Upload
hoangduong
View
217
Download
2
Embed Size (px)
Citation preview
AS Lawing NCCAVS CMPUG 5/5/2004
Pad Conditioning Effects in Chemical Mechanical Polishing
Pad Conditioning Effects in Chemical Mechanical Polishing
A. Scott LawingRohm and Haas Electronic Materials CMP Technologies
Phoenix, AZ
AS Lawing NCCAVS CMPUG 5/5/2004
IntroductionIntroduction
BackgroundPad Conditioning and Pad Surface Structure
Competing effects of wear and restoration Conditioner design and cutting behaviorVoid-filled vs. solid padsIntrinsic vs. induced structureContact areaProcess effects
Pad Conditioning and Polish RateEx situ responseIn situ responseRate maximization behaviorHydrodynamic effects
Pad Conditioning and PlanarizationPlanarization efficiencyDishing
Conclusions
AS Lawing NCCAVS CMPUG 5/5/2004
BackgroundBackground
Polishing PadWafer
ConditionerSlurry
Pad
AsperityPad Pore
Diamond Crystal
Pad conditioning is the process of “dressing” the polishing padPad is contacted with an abrasive medium, typically a diamond abrasive discThe conditioning process involves the removal of a thin layer of pad material
Conditioning determines the intrinsic asperity structure of the padConditioning acts to maintain surface stability through the removal of worn surface material and restoration of the intrinsic structure
AS Lawing NCCAVS CMPUG 5/5/2004
Extremes of Pad SurfacesExtremes of Pad Surfaces
Conditioning Dominated
Wafer Dominated
Abrasive wear results in truncation of asperity tips
Pad conditioning restores asperities to their intrinsic structure
-40 -20 0 20Pad Height (um)
Frequency (a.u.)
-40 -20 0 20Pad Height (um)
Frequency (a.u.)
Conditioning dominated surface exhibits the intrinsic structure imparted by a medium aggressive conditionerWafer dominated surfaces exhibits the truncated asperity structure (red component in the distribution at lower center) induced by polishing many wafers in the absence of conditioning
Intrinsic contribution to distribution in blue at lower center
AS Lawing NCCAVS CMPUG 5/5/2004
Competing EffectsCompeting Effects
Pad Wear Rate
Con
ditio
ner C
ut R
ate
Over-conditioning
Severe GlazingIntrin
sic structure restored above horizontalMore stable
Less stable
Cut Rate = Wear Rate
The pad surface structure is determined by the balance between the competing effects of pad wear due to pad-wafer contact and pad surface restoration due to conditioning
The pad wear rate is a function of process conditions and the consumable setThe conditioner cut rate is a function of the conditioner design and the process conditions
· The intrinsic structure of the pad surface is also dependent on the conditioner cutting characteristics
AS Lawing NCCAVS CMPUG 5/5/2004
Competing EffectsCompeting Effects
In this example, conditioner cut rate is varied by manipulating the intrinsic aggressiveness of the pad conditioner through changes in diamond configuration
Pad wear rate is held constant by using an identical process condition and consumable set
The effect of increasing cut rate is to reduce the amount of steady-state glazing on the pad surface
-40 -30 -20 -10 0 10 20 30
-30 -20 -10 0 10 20
Pad Wear Rate
Con
ditio
ner C
ut R
ate
Increasing Conditioner Aggressiveness
AS Lawing NCCAVS CMPUG 5/5/2004
Competing EffectsCompeting Effects
In this example, pad wear rate is varied by manipulating wafer down-force
The point at far left represents a pad surface after break-in (i.e. zero pad wear)
The effect of increasing wear rate is to increase the amount of steady state glazing on the pad surface
Pad Wear Rate
Con
ditio
ner C
ut R
ate
Increasing Wafer
