<%BANNER%>

Externally Induced Agglomeration during Chemical Mechanical Polishing of Metals and Dielectrics

Permanent Link: http://ufdc.ufl.edu/UFE0022879/00001

Material Information

Title: Externally Induced Agglomeration during Chemical Mechanical Polishing of Metals and Dielectrics
Physical Description: 1 online resource (122 p.)
Language: english
Creator: Chang, Feng
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2008

Subjects

Subjects / Keywords: agglomeration, cmp, defectivity, interparticle, pump, slurry, stability
Materials Science and Engineering -- Dissertations, Academic -- UF
Genre: Materials Science and Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: With the miniaturization of semiconductor devices, chemical mechanical planarization (CMP) is now commonly employed for both the front and back end processing of such devices due to its unique global planarization capability. Highly stable slurry plays an important role in minimizing the process-induced particle agglomeration and ensuring a superior polishing performance. However, abrasive particles can be agglomerated during the slurry formulation and handling processes. The presence of oversize particles in the slurry is one of the main causes of defectivity during CMP of metals and dielectrics. We investigated the stress effects on slurry agglomeration and particle-induced defectivity during CMP of metals and dielectrics. Our results indicate that the agglomeration of the slurries was found to depend both on the external shear stress and the interparticle forces acting on the slurries. The magnetically levitated centrifugal pump caused low shear flow and insignificant increase in agglomerated particles during the handling process as compared to the positive displacement pump. In addition, the interparticle forces were determined by the colloidal probe technique. An increase of repulsive interparticle force corresponded to a considerable decrease in agglomerated particles during the handling process. The addition of chemicals (e.g., salt, pH, and surfactant) into ceria and silica slurries can influence the slurry stability and the polishing performance. A positive correlation was established between the roughness/defect density and the degree of agglomeration. Slurry stability is qualitatively assessed through interparticle force and/or zeta potential measurements. However, these types of measurements do not fully describe agglomeration phenomena in CMP slurries. Therefore, we developed a novel experimental/theoretical approach for quantifying the degree of agglomeration in CMP slurries. This method involves subjecting the slurry to high shear forces and measuring particle agglomeration characteristics in their tail distribution. By modeling changes in tail distribution using Smoluchowski?s slow aggregation theory, the agglomeration index can be used to quantify the degree of agglomeration caused by external shear stress and internal slurry chemistry. Our modeled and experimental results revealed that the shear stress in a positive displacement pump was 100 times greater than that in a magnetically levitated centrifugal pump. In addition, our numerous polishing experiments revealed that slurries with higher agglomeration index values (AI > 1.8) contained more agglomerated particles during the handling process and caused more surface defectivity during polishing. Our novel method can be applied to determine slurry stability, and thus further minimize particle-induced defectivity during CMP of metals and dielectrics.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Feng Chang.
Thesis: Thesis (Ph.D.)--University of Florida, 2008.
Local: Adviser: Singh, Rajiv K.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2009-06-30

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2008
System ID: UFE0022879:00001

Permanent Link: http://ufdc.ufl.edu/UFE0022879/00001

Material Information

Title: Externally Induced Agglomeration during Chemical Mechanical Polishing of Metals and Dielectrics
Physical Description: 1 online resource (122 p.)
Language: english
Creator: Chang, Feng
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2008

Subjects

Subjects / Keywords: agglomeration, cmp, defectivity, interparticle, pump, slurry, stability
Materials Science and Engineering -- Dissertations, Academic -- UF
Genre: Materials Science and Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: With the miniaturization of semiconductor devices, chemical mechanical planarization (CMP) is now commonly employed for both the front and back end processing of such devices due to its unique global planarization capability. Highly stable slurry plays an important role in minimizing the process-induced particle agglomeration and ensuring a superior polishing performance. However, abrasive particles can be agglomerated during the slurry formulation and handling processes. The presence of oversize particles in the slurry is one of the main causes of defectivity during CMP of metals and dielectrics. We investigated the stress effects on slurry agglomeration and particle-induced defectivity during CMP of metals and dielectrics. Our results indicate that the agglomeration of the slurries was found to depend both on the external shear stress and the interparticle forces acting on the slurries. The magnetically levitated centrifugal pump caused low shear flow and insignificant increase in agglomerated particles during the handling process as compared to the positive displacement pump. In addition, the interparticle forces were determined by the colloidal probe technique. An increase of repulsive interparticle force corresponded to a considerable decrease in agglomerated particles during the handling process. The addition of chemicals (e.g., salt, pH, and surfactant) into ceria and silica slurries can influence the slurry stability and the polishing performance. A positive correlation was established between the roughness/defect density and the degree of agglomeration. Slurry stability is qualitatively assessed through interparticle force and/or zeta potential measurements. However, these types of measurements do not fully describe agglomeration phenomena in CMP slurries. Therefore, we developed a novel experimental/theoretical approach for quantifying the degree of agglomeration in CMP slurries. This method involves subjecting the slurry to high shear forces and measuring particle agglomeration characteristics in their tail distribution. By modeling changes in tail distribution using Smoluchowski?s slow aggregation theory, the agglomeration index can be used to quantify the degree of agglomeration caused by external shear stress and internal slurry chemistry. Our modeled and experimental results revealed that the shear stress in a positive displacement pump was 100 times greater than that in a magnetically levitated centrifugal pump. In addition, our numerous polishing experiments revealed that slurries with higher agglomeration index values (AI > 1.8) contained more agglomerated particles during the handling process and caused more surface defectivity during polishing. Our novel method can be applied to determine slurry stability, and thus further minimize particle-induced defectivity during CMP of metals and dielectrics.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Feng Chang.
Thesis: Thesis (Ph.D.)--University of Florida, 2008.
Local: Adviser: Singh, Rajiv K.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2009-06-30

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2008
System ID: UFE0022879:00001


This item has the following downloads:


Full Text

PAGE 1

1 EXTERNALLY INDUCED AGGLOMERATION DURING CHEMICAL MECHANICAL POLISHING OF METALS AND DIELECTRICS By FENG-CHI CHANG 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 2008

PAGE 2

2 2008 Feng-Chi Chang

PAGE 3

3 To my family

PAGE 4

4 ACKNOWLEDGMENTS First I am deeply appreciative of my adviso r, Dr. Rajiv Singh, for his research advice, support, and encouragement. His creative thinki ng and enthusiastic attitude for fundamental material research are deeply impressed on my mi nd. I would also like to thank my committee, Dr. Stephen Pearton, Dr. David Nort on, Dr. Wolfgang Sigmund, and Dr Fen Ren, for their advice and support. I am also very thankful to Dr. Brij Moudgil for his many insightful comments in particle science and CMP research. I am extremely grateful to Levitronix LLC for their financial and experimental support, and the industrial partners of the Particle E ngineering Research Center (PERC) for financial support throughout my graduate program. Many thanks are given to all of my friends a nd colleagues for their friendship and helpful discussion. I would like to thank my past a nd present group members, Siddharth Tanawade, Purushottam Kumar, Sejin Kim, Jaeseok Lee, Sushant Gupta, Taekon Kim, Yoonjae Moon, Aniruddh Khanna, and Myoung Hwan, for their contributions to this research with valuable discussions and friendship, and my good friends, Jiahau Yan, Ti enyu Chang, Yu-Ping Huang, for their friendship and memorable time in Gainesville. I would like to thank Gary Scheiffele and G ill Brubaker in PERC for their dedication and experimental support throughout my research. I would also like to thank Kyoung-Ho Bu for his helpful research support and discussion. I would like to express my gr atitude to family members: my parents, Cheng-Tsung Chang and Li-Yu Li, and my brothers, Feng-Shih Chang and Chung-Hao Chang, for their love, encouragement, and great anticipation for me. Fina lly, I would like to tha nk my cutest wife, JiaYuan Chang, for her love and optimistic attit ude to encourage and accompany me go through my graduate studies.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS...............................................................................................................4LIST OF TABLES................................................................................................................. ..........7LIST OF FIGURES.........................................................................................................................8ABSTRACT...................................................................................................................................11CHAPTER 1 INTRODUCTION..................................................................................................................13Chemical Mechanical Polishing............................................................................................. 13Slurry Handling......................................................................................................................142 LITERATURE REVIEW.......................................................................................................19The CMP Process and Slurry.................................................................................................. 19The Cu/Low K CMP....................................................................................................... 19Process flow of Cu/low k CMP................................................................................21Mechanisms of Cu/low k slurries............................................................................. 22Shallow Trench Isolation CMP.......................................................................................24Process flow of STI CMP.........................................................................................25Mechanisms of STI slurries...................................................................................... 25Slurry Stability............................................................................................................... .........27Electrostatic Stability.......................................................................................................27Polymer Stabilization...................................................................................................... 28Particle Agglomeration......................................................................................................... ..30Slurry Aggl omera tion......................................................................................................30Effect of chemicals on pa rticle agglomeration......................................................... 30Effect of pump-induced pa rticle agglomeration....................................................... 31Model of the Particle Agglomeration.............................................................................. 32Particle-Induced Defectivity............................................................................................ 333 CMP CHARACTERIZATIONS............................................................................................ 44CMP System and Consumables.............................................................................................. 44Particle Characterization.........................................................................................................45Particle Size Measurement.............................................................................................. 45Interparticle For ce Measurement..................................................................................... 48Measurement of Electrical Po tential (Zeta Potential).....................................................49Polishing Characterization......................................................................................................50Polishing Parameter for Cu/Low K CMP........................................................................ 50Characterization of Polished Wafer................................................................................. 51

PAGE 6

6 4 EFFECTS OF STRESS-INDUCED PARTICLE AGGLOMERATION ON DEFECTIVITY DURING CMP OF LOW -K DIELECTRICS............................................. 59Introduction................................................................................................................... ..........59Experimental................................................................................................................... ........60Results and Discussion......................................................................................................... ..61Characterization of Oversi ze Particle Distribution.......................................................... 61Characterization of Low-k CMP.....................................................................................64Summary.................................................................................................................................655 ROLE OF INTERPARTICLE FORCES DURING PUMP-INDUCED AGGL OMERATION OF CMP SLURRIES.......................................................................... 74Introduction................................................................................................................... ..........74Experimental................................................................................................................... ........75Results and Discussion......................................................................................................... ..77Summary.................................................................................................................................806 ROLE OF SLURRY CHEMISTRY ON ST RESS-INDUCE D AGGLOMERATION.........88Introduction................................................................................................................... ..........88Experimental................................................................................................................... ........89Results and Discussion......................................................................................................... ..90Summary.................................................................................................................................927 NOVEL METHOD TO QUANTIFY THE DE GREE OF AGGLOMERATION IN HIGHLY STABLE CMP SLURRIES.................................................................................. 101Introduction................................................................................................................... ........101Experimental................................................................................................................... ......102Determining the Agglomeration Index.......................................................................... 102Effect of Agglomeration Index on Polishing Performance...........................................104Results and Discussion......................................................................................................... 104Summary...............................................................................................................................1068 CONCLUSIONS AND SUGGESTI ONS FOR FUTURE WORK ...................................... 114Conclusions...........................................................................................................................114Suggestions for Future Work................................................................................................ 115LIST OF REFERENCES.............................................................................................................117BIOGRAPHICAL SKETCH.......................................................................................................122

PAGE 7

7 LIST OF TABLES Table page 1-1 Various systems-level variables in CMP........................................................................... 172-1 Comparison between LOCOS and STI processes............................................................. 402-2 Properties of dielectric films............................................................................................. .432-3 Interconnect surface requirements at near-term technology nodes.................................... 433-1 Various measurement me thods for particle size................................................................ 557-1 Agglomeration indexes of typical chem ical mechanical polishing (CMP) slurries........ 110

PAGE 8

8 LIST OF FIGURES Figure page 1-1 Basic process of chemical mechanical polishing............................................................... 16 1-2 Typical slurry distribution system..................................................................................... 18 2-1 Role of CMP process in the manuf ac ture of a microprocessor device.............................. 36 2-2 Possibilities for reducing the dielectric constant. ..............................................................37 2-3 The classification of low k m aterials.................................................................................37 2-4 Daul-Damesence Process Flow for m aking MLM structure............................................. 38 2-5 Three polishing stages during Cu/Low k CMP.................................................................. 39 2-6 STI process flow........................................................................................................... .....41 2-7 Overlapping of two electrical double layers. ..................................................................... 42 3-1 Schematic illustration of a slurry distribution system........................................................54 3-2 Schematic illustration of the colloidal probe technique..................................................... 56 3-3 Polishing parameters....................................................................................................... ...57 3-4 Contact mode AFM: the cantilever deflection is measured by positive-sensitive photodiodes, which is utilized to detect lase r beam reflections from the back of the cantilever............................................................................................................................58 3-5 Topping mode AFM: the surface topography is measured by detecting the amplitude of the cantilever oscillation. ...............................................................................................58 4-1 Primary particle size as a function of sl urry turnovers: low-k slurries circulatedby m aglev, bellows, and diaphragm pumps............................................................................67 4-2 Cumulative concentration vs. particle size at 0, 250, and 500 turnovers for Bellows, diaphragm and magnetically levitated centrifugal pump systems.................................... 68 4-3 Normalized oversize partic le distribution for positive di splacem ent and magnetically levitated centrifugal pumps................................................................................................ 69 4-4 Optical images............................................................................................................. ......70 4-5 Scratch density as a function of turnovers:........................................................................ 71 4-6 Comparison of surface roughness...................................................................................... 72

PAGE 9

9 4-7 Defectivity vs. norm alized oversize particles.................................................................... 73 5-1 SEM image of 5 m silica probe........................................................................................ 82 5-2 Circulated silica slurries for 1000 turnovers at pH 2, 7, and 11:....................................... 83 5-3 Force versus distance between a silica substrate and silica pro be in supernatant slurries at pH 2, 7, and 11, with and without 0.1 M KCl................................................... 84 5-4 Zeta potential of 30 nm silica slurries as a function of pH, wi th and without 0.1 M KCl.....................................................................................................................................85 5-5 Correlations between repulsive interacti on forces and m ean values of normalized oversize particles as a function of pH, with and without 0.1 M KCl................................. 86 5-6 Surface roughness as a function of pH.............................................................................. 87 6-1 Effects of stress-induced particle agglom eration in unformulated ceria and silica slurries................................................................................................................................93 6-2 Plot of zeta potential vs. pH: the potentia l curv es of un-formulated silica and ceria slurries................................................................................................................................94 6-3 Plot of zeta potential vs. pH: the potential curv es of formulated ceria slurries with varying surfactants.............................................................................................................94 6-4 Comparison of particle agglomeration in form ulated and un-formulated ceria slurries due to stress effect..............................................................................................................95 6-5 Plot of zeta potential vs. pH in cer ia slurry with cationic surfactant. ................................ 96 6-6 Dynamic light sca tte ring (DLS) experiment: the part icle size distri butions of 80nm silica slurries with surfactants............................................................................................ 97 6-7 Comparison of particle agglomeration in form ulated and un-formulated silica slurries. ............................................................................................................................................98 6-8 Plot of zeta potential vs. pH in silica slurry with surfactants.............................................99 6-9 Silica CMP................................................................................................................. ......100 7-1 Schematic illustration of how the agglom eration index is determined............................107 7-2 Tail distributions of as-received slur ry and circulated s lurries by positive displacement and maglev centrifugal pumps................................................................... 108 7-3 Effect of external shear stress on the a gglom eration index: experimental and modeled tail distributions in positive displace ment and maglev centrifugal pumps...................... 109

PAGE 10

10 7-4 Effect of internal slurry chemistry on the agglo meration inde x: experimental and modeled tail distributions in circulated sl urries (80 nm silica) at pH 9 and 3................. 111 7-5 Cu chemical mechanical polishing.................................................................................. 112 7-6 Plot of the agglomeration index ( AI ) versus typ ical chemical mechanical polishing (CMP) slurries in determining particle-induced defectivity............................................ 113

PAGE 11

11 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 EXTERNALLY INDUCED AGGLOMERATION DURING CHEMICAL MECHANICAL POLISHING OF METALS AND DIELECTRICS By Feng-Chi Chang December 2008 Chair: Rajiv K. Singh Major: Materials Science and Engineering With the miniaturization of semiconductor de vices, chemical mech anical planarization (CMP) is now commonly employed for both the front and back end processing of such devices due to its unique global planarization capability. Highly stable slurry plays an important role in minimizing the process-induced particle aggl omeration and ensuring a superior polishing performance. However, abrasive particles can be agglomerated during the slurry formulation and handling processes. The presence of oversize particles in the slurry is one of the main causes of defectivity during CMP of me tals and dielectrics. We investigated the stress effects on slurry a gglomeration and particle-induced defectivity during CMP of metals and dielectrics. Our result s indicate that the agglomeration of the slurries was found to depend both on the external shear st ress and the interparticle forces acting on the slurries. The magnetically levita ted centrifugal pump caused low shear flow and insignificant increase in agglomerated particles during the handling process as compared to the positive displacement pump. In addition, the interparticle fo rces were determined by the colloidal probe technique. An increase of repulsive interparticl e force corresponded to a c onsiderable decrease in agglomerated particles during the handling process. The addition of chemicals (e.g., salt, pH, and surfactant) into ceria and silica slurries can influence the slurry stability and the polishing

PAGE 12

12 performance. A positive correlation was establis hed between the roughness/defect density and the degree of agglomeration. Slurry stability is qualitativel y assessed through interparticle force and/or zeta potential measurements. However, these types of measur ements do not fully describe agglomeration phenomena in CMP slurries. Therefore, we devel oped a novel experimental/t heoretical approach for quantifying the degree of agglomeration in CMP slurries. This method involves subjecting the slurry to high shear forces and measuring particle agglomeration characteristics in their tail distribution. By modeling changes in tail distribution using Sm oluchowskis slow aggregation theory, the agglomeration index can be used to quantify the degree of agglomeration caused by external shear stress and internal slurry chem istry. Our modeled and experimental results revealed that the shear stress in a positive displa cement pump was 100 times greater than that in a magnetically levitate d centrifugal pump. In addition, our numerous polishing experiments revealed that slurries with higher agglomeration index values ( AI > 1.8) contained more agglomerated particles during the handling proc ess and caused more su rface defectivity during polishing. Our novel method can be applied to de termine slurry stability, and thus further minimize particle-induced defectivity dur ing CMP of metals and dielectrics.

