LJB - MPI 2005

Lubrication Layer Perturbations inChemical-Mechanical Polishing

Dr. Len BoruckiCTO, Araca, Inc.

LJB - MPI 2005

Outline

• A quick tutorial on chemical-mechanical polishing (CMP).

• Elastohydrodynamic lubrication with a pure lubricant.

• Questions posed by the presence of slurry particles.

LJB - MPI 2005

Quick CMP Tutorial

LJB - MPI 2005

Wafer Planarization in Integrated Circuit Fabrication•Integrated circuits are made by deposition and modification of numerous material layers.•Photolithography, the main method of creating patterns, works best on flat surfaces.•Chemical-Mechanical Polishing (CMP) is currently the leading method for planarizing surfaces.

Schematic of an integrated circuitproduced without CMP

Schematic of an integrated circuitproduced with CMP

Peter Wolters

LJB - MPI 2005

Generic Rotary Polishing Tool

Platen

Polishing Pad

SlurryStream,Puddle

Wafer

Wafer CarrierConditioner

AMAT ReflexionCommercial tool

Bench top polisherStruers. Photo by Rob HightGeorgia Tech.

A rotating tool with adiamond-covered facemaintains pad surfaceroughness, counteractingabrasive wear, removingdebris, exposing newpad surface.

The wafer is heldupside down by arotating carrier.The wafer surfacereacts with chemicalsin the slurry and isabraded by slurryparticles. The padalso experiencesabrasive wear.

LJB - MPI 2005

Polishing PadsThe most commonlyused pad, IC-1000TM,(Rohm and Haas) isa closed cellpolyurethane foamwith a mean voiddiameter of about30 microns. Voidsoccupy about 35%of the volume andare not interconnectedexcept at the surface.

The pad is shownhere next to a scaleddrawing of a 100 µmwide, 2 µm deeptrench. The padsurface roughnessis large comparedwith typical waferfeatures.

Wafer Topography

Letitia Molina, Motorola

LJB - MPI 2005

C.Elmufdi et al., 2004 CAMP Symposium.

Polishing pads are soft compared with most of the materials they polish.

Copper: 110 GPa Silicon Dioxide: 43-77 GPa

The pad elastic modulus generally decreases with increasing temperatureand water content. A wide range is possible: ~100-550 MPa.

LJB - MPI 2005

Polishing Pad Surfaces

Kojima andNishiguchi,2003 CMP-MIC.

The pad surface is not static, but evolves under conditioning and abrasive wear.

Freshly conditioned After polishing without conditioning

Scott Lawing,Rohm and HaasECS 2003

This shows thatthere is pad/wafercontact. The wearrate varies withthe abrasive type.Spherical particlesproduce less wear,irregular particlesmore wear.

LJB - MPI 2005

Slurry ParticlesSlurry particles are very much smaller than pad asperities. Typical mean diameters forspherical colloidal particles range from a few tens to a few hundreds of nanometers.Solid loadings vary from ~0.3% to ~30% by weight. Slurry viscosity is similar to water.

S. Lawing, ECS 2003

We’ll assume spherical colloidal particles

LJB - MPI 2005

Polishing Mechanisms (SiO2)

Hypothesizedmechanism forSiO2 removalby ceria. Mostremoved silica isfound on the slurryparticles andin colloidalsuspension.(W. America,CAMP 2003).

Particle surfacechemistry matters

Polishing pad

Wafer (held upside down by carrier)Chemically-softened layer with surface charge.Reactive fluid and charged slurry particles(Particle size greatly exaggerated)

Pressure p

Relative sliding velocity V

Chemicalreaction/softeningand mechanicalremoval.

- - - - - - - - - - - - - - - - - - - -+

+

+ + +

++ + +

++

+ + ++

++

LJB - MPI 2005

Pad

Wafer

Pad asperitiesand fluidthickness –tens of microns

...

.. .

.... . ..

Slurry particlestens-hundredsof nm.

Nominal pressure p=7 - 49 kPa (1-7 PSI)Relative sliding speed ~1 m/sec

Summary of Scales and Numerical Values

≈

1 GPa

E 100-500 MPa

≈E

LJB - MPI 2005

Elastohydrodynamic Lubrication (EHL)

LJB - MPI 2005

Dry Contact

Hertzian Theory

o Undeformed asperity tips arespherical.

o Contact area is circular.

o Pressure is elliptic.2/12

0 ))/(1( arpp −=

δππ RaA == 2

++

K.L. Johnson, Contact Mechanics

δ

R

2a

2/12/10 )1(

2 δπν REp

−=

Asperity tip

LJB - MPI 2005

WaferEHL Theory

o A thin lubrication layer forms.

o Hydrodynamic pressures deformasperity tips.

o Positive hydrodynamic pressures inthe lubrication layer support the load.