Down-force
AS Lawing NCCAVS CMPUG 5/5/2004
Conditioner DesignConditioner Design
Pad surface structure is the product of many individual pad-diamond crystal interactions
Structure is driven by the cutting characteristics of diamond crystalsIntrinsic pad surface structure can be manipulated through conditioner design
Size, shape and density of diamonds on conditioner surface drive cutting characteristics and intrinsic surface structureMore aggressive diamonds tend to remove more pad per unit interaction and result in more “rough” surface
Diamond Crystal TypesSolid Pad Surface
DiamondFurrows
AS Lawing NCCAVS CMPUG 5/5/2004
Conditioning and Intrinsic StructureConditioning and Intrinsic Structure
-50 -30 -10 10 30
Height (µm)
Freq
uenc
y (a
rb. u
nits
) HighMediumLow
50 -30 -10 10 30
Height (µm)
-50-40-30-20-10
010203040
Hei
ght (
µm)
-50-40-30-20-10
010203040
1 1.2 1.4 1.6 1.8 2Lateral Dimension (mm)
Hei
ght (
µm)
Aggressiveness Void-filled SolidVoid-filled
Solid
Pad surface height statistics are significantly affected by conditioning aggressiveness
Aggressiveness adjusted through changes in conditioning designIn solid pad, aggressiveness affects roughness
Solid pad has no inherent textureIntrinsic solid pad surface height distribution nearly Gaussian
In void-filled pad aggressiveness affects near surface regionRoughness of void-filled pad determined by filler materialConditioning superimposes additional roughness frequency
AS Lawing NCCAVS CMPUG 5/5/2004
Intrinsic Structure vs. Pad WearIntrinsic Structure vs. Pad Wear
0102030405060708090
100 c) High
0102030405060708090
100 b) Medium
0102030405060708090
100 a) Low
D/h
(Dim
ensi
onle
ss)
Pad Height PDF Asperity Aspect Ratio (D/h)
-30 -20 -10 0 10
Freq
uenc
y (a
.u.)
Intrinsic ComponentWear ComponentOverall FitDepth of ContactData
-40 -30 -20 -10 0 10 20 30-30 -20 -10 0 10 20
a) Low b) Medium c) High
In steady state polishing, surface structure is determined by balance between restorative effect of conditioning and destructive effect of wear due to pad-wafer contact
Conditioner acts to restore intrinsic structure by removing worn surface materialLess aggressive conditioner characterized by lower cut-rate such that more wear is evident on the steady-state surface as aggressiveness is reduced
Asperity structure becomes more truncated (lower Diameter/heightratio) as aggressiveness is decreasedEstimated contact area increases as aggressiveness is decreased
Contact area estimated at 11.3 %, 7.7 % and 2.2 % for the low, medium and high aggressive conditioner respectively
AS Lawing NCCAVS CMPUG 5/5/2004
Process EffectsProcess Effects
Increased conditioning down-force results in the removal of more pad material shifting the balance between wear and restoration
Below a critical down-force, polish rate will drop with decreasing down-forceAbove critical down-force rate saturates with increasing conditioner down-force
Intrinsic structure does not change significantly with increasing conditioner down-force
Once intrinsic structure is restored (wear component is accounted for) additional conditioning does not change the structure of the pad surface
0 5 10 15 20Conditioning Load (lb-hours)
Rou
ghne
ss
R = 2"R = 5"R = 8"
30 min @ 5 lb
+30 min @ 8 lb
+30 min @ 11 lb
+30 min @ 14 lb
"Broken-In"
20002200240026002800300032003400360038004000
0 5 10 15
Conditioning Down Force (lbs)
Pol
ish
Rat
e (Å
/min
)
Fumed, HighFumed, HighColloidal, MediumColloidal, Medium
Solid Pad Texture with Break-inConditioning DF Effect on Rate
AS Lawing NCCAVS CMPUG 5/5/2004
Ex SituEx Situ Rate DecayRate Decay
