Title: Mixed surfactant systems to control dispersion stability in severe environments for enhancing chemical mechanical polishing (CMP) of metal surfaces
CITATION PDF VIEWER THUMBNAILS PAGE IMAGE ZOOMABLE
Full Citation
STANDARD VIEW MARC VIEW
Permanent Link: http://ufdc.ufl.edu/UF00100770/00001
 Material Information
Title: Mixed surfactant systems to control dispersion stability in severe environments for enhancing chemical mechanical polishing (CMP) of metal surfaces
Physical Description: Book
Language: English
Creator: Palla, Byron Joseph, 1972-
Publisher: University of Florida
Place of Publication: Gainesville Fla
Gainesville, Fla
Publication Date: 2000
Copyright Date: 2000
 Subjects
Subject: Surface active agents   ( lcsh )
Stabilizing agents   ( lcsh )
Colloids   ( lcsh )
Chemical Engineering thesis, Ph. D   ( lcsh )
Dissertations, Academic -- Chemical Engineering -- UF   ( lcsh )
Genre: government publication (state, provincial, terriorial, dependent)   ( marcgt )
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )
 Notes
Summary: ABSTRACT: The stability of colloidal dispersions is a critical parameter in many industries such as paints and pigments, minerals processing and electronics. Particle settling is often caused by the shielding of surface charges on the particles which otherwise would prevent coagulation and subsequent settling. This is particularly a problem in high ionic strength dispersions, where large amounts of ions serve to enhance the charge shielding and compression of the electrical double layer around the particles. This phenomenon has been investigated for industrially significant slurries used for tungsten and copper chemical mechanical polishing (W-CMP and Cu-CMP). It has been found that the effects of addition of conventional stabilizing agents (e.g. ionic surfactants, polymers) to these high ionic strength slurries are neutralized by the electrolytes in solution. However, the synergistic combination of a properly chosen ionic and nonionic surfactant has been found to be a suitable stabilizing agent for this type of system. The stabilization observed for these mixed surfactant systems has been explained in terms of adsorption of ionic surfactant on particle surfaces and nonionic surfactant molecules penetrating the film of the ionic surfactant due to hydrocarbon chain interactions. The enhanced adsorption of nonionic surfactant in this mechanism brings about the steric stabilization of the slurry.
Summary: ABSTRACT (cont.): The factors influencing this stabilization mechanism have been examined, yielding a robust model for stabilization of chemically complex slurries. The use of a relatively hydrophobic nonionic surfactant in the mixture yields optimal stability, with increasing hydrophobicity originating from either an increase in the hydrocarbon chain length or a decrease in the length of the ethoxylated chain. The increased stability with hydrophobicity of nonionic surfactant suggests that the partitioning of nonionic surfactant out of aqueous solution is a more important factor than the enhanced steric stabilization brought about by increasing the length of the polymeric polar group. The effect of surfactant concentration is examined and shows that maximum stabilization occurs over a range of concentration, which is dependent on the chosen surfactants. The effect of ratio of ionic to nonionic surfactant is a more complex correlation, with solubility of each surfactant becoming an issue at ratios that favor that surfactant. The various factors influencing dispersion stability are verified by novel surfactant adsorption measurements for high ionic strength environments that utilize surface tension measurements and absorbance of dye into surfactant micelles.
Summary: ABSTRACT (cont.): The influence of dispersion stability on polishing performance has been correlated. First, the use of stable dispersions is found to prevent particle agglomeration. Next, the use of stable dispersions is found to have little effect on the polishing rate of blanket tungsten wafers. The surface quality (or planarization) is found to increase by adding surfactant, although it does not necessarily correlate with dispersion stability. The polishing performance is explained as due to a lubricating layer of surfactant film on the particles. Finally, the particulate contamination of polished wafers is found to decrease with added surfactant, with either single surfactants or mixtures of surfactants leading to enhanced particle removal efficiency. All of these results suggest that the use of stable dispersions and surfactant additives in CMP slurry formulations can enhance the polishing performance, particularly for metal substrates.
Summary: KEYWORDS: dispersions, stabilizing agents, mixed surfactant systems, chemical mechanical polishing (CMP), tungsten, copper, severe environments, particle size
Thesis: Thesis (Ph. D.)--University of Florida, 2000.
Bibliography: Includes bibliographical references (p. 165-173).
System Details: System requirements: World Wide Web browser and PDF reader.
System Details: Mode of access: World Wide Web.
Statement of Responsibility: by Byron Joseph Palla.
General Note: Title from first page of PDF file.
General Note: Document formatted into pages; contains xvii, 174 p.; also contains graphics.
General Note: Vita.
 Record Information
Bibliographic ID: UF00100770
Volume ID: VID00001
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: oclc - 50750885
alephbibnum - 002639581
notis - ANA6408

Downloads

This item has the following downloads:

byronpalla ( PDF )


Full Text










MIXED SURFACTANT SYSTEMS TO CONTROL DISPERSION STABILITY IN
SEVERE ENVIRONMENTS FOR ENHANCING CHEMICAL MECHANICAL
POLISHING (CMP) OF METAL SURFACES

















By

BYRON JOSEPH PALLA


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


2000




























Copyright 2000

by

Byron Joseph Palla



























I dedicate this dissertation to my father, Arthur Palla, my fiance Lori, my sisters Debbie

and Cindy, my brother Tim, and especially my late mother, Martha Palla, who would

have been so proud...















ACKNOWLEDGMENTS

I would like to express my sincere thanks and appreciation to my advisor,

Professor Dinesh Shah, chairman of my supervisory committee, for his support,

guidance, enthusiasm, and philosophical lessons on love, life and the pursuit of

happiness. May these lessons never be overlooked nor forgotten as we continue our

journey. Thanks also to the other supervisory committee members, including Professors

Rajiv Singh, Raj Rajagopalan, Spyros Svoronos and Chang Wong Park for their valuable

time and suggestions.

I have also greatly appreciated the financial support and the collaboration

opportunities provided by the Engineering Research Center (ERC) for Particle Science

and Technology at the University of Florida. The opportunities to discuss research goals

and directions have been the primary guidance for my research, and this dissertation is

the result. I would also like to sincerely thank the ERC faculty, particularly the director

Professor Brij Moudgil, for critical analysis of my research that has kept me on track

through the years. I would also like to acknowledge all the undergraduate students

sponsored by the ERC for their experimental contributions to this thesis: Jennifer Hite,

Jason Shaw, Robel Vina, Shane Todd, Jason Shoemaker, Augustine Jeyakumar, Matthew

Brinkman, Wil Companioni and Diana Widjaya (masters student). I would also like to

acknowledge the members of the Goal II: Dispersion and CMP group for their guidance

and stimulating discussions, particularly Bahar Basim, Joshua Adler, Uday Mahajan,

Seung-Mahn Lee, Pankaj Singh, and Kimberly Christmas.









I would also like to thank my colleagues from the Department of Chemical

Engineering and the Center for Surface Science and Engineering for their help and good

cheers, including Dr. Alex Patist, Dr. Steve Truesdail (and Vikki), Dr. Paul Huibers, Dr.

Michael Free, Dr. Rahul Bagwe, Dr. Dibakhar Dhara, James Kanicky, Linda Jacoby,

Brian Burgess, and many others with whom I have had the pleasure of working.

Finally I would like to acknowledge the assistance of Dr. Bhavani Sankar from

the Department of Aerospace Engineering, Mechanics and Engineering Sciences and Dr.

Jose Matutes-Aquino from the Advanced Materials Research Center in Chihuahua,

Mexico, for their guidance and assistance on previous projects, and also the faculty of

The University of Texas at Austin and my colleagues at 3M Corporation for the

knowledge and inspiration that started this quest.
















TABLE OF CONTENTS

page

A C K N O W L E D G M E N T S ................................................................................................. iv

LIST OF TABLES .................................... .. .... .... .................ix

LIST OF FIGURES ................................... ...... ... ................. .x

A B ST R A C T ............. ............................................................. ......... xv

CHAPTERS

1 IN TR O D U C TIO N ....................... ........................... .. ........ ..............

1.1 Dispersion Stability ............................... ......... 1
1.2 Long Range Forces in Solution ........................................................ ....... ... ... 3
1.2.1 Charge Screening in High Ionic Strength Dispersions .................................... 4
1.2.2 Van der W aals Attractive Forces ........................................... .. ........... 8
1.2.3 D LV O Theory ............... ............. ...... ...... .. .. .......... 9
1.3 Chemical M echanical Polishing (CM P) .............. ................................... ........ 11
1.3.1 CM P Process Com ponents ................................................... ...... ........ 12
1.3.2 CMP Polishing Materials and Slurry Chemistries ............................ 16
1.3.3 Tungsten CM P .. ................. .......................... .. .......... .. 19
1.3.4 C opper C M P .... ...................... ...... ................................ .............. 20
1.3.5 Aluminum CM P ................ ...................................... .... ............ .. 22
1.3.6 CM P of Other M etals.................................................................. .............. 23
1.3.7 CMP of Silica (SiO2) and low-K Dielectric Materials............................ 24
1.4 Surfactants and Polymers as Stabilizing Agents ......................... .............. 26
1.4.1 Ionic Surfactants ... ................................................ ............ .. 28
1.4.2 Nonionic Surfactants .............................................. 28
1.4.3 Polymers ........................................... .......... 29
1.4.4 Applications as Stabilizing Agents .............. .................................... ........ 30
1.5 Surfactants in Solution ................................................... .... .......... ........ 31
1.5.1 Monolayers at the Air/Liquid Interface ............................................... 33
1.5.2 M icelles ................................ ... ....... ......... .......... 34
1.5.3 Adsorption of Surfactant on Particles............................. 36
1.6 R ationale of the Proposed R esearch............................................. .... .. .............. 38









2 STABILIZATION OF HIGH IONIC STRENGTH DISPERSIONS USING THE
SYNERGISTIC BEHAVIOR OF A MIXED SURFACTANT SYSTEM.....................40

2.1 Synergistic Behavior of Surfactant M ixtures............... ............... ................... 40
2.2 Methods for Sedimentation Experiments ............................. ... ............ 43
2.2.1 Method of Dispersion Preparation ....................................................... 44
2.2.2 M ethod of Sedimentation Characterization ................................................. 45
2.3 Effects of Addition of Ionic or Nonionic Surfactants on Slurry Stability ........... 48
2.3.1 Ionic Surfactants ... .... ........................................ .... ............ .. 49
2.3.2 N onionic Surfactants ................................................................................. 50
2.3.3 Verification by Sedimentation Experiments ............................................ 51
2.3.4 Adsorption Verification by Surface Tension Measurements.......................... 53
2.4 Effects of Addition of Surfactant Mixtures on Slurry Stabilization ..................... 56
2.5 Explanation of Stabilization M echanism .................................... .................... 60

3 MOLECULAR FACTORS THAT OPTIMIZE DISPERSION STABILITY USING
MIXED SURFACTANT SYSTEMS AS STABILIZERS FOR DISPERSIONS IN
SEV ER E EN V IR O N M EN T S ................................................................. .....................64

3.1 Molecular Factors Influencing Stability of Dispersions ...................................... 64
3.2 Nonionic Surfactants Used in This Investigation .............................................. 66
3.3 Optimization of Dispersion Stability Using Various Nonionic Surfactants for
T u n g sten C M P ............................ .................................................... 6 7
3.4 Verification of Results for M odel Copper CMP Slurry......................................... 71
3.5 Correlation of Dispersion Stability with HLB Number of Nonionic Surfactant ... 73
3.6 Effect of Total Surfactant Concentration on Dispersion Stability ..................... 76
3.7 Effect of Ratio of Surfactants in Mixtures on Dispersion Stability ....................... 79
3.8 Explanation of Observed Instability in Surfactant Mixtures at High Concentrations
and at Ratios Favoring N onionic Surfactant.............................. .................... 82

4 ADSORPTION OF SURFACTANTS ON PARTICLES FROM MIXED
SURFACTANT SOLUTIONS IN SEVERE ENVIRONMENTS ..................................85

4.1 Review of Previous M ethods and Results........................................................... 85
4.2 Novel Methods and Apparatus for High Ionic Strength Environments .............. 87
4.2.1 Total Adsorption Measurement .......................... ..... ............ 88
4.2.2 SDS Adsorption Measurement Using a Titration Method............................. 90
4 .2 .3 M materials and M ethods..................................... .. ..................................... 92
4.3 Application of Eosin-Y Dye in Total Surfactant Adsorption Measurements ....... 94
4.4 The Use of Merocyanine 540 Dye to Study the Adsorption of Mixed Surfactants
on A lum ina Particles................. .. ............ ........ ....... .............. 96
4.5 Total Surfactant Adsorption Isotherm of SDS/Tween 80 Mixture on Alumina
P articles............................. ... ... ..................... ... ...... 99
4.6 SDS Adsorption Results Using the Titration Method..................................... 101
4.7 Comparison of Adsorption from SDS/Tween 80 and SDS/Symperonic A4
M ix tu re ............... .................................. ..................................... 10 4









5 CORRELATION OF DISPERSION STABILITY TO PARTICLE SIZE
M E A SU R E M E N T S .............................................................................. ..................... 108

5.1 Importance of Particle Size versus Dispersion Stability Correlation.................. 108
5.2 Materials and Methods for Particle Size Analysis.................... .............. 110
5.3 Particle Size of Alumina Dispersions with and without Salt............................ 115
5.4 Correlation of Dispersion Stability to Particle Size ........................................... 117
5.5 Correlation of Dispersion Stability to Surface Area of Particles....................... 121
5.6 Verification of Particle Size Results with Other Systems ............................ 123
5.7 Implications for Metal CMP Applications................. .................... 126

6 A CORRELATION OF DISPERSION STABILITY TO POLISHING
PERFORMANCE OF CMP SLURRIES ..................................................127

6.1 Focus of the Correlation of Dispersion Stability with Polishing Performance... 127
6.2 M ethods for Polishing Performance Analysis ........................................ ...... 129
6.3 Effects of Dispersion Stability on Surface Quality............................................. 131
6.4 Effects of Dispersion Stability on CMP Polishing Rate .............. .............. 135
6.5 Lubrication Mechanism of Adsorbed Organic Layer ..................... .............. 139
6.6 Effect of Adsorbed Surfactant and Dispersion Stability on Particulate
Contamination of Polished Wafers for Post-CMP Cleaning Applications ......... 140
6.6.1 Rationale and M ethodology .......................................... ............ .............. 140
6.6.2 Correlation of Dispersion Stability with Particulate Contamination ............ 142
6.6.3 Effect of Single Surfactants on Particulate Contamination .......................... 145

7 SUMMARY AND RECOMMENDATIONS FOR FUTURE RESEARCH ..............148

7.1 D ispersion Stability in Severe Environm ents ..................................................... 148
7.2 Stabilization of High Ionic Strength Dispersions Using the Synergistic Behavior of
a M ixed Surfactant System ....... ....... ..................... ..... .. ................... 149
7.3 Molecular Factors that Optimize Dispersion Stability Using Mixed Surfactant
Systems as Stabilizers for Dispersions in Severe Environments ................... 151
7.4 Adsorption of Surfactants on Particles from Mixed Surfactant Solutions in Severe
Environm ents ................. ....................... ........................ ....... .......... 152
7.5 Correlation of Dispersion Stability to Particle Size Measurements.................... 153
7.6 Correlation of Dispersion Stability to Polishing Performance for CMP
A applications ......................... ..... ... .................... ... ........... 154
7.7 Recommendations for Future Research..... ...................... ........ 156
7.7.1 High Priority Recommendations........................ ........... ... 156
7.7.2 Long-term Recommendations .............. ...... ...................................... 161

REFEREN CES .................................. .. ........... .............. 165

BIO GRAPH ICAL SK ETCH ................................................. .............................. 174
















LIST OF TABLES


Table Page

1-1. Current applications in which dispersion stability is an issue..........................2

1-2. Properties of low resistivity metals (from Li et al., 1994 and Wolf and Tauber,
1 9 8 6 ) ...................... .. .. ......... .. .. ............................................. . . 1 7

1-3. Slurry chemistries applied to CMP applications. ............................................18

1-4. Dielectric constants of selected candidate low-K ILD materials .......................24

2-1. Fractions used for determination of total settled volume based on degree of
cloudiness in sedimentation experiments. ....................................................... 47















LIST OF FIGURES


Figure Pag(

1-1. Effect of a) salt concentration and b) counterion valency on electrical potential
decay from a plane charged surface according to Equation (1.3) .....................6

1-2. Illustration of the effect of salt (K3Fe(CN)6 oxidizing agent) on the stability of
dispersions of charged alumina particles: a) particles in water at pH 4, b)
w ith addition of salt. ........................... .................. .............. .......... ...... .