PAGE 13

13 CHAPTER 1 INTRODUCTION Chemical Mechanical Polishing Che mical mechanical polishi ng (CMP) was first introduced by IBM in the 1980s to fabricate multilevel interconnect structure in silicon integrated ci rcuits (IC) [Z an04]. Multilevel metallization (MLM) is established by depositing th e interlayer dielectrics (ILD), the tungsten as the inlaid metal, and the meta l interconnect as the planar interconnect. To improve the device speed, the semiconductor industry had replaced SiO2 ILD and Al interconnects with low k dielectrics and Cu interconnects, respectively. Curre ntly, the nine layers of Cu interconnect with low k ILD have developed for semiconductor manufacturing at 45nm technology node [Mis07]. With the shrinking of feature size in devices the semiconductor industry has reported many challenges. For example, to fabricate miniature and dense interconne cts, a low step height of the surface topography is an important requirement to accomplish the high resolution photolithography. To meet challenges such as this CMP has become an unavoidable solution for the processing of MLM structure due to its unique global planarization cap ability. Advantages of the CMP process in semiconductor manu facturing include the ability to Achieve better global planarization Planarize a large variety of materials ( e.g., Cu, W, Al, Ta, TaN, SiC, SiO2, and Si3N4) Reduce prior step defects Achieve multilevel metallization Improve metal step coverage Provide higher photolithography Replace reactive ion etching (RIE) and plasma etching with etching metals and alloys. Typically, CMP is a complicated process to control and understand because there are more than 20 input variables (e.g., particle types, chem icals in the slurries, physical conditions, and pad characteristics) necessary to achieve corr esponding output parameters (e.g. removal rate and planarization), as shown in Ta ble1-1 [Sin02]. The CMP system comprises the polishing pad,

PAGE 14

14 carrier head, and slurry distribut ion system, as shown in Figure 1-1. During the polishing process, CMP slurry is supplied from the slurry distribution system and the wafer is pressed against the polishing pad. The material removal rate is c ontrolled by adjus ting the rotation speed of the platen and down pressure of the carrier head. Typically, the angul ar velocity of the carrier and platen keep the same direction. Slurry Handling CMP slurry is a synergistic combination of particles and chemicals, and is perhaps the most critical consumable product in the semic onductor industry as it cont rols the uniqueness and technical performance of the CMP process. Abrasi ve particles in slurries typically provide the mechanical component of the polishing pro cess. To reduce/minimize process dependent defectivity is to ensure that the particle size di stribution, especially in the oversize tail region, is as small as possible and does not increase with time during the slurry delivery and polishing process. The slurry distribution system is co mprised of daytanks, chemical blending vessels, slurry loops, and distribution pumps, as can be seen in Figure 1-2. As-re ceived slurries were diluted and mixed with chemicals in the daytank a nd then distributed in the slurry loop using a pumping device. Some studies have shown th at positive displacement pumps (e.g., bellows and diaphragm) generate high shear stress and tend to agglomerate particles during slurry handling [Bar99, Lit04, Sin04]. Agglomerated particles can settle on the surface of wafers or increase the mechanical stress that can lead to increased defectivity (e.g., scratches, partic le residues, and pits) during the polishing. Therefore, th e control and reduction of oversiz e particles are critical to further reduce surface defectivity during CMP. This especially pertains to the 45 nm semiconductor technology manufacturing node, wh ich is expected to be less tolerant to defects. This study fundamentally focuses on the effect of stress-induced particle agglomeration and defectivity during CMP of metals and dielectrics. The magnitude of agglomerate

PAGE 15

15 concentrations is determined by the particle type s, slurry chemicals, and external mechanical forces applied on the slurry itself during the slu rry handling process. The fundamental of slurry stability and stress-induced part icle agglomeration has studie d. To minimize particle-induced defectivity, we also develop a novel approach to measure the robustness/stability of CMP slurries. Chapter 4 investigates the effects of shear stress induced by positive displacement pumps and centrifugal pumps on particle agglomerati on and polished defectivity. These studies were carried out by measuring the oversize particle distribution and the surface defectivity generated during the CMP of low k dielectrics. Chapter 5 delineates the chem ical effects on pump-induced particle agglomeration. The agglomeration of the slurries was found to de pend both on the external shear stress and the interparticle forces acting on the slurries. Th e correlations were es tablished between the repulsive interaction forces and surface defectivity due to oversize particles. Chapter 6 studies the slurry formulation in order to stabilize ceri a and silica colloidal particles to minimize the process-induced par ticle agglomeration. The slurry stability was investigated from the polishing performance. Chapter 7 develops a novel expe rimental/theoretical approach to determine the degree of agglomeration in highly stable CM P slurries. This approach can be applied to minimize particleinduced defectivity during CMP. Chapter 8 summarizes the conclusions of th is study and suggests future research.

PAGE 16

16 Down Pressure Conditioner Platen Polishing Pad Slurry Feed Sample Holder Wafer Down Pressure Conditioner Platen Polishing Pad Slurry Feed Sample Holder Wafer Figure 1-1. Basic process of chemical mechanical polishing.

PAGE 17

17 Table 1-1. Various systems-level variables in CMP [Sin02]. Input variables Microscale para meters Nanoscale interaction Output parameters Particle Characteristics Size and distribution Shape Mechanical properties Chemistry Concentration Agglomeration Slurry Chemistry Oxidizers Buffering agents Complexing agents Dispersants Down Pressure and Linear Velocity Pad Characteristics Mechanical properties Topography Conditioning Substrate Characteristics Feature size Feature density Pad Contact angle Pressure on the pad Particles on the Pad Pressure Coverage Chemical Concentration and Distribution Contact Mode Direct Mixed Hydroplaning Chemomechanical Dynamics of surface layer formation Layer removal mechanism Abrasion frequency Chemical and Mechanical Etching Mechanical removal Removal Rate Planarization Surface Finish Defectivity

PAGE 18

18 Day Tank Recirculation Dilution &Mixing Station Pumping DeviceSlurry Feed CMP Machine Nitrogen Day Tank Day Tank Recirculation Dilution &Mixing Station Dilution &Mixing Station Pumping DeviceSlurry Feed CMP Machine CMP Machine Nitrogen Figure1-2. Typical slurry distribution system.

PAGE 19

19 CHAPTER 2 LITERATURE REVIEW The CMP Process and Slurry Chemical mechanical planar ization (CMP) is now commonl y employed for both the front and back end processing of IC devices due to its unique global planar ization capability. This process includes shallow trench isolation (STI), pre-metal dielectrics (PMD), tungsten, interlayer dielectric (ILD), and copper CMP, as shown in Figure 2-1. Each specific CMP process uses slurries that are specifically designed for that process to ensu re the polishing performance (e.g., removal rate and selectivity) a nd minimize the process-dependent defectivity (e.g., scratches and delamination). In this chapter, we discuss the CMP processes ( e.g., Cu/ low k and dielectric polishing), fundamentals of slurry stability, a nd process-dependent def ectivity due to oversize particles. The Cu/Low K CMP By m iniaturizing the feature size of semi conductor devices, the number of transistors increases: the Semiconductor Industry Associat ion (SIA) has projected that high performance chips will contain more than 3 billion transistors in 2011 [Nai02]. Thus, more reliable and efficient interconnect system in the back-end-of-line (BEOL) technology is essential for manufacturing the new generation of IC devices. Consequently, the multile vel interconnect is a reliable solution to provide dire ct routing and improve the perf ormance of IC devices, and the CMP process is a dependable technology to manuf acture multilevel metallization structures due to its unique global plan arization capability. It is well know that the in terconnect delay becomes the dom inating issue affecting device speed, called RC delay time, as the feature size of devices continues to scale down [Cha04a and Mur93]. RC time ( ) is the required time to charge nearly 63% of the total volta ge and is affected

PAGE 20

20 by the resistance (R ) of interconnect and the capacitance ( C ) of dielectric materials [Sze01]. The RC time is given by tw l R [2-1] d wl Ci [2-2] CR [2-3] where is the resistivity of interconnect, i is the dielectric permittivity, l w and t are length, width, thickness of materials, and d is the thickness of dielectric film. The thinner and longer interconnection and larger c oupling capacitance can increase the RC delay time due to the decrease in the distance between adjusted metal lines [Sak93]. However, it is difficult to lower the parasitic capacitance by decreasing the thickness of the diel ectric film as the device shrinks. Thus, to lower the interconnect de lay, an alternative method is to replace aluminum interconnect and silicon dioxide with copper and low k materials, respectively. The dielectric constant (k) has represented the relative permittivity ( r) in the semiconductor industry. When applied to the voltag e in the device, the dielectric constant is determined by the polarization of materials. The Debye equation desc ribes the relationship between the relative permittiv ity and the polarization, kT Nde r r3 32 12 0 [2-4] where N is the number of molecules per cubic meter and 0 is the permittivity of vacuum. The polarization phenomena are represente d by the electronic polarization (e), the distortion polarization (d), and orientation polarization () [Mae03]. Thus, the dielectric constant can be reduced by introducing the materials, which have smaller molecule densities and less polar bonds, as shown in Figure 2-2 [Sha04].

PAGE 21

21 Furthermore, low k materials can be simply cl assified as the silica -based, silsesquioxane (SSQ)-based materials, and organic polymers, as shown in Figure 2-3 [Sha04]. The silica-based films, deposited by plasma-enhanced chemical vapor deposition (PECVD) or CVD, have a tetrahedral structure. To lower the k value of silicon dioxide, Si-O bonds are replaced with low polar bonds, such as Si-F and Si-CH3 bonds. Historically, the fluorinated silicon dioxide glass (FSG) is the first low k dielectric introduced into an ILD layer. The k values range from 3.5 to 3.7. In addition, organosilicate gl asses (OSG) are the materials of silicon dioxide doped with carbon or CH3 to reduce the film density and a number of polar bonds. Their dielectric constants are closed to 2.7 to 3.0. In regard to SSQ-based materials, these films are deposited by spincoating, and are cured for forming stable pol arization. The formula of the SSQ is (R-SiO3/2)n, where R can be H or CH3. Silicon and hydrogen form a cubic structure and hydrogen and methyl group are connected to the corner of the cubi c structure to form hydrogensilsesquioxane (HSSQ) and methyl-silsesquioxane (MSSQ) structures. Typically, the k-value range of these materials is 2.9 to 3.6. In addition, the organic po lymers have low dielectric constants (2.0-3.0) due to the non-polar bonds (e.g. C-C) and asymmetric charge distribution (e.g., C-F and C-F3). These materials can be deposited by CVD or spin-on polarization [Mae03 and Bor02]. Process flow of Cu/low k CMP As copper has excellent electr onic conductivity and re sistance to electromigration, it has successfully replaced aluminum in the process of interconnection metallization. However, a lack of suitable dry-etching solution for removing unwanted Cu materi als makes it more difficult to manufacture MLM structure. To address this pr oblem, the dual-damascene process is introduced into Cu/Low k process flow and successfully manufactures the multilevel metallization. The dual-damascene process includes two photolithography and reactive ion etch ing (RIE) steps to make the vias and trenches, as shown in Figure 2-4 [Sze01]. Afterwards, Ta/TaN and Cu layers

PAGE 22

22 are deposited onto via and trench structures, and the excess copper bulk is removed by CMP. As this structure contains copper bulk, barrier (Ta or TaN), and lo w k materials, three polishing stages during Cu CMP are required to achieve pl anarization. The first plat en is used mainly for removal of unwanted copper bulk above the pattern ed features. The Cu slurry used in the first platen must have a high removal rate and hi gh selectivity between th e barrier and copper materials. After removing unwanted Cu materials, the nonselective slurry is used in the second platen to remove residual Cu and barrier materials simultaneousl y. This barrier slurry, composed of silica abrasive particles, can remove copper and barrier metals at approximately the same removal rate to achieve the flat surface. Finally, the third platen is subjected to a buffing step for smoothing the dielectric surface and removing residual barrier metal, as shown in Figure 2-5 [Bal04]. Mechanisms of Cu/low k slurries Typically, two kinds of slurries Cu and barrier/low k slurrie s, are utilized in Cu/low k polishing. First, Cu slurry is designed for a hi gh removal rate and selec tivity between barrier and copper bulk materials. The chemicals in Cu sl urry are comprised of corrosion inhibitors, complexing agents, and oxidizers. The oxidizer, such as hydrogen peroxide [Ein07 and Den06] and ferric nitride [Ber08], reacts with Cu film to fo rm a soft oxide layer, which is easily removed by abrasive particles. The addition of the comple xing agent, such as glycine [Lin08 and Tam02] and oxalic acid [Pan07], can increase the solubility of Cu films or Cu oxides into Cu ions. The addition of corrosion inhibitor into Cu slurry, such as be nzotriazole (BTA) [Kim08a and Tam02], forms a protective film to defend against isotropic etching of aggressive chemicals. The phenomena of Cu dissolution and passivation in the Cu-water system can be illustrated from the electrochemical behavior. The copper can dissolve into Cu ions in the acidic solutions, or form Cu oxides in neutral or alkalin e solutions. The oxidizer, such as hydrogen peroxide, can further

PAGE 23

23 make a growth of Cu oxide films to obstacle th e ion transportation from the Cu surface. Eom et al. [Eom07] indicates that the co ncentration of Cu dissolution is proportional to the thickness of Cu oxide, which means that Cu oxides can preven t Cu dissolution in the solution. Furthermore, the addition of BTA in Cu slurries can form a monolayer of the Cu-B TA surface to prevent chemical attack on the Cu surface; however, Cu-BTA surface may cause a decrease of the removal rate during Cu CMP [Kim08a and Li08]. Consequently, electrochemical behavior plays an important role in achieving material remova l and preventing surface defectivity in metal CMP. Low k slurries, or Cu barrier slurries, are ut ilized in polishing ba rrier and dielectric materials to achieve excellent uniformity and surface finish. The mechanism of dielectric polishing is an extension of gla ss polishing. Glass polishing for optical applications, such as lens and window sheets, was developed for smooth and planarized surfaces a decade ago. The mechanism of glass polishing is based on the water-oxide interaction [Coo90, Cum95]. During the polishing, the water-silica reaction between siloxane bonds (Si-O-Si) and water forms soft SiOH layers. The reversible depolymerization r eaction can be described as [Ile79] 4 1 2 2 22 OHSi SiOOHSiOx x [2-5] The modified layers are removed by abrasive par ticles, and the material removal is determined by the hydration rate of siloxane ne twork. The hydration of silanol group ( SiOH) on silica surface, at pH below and above the isoelectric point of silica (IEP 2.2), creates positive ( SiOH2 +) and negative ( SiO) charges. 2SiOH HSiOH [2-6] OHSiO OHSiOH2 [2-7]

PAGE 24

24 Thus, the material removal rate in creases rapidly in alkaline slurri es due to a significant increase in the dissolution of the silanol group. Choi [Cho04a] studied the pH effect on the removal rate of silicon dioxide during CMP. He found that th e surface charges of sili ca particles increased rapidly with increasing the slurry pH, resulting in raised repulsive forces between silica particle and silica wafer. These repulsive forces not only increased the spatial distance of silica surfaces, but also provided the lubrica tion effect during the polishing, which caused a decrease in the material removal rate. When the threshold pH re ached 9, it caused a dramatic increase in silica polishing due to forming soft layers on the silic a surface. The abrasive particles can easily remove these modified layers during the polishing. The reduction of the dielectric constant is achieved by doping the materials that have low molecule density or polar bonds. These doping materials may affect the removal rate during CMP. For example, the FSG has a higher remova l rate than that of an un-doped TEOS oxide because fluorinated doped silica exhibits a lower elastic modulus and hardness [Tse97]. Conversely, the carbon-doped silica films (e.g., OSG, HSSQ, and MSSQ) have a low removal rate during CMP due to the hydrophobicity of the inorganic-organic films [Bor02]. Shallow Trench Isolation CMP Shallow tr ench isolation is the device isolati on process that achieves a high device-packing density in the front-end-of-line (FEOL) ma nufacturing [Asa95, Cha96, and Ois00]. As the feature size of the device scales down, the semiconductor industry has replaced the local oxidation of silicon (LOCOS) process with th e STI process because the LOCOS process is susceptible to causing a birds b eak structure in the region of th e active-isolation transition. This structure reduces the active area and results in a low device-packing density. Consequently, the

PAGE 25

25 STI process has been commonly empl oyed for the technology nodes at 0.25 m and beyond. The comparison between STI and LOCOS processes is listed in Table 2-1[Li08 and Ste96]. Process flow of STI CMP Figure 2-6 shows a typical process flow of STI CMP. STI process starts with the growth of the oxide thin pad and then the deposition of the nitr ide film on a raw silicon wafer. The thickness of the oxide thin pa d and nitride film are around 15-25 and 100-150 nm, respectively [Kah08]. The following lithography process patterns the isolation area where the nitride and oxide are etched by the reactive ion etching (R IE). Subsequently, the CVD oxide refills the trenches and the over-filled oxides are removed by CMP. Finally, nitride layers are removed by phosphoric acid [Sez01]. Mechanisms of STI slurries The strategy for designing STI slurry is to consider the slurry selectivity, which is the ratio of material removal rate of silicon dioxide to nitride, during the polis hing. If slurry contains high selectivity properties, not only can the unw anted silica be removed quickly, but also the surface uniformity can be achieved by using the endpoint detection [Kim08b and Bib98]. Bu et al. [Bu05 and Bu07] studied the e ffects of pH and surf actant concentration in silica slurries on polishing selectivity. They studied the effect of sodium dodecyl sulfate (SDS) concentration on the adsorption of silicon dioxide and silicon ni tride wafers. Because the isoelectric point of silicon nitride and silicon dioxide wafers are pH 4.5 and 2, silicon nitride exhibited a higher SDS surfactant adsorption at low pH as compared to silicon dioxide wafers due to the electrostatic attraction. Consequently, the maximum SDS adso rption on silicon nitride wafer was found at pH 2; however, an insignif icant increase in surfactant adsorpti on was found on the silicon dioxide wafer. Furthermore, the maximum polishing selec tivity of silica slurries was reached when the SDS concentration was beyond the critic al micelle concentration (~8mM).

PAGE 26

26 In addition, ceria particles have been comm only employed for glass polishing due to their high removal rate on silicon dioxide. The mech anisms of ceria polishing focus on chemical interactions between abrasive particles and surface groups on the substrate [Oss02 and Sup04]. Cook [Coo90] proposed the following chemical tooth for the glass polishing, listed below: 1. Water penetrates into the glass surface 2. Water reacts with the surface, which lead s to the dissolution under particle load 3. Abrasive particles adsorb some dissoluti on products and leave from the substrate 4. Some dissolution products re deposit onto the substrate 5. Surface dissolution happens between particle impacts CeO2 has a high polishing rate because the free energy of the formation of cerium oxide ( Hf = -260 kcal/mole) is much less than the fr ee energy of the formation of silicon dioxide ( Hf = -216 kcal/mole) [Ste96 and Kub67]. Maximum ma terial removal happens when a neutrally charged CeO2 particle approaches a sili ca substrate with negative surf ace charges to form surface chemical bonds. The condensation reaction can be expressed as OHCeOSiOHCeOSi [2-8] The material removal occurs when a silica tetrahed ron structure is broken from the silica surface because the strength of Ce O bonding is greater than Si O bonding [Oss02]. CeOSiOSiCeOSiOSi [2-9] Furthermore, the bonding effect on ceria polishing can be proven by measuring particle size and zeta potential in ceria slurries. Abiade et al. [Abi05] studied the material removal rate as a function of ceria slurry pH in silica polishing. They concluded that the maximum material removal rate occurred near neut ral slurry pH. The polished slurry at nearly neutral pH had a larger particle size, and the zeta potential show ed much more negative potential as compared to un-polished slurry. That means Si-O is adsorbed onto ceria pa rticles during the polishing.

