Machine learning-guided ATR-FTIR for in-depth analysis of graphene oxide dispersions

Abstract

The variation of graphene oxide preparation techniques and the often occurring similarity of spectral information in molecular spectroscopy data for tested samples pose challenges for reliable data interpretation, especially when conservative “manual” analysis methods are used. This work employs a machine learning (ML)–based approach to develop an algorithm to solve cluster analysis issues of the infrared spectroscopy data for the graphene oxide: as–prepared, purified (by dialysis bag), and reduced samples. We propose an ML–based model to provide fully–automated qualitative analysis and a semi–automated pipeline for functional groups speciation analysis on graphene oxide, developed by simultaneously combining statistical analysis and data processing, optimization algorithms, and applying unsupervised learning techniques. Also, the study examines the possibilities of applying ML to analyze and cluster data from UV/vis and Dynamic Light Scattering (DLS).