<%BANNER%>

Chemical Mechanical Polishing of Copper Using Nanoparticle-Based Slurries

Permanent Link: http://ufdc.ufl.edu/UFE0010076/00001

Material Information

Title: Chemical Mechanical Polishing of Copper Using Nanoparticle-Based Slurries
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0010076:00001

Permanent Link: http://ufdc.ufl.edu/UFE0010076/00001

Material Information

Title: Chemical Mechanical Polishing of Copper Using Nanoparticle-Based Slurries
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0010076:00001


This item has the following downloads:


Full Text

PAGE 1

CHEMICAL MECHANICAL PO LISHING OF COPPER USING NANOPARTICLE-BASED SLURRIES By SU-HO JUNG A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2005

PAGE 2

Copyright 2005 by Su-Ho Jung

PAGE 3

This study is dedicated to my family.

PAGE 4

iv ACKNOWLEDGMENTS I would like to sincerely a ppreciate my advisor, Dr. Rajiv Singh, for his guidance and support throughout this study. Not only his fi nancial support help ed me pursue this research work but also his sincere dedicat ion and great enthusiasm to advance this research area have inspired th is research much. I also woul d like to acknowledge Dr. Brij Moudgil and Dr. Chang-Won Park for their cons tant and valuable discussions, during the weekly CMP meetings. I also would like to acknowledge the members of my advisory committee, Dr. Hassan El-Shall, Dr. Stephen Pearton, and Dr. Dinesh Shah, for their consideration and time to discuss this research. I would like to acknowledge the National Science Foundati on/Particle Engineering Research Center for financial support of part of this research. I also would like to acknowledge staff members of Particle E ngineering Research Center, Materials Analytical Instrumentation Center, and th e Department of Materials Science and Engineering: Wayne Acree, Edward Bailey, Mike Beasly, Jerry Bourne, Gill Brubaker, Valentin Craciun, Luisa Amelia Demper e, Anne Donnelly, Doris Harlow, John Henderson, Jennifer Horton, Eric Lambers, Sophie Leone, Martha McDonald, Gary Scheiffele, Kerry Sieben, Jo-Ann Standridge, and Mike Tollon, who helped me much in this research. Special thanks go to our secr etary, Margaret Rathf on, who took care of me like my family does, especially in a situati on in which I was far away from my family across a huge ocean.

PAGE 5

v Many thanks go to my old and present colleagues Jeremiah Abiade, Bahar Basim, Nabil Bassim, Scott Brown, Kyoung-Ho B u, Jaeyoung Choi, Won-Seop Choi, Madhavan Esaynaur, Chad Essary, Josh Howard, Fra nk Kelly, Won-Seok Kim, Vishal Kosla, Seung-Mahn Lee, Mike Ollinger, Karthik Rama ni, Danielle Stodilka, Suresh Yeruva, and Sang Yoon, who not only helped and encouraged me in this research but also made my graduate study years lots of fun. Finally I would like to thank my family members: my parents, Ok-Suk Jung and Hee-Jin Jung, who have given me this enormous love that made this thesis possible, and my sister, Su-Yeoun Jung, and her husband, Kyung-Hwan Kim, who have constantly cheered me with their support.

PAGE 6

vi TABLE OF CONTENTS page ACKNOWLEDGMENTS.................................................................................................iv LIST OF TABLES...........................................................................................................viii LIST OF FIGURES...........................................................................................................ix ABSTRACT......................................................................................................................x ii CHAPTER 1 INTRODUCTION........................................................................................................1 2 LITERATURE REVIEW.............................................................................................5 Need of Chemical Mechanical Polishing.....................................................................5 Multilevel Interconnect and Damascene...............................................................5 Challenges in Integrating Copper and Low k Dielectrics.....................................8 The CMP Process..................................................................................................9 The CMP Mechanisms........................................................................................13 Wafer-Slurry-Pad Interactions in Microand Nano-Scales........................................15 Real Area of Pad-Wafer Contact.........................................................................15 Fractional Surface Coverage...............................................................................18 Formation of Chemically Modified Surface Layer.............................................20 Indentation Depth of Single Particle ( W)...........................................................21 3 EXPERIMENTALS...................................................................................................24 Sample Preparation.....................................................................................................24 Wafer Preparation................................................................................................24 Particle Characterization and Slurry Synthesis...................................................24 Light Scattering ...................................................................................................25 Imaging Techniques.............................................................................................25 Slurry Preparation................................................................................................25 Table-Top Polishing Equipment.................................................................................26 Substrate Holder..................................................................................................26 Polishing Equipment Calibration........................................................................27 Polishing Experiments.........................................................................................28 In Situ Friction Force Measurements..................................................................29 Film Thickness Measurements...................................................................................30

PAGE 7

vii Four Point Probe Method....................................................................................30 Various Angle Spectroscopic Ellipsometry (VASE)..........................................32 Surface Topography Measurements: Atom ic Force Microscopy (AFM)...................33 4 SYNERGISTIC CHEMICAL-MECHANICAL EFFECTS.......................................35 Particle Characterization.............................................................................................36 Colloidal Silica Nanoparticles....................................................................................38 Micron-size Silica and 200 nm Alumina Particles.....................................................42 5 FORMATION OF REMOVA BLE SURFACE LAYER...........................................46 Experimentals.............................................................................................................47 Removal Rate and Static Etch Rate Measurements............................................47 Electrochemical Tests..........................................................................................48 X-ray Photoelectron Spectroscopy (XPS)...........................................................48 Nanoindentation and X-ray Reflectivity.............................................................49 Synergistic Chemical Effect.......................................................................................49 Dissolution / Passivation.............................................................................................50 Surface Chemical Composition..................................................................................52 Reaction Kinetics........................................................................................................53 Physical Properties of Surface Modified Surface.......................................................54 Discussion...................................................................................................................56 Conclusion..................................................................................................................58 6 ROLE OF NANOPARTICLE SIZE AND CONCENTRATION..............................60 Results and Discussion...............................................................................................61 Conclusion..................................................................................................................68 7 CONCLUSION...........................................................................................................70 LIST OF REFERENCES...................................................................................................72 BIOGRAPHICAL SKETCH.............................................................................................75

PAGE 8

viii LIST OF TABLES Table page 3-1 Comparison of process variables a nd slurry ingredients for calibration..................29 3-2 Process parameters for calibration...........................................................................29 4-1 Details of different types of particles.......................................................................37

PAGE 9

ix LIST OF FIGURES Figure page 2-1 Schematic of the interconnect that sh ows metal lines and insulating dielectrics.......6 2-2 Multi-level interconnect structure inte grated with copper and low k dielectrics.......7 2-3. Flow chart of interconnect fabrica tion by damascene proc ess (a) copper is deposited after the insulating dielectric is patterned with a trench (b) overburden copper is removed with CMP process (c) planarized single interconnect layer. These processes are repeated to build the MLI structure...........................................8 2-4 Schematic of typical CMP apparatus that includes wafer, pad, and slurry................9 2-5 Typical Pourbaix diagram. Boundary lin es indicate equilibrium between either a solid phase and an ion or two solid phases. Within the boundary, regions of corrosion, passivation, and immunity are shown.....................................................12 2-6 Schematic illustration of micros cale and nanoscale phenomena during CMP........15 2-7 Percent pad contact area on wafer surf ace as a function of applied load obtained by FTIR/ATR technique...........................................................................................17 2-8 Estimation of fractional surface c overage using in situ friction force measurements (a) in situ friction forc e as a function of solids loading for different down pressure and particle size (b) normalized fractional surface coverage of particles in contact with wafer surface, converted from (a).................20 2-9 The wafer-particle-pad interaction during polishing. This schematic shows a single particle that is tra pped between the pad and the wafer surface (a) when the indentation depth, w, of the particle is smaller than the surface layer thickness, t, and (b) when the indentation depth is greater than the surf ace layer thickness....23 3-1 Schematics of (a) table-top polishing equipment and (b) the substrate holder, bottom view and top view........................................................................................27 3-2 The polishing equipment calibration (a) down force calibration using weight balance (b) benchmarking of the tabletop polisher by comparing removal rate values as a function of pressure*velocity with data from open literature and commercial slurry.....................................................................................................28

PAGE 10

x 3-3 Schematic of apparatus for in situ friction force measurement................................30 3-4 Schematic of four point probe measurement............................................................31 3-5 Schematic illustration of spectroscopic ellipsometry...............................................32 3-6 Schematic diagram of atomic force microscopy......................................................34 4-1 Particle analysis (a) particle size di stribution obtained by light scattering method (b) image of 100 nm silica particle obtained by SEM (c) image of 200 nm alumina particle obtained by TEM (d) image of 1 m silica particle obtained by SEM..........................................................................................................................37 4-2 Removal rate of copper using 10 wt. % so lids loading of colloi dal silica particles at pH 7.0 (a) effect of hydrogen peroxide concentration in de-ionized water (b) effect of benzotriazole concentration in 5 % hydrogen peroxide.............................39 4-3 Removal rate of copper versus citr ic acid concentration, pH: 7.0. 100 nm silica particles are used. Open ci rcle indicates the removal rates measured with 5 % H2O2 and 10 mM benzotriazole. Open tr iangle indicates the removal rate measured in water.....................................................................................................39 4-4 Removal rate of copper as a functio n of down pressure w ith and without 100 mM citric acid. 5 % hydrogen peroxide, 10 mM benzotriazole 100 nm silica at 10 wt. %, pH: 7.0......................................................................................................40 4-5 Down pressure effect on surface image and roughness after polishing with 100 nm silica particles at 10 wt. %: (a) 1.5 psi, (b) 4.5 ps i, (c) 7.5 psi, (d) 9.0 psi.........41 4-6 Removal rate and surface roughness after polishing with various particles: slurry condition: 5 % H2O2, 10 mM BTA, 100 mM citric acid at pH 7.0..........................43 4-7 Surface roughness copper specimen (a) as received copper wafer (b) polished with 100 nm silica at 10 wt % (c) polished with 1 m silica at 10 wt. % (d) polished with 200 nm alumina at 5 wt. %................................................................45 5-1 Removal rate of copper vs. citric acid concentration in the presence of 5 % hydrogen peroxide and 10 mM BTA at pH 7...........................................................50 5-2 Effect of citric acid concentration on copper etch rate in various aqueous media; deionized water, 5 % H2O2, 5 % H2O2 and 10 mM benzotriazole at pH 7..............51 5-3. Effect of citric acid on polarization curves in various aqueous media at pH 7........52 5-4 (a) The XPS spectra of copper surface treated with 5 % H2O2, 10 mM benzotriazole, and 100 mM citric acid at pH 7 with different sputtering time, (b) O1s line, (c) C1s line................................................................................................53

PAGE 11

xi 5-5 Transient response of layer growth on copper in the millisecond regime (a) effect of citric acid in various media (b) effect of citric acid concentration in the aqueous media that contains 5 % H2O2 and 10 mM BTA.......................................55 5-6 Effect of citric acid addition on the ha rdness and the density of the surface layer; 5 % H2O2 and 10 mM benzotriazole at pH 7...........................................................56 5-7 Potential-pH diagram for copper-ci tric acid-water system. Redox potential values of 0.1 M citric acid solution and 0.1 M citric acid with 4 % hydrogen peroxide are also displayed......................................................................................58 6-1 Particle analysis (a) particle size dist ribution of colloidal s ilica obtained by light scattering method (b) image of 30 nm colloidal silica obtained by TEM (c) image of 100 nm colloidal silica obtained by SEM.................................................62 6-2 Effect of particle size and concentr ation on removal rate of copper with fixed chemical agents at pH 7...........................................................................................63 6-3 Effect of particle concentrati on on surface topography after polishing with 100 nm colloidal silica part icle (a) 3 wt. % (b) 5 wt. % (c) 10 wt. %.............................64 6-4 Effect of particle size on surface topography after polishing with 10 wt. % solids loading (a) 30 nm (b ) 80 nm (c ) 100 nm..................................................................65 6-5 Effect of particle size and con centration on surface roughness of copper specimen after polishing...........................................................................................65 6-6 Effect of particle size and con centration on friction force during copper polishing (a) without citric ac id in the slurry (b) with addition of 100mM citric acid........................................................................................................................... 66 6-7 Schematic illustration of nanoscale pa rticle-wafer surface in teractions. Dashed line shows the boundary of surface layer formed by chemical agents during polishing. The figure shows that the bigger pa rticles (100 nm) can have larger indentation volume on the surface layer whic h leads to higher friction force and more efficient material removal...............................................................................68

