Autonomous identification of Lindemann atoms based on deep learning

YK Peng and Z Tian and LL Liu and Q Zheng, MATERIALS TODAY COMMUNICATIONS, 35, 106053 (2023).

DOI: 10.1016/j.mtcomm.2023.106053

Deep learning is one of the most popular fields in computer science and has a vast number of applications. In this work, we present a method to identify Lindemann atoms during melting of gold, silver, and copper nanoparticles using neural network. The feature vector of dataset is obtained based on the processing of LaSCA. Both the training set and the test set of the model have good accuracy, precision and recall. Therefore, a new parameter, the Lindemann atomic probability (LAP), can be established. The comparison between LAP and the kinetic and structural parameters shows that LAP can not only describe the dynamic properties, but also reveal the micro-structural changes during the phase transition. This technology overcomes the shortcomings of traditional methods of calculating Lindemann index, which require a sequence of coordinates for each atom over a period. It also provides a fast and accurate method for studying the thermodynamics process of metals.

Return to Publications page