From Structure Search to Autonomous Experiments: Towards Function by Design at the Atomic Scale

Recent progress in machine learning and automation is beginning to fundamentally change how we explore and engineer materials at surfaces. In this talk, I will discuss two complementary directions towards a more predictive and autonomous surface science workflow.

First, I will present our work on structure prediction for molecule–surface interfaces, combining density-functional theory with machine-learning approaches such as SAMPLE to efficiently explore complex configurational spaces. These approaches enable the identification of experimentally relevant polymorphs and provide insight into the interplay between thermodynamics, kinetics, and growth conditions in hybrid inorganic–organic systems.

In the second part, I will discuss our efforts towards autonomous scanning probe microscopy experiments, where individual atoms or molecules are manipulated with atomic precision using an STM. Here, automation and AI-based decision making offer a path towards substantially increased throughput and complexity, potentially enabling the autonomous fabrication of artificial quantum structures and tailored nanoscale functionalities.

Finally, I will outline a broader perspective in which predictive modeling, real-time experimental feedback, and autonomous control are combined into “computer experiments” for atomic-scale materials design.