Colloidal Aspects of Chemical Mechanical Polishing (CMP)

1
Colloidal Aspects of Chemical Mechanical
Polishing (CMP)
Tanuja Gopal Jan Talbot Chemical Engineering
Program University of California, San Diego May
10, 2004
2
Outline

• Introduction
• Background Motivation
• Research Approach
• Experimental Results
• Conclusions
• Future Work

3
What is CMP?
CMP is a method through synergistic effects of
chemical and mechanical forces to achieve local
and global planarization of Integrated Circuit
(IC) structures.
Unplanarized
Surface smoothing
Local planarization
Global planarization
Ref. Steigerwald, J. M., Murarka, S. P. and R.
Gutmann, Chemical Mechanical Planarization of
Microelectronic Materials, Wiley and Sons, New
York (1997).
4
CMP Applications

• Oxide CMP
• Metal CMP

Ta
5
CMP Schematic
P 1.5-13 psi
V 20-60 rpm
wafer carrier
slurry
(100-300 ml/min)
wafer
polishing pad
(polyurethane)
platen head
wafer
slurry
polishing pad
6
CMP Parameters

• Process Variables
• Wafer down pressure
• Wafer velocity
• Pad characteristics
• Particle characteristics
• Slurry chemistry
• Substrate characteristics

• Process Results
• Material Removal Rate
• Planarization
• Surface finish

7
Typical Process Conditions

• Wafer
• Wafer rotational speed 20 - 60 rpm
• Applied pressure 1.5-13 psi
• Slurry
• Flow rate 100 - 300 ml per min
• Particle type silica, alumina, ceria, titania,
  etc.
• Particle concentration 1 - 30 by weight
• Particle size 50 - 1000 nm diameter
• Removal Rate
• SiO2 200 - 300 nm per minute
• Cu or W 300 - 600 nm per minute
• Planarization time 1- 3 min
• RMS roughness lt 1 nm

8
Mass Transfer Process

• (a) movement of solvent into the surface layer
  under load imposed by abrasive particle
• (b) surface dissolution under load
• (c) adsorption of dissolution products onto
  abrasive particle surface
• (d) re-adsorption of dissolution products
• (e) surface dissolution without a load
• (f) dissolution products washed away or dissolved

Dissolution products
Abrasive particle
Surface
Surface dissolution
Ref. L. M. Cook, J. Non-Crystalline Solids,
120, 152 (1990).
9
CMP Defects
Surface Particle
Residual Slurry
Ripout
Dishing
Embedded Particle
Micro-scatch
Ref. Philipossian et al. (2001)
10
Why CMP ?

• Multi-material surfaces
• Global planarization
• 200 and 300 mm (8 and 12 inch) wafers
• ICs have feature sizes lt0.2 ?m
• RMS roughness lt 1nm
• Disadvantages
• Large water consumption
• CMP defects
• End point detection

11
Motivation for Research

• Fundamental understanding of chemical effects in
  CMP
• Role of slurry chemistry not understood
  (additives, ionic strength, pH)
• Optimize slurries -high removal rates w/ adequate
  planarity
• Reduce consumables (slurries are expensive,
  mostly not recycled)
• Enhance post CMP cleaning large water usage
• Focus on Copper CMP Cu interconnect of choice
• Lack of comprehensive CMP model
• Lou and Dornfeld CMP mechanical model- add
  colloidal effects

12
Research Approach

• Experimental study of colloidal behavior of CMP
  slurries
• Zeta potential and particle size distribution
  measurements
• Function of pH, ionic strength, additives
• Commercial alumina slurries
• Alumina no additives
• Alumina in presence of common Cu CMP additives
• Agglomeration during CMP
• Incorporate colloidal chemistry into existing
  mechanical model by Lou and Dornfeld
• Average particle size, standard deviation
  parameters
• Comparison to literature material removal rates

13
Cu CMP Chemical Reactions
Dissolution Cu(s) HL ? CuL(aq) H e
Oxidation 2Cu H2O ? Cu2O 2H
2e Oxide dissolution Cu2O 3H2O ? 2CuO22-
6H 2e Complexation (to enhance
solubility) Cu2 HL ? CuL H
14
Pourbaix Diagrams

• Pourbaix diagrams-predicts stable phases in
  aqueous systems at equilibrium

copper-water-glycine system, LT10-1M
CuT10-5M
copper-water system, CuT10-5M
Ref. Aksu and Doyle (2002)
15
Colloidal Aspects of CMP
Interaction forces influence particle stability,
aggregation,deposition

1. Particle particle
2. Particle surface
3. Particle dissolution product
4. Surface dissolution product

16
Electrical Double Layer
Diffuse Layer
Shear Plane
Particle Surface

• Potential at surface usually stems from
  adsorption of lattice ions, H or OH-
• Potential is highly sensitive to chemistry of
  slurry
• Slurries are stable when all particles carry same
  charge electrical repulsion overcomes Van der
  Waals attractive forces
• Agglomeration may occur for z lt 5mV.

