betaFeedback

WorksPhysics and Astronomy

The magnetic field does not perform work – or does it?

Abstract

We investigate the question discussed in the literature as to whether the magnetic field can perform work using two models that describe interacting magnetic dipoles. In the first model, the dipoles are realized by rigidly rotating charge clouds, whereas in the second model, one of the two dipoles is described by a real macroscopic spin density. The theoretical foundations of the second model are formulated in a recently published paper. We obtain equations of motion and detailed energy balance equations for both cases, but the answer to the title question depends on the choice of criteria for ``magnetic work".

Introduction

According to a widespread opinion it holds that “the magnetic field does no work, since the magnetic force is perpendicular to the velocity,” see, e. g., [1], section 6.7. On the other hand, you can accelerate a magnet through the field of another magnet, and do work in the process. Isn’t that a contradiction?

This problem has been discussed frequently in the literature, see, e. g., [2, 3, 4, 5, 6] and on internet platforms. Recently it has been taken up again by J. A. Barandes [7, 8], who considers an electromagnetic field coupled to a relativistic charged particle with spin and derives expressions for the force on this particle which include the magnetic field. Hence he answers the question ”Can the magnetic field do work?” in the affirmative. Our work can be seen as a complementary contribution to the discussion in [7, 8], as we will elaborate below.

The mentioned paradox refers to the energy exchange between the electromagnetic field and matter. If we want to analyze it, it is inevitable to include theoretical descriptions of matter besides Maxwell’s equations for the electromagnetic field, even if one restricts oneself to strong simplifications. Electrodynamics knows only the reduced description of matter in terms of charges and currents (plus possibly spin densities, see below), but has no possibility to represent interactions and energy flows within matter. For a theoretical description of matter, in principle, both classical or relativistic mechanics and quantum mechanics come into question.

But first we want to reduce the paradox to its essential core. The situation where two magnets attract (or repel) each other and perform work in the process shall be described simplistically by the interaction of two magnetic dipoles, a “small” and a “large” one. Suitable assumptions are made to ensure that the large dipole retains its position and magnetic moment, so that its magnetic field remains constant over time and can be considered as the “external” field. The restriction to temporally constant external fields makes sense, as time-variable magnetic fields are associated with electric fields, which generally lead to a flow of energy between the field and matter. In this sense, time-dependent magnetic fields perform work, which is completely undisputed in the above-mentioned debate. The question there is rather whether time-constant magnetic fields perform work.

As a further simplification, let us refrain from considering fully relativistic equations including radiation effects caused by accelerated charges. The force effects contributing to the apparent paradox can already be calculated non-relativistically and quasi-statically to a good approximation. Correspondingly, we will neglect relativistic effects in the motion of the small dipole and adopt a description in terms of classical mechanics.

There are two possibilities for the theoretical description of the two magnetic dipoles, which now stand for “matter”. In the approach commonly used in electrodynamics, the magnetic moments are realized by current distributions (model A, or “Ampère model”), see § 2, The well-known equation for the force on a magnetic dipole in a weakly inhomogeneous magnetic field can be reproduced if we assume the current distribution of the small dipole to be sharply localized, see subsection 2.2. In order to obtain the simplest possible description in terms of classical mechanics, we will realize the dipoles by rigidly rotating charged clouds with, say, negative charge and a rotationally invariant Gaussian density. Both magnetic moments are initially, and hence for all times, oriented parallel to the zz-axis, see Figure 1. Since we do not want to consider any disturbing effects caused by the electric fields of the charged clouds, we additionally assume rigid clouds with the opposite, positive charge density, which have the same center of mass as the negatively charged clouds and initially do not rotate. The center of mass of the two clouds representing the small dipole will move into zz-direction under the influence of the force exerted by the magnetic field of the large dipole, see subsection 2.1. In summary, we therefore consider four clouds that realize two electrically neutral dipoles in pairs.

Sketch of two magnetic dipoles realized by rotating clouds

Figure 1. Sketch of two magnetic dipoles realized by rotating clouds, a “large” one (red color) and a “small” one (green color), with opposite magnetic moments oriented parallel to the vertical zz-axis.

However, this description is unrealistic when dealing with magnetic dipoles that are physically caused in whole or in part by spin densities. Therefore, we will treat this case separately, see section 3, relying on a recently published proposal [9] to supplement the classical Maxwell equations with spin densities as possible sources of magnetic fields (model B). However, it would not be appropriate to describe also the large dipole by a real spin density. In general, both magnetic moments would precess around each other and the magnetic field of the large dipole could not be kept constant in time. This can be illustrated by studying the interaction of two magnetic dipoles fixed in space, see [10]. Therefore, we will describe the large dipole by two clouds, one of which is rotating, as in model A, ensuring that the position and angular velocities of these clouds remain constant in time, see subsection 3.1.

Only the small dipole is represented by a rigid, electrically neutral cloud with a genuine magnetic moment density. Its center of mass can move in the zz-direction, but it turns out that the cloud will not rotate. Also in model B the small dipole will be extended and not considered as a point dipole, which has the advantage that its field energy will be finite.

Therefore, our treatment of the problem is complementary to the discussion in [8], since we (1) consider only a case study and not a complete theory, (2) restrict ourselves to non-relativistical calculations and (3) consider extended dipoles with finite field energy.

Interestingly, the analysis of the paradox is different in the two cases mentioned. In both cases, we derive equations of motion (eom) for the position of the small dipole, and, in case A, for the angular velocities of its two clouds, see subsection 2.3 and Appendix § 7. For these eom and for the energy balance, it is necessary to take into account the electric field E\mathbf{E} induced by the temporal change of the magnetic field of the small dipole. If the two magnetic dipoles are described by rigidly rotating charged clouds, the argument cited at the beginning comes into full effect, namely that the Lorentz force is perpendicular to the velocity of each charge element and therefore cannot perform any work. However, this does not apply to the E\mathbf{E} field, which causes an exchange of energy between the two dipoles and the field. It can be proven that the sum of the field energy and the total kinetic energy of the mechanical system remains constant, see subsection 2.4. Surprisingly, the kinetic rotational energy of the large dipole is variable and its change must also be taken into account. Nevertheless, if the mass (and thus the moment of inertia) of the large dipole is chosen to be very large compared to the mass of the small dipole, this result is compatible with the assumption of a constant external field.

If, on the other hand, we are dealing with “real” magnetic dipoles, a mechanical description using a classical spin density and a so-called Zeeman term, which acts like a potential energy, is appropriate, see the subsections 3.1 and 3.2. The Zeeman term is not ad hoc, but can be justified, for example, by the minimal coupling of a Dirac spinor to a classical electromagnetic field, see [11]. In this case, case B, there is an energy exchange between kinetic translational energy and Zeeman energy without tapping the field energy. On the other hand, the field energy is not conserved but increases at the expense of the kinetic rotational energy of the large dipole, similar to model A, see the subsection 3.3 and Appendix § 6. We emphasize that this result is restricted to the case of (anti-)parallel magnetic moments of the two dipoles and may be more complicated in the general case where the moment of the small dipole precesses around the external magnetic field.

This results in a, perhaps unexpected, answer to the title question “The magnetic field does not perform work - or does it?”: It depends, the question must be made more precise. In case A, there is an exchange of magnetic field energy and kinetic energy of both dipoles. This could be described as “magnetic work”; however, the energy exchange is mediated by the electric field. In case B, the acceleration of the small dipole is associated with a decrease in Zeeman energy, which in turn depends on the external magnetic field. In addition, magnetic field energy and rotational energy of the large dipole are converted into each other. But here, too, the total magnetic field is time-dependent and the energy flow is mediated by the induced electric field. In any case, we can state that the energetic relations and the motion of the small dipole strongly depend on whether we use model A or B. This will be outlined in the Summary.