PAGE 12

xii Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy CHEMICAL MECHANICAL POLISHING OF COPPER USING NANOPARTICLE BASED SLURRIES By Su-Ho Jung May 2005 Chair: Rajiv K. Singh Major Department: Materials Science and Engineering Chemical mechanical polishing (CMP) is a vital step for planarizing multi-level interconnect structures in ultr a large-scale integrated circ uit applications. The CMP has become the fastest growing semiconductor ma nufacturing operation in the past decade and is expected to continue its high growth rate with the emergence of next generation interconnect materials such as copper and ultralow dielectric constant insulators in the coming decade. However, these next generation interconnects, due to their fragility and poor adhesion, are susceptible to CMP-induced defect formation such as microscratches, copper and barrier peeling, low k damage, di shing, and erosion. The state-of-the-art slurries presently designed for polishing c opper/silica dielectric use hard aggregate particles (fumed alumina, 100-300 nm in diamet er), which, we believe, may not be easily extended to polishing of copper/low k or ultra low k dielectrics. In this study, we investig ate copper CMP using nanopart icle based slurries to reduce the defect formation. The reduction of defect formation, however, is among other

PAGE 13

xiii considerations such as high removal rate. We examine the nanoscale synergistic chemical and mechanical interactions to determine c ontrolling factors in defectivity and removal rate. Our experimental results indicate that the synergistic effect that is, the rapid formation of surface passive layer that can be subsequently removed by the nanoparticles without deforming underlying bare copper, is needed to obtain the gentle copper CMP. The removal rate is synergistic, but more dominated by the chemical reaction than by the mechanical abrasion. The formation mechan ism of the removabl e surface layer is investigated. It is suggested that the enhan ced the reaction kinetics of the layer formation by addition of chelating agent in the slurry leads to a less dens e oxide layer on copper surface that can be removed by the nanopartic les. The role of nanoparticle size and concentration is also studied to understand in what manner the material removal occurs. The results show that the indentation volume of the particle onto th e surface layer plays an important role in material removal.

PAGE 14

1 CHAPTER 1 INTRODUCTION The constant device miniaturization driv en by semiconductor industry in the last decade has required advanced interconnect sc hemes. As the device gets smaller and smaller, the delay of interconnects that wi re the transistors no n-linearly increases, whereas the delay of gates th at switch the transistors significantly decreases. Below a micrometer feature size, the interconn ect delay measured by RC (resistance x capacitance) of a circuit is much larger than the gate delay and becomes a rate limiting factor for the total device speed. To reduce the interconnect delay, new multilevel interconnect (MLI) schemes, which span several planes, have been deve loped. To further reduce the interconnect delay, the conventional interc onnect materials such as aluminum (metal) and silica (dielectric) can be replaced by low resistiv ity metal such as copper and low dielectric insulating materials such as fluorinated si lica, carbon-doped silica, and polymers. By integrating copper and the low k dielectrics, the MLI structur e can provide (i) significant reduction in interconnect delay that leads to increased device speed (ii) enhanced electromigration resistance, and ( iii) reduced number of layers. Due to the lack of dry gas phase etching chemistries for copper, however, chemical mechanical polishing (CMP) is the only method for nanomachini ng the copper-based MLI. The method known as damascene is utili zed to fabricate the copper-based MLI, which deposits excess amount of copper on diel ectric patterned with trenches and then

PAGE 15

2 employs the CMP step to rapidly remove th e overburden copper a nd globally planarize the structure. The advent of copper and the low k dielectr ics as interconnect materials has indeed accelerated the need of CMP process but also introduced a new set of challenges. These next generation interconnect materials ar e quite fragile and exhibit poor adhesion property and thus are very susceptible to CM P induced normal and shear stresses. It is also noted that mechanical strengths of di electric films significantly decrease with a decrease of dielectric constant s of the films. Consequently the defect formation such as microscratches, copper-barrier peeling, low k damage, dishing, erosion, and so on have been often observed during the CMP process. Such defect formations not only negatively impact the successive process loop but also are ultimately detrimental to the device performance [Sin03, Wan01], which however still remains as one of the biggest challenges of the CMP for copper an d low k dielectric integration. To reduce the defect formation, various i nvestigations such as reduction in down pressure [Kon03], abrasive-free polishing [P ad03], electro-polishing [Cha03], and so on have been conducted. The challenging aspects of these approaches towards the gentle copper CMP, however, appeared to be balanc ing with other demanding output parameters such as removal rate and planarity. This is primarily because these approaches are based on the cause-and-effect relationship and are lack of thorough understa nding of the timedependent wafer-slurry-pad inte ractions that are complicat ed by the large number of slurry variables and thei r synergistic effects.

PAGE 16

3 To mitigate these challenges, we have earlier proposed a mechanistic methodology of understanding the wafer-pad-slurry inte ractions, namely microand nano-scale interactions [Sin02]. Mechanical stresses during the CMP process can be related to particle indentation (penetration) into the wafer surface. According to the cont act mechanics theory [Zha02, Qin04], the particle indentat ion during the CMP process may depend upon (i) particle properties, (ii) properties of chemically modified surface layer, and (iii) pad modulus. Hardness and size of the particle can be reduced to lower th e penetration, thus avoiding the damaging effects. The state-of-the-art slurries presently designed for polishing copper/silica use hard aggregate particles (f umed alumina, 100-300 nm in diameter). We believe that this may not be easily extended to polishing of copper/low k or ultra low k dielectrics. This thesis describes an investigation into the chemical mechanical polishing of copper using sub 100 nm silica particles. The objective of this work is to examine synergistic chemical-mechanical effects dur ing the copper CMP process. The influence of synergistic interactions on polishing ch aracteristics (or polis hing output parameters) such as defectivity and removal rate, and th e role of each individual component, chemical and mechanical, in the synergy are studied throughout. In chapter 2, more details of current understanding are reviewed. The experime ntal procedures used in this thesis are described in chapter 3. In chapter 4, s ynergistic chemical-mechanical effects on controlling the defect formation and remova l rate are examined. This chapter also compares and contrasts the results obtained by the nanoparticles with the results obtained by two other types of particles, one micron-si ze silica and 200 nm al umina particles. In

PAGE 17

4 chapter 5, the formation mechan ism of a surface passive layer that can be removed by the nanoparticles is investigated. In chapter 6, the removal mech anism by studying the role of nanoparticle size and concentrati on is investigated. The conclusi ons is given in chapter 7.

PAGE 18

5 CHAPTER 2 LITERATURE REVIEW Chemical mechanical polishing (CMP) is a technique that removes excess material and planarizes an uneven surface. Alt hough adopted by semiconductor industry for silicon wafer polishing around 40 years ago, the technique has long been used to obtain a mirror surface. In the last decade, the CMP process has become a vital step for fabricating microelectronic devices in u ltra large-scale integrated circuit (ULSI) applications, such as shallow trench isolat ion (STI) and multi-level interconnects (MLI). This enabling process to significantly redu ce topographical varia tion and to provide excellent surface finish is also rapidly stretc hing to other numerous applications such as micro-electro mechanical systems (MEMS), compound semiconductors, and ferroelectric and magnetic random access memories (FeRAM). Need of Chemical Mechanical Polishing Rapid advances in ultra large-scale integrat ed circuits (ULSI) fo r miniaturized chip size and improved performance have raised demand for enhanced metallization schemes. Multilevel Interconnect and Damascene Fig. 1 illustrates the schema tic of interconnect that in cludes metal lines, insulator and substrate. The delay of interconnect is represented by RC (resistance x capacitance) of the circuit components. By taking into account the metal resist ance and capacitances (metal-metal and metal-substrate), the tota l RC delay can be expressed as following,

PAGE 19

6 D Mt t wd L RC 1 12 0 (2.1) With decrease in the feature size (< 0.25 m), the first term in the parenthesis (1/ wd ) becomes dominant. The RC delay of a circ uit increases non-linearly and is much larger than the delay in switching the transi stor (gate delay) [Rya95]. To reduce the RC time delay, new multilevel interconnect (MLI) schemes, which span several planes and are isolated by dielectric insu lating layers, have been deve loped. To further reduce the RC delay, low resistivity ( ) materials such as copper (alternative for aluminum) and low dielectric constant ( ) materials such as fluorinated silica, carbon-doped silica and polymers (alternative for silicon dioxide) can be used [Sho00]. Therefore, the MLI scheme based on Cu/low-K promises significant improvements in operating the miniaturized devices. Figure 2-1 Schematic of the inte rconnect that shows metal lines and insulating dielectrics Due to the lack of dry gas phase etching chemistries for copper, chemical mechanical polishing (CMP) has been almost th e sole method to achieve planarization of

PAGE 20

7 copper and eventually to build the MLI stru cture. This method known as the damascene process deposits blanket metal on dielectric which has been patterned with trenches and then employs CMP to polish the top con tinuous overburden metal layer [Mur00]. To make very thin copper lines, the following steps are conducted: (i ) deposition of low k dielectric (or silica) layer, (ii) patterning of the silica (or low k diel ectric layer) by dry gas phase etching, (iii) deposition of a thin Ta (or TaN layer) wh ich acts as an adhesion layer and also a diffusion barrier, (iv) deposition of a thin copper laye r as seed layer by physical vapor deposition, (v) thick coppe r deposition by electroplating, and (vi) chemical mechanical polishing to remove the overburden regions so that nanoscale copper lines are exposed (Fig. 2-3). Figure 2-2 Multi-level interconnect (MLI) stru cture integrated with copper and low k dielectrics (fluorinat ed SiO2) [Intel].

PAGE 21

8 Challenges in Integrating Copper and Low k Dielectrics The next generation interconnect material s such as copper and low k dielectrics, due to their fragility and poor adhesion, are susceptible to CMP-i nduced contact stresses (normal and shear) and defect formation such as microscratches, copper and barrier layer delamination, low k damage, dishing, and erosion [Sin03, Wan01]. Figure 2-3. Flow chart of interconnect fabric ation by damascene proc ess (a) copper is deposited after the insulating dielectric is patterned with a trench (b) overburden copper is removed with CM P process (c) planarized single interconnect layer. These processes are repeated to build the MLI structure. Principle of Chemical Mechanical Polishing

PAGE 22

9 The CMP Process A schematic diagram of the CMP process is shown in Fig. 2-4. A typical rotary CMP tool consists of a polishing pad affixed to a circular polishing plate, a carrier to hold a wafer against the pad, and slurry. Both the ca rrier and the plate are rotated as the front of the wafer is pressed down against the pa d, which is covered with the polishing slurry. The polishing slurry provides the mean s by which both chemical and mechanical (tribological) actions are used to remove and subsequently planarize the wafer surface. A typical CMP process mainly consists of th ree components; wafer, slurry, and pad. As illustrated in Fig. 2-4, the wafer is pressed against the pad with a normal force and moves along the spinning pad with a relative velocity. The polishi ng slurry that contains chemical additives and abrasive particles is fed onto the pad and travels between the pad and the wafer surface. The pad is normally made of porous polymer (polyurethane) and has several functions: (i) uniform slurry trans port and distribution, (ii) removal of reacted products (debris, etc.), and (iii) unifor m pressure distribution across the wafer. Figure 2-4 Schematic of typical CMP appara tus that includes wafer, pad, and slurry.

