Ligand Optimization of Exchange Interaction in Co(II) Dimer Single Molecule Magnet by Machine Learning.

Area of Science:

• Quantum Chemistry
• Materials Science
• Computational Materials Science

Background:

• Single-molecule magnets (SMMs) are crucial for quantum computing and data storage.
• Tuning magnetic properties, particularly exchange energy, is key for SMM applications.
• Density functional theory (DFT) is a primary method for characterizing SMM magnetic properties.

Purpose of the Study:

• To control the exchange energy (ΔE) in a Cobalt(II) dimer SMM by modifying capping ligands.
• To explore a wide range of ligand substitutions for tuning magnetic interactions.
• To identify predictive descriptors for exchange energy using machine learning.

Main Methods:

• Assembled a dataset of 1081 ligand substitutions for a Co(II) dimer using DFT.
• Employed gradient boosting machine learning models for classification and regression of ΔE.
• Compared various descriptors: one-hot encoded, structure-based, and chemical (HOMO/LUMO, electronegativity, bond order).

Main Results:

• Achieved a broad range of exchange energies from +50 to -200 meV through ligand tuning.
• 80% of ligand modifications resulted in small exchange energies (< 10 meV).
• Chemical descriptors demonstrated superior performance in predicting ΔE compared to others.
• Correlated exchange coupling (ΔE) with bridging angles, aligning with Goodenough-Kanamori rules.

Conclusions:

• Ligand tuning is an effective strategy to control exchange energy in Co(II) dimer SMMs.
• Machine learning models, particularly using chemical descriptors, can accurately predict magnetic properties.
• The findings provide a pathway for designing SMMs with desired magnetic characteristics for advanced applications.