Model A

General notations and assumptions

As explained in the Introduction, we model each of the two dipoles by two rigidly rotating clouds with opposite charge densities. These four clouds are distinguished by a superscript (0)(0) for the small dipole or (1)(1) for the large dipole and additionally, by a subscript ±\pm, which refers to the total positive/negative charge of the cloud. The center of mass of either dipole will be denoted by r0(i)\mathbf{r}_0^{(i)}, i=0,1i=0,1, and can be chosen of the special form

r0(1)=0andr0(0)=(0,0,z0(t))such thatz0(0)=a>0(1)\mathbf{r}_0^{(1)}=\mathbf{0} \qquad\text{and} \qquad\mathbf{r}_0^{(0)}=(0,0,z_0(t)) \quad\text{such that} \quad z_0(0)=a>0 \tag*{(1)}

anticipating the result that the motion of the small dipole is restricted to the zz-axis. The various densities are proportional to a normalized, rotationally invariant, Gaussian density

ρ(r)=α3π3/2e−α2r2(2)\rho(r)=\frac{\alpha^3}{\pi^{3/2}}e^{-\alpha^2r^2} \tag*{(2)}

depending on a parameter α>0\alpha>0. For the radius rr we have to insert

r=∣r−r0(i)(t)∣(3)r=\left|\mathbf{r}-\mathbf{r}_0^{(i)}(t)\right| \tag*{(3)}

depending on which cloud is considered. If necessary, we distinguish between the two dipoles by writing ρ(0)\rho^{(0)} or ρ(1)\rho^{(1)}. For later use we note

∇ρ=−2α2rρ.(4)\nabla\rho=-2\alpha^2r\rho. \tag*{(4)}

The four clouds are rotating with angular velocity ω±(i)(t)\omega_{\pm}^{(i)}(t) around the zz-axis, where for the large dipole we set

ω+(1)(t)=0andω−(1)(t)=const.(5)\omega_+^{(1)}(t)=0 \qquad\text{and} \qquad\omega_-^{(1)}(t)=\mathrm{const}. \tag*{(5)}

Moreover, the clouds will have mass densities

M±(i)ρ(∣r−r0(i)(t)∣),(6)M_{\pm}^{(i)}\rho\left(\left|\mathbf{r}-\mathbf{r}_0^{(i)}(t)\right|\right), \tag*{(6)}

charge densities

e±(i)ρ(∣r−r0(i)(t)∣)satisfyinge−(i)=−e+(i)<0,(7)e_{\pm}^{(i)}\rho\left(\left|\mathbf{r}-\mathbf{r}_0^{(i)}(t)\right|\right) \qquad\text{satisfying} \qquad e_-^{(i)}=-e_+^{(i)}<0, \tag*{(7)}

and magnetic moment densities

μ±(i)=(00m±(i)(t))ρ(∣r−r0(i)(t)∣)(8)\boldsymbol{\mu}_{\pm}^{(i)} = \begin{pmatrix} 0 \\ 0 \\ m_{\pm}^{(i)}(t) \end{pmatrix} \rho\left(\left|\mathbf{r}-\mathbf{r}_{0}^{(i)}(t)\right|\right) \tag*{(8)}
=ϵ±(i)2α2(00ω±(i)(t))ρ(∣r−r0(i)(t)∣).(9)= \frac{\epsilon_{\pm}^{(i)}}{2\alpha^{2}} \begin{pmatrix} 0 \\ 0 \\ \omega_{\pm}^{(i)}(t) \end{pmatrix} \rho\left(\left|\mathbf{r}-\mathbf{r}_{0}^{(i)}(t)\right|\right). \tag*{(9)}

The total magnetic dipole moments of the respective dipoles are designated as:

m(i)=m+(i)+m−(i)=(00m(i)(t)),for i=0,1.(10)\mathbf{m}^{(i)}=\mathbf{m}_{+}^{(i)}+\mathbf{m}_{-}^{(i)}= \begin{pmatrix} 0 \\ 0 \\ m^{(i)}(t) \end{pmatrix}, \qquad\text{for } i=0,1. \tag*{(10)}

Recall that m(1)(t)=const.m^{(1)}(t)=\text{const.} To simplify the discussion we choose the case where the two dipoles repel, say,

m(1)>0andm(0)<0.(11)m^{(1)}>0 \qquad\text{and} \qquad m^{(0)}<0. \tag*{(11)}

Force and torque

The magnetic field generated by a cloud with charge density ϵρ(η)\epsilon\rho(\eta), its center of mass located at r0=0\mathbf{r}_{0}=0, and rotating with angular velocity ω=(0,0,ω(t))\boldsymbol{\omega}=(0,0,\omega(t)) is given by

B=ϵμ0ω(t)4π3/2αr4e−α2r2B(a)+ϵμ0ω(t)8παr5erf⁡(αr)B(b),(12)\mathbf{B}= \frac{\epsilon\mu_{0}\omega(t)}{4\pi^{3/2}\alpha r^{4}}e^{-\alpha^{2}r^{2}}\mathbf{B}^{(a)} + \frac{\epsilon\mu_{0}\omega(t)}{8\pi\alpha r^{5}}\operatorname{erf}(\alpha r)\mathbf{B}^{(b)}, \tag*{(12)}

where

B(a)=(−xz(2α2r2+3)−yz(2α2r2+3)2r2(α2(x2+y2)−1)+3(x2+y2)),(13)\mathbf{B}^{(a)}= \begin{pmatrix} -xz\left(2\alpha^{2}r^{2}+3\right) \\ -yz\left(2\alpha^{2}r^{2}+3\right) \\ 2r^{2}\left(\alpha^{2}\left(x^{2}+y^{2}\right)-1\right)+3\left(x^{2}+y^{2}\right) \end{pmatrix}, \tag*{(13)}
B(b)=(3xz3yz3z2−r2).(14)\mathbf{B}^{(b)}= \begin{pmatrix} 3xz \\ 3yz \\ 3z^{2}-r^{2} \end{pmatrix}. \tag*{(14)}

We have derived (12–14) by different methods, one being described in [12], and have confirmed it by checking the equations ∇⋅B=0\nabla\cdot\mathbf{B}=0 and ∇×B=μ0j=μ0ϵρ ω×r\nabla\times\mathbf{B}=\mu_{0}\mathbf{j}=\mu_{0}\epsilon\rho\,\boldsymbol{\omega}\times\mathbf{r}. This result will be used to calculate the magnetic field B(1):=B−(1)\mathbf{B}^{(1)}:=\mathbf{B}_{-}^{(1)} generated by the large dipole and B(0):=B+(0)+B−(0)\mathbf{B}^{(0)}:=\mathbf{B}_{+}^{(0)}+\mathbf{B}_{-}^{(0)} generated by the two clouds of the small dipole.

For large distances from the center of mass of the rotating cloud, i. e., for αr→∞\alpha r\to\infty we have e−α2r2→0e^{-\alpha^{2}r^{2}}\to0 and erf⁡(αr)→1\operatorname{erf}(\alpha r)\to1. Then (12) asymptotically assumes the form of the field of a magnetic point dipole. This asymptotic form will always be used when calculating the force or the torque exerted from the large dipole upon the small one. Moreover, for these calculations we will use the linear approximation of B(1)\mathbf{B}^{(1)} at the position r0(0)(t)\mathbf{r}_{0}^{(0)}(t) of the small dipole.

The result of the action of the external field B(1)\mathbf{B}^{(1)} on the small dipole is, firstly, a total force given, in accordance with the textbook equation, by

F(0)=∇(m(0)⋅B(1))=(0,0,−3Bm(0)a+z0(t)),(15)\mathbf{F}^{(0)} = \nabla\left(\mathbf{m}^{(0)} \cdot\mathbf{B}^{(1)}\right) = \left(0,0,-\frac{3 B m^{(0)}}{a+z_0(t)}\right), \tag*{(15)}

where BB is the zz-component of B(1)\mathbf{B}^{(1)} in the point dipole approximation at the position of the small dipole, i. e.,

B=μ0m(1)2π(a+z0(t))3>0.(16)B = \frac{\mu_0 m^{(1)}}{2\pi\left(a+z_0(t)\right)^3} > 0 . \tag*{(16)}

In this context it is remarkable that T. H. Boyer derives two different expressions for the force (15), depending on whether one chooses the electric current model or the separated magnetic charge model [13], see also [14]. However, if ∇×B(1)=0\nabla\times\mathbf{B}^{(1)} = 0, which is approximately satisfied in our case, both expressions coincide.

