Flexcompute-Hackathon

Physics-informed inverse design of integrated photonic filters using coupled waveguides. Maps arbitrary target spectra to realizable circuits via transfer-matrix modeling and ML optimization.

Physics-Informed Design of Optical Filters using Coupled Waveguides

This repository contains a complete pipeline for designing integrated photonic filters using a combination of:

• Waveguide mode simulations
• Supermode-based coupling extraction
• Physics-informed compact modeling
• AI-assisted optimization (PIGNN)
• Full-wave FDTD validation

---

🚀 Project Overview

We design arbitrary optical filters using cascaded directional couplers. The workflow is structured into three main stages:

1. Waveguide characterization → extract propagation constants (β)
2. Coupling characterization → extract coupling coefficient (κ) and Ω
3. Filter design & validation → compact model + AI + FDTD

---

📁 Repository Structure

waveguide_design/
filter_model_PIGNN/
fdtd_simulation/

---

1️⃣ Waveguide Design

Workflow (in order)

1. Mode Simulation

Run: notebooks/waveguide_mode.ipynb

2. Width Sweep

Run: notebooks/width_sweep.ipynb

3. Clean Width Sweep Data

Run: notebooks/width_sweep_data_cleaner.ipynb

4. Frequency Sweep

Run: notebooks/frequency_sweep.ipynb

5. Clean Frequency Sweep Data

Run: notebooks/frequency_sweep_data_cleaner.ipynb

Outputs:

• β_A(λ)
• β_B(λ)

---

2️⃣ Coupling Coefficient Extraction (κ, Ω)

1. Single Coupling Simulation

notebooks/supermode_coupling.ipynb

2. Gap + Frequency Sweep

notebooks/supermode_coupling_gap_frequency_sweeping.ipynb

3. Data Cleaning

notebooks/coupling_gap_frequency_sweep.ipynb

Outputs:

• Ω(g, λ)
• κ(g, λ)

---

3️⃣ Filter Model + AI (PIGNN)

Folder: filter_model_PIGNN/

Compact Model

notebooks/filter_compact_model.ipynb

Interactive Simulator

scripts/filter_live.py

AI Model

Maps: Target spectrum → {g_i, L_i}

---

4️⃣ FDTD Simulation

Folder: fdtd_simulation/

Geometry Builder

notebooks/fdtd_geometry_builder.ipynb

Single Coupling

notebooks/fdtd_single_coupling.ipynb

Multi Coupling (Example N=4)

notebooks/fdtd_multi_coupling.ipynb

---

📊 Data

Each stage contains:

• data/raw/
• data/processed/

---

🧠 Key Idea

Use physics-informed modeling + AI instead of brute-force FDTD.

---

🛠 Requirements

pip install -r requirements.txt