1-3. Settling of alumina particles after addition of oxidizer, potassium ferricyanide.
Slurries are 1 wt. % AKP-50 (100-300 nm) alumina particles at pH 3,4, and 5
(left to right), allowed to settle for 24 hours.................... ..... .........8....8

1-4. Processing steps involved in multilevel integrated circuit manufacturing.............11

1-5. Schematic of the CMP process, including an illustration of the interaction
between the wafer and slurry in tungsten CMP (W-CMP). ................................12

1-6. Illustration of common problems associated with CMP of metal surfaces............15

1-7. Illustration of surfactants and polymers as stabilizing agents on hydrophilic
particles ............. ......... ........... ...................................... 27

1-8. Illustration of the preferential locations of surfactant molecules in solution.........32

1-9. Mechanisms for the two relaxation times, TI and c2, involved in a surfactant
solution above CM C.................... ....................... .. ......34

1-10. Stages of aggregation of adsorbed ionic surfactant on oppositely charged
particle surfaces, showing reverse orientation model (left) and bilayer model
(right) (from [Somasundaran and Krishnakumar, 1997]). ...............................37

2-1. Examples of sedimentation experiments along with the total settled volume
determined using the procedure outlined below (f = fraction settled as defined
in T able 2-1) .................................................. ................. . 46

2-2. Illustration of the effect of single ionic or nonionic surfactant addition on the
stability of high ionic strength slurries.............. .................................................48









2-3. The effects of ionic or nonionic surfactant addition on a low ionic strength
slurry containing 0.001 M potassium ferricyanide oxidizing agent. The
slurries are 1 wt. % AKP-50 alumina at pH 4 with 10 mM surfactant added.
The photographs were taken after 24 hours of settling .......................................51

2-4. The effects of ionic or nonionic surfactant addition on a high ionic strength
slurry containing 0.1 M potassium ferricyanide oxidizing agent. The slurries
are 1 wt. % AKP-50 alumina at pH 4 with 10 mM surfactant added. The
photographs were taken after 24 hours of settling.............................................. 53

2-5. Surface tension of solutions before and after addition of alumina particles, as a
function of surfactant concentration, achieved through dilution with water.
Surfactants are a) CPC, b) SDS and c) Tween 80. .............................................55

2-6. Sedimentation results of 0.1 M Fe(NO3)3 slurries with mixtures of surfactants
add ed ..............................................................................57

2-7. Sedimentation results of 0.1 M K3Fe(CN)6 slurries with mixtures of anionic
and nonionic surfactants added ........................................ ......................... 59

2-8. The mechanism of high ionic strength slurry stabilization by the synergistic
mixture of anionic and nonionic surfactants ......................................................60

2-9. Other possible conformations of stabilizing surfactant that result from the
adsorption of a mixture of ionic and nonionic surfactants on particle surfaces.
Conformations are a) full bilayer and b) penetrated bilayer.............................62

3-1. Schematic of a general ethoxylated nonionic surfactant and details about the
types used in this investigation. ........................................ ........................ 67

3-2. Sedimentation results of dispersions with 1:1 mixtures of SDS and the C12
nonionic surfactant indicated. Oxidizing agent is 0.1 M potassium
ferricyanide. Ex represents the mean number of ethoxy groups on the
nonionic surfactant m olecule. ........................................ ......................... 68

3-3. Sedimentation results of dispersions with 1:1 mixtures of SDS and the C18
nonionic surfactant indicated. Oxidizing agent is 0.1 M potassium
ferricy anide. ..................................................... ................. 6 8

3-4. Correlation of dispersion stability with hydrophobicity of nonionic surfactant,
as given by its HLB number. Data is from 50 mM 1:1 mixtures of SDS and
the given nonionic surfactant adsorbed on 10 wt. % AKP-50 particles in 0.1 M
K3Fe(CN)6 at pH 4.................... .................................. 70

3-5. Variation in dispersion stability vs. HLB number of nonionic surfactant for a
model copper CMP slurry containing 0.1 M KIO3, 0.01 M KI and 0.01 M
EDTA (Nonionic surfactants are: 1 = Span 80, 2 = Brij 93, 3 = Brij 52, 4 =
Span 20, 5 = Brij 30, 6 = Symperonic A4, 7 = Tween 81, 8 = Symperonic A7,









9 = Brij 97, 10 = Tween 21, 11 = Symperonic Al 12 = Tween 80, 13 = Brij
98, 14 = Tween 20, 15 = Brij 35, 16 = Symperonic A50, 17 = Brij 700). ...........72

3-6. Schematic of dispersion environment showing adsorption density differences
between a) high HLB hydrophilicc) nonionic surfactant and b) low HLB
(hydrophobic) nonionic surfactant. ............................................ ............... 74

3-7. Sedimentation experiments showing dependence of dispersion stability on
surfactant concentration in 0.1 M potassium ferricyanide slurries at pH 4...........77

3-8. Effect of surfactant concentration on dispersion stability in 0.1 M KIO3, 0.01
M KI and 0.01 M EDTA slurries at pH 6. .....................................................78

3-9. The effect of ratio of surfactants on dispersion stability for slurries containing
0.1 M K3Fe(CN)6 at pH 4. Total surfactant concentration is 50 mM. ..................80

3-10. The effect of ratio of surfactants on dispersion stability for slurries containing
0.1 M KIO3, 0.01 M KI and 0.01 M EDTA. ................................. ............... 82

3-11. The proposed mechanism of induced instability in mixed surfactant systems;
a) low concentrations/ratios of nonionic surfactant and b) high
concentrations/ratios of nonionic surfactant ............... .............. ..................... 84

4-1. The chemical structure of Merocyanine 540 (MC 540) anionic dye.................89

4-2. Equilibria involved in the two-phase, mixed indicator titration method for
anionic surfactants (from Li and Rosen, 1981). ............. .................................... 91

4-3. UV-Vis spectra using 0.02 mM Eosin-Y dye and solutions containing 0.1 M
K3Fe(CN)6 and varying concentrations of a) SDS + Tween 80, b) SDS only
and c) Tw een 80 only. ............................................. ................ ............. 95

4-4. UV-Vis spectra for SDS/Tween 80 mixtures at the indicated concentrations
(in mM), containing 0.02 mM MC 540 dye and 0.1 M K3Fe(CN)6. ...................... 97

4-5. Absorbance at 563 nm for SDS/Tween 80 solutions shown in Figure 4-4, with
C M C indicated. .....................................................................97

4-6. Absorbance at Xmaxfor (a) SDS only solutions with no K3Fe(CN)6 (Xmax= 557
nm) and (b) Tween 80 only solutions with 0.1 M K3Fe(CN)6 (Xmax= 563 nm),
w ith CM C indicated. ............................... .... .......... ................ ............. 98

4-7. Total adsorption isotherm of SDS/Tween 80 mixture on 10 wt. % AKP-50
alumina particles in 0.1 M K3Fe(CN)6 solution at pH 4: a) dye absorbance
method and surface tension method and b) correlation of adsorption data with
dispersion stability ..................... .................. .......................... 100









4-8. Verification of anionic surfactant titration method using known SDS
concentrations with and without both Tween 80 (T80) and K3Fe(CN)6 (PoFe).
Corresponding concentrations calculated assuming a 1:1 reaction between
SDS and Hyamine 1622 are given above each bar. ...................... ...............102

4-9. SDS adsorption and total surfactant adsorption on alumina particles from
mixtures of SDS and Tween 80 in 0.1 M K3Fe(CN)6 at pH 4 ....................... 103

4-10. Adsorption of SDS/Symperonic A4 mixture on 10 wt. % AKP-50 particles in
0.1 M K3Fe(CN)6 at pH 4. ................................... 105

4-11. Correlation of adsorption with dispersion stability for both SDS/Tween 80 and
SDS/Symperonic A4 mixtures. Results include total surfactant adsorption
(0), SDS adsorption (*), and dispersion stability (A). ................. .................106

5-1. Electron micrographs of two types of alumina particles used in particle size
correlation: a) Leco y-alumina and b) Sumitomo AKP-50 a-alumina ..............110

5-2. Viscosity of standard solutions of SDS and Tween 80 mixtures (1:1 molar
ratio) measured immediately (0) and 6 days (A) after preparation using a
Brookfield cone and plate viscometer. ......... ................... ..... ..............113

5-3. Acoustosizer particle size results for 10 wt. % alumina dispersions with and
without 0.1 M potassium ferricyanide salt added...............................115

5-4. Microtrac UPA 150 dynamic light scattering particle size results for 10 wt. %
a) AKP-50 a-alumina and b) Leco y-alumina dispersions with and without 0.1
M potassium ferricyanide salt added.................. ..............116

5-5. Correlation of dispersion stability to particle size for Leco 50 nm alumina
dispersions with 0.1 M potassium ferricyanide at pH 4. ....... ............118

5-6. Correlation of dispersion stability to particle size for AKP-50 100-300 nm
dispersions with 0.1 M potassium ferricyanide at pH 4. .............................120

5-7. Correlation of dispersion stability and particle size with surface area of the
abrasive particles. Data is taken from Figures 5-5 and 5-6 ..............................123

5-8. Mean particle size and dispersion stability for Fe(N03)3 based slurries
stabilized using various mixed surfactant systems. ....... ................ ............... 124

5-9. Mean particle size of copper CMP dispersions stabilized by mixed surfactant
systems (B = Brij, A = Symperonic, S = Span, T = Tween) ..............................125

6-1. AFM surface plots obtained for a) as deposited (unpolished) tungsten wafer;
and wafers polished with slurry containing b) no surfactant and c) 50 mM
SDS/Tween 80 surfactant mixture. Scan size = 1 |tm ......................................132









6-2. Surface quality of polished tungsten wafers from AFM surface plots as a
function of dispersion stability. Each polished wafer is shown as a separate
bar along with standard deviations from each of the five scans per wafer.
Surfactants are 1:1 mixtures of SDS and nonionic surfactant indicated ...............133

6-3. Surface quality results from AFM for varying concentration and ratio of
surfactants in SDS/Tween 80 mixture added to each slurry..............................134

6-4. Polishing rate for tungsten wafers polished with slurries containing the
indicated nonionic surfactant in a mixture with SDS. Values are the average
polishing rate obtained for two wafers.............. .............................................. 136

6-5. Evidence of lubrication mechanism from polishing results of tungsten using
slurry with and without SDS/Tween 80 mixture...............................138

6-6. Illustration of the lubrication mechanism of adsorbed surfactant and its effect
on depth of cut (6) and overall polishing rate......................................................140

6-7. SEM micrographs of W wafers dipped in 0.1 wt. % AKP-50 alumina slurries
with 0.1 M potassium ferricyanide, containing a) no surfactant; b) 10 mM
SDS/Brij 52 mixture; c) 50 mM SDS/Symperonic A4 mixture; and d) 10 mM
Sym peronic A 4 only. ............................... .... .......... ................ ............. 143

6-8. Particulate contamination of W wafers dipped in dispersions containing 1:1
mixtures of SDS and indicated nonionic surfactant, viewed by SEM .................144

6-9. Particulate contamination of W wafers dipped in dispersions containing a
variety of single surfactants, viewed by SEM ......................................................146















Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy



MIXED SURFACTANT SYSTEMS TO CONTROL DISPERSION STABILITY IN
SEVERE ENVIRONMENTS FOR ENHANCING CHEMICAL MECHANICAL
POLISHING (CMP) OF METAL SURFACES



By

Byron Joseph Palla

August 2000



Chairman: Dinesh O. Shah
Major Department: Chemical Engineering


The stability of colloidal dispersions is a critical parameter in many industries

such as paints and pigments, minerals processing and electronics. Particle settling is

often caused by the shielding of surface charges on the particles which otherwise would

prevent coagulation and subsequent settling. This is particularly a problem in high ionic

strength dispersions, where large amounts of ions serve to enhance the charge shielding

and compression of the electrical double layer around the particles.

This phenomenon has been investigated for industrially significant slurries used

for tungsten and copper chemical mechanical polishing (W-CMP and Cu-CMP). It has

been found that the effects of addition of conventional stabilizing agents (e.g. ionic









surfactants, polymers) to these high ionic strength slurries are neutralized by the

electrolytes in solution. However, the synergistic combination of a properly chosen ionic

and nonionic surfactant has been found to be a suitable stabilizing agent for this type of

system. The stabilization observed for these mixed surfactant systems has been explained

in terms of adsorption of ionic surfactant on particle surfaces and nonionic surfactant

molecules penetrating the film of the ionic surfactant due to hydrocarbon chain

interactions. The enhanced adsorption of nonionic surfactant in this mechanism brings

about the steric stabilization of the slurry.

The factors influencing this stabilization mechanism have been examined,

yielding a robust model for stabilization of chemically complex slurries. The use of a

relatively hydrophobic nonionic surfactant in the mixture yields optimal stability, with

increasing hydrophobicity originating from either an increase in the hydrocarbon chain

length or a decrease in the length of the ethoxylated chain. The increased stability with

hydrophobicity of nonionic surfactant suggests that the partitioning of nonionic surfactant

out of aqueous solution is a more important factor than the enhanced steric stabilization

brought about by increasing the length of the polymeric polar group. The effect of

surfactant concentration is examined and shows that maximum stabilization occurs over a

range of concentration, which is dependent on the chosen surfactants. The effect of ratio

of ionic to nonionic surfactant is a more complex correlation, with solubility of each

surfactant becoming an issue at ratios that favor that surfactant. The various factors

influencing dispersion stability are verified by novel surfactant adsorption measurements

for high ionic strength environments that utilize surface tension measurements and

absorbance of dye into surfactant micelles.









The influence of dispersion stability on polishing performance has been

correlated. First, the use of stable dispersions is found to prevent particle agglomeration.

Next, the use of stable dispersions is found to have little effect on the polishing rate of

blanket tungsten wafers. The surface quality (or planarization) is found to increase by

adding surfactant, although it does not necessarily correlate with dispersion stability. The

polishing performance is explained as due to a lubricating layer of surfactant film on the

particles. Finally, the particulate contamination of polished wafers is found to decrease

with added surfactant, with either single surfactants or mixtures of surfactants leading to

enhanced particle removal efficiency. All of these results suggest that the use of stable

dispersions and surfactant additives in CMP slurry formulations can enhance the

polishing performance, particularly for metal substrates.


xvii














CHAPTER 1
INTRODUCTION


1.1 Dispersion Stability


Dispersion is a term generally applied to a system of insoluble or partially soluble

solid particles dispersed in a liquid medium. Dispersion stability refers to the kinetic

stability of the particles in the medium, or in other words the ability of the particles to

remain "dispersed" over a relevant time scale. Dispersions are not thermodynamically

stable systems except in the rare case of exact density matched phases. The relevant time

scales of dispersion stability may vary from tenths of seconds to years, depending on the

application [Hiemenz and Rajagopalan, 1997]. The stability of a dispersion is an issue in

any industry in which settling of particles can result in poor performance. For example,

if particles settle during shipment of a slurry, the product received by the consumer will

appear undesirable, and often the settled particles cannot be re-dispersed by means

available to the consumer. In other industries, the stability of slurries can be more

susceptible to even a slight increase in mean particle size. For example, in an application

where particle size is crucial to the process, the product or process can be completely

changed by agglomeration of slurry particles yielding a higher mean particle size in the

slurry, even if the agglomeration is not enough to cause significant visible settling.

In most industrial applications, dispersions are complex formulations of particles

(sometimes multiple types) dispersed in an aqueous or organic phase, with chemical

agents added to promote theological behavior or some desired reaction or interaction with











a substrate. Dispersion stability can have a significant effect on performance, resulting in


insufficient film thickness [Pineiro and Himics, 1996], poor adhesion and insufficient


reactivity [Viesturs et al., 1999] amongst many other problems depending on the


application.





Table 1-1. Current applications in which dispersion stability is an issue.

Industry / Terms Used Example Systems Stability Issues
Application
Paints, Pigments Dispersions Titania with PMMA, MMA [Hegedus, 1993]; Rub-up, tint strength, transparency and
and Dyes Carbon black and copper phthalocyanine in hue, hiding, gloss and rheology
various solvents [Li, 1997] [Pineiro, 1996 and Himics, 1998]

Paper Pulp, Muds, Inks, paper, calcium carbonate, enzyme, Transport of pulp [Kankaanpaa, 1999];
Processing Tailings surfactant, oil in recycle process [Viesturs, Detachment and hydrophobicity of
1999]; Glucomannans, calcium chloride and toner, Microink agglomeration,
colloidal wood resin [Sihvonen, 1998] Redeposition of ink [Viesturs, 1999]
Minerals and Concentrate, Sodium polyacrylate and PVA with kaolin Flocculation and distribution in clays
Ceramics Tailings, Slips [Sjoberg, 1999]; Humic acid adsorbed on and aquifers [Kretzschmar, 1997]
ProNa-kaolinite in NaCIO4 for soils and aquifers
Processin[Kretzschmar, 1997]
Oil Refining Drilling Muds, Glycol and polyamines in water-based muds Gumbo problems on wells; Aquatic
Sales, Slips with salts [Aston, 1994]; Polysaccharide and toxicity and swelling [Welch, 1992];
potassium in lime muds with bentonite Deflocculation effectiveness [Weldes,
particles [Walker, 1984]; Sodium silicates 1969]
[Weldes, 1969]
Electonics Slurries, Alumina and silica with various chemical Polish rate, surface damage and
Dispersions additives for chemical mechanical polishing particulate contamination in CMP
(CMP) [Wang, 1996]; Titania in Pb-Sn [Bielmann, 1999]; Microstructural
eutectic solder for surface mount technology stability and creep resistance in SMT
(SMT) [Mavoori, 1998]; Powder and polymer [Betrabet, 1992]; Discharge capacity in
binder in lithium-ion batteries [Kim, 1999] batteries [Kim, 1999]
Biotechnology Broth Latex microspheres with adsorbed proteins Separation and collection of targeted
and other biomolecules [Kawaguchi, 1999]; molecules [Kawaguchi, 1999]; Particle
Salbutamol base particles and oleic acid morphology and particle size
dispersed in freon 11 and 12 [Eriksson, distribution [Barber, 1998]; Distribution
1995] of components in metered-dose
inhalers [Eriksson, 1995]
General Sludge PCB's, lead, oil and grease mixed with Retrieval transport and solid-liquid
Industrial additives and water [Grube, 1990]; Fly ash separations [Tingey, 1999]; Short- and
(heavy metals) and sewage sludge in ocean long-term extraction and leaching
dispersion [Young, 1997] [Grube, 1990]; Distribution,
modification by organisms in ocean


Industries in which dispersion stability is a critical issue are summarized in Table


1-1, along with some of the many terms synonymous with dispersion. Table 1-1 also lists









representative components of slurries and some of the critical issues in each industry

which can be affected by dispersion stability.

An important characteristic of many of the example systems listed in Table 1-1 is

that many of the dispersions are formed or are processed under severe environments.

Severe environments include complex chemical environments which contribute to

extreme pH or chemical reactivity conditions. Examples of these environments in Table

1-1 include acidic conditions found in the gastrointestinal tract in biotechnology

applications and extreme oxidizing or dissolving conditions found in electronics

applications for batteries or polishing. Another type of severe environment includes

extreme pressure or temperature conditions. Examples of these environments in Table 1-

1 include minerals and ceramics processing and oil refining. All of these severe

environments can make the stabilization of dispersions a much more formidable

challenge. This dissertation focuses on the chemically complex type of severe

environment. These systems pose a challenge for the technological solution due to the

effects of charge screening on the dispersed particles.


1.2 Long Range Forces in Solution


The difficulty in achieving dispersion stability in chemically complex

environments can be due to either severe pH, reactivity, or high ionic strength conditions.

The surface charge on particles in aqueous solution is pH dependent, and the nature of the

chemical species present on particle surfaces will depend on the chemicals in solution.

Likewise, stabilizing agents added to dispersions may be ineffective under extreme pH

and reactivity conditions. Most importantly for the present discussion, the chemical

agents present as ions in dispersions add to the ionic strength of the medium.











