In the world of metal oxide particle synthesis, researchers have long relied on traditional methods that involve intuition or trial-and-error. These approaches are often inefficient and time-consuming. However, a team of researchers from PNNL has developed a new approach that leverages data science and machine learning techniques to streamline the development process.
Their innovative approach addresses two main issues: identifying feasible experimental conditions and predicting potential particle characteristics based on synthetic parameters. To achieve this, the researchers developed an ML model that can accurately predict iron oxide outcomes based on synthesis reaction parameters.
This groundbreaking approach represents a paradigm shift in metal oxide particle synthesis and has the potential to significantly reduce the time and effort required for ad hoc iterative synthesis approaches. By training the ML model on careful experimental characterization, the approach demonstrated remarkable accuracy in predicting iron oxide outcomes based on synthesis reaction parameters. Additionally, the search and ranking algorithm used revealed previously overlooked factors that play a crucial role in determining phase and particle size, such as pressure applied during synthesis.
Juejing Liu et al’s study, “Machine learning assisted phase and size-controlled synthesis of iron oxide particles,” was published in the Chemical Engineering Journal (2023) with DOI: 10.1016/j.cej.2023.145216