69 68 Principal Scientist Profiles Mario Chemnitz Mario Chemnitz Principal Scientist Profiles MARIO CHEMNITZ RESEARCH AREAS The research group of Dr. Chemnitz investigates the application potential of emerging programmable optical devices in combination with machine learning algorithms and nonlinear photonics. Through the exploration and functionalization of novel nonlinear wave dynamics, the research group thrives in the exploration of new imaging and sensing solutions, new nonlinear states of light, and neuromorphic (brain-like) processor hardware. Current topics include: • Complex nonlinear phenomena in waveguides and their coherent control. • Liquid-core optical fibers and optofluidic platforms. • Neuromorphic information processing using nonlinear wave dynamics. • Hyperspectral sensing and imaging techniques. TEACHING FIELDS Dr. Chemnitz aims to introduce students to the foundations of machine learning and their applications in optics. Courses will cover the mathematical foundations of various regression, optimization, and learning algorithms as well as their applications in optics, focusing on design strategies for programmable optics, autonomous devices, and optical neural networks. RESEARCH METHODS • Design and fabrication of hybrid-material optical fibers and waveguides with specific linear and nonlinear properties. • Nonlinear phenomena in all-fiber and on-chip systems with a specific focus on optical soliton dynamics and modulation instabilities. • Machine learning algorithms in conjunction with programmable optical devices and reconfigurable waveguides for advanced nonlinear system control. PROFESSOR FOR INTELLIGENT OPTICAL SYSTEMS Dr. Chemnitz holds a PhD in Physics from the FSU in 2019 and pursued his postdoctoral studies at the INRS-EMT in Montreal, Canada. In 2022, he became an independent research group leader at the Leibniz Institute of Photonic Technology, specializing in nonlinear optics and photonics. He holds a professorship for Intelligent Optical Systems in Jena since 12/2024. With over ten years of research experience, he focuses on harnessing nonlinear optical functionalities in optical fibers and integrated photonic devices for applications in biophotonics and neuromorphic information processing. He has received a NEXUS Nachwuchsgruppenförderung from the Carl-Zeiss Stiftung, to build up a dedicated research laboratory and a highly interdisciplinary team at the interface of physics, computer sciences, and engineering. RECENT RESEARCH RESULTS Dr. Chemnitz’ research has been groundbreaking in several key areas: • Reconfigurable optical fibers, where the PI matured liquid-core fibers as a dynamic platform for nonlinear optics featuring local dispersion tunability, advanced mode control, and non-local nonlinearity [1]. • Smart photonics, where the PI demonstrated autonomous on-chip multi-path interferometry for reconfigurable picosecond waveform shaping empowered by meta-heuristic optimization algorithms and an all-optical sampling scheme [2]. • Neuromorphic computing, where the PI showcased fission-based broadband frequency generation as a resource to replicate the computational capabilities of simple neural networks in the optical domain [3]. • Soliton dynamics, where the PI experimentally demonstrated the impact of temporal non-locality in nonlinear media leading to unique noise-resilient localization effects that significantly differ from the soliton phenomena known from glass fibers [4]. HARNESSING NONLINEAR OPTICAL DYNAMICS FOR NEUROMORPHIC COMPUTING In the age of rapidly advancing artificial intelligence, energy efficiency is of global interest. Dr. Chemnitz and his team recently presented a non-intuitive way of how nonlinear optics may be used for low-power emulation of not only one artificial neuron but entire neural networks within a single optical fiber. Unlike electronic hardware that evolves by adding computational units, their system is characterized by its unique scalability of optical wave dynamics: By increasing the nonlinearity of the optical wave dynamics, the computational power increases within the same fiber. Such scaling of computations within a single physical unit may not only solve energy and scalability problems of neuromorphic hardware but also open new avenues of research on processing optical information directly in the optical domain. The figure shows the operating principle of the fiber processor exemplarily for handwritten digit recognition – a key benchmark for neural networks. Images of various handwritten digits are encoded on the phase of the input pulse, projected by spectral broadening inside a fiber, and read out from selected spectral windows. The plot shows a classification result of three types of digits predicted by the experimental system from over 500 handscripts. Illustration of the non-instantaneous light-matter interaction leading to a modification of the classical soliton decay known as supercontinuum generation. Resulting hybrid states exhibit dispersive characteristics in time-frequency space. [1] Chemnitz et al., Laser Phot. Rev. 17, 2300126 (2023). [2] Fischer et al., Optica 8, 1268-1276 (2021). [3] Fischer et al., Adv. Sci., 2303835 (2023). [4] Chemnitz et al., Nat Commun 8, 42 (2017). Contact: Phone: + 49 3641 206 145 Email: mario.chemnitz@leibniz-ipht.de
RkJQdWJsaXNoZXIy OTI3Njg=