Carbon-based materials have enormous potential for constructing a sustainable future; however, material scientists require tools to properly analyze their atomic structure, which determines their functional properties. One of the tools used for this is X-ray photoelectron spectroscopy (XPS), but the results can be challenging to interpret. Now, researchers have created the XPS Prediction Server, a machine-learning tool for improving XPS analyses.
XPS spectra are graphs containing a collection of peaks representing the binding energy of electrons deep within the atoms that comprise a material. Because binding energies are affected by the atomic environment, they can deduce the connection of atoms in a given material or molecule. However, because binding energies are affected by various factors, XPS spectra are difficult to interpret. The binding energies of various atomic features can also overlap, complicating the analysis further.
Based on a computer-generated structural model, the researchers created a computational method for predicting the binding energy spectrum of a material. It simplifies data interpretation (X-ray photoelectron spectroscopy) by allowing the comparison of the experimentally observed binding energies and the computational predictions. The researchers trained an inexpensive computer algorithm to predict the outcome of a computationally expensive reference method. The model used an efficient combination of computationally cheap and expensive quantum mechanical data.