Home› Knowledge Base› The Navier-Stokes equations govern momentum transport in the reactor and determine the velocity field through which precursors travel.

CVD modeling turns a deposition recipe into a testable chain of conservation laws, chemical mechanisms, surface boundary conditions, and scale-bridging assumptions, so its purpose is not merely to reproduce film thickness but to explain why rate, uniformity, composition, conformality, stress, and defects move together.

<svg viewBox="0 0 760 470" xmlns="http://www.w3.org/2000/svg" font-family="-apple-system,BlinkMacSystemFont,Segoe UI,Roboto,sans-serif">
  <rect x="0" y="0" width="760" height="470" rx="18" fill="#0d1117"/>
  <text x="380" y="32" fill="#e6edf3" font-size="20" font-weight="700" text-anchor="middle">Chemical Vapor Deposition (CVD) — Boundary Layer Kinetics &amp; Conformality</text>
  <text x="380" y="52" fill="#8b98a5" font-size="12" text-anchor="middle">Gas-Phase Precursor Transport, Surface Adsorption, Desorption &amp; Aspect-Ratio Step Coverage</text>

  <!-- Left Panel: Reaction Chamber & Boundary Layer Transport -->
  <g transform="translate(25, 70)">
    <rect width="345" height="330" fill="#161b22" stroke="#30363d" stroke-width="1.5" rx="8"/>
    <text x="172.5" y="24" fill="#60a5fa" font-size="13" font-weight="700" text-anchor="middle">1. Gas-Phase Transport &amp; Surface Reactions</text>
    
    <!-- Gas Inflow -->
    <rect x="25" y="40" width="295" height="22" fill="#1e293b" stroke="#60a5fa" stroke-width="1.2" rx="3"/>
    <text x="172.5" y="55" fill="#93c5fd" font-size="10.5" font-weight="600" text-anchor="middle">Main Bulk Gas Flow (SiH₄ + NH₃ / TEOS Inflow)</text>
    
    <!-- Main Flow Velocity Vector -->
    <line x1="40" y1="75" x2="270" y2="75" stroke="#38bdf8" stroke-width="2"/>
    <polygon points="275,75 267,71 267,79" fill="#38bdf8"/>
    <text x="150" y="70" fill="#7dd3fc" font-size="9" text-anchor="middle">Bulk Flow Velocity U∞</text>

    <!-- Stagnant Boundary Layer -->
    <path d="M 25 120 Q 150 95 320 95 L 320 120 L 25 120 Z" fill="#1e1b4b" fill-opacity="0.6"/>
    <line x1="25" y1="95" x2="320" y2="95" stroke="#818cf8" stroke-width="1" stroke-dasharray="3 3"/>
    <text x="172.5" y="108" fill="#c7d2fe" font-size="9.5" font-weight="700" text-anchor="middle">Stagnant Boundary Layer (Thickness δ)</text>

    <!-- Transport Step Annotations -->
    <!-- 1 Diffusion -->
    <line x1="80" y1="120" x2="80" y2="160" stroke="#f59e0b" stroke-width="1.5" stroke-dasharray="2 2"/>
    <polygon points="80,165 77,157 83,157" fill="#f59e0b"/>
    <text x="80" y="140" fill="#fbbf24" font-size="8.5" font-weight="700" text-anchor="end">1. Mass Transport  </text>

    <!-- 2 Adsorption -->
    <circle cx="140" cy="172" r="5" fill="#34d399"/>
    <text x="140" y="160" fill="#6ee7b7" font-size="8.5" font-weight="700" text-anchor="middle">2. Adsorption</text>

    <!-- 3 Surface Migration & Reaction -->
    <path d="M 145 172 Q 170 168 195 172" fill="none" stroke="#38bdf8" stroke-width="1.5"/>
    <polygon points="200,172 193,169 193,175" fill="#38bdf8"/>
    <text x="172.5" y="185" fill="#7dd3fc" font-size="8.5" font-weight="700" text-anchor="middle">3. Reaction &amp; Film Growth</text>

    <!-- 4 Byproduct Desorption -->
    <line x1="260" y1="170" x2="260" y2="130" stroke="#f87171" stroke-width="1.5" stroke-dasharray="2 2"/>
    <polygon points="260,125 257,133 263,133" fill="#f87171"/>
    <text x="265" y="145" fill="#fca5a5" font-size="8.5" font-weight="700">4. Byproducts ↑ (H₂)</text>

    <!-- Substrate & Growing Film -->
    <rect x="25" y="172" width="295" height="8" fill="#0284c7" stroke="#38bdf8" rx="1"/>
    <text x="315" y="168" fill="#38bdf8" font-size="8" text-anchor="end">Deposited Solid Film (SiO₂/Si₃N₄)</text>
    <rect x="25" y="180" width="295" height="40" fill="#1e293b" stroke="#64748b" rx="2"/>
    <text x="172.5" y="204" fill="#94a3b8" font-size="10" font-weight="700" text-anchor="middle">Heated Wafer Substrate (300°C – 800°C)</text>

    <!-- Susceptor / Heater -->
    <rect x="55" y="235" width="235" height="75" fill="#0f172a" stroke="#fbbf24" stroke-width="1" rx="4"/>
    <text x="172.5" y="255" fill="#fbbf24" font-size="10" font-weight="700" text-anchor="middle">Thermal / RF Plasma Susceptor</text>
    <text x="172.5" y="273" fill="#8b98a5" font-size="9" text-anchor="middle">LPCVD: Low Pressure ~0.1-1 Torr (Reaction-Limited)</text>
    <text x="172.5" y="290" fill="#8b98a5" font-size="9" text-anchor="middle">PECVD: Plasma Accelerated at 200°C-400°C</text>
  </g>

  <!-- Right Panel: Step Coverage & Conformality Comparison -->
  <g transform="translate(390, 70)">
    <rect width="345" height="330" fill="#161b22" stroke="#30363d" stroke-width="1.5" rx="8"/>
    <text x="172.5" y="24" fill="#34d399" font-size="13" font-weight="700" text-anchor="middle">2. Trench Conformality &amp; Step Coverage</text>

