← Back to AI Factory Chat

AI Factory Glossary

44 technical terms and definitions

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z All
Showing page 1 of 1 (44 entries)

raman mapping, metrology

Spatial stress or composition mapping.

raman spectroscopy,metrology

Analyze molecular vibrations and stress.

ramp rate, packaging

Heating/cooling speed.

random defects,metrology

Unpredictable particle-caused defects.

random signature, metrology

No clear pattern in failures.

reactive ion etching (sample prep),reactive ion etching,sample prep,metrology

Etch samples for analysis.

recombination parameter extraction, metrology

Determine SRH parameters.

redistribution layer (rdl),redistribution layer,rdl,advanced packaging

Reroute connections from die pads to larger pitch for packaging.

redistribution layer for tsv, rdl, advanced packaging

Routing layer connecting TSVs.

reel diameter, packaging

Size of component reel.

reference material,metrology

Standard sample with certified properties for tool calibration.

reference standard,metrology

Certified artifact for calibration.

reflection high-energy electron diffraction (rheed),reflection high-energy electron diffraction,rheed,metrology

In-situ surface crystallography.

reflection interferometry,metrology

Monitor etch depth using interference.

reflective optics (euv),reflective optics,euv,lithography

Mirrors instead of lenses for EUV light.

reflectometry,metrology

Measure film thickness from interference of reflected light.

reflow profile, packaging

Temperature vs time during reflow.

reflow soldering for smt, packaging

Solder paste melted to attach.

regression analysis,regression,ols,least squares,pls,partial least squares,ridge,lasso,semiconductor regression,process regression

