MTC_CMP

The Microelectronics Training Center, IMEC v.z.w.www.imec.be/mtc

delfi.imec.be

MTC 2003 : Silicon Processing courseIMEC© 2003

David J. SteinPage 1

CMP

David J. Stein, Ph.D.Sandia National Laboratories

andUniversity of New MexicoAlbuquerque, NM USA

[email protected]

Acknowledgements

Sandia National LabsDale Hetherington, Reid Bennett

IMECKatia Devriendt

RodelMike Oliver

CabotPaul Feeney

Sharp Microelectronics USADave Evans

Laredo TechnologiesTom Tucker


delfi.imec.be



Outline

IntroductionWhat is CMP?– Overview – CMP Consumables (focus on slurries and pads)

CMP processes and applicationsGlass polishing basicsDielectric CMP– ILD, STI

Metal CMP– W, Cu

Novel applications– MEMS, Ag, replacement gates…

Chemical-Mechanical Polishing (CMP)

Film topography

Post CMP

Process ParametersPad and slurry typePolishing tool

Pressure (carrier/backside)Velocity (carrier/platen)Carrier design Flow rate of slurryTemperature (slurry/pad)

Pad conditioning

Down Force

CarrierBacking filmWaferPad

SlurryRotation

Platen

Pad Conditioner


delfi.imec.be



CMP Images

Polish head (holds wafer)

Conditioning disk

Slurry dispense

IR probe

Technology enabled by CMP

1985-90 19971980

Spin-on resist& etchback

CMP of PECVD

dielectrics

Damascene Cu CMP

1995

STI CMP

2.5 µm

LOCOS

LOCOS isolation

STI

2000+

IBM, 1997

1987-92

CMP of WPolyNoble metalsMEMS…


delfi.imec.be



Driving Forces Behind Planarization

A) Film DepositionStep coverage

B) LithographyDepth of focus

C) EtchEtch stringers

Diagram - C. Paumier, SEMICON/Europa 95 Technical Conference

D) New materials and manufacturingex. damascene Cu CMP

Moore’s Law and CMP Applications

1970 1980 1990 2000 20100.01

0.1

1

10

Copper CMP

STI CMP

Tungsten CMP

ILD CMP

Mic

ron

Feat

ure

YearMinimum Feature Size Data: IEDM Short Course, 1999

New Applications

Poly CMP


delfi.imec.be



Planarization Roadmap

1999 INTERNATIONAL TECHNOLOGY ROADMAP FOR SEMICONDUCTORS

1999 2000 2001 2002 2003 2004 2005

0.1

0.2

0.3

0.4

Year

Min

imum

Fea

ture

(µm

)

0.2

0.4

0.6

0.8

1.0

Dep

th o

f Foc

us (µ

m)

Feature size decreasingChip size increasingDecreasing depth of focus budget:

stepper issues (~ 80%)wafer flatness (~20 %)

100 nm technology nodedesire < 50 nm of planarization within the die*

Global Planarity Issues

* CMPMIC Conference Roadmap Discussion March, 2001.

CMP Infrastructure

Process Technology

Process parameters:film & substrate type, polish speed, pressure, slurry flow

Process Specs:Uniformity, Planarity, Final Thickness

Pad Technology

Material properties:density, filler, porosity mechanical modulus

Pad Engineering:thickness, subpad, adhesive, grooves, pad life, conditioning

Slurry Technology

Material properties:colloid type, size, shape, stability

Formulation:pH, % solids, buffer, viscosity

Distribution:bulk mixing, recycling

Polisher Technology

Polisher Design:rotary, linear, orbital

Carrier Design:gimballing, floating, bellows, membrane

Throughput Design:single/multiple heads,multiple platens

Manufacturing Technology

Endpoint techniques:pad temperature, motor current, vibration, film thickness

CMP Process

Manufacturing Technology

Pad Conditioning:end effector type, cut rate, pad break-in, removal rate stability


delfi.imec.be



CMP Market

Tom Tucker, Laredo Technologies Inc. email: [email protected]


delfi.imec.be






delfi.imec.be






delfi.imec.be



CMP Process Sector

Polisher

CMP Process

Wafers In

Slurry Mixand

Distribution

Pad & InsertTechnology

SlurryReclaim

ConditioningTechnique

End PointDetection

ProcessMonitoring Post CMP Cleaning

Brush Scrubber

MegasonicClean

Metrology Defects and Film Thickness

Wafers Out

The current generation CMP tool is a dry-in dry-out system.

2 of 3 Basic Types of CMP Processes

Polish oxide

Stop on oxide

METAL-VIA FILLOXIDE

M

OXIDE

Polish metal

Stop on oxide

M

Continuous Layer CMP Damascene CMP

Tungsten, Copper, Aluminum,Shallow Trench, and Trench DRAM Poly

Dielectric (backend ILD)


delfi.imec.be



3rd Type (not discussed in presentation)

CMP smoothing process

Surface finishing

surface roughness

Examples: silicon surface finishing, polysilicon smoothing.

Polishing Terminology

Removal Rate

pre,post = thickness measurementsn = number of measurementst = polish time

Within Wafer Non-Uniformity (WIWNU)

RR implied; equals standard deviation of RR

Wafer To Wafer Non-Uniformity (WTWNU)

standard deviation of wafer-level RR means

Edge exclusiondistance from edge that measurements are excluded fromfor blanket wafers typically wafer diameter is reduced by 2 X edge exclusion for calculation of measurement point locations

Within Die Non-Uniformity (WIDNU)

standard deviation, across a wafer, of RR or TTV within a die

( )

nt

postpreRR

n

nn∑ −= 1


delfi.imec.be



Polishing Characterization

Wafer Level(due to tool and

consumable issues)Die Level

Wafer edge effects

Within-die effects (due to pattern density layout)

CMP Consumables

SlurriesPadsSlurry filtersPad conditioning disksCarrier insert films (backing films)Carrier retaining ringsChemicalsPolyvinyl alcohol (PVA) brushesDe-ionized waterMonitor wafers

√√

Subject of remainder of introduction presentation√


delfi.imec.be



CMP Slurries

Ultra fine solid particles dispersed in a solutionnormally water basedcolloid stability is important

Particle size and concentration varies10 nm < primary particle size < 1000 nm0.1 % < weight percent concentration < 30 %

Abrasive typesSilica, alumina, ceria

AdditivesOxidizers, buffers, surfactants, chelating agents, corrosion inhibitors, (other proprietary stuff).

Important Slurry Properties

Average particle size and distributionSpecific surface areaPercent of abrasive in solutionChemical additivesSpecific gravity, viscosity, and pHImpuritiesSettling characteristicsManufacturing processes and other details in Appendix


delfi.imec.be



TEM of Fumed Silica Particles*

* Courtesy of Cabot Microelectronics

TEM of Colloidal Silica Suspensions*

*Klebosol Colloidal Silicas


delfi.imec.be



Important Pad Properties

ThicknessSpecific GravityPore Structure

sizedensity or volume

CompressibilityHardnessChemical durability

IC1000specific gravity 0.7

IC1000specific gravity 0.9

Closed Cell Foam Pads - IC1400 (Class III)

top pad

bottom sub-pad

100 µm

IC1400 pad is manufactured by Rodel Corp.

glue layer


delfi.imec.be



Images of Molded Urethane Pad*

*Cabot EPICTM Pad

surface Open (interconnected) cells

Fixed Abrasive Polishing Pad*

PRESSURE-SENSITIVE ADHESIVE

Resilient Foam ( 50 - 90 mils)

ADHESIVE

ADHESIVE

Polycarbonate ( 20 - 60 mils)

BACKING

Structured Abrasive

SUBPAD

~ 250 µm

* 3M/Rodel Fixed Abrasive Pad

SEM images - A. L. Moy, SNL


delfi.imec.be



Micro-replicated Surface Contains Ceria Abrasive

0.001 0.01 0.1 1.0 10.0 100.0 103 104 µm

Aqueous Salts

Metal Ion

Atomic Radii

Colloidal Silica/Particles

Feature Size

Pad AsperityWafer

Die SizePad Groove

Polishing mechanisms Manufacturing Yield

CMP Length Scales - 9 Orders of Magnitude!

