Synthesis and Characterization of Machine Learning Predicted Thermoelectric Materials.

다국어 초록 (Multilingual Abstract)

Thermoelectric materials have the capacity to revolutionize multiple fields of engineering. From focused cooling such as computer’s central processing unit (CPU) to energy harvesting for wearable devices. However, these materials have historically been hindered by low performance and expensive materials. The ideal thermoelectric material has a high electric conductivity and low thermal conductivity, a so called ”phonon-glass electron crystal.” The discovery of a low cost, nontoxic thermoelectric material could help increase efficiency as well as reduce the size of known materials. The difficulty of material discovery is not restricted to the field of thermoelectric materials. Many fields such as batteries and high strength materials require new materials to overcome bottlenecks within their fields. A new approach is the use of computational recommendations utilizing machine learning. By taking a large dataset, models can be trained to recommend materials with predicted properties. These algorithms use multiple statistical methods to verify accuracy such as k-fold cross validation as well as precision and recall. Yet, experimental verification is king. In this paper, we will explore the physical properties of a thermoelectric materials as well as some choice papers using machine learning to predict new thermoelectric materials of interest. We will then discuss the methods of synthesis and characterization of thermoelectric materials in this study. We will then conclude with three papers detailing the investigation of these materials and their measurements.