    <!-- Top Sub-panel: Conformal (LPCVD) -->
    <g transform="translate(15, 38)">
      <rect width="315" height="128" fill="#0d1117" stroke="#10b981" stroke-width="1.2" rx="6"/>
      <text x="157.5" y="20" fill="#3fb950" font-size="11" font-weight="700" text-anchor="middle">A. Conformal Deposition (LPCVD / High Migration)</text>

      <!-- Substrate with Trench -->
      <path d="M 30 35 L 110 35 L 110 100 L 205 100 L 205 35 L 285 35 L 285 118 L 30 118 Z" fill="#1e293b" stroke="#64748b"/>
      
      <!-- Conformal Film Coating -->
      <path d="M 30 35 L 110 35 L 110 100 L 205 100 L 205 35 L 285 35" fill="none" stroke="#34d399" stroke-width="7" stroke-linejoin="round"/>

      <text x="157.5" y="65" fill="#e6edf3" font-size="9" font-weight="700" text-anchor="middle">Uniform Wall &amp; Bottom Thickness</text>
      <text x="157.5" y="80" fill="#34d399" font-size="8.5" font-weight="700" text-anchor="middle">Step Coverage = t_bottom / t_top ≈ 100%</text>
      <text x="157.5" y="112" fill="#8b98a5" font-size="8" text-anchor="middle">Reaction-Limited Regime (Da &lt;&lt; 1) | Ideal for STI &amp; Liners</text>
    </g>

    <!-- Bottom Sub-panel: Non-Conformal / Overhang (PECVD / High Flux) -->
    <g transform="translate(15, 180)">
      <rect width="315" height="135" fill="#0d1117" stroke="#f87171" stroke-width="1" rx="6"/>
      <text x="157.5" y="20" fill="#f87171" font-size="11" font-weight="700" text-anchor="middle">B. Non-Conformal &amp; Overhang (Transport-Limited)</text>

      <!-- Substrate with Trench -->
      <path d="M 30 35 L 110 35 L 110 105 L 205 105 L 205 35 L 285 35 L 285 122 L 30 122 Z" fill="#1e293b" stroke="#64748b"/>
      
      <!-- Overhang Film Profile -->
      <path d="M 30 35 L 122 35 L 112 105 L 203 105 L 193 35 L 285 35" fill="none" stroke="#ef4444" stroke-width="10" stroke-linecap="square"/>
      
      <!-- Keyhole Void Callout -->
      <circle cx="157.5" cy="70" r="10" fill="#0d1117" stroke="#f87171" stroke-dasharray="2 2"/>
      <text x="157.5" y="73" fill="#fca5a5" font-size="8" font-weight="700" text-anchor="middle">Void</text>

      <text x="157.5" y="48" fill="#fca5a5" font-size="9" font-weight="700" text-anchor="middle">Top Overhang Pinch-Off</text>
      <text x="157.5" y="114" fill="#8b98a5" font-size="8" text-anchor="middle">Transport-Limited Regime (Da &gt;&gt; 1) | Risk of Keyhole Voiding</text>
    </g>
  </g>

  <!-- Bottom Key Insight Bar -->
  <rect x="25" y="415" width="710" height="22" rx="3" fill="#0b1220" stroke="#233043" stroke-width="0.8"/>
  <text x="380" y="430" fill="#fbbf24" font-size="9.5" font-weight="700" text-anchor="middle">Damköhler Number Da = k_s / h_G | Reaction-Limited (Da &lt;&lt; 1) yields 100% Conformality vs Transport-Limited (Da &gt;&gt; 1) Pinch-Off</text>

  <!-- Footer -->
  <text x="380" y="458" fill="#6b7684" font-size="11" text-anchor="middle">Essential thin-film deposition standard for ILD, STI liners, tungsten plugs &amp; GAA spacer dielectrics</text>
</svg>

Introduction

Chemical Vapor Deposition (CVD) is a critical thin-film deposition technique in semiconductor manufacturing. Gaseous precursors are introduced into a reaction chamber where they undergo chemical reactions to deposit solid films on heated substrates.

Key Process Steps

Transport of reactants from bulk gas to the substrate surface Gas-phase chemistry including precursor decomposition and intermediate formation Surface reactions involving adsorption, surface diffusion, and reaction Film nucleation and growth with specific microstructure evolution Byproduct desorption and transport away from the surface

Common CVD Types

APCVD — Atmospheric Pressure CVD LPCVD — Low Pressure CVD (0.1–10 Torr) PECVD — Plasma Enhanced CVD MOCVD — Metal-Organic CVD ALD — Atomic Layer Deposition HDPCVD — High Density Plasma CVD

Governing Equations

Continuity Equation (Mass Conservation)

$$\frac{\partial \rho}{\partial t} + \nabla \cdot (\rho \mathbf{u}) = 0$$

Where:

$\rho$ — gas density $\left[\text{kg/m}^3\right]$ $\mathbf{u}$ — velocity vector $\left[\text{m/s}\right]$ $t$ — time $\left[\text{s}\right]$

Momentum Equation (Navier-Stokes)

$$\rho \left( \frac{\partial \mathbf{u}}{\partial t} + \mathbf{u} \cdot \nabla \mathbf{u} \right) = -\nabla p + \mu \nabla^2 \mathbf{u} + \rho \mathbf{g}$$

Where:

$p$ — pressure $\left[\text{Pa}\right]$ $\mu$ — dynamic viscosity $\left[\text{Pa} \cdot \text{s}\right]$ $\mathbf{g}$ — gravitational acceleration $\left[\text{m/s}^2\right]$

Species Conservation Equation

$$\frac{\partial (\rho Y_i)}{\partial t} + \nabla \cdot (\rho \mathbf{u} Y_i) = \nabla \cdot (\rho D_i \nabla Y_i) + R_i$$

Where:

