From Event: SPIE Nanoscience + Engineering, 2018
Over twenty years have elapsed since the first report of single molecule surface enhanced Raman spectroscopy (SERS) yet quantitative sensing in the single molecule regime remains elusive. We have recently introduced a new self-assembly method for fabricating sensor surfaces capable of single-molecule SERS with uniform SERS enhancements of 109 over 100s of μm. Yet, the main challenge in quantitative single molecule SERS is the complex, system dependent, hotspot occupancy rate of analyte. In this work, we solve this problem using a new big data method for quantifying analyte concentration in the single molecule regime. Specifically, deep convolutional neural networks are trained with SERS spectra acquired across large sensor surfaces. The universal approximation property of neural networks is used to automatically fit the hotspot occupancy rate of any molecule. We demonstrate quantitative detection of rhodamine 800 as low as 1 femtomolar, where the single molecule regime begins at approximately 100 picomolar concentrations. This method is validated by comparison with traditional, non-quantitative methods of single molecule SERS detection. Further, we use SERS’s rich spectral information and label free detection to demonstrate the simultaneous quantification of multiple analyte molecules. Finally, this new quantification method is used to sense small molecules produced by bacterial biofilms. The proposed method is not system specific and is thus broadly applicable to any SERS sensor capable of large-area, uniform single molecule detection.
© (2018) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Regina Ragan and William Thrift, "Quantitative single molecule SERS sensing enabled by machine learning (Conference Presentation)," Proc. SPIE 10728, Biosensing and Nanomedicine XI, 107280C (Presented at SPIE Nanoscience + Engineering: August 19, 2018; Published: 17 September 2018); https://doi.org/10.1117/12.2320297.5836012299001.