PAGE 23

10 The performance of CMP may be categor ized as removal rate, planarization, defectivity, and selectivity. Rem oval rate is related to throug hput issue of the process. For industrial perspective, fast removal rate is needed for mass production of wafers, which is usually estimated 400-600 nm/min for coppe r. The deposited overburden films on patterned surfaces have large la teral dimensions (currently 300 mm) and wide variations of pattern-dependent topography. The pitch of the pattern (the sum of the width of the patterned lines and the spacing between them) as well as its density can vary significantly across the die. The removal of such a co mplex structured surface requires global planarization. Defect formation during the process is another issue and become quite significant when next genera tion interconnects such as copper/low k dielectrics are integrated. A variety of def ects such as microscratches, metal-barrier layer peeling, dielectric damage, dishing, and erosion ar e often observed and must be minimized. Selectivity is defined as the ratio of top layer removal rate to underlying layer removal rate. In copper CMP, as seen in damascene process, the selectivity refers removal of copper vs. barrier (Ta or TaN) vs. dielectric (SiO2 or low k dielectrics). A high selectivity value is desired because the CMP process n eeds to stop once the top layer is removed. The performance quality of the CMP pr ocess is, of course, determined by optimization of the process variables includ ing the polishing slurry. One of the unique features of the CMP is that the material re moval and planarization normally take place via the synergistic interactions. These combin atorial effects that combine both chemical and mechanical effects are intuitively diffe rent from chemical dissolution-type or mechanical polishing-type of material remova l. In general, the chemical agents are responsible for surface reaction, while the abrasive particles are responsible for

PAGE 24

11 mechanical action. Although the to ol operation itself is relatively simp le, strict control and deep understanding of th e process are often complicated by the exceedingly large number of process variables (> 20), their synergis tic interactions, and the dynamic nature of the process. Since metal CMP involves electrochemical or igins, a Pourbaix diagram (Fig. 2-5) usually illustrates a concisely visual and ra tional approach to the slurry chemistry. The Pourbaix diagram is a member of a general class of thermodynamic diagrams that are termed aqueous stability or predominan ce area diagrams (potential vs. pH) [Pou74]. Slurries employed for planarization of metal typically include an oxidizer to form metal oxides, a corrosion inhibitor to prevent high st atic etch rates, and a complexing agent to enhance solubility of metals or to soften the layer. The dissolution and passivation processes resulted from the combination of these chemical additives are very competitive, and must be tunable to attain ideal planari zation. Therefore, it also should be pointed out that the chemically modified surface laye r with different char acteristics (composition, harness, thickness) may form in various slurries depending upon thermodynamic basis, the pH and electrochemical potential of the me tal in the slurry, and also kinetics of the various reactions that are thermodynamically favorable at the give n pH and potential [Ste97]. The state-of-the-art slurry chemicals employed for planarization of copper typically include hydrogen pe roxide as the oxidizer, be nzotriazole as the corrosion inhibitor, and citric acid or glycine as the complexing agent. The dissolution and passivation processes resulted from the combination of these chemical additives are very competitive, and must be tunable to attain ideal planarization. Ther efore, a chemically modified surface layer with different charac teristics (composition, hardness, thickness)

PAGE 25

12 may form in various slurries dependi ng upon thermodynamic basis, the pH and electrochemical potential of the metal in th e slurry, and also kinetics of the various reactions that are thermodynamically favorable at the given pH and potential [Lee03]. Figure 2-5 Typical Pourbaix diagram. Boundary lines indicate equili brium between either a solid phase and an ion or two solid phases. Within the boundary, regions of corrosion, passivation, and imm unity are shown [Pou74]. In conjunction with slurry chemistry, abrasi ves also play an important role in CMP performance such as material removal rate and surface finish. A variety of abrasive materials have been utilized in industry, wh ich typically include commercially available Al2O3 and SiO2 particles for metal CMP. A large number of characteristics are used to

PAGE 26

13 determine their roles in the po lishing process. The abrasive s are mainly responsible for mechanical action of the slurry. Thus th e most attention-grabbing issues for the semiconductor industry are their causes of surface scratching, denting, and peeling during polishing soft metal like coppe r [Sin03]. In addition, the la rge differences between the hardness of the metal and the slurry abrasi ve can further accentuat e special type of defects known as dishing and erosion th at results in non-planarity. These defects addressed are detrimental to both fabr icating and operating IC devices. Despite their importance, effects of ab rasive particle on CMP performance are somewhat inconclusive. Experimental results reported are quite contradicting [Mah99, Bie98, Zho02, Bou02, Lu03], and its roles in modeling work are not experimentally proven yet [Coo90, Qin04, Zha02]. This is primaril y due to limitations and difficulties of experimental approach that lead to lack of understanding the fundamental aspects of small-scale interactions that take place dur ing the polishing. It is very difficult to precisely resolve the individual effect because (i) a large number of other input variables is introduced, (ii) the entire process is exceed ingly dynamic, transient, and synergistic, and (iii) the actual removal takes place on a very small scale ( to nm). The CMP Mechanisms Due to its simplicity, Prestons law has been the most extensivel y used equation in describing the mechanism of CMP, which states that the removal rate of a material, that is, the thickness decrease ( h) over time ( t) is directly proportional to the applied pressure (P) and relative ve locity (V) of the particle s across the wafer [Pre27]. RR =h / t= KpPV (2.2)

PAGE 27

14 Prestons constant (Kp) depends on properties of the polis hing pad, slurry, and materials to be polished. Although Prestons equation ha s been a basis for illustrating general polishing, it does not provide profound insight of synergistic phe nomena (chemical and tribological dynamics of surface layer forma tion, contact mechanics, etc.) during CMP, due to its initial development fo r mechanical polishing of glass. The electrochemical effect and material removal mechanism in metal CMP were first proposed by Kaufmann [Kau91]. In metal CMP, the chemical action by the slurry chemical dissolves the metal surface and form s a passivating film preventing the isotropic chemical etching process on the wafer surface. By the mechanical action of the abrasive particles and the polishing pad, the passiva ted film is removed, achieving a degree of global planarization that is unmatched by the chemical etching process. In general, the dissolution rate of the metal surface was found to be two orde rs of magnitude lower than the polishing rate [Sin02]. Singh et al. [Sin02] outlined mechanistic methodology for understanding the waferpad-slurry interactions during the CMP proce ss, namely microand nanoscale effects. The large number of process variables with microand nano-scale effects and output parameters are summarized in Table 1. The micro-scale perception involves identification of polishing mode in which the particle base d slurry interacts with the comparatively rough pad and the flat wafer surface. It is genera lly believed that the majority of material removal takes place on a contact mode, that is the active numbers of particles that are trapped between the pad asperity and the wafer surface under certain down pressure participate mostly in material removal. It is therefore critical to be aware of the real area of wafer-pad contact, the fraction of partic les that cover the pad surface (fractional

PAGE 28

15 surface coverage), and so on. The nano-scale pe rception takes into account the formation of surface layer by chemical a dditives and its subsequent re moval by particle abrasion, which generally occurs in nanoor atomic scale within a very short time (~ millisecond). Therefore, it is important of understand the dynamics of surface layer formation (transient characteristics) and its physical interaction with th e abrasive particle such as indentation depth. Figure 2-6 Schematic illustration of mi croscale and nanoscale phenomena during CMP [Sin02] Wafer-Slurry-Pad Interactions in Microand Nano-Scales Real Area of Pad-Wafer Contact One of the distinctive features of CM P phenomena from conventional tribology is the contact of two solids, the pad and the wa fer. The pad is typically made of porous

PAGE 29

16 polymeric materials, which is more complia nt than the oxideor metal-wafers. The roughness of the pad can be characterized to be much higher than that of the wafer (the difference can be up to 3-4 orders of magnit ude). Therefore, when the pad and the wafer are in physical contact and pressed against each other, it is generally accepted that only certain portion of the pad can be deformed on the wafer surface. Fig. 2 illustrates the contact between a polishing pad and the wafer, first proposed by Yu et al. [Yu93]. Yu et al statistically analyzed roughne ss of the polishing pad using the Greenwood-Williamson elastic model [Gre66, Gre67]. The assumptions in the model were (i) spherical shaped asperity and (ii) Ga ussian distribution of va riations in asperity height and radius of the sphere. Assuming th e direct contact mode of polishing regime, the applied load can be carried solely by th e pad asperities. The total contact area (Acon) and load (L) over a nominal pad area (A) was calculated as d z condz d a A0 (2.3) d zdz d l A L0 (2.4) where is the asperity density, is the asperity radius, z is the asperity height, and z are the Gaussian distribution f unctions of the asperity radi us and the asperity height, respectively. The contact pressure is Pcon = L/ Acon. Fig. 3 shows the calculated Pcon and Acon against P. As shown in the figure, the pa d contact area increases with an increase of down load while the pad contact pressure is almost constant with the increase of down load due to the increase in the contact area. This result suggests that applied down pressure controls the polishing via the pa d contact area rather than the pad contact

PAGE 30

17 pressure. At 7 psi, Acon/A varies from 0.05 to 0.54 %, depending on the pad modulus (10100 MPa). Basim et al. [Bas03] experimentally verified this model using the Fourier transform infrared spectroscopy/attenuat ed total internal reflection spectroscopy (FTIR/ATR) technique. A sample pad (IC1000) was placed on the (ZnSe) ATR crystal and loaded from 3.4 to 11.4 psi. The intensities of the CH2 peaks of the polyurethane were recorded at wave number range 3000-2750 cm 1 by measuring the area under the detected peak. As shown in Fig. 4a, five spec tra were collected at each pressure level. The percent pad contact area was then determ ined by comparing these intensity values with the value of a 100 % pad contact with non-porous, defect-free and flat surface. The real contact area of the pa d (IC-1000, E=30-40 MPa [Oer85, Tic99]) was experimentally estimated as 0.33 % at 7 psi which is in good ag reement with the Yus calculated result. Figure 2-7 Percent pad (IC 1000) contact area on wafer surface as a function of applied load obtained by FRIR/ATR technique [Bas03].

PAGE 31

18 Fractional Surface Coverage As we assume in this work that the majority of material removal takes place on the direct contact mode, the number of partic les placed at the real area of the pad-wafer contact should be understood. Cook and Brown et al. [Bro81, Coo90] descri bed the fractional su rface coverage of particles during polishing using the c oncept of particle fill fraction ( K ) on the entire wafer surface. The model was premised on the flat and hard pad surface without asperity. Then the number of particles per unit area can be expressed by the following: 23 2 /K A N (2.5) K is unity for fully hexagonal close packing. The force per single particle when trapped between two hard surfaces can be expressed as follows: K P F 2 32 (2.6) where P is the polishing pressure, is the diameter of particle, and F is the force per individual particle. Choi et al. [Cho04] used in situ friction force measurements to estimate fractional surface coverage. To estimate the real number of particle-pad contac t, experiments were conducted with relatively soft silica abrasi ve particle (Vicke rs hardness = 540 kg/mm2) and hard sapphire wafer (2370 kg/mm2). The fractional surface c overage that is actually in contact with both the pad asperity and th e wafer surface is calculated by measuring the in situ friction force. According to the Amont ons law, friction force (F) is defined as product of the friction coefficient ( ) and normal force (FN). NF F (2.7)

PAGE 32

19 The normal force can be written as follows: PA FN (2.8) Combining Eq. (5) and (6), friction force can be determined by following relationship, PA F (2.9) Total F equals the sum of the friction force due to the particles and the wafer surface (F) and the friction force due to the pad and the wafer surface (Fp): p p p pA P A P F F F (2.10) Since the typical CMP pads are compliant, it may be assume that the real contact pressure (Pcon) remains unchanged (Pcon=P=Pp): ) ) 1 ( (p con con p p conf f P A A A P F (2.11) where Acon is the real contact area and f is the fractional surface coverage (A/Acon). Then an equation for fractional surface coverage for CMP with soft pads can be established: ) /( ) (con con p con con con con pA P A P A P F f (2.12) The experimental results showed that the fractional surface covera ge of particles is independent of down load and in creases with solids loading. The decrease of particle size leads to an increase in the tota l contact area of particles in co ntact with wafer, leading to a higher fraction of slurry particle s in contact with the wafer.

