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

We present MOFSynth-ADV, an advanced iteration of the MOFSynth tool designed to evaluate the synthetic feasibility of Metal-Organic Frameworks (MOFs). By integrating the Atomic Simulation Environment (ASE) to leverage extended tight-binding (xTB) and machine learning interatomic potential (MLIP) calculations, MOFSynth-ADV achieves geometry optimization and energy evaluations with significantly higher accuracy than classical force field approaches. Crucially, this update eliminates reliance on proprietary software, delivering a fully open-source solution optimized for high-throughput screening. Performance benchmarks reveal that the xTB module reduces unconverged cases by 73% while decreasing computational time by 13%, with MLIPs (specifically MACE-OFF) yielding comparable improvements. Validated against a benchmark set of experimentally reported MOFs, MOFSynth-ADV demonstrates superior structural and energetic prediction quality, effectively capturing the subtle chemical interactions essential for accurate synthesis prediction. Fast, modular, and easily deployable, this upgraded framework establishes a robust foundation for future machine learning and quantum-classical hybrid workflows in MOF discovery.