Use this URL to cite or link to this record in EThOS:
Title: Merging data-driven and computational methods to understand ice nucleation
Author: Fitzner, Martin
ISNI:       0000 0004 7661 0673
Awarding Body: UCL (University College London)
Current Institution: University College London (University of London)
Date of Award: 2018
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
Heterogeneous ice nucleation (IN) is one of the most ubiquitous phase transitions on earth and impacts a plethora of fields in industry (e.g. air transport, food freezing and harsh-weather operations) and science (e.g. freeze avoidance of animals, cryobiology, cloud research). Still to date, we are lacking reliable answers to the question: What is it at the molecular scale that causes an impurity to facilitate the freezing process of supercooled liquid water? In this thesis we make headway towards identifying such microscopic principles by performing computational studies combined with data-driven approaches. In chapter 3 we screen a range of model substrates to disentangle the contributions of lattice match and hydrophobicity and find that there is a complex interplay and an enormous sensitivity to the atomistic details of the interface. In chapter 4 we show that the heterogeneous setting can alter the polymorph of ice that forms and introduce the concept of pre-critical fluctuations, yielding new ideas to design polymorph-targeting substrates. Chapter 5 deals with the liquid dynamics before and during the nucleation event, an aspect of nucleation that mostly goes unrecognized. We show that the homogeneous nucleation event happens in relatively immobile regions of the supercooled liquid, a finding that opens new avenues to understand and influence heterogeneous nucleation by targeting dynamics rather than structure. Finally, Chapter 6 builds on the large amount of data created during this project in that we combine all previously simulated systems and devise a machine-learning approach to find the most important descriptors for their ice nucleation activity (INA). With this we identify new microscopic guidelines and demonstrate that the quantitative prediction of heterogeneous INA is in reach. The unveiling of a computational artifact that potentially affects many computational interface studies is also part of this thesis.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available