|USFDC Home | USF Electronic Theses and Dissertations||| RSS|
This item is only available as the following downloads:
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 001914307
007 cr mnu|||uuuuu
008 071022s2007 flu sbm 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0001973
Kakireddy, Veera Raghava R.
Effect of temperature on copper chemical mechanical planarization
h [electronic resource] /
by Veera Raghava R. Kakireddy.
[Tampa, Fla.] :
b University of South Florida,
ABSTRACT: The effects of different process parameters on tribology and surface defects were studied till date, but there has been a very minimal study to understand the effect of slurry temperature during Copper Chemical Mechanical Polishing (CMP). The surface defects such as dishing, erosion and metal loss amount for more than 50% of the defects that hamper the device yield and mainly the electrical properties during the manufacturing process. In this research, the effect of slurry temperature on tribology, surface defects and electrical properties during copper CMP employing different pad materials and slurries has been explored. Experiments were conducted at different slurry temperatures maintaining all the other process parameters constant. Post polished copper samples were analyzed for their dishing and metal loss characteristics. From the results, it was seen that the coefficient of friction and removal rate increased with increase in slurry temperature during polishing with both types of polishing pads. This increase in removal rate is attributed to a combined effect of change in pad mechanical properties and chemical reaction kinetics. The experimental data indicated that the increase in slurry temperature results in an increase in amounts of metal dishing and copper metal loss for one type of slurry and defects decrease with increase in slurry temperature for other type of slurry. This phenomenon indicates the effect of temperature on chemical reaction kinetics and its influence on defect generation. This can be attributed due to the change in pad asperities due to change in pad mechanical properties and chemical kinetics with change in slurry temperature. The slurry temperature has an effect not only on the surface defects and tribology but also on the change in pad mechanical properties. The copper thin films peeled off at higher polishing temperatures, leading to adhesion failure. With increase in temperature the copper crystallinity, hardness and modulus increased. Further with increase in the defects the electrical properties of the devices also degraded drastically and even failed to operate at higher levels of dishing and metal loss. This research is aimed at understanding the physics governing the defect generation during CMP.
Thesis (M.S.)--University of South Florida, 2007.
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 90 pages.
Advisor: Ashok Kumar, Ph.D.
x Electrical Engineering
t USF Electronic Theses and Dissertations.
Effect of Temperature on Copper Ch emical Mechanical Planarization by Veera Raghava R. Kakireddy A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science in Electrical Engineering Department of Electrical Engineering College of Engineering University of South Florida Co-Major Professor: Ashok Kumar, Ph.D. Co-Major Professor: Sh ekhar Bhansali, Ph.D. Wilfrido Moreno, Ph.D. Date of Approval: March 28, 2007 Keywords: cof, dishing, defects, pad, metal Copyright 2007, Veera Raghava R. Kakireddy
Dedication I dedicate this work to my belove d parents, family and my friends.
ACKNOWLEDMENTS I am thankful to everyone who helped me throughout my research work to make this work successful. I thank my family for th eir love and constant support. I express my deep gratitude and thankfulness to Dr. Ashok Kumar, Major Professor, for providing me with this opportunity to c onduct the thesis and also for his guidance and encouragement throughout my research work. I am grateful to Dr. Shekhar Bhansali, Co-major Professor and Dr. Wilfrido Moreno for accepting to be on the committee. I am very thankful to my colleagues and friends in the group, Raghu Mudhi varthi for his valuable suggestions and help during the research work and Sathyaharish Jeedigunta for his support and encouragement. I thank all my friends for their encouragement and moral support during the research period.
i Table of Contents List of Tables iii List of Figures iv Abstract vii Chapter 1 Introduction 1 1.1 Multilevel Metallization and I.Cs 1 1.2 Why CMP 2 1.3 Literature Overview 6 1.4 Thesis Outline 10 Chapter 2 Understanding CMP 11 2.1 The CMP Process 11 2.2 Factors Affecting CMP 13 2.2.1 Polishing Pad 13 2.2.2 Slurry 14 2.2.3 Surface Under Polish 16 2.2.4 Pressure 16 2.2.5 Velocity 17 2.2.6 Temperature 17 2.3 Copper Significance 18 2.4 Copper Deposition Techniques 20 2.4.1 Chemical Vapor Deposition (CVD) 20 2.4.2 Physical Vapor Deposition (PVD) 21 2.4.3 Copper Electroplating 22 2.4.4 Electroless-Plating 23 2.5 Copper Removal Mechanism 24 2.5.1 Formation of Surface Layer 25 2.5.2 Abrasion 26 2.5.3 Removal of Abraded Material 26 2.6 Pattern Geometry Effects on Rem oval Mechanism 27 2.7 Defects During CMP 27 2.7.1 Dishing 28 2.7.2 Erosion 29 2.7.3 Total Copper/Metal Loss 30 2.7.4 Scratches 30 2.7.5 Wafer to Wafer Non-uniformity 31
ii 2.7.6 Within Wafer Non-uniformity 31 2.8 Summary 31 Chapter 3 Experimental 33 3.1 CMP Testing Tool 33 3.2 CMP Consumables 34 3.3 Experimental Parameters 35 3.4 Estimation of Removal Rate 37 3.5 Post CMP Analysis/Characteri zation Tools 39 3.5.1 Profilometer 39 3.5.2 Atomic Force Microscopy (AFM) 40 3.5.3 Scanning Electron Microsc opy (SEM) 41 3.5.4 Dynamic Mechanical Analysis (DMA) 43 3.5.5 X-Ray Diffraction (XRD) 45 3.6 Summary 47 Chapter 4 Results and Discussion 48 4.1 Dynamic Mechanical Analysis (DMA) of Pads 48 4.2 Scanning Electron Microscopy (SEM ) Analysis 50 4.3 Analysis of COF and Removal Rate with Slurry Temperature and Pad 52 4.3.1 Coefficient of Friction (COF) 52 4.3.2 Removal Rate 59 4.4 Analysis of Dishing with Slurry Temperature and Pad 62 4.5 Analysis of Metal Loss with Slurry Temperature and Pad 69 4.6 Analysis of Adhesion Failure and M echanical Properties with Slurry Temperature 71 4.7 Analysis of Electrical Pr operties with Slurry Temperature 76 4.8 Summary 80 Chapter 5 Conclusions and Future Work 82 5.1 Conclusions 82 5.1.1 Friction (COF) and Removal Rate Studies 82 5.1.2 Dishing and Metal Loss 83 5.1.3 Film Mechanical and Electrical Properties 85 5.2 Future Work 85 References 87
iii List of Tables Table 2.1 Comparison of Properties of Metals 18 Table 3.1 Consumables and Process Parameters for Polishing 35 Table 4.1 Coefficient of Friction at Different Slurry Temperatures Using Slurry-1 57 Table 4.2 Coefficient of Friction at Different Slurry Temperatures Using Slurry-2 58 Table 4.3 Removal Rates at Different Slurry Temp eratures Using Slurry 1 and 2 60 Table 4.4 Normalized Dishing Data on Metal Lines and Bond Pads 62 Table 4.5 Mechanical Properties of Copper Thin Films Before and After Polishing 74
iv List of Figures Figure 1.1 Cross Section of a Device Showing Various Levels of Metallization 2 Figure 1.2 2005 ITRS Half Pitch and Gate Length Trends 3 Figure 1.3 Schematic of a CMP Setup 6 Figure 2.1 Schematic of a CVD Process 21 Figure 2.2 Schematic of a PVD Process 22 Figure 2.3 Schematic of a Copper Electropl ating System 23 Figure 2.4 Schematic of Copper Removal Mechanism 24 Figure 2.5 Defects Formed During CMP 28 Figure 2.6 Schematic of Copper Dishing During CMP 29 Figure 2.7 Schematic of Erosion Profile 30 Figure 3.1 CMP Bench Top Tester 34 Figure 3.2 Process Flow Diagram for Fabrica ting the PMOS Devices 37 Figure 3.3 COF vs. Time Graph Used to Estimate Removal Rate 39 Figure 3.4 Schematic of a Profilometer Scan 40 Figure 3.5 Schematic of Operation of an AFM 41 Figure 3.6 SEM Image/Schematic of an AFM Cantilever Tip 41 Figure 3.7 Block Diagram Showing the Operation of a SEM 43 Figure 3.8 DMA 2980 Instrument at USF 45 Figure 3.9 Schematic of an X-Ray Diffractometer 46 Figure 4.1 Storage Modulus and Tan Delta vs. Temp erature at 10 Hz Frequency 49
v Figure 4.2 Loss Modulus vs. Temperature at 10 Hz Frequency 50 Figure 4.3 SEM Image of Pad A a) Unconditioned b) Conditioned 51 Figure 4.4 SEM Image of Pad B a) Unconditioned b) Conditioned 51 Figure 4.5 COF vs. Time of a Sample at 15 C on Pad A 52 Figure 4.6 COF vs. Time of a Sample at 20 C on Pad A 53 Figure 4.7 COF vs. Time of a Sample at 25 C on Pad A 53 Figure 4.8 COF vs. Time of a Sample at 30 C on Pad A 54 Figure 4.9 COF vs. Time of a Sample at 35 C on Pad A 54 Figure 4.10 COF vs. Time of a sample at 15 C on Pad B 55 Figure 4.11 COF vs. Time of a sample at 20 C on Pad B 55 Figure 4.12 COF vs. Time of a sample at 25 C on Pad B 56 Figure 4.13 COF vs. Time of a sample at 30 C on Pad B 56 Figure 4.14 COF vs. Time of a sample at 35 C on Pad B 57 Figure 4.15 Change in COF vs. Slurry Temper ature for Slurry-1 58 Figure 4.16 Change in COF vs. Slurry Temper ature for Slurry-2 59 Figure 4.17 Removal Rate vs. Slurry Temperat ure Using Slurry-1 60 Figure 4.18 Removal Rate vs. Slurry Temperat ure Using Slurry-2 61 Figure 4.19 Dishing on a 10 m Wide Meta l Line vs. Slurry Temperature Using Slurry-1 63 Figure 4.20 Dishing on a 10 m Wide Meta l Line vs. Slurry Temperature Using Slurry-2 64 Figure 4.21 Dishing on a 20 m Wide Meta l Line vs. Slurry Temperature Using Slurry-1 65 Figure 4.22 Dishing on a 20 m Wide Meta l Line vs. Slurry Temperature Using Slurry-2 65
vi Figure 4.23 Dishing on a 50 m Wide Meta l Line vs. Slurry Temperature Using Slurry-1 66 Figure 4.24 Dishing on a 50 m Wide Meta l Line vs. Slurry Temperature Using Slurry-2 67 Figure 4.25 Dishing on a 100 m Bond Pad vs. Slurry Temperature Using Slurry 1 68 Figure 4.26 Dishing on a 100 m Bond Pad vs. Slurry Temperature Using Slurry 2 68 Figure 4.27 Dishing Profiles of Metal Lines of Different Widths 69 Figure 4.28 Metal Loss vs. Slurry Temperatur e Using Slurry-1 70 Figure 4.29 Metal Loss vs. Slurry Temperatur e Using Slurry-2 71 Figure 4.30 Pictures Showing Pee ling on Wafers Polished at Different Temperatures 72 Figure 4.31 Graphs Showing XRD Peak s on Copper Wafers Polished at Different Temperatures 73 Figure 4.32 Hardness vs. Displacement into Surface for Unpolished and Polished Copper Samples 75 Figure 4.33 Modulus vs. Displacement into Surface for Polished and Unpolished Copper Samples 75 Figure 4.34 Graphs Showing Transfer Curv es (I-V Characteristics) of Wafers Polished at Different Temperatures 77 Figure 4.35 Graphs Showing Sub-Thre shold Log Plot Curves (I-V Characteristics) of Wafers Polishe d at Different Temperature 79
vii Effect of Temperature on Copper Ch emical Mechanical Planarization Veera Raghava R. Kakireddy ABSTRACT The effects of different process parame ters on tribology and surface defects were studied till date, but there has been a very minimal study to understand the effect of slurry temperature during Copper Chemical Mechanic al Polishing (CMP). The surface defects such as dishing, erosion and metal loss amount for more than 50 % of the defects that hamper the device yield and mainly the el ectrical properties dur ing the manufacturing process. In this research, the effect of slurry temperature on tribology, surface defects and electrical properties during copper CMP employing differe nt pad materials and slurries has been explored. Experiments were condu cted at different slurry temperatures maintaining all the other pro cess parameters constant. Post polished copper samples were analyzed for their dishing and metal loss characteristics. From the results, it was seen that the coefficient of friction and removal rate in creased with increase in slurry temperature during polishing with both type s of polishing pads. This in crease in removal rate is attributed to a combined effect of change in pad mechanical properties and chemical reaction kinetics. The experimental data indicat ed that the increase in slurry temperature results in an increase in amounts of metal dishing and copper metal loss for one type of slurry and defects decrease with increase in slurry temperature for other type of slurry. This phenomenon indicates the effect of temper ature on chemical reaction kinetics and its
viii influence on defect generation. This can be at tributed due to the change in pad asperities due to change in pad mechanic al properties and chemical kine tics with change in slurry temperature. The slurry temperature has an effect not only on th e surface defects and tribology but also on the change in pad mechanical properties. The copper thin films peeled off at higher polishing temperatures, l eading to adhesion failure. With increase in temperature the copper crystallinity, har dness and modulus increased. Further with increase in the defects the electrical propert ies of the devices also degraded drastically and even failed to operate at higher levels of dishing and metal loss. This research is aimed at understanding the physics govern ing the defect generation during CMP.
