betaFeedback

WorksPhysics and Astronomy

A Kirchhoff-like theory for hard magnetic rods under geometrically nonlinear deformation in three dimensions

Abstract

Magneto-rheological elastomers (MREs) are functional materials that can be actuated by applying an external magnetic field. MREs comprise a composite of hard magnetic particles dispersed into a nonmagnetic elastomeric matrix. By applying a strong magnetic field, one can magnetize the structure to program its deformation under the subsequent application of an external field. Hard MREs, whose coercivities are large, have been receiving particular attention because the programmed magnetization remains unchanged upon actuation. Hence, once a structure made of a hard MRE is magnetized, it can be regarded as magnetized permanently. Motivated by a new realm of applications, there have been significant theoretical developments in the continuum description of hard MREs. By reducing the 3D description into 1D or 2D via dimensional reduction, several theories of hard magnetic slender structures such as linear beams, elastica, and shells have been recently proposed. In this paper, we derive an effective theory for MRE rods under geometrically nonlinear 3D deformation. Our theory is based on reducing the 3D magneto-elastic energy functional for the hard MREs into a 1D Kirchhoff-like description. Restricting the theory to 2D, we reproduce previous works on planar deformations. For further validation in the general case of 3D deformation, we perform precision experiments with both naturally straight and curved rods under either constant or constant-gradient magnetic fields. Our theoretical predictions are in excellent agreement with both discrete simulations and precision-model experiments. Finally, we discuss some limitations of our framework, as highlighted by the experiments, where long-range dipole interactions, which are neglected in the theory, can play a role.

Introduction

Leveraging the coupling between internal degrees of freedom and material strains to open up new avenues for structural instabilities has been receiving much recent attention in the mechanics and advanced functional material communities [1, 2]. The magneto-elastic coupling is particularly interesting since it can be exploited to actuate soft magnetic structures without any mechanical contact. Magneto-rheological elastomers (MREs) comprise a composite of magnetized (metal) particles and a soft elastic matrix that can respond to an applied magnetic field. MREs have long been studied, from both mathematical modeling and application perspectives [2, 3]. For instance, tunable mechanical functionalities of magnetic structures have been used in the applications such as sensing medical devices [1], engine mounts [4], soft robotics [5], actuators [6], and minimally invasive procedures [7].

The mechanical responses of MREs to an applied field can be categorized into the following (i)-(v) cases, depending on their ingredients, such as the coercivity, which is the necessary field strength to erase the magnetization. (i) Super-para-magnetic and soft-ferromagnetic materials have large magnetic susceptibility, enabling the temporary magnetization by an external magnetic field [8, 9, 10, 11, 12, 13, 14]; a super-para-magnetic material loses its magnetization, once the external field is removed. As a result, the magnetization of these materials depends strongly on the external fields. (ii) Hard-ferromagnetic materials have a high coercivity. An external field does not change the magnetization within the range of practical actuation. Hence, their magnetization is independent of external fields. (iii) Chains of hard-ferromagnetic spheres are purely magnetic systems without elasticity [15, 16, 17]. The complex internal interaction among beads yields a variety of equilibrium configurations. Interestingly, these chains posses an effective bending stiffness [16]. (iv) MREs embedded with soft-ferromagnetic particles or soft magnetic elastomer [1, 4, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27] deform upon the application of an external magnetic field due to elasto-magnetic couplings. The magnetic particles with low coercivity dispersed in a soft matrix align themselves with the applied field, thereby forming particle chains. To enhance the elasto-magnetic coupling, it is common to apply the magnetic field to MRE during the curing processes. (v) MREs embedded with hard-ferromagnetic particles or hard magnetic elastomer have both high coercivity and flexibility in shape-programming [7, 28, 29, 30, 31, 32, 33, 34, 35]. Once they are saturated magnetically, they retain permanent magnetization. For the remainder of this paper, we will focus on this latter type (v).

The high coercivity of hard MREs enables the design of highly functional mechanical systems, such as auxetic metamaterials [29], programmable materials [33], micro-swimmers [36, 37, 38], micro or soft-robotics [5, 6, 39] and haptic devices [40]. Motivated by these emerging applications of hard MREs, it is timely to formulate structural theories for elementary building blocks made of hard MREs to serve as predictive tools in the design process and aid in subsequent analysis.

During the past decade, there have been significant developments in the fundamental theory of hard magnetic elastomers. In a recent pioneering work by Zhao et al. [30], a nonlinear theory for finite deformation of 3D (bulk) hard MRE was developed, based on the nonlinear elasticity complemented by the magneto-mechanical constitutive laws. In this framework, the Helmholtz free energy density function comprises elastic (neo-Hookean) and magneto-elastic parts. The relationship between the induced magnetic flux density in the material and the external field strength is assumed to be linear, with a permeability close to that of vacuum. The free energy density for the magneto-elastic part is modeled such that the magnetic moment is embedded in a soft matrix. A simulation framework using the finite element method developed by the same authors, based on nonlinear elasticity in 3D, was shown to be in quantitative agreement with their experimental results [30].

Based on the framework reported in Ref. [30], theories for hard magnetic linear beams and elastica have been derived and validated against experiments under either constant magnetic field [31] or a field with constant gradient [35]. Furthermore, it was shown how a hard magnetic beam (or elastica) can buckle under an applied field, analogous to an elastica under compression [28, 31, 32, 35, 41]. Methodologically, the derivations of all of the above reduced models for hard magnetic linear beam and elastica always boils down to the formal procedure of dimensional reduction [42, 43], which is summarized next. More specifically, the total energy functional of a beam is modeled as Ref. [30], with elastic and magneto-elastic parts. The integrand is expanded with respect to the thickness of the cross-section, only retaining the leading order terms. Upon integration over the cross-section, a one-dimensional (1D)-reduced energy is derived. The reduced framework is consistent with the analogous procedure based on force and moment balance [28], which was also implemented in the study of the inverse problem [32]. A similar strategy based on dimensional reduction has also been employed to predict the axisymmetric deformation of pressurized hard magnetic shells [34].

While theoretical frameworks for one-dimensional hard magnetic slender structures are now well established for planar 2D deformations (e.g., beams and elastica), the modeling of hard magnetic rods, with natural curvatures, undergoing 3D deformation is still lacking. However, the extension of the formulation in 2D toward 3D deformation is not straightforward because of the complex interplay between elasticity and geometry [42, 43, 44, 45, 46, 47]. The precise modeling of a hard magnetic elastomeric rod is crucial to simulate, for example, micro-magnetic-swimmers [36, 37, 38] in a 3D complex channel, and endovascular probes [7].

In this paper, we establish a theory for the deformation of hard magnetic elastomeric rods, combining dimensional reduction and simulations, both of which are validated against precision experiments. For our theory, we start from the total energy functional comprising the sum of the Hookean elastic energy [43] and the elasto-magnetic free-energy proposed in Ref. [30]. We then employ the centerline-based kinematics utilizing the Cosserat frame and Darboux vector [47]. Along the (circular) cross-section, we span the (local) polar coordinate around the centerline position to parametrize the material points of the hard magnetic rod. Subsequently, we expand the integrand of the elasto-magnetic energy for the rod diameter, retaining only the leading order terms. After integrating the integrand over the cross-section, we obtain the reduced elasto-magnetic functional. Based on this reduced elasto-magnetic functional, we then apply the principle of virtual work to arrive at the governing nonlinear ordinary differential equations (strong form of the problem). The derived equilibrium equations are analogous to the Kirchhoff rod equations, but with additional elasto-magnetic terms; hence, we call them the magnetic Kirchhoff rod equations. The nonlinear deformations based on the magnetic Kirchhoff rod equations are computed with the discrete simulation method [48, 49, 50, 51, 52, 53, 54, 55, 56, 57]. Both the dimensional reduction procedure and the simulations are then validated against precision experiments. Since our theory is based on the assumption that the magnetic torque is induced by the misalignment between the magnetization and the applied field, the long-range dipole interactions are not modeled. Still, we find excellent agreement between theory and experiments as long as the long-range dipole interactions are negligible. We quantify when and how the long-range interactions become important in the experiments.

Our paper is organized as follows. In § 2, we define the problem. The derivation of the magnetic Kirchhoff equations follows in § 3, where we also show that our theory reproduces the previous results of planar deformation of hard magnetic elastica and beams. To compute 3D geometrically nonlinear deformation, we numerically solve the derived set of equations as detailed in § 4. In § 5, the experimental fabrication, apparatus, and protocols are presented. We report the comparison between theory and experiment in § 6. We further discuss the limitation of our theory systematically in § 7. In § 8, we discuss and summarize our findings.

Definition of the problem

In this section, we define the problem and identify the relevant variables. We consider a naturally curved and twisted (inextensible) rod of circular cross-section with diameter dd (area A=πd2/4A = \pi d^2/4) and total length LL (Figure 1(a)) [43, 44, 45, 46, 47]. The goal of our study was to derive a reduced-order theoretical framework to describe the deformation of the centerline, r(s)\boldsymbol{r}(s), of this hard magnetic rod, upon magnetic actuation. The arc-length of the centerline is ss (0≤s≤L0 \le s \le L). In particular, we seek to derive the set of governing ordinary differential equations (ODEs); i.e., the strong form of the problem depicting force and moment balance of the system. This section details the kinematics and constitutive relations of our system, in § 2.1 and 2.2, respectively.

Definition of the problem

Figure 1. Definition of the problem. (a) A naturally curved and twisted rod is magnetized permanently along Br\mathbf{B}^{r}. (b) Subsequently, when a magnetic field Ba\mathbf{B}^{a} is applied, the rod deforms due to the induced magnetic force and torque. The centerline of the rod is described uniquely by the rotation of the Cosserat frame (d^1,d^2,d^3)(\hat{\mathbf{d}}_1,\hat{\mathbf{d}}_2,\hat{\mathbf{d}}_3) along the arc-length ss. (c) The cross section of the rod is assumed circular. Any material point of the 3D rod is identified by the cylindrical coordinates (s,ρ,Θ)(s,\rho,\Theta), where polar coordinates (ρ,Θ)(\rho,\Theta) are set on the d^1−d^2\hat{\mathbf{d}}_1-\hat{\mathbf{d}}_2 plane for each value of ss.

Kinematics

Our rod is slender such that the diameter is significantly smaller than its length; d/L≪1d/L \ll1. Throughout this paper, we denote the quantities in the reference configuration with (⋅)(\cdot). For example, we write the centerline position in the reference configuration as r∘(s)\overset{\circ}{\boldsymbol{r}}(s). The 2D orthogonal basis d^1\hat{\mathbf{d}}_1-d^2\hat{\mathbf{d}}_2 is defined on the cross-section normal to the tangent vector d^3=r′(s)\hat{\mathbf{d}}_3 = \boldsymbol{r}'(s), where (⋅)′(\cdot)' represents differentiation with respect to ss. With the help of the Cosserat frame basis d^a\hat{\mathbf{d}}_a (a=1,2,3a = 1, 2, 3) defined here, one can uniquely identify the configuration of the rod centerline (other than rigid body translations and rotations). The rotation of the Cosserat frame can be represented using the Darboux vector Ω=Ωad^a\boldsymbol{\Omega}=\Omega_a\hat{\mathbf{d}}_a as

d^˙a=Ω×d^a(1)\dot{\hat{\mathbf{d}}}_a=\boldsymbol{\Omega}\times\hat{\mathbf{d}}_a \tag*{(1)}

where Ωa\Omega_a are the rotation rates of the Cosserat frame around d^a\hat{\mathbf{d}}_a [43, 44, 45, 46, 47]. We can then define the slenderness parameters as Ω˙ad\dot{\Omega}_a d.

The 3D reference configuration of the rod can be described uniquely by introducing the cylindrical coordinates ξμ=s,ρ,Θ\xi_\mu=s,\rho,\Theta (0≤ρ≤d/20\leq\rho\leq d/2, 0≤Θ≤2π0\leq\Theta\leq2\pi), whose origin corresponds to r(s)\mathbf{r}(s) as shown in Figure 1 (c). For example, the position vector of the 3D rod is

R(s,ρ,Θ)=r(s)+ρcos⁡Θ d^1(s)+ρsin⁡Θ d^2(s).(2)\mathbf{R}(s,\rho,\Theta)=\mathbf{r}(s)+\rho\cos\Theta\,\hat{\mathbf{d}}_1(s)+\rho\sin\Theta\,\hat{\mathbf{d}}_2(s). \tag*{(2)}

Once the geometry is set up, we can obtain information about the deformation of the 3D body, which is characterized by the deformation gradient F\mathbf{F}, defined as

F≡∂R∂R˚=tμ⊗t˚μ.(3)\mathbf{F}\equiv\frac{\partial\mathbf{R}}{\partial\mathring{\mathbf{R}}}=\mathbf{t}_\mu\otimes\mathring{\mathbf{t}}^\mu. \tag*{(3)}

The tangent vector along the curvilinear coordinate, tμ\mathbf{t}_\mu, is expressed as tμ=∂R/∂ξμ\mathbf{t}_\mu=\partial\mathbf{R}/\partial\xi_\mu, where ξμ=s,ρ,Θ\xi_\mu=s,\rho,\Theta for μ=1,2,3\mu=1,2,3, respectively. The reciprocal basis of tμ\mathbf{t}_\mu denoted by, t˚μ\mathring{\mathbf{t}}^\mu, is defined as t˚μ⋅tν=δνμ\mathring{\mathbf{t}}^\mu\cdot\mathbf{t}_\nu=\delta^\mu_\nu with the Kronecker delta δνμ\delta^\mu_\nu, which is equivalently written as t˚μ≡∂ξμ/∂R\mathring{\mathbf{t}}^\mu\equiv\partial\xi^\mu/\partial\mathbf{R}. By using the kinematic relations, we can calculate the tangent vectors. Note, for example, that when t˚μ\mathring{\mathbf{t}}^\mu and tμ\mathbf{t}_\mu are expressed in terms of the Cartesian basis e^i\hat{\mathbf{e}}_i (i=x,y,zi=x,y,z), each component of F\mathbf{F} is calculated as Fij=(tμ⋅e^i)(t˚μ⋅e^j)F_{ij}=(\mathbf{t}_\mu\cdot\hat{\mathbf{e}}_i)(\mathring{\mathbf{t}}^\mu\cdot\hat{\mathbf{e}}_j). Throughout this paper, a,b,ca,b,c and i,j,ki,j,k are indices of the Cosserat frame and the Cartesian basis, respectively, if not specified.

Constitutive relation and Elasto-magnetic potential

We assume that the rod is made of a Hookean material, where the internal stress is linearly proportional to the strain [43]. The rod is permanently magnetized in the reference configuration with a magnetic flux density Ba\mathbf{B}^a. We investigate the deformation of the rod upon the application of an external applied magnetic flux density vector Ba(r)\mathbf{B}^a(\mathbf{r}) (Figure 1(b)). Note that the applied field Ba\mathbf{B}^a is set to vary in space. We will specialize our theory in the cases of either homogeneous ∂Ba/∂ri=0\partial B^a/\partial r_i=0 or inhomogeneous fields ∂Ba/∂ri≠0\partial B^a/\partial r_i\ne0, making use of the precision experimental framework introduced in section 5.2.

As a starting point for the derivation of our Kirchhoff-like theory for hard magnetic rods, we will make use of the Helmholtz free energy for ideal hard-magnetic soft materials that was introduced recently in Refs. [29, 30]; which we review next. Once the rod is magnetized, all magnetized moments are parallel with Br\mathbf{B}^r (fully saturated). Hence, it is reasonable to assume that the magnetic permeability of the hard magnetic material is close to that of vacuum μ0\mu_0. The magnetized moments will not affect the surrounding applied field Ba\mathbf{B}^a upon actuation. Besides, the permeability of the elastomer that we will use in the experiments (section 5.2) is nearly the same as that of the vacuum μ0\mu_0. The magnetization density vector per unit volume in the deformed configuration is modeled as the “rotation” of Br\mathbf{B}^r as M3D=FBr/μ0\mathcal{M}_{3D}=\mathbf{F}\mathbf{B}^r/\mu_0. Thus, the total elasto-magnetic energy for a hard magnetic rod can be simplified as the work to align the residual magnetic moment of the material with Ba\mathbf{B}^a:

Emag=−∭M3DBaρ dρ dΘ ds=−∭1μ0(FBr)⋅Baρ dρ dΘ ds.(4)E_{\mathrm{mag}}=-\iiint\mathcal{M}_{3D}\mathbf{B}^a\rho\,d\rho\,d\Theta\,ds=-\iiint\frac{1}{\mu_0}(\mathbf{F}\mathbf{B}^r)\cdot\mathbf{B}^a\rho\,d\rho\,d\Theta\,ds. \tag*{(4)}

Note here that we do not consider the internal stress or long-range forces induced by a dipole-dipole interaction [15, 16, 58, 59]. The validity of this simplifying assumption will be evaluated thoroughly in § 7. In the following, by integrating Eq. (4) over the cross-section, we seek to obtain the reduced expression for the elasto-magnetic energy density of the rod per a unit length, up to the leading order in the slenderness.

Theory of a Kirchhoff-like equation for hard magnetic elastic rods

In this section, we derive the equilibrium equations (ODEs) describing the geometrically nonlinear deformation of hard magnetic rods. The main results of this section are Eqs. (20) and (21), that is the magnetic Kirchhoff rod equations. To arrive at this set of ODEs, in § 3.1, we will follow the procedure of dimensional reduction. In § 3.2, we review the derivation of the non-magnetic Kirchhoff rod equations based on the principle of virtual work. We derive the elasto-magnetic force pmag\mathbf{p}_{\mathrm{mag}} and torque qmag\mathbf{q}_{\mathrm{mag}} through variational calculus in § 3.3, thus obtaining the magnetic Kirchhoff rod equations for hard magnetic rods. In § 3.4, we show that our framework reproduces previous works on the planar deformation of hard magnetic beams and elastica [28, 31, 32, 35], whereas in § 6 we will employ our framework to study cases where the deformation is three-dimensional.

