Numerical methods for chemical analysis of complex surfaces

Abstract

Temperature programmed desorption (TPD) is a classic surface science experiment that provides information regarding gas adsorption thermodynamics and desorption kinetics. Conventional TPD data analysis methods are inadequate for desorption processes involving complex surfaces, such as those involving porous materials or heterogenous mixtures. To address this, numerical methods were developed for analysis of TPD from such materials. These signals, which consist of overlapping time-resolved mass spectra, are governed by dynamic chemical processes. The core innovation involves nonlinear regression of numerically integrated kinetic models in a method inspired by deconvolution analysis of X-ray photoelectron spectra. Fundamental details of implementation and accuracy are discussed, and the modeling approach is extended to address porous samples and amorphous surfaces. A probabilistic approach for TPD analysis based on Bayesian machine learning theory is introduced to quantify uncertainty in extracted kinetic parameters. This approach is also modified for application to Gaussian mixture decomposition. These methods are applied to surface analysis of metal-organic frameworks materials with an emphasis on application to membrane separation technologies.