The effect of high ionic strength environments on dispersion stability comes about

as the result of the electrical double layer developed around a charged particle in solution.

The double layer refers to the electrical potential function surrounding the particle and its

common division into two layers, one with rapid potential decay near the surface and the

other with a more gradual decay farther from the surface. The following discussion

derives expressions which govern the magnitude of the potential decay in the region

farther from the surface, as derived in Adamson [1982], and then applies those

expressions to high ionic strength environments.


1.2.1 Charge Screening in High Ionic Strength Dispersions

If a plane surface bearing a uniform charge density is placed in contact with a

solution phase containing both positive and negative ions, the probability of finding an

ion at some particular point will be proportional to the Boltzmann factor e-e ."kT, where z

is the valence of the ion, e is the charge on the electron (1.6021 x 10-19 C), k is the

Boltzmann constant (1.3805 x10-23 J/molecule (K)), T is temperature and i is the

potential at that point in solution. The integration to infinity of the net charge density (p)

at any point must be equal in magnitude but opposite in sign to the surface charge density

(c) according to the equation:




= -fpdx (1.1)









This integration has been carried out using the additional assistance of Poisson's

equation, which relates the divergence of the gradient of the electrical potential at a given

point to the charge density at that point. The treatment of a plane charged surface and the

resulting diffuse double layer is due mainly to Gouy [1917] and Chapman [1913]. This

treatment is simplified by the assumption that x is now only a function of distance

normal to the surface. It is convenient to define the quantities y and yo as


zeo zeo o (1.2)
kT and kT


where y, is the potential at the surface. The solution to Equation (1.2) with the boundary

conditions (y = 0 and dy/dx = 0 for x = -; y = yo at x = 0) is


ey +I1+(eY -1)e"X
ey e ++ e 1 (1.3)
ey.2 + 1- (ey2 )eex



where K has units of length-1 and is defined by the equation

2
K2 -niz2 (1.4)
s;okT i


where s is the dielectric constant of the medium, So is the permittivity of vacuum (8.854 x

10-12), and niis the ion concentration.

Equation (1.3) can be simplified for the case of yo I 1 (or, for singly charged ions

and room temperature, Wo 1 25) as


y = oKe (1.5)









The quantity 1/K is the distance at which the potential has reached the 1/e fraction

of its value at the surface. This quantity is instrumental in defining the potential decay of

a charged surface, and is therefore often called the Debye screening length. To illustrate

the importance of ionic strength on the potential decay from a charged surface, this

quantity has been graphed as a function of both salt molar concentration (c) and

counterion valency (z) in Figure 1-1.



a) 5 b)
25 _-------------- b 50 i_______________

V.o= 25 mV V = 50 mV
20 z=l 40\ c= 0.1M -

> 15 > 30
E E
> *c = 0.001 M \
10 > 20 -l
zl

0.00
0 0.1 0
0 50 ) 100 150 0 10 20 30 40
xx (A)


Figure 1-1. Effect of a) salt concentration and b) counterion valency on electrical
potential decay from a plane charged surface according to Equation (1.3).



In Figure 1-1, it is evident that both salt concentration (part a) and counterion

valency (part b) have a significant effect on the electrostatic repulsion felt between two

equally charged surfaces. The electrostatic repulsion then depends significantly on the

ionic strength, I, of the solution, defined as I=1/2Zcizi2. Hence, both ionic strength and

Debye length are good indications of the extent of charge screening in a given

environment.

The decreased electrostatic repulsion associated with charge screening, if

significant enough, can bring about agglomeration of particle in a dispersion, as









illustrated in Figure 1-2 for high ionic strength slurries. This figure shows alumina

particles, which usually have a substantial positive charge at low pHs. However, the

addition of 0.1 M potassium ferricyanide, a strong oxidizing agent, shields the positive

charge with the multivalent ferricyanide anions, thus causing agglomeration of particles

due to van der Waals forces. The Debye length for this system has been calculated using

Equation (1.4). With a surface charge of 50 mV and a counterion valency of 3, the

Debye length in this system is 4.3 A.



a) o o o o o
0 0 0 00

o o
0 0 0 o
OO o O
00 0 0
O0 O


S With addition of salt

b)

K+0








at pH 4, b) with addition of salt.
K + K +0++ K+
Fc(CN)6 ^ Fe(CN) 3-

C9pococ 8 cMckbo


Figure 1-2. Illustration of the effect of salt (K3Fe(CN)6 oxidizing agent) on the
stability of dispersions of charged alumina particles: a) particles in water
at pH 4, b) with addition of salt.



Figure 1-3 shows the settling behavior of slurries with and without addition of

potassium ferricyanide oxidizing agent. The slurry without oxidizer (part a) is clearly

stable since the white particles are dispersed throughout the cylinder. The slurry

containing oxidizer (part b) is clearly unstable with all particles having settled in the 24









hours at which time this picture was taken. From Figures 1-2 and 1-3, it should now be

evident that in a dispersion of charged particles, the increase in ionic strength of the

medium can significantly decrease the dispersion stability.



'" "' -
r ".



Alumina Alumina +
Alumina
in Water 0.1 M Potassium
Ferricyanide



i




Figure 1-3. Settling of alumina particles after addition of oxidizer, potassium
ferricyanide. Slurries are 1 wt. % AKP-50 (100-300 nm) alumina particles
at pH 3,4, and 5 (left to right), allowed to settle for 24 hours.



1.2.2 Van der Waals Attractive Forces

The forces which are responsible for the agglomeration of particles under high

ionic strength conditions are commonly called van der Waals forces due to the fact that it

is these forces that give rise to condensation of a vapor to a liquid [Adamson, 1982]. The

van der Waals forces include the sum of all electrostatic interactions between two

molecules having dipole moments (t,[t), a dipole and its induced effect on a polarizable

molecule ([t,a)u, and the dispersion force between all atoms due to the polarizability of the

positive nucleus with respect to the negative electrons (a,a). The energy of each of these









interactions, at the molecular level, is a function of x-6, where x is the distance of

separation, and this allows all of these energies to be grouped as van der Waals forces.

The total interaction between two infinite slabs can then be obtained by a

summation over all atom-atom interactions, if the forces can be assumed to be additive.

This summation is made for surfaces which are large relative to the atomic diameter by a

triple integration, and the summation over the second slab can be made by another

integration over the depth of the second slab. The energy is then given by



(X)- slIa-sab (1.6)



where s(x) is the potential at distance x, n is the number of atoms per cm3, C1 is an

integration constant, and A is the Hamaker constant, which incorporates all other

constants [Hamaker, 1937]. The force can then be obtained simply by calculating

dE(x)/dx. Due to geometry effects on the integration, the sphere-sphere interaction energy

is different and is given by the equation

rA
(x) = (1.7)
F(X)ssphere-sphere (1.7)
12x


where r is the sphere radius and the interaction energy is shown to vary with x-1.


1.2.3 DLVO Theory

The combination of the treatment of van der Waals forces between objects in a

condensed medium and the electrostatic repulsion due to the electrical double layer is

called DLVO theory, after its developers Derjaguin, Landau, Verwey and Overbeek

[Adamson, 1982]. It can be seen from this derivation that high ionic strength of solution









can promote agglomeration. If the condition for rapid agglomeration is taken to be that

no barrier exists and hence s(x) = 0 and dg(x)/dx = 0, then for two slabs DLVO theory

predicts that the electrolyte concentration needed for agglomeration (no) is given by


S2732 D k5T5y4 ] 1.8)
L p = --- L eA j (1.8)
S exp(4) e6A2 z6


where y is the surface free energy per cm2 of the solid-liquid interface and all other

variables are as described before [Verwey and Overbeek, 1948]. For a z-z balanced

electrolyte, equivalent conditions of concentration are in the order 100 : 1.6 : 0.13 for a 1-

1, 2-2 and 3-3 electrolyte, respectively. This dependence is quite close to the prediction

of the well-known Schulze-Hardy rule, which states that an increase in valency of one

produces an order of magnitude decrease in the concentration needed for flocculation

[Langmuir, 1938].

In one of the industries mentioned in Table 1-1, the chemical environments found

in dispersions can be particularly severe, and that is the chemical mechanical polishing

(CMP) industry. For metal polishing, strong oxidizing agents are added at high

concentrations in order to change the surface of the substrate to make polishing easier.

The oxidizing agents or other salts which are added into the polishing slurries make the

process particularly vulnerable to the issues concerning high ionic strength environments

discussed here. The next section gives an introduction to this exciting field and the many

challenges faced by researchers in this area.










1.3 Chemical Mechanical Polishing (CMP)


In the microelectronics industry, the historic goal has been to achieve increasing

complexity in a smaller size device. This goal has been achieved in recent years by the

implementation of multilevel processing. In order to fabricate high-performance

multilevel devices, planarization of the interlayer connection metals as well as the

interlayer dielectric material is essential. CMP is the preferred process by which thin

films of metals and dielectric materials for multilayer integrated circuit manufacturing are

planarized [Stiegerwald et al., 1996]. Figure 1-4 illustrates the processing steps involved

in multilevel integrated circuit manufacturing. This processing scheme is called

"damascene" processing, and it involves essentially two planarization steps, achieved by

CMP, for each processing level. These steps are illustrated in part b) showing SiO2

planarization and part c) showing metal CMP.



a) Deposition of SiO2 c) Deposition of Interconnect
on Metal Circuits Metal and Metal CMP


b) Planarization of SiO2 and
Etching of Interconnects


d) Repetition of Process
for Additional Levels


Figure 1-4. Processing steps involved in multilevel integrated circuit manufacturing.









1.3.1 CMP Process Components

The CMP process generally consists of rotating a polishing media, referred to as a

pad, against the wafer while polishing slurry is deposited between the pad and the wafer.

The pad is used to provide support against the sample surface and to carry slurry between

the sample surface and pad [Golini and Jacobs, 1991]. A schematic of the CMP process

is illustrated in Figure 1-5. This figure also includes an illustration of the interaction

between the wafer and slurry in tungsten CMP. This figure illustrates the movement of

abrasive particles across the tungsten surface, which is simultaneously undergoing

chemical conversion to tungsten oxide.

Downward Force

Slurry Feed 300 mm Wafer




Polishing
Pad Platen


Fe(CN)63- Fe(CN)64-
ST) -Fe(CN)63-
Slurry Flow 0 1 &
kh, F 4 Fe(CN)64-
W Fe(CN)63- Fe(CN)63- Fe(CN)63- Fe(CN)64


Figure 1-5.


Schematic of the CMP process, including an illustration of the interaction
between the wafer and slurry in tungsten CMP (W-CMP).


The polishing slurry provides the means by which both chemical and mechanical

action is used to remove and subsequently planarize the wafer surface. Mechanical









action is accomplished by the use of abrasive particles in the slurry. Chemical action is

achieved by the incorporation of chemical agents that aid in planarization into the slurry

[Carr, 1990]. The chemical component of the slurry, which acts isotropically on the

wafer surface, is necessary to enhance the mechanical action of the slurry. The chemical

components which may be present include oxidizing agents, completing agents,

dissolution enhancing agents, corrosion inhibitors and buffering agents. However, the

mechanical component of the slurry must also be present to preferentially abrade material

at the asperities, or high points on the surface, thus providing the driving force for

planarization [Carr, 1990].

It is both the chemical and mechanical action achieved by the slurry that achieves

global planarization of the wafer surface. The chemical component of the polishing

slurry, although crucial in achieving the required planarization, can also have detrimental

effects on the particles used in the slurry. As discussed previously, the addition of ions to

a particle slurry will decrease the electrostatic repulsion between the charged particles by

shielding the charges. This decreased electrostatic repulsion, if significant enough, can

bring about agglomeration of the particles.

The presence of large particles in the polishing slurry which contacting the wafer

during polishing has been determined to cause defects in the wafers. However, the

particles used in CMP slurries usually have a mean particle size of less than 0.5 jtm

[Golini and Jacobs, 1991], which is too small to cause the scratches observed on

defective wafers. In addition, filtration devices are used in the input lines to the polisher

which prevent particles much larger than the mean particle size from entering the

polishing apparatus. Thus, the large particles that cause the observed scratches in









defective wafers are being formed during the CMP process as the result of agglomeration

of small particles or chunks or pieces of debris produced by the polishing process.

In recent years, CMP has been applied to shallow trench isolation (STI)

technology for deep-sub-micron processes. Poly silicon (Si), CVD Si or silicon dioxide

(SiO2) can be grown or deposited in the trench and planarized by a CMP process [Wang

et al., 1998b]. The main area of current interest in the CMP industry is in improving

methodologies for inlaid metallization applications. The two materials which have

received the most attention for integration into the inlaid scheme are copper and

aluminum, both of which offer significantly lower resistivities over the conventional

metal, tungsten. The inlaid scheme offers greater challenges for CMP performance,

primarily due to the need to provide sufficiently planar surfaces for broad topographic

ranges at each metal level [Farkas et al., 1998]. Nonplanarized surface topography is a

result of the fabrication process that ends up with a deposition of the film on a previously

patterned surface, with a pattern generated by etching. Loss of planarity also arises

during lithography due to autofocus errors, residual lens aberrations, resist thickness

variations and wafer curvature associated with wafer preparations and with film stresses.

[Steigerwald et al., 1996].

As can be seen from this discussion, there are many variables which can change

the performance of the CMP process. Since there are many substrates which are either

presently used or are candidate materials for CMP processing, there are likewise many

different slurry chemistries which have been developed in order to overcome the unique

challenges that each substrate presents. Within each subclass of CMP, there are also

many strategies which have been taken in slurry formulation to attempt to enhance the









particular characteristics which are crucial to that subclass. For example, the issues of

importance in metal CMP include planarization rate, surface quality, selectivity, dishing

and erosion. The last three parameters are issues in patterned wafers, and often end up

being the critical parameters influencing slurry design. Selectivity is the ratio of the

polishing rate of the metal to that of the interlayer dielectric (ILD) material, which is

usually silicon dioxide (SiO2). Since polishing should stop once the insulating layer is

reached, a high selectivity is desired in metal CMP. Another common problem in metal

CMP is dishing, which refers to poor step coverage of deposited metal in etched trenches

and other features. This problem is illustrated in Figure 1-6, which shows a properly

polished feature with good step coverage (a), a feature which shows dishing (b) and a

feature which shows erosion of the ILD material (c). As can be seen, dishing refers to a

curved profile of the polished feature which is then translated to succeeding layers and

can cause defects or failure. Finally, erosion of the ILD material refers to the

overpolishing of insulating material near features, which results in loss of planarity near

features. Erosion is enhanced by using a slurry with low selectivity towards the metal

surface. Both dishing and erosion should be minimized by appropriate slurry design.


a) Good step coverage b) Dishing c) Erosion of ILD Material
r-- w


ID Meal~ -In im


ID Meal~ -In im


ILD etalI.I


I I II I I
Figure 1-6. Illustration of common problems associated with CMP of metal surfaces.









1.3.2 CMP Polishing Materials and Slurry Chemistries

The successful applications of the CMP process in silicon integrated circuits (IC)

has started with building multilevel (greater than 2) interconnection structures employing

deposited silica (SiO2) as the ILD, chemical vapor deposited (CVD) tungsten as the via

fill metal, and sputtered aluminum as the planar interconnection metal [Moy et al., 1989].

Thus initial process developments in the CMP industry have focused on the CMP of SiO2

and tungsten layers [Kaufman et al, 1991]. Since these developments, the use of CMP

has expanded to a large variety of materials including metals (Al, Cu, Ta, Ti, TiN, W, and

their alloys), insulators (SiO2 and doped SiO2 glasses, Si3N4, and polymers), and

polysilicon [Steigerwald et al, 1996].

Advanced metallization schemes are required to obtain the performance benefits

of scaling device dimensions into the sub-0.5 |tm regime [Steigerwald et al., 1996].

Interconnect delay is a critical parameter that determines the flow of current through the

IC, and it is defined by

12 (1.9)
RC = pE- (1.9)
td


where RC is the interconnect delay, p is the metal resistivity, ; is the permittivity and t is

the thickness of the insulator, and I and d are the length and thickness of the metal line,

respectively. The interconnect delay can be thus be decreased by decreasing p, s, or /, or

by increasing d or t [Wilson et al., 1993]. As discussed below, the use of multilevel

metallization decreases line length / and allows for interconnections to be scaled less

aggressively than the gate level. The combination of low resistivity (p) metal, low









dielectric constant (s) ILD, and multilevel metallization should yield high performance

interconnections [Steigerwald et al., 1996].

Conventional IC metallization schemes utilize aluminum alloys, in part or in full,

as the interconnection metal. While aluminum is considered a good conductor, with a

resistivity of 2.66 [DQ-cm, other metals possess even lower resistivities. Table 1-2 lists

several new candidates for IC metallization as well as metals currently utilized.




Table 1-2. Properties of low resistivity metals (from Li et al., 1994 and Wolf and
Tauber, 1986)

Ag Al Al Alloy Au Cu W

Resistivity (p2-cm) 1.59 2.66 3.5 2.35 1.67 5.65
Electromigration
esinPcat oor Poor Fair- Poor Very Good Good Very Good
Resistance (at 0.5 lm)
Corrosion Resistance Poor Good Good Excellent Poor Good
Adhesion to SiO2 Poor Good Good Poor Poor Poor
Si Deep Levels Yes No No Yes Yes No



From Table 1-2, of the metals with lower resistivity than aluminum, copper appears to be

the most attractive. Copper has a resistivity only slightly greater than silver and

approximately 50% lower than conventional Al alloys. Copper has a higher melting

point (1356 K) than aluminum (933 K) which leads to greater electromigration resistance

[Murarka, 1993]. As interconnection dimensions are scaled, the metal interconnections

are required to carry greater electron current densities. Copper is expected to handle

current densities of up to 5 x 106 A/cm2 before the onset of electromigration failure,

which is over an order of magnitude higher than the current densities which Al can

handle [Li et al., 1994].








18



According to Table 1-2, there are also several challenges with using copper metal,


including the fact that copper exhibits deep levels in the silicon band gap and copper


impurities in SiO2 lead to leakage [Sze, 1981]. However, several materials are effective


barriers to copper diffusion and their use as liner films between the copper and the silicon


or SiO2 will prevent degradation in the electronic properties of the silicon and SiO2


[Wang, 1994]. Another challenge with copper is that it is susceptible to corrosion and


therefore must be passivated. The use of corrosion inhibitors has hence been


investigated for copper CMP applications [Steigerwald et al, 1996].