Secondly, the torque N(10)\mathbf{N}^{(10)} exerted by B(1)\mathbf{B}^{(1)} on the small dipole is calculated as

N(10)=(0,0,3Bm(0)z˙0(t)a+z0(t)),(17)\mathbf{N}^{(10)} = \left(0,0,\frac{3 B m^{(0)} \dot{z}_0(t)}{a+z_0(t)}\right), \tag*{(17)}

This torque leads to a slowdown in the rotation of the negatively charged cloud and thus to a reduction in its magnetic moment. For a qualitative explanation of these results see Figure 2.

Image not extracted

Figure 2. Left panel: Force dF=de v×B\mathrm{d}\mathbf{F} = \mathrm{d}e\,\mathbf{v} \times\mathbf{B} on a charge element de<0\mathrm{d}e < 0 rotating anticlockwise on the blue circle about the zz-axis (dashed black line). The initial velocity v\mathbf{v} is tangent to the blue circle and the force dF\mathrm{d}\mathbf{F} has a constant zz-component. Right panel: Force dF′=de v′×B\mathrm{d}\mathbf{F}' = \mathrm{d}e\,\mathbf{v}' \times\mathbf{B} on a charge element de<0\mathrm{d}e < 0, with the constant component v′\mathbf{v}' of the velocity into zz-direction. Consequently, dF′\mathrm{d}\mathbf{F}' represents the constant tangential component of dF\mathrm{d}\mathbf{F}.

The induced electric vortex fields E±(0)\mathbf{E}^{(0)}_{\pm} are obtained as E±(0)=−∂∂tA±(0)\mathbf{E}^{(0)}_{\pm}=-\frac{\partial}{\partial t}\mathbf{A}^{(0)}_{\pm}, where A±(0)\mathbf{A}^{(0)}_{\pm} is the vector potential of B±(0)\mathbf{B}^{(0)}_{\pm}, the latter calculated according to (12–14). Note that E±(1)=0\mathbf{E}^{(1)}_{\pm}=0 since B±(1)\mathbf{B}^{(1)}_{\pm} is assumed to be constant in time. The final result reads

E±(0)=ω±(0)(t)(z−z0(t))z˙0(t)U+ω˙±(0)(t)V,(18)\mathbf{E}^{(0)}_{\pm}=\omega^{(0)}_{\pm}(t)\left(z-z_0(t)\right)\dot{z}_0(t)\mathbf{U}+\dot{\omega}^{(0)}_{\pm}(t)\mathbf{V}, \tag*{(18)}

where

U=(y(3πerf⁡(αw)−2αwe−α2w2(2α2w2+3))−x(3πerf⁡(αw)−2αwe−α2w2(2α2w2+3))0),(19)\mathbf{U}=\begin{pmatrix} y\left(3\sqrt{\pi}\operatorname{erf}(\alpha w)-2\alpha w e^{-\alpha^2w^2}\left(2\alpha^2w^2+3\right)\right)\\ -x\left(3\sqrt{\pi}\operatorname{erf}(\alpha w)-2\alpha w e^{-\alpha^2w^2}\left(2\alpha^2w^2+3\right)\right)\\ 0 \end{pmatrix}, \tag*{(19)}
V=(y(w2(πerf⁡(αw)−2αwe−α2w2))−x(w2(πerf⁡(αw)−2αwe−α2w2))0),(20)\mathbf{V}=\begin{pmatrix} y\left(w^2\left(\sqrt{\pi}\operatorname{erf}(\alpha w)-2\alpha w e^{-\alpha^2w^2}\right)\right)\\ -x\left(w^2\left(\sqrt{\pi}\operatorname{erf}(\alpha w)-2\alpha w e^{-\alpha^2w^2}\right)\right)\\ 0 \end{pmatrix}, \tag*{(20)}

and

w=r2+z0(t)2−2zz0(t).(21)w=\sqrt{r^2+z_0(t)^2-2zz_0(t)}. \tag*{(21)}

The total force exerted by the electric fields E±(0)\mathbf{E}^{(0)}_{\pm} on the two clouds of the small dipole vanishes, but there will appear some non-zero torque N±(00)\mathbf{N}^{(00)}_{\pm} that leads to a change of ω±(0)\omega^{(0)}_{\pm} according to the equation N=θω˙\mathbf{N}=\theta\dot{\boldsymbol{\omega}} and using the values of the inertial moments

θ±(0)=M±(0)α2.(22)\theta^{(0)}_{\pm}=\frac{M^{(0)}_{\pm}}{\alpha^2}. \tag*{(22)}

of the two clouds representing the small dipole.

Altogether, the above considerations result in a coupled system of differential equations for the motion of the magnetic dipole, where the variables are z0(t)z_0(t) and ω±(0)(t)\omega^{(0)}_{\pm}(t). This system represents a genuine interaction between matter and field. In most physical applications, only a one-sided action is considered, either the action of the given electromagnetic fields on the charges and currents, or the generation of electromagnetic fields by given charges and currents. In our case, the two charged clouds move in the external magnetic field and their movement generates an electric field which acts back on them, and so on.

Equation of motion

It is advisable to formulate the equation of motion (eom) in dimensionless variables. For this purpose, we introduce the following problem-related physical units.

  • Length: aa (initial distance between the large and the small dipole)

  • Mass: M−(0)M^{(0)}_{-} (the mass of the negative cloud of the small dipole)

  • Charge: e=∣e−(0)∣e=\lvert e^{(0)}_{-}\rvert (the absolute charge of the negative cloud of the small dipole)

  • Time: 1/ω11/\omega_1 (inverse Larmor frequency).

The Larmor frequency ω1\omega_1 is given by

ω1=eBM−(0)=eμ0m(1)2πa3M−(0).(23)\omega_1=\frac{eB}{M^{(0)}_{-}}=\frac{e\mu_0m^{(1)}}{2\pi a^3M^{(0)}_{-}}. \tag*{(23)}

Further we can form the following dimensionless ratios:

ε:=1αa(24)\varepsilon:= \frac{1}{\alpha a} \tag*{(24)}
μ1:=M−(0)M+(0)(25)\mu_1 := \frac{M_-^{(0)}}{M_+^{(0)}} \tag*{(25)}
μ2:=M−(0)e2μ0α(26)\mu_2 := \frac{M_-^{(0)}}{e^2\mu_0\alpha} \tag*{(26)}
γ1:=m(0)m(1)(27)\gamma_1 := \frac{m^{(0)}}{m^{(1)}} \tag*{(27)}
ω0ω1=4πγ1μ2ε3(28)\frac{\omega_0}{\omega_1} = \frac{4\pi\gamma_1\mu_2}{\varepsilon^3} \tag*{(28)}

In (28) we have introduced the initial angular velocity ω0:=ω−(0)(0)\omega_0 := \omega_-^{(0)}(0) of the negative cloud of the small dipole. Note that γ1<0\gamma_1 < 0 for the case of anti-parallel magnetic moments of both clouds as chosen in this paper. The eom for the dimensionless variables (denoted by the same letters and omitting the superscript (0)(0)) then read:

z¨0(t)=3μ1ε22(μ1+1)ω−(t)−ω+(t)(z0(t)+1)4(29)\ddot{z}_0(t) = \frac{3\mu_1\varepsilon^2}{2(\mu_1+1)}\frac{\omega_-(t)-\omega_+(t)}{(z_0(t)+1)^4} \tag*{(29)}
ω˙+(t)=36π3/2μ1μ22μ1+24π3/2μ2+2z˙0(t)(z0(t)+1)4(30)\dot{\omega}_+(t) = \frac{36\pi^{3/2}\mu_1\mu_2}{\sqrt{2}\mu_1+24\pi^{3/2}\mu_2+\sqrt{2}}\frac{\dot{z}_0(t)}{(z_0(t)+1)^4} \tag*{(30)}
ω˙−(t)=−36π3/2μ22μ1+24π3/2μ2+2z˙0(t)(z0(t)+1)4(31)\dot{\omega}_-(t) = -\frac{36\pi^{3/2}\mu_2}{\sqrt{2}\mu_1+24\pi^{3/2}\mu_2+\sqrt{2}}\frac{\dot{z}_0(t)}{(z_0(t)+1)^4} \tag*{(31)}

The equations (30) and (31) can be integrated, using the initial conditions ω+(0)=0\omega_+(0) = 0 and ω−(0)=ω0\omega_-(0) = \omega_0, and inserted into (29) with the result

z¨0(t)=3ε22(μ1+1)(z0(t)+1)7(ω0+[2(μ1+1)ω0−12π3/2(μ1−1)μ2]z0(t)(z0(t)(z0(t)+3)+3)2(μ1+1)+24π3/2μ2)(32)\ddot{z}_0(t) = \frac{3\varepsilon^2}{2(\mu_1+1)(z_0(t)+1)^7}\left(\omega_0+\frac{\left[\sqrt{2}(\mu_1+1)\omega_0-12\pi^{3/2}(\mu_1-1)\mu_2\right]z_0(t)\left(z_0(t)(z_0(t)+3)+3\right)}{\sqrt{2}(\mu_1+1)+24\pi^{3/2}\mu_2}\right) \tag*{(32)}

It is not analytically solvable but can be studied numerically.