# Regression Analysis Semiconductor fabrication involves hundreds of sequential process steps, each governed by dozens of parameters. Regression analysis serves critical functions: - Process Modeling: Understanding relationships between inputs and quality outputs - Virtual Metrology: Predicting measurements from real-time sensor data - Run-to-Run Control: Adaptive process adjustment - Yield Optimization: Maximizing device performance and throughput - Fault Detection: Identifying and diagnosing process excursions Core Mathematical Framework Ordinary Least Squares (OLS) The foundational linear regression model: $$ \mathbf{y} = \mathbf{X}\boldsymbol{\beta} + \boldsymbol{\varepsilon} $$ Variable Definitions: - $\mathbf{y}$ — $n \times 1$ response vector (e.g., film thickness, etch rate, yield) - $\mathbf{X}$ — $n \times (k+1)$ design matrix of process parameters - $\boldsymbol{\beta}$ — $(k+1) \times 1$ coefficient vector - $\boldsymbol{\varepsilon} \sim N(\mathbf{0}, \sigma^2\mathbf{I})$ — error term OLS Estimator: $$ \hat{\boldsymbol{\beta}} = (\mathbf{X}^\top\mathbf{X})^{-1}\mathbf{X}^\top\mathbf{y} $$ Variance-Covariance Matrix of Estimator: $$ \text{Var}(\hat{\boldsymbol{\beta}}) = \sigma^2(\mathbf{X}^\top\mathbf{X})^{-1} $$ Unbiased Variance Estimate: $$ \hat{\sigma}^2 = \frac{\mathbf{e}^\top\mathbf{e}}{n - k - 1} = \frac{\sum_{i=1}^{n}(y_i - \hat{y}_i)^2}{n - k - 1} $$ Response Surface Methodology (RSM) Critical for semiconductor process optimization, RSM uses second-order polynomial models. Second-Order Model $$ y = \beta_0 + \sum_{i=1}^{k}\beta_i x_i + \sum_{i=1}^{k}\beta_{ii}x_i^2 + \sum_{i n$) - Addresses multicollinearity - Captures latent variable structures - Simultaneously models X and Y relationships NIPALS Algorithm 1. Initialize: $\mathbf{u} = \mathbf{y}$ 2. X-weight: $$\mathbf{w} = \frac{\mathbf{X}^\top\mathbf{u}}{\|\mathbf{X}^\top\mathbf{u}\|}$$ 3. X-score: $$\mathbf{t} = \mathbf{X}\mathbf{w}$$ 4. Y-loading: $$q = \frac{\mathbf{y}^\top\mathbf{t}}{\mathbf{t}^\top\mathbf{t}}$$ 5. Y-score update: $$\mathbf{u} = \frac{\mathbf{y}q}{q^2}$$ 6. Iterate until convergence 7. Deflate X and Y, extract next component Model Structure $$ \mathbf{X} = \mathbf{T}\mathbf{P}^\top + \mathbf{E} $$ $$ \mathbf{Y} = \mathbf{T}\mathbf{Q}^\top + \mathbf{F} $$ Where: - $\mathbf{T}$ — score matrix (latent variables) - $\mathbf{P}$ — X-loadings - $\mathbf{Q}$ — Y-loadings - $\mathbf{E}, \mathbf{F}$ — residuals Spatial Regression for Wafer Maps Wafer-level variation exhibits spatial patterns requiring specialized models. Zernike Polynomial Decomposition General Form: $$ Z(r,\theta) = \sum_{n,m} a_{nm} Z_n^m(r,\theta) $$ Standard Zernike Polynomials (first few terms): | Index | Name | Formula | |-------|------|---------| | $Z_0^0$ | Piston | $1$ | | $Z_1^{-1}$ | Tilt Y | $r\sin\theta$ | | $Z_1^{1}$ | Tilt X | $r\cos\theta$ | | $Z_2^{-2}$ | Astigmatism 45° | $r^2\sin 2\theta$ | | $Z_2^{0}$ | Defocus | $2r^2 - 1$ | | $Z_2^{2}$ | Astigmatism 0° | $r^2\cos 2\theta$ | | $Z_3^{-1}$ | Coma Y | $(3r^3 - 2r)\sin\theta$ | | $Z_3^{1}$ | Coma X | $(3r^3 - 2r)\cos\theta$ | | $Z_4^{0}$ | Spherical | $6r^4 - 6r^2 + 1$ | Orthogonality Property: $$ \int_0^1 \int_0^{2\pi} Z_n^m(r,\theta) Z_{n'}^{m'}(r,\theta) \, r \, dr \, d\theta = \frac{\pi}{n+1}\delta_{nn'}\delta_{mm'} $$ Gaussian Process Regression (Kriging) Prior Distribution: $$ f(\mathbf{x}) \sim \mathcal{GP}(m(\mathbf{x}), k(\mathbf{x}, \mathbf{x}')) $$ Common Kernel Functions: *Squared Exponential (RBF)*: $$ k(\mathbf{x}, \mathbf{x}') = \sigma^2 \exp\left(-\frac{\|\mathbf{x} - \mathbf{x}'\|^2}{2\ell^2}\right) $$ *Matérn Kernel*: $$ k(r) = \sigma^2 \frac{2^{1-\nu}}{\Gamma(\nu)}\left(\frac{\sqrt{2\nu}r}{\ell}\right)^\nu K_\nu\left(\frac{\sqrt{2\nu}r}{\ell}\right) $$ Where $K_\nu$ is the modified Bessel function of the second kind. Posterior Predictive Mean: $$ \bar{f}_* = \mathbf{k}_*^\top(\mathbf{K} + \sigma_n^2\mathbf{I})^{-1}\mathbf{y} $$ Posterior Predictive Variance: $$ \text{Var}(f_*) = k(\mathbf{x}_*, \mathbf{x}_*) - \mathbf{k}_*^\top(\mathbf{K} + \sigma_n^2\mathbf{I})^{-1}\mathbf{k}_* $$ Mixed Effects Models Semiconductor data has hierarchical structure (wafers within lots, lots within tools). General Model $$ y_{ijk} = \mathbf{x}_{ijk}^\top\boldsymbol{\beta} + b_i^{(\text{tool})} + b_{ij}^{(\text{lot})} + \varepsilon_{ijk} $$ Random Effects Distribution: - $b_i^{(\text{tool})} \sim N(0, \sigma_{\text{tool}}^2)$ - $b_{ij}^{(\text{lot})} \sim N(0, \sigma_{\text{lot}}^2)$ - $\varepsilon_{ijk} \sim N(0, \sigma^2)$ Matrix Notation $$ \mathbf{y} = \mathbf{X}\boldsymbol{\beta} + \mathbf{Z}\mathbf{b} + \boldsymbol{\varepsilon} $$ Where: - $\mathbf{b} \sim N(\mathbf{0}, \mathbf{G})$ - $\boldsymbol{\varepsilon} \sim N(\mathbf{0}, \mathbf{R})$ - $\text{Var}(\mathbf{y}) = \mathbf{V} = \mathbf{Z}\mathbf{G}\mathbf{Z}^\top + \mathbf{R}$ REML Estimation Restricted Log-Likelihood: $$ \ell_{\text{REML}}(\boldsymbol{\theta}) = -\frac{1}{2}\left[\log|\mathbf{V}| + \log|\mathbf{X}^\top\mathbf{V}^{-1}\mathbf{X}| + \mathbf{r}^\top\mathbf{V}^{-1}\mathbf{r}\right] $$ Where $\mathbf{r} = \mathbf{y} - \mathbf{X}\hat{\boldsymbol{\beta}}$. Physics-Informed Regression Models Arrhenius-Based Models (Thermal Processes) Rate Equation: $$ k = A \exp\left(-\frac{E_a}{RT}\right) $$ Linearized Form (for regression): $$ \ln(k) = \ln(A) - \frac{E_a}{R} \cdot \frac{1}{T} $$ Parameters: - $k$ — rate constant - $A$ — pre-exponential factor - $E_a$ — activation energy (J/mol) - $R$ — gas constant (8.314 J/mol·K) - $T$ — absolute temperature (K) Preston's Equation (CMP) Basic Form: $$ \text{MRR} = K_p \cdot P \cdot V $$ Extended Model: $$ \text{MRR} = K_p \cdot P^a \cdot V^b \cdot f(\text{slurry}, \text{pad}) $$ Where: - MRR — material removal rate - $K_p$ — Preston coefficient - $P$ — applied pressure - $V$ — relative velocity Lithography Focus-Exposure Model $$ \text{CD} = \beta_0 + \beta_1 E + \beta_2 F + \beta_3 E^2 + \beta_4 F^2 + \beta_5 EF + \varepsilon $$ Variables: - CD — critical dimension - $E$ — exposure dose - $F$ — focus offset Bossung Curve: Plot of CD vs. focus at various exposure levels. Virtual Metrology Mathematics Predicting quality measurements from equipment sensor data in real-time. Model Structure $$ \hat{y} = f(\mathbf{x}_{\text{FDC}}; \boldsymbol{\theta}) $$ Where $\mathbf{x}_{\text{FDC}}$ is Fault Detection and Classification sensor data. EWMA Run-to-Run Control Exponentially Weighted Moving Average: $$ \hat{T}_{n+1} = \lambda y_n + (1-\lambda)\hat{T}_n $$ Properties: - $\lambda \in (0,1]$ — smoothing parameter - Smaller $\lambda$ → more smoothing - Larger $\lambda$ → faster response to changes Kalman Filter Approach State Equation: $$ \mathbf{x}_{k} = \mathbf{A}\mathbf{x}_{k-1} + \mathbf{w}_k, \quad \mathbf{w}_k \sim N(\mathbf{0}, \mathbf{Q}) $$ Measurement Equation: $$ y_k = \mathbf{H}\mathbf{x}_k + v_k, \quad v_k \sim N(0, R) $$ Update Equations: *Predict*: $$ \hat{\mathbf{x}}_{k|k-1} = \mathbf{A}\hat{\mathbf{x}}_{k-1|k-1} $$ $$ \mathbf{P}_{k|k-1} = \mathbf{A}\mathbf{P}_{k-1|k-1}\mathbf{A}^\top + \mathbf{Q} $$ *Update*: $$ \mathbf{K}_k = \mathbf{P}_{k|k-1}\mathbf{H}^\top(\mathbf{H}\mathbf{P}_{k|k-1}\mathbf{H}^\top + R)^{-1} $$ $$ \hat{\mathbf{x}}_{k|k} = \hat{\mathbf{x}}_{k|k-1} + \mathbf{K}_k(y_k - \mathbf{H}\hat{\mathbf{x}}_{k|k-1}) $$ Classification and Count Models Logistic Regression (Binary Outcomes) For pass/fail or defect/no-defect classification: Model: $$ P(Y=1|\mathbf{x}) = \frac{1}{1 + \exp(-\mathbf{x}^\top\boldsymbol{\beta})} = \sigma(\mathbf{x}^\top\boldsymbol{\beta}) $$ Logit Link: $$ \text{logit}(p) = \ln\left(\frac{p}{1-p}\right) = \mathbf{x}^\top\boldsymbol{\beta} $$ Log-Likelihood: $$ \ell(\boldsymbol{\beta}) = \sum_{i=1}^{n}\left[y_i \log(\pi_i) + (1-y_i)\log(1-\pi_i)\right] $$ Newton-Raphson Update: $$ \boldsymbol{\beta}^{(t+1)} = \boldsymbol{\beta}^{(t)} + (\mathbf{X}^\top\mathbf{W}\mathbf{X})^{-1}\mathbf{X}^\top(\mathbf{y} - \boldsymbol{\pi}) $$ Where $\mathbf{W} = \text{diag}(\pi_i(1-\pi_i))$. Poisson Regression (Defect Counts) Model: $$ \log(\mu) = \mathbf{x}^\top\boldsymbol{\beta}, \quad Y \sim \text{Poisson}(\mu) $$ Probability Mass Function: $$ P(Y = y) = \frac{\mu^y e^{-\mu}}{y!} $$ Model Validation and Diagnostics Goodness of Fit Metrics Coefficient of Determination: $$ R^2 = 1 - \frac{\text{SSE}}{\text{SST}} = 1 - \frac{\sum_{i=1}^{n}(y_i - \hat{y}_i)^2}{\sum_{i=1}^{n}(y_i - \bar{y})^2} $$ Adjusted R-Squared: $$ R^2_{\text{adj}} = 1 - (1-R^2)\frac{n-1}{n-k-1} $$ Root Mean Square Error: $$ \text{RMSE} = \sqrt{\frac{1}{n}\sum_{i=1}^{n}(y_i - \hat{y}_i)^2} $$ Mean Absolute Error: $$ \text{MAE} = \frac{1}{n}\sum_{i=1}^{n}|y_i - \hat{y}_i| $$ Cross-Validation K-Fold CV Error: $$ \text{CV}_{(K)} = \frac{1}{K}\sum_{k=1}^{K}\text{MSE}_k $$ Leave-One-Out CV: $$ \text{LOOCV} = \frac{1}{n}\sum_{i=1}^{n}(y_i - \hat{y}_{(-i)})^2 $$ Information Criteria Akaike Information Criterion: $$ \text{AIC} = 2k - 2\ln(\hat{L}) $$ Bayesian Information Criterion: $$ \text{BIC} = k\ln(n) - 2\ln(\hat{L}) $$ Diagnostic Statistics Variance Inflation Factor: $$ \text{VIF}_j = \frac{1}{1-R_j^2} $$ Where $R_j^2$ is the $R^2$ from regressing $x_j$ on all other predictors. Rule of thumb: VIF > 10 indicates problematic multicollinearity. Cook's Distance: $$ D_i = \frac{(\hat{\mathbf{y}} - \hat{\mathbf{y}}_{(-i)})^\top(\hat{\mathbf{y}} - \hat{\mathbf{y}}_{(-i)})}{k \cdot \text{MSE}} $$ Leverage: $$ h_{ii} = [\mathbf{H}]_{ii} $$ Where $\mathbf{H} = \mathbf{X}(\mathbf{X}^\top\mathbf{X})^{-1}\mathbf{X}^\top$ is the hat matrix. Studentized Residuals: $$ r_i = \frac{e_i}{\hat{\sigma}\sqrt{1 - h_{ii}}} $$ Bayesian Regression Provides full uncertainty quantification for risk-sensitive manufacturing decisions. Bayesian Linear Regression Prior: $$ \boldsymbol{\beta} | \sigma^2 \sim N(\boldsymbol{\beta}_0, \sigma^2\mathbf{V}_0) $$ $$ \sigma^2 \sim \text{Inverse-Gamma}(a_0, b_0) $$ Posterior: $$ \boldsymbol{\beta} | \mathbf{y}, \sigma^2 \sim N(\boldsymbol{\beta}_n, \sigma^2\mathbf{V}_n) $$ Posterior Parameters: $$ \mathbf{V}_n = (\mathbf{V}_0^{-1} + \mathbf{X}^\top\mathbf{X})^{-1} $$ $$ \boldsymbol{\beta}_n = \mathbf{V}_n(\mathbf{V}_0^{-1}\boldsymbol{\beta}_0 + \mathbf{X}^\top\mathbf{y}) $$ Predictive Distribution $$ p(y_*|\mathbf{x}_*, \mathbf{y}) = \int p(y_*|\mathbf{x}_*, \boldsymbol{\beta}, \sigma^2) \, p(\boldsymbol{\beta}, \sigma^2|\mathbf{y}) \, d\boldsymbol{\beta} \, d\sigma^2 $$ For conjugate priors, this is a Student-t distribution. Credible Intervals 95% Credible Interval for $\beta_j$: $$ \beta_j \in \left[\hat{\beta}_j - t_{0.025,\nu}\cdot \text{SE}(\hat{\beta}_j), \quad \hat{\beta}_j + t_{0.025,\nu}\cdot \text{SE}(\hat{\beta}_j)\right] $$ Design of Experiments (DOE) Full Factorial Design For $k$ factors at 2 levels: $$ N = 2^k \text{ runs} $$ Fractional Factorial Design $$ N = 2^{k-p} \text{ runs} $$ Resolution: - Resolution III: Main effects aliased with 2-factor interactions - Resolution IV: Main effects clear; 2FIs aliased with each other - Resolution V: Main effects and 2FIs clear Central Composite Design (CCD) Components: - $2^k$ factorial points - $2k$ axial (star) points at distance $\alpha$ - $n_0$ center points Rotatability Condition: $$ \alpha = (2^k)^{1/4} $$ D-Optimal Design Maximizes the determinant of the information matrix: $$ \max_{\mathbf{X}} |\mathbf{X}^\top\mathbf{X}| $$ Equivalently, minimizes the generalized variance of $\hat{\boldsymbol{\beta}}$. I-Optimal Design Minimizes average prediction variance: $$ \min_{\mathbf{X}} \int_{\mathcal{R}} \text{Var}(\hat{y}(\mathbf{x})) \, d\mathbf{x} $$ Reliability Analysis Cox Proportional Hazards Model Hazard Function: $$ h(t|\mathbf{x}) = h_0(t) \cdot \exp(\mathbf{x}^\top\boldsymbol{\beta}) $$ Where: - $h(t|\mathbf{x})$ — hazard at time $t$ given covariates $\mathbf{x}$ - $h_0(t)$ — baseline hazard - $\boldsymbol{\beta}$ — regression coefficients Partial Likelihood $$ L(\boldsymbol{\beta}) = \prod_{i: \delta_i = 1} \frac{\exp(\mathbf{x}_i^\top\boldsymbol{\beta})}{\sum_{j \in \mathcal{R}(t_i)} \exp(\mathbf{x}_j^\top\boldsymbol{\beta})} $$ Where $\mathcal{R}(t_i)$ is the risk set at time $t_i$. Challenge-Method Mapping | Manufacturing Challenge | Mathematical Approach | |------------------------|----------------------| | High dimensionality | PLS, LASSO, Elastic Net | | Multicollinearity | Ridge regression, PCR, VIF analysis | | Spatial wafer patterns | Zernike polynomials, GP regression | | Hierarchical data | Mixed effects models, REML | | Nonlinear processes | RSM, polynomial models, transformations | | Physics constraints | Arrhenius, Preston equation integration | | Uncertainty quantification | Bayesian methods, bootstrap, prediction intervals | | Binary outcomes | Logistic regression | | Count data | Poisson regression | | Real-time control | Kalman filter, EWMA | | Time-to-failure | Cox proportional hazards | Equations Quick Reference Estimation $$ \hat{\boldsymbol{\beta}}_{\text{OLS}} = (\mathbf{X}^\top\mathbf{X})^{-1}\mathbf{X}^\top\mathbf{y} $$ $$ \hat{\boldsymbol{\beta}}_{\text{Ridge}} = (\mathbf{X}^\top\mathbf{X} + \lambda\mathbf{I})^{-1}\mathbf{X}^\top\mathbf{y} $$ Prediction Interval $$ \hat{y}_0 \pm t_{\alpha/2, n-k-1} \cdot \sqrt{\text{MSE}\left(1 + \mathbf{x}_0^\top(\mathbf{X}^\top\mathbf{X})^{-1}\mathbf{x}_0\right)} $$ Confidence Interval for $\beta_j$ $$ \hat{\beta}_j \pm t_{\alpha/2, n-k-1} \cdot \text{SE}(\hat{\beta}_j) $$ Process Capability $$ C_p = \frac{\text{USL} - \text{LSL}}{6\sigma} $$ $$ C_{pk} = \min\left(\frac{\text{USL} - \mu}{3\sigma}, \frac{\mu - \text{LSL}}{3\sigma}\right) $$ Reference | Symbol | Description | |--------|-------------| | $\mathbf{y}$ | Response vector | | $\mathbf{X}$ | Design matrix | | $\boldsymbol{\beta}$ | Coefficient vector | | $\hat{\boldsymbol{\beta}}$ | Estimated coefficients | | $\boldsymbol{\varepsilon}$ | Error vector | | $\sigma^2$ | Error variance | | $\lambda$ | Regularization parameter | | $\mathbf{I}$ | Identity matrix | | $\|\cdot\|_1$ | L1 norm (sum of absolute values) | | $\|\cdot\|_2$ | L2 norm (Euclidean) | | $\mathbf{A}^\top$ | Matrix transpose | | $\mathbf{A}^{-1}$ | Matrix inverse | | $|\mathbf{A}|$ | Matrix determinant | | $N(\mu, \sigma^2)$ | Normal distribution | | $\mathcal{GP}$ | Gaussian Process |

