Abstract
Biological organisms and artificial active particles selforganize into swarms and patterns. Open questions concern the design of emergent phenomena by choosing appropriate forms of activity and particle interactions. A particularly simple and versatile system are 3Dprinted robots on a vibrating table that can perform selfpropelled and selfspinning motion. Here we study a mixture of minimalistic clockwise and counterclockwise rotating robots, called rotors. Our experiments show that rotors move collectively and exhibit superdiffusive interfacial motion and phase separate via spinodal decomposition. On long time scales, confinement favors symmetric demixing patterns. By mapping rotor motion on a Langevin equation with a constant driving torque and by comparison with computer simulations, we demonstrate that our macroscopic system is a form of active soft matter.
Introduction
Swarms of robots mimic collaboration of animals and can be programmed to arrange into shapes by information exchange among individuals^{1}. A purely physical approach to achieve collective behavior is selforganization, where patterns emerge outside of thermal equilibrium^{2}. Particularly simple examples are walking robots known as bristlebots, artificial bugs, or vibrobots. To excite motion, energy is input by a motor carried on top of the robot vibrating its body^{3} or, alternatively, by a vibrating baseplate^{4, 5}. Each robot consists of a rigid body and tilted elastic legs. The legs act as elastic springs and convert momentum of the baseplate into motion of the robot at each impact^{6,7,8,9}. In general, parallel legs apply translational forces (walker), while legs arranged in a circular manner apply torques (rotor, spinner or vibrot). As a result, these robots move, similar to mechanically vibrated polar disks^{10, 11}. Together with biological^{12} and synthetic microswimmers^{13}, they belong to the class of active soft matter. Collectives of such particles exhibit phase separation into dense clusters and a surrounding gas phase^{14,15,16,17,18,19} or selforganize into swarms and flocks^{3, 10, 11}.
In this work, we focus on 3Dprinted rotors of gearlike shape, which are expected to selforganize in ways not possible in the absence of energy input^{20,21,22,23,24,25}. Existing experimental realizations of rotors depend on the application of electromagnetic fields^{20,21,22}. This approach does not allow rotation control of individual rotors because the fields act equally on all particles. As an alternative, we use 3D printing to fabricate rotors with propulsion mechanisms that can be varied for each particle individually. We demonstrate that our setup is suitable for observing phase separation driven by active rotation^{26, 27}, a phenomenon reminiscent of spinodal decomposition in binary fluids with a high symmetry between both phases. Our experiment is a simple yet nontrivial example for chiral symmetry breaking in a classical system, far beyond the wellknown swarming and flocking behavior of active agents.
Results
Phase separation of opposite rotors in experiment
A mixture of 210 clockwise and 210 counterclockwise spinning rotors (Fig. 1a, b) with area filling fraction of 52% is prepared by placing the particles in a fully mixed checkerboard configuration (Fig. 1c). Once the vibration is turned on, the initial state quickly decays into a disordered configuration (Fig. 2a, Supplementary Fig. 1, and Supplementary Movie 1). After about 60 s, domains of equally rotating particles are visible and then grow until the system is almost fully segregated (Fig. 2b, c). During the segregation process, we observe superdiffusive motion along the interface, also known as edge currents (Fig. 2g)^{26, 28}.
Particle segregation is robust in a wide range of packing fractions ϕ. To show this, we vary the density of a smaller system (enclosing ring 300 mm diameter) in the range ϕ = 0.3–0.7, where the system is stable against gravitational drift and still below the transition to crystallization (Fig. 2h). We start at ϕ = 0.3 and successively add particles and wait for the system to relax (1–2 h per step). After reaching ϕ = 0.7, we reverse the process by removing particles to test the reversibility of phase separation. Domains of likewise spinning rotors form at packing fractions up to ϕ = 0.6. To demonstrate the efficiency of the demixing process, we calculate the relative difference of equal and opposite nearest neighbars (N_{op} − N_{eq}) / N_{tot} for each packing fraction, which in the absence of system boundaries is 0 for the fully mixed state and 1 for the fully demixed state. A plateau is found in the range \(\phi \ \lesssim \ 0.55\) before the system starts mixing again at higher packing fraction (Fig. 2i).
