Stochastic modelling of frequency fluctuations in power grids of the future

 Predictions of the relative stabilities of polymorphs – crystals that differ only in the arrangement of the molecule – can be accurately made if we develop precise models for the energetics of interactions. Similarly, in studies of the nucleation, or of the transformation of one kind of polymorph into, small errors in the energy model can result in exponentially large errors in nucleation rates. Key to both phenomena are accurate energy models – the force-fields (FFs) – that are computationally inexpensive enough to be use for the long simulation times needed for these phenomena.

But there is a problem: We can derive accurate FFs based on first-principles calculations and physical models, but these still take too long to create. Moreover, we have not yet found an efficient way of including the effects of molecular flexibility in these models. In this project we will combine physical models with machine learning
to make the development of accurate physical FFs easy, and we will work with the developers of Tinker-HP to make these models available in a molecular dynamics package so that they can be used in simulations of the condensed phase.

The PhD students in these projects will need to have reasonably good theoretical and computational skills. They will be expected to develop Python code to automate calculations, and to bring large-data, machinelearning ideas into the research work. The students will benefit from the training programme and conferences held in the PHYMOL network. 

There are several PhD projects open in the group of Dr Alston J. Misquitta in the theory and applications of intermolecular interactions and force fields. The project on force-fields from machine-learned densities will be the co-supervision of Dr Devis Di Tommaso an expert on coarse-graining methods. The student will
be part of the School of Physics and Chemical Sciences and the London Centre for Theory and Simulation of Materials and Molecules, and will additionally be able to participate in the PHYMOL doctoral network: a €2.9M Marie Skłodowska–Curie Actions Doctoral Network on Intermolecular Interactions that was
awarded to Dr Misquitta as coordinator, and which involves 9 other European Institutes. 

Training & Development: Students can expect to be trained to interact with researchers between scientific communities. They will be trained in computer simulations, quantum chemistry, machine
learning, and scientific programming. The PhD project will provide opportunities for training in a wide range of method development and techniques and will equip the successful applicants with a highly desirable portfolio of scientific skills and associated transferable skills.