Fiona M. Doyle and Shantanu Tripathi*

1
TRIBO-CHEMICAL MECHANISMS AND MODELING IN COPPER
CMP

• Fiona M. Doyle and Shantanu Tripathi
• University of California at Berkeley
• Department of Materials Science and Engineering
• 210 Hearst Mining Building 1760
• Berkeley, CA 94720-1760
• fmdoyle_at_berkeley.edu
• Department of Mechanical Engineering

2
FLCC CMP Approach

• Our approach is to develop integrated
  feature-level process models linked to basic
  process mechanics
• These models will drive process optimization and
  the development of novel consumables to minimize
  feature-level defects and pattern sensitivity
• Current effort aims to integrate mechanical and
  chemical phenomena
• Need to capture synergism between the two

3
CMP Overview
Slurry feeder
Pressure
ALUMINA PARTICLES average size 120 nm from
EKC Tech.

• SLURRY
• Abrasive particles
• Chemicals

Carrier
Wafer
Rotation
Patterned wafer
Polishing Plate
POLISHING PAD
Cross-sectional View of SUBA 500 Pad, Rodel Corp.
(courtesy Y. Moon)
Polishing pad
Pad asperities
4
Passive films, or corrosion inhibitors, are key
to attaining planarization
Kaufmans Model for Planarization For effective
planarization, must maintain higher removal at
protruding regions and lower removal at recessed
regions on the wafer
1- removal of passivating film by mechanical
action at protruding areas
2- wet etch of unprotected metal by chemical
action. passivating film reforms
Metal
Passivating film
3- planarization by repetitive cycles of (1) and
(2)
5
Chemical Mechanical Planarization
Mechanical Phenomena
Chemical Phenomena
Interfacial and Colloid Phenomena
6
Chemistry interacts synergistically with
mechanical/colloidal phenomena
Chemistry affects degree of aggregation of
abrasive particles.
Mechanical forces on copper introduce defects,
increasing reactivity
Copper nanoparticles have dramatic effect
Mechanical properties of films appear to be
strongly dependent on chemistry, and probably
potential
7
The Problem
Needed an Integrated Copper CMP Model

• Interactions
• Asperity-copper
• Abrasive-copper

Pad Pressure/ Velocity Abrasive
Oxidizer Inhibitor Complexing agent Surface Film
Integrated Cu CMP Model
Colloid Agglomeration
The Problem
Fluid Mechanics Mass Transfer
Needed understanding of the synergy between
different components
Needed an Integrated Copper CMP Model

• Interactions
• Asperity-copper
• Abrasive-copper

Fluid pressure Contact pressure
Pad Pressure/ Velocity Abrasive
Oxidizer Inhibitor Complexing agent Surface Film
Integrated Cu CMP Model
Colloid Agglomeration
Fluid Mechanics Mass Transfer
Needed understanding of the synergy between
different components
Fluid pressure Contact pressure
8
Tribo-Chemical Model of Copper CMP

• Synergism between frequent mechanical
  interactions and action of chemical slurry make
  copper CMP process electrochemically TRANSIENT
  but to date
• NO study of transient behavior, focus on steady
  state.
• NO mechanistic models of tribo-chemical
  synergism.

• We must study
• Transient passivation behavior of copper first
  few moments of copper passivation.
• Abrasive-copper interactions frequency, duration
  and force.
• Properties of passive film mechanical,
  electrical, chemical

9
Average removal rate between abrasive-copper
contacts
10
Transient Passivation Behavior

• No direct study on copper CMP slurry
  constituents.
• Observed behavior for other metal-chemical
  combinations log-log (oxidation rate time)
  Jones DA Principles and prevention of
  corrosion Prentice Hall 2nd edition, 1995
• Complex behavior observed for Cu-AHT (inhibitor)
  behavior Beier M, Schultze JW, Electrochimica
  Acta 37 (12) 2299-2307 1992
• Wide variation observed in decay kinetics for
  different systems milliseconds to minutes.

Beier Schultze
11
Parabolic Rate Law for Corrosion Kinetics?
oxidant in slurry (fixed)
Cu
Passive film
CMP Slurry containing oxidant
oxidant in copper (fixed)
Film thickness x(t)
Flux of oxidant
12
Mechanical Interactions
Duration between contact events.

• Passive film thickness ? corresponding oxidation
  rate
• Duration/Force of contact ? Thickness of Passive
  film removed

13

• Interval between asperity-copper contact 1ms
• Duration of contact 10µs
• Needed study of abrasive-copper interactions

Interaction Frequency Duration
C-RICM image of real contact area
Elmufdi Muldowney, Mater. Res. Soc. Symp. Proc.
Vol. 91, 2006 Spring
14
Tribological Properties of Passive Films

• Wear of passive film depends on mechanical
  properties of passive film and abrasive particle,
  and force of contact.
• Mechanical properties of passive film affected by
  chemical conditions (inhibitor, oxidation
  potential)

Passive film properties varying with slurry
chemistry
Linear wear till passive film removed
Bi-layer passive film
Loading of abrasive
15
Quartz Crystal Microbalance

• Sauerbrey equation
• where ?q is the shear modulus of the quartz
  crystal, ?q the density, and f0 the resonant
  frequency
• for an AT-cut quartz crystal with a resonant
  frequency of 5 MHz gives that ?m/?f is 1.77 x
  10-8 g/cm2Hz

• The changes in frequency of a piezoelectric
  quartz crystal, ?f, are related to changes in
  mass, ?m, of a substrate (e.g. Cu) that is
  attached to the quartz crystal

