Low-energy electron microscopy intensity-voltage data - Factorization, sparse sampling and classification.

Area of Science:

• Surface science
• Materials science
• Spectroscopy

Background:

• Low-energy electron microscopy (LEEM) I-V curves offer hyperspectral surface imaging.
• Analyzing LEEM I-V data for distinct surface phases is challenging.
• Complex surface compositions and structures require advanced analytical methods.

Purpose of the Study:

• To introduce and validate a novel algorithm, factorizing data into spectra and concentrations (FSC³), for analyzing LEEM I-V data.
• To demonstrate the algorithm's capability in identifying distinct physical surface phases.
• To enhance the efficiency and accuracy of surface characterization techniques.

Main Methods:

• Application of the unsupervised and fast factorizing data into spectra and concentrations (FSC³) algorithm.
• Utilizing experimental LEEM I-V data from praseodymium oxide and ruthenium oxide growth on ruthenium substrates.
• Employing sparse sampling to reduce measurement time.
• Implementing a support vector machine for supervised classification based on FSC³ outputs.

Main Results:

• The FSC³ algorithm successfully factorizes complex LEEM I-V data into characteristic spectral and concentration components.
• Demonstrated a reduction in measurement time by 1-2 orders of magnitude using sparse sampling.
• Achieved reliable classification of surface types using FSC³-derived features and a support vector machine.

Conclusions:

• FSC³ provides an efficient and unsupervised method for analyzing hyperspectral LEEM I-V data.
• The developed approach significantly accelerates surface analysis and enables precise identification of surface phases.
• This methodology is highly relevant for dynamic surface studies and advanced materials characterization.