Use of artificial intelligence models for the reduction of nanoparticle size in the synthesis of ZnO

BACKGROUND: This study evaluates the effectiveness of an artificial intelligence (AI) model for predicting the best experimental conditions to reduce particle size during the synthesis of ZnO nanoparticles. Firstly, an artificial neural networks (ANN) was trained using 52 experimental data from the synthesis of ZnO nanoparticles. The selected input variables were temperature, experimental time, and NaOH concentration, and the output variable was nanoparticle size. The performance of the ANN was measured with the root mean square error and mean absolute percentage error, and the obtained values for the selected ANN were 0.67% and 9.87%, respectively. These values were calculated by using real and predicted values. RESULTS: A genetic algorithm (GA) model was coupled with the ANN to find the best operational conditions for the reduced size of ZnO nanoparticles. According to the AI model, a temperature of 59 °C, an experimental time of 56 min, and a concentration of NaOH of 0.08 should be tested to obtain ZnO nanoparticles with 5.67 nm of diameter. After applying the conditions predicted by the model, ZnO nanoparticles with a mean diameter of 5.3 ± 0.4 nm were obtained. The results were confirmed by using several characterization methods, such as approximation of effective masses (5.3 nm), equation Debye–Scherrer (5.2 nm), and high-resolution transmission electron microscope (HR-TEM) selected micrograph (5.5 nm). The photocatalytic activity of the synthesized nanoparticles was evaluated using the synthetic dye thymol blue. The best discoloration efficiency was reached by the synthesized ZnO nanoparticles at 74%, while the commercial ZnO only achieved 58%. CONCLUSION: According to the ANN-GA model, it was possible to predict the experimental conditions needed to obtain ZnO nanoparticles with reduced sizes and excellent photocatalytic activity.