PAGE 33

20 (a) (b) Figure 2-8 Estimation of fractional surface coverage using in situ friction force measurements [Cho04] (a) in situ fricti on force as a function of solids loading for different down pressure and particle size (b) normalized fractional surface coverage of particles in contact with wafer surface, converted from (a). Formation of Chemically Modified Surface Layer It is well established that the slurry chemistry significantly influences copper surface properties such as dissolution and passi viation. These parameters are integral for controlling the properties and kinetics of chemically modi fied surface layer formation and therefore will ultimately influence polishing characteristics such as removal rate, planarity, surface quality, and defectivity. Po lishing slurries for copper CMP typically contain the following [Tam02]: 1. Oxidizer to form a surface oxide layer, 2. Corrosion inhibitor to preven t dissolution of copper, 3. Chelating agent to increase the copper removal rate. Earlier investigations on copper CMP larg ely utilized tools that were used for corrosion-related studies. Traditional methods of measuring corrosion effects such as potentiodynamic scans, and a.c. impedance provide important information on the

PAGE 34

21 dissolution/passivation behavior but, due to the dynamic nature of particle-surface interactions during CMP, are often inadequate for simulati ng or explaining CMP removal rates [Ste99, Kne97]. Instead, transient t echniques such as chronoamperometry that resolves millisecond range can provide better correlation with CMP removal rates [Sin02]. Lee extensively characterized ch emical interactions during copper CMP [Lee03]. The rapid formation of a thin passive laye r is one of the key aspects for achieving planarization [Sin02]. For this reason, an oxidizer is normally added to form a passive layer for metal CMP. However, hydrogen peroxide (H2O2), an oxidizer itself, is known to be inefficient in passivating copper surfaces. A corrosion inhibitor such as benzotriazole can be added to form a thin passive layer. By properly adjusting concentration of peroxide and benzotriazole, we can control the thickness of the surface passvation layer. Benzotriazole (BTA, C6H5N3), Citric acid (CA, HO2CCH2C(OH)(CO2H)CH2CO2H). Indentation Depth of Single Particle (W) A model describing the indentation de gree of particles onto the surface during polishing was first developed by Cook and Brow n et al. [Bro81] for the super polishing of metals. The particle indentation can be described by a model in which a spherical particle of diameter under uniform load P penetrates the surface with a force L and moves along the surface at some velocity re moving a volume of material of dimensions proportional to the penetration. They utilized a standard Hertzian in dentation to describe the penetration degree of the particle into the substrate. The indentat ion depth of a single particle was expressed as

PAGE 35

22 3 / 22 4 3 E K Pw (2.13) This equation assumes that the particles are much harder than the surface layer and interfacial contact betw een the particles and the surface is elastic. Zhao et al. and Qin et al. [Zha02, Qin04] de fined the indentation depth of a particle into the wafer by making use of the theory of contact mechanics in conjunction with the force equilibrium of a particle participati ng in the wear process. Different modes of deformation caused by a single particle into th e pad (elastic) and wafer (plastic) due to their very different mechanical properties. Based on these modes, th e wafer/particle/pad contact within the real area of contact between the pad and th e wafer can be visualized in Fig. 2-9, where the particle is largely embe dded into the pad surface. The elastic contact force between the pad and the particle (FSP) is given by 2 / 3 2 / 12 3 4P SP SPE F (2.14) where ESP is the reduced Youngs m odulus of contacting bodies, pad and particle, and P is indentation depth of a particle in polishing pad. The plastic contact force (FSW) between the wafer and the particle by W W SWH F (2.15) where HW is wafer hardness, and W is indentation depth of a particle in wafer. This force equilibrium can be used to cal culate the indentation depth (W) of the particle into the wafer surface with knowledge of wafer hardne ss, particle diameter and Youngs moduli of the particle and pad.

PAGE 36

23 (a) When t > (b) When t < w t a R b Pad Wafer Fa w t Pad Wafer a R b Fa(a) When t > (b) When t < w t a R b Pad Wafer Fa w t Pad Wafer a R b Fa w t Pad Wafer a R b Fa Figure 2-9 The wafer-particle-pad interacti on during polishing. This schematic shows a single particle that is trapped betwee n the pad and the wafer surface (a) when the indentation depth, w, of the particle is sm aller than the surface layer thickness, t, and (b) when the indenta tion depth is greater than the surface layer thickness [Qin04].

PAGE 37

24 CHAPTER 3 EXPERIMENTALS This chapter illustrates various experimental methods used in this investigation. Preparation processes for polishing experiments such as slurry characterization, polisher calibration, and various analysis methods for polishing experiments such as removal rate measurements, surface roughness m easurements, and in situ friction force measurements are described in detail. Brief descriptions of measurement principles for each equipment are presented. Sample Preparation Wafer Preparation 8-inch blanket PVD copper/Ta/TEOS silica wa fers were cleaved into 1.5 in. x 1.5 in. square specimens. To prevent edge chi pping during the polishing, the specimen edges were rounded with silicon carbide gri nding papers (Buehler, Grit 400/P800). The specimens were then cleaned with methanol and acetone and then dried with blown air. Particle Characterization and Slurry Synthesis The size, size distribution, and shape of the abrasive particle s were characterized using various techniques. The light scattering method was us ed to measure the size and distribution of the particles, and imagi ng techniques such as scanning electron microscopy (SEM) and transmission electron microscopy (TEM) were used to verify the size and shape of the particles.

PAGE 38

25Light Scattering A Microtrac UPA 150, the light scatteri ng particle size an alyzer, supplied by Honeywell was used to measure the particle size distribution. The light scattering is a relatively simple method that provides accurate information (0.003 m to 6.54 m) on the particle size and the size distribution. Light from laser diode is coupled to the sample, scattered from each particle, and shifted by pa rticle motion. The velo city distribution of particles suspended in a dispersing medium is a known function of particle size. The shifted scattered light is mixe d with coherent unshifted light in a silicon photodetector and mathematically analyzed with interfaced software. The particles were diluted in deionized water and sonicated for 10 minutes before the measurements. The measurements were repeated multiple times to obtain reproducible results. Imaging Techniques Particle size and shape were confirme d by microscopic techniques such as transmission electron microscopy (TEM) a nd scanning electron microscopy (SEM). The TEM was utilized for small-size particles (30 nm diameter) that are not easily observable with the SEM technique. Corresponding histog ram of particle distribution to the TEM images was also plotted by numbering and sizing particles in Photoshop. Slurry Preparation Any visible reactions while mixing different chemicals and particles were carefully monitored. Proper handling, such as suffici ent stirring rate and time when mixing the polishing additives, was done. First, citric ac id and benzotriazole (BTA) were mixed in deionized (DI) water. Abrasive particles were diluted for a desired solids loading (weight %) in separate DI water and were sonicated for approximate ly 10 min. to break down the

PAGE 39

26 aggregated particles to primary particles. The chemicals (citric acid and BTA) and abrasive particles were mixed into one c ontainer, and its pH was adjusted by KOH. Hydrogen peroxide was always added prior to each polishing in order to minimize the effect of time depende nt degradation. Table-Top Polishing Equipment Polishing experiments were performed on tabletop polishing equipment (Struers RotoPol-15 and RotoForce-1) using IC 1000/Suba IV stacked pads (Rodel). Such systems realistically represent scaled-down ve rsions of industrial scale system. Substrate Holder Figure 3-1 shows schematics of the tabletop polishing equipment and the substrate holder. The sample holder was specially machin ed for the RotoForce-1. As shown in Fig. 3-1b, the holder was designed to have three r ods (the fixation rods to the polisher head, RotoForce-1) for obtaining even pressure dist ribution. An adhesive carrier film (Rodel) was attached in the square-shape recessed area to hold the wafer against the pad.

PAGE 40

27 (a) (b) Figure 3-1 Schematics of (a) table-top polishi ng equipment and (b) the substrate holder, bottom view (upper image) and top view (lower image)*. Polishing Equipment Calibration An operational pressure range of a benchtop polishing apparatus was calibrated using a weight balance to provide accurate control of down pressure (Fig.3-2a), and the apparatus was benchmarked using commercially available CMP slurries and comparing our experimental results with results fr om open literature [Zho02] (Fig.3-2b). TEOS Silica wafers were used as standard samples. The bench-top polisher turned out to be reliable. Inch scale. The figure does not represent the real scale.

PAGE 41

28 (a) (b) Figure 3-2 The polishing equipment calibrati on (a) down force calibration using weight balance (b) benchmarking of the tabletop polisher by comparing removal rate values as a function of pressure*velocity (open circles represent in this study (open circle) with data from open literature (open triangle [Zho03]) and commercial slurry (open s quare, Klebosol, Rodel). Polishing Experiments The pad was conditioned prior to each polis hing experiment using a grid-abrade diamond pad (Rodel). The down pressure was varied from 1.5 psi to 9.0 psi. The speed of both platen and head were fixed at 150 RPM (0.88 m/s). The flow rate was fixed at 100 mL/min. Polishing was carried out for 30 s econds, and repeated three times. One-hour static etch rate was measured with the four-point probe me thod before and after dipping specimens in the aqueous media.

PAGE 42

29 Table 3-1 Comparison of process variables a nd slurry ingredients for calibration. Zhou et al. [Zho02] In this study Klebosol (Rodel ILD) Abrasive type Colloidal silica Colloidal silica Colloidal silica Abrasive size 80 nm 80 nm ~ 100 nm Solids loading 30 wt.% 30 wt.% 30 wt.% pH 11.0 10.5 10.8 Media DI water DI water ----Polishing apparatus RotoPol-35 RotoPol-15 RotoPol-15 Prestonian constant (Kp) 107.94 105.65 127.83 Table 3-2 Process parameters for calibration. Down Pressure 1.5 to 9.0 psi Head Speed 150 RPM Platen Speed 150 RPM Slurry Flow Rate 100 mL/min Polish Time 30 sec In Situ Friction Force Measurements The in-situ friction force measurement is a powerful technique for examining polishing mechanisms. Fig. 3-3 shows the schematic of the force measurement apparatus. The force transducer was connected to a data acquisition system and data was recorded every 250 ms for duration of 60 seconds. This ap paratus has been extensively used by our CMP group to investigate polishing-scale frict ion force behavior [Mah99a, Mah99b, and Cho04] especially in silica polishing. To expl ore various interactions in copper polishing, experiments were conducted as a function of particle size, concentration, and chemical additives with fixed values of the rest of the polishing parameters, such as pressure, velocity etc. The friction force values were correlated with the copper removal rate values.

PAGE 43

30 Figure 3-3 Schematic of apparatus for in situ friction force measurement [Mah99]. Film Thickness Measurements Removal rates were determined by measur ing the film thickness before and after polishing. The copper film thickness was measur ed using sheet resistance measurements and the thickness of insulating silica dielectrics used for calibrating the table-top polisher was measured using Various Angle Sp ectroscopic Ellipsometry (VASE). Four Point Probe Method The four point probe technique (Jandel) was used to measure sheet resistance and ultimately to determine the thickness of copper films. As shown in Fig. 3-4, a four point probe setup is mainly composed of current source, digital voltmeter, and four point probes. The four point probes consist of two curre nt carrying probes (1 and 4) and two voltage measuring probes (2 and 3). The meta l probes are lightly pressed onto the copper surface. A constant current ( I ) is passed through probes 1 and 4 and then the resulting voltage drop ( V ) between probes 2 and 3 is measured The sheet resistance of film ( Rs) is expressed as

PAGE 44

31 I V K Rs (3.1) where K is a correction factor dependent on the specimen diameter and probe spacing. If the probe spacing is larger than the film th ickness (t) and smaller than the distance to edge of the film, K=4.53. Figure 3-4 Schematic of four point probe measurement. The sheet resistance is film-dependent quantity which is rearranged from resistance. Let us consider resistance of a bar of materials with resistivity In this rectangular bar with thickness t, width W, and length L, the resistance is given by W L R W L t t W L A L Rs where Rs = /t is the sheet resistance of a layer of this material. Thus one can obtain the thickness of the copper film by measuring the sheet resist ance of the copper film.