1 Chapter 1 Introduction 1.1 Multilevel Metallization and I.Cs Shrinking device dimensions associated with ultra large scale integrated circuits is highly effective in achieving high speed pe rformance and in increa sing yields at lower cost per chip With the ever increasing thirst to build faster chips the industry is facing the critical problem of integrating the new desi gn rules and also in advancing the Moores law further. In order to achie ve higher circuit density the number of interconnect layers are increased. However when the devices are scaled down the performance falls down as a result for higher circuit de nsity. The higher circuit density causes the devices to slow down due to the increase in interconnect RC time delay of the circuit. As more and more faster devices are built the complexity of fa bricating the device with more interconnect layers also increases . In order to assu re the performance of the high speed circuits, continuous efforts have been devoted for incorp orating copper or low dielectric constant materials into multilevel interconnections for reducing major part of circuit delay, cross talk and power consumption. Till date Alum inum has been the choice of semiconductor manufacturers for the metal contacts, but as aluminum has high resi stivity the industry had shifted to copper which has very low resitivity. Also copper has lower electromigration when compared to alum inum. Unlike aluminum, copper cannot be etched and has to be planarized in a different form. Figure.1 shows a typical multilevel metallization circuit consisting of various metallization layers.
2 Figure 1.1 Cross Section of a Device Showi ng Various Levels of Metallization  1.2 Why CMP With the day-to-day advancement in th e field of semiconductor technology and a constant thirst driven by competition for the miniaturization of devices the International Technology Roadmap for Semiconductors (ITRS) which is an association for semiconductor industry has set a goal for achieving various leve ls of miniaturization in the field of semiconductor industry. Figur e.2 shows the levels of metallization and miniaturization set by ITRS. As the number of levels in an interconnect technology is increased, the stacking of add itional layers on top of one another produces a more and more rugged topography. The surface of the wafe r must be planarized in some fashion to prevent topography roughness from growing with each level. Without such planarization stacking of device features can lead to topography conditions that would eventually reduce the yield of circuits . Ca p la y er HM (hard mask) Low-k Low-k Cu Barrier Cu Low-k Low-k Cu Cu
3 Figure 1.2 2005 ITRS Half Pitch and Gate Length Trends  Chemical mechanical polishing takes it r oots from the glass polishing industry, where it is used to produce ultra smooth finish ed glass surfaces. Much before the use in semiconductor industry CMP is being used ex tensively in the glass polishing industry. From just a simple glass polishing technique CMP has been the major process technique to obtain various levels of planarization in semiconductor industry. Chemical Mechanical Planarization (CMP), a process that was pioneered at IBM in the 1980s is the globally used technique to planarize both dielectric and metal layers. With the advancement in semiconductor technology there is a need for global planarization through CMP in order to ensure multilevel copper interconnects. Chemical Mechanical Polishing offers various advantages over other available processe s. The advantages of CMP are:  Universal or materials insensitive al l types of surfaces can be planarized. Achieves global planarization.
4 Reduces severe topography allowing for fa brication with tighter design rules and multilevel interconnects. Provides an alternate means to pattern metal thereby eliminating the need for reactive ion etching or plasma etch for di fficult to etch metals and alloys (Eg: Copper). Leads to improved metal step coverage. Helps in improving reliability, speed and yield of devices. Does not use hazardous gases as in the case of dry etching process. Low cost process. The CMP process even with many adva ntages and desired properties is not without drawbacks such as di shing, erosion etc. Very littl e has been understood about the effects of temperature during the CMP proce ss. Even though research has been done to study the temperature effect on pad not much has been done to study its affect on surface defects and tribology . Chemical mechanical polishing which is a process of chemical and mechanical action on the surface being pro cessed is dependent on variou s input process parameters. CMP was initially developed for polishing oxide (SiO2), which was then used as the interlevel dielectric (ILD) in the multile vel metallization. However with unprecedent advancement in semiconductor technology an d introduction of various new dielectric materials, barrier layer mate rials and copper as metal the CMP development was forced to follow a two fold approach to achieve th e required level of plan arization of various layers. In this technique, high elevation f eatures are selectively removed resulting in surface with improved planarity .
5 The process of chemical mechanical polishing is performed by mounting the wafer face down on a carrier, which is then fo rced against a platen with a polishing pad of a polymer such as polyurethane. The carri er and the platen rotate relative motion to each other . A continuous flow of slurry c ontaining abrasive part icles is continuously fed onto the surface of polishing pad while the wafer and pad rotate relative to each other. The interaction of both chemical and mechani cal effects results in material removal from the wafer surface. The removal rate of the material during CMP is governed by the Prestons equation:  Removal Rate (RR) = Kp P V (1.1) Where P is the pressure applied on th e wafer surface, Kp is the Prestons coefficient and V is the linear velocity of th e pad relative to the wafer. Prestons equation is the most frequently referenced expression for polishing rates. Figure 1.3 shows the basic schematic of a CMP setup. The polishing pad usually a polymer is placed on the platen, which rotates a set speed typically in the range of 100 200 RPM. The wafer to be planarized is pl aced on a carrier face down and is pressed down onto the pad surface. The force is applied onto the wafer through the carrier. Proper care has to be taken so that the pressure is uniform at all points on the wafer combined with controlled table motion and carrier rotation rates. The slurry is fed onto the interface of the wafer and pad through a feed pipe at a constant volume. The constant slurry flow allows for uniform planarization and also dissolves the removed material in the slurry from the pad wafer interface. The pads are generally porous in nature and are either grooved or perforated for different applicatio ns. The pad is constantly conditioned while
6 polishing in order to main tain its roughness. The conditi oner is usually made of a diamond grid. Figure 1.3 Schematic of a CMP Setup 1.3 Literature Overview With the ever increasing demand for fa ster performing devices and a constant competition driven industry, till date vast re search has been done on various aspects of chemical mechanical polishing (CMP) and st ill research is being done investigating various aspects of CMP which have been either not looked into or the aspects which couldnt be understood. Various materials are be ing investigated for use in semiconductor devices. Different low-k dielectric materials ha ve replaced the tradit ional silicon dioxide (SiO2) which has a dielectric constant ~ 4 wi th materials of diel ectric constant ~ 2.2. Copper has replaced aluminum as the materi al for contact metal and interconnects. Various reliability studies have been conducted on copper for multilevel metallization Carrier Wafer Slurry Platen (Pad Holder) Pad Pad Conditioner Downforce
7 schemes. Chemical mechanical polishing ha s been the choice of semiconductor industry and researchers for copper removal, as copper cannot be removed by conventional techniques such as wet and dry etching. Till date research has been done on the CMP of c opper looking into various aspects of polishing such as the consumab les, process parameters and other input parameters. The chemical mechanisms by wh ich the planarization and removal occur during the chemical mechanical polishing (CMP ) of copper have been investigated long back by Steigerwald et al . They proposed that removal occurs as mechanical abrasion of the surface under polish followed by dissol ution of abraded particles. Chemical Mechanical Polishing of copper using silica an d alumina abrasive particles in different chemistries has been studied by Carpio et al . In their study the problems with formulation of copper slurries were discussed. Slurries with different chemistries have been studied for their reactivity with pad surf aces by Obeng et al . The results showed that the polyurethane pad material is incompatible with some of the chemicals used in CMP, such as hydrogen peroxide. The influence of the slurry chemistry on frictional force in copper during the CMP has been st udied by Ishikawa et al . Their study showed that polishing rate increased nonlinearly with frictional force, which is controlled by the concentration of slurry chemical constituents. The effect of slurry surfactant, abrasive size on tribology and kinetics in copper CMP has been determined by Li et al  The abrasive size was a significant factor while surfactant containing formulations show ed significant reducti on in coefficient of friction (COF). The effects of amount of concentration of ab rasive particles were studied by Nomura et al  and were found to have a significant affect on the polishing process.
8 CMP process of two different substrates us ing copper discs and copper were compared by analyzing their COFs, remova l rate and pad temperature an alysis by Li et al . Their studies showed that the coefficient of friction and material removal rates were higher for copper deposited wafers. The wear an d tear in CMP polyurethane pads due the physical and chemical changes have been analyz ed by Lu et al . Their results showed changes in pore geometry of the CMP pads af ter use due to wear and tear. Properties of CMP polishing pads before and after CMP have been studied using Dynamic Mechanical Analysis (DMA) by Lu et al . The studies s howed that mechanical force has shown more significance than chemical action in pad degradation. The effect of pattern characteristics duri ng copper CMP has been investigated by Wang et al . The copper and ILD CMP ha ve been characterized by looking into some thermal aspects by Sorooshian et al . In their study a modified Prestons equation has been proposed by taking the temperature eff ect into consideration. A two step copper removal method has been studied for removal of thick copper by Mira nda et al . The use of urea and hydrogen peroxide based slurries were investig ated by Tsai et al . Their results showed that urea based slurries can achieve better cu CMP. The effects of slurry pH and oxidizer concentration on coppe r CMP process were studied by Miranda et al . Their study using the design of expe riments (DOE) approach showed that the interaction factor of pH and H2O2 has dominant effect on CMP process. Delamination during the CMP process in integrating ultra low-k/Cu has been studied by Leduc et al . Chemical Mechanical Polishing of copper in alkaline media such as ammonium hydroxide has been investigated in . Their studies show ed that the removal rate
9 increases with increase in NH4OH concentration and then at a certain point saturates. Surface generated defects such as pitting co rrosion, galvanic corrosion and excessive etching to chemical action has been studied by Miller et al  Their investigation showed that a two step CMP process for Cu/T a could reduce the def ects. Electrochemical characterization of CMP process using hydroge n peroxide as oxidizer and alumina as abrasive particles has been done on coppe r in  and was found that at 1% H2O2 concentration the removal rate was maximu m and decreased further with increase in H2O2 concentration. The effects of slurry flow rate, pad surface temperature and pad temperature during conditioning on surface tribology and pattern related defects we re investigated by Mudhivarthi et al . In their study th e pad surface temperature and conditioning temperature were shown to have a significan t effect on metal dishing, dielectric erosion and on the nature of conditioning. Their study also indicated that dishing and erosion decreased with increase in sl urry flow rate. The increase in amount of dishing was correlated with the increase in pad surface temperature. A study by Mudhivarthi et showed that a change in chemical dissoluti on rate of copper coul d be influencing the dishing characteristics . These researches and many others dealt mostly with the fundamental aspects of CMP related to process parameters like pressure, velocity and other consumable characteristics but research based on thermal aspects has not been done extensively. Even though there are some modeling works and a ve ry few experimental works on the effect of temperature on CMP [16, 26, 27, 28], there is a lack of understand ing on the effect of
10 temperature on various aspects of CMP process defects such as dish ing, erosion etc. and on tribology during copper CMP. 1.4 Thesis Outline The aim of this study/thesis is to study the effects of slurry temperature on surface defects such as dishing, di electric erosion, metal loss a nd on tribology during copper Chemical Mechanical Polishing (CMP) process. This research is aimed at understanding the physics governing the def ect generation during CMP. The introduction to use of copper, multilevel metallization and use of CMP process to planarize copper, various consumables effecting CMP process and literature overview of copper CMP process has already be en discussed in Chapter 1. The Chapter 2 gives a detailed insight about the significan ce of copper, copper removal mechanism and the available deposition methods for copper. Th is chapter also gives a detailed overview of various factors effecting CMP process and the consumables used. The CMP polisher used and the experimental setu p along with the surface def ect generation and effect of slurry temperature on tribology are analyzed in chapter 3. Chapter 4 describes the PostCMP analysis of polished cu samples using Atomic Force Microscopy (AFM), analysis of pad materials before and after conditio ning using Scanning Electron Microscopy (SEM), Dynamic Mechanical Analysis (DMA). Also analysis of polished wafers for device electrical characteristic s, due to surface de fects and structural changes in copper due to polishing at different temperatures are described. Changes in copper structural characteristics were analyzed using X-ray diffraction (XRD) and na noindentation studies. Chapter 5 gives an outline about the conclusions from the research work and future work.
11 Chapter 2 Understanding CMP Semiconductor industry which is constantly looking for new and better materials to replace the existing interconnect materials in order to reduce the RC delay and size of the device has shifted towards copper as in terconnect material fr om the traditional aluminum. Copper posses several advantages ov er aluminum and this led the industry to shift towards the use of copper. Copper has properties such as lo w resistivty and high electromigration resistance required for a good conductor and a interconnect material. The semiconductor industry shifting towards the use of copper replacing aluminum faced a problem as unlike aluminum, copper cannot be etched and has to be planarized in a different form. Deriving fr om just a simple glass polishing technique, Chemical Mechanical Planarization (CMP) is the only and most widely used method to planarize copper in the semi conductor industry. Chemical Mechanical Planarization (CMP), a process that was pione ered at IBM is the globally used technique to planarize both dielectric and metal layers. 2.1 The CMP Process The Chemical Mechanical Polishing (CMP ) process was started taking the idea of glass polishing which has been in practice since centuries. Unlik e the glass polishing CMP has many more advantages and is a cont rolled process. The process of CMP was initially used for planarizing oxide, but with the introduction of copper for multilevel
12 metallization schemes use of Chemical Mechanical Polishing (CMP) has become inevitable. With rapid increase in advancem ent of semiconductor processing technologies CMP has become a global planarization technique in semiconductor manufacturing. As said earlier in chapter 1 the proce ss of chemical mechanical polishing is performed by mounting the wafer face down on a carrier, which is then forced against a platen with a polishing pad of a polymer such as polyurethane material. The carrier and the platen rotate relative motion to each other . Slurry containing abrasive particles is constantly fed onto the surface of polishing pad while the wafer and pad rotate relative to each other. The chemical and mechanical inte raction then results in material removal from the wafer surface. The re moval rate of the material is governed by the Prestons equation . The CMP process which takes place due to the interaction of pad, wafer and slurry form the main consumables and play a significant role. Along with the consumables other factors such as pressure, ve locity, temperature, constituents of slurry and various other factors play a considerab le role. The significance of each consumable and various factors affecting the Chemical Mechanical Polishing (CMP) process are explained in detail in the fl owing sections. The CMP proces s with many advantages also has some disadvantages such as surface def ects (dishing, dielectric erosion, metal loss and erosion), scratches due to particles, pattern dependent nonuniformity, wafer to wafer and within wafer non-uniformity caused due to various process parameters and process consumables.