*M. Verhoff, Sandia National Labs


delfi.imec.be



Relative Scale of Asperities

10 µm

Oxide Fill Material Silicon Wafer

Polyurethane Pad

Al/Cu Metal

1µm step heights

20 µm pore diameter

Summary - Introduction Section

CMP continues to enable manufacturing of advanced IC’s

Driving forces for improved planarization– Photolithography (depth of focus)– Etching (clearing)– Deposition (step coverage)– New materials and manufacturing methods that involve

CMPLarge infrastructure (worldwide)– equipment, consumables, manufacturing methods and

controlApplications include:– STI, polysilicon, ILD, tungsten, copper, etc.


delfi.imec.be



Dielectric CMP

Mechanisms and Processing

Empirical wear equation

Brown(2) and Cook(3) showed derivationelastic Hertzian indentationmaterial removed equals indentation cross-sectional area * distance traveledwhere E = Young’s modulus

Cook(3) noted experimental kp and (2E)-1 don’t agree

E2PvPR =

Preston - Removal rate(1)

vPkRR rp ⋅⋅=

PR - Glass polish ratekp - Preston coefficient

(process dependent)P - Applied pressurev - Linear velocity

PR

(1) F. Preston, J. Soc. Glass Tech. 11 (1927) 214.(2) N. J. Brown, P. C. Baker, and R. T. Maney, SPIE 306 Contemporary Methods of

Optical Fabrication, 1981, 42.(3) L. M. Cook, J. Non-Cryst. Solids 120, 1990, 152.


delfi.imec.be



travel

compressiontension

particle oxide surface

Stress Contours - Travelling Hertzian Indenter*

SiO2 material in tensiondiffusion processes (layer becomes hydrated)

SiO2 material in compressiondissolution processes (pressure dependent dissolution)

*L. M. Cook, J. Non-Cryst. Solids 120, 1990, 152.

Water Reaction with SiO2

Water is necessary for effective polishing.Penetration of water into film is governed by diffusion.

a few atomic layers (Jaso, et al., PMIC 1997) (experiment Hydrogen profiling)1.2 nm (Cook, J. Non-Crystalline Solids, 1990) (calculation)75 -150 nm (Trogolo, et al., MRS Symp.1992) (experiment FTIR/TEM)

Reaction may produce thin hydrated (gel) layer which is abraided away during polishing exposing a fresh surface (process repeats itself).Dissolution of SiO2 enhanced at high pH (~ 10).Contact (wetting) angle on oxide surface decreases to 0 after CMP.

Si - O - Si + H2O 2SiOHSi

Si

O O

H

H

Si

Si

O

O

H

H

Si

Si

O

O

H

H


delfi.imec.be



0 1 2 3 4 5 6 7 8 9 10 11 120

1000

2000

3000

4000

5000

6000

7000

20 rpm

50 rpm

Carrier and platen speed - 90 rpmR

emov

al R

ate

(Å/m

in)

Carrier Pressure (psi)

rp vPkRR ∗∗=

Polishing Kinetics

kp varies with slurry, glass type, etc.

Oxide CMP Removal Rate ModelsPreston

wear equationCook

inclusion of elastic properties of particleMore recent

Zhang, et al. (hertzian-type plow describes removal)Tseng, et al. (expanded hertzian indenter to include shear stress)Zao, et al. (proposed pressure threshold mechanism but never confirmed experimentally)

See appendix I


delfi.imec.be



Degrees of Planarization

NoPlanarization

Smoothing

CompleteLocalPlanarization

CompleteGlobalPlanarization

Initial Step Height = t

Final Step

Final Step

t

Near GlobalPlanarization

Final Stepmm distances

CMP Processes

Planarization Length

Final Surface

Uniform polish Uniform polishNon-uniform polish region

Planarization Length

Typical polishing processes result in 3 - 5 mm planarization lengths.


delfi.imec.be



Back End of Line Dielectric CMP

Al/Cu MetalTungsten plug

Intermetal Dielectric Deposit Dielectric

Polish

CMP Issues:• No automatic endpoint• Global pattern density

variation• Deposition technique• Deposition thickness• Tight thickness control

Premetal Dielectric

Back-end Dielectric CMP

PreMetal Dielectric (PMD) CMP

Typically a phosphorous doped layerCritical layer for depth of focus (pattern density and planarity are key concerns)

Intermetal Dielectric (IMD) CMP

HDP commonly usedDielectric roadmap - low k dielectric materials

M1

M2

M3

WPMD

IMD1

IMD2


delfi.imec.be



Surface Profile and Gap-fill Technique

VoidingSiO2 Plasma Enhanced CVD TEOS(1)

Faceted topography

(1) L. Roherty-Osmun, Sandia Labs (2) J. T. Pye et al., Solid St. Tech., Dec. 1995.

Simultaneous dep-etch

PETEOS deposited with 5x dep/etch cycles

SiO2High Density Plasma(2)

PETEOS deposited without any dep/etch cycles

No Voiding

Pattern Density Variations - Measurement

84 % density - thickest oxide post CMP

4% density - thinnest oxide post CMP

Post-CMP Die Image4 8 16 28 44

12 20 32 48 64

24 36 52 68 80

40 56 72 84 92

60 76 88 96 100

aa

84% density

Section a-a

210 µm 40 µm

2 µm oxide

MIT/Sandia Density Mask Layout

2 mm

2 mm

10 mm


delfi.imec.be



Final Dielectric Thickness Vs. Pattern Density

Within-die range 520 nm

Target thickness

MIT/Sandia Density Mask

Case Study 1

Role of the Sub-pad on Die-level and Wafer-level Nonuniformity


delfi.imec.be



Sub-pad Experiment

foam sub-pad

(a) IC1000/foam sub-pad (b) IC1000/fiberglass sub-pad

Investigate 2 extreme cases of variations in sub-pad stiffnessCompare polishing results– within-die uniformity using Sandia/MIT density mask

Perform mechanical measurements– 3-point bending, compression, micro-indentation

IC1000 K groove

soft foam sub-pad

IC1000 K groove

fiberglass sub-pad

50 mil

50 mil

50 mil

50 mil

Bulk Mechanical Properties

0.000 0.002 0.004 0.006 0.008 0.010

0

200

400

600

800

1000 IC1000/fiberglass sub-pad IC1000/foam sub-pad

Flex

ural

Stre

ss (p

si)

Strain (in/in)0 5 10 15 20

0

200

400

600

800

1000

1200

1400 IC1000/fiberglass sub-pad IC1000/foam sub-pad

Com

pres

sive

Stre

ss (p

si)

Strain (%)

3-point bending test Compression test


delfi.imec.be



Micro-indentation Effects from Subpad

0 100 200 300 400 5000

5

10

15

20 IC1000/foam sub-pad IC1000/fiberglass sub-pad

load 10.2g

Pene

tratio

n de

pth

(µm

)Time (s)

Measurement data

The tip penetrates ~ 5 µm further into the IC1000/foam sub-pad compared to the IC1000/fiberglass under a 10 g load.

