Cavitation-enhanced carbonation for nano-ZnO synthesis via an ultrasonic-jet coupled reactor: Machine learning...

Nano-sized zinc oxide (Nano-ZnO) is of significant interest in catalysis, adsorption, rubber reinforcement, and electronic devices due to its high specific surface area, tunable crystal structure, and superior interfacial properties. However, the traditional carbonization process is often constrained by inefficient gas-liquid-solid mass transfer, the formation of a dense "self-passivation layer" on particle surfaces, and severe agglomeration during crystal growth. Consequently, achieving a synergistic enhancement in reaction yield, specific surface area, and crystallite size reduction remains a significant challenge. In this study, a 20 kHz clamp-mounted horn-type ultrasonic vibration unit coupled with an ultrasonic-jet cavitation reactor was developed to address these issues. Through multi-scale synergistic intensification involving macroscopic turbulent shear and microscopic cavitation effects, the reactor significantly enhances the interfacial renewal rate, mass transfer efficiency, and nucleation density, thereby modulating carbonization kinetics and improving product microstructure. Based on the Box-Behnken Design (BBD), the effects of four key operating parameters-ultrasonic axial distance, solid-liquid ratio, incident pressure, and jet outlet height-on reaction yield, BET specific surface area, and crystallite size were systematically investigated. Four machine learning models-BP-ANN, SVR, RF, and XGBoost-were constructed and evaluated using the BBD experimental dataset. Comparative analysis revealed that the XGBoost model exhibited superior predictive performance (R² = 0.956), significantly outperforming the other three models. Furthermore, a multi-objective integrated optimization framework was established by coupling XGBoost with a genetic algorithm. The optimal process parameters were determined as follows: t = 90 min, T = 80°C, ultrasonic power = 700 W, ultrasonic transducer axial distance = 60.26 mm, solid-liquid ratio = 5.72:100, incident pressure = 0.756 MPa, and jet outlet height = 338.41 mm. Experimental validation demonstrated high consistency with model predictions, achieving a reaction yield of 94.92%, a specific surface area of 62.12 m² g^-1, and a crystallite size of 18.89 nm. Calorimetric measurements further showed that the net ultrasonic calorimetric power delivered to the liquid phase was 194.90 W. Based on the optimized treatment time of 90 min, the net ultrasonic energy input was estimated to be 1052.46 kJ. Mechanism analysis indicated that the synergistic ultrasonic-jet cavitation effectively disrupts the self-passivation layer, promotes efficient CO₂ mass transfer, enhances nucleation density, and inhibits secondary agglomeration. Consequently, the synthesized product exhibits higher crystallinity, a well-developed mesoporous structure, and a narrower crystallite size distribution. The proposed machine-learning-assisted genetic algorithm optimization strategy successfully addresses the challenges associated with multivariable nonlinear coupling. This work demonstrates the potential of this strategy in complex chemical process intensification and provides a novel technical route and theoretical basis for the green, controllable, and efficient preparation of high-performance Nano-ZnO.