betaFeedback

WorksPhysics and Astronomy

Emergence of beating in a magnetic flagellum consisting of active bots

Abstract

We investigate the emergence of flagellar beating in chains of magnetic self–propelled particles (MSPPs) built from centimeter–scale vibrating robots (Hexbugs) with embedded neodymium dipoles. When one end of the chain is anchored and self–propulsion is activated, longitudinal stress accumulates along the chain until it overcomes the magnetic bending stiffness, triggering a buckling instability that drives sustained flagellar beating. Using a combination of experiments and numerical simulations, we identify three distinct dynamical regimes straight chain, stable flagellar beating, and fission governed by the competition between active force, chain length, and magnetic bending stiffness. The onset of beating requires a seed misalignment set by the balance between magnetic torques and rotational noise, and we show that the transition corresponds to a supercritical Hopf bifurcation. A kinematic model reproduces the observed orientation dynamics with excellent agreement. The magnetic bending stiffness, which arises directly from dipole–dipole interactions, is fully tunable via dipole strength and chain length, offering independent experimental control over both activity and rigidity. Our results establish a macroscopic platform for studying force-induced buckling and self–oscillations in active filaments, with direct connections to flagellar motion in biological and synthetic microswimmers.

Among the various interactions governing collective behavior in physical systems, magnetic dipole–dipole interactions appear across a broad range of scientific and technological contexts, from condensed matter and colloidal physics to biophysics and materials science [1, 2, 3, 4, 5, 6]. In active matter systems, these interactions enable programmable self-assembly and long-range orientational alignment [7, 8, 9, 10], with magnetotactic bacteria serving as natural models for directional guidance and collective behavior [11, 12, 13], while magnetically driven colloids offer synthetic analogs that can be precisely controlled via external fields [1, 14, 15]. These systems have proven effective in applications such as soft robotics, microfluidics, and biomedical devices. However, they often exhibit dynamic, transient metastable states, making long-term control a persistent challenge.

Active magnetic particles have thus emerged as a compelling class of systems, both in biological and synthetic contexts, where their complex behavior arises from the interplay between self-propulsion, magnetic dipolar interactions, and intrinsic polarization [8, 16, 17, 18, 19]. A particularly intriguing case emerges when, in the presence of external fields or elastic bonding between particles, the head-to-tail symmetry of a linear configuration of active particles is broken, leading to flagellar-like beating driven by activity [20, 21, 22]. This phenomenon, known as beating by propulsion, mirrors the locomotion of spermatozoa and eukaryotic flagella [23, 24, 25, 26, 27, 28, 29], and has also been observed in synthetic systems such as active filaments and elastohydrodynamic simulations [9, 30, 31, 32, 33, 34, 35, 36], as well as in magnetic Janus particles and non-reciprocal active colloids [15, 17, 37]. Recently, this effect has been reproduced at the macroscopic scale using centimeter-scale Hexbugs: battery-powered robots propelled by internal vibrations [35, 38]. When connected via an elastic membrane, a Hexbug chain spontaneously exhibits oscillatory motion and synchronization, demonstrating the universality of this phenomenon across scales [39, 40]. Synchronization between flagella and cilia is a fundamental feature of biological locomotion and fluid transport [41], and has recently been reproduced in minimal synthetic designs [42].

In this work, we investigate the emergence of beating by propulsion in a chain of inertial, active disk-shaped particles with permanent magnetic dipole–dipole interactions [43, 44, 45]. Particles are arranged in a linear configuration with one end anchored to a rigid wall [33], remaining bound via dipole–dipole interactions. Upon activation, self-propulsion induces longitudinal stress along the chain that, depending on the interplay between activity, magnetic interaction, and orientational fluctuations, drives the system toward one of three regimes: a static straight chain, stable flagellar-like beating, or fission [16, 46]. We develop an analytical model for the tangential misalignment that triggers buckling [20, 34], and show that beating onset arises from the competition between active force F0F_0, dipolar attraction FmagF_{\mathrm{mag}}, and orientational fluctuations DRD_R. We further characterize beating amplitude and frequency as functions of magnetic strength and chain length through bending rigidity, bond angles, and three-body bending energy [9], providing a controllable platform for exploring active solids and self-oscillating filament dynamics [39, 47].

Results and discussion

Experimental details

We experimentally study the dynamics of a chain composed of NN magnetic self-propelled particles (MSPPs), constructed from Hexbug Nano robots [48] enclosed within 3D-printed disk-shaped armors of diameter σ=5.0±0.1\sigma= 5.0 \pm0.1 cm and mass M=0.018M = 0.018 kg. Each disk includes a cylindrical compartment along its diameter that houses a neodymium rod magnet of diameter d=0.6±0.001d = 0.6 \pm0.001 cm and length L=3.5±0.1L = 3.5 \pm0.1 cm, whose permanent magnetic dipole moment m=0.6±0.1m = 0.6 \pm0.1 A m2^2 was determined via gaussmeter measurements [44]. Each MSPP self-propels along its body axis, oriented at angle φi\varphi_i with respect to the horizontal, with a magnetic dipole moment mi=mp^i\mathbf{m}_i = m\hat{\mathbf{p}}_i parallel to the propulsion direction (see Figure 1).

Magnetic self-propelled particle (MSPP)

Figure 1. Magnetic self-propelled particle (MSPP). Left: components of an MSPP a neodymium cylindrical magnet (S–N poles indicated), a 3D-printed disk-shaped armor of diameter σ=5\sigma= 5 cm, and a Hexbug Nano robot providing self-propulsion via internal vibratory motion. Right: schematic of the assembled MSPP, modeled as a disk of diameter σ\sigma with orientation angle φ\varphi measured from the chain axis (green dashed line, corresponding to the x^\hat{x} axis). The self-propulsion direction p^=(cos⁡φ,sin⁡φ,0)\hat{\mathbf{p}} = (\cos\varphi,\sin\varphi,0) is parallel to the permanent magnetic dipole moment m=mp^\mathbf{m} = m\hat{\mathbf{p}}, with magnitude m=0.6±0.1 A m2m = 0.6 \pm0.1\ \mathrm{A\,m^2} determined via gaussmeter measurements [44].