M

Quartz probe

60 mil radius+

1 mm 50 mil

50 mil

Penetration depth

WIDNU - Effect of Sub-pad

• IC1000/foam sub-pad shows highest die-level NU.• Speed has no effect on WIDNU.• Pressure shows effect on WIDNU but less than IC1000/foam sub-pad.

0.2 0.4 0.6 0.8 1.0 1.2

3 psi 30 rpm 3 psi 90 rpm 9 psi 30 rpm 9 psi 90 rpm

Normalized Polish Time (s)

0.2 0.4 0.6 0.8 1.0 1.2 1.40

1000

2000

3000

4000

5000

6000

7000


Oxi

de T

hick

ness

Var

iatio

n (Å

)


IC1000/foam sub-pad IC1000/fiberglass sub-pad

SS-12 slurry SS-12 slurry


delfi.imec.be



WIW Oxide CMP Thickness Uniformity - Subpad Effect

Within-wafer non-uniformity shows opposite trend from within-die non-uniformity. The mechanical

properties of the sub-pad must be optimized to produce the lowest WIDNU and WIWNU.

02468

1012141618

90 rpm30 rpm90 rpm30 rpm

9 psi3 psi

Waf

er-le

vel n

on-u

nifo

rmity

(%1σ

) IC1000/fiberglass sub-padIC1000/foam sub-pad

Note: SS-12 slurry used in all processes.

Case Study 2

Role of Pad Thickness on WIDNU and WIWNU


delfi.imec.be



Experimental Description

Thicker pad showed faster time to planarization.Stiffer pad (no sub-pad) showed faster time to planarization.Increase in time to planarize larger step heights is not proportional to the step height.

Pad type 800 nm initial stepheight 2000 nm initial stepheight55 mils FX-9/subpad 300 s 450 s80 mils FX-9/subpad 250 s 400 s

80 mils FX-9/no subpad 200 s 350 s

K. Devriendt, Proceedings of the 1999 CMP-MIC Conference, VMIC, Tampa, 227

Effect of Top Pad Thickness on WIDNU*

* K. Devriendt, Proceedings of the 1999 CMP-MIC Conference, p. 227.

WIDNU defined as the level difference between the up features of the 75% dense structures and the down feature of an isolated structure (0% density).

0

500

1000

1500

2000

2500

0 100 200 300 400 500

WIDNU - Effect of pad thickness

55 mils FX-9/subpad80 mils FX-9/subpad

WID

NU

(nm

)

Polish time (s)

Klebosol 30N50 PHN slurry


delfi.imec.be



Wafer-level Non-uniformity: Down Features

* K. Devriendt, Proceedings of the 1999 CMP-MIC Conference, p. 227.

0

50

100

150

200

250

0 50 100 150 200 250 300 350

800 nm initial stepheight50 % pattern density

55 mils FX-9/subpad80 mils FX-9/subpad80 mils FX-9/no subpad

WIW

NU

(nm

)

Polish time (s)

Klebosol 30N50 PHN slurry

Case Study 3

Conditioning and Pad Wear Effects


delfi.imec.be



10 20 30 40 50 60500

1000

1500

2000

Rem

oval

Rat

e (Å

/min

)

Cumulative Polishing Time (min)

Removal Rate Reduction - No Conditioning

SEMs of pad surface

No Pad Conditioning IC1400 Pad

200 µm

Mesas

100 µm

Removal Thickness Vs. Time

1 2 3

2000

4000

6000

8000

10000

12000

14000

Pad Conditioningin-situex-situ

in-situex-situ

Pad Conditioning

Process B

Process A

4475 Å/min

3735 Å/min

973 Å/min1146 Å/min

Oxi

de T

hick

ness

Rem

oved

(Å)

Polish Time (min)

A: 9psi, 90 rpm

B: 5 psi, 20 rpm

Process

IC1400 PadSS-12 SlurryThermal OxideIPEC 472 Rotary


delfi.imec.be



Within-Die Uniformity - Pad Wear Effects

Oxi

de T

hick

ness

Var

iatio

n (Å

)

Data: K. Achuthan, Clarkson U. Thesis, 1998.

MIT/Sandia pattern density mask

Each data point represents a wafer polished to planarization at a particular moment of the pad life.

Case Study 4

Silica versus Ceria slurry


delfi.imec.be



Blanket Polish Rates - PECVD oxide

0

2000

4000

6000

8000Speed Slurry

9 psi3 psi

Rem

oval

Rat

e (Å

/min

)

Pressure

30 rpm, silica, SS12 90 rpm, silica, SS12 30 rpm, ceria, STS1000DV 90 rpm, ceria, STS1000DV

Pad type IC1400

Die-level Uniformity - IC1000/foam Sub-pad

0.2 0.4 0.6 0.8 1.0 1.2 1.40

1000

2000

3000

4000

5000

6000

7000


Die

-leve

l Oxi

de T

hick

ness

Var

iatio

n (Å

)


SS-12 Silica slurry STS1000DV Ceria slurry

0.2 0.4 0.6 0.8 1.0 1.2 1.4



Ceria slurry shows better die-level uniformity.


delfi.imec.be



Wafer-level Non-uniformity*

0

5

10

15

20

90 rpm

90 rpm

30 rpm

30 rpm

9 psi3 psi

Waf

er-le

vel n

on-u

nifo

rmity

(% 1σ)

Silica slurry Ceria slurry

IC1000/foam sub-pad

Summary of Results*

Die-level uniformity Wafer-level uniformityBest Worst Best Worst

PAD IC1000/fiberglass

IC1000/foam IC1000/foam IC1000/fiberglass

SLURRY Ceria Silica Ceria or Silica Ceria

SETTING 3 psi 90 rpm 9 psi 30 rpm 9 psi 30 rpm 9 psi 90 rpm

* D. L. Hetherington, et al., ECS Symposium on CMP in IC DeviceManufacturing, Oct., 1999.


delfi.imec.be



Shallow Trench Isolation CMP

Device Isolation Techniques - LOCOS vs. STI

LOCOS-based techniques in 0.25 µm technology

+ Manufacturable, simple,good know-how

- Large ‘bird’s beak’- Deep submicron scalability

Shallow Trench Isolation with Chemical Mechanical Polishing

+ Small active area pitch possible+ Denser device packing+ Superior planarity!

Further down-scaling

LOCOS LOCOS

p-well

n-wellp-wellSTI STI

n-well

Lateral isolation candidates


delfi.imec.be



Si

Si

Si3N4

Si

Si

Si3N4

Si3N4

Si3N4

Isolation stack :Thermal oxide growth (pad oxide)Active protecting nitride

Pattern and etch ofSi3N4 and SiO2

Trench etch into Si

Thermal oxidation of trench sidewall,corner rounding

SiO2

SiO2

SiO2

SiO2

SiO2

STI Process Flow

STI Process Flow

Si3N4

Si

SiO2

Si3N4

Si

SiO2

Si

SiO2

Deposition of trench fillingoxide (f.ex. HDP-CVD)

Oxide CMP Polish step with stop on nitride

Removal of nitride layer(wet etch in phosphoric acid)

SiO2

SiO2

SiO2


delfi.imec.be



CMP STI Polishing Regimes

I. Removing step height and oxide over the active area

II.Over-polish to ensure that all oxide is removed over active area everywhere on the die.

