The Effects of Slurry Chemistry on theThe Effects of Slurry Chemistry on theColloidal Behavior of Alumina Slurries andColloidal Behavior of Alumina Slurries and
Copper Copper Nanohardness Nanohardness for Copper Chemicalfor Copper ChemicalMechanical Mechanical Planarization Planarization (CMP)(CMP)
Jan TalbotJan Talbot && Robin Robin Ihnfeldt Ihnfeldt (PhD 2008)*(PhD 2008)*
University of California, San DiegoUniversity of California, San DiegoSept. 22, 2008Sept. 22, 2008
*Apex Medical Technologies, Inc., San Diego
22
CMP - Faculty TeamCMP - Faculty Team
3
Outline- Background
- CMP process- Slurry chemistry and role of abrasives- Research objectives and motivation
- Measuring Colloidal Behavior – Effects of Chemistry- Previous work from T. Gopal (Ph.D. 2004)
- Measuring Nanohardness and Etch Rate – Effects of Chemistry- Review of Luo and Dornfeld CMP Model
- Do agglomerates break up?- Model sensitivity to agglomerate size, distribution, and hardness
- Modeling Results Compared to Experiment- Conclusions- Future Work- Acknowledgments
4
Integrated Circuit (IC) manufacturing requires material removaland global planarity – Chemical Mechanical Planarization (CMP)
CMP slurries provide materialremoval by:Mechanical Abrasion
Nanometer sized abrasiveparticles (alumina)
Chemical ReactionChemical additives (glycine, H2O2,etc.)
Material Removal Rate (MRR)is affected by: Abrasive size and sizedistribution Wafer surface hardness
Cu is interconnect ofchoice- our research focus
Introduction
5
CMP Schematic
Particle concentration = 1 - 30 wt%Particle size = 50 - 1000 nm dia
slurry(100-300 ml/min)
platen
polishing pad
wafer
slurry
wafer carrier
P = 1.5-13 psi
V= 20-60 rpm
Cu MRR= 50 - 600 nm/minPlanarization time = 1- 3 minRMS roughness = < 1 nm
polishing pad
(polyurethane)
6
CMP History-Introduced in 1980’s by IBM-Shrinking device dimensions in IC manufacturing lead to uneven surfaces
-Caused depth of focus problems for photolithography-CMP provided local and global planarization that could not previously be obtained bychemical etching
1http://www.tf.uni-kiel.de/matwis/amat/elmat_en/kap_5/illustr/i5_1_1.html accessed April 20082K. Cadien, “Is There a Future For CMP?” Proceedings for the 13th International CMP for ULSI Multilevel Interconnection
Conference (CMP-MIC), Keynote presentation, Fremont, California, March 4, 2008.
16M DRAM1
8 metal layers2
(state of the art)65nm
1µm
No CMP
7
Copper CMP SlurryAbrasives Most CMP slurries contain alumina or silica abrasives Abrasives may scratch and cause defects
Slurries with abrasives need to be tailored so particle sizes and distributions aresmall enough to not cause scratches
Chemical additives – Surfactants, Complexing agents, Inhibitors, OxidizersChemicals may cause corrosion and low MRRs
Studies of chemistry effects on colloidal behavior of abrasives and thewafer surface are needed
8
Research Objectives
Understand effects of slurry chemistry on Cu CMPprocess Colloidal behavior measured by zeta potential and
agglomerate size distribution - effects of chemicaladditives and presence of copper
Copper surface hardness and etch rate as function ofchemistry
Infer state of the Cu (Cu, CuO, Cu2+, etc.) in slurry andon surface
Use colloidal behavior, nanohardness and etch rates inmodel of CMP
9
Motivation Better process control
Understand role of slurry chemistry (additives, pH, etc.) Develop slurries to provide adequate removal rates and global
planarity
Prediction of material removal rates (MRR) Predictive CMP models - optimize process consumables Improve understanding of effects of CMP variables Reduce cost of CMP
