Orlando J Silveira1, Markus Junttila1, Lauri Kurki1, Shawulienu Kezilebieke2, Peter Liljeroth1, Adam S Foster1,3
1Department of Applied Physics, Aalto University, Aalto, Helsinki, Finland
2Department of Physics, Department of Chemistry and Nanoscience Center, University of Jyväskylä, Jyväskylä, Finland
3WPI Nano Life Science Institute (WPI-NanoLSI), Kanazawa University, Kanazawa, Japan
Tip-enhanced Raman spectroscopy (TERS) is a method capable of mapping vibrational modes of molecules in a unique way with Å-resolution, capturing not only the physical structure of a sample but also its chemical bonds [1]. Both effects, in combination with the Raman fingerprints of different chemical groups, can fully define the structural arrangement and the constituents of a single molecule. As TERS is promising for being a widely adopted method in materials science, a full understanding of the produced images and the physics behind the different contrasts is necessary. In this work, we propose image interpretation tools to extract physical information from TERS images using machine learning. Image analysis methods using machine learning have been already proven effective in scanning tunnelling microscopy (STM) [2] and atomic force microscopy (AFM) [3]. In here, we bring similar concepts to demonstrate the extraction of structural and chemical information directly from simulated TERS images. We also analysed different possibilities of simulating the TERS signal starting from density functional theory (DFT) calculation and investigated how they correspond to experimental images.
[1] Zhang, R., Zhang, Y., Dong, Z. et al. Chemical mapping of a single molecule by plasmon-enhanced Raman scattering. Nature 498, 82–86 (2013)
[2] Lauri Kurki, Niko Oinonen, and Adam S. Foster. ACS Nano 2024 18 (17), 11130-11138
[3] Oinonen, N., Kurki, L., Ilin, A., Foster, Adam S. MRS Bulletin 47, 895–905 (2022)