PAGE 27

27 Slurry Stability Highly stab le colloidal suspension plays an important role in chemical mechanical polishing (CMP) slurries as it de termines the quality of polis hing performance. To minimize process-dependent defectivity (e.g., scratches and particle residue), abrasive particles should not agglomerate or settle down during the slurry blending and handling process. The dispersion stability can be achieved by two methods: the electrostatic inte raction between particles and polymer stabilization. Electrostatic Stability When two colloid al particles with the same ch arges approach each other, osmotic pressure tends to push particles away due to its high c ounterion concentrations, as shown in Figure 2-7 [Hun01, Ros04]. The overlapping of the electric al double layer potential causes a raise of repulsive potential. The repulsi ve double layer potential can be determined by Debye length 1/: n i ii rZCF RT k1 2 2 04 /1 [2-10] where r is the relative permittivity, 0 is the permittivity of a vacuum, R is a gas constant, T is temperature, F is the Faraday constant, Ci is the molar concentration of any ion in the solution phase, and Zi is the valence of ions in the solution phase For two spherical particles of radius a, when ka << 1 (i.e., small particles and a relativ ely thick electrical double layer), the repulsive potential can be expressed as kH r Re R a V2 0 2 [2-11] where 0 is the surface potential. Thus, the magnitude of the repulsive double layer potential is dependent on surface potential and chemical in the solution.

PAGE 28

28 Further, total surface potential in colloidal systems is illustrated by the Derjaguin-LandauVerwey-Overbeek (DLVO) theory, which demonstrates that the stability of colloidal particles is dependent on a balance between attractive Van der Waals poten tial and repulsive double layer potential. The total potential energy between coll oidal particles is expr essed as the sum of repulsive (VR) and attractive (VA) potential: RA totalVVV [2-12] Thus, the potential attraction energy for si milar spherical particles with radius a whose centers are separated by distance R can be expressed as H Aa VA12 [2-13] where H is the nearest distance be tween particle surfaces and A is the Hamaker constant. Polymer Stabilization Highly stable dispersions in an aqueous system can also be achiev ed by the addition of surfactants (e.g., ionic and nonioni c polymers) in the presence or absence of electrostatic interaction. Basim et al. [Bas03] indicate that the addition of cationic surfactants into silica slurries can provide barrier forces to stabilize colloidal suspensi ons. When the concentrations of CnTAB (trimethylammonium bromide) exceeded a cr itical micelle concentr ation, micelles were formed and adsorbed onto silica surfaces to provi de repulsive electrostati c interactions between particles to defend against the coagulation. The magnitude of repulsive forces was also dependent upon the polymer chains. They indicate that the potential ba rriers between silica particles were observed while the carbon chains of the CnTAB were larger than 8 (n > 8). Although higher repulsive barrier fo rces can stabilize colloidal par ticles, lower material removal rates were found during CMP because of the lubricant effect.

PAGE 29

29 The other polymer stabilization, steric stab ilization, utilizes non-ionic surfactant to stabilize colloidal particles [Mor02 and Hun01] The repulsion occurs when two adsorbed polymer layers on the particles approach each ot her. Napper [Nap83] demonstrated that polymer stabilization occurs due to di sfavored thermodynamics. The cha nge of the free energy of two approaching polymer particles, GF, can be expressed F F FSTHG [2-14] where subscript F represents the flocculation, HF and SF are the change of the enthalpy and entropy, and T is the temperature. Three conditions may achieve steric stabilization. The first approach is the enthalpic stabilization : positive changes of the enthalpy and entropy. If HF is larger than SF, the flocculation is disfavored. The fl occulation can be caused by raising the temperature. The second approach is entropic stabilization : negative changes of the enthalpy and entropy. SF is dominated by the change of energy. By lowering the temperature, the flocculation can occur. The third method is enthalpic-entropic stabilization : a positive change of enthalpy and a negative change of entropy. Th e flocculation can not be caused by changing temperature. Furthermore, Palla et al. [Pal00] investigat ed the stability of alumina particles in the presence of a mixture of surf actants. They indicate that a mixture of anionic and nonionic surfactants can efficiently stab ilize alumina particles because the anionic surfactants were attracted by charge surface of alumina particles and nonionic surfactants penetrated into anionic surfactant layers due to hydrocar bon chain interactions. The mech anisms of surfactant adsorption involved three phenomena. First, the driving force of the ad sorption of ionic surfactants depended on the length and concentration of hydrocarbon chains. Second, the hydrocarbon chains interact with an ionic and nonionic surfactants. Finally the magnitude of polymer-solvent

PAGE 30

30 interaction should be larger th an that of the polymer-polymer interaction to prevent coiled polymers, and the steric stabi lization occurred when nonionic surfactants were adsorbed onto particle surfaces. Particle Agglomeration CMP slurry is a synergistic combination of particles and chemicals, and is perhaps the most critical consumable in the semiconducto r industry as it controls the uniqueness and technical performance of the CMP process. Slurry stability plays an important role in ensuring the polishing performance. If the agglomerati on happens in the slurry storage or delivery, agglomerated particles not only can reduce the life time of filter in the slurry distribution system, but can also result in the increase of surface defectivity during the polishing process. Thus, the understanding of fundamental of particle agglomeration helps a ddress these issues. Slurry Agglomeration Slurry agglomeration is influenced by particle types, slurry chemicals, and the external mechanical forces applied on the slurry itself during the slurry handling pr ocess. Typically silica, ceria, and alumina are the most common abrasive particles in the CMP slurries. These solid oxides not only provide the mechanical behavi or to remove unwanted materials during the polishing, but also exhibit the interparticle forces to stabilize the suspensions. However, these abrasive particles along with numerous chemical s could deteriorate the slurry stability and influence the polishing performance. Effect of chemicals on particle agglomeration Chemicals in CMP slurries include oxidizer s, buffering agents, dispersants, complexing agents, and pH [Sin02 and Ste96]. Each chemical has its unique function applied in the polishing process. For example, adjusting pH in CMP slur ries is used to dissolve surface wafers for an increase in material removal rate. However, ag glomerated particles incr ease significantly when

PAGE 31

31 slurry pH is close to the isoe lectric point of the solid oxide due to weak interaction forces between particles [Par65, Ram00, and Fran00]. Furthermore, slurries have varying ionic strengths that can change the el ectrostatic interaction between particle-particle and particlewafer. The strength of the interaction depends on the concentration of salt addition in the slurry. The salt addition in CMP slurries can reduce the repulsive forces betwee n particles and wafer, leading to an increased material removal rate during CMP [Hay95]. However, the introduction of salt in CMP slurries can also reduce the strength of electrostatic interaction between particles, resulting in increased particle agglomeration and surface defectivity during CMP [Bas02]. Choi [Cho04b] studied the effect of i onic salts on slurry stability a nd the polishing rates. He found that the polishing rate increased with ionic streng th in stable slurry due to the uniform contact area between agglomerated partic les and wafer. However, larger agglomerated particles caused non-uniform contact in unstable slurries, resulting in a decrease in material removal rate. In addition, Palla et al. [Pal99] st udied the effect of oxidizer add ition on the stability of alumina slurry in tungsten CMP. Typically, oxidizer can form a passive layer on tungsten surface to prevent the underlying tungsten su bstrate from corrosion. He found that, after adding 0.1 M potassium ferricyanide to alumina slurry, part icles were agglomerated and settled due to numerous counter-ions shielding th e surface charges on the particles. Effect of pump-induced particle agglomeration To minimize the process-dependent defectivity, particle size dist ribution, especially the tail region representing oversized particles, should be as small as possible and should not increase with time during slurry delivery. Some studies have shown that positive displacement pumps (e.g., bellows and diaphragm) may generate high shear stress and cause particles to agglomerate during the slurry handling [Bar99 and Lit04]. Singh et al. [Sin01] investigated the pump effect on slurry agglomeration. Silica, ceria, and alumina sl urries were utilized to circulate in the slurry

PAGE 32

32 delivery systems, which were comprised of a bellows pump and a v acuum-pressure-dispensetechnology pump. They indicate that circulated silica-base d slurries contained more agglomerated particles as compared to ceria and alumina-based slurri es. Furthermore, more agglomerated particles were found in the bellows pump system than those of a vacuum-pressuredispense-technology pump system. Model of the Particle Agglomeration Colloidal particles present a behavior of Brownian random motion in an aqueous system. The coagulation of colloidal particles occurs due to particle collisions. If we do not consider repulsive forces between colloidal particles, the collision frequency increases with raising shear flow, which is called rapid coagul ation. On the other hand, if electrostatic interactions present between colloidal particles, the potential barrie r can defend against shear flow and then reduce the frequency of particle coll ision. This phenomenon is called slow coagulation. The kinetic theory of particle coagulation was first wo rked by von Smoluchowski [Smo17, Rus89, and Hun01]. His kinetic model follows the Brownian random motion and the rapid coagulation. The model assumes that the particle collisions are binary and proportional to the particle concentration. The aggregation rate of the k-fold aggregates, dNk/dt is given by the time evolution of the cluster size aggregates, i and j -folds, 1 12 1k ikik kl kji jiij kNkNNNk dt dN [2-15] where kij is the second-order aggregation constant. If colloidal particles are subjected to the uniform laminar flow, this aggregation constant, kij, can be expressed as a function of shear rate, G and particle size, a )( 3 4ji ijaaGk [2-16]

PAGE 33

33 Furthermore, the efficiency of particle collisi on is influenced by the el ectrostatic interaction between particles in the aqueous system. Total electrostatic interaction provides a potential barrier to hinder particle agglomeration, resulting in a decreas e of the aggregation rate. Thus, the rate of k-fold aggregates can be revi sed by introducing the stability ratio ( W), which is the ratio of rapid aggregation rate to the slow aggregation ratio. The aggreg ation rate can be expressed as 1 1)/( )/( 2 1k ikiki k kl kji ji j iij kNWkNNNWk dt dN [2-17] Thus, the Smoluchowski theory of shear aggregation [Hig82] is an applicable method to model the effect of pump-induced particle agglomeration in CMP slurries. Particle-Induced Defectivity Abrasive particles in the CMP slurry typica lly provide the mechanic al component of the polishing process. Some polishing models based on the interaction betwee n particles and wafer surfaces have been developed. Cook [Coo90] fi rst proposed the indentation model for glass polishing. In this model, the indentation depth or material removal is taken to be in proportion to the particle size and down pressure, but is inversely proportional to material properties (e.g., elastic modulus and hardness). Thus, an indentation depth ( Rs) as a function of particle size () is given by 3/224 3 KE P Rs [2-18] where P is the down pressure, K is the surface particle fill factor, a nd E is Youngs modulus. Thus, large particles or agglomerate are more lik ely to cause an increase in the defectivity of fragile materials. Mahajan et al. [Mah00 and Bas00] reported that th e presence of larger particles (>0.5 m) in the slurry not only changed the rem oval rate, but also cau sed surface defectivity (e.g., pits and micro-sc ratches) during CMP. As the particle size increases, the polishing

PAGE 34

34 mechanism changes from a contact area-based mechanism to an indentation-based mechanism. Thus, the presence of agglomerated particles (>0.5 m) in the slurry is the dominant factor contributing to the deterioration in the quality of the polished surface and the material removal rate during CMP. Typically, four different types of particle-induced defectivity (i.e., killer scratch, triangle scratch, shallow scratch, and embedded partic le) were found during Cu CMP [Teo04]. The killer scratches can cause significant damage or cu t through the copper interc onnects. The triangle scratches could be as deep as 2000 resulting in the formation of metal residual on a sequent metal layer, leading to an electrical short circ uit. Furthermore, shallow scratches and embedded particles may influence the dielectric reliability. In addition, the introduction of low k materials into the interlayer dielectrics is expected to ha ve more challenges in th e device integration during CMP of Cu/low-k dielectrics because the modulus and hardness of low-k materials decrease with decreasing the dielectric constant [Xu02]. Those ma terials with poor mechanical properties have a tendency to form more surface defects (e.g scratches, embedded particles, and film delamination) during CMP [Klo02 and Hij04]. Cha ndrasekaran et al. [Cha04b] investigated polishing conditions on surface defects in three different types of low k materials. The dielectric constants and mechanical properties (elastic modulus and hardness) of those materials are shown in Table 2-2. They indicate that black diam ond (BD) wafer, with a dielectric constant ( k) of 3, had smooth surface finishing after the polishing. However, porous methyl silsesquioxane (MSQ) wafer, with a k of 2.5 and 2.2, exhibited rough surface finishing and some surface defects (e.g., pits and microscratches) after polishing due to an exposed underl ying pore structure. As manufacturing nodes continue to scale down, less toleranc e for defect and particle densities on the wafer is expected to ensure de vice reliability. The International Technology

PAGE 35

35 Roadmap for Semiconductors (ITRS) has predicte d the required reduction of surface defects, such as killer defect density and critical particle diameter, in the near-term technology nodes, as shown in Table 2-3 [ITRS07]. Thus, robust CMP slurries may play an important role in minimizing process-induced agglomeration and particle-induced defec tivity during CMP of metals and dielectrics.

PAGE 36

36 p type substrate n-well region n+ n+ p+ p+ Cu/Low K CMP STI CMP W/PMD CMP Poly Opening/Metal Gate Polish p type substrate n-well region n+ n+ p+ p+ p type substrate n-well region n+ n+ p+ p+ Cu/Low K CMP STI CMP W/PMD CMP Poly Opening/Metal Gate Polish Figure 2-1. Role of CMP pr ocess in the manufacture of a microprocessor device.

PAGE 37

37 Dielectric Reduction Decreased Polarizability Decreased Density Less Polar Bonds: Si-F Si-C C-C C-H Constitutive Porosity: Self-organized free volume Subtractive Porosity: Selectivity Removed materials Dielectric Reduction Decreased Polarizability Decreased Density Less Polar Bonds: Si-F Si-C C-C C-H Constitutive Porosity: Self-organized free volume Subtractive Porosity: Selectivity Removed materials Figure 2-2. Possibilities for reducing the dielectric constant [Sha04]. Low K Materials Non-Silicon Silicon-Based Polymers: Amorphous carbon Silica-Based: SiOF SiOCH SSQ-Based: HSSQ MSSQ Low K Materials Non-Silicon Silicon-Based Polymers: Amorphous carbon Silica-Based: SiOF SiOCH SSQ-Based: HSSQ MSSQ Figure 2-3. The classification of low k materials [Sha04].

PAGE 38

38 Si3N4Dielectric Dielectric Resist Resist Dielectric Dielectric Resist Resist Via Dielectric Dielectric Trench Via (b) (c) (a) Dielectric Dielectric Cu Line CuVia(d) TaN Si3N4Dielectric Dielectric Resist Resist Dielectric Dielectric Resist Resist Via Dielectric Dielectric Trench Via (b) (c) (a) Dielectric Dielectric Cu Line CuVia(d) TaN Figure 2-4. Daul-Damesence Process Flow for ma king MLM structure: (a) resist patterning, (b)RIE and resist patterning, (c) RIE ach ieved trench, and (d) Cu deposition and CMP [Sze01].

PAGE 39

39 Barrier Materials Cu Bulk Dielectric 1stStage: Cu removal 2ndor 3rdStage: barrier and dielectric removal Barrier Materials Cu Bulk Dielectric Cu Bulk Dielectric 1stStage: Cu removal 2ndor 3rdStage: barrier and dielectric removal Figure 2-5. Three polishing st ages during Cu/Low k CMP.

PAGE 40

40 Table 2-1. Comparison between LOCOS and STI processes. Process LOCOS STI Advantages 1. Simple Process 2. Low Cost 1. Eliminated Birds Beak 2. High Device Packing Density (node 0.25 m) Disadvantages 1. Uncontrollable geometry 2. Birds Beak 3. Low Device Packing Density (node 0.25 m) 1. Defects (e.g. micro-scratches and dishing) 2. Increased Leakage Current

PAGE 41

41 Silicon Silicon Dioxide Nitride/ Pad Oxide Deposit Nitride/Oxide Stack Etch Isolation Trench SiO2Deposition Un-wanted SiO2Removal Nitride Removal Silicon Silicon Dioxide Nitride/ Pad Oxide Deposit Nitride/Oxide Stack Etch Isolation Trench SiO2Deposition Un-wanted SiO2Removal Nitride Removal Figure 2-6. STI process flow.

PAGE 42

42 DistancePotential Overlapping electrical double layers DistancePotential Overlapping electrical double layers Figure 2-7. Overlapping of tw o electrical double layers [Hun01].

PAGE 43

43 Table 2-2. Properties of di electric films [Cha04]. Types of DielectricsDielectric ConstantHardness (GPa)Elastic Modulus (GPa) Organosilicate glass (OSG)34.624.9 Silsesquioxane (SSQ)2.51.248.63Silsesquioxane (SSQ)2.20.684.98 Table 2-3. Interconnect surface requirements at near-term technology nodes [ITRS07]. Year of Production 2008 2010 2012 2014 Technology Node (nm) 59 45 36 28 # of Metal levels 12 12 12 13 Average dielectric constant 2. 7-3.1 2.4-2.8 2.2-2.6 2.2-2.6 Dishing Planarity (nm) 20 16 14 11 Min. defect particle size (nm) 28.5 22.5 17.5 14

PAGE 44

44 CHAPTER 3 CMP CHARACTERIZATIONS CMP System and Consumables A CMP system basically consists of CMP polishe r and slurry distributi on. In our studies, different types of slurries, incl uding silica slurries for Cu/ low k dielectric polishing and ceria based slurries for STI polishing, were utilized to polish 1-inch square BD1 low-k (k=3.0), JSR ultra low-k (k=2.2), copper, and tetraethyl or thosilicate (TEOS) wafers from a TegraPol-35 Table Top Polisher, Struers Co. The 1-inch s quare sample holder was made by a stainless cylinder of 2.25 inches in diam eter and 1.125 inches in height The following conditions were applied to each run: a down pressure of 3 psi, polishing time of 1 min, slurry flow rate of 100 ml/min, and rotation speed of 150 rpm. During the polishing process, CMP slurries with varying chemicals were poured down a soft polymeric polis hing pad, which provides a transport of slurry to achieve a local and global planarization. Furthermore, Figure 3-1 shows a slurry dist ribution system, which consisted of pumping devices and a slurry delivery l oop. A distribution loop consiste d of a 12-foot long PFA tubing (Teflon poly-fluoroalkoxy), slurry tank (10 gallons), pressure gauge flowmeter, pressurized air supply outlet/inlet, and 220 volt power supply system Pumping devices can be simply classified as positive displacement pumps (e.g., bellows or diaphragm pumps) and centrifugal pumps (e.g., a magnetically levitated centrifugal pump). Bo th pumping devices ha ve different operating principles, causing to different magnitudes of shear stress and degr ees of particle agglomeration during slurry delivery. For example, a positive di splacement pump can be considered a constant volume pump because the slurry is transported from the suction nozzle to the discharge nozzle. The volume of captured slurry depends on the ch amber size. However, this pumping device may generate highly localized shear stresses near the wall during the pump stroke, resulting in a

PAGE 45

45 significant increase in agglomerated particles. On the other hand, a magnetically levitated centrifugal pump, which is a constant pressure head device, provides a smooth pulseless flow due to a contact-free impeller in the pump housing, resulting in an insignificant increase in agglomerated particles. Thus, the magnetically levitated centrifugal pump can be considered a low shear pumping device. Particle Characterization As CMP slurries are a synergistic combination of particles and chemicals, the properties of abrasive particles can influe nce the slurry stability and po lishing performance (e.g., removal rate and roughness). In the pres ent studies, particle characteri zation involves the measurements of particle size, interparticle force, and su rface charge. These measurements can help in understanding the particle propert ies for the CMP application. Particle Size Measurement Numerous particle size instrument s can be utilized to measure particle size distribution in the slurry, as listed in Table 3-1 [Bow02]. These instruments are base d on several principles (e.g., microscopy, sedimentation, sievi ng, single particle counter, light scattering, and gas adsorption) and their limitation to determine particles in the size range of a few nanometers to thousand microns. As the abrasive particles utilized in the polishing are below 1 m, some particle size instruments, such as dynamic light scattering, laser diffracti on, photo-centrifuge, and a singleparticle optical sensor (SPOS) system, are employed for measuring the particle size and distribution. To detect the agglomerates or oversize pa rticles in CMP slurries, the particle size measurement system AccuSizer 780, consisting of a single-particle opti cal sensor (SPOS), was used to measure the oversize particle di stribution in the size range of 0.51 to 200 m [Pro98].