Phase separation is counterintuitive at first glance because particles interact purely repulsively and the spinning motion of equal rotors is blocked when in contact (Fig. 3a–d). In contrast, opposite rotors come apart more easily after collision (Fig. 3e–g). As a result of the longer contact time of equal rotors, the system exhibits spontaneous separation^{26}. We observe that the segregation efficiency decreases for high packing fraction (Fig. 2i), possibly because there is less freedom for likewise spinning particles to perform collective rotation. In the following experiments and simulations, we fix the packing fraction again to 52% by using the large enclosing ring with 420 particles. At this packing fraction pronounced segregation is generally observed.
We quantify edge currents by measuring the mean squared displacement (MSD) of particles moving close to the interface between domains of opposite rotors. MSD is recorded after a fixed time interval in dependence on the distance to the interface (Fig. 4a) as well as at fixed initial distance to the interface in dependence on time (Fig. 4b). We observe that particles close to the interface are more mobile and for distances greater than three particle diameters the displacement is independent of the distance. Furthermore, the time dependence of the MSD is quadratic at the interface and linear at larger distances.
Mapping of singleparticle motion to Langevin dynamics
Numerous results in the field of active matter have been discovered in simulations using effective equations of motion, typically with colloidal applications in mind in which active motion is driven by chemical gradients^{29}. In our system, particle rotation is the result of a purely mechanical ratcheting mechanism^{5, 8}. Interactions are extremely shortranged, collisions dissipative, and hydrodynamics negligible. We test whether our macroscopic experiment agrees with the approach in ref.^{26} and map it to a system of coupled Langevin equations with a driving torque of constant magnitude.
The equations of motion for the ith particle with velocity v_{ i } and angular momentum ω_{ i } read
where M and I are mass and moment of inertia of a single rotor, γ^{T} and γ^{R} are translational and rotational damping coefficients, respectively, \(\tau _i^{\mathrm{D}} = \pm {\kern 1pt} {\mathrm{const}}\) is a driving torque, and F_{ i } and τ_{ i } are forces and torques resulting from collisions with other particles and the enclosure. The random forces and torques are uncorrelated Gaussian processes with \(\langle \eta _i^{(m)}(t)\eta _j^{(n)}(t^\prime )\rangle = \delta _{ij}\delta (t  t^\prime )\delta _{mn}\), where m,n ∈ {x,y} and \(\langle \zeta _i(t)\zeta _j(t^\prime )\rangle = \delta _{ij}\delta (t  t^\prime )\). In the model, their amplitudes are functions of the diffusion coefficients D^{T} and D^{R}. However, in contrast to colloidal systems, the random forces do not describe thermal noise but an instability that arises from the vibration mechanics^{8}. This means diffusion and damping constants are not related through the Stokes–Einstein relation.
We measure the mass M and calculate I from particle shape and density. All other parameters are extracted from vibration experiments. The diffusion coefficients D^{T} and D^{R} are obtained from stochastic velocity fluctuations of a singleparticle trajectory (Fig. 5a, b) by fitting the slope of the MSD (Fig. 5c). The driving torque is given by \(\tau ^{\mathrm{D}} = \gamma ^{\mathrm{R}}\bar \omega\) in the steady state. It is only weakly dependent on the vibration amplitude A for A > 70 μm, where we observe an average angular velocity \(\bar \omega = \frac{1}{T}{\int}_0^T \omega (t){\rm{d}}t \sim 10{\text{ rad s}^{1}}\) (Fig. 5d). The translational damping coefficient γ^{T} is obtained from the response of an isolated rotor to an instantaneous kick. A rotor is perturbed and the resulting velocity decay measured (Fig. 5e). We find a good fit with an exponential decay, \(v \propto {\rm{exp}}(  t\gamma ^{\mathrm{T}}{\mathrm{/}}M)\), similar to a particle suspended in a viscous fluid, which confirms that the assumption of a velocitydependent frictional force in the Langevin equations (1) and (2) is justified. Finally, the rotational damping coefficient γ^{R} is determined from the acceleration of a stopped rotor due to the driving torque (Fig. 5f). Again, we find a good fit with an exponential function, \(\bar \omega  \omega \propto {\rm{exp}}(  t\gamma ^{\mathrm{R}}{\mathrm{/}}I)\).
