MAP-DESIGN

Many properties of materials, such as the taste and appearance of chocolate, the solubility of artificial sweetener, or the effectivity of pharmaceutical drugs, depends sensitively on the way the building blocks of that material arrange with respect to each other. One of the fundamental challenges when developing new materials, therefore, is to optimize that arrangement. Here, the main difficulty is that the structure, also called polymorph itself is strongly dependent on the processing conditions of the material, including e.g., the temperature, the employed solvent, etc. Since there are only very few systematic rules that describe how the processing conditions impact the polymorph formation, so far most processing conditions need to be found experimentally using trial and error. Of course, this is very cost- and time-intensive. The purpose of the present project is, therefore, to predict optimal processing conditions using quantum mechanical simulations and advanced data analysis methods. Exemplarily, we will focus on thin films of organic molecules deposited on inorganic substrates. Such systems are highly relevant for the field of organic electronics and are used, e.g., in OLED-TVs. Such system show a particularly rich polymorphism, often including phases that would not be stable in the isolated components, but that show significantly superior properties. Although a full determination of all possible polymorphs is in principle an intractable problem, by limiting to that class of materials we can invoke assumptions that simply the problem. We can then employ a new, innovative structure search algorithm to determine the most important polymorphs. The data from this step is subsequentially analyzed with machine learning methods in order to map their properties onto experimental handles, such as temperature, stress, or solvent polarity. This map is then explored using routing algorithms, similar to those used by Google Maps, in order to provide ideal recpies (routes) towards the desired structure.

Public Summary Statement - FWF Project MAP-DESIGN (Y1157) Designing Metastable Polymorphs for Organic Electronics The FWF-funded project MAP-DESIGN, led by Prof. Oliver T. Hofmann at Graz University of Technology, has made groundbreaking contributions to the design and control of organic materials used in flexible electronics. The project's primary goal was to understand and steer metastable polymorphs-different structural forms that organic molecules can adopt-to improve the performance and reliability of devices like organic solar cells, transistors, and sensors. In organic electronics, the precise arrangement of molecules in thin films critically influences device performance. However, these molecules often crystallize into multiple polymorphs, some of which may not exhibit optimal electronic properties. Traditional experimental approaches to control these structures are time-consuming and often rely on trial and error. MAP-DESIGN addressed this challenge by using computational tools to predict and direct the formation of desirable polymorphs. The project developed innovative machine learning algorithms to explore the complex energy landscapes that govern molecular arrangements. These tools simulate how external factors-such as temperature, electric fields, and mpressure -affect how molecules organize themselves. By identifying energetically favorable pathways, the researchers could suggest processing conditions that reliably guide materials into targeted polymorphic forms. A particularly novel achievement was the adaptation of route-finding algorithms (similar to GPS navigation systems) to steer materials through complex energy landscapes toward desired polymorphs. This concept allows scientists to design processing steps that avoid unwanted structures and ensure reproducibility in device manufacturing. Key advances and publications from MAP-DESIGN have demonstrated: How electric fields can selectively stabilize certain polymorphs at organic/inorganic interfaces, opening new avenues for material tuning. The use of reinforcement learning for intelligent manipulation of single molecules, showcasing the potential for AI-driven material design. Insights into the role of adatoms (individual atoms on surfaces) in facilitating or hindering the adsorption of organic molecules-crucial for understanding and optimizing device interfaces. By bridging theoretical predictions with experimental realities, the MAP-DESIGN project has helped establish a more predictive, efficient, and rational approach to material design. The integration of machine learning, quantum simulations, and data-driven modeling marks a shift from passive material discovery to active design, where desired properties can be engineered from the atomic level upward. For more details and a list of resulting publications, visit the official project website: https://www.tugraz.at/en/projekte/map-design