Publications Repository - Helmholtz-Zentrum Dresden-Rossendorf

Systems of self-propelled particles exhibit self-organized collective behavior that leads to the formation of complex spatio-temporal patterns that can be observed all over nature — from the active self-assembly of microtubules in cells through the action of kinesin motor proteins, to flocking birds. Because of their abundance, the question of how these rich macroscopic structures emerge from the microscopic interactions of their constituents remains of central interest. While there exist several hydrodynamic theories that help better understand the physical mechanisms, it is often difficult to determine which local microscopic interactions shape and regulate self-organized structures in active particle systems. Using a combination of unsupervised clustering algorithms and sparsity-promoting inference, we learn from data dominant force balance laws that locally drive the emergence of macroscopic patterns in active particle systems. We consider a classic hydrodynamic model of self-propelled particle systems that hosts solutions composed of spatiotemporal patterns like asters and propagating stripes. We show that 1) propagating stripes are formed by local alignment interactions and driven by gradients in polarization density and 2) steady-state asters are shaped by a mechanism of splay-induced negative compressibility arising from strong particle interactions. These data-driven discoveries are in excellent agreement with analytical predictions. We therefore believe that the presented data-driven strategy, in combination with physical modeling, can help our mechanistic understanding of active material systems as well as the design of biomimetic materials.