To ensure uniform propulsion, all experiments were carried out using new batteries, which allowed each MSPP to self-propel at a consistent average speed u0hexbugs=2u_0^{\mathrm{hexbugs}} = 2 cm/s [44], leading to an active force F0=γTu0hexbugs=2×10−4F_0 = \gamma_T u_0^{\mathrm{hexbugs}} = 2 \times10^{-4} N. The translational friction coefficient γT=10−2\gamma_T = 10^{-2} kg/s, rotational friction coefficient γR=10−3\gamma_R = 10^{-3} kg m2^2/s, and rotational diffusion coefficient DR=10−5D_R = 10^{-5} rad2^2/s were not measured directly but were selected to be consistent with the known dynamical behavior of Hexbug-based MSPPs in previous experiments [39, 43, 44], and validated by matching simulation trajectories to experimental observations (see Supplemental Material [49]). Due to the limited self-propulsion speed of the Hexbugs, we focus on chain lengths of N=4N = 4 and N=5N = 5 MSPPs. Throughout the main text we present results for the N=4N = 4 case; details for the N=5N = 5 configuration are provided in the Supplemental Material [49].

We break the symmetry of the magnetic chain by anchoring the head particle to a 3D-printed wall that fixes its position (r1=(0,0,0)\mathbf{r}_1 = (0,0,0)) and orientation (p^=−x^\hat{\mathbf{p}} = -\hat{\mathbf{x}}), forming a magnetically bound chain via dipole–dipole interactions (see Figure 2). Due to the Hexbug’s internal propulsion mechanism, the head exhibits minimal residual vibrations. All particles are activated simultaneously at the start of each experiment.

Flagellar beating of a magnetic flagellum consisting of active bots

Figure 2. Flagellar beating of a N=4N = 4 MSPP chain. (a) Experimental photograph showing the trajectories of individual MSPPs during flagellar beating. Colors indicate particle index: green for i=2i = 2, orange for i=3i = 3, and blue for i=4i = 4 (tail); the head particle (i=1i = 1) is anchored and not shown. (b) Schematic of the N=4N = 4 anchored chain defining the oscillation angle θ2\theta_2 between the position vector of the second particle and the chain axis x^\hat{\mathbf{x}} (green dashed line). Red circles indicate the excluded-volume diameter σ\sigma of each MSPP. (c) Arc displacement Δsi=(i−1)σθi(t)\Delta s_i = (i - 1)\sigma\theta_i(t) as a function of time from experiments. (d) Transverse displacement Δyi(t)\Delta y_i(t) as a function of time measured during numerical simulations. In both (c) and (d), amplitude increases with distance from the anchored head, consistent with flagellar beating [27, 33]. (e) Oscillation amplitude ΔSi=(i−1)σAi\Delta S_i = (i - 1)\sigma A_i across multiple realizations. Filled circles represent simulations and triangles represent experiments. Amplitudes are consistent across realizations, indicating they are set by intrinsic system parameters. (f) Normalized oscillation period Ti/T2T_i/T_2 across multiple realizations. Periods vary between realizations, reflecting sensitivity to initial orientations and rotational noise at the moment of buckling.

The motion of the chain is recorded using a standard cellphone camera. Post-processing was conducted using a custom Python code to extract the trajectories and orientations of individual particles (see Supplemental Material [49]).

The onset of buckling

When particles are self-propelled along p^=−x^\hat{\mathbf{p}} = -\hat{\mathbf{x}}, head-to-tail magnetic dipoles generate attractive forces and torques that bind the particles together, while active forces introduce a longitudinal pressure Σlong=F0N/σ2\Sigma_{\mathrm{long}} = F_0N/\sigma^2 that increases with chain length ℓchain=Nσ\ell_{\mathrm{chain}} = N\sigma, where σ\sigma is the particle diameter and F0F_0 is the active force (see Figure 1).

This stress accumulation generates a symmetry breaking between the head and the tail of the chain. Whether buckling occurs depends on the competition between this accumulated stress and the magnetic bending stiffness kmag=μ0m2/2πσ2k_{\mathrm{mag}} = \mu_0m^2/2\pi\sigma^2. Following Euler buckling theory [17, 21, 33], we define a critical force Fc=kmag/ℓchain2F_c = k_{\mathrm{mag}}/\ell_{\mathrm{chain}}^2 and a total active force Fact=F0NF_{\mathrm{act}} = F_0N. Their ratio defines the dimensionless buckling parameter

Πbuck=FactFc=F0N3σ2kmag,(1)\Pi_{\mathrm{buck}} = \frac{F_{\mathrm{act}}}{F_c} = \frac{F_0N^3\sigma^2}{k_{\mathrm{mag}}}, \tag*{(1)}

which sets the balance between the active force F0F_0, the magnetic interaction strength kmagk_{\mathrm{mag}}, and the chain length NN. When Πbuck≪1\Pi_{\mathrm{buck}} \ll1 the chain remains straight; when Πbuck∼1\Pi_{\mathrm{buck}} \sim1 flagellar-like beating develops; and when Πbuck≫1\Pi_{\mathrm{buck}} \gg1 fission is possible [16, 46] (see phase diagram in Supplemental Material [49]).

Substituting the experimental parameters (σ=0.05\sigma= 0.05 m, m=0.6m = 0.6 A m2^2, F0=2×10−4F_0 = 2 \times10^{-4} N) gives kmag=μ0m2/2πσ2=2.88×10−5k_{\mathrm{mag}} = \mu_0m^2/2\pi\sigma^2 = 2.88 \times10^{-5} N m2^2 and

Πbuck=F0σ2kmagN3=1.736×10−2N3.(2)\Pi_{\mathrm{buck}} = \frac{F_0\sigma^2}{k_{\mathrm{mag}}}N^3 = 1.736 \times10^{-2}N^3. \tag*{(2)}

The resulting values for each chain length are summarized in Table I:

NNΠbuck\Pi_{\mathrm{buck}}Observed behavior
30.47chain remains straight
41.11stable flagellar beating
52.17stable flagellar beating
63.75chain breaks (fission)

Table I. Buckling parameter Πbuck\Pi_{\mathrm{buck}} for each chain length NN, computed from Eq. (1) using the experimental parameters. The criterion Πbuck∼1\Pi_{\mathrm{buck}} \sim1 accurately separates the three observed regimes.

