Russian and Chinese researchers have developed a method for optimizing microdisk lasers that could help advance photonic chips for artificial intelligence systems and data centers, achieving a measured maximum modulation bandwidth of 4.15 gigahertz.
The research was conducted by scientists at the International Laboratory of Quantum Optoelectronics at HSE University in St. Petersburg in collaboration with researchers at the Laser Institute of Qilu University of Technology (Shandong Academy of Sciences) in China. The findings were published in the Journal of Lightwave Technology.The researchers carried out what they described as the first direct comparison of the response speed of a microdisk laser under optical and electrical excitation, allowing them to determine the upper limit of its operating speed more precisely.
Their experiments showed a maximum −3 dB modulation bandwidth of 4.15 GHz. The study used a 12-micrometer microdisk laser operating at room temperature and based on InAs/InGaAs quantum dots.
The researchers said the ability to directly compare the laser's response under different excitation methods could help improve the design of photonic devices and accelerate the development of next-generation photonic chips.
Such chips use light rather than conventional electrical signals to transmit and process information, offering potential advantages for the high-speed, high-volume data movement required by modern AI systems and data centers.
The researchers said the technology could eventually enable multiple computing circuits to be integrated on a single chip and allow signals to move rapidly between them. They also pointed to the possibility of networks of optically interconnected microlasers operating in ways inspired by neural networks or reservoir computing.
The work comes as researchers worldwide explore photonic computing as a way to address the growing computational and data-transfer demands of artificial intelligence. Photonic accelerators are increasingly being studied for their potential to improve throughput, latency and energy efficiency compared with conventional electronic systems.