PAGE 45

32Various Angle Spectroscopic Ellipsometry (VASE) Ellipsometry is a non-destructive technique for determining optical constants, film thicknesses in multilayered systems, surface and interfacial roughness, etc [Bru92]. In this study, the ellipsometry was used to measure the thickness of silica films that were used for calibrating the table-top polisher. Fi g. 3.5 shows the schema tic of spectroscopic ellipsometry (J.A. Woollam Co. Inc.). Figure 3-5 Schematic illustration of spectroscopic ellipsometry [Bru92]. Ellipsometry measures relative changes in the amplitude,, and the phase difference, of a linearly polarized monochromatic incident light upon an oblique reflection from a sample surface. In the ellipsometry, the complex ratio, is defined by the following equation, i R Rs pexp tan where Rp and Rs are the complex reflection coefficients of the polarized light parallel and perpendicular to the incident plane, respectiv ely. For a layered samp le, measured spectra may be analyzed using an appropriate fitting model constructed based on the sample

PAGE 46

33 structure. In the model, unknown parameters such as thickness, refractive index and absorption index of a layer may be then determined as fitting parameters. These parameters are then determined by minimi zing the mean-squared errors (MSE) between measured and calculated data. In order to obtain thickness information, the technique requires a model that contains wavelength of incident light, incident beam polarization state, angle of incidence, thickness of la yer, optical constant s (refractive index, absorption index, etc.). The thickness input va lues are varied until we meet perfect match with the model. Surface Topography Measurements: Atomic Force Microscopy (AFM) Atomic force microscopy (Digital Instrume nt Nanoscope III) was used to measure the surface roughness of polished copper specimens. The schematic diagram of the AFM apparatus is shown in Fig. 3-4. The AFM utilizes a sharp probe (a tip mounted on a flexible cantilever) moving along the specimen su rface in a raster scan. Repulsive van der Waals forces between atoms on the tip and s ubstrate cause deflection of the cantilever. The force between the tip and sample are a f unction of the separation distance. Therefore, surface topography is obtained by monitoring th e cantilever deflection. The light from the laser diode is reflected u pon the cantilever and detected on the quadruple photodiode. By measuring the light position on the photodiode changes in the bendi ng of the cantilever can be measured. The surface roughness is quantitatively expressed as n i iZ Z n RMS1 2) ( 1 (3.1) where Zi is the height values of single data points in the image, Z is the mean value of all height values in the image, and n, the number of data points within the image.

PAGE 47

34 Figure 3-6 Schematic diagram of atom ic force microscopy (AFM) [Bru92]. The measurements were conducted in tapping mode. Prior to all surface roughness measurements, excess particles adhered on th e surfaces after polishing were removed by sonicating for a minute in isopropanol.

PAGE 48

35 CHAPTER 4 SYNERGISTIC CHEMICAL-MECHANICAL EFFECTS As addressed in chapter 1 and 2, due to their fragility and poor adhesion next generation interconnect materials such as coppe r and low k dielectrics are susceptible to CMP-induced contact stresses (normal and shea r) which may lead to defect formation such as microscratches, copper and barrier la yer delamination, low k damage, dishing and erosion [Sin03, Wan01]. To reduce the defect formation, investiga tions such as reduction in down pressure [Kon03], abrasive-free polishing [Pad03], el ectro-polishing [Cha03], etc. have been conducted. Material removal by these methods however, is often dependent on chemical dissolution for high removal rate, which is known to be ineffective for obtaining high planarity due to its isotropic nature. Our methodology in this study to address the defect reduction is based on contact mechanics. In this mode of polishing regime, it is believed that active abrasive particles are trapped between pad asperity and wafer su rface, and therefore participate in material removal. As shown in Fig. 2, the particle indentation during po lishing is a nano-scale phenomenon and its depth into the wafer su rface can be theoretically estimated using force equilibrium between particle-pad in terface and particle-waf er interface [Zha02, Qin04]. This penetration depth depends on (i) particle properties (size, hardness, shape etc.), (ii) surface layer properties (thickness, hardness, density, etc.), and (iii) pad modulus. It should be noted that the penetrati on depth could be either larger or smaller than the surface layer thickness. Due to the low hardness of bare copper underneath the

PAGE 49

36 surface layer, a penetration depth greater than the surface layer thickness can cause significant micro-scratches and other defect. Therefore, the defect formation can be minimized by decreasing the penetration dept h of particles. Acco rding to the concept above, however, the penetration depth is not de termined solely by the particle properties, but is determined by their synergistic interaction with surface layer that can be influenced by various factors. A study of the particle effects on polishing characteristics with systematic variation in slurry chemistry is critical, but has yet to be investigated. Our objective in this chapter is to exam ine the synergistic chemical-mechanical effects on defect formation during copper CMP. To explore the effect s of particle size and hardness, various types of particles such as 100 nm silica particle, 1 micron silica particle, and 200 nm alumina particle were used. Slurry chemistries such as hydrogen peroxide as an oxidizer, benz otriazole as a corrosion inhibitor, and citric acid as a chelating agent were also systematically vari ed to alter the copper surface properties (pH: 7.0). Removal rate and surface roughness were measured as function of these slurry parameters. The surface roughness and micro-scratc hes were used as indicators of defect formation. Based on measurement results, th e proper method to obtain gentle polishing behavior of copper is suggested. Particle Characterization To study the effect of particle size and ha rdness, three different types of particles (nano-size silica, micron-size silica, 200 nm alum ina) are used in this chapter. The details of the particles used in this ch apter are given in Table 4-1.

PAGE 50

37 Table 4-1 Details of differe nt types of particles Materials Approx. mean diameter (m) Manufacturer Type SiO2 0.1 H.C. Starck Levasil SiO2 1.0 Geltech Geltech 1.0 m -Al2O3 0.2 Sumitomo AKP 50 Figure 4-1 Particle analysis (a) particle size distribution obtained by light scattering method (b) image of 100 nm silica particle obtained by SEM (c) image of 200 nm alumina particle obtain ed by TEM (d) image of 1 m silica particle obtained by SEM.

PAGE 51

38 Prior to polishing experiments, the partic les used were characterized using several techniques. The light scatteri ng method was used to determine the particle size and the particle size distribution. Th e imaging techniques such as SEM and TEM were used to confirm the particle size and shape. The analysis results are shown in Fig. 4-1. Fig. 4-1a shows the particle size distribut ion of three different partic les obtained by light scattering method. Figures 4-1b to 4-1d show the im ages of the particles obtained by SEM and TEM. The silica particles (100 nm and 1 m in mean diameter) are quite spherical whereas the alumina particles are somewh at irregular and faceted in shape. Colloidal Silica Nanoparticles Figure 4-2 shows the removal rate of copper using neutral slurries (pH 7.0) with 10 wt. % solids loading of colloidal silica particles. As shown in Fig. 4-2a, the removal rate obtained by the nanoparticles dispersed in de -ionized water is approximately 15 nm/min, which is quite negligible. The addition of hydrogen peroxide does not significantly influence the removal rate. Figure 4-2b shows the effect of benzotri azole addition in 5 % hydrogen peroxide on removal rate, which suggest s that the removal rate is still quite low. Figure 4-3 shows the removal rate measured as a function of citric acid concentration in the same slurry used for Fig. 4-2b. In the presence of 5 % hydrogen peroxide and 10 mM benzotriazole, the increase in citric acid co ncentration in the slurry exhibits a linear increase in removal rate. The figure also shows that the removal rate is quite low when only citric acid is added in de-ionized water without addi tion of hydrogen peroxide and benzotriazole.

PAGE 52

39 (a) (b) Figure 4-2 Removal rate of copper using 10 wt. % solids loading of colloidal silica particles at pH 7.0 (a) effect of hydrogen peroxide concentration in de-ionized water (b) effect of benzotriazole con centration in 5 % hydrogen peroxide. Figure 4-3 Removal rate of copper versus citric acid concentration, pH: 7.0. 100 nm silica particles are used. Open ci rcle indicates the removal rates measured with 5 % H2O2 and 10 mM benzotriazole. Open tr iangle indicates the removal rate measured in water.

PAGE 53

40 Figure 4-4 shows the root mean square (RMS) values of copper surface after polishing with 10 mM and 200 mM of citric acid concentration in the slurry used for Fig. 4-3. The result shows that th e surface roughness is fairly i ndependent of citric acid concentration Figure 4-4 Removal rate of copper as a func tion of down pressure with and without 100 mM citric acid. 5 % hydrogen peroxide, 10 mM benzotriazole 100 nm silica at 10 wt. %, pH: 7.0. Figure 4-4 shows the removal rate of copper as a function of down pressure using 100 nm silica particles at 10 wt. % along with 5 % H2O2, 10 mM benzotriazole (BTA) and 100 mM citric acid at pH 7. Without citric acid, the removal rate is low, below 30 nm/min, over the entire down pressure range. Ho wever, as 100 mM of citric acid is added to the slurry, the removal rate increases considerably. The linear increase in removal rate

PAGE 54

41 with increasing down pressure is attributed to an increa se in the number of active particles contacting the wafer su rface due to the compliant nature of the polymeric pad. Based on these results, it is suggested that the addition of citr ic acid may enhance delamination between the surface layer and the bulk copper, and may cause brittle fracture by the abrasives maki ng it possible to form a layer removable by the nano-size silica abrasives. As shown in Fig. 4-5, the surface topography after polishing with nanosize silica abrasives is almost scratch-free, indicating that material removal takes place only within the surface layer. Slight increase in root mean square (RMS) value is shown as down pressure increases. The linear increas e in removal rate and slight increase in RMS as a function of the down pressure are at tributable to an increase in the active number of abrasives in contact with the wafe r surface, due to the compliant nature of the polymer pad [4]. (a) (b) (c) (d) Figure 4-5 Down pressure effect on surface image and roughness after polishing with 100 nm silica particles at 10 wt. %: (a) 1.5 psi (RMS: 0.350 nm), (b) 4.5 psi (RMS: 0.427 nm), (c) 7.5 psi (RMS: 0.711 nm), (d) 9.0 psi (RMS: 1.189 nm).

PAGE 55

42 The rapid formation of a thin passive laye r is one of the key aspects for achieving excellent planarity by the CMP process [Sin02]. For this reason, an oxidizer is normally added in copper polishing slurries. The oxidize r such as hydrogen peroxide is responsible for rapidly forming the passive layer. Howeve r, hydrogen peroxide itself is known to be inefficient to passivate the copper surfaces. A corrosion inhibitor such as benzotriazole can be added to form a thin passive layer. As shown in chapter 2, the thickness of the surface passive layer can be controlled by prope rly adjusting concentration of peroxide and benzotriazole. However, using nano-size si lica particles, the removal rate for this slurry condition is very low. This imp lies that the surface layer formed by hydrogen peroxide and benzotriazole is mechanically too rigid to be rem oved by nano-size silica particles. On the other hand, th e synergistic effect is found with the addition of citric acid in the slurry and serves to increase the removal rate considerably. Micron-size Silica and 200 nm Alumina Particles Unlike in nano-silica based polishing, mi croscratches are often observed after polishing of copper with hard or large abrasi ve-based conventional slurries most likely due to the penetration deeper than the su rface layer thickness. This removal behavior resembles conventional mechanical polishing wh ere the removal of material results from direct interaction between ab rasives and the bare copper, i.e. the indentation of the surface and the removal of material by a sc ratching process is prevalent [Sin02b]. As shown in Fig. 4-6, polishing with 200 nm alumina abrasive generates a much higher removal rate than that of 100 nm silic a abrasive, but shows very aggressive microscratches (Fig. 5) with high RMS value.

PAGE 56

43 Figure 4-6 Removal rate and surface roughness after polishing with various particles (slurry condition: 5 % H2O2, 10 mM BTA, 100 mM citr ic acid at pH 7.0) The results indicate that the hard abrasi ves not only penetrate through the surface layer, but also remove bare copper effectively with the sc ratching process that requires higher energy to break copper bonds. On the other hand, polishing with 1 m silica abrasives does not exhibit much higher removal rate than that of 100 nm silica abrasives. The RMS value of the copper surface afte r polishing with the micron-size silica is slightly lower than that obtained from 200 nm alumina abrasives, but much higher than that obtained from 100 nm silica abrasives. Th is suggests that the la rge silica abrasive can penetrate through the surface laye r but cannot effectively remove bare copper underneath the layer due to its low hardness. As such, the removal mechanism can be quite different when indentation depth by the abrasive is larger or smaller than the surface layer thickness. In order to achieve low stress polishing of copper, it is preferable to cause

PAGE 57

44 brittle fracture only within the surface laye r because it requires low energy to generate material removal due to natu re of brittle fracture. Conclusion We have investigated copper CMP using na noparticle-based slurries. With addition of citric acid in the presence of hydrogen pe roxide and benzotriazole at pH 7, we obtained moderate removal rates with very low rms values, less than 0.7 nm. Micron-size silica and 200 nm alumina particles were also used as comparison. 1. The hardness and size of the particles can be reduced to lower the penetration depth, thus avoiding damaging effects. Howeve r, reducing particle hardness and size without use of synergistic chemistries ma y lead to much reduced removal rates. 2. The addition of complexing agent, citric acid, in the slurry is critical to the formation of surface layer removable by the colloidal silic a nanoparticles. 3. In contrast, larger or harder particles resu lt in greater surface defects suggesting that removal takes place via scratching process. We propose that citric acid softens the surface layer formed by hydrogen peroxide and benzotriazole, and results in larger indentation depth of nano-size silica particles, thus increasing the removal rate. Combined mechanical-chemical effects are needed to achieve low-defectivity polishing of copper.

