Semiconductor Manufacturing: Etch Equipment Mathematical Modeling

Keywords: etch equipment, plasma etch equipment, icp etch, ccp etch, reactive ion etch equipment, plasma reactor, etch chamber, etch simulator

Semiconductor Manufacturing: Etch Equipment Mathematical Modeling

1. Introduction

Plasma etching is a critical process in semiconductor manufacturing where material is selectively removed from wafer surfaces using reactive plasmas. Mathematical modeling spans multiple scales and physics domains:

• Plasma physics — Generation and transport of reactive species
• Surface chemistry — Reaction kinetics at the wafer surface
• Transport phenomena — Gas flow, heat transfer, species diffusion
• Feature evolution — Nanoscale profile development
• Process control — Run-to-run optimization and fault detection

1.1 Etch Process Types

2. Plasma Discharge Modeling

2.1 Electron Kinetics

The electron energy distribution function (EEDF) governs ionization and dissociation rates. It is described by the Boltzmann transport equation:

Where:

• $f(\vec{r}, \vec{v}, t)$ — Electron distribution function
• $\vec{E}$ — Electric field vector
• $m_e$ — Electron mass
• $C[f]$ — Collision integral

Two-Term Approximation

For weakly anisotropic distributions:

2.2 Species Continuity Equations

For each species $i$ (electrons, ions, neutrals, radicals):

Where:

• $n_i$ — Number density of species $i$ (m⁻³)
• $\vec{\Gamma}_i$ — Flux vector (m⁻² s⁻¹)
• $S_i$ — Source/sink term from reactions (m⁻³ s⁻¹)

Flux Expressions

• Neutral species (diffusion only):

• Charged species (drift-diffusion):

Where:

• $D$ — Diffusion coefficient (m² s⁻¹)
• $\mu$ — Mobility (m² V⁻¹ s⁻¹)

Einstein Relation

2.3 Reaction Rate Coefficients

Rate coefficients are computed by integrating cross-sections over the EEDF:

Where:

• $\sigma(\varepsilon)$ — Energy-dependent cross-section (m²)
• $v(\varepsilon) = \sqrt{2\varepsilon/m_e}$ — Electron velocity
• $f(\varepsilon)$ — Normalized EEDF

Key Reactions in Fluorine-Based Plasmas

2.4 Electron Energy Balance

Where:

• $P_{abs}$ — Power absorbed from RF field (W m⁻³)
• $P_{loss}$ — Power lost to collisions (W m⁻³)

2.5 Electromagnetic Field Equations

Capacitively Coupled Plasma (CCP)

Poisson's equation:

Inductively Coupled Plasma (ICP)

Wave equation for the azimuthal electric field:

With plasma conductivity:

3. Sheath Physics

The plasma sheath is a thin, ion-rich region at the wafer surface that accelerates ions for bombardment.

3.1 Bohm Criterion

Ions must reach the sheath edge with minimum velocity:

Where:

• $k_B$ — Boltzmann constant (1.38 × 10⁻²³ J K⁻¹)
• $T_e$ — Electron temperature (K or eV)
• $M_i$ — Ion mass (kg)

3.2 Child-Langmuir Law

Maximum ion current density through a collisionless sheath:

Where:

• $V_s$ — Sheath voltage (V)
• $d$ — Sheath thickness (m)
• $\varepsilon_0$ — Permittivity of free space (8.85 × 10⁻¹² F m⁻¹)

3.3 Sheath Thickness

Approximate expression:

Where Debye length:

3.4 Ion Energy Distribution Function (IEDF)

The IEDF depends critically on the ratio:

Where:

• $\xi \gg 1$ (high frequency): Ions see time-averaged sheath voltage → narrow IEDF
• $\xi \ll 1$ (low frequency): Ions respond to instantaneous voltage → bimodal IEDF

Bimodal IEDF Expression

For RF sheaths:

With:

• $E_{max} = e(V_{dc} + V_{rf})$
• $E_{min} = e(V_{dc} - V_{rf})$

3.5 Collisional Sheath Effects

When $d > \lambda_{mfp}$ (ion mean free path), ion-neutral collisions broaden the IEDF:

4. Surface Reaction Kinetics

4.1 General Etch Rate Model

Where:

• $ER$ — Etch rate (nm min⁻¹ or Å s⁻¹)
• $Y_{phys}$ — Physical sputtering yield (atoms/ion)
• $Y_{chem}$ — Chemical etch yield coefficient
• $\Gamma$ — Flux (m⁻² s⁻¹)
• $\theta_{ads}$ — Surface coverage fraction (0–1)
• $f(E_{ion})$ — Ion enhancement function