Energy balance

As a first approach to the energy balance we consider the Poynting theorem in its quasi-static approximation:

ddt∫12μ0∣B∣2 dV=−∫j⋅E dV(33)\frac{d}{dt}\int\frac{1}{2\mu_0}|\mathbf{B}|^2\,dV=-\int\mathbf{j}\cdot\mathbf{E}\,dV \tag*{(33)}

which can be derived in the usual way from the quasi-static Maxwell equations

∇×E+∂∂tB=0(34)\nabla\times\mathbf{E}+\frac{\partial}{\partial t}\mathbf{B}=0 \tag*{(34)}
∇⋅B=0(35)\nabla\cdot\mathbf{B}=0 \tag*{(35)}
∇⋅E=0(36)\nabla\cdot\mathbf{E}=0 \tag*{(36)}
∇×B=μ0j(37)\nabla\times\mathbf{B}=\mu_0\mathbf{j} \tag*{(37)}

Here B=B(1)+B(0)\mathbf{B}=\mathbf{B}^{(1)}+\mathbf{B}^{(0)} and E=E+(0)+E−(0)\mathbf{E}=\mathbf{E}^{(0)}_{+}+\mathbf{E}^{(0)}_{-} as defined in subsection 2.2. j=j−(1)+j+(0)+j−(0)\mathbf{j}=\mathbf{j}^{(1)}_{-}+\mathbf{j}^{(0)}_{+}+\mathbf{j}^{(0)}_{-} comprises all sources of the total magnetic field B\mathbf{B}. Note that the induced electric field does not contribute to the field energy in this approximation.

The field energy can be decomposed into three terms according to

∫12μ0∣B(1)+B(0)∣2 dV(38)\int\frac{1}{2\mu_0}\left|\mathbf{B}^{(1)}+\mathbf{B}^{(0)}\right|^2\,dV \tag*{(38)}
=∫12μ0∣B(1)∣2 dV+∫1μ0B(1)⋅B(0) dV+∫12μ0∣B(0)∣2 dV(39)=\int\frac{1}{2\mu_0}\left|\mathbf{B}^{(1)}\right|^2\,dV+\int\frac{1}{\mu_0}\mathbf{B}^{(1)}\cdot\mathbf{B}^{(0)}\,dV+\int\frac{1}{2\mu_0}\left|\mathbf{B}^{(0)}\right|^2\,dV \tag*{(39)}
=:Efield11+Efield10+Efield00.(40)=:E_{\mathrm{field}}^{11}+E_{\mathrm{field}}^{10}+E_{\mathrm{field}}^{00}. \tag*{(40)}

The first term Efield11E_{\mathrm{field}}^{11} is constant and can be neglected when considering energy balance. For the time derivative of the second “interference term” we obtain

ddtEfield10=3ϵ2z˙0(t)2(2(μ1+1)+24π3/2μ2)(2(μ1+1)ω0(z0(t)+1)3−12π3/2μ2(−μ1+(μ1−2ω0+1)z0(t)(z0(t)(z0(t)+3)+3)−2ω0−1))(z0(t)+1)7.(41)\frac{d}{dt}E_{\mathrm{field}}^{10}= \frac{3\epsilon^2\dot{z}_0(t)} {2\left(\sqrt{2}(\mu_1+1)+24\pi^{3/2}\mu_2\right)} \frac{\left(\sqrt{2}(\mu_1+1)\omega_0(z_0(t)+1)^3-12\pi^{3/2}\mu_2\left(-\mu_1+(\mu_1-2\omega_0+1)z_0(t)\left(z_0(t)(z_0(t)+3)+3\right)-2\omega_0-1\right)\right)} {(z_0(t)+1)^7}. \tag*{(41)}

and, analogously,

ddtEfield00=−3(μ1+1)ϵ2z˙0(t)2(2(μ1+1)+24π3/2μ2)2(2(μ1+1)ω0(z0(t)+1)3+12π3/2μ2(2ω0−(μ1−2ω0+1)z0(t)(z0(t)(z0(t)+3)+3)))(z0(t)+1)7.(42)\frac{d}{dt}E_{\mathrm{field}}^{00}= -\frac{3(\mu_1+1)\epsilon^2\dot{z}_0(t)} {\sqrt{2}\left(\sqrt{2}(\mu_1+1)+24\pi^{3/2}\mu_2\right)^2} \frac{\left(\sqrt{2}(\mu_1+1)\omega_0(z_0(t)+1)^3+12\pi^{3/2}\mu_2\left(2\omega_0-(\mu_1-2\omega_0+1)z_0(t)\left(z_0(t)(z_0(t)+3)+3\right)\right)\right)} {(z_0(t)+1)^7}. \tag*{(42)}

Numerical examples show that Efield00E_{\mathrm{field}}^{00} decreases with time due to the weakening of the dipole moment m(0)\mathbf{m}^{(0)} but this is overcompensated by the increase of Efield10E_{\mathrm{field}}^{10}, see Figure 3.

Four quantities of the energy balance

Figure 3. Four quantities of the energy balance (46) are numerically calculated and shown as functions of tt. The total kinetic energy of the small dipole is slightly increasing (upper left panel) as well as the field energy Efield:=Efield10+Efield00E_{\mathrm{field}} := E_{\mathrm{field}}^{10} + E_{\mathrm{field}}^{00} (upper right panel). This is compensated by the decrease of rotational kinetic energy of the large dipole (lower left panel) such that the total energy E(t)E(t) is constant (lower right panel).

For the two parts Etrans(0)E_{\mathrm{trans}}^{(0)} and Erot(0)E_{\mathrm{rot}}^{(0)} of the kinetic energy of the small dipole we obtain

ddtEtrans(0)=3ϵ2z˙0(t)2(z0(t)+1)7(z0(t)(z0(t)(z0(t)+3)+3)(2(μ1+1)ω0−12π3/2μ2(μ1−2ω0+1))2(μ1+1)+24π3/2μ2+ω0).(43)\frac{d}{dt}E_{\mathrm{trans}}^{(0)} = \frac{3\epsilon^2\dot{z}_0(t)}{2(z_0(t)+1)^7} \left( \frac{z_0(t)\left(z_0(t)(z_0(t)+3)+3\right)\left(\sqrt{2}(\mu_1+1)\omega_0-12\pi^{3/2}\mu_2(\mu_1-2\omega_0+1)\right)} {\sqrt{2}(\mu_1+1)+24\pi^{3/2}\mu_2} +\omega_0 \right). \tag*{(43)}

and

ddtErot(0)=36π3/2μ2ϵ2z˙0(t)(2(μ1+1)+24π3/2μ2)2(−2(μ1+1)ω0(z0(t)+1)3−12π3/2μ2(2ω0−(μ1−2ω0+1)z0(t)((z0(t)+3)z0(t)+3)))(z0(t)+1)7.(44)\frac{d}{dt}E_{\mathrm{rot}}^{(0)} = \frac{36\pi^{3/2}\mu_2\epsilon^2\dot{z}_0(t)} {\left(\sqrt{2}(\mu_1+1)+24\pi^{3/2}\mu_2\right)^2} \frac{\left(-\sqrt{2}(\mu_1+1)\omega_0(z_0(t)+1)^3-12\pi^{3/2}\mu_2\left(2\omega_0-(\mu_1-2\omega_0+1)z_0(t)\left((z_0(t)+3)z_0(t)+3\right)\right)\right)} {(z_0(t)+1)^7}. \tag*{(44)}