13 2.2 Factors Affecting CMP In a CMP process three main consumab les play an import ant role. Although various other factors such as temperature, pressure and velo city play an important role, the three consumables mentioned below play a major role in a CMP process. They are: 1 Polishing pad. 2 Slurry. 3 Wafer or surface to be polished. 2.2.1 Polishing Pad Polishing pads are generally made of either a matrix of polyurethane foam with filler material to control hardness or polyurethane impregnated felts. Polyurethane is utilized because urethane chemistry allows th e pad characteristics to be varied to meet specific mechanical properties needs. There are various types of polishing pads that are available commercially. Most commonly used pads are the ones that have perforations or groves on the pads. Rodel and Cabot Microe lectronics are the two main commercial manufacturers of CMP pads. The polishing pads are porous and pores are introduced intentionally. Pads are made por ous as they aid in slurry tr ansport or distribution over the pad surface. Pads used for polishing either ha ve perforations or k-grooves. Researchers and industry are coming out with different kinds of pad materi als and pad surface structures. The pad hardness also effects the pl anarization to great extent. The harder the pads the better is its planarizing ability. Pa ds with different hardnesses are used for planarizing different material s. The effect of pad hardne ss along with temperature on planarization process is investigated in this research. To study th e pad properties studies such as Dynamic Mechanical Analysis (DMA), Scanning Electron Microscopy (SEM)
14 and Fourier Transform Infrared Spectroscopy (FTIR) are performed. The DMA explains the change in pad hardness and properties with change in temperature. This helps in understanding the affect of pad on defect generation during CMP with increase in temperature. The SEM images show the pad pore size and wear in pads, as CMP process is a constant wear and tear process due to the friction between surfaces which is a mechanical component. Also the pad degrades due to the chemical effect. Studies by Obeng et al have showed that polyurethane pads are incompatible with chemicals, such as hydrogen peroxide (H2O2) . FTIR studies give an understanding about the affect of chemicals on pad. 2.2.2 Slurry The slurry is one of the main constituent in the chemical mechanical polishing process as the slurry forms the chemical co mponent of the CMP process. The slurry is made up of abrasive particles, oxidizers, buffering agents and complexing agents. Each and every constituent of the slurry plays an important role in polishing process. The principles of electrochemistry are very useful in explaining the chemical mechanisms of metal CMP. The electrochemistry of the slurry explains the proces s of metal solubility, metal dissolution, and surface layer formation. since CMP is used for a large variety of materials such as oxides, barrier metals and copper, each of these materials has a different chemistry associated with the polishing pro cess. The chemical reaction between the pad and wafer surface modify the mechanical propert ies of the film, abrasive particles, pad, which in turn affects the mechanical component .
15 The pH of the slurry affects the solubility and dissolution rate of the surface being polished. Various types of buffering agents de pending upon the type of slurry are used to maintain the pH constant through out the sl urry volume over time. Without the use of which the pH may vary considerably at th e wafer surface and in turn may affect the polish rate. In a metal CMP process such as in copper CMP process the chemical reactions are electrochemical in nature. Oxidiz ers react with the metal surface in order to raise the oxidation state of the metal through a reduction-oxidation re action, resulting in formation of a surface film on the metal or dissolution of metal. The use of complexing agents increases the solubility of the film being polished. The increase in solubility rate increases the metal removal rate. Hydrogen peroxide (H2O2) is the most commonly used oxidizers in a CMP process. The use of various complexi ng agents such as ammonium hydroxide (NH4OH) , citric acid , oxalic acid , etc., have been investigated. The abrasive or the particles used in th e slurry provide the necessary mechanical action in a CMP process. The abrasives are ma de of different materi als and are spherical in shape. The diameter of abrasive particle s may vary in the range of nanometers to a couple of microns. Silica (SiO2) particles are the mostly used abrasive particles in oxide CMP, also used in metal CMP such as in copper polishing. Alumina (Al2O3) particles are the most commonly used for metal CMP. Othe r than alumina and silica various other particles such as ceria (CeO2), magnesium oxide (MgO), titanium dioxide (TiO2), zirconia (ZrO2) are used for polishing different mate rials. The size and hardness of the abrasives also plays an important role as the size of the particle affects the removal rate and surface damage. The abrasive concentrat ion also affects the polishing rate, the material removal rate increases with increase in abrasive concentration.
16 2.2.3 Surface Under Polish Wafer or the surface to be polished is also one of the major factors as the feature size, pattern density and material being polishe d affects the localized pressure distribution which in turn affects the removal rate. The patte rn density and feature size thus affect the polishing rate and may cause metal dishing a nd dielectric erosion. Small features polish quicker than larger features . In plan arizing metal over ILD patterned for vias and trenches pattern dependence is observed. The hardness of the wafer surface being polished may also cause scratches or some su rface defects. The surface quality affects the yield and also the reliability of the device. The curvature of the wafer also plays a major role in pressure distribution across the wafer surface. This may result in non-uniformity in polishing across the wafer. The stresses in the different film layers may affect planarization, in case of some low-k dielectrics the film might delaminate due to low mechanical strength. 2.2.4 Pressure Mechanical load or pressure is appl ied on the wafer surface pressing it downwards during the polishing process. If the surface is rough or has topography, the contact area is less than the geometric area. Due to the topography or rough/uneven surface the pressure is increased until the surface is made smooth. Based on Prestons equation and earlier studies have shown that the material removal rate is proportional to the applied load or pressure .
17 2.2.5 Velocity The platen velocity plays a significant ro le in removal rate. The material removal rate is proportional to the pad platen veloc ity . Slurry trans port across the wafer is dependent on the pad velocity. As the pad ve locity increases the slurry film thickness decreases and there is an increa se in material removal rate. 2.2.6 Temperature Since the polishing takes place as a result of mechanical interaction of the surface being polished, slurry and the pa d, this process in part can be termed as a wear process. Due to the rubbing of two surfaces there is an increase in te mperature at the pad wafer surface interface due to friction. Major part of the temperature increase or heat generated is attributed to the interaction of wafer and slurry interaction and very less due to pad. The increase in the temperature mainly affe cts the reaction rates. The temperature has considerable effects on the surface defect generation and on tribology during the CMP process. Mudhivarthi et al has investigated the affect of pad conditioning temperature on the removal rate and surface defects . Till date research has been done looking into various aspects such as CMP consumables, pre ssure, velocity and othe r factors, but little has been understood about the effect of slurry temperature on the polishing and the tribology. This research aids in understanding the effect of slurry temperature on removal rate, tribology and surface defects such as dishing and erosion during copper CMP. Also other factors such as thickness, elas tic and shear modulus, aging effects of pad, post CMP cleaning, surface quality etc., have a considerable affect on the Chemical
18 Mechanical Polishing (CMP) process, this resear ch mainly focuses on the effect of slurry temperature and to study the physic s governing the copper CMP process. 2.3 Copper Significance Copper has become the most favorite ma terial in the semiconductor industry for manufacturing the devices. W ith constant decrease in th e size of the semiconductor devices in accordance with the Moores law the industry s earch for alternate metal to replace aluminum has resulted with the exte nsive use of copper as interconnect material. The properties of copper such as good electr ical conductivity due to low resistivity of 1.67 x 10-8 -m and high resistance to electromigration has become the ultimate choice of semiconductor industry. The table 2.1 comp ares various propert ies of copper with other metals and explains the significance of copper. Table 2.1 Comparison of Pr operties of Metals  Cu Al Ag Au Al alloy W Resistivity (10-8 -m) 1.67 2.66 1.59 2.35 3.5 5.65 Electromigration Resistance Good Poor Poor Very Good Poor Very Good Corrosion Resistance Fair Good Poor Excellent Good Good Adhesion to Oxide/Low-k Poor Good Poor Poor Good Poor CVD Processing Available None None None None Available RIE Etch None Available None None Available Available From table 2.1 it can be observed that of a ll the metals listed with low values of resistivity only copper (Cu), silver (Au) and gold (Ag) ha ve low resistivity values compared to aluminum (Al). It could be seen th at silver has lower resistivity than that of
19 copper, but silver diffuses eas ily into silicon dioxide (SiO2) and also has poor resistance to electromigration. Whereas copper has good el ectromigration resistance and search for proper barrier layers to prevent Ag diffusion into oxide couldnt yield any success. Gold (Au) has high resistance to corrosion and electromigration but has higher resistivity compared to copper. Also copper has high melting and boiling points (1357.77, 2835 K) compared to silver (1235, 2435 K) and alum inum (933.5, 2792 K)  Current densities in the lower interconnect levels can lead to electromigration failure. Copper can carry higher current densities compared to al uminum and its alloys and has high electromigration resistance . Copper with all the above properties compar ed to other available low resistivity metals make it an inevitable choice as a in terconnect metal. Even though copper has good properties to be a interconnect metal it ha s its own drawback and is not without challenges. Copper low adhesion and di ffuses easily into oxide/low-k. However researchers have come up with various barrie r metals such as tantalum (Ta), tantalum nitride (TaN), titanium (Ti), titanium dioxide (TiO2) etc., to improve copper adhesion and avoid diffusion into oxide. Copper also is not resistant to corrosion and hence a passivation layer is required. Researchers were successful in finding methods for passivating copper to avoid corrosion . Of all these the major challenge for use of copper as a interconnect metal is copper cannot be etched or patterned using the available wet/dry etch techniques. With the introducti on of Chemical Mechanical Polishing (CMP) the problem of patterning coppe r could be eliminated and copper has found a widespread use in semiconductor industry as a interconnect metal.
20 2.4 Copper Deposition Techniques Copper films can be deposited in a number of ways, such as chemical vapor deposition (CVD), physical vapor depositio n (PVD), electroplating and electroless plating. 2.4.1 Chemical Vapor Deposition (CVD) The CVD process uses precursors for copper deposition and organo-mettalic sources are the most commonly used precursor s for CVD of copper. Precursors such as copper (II) hexa-fluoroacetyl-acetonate [CuII(hfac)2] and CuI(hfac)L, where L is a weakly bonded neutral ligand are used. Copper f ilms of higher quality can be obtained by hydrogen reduction of CuII(hfac)2. In order to obtain copper fi lms with low resistivity the deposition temperature must be in the rang e of 350-450 C, but sinc e use of copper with low-k dielectric materials which are thermally fragile such high deposition temperatures are undesirable. Various deposition process have been developed of which a low-pressure plasma process has been reported which allows for copper deposition at 170 C and films with comparatively low resistivity can be obtained. Chemical vapor deposition process has issues with depositing void free filling of contact holes and vias. Figure 2.1 shows the schematic of a CVD process.
