Development of Droplet-Based Microfluidics to Assess the Stochastic Behavior of Crystal Nucleation

Abstract

Nucleation, the very early stage of crystallization processes, is of fundamental importance on defining crystal properties such as purity, size distribution and polymorphism. Since it is a stochastic phenomenon, a number of challenges to obtaining experimental reproducibility exists. In this thesis we have proposed to study nucleation in microdroplets. Droplet-based microfluidics is a high-throughput technique that enables generating large numbers of discrete and nearly identical micro-crystallizers, a feature which is highly desirable for constructing representative statistics, thereby reliably characterizing nucleation kinetics. We have studied primary nucleation of an organic substance from aqueous solutions. Information about the stochastic nature of nucleation was experimentally obtained from thousands of droplets, generated in different microfluidic devices, developed within the scope of this thesis. Despite the number, error due to the impossibility of sampling the entire population will invariably be made. We have therefore proposed a thorough statistical analysis to account not only for the stochastic nature of nucleation, but also for the variability in microdroplet volumes and the uncertainty of the automated image analyisis procedure in the estimation of the physical quantities of the process. As source of uncertainty, it is important to measure and report droplet volume distribution in detail and to control it tightly; a chapter of this thesis was dedicated to characterize volume and shape of droplets generated in microfluidic T-junctions to evaluate whether their inherent volume polydispersity was reproducible and suitable in order to minimize its impact on the propagation of experimental uncertainties. Using the developed statistical methodology, we have also studied nucleation in microdroplets at different fluid dynamic regimes. We have first analyzed the importance of having a lubricating film surrounding the flowing droplet, since it provokes counter-flow patterns which enhance fluid shear improving mass transfer. We have observed that nucleation in flow is much faster than nucleation at quiescent conditions; a difference in nucleation rates of about three orders of magnitude was reported. We have discussed reasons for this occurrence by calling upon sources of mesostructured clusters ordering, which in view of the two-step mechanism theory is the initiation to nucleation. Summarizing, this thesis comes to contribute with a statistical methodology to cope with distribution of droplet volumes and detection techniques via automated image analysis, which makes microfluidic platforms even more attractive to study stochastic behavior of processes as nucleation. In addition, as long as this technique is employed prudently and attention is given to its possible pitfalls, it provides an unique framework which represents a step forwards to the realization of a fully versatile microfluidic platform for applications in crystallization.