PAGE 46

46 This instrument has two stages of the autodi lution system to obtain an acceptable particle concentration for the SPOS sensor. The total extent dilution of initial sample suspension is equal to the dilution factor 1 (DF1) of first stage mu ltiplied by the dilution fact or 2 (DF2) of the second stage, defined as 21///211 2DFDFCCCCCCi i [3-1] where Ci, C1, and C2 are defined as initial particle con centration and first and second stage of particle concentrations. Thus, the advantage of this instrument is that it is capable of starting a concentrated particle concentra tion. In our experiment, 1 ml of circulated slurry was dropped into the solution chamber and diluted with deionized water. Particles in the dilute slurry flow through a sensor zone, a laser diode, to cause a detected pulse. The magn itude of pulse depends on the particle diameter and the physical method, light extinction and light scattering. Figure 3-3 shows the cumulative concentration tails of oversi ze particles in 60nm silica slurry (5 wt%) at 0 and 1,000 turnovers. The cumulative concentration ta il can clearly demonstr ate the stress effect on particle agglomeration during the handling process. Furthermore, the primary sizes for sub-micr on particles in CMP slurry are measured by dynamic light scattering, photo-cen trifuge, and laser diffraction t echniques [Bow02 and Sta02]. The principle of photo-centrifuge is a combinati on of sedimentation theory and light adsorption. The centrifugal sedimentation can be expressed by the Stokes law: the particle diameter ( d ) is dependent on the time ( t ) required to settle down a known distance. t xxIn d2 0 210/18 [3-2] where is the sample density, 0 is the dispersing liquid density, x1 is the starting radius, x2 is ending radius, and is the rotational angular velocity. During the measurement, small amounts

PAGE 47

47 of particles at a particular size pass through the light beam and cau se an intensity reduction. The particle concentration can be obt ained by the Beer-Lamberts Law. n i iii idNdKAII1 2 0)( loglog [3-3] where I0 is the maximum intensity of the initial light beam, Ii is the reduction intensity while detecting particles, A is optical coefficient of the cell, Ni is the number of particles, and K(di) is the efficiency of light extinction. For example, Figure 3-4 shows the particle size distribution of 80 nm silica slurry measured by Disc Centrifuge from CPS instruments. The measurable particle size range of Disc Centrifuge is 20nm to 30 m. Furthermore, a light scattering technique has been commonly employed for measuring particle size. This technique can be divided into two groups, dyna mic light scattering (or quasielastic light scattering or photon co rrelation spectroscopy) and low angle laser light scattering (or laser diffraction). Dynamic light scattering can quickly and accura tely measure the particle size range from 0.8 nm to 6.5 m (Nanotrac). When the particle size is smaller than a sub-micron, colloidal particles have a behavi or of Brownian motion due to thermal vibration. Thus, for a monodisperse distribution of spherical particles, an intensity of autocorrelation function is employed for analyzing the intensity fluctuations. )exp()(2tDq tg [3-4] where g(t) is the normalized autocorrection function, and D q and t are the Brownian diffusion coefficient (Stokes-Einstein rela tion), scattering vector, and delay time of the autocorrelation function. dTKDB 3/ [3-5]

PAGE 48

48 where KB is Boltzmann constant, T is the absolute temperature, is the liquid viscosity, and d is the particle diameter. Thus, smaller particles move more rapidly and have more intensity fluctuations than bigger partic les. Furthermore, the laser diffraction has been commonly employed for detecting large partic les. When a particle size is larger than incident wavelength, the diffraction phenomenon occurs while light b eam interacts with particles. The smaller particles have a higher angle of diffr action than bigger particles [Mor02]. Interparticle Force Measurement When particles approach each other closely, the overlapping of the repulsive electrical double layer forces have the capability to defend ag ainst the external mechanical forces applied on slurry itself. In this experi ment, the silica-silica and silica-ceria interparticle forces were measured using the wet cell arrangement in the Digital Instruments Nanoscope III atomic force microscope (AFM). Before the force measuremen t, colloidal probes and substrates must be prepared. A silica micro-spherical particle can be attached to the end of the cantilever using epoxy resins to simulate the interparticle forces The liquid used in the measurement will be the supernatant of the specific slurry used (e.g., silica slurry, low k slu rry etc.). The strength of the interaction will depend on the chemistry and add itives in the slurry and the nature of the particles. Such measurements will be conducted on respective slurries. It should be noted that the defectivity in CMP can be simulated with particle interactions with the respective substrate (e.g. copper, tungsten etc.). In this case, the plate for AFM measurem ent would be the same as the surface to be polished. The AFM force measurement was conducted by a colloidal probe technique, as shown in Figure 3-2 [But91 and But95]. The force-dist ance curve was the cantilever deflection ( zc) versus the height posit ion of the sample ( zp). During the measurement, the cantilever deflection

PAGE 49

49 was measured by a laser diode beam, reflecting from the back of the cantilever to a positionsensitive photodiode, and the sample position wa s controlled by a piezoelectric translator. The force measurement began with a large tip-surface separation in an aqueous electrolyte, as shown in Figure 3-2 (top). When the tip approaches the wafer surface, the interaction forces between the silica probe and substrate can bend the cantilever, resulting in th e cantilever deflections. When a colloidal probe moved closely at a certain distan ce, where the attractive Van der Waals force was larger than the repulsive electrical double laye r force, the tip rapidly jumped onto the sample surface. Once the tip was in contact with the surface, the cantilever de flection increased as zp decreases, pulling the cantilever upward. When the sample was retracted from the substrate, the silica probe may stick on the wafer surface due to the adhesion force. When the bending force was greater than the adhesion force, the sample ab ruptly jumped from the surface and returned to its starting position. The interaction force betw een silica probe and substrate was calculated by multiplying the spring constant of the cantile ver with the separation distance. Thus, the magnitude of the interparticle force can be conve rted from the interact ion force between silica probe and substrate because the inte rparticle force is half the valu e of the particle-plate force. Measurement of Electrical Po tential (Zeta Potential) The electrical double layer poten tial plays an important role in stabilizing the colloidal particles. However, there is no experimental approach to measure such potentials. The alternative method is to measure the zeta potential, which is the surface potential at the shear plane in an aqueous system, by means of the electrokinetic experiments. In this study, the zeta potential of abrasive particles in CMP slurries was measured by the Brookhaven ZetaPlus, which is utilized an electr ophoresis method to measure the electrophoretic velocity (E ) of colloidal particle s [Ste05 and Hun01]. When th e electrical double layer ( 1/k ) is

PAGE 50

50 much smaller than particle size, the Smoluchowski equation can be applied to measure the zeta potential (). 0EE [3-6] where and are the values of viscosity and diel ectric constant in the solution and E0 is the electric field. If the electrical double layer ( 1/k ) is much larger than the particle size, the Huckel equation can be applie d, and expressed as 02 3 EE [3-7] Polishing Characterization Polishing Parameter for Cu/Low K CMP The process of Cu/low CMP is a depe ndable technology to manufacture multilevel metallization structures. The introduction of low pol ar bonds or low density materials into silicon dioxide for reducing dielectric constant causes mech anical properties to deteriorate and presents many challenges (e.g., Cu/low-k delamination and scratching) in the de vice integration during CMP. In our experiment, 5 wt% 80nm silica slurries at pH 9 were used to polish low k wafers (BD1) in the TegraPol-35 Table Top Polisher at down pressure of 1.7, 3.4, and 5.0 psi to determine the best polish parameters in low k CMP. Figure 3-3(a) shows the material removal rate as a function of down pr essure. Although removal rate wa s proportional to down pressure, broken wafers were found at down pressure of 5 ps i due to the poor mechan ical strength of low k wafer. Thus, the down pressure for polishing was subjected to 3.4 psi. For metal CMP, chemical dissociation and pa ssivation on a metal surface can dramatically affect the polishing performance. According to the Pourbaix diagram, Cu starts to form a passivation layer above pH 6. In our experiment, 75n m silica slurries at pH 6, 6.5, and 7.0 were

PAGE 51

51 used to polish Cu wafers at a down pressure of 3.4 psi to investigate the effect of passivation layer on material removals. Figure 3-3(b) shows the material removal rate as a function of slurry pH. The Cu removal rates decrease d with slurry pH due to the gr owth of the passivation layers on Cu surface. Thus, acidic slurries with a corrosi on inhibitor were utilized to polish Cu wafers to get a higher removal rate. Although a corro sion inhibitor can protect the Cu layer from adramatic chemical dissociation below pH 6, it can decrease the Cu rem oval rate, as shown in Figure 3-3(c). Characterization of Polished Wafer Polished Cu, low k, and TEOS wafers were characterized by using an atomic force microscope (AFM), ellipsometry, and a scanning electron microsc ope (SEM), optical microscope, and four-point probe to measure th e material removal rate, surface roughness, and polished defectivity. AFM has been commonly employed for meas uring the root mean square, RMS = ( hi/N)1/2 (h and N are height values and numbers of measurements), and maximum height (Rmax) of polished substrates to determine the surface rough ness and defectivity. Typically, the principle of AFM is to measure the near-filed forces betwee n cantilever tip and substr ate, and its operational modes can simply divide into the contact and tapping modes [Len08]. In the case of contact mode, the AFM tip constantly contacts with subs trate during the measurement, and the cantilever deflection is measured by positive-sensitive photodiodes, which is utilized to detect laser beam reflections from the back of a cantilever, as illustrated in Figure 3-4. On the other hand, the tapping mode AFM measures the surface topography by means of the amplitu de of the cantilever oscillation. The decrease and increase in th e oscillation amplitude co rresponds to a protruding and a concaving surface, respectively, as shown in Figure 3-5. Furthermore, to obtain an

PAGE 52

52 accurately scanned position, the three dimensiona l scanning is controlled by X-Y-Z piezoelectric tubes. The surface images of polished defectivity can also be detected by a SEM, which has a resolution up to1-10nm, and magnification range from 10x to 106x [Len08]. The image resolution is determined by the beam brightness, which is dependent on the electron gun (i.e., field emission gun, LaB6 thermionic gun, and tungsten thermi onic gun). Typically, the secondary electrons and the backscattered electrons can be applied to the surface topographic images and compositional images, respectively. When a high en ergy electron beam interacts with the sample, the secondary electrons escape from the near-surf ace region in the depth range of 5 to 50 nm due to its weak electron energy (< 50 eV). Because the backscattered electrons have a greater electron energy (>50 eV), they can escape from a deeper depth range of 50 to 300nm [Eva92]. Thus, the secondary electron image is the best choice to observe the su rface defec tivity of polished wafers. The thickness of oxide films can be measured by ellipsometry, which is a nondestructive technique. The advantages of this instrument are that it requires no vacuum and has a large detected range (1-1000nm) [Eva92]. Typically, the principle of ellipsometry is to measure the change in the polarization of thin films after re flection or transmission. By measuring the change of the amplitude factor () and phase shift ( ), the complex ratio (), defined as a ratio of reflection coefficient rp (electric field parallel to the plane of incidence) to rs (electric field perpendicular to the plane of incidence), can be obtained. i s pe r r tan [3-8] In this study, the thickness of silica and lo w k wafers were measured by a Woollam EC110 Ellipsometer to obtain the material removal rate.

PAGE 53

53 In addition, four-point probe technique has been commonly used to measure the sheet resistance for determining the thickness of the Cu film due to its accurate and reliable measurement [Sze01]. During measurements, two outer probes were passed through applied current and inner probes were measured a voltage drop by voltmeter. By se parating the probes of supplied current and measured voltage using two pa irs of probes, this instrument can eliminate the measurement errors from probe and contact re sistances because large impedance of voltmeter caused to negligible parasitic resistances (probe and contact resistances) in inner probes [Doe08]. Thus, it accurately measured voltage difference of tested sheet. When the probe spaces were equal, the sheet resistance ( Rs) can be expressed: I V t Rs2ln [3-9] where is the resistivity, t is the thickness of thin film, V and I are measured voltage and applied current. By determining the sheet resistance, the thickness of the Cu film can be obtained. In addition, the geometrical error, arising from s light variation of probe spacing and closing to wafer edges, can be eliminated by its dual-configur ation technique.

PAGE 54

54 Drain Tank Liquid Pressure Gauge Filter Flowmeter PD Pump PD Pump System Maglev Centrifugal Pump System Power Supply Maglev Pump Dampener Gas Pressure Gauge Air Supply Replaceable DI Water PFA Tube Drain Tank Liquid Pressure Gauge Filter Flowmeter PD Pump PD Pump System Maglev Centrifugal Pump System Power Supply Maglev Pump Power Supply Maglev Pump Dampener Gas Pressure Gauge Air Supply Replaceable DI Water PFA Tube Figure 3-1. Schematic illustration of a slurry distribution system.

PAGE 55

55 Table 3-1. Various measurement methods for particle size [Bow02]. Method Medium Size Range ( m) Diameter Measured Microscopy Optical Electron Sieving Sedimentation Gravity Centrifuge Ultracentrifuge Single particle counters Electrical sensing Zone Time of flight Light scattering Diffraction (LALLS) Dynamic (PCS) Gas adsorption Liquid/Gas Vacuum Air Liquid Liquid Liquid Liquid Liquid Gas Liguid/Gas Liquid Gas/Vacuum 400.5 400.001 8000 5000 100.5 300.01 300.001 1200.3 700.2 1800.5(0.1) 0.5(1).002 5.005 Projected area Feret Sieve Stokes Volume Aerodynamic Volume Hydrodynamic Surface-Volume

PAGE 56

56 Position zpDeflection zc Distance DForce F Position zpDeflection zc Distance DForce F Figure 3-2. Schematic illustration of the coll oidal probe technique: ca ntilever shapes and cantilever deflection zc versus height position of sample zp (top); forcedistance curve (bottom) [But91].

PAGE 57

57 Low k polishing 5% 80nm Silica at pH 967.9 117.5 188.2 0 40 80 120 160 200 240 0123456Pressure (psi)RR (A/min) Broken Wafer Low k polishing 5% 80nm Silica at pH 967.9 117.5 188.2 0 40 80 120 160 200 240 0123456Pressure (psi)RR (A/min) Broken Wafer Cu Polishing0 2 4 6 8 5.566.577.5pHRR (nm/min ) 2% 75nm silica-pH 4.60 4 8 12 16 20 240 Mm BTA10mM BTA20mM BTARR (nm/min) Figure 3-3. Polishing parameters: (a) the material removal rate as a function of down pressure in low k CMP, (b) the material removal rate as a function of pH in Cu CMP, (c) the material removal rate as a function of BTA concentrations in Cu CMP. (c) (b) (a)

PAGE 58

58 positive-sensitive photodiodes positive-sensitive photodiodes Figure 3-4. Contact mode AFM: the cantileve r deflection is measured by positive-sensitive photodiodes, which is utilized to detect lase r beam reflections from the back of the cantilever [Len08]. Figure 3-5. Topping mode AFM: the surface topogr aphy is measured by detecting the amplitude of the cantilever oscillation [Len08].

PAGE 59

59 CHAPTER 4 EFFECTS OF STRESS-INDUCED PARTICLE AGGLOMERATION ON DEFECTIVITY DURING C MP OF LOW-K DIELECTRICS Introduction Chemical mechanical planar ization (CMP) is now commonl y employed for both the front and back end processing of IC devices due to its unique global planar ization capability [Ste96 and Sin02b]. With the miniaturization of semic onductor devices, multilevel interconnect plays an important role in providing a reliable and an efficient interconnect system for new generation of IC devices. The CMP process of Cu/low-k dielec trics is a dependable technology to manufacture multilevel metallization structures. This process is used for removing the excess copper, barrier layer, and underlying lowk dielectrics. As low-k films exhi bit poor mechanical properties, many remaining challenges ( e.g., Cu/low-k delamination and scratc hing) in the device integration during CMP of Cu/low-k dielectrics need to be solved [Cha04b, Bal04, and Xu02]. The presence of larger particles in CMP slurries is among the foremost factors that introduces defects ( e.g., scratches and embedded particle s) during the polishing process [Rem06]. Mahajan et al. [Mah00 and Bas00] studied the eff ect of particle size on the surface roughness and removal rate during CMP of oxide thin films. They reported that large particles (>0.5 m) in the slurry not only changed the removal ra te, but also deteriorated the quality of the oxide surface ( e.g., pits and micro-scratches) during CMP. These large particles in CMP slurries can increase mechanical stress that can lead to increase in surface defectivity during the polishing; especially for low-k dielectrics with low elastic modulus and hardness [Hij04]. These surface damages have a significant negative im pact on the manufacturing process and the reliability of dielectrics. Large particles can exist in CMP slurry, or be generated from th e slurry blending and slurry distribution system, which utilizes a pumpi ng device to circulate CMP slurry in the global

PAGE 60

60 distribution loop [Joh03]. Some studies have shown that positive displacement pumps ( e.g., bellows and diaphragm) generate high shear st ress and tend to agglomerate particles during slurry handling [Sin01, Sin04, and Lit04]. Litchy et al. [lit05] reported clogging of the filters in a slurry distribution system by agglomerated particle s, resulting in a pressure drop across the filter. Diaphragm and bellows pumps caused a larger pre ssure drop as compared to a magnetically levitated centrifugal pump as la rger amounts of agglomerated particles were produced during slurry handling. As some CMP slurries are shear-sensitive, whic h means that particles tend to agglomerate under the shear flow, low shear pump is in urgent need in slurry distribution system to prevent defect generation due to agglom erated particles during polishi ng. In the present study, we investigated the stress effects of three different types of pump on part icle agglomeration and particle-induced defectivity duri ng CMP of low-k dielectrics, and established a correlation between the roughness/defect density and the degree of agglomeration. Experimental A slurry distribution system was designed a nd built to observe the effects of stress on particle agglomeration. This system consisted of a 12-foot long tubing distribution loop (Teflon poly-fluoroalkoxy), a slurry tank, a pressure gauge, a flow mete r, and a pressurized air supply outlet/inlet. Positive displacement pumps ( e.g., bellows or diaphragm pumps) and magnetically levitated centrifugal pumps, provided by Levitronix LLC, were placed in the system to circulate the CMP slurry. The slurry used was Cu barrier slurry comprised of 10 wt% silica, designed to remove the residual Cu, the barrier metal, and a po rtion of the low-k dielectric. The flow rate of the silica slurry was held constant at 12 L/min by fixing the gas pr essure at 30 psi for the positive displacement pumps and 5900 rpm for the magnetica lly levitated centrifugal pump. The effects of pump-induced particle agglomeration were examined at 250, 500, and 1000 turnovers.