Trench OxideNitride

Silicon Active Area

Trench Oxide

Trench OxideSilicon Active Area

Nitride

Incomplete Polish or Erosion?→ Dependent on fill and integration approach ! Over-polished

Si

SiO2

SiO2Si3N4

densitypattern1

STI-CMP Issues

Pattern layout effectslocal removal rate ~ local pressure ~

variations in the active area to field step height after nitride strip – problems at subsequent lithography steps!

dishing in large field areas influence on surface planarity!


delfi.imec.be



The Problem

overpolished open area

underpolished dense area

Residual nitride in dense SRAM area

TEOS-Ozone HDP-CVD OxideLocal planarization as a function of active area width:

+/- Same oxide thickness everywhere

Different pyramids dependenton active area width

Difference in CMP behavior! (pattern density effects)

Trench Fill Method


delfi.imec.be



Solutions to Pattern Density CMP STI Problem

(1) Dummy fill(2) Oxide reverse etch (blockout) mask(3) Dual Nitride(4) Resist Block/Etchback Assisted CMP(5) Without Additional Litho Step

approaches(6) Consumables(7) Oxide dummy structures(8) Combinations of the above

Dummies widely used in oxide CMP BEOLeasiest way to solve pattern density variations

Similar idea for STI levelLarge field areas around isolated active areas can be filled up with dummy active areasrisk of overpolishing is reduced

Dummies offers a lot of design issuesdesigners don’t like itmixed signal technologies are especially sensitive to dummy charging–capacitive coupling and noise

Use of Dummy Structures


delfi.imec.be



Examples of Dummy Metal-FillGrounded metal-fill for DEC 600 Mhz Alpha chip(a)

Active Lines Metal fill(a) P. Gronowski, et al. IEEE J. Solid-State

Circuits, vol. 33, 5, p. 676, (1998).

Floating metal-fill

Active Lines Metal fill

STI Dummy Fill

STI dummy fill

Courtesy of J. Soden Sandia National Labs


delfi.imec.be



Modifying Existing Layouts

Requires analysis of within-chip layout density variationsdummy fill optimization

Optimized fill structure reduces density variation from 29% to 18%

Dummy FillPhysical process model-based dummy structure design:

Without dummystructure:

With dummystructure:

A

R Tian and X. Tang, IEEE TRANSACTIONS ON COMPUTER-AIDED DESIGN OF INTEGRATED CIRCUITS AND SYSTEMS, VOL. 21, NO. 1, JANUARY 2002


delfi.imec.be



Oxide Dummies - 6.25% Density

2000

3000

4000

5000

6000

7000

8000

40 60 80 100 120 140 160 180

1 - Field1 - Dummy4 - Field

Thic

knes

s (Å

)

Time (s)

yellow = dummy

blue = active

3000

4000

5000

6000

7000

8000

9000

40 60 80 100 120 140 160 180

2 - Field3 - Field3 - Dummy

Thic

knes

s (Å

)

Time (s)

Oxide Dummies - 25% Density

yellow = dummy

blue = active


delfi.imec.be



Oxide Dummies - Field to Dummy Comparison

2 3

4000

5000

6000

7000

8000

9000

10000

11000

12000

20 40 60 80 100 120 140 160 180

ata 5 38 8/8/ 00

2 - Dummy2 - Field3 - Dummy3 - Field

Thic

knes

s (Å

)

Time (s)

yellow = dummy

blue = active

Oxide Reverse Etch Approach

Oxide Reverse Etch (ORE), followed by CMP :

pad oxide / nitride depositiontrench patterning and etchingsidewall oxidation / deposition of trench filling oxideremove (most of) the oxide on large active areas with additional litho and etch stepremove remaining oxide from the active areas during the CMP stepnitride removed during etch step

Parameters to be optimized:• Oversize (litho misalignment)• Amount of oxide etched back?

Etch of (part of the) oxidein large active areas

Process flow : HDP oxideNitride

Oversize


delfi.imec.be



Dual Nitride Approach*

Protection of isolated active areas and wide field regions

No field protection in dense areas!

Oversize of 0.4 µm

Patterning of 2nd nitride (DN), followed by CMP:

Parameters to be optimized:• Oversize (litho misalignment)• Topography nitride on active to nitride on field

* G. Badenes, et al. JECS October 2000

– Isolation patterning – Etching of trenches / filling trenches with oxide– Patterning of a block resist over the large field areas to

bring up the surface to about the same level as oxide on the active areas / hardening of block resist

– Spinning of planarizing resist and cure– Reactive Ion Etch (RIE with 1:1 selectivity of resist to

oxide): flat resist surface transformed to flat oxide surface– (stopped before removing all oxide on active nitride)– Removal of spikes during short CMP step– Excellent planarization performance– RIE process and a planarizing resist required

Resist Block Approach

Process flow :

Planarizing resist

*B. Davari et al., IEDM Tech. Digest, p 61 (1989)

Block resistHDP Oxide


delfi.imec.be



Deposit thin nitride overcoat

HDP Oxide

Without Additional Litho Steps

Trench filling with oxide and deposition of a thin nitride layer

CMP with low selectivity slurryNitride overcoat is removed after a short polishing time in the dense active areas and on the small isolated ones

The large field areas are protected until the nitride overcoat above the dense areas is removed

*J.M. Boyd, J. P. Ellul, J. Electrochem. Soc, 143, p 3718 (1996) J.M. Boyd, J. P. Ellul, J. Electrochem. Soc, 144, p 1838 (1997)

approach applicable? Depending on the design!

Fixed Abrasive Pad

Abrasive: Cerium oxideSelectivity: (topography vs planarized film)

200 : 1Selectivity: oxide to nitride (planarized film)

1.2 : 1.0Low overfill thickness required:

CVD oxide higher throughputLess polishing time, higher throughput

Other advantages:Selective toward topography self stopping behavior!

Issues:Very dependent on mask set layout!Tight control on overfill thickness

M3100 3M/Rodel pad


delfi.imec.be



Fixed Abrasive Pad

0

200

400

600

800

1000

1200

1400

1600

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

Pattern Density (100 µm Line Pitch)

Low selectivity slurry

High selectivity slurry

M3100 Fixed Abrasive

Nitr

ide

Loss

(Å)

Fixed Abrasive Pad

0

500

1000

1500

2000

2500

3000

3500

4000

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

Pattern Density (100 µm Line Pitch)

Tren

ch O

xide

Los

s (Å

) Low selectivity slurry

High selectivity slurry

M3100 Fixed Abrasive


delfi.imec.be



Defects

Defects


delfi.imec.be



Defect Wafer Map

Die X Value

Die

Y Va

lue

STI

Trench Cut

Defect Source


delfi.imec.be



Edges

Edges can’t be ignored!

Edge ComparisonUnprotected Protected


delfi.imec.be



EBR to Bevel

Summary - Dielectric CMP

Pattern density effects is the key problem to solve for dielectric CMPDefectivity is also importantSlurry and pad engineering is focused on solutions for improving planarity and minimizing defectsChip layout can be optimized for CMP process


delfi.imec.be



Developments In Metal CMP Processes

CMP Removal to Stop Layer

Material removal and planarizationW plugs, copper dual damascene, aluminum (dual) damascene, STI, polysilicon-filled trenches

non-selective materialdeposition (conformal

or non-conformal)

CMP is selectiveto the stop layer


delfi.imec.be




Tungsten CMP

Process Overview

Multiple level etch stop

Photo Pattern and Etch Contacts

Sputter Ti/TiN & CVD W

Tungsten CMP

W

Plug erosion/ ILD scratching

CVD W

High aspect ratio contact fill

CVD Tungsten Filled Vias

Post Polished Tungsten Via

Process Sequence Cross-section of Via Fill

Post W CMP Surface


delfi.imec.be



Process and Technology ConsiderationsW CMP (vs. etchback) is the preferred method for sub - 0.5 µm technologiesW dual damascene has also been integrated into DRAM manufacturingIssues

removal rate and non-uniformitypad life and pad conditioningbarrier stack polish rateselectivity W to oxide removal ratesW plug key-holing, corrosion, and dishingdeep scratches in oxideoxide erosionpost CMP cleaningprocess stabilityprocess end-point

