Interactions between uniformly magnetized spheres

Boyd F. Edwards, D. M. Riffe, Jeong-Young Ji, and William A. Booth Department of Physics Utah State University Logan, UT 84322

(Dated: May 27, 2016)

We demonstrate the magnetic interactions (both and ) between two uniformly mag- netized spheres are identical to the magnetic interactions between two point magnetic .

I. INTRODUCTION magnetized sphere and the field of an equivalent point . Unlike the previous calculations, ours rely on Let’s consider two permanent of arbitrary simple symmetry arguments and pertain to spheres of shape. If each has a nonzero dipole moment, arbitrary sizes, positions, , and magnetic then the dipole moments of these magnets will dominate orientations. their interactions at separations that are large compared The between two dipoles is noncentral. Namely, with their sizes, and each magnet may be treated as if it this force is not generally directed along the line through were a point . Dipolar fields and forces the dipoles. While other noncentral magnetic forces vio- are often used as the starting point for analytical and nu- late ’s third law [31–33], we show that the paired merical approximations of the forces between permanent forces between magnet dipoles obeys this law. These magnets of various shapes [1–10]. forces therefore exert a net r × F on an isolated two-dipole system. This torque cancels the paired m×B A uniformly magnetized sphere produces a magnetic torques, which are not equal and opposite. Thus angular field that is identical to its dipole field, not just at large momentum is conserved. distances, but everywhere outside of the sphere [11, 12]. One is thus naturally lead to ask whether the forces and torques between two uniformly magnetized spheres are II. POINT DIPOLE INTERACTIONS identical to those between two point dipoles, independent of their separation. Here we show this is indeed the case. In this section we review the interactions between point This result has practical applications. Dipolar fields magnetic dipoles. The magnetic field at position r pro- and forces have been used to approximate the interac- duced by a point dipole m located at the origin is given tions among assemblies of spherical nanoparticles [13] by [34, 35] and magnetic microspheres [14]. Our results show that these approximations are, in reality, exact. In addition, µ0 3m · r m B(m; r)= r − , (1) small rare-earth magnetic spheres are used both in and 4π r5 r3 out of the classroom to teach principles of mathemat-   ics, physics, chemistry, biology, and engineering [15–19]. for r = |r| satisfying r > 0. This field can be obtained Our results enable simple dipole interactions to be used from the scalar potential to model the dynamical interactions between these mag- nets [20, 21]. µ0 m · r ϕ(m; r)= 3 (2) Previous calculations of the force between two uni- 4π r formly magnetized spheres have been carried out in three via limiting geometries: (i) for magnetizations that are per- pendicular to the line through the sphere centers [22], B(m; r)= −∇ϕ(r). (3) (ii) for parallel magnetizations that make an arbitrary angle with this line [23], and (iii) for configurations with Because ∇ · B = 0, ϕ satisfies Laplace’s equation, one parallel to this line and the other in 2 an arbitrary direction [24]. All three calculations yield ∇ ϕ =0. (4) forces that are identical to the force between two point We consider two dipoles, m1 and m2, which are re- magnetic dipoles [25–30]. spectively located at positions r1 and r2. From Eq. (1), In this paper, we show generally that the force be- the field produced by dipole mi is tween two uniformly magnetized spheres is identical to the force between two point magnetic dipoles, and show Bi(r)= B(mi; r − ri), (5) the same equivalence for the torque. Like the calcula- tions for the three limiting geometries, our calculations where i = 1, 2, and where r − ri is the position vector exploit the equivalence of the field outside of a uniformly relative to dipole mi (Fig. 1). Accordingly, the field of 2

r r r r mi evaluated at the location of mj is - 1 - 2 F 21 r M M 2 µ0 3mi · rij mi 1 Bi(rj )= 5 rij − 3 , (6) r 4π rij rij ! a 12 F a 1 z 12 2 r r 1 2 where rij = rj − ri is the position of mj relative to mi, and rij = |rj − ri|. y The interaction energy between mj and the magnetic x field of mi is FIG. 1. Diagram showing two uniformly magnetized spheres Uij = −mj · Bi(rj ). (7) with positions r1 and r2, radii a1 and a2, magnetizations M1 and M2, and paired non-central magnetic forces F12 and F21. Inserting Eq. (6) gives Shown also are an arbitrary position vector r and the relative position vectors r − r1, r − r2, and r12 = r2 − r1. The same µ0 mi · mj (mi · rij )(mj · rij ) diagram applies for the forces between two point dipoles if Uij = 3 − 3 5 . (8) spheres M1 and M2 are replaced by dipoles m1 and m2 at 4π " rij rij # the same locations.