Table 1-3. Slurry chemistries applied to CMP applications.


Polishing Abrasive Chemical Agents References
Substrate Particle
Potassium ferricyanide (K3Fe[CN]6), ferric
Tungsten Alumina (A1203) nitrate (Fe(NO3)3) or potassium lodate (KIO3) as [Kaufmann, 1991]
oxidizing agent

Copper Alumina Nitric acid (HNO3) with citric acid as an inhibitor [Feng, 1999]
I Alumina Nitric acid (HNO3) with 1H-benzotriazole (BTA) [Stegerwald 1994]
Alumina [Stlegerwald, 1994]
as an inhibitor
Alumina Ferric nitrate (Fe[NO3]3) with BTA as inhibitor [Luo 1998] and
S_ F[Luo, 1996]
Silica (SIO2) Hydrogen peroxide (H202) with citric acid and [Kondo, 1998]
BTA as inhibitors
An oxidizing agent such as urea H202, a [Kaufman, 1999
Alumina or other complexing agent such as ammonium oxalate and Kaufman,
or tartaric acid, and an optional surfactant 1998]
Oxidizing solution containing peroxides, amino [Prendergast
Alumina or other acids or organic amines, and a metal or metal 1999]
Compound such as Cu, Co, Fe, Pb, or Ni
One of the following a hydroxylamine,
Alumina or other ammonium persulfate, a peracetic acid or [Small, 1998]
periodic acid
V Alumina Ammonium hydroxide (NH40H) with NaCIO3 as [Luo 1997]
extra oxidizer and BTA as inhibitor
Aluminum Alumina H202 [Wrschka, 1999]
SAlumina Phosphoric acid and citric acid [Kuo, 2000]
SSilica Fluorine or other halogen and citric acid as [Feller, 1997]
chelating agent
TiN (barrier Alumina or other Urea hydrogen peroxide (oxidizer 1) and an [Kaufman 2000]
layer) organic acid (oxidizer 2)
Alumina H202with a commercial slurry [Hernandez, 1999]
An oxidizing agent and complexing agents [Wang 2000 and
Alumina or other including a phthalate compound and a dl- or trl- Wang 19
carboxylic acid 1998]
Silica (SiO2) Silica Ammonium hydroxide (NH40H) to adjust pH [Mahajan, 1999]
FLARE 2.0 Zirconium oxide
(l (ZrO2) None [Towery, 1998]
(low-K) (ZrO)












The slurry chemistries being implemented in the present CMP market have been

thoroughly reviewed and many examples are listed in Table 1-2. This table identifies the

type of surface, the abrasive particles and the chemical agents incorporated along with the

source. Note that the references are very recent, being primarily from the years 1999-

2000. The recent research efforts in each area of the CMP industry are summarized in

the following sections.


1.3.3 Tungsten CMP

Tungsten polishing (W-CMP) will be discussed first here because it is the most

"conventional' metal used in CMP processes. Conventional in this sense means it has

been around more than five years, has been researched thoroughly by CMP standards and

is being slowly phased out and replaced by more conductive metals such as copper and

aluminum. Tungsten CMP has been shown by Kaufmann et al. [1991] to proceed by

oxidation of the metal surface followed by removal of the oxide by the abrasive material.

Hence an oxidizing agent is incorporated in all tungsten CMP slurry chemistries used to

date. The slurry chemist is assisted in tungsten CMP by the fact that tungsten not only

forms a passivating oxide layer, but also that the oxide layer is a softer, easier to polish

material relative to the metal. Bielmann et al. [1999a] have verified the passive layer

formation on a W surface using A1203 and K3Fe(CN)6 slurry by electrochemical

measurements. Kneer et al. [1996] have shown that the thin oxide layer formed in W-

CMP is composed of both W02 and W03 phases. In a review of metal CMP processes,

Yu et al. [1999] report that in the period from 1994 to 1998, defectivity, which refers to

the percentage of defective wafers, and oxide erosion in W-CMP have been reduced by









factors of 20x and 5x, respectively. The colloidal stability and purity of the slurry have

also been improved by using better materials and equipment.

The isoelectric point (IEP) of alumina particles in water is near pH 8.5. However,

Osseo-Asare and Khan [1998] have shown that in the presence of tungstate ions removed

from W wafers by CMP, the IEP of alumina shifts to lower pH values. In fact, with a

tungstate ion concentration of 10-5 M, the IEP of alumina is less than 3, indicating a

negatively charged surface throughout the experimental range for W-CMP. This result

could have serious implications on colloidal stability and post-CMP cleaning. The

defectivity performance of Cabot Corporation's SEMI-SPERSE W2000 alumina based

slurry has been investigated using atomic force microscopy (AFM) by Grumbine et al.

[1998]. Although highest levels of defectivity were observed using a hard pad with high

downforce, the choice of slurry was the largest factor governing the defectivity level with

W2000 leaving a virtually undamaged oxide surface. The defectivity presumably reflects

the damage done to the surface by slurry particles during the abrasion process.


1.3.4 Copper CMP

Copper polishing (Cu-CMP) has introduced new challenges for the CMP industry

due to differences between copper and tungsten. Copper, unlike tungsten, does not form

a passivating oxide layer but rather dissolves quite readily at low pH's. As a result,

inhibitors such as 1H-benzotriazole (BTA) have been introduced to copper slurry

chemistries in order to control corrosion by forming a protective and stable film which

can withstand chemical and thermal environments [Brusic, 1991]. Concentrations of

BTA as low as 0.01 M have been found to reduce the corrosion rate of copper to near

zero while not reducing the polishing rate significantly [Wang et al., 1997].









Another method that has been introduced to combat this problem has been to use

less harsh chemical environments that are still capable of achieving an acceptable

removal rate. The passivation regime in which copper will form an oxide layer is in the

region of pH greater than 6. In this regime, oxidizing agents are added to enhance

oxidation of the substrate and completing agents may be added to enhance removal of

abraded copper. The completing agents are necessary because copper removed during

the polishing process is often redeposited on the metal surface, which can cause

significant defects due to lack of planarity. The use of slurries above pH 6, however, are

problematic in that the selectivity to SiO2 polishing is rather low [Carpio et al., 1995].

This is not a problem in tungsten CMP since pH's lower than 4 combine both passivation

chemistry due to the formation of an oxide layer and a high selectivity towards the SiO2

insulating layer. Another problem with copper CMP in the oxidation regime is that the

oxide layer, which may consist of CuO, Cu2O, Cu(OH)2 or combinations thereof, is

usually porous and may not provide sufficient protection of the metallic copper from the

chemical species. As a result, inhibiting layers that prevent corrosion are still added in

the oxidation regime, as in the dissolution regime [Luo, 1997]. These inhibiting layers

slow the dissolution rate of copper and allow the CMP process to control the polishing

rate, rather than the dissolution rate.

Another challenge facing formulators is that copper is a softer material than

tungsten, creating the need for using either softer abrasives or smaller, more uniform

particle sizes which reduce scratching for copper CMP [Carpio et al., 1995]. Kondo et al.

[1998] has proposed a model which shows that Cu-CMP proceeds according to Cu

surface oxidization, oxidized layer protection by an inhibitor, polishing of the protection









layer on top of protrusions by fine abrasive, and etching of the oxide by an acidic media.

This model is similar to the passivation model proposed for tungsten CMP by Kaufman et

al. [1991], except that the copper model includes an inhibiting agent and the ability of the

slurry to etch the oxide. The proposed silica-based polishing slurry yields a removal rate

of 150 nm/min with reduced scratches. However, a large ratio of the CMP removal rate

to the etching rate is required for reducing Cu dishing [Kondo et al., 1998]. Luo et al.

[1998] has shown for Cu-CMP that the polish rate increases linearly with increasing

downward pressure and rotational speed, but the Preston equation needs to be modified to

represent the data. The inclusion of an additional velocity term representing the greater

dependence of removal rate on the velocity, and a constant representing the purely

chemical reactivity of the slurry, provides a satisfactory model. Although the challenges

of Cu-CMP are difficult, it has been established in recent years as a viable approach for

Cu metallization technology [Yu et al., 1999].


1.3.5 Aluminum CMP

Aluminum polishing (Al-CMP) is proposed to proceed much like tungsten

polishing, with oxidizing agents added to the slurry to yield a surface passivated oxide

layer which is removed stepwise [Wang et al., 1998c]. Hernandez et al. [1999] showed

using scanning electron microscopy (SEM) and X-ray photoelectron spectroscopy (XPS)

that the removal rate depends critically on the surface saturation of the pad with A1203

abrasive particles. Wrschka et al. [1999] showed for a HO2 oxidizing agent that the

oxidizing agent concentration only has a weak effect on the removal rate of Al. The

Preston equation is the most frequently referenced expression for polishing rate, and it

predicts that the polishing rate is directly proportional to the pressure and the linear









velocity of the pad relative to the wafer [Preston, 1927; Brown et al., 1981]. The study

by Wrschka et al. [1999] fails to describe the dependence of the removal rate on pressure

and velocity, and a power law function is proposed instead. An Al-CMP technology has

been reported by Yu et al. [1999] to yield good defectivity and excellent patterned wafer

planarity. Wang et al. [1998c] have shown using Al and alloys with Cu and Si that alloy

content and grain size have an effect on removal rate. As can be seen, Al-CMP

technology is more established and seems to present less challenges than Cu-CMP. Even

though copper CMP is a more difficult process than aluminum CMP, the lower resistivity

of copper versus aluminum makes up for the difficulties in polishing.


1.3.6 CMP of Other Metals

Polishing of a TiN barrier layer should proceed at similar rates to the primary

metal, so a selectivity of close to 1 is desired in this case. Hernandez et al. [1999] found

that the incorporation of hydrogen peroxide (H202) into a commercial slurry for Al-CMP

dramatically increased the dissolution of TiN, as desired. Many of the slurries being

investigated for copper CMP are also useful for polishing of TiN and other barrier

materials.

Many investigations have also attempted to explain and improve polishing

through the use of electrochemical measurements. For example, the role of metal

concentration gradient in CMP has been examine by Osseo-Asare [1998] with the aid of

pH-potential diagrams, also known as Pourbaix diagrams, which indicate the various

phases (reactions and reaction products) that are stable in an aqueous solution at

equilibrium [Pourbaix, 1975]. Process conditions have been identified that favor









mechanical polishing, etching, or polishing that is both chemical and mechanical for Cu-

CMP, W-CMP and titanium CMP.


1.3.7 CMP of Silica (SiO2) and low-K Dielectric Materials

As illustrated in Figure 1-4, CMP of the interlayer dielectric material is another

important application of CMP. The conventional dielectric material is silica, which is

usually polished with silica particles at a high pH in order Io enhance silica dissolution

[Mahajan et al., 1999]. This slurry chemistry does not yield a high ionic strength

environment, and hence the instability shown in Figure 1-2 is not expected to occur in

this system. This is because silica has sufficient negative charge at high pH to promote

stability in these low ionic strength environments.



Table 1-4. Dielectric constants of selected candidate low-K ILD materials.

Dielectric
Material
Material Constant (x)

SiO2 3.9 6.0
SiO2 F, SiO2 B 3.0 -3.9
Polyimides 2.9 3.9
Fluorinated Polyimides 2.3 2.8
Parylenes 2.3 2.7
Fluoro-polymers 1.8- 2.2
Teflon-AF 1.9
Micro-porous Polymers, Aero-gels 1.3 1.7





Motivated by equation (1.9), strategies to reduce capacitance effects associated

with shrinking design parameters include incorporating not only low resistivity metals

such as copper but also insulators with low dielectric constants, or low-K materials.









Several inorganic and organic materials are currently being investigated as low-K

materials [Pai and Ting, 1989]. Conventional Si02 deposited by CVD has a K of between

3.9 and 6.0 depending on the H20 concentration in the SiO2 [Ting, 1994]. By doping the

SiO2 with fluorine or boron and lowering the H20 content, K may be reduced to between

3.0 and 3.9. Organic materials such as those listed in Table 1-4 exhibit even lower

dielectric constants and are excellent candidates for low-K ILD's [Steigerwald et al.,

1996].

It follows that CMP of novel low-K ILD materials has been investigated as a

processing issue. Some of the recent research in this area is highlighted in the following

discussion. FLARE 2.0, a poly(arylene) ether from AlliedSignal, Inc., with a K of 2.8, is

a candidate low-K ILD material [Towery and Fury, 1998]. Other FLARE polymers have

been investigated by Hendricks and Lau [1996] and have been shown to have sufficiently

high thermal and thermomechanical stability to withstand current process temperatures

associated with metal deposition and annealing steps. CMP of FLARE films yield

uniform, defect-free and scratch-free surfaces. A recent attempt to develop a CMP

process for stacked low-K CVD oxide films has been investigated by Hartmannsgruber et

al. [2000]. The dielectric layer is composed of a low-K, methyl-doped Si02 layer

(tradename Flowfill) embedded between thin SiO2 layers. A higher removal rate of the

low-K Flowfill layer in comparison to that of the cap (SiO2) layer results in a significant

increase in the degree of planarization. It also leads to a reduction in the polishing time

required to achieve planarization of the entire wafer once the first area of the wafer is

planarized, also called the overpolish time. Another low-K material which has been









successfully planarized with a commercial CMP slurry is the spin-on SiLK resin

[Kuchenmeister et al., 2000].

A study by Zhang et al. [1998a] identifies the mechanism in polymer polishing as

mostly mechanical, not chemical in nature. As a result, the slurry particle size plays the

major role in controlling the polish rate. Dishing in patterned wafers is minimized, as

expected, by use of a hard pad surface. Hard pads do not conform to the topography on

the wafer surface, hence reducing dishing effects. Another class of polymers with

potential application in low-K dielectric applications is the benzocyclobutenes (BCB's).

Neirynck et al. [1996] have investigated the polishing of a BCB using surfactant additives

to enhance polymer dissolution, with other additives including HNO3 and BTA. It is

proposed that the surfactant improves the polymer wettability by making the surface

hydrophilic, hence yielding a higher dissolution rate. Other polymers which have been

investigated as ILD materials with copper interconnects include divinylsiloxane and

parylene-n [Price et al., 1997].

From the preceding discussion, it is evident that there are many issues in CMP

which are yet to be fully addressed. Among these issues is dispersion stability, which

may have a significant effect on slurry performance. The present investigation focuses

on the issues concerning dispersion stability in model systems for CMP of metal surfaces.


1.4 Surfactants and Polymers as Stabilizing Agents


The work presented here focuses on dispersion stabilization and its application in

CMP, so it is important to discuss the general approaches taken to stabilize dispersions.

The two primary classes of molecules which are applied as stabilizing agents are








surfactants and polymers. These two classes of molecules differ, sometimes

significantly, in the way in which they bring about stabilization. To illustrate this point,

the use of surfactants and polymers as stabilizing agents is shown schematically in Figure

1-7.


a) Surfactants


Ionic


Nonionic


b) Polymers


Figure 1-7. Illustration of surfactants and polymers as stabilizing agents on
hydrophilic particles.


Surfactants can be divided into two general types: ionic and nonionic. This

designation refers to the charge on the hydrophilic head group of the surfactant. Ionic

surfactants can be further categorized into cationic, anionic and zwitterionic types, with

the latter referring to a molecule that has both a positive and a negative charge present on

the head group. The most widely used anionic groups are sulfate (-S04), sulfonate (-









SO3-) and carboxylate (-CO2-), while the most widely used cationic groups are the

quaternary ammonium groups (-NR3), where R can be -CH3, -(CH2)nCH3, or a variety

of other alkyl groups.


1.4.1 Ionic Surfactants

As can be seen from Figure 1-7, ionic surfactants are generally used as

electrostatic stabilizing agents. By adsorbing ionic surfactant onto an oppositely charged

surface in sufficient quantity that a bilayer of molecules is formed due to hydrocarbon

chain interactions between adjacent molecules, a layer of charged hydrophilic groups will

be extending into the aqueous solution. Since these charges are opposite in sign to the

surface charge, this procedure has the effect of reversing the sign of the electrical

potential at some distance prior to the second layer of adsorbed ionic head groups. As a

result, the long range electrostatic forces between adjacent particles will be repulsive due

to the repulsion of like charges opposite in sign from the surface charge. Due to the

significant number of ionic surfactant molecules which can be adsorbed on a particle in a

bilayer configuration, which is limited only by the area per molecule of a tightly packed

layer of the given molecule, the apparent charge of the particle can be made significant

enough to produce a stable dispersion. This is true even if the initial surface charge is

insufficient to electrostatically stabilize the particles, and therefore ionic surfactants can

be used to stabilize particles of little or no surface charge.


1.4.2 Nonionic Surfactants

Nonionic surfactants have no charge associated with them. As a result, DLVO

theory predicts that nonionic surfactants will not provide a significant repulsive barrier to

agglomeration. However, nonionic surfactants have been successfully implemented as









stabilizing agents in a number of applications. This is because nonionic surfactants are a

class of molecules that may act as steric stabilizing agents, which is a type of stabilization

not considered by DLVO theory. Nonionic surfactants that are used as stabilizing agents

are typically polymeric in nature, meaning that the hydrophilic head group of the

surfactant is a hydrophilic polymer chain, and therefore they act to provide steric

stabilization in the same fashion as pure polymer molecules, which will be considered in

the following section.

Nonionic surfactants differ from polymers in that, in Figure 1-7, they must form a

bilayer if adsorbed on a hydrophilic surface in an aqueous solution, similar to ionic

surfactants. Nonionic surfactant molecules offer an advantage over polymers in that they

are amphiphilic surfactant molecules and typically have a higher adsorption potential

than an equal chain length polymer. The tendency of surfactant molecules to adsorb at

interfaces will be discussed in more detail in the next section. The discussion of steric

stabilization naturally leads to the discussion of polymers as dispersion stabilizers.


1.4.3 Polymers

Polymers can be used as stabilizing agents for dispersions as well, with the type

of stabilization mechanism being steric in nature, particularly for uncharged polymers.

