Manifold learning and clustering of spatially resolved high-dimensional Raman spectra for surface heterogeneity analysis of polymer solid-state electrolytes.

Abstract

Here, we introduce a novel computational framework for analyzing spatially resolved spectral imaging on the basis of a manifold learning technique coupled with a clustering algorithm. A total of 57 600 Raman spectra collected from a series of sodium trifluoromethanesulfonate (NaOTf) and polyethylene oxide (PEO) mixtures were used to verify the framework. From the framework analysis, the 1 : 12 (NaOTf : PEO) sample showed the highest degree of salt dissolution, but high surface heterogeneity. The 1 : 4 and 1 : 12 samples displayed undissolved salts with amorphism and a heterogeneous surface with low concentration pockets, respectively. The proposed framework enables decision-making based on high-dimensional spectroscopy data and is a potential candidate for spatially and temporally resolved spectral data analysis applied to material design-space investigations.

Abstract as supplied by the indexing source. © the authors / publisher.