Three-dimensional, multimodal synchrotron data for machine learning applications.

Area of Science:

• Materials Science
• Data Science
• Scientific Imaging

Background:

• Machine learning (ML) is increasingly used in scientific imaging, but requires high-quality training data.
• Developing advanced ML tools for multimodal imaging data presents a significant challenge due to data limitations.

Purpose of the Study:

• To present a unique, multimodal synchrotron dataset for developing ML algorithms.
• To provide a resource for advancing deep learning and data fusion techniques in scientific imaging.

Main Methods:

• Multi-resolution micro X-ray computed tomography (CT) was used to characterize zeolite pores.
• Spatially resolved X-ray diffraction computed tomography (XRD-CT) mapped sodium and zinc phases.
• Controlled zinc absorption created a distinct two-phase material for analysis.

Main Results:

• A spatially resolved, 3D, multimodal, multi-resolution dataset was generated.
• The dataset features a bespoke zinc-doped Zeolite 13X sample.
• Both raw and processed data are publicly available through Zenodo.

Conclusions:

• The dataset facilitates the development of novel ML techniques.
• Applications include super-resolution, multimodal data fusion, and 3D reconstruction algorithms.
• This resource advances ML applications in materials science and medical/physical imaging.