Silicon Diffusion & Ion Implantation Optimization Using Silvaco TCAD

Author

Amirhossein Zahedi
Electrical Engineering
Micro- and Nanoelectronics

This project is designed by Prof. Bijan Rashidian and his TA Reza Khoddami for the Fabrication Course, Winter 2026.

---

Project Overview

This repository presents a comprehensive TCAD-based analysis of phosphorus diffusion and ion implantation in silicon using Silvaco ATHENA.

The project investigates the trade-offs between:

• Junction depth (Xj)
• Sheet resistance (Rs)
• Surface concentration (Ns)

under various processing conditions, including:

• Pre-deposition time
• Drive-in time
• Temperature
• Ion implantation energy
• Implant dose

The work includes process simulation and sensitivity analysis.

---

Objectives

The main goals of this project are:

1. Simulate constant-source pre-deposition diffusion at 900°C.
2. Perform drive-in diffusion and analyze the time evolution of doping profiles.
3. Extract and analyze:
   • Junction depth (Xj)
   • Sheet resistance (Rs)
   • Surface concentration (Ns)
4. Study scaling effects:
   • Rs → Rs/m (m = 2,3,4,10)
   • Xj → nXj (n = 2,3,4,10)
5. Propose realistic, hard, and impossible device targets.
6. Implement multi-step diffusion to achieve challenging targets.
7. Replace diffusion with ion implantation (≤ 200 keV).
8. Compare diffusion vs implantation approaches.
9. Perform sensitivity analysis.
10. Verify results using published semiconductor processing literature.
11. Study meshing strategies and numerical stability.

---

Initial Conditions

• Substrate: ⟨100⟩ Silicon
• Background doping: Boron 1×10¹⁵ cm⁻³
• Pre-deposition temperature: 900°C
• Dopant: Phosphorus
• Surface concentration set to solid solubility at 900°C

---

Diffusion Modeling

Simulations include:

• Constant-source pre-deposition
• Limited-source drive-in diffusion
• Multi-step diffusion sequences

Time sweeps: 10, 30, 90, 300 minutes

Extracted parameters:

• Xj
• Rs
• Ns

---

Ion Implantation Study

Implantation simulations include:

• Energies up to 200 keV
• Dose optimization
• Post-implant annealing
• Nitride barrier analysis

Secondary effects observed:

• Transient Enhanced Diffusion (TED)
• Dopant clustering
• Channeling
• Concentration-dependent diffusion
• Junction smoothing

---

Sensitivity Analysis

Studied dependence of:

• Xj vs drive-in time
• Rs vs temperature
• Ns vs pre-deposition time

Key observations:

• Xj ∝ √(Dt)
• Rs is inversely related to integrated dopant dose
• Temperature exponentially affects the diffusion coefficient

---

Verification of Results

Simulation results were compared against published processing data for:

• BJT diffusion
• MOS well formation
• Solar cell phosphorus diffusion

Findings:

• Profile shapes matched analytical solutions.
• Absolute values slightly deviated due to:
  • Concentration-dependent diffusivity
  • Solid solubility modeling
  • Mesh resolution
  • Numerical solver tolerances
  • Boundary conditions
  • Clustering models

Silvaco solves nonlinear PDEs using discretization and Newton-based numerical solvers, which introduce realistic physical modeling beyond ideal analytical solutions.

---

Meshing Strategy

Meshing is critical for:

• Accurate Xj extraction
• Correct Rs calculation
• Stability in high-gradient regions

Strategies used:

• Fine mesh near the surface
• Adaptive refinement near junction
• Coarser mesh in bulk
• Controlled aspect ratio

Improper meshing leads to:

• Artificial junction shifts
• Oscillatory concentration profiles
• Convergence failures

---

Engineering Target Analysis

Targets categorized as:

• Possible
• Hard/Expensive
• Impossible

Strategies evaluated:

• Increasing pre-dep time
• Multi-step diffusion
• Switching to implantation
• Barrier layer introduction

Multi-step diffusion provided better thermal budget control.

---

Diffusion vs Implantation Comparison

Feature	Diffusion	Implantation
Profile shape	Error function	Gaussian
Dose control	Indirect	Precise
Thermal budget	High	Lower
Damage	Minimal	Requires anneal
Depth control	Limited	Accurate

Implantation provides improved Xj–Rs trade-off control.

---

Tools Used

• Silvaco ATHENA
• TonyPlot
• MATLAB

---

Key Insights

1. Junction depth scales approximately with √t.
2. Rs does not scale linearly with diffusion time.
3. Increasing Xj while maintaining constant Rs is physically constrained.
4. Solver and meshing strongly influence simulation accuracy.
5. Multi-step diffusion improves manufacturability.

---

How to Run

1. Open DeckBuild.
2. Load input file from /silvaco.
3. Run simulation.
4. Use TonyPlot for extraction.
5. Post-process data in /analysis.

---