W CMP ConsumablesSlurries

abrasives– alumina– silica– ceria (only commercial in fixed-abrasive pad)

oxidizers– potassium ferrocyanide– potassium iodate – ferric nitrate– hydrogen peroxide– hydrogen peroxide with catalyst

pH range– acidic

pH 1.5 for ferric-based solutionspH 4.0 (buffered) iodate-based solutionsup to ~ pH 6


delfi.imec.be



W CMP ConsumablesPads

polyurethane– hard pad (closed pore, /w or w/o grooves)

Rodel IC1400, Thomas West P777, Freudenberg FX9 & PANW, Universal Photonics LP line

– soft pad (/w or w/o embossed channels)Rodel Politex, Universal Photonics UNI-NAP 12

new pad developments– open-celled polishing pad material (Cabot)– fixed abrasive pad material (3M)

conditioning disksNi-diamond end effectors

– higher corrosion rate in metal CMP slurry– reliability is a concern (diamonds may fall out and stick in the pad)

CVD diamond end effector (e.g. Diamonex)nylon brush for soft pads

Grain Size EffectAl

W

W-T. Tseng, Y-L. Wang, J. Niu, Thin Solid Films 370

Al

W


delfi.imec.be



Defects

erosion

Electrical contactnot made

Tungsten CMP

Metal patternand etch

Formation of nextmetal level

tungsten

PETEOStungsten plugs

Tungstenkeyhole

Locallyplanar

Locallyunplanar

Dishing

dishing and planarity

erosion

Dishing and Erosion Examples

P-11 profile scan location

4x4 array of tungsten vias (diameter 0.9 µm space 1.5 µm)

D. J. Stein, 1998 CMP-MIC, p. 161

Thin oxide lines


delfi.imec.be



Pattern Erosion Comparison

-2500

-2000

-1500

-1000

-500

0

500

0 1000 2000 3000 4000 5000

Prof

ile (Å

)

Distance (µm)

Politex pads

IC pads

IC1400 (24 °C)

IC1400 (44 °C)

Politex embossed (24 °C)

Politex regular (24 °C)


Relationship Between Erosion / Pattern Density

densitypattern area totalarea metal

=Φ

=Φ ( )ott −Φ−

Φα=ε

1

KvPerosion

=α=ε

N. Elbel at al., J. Electrochem. Soc. 145 (5), May 1998

t,,P,v)erosion( Φ↑ε


delfi.imec.be



Dishing - AFM Scans

IC1400 pad Politex pad


Pattern Density Effects on Dishing

area totalarea metal

=ΦDishing decreases withline width (thinner lines less dishing)increase in metal pattern density (more erosion)oxide polish rate (more erosion)

more metal

N. Elbel at al., J. Electrochem. Soc. 145 (5), May 1998


delfi.imec.be



Comparison of Metal CMP End-Point Methods

Study focuses on W CMP process

End-PointMethod

Detector Technique Issues

Padtemperature

IRtemperaturesensor

Sense change in padtemperature

pad wear

Carriercurrent

Motor currentsensor

Sense change infriction

motor currentnoise

Platen current Motor currentsensor

Sense change infriction

motor currentnoise

Filmreflectivity

Spectrometer Sense change inwafer surfacereflectivity

positioning ofspectrometer

Ming-Cheng Yang, et al., Proceedings CMP-MIC, 1998, p. 216

Pad Temperature and Motor Current Endpoint

Ferric nitrate based alumina slurry

G. Springer, Proceedings CMP-MIC, 1999, p. 45.


delfi.imec.be



Reflectivity End-point

Ref

lect

ivity

wavelength (λ)

time (s)Det

ectio

n al

gorit

hm

Ming-Cheng Yang, et al., Proceedings CMP-MIC, 1998, p. 216


Cu CMP


delfi.imec.be



Process Flow

previous level metal nitride hard mask via ILDpattern and etch hard mask

for vias

deposit line ILD over viahard mask, deposit, pattern,

and etch line hard mask

Process Flowdry etch (complicated!) to openspaces for damascene metal

seed layer deposition and ECDgrowth of Cu


delfi.imec.be



Process Flow - CMP

Cu dual-damascene CMP is similar to W CMP

CMP of conductor and barrier materialsstop on ILD/barriersimilar defect issues - dishing and erosion

Some issues in Cu CMP are different and harder to solve

CMP removal of excess materialto leave damascene vias and lines

Comparison to W CMPBoth W and Cu CMP require:

polish deposited metal through nitride barrier to oxide stopminimal dishing and erosiongood selectivity to oxide polish stoplow defects

Cu CMP is complicated by:polishing lines instead of plugs - electrical performance directly impacted by overpolishvery non-uniform deposition profile - can be worse than oxide CMPCu corrosion


delfi.imec.be



Cu and Dual-damascene CMP IssuesCu as a ECD material

very soft and susceptible to corrosionspontaneously anneals after deposition (up to or over 48 hours for stabilization at room T)

Cu as line metal in dual damasceneline thickness now determined by CMPline thickness directly affects electrical performance

Cu deposition profile affects CMPBarrier material

can it be removed using Cu slurry?

Approach 1 - 1:1:1 Selective Slurry

1:1:1 slurry may create die- and wafer-levelnonuniformity.


delfi.imec.be



Approach 2 - 2-step Stop on Barrier

2-step (stop on barrier) may dish copper.Too little barrier in isolated areas may beremoved (similar to STI)

Approach 3 - 2-step Stop on Oxide

1 step (stop on oxide) may dish copper anderode oxide.

st

2 step (1:1:1) creates die- and wafer-levelnonuniformity.

nd


delfi.imec.be



Topography After 1st Step of 2-step Slurry

The dishing is very small at the clear time

Dishing as a function of the polishing timeMeanders with a line width of 10 um

Different pattern densities

10 %20 %50 %67 %96 %

Pattern density

0

200

400

600

800

1000

0 50 100 150 200

Dis

hing

(nm

)

Polishing time (s)

clear time

J. Grillaert et al., 2000 CMP-MIC Proceedings, p. 75 and presentation at SEMICON Europa, April 2000.


0.4 µm spacing - no dishing but excessive erosion

Dishing increasesas pattern densitydecreases

Dishing as a function of the polishing timeMeanders with a line width of 10 um

Different pattern densities(Enlargement)

0

50

100

150

200

250

300

0 50 100 150 200

10 %20 %50 %67 %96 %

Dis

hing

(nm

)

Polishing time (s)

clear time



delfi.imec.be




Pattern densityincrease

•Lowest density: no erosion•Highest density: erosion increases immediately after the clear time•Intermediate density: first no erosion, then increase

Erosion as a function of the polishing timeMeanders with a line width of 10 um


0

50

100

150

200

250

300

350

400

0 50 100 150 200

96 %67 %50 %20 %10 %

Eros

ion

(nm

)

Polishing time (s)

Pattern densityclear time


Topography Issues

Dishing and erosion increasing with polishing timelow density features – heavy dishing but low erosionhigh density features – low dishing but heavy erosion

Accurate end-point detectionminimize dishing and erosionminimize process time thus increase throughput

Most uniform process is desiredleast WIW and WID NU


delfi.imec.be



Cu 112"Cu 122"Cu 132"

Electrical Results - Various Pattern Densities

Increase in time

20

25

30

35

40

96916650332010

Sheet resistance as a function of polishing timeMeanders with a line width of 10 um


Shee

t res

ista

nce

(moh

m/s

q)

Pattern density (%)Increase in erosionIncrease in dishing

Pattern density has a large influence on sheet resistanceJ. Grillaert et al., 2000 CMP-MIC Proceedings, p. 75 and presentation at SEMICON Europa, April 2000.