regression-based ocd, metrology

Fit model parameters directly.

resin bleed, packaging

Compound on surfaces.

resist profile simulation,lithography

Predict 3D resist shape after develop.

resist sensitivity,lithography

Amount of energy needed to expose resist.

resist spin coating,lithography

Apply liquid resist by spinning wafer at high speed.

resist strip / ashing,lithography

Remove resist after etching using plasma or solvents.

resolution,lithography

Smallest feature size that can be printed.

resolution,metrology

Smallest measurable difference.

resonant ionization mass spectrometry, rims, metrology

Selective ionization for ultra-sensitive detection.

resonant raman, metrology

Enhanced scattering at absorption resonance.

resonant soft x-ray scatterometry, metrology

Element-specific CD measurement.

reticle / photomask,lithography

Glass plate with circuit pattern for lithography exposure.

reticle handling, lithography

Procedures for moving and storing masks.

reticle lifetime, lithography

Number of wafer exposures before replacement.

reticle management, lithography

Track and control mask inventory.

reticle, lithography

Another term for photomask.

reverse bonding, packaging

Bond to substrate first.

reverse tone imaging,lithography

Use negative resist with positive mask (or vice versa).

review sem,metrology

High-resolution follow-up of detected defects.

rga (residual gas analyzer),rga,residual gas analyzer,metrology

Mass spec analyzing chamber gases.

rie, reactive ion etch, reactive ion etching, dry etch, plasma etch, etch modeling, plasma physics, ion bombardment

