Machine-Learning-Assisted Synthesis of Polar Racemates.

Area of Science:

• Materials Science
• Crystallography
• Computational Chemistry

Background:

• Racemates are gaining interest for their nonlinear optical and piezoelectric properties.
• The synthesis of specific metal-hexafluoride compounds (M = Ti, Zr, Hf) with 2,2'-bipyridine (bpy) ligands was explored.

Purpose of the Study:

• To apply a machine-learning-assisted approach to synthesize missing titanium (Ti) and zirconium (Zr) members of a specific metal-hexafluoride family.
• To understand and predict the formation of polar noncentrosymmetric racemates versus centrosymmetric chain compounds.

Main Methods:

• Utilized a machine-learning-assisted composition space approach to guide synthesis.
• Trained machine learning models on reaction parameters (e.g., molar ratios, pH) to identify trends.
• Employed fluorine K-edge X-ray absorption spectroscopy to investigate chemical bonding differences.

Main Results:

• Successfully synthesized the Ti and Zr members of the [Cu(bpy)₂(H₂O)]₂[MF₆]₂·3H₂O family.
• Identified the 2,2'-bipyridine to copper oxide molar ratio as a primary driver for phase selection in Zr and Hf systems.
• Determined that decreased hydrofluoric acid (HF) concentration (increased pH) is crucial for Ti-based polar racemate formation.
• Spectroscopic analysis revealed distinct Ti-F bonding compared to Zr-F and Hf-F.

Conclusions:

• Machine learning effectively guided the synthesis of targeted polar materials.
• Phase selection in these systems is controllable via reaction parameters.
• The distinct electronic structure of TiF₆^2- influences the material properties and synthesis outcomes.