UC Santa Barbara

The quest for sustainable, efficient energy systems is more pressing than ever as we step into the 21st century. Finding the suitable materials for a specific technology is a very challenging task. The required properties can be unique to particular chemical compositions and structures. Furthermore, achieving high performance often demands the precise alignment of multiple interacting properties, which can have competing effects. Additional economic constraints, such as the cost of raw materials and the manufacturing process, further add to the challenges. Traditionally, this journey has been arduous, relying on slow and laborious experimental techniques. One pivotal approach that has revolutionized the pace and efficiency of this discovery process is High-throughput screening (HTS). HTS is a powerful and systematic methodology that enables the rapid evaluation of a large number of materials in a relatively short period by leveraging large databases, accurate and efficient computational techniques including ab initio calculations and machine learning. When large libraries of materials are screened based on specific properties or descriptors, only a limited amount of the most promising ones pass to the final stage for further validation through expensive calculations and experiments. This thesis aims to develop HTS pipelines on finding promising candidates for two energy applications: thermoelectric materials that convert heat into electricity, allowing for waste heat recovery, and magnetocaloric materials that serve an environmental friendly cooling technology. The integration of predictive models and screening pipelines significantly reduces the time and cost associated with material discovery, transforming the traditional methods of discovering and developing novel materials.

In the pursuit of innovative solutions for sustainable energy, thermoelectric materials hold undeniable potential to transform energy generation. They possess the unique ability to convert thermal energy into electricity, offering promising avenues for harnessing waste heat and enhancing energy efficiency. Achieving high thermoelectric performance requires efficient manipulation of thermal conductivity and a fundamental understanding of the microscopic mechanisms of phonon transport in crystalline solids. One of the major challenges in thermal transport is achieving ultralow lattice thermal conductivity. We use the anti-bonding character of the highest-occupied valence band as an efficient descriptor for discovering new materials with an ultralow thermal conductivity. We first examine the relationship between anti-bonding valence bands and low lattice thermal conductivity in model systems PbTe and CsPbBr3. Then, we perform a high-throughput search in the Materials Project database and identify over 600 experimentally stable binary semiconductors and report the anti-bonding strength in their valence bands. From our candidate list, we conduct a comprehensive analysis of the chemical bonds and the thermal transport in the XS family, where X=K, Rb, and Cs are alkaline metals. These materials all exhibit ultralow thermal conductivities less than 1~W/(m K) at room temperature despite simple structures. We attribute the ultralow thermal conductivity to the weakened bonds and increased phonon anharmonicity due to their anti-bonding valence bands. Our results provide chemical intuitions to understand lattice dynamics in crystals and open up a convenient venue towards searching for materials with an intrinsically low lattice thermal conductivity.

In another search for new energy materials, magnetocaloric materials are requested as promising candidates for energy-efficient and eco-friendly cooling technologies. These materials exhibit the magnetocaloric effect, where their temperature changes in response to the addition or removal of an external magnetic field. This unique property opens doors to environmentally friendly refrigeration systems since it eliminates the need for refrigerants with high global warming potential. Magnetic cooling based on the magnetocaloric effect is a promising solid-state refrigeration technology for a wide range of applications in different temperature ranges. Previous studies have mostly focused on near room temperature (300K) and cryogenic temperature (<10K) ranges, while important applications such as hydrogen liquefaction call for efficient magnetic refrigerants for the intermediate temperature 10K to 100K. For efficient use in this range, new magnetocaloric materials with matching Curie temperatures need to be discovered, while conventional experimental approaches are typically time-consuming and expensive. In this work, we report a computational material discovery pipeline based on a materials database containing more than 6000 entries auto-generated by extracting reported material properties from literature using a large language model. We then use this database to train a machine learning model that can efficiently predict magnetocaloric properties of materials based on their chemical composition. We further verify the magnetocaloric properties of predicted compounds using ab initio atomistic spin dynamics simulations to close the loop for computational material discovery. Using this approach, we identify 11 new promising magnetocaloric materials for the target temperature range. The joined force of AI advances and high-throughput search pipelines enables more efficient prediction and iteration. Together they are revolutionizing the conventional ways to discover new materials.