PAGE 61

61 The particle size measurement system Accu Sizer 780, consisting of a single-particle optical sensor (SPOS), was used to characterize the oversize particle dist ribution in circulated slurries and can detect particle in the size range of 0.51 to 200 m. To measure the particle size, 1 ml of circulated low-k slurry was dropped into the solution cham ber and diluted with deionized water to prevent slurry re-agglomeration. Single oversize particle was detected by the photozone, a narrow laser diode, and the cumulative oversize particle tail was obtained after measurements had been taken. Furthermore, the primary pa rticle size can be measured by dynamic light scattering technique. In this experiment, the as-received and circulat ed slurries were measured by Nanotrac, which can detect the particle size range from 0.8 nm to 6.5 m. Subsequently, silica slurries circulated by positive displacement and magnetically levitated centrifugal pumps were used to polish 1-inch square BD1 low-k (k = 3.0, Hardness/Elastic modulus = 4.6/24.9 GPa) and JSR ultra low k (k = 2.2, Hardness/Elas tic modulus = 0.68/4.98 GPa) wafers [Cha04] in a Struers RotoPol-31 Tabl e Top Polisher with a dow n pressure of 3 psi, polishing time of 1 min, slurry flow rate of 100 ml/min, and rotation speed of 150 rpm. The surface roughness of polished low-k wafers was characterized using a Digital Instruments Nanoscope III atomic force microscope (AFM). The defect density was determined by counting the number of defects per square millimeter by optical microscopy at 200 magnification. Results and Discussion Characterization of Oversize Particle Distribution Silica slurries (10 wt%) were circulated using positive displacement and magnetically levitated centrifugal pumps (Magle v pump) in the slurry distri bution system to observe the effects of stress on particle agglomeration. Figu re 4-1 shows the primary particle size of low-k slurries as a function of slurry turnovers. The results indicate that the primary particle sizes

PAGE 62

62 remained constant in all pumping devices and we re independent of numbers of slurry turnovers. Thus, the pumping devices would not change the pr imary particle size in highly stable slurries during the handling process. Figure 4-2 shows the cumulative distribution curves of oversize particles in low-k slurries at 0, 250, and 500 turnovers from different pumps obtained by cumulatively summing the number of particles above a certain size from 10 to 0.51 m. The cumulative concentration at 0.51 m was 15,377 (particles/ml) in as-received slurry (0 turnovers). The increase in the concentration of oversize particle s depends on the shear stress, wh ich arises from the different types of pumps used, and the number of turnover s. The cumulative concentrations of oversize particles for 250 and 500 turnovers at 0.51 m were 44,613 and 66,916 (particles/ml) in the bellows pump system, 41,435 and 61,481 (particles /ml) in the diaphragm pump system, and 16,485 and 18,809 (particles/ml) in the magnetica lly levitated centrifuga l pump system. Based on experimental results, the cumulative concen trations of oversize particles were shown to significantly increase with slurry turnovers in the positive displacement pump system. In contrast, increasing the number of turnovers did not increase the concentration of oversize particles significantly in the case of magne tically levitated centrifugal pump system. The normalized oversize particle distribution, which is defined as the ratio of cumulative concentrations at 500 slurry tur novers to cumulative concentrati ons at 0 slurry turnovers, can clearly determine the effect of stress on particle agglomeration, as shown in Figure 4-3. In the positive displacement pump system, the normalized distributions exhibited greater and wider distributions as compared to the magnetically levitated centrifugal pump system. Furthermore, the mean value of normalized oversize concentrat ions in the positive displacement pump system was 6 times higher than that of a magnetically levitated centrifugal pump system. The mean

PAGE 63

63 value of normalized number of ove rsize concentrations was calculated in the particle size range of 0.51 to 1 m because most oversize particles were generated in this region during slurry delivery. The magnetically levitated centrifugal pump did not increase the normalized oversize particle distribution signifi cantly at 500 turnovers. The phenomena of stress-induced particle agglomeration can be explained by the Smoluchowski theory [Smo17, Hig82, and Ste05] in which a model is proposed that considers the shear flow and the electrosta tic interaction between particles. This model assumes that the particle collisions are binary a nd proportional to the particle con centration. The aggregation rate of k-fold aggregates, dNk/dt is given by the time evolution of the cluster size aggregates, i and j fold, 1 1)/( )/( 2 1k ikiki k kl kji ji j iij kNWkNNNWk dt dN [4-1] )( 3 4ji ijaaGk [4-2] where the aggregation constant, kij, is a function of the shear rate ( G ) and particle size ( a). The stability ratio ( W ) is the ratio of the rapid aggregation rate in the absence of electrostatic interaction to the slow aggregation rate when there are electrostatic interactions between particles. According to this model, during slurry delivery, the shear flow causes particles to approach each other. When the particles are suffi ciently close to each other that the attractive Van der Waals force is greater than the repulsive interparticle force, particle agglomeration occurs. The degree of particle agglomeration that occurs is determined by the slurry properties ( e.g., interparticle forces), external shear stress ( i.e. type of pump), and the number of turnovers of the slurry. In our experiment, the normalized oversize particle distribu tion was measured from the same slurry at fixed slurry turnovers (500 turnovers). Thus, the magnitude of shear stress induced by pumps can be distinguished by the degree of particle agglomeration. Lower

PAGE 64

64 agglomeration of particles during slurry handl ing by a magnetically levitated centrifugal pump was because of its contact-free impeller in th e pump housing which provided a smooth pulseless flow, whereas the positive displacement pump ge nerated highly localized shear stresses near the wall during the pump stroke, resultin g in a significant increase in oversize particles. Thus, the magnetically levitated centrifugal pu mp is a low shear pumping device. Characterization of Low-k CMP The effects of oversize particles present in the slurry on the defectivity during CMP of low-k dielectrics were investigated by analyzing various operating conditions ( e.g., turnovers and pump types) and corresponding output parameters ( e.g., RMS roughness and defect density) to investigate the correlation be tween the roughness/defect density and the degree of particle agglomeration. First, low-k and ultra low-k wafers polished by circul ated slurries at 1000 turnovers were characterized by optical microsco py at 200 magnification, as shown in Figure 44. As ultra low-k wafers exhibited poor mechani cal properties (i.e., elastic modulus and hardness) as compared to BD1 wafers, more surface defect s were found on polished ultra low-k wafers. In addition, the defect density, determined as th e number of defects per square millimeter, was calculated from those images. Figure 4-5 shows th e defect densities of BD1 and ultra low k wafers as a function of slurry turnovers. Insigni ficant increase in defect density with slurry turnover was observed in the case of maglev pump processed slurri es due to less agglomerated particles. Consequently, the defect density or surface defects increased with increasing agglomerated particles caused by pumping device s and decreasing with mechanical properties of thin films. Furthermore, the example of the surface roughness (RMS and Rmax characterized by contact mode AFM) of polished BD1 wafers as a f unction of turnovers is shown in Figure 4-6. In the positive displacement pump system ( i.e. bellows and diaphragm pumps), the surface

PAGE 65

65 roughness increased with slurry turnover, but the increase was not significant in the magnetically levitated centrifugal pump system. As a result, a positive correlation was observed between the roughness/defect densities and mean value of nor malized oversize concentr ations, as shown in Figure 4-7. The normalized oversize concentration in this figure was defined as the mean value of normalized oversize concentrations in the particle size range of 0.51 to 1 m because a large number of oversize particles in this range can cause most surface defect ivity during CMP of lowk dielectrics. In both graphs, poi nts closer to and further from the y-axis correspond to slurries circulated by the magnetically levitated cen trifugal pump and positive displacement pump, respectively. The maximum defect density was 4.4 (numbers/mm2) on low-k wafers polished by circulated slurries from the positive disp lacement pump system, and 1.1 (numbers/mm2) in the case of the magnetically levitate d centrifugal pump system. Cons equently, positive displacement pump system caused more particle agglomeration with slurry turnovers, resulting in significant increases in the surface roughness an d defect density as compared to the magnetically levitated centrifugal pump system. Summary We examined the effects of stress on particle agglomeration occurring in silica slurries circulated by both positive displacement a nd magnetically levitate d centrifugal pumps. Our results indicate that the magnetically levitated centrifugal pump had less effect of stress on particle agglomeration and did not increase the concentration of oversize particles significantly with slurry turnovers. The mean value of normalized oversize concentrations in the positive displacement pump system was 6 times higher than that of a magnetically levitated centrifugal pump system. The roughness/defect density and the degree of a gglomeration were found to be positively correlated. In addition, more surf ace defects were found on polished ultra low-k

PAGE 66

66 wafers because ultra low-k wafers exhibited po or mechanical propertie s (i.e., elastic modulus and hardness) as compared to BD1 wafers. Cons equently, the magnetically levitated centrifugal pump was a low shear device and caused less proc ess-dependent defectivity (e.g., scratches and embedded particles) during CMP of low k dielectrics.

PAGE 67

67 Maglev Pump0 0.05 0.1 0.15 0.2 0.25 02505001000No. of TurnoversParticle Size (m) Bellows Pump0 0.05 0.1 0.15 0.2 0.25 02505001000 No. of TurnoversParticle Size (m) Diaphragm Pump0 0.05 0.1 0.15 0.2 0.25 02505001000 No. of TturnoversParticle Size (m) Figure 4-1. Primary particle size as a function of slurry turnovers: low-k slurries circulatedby (a) maglev, (b) bellows, and (c) diaphragm pumps. (a) (b) (c)

PAGE 68

68 11 0 101102103104105 Cum. Conc. (#/ ml)Particle Size (m) 0 turnovers 250 turnovers 500 turnovers(a) 11 0 101102103104105 (b) Cum. Conc. (#/ ml)Particle Size (m) 0 turnovers 250 turnovers 500 turnovers 11 0 101102103104105 (c) Cum. Conc. (#/ ml)Particle Size (m) 0 turnovers 250 turnovers 500 turnovers Figure 4-2. Cumulative concentration vs. particle size at 0, 250, and 500 turnovers for (a) Bellows, (b) diaphragm, and (c) magnetically levitated centrifugal pump systems. The magnetically levitated centrifugal pump cause d less particle agglomeration as the tail portion did not change signifi cantly with increase in the number of slurry turnovers.

PAGE 69

69 11 0 0 4 8 12 16 Bellows Pump Diaphragm Pump Magnetically Levitated Centrifugal PumpC500/C0 Turnovers (Conc./ Conc.)Particle Size (m) Figure 4-3. Normalized oversize particle distribution for positiv e displacement and magnetically levitated centrifugal pumps.

PAGE 70

70 Figure 4-4. Optical images: (a) BD1 and (c) ultra low k wafers polished by maglev pump processed slurries, and (b) BD1 and (d) ultra low k wafers polished by positive displacement pump processed slurries. (a) (b) (c) (d)

PAGE 71

71 0 0.5 1 1.5 2 2.5 3 3.5 4 2505001000TurnoversScratch density (#/sq. mm ) Maglev pump Positive displacement pump 0 1 2 3 4 5 2505001000TurnoversScratch density (#/sq. mm ) Maglev pump Positive displacement pump Figure 4-5. Scratch density as a function of turnovers: (a) BD1 wa fers and (b) ultra low k wafers. (a) (b)

PAGE 72

72 Figure 4-6. Comparison of surface roughness (a) RMS and (b) Rmax of low-k wafers polished by circulated slurries from diaphragm, bello ws, and magnetically le vitated centrifugal pumps. 0 0.5 1 1.5 2250 turnovers1000 turnoversRmax (nm) Magnetically Levitated Centrifugal Pump Diaphragm Pump Bellows Pump(b) 0 0.05 0.1 0.15 0.2 0.25250 turnovers1000 turnoversRMS (nm) Magnetically Levitated Centrifugal Pump Diaphragm Pump Bellows Pump(a)

PAGE 73

73 Figure 4-7. Defectivity vs. normalized oversize particles. (a) Scratch density vs. normalized oversize particles. (b) RMS roughness vs. normalized oversize particles. 0 1 2 3 4 5 6 7 12345678910Normalized Oversize ParticlesScratch Density (#/sq. mm) Magnetically Levitated Centrifugal Pump Positive Displacement Pump 0 1 2 3 4 5 6 7 12345678910Normalized Oversize ParticlesScratch Density (#/sq. mm) Magnetically Levitated Centrifugal Pump Positive Displacement Pump(b) 0.00 0.05 0.10 0.15 0.20 0.25 0.30 12345678910Normalized Oversize ParticlesRMS Roughness (nm) Positive Displacement Pump Magnetically Levitated Centrifugal Pump 0.00 0.05 0.10 0.15 0.20 0.25 0.30 12345678910Normalized Oversize ParticlesRMS Roughness (nm) Positive Displacement Pump Magnetically Levitated Centrifugal Pump(a)

PAGE 74

74 CHAPTER 5 ROLE OF INTERPARTICLE FORCES DURI NG P UMP-INDUCED AGGLOMERATION OF CMP SLURRIES Introduction With the miniaturization of semiconductor de vices, chemical mechanical polishing (CMP) is commonly employed for both the front and back end processing of such devices for local and global planarization. The CMP slurry is a syne rgistic combination of abrasive particles and chemicals, and is perhaps the most critical consumable in the semiconductor industry as it controls the uniqueness and tech nical performance of the CMP process [Sin02a and Ste96]. Thus, highly stable particles in chemical mechan ical polishing (CMP) slu rry play an important role to ensure a superior polishing performance and minimize the process-dependent defectivity. Stable suspensions in CMP slurry are achieved by overlapping the similar surface charges from the dissociation of the metal oxide groups (M+OH). The degree of slurry stability depends on intrinsic properties of particles and slurry pH. However, abrasive particles ( e.g., ceria, silica, and alumina) along with numerous chemicals (e .g., oxidizers, salts, and complexing agents) can deteriorate the stability of ab rasive particles and cause surf ace defectivity during CMP [Sin02b and Bie99]. Typically, the introduction of salt in CMP slurries can increase the polishing rate due to screening the surface charges be tween abrasive particles and subs trate, leading to an increase in friction force during the polishing [Cho04b]. Howe ver, slurries containing high ionic strengths can also diminish the repulsive forces between abrasive particles, resulting in a significant increase in agglomerated particles. Basim et al. [Bas02] investigated the effect of salt concentration on particle agglomeration and surface quality in silica polishing. They indicate that the abrasive particles tended to agglomerate each other at high salt concentration (NaCl > 0.2 M). These soft agglomerates deteriorated the quality of polished surface during CMP.

PAGE 75

75 To minimize the polishing defectivity, the par ticle size distribution, especially the tail region representing oversized particles, should be as small as possible and should not increase with time during slurry delivery. Our recent study indicates that some pu mping devices in the slurry distribution system may generate high sh ear stress and cause to a significant increase in agglomerated particles during slurry hand ling [Cha08]. The phenomenon of pump-induced particle agglomeration can be explained by the Smoluchowski theory of shear aggregation, as listed in Equation 2-17. The degree of particle agglomeration that occurs is determined by the slurry properties ( e.g., interparticle forces) and the extern al shear stress applied on the slurry itself during the slurry handling process. In our study, formulated slurri es were recirculated in the slurry distribution system to examine the slurry stability. The agglomeration of the slurries was found to depend both on the external shear stress and the interparticle forces acting on the slurries. The interparticle forces were determined by means of the interaction forces between silica probe and silica wafer, measured by the colloidal probe technique in the supernatant sl urries with varying chemicals. Experimental We circulated 30 nm silica slurries (5 wt %) at various pH values (2, 7, and 11), with and without 0.1 M KCl in the slurry distribution sy stem for 1000 turnovers to observe the stress effect on particle agglomeration. The slurry distribution system consisted of a positive displacement pump (e.g., bellows or diaphragm pumps), a 12-foot long tubing distribution loop (Teflon poly-fluoroalkoxy), a slurry tank, a pressure gauge, a flow meter, and a pressurized air supply outlet/inlet. The flow rate of the slurry was held constant at 12 L/min by fixing the gas pressure at 30 psi. Agglomerated particles in circulated slurries were characterized us ing the particle size measurement system AccuSizer 780, consisting of a single-particle opt ical sensor (SPOS).

PAGE 76

76 This system can detect part icle sizes from 0.51 to 200 m. To measure the particle sizes, 1 ml of circulated silica slurry was dropped into the so lution chamber and diluted with deionized water to prevent slurry re-agg lomeration. The single ov ersize particle was de tected by the photozone, a narrow laser diode, and the cumulative oversize particle tail was obtained after measurements had been taken. Furthermore, the surface potenti al of silica particles was characterized by zeta potential (Brookhaven ZetaPlus), me asuring the velocity of charge d particles at shear plane in the solution. The interparticle force measurement can be achieved by means of the colloidal probe technique using the atomic force microscope (AFM). Ducker et al. [Duc91 and Duc92] and Butt et al. [But91 and Kap02] first used co lloidal probe technique to m easure directly interaction forces between colloidal probe and samp le surface. In this experiment, a 5 m spherical silica particle from Bangs Laboratories Inc. was atta ched to the end of cantilever using epoxy resins and the micromanipulator under the control of an optical microscope. Figure 5-1 shows the scanning electron microscope (SEM) image of 5 m spherical silica particle glued on the top of the AFM tip. This experiment was performed with the wet cell arrangement by AFM force measurement in the supernatant slur ries at various pH values (2, 7, and 11), with and without 0.1 M KCl. The surface forces were measured by th e cantilever deflection as a function of the distance between silica probe and plate surface. The plate used was silica wafer deposited by a plasma-enhanced chemical vapor deposition (PEC VD) technique. The definition of zeros of both force and tip-substrate separation is necessary to analyze deflection data. The zero of force was chosen where the deflection was constant, and th e zero of tip-substrate se paration was defined as the region of constant compliance. The surf ace force was calculated by multiplying the tip deflection with the spring constant (0.58 N/m) of the cantilever.