Electrical Results - Various Pitch Constant Density

Sheet resistance as a function of polishing timeMeanders with a pattern density of 50 %

Different line widths

20

25

30

35

40

0.3 0.4 0.5 0.7 1 2 5 10 30

Cu 112"Cu 122"Cu 132"

Shee

t res

ista

nce

(moh

m/s

q)

Line width (um)The line width has only a minor influence

Increase in time



delfi.imec.be



Electrical Results (Density)

The variation in the sheet resistance is the largest for the lowest andthe highest pattern density, i.e. the structures that are most sensitivefor dishing or erosion. ==> crucial importance of a uniform polishing and deposition processes.

20 25 30 35 40

96 %

91 %

67 %

50 %

33 %

20 %

10 %

Distribution of the sheet resistanceMeanders with a line width of 10 um


Sheet resistance (mohm/sq)

Pattern densityincreases


20 25 30 35 40

0.3 um0.4 um0.5 um0.7 um

1 um2 um5 um

10 um30 um

Distribution of the sheet resistanceMeanders with a pattern density of 50 %

Different line widths

Sheet resistance (mohm/sq)

The distribution of the sheet resistance across the waferis independent of the line width.

Line widthincreases

Electrical Results (Pitch)



delfi.imec.be



Time and Chemistry Effects on Dishing

SlurryAluminaH2O2carboxylic acid complexantproprietary stabilizer

Wafers6000 Å trenches in thermal SiO2500 Å sputtered TiCu by sputter

Timed endpointDifferent slurry than previous case study!

V. Nguyen et al., “Dependency of Dishing on Polish Time and Slurry Chemistry In Cu CMP”, Microelectronic Engineering 50, 403.

Linewidth and Pattern Density

Dishing increases with linewidth.

Dishing is roughly independentof pattern density.

(note difference with previous example)


Without overpolish


delfi.imec.be



Overpolish


Dishing increases with overpolish.

Dishing increases with linewidth.

[Oxidizer] and Cu Dep Thickness

Dishing decreases with H2O2 concentration.

Dishing decreases with thicker Cu depositions (1.5 µm vs. 0.8 µm).



delfi.imec.be



Cu CMP Consumables

Padssame as for W except 3M fixed abrasive also workssimilar issues as W CMP pads (discoloration, short lifetimes, conditioning)

Slurriesdepends on if one or two step process is usedall have to, at a minimum, meet these requirements– low Cu and barrier corrosion rates– low defect generation

Cu Slurry Chemical Effects

M. Pourbaix, Atlas of Electrochemical Equilibira in Aqueous Solutions, NACE, Houston, 1974


delfi.imec.be



Example Cu SlurriesFe(NO3)3 or HNO3 and inhibitor (BTA)

acidic pH (<3) - inhibitor needed to prevent Cu corrosionhigh Cu / low SiO2 polish ratecolloidal alumina is stable

H2O2 and amino acidsacidic pH (<7) - inhibitor needed high Cu / moderate SiO2 polish ratecolloidal alumina is marginally stable

Neutral oxidizerneutral to akaline pH (>7) - inhibitor not needed to control Cu corrosion ratelow Cu / high SiO2 olispolish ratecolloidal alumina is not stable

Grain Structure Evolution

48:000:45

SΩ

0 10 20 30 4015

16

17

18

19

20

R(m

/sq)

Time (hours)

Similar time dependence as RS !

24:003:45

10 µm

5:00 8:00

Source: S. Brongersma (Imec)


delfi.imec.be



Hardness and Resistivity vs. Anneal Time

room temperature anneal

K. Smekalin et al. MRS Proceedings 566, p. 143, April 1999

PR and Resistivity vs. Anneal Time

room temperature anneal



delfi.imec.be



Elevated Temperature Anneals

5 minutes in inert gas


Novel CMP Consumable


delfi.imec.be



Slurry-free Cu CMP

V. H. Nguyen et al., Microelectronic Engineering 55, 2001, p 305.

Slurry-free Cu CMP


A = slurry free

B = IC1000/Suba IVand QCTT1010 slurry


delfi.imec.be



Slurry-free Cu CMP

A = slurry free


B = IC1000/Suba IVand QCTT1010 slurry

Abrasive-Free (AF) Cu CMP

RR vs. downforce30 rpmnon-linear (saturates)WIWNU decreases to asymptote

RR vs. platen speed22 KPa fixed downforceremoval rate saturatesWIWNU decreases to asymptote

S. Kondo et al., JECS 147 (10), 3907


delfi.imec.be



Role of Grooves in AF CMP

Erosion and Dishing in Abrasive-Free Cu CMP

Both dishing and erosion are significantly reduced by using abrasive-free CMP

S. Kondo et al., JECS 147 (10), 3907


delfi.imec.be



Erosion and Dishing, Slurry vs. AF CMP

Slurry-based CMP (top left) shows significant dishing in the bond pads while AF CMP shows no dishing (bottom left)Less dishing results in lower line resistance (above)

S. Kondo et al., JECS 147 (10), 3907

Low-K


delfi.imec.be



Process Needs: Pre-metal Dielectric (PMD)

PMD layer issuesCritical layer for lithography depth of focusPlanarity/pattern density improvements requiredAt 100 nm Technology node: WIDNU < 50 nm.

Short TermConventional SiO2 materials

Long TermLow k (which materials?)

Process Needs: Low k / Ultra low k Low k and Ultra Low k CMP processes must be compatible with copper material

Current scenario: No direct CMP of low k– Capping layer CMP (Conventional SiO2)– Touchup post Copper CMP (dielectric exposed)

CMP of low k materialsNew consumables / process parametersDefectivity/Reliability issues– poor adhesion, scratching– change in surface properties

low friction low k CMP process– see S. M. Jang, et al. VLSI Tech. Symp., p. 18,

2002.


delfi.imec.be



1 kÅ

4 kÅ1 kÅ

5.5 kÅ

1 kÅ PVD Cu seed250 Å PVD Ta

Sematech CMP Study: low - k films*

*J. T. Wetzel, et. al. IEDM 2001, p. 73

CMP adhesion test for various low-k dielectric films

CMP test blanket film stack

CMP Failures vs. Hardness/Modulus*

Note: Process of record for Cu CMP used

at facility

*J. T. Wetzel, et. al. IEDM 2001, p. 73


delfi.imec.be



Cu/CVD Low-k (k< 2.5) Interconnects*

1 µm

~1000 Å

Cu pad AFM scan AFM results

8 levels of Cu/Low-k (k< 2.5) 1 level Cu/USG

*T. I. Bao, et al., IEDM Tech. Digest, Dec. 2002.

