SurfaceScience21 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

O: Fachverband Oberflächenphysik

O 86: Mini-Symposium: Machine learning applications in surface science II

O 86.2: Invited Talk

Thursday, March 4, 2021, 11:00–11:30, R1

Chemisorbed or Physisorbed? Resolving surface adsorption with Bayesian inference and atomic force microscopy — •Milica Todorović — Department of Applied Physics, Aalto University, P.O. Box 11100, Aalto 00076, Finland

The knowledge on structure, bonding and properties of organic adsorbates to inorganic substrates underpins key technologies from catalysis through coatings to optoelectronics. Atomic force microscopy (AFM) and ab initio simulations are powerful tools for characterising molecular adsorption, but both struggle with complex bulky adsorbates where lack of chemical intuition and inconclusive imaging render the structure identification problematic. We address this challenge with Bayesian Optimization Structure Search (BOSS), a computational tool for global configurational structure search at surfaces and interfaces.

We employed BOSS to study the adsorption of (1S)-camphor on the Cu(111) surface, where AFM experiments recorded an unusual variety of images. From a single configurational search we retrieved 8 unique stable adsorbates. We discovered that camphor undergoes both covalent and dispersive binding to this substrate. Matching our findings to experimental data allowed us to categorise the AFM images into those associated with chemisorbed and physisorbed molecules. By simulating AFM images of the chemisorbed model structures, we identified three distinct adsorbates in the experimental images, further clarifying AFM image interpretation. This study illustrates how machine learning applications advance understanding in surface science by complementing both computation and experiment.