Reduce defects Control of abrasive particle size Control of interactions between the wafer surface and the slurry
10
Slurry Abrasives40 wt% α-alumina slurry (from Cabot Corp.)150nm average aggregate diameter – 20nm primary particle diameter
Common Copper CMP Slurry AdditivesGlycine, ethylene-diamene-tetra-acetic acid (EDTA), hydrogen peroxide(H2O2), benzotriazole (BTA), sodium-dodecyl-sulfate (SDS)
Copper nano-particlesAdded to simulate removal of copper surface during CMP<100 nm in diameter (from Aldrich)
Zeta Potential and Agglomerate Size DistributionBrookhaven ZetaPlus
Zeta Potential – Electrophoretic light scattering technique (±2%)Agglomerate Size – Quasi-elastic light scattering (QELS) technique (±1%)
All samples diluted to 0.05 wt% in a 1 mM KNO3 solutionSolution pH adjusted using KOH and HNO3 and ultrasonicated for 5 minprior to measuring
Measuring Colloidal Behavior –Experimental Procedure
11
Zeta PotentialDiffuse Layer
ShearPlane
ParticleSurface
1/κ
Zeta Potential - Potential at the Stern LayerElectrophoresis – Zeta potential estimated byapplying electric field and measuring particlevelocity
+ +
++
+
+ +++++
++
++
++
+
+
++
a
+
+
+
Distance
Pot
entia
l
ζ
Stern Layer
-60
-40
-20
0
20
40
60
80
2 4 6 8 10 12
pH
t Z
eta
Po
ten
tial (m
V)
0
1000
2000
3000
4000
g A
gg
lom
era
te S
ize (
nm
)Surface charge on metal
oxides is pH dependant:
Isoelectric point (IEP) at ζ = 0 Slurries are stable when |ζ | > 25 mV
M-OH + H+ → M-OH2+
M-OH + OH- → M-O- + H2O
Cabot alumina without additives in 10-3M KNO3 solution(bars indicate standard deviation of agglomeratesize distribution)
12
Previous WorkPrevious work done by T. Gopal (PhD_2004). Colloidal Behavior - Zeta potential and particle size distribution
measurements were taken for EKC Tech alumina Various common copper slurry additives were investigated Solution pH varied from 3-10
Experimental Results1: pH had the largest effect on particle size Additives typically decreased magnitude of zeta potential, and increased
particle sizes Modeling - Incorporated particle size distributions into Luo and
Dornfeld model and compared to experimental CMP from literatureModeling Results2:
Model capable of predicting trends of MRR with changes in chemistry asobserved in literature
1T. Gopal and J. B. Talbot, J. Electrochemical Soc., 153, G622 (2006). 2T. Gopal and J. B. Talbot, J. Electrochemical Soc., 154, H507 (2007).
13
Zeta PotentialCabot alumina in 10-3M KNO3 solution with and without 0.12mM copper
IEP ~6.5 with and without copper IEP~9.2 for α-alumina from literature* Presence of NO3
-, SO42-, etc., can lower IEP**
At high pH values magnitude of zeta potential is smaller with copperthan without
*M.R. Oliver, Chemical-Mechanical Planarization of Semiconductor Material, Springer-Verlag, Berlin (2004).**G.A. Parks, Chem. Tevs., 65, 177 (1965).
-80
-40
0
40
80
0 2 4 6 8 10 12
pH
Zeta
Po
ten
tial (m
V)
Without Cu
With Cu
14
Agglomerate Size Distribution
pH 2 – presence of copper causes decrease in agglomeration pH 7 – presence of copper causes increase in agglomeration
Cabot alumina dispersion in 1mM KNO3 solution with (red) and without(blue) 0.12 mM copper and without chemical additives
pH 2
0
10
20
30
0 2 4
Agglomerate size (µm)
% in
so
luti
on
pH 7
0
10
20
30
40
0 5 10
Agglomerate size (µm)
% in
so
lutio
n
15
Potential-pH forCopper-waterSystem[Cu]=10-4M at 250C and1atm (M. Pourbaix 1957)
■ Agglomeration behavior is consistent with the Pourbaix diagram
Copper-Alumina-Water System
Average agglomerate size of bimodal distributions in a 1 mM KNO3 solution
IEP of CuO ~ 9.5*
*G.A. Parks, Chem. Tevs., 65, 177 (1965).