Reduced elasto-magnetic and elastic energy

Integrating Eq. (4) over the cross-section will yield the 1D reduced energy of the rod system, to leading order of slenderness O(1)O(1) (Ωad∗,Ωad≪1\Omega^{*}_{a d}, \Omega_{a d} \ll1). To proceed, we first expand the deformation gradient as F(s,ρ,Θ)=D(s)+O(Ωad∗,Ωad)\mathbf{F}(s,\rho,\Theta)=\mathbf{D}(s)+O(\Omega^{*}_{a d},\Omega_{a d}), where

D(s)≡d^a(s)⊗a^d(s)(5)\mathbf{D}(s) \equiv\hat{\mathbf{d}}_{a}(s) \otimes\hat{\mathbf{a}}_{d}(s) \tag*{(5)}

is the reduced deformation gradient, mapping the undeformed tangent space to the deformed tangent space along the centerline of the rod. The deformation gradient is now reduced to the tensor product between deformed and undeformed directors, as given by Eqs. (3) and (5). Given that D\mathbf{D} is not a function of ρ\rho nor Θ\Theta, the integral along the cross-section is replaced by the constant cross-section area AA. Hence, from Eq. (4), we obtain the reduced elasto-magnetic energy functional as

Emag=−∫0LM(s)⋅Ba(r(s)) ds,(6)E_{\mathrm{mag}}=-\int_{0}^{L}\boldsymbol{\mathcal{M}}(s)\cdot\mathbf{B}^{a}(\mathbf{r}(s))\,\mathrm{d}s, \tag*{(6)}

where we have introduced the magnetization density vector per a unit length M\boldsymbol{\mathcal{M}} (by contrast to the 3D counterpart, M3D\boldsymbol{\mathcal{M}}_{\mathrm{3D}}, introduced above) defined as

M(s)≡Aμ0(D(s)Br).(7)\boldsymbol{\mathcal{M}}(s)\equiv\frac{A}{\mu_{0}}\left(\mathbf{D}(s)\mathbf{B}^{r}\right). \tag*{(7)}

It is important to note that, when the applied field Ba\mathbf{B}^{a} is not uniform, Ba\mathbf{B}^{a} at the material point R(s,ρ,Θ)\mathbf{R}(s,\rho,\Theta) is represented by the value at the centerline position r(s)\mathbf{r}(s) as Ba(R(s,ρ,Θ))≃Ba(r(s))\mathbf{B}^{a}(\mathbf{R}(s,\rho,\Theta))\simeq\mathbf{B}^{a}(\mathbf{r}(s)). The spatial variation of Ba\mathbf{B}^{a} along the cross section is at higher order of slenderness O(Ωad∗,Ωad)O(\Omega^{*}_{a d},\Omega_{a d}) as Ba(R)=Ba(r)+O(Ωad∗,Ωad)\mathbf{B}^{a}(\mathbf{R})=\mathbf{B}^{a}(\mathbf{r})+O(\Omega^{*}_{a d},\Omega_{a d}), and can, therefore, be neglected.

We now want to derive the equilibrium equations, augmenting the non-magnetic Kirchhoff rod equations. In the absence of magnetic effects, we recall that, from the classic literature [43, 44, 45, 46, 47], the total elastic energy for a naturally curved and twisted elastic (Hookean) rod in the absence of a magnetic field is given by

Eel=∫0LEI2{(Ω1−Ω˙1)2+(Ω2−Ω˙2)2}+GJ2(Ω3−Ω˙3)2 ds,(8)E_{\mathrm{el}}=\int_{0}^{L}\frac{EI}{2}\left\{(\Omega_{1}-\dot{\Omega}_{1})^{2}+(\Omega_{2}-\dot{\Omega}_{2})^{2}\right\}+\frac{GJ}{2}(\Omega_{3}-\dot{\Omega}_{3})^{2}\,ds, \tag*{(8)}

where EIEI and GJGJ are the bending and twist moduli, respectively, with EE as the Young modulus, G=E/{2(1+ν)}G=E/\{2(1+\nu)\} as the shear modulus with the Poisson ratio ν\nu, I=πd4/64I=\pi d^{4}/64 as the second moment of area and J=πd4/32J=\pi d^{4}/32 as the torsional constant.

The set of equations of mechanical equilibrium is derived through the principle of virtual work as

δEel+δEmag=δWex,(9)\delta E_{\mathrm{el}}+\delta E_{\mathrm{mag}}=\delta W_{\mathrm{ex}}, \tag*{(9)}

where WexW_{\mathrm{ex}} is the work done by the external forces and torques, (Fex(0),Mex(0))(\boldsymbol{F}_{\mathrm{ex}}(0),\boldsymbol{M}_{\mathrm{ex}}(0)) and (Fex(L),Mex(L))(\boldsymbol{F}_{\mathrm{ex}}(L),\boldsymbol{M}_{\mathrm{ex}}(L)), applied at s=0s=0 and s=Ls=L, respectively, together with the external forces p(s)\boldsymbol{p}(s) and torques q(s)\boldsymbol{q}(s) per unit length (e.g., gravity, viscous force and torque).

The magnetic Kirchhoff rod equations can be derived directly from Eq. (9), which will be tackled in § 3.3. Before describing our theory for magnetic rods, we review the well-established and classic (non-magnetic) Kirchhoff rod equations. Note that δEel=δWex\delta E_{\mathrm{el}}=\delta W_{\mathrm{ex}} corresponds to the classic case in the absence of the magnetic field. Given that EmagE_{\mathrm{mag}} is additive in the total energy, the variation of EmagE_{\mathrm{mag}} alone will yield the expressions of the elasto-magnetic force pmag\boldsymbol{p}_{\mathrm{mag}} and torque qmag\boldsymbol{q}_{\mathrm{mag}} per unit length, which must be added to the force and moment balance equations as an external force and torque, respectively.

Review of the (non-magnetic) Kirchhoff rod equations

In the previous section, we have derived the reduced expressions for the elasto-magnetic potential. In the absence of the magnetic field Emag=0E_{\mathrm{mag}}=0, the variations of EelE_{\mathrm{el}} and WexW_{\mathrm{ex}} follow the classic (non-magnetic) Kirchhoff rod equations. In this section, for completeness, we review their classic derivation. The Kirchhoff rod equations can be derived by considering the restricted variation of the total energy functional such that the centerline does not stretch: δ(ds)=0\delta(ds)=0 (i.e., the differentiation with respect to ss, d/dsd/ds, and the variation, δ\delta, commute). A detailed account of this derivation can be found in Ref. [45]; in what follows, we provide just a brief sketch.

The variation of the elastic energy functional δEel\delta E_{\mathrm{el}} is written using the variation of Ωa\Omega_{a}, δΩa\delta\Omega_{a}:

δEel=∫0LEI(Ω1−Ω˙1)δΩ1+EI(Ω2−Ω˙2)δΩ2+GJ(Ω3−Ω˙3)δΩ3 ds.(10)\delta E_{\mathrm{el}}=\int_{0}^{L}EI(\Omega_{1}-\dot{\Omega}_{1})\delta\Omega_{1}+EI(\Omega_{2}-\dot{\Omega}_{2})\delta\Omega_{2}+GJ(\Omega_{3}-\dot{\Omega}_{3})\delta\Omega_{3}\,ds. \tag*{(10)}

To rewrite Eq. (10) with respect to the angular parameters and infinitesimal angles ϵ\epsilon [45], the infinitesimal rotation vector of the Cosserat frame ϵ\boldsymbol{\epsilon} is introduced as ϵ≡δχad^a\boldsymbol{\epsilon}\equiv\delta\chi_{a}\hat{\boldsymbol{d}}_{a} whose components are expressed as δχ1≡(δd^2)⋅d^3=−(δr)′⋅d^2\delta\chi_{1}\equiv(\delta\hat{\boldsymbol{d}}_{2})\cdot\hat{\boldsymbol{d}}_{3}=-(\delta\boldsymbol{r})'\cdot\hat{\boldsymbol{d}}_{2}, δχ2≡(δd^3)⋅d^1=(δr)′⋅d^1\delta\chi_{2}\equiv(\delta\hat{\boldsymbol{d}}_{3})\cdot\hat{\boldsymbol{d}}_{1}=(\delta\boldsymbol{r})'\cdot\hat{\boldsymbol{d}}_{1}, and δχ3≡(δd^1)⋅d^2=−(δd^2)⋅d^1\delta\chi_{3}\equiv(\delta\hat{\boldsymbol{d}}_{1})\cdot\hat{\boldsymbol{d}}_{2}=-(\delta\hat{\boldsymbol{d}}_{2})\cdot\hat{\boldsymbol{d}}_{1}. The end-forces and moments are applied at s=0s=0 and s=Ls=L as Fex(0)\boldsymbol{F}_{\mathrm{ex}}(0), Mex(0)\boldsymbol{M}_{\mathrm{ex}}(0), Fex(L)\boldsymbol{F}_{\mathrm{ex}}(L), and Mex(L)\boldsymbol{M}_{\mathrm{ex}}(L), respectively. The rod is subjected to the external force and torque per unit length p(s)\boldsymbol{p}(s) and q(s)\boldsymbol{q}(s). The variation of the work done by the external forces and torques (in the absence of magnetic fields) is

δWex=Fex(0)⋅δr(0)+Fex(L)⋅δr(L)+Mex(0)⋅ϵ(0)+Mex(L)⋅ϵ(L)+∫0L{p(s)⋅δr(s)+q(s)⋅ϵ(s)} ds.(11)\delta W_{\mathrm{ex}}=\boldsymbol{F}_{\mathrm{ex}}(0)\cdot\delta\boldsymbol{r}(0)+\boldsymbol{F}_{\mathrm{ex}}(L)\cdot\delta\boldsymbol{r}(L)+\boldsymbol{M}_{\mathrm{ex}}(0)\cdot\boldsymbol{\epsilon}(0)+\boldsymbol{M}_{\mathrm{ex}}(L)\cdot\boldsymbol{\epsilon}(L) +\int_{0}^{L}\{\boldsymbol{p}(s)\cdot\delta\boldsymbol{r}(s)+\boldsymbol{q}(s)\cdot\boldsymbol{\epsilon}(s)\}\,ds. \tag*{(11)}

The internal and external virtual works expressed with respect to δr\delta\mathbf{r} and ϵ\boldsymbol{\epsilon} are then equated as δEel=δWex\delta E_{\mathrm{el}}=\delta W_{\mathrm{ex}}. Collecting the terms associated with δr(s)\delta\mathbf{r}(s) and ϵ(s)\boldsymbol{\epsilon}(s) yields the (non-magnetic) Kirchhoff rod equations

F′(s)+p(s)=0,(12)\mathbf{F}'(s)+\mathbf{p}(s)=0, \tag*{(12)}
M′(s)+d^3(s)×F(s)+q(s)=0,(13)\mathbf{M}'(s)+\hat{\mathbf{d}}_{3}(s)\times\mathbf{F}(s)+\mathbf{q}(s)=0, \tag*{(13)}

with the internal moment vector M(s)=Mad^a=EI{(Ω1−Ω˚1)d^1+(Ω2−Ω˚2)d^2}+GJ(Ω3−Ω˚3)d^3\mathbf{M}(s)=M_{a}\hat{\mathbf{d}}_{a}=EI\{(\Omega_{1}-\mathring{\Omega}_{1})\hat{\mathbf{d}}_{1}+(\Omega_{2}-\mathring{\Omega}_{2})\hat{\mathbf{d}}_{2}\}+GJ(\Omega_{3}-\mathring{\Omega}_{3})\hat{\mathbf{d}}_{3}, and F(s)\mathbf{F}(s) is the internal force acting over the cross section at ss. Equations (12) and (13) are the equilibrium equations of forces and moments for the rod, respectively. The boundary conditions can be also derived from variational calculus [45] as

M(0)=−Mex(0),M(L)=Mex(L),F(0)=−Fex(0),F(L)=Fex(L).(14)\mathbf{M}(0)=-\mathbf{M}_{\mathrm{ex}}(0),\quad\mathbf{M}(L)=\mathbf{M}_{\mathrm{ex}}(L),\quad\mathbf{F}(0)=-\mathbf{F}_{\mathrm{ex}}(0),\quad\mathbf{F}(L)=\mathbf{F}_{\mathrm{ex}}(L). \tag*{(14)}

Next, we will tackle the derivation of the magnetic Kirchhoff rod equations based on the same procedure sketched above, while the elastic terms Eqs. (12) and (13) remain unchanged. The last task is to compute the variation of the elasto-magnetic potential, EmagE_{\mathrm{mag}}, with respect to δr(s)\delta\mathbf{r}(s) and ϵ(s)\boldsymbol{\epsilon}(s), δEmag\delta E_{\mathrm{mag}}, which is addressed next.

Derivation of the magnetic Kirchhoff rod equations using the principle of virtual work

We now set out to derive the magnetic Kirchhoff rod equations. First, we will compute the variation of the elasto-magnetic potential, δEmag\delta E_{\mathrm{mag}}. Equating the total internal and external virtual works as Eq. (9), we will obtain the magnetic Kirchhoff rod equations. Given that δEel\delta E_{\mathrm{el}}, δWex\delta W_{\mathrm{ex}}, and δEmag\delta E_{\mathrm{mag}} are additive in Eq. (9), the variation of the elasto-magnetic potential δEmag\delta E_{\mathrm{mag}} provides the elasto-magnetic force pmag\mathbf{p}_{\mathrm{mag}} and torque qmag\mathbf{q}_{\mathrm{mag}} explicitly, which will be readily added to Eqs. (12) and (13) to capture magnetic effects.

Before calculating δEmag\delta E_{\mathrm{mag}}, we compute the variation of the Cosserat frame basis d^a\hat{\mathbf{d}}_{a}, δd^a\delta\hat{\mathbf{d}}_{a}. Given that ϵ\boldsymbol{\epsilon} represents the infinitesimal rotation around d^a\hat{\mathbf{d}}_{a}, the variation of d^a\hat{\mathbf{d}}_{a} is δd^a=ϵ×d^a=−εabcδχbd^c\delta\hat{\mathbf{d}}_{a}=\boldsymbol{\epsilon}\times\hat{\mathbf{d}}_{a}=-\varepsilon_{abc}\delta\chi_{b}\hat{\mathbf{d}}_{c}, where εabc\varepsilon_{abc} is the Eddington epsilon. We aim to express δEmag\delta E_{\mathrm{mag}} with respect to δr\delta\mathbf{r} and ϵ\boldsymbol{\epsilon}. Note that δBr\delta\mathbf{B}^{r} and δd^a\delta\hat{\mathbf{d}}_{a} are zero, since they do not depend on the state variables. We first determine the variation of the reduced deformation gradient D\mathbf{D} for both δr\delta\mathbf{r} and ϵ\boldsymbol{\epsilon} as

δD=δ(d^a⊗d^a∘)=(δd^a)⊗d^a∘={(δr′⋅d^2)}Δ1−{(δr′⋅d^1)}Δ2−{(ϵ⋅d^3)}Δ3=−{(ϵ⋅d^a)}Δa,(15)\begin{aligned} \delta\mathbf{D} &=\delta(\hat{\mathbf{d}}_{a}\otimes\hat{\mathbf{d}}_{a}^{\circ})=(\delta\hat{\mathbf{d}}_{a})\otimes\hat{\mathbf{d}}_{a}^{\circ}\\ &=\{(\delta\mathbf{r}'\cdot\hat{\mathbf{d}}_{2})\}\boldsymbol{\Delta}_{1}-\{(\delta\mathbf{r}'\cdot\hat{\mathbf{d}}_{1})\}\boldsymbol{\Delta}_{2}-\{(\boldsymbol{\epsilon}\cdot\hat{\mathbf{d}}_{3})\}\boldsymbol{\Delta}_{3}\\ &=-\{(\boldsymbol{\epsilon}\cdot\hat{\mathbf{d}}_{a})\}\boldsymbol{\Delta}_{a}, \tag*{(15)} \end{aligned}

where we have defined the new tensor Δa\boldsymbol{\Delta}_{a} (a=1,2,3a=1,2,3) associated with the magnetic torque as

Δa(s)≡εabcd^b(s)⊗d^c∘(s).(16)\boldsymbol{\Delta}_{a}(s)\equiv\varepsilon_{abc}\hat{\mathbf{d}}_{b}(s)\otimes\hat{\mathbf{d}}_{c}^{\circ}(s). \tag*{(16)}

To obtain Eq. (15), we have used the orthogonality of the basis d^b×d^c=εabcd^a\hat{\mathbf{d}}_{b}\times\hat{\mathbf{d}}_{c}=\varepsilon_{abc}\hat{\mathbf{d}}_{a}. We also used the fact that, for an arbitrary vector a=a(s)\mathbf{a}=\mathbf{a}(s), δr′⋅a=δ(d^3)⋅a=(ϵ×d^3)⋅a=ϵ⋅(d^3×a)\delta\mathbf{r}'\cdot\mathbf{a}=\delta(\hat{\mathbf{d}}_{3})\cdot\mathbf{a}=(\boldsymbol{\epsilon}\times\hat{\mathbf{d}}_{3})\cdot\mathbf{a}=\boldsymbol{\epsilon}\cdot(\hat{\mathbf{d}}_{3}\times\mathbf{a}) holds, where we set a=d^1\mathbf{a}=\hat{\mathbf{d}}_{1} or a=d^2\mathbf{a}=\hat{\mathbf{d}}_{2}. Given that the variation of Br(r)\mathbf{B}^{r}(\mathbf{r}) acts as δBr=(∂Br/∂ri)δri≡(∇Br)δr\delta\mathbf{B}^{r}=(\partial\mathbf{B}^{r}/\partial r_{i})\delta r_{i}\equiv(\nabla\mathbf{B}^{r})\delta\mathbf{r}, we compute the variation of EmagE_{\mathrm{mag}}, δEmag\delta E_{\mathrm{mag}} as

δEmag=−∫0L[δr⋅pmag(s)+ϵ⋅qmag(s)]ds,(17)\delta E_{\mathrm{mag}}=-\int_{0}^{L}\left[\delta\mathbf{r}\cdot\mathbf{p}_{\mathrm{mag}}(s)+\boldsymbol{\epsilon}\cdot\mathbf{q}_{\mathrm{mag}}(s)\right]\mathrm{d}s, \tag*{(17)}
pmag(s)≡Aμ0(D(s)Br)⋅(∇Ba)=M⋅(∇Ba),(18)\mathbf{p}_{\mathrm{mag}}(s)\equiv\frac{A}{\mu_{0}}\left(\mathbf{D}(s)\mathbf{B}^{r}\right)\cdot\left(\nabla\mathbf{B}^{a}\right)=\mathbf{M}\cdot\left(\nabla\mathbf{B}^{a}\right), \tag*{(18)}
qmag(s)≡−Aμ0{(Δa(s)Br)⋅Ba}d^a(s)=M×Ba.(19)\mathbf{q}_{\mathrm{mag}}(s)\equiv-\frac{A}{\mu_{0}}\left\{\left(\Delta\mathbf{a}(s)\mathbf{B}^{r}\right)\cdot\mathbf{B}^{a}\right\}\hat{\mathbf{d}}_{a}(s)=\mathbf{M}\times\mathbf{B}^{a}. \tag*{(19)}

We have introduced pmag\mathbf{p}_{\mathrm{mag}} and qmag\mathbf{q}_{\mathrm{mag}} as the elasto-magnetic force and torque, respectively. When the applied field is homogeneous ∂Ba/∂ri=0\partial\mathbf{B}^{a}/\partial r_{i}=0, the magnetic torque qmag≠0\mathbf{q}_{\mathrm{mag}}\ne0 drives the deformation (i.e., pmag=0\mathbf{p}_{\mathrm{mag}}=0), while both pmag\mathbf{p}_{\mathrm{mag}} and qmag\mathbf{q}_{\mathrm{mag}} contribute to the deformation if the applied field is inhomogeneous.

