skip to content


Contact me on LinkedIn at Michael Davies


In my PhD I use computer simulations and artificial intelligence techniques such as deep learning to understand water and ice at the atomic scale. I’m particularly interested in the formation of ice, which may sound like an everyday phenomenon but is far more complicated than you might think.



Here is a simple observation: ice always melts at 0°C. Now here is a perhaps more surprising one: water doesn’t necessarily freeze at 0°C. Water can in fact remain in its liquid form all the way down to -40oC. This is because ice formation begins at the atomic scale, and here it is a “rare event”. Our everyday experience of ice forming closer to 0oC is due to presence of foreign materials in water. These facilitate the formation of ice much in the same way that catalysts facilitate chemical reactions.

So why does understanding how ice forms matter? Well in the atmosphere ice affects cloud albedo, lifetime and composition. This in turn makes it vital to Earth’s radiation budget and the modelling of future changes in climate. In fact, the biggest uncertainty in our climate models is the presence and properties of clouds! In industry, controlling ice formation presents great opportunity: cryopreservation is vital to numerous clinical applications, US state and local agencies spend over 2.3 billion p.a. on snow and ice control, and de-icing a single Boeing 747 airplane can cost up to $50,000!


During my PhD I have used high-throughput computational studies combined with data-driven analysis to explore how materials promote the formation of ice. Using deep learning we produced a model named IcePic which could accurately predict a materials ability to promote ice formation simply from images of room temperature water. What’s more it greatly outperformed human scientists in a head-to-head challenge. We have also investigated how materials can control the polymorph of ice produced. We designed substrates to nucleate desired polymorphs of ice, including the elusive cubic ice which has yet to be created from liquid water in the laboratory.



Selected Publications 

Routes to cubic ice through heterogeneous nucleation
MB Davies, M Fitzner, A Michaelides – Proc Natl Acad Sci U S A (2021) 118, e2025245118


Telephone number

01223 336530 (shared)