Design and analysis of on-chip reconfigurable photonic components for photonic multiply and accumulate operation

Photonic computing plays a significant role in high-performance computing applications. The high speed and capacity of processing larger information by photonic signals assist the high-performance computing applications such as hardware accelerators, machine learning application and deep learning applications. In this work, we propose a photonic MAC (PMAC) based on reconfigurable photonic components such as reconfigurable Mach–Zehnder interferometer (RMZI), reconfigurable directional coupler (RDC) and reconfigurable micro-ring resonator (RMRR). Theoretical analysis and simulations are carried out based on MATLAB R2023a software package and Ansys Lumerical 2018a software suits. Based on the analysis it is evident that the PMAC realization, based on RDC is more suitable for MAC operations due to its smaller footprint and less sensitive (2%) to fabrication variations. Comparatively RMZI results in larger footprint and RMRR shows more sensitive (11%) to fabrication variations. The photonic MAC proposed in this work acts as the key component for machine learning and deep learning applications.