Adding the elasto-magnetic force and torque terms from Eqs. (18) and (19) into the classic (non-magnetic) Kirchhoff rod equations (12) and (13), yields the force and moment balance equations:

F′(s)+p(s)+pmag(s)=0,(20)\mathbf{F}'(s)+\mathbf{p}(s)+\mathbf{p}_{\mathrm{mag}}(s)=0, \tag*{(20)}
M′(s)+d^3(s)×F(s)+q(s)+qmag(s)=0,(21)\mathbf{M}'(s)+\hat{\mathbf{d}}_{3}(s)\times\mathbf{F}(s)+\mathbf{q}(s)+\mathbf{q}_{\mathrm{mag}}(s)=0, \tag*{(21)}

where the magnetic force pmag\mathbf{p}_{\mathrm{mag}} and torque qmag\mathbf{q}_{\mathrm{mag}} were defined in Eqs. (18) and (19), respectively. Note that the boundary conditions specified in Eq. (14) remain unchanged in the elasto-magnetic case because they do not appear in Eq. (17). We refer to the set of Eqs. (20) and (21) as the magnetic Kirchhoff rod equations, where the coupling between elasticity and magnetism is captured by pmag\mathbf{p}_{\mathrm{mag}} and qmag\mathbf{q}_{\mathrm{mag}}.

Reduction of the magnetic Kirchhoff rod equations to planar deformations

In this subsection, for verification purposes, we show that the magnetic Kirchhoff equations derived above can reproduce existing results in the literature for the hard magnetic beams and elastica [31, 35], loaded under a constant or constant gradient magnetic fields. In Ref. [31], a theoretical model under constant field was derived within the continuum framework and validated against experiments. The governing equations for a more general description that can tackle either under constant or constant gradient fields were developed through dimensional reduction and validated experimentally in Ref. [35].

Consider a naturally straight rod clamped at s=0s=0 (r(0)=0\mathbf{r}(0)=0), while the other end at s=Ls=L is set to be force and momentum free. The rod is free from any external forces and torques other than magnetic fields, that is p=q=0\mathbf{p}=\mathbf{q}=0. We will derive the governing equation for the bending angle θ(s)\theta(s) in the xx-yy plane such that θ(s)=0\theta(s)=0 holds in the reference configuration. The constitutive law is M(s)=EIθ′(s)e^z\mathbf{M}(s)=EI\theta'(s)\hat{\mathbf{e}}_{z}. We will study two cases: a naturally straight hard magnetic rod loaded under a constant or constant-gradient fields, which have been studied in Ref. [31] and Ref. [35], respectively.

First, we consider the case of a rod under a constant magnetic field Ba=Bae^y\mathbf{B}^{a}=B^{a}\hat{\mathbf{e}}_{y}; see Figure 2(a). The rod is magnetized as Br=−Bre^y\mathbf{B}^{r}=-B^{r}\hat{\mathbf{e}}_{y} along the center-line (d^1=e^x\hat{\mathbf{d}}_{1}=\hat{\mathbf{e}}_{x}, d^2=e^z\hat{\mathbf{d}}_{2}=\hat{\mathbf{e}}_{z}, d^3=−e^y\hat{\mathbf{d}}_{3}=-\hat{\mathbf{e}}_{y}) with Br>0B^{r}>0 and Ba>0B^{a}>0. In the absence of the applied field, the rod is naturally straight along e^y\hat{\mathbf{e}}_{y} as d^3=−e^y\hat{\mathbf{d}}_{3}=-\hat{\mathbf{e}}_{y}. The Cosserat frame in the reference configuration is given by d^1=(cos⁡θ,sin⁡θ,0)\hat{\mathbf{d}}_{1}=(\cos\theta,\sin\theta,0), d^2=e^z\hat{\mathbf{d}}_{2}=\hat{\mathbf{e}}_{z} and d^3=(sin⁡θ,−cos⁡θ,0)\hat{\mathbf{d}}_{3}=(\sin\theta,-\cos\theta,0). Since the applied field is homogeneous, we have pmag=0\mathbf{p}_{\mathrm{mag}}=0 and F(s)=0\mathbf{F}(s)=0. To calculate qmag\mathbf{q}_{\mathrm{mag}} in Eq. (19), we first need to express M\boldsymbol{\mathcal{M}} with respect to θ\theta. From Eq. (7), M\boldsymbol{\mathcal{M}} is simplified as

Reduction of magnetic Kirchhoff rod equations to planar (2D) cases

Figure 2. Reduction of magnetic Kirchhoff rod equations to planar (2D) cases. (a) Schematics of a hard magnetic rod under constant field that deforms in 2D. The rod is magnetized along the tangent of the rod d^3\hat{\mathbf{d}}_{3}. (b) Schematics of a hard magnetic rod under constant gradient field. The rod is magnetized along d^1\hat{\mathbf{d}}_{1}.

M=ABrμ0d^3.(22)\boldsymbol{\mathcal{M}}=\frac{AB^{r}}{\mu_{0}}\hat{\mathbf{d}}_{3}. \tag*{(22)}

Substituting F(s)=0\mathbf{F}(s)=0 and Eq. (22) into the moment balance equation (Eq. (21)), we can reproduce the equation for the hard magnetic elastica [31] as

EIθ′′(s)+ABrBaμ0sin⁡θ(s)=0,(23)EI\theta^{\prime\prime}(s)+\frac{AB^{r}B^{a}}{\mu_{0}}\sin\theta(s)=0, \tag*{(23)}

with the boundary conditions θ(0)=0\theta(0)=0 and θ′(L)=0\theta^{\prime}(L)=0. As discussed in Ref. [31], Eq. (23) is mathematically equivalent to the governing equation of the clamped-free elastica. In this configuration, the magnetic rod is known to buckle at the critical applied field [31]

Bbenda∗=(π2)2μ0EIABrL2,(24)B_{\mathrm{bend}}^{a*}=\left(\frac{\pi}{2}\right)^{2}\frac{\mu_{0}EI}{AB^{r}L^{2}}, \tag*{(24)}

a result that will be used in § 6.2 to connect our 3D analytical results with the existing 2D results in literature [31].

Second, we consider the case of a rod in a constant gradient magnetic field: Ba(r)=bye^y\mathbf{B}^{a}(\mathbf{r})=by\hat{\mathbf{e}}_{y}, where bb is a constant. The rod is clamped along e^x\hat{\mathbf{e}}_{x} (y=0y=0) and magnetized as Br=Bre^y\mathbf{B}^{r}=B^{r}\hat{\mathbf{e}}_{y} (Figure 2(b)). The Cosserat frame in the reference configuration is given by d^˚1=e^y\mathring{\hat{\mathbf{d}}}_{1}=\hat{\mathbf{e}}_{y}, d^˚2=e^z\mathring{\hat{\mathbf{d}}}_{2}=\hat{\mathbf{e}}_{z}, and d^˚3=e^x\mathring{\hat{\mathbf{d}}}_{3}=\hat{\mathbf{e}}_{x}. We can write the Cosserat frame in the deformed configuration as d^1=(−sin⁡θ,cos⁡θ,0)\hat{\mathbf{d}}_{1}=(-\sin\theta,\cos\theta,0), d^2=e^z\hat{\mathbf{d}}_{2}=\hat{\mathbf{e}}_{z}, and d^3=(cos⁡θ,sin⁡θ,0)\hat{\mathbf{d}}_{3}=(\cos\theta,\sin\theta,0). Hence, the magnetization vector is simplified from Eq. (7) as

M=ABrμ0d^1.(25)\boldsymbol{\mathcal{M}}=\frac{AB^{r}}{\mu_{0}}\hat{\mathbf{d}}_{1}. \tag*{(25)}

Given that the rod is in a constant gradient field, pmag\mathbf{p}_{\mathrm{mag}} is non-zero. From Eqs. (18) and (25), we find pmag(s)=(ABrb/μ0)cos⁡θ(s) e^y\mathbf{p}_{\mathrm{mag}}(s)=(AB^{r}b/\mu_{0})\cos\theta(s)\,\hat{\mathbf{e}}_{y}. The force balance in Eq. (20) can now be integrated as

F(s)=F(L)+∫sLpmag(s′) ds′=ABrbμ0e^y∫sLcos⁡θ(s′) ds′.(26)\begin{aligned} \mathbf{F}(s) &= \mathbf{F}(L)+\int_{s}^{L}\mathbf{p}_{\mathrm{mag}}(s')\,\mathrm{d}s'\\ &=\frac{AB^{r}b}{\mu_{0}}\hat{\mathbf{e}}_{y}\int_{s}^{L}\cos\theta(s')\,\mathrm{d}s'. \tag*{(26)} \end{aligned}

Plugging this result into the moment balance in Eq. (21), we obtain the following integrodifferential equation

EId2θds2−ABrbμ0{y(s)sin⁡θ(s)−cos⁡θ(s)∫sLcos⁡θ(s′) ds′}=0,(27)EI\frac{\mathrm{d}^{2}\theta}{\mathrm{d}s^{2}}-\frac{AB^{r}b}{\mu_{0}}\left\{y(s)\sin\theta(s)-\cos\theta(s)\int_{s}^{L}\cos\theta(s')\,\mathrm{d}s'\right\}=0, \tag*{(27)}

whose second and third terms originate from magnetic torque and force, respectively. Equation (27) has been recently derived and validated experimentally in Refs. [31, 35], verifying our reduced framework presented in this section.

We note that the inhomogeneous applied field can be interpreted as a distributed load on the magnetic rod. Indeed, in the limit of small bending angle θ≪1\theta\ll1, Eq. (27) is identical to that of an elastica under gravity [41]. By analogy with the gravitational case, we can introduce the magneto-bending length ℓm\ell_m as

ℓm≡(μ0EIABrb)1/3,(28)\ell_m\equiv\left(\frac{\mu_{0}EI}{AB^{r}b}\right)^{1/3}, \tag*{(28)}

which can be regarded as a (persistence) length quantifying the relative importance of the magnetic and elastic effects. When λm≡L/ℓm≪1\lambda_m\equiv L/\ell_m\ll1, the gradient of the magnetic field is negligible, while in the case of λm≫1\lambda_m\gg1, the magnetic gradient plays an important role in the bending of the magnetic rod.

We have shown that the magnetic Kirchhoff rod equations derived in § 3.3 are able to reproduce existing results reported in Refs. [31, 35] for the planar (2D) deformation of the magnetic rod. The specialization to the 2D cases verifies our theory, even if only partially. In the following sections, we will combine discrete simulation and precision experiments to study non-planar (3D) deformations of the rod, towards the validation of our theoretical framework.

Discrete simulations of magneto-elastic rods

Although our magnetic Kirchhoff rod model reproduces the governing equation of planar hard magnetic beams and elastica, computing the 3D nonlinear geometric deformation requires, in general, a numerical approach. Specifically, we discretize the rod-centerline as a set of connected particles via straight rigid segments [48, 49, 50, 51, 52, 53, 54, 55, 56, 57]. This discrete method is widely used to simulate the large deformation of rod-like structures such as nano-springs [49], DNA [48], helical bacteria [50], flagella [51, 52], human hairs [53], and gift-wrapping ribbons [54, 55, 56, 57]. Within this computational framework, the rod centerline is regarded as the chain of straight segments of natural length ℓ0\ell_0 and an equal mass m=ρAL/Nm=\rho AL/N. At each node i=1,2,…,Ni=1,2,\ldots,N, we assigned a discrete Cosserat frame basis vectors (d^1,i,d^2,i,d^3,i)(\hat{\mathbf{d}}_{1,i},\hat{\mathbf{d}}_{2,i},\hat{\mathbf{d}}_{3,i}), corresponding to d^a\hat{\mathbf{d}}_{a} in the continuum model. The elastic bending and twist deformations are represented by the Euler angles between adjacent discretized Cosserat frame basis vectors [48, 49, 50, 51, 52, 53, 54, 55, 56, 57].

In the present section, we detail the simulation method, with a particular focus on the discrete versions for the elasto-magnetic forces, P\mathcal{P}, and torques, Q\mathcal{Q}, in § 4.1. The equations of motion for the discrete segments follow in § 4.2.

Elasto-magnetic force and torque in the discrete simulation

To compute the nonlinear geometric deformation of our rod-like magnetic structures, we adopt dynamic simulations, instead of solving the static equilibrium equations. Static analysis packages for ODEs (e.g., AUTO [60]) might have been able to compute the deformation of our MRE rods. However, it is not trivial to implement the elasto-magnetic interactions into such packages due to the geometric non-linearity of our MRE rods. To facilitate numerical convergence to the mechanical equilibrium, we perform the dynamic simulation by including the inertial terms in Eqs. (20) and (21). Specifically, we will derive the equations of motion for r\mathbf{r} and the twist angle χ3\chi_3 (see § 3.2 for the definition of χ3\chi_3). We recall that the expressions of the elasto-magnetic force pmag\mathbf{p}_{\mathrm{mag}} and torque qmag\mathbf{q}_{\mathrm{mag}}, provided in Eqs. (17)–(19), have been derived from the continuum framework, through the variation δEmag\delta E_{\mathrm{mag}} with respect to the set of (δr,ϵ)(\delta\mathbf{r},\boldsymbol{\epsilon}). The set of infinitesimal bending and twist angles were represented by ϵ=δχad^a\boldsymbol{\epsilon}=\delta\chi_a\hat{\mathbf{d}}_a as in § 3.2. In the discrete simulation framework [48, 49, 50, 51, 52, 53, 54, 55, 56, 57], it is sufficient to derive the equation of motion of the centerline position and the twist angle (r,χ3)(\mathbf{r},\chi_3) alone. In other words, the remaining bending angles χ1\chi_1 and χ2\chi_2 are determined from the centerline positions r\mathbf{r}, as we explain below [45]. To include the elasto-magnetic force and torque in the equations of motion for (r,χ3)(\mathbf{r},\chi_3), we will introduce the discrete version of (pmag,qmag)(\mathbf{p}_{\mathrm{mag}},\mathbf{q}_{\mathrm{mag}}), as (P,Q)(\mathcal{P},\mathcal{Q}), by rewriting δEmag\delta E_{\mathrm{mag}} (Eq. (17)) with respect to the set of (r,χ3)(\mathbf{r},\chi_3). The discrete version of the elasto-magnetic torque Q\mathcal{Q} (a scalar) will correspond to the elasto-magnetic torque around d^3\hat{\mathbf{d}}_3, while the remaining components of qmag\mathbf{q}_{\mathrm{mag}} (i.e., qmag1q_{\mathrm{mag}1} and qmag2q_{\mathrm{mag}2}) will be included in the equation of motion for r\mathbf{r}, thereby defining P\mathcal{P}, as we detail below.

Expressing δχ1\delta\chi_1 and δχ2\delta\chi_2 from Eq. (17) with the aid of δχ1=−(δr)′⋅d^2\delta\chi_1=-(\delta\mathbf{r})'\cdot\hat{\mathbf{d}}_2 and δχ2=(δr)′⋅d^1\delta\chi_2=(\delta\mathbf{r})'\cdot\hat{\mathbf{d}}_1, we find the following equation after the partial integration of (δr)′(\delta\mathbf{r})':

δEmag=−∫0L[δr(s)⋅pmag(s)+ϵ(s)⋅qmag(s)] ds=−∫0L[δr⋅pmag+δχ1qmag1+δχ2qmag2+δχ3qmag3] ds=−∫0L[δr⋅pmag+{−(δr)′⋅d^2}qmag1+{(δr)′⋅d^1}qmag2+δχ3qmag3] ds=−∫0L[δr⋅P+δχ3Q] ds+[δr⋅{qmag1d^2−qmag2d^1}]0L,(29)\begin{aligned} \delta E_{\mathrm{mag}}={}&-\int_0^L\left[\delta\mathbf{r}(s)\cdot\mathbf{p}_{\mathrm{mag}}(s)+\boldsymbol{\epsilon}(s)\cdot\mathbf{q}_{\mathrm{mag}}(s)\right]\,ds\\ ={}&-\int_0^L\left[\delta\mathbf{r}\cdot\mathbf{p}_{\mathrm{mag}}+\delta\chi_1q_{\mathrm{mag}1}+\delta\chi_2q_{\mathrm{mag}2}+\delta\chi_3q_{\mathrm{mag}3}\right]\,ds\\ ={}&-\int_0^L\left[\delta\mathbf{r}\cdot\mathbf{p}_{\mathrm{mag}}+\{-(\delta\mathbf{r})'\cdot\hat{\mathbf{d}}_2\}q_{\mathrm{mag}1}+\{(\delta\mathbf{r})'\cdot\hat{\mathbf{d}}_1\}q_{\mathrm{mag}2}+\delta\chi_3q_{\mathrm{mag}3}\right]\,ds\\ ={}&-\int_0^L\left[\delta\mathbf{r}\cdot\mathcal{P}+\delta\chi_3\mathcal{Q}\right]\,ds+\left[\delta\mathbf{r}\cdot\{q_{\mathrm{mag}1}\hat{\mathbf{d}}_2-q_{\mathrm{mag}2}\hat{\mathbf{d}}_1\}\right]_0^L, \tag*{(29)} \end{aligned}

where we introduce the a(=1,2,3)a(=1,2,3)-th component of the magnetic torque qmaga=qmag⋅d^aq_{\mathrm{mag}a}=\mathbf{q}_{\mathrm{mag}}\cdot\hat{\mathbf{d}}_a. Here, we define the discrete version of the elasto magnetic force and torque as

P=pmag+(qmag1d^2−qmag2d^1)′,(30)\mathcal{P}=\mathbf{p}_{\mathrm{mag}}+\left(q_{\mathrm{mag}1}\hat{\mathbf{d}}_2-q_{\mathrm{mag}2}\hat{\mathbf{d}}_1\right)', \tag*{(30)}
Q=qmag3,(31)\mathcal{Q}=q_{\mathrm{mag}3}, \tag*{(31)}

which will be used to formulate the discrete version of the equations of motion. Note that qmag1q_{\mathrm{mag}1} and qmag2q_{\mathrm{mag}2} are now included in P\mathcal{P}, appropriately. In the next subsection, we incorporate the elasto-magnetic force and torque through P\mathcal{P} and Q\mathcal{Q} into the equations of motion for the centerline position and the twist angle (r,χ3)(\mathbf{r},\chi_3).