The table confirms that the criterion Πbuck∼1\Pi_{\mathrm{buck}} \sim1 cleanly separates the three regimes. When the magnetic interaction is kept constant, a chain of N=3N = 3 cannot reach the buckling threshold and remains straight (see Supplemental Movie SM1). For N=6N = 6, the active force overcomes magnetic cohesion and the chain breaks shortly after buckling (see Supplemental Movie SM4). Stable flagellar beating is therefore observed only for N=4N = 4 and N=5N = 5, which bracket the critical value Πbuck≈1\Pi_{\mathrm{buck}} \approx1 (see Supplemental Movie SM2, SM3).

The dimensionless criterion Πbuck∼1\Pi_{\mathrm{buck}} \sim1 is the discrete-chain analog of buckling parameters previously introduced for continuous active filaments [17, 20, 33], where the magnetic dipole-dipole interaction provides the effective bending rigidity kmagk_{\mathrm{mag}} directly from particle properties, enabling independent experimental control of both activity and stiffness.

Flagellar beating

We performed numerical simulations modeling MSPPs as disk-shaped particles with point dipoles located at their centers. The position of the ii-th particle is ri(t)=(xi(t),yi(t),0)\mathbf{r}_i(t)=(x_i(t),y_i(t),0) and its orientation is p^i(t)=(cos⁡φi(t),sin⁡φi(t),0)\hat{\mathbf{p}}_i(t)=(\cos\varphi_i(t),\sin\varphi_i(t),0). The magnetic dipole moment is mi=mp^i(t)\mathbf{m}_i=m\hat{\mathbf{p}}_i(t), where mm is its magnitude. Once buckling is triggered, the chain develops stable flagellar beating. The equations of motion are

Mr¨i=F0p^i(t)−γTr˙i−∇ri∑j≠i(UijWCA+UijD),(3)M\ddot{\mathbf{r}}_i=F_0\hat{\mathbf{p}}_i(t)-\gamma_T\dot{\mathbf{r}}_i-\nabla_{\mathbf{r}_i}\sum_{j\ne i}\left(U_{ij}^{\mathrm{WCA}}+U_{ij}^{\mathrm{D}}\right), \tag*{(3)}
p^˙i=1γR(ξi,R(t)−Ti)×p^i,(4)\dot{\hat{\mathbf{p}}}_i=\frac{1}{\gamma_R}\left(\boldsymbol{\xi}_{i,R}(t)-\mathbf{T}_i\right)\times\hat{\mathbf{p}}_i, \tag*{(4)}

where γT\gamma_T and γR\gamma_R are the translational and rotational friction coefficients, respectively, and F0F_0 is the active force. The term ξi,R(t)\boldsymbol{\xi}_{i,R}(t) is a rotational Gaussian white noise of zero mean satisfying ⟨ξj,R(t1)⟩=0\langle\boldsymbol{\xi}_{j,R}(t_1)\rangle=0 and ⟨ξj,R(t1)ξj,R(t2)⟩=2γR2DRδ(t1−t2)\langle\boldsymbol{\xi}_{j,R}(t_1)\boldsymbol{\xi}_{j,R}(t_2)\rangle=2\gamma_R^2D_R\delta(t_1-t_2), where DRD_R is the rotational diffusion coefficient. The magnetic torque acting on particle ii is Ti=p^i×∇p^iUD\mathbf{T}_i=\hat{\mathbf{p}}_i\times\nabla_{\hat{\mathbf{p}}_i}U^{\mathrm{D}}.

We note that a self-alignment torque β(p^i×v^i)\beta(\hat{\mathbf{p}}_i\times\hat{\mathbf{v}}_i), which couples orientation to velocity, has been included in previous models of active magnetic particles [16, 44, 46] and shown to influence rotational dynamics in the absence of magnetic interactions. In the present system, however, magnetic torques dominate the orientational dynamics, and we verified that self-alignment plays no significant role in the flagellar beating observed here. Furthermore, since we study anchored chain configurations with fixed initial conditions, and since our analytical model (Eq. (8)) reproduces the observed dynamics without this term, we omit it throughout.

Unlike colloidal systems typically modeled in the overdamped limit, Hexbug-based MSPPs operate at the macroscopic scale where inertial effects are non-negligible. The mass MM therefore appears explicitly in Eq. (3), placing this system in the regime of inertial magnetic active matter [43, 44, 50], where translational dynamics retains inertia while orientational dynamics remains overdamped.

Each MSPP is subject to two interactions: an excluded-volume potential UWCAU^{\mathrm{WCA}} modeled by the Weeks–Chandler–Anderson (WCA) potential, and a magnetic dipole–dipole interaction UDU^{\mathrm{D}}, given by

UijWCA={4ε[(σrij)12−(σrij)6]rij≤rm,0otherwise,(5)U_{ij}^{\mathrm{WCA}}= \begin{cases} 4\varepsilon\left[\left(\frac{\sigma}{r_{ij}}\right)^{12}-\left(\frac{\sigma}{r_{ij}}\right)^6\right] & r_{ij}\le r_m,\\ 0 & \text{otherwise}, \end{cases} \tag*{(5)}
UijD=μ0m24πrij3[p^i⋅p^j−3(p^i⋅rij)(p^j⋅rij)rij2],(6)U_{ij}^{\mathrm{D}}=\frac{\mu_0m^2}{4\pi r_{ij}^3}\left[\hat{\mathbf{p}}_i\cdot\hat{\mathbf{p}}_j-\frac{3(\hat{\mathbf{p}}_i\cdot\mathbf{r}_{ij})(\hat{\mathbf{p}}_j\cdot\mathbf{r}_{ij})}{r_{ij}^2}\right], \tag*{(6)}

where rij=rj−ri\mathbf{r}_{ij}=\mathbf{r}_j-\mathbf{r}_i, rij=∣rij∣r_{ij}=|\mathbf{r}_{ij}|, rm=21/6σr_m=2^{1/6}\sigma is the WCA cutoff, and ε=100 kBT\varepsilon=100\,k_{\mathrm{B}}T is the excluded-volume energy scale.

