Overlapping XAFS L Spectra of 3d Metals A New Application of the Regularization Method

Abstract

L XAFS spectra of polycrystalline Fe and Cr films are investigated by EXAFS and TEY (Total Electron Yield) techniques. A new method of obtaining local structure information from overlapping L XAFS spectra for 3d metals is proposed. Tikhonov regularization method of solving ill-posed problem is used. In contrast to the conventional methods (Fourier transformation and fitting procedures) this method does not demand any model assumption and any special procedure of deconvolution of L 1 -, L 2 -, L 3 -contributions. The parameters obtained from L XAFS spectra of polycrystalline Fe and Cr films (interatomic distance and coordination number for the first and the second coordination shells) agree with the expected crystallographic values. Up to now the regularization method was applied only to nonoverlapping XAFS spectra.

Decoding the Pair Distribution Function of Uranium in Molten Fluoride Salts from X-Ray Absorption Spectroscopy Data by Machine Learning

“…One direction that can offer a solution to this problem is the use of an “inverse” modeling approach, such as Reverse Monte Carlo (RMC), − regularization method, − and one enabled by the application of supervised machine learning methods. , Recently, a method of extracting the pair distribution function ( g ρ ( r )) around an X-ray absorbing atom from an EXAFS spectrum was developed. , However, the theoretical training set for the artificial neural network (NN) that maps EXAFS on g ρ ( r ) was constructed using MD-EXAFS; hence, the resultant g ρ ( r ) function extracted using the neural network approach was model-dependent. A recently developed follow-up method, utilizing an “objective” training approach, was shown to provide reliable results for Ni complexes in molten salt mixtures, and thus is a good starting point for extending this method to the molten actinide salt research.…”