Equations of motion used in the discrete simulations

In this subsection, we derive the discrete version of the equation of motion for the MRE rod by adding (P,Q)(\mathcal{P},\mathcal{Q}) into the equation of motion for the centerline position and the twist angle (r,χ3)(\boldsymbol{r},\chi_3). We start from the magnetic Kirchhoff equations that we derived earlier, with the sole addition of the inertial terms

ρA∂v∂t=∂F∂s+p+P,(32)\rho A \frac{\partial\boldsymbol{v}}{\partial t} = \frac{\partial\boldsymbol{F}}{\partial s} + \boldsymbol{p} + \mathcal{P}, \tag*{(32)}
∂L∂t=∂M∂s+d^3×F+q+Qd^3,(33)\frac{\partial\boldsymbol{L}}{\partial t} = \frac{\partial\boldsymbol{M}}{\partial s} + \hat{\boldsymbol{d}}_3 \times\boldsymbol{F} + \boldsymbol{q} + \mathcal{Q}\hat{\boldsymbol{d}}_3, \tag*{(33)}

where (pmag,qmag)(\boldsymbol{p}_{\mathrm{mag}},\boldsymbol{q}_{\mathrm{mag}}) are replaced by (P,Qd^3)(\mathcal{P},\mathcal{Q}\hat{\boldsymbol{d}}_3), as derived in the previous section. The left hand sides of Eq. (33) represent the inertia terms. Here, we define the velocity of the centerline v(s,t)=∂r/∂t\boldsymbol{v}(s,t)=\partial\boldsymbol{r}/\partial t, the angular momentum per unit length L=Iaωad^a\boldsymbol{L}=I_a\omega_a\hat{\boldsymbol{d}}_a, and the angular velocity vector ω(s,t)=ωad^a\boldsymbol{\omega}(s,t)=\omega_a\hat{\boldsymbol{d}}_a, with the mass density per unit volume ρ\rho and the principal moments of inertia of the cross section I1=I2=II_1=I_2=I and I3=I1+I2=2II_3=I_1+I_2=2I [61]. The angular velocity vector ω\boldsymbol{\omega} describes the rotation rates of d^a\hat{\boldsymbol{d}}_a in time as

∂d^a∂t=ω×d^a.(34)\frac{\partial\hat{\boldsymbol{d}}_a}{\partial t} = \boldsymbol{\omega} \times\hat{\boldsymbol{d}}_a. \tag*{(34)}

To ensure that the rod relaxes to the mechanical equilibrium, we include numerical drag forces and torque acting on the rod-centerline per unit length through p=−γtv\boldsymbol{p}=-\gamma_t\boldsymbol{v} and q=−γrω3d^3\boldsymbol{q}=-\gamma_r\omega_3\hat{\boldsymbol{d}}_3. We readily derive the dynamic equations of twist around d^3\hat{\boldsymbol{d}}_3 by taking the inner product between the moment balance equation and d^3\hat{\boldsymbol{d}}_3. Then, introducing the internal elastic forces f\boldsymbol{f} and the axial torque TT per unit length as

f=∂F∂s,T≡∂M3∂s+M2Ω1−M1Ω2,(35)\boldsymbol{f}=\frac{\partial\boldsymbol{F}}{\partial s}, \qquad T\equiv\frac{\partial M_3}{\partial s}+M_2\Omega_1-M_1\Omega_2, \tag*{(35)}

we rewrite Eqs. (32) and (33) as

ρA∂2r∂t2=f+p+P,(36)\rho A\frac{\partial^2\boldsymbol{r}}{\partial t^2}=\boldsymbol{f}+\boldsymbol{p}+\mathcal{P}, \tag*{(36)}
I3∂2χ3∂t2=T−γr∂χ3∂t+Q,(37)I_3\frac{\partial^2\chi_3}{\partial t^2}=T-\gamma_r\frac{\partial\chi_3}{\partial t}+\mathcal{Q}, \tag*{(37)}

respectively. We solve Eq. (37) for the discrete versions of the position vector ri\boldsymbol{r}_i and the twist angle χ3,i\chi_{3,i}. Choosing the units of mass, length, and force as mℓ0m\ell_0, ℓ0\ell_0 and EI/ℓ02EI/\ell_0^2, respectively, we non-dimensionalize Eq. (37) as

∂2r~i∂t~2=f~i+p~i+P~i,(38)\frac{\partial^2\tilde{\boldsymbol{r}}_i}{\partial\tilde{t}^2}=\tilde{\boldsymbol{f}}_i+\tilde{\boldsymbol{p}}_i+\tilde{\mathcal{P}}_i, \tag*{(38)}
I~3∂2χ3,i∂t~2=T~i−γ~r∂χ3,i∂t~+Q~i,(39)\tilde{I}_3\frac{\partial^2\chi_{3,i}}{\partial\tilde{t}^2}=\tilde{T}_i-\tilde{\gamma}_r\frac{\partial\chi_{3,i}}{\partial\tilde{t}}+\tilde{\mathcal{Q}}_i, \tag*{(39)}

where quantities with (⋅~)(\tilde{\cdot}) represent the corresponding dimensionless variables.

The numerical scheme for calculating the elastic force fi\boldsymbol{f}_i and axial torque TiT_i at each node ii follows that of Refs. [48, 49, 50, 51, 52]. In short, we employ the Euler angle representation to describe the configuration of the discrete director frames at each point. The Darboux vector components in the discretized model, Ωa,i\Omega_{a,i} (a=1,2,3a = 1, 2, 3), and the corresponding discretized elastic energy can be expressed in terms of the three Euler angles. The variation of the elastic energy EelE_{\mathrm{el}} is related to the variations of ri\mathbf{r}_i and χ3,i\chi_{3,i} through the variations of the corresponding Euler angles, from which we find the bending and twisting forces acting on each node of the chain [48, 49, 50, 51, 52]. All values of the dimensionless parameters are taken from the experiments, which will be presented in § 5, such that our simulations do not contain any free parameters. We performed the simulation for a sufficiently long time such that the rod relaxes to the mechanical equilibrium (typically 107−10910^7 - 10^9 discrete time steps), where the number of nodes lies within the range 40≤N≤12040 \le N \le120. We use the two-step Adams-Bashforth method [62] to numerically integrate the re-scaled dynamical equations in Eq. (39) with non-dimensional time steps, typically ranging between 10−3−10−110^{-3} - 10^{-1}, to ensure sufficient numerical accuracy. The director frames at each node are also updated at each time step, and the corresponding Euler angles are calculated for a new configuration.

Experimental fabrication, apparatus, and protocols

We perform three different sets of experiments to validate our theory presented in § 3: (i) a naturally straight rod under a constant field, (ii) a helical rod under a constant field, and (iii) a helical rod under a constant gradient field. We chose these three configurations because previous works [31, 32, 35, 41] were limited to the case of straight beams, without twist deformations. For case (i), we still consider a straight rod but also include twist deformation. Through case (ii), we can individually validate the elasto-magnetic torque qmagq_{\mathrm{mag}}. In case (iii), both qmagq_{\mathrm{mag}} and pmagp_{\mathrm{mag}} can be validated. In this section, we present the details of our experiments, whose results will be provided in § 6. To perform the experiments, in § 5.1, we detail the procedure that we developed to manufacture our MRE rods. In § 5.2, we present the design of the magnetic coils used to generate the magnetic field. The detailed protocols for the cases (i)-(iii) follow in § 5.3.

Fabrication of hard magnetic rods

Our rods were made of a magnetorheological elastomer (MRE), a composite of hard-magnetic NdPrFeB particles and vinylpolysiloxane (VPS) polymer (Young’s modulus Evps=1.16±0.03E_{\mathrm{vps}} = 1.16 \pm0.03 MPa and mass density ρvps=1.17±0.26\rho_{\mathrm{vps}} = 1.17 \pm0.26 g/cm3^3). The fabrication of the MRE rods involved the following steps. First, the non-magnetized NdPrFeB particles (average size of 5 μm5\,\mu\mathrm{m}, mass density ρmag=7.61\rho_{\mathrm{mag}} = 7.61 g/cm3^3 MQP-F-15-7-20065-089, Magnequench) of weight mmagm_{\mathrm{mag}} were mixed with the VPS base liquid solution (see tbl:32, Zhermack) of weight mbasem_{\mathrm{base}} using a centrifugal mixer (ARE-250, Thinky Corporation). Secondly, the VPS catalyst (VPS catalyst, weight mcat=mbasem_{\mathrm{cat}} = m_{\mathrm{base}}) was added to the mixed solution with a ratio of 1:1 in weight to the VPS base. The solution of total weight mtot=mmag+mbase+mcatm_{\mathrm{tot}} = m_{\mathrm{mag}} + m_{\mathrm{base}} + m_{\mathrm{cat}} was then mixed using the centrifugal mixer for 40 s at 2000 rpm (clockwise), and another 20 s at 2200 rpm (counterclockwise). We further degassed the solution in a vacuum chamber (absolute pressure below 8 mbar) to remove any air bubbles that could otherwise compromise the homogeneity of the MRE. The final solution was injected into the molds (detailed in the next paragraph) using a syringe to cast either straight or helical rods. We varied the mass concentration ratio for NdPrFeB particles in the range c=mmag/mtot=10−30%c = m_{\mathrm{mag}}/m_{\mathrm{tot}} = 10 - 30\%.

The schematic and a photograph of a mold used to fabricate a straight rod are shown in Figure 3 (a) and (b), respectively. We injected the mixed solution into a straight acrylic tube (Plexiglas XT tubes incolore 0A070, Röhm, Switzerland, inner diameter 2 or 4 mm) threaded into two coaxial holes in acrylic plates. One of the inlets of the tube was designed to have a convex shape pointing to the direction of the residual flux density. The convex inlet is utilized to align the magnetization vector Br\mathbf{B}^{r} precisely.

Fabrication procedure for straight and helical MRE rods

Figure 3. Fabrication procedure to manufacture straight and helical rods made out of a MRE. (a) Schematic diagram of the mold for a straight rod. A straight acrylic tube (green) is threaded into two parallel supporting plates. (b) Photograph of the mold used to produce a straight rod. (c) Schematic diagram of the mold for a helical rod. A rigid cylinder with a helical groove (inner mold) is 3D printed. (d) Photograph of the mold for a helical rods. (e) The mold with the cured elastomer is placed into an impulse magnetizer such that Br\mathbf{B}^{r} in (a) and (c) is aligned with the axis of the coil. Schematics of a magnetized (f) straight and (g) helical rod. The local magnetization vectors are depicted by the embedded “magnets,” where the red and blue ends correspond to the north and south poles, respectively.

To fabricate helical rods, we prepared a rigid cylinder of diameter 20 mm with a helical half-piped groove (diameter 4.5 mm) using a 3D printer (Printer: Form 2, Formlabs, Material: Clear Resin (RS-F2-GPCL-04)). The schematic and the photograph of the mold with a helical groove are shown in Figure 3 (c) and (d), respectively. A flexible tube (diameter 4 mm, Misumi) is placed along the groove, injecting the mixed solution into it. The inner mold and the tube were then positioned inside a 3D printed (outer) cylinder (inner diameter 12 mm and outer diameter 16 mm). The central axis of the helical groove and that of the outer cylinder were set co-axially. The 3D printed molds allowed us to control the pitch angle ψ\psi of a helical MRE rod up to ψ≲1.49\psi\lesssim1.49 rad. For larger values of ψ\psi, we prepared the acrylic cylinder (without a groove) and tightly spooled a flexible tube around it with a pitch angle of ψ=1.51\psi= 1.51 rad.

After injection of the mixed solution into either the straight or helical molds, the curing of the polymer mixture occurred in approximately 20 min, at room temperature. The (projected) radius of the helical centerline was chosen to be R=10R = 10 mm to minimize the deformation of the tube, allowing us to have a cross-section of the helix close to circular, which could otherwise deviate from the circular shape for tighter helices due to the Brazier instability [63].

Before demolding, the rods were magnetized permanently by saturating the NdPrFeB particles in the MRE using an applied uniaxial magnetic field (4.4 or 2.5 T) generated by an impulse magnetizer (IM-K-010020-A Magnet-Physik Dr. Steingroever GmbH), as shown in Figure 3 (e). The directions of Br\mathbf{B}^{r} for either a straight or helical rods were set normal to the cross section or parallel to the central axis of a helix, respectively. Schematic diagrams for the magnetized straight and helical rods are shown in Figure 3 (f) and (Figure 3g), respectively. The fabricated MRE rods possessed the residual flux densities of ∣Br∣=0.90cv|\mathbf{B}^{r}| = 0.90c_{\mathrm{v}} T (§ 6.1 and § 6.3) and ∣Br∣=0.86cv|\mathbf{B}^{r}| = 0.86c_{\mathrm{v}} T (§ 6.2). The Young’s modulus was E=Evps/(1−cv1/2)E = E_{\mathrm{vps}}/(1-c_{\mathrm{v}}^{1/2}) with the volume concentration of the particles equal to cv=(1+(ρmag/ρvps)(mtot/mmag−1))−1c_{\mathrm{v}} = \left(1+\left(\rho_{\mathrm{mag}}/\rho_{\mathrm{vps}}\right)(m_{\mathrm{tot}}/m_{\mathrm{mag}}-1)\right)^{-1} [64]. After magnetization, the rods were demolded and cut to the desired length LL (within an experimental uncertainty of ±1\pm1 mm).

Experimental apparatus

During the experimental tests, the samples are loaded magnetically by placing them in between a set of two coaxial coils, which induce a steady axial symmetric magnetic flux Ba(r,z)\mathbf{B}^{a}(r,z) [35]. Each coil is manufactured by winding an aluminum circular spool with an insulated magnet wire (Repelec Moteurs S.A.). The magnet wire (enameled wire, Isomet AG) has a circular cross-section of diameter 1.32 mm for the copper core and thickness 0.102 mm for the outer insulation layer. The final dimensions of the coil are 86 mm in inner diameter, 152 mm in outer diameter, and 43 mm (33 mm for wires and 10 mm for the spool) in height; the mean radius of the coil R\mathcal{R} is R=59.5\mathcal{R} = 59.5 mm. The coils are powered by a DC power supply providing a maximum current to power ratio of 25 A/1.5kW (EA-PSI 9200-25T, EA-Elektro-Automatik GmbH).

The two identical coils are set co-axially along e^z\hat{\mathbf{e}}_{z} and their current can flow in either the same or opposite directions. To realize a constant (homogeneous) field, the center-to-center axial distance is set to be 59.5 mm (=R=\mathcal{R}), and the current in the coils is set to flow in the same direction; this is known as the Helmholtz coil configuration. In the central region between the coils, the field generated by the Helmholtz coil is

Ba=Bae^z.(40)\mathbf{B}^{a}=B^{a}\hat{\mathbf{e}}_{z}. \tag*{(40)}

In contrast, to realize the constant gradient field, the center-to-center axial distance is set to be 103 mm (=3R=\sqrt{3}\mathcal{R}), and the current in the coils is set to flow in opposite directions. This configuration is known as the Maxwell coil, where the magnetic field induced near the coil is

Ba(r,z)=−b2re^r+bze^z.(41)\mathbf{B}^{a}(r,z)=-\frac{b}{2}r\hat{\mathbf{e}}_{r}+bz\hat{\mathbf{e}}_{z}. \tag*{(41)}

In this Maxwell coil configuration, the gradient of the magnetic field is b=(∂zBa)⋅e^zb=(\partial_{z}\mathbf{B}^{a})\cdot\hat{\mathbf{e}}_{z}, i.e. constant. From the Gauss law of magnetism, ∇⋅Ba=0\nabla\cdot\mathbf{B}^{a}=0, there are gradient components in both the radial, ere_{r}, and the eze_{z} directions.