For N=4N=4 MSPPs (and N=5N=5 in the Supplemental Material [49]), with parameters m=0.6 A m2m=0.6\ \mathrm{A\,m^2} and F0=2×10−4 NF_0=2\times10^{-4}\ \mathrm{N}, we observe in both experiments and simulations that each particle in the chain undergoes flagellar beating (see Supplementary Movies S2, S3, S5). The beating is characterized by the angle θi\theta_i, defined between the position vector of the ii-th particle’s center and the horizontal axis (see Figure 2(b)). Each particle follows a closed orbit whose amplitude increases with distance from the clamped head, as observed in sperm flagella and elastoactive simulations [26, 27, 39]. This motion can be quantified by the arc length Δsi=(i−1)σi\Delta s_i=(i-1)\sigma_i (valid when θi\theta_i is small) or equivalently by the transverse displacement Δyi\Delta y_i, where i=1,2,…,Ni=1,2,\ldots,N and i=1i=1 corresponds to the head particle.

We plot both Δy\Delta y and Δs\Delta s as a function of time in Figure 2(c),(Figure 2d) for experiments and simulations, respectively. Stable periodic oscillations are observed, with amplitude increasing with distance from the clamped head. This behavior has been reported in previous experimental realizations in colloidal and inertial systems [17, 39] and is consistently reproduced in our simulations. To quantify the oscillations, we fit a sinusoidal function Δyi(t)≈Δsi(t)=Aisin⁡(2πt/Ti)\Delta y_i(t)\approx\Delta s_i(t)=A_i\sin(2\pi t/T_i) to each particle’s trajectory, extracting the oscillation amplitude AiA_i and period TiT_i across multiple experimental and numerical realizations.

To analyze and compare these results, we plot the arc swept Δsi=(i−1)σi\Delta s_i=(i-1)\sigma_i and the normalized period ratio Ti/T2T_i/T_2 for particles i=3i=3 (orange) and i=4i=4 (blue) in Figure 2(e),(Figure 2f). Experimental data are shown as triangles and simulation results as circles, with all periods normalized by T2T_2, the period of the second particle in the chain.

Interestingly, the oscillation amplitudes remain consistent across realizations, indicating that they are set by intrinsic physical parameters such as the dipolar interaction strength mm and the active force F0F_0 [16, 33]. In contrast, the oscillation periods show large variations. On average, we measure ⟨T3/T2⟩=1.6\langle T_3/T_2\rangle=1.6 and ⟨T4/T2⟩=1.9\langle T_4/T_2\rangle=1.9, with the last particle generally exhibiting a longer period than its neighbors, a trend observed consistently in both simulations and experiments. This variability suggests that the oscillation frequency is not deterministic but is sensitive to rotational noise and the relative initial orientation of each particle at the moment of buckling.

We further investigate the role of magnetic interactions in shaping the flagellar beating.

A simple theory for tangential misalignment

During the experiments, we tracked the orientation angle φi\varphi_i of each particle. Initially, all particles are aligned along the −x^-\hat{x} direction, forming a straight magnetic chain. Upon activation, a small misalignment between neighboring particles, induced by the accumulation of longitudinal stress and rotational diffusion [33], triggers buckling and the onset of flagellar beating. In the absence of rotational diffusion—as modeled in simulations—the system fails to break the initial symmetry, and even strongly active particles remain confined to a straight chain configuration [16].

The orientation time evolution is shown in Figure 3(a),(Figure 3b), where the amplitude of angular oscillation increases with the particle’s distance from the anchored head. The angular displacements are limited by dipole–dipole magnetic torques, which tend to restore tangential alignment between neighboring dipoles. In Eq. (4), this torque counteracts rotational noise, acting as a restoring torque. The balance between the dipolar restoring torque and rotational noise defines a characteristic dimensionless misalignment parameter

Orientation dynamics and onset of flagellar beating

Figure 3. Orientation dynamics and onset of flagellar beating. (a) Time evolution of the orientation angle φi\varphi_i for each particle during experiments. (b) Same quantity from numerical simulations (triangles). (c) Theoretical orientation time evolution from Eq. (8) for m=1m=1, σ=1\sigma=1, δy=0.45\delta_y=0.45, ω=0.72\omega=0.72, and ϕ=−π/4\phi=-\pi/4 for each particle, with amplitude growing with distance from the anchored head. Colors in (a)–(c) indicate particle index: green for i=2i=2, orange for i=3i=3, and blue for i=4i=4 (tail). (d) Rotational noise parameter ΠR=4πγRDRσ/μ0m2\Pi_R=4\pi\gamma_R D_R\sigma/\mu_0m^2 as a function of the magnetic dipole strength mm for fixed σ\sigma, shown for different rotational diffusion coefficients DRD_R. Larger DRD_R (blue) leads to ΠR≫1\Pi_R\gg1, where noise dominates and stable flagellar beating is disrupted; smaller DRD_R (red, as in our setup with DR=10−5D_R=10^{-5}) gives ΠR≪1\Pi_R\ll1, allowing magnetic torques to maintain chain cohesion. (e) ΠR\Pi_R as a function of particle diameter σ\sigma for fixed mm. Increasing σ\sigma weakens the dipole-dipole interaction, raising ΠR\Pi_R and reducing chain stability, consistent with the observation that larger particles require stronger dipole moments to sustain flagellar beating. (f) Stability diagram for the kinematic model: δy/σ\delta_y/\sigma as a function of angular frequency ω\omega for different phase angles ϕ\phi, showing the range of (δy,ω,ϕ)(\delta_y,\omega,\phi) values that sustain flagellar beating. The shaded region is bounded below by δy,min⁡=0.1σ\delta_{y,\min}=0.1\sigma and above by δy,max⁡=0.5σ\delta_{y,\max}=0.5\sigma. Dashed lines indicate values observed in experiments (gray) and simulations (red). Inset: schematic of the transverse displacement δy\delta_y imposed on the tail particles to initiate beating.