PAGE 77

77 Subsequently, circulated silica slurries (30nm) were used to polish 1-inch square BD1 lowkwafers in a Struers RotoPol-31 Table Top Polisher with a down pressure of 3 psi, polishing time of 1 min, slurry flow rate of 100 ml/min, and rotation speed of 150 rpm. The surface roughness of polished low-k wafers was characteri zed using a Digital Instruments Nanoscope III atomic force microscope (AFM). Results and Discussion To investigate the stability of formulated slurri es, 30nm silica slurries (5 wt%) at pH 2, 7, and 11, with and without 0.1 M KCl, were circulated in the slurry distribution system for 1000 turnovers. Figure 5-2 shows the cumulative distri bution curves of oversize particles in silica slurries obtained by cumulatively summing the numb er of particles above a certain size from 10 to 0.51 m. The cumulative concentrations of oversize particles increased ra pidly as pH values decreased, as shown in Figure 5-2(a). The particle concentrations at 0.51 m were 95,717, 59,577, and 45,373 (particles/ml) at pH 2, 7, a nd 11, respectively. The introduction of salt in silica slurries further accelerate d slurry agglomeration at all pH values, as shown in Figure 52(b). The particle co ncentrations at 0.51 m were 118,457, 69,811, and 54,150 (particles/ml) at pH 2, 7, and 11 with 0.1 M KCl, respectively. In addition, the mean value of normalized oversize particle distributions, defined as the ratio of cumulative concen trations at 1000 slurry turnovers to 0 slurry turnovers in the pa rticle size range of 0.51 to 1 m, can clearly determine the effect of pump-induced particle agglomeratio n, as shown in Figure 5-2(c). We found the mean values of normalized oversize concentrations at pH 2, 7, a nd 11 to be 4.5, 2.6, and 2.1 times higher than those of as-received slurries. In the case of salt addition, the m ean values of normalized oversize concentrations at pH 2, 7, and 11 with 0.1 M KCl were 6.6, 3.9, and 3.3 times higher. These

PAGE 78

78 results indicate that silica slurri es at acidic pH, and with salt addi tion, were unstable, resulting in a significant increase in agglomerated particles during the handling process. The agglomeration of the slur ries was found to depend both on the external shear stress acting on the slurries and the stab ility of colloidal particles, e xpressed from Equation 2-17. The stability ratio ( W ) is a criterion for slurry stability and dependent on the total electrostatic interaction ( Vtotal) at a particle se paration distance ( d), can be expressed as du u KTV Wtotal 0 22 /exp 2 [5-1] where u is equal to d/a for same particle size ( a), K is the Boltzmanns constant, and T is temperature [Eli98]. Based on De rjaguin-Landau-VerweyOverbeek (DLVO) theory, the total electrostatic interaction, a sum of repulsive electri cal double layer potential ( Vr) and attractive Van der Waals potential ( Va), provides a potential barrier to prevent particle agglomeration [Hun01]. However, some pumping devices may generate high shear st ress that causes particles to approach each other. When the particles are suffi ciently close to each other that the attractive Van der Waals force (Va/d) is greater than the repulsive interparticle force (Vr/d), particle agglomeration occurs. Thus, the repulsive interparticle forces play an important role to stabilize the colloidal particles. The streng th of repulsive interparticle fo rces can be controlled by varying chemicals (e.g., pH and salt) and colloidal particles ( e.g., ceria and silica). Based on Derjaguin approximation, the force of sphere-sphere ( Fs-s) is half the value for sphere-plate interaction forces ( Fs-p) [Isr91]. Thus, the interparticle force is determined by means of the interaction forces between a silica probe and a silica wafer measured by the colloidal probe technique in the supe rnatant slurries. The normalized interaction force ( F/R ) as a function of separation distance between the silica probe a nd the silica wafer in s upernatant slurries at various pH values, with and wit hout salt addition (0.1 M KCl), was shown in Figure 5-3. When

PAGE 79

79 the silica probe was far away from the silica wa fer (separation distance > 25 nm), no interaction force was observed from all supernatant slurri es. As the separation distance consecutively decreased, the cantilever bent upwar d due to repulsive surface forces, generated by overlapping the electrical double layers betwee n the silica probe and the silica substrate. Afterwards, when the silica probe approached to the silica wafer at a certain interactive di stance, the silica probe suddenly jumped onto the substrate (jump-in) becau se the gradient of at tractive Van der Waals force exceeded the gradient of repulsive surface fo rce plus the spring constant of the cantilever, and the normalized repulsive forces became vertical at this contact region, defined as zero of separation distance. As the isoelectric point (IEP) of sili ca is around pH 2, the increase of the slurry pH raised a repulsive surface fo rce due to increasing negative charges ( SiO) on silica surface [Par65]. The maximum normalized repulsive in teraction forces in s upernatant slurries at pH 2, 7, and 11 were 0.66, 1.0, and 1.82 (mN/m), respectively. However, smaller repulsive interaction forces were observed from all supern atant slurries with adding 0.1 M KCl because of the screening effect. The maximum normalized repulsive interaction forces at pH 2, 7, and 11 with 0.1 M KCl were 0.33, 0.79, and 1.49 (mN/ m), respectively. Consequently, alkaline silica slurries without salt addition exhibited larger re pulsive interaction forces between silica surfaces. In addition, the screening effect caused by the salt addition in silica slurries can also be proved by zeta potential measurement, as show n in Figure 5-4. The zeta potentials of silica slurries at pH 2, 7, and 11 were -3.5, -33, and -39.6 (mV), respectively. In the case of salt addition, smaller surface potentials were observed. The zeta potentials of silica slurries at pH 2, 7, and 11 with 0.1 M KCl were -0.45, -27.8, and -34.5 (mV), respectively. Based on experimental results, excellent co rrelations were established between the repulsive interaction forces and the mean values of normalized oversize par ticles as a function of

PAGE 80

80 pH, with and without 0.1 M KCl, as shown in Fi gure 5-5. The increase in slurry pH raised a repulsive interaction force and co rresponded to a considerable decrea se in agglomerated particles during the handling process. Sili ca slurries at pH 11 without sa lt addition exhibited the largest repulsive interaction forces and the smallest mean values of oversize par ticles. The addition of salt in silica slurries further reduced the re pulsive interaction forces and caused more agglomerated particles during slurry handling. These results can prove that interparticle forces in colloidal suspension play an important role to prevent process-induced particle agglomeration. In addition, agglomerated partic les in slurries may increase th e mechanical stress that leads to increase surface defectivity during the low k po lishing. Circulated silica slurries (5wt% 30nm) at varying pH and salt addition were used to polish low k wafers. Figure 5-6 show the surface roughness (i.e., RMS and Rmax) as a function of slurry pH and the AFM images of polished low k wafers at pH 2, 7, and 11. Our results indica te that larger surface roughness and more surface defectivity (e.g., micro-scratches) were found in low k wafers polished by circulated silica slurries at acidic and neutral pH, with and without salt addition because lo wer interparticle forces between silica particles caused mo re agglomerated particles during the handling process. The maximum surface roughness (RMS = 1.47nm, Rmax = 35.96nm) was found at silica slurry at pH 2 with 0.1 M KCl; whereas less surface def ectivity and minimum surface roughness (RMS = 0.79nm, Rmax = 13.01nm) was observed at silica slurry at pH 11. Summary Our results indicate that silica slurries at acidic pH, and with salt addition, tended to agglomerate more oversize particles during the handling proce ss. The agglomeration of the slurries was dependent both on the external shea r stress acting on the slurries and interparticle forces in a colloidal suspension. The interpar ticle forces were determined by means of the interaction forces between a silica probe and a silica wafer measured by the colloidal probe

PAGE 81

81 technique in the supernatant slurries with varyin g chemicals. Larger repulsive interaction forces were found in alkaline silica slurries. However, the addition of sa lt in silica slurries reduced the strength of repulsive interaction forces because of the screening effect. As a result, excellent correlations were established between the repulsiv e interaction forces and the mean values of normalized oversize particles. The increase in slurry pH raised a repulsive interparticle force to defend against shear flow, resulti ng in less agglomerated particle s during slurry handling. Least particle-induced defectivity wa s observed on polished low k wafe rs, utilized highly alkali slurries. Thus, alkaline silica slurri es would be the best condition for robust slurries to minimize the process-induced particle agglomeration a nd surface defectivity during CMP of metals and dielectrics.

PAGE 82

82 Figure 5-1. SEM image of 5 m silica probe.

PAGE 83

83 11 0 0.0 3.0x1046.0x1049.0x1041.2x1051.5x105 (a) pH 2 pH 7 pH 11 As-received slury Cum. Conc. (#/ml)Particle Size (m) 11 0 0.0 3.0x1046.0x1049.0x1041.2x1051.5x105 (b) pH 2+0.1M KCl pH 7+0.1M KCl pH 11+0.1M KCl As-received slurry with 0.1M KClCum. Conc. (#/ml)Particle Size (m) pH 2pH 7pH 11 0 2 4 6 8 10 C1000/C0 Turnovers (Conc./Conc.) 0 M KCl 0.1 M KCl(c) Figure 5-2. Circulated silica sl urries for 1000 turnovers at pH 2, 7, and 11: (a) cumulative concentration without salt, (b ) cumulative concentration w ith 0.1 M KCl, and (c) the mean value of normalized oversize particles in the particle size range of 0.51 to 1 m.

PAGE 84

84 051015202530 0 1 2 3 4 5 Jump-In Maximum Repulsive Force F/R(mN/m)Tip-Substrate Distance (nm) pH 2 pH 2 + 0.1MKCl pH 7 pH 7 + 0.1MKCl pH 11 pH 11+0.1MKCl Figure 5-3. Force versus distance between a silica substrate and silica probe in supernatant slurries at pH 2, 7, and 11, with and without 0.1 M KCl.

PAGE 85

85 -50 -40 -30 -20 -10 0 10 20 02468101214pHZeta Potential (mV ) 30nm Silica 30nm Silica-0.1 M KCl Figure 5-4. Zeta potential of 30 nm silica slurries as a function of pH, with and without 0.1 M KCl.

PAGE 86

86 0 1 2 3 4 5 6 7pH 2pH 7pH 11Mean values of C1000/C0 (Conc./Conc.)0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2F/R(mN/m) 0 M KCl-Mean values 0.1 M KCl-Mean values 0 M KCl-F/R 0.1 M KCl-F/R Figure 5-5. Correlations between repulsive intera ction forces and mean values of normalized oversize particles as a function of pH, with and without 0.1 M KCl.

PAGE 87

87 Low K CMP-RMS0 0.5 1 1.5 2pH 2pH 7pH 11RMS (nm) No salt 0.1M KCl Low K CMP-Rmax0 10 20 30 40 50pH 2pH 7pH 11Rmax (nm) No salt 0.1M KCl Defectivity Defectivity Figure 5-6. Surface roughness as a function of pH : (a) RMS and (b) Rmax, and AFM images of low k wafers polished by circulated silica sl urries at (c) pH 2, (d) pH 2 with 0.1M KCl, (e) pH 7, (f) pH 7 with 0.1M KCl, (g) pH 11, and (h) pH 11 with 0.1M KCl. (a) (b) (e) (c) (d) (g) (f) (h)

PAGE 88

88 CHAPTER 6 ROLE OF SLURRY CHEMISTRY ON STRESS-INDUCED AGGLOMERATION Introduction The presence of colloidal dispersions is ther modynamically disfavored due to large surface area of colloidal particles and particle collisions (Brownian motion). If a lack of potential barriers between colloidal dispersions, particles have a tendency to agglomerate during the storage and the slurry blending. Thus, electros tatic and polymeric stab ilization are commonly employed for stabilizing abrasive particles in CMP slurries. Palla et al. [Pal 99 and Pal02] studied the effect of the surfactant addition on the stability of alumina slurry with oxidizer agent. Typica lly, oxidizer can form a passive layer on tungsten surface to prevent underlying tungsten substrate fr om corrosion. However, with adding 0.1 M potassium ferricyanide to alumina slurry, partic les were agglomerated and settled because of numerous counter-ions shielding surface charges on the particles. On the other hand, the addition of a mixture surfactant (nonionic an d anionic surfactants) into the sl urry can efficiently stabilize alumina particles because anionic surfactants were attracted by charge surface of alumina particles and nonionic surfactants penetrated in to anionic surfactant la yers due to hydrocarbon chain interactions. The polishing results in tungst en CMP indicate that the addition of a mixture surfactant in alumina slurry can reduce surf ace roughness and particle residues during W CMP [Bie99]. In addition, Hackley [Hac97] demonstrat ed the role of polymer dispersant in the dispersion of silicon nitride particles. The polyacrylic acid (PAA) formed a protective film onto the surfaces of silicon nitride particles, resulting in a wide pH range of the colloidal stabilization due to shifting the isoelectric points of silicon nitr ide particles induced by the ionization of PAA. Thus, the addition of polymer su rfactants in the absence or pr esence charge effects can form protective barriers to stab ilize colloidal suspensions.

PAGE 89

89 In our study, formulated and un-formulated ceria and silica slurries were subjected to high shear forces and then measured the particle agglomeration characte ristic of their tail distribution to investigate the eff ect of surfactants on the slurry stability. The slurry stability was investigated from the polishing performance. Experimental During the slurry handling, some pumping de vices in a slurry distribution system may generate high shear flow and cause to agglomer ate particles in the tail distribution of oversize particles. In this study, as-received and formulated slurries were used to circulate in the slurry distribution system, utilized by high shear pumping device (e.g., positive displacement pump), to examine the slurry stability. The slurries used for shear testing incl uded different types of particles (e.g., ceria and silica) and surfactants (e.g., anionic, nonionic, an d cationic surfactants). The primary particle size and size distri bution were measured by the dynamic light scattering technique, Nanotrac from Microtrac Inc., can quickly and accurately measure the particle size in the ra nge of 0.8 nm to 6.5 m. In addition, the tail distribution of oversize particles in circulated slurries can be characterized using the particle size measurement system AccuSizer 780, consisting of a si ngle-particle optical sensor ( SPOS). This system can detect particle sizes from 0.51 to 200 m. To measure the particle sizes, 1 ml of circulated silica slurry was dropped into the solution chamber and diluted with deionized water to prevent slurry reagglomeration. The single oversize particle wa s detected by the photozone, a narrow laser diode, and the cumulative oversize particle tail was obtained after measurements had been taken. The surface potentials of abrasive particles we re characterized by zeta potential, ZetaPlus from Brookhaven Instruments Corp oration, utilizing an electr ophoresis method to measure the

PAGE 90

90 electrophoretic velocity of colloidal particles. In this ex periment, the Smoluchowski model was applied to measure the zeta potentials of CMP slurries. Results and Discussion Typically, to obtain high removal rates of dielectric substrates the ceria slurry was taken near its isoelectric point (IEP) ( i.e. pH 6~7), and silica slurry was at alkaline solutions (i.e., pH 9~11). Thus, un-formulated ceria slurry at pH 6 and silica slurry at pH 9 were circulated in the slurry distribution system for 1000 turnovers to determine the stabili ty of ceria and silica slurries. Figure 6-1 shows the normalized distribution curves of oversize particles, de fined as the ratio of cumulative particle concentrations at 1000 turnovers to 0 turnovers. A significant increase in oversize particle distribution was found in ceria slurry as compared to silica slurry. This phenomenon can be demonstrated by measuring the surface potential. Figure 6-2 shows the zeta potentials of un-formulated ceria and silica slurries as a function of slurry pH. The zeta potentials of ceria slurry at pH 6 and silica slurry at pH 9 were 0.41 (mV) and -46.4 (mV), respectively. A weakly repulsive electrostatic interaction betwee n ceria particles caused to a significant increase in particle agglomeration under the external shear stress applied on slurry itself. To improve the stability of CMP slurries, va rying surfactants (e.g., anionic and cationic surfactants) were introduced into ceria and silica slurries for pr oviding the polymer stabilization. Figure 6-3 shows the zeta potential s of formulated ceria slurri es as a function of pH. The addition of anionic ionic surfact ant (sodium dodecyl sulfate) and polymer dispersant (PAA) to ceria slurry can shift the IEP of ceria to more aci dic pH due to lots of negative charges adsorbed onto ceria surface. On the other hand, the addi tion of cationic surfactants, decyltrimethylammonium bromide (C10TAB) and tetramethylammonium hydr oxide (TMAH), can shift the IEP to more alkaline pH due to the positive charges of hydrophilic heads. A comparison of normalized tail distributions betw een the circulation of formulat ed (cationic surfactant) and un-

PAGE 91

91 formulated ceria slurries at pH 6 is shown in Figure 6-4. The formulated ceria slurry can significantly minimize the magnitude of stressinduced particle aggl omeration during the handling process because the adsorption of cati onic surfactant onto ceria surfaces provided the stronger repulsive electrostatic in teraction between ceria particles as compared to un-formulated ceria slurry (Figure 6-5). Furthermore, the effect of surfactant addi tion on the stability of silica slurries was measured by dynamic light scattering and SPOS. Figure 6-6 shows the primary particle size distribution of 80nm silica slurries with cationic (CTAB) and anioni c surfactants (SDS) at pH 9. The addition of cationic surfact ant can neutralize the charges on silica surface that can reduce the magnitude of repulsive forces between silica. The agglomeration behavior was found during the slurry formulation and the handling process. Howe ver, an insignificant change in particle size distribution was observed in silica slurries with anionic surfactan t. Furthermore, the magnitude of stress-induced particle agglomeration can be clearly determined by the normalized tail distribution of oversize particles, as shown in Figure 6-7. The a ddition of anionic surfactant into silica slurry can provide the str onger electrostatic interactions be tween silica particles to defend against shear stress during slurry handling. Th e magnitude of surface charges was observed by zeta potential (Figure 6-8). Cons equently, alkaline silica slurry with SDS (anionic) surfactant exhibited better slurry stability due to increasing the strength of the electr ical double layer. In addition, the stability of un-formulated and formulated sl urries can be investigated by the polishing performance. For example, Figure 6-9 shows the surface roughness and AFM images of silica wafers polished by un-formulated and formulated ceria slu rries at pH 6. Silica wafers polished by un-formulated ceria slurri es showed a smooth surface and low surface roughness (RMS = 2.06 nm and Rmax= 10.6 nm), as shown in Figure 6-9(c). However, higher

PAGE 92

92 surface roughness (RMS = 10.3 nm and Rmax= 272 nm and deep micro-scratch were observed on silica wafers polished by un-formulated ceria sl urries circulated for 1000 turnovers because high shear stress generated by pumping device can significantly increase agglomerated particles during the handling process, as shown in Figure 6-9(e). Conversely, th e addition of cationic surfactant into ceria slurry can provide a stronger potential barrier to defend against shear stress and significantly reduce the magnitude of stress -induced particle agglomeration, as shown in Figure 6-4. Thus, a small increase in surface roughness (RMS = 2.18 nm and Rmax= 40.8 nm) was found in formulated ceria slurry circulated for 1000 turnovers, as shown in Figure 6-9(d). Summary Highly stable slurry plays an important role to minimize the process-induced particle agglomeration and ensure a superior polishing performance. The stability of slurry can be obtained by adding surfactants (e .g., anionic and cationic surfactants) to provide the polymer stabilization between particles. In this study, the stability of unformulated and formulated ceria and silica slurries was determined by their pr imary particle size di stribution and oversize distribution. As the ceria slurry near its isoelectric point (IEP = 6~7) exhibited the amphoteric properties, the addition of anionic and cationic surfactants into ceria slurries at pH 6 can provide a stronger potential barrie r to defend against the stress-induced particle agglomeration, resulting in the minimization of surface defectivity duri ng CMP of dielectric materials. However, the addition of cationic surfactant in to silica slurries deteriorated the stability and caused more agglomerates during the handling due to neutralizing the surface charges.

