Optimisation of microfluidic synthesis of silver nanoparticles via data-driven inverse modelling

The informed choice of conditions to produce nanoparticles with specific properties for targeted applications is a critical challenge for nanoparticle manufacture. In this study, this problem is addressed taking as an example the synthesis of silver nanoparticles (AgNPs) using an inverse modelling approach, where a polynomial function was constructed using synthesis parameters, including nucleation ($k_1$) and growth ($k_2$) constants, collection/storage temperature (T), Reynolds number ($\mathrm{Re}$), and the ratio of Dean number to Reynolds number ($\mathrm{De}/\mathrm{Re}$). This function was used to identify the parametric space for hydrodynamic conditions, with other parameters being held constant while employing Latin Hypercube Sampling (LHS) to explore initial guesses in the $\mathrm{Re}$ and $\mathrm{De}/\mathrm{Re}$ domain. Data assimilation techniques were then applied to incorporate experimental data into the model, facilitating parameter identification and optimization, which resulted in improved predictions and reduced uncertainty. The inverse model was evaluated against unseen data, demonstrating good consistency between forward and inverse modelling paths for AgNP size prediction. Experimental data was used to validate the capability of the model to design AgNPs of a targeted size using specific set of chemicals in a microfluidic system. The integration of LHS and inverse modelling through data assimilation is shown to provide a robust framework for addressing uncertainty in nanoparticle manufacture.