). 223. C. Kourouklis, T. Kohlmeier and H.H. Ga tzen, Sens. Actuators A: Phys. 106, 263 (2003) 224. J. Luo and D.A. Dornfeld, Material rem oval mechanism in chemical mechanical polishing: theory and modeling, IEEE Tran s. Semicond. Manuf. 14, 2, 112 (2001). 225. S.R. Runnel and L.M. Eyman, J. Electrochem. Soc. 141, 1698 (1994). 226. Y.-R. Jeng and P.-Y. Huang, Electrochem. Solid-State Lett. 7, 2, G40 (2004). 227. N.J. Brown, P.C. Baker and R.T. Maney, Proc. SPIE 306, 42 (1981). 228. J. Sorooshian, D. Hetherington and A. Ph ilipossian, Electrochem. Solid State Lett., 7 (10), G222, (2004). 229. I. Kojima and M.Kurahashi, J. Electron. Spectr. Rel. Phen. 42, 177 (1987). 230. W. Li, D. W. Shin, M. Tomozawa, S. P. Murarka, Thin Solid Films 270, 601, (1995). 231. B. Mullany, G. Byrne, J. Mater. Process. Technol 132, 28, (2003). 232. S. Mudhivarthi, P. B. Zantye, A. Kumar, A. Kumar, M. Beerbom and R. Schlaf, Electrochemical and Solid State Lette rs, Accepted for Publication, (2005). 233. P.B. Zantye, A. Kumar and A. Kumar, Novel Nanoparticle based CMP slurry for Cu planarization, US Patent Pro cedures Pending (Ref No.: 05A023PR).
ABOUT THE AUTHOR The author received his MasterÂ’s degr ee in field of Mechanical Engineering in May 2002 at University of South Florida, Tampa, FL, USA. He will be receiving his Doctoral degree in Mechanical Engineering in August 2005 from the same university, for his work in the area of Chemical Mechanic al Planarization. Apart from CMP, the authorÂ’s research interests include: MEMS, microelectronic thin film synthesis and metrology, and chemical and biological sens ors. He has also pur sued his research interests overseas at the Institute of Micr oelectronics, Singapore during his internship. The author currently has filed 2 Patents, ha s 8 journal articles either published or under the process of being published, besides 21 c onference papers. He received the Sigma Xi Outstanding Graduate Researcher Award (T ampa Bay Chapter) for the year 2004. The author has been elected a full member of th e American Society of Mechanical Engineers and is also affiliated to the Materials Resear ch Society and the Electrochemical Society.