Potential
?
Distance
1/?
17
Measurement of Zeta Potential

• Particle velocity measured through microscope
  using rotating prism technique
• Pen Kem Lazer Zee Meter
• accuracy 5mV
• Brookhaven ZetaPlus
• accuracy 2
• particle size-light scattering

• calculated using Smoluchowski eqn
• (valid for ? a gtgt1)
• z v?/?E

z 30 mV stable z lt 5 mV agglomeration
18
Background Colloidal Effects

• Zeta potential and iso-electric point (IEP, pH
  where surface charge is neutral) of polished
  surface and abrasive particle is important

Polishing Regime
Zeta Potential (mV)
pH
Ref.Malik et al. (1997)
19
Colloidal effects

• Maximum polishing rates for glass observed
  compound IEP solution pH gt surface IEP
• (Cook, 1990)
• Polishing rate dependent upon colloidal
  particle - W in KIO3 slurries
• (Stein et al., J. Electrochem. Soc. 1999)

20
Agglomeration

• Agglomeration process of the slurry versus pH,
  additive concentration, and ion concentration

(Bellman et al., 2002)
21
Removal Rate in CMP

• Prestons Equation - most widely used model in
  CMP
• MRR KVP
• MRR Material removal rate K Preston
  constant P Pressure in the wafer- pad space
  V Linear pad- wafer velocity

• Drawbacks of Prestons Eqn
• Does not take into account chemical synergistic
  effects
• Fails to provide insight into the interaction
  process (particle size, concentration, pad
  variables etc.)

Ref. Luo and Dornfeld (1998)
22
Model Review

• Mechanical Models
• Boning (2001)
• ParametersP,V, pattern density, step height
• Discretize the chip to create a P profile then
  use Prestons Eqn. to calculate removal rate.
• Dornfeld (2001)
• Parameters P, V, pad hardness, pad roughness,
  abrasive size, abrasive geometry, wafer hardness
• MRR ?w N Vol
• ?w density of wafer
• N number of active abrasives
• Vol volume removed by single abrasive

23
Model Review

• Chemical Models
• Stein model (1999) MRR kPV/(1kPV)
• Main variables type of colloidal species and
  concentration
• Chemistry, particle size, P, V constant
• Found that MRR and temperature were functions of
  colloid species concentration
• Subramanian model (1999) mass transport model
• Chemical removal of material coupled with mass
  transport
• MRR lower than observed rates because excludes
  mechanical action
• Gutman (2000) MRR kO/(1kO)
• Main variable Oxidizer concentration
• MRR increases with oxidizer concentration upto
  saturation point (2 wt )

24
Model Review

• Synergistic Model
• Gokis (2000)- MRR results from abrasive and
  chemical action
• MRR kchem (RRmech)o kmech (RRchem)o
  (RRmech)o mechanical wear Ke PV (RRchem)o
  chem. dissolution kr exp(-E/RT)PCin
• kchem factor accounting for chemical
  modification
• kmech factor accounting for abrasive activation

25
Effects of glycine and H2O2 on Cu removal
rate
(Seal et al., 2003)
26
Experimental Study
A) Measurement of Zeta Potential

• Alumina, silica
• pH
• Ionic strength
• Ultrasonication
• Cu CMP additives
• Stability of colloidal particles

B) Measurement of particle size and distribution
as function of slurry chemistry

• Coagulation/ well-dispersed
• Bi-modal near IEP

27
Research Study

• Experiments
• Ceralox alumina
• DI H2O
• w/ KCl to alter ionic strength (Babu et al.,
  2000)
• Commercial alumina slurries from Stein (Sandia
  National Laboratories)
• EKC Tech slurry (Doyle, UCB)- Cu CMP slurry
  additives
• Model MRR predictions vs. literature experimental
  polishing data
• Average particle size and standard deviations
  used in Lou and Dornfeld model

28
Alumina particles in DDI H2O
(Sumitomo Chem. Co.,250 nm)
(Ceralox, 300 nm)
IEP ? 9
29
Ceralox alumina ionic strength
Ionic Strength 10-4 to 10-7M
30
z vs. pH for Ceralox alumina particles with 10-3M
KNO3

