Nanohardness from First Principles with Active Learning on Atomic Environments.

Area of Science:

• Materials Science
• Computational Materials Science
• Nanotechnology

Background:

• Predicting material hardness at the nanoscale is challenging due to experimental difficulties and the limitations of traditional models.
• Atomistic simulations offer a potential route for nanoscale material characterization, but require accurate interatomic potentials.

Purpose of the Study:

• To introduce a novel methodology for calculating nanohardness using atomistic simulations.
• To enable the design of materials with exceptional hardness from first principles, particularly at the nanoscale.

Main Methods:

• Developed a methodology combining machine-learning interatomic potentials with atomistic simulations of nanoindentation.
• Fitted machine-learning potentials on-the-fly to quantum-mechanical calculations of local fragments.
• Calculated nanohardness for diamond, AlN, SiC, BC2N, and Si as a function of load and crystallographic orientation.

Main Results:

• The methodology accurately predicts nanohardness for various materials, showing good agreement with experimental data from literature.
• Demonstrated the predictive power of the method across different materials and conditions.
• Validated the approach by comparing simulation results with calibrated macro- and microhardness values.

Conclusions:

• The proposed methodology provides a reliable tool for calculating nanohardness with high predictive power.
• This approach facilitates the *in silico* design of novel, ultra-hard materials.
• The method is particularly valuable for nanoscale applications where experimental measurements are difficult and empirical models fail.