Device Physics & Mathematical Modeling

Keywords: device physics mathematics,device physics math,semiconductor device physics,TCAD modeling,drift diffusion,poisson equation,mosfet physics,quantum effects

Device Physics & Mathematical Modeling

1. Fundamental Mathematical Structure

Semiconductor modeling is built on coupled nonlinear partial differential equations spanning multiple scales:

Multiscale Hierarchy

The mathematics forms a hierarchy of models through successive averaging:

2. Process Physics & Models

2.1 Oxidation: Deal-Grove Model

Thermal oxidation of silicon follows linear-parabolic kinetics :

where:

• $x_{ox}$ = oxide thickness
• $B/A$ = linear rate constant (surface-reaction limited)
• $B$ = parabolic rate constant (diffusion limited)

Limiting Cases:

• Thin oxide (reaction-limited):

• Thick oxide (diffusion-limited):

Physical Mechanism:

1. O₂ transport from gas to oxide surface 2. O₂ diffusion through growing SiO₂ layer 3. Reaction at Si/SiO₂ interface: $\text{Si} + \text{O}_2 \rightarrow \text{SiO}_2$

Note: This is a Stefan problem (moving boundary PDE).

2.2 Diffusion: Fick's Laws

Dopant redistribution follows Fick's second law :

For constant $D$ in 1D:

Analytical Solutions (1D, constant D):

• Constant surface concentration (infinite source):

• Limited source (e.g., implant drive-in):

where $Q$ = dose (atoms/cm²)

Complications at High Concentrations:

• Concentration-dependent diffusivity: $D = D(C)$
• Electric field effects: Charged point defects create internal fields
• Vacancy/interstitial mechanisms: Different diffusion pathways

2.3 Ion Implantation: Range Theory

The implanted dopant profile is approximately Gaussian :

where:

• $\Phi$ = implant dose (ions/cm²)
• $R_p$ = projected range (mean depth)
• $\Delta R_p$ = straggle (standard deviation)

LSS Theory (Lindhard-Scharff-Schiøtt) predicts stopping power:

where:

• $S_n(E)$ = nuclear stopping power (dominant at low energy)
• $S_e(E)$ = electronic stopping power (dominant at high energy)
• $N$ = target atomic density

For asymmetric profiles , the Pearson IV distribution is used:

Modern approach: Monte Carlo codes (SRIM/TRIM) for accurate profiles including channeling effects.

2.4 Lithography: Optical Imaging

Aerial image formation follows Hopkins' partially coherent imaging theory :

where:

• $TCC$ = Transmission Cross-Coefficient
• $\tilde{M}(f)$ = mask spectrum (Fourier transform of mask pattern)
• $\mathbf{r}$ = position in image plane

Fundamental Limits:

• Rayleigh resolution criterion:

• Depth of focus:

where:

• $\lambda$ = wavelength (193 nm for ArF, 13.5 nm for EUV)
• $NA$ = numerical aperture
• $k_1, k_2$ = process-dependent factors

Resist Modeling — Dill Equations:

where $M$ = photoactive compound concentration.

2.5 Etching & Deposition: Surface Evolution

Topography evolution is modeled with the level set method :

where:

• $\phi(\mathbf{r}, t) = 0$ defines the surface
• $V$ = local velocity (etch rate or deposition rate)

For anisotropic etching:

CVD in High Aspect Ratio Features:

Knudsen diffusion limits step coverage:

where:

• $D_K = \frac{d}{3}\sqrt{\frac{8k_BT}{\pi m}}$ (Knudsen diffusivity)
• $d$ = feature width
• $k_s$ = surface reaction rate

ALD (Atomic Layer Deposition):

Self-limiting surface reactions follow Langmuir kinetics:

where $\theta$ = surface coverage, $P$ = precursor partial pressure.

3. Device Physics: Semiconductor Equations

The core mathematical framework for device simulation consists of three coupled PDEs :

3.1 Poisson's Equation (Electrostatics)

where:

• $\psi$ = electrostatic potential
• $n, p$ = electron and hole concentrations
• $N_D^+, N_A^-$ = ionized donor and acceptor concentrations

3.2 Continuity Equations (Carrier Conservation)

Electrons:

Holes:

where:

• $G$ = generation rate
• $R$ = recombination rate

3.3 Current Density Equations (Transport)

Drift-Diffusion Model:

Einstein Relation:

3.4 Recombination Models

Shockley-Read-Hall (SRH) Recombination:

Auger Recombination:

Radiative Recombination:

3.5 MOSFET Physics

Threshold Voltage:

where:

• $V_{FB}$ = flat-band voltage
• $\phi_B = \frac{k_BT}{q} \ln\left(\frac{N_A}{n_i}\right)$ = bulk potential
• $C_{ox} = \frac{\varepsilon_{ox}}{t_{ox}}$ = oxide capacitance

Drain Current (Gradual Channel Approximation):

• Linear region ($V_{DS} < V_{GS} - V_T$):

• Saturation region ($V_{DS} \geq V_{GS} - V_T$):

4. Quantum Effects at Nanoscale

For modern devices with gate lengths $L_g < 10$ nm, classical models fail.

4.1 Quantum Confinement

In thin silicon channels, carrier energy becomes quantized :

where:

• $n$ = quantum number (1, 2, 3, ...)
• $m^*$ = effective mass
• $t_{Si}$ = silicon body thickness

Effects:

• Increased threshold voltage
• Modified density of states: $g_{2D}(E) = \frac{m^*}{\pi \hbar^2}$ (step function)

4.2 Quantum Tunneling

Gate Leakage (Direct Tunneling):

WKB approximation:

where $\kappa = \sqrt{\frac{2m^*(\Phi_B - E)}{\hbar^2}}$

Source-Drain Tunneling:

Limits OFF-state current in ultra-short channels.

