Causal inference machine learning leads original experimental discovery in CdSe/CdS core/shell nanoparticles

Rulin Liu (1) , Junjie Hao (2, 3) , Jiagen Li (1) , Shujie Wang (1) , Haochen Liu (2) , Ziming Zhou (2) , Marie-Hélène Delville (3) , Jiaji Cheng (4) , Kai Wang (2) , Xi Zhu (1)

1 AIRS - Shenzhen Institute of Artificial Intelligence and Robotics for Society
2 Department of Electrical and Electronic Engineering
3 ICMCB - Institut de Chimie de la Matière Condensée de Bordeaux
4 School of Materials Science and Engineering

The synthesis of CdSe/CdS core/shell nanoparticles was revisited with the help of a causal inference machine learning framework. The tadpole morphology with 1–2 tails was experimentally discovered. The causal inference model revealed the causality between the oleic acid (OA), octadecylphosphonic acid (ODPA) ligands, and the detailed tail shape of the tadpole morphology. Further, with the identified causality, a neural network was provided to predict and directly lead to the original experimental discovery of new tadpole-shaped structures. An entropy-driven nucleation theory was developed to understand both the ligand and temperature dependent experimental data and the causal inference from the machine learning framework. This work provided a vivid example of how the artificial intelligence technology, including machine learning, could benefit the materials science research for the discovery.

Mots clés: Morphology; Ligands; Theoretical and computational chemistry; Transmission electron microscopy; Neural networks

Domaines: Matériaux

Rulin Liu, Junjie Hao, Jiagen Li, Shujie Wang, Haochen Liu, et al.. Causal inference machine learning leads original experimental discovery in CdSe/CdS core/shell nanoparticles. Journal of Physical Chemistry Letters, 2020, 11, pp.7232-7238. ⟨10.1021/acs.jpclett.0c02115⟩. ⟨hal-02922356⟩