The validity of the equations of motion and the parameters extracted from experiment (Table 1) are tested with Langevin dynamics simulations. A typical simulation evolution is shown in Fig. 2d–f (Supplementary Fig. 4 and Supplementary Movie 4). Visual inspection shows good agreement with the experiment.
Collective behavior on long time scales
We compare the domain size evolution of seven experimental runs and 128 Langevin dynamics simulations for a duration of 1000 s (Fig. 6a). Variations among experimental runs appear to be slightly larger than variations among the simulation runs, possibly caused by mass polydispersity and shape imperfections of our 3Dprinted rotors. Still, overall good agreement in the average coarsening behavior is observed. This demonstrates that the system of Langevin equations fitted to singleparticle behavior also reproduces collective dynamics well.
Dimensional analysis and numerical solutions of the Cahn–Hilliard model for spinodal decomposition predict that domain size growth follows s(t) = t^{1/3}g^{30, 31} with growth coefficient g on the order of (Dw)^{1/3} ^{32, 33}. Here, D is the particle selfdiffusion coefficient in the dense rotor system and w is the width of the transition zone between domains. The selfdiffusion coefficient is approximated from experiment and simulation as D = 2 × 10^{−5} m^{2} s^{−1} and the transition zone width is assumed to be about one particle diameter, w = 15 mm. Thus, g should be on the order of 10^{−2} m s^{−1/3}. Our data is consistent with g = 1.5(5) × 10^{−2} m s^{−1/3} (thick black lines in Fig. 6a). Nevertheless, the comparison to a powerlaw is only qualitative and is affected by the confinement condition. Simulations of large systems of rotors without confinement have already shown with high accuracy that the Langevin model used here leads to spinodal decomposition with exponent 1/3^{26}. Larger setups than ours are necessary to accurately determine the growth exponent in experiment.
Because random forces and the viscous term in the contact force are frequently neglected in the literature, but could be relevant in our experiment, we test simulations with D^{T} = D^{R} = t_{vis} = 0 (Fig. 6b). The deviation of these simulations from the ones with the full model is small, justifying the simplification. We use this simplification to reduce computational cost and study a large ensemble of realizations over longer times than attainable from experiment in a reasonable time.
At longer times, confinement affects the kinetics of phase separation strongly^{28, 34}. Clear deviations from a power law occur when we extend the observation time to 10000 s (Fig. 6c). The system remains trapped in metastable states for a certain time period, a phenomenon that is apparent as plateaus in the domain size growth plots, before coarsening eventually saturates in the completely phaseseparated state. From the analysis of snapshots, we discover that the plateaus correspond to the following highsymmetry states: the completely phaseseparated state (Fig. 6d), a symmetric state with three domains (Fig. 6e) and two types of metastable states with four domains (Fig. 6f, g). Figure 6h–i show corresponding highsymmetry states observed in experiment. Domain interfaces in the symmetric states are curved such that they intersect the boundary at angles close to 90° (dashed red lines in Fig. 6i). Domains with more straight interfaces are harder to break up because the interfacial motion exerts a lower pressure, which means the system remains in these metastable state for longer time. Surprisingly, this observation is reminiscent of phase segregation in immiscible fluids under strong confinement^{35, 36}, even though all essential preconditions applicable to molecular fluids are not fulfilled in our system. In particular, pressure is generally not a state variable in active systems^{37}.
While the simulation typically reaches a fully demixed symmetric steady state, the experimental system is more sensitive and the fully segregated symmetrical state is found only for small system sizes (Fig. 2h). For the large system of Fig. 6, the time to reach the final state is so long that gravitational drift eventually leads to asymmetry and we only observe nearly full demixing. Despite our attempts to correct the tilt down to 0.001 degree and careful readjustment before each run, tilt is gradually introduced due to the constant vibrations of the system restricting the maximal experimental time. Edge currents at the boundary appear to prevent the achievement of the symmetric fully segregated state (Fig. 7).