21 Figure 2.1 Schematic of a CVD Process  2.4.2 Physical Vapor Deposition (PVD) PVD or sputtering is one of important deposition techniques available to deposit copper. Copper is electroplated for depositing the bulk copper layer. PVD is used to deposit the Cu seed layer which is required to electroplate copper. Sputtering is a physical process whereby atoms in a solid targ et material are ejected into the gas phase due to bombardment of the material by energetic ions. The impact of an atom or ion on a surface produces sputtering from the surface as a result of the momentum transfer from the in-coming particle. Unlike many other va por phase techniques there is no melting of the material . Figure 2.2 shows a sc hematic of a PVD deposition process. Inlet Deposition Chamber Plasma Gaseous Precursors
22 Figure 2.2 Schematic of a PVD Process  2.4.3 Copper Electroplating Deposition of copper by electroplating t echnique is the most widely used deposition technique in the semiconductor industry. Since electroplating is a low temperature process it has become the ultimat e choice of industry to use it for depositing copper on low-k materials. Elec troplating process has high yi eld output as it has high throughput and also this process can fill high as pect ratio structures. Solutions containing the ions of metal to be deposited are uti lized in electroplating process. For copper electroplating a solution cont aining copper sulfate (CuSO4), sulfuric acid (H2SO4), and water is used. The CuSO4 in the solution breaks up into Cu2+ and SO4 2ions. The wafer with a copper seed layer is immersed into the solution and the wafer surface is connected electrically to negative side of an external DC supply source, this acts as a cathode. The positive copper ions are attracted towards the cathode and acquire two electrons and reduces to copper metal, thus depositing (p lating) on wafer surface [34, 36]. The reaction that takes place at the cathode is:
23 Cu2+ + 2eCu (2.1) The anode is attached to th e positive side of the exte rnal DC supply source. The rate of electroplating is a direct function of current de nsity. Electropla ting process has very high deposition rates. Figure 2.3 show s the schematic of copper electroplating system. Figure 2.3 Schematic of a Copper Electroplating System  2.4.4 Electroless-Plating The process of copper deposition thr ough electroless-plat ing involves the formation of a thin film from an electrolytic solution without externally applied voltage as in case of electroplating. In the electro less process the depositi on takes place due to electrochemical reaction betw een the metal ions, complexing, reducing agent, and pH buffers on catalytic surfaces. The chemical process involves CuO4 and formaldehyde reduction: [34, 38] Cu2+ +2HCHO + 4OHCuo (solid) + 2H2O (liquid) + H2 (gas) (2.2)
24 The main challenge in using both electr oplating and electroless-plating is to obtain a void free surface. 2.5 Copper Removal Mechanism Each material has different removal m echanisms as different materials being polished have different hardnesses and remo val chemistry. The chemical mechanical polishing (CMP) of copper several differences compared to CMP of oxide/low-k, barrier metals and tungsten. Copper with a hardness of ~ 3 GPa  is softer in nature and is easily abraded compared to other materials. The removal of copper basically takes place in about three steps viz., 1) ch emical reaction of slurry with copper to form surface layer. 2) Followed by mechanical abrasion of coppe r surface by abrasives pr esent in slurry. 3) Removal of the abraded copper material from the copper surface. The abraded particles are mostly washed away by the constant flow of slurry onto the pad. Figure 2.4 shows the schematic of copper removal mechanism. Figure 2.4 Schematic of Coppe r Removal Mechanism 
25 2.5.1 Formation of Surface Layer Initially when the wafer surface comes in to contact with slurry film it forms a native oxide, hydroxide films, Cu2O, CuO and Cu(OH)2 surface layer depending upon the slurry being used. Film growth on the surfaces being abraded is controlled by both growth kinetics as well as abrasion dynamics Copper CMP is generally performed using basic and acidic slurries. Removal rate of c opper is low in basic slurries compared to acidic slurries. Basic slurries usually consis t of ammonium hydroxide (NH4OH), 1 vol% NH4OH . While using basic slurries a Cu2O surface layer is formed. The removal rate is low in ammonium hydroxide based slurries. Even t hough high solubility of copper in ammonia based slurries the low removal rate is due to the formation of Cu2O layer on the surface. The surface layer acts as a ba rrier to etching and slows the dissolution of copper. Acidic slurries usually consist of nitric acid (HNO3). The acidic slurries generally have low pH values and at low pH values co pper doesnt form a surface layer. Studies have shown that as HNO3 concentration increases the etch rate also increases linearly. This suggests that there no form ation of a surface layer. The disadvantage of using acidic slurries is that the high etch rate results in a high removal rate in the recessed areas. The surface is susceptant to corrosion in acidic slurries. In order to avoid corrosion during polishing with acidic slurries a corrosion inhi bitor such as benzotriazole (BTA) is added . With addition of BTA a m onolayer of Cu-BTA is formed which protects the surface from corrosion.
26 Due to the corrosive nature of the acidic slurries surf ace defects such as dishing might take place in the recessed areas. The NH4OH based does not etch copper and therefore dishing might be only due to th e pad reaching into the recessed region. 2.5.2 Abrasion The abrasive component, whether delivered in a solution phase or abraded form a solid phase, generally provides the mechanical part of CMP. The abrasive particle impacts the surface and abrades the chemically treated surface exposing new material for chemical attack . The par ticles are either colloidal or non-colloidal in nature. The most commonly used abrasive particles are fumed, colloidal silica, aluminum oxide (Al2O3), ceria etc., Copper CMP does not require chemical activity from the particle. In general the size of abrasive particles may va ry between 25 nm 1 micron in diameter. 2.5.3 Removal of Abraded Material Once the material is abraded from the copper surface the abraded material must be removed from the vicinity of the surface so that it does not redeposit. Since the pad makes constant rotation the chance of the abra ded material returning back to surface is high. There have been several ways by which abraded material can be removed. The best method is by passing a constant flow of slurry, due the turbulent motion of the slurry the abraded material is washed away with cons tant inflow of new slurry. Also different oxidizing agents could be used with basic and acidic slurri es so that the copper polish rate may be increased by increasing the dissolution rate of abraded material. This primarily
27 serves two purposes by increasing the remova l rate and by removing the abraded material . 2.6 Pattern Geometry Effect s on Removal Mechanism Copper dishing and ILD erosion are th e two main undesired defects in CMP process. The dishing and eros ion occur during the over polish step which is required to ensure complete copper removal across the entire wafer. The copper dishing can be defined as the difference in height between the center of the coppe r line, which is the lowest point of the dish, and the point wher e ILD levels off. Dishing in copper mainly occurs due to bending of polishing pad slightly into the recess to remove the copper from within the recess. Dielectric erosion is a thinning of dielectric layer, resulting because of the reason that the polish rate of oxide is non-zero during the over polish time. The dishing leads to reduction in final thickness of copper line and resulting in complications while adding multiple layers of metal  the dishing increases with increase in metal line width. 2.7 Defects During CMP The chemical mechanical planarization process with many advantages also has some disadvantages in the form of surf ace defects caused during the planarization process. These defects cause complications while adding multiple layers of metal and other dielectric layers. Figure 2.5 shows some of the defects formed during metal CMP. The various defects formed dur ing CMP are listed below:
28 1. Dishing 2. Erosion 3. Metal Loss 4. Scratches 5. Wafer to wafer non-uniformity 6. Within wafer non-uniformity Figure 2.5 Defects Formed During CMP  2.7.1 Dishing Dishing during copper CMP occurs due to the difference in height between the center of the copper line, which is the lowest point of the dish, and the point where the oxide levels off, which is the highest point of oxide/dielectri c. Dishing occurs mainly due to the reason that the pad bends slightly into the recess/interconnect line to remove copper from within the recess. The dishing occurs during the over polish step, which is required to ensure complete copper removal across the wafer. Dishing is undesirable because it reduces the final th ickness of the copper line, affect s the electrical properties of the device and adds to complexity in adding additional layers of metal. The amount of dishing is dependent or is a f unction of line width. The dishi ng also occurs due to nature
29 of pad physical and mechanical properties. Ri gid pads are founds to cause less amount of dishing on patterned wafers. The removal rate of the metal is lower inside the recess because the pad does not exert the same pressu re on the recessed surface as on the flat metal surface . The dishing can be redu ced by controlling the over polish time closely. Figure 2.6 shows the schema tic of copper dishing. Figure 2.6 Schematic of Copper Dishing During CMP  2.7.2 Erosion Erosion is the thinning of oxide/dielectri c layer caused due to the difference in polish rate between copper and oxide/dielectric during the over polish step. It can be said as the difference in thickness before and af ter polishing. Oxide erosion is dependent on the pattern density of the feat ures under polish. The removal rate of oxide increases with increase in copper pattern density. This is du e to the reason that as the copper pattern density increases there is le sser amount of oxide availabl e to support the force and increase in oxide removal rate. With increase in pattern density tends to cause an increase in pressure on oxide thus increasing the oxide polish rate. Erosion during polishing is undesirable as it degrades the el ectrical properties of the devi ce and increases the process complexity while adding more layers over it and finally may short the device electrically.
30 The oxide erosion can be reduced or avoide d by reducing the copper over polish time . Figure 2.7 shows the schematic of an interl ayer dielectric (IL D) erosion profile. Figure 2.7 Schematic of Erosion Profile 2.7.3 Total Copper/Metal Loss The total copper loss or metal loss is the lo ss or removal of metal in the patterns of higher copper widths. The copper loss occu rs due to difference in removal rates between the adjacent materials and also due to pattern dependency. This causes to affect the device operating characteristics a nd is undesirable in a CMP process. 2.7.4 Scratches Scratches during polishing mainly occur due to the abrasive particles used. Since copper is softer in nature the abrasive part icle due to their prope rties and mechanical properties may cause scratches on the copper surf ace. The scratch can be identified by the depth and no of the scratches caused by the abra sive particles. In or der to avoid scratches abrasive particles softer in nature may be used. Scratches may have an affect on the overlying layers during processing and next step of polishing.
31 2.7.5 Wafer to Wafer Non-uniformity Wafer to wafer non-uniformity (WTWNU) is the difference in the amount of the polishing observed between wafer to wafer. This may be caused due to various process parameters controllable and uncontrollable. The WTWNU is caused due to decay in material removal rate with polishing more num ber of wafers on the same pad for a longer time . This may also due to the change in wafer surface properties. 2.7.6 Within Wafer Non-uniformity The within wafer non-uniformity (WIWNU) is due to the varia tion in material removal rate in CMP process. It has been ob served that even with a uniform down force the material removal rate is not uniform across the wafer . The non-uniformity in polishing may be caused due to degrading pa d polishing properties or may be due to nonuniformity in the applied down force across the wafer surface. 2.8 Summary With shrinking device dimensions and increasing RC delay there has been an extensive search for various new metals to replace aluminum as a interconnect material. Various materials as potential replacements were considered but copper a good electrical conductor with low resistivity, high current de nsity and high electromigration resistance has replaced aluminum as the interconnect material. A lthough copper possesses all the characteristics of a good electrical conduc tor it also has some disadvantages or complications. Unlike aluminum copper canno t be etched, which was overcomed by introduction of Chemical Mechanical Polish ing (CMP) where copper is etched due to
32 interaction of chemical and mechanical components. First the surface is altered chemically and then abraded away by the abra sive particles present in the slurry. Copper also diffuses easily into oxide/low -k dielectrics and in order to avoid this barrier layer has to be deposited. Copper can be deposited by various methods, such as CVD, PVD, electroplating and electr oless-plating. Copper with all ch aracteristics of a good electrical conductor has assumed a significant place in semiconductor manufacturing as an interconnect material. Copper CMP with its ad vantages also has some problems that are caused due to surface defects such as dish ing, erosion, scratches, wafer to wafer nonuniformity and within wafer non-uniformity. Th ese defects can be minimized by reducing the over polish time and by selectively choosing the polishing pads and other consumables.
33 Chapter 3 Experimental 3.1 CMP Testing Tool The experimental work for this research work has been carried on a CETR CP-2 CMP bench top polisher provided by CETR In c., USA. The CMP process on this bench top polisher closely imitates large wafer fa brication production processes. The CMP tester generates real time lateral and norma l force data. These signals can be monitored and analyzed in-situ. Figure 3.1 shows the pict ure of the CMP bench top polisher used at University of South Florida. The bench top tester has an upper and lowe r platen which can rotate at speeds from 0.001-5000 rpm. The lower platen is used to hold a polishing pad of 6 in diameter. The upper platen can hold a wafer upto 2 in diameter. The upper platen is connected to a vertical linear motor that has a travel length upto 150 mm. th e tester has ultra accurate strain-gauge sensors which can measure the load (0-10 Kg) and torque in two to six axes. The forces can be measured with a resolu tion of 0.1% and precise ly in the range of milligrams to kilograms with very high repeat ability. The upper platen is positioned on a slider assembly, which can slide in the X-Y direction. The tester has a fully automated PC based interface for motor control and data acquisition.