The force of mi on mj follows from ∇ is generally unequal to zero. Therefore, unlike the paired Fij = − jUij , (9) A A forces F12 and F21, the paired torques τ12 and τ21 are where ∇j is the gradient with respect to rj . Inserting not generally equal and opposite. Eq. (8) yields The second contribution to the torque arises from the force of mi on mj , 0 3µ B Fij = 5 (mi · rij ) mj + (mj · rij ) mi τ = r × F , (16) 4πr ij j ij ij  (m · r ) (m · r ) for which the sum + (m · m ) r − 5 i ij j ij r . (10) i j ij r2 ij τ B τ B ij  12 + 21 = r12 × F12 (17) The first two terms in the square brackets are respec- is also generally unequal to zero. tively parallel to mj and mi. Consequently, Fij is not Equations (15) and (17) and a little algebra reveal that central, namely, it is not generally parallel to the vector the net torque on an isolated pair of dipoles vanishes rij between the dipoles. identically; Equations (8) and (10) imply that A A B B τ12 + τ21 + τ12 + τ21 =0. (18) U21 = U12 (11) A A The torque supplied by τ12 and τ21 therefore cancels the torque supplied by τ B and τ B , and angular momentum and 12 21 is conserved. Thus, an isolated dipole-dipole system does F21 = −F12, (12) not spontaneously rotate, and there is no exchange be- tween mechanical and electromagnetic momentum. Such confirming that Newton’s third law applies to the mag- exchanges have been the subject of considerable study netic force between point magnetic dipoles, and ensuring [32, 37–40]. that linear momentum is conserved in an isolated two- dipole system. III. FORCE BETWEEN SPHERES We now investigate the torque of mi on mj, which has two contributions [36], We now come to the crux of this paper, the force be- τ τ A τ B ij = ij + ij . (13) tween two uniformly magnetized spheres. We present four separate arguments that show that the force be- The first arises from mj residing in the field of mi, tween uniformly magnetized spheres is identical to the force between point dipoles. While each proof is sufficient τ A = m × B (r ). (14) ij j i j to show this equivalence, each utilizes different concepts The sum of paired torques from mechanics and , and each has ped- agogical value. τ A τ A 3µ0 As seen in Fig. 1, we take sphere i to have position ri, 12 + 21 = 5 (m1 · r12) m2 × r12 4πr12 radius ai, magnetization Mi, and total dipole moment h 4 + (m2 · r12) m1 × r12 (15) m = πa3 M . (19) i 3 i i i 3

(a) Fig. 2(b): Sphere 1 produces the same field B1, and m therefore exerts the same force F12 on dipole 2. m F 2 1 12 Fig. 2(c): Newton’s third law gives the force F21 = −F12 of dipole 2 on sphere 1. This force is pro- duced by the field B2 of dipole 2.

(b) Fig. 2(d): Sphere 2 produces the same field B2, and therefore exerts the same force F21 on sphere 1. m F 2 M 12 1 Fig. 2(e): To complete the argument, we again apply Newton’s third law to show that the force F12 = −F21 of sphere 1 on sphere 2 is identical to the force of dipole 1 on dipole 2 (Fig. 2a). (c) m M 2 1 F B. Direct integration 21 The force between two uniformly magnetized spheres can be determined by integrating the energy density (d) −Mj · Bi(r) associated with sphere j sitting in the mag- netic field Bi(r), giving the total interaction energy M M 2 1 F 21 Uij = − Mj · Bi dV = −Mj · Bi dV. (20) Zj Zj Here, the is over the volume of sphere j and (e) the second equality exploits the uniformity of Mj. If all M sources of a magnetic field lie outside a particular sphere, M F 2 1 12 then the spatial average of the field over the sphere is given by the value of the field at the sphere center [43]. That is,