Discussion
Our findings suggest an analogy of our granular experiment and the behavior of microscopic soft matter systems. Past simulations predicted that phase separation of rotors is still possible in the presence of hydrodynamic interactions^{27}. We therefore expect that the evolution observed in our experiments could agree with that of smaller active rotors, like colloidal and molecular spinners in liquid crystals or biological motors^{38,39,40,41,42,43}, even though the confirmation of this phenomenon remains an ongoing experimental challenge. Additionally, our joined experimental and numeric approach strongly suggests that analytic descriptions of selforganization for microscopic systems are also applicable to active matter on large length scales. In future, effects characteristic for binary fluids, such as emulsification or critical Casimir forces, could be studied on the macroscopic scale with active rotors or other 3Dprinted vibrated particles.
Methods
Particle fabrication
The rotor geometry consists of two connected central cylinders, a cap (diameter 15 mm, height 2 mm) and a body (diameter 11 mm, height 6 mm). Four satellites (sidelength 3 mm) and seven tilted legs (length 8.5 mm, diameter 1 mm, angle 18degrees) are attached to the cap. Rotors are manufactured by 3D printing using fused deposition modeling from acrylonitrile butadiene styrene (ABS) for experiments on the large plate and stereolithography using acrylatebased photopolymer for experiments on the small plate. To simplify the detection of positions and orientations from video recordings, we mark the rotors with a cross at the top (either painted or with a printed label). To avoid alternative segregation mechanisms commonly observed in granular systems, such as sizedependent or weightdependent segregation, the particles must be manufactured with sufficiently high accuracy. We also need to consider other effects that might lead to additional inaccuracy, like positioning on the build plate (due to misalignment, particles close to the edges of the build plate can slightly differ in height) and the role of secondary manufacturing steps (painting the cross at the top). The 3D printing fabrication via fused deposition modeling employed in this work achieves a maximum accuracy of 0.1 mm. Overall, 300 particles of each species are fabricated. Particles with a mass that deviates more than 5% from the average mass (Fig. 8) were successively discarded until 210 particles of each species remained. The resulting distribution has a standard deviation of <2%. The mean weight of each species deviates by <0.5%. Particles created by stereolithography, which achieves a higher accuracy, do not require postprocessing.
Vibrating table
The active rotors are excited by vertical vibrations of circular acrylic baseplates (diameter 300 and 480 mm, height 15 and 30 mm) attached to an electromagnetic shaker. The tilt of the plate is adjusted with an accuracy of 10^{−3} degrees. Motion is restricted by an enclosing ring (diameter 470 mm). The shaker is attached to heavy concrete blocks to suppress resonances. Production experiments are performed with vibration frequency f = 80 Hz and vibration amplitude A = 84(2) μm, at which stable excitation is guaranteed. Experiments are recorded using a highspeed camera system at up to 500 frames per second with spatial resolution of 1024 × 1024 pixels. The rotors are tracked using standard image recognition methods with subpixel accuracy, which lies below the manufacturing tolerance of the 3d printer of 0.1 mm.
Damping constants
To measure the damping constants γ^{T} and γ^{R}, we assume that, on average, particle motion (translation and rotation) relaxes according to Newton’s equations of motion with a viscous damping term and external torque τ,
These equations are solved by an exponential with relaxation times t^{T} = M/γ^{T} and t^{R} = I/γ^{R}.
