Application-Specific AI-Assisted Spectral Raman Detection

• Applicants: Dr.-Ing. Reinhard Caspary; Professor Dr.-Ing. Roland Lachmayer; Professor Dr.-Ing. Bodo Rosenhahn; Professor Dr. Bernhard Roth
• Subject Area: Measurement Systems ; Engineering Design, Machine Elements, Product Development
• Term: since 2023
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 517733257

The aim of this proposal is to use unconventional optics for a new high-resolution wave-dependent Raman spectroscopy approach. Compared to conventional Raman systems, which measure a continuous spectrum, we plan to use highly accurate detection of individually selectable wavelength ranges to provide a barcode-like fingerprint of the substances to be detected. The approach can be adapted to polymers, special chemical substances such as toxins or certain gases in a task-specific manner and realize real-time detection. One potential application that we will pursue during instrument development and use to validate the approach is the detection of microplastics in liquid samples such as drinking or industrial water, in which the sensitive detection of low concentrations down to single particles will be demonstrated. For the first time, methods of additive manufacturing will be combined with machine learning for design and data evaluation to realize the system. Our approach is to direct the measurement signal selectively to the light-sensitive regions of a detector via a holographic diffractive optical element (DOE), tailored specifically by using two-photon polymerization. For this purpose, we will employ highly sensitive APD cameras, which usually cannot be used for Raman detection in conventional systems due to the large gaps between the pixels. At the same time, the photon yield will be further increased by using additively manufactured free-form lenses. Machine learning will be used for the reliable evaluation of the measurement signal, which is reduced compared to a full spectrum, as well as for the calibration and optimization of the system. This makes it possible to tune the interdependent components of the specific Raman system, i.e. the lens, the DOE and signal strength. The goal is to manufacture the entire system via an additive manufacturing process chain and demonstrate it through microplastic detection, so that in the future cost-effective and individualized Raman systems can be used for further applications in research, for example in chemistry, medicine, environmental analysis or forensics.

• DFG Programme: New Instrumentation for Research
• Major Instrumentation: Additive Fertigungsanlage, MJM-Verfahren