Real projects. Real results.
These are not sandboxes or proof-of-concept demos. These are real R&D projects where structured, AI-driven experimentation made a measurable difference.
Case Study 1: Optimizing SURMOF Coatings
THE CHALLENGE
Creating highly crystalline, oriented metal-organic framework (MOF) coatings requires optimizing multiple interacting synthesis parameters — metal and linker concentrations, modulator amounts, cleaning times. Without systematic guidance, scientists face a combinatorial explosion of conditions to test, with no clear path to optimal SURMOF growth.
THE APPROACH
Using Edison 4.0’s adaptive optimization, the team defined multiple synthesis parameters and ran autonomous, multi-objective optimization via genetic algorithms. The system identified the best synthesis conditions while simultaneously analyzing which parameters actually mattered most.
THE RESULT
Beyond the immediate results, the team developed a general methodology applicable to various material synthesis processes — demonstrating that the approach works across different MOF systems and substrates.
KEY INSIGHT
The biggest value wasn’t just the optimized coating — it was understanding which parameters actually drive performance. That knowledge transfers to every future project.
PUBLICATIONS
Fully Automated Optimization of Robot-Based MOF Thin Film Growth via Machine Learning Approaches. doi.org/10.1002/admi.202201771
Enhancing the Quality of MOF Thin Films for Device Integration Through Machine Learning: A Case Study on HKUST-1 SURMOF Optimization. doi.org/10.1002/adfm.202404631
Case Study 2: Predicting Synthesis Conditions for Novel MOFs
THE CHALLENGE
Before synthesizing a new metal-organic framework, researchers typically spend weeks searching literature for relevant synthesis conditions. This process is tedious, inconsistent, and often overlooks key insights buried across thousands of papers — resulting in redundant lab work and missed opportunities.
THE APPROACH
Aixelo used natural language processing (NLP) and machine learning to automatically extract synthesis data from over 10,000 research papers, creating the SynMOF database — a structured knowledge base containing synthesis parameters and structural information for over 900 MOFs. Machine learning models then predicted optimal synthesis conditions for new MOFs based on this data.
THE RESULT
MOF synthesis records extracted
prediction accuracy than human experts
The system accelerated the identification of promising MOF candidates, significantly reduced trial-and-error, and enabled the discovery of MOFs with unique properties — opening doors to breakthroughs in material science that would have taken years of manual literature review.
KEY INSIGHT
The bottleneck in materials R&D isn’t always the experiments — sometimes it’s the knowledge that’s already out there but trapped in thousands of unstructured papers. AI can unlock it.
PUBLICATIONS
MOF Synthesis Prediction Enabled by Automatic Data Mining and Machine Learning. doi.org/10.1002/anie.202200242
Functional Material Systems Enabled by Automated Data Extraction and Machine Learning. doi.org/10.1002/adfm.202302630
Research Highlight: AI-Driven Perovskite Solar Cell Discovery
In a study published in Science, Aixelo co-founder Prof. Pascal Friederich and advisory board member Prof. Christoph Brabec demonstrated how advanced machine learning can transform materials discovery.
The research explored a virtual library of one million candidate molecules for perovskite solar cells. Using AI-driven, closed-loop design strategies, the team identified new materials capable of boosting solar cell efficiencies to over 26% — with only around 150 experiments, rather than thousands.
candidate molecules screened
experiments to reach target
solar cell efficiency achieved
Why this matters for your R&D
The methodologies behind this breakthrough — adaptive optimization, closed-loop experimentation, virtual screening — are the same principles built into Edison 4.0. Whether you work with solar cells, formulations, polymers, or catalysts, the approach is transferable: let AI navigate the search space so your team focuses on what matters.
Published in Science: DOI 10.1126/science.ads0901
What These Projects Have in Common
Real projects, not demos
Every case started with a real R&D challenge — not a sandbox or proof of concept. The outcome mattered.
Fewer experiments, deeper insights
AI-guided experimentation consistently reduced the number of experiments needed while providing understanding of why results happened — not just what happened.
Knowledge that transfers
The methodologies developed in each project apply beyond the original problem — to new materials, new systems, new teams.
Bring your own project.
Tell us about your R&D challenge. We’ll walk you through Edison 4.0 with your use case in mind.