Small
Average (nm)
Large
Average (nm)
Small
Average (nm)
Large
Average (nm)
2 Cu, Cu+
170 5000 160 810
7 Cu, Cu 2O, CuO, Cu(OH) 2 580 3300 1700 9400
10 Cu, Cu 2O, CuO, Cu(OH) 2 150 720 300 1600
Without Copper With Copper
Possible State of CopperpH
16
Zeta PotentialCabot alumina in 0.1M glycine and 10-3M KNO3 solution with and
without 0.12mM copper
IEP ~6.5 without copper IEP~9.2 increased with copper
*M.R. Oliver, Chemical-Mechanical Planarization of Semiconductor Material, Springer-Verlag, Berlin (2004).**G.A. Parks, Chem. Tevs., 65, 177 (1965).
-80
-40
0
40
80
0 2 4 6 8 10 12
pH
Zeta
Po
ten
tial (m
V) Without Cu
With Cu
17
Potential-pH for Copper-Glycine-Water System*[Cu]=10-4M, [Glycine]=10-1M at 250Cand 1atm
Agglomeration behavior is consistent with Pourbaix diagram
Average agglomerate size of bimodaldistributions in a 1 mM KNO3 solution withvarious additives
Copper-Glycine-Water System
*S. Aksu and F. M. Doyle, J. Electrochemical Soc., 148, 1, B51 (2006).
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
0 2 4 6 8 10 12 14 16pH
E, V
vs. S
HE
Cu2+
CuO/
Cu(OH)2
CuCu2O
CuO22-CuHL2+
CuL+
CuL2
CuL2-
Small
Average (nm)
Large
Average (nm)
Small Average
(nm)
Large
Average (nm)
2 Cu, CuHL2+
310 8100 220 700
7 Cu, CuL 2 2000 1900
10 Cu, CuL 2-, CuL 2 1030 6300 350 2100
Possible State of
Copper
Without Copper With Copper
Solution
0.1M Glycine
pH
18
Hardness Measurements -TriboScope Nanomechanical Testingsystem, Hysitron Inc.1000 nm Cu sputter deposited on 30nm Ta on 1 cm2 silicon wafer pieces10min exposure in 100ml of solution(without abrasives), removed, driedwith air and measuredMaximum applied load varied from50-3000 µNEtch Rates - wafer pieces weighedbefore and after immersion in solution
Experimental Procedure – MeasuringHardness and Etch Rates
19
Measuring Hardness and Etch Rates
■ Considerations* - δmin - Applied load must be large enough to induce plastic deformation
Indentations depths must be >5nm δmax - Applied load must be small enough to not have effects of
underlying layer – even with high etch rate slurry solutions whichdecrease copper thickness
General Rule δmax < 0.1 x film thickness ~100nm
*N. Ye and K. Komvopoulos, J. Tribology, 125, 685 (2003).
20
Nanohardness Before Chemical Exposure
1D. Beegan, S. Chowdhury, and M. T. Laugier, Surface and Coatings Technology, 210, 5804 (2007).2A. Jindal and S.V. Babu, J. Electrochemical Soc., 151 (10), G709-G716 (2004).3M. Ueda, C. M. Lepienski, E.C. Rangel, N. C. Cruz, and F. G. Dias, Surface and Coatings Technology, 156, 190 (2002).4A. Szymanski and J. M. Szymanski, Hardness Estimation of Minerals Rocks and Ceramic Materials, Elsevier Science Publishers B.
V., New York, NY (1989).
