Computational reconstruction on-chip spectrometer based on reconfigurable silicon photonc filters

Spectroscopic analysis techniques are indispensable tools in a multitude of disciplines such as biomedical research, material science, and remote sensing. Traditional benchtop spectrometers suffer from several drawbacks; they are generally bulky, complex, and expensive, rendering them ineffective for emerging applications such as wearable health monitoring and Lab-on-Chip systems. In comparison to bulky desktop spectrometers, integrated chip-level spectrometers find application in portable health monitoring, environmental sensing, and other scenarios. We designed an on-chip spectrometer based on a silicon photonics platform. The device consists of a silicon photonic filter with a reconfigurable transmission spectrum. By altering the transmission spectrum of the filter, multiple and diverse sampling of the input spectrum can be obtained. Leveraging an artificial neural network algorithm, the incident spectrum is reconstructed from the sampled signals. The reconfigurable silicon photonic filter is composed of intercoupled Mach-Zehnder interferometers and micro-ring resonators. The introduction of thermal-optic phase shifters facilitates reconstruction of the transmission spectrum of the filter. Through this approach, a response function encompassing diverse features of broad and narrow spectra can be obtained from a single reconfigurable filter, eliminating the need for a filter array and significantly reducing the footprint of the spectrometer. Simulation results demonstrate that the designed device can achieve continuous and sparse spectrum reconstructions within the wavelength range of 1500-1600 nm, with a resolution of approximately 0.2 nm. On a test set composed of synthetic spectra, the calculated average RMSE for the reconstructed spectra is 0.0075, with an average relative error of 0.0174. Benefiting from the reconfigurable nature of the silicon photonic filter, this device exhibits the capability to flexibly adjust the number of sampling channels, enabling users to configure the chip according to specific application scenarios. This device holds significant potential in applications such as wearable optical sensors and portable spectrometers.