Steric stabilization occurs due to the presence of physical barriers adsorbed on particles

that prevent the particles from coming close enough to allow the van der Waals attractive

forces between particles to dominate [Napper, 1970; Smitham and Napper, 1979]. The

van der Waals attractive potential between two spheres has been given in Equation (1.7),

and can be seen to decrease with the inverse of the separation distance. Polymers can be

effective stabilizing agents only if the polymer solvent interactions are more favorable









than the polymer polymer interactions [Rosen, 1989]. These interactions determine the

conformation of the polymer once adsorbed on a surface, and hence determine the

minimum separation distance which can be achieved through the adsorbed layers.

The key to steric stabilization is to increase the closest distance of approach

between particles in order to minimize van der Waals attractive forces [Tadros, 1986; Lee

et al., 1986]. Without steric stabilizing agents, the closest distance of approach is limited

only by the weak electrostatic forces, so the particles easily enter the strong potential well

near the surface. With stabilizing agents, the attractive potential is greatly reduced at the

closest distance of separation, making the barrier to agglomeration much easier to

manipulate. A key modification to DLVO theory has been the estimation of so-called

steric forces which come into play in a system where steric stabilization dominates.

These forces then supply the repulsive barrier that prevents particle agglomeration in

such sterically stabilized systems. There are at least three advantages that steric

stabilization possesses over electrostatic stabilization: Its relative insensitivity to high

concentrations of electrolytes; the fact that high solids dispersions display relatively low

viscosities; and it is equally effective in both aqueous and nonaqueous dispersion media

[Napper, 1983].


1.4.4 Applications as Stabilizing Agents

There have been numerous investigations in previous decades devoted to

stabilization of particles using either surfactants or polymers. The following discussion

serves to give an example of some of the research in this area, although a more thorough

review is focused on the study of mixtures of surfactants as stabilizing agents in the next

chapter. Numerous investigations, as discussed below, have focused on adsorption of









ionic surfactants on oppositely charged particle surfaces. Such adsorption has been

studied extensively by Somasundaran and Feurstenau [1966], Gaudin and Feurstenau

[1955] and Somasundaran and Kunjappu [1989], primarily for the system of sodium

dodecyl sulfate (SDS) on alumina particles. The stability of silver iodide and iron (III)

oxide dispersions has been found by Gotoh et al. [1998] to be dependent on both the

concentration and hydrocarbon chain length of monoamino-monocarboxylic acid, a

zwitterionic dispersing agent. Catalytic colloidal rhodium particles have been stabilized

using a water-soluble surfactant by Larpent et al. [1991]. The stabilizing mechanism of

calcite dispersions in the presence of sodium polyacrylate has been found by Rogan et al.

[1994] to change from electrostatic to steric depending on the dispersant concentration.

Steric stabilization by nonionic polymers at high concentrations for silica (SiO2) systems

has been observed by Giordano-Palmino et al. [1994]. A recent paper by Moudgil and

Prakash [1998] discusses the application of polymers and surfactants as agents for

selective flocculation and the parameters which influence adsorption. Another recent

paper by Sjoberg et al. [1999] discusses adsorption of surfactants and polymers on

stability and rheology of concentrated kaolin dispersions. As discussed, both ionic and

nonionic surfactants can be effective stabilizing agents for different types of dispersions,

and this class of molecules will be the subject of the remainder of this investigation. As a

result, an introductory discussion into this unique class of molecules is imperative.


1.5 Surfactants in Solution


Surfactant molecules are amphiphilic in nature, meaning they have both a

hydrophilic and a hydrophobic component. This characteristic is responsible for making

surfactants "surface-active", and often they are referred to as surface-active agents. This








surface activity is derived from thermodynamic considerations of the minimum free

energy locations for the molecules. When placed in solution, whether polar or nonpolar,

a surfactant will have a positive interaction between one of its sides and the solvent, and a

negative interaction between the other side and the solvent. Although these individually

solubilized surfactant molecules, termed monomers, are present in all surfactant

solutions, they will quickly adsorb at lower free energy locations such as interfaces. At

interfaces where a difference in hydrophobicity occurs on the two sides of the interface,

surfactant molecules will adsorb with the hydrophilic side toward the more hydrophilic

side of the interface. Interfaces which may have adsorbed surfactant molecules from

solution include the gas-liquid interface (i.e. bubbles), the liquid-liquid interface (i.e.

emulsions) and the solid-liquid interface (i.e. surfaces of particles or container walls).





000 000000000000000 00

4 Hydrophobic Q
Particles






C > CMC >
OB <--\- -r
,- More "-
Hydrophilic \ Mr
Particles Surf

Figure 1-8. Illustration of the preferential locations of surfactant molecules in solution.








The ability of surfactant molecules to adsorb at all of these interfaces is illustrated in

Figure 1-8.


1.5.1 Monolayers at the Air/Liquid Interface

The adsorption of surfactant monomers in a monolayer configuration at the

air/water interface, shown in Figure 1-8 as location A, has been extensively studied. The

adsorption of surfactant decreases the surface tension (y), also called the surface free

energy, due to orientation of the surfactant molecules in an entropically favored

configuration with hydrophilic head group towards the water and hydrophobic tail

towards the air [Adamson, 1982]. The thermodynamic treatment of the variation of

surface tension with interface composition has been given by Gibbs [1931], with further

amplification by Guggenheim and Adam [1933]. The fundamental Gibbs equation,

which relates the surface excess of the surface-active species in solution, F2, to the

surface tension y is given by


a dy)
F =2 = a (1.10)
RT da


where a is the activity of the solute, which is equal in dilute solutions to its concentration,

R is the gas constant and T is the temperature. It is important to note that the adsorbed

molecules of surfactant remain in equilibrium with surfactant monomers and other

species in solution.









1.5.2 Micelles

Above a concentration called the critical micelle concentration (CMC), surfactant

monomers will self-aggregate into structures called micelles (location B in Figure 1-8),

first observed by McBain [1913]. In aqueous solution, the micelles consist of a

hydrophobic core with surfactant tails associated through hydrophobic bonding, with

head groups facing outward into solution. This preferred arrangement yields a lower free

energy system and has been described thermodynamically as exhibiting phase

equilibrium behavior [Hiemenz and Rajagopalan, 1997]. By this analysis, some

minimum value of concentration n is necessary before the exclusion of hydrophobic tails

from the aqueous medium is effective, and once n is sufficient to form micelles, any

additional surfactant added to the solution goes into the micelles. A review of

experimental methods for determining CMC as well as the kinetic aspects of micellar

breakup and equilibrium processes has been published by Patist [1999].






Fast relaxation time, microseconds

11-0 al-




2
Slow relaxation time, milliseconds to minutes


Figure 1-9. Mechanisms for the two relaxation times, TI and c2, involved in a
surfactant solution above CMC.









Micelles are dynamic structures that have two relaxation processes associated

with them. First, there is a fast relaxation process referred to as TI, which is associated

with the fast exchange of monomers between micelles and the surrounding bulk phase.

Second, there is a slow relaxation process referred to as c2 which is attributed to the

micelle formation and dissolution process [Patist, 1999]. Figure 1-9 shows the two

characteristic relaxation times TI and c2 associated with micellar solutions (from Patist,

1999).

Micellar relaxation kinetics show dependence on temperature, pressure and

concentration and have been studied by various techniques such as stopped-flow,

temperature-jump, pressure-jump and ultrasonic absorption [James et al., 1977; Hoffman

et al., 1976; Kato et al., 1995; Tondre et al., 1975]. The stability of micelles, as

determined by measurement of c2, has been found to correlate with a number of

technological processes, including foamability, wetting time of textiles, bubble volume,

emulsion droplet size and solubilization rate of benzene [Shah, 1998]. The micellar

stability of sodium dodecyl sulfate (SDS) solutions has been determined by Lessner et al.

[1981] and Oh et al. [1993] using pressure-jump with electrical conductivity detection.

This technique is described in detail by Huibers et al. [1996]. The maximum stability of

micelles has been found to correlate with lower foamability, larger single bubble

volumes, minimum frequency of bubble generation, and maximum single film stability

[Oh and Shah, 1991; Oh et al., 1992; Patel et al., 1996]. The correlation of micellar

stability, c2, to dispersion stability is an important consideration, and its implications are

discussed in the recommendations for future research in Chapter 7.









1.5.3 Adsorption of Surfactant on Particles

As can be seen in Figure 1-8, surfactant molecules will adsorb at interfaces

present in solution, particularly if there is a driving force (e.g. hydrophobic forces or

electrostatic forces) of some type to put the surfactant at the interface. This driving force

is generally the difference in hydrophobicity across the interface, as in the air-water

interface, where surfactants can orient in a way which minimizes the free energy of the

system. In aqueous solutions, surfactants will adsorb due to this driving force on

hydrophobic particles with the surfactant tail towards the particle (location C in Figure 1-

8). This configuration, which makes the particle more hydrophilic, can be used to

stabilize hydrophobic particles which would otherwise agglomerate.

On the other hand, many types of particles are too hydrophilic to promote this

type of driving force. In this case, the hydrophilic head group of the surfactant can

adsorb on the particles if there is an electrostatic attraction caused by opposite charges.

In this case, a single layer of surfactant will make the particles hydrophobic, which will

promote agglomeration (location D in Figure 1-8). However, hydrophobic bonding can

attract more surfactant to a fully covered interface, causing a second layer of surfactant to

adsorb with tails towards the surface. By this configuration, a bilayer of surfactant

molecules minimizes the free energy of the system by excluding a significant number of

hydrophobic tails in contact with water from the solution as well as from the surface

(location E in Figure 1-8).

The concentration at which surfactant aggregates begin to form on a particle

surface, deemed the critical aggregate concentration (CAC), is often orders of magnitude

lower than the bulk counterpart, the CMC. These aggregates have been termed

hemimicelles [Gaudin and Fuerstenau, 1955], admicelles [Cases et. al., 1985] and









solloids [Somasundaran and Kunjappu, 1989], and they may differ significantly in nature

from their bulk counterparts. Figure 1-10, from Somasundaran and Krishnakumar [1997]

illustrates the various stages of aggregate growth of surfactant on a particle surface. In

this example, an anionic surfactant (sodium dodecyl sulfonate) is adsorbing onto a

positively charged alumina surface, with the adsorption density as a function of surfactant

concentration shown as the solid line. Similar isotherms have been reported in literature

for ionic surfactants adsorbing on oppositely charged surfaces.


+++ .+
54"4


+ ++ L-


G 9


Figure 1-10.


Stages of aggregation of adsorbed ionic surfactant on oppositely charged
particle surfaces, showing reverse orientation model (left) and bilayer
model (right) (from [Somasundaran and Krishnakumar, 1997]).


The four stages illustrated are represented on the solid line as stages I, II, III and

IV, and the corresponding state of aggregation is shown schematically for both the

reverse orientation model and the bilayer model. The CAC on this figure is at the


Ad;&









boundary between regions I and II. As shown, the aggregates grow in size and number

rapidly in region II, then the growth slows in region III as the surface approaches

saturation. Region IV then represents a full bilayer of adsorbed surfactant. This type of

aggregation of surfactants on surfaces has been reported as well for nonionic surfactant

adsorption, although concentrations are much lower for nonionic surfactants due to a lack

of electrostatic repulsion between head groups [Giordano-Palmino et al., 1994].


1.6 Rationale of the Proposed Research


This investigation focuses on the issues affecting dispersion stability in severe

environments. From preceding discussions, it can be seen that this is an issue of great

interest in current industry. The use of surfactants and polymers as stabilizing agents has

been introduced in this chapter, and the use of mixtures of these surfactants as stabilizing

agents will be considered in Chapter 2. The mechanism of stabilization is discussed in

this chapter in detail, along with literature that supports the molecular mechanism.

Chapter 3 then goes beyond the current knowledge base and investigates the effects of

various molecular factors on dispersion stability using the proposed mechanism. The

effects of surfactant choice, concentration, and the ratio of surfactants in the mixture are

all considered in this chapter. In order to verify that the stabilization mechanism and

some of the nuances in the data are justified, the adsorption of surfactant is determined

for various mixed systems in severe environments in Chapter 4. This determination

involves the use of novel techniques for high ionic strength environments.

The remainder of this work is intended to assess the relevance of the stabilization

mechanism presented to the primary application for which the model dispersions are

intended, CMP. Since abrasive particle size is known to be a major factor governing









many of the parameters in the CMP process, Chapter 5 is focused on correlating the

dispersion stability to the particle size of the slurry particles. This chapter includes a

comparison of available techniques for particle sizing in severe environments as well as

verification of the results for many different systems. Finally, the effect of dispersion

stabilization using mixed surfactant systems on CMP polishing performance is

ascertained in Chapter 6. These results determine if the stability of the slurry influences

the important output parameters such as polishing rate, surface quality and particulate

contamination of the polished wafer. The implications of the entire investigation are then

considered in Chapter 7, along with suggestions for future directions to continue the

growth of knowledge in the dispersion area.














CHAPTER 2
STABILIZATION OF HIGH IONIC STRENGTH DISPERSIONS USING THE
SYNERGISTIC BEHAVIOR OF A MIXED SURFACTANT SYSTEM



2.1 Synergistic Behavior of Surfactant Mixtures


As discussed in Chapter 1, surfactants can stabilize dispersions through either

electrostatic or steric effects or some combination of both. In some instances, mixtures of

surfactants can show some synergistic behavior which may be beneficial for stabilization

purposes, particularly in cases where single surfactants do not perform well. This

synergistic behavior is typically brought about by the interaction between different types

of surfactant molecules, which provides a driving force for the mixtures to form mixed

aggregates or structures, as in mixed monolayers, mixed bilayers or mixed micelles.

These mixed structures can often have a more fluid-like or solid-like behavior than either

of their single surfactant aggregates.

Mixed surfactant systems and mixtures of surfactants and polymers have been

investigated as stabilizing agents for dispersions by a number of investigators, discussed

below. Often the synergistic effect of the mixed systems has been shown to be due to the

enhanced adsorption from the mixture of a weakly adsorbing component. For example,

sodium p-octylbenzenesulfonate adsorbs three orders of magnitude more than CQ2E on

alumina. However, when adsorption is conducted from a 1:1 mixture of these surfactants

the ethoxylated alcohol has been observed by Somasundaran et al. [1992] to adsorb to a

greater extent than the sulfonate. The explanation for this phenomenon is that the
Reproduced in part with permission from Journal of Colloid and Interface Science 223, 102 (2000).
Copyright 2000 Academic Press.
40









adsorption of the sulfonate provides a sufficient number of hydrophobic sites for the

solloid-type adsorption of the ethoxylated alcohol. Similar results have been observed

for SDS and C12EOs [Somasundaran et al., 1992].


Another example of this phenomenon has been observed in the case of adsorption

of tetradecyl trimethyl ammonium chloride (TTAC) cationic surfactant and

pentadecylethoxylated nonyl phenol (NP-15) nonionic surfactant on alumina. The NP-15

does not adsorb on alumina by itself, but in the mixture with strongly adsorbing TTAC,

the NP-15 does adsorb significantly and also enhances TTAC adsorption [Huang et al.,

1996]. In Na-kaolinite dispersions, nonionic surfactants adsorb to a greater extent than

ionic surfactants. The presence of nonionic surfactant in a mixture with either anionic or

cationic surfactant has been found by Xu et al. [1991] to greatly enhance the adsorption

of the ionic surfactant. Hydrocarbon chain interaction between adsorbed nonionic

surfactant and ionic surfactant in solution has been proposed to be responsible for the

enhanced adsorption. As seen from these examples, the enhanced adsorption of a weakly

adsorbing component is due to its ability to adsorb in a reverse orientation. The strong

hydrophobic chain interactions are then the driving force behind the adsorption of that

component.


Different effects on dispersion stability have been observed with adsorption of

mixtures of a surfactant and a polymer on particles. The adsorption of polyethylene

oxide (PEO) and cationic surfactant dodecyltrimethylammonium bromide (DTAB) on

silica has been shown by Esumi et al. [1998] to exhibit competitive adsorption between

the two species. The amount of PEO adsorbed decreases with increasing surfactant

concentration, while that of the surfactant increases with surfactant concentration but was









lower in the presence of PEO than in the absence of PEO. Similar competitive adsorption

effects have been observed by Esumi and Matsui [1993] using mixtures of

polyvinylpyrrolidone (PVP) and poly(dimethyldiallylammonium chloride) (PDC)

adsorbed on silica. The adsorption of PVP is almost unchanged with increasing

concentration of PVP, while the adsorption of PDC decreases with increasing PVP

concentration. The results have been correlated to adsorbed polymer conformation on the

particle. Simultaneous adsorption of PVP and either lithium dodecyl sulfate (LiDS) or

lithium perfluorooctanesulfonate (LiFOS) anionic surfactant on alumina has been studied

by Otsuka and Esumi [1994]. In the binary PVP-LiDS system, the adsorption of PVP

increases remarkably due to the presence of LiDS at low LiDS concentration, followed

by a decrease at high LiDS concentration. A similar trend is observed in the PVP-LiFOS

system. The adsorption results have been correlated with dispersion stability for this

system. Finally, the adsorption of hydroxyethylcellulose (HEC) or hydrophobically

modified HEC (HMHEC) together with SDS on alumina has been investigated by

Yamanaka and Esumi [1997]. The adsorption of HEC increases due to the presence of

SDS, while the amount of adsorbed SDS is unaffected by the presence of HEC. Similar

results have been obtained for the SDS-HMHEC system.


As can be seen from these investigations, mixtures of surfactants and polymers

can significantly enhance or reduce adsorption of either component. However, the ability

of a carefully chosen ionic and nonionic surfactant mixture to synergistically stabilize

dispersions has been demonstrated in the discussion above. The application of ionic and

nonionic surfactant mixtures to stabilization of model slurries for CMP will be









investigated in this chapter after a discussion of single surfactant adsorption in severe

environments.


2.2 Methods for Sedimentation Experiments


Sedimentation experiments are used in this investigation and in the following

chapter to investigate dispersion stability. The sedimentation method involves preparing

a dispersion and allowing it to sediment undisturbed under gravity in a long vertical

column such as a graduated cylinder (e.g. 100 mL volume) while monitoring the height

of the dispersed phase. Both the dispersion preparation method and the monitoring

procedure are variables which are addressed in the following section. This method has

been used conventionally as one of the primary means of determining stability in systems

where destabilization occurs on macroscopic time scales and within a reasonable amount

of time for measurement (e.g. in 1 hour up to 30 days). The advantages of using

sedimentation over alternative techniques include the ability to directly observe the entire

sedimenting layer, which allows the experimentalist to observe any unusual changes in

sedimentation behavior as a whole. The other advantage of sedimentation is that

complete data is taken as a function of time, thus allowing any unusual kinetic behavior

to be observed. One can also visually determine the volume of sedimented particles at

the bottom of the cylinder.