We characterized the relationship between the flux density and the current in the coils, IcI_{c}, by using a Teslameter (FH 55, Magnet-Physik Dr. Steingroever GmbH) and the results are plotted in Figure 4. The fields described by Eqs. (40) and (41) are realized in the region z,r≲0.6Rz,r\lesssim0.6\mathcal{R}. For the Helmholtz configuration, the constant field is induced near the center, for several values of IcI_{c}. In Figure 4 (a), we plot the profiles of the measured field Ba\mathbf{B}^{a} divided by IcI_{c}, which collapse onto a single curve. This collapse confirms that Ba\mathbf{B}^{a} is proportional to IaI_{a}. By fitting the constant BaB^{a} near the center, we find BaB^{a} and IaI_{a} as Ba/Ic=8.9B_{a}/I_{c}=8.9 mT/A. For the Maxwell configuration, a constant gradient field is generated near the center of the coils (Figure 4 (b) and (c)). We measured the profiles of the induced field both along with eze_{z} and ere_{r} for different values of IcI_{c}. The data again collapse on a single curve as Ba/IcB^{a}/I_{c}, which implies that the gradient of the field bb is linearly proportional to IcI_{c}. By fitting the field along eze_{z}, we find b/Ic=0.133b/I_{c}=0.133 T/(A · m). From Eq. (41), the gradient along ere_{r} is half of that along eze_{z}. In Figure 4 (c), we plot the slope of b/2b/2 as the dashed line, which is in an excellent agreement with the measurement. This experimental characterization confirms that our coils accurately produce the fields according to Eqs. (40) and (41) when set in either the Helmholtz or the Maxwell configurations, respectively.

Experimental characterization of the field generated by the magnetic coils in the Helmholtz and Maxwell configurations

Figure 4. Experimental characterization of the field generated by the magnetic coils in the (a) Helmholtz (constant field) and (b, c) Maxwell (constant gradient field) configurations. (a) Profile of the (rescaled) applied field along x=y=0x=y=0. Profile of the applied field along (b) x=y=0x=y=0 and (c) y=z=0y=z=0. The dashed lines in the plot correspond to the (a) constant and (b) (c) linear fits within the shaded regions where the experiments were performed.

Experimental protocols

Having described the experimental apparatus above, we proceed by presenting the protocols that we followed during the experiments. We placed the samples (prepared through the procedure detailed in § 5.1) between the coils (whose field was characterized in the previous section). First, we will study the deformation of a straight or curved rod under the constant field (Helmholtz coil) followed by the study of the deformation of a helix under the constant gradient field (Maxwell coil), as shown in Figure 5 (a)-(Figure 5d). A photograph of our apparatus is shown in Figure 5 (a). All the experimental results for the corresponding setups will be presented in § 6. The protocols for each experiment are detailed next.

Experimental apparatus and setups

Figure 5. (a) Photograph of the full experimental apparatus. A single power supply generates the electric current provided to the two coils. A digital camera set underneath the coils captures the deformation of the rod. (b) Experimental setup used for the twist instability of a straight rod (§ 6.1). The straight rod magnetized along −e^z-\hat{e}_z is clamped along e^x\hat{e}_x. The applied field Ba\mathbf{B}^{a} is set as shown in Eq. (40). (c) The schematic of the experimental setup for a helix under constant field (right panel, See § 6.2), with the uniformly helical magnetic rod (left panel). The helix magnetized along e^z\hat{e}_z is clamped such that its central axis is parallel to that of the coils. The magnetic helix is immersed in a glycerol bath. (d) Photograph of a magnetic helical rod (in the glycerol bath) under constant gradient field (§ 6.3).

In § 6.1, we will present results on the deformation of a straight rod under constant field. For these experiments, our MRE rods were magnetized along −e^z-\hat{\mathbf e}_z, and a constant magnetic field was applied along e^z\hat{\mathbf e}_z; see Figure 5 (b). An acrylic clamp, engraved to have the convex shape at one of the ends of the rod, hanged the rod from above. This clamp was also designed such that its width matched the gap between the coils (13.5 mm). We varied the total length of the rod in the ranges L=40−57L=40-57 mm, the mass concentration ratio of NdPrFeB particles is c=10,20,30%c=10, 20, 30\%, and the rod diameter was d=3,5d=3, 5 mm. We measured the twist deformation of the rod (detailed in the next section) using a digital camera, placed underneath the setup and directed upwards. The measured twisting angles were averaged over 4 or 5 samples.

In § 6.2, we will study the deformation of a helical rod under the constant field. For these experiments, the helix was clamped vertically by a 3D-printed rigid cylinder such that its central axis was parallel to the coil axis (e^z\hat{\mathbf e}_z). The constant magnetic field (Eq. (40)) was applied (anti-parallel to Br\mathbf B^r) to the helical rod magnetized along e^z\hat{\mathbf e}_z, as shown in Figure 5 (c). The experiments were performed in a glycerol bath (85%, 1.23 g/cm3^{3}, Sigma-Aldrich), which density-matched the rods, thereby minimizing the effect of gravity. We performed experiments for the MRE rod of diameter 2 mm, pitch angle ψ=1.28\psi= 1.28 rad, and radius of curvature R=10R = 10 mm. The total length LL and the concentration ratio cc were varied in the ranges L=40−120L = 40 - 120 mm and c=10,20%c = 10, 20\%, respectively.

For the results in § 6.3 and 7, we applied a constant gradient field (according to Eq. (41)) to a helical rod, whose central axis in the reference configuration was chosen to be aligned with the axis of the coils (zz-axis, perpendicular to gravity). As shown in Figure 5 (d), we clamped the helix and perform the experiment in a glycerol bath. During the experiments, we imaged the deformed shape of the helix using a digital camera. In § 6.3, we measured the displacement of the free-end as a function of the applied field. The total length LL and the concentration ratio cc were varied as L=74,81,103L = 74, 81, 103 mm and c=10,20%c = 10, 20\%, respectively, while the diameter d=2d = 2 mm, the radius of natural curvature R=10R = 10 mm and the pitch angle ψ=1.51\psi= 1.51 rad were fixed. In § 7, the deformation of the magnetic helix were studied, varying the pitch angle ψ\psi and total length LL in the range of ψ=1.26−1.49\psi= 1.26 - 1.49 rad and L=65−140L = 65 - 140 mm, while d=2d = 2 mm, R=10R = 10 mm, and c=20%c = 20\% were kept fixed throughout § 7.

Validation of the theory of hard magnetic rods vs experiments

In this section, we will use precision experiments on specific test configurations to perform a detailed validation of the theoretical framework introduced in § 3.3 for hard MRE rods. In § 6.1 and 6.2, the twist instabilities for a straight rod and a helix are investigated under a constant field. Given that pmag=0p_{\mathrm{mag}} = 0 under the constant field, we can show that the elasto-magnetic torque qmagq_{\mathrm{mag}} in both naturally straight and curved configurations captures the corresponding experimental results in excellent agreement. Then, in § 6.3, we study the deformation of a helix under a constant gradient field to test the validity of both pmagp_{\mathrm{mag}} and qmagq_{\mathrm{mag}}.

Twist instability of a straight magnetic rod under magnetic loading

Consider a straight rod aligned with e^x\hat{\boldsymbol{e}}_x (clamped at s=0s = 0), that is magnetized normal to the tangent, i.e., Br=Bre^z\boldsymbol{B}^r = B^r\hat{\boldsymbol{e}}_z. We apply the constant flux density as Ba=−Bae^z\boldsymbol{B}^a = -B^a\hat{\boldsymbol{e}}_z (see Figure 6 (a)-(d)). Figures 6 (a) and (c) are photographs from the experiments of the rod cross-sections at their free-ends, parallel with yy-zz planes. Three-dimensional representations of the straight MRE rods obtained from the simulations are shown in Figs. 6 (b) and (d). The rod remains straight when the magnitude of the applied field BaB^a is sufficiently small enough (Figure 6 (a) and (b)), whereas the rod twists, with the centerline still straight, above the critical flux density BtwistaB^a_{\mathrm{twist}} (Figure 6 (c) and (d)). In the experiments, we measured the twist angle ϕL\phi_L at s=Ls = L by tracking the rotation angle of a nitinol rod embedded near the free-end perpendicular to the centerline (as shown by n^(L)\hat{\boldsymbol{n}}(L) in Figs. 6 (a) and (c)). In the simulation, we also measured ϕL\phi_L by computing the rotation angle for one of the basis vectors of the Cosserat frame d^1\hat{\boldsymbol{d}}_1. In Figure 6 (e), we superpose the experimental and numerical results of ϕL\phi_L as a function of Ba/Btwista∗B^a/B^{a*}_{\mathrm{twist}}. The error bars in the experimental data correspond to the uncertainty in the setting of the angle of the clamp. Experimental and numerical results are in excellent agreement with each other across the full parameter range in Ba/Btwista∗B^a/B^{a*}_{\mathrm{twist}}.

Twist instability of straight MRE rods

Figure 6. Twist instability of straight MRE rods. (a) Experimental photograph of the cross section at s=Ls=L with Ba<Btwista∗B^{a}<B_{\mathrm{twist}}^{a*}. The symbols n^(L)\hat{\boldsymbol{n}}(L) and n^∘(L)\hat{\boldsymbol{n}}^{\circ}(L) represent the direction of the nitinol rod in the deformed and reference configuration, respectively. The corresponding simulation snapshot is shown in (b). (c) Experimental photograph of the cross section at s=Ls=L with Ba>Btwista∗B^{a}>B_{\mathrm{twist}}^{a*}. The corresponding simulation snapshot is shown in (d). (e) ϕL\phi_L as a function of Ba/Btwista∗B^{a}/B_{\mathrm{twist}}^{a*}. The data points are experimental results and the solid line is the simulation result. (f) Logarithmic plot of ϕL\phi_L as a function of Ba/Btwista∗−1B^{a}/B_{\mathrm{twist}}^{a*}-1. The legend is the same as (e). The dashed line is the fitting of the power law with the exponent 1/21/2, Eq. (48), obtained through scaling arguments.

We rationalize the experimental and numerical results using the magnetic Kirchhoff equations that we derived in § 3.3 by the perturbation against the twist instability. Following [65], we expand the Cosserat frame with respect to the straight configuration (d^1(0),d^2(0),d^3(0))=(e^y,e^z,e^x)(\hat{\boldsymbol{d}}_1^{(0)}, \hat{\boldsymbol{d}}_2^{(0)}, \hat{\boldsymbol{d}}_3^{(0)}) = (\hat{\boldsymbol{e}}_y, \hat{\boldsymbol{e}}_z, \hat{\boldsymbol{e}}_x) as

d^a=(δab+εabcαc)d^b(0),(42)\hat{\boldsymbol{d}}_a = \left(\delta_{ab}+\varepsilon_{abc}\alpha_c\right)\hat{\boldsymbol{d}}_b^{(0)} , \tag*{(42)}

where εabcαc\varepsilon_{abc}\alpha_c is the asymmetric tensor and αa=αa(s)\alpha_a = \alpha_a(s) (a=1,2,3a = 1, 2, 3) is the small (angular) perturbation parameter αa≪1\alpha_a \ll1. The coefficient εabcαc\varepsilon_{abc}\alpha_c enforces the orthogonality of d^a\hat{\boldsymbol{d}}_a (a=1,2,3a = 1, 2, 3) even up to the 1st order of the perturbation. Plugging the perturbed form of d^a\hat{\boldsymbol{d}}_a, Eq. (42), into the kinematic equation, Eq. (1), (see Appendix § 10 for details), we find

αa′=Ωa,(43)\alpha'_a = \Omega_a , \tag*{(43)}

from which we conclude that αa\alpha_a represents the infinitesimal rotation angle around d^a\hat{\boldsymbol{d}}_a. Given that the rod is clamped at s=0s = 0 and free at s=Ls = L, the corresponding boundary conditions are αa(0)=0\alpha_a(0) = 0 and αa′(L)=Ω(L)=0\alpha'_a(L) = \Omega(L) = 0. Furthermore, pmag=0p_{\mathrm{mag}} = 0 and thus F(s)=0\boldsymbol{F}(s) = 0. Substituting Eq. (42) into Eq. (21) and linearizing the result for αa\alpha_a yields

EIα1′′=−ABrBaμ0α1,(44)EI\alpha_1''=-\frac{AB^{r}B^{a}}{\mu_0}\alpha_1, \tag*{(44)}
EIα2′′=0,(45)EI\alpha_2''=0, \tag*{(45)}
GJα3′′=−ABrBaμ0α3.(46)GJ\alpha_3''=-\frac{AB^{r}B^{a}}{\mu_0}\alpha_3. \tag*{(46)}

Equations (44)-(46) are the governing equations for the infinitesimal rotation angles αa\alpha_a along the rod. To determine the conditions for the existence of nontrivial solutions (αa(s)≠0)(\alpha_a(s)\ne0) for Eqs. (44)-(46), we will substitute αa∝sin⁡(πs/2L)\alpha_a\propto\sin(\pi s/2L) (a=1,2,3a=1,2,3), which satisfies the boundary conditions αa(0)=αa′(L)=0\alpha_a(0)=\alpha_a'(L)=0, into Eqs. (44)-(46). Equation (45) gives us the null amplitude α2(s)=0\alpha_2(s)=0, while Equations (44) and (46) are satisfied if ABrBa/μ0=EI(π/2L)2AB^{r}B^{a}/\mu_0=EI(\pi/2L)^2 or ABrBa/μ0=GJ(π/2L)2AB^{r}B^{a}/\mu_0=GJ(\pi/2L)^2 hold, respectively. The smaller value of BaB^{a} derived here corresponds to the critical value of BaB^{a} for the twist instability, Btwista∗B_{\mathrm{twist}}^{a*}. Since GJ/EI=1/(1+ν)<1GJ/EI=1/(1+\nu)<1, the eigenvalue of Equation (46) corresponds to the critical flux density, above which the rod twists: Hence, we find

Btwista∗=(π2)2μ0GJABrL2.(47)B_{\mathrm{twist}}^{a*}=\left(\frac{\pi}{2}\right)^2\frac{\mu_0GJ}{AB^{r}L^2}. \tag*{(47)}

Notice that EIEI in Eq. (24) is now replaced by GJGJ in Eq. (47). Given that α3=α3(s)\alpha_3=\alpha_3(s) is the profile of the (infinitesimal) twisting angle (from Eq. (43)), α3(L)\alpha_3(L) corresponds to the twist angle of the free-end measured in our experiment and ϕL=α3(L)\phi_L = \alpha_3(L) in the limit of ϕL≪1\phi_L \ll1. By analogy with the equation of clamped-free elastica [43] (replacing α3(s)\alpha_3(s) with the bending angle in elastica), Equation (46) is now mathematically equivalent to the amplitude equation for the pitch-fork bifurcation. In the neighborhood of Ba/Btwista∗∼1B^{\mathrm{a}}/B_{\mathrm{twist}}^{\mathrm{a}*} \sim1, we thus anticipate that ϕL\phi_L obeys the following scaling law, where the twist angle at the tip ϕL\phi_L evolves as

ϕL∝BaBtwista∗−1.(48)\phi_L \propto\sqrt{\frac{B^{\mathrm{a}}}{B_{\mathrm{twist}}^{\mathrm{a}*}} - 1}. \tag*{(48)}

We find an excellent agreement between experiments and the theoretical description developed above ϕL\phi_L, although the transition is not sharp due to the inevitable imperfection of the system below the critical field Ba/Btwista∗<1B^{\mathrm{a}}/B_{\mathrm{twist}}^{\mathrm{a}*} < 1 [66]. Furthermore, from the logarithmic plot in Figure 6 (f), we confirm that ∣ϕL∣\lvert\phi_L\rvert is consistent with the predicted scaling law Eq. (48), in the region of ∣Ba/Btwista∗−1∣≪1\lvert B^{\mathrm{a}}/B_{\mathrm{twist}}^{\mathrm{a}*} - 1\rvert\ll1. The above results demonstrate that our magnetic Kirchhoff equations correctly predict the twist instability of straight rods. In particular, we validated the elasto-magnetic torque qmagq_{\mathrm{mag}} in the case of straight rods. Combining the fact that our framework reproduces the previous results on the planar deformation (§ 3.4) [28, 31, 32, 35] and the results presented in this subsection, the magnetic Kirchhoff equation (20) and (21) correctly predict the large deformation of naturally straight hard MRE rods. Next, we focus on the deformation of hard MRE rods that are naturally curved .

Buckling of a magnetic helix under constant field

We proceed by considering a naturally curved and twisted rod, i.e., a helix, which undergoes structural (buckling) instability of a helix at a critical field Ba∗B^{\mathrm{a}*}, but with an inevitable twist-bend coupling. We will show that our simulations and theoretical analyses rationalize our experimental results with excellent quantitative agreement.

We fabricate a hard magnetic helix with the pitch angle ψ=1.28\psi= 1.28 rad, which is clamped such that the central axis is aligned along the coil axis. We apply the magnetic field Ba=−Bae^z\mathbf{B}^{\mathrm{a}} = -B^{\mathrm{a}}\hat{\mathbf{e}}_z on a helical rod with lengths in the range of 40≤L [mm]≤12040 \le L\ [\mathrm{mm}] \le120. These helices are magnetized along their central axis as Br=Bre^z\mathbf{B}^{\mathrm{r}} = B^{\mathrm{r}}\hat{\mathbf{e}}_z. Below the critical field Ba<Ba∗B^{\mathrm{a}} < B^{\mathrm{a}*}, the rod remains uniformly helical, as shown in the filled data points in Figure 7 (a), buckling occurs when Ba>Ba∗B^{\mathrm{a}} > B^{\mathrm{a}*}, as depicted by the open symbols. We regarded the rod as helical if the tip moves by less than the diameter dd in the experiments.

Buckling of a magnetic helix

Figure 7. Buckling of a magnetic helix. (a) Phase diagram of the buckling instability of a magnetic helix in experiments. Filled and empty data points are uniform or buckled helix states, respectively. Buckling of a magnetic helix under constant field. In the previous section. The dashed line is the analytic prediction in the limit of L/R≪1L/R \ll1. The simulation snapshots for L/R=12L/R = 12 correspond to Ba/Ba∗=0.25B^{a}/B^{a*} = 0.25 (b) and Ba/Ba∗=2.5B^{a}/B^{a*} = 2.5 (c). (d) Rescaled phase diagram of our experimental results. The simulation results and asymptotic solutions are superposed with ψ=1.28\psi= 1.28 rad.

We now specialize the magnetic Kirchhoff equations Eqs. (20) and (21) for a magnetic helix to rationalize the experimental observations. The critical flux density for buckling of the magnetic helix Ba∗B^{\mathrm{a}*} can be derived via rigorous analysis of Eqs. (20) and (21). Here, to simplify the analysis and emphasize the importance of a twist-bend coupling in the system, we will compute the critical field strength Ba∗B^{\mathrm{a}*} in the limit of small curvature KL≪1KL \ll1 (KK is total curvature for a helix defined later) but with ψ<π/2\psi< \pi/2, which indeed provides a good prediction for our observations as we detail later.

