GENUFASD: Generative Understanding of Ultrafast Fluid Dynamics (*1)

Ultra-fast fluid dynamics (UFD) happen on nano- to micro-second time scales (figure 1). They are characterized by competing dynamics during relaxation from sudden localized disruption of equilibrium. The key to engineering application of UFD for nano-particle production, surface structuring, droplet breakup and bubble collapses is to understand how UFD can be tailored such that the competing dynamics during relaxation towards equilibrium result in the generation of targeted patterns and structures. Tailoring requires control and optimization, which again require understanding. We understand what we can reproduce and generate from models. The complexity of UFD and its lack of quantitative experimental accessibility constitute a serious barrier for conventional model- and simulation-driven research.

Exploration and technological deployment of UFD requires a new paradigm of model-based and data-driven research on generative prediction and understanding. In this project, we investigate ultrafast fluid dynamics after sudden exposure to thermal energy at interfaces and aim to exploit their ability to create patterns and structures for microfabrication and energy conversion. We want to understand how ultrafast fluid dynamics can be predicted by generative models and what confidence can be placed in their optimization.

The first step is to generate high-quality reference datasets that can be used both for process optimization and as training datasets for the machine learning methods under investigation.