91福利

Biography

Albert P. Barto虂k is an Associate Professor at the 91福利. He earned his Ph.D. degree in physics from the University of Cambridge in 2010, his research having been on developing interatomic potentials based on ab initio data using machine learning. He was a Junior Research Fellow at Magdalene College, Cambridge, and later a Leverhulme Early Career Fellow. Before taking up his current position, he was a Research Scientist at the Science and Technology Facilities Council. His research focuses on developing theoretical and computational tools to understand atomistic processes.

Research Interests

My main areas of research are materials modelling on the atomistic scale, developing methods to accelerate simulations while making accurate predictions. I am also interested in electronic structure calculations, density functional theory and related developments. I am using tools from machine learning to build fast and accurate models that are based on large amounts of ab initio data.

I am developing physics-inspired machine learning models for atomistic simulations. When the very flexible ML models are suitable constrained by embedding physics into them, we can expect results which are not only predictive but give us in-depth understanding of matter at the microscopic level. I was an early adopter: I am one of the main developers of the (GAP) method, one of the first machine-learned interatomic potential (MLIP) frameworks which has been successfully applied to a broad range of materials science problems. In addition, the software package and the method has been licensed to various companies.

I also invented (SOAP), a descriptor formalism to characterise chemical environments around atoms, which has inspired many other similar tools, spilling over to unrelated areas such as computer vision. SOAP and its descendants are widely used in machine learned interatomic potentials and simulation analysis tools.

Having worked in ML, where regularisation is essential, made me recognise the importance of the smoothness of density functional approximations. So we the Strongly Constrained and Appropriately Normed exchange-correlation functional which inspired a whole family of RSCAN functionals. I am part of the CASTEP Developer Group and contribute to the CASTEP DFT package.

Apart from method development, I apply all of this in my research. I am interested in defects and how they influence mechanical and electronic properties of materials. We have studied the plastic deformation mechanisms of various metals, as well as defect physics and migration of diamond. I am keen using MLIPs to predict the phase diagram of materials using thermodynamics calculations. I have an interest in molecular crystals and their characterisation, in particular how atomistic modelling can aid the interpretation of experimental techniques. I am working with solid-state NMR experimentalists to generate accurate structure models of various industrially relevant systems.

For an up-to-date overview of my currently active projects, publications and teaching activities, check my or my .