Abstract
Our planet has massive ice resources that play a crucial role in mitigating the greenhouse effect and providing an environment for atmospheric chemical reactions. These reactions usually occur on the ice surfaces and are accompanied by the "quasi-liquid layer" (QLL) generated below the ice melting temperature. This QLL is also involved in the ice nucleation and growth on different ice surfaces at low vapour pressures. The most common ice crystal in our ambient environment is hexagonal ice (Ih). The growth competition between ice Ih basal and prismatic surfaces results in the ice morphologies varying between plates and columns. Therefore, our primary goals are to characterize the properties of the QLL and investigate the ice crystal growth on different ice surfaces.To unravel the mechanisms of ice growth from the vapour at the molecular scale, the initial work of this project was to systematically study the characteristics of the QLL on ice surfaces in a wide range of supercooling temperatures using molecular dynamics (MD) simulations. It analyzed the results using a range of order parameters and a deep-learning neural network framework (DeepIce) to distinguish the ice-like and liquid-like atomic environments. Then, the ice crystallisation process was investigated using the MD and well-tempered metadynamics methods to give insights into the QLL/ice interface dynamics. Through selecting an efficient collective variable (environment similarity), the free energy surfaces associated with QLL melting and recrystallising were recovered. The relationship between ice growth rates on different ice surfaces and a range of temperatures was obtained. The growth kinetics of varying ice surfaces were compared and discussed. Based on the direct coexistence MD method, the free energy and chemical potential differences between ice Ih and water in the ice bulk coexistence systems were investigated using the on-the-fly probability enhanced sampling (OPES) simulations. The melting temperature of ice Ih was also confirmed by fitting the curve of the chemical potential differences. In addition, by combining the direct coexistence MD technique with an off-the-shelf deep learning neural network potential for water (DNNP), the ice Ih melting temperature for this DNNP model was validated and compared with the TIP4P/Ice water model. The investigation of the QLL/ice interface’s dynamic equilibrium and the ice crystal growth mechanisms was expected to support our community in understanding ice further.
| Date of Award | 1 Apr 2024 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Carla Molteni (Supervisor) & affiliated academic (Supervisor) |