At the PHOCUS Lab (Department of Electrical Engineering, IIT Hyderabad), my students and I work along two major lines: beam-steering using photonics, and photonic computing for AI/ML and optical communication.
Beam-Steering Using Photonics:
Modern 5G and satellite antennas don't physically move. Instead, they steer their signal beam electronically. But the conventional way of doing this has a flaw: when the signal covers a wide range of frequencies, different frequencies end up pointing in slightly different directions, weakening the link.
What we're building: Photonic beam-steering systems where the antenna signals are carried and controlled by light, so the beam points in the same direction at every frequency. We are exploring several approaches, including specially designed multi-core optical fibers, and quasi-time-delay techniques based on optical frequency combs and frequency modulation (studies underway), among others.
How it works: Instead of shifting the phase of each antenna's signal electronically, we give each antenna element exactly the right time delay in the optical domain, for example by routing signals through different cores of a multi-core fiber and simply tuning a laser to set the delays. Because true time delays affect all frequencies equally, the beam stays locked on target across the entire band.
Why this matters: These approaches let antennas be fed remotely over fiber and steered cleanly across the full bandwidth, important for 5G networks, satellite terminals, and ground stations.
Photonic Computing for AI/ML Applications:
AI systems spend most of their time and energy on a few basic mathematical operations, multiplying and adding large arrays of numbers. Electronic chips doing this at scale are hitting limits in speed and power consumption.
What we're building: Processors that do this math with light instead of electricity: reconfigurable optical engines for multiplication, addition, and convolution, along with optical pre-processing stages for communication receivers.
How it works: Numbers are encoded onto light using compact optical components, and the multiplication and addition happen directly in the optical domain as the light propagates and is detected. The same hardware can be reconfigured to handle different problem sizes. We apply the same idea to optical communication: by processing signals while they are still light, before they are converted to electronics, we take load off the power-hungry digital processing chips in receivers.
Why this matters: Light offers enormous bandwidth and natural parallelism, so these operations run faster and with far less energy than in electronics. Together, these efforts point toward light-based accelerators for future computing platforms and lower-latency, more energy-efficient communication systems.