Perspective: Web-based machine learning models for real-time screening of thermoelectric materials properties

Abstract

The experimental search for new thermoelectric materials remains largely confined to a limited set of successful chemical and structural families, such as chalcogenides, skutterudites, and Zintl phases. In principle, computational tools such as density functional theory (DFT) offer the possibility of rationally guiding experimental synthesis efforts toward very different chemistries. However, in practice, predicting thermoelectric properties from first principles remains a challenging endeavor [J. Carrete et al., Phys. Rev. X 4, 011019 (2014)], and experimental researchers generally do not directly use computation to drive their own synthesis efforts. To bridge this practical gap between experimental needs and computational tools, we report an open machine learning-based recommendation engine (http://thermoelectrics.citrination.com) for materials researchers that suggests promising new thermoelectric compositions based on pre-screening about 25 000 known materials and also evaluates the feasibility of user-designed compounds. We show this engine can identify interesting chemistries very different from known thermoelectrics. Specifically, we describe the experimental characterization of one example set of compounds derived from our engine, RE12Co5Bi (RE = Gd, Er), which exhibits surprising thermoelectric performance given its unprecedentedly high loading with metallic d and f block elements and warrants further investigation as a new thermoelectric material platform. We show that our engine predicts this family of materials to have low thermal and high electrical conductivities, but modest Seebeck coefficient, all of which are confirmed experimentally. We note that the engine also predicts materials that may simultaneously optimize all three properties entering into zT; we selected RE12Co5Bi for this study due to its interesting chemical composition and known facile synthesis.

Understanding and designing magnetoelectric heterostructures guided by computation: progresses, remaining questions, and perspectives

“…Specifically, although big-data approaches have been utilized to discover predictive models and design principles from the data of atomic structures and corresponding responses generated by quantum computations, [95][96][97][98][99][100][101][102][103] structural and functional imaging data obtained by scanning probe microscopy, 1, 169-171 X-ray diffraction data, 172,173 and even data from archived laboratory notebooks, 174 they have been rarely used to understand and harness the data of materials microstructure and properties generated by mesoscale materials modeling. We believe that the latter will offer opportunities as many as (and possibly larger than) what integrating big-data approaches with quantum computations have brought and will bring.…”

“…In fact, using machine learning to discover statistically correct predictive model from quantum computations data has been explored since the year 2003. [95][96][97][98][99][100][101][102][103] Furthermore, a statistically correct predictive model can also be discovered from high-quality experimental data. More discussions of such datadriven computational materials design (the bottom layer in Fig.…”

Data-Driven Materials Investigations: The Next Frontier in Understanding and Predicting Fatigue Behavior

“…The models may be trained directly on experimental data 13,14 or act as very fast surrogates for more expensive physics-based simulation such as density functional theory 15 . ML has successfully guided the experimental discovery of novel Heusler alloys, 16 Ni-based superalloys, 17 and shape-memory alloys, 18 among many other application areas.…”

Human versus Robots in the Discovery and Crystallization of Gigantic Polyoxometalates

“…To our knowledge,itis the first time that such an active data acquisition strategy is applied in this context and compared with human experimenters.M achine learning methods have previously been used as at ool of optimization [9] and af aster data mining technique for extensive databases. [10][11][12][13][14] It is important to note that our approach should not be mistaken for high-throughput screening as it uses machine learning techniques capable of abstracting problems rather than abrute force increase of processing speed. We instead suggest this approach should be viewed as "intelligent throughput" since not all the possible experiments are done,and only those chosen by the algorithm are explored and the system effectively learns as the experiment continues similar to how an expert chemist would work.…”