Advancing AI-Driven Analysis in X-ray Absorption Spectroscopy: Spectral Domain Mapping and Universal Models

TL;DR

This paper advocates an AI-driven pipeline for X-ray Absorption Spectroscopy (XAS) analysis built from benchmarks, workflows, databases, and ML models. It demonstrates two key advances: Spectral Domain Mapping (SDM), which transforms experimental XANES spectra into a simulation-like space to bridge gaps with training data, and a universal XAS model trained across 47 elements to capture global spectral trends, evidenced by the XANES Bird latent-space structure. The first case study shows SDM reconciles discrepancies in oxidation-state predictions for Ti in a combinatorial zinc titanate film, while the second case study outlines data-curation steps and latent-space analyses that underpin cross-element XANES modeling. Collectively, these contributions point to real-time XAS analysis capabilities that reduce reliance on expert interpretation and scale to diverse material systems.

Abstract

In recent years, rapid progress has been made in developing artificial intelligence (AI) and machine learning (ML) methods for x-ray absorption spectroscopy (XAS) analysis. Compared to traditional XAS analysis methods, AI/ML approaches offer dramatic improvements in efficiency and help eliminate human bias. To advance this field, we advocate an AI-driven XAS analysis pipeline that features several inter-connected key building blocks: benchmarks, workflows, databases, and AI/ML models. Specifically, we present two case studies for XAS ML. In the first study, we demonstrate the importance of reconciling the discrepancies between simulation and experiment using spectral domain mapping (SDM). Our ML model, which is trained solely on simulated spectra, predicts an incorrect oxidation state trend for Ti atoms in a combinatorial zinc titanate film. After transforming the experimental spectra into a simulation-like representation using SDM, the same model successfully recovers the correct oxidation state trend. In the second study, we explore the development of universal XAS ML models that are trained on the entire periodic table, which enables them to leverage common trends across elements. Looking ahead, we envision that an AI-driven pipeline can unlock the potential of real-time XAS analysis to accelerate scientific discovery.

Figures (7)

• Figure 1: Schematic of the AI-driven XAS analysis pipeline.
• Figure 2: Schematic of spectral domain mapping (SDM). (a) SDM learns a transformation between corresponding regions in the experimental and simulated XAS domains. Sub-regions correspond to different chemical environments. (b) Workflow of ML applications with and without the SDM.
• Figure 3: (a) The combinatorial thin film of zinc titanate synthesized with pulsed laser deposition. (b) Sample Ti K-edge XANES spectra measured along the Ti concentration gradient.
• Figure 4: The RankAAE autoencoder architecture. The RankAAE loss function $\mathcal{L}$ combines a standard adversarial autoencoder (AAE) loss with a rank constraint that aligns each latent dimension with a chosen chemical descriptor, such as oxidation state (OS) or coordination number (CN).
• Figure 5: (a) Predicted CN from RankAAE without and with SDM. The expected trend, based on the proportion of under-coordinated or distorted Ti atoms, is shown as $1-P_{UD}$. (b) Predicted OS without and with SDM; side panel shows box plots of distribution of these OS values. The expected trend is that the OS should stay constant as Ti fraction varies. Shaded areas indicate the standard deviation across the ensemble of neural networks.