(436f) Heterogeneous Ice Nucleation Using Forward Flux Sampling

Rare events are difficult to study in simulations. Rare events are called such because the length and time scales needed to observe the event are significantly longer than the event itself. This means that the expectation time for many rare events exceeds current computational capabilities by many orders of magnitude. One such rare event is ice nucleation. Understanding the role played by solid surfaces in ice nucleation is a significant step toward designing surfaces that can control ice nucleation. Our goal is to explain the mechanisms through which surfaces affect ice nucleation and growth. The challenge in this is to generate statistically significant number of liquid-to-ice transition paths. In our group, we use Forward Flux Sampling (FFS) to address this challenge. However, the adoption of FFS to complex, more realistic, systems requires significant job and data management that can be difficult to do using standard computational scripting tools. We have developed a unique computational framework that combines state-of-the-art Big Data technology with FFS to enable large scale calculations. Using this platform we have studied ice nucleation on silver iodide (AgI) and a modified kaolinite surface. Both these surfaces are a good atmospheric ice nucleating agents and have the same lattice match to ice. Traditionally, lattice match has been used a governing parameter to determine the ice nucleating efficiency of surface. Our extensive calculations show that the rate of ice nucleation on AgI and the modified kaolinite surface is different. In addition, the mechanism of nucleation is also different. This clearly demonstrates that lattice spacing is not a sufficient measure of ice nucleating efficiency. We have also performed committor analysis to identify the reaction coordinates relevant to this transition. To the best of our knowledge, these studies are the first large scale FFS calculations of heterogeneous ice nucleation using all-atom water models.