Moment of inertia
It is difficult to measure the moment of inertia I directly. However, since we know the mass and shape of the particle, I can be calculated from shape. The particle consists of two central cylinders (c1 and c2 with mass m_{c1,c2} and radius r_{c1,c2}), four cubes (satellites, s with mass m_{s}, edge length a_{s} and distance d_{s}) on the outside and seven cylinders (legs l with mass m_{l}, radius r_{l}, length l_{l}, distance from the center d_{l} and inclination angle θ with respect to the vertical). The total moment of inertia is the sum of individual moments considering offcentral rotation for the satellites and the legs,
Langevin dynamics simulations
Rotors are modeled by a large central disk (diameter 15 mm) and four satellite disks (diameter 3.4 mm). The diameter of the satellite disks is chosen such that the area covered by a particle equals that of the experiment. As typical for granular matter, we assume a viscoelastic normal contact force of the form \(F_{{\mathrm{vis}}} = F_{{\mathrm{el}}}(\xi ) + t_{{\mathrm{vis}}}\dot \xi \frac{{\partial F_{{\mathrm{el}}}}}{{\partial \xi }}\), where ξ is the compression or indentation depth of neighboring particles (which is only microscopic in our system) and t_{vis} is the viscous relaxation time describing the dissipative part of the force. Tangential forces between colliding discs are neglected. The elastic contact force of contacting cylinders is given by
with R_{eff} = 2(R_{1}R_{2}) / (R_{1} + R_{2})^{44} and ξ = R_{1} + R_{2} − d. Here, Y = 2.3 GPa is the Young modulus of the 3D printing material (ABS), ν = 0.35 is the Poisson ratio, h = 2 mm is the cylinder height and R_{1}, R_{2} are the cylinder radii, either 1.7 or 7.5 mm, depending on the colliding parts and d is the distance between particle centers. We either set a small value of t_{vis} = 10^{−5} s as typical for stiff materials or t_{vis} = 0 s for purely elastic collisions. Langevin dynamics simulations are performed using custom code that is included as part of the Supplementary Information.
Characteristic domain size
For two particle species A and B with particle densities
and the difference ρ_{AB}(r) = ρ_{A}(r) − ρ_{B}(r), we define the relative radial distribution function for the binary mixture as the autocorrelation function
If the system is homogeneous and isotropic, this expression reduces to the relative radial distribution function g_{AB}(r) that only depends on the distance r = r_{1}−r_{2}. It is positive if at distance r mainly particles from equal species contribute to the average and negative if mainly particles of the opposite species contribute to the average. The roots of g_{AB}(r) are found at distances, where particle species A and B appear with equal probability. A characteristic domain size is defined from the first nontrivial root^{26, 32, 33}. Measured examples at different times are shown in Fig. 9.
Code availability
The Langevin dynamics simulation source code that was used to generate all simulation data for this study is included as Supplementary Information. An OpenSCAD and STL file of the rotor design is included as Supplementary Information.
Data availability
The data that support the plots within this paper and other findings of this study are available from the corresponding author upon request.
Change history
11 April 2018
The original version of this Article contained an error in Fig. 5. In Fig. 5e, the scale on the yaxis originally incorrectly went from ‘0’ to ‘1.10^{−4}’. The correct scale goes from ‘0’ to ‘0.4’. In Fig 5f, the xaxis labels were incorrectly shifted to the right. This has been corrected in both the PDF and HTML versions of the Article.
References
 1.
Rubenstein, M., Cornejo, A. & Nagpal, R. Programmable selfassembly in a thousandrobot swarm. Science 345, 795–799 (2014).
 2.
Cross, M. C. & Hohenberg, P. C. Pattern formation outside of equilibrium. Rev. Mod. Phys. 65, 851–1112 (1993).
 3.
Giomi, L., HawleyWeld, N. & Mahadevan, L. Swarming, swirling and stasis in sequestered bristlebots. Proc. R. Soc. A 469, 20120637 (2013).
 4.
Tsai, J.C., Ye, F., Rodriguez, J., Gollub, J. P. & Lubensky, T. C. A chiral granular gas. Phys. Rev. Lett. 94, 214301 (2005).
 5.
Altshuler, E., Pastor, J. M., Garcimartín, A., Zuriguel, I. & Maza, D. Vibrot, a simple device for the conversion of vibration into rotation mediated by friction: preliminary evaluation. PLoS ONE 8, e67838 (2013).
 6.
Scholz, C. & Pöschel, T. Actively rotating granular particles manufactured by rapid prototyping. Rev. Cuba. Física 33, 37–38 (2016).
 7.
Torres, H., Freixax, V. M. & Perez, D. The Newtonian mechanics of a vibrot. Rev. Cuba. Física 33, 39–43 (2016).
 8.
Scholz, C., D’Silva, S. & Pöschel, T. Ratcheting and tumbling motion of vibrots. New J. Phys. 18, 123001 (2016).
 9.
Koumakis, N., Gnoli, A., Maggi, C., Puglisi, A. & Leonardo, R. D. Mechanism of selfpropulsion in 3Dprinted active granular particles. New J. Phys. 18, 113046 (2016).