ΠR=4πγRDRσ3μ0m2,(7)\Pi_R = \frac{4\pi\gamma_R D_R \sigma^3}{\mu_0 m^2}, \tag*{(7)}

where ΠR=4πγRDRσ3/μ0m2\Pi_R = 4\pi\gamma_R D_R \sigma^3/\mu_0m^2 measures the typical noise-induced angular misalignment between neighboring MSPPs. When ΠR≪1\Pi_R \ll1, magnetic torques dominate and the chain remains well-aligned; when ΠR≳1\Pi_R \gtrsim1, rotational noise overcomes magnetic cohesion and stable flagellar beating is disrupted.

In Figure 3(d), we plot ΠR\Pi_R as a function of mm for different DRD_R values; red curves correspond to low DR∼10−5D_R \sim10^{-5}, while

blue curves correspond to high DR∼10−2D_R \sim10^{-2}. For large DRD_R, ΠR>1\Pi_R > 1 for most values of mm, meaning rotational noise dominates and stable flagellar beating cannot be maintained. For small DRD_R, as in our setup (DR=10−5D_R = 10^{-5}), the noise-induced misalignment is small (ΠR≈0.006\Pi_R \approx0.006 at m=0.6m = 0.6 A m2^2), allowing magnetic torques to sustain a stable chain configuration from which flagellar beating can develop once Πbuck∼1\Pi_{\mathrm{buck}} \sim1.

In Figure 3(e), we show ΠR\Pi_R as a function of σ\sigma for fixed mm. As σ\sigma increases, the dipole–dipole interaction weakens (UD∝σ−3U^D \propto\sigma^{-3}) while the noise-induced misalignment grows (ΠR∝σ3\Pi_R \propto\sigma^3), eventually exceeding the stability threshold. This confirms that larger particles require stronger dipole moments to sustain flagellar beating, consistent with our experimental observations.

To better understand the onset of buckling, we developed a theoretical model describing the time evolution of the orientation angle φi\varphi_i due to magnetic torques,

φ˙i=μ0m24πγR∑j≠ip^i×(3r^ij(p^j⋅r^ij)rij5−p^jrij3).(8)\dot{\varphi}_i = \frac{\mu_0m^2}{4\pi\gamma_R}\sum_{j\ne i}\hat{\mathbf{p}}_i \times\left(\frac{3\hat{\mathbf{r}}_{ij}(\hat{\mathbf{p}}_j\cdot\hat{\mathbf{r}}_{ij})}{r_{ij}^5}-\frac{\hat{\mathbf{p}}_j}{r_{ij}^3}\right). \tag*{(8)}

We assume particle orientations p^1=(1,0,0)\hat{\mathbf{p}}_1 = (1,0,0) and, for

i=2,3,4i=2,3,4, p^i=(cos⁡φi(t),sin⁡φi(t),0)\hat{\mathbf{p}}_i=(\cos\varphi_i(t),\sin\varphi_i(t),0), with positions r1=(0,0,0)\mathbf{r}_1=(0,0,0), r2=(σ,0,0)\mathbf{r}_2=(\sigma,0,0), r3=(2σ,δycos⁡(ωt),0)\mathbf{r}_3=(2\sigma,\delta_y\cos(\omega t),0), and r4=(3σ,δycos⁡(ωt+ϕ),0)\mathbf{r}_4=(3\sigma,\delta_y\cos(\omega t+\phi),0), where δy\delta_y is the amplitude of the transverse oscillation imposed on particles 3 and 4, ω\omega is an angular frequency, and ϕ\phi a phase. Note that δy\delta_y is the steady-state beating amplitude used as input to the kinematic model. When δy=0\delta_y=0 the chain remains stable; for δy≠0\delta_y\ne0, flagellar beating emerges depending on ϕ\phi and ω\omega. In Figure 3(c), filled curves show solutions to Eq. (8) for m=1m=1, σ=1\sigma=1, δy=0.45\delta_y=0.45, ω=0.72\omega=0.72, and ϕ=−π/4\phi=-\pi/4, showing excellent agreement with simulations.

We find that a range of (δy,ϕ)(\delta_y,\phi) values leads to sustained flagellar beating for fixed σ\sigma and mm, as shown in Figure 3(f). The colored lines show δy/σ\delta_y/\sigma as a function of ω\omega for different phase angles ϕ\phi, bounded below by the minimum amplitude required to sustain beating, δy,min⁡=0.1σ\delta_{y,\min}=0.1\sigma, and above by the maximum, δy,max⁡=0.5σ\delta_{y,\max}=0.5\sigma.

How soft is the magnetic flagellum?

The onset of these oscillations corresponds to a supercritical Hopf bifurcation [32, 34, 39], leading to a limit cycle in terms of the system’s mean polarization Ω(t)=1N∑i=1Nφi(t)\Omega(t)=\frac{1}{N}\sum_{i=1}^{N}\varphi_i(t) and mean bond angle Φ(t)=1N−1∑i=1N−1Θi(t)\Phi(t)=\frac{1}{N-1}\sum_{i=1}^{N-1}\Theta_i(t), where Θi=∠(bi,bi+1)\Theta_i=\angle(\mathbf{b}_i,\mathbf{b}_{i+1}) is the angle between consecutive bond vectors bi=ri+1−ri\mathbf{b}_i=\mathbf{r}_{i+1}-\mathbf{r}_i, as shown in Figure 4(a) (red for experiments, gray for simulations). This limit cycle is analogous to that observed in elastoactive structures [39] and beating biological flagella [26, 27], where mean curvature and polarization serve as natural order parameters for the oscillatory state.

