High-performance, intelligent, on-chip speckle spectrometer using 2D silicon photonic disordered microring lattice

Abstract

High-performance integrated spectrometers are highly desirable for applications ranging from mobile phones to space probes. Based on silicon photonic integrated circuit technology, we propose and demonstrate an on-chip speckle spectrometer consisting of a 15×15, 2D disordered microring lattice. The proposed 2D, disordered microring lattice was simulated by the transfer-matrix method. The fabricated device featured a spectral resolution better than 15 pm and an operating bandwidth larger than 40 nm. We also demonstrated that, based on the speckle patterns, our device can perform a spectrum classification using machine learning algorithms, which will have a huge potential in fast, intelligent material and chemical analysis.

Integrated Microring Spectrometer with In‐Hardware Compressed Sensing to Break the Resolution‐Bandwidth Limit for General Continuous Spectrum Analysis

“…Dual MRR [32] 0.04 100 2501 1 2501 Active 60×60 μm 2 Sparse 2D disordered microring lattice [33] 0.015 nm 40 nm 2667 4096 0.65 Passive 1000×1000 μm 2 -Multimode spiral Waveguide [34] 0 of the random grating is nearly orthogonal, the sparsity limit approaches the number of random gratings M. It can always attain a larger spectral bandwidth consisting of more resonant peaks by simply using more random gratings, which does not increase the reconstruction error (Figure S4d in Note S3, Supporting Information). Therefore, to effectively increase spectral channels, we can independently enhance the spectral resolution (using a higher quality-factor MRR) and the spectral bandwidth (via a larger FSR, tuning range, and sparsity).…”

“…Otherwise, the number of physical channels required would be significantly greater than the number of spectral channels, increasing system complexity and power consumption. [ 33 ] Our scheme utilizes the MRR to achieve CS in hardware by decomposing any complex continuous spectrum into a series of simple comb spectra. By precalibration of the MRR and response matrix, the general input continuous spectrum can be successfully retrieved.…”

“…Moreover, there is an inherent dilemma of balancing power consumption and device footprint. In the past decade, a new type of spectrometer based on computational spectral reconstruction has emerged, [17][18][19][20][21][22][23][24][25][26][27][28][29][30][31][32][33][34] since it breaks through the design limitation of conventional spectrometers. Unlike conventional spectrometers based on dispersive optics and narrowband filters by spectrum-to-space mapping principle (one frequency mapped to one spatial position), reconstructive spectrometers simultaneously address a full range of spectral components at each detector, which has a robust performance to fabrication imperfections.…”

An Inversely Designed Reconstructive Spectrometer on SiN Platform

“…However, due to the large number of design parameters involved, conventional design approaches for SWFs often rely on a brute-force random approach. While this approach, which is also commonly used in other types of broadband filters for reconstructive spectrometers, [7][8][9] can indeed generate filters with highly random and distinct spectral responses, it also leads to unpredictable and non-reproducible performance of the realized spectrometer. Moreover, without fine-tuning the design parameters, the realized spectrometer may fall short of its optimal capabilities.…”

“…[ 6 ] However, previously demonstrated SWFs are designed in a brute‐force random manner, and have no optimized and deterministic selection of the structural parameters, leading to suboptimal and non‐reproducible performance of the spectrometer. This problem also applies to other broadband filters, such as disordered photonics media, [ 7 ] random photonic crystal cavities, [ 8 ] disordered microring lattice [ 9 ], and multi‐point self‐coupling waveguide filters. [ 5 ] On the other hand, using an inverse design algorithm to optimize a large number of structural parameters across a large space has been extensively applied to integrated photonics components, but limited to single device with simple spectral response, such as waveguide bends, [ 10 ] tapers/transitions, [ 11 ] beam splitters [ 12 ] waveguide crossings [ 13 ] etc.…”