Free-Space Photonic Synaptic Modulators Based on Tamm Plasmon for Adaptive Multispectral Image Processing

The rapid advancement of computing technologies has underscored the increasing need for systems capable of efficiently processing and improving large datasets of visual information. Traditional electronic computing architectures, while robust, are facing growing challenges related to speed, power consumption, and adaptability in image processing tasks. In response, neuromorphic computing, inspired by the brain’s complex neuronal and synaptic functions, has gained attention as a promising solution. Optical technologies, in particular, have garnered significant interest for their ability to deliver superior speed and parallelism, making them well-suited for modern image processing. Optical computing, especially through photonic integrated circuits (PICs), combines high-bandwidth optical communication with localized processing, significantly improving speed and energy efficiency [1]. Additionally, diffractive deep neural networks (D²NN), a form of free-space optical computing, enable efficient processing of complex tasks by guiding light through diffractive layers, leveraging the advantages of speed, parallelism, and energy efficiency in optical systems [2].<br/>Reconfigurable photonic structures are also gaining recognition for their role in adaptive image filtering within free-space optical systems. However, despite these advancements, challenges such as denoising, spectral filtering, and contrast enhancement in multispectral data still require further development. Addressing these challenges is essential for improving the management of complex visual information.<br/>In this work, we propose the use of active Tamm plasmon resonators to create a highly adaptable and efficient photonic structure for multispectral image processing. We have developed a reconfigurable device with fine control over resonant wavelengths by adjusting the impedance matching condition of the Tamm plasmon resonator, allowing it to target specific spectral components with high quality factor (Q-factor). To achieve adaptive control with a gradual on/off function, we integrate poly(3,4-ethylenedioxythiophene):polystyrene sulfonate (PEDOT:PSS) into the resonator. Depending on its doping state, PEDOT:PSS exhibits either metallic or dielectric properties, enabling high absorptance or reflectance at wavelengths longer than its plasma frequency. The coexistence of doped and undoped domains within PEDOT:PSS is crucial for achieving long-term memorizing functions, which are necessary for precise neuromorphic computations across multiple memory stages [3].<br/>To ensure stable synaptic weights, we combine the large modulation depth of the Tamm plasmon, which theoretically reaches 99%, with the non-volatile characteristics of PEDOT:PSS, enabling the creation of multiple synaptic states without saturation. This results in stable synaptic weights with 256 distinct levels, achieved through continuous electrical pulse inputs. Additionally, the high Q-factor of the Tamm plasmon allows for effective multispectral filtering, enhancing target signals while minimizing noise interference in dynamic environments, leading to more scalable and efficient technologies for complex image processing tasks.<br/><br/>References<br/><br/>[1] B.J. Shastri, A.N. Tait, T. Ferreira de Lima, W.H. Pernice, H. Bhaskaran, C.D. Wright, P.R. Prucnal " Photonics for artificial intelligence and neuromorphic computing," Nature Photonics. <b>15</b>, 102-114 (2021).<br/>[2] X. Lin, Y. Rivenson, N.T. Yardimci, M. Veli, Y. Luo, M. Jarrahi, and A. Ozcan, “All-optical machine learning using diffractive deep neural networks” Science <b>361</b>, 1004-1008 (2018).<br/>[3] J. H. Ko, D. H. Seo, H.-H. Jeong, S. Kim, Y. M. Song, “Sub-1-volt electrically programmable optical modulator based on active Tamm plasmon,” Advanced Materials <b>36</b>, 2310556 (2024).