Deep learning has fostered many novel applications in materials informatics. However, the inverse design of inorganic crystals, i.e. generating new crystal structure with targeted properties, remains a grand challenge. An important ingredient for such generative models is an invertible representation that accesses the full periodic table. This is challenging due to limited data availability and the complexity of 3D periodic crystal structures. In this paper, we present a generalized invertible representation that encodes the crystallographic information into the descriptors in both real space and reciprocal space. Combining with a generative variational autoencoder (VAE), a wide range of crystallographic structures and chemistries with desired properties can be inversedesigned. We show that our VAE model predicts novel crystal structures that do not exist in the training and test database (Materials Project) with targeted formation energies and band gaps. We validate those predicted crystals by first-principles calculations. Finally, to design solids with practical applications, we address the sparse label problem by building a semi-supervised VAE and demonstrate its successful prediction of unique
top of page
Direct patterning of thermoelectric metal chalcogenides can be challenging and is normally constrained to certain geometries and sizes. Here we report the synthesis, characterization, and direct writing of sub-10 nm wide bismuth sulfide (Bi2S3) using a single-source, spin-coatable, and electron-beam-sensitive bismuth(III) ethylxanthate precursor. In order to increase the intrinsically low carrier concentration of pristine Bi2S3, we developed a self-doping methodology in which sulfur vacancies are manipulated by tuning the temperature during vacuum annealing, to produce an electron-rich thermoelectric material. We report a room-temperature electrical conductivity of 6 S m–1 and a Seebeck coefficient of −21.41 μV K–1 for a directly patterned, substoichiometric Bi2S3 thin film. We expect that our demonstration of directly writable thermoelectric films, with further optimization of structure and morphology, can be useful for on-chip applications.
Hybrid (organic-inorganic) materials have emerged as a promising class of thermoelectric materials, achieving power factors (S2σ) exceeding those of either constituent. The mechanism of this enhancement is still under debate, and pinpointing the underlying physics has proven difficult. In this work, we combine transport measurements with theoretical simulations and first principles calculations on a prototypical PEDOT:PSS-Te(Cux) nanowire hybrid material system to understand the effect of templating and charge redistribution on the thermoelectric performance. Further, we apply the recently developed Kang-Snyder charge transport model to show that scattering of holes in the hybrid system, defined by the energy-dependent scattering parameter, remains the same as in the host polymer matrix; performance is instead dictated by polymer morphology manifested in an energy-independent transport coefficient. We build upon this language to explain thermoelectric behavior in a variety of PEDOT and P3HT based hybrids acting as a guide for future work in multiphase materials.
bottom of page