A complementary technique for observing dispersion stability is centrifugation, in

which the dispersion is allowed to sediment under centrifugal forces greater than gravity

[Coelfen, 1998; Gafford, 1985]. The basic concepts of this technique are the same and

allow for faster observation of dispersions which settle very slowly under gravity, but the

technique does not allow for observation of unusual settling behavior during the









sedimentation process. Other methods for determining dispersion stability involve light

transmittance measurements [Hyde, 1978]. These techniques allow for exact

quantification of stability data through the use of a turbidity parameter, yet the data is

taken at only one place and at one time during the experiment. For turbidity data to be as

useful as a complete set of sedimentation data, the data must be taken at points located

throughout the holding container and as a function of time.


2.2.1 Method of Dispersion Preparation

For the following investigation, all materials were used without further

purification. The slurry particles used in this study were high purity AKP-50 oa-alumina

particles with particle size of 100-300 nm obtained from Sumitomo Chemical Co., Ltd.

Potassium ferricyanide and iron (III) nitrate were investigated as oxidizing agents, and

they were obtained from Fisher, Inc. Surfactants used were from various suppliers.

Sodium dodecyl sulfate (SDS), cetyl pyridinium chloride (CPC), sodium salt of capric

acid and sodium salt of lauric acid were obtained from Sigma Chemical Co.

Hexadecyltrimethylammonium bromide (C16TAB) was obtained from Fisher, Inc.

Dodecyltrimethylammonium bromide (C12TAB) was obtained from Acros, Inc. Triton

X-100 was obtained from Aldrich, Inc. Tween 20, Tween 40, Tween 60, Tween 80 and

Symperonic A4, A7, All and A20 were obtained from ICI Surfactants. Decyl sodium

sulfate and hexadecyl sodium sulfate were obtained from Eastman Kodak Co. Ultrapure

water was obtained from a Milli-Q Gradient A10 filtration system supplied by Millipore

Corp.

Dispersions were prepared by first dissolving surfactants cationicc, anionic, or

nonionic) or mixtures of surfactants in the desired concentrations in water. Alumina









particles in a concentration of 10 wt. % were then added into the surfactant solution and

the pH was adjusted to a value less than 4. In order to attempt to break up agglomerates

that form either in the particle preparation process or when dry particles are stored in air

(in plastic containers) for long periods of time, the dispersions were then sonicated for 30

minutes. The oxidizing agent was then added and the pH was carefully readjusted to a

value between 3.5 and 4.0. This pH range was used in order to model CMP slurries used

for tungsten polishing, which have been shown to have a high selectivity of tungsten over

silicon dioxide in this range [Kaufmann, 1991].


2.2.2 Method of Sedimentation Characterization

Dispersion stability was then characterized by sedimentation testing.

Sedimentation experiments were performed by thoroughly dispersing the slurries prior to

pouring them into 100 mL graduated cylinders, where they were then undisturbed

throughout the length of the experiment. Room temperature was maintained at 25+10C

throughout the experiment time. The amount of settled slurry as well as the sediment

height was recorded over time and the results converted to fractional volume to normalize

the results. The amount of settled slurry was determined as the sum of 1) the amount of

clear liquid on top and 2) a fraction of the amount of cloudy layer below the clear liquid.

The cloudy layer was the transition region between fully dispersed slurry and clear liquid.

Examples of this calculation procedure for total settled volume are given in Figure 2-1 in

order to make the procedure more clear. The fractions used for conversion of the cloudy

layer to the clear liquid equivalent are given in Table 2-1. These fractions (f) are used to

calculate the total settled volume in the examples in Figure 2-1.










20 20 C 20
mL n mL L q mL
f = m Slightly
20 Hazy
mL Slurry
40 Cloudy (0.1 0.3 AU)
Slurry f = 0.8
mL
Fully 1 2 AU)
Disperse f= 0.3 40 Cloudy
40 Cloudy
Slurry
Slurry mL Slurry
f = 1 Fully (2-3AU)
Disperse f = 0.1
Slurr



Total Settled Total Settled Total Settled
Volume = 20 mL Volume = 32 mL Volume = 40 mL

Figure 2-1. Examples of sedimentation experiments along with the total settled
volume determined using the procedure outlined below (f = fraction
settled as defined in Table 2-1).



In Figure 2-1, the slurry on the left is an example of the simplest type of settling

to quantify. In this figure, 20 mL of clear liquid is on top of slurry that is fully dispersed,

making the total settled volume 20 mL. This type of characterization procedure is

commonly used to obtain a single parameter for settling called the initial settling rate,

which is the rate of settling extrapolated to time 0. As can be seen by the other example

slurries in Figure 2-1, however, this method does not always give a clear picture of the

settling results. The slurries in the middle and on the right in Figure 2-1 are clearly less

stable than the slurry on the left, and yet the above method would give the same settled

amount for each of these slurries. In response to this problem with the characterization

method, the fractions used to calculate total settled volume for the cloudy to hazy layers

have been developed in Table 2-1.









Table 2-1. Fractions used for determination of total settled volume based on degree
of cloudiness in sedimentation experiments.


Classification Range of Absorbance Values Fraction Settled
Very cloudy 2.0 3.0 AU 0.1
Cloudy 1.0 2.0 AU 0.3
Hazy 0.3 1.0 AU 0.6
Slightly hazy 0.1 0.3 AU 0.8


The slurry in the middle in Figure 2-1 has the same level of clear liquid (20 mL)

as the slurry on the left, but below this clear layer is a layer of slurry that is cloudy but

not fully dispersed. In order to quantify the degree of opacity in this type of situation, the

scale in Table 2-1 for the different levels of opacity has been developed. The different

levels are arbitrarily assigned (there easily could be 10 different levels instead of the 4

levels indicated) and the fractions used are also arbitrary, although they indicate the

degree of settling based on experience with sedimentation results. In order to calculate

the total settled volume for the slurry in the middle of Figure 2-1, the fraction for the

cloudy layer (f=0.3) is multiplied by the volume of that layer (40 mL). This makes the

total settled volume for this slurry 20 + 40(0.3) = 32 mL. For the slurry on the right, the

slightly hazy layer under the clear liquid layer is assigned a fraction of 0.8 to take into

account the fact that the slurry is almost fully settled in this region. The region below the

slightly hazy layer is also not fully dispersed, so it is classified as a very cloudy layer and

gets assigned a fraction of 0.1, taking into account the fact that this layer is almost fully

dispersed. The total settled volume for the slurry on the right in Figure 2-1 is then

calculated to be 20 + 20(0.8) + 40(0.1) = 40 mL. This slurry is therefore assigned the

same total settled volume as a slurry with 40 mL of clear liquid atop a fully dispersed

slurry. The three slurries in Figure 2-1 are then placed in the proper relative order in









terms of their settling behavior shown. Using the method outlined above, all slurry that is

settled is taken into account in this investigation, and not just the fully settled layer of

clear liquid. Although this method has arbitrarily assigned values for the different levels

of opacity, the settling data obtained with this method is shown in Chapter 3 to correlate

well with the hydrophobicity of the nonionic surfactant, indicating the validity of the

method. Two cylinders were prepared of each slurry and settling data reported is the

average value.


2.3 Effects of Addition of Ionic or Nonionic Surfactants on Slurry Stability


The simplest methodology for stabilizing dispersions using surfactants is through

the addition of single surfactants. When an ionic or nonionic surfactant stabilizing agent

is added alone to a dispersion in a high ionic strength environment, the expected effect on

the dispersed particles is illustrated in Figure 2-2.


Anionic (or Cationic) Surfactant


0 0 0


Figure 2-2.


Nonionic Surfactant
0 n o

0o 0
0 0 0


0




+ oo 0
+ 0 0



Unstable Slurry Unstable Slurry
0 = positive ions 0 = negative ions
Illustration of the effect of single ionic or nonionic surfactant addition on
the stability of high ionic strength slurries.









This figure illustrates particles with a net positive charge (e.g. alumina) after

addition of either anionic surfactant or ethoxylated nonionic surfactant. The explanation

of the expected effects is in the following discussion.


2.3.1 Ionic Surfactants

The shielding of surface charges on the particles due to the large number of ions

does not prevent the adsorption of oppositely charged ionic surfactant. The charged sites

on the particles still act as anchoring sites for the ionic surfactant head groups, even in the

high ionic strength environment. In fact, since the surface charge represents a net charge,

there will still be negative sites where surfactant with a positive sign (same sign as the

surface charge) can adsorb, but these negative sites will be fewer in number than positive

sites, resulting in a lower adsorption density of cationic surfactant. Adsorbed ionic

surfactants may form a bilayer due to hydrophobic chain interactions, but this depends on

the surfactant. As discussed in Chapter 1, ionic surfactants that have no polymeric

component primarily stabilize slurries by increasing surface charge and hence causing

electrostatic repulsion. However, in high ionic strength slurries, even the charges placed

on the surface by adsorption of ionic surfactant are shielded by the ions in the slurry and

the hence electrostatic repulsion will not be effective. Another way of stating this is that

in high ionic strength environments the Debye length, 1/K, is so small that van der Waals

attractive forces will predominate and cause particle agglomeration, practically regardless

of the concentration of surfactant placed on the surface.

The adsorbed ionic surfactants do, however, have a steric component to

stabilization that can be effective. The adsorbed surfactants do present a physical barrier

to approach, hence reducing the effect of van der Waals forces. This effect was found to









be significant in systems of cetyl trimethylammonium bromide (CTAB) cationic

surfactant adsorbed on negatively charged silica (SiO2) surfaces by Singh et al. [2000].

In this system, with 0.1 M NaCl providing a high ionic strength environment, significant

CTAB adsorption was found to stabilize these slurries, presumably due to steric effects

since electrostatic effects should be minimal in this system. This study verifies that

steric effects can have a significant contribution even in systems of nonpolymeric ionic

surfactants. However, in the present investigation, the van der Waals attraction between

alumina particles is much greater than that between silica particles. This can be seen by

the difference in non-retarded Hamaker constant for the two materials interacting across

vacuum (air), which is 14 x 10-20 J for alumina [Velamakanni et al., 1994] and about 6.3

x 10-20 J for silica [Israelachvili, 1992]. The greater van der Waals attraction for alumina

means that the steric layer needed to stabilize these particles will be larger than that in the

silica system, and hence common ionic surfactants will not be sufficient to provide this

layer.


2.3.2 Nonionic Surfactants

The polar group in most nonionic surfactants is polymeric in nature and hence

causes stabilization of slurries by causing steric repulsion between particles. This

stabilization mechanism should be significant, even in high ionic strength slurries.

Although the polymer conformation may be slightly different under high ionic strength

conditions, this difference should not be significant enough to cause the steric

stabilization effects to change. In aqueous environments, the most common nonionic

surfactants used for stabilization are of the poly(oxyethylene) (PEO) variety. This is

because PEO polymer chains are very hydrophilic in nature and therefore tend to extend






51

into aqueous solution and not coil at the surface. In this way, ethoxylated nonionic

surfactants have been shown to stabilize numerous aqueous dispersions, even under high

ionic strength conditions [Rosen, 1989]. However, numerous studies have shown that

PEO has a very low adsorption density on alumina [Somasundaran and Krishnakumar,

1997]. Since nonionic surfactants must adsorb on alumina with the head group towards

the surface since alumina is a hydrophilic particle, ethoxylated nonionic surfactants

likewise do not adsorb significantly on alumina surfaces and hence would not be

expected to stabilize these slurries.


2.3.3 Verification by Sedimentation Experiments

The theories of ionic and nonionic surfactant adsorption presented above have

been verified using sedimentation experiments in Figures 2-3 and 2-4. In Figure 2-3, the

addition of surfactant to a low ionic strength slurry containing 0.001 M potassium

ferricyanide (K3Fe(CN)6) is investigated.





rI N ...u.. I"'I P"( | tEB
iB "-I I"' I H











slurry containing 0.001 M potassium ferricyanide oxidizing agent. The
slurries are I wt. % AKP-50 alumina at pH 4 with 10 mM surfactant
addeFigure 2-3. The effects of ionic or nonionic surfactant additionwere taken after 24 hours of settlionic strength
slurry containing 0.001 M potassium ferricyanide oxidizing agent. The
slurries are 1 wt. % AKP-50 alumina at pH 4 with 10 mM surfactant
added. The photographs were taken after 24 hours of settling.









In this figure, the control sample with no surfactant added shows instability,

verifying that the electrostatic repulsion between particles has been screened even under

low ionic strength conditions. Cationic, nonionic and anionic surfactants in a

concentration of 10 mM each have been added to these slurries to investigate

stabilization. Adding either cationic surfactant (CPC) or anionic surfactant (SDS)

stabilizes the slurries, which are unstable when no surfactant is added. This is shown by

the white alumina particles being dispersed throughout the slurry. These results verify

that both cationic and anionic surfactants can adsorb on these particles. At this low ionic

strength, the Debye length (1/K) of the system is sufficiently large that ionic surfactant

can have an effect on stabilization. This effect is presumably a combination of both steric

and electrostatic effects from the adsorbed layer of ionic surfactant.

The addition of nonionic surfactant (Triton X-100) does not stabilize the slurry in

Figure 2-3 under these conditions. Other nonionic surfactants investigated, including

Tween 80, Tween 20 and Symperonic A7 show similar results. The lack of stability in

these systems is presumably due to insufficient or lack of adsorption of nonionic

surfactant on the alumina particles.

Figure 2-4 investigates the addition of the same surfactants to high ionic strength

dispersions containing 0.1 M potassium ferricyanide. This amount of potassium

ferricyanide is commonly used to achieve the required amount of oxidation of tungsten or

copper surfaces in metal CMP slurries. Under these conditions, the control sample with

no surfactant remains unstable. The addition of cationic surfactant (CPC) shows a higher

fractional volume than the other surfactants, but this is presumably due to very rapid









settling of the slurry and formation of loose agglomerates between particles, resulting in a

loose packing efficiency.




t. ,



No Surfl''l; m Ia








Figure 2-4. The effects of ionic or nonionic surfactant addition on a high ionic
strength slurry containing 0.1 M potassium ferricyanide oxidizing agent.
The slurries are 1 wt. % AKP-50 alumina at pH 4 with 10 mM surfactant
added. The photographs were taken after 24 hours of settling.



The addition of anionic surfactant (SDS) and nonionic surfactant (Triton X-100)

show no significant stabilization of the slurries under these conditions. The high ionic

strength of these slurries prevents the cationic or anionic surfactants from stabilizing the

slurries electrostatically. The nonionic surfactant does not appear to adsorb on the

particles under these conditions, as is the case in the low ionic strength environment

(Figure 2-3). These results have been verified using other cationic, anionic and nonionic

surfactants.


2.3.4 Adsorption Verification by Surface Tension Measurements

Static surface tension measurements were made using a Rosano surface

tensiometer made by Biolar Corporation. Surface tension measurements were made in









air using the Wilhelmy plate method [Becher, 1965] with a 10 x 20 mm platinum plate,

allowed to equilibrate in solution for 60 seconds before taking a measurement. Slurries

were centrifuged to obtain supematants for surface tension measurements using a Sorvall

RC-5B refrigerated superspeed centrifuge operating at 10000 rpm.

The adsorption of ionic and nonionic surfactants has also been investigated using

static surface tension measurements. By comparing the surface tension of surfactant

solutions to that of the supernatant solutions after surfactant adsorption on particles,

adsorption of surfactant can be determined. For these measurements, 10 wt. % AKP-50

particle slurries were prepared at pH 4 using 0.1 M potassium ferricyanide oxidizing

agent. SDS, CPC and Tween 80 surfactants were added at a concentration of 20 mM

prior to dilution. The slurries were centrifuged and the supematants compared to the

same three solutions before addition of particles. The solutions were diluted several

times with water and the surface tension at each dilution was measured. The results are

given in Figure 2-5, which shows the surface tension as a function of surfactant

concentration for all three surfactant solutions before and after adsorption of surfactant on

particles.

The surface tension of surfactant solutions decreases with concentration up to the

critical micelle concentration (CMC). The adsorption of surfactant on particles will

result in a lower amount of residual surfactant in the supematants of solutions containing

particles. Hence, at a given dilution, the solution with particles will have a higher surface

tension than that without particles if the surfactant is adsorbing on the particles.

In Figures 2-5 a) and b), which show CPC and SDS solutions, respectively, the

surface tension of solutions with particles is higher than the surface tension of solutions









without particles. These results verify that CPC and SDS adsorb on alumina particles and

hence lower the concentration of surfactant in solution. However, the adsorption density

of surfactant is presumably insufficient to stabilize alumina dispersions.


80

60

40

20 -
0.(


0.01


0.021


SDS
0- 0..1..
b)

0 0.1 0.;


50 -
40 -
30 -
0.00


1


S= Without
Addition
of Particles

A = After
Addition
of Particles


1000


Figure 2-5.


Concentration Before
Adsorption (mM)

Surface tension of solutions before and after addition of alumina particles,
as a function of surfactant concentration, achieved through dilution with
water. Surfactants are a) CPC, b) SDS and c) Tween 80.


Figure 2-5 c) shows that for Tween 80 solutions, there is a negligible difference

between surface tensions of solutions with and without particles. This verifies that this


CPC
a)-


Tween 80

c)


O0


6n









nonionic surfactant does not adsorb on alumina and hence does not lower the surfactant

concentration in solution.


2.4 Effects of Addition of Surfactant Mixtures on Slurry Stabilization


The addition of mixtures of cationic, anionic and nonionic surfactants to the high

ionic strength metal CMP slurries has also been investigated. This has been done in order

to determine if any synergistic effects of surfactant combinations are present in these

slurries. The effects of addition of surfactant mixtures on stabilization of high ionic

strength slurries are shown in Figures 2-6 and 2-7. These figures show the results of

sedimentation tests on 10 wt. % AKP-50 (100-300 nm) alumina slurries containing 0.1 M

oxidizing agent. Mixtures of two surfactants were added to these slurries at a molar ratio

of 1:1, with a total surfactant concentration of 50 mM. This value was chosen because it

is well above the CMC of any of the surfactants used in this study [Dahanayake, 1984;

Rosen, 1982]. The CMC of SDS in water is 8.0 mM while that of ethoxylated nonionic

surfactants can vary from about 0.005 mM to 0.2 mM [Patist, 2000]. More importantly,

the amount of residual surfactant in these slurries after equilibrium adsorption is reached

is also greater than the CMC of any of the surfactants investigated. Hence, the amount of

surfactant adsorption will have reached a saturation value at the concentration used in this

study, and no further stabilization effects are expected to occur at higher concentrations,

unless multilayer adsorption effects are significant.