PAGE 58

45 Figure 4-7 Surface roughness copper specimen (a) as received (RMS: 6.963 nm) copper wafer (b) polished with 100 nm silic a at 10 wt. % (RMS: 0.711 nm) (c) polished with 1 m silica at 10 wt. % (RMS: 6. 985 nm) (d) polished with 200 nm alumina at 5 wt. % (RMS: 8.793 nm)

PAGE 59

46 CHAPTER 5 FORMATION OF REMOVABLE SURFACE LAYER In chemical mechanical polishing (CMP) of copper, chelating agents are often added in the hydrogen peroxide, the most commonly used oxid izer, -based slurries to increase the removal rate [Sin02b]. It has b een reported that simu ltaneous use of some chelating agents and the peroxi de synergistically increases the removal rate of copper. These results are normally inferred to the abil ity of the chelating agents to form copper complexes and its enhanced solubility [Aks02, Aks03, Du04, Che04]. Citric acid is of particular interest in this work because, although it is often utilized in copper CMP due to its chelating ability, it is relatively le ss explored among other chelating agents such as glycine. Tamilman i et al. [Tam02] constructed potential-pH diagrams based on thermodynamic calculations a nd suggested that presence of the citrate ions in the slurry would tend to chemically dissolve copper through complex formations. Chen et al. [Che04] studied the effect of citric acid on coppe r polishing using hydrogen peroxide-based acidic slurries that contain 50 nm alumina particles. The removal rate in their study was exceedingly high, which was up to 5.5 m at 3 psi. It is conceivable that the enhanced material removal in acidic re gime without an inhi bitor may include an unwanted level of chemical dissolution, which would not help in achieving efficient planarization due to its isotropic nature and al so raises a corrosion i ssue. In addition, we demonstrated in chapter four that use of such hard particles may result in significant micro-scratches due to their se vere penetration and direct interaction with soft bare

PAGE 60

47 copper underneath the surface layer. Therefore, further aspect s of citric acid addition under more passivating and mechanically gentle ambient should be addressed. Our objective in this chapter is to unde rstand the role of ci tric acid during the copper CMP using silica nanoparticle based neutral slurries. To study its effect on polishing behavior, removal rate was measured as a function of citric acid in the presence of hydrogen peroxide and benzot riazole at pH 7. Static et ch rate measurements and potentiodynamic scans were conducted to st udy dissolution / passiva tion behavior. To study the effect of citric acid on chemical composition of the surface layer, x-ray photoelectron spectrosco py (XPS) was used. Reaction kinetics of the layer formation was studied using chronoamperometry technique. Ph ysical properties of the surface layer such as hardness and density we re obtained using nanoindent ation and x-ray reflectivity (XRR), respectively. Various aspects of citr ic acid addition on copper surface and its influence on polishing performance are discussed. Experimentals Removal Rate and Static Etch Rate Measurements Eight-inch blanket PVD copper/Ta/TEOS silica wa fers were cleaved into 1.5 in. x 1.5 in. samples. Polishing experiments were conducted on tabletop polishing equipment (Struers RotoPol-15 and RotoForce-1) using IC 1000/Suba IV stacked pads (Rodel). The pad was conditioned prior to each polishing experiment using a grid-abrade diamond pad (Rodel). Down pressure was fi xed at 7.5 psi, with 150 RP M of both platen and head speed (0.88 m/s). Flow rate was fixed at 100 mL/min. The sample holder was designed and machined for the equipment. An adhesive carrier film (Rodel) was attached in the square-shape recessed area to hold the wafe rs against the pad. Removal rates were determined by measuring the thickness of th e copper layer using the four-point probe

PAGE 61

48 method before and after polishing. Polishing wa s carried out for 30 seconds, and repeated three times. Chemicals (hydrogen peroxide, ci tric acid and benzotriazole) and abrasive particles were mixed in deionized water, and further adjusted to pH 7.0 by KOH. Onehour static etch rate was measured with th e four-point probe me thod before and after dipping specimens in the aqueous media. Electrochemical Tests The three electrode cell was used. Hi gh purity copper coupons (99.9985%), the working electrode, were pre-polished and mount ed in a Teflon holder that exposed 1 cm2 surface area. A saturated calomel electrode (S CE) and two graphite rods was used as a reference electrode and counter electrodes, respectively. The Autolab PGSTAT 30 was used for the potentiodynamic scans. The scan s were performed at the same rate of 1 mV/s, beginning at -0.75 V a nd ending at 0.5 V (respect to SCE). To examine the transient electrochemical beha vior, the FRA2 module was adde d to the potentiostat. This technique, so-called Chronoamperometry, measur es the decrease in curr ent with time at a given electrical potential, which provides valuable time-resolved information. Prior to the measurement, the copper specimens were kept at a cathodic potential in the test solution to prevent oxidation reaction on the surface. Then an open-circuit potential, measured from potentiodynamic scans, was applied. The current drop was monitored in the millisecond range and converted to the surf ace layer thickness using Faradays law of electrolysis. X-ray Photoelectron Spectroscopy (XPS) XPS experiments were performed to determine the chemical composition of the chemically modified surface layer. A specime n was dipped in the aqueous solution for a minute and moved over to the ultra high vac uum (UHV) chamber that was less than 10-8

PAGE 62

49 Torr. Photoelectrons were excite d by using non-monochromatized Mg K x-ray irradiation (1253.6 eV), and the detection angle was normal to the sample surface. Nanoindentation and X-ray Reflectivity Nanoindentation experiments were co nducted using a cube corner diamond indenter with Triboindent er and Triboscope manufactured by Hysitron. Copper specimens were dipped in the aqueous solutions for a minute and moved to the sample stage. For each specimen, the experiments we re conducted at 64 separate maximum loads starting from 100 N to 5 N with successive 4 % decreas e. All indentations were performed using a load-time sequence. To ch aracterize the density of the surface layer modified by the chemical additives, x-ra y reflectivity (XRR) was used. XRR is a nondestructive and non-contact technique for de nsity measurements. A Philips XPert x-ray diffraction system was used for this study. For x-ray radiation of incident angle I < critical angle (c) total external reflection occurs at a certain reflection angle (r). The critical angle is proportional to the half power of the layer dens ity. For conditions in which the incident and reflected angle are equal, the density can be obtained for the chemically modified surface layer. Synergistic Chemical Effect Figure 5-1 shows the removal rate of copper vs. citric acid concen tration (0-200 mM) using 10 wt. % of 100 nm sili ca particles along with 5 % H2O2 and 10 mM benzotriazole (BTA) at pH 7. The peroxide and BTA are a dded in the slurry for the purpose of rapid formation of a passive layer which is necessa ry for achieving the effective planarization [1]. As shown in Fig. 5-1, the removal rate of these two chemical agen ts without citric acid is negligible but linearly increases with an increase of citric acid concentration in the

PAGE 63

50 slurry. The figure shows that the removal rate is also influenced by the down pressure. To identify the role of citric acid, as the third chemical ingredient that increases the removal rate, several experiments were further conducted. Figure 5-1 Removal rate of copper vs. citric acid concentration in the presence of 5 % hydrogen peroxide and 10 mM BTA at pH 7. Dissolution / Passivation First, static etch rate was measured as a function of citric acid concentration in several aqueous media at pH 7. The rate va lues were taken from the thickness loss of copper after dipping in the test solutions. As shown in Fig. 5-2, the etch rate slightly increases with increasing citric acid content in deionized water. The etching effect is enhanced with the addition of 5 % hydrogen peroxide but shows saturation with an excess amount of citric acid (above 100 mM). It is also shown that the chemical dissolution effect enhanced by the mixture of hydrogen peroxide and citric acid can be completely suppressed by addition of 10 mM BTA due to its inhibitive action.

PAGE 64

51 Figure 5-2 Effect of citric acid concentra tion on copper etch rate in various aqueous media; deionized water (open square), 5 % H2O2 (open circle), 5 % H2O2 and 10 mM benzotriazole (ope n triangle) at pH 7. Potentiodynamic scans of copper were also conducted in several aqueous media. The plots of SCE vs. logarithmic current dens ity are shown in Fig. 5-3. Much higher opencircuit potentia l of copper is observed in the 5 % hydr ogen peroxide solution than that in the 200 mM citric acid solution. Mixed soluti on of these two agents exhibits both higher open-circuit potential and highe r corrosion current density th an those in individual ones. Similarly to the static etch rate result, the a ddition of BTA is also shown to be effective in surface passivation. The values of open-circui t potential and corrosion current density are summarized in Table 1. Dissolution rate of copper was calculated from the current density and compared with the values from thickness loss method. The dissolution rate values taken from the potentiodynamic scans we re higher than that from the thickness loss measurements. The limited resolution of thickness measurement by four point probe

PAGE 65

52 method and different time scale of measur ements may account for the disagreement. Nevertheless, both results show similar tre nd in dissolution/passivation behavior with chemistry variations. Figure 5-3. Effect of citric acid on polarizati on curves in various aqueous media at pH 7. Surface Chemical Composition X-ray photoelectron spectros copy (XPS) was used to examine how citric acid influences the chemical composition of the copper surface modified by hydrogen peroxide. Fig. 4a shows the XPS spectra of a copper specimen after dipping in 5% hydrogen peroxide and 100mM citric acid. The Cu2P3/2 line and its satellite are observed at 933.6 eV and 943.3 eV with very weak intensity, indicating relatively thin CuO formation on the surface. Cu2O and Cu(OH)2 are also revealed from O1s binding energy of 530.4 eV and 531.3 eV, respectively. Moreover, C1s of carboxylic acid group (COOH-) is found at 288 eV, indica ting that citric acid comple x exists on the surface. To further study the inner layer composition, th e specimen was sputtered with Ar+ ions for 12 and 48 seconds.

PAGE 66

53 Figure 5-4 (a) The XPS spectra of copper surface trea ted with 5 % H2O2, 10 mM benzotriazole, and 100 mM citric acid at pH 7 with different sputtering time, (b) O1s line, (c) C1s line. After sputtering for 12 seconds, the carboxylic acid group starts to vanish as shown in Fig 4c. After 48 second-sputtering, the ca rboxylic acid group is mostly disappeared while Cu2O is still observed with the same intensity (Fig. 5-4b), indicating that the oxide is the main component of the surface layer. Reaction Kinetics An important aspect relevant to the layer fo rmation is the rate at which the particle abrasion takes place, which is generally cons idered in the range of 10 to 400 milliseconds depending upon the slurry variables [Sin02b]. To study the effect of citric acid on the

PAGE 67

54 dynamic nature of layer growth in the shor t-time scale, chronoamperometry technique was used. Figure 5-5a shows the transient behavi or of surface layer formation on copper in the millisecond range (< 0.07 sec). When only citric acid is used, rate of the layer growth is relatively low. When hydrogen pe roxide is added to the citr ic acid solution, the growth becomes much faster, and the dependence of la yer thickness on time is almost linear. The addition of BTA in the mixed solution decrease s the growth rate, bu t the rate is still higher than that in the citric acid solution. The layer growth was also monitored as a function of citric acid concentration in the presence of the peroxide and BTA, which is shown in Fig. 5b. The figure shows that the in crease of citric acid concentration enhances the reaction kinetics. Physical Properties of Surface Modified Surface The hardness and density of the layer were measured using nanoindentation and XRR, respectively, as a function of citric acid concentration in neutral solutions that contain 5 % hydrogen peroxide a nd 10 mM BTA. Figure 5-6 shows that the both the hardness and density of the surface laye r is significantly decreased by the addition of citric acid in the slurry; approximately 40 % decrease in hardness and 21 % decrease in density with 100 mM of citric acid content.

PAGE 68

55 (a) (b) Figure 5-5 Transient response of layer grow th on copper in the millisecond regime (a) effect of citric acid in various media (b) effect of citric acid concentration in the aqueous media that contains 5 % H2O2 and 10 mM BTA.