CMP of porous SiOC:H films (low friction polishing)

Cu Summary

Cu CMP is not easy!Similar to W CMP except– highly corrosive– soft (susceptible to dishing and erosion)– dishing and erosion enhanced because polishing lines

not plugs (higher density Cu exposed at end of process)

– hard to find a slurry that will polish Cu and barrier - 2 step slurry systems common

Pattern density references (ILD also)– http://www-mtl.mit.edu/Metrology/Publications.html


delfi.imec.be



Novel Metal CMP Processes

Silver Metallization

R. Manepallei et al., IEEE Transactions on Advanced Packaging 22 (1), 4.


delfi.imec.be



Silver Metallization

M. Hauder, Microelectronic Engineering 64, 73.

Noble Metal CMP

Enabling New Memory Devices


delfi.imec.be



DC icorr Measurements

Tafel plots are used to determine icorr

Electrode i-E behavior under activation polarization is described by the Butler-Volmer equationDetermine icorr by fit of i-E data to Butler-Volmer equation

i i e ecorr

nFRT

nFRT= −

−α η β ηButler-Volmer

Equation 10-6 10-5 10-4 10-3 10-2-0.25

-0.20

-0.15

-0.10

-0.05

0.00

0.05

0.10measuredfit

abs(i) (A cm-2)

E vs

. SC

E (V

)

OCPicorr

Noble Metal CMP - Pt

Pt w/ abrasion250 Å min-1

Pt w/o abrasion8 Å min-1

Pt w/ abrasion, no chemistry300 Å min-1 on polisher

Pt w/o abrasion, no chemistry

Actual polish rate610 Å min-1

K. Moeggenborg et al., CMP-MIC 2001 Proceedings, Santa Clara, VMIC, 150.


delfi.imec.be








delfi.imec.be



Noble Metal CMP – Pt/Ir

0

500

1000

1500

2000

0 20 40 60 80 100

Polis

h ra

te (Å

min

-1)

Fraction Ir

D. Evans, personal communication.

Novel Gate Structures


delfi.imec.be



Damascene Metal Gate

Why damascene metal gates instead of poly?

high poly gate resistance limits speedboron penetration from doped poly changes threshold Vgate leakage degrades device reliabilitythreshold voltage deviation

A. Yagishita et al., IEEE Transactions on Electron Devices 47 (5), 1028.

Damascene Metal GateSi N3 4

PolyCombination of wet and dry etch



delfi.imec.be



Damascene Metal Gate

Damascene metal gatesallow for high-K gate dielectricsMinimizes plasma damagecomplete planarity and high scalabilitylow sheet resistivity thus lower RC time delayno depletiondrastic improvement in GOI


Acknowledgments

Dale Hetherington, Katia Devriendt, Mike Oliver, Dave Evans


delfi.imec.be



And Finally…

Si-Ge Smoothing with CMP*Device structure

AFM Results

*N. Sugii, et. al., IEDM 2001, p. 737.

CMP surface

CMP applied after the Si0.7Ge0.3layer grown.


delfi.imec.be



CMP Eliminates MEMS Design Constraints

Without CMP Planarized with CMP

Microgear, Linkage, and Hub Assembly

Planar linkage arm

Design Interference

Advanced MEMS Structures

Latch Mechanism Gear and Track Assembly - 5 Levels of Poly

Torque ConverterPop-up Mirror


delfi.imec.be



CMP really is rocket science!

Appendix IOxide CMP Removal Rate

Models and Studies Adapted from: D. J. Stein, D. L. Hetherington,

ECS Third International Symposium on Chemical Polishing in IC Device Manufacturing,October 17-22, 1999, Honolulu, Hawaii.


delfi.imec.be



Oxide CMP Removal Rate Models

Prestonwear equation (1)

Cookinclusion of elastic properties of particle (2), (3)

More recent Zhang, et al. (hertzian-type plow describes removal) (4)Tseng, et al. (expanded hertzian indenter to include shear stress) (7)Zao, et al. (proposed pressure threshold mechanism but never confirmed experimentally) (8)

Empirical wear equation

Brown(2) and Cook(3) showed derivationelastic Hertzian indentationmaterial removed equals indentation cross-sectional area * distance traveledwhere E = Young’s modulus

Cook(3) noted experimental kp and (2E)-1 don’t agree

E2PvPR =

Preston - Wear Rate(1)

vPkRR rp ⋅⋅=

PR - Glass polish ratekp - Preston coefficient

(process dependent)P - Applied pressurev - Linear velocity

PR


delfi.imec.be



Zhang and Busnaina(4) - PR=K(Pv)1/2

Total force is applied force plus electrostatic Van der Waals attractive force

either particle or surface deforms

A = Hamaker constant, zo is particle-surface separation distance

They assume Hertzian plowing removal, surface roughness equals particle indentation depth, and MP model(5):

2o

o z6ARF =

+=

o

2

oa Rza1FF

HaFF 2a π=+

Zhang and Busnaina(4) - PR=K(Pv)1/2

Normal stress = contact pressure:

From MP model(5) :

Use tribological approximation

Runnels(6) model

They obtain:

2a

c aFFP

π+

==σ

HaFF 2a π=+ HPc ==σ

snCPR τσ=

( )21

s PvC µ=τ

( )21

sn Pv'KKPR =τσ= 21

KH'K µ=


delfi.imec.be



Tseng et al.(7) - PR=KP5/6v1/2

Started with Runnels(6) PR=Cσnτs

Hertzian indenter

soUses Runnels shear stress approximation

Combining yields

KPD

23

NLF

2

== 2n aFπ

=σ31

2

22

1

21 r

E1

E1L

43a

ν−+

ν−=

31

Pn ∝σ

21

s PvPC ∝µ=τ

21

65

vKPPR =

Shi and Zhao(8,9,10) - PR=KP2/3v

Pad material is softer than particles or wafer hence particles must sink into padAssuming elastic deformation of pad

Number of particles per area of pad assumed constantHertzian plowing removalEach particle removes the same amount regardless of applied pressure hence

32

32

PFA ∝∝

vKPPR 32

=


delfi.imec.be



Zhao and Shi (8,9,10) - Pressure Threshold

Modification of previous derivationParticles roll in pad if applied pressure below threshold value

−= 3

2

threshold32

PPKPR

Polish Rate Experiments(11)

Thermal oxide

PECVD TEOS

VTR TEOS(710 °C)

HDP(SiH4 and

O2)varying

[Si]

BPSG5 flavors

Klebosol 1498-50

neutral pH 50 nm silica

neutral pH and pH 10,

5.5, 2.5

Klebosol 1508-50

NH4OH stabilized pH 10.8

50 nm silica√

Klebosol 30H50

pH 2.5 50 nm silica tried twice

Cabot SS12 pH 10.9-11.2130-180 nm √ √ √ Limited √

Rodel STS1000DV

pH 6.1 300 nm

predominately ceria

√ √ Limited

Films investigated

Slurry Description


delfi.imec.be



Polish rate (K P v) Comparison- Silica Slurries(11)

0

1000

2000

3000

4000

5000

6000

0 20 40 60 80 100

SS121508-50 pH 10.81498-50 pH 7.0

Polis

h ra

te (Å

min

-1)

P*v (kPa m s-1)

PR = K P v K = 72.1, R2 = 0.8891K = 64.1, R2 = 0.9626

K = 51.9, R2 = 0.9790

Thermal Oxide

Polish Rate - Silica Slurries Variable pH(11)

0

1000

2000

3000

4000

5000

0 20 40 60 80 100

1498-50 pH 7.01498-50 pH 10.01498-50 pH 5.51498-50 pH 2.5

Polis

h ra

te (Å

min

-1)

P*v (kPa m s-1)

PR = K P v

K = 51.9, R2 = 0.9790K = 55.6, R2 = 0.9701

K = 46.8, R2 = 0.9792

K = 50.6, R2 = 0.9156

Thermal Oxide


delfi.imec.be



Silica and Ceria Slurries(11)

0

1000

2000

3000

4000

5000

6000

7000

8000

0 20 40 60 80 100

ThOx SS12ThOx STS1000DVVTR TEOS SS12PECVD TEOS SS12

Polis

h ra

te (Å

min

-1)

P*v (kPa m s-1)