# Mathematical Modeling of Plasma Etching in Semiconductor Manufacturing ## Introduction Plasma etching is a critical process in semiconductor manufacturing where reactive gases are ionized to create a plasma, which selectively removes material from a wafer surface. The mathematical modeling of this process spans multiple physics domains: - **Electromagnetic theory** — RF power coupling and field distributions - **Statistical mechanics** — Particle distributions and kinetic theory - **Reaction kinetics** — Gas-phase and surface chemistry - **Transport phenomena** — Species diffusion and convection - **Surface science** — Etch mechanisms and selectivity ## Foundational Plasma Physics ### Boltzmann Transport Equation The most fundamental description of plasma behavior is the **Boltzmann transport equation**, governing the evolution of the particle velocity distribution function $f(\mathbf{r}, \mathbf{v}, t)$: $$ \frac{\partial f}{\partial t} + \mathbf{v} \cdot \nabla f + \frac{\mathbf{F}}{m} \cdot \nabla_v f = \left(\frac{\partial f}{\partial t}\right)_{\text{collision}} $$ **Where:** - $f(\mathbf{r}, \mathbf{v}, t)$ — Velocity distribution function - $\mathbf{v}$ — Particle velocity - $\mathbf{F}$ — External force (electromagnetic) - $m$ — Particle mass - RHS — Collision integral ### Fluid Moment Equations For computational tractability, velocity moments of the Boltzmann equation yield fluid equations: #### Continuity Equation (Mass Conservation) $$ \frac{\partial n}{\partial t} + \nabla \cdot (n\mathbf{u}) = S - L $$ **Where:** - $n$ — Species number density $[\text{m}^{-3}]$ - $\mathbf{u}$ — Drift velocity $[\text{m/s}]$ - $S$ — Source term (generation rate) - $L$ — Loss term (consumption rate) #### Momentum Conservation $$ \frac{\partial (nm\mathbf{u})}{\partial t} + \nabla \cdot (nm\mathbf{u}\mathbf{u}) + \nabla p = nq(\mathbf{E} + \mathbf{u} \times \mathbf{B}) - nm\nu_m \mathbf{u} $$ **Where:** - $p = nk_BT$ — Pressure - $q$ — Particle charge - $\mathbf{E}$, $\mathbf{B}$ — Electric and magnetic fields - $\nu_m$ — Momentum transfer collision frequency $[\text{s}^{-1}]$ #### Energy Conservation $$ \frac{\partial}{\partial t}\left(\frac{3}{2}nk_BT\right) + \nabla \cdot \mathbf{q} + p\nabla \cdot \mathbf{u} = Q_{\text{heating}} - Q_{\text{loss}} $$ **Where:** - $k_B = 1.38 \times 10^{-23}$ J/K — Boltzmann constant - $\mathbf{q}$ — Heat flux vector - $Q_{\text{heating}}$ — Power input (Joule heating, stochastic heating) - $Q_{\text{loss}}$ — Energy losses (collisions, radiation) ## Electromagnetic Field Coupling ### Maxwell's Equations For capacitively coupled plasma (CCP) and inductively coupled plasma (ICP) reactors: $$ \nabla \times \mathbf{E} = -\frac{\partial \mathbf{B}}{\partial t} $$ $$ \nabla \times \mathbf{H} = \mathbf{J} + \frac{\partial \mathbf{D}}{\partial t} $$ $$ \nabla \cdot \mathbf{D} = \rho $$ $$ \nabla \cdot \mathbf{B} = 0 $$ ### Plasma Conductivity The plasma current density couples through the complex conductivity: $$ \mathbf{J} = \sigma \mathbf{E} $$ For RF plasmas, the **complex conductivity** is: $$ \sigma = \frac{n_e e^2}{m_e(\nu_m + i\omega)} $$ **Where:** - $n_e$ — Electron density - $e = 1.6 \times 10^{-19}$ C — Elementary charge - $m_e = 9.1 \times 10^{-31}$ kg — Electron mass - $\omega$ — RF angular frequency - $\nu_m$ — Electron-neutral collision frequency ### Power Deposition Time-averaged power density deposited into the plasma: $$ P = \frac{1}{2}\text{Re}(\mathbf{J} \cdot \mathbf{E}^*) $$ **Typical values:** - CCP: $0.1 - 1$ W/cm³ - ICP: $0.5 - 5$ W/cm³ ## Plasma Sheath Physics The sheath is a thin, non-neutral region at the plasma-wafer interface that accelerates ions toward the surface, enabling anisotropic etching. ### Bohm Criterion Minimum ion velocity entering the sheath: $$ u_i \geq u_B = \sqrt{\frac{k_B T_e}{M_i}} $$ **Where:** - $u_B$ — Bohm velocity - $T_e$ — Electron temperature (typically 2–5 eV) - $M_i$ — Ion mass **Example:** For Ar⁺ ions with $T_e = 3$ eV: $$ u_B = \sqrt{\frac{3 \times 1.6 \times 10^{-19}}{40 \times 1.67 \times 10^{-27}}} \approx 2.7 \text{ km/s} $$ ### Child-Langmuir Law For a collisionless sheath, the ion current density is: $$ J = \frac{4\varepsilon_0}{9}\sqrt{\frac{2e}{M_i}} \cdot \frac{V_s^{3/2}}{d^2} $$ **Where:** - $\varepsilon_0 = 8.85 \times 10^{-12}$ F/m — Vacuum permittivity - $V_s$ — Sheath voltage drop (typically 10–500 V) - $d$ — Sheath thickness ### Sheath Thickness The sheath thickness scales as: $$ d \approx \lambda_D \left(\frac{2eV_s}{k_BT_e}\right)^{3/4} $$ **Where** the Debye length is: $$ \lambda_D = \sqrt{\frac{\varepsilon_0 k_B T_e}{n_e e^2}} $$ ### Ion Angular Distribution Ions arrive at the wafer with an angular distribution: $$ f(\theta) \propto \exp\left(-\frac{\theta^2}{2\sigma^2}\right) $$ **Where:** $$ \sigma \approx \arctan\left(\sqrt{\frac{k_B T_i}{eV_s}}\right) $$ **Typical values:** $\sigma \approx 2°–5°$ for high-bias conditions. ## Electron Energy Distribution Function ### Non-Maxwellian Distributions In low-pressure plasmas (1–100 mTorr), the EEDF deviates from Maxwellian. #### Two-Term Approximation The EEDF is expanded as: $$ f(\varepsilon, \theta) = f_0(\varepsilon) + f_1(\varepsilon)\cos\theta $$ The isotropic part $f_0$ satisfies: $$ \frac{d}{d\varepsilon}\left[\varepsilon D \frac{df_0}{d\varepsilon} + \left(V + \frac{\varepsilon\nu_{\text{inel}}}{\nu_m}\right)f_0\right] = 0 $$ #### Common Distribution Functions | Distribution | Functional Form | Applicability | |-------------|-----------------|---------------| | **Maxwellian** | $f(\varepsilon) \propto \sqrt{\varepsilon} \exp\left(-\frac{\varepsilon}{k_BT_e}\right)$ | High pressure, collisional | | **Druyvesteyn** | $f(\varepsilon) \propto \sqrt{\varepsilon} \exp\left(-\left(\frac{\varepsilon}{k_BT_e}\right)^2\right)$ | Elastic collisions dominant | | **Bi-Maxwellian** | Sum of two Maxwellians | Hot tail population | ### Generalized Form $$ f(\varepsilon) \propto \sqrt{\varepsilon} \cdot \exp\left[-\left(\frac{\varepsilon}{k_BT_e}\right)^x\right] $$ - $x = 1$ → Maxwellian - $x = 2$ → Druyvesteyn ## Plasma Chemistry and Reaction