$Y_i$ — mass fraction of species $i$ $\left[\text{dimensionless}\right]$ $D_i$ — diffusion coefficient of species $i$ $\left[\text{m}^2/\text{s}\right]$ $R_i$ — net production rate from reactions $\left[\text{kg/m}^3 \cdot \text{s}\right]$

Energy Conservation Equation

$$\rho c_p \left( \frac{\partial T}{\partial t} + \mathbf{u} \cdot \nabla T \right) = \nabla \cdot (k \nabla T) + Q$$

Where:

$c_p$ — specific heat capacity $\left[\text{J/kg} \cdot \text{K}\right]$ $T$ — temperature $\left[\text{K}\right]$ $k$ — thermal conductivity $\left[\text{W/m} \cdot \text{K}\right]$ $Q$ — volumetric heat source $\left[\text{W/m}^3\right]$

Key Dimensionless Numbers

NumberDefinitionPhysical Meaning
Reynolds$Re = \frac{\rho u L}{\mu}$Inertial vs. viscous forces
Péclet$Pe = \frac{u L}{D}$Convection vs. diffusion
Damköhler$Da = \frac{k_s L}{D}$Reaction rate vs. transport rate
Knudsen$Kn = \frac{\lambda}{L}$Mean free path vs. length scale

Where:

$L$ — characteristic length $\left[\text{m}\right]$ $\lambda$ — mean free path $\left[\text{m}\right]$ $k_s$ — surface reaction rate constant $\left[\text{m/s}\right]$

Chemical Kinetics

Arrhenius Equation

The temperature dependence of reaction rate constants follows:

$$k = A \exp\left(-\frac{E_a}{R T}\right)$$

Where:

$k$ — rate constant $\left[\text{varies}\right]$ $A$ — pre-exponential factor $\left[\text{same as } k\right]$ $E_a$ — activation energy $\left[\text{J/mol}\right]$ $R$ — universal gas constant $= 8.314 \, \text{J/mol} \cdot \text{K}$

Gas-Phase Reactions

Example: Silane Pyrolysis

$$\text{SiH}_4 \xrightarrow{k_1} \text{SiH}_2 + \text{H}_2$$
$$\text{SiH}_2 + \text{SiH}_4 \xrightarrow{k_2} \text{Si}_2\text{H}_6$$

General reaction rate expression:

$$r_j = k_j \prod_{i} C_i^{ u_{ij}}$$

Where:

$r_j$ — rate of reaction $j$ $\left[\text{mol/m}^3 \cdot \text{s}\right]$ $C_i$ — concentration of species $i$ $\left[\text{mol/m}^3\right]$ $u_{ij}$ — stoichiometric coefficient of species $i$ in reaction $j$

Surface Reaction Kinetics

Hertz-Knudsen Impingement Flux

$$J = \frac{p}{\sqrt{2 \pi m k_B T}}$$

Where:

$J$ — molecular flux $\left[\text{molecules/m}^2 \cdot \text{s}\right]$ $p$ — partial pressure $\left[\text{Pa}\right]$ $m$ — molecular mass $\left[\text{kg}\right]$ $k_B$ — Boltzmann constant $= 1.381 \times 10^{-23} \, \text{J/K}$

Surface Reaction Rate

$$R_s = s \cdot J = s \cdot \frac{p}{\sqrt{2 \pi m k_B T}}$$

Where:

$s$ — sticking coefficient $\left[0 \leq s \leq 1\right]$

Langmuir-Hinshelwood Kinetics

For surface reaction between two adsorbed species:

$$r = \frac{k \, K_A \, K_B \, p_A \, p_B}{(1 + K_A p_A + K_B p_B)^2}$$

Where:

$K_A, K_B$ — adsorption equilibrium constants $\left[\text{Pa}^{-1}\right]$ $p_A, p_B$ — partial pressures of reactants A and B $\left[\text{Pa}\right]$

Eley-Rideal Mechanism

For reaction between adsorbed species and gas-phase species:

$$r = \frac{k \, K_A \, p_A \, p_B}{1 + K_A p_A}$$

Common CVD Reaction Systems

Silicon from Silane: $\text{SiH}_4 \rightarrow \text{Si}_{(s)} + 2\text{H}_2$

Silicon Dioxide from TEOS: $\text{Si(OC}_2\text{H}_5\text{)}_4 + 12\text{O}_2 \rightarrow \text{SiO}_2 + 8\text{CO}_2 + 10\text{H}_2\text{O}$

Silicon Nitride from DCS: $3\text{SiH}_2\text{Cl}_2 + 4\text{NH}_3 \rightarrow \text{Si}_3\text{N}_4 + 6\text{HCl} + 6\text{H}_2$

Tungsten from WF₆: $\text{WF}_6 + 3\text{H}_2 \rightarrow \text{W}_{(s)} + 6\text{HF}$

Process Regimes

Transport-Limited Regime

Characteristics:

High Damköhler number: $Da \gg 1$ Surface reactions are fast Deposition rate controlled by mass transport Sensitive to: Flow patterns Temperature gradients Reactor geometry

Deposition rate expression:

$$R_{dep} \approx \frac{D \cdot C_{\infty}}{\delta}$$

Where:

$C_{\infty}$ — bulk gas concentration $\left[\text{mol/m}^3\right]$ $\delta$ — boundary layer thickness $\left[\text{m}\right]$

Reaction-Limited Regime

Characteristics:

Low Damköhler number: $Da \ll 1$ Plenty of reactants at surface Rate controlled by surface kinetics Strong Arrhenius temperature dependence Better step coverage in features

Deposition rate expression:

$$R_{dep} \approx k_s \cdot C_s \approx k_s \cdot C_{\infty}$$

Where:

$k_s$ — surface reaction rate constant $\left[\text{m/s}\right]$ $C_s$ — surface concentration $\approx C_{\infty}$ $\left[\text{mol/m}^3\right]$

Regime Transition

The transition occurs when:

$$Da = \frac{k_s \delta}{D} \approx 1$$

Practical implications:

