Small fire ant rafts are unstable

In this combined experimental and numerical study, we film the formation of fire ant rafts to determine how they cohere together. Surprisingly, we discover that ants prioritize separation and exploration: they flail legs and bounce off neighbors when they collide. Despite the active repulsion, fire ants cohere by the Cheerios effect, a capillary force that attracts small floating objects such as breakfast cereal. Experiments reveal that rafts consisting of fewer than ten ants disintegrate within minutes. Predictions by a Langevin model reproduce the stability transition and the critical raft size, which emerges from the balance between their mutual repulsion and the Cheerios effect. This work may inspire physically grounded models for the behavior of natural swarms.

Figure 1

Small rafts are unstable. (a) A raft of 20 ants on the surface of water. (Photo credits: Andre L. Magyar and Candler Hobbs). (b) Proportion of stray ants $P$ as a function of the raft size $N$. Orange triangles indicate experimental results ($n=72$) and gray circles simulation results ($n=150$). Orange and gray dashed lines represent the respective best-fit logistic functions for experiments and simulation. (c) A time series showing the destabilization of raft of five ants over 100 s. Scale bar is 1 cm.

Figure 2

Ants on the water surface behave as inertial random walkers. (a) Ten representative ant trajectories. (b) Probability distribution of the velocity magnitude $U$. Velocity components $u$ and $v$ are shown in the insets. The black solid line is the two-dimensional Maxwell-Boltzmann distribution with a standard deviation of 3.5 mm/s. (c) Mean-squared displacement of ant trajectories. The blue dashed line has a slope of 2 and the red a slope of 1. Both are offset from the best fit for clarity. The vertical gray dashed line marks the timescale we obtain through a separate series of experiments $\tau =m/k_f$ 4.3 s. The shaded area represents one standard deviation, which appears distorted under the logarithmic scale. For panels (b) and (c), $n$ = 96.

Figure 3

Destabilization of simulated ant rafts. (a) Trajectories of simulated ants as they start from a dense, ordered raft. Initial state of ants shown in gray circles and end state shown with black circles. Note that many of the ending states have gone beyond boundaries of the figure. As the number of ants on the raft increases from 7 ants (left) to 9 ants (middle), and to 13 ants (right), rafts become more stable. (b) Schematic of our Langevin model [Eq. (2)]. (c) Raft stability diagram. Color represents the proportion of stray ants $P$ after 5 min. Black dashed line marks the transition point where half of the ants go astray ($P=0.5$).

Figure 4

Assembly of ant rafts from different initial spacings. (a) Proportion of stray ants $P$ as a function of activity level $k_a$ and initial spacing $L$. Top left inset show the definition of initial ant spacing $L$. The remaining insets show representative trajectories of ants across a time frame of 5 min. Time series of $P$ for (b) small initial spacing $L=18.75\mathrm{mm}$ and (c) large initial spacing 30 mm. Color bar legend from part (a) holds for all subplots.