4.2 Physical Sputtering Yield

Sigmund Theory

For normal incidence:

Where:

• $U_s$ — Surface binding energy (eV)
• $M_1$, $M_2$ — Ion and target atom masses
• $\alpha$ — Dimensionless parameter (~0.2–0.4)

Threshold Energy

Sputtering occurs only above threshold:

4.3 Angular Dependence of Sputtering

Yamamura formula:

Where:

• $\theta$ — Ion incidence angle from surface normal
• $f$ — Fitting parameter (~1.5–2.5)
• $b$ — Fitting parameter (~0.1–0.5)

Physical interpretation:

• $\cos^{-f}(\theta)$ term: Enhanced yield at grazing angles (energy deposited closer to surface)
• $\exp[-b(\cdot)]$ term: Suppression at very grazing angles (reflection)

4.4 Surface Coverage Dynamics

Langmuir adsorption kinetics:

Where:

• $s$ — Sticking coefficient (0–1)
• $k_d =

u_0 \exp(-E_d/k_B T)$ — Desorption rate

• $

u_0$ — Attempt frequency (~10¹³ s⁻¹)

• $E_d$ — Desorption activation energy (eV)

Steady-State Coverage

4.5 Ion-Enhanced Etching Mechanisms

Damage Model

Ion bombardment creates reactive sites:

Chemically Enhanced Physical Sputtering

Product species have lower binding energy:

Where typically $Y_{product} > Y_{substrate}$.

4.6 Silicon Etching in Fluorine Plasmas

Simplified mechanism:

1. Adsorption: $F_{(g)} + Si^* \rightarrow SiF_{(ads)}$ 2. Fluorination: $SiF_{(ads)} + F \rightarrow SiF_2 \rightarrow SiF_3 \rightarrow SiF_4$ 3. Desorption: $SiF_4 \xrightarrow{ion} SiF_4 (g)\uparrow$

Etch rate expression:

Where:

• $N_0$ — Avogadro's number
• $\rho_{Si}$ — Silicon atomic density (5 × 10²² cm⁻³)

4.7 Oxide Etching in Fluorocarbon Plasmas

More complex due to polymer competition:

Where:

• $t_{poly}$ — Polymer thickness
• Balance between etching and deposition determines regime

Regime boundaries:

• High F/C ratio → Etching dominant
• Low F/C ratio → Deposition dominant (polymerization)

5. Feature-Scale Modeling

5.1 Level Set Method

The surface is represented implicitly as the zero level set of $\phi(\vec{x}, t)$:

Evolution Equation

Where $V_n$ is the velocity in the normal direction:

Advantages

• Handles topological changes naturally (merging, splitting)
• No explicit surface tracking required
• Curvature easily computed: $\kappa =

abla \cdot \vec{n}$

5.2 Flux Calculation at Surface Points

Local etch velocity depends on incident fluxes:

Where $\Omega$ is the atomic volume.

5.3 Knudsen Transport in High Aspect Ratio Features

At low pressure, neutral mean free path > feature dimensions → free molecular flow.

View Factor Method

Flux at surface point P:

Where:

• $\Gamma_0$ — Flux from plasma (at feature opening)
• $\Omega(P)$ — Solid angle subtended by opening at P
• $K(P', P)$ — Kernel for re-emission from P' to P

Cosine Re-emission Law

For diffuse reflection:

Where:

• $\theta'$, $\theta$ — Angles from surface normals
• $r$ — Distance between points
• $s$ — Sticking coefficient

5.4 Clausing Factor for Tubes

Transmission probability through a cylindrical hole:

Where $L$ = length, $r$ = radius.

For aspect ratio $AR = L/(2r)$:

5.5 Aspect Ratio Dependent Etching (ARDE)

Empirical model:

Where:

• $ER_0$ — Etch rate at open area
• $\beta$, $n$ — Fitting parameters (typically $n \approx 1$–2)

Physical causes:

• Neutral transport limitation (Knudsen diffusion)
• Ion angular distribution effects
• Charging effects in dielectric etching

5.6 Ion Angular Distribution Effects

Ions have finite angular spread due to:

• Thermal velocity at sheath edge
• Collisions in sheath
• Non-vertical electric fields

Distribution often modeled as:

Typical $\sigma_\theta \approx 2°$–5°

5.7 Monte Carlo Feature-Scale Methods

Algorithm:

1. Launch particle from plasma with appropriate energy/angle distribution 2. Track trajectory to surface 3. Evaluate reaction probability based on local conditions 4. If reaction occurs, remove material; else reflect particle 5. Repeat for statistical convergence 6. Advance surface based on accumulated removal

Advantages:

• Naturally handles stochastic effects
• Easy to incorporate complex physics
• Parallelizable

6. Equipment-Scale Transport

6.1 Gas Flow Regimes

Characterized by Knudsen number:

Where $\lambda$ is mean free path, $L$ is characteristic length.