K = 72.1, R2 = 0.8891

K = 61.8, R2 = 0.8738K = 66.5, R2 = 0.8985

K = 88.9, R2 = 0.9353

IV-67

BPSG with Silica Slurry(11)

0

2000

4000

6000

8000

10000

12000

14000

16000

0 20 40 60 80 100

0/00/55/55/02.5/2.5

Polis

h ra

te (Å

min

-1)

P*v (kPa m s-1)

K = 167.2, R2 = 0.8817

K = 156.8, R2 = 0.9207K = 165.6, R2 = 0.8973

K = 261.0, R2 = 0.6563

K = 99.5, R2 = 0.9373

IV-68


delfi.imec.be



HDP Oxide with Ceria Slurry(11)

0

1000

2000

3000

4000

5000

6000

0 20 40 60 80 100

1.4551.4791.5141.562

Polis

h ra

te (Å

min

-1)

P*v (kPa m s-1)

K = 29.9, R2 = 0.3842

K = 7.2, R2 = n/aK = 8.4, R2 = n/a

K = 77.3, R2 = 0.4920

R2 Fit - All Models(11)

Preston Zhang, et al Tseng, et al Shih, et alK*P*v K*(P*v)1/2 K*P5/6*v1/2 K*P2/3*v

SS12 ThOx 0.8891 0.7549 0.9299 0.78501508-50 ThOx 0.9629 0.8146 0.9454 0.90741498-50 ThOx 0.9790 0.7896 0.9165 0.9192

1498-50 pH 10.0 ThOx 0.9701 0.8119 0.9386 0.91531498-50 pH 5.5 ThOx 0.9792 0.7925 0.9217 0.91851498-50 pH 2.5 ThOx 0.9156 0.8647 0.9529 0.9168

STS1000DV ThOx 0.8738 0.8179 0.9410 0.8356SS12 VTR TEOS 0.8985 0.7567 0.9326 0.7832SS12 PECVD TEOS 0.9353 0.7583 0.9211 0.8338SS12 0/0 BPSG 0.9373 0.7547 0.9123 0.8370SS12 0/5 BPSG 0.9207 0.7656 0.9323 0.8180SS12 5/5 BPSG 0.6563 0.6534 0.9002 0.4167SS12 5/0 BPSG 0.8973 0.7330 0.9235 0.7159SS12 2.5/2.5 BPSG 0.8817 0.7220 0.9221 0.6916

STS1000DV HDP n=1.455 0.4920 0.6649 0.7090 0.5550STS1000DV HDP n=1.479 0.3842 0.5321 0.4929 0.5071STS1000DV HDP n=1.514 0.2367 0.1039

Slurry Film


delfi.imec.be



ANOVA - All Models(11)

Anova: Single Factor α=0.05

SUMMARYGroups Count Sum Average VarianceK*P*v 16 13.57 0.85 0.03

K*(P*v)1/2 18 12.35 0.69 0.04K*P5/6*v1/2 18 14.69 0.82 0.06K*P2/3*v 16 12.36 0.77 0.02

ANOVASource of Variation SS df MS F P-value F crit

Between Groups 0.26 3 0.09 2.20 0.10 2.75Within Groups 2.50 64 0.04

Total 2.76 67

ANOVA - All Models(11) (HDP Excluded)

Anova: Single Factor α=0.05

SUMMARYGroups Count Sum Average VarianceK*P*v 14 12.70 0.91 0.01

K*(P*v)1/2 14 10.79 0.77 0.00K*P5/6*v1/2 14 12.99 0.93 0.00K*P2/3*v 14 11.29 0.81 0.02

ANOVASource of Variation SS df MS F P-value F crit

Between Groups 0.24 3 0.08 11.83 0.00 2.78Within Groups 0.36 52 0.01

Total 0.60 55


delfi.imec.be



Analysis of Pressure Threshold - Experimental(11)

0

1000

2000

3000

4000

5000

6000

0 5 10 15 20 25

PECVD TEOS SS12ThOx STS1000DV

Polis

h ra

te (Å

min

-1)

P2/3 * v (kPa2/3 m s-1)

3 psi30 rpm

1.5 psi90 rpm

10

100

1000

10000

0 5 10 15 20 25

PECVD TEOS SS12ThOx STS1000DV

Polis

h ra

te (Å

min

-1)

P2/3 * v (kPa2/3 m s-1)

3 psi30 rpm

1.5 psi90 rpm

Analysis of Pressure Threshold - Experimental(11)


delfi.imec.be



Conclusions From Oxide Models(11)

All models with predicting equationsare based on Hertzian indentationignore slurry and particle chemistry

Performance of each model studied R2 term was good for some but not all process conditionsR2 KPv and KP5/6v1/2 best when ceria and high [Si] HDP oxide removed from analysisPressure threshold model did not agree with experimental polish results

Simplicity of models studied indicates extensive work still needs to be done

Selected Bibliography for CMP Oxide Models

1. F. Preston, J. Soc. Glass Tech. 11, 1927, 214.2. N. J. Brown, P. C. Baker, and R. T. Maney, SPIE 306 Contemporary

Methods of Optical Fabrication, 1981, 42.3. L. M. Cook, J. Non-Cryst. Solids 120, 1990, 152.4. F. Zhang and A. Busnaina, Electrochem. Solid-state Lett. 1 (4), 1998, 184.5. D. Maugis and H. M. Pollock, Acta Matall. 32, 1984, 1323.6. S. R. Runnels and L. M. Eyman, J. Electrochem. Soc. 141 (6), 1994, 1698.7. W-T. Tseng, C-W Liu, B-T. Dai, and C-F Yeh, Thin Solid Films 290-291,

1996, 458.8. F. G. Shi and B. Zhao, Appl, Phys. A 67, 1998, 2499. B. Zhao and F. G. Shi, 1998 IEDM, 341.10. B. Zhao and F. G. Shi, 1999 CMP-MIC, 13.11. D. J. Stein, D. L. Hetherington, ECS Third International Symposium on

Chemical Polishing in IC Device Manufacturing, October 17-22, 1999, Honolulu, Hawaii.


delfi.imec.be



Appendix II: Polishing Motions

Kinematics of the relative velocity of a point on the wafer with respect to the polishing pad.

Rotary - Relative Velocity of Point on Wafer

The linear velocity of a arbitrary point Q on the wafer in the coordinate frame of the platen P is derived*:

*W. J. Patrick, et al., J. Electrochem. Soc., Vol. 138, No. 6, June (1991).

Platen/P

Wafer/H

•Q rq

rp

ωc ωp

rc

cpq rrrrrr

+=

)]([)()()( pccppccqpHQPCPQ xrrxrxrxVVV ωωωωω −−−=+−=+= ′

rrrrrrr

C’


delfi.imec.be



Linear - Relative Velocity of Point on Wafer

Wafer/H

•Qrc

ωc

VB

)( ccBHQBBQ rxVVVVrrrrr

ω+=+=

The linear velocity of a arbitrary point Q on the wafer in the coordinate frame of the platen B:

Platen/B

Orbital - Relative Velocity of Point on Wafer

The linear velocity of a arbitrary point Q on the wafer in the coordinate frame of the platen O:

)()(' ccooHQOCOQ rxrxVVVrrrrr

ωω −=+=

Wafer/H

• Q

ωc

rc

Orbital Motion

ro

ωo

C’


delfi.imec.be



About Sandia

Sandia National Laboratories is a multiprogram laboratory operated by Sandia Corporation, a Lockheed Martin Company, for the United States Department of Energy’s National Nuclear Security Administration under contract DE-AC04-94AL85000

Documents

MTC_CMP