Similarly to § 6.1, we expand the Cosserat frame with respect to the uniform helix as we did in Eqs. (42) and (43). The Darboux vector in the uniform helix is Ω=κd^2+τd^3\boldsymbol{\Omega} = \kappa\hat{\mathbf{d}}_2 + \tau\hat{\mathbf{d}}_3 with the natural curvature κ\kappa and twist τ\tau. By integrating the kinematic relation Eq. (1), d^˙a=Ω×d^a\dot{\hat{\mathbf{d}}}_a = \boldsymbol{\Omega} \times\hat{\mathbf{d}}_a, the

Cosserat frame d^a\hat{\mathbf{d}}_a (a=1,2,3a = 1, 2, 3) for a helix in the reference configuration is computed as

d^1(0)(s)=cos⁡(Ks)e^x−sin⁡(Ks)e^y,(49)\hat{\mathbf{d}}_1^{(0)}(s) = \cos(Ks)\hat{\mathbf{e}}_x - \sin(Ks)\hat{\mathbf{e}}_y, \tag*{(49)}
d^2(0)(s)=−cos⁡ψ(sin⁡(Ks)e^x+cos⁡(Ks)e^y)−sin⁡ψe^z,(50)\hat{\mathbf{d}}_2^{(0)}(s) = -\cos\psi\left(\sin(Ks)\hat{\mathbf{e}}_x + \cos(Ks)\hat{\mathbf{e}}_y\right) - \sin\psi\hat{\mathbf{e}}_z, \tag*{(50)}
d^3(0)(s)=sin⁡ψ(sin⁡(Ks)e^x+cos⁡(Ks)e^y)−cos⁡ψe^z.(51)\hat{\mathbf{d}}_3^{(0)}(s) = \sin\psi\left(\sin(Ks)\hat{\mathbf{e}}_x + \cos(Ks)\hat{\mathbf{e}}_y\right) - \cos\psi\hat{\mathbf{e}}_z. \tag*{(51)}

We choose the central axis to be parallel to z^\hat{\mathbf{z}} and define the pitch angle ψ\psi as κ=Ksin⁡ψ\kappa= K\sin\psi, τ=κcos⁡ψ\tau= \kappa\cos\psi, and K≡κ2+τ2K \equiv\sqrt{\kappa^2 + \tau^2}. The radius of curvature on the x−yx-y plane RR and ψ\psi are related as κ=sin⁡2ψ/R\kappa= \sin^2\psi/R and τ=sin⁡2ψ/2R\tau= \sin2\psi/2R (see Appendix B for the detailed derivation).

Plugging Eqs. (49)–(51) into the magnetic Kirchhoff equations (Eqs. (20) and (21)) and using Eqs. (42) and (43), we obtain the set of equilibrium equations for the linear perturbation of a helix in the limit of KL≪1KL \ll1:

EIα1′′=−ABrBaμ0α1(52)EI\alpha_1'' = -\frac{AB^rB^a}{\mu_0}\alpha_1 \tag*{(52)}
EIα2′′=ABrBaμ0cos⁡ψ(α2cos⁡ψ−α3sin⁡ψ)(53)EI\alpha_2'' = \frac{AB^rB^a}{\mu_0}\cos\psi\left(\alpha_2\cos\psi-\alpha_3\sin\psi\right) \tag*{(53)}
GJα3′′=−ABrBaμ0sin⁡ψ(α2cos⁡ψ−α3sin⁡ψ),(54)GJ\alpha_3'' = -\frac{AB^rB^a}{\mu_0}\sin\psi\left(\alpha_2\cos\psi-\alpha_3\sin\psi\right), \tag*{(54)}

with boundary conditions that are the same as those given in § 6.1: αa(0)=αa′(L)=0\alpha_a(0)=\alpha_a'(L)=0. Equations (52)–(54) are the linearized moment balance equations of Eq. (21).

Assuming the functional form αa∝sin⁡(ks)\alpha_a \propto\sin(ks) with k=π/2Lk=\pi/2L (that satisfies the boundary conditions αa(0)=αa′(L)=0\alpha_a(0)=\alpha_a'(L)=0), the critical magnetic field Ba∗B^{a*} satisfying αa≠0\alpha_a \ne0 can be obtained from Eqs. (53) and (54) as

Ba∗=(π2L)2μ0ABrEIcos⁡2ψ+(EI/GJ)sin⁡2ψ.(55)B^{a*} = \left(\frac{\pi}{2L}\right)^2 \frac{\mu_0}{AB^r}\frac{EI}{\cos^2\psi+(EI/GJ)\sin^2\psi}. \tag*{(55)}

Equation (55) captures the previous results on planar bending in Eq. (24) in § 3.4, as well as the twist instabilities in Eq. (47), which we investigated in § 6.1. Indeed, by taking the limit ψ→0\psi\to0, (i.e., rod clamped vertically and magnetized along its tangent), we get Bbenda∗→Ba∗B_{\mathrm{bend}}^{a*} \to B^{a*}. By contrast, in the limit of ψ→π/2\psi\to\pi/2, (i.e., rod clamped horizontally but magnetized as in § 6.1), we recover Btwista∗→Ba∗B_{\mathrm{twist}}^{a*} \to B^{a*}.

We performed numerical simulations corresponding to the buckling of a hard magnetic helix under the constant magnetic field. When the applied field is sufficiently small Ba<Ba∗B^{a}<B^{a*}, the rod remains helical, but the helix buckles when Ba>Ba∗B^{a}>B^{a*}. In Figure 7 (b) and (c), we show representative snapshots of the numerical simulations for Ba<Ba∗B^{a}<B^{a*} and Ba>Ba∗B^{a}>B^{a*}, respectively. Based on Eq. (55), we summarize experimental and numerical results in the rescaled phase diagram shown in Figure 7 (d). In the simulations, we remark that the rod remains helical if the displacement of the free-end is less than 10−3ℓ010^{-3}\ell_{0}, where ℓ0\ell_{0} is the natural length of the spring in the simulation (unit length of the simulation). The phase boundary between the undeformed (helical) and buckled states is in an excellent agreement between experiments and simulations. Furthermore, the analytic prediction (Eq. (55)) correctly predicts both experimental and numerical results. It should be noted that not only Eq. (55) predicts the phase boundary for L/R≪1L/R\ll1, but also that for L/R=O(1)L/R=O(1), without any adjustable parameters. A more rigorous analysis will be able to derive the full analytic prediction of the phase boundary, whereas it is beyond the scope of the current paper.

In this subsection, we have confirmed that the elasto-magnetic torque qmagq_{\mathrm{mag}} Eq. (19) derived in § 3.1 correctly predicts the experimental results for a naturally curved and twisted rod. Combining the results from § 6.1 and those in this subsection, we validated the magnetic Kirchhoff equations (Eqs. (20) and (21)) in 3D deformation under the constant field. In the next subsection, we consider the case of loading in a constant gradient magnetic field to validate the elasto-magnetic force pmagp_{\mathrm{mag}} (Eq. (18)) to fully validate our theoretical framework.

Deformation of a magnetic helix under constant gradient field

In this subsection, we will validate the framework for magnetic helices under the constant gradient field (pmag,qmag≠0p_{\mathrm{mag}},q_{\mathrm{mag}}\ne0). For this third and final test configuration, the magnetic helix is clamped at s=0s=0, while setting the other end (s=Ls=L) to be free. As we increase the gradient of the field bb (or equivalently, the rescaled gradient λm=L/ℓm\lambda_m=L/\ell_m, see Eq. (28)), the magnetic helix stretches. We will present results on the displacement of the free-end as a function of λm\lambda_m and find excellent agreement between experiments and simulations. In this subsection, we study the magnetic helix of a pitch shorter than that in § 6.2. Hence, the helix interacts with itself non-locally via long-range interaction. Although long-range interactions are not considered in our theory, we can predict the stretching of the helix as long as the pitch becomes larger upon increasing λm\lambda_m.

As an illustrative configuration, the magnetic helix is placed with its central axis aligned with the coil axis, and the clamped-end is located on the z=0z=0 plane (see Figure 8 for a photograph of the setup), within the error of ±1\pm1 mm. The magnetic helix has geometries of (L,R)=(10,103)(L,R)=(10,103) mm and pitch angle ψ=1.51\psi=1.51 rad. We suspend the magnetic helix in a glycerol bath, horizontally. The gradient of the magnetic field bb is increased from zero up to b=1.06b=1.06 T/m (maximum value in our setup), and then bb is decreased back to zero.

Deformation of a hard magnetic helix

Figure 8. Deformation of a hard magnetic helix. Snapshots of (a)-(c) experiments and (d)-(f) simulations with Br=13B^{r} = 13 mT, R=10R = 10 mm and L=103L = 103 mm. The corresponding values of rescaled gradient are (a) (d) λm=3.0\lambda_m = 3.0, (b) (e) λm=2.5\lambda_m = 2.5, (c) (f) λm=2.0\lambda_m = 2.0.

In the absence of the applied field, λm=0\lambda_m=0 (i.e. b=0b=0), the helix is fully contracted due to the long-range dipole-dipole attractive interaction. As discussed in [30, 35] and § 3, the applied field gradient is analogous to a unidirectional body force of constant magnitude such as gravity. Hence, the helix extends (or contracts) when λm\lambda_m is increased or decreased, respectively. We present the photograph of the extended helix with λm=3.0\lambda_m=3.0 in Figure 8 (a). The rescaled gradient is then decreased to λm=2.5\lambda_m=2.5 and λ=2.0\lambda=2.0, as shown in Figure 8 (b) and (c), respectively.

We performed the numerical simulations of the magnetic helix under the constant gradient field to compare its deformed configurations with the experimental ones. Figure 8 (d), (Figure 8e) and (Figure 8f) are the snapshots of our discrete simulation with λm=3.0\lambda_m = 3.0, 2.5 and 2.0, respectively. Similar to our experimental observations, the helix contracts as λm\lambda_m is decreased. The predicted shapes in the numerical simulation agree with the configurations in experiments qualitatively.

To quantify the deformation of the helix in the present configuration, we measured the end-to-end distance along zz between the clamp and the tip of the helix: zL≡z(L)−z(0)z_L \equiv z(L) - z(0) defined in Figure 8 (a). In Figure 9 (a)-(Figure 9c), we plot zLz_L as a function of the rescaled applied field gradient λm=L/ℓm\lambda_m = L/\ell_m for different values of LL or BrB^{r}, where ℓm\ell_m is the magneto-bending length defined in Eq. (28). The data points correspond to the experimental results. As an example, we describe the experimental results presented in Figure 9 (a). In the absence of the applied field λm=0\lambda_m = 0, the helix is in self-contact due to long-range attractive interactions. Upon increasing λm\lambda_m, the helix loses self-contact when λm=λm+\lambda_m = \lambda_m^{+}. Past this point, zLz_L increases as the elasto-magnetic force stretches the helix. When λm\lambda_m is decreased, the helix contracts until self-contact at λm=λm−\lambda_m = \lambda_m^{-}. For the sake of convenience, we refer to the branch where the helix is in self-contact self-contact branch, and to the branch without self-contact as the elasto-magnetic branch.

The z-position of the free-end as a function of the rescaled gradient for different magnetic fields and rod lengths

Figure 9. The zz-position of the free-end as a function of the rescaled gradient λm=L/ℓm\lambda_m = L/\ell_m for different BrB^{\mathrm{r}} and length LL. The data points and solid lines are experimental and simulation results, respectively. The shaded areas are errors due to the clamped position computed from the simulation. (a) Br=13B^{\mathrm{r}} = 13 mT, L=103L = 103 mm (b) Br=13B^{\mathrm{r}} = 13 mT, L=74L = 74 mm (c) Br=26B^{\mathrm{r}} = 26 mT, L=81L = 81 mm.

We find that the values of λm±\lambda_m^{\pm} depend on BrB^{r} and LL. Indeed, in Figure 9 (a) and (Figure 9c), λm−≠0\lambda_m^{-} \ne0, while, in Figure 9 (b), λm−=0\lambda_m^{-} = 0. The fact that λm−=0\lambda_m^{-} = 0 implies that the helical shape (without contact) is stable but it is not a unique stable state due to the long-range interactions. The value of λm±\lambda_m^{\pm} can be used to classify the hysteric behavior, depending on the magnitude of residual flux density ∣Br∣|B^{r}|, the pitch angle ψ\psi, and the rescaled total length L/RL/R. We will discuss the hysteric behavior in zLz_L more in detail, in § 7.

Complementing the experimental data, we performed the discrete simulations corresponding to our experimental settings to measure the end-to-end distance zLz_L as a function of λm\lambda_m. Given that our theory does not include the self long-range interaction, zLz_L is not hysteretic (the self-contact branch does not appear). Still the predictions from our model are in excellent agreement with the experimental results in the elasto-magnetic branches. In Figure 9 (a)-(Figure 9c), we plot the predictions from the simulations as solid lines. The width of the shaded region in Figure 9 (a)-(Figure 9c) correspond to the propagation of errors due to the ±1\pm1 mm, uncertainty of the total length LL measurement. From the excellent agreement between the experimental data and numerical results, we conclude that the magnetic Kirchhoff equations (Eqs. (20) and (21)) provide an accurate quantitative prediction for the deformation of the magnetic helix under a constant gradient field, when the self long-range interaction is not dominant.

In this section, we combined the discrete simulations with experiments, to validate the magnetic Kirchhoff equations (Eqs. (20) and (21)) derived in § 3.3, for specific cases where the long-range interaction can be neglected compared with the elasto-magnetic forces. We showed that the numerical simulations can predict the behavior of the elasto-magnetic branch observed experimentally. We have validated the elasto-magnetic torque qmagq_{\mathrm{mag}} and force pmagp_{\mathrm{mag}} in a naturally straight and curved rod, thereby allowing us to validate Eqs. (20) and (21) in 3D geometrically nonlinear deformations, at least when the long-range interaction is negligible. Next, in § 7, we perform a more systematic experimental study of when and how the long-range interaction affects the deformation of a magnetic helix.

Limitations of the theory

In this section, we systematically quantify the limitation of the theory through precision experiments alone, because the self long-range interactions is not included in Eqs. (20) and (21); doing so is a challenging endeavour that goes beyond the scope of the present study. We classify the hysteretic behavior reported in Figure 9 based on the value of λm±\lambda_m^{\pm}; the critical field strength at which zLz_L jumps discontinuously. We will discuss the classification by considering the structural instability of the magnetic helix in the absence of the applied field, that is λm=0\lambda_m = 0.

The long-range interactions in the magnetic helix depend on the combined effects of the magnitude of the magnetization ∣M∣∝∣Br∣|\mathcal{M}| \propto|B^{\mathrm{r}}| and the geometry of the helix (e.g., the pitch angle ψ\psi and the rescaled total length L/(2πR)L/(2\pi R)). The magnitude of the magnetization ∣M∣|\mathcal{M}| is programmed at the fabrication stage by the concentration ratio of the NdPrFeB particles cc. When cc is small, the long-range interaction will be suppressed (and vice versa). On the other hand, the dependence of ψ\psi and L/(2πR)L/(2\pi R) is highly nontrivial, as we discuss below.

The constant gradient field is applied to the magnetic helix as in § 6.3. The helix is extended and contracted under the constant gradient magnetic field, for several pitch angles (in the natural configuration). We introduce the normalized pitch angle as ψ~≡ψ/(π/2)\widetilde{\psi}\equiv\psi/(\pi/2) such that ψ~=1\widetilde{\psi}=1 corresponds to the planar circle with zero twist (experimentally non-realizable due to the finite diameter of the rod) and ψ~=0\widetilde{\psi}=0 corresponds to a straight rod. As ψ~\widetilde{\psi} decreases, the pitch of the helix increases. Practically, ψ~max⁡=0.96\widetilde{\psi}_{\max}=0.96 is the largest value of ψ~\widetilde{\psi} realizable in our experimental conditions with (d,R)=(2,10) mm(d,R)=(2,10)\,\mathrm{mm}. Our rods are fabricated systematically through the procedure in § 5.1 with different pitch angles in the range 0.80≤ψ~≤0.950.80\le\widetilde{\psi}\le0.95. We clamp the magnetized helix with Br=26 mT\mathbf{B}^{r}=26\,\mathrm{mT} at z(0)=0z(0)=0, while s=Ls=L is set to be free. Through the cyclic protocols consisting of increasing and decreasing the flux density, the values of λm\lambda_m characterizing the hysteresis λm±\lambda_m^{\pm} are identified (see Figure 9 for the definition of λm±\lambda_m^{\pm}).

In Figure 10, we present the experimental phase diagram in the ψ~\widetilde{\psi}-L/2πRL/2\pi R parameter space to classify the response against λm\lambda_m, finding that there are three different possible responses (i)-(iii), depending on the geometry of the magnetic helix. (i) When L/2πR≃1L/2\pi R\simeq1 (length of the helix is nearly a single turn) or ψ~\widetilde{\psi} is small, (i.e. when the pitch p=2πR/tan⁡ψp=2\pi R/\tan\psi is long enough), no hysteresis is observed (circles in Figure 10). When we plot zLz_L as a function of λm\lambda_m, zLz_L changes continuously and reversibly for any λm\lambda_m. The tip of the helix follows the same curve upon increasing or decreasing λm\lambda_m. In this case, we have λm+=λm−=0\lambda_m^{+}=\lambda_m^{-}=0 (Figure 10). For higher values of ψ\psi, we observe hysteresis, which is divided into two further categories; λm+≠0\lambda_m^+ \ne0, λm−=0\lambda_m^- = 0 or λm+,λm−≠0\lambda_m^+, \lambda_m^- \ne0, which are shown as triangles and cross marks in Figure 10, respectively. (ii) In the case of λm+≠0\lambda_m^+ \ne0, λm−=0\lambda_m^- = 0, zLz_L changes discontinuously from the contracted helix to the stretched helix at λm=λm+\lambda_m = \lambda_m^+, while zLz_L decreases smoothly as λm→0\lambda_m \to0, i.e. λm−=0\lambda_m^- = 0 (see Figure 9 (b) for the corresponding behavior). (iii) By contrast, in the case of λm+,λm−≠0\lambda_m^+, \lambda_m^- \ne0, zLz_L jumps at both λm+\lambda_m^+ and λm−\lambda_m^-. Thus, if we plot zLz_L as a function of λm\lambda_m, we obtain similar curves as those in Figure 9 (a) and (Figure 9c).

