Reconfigurable Optical Computer for Solving Partial Differential Equation Open Access
Downloadable ContentDownload PDF
Nowadays, digital processors are overwhelmed by vast data and complex iterative operation. Thus, some of the workloads can be taken by more specific processor whose hardware is engineered to map specific functionalities and operations such as TPU and FPGAs. In this category, various photonic engines1 which leverage on the absence of RC delay, distributed network and possibility to do weighted addition inherently by interference and concurrently exploit attojoule efficient high-speed modulators. Here we introduce Reconfigurable Optical Computer (ROC), a research project undertaken by Professor Volker J. Sorger’ OPEN Lab Team and professor Tarek El-Ghazawi’s HPCL Team both at the George Washington University, through the funding of NSF RAISE Grant. This project is mainly working on the physics theory and fabrication technology of analog co-processors in silicon photonics and optical metatronics. The goal of ROC is to investigate, model and ultimately develop an optical solution that can numerically and experimentally solve various partial differential equations (PDE). My contribution to ROC has been to prove the reconfigurability of the photonic integrated version of ROC, by numerically modeling a programmable photonic chip in a photonic interconnect simulation framework (Lumerical-Interconnect) photonic integrated aiming to map different partial differential equation problems which describe heat transfer problems in different materials. I compared the solutions obtained at each node of the photonic circuit to the discretized solutions obtained using commercially available heat transfer simulation (COMSOL) which uses finite element modeling. Since each material has precise thermal property, the characteristic heat transfer map is uniquely represented, thus attaining a library of photonic chip configuration for solving the Laplace equation applied at different domains. Therefore, the designed chip can be reprogrammed to solve the Laplace heat transfer equation for different materials and holds the potential to solve other kinds of PDE.