PAGE 69

56 Figure 5-6 Effect of citric acid addition on the hardness and the density of the surface layer; 5 % H2O2 and 10 mM benzotriazole at pH 7. Discussion Citric acid is a chelating agent with tetra-dentate car boxylic groups that can lose protons and become a citrate ion. Tamilman i et al. [Tam02] suggested that, based on thermodynamic calculations (pH-potential diagrams ), presence of the citrate ions in the slurry would tend to chemically dissolve copper through complex formations. According to the pH-potential diagram, doubl y charged anionic complex (Cu(H-1L)2-, where L = C6H5O7 3-) is predominant at the neutral pH. Despit e the rational identif ication of stable species under given conditions, only the th ermodynamically favorable and predominant species appear on the diagram. Our result of X PS with sputtering, in fact, shows that top surface of the copper specimen treated with the peroxide and citric ac id at neutral pH is mainly composed of copper oxide, not the copp er-citric acid complex. This implies that the enhanced dissolution effect from the mixtur e of the peroxide and citric acid in Fig. 5-2

PAGE 70

57 and Fig. 5-3 is attributed to oxide etching, which may take place by two main reactions; oxidation of copper (1), responsible for th e oxide formation, followed by complexation (2), responsible for the oxide dissolution. Cu = Cu2+ + 2e(1) Cu2+ + citrate = Cu-citrate complex (2) It is believed that above 100 mM of citric acid the total di ssolution rate is limited by the oxidation rate, thus showing the saturation (open circle) in Fig. 5-2. The lower etch rate in deionized water is most likely due to relatively thin-native oxide formation by less oxidizing power of water than that of hydroge n peroxide, which is demonstrated by the potentiodynamic scans (Fig. 5-3). It is noted here that the etch ing effect by the peroxide and citric acid is efficiently suppressed by the a ddition of BTA over the en tire range of citric acid concentration. Regardless of the minimal etch rate with the BTA addition, however, the removal rate is linearly pr oportional to the citric acid conc entration, as shown in Fig. 5-1. Such a discrepancy clearly indicates th at the polishing rate of copper under this condition is fully attributed to chemo-mechan ical effects, and the polishing mechanism can be better correlated with the transient electrochemical study. Figure 5-5 shows that the reaction kinetics in the millisecond regime is indeed influenced by the slurry chemistries. A notable feature in this transient study is that, as shown in Fig. 5-5b, the addition of citric ac id in the slurry that contains 5 % hydrogen peroxide and 10 mM BTA enhances the layer gr owth rate, which is not indicated by the static dissolution/passivation study. In addition, the adsorption of BTA is seen to play a role not only in that it prohibits copper from direct dissolution but also in that the layer growth shows parabolic dependence on time. One may reason that the enhanced reaction

PAGE 71

58 kinetics leads to less dense oxi de layer, thus decreasing the layer hardness. This is also confirmed by the results of XRR and nanoindentation. The ad dition of citric acid does degrade the density and the hardness of the surface layer. Based on these observations, we believe that, using nanoparticle-based sl urries, the synergistic mechanical-chemical effect by the addition of citric acid that serves to increase the removal rate is solely attributable to the soft layer formation th at allows considerable indentation of nano particles, rather than enhanc ing solubility of copper. The absence of chelating action of citric acid that leads to enhanced copper dissolution is ascribed to (1) effective passivation by use of neutral slurry and by addition of BT A and (2) relatively small amount of debris generated by nano particles. Figure 5-7 Potential-pH diagram for coppercitric acid-water system. Redox potential values of 0.1 M citric acid solution (square) and 0.1 M citric acid with 4 % hydrogen peroxide (circle) are also displayed. (Cu(H-1L)2-, where L = C6H5O7 3-) [Tam03] Conclusion We investigated the role of citric acid during copper CMP using silica nanoparticlebased slurries. Linear increase in removal ra te was found with an increase of citric acid

PAGE 72

59 content in the presence of hydrogen peroxide and benzotriazole at neutral pH. Static dissolution studies and x-ray photoelectron spectroscopy sugges ts that coppe r dissolution is enhanced by the mixture of hydrogen per oxide and citric acid mostly via oxide dissolution. However, the static etch rate of copper is essentially minimal with addition of BTA but is not well correlated with the removal rate result. In contrast, the transient electrochemical study suggests that reacti on kinetics in the m illisecond regime is enhanced by citric acid add ition even in the presence of BTA, which we believe leads to less dense oxide formation and therefore allows considerable indent ation of silica nano particles. This is also confirmed by our results of nanoindentation and XRR, which exhibit that the addition of citric acid in the presence of hydrogen peroxide and benzotriazole significantly decreases the ha rdness and density of the surface layer. Therefore, using silica nanoparticle-based neut ral slurries, the linea r increase in removal rate as a function of citric acid concentra tion can be explained by formation of weakly bonded surface layer, rather than high solubility of copper.

PAGE 73

60 CHAPTER 6 ROLE OF NANOPARTICLE SIZE AND CONCENTRATION The particle size and concentration play important roles in determining polishing characteristics such as remova l rate, surface roughness, etc., a nd also have been used to help indirectly understand in what manner the material removal occurs during polishing [Sin02]. Nevertheless, experiment al results that have been re ported with respect to the particle size for polishing of different materi als are quite contradicting [Bie99, Mah00]. It has been reported that such contradicting results are also found when the same materials are polished in different particle size ranges such as nano-size part icles vs. micron-size particles. Bielmann et al. [Bie99] studied tungsten CMP using alumina particles ranging 0.313 m and showed that removal rate was invers ely dependent of particle size. On the other hand, Bouvet et al. [Bou02 ] reported that, using sub 100 nm silica particles (12-75 nm), tungsten removal rate was fairly constant with particle size. The conflicting results from different particle size regimes are also seen in oxide CMP. Ma hajan et al. [Mah00] used 0.2-1.5 m silica particles and showed that op timum concentration for high removal rate existed for each particle size, which shif ted to a lower concentration with increasing particle size. Such an optim um concentration was not obser ved when Zhou et al. [Zho02] investigated silica CMP using relatively sm all silica particles (10-140 nm). Their result suggested that the variation of removal rate with nano-sc ale particle size was non-linear and the highest removal rate was obtained using 80 nm particle based slurry.

PAGE 74

61 In copper CMP, to our best knowledge, onl y Lu et al. [Lu03] studied effect of particle size, using si lica particles ranging from 50 to 300 nm in hydrogen peroxide and glycine-based acidic slurries. They showed that removal rate increased with a decrease in particle size and suggested that material re moval would be controlled by specific surface area of particles. However, as observed in the tungsten and silica CMPs above, use of particles in different size regimes may l ead to different polishing phenomena. The study of particle size in copper CMP discretely for the nano-size regime has not yet been conducted. In this chapter, we investigate the eff ects of nano-particle size (30-100 nm) and concentration (3-10 wt. %) in copper CMP. The particles are charac terized using several techniques. The removal rate and CMP i nduced-surface roughness are measured with fixed concentration of slurry chemical agents (5 % H2O2, 10 mM benzotriazole (BTA) and 100 mM citric acid) at pH 7. In order to understand the wafer-slu rry-pad interactions, in situ friction force is measured. Based on these measurements, the role of nano-size silica particles for copper polishing is examined. Results and Discussion Particles are characterized using severa l techniques. Fig. 6-1a shows the size distribution of three different particles (30, 80, 100 nm in approximate mean diameter) obtained by light scattering method, which indicates that each particle size is narrowly distributed. In Fig. 6-1b to c, the images, of particles measured by TEM and SEM are shown, which confirm the particle size and sugge st that the particles are quite spherical in shape.

PAGE 75

62 (a) (b) (c) Figure 6-1 Particle analysis (a) particle si ze distribution of colloidal silica obtained by light scattering method (b) image of 30 nm colloidal silica obtained by TEM (c) image of 100 nm colloidal silica obtained by SEM. To study particle size and concentration e ffects, we selected a specific slurry chemical composition (5 % H2O2, 10 mM BTA and 100 mM citric acid at pH 7). Based on our results in chap ter four, such a slurry compos ition exhibits a reasonably high

PAGE 76

63 removal rate without involving chemical disso lution. Figure 6-2 shows the removal rate of copper as function of nanopart icle size (30-100 nm) and con centration (3-10 wt. %). The figure shows the removal rate increases w ith the increase of both particle size and concentration. Figure 6-2 Effect of particle size and concen tration on removal rate of copper with fixed chemical agents (5 % H2O2, 10 mM BTA and 100 mM citric acid) at pH 7 (open circle; 100 nm, open square; 80 nm, open triangle; 30 nm) The surface topographic image and roughness of the polished samples measured by AFM are shown in Fig. 6-3. As shown in Fi g. 6-3a, micro-scratches that are commonly observed after copper CMP process are not exhi bited. Figure 6-3b shows that root mean square (RMS) values are very low (> 0.7 nm) and fairly independent of particle size and concentration.

PAGE 77

64 (a) (b) (c) Figure 6-3 Effect of particle concentra tion on surface topography after polishing with 100 nm colloidal silica particle (a) 3 wt. % (b) 5 wt. % (c) 10 wt. % To understand the role of particle size in copper CMP, in situ friction force measurements were conducted on these samp les. Figure 6-6 shows the friction force response vs. particle concen tration and particle size ( 30 and 100 nm). Two sets of experiments were conducted: one without citric acid in the sl urry (5 % hydrogen peroxide and 10 mM BTA at pH 7, Fig. 6-6a), the other with 100 mM citric acid in the same slurry (Fig. 6-6b). In both cases, friction force in creases with increasing particle size and concentration. However, friction force response becomes much more dependent on particle size and increases more rapidly with increasing particle c oncentration as 100 mM citric acid is added in the polishing slurry (Fig. 6-6b).

PAGE 78

65 (a) (b) (c) Figure 6-4 Effect of particle size on su rface topography after polishing with 10 wt. % solids loading (a) 30 nm (b) 80 nm (c) 100 nm. Figure 6-5 Effect of particle size and concentration on surf ace roughness (RMS) of copper specimen after polishing.

PAGE 79

66 Figure 6-6 Effect of particle size and c oncentration on friction force during copper polishing (a) without citric acid in the slurry (5% H2O2, 10mM BTA at pH 7) (b) with addition of 100mM citric acid. In contrast to earlier results by Lu et al. [Lu03], the removal rate obtained in this study increases with increasing pa rticle size. It should be not ed that, in copper CMP, the nano-particle polishing results are quite different compared to sub-micron size particle polishing in which we observed much high rms values and deeper scratches in chapter four. As shown in Fig. 6-3 and Fig. 6-4, the copper surface polished with nanoparticles does not involve the microscratches and the RMS value is independe nt of the particle size, which suggest that, due to nanoparticle characteristics of shallow indentation, the material removal occurs wit hout significantly damaging the sub bare copper underneath the chemically modified surface layer. The slurry chemistry also significantly in fluences copper surface properties such as dissolution/passivation behavior or surface laye r formation mechanism, and consequently influences the polishing characteristics. It is conceivable that the material removal in the work by Lu et al. includes high level of ch emical dissolution. The removal rate obtained

PAGE 80

67 without use of abrasives was exceedingly hi gh (> 200 nm/min), which may not be the appropriate slurry environment to study the pure effect of particle size. On the other hand, we earlier showed that the slurry chemical condition selected for this study did not involve chemical dissolution, but rather form a passive surface layer that can be indented and removed by the nanoparticles. Detailed discussion of the removable layer formation is presented in chapter 5. The friction force result shown in Fig. 66 implies that the particle indentation volume plays an important role in material removal. Choi et al. [Cho04] utilized the in situ friction force approach to estimate the fr actional surface coverage of particle at the pad-wafer contact area. To minimize the eff ect of particle indent ation on the friction force measurement, the experiments were conducted using soft silica particles (*540 kg/mm2) and hard sapphire wafer (*2370 kg/mm2). Their experimental result showed that the friction force increased as the particle si ze decreased and as the particle concentration increased. The larger friction force responses for the smaller particles were explained by their larger surface area (or larger number of contact). Note that the number of smaller particles is greater than that of bigger particle s at the same particle concentration (wt. %). Our measurement result obtained in this study is similar to Chois result in that the friction force increases as particle concentr ation increases, but shows disagreement in the particle size effect on the fric tion force responses. Such conflic ting results in the particle size effect on friction force are attributable to the particle indentation on the wafer surface. A proposed schematic of nanoscale particle-copper surface interactions is illustrated in Fig. 6-7. It is believed that the particle indentation is limited within the Vickers hardness [Ric71]

PAGE 81

68 surface layer because of the characteristics of nanoparticles, and the larger particles (100 nm) with larger indentation volume may lead to higher friction force and more efficient material removal. Figure 6-7 Schematic illustration of nanoscale particle-wafer surface interactions. Dashed line shows the boundary of surface layer formed by chemical agents during polishing. The figure shows that the bigger pa rticles (100 nm) can have larger indentation volume on the surface layer wh ich leads to higher friction force and more efficient material removal. Conclusion We have investigated the role of part icle size below 100 nm in mean diameter during chemical mechanical polishing (CMP) of copper. Our experiment results exhibit following; 1. Copper removal rate increases as the nanoparticle size a nd concentration increase.