Transport-limited: Optimize flow, temperature uniformity Reaction-limited: Optimize temperature, precursor chemistry Mixed regime: Most complex to control and model

Multiscale Modeling

Scale Hierarchy

ScaleLengthTimeMethods
Reactorcm – ms – minCFD, FEM
Featurenm – μmms – sLevel set, Monte Carlo
Surfacenmμs – msKMC
AtomisticÅfs – psMD, DFT

Reactor-Scale Modeling

Governing physics:

Coupled Navier-Stokes + species + energy equations Multicomponent diffusion (Stefan-Maxwell) Chemical source terms

Stefan-Maxwell diffusion:

$$\nabla x_i = \sum_{j eq i} \frac{x_i x_j}{D_{ij}} (\mathbf{u}_j - \mathbf{u}_i)$$

Where:

$x_i$ — mole fraction of species $i$ $D_{ij}$ — binary diffusion coefficient $\left[\text{m}^2/\text{s}\right]$

Common software:

ANSYS Fluent COMSOL Multiphysics OpenFOAM (open-source) Silvaco Victory Process Synopsys Sentaurus

Feature-Scale Modeling

Key phenomena:

Knudsen diffusion in high-aspect-ratio features Molecular re-emission and reflection Surface reaction probability Film profile evolution

Knudsen diffusion coefficient:

$$D_K = \frac{d}{3} \sqrt{\frac{8 k_B T}{\pi m}}$$

Where:

$d$ — feature width $\left[\text{m}\right]$

Effective diffusivity (transition regime):

$$\frac{1}{D_{eff}} = \frac{1}{D_{mol}} + \frac{1}{D_K}$$

Level set method for surface tracking:

$$\frac{\partial \phi}{\partial t} + v_n |\nabla \phi| = 0$$

Where:

$\phi$ — level set function (zero at surface) $v_n$ — surface normal velocity (deposition rate)

Atomistic Modeling

Density Functional Theory (DFT):

Calculate binding energies Determine activation barriers Predict reaction pathways

Kinetic Monte Carlo (KMC):

Stochastic surface evolution Event rates from Arrhenius:

$$\Gamma_i = u_0 \exp\left(-\frac{E_i}{k_B T}\right)$$

Where:

$\Gamma_i$ — rate of event $i$ $\left[\text{s}^{-1}\right]$ $u_0$ — attempt frequency $\sim 10^{12} - 10^{13} \, \text{s}^{-1}$ $E_i$ — activation energy for event $i$ $\left[\text{eV}\right]$

CVD Process Variants

LPCVD (Low Pressure CVD)

Operating conditions:

Pressure: $0.1 - 10 \, \text{Torr}$ Temperature: $400 - 900 \, °\text{C}$ Hot-wall reactor design

Advantages:

Better uniformity (longer mean free path) Good step coverage High purity films

Applications:

Polysilicon gates Silicon nitride (Si₃N₄) Thermal oxides

PECVD (Plasma Enhanced CVD)

Additional physics:

Electron impact reactions Ion bombardment Radical chemistry Plasma sheath dynamics

Electron density equation:

$$\frac{\partial n_e}{\partial t} + \nabla \cdot \boldsymbol{\Gamma}_e = S_e$$

Where:

$n_e$ — electron density $\left[\text{m}^{-3}\right]$ $\boldsymbol{\Gamma}_e$ — electron flux $\left[\text{m}^{-2} \cdot \text{s}^{-1}\right]$ $S_e$ — electron source term (ionization - recombination)

Electron energy distribution:

Often non-Maxwellian, requiring solution of Boltzmann equation or two-temperature models.

Advantages:

Lower deposition temperatures ($200 - 400 \, °\text{C}$) Higher deposition rates Tunable film stress

ALD (Atomic Layer Deposition)

Process characteristics:

Self-limiting surface reactions Sequential precursor pulses Sub-monolayer control

Growth per cycle:

$$\text{GPC} = \frac{\Delta t}{\text{cycle}}$$

Typically: $\text{GPC} \approx 0.5 - 2 \, \text{Å/cycle}$

Surface coverage model:

$$\theta = \theta_{sat} \left(1 - e^{-\sigma J t}\right)$$

Where:

$\theta$ — surface coverage $\left[0 \leq \theta \leq 1\right]$ $\theta_{sat}$ — saturation coverage $\sigma$ — reaction cross-section $\left[\text{m}^2\right]$ $t$ — exposure time $\left[\text{s}\right]$

Applications:

High-k gate dielectrics (HfO₂, ZrO₂) Barrier layers (TaN, TiN) Conformal coatings in 3D structures

MOCVD (Metal-Organic CVD)

Precursors:

Metal-organic compounds (e.g., TMGa, TMAl, TMIn) Hydrides (AsH₃, PH₃, NH₃)

Key challenges:

Parasitic gas-phase reactions Particle formation Precise composition control

Applications:

III-V semiconductors (GaAs, InP, GaN) LEDs and laser diodes High-electron-mobility transistors (HEMTs)

Step Coverage Modeling

Definition

Step coverage (SC):

$$SC = \frac{t_{bottom}}{t_{top}} \times 100\%$$

Where:

$t_{bottom}$ — film thickness at feature bottom $t_{top}$ — film thickness at feature top

Aspect ratio (AR):

$$AR = \frac{H}{W}$$

Where:

$H$ — feature depth $W$ — feature width

Ballistic Transport Model

For molecular flow in features ($Kn > 1$):

View factor approach:

$$F_{i \rightarrow j} = \frac{A_j \cos\theta_i \cos\theta_j}{\pi r_{ij}^2}$$

Flux balance at surface element:

$$J_i = J_{direct} + \sum_j (1-s) J_j F_{j \rightarrow i}$$

Where:

$s$ — sticking coefficient $(1-s)$ — re-emission probability

Step Coverage Dependencies

Sticking coefficient effect:

$$SC \approx \frac{1}{1 + \frac{s \cdot AR}{2}}$$

Key observations:

