MOFSynth-ADV: An Open-Source Engine for Synthesizability Evaluation of Metal-Organic Frameworks

Area of Science:

• Materials Science
• Computational Chemistry
• Crystallography

Background:

• Accurate prediction of Metal-Organic Framework (MOF) synthesis feasibility is crucial for materials discovery.
• Existing computational tools often rely on proprietary software or less accurate classical force fields.
• High-throughput screening requires efficient and reliable methods for evaluating potential MOF structures.

Purpose of the Study:

• To introduce MOFSynth-ADV, an upgraded, open-source tool for assessing the synthetic feasibility of MOFs.
• To integrate advanced computational methods, including extended tight-binding (xTB) and machine learning interatomic potentials (MLIPs), for improved accuracy.
• To provide a robust platform for high-throughput screening and accelerate MOF discovery.

Main Methods:

• Integration of the Atomic Simulation Environment (ASE) for geometry optimization and energy calculations.
• Leveraging extended tight-binding (xTB) calculations to enhance accuracy and reduce unconverged cases.
• Utilizing machine learning interatomic potentials (MLIPs), specifically MACE-OFF, for efficient and accurate energy evaluations.
• Development of a fully open-source framework, eliminating reliance on proprietary software.

Main Results:

• MOFSynth-ADV demonstrates significantly higher accuracy in structural and energetic predictions compared to classical force fields.
• The xTB module reduced unconverged cases by 73% and computational time by 13%.
• MLIPs (MACE-OFF) provided comparable improvements in accuracy and efficiency.
• Validation against experimentally reported MOFs confirmed the tool's superior predictive quality.

Conclusions:

• MOFSynth-ADV offers a fast, modular, and deployable solution for predicting MOF synthetic feasibility.
• The open-source nature and improved accuracy establish a strong foundation for future computational workflows in MOF research.
• This advanced framework facilitates the discovery of novel MOFs through enhanced machine learning and quantum-classical hybrid approaches.