Gianluca Seaford
PhD Title: Atomistic insight into nucleation and electrochemistry: Machine learning multiscale simulation
PhD Supervisors: Prof. Nicholas Hine (Physics), Prof. David Quigley (Physics) and Dr. Alex Robertson (Physics)
Background:
I am a first-year PhD student at the CDT for Modelling Heterogeneous Systems (HetSys) focusing on simulating nucleation and electrochemical processes occuring across several lengthscales and timescales via machine-learned interatomic potentials (MLIPs). My interests lie in the application of equivariant message-passing neural networks (MPNNs) to generate machine-learned interatomic force fields (ML-FFs) to simulate systems too complex to be accessed by traditional DFT approaches.
Before joining HetSys I completed my Master's Degree in Mathematics and Physics at Warwick with my dissertation focusing on performing constrained minimum-energy conical intersection (MECI) searches on machine-learned potential energy surfaces (ML-PESs) of Mycosporine-glycene. The MACE model, an O(3)-equivariant many-body neural network, was used to generate PESs for the constrained MECI search. This work was supervised by Prof. Nicholas Hine and resulted in a high first-class award for this work. The poster accompanying this dissertation is available here.
I am currently part of the Hine research groupLink opens in a new window (Warwick Theory Group) and the Robertson research groupLink opens in a new window (Warwick Microscopy Group).
| Year | Program | Institution | Result |
| 2024-2028 | PhD, Modelling of Hetrogeneous Systems | University of Warwick | in progress |
| 2024-2025 | PGDip, Modelling of Hetrogeneous Systems | University of Warwick | in progress |
| 2020-2024 | MMathPhys (with Honours), Mathematics and Physics | University of Warwick | 2.i |
Current Research:
Developing battery technologies requires atomistic insight into electrochemistry, nucleation, and degradation, but simulation is presented with a challenging combination of lengthscale, timescale and accuracy demands. This presents a great opportunity for Scientific Machine Learning to work closely with experimental techniques such as transmission electron microscopy, and to learn to simulate nucleation and electrochemistry processes.
This project aims to develop and apply MLIPs to generate 2-dimensional simulated training data in order to construct multiscale ML models for nucleation. The machine-learning process will be constrained and informed by in-situ S-TEM images capturing atomic-level electrochemical processes occuring during the electrodeposition of Au onto 2-dimensional materials. These S-TEM images will be taken by the Robertson research group.
This work is made possible through an EPSRC grant to the EPSRC CDT in Modelling Heterogeneous Systems (HetSys).