Numerical examples and analytical considerations show that the sum of field energy and kinetic energy of the small dipole is not constant, see Figure 3. Surprisingly, also the rotational kinetic energy Erot(1)E_{\mathrm{rot}}^{(1)} of the large dipole has to be taken into account. We cannot calculate it directly, but its time derivative can be obtained by means of the torque N(01)\mathbf{N}^{(01)} exerted by E(0)\mathbf{E}^{(0)} upon the negatively charged cloud of the large dipole. For this we use the equation ddtErot=ddt12θω2=θωω˙=ωN\frac{d}{dt}E_{\mathrm{rot}}=\frac{d}{dt}\frac{1}{2}\theta\omega^2=\theta\omega\dot{\omega}=\omega N which also shows that the rotational energy of the positively charged cloud of the large dipole retains its initial value 0 since ω+(1)=0\omega^{(1)}_{+}=0. Thus we obtain

ddtErot(1)=3ε2z˙0(t)2(2(μ1+1)+24π3/2μ2)12π3/2μ2(−μ1+(μ1−2ω0+1)z0(t)(z0(t)(z0(t)+3)+3)−2ω0−1)−2(μ1+1)ω0(z0(t)+1)3(z0(t)+1)7.(45)\frac{d}{dt}E_{\mathrm{rot}}^{(1)} = \frac{3\varepsilon^{2}\dot{z}_{0}(t)} {2\left(\sqrt{2}(\mu_{1}+1)+24\pi^{3/2}\mu_{2}\right)} \frac{12\pi^{3/2}\mu_{2}\left(-\mu_{1}+(\mu_{1}-2\omega_{0}+1)z_{0}(t)\left(z_{0}(t)(z_{0}(t)+3)+3\right)-2\omega_{0}-1\right)-\sqrt{2}(\mu_{1}+1)\omega_{0}(z_{0}(t)+1)^{3}} {(z_{0}(t)+1)^{7}} . \tag*{(45)}

Finally, the following energy balance can be proven by computer-algebraic means:

Proposition 1 Under the preceding assumptions there holds

ddt(Efield10+Efield00+Etrans(0)+Erot(0)+Erot(1))=0.(46)\frac{d}{dt}\left(E_{\mathrm{field}}^{10}+E_{\mathrm{field}}^{00}+E_{\mathrm{trans}}^{(0)}+E_{\mathrm{rot}}^{(0)}+E_{\mathrm{rot}}^{(1)}\right)=0 . \tag*{(46)}

We add a remark as to whether the change in rotational energy of the large dipole is compatible with the assumption of a constant external field. Recall that ddtErot=ωN\frac{d}{dt}E_{\mathrm{rot}}=\omega N and ω˙=Nθ\dot{\omega}=\frac{N}{\theta}. If the inertial moment θ−(1)\theta_{-}^{(1)} is very large compared with θ−(0)\theta_{-}^{(0)} and θ+(0)\theta_{+}^{(0)} then a finite value of ddtErot(1)=ω−(1)N(01)\frac{d}{dt}E_{\mathrm{rot}}^{(1)}=\omega_{-}^{(1)}N^{(01)} will nevertheless lead to a negligible value of ω˙−(1)=N(01)θ−(1)\dot{\omega}_{-}^{(1)}=\frac{N^{(01)}}{\theta_{-}^{(1)}}. To this end it suffices to assume

M−(1)≫M−(0),M+(0),(47)M_{-}^{(1)} \gg M_{-}^{(0)}, M_{+}^{(0)} , \tag*{(47)}

thereby legitimizing talk of “large” and “small” dipole. This assumption also guarantees that the force exerted by the magnetic field of the small dipole on the large one (which has the same absolute value as the force exerted by the large dipole on the small one in accordance with the principle actio=reactioactio = reactio) does not set the center of mass of the large dipole in motion. Both arguments together mean that the external field generated by the large dipole will be constant to an excellent approximation.

In order to illustrate the above results we have numerically solved the eom for the choice of parameters ϵ=μ1=μ2=0.01\epsilon= \mu_1 = \mu_2 = 0.01 and ω0=ω1\omega_0 = \omega_1. Some interesting terms of the energy balance (46) are shown as functions of tt in Figure 3.

Model B

General notations and assumptions

In this section we model the small dipole by a rigid cloud carrying a genuine dipole density

μ=m(0)ρ(0)=(00m(0))ρ(0),(48)\boldsymbol{\mu} = \mathbf{m}^{(0)}\rho^{(0)} = \begin{pmatrix} 0 \\ 0 \\ m^{(0)} \end{pmatrix}\rho^{(0)}, \tag*{(48)}

with ρ(0)=ρ(∣r−r0(t)∣)\rho^{(0)} = \rho(\lvert\mathbf{r} - \mathbf{r}_0(t)\rvert) given by (2). It turns out that the cloud will not rotate but will move along the zz-axis. Hence the normalized density will be written as ρ(0)=ρ(∣(x,y,z−(a+z0(t)))∣)\rho^{(0)} = \rho(\lvert(x,y,z-(a+z_0(t)))\rvert), similarly as in Section 2. We largely adopt the notation of Section 2.

Such genuine magnetic dipole densities have been considered in [9] as additional sources of the electromagnetic field resulting in the following extended form of Maxwell’s equations:

∇×E+∂∂tB=0,(49)\nabla\times\mathbf{E} + \frac{\partial}{\partial t}\mathbf{B} = 0, \tag*{(49)}
∇⋅B=0,(50)\nabla\cdot\mathbf{B} = 0, \tag*{(50)}
∇⋅(ϵ0E+Q)=ρ,(51)\nabla\cdot(\epsilon_0\mathbf{E}+\mathbf{Q}) = \rho, \tag*{(51)}
1μ0∇×B−∂∂t(ϵ0E+Q)=j+∇×μ.(52)\frac{1}{\mu_0}\nabla\times\mathbf{B} - \frac{\partial}{\partial t}(\epsilon_0\mathbf{E}+\mathbf{Q}) = \mathbf{j}+\nabla\times\boldsymbol{\mu}. \tag*{(52)}

Here Q\mathbf{Q} is a kind of “quasi-electric polarization”, which must be added because of the relativistic covariance, but is not considered further in this paper.

Analogously to the usual electromagnetic theory a Poynting theorem can be derived from (49 – 52) that describes the energy flow between field and matter:

ddt∫(12μ0∣B∣2+ϵ02∣E∣2) dV=−∫(j+∇×μ+∂∂tQ)⋅E dV,(53)\frac{d}{dt}\int\left(\frac{1}{2\mu_0}\lvert\mathbf{B}\rvert^2+\frac{\epsilon_0}{2}\lvert\mathbf{E}\rvert^2\right)\,dV=-\int\left(\mathbf{j}+\nabla\times\boldsymbol{\mu}+\frac{\partial}{\partial t}\mathbf{Q}\right)\cdot\mathbf{E}\,dV, \tag*{(53)}

where we have assumed that all fields decay sufficiently fast at infinity thereby neglecting radiation effects.

In general, a local magnetic moment μdV\boldsymbol{\mu}dV will precess around the external magnetic field B(1)\mathbf{B}^{(1)}. This precession only does not occur on the zz-axis, as the direction of μ\boldsymbol{\mu} is parallel to B(1)\mathbf{B}^{(1)}. If one deviates slightly from the zz-axis, the direction of μdV\boldsymbol{\mu}dV is no longer parallel to B(1)\mathbf{B}^{(1)} and the precession generates a position- and time-dependent direction field m(0)(r,t)\mathbf{m}^{(0)}(\mathbf{r},t), which replaces the constant direction field m(0)\mathbf{m}^{(0)} in (48). This effect, at least in linear order w. r. t. r, should be taken into account for a complete description of the energy balance. However, since it has no influence on the basic results, we move the “inhomogeneity correction” to Appendix A and neglect this effect in the present section.

Force and torque

The large dipole is still realized by two massive rigid, charged clouds, so that only the negatively charged one rotates, as in § 2. As shown in [9] the force density f\mathbf{f} within the small dipole due to the external magnetic field B(1)\mathbf{B}^{(1)} is given by

f=(∇×μ)×B(1)(54)\mathbf{f} = (\nabla\times\boldsymbol{\mu}) \times\mathbf{B}^{(1)} \tag*{(54)}

and the integration of (54) over the whole space gives

F=(m(0)⋅∇)B(1)∣r=r0(55)\mathbf{F} = \left.(\mathbf{m}^{(0)} \cdot\nabla)\mathbf{B}^{(1)}\right|_{\mathbf{r}=\mathbf{r}_0} \tag*{(55)}
=(0,0,−3Bm(0)a+z0)=(0,0,−3m(0)m(1)μ02π(a+z0)4)(56)= \left(0,0,-\frac{3Bm^{(0)}}{a+z_0}\right)=\left(0,0,-\frac{3m^{(0)}m^{(1)}\mu_0}{2\pi(a+z_0)^4}\right) \tag*{(56)}

in accordance with (15) and eq. (37) of [9].

