models and simulations computer integrated manufacturing (cim) goal: monitor and control machines,...

15
Models and Simulations Computer Integrated Manufacturing (CIM) Goal: monitor and control machines, recipes to improve YIELDS ! !

Upload: merry-jennings

Post on 31-Dec-2015

214 views

Category:

Documents


0 download

TRANSCRIPT

Models and Simulations

Computer Integrated Manufacturing (CIM) Goal:monitor and control machines, recipes to improve YIELDS

!!

Modeling Particle Contamination and Yield

• Particles of feature size cause defects • ≈ 75% of yield loss in modern VLSI fabs is due to particle contamination.• Yield models depend on information about the distribution of particles. Yields

models use measured defect density N(dp) and size (dp)

• Particles on the order of 0.1 - 0.3 µm are the most troublesome: • larger particles precipitate easily • smaller ones coagulate into larger particles

• Yields are described by Poisson statistics in the simplest case.

Y = exp −ACDO

where AC is the critical area and DO the defect density.

• This model assumes independent randomly distributed defects and often underpredicts yields.

Even very small particles lead to failure (pinhole in the oxide )

(Empirical)

Contamination Reduction

Particles 10nm-10 µm

N(dp)=k(dp)-3

• Use of negative binomial statistics eliminates these assumptions and is more accurate.

where C is a measure of the particle spatial distribution (clustering factor).

Yields depend on defects (D0) density and chip size (Ac) (excluding areas of usual low yields (perimeter)

Spacial distribution of defects

Yields in ICs

C ∞

NB P

•Defects are random and independent•Predicts too low yields

NB - less random/independent (ex. Clustering)

Fraction of failed

ITRS needs

• Vertical lines are estimated chip sizes (from the ITRS).

• Note that defect densities will need to be extremely small in the future to achieve the high yields required for economic IC manufacturing.

Yields in ICs

NB Yield = f(DO)

MONITOR DEFECTS€

Y = (1+AC iDO i

Ci

)i=1

levels

∏ ≅ 1+ AC iDO ii=1

evels

∑ ⎛

⎝ ⎜

⎠ ⎟/Ci

⎧ ⎨ ⎩

⎫ ⎬ ⎭

−C i

≈40x40mm2

Overall yield depends on each step in the manufacturing cycle

C=2 in all plots

Defect density

Modeling Wafer Cleaning

• Cleaning involves removing particles, organics (photoresist) and metals from wafer surfaces.• Particles are largely removed by ultrasonic agitation during cleaning.• Organics like photoresists are removed in an O2 plasma or in H2SO4/H2O2 solutions.• The “RCA clean” is used to remove metals and any remaining organics.• Metal cleaning can be understood in terms of the following chemistry.

Si + 2H 2O↔ SiO2 + 4H + + 4e−

M↔ M z+ + ze−

(5)

(6)

• If we have a water solution with a Si wafer and metal atoms and ions, the stronger reaction will dominate.• Generally (6) is driven to the left and (5) to the right so that SiO2 is formed and M plates out on the wafer.• Good cleaning solutions drive (6) to the right since M+ is soluble and will be desorbed from the wafer surface.

• The strongest oxidants are at the bottom (H2O2 and O3). These reactions go to the left grabbing e- and forcing (6) to the right.

• Fundamentally the RCA clean works by using H2O2 as a strong oxidant.

Contamination Reduction: Wafer Cleaning

Contamination Reduction: Wafer Cleaning

No models exist but good understanding of cleaning steps

Remove metals by oxidizing (=removing e- -> ions) and by dissolving in cleaning solutionsSi+2H2O <->SiO2 +4H++4e-

M <-> MZ++ Ze-

Reduction<-> oxidation

Example:

In water: SiO2/Si Fe3+/Fe stronger potential Fe3+ -> Fe. Fe is plating on Si, Si is oxidizing •Add H2O2 • stronger potential -> Takes e- from M -> ions soluble in aqueous sol.

H2O2/H2O: will dominate Fe3+/Fe reaction Oxidizes Si

Larger (stronger) potential -> reaction to the left all others go to the rightUse OZONE

Take e- from M -> ions are produced

Atoms IonsStrongeroxidants

Dominantimpurities

Manufacturing Methods and Equipment

Wafer Cleaning

High-pH Oxidizes organics -> water soluble compounds and complexes IB IIB and other metals Au, Ag, Cu, Ni, Zn etc. Ex. CuCu(NH3)4

2+ in SC1

Low-pHInsoluble in NH4OH

Al3+, Fe3+

Modeling Gettering• Gettering consists of

1. Making metal atoms mobile.2. Migration of these atoms to trapping sites.

3. Trapping of atoms.• Step 1 generally happens by kicking out the substitutional atom into an interstitial site. One possible reaction is:

• Step 2 usually happens easily once the metal is interstitial since most metals diffuse rapidly in this form.