FIG. 2. Diagram illustrating the five steps of the Newton’s 4 3 Bi dV = πaj Bi(rj ). (21) third law argument showing the force between two uniformly j 3 magnetized spheres is identical to the force between two point Z dipoles. Equations (19), (20), and (21) give Uij = −mj · Bi(rj ), which replicates Eq. (7). Thus, the energy of interaction between the two spheres is identical to the energy of in- Its magnetic field is given by Bi = 2µ0Mi/3 inside the teraction between two point dipoles. The associated force sphere (for |r − ri|

A. Newton’s third law We can also show the force equivalence by integrating 2 the magnetic energy density B /2µ0 over all space, giving A five-step argument involving Newton’s third law the total magnetic energy of interaction shows that the force between two spheres with uniform 1 2 1 2 magnetizations M and M is identical to the force be- U(r1, r2)= (B1 + B2) dV tween two point dipoles with corresponding equivalent 2µ0 Z magnetic moments m1 and m2, located at the same po- 1 2 2 = B1 +2 B1 · B2 + B2 dV. (22) sitions as the spheres. Figure 2 illustrates this argument. 2µ0 Z  Fig. 2(a): As in the preceding discussion, F12 represents Because the magnetic energy of a single dipole does not the force of dipole 1 on dipole 2. This force is pro- depend on its location in space, the self-energy −1 2 duced by the field B1 of dipole 1. (2µ0) Bi dV do not depend on ri. Therefore the force R 4

∇ τ A Fij = − j U on sphere j depends only on the interaction Invoking Eqs. (19) and (21) gives ij = mj × Bi(rj ), energy τ A which replicates Eq. (14). Thus, ij between two uni- 1 formly magnetized spheres is identical to the torque be- Uint = B1 · B2 dV tween two point magnetic dipoles. µ0 Z The second contribution arises from the force of sphere 1 i on sphere j, which can be obtained from the interaction = + + ∇ϕ1 · ∇ϕ2 dV, (23) µ0 1 2 outside energy density u = −M · B according to Z Z Z  ij j i where we have inserted Eq. (3), and we have separated Fij = − ∇uij dV. (26) the integral over all space into integrals over sphere 1, j sphere 2, and the region outside of both spheres. Z 2 For the integral over sphere 1, we use ∇ ϕ2 = 0 and This gives the same force as Fij = −∇j Uij , with Uij the divergence theorem to write given by Eq. (7). The torque follows by integrating the torque density r × (−∇uij) over the volume of sphere j; ∇ϕ1 · ∇ϕ2 dV = ∇ · (ϕ1∇ϕ2) dV 1 1 τ B ∇ Z Z ij = − r × uij(r)dV. (27) j = ϕ1∇ϕ2 · nˆ1 dA, (24) Z ′ ZS1 Rewriting using the position r = r − rj relative to the center of sphere j gives where S1 denotes the integral over the surface of sphere 1, and nˆ1 is the unit vector directed normally outward τ B ′ ∇ from the sphere’s surface. The quantity ϕ1∇ϕ2 is con- ij = − (r + rj ) × uij (r)dV j tinuous across this surface. The surface values of ϕ1 and Z 2 ′ ′ ′ ′ ϕ are identical to the potentials of point dipoles, and = − r × ∇ Ψ(r )dV − rj × ∇uij(r)dV, (28) so the integral over sphere 1 is the same as it would be Zj Zj if the spheres were replaced by equivalent point dipoles. where Ψ(r′)= u (r′ + r ). Converting the first integral Similarly, the integral over sphere 2 is the same as for ij j into a surface integral gives equivalent point dipoles. Since the fields outside of both spheres match the fields of equivalent point dipoles, all ′ ∇′ ′ ∇′ ′ ′ three integrals Eq. (23) are the same for point dipoles as r × ΨdV = − × (Ψr )dV (29) j j for uniformly magnetized spheres. Therefore, the force Z Z ∇ ′ ′ Fij = − j Uint must also be the same. = − nˆj × r ΨdA . (30) ZSj ′ D. Stress tensor Since r is parallel to the unit normal vector nˆj , their cross product is zero and the integral vanishes. Combin- ing Eqs. (26) and (28) gives The total force on an object can be calculated by inte- τ B grating the Maxwell stress tensor over an arbitrary sur- ij = rj × Fij , (31) face surrounding the object [44, 45]. The magnetic stress tensor depends only on the magnetic field. The field out- which is identical to Eq. (16). Thus, the torque asso- side a uniformly magnetized sphere is the same as the ciated with the force on a uniformly magnetized sphere field of a point dipole located at its center. The total in a dipole field is the same as the torque on the corre- field produced by and outside of two uniformly mag- sponding point dipole. These results ensure that the net netized spheres, and the associated stress tensor, is the torque on an isolated two-sphere system is zero, as seen same as that of two point dipoles. Therefore, the force earlier for point dipoles. between spheres must be the same as the force between point dipoles. V. SUMMARY

