Application of machine learning for nanodiamonds shape and surface classification based on X-ray pattern analysis

Area of Science:

• Materials Science
• Computational Chemistry
• Nanotechnology

Background:

• Characterizing nanoparticle shape and surface structure is crucial for understanding their properties.
• Powder diffraction data offers insights into atomic arrangements but requires advanced analysis for complex nanostructures.
• Distinguishing between different nanodiamond morphologies and surface terminations is challenging.

Purpose of the Study:

• To develop and apply Machine Learning (ML) algorithms for classifying nanodiamond shape and surface structure.
• To train ML models using simulated diffraction data and validate them with experimental results.
• To determine the predominant shape and surface characteristics of adamantane-synthesized nanodiamonds.

Main Methods:

• Utilized three ML algorithms: Random Forest, Neural Networks, and Extreme Gradient Boosting.
• Trained classifiers on structure functions S(Q) derived from Molecular Dynamics simulations of nanodiamond models.
• Applied trained classifiers to experimental powder diffraction data of nanodiamonds (1.2–3.3 nm).

Main Results:

• All three ML algorithms demonstrated high accuracy in classifying nanodiamond shapes (rods, plates, superspheres) and surface structures (one or three dangling bonds).
• ML classification results closely matched those obtained via Pair Distribution Function analysis.
• The study identified plate-like structures with (111) surfaces featuring three dangling bonds as the dominant form in adamantane-synthesized nanodiamonds.

Conclusions:

• Machine learning provides a robust and efficient method for analyzing nanostructure from diffraction data.
• The developed ML models successfully characterize nanodiamond morphology and surface chemistry.
• Adamantane synthesis predominantly yields plate-shaped nanodiamonds with specific surface terminations.