• IEP 9, agglomeration
• Broader distribution near IEP
• Average size 300 nm

• Standard deviation
• pH 3.5-7 10 nm
• pH 9 300 nm

31
Common Cu slurry additives
Additives Name Concentration
Buffering agent NH4OH, KOH, HNO3 bulk pH 3-8
Complexing agent Glycine Ethylene-diamine-tetra-acetate (EDTA) citric acid 0.01-0.1M
Corrosion inhibitor Benzotriazole (BTA) 3-amino-triazole (ATA) KI 0.01-1wt
Oxidizer H2O2, KIO3, K3Fe(CN) citric acid 0-2 wt
Surfactant Sodium-dodecyl-sulfate (SDS), cetyltrimethyl-ammonium-bromide (CTAB) 1-20 mM
32
z and particle size vs. pH for EKC Tech alumina
with 10-3M KNO3

• IEP 9 ? agglomeration
• z varied by15
• 200 nm - pHlt8

• particle size standard deviation
• z lt 5nm for pHgt8
• z gt 300 nm for pHlt8

33
z and particle size vs. pH for EKC Tech alumina
with 10-3M KNO3 and glycine

• IEP 9, agglomeration
• z varied by 2
• 200 nm pHlt8

34
z and particle size vs. pH for EKC Tech alumina
with 10-3M SDS and 10-3M KNO3
pH 6

• z ranged from -34 to -46 mV
• Average particle size 220nm (approximately
  double stated size)
• Particle size standard deviation small (lt 5nm)

35
z and particle size vs. pH for EKC Tech alumina
with 0.01 wt BTA or 0.01M EDTA 10-3M KNO3

• BTA - no effect
• EDTA - shifted IEP to pH 5, large particles

36
Lou and Dornfeld Mechanical Model
Basic Eqn. of Material Removal MRR N x Vol
Ref. Lou and Dornfeld (2001)
37
Overall Research Approach
Comprehensive Model (Dornfeld, 2003) a)
Mechanical effects (Dornfeld et al., UCB) b)
Electrochemical effects (Doyle et al., UCB) c)
Colloidal effects (Talbot Gopal, UCSD)

• Si Wafer
• Pressure 1.5 psi
• Velocity 2-12 rpm
• Polishing time 2-4 hours

Slurry film thickness (mm)
(Moon and Dornfeld et al. 1999)
38
Model Sensitivity to Standard Dev.

• Simplified dependency on standard deviation
• For xavg lt500 nm small variation s results in
  large change in MRR

39
Collision Efficiency

• CMP 104-106 s-1
• Collison Efficiency(ao)-fraction collisions ?
  permanent attachment
• Most particles do not agglomerate

106
104
105
40
Maximum Aggregate Size
Effective Particle Size (nm) Max. Aggregate Size (nm)
Shear rate 104s-1 Shear rate 104s-1
100 180
200 or greater Total aggregate break up
Shear rate 103s-1 Shear rate 103s-1
100 1800
200 900
300 600
400 or greater Total aggregate break up

Rmax
41

• P 1 psi, 4 inch blanket wafer, wafer carrier
  platen velocity 100 rpm, pad hardness 100 MP,
  passivation rate 100 nm/min

42
MRR prediction and particle size for alumina with
and without glycine
0.1 M glycine
No additives

• Max. MRR 160 nm/min without additives
• Max. MRR 120nm/min with 0.1M glycine

43
MRR prediction and particle size for alumina with
glycine and hydrogen peroxide
0.1M glycine, 2 wt H2O2
0.1M glycine, 0.1wt H2O2

• Max. MRR 170 nm/min with 0.1 wt H2O2
• Max. MRR 220 nm/min with 2 wt H2O2

44
MRR prediction and particle size for alumina with
Cu slurry additives
0.01wt BTA, 10-3M SDS, 0.01M EDTA, 0.1wt H2O2,
0.01wt BTA, 10-3M SDS, 0.1M glycine, 0.1wt
H2O2,

• MRR 1-10 nm/min
• Particle size 0.5 -3 microns

45
Summary- effects of additives
Additive Effect
Glycine z stabilizing agent
BTA No effect
EDTA Unstable, agglomeration
SDS 2x agglomeration, stable, negative ?
46
Conclusions

• Background electrolyte
• Particle size distribution vs. IEP
• Effects of Cu polishing rates w/different
  chemistries
• Cu-glycine complexes in presence of H2O2 result
  in increased MRR
• Slurry additives affect colloidal behavior pH
  largest effect
• Lou and Dornfeld model
• Can predict trends well
• Model is sensitive to variation of z

47
Future Work

• Cu CMP Experiments
• Slurry additives glycine, hydrogen peroxide
• Zeta potential w/ dissolved Cu or Cu particles
• Model improvements Use actual particle
  distribution Surface hardness link to
  chemistry Passivation rate of Cu (Doyle)
• Adhesion tests post-CMP cleaning