1400
1600
1800
2000
2200
2400
2600
2800
3000
3200
0 5 10 15 20 25 30Time Without Conditioning(min)
Pol
ish
Rat
e (Å
/min
)
Medium Aggressive
High AggressiveFumed
Colloidal
Pad Height (tick = 10 mm)
Freq
uenc
y (a
rb.u
nits
)
Pad Height (tick = 10 µm)
Fumed - Medium Colloidal - Medium
Break-In 0 min3 min
15 min
31 min
Break-In0 min 31 min
Rate Data Pad Surface Data
Typical “logarithmic” decay of rate after conditioning is suspendedIncrease in asperity wear with time corresponds to decrease in polish rateColloidal slurries induce less significant asperity wear and result in less significant rate decayConditioning sets initial rate but has little influence after conditioning is suspended
AS Lawing NCCAVS CMPUG 5/5/2004
Polish Rate, Slurry and Conditioning Polish Rate, Slurry and Conditioning
Steady-State Polishing with 100% In Situ Conditioning
0500
1000150020002500300035004000
0 5 10 15Wafer Pressure (psi)
Polis
h R
ate
(Å/m
in)
Modified Preston fitHigh fumedMedium fumedLow fumedMedium colloidalLow colloidal
0
500
1000
1500
2000
2500
3000
3500
0 20 40 60Linear Velocity (m/min)
Polis
h R
ate
(Å/m
in)
V = 30 m/min DF = 6 psi
Less aggressive conditioners exhibit more significant deviation from maximum polish rateColloidal slurries exhibit less significant deviation from maximum rate compared to fumed silica slurriesMore significant deviation from maximum rate with increased velocity as opposed to increased down-force
Low aggressive-fumed silica data set exhibits rate maximization with respect tovelocity
AS Lawing NCCAVS CMPUG 5/5/2004
Polish Rate, Slurry and Conditioning Polish Rate, Slurry and Conditioning Steady State Pad Surfaces after In Situ Conditioning with Fumed Slurry
Z (tick mark = 10 µm)
Freq
uenc
y (a
.u.) Break-in
2 psi 10 psi 14 psi6 psi
Pad Height (tick = 10 µm)
Break-in 2 psi6 psi
10 psi14 psi
Z (tick mark = 10 µm)
Break-in
15 rpm 30 rpm 45 rpm 60 rpm
High Aggressive Low Aggressive
High aggressive conditioner exhibits the least surface wear and the maximum polish rateLow aggressive conditioner exhibits the most surface wear and the most significant polish rate deviationMore significant wear as a result of increase in down-force compared to increase in velocity
Increasing down-force driving abrasive wearThis contradicts trend of more significant rate reduction with increase in velocity compared to increase in rate
Colloidal slurries induce less significant asperity wear (not shown) corresponding to less significant deviation from maximum rate
AS Lawing NCCAVS CMPUG 5/5/2004
Polish Rate and Hydrodynamics Polish Rate and Hydrodynamics
02468
10121416
1.E-05 1.E-04 1.E-03Sommerfeld Number So
Red
uced
Pol
ish
Rat
e(Å
/min
)/(ps
i*m/m
in)
02468
10121416
0.000 0.002 0.004 0.006Frictional Loading Factor µV/Pc (µm)
Red
uced
Pol
ish
Rat
e(Å
/min
)/(ps
i*m/m
in)
High fumed Series6Medium fumed Medium ColloidalLow fumed Low Colloidal σ
µ
co P
VS =Boundary
Lubrication PartialLubrication
Influence of partial lubrication with increasing Sommerfeld number explains the more significant rate decrease with increasing relative velocity
Hydrodynamic effects become more significant at low contact pressure (high contact area) and high velocity
Aggressively conditioned surfaces are dominated by solid contact
AS Lawing NCCAVS CMPUG 5/5/2004
Estimation of PadEstimation of Pad--Wafer Contact AreaWafer Contact Area
-30 -20 -10 0 10Pad Height (µm)
Pad
Hei
ght F
requ
ency
1.E+00
1.E+01
1.E+02
1.E+03
1.E+04
1.E+05
1.E+06
1.E+07
1.E+08
Vol
ume
(µm
3 )
Pad Height in Contact
Pad Height not in Contact
Cumulative Volume
-30-25-20-15-10
-505