Low $s$ → better step coverage High AR → poorer step coverage ALD achieves ~100% SC due to self-limiting chemistry

Aspect Ratio Dependent Deposition (ARDD)

Local loading effect:

Reactant depletion in features Aspect ratio dependent etch (ARDE) analog

Modeling approach:

$$R_{dep}(z) = R_0 \cdot \frac{C(z)}{C_0}$$

Where:

$z$ — depth into feature $C(z)$ — local concentration (decreases with depth)

Thermal Modeling

Heat Transfer Mechanisms

Conduction (Fourier's law):

$$\mathbf{q}_{cond} = -k \nabla T$$

Convection:

$$q_{conv} = h (T_s - T_{\infty})$$

Where:

$h$ — heat transfer coefficient $\left[\text{W/m}^2 \cdot \text{K}\right]$

Radiation (Stefan-Boltzmann):

$$q_{rad} = \varepsilon \sigma (T_s^4 - T_{surr}^4)$$

Where:

$\varepsilon$ — emissivity $\left[0 \leq \varepsilon \leq 1\right]$ $\sigma$ — Stefan-Boltzmann constant $= 5.67 \times 10^{-8} \, \text{W/m}^2 \cdot \text{K}^4$

Wafer Temperature Uniformity

Temperature non-uniformity impact:

For reaction-limited regime:

$$\frac{\Delta R}{R} \approx \frac{E_a}{R T^2} \Delta T$$

Example calculation:

For $E_a = 1.5 \, \text{eV}$, $T = 900 \, \text{K}$, $\Delta T = 5 \, \text{K}$:

$$\frac{\Delta R}{R} \approx \frac{1.5 \times 1.6 \times 10^{-19}}{1.38 \times 10^{-23} \times (900)^2} \times 5 \approx 10.7\%$$

Susceptor Design Considerations

Material: SiC, graphite, quartz Heating: Resistive, inductive, lamp (RTP) Rotation: Improves azimuthal uniformity Edge effects: Guard rings, pocket design

Validation and Calibration

Experimental Characterization Techniques

TechniqueMeasurementResolution
EllipsometryThickness, optical constants~0.1 nm
XRFComposition, thickness~1%
RBSComposition, depth profile~10 nm
SIMSTrace impuritiesppb
AFMSurface morphology~0.1 nm (z)
SEM/TEMCross-section profile~1 nm
XRDCrystallinity, stress—

Model Calibration Approach

Parameter estimation:

Minimize objective function:

$$\chi^2 = \sum_i \left( \frac{y_i^{exp} - y_i^{model}}{\sigma_i} \right)^2$$

Where:

$y_i^{exp}$ — experimental measurement $y_i^{model}$ — model prediction $\sigma_i$ — measurement uncertainty

Sensitivity analysis:

$$S_{ij} = \frac{\partial y_i}{\partial p_j} \cdot \frac{p_j}{y_i}$$

Where:

$S_{ij}$ — normalized sensitivity of output $i$ to parameter $j$ $p_j$ — model parameter

Uncertainty Quantification

Parameter uncertainty propagation:

$$\text{Var}(y) = \sum_j \left( \frac{\partial y}{\partial p_j} \right)^2 \text{Var}(p_j)$$

Monte Carlo approach:

Sample parameter distributions Run multiple model evaluations Statistical analysis of outputs

CVD Reactor Geometries and Flow Patterns Four canonical single-wafer and batch configurations with dominant transport mechanisms Showerhead Reactor Wafer Heated susceptor Quasi-1D axial flow Da-tunable uniformity Cross-Flow Reactor Wafer Gas in Out Precursor depletion along flow direction Rotating-Disk Reactor ω rotation von Kármán BL δ = √(ν/Ω) LPCVD Tube Furnace wafers Gas Batch, hot-wall Knudsen diffusion Reactor Comparison Matrix Parameter Showerhead Cross-Flow Rotating Disk Tube Furnace Flow symmetry Axisymmetric 2D lateral Axisymmetric Quasi-1D axial Pressure range 0.1–760 Torr ~760 Torr 10–760 Torr 0.1–2 Torr Throughput Single wafer Single wafer Single wafer Batch 25-150 Key model Stagnation flow Depletion integral von Kármán BL Plug + Knudsen Uniformity lever Gap, hole pattern Tilt, flow rate Rotation speed Spacing, T profile Convection risk Low (top-down) High (buoyancy) Suppressed by ω Minimal at LPCVD

The Navier-Stokes equations govern momentum transport in the reactor and determine the velocity field through which precursors travel. The momentum equation $\rho (\partial \mathbf{v}/\partial t + \mathbf{v} \cdot \nabla \mathbf{v}) = -\nabla p + \nabla \cdot \boldsymbol{\tau} + \rho \mathbf{g}$ includes a gravitational body force that can drive natural convection when temperature gradients create density differences. The Grashof number $Gr = g \beta \Delta T L^3 / \nu^2$ quantifies buoyancy relative to viscous forces, and Evans and Greif demonstrated that when $Gr/Re^2 > 1$ in horizontal reactors, buoyancy-driven recirculation rolls degrade uniformity, motivating top-down showerhead geometries.

The energy equation couples to momentum through temperature-dependent density and to chemistry through reaction enthalpies. The general form $\rho c_p (\partial T / \partial t + \mathbf{v} \cdot \nabla T) = \nabla \cdot (k \nabla T) + Q_{rxn} + Q_{rad}$ includes heat from gas-phase reactions and radiative transfer. In hot-wall LPCVD furnaces, radiation between wafers, boat, and tube wall can be significant; in cold-wall single-wafer reactors, steep temperature gradients exist between the hot wafer and the cooled chamber walls. Many CVD gases are optically thin, so radiation must be treated as surface-to-surface exchange using view factors rather than through continuum approximations.