Image not extracted

Figure 4. Chain flexibility and limit cycle dynamics. (a),(b) Kymographs showing the temporal evolution of the bond angles Θ1\Theta_1 and Θ2\Theta_2 (defined in the schematic inset) for simulations (a) and experiments (b), respectively. Colors represent the bond angle normalized by the maximum deflection detected, from blue (minimum) to red (maximum). The periodic banding confirms sustained flagellar beating, with simulations showing more regular oscillations than experiments due to the absence of environmental noise. (c) Phase space portrait of the mean curvature Φ(t)\Phi(t) versus mean polarization Ω(t)\Omega(t), showing the limit cycle structure characteristic of a supercritical Hopf bifurcation [34, 39]. Experimental trajectories (cyan) exhibit larger variability than simulations (red), reflecting sensitivity to rotational noise and initial conditions. (d) Time evolution of the three-body bending energy UbU_b for flagella composed of N=4N = 4 (top) and N=5N = 5 (bottom) particles, from simulations (red) and experiments (cyan). The N=5N = 5 chain exhibits larger oscillation amplitudes and more negative mean UbU_b, consistent with its greater flexibility as predicted by the kmag/(Nσ)2k_{\mathrm{mag}}/(N\sigma)^2 scaling [33, 51].

Next, we investigate the flexibility of the magnetic flagellum. The magnetic bending stiffness per particle pair, κmag=μ0m2/2πσ2\kappa_{\rm mag}=\mu_0m^2/2\pi\sigma^2, sets the resistance to relative angular deflection between neighboring MSPPs [9, 51]. The effective rigidity of the chain as a whole, however, is governed by the Euler critical force Fc=κmag/(Nσ)2F_c=\kappa_{\rm mag}/(N\sigma)^2, which decreases as N−2N^{-2} with chain length [21, 33]. Consequently, shorter chains are stiffer and harder to buckle, while longer chains are more flexible and buckle under weaker active forcing, consistent with our experimental observations (see Supplementary Movies SM1–SM4).

To quantify this flexibility, we analyze the bond angles Θi\Theta_i, defined in the schematic of Figure 4, and compute kymographs showing their time evolution for simulations and experiments in Figure 4(b),(Figure 4c), respectively. Colors represent the bond angle normalized by the maximum deflection detected in each realization.

We model the chain bending using a three-body potential based on the unit bond vectors bi=(ri−ri−1)/∣ri−ri−1∣\mathbf{b}_i = (\mathbf{r}_i - \mathbf{r}_{i-1})/\lvert\mathbf{r}_i - \mathbf{r}_{i-1}\rvert, yielding the total bending energy [31, 33, 51]

Ub=kmag2∑i=2N−1(bi+1−bi)2(9)U_b = \frac{k_{\mathrm{mag}}}{2} \sum_{i=2}^{N-1} (\mathbf{b}_{i+1} - \mathbf{b}_i)^2 \tag*{(9)}

where kmag=μ0m2/2πσ2k_{\mathrm{mag}} = \mu_0 m^2/2\pi\sigma^2 is the magnetic bending stiffness, which here plays the role of the bending rigidity κ\kappa in the discrete worm-like chain model [51]. Unlike passive polymer chains where κ\kappa is fixed by molecular architecture, here kmagk_{\mathrm{mag}} is tunable via the dipole moment mm, offering direct experimental control over chain flexibility [9]. The bending energy over time is shown in Figure 4(d),(Figure 4e), where oscillation amplitudes grow with NN in agreement with the predicted kmag/(Nσ)2k_{\mathrm{mag}}/(N\sigma)^2 scaling [47].

Our work provides a physical realization of beating by propulsion [20, 21] in a model system where both bending rigidity and activity arise from controllable, experimentally tunable parameters. By combining magnetic dipolar interactions with self-propelled motion in centimeter-scale robots (Hexbugs) [43, 44], we construct magnetic self-propelled particles (MSPPs) that self-assemble into flexible chains and undergo spontaneous flagellar beating when one end is clamped. Unlike microscale systems where flagellar motion typically results from internal motor activity or elastic deformations in slender filaments [31, 32, 33], here the oscillations are driven purely by the interplay between active forces, magnetic cohesion, and orientational fluctuations. Elasticity emerges effectively from the magnetic dipole-dipole interactions, giving rise to a bending rigidity kmag=μ0m2/2πσ2k_{\mathrm{mag}} = \mu_0m^2/2\pi\sigma^2 that can be modulated via dipole strength [9, 51], while the effective chain flexibility grows with particle number as Fc=kmag/(Nσ)2F_c = k_{\mathrm{mag}}/(N\sigma)^2, consistent with Euler buckling theory [17, 21, 33]. The onset of beating is initiated by a critical tangential misalignment δ\delta that triggers buckling, and rotational diffusion plays an essential role in symmetry breaking without it, as shown in simulations, the chain remains in a metastable straight configuration regardless of propulsion strength [16, 33].

By systematically analyzing both experiments and simulations, we demonstrate that flagellar beating is robust and reproducible, with amplitude and frequency depending nonlinearly on dipolar strength and chain length [17, 39]. The macroscopic nature of our model offers unique advantages: trajectories and orientations can be tracked with high spatiotemporal resolution using simple imaging tools [44]; propulsion strength, dipolar coupling, and chain architecture can be modified in a controlled manner; and long timescales allow direct observation of transient and steady-state dynamics, including transitions between linear and nonlinear regimes [39, 47].

The observed flagellar beating is reminiscent of locomotion in spermatozoa and cilia [23, 26, 28], suggesting that the underlying principles force-induced buckling coupled to orientational dynamics may be generic across scales [27, 34]. While our current system focuses on a single flagellum, future studies could explore collective effects in systems of multiple interacting chains, where flagellar synchronization via magnetic or elastic coupling may give rise to behaviors analogous to those observed in biological arrays of cilia and flagella [40, 41, 42].

Our system further provides a platform for testing reduced models of active filaments, including symmetry-breaking bifurcations, limit cycles, and mode selection in non-linear oscillators [32, 34, 39] offering insights directly transferable to synthetic microswimmers [15, 37] and biological systems governed by active forces and mechanical constraints.

