Multi elements approach tuning defect, phase structure, and thermoelectric properties of M3Q2-formula thermoelectric materials

Abstract

Thermoelectric (TE) materials with the formula of M3Q2 exhibit excellent performance near moderate temperatures, showing promise for applications in self-powered wearable smart devices or Internet of Things sensors. In recent years, enormous efforts have been made to tune TE properties through electronic engineering and phonon engineering projects. Most of these works have relied on one or two doping elements to generate defects or nano-inclusions. In this dissertation, the effects of doping multiple elements on M3Q2-formula TE materials are studied. First, the enhanced TE performance levels of Mg3Sb2-based materials after doping multiple elements are investigated, including multiple transition metals (TM = CrMnFeCoCu) codoped at the interstitial site and Mg3Sb1.5Bi0.5 polycrystalline bulk materials codoped with Ti and Te. The multiple interstitial dopants significantly suppress the formation of Mg vacancies, leading to a high power factor of 2799 μW m-1 K-2 and a high ZT value of 0.76 at room temperature. Moreover, the Young’s modulus, hardness, and compressive strength values of the TM0.01Mg3Sb1.5Bi0.5 sample are significantly better than those of the TM-free Mg3Sb1.5Bi0.5 sample. The Ti dopant at the Mg sublattice site promotes the growth of grain size, favoring high carrier mobility. In addition, the impurity level in the band gap induced by the Ti dopant optimizes the carrier concentration near room temperature, leading to a high power factor of 3000 μW m-1 K-2 and a promising ZT value of 0.82 at room temperature. Furthermore, a doping diagram is constructed based on the defect formation energies and physical elemental properties (electronegativity and atomic radius values) to explore the preferred doping sites of Mg3Sb2 materials to select proper multiple dopants. Second, a p-type Bi2Se3-based material is obtained by alloying multiple elements at the cationic site of a Bi2Se3 matrix. The intrinsic Se vacancy is significantly suppressed with the increased configuration entropy, and a single phase is obtained in the as-designed composition of Bi0.8Sb0.8In0.4Se3. In addition, a Bi2Se3-Sb2Se3-In2Se3 diagram is depicted to tune the chemical components within a narrow range to further improve the TE performance of high-entropy selenides. Furthermore, multiple elements are found to broaden the solution limit of magnetic elements in Bi2Se3-based materials, and the exploration of magneto-thermoelectric effects is promoted due to the active research on fabricating a single-phase magnetic matrix. Finally, to discover new potential TE material family, multiple elements from different groups are randomly distributed in formulas for 1M:2M:1X:2X:3X and A:1B:2B:1X:2X to generate a material family of 720 Bi2Te3-type compounds and 360 Mg3Sb2-type compounds. Compounds with known crystal structures from the database are treated as a training set. A random forest method and Bayesian optimization are applied to train the model. Finally, a high prediction accuracy is achieved in the machine learning for predicting the compounds with structural similarities to Bi2Te3 and Mg3Sb2. Furthermore, an empirical rule based on the electronegativity differences of cationic and anionic elements is discussed. In this work, it was verified that the multiple elements approach is a fruitful and effective strategy for improving the performance of TE materials.