It can be shown that the total torque N(10)\mathbf{N}^{(10)} exerted by B(1)\mathbf{B}^{(1)} on the small dipole vanishes, i. e.,

N(10)=∫r×((∇×μ)×B(1)) dV=0.(57)\mathbf{N}^{(10)}=\int\mathbf{r}\times\left((\nabla\times\boldsymbol{\mu})\times\mathbf{B}^{(1)}\right)\,dV=0. \tag*{(57)}

Energy balance

The force (55) can be derived by F=−∇EZ\mathbf{F}=-\nabla E_Z from a potential energy EZE_Z, usually called “Zeeman energy”, given by

EZ(r0)=−m(0)⋅B(1)(r0)=−m(0)m(1)μ02π(a+z0)3.(58)E_Z(\mathbf{r}_0)=-\mathbf{m}^{(0)}\cdot\mathbf{B}^{(1)}(\mathbf{r}_0)=-\frac{m^{(0)}m^{(1)}\mu_0}{2\pi(a+z_0)^3}. \tag*{(58)}

To some extent it plays the role of the rotational kinetic energy of the small dipole in case A, but now there holds a strict energy conservation law of the form

ddt(Etrans(0)+EZ)=0,(59)\frac{d}{dt}\left(E_{\mathrm{trans}}^{(0)}+E_Z\right)=0, \tag*{(59)}

since m(0)\mathbf{m}^{(0)} is constant. Interestingly, the resulting equation of motion for the vertical coordinate z0(t)z_0(t) can be analytically solved, see Appendix B.

To complete the energy balance we have to consider the field energy. Of the three terms of (38–40) only the “interference term” Efield10E_{\mathrm{field}}^{10} is not constant in time. From the comparison with model A we expect that the increase of Efield10E_{\mathrm{field}}^{10} is balanced by a corresponding decrease of Erot(1)E_{\mathrm{rot}}^{(1)}. In fact, we can prove the following:

Proposition 2 Under the preceding conditions there holds:

ddt(Efield10+Erot(1))=0(60)\frac{d}{dt}\left(E_{\mathrm{field}}^{10}+E_{\mathrm{rot}}^{(1)}\right)=0 \tag*{(60)}

Proof: We only consider the case without inhomogeneity corrections and refer the reader to Appendix A for the remaining part of the proof. The magnetic field B(0)\mathbf{B}^{(0)} satisfies the quasi-static equations ∇⋅B(0)=0\nabla\cdot\mathbf{B}^{(0)} = 0 and ∇×B(0)=μ0∇×μ\nabla\times\mathbf{B}^{(0)} = \mu_0 \nabla\times\boldsymbol{\mu}, resulting from (50) and (52). It follows that

B(0)=μ0(μ+∇u),(61)\mathbf{B}^{(0)} = \mu_0 \left(\boldsymbol{\mu} + \nabla u\right), \tag*{(61)}

with a scalar field uu satisfying

Δu=−∇⋅μ,(62)\Delta u = -\nabla\cdot\boldsymbol{\mu}, \tag*{(62)}

and vanishing at infinity. Then

Efield10=1μ0∫B(1)⋅B(0) dV(63)E_{\mathrm{field}}^{10} = \frac{1}{\mu_0} \int\mathbf{B}^{(1)} \cdot\mathbf{B}^{(0)} \, dV \tag*{(63)}
=(61)∫B(1)⋅(μ+∇u) dV(64)\overset{(61)}{=} \int\mathbf{B}^{(1)} \cdot\left(\boldsymbol{\mu} + \nabla u\right) \, dV \tag*{(64)}
=∫B(1)⋅μ dV+∫B(1)⋅(∇u) dV(65)= \int\mathbf{B}^{(1)} \cdot\boldsymbol{\mu} \, dV + \int\mathbf{B}^{(1)} \cdot(\nabla u) \, dV \tag*{(65)}
=∫B(1)⋅μ dV+∫(∇⋅(B(1)u)−u(∇⋅B(1))) dV(66)= \int\mathbf{B}^{(1)} \cdot\boldsymbol{\mu} \, dV + \int\left(\nabla\cdot\left(\mathbf{B}^{(1)}u\right) - u\left(\nabla\cdot\mathbf{B}^{(1)}\right)\right) \, dV \tag*{(66)}
=∫B(1)⋅μ dV(67)= \int\mathbf{B}^{(1)} \cdot\boldsymbol{\mu} \, dV \tag*{(67)}
=(48)m(0)⋅∫B(1)ρ(0) dV(68)\overset{(48)}{=} \mathbf{m}^{(0)} \cdot\int\mathbf{B}^{(1)} \rho^{(0)} \, dV \tag*{(68)}
=m(0)⋅B(1)∣r=r0,(69)= \mathbf{m}^{(0)} \cdot\mathbf{B}^{(1)}\big|_{\mathbf{r}=\mathbf{r}_0}, \tag*{(69)}

using ∇⋅B(1)=0\nabla\cdot\mathbf{B}^{(1)} = 0 and Gauss’ integral theorem in (67) as well as the point dipole approximation in (69). Interestingly, Efield10E_{\mathrm{field}}^{10} equals the negative Zeeman energy (58) which shows that Efield10E_{\mathrm{field}}^{10} also increases if the small dipole is accelerated and its kinetic energy increases, similarly as in the model A.

In order to evaluate

ddtErot(1)=ω(1)⋅N(01)(70)\frac{d}{dt} E_{\mathrm{rot}}^{(1)} = \boldsymbol{\omega}^{(1)} \cdot\mathbf{N}^{(01)} \tag*{(70)}

we consider

N(01)=∫r×(e−(1)ρ(1)E(0)) dV(71)\mathbf{N}^{(01)} = \int\mathbf{r} \times\left(e_-^{(1)} \rho^{(1)} \mathbf{E}^{(0)}\right) \, dV \tag*{(71)}
=(4)−e−(1)2α2∫∇ρ(1)×E(0) dV(72)\overset{(4)}{=} -\frac{e_-^{(1)}}{2\alpha^2} \int\nabla\rho^{(1)} \times\mathbf{E}^{(0)} \, dV \tag*{(72)}
=−e−(1)2α2∫(∇×(ρ(1)E(0))−(∇×E(0))ρ(1)) dV(73)= -\frac{e_-^{(1)}}{2\alpha^2} \int\left(\nabla\times\left(\rho^{(1)} \mathbf{E}^{(0)}\right) - \left(\nabla\times\mathbf{E}^{(0)}\right)\rho^{(1)}\right) \, dV \tag*{(73)}
=e−(1)2α2∫(∇×E(0))ρ(1) dV(74)= \frac{e_-^{(1)}}{2\alpha^2} \int\left(\nabla\times\mathbf{E}^{(0)}\right)\rho^{(1)} \, dV \tag*{(74)}
=(49)−e−(1)2α2∫(ddtB(0))ρ(1) dV(75)\overset{(49)}{=} -\frac{e_-^{(1)}}{2\alpha^2} \int\left(\frac{d}{dt}\mathbf{B}^{(0)}\right)\rho^{(1)} \, dV \tag*{(75)}
=−e−(1)2α2(ddtB(0))∣r=0,(76)= -\frac{e_-^{(1)}}{2\alpha^2}\left(\frac{d}{dt}\mathbf{B}^{(0)}\right)\bigg|_{\mathbf{r}=0}, \tag*{(76)}

where we have used Gauss’ integral theorem in (74) and the point dipole approximation in (76). Hence

ddtErot(1)=(70,76)ddt(−m(1)⋅B(0))∣r=0.(77)\frac{d}{dt}E_{\mathrm{rot}}^{(1)} \overset{(70,76)}{=} \frac{d}{dt}\left(-\mathbf{m}^{(1)} \cdot\mathbf{B}^{(0)}\right)\bigg|_{\mathbf{r}=0}. \tag*{(77)}

The claim (60) then follows from

m(1)⋅B(0)∣r=0=m(0)⋅B(1)∣r=r0=m(0)m(1)μ04π2(a+z0)3.(78)\mathbf{m}^{(1)} \cdot\mathbf{B}^{(0)}\bigg|_{\mathbf{r}=0}=\mathbf{m}^{(0)} \cdot\mathbf{B}^{(1)}\bigg|_{\mathbf{r}=\mathbf{r}_0}=m^{(0)}m^{(1)}\frac{\mu_0}{4\pi}\frac{2}{(a+z_0)^3}. \tag*{(78)}

Summary

The problem of the interaction between two magnetic dipoles is interesting in itself and also as a case study to investigate the title question of whether the magnetic field can perform work. In this paper we have considered the special case in which both magnetic moments are initially aligned parallel to the line connecting the positions of the two dipoles, and, moreover, the limiting case in which the “large” dipole is so massive that it practically does not move and generates an external magnetic field, constant in time, in which the “small” dipole is accelerated.

