Understanding the intricate interactions between individual molecules and mastering their manipulation are pivotal for constructing atomically precise nanostructures with novel material properties. While chemical methods from solution-based synthesis exploit intermolecular interactions to form structures, they do not offer the flexibility needed for arbitrary arrangements. Scanning probe microscopy (SPM) stands out as a unique tool capable of imaging and manipulating individual molecules and atoms. However, the interaction processes involved are complex and vary significantly across different molecules and substrates. This variability poses challenges for constructing even moderately sized nanostructures.

This thesis will focus on leveraging reinforcement learning (RL) agents to study molecular manipulation on various substrates and to compare the agent’s learning effectiveness across different surfaces. Special emphasis will be placed on using transfer learning, where an RL agent is pre-trained on a simpler substrate and subsequently applied to a more complex substrate. This approach aims to explore the potential of utilizing previously acquired knowledge to enhance learning and adaptability in complex surface environments.

COMPENSATION: € 440,-- Forschungsbeihilfe for 6 months

CONTACT: Oliver Hofmann (o.hofmann@tugraz.at)