ENHANCEMENT OF THERMOELECTRIC PROPERTIES OF VAPOR-PHASE CONDENSATES SnTe USING AI METHODS

Abstract

The article explores the potential for enhancing the thermoelectric properties of tin telluride (SnTe) using modern artificial intelligence methods. SnTe, a semiconductor material with a NaCl crystal structure, a narrow bandgap (~0.2 eV), and a high carrier concentration, is widely used to create thermoelectric elements with p-type conductivity. However, the efficiency of SnTe is limited by structural defects, particularly tin vacancies, which affect electrical conductivity (σ), the Seebeck coefficient (S), and thermoelectric power (S²σ). Machine learning methods and generative algorithms enable the prediction of material properties and optimization of synthesis conditions, significantly reducing time and costs. Using neural networks, complex relationships between composition, morphology, and thermoelectric characteristics were modeled. This approach resulted in new SnTe compositions with doping elements (gallium, bismuth), demonstrating improved performance. The methodology included the deposition of SnTe films (40–800 nm), morphological analysis using atomic force microscopy and diffractometry, and the creation of training datasets. Genetic algorithms were applied to identify optimal compositions, while computer vision models automated surface analysis, determining the optimal orientation of nanocrystals to minimize thermal resistance. The results showed an increase in electrical conductivity up to ~12·10³ Ω⁻¹·cm⁻¹, a Seebeck coefficient of ~85 μV/K, and thermoelectric power of ~25 μW/K²·cm (an improvement of 39–50%). Optimizing film thickness and nanocrystal distribution significantly reduced thermal losses. The findings open new opportunities for the application of SnTe in highly efficient thermoelectric generators, cooling systems, industrial heat recovery, and renewable energy.