Quantifying the dynamics of the 3D atomic structure using hidden Markov models in scanning transmission electron microscopy.

The aim of this project is to quantify the 3D dynamics of complex nanostructures at the atomic scale when they evolve over time via adatom dynamics, surface diffusion or during in situ experiments. This highly challenging and innovative objective will be reached by combining novel data-driven statistical methods with new image detection capabilities in aberration corrected scanning transmission electron microscopy. The ability to follow the motion of individual atoms in 3D in a realistic environment will clearly take the characterisation of nanomaterials to the next level. Quantitative 3D characterisation of nanostructures can nowadays be achieved with high reliability for systems under stationary conditions. Yet, major problems exist to get insight into the 3D dynamics because of the lack of physics-based models, detailed statistical analyses, and optimal design of experiments in a self-consistent computational framework. Machine learning using a hidden Markov model will enable us to explicitly describe structural changes as a function of time and to fully exploit the temporal information available in the observations. This unique approach will result in a precise characterisation of complex nanostructures in response to environmental stimuli such as temperature, pressure or gas composition. Clearly this is a prerequisite to understand the unique link between a material's structure and its properties, which is important for the design of a broad range of nanomaterials.

Keywords:METROLOGY, ELECTRON MICROSCOPY, NANOSTRUCTURES

Disciplines:Statistics, Modelling and simulation, Classical physics not elsewhere classified, Metrology, Quantum physics not elsewhere classified