MATTER OF OPINION
The Most Active Matter of All Nicholas T. Ouellette1,* The term ‘‘matter’’ encompasses everything from molecules to mountains. It also includes living, sentient beings. If matter composes all physical things, and materials science considers the behavior of such things, can materials science describe the most active matter of all? Active matter has emerged in recent years as an exciting new research area at the forefront of materials science. Generally speaking, matter is considered to be ‘‘active’’ when there are energy production and dissipation mechanisms operating inside what we consider to be its constituent elements,1 just as ‘‘soft’’ matter is built out of elements with internal mechanical and thermal properties. In part because many biomaterials can be considered active, there has been an explosion of research over the past few decades aimed at understanding and controlling active matter. Typical experimental active-matter systems include microtubules, bacterial colonies, and active colloids. These systems can be considered ‘‘matter’’ in the same way that granular materials such as sand piles can be thought of as matter, in that the microscopic constituent elements interact via well-understood forces such as mechanical contact interactions or fluid-mediated stresses. However, many researchers in active matter point to a wholly different phenomenon to motivate their work: the collective behavior of groups of social animals, such as flocks of birds, schools of fish, swarms of insects, or even human crowds (Figure 1). On its surface, animal behavior seems to be a field of research completely distinct from
materials science, but there are potentially important links between the two. Although animal groups are generally tied together via social interactions rather than real forces, they exhibit behavior that is certainly reminiscent of moving materials.2 A flock of starlings, for example, undulates and pulsates in ways that appear fluid-like. Herds of migrating wildebeest exhibit frontal instabilities that look like interfacial waves. And, just like in more common kinds of materials, the macroscopic morphology of the group and the coherent, cohesive behavior do not generally arise from specially informed leaders but rather are thought to be emergent properties. Despite these qualitative similarities and the trend of looking to animal groups as motivation for studying active matter on much smaller scales, many scientists still balk at taking the analogy seriously and thinking about using ideas and methods from materials science to characterize animal groups. Why? There are perhaps a few different answers to this question. Some are technical. For one thing, the lack of contact or field-mediated interactions makes it difficult to apply traditional physicsbased methods rigorously. Since interactions between animals arise from the integration and processing of information gleaned from a host of complex signal transduction mechanisms, they cannot really be cast as forces, and concepts such as Newton’s third law that are ingrained in our intuition need not apply. The behavior and dynamics can vary between individuals but in a way that is likely not well modeled by random, thermal-like fluctuations. And, more seriously, there are essentially no conservation laws that groups of animals have to obey, robbing us of our most powerful tool for analyzing physical systems. But the ultimate
reason that a true active-matter approach to collective animal behavior has not taken off may be that it feels strange to neglect cognitive processing or agency for macroscopic animals. We can see individual animals move independently, and we know that they often spend time in isolated contexts, so there is cognitive dissonance involved in treating animals as mere constituents of some larger object. Thus, understanding collective animal motion has typically been cast as a question of behavior rather than of materials science. Even though the central idea of models of collective behavior—that interactions at the local, individual level should scale up to produce net group motion and behavior—has all the hallmarks of a statistical physics question, the emphasis in the field has been on the behavioral rules followed by individuals rather than on the properties of the collective. This is not an a priori unreasonable approach and is directly applicable to exciting applications such as the design of swarm robotics systems. But focusing on individual rules in a field that is by its nature empirical (given that the lack of conservation laws renders typical first-principles theory impossible) presupposes that it is possible to infer these rules from measurable data. Extracting the behavior of individuals from observations of the group, however, is a challenging inverse problem—and a fundamentally nonlinear one at that, given the existence of interactions. Thus, it is inherently ill posed and requires some kind of modeling at the outset that will inevitably be convolved into the results of any analysis of the empirical data. And without incontrovertible first principles for behavior, it is hard to tell how much these modeling assumptions will color the answers one arrives at. One potential resolution to this conundrum is to take the notion of animal groups as active matter as more than
Matter 1, 297–299, August 7, 2019 ª2019 Elsevier Inc.
297
Figure 1. Blurring between Individual Animals to a Collective Group as a Whole By blurring our perspective on collective animal behavior away from individuals to the group as a whole, we can apply tools from materials science to deepen our understanding of these active systems.
just an analogy or a way to motivate research on physics-based models of active colloids but rather as a serious framework for advancing our understanding of animal behavior. Such an approach also creates opportunities for deepening our knowledge of more traditional active materials. The key in both cases is to do a better job of characterizing the properties of the macroscopic group state rather than focusing entirely on the microscopic interactions. With very few exceptions, however, animal groups have been characterized almost exclusively by their morphology and degree of directional order. A bird flock, for example, may be described qualitatively as being dense or loose, being elongated or spherical, or containing holes or necks. Flocks also exhibit a high degree of polarization, in that the velocity vectors of all of the individuals point in roughly the same direction, while a disordered group like an insect swarm does not. However, these qualitative descriptions give very little specific information. Bird flocks and fish schools, for example, can both be highly polarized, but that does not at all mean that they are the same. More subtly, simple morphological metrics cannot tell us how similar a flock of starlings and a flock of pigeons are,
298
Matter 1, 297–299, August 7, 2019
because such metrics contain no information about the actual properties of the group. In a materials context, the ways that animal groups are currently described would be akin to saying that solids and liquids are different, but all solids are more or less the same—that is, considering phase alone with no notion of material properties. Thinking like a materials scientist thus presents a potential roadmap forward for improving characterizations of animal groups. Materials possess a wide range of emergent properties that distinguish them even when they are visually similar. These properties can be related to the mechanics of the material, such as the elastic modulus or the viscosity, or to the way the material internally transports external stimuli, such as thermal or electrical conductivity. The way these properties emerge from the underlying properties of atoms or molecules is highly nontrivial, but importantly, they are definable and measurable without knowing anything about the underlying material structure. Material properties thus represent exactly the kind of nontrivial information about the group state that will help us move beyond current descriptions of animal groups.