1 -

0.9 -

0.8 -

0.7

0.6 -

0.5 -

0.4 -

0.3

0.2 -

0.1

0 -


E F


Oxidizing Agent =
0.1 M Ferric Nitrate






D







B C


Cationic + Anionic
Surfactants


No Cationic + Nonionic Anionic + Nonionic
Surf Surfactants Surfactants


Mixed Surfactant Compositior
(50 mM Total Concentration)


* A -FeNit Only
O B -CTAB + T80
E]C -CPC + T80
OD CTAB + A7
OE -SDS + T80
-]F -SDS + T20
G SDS + X100
CH -SDS + A7
S- SDS + C12TAB (5 1)
*J -SDS + CTAB (5 1)


Figure 2-6.


Sedimentation results of 0.1 M Fe(N03)3 slurries with mixtures of
surfactants added.


Figure 2-6 shows the sedimentation behavior of slurries containing 0.1 M ferric


nitrate (Fe[N03]3), another common oxidizing agent used in metal CMP slurries. The


settling behavior is reported as fractional volume of dispersion, which is defined as the


ratio of the volume of dispersed slurry remaining at a given time to the initial volume of


slurry. Hence, this parameter has a value of 1.0 for stable slurries. For unstable slurries,


the values of this parameter can be as low as 0.1, which represents the fractional volume


of tightly packed 10 wt. % dispersions of AKP-50 particles. In Figure 2-6, the instability


of slurries with no surfactant stabilizing agents added is once again verified by the ferric


nitrate only (FeNit only) results, which show that after only 3 hours of settling the


G









fractional volume of dispersion is 0.3. The combination of cationic and nonionic

surfactants gives some stabilization, but significant particle settling is still observed. The

combination of anionic and nonionic surfactants shows a wide variety in extent of

stabilization. In this case, the combination of SDS and either Tween 20 or Tween 80

results in fully stabilized slurry after 3 hours. The other nonionic surfactants (Triton X-

100 and Symperonic A7) mixed with SDS do not significantly stabilize the slurries. For

the mixtures of cationic and anionic surfactants (C12TAB or C16TAB and SDS), a molar

ratio of 5:1 anionic:cationic surfactant was chosen because higher amounts of these

cationic surfactants results in precipitation due to electrostatic interaction between

surfactants. Figure 2-6 shows that the mixture of these surfactants actually results in

destabilization of these slurries, since the fractional volume of dispersion is actually less

than that of the slurry without surfactant.

A more detailed investigation of the effects of minor changes in surfactant type on

slurry stabilization was conducted for the combination of anionic and nonionic

surfactants. For this investigation, 0.1 M potassium ferricyanide was used as the

oxidizing agent in order to compare the differences in stabilization behavior for surfactant

mixtures in different chemical environments. Surfactant mixtures were added in a 1:1

molar ratio with a total concentration of 50 mM, as with the ferric nitrate slurries. Figure

2-7 shows the fractional volume of dispersion of these slurries after 3 days of settling.

The different anionic surfactants added with Tween 80 (T80) show that sodium sulfate

surfactants stabilize these slurries better than sodium salts of fatty acids. Of the chain

lengths investigated, the 12-carbon chain (SDS) performed the best, with a fractional

volume of dispersion near 1.







59





1.00
O
.0 0.90
SA PotFer Only
0.80 -- DB Na(C10)Sulf + T80
C 08Oxidizing Agent =
0 1 M Potassium OC Na(C12)Sulf + T80
*- I 0.70 --J K
SFerricyanide D Na(C16)Sulf + T80
0 0.60 -- DE -Na(C10)Acid + T80
o) 0 LF Na(C12)Acid + T80
E 0.50 -- G -SDS+T20
| B CD -SDS + T40
o *0 0.40
S 0.40 SDS + T60
0.30 G H M DS + A4
K SDS + A7
O 0.20 -- L-SDS+All
SOM -SDS+A20
0.10 E F
LL 0.00
No Different Anionics Different Nonionics
Surf with T80 with SDS

Mixed Surfactant Composition
(50 mM Total Concentration)

Figure 2-7. Sedimentation results of 0.1 M K3Fe(CN)6 slurries with mixtures of
anionic and nonionic surfactants added.



The different nonionic surfactants investigated with SDS show effects of chain

length on stabilization of these slurries. Tween 20, 40 and 60 are identical surfactants

except that they have chain lengths of 12, 16 and 18 carbons, respectively. Tween 60

(T60) shows significantly better stability than Tween 20 (T20) or Tween 40 (T40).

However, the SDS + Tween 80 (C18 oleyl surfactant) combination still shows the best

results. The Symperonic surfactants investigated are identical except that the hydrophilic

polymer group length varies, with A4, A7, Al 1 and A20 corresponding to 4, 7, 11 and 20

ethylene oxide (EO) groups, respectively. Figure 2-7 shows that the stabilization

increases with decreasing number of EO groups. This trend will be considered in great

detail in the following chapter.










2.5 Explanation of Stabilization Mechanism


The stabilization of high ionic strength slurries shown in Figures 2-6 and 2-7

using the combination of SDS and either Tween 20, Tween 80, Symperonic A4 or A7 can

be explained by the generic adsorption model proposed in Figure 2-8. The stabilization is

shown to occur by a two step mechanism: 1) The adsorption of a strongly adsorbing

ionic surfactant. 2) The adsorption by hydrocarbon chain interaction of a nonionic

surfactant capable of steric stabilization.


K+

[Fe(CN),]-


[Fe(C NL I


[Fe(C N I


I ~


[Fe(CN),]-


S0


[Fe(CN)]3-


Figure 2-8.


The mechanism of high ionic strength slurry stabilization by the
synergistic mixture of anionic and nonionic surfactants.


-o oP
MO, 4*0,0,


a
44









This process results in the adsorption of a mixed surfactant layer on the surface of the

alumina particles. The presence of polymeric chain segments from the nonionic

surfactant extending into the water results in steric repulsion when two such particles

approach each other.

Although this type of stabilization has only been observed in these systems for an

anionic and nonionic surfactant mixture, it is suspected that the same mechanism is

possible for a cationic and nonionic surfactant system. This is because electrostatic

effects are not expected to influence ionic surfactant adsorption in high ionic strength

environments. Further, the adsorption of nonionic surfactant is governed in this

mechanism by the interaction of the hydrocarbon chains, and these interactions should be

similar for cationic and anionic surfactants.

The mechanism in Figure 2-8 only depicts one possible conformation of the

adsorbed mixed surfactant layer, which in fact may be present as either a monolayer or a

bilayer, as explained below. The conformation presented in this figure is a mixed

monolayer conformation, in which the hydrocarbon chains of the two surfactants are

lined up adjacent to each other. This type of conformation will be favored if the amount

of adsorbed ionic surfactant is low enough that there are gaps between molecules that are

large enough for the nonionic surfactant hydrocarbon chains to fully penetrate, as shown

in the figure. The mixed monolayer conformation will result in the lowest surfactant

adsorption density of any of the conformations possible in this mechanism.










b)


Figure 2-9. Other possible conformations of stabilizing surfactant that result from the
adsorption of a mixture of ionic and nonionic surfactants on particle
surfaces. Conformations are a) full bilayer and b) penetrated bilayer.


Another type of conformation that the adsorbed mixed surfactant layer may prefer

is a full bilayer conformation. This type of conformation is illustrated in Figure 2-9 a),

which shows that in this conformation the anionic surfactant has adsorbed in a full

monolayer. The nonionic surfactant then adsorbs via hydrocarbon chain interactions as a

full second monolayer. This conformation results in twice as many molecules of each

surfactant adsorbing on the surface than in the mixed monolayer conformation. The full

bilayer conformation will be favored when there is enough anionic surfactant in solution

to adsorb as a full monolayer. This conformation will result in the highest surfactant

adsorption density of any of the conformations possible in this mechanism.

The final type of conformation, a penetrated bilayer, results in a surfactant

adsorption density that is in between the two extremes already discussed. This type of

conformation is shown in Figure 2-9 b), which shows that the hydrocarbon chains of the

nonionic surfactant penetrate the monolayer of anionic surfactant adsorbed on the particle









surface, resulting in an adsorbed layer that is neither a mixed monolayer nor a complete

bilayer. The degree of penetration of the nonionic surfactant will depend on the amount

of space available between adsorbed anionic surfactant molecules, as well as the strength

of the hydrocarbon chain interactions between adjacent molecules.

The results of Figures 2-6 and 2-7 suggest that this stabilization process is highly

dependent on the choice of both ionic and nonionic surfactant. The stabilizing effect of

surfactant mixtures by this mechanism is dependent on several factors, which are

explained in the following chapter. This chapter has shown that mixed surfactant systems

can show synergistic behavior as a stabilizing agent for dispersions. The next chapter

will examine the factors that effect the stabilizing ability of a mixed ionic and nonionic

surfactant system. With this knowledge, the formulator is given the ability to choose

surfactants which will maximize stabilization, even in severe environments such as those

considered here.














CHAPTER 3
MOLECULAR FACTORS THAT OPTIMIZE DISPERSION STABILITY USING
MIXED SURFACTANT SYSTEMS AS STABILIZERS FOR DISPERSIONS IN
SEVERE ENVIRONMENTS



3.1 Molecular Factors Influencing Stability of Dispersions


In this investigation, the factors which effect the stabilization mechanism shown

in Figure 2-7 are investigated. Specifically, the types of surfactants which are preferred

for this type of mechanism are determined, as well as the concentrations and ratios of

surfactants which are necessary for the stabilization to be successful. Finally, this

mechanism is applied to another slurry which utilizes chemistry for copper CMP

applications, and the differences in surfactant type, concentration and ratio for optimal

stability are explained.

The factors which influence the stabilization mechanism presented in Figure 2-7

have been deduced and are considered here. They are:

1) Adsorption of ionic surfactant onto alumina surface with ample surface coverage.

As explained in Chapter 1, surfactants with sign opposite of the surface charge are

expected to have a higher affinity to the surface, although the magnitude of the

surface charge and heterogeneity will ultimately determine the adsorption density.

2) Partitioning of nonionic surfactant out of the aqueous phase and onto the alumina

surface by hydrocarbon chain interaction with the film of ionic surfactant.

Hydrocarbon chain interactions have been shown to be maximum for equal chain
Reproduced in part with permission from Langmuir, submitted for publication.
Unpublished work copyright 2000 American Chemical Society.

64









lengths of ionic and nonionic surfactant [Shiao et al., 1997; Shiao, 1996; Patist et

al., 1997]. However, the driving force for partitioning of nonionic surfactant out of

the aqueous phase and onto the surface increases with increasing hydrophobicity of

nonionic surfactant according to a semiquantitative extension of Traube's rule

[Traube, 1891] made by Freundlich [1926]. As a result, the adsorption affinity of

the nonionic surfactant increases with long hydrocarbon chains or short hydrophilic

segments of the nonionic surfactant.

3) Steric stabilization via the mixed surfactant layer on adjacent particles. This

depends primarily on the ability of the hydrophilic ethylene oxide (EO) segments of

the nonionic surfactant to extend into the aqueous phase upon adsorption and not

tend to form flocs. As discussed in Chapter 1, steric stabilization occurs in mixed

layers due to the presence of steric barriers from adsorbed nonionic molecules that

prevent the particles from coming close enough to allow van der Waals attractive

forces between particles to dominate [Tadros et al., 1986; Lee et al., 1986].

However, the poly(ethylene oxide) segments must not interact attractively with

segments on adjacent particles to induce flocculation, causing destabilization of the

slurry. The tendency for flocculation will increase as the adsorption density of

polymer segments is increased [Rosen, 1989].

This investigation determines the influence of molecular factors and parameters of

the surfactants used in this stabilization mechanism. The results presented in this chapter

are entirely from sedimentation experiments. Although the use of this type of enhanced

adsorption mechanism has been reviewed in the literature presented in Chapter 2, the

investigation of detailed influences of various factors such as presented in this chapter









has not been performed to date. This investigation has even greater implications in that it

investigates dispersion stability in severe environments, which is an issue in many current

industries.


3.2 Nonionic Surfactants Used in This Investigation


As can be seen from the factors influencing stabilization, the nonionic surfactant

plays a major role in determining the outcome in this mechanism. As a result, a large

number of commercial nonionic surfactants has been investigated using sedimentation

experiments. Figure 3-1 illustrates the ethoxylated nonionic surfactants used in this

investigation and lists those trade names which represent each linkage. As can be seen,

ethoxylated nonionic surfactants consist of a hydrocarbon chain, which may be branched,

a poly(ethylene oxide) (PEO) segment as the head group and a linkage which connects

the two sides of the molecule. Three types of linkages have been investigated in this

chapter, including ethers, esters and sorbitans.

The sedimentation experiments have been performed using the same methodology

as described in Chapter 2. All nonionic surfactants investigated, including Span 20, Span

80, Brij 30, Brij 35, Brij 52, Brij 93, Brij 97, Brij 98, Brij 700, Tween 21, Tween 80,

Tween 81, Tween 85, Symperonic A4, Symperonic A7, Symperonic All, Symperonic

A50 and Myrj 45, have been obtained from ICI Surfactants, Ltd. For this investigation,

dispersions of 10 wt. % AKP-50 alumina have been used at pH's appropriate for the

CMP application desired.











I 1 ------ oMc

Tail Group: Linkages =
C12- C18 Ether
(C12 C18' (Brij)
C18 oleatee) Ester
and tri-C18) C O (Symperonic,

C Myrj)
x Sorbitan
Ey (Tween, Span'


Figure 3-1.


OOH
Head Group:
Eo Eloo00
(Eo, E2, E4, E7, E8, E10,
Ell, E20, E23, E50, E100)


Schematic of a general ethoxylated nonionic surfactant and details about
the types used in this investigation.


3.3 Optimization of Dispersion Stability Using Various Nonionic Surfactants for
Tungsten CMP

In order to determine the effects of nonionic surfactant type on dispersion stability

in the tungsten CMP slurry investigated primarily in Chapters 1 and 2, the same slurry

has been investigated using sedimentation experiments and the nonionic surfactants listed

previously. Nonionic surfactants have been divided into groups with carbon chain

lengths of 12 and 18. The sedimentation results are given in Figures 3-2 and 3-3 for C12

and C18 nonionic surfactants, respectively.

As can be seen for both the C12 and C18 surfactants, there appears to be some

correlation between degree of polymerization, as indicated for each series, and dispersion

stability.




























0 5 10 15
Settling Time (days)


Sedimentation results of dispersions with 1:1 mixtures of SDS and the C12
nonionic surfactant indicated. Oxidizing agent is 0.1 M potassium
ferricyanide. Ex represents the mean number of ethoxy groups on the
nonionic surfactant molecule.


5 10 15
Settling Time (days)


Sedimentation results of dispersions with 1:1 mixtures of SDS and the C18
nonionic surfactant indicated. Oxidizing agent is 0.1 M potassium
ferricyanide.


Figure 3-2.


Figure 3-3.









The most stable combinations in both Figures 3-2 and 3-3 have only 2 EO groups (E2) on

the nonionic surfactant, while the highest degree of polymerization (E50 for C12 and E00oo

for C8i) results in the most unstable slurries. These two extremes show a general trend of

increasing stability with decreasing degree of ethoxylation of the nonionic surfactant.

Additionally, Span surfactants (Eo) are unstable in both figures. However, in both Figure

3-2 and 3-3 the Span surfactants show a substantial particle layer (F (p 0.4) remaining in

the unstable slurries. These surfactants promote the formation of loose particle

agglomerates rather than tightly packed agglomerates. As a result, the final packed

volume of these slurries after settling is complete is much larger than the packed volume

of other slurries that form tightly packed agglomerates. Also in Figures 3-2 and 3-3,

there appears to be a significant variation with type of nonionic surfactant linkage, since

for the C18 surfactants, Tween 80 (sorbitan linkage) and Brij 98 (ether linkage) all are E20

surfactants, but their results do not follow the same trend.

There appears to be some correlation in Figures 3-2 and 3-3 of dispersion stability

with the molecular factors of the surfactant, including the carbon chain length, degree of

polymerization and linkage segment. All of these parameters contribute to the overall

hydrophobicity of the surfactant, which can be summed up for surfactants using the

empirical hydrophile-lipophile balance (HLB) number. HLB numbers are low for

hydrophobic surfactants and high for hydrophilic surfactants, and an equal hydrophobic-

hydrophilic balance corresponds to a value near 10. The hydrophobicity of a nonionic

surfactant is expected to be a significant factor in dispersion stability, since ultimately

this factor and the surface architecture will determine the driving force for surfactant to

leave the bulk and adsorb at the interface.







70


In order to show the effect hydrophobicity of nonionic surfactant on dispersion

stability, the data in Figures 3-2 and 3-3 has been plotted in Figure 3-4 as a function of

HLB number of nonionic surfactant. The surfactants are grouped in this figure according

to both linkage type and chain length (C12 or C18). Figure 3-4 shows that both C12 and

Ci8 ethers and the esters (all C12) show a decrease in dispersion stability with an increase

in HLB number of nonionic surfactant. The sorbitan surfactants remain stable at HLB

numbers greater than 15, but it is expected that if it were possible to obtain very high

HLB number sorbitan surfactants, they would also show a decrease in dispersion stability

at higher HLB numbers.


I I I I I
0 5 10 15 20
Hydrophile-Liphophile Balance (HLB)


Correlation of dispersion stability with hydrophobicity of nonionic
surfactant, as given by its HLB number. Data is from 50 mM 1:1 mixtures
of SDS and the given nonionic surfactant adsorbed on 10 wt. % AKP-50
particles in 0.1 M K3Fe(CN)6 at pH 4.


1.0 -









0.4
0.2

,0.
(0
* 0.4 -
o

,2 0.2 -
(o

0.0 -


Figure 3-4.