• Step 3 happens because heavy metals segregate preferentially to damaged regions or to N+ regions or pair with effective getters like P (AuP pairs). (See text.)

• In intrinsic gettering, the metal atoms segregate to dislocations around SiO2 precipitates.

Devices in nearsurface region

Denuded Zoneor Epi Layer

IntrinsicGettering

Region

BacksideGettering

Region

10 - 20 µm

500+ µm

PSG Layer

1

2

3

*3

Trapping

*Trapping

Aus → Aui

Diffusion

AuS + I↔ Au i

Gettering

All metal atoms mobile (DMi > DMs 10x)

(Fig. 4.8!) DM>> DDopantss

Sol. Sol MI>>MS (Cu, Ni)Sol.Sol. MI<<MS (Au, Pt)

Except of Ti, Mo, etc.

AuS+I Aui kick-out mechanism, then getterAus Aui+V dissociative or Frank-Turnbull mech.

I increase improves gettering of AuV increase hinders gettering

Ex. P diffusion, Ion Implant=damage, intrinsic gettering (=I )

“I” are closer to wafer surface

Metal Diffusion to the Gettering Sites

Long time

Au diffuses to the wafer back side and is trapped

t

Back side

Gettering of Au - the Role of the Back Side Injection of Si-I

Metals diffuse much faster than I (silicon) DSi-I >> DDopants

I are generated at the back and diffuse -> Aui form, diffuse and get trapped

At high T I -> gettering more effective, not limited by the backside injection

I

I @high P concentrations, Ion Impl., diffusion

Trapping the Metal Atoms at the Gettering SitesTrapped by: ion implantation, P diffusion, laser damage, poly-Si films, mechanical damage, etc.

But HOW?

*Physical damage -> metal trapped at defect sites; binding energy Eg depends on T; Fraction Bound=(1-K1exp-Eg/kT)

*Segregation, related to solubility in the silicon perfect crystal and in the gettering regionCAu,Si=NSiexp(-EA1/kT) CAu,G=NGexp(-EA2/kT) gettered -> k0=(CAu,G+CAu,Si)/CAu,Si=1+K2exp-[(EA1-EA2)/kT] --> k0=1+NG/5x1022exp(0.82eV/kT) gives fraction of Au

bound in gettered region

*Enhances sol.sol by high dopant concentrations: in “n” Au=acceptor, “p” - Au=donorAu+e- Au-, Keq=[Au-]/[Au][e-]=constant, [Au-

i]/[Au]ni= [Au-n]/[Au]n or [Au-

n]/[Au-i] = n/ni

Au acceptor in “n+” Si (100x if ni(1000°C)=7.14x1018 -> 1021cm-3 doping level)

*Ion pairing model -> AuP less strain

* Coulombic attraction Au+P -> Au-P+

* Interaction with point defects V- in “n+” Aui+V- Aus at the trapepd site

* Intrinsic gettering - trapping on dislocations and SF which surround precipitates. Dislocations have compressive and tensile stress - accommodate smaller and larger atoms

Present in Si

Increases with P concentration

Limits and Future Trends in Technology and Modeling; Environment

Eliminate defects from wafers: particles, contaminants, clean room -> local=SMIF (standard mechanical interfaceWafer cleaning in future ICs-> less chemicals (liquids, vapors), more diluted (disposal)

New cleaning: •Use ozone• Dry and vapor phase, (Vapors, Plasmas) environment! Cluster tools• Low Energy Physical Processes (sputtering)•Photochemically enhanced clean

Gettering -> intrinsic (less extrinsic), control Oi, Cs, use low T processing, use modeling tool -> point defects engineering, release, diffuse, entrap.

Watch for surface roughness

Summary of Key Ideas

• A three-tiered approach is used to minimize contamination in wafer processing.

• Particle control, wafer cleaning and gettering are some of the "nuts and bolts" of chip manufacturing.

• The economic success (i.e. chip yields) of companies manufacturing chips today depends on careful attention to these issues.

• Level 1 control - clean factories through air filtration and highly purified chemicals and gases.

• Level 2 control - wafer cleaning using basic chemistry to remove unwanted elements from wafer surfaces.

• Level 3 control - gettering to collect metal atoms in regions of the wafer far away from active devices.

• The bottom line is chip yield. Since "bad" die are manufactured alongside "good" die, increasing yield leads to better profitability in manufacturing chips.