Dec 28, 2015
Overview
Need for CMP Process
Detailed process steps Production: Relevant information
Dishing and Erosion Integration issues
Post-CMP Clean Review
Need for CMP
Where is CMP used? To planarize oxide, tungsten or copper
Why planarize? Oxide: need planar surface for photolithography Copper: no other way to remove excess copper and form the wires (shape) W: relatively easy method
Details: (next page)
Need for CMP
Oxide: (or any other insulator/ dielectric)
Consider the following example: Process steps for ‘tapping’ the ‘gate’ and ‘source/ drains’
(Source)P+ (Drain)P+
Gate(P+)
Need for CMP
1. Deposit oxide (insulator/dielectric)
(Source)P+ (Drain)P+
Gate(P+)
2. Coat photo resist and .... bring the mask to focus Focus on which plane?
Top Plane
Bottom plane
Need for CMP
Solution: Deposit oxide and then PLANARIZE it
(Source)P+ (Drain)P+
Gate(P+)
Planarization removes or decreases ‘topography’
It also removes some material from the ‘lower’ regions
Sample Image
TableQuill (Wafer Carrier)
Animation
Schematic
Background
Used to polish ‘bare wafer’ IBM pioneered Uses particles. Class 100 or Class 1000 clean room
Animation
Pad soft pad, hard pad, stacked pad hard pad with perforations, grooves
Quill surface tension, vacuum
Slurry abrasive, chemicals
Copper
Pad
Removal in these regions
No Removal in this region
• ==> non planar surface --> planar surface
Removal Mechanism
CMP: Pad
Pad is softer than wafer whether it is Cu or oxide or W
Pure chemical dissolution is rare A complexing or oxidizing film usually forms
Pure mechanical removal is possible, but not preferred Scratches
==> shorts Pad has asperities
helps in holding particles
CMP: Pad
Common pads from Rodel Inc soft pads: SUBA IV hard pads: IC1000 stacked pad: IC1400, IC1000/SUBA IV
Slurry Transportation soft pad doesn’t have grooves has ‘hair like’ surface hard pads have grooves/perforations
CMP: Pad
Grooves add mechanical flexibility Soft pad, tend to planarize less Tend to polish more Scratch less
Pad conditioning loss of asperities, roughness use diamond particle conditioner between polishes initial conditioning with few blank wafers
Sample conditioner surface©Rodel Inc
CMP: Consumable
Pad can be used for 150 polishes (for example) Polish rate and within wafer non uniformity will go worse (WIWNU)
PETEOS- Plasma Enhanced TEOS based oxide
CMP: Consumable
Mostly patented slurries (expensive) Oxide polish with KOH or Ammonia slurry
silica (and sometimes ceria) abrasives ceria has very high polish rate (considering its hardness)
Copper polish with acid based or neutral slurries Peroxide, glycine, BTA (benzo triazole) inhibitor silica or alumina abrasives
W polish with Ferric nitrate, peroxide, iodate slurries alumina abrasives
In general, 1 to 3 min polish, 100 to 300 ml slurry used
CMP: Consumable
Silica particles are small Removal rate is low lower scratches fumed or colloidal mono dispersed or poly dispersed surfactants (general)
Alumina harder, more removal, more likelyhood of scratches
ceria soft, some scratch, more removal (for silica polish)
Usually, removal rates in few hundred nm/min
CMP: Quill (Head) control
Pressure application is expected to be uniform on the wafer To ensure it, ‘back pressure’ is applied on the wafer
vacuum used mainly in ‘picking up’ the wafer from the load station and after polish
CMP: Removal Rate
Both pad and quill (wafer holder) rotating at the same rate Quill off-center Relative velocity at all the locations will be same
Practically quill will be ‘swept’ to the front and back (i.e. Edge and center of the pad) quill usually rotates at a slightly different rpm
based on experimental optimization Done to ensure uniformity
and to utilize maximum pad area
CMP: Removal Rate
First order approximation Preston Equation
pRR k PV
Correction for chemical dissolution
Kp Preston Constant, depends on slurry,pad, temp, material to be removed etc
For the same system, different materials will be removed at different rates
0 1 2RR K K V K PV
CMP: Removal Rate
Proposed Mechanisms
Fluid layer holds the pressure Model not able to explain observation good to explain within wafer non uniformity
Particles hold the wafer Increase in pressure increases number of particles in contact
and not the friction coefficient and not the indentation depth
Removal rate == # particles, indentation depth, shear length
Dishing Erosion Losses
Soft pad ==> more dishing. Erosion Dishing/Erosion ==> higher resistance, more variation in resistance Dishing/Erosion ==> DOF in the next level!
Dishing Erosion Losses
eg. Copper polishLiner (barrier), oxide and copper ‘bulk’ removal sees only copper liner and oxide ‘touch up’ CMP needed to remove liner usually 2 step CMP
Selectivity too high a selectivity leads to drastic dishing/erosion controllable selectivity is the key sometimes 1:1:1 slurry used
CMP: Dummy, Slotting
Copper CMP: Very wide line ==> very high dishing Resistance increase too much Due to erosion, some lines may disappear!
Simple schematics below
CMP: Dummy, Slotting
Aerial Density important Max density and min density in a ‘window’ is specified
M1 layout
Window size depends on characteristic planarization length
function of material, slurry, pad can be mm for oxide, 100 um for Cu, W
CMP: Dummy, Slotting
OxideCu Cu
OxideCu CuCu Cu
Effective density of copper = 40% ( Oxide density = 60%)
Effective density of copper = 20% ( Oxide density = 80%)
CMP: Dummy, Slotting
Some areas may not have any metal Will cause non uniformity
“stretch” the process too much Use “dummy fill” to bring the density to minimum level
Some lines will be too wide slot them , for better manufacturability
These are examples are “design for manufacturability” or DFM
Both dummy and slotting are used in copper
Characterization with MIT / Sematech mask Set
CMP: Copper
Single step vs Multi step (cost of equipment, slurry) Barrier CMP Ta harder, less reactive Traces of barrier ==> short More mechanical removal ==> scratches ==> short More chemical removal ==> Cu loss Need to strike a balance
CMP: Oxide
Blank Oxide vs STI LOCOS vs STI LOCOS: semi recessed/ fully recessed Issues:
Dummy Fill / Reverse Active Mask Dishing Notch Stress Nitride Polish
CMP: W
Well established Liner removal is easier (Ti/TiN)
lingering liner can be removed (for example) with plasma etch / RIE for a short time (touch up RIE)
Need to have seemless dep (to avoid corrosion issues) Good selectivity is achievable W plug is never large
==> dishing not an issue Erosion still an issue
not for resistance however for DOF
CMP: Clean
Particles to be removed
Brush Scrub Boundary Layer
Mega Sonic/ ultra sonic
Copper sometimes cleaned in dark!
CMP: Clean
CMP: New Nodes
Abrasive Free, Fixed abrasive, Micelle Electro polish