This project set out to speed up how we study tiny chemical interactions. It specifically demonstrates how ethanol molecules attach to aluminium surfaces by replacing long, power-hungry simulations with fast, accurate machine-learning predictions. Traditionally, scientists rely on detailed atomic-level simulations that can take days or weeks to run for every new set of conditions. In this Mansim project, those rich simulations were used once to create a training dataset, and an AI model was then taught to predict the same outcomes in a fraction of a second.
The result is a practical, data-driven tool that preserves the insight of heavy simulations while slashing turnaround time and computing cost. It can forecast how quickly molecules will stick, how many will bind, and how this changes with temperature, speed of impact, or concentration, without having to re-run the original physics every time. Because the predictions are both rapid and accurate, engineers can try many “what-if” scenarios, compare designs, and choose promising options far earlier in a project.
The impact reaches well beyond a single molecule–surface pair. By plugging this kind of AI surrogate into larger design workflows, teams can screen protective coatings, sensor surfaces, and catalytic treatments much more quickly. It enables rapid iteration in digital design studies, reduces the need for supercomputing resources, and lowers the environmental footprint of computation. The approach also supports automated optimisation: the model can guide searches towards the best operating conditions or materials, making the overall process smarter and more efficient.
In summary, this project demonstrated how Mansim’s AI-enabled simulations turn slow, expensive studies into fast, exploratory tools, accelerating discovery and helping researchers and industry move from ideas to tested designs with far less effort.