 10.
Deseigne, J., Dauchot, O. & Chaté, H. Collective motion of vibrated polar disks. Phys. Rev. Lett. 105, 098001 (2010).
 11.
Deseigne, J., Léonard, S., Dauchot, O. & Chaté, H. Vibrated polar disks: spontaneous motion, binary collisions, and collective dynamics. Soft Matter 8, 5629–5639 (2012).
 12.
Berg, H. C. E. coli in Motion (Springer, New York, NY, 2004).
 13.
Golestanian, R., Liverpool, T. B. & Ajdari, A. Designing phoretic micro and nanoswimmers. New J. Phys. 9, 126 (2007).
 14.
Palacci, J., Sacanna, S., Steinberg, A. P., Pine, D. J. & Chaikin, P. M. Living crystals of lightactivated colloidal surfers. Science 339, 936–940 (2013).
 15.
Buttinoni, I. et al. Dynamical clustering and phase separation in suspensions of selfpropelled colloidal particles. Phys. Rev. Lett. 110, 238301 (2013).
 16.
Stenhammar, J., Tiribocchi, A., Allen, R. J., Marenduzzo, D. & Cates, M. E. Continuum theory of phase separation kinetics for active Brownian particles. Phys. Rev. Lett. 111, 145702 (2013).
 17.
Cates, M. E. & Tailleur, J. Motilityinduced phase separation. Annu. Rev. Condens. Matter Phys. 6, 219–244 (2015).
 18.
Bialké, J., Speck, T. & Löwen, H. Active colloidal suspensions: clustering and phase behavior. J. Non Cryst. Solids 407, 367–375 (2015).
 19.
Blaschke, J., Maurer, M., Menon, K., Zöttl, A. & Stark, H. Phase separation and coexistence of hydrodynamically interacting microswimmers. Soft Matter 12, 9821–9831 (2016).
 20.
Grzybowski, B. A., Stone, H. A. & Whitesides, G. M. Dynamic selfassembly of magnetized, millimetresized objects rotating at a liquidair interface. Nature 405, 1033–1036 (2000).
 21.
Yan, J., Bloom, M., Bae, S. C., Luijten, E. & Granick, S. Linking synchronization to selfassembly using magnetic Janus colloids. Nature 491, 578–581 (2012).
 22.
Driscoll, M. et al. Unstable fronts and motile structures formed by microrollers. Nat. Phys. 13, 375–379 (2016).
 23.
Petroff, A. P., Wu, X.L. & Libchaber, A. Fastmoving bacteria selforganize into active twodimensional crystals of rotating cells. Phys. Rev. Lett. 114, 158102 (2015).
 24.
Snezhko, A. Complex collective dynamics of active torquedriven colloids at interfaces. Curr. Opin. Colloid Interface Sci. 21, 65–75 (2016).
 25.
Scholz, C. & Pöschel, T. Velocity distribution of a homogeneously driven twodimensional granular gas. Phys. Rev. Lett. 118, 198003 (2017).
 26.
Nguyen, N. H. P., Klotsa, D., Engel, M. & Glotzer, S. C. Emergent collective phenomena in a mixture of hard shapes through active rotation. Phys. Rev. Lett. 112, 075701 (2014).
 27.
Yeo, K., Lushi, E. & Vlahovska, P. M. Collective dynamics in a binary mixture of hydrodynamically coupled microrotors. Phys. Rev. Lett. 114, 188301 (2015).
 28.
van Zuiden, B. C., Paulose, J., Irvine, W. T. M., Bartolo, D. & Vitelli, V. Spatiotemporal order and emergent edge currents in active spinner materials. Proc. Natl Acad. Sci. USA 113, 12919–12924 (2016).
 29.
Bechinger, C. et al. Active particles in complex and crowded environments. Rev. Mod. Phys. 88, 045006 (2016).
 30.
Lifshitz, I. & Slyozov, V. The kinetics of precipitation from supersaturated solid solutions. J. Phys. Chem. Solids 19, 35–50 (1961).
 31.