Kinetics ### Species Balance Equation For species $i$: $$ \frac{\partial n_i}{\partial t} + \nabla \cdot \mathbf{\Gamma}_i = \sum_j R_j $$ **Where:** - $\mathbf{\Gamma}_i$ — Species flux - $R_j$ — Reaction rates ### Electron-Impact Rate Coefficients Rate coefficients are calculated by integration over the EEDF: $$ k = \int_0^\infty \sigma(\varepsilon) v(\varepsilon) f(\varepsilon) \, d\varepsilon = \langle \sigma v \rangle $$ **Where:** - $\sigma(\varepsilon)$ — Energy-dependent cross-section $[\text{m}^2]$ - $v(\varepsilon) = \sqrt{2\varepsilon/m_e}$ — Electron velocity - $f(\varepsilon)$ — Normalized EEDF ### Heavy-Particle Reactions Arrhenius kinetics for neutral reactions: $$ k = A T^n \exp\left(-\frac{E_a}{k_BT}\right) $$ **Where:** - $A$ — Pre-exponential factor - $n$ — Temperature exponent - $E_a$ — Activation energy ### Example: SF₆/O₂ Plasma Chemistry #### Electron-Impact Reactions | Reaction | Type | Threshold | |----------|------|-----------| | $e + \text{SF}_6 \rightarrow \text{SF}_5 + \text{F} + e$ | Dissociation | ~10 eV | | $e + \text{SF}_6 \rightarrow \text{SF}_6^-$ | Attachment | ~0 eV | | $e + \text{SF}_6 \rightarrow \text{SF}_5^+ + \text{F} + 2e$ | Ionization | ~16 eV | | $e + \text{O}_2 \rightarrow \text{O} + \text{O} + e$ | Dissociation | ~6 eV | #### Gas-Phase Reactions - $\text{F} + \text{O} \rightarrow \text{FO}$ (reduces F atom density) - $\text{SF}_5 + \text{F} \rightarrow \text{SF}_6$ (recombination) - $\text{O} + \text{CF}_3 \rightarrow \text{COF}_2 + \text{F}$ (polymer removal) #### Surface Reactions - $\text{F} + \text{Si}(s) \rightarrow \text{SiF}_{(\text{ads})}$ - $\text{SiF}_{(\text{ads})} + 3\text{F} \rightarrow \text{SiF}_4(g)$ (volatile product) ## Transport Phenomena ### Drift-Diffusion Model For charged species, the flux is: $$ \mathbf{\Gamma} = \pm \mu n \mathbf{E} - D \nabla n $$ **Where:** - Upper sign: positive ions - Lower sign: electrons - $\mu$ — Mobility $[\text{m}^2/(\text{V}\cdot\text{s})]$ - $D$ — Diffusion coefficient $[\text{m}^2/\text{s}]$ ### Einstein Relation Connects mobility and diffusion: $$ D = \frac{\mu k_B T}{e} $$ ### Ambipolar Diffusion When quasi-neutrality holds ($n_e \approx n_i$): $$ D_a = \frac{\mu_i D_e + \mu_e D_i}{\mu_i + \mu_e} \approx D_i\left(1 + \frac{T_e}{T_i}\right) $$ Since $T_e \gg T_i$ typically: $D_a \approx D_i (1 + T_e/T_i) \approx 100 D_i$ ### Neutral Transport For reactive neutrals (radicals), Fickian diffusion: $$ \frac{\partial n}{\partial t} = D\nabla^2 n + S - L $$ #### Surface Boundary Condition $$ -D\frac{\partial n}{\partial x}\bigg|_{\text{surface}} = \frac{1}{4}\gamma n v_{\text{th}} $$ **Where:** - $\gamma$ — Sticking/reaction coefficient (0 to 1) - $v_{\text{th}} = \sqrt{\frac{8k_BT}{\pi m}}$ — Thermal velocity ### Knudsen Number Determines the appropriate transport regime: $$ \text{Kn} = \frac{\lambda}{L} $$ **Where:** - $\lambda$ — Mean free path - $L$ — Characteristic length | Kn Range | Regime | Model | |----------|--------|-------| | $< 0.01$ | Continuum | Navier-Stokes | | $0.01–0.1$ | Slip flow | Modified N-S | | $0.1–10$ | Transition | DSMC/BGK | | $> 10$ | Free molecular | Ballistic | ## Surface Reaction Modeling ### Langmuir Adsorption Kinetics For surface coverage $\theta$: $$ \frac{d\theta}{dt} = k_{\text{ads}}(1-\theta)P - k_{\text{des}}\theta - k_{\text{react}}\theta $$ **At steady state:** $$ \theta = \frac{k_{\text{ads}}P}{k_{\text{ads}}P + k_{\text{des}} + k_{\text{react}}} $$ ### Ion-Enhanced Etching The total etch rate combines multiple mechanisms: $$ \text{ER} = Y_{\text{chem}} \Gamma_n + Y_{\text{phys}} \Gamma_i + Y_{\text{syn}} \Gamma_i f(\theta) $$ **Where:** - $Y_{\text{chem}}$ — Chemical etch yield (isotropic) - $Y_{\text{phys}}$ — Physical sputtering yield - $Y_{\text{syn}}$ — Ion-enhanced (synergistic) yield - $\Gamma_n$, $\Gamma_i$ — Neutral and ion fluxes - $f(\theta)$ — Coverage-dependent function ### Ion Sputtering Yield #### Energy Dependence $$ Y(E) = A\left(\sqrt{E} - \sqrt{E_{\text{th}}}\right) \quad \text{for } E > E_{\text{th}} $$ **Typical threshold energies:** - Si: $E_{\text{th}} \approx 20$ eV - SiO₂: $E_{\text{th}} \approx 30$ eV - Si₃N₄: $E_{\text{th}} \approx 25$ eV #### Angular Dependence $$ Y(\theta) = Y(0) \cos^{-f}(\theta) \exp\left[-b\left(\frac{1}{\cos\theta} - 1\right)\right] $$ **Behavior:** - Increases from normal incidence - Peaks at $\theta \approx 60°–70°$ - Decreases at grazing angles (reflection dominates) ## Feature-Scale Profile Evolution ### Level Set Method The surface is represented as the zero contour of $\phi(\mathbf{x}, t)$: $$ \frac{\partial \phi}{\partial t} + V_n |\nabla \phi| = 0 $$ **Where:** - $\phi > 0$ — Material - $\phi < 0$ — Void/vacuum - $\phi = 0$ — Surface - $V_n$ — Local normal etch velocity ### Local Etch Rate Calculation The normal velocity $V_n$ depends on: 1. **Ion flux and angular distribution** $$\Gamma_i(\mathbf{x}) = \int f(\theta, E) \, d\Omega \, dE$$ 2. **Neutral flux** (with shadowing) $$\Gamma_n(\mathbf{x}) = \Gamma_{n,0} \cdot \text{VF}(\mathbf{x})$$ where VF is the view factor 3. **Surface chemistry state** $$V_n = f(\Gamma_i, \Gamma_n, \theta_{\text{coverage}}, T)$$ ### Neutral Transport in High-Aspect-Ratio Features #### Clausing Transmission Factor For a tube of aspect ratio AR: $$ K \approx \frac{1}{1 + 0.5 \cdot \text{AR}} $$ #### View Factor Calculations For surface element $dA_1$ seeing $dA_2$: $$ F_{1 \rightarrow 2} = \frac{1}{\pi} \int \frac{\cos\theta_1 \cos\theta_2}{r^2} \, dA_2 $$ ## Monte Carlo Methods ### Test-Particle Monte Carlo Algorithm ``` 1. SAMPLE incident particle from flux distribution at feature opening - Ion: from IEDF and IADF - Neutral: from Maxwellian 2. TRACE trajectory through feature - Ion: ballistic, solve equation of motion - Neutral: random walk with wall collisions 3. DETERMINE reaction at surface impact - Sample from probability distribution - Update surface coverage if adsorption 4. UPDATE surface geometry - Remove material (etching) - Add material (deposition) 5. REPEAT for statistically significant sample ``` ### Ion Trajectory Integration Through the