Species transport carries precursor from the inlet to the wafer surface through the conservation equation $\partial C_i / \partial t + \nabla \cdot (C_i \mathbf{v}) = \nabla \cdot (D_i \nabla C_i) + R_i$. In multicomponent mixtures the binary Fickian approximation breaks down and the Stefan-Maxwell equations $\nabla x_i = \sum_{j \neq i} x_i x_j ({\mathbf{v}_j - \mathbf{v}_i})/{D_{ij}}$ must be solved, with binary diffusion coefficients estimated from Chapman-Enskog theory. Coltrin, Kee, and Rupley at Sandia implemented multicomponent transport in the CHEMKIN framework that became the standard tool for CVD gas-phase modeling.

The boundary layer between the bulk gas and the wafer surface is where transport and reaction compete most intensely. In a stagnation-flow showerhead reactor $\delta \sim \sqrt{\nu L / v_0}$; in a rotating-disk reactor $\delta \sim \sqrt{\nu / \Omega}$. The Sherwood number $Sh = k_m L / D$ characterizes convective mass transfer efficiency, and for laminar stagnation flow $Sh \approx 0.62 Re^{1/2} Sc^{1/3}$, connecting deposition rate to the dimensionless groups that define the flow state.

Gas-phase chemistry transforms precursor molecules into reactive intermediates before they reach the surface. The primary silane decomposition $\text{SiH}_4 \rightarrow \text{SiH}_2 + \text{H}_2$ produces silylene, which inserts into other silane molecules to form disilane and higher oligomers. Ho, Breiland, and Coltrin at Sandia showed that $\text{SiH}_2$ is the dominant growth precursor in LPCVD, not intact $\text{SiH}_4$. Each elementary reaction is parameterized by the Arrhenius rate expression $k(T) = A T^n \exp(-E_a / (R T))$, and the net production rate sums over all reactions: $R_i = \sum_{r=1}^{N_r} \nu_{i,r} k_r \prod_{j=1}^{N_s} C_j^{\alpha_{j,r}}$.

Gas-Phase and Surface Reaction Pathways in Silane CVD From precursor decomposition through intermediate species to film incorporation GAS PHASE (bulk + boundary layer) SiH₄ k₁, Δ SiH₂ +SiH₄ Si₂H₆ dominant precursor H₂ Si₃H₈, ... particles! minor SURFACE REACTIONS Adsorption SiH₂ → SiH₂(ads) s = sticking coeff Surface Diffusion Migration to step edge or kink site Incorporation Si lattice bond + H₂ desorption Desorption Weakly-bound species leave DEPOSITED FILM (Si, SiO₂, Si₃N₄, W, ...) Langmuir-Hinshelwood: Rₛ = kₛKₐKᵦCₐCᵦ / (1+KₐCₐ+KᵦCᵦ)² Eley-Rideal: Rₛ = kₛθₐCᵦ θ = fractional coverage

Surface reaction kinetics determine the actual film growth rate and are the hardest part of the model to parameterize from first principles. The Langmuir-Hinshelwood mechanism gives $R_s = k_s K_A K_B C_A C_B / (1 + K_A C_A + K_B C_B)^2$, while the Eley-Rideal mechanism gives $R_s = k_s \theta_A C_B$. The sticking coefficient $s$ encodes all surface physics into a single number: Gates, Kulkarni, and Scott showed that for TEOS-based oxide deposition, $s$ drops by orders of magnitude below 300 degrees C, explaining why TEOS gives excellent step coverage at low temperatures where precursor diffuses deep into features before reacting.

The local film growth rate connects surface reaction flux to thickness as $dh/dt = M_w R_s / \rho_{film}$. When reaction-limited ($Da \ll 1$), the rate is exponentially sensitive to temperature: Jensen quantified this as $\delta R / R = (E_a / (R T^2)) \delta T$, meaning a 1 degree C non-uniformity at 700 degrees C in LPCVD polysilicon with $E_a \approx 1.5$ eV produces roughly 1.8% thickness non-uniformity. When transport-limited ($Da \gg 1$), the rate is controlled by the mass transfer coefficient, which depends on flow patterns and diffusion coefficients rather than on temperature.

Precursor depletion along the flow direction is the dominant source of non-uniformity in cross-flow and tube reactors. The concentration drops as $C(x) = C_0 \exp(-k_s W x / Q)$, and Hitchman and Jensen showed that axial depletion in LPCVD tube furnaces can produce 10-20% thickness variation unless a temperature-tilt strategy compensates by running downstream zones hotter to offset lower precursor concentration.

The showerhead is a gas distribution device whose modeling requires fluid mechanics at two scales. At the macro scale, the pressure drop through individual holes follows $\Delta P = \rho v^2 / (2 C_d^2)$, and a well-designed showerhead achieves a uniformity index above 0.98. At the micro scale, gas jets must merge into uniform flow before reaching the wafer, and the showerhead-to-wafer gap controls the merging. Natural convection threatens uniformity in atmospheric-pressure CVD when the mixed-convection parameter $Gr/Re^2$ exceeds unity, creating buoyancy-driven recirculation cells; Moffat and Jensen showed that critical Rayleigh numbers for this transition depend on aspect ratio and temperature difference. LPCVD largely avoids this problem because at sub-Torr pressures buoyancy forces are negligible.

The Knudsen number $Kn = \lambda / L$ determines whether the continuum Navier-Stokes equations are valid. The mean free path $\lambda = k_B T / (\sqrt{2} \pi d^2 P)$ is about 0.1 $\mu$m at atmospheric pressure and 500 degrees C but increases to 0.5 mm at 0.1 Torr, where slip corrections become necessary. Inside high-aspect-ratio features at low pressure, the local Knudsen number can exceed unity, pushing transport into the free-molecular regime where Knudsen diffusion replaces Fickian diffusion.

