characterization and modeling of cmp pad asperity propertiescharacterization and modeling of cmp pad...
Post on 09-Feb-2020
21 Views
Preview:
TRANSCRIPT
Characterization and Modeling of CMP Pad Asperity Properties
Wei FanWei FanMicrosystems Technology Laboratories, EECS, MIT
December 16 2010December 16, 2010
CMP System and ConsumablesSurface Asperities
Cells
IC1000 pad grooves IC1000 pad micro structure
Abrasive particles in slurry Conditioning disk Diamonds on conditioning disk
December 16, 20102
Abrasive particles in slurry Conditioning disk Diamonds on conditioning disk
Polishing Pad Properties
• Polishing pad is typically made of polyurethaneade o po yu et a e
• Pad is designed to– Provide elastic responsep– Transport slurries
• Pad surfaceSEM of conditioned pad
L. Borucki, 2004
– Consists of asperities with range of sizes around 50mTail of the pad surface– Tail of the pad surface height is ≈ exponentiallydistributed
December 16, 2010
Surface Height Distribution L. Borucki, 2006, ICPT
3
Pad Response Model:Bulk and SurfaceBulk and Surface
• Pad is assumed to consist of a bulk material and asperities
Pada bu a e a a d aspe es
• Bulk material is assumed to be elastic
=
be elastic– Model with effective pattern
density Bulk
• Asperities are assumed to – Observe Hook’s law, i.e., the
+
w(x,y)
, ,force asperity exerts is proportional to its compression
– Have a statistical distribution of it h i ht
Asperities
December 16, 2010 4
asperity heights
Pad-Wafer Interaction:Contact-Based Removal Mechanisms
• Material removal is believed to be due
Pad Asperity
FluidPad/Particle Motion
believed to be due primarily to 3-body contact
• Surface modification by
Wafer
Pad Asperity
AbrasiveParticle
ythe slurry is necessary
• Different nanoscale removal mechanisms have
Wafer Motion
Waferremoval mechanisms have been proposed
– Indentation models– Chemical tooth models– Pressure-driven dissolution
models
December 16, 2010 5
Outline
• Physical measurements of CMP pad properties– Pad modulus– Asperity height distribution
• Pad aging experiment and property testsCu wafer polishing and pad sample collection– Cu wafer polishing and pad sample collection
– Time dependence of pad properties– Spatial variation in pad properties
• Model for pad-wafer contact – Based on mechanical response of pad asperity– Assumptions and mathematical derivationAssumptions and mathematical derivation– Contact area and pressure predictions and trends
• Conclusions and future work
December 16, 2010 6
Pad Modulus Measurement:NanoindentationNanoindentation
• Pad slice nanoindentation:
• Pad asperity nanoindentation:Max Force 10 mNForce Resolution 2 nN
Indent Working Mode
Force Resolution 2 nNMin Contact Force <100 nNForce Load Rate >50 mN/sMax Displacement 5 µm
December 16, 2010
Hysitron TriboIndenter
7
Displacement Resolution 0.04 nm
Pad Sample
• JSR water soluble ti l (WSP) dparticle (WSP) pad
– Soft surface for less scratch
– Hard bulk consist of soft matrix and WSP for betterWSP for better planarization
– Surface porosity controlled by WSPcontrolled by WSP size
http://www.jsr.co.jp/jsr_e/pd/images/pad.pdf
December 16, 2010 8
Pad Slice: Contour Plot of Reduced ModulusReduced Modulus
Test Pattern: slice, multiple points
Example: JSR pad slice, same test area, repeat twice
10µm
1st test
Mean: 462.69MPa
Standard Deviation: 272.44MPa
2nd test (same position on the sample)
Mean: 460.93MPa
Standard Deviation: 270.58MPa
200300
400500
600
400
400
400
500
500 600
600
10
15
2010µm
200
300
300
400
400
400
400500
00
500
500 600
600
10
15
20
300
300
400
400
400
500
500
500
600
600
600
700
700
70040
800800
800
800
900
900900
10001100
Y(m
)
10
-5
0
5
300
300
400
400
0
0
500
500
600
600
600
0700
700
700
800800
800
800
900
900900
10001100
Y(m
)
10
-5
0
5
300300 400
400
400 500
500
500
500
600600
700
700800
900901000
X(m)-20 -15 -10 -5 0 5 10 15 20-20
-15
-10
300 400
400
500
500
500
500
600600
700
700
7
800
900901000
X(m)-20 -15 -10 -5 0 5 10 15 20-20
-15
-10
December 16, 2010
( )
9• There is spatial variation in pad mechanical properties
( )
Pad Asperity: Depth Dependenceof Reduced Modulus
800
1000
MPa
)
JSR Pad 200
300
300
300
400
400
400
400
500
500
600700
500
500
800
600
600
m)
5
10
15
20
400
600d
Mod
ulus
(M 300
300
400
400
400
400
500
500
500
500
500
5
600600
600
600
700
700
700
700
800
800
800
800 900
900900
10001100
1000
X(um)
Y(um
-20 -10 0 10 20-20
-15
-10
-5
0
0
200
Red
uced
Boning and Fan, MRS Spring Meeting April
X(um)
0 100 200 300 4000
Indentation Depth (nm)
• Deep indentation ( > 300 nm):A it d l h b lk d l
Spring Meeting, April 2010.