0
2
4
6
8
10
0 50 100 150 200 250
Indentation Depth (nm)
Hw
(G
Pa)
Average Hw=2.6 GPa Hardness technique Material Value
Nanohardness (GPa) Cu1
2.5 ± 0.3
Ta2
3.3 ± 1.5
Si3
12 ± 2
Cu2O* 17 ± 5
Moh's hardness4
Cu(OH)2 2.0-2.5
Cu metal 3
CuO 3.5
Cu2O 4
Ta 6.5
Si 6.5
*as measured in this study
Hardness values
Nanohardness vs indentation depthNanohardness near surface (<20nm) >Cu metalPossibly formation of copper oxide
At indentation depth >30nm, hardness of Cu metal
21*M. Pourbaix, Atlas of Electrochemical Equilibria in Aqueous Solutions, National Association of CorrosionEngineers, Houston, Texas (1974).
Hardness values near surface (<40nm) increase as the pH increasesConsistent with potential-pH equilibrium diagrams which indicate that copperoxides are more stable at higher pH*Nanohardness is that of Cu metal for indentation depths >70nm
0
1
2
3
4
5
6
7
8
0 20 40 60 80 100 120 140
Indentation Depth (nm)
Hw
(G
Pa
)pH 2.9
pH 8.3
pH 11.7
Effect of pH on Hardness
22
Nanohardness vs indentation depth after exposure tosolutions with 0.1M glycine and 2wt% H2O2 at various pH Possible Surface Reactions:
passivation2Cu + H2O → Cu2O + 2H+ + 2eCu2O + H2O → 2CuO + 2H+ + 2e
complex formationCu2O+4HL→2CuL2(s)+H2O+2H++2eCuO + 2HL →CuL2 (s) + H2O
dissolutionCuL2 (s) → CuL2 (l)
decompositionH2O2 + e- → OH* + OH-
* G. Xu, H. Liang, J. Zhao, and Y. Li, J. Electrochemical Soc., 151, (10) G688 (2004).
Effect of Additives on Hardness
At pH 8.3 film is very soft & etch rate is very large (56 nm/min)H2O2 decomposition occurs faster at higher pH.*At pH 10.0, large etch rate, 33 nm/min, but a thick passivation layer possiblyforms which inhibits Cu-glycine complex formation.
0
2
4
6
8
10
12
14
16
18
20
0 25 50 75 100 125 150Indentation Depth (nm)
Hw
(G
Pa)
0.1M Glycine, 2.0wt% H2O2 pH 8.3
0.1M Glycine, 2.0wt% H2O2 pH 10.0
23
Hardness sensitiveto chemistry!(depends on pH, etchrate, and chemicaladditives)
Solutions withoutetching had harderfilms
Solutions with largeetch rates (>14nm/min) have soft(possibly porous) films
For most cases thehardness was that of Cumetal at indentationdepths >70nm
Aqueous solution
with 1mM KNO 3
plus: pH
Etch rate
(±4nm/min)
Hardness for <80nm
indentation depth
(±0.3GPa)
2.9 0.7 1.6-4.6
8.3 0.0 1.8-5.7
11.7 2.6 2.0-7.3
3.1 1.2 1.0-3.5
8.5 7.6 1.6-5.2
10.0 0.0 3.5-16
3.0 45 1.0-4.4
8.3 33 2.1-5.6
10.1 14 0.34-4.7
3.0 38 2.8-8.2
8.3 56 0.1-5.5*
10.0 33 2.3-18
3.0 1.6 1.2-3.3
8.4 0.0 1.7-5.3
10.8 8.6 2.0-4.0*
2.6 0.0 0.9-11*
9.0 0.0 2.5-8.0
10.9 0.0 2.1-13
*Values are different than Cu metal for indentation depths >80nm
0.1 M glycine, 2.0
wt% H 2O2
Combination w/
glycine
Combination w/
EDTA
Etch rates and nanohardness with indentation
depth <80nm
none
0.1 M glycine
0.1 M glycine, 0.1
wt% H 2O2
Nanohardness and Etch RatesFor some cases:
24
Experimental Copper CMPToyoda Polishing Machine1000nm Cu film sputter deposited on 30nm Ta on siliconwafers (100mm diameter)2 min polishing time
Slurry -150 ml/min
IC1000 polishingpadPlaten -
30 rpm
wafer carrier -30 rpm
P0 = 1 psi
25
Luo and Dornfeld CMP Model(IEEE Trans. on Semi. Manuf., 2001)
Overall mass material removal (MRR) rate during CMP:
0CNVMRR
W+= !
where ρw = density of copper surface (g/m3)N = number of active abrasivesV= average volume removed by a single abrasive (m3/min)C0 = removal due to chemical etching (g/min) (negligible?)