Dimensionless NumberDefinitionPhysical MeaningTypical CVD RangeImpact on Model Choice
Damköhler ($Da$)$k_s L / D$reaction rate / diffusion rate$10^{-2}$ to $10^2$determines rate-limiting step
Reynolds ($Re$)$\rho v L / \mu$inertial / viscous forces1 to 100laminar flow assumed
Grashof ($Gr$)$g \beta \Delta T L^3 / \nu^2$buoyancy / viscous forces$10^0$ to $10^6$convection cell risk
Péclet ($Pe$)$v L / D$convection / diffusion1 to 50advection vs diffusion
Knudsen ($Kn$)$\lambda / L$mean free path / length scale$10^{-5}$ to $10^1$continuum vs rarefied
Schmidt ($Sc$)$\nu / D$momentum / mass diffusivity0.2 to 2BL thickness ratio
Prandtl ($Pr$)$\mu c_p / k$momentum / thermal diffusivity0.5 to 1thermal BL shape
Thiele ($\phi$)$L \sqrt{k_s / D_{Kn}}$reaction / pore diffusion$10^{-1}$ to $10^2$step coverage quality

Feature-scale modeling addresses what happens inside the trench, via, or high-aspect-ratio hole where reactor-scale models cannot resolve the geometry. The Thiele modulus $\phi = L \sqrt{k_s / D_{Kn}}$ compares feature depth to the diffusion-reaction length. When $\phi \ll 1$ the step coverage is conformal; when $\phi \gg 1$ bread-loafing or keyhole formation occurs. Knudsen diffusion $D_{Kn} = (d_{feature}/3) \sqrt{8 R T / (\pi M)}$ governs transport inside features where the mean free path exceeds the feature width, and the coefficient decreases linearly with width, which is why high-aspect-ratio structures present extreme step-coverage challenges.

Feature-Scale Deposition: Step Coverage vs Thiele Modulus Conformal, moderate, and non-conformal profiles in high-aspect-ratio trenches φ ≪ 1 (Conformal) Reaction-limited regime SC ≈ 95-100% LPCVD, ALD φ ≈ 1 (Moderate) Transition regime SC ≈ 50-80% PECVD typical φ ≫ 1 (Non-conformal) Transport-limited regime VOID SC ≈ 0-30% APCVD, fast PECVD Knudsen diffusion: D_Kn = (d/3)√(8RT/πM) Thiele modulus: φ = L√(kₛ/D_Kn) Step coverage ≈ 1/cosh(φ) for first-order kinetics in a rectangular trench
PECVD Plasma Physics: From Electron Kinetics to Film Deposition Sheath structure, electron-impact chemistry, and ion bombardment effects RF POWERED ELECTRODE (showerhead) SHEATH (ion acceleration zone) BULK PLASMA Quasi-neutral: nₑ ≈ nᵢ ≈ 10⁹-10¹¹ cm⁻³ Tₑ = 2-5 eV, Tᵢ ≈ T_gas ≈ 0.03 eV e⁻ e⁻ e⁻ e⁻ Ar⁺ SiH₃⁺ SHEATH (V_sheath ~ 10-500 V) GROUNDED ELECTRODE (wafer on susceptor) Electron-Impact Reactions in SiH₄/N₂O PECVD Dissociation: e + SiH₄ → SiH₃ + H + e e + SiH₄ → SiH₂ + H₂ + e e + N₂O → N₂ + O + e Ionization: e + SiH₄ → SiH₃⁺ + H + 2e e + Ar → Ar⁺ + 2e Excitation: e + N₂ → N₂* + e e + Ar → Ar* + e (metastable) kₑ = ∫σ(ε)√(2ε/mₑ) f(ε) dε Key Sheath Relations Bohm velocity: v_B = √(kT_e / m_i) Child-Langmuir: J = (4ε₀/9)√(2e/m_i)·V^(3/2)/d² Floating potential: V_f = -(T_e/2e)·ln(m_i/2πm_e) Ion energy at substrate ≈ V_plasma - V_substrate Ion Bombardment Effects on PECVD Film Properties ↑ Density ↓ H content Stress control ↑ Damage risk ↓ Wet etch rate
Wafer Temperature Uniformity in Cold-Wall CVD Reactor Heat transfer mechanisms, edge effects, and multi-zone compensation Reactor Cross-Section (Cold-Wall Design) Cold wall (water-cooled) Showerhead (cooled) Process gas Boundary layer δ Wafer (300mm) Backside He/Ar gap (5-20 Torr) Edge Center zone Edge Multi-zone resistive heater (susceptor) Radiation to cold wall Wafer Temperature Profile Temperature Radial position Edge Center Edge Uncompensated Compensated ΔT Heat Flux to Wafer q = h_conv(T_susc - T_w) + εσ(T⁴_susc - T⁴_w) h_conv depends on backside gas, pressure, gap Temperature Sensitivity in Reaction-Limited CVD Sensitivity formula: δR/R = (E_a / RT²) · δT LPCVD poly-Si (E_a≈1.6eV) ~1.8%/°C at 620°C LPCVD Si₃N₄ (E_a≈1.8eV) ~1.5%/°C at 780°C PECVD SiO₂ (E_a≈0.3eV) ~0.3%/°C at 400°C (less sensitive)
Advanced CVD Process Modeling: Regime-Specific Challenges From thermal equilibrium to plasma-assisted and self-limiting deposition Thermal LPCVD Reaction-limited, batch Key equation: R = A·exp(-E_a/RT)·C_surface Model focus: • T uniformity across wafer boat • Axial depletion C(x)=C₀e^(-kWx/Q) • Wafer-to-wafer loading effect Jensen, Hitchman (tube models) Coltrin, Kee (CHEMKIN kinetics) PECVD / HDP-CVD Plasma-assisted, single-wafer Key equations: Boltzmann EEDF + Navier-Stokes Model focus: • Electron kinetics → radical generation • Ion energy/angular distributions • Dep + etch balance (HDP) Lieberman, Lichtenberg (plasmas) Godyak, Piejak (EEDF meas.) ALD (Thermal + Plasma) Self-limiting, atomic control Key equation: θ(t) = θ_sat(1-e^(-k_ads·p·t)) Model focus: • Saturation dose vs AR • Purge efficiency, dead volumes • Nucleation delay on surfaces George (ALD framework) Puurunen (nucleation review) Model Validation: Experimental Diagnostics for CVD In-situ Spectroscopy FTIR, Raman, OES Laser Diagnostics LIF, TDLAS, CARS Film Characterization Ellipsometry, XRR, XPS RGA / QMS Gas composition SEM / TEM / AFM Feature profiles Critical Modeling Challenges at Advanced Nodes 3D NAND (AR > 100:1) Precursor penetration <1% flux at bottom GAA Transistors Conformal wrap-around Multi-material stacks Area-Selective Deposition Nucleation modeling Selectivity loss prediction Backside Power Delivery Through-wafer vias Conformal fill required

The quantity of interest determines the minimum credible model. Radial thickness, chamber matching, feature conformality, and particle risk require different state variables and resolution, so the decision and error tolerance must be stated before equations are selected.