Classification of the hysteretic behavior in the stretching and contraction of a magnetic helix

Figure 10. Classification of the hysteretic behavior in the stretching and contraction of a magnetic helix. The phase diagram of hysteresis is obtained from experiments with Br=26 mT\mathbf{B}^{r}=26\,\mathrm{mT} and R=10 mmR=10\,\mathrm{mm}; ○: No hysteresis (λm+,λm−=0\lambda_{m}^{+},\lambda_{m}^{-}=0), △: Meta-stable helix (λm+≠0,λm−=0\lambda_{m}^{+}\ne0,\lambda_{m}^{-}=0) and ×: Fully contracted helix (λm+,λm−≠0\lambda_{m}^{+},\lambda_{m}^{-}\ne0).

The classification of the hysteretic responses in the zLz_L-λm\lambda_m plots (i)-(iii) can be understood by considering the configuration at λm=0\lambda_m = 0. In the following, we discuss the stability of the uniform helix at λm=0\lambda_m = 0 against mechanical perturbations. (i) When λm+=λm−=0\lambda_m^+ = \lambda_m^- = 0 (no hysteresis), the uniform helical shape corresponds to a unique energy minimum. The reference configuration is mono-stable upon any mechanical perturbations, because self-interactions have minimal effects. (ii) For λm+≠0\lambda_m^+ \ne0 and λm−=0\lambda_m^- = 0, the uniform helix is stable but it is not a unique minimum. In this case, the helix can contracts to the (stable) state with the normalized pitch angle of ψ~=ψ~max⁡\tilde{\psi} = \tilde{\psi}_{\max} if perturbed; we call this state meta-stable helix. Lastly, (iii) for λm+,λm−≠0\lambda_m^+, \lambda_m^- \ne0, the uniform helix is no longer stable. Indeed, the contracted state is the unique energy minimum. In this third case, the helix remains fully-contracted against any mechanical perturbation.

The nature of the hysteretic behavior in magnetic helices depends on the concentration ratio of NdPrFeB particles cc and the geometry of the helix as described by ψ\psi and L/(2πR)L/(2\pi R). We classified the deformation due to long-range self-interactions based on the value of λm\lambda_m at which zL(λm)z_L(\lambda_m) changes discontinuously. The hysteretic behavior in the magnetic helix originates from the bi-stability due to the long-range interactions. In the previous sections, we had validated the magnetic Kirchhoff rod equations when the long-range interaction can be neglected. At the same time, we clarified the limitation of our theory systematically. The hysteretic behavior, which our theory cannot capture, is observed only when the pitch angle ψ~\tilde{\psi} is large and when the rod is long L/(2πR)≳1L/(2\pi R) \gtrsim1. We can conclude that the magnetic Kirchhoff rod equations (Eqs. (20) and (21)) are valid as long as ψ\psi, L/(2πR)L/(2\pi R), or cc remains small.

Conclusion

The goal of the present study was to develop a Kirchhoff-like theory for hard magnetic rods based on dimensional reduction. Our theoretical framework, rooted in 3D elasticity, yielded a reduced (centerline-based) description of the magnetic rod (§ 3). The set of governing equations contain elasto-magnetic forces and torques (Eqs. (18) and (19)), in addition to the purely elastic ones for a (non-magnetic) rod. We validated the magnetic Kirchhoff rod equations (Eqs. (20) and (21)) through a set of precision experiments. Our theoretical results supplemented by the discrete simulation are in excellent agreement with the experimental results. Moreover, the magnetic Kirchhoff rod equations reproduce the previous results on the planar deformation of hard magnetic beams or elastica [28, 30, 32, 35]. We confirmed that the force and moment balance equations in our theory for hard magnetic rods reduce to those of Refs. [28, 30, 32, 35] for beams and elastica.

To validate our theory for the 3D deformation of magnetic rods, we performed three different sets of experiments; (i) a straight magnetic rod under constant external magnetic field (§ 6.1), (ii) a helical rod under constant field (§ 6.2), and (iii) a helical rod under the constant gradient field (§ 6.3), with the clamped-free boundary conditions. (i) In the case of a straight rod under the constant field, we studied the twist instability. When the rod is magnetized along d^1\hat{\boldsymbol{d}}_1 (perpendicular to the tangent d^3\hat{\boldsymbol{d}}_3) and we apply an anti-parallel field, the rod twists, while its centerline remains straight. We derived an analytical prediction for the critical applied field for this twist instability Btwista∗B_{\mathrm{twist}}^{a*}, which is in excellent agreement with experimental results. (ii) We studied the buckling instability of a helical rod under constant external magnetic field. The helix was magnetized along its central axis. When we apply the field to the magnetic helix, it buckles above the critical applied field Ba∗B^{a*}. We showed that this instability is triggered by a balance between elastic and elasto-magnetic torques. The critical applied fields were predicted analytically from a linear stability analysis. The simulation results correctly capture the experimental observations. (iii) Under a constant gradient field, the helix magnetized along the central axis stretches (or contracts) as we increase (or decrease) the applied field strength. Due to the long-range self-interaction between magnetic dipoles within the rod, the deformation is hysteretic upon extension and contraction.

Our framework correctly predicts the deformation of the magnetic helix when its pitch is large enough or when the long-range self-interactions are negligible. The hysteretic behavior was studied systematically through experiments by controlling the pitch angle ψ\psi of the helix. We revealed that there are three types of behavior; no hysteresis state, meta-stable, and fully contracted helices, depending on the total length L/RL/R and the pitch angle ψ\psi.

Although our theory is in excellent agreement with the experimental results when the elasto-magnetic force and torque are dominant, the non-local interactions between the magnetization vectors M\boldsymbol{\mathcal{M}} are necessary to describe the hysteretic behavior. In Refs. [15, 16, 17], the continuum mechanics of magnetic chains has been studied by considering the self-interactions between magnetic beads. In the future, combining the continuum mechanics of the magnetic dipole interactions and our magnetic Kirchhoff rod equations, the hysteretic behavior of the magnetic helix could be studied in detail theoretically. Extending the magnetic Kirchhoff equations to include the dipole-dipole interactions is one of the interesting directions of future works. Also, extending our theoretical framework toward dynamics of hard magnetic rods is also another exciting opportunity for future work. The dynamics could be simulated by considering the (external) viscous force and torque of a surrounding medium in Eqs. (20) and (21).

The framework established in this paper would be valuable to simulate the large deformation of hard magnetic rods used in micro or soft-robotics [5, 32, 36, 39] or haptic devices [40]. In the future, novel functional devices of complex geometries could be developed by combining the simple building blocks studied in this paper.

Acknowledgement

T.G.S. was supported by Grants-in-Aid for Japan Society for the Promotion of Science Overseas Research Fellowship 2019-60059 and MEXT KAKENHI 18K13519.

Derivation of Equation (43)

In this appendix, we provide the detailed derivation of Eq. (43);

αa′=Ωa.(56)\alpha'_{a} = \Omega_{a}. \tag*{(56)}

We recall that our rod is naturally curved and twisted: Ωa(0)=Ωa(0)d^a(0)\boldsymbol{\Omega}_{a}^{(0)} = \Omega_{a}^{(0)} \hat{\boldsymbol{d}}_{a}^{(0)}. We expand the Cosserat frame basis in Eq. (42) around the reference configuration d^a(0)\hat{\boldsymbol{d}}_{a}^{(0)}. We will determine the relationship between the first-order perturbation of the components of the Darboux vector Ωa(1)\Omega_{a}^{(1)} defined as

Ωa=Ωa(0)+Ωa(1)+⋯ .(57)\Omega_{a} = \Omega_{a}^{(0)} + \Omega_{a}^{(1)} + \cdots. \tag*{(57)}

and the first order perturbation of d^a\hat{d}_a, αa\alpha_a. Note that Ωa(1)\Omega_a^{(1)} and αa\alpha_a are infinitesimal quantities of the same order Ωa(1)∼αa\Omega_a^{(1)} \sim\alpha_a. We substitute Eqs. (42)

d^a=(δab+εabcαc)d^b(0),(58)\hat{d}_a = (\delta_{ab} + \varepsilon_{abc}\alpha_c)\hat{d}_b^{(0)}, \tag*{(58)}

and (A.2) into Eq. (1); d^a′=Ω×d^a\hat{d}'_a = \boldsymbol{\Omega} \times\hat{d}_a, to derive Eq. (43). The final result is consistent with that in Ref. [65] (which is based on matrix representation), while our derivation is based on a tensor representation.

First, the kinematic equation Eq. (1) is rewritten using the Eddington epsilon as

d^a′=−εabcΩbd^c,(59)\hat{d}'_a = -\varepsilon_{abc}\Omega_b\hat{d}_c, \tag*{(59)}

which also holds for the base solution as d^a(0)′=−εabcΩb(0)d^c(0)\hat{d}_a^{(0)\prime} = -\varepsilon_{abc}\Omega_b^{(0)}\hat{d}_c^{(0)}. The left-hand side of Eq. (A.4) is computed as

d^a′={(δab+εabcαc)d^b(0)}′=−εabcΩb(0)d^c(0)+εabcαc′d^b(0)−εabcεbdeαcΩd(0)d^e(0)=−εabcΩb(0)d^c(0)+εabcαc′d^b(0)−(δcdδae−δceδad)αcΩd(0)d^e(0)=−εabcΩb(0)d^c(0)+εabcαc′d^b(0)−(αcΩc(0))d^a(0)+Ωa(0)(αcd^c(0)).(60)\begin{aligned} \hat{d}'_a &= \left\{(\delta_{ab}+\varepsilon_{abc}\alpha_c)\hat{d}_b^{(0)}\right\}' \\ &= -\varepsilon_{abc}\Omega_b^{(0)}\hat{d}_c^{(0)}+\varepsilon_{abc}\alpha'_c\hat{d}_b^{(0)}-\varepsilon_{abc}\varepsilon_{bde}\alpha_c\Omega_d^{(0)}\hat{d}_e^{(0)} \\ &= -\varepsilon_{abc}\Omega_b^{(0)}\hat{d}_c^{(0)}+\varepsilon_{abc}\alpha'_c\hat{d}_b^{(0)}-(\delta_{cd}\delta_{ae}-\delta_{ce}\delta_{ad})\alpha_c\Omega_d^{(0)}\hat{d}_e^{(0)} \\ &= -\varepsilon_{abc}\Omega_b^{(0)}\hat{d}_c^{(0)}+\varepsilon_{abc}\alpha'_c\hat{d}_b^{(0)}-(\alpha_c\Omega_c^{(0)})\hat{d}_a^{(0)}+\Omega_a^{(0)}(\alpha_c\hat{d}_c^{(0)}). \tag*{(60)} \end{aligned}

We expand the right-hand side of Eq. (A.4) as

−εabcΩbd^c=−εabc(Ωb(0)+Ωb(1))(δcd+εcdeαe)d^d(0)=−εabcΩb(0)d^c(0)−εabcΩb(1)d^c(0)−εabcεcdeαeΩb(0)d^d(0)=−εabcΩb(0)d^c(0)−εabcΩb(1)d^c(0)−(δadδbe−δaeδbd)αeΩb(0)d^d(0)=−εabcΩb(0)d^c(0)−εabcΩb(1)d^c(0)−(αcΩc(0))d^a(0)+αa(Ωc(0)d^c(0)).(61)\begin{aligned} -\varepsilon_{abc}\Omega_b\hat{d}_c &= -\varepsilon_{abc}(\Omega_b^{(0)}+\Omega_b^{(1)})(\delta_{cd}+\varepsilon_{cde}\alpha_e)\hat{d}_d^{(0)} \\ &= -\varepsilon_{abc}\Omega_b^{(0)}\hat{d}_c^{(0)}-\varepsilon_{abc}\Omega_b^{(1)}\hat{d}_c^{(0)}-\varepsilon_{abc}\varepsilon_{cde}\alpha_e\Omega_b^{(0)}\hat{d}_d^{(0)} \\ &= -\varepsilon_{abc}\Omega_b^{(0)}\hat{d}_c^{(0)}-\varepsilon_{abc}\Omega_b^{(1)}\hat{d}_c^{(0)}-(\delta_{ad}\delta_{be}-\delta_{ae}\delta_{bd})\alpha_e\Omega_b^{(0)}\hat{d}_d^{(0)} \\ &= -\varepsilon_{abc}\Omega_b^{(0)}\hat{d}_c^{(0)}-\varepsilon_{abc}\Omega_b^{(1)}\hat{d}_c^{(0)}-(\alpha_c\Omega_c^{(0)})\hat{d}_a^{(0)}+\alpha_a(\Omega_c^{(0)}\hat{d}_c^{(0)}). \tag*{(61)} \end{aligned}

Note that, in Eqs. (A.5) and (A.6), we dropped higher order terms (e.g. the terms proportional to αaΩb(1)\alpha_a\Omega_b^{(1)}).

Substituting Eqs. (A.5) and (A.6) into Eq. (A.4), we obtain the tensorial relation

εabcαc′d^b(0)=−εabcΩb(1)d^c(0)+αa(Ωc(0)d^c(0))−Ωa(0)(αcd^c(0)).(62)\varepsilon_{abc}\alpha'_c\hat{d}_b^{(0)} = -\varepsilon_{abc}\Omega_b^{(1)}\hat{d}_c^{(0)}+\alpha_a(\Omega_c^{(0)}\hat{d}_c^{(0)})-\Omega_a^{(0)}(\alpha_c\hat{d}_c^{(0)}). \tag*{(62)}

Next, by taking the inner product between Eq. (A.7) and d^d(0)\hat{d}_d^{(0)}, and by arranging the indices, we arrive at

εabcαc′=εabcΩc(1)+(αaΩb(0)−Ωa(0)αb).(63)\varepsilon_{abc}\alpha'_c = \varepsilon_{abc}\Omega_c^{(1)}+(\alpha_a\Omega_b^{(0)}-\Omega_a^{(0)}\alpha_b). \tag*{(63)}

The presence of the term inside the parenthesis in the right-hand side of Eq. (A.8) encodes the geometric non-linearity intrinsic to a naturally curved rod. When the rod is naturally straight, Ωa(0)=0\Omega_a^{(0)}=0, we recover Eq. (43).

Appendix B Derivation of Equations (49)-(51)

In this appendix, we derive the Cosserat frame basis for the uniform helix Eqs. (49)-(51). To do so, we will integrate the kinematic relation Eq. (1) with Ω=κd^2+τd^3\boldsymbol{\Omega}=\kappa\hat{d}_2+\tau\hat{d}_3. The goal is to express r(s)\mathbf{r}(s) and d^a\hat{\mathbf{d}}_a (a=1,2,3a=1,2,3) in terms of the Cartesian basis e^i\hat{\mathbf{e}}_i. From the kinematic relation, we get

d^1′=−κd^3+τd^2,(64)\hat{\mathbf{d}}'_1=-\kappa\hat{\mathbf{d}}_3+\tau\hat{\mathbf{d}}_2, \tag*{(64)}
d^2′=−τd^1,(65)\hat{\mathbf{d}}'_2=-\tau\hat{\mathbf{d}}_1, \tag*{(65)}
d^3′=κd^1,(66)\hat{\mathbf{d}}'_3=\kappa\hat{\mathbf{d}}_1, \tag*{(66)}

which are the same as the Frenet-Serret equations [43]. Differentiating Eq. (B.1) with respect to ss, and with the help of Eqs. (B.2) and (B.3), we find the ODE only for d^1\hat{\mathbf{d}}_1:

d^1′′=−K2d^1.(67)\hat{\mathbf{d}}''_1=-K^2\hat{\mathbf{d}}_1. \tag*{(67)}

Here, we introduced the normalized curvature KK as K=κ2+τ2K=\sqrt{\kappa^2+\tau^2}, or equivalently

κ=Ksin⁡ψ,τ=Kcos⁡ψ.(68)\kappa=K\sin\psi,\qquad\tau=K\cos\psi. \tag*{(68)}

After integrating Eq. (B.4) twice, we compute d^1\hat{\mathbf{d}}_1 as

d^1(s)=cos⁡(Ks)d^1(0)+τKsin⁡(Ks)d^2(0)−κKsin⁡(Ks)d^3(0).(69)\hat{\mathbf{d}}_1(s)=\cos(Ks)\hat{\mathbf{d}}_1(0)+\frac{\tau}{K}\sin(Ks)\hat{\mathbf{d}}_2(0)-\frac{\kappa}{K}\sin(Ks)\hat{\mathbf{d}}_3(0). \tag*{(69)}

Following a similar procedure for d^2\hat{\mathbf{d}}_2 and d^3\hat{\mathbf{d}}_3, we obtain

d^2(s)−d^2(0)=−τKsin⁡(Ks)d^1(0)−τK2(1−cos⁡(Ks))(τd^2(0)−κd^3(0)),(70)\hat{\mathbf{d}}_2(s)-\hat{\mathbf{d}}_2(0)=-\frac{\tau}{K}\sin(Ks)\hat{\mathbf{d}}_1(0)-\frac{\tau}{K^2}(1-\cos(Ks))(\tau\hat{\mathbf{d}}_2(0)-\kappa\hat{\mathbf{d}}_3(0)), \tag*{(70)}
d^3(s)−d^3(0)=κKsin⁡(Ks)d^1(0)+κK2(1−cos⁡(Ks))(τd^2(0)−κd^3(0)).(71)\hat{\mathbf{d}}_3(s)-\hat{\mathbf{d}}_3(0)=\frac{\kappa}{K}\sin(Ks)\hat{\mathbf{d}}_1(0)+\frac{\kappa}{K^2}(1-\cos(Ks))(\tau\hat{\mathbf{d}}_2(0)-\kappa\hat{\mathbf{d}}_3(0)). \tag*{(71)}

