High-throughput Computations of X-ray Spectroscopy and Machine Learning-assisted Chemical Environment Identification - Doctor of Philosophy Thesis

Aug 5, 2026

High-throughput computations were developed for X-ray spectroscopy using ab initio L-edge X-ray absorption near-edge structure (XANES) workflows, enabling scalable data generation and processing for chemical environment discovery. The research integrates machine learning models—such as random forests—to identify coordination environments from spectral features extracted from XANES. Featurization strategies and benchmarking on computational and experimental spectra evaluate model accuracy, element-wise performance, and feature importance across spectral regions. Results support reliable chemical environment classification while clarifying which spectral information drives predictions.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages103

Reading Time4 h 20 min

AccessFree to download

Chapter 1 Introduction

Chapter 2 Database of ab initio L-edge X-ray absorption near edge structure

Chapter 3 Random Forest Models for Accurate Identification of Chemical Environment from X-ray Absorption Near-Edge Structure

Chapter 4 Featurization Approaches for Machine Learning of X-ray Absorption Spectra

UC San Diego UC San Diego Electronic Theses and Dissertations Title High-throughput Computations of X-ray Spectroscopy and Machine Learning-assisted Chemical Environment Identification Permalink [https://escholarship.org/uc/item/7f35h4dn](https://escholarship.org/uc/item/7f35h4dn) Author Chen, Yiming Publication Date 2023 Peer reviewed|Thesis/dissertation [eScholarship.org](eScholarship.org) Powered by the California Digital Library University of California UNIVERSITY OF CALIFORNIA SAN DIEGO High-throughput Computations of X-ray Spectroscopy and Machine Learning-assisted Chemical Environment Identification A dissertation submitted in partial satisfaction of the requirements for the degree Doctor of Philosophy in NanoEngineering by Yiming Chen Committee in charge: Professor Shyue Ping Ong, Chair Professor Renkun Chen Professor Zheng Chen Professor Alex Frano Pereira Professor Kesong Yang 2023 Copyright Yiming Chen, 2023 All rights reserved. The dissertation of Yiming Chen is approved, and it is acceptable in quality and form for publication on microfilm and electronically. University of California San Diego 2023 TABLE OF CONTENTS Dissertation Approval Page ................................. iii Table of Contents ...................................... iv List of Figures ........................................ v List of Tables ........................................ vii Acknowledgements ..................................... viii Vita ............................................. x Abstract of the Dissertation ................................. xii Chapter 1 Introduction ................................. 1 Chapter 2 Database of ab initio L-edge X-ray absorption near edge structure .... 7 Chapter 3 Random Forest Models for Accurate Identification of Chemical Environment from X-ray Absorption Near-Edge Structure ............. 22 Chapter 4 Featurization Approaches for Machine Learning of X-ray Absorption Spectra 57 Bibliography ........................................ 76 LIST OF FIGURES Figure 1.1: Schematic diagram for ML applications in spectroscopy........... 4 Figure 1.2: Examples of feature importance analysis................... 5 Figure 2.1: Schematic diagram of high-throughput workflow for L-edge XANES computation and data processing........................... 16 Figure 2.2: Data distribution for the L-edge XANES database. The upper and lower numbers for each element correspond to the number of site-averaged (upper) and site-wise count (lower), respectively................... 17 Figure 2.3: Benchmarking results for (a) cluster radius for SCF, and (b) core-hole treatment parameters for FEFF9 L-edge XANES calculations.......... 17 Figure 2.4: Comparison between experimental spectra, ocean computed spectra and FEFF9 computed spectra........................... 18 Figure 2.5: L2,3 ~~ ~~ edge XANES spectra for 3d transition metal elements......... 18 Figure 3.1: Workflow schema of the coordination environment identification algorithm. 43 Figure 3.2: Performance of five machine learning (ML) classifiers on coordination environment classification............................. 44 Figure 3.3: Normalized feature importance for different regions of spectra........ 45 Figure 3.4: Accuracy of baseline models for each element................ 45 Figure 3.5: Top coordination environment classification Jaccard scores of random forest classifier with respect to (a) datasets’ label entropy categorized by elemental groups and (b) training dataset size...................... 46 Figure 3.6: Boxplots of dataset size distribution per distinct coordination environment ranking label.................................. 46 Figure 3.7: Overview of convolutional neural network classifier’s classification performance with respect to datasets’ label entropy and training dataset size.... 47 Figure 3.8: The random forest classifier’s element-w

FAQs

What computational approach is used to generate L-edge XANES data at high throughput?

The dissertation uses ab initio L-edge X-ray absorption near-edge structure computations within a high-throughput workflow, followed by data processing to build a usable spectral database.

How does the work identify chemical environments from X-ray absorption spectra?

Machine learning, particularly random forest models, is applied to X-ray absorption near-edge structure features to classify coordination environments accurately.

Why are featurization approaches important in the presented machine learning pipeline?

Featurization determines how spectral information is transformed into model inputs; the dissertation evaluates different featurization strategies to improve learning from X-ray absorption spectra.

Aug 5, 2026

Upload by Emma Wilson