-W-C18 Ethers
-A-C12 Ethers
-*-C18 Sorbitans
-*-C12 Sorbitans
--C12 Esters


I









The data in Figure 3-4 shows that increasing hydrophobicity of nonionic

surfactant leads to increased dispersion stability, which is contrary to general

expectations. In most dispersions stabilized by steric layers, the stability should increase

with increasing length of the steric stabilizing component, which is the degree of

ethoxylation in the present investigation. This figure shows the reverse trend in mixed

surfactant systems, in which increasing the length of the steric stabilizing layer actually

contributes to a decrease in dispersion stability.


3.4 Verification of Results for Model Copper CMP Slurry


In order to determine the generality of the trends presented above, the effect of

nonionic surfactant on dispersion stability was determined using similar sedimentation

experiments for a type of slurry with application in copper CMP. This slurry contains

primarily monovalent ions, unlike the trivalent ferricyanide ion. The slurry is composed

of 10 wt. % alumina particles and 0.1 M KIO3, 0.01 M KI and 0.01 M EDTA. This

slurry combination has been shown to provide substantial oxidation of the copper surface

due to KIO3 while the corrosion is inhibited by the presence of I from KI, which forms a

surface layer on Cu. Finally, the driving force for removal of copper is maximized by

addition of the completing agent EDTA [Kaufman et al., 1999].

Figure 3-5 shows the trend of dispersion stability for this copper CMP slurry with

HLB number of nonionic surfactant for all of the nonionic surfactants investigated in

Figure 3-4. In this figure, the surfactants are grouped according to carbon chain length

(C12 and C18), and the C12 surfactants are investigated at total surfactant concentrations of

both 50 mM and 100 mM to observe any differences. The data in Figure 3-5 appears to

be very well correlated with HLB number of nonionic surfactant. For both C12 and C8i






72

surfactants, there appears to be a rapid stable unstable transition in a range of HLB's

from 10-12. It is important to note that this stability data is presented after only 1 day of

settling, indicating that the speed of settling in the unstable slurries is rapid. However,

the stable slurries with HLB numbers less than 10 are fully stable after a period of 10

days, indicating a high degree of stability.


1.0


0.8


0.6


0.4


0.2


0.0


1 3


5 10 15
HLB # of Nonionic Surfactant


Variation in dispersion stability vs. HLB number of nonionic surfactant
for a model copper CMP slurry containing 0.1 M KIO3, 0.01 MKI and
0.01 M EDTA (Nonionic surfactants are: 1 = Span 80, 2 = Brij 93, 3 =
Brij 52, 4 = Span 20, 5 = Brij 30, 6 = Symperonic A4, 7 = Tween 81, 8 =
Symperonic A7, 9 = Brij 97, 10 = Tween 21, 11 = Symperonic All, 12 =
Tween 80, 13 = Brij 98, 14 = Tween 20, 15 = Brij 35, 16 = Symperonic
A50, 17 = Brij 700).


-'
2



5






mC12 Nonionics (50 mM)
899 12,13 17
-- I C12 Nonionics (100 mM) j

SC18 Nonionics 11 1
14,15,16


Figure 3-5.


4 67









The primary outlier point is a Q2 nonionic surfactant with an HLB number of

13.3, which yields a stable dispersion. This surfactant is Tween 21, which has a sorbitan

linkage, unlike most of the C12 surfactants investigated. It appears that, as in Figure 3-4,

the sorbitan linkage contributes more significantly to steric stabilization at high HLB

numbers than its ether or ester counterparts. The results for the dispersions at 100 mM

concentration mimic those at 50 mM concentration, indicating that there is no change in

stability in this concentration range. The effects of surfactant concentration on dispersion

stability will be discussed in greater detail in section 3.6. The results of Figure 3-5

verify the robust nature of the stabilizing scheme presented in this investigation. The

general trend of increasing dispersion stability with decreasing HLB number of nonionic

surfactant holds for both the tungsten and copper CMP slurries investigated, indicating

that the underlying stabilization mechanism is the same in both systems.

The results in Figure 3-5 do include the Span surfactants (Eo), which produce

stable slurries in contrast to the results in Figure 3-4. This is true even though the Span

surfactants have 0 EO groups, and hence their hydrophilic segment is composed of only

the sorbitan linkage. This contrasting element will be considered and explained in the

next section after an explanation of the general trend of increasing stability with

decreasing HLB number of nonionic surfactant.


3.5 Correlation of Dispersion Stability with HLB Number of Nonionic Surfactant


The trend that stands out in both Figures 3-4 and 3-5 is that increasing

hydrophobicity of nonionic surfactant, with all other factors constant, brings about

increasing dispersion stability. This is contrary to the popular thought that longer EO

chains should provide better dispersion stability due to increasing size of the steric barrier









to particle approach. A schematic of the environment in which a hydrophilic nonionic

surfactant with a long polymer chain is present in solution is presented in Figure 3-6 a),

while the same situation with a hydrophobic short chain surfactant is presented in Figure

3-6 b).




a) I*


High CMC -mmO
High Monomer
Concentration
Less Adsorption
at S/L Interface


b)




Low CMC
Low Monomer
Concentration
More Adsorption
at S/L Interface


Figure 3-6. Schematic of dispersion environment showing adsorption density
differences between a) high HLB hydrophilicc) nonionic surfactant and b)
low HLB (hydrophobic) nonionic surfactant.


The reason for the greater degree of stabilization with the hydrophobic (low HLB)

surfactant is clearly the number of surfactant molecules present on the surface. With the

hydrophobic surfactant, the driving force to place monomers onto the particle surface is

higher than for the more hydrophilic surfactant. As a result, the hydrophilic surfactant,









although capable of greater steric stabilization if adsorbed, does not adsorb in sufficient

quantity to stabilize the particles. The key to understanding this stabilization mechanism

is that there are opposing forces governing the choice of nonionic surfactant-the driving

force to adsorb vs. the degree of polymer extension into solution once adsorbed. Clearly

in the example slurries presented in Figures 3-4 and 3-5, the repulsive barrier is sufficient

even with only 2 or even 0 E groups on the surface.

This leads to the discussion of the contrasting element between Figures 3-4 and 3-

5. The Span surfactants (Eo) stabilize the KIO3 slurries but not the K3[Fe(CN)6] slurries.

The reason for this observation is that in the KIO3 slurries the charge screening effects

are not as overwhelming as in the K3[Fe(CN)6] slurries. Quantitatively, the Debye length

(1/K) in electrolyte solution containing 0.1 M KIO3 and 0.01 M KI is 9.17 A, while in

solution containing 0.1 M K3[Fe(CN)6] it is only 4.30 A. Therefore, the alumina particles

in KIO3 dispersions experience a greater electrostatic barrier, and this combined with the

steric barrier provided by the surfactant leads to greater dispersion stability in KIO3

dispersions. The strongly adsorbed hydrophobic Span (Eo) surfactants provide a

sufficient steric barrier in KIO3 dispersions to impart stability. The K3[Fe(CN)6] slurries,

with a trivalent counterion, provide more complete charge screening and therefore reduce

the electrostatic barrier to agglomeration and increase the requirements of the steric

barrier needed. The E2 or E4 surfactants, however, are considerably more hydrophilic

than the Eo surfactants and possess the polymer chain necessary to provide steric

repulsion in these slurries. Since the hydrophobicity of these surfactants is also

sufficiently high, they provide ample surface coverage and maximum stabilization for the

K3[Fe(CN)6] environment.










3.6 Effect of Total Surfactant Concentration on Dispersion Stability


The results in Figures 3-4 and 3-5 are for a total surfactant concentration of 50

mM, but the dispersion stability is expected to vary significantly with surfactant

concentration. As a result, the surfactant concentration has been varied in sedimentation

experiments presented in Figure 3-7. In this figure, the dispersions are 10 wt. % AKP-50

alumina at pH 4 with 0.1 M potassium ferricyanide and the surfactant in these systems is

a mixture of SDS and the indicated nonionic surfactant in a 1:1 molar ratio.

The results in Figure 3-7 show interesting, unexpected results, particularly for the

SDS/Tween 80 surfactant mixture. The data shows that, as expected, a minimum

surfactant concentration is needed for stabilization, indicative of the large surface area of

the particles that must be sufficiently covered with surfactant to achieve stabilization.

The unexpected result is that the dispersion stability once again decreases at high

concentrations of SDS/Tween 80. This result may be due to a number of factors, the first

being the formation of nonionic surfactant domains or micelles lining the adjacent

particles. The Tween 80 domains formed by E groups then serve as glue (adhesive)

connecting the particles in a loose agglomerate or floc. This mechanism will be

explained in detail in section 3.8. The other possible explanation for this result is that

multilayer effects are dominant in this system, resulting in the adsorption of a third

surfactant layer of either SDS or Tween 80 at high concentrations. This third layer, with

hydrophobic chains extending into solution, makes the particles once again susceptible to

agglomeration due to hydrophobic interactions between adjacent particles.











r 1.0
0

o 0.8
U)= 0= 8

U --SDS/Brij 93 (E2)
0o 0.6
E -A-SDS/Symp A4 (E4)
o 0.4 SDS/Tween 80 (E20)


o 0.2


.L 0.0

0 25 50 75 100 125 150

Surfactant Concentration (mM)


Figure 3-7. Sedimentation experiments showing dependence of dispersion stability on
surfactant concentration in 0.1 M potassium ferricyanide slurries at pH 4.



The results of the other surfactant systems (SDS/Brij 93 and SDS/Symperonic

A4) which are shown in Figure 3-7 do not show the same trends as the SDS/Tween 80

results. Although SDS/Brij 93 mixtures do show a minimum concentration required for

stability, there is no decrease in dispersion stability at higher concentrations. However,

it is important to note that Brij 93 is only slightly soluble in water due to its

hydrophobicity. The solubility of Brij 93 is significantly enhanced by the presence of

SDS, but there is some cloudiness present at high concentrations (,t 75 mM), indicating

that the solubility limit has been reached. As a result, it is expected that the reason the

SDS/Brij 93 mixtures do not become unstable at higher concentrations is that not enough

Brij 93 surfactant is solubilized to cause instability at high concentrations. The









SDS/Symperonic A4 mixture is likewise less soluble than the SDS/Tween 80 mixture, so

the results for this system are expected to be similar to those of the SDS/Brij 93 system,

as shown in Figure 3-7.

The effects of surfactant concentration on dispersion stability have also been

investigated for the copper CMP slurry consisting of 0.1 M KIO3, 0.01 M KI and 0.01 M

EDTA. It is important to note that the SDS/Tween 80 mixture does not significantly

stabilize this slurry, so a direct comparison with the K3Fe(CN)6 slurry results is not

possible. However, the effects of concentration are given in Figure 3-8 for two mixtures

that have been found to provide optimum stabilization for this slurry at 50 mM

concentration.



1.0
0


0.8 --SDS/Symp A4 (E4)

S\ --SDS/Tween 21 (E4)
0 0.6
0) 0
Eo

o "0 0.4


4.. 0.2


LL 0.0 -
0 25 50 75 100 125 150

Surfactant Concentration (mM)

Figure 3-8. Effect of surfactant concentration on dispersion stability in 0.1 M KIO3,
0.01 M KI and 0.01 M EDTA slurries at pH 6.









This figure shows that both the SDS/Symperonic A4 and SDS/Tween 21 mixtures

impart stability to alumina particles in the KIO3 environment at concentrations as low as

25 mM, but at concentrations higher than 75 mM, both dispersions are clearly unstable.

This result is similar to the result of the SDS/Tween 80 mixture in Figure 3-7. The

SDS/Symperonic A4 mixture shows these high concentration effects in KIO3

environment but not in K3Fe(CN)6 environment because the solubility of the

SDS/Symperonic A4 mixture is greater in KIO3 solution than in K3Fe(CN)6 solution

since the prevalent ion is monovalent, not trivalent. Both mixtures of SDS/Symperonic

A4 and SDS/Tween 21 in KIO3 solution become cloudy at concentrations greater than

100 mM, indicating that the solubility limit has been reached, but not before high

concentration effects cause instability of the dispersion.

The results presented in Figures 3-7 and 3-8 show that surfactant concentration is

a parameter which can significantly effect dispersion stability. The results have shown

the presence of a maximum in dispersion stability over a concentration range which

depends upon a number of factors. The results have also shown that mixtures of

surfactants will show a solubility limit that depends upon the ion concentration and

valence as well as the hydrophobicity of the surfactant system. Finally, the results have

shown that the solubility limit of the surfactant may keep the surfactant from dissolving

at high enough concentrations to cause instability.


3.7 Effect of Ratio of Surfactants in Mixtures on Dispersion Stability


The final parameter regarding the surfactant mixture in the complex systems that

have been investigated is the ratio of ionic to nonionic surfactant. To this point, all of the

mixtures investigated have been at a 1:1 molar ratio. However, it has been shown that






80


ionic or nonionic surfactant alone yield unstable slurries, so there is expected to be at

least two stable unstable transitions on a plot of stability versus fraction of one

component. This parameter has therefore been studied in order to make this investigation

of the molecular factors influencing stabilization complete.

For the K3Fe(CN)6 chemistry, ratios of SDS:Tween 80 and SDS:Symperonic A4

have been investigated using 50 mM total surfactant and 10 wt. % AKP-50 particles. The

sedimentation results are given in Figure 3-9 for these systems as a function of mole

fraction of SDS in the total mixture.


1.0


0.8


0.6


0.4


0.2


0.0


Figure 3-9.


0 0.2 0.4 0.6 0.8 1
Mole Fraction SDS in SDS/Nonionic Mixture


The effect of ratio of surfactants on dispersion stability for slurries
containing 0.1 M K3Fe(CN)6 at pH 4. Total surfactant concentration is 50
mM.


Both systems investigated are unstable when only one surfactant is used (SDS, Tween 80,

or Symperonic A4 alone), as illustrated at the endpoints in Figure 3-9. The SDS/Tween









80 mixture is shown to impart stability to dispersions at mole fractions of SDS at or

greater than 0.5 and all the way up to 0.97 (corresponding to a molecular ratio of 30:1

between SDS and Tween 80). At ratios that favor Tween 80, these dispersions are

unstable. The SDS/Symperonic A4 mixture is shown to be unstable only at the endpoints

(mole fractions of 0 and 1), while it is stable throughout the range of ratios investigated.

The differences in behavior between SDS/Tween 80 and SDS/Symperonic A4

mixtures is presumably due to the inability of low ratios of SDS/Symperonic A4 to

produce nonionic surfactant domains which act as adhesive between particles. This

explanation is feasible since the mechanism, which is discussed in the next section,

depends on long hydrophilic polymer segments on the nonionic surfactant aggregating to

form the domains responsible for instability. Since Symperonic A4 is an F4 surfactant

and Tween 80 is an E20 surfactant, the tendency for Tween 80 systems to exhibit this

behavior is much greater than for Symperonic A4 systems.

The effect of ratio of surfactants in mixtures has also been investigated for the

copper CMP slurry based on KIO3 chemistry. The results of sedimentation experiments

showing the effects of ratio of surfactants on dispersion stability are presented in Figure

3-10. Since the SDS/Tween 80 mixture does not stabilize the copper CMP slurry

significantly, only one mixture is investigated in this figure, SDS/Symperonic A4, but the

results are shown after both 2 days and 10 days of settling. The results show that for

SDS/Symperonic A4 mixtures, as with the K3Fe(CN)6 slurries, the dispersions are stable

almost throughout the full range of SDS mole fractions investigated, except at high mole

fractions greater than 0.9. The results after 10 days show slightly lower stability than

those after 2 days, but the dispersions are still relatively stable. Limited solubility of









SDS in high ionic strength environments is proposed to be the reason that very high ratios

of SDS:A4 are unstable. The overall behavior in Figure 3-10 is very similar to the

behavior of SDS/Symperonic A4 in Figure 3-9, as expected.


.2 1.0 -
O ^
.I-U

.- 0.8

4-
o 0.6
E
| 0.4

c 0.2
0

2 0.0
.U
0.0


0.2 0.4 0.6 0.8

Mole Fraction SDS in SDS/A4 Mixture


Figure 3-10. The effect of ratio of surfactants on dispersion stability for slurries
containing 0.1 M KIO3, 0.01 M KI and 0.01 M EDTA.



3.8 Explanation of Observed Instability in Surfactant Mixtures at High Concentrations
and at Ratios Favoring Nonionic Surfactant


The results of dispersion stability as a function of concentrations and ratios of

SDS/Tween 80 mixtures in Figures 3-7 and 3-9, respectively, both exhibit unexpected

instability in the dispersions. Likewise, the results of dispersion stability as a function of

concentration for the KIO3 chemistry shown in Figure 3-8 shows instability at high

concentrations. The results will be explained in terms of the SDS/Tween 80 system, but

the same mechanism applies to other systems for the KIO3 chemistry. The observed


1.0









instability is presumably due to a change in conformation of the adsorbed polar segments

of Tween 80 in the presence of excess Tween 80. Since Tween 80 is an Ezo surfactant,

any change in conformation of these steric stabilizing segments could lead to a change in

stabilization behavior to a greater extent than the surfactants with lower degrees of

ethoxylation. This is presumably the reason that the SDS/Tween 80 mixture exhibits

instability in the ratio of surfactant results, but the SDS/Symperonic A4 mixture does not.

Similarly, in the concentration of surfactant results, the SDS/Tween 80 mixture exhibits

instability in the K3Fe(CN)6 solution while the other mixtures, with lower degrees of

ethoxylation, do not.

The change in conformation of adsorbed Tween 80 segments is presumably due to

a greater flocculation potential between adjacent particles at higher concentrations of

Tween 80. This type of flocculation, however, is of a different origin than normal

flocculation, which is favored in systems using low concentrations of high molecular

weight polymers that are not well solvated. High concentrations of the highly solvated

Tween 80 molecules can lead to the formation of admicelles, or nonionic surfactant

domains, of Tween 80 near the particle surfaces. These admicelles are aggregates of

adsorbed Tween 80 and Tween 80 in bulk solution that interacts with the adsorbed

molecules. When two particles surrounded by these nonionic surfactant domains

approach one another, the Tween 80 molecules have a tendency to aggregate with Tween

80 molecules on the adjacent particles, causing a loose floc to form, with the region

between particles having a high density of Tween 80 molecules. This type of

flocculation is illustrated in Figure 3-11. The region between particles may or may not

include additional Tween 80 molecules bound to the floc as shown in Figure 3-11 b). If




University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - - mvs