Process optimization of preparing silica particles with uniform distribution in sub-micron sizes using artificial neural networks

Optimal concentrations of reactants for preparing sub-micron uniform silica particles were predicted using a process optimization algorithm. Silica particles were synthesized from a mixture of tetraethyl orthosilicate (TEOS), NH3, and H2O using a sol–gel reaction. Different sets of the reactant concentrations were selected within an operating range and were designed to evaluate the particle size (PS) and the particle size distribution (PSD) data. The relationship between the reactant concentrations and PS and PSD of the synthesized samples can be constructed by an artificial neural networks (ANN) modeling approach. The built ANN model was then used to predict PS and PSD values corresponding to the reactant concentrations. The optimization problem of determining a combination of the reactant concentrations to fabricate a certain PS with narrow PSD was solved by the ANN model and an optimization method. The optimal reactant concentrations were estimated and it showed that [TEOS] and [NH3] needed to be controlled in smaller values; meanwhile, the produced particles between 200 and 500 nm can be obtained by fixing [TEOS] and [NH3] and only adjusting [H2O].