PAGE 82

69 2. Microscratches were not found on surface of nay copper specimen after polishing. The RMS values were very low (< 0.7 nm) and were independent of nanoparticle size and concentration. 3. In situ friction force also increases as the nanoparticle size and concentration increase. It is believed that, due to the shallow indentation characteristics of the nanoparticles, the material removal occurs within the surface layer formed by chemical additives without severely damaging the sub ba re copper. In situ friction force result suggests that the increased copper removal rate with increasing nanoparticle size is attributable to the larg er indentation volume.

PAGE 83

70 CHAPTER 7 CONCLUSION We have investigated copper CMP using na noparticle-based slurries. With addition of citric acid in the presence of hydrogen pe roxide and benzotriazole at pH 7, we obtained moderate removal rates with very low rms values, less than 0.7 nm. Micron-size silica and 200 nm alumina particles were also used as comparison. 1. The hardness and size of the particles can be reduced to lower the penetration depth, thus avoiding damagi ng effects. However, reducing particle hardness and size without use of synergistic chemistries may lead to much reduced removal rates. 2. The addition of complexing agent, citric acid, in the slurry is critical to the formation of surface layer removable by the colloidal silic a nanoparticles. 3. In contrast, larger or harder particles result in greater surface defects suggesting that removal takes place via scratching process. We propose that citric acid softens the surface layer formed by hydrogen peroxide and benzotriazole, and results in larger indentation depth of nano-size silica particles, thus increasing the removal rate. Combined mechanical-chemical effects are needed to achieve low-defectivity polishing of copper. We investigated the role of citric acid during copper CMP using silica nanoparticlebased slurries. Linear increase in removal ra te was found with an increase of citric acid content in the presence of hydrogen peroxide and benzotriazole at neutral pH. Static dissolution studies and x-ray photoelectron spectroscopy sugges ts that coppe r dissolution is enhanced by the mixture of hydrogen per oxide and citric acid mostly via oxide dissolution. However, the static etch rate of copper is essentially minimal with addition of

PAGE 84

71 BTA but is not well correlated with the removal rate result. In contrast, the transient electrochemical study suggests that reacti on kinetics in the m illisecond regime is enhanced by citric acid add ition even in the presence of BTA, which we believe leads to less dense oxide formation and therefore allows considerable indent ation of silica nano particles. This is also confirmed by our results of nanoindentation and XRR, which exhibit that the addition of citric acid in the presence of hydrogen peroxide and benzotriazole significantly decreases the ha rdness and density of the surface layer. Therefore, using silica nanoparticle-based neut ral slurries, the linea r increase in removal rate as a function of citric acid concentra tion can be explained by formation of weakly bonded surface layer, rather than high solubility of copper. We have investigated the role of part icle size below 100 nm in mean diameter during chemical mechanical polishing (CMP) of copper. Our experiment results exhibit following; 1. Copper removal rate increases as the nanoparticle size a nd concentration increase. 2. Microscratches were not found on surface of nay copper specimen after polishing. The RMS values were very low (< 0.7 nm) and were independent of nanoparticle size and concentration. 3. In situ friction force also increases as the nanoparticle size and concentration increase. It is believed that, due to the shallow indentation characteristics of the nanoparticles, the material removal occurs within the surface layer formed by chemical additives without severely damaging the sub ba re copper. In situ friction force result suggests that the increased copper removal rate with increasing nanoparticle size is attributable to the larg er indentation volume.

PAGE 85

72 LIST OF REFERENCES [Aks02] S. Aksu and F.M. Doyle, J. Electrochem. Soc., 149, 6 (2002) p. G352 [Aks03] S. Aksu, L. Wang, and F.M. Doyle, J. Electrochem. Soc., 150, 11 (2003) p. G718 [Bas00] G.B. Basim, J.J. Adler, U. Mahajan, R.K. Singh, and B.M. Moudgil, J. Electrochem. Soc., 147, 9, (2000) p. 3523 [Bas03] G.B. Basim, I.U. Vakarelski, and B.M. Moudgil, J. of Colloid. And Interface. Sci., 263 (2003) p. 506 [Bie99] M. Bielmann, U. Ma hajan, and R.K. Singh, Electrochem. Solid-State Lett., 2, 8 (1999) p. 401 [Bou02] D. Bouvet, P. Beaud, P. F azan, R. Sanjines, and E. Jacquinot, J. Vac. Sci. Technol. B, 20, 4 (2002) p. 1556 [Bro81] N.J. Brown, P.C. Baker and R.T. Maney, Proc. SPIE, 306 (1981) p. 42 [Bru92] C.R. Brundle, C.A. Evans Jr. and S. Wilson, “Encyclopedia for Materials Analysis-Surfaces, Interfaces, Thin Films,” Elsevior (1992) [Cha03] Chang, J. Shieh, B. Dai, M. Feng, Y. Li, C.H. Shih, M.H. Tsai, S.L. Shue, R.S. Liang, and Y. Wang, Electrochem. and Solid-State Lett., 6, 5 (2003) p. G72 [Che04] J. Chen and W. Tsai, Mater. Chem. Phys., 87, 2-3 (2004) p. 387 [Cho04] W. Choi, S. Lee, J. T. Abiade, and R.K. Singh, J. Electrochem. Soc. 151, 5 (2004) p. G368 [Coo90] L.M. Cook, J. Non-Cryst. Solids, 120 (1990) p. 152 [Du04] T. Du, A. Vijayakumar, and V. Desai, Electrochimica Acta, 49, 25 (2004) p. 4505 [Gre66] J.A. Greenwood and J.B. Williamson, Proc. Royal Soc. A, 395 (1966) p. 300 [Gre67] J.A. Greenwood, J. Lubrication Tech. Jan. (1967) p. 81 [Intel] www.intel.com

PAGE 86

73 [Kau91] F.B. Kaufmann, D.B. Thompson, R.E. Broadie, M.A. Jaso, W.L. Githrie, D.J. Pearson, and M.B. Small, J. Electochem. Soc. 138 (1991) p. 3460 [Kne97] E.A. Kneer, C. Raghunath, J.S. Jeon, and S. Raghavan, J. Electrochem. Soc., 144 (1997) p. 1934 [Kon03] S. Kondo, B.U. Yoon, S. Tokitoh, K. Misawa, S. Sone, H.J. Shin, N. Ohashi and N. Kobayashi, IEEE Technical Digest Interna tional Electron Devices Meeting, (2003) p. 151 [Lee03] S. Lee, dissertation, “Characterization of Chemic al Interactions during Chemical Mechanical Polishing of Copper,” University of Florida (2003) [Lu03] Z. Lu, S. Lee, S.V. Babu, and E. Matijevic, J. Colloid and Interface Sci. 261 (2003) p. 55 [Mah99a] U. Mahajan, M. Bielmann, and R.K. Singh, Electrochem. and Solid-State Lett. 2, 1 (1999) p. 46 [Mah99b] U. Mahajan, M. Bielmann, and R.K. Singh, Electrochem. and Solid-State Lett. 2, 2 (1999) p. 80 [Mah00] U. Mahajan, M. Bielmann, and R.K. Singh, Materials Research Society Symposium, 566 (2000) p. 27 [Mur00] S.P. Murarka, I.V. Verner, and R.J. Gutmann, “Copper-fundamental mechanisms for microe lectronic applications” (2000) John Wiley, New York [Oer85] G. Oertel, “Polyurethane Handbook,” Macmillan, New York, 1985 [Pad03] D. Padhi, J. Yahalom, S. Gandikota, and G. Dixit, J. of Electrochem. Soc., 150, 1 (2003) p. G10S [Pou74] M. Pourbaix, “Atlas of Electrochemical Equilibria in Aqueous Solution”, NACE, Houston, TX (1974) [Pre27] F. Preston, J. Soc. Glass Technol., 112, (1927) p. 14 [Qin04] K. Qin, B. Moudgil, and C. Park, Thin Solid Films, 446, 2 (2004) p. 277 [Ric71] R. W. Rice, in Materials Science Research, Vol. 5: Ceramics in Severe Environments (1971) p. 195, edited by W. W. Kriegel and H. Palmour III, Plenum, New York [Rya05] J.G. Ryan, R.M. Geffken, N.R. Poulin, and J.R. Paraszczak, IBM J. Res. Dev., 39, 4 (1995) p. 371 [Sho00] L. Shon-Roy, “CMP: Market trends and tec hnology,” Solid State Technology, 43, 6 (2000) p. 67

PAGE 87

74 [Sin02a] R.K. Singh and R. Bajaj, Material Research Society Bulletin, 27 (2002) p. 743 [Sin02b] R.K. Singh, S. Lee, K. Choi, D.B. Basim, W. Choi, Z. Chen, and B.M. Moudgil, Material Research Society Bulletin, 27 (2002) p. 752 [Sin03] R.K. Singh and S. Lee, 8th International Proceedings, Chemical Mechanical Planarization for ULSI Interc onnection Conference (CMP-IMIC), Feb. (2003) p. 128 [Ste97] J. M. Steigerwald, S.P. Murarka, and R.J. Gutmann, “Chemical Mechanical Planarization of Microe lectronic Materials” (1997) John Wiley & Sons, New York [Ste99] D.J. Stein, D.L. Hetherington, and J.L. Cecchi, J. Electrochem. Soc., 146 (1999) p. 1934 [Tam02] S. Tamilmani, W. Huang, S. Raghavan, and R. Small, J. Electrochem. Soc., 149, 12 (2002) p. G638 [Tic99] J. Ticky, J.A. Levert, L. Shan, and S. Danyluk, J. Electrochem. Soc., 146 (1999) p. 1523 [Yu93] Y. Yu, C.C. Yu, and M. Orlowski, IEEE IEDM, 865, 65 (1993) p. 35.4.1 [Wan01] S. Wang, G. Grover, C. Ba ker, J. Chamberlain, and C. Yu, Solid State Technology, September (2001) p. S9 [Zha02] Y. Zhao and L. Chang, Wear, 252 (2002) p. 220 [Zho02] C. Zhou, L. Shan, J.R. Hight, S. Danyluk, S.H. Ng, and A.J. Paszkowski, Tribology Transactions, 45, 2 (2002) p. 232

PAGE 88

75 BIOGRAPHICAL SKETCH Su-Ho Jung was born on May 9, 1973, in Se oul, South Korea. He graduated in 1999 with a bachelor’s degree in material s science and engineering from Kangwon National University in Chunchon, South Kor ea. While he was pursuing the bachelor’s degree, he served in Korean National Army for a year and a half, and also spent two semesters at Cornell University, Ithaca, New York, attending an intensive English program. He started his graduate study in the department of materials science and engineering at the University of Florida, Ga inesville, Florida, in fall 1999. His master’s research was supported by the Department of Energy and focused on the oxide ionic conductors for the fuel cell application under the advisory of Professor Eric Wachsman. He graduated in fall 2001 with a master’s degree and a thesis titled “Cubic Bismuth Oxides and Their Application in Low Temper ature Solid Oxide Fuel Cells.” In spring 2002, he joined the Professor Rajiv Singh’s re search group for his Ph.D. work and started an investigation mainly on understanding the copper chemical-mechanical polishing (CMP) mechanism towards low-defect purposes His work also ranged over versatile subjects such as compound semiconductor CM P and shallow trench isolation CMP. He graduated from the University of Florida with a doctoral degree in materials science and engineering in May 2005.