16
EQCM Experimental apparatus and materials

1. Maxtek Research Quartz Crystal Microbalance
2. Maxtek 1-inch diameter quartz crystals and the
   electrode configuration
3. Maxtek crystal holder
4. Schematic diagram of experimental setup for EQCM
   measurements. (left) chemical reagents introduced
   against the wall of cell, (right) a tube 10 mm
   from the crystal) for injecting chemicals

17
pH 4, OCP, 0.01 M glycine premixed in acetate
buffer
Temporary loss in weight, followed by significant
gain in weight, more pronounced at higher
concentration of H2O2.
18
pH 9, OCP, 0.01 M glycine added to carbonate
buffer after stabilization
Slow loss in weight upon adding glycine.
Temporary sharp loss in weight after adding
peroxide, followed by significant gain in weight.
19
Effect of adding additional glycine, after adding
2.09 hydrogen peroxide
Deionized water
pH 9
20
Open circuit potential of copper, pH 9, 0.01 M
glycine and 2.09 hydrogen peroxide
No H2O2. Potential same as that induced by H2O2
No passivation without H2O2. See that behavior
is strongly dependent on history of glycine
additions oxidized layers must resist dissolution
21
Effect of glycine and H2O2 additions at different
potentials, pH 9, 0.01 M glycine
S. Zecevic, D.M. Dražic, S. Gojkivic J.
Electroanal. Chem, 265 (1989) 179
Iron disk-Au ring electrode. H2O2 produced
during reduction of O2 is rapidly reduced at high
and low potentials, but can escape electrode at
intermediate potentials
At controlled potentials, either oxidizing or
reducing, H2O2 does NOT lead to weight increase.
Protective film must be sensitive to potential
However, this is not consistent with passivation
at high concentrations of H2O2
22
Environmental AFM

• In-situ flow through experiment (flow rate
  0.675ml/min)
• Slurry constituents
• DI water (introduced at time t 0min)
• Glycine in pH 4 acetic acid/acetate buffer
• (at time t 22 min)
• Glycine Hydrogen Peroxide in pH 4 acetic
  acid/acetate buffer (at time t 56 min)

AFM scanner
3
2
1
out
in
Cu sample in a flow through cell
6 port valve
Peristaltic pump
23
AFM in Air of Copper Pre-exposed to Different
Slurry Components Ex-situ
z range 320.1nm
z range 4.93nm
z range 47.6nm
z range 0.74nm
Topography
Deflection
xy1.94µm
Topography
Deflection
xy1.13µm
Copper in Air
Copper pre-exposed to 2 H2O2 and 0.01M glycine
_at_ pH4 for about 1 hour
z range 31.3nm
z range 0.59nm
Glycine at pH 4 (albeit short exposure) does not
affect surface morphology significantly With
peroxide, original surface morphology is changed
dramatically Although there is some ambiguity,
peroxide is much more likely to be adding a
surface film rather than etching, which would
affect grain boundaries preferentially
Topography
Deflection
xy1.13µm
Copper pre-exposed to 0.01M glycine _at_ pH4 for 1
minute
24
Corrosion of Copper in 0.01M glycine, pH 4
In-situ imaging Buffered glycine solution
introduced at t22 min. See slight etching,
correlates with very slightly negative gradient
in EQCM work before peroxide addition
t29min
t32min
t35min
xy1.13µm, Deflection images
t47min
t44min
t41min
25
Copper in 2 H2O2, 0.01M Glycine at pH 4

• Effect of changing the flow through constituent
• Instant drift and noise in AFM imaging, then
  stabilization.
• Transient noise prevents capturing any transient
  material removal upon adding peroxide

z range 98.4nm
z range 1.2nm
t68min

• Contact mode imaging gives very noisy AFM images
• Consistent with presence of very porous and
  mechanically weak film on copper
• Possible deterioration of AFM probe tips in this
  chemistry

Deflection
xy1.13µm
Topography
Flow through imaging, H2O2 solution introduced at
t56min, after solution 1 2
z range 65nm
z range 0.66nm
Consistent with plateau in weight gain after
adding peroxide, with passivation seen in
glycine/peroxide chemistries, and with signficant
acceleration of material removal
Topography
Deflection
xy2.09µm
Imaging in standing solution, no pre-exposure to
solutions 1 2
26
Future AFM Work

• Use of AFM tip to damage existing passive films
• Observe effect of chemistry on mechanical
  properties of films
• Observe transient currents, and correlate with
  area of damaged surface to obtain current
  densities as a function of time
• Study passive film formation kinetics, to
  identify best model for transient behavior

27
Testing of Model

• Earlier electrochemical studies (under SFR, by
  Serdar Aksu) will be used to test model
  predictions
• Synergy between mechanical and chemical factors
  of particular interest
• EQCM work done under FLCC by Ling Wang will
  provide reference for short time frames

28
Polarization Curves in Cu-Glycine-H2O
CuT 10-5, LT 10-2
LT 10-2
29
In-situ Polarization
pH 4
pH 12
pH 9
Aqueous 10-2 M glycine, 27.6 kPa, 200 rpm
30
Conclusions

• Earlier mechanistic studies of copper CMP are
  providing insight for coupling of chemical and
  mechanical models
• Mechanistic approach is designed to capture the
  synergy between the two
• Work on colloidal properties of abrasives will
  also be invoked
• FLCC CMP team well positioned to capture relevant
  developments in other fields
• In addition to the intrinsic utility of a
  combined chemical/mechanical model for CMP, this
  should resolve remaining questions on material
  removal mechanisms
• This in turn will allow more efficient
  developments in future