Data-driven discovery of complex-fluid physics

Differentiable models learn particle-scale interactions and continuum constitutive dynamics from complex-fluid flows — Learning physical models of complex-fluid behavior across scales.

We develop theory, algorithms, and open-source software to learn and represent complex-fluid physics directly from experimental data or finer-scale simulations. Our primary tools are differentiable simulations, which explicitly encode physical laws and efficiently compute gradients of physical observables with respect to model parameters, design variables, and control inputs. At the particle scale, we seek tractable descriptions of many-body interactions that are difficult to measure or compute directly. At the continuum scale, we develop constitutive relations that connect microstructure, memory, and activity to the mechanical response of complex fluids and soft materials. Beyond fitting models to data, gradients from these solvers enable inverse design and inform automated experimentation platforms, supporting applications ranging from in situ characterization of cell mechanics to tailoring fluid formulations and processing conditions.

Selected work

Learning constitutive models and rheology from partial flow measurements

Alp M. Sunol, James V. Roggeveen, Mohammed G. Alhashim, Henry S. Bae, and Michael P. Brenner

arXiv · 2025

Inferring interaction potentials from stochastic particle trajectories

Ella M. King, Megan C. Engel, Caroline Martin, Alp M. Sunol, Qian-Ze Zhu, Sam S. Schoenholz, Vinothan N. Manoharan, and Michael P. Brenner

Physical Review Research · 2025

MPC and System Identification with Differentiable Physics: Fluid System and Particle Beam Control

Alan Williams and Alp Sunol

arXiv · 2026