组成驱动的无机材料XPS O1s结合能与化学位移智能预测OA

The O 1s XPS binding energy is a key indicator for identifying oxygen chemical environ-ments in inorganic materials with broad applications in energy materials,catalysis,and synthesis monitoring.However,the cross-system peak assignment relies heavily on extensive validation experi-ments.In this study,we developed a composition-driven machine learning framework based on O 1s binding energy data from 514 oxygen-containing inorganic compounds.This framework exclusively us-es chemical formula inputs for low-cost pre-experimental peak-position estimation and trend assess-ment.Using the original compositional descriptors,the Gradient Boosting model achieved a test MAE of 0.505 eV.After introducing physically inspired descriptors,including the cation field strength and transition metal fraction,the MAE decreased to 0.481 eV and R2 increased to 0.716.The mean cation radius was identified as a key variable,a finding further supported by the XPS mea-surements of Ca14Mg5M4O25(M=Al,Sc,Y).For the paired salt/oxide systems,the model predicted the relative chemical shifts with an MAE of 0.339 eV and a R2 of 0.822.This method provides a low-cost and interpretable auxiliary reference for XPS analysis of complex oxygen-containing inorganic ma-terials.