By integrating r′=d^3\mathbf{r}'=\hat{\mathbf{d}}_3, the centerline position is obtained as

r(s)−r(0)=κK2(1−cos⁡(Ks))d^1(0)−κK3sin⁡(Ks)(τd^2(0)−κd^3(0))+τsK2(κd^2(0)+τd^3(0)),(72)\mathbf{r}(s)-\mathbf{r}(0)=\frac{\kappa}{K^2}(1-\cos(Ks))\hat{\mathbf{d}}_1(0)-\frac{\kappa}{K^3}\sin(Ks)(\tau\hat{\mathbf{d}}_2(0)-\kappa\hat{\mathbf{d}}_3(0))+\frac{\tau s}{K^2}(\kappa\hat{\mathbf{d}}_2(0)+\tau\hat{\mathbf{d}}_3(0)), \tag*{(72)}

where the unit vector of the central axis of the helix N^\hat{\mathbf{N}} is

N^≡κKd^2(0)+τKd^3(0).(73)\hat{\mathbf{N}}\equiv\frac{\kappa}{K}\hat{\mathbf{d}}_2(0)+\frac{\tau}{K}\hat{\mathbf{d}}_3(0). \tag*{(73)}

Finally, let us rewrite Eq. (B.9) in a more compact form by choosing d^a(0)\hat{\mathbf{d}}_a(0) and r(0)\mathbf{r}(0) such that the central axis lies along −e^z-\hat{\mathbf{e}}_z; i.e., we select N^=−e^z\hat{\mathbf{N}}=-\hat{\mathbf{e}}_z, d^1(0)=e^x\hat{\mathbf{d}}_1(0)=\hat{\mathbf{e}}_x, and e^y=e^z×e^x=−(τd^2(0)−κd^3(0))/K\hat{\mathbf{e}}_y=\hat{\mathbf{e}}_z\times\hat{\mathbf{e}}_x=-(\tau\hat{\mathbf{d}}_2(0)-\kappa\hat{\mathbf{d}}_3(0))/K. Then, using Eq. (B.5), the centerline position is calculated as

r(s)=−sin⁡ψKcos⁡(Ks)e^x+sin⁡ψKsin⁡(Ks)e^y−scos⁡ψe^z,(74)\mathbf{r}(s)=-\frac{\sin\psi}{K}\cos(Ks)\hat{\mathbf{e}}_x+\frac{\sin\psi}{K}\sin(Ks)\hat{\mathbf{e}}_y-s\cos\psi\hat{\mathbf{e}}_z, \tag*{(74)}

which implies

R=sin⁡ψK.(75)R=\frac{\sin\psi}{K}. \tag*{(75)}

The normalized curvature KK is now rewritten as a function of κ\kappa and ψ\psi only; K=κ1+tan⁡−2ψ=κ/sin⁡ψK=\kappa\sqrt{1+\tan^{-2}\psi}=\kappa/\sin\psi, i.e. κ=sin⁡2ψ/R\kappa=\sin^2\psi/R and τ=sin⁡ψcos⁡ψ/R=sin⁡2ψ/2R\tau=\sin\psi\cos\psi/R=\sin2\psi/2R. The choice of N^\hat{\mathbf{N}} allows us to obtain the Cosserat frame basis vectors as

d^1(s)=cos⁡(Ks)e^x−sin⁡(Ks)e^y,(76)\hat{\mathbf{d}}_1(s)=\cos(Ks)\hat{\mathbf{e}}_x-\sin(Ks)\hat{\mathbf{e}}_y, \tag*{(76)}
d^2(s)=−cos⁡ψ(sin⁡(Ks)e^x+cos⁡(Ks)e^y)−sin⁡ψe^z,(77)\hat{\mathbf{d}}_2(s)=-\cos\psi\left(\sin(Ks)\hat{\mathbf{e}}_x+\cos(Ks)\hat{\mathbf{e}}_y\right)-\sin\psi\hat{\mathbf{e}}_z, \tag*{(77)}
d^3(s)=sin⁡ψ(sin⁡(Ks)e^x+cos⁡(Ks)e^y)−cos⁡ψe^z,(78)\hat{\mathbf{d}}_3(s)=\sin\psi\left(\sin(Ks)\hat{\mathbf{e}}_x+\cos(Ks)\hat{\mathbf{e}}_y\right)-\cos\psi\hat{\mathbf{e}}_z, \tag*{(78)}

thereby reproducing Eqs. (49)-(51) that we set out to derive.

References

  1. [1]W. Li, K. Kostidis, Xianzhuo Zhang, Yang Zhou, Development of a force sensor working with mr elastomers, in: 2009 IEEE/ASME International Conference on Advanced Intelligent Mechatronics, pp. 233–238.DOI
  2. [2]L. Dorfmann, R. W. Ogden, Nonlinear Theory of Electroelastic and Magnetoelastic Interactions, Springer, New York, 2014.DOI
  3. [3]A. M. Menzel, Mesoscopic characterization of magnetoelastic hybrid materials: magnetic gels and elastomers, their particle-scale description, and scale-bridging links, Archive of Applied Mechanics 89 (2019) 17–45.DOI
  4. [4]J. M. Ginder, M. E. Nichols, L. D. Elie, J. L. Tardiff, Magnetorheological elastomers: properties and applications, in: M. R. Wuttig (Ed.), Smart Structures and Materials 1999: Smart Materials Technologies, International Society for Optics and Photonics, 1999, pp. 131–138.DOI
  5. [5]H.-W. Huang, M. S. Sakar, K. Riederer, N. Shamsudhin, A. Petruska, S. Pané, B. J. Nelson, Magnetic Microrobots with Addressable Shape Control , IEEE International Conference on Robotics and Automation ICRA Stockholm, Sweden (2016).DOI
  6. [6]W. Hu, G. Z. Lum, M. Mastrangeli, M. Sitti, Small-scale soft-bodied robot with multimodal locomotion, Nature 554 (2018) 81–85.DOI
  7. [7]L. Pancaldi, P. Dirix, A. Fanelli, A. M. Lima, N. Stergiopulos, P. J. Mosimann, D. Ghezzi, M. S. Sakar, Flow driven robotic navigation of microengineered endovascular probes, Nature Communications 11 (2020) 1–14.DOI
  8. [8]F. C. Moon, Y.-H. Pao, Magnetoelastic Buckling of a Thin Plate, Journal of Applied Mechanics 35 (1968) 53–58.
  9. [9]A. Cebers, Dynamics of a chain of magnetic particles connected with elastic linkers, Journal of Physics: Condensed Matter 15 (2003) S1335–S1344.DOI
  10. [10]A. Čēbers, I. Javaitis, Bending of flexible magnetic rods, Physical Review E 70 (2004) 021404.pubmed.ncbi.nlm.nih.gov/15447488
  11. [11]R. Dreyfus, J. Baudry, M. L. Roper, M. Fermigier, H. A. Stone, J. Bibette, Microscopic artificial swimmers, Nature 437 (2005) 862–865.pubmed.ncbi.nlm.nih.gov/16208366
  12. [12]M. Roper, R. Dreyfus, J. Baudry, M. Fermigier, J. Bibette, H. A. Stone, On the dynamics of magnetically driven elastic filaments, Journal of Fluid Mechanics 554 (2006) 167–190.DOI
  13. [13]A. Čēbers, T. Cirulis, Magnetic elastica, Physical Review E 76 (2007) 031504.pubmed.ncbi.nlm.nih.gov/17930248
  14. [14]F. Gerbal, Y. Wang, F. Lyonnet, J.-C. Bacri, T. Hocquet, M. Devaud, A refined theory of magnetoelastic buckling matches experiments with ferromagnetic and superparamagnetic rods, Proceedings of the National Academy of Sciences 112 (2015) 7135–7140.
  15. [15]C. L. Hall, D. Vella, A. Goriely, The Mechanics of a Chain or Ring of Spherical Magnets, SIAM Journal on Applied Mathematics 73 (2013) 2029–2054.DOI
  16. [16]D. Vella, E. du Pontavice, C. L. Hall, A. Goriely, The magneto-elastica: from self-buckling to self-assembly, Proceedings of the Royal Society of London A: Mathematical, Physical and Engineering Sciences 470 (2014) 20130609.pubmed.ncbi.nlm.nih.gov/24511257
  17. [17]J. Schönke, E. Fried, Stability of vertical magnetic chains, Proceedings of the Royal Society of London A: Mathematical, Physical and Engineering Sciences 473 (2017) 20160703.DOI
  18. [18]Z. Rigbi, L. Jilkén, The response of an elastomer filled with soft ferrite to mechanical and magnetic influences, Journal of Magnetism and Magnetic Materials 37 (1983) 267–276.DOI
  19. [19]A. Dorfmann, R. W. Ogden, Magnetoelastic modelling of elastomers, European Journal of Mechanics - A/Solids 22 (2003) 497–507.DOI
  20. [20]K. Danas, S. V. Kankanala, N. Triantafyllidis, Experiments and modeling of iron-particle-filled magnetorheological elastomers, Journal of the Mechanics and Physics of Solids 60 (2012) 120–138.
  21. [21]E. G. Loukaides, S. K. Smoukov, K. A. Seffen, Magnetic actuation and transition shapes of a bistable spherical cap, International Journal of Smart and Nano Materials 5 (2015) 270–282.
  22. [22]K. A. Seffen, S. Vidoli, Eversion of bistable shells under magnetic actuation: a model of nonlinear shapes, Smart Materials and Structures 25 (2016) 065010.DOI
  23. [23]M. M. Schmauch, S. R. Mishra, B. A. Evans, O. D. Velev, J. B. Tracy, Chained Iron Microparticles for Directionally Controlled Actuation of Soft Robots, ACS Applied Materials & Interfaces 9 (2017) 11895–11901.DOI
  24. [24]J. Ciambella, A. Favata, G. Tomassetti, A nonlinear theory of fibre-reinforced magneto-elastic rods, Proceedings of the Royal Society of London A: Mathematical, Physical and Engineering Sciences 474 (2018) 20170703.DOI
  25. [25]R. P. Singh, P. R. Onck, Magnetic field induced deformation and buckling of slender bodies, International Journal of Solids and Structures 143 (2018) 29–58.
  26. [26]E. Psarra, L. Bodelot, K. Danas, Wrinkling to crinkling transitions and curvature localization in a magnetoelastic film bonded to a non-magnetic substrate, Journal of the Mechanics and Physics of Solids 133 (2019) 103734.DOI
  27. [27]Y. Alapan, A. C. Karacakol, S. N. Gultehan, I. Isik, M. Sitti, Reprogrammable shape morphing of magnetic soft machines, Science Advances 6 (2020) eabc6414.DOI
  28. [28]G. Z. Lum, Z. Ye, X. Dong, H. Marvi, O. Erin, W. Hu, M. Sitti, Shape-programmable magnetic soft matter, Proceedings of the National Academy of Sciences 113 (2016) E6007–E6015.pubmed.ncbi.nlm.nih.gov/27671658
  29. [29]Y. Kim, H. Yuk, R. Zhao, S. A. Chester, X. Zhao, Printing ferromagnetic domains for untethered fast-transforming soft materials, Nature 558 (2018) 274–279.DOI
  30. [30]R. Zhao, Y. Kim, S. A. Chester, P. Sharma, X. Zhao, Mechanics of hard-magnetic soft materials, Journal of the Mechanics and Physics of Solids 124 (2019) 244–263.
  31. [31]L. Wang, Y. Kim, C. F. Guo, X. Zhao, Hard-magnetic elastica, Journal of the Mechanics and Physics of Solids 142 (2020) 104045.DOI
  32. [32]J. Ciambella, G. Tomasetti, A form-finding strategy for magneto-elastic actuators, International Journal of Non-Linear Mechanics 119 (2020) 103297.DOI
  33. [33]T. Chen, M. Chelly, P. M. Reis, A reprogrammable mechanical metamaterial with stable memory, Nature 589 (2021) 386–390.DOI
  34. [34]D. Yan, M. Pezzulla, L. Cruveiller, A. Abbasi, P. M. Reis, Magneto-active elastic shells with tunable buckling strength, Nature Communications 12 (2021) 1–9.DOI
  35. [35]D. Yan, A. Abbasi, P. M. Reis, A comprehensive framework for hard-magnetic beams: reduced-order theory, 3D simulations, and experiments, arXiv:2106.14878[cond-mat.soft] (2021).arxiv.org/abs/2106.14878
  36. [36]E. Diller, J. Zhuang, G. Z. Lum, M. R. Edwards, M. Sitti, Continuously distributed magnetization profile for millimeter-scale elastomeric undulatory swimming, Applied Physics Letters 104 (2014) 174101.DOI
  37. [37]N. Hu, L. Wang, W. Zhai, M. Sun, H. Xie, Z. Wu, Q. He, Mechanically Actuated Rolling of Star-Shaped Hydrogel Microswimmer, Macromolecular Chemistry and Physics 219 (2018) 1700540.DOI
  38. [38]J. Zhang, E. Diller, Untethered Miniature Soft Robots: Modeling and Design of a Millimeter-Scale Swimming Magnetic Sheet, Soft Robotics 5 (2018) 761–776.DOI
  39. [39]F. Tsumori, A. Saijou, T. Osada, H. Miura, Development of actuation system for artificial cilia with magnetic elastomer, Japanese Journal of Applied Physics 54 (2015) 06FP12.DOI
  40. [40]F. Pece, J. J. Zarate, V. Vechev, N. Besse, O. Gudozhnik, H. Shea, O. Hilliges, MagTics, in: UIST ’17: The 30th Annual ACM Symposium on User Interface Software and Technology, ACM, New York, NY, USA, 2017, pp. 143–154.DOI
  41. [41]C. Y. Wang, A critical review of the heavy elastica, International Journal of Mechanical Sciences 28 (1986) 549–559.DOI
  42. [42]E. H. Dill, Kirchhoff’s theory of rods, Archive for History of Exact Sciences 44 (1992) 1–23.DOI
  43. [43]B. Audoly, Y. Pomeau, Elasticity and Geometry, Oxford Univ. Press, 2010.
  44. [44]L. D. Landau, E. M. Lifshitz, Theory of Elasticity, Pergamon Press, 1980.
  45. [45]T. R. Powers, Dynamics of filaments and membranes in a viscous fluid, Reviews of Modern Physics 82 (2010) 1607–1631.DOI
  46. [46]D. Bigoni, Extremely Deformable Structures (CISM Lecture Notes No. 562), Springer-Verlag, 2015.
  47. [47]A. Goriely, The Mathematics and Mechanics of Biological Growth, Springer-Verlag, 2017.DOI
  48. [48]G. Chirico, J. Langowski, Kinetics of DNA supercoiling studied by Brownian dynamics simulation, Biopolymers 34 (1994) 415–433.DOI
  49. [49]H. Wada, R. R. Netz, Stretching helical nano-springs at finite temperature, Europhysics Letters 77 (2007) 68001.DOI
  50. [50]H. Wada, R. R. Netz, Model for Self-Propulsive Helical Filaments: Kink-Pair Propagation, Physical Review Letters 99 (2007) 108102.pubmed.ncbi.nlm.nih.gov/17930410
  51. [51]R. Vogel, H. Stark, Force-extension curves of bacterial flagella, The European Physical Journal E 33 (2010) 259–271.DOI
  52. [52]M. Reichert, Hydrodynamic Interactions in Colloidal and Biological Systems, Ph.D. thesis, University Konstanz, 2006.
  53. [53]M. Bergou, M. Wardetzky, S. Robinson, B. Audoly, E. Grinspun, Discrete elastic rods, in: ACM SIGGRAPH 2008 papers, ACM Press, New York, New York, USA, 2008, p. 1.
  54. [54]Y. Morigaki, H. Wada, Y. Tanaka, Stretching an Elastic Loop: Crease, Helicoid, and Pop Out, Physical Review Letters 117 (2016) 198003.pubmed.ncbi.nlm.nih.gov/27858442
  55. [55]T. G. Sano, T. Yamaguchi, H. Wada, Slip Morphology of Elastic Strips on Frictional Rigid Substrates, Physical Review Letters 118 (2017) 178001.DOI
  56. [56]T. G. Sano, H. Wada, Snap-buckling in asymmetrically constrained elastic strips, Physical Review E 97 (2018) 013002.DOI
  57. [57]T. G. Sano, H. Wada, Twist-Induced Snapping in a Bent Elastic Rod and Ribbon, Physical Review Letters 122 (2019) 114301.DOI
  58. [58]J. D. Jackson, Classical electrodynamics; 2nd ed., Wiley, New York, NY, 1975.
  59. [59]D. Garcia-Gonzalez, M. Hossain, Microstructural modelling of hard-magnetic soft materials: Dipole–dipole interactions versus zeeman effect, Extreme Mechanics Letters 48 (2021) 101382.DOI
  60. [60]E. Doedel, AUTO: A program for the automatic bifurcation analysis of autonomous systems, Congressus Numerantium 30 (1981).
  61. [61]H. Goldstein, C. Poole Jr., J. Safko, Classical mechanics 3rd. Edition, Addison-Wesley, 2002.
  62. [62]E. Hairer, S. P. Nørsett, G. Wanner, Solving ordinary differential equations I: Nonstiff problems (2nd ed.), Springer Verlag, Berlin, 1993.
  63. [63]L. G. Brazier, On the flexure of thin cylindrical shells and other ”thin” sections, Proceedings of the Royal Society of London. Series A, Containing Papers of a Mathematical and Physical Character 116 (1927) 104–114.DOI
  64. [64]U. J. Counto, The effect of the elastic modulus of the aggregate on the elastic modulus, creep and creep recovery of concrete, Magazine of Concrete Research 16 (1964) 129–138.
  65. [65]A. Goriely, M. Tabor, Nonlinear dynamics of filaments. III. Instabilities of helical rods, Proceedings of the Royal Society of London A: Mathematical, Physical and Engineering Sciences 453 (1997) 2583–2601.DOI
  66. [66]S. H. Strogatz, Nonlinear Dynamics and Chaos: With Applications to Physics, Biology, Chemistry, and Engineering, 2nd Ed., CRC Press, Boca Raton, FL, 2018.

Paper details

vlnc:W_fmb7c6ds
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

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