Asperity

V

Lubricated Contact

LubricationLayer

LJB - MPI 2005

EHL TheoriesElastohydrodynamic lubrication has been studied for almost a century. Some early citationsfrom Szeri, Fluid Film Lubrication, Theory and Design, Cambridge (1998) Ch. 8:

1916 H.M. Martin Assumed rigid bodies. Predicted thinner lubrication layer than observed.1936 W. Peppler Allowed contacts to deform elastically.1945 Gatcombe Generalized to pressure-dependent viscosity, µ=µ(p)1949 A.N. Grubin First satisfactory solution accounting for elastic deformation and µ=µ(p).

Reynolds equationV

Elasticity theory

http://www.lsbu.ac.uk/water/explan2.html

LJB - MPI 2005

V

Fluid Pressure

Asperity deformation

Hertz theory

From Szeri, Fluid Film Lubrication, Cambridge University Press, 1998

Fluid pressures are approximatelythe same as Hertzian pressuresexcept for a pressure spike near thetrailing edge. Disagreement becomesmore pronounced at lower loads.

The lubrication layer is nearlyuniform in thickness except fora constriction at the trailing edgethat produces the pressurespike.

EHL Theory Example (Roller)

V

LJB - MPI 2005

EHL Compact FormulasSimple formulas are often available for estimating the average lubricationlayer thickness. Some fitting to more complex solutions is involved.

Pad Asperity

hmin hc

22.042.08.064.0)(5.1 −−= avgc pERVh µ

cavg Npp /=

∫∞

=d ssc dzzN )(φη

ηs = summit area densityφs = summit height PDF.

ηs =2x108/m2

φs Gaussian with σ=6µmR=50µmp=1 kPa

V=1 m/secµ=2.5x10-3 Pa-sec

hc=29 nmabout the same as the mean diameterof some types of slurry particles.

LJB - MPI 2005

Slurry Particles

LJB - MPI 2005

Active ParticlesPolish rates are generally low for slurries that do not contain particles. When particles areadded, some are evidently trapped between the wafer and pad asperities and increase theremoval rate by mechanical or chemical means. These are active particles. Experimentalestimates of slurry residence time and utilization suggest that most particles never becomeactive.

Active

Inactive

Some active particlesalso abrade the pad.

LJB - MPI 2005

0

500

1000

1500

2000

2500

0 5 10 15 20 25 30 35

Concentration (wt.%)

MR

R (Ă

/min

)50nm

80nm

140nm

Removal Rate and Solids LoadingRemoval rate increases with weight fraction for particles of a given size.

C. Zhou et al., STLE 2002

LJB - MPI 2005

Removal Rate and Particle Size

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

0 50 100 150

Abrasive Size (nm)

MR

R (Ă

/min

)

18 kPa, 0.4 m/s

32 kPa, 0.6 m/s

45 kPa, 0.9 m/s

C. Zhou et al., STLE 2002

At a fixed weight fraction, the removal rate has a peak at a size comparable toa possible lubrication layer thickness.

30 wt% In this figure,the number ofabrasive particlesdecreases asthe size increases.

LJB - MPI 2005

Questions

Slurry particle diameter

Smallparticle

Which portion of the particledistribution becomes active?

If we start out with no particlesin the lubrication layer, then howfast do they accumulate?

How much does the accumulation ofa few particles affect the probabilityof capture of additional particles?

MediumLargeactive?

activenotactive

LJB - MPI 2005

Questions

What happens to the thicknessand shape of the lubricationlayer as the solids loadingincreases from zero to typicalupper weight fraction limits?

Are hydrodynamic pressuresthe main determinant of thelayer geometry or is therea point at which particle sizeand loading are the main factors?

Can the compact models oflubrication layer thicknessbe generalized to includeslurry particles?

No solids

High solids loading

?

These are the main questions that I would like to address.

LJB - MPI 2005

I have with me …1. A 2D finite element Reynolds

equation solver.2. A 2D/3D linear elasticity solver.3. Some literature.

Fluid pressures fromthe Reynolds solver