Abstract
AbstractFor the combination of big data mining and high-performance porous materials, it is necessary to establish the accurate descriptors, especially when it comes to capturing the pore topology of crystalline materials. This study developed persistent homology-based topology fingerprints for metal-organic frameworks (MOFs) and automated software to generate topological fingerprints. An Machine learning-based strategy was primarily applied to screen MOFs with high adsorption properties for xylenes from a database of 137,953 hypothetical MOFs. The approach was then extended to two additional systems (methane/ethane/propane, and chemical warfare agents) to demonstrate its universality. By incorporating topology fingerprints, the model accuracy (R2) was improved from 0.7798 to 0.9205, and the generality was validated. Topology fingerprints identified beneficial pore shapes, sizes, and interpenetration, with rhombic pores being particularly effective for xylene separation. This method offers a new approach for designing high-performance MOFs by focusing on pore topology.
| Original language | English |
|---|---|
| Article number | 123621 |
| Journal | Chemical Engineering Science |
| Volume | 327 |
| DOIs | |
| Publication status | Published - 1 Jun 2026 |
Keywords
- Adsorption
- Metal-organic framework
- Persistent homology
- Porestructure
ASJC Scopus subject areas
- General Chemistry
- General Chemical Engineering
- Industrial and Manufacturing Engineering
Fingerprint
Dive into the research topics of 'Pore topology fingerprints and big-data mining to accelerate the design of high-performance metal-organic frameworks'. Together they form a unique fingerprint.Cite this
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver