Endpoint Detection Simulator

Real-time simulation of optical emission spectroscopy (OES), laser interferometry, and mass spectrometry endpoint detection for plasma etching. Model time-varying emission intensities, interference patterns, and reaction product concentrations as etch progresses through material layers.

Simulation Status: AUTO-RUNNING

Watch in real-time as the simulation etches through material layers and detects the endpoint using multiple sensing modalities. OES tracks plasma emission lines, interferometry measures film thickness, and mass spec monitors volatile products.

ENDPOINT DETECTED! Underlying layer exposed.

Simulation Controls

Process Parameters

Remaining Film Thickness

500
nm

Etch Time Elapsed

0.0
seconds

Interferometer Signal

0.50
normalized

OES Si Intensity

250
counts

Mass Spec SiF₄

4.2e15
mol/s

Endpoint Confidence

0
%

3D Wafer Etch Depth Profile

Endpoint Detection Physics & Equations

Optical Emission Spectroscopy (OES)

Plasma emission intensity from excited species follows excitation-de-excitation balance:

I(λ) = (h·c/λ) · n* · A where: I(λ) = emission intensity at wavelength λ n* = excited state density = (k_exc · n_e · n_gas) / (A + k_quench · n_gas) A = Einstein coefficient (spontaneous emission rate) k_exc = electron impact excitation rate k_quench = quenching rate coefficient n_e = electron density n_gas = ground state gas density For Si atoms etched from wafer: Si: 288 nm (3P → 1D transition), A = 2.17×10⁷ s⁻¹ Si emission ∝ Si etch product concentration in plasma

Emission Line Ratio Method

Ratio of product emission to reactant emission indicates material transition:

R(t) = I_product(t) / I_reactant(t) For SiO₂ → Si transition: Product: Si emission (288 nm) - increases at endpoint Reactant: F emission (704 nm) - decreases at endpoint Endpoint criterion: dR/dt > threshold AND R > R_critical R_critical typically 0.5-2.0 depending on process

Laser Interferometry

Interference from reflections at film surfaces creates periodic intensity pattern:

I_detector = I₀ · [1 + V · cos(Φ)] where phase difference: Φ = (4π · n · d) / λ + Φ₀ n = refractive index of film d = film thickness λ = laser wavelength V = fringe visibility (0-1) Φ₀ = phase offset from substrate reflectivity difference As film etches: Φ(t) = (4π·n/λ) · [d₀ - R_etch·t] One complete fringe cycle = λ/(2n) thickness change For SiO₂ (n=1.46) at 633nm: one cycle = 217 nm

Etch Rate from Interferometry

Real-time etch rate determined from fringe frequency:

R_etch = (λ / 2n) · f_fringe where: f_fringe = frequency of intensity oscillations (Hz) Example: f = 0.1 Hz, λ = 633 nm, n = 1.46 → R_etch = 21.7 nm/s = 1300 nm/min Endpoint detected when fringes cease (film cleared)

Mass Spectrometry

Volatile etch products detected by quadrupole mass analyzer:

Signal(m/z) ∝ partial pressure ∝ etch rate × sticking coefficient For SiO₂ etching in CF₄/O₂: Primary products: SiF₄ (m/z = 104), CO₂ (m/z = 44), CO (m/z = 28) Si etching in CF₄: Primary products: SiF₄ (m/z = 104), SiF₃ (m/z = 85) Signal intensity: I(m/z) = σ_ion · n(m/z) · I_electron σ_ion = ionization cross-section n(m/z) = species partial density I_electron = electron beam current (70 eV standard)

Product Formation Rate

Etch product generation linked to material removal:

Γ_product = (R_etch · A_wafer · ρ_material) / M_product where: Γ_product = molecular flow rate (molecules/s) R_etch = etch rate (cm/s) A_wafer = wafer area (cm²) ρ_material = material density (g/cm³) M_product = molecular weight (g/mol) For 300mm wafer, SiO₂ at 100 nm/min: Γ_SiF₄ = 2.4×10¹⁷ molecules/s = 4×10⁻⁷ mol/s

Signal Derivative Detection

First derivative of OES or mass spec signal maximizes at endpoint:

S'(t) = dS/dt = [S(t+Δt) - S(t-Δt)] / (2Δt) Endpoint criterion: |S'(t)| > threshold AND S''(t) changes sign Second derivative for noise rejection: S''(t) = [S(t+Δt) - 2S(t) + S(t-Δt)] / Δt² Moving average filter applied for SNR > 10

Multi-Modal Fusion

Combine multiple detection methods for robust endpoint determination:

Confidence(t) = w₁·C_OES(t) + w₂·C_interf(t) + w₃·C_MS(t) where weights sum to 1: w₁ + w₂ + w₃ = 1 Individual confidences based on: C_OES = sigmoid[(R(t) - R_threshold) / σ_R] C_interf = 1 if fringes stopped, 0 otherwise C_MS = sigmoid[(I_product(t) - I_threshold) / σ_I] Typical weights: w₁=0.5 (OES), w₂=0.3 (interf), w₃=0.2 (MS) Endpoint declared when Confidence > 0.8