*Assume Solid-Solid contact mode.
26
Elastic Deformation of Pad Soft pad Abrasives are small Uniform distribution of asperities Asperities have same radius and
heightPlastic deformation of wafer by
abrasive Small cutting depths (<1µm) cause
brittle materials to be removed inductile manner
Abrasives - spherical geometry Removal due to sliding indentation of
abrasive Normal distribution xact > xavg
Model Assumptions
*J. Luo and D. Dornfeld, IEEE Trans. Semicond. Manuf., 14, 2 (2001).
27
Luo and Dornfeld CMP ModelMaterial removal rate based on thickness:
Thickness Removed (nm/min) Probability of ActiveAbrasives (value = 0-0.5)
*J. Luo and D. Dornfeld, IEEE Trans. Semicond. Manuf., 14, 2 (2001).
Model requires input values for:xavg – average abrasive sizeσ – standard deviation of abrasive size distributionHW – wafer surface hardnessC0 – etch rate
xavg and σ measured under quiescent conditionsDo agglomerates break up during CMP?
( )023
2
223
1321
31961
Cx
HHB
xxxH
BMRR
avg
WPavgavgavgW
+!!"
#
$$%
&
''(
)**+
, +
''(
)**+
,+-.-
''
(
)
**
+
,++=
/
///
vP
R
alEDmdB
a
PSUMasss
0
25
21
1
3
28
!"
" #=31
0
32
34
2
3
4
4
1
!!"
#$$%
&!"
#$%
&!"
#$%
&=
SUM
P
D
P
R
EB
'
EtchRate
28
Shear Force on AbrasivesShear force is directly
proportional to normalforce1,2
µf – coefficient of frictionbetween wafer andpad asperities tips ~0.561 (at vel=0.3m/s)
FN – normal force = PC AS
PC – contact pressure asgiven in L&D model3
AS-surface area of contacton abrasive
For x = 200nmFS = 3.6x10-10 N
1J. Seok, C.P.Sukam, A. T. Kim, J.A. Tichy, T. S. Cale, Wear, 257, 496 (2004).2S. Tsai, L. Chen, L. Sun, R. Mavliev, W. Hsu, L. Xia, R. Morad, IEEE, 7803, 7216 (2002)3J. Luo and D. Dornfeld, IEEE Trans. Semi. Manuf., 14, 112 (2001).
22
42C
W
fWCfSCfNfS Px
H
xPAPFF
µ!!µµµ ="
#
$%&
' (===
29
Force to Break Up AgglomeratesForce required to break up an agglomerate1:
( )SNsBCFF += µ
µs – static coefficient of frictionbetween particles = 0.681
FN – normal force pressing particlestogether = PC AB
CS – cohesive strength of theagglomerate
1R. L. Brown and J. C. Richards, Principles of Powder Mechanics, Pergamon Press, Elmsford, New York (1970).2W. Pietsch, Agglomeration Processes, Wiley-VCH Verlag GmbH, Weinheim, 2002.3J. K. Beddow and T. P. Meloy, Advanced Particluate Morphology, CRC Press, Boca Raton, Florida, 1980.
Cohesive strength of anagglomerate filled with liquid
(Rumpf model)2: a’ – correction factor, has values between 6-82
ε – porosity of agglomerate – (random loosepacking of spheres ε = 0.399)3
α – surface tension of liquid (H20) 2 = 72 dyn/cmdP – primary particle diameter = 20 nm
( )
p
Sd
aC!
"
"#=
1'
( )SCC
W
s
BCPP
H
xF +=
2
2!µFor x = 200 nm
FB = 5.0 x 10-9N
Agglomerates Do NOT Break Up! FB > FS
30
Maximum Agglomerate Size
0
5
10
15
20
25
30
0 2000 4000 6000 8000
Particle diameter (nm)
% in
so
luti
on
Someagglomerates arelarge (>1µm)
Some distributionsare bimodal
Maximumagglomerate sizestable in fluid1,2:
2
1
2max
182 !
"
#$%
&=
'()r
HrD
1M.Elimelech, J. Gregory,X. Jia, and R. Williams, Particle Deposition and Aggregation, Butterworth-Heinemann, Oxford, Great Britian (1995).
2R. Ihnfeldt and J. B. Talbot, J. Electrochemical Soc., to be published (2007).
Alumina dispersion in 0.1M glycine, 0.1wt% H2O2, 0.12mM Cu and 1mM KNO3solution at pH (♦) 8.3 or (■) 10.0.
DMAX ~ 4.8 µm
Parameter Value
Hamaker Constant, H 4.17 x 10-20
J
Primary particle radius, r 10 nm
separation of particles, ! 0.4 nm
shear stress, " 2.1 Pa
31
Luo and Dornfeld CMP Model*Material removal rate based on thickness:
* J. Luo and D. Dornfeld, IEEE Trans. on Semi. Manuf., 14, 2, 2001.
( )023
2
223
1321
31961
Cx
HHB
xxxH
BMRR
avg
WPavgavgavgW
+!!"
#
$$%
&
''(
)**+
, +
''(
)**+
,+-.-
''
(
)
**
+
,++=
/
///
For our Copper CMP experimental conditions:B1 = 3.18 x 107 nm2GPa1.5/min B2 = 1720 kPa
Complicated function of xavg, σ and HW
To predict MRR:For bimodal distributions, xavg and σ from distributions with xavg<DMAX
HW is average of values with applied loads between 50-100µNExperimentally measured CO is incorporated
Thickness Removed (nm/min) Probability of ActiveAbrasives (value = 0-0.5)
EtchRate
32
0
100
200
300
400
0 2 4 6 8 10 12pH
MR
R (
nm
/min
)
ExperimentModel w/ constant HwModel w/ Hw and Co
Experimental and model MRR predictions using 1mM KNO3 solution
Model overpredicts MRR at low pH Predictions at 8.3 and 11.7 agree well with experiment – low etch rate – not
sensitive to HW
Model Predictions
33
0
100
200
300
400
0 2 4 6 8 10 12pH
MR
R (
nm
/min
)
Experiment
Model w/ const Hw
Model w/ Hw and Co
Experimental and model MRR predictions using 1mM KNO3, 0.1M glycine and 0.1wt% H2O2solution
Model predictions do not agree with experiment for all pH values At low pH, meas. HW is too small At pH>8, solution in very chemically active (large etch rates >8nm/min),
model is very sensitive to HW
HW under quiescent conditions different from HW during CMP Surface film to thin to measure using our technique
Model Predictions
34
Model Sensitivity
!!
"
#
$$
%
&++!!
"
#$$%
&3
2
2
2
6 961
min10
avgavgavgxxx
nm ''
0
0.1
0.2
0.3
0.4
0.5
0.6
1 10 100 1000 10000nm
Pro
ba
bil
ity
of
AA
xavg!
0
0.01
0.02
50 150 250 350
xavg → large, xact → xavg and AA→0.5σ → small, xact → xavg and AA →0.5
xavg=270 nmσ=7 nm
MRR predictions exhibit sensitive dependence on xavg and σ Small changes in xavg or σ can cause large differences in predicted
MRR
1
10
100
1000
10000
1 10 100 1000 10000xavg- solid, !- dashed (nm)
MR
R (
nm
/min
)
!=7
!=20
!=100
!