– Asperity modulus approaches bulk modulus– Bulk estimate = 291 MPa (depth > 300 nm)
• Shallow indentation ( < 100 nm):S b t ti ll hi h d l 2 t th b lk l
December 16, 2010 10
– Substantially higher modulus, ~2x or greater the bulk value – Surface estimate = 572 MPa (depth < 100nm)
Apparent Stiffness of Polymer Surfaces under Contact (Tweedie et al.)
December 16, 2010 11
C. Tweedie et al., Adv. Mat., 19, 2540-2546, 2007.
Depth Dependence ofAsperity Modulusp y
1000
400
600
800
uced
Mod
ulus
(MPa
)
0 100 200 300 4000
200
Indentation Depth (nm)
Red
u
• Surface structure effect or material propertyN d b ifi d b i fl d l
Boning and Fan, MRS Spring Meeting, April 2010. C. Tweedie et al., Adv. Mat., 19, 2540-2546, 2007.
December 16, 2010 12
– Needs to be verified by experiments: e.g., flat pad sample test
Pad Asperity Height Distribution
0 9990.9995
0.9999Pad Surface Profilometry
JSR Pad
0.990.995
0.999
roba
bilit
y
Extracted top 1%λ = 12.7 µm
0 050.250.5
0.750.9
0.95
PExtracted top 5%
λ = 9.9 µm
• Consistent with an exponential height distribution
9 10 11 12 13 14 150.050.25
Height Difference from Mean (m)
– Exponential in the tail of the distribution, i.e., for heights substantially greater than the mean height
– A very small number of very tall asperities (i.e. fewer than 0.02%). We ignore these.
December 16, 2010 13
these.– Possibility of a bimodal (exponential) distribution; useful to extract both.
Outline
• Physical measurements of CMP pad properties– Pad modulus– Asperity height distribution
• Pad aging experiment and property testsCu wafer polishing and pad sample collection– Cu wafer polishing and pad sample collection
– Time dependence of pad properties– Spatial variation in pad properties
• Model for pad-wafer contact – Based on mechanical response of pad asperity– Assumptions and mathematical derivationAssumptions and mathematical derivation– Contact area and pressure predictions and trends
• Conclusions and future work
December 16, 2010 14
Pad Aging Study
• Motivations– Understand how pad properties change during CMPUnderstand how pad properties change during CMP
process– Evaluate pad conditioning effect
• Physical measurements of pad properties– Asperity reduced modulus: nanoindenter
A it h i ht di t ib ti i fil t– Asperity height distribution: micro profilometer– Pad groove depth: microscope and positioning
system on nanoindentery
December 16, 2010 15
Pad Aging Experiment
• Cu wafer polishing with JSR WSP pad– Polisher: Araca APD-800
P li hi h d d 25– Polishing head speed: 25 rpm– Reference pressure: 1.5 psi– Condition head speed: 95 rpm– Conditioner down force: 8 lbF
• Pad sample collection• Aging samples: • Spatial samples after 16 hours:g g p Sp p 6
December 16, 2010 16
Sample size: 2.5cm×2.5cm Sample size: 1.5cm×1.5cm
Pad Properties:Pad Asperity Indentation
• Pad asperity nanoindentation:
Hysitron TriboIndenter Test Pattern Applied
• Indentation curves:F il d t t i d t ti lidi S f l t t lid t tFailed test: indenter tip sliding Successful test: solid contact
December 16, 2010 17
Pad Aging Results
Asperity Modulus Asperity Height Groove Depth
• Asperity modulus and asperity height distribution are both consistent across li hi / diti i tipolishing/conditioning times
• Depth dependence of modulus– Deep indentation: asperity modulus approaches bulk modulus (<200 MPa)– Shallow indentation: substantially higher modulus, ~2x or greater the bulk value
December 16, 2010 18
y g , g• Substantial pad wear during CMP process: groove depth decreases linearly
with polish time
Spatial Results: Asperity Modulus
OA direction: OB direction:
• No clear radial dependence of asperity reduced modulus• Depth dependence of modulus
– Deep indentation: asperity modulus approaches bulk modulus (<200 MPa)
December 16, 2010 19
Deep indentation: asperity modulus approaches bulk modulus ( 200 MPa)– Shallow indentation: substantially higher modulus, ~2x or greater the bulk value
Spatial Results: Asperity Height
OA direction: OB direction:
