An Optical Computing Breakthrough
Photonic memory development for AI processing has been problematic. It gains speed by sacrificing energy usage. For the first time, however, an international team of electrical engineers developed a new method for photonic in-memory computing with the potential to make optical computing a reality.
The team demonstrated a unique solution addressing current limitations of optical memory that have yet to combine non-volatility, multi-bit storage, high switching speed, low switching energy, and high endurance in a single platform. The researchers from the University of Pittsburgh Swanson School of Engineering, the University of California – Santa Barbara, the University of Cagliari, and the Tokyo Institute of Technology published their results in Nature Photonics.
The materials used to develop the cells were primarily used for static optical applications such as on-chip isolators rather than a platform for high-performance photonic memory. Their work represents a key move toward a faster, more efficient, and more scalable optical computing architecture that can be directly programmed with CMOS circuitry and integrated into today’s computer technology.
The team claims its technology showed three orders of magnitude better endurance than other non-volatile approaches, with 2.4 billion switching cycles and nanosecond speeds. They propose a resonance-based photonic architecture to leverage the non-reciprocal phase shift in magneto-optical materials to implement photonic in-memory computing. By using magneto-optic memory cells comprised of heterogeneously integrated cerium-substituted yttrium iron garnet (Ce:YIG) on silicon micro-ring resonators, the cells cause light to propagate bidirectionally, like sprinters running opposite directions on a track.
By applying a magnetic field to the memory cells, they controlled the speed of light differently depending on whether the light is flowing clockwise or counterclockwise around the ring resonator. They are now working to scale up from a single memory cell to a large-scale memory array, supporting even more data for computing applications.