PAGE 93

93 11 0 0 10 20 30 40 50 C1000/C0 Turnovers (Conc./Conc.)Particle Size (m) Ceria slurry at pH 6 Silica slurry at pH 9 Stre ss Effect on Particle Agglomeration in Un-formulated Slurries Figure 6-1. Effects of stress-i nduced particle agglomeration in un-formulated ceria and silica slurries.

PAGE 94

94 -80 -60 -40 -20 0 20 40 60 80 02468101214pHZeta Potential (mV ) Silica Slurry Un-formulated Ceria Slurry Figure 6-2. Plot of zeta potential vs. pH: the potential curves of un-formulated silica and ceria slurries. -100 -80 -60 -40 -20 0 20 40 60 80 100 02468101214pHZeta Potential (mV ) Ceria Ceria + C10TAB (Cationic) Ceria + TMAH (Cationic) Ceria +SDS (Anionic) Ceria + PAA (Dispersant) Figure 6-3. Plot of zeta potential vs. pH: the potential curves of formulated ceria slurries with varying surfactants.

PAGE 95

95 11 0 0 10 20 30 40 50 Formulated slurry (cationic) Un-formulated slurryC1000/C0 Turnovers (Conc./Conc.)Particle Size (m) Ceria Slurries at pH 6 Figure 6-4. Comparison of particle agglomeration in formulated and un-formulated ceria slurries due to stress effect.

PAGE 96

96 -80 -60 -40 -20 0 20 40 60 80 02468101214pHZeta Potential (mV ) Formulated Ceria Slurry Un-formulated Ceria Slurry Figure 6-5. Plot of zeta poten tial vs. pH in ceria slurry with cationic surfactant.

PAGE 97

97 0 10 20 30 40 0500100015002000Particle Size (nm)Volume % pH9 pH9-1000t pH9+SDS pH9+SDS-1000t pH9+CTAB pH9+CTAB-1000t Figure 6-6. Dynamic light scatte ring (DLS) experiment: the part icle size distributions of 80nm silica slurries with surfactants.

PAGE 98

98 11 0 0 2 4 6 8 10 Un-formulated slurry Formulated slurry with CTAB Formulated slurry with SDSC1000/C0 Turnovers (Conc./Conc.)Particle Size (m)Silica Slurries at pH 9 Figure 6-7. Comparison of particle agglomeration in formulated and un-formulated silica slurries.

PAGE 99

99 -60 -40 -20 0 20 40 02468101214pHZeta Potential (mV ) Silica+SDS Silica+CTAB Silica Figure 6-8. Plot of zeta pot ential vs. pH in silica slurry with surfactants.

PAGE 100

100 Silica Polishing-RMS0 2 4 6 8 10 12 Un-formulated ceria slurry Formulated ceria slurry1000 turnovers Unformulated ceria slurry1000 turnoversRMS(nm) Silica Polishing-Rmax0 50 100 150 200 250 300 Un-formulated ceria slurry Formulated ceria slurry1000 turnovers Unformulated ceria slurry1000 turnoversRmax (nm) Figure 6-9. Silica CMP: (a) surface roughness (RMS), (b) Rmax, (c) AFM images of silica wafers polished by un-formulated ceria slurry at pH 6, (d) formulated ceria slurry (cationic surfactant) at pH 6 (1000 turnovers), and (d ) un-formulated ceria slurry at pH 6 (1000 turnovers). (a) (b) (c) (d) (e)

PAGE 101

101 CHAPTER 7 NOVEL METHOD TO QUANTIFY THE DEGREE OF AGGLOME RATION IN HIGHLY STABLE CMP SLURRIES Introduction Highly stable slurry plays an important role in chemical mechanical polishing (CMP), as it minimizes particle-induced defectiv ity. The presence of agglomerated particles in a CMP slurry is one of the main causes of defects during polishing [Bas02]. The degree of agglomeration depends on external shear stress (e.g., slurry delivery) and inte rnal slurry chemistry (e.g., interparticle force) [Cha08]. As manufacturing nodes continue to scale down, it is becoming increasingly important to cont rol and reduce agglomerated particles to minimize defectivity during CMP. Thus, a reliable and accurate met hod for measuring agglomeration phenomena in CMP slurries is important to prevent particle-induced defectivity. Typically, measurements of interparticle force and/or zeta potential ar e used to assess the stability of CMP slurries qualitatively. Basim et al. [Bas03] studied surfactant effects on repulsive interaction forces be tween silica particles at high io nic strengths. They found that repulsive interaction forces, correlated to the leng th of surfactant chains, determined the stability of colloidal particles. Slurries with shorter surfactant chain lengths (less repulsive forces) were unstable (large particle size). Ot her researchers have measured zet a potential and light scattering (e.g., dynamic light scattering and laser diffraction) to investigat e how slurry chemicals (e.g., organic acids, ionic concentrati ons, and surfactants) a ffect colloidal disper sion behavior [Gop06, Vid05, and Eom02]. They observed agglomeration behavior when colloidal particles exhibited low zeta potential values. These measurements, however, do not fully describe agglomeration phenomena in CMP slurries because of complicating factors such as high ionic strength and multiple additives. In addition, light scattering techni ques are inadequate for determin ing the degree of agglomeration

PAGE 102

102 in CMP slurries, due to their inability to dete ct changes in slurry characteristics [Nic01]. Therefore, we developed a novel method for quan tifying the degree of agglomeration in a CMP system. We circulated typical CMP slurries us ing a high shear pumping device, and measured particle agglomeration characteri stics using a single-particle op tical sensor (SPOS). We used Smoluchowskis theory to model changes in tail distribution, and deve loped an agglomeration index to quantify the degree of agglomeration. Experimental Determining the Agglomeration Index The agglomeration index ( AI ) of CMP slurries can be determin ed in three steps: (1) subject CMP slurries to high shear forces, (2) measur e the slurry characteristics of oversize tail distributions, and (3) model change s in oversize tail distribution. High shear stress, generated by pumping devices, causes particle agglomeration during the handling process [Cha09]. We used magnetically levitated (Maglev) centrifugal and positive displacement pumps to circulate various types of sl urries (e.g., various surfactants, particle sizes and types, ionic concentrations and pH) in a slurry distribut ion system, to investigate how external shear stress and internal slurry chemistry affect the tail distributio n of oversize particles. In this experiment, slurries were subjected to a constant slurry flow rate of 12 L/min, and were circulated in a slurry loop to a maximum of 1000 turnovers, where a turn over was defined as the cycling of the total volume of each slurry (2 L) through the entire slurry loop once. We measured the particle agglomeration ch aracteristics of tail distribution using the AccuSizer 780 particle size measurement system, which consists of a single-particle optical sensor (SPOS). Its auto-dilution component can obta in a particle concentration that allows its photozone sensor to detect single par ticles ranging in size from 0.51 to 200 m.

PAGE 103

103 To calculate the agglomeration index, we used Smoluchowskis slow coagulation theory to model changes in particle tail distribution under external shear stress [Eli98 and Rus89]. Slow coagulation is defined as the presence of electrostatic inter action between particles, which defends against coagulation. We assumed that par ticle collisions were binary and proportional to particle concentration. The total change in the ra te of agglomerate concentr ations, such as singlet ( dN1/dt ), doublets ( dN2/dt ), and triplets ( dN3/dt ), can be expressed as: 311313211212 2 11111 1)/()/()/( NNWkNNWkNWk dt dN [7-1] 322323 211212 2 11111 2)/()/( 2 )/( NNWkNNWk NWk dt dN [7-2] 322323311313211212 3)/()/()/( NNWkNNWkNNWk dt dN [7-3] where the aggregation constant, kij = 4G(ai + aj)/3 is a function of the particle size (a ) aggregates, i and j-folds, and the shear rate ( G ), described by the orthok inetic theory. Individual agglomerate concentrations can be derive d from the above equations as follows: 1 1 0/1 )/( k k kt tN N [7-4] 0Nk Wij ij [7-5] where N0 is the total particle concentration, t is the aggregation time, and W is the stability ratio, defined as the ratio of the rapid aggregatio n rate in the absence of electrostatic interaction to the slow aggregation rate when electrostati c interactions occur between particles [Ste05]. Thus, we can define the agglomeration index ( AI ) as a logarithm ratio of external shear stress to the stability ratio (Log ( G/W)) Therefore, a slurry with a lower agglomeration index will have a more stable colloidal suspension. We determined the agglomeration index of CM P slurries by modeling changes in the tail distribution. Figure 7-1 schematically illustrates how we determined the agglomeration index.

PAGE 104

104 First, we used the initial tail distribution of oversi ze particles to simulate th e growth of individual agglomerates. By fitting both experimental and m odeling changes to tail distributions, we were able to determine the value of the agglomeration index. Effect of Agglomeration Index on Polishing Performance To investigate how the agglomeration index affected polishing performance, we used circulated slurries to polish 1-inch square Cu TEOS, and low k wafers in a TegraPol-35 Table Top Polisher, from Struers Co., w ith a down pressure of 3 psi, po lishing time of 1 min, slurry flow rate of 100 ml/min, and rotation speed of 150 rpm. The surface roughness (RMS) of polished Cu wafers was characterized using a Digital Instruments Nanoscope III atomic force microscope (AFM). Results and Discussion To investigate how external shear stress a ffected the agglomeration index, we used pumping devices to circulate sili ca slurries at pH 10 (primary particle size: 150 nm). As-received slurries were circulated by maglev centrifugal and positive displacement pumps for 500 turnovers (5000 seconds). Figure 7-2 illustrates th e changes in the fraction of oversize tail distributions, defined as a ratio of particle concentration at a specific particle size to the total particle concentration (1.41 1013 particles/ml). As the posi tive displacement pump generated highly localized shear stresses near the wall during the pump stroke, the tail distribution increased significantly compared to the maglev centrifugal pump. Subsequently, we used as-received particle c oncentrations (ranging in size from 0.51 to 1 m) to simulate the change in tail distribution at 5000 seconds. We used Equation 4 to simulate agglomeration concentrations (e.g., singlet, doublet, triplet, etc.) at their correlated particle sizes. We were able to obtain the change in tail di stribution by cumulating the simulated agglomerates.

PAGE 105

105 Consequently, we determined the agglomera tion index of CMP slurries by fitting both experimental and modeled tail di stributions, as shown in Figure 7-3. Circulated slurries had AI values of 2.48 and 4.48 in the maglev centrifuga l and positive displacement pumps, respectively. As AI can be defined as Log ( G/W ), the relative shear stress between pumping devices can be determined using identical slurry stability ( W ). As a result, the shear stress in the maglev centrifugal pump system was 100 times lower than that in the positive displacement pump system. To investigate how slurry chemistry affected the agglomeration index, we circulated typical CMP slurries with varying chemicals (e.g., pH, salt, and surfactant) and abrasive particles (e.g., size and type) in the slurry distribution using the positive displacement pump. We used the as-received slurries to simulate changes in tail distributions. Table 7-1 lists the AI values of those slurries; values ranged from 0.6 to 5. Slurries with lower AI values had more stable abrasive particles. The AI values for silica slurry (80 nm) were 1.76 at pH 9 and 2.46 at pH 3. Circulated slurries at pH 9 had less oversize tail distribution compared to circul ated slurry at pH 3, as shown in Figure 7-4. To assess how the agglomeration index affect ed polishing performance, we used these slurries (80 nm silica) to polish Cu wafers; the goal was to inves tigate particle-induced defectivity. Figure 7-5 shows typical AFM images of polished Cu wafers using as-received and circulated slurries. Figure 7-5 (a) shows the corresponding surface roughness (RMS). A Cu wafer that was not subjected to the polishing process had an uneven surface and high surface roughness (RMS = 6.14 nm), as shown in Figure 7-5(b). In contrast, a Cu wafer that was polished by as-received slurries had a flat su rface and lower surface roughness (RMS = 2.64 nm), as shown in Figure 7-5(c). As circulated slurries at pH 3 had more agglomerated particles,

PAGE 106

106 they resulted in high surface roughness (RMS = 7.98 nm) and large micro-scratches on polished Cu wafers, as shown in Figure 7-5(e); however, circ ulated slurries at pH 9 resulted in less surface roughness (RMS = 2.99 nm) on polished Cu wafers, as shown in Figure 7-5(d). Our numerous polishing tests (e.g ., using Cu, low k, and silica CMP) revealed that slurries with higher agglomeration index values (AI > 1.8) tended to contain mo re agglomerated particles during the handling process and to cause more surf ace defectivity (e.g., pits and micro-scratches) during CMP, as shown in Figure 7-6. Consequen tly, the agglomeration in dex can be used to quantify the degree of agglomer ation in highly stable CMP sl urries, and the use of low AI slurries during polishing can further minimize particle-induced defectivity. Summary The presence of agglomerated particles in a CMP slurry is one of the main causes of defects during polishing, particul arly as the manufacturing nodes continue to scale down. The degree of agglomeration depends on external (shear stress) and inte rnal (slurry chemistry) forces. Therefore, we developed the a gglomeration index (AI) to quantify the agglomeration phenomena caused by both external and internal forces. In typical CMP slurries AI values range from 0.6 to 5; slurries with a lower agglom eration index have more stable abrasive particles. Our modeled and experimental results revealed that the sh ear stress in a positive displacement pump was 100 times greater than that in a magnetically levita ted centrifugal pump. In addition, our numerous polishing experiments revealed that slurries with higher agglomer ation index values (AI > 1.8) contained more agglomerated particles during the handling process and caused more surface defectivity during polishing. Our novel method can be applied to determine slurry stability, and thus further minimize particle-induced defec tivity during CMP of metals and dielectrics.

PAGE 107

107 11 0 10-1310-1210-1110-1010-910-810-7 Part. Conc./Total Part. Conc.Particle Size (m) As-received Slurry Modeled the change of tail distribution Modeling Curve(a)11 0 10-1310-1210-1110-1010-910-810-7 Part. Conc./Total Part. Conc.Particle Size (m) As-received Slurry Modeled the change of tail distribution Modeling Curve11 0 10-1310-1210-1110-1010-910-810-7 Part. Conc./Total Part. Conc.Particle Size (m) As-received Slurry Modeled the change of tail distribution Modeling Curve(a) 11 0 10-1210-1110-1010-910-8 Part. Conc./Total Part. Conc.Particle Size (m) Positive Displacement Pump Fitting experimental and modeled tail distributions(b)11 0 10-1210-1110-1010-910-8 Part. Conc./Total Part. Conc.Particle Size (m) Positive Displacement Pump Fitting experimental and modeled tail distributions11 0 10-1210-1110-1010-910-8 Part. Conc./Total Part. Conc.Particle Size (m) Positive Displacement Pump Fitting experimental and modeled tail distributions(b) Figure 7-1. Schematic illustration of how the a gglomeration index is determined: (a) modeling the change in tail distribu tion of oversize particles; a nd (b) fitting experimental and modeled tail distributions to obt ain the agglomeration index.

PAGE 108

108 11 0 10-1210-1110-1010-910-8 Part. Conc./Total Part. Conc.Particle Size (m) As-received Slurry Magnetically levitated centrifugal pump Positive displacement pump Figure 7-2. Tail distributions of as-received slurry and circulated slurries by positive displacement and maglev centrifugal pumps.

PAGE 109

109 0.50.60.70.80.91.0 10-1110-1010-910-810-7 Part. Conc./Total Part. Conc.Particle Size (m) Maglev centrifugal pump Simulated Maglev centrifugal pump Positive displacement pump Simulated positive displacement pumpSilica Slurries (150nm) Figure 7-3. Effect of external shear stress on the agglomeration i ndex: experimental and modeled tail distributions in positive displ acement and maglev centrifugal pumps. AI = 4 4 8 AI = 2 4 8

PAGE 110

110 Table 7-1. Agglomeration indexes of typical chemical mechanical polishing (CMP) slurries. Slurry Agglomeration Index ( AI ) A 150nm Silica: pH 10 4.48 B 80nm Silica: pH 9 1.76 C 80nm Silica: pH 9+Anionic 0.66 D 80nm Silica: pH 3 2.46 E 30nm Silica: pH 11 1.36 F 30nm Silica: pH 11+Salt 1.66 G 30nm Silica: pH 7 3.92 H 30nm Silica: pH 7+Salt 4.26 I 30nm Silica: pH 2 4.36 J 30nm Silica: pH 2+Salt 4.96 K 70nm Ceria: pH 5+Cationic 1 2.36 L 70nm Ceria: pH 5+Cationic 2 1.96

PAGE 111

111 0.50.60.70.80.91.0 10-1310-1210-1110-10109 Part. Conc./Total Part. Conc.Particle Size (m) pH 9 Simulated pH 9 pH 3 Simulated pH 3Silica Slurries (80nm) Figure 7-4. Effect of internal slurry chemistr y on the agglomeration in dex: experimental and modeled tail distributions in circulated sl urries (80 nm silica) at pH 9 and 3. AI = 2 4 6 AI = 1 .76

PAGE 112

112 Cu CMP-RMS0 2 4 6 8 10 Un-polished wafer As-received slurry Circulated slurry at pH 9 Circulated slurry at pH 3RMS (nm) Cu CMP-Rmax0 50 100 150 200 250 Un-polished wafer As-received slurry Circulated slurry at pH 9 Circulated slurry at pH 3Rmax (nm) Figure 7-5. Cu chemical mechanical polis hing (CMP): (a) surface roughness (RMS and Rmax); and AFM images of Cu wafers (b) un-polis hed, (c) polished by as-received slurry, (d) polished by circulated slurry at pH 9, and (e) polished by circulated slurry at pH 3. (a) (b) (c) (d) (e)

PAGE 113

113 ABCDEFGHIJKL 0 1 2 3 4 5 6 Defectivity AISlurries Figure 7-6. Plot of th e agglomeration index (AI ) versus typical chemical mechanical polishing (CMP) slurries in determining particle-induced defectivity.

