Machine learning accelerated analysis of microwave-assisted synthesis parameters on calcium carbonate particles

The synthesis of calcium carbonate (CaCO₃) particles is crucial for tailoring their structural and chemical properties for specific advanced applications. However, precision synthesis is challenging due to bulk crystallization effects, which makes it difficult to predict the outcomes of chemical synthesis. Microwave-assisted synthesis offers significant advantages, including faster reaction times, energy efficiency, and potential for solventless reactions, making it valuable in various chemical fields. The advent of machine learning has shifted materials research from traditional trial-and-error methods to precise computational science. In this study, we use microwave technology to examine the effects of temperature, pressure, concentration, and time on CaCO₃ particle synthesis, focusing on morphology, size, shape, and phase composition. Machine learning models, such as decision trees, random forests, and gradient boosting, enable us to predict particle phases based on initial parameters and identify key factors influencing particle formation. Results indicate that temperature is more critical than concentration, pressure, and time for CaCO₃ polymorph formation. The dominant phase transitions from vaterite to aragonite at 100 °C. At temperatures above 100 °C, higher concentrations result in increased pressure, leading to the synthesis of smaller particles. This is due to enhanced crystal nucleation density and higher kinetic energy, which inhibit crystal growth. SHapley Additive ExPlanations (SHAP) analysis was used to enhance the interpretability of regression models and understand the contributions of each variable to CaCO₃ phase composition predictions. This research pioneers the integration of microwave-assisted synthesis and machine learning to precisely control and understand the properties of micro- and nano-sized CaCO₃ particles.