34 Figure 3.1 CMP Bench Top Tester 3.2 CMP Consumables The experiments were carried out on 2 cm X 2 cm patterned copper wafers having MIT 854 structure. The patterned copper samples had a barrie r and low-k layers beneath the copper. The samples were polished using two different kinds of polishing pad and copper slurry containing colloidal silica as abrasive particles. The polishing pads used were Pad A with SUBA IV subpad and Pad B. Subpads were used with both the polishing pads in order to reduce th e non-uniformity during polishing. Hydrogen peroxide (H2O2) was used as the oxidizer in the slurry for polishing. The pad was conditioned using deionized (DI) water for 10 minutes prior to polishing each sample to Slider Assembly AE Sensor Slurry Flow Force Sensor Wafer Carrier Polishing Pad Slurry Outlet
35 increase the roughness of the pad. The in itial thickness of copper on the unpolished samples was 1 m. 3.3 Experimental Parameters The consumables and the process paramete rs used for the polishing experiments are listed in table 3.1. Table 3.1 Consumables and Process Parameters for Polishing Parameters and Consumables Description 1. Polishing Pressure 2. Velocity 3. Slider Motion 4. Pad Conditioning Pressure 5. Slurry Flow Rate 6. Polishing Pad 7. Slurry 8. Oxidizer 9. Slurry Temperature 10. Wafer/Sample 3 psi 150 rpm (0.5 m/sec) 5 mm offset at a speed of 1 mm 1 psi 75 mL/min Two types of pads with k-grooves Three types of commer cial copper slurry (Two for first two series of experiments & other one for the third series of experiments) Hydrogen Peroxide (f or two types of slurries) and APS for third slurry. 15, 20, 25, 30 and 35 C 2 inch wafers with 10000 A copper over patterned low-k dielectric for dishing experiments with MIT 854 structure for dishing and metal loss experiments 3 inch wafers with 10000 A copper over patterned silicon dioxide for determination of device electrical properties 2 inch blanket coppe r wafers with 10000 A copper for determination of any structural changes
36 The patterned copper samples were 2 in ch wafers with MIT 854 structure and were polished using two different pads, pa d A and pad B with subpads. The pad A has perforations in it and pad B polishing pad has k-grooves in it. Pad B is softer in nature compared to pad A. The pads were conditi oned for 10 minutes prior to polishing each sample to ensure that the pad roughness remained constant throughout the polishing process. The pressure and platen velocity we re maintained constan tly at 3psi and 150 rpm through out the polishing. The sl ider was allowed to oscillate 5mm in X-Y direction at a velocity of 1mm/sec. The slurry temperature was varied at 15 C, 20 C, 25 C, 30 C and 35 C. The slurry temperature was controlled for each experiment during the whole project by monitoring and main taining the temperature at a specific value within a 0.2 oC variation using a temp erature controller. Temperatures below and above room temper ature were investigated as there is a very minimal probability of slurry temperatur e going down very low or very high; also as heat is generated during polishing to rubb ing of two surfaces th ere is a raise in temperature. As recommended by the manufacturer the temperatures above 40 C were avoided as the slurry would undergo changes chemically and would be rendered useless. Hydrogen peroxide (H2O2) was added to the slurry as an oxidizer. Figure 3.2 shows the process steps for the PMOS device fabricated for testing the electrical properties. The wafers used were 3 inch, type <100> orient ation and N-type silicon wafers. In order to make it a P-type device it was doped with a Ptype (boron) spin on dopant. The process was a three mask process, used for defining the diffusion areas, gate and regions for the metal interconnects. A thin layer of Tantalum of 25 nm thickness was used as a barrier and adhesion layer for copper. A 1 m thick copper layer was deposited using
37 electroplating method. The wafers were than polished and tested for the electrical properties. Figure 3.2 Process Flow Diagram fo r Fabricating the PMOS Devices 3.4 Estimation of Removal Rate During the polishing the coefficient of friction (COF) is continuously monitored in-situ at a sampling rate of 20 kHz and r ecorded. The COF at the pad-wafer interface, which changes between the wafer layers (met al, barrier and oxide) in contact with pad, indicates the material removal time and allo ws for calculation of material removal rate [52,53], friction force and coefficient in the wa ferpad interface, changes which allow for precise detection of the startpoint and end-point of remova l;  pad wear, temperature of the slurry and/or on the pa d; and contact acoustic emissi on in the waferpad interface, N-type Si Substrate Grow Field Oxide Define areas for diffusion Etch oxide Spin Boron Boron PredepDefine areas for Gate Oxide Etch oxide Grow Gate Oxide Define areas for metal interconnects Etch oxide Deposit Ta barrier layer Electroplate Coppe r Cu CMP & Test Device Field Oxide Boron Gate Oxide Copper
38 the average level of which allows for eval uation of polishing intensity and peaks allow for detection of wafer scratching and layer delamination processes.  The in situ monitoring of the coefficient of friction between the wafer and the pad provides understanding of the tr ibological mechanisms occurring at the interface and also facilitates precise end-point de tection of the polishing process. This method of end-point detection overcomes challenges posed by ot her methods like normal reflection, eddy current systems, etc. These other methods are applicable on ly to a limited set of material types used depending on the material optical an d electrical properties. As the coefficient of friction is specific to the material type and the characteristics of the interface, it changes as soon as the material type at th e interface changes upon the initial exposure of the underlying film. Thus, in s itu monitoring of the coeffi cient of friction during CMP provides precise end-point detect ion irrespective of the optical and electrical properties of the material being polished. The in situ monitori ng of the coefficient of fric tion between the conditioner and the pad facilitates precise detection of star t and end points of pol ishing and allows for optimization of the material removal process. The details and capabilities of the instrument are discussed elsewhere extens ively. [52, 55, 56] Figur e 3.3 shows the graph of COF vs. time used to estimate the removal rate. The removal rate (RR) in nm/min ( units) was calculated as follows (3.1): Removal Rate (RR) = Thickness of Copper Time taken to remove whole copper (3.1)
39 Figure 3.3 COF vs. Time Graph Used to Estimate Removal Rate Initially at each temperature a sample was polished for total removal and using the change in COF the removal rate was calculated. Using the calculated time for removal the experimental patterned copper samp les were polished for removal rate, COF, dishing and erosion values with a 10% over polishing time to ensure that the copper is completely removed. 3.5 Post CMP Analysis/Characterization Tools 3.5.1 Profilometer The Profilometer uses a diamond tip to scan a surface to get information about surface topography. Like any scanning probe, the tip has a finite surface area which interacts with the sample being scanned. Figur e 3.4 shows the schema tic of operation of a profilometer.
40 Figure 3.4 Schematic of a Profilometer Scan The tip scans across the surface of the samp le, and an inductive sensor registers the vertical motion of the tip. The signal ge nerated by the motion of the tip is used to create a two-dimensional profile of the su rface from which the step height can be calculated . 3.5.2 Atomic Force Microscopy (AFM) The atomic force microscopy (AFM) works similar to the same way as a profilometer works, only on a much, much smaller scale. The AFM consists of a microscale cantilever with a sharp tip (probe) at its end that is used to scan the specimen surface. The cantilever is typically made of silicon or silicon nitride with a tip radius of curvature on the order of about 10 nanometers When the tip is brought into proximity of a sample surface, forces between the tip a nd the sample lead to a deflection of the cantilever according to Hooke's law. The change in the vertical posi tion (due to repulsive or attractive Van der Waals force) reflects the topography of the surface. By collecting the height data for a succession of lines it is possible to form a three dimensional map of
41 the surface features. A tip or can tilever which drags across th e surface of the sample is specially constructed for surface structure investigation. Figure 3.5 shows the schematic of an AFM and figure 3.6 shows the SEM pi cture/schematic of an AFM cantilever/tip [33, 42]. Figure 3.5 Schematic of Operation of an AFM  Figure 3.6 SEM Image/Schematic of an AFM Cantilever Tip 
42 3.5.3 Scanning Electron Microscopy (SEM) The scanning electron microscope (SEM) is used to produce high resolution images of high magnification upto atomic le vel. In a SEM electrons are thermionically emitted from tungsten which acts as a cathode and are accelerated towards an anode; alternatively electrons can be emitted via field emission (FE). The electron beam, which typically has an energy ranging from a few 100 eV to 50 keV, is focused by one or two condenser lenses into a beam w ith a very fine focal spot sized in the order of 1 nm to 5 nm. The electron beam then passes through a pair of scanning coils in the objective lens, which deflect the beam in a raster fashion over a rectangular area of the sample surface. As the primary electrons strike the surface they are inelastically scattered by atoms in the sample. Through these scattering events, the primary electron beam effectively spreads and fills a teardrop-shaped volume, known as the interaction volume, extending from less than 100 nm to around 5 m into the surface. Interactions in this region lead to the subsequent emission of electrons which are then detected to produce an image . Figure 3.7 shows the basic block diagra m of a scanning electron microscope.
43 Figure 3.7 Block Diagram Showing the Operation of a SEM  3.5.4 Dynamic Mechanical Analysis (DMA) Dynamic mechanical analysis (DMA) is a technique th at is used to study and characterize materials. The DMA technique is mostly used for observing the viscoelastic nature of polymers. An oscillating force is applied to a sample of material and the resulting displacement of the sample is measur ed, from which the stiffness of the sample can be determined, and the sample modulus can be calculated. The damping properties of the material can be determined by measuring the time lag in the displacement compared to the applied force. Polymers with viscoelastic properties typically exist in two distinct states. These polymers exhibit the properties of a glass (h igh modulus) at low temperatures and those of a rubber (low modulus) at higher temper atures. This change of state, the glass Electron Gun Illuminating Lens System Scan Coils Final Lens Display To Pumps S p ecimen Detector
44 transition or alpha relaxation, can be observed during DMA run by scanning the temperature. The DMA technique is most wi dely used to investigate viscoelastic properties of CMP polishing pads as it can yi eld better data and can also be used to investigate the frequency dependant nature of the glass transition . The DMA operation of DMA is based on the equation 3.2 : G* = G + iG (3.2) Where G* is the complex modulus, G is loss modulus, i is imaginary root of -1 and G is storage modulus. Storage modulus (G) gives the elastic properties of the material and is measured directly. The loss modulus factor (iG) gives the details about the viscous properties of the material. The ra tio of loss modulus to storage modulus gives the tan delta or the damping ratio . The DMA were performed using a DMA 2980 in the single cantilever bending mode. The pad samples were fixed using th e DMA fixture clamps. The polyurethane pad samples with dimensions of 21.93x5.2x1.32 mm (pad A) and 21.83x5.75x1.37 mm (pad B) were tested using a multi frequency deformation mode at frequencies ranging from 1 100 Hz. The temperature was increased at a rate of 5 C/min and was tested from -150 C to 100 C. The samples were tested in the te nsion mode and liquid ni trogen was used for sub-ambient testing. Generally, in a DMA test storage modulus (G), loss modulus (G) and tan delta (tan ) are measured. Figure 3.8 show s the picture of the DMA 2980 instrument at USF.
45 Figure 3.8 DMA 2980 Instrument at USF  3.5.5 X-Ray Diffraction (XRD) X-rays are electromagnetic radiation similar to light, but with a much shorter wavelength. They are produced when electrical ly charged particles of sufficient energy are decelerated. Typically in an X-ray t ube, the high voltage maintained across the electrodes draws electrons toward a metal ta rget. X-rays are produced at the point of impact, and radiate in all directions.  If an incident X-ray beam encounters a crystal lattice, gene ral scattering occurs. Although most scattering interferes with itsel f and is eliminated, diffraction occurs when scattering in a certain direction is in phase w ith scattered rays from other atomic planes. Under this condition the reflect ions combine to form new enhanced wave fronts that mutually reinforce each other. The relation by which diffraction occurs is known as the
46 Bragg law or equation. Because each crystalline material has a characteristic atomic structure, it will diffract X-rays in a unique characteristic pattern. The characterization of material structure using X-ray diffraction follows the Braggs law described below: Sin d ns2 (3.3) Where d is lattice interplanar spacing of the crystal, is X-ray incident angle, is wavelength of characteristic X-rays. The basic geometry of an X-ray diffractometer (fig. 3.9) involves a source of monochromatic radiati on and an X-ray detector situated on the circumference of a graduated circle centere d on the specimen. Divergent slits, located between the X-ray source and the specimen, a nd divergent slits, located between the specimen and the detector, limit scattered (non-diffracted) radiation, reduce background noise, and collimate the radiation.  Figure 3.9 Schematic of an X-Ray Diffractometer. 
47 The detector and specimen holder are mech anically coupled with a goniometer so that a rotation of the detector through 2x degr ees occurs in conjunction with the rotation of the specimen through x degrees, a fixed 2:1 ratio. A curved-cry stal monochromator containing a graphite crystal is normally used to ensure that the detected radiation is monochromatic. When positioned properly just in front of the detector, only the K radiation is directed into the detector, and the K radiation, because it is diffracted at a slightly different angl e, is directed away. 3.6 Summary The CMP tool used for the experiments closely imitates large wafer fabrication production processes. The CMP tester generate s real time lateral and normal force data. These signals can be monitored and analy zed in-situ. The polishing experiments were carried on CETR Cp-2 bench top polisher. The patterned copper samples were polished at different slurry temperatures keeping the pressure and velocity throughout the experiments. Two different kind s of pads and one type of polishing slurry were used for polishing experiments. The patterned copper samples polished at different condition were analyzed for the amount of dishing and erosion using atomic force microscopy (AFM) and profilometer. The change in pad physical properties with increase in temperature was analyzed using dynamic mechanical analysis (DMA). The surface morphology of the pad before and after conditioning was analyzed using scanning electron microscopy (SEM). The pad roughness before and after conditioning was measured using the profilometer. The structural changes in the copper blanke t films was analyzed using X-Ray diffraction (XRD) analysis method for any change in crystallinity after polishing at different temperatures.