1015
0 0.5 1 1.5Scan Length (mm)
Pad
Hei
ght (
µm)
Using simple elastic model, equate volumetric displacement of ideal bulk material with the pad height probability distribution function (pdf) at a given applied pressure
Effective spring constant used was obtained by fitting patterned wafer polish data to elastic pad model (Lawing & Merchant ECS 2000)Similar length scale for pad compression
“Asperities” are defined as the portion of the pad lying above the contact plane (dotted line in the figures above)Estimate of contact coincides with maximum in the component of the distribution due to abrasive wear
AS Lawing NCCAVS CMPUG 5/5/2004
Polish Rate and Contact AreaPolish Rate and Contact Area
1500
1700
1900
2100
2300
2500
2700
2900
3100
0 10 20 30
Time w/o conditioning (minutes)
Rat
e (Ǻ
/min
ute)
Colloidal in situ #1
Colloidal in situ #2
Fumed in situ
Colloidal ex situ
Polish rate is maximized with respect to estimated contact areaIncreasing ex situ rate trends support this observation
· Wear model demands that surface area increases as a function of time without conditioning
Above a critical point rate is proportional to contact area (applied pressure)Working model is that below a critical contact area rate is limited by absolute contact area and not contact pressure
1000
1500
2000
2500
3000
3500
Model Fractional Contact Area (Relative)
Pol
ish
Rat
e (Å
/min
)
Fumed #190% Colloidal/10% FumedColloidalFumed #2Asst. Conditioners in situ
Rate as a Function Estimated Contact Area Rate Trends with Various Initial Conditions
Increasing AggressivenessIncreasing Contact Area
AS Lawing NCCAVS CMPUG 5/5/2004
-0.10
0.10.20.30.40.50.60.70.80.9
11.1
0 2000 4000 6000 8000 10000 12000
Amount Removed from Hi Area (Å)
Pla
nariz
atio
n E
ffici
ency
Low AggressiveHigh Aggressive
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
0 2000 4000 6000 8000 10000 12000Amount Removed from High Area (Å)
Ste
p H
eigh
t Rem
oved
(Å)
Low AggressiveHigh Aggressive
Conditioning and Planarization Conditioning and Planarization
Low aggressive conditioning results in more efficient planarization
More truncated asperity structure results in preferential polishing on high areas and more efficient step removal
RRhi
RRlo
Plo < Phi
Planarization Efficiency of 10% Density Features
( )Hi
LoHi
RRRRRRPE −
=
Step Height of 10% Density Features
Dotted lines indicate ideal planarization
AS Lawing NCCAVS CMPUG 5/5/2004
Conditioning and DishingConditioning and Dishing
-0.5
0
0.5
1
0 0.3 0.6 0.9 1.2 1.5
Fractional Polish Time
Nor
mal
ized
Ste
p H
eigh
t
0 0.3 0.6 0.9 1.2 1.5
Fractional Polish Time
90%
10%
50% 5 µm
50% 50 µm
50% 500 µm
Over-polish
Dishing
Medium/High Aggressive Low Aggressive
Low aggressive conditioning results in less dishingMore truncated asperity structure penetrates less deeply into low lying areas
AS Lawing NCCAVS CMPUG 5/5/2004
ConclusionsConclusions
Pad conditioning defines the intrinsic structure of the pad surface
Intrinsic pad structure can be adjusted through changes in conditioner design
Process and consumable variables adjust the balance between pad surface restoration due to conditioning and asperity wear due to pad-wafer contact
Conditioner design, conditioner and wafer down-force, relative linear velocity, conditioning time and slurry type are all significant factors in defining equilibrium pad surface structure
Pad surface structure and conditioning have a significant effect on CMP process response
Polish rate, planarization performance and hydrodynamic effects are heavily influenced by pad conditioning
AS Lawing NCCAVS CMPUG 5/5/2004