Project Introduction:

The terahertz (THz) band offers great potential for high-speed communication and sensing, yet receiver development is limited by electronic architectures suffering from bandwidth, latency, and power constraints. Conventional THz receivers depend on frequency down-conversion and digital processing, which hinder real-time performance. Advances in diffractive optical neural networks and analog optical computing provide a compelling alternative, enabling light-based computation at propagation speed. Using artificial intelligence (AI) to design such architectures allows complex inference and signal recognition to be performed passively, offering a pathway toward ultrafast, energy-efficient THz detection and demodulation.

Project Description:

This project develops AI-enhanced diffractive architectures for real-time THz signal recognition. It focuses on designing diffractive neural networks that map spatial and temporal features of incident THz waves directly to symbol or modulation classifications without digital processing. Deep-learning optimization will train the diffractive layers, while analytical modeling and full-wave simulations will assess performance under realistic propagation and noise conditions. The study will also examine feasible implementation using reconfigurable metasurfaces or THz-compatible materials and evaluate fabrication tolerances. Ultimately, it aims to establish a foundation for analog neural receivers that merge photonic speed with AI intelligence for next-generation THz communication systems.

Objectives:

• Develop analytical and computational models for THz wave propagation through multi-layer diffractive neural networks, incorporating realistic parameters such as noise, diffraction limits, and material dispersion.

• Design and train AI-optimized diffractive architectures to classify and decode THz signal patterns (e.g., modulation symbols or waveforms) in real time via passive optical inference.

• Evaluate performance in terms of recognition accuracy, SNR robustness, bandwidth tolerance, and latency, and benchmark against conventional digital THz receivers.

• Assess implementation feasibility using available diffractive materials or programmable metasurfaces, including tolerance to fabrication and alignment errors.

• Demonstrate potential real-time operation through simulation or experimental validation, and establish design guidelines for scalable analog computing architectures for THz communication and sensing.

Application Deadline

9/11/2025

Interview Date

17/11/2025

Entry Date

January/April 2026