48 Chapter 4 Results and Discussion The values of coefficient of friction CO F and removal rates at different slurry temperatures were monitored and analyzed in -situ. The removal rate was calculated using the change in coefficient of friction (COF). The amount of dishing and metal loss during polishing was analyzed using profilometer and atomic force microscope (AFM). The change in pad physical properties with ch ange in temperature was analyzed using dynamic mechanical analysis (DMA) technique. 4.1 Dynamic Mechanical Analysis (DMA) of Pads The CMP polishing pad is subjected to high temperatures due to friction as a result of interaction between the polishing pad and wafer surface. This rise in temperature is partially brought down due to slurry flow. Tregub et al had invest igated the effect of rise in temperature due to friction. The studies showed that there is a temperature rise in the order of about 20 C to 30 C during CMP . During the polishing process the local pad temperature can increase significantl y, especially at the point of contacts between the trench edges and pad. The substa ntial increase in the pad temperature during the CMP process can change the physical and mechanical properties of the pad. The pads used for this re search were analyzed using DMA in a frequency range of 1 to 100 Hz. Figure 4.1 shows the storage m odulus and tan delta of the two polishing pads at different temperatures It could be seen that in the temperature range between 15
49 C to 35 C the storage modulus decreased by 27 % for pad A and by around 50 % for pad B. The decrease in the storage modulus as seen from figure 4.1, results in the change of pad hardness and as such the pad polishing performance . 0 500 1000 1500 2000 2500 3000 3500 4000 -200-150-100-50050100150 Temperature (oC)Storage Modulus (MPa)0 0.05 0.1 0.15 0.2 0.25 0.3Tan Delta Pad A Storage Mod Pad B Storage Mod Pad A Tan Delta Pad B Tan Delta Figure 4.1 Storage Modulus and Tan Delta vs. Temperature at 10 Hz Frequency In an ideal situation a polishing pad should show no change in the storage modulus within the range of operating temperat ure. With the increase in temperature the pad material softens and exhi bits lower storage modulus. Th e change in percentage of storage modulus within the operating temperature shows that with increase in temperature pad B softens more quickl y than pad A. The tan delta (tan ) gives the energy dissipation of the pad mate rials. The peaks from the tan give the information about the transition temperatur e (Tg). It could be seen th at pad A and pad B have subambient transitions at -130 C and -130 C, -7 0 C respectively. Also pad B has a sharper transition at 55 C in tan whereas pad A doesnt show any peaks below 100 C. This 55 C 70 C 130 C 15 35
50 sharp peak represents the glass transition temperature (Tg). The glass transition temperature represents a major transition, such as change in physical properties drastically as the material goes from hard glassy to a rubbery state. It could be interpreted that as the temperature increases pad B lo ses its physical properties drastically as compared to pad A. Thus, it could be stated that with in increase in temperature over the operating range pad B becomes softer than pad A, loses its stiffness. The area under the tan gives the ability of the pad to absorb the mechanical energy or its toughness. Figure 4.2 shows the ch ange in loss modulus with change in temperature. The loss modulus (G) is the product of the storage modulus (G) and tan 0 50 100 150 200 250 -200-150-100-50050100150Temperature (oC)Loss Modulus (MPa) Pad A Pad B Figure 4.2 Loss Modulus vs. Temp erature at 10 Hz Frequency 4.2 Scanning Electron Microscopy (SEM) Analysis The scanning electron microscopy (SEM ) analysis was performed on both kinds of polishing pads, before and after condition for same amount of time. The SEM analysis gives the surface topography of the pad, wh ich improves after the conditioning. The
51 pores present on the pad material open af ter the conditioning of the pad, enabling constant removal rate between sample to sample and also improves the uniformity in polishing within and also from wafer to wa fer. Figures 4.3 and 4.4 show the pad A and pad B respectively, before and after conditioning. Figure 4.3 SEM Image of Pad A a) Unconditioned b) Conditioned Figure 4.4 SEM Image of Pad B a) Unconditioned b) Conditioned From the figures it can be seen that the pas surface roughness improves after conditioning the pad. Also it can be seen th at after conditioning more number of pores open up. With the increase in pas roughness after condi tioning the same amount of a b a b
52 removal rate can be maintained from sample to sample and also the uniformity in polishing also increased with conditioning. 4.3 Analysis of COF and Removal Rate with Slurry Temperature and Pad 4.3.1 Coefficient of Friction (COF) The patterned copper samples were polis hed at various process conditions as mentioned in table 3.1. The graphs of COF vs. polishing time at five different temperatures on pad A using two types of slurries are shown in figures 4.5 4.9. Figure 4.5 COF vs. Time of a Sample at 15 C on Pad A Slurry-1 Slurry-2
53 Figure 4.6 COF vs. Time of a Sample at 20 C on Pad A Figure 4.7 COF vs. Time of a Sample at 25 C on Pad A Slurry-1 Slurry-2 Slurry-1 Slurry-2
54 Figure 4.8 COF vs. Time of a Sample at 30 C on Pad A Figure 4.9 COF vs. Time of a Sample at 35 C on Pad A The influence of slurry temperature on th e removal rate and coefficient of friction (COF) was analyzed. The change in the COF with change in slurry temperature is Slurry-1 Slurry-2 Slurry-1 Slurry-2
55 presented in graphs above. The graphs of COF vs. polishing time at five different temperatures and using two types of slur ries on pad B are shown in figures 4.10 4.14. Figure 4.10 COF vs. Time of a Sample at 15 C on Pad B Figure 4.11 COF vs. Time of a Sample at 20 C on Pad B Slurry-1 Slurry-2 Slurry-1 Slurry-2
56 Figure 4.12 COF vs. Time of a Sample at 25 C on Pad B Figure 4.13 COF vs. Time of a Sample at 30 C on Pad B Slurry-1 Slurry-2 Slurry-1 Slurry-2
57 Figure 4.14 COF vs. Time of a Sample at 35 C on Pad B Table 4.1 and 4.2 show the COF values of samples polished at five different slurry temperatures on two pads (pad A and pa d B) and two slurries. Three replicates of samples were taken at each slurry temperatur e. The average of the three replicates was taken to plot the COF vs. slurry temperat ure graph. Figure 4.15 and 4.16 show the graphs between COF at five different slurry temper atures taken on two different polishing pads and using slurry 1 and slurry 2 respectively. Table 4.1 Coefficient of Fric tion at Different Slurry Te mperatures Using Slurry-1 15 C 20 C 25 C 30 C 35 C 1 0.4682 0.48220.48240.4856 0.4878 2 0.4785 0.47510.48210.4864 0.4872 3 0.4587 0.47360.47950.4803 0.4869 COF on Pad A Average 0.4675 0.47690.48130.4841 0.4873 1 0.5048 0.51360.51990.5337 0.5449 2 0.5039 0.51430.52180.5348 0.5441 3 0.5044 0.51270.52280.5341 0.5435 COF on Pad B Average 0.5043 0.51350.52150.5342 0.5441 Slurry-1 Slurry-2
58 Table 4.2 Coefficient of Fric tion at Different Slurry Te mperatures Using Slurry-2 15 C 20 C 25 C 30 C 35 C 1 0.2241 0.25010.253 0.2607 0.2711 2 0.2446 0.24950.25140.2597 0.2704 3 0.2457 0.24980.25270.2601 0.2709 COF on Pad A Average 0.2448 0.24980.25230.2601 0.2708 1 0.2897 0.29510.30310.3217 0.3431 2 0.2888 0.29370.30250.3221 0.3425 3 0.2894 0.29370.30210.3211 0.3425 COF on Pad B Average 0.2893 0.29410.30250.3216 0.3427 0.45 0.46 0.47 0.48 0.49 0.5 0.51 0.52 0.53 0.54 0.55 0.56 10152025303540 Temperature (oC)Coefficient of Friction (COF) Pad A Pad B Figure 4.15 Change in COF vs. Slur ry Temperature for Slurry 1 From tables 4.1, 4.2 and figures 4.15, 4.16 it can be seen that for pad B the coefficient of friction (COF) increases with increase in slurry temperature with both slurries. For pad A the change in COF with in crease in slurry temper ature is very minimal for both slurries. This change in coefficient of friction can be attrib uted to increased area of contact of pad-wafer surface with increase in slurry temperature, resulting in higher
59 shear force at the interface, which can be supported by the change in COF with change in temperature (see fig. 4.15). The change in coefficient of fricti on during CMP employing pad 2 was found to be higher compared to pad A, which can be explained by the DMA analysis (see fig. 4.1) that with increase in temperature there is a change in pad hardness and material characteristics, resulting in increa sed area of contact. It can be seen that of the two pads used the second pads undergoes changes at a lower temperature and does expand elastically, explaining the reason behi nd the higher COF on one of the pads used. 0.2 0.22 0.24 0.26 0.28 0.3 0.32 0.34 0.36 10152025303540 Temperature (oC)Coefficient of Friction Pad A Pad B Figure 4.16 Change in COF vs. Slur ry Temperature for Slurry 2 4.3.2 Removal Rate Tables 4.3 shows the removal rate of patterned copper samples polished at five slurry temperatures on two different slurries. Three replicates of removal rate (nm/min) of copper samples were taken at each slurry temperature. The average of the three replicates was taken to plot the removal rate vs. slurry te mperature graph. Figures 4.17
60 and 4.18 shows the graphs between removal rate s at three different slurry temperatures taken on two different polishing pads. Table 4.3 Removal Rates at Different Slu rry Temperatures Using Slurry 1 and 2 15 C 20 C 25 C 30 C 35 C Slurry 1 190.5 222.2 255.3 285.7 333.3 Removal Rate on Pad A (nm/min) Slurry 2 137.9 153.8 160 173.9 218.2 Slurry 1 333.3 400 500 600 666.7 Removal Rate on Pad B (nm/min) Slurry 2 153.8 181.8 200 222.3 272.7 0 100 200 300 400 500 600 700 800 10152025303540Temperature (oC)Removal Rate (nm/min) Pad A Pad B Figure 4.17 Removal Rate vs. Slurry Temperature Using Slurry-1 From table 4.3 and figures 4.17 and 4.18 it can be seen that rem oval rate increases with increase in slurry temperature for pad A and pad B polishing pads. The removal rate
61 of copper is higher on pad B compared to pad A. This variation in removal rates on two polishing pads can be attributed to pad har dness and other pad mechanical properties. 100 120 140 160 180 200 220 240 260 280 30010152025303540 Temperature (oC)Removal Rate (nm/min) Pad A Pad B Figure 4.18 Removal Rate vs. Slurry Temperature Using Slurry-2 The increase in removal rate without any significant change in coefficient of friction during CMP using pad A is attributed mainly to the change in chemical reaction kinetics. Also it can be concl uded that the changes in the m echanical properties of pad B are not influencing the tribological mechanis m at the interface. The difference in the removal rates between the pads can be attr ibuted to the difference in mechanical properties of pads such as hardness and stor age modulus. Pad B was found to have higher removal rates with increase in temperature compared to pad A. The increase in removal rates and differences in removal rates on tw o types of polishing pa ds can be correlated with the coefficient of friction values, incr ease in COF aiding in increased removal rate along with increase in the metal dissolution rates with increase in slu rry temperature. The
62 increase in the coefficient of friction and ma terial removal rate with increase in slurry temperature compares well with literature . 4.4 Analysis of Dishing with Slurry Temperature and Pad The patterned copper samples polished at five different slurry temperatures were analyzed for the amount of dishing on 10 m, 20 m and 50 m wide metal lines and on 100 m square bond pads using profilometer a nd atomic force microscopy (AFM). Table 4.4 gives the values of the nor malized dishing data on the metal lines and the bond pads using two slurries. Figures 4.19 and 4.20 show the amount of dishing on a 10 m wide metal line at different slurry temperatures on different pads us ing two slurries. The amount of dishing (nm) has been normalized for analysis. Table 4.4 Normalized Dishing Data on Metal Lines and Bond Pads 10 m wide metal line 20 m wide metal line 50 m wide metal line 100 m bond pads Slurry Temperatur e (C) Pad A Pad B Pad A Pad B Pad A Pad B Pad A Pad B 15 C 0.0307 0.075 0.1194 0.1358 0.5881 0.5120 0.6336 0.481 20 C 0.0848 0.114 0.1441 0.1619 0.7586 0.6003 0.7108 0.564 25C 0.1364 0.1531 0.1839 0.191 0.8651 0.698 0.9132 0.651 30C 0.1657 0.1868 0.2146 0.225 0.9098 0.8218 1 0.747 Slurr y 1 35C 0.1502 0.1647 0.1927 0.2078 0.8825 0.7791 0.98 0.665 15C 0.0282 0.0986 0.0594 0.1425 0.483 0.8786 0.5204 1 20C 0.0184 0.0804 0.0409 0.1146 0.4005 0.8199 0.2134 0.933 25C 0.0116 0.0685 0.0294 0.0977 0.254 0.7981 0.167 0.889 30C 0.0084 0.0525 0.0238 0.0746 0.1615 0.7779 0.1543 0.843 Slurr y 2 35C 0.0078 0.035 0.020 0.0595 0.066 0.7592 0.1202 0.779
63 From table 4.3 and figure 4.19 it can be seen that the dishing on the 10 m metal line increased with rise in slur ry temperature from 15 C to 30 C and then decreased with further rise in slurry temperature to 35 C using slurry 1. The main reason for the change in dishing due to temperature can be attributed to the change in the properties of surface asperities of the pad with the change in slurry temperature. 0 0.05 0.1 0.15 0.2 0.2510152025303540Temperature oCNormalized Dishing Data Pad A Pad B Figure 4.19 Dishing on a 10 m Wide Metal Line vs. Slurry Temperature Using Slurry-1 Also with the increase in slurry temperat ure a change in chemical dissolution rate of copper  could be influe ncing the dishing characteristic s. It can also be seen from figure 4.19 that amount of dishing is higher on pad B compared to pad A on certain features and vice versa on others. This di fference in dishing characteristics as demonstrated by different pads suggests that there is more than one mechanism that can explain the dependence of dishing on rise in temperature. The high dishing on 10 m on pad B compared to low dishing on pad A (fig 4.19) could be due to local softening of the pad because of relatively high temperatures at the interface. The asperities of a softer pad (pad B) reach deeper into the metal lines of smaller widths compared to the harder pads
64 (pad A), resulting in increase of dishing. Th is change in pad physical properties with increase in slurry temperature is well expl ained with the help of dynamic mechanical analysis (DMA) of polishing pad, presented in the following sections. From figure 4.20 it can be seen that the dishing values on two t ypes of polishing pads decrease with increase in slurry temperature using slurry 2. 0 0.02 0.04 0.06 0.08 0.1 0.12 10152025303540Temperature oCNormalized Dishing Data Pad A Pad B Figure 4.20 Dishing on a 10 m Wide Metal Line vs. Slurry Temperature Using Slurry-2 This may be due to the reason that the slurry 2 had a different oxidizer (Ammonium per sulfate) and when compared w ith slurry -1 the oxidation rate is low and also reactivitys of both the slur ries with pad is different. Also the chemical nature of the slurries is suspected to play an important ro le in determining the change in dishing with change in temperature. It can be noted that also with slurry 2 the pad B has higher amount of dishing when compared to pad A. Figures 4.21 and 4.22 show the normalized dishing on a 20 m wide metal line with increase in slurry temperature. Howe ver, as the line width increases, on a 20 m
65 metal line (fig 4.21, 4.22) the amount of dishi ng increases with increase in line width compared to 10 m interconnect line. As w ith the 10 m interconnect line the dishing with pad B is higher than pad A on 20 m interconnect line. 0.1 0.12 0.14 0.16 0.18 0.2 0.22 0.24 10152025303540Temperature (oC)Normalized Dishing Data Pad A Pad -B Figure 4.21 Dishing on a 20 m Wide Metal Line vs. Slurry Temperature Using Slurry-1 0.018 0.028 0.038 0.048 0.058 0.068 0.078 0.088 0.098 0.108 0.118 0.128 0.138 0.148 10152025303540Temperature (oC)Normalized Dishing Data Pad A Pad B Figure 4.22 Dishing on a 20 m Wide Metal Line vs. Slurry Temperature Using Slurry-2
66 Figures 4.23 and 4.24 show the normalized dishing on a 50 m wide metal line with increase in slurry temperature. Howe ver, as the line width increases, on a 50 m metal line (fig 4.23, 4.24) the amount of dishi ng increases with increase in line width compared to 10 and 20 m interconnect lines. As compared with the 10 and 20 m interconnect lines the dishing with pad A is higher than pa d B on 50 m interconnect line with slurry 1, but where as with slurry 2 the dishing with pad B is higher than pad A. 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1 10152025303540Temperature (oC)Normalized Dishing Data Pad A Pad B Figure 4.23 Dishing on a 50 m Wide Metal Line vs. Slurry Temperature Using Slurry-1 This might be due to the reason that th e pad asperities reachi ng deeper at higher temperature and also as pad A is stiffer than pad B, on larger line widths it might reach more deeper into the interconnect line cause more amount of dishing. With slurry 2 the pad B has dishing than pad A (see fig 4.24) and this might be due to the pad slurry interaction playing an important role. Thus this observation confirms that the amount of dishing changes with line width , nature of the pad and slurry. Also this confirms that the rather than only the effect of temperatur e there is a combinator ial effect of slurry
67 temperature, slurry type and pad on the dishing generated rath er than the effect of only one single factor. 0.06 0.11 0.16 0.21 0.26 0.31 0.36 0.41 0.46 0.51 0.56 0.61 0.66 0.71 0.76 0.81 0.86 10152025303540Temperature (oC)Normalized Dishing Data Pad A Pad B Figure 4.24 Dishing on a 50 m Wide Metal Line vs. Slurry Temperature Using Slurry-2 Figures 4.25 and 4.26 shows the normalized dishing on a 100 m bond pads with increase in slurry temperature using two slu rries. The dishing increases with increase in temperature from 15 C to 30 C and then decr eased with rise slurry temperature to 35 C using slurry 1, whereas with slurry 2 the amou nt of dishing decreases with increase in slurry temperature.
