An Automated Adsorption Workflow for Semiconductors

Adsorption is a crucial step in a broad range of processes, including heterogeneous catalysis, energy storage, and solid-state synthesis. Because of the complexity of the adsorption mechanism, the study of such processes using first-principles calculations often requires a high-throughput approach, including exhaustive surface and adsorption-site search algorithms. This can become an involved and expensive undertaking, due to the large number of possible adsorption structure configurations and the elevated computational cost of calculations performed on such systems.<br/> <br/>We present here the development of a new and improved automated adsorption workflow, which employs density functional theory (DFT) calculations to generate a comprehensive set of adsorption properties for a given bulk material and set of adsorbates. Starting from a bulk structure and a set of adsorbate structures, and requiring minimal user supervision, the workflow performs DFT geometry optimizations for the bulk, the slabs corresponding to all the unique surfaces identified in the output bulk structure, and the adsorption structures generated as a result of the adsorption site searching step for each surface and each input adsorbate. Additionally, a set of more accurate static calculations is performed for each surface and adsorption structure with the purpose of obtaining a more reliable set of density of states data. The final step in the workflow for each adsorption structure is an analysis step which collects and organizes the data computed throughout the workflow and saves it to a personal user database that can be easily queried. <br/> <br/>The workflow is particularly designed to be suitable for the study of semiconductors, a class of materials previously overlooked in the development of high-throughput adsorption methodologies. The less homogeneous electron density at the surface compared to metallic systems, the potential dipoles due to polar terminations, and the generally more complex surfaces which exhibit stronger site dependence of computed properties render the modelling of adsorption processes on semiconductor surfaces a particularly challenging task. To address these problems and ensure the applicability of the workflow to semiconducting materials, the input adsorption structures are rigorously designed such that the success of the geometry optimization calculation is maximized, even in cases where surface interactions are weak. Moreover, the workflow calculates a series of properties that are particularly relevant to semiconductors and photocatalysis: energy-based descriptors (such as the adsorption energy), geometric descriptors (including translation vectors and coordination numbers), electronic descriptors (for example, the orbital and elemental make-up of the conduction band minimum and valence band maximum), and charge-specific descriptors (which quantify the amount of charge transferred from the surface to the adsorbate). Our group has successfully employed this method to study the adsorption of relevant species on potential CO₂ reduction photocathodes (see abstract from Siron et al.) and we present it here as a robust materials discovery tool for any applications that involve adsorption processes.