Acknowledgements

We thank Professor Eric Clement for helpful discussions. N. Q, D.H and F.G.-L. have received support from Fondecyt Regular No. 1250913.

Author contributions

All authors contributed equally to the analysis and writing of the manuscript; F.G.-L, E.L. and E.B. designed the research, F.G.-L. performed the simulations. N.Q, D.H and E.B performed the experiments. N.Q. and D.H wrote the scripts to analyze the experimental data.

Additional information

There are no conflicts to declare.

References

  1. [1]H. Xie, M. Sun, X. Fan, Z. Lin, W. Chen, L. Wang, L. Dong, and Q. He, Reconfigurable magnetic microrobot swarm: Multimode transformation, locomotion, and manipulation, Science Robotics 4, eaav8006 (2019).DOI
  2. [2]K. J. Bishop, S. L. Biswal, and B. Bharti, Active colloids as models, materials, and machines, Annual Review of Chemical and Biomolecular Engineering 14, 1 (2023).DOI
  3. [3]R. McCallum, L. H. Lewis, R. Skomski, M. Kramer, and I. Anderson, Practical aspects of modern and future permanent magnets, Annual Review of Materials Research 44, 451 (2014).DOI
  4. [4]L. Yang and L. Zhang, Motion control in magnetic microrobotics: From individual and multiple robots to swarms, Annual Review of Control, Robotics, and Autonomous Systems 4, 509 (2021).DOI
  5. [5]F. Meng, D. Matsunaga, J. M. Yeomans, and R. Golestanian, Magnetically-actuated artificial cilium: a simple theoretical model, Soft Matter 15, 3864 (2019).DOI
  6. [6]V. V. Khatavkar, P. D. Anderson, J. M. den Toonder, and H. E. Meijer, Active micromixer based on artificial cilia, Physics of Fluids 19 (2007).DOI
  7. [7]F. R. Koessel and S. Jabbari-Farouji, Emergent pattern formation of active magnetic suspensions in an external field, New Journal of Physics 22, 103007 (2020).DOI
  8. [8]F. Martinez-Pedrero, A. Ortiz-Ambriz, I. Paganobarraga, and P. Tierno, Colloidal microworms propelling via a cooperative hydrodynamic conveyor belt, Physical review letters 115, 138301 (2015).
  9. [9]B. Kiani, D. Faivre, and S. Klumpp, Elastic properties of magnetosome chains, New Journal of Physics 17, 043007 (2015).
  10. [10]P. Tierno, R. Golestanian, I. Pagonabarraga, and F. Sagués, Magnetically actuated colloidal microswimmers, The Journal of Physical Chemistry B 112, 16525 (2008).DOI
  11. [11]S. Klumpp, C. T. Lefèvre, M. Bennet, and D. Faivre, Swimming with magnets: from biological organisms to synthetic devices, Physics Reports 789, 1 (2019).DOI
  12. [12]A. Théry, A. Chamolly, and E. Lauga, Controlling confined collective organization with taxis, Physical Review Letters 132, 108301 (2024).DOI
  13. [13]B. Vincenti, G. Ramos, M. L. Cordero, C. Douarche, R. Soto, and E. Clement, Magnetotactic bacteria in a droplet self-assemble into a rotary motor, Nature communications 10, 5082 (2019).
  14. [14]X. Wang, P.-C. Chen, K. Kroy, V. Holubec, and F. Cichos, Spontaneous vortex formation by microswimmers with retarded attractions, Nature Communications 14, 56 (2023).DOI
  15. [15]H. R. Vutukuri, B. Bet, R. Van Roij, M. Dijkstra, and W. T. Huck, Rational design and dynamics of self-propelled colloidal bead chains: from rotators to flagella, Scientific Reports 7, 16578 (2017).
  16. [16]F. Guzman-Lastra, A. Kaiser, and H. Löwen, Fission and fusion scenarios for magnetic microswimmer clusters, Nature Communications 7, 13519 (2016).
  17. [17]D. Nishiguchi, J. Iwasawa, H.-R. Jiang, and M. Sano, Flagellar dynamics of chains of active janus particles fueled by an ac electric field, New Journal of Physics 20, 015002 (2018).
  18. [18]G.-J. Liao and S. H. Klapp, Emergent vortices and phase separation in systems of chiral active particles with dipolar interactions, Soft Matter 17, 6833 (2021).DOI
  19. [19]N. Vanesse, E. Opsomer, G. Lumay, and N. Vandewalle, Collective dynamics of dipolar self-propelled particles, Physical Review E 108, 024608 (2023).DOI
  20. [20]G. Jayaraman, S. Ramachandran, S. Ghose, A. Laskar, M. S. Bhamla, P. S. Kumar, and R. Adhikari, Autonomous motility of active filaments due to spontaneous flow-symmetry breaking, Physical review letters 109, 158302 (2012).DOI
  21. [21]S. Camalet, F. Jülicher, and J. Prost, Self-organized beating and swimming of internally driven filaments, Physical review letters 82, 1590 (1999).DOI
  22. [22]K. Sekimoto, N. Mori, K. Tawada, and Y. Y. Toyoshima, Symmetry breaking instabilities of an in vitro biological system, Physical review letters 75, 172 (1995).pubmed.ncbi.nlm.nih.gov/10059143
  23. [23]W. Gilpin, M. S. Bull, and M. Prakash, The multiscale physics of cilia and flagella, Nature Reviews Physics 2, 74 (2020).DOI
  24. [24]S. Werner, J. C. Rink, I. H. Riedel-Kruse, and B. M. Friedrich, Shape mode analysis exposes movement patterns in biology: flagella and flatworms as case studies, PLoS one 9, e113083 (2014).DOI
  25. [25]M. T. Gallagher, G. Cupples, E. H. Ooi, J. Kirkman-Brown, and D. Smith, Rapid sperm capture: high-throughput flagellar waveform analysis, Human Reproduction 34, 1173 (2019).