• No strong radial dependence of asperity height distribution: good spatial uniformity of asperity heights with conditioning
After 16 hours polishing (with conditioning)
December 16, 2010 20
spatial uniformity of asperity heights with conditioning
Spatial Results: Groove Depth
Pad sample
Groove depth
Pad sample
• Groove depth has a strong radial dependence: more pad wear near the center (non optimized pad conditioning in this case)
After 16 hours polishing (with conditioning)
December 16, 2010 21
near the center (non-optimized pad conditioning in this case)
Summary of Pad Aging Results
• Depth dependence of asperity modulus• Pad conditioning keeps asperity properties g p p y p p
consistent during CMP process– Asperity modulus and asperity height distribution are both
consistent across polishing/conditioning timesconsistent across polishing/conditioning times– No strong radial dependence of asperity modulus of asperity
height
• Pad wears linearly with polish time (as measured• Pad wears linearly with polish time (as measured by groove depth)
• Pad wear and thickness does not strongly affectPad wear and thickness does not strongly affect asperity properties: conditioning effective in maintaining pad asperity structure
December 16, 2010 22
Outline
• Physical measurements of CMP pad properties– Pad modulus– Asperity height distribution
• Pad aging experiment and property testsCu wafer polishing and pad sample collection– Cu wafer polishing and pad sample collection
– Time dependence of pad properties– Spatial variation in pad properties
• Model for pad-wafer contact – Based on mechanical response of pad asperity– Assumptions and mathematical derivationAssumptions and mathematical derivation– Contact area and pressure predictions and trends
• Conclusions and future work
December 16, 2010 23
Motivation: CMP ModelsDie-Level Non-UniformityWafer-Level Non-Uniformity
CMP Tool & Process Setup
Chip Layout Design
W f L l
Better CMP Fab-friendly
Wafer-Level CMP Model
Die-Level CMP Model
Processes, Tools, & Consumables
yLayout Design
December 16, 2010
Physics of CMP / Particle-Level CMP Model
24
State of Modeling in CMP
• Wafer-level modeling– Velocity, wafer edge pressure distributions
• Die-level modeling– Pattern density– Step height– Pad bulk vs. asperities: height and diameter distributions
• Particle-level modeling: basic mechanisms– What part of pad participates in polishing? Focusp p p p p g– What is nature of pad/particle/wafer interaction?
• Challenges for the CMP community– Have: effective chip-scale models for a fixed process – useful in chip
Focus
Have: effective chip scale models for a fixed process useful in chip design and optimization
– Need: fundamental mechanisms and models for varying process, pad, slurry, tool, and wafer materials and patterns – useful in process design and optimization and in tool/consumable design and
December 16, 2010 25
process design and optimization, and in tool/consumable design and optimization
Particle-Level CMP Model
Particle-Level
Input variables of CMP
• Applied pressure
R l ti l it
Output variables of CMP
• Blanket removal rate
CMP Model• Relative velocity
• Abrasive size
• etc
• Surface quality
• etc
Microscopic Mechanism
Physics of CMPMacroscopic Phenomenon Microscopic Mechanismp
December 16, 2010
50 nm300 mm ~107 difference in scale26
Overall Physical Modeling Approach• Pad-wafer interaction• Pad-abrasive interaction
Ab i f i t ti• Abrasive-wafer interactionPad
Dielectric
11mmPPad
PAbrasive
December 16, 2010~50nm
27
Pad-Wafer Interaction• Assume “fully supported” asperities• Asperity compression and asperity contactp y p p y• Can predict local asperity contact pressure
December 16, 2010 28
Have Physical Model… But Many Assumptions!
• Many of these assumptions have not been experimentally nailed down:– What part of pad participates in polish? Asperity
height and size distributions? Mechanical properties of asperities?
Focus
– What slurry particles participate? Large particles only? What particle/asperity loading occurs?
– Effects of slurry chemistry temperature– Effects of slurry chemistry, temperature, …• Alternative physical assumptions and models are
possible and have been proposed• If we can find the correct physics, then model can be
used to predict results for different pads, slurries (e.g. particle sizes), pressures, velocities, etc., …
December 16, 2010 29
p ), p , , ,
Summary ofPad-Wafer Interaction Model
• Greenwood Williamson approach– Asperities have spherical surfaces with same radius– Elastic Hertzian contact– Elastic Hertzian contact
• Single asperity compression
• Exponential asperity height distributionResult: Predict contact area fraction f• Result: Predict contact area fraction f
E: asperity reduced modulusP0: reference pressure
December 16, 2010 30
P0: reference pressure
Boning and Fan, MRS Spring Meeting, April 2010.
Model Trend: Contact Area vs. Reference Pressure
0 06
0.08
ShallowDeep
E = 291 MPa
0.04
0.06
f (%
)E = 572 MPa
0 2 4 6 8 100
0.02
• Contact area increases linearly with P0
0 2 4 6 8 10P0 (psi)
– Depends on reduced pad modulus– Using shallow modulus average (stiffer asperities):
smaller f% for same pressure– Using deep modulus average (asperities same as bulk):
December 16, 2010 31
– Using deep modulus average (asperities same as bulk): predicts larger f% for same pressure
Model Trend: Contact Area vs. Characteristic Asperity Heightp y g
0.026
0.028
P0 = 4 psi
0 02
0.022
0.024(%
)
P0 4 psi
0.016
0.018
0.02f
5 6 7 8 9 10 11 12 13 14 150.014
(m)
• Contact area decreases with larger λ– Larger λ implies wider distribution (more taller asperities)– For wider distribution, a smaller number of tall asperities bear the load,
December 16, 2010 32
For wider distribution, a smaller number of tall asperities bear the load, reducing the contact area percent
Consistent with Conditioning/Contact Area Data
C diti A i d C t t A
4 PSI A: Less aggressiveB: More aggressive (3X cut rate of A)
Conditioner Aggressiveness and Contact Area
gg ( )
December 16, 2010
L. Borucki et al., CSITC, March 2010.
33
Example Polishing Pad Topography5050 m
L. Borucki et al., CSITC, March 2010
December 16, 2010
March 2010.
34
Simulated Pad-Wafer InteractionP0 = 5 psi
λ = 11.8 µm
E = 460 MPa
December 16, 2010 35
Simulated Pad-Wafer InteractionP0 = 10 psi
λ = 11.8 µm
E = 460 MPa
December 16, 2010 36
Simulated Pad-Wafer InteractionP0 = 50 psi
λ = 11.8 µm
E = 460 MPa
December 16, 2010 37
Simulated Pad-Wafer InteractionP0 = 150 psi
• Contact area changes with• Contact area changes with overall applied pressure
• There is also a distribution of asperity contact pressures: has
λ = 11.8 µm
E = 460 MPa
December 16, 2010 38
asperity contact pressures: has implications for modeling
Model Parameters andMeasured Results
• Model parameters– E and λ are single fixed values
• Physical measurements– Depth dependence of E– Depth dependence of E– Height range of extracted λ
• How to utilize the measure results in the model– Modify the model to include nonlinear effect?– Take the mean value of measured result in a certain range?
December 16, 2010 39
Take the mean value of measured result in a certain range?
Conclusions
• Measurement approaches developed:– Pad slice indentation test– Asperity reduced modulus– Asperity height distribution
Meas rement obser ations• Measurement observations:– See strong depth dependence of pad asperity
modulus– Pad aging evaluation: surface properties remain
consistent with conditioning
• Model for pad-wafer contact• Model for pad-wafer contact – Based on mechanical response of pad asperities– Contact area predictions and trends
December 16, 2010 40
Future Work• Compare predicted and measured contact
fractions
• Understand distribution of asperity size, heights, and mechanical propertiesand mechanical properties
• Consider implications of shallow indentation modulus 1000 0.08
Shallow
400
600
800
duce
d M
odul
us (M
Pa)
0 02
0.04
0.06f (
%)
ShallowDeep
December 16, 2010 41
0 100 200 300 4000
200
Indentation Depth (nm)
Red
0 2 4 6 8 100
0.02
P0 (psi)
Acknowledgements
• SRC/SEMATECH Engineering Research Center for Environmentally Benign Semiconductor y gManufacturing
• Y. Zhuang, Y. Sampurno, and A. Philipossian at D f Ch i l d E i lDepartment of Chemical and Environmental Engineering, Univ. of Arizona; Araca Inc.
• D Hooper and M Moinpour at Intel Corp• D. Hooper and M. Moinpour at Intel Corp.
December 16, 2010 42
top related