Exploring Monodisperse Condition for Ceramic Nanoparticles Using Hydrothermal Synthesis and Machine Learning

Abstract

In this research, near-monodispersed iron oxide nanoparticles were synthesized using the alkaline-treated hydrothermal method under various conditions. Machine learning, specifically support vector regression (SVR), was employed with the hydrothermal temperature, time, ammonia solution amount, and Ole/M ratio as explanatory variables, and CV values as the response variable. The experimental results showed a CV value of 11.4% under conditions of hydrothermal temperature 220°C, hydrothermal time 24 hours, ammonia solution amount 3.00 mL, and Ole/M ratio 1.00. SVR was performed to estimate the relationship between CV values and experimental conditions. However, the obtained CV value was 25% which was higher than expected. To address this, a new evaluation index that combines particle size and CV value was introduced for estimation, however, the predicted value did not exceed the actual measured value. Finding conditions that yield lower CV values is challenging with the current method. On the other hand, the CeO₂ nanoparticles synthesized using the alkaline-treated hydrothermal method showed a broader particle distribution. This highlights the importance of selecting appropriate initial materials. As a next step, further exploration of synthesis conditions near the lowest CV, validation of the experimental results, and refinement of the dataset will be conducted.