xact=xavg +3σ
35
Model Sensitivitya) HW decreases with increase in
xavg from 100 to 200 nm HW increases with increase in
xavg from 200 to 1000 nm
a)
0
100
200
300
400
0 2 4 6 8 10 12 14 16 18 20
HW (GPa)
Pre
dic
ted
MR
R (
nm
/min
)
100nm
200nm
1000nm
xavg
b)
0
100
200
300
400
0 2 4 6 8 10 12 14 16 18 20
HW (GPa)
Pre
dic
ted
MR
R (
nm
/min
)
10nm
50nm
200nm
!
b) For small MRR predictions modelis insensitive to HW, and for largeMRR predictions model is verysensitive to HW
36
a)
0
50
100
150
0 10 20
HW (GPa)
MR
R (
nm
/min
)
pH 2.9
pH 8.3
pH 11.7
a) xavg and σ values smaller at low pH compared to alkaline conditionsb) Meas. HW to small, need larger HN to achieve exp. MRR.
Thin harder film on surface which cannot be measured using our technique Surface film under quiescent conditions different than during CMP
Model Predictions
b)
0
50
100
150
0 10 20
HW (GPa)
MR
R (
nm
/min
) pH 2.9
Meas.HW
MRR
Prediction
Exp. MRR
Range
HN
MRR model predictions using measured xavg, σ, and C0 for 1mM KNO3 solution, withvarying HW
At pH 8.3 and 11.7 solutions are less chemically active (low etch rate) – not sensitive to HW
37
Model Sensitivity
38
ConclusionsColloidal Behavior pH has greatest effect on colloidal behavior Presence of Cu nanoparticles will increase or decrease
agglomeration depending on state of copper in solution Agglomeration behavior with copper is consistent with
potential-pH diagramsNanohardness and Etch Rates of Copper Surface HW can vary from 0.1-20 GPa depending on chemistry (pH,
additives) due to: formation of different films (CuO, Cu2O, etc.) and/or changes in
compactness from dissolution or complexing reactions HW very sensitive to chemistry
depends on pH, etch rate, and additives For most cases, state of copper on surface is consistent with
potential-pH diagrams
39
Modeling Agglomerated abrasive particles do not break up during CMP Agglomerate size measurements can be used in Luo and
Dornfeld model Model predictions improved using measured HW and C0
compared to using constant HW of Cu metal and neglecting C0 Model only agrees with experiment for solutions with small
etch rates (<8nm/min) and pH>8 MRR is overpredicted for acidic solutions MRR does not agree with experiment for solutions with large
etch rates (>8nm/min) and pH>8 HW under quiescent conditions is different than HW during CMP film is too thin to measure using our technique
Luo and Dornfeld model is useful for Cu CMP
Conclusions
40
Future WorkContinue to develop basic understanding of agglomeration/
dispersion effects on CMP Study rate of agglomeration as a function of chemistry Investigate effects of heating (temperature) on agglomeration and surface
hardness as a function of chemistrySurface Hardness Study effects of exposure time of solution on hardness (film formation due to
fast reaction or slow reaction) Measure nanohardness after exposure to flowing slurry How to probe 1-5nm range? Adjust equipment parameters, use different
AFM tip?Modeling Develop rate of agglomeration model as a function of chemistry which
includes the effects of heating (temperature) Incorporate rate of agglomeration and heating effects into Luo and Dornfeld
CMP model Other Models?
41
Funded by FLCC Consortium through a UC Discovery grant. Wegratefully acknowledge the companies involved in the UC Discoverygrant: Advanced Micro Devices, Applied Materials, Atmel, Cadence,Canon, Cymer, DuPont, Ebara, Intel, KLA-Tencor, Mentor Graphics,Nikon Research, Novellus Systems, Panoramic Technologies,Photronics, Synopsis, Tokyo Electron
Collaboration of Prof. Komvopoulos and Doyle and their researchgroups from UC Berkeley
Prof. Nesterenko from UC San Diego provided expertise indeveloping criteria to determine when agglomerates break apart
Prof. Dornfeld and his research group at UC Berkeley for use ofthe CMP apparatus and model program
Prof. Bavarian at CalStateU-Northridge, and Prof. Talke at UCSDfor the use of the Hysitron Instrument.