sheath/feature: $$ m\frac{d^2\mathbf{r}}{dt^2} = q\mathbf{E}(\mathbf{r}) $$ **Numerical integration:** Velocity-Verlet or Boris algorithm ### Collision Sampling Null-collision method for efficiency: $$ P_{\text{collision}} = 1 - \exp(-\nu_{\text{max}} \Delta t) $$ **Where** $\nu_{\text{max}}$ is the maximum possible collision frequency. ## Multi-Scale Modeling Framework ### Scale Hierarchy | Scale | Length | Time | Physics | Method | |-------|--------|------|---------|--------| | **Reactor** | cm–m | ms–s | Plasma transport, EM fields | Fluid PDE | | **Sheath** | µm–mm | µs–ms | Ion acceleration, EEDF | Kinetic/Fluid | | **Feature** | nm–µm | ns–ms | Profile evolution | Level set/MC | | **Atomic** | Å–nm | ps–ns | Reaction mechanisms | MD/DFT | ### Coupling Approaches #### Hierarchical (One-Way) ``` Atomic scale → Surface parameters ↓ Feature scale ← Fluxes from reactor scale ↓ Reactor scale → Process outputs ``` #### Concurrent (Two-Way) - Feature-scale results feed back to reactor scale - Requires iterative solution - Computationally expensive ## Numerical Methods and Challenges ### Stiff ODE Systems Plasma chemistry involves timescales spanning many orders of magnitude: | Process | Timescale | |---------|-----------| | Electron attachment | $\sim 10^{-10}$ s | | Ion-molecule reactions | $\sim 10^{-6}$ s | | Metastable decay | $\sim 10^{-3}$ s | | Surface diffusion | $\sim 10^{-1}$ s | #### Implicit Methods Required **Backward Differentiation Formula (BDF):** $$ y_{n+1} = \sum_{j=0}^{k-1} \alpha_j y_{n-j} + h\beta f(t_{n+1}, y_{n+1}) $$ ### Spatial Discretization #### Finite Volume Method Ensures mass conservation: $$ \int_V \frac{\partial n}{\partial t} dV + \oint_S \mathbf{\Gamma} \cdot d\mathbf{S} = \int_V S \, dV $$ #### Mesh Requirements - Sheath resolution: $\Delta x < \lambda_D$ - RF skin depth: $\Delta x < \delta$ - Adaptive mesh refinement (AMR) common ### EM-Plasma Coupling **Iterative scheme:** 1. Solve Maxwell's equations for $\mathbf{E}$, $\mathbf{B}$ 2. Update plasma transport (density, temperature) 3. Recalculate $\sigma$, $\varepsilon_{\text{plasma}}$ 4. Repeat until convergence ## Advanced Topics ### Atomic Layer Etching (ALE) Self-limiting reactions for atomic precision: $$ \text{EPC} = \Theta \cdot d_{\text{ML}} $$ **Where:** - EPC — Etch per cycle - $\Theta$ — Modified layer coverage fraction - $d_{\text{ML}}$ — Monolayer thickness #### ALE Cycle 1. **Modification step:** Reactive gas creates modified surface layer $$\frac{d\Theta}{dt} = k_{\text{mod}}(1-\Theta)P_{\text{gas}}$$ 2. **Removal step:** Ion bombardment removes modified layer only $$\text{ER} = Y_{\text{mod}}\Gamma_i\Theta$$ ### Pulsed Plasma Dynamics Time-modulated RF introduces: - **Active glow:** Plasma on, high ion/radical generation - **Afterglow:** Plasma off, selective chemistry #### Ion Energy Modulation By pulsing bias: $$ \langle E_i \rangle = \frac{1}{T}\left[\int_0^{t_{\text{on}}} E_{\text{high}}dt + \int_{t_{\text{on}}}^{T} E_{\text{low}}dt\right] $$ ### High-Aspect-Ratio Etching (HAR) For AR > 50 (memory, 3D NAND): **Challenges:** - Ion angular broadening → bowing - Neutral depletion at bottom - Feature charging → twisting - Mask erosion → tapering **Ion Angular Distribution Broadening:** $$ \sigma_{\text{effective}} = \sqrt{\sigma_{\text{sheath}}^2 + \sigma_{\text{scattering}}^2} $$ **Neutral Flux at Bottom:** $$ \Gamma_{\text{bottom}} \approx \Gamma_{\text{top}} \cdot K(\text{AR}) $$ ### Machine Learning Integration **Applications:** - Surrogate models for fast prediction - Process optimization (Bayesian) - Virtual metrology - Anomaly detection **Physics-Informed Neural Networks (PINNs):** $$ \mathcal{L} = \mathcal{L}_{\text{data}} + \lambda \mathcal{L}_{\text{physics}} $$ Where $\mathcal{L}_{\text{physics}}$ enforces governing equations. ## Validation and Experimental Techniques ### Plasma Diagnostics | Technique | Measurement | Typical Values | |-----------|-------------|----------------| | **Langmuir probe** | $n_e$, $T_e$, EEDF | $10^{9}–10^{12}$ cm⁻³, 1–5 eV | | **OES** | Relative species densities | Qualitative/semi-quantitative | | **APMS** | Ion mass, energy | 1–500 amu, 0–500 eV | | **LIF** | Absolute radical density | $10^{11}–10^{14}$ cm⁻³ | | **Microwave interferometry** | $n_e$ (line-averaged) | $10^{10}–10^{12}$ cm⁻³ | ### Etch Characterization - **Profilometry:** Etch depth, uniformity - **SEM/TEM:** Feature profiles, sidewall angle - **XPS:** Surface composition - **Ellipsometry:** Film thickness, optical properties ### Model Validation Workflow 1. **Plasma validation:** Match $n_e$, $T_e$, species densities 2. **Flux validation:** Compare ion/neutral fluxes to wafer 3. **Etch rate validation:** Blanket wafer etch rates 4. **Profile validation:** Patterned feature cross-sections ## Key Dimensionless Numbers Summary | Number | Definition | Physical Meaning | |--------|------------|------------------| | **Knudsen** | $\text{Kn} = \lambda/L$ | Continuum vs. kinetic | | **Damköhler** | $\text{Da} = \tau_{\text{transport}}/\tau_{\text{reaction}}$ | Transport vs. reaction limited | | **Sticking coefficient** | $\gamma = \text{reactions}/\text{collisions}$ | Surface reactivity | | **Aspect ratio** | $\text{AR} = \text{depth}/\text{width}$ | Feature geometry | | **Debye number** | $N_D = n\lambda_D^3$ | Plasma ideality | ## Physical Constants | Constant | Symbol | Value | |----------|--------|-------| | Elementary charge | $e$ | $1.602 \times 10^{-19}$ C | | Electron mass | $m_e$ | $9.109 \times 10^{-31}$ kg | | Proton mass | $m_p$ | $1.673 \times 10^{-27}$ kg | | Boltzmann constant | $k_B$ | $1.381 \times 10^{-23}$ J/K | | Vacuum permittivity | $\varepsilon_0$ | $8.854 \times 10^{-12}$ F/m | | Vacuum permeability | $\mu_0$ | $4\pi \times 10^{-7}$ H/m |

robot (wafer handling),robot,wafer handling,automation

Automated arm that picks and places wafers in tools.

runner system, packaging

Channels feeding compound.

runner waste, packaging

Compound in runners.

rutherford backscattering spectrometry (rbs),rutherford backscattering spectrometry,rbs,metrology

Elemental depth profiling.