Public defence in Engineering Physics, M.Sc. (Tech) Niko Oinonen

Title of the doctoral thesis: Automating high-resolution atomic force microscopy image interpretation

Doctoral student: Niko Oinonen
Opponent: Prof. Philip Moriarty, University of Nottingham, England
Custos: Prof. Adam Foster, Aalto University School of Science, Department of Applied Physics

Thesis available for public display 10 days prior to the defence at: https://aaltodoc.aalto.fi/doc_public/eonly/riiputus/

Machine learning enhances atomic force microscopes' ability to resolve atomic structures of molecules

Increasing our understanding of the structure of matter at the nanoscale is important both for fundamental physics research and for the future development of nanotechnologies. An important component for achieving this goal is the development of microscopy techniques capable of resolving the structure of materials down to the atomic scale. Atomic force microscopy is among the leading techniques in this direction, being capable of distinguishing individual atoms in individual molecules. However, the usefulness of this technique is currently limited mostly to imaging of relatively flat molecules due to the difficulty in interpreting the measured signal for more complex structures. 

This thesis addresses the issue of interpretability in high-resolution atomic force microscopy experiments through the use of machine learning models. While real experimental data is too scarce to use in training machine learning models, it is possible to do realistic simulations of atomic force microscopy images at a very high through-put rate using the so-called probe-particle model. The work in this thesis uses the probe-particle model to construct a large database of simulated atomic force microscopy images and subsequently uses this data to train several machine learning models for predicting the atomic structure and electrostatic properties of molecules directly from atomic force microscopy images. The simulation-trained models are also applied to real atomic force microscopy images. The results are promising and represent a first step in automating the analysis of experimental data in atomic force microscopy, which is likely to play a significant role in increasing the applicability and ease of use of the measurement technique.

Contact details:

Doctoral theses in the School of Science: https://aaltodoc.aalto.fi/handle/123456789/52