Acknowledgments
Prof. Arrhenius from UC San Diego providedthe cuprite samples
Evelyn York for the EDX analysis
42
PublicationsJournal Papers Modeling Material Removal Rates for Copper CMP Using Copper Nanohardness and Etch Rates,
R.V. Ihnfeldt and J.B. Talbot, J. Electrochemical Soc., submitted March 2008. Effect of CMP Slurry Chemistry on Copper Nanohardness, R.V. Ihnfeldt and J. B. Talbot, J.
Electrochemical Soc., 155, H412 (2008). Modeling of Copper CMP Using the Colloidal Behavior of an Alumina Slurry with Copper
Nanoparticles, R.Ihnfeldt and J.Talbot, J. Electrochemical Soc., 154, (12) H1018 (2007). The Effects of Copper CMP Slurry Chemistry on the Colloidal Behavior of Alumina Abrasives, R.
Ihnfeldt and J.B. Talbot. J. Electrochemical Soc., 153, G948 (2006).
Conference Papers and Presentations Modeling Copper CMP Material Removal Rates Using Copper Surface Nanohardness, R. V.
Ihnfeldt and Jan B. Talbot, to be presented at the 213th Meeting of the Electrochemical Society,Phoenix, Arizona, May 18-23, 2008.
Copper CMP Removal Rate Predictions Using Alumina Agglomerate Size Distributions, R.V.Ihnfeldt and J.B. Talbot, VLSI Multilevel Intercon. Conf.(VMIC), Sept.25-27, 2007, Fremont, CA.
Effects of CMP Slurry Chemistry on Agglomeration of Alumina and Copper Surface Hardness, R.Ihnfeldt and J.B. Talbot, The 210th Meeting of the Electrochemical Society, Cancun, Mexico, Oct.29-Nov. 3, 2006, 3 (41), 21.
Conference Poster Presentations 2nd Place ICPT 2006 Outstanding Poster Award - Copper Removal Rate Predictions Using
Alumina Agglomerate Size Distribution and Copper Nanohardness Measurements, R. Ihnfeldt andJ.B. Talbot. 2006 International Conference on Planarization/CMP Technology, Oct. 12-13, FosterCity, CA.
43
Solutions
a) 1 mM KNO 3
b) 1 mM KNO 3, 0.1 M Glycine
c) 1 mM KNO 3, 0.1 M Glycine, 0.1 wt% H 2O2
d) 1 mM KNO 3, 0.1 M Glycine, 2.0 wt% H 2O2
e) 1 mM KNO 3, 0.1 M Glycine, 0.1 wt% H 2O2, 0.01 wt% BTA, 0.1 mM SDS
f) 1 mM KNO 3, 0.01 M EDTA, 0.1 wt% H 2O2, 0.01 wt% BTA, 0.1 mM SDS
Concentration of additives
6 different solutions studied Glycine and EDTA – complexing agents H2O2 – oxidizing agent BTA- corrosion inhibitor SDS – anionic surfactant KNO3 added to maintain constant ionic strength
44
Indentation Depth vs Load BeforeChemical Exposure
Indentation depth versus load Indentation depth increases as load increases Film is spatially uniform across the surface
0
50
100
150
200
250
0 500 1000 1500 2000 2500 3000
Load (µN)
Ind
en
tati
on
De
pth
(n
m)
fd
45
Indentation depth does not always increase as the load increasessurface may be nonuniform with different types of films (oxides, hydroxides,etc.)
Non-uniform Surface
0
20
40
60
80
100
120
140
0 200 400 600 800 1000
Load (µN)
Ind
en
tati
on
Dep
th (
nm
)
fd
pH 2.9
pH 8.3
pH 11.7
46
0
100
200
300
400
0 2 4 6 8 10
HW (GPa)
MR
R (
nm
/min
)
pH 8.3
Meas.HW
MRR
Prediction
Exp. MRR +/-14
HN
Model Predictions