  26. [26]J. S. Guasto, J. B. Estrada, F. Menolascina, L. J. Burton, M. Patel, C. Franck, A. Hosoi, R. K. Zimmer, and R. Stocke, Flagellar kinematics reveals the role of environment in shaping sperm motility, Journal of the Royal Society Interface 17, 20200525 (2020).DOI
  27. [27]R. Ma, G. S. Klindt, I. H. Riedel-Kruse, F. Jülicher, and B. M. Friedrich, Active phase and amplitude fluctuations of flagellar beating, Physical review letters 113, 048101 (2014).DOI
  28. [28]J. Cammann, H. Laeverenz-Schlogelhofer, K. Y. Wan, and M. G. Mazza, Form and function in biological filaments: A physicist’s review, Philos. Trans. A Math. Phys. Eng. Sci. 383, 20240253 (2025).arxiv.org/abs/2502.12731
  29. [29]M. Lisicki, M. F. V. Rodrigues, and E. Lauga, Eukaryotic swimming cells are shaped by hydrodynamic constraints, Journal of Fluid Mechanics 978, R1 (2024).arxiv.org/abs/2310.03345
  30. [30]Y. Fily, P. Subramanian, T. M. Schneider, R. Chelakkot, and A. Gopinath, Buckling instabilities and spatio-temporal dynamics of active elastic filaments, Journal of The Royal Society Interface 17, 20190794 (2020).DOI
  31. [31]A. Laskar and R. Adhikari, Brownian microhydrodynamics of active filaments, Soft matter 11, 9073 (2015).arxiv.org/abs/1508.02376
  32. [32]G. De Canio, E. Lauga, and R. E. Goldstein, Spontaneous oscillations of elastic filaments induced by molecular motors, Journal of The Royal Society Interface 14, 20170491 (2017).
  33. [33]R. Chelakkot, A. Gopinath, L. Mahadevan, and M. F. Hagan, Flagellar dynamics of a connected chain of active, polar, brownian particles, Journal of The Royal Society Interface 11, 20130884 (2014).
  34. [34]F. Ling, H. Guo, and E. Kanso, Instability-driven oscillations of elastic microfilaments, Journal of The Royal Society Interface 15, 20180594 (2018).arxiv.org/abs/1808.01583
  35. [35]Y. Xi, T. Marzin, R. B. Huang, T. J. Jones, and P.-T. Brun, Emergent behaviors of buckling-driven elasto-active structures, Proceedings of the National Academy of Sciences 121, e2410654121 (2024).DOI
  36. [36]S. Al-Izzi, Y. Du, J. Veenstra, R. Morris, A. Souslov, A. Carlson, C. Coulais, and J. Binysh, Nonreciprocal buckling makes active filaments polyfunctional, Proceedings of the National Academy of Sciences 123, e2531723123 (2026).
  37. [37]S. Gonzalez and R. Soto, Active colloidal chains with cilia- and flagella-like motion, New Journal of Physics 20, 053014 (2018).arxiv.org/abs/1803.08441
  38. [38]L. L. Gutierrez-Martinez and M. Sandoval, Inertial effects on trapped active matter, The Journal of Chemical Physics 153 (2020).
  39. [39]E. Zheng, M. Brandenbourger, L. Robinet, P. Schall, E. Lerner, and C. Coulais, Self-oscillation and synchronization transitions in elastoactive structures, Physical Review Letters 130, 178202 (2023).DOI
  40. [40]Y. Xia, Z. Hu, D. Wei, K. Chen, Y. Peng, and M. Yang, Biomimetic synchronization in biciliated robots, Physical Review Letters 133, 048302 (2024).DOI
  41. [41]K. Y. Wan and R. N. Poon, Mechanisms and functions of multiciliary coordination, Current Opinion in Cell Biology 86, 102286 (2024).DOI
  42. [42]C. Moreau, B. J. Walker, R. N. Poon, D. Soto, D. I. Goldman, E. A. Gaffney, and K. Y. Wan, Minimal design of a synthetic cilium, Physical Review Research 6, L042061 (2024).DOI
  43. [43]M. Musacchio, M. Felber, M. Paoluzzi, A. Gnoli, A. Puglisi, and L. Angelani, Fluidization induced by magnetic interactions in confined active matter, Physical Review E 113, 055413 (2025).arxiv.org/abs/2511.21472
  44. [44]P. M. Obreque, O. Garrido, D. Romero, H. Löwen, and F. Guzmán-Lastra, Dynamics of magnetic self-propelled particles in a harmonic trap, arXiv preprint arXiv:2403.02569 (2024).arxiv.org/abs/2403.02569
  45. [45]N. Sepúlveda, E. Guzmán-Lastra, M. Carrasco, B. González, E. Hamm, and A. Concha, Bioinspired magnetic active matter and the physical limits of magnetotaxis, arXiv preprint arXiv:2111.04889 (2021).arxiv.org/abs/2111.04889
  46. [46]A. Kaiser, K. Popova, and H. Löwen, Active dipole clusters: From helical motion to fission, Physical Review E 92, 012301 (2015).DOI
  47. [47]A. R. Tejedor, J. Ramírez, and M. Ripoll, Progressive polymer deformation induced by polar activity and the influence of inertia, Physical Review E 6, L032002 (2024).
  48. [48]Hexbug, Hexbug is a brand of toy automata developed and distributed by innovation first, https://www.hexbug.com/.
  49. [49]See Supplemental Material at [URL will be inserted by publisher] for details of the numerical methods, additional figures, and supplementary analyses.
  50. [50]M. Leoni, M. Paoluzzi, S. Eldeen, A. Estrada, L. Nguyen, M. Alexandrescu, K. Sherb, and W. V. Ahmed, Surfing and crawling macroscopic active particles under strong confinement: Inertial dynamics, Physical Review Research 2, 043299 (2020).arxiv.org/abs/2005.10768
  51. [51]D. Li, S. Banon, and S. L. Biswal, Bending dynamics of dna-linked colloidal particle chains, Soft Matter 6, 4197 (2010).DOI

Paper details

vlnc:W_6k9jwnc7
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 2

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents