Homogeneous Freezing of Water Using Microfluidics.
Mark D TarnSebastien N F SikoraGrace C E PorterJung-Uk ShimBenjamin J MurrayPublished in: Micromachines (2021)
The homogeneous freezing of water is important in the formation of ice in clouds, but there remains a great deal of variability in the representation of the homogeneous freezing of water in the literature. The development of new instrumentation, such as droplet microfluidic platforms, may help to constrain our understanding of the kinetics of homogeneous freezing via the analysis of monodisperse, size-selected water droplets in temporally and spatially controlled environments. Here, we evaluate droplet freezing data obtained using the Lab-on-a-Chip Nucleation by Immersed Particle Instrument (LOC-NIPI), in which droplets are generated and frozen in continuous flow. This high-throughput method was used to analyse over 16,000 water droplets (86 μm diameter) across three experimental runs, generating data with high precision and reproducibility that has largely been unrepresented in the microfluidic literature. Using this data, a new LOC-NIPI parameterisation of the volume nucleation rate coefficient (JV(T)) was determined in the temperature region of -35.1 to -36.9 °C, covering a greater JV(T) compared to most other microfluidic techniques thanks to the number of droplets analysed. Comparison to recent theory suggests inconsistencies in the theoretical representation, further implying that microfluidics could be used to inform on changes to parameterisations. By applying classical nucleation theory (CNT) to our JV(T) data, we have gone a step further than other microfluidic homogeneous freezing examples by calculating the stacking-disordered ice-supercooled water interfacial energy, estimated to be 22.5 ± 0.7 mJ m-2, again finding inconsistencies when compared to theoretical predictions. Further, we briefly review and compile all available microfluidic homogeneous freezing data in the literature, finding that the LOC-NIPI and other microfluidically generated data compare well with commonly used non-microfluidic datasets, but have generally been obtained with greater ease and with higher numbers of monodisperse droplets.