IV. TORQUE BETWEEN SPHERES We have demonstrated that the forces and torques be- tween two uniformly magnetized spheres are identical

Here we calculate the torque τij of sphere i on sphere to the forces and torques between two point magnetic j. This torque has two contributions, as before. dipoles. The first arises from sphere j residing in the field of This equivalence extends to uniformly polarized sphere i, and can be obtained by integrating the torque spheres and point electric dipoles. This is because the density Mj × Bi over the volume of sphere j, giving fields, forces, and torques of electric dipoles have the same mathematical forms as their magnetic counterparts, τ A = M × B dV = M × B dV. (25) and the field outside of a uniformly polarized sphere is ij j i j i identical to the electric dipole field. Zj Zj 5

ACKNOWLEDGMENTS cases, correspondence with Marco Beleggia about the po- tential theory argument, and discussions with Kirk Mc- We gratefully acknowledge support from NSF Grant Donald and David Griffiths regarding the point-sphere No. 1332265, discussions with W. Farrell Edwards on the equivalence. We are especially grateful to David Grif- applicability of Newton’s third law to magnetostatic in- fiths for suggesting the stress tensor argument to us, and teractions, correspondence with David Vokoun about cal- to Kirk McDonald for pointing out Ref. [43], which sim- culations of the force between magnet spheres in special plifies some of the calculations.

