Speaker
Description
Abstract:
Particle-laden flows are everywhere: a fluid carries a dispersed population of small rigid bodies whose shapes can be as simple as spheres or as intricate as helices. This talk asks a basic question at the interface of geometry and dynamics. In the limit where inertia is negligible (Re → 0), can the shape of a particle, together with a time-dependent forcing that averages to zero, produce a net transport of the particles through the fluid, and can this transport distinguish between left- and right-handed (enantiomeric) shapes?
At zero Reynolds number, the governing equations are linear and kinematically reversible, so motion cannot rely on inertia or "memory" in the usual sense. Net effects must instead come from broken symmetries and from the geometry of the path traced by the system through its configuration space, for rigid bodies the space of orientations is the rotation group. This viewpoint echoes an old theme in physics. Einstein’s early work on suspensions already showed how macroscopic observables, such as effective viscosity, encode microscopic structure even in the simplest dilute limit [1]. The challenge is that for genuinely three-dimensional shapes, the microscopic response to the local flow is hard to compute at the scale needed to study collective transport in suspensions.
I will present a multiscale route around this bottleneck [2]: a neural surrogate closure that learns the geometry-dependent single-particle response to local linear Stokes flow (including stress and rotational couplings, and chirality-sensitive contributions for helices). The result is a shape-agnostic building block designed to be embedded into suspension-level solvers, enabling systematic studies of geometry-induced pumping and enantioselective transport once the coupling is in place. The goal of the talk is to frame the fundamental physics problem clearly, and to show how modern surrogate modelling makes it tractable without giving up geometric fidelity.
[1] Einstein, A. (1906). A new determination of molecular dimensions. Annln., Phys., 19, 289-306.
[2] Laudato, M. (2025). Neural-Network Closures for Complex-Shaped Particles in the Force-Coupling Method. arXiv preprint arXiv:2512.14532.