Band-to-Band Tunneling:

Enables Tunnel FETs (TFETs):

4.3 Ballistic Transport

When channel length $L < \lambda_{mfp}$ (mean free path), the Landauer formalism applies:

where:

• $T(E)$ = transmission probability
• $f_S, f_D$ = source and drain Fermi functions

Ballistic Conductance Quantum:

4.4 NEGF Formalism

The Non-Equilibrium Green's Function method is the gold standard for quantum transport:

where:

• $H$ = device Hamiltonian
• $\Sigma_1, \Sigma_2$ = contact self-energies
• $G^R$ = retarded Green's function

Observables:

• Electron density: $n(\mathbf{r}) = -\frac{1}{\pi} \text{Im}[G^<(\mathbf{r}, \mathbf{r}; E)]$
• Current: $I = \frac{q}{h} \text{Tr}[\Gamma_1 G^R \Gamma_2 G^A]$

5. Numerical Methods

5.1 Discretization: Scharfetter-Gummel Scheme

The drift-diffusion current requires special treatment to avoid numerical instability:

where the Bernoulli function is:

Properties:

• $B(0) = 1$
• $B(x) \to 0$ as $x \to \infty$
• $B(-x) = x + B(x)$

5.2 Solution Strategies

Gummel Iteration (Decoupled):

1. Solve Poisson for $\psi$ (fixed $n$, $p$) 2. Solve electron continuity for $n$ (fixed $\psi$, $p$) 3. Solve hole continuity for $p$ (fixed $\psi$, $n$) 4. Repeat until convergence

Newton-Raphson (Fully Coupled):

Solve the Jacobian system:

5.3 Time Integration

Stiffness Problem:

Time scales span ~15 orders of magnitude:

Solution: Use implicit methods (Backward Euler, BDF).

5.4 Mesh Requirements

Debye Length Constraint:

The mesh must resolve the Debye length:

For $n = 10^{18}$ cm⁻³: $\lambda_D \approx 4$ nm

Adaptive Mesh Refinement:

• Refine near junctions, interfaces, corners
• Coarsen in bulk regions
• Use Delaunay triangulation for quality

6. Compact Models for Circuit Simulation

For SPICE-level simulation, physics is abstracted into algebraic/empirical equations.

Industry Standard Models

Key Physics Captured

• Short-channel effects: DIBL, $V_T$ roll-off
• Quantum corrections: Inversion layer quantization
• Mobility degradation: Surface scattering, velocity saturation
• Parasitic effects: Series resistance, overlap capacitance
• Variability: Statistical mismatch models

Threshold Voltage Variability (Pelgrom's Law)

where $A_{VT}$ is a technology-dependent constant.

7. TCAD Co-Simulation Workflow

The complete semiconductor design flow:

┌─────────────────────────────────────────────────────────────┐
│  ┌───────────────┐   ┌───────────────┐   ┌───────────────┐  │
│  │   Process     │──▶│    Device     │──▶│   Parameter   │  │
│  │  Simulation   │   │  Simulation   │   │  Extraction   │  │
│  │  (Sentaurus)  │   │  (Sentaurus)  │   │ (BSIM Fit)    │  │
│  └───────────────┘   └───────────────┘   └───────────────┘  │
│         │                   │                   │           │
│         ▼                   ▼                   ▼           │
│  ┌───────────────┐   ┌───────────────┐   ┌───────────────┐  │
│  │• Implantation │   │• I-V, C-V     │   │• BSIM params  │  │
│  │• Diffusion    │   │• Breakdown    │   │• Corner extr. │  │
│  │• Oxidation    │   │• Hot carrier  │   │• Variability  │  │
│  │• Etching      │   │• Noise        │   │  statistics   │  │
│  └───────────────┘   └───────────────┘   └───────────────┘  │
│                                                │            │
│                                                ▼            │
│                                         ┌───────────────┐   │
│                                         │    Circuit    │   │
│                                         │  Simulation   │   │
│                                         │(SPICE,Spectre)│   │
│                                         └───────────────┘   │
└─────────────────────────────────────────────────────────────┘

Key Challenge: Propagating variability through the entire chain:

• Line Edge Roughness (LER)
• Random Dopant Fluctuation (RDF)
• Work function variation
• Thickness variations

8. Mathematical Frontiers

8.1 Machine Learning + Physics

• Physics-Informed Neural Networks (PINNs):

where $\mathcal{L}_{physics}$ enforces PDE residuals.

• Surrogate models for expensive TCAD simulations
• Inverse design and topology optimization
• Defect prediction in manufacturing

8.2 Stochastic Modeling

Random Dopant Fluctuation:

Approaches:

• Atomistic Monte Carlo (place individual dopants)
• Statistical impedance field method
• Compact model statistical extensions

8.3 Multiphysics Coupling

Electro-Thermal Self-Heating:

Stress Effects on Mobility (Piezoresistance):

Electromigration in Interconnects:

8.4 Atomistic-Continuum Bridging

Strategies:

• Coarse-graining from MD/DFT
• Density gradient quantum corrections:

• Hybrid methods: atomistic core + continuum far-field

The mathematics of semiconductor manufacturing and device physics encompasses:

Each level trades accuracy for computational tractability . The art lies in knowing when each approximation breaks down—and modern scaling is pushing us toward the quantum limit where classical continuum models become inadequate.

Source: ChipFoundryServices — Search this topic — Ask CFSGPT