Delft, 20th November 2024: In a bid to accelerate the transition to clean energy in the fight against climate change, VSParticle (VSP) – a Dutch nanotechnology engineering company – today announced the first results from a collaboration with Meta’s Fundamental AI Research (FAIR) team, and the University of Toronto (UofT).
The collaboration brings together VSP’s nanoporous layers printing technology, with UofT’s testing platform and Meta AI’s models, to rapidly produce, print and test the next-generation materials needed for clean energy technologies. Through its first Open Catalyst Experiments 2024 (OCx24), the collaboration has identified, synthesized and tested hundreds of electrocatalysts that are critical for clean energy solutions, and in doing so has built the first and largest open-source experimental catalyst database.
This is a milestone needed to help turn today’s AI-driven predictions into scalable, real-world products. The findings mark a major breakthrough in bridging the gap between computational models and experimental studies, bringing us closer to viable clean energy solutions at scale.
Electrocatalysts are critical to decarbonizing industries and achieving global climate targets due to the role they play in clean energy processes like carbon dioxide reduction reactions (CO2RR), hydrogen production and next-generation batteries. To accelerate the discovery of these catalysts, Meta’s FAIR team has been developing AI models to identify candidates for energy conversion processes in hours, rather than months. However, translating these predictions into scalable applications remains a complex challenge, typically taking up to 15 years. At the same time, training AI models to predict the best electrocatalyst materials requires large and diverse experimental datasets which simply don’t exist today.
To bridge this gap and accelerate the path of discovery to manufacturing, VSP, Meta and UoT came together to test datasets of hundreds of unique and diverse materials in the lab – creating the open-source database. Using a process called spark ablation, the VSP-P1 nanoprinter synthesized 525 materials that AI had predicted as the best candidates for CO2 Reduction Reactions (CO2RR) by vaporizing each one into nanoparticles.
These nanoparticles were then deposited as thin, nanoporous films and shared with the University of Toronto where its high-throughput pipeline tested how well each one performed under a range of industrial conditions. VSP’s unique nanoparticle approach gave researchers greater control over particle size and composition, with the high levels of automation and speed needed to create nanoporous materials at the scale required. Other technologies would need decades to synthesize such a high number of new nanoporous materials, which would have made the project impossible.
The findings were fed into an experimental database, from which researchers were able to validate the AI predictions against real-world results; identify hundreds of potential low-cost catalysts for key reactions; and which can now be used to train and further refine the AI and ML predictions. Next to building the largest experimental dataset, the project ran a record 20 million computer simulations – the largest computation of its kind to date – which can now be used to build even larger databases for scaling up the processes.
Aaike van Vugt, co-founder and CEO of VSParticle said: “By producing unique electrocatalysts at unprecedented speed, our partnership with Meta and the University of Toronto is not only helping to validate years of theory, but it’s shortening the discovery-to-application timeline; clearing a bottleneck that has held advanced materials back for decades. We have the only technology worldwide which is capable of delivering such a high number of unique nanoporous materials in a short period, to bring to life the vital work of Meta and UoT. Together, we’re proving that the materials needed to power this next generation of clean energy systems can be discovered and deployed at a pace that meets the urgency of the climate crisis.”
Larry Zitnick, Research Director at Meta AI said: “Through this collaboration, we’re breaking new ground in material discovery. It marks a significant leap in our ability to predict and validate materials that are critical for clean energy solutions. The results we’re seeing with electrocatalysts demonstrate the real-world potential of AI in addressing urgent climate challenges.”
To really crack the code for material discovery, AI models need to be trained on a much larger experimental dataset of between 10,000 to 100,000 unique tested materials. Since VSP’s technology is the only technology that could synthesize such a large number of thin-film nanoporous with high electrocatalytic performance, the company is working with many more organisations, including the Sorbonne University Abu Dhabi, the San Francisco-based Lawrence Livermore National Laboratory, the Materials Discovery Research Institute (MDRI) in the Chicago area, and the Dutch Institute for Fundamental Energy Research (DIFFER).
Alongside this project, VSP has been scaling up its own technology to be faster and more efficient in the future. The current VSP-P1 printer is powered by 300 sparks per second, but the team is also working on a new printer that would increase this output time to 20,000 sparks per second, which could supercharge this type of research even further. In particular, this would enable it to scale its core technology to support green hydrogen production through printing the necessary components for the porous transport electrode, something industrial customers are requesting. This means VSP will be able to reduce current production costs by 85%, through using fewer pieces of equipment, less energy and more automation, making it the most cost-competitive production technology for this critical aspect of green hydrogen production.
