RIE LAG & ARDE Physics
Knudsen Number and Transport Regimes
The Knudsen number determines transport mechanisms:
Knudsen number:
Kn = λ_mfp / D_trench
λ_mfp: Mean free path ≈ 5×10⁻³ cm / P(mTorr) at 300K
D_trench: Trench diameter/width
Transport regimes:
Kn < 0.01: Continuum (viscous flow)
- Collisions dominate
- Bulk gas transport
- Pressure-driven flow
0.01 < Kn < 10: Transition
- Mixed regime
- Both mechanisms
Kn > 10: Molecular (Knudsen diffusion)
- Collision-free
- Ballistic transport
- Geometric shadowing dominates
Gotoh ARDE Model
Empirical model for aspect ratio dependent etching:
R(AR) = R₀ / (1 + K × AR^n)
Where:
R(AR): Etch rate at aspect ratio AR
R₀: Etch rate for shallow/open areas
AR: Aspect ratio = depth / width
K: ARDE coefficient (0.1-0.3 typical)
n: Exponent (1-2, typically 1.5)
Physical interpretation:
K and n depend on:
- Pressure (lower P → higher K)
- Sticking coefficient
- Neutral/ion ratio
- Reaction probability
At high AR:
R(AR) → R₀/(K × AR^n) ~ 1/AR^n
Etch slows dramatically
Neutral Transport Limitation
Radical flux decreases with depth:
Concentration at depth z:
C(z) = C₀ × exp(-z / λ_diff)
λ_diff: Diffusion length
λ_diff = D_trench × √(D / k_rxn)
D: Diffusion coefficient
k_rxn: Surface reaction rate
For molecular regime (high Kn):
Flux ~ solid angle = Ω / 4π
Ω decreases with depth
Angular distribution:
F(θ) = F₀ × cos(θ)
Trenches see reduced effective flux
Neutral shadowing:
Fraction reaching bottom ~ (D/depth)²
Strong AR dependence
Ion Shadowing and Beam Divergence
Ion angular distribution limits deep etch:
Ion angular distribution:
f(θ) ~ exp(-θ² / 2σ²)
σ: Angular spread ≈ 2-5° typical
Maximum etchable AR:
AR_max ≈ 1 / (2 × tan(σ))
For σ = 3°: AR_max ≈ 10
Beyond AR_max:
Ions hit sidewalls, not bottom
Sidewall damage, no bottom etch
Sheath collisions increase σ:
High pressure → more collisions → larger σ
Lower AR_max at high P
Microloading Effect
Local pattern density affects etch rate:
Microloading:
R_local = R_global × (1 - β × ρ_local)
ρ_local: Local open area fraction (0-1)
β: Microloading coefficient (0.2-0.5)
Physical mechanism:
Dense patterns → more trenches competing
Local reactant depletion
Byproduct accumulation
Scale lengths:
Local: < 100 μm (diffusion)
Global: Wafer-scale (mm-cm)
Mitigation:
- Higher pressure (shorter λ_mfp)
- Dummy fills
- Local gas injection
- Temperature control
Reaction-Diffusion Competition
Balance between transport and reaction:
Damköhler number:
Da = k_rxn × L / D
Da << 1: Reaction-limited
- Slow surface reaction
- Good transport
- Uniform profiles
Da >> 1: Diffusion-limited
- Fast reaction
- Poor transport into trenches
- Strong ARDE
Optimization:
Want Da ≈ 1 for balance
Adjust T, P, chemistry
Trade rate vs uniformity
Pressure Dependence
Pressure dramatically affects LAG:
Low pressure (2-5 mTorr):
- Long λ_mfp (> 1 cm)
- Kn > 10 (molecular)
- Severe neutral shadowing
- Strong ARDE (K large)
- Better anisotropy
High pressure (50-100 mTorr):
- Short λ_mfp (< 1 mm)
- Kn < 0.1 (continuum)
- Better neutral transport
- Weaker ARDE (K small)
- More isotropic
Optimal pressure:
Balance ARDE vs anisotropy
Typically 10-30 mTorr
Depends on AR requirements
Sticking Coefficient Effects
Surface reaction probability affects depth penetration:
Sticking coefficient s:
Probability that incident species reacts
Low s (0.01-0.1):
- Multiple bounces possible
- Better depth penetration
- Reduced ARDE
- Example: F radicals on Si
High s (0.5-1.0):
- First surface reaction
- Poor depth penetration
- Strong ARDE
- Example: Cl on Al
Effective sticking in trenches:
s_eff increases with AR
Multiple wall collisions
Each has probability s
Mitigation Strategies
Techniques to reduce RIE LAG:
1. Higher pressure:
- Reduce Kn → continuum regime
- Better neutral transport
- Trade-off: Less anisotropic
2. Pulsed plasma:
- ON: Generate radicals
- OFF: Allow diffusion
- Separates generation and transport
3. Neutral-rich chemistry:
- Higher F/ion ratio
- Chemical component stronger
- Less AR dependence
4. Multi-step process:
- Fast initial etch
- Slow deep etch (higher P)
- Adaptive conditions
5. Overetch compensation:
- Small features: Shorter time
- Large features: Longer time
- CD-based recipes