Jacobus Dijkman

🤹 Excited to share Erwin: A Tree-based Hierarchical Transformer for Large-scale Physical Systems

joint work with @wellingmax.bsky.social and @jwvdm.bsky.social

preprint: arxiv.org/abs/2502.17019
code: github.com/maxxxzdn/erwin

Our efficient method could accelerate research into molecular systems for critical applications like hydrogen storage and direct air capture—enabling scientists to explore far more scenarios than traditional simulations allow. 🌎

Want to learn more? Read the full paper here: doi.org/10.1103/Phys...

This approach lets us skip time-intensive simulations of complex systems, which could become prohibitively expensive for larger, real-world applications.

The key insight: our model learns by observing molecular interactions in simple uniform bulk systems. Once it grasps these patterns, it can predict behavior in complex environments like pores—despite never encountering non-uniform conditions during training.

The neural free energy functional estimates the particle density much faster.

We developed a novel ML approach that rapidly predicts molecular behavior—without running lengthy simulations. 🏎️

Sampling the particle density from molecular simulation is expensive.

Scientists traditionally rely on computer simulations to understand molecular-level behavior of liquids and gases. However, these simulations can be incredibly time-consuming. ⏳

🚨 Excited to share our work just published in Physical Review Letters with @wellingmax.bsky.social, @jwvdm.bsky.social, @berndensing.bsky.social, Marjolein Dijkstra and René van Roij: doi.org/10.1103/Phys....

Details below 👇

Our efficient method could accelerate research into molecular systems for critical applications like hydrogen storage and direct air capture—enabling scientists to explore far more scenarios than traditional simulations allow. 🌎

Want to learn more? Read the full paper here: doi.org/10.1103/Phys...

This approach lets us skip time-intensive simulations of complex systems, which could become prohibitively expensive for larger, real-world applications.

The key insight: our model learns by observing molecular interactions in simple uniform bulk systems. Once it grasps these patterns, it can predict behavior in complex environments like pores—despite never encountering non-uniform conditions during training.

We developed a novel ML approach that rapidly predicts molecular behavior—without running lengthy simulations. 🏎️

Scientists traditionally rely on computer simulations to understand molecular-level behavior of liquids and gases. However, these simulations can be incredibly time-consuming. ⏳

🙋‍♂️