[1] G. Akoun and J. Yonnet, IEEE Trans. Magn. 20, 1962 at https://drive.google.com/file/d/0Bw7DdocNZGQgW (1984). ThTb3VvUHYza2s/view (accessed May 10, 2016). [2] A. Kruusing, J. Magn. Magn. Mater. 234, 545 (2001). [17] The Zen Gallery, curated by Shihan Qu, shows photos of [3] D. Vokoun, M. Beleggia, L. Heller, and P. Sittner, “Mag- various magnetic sculptures including models of fractals, netostatic interactions and forces between cylindrical molecules, lattices, and Platonic solids, http://zenmag permanent magnets,” J. Magn. Magn. Mater. 321, 3758- nets.com/gallery/ (accessed Feb. 9, 2016). 3763 (2009). [18] Typing “Zen Magnets” into the YouTube search field at [4] J. S. Agashe and D. P. Arnold, J. Phys. D. Appl. Phys. https://www.youtube.com identifies over 90,000 videos 42, 099801 (2009). describing various (accessed Mar. 23, 2016). As of August [5] S. Sanz and L. Garcia-Tabares, IEEE Trans. Appl. Su- 22, 2014, the most popular of these had a total view count percond. 20, 846 (2010). exceeding 145 million (Ref. [16], Appendix D). [6] J. Camacho and V. Sosa, Rev. Mex. Fisica E 59, 8 (2013). [19] Several countries, including the United States, have [7] M. Beleggia, S. Tandon, Y. Zhi, M. De Graef, “On the banned the sale of 5-mm diameter nickel-coated magnetostatic interactions between nanoparticles of ar- neodymium magnets marketed as a desk toys under bitrary shape,” J. Magn. and Magn. Mater. 278, 270-284 the trade names BuckyBalls, Zen Magnets, Neoballs, (2004). etc. following reports of intestinal injuries from ingestion [8] Wikipedia, Force between magnets, https://en.wikipedia of these magnets. But 2.5-mm magnets may be still .org/wiki/Force between magnets (accessed Feb. 10, be purchased at http://micromagnets.com (accessed 2016). Feb. 16, 2016), and magnet spheres of various sizes, [9] Attraction/repulsion force between two equal magnets?, including 5-mm spheres, may still be purchased from http://www.eng-tips.com/viewthread.cfm?qid=341932 industrial suppliers (http://www.kjmagnetics.com/, (accessed Feb. 10, 2016). http://www.alibaba.com, http://www.magnet4less.com, [10] Commercial magnetic-field mapping software can be accessed Feb. 16, 2016). found at phttp://www.gmw.com/magnetic measure [20] Boyd F. Edwards and John M. Edwards, “Dynamical ments/magmap.html?gclid=CLuJiK NvssCFUKUfgod interactions between two uniformly magnetized spheres,” K9AAJg, https://www.integratedsoft.com, http://www. to be published (2016). infolytica.com (accessed Feb. 16, 2016). [21] John M. Edwards, MagPhyx Simulation and Visu- [11] Roald K. Wangsness, Electromagnetic Fields, 2nd edition alization Software, http://www2.cose.isu.edu/ edwa- (Wiley, New York, NY, 1986), p. 326. john/MagPhyx (accessed Mar. 11, 2016). This web-based [12] J. D. Jackson, Classical Electrodynamics, 2nd edition software simulates the 2D motion of a magnet sphere in (Wiley, New York, NY, 1975), p. 195. response to the forces and torques supplied by a second [13] M. Var´on, M. Beleggia, T. Kasama, R. J. Harrison, R. E. magnet sphere, held fixed. It is provided freely to the Dunin-Borkowski, V. F. Puntes, & C. Frandsen, “Dipolar physics community for education and exploration. Magnetism in Ordered and Disordered Low-Dimensional [22] M. Beleggia and M. De Graef, “General magnetostatic Nanoparticle Assemblies,” Scientific Reports 3 (1234), 1- shape-shape interactions,” J. Magn. and Magn. Mater. 5 (2013). 285, L1-L10 (2005). [14] G. Helgesen, T. T. Skjeltorp, P. M. Mors, R. Botet, and [23] D. Vokoun and M. Beleggia, “Forces between arrays of R. Jullien, “Aggregation of Magnetic Microspheres: Ex- permanent magnets of basic geometric shapes,” J. Magn. periments and Simulations,” Phys. Rev. Lett. 61 (15), and Magn. Mater. 350, 174-178 (2014). This publication 1736-1739 (1988). quotes the result of an unpublished calculation of the [15] David A. Richter, Expert Report, “Teaching Geometry magnetic interaction between two identical spheres with with Magnet Sphere Kits,” in the matter of Zen Mag- parallel magnetizations. Details of this calculation were nets, LLC, CPSC Docket No. 12-2, Item 124, Exhibit sent privately to us by D. Vokoun on Feb. 8, 2016. 3, 10/20/2014, http://www.cpsc.gov/en/Recalls/Recall- [24] W. Booth and B. Edwards, unpublished. Lawsuits/Adjudicative-Proceedings/ (accessed Feb. 9, [25] David Griffiths, Introduction to Electrodynamics, 3rd edi- 2016). tion (Pearson Education, Delhi, India, 1999), p. 282. [16] Boyd F. Edwards, “Expert Report: Educational Value of [26] Kar W. Young, Peter B. Landecker, and Daniel D. Vil- Neodymium Magnet Spheres in the matter of Zen Mag- lani, “An analytic solution for the force between two nets, LLC, CPSC Docket No. 12-2,” Item 124, Exhibit magnetic dipoles,” Magnetic and Electrical Separation, 4, 10/20/2014, http://www.cpsc.gov/en/Recalls/Recall- 9, 39-52 (1998). Lawsuits/Adjudicative-Proceedings/ (accessed Feb. 9, [27] Timothy H. Boyer, “The force on a magnetic dipole,” 2016). Redacted, high-resolution, color version available Am. J. Phys. 56, 688-692 (1988). 6