68 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1 1.05 1.1 10152025303540 Temperature (oC)Normalized Dishing Pad A Pad B Figure 4.25 Dishing on a 100 m Bond Pad vs. Slurry Temperature Using Slurry 1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 10152025303540Temperature (oC)Normalized Dishing Dtaa Pad A Pad B Figure 4.26 Dishing on a 100 m Bond Pad vs. Slurry Temperature Using Slurry 2 Dishing increases with incr ease in line width is also confirmed again. The above phenomenon shows that softer pads get into the smaller metal lines more easily than stiffer or harder pads and cause more def ects. Pad B which is less stiffer in nature
69 compared to pad A, confirmed by the DMA analysis tends to get into smaller metal lines and removes more amount material than pad A. With increase in line widths pad A gets easily into metal lines and removes more amoun t of material even though pad B also gets into metal lines because due to the softer na ture of pad B it causes less amount of dishing. Figure 4.27 shows the atomic force microscope (AFM) dishing profiles of metal lines of different widths. Figure 4.27 Dishing Profiles of Meta l Lines of Different Widths 4.5 Analysis of Metal Loss with Slurry Temperature and Pad The patterned copper samples polished at different slurry temperatures were analyzed for the amount of metal loss on me tal lines of 100 m line width and 99 % pattern density as a function of slurry temperature and pa d type using profilometer and atomic force microscopy (AFM). Figures 4.28 and 4.29 show the metal loss at different slurry temperatures on two types of polishing pads using two types of polishing slurries.
70 It can be seen from figure 4.28 that the amount of metal loss is higher with pad A than with pad B. The area of the feature on whic h the metal loss was measured has dimensions in the order of 1500 m. The r eason for higher metal loss with pad A could be that the reaching of asper ities of pad A well into the metal lines as well as the dielectric material is the dominating factor as compared to the ha rdness of the pad. From this we can infer that pad asperity interaction with the feature and the bulk hardness of the pad both have a combinatorial effect on dishing. 0.3 0.33 0.36 0.39 0.42 0.45 0.48 0.51 0.54 0.57 0.6 0.63 0.66 10152025303540Temperature (oC)Normalized Metal Loss Dataa Pad A Pad B Figure 4.28 Metal Loss vs. Slurry Temperature Using Slurry 1 Figure 4.29 shows the metal loss on the two types of polishing pads using slurry 2. The amount of metal loss decreases with in crease in slurry temperature, pad B having more amount of metal loss than pad A
71 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 10152025303540 Temperature (oC)Normalized Metal Loss Pad 1 Pad 2 Figure 4.29 Metal Loss vs. Slurry Temperature Using Slurry 2 From the above observations it can be c oncluded that the slurry temperature has an effect not only on the surface defects a nd tribology but also on the change in pad physical and mechanical properties. Further it could be inferred that the surface defects and change in coefficient of friction (COF) is not solely due to the effect of slurry temperature but could be termed due to the in teraction factor of slurry temperature and pad properties. Also it can be observed that th e slurry type and the chemical constituents of the slurry along with temperature pl ays a major role in defect generation. 4.6 Analysis of Adhesion Failure and Mechanical Properties with Slurry Temperature The copper films on some samples got peel ed off during polishing with increase in temperature. Figure 4.30 shows the copper films after polishing at various temperatures. The films remained undamaged at temperatures in the range of 15-20 C and then started to peel with rise in temper ature. The most damaged films were at 35 C.
72 Figure 4.30 Pictures Showing Peeling on W afers Polished at Diffe rent Temperatures. A) Wafer Polished at 15 C. B) Wafer Polis hed at 20 C. C) Wafer Polished at 25 C. D) Wafer Polished at 30 C. E) Wafer Polished at 35 C This might be due to the reason that the wafers have a low-k material beneath the copper, which have low adhesion and mechan ical strength causing co pper to peel off. This proves that the material (low-k) prope rties degrade with in crease in polishing temperature. Further in order to investigate the structural changes if any with increase in polishing temperature using X -ray diffraction (XRD) analysis a series of experiments were conducted. These experiments were perfor med at three different temperatures (15, 25 and 30 C) using pad A and different copper polishing slurry. The wafers used were 2 inch blanket copper wafers having 10000 A thick copper films. The wafers were polished for different times depending on the temperature at which they were being polished so that same amount of copper is re moved from all the wafers. The removal rate and time was calculated using the coefficient of friction graphs as in first and second A B C D E
73 series of experiments. Figure 4.31 shows the XRD graph of the wafers polished at different temperatures and also on an as-is copper blanket wafer. The XRD spectra as collected were co mpared on unpolished (as-is) and polished samples. It can be seen from Fig. 4.31 that all the films exhibit X-ray reflection from Cu (111), (200), (220), and (222) planes. For all the films, strongest x-ray reflections are visible from Cu (111) planes. The intensity of reflections from Cu (220) is second highest. This indicates that crystallizati on occurs preferentially in (111) and (220) directions. In order to investig ate the evolution of crystallit e orientations due to polishing at different temperatures the peak intensitie s can be compared. It can be seen from the figure that that the evolution of peaks in all th e orientations increases with the increase in slurry temperature. The evolution is so promin ent with temperature that the peak in (2 1 1) direction does not even ex ist in the sample polished at 15 C slurry temperature. Figure 4.31 Graphs Showing XRD Peaks on Copper Wafers Polished at Different Temperatures
74 Also from the XRD measurements as show n in figure 4.31, an additional peak at 32 deg diffraction angle was observed on the sample polished at 30 oC temperature. This indicates that the copper crys tallinity increases with increasing polishing temperature. As the crystallinity increases, the mechanical characteristics improve as well and this was confirmed with the hardness tests conducte d on the samples using a nano-indenter. Figure 4.32 presents the ha rdness of the unpolished film and for the films post CMP at different temperatures. From figure 4.32 it can be noted that the hardness of the polished copper surface increased with incr easing slurry temperature. The unpolished sample had significantly lower hardness than the polished ones. This change can be attributed to the work hardening phenomenon during CMP. The work hardening phenomenon appears to be more influential at elevated temperatures. Similarly the modulus of the polished and unpolished th in films along the penetration depth is presented in figure 4.33. It can be seen that the modulus of elasti city increases with increasing slurry temperature. Also, the modul us of elasticity of the unpolished film is significantly lower than that of the polished samples. The numerical data from the nanoindentation experiments is presented in ta ble 4.5 which presents in detail the change in mechanical properties of thin films with a change in slurry temperature. Table 4.5 Mechanical Propert ies of Copper Thin Films Before and After Polishing Sample Modulus (Gpa) Hardness (Gpa) Unpolished Cu 118.51 3.72 1.22 0.06 Polished at 15 C 136.38 5.45 1.57 0.087 Polished at 25 C 136.29 4.77 1.66 0.093 Polished at 30 C 143.45 2.98 1.72 0.068
75 0 1 2 3 4 5 6 050100150 Displacement Into Surface (nm)Hardness (GPa) 15 C 30 C 25 C unpolished Figure 4.32 Hardness vs. Displacement into Surface for Unpolished and Polished Copper Samples 50 100 150 200 250 050100150 Displacement Into Surface (nm)Modulus (GPa) 15 C 30 C 25 C unpolished Figure 4.33 Modulus vs. Displacement into Surface for Polished and Unpolished Copper Samples
76 4.7 Analysis of Electrical Prop erties with Slurry Temperature In order to further investigate the effects of surface defects, due rise in polishing temperature on device operating ch aracteristics the second series of experiments were carried out. These experiments were carried on same types of polishi ng pads (pad 1 & 2) and with same slurry, other parameters remaining the same. The fabricated PMOS devices had similar feature sizes of 10, 20 & 50 m and bond pads of 100 m in dimensions as in first set of experiments. The dishing values and metal loss values were almost numerically similar to the values obtained in the first set of experiments. Then the polished samples were tested for the I-V characteristics using HP 4145B para meter analyzer. Figures 4.34, 4.35 show the transfer curves and sub threshold log plots I-V characteristics of the devices polished at different temperatures. For the drain characteristics the drain to source voltage (VDS) was swept from 0 to -15 V and the gate to source voltage (VGS) was increased in steps from 0 to -15 V and the corresponding values of drain current (ID) were recorded. In order to explain the shape of drain characteristics the curve may be divide d into four regions as shown in figure 13a. The first region in the curve shown as 0A is the ohmic region. In th is region the drain current increases linearly with increase in dr ain voltage, obeying Ohm s law. In the next region (AB) the drain current increases accord ing to the reverse square law. Here the drain current increases slowly as compared to the ohmic region. This occu rs due to the fact that with increase in VD, the ID increases. As a result of th is, the size of the depletion region becomes bigger, reducing the effective width of the channel. The region under the curve BC is the saturation or the constant current region. He re the drain current reaches
77 its maximum value and does not increase furthe r even with increase in drain voltage. The region beyond this is known as the breakdow n region, in the region the drain current increases rapidly as the drain voltage is in creased leading to br eakdown of the device. Figure 4.34 Graphs Showing Transfer Curves (I -V Characteristics) of Wafers Polished at Different Temperatures. A) Wafer Polished at 15 C. B) Wafer Polished at 20 C. C) Wafer Polished at 25 C. D) Wafer Polished at 30 C. E) Wafer Polished at 35 C It has already been observ ed that with increase in temperature the amount of surface defects i.e., dishing and metal loss also increase. From figure 4.34 it can be seen that wafers polished at temperatures 15 a nd 20 C operated well in accordance with the operating characteristics, further on wafers polished at higher temperatures 25 and 35 C the devices either reached the breakdown region at very low values of drain current or did not operate in the ohmic region, which are not the desired operating characteristics. Whereas on the wafer polished at 30 C, the wafer with highest amount of dishing values, the device failed to operate completely (see fig. 4.34). A B C D E A B C
78 The transfer curve also called as transconductance curves gives the relationship between drain current and gate voltage. For the sub threshold (t ransfer) characteristics the gate to source voltage (VGS) was swept from 0 to -25 V and the corresponding values of current (ID) were recorded. In order to explain the shape of drain characteristics the curve may be divided into three regions as shown in figure 14a. The first region is known as the threshold region. A minimum gate voltage is re quired in practical to produce a inversion layer, known as the gate threshold voltage (VG(th)). In a ideal situation when the gate voltage is less than the thres hold voltage no current flows fr om drain to source, but in practical a very less amount of current flows from drain to source. When the gate voltage is greater than threshold volta ge, the inversion layer connects the drain and source and we get significant value of current From figure 4.35 it could be s een that at wafers polished at lower temperatures 15 and 20 C the device the threshold voltage was around 3 volts, further on wafer polished at higher temperatur es 25 and 35 C the threshold voltage was in between 14-16 volts. Whereas on the wafer polished at 30 C, the wafer with highest amount of dishing values, the device failed to operate completely (see fig. 4.35). After crossing the threshold voltage the current increases sharply and then reaches the saturation point and then reaches the breakdown voltage. From these results it can be said that the increase in temperature increases th e surface defects which in turn results in device failure.