As a practical matter, having a better description of the group state would allow us to flip the current paradigm for modeling and understanding collective behavior. Instead of trying to use observational data to constrain models by measuring the behavior of a group and attempting to infer the interaction rules, one could instead posit rules, run the model, and compare the model output to the data. Doing so with a simple order parameter like polarization is not informative enough for this approach to work, but comparing emergent transport coefficients and mechanical moduli may be. There are additional, deeper benefits to taking a materials approach to collective behavior. There remain longstanding biological questions as to the purpose of collective behavior, with suggestions ranging from predator avoidance to improved sensing and decision making to robustness in uncertain environments. These questions are very difficult to answer definitively, since many hypotheses are fundamentally untestable. But knowledge of the ‘‘material’’ properties of a group may be able to provide new insight. An animal group that behaves as a highly elastic and stiff material is likely to
transmit information rapidly and without loss throughout the group— the better to share information about external predators or other features of interest. In contrast, a group that is highly viscous and floppy is likely to be able to damp out external perturbations—the better to remain stable in the face of strong environmental perturbations. It is reasonable to hypothesize that evolution will have shaped animal behavior so that the group is effective at carrying out its purpose—and thus that this purpose may be encoded in the material properties, just as engineers design structural or functional materials with particular properties to suit their intended use. And along these lines, understanding the material-like properties of a collectively behaving group may also find application in the design of artificial multi-agent systems like swarms of robots, where one could think about tuning the low-level interactions to achieve particular kinds of large-scale response. Now, it is certainly true that knowledge of macroscopic properties does not uniquely constrain the underlying microscopic interactions, either in animal groups or real materials. In that sense, measuring the group properties alone does not tell us what kinds of individual-level behaviors animals execute. However, more than a century of study has built up intuition for how macroscopic properties and microscopic interactions are linked, which we may be able to use to guide the
development of appropriate models. These relationships will likely be somewhat different for active matter, but the ways in which they might change can be explored in more traditional active-matter systems that are easier to control and specify. Finally, there is an appealing similarity with the historical development of materials science in taking this approach, where macroscopic material properties were defined and characterized well before their underlying structure was understood. So what will it take to treat animal groups seriously as active matter? For theory and modeling, the primary challenges are grappling with the lack of true forces and conservation laws. For example, the possibility of nonmutual interactions (due to the lack of a third law) removes constraints on the governing equations that we tend to take for granted, with potentially unusual consequences. And without microscopic conservation laws, our typical starting point for theory and modeling is absent. Starting at the macroscopic (‘‘thermodynamic’’) level may be a viable approach to solving both of these problems, since assumptions such as approximate conservation laws may be more forgiving at the group level. And a side benefit of constructing models at the group level may be a better characterization of more traditional active-matter systems, just as classical thermodynamics is still tremendously useful even in the era of first-principles statistical mechanics. On the empirical
side, the primary challenge is to learn to conduct real experiments with tunable external stimuli rather than simply passive observations. As any materials scientist knows, detailed materials characterization requires applying controlled stimuli to materials and measuring the response. Although a few such measurements have been done on animal groups, primarily with insects, as they are small enough to keep in a laboratory,3,4 there is tremendous opportunity for progress in this area. As research in animal behavior broadens and becomes ever more interdisciplinary, it is time for materials scientists to become involved and contribute their unique skills. By viewing animal groups as the most active of active matter, materials scientists will bring a valuable new perspective to this vibrant field. 1. Marchetti, M.C., Joanny, J.F., Ramaswamy, S., Liverpool, T.B., Prost, J., Rao, M., and Aditi Simha, R. (2013). Hydrodynamics of soft active matter. Rev. Mod. Phys. 85, 1143–1189. 2. Couzin, I.D., and Krause, J. (2003). Selforganization and collective behaviour in vertebrates. Adv. Stud. Behav. 32, 1–75. 3. Tennenbaum, M., Liu, Z., Hu, D., and Fernandez-Nieves, A. (2016). Mechanics of fire ant aggregations. Nat. Mater. 15, 54–59. 4. Ni, R., Puckett, J.G., Dufresne, E.R., and Ouellette, N.T. (2015). Intrinsic fluctuations and driven response of insect swarms. Phys. Rev. Lett. 115, 118104. 1Department
of Civil and Environmental Engineering, Stanford University, Stanford, CA 94305, USA *Correspondence:
[email protected] https://doi.org/10.1016/j.matt.2019.07.012
Matter 1, 297–299, August 7, 2019
299