CUDA simulations of active dumbbell suspensions

TL;DR

This paper presents GPU-accelerated simulations of hydrodynamic interactions in active dumbbell suspensions, demonstrating the effectiveness of stroke-averaged models in capturing collective dynamics at short distances.

Contribution

It introduces CUDA-based simulations to study active dumbbell suspensions and evaluates the accuracy of stroke-averaged models against detailed hydrodynamic interactions.

Findings

Stroke-averaged models are accurate down to a few times the dumbbell length.

GPU simulations enable analysis of hundreds of swimmers in real time.

Stroke-averaged far-field equations are useful for deriving hydrodynamic field equations.

Abstract

We describe and analyze CUDA simulations of hydrodynamic interactions in active dumbbell suspensions. GPU-based parallel computing enables us not only to study the time-resolved collective dynamics of up to a several hundred active dumbbell swimmers but also to test the accuracy of effective time-averaged models. Our numerical results suggest that the stroke-averaged model yields a relatively accurate description down to distances of only a few times the dumbbell's length. This is remarkable in view of the fact that the stroke-averaged model is based on a far-field expansion. Thus, our analysis confirms that stroke-averaged far-field equations of motion may provide a useful starting point for the derivation of hydrodynamic field equations.