[28] K. R. Brownstein, “Force exerted on a magnetic dipole,” [35] David J. Griffiths, “Dipoles at rest,” Am. J. Phys. 60, Am. J. Phys. 61, 940-941 (1993). 979-987 (1992). [29] Jack B. Greene and Frank G. Karioris, “Force on a Mag- [36] We ignore the torque on a sphere due to the electric netic Dipole,” Am. J. Phys. 39, 172-175 (1971). dipole moment that it acquires from its motion. Point [30] Lev Vaidman, “Torque and force on a magnetic dipole,” magnetic dipoles in uniform motion through a uniform Am. J. Phys. 58, 978-983 (1990). magnetic field acquire an electric dipole moment that [31] Richard P. Feynman, R. B. Leighton, and M. Sands, The interacts with the magnetic field to give an additional Feynman Lectures on Physics Vol. II, (Addison-Wesley, torque on the dipole [38]. But this torque is proportional Boston, Massachusetts, 1964), p. 26-5. Here, Feynman to v2/c2, and we ignore it for our non-relativistic inves- discusses the example of the non-central force between tigations. two positively charged particles, one moving in the +x [37] David J. Griffiths and V. Hnizdo, “Mansuripur’s para- direction and the other moving in the +y direction. At dox,” Am. J. Phys. 81, 570-574 (2013). the instant that particle 1 is at the coordinate origin and [38] David J. Griffiths and V. Hnizdo, “The torque on a dipole particle 2 is on the +y axis, the electric forces are central in uniform motion,” Am. J. Phys. 82, 251 (2014). and repulsive, but particle 1 exerts a non-central mag- [39] V. Hnizdo, “Hidden mechanical momentum and the field netic force on particle 2 in the +x direction, while particle momentum in stationary electromagnetic and gravita- 2 exerts no magnetic force on particle 1. Thus, Newton’s tional systems,” Am. J. Phys. 65, 515-518 (1997). third law does not apply, raising a paradox about the [40] Timothy H. Boyer, “Classical interaction of a magnet origin of the extra mechanical momentum. and a point charge: The Shockley-James paradox,” Phys. [32] Adam Caprez and Herman Batelaan, “Feynman’s Rel- Rev. E 91, 013201 (2015). ativistic Electrodynamics Paradox and the Aharonov- [41] Roald K. Wangsness, Electromagnetic Fields, 2nd edition Bohm Effect,” Found. Phys. 39, 295-306 (2009). Here, (Wiley, New York, NY, 1986), p. 340. the authors show that the extra mechanical momentum [42] Coercivity, Wikipedia, http://en.wikipedia.org/wiki/ of Feynman’s paradox is hidden in the electromagnetic Coercivity (accessed Feb. 16, 2016). fields. [43] B. Y. Hu, “Averages of static electric and magnetic fields [33] Jerry B. Marion and Stephen T. Thornton, Classical Dy- over a spherical region: a derivation based on the mean- namics of Particles and Systems, 4th edition (Saunders value theorem,” Am. J. Phys. 68, 1058 (2000). College Publishing, Fort Worth, Texas, 1995), p. 50. [44] Jack Vanderlinde, Classical Electromagnetic Theory, 1st [34] David Griffiths, Introduction to Electrodynamics, 3rd edi- edition (John Wiley & Sons, New York, New York, 1993), tion (Pearson Education, Delhi, India, 1999), p. 246. pp. 81-84. [45] David Griffiths, Introduction to Electrodynamics, 3rd edi- tion (Pearson Education, Delhi, India, 1999), pp. 351- 354.