For the detailed investigation, we have considered two cases, which we refer to as cases A and B. In the realm of Maxwell theory, it may be a matter of taste whether the magnetic moments are simulated by currents (model A) or considered as genuine quantities (model B). But for the energy balance we have to consider mechanical models for the dipoles, and then it makes a considerable difference which model is used.

We have derived energy balance equations for both models. For case A, the sum of the total kinetic energy of the two dipoles and the field energy is constant. Surprisingly, the rotational kinetic energy of the large dipole must also be taken into account as the source of the increase in both the field energy and the total kinetic energy of the small dipole. For case B, the role of the rotational kinetic energy of the small dipole is largely taken over by the Zeeman energy. In contrast to model A, here we have two separate energy conservation equations, one for the small dipole and another one for the field energy and the large dipole.

However, the question whether the magnetic field can do work cannot be answered unambiguously, because it is not clear what it means exactly. For example, J. A. Barandes argues that the magnetic field can do work since in his approach an energy balance analogous to our (59) is satisfied, see [8], eq. (40). On the other hand, the r. h. s. of the Poynting theorem (53), which summarizes the energy flow from matter to the electromagnetic field, is linear in E\mathbf{E} and does not contain B\mathbf{B}, also in the extended version based on [9]. Hence the magnetic field energy cannot be transformed into matter energy unless E≠0\mathbf{E} \ne0.

It is instructive to compare the repulsion of two magnetic dipoles with the repulsion of two charges of the same sign, which are similarly modeled by rigid charge clouds of extremely different mass. Here it is certainly legitimate to say that the electric field of the ”large” charge performs work, according to whatever criteria. The increase in kinetic energy of the “small” charge is compensated for by a decrease in potential energy, analogous to our case B. However, in the case of electric charges, the potential energy can be interpreted as the field energy of the total electric field. This field energy therefore decreases, while the magnetic field energy increases in both of our cases, A and B. In our opinion, the detailed description of the differences between the two examples of repulsion is much more revealing than the reduction to the simple alternative “the magnetic field performs work - or it does not”.

Acknowledgment

We thank Bruno Klajn for bringing the topic of this paper and recent literature to our attention, and Jacob A. Barandes for insightful discussions and comments on earlier versions of the manuscript. H.-J. S. would also like to express his gratitude to Jürgen Schnack for organizing a mini-symposium at the University of Bielefeld, where he was allowed to present parts of the present work.

Inhomogeneity correction

We replace (48) by

μ=mIρ:=(I1(t)x+I2(t)y−I2(t)x+I1(t)ym(0))ρ(∣r−r0(t)∣),(79)\boldsymbol{\mu}=\mathbf{m}_{I}\rho:=\begin{pmatrix} I_{1}(t)x+I_{2}(t)y \\ -I_{2}(t)x+I_{1}(t)y \\ m^{(0)} \end{pmatrix}\rho(\lvert\mathbf{r}-\mathbf{r}_{0}(t)\rvert), \tag*{(79)}

taking into account the SO(2)SO(2) symmetry w. r. t. rotations around the zz-axis and the condition that the direction field mI\mathbf{m}_{I} has the same absolute value everywhere in the linear order O(r)O(\mathbf{r}). The integral over the linear parts in μ\boldsymbol{\mu} vanishes so that the total magnetic moment is unchanged:

∫μ dV=m(0)=(00m(0)).(80)\int\boldsymbol{\mu}\,dV=\mathbf{m}^{(0)}=\begin{pmatrix} 0 \\ 0 \\ m^{(0)} \end{pmatrix}. \tag*{(80)}

Consequently, the equation for the force on the small dipole (55) still holds. Also the total torque vanishes, as in (57), due to SO(2)SO(2) symmetry. Hence the first energy balance equation (59) is left unchanged.

We introduce the “gyromagnetic ratio” χ\chi as the proportionality factor between moment density μ\boldsymbol{\mu} and spin density σ\boldsymbol{\sigma}, i. e.,

μ=χσ.(81)\boldsymbol{\mu}=\chi\boldsymbol{\sigma}. \tag*{(81)}

Then the precession of μ\boldsymbol{\mu} around the local magnetic field B(0)\mathbf{B}^{(0)} is given by the equation

ddtμ=χμ×B(0).(82)\frac{d}{dt}\boldsymbol{\mu}=\chi\boldsymbol{\mu}\times\mathbf{B}^{(0)}. \tag*{(82)}

This entails the equations of motion for the functions I1(t)I_1(t) and I2(t)I_2(t) which read in first order O(r)O(\mathbf{r}):

ddtI1(t)=−m(1)μ0χI2(t)2π(a+z0(t))3(83)\frac{d}{dt}I_1(t)=-\frac{m^{(1)}\mu_0\chi I_2(t)}{2\pi\left(a+z_0(t)\right)^3} \tag*{(83)}
ddtI2(t)=−3m(0)m(1)μ0χ4π(a+z0(t))4+m(1)μ0χI1(t)2π(a+z0(t))3.(84)\frac{d}{dt}I_2(t)=-\frac{3m^{(0)}m^{(1)}\mu_0\chi}{4\pi\left(a+z_0(t)\right)^4}+\frac{m^{(1)}\mu_0\chi I_1(t)}{2\pi\left(a+z_0(t)\right)^3}. \tag*{(84)}

It remains to calculate the inhomogeneity corrections to ddtEfield10\frac{d}{dt}E^{10}_{field} and to ddtErot(1)\frac{d}{dt}E^{(1)}_{rot}. For Efield10E^{10}_{field} we obtain additional terms in the integral (68) of order O(r2)O(\mathbf{r}^2) times ρ(0)\rho^{(0)} which result in:

Efield10=m(0)⋅B(1)(r0)+3m(1)μ0I1(t)4πα2(a+z0(t))4.(85)E^{10}_{field}=\mathbf{m}^{(0)}\cdot\mathbf{B}^{(1)}(\mathbf{r}_0)+\frac{3m^{(1)}\mu_0I_1(t)}{4\pi\alpha^2\left(a+z_0(t)\right)^4}. \tag*{(85)}

The time derivative of Efield10E^{10}_{field} hence reads:

ddtEfield10=m(0)⋅ddtB(1)(r0(t))+3m(1)μ0((a+z0(t))I˙1(t)−4I1(t)z˙0(t))4πα2(a+z0(t))5.(86)\frac{d}{dt}E^{10}_{field}=\mathbf{m}^{(0)}\cdot\frac{d}{dt}\mathbf{B}^{(1)}(\mathbf{r}_0(t))+\frac{3m^{(1)}\mu_0\left(\left(a+z_0(t)\right)\dot{I}_1(t)-4I_1(t)\dot{z}_0(t)\right)}{4\pi\alpha^2\left(a+z_0(t)\right)^5}. \tag*{(86)}

For ddtErot(1)\frac{d}{dt}E^{(1)}_{rot} we have to calculate the inhomogeneity correction to the magnetic field B(0)\mathbf{B}^{(0)}. To this end we recall (61) and (62) and note that