Zhu, J., Chen, L.Q., Shen, J. & Tikare, V. Coarsening kinetics from a variablemobility CahnHilliard equation: application of a semiimplicit Fourier spectral method. Phys. Rev. E 60, 3564–3572 (1999).
 32.
Toral, R., Chakrabarti, A. & Gunton, J. D. Numerical study of the CahnHilliard equation in three dimensions. Phys. Rev. Lett. 60, 2311–2314 (1988).
 33.
Toral, R., Chakrabarti, A. & Gunton, J. D. Large scale simulations of the twodimensional CahnHilliard model. Phys. A Stat. Mech. Appl. 213, 41–49 (1995).
 34.
Spellings, M. et al. Shape control and compartmentalization in active colloidal cells. Proc. Natl Acad. Sci. USA 112, E4642–E4650 (2015).
 35.
Avalos, E., Higuchi, T., Teramoto, T., Yabu, H. & Nishiura, Y. Frustrated phases under threedimensional confinement simulated by a set of coupled CahnHilliard equations. Soft Matter 12, 5905–5914 (2016).
 36.
Basu, S., Majumder, S., Sutradhar, S., Das, S. K. & Paul, R. Phase segregation in a binary fluid confined inside a nanopore. Europhys. Lett. 116, 56003 (2016).
 37.
Junot, G., Briand, G., LedesmaAlonso, R. & Dauchot, O. Active versus passive hard disks against a membrane: mechanical pressure and instability. Phys. Rev. Lett. 119, 028002 (2017).
 38.
Riedel, I. H. A selforganized vortex array of hydrodynamically entrained sperm cells. Science 309, 300–303 (2005).
 39.
Sumino, Y. et al. Largescale vortex lattice emerging from collectively moving microtubules. Nature 483, 448–452 (2012).
 40.
SchwarzLinek, J. et al. Phase separation and rotor selfassembly in active particle suspensions. Proc. Natl Acad. Sci. USA 109, 4052–4057 (2012).
 41.
Maggi, C., Saglimbeni, F., Dipalo, M., De Angelis, F. & Di Leonardo, R. Micromotors with asymmetric shape that efficiently convert light into work by thermocapillary effects. Nat. Commun. 6, 7855 (2015).
 42.
Vizsnyiczai, G. et al. Light controlled 3D micromotors powered by bacteria. Nat. Commun. 8, 15974 (2017).
 43.
Kokot, G. et al. Active turbulence in a gas of selfassembled spinners. Proc. Natl Acad. Sci. USA 114, 12870–12875 (2017).
 44.
Schwager, T. Coefficient of restitution for viscoelastic disks. Phys. Rev. E 75, 051305 (2007).
Acknowledgements
We acknowledge funding by Deutsche Forschungsgemeinschaft through the Cluster of Excellence Engineering of Advanced Materials (EXC 315/2) and through the grant SCHO 1700/11. Support from the Central Institute for Scientific Computing (ZISC), the Interdisciplinary Center for Functional Particle Systems (IZFPS) and compute resources provided by the Erlangen Regional Computing Center (RRZE) at FAU ErlangenNürnberg are gratefully acknowledged.
Author information
Affiliations
Contributions
C.S. and T.P. designed the experimental setup. C.S. carried out and analyzed the experiments. M.E. and C.S. wrote the simulation code. M.E. performed and analyzed the simulations. All authors discussed the results and wrote the manuscript.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing financial interests.
Additional information
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Electronic supplementary material
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Scholz, C., Engel, M. & Pöschel, T. Rotating robots move collectively and selforganize. Nat Commun 9, 931 (2018). https://doi.org/10.1038/s41467018031547
Received:
Accepted:
Published:
Further reading

Reconfigurable structure and tunable transport in synchronized active spinner materials
Science Advances (2020)

Rectification and separation of mixtures of active and passive particles driven by temperature difference
The Journal of Chemical Physics (2020)

Inertial effects of selfpropelled particles: From active Brownian to active Langevin motion
The Journal of Chemical Physics (2020)

Oscillating collective motion of active rotors in confinement
Proceedings of the National Academy of Sciences (2020)

Binary mixtures of active and passive particles on a sphere
Soft Matter (2020)
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.