6.2 Navier-Stokes Equations

Continuity:

Momentum:

Energy:

Where $\Phi$ is viscous dissipation.

6.3 Slip Boundary Conditions

For Knudsen numbers 0.01–0.1:

Where $\sigma_v$, $\sigma_T$ are accommodation coefficients.

6.4 Wafer Temperature Model

Energy balance at wafer surface:

Components:

• Ion bombardment: $Q_{ion} = \Gamma_{ion} \cdot E_{ion}$

• Chemical reactions: $Q_{chem} = \Gamma_{etch} \cdot \Delta H_{rxn}$

• Radiation: $Q_{rad} = \varepsilon \sigma (T_w^4 - T_{wall}^4)$

• Conduction to chuck: $Q_{cond} = h_c (T_w - T_{chuck})$

The contact conductance $h_c$ depends on:

• Backside gas pressure
• Surface roughness
• Clamping force

6.5 Uniformity Modeling

Radial etch rate profile:

Where $R_w$ is wafer radius.

Uniformity metric:

6.6 Loading Effect

Etch rate depends on exposed area:

Or in terms of pattern density $\rho_p$:

7. Multiscale Coupling

7.1 Scale Hierarchy

7.2 Coupling Strategies

Hierarchical Approach

1. Solve equipment-scale flow → boundary conditions for plasma 2. Solve plasma model → fluxes to sheath 3. Solve sheath model → IEDF to surface 4. Solve feature-scale model → local etch rates

Embedded Multiscale

Feature-scale model embedded in equipment simulation:

• Sample representative features across wafer
• Compute local etch rates from local plasma conditions
• Interpolate for full wafer prediction

7.3 Reduced-Order Models

Plasma model simplification:

Where P is pressure, W is power.

Response surfaces:

8. Process Control Mathematics

8.1 Run-to-Run (R2R) Control

EWMA Controller

Where:

• $u_k$ — Recipe parameter at run $k$
• $y_k$ — Measured output at run $k$
• $K$ — Controller gain

Double EWMA (for drift)

8.2 Model Predictive Control (MPC)

Optimize over horizon N:

Subject to:

• Process model: $y_{k+1} = f(y_k, u_k)$
• Input constraints: $u_{min} \leq u \leq u_{max}$
• Output constraints: $y_{min} \leq y \leq y_{max}$
• Rate constraints: $|\Delta u| \leq \Delta u_{max}$

8.3 Virtual Metrology

Predict wafer-level results from equipment data:

Where $\vec{x}_{sensor}$ includes:

• Optical emission spectroscopy (OES) signals
• RF impedance (voltage, current, phase)
• Pressure, flow rates
• Chamber wall temperature
• Endpoint detection signals

PLS (Partial Least Squares) Model

Neural Network Model

8.4 Fault Detection and Classification (FDC)

Hotelling's T² Statistic

Alarm if $T^2 > T^2_{critical}(\alpha, p, n)$

Q-Statistic (SPE)

Where $\hat{\vec{x}}$ is PCA reconstruction.

8.5 Endpoint Detection

OES Endpoint

Monitor emission intensity ratio:

Endpoint when:

9. Emerging Frontiers

9.1 Atomic Layer Etching (ALE)

Self-limiting process:

1. Modification step: Surface layer modified (e.g., chlorination) 2. Removal step: Modified layer removed by low-energy ions

Where:

• $EPC$ — Etch per cycle (typically 0.5–2 Å)
• $\Gamma_{sat}$ — Saturation coverage
• $\delta_{modified}$ — Modified layer thickness

Synergy Parameter

High synergy indicates good self-limiting behavior.

9.2 Machine Learning Integration

Physics-Informed Neural Networks (PINNs)

Loss function includes physics constraints:

Where:

Gaussian Process Regression

For process optimization with uncertainty quantification:

Posterior mean:

9.3 Stochastic Effects at Nanoscale

Line Edge Roughness (LER)

At sub-10 nm features, discrete nature of reactions matters:

Where $a$ is atomic spacing, $L$ is line length.

Kinetic Monte Carlo (KMC)

Event selection probability:

Time advance:

Where $u \in (0,1)$ is uniform random.

Physical Constants

Typical Process Parameters

Source: ChipFoundryServices — Search this topic — Ask CFSGPT