New material tackles one of the biggest barriers to scaling artificial intelligence

Photonic chips are crucial for the internet, but they lack the ability to convert, amplify or reshape light signals efficiently.
Generative AI systems increase data traffic, adding pressure to current photonic infrastructure and raising energy costs.
Researchers from Polytechnique Montréal have developed a material that can be added to silicon to allow light processing directly on the chip.
The material, triphenylamine-dicyanoquinoxaline (TPA-QCN), can convert infrared light to visible red light directly on the chip using a process called second-harmonic generation.
The material does not require electric field poling and works at low temperatures and low costs, thus simplifying fabrication.
The approach promises higher efficiency and thermal stability, making it a promising candidate for integrating additional optical functions directly onto photonic chips.
The advancement could lead to reduced heat and fewer conversion steps in data centers, addressing scalability challenges posed by AI systems.
The compatibility with back-end CMOS processing may make scaling this technology more accessible and easier to integrate into existing manufacturing lines.