PAGE 114

114 CHAPTER 8 CONCLUSIONS AND SUGGESTIONS FOR FUTURE WORK Conclusions In the chemical mechanical planarization of dielectrics and metals, the presence of oversize particles in the slurry is one of the main cau ses of defectivity. Pumps within the slurry distribution system play a signi ficant role in increasing both the number and distribution of oversize particles. Therefore, th e effect of stress induced by diffe rent types of pump on particle agglomeration was investigated. We found the mean value of normalized oversize concentrations in the positive displacement pump system at 500 turnovers to be 6 times higher than that of a magnetically levitated centrifugal pump syst em. The magnetically levitated centrifugal pump was a low shear pump and did not increase the con centration of oversize particles significantly with slurry turnovers. A positive correlation was established between the roughness/defect density and the degree of agglomeration. Thus, a stable suspension in chemical mechan ical polishing (CMP) slur ry is an essential requirement to ensure superior polishing performance. The aggl omeration of the slurries was found to depend both on the extern al shear stress and the inte rparticle forces acting on the slurries. Our results indicate that alkaline silica slurries exhibited larger repulsive interparticle forces that successfully defended against shear fl ow and resulted in less agglomerated particles during the handling process. The least particle-induced defectivity was observed on polished low k wafers that utilized highly alkali slurries. Thus, alkaline silica slurries are the best condition for robust slurries to minimize the process-induced particle agglomeration and surface defectivity during CMP of metals and dielectrics. Furthermore, the stability of slurry can be obtained by adding surfactants (e.g., anionic and cationic surfactants) to provi de the polymer stabilization be tween particles. Typically, to

PAGE 115

115 obtain high removal rates of dielec tric substrates, the ceria slurry is taken near its isoelectric point (IEP) (e.g., pH 6~7), and th e silica slurry is alkaline (e .g., pH 9~11). The addition of anionic and cationic surfactants into ceria slurries at pH 6 and anionic surfactants into silica slurries at pH 9 can provide a stronger potential ba rrier to defend agains t the stress-induced particle agglomeration, resul ting in the minimization of su rface defectivity during CMP of dielectric materials. Highly stable slurry plays an important role in chemical mechanical polishing (CMP) as it decreases particle-induced defectivity. Typically th e stability of the slurry is measured by light scattering measurements. However, such measurem ents are inadequate in determining the degree of agglomeration. We have a novel experimental/theoretical approach to determine the degree of agglomeration in typical CMP slurries. This me thod is based on subjecting the slurry to high shear forces and measuring the particle agglomerati on characteristic of thei r tail distribution. By modeling the change in the tail distribution, the degree of agglomeration can be determined. Consequently, the agglomeration index ( AI ) was established by this approach, and distributed in the range of 0.6 to 5; the lower agglomeration index of the slurry, the more stability of colloidal suspensions. Based on numerous polishing results, slurries with a high agglomeration index ( AI > 1.8) were unstable and tended to agglomerat e more oversize partic les during the handling process that caused a significan t increase in surface defectivity (e.g., micr o-scratches) during CMP. This novel method can be applied to determin e the slurry stability for minimizing particleinduced defectivity during CMP of metals and dielectrics. Suggestions for Future Work As manufacturing nodes continually scale down, new challenges emerge from both the front and back end processes. For example, shri nking the transistor feature size is expected to apply atomic scale polishing to polysilicon polishi ng. Thus, the trend in CMP slurries is to

PAGE 116

116 reduce particle sizes with tunable chemicals in order to ensure slurry stability and protect a device from damages. To minimize particle-induced defectivity, slurry stability/robustness is essential. Achieving robustness is made possi ble by introducing polymer surfactants (ionic and nonionic surfactants) into colloidal suspensions. The mechanism of surfactant adsorption onto me tal oxide surfaces (e.g ., silica, ceria, and alumina) is the key factor in achieving polymer stabilization. Thus, four ier transform infrared spectroscopy (FTIR) can be employ ed for studying the surfactant ad sorption due to its sensitivity in detecting chemical bonding. Furthermore, the effect of surfactant concentr ations on the magnitude of surface charges of colloidal particles should also be studied. The addition of ionic surfactants into reverse charge particles could worsen the slurry stability due to screening the surface charges, but then improve the slurry stability because more surfactants adso rb onto particle surfaces. Thus, the quantity of surfactant concentration could be determined. Finally, the agglomeration index can be employed for determining the degree of agglomeration in commercial CMP slurries. More extensive slurry studies will be needed to be conducted as a function of particle types (e.g., al umina, silica, and ceria), particle sizes, and chemicals to build up the agglomeration index. Th e slurry stability will be confirmed by the polishing performance.

PAGE 117

117 LIST OF REFERENCES Abi05 J. T. Abiade, W. Choi, R. K. Singh, J. Mater. Res ., 20, 1139 (2005). Asa95 A. H. Perera, J.-H. Lid, Y.-C. Ku, M. Azrak, B. Taylor, J. Hayden, M. Thompson, M. Blackwell, EDM Tech. Dig. 679 (1995). Bal04 S. Balakumar, X. T. Chen, Y. W. Chen, T. Selvaraj, B. F. Lin, R. Kumar, T. Hara, M. Fujimoto, and Y. Shimura, Thin Solid Films 462, 161 (2004). Bar99 J. Bare, B. Johl, T. Lemke, Semiconductor International January (1999). Bas00 G. B. Basim, J. J. Adler, U. Mahajan, R. K. Singh, B. M. Moudgil, J. Electrochem. Soc ., 147, 3523 (2000). Bas02 G. B. Basim and B. M. Moudgil, J. Colloid Interface Sci. 256, 137 (2002). Bas03 G. B. Basim, I. U. Vakarelski, and B. M. Moudgil, J. Colloid Interface Sci., 263, 506 (2003). Ber08 P. Bernard, S. Valette, S. Daveau, J.C. Abry, P. Tabary, Ph. Kapsa, Tribology International 41, 416 (2008). Bib98 T. Bibby and K. Holland, J. Electron. Mater ., 27, 1073 (1998). Bie99 M. Bielmann, U. Mahajan, R. K. Singh, D. O. Shah, and B. J. Pallab, Electrochem.and Solid-StateLetters 2, 148 (1999). Bor02 C. L. Borst, W. N. Gill, R. J. Gutmann, ChemicalMechanical Polishing of Low Dielectric Constant Polymers and Organosili cate Glasses: Fundamental Mechanisms and Application to IC Interconnect Technology, Kluwer Academic Publishers, Boston (2002). Bow02 P. Bowen, J. Dispers. Sci. Technol ., 23, 631 (2002). Bu05 K. H Bu and B. M. Moudgil, Mater. Res. Soc. Symp. Proc ., 867, W8.5.1 (2005). Bu07 K. H. Bu and B. M. Moudgil, J. Electrochem. Soc. 154, H631 (2007). But91 H.-J. Butt, Biophysical Journal 60, 1438 (1991). But95 H.-J. Butt, M. Jaschke, and W. Ducker, Bioelectrochemistr y and Bioenergetics 38, 191 (1995). Cha96 A. Chatterjee, D. Rogers, J. McKee, I. Ali, S. Nag, and I.-C. Chen, IEDM Tech. Dig ., 829 (1996). Cha04a B. Chandran, R. Mahajan, M. Bohr, C.-H. Jan, and Q. T. Vu, Future Fab International 17 (2004).

PAGE 118

118 Cha04b N. Chandrasekaran, S. Ramarajan, W. Lee, G. M. Sabde, S. Meikle, J. Electrochem. Soc ., 151, G882 (2004). Cha08 F. C. Chang, S. Tanawade and R. Singh, Semiconductor International 31, 46 (2008). Cha09 F.-C. Chang, S. Tanawade, and R. K. Singh, J. Electrochem. Soc., 156, H39 (2009). Cho04a W. Choi, S.-M. Lee, and R. K. Singh, Electrochem. Solid-State Lett. 7, G141 (2004). Cho04b W. Choi, U. Mahajan, S.-M. Lee, J. Abiade, and R. K. Singh, J. Electrochem. Soc ., 151, G185 (2004). Coo90 L.M. Cook, J. Non-Crystall. Solids, 120, 152 (1990). Cum95 M. J. Cumbo, D. Fairhurst, S. D. Jacobs, and B. E. Puchebner, Applied Optics 34, 3743 (1995). Den06 D. Denardis, D. Rosaled-Yeomans, L. Borucki, and A. Philipossian, Thin Solid Films, 513, 311 (2006). Doe08 R. Doering and Y. Nishi, Handbook of semiconducto r manufacturing technology, CRC Press, Boca Raton, (2008). Duc91 W. A. Ducker, T. J. Senden, and R. M. Pashley, Nature, 353, 239 (1991). Duc92 W. A. Ducker, T. J. Senden, and R. M. Pashley, Langmuir 8 1831 (1992). Ein07 Y. Ein-Eli and D. Starosvetsky, Electrochim. Acta 52, 1825 (2007). Eli98 M. Elimelech, J. Gregory, X. Jia, R. A. Williams, Particle Deposition and Aggregation Butterworth-Heinemann, Oxford, (1998). Eom02 D.-H. Eom, J.-G. Park, and E.-S. Lee, Jpn. J. Appl. Phys., 41, 1305 (2002). Eom07 D. H. Eom, I. K. Kim, J. H. Han, and J. G. Park, J. Electrochem. Soc. 154, d38 (2007). Eva92 C. Evans, R. Brundle, and S. Wilson, Encyclopedia of Materials Characterization Encyclopedia of Materials Characterization, Elsevier Butterworth-Heinemann, Boston, (1992). Fran00 G. V.Franks, Z. Zhou, N. J. Duin, and D. V. Boger, J. Rheol 44, 759 (2000). Gop06 T. Gopal and J. B. Talbot, J. Electrochem. Soc., 153, G622 (2006). Hac97 V. A. Hackley, J. Am. Ceram. Soc. 80, 2315 (1997). Hay95 Y. Hayashi, M. Sakurai, T. Nakajima, K. Hayashi, Jpn. J. Appl. Phys., 34, 1037 (1995).

PAGE 119

119 Hig82 K. Higashitani, R. Ogawa, G. Hosokawa, and Y. Matsuno, J. Chem. Eng. Jpn ., 15, 299 (1982). Hij04 K.-I. Hijioka, F. Ito, M. Tagami, H. Ohta ke, Y. Harada, T. Takeuchi, S. Saito and Y. Hayashi, Jpn. J. Appl. Phys. 43, 1807 (2004). Hun01 R. J. Hunter, Foundations of colloid science Oxford University Press, (2001). Ile79 R. Iler, The Chemistry of Silica Wiley, New York (1979). Isr91 J. Israelachvili, Intermolecular and Surface Forces, Academic Press, London, (1991). ITRS07 International Technology Roadmap for Semiconductors, interconnect (2007). Joh03 B. Johl and R.K. Singh, Solid State Technology, 46, 63 (2003). Len08 Y. Leng, Materials Characterization: Introductio n to Microscopic and Spectroscopic Methods, John Wiley & Sons, Singapore, (2008) Li08 Y. Li, Microelectronic Applications of Chemical Mechanical Planarization John Wiley & Sons, New Jersey, (2008) Lin08 J.-Y. Lin, A. C. West, and C.-C. Wan, J. Electrochem. Soc. 155, H396 (2008). Lit04 M. Litchy and R. Schoeb, Semiconductor International 27, 87 (2004). Lit05 M. Litchy and R. Schoeb, MRS Symp. Proc ., 867, W2.8.1 (2005). Kah08 A.B. Kahng, IEEE Transactions on Computer-Aided De sign of Integrated Circuits and Systems, 27, 3 (2008). Kap02 M. Kappl, H.-J. Butt, Part. Part. Syst. Charact., 19, 129 (2002). Kim08a I.-K. Kim, Y.-J. Kang, T.-G. Kim, and J.-G. Park, Jpn. J. Appl. Phys. 47, 108 (2008). Kim08b S. Kim, J. H. So, D. J. Lee, and S. M. Yang, J. Colloid Interface Sci., 319, 48 (2008). Klo02 S. G. Kloster, T. Xu, G. Blaine, J. Sun, Y. Zhou, Interconnect Technology Conference, Proc. IEEE 242 (2002). Kub67 O. Kubaschewski, E. Evans, and C. Alcock, Metallurgical Thermochemistry Pergamon Press, Oxford (1967). Mae03 K. Maex, M. R. Baklanov, D. Shamirya n, F. Iacopi, S. H. Brongersma, and Z. S. Yanovitskaya, J. Appl. Phys. 93, 8793 (2003). Mah00 U. Mahajan, M. Bielman, and R. K. Singh, Mater. Res. Soc. Proc. 566, 27 (2000).

PAGE 120

120 Mis07 K. Mistry, C. Allen, C. Auth, B. Beattie, and D. Bergstrom, Proceedings of International Electron Devices Meeting 247 (2007). Mor02 I. D. Morrison and S. Ross, Colloidal suspensions: Suspensions, Emulsions, and Foams John Wiley and Sins, New York, (2002). Mur93 S. P. Murarka, Metallization: Theory and Practice for VLSI and ULSI Butterworth Heinemann, Boston, MA, (1993). Nai02 R. Nair, IBM J. Res. Develop. 223 (2002). Nic01 K. Nicholes, R. K. Singh, D. Grant, and M. R. Litchy, Semiconductor International July, 201 (2001). Ois00 T. Oishi, K. Shiozawa, A. Furukawa, Y. Abe, and Y. Tokuda, IEEE Transactions on Electron Devices 47, 822 (2000). Oss02 K. Osseo-Asare, J. Electrochem. Soc., 149, G651 (2002). Pal99 B. J. Palla and D. O. Shah, IEEE/CPMT Int. Electron. Manufact. Tech. Symp. 362 (1999). Pal00 B. J. Palla and D. O. Shah, J. Colloid Interface Sci. 223, 102 (2000). Pal02 B. J. Palla and D. O. Shah, J. Colloid Interface Sci. 256, 143 (2002). Pan07 S. Pandija, D. Roy, and S. V. Babu, Mater. Chem. Phys. 102, 144 (2007). Ram00 S. Ramarajan, Y. Li, M. Hariharaputhiran, Y. S. Her, and S. V. Babu, Electrochem. Solid-State Lett. 3, 232 (2000). Par65 G. A. Parks, Chem. Rev., 65, 177 (1965). Pro98 T. Provder, Particle Size Distributio n III: Assessment and Characterization, Chapter 6 American Chemical Society (ACS) Symposium Series, Orlando, FL. Ram00 S. Ramarajan, Y. Li, M. Harihara puthiran, Y.-S. Her, and S. V. Babua, Electrochem. Solid-State Lett ., 3, 232 (2000). Rem06 E. E. Remsen, S. Anjur, D. Boldridge, M. Kamiti, S. Li, T. Johns, C. Dowell, J. Kasthurirangan, and P. Feeney, J. Electrochem. Soc. 153, G453 (2006). Ros04 M. J. Rosen, Surfactants and Interfacial Phenomena Wiley, New York, NY, (2004). Rus89 W. B. Russel, D. A. Saville, and W. R. Schowalter, Colloidal Dispersions Cambridge University, United Kingdom, (1989). Sak93 T. Sakurai, IEEE Transactions on Electron Devices 40, 118 (1993).

PAGE 121

121 Sez01 S. M. Sze, Semiconductor Devices: Physics and Technology Wiley, New York, NY, (2001). Sha04 D. Shamiryan, T. Abell, F. Lacopi, K. Maex, Mater. Today, 7, 34 (2004). Smo17 M. von Smoluchowski, Z. Phys. Chem. 92, 129 (1917). Sin01 R. K. Singh and B. R. Roberts, IEEE/SEMI Adv. Semic onduct.Manufact. Conf. p. 107 (2001). Sin02a R. K. Singh, R. Bajaj, MRS Bull ., 27, 743 (2002). Sin02b R. K. Singh, S.-M. Lee, K.-S. Choi, G.. B. Basim, W. S. Choi, Z. Chen, and B. M. Moudgil, MRS Bull. 27, 752 (2002). Sin04 R. K. Singh, G. Conner, and B. R. Roberts, Solid State Technol ., 47, 61 (2004). Sta02 M. Staiger, P. Bowen, J. Ketterer, and J. Bohonek, J. Dispers. Sci. Technol ., 23, 619 (2002). Ste96 J. M. Steigerwald, S.P. Murarka, R.J. Gutmann, Chemical Mechanica l Planarization of Microelectric Materials New York (1996). Ste05 H. Stechemesser and B. Dobi, Coagulation and flocculation Taylor & Francis, Boca Raton (2005). Sup04 P. Suphantharida and K. Osseo-Asare, J Electrochem Soc. 151, G658 (2004). Tam02 S. Tamilmani, W. Huang, S. Raghavan, and R. Small, J. Electrochem. Soc. 149, G638 (2002). Teo04 T. Y. Teo, W. L. Goh, V. S. K. Lim, L. S. Leong, T. Y. Tse, and L. Chan, J. Vac. Sci. Tech., B 22, 65 (2004). Tse97 W.-T. Tseng, Y.-T. Hsieh, C.-F. Lin, M.-S. Tsai, and M.-S. Feng, J. Electrochem. Soc. 144, 1100 (1997). Vid05 G. Vidrich, J.-F. Castagnet, and H. Ferkelz, J. Electrochem. Soc.,152, C294 (2005). Xu02 G. Xu, E. Andideh, J. Bielefeld, and T. Scherban, Interconnect Technology Conference, Proc. IEEE 57 (2002). Zan04 P. B. Zantye, A. Kumar, and A. K. Sikder, Mater. Sci. Eng. R 45, 89 (2004). Zen05 T. F. Zeng, and T. Sun, IEEE Trans. Trans. Semiconduct. Manufact. 18, 655 (2005).

PAGE 122

122 BIOGRAPHICAL SKETCH Feng-Chi Chang was born in 1976, in Taiwan. He received his B.S. in m aterials science from Tatung University, Taipei, Taiwan, in 1999. Feng-Chi pursued furthe r studies at National Cheng Kung University, Tainan, Taiwan, and obt ained his M.S. in 2001. In January 2004, he was a process engineer at the chemical mechanical polishing (CMP) department in Taiwan Semiconductor Manufacturing Company (TSMC), where he worked in oxide and tungsten CMP and process development. In fall 2004, Feng-Chi began his Ph.D. studies at the University of Florida. In August 2005, he joined Dr. Singhs group for CMP studies. His dissertation research focu sed on the effects of external forces on particle aggl omeration during CMP of metals a nd dielectrics. In May 2007, he interned at Sinmat Inc., an R&D company in CM P slurry, where he studied slurry formulation for material removal, slurry stability, and po lishing uniformity. In December 2008, he graduated from the University of Florida with a doctorate in the Department of Materials Science and Engineering.