Every scale handoff needs an explicit physical contract. Reactor models should pass temperature and resolved species fluxes to surface or feature models with units, averaging interval, angular information where needed, and uncertainty rather than passing an unexplained scalar rate.

Elemental and site balances are stronger checks than attractive contours. Integrated inlet, outlet, wall loss, solid incorporation, and accumulation must close for every conserved element, while adsorbate fractions and vacant sites must sum to available surface sites.

Reaction mechanisms should be reduced against target predictions. Reaction-path analysis and sensitivity tests can remove expensive species only after growth rate, composition, depletion, and particle precursors remain accurate across the claimed recipe window.

Residence-time distributions expose chemistry hidden by average flow. Recirculation, bypass, and dead zones give molecules different thermal histories, so tracer transients and age-of-fluid fields constrain decomposition better than nominal chamber volume divided by flow.

Particle models must couple birth, growth, forces, and wall interaction. Nucleation alone cannot predict contamination because thermophoresis, drag, gravity, charging, coagulation, pumping, and sticking determine whether a cluster reaches wafer or wall.

Calibration cannot identify parameters that move predictions identically. Arrhenius prefactor and activation energy, sticking and mass transfer, or wall loss and homogeneous consumption can be correlated, requiring mechanism-separating experiments and confidence intervals.

Validation must use evidence withheld from parameter fitting. A new pressure, temperature, wafer loading, reactor spacing, feature aspect ratio, or chamber state is stronger than withholding nearby points from the same recipe.

Uncertainty must propagate through the full hierarchy. Flow calibration, geometry, temperature, transport data, kinetic rates, wall state, numerical error, and model discrepancy should reach prediction intervals for thickness, composition, conformality, and defects.

Measurements require their own forward models. Ellipsometry, optical emission, mass spectrometry, XRF, SEM, and endpoint traces average space and time differently, so simulation should be compared with instrument response rather than an imagined exact state.

Structured residuals reveal which physics is missing. Radial error suggests thermal or delivery fields, loading dependence suggests depletion, feature-depth error suggests molecular transport, and wafer-sequence drift suggests wall state.

Surrogate models must advertise their validity domain. Gaussian processes, reduced bases, and neural networks require distance-to-training checks, physical constraints where available, and fallback to the verified high-fidelity model outside their trusted region.

Reproducibility is part of model credibility. Geometry, properties, chemistry, boundaries, mesh, tolerances, calibration data, validation data, and scripts should be versioned so a changed prediction can be traced to a changed assumption.

Modeling claimMinimum physicsCalibration evidenceWithheld validation
Blanket growth rateSurface kinetics and wafer temperatureRate versus temperature and partial pressureNew pressure or carrier gas
Radial uniformityFlow, heat, species, and surface sinkThickness and temperature mapsChanged spacing or rotation
Batch depletionAxial transport and distributed consumptionWafer-position and load-size profilesDifferent boat loading
PECVD responseRadical source, sheath inputs, surface chemistryPlasma diagnostics and film propertiesIndependent source-bias split
MOCVD compositionSpecies-specific gas and surface mechanismThickness and composition mapsChanged precursor ratio
Feature conformalityMolecular transport, sticking, saturation, moving wallCross sections over aspect ratioNew feature geometry
Particle riskNucleation, size evolution, forces, wall lossParticle monitor and deposit mapsChanged thermal gradient
Chamber matchingAs-built geometry, boundaries, wall stateMatched sensor and wafer datasetsPost-maintenance wafer sequence
start: Define the decision and quantity of interest
scale: Choose reactor boundary layer feature surface or coupled scales
balances: Close mass elements energy sites and charge where applicable
regime: Evaluate Reynolds Peclet Damkohler and Knudsen regimes
inputs: Version geometry properties chemistry and boundary conditions
verify: Verify balances mesh time step and benchmark cases
identify: Test sensitivity correlation and identifiability
calibrate: Calibrate only identifiable parameters
validate: Predict a withheld mechanism-sensitive condition
residual: Are residuals unstructured and within acceptance limits?
deploy: Propagate uncertainty and guard the validity envelope
classify: Classify residuals by radius loading temperature feature and sequence
revise: Replace the falsified mechanism
start->scale->balances->regime->inputs->verify->identify->calibrate->validate->residual
residual->deploy
residual->classify
classify->revise
revise->verify

A closure test should predict a condition the model has never seen. Specify the expected direction and tolerance for a new temperature, loading, pressure, geometry, or chamber state before running it; success supports transportability, while failure identifies the next falsified assumption.

CVD modeling becomes trustworthy when conservation, calibration, and prediction agree across scales. Reactor flow and heat determine chemical histories, surface state converts those histories into incorporation, feature transport converts incident flux into conformality, and measurement models connect predictions to observations with stated uncertainty. Read CVD modeling through a conservation-and-validation lens rather than a contour-generation lens.

cvd modelingchemical vapor depositioncvd processlpcvdpecvdhdp-cvdmocvdaldthin film depositioncvd equipmentcvd simulation

Related Topics

Cvd Equipment ModelingCvd Process Modeling

Explore 500+ Semiconductor & AI Topics

From EUV lithography to CUDA optimization — search the full knowledge base or chat with our AI assistant.

🔍 Search Topics 💬 Ask CFSGPT 📚 Browse All