Designing Pb-Free High-Entropy Relaxor Ferroelectrics with Machine Learning Assistance for High Energy Storage.

Area of Science:

• Materials Science
• Solid State Physics
• Dielectric Materials

Background:

• High-entropy materials offer potential for enhanced dielectric energy storage in relaxor ferroelectrics.
• The vast compositional space of these materials complicates rational design for optimal performance.
• Relaxor ferroelectrics are crucial for pulsed-power electronic systems.

Purpose of the Study:

• To develop a machine learning-supplemented strategy for designing high-entropy relaxor ferroelectrics.
• To achieve ultrahigh energy-storage density and efficiency in these materials.
• To identify key intrinsic features influencing material performance.

Main Methods:

• Utilized a machine learning approach, specifically a random forest regression model, to identify critical intrinsic features of constituent ions.
• Integrated six A-site and one B-site ionic features for material design.
• Performed atomic-level local structural analysis to understand polarization behavior.

Main Results:

• Identified the (Bi2/5Na1/5K1/5Ba1/5)(Ti,Hf)O3 high-entropy system with exceptional dielectric energy storage properties.
• Achieved an ultrahigh energy-storage density of 17.2 J cm⁻³, 87% efficiency, and 79 kV mm⁻¹ breakdown strength.
• Observed a highly fluctuating local polarization structure with pronounced orientation disorder and distributed polarization vectors.

Conclusions:

• The data-driven, machine learning-assisted strategy effectively navigates the complex composition space of high-entropy relaxors.
• The optimized relaxor exhibits excellent dielectric properties, high discharge energy density (5.8 J cm⁻³), and power energy density (447 MW cm⁻³).
• This approach facilitates the rational design of high-performance dielectric materials for advanced energy storage applications.