∇⋅μ=−2ρ(0)(I1(−1+α2(x2+y2))+m(0)α2u).(87)\nabla\cdot\boldsymbol{\mu}=-2\rho^{(0)}\left(I_1\left(-1+\alpha^2\left(x^2+y^2\right)\right)+m^{(0)}\alpha^2u\right). \tag*{(87)}

We consider the well-known equation

Δ(−14πrerf⁡(αr))=ρ(88)\Delta\left(-\frac{1}{4\pi r}\operatorname{erf}(\alpha r)\right)=\rho \tag*{(88)}

and apply differential operators ∂∂x\frac{\partial}{\partial x}, ∂∂y\frac{\partial}{\partial y}, …\ldots on both sides, using that these differential operators commute with Δ\Delta. This gives polynomials of xx, yy, zz times ρ\rho at the r. h. s., and thus we can obtain a solution of Δu=−∇⋅μ\Delta u=-\nabla\cdot\boldsymbol{\mu}, and, further the magnetic field B(0)\mathbf{B}^{(0)} with inhomogeneity correction. However, it is too complex to be shown here. Following the arguments that lead from (71) to (76) we finally obtain

ddtErot(1)=ddt(−m(1)⋅B(0))∣r=0−3m(1)μ0((a+z0(t))I˙1(t)−4I1(t)z˙0(t))4πα2(a+z0(t))5.(89)\frac{d}{dt}E^{(1)}_{rot}=\left.\frac{d}{dt}\left(-\mathbf{m}^{(1)}\cdot\mathbf{B}^{(0)}\right)\right|_{\mathbf{r}=0}-\frac{3m^{(1)}\mu_0\left(\left(a+z_0(t)\right)\dot{I}_1(t)-4I_1(t)\dot{z}_0(t)\right)}{4\pi\alpha^2\left(a+z_0(t)\right)^5}. \tag*{(89)}

Comparison of (A.11) and (A.8) proves that the energy balance (60) also holds if taking the inhomogeneity correction into account.

Analytical solution of the eom for model B

Eq. (A.4) shows that ω1=χB\omega_1=\chi B is a natural unit for the angular velocity in case of model B. It can be identified with ω1\omega_1 defined by (23) for model A if we introduce a formal charge

e(0):=χM(0)(90)e^{(0)}:=\chi M^{(0)} \tag*{(90)}

also for model B and identify M(0)M^{(0)} with M−(0)M_-^{(0)}. Then we can adopt the dimensionless quantities from § 2.3 and write the eom for the model B in the form

z¨0(t)=3(2π∣γ1∣μ22ϵ)1(1+z0(t)4)(91)\ddot{z}_0(t)=3\left(2\pi|\gamma_1|\frac{\mu_2^2}{\epsilon}\right)\frac{1}{\left(1+z_0(t)^4\right)} \tag*{(91)}

Using these quantities the first energy balance (59) assumes the form

12z˙0(t)2+λ21(1+z0(t)3)=E.(92)\frac{1}{2}\dot{z}_0(t)^2+\frac{\lambda}{2}\frac{1}{\left(1+z_0(t)^3\right)}=E. \tag*{(92)}

With the initial conditions z0(0)=0z_0(0)=0 and z˙0(0)=0\dot{z}_0(0)=0 we obtain for the total energy E=λ2E=\frac{\lambda}{2} and hence

z˙0(t)2=λ(1−1(1+z0(t))3).(93)\dot{z}_0(t)^2=\lambda\left(1-\frac{1}{\left(1+z_0(t)\right)^3}\right). \tag*{(93)}

Obviously, for t→∞t\to\infty we have z0(t)→∞z_0(t)\to\infty and z˙0(t)→λ\dot{z}_0(t)\to\sqrt{\lambda}. Solving (B.4) for tt yields

t=∫0z0dzλ(1−1(1+z)3).(94)t=\int_0^{z_0}\frac{dz}{\sqrt{\lambda\left(1-\frac{1}{(1+z)^3}\right)}}. \tag*{(94)}
=−13λ∫1(1+z0)−3u−4/3(1−u)−1/2 du(95)=-\frac{1}{3\sqrt{\lambda}}\int_1^{(1+z_0)^{-3}}u^{-4/3}(1-u)^{-1/2}\,du \tag*{(95)}
=−13λ(B(−13,12,(1+z0)−3)−B(−13,12,1))(96)=-\frac{1}{3\sqrt{\lambda}}\left(B\left(-\frac{1}{3},\frac{1}{2},(1+z_0)^{-3}\right)-B\left(-\frac{1}{3},\frac{1}{2},1\right)\right) \tag*{(96)}
=1λ(1+z0)(2F1(−13,12;23;(1+z0)−3)−π Γ(23)Γ(16)).(97)=\frac{1}{\sqrt{\lambda}}(1+z_0)\left({}_2F_1\left(-\frac{1}{3},\frac{1}{2};\frac{2}{3};(1+z_0)^{-3}\right)-\frac{\sqrt{\pi}\,\Gamma\left(\frac{2}{3}\right)}{\Gamma\left(\frac{1}{6}\right)}\right). \tag*{(97)}

In (B.6) we have substituted u=(1+z)−3u=(1+z)^{-3} in order to write the integral in terms of the incomplete beta function B(a,b,x)B(a,b,x), see [15], (8.17.1). The latter can also be expressed by the hypergeometric function 2F1(a,b;c;x){}_2F_1(a,b;c;x)., see [15], (8.17.7), which gives (B.8). Thus we have obtained an analytical solution of the eom in terms of the inverse function t↦t(z0)t \mapsto t(z_0).

We have compared this solution with a numerical solution of the eom choosing the special value of λ=1\lambda= 1 and found perfect agreement, see Figure B1.

Figure B1. Numerical solution of the eom (B.2) with the special value of λ=1\lambda= 1 (blue curve) and analytical solution according to (B.8) (red, dashed curve).

References

  1. [1]J. D. Jackson, Classical Electrodynamics, 3rd edition, Wiley, New York, 1999.
  2. [2]E. P. Mosca, Magnetic Forces Doing Work?, Am. J. Phys. 42, 295 – 297 (1974)DOI
  3. [3]C. A. Coombes, Work done on charged particles in magnetic fields, Am. J. Phys. 47, 915 – 916 (1979)DOI
  4. [4]P. Onorato and A. De Ambrosis, How can magnetic forces do work? Investigating the problem with students, Phys. Educ. 48 (6), 766 (2013)DOI
  5. [5]J. Gates, Magnetic force and work: an accessible example, Phys. Educ. 49 (3), 299 – 302 (2014)DOI
  6. [6]S. C. Veetil, H. X. Sim, and B. Ricardo, Magnetic Forces: Do They Really Work?, The Phys. Educ. 5 (1), 2350001 (2023)DOI
  7. [7]J. A. Barandes, Can magnetic forces do work?, Preprint arXiv:1911.08890 (2019)arxiv.org/abs/1911.08890
  8. [8]J. A. Barandes, On magnetic forces and work, Found. Phys. 51 (4), 79, (2021)DOI
  9. [9]H.-J. Schmidt and T. Bröcker, Maxwell equations with spin density, Eur. J. Phys. 44(3), 035201, (2023)DOI
  10. [10]H.-J. Schmidt, C. Schröder, E. Hägele, and M. Luban, Dynamics and thermodynamics of a pair of interacting magnetic dipoles, J. Phys. A, Math. Theor. 48(18), 185002, (2015)DOI
  11. [11]J. D. Bjorken and S. D. Drell, Relativistic Quantum Mechanics, McGraw-Hill, New York, 1964.
  12. [12]J. S. Marsh, Magnetic and electric fields of rotating charge distributions, Am. J. Phys. 50, 51 – 53, (1982)DOI
  13. [13]T. H. Boyer, The force on a magnetic dipole, Am. J. Phys. 56, 688 – 692, (1988)DOI
  14. [14]L. Vaidman, Torque and force on a magnetic dipole, Am. J. Phys. 58, 978 – 983, (1990)
  15. [15]NIST Digital Library of Mathematical Functions. http://dlmf.nist.gov/, Release 1.1.1 of 2023-09-15. F. W. J. Olver, A. B. Olde Daalhuis, D. W. Lozier, B. I. Schneider, R. F. Boisvert, C. W. Clark, B. R. Miller, B. V. Saunders, H. S. Cohl, and M. A. McClain, eds.

Paper details

vlnc:W_nzyae66k
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