Computational modelling workflows for metal-organic polyhedra in The World Avatar

Area of Science:

• Materials Science
• Computational Chemistry
• Supramolecular Chemistry

Background:

• Metal-organic polyhedra (MOPs) show great promise for host-guest chemistry.
• Discovering new MOPs is challenging due to the vast number of possible building units.
• Computational screening is a key strategy for MOP discovery.

Purpose of the Study:

• To improve computational screening of MOPs by incorporating post-assembly computational modeling.
• To benchmark different computational methods for MOP structure refinement.
• To create a refined dataset of MOPs for host-guest chemistry applications.

Main Methods:

• Assembling computation-ready MOPs using geometric operations.
• Benchmarking geometry optimization methods (MLIPs, tight-binding DFT) against 85 experimental MOP structures.
• Applying the most accurate method to a large dataset and integrating it into The World Avatar via OntoMOPs.

Main Results:

• Geometry optimization significantly refines MOP cavity and pore properties compared to initial structures.
• Machine learning interatomic potentials (MLIPs) demonstrated high accuracy and convergence.
• A refined MOP dataset was created and made accessible for natural language querying.

Conclusions:

• Post-assembly computational modeling, particularly using MLIPs, is crucial for accurate MOP property prediction.
• The refined MOP dataset enhances computational screening for host-guest chemistry.
• The developed workflow enables efficient discovery of MOPs for specific guest molecules, like urea.