Apr 8, 2025
11:00am - 11:30am
Summit, Level 4, Room 424
Christoph Kreisbeck1,Manuel Tsotsalas1,2,Yi Luo1,2
Aixelo, Inc1,Karlsruhe Institute of Technology (KIT)2
Christoph Kreisbeck1,Manuel Tsotsalas1,2,Yi Luo1,2
Aixelo, Inc1,Karlsruhe Institute of Technology (KIT)2
Materials AI can accelerate material discovery by up to 10x, but integrating AI into lab operations remains a challenge. Advanced tools like Bayesian optimizers and deep learning models are readily available through public git-repositories. However, true differentiation and value creation depend on successful adoption and how effectively these tools are applied in lab workflows.
In this presentation, I will share my firsthand experiences, focusing on the benefits, challenges, and obstacles of integrating AI with established experimental practices in MOF research. Specifically, we showcase the adoption of AI for optimizing Metal-Organic-Framework (MOF) thin-film growth. Precise control over MOF film fabrication is critical for achieving desired performance in applications such as sensors, photodetectors, catalysts etc.
We demonstrate the successful application of closed-loop AI strategies to optimize growth conditions for surface-anchored MOFs (SURMOFs). We reached high-quality films while meeting specific target properties such as controlled orientation, low surface roughness, or adherence to green synthesis principles outlined by Paul Anastas and John Warner. Our results highlight that AI-enhanced experimental planning consistently outperforms conventional techniques like one-factor-at-a-time (OFAT), Design of Experiment, or expert-guided trial and error.
Key technologies introduced include (i) automated data extraction combined with machine learning to predict synthesis conditions, (ii) adaptive experimental planning for autonomous optimization, and (iii) large language models for efficient data management and extraction. We provide these technologies to lab teams through Aixelo’s research management and materials informatics platform, EDISON-4.0™.