79 Figure 4.35 Graphs Showing Sub-Threshold L og Plot Curves (I-V Characteristics) of Wafers Polished at Different Temperatur es. A) Wafer Polished at 15 C. B) Wafer Polished at 20 C. C) Wafer Polished at 25 C. D) Wafer Polished at 30 C. E) Wafer Polished at 35 C From figures 4.34 and 4.35 it could be interpre ted that at wafers polished at lower temperatures (15 and 20 C) the device operated normally, further on wafer polished at higher temperatures (25 and 35 C) the devi ce properties degraded (see fig. 4.34 & 4.35). Whereas on the wafer polished at 30 C, the wafer with highest amount of dishing values, the device failed to operate co mpletely (see fig. 4.34). From these results it can be said that the increase in temperature increases th e surface defects which in turn results in device failure. A B C D E VG(Th)
80 4.8 Summary The effect of slurry temperature at three temperatures on two different pads were investigated, keeping the pressure and platen velocity constant. The patterned copper samples were characterized using the AFM a nd profilometer. The characteristics of two different kinds of pads were analyzed usi ng DMA. From the DMA analysis it could be inferred that pad softens with increase in temperature and pad B is softer in nature compared to pad A. Within the operating range there is greater percen tage of decrease in storage modulus of pad B compared to pad A. From SEM analysis it was found that pad conditioning improved the uniformity of polis h and also increased the roughness of the polishing pad. It was found that the COF on pad A increased with increase in slurry temperature and on pad B the variation in CO F with increase in slurry temperature was found to be statistically insignifi cant. This change in COF on th e pads can be attributed to the pad physical properties. The removal rates on both the pads increased with increase in slurry temperature. However the removal rate on pad A was higher than the removal rates on pad B. This increase in removal rate and the difference in the amount of removal rates on both pads can be attributed to the change in the rate of kinetics of chemical reactions and also to the mechanical properties of the pad. The dishing initially increased with increase in temperature upto 30 oC and then decreased with further incr ease in temperature using slurry 1, whereas the dishing decreased with increase in slurry temperature using slurry 2. This could be attributed to the change in the properties of surface asperiti es of the pad. This could be well explained with the DMA analysis of pad materials. It further concluded that the dishing was dependent on the interaction f actor of slurry temperatur e, slurry type (chemical
81 constituents of the slurry) and change in pad properties with temperature, than as initially thought would be due to slurry temperature. Also changes in copper structural properties were observed with increase in slurry temp erature. The device electrical properties degraded drastically with increasing slurry te mperature and even leading to device failure at higher polishing temperatures.
82 Chapter 5 Conclusions and Future Work The main objective of this research was to study the effect of slurry temperature on defect generation and tribology during c opper CMP. Surface defect generation on patterned copper wafers with increase in slur ry temperatures were tested using the CMP tester at different temperatures and with tw o types of polishing pads. Post CMP analysis of the polished samples was conducted using atomic force microscopy (AFM) and profilometer. The mechanical and physical prop erties of the polishi ng pads used were analyzed using dynamic mechanical analysis (DMA). Change in copper structural properties was studied using X-ray diffraction (XRD) analysis. Also the device electrical working characteristics were used on fabric ated devices polished at different slurry temperatures. Conclusions based on the res earch work and the re sults from previous chapters are presented here. 5.1 Conclusions 5.1.1 Friction (COF) and Removal Rate Studies The coefficient of friction (COF) on pad A and pad B increased with the increase in slurry temperature. This increase in coefficient of friction on pad A and pad B can be attributed to increased area of contact of pad-wafer surf ace with increase in slurry temperature, resulting in higher shear force at the interface.
83 The removal rate increased with increase in slurry temperature on both pad A and pad B. With increase in slurry temperature the removal rate on pad A was higher compared to pad B depending on the interconnect line width. This variation in removal rates on two polishing pads can be attributed to pad hardness and other pad mechanical prop erties. Also this variation can be attributed to change in the rate of kine tics of chemical reactions with increase in slurry temperature. 5.1.2 Dishing and Metal Loss The dishing and metal loss initially increa sed with increase in temperature from 15 C to 30 C and then decreased with further increase in temperature on both pads with slurry-1. This phenomenon of increase in dishing and metal loss initially and then decrease can be attributed to local softening of pad and change in pad physical and mechanical properties. The dishing decreased with increase in sl urry temperature using slurry-2. This phenomenon can be attributed to the chemical nature (oxidizer and other constituents of the slurry) of the slurry and its interaction with pad and also change in its chemical properties with temperature rise. The different trend in dishing pattern with two different types of slurries is due to change in pH of slurry with increase in temperature.
84 The amount of dishing on 10 m metal line was higher on pad B compared to pad A, but whereas with increase in the line width the dishing with pad A is higher than dishing on pad B. This variation in dishing with different ty pes of pads can be attributed to change in the properties of surface asperities of the pad and also change in chemical dissolution rate. The DMA analysis confirms that with increase in temperature the pad becomes softer. The amount of decrease in percentage of storage modulus of pad material is higher for pad B compared to pad A. But in an ideal case there should be no change in storage modulus for a polishing pad within the operating temperature range. Pad B has lower glass transition temperat ure (Tg) compared to pad A. After reaching Tg the pad transform from a glassy state into a rubbery state. Due to the softening of the pad with the increase in temperature the pad asperities reach deeper into the metal lines and cause dishing and metal loss. Dishing and metal loss were dependent on the interaction factor of slurry temperature and change in pad properties rather than only with the slurry temperature. Also the uniformity in polishing improved with pad conditioning. The conditioning also improved the roughness of the polishing pad.
85 5.1.3 Film Mechanical and Electrical Properties The films peeled off at higher temperatures At 35 C the copper films on some of the samples completely peeled off due to adhesion problems. The structural properties improved with in crease in slurry temperature. There was an increase in crystallinity of the coppe r thin films and also hardness and modulus increased with increase in polishing temperature. An extra peak at (2 1 1) was observed on the film polished at 30 C indicating a change in crystallinity. The electrical properties degrad ed drastically with increase in slurry temperature. At the highest temperature the devices even failed to operate. The devices with minimal or mo dishi ng operated perfectly and exhibited good IV characteristics. The devices/wafers polished at 35 C failed to operate completely. 5.2 Future Work The effect of interaction of other process conditions such as pr essure and velocity with slurry temperature can be investigated Also the effect of slurry temperature on surface defects and tribology can be studied us ing various acidic and basic slurries and also with various types of oxi dizers and concentration of oxidi zers. The effect of various particles sizes, particle con centrations and different kinds of abrasive particles can be studied. The effect of various orientati ons on pad surfaces such as grooves and perforations along with different kinds of pad materials can be investigated. Also various methods to reduce the temperature during polis hing can be studied in order to reduce the
86 surface defects. The electrochemical analysis on different slurries can be done to get a better understanding of change in slurry char acteristics with cha nge in temperature.
87 References 1. Joseph M. Steigerwald, Shyam P. Mura rka, Ronald J. Gutmann, Chemical Mechanical Planarization of Microelectronic Materials, John Wiley and Sons, Inc. New York, (1997). 2. J.G. Ryan, R.M. Geffken, N.R. Pouli n, J.R. Paraszczak, The Evolution of Interconnection Technology at IBM, IBM J. Res. Dev., 39, 371, (1995). 3. S. Wolf, Silicon Processing For the VLSI Er a, Vol. 4, Lattice Press, California, (2002). 4. Internet Website: http://www.itrs.net/Links/2005ITRS/Home2005.htm ITRS, (2005). 5. V. R. Kakireddy, R. Mudhivart hi, A. Kumar, P. Lefevr e, Proc. VLSI Multilevel Interconnect Conference, 050, 415-420 (2006). 6. J. M. Steigerwald, S. P. Murarka, R. J. Gutmann, D. J. Duquette, Materials Chemistry and Physics, 41, 217-228, (1995). 7. R. Carpio, J. Farkas, R. Jairath, Thin Solid Films, 266, 238-244, (1995). 8. Y. S. Obeng, J. E. Ramsdell, S. Deshpande, S. C. Kuiry, K. Chamma, K. A. Richardson, and S. Seal, IEEE Transactions on Semiconductor Manufacturing, 18, 4, (2005). 9. A. Ishikawa, H. Matsuo, T. Kikkawa, J. Electrochem. Soc. 152, (9), G695-G697, (2005). 10. Z. Li, K. Ina, P. Lefevre, I. Koshiyam a, A. Philipossian, J.Electrochem. Soc. 152, (4), G299-G304, (2005). 11. Y. Nomura, H. Ono, H. Terazaki, Y. Kami gata, M. Yoshida, Materials Research Society Symp. Proc., 816, (2004). 12. Z. Li, P. Lefevre, I. Koshiyama, K. Ina, D. Boning, and A. Philipossian, IEEE Transactions on Semiconductor Manufacturing, 18, 4, (2005). 13. H. Lu, B. Fookes, Y. Obeng, S. M achinski, K.A. Richardson, Materials Characterization, 49, 35-44, (2002).
88 14. H. Lu, Y. Obeng, K.A. Richardso n, Materials Characterization, 49, 177-186, (2003). 15. L. Wang, and F. M. Doyle, Proc. CMP Mu ltilevel Interconnections Conference, (2004). 16. J. Sorooshian, D. DeNardis, L. Char ns, Z. Li, F. Shadman, D. Boning, D. Hetherington, and A. Philipossian, J. Electrochem. Soc., 151(2), G85-G88, (2004). 17. P. A. Miranda, J. A. Imonigie, A. L. Erbe, A. J. Moll, IEEE WMED, (2005). 18. T. H. Tsai, Y. F. Wu, S. C. Yen, A pplied Surface Science, Article in press, (2003). 19. P. A. Miranda, J. A. Imonigie, and A. J. Moll, J. Electrochem. Soc., 153(3), G211-G217, (2006). 20. P. Leduc, M. Savoye, S. Maitrejean, D. Scevola, V. Jousseaume, G. Passemard, IEEE, (2005). 21. Q. Luo, D. R. Campbell, S. V. Ba bu, Thin Solid Films, 311, 177-182, (1997). 22. A. E. Miller, P. B. Fischer, A. D. Feller, K. C. Cadien, IEEE, 143-145, (2001). 23. T. Du, D. Tamboli, V. Desai, Mi croelectronic Engineering, 1, (2003). 24. S. Mudhivarthi, N. Gitis, S. Kuir y, M. Vinogradov, and A. Kumar, J.Electrochem. Soc. 153, (5), G372-G378, (2006). 25. S. Mudhviarthi et al, Electrochem Solid-State Lett., 8, (9), G241-G245, (2005). 26. L. Borucki, Z. Li, and A. Philipossian, J. Electrochem. Soc 151, G559 (2004) 27. Len Borucki, Leslie Charns and Ara Phili possian, J. Electroche m. Soc., 151, (12), (2004), G809-G813. 28. Z. Li, L. Borucki, I. Koshiyama, and A. Philipossian, J. Electrochem. Soc., 151, (7), G482-G487, (2004). 29. V. R. K. Gorantla, K. A. Assiongbon, S. V. Babu, a nd D. Roy, J. Electrochem. Soc., 152, (5), G404-G410, (2005). 30. V. R. K. Gorantla, A. Babel, S. Pandija and S. V. Babu, Electrochem. Solid-State Lett., 8, (5), G131-G134, (2005).
89 31. J. M. Steigerwald, R. Zirpoli, S. P. Mu rarka, D. Price, a nd R. J. Gutmann, J. Electrochem. Soc., 141, 2842, (1994). 32. F. Preston, J. Soc. Glass Tech., 11, 214, (1927). 33. Internet Website: www.wikipedia.org 34. S. Wolf, and R. N. Tauber, Silicon Pro cessing For the VLSI Era, Vol. 1, Lattice Press, California, (2000). 35. Internet Website: www.pubs.acs.org American Chemical Society. 36. J. D. Reid, Fundamentals of C opper Electroplating, Semiconductor International, June 1998, p. 94. 37. Internet Website: www.mems-exchange.org 38. V. M. Dubin et al, Proc. VLSI Multi level Interconnect Conference, p. 315, (1995). 39. P. B. Zantye, A. Kumar and A.K. Sikder Chemical mechanical planarization for microelectronics applications, Materials Science and Engineering: R: Reports, 45 (3-6) pp. 89 (2004). 40. H. H. Uhlig, Corrosion and Corrosion C ontrol, John Wiley and Sons Inc., New York, (1985). 41. M. R. Oliver, Chemical Mechanical Pl anarization of Semiconductor Materials, Springer Series in Material s Science, New York, (2004). 42. Internet Website: http://nnrc.eng.usf.edu University of South Florida. 43. Internet Website: http://acept.la.asu.edu/PiN/rdg/elmicr/elmicr.shtml Arizona State University. 44. I. Li, K. M. Forsthoefel, K. A. Richar dson, Y. S. Obeng, W. G. Easter, and A. Maury, Mat. Res. Soc. Symp., 613, (2000). 45. Kevin P. Menard, Dynamic Mechanical Analysis, CRC Press Inc., Florida, (1999). 46. A. Tregub, M. Moinpur, and J. Sorooshian, Effect of Temperature on Thermoanalytical Propertie s of Polishing Pads. 47. J. Luo, D. A. Dornfeld, IEEE Transactions on Semiconductor Manufacturing, 14 (2), 112-133, 2001.
90 48. Internet Website: http://uweb.cas.usf.edu/~harmon University of South Florida. 49. P. Singer, Copper CMP: Taking aim at di shing, Semiconducto r International, October 2004. 50. G. Fu and A. Chandra, J. Electr onic Materials, 31 (10), 1066-1073, 2002. 51. Internet Website: http://pubs.usgs.gov/of/2001/of01041/htmldocs/images/xrdschem.jpg 52. N. Gitis, Semiconductor Fabtech 18, 125 ( 2003 ) 53. N. Gitis, U.S. Pat. 6,494,765 ( 2002). 54. N. Gitis, U.S. Pat. 6,702,646 (2004). 55. N. Gitis and M. Vinogradov, in Proceedings of 2nd ICMI Santa Clara, CA (2001). 56. A. Sikder, A. Kumar J. Electron. Mater. 30, 1520 (2001).