Download Dielectric and Magnetic Image Methods: A Comprehensive Guide with Exercises and more Exercises Electrodynamics in PDF only on Docsity!
Dielectric (and Magnetic) Image Methods
Kirk T. McDonald Joseph Henry Laboratories, Princeton University, Princeton, NJ 08544 (November 21, 2009; updated February 7, 2020)
1 Problem
The method of images is most often employed in electrostatic examples with point or line charges in vacuum outside conducting planes, cylinders or spheres [1]. Develop similar prescriptions for the electric scalar potential in examples where the conductor is a linear, isotropic dielectric medium with relative permittivity . Discuss also media with relative permeability μ when a line current or “point” magnetic dipole is present. Use the results to discuss the forces on these static systems.
2 Image Solutions^1
2.1 Dielectric Media
In a linear isotropic dielectric medium with relative permittivity , the electric field E and the displacement field D are related by D = E (in Gaussian units). We assume that there are no free charges in/on the dielectric medium, such that ∇ · D = 0, and hence ∇ · E = 0 within the dielectric medium. The net polarization charge density, if any, resides only on the surface of the dielectric medium. Of course, ∇ × E = 0 in static examples, so the electric field can be related to a scalar potential V according to E = −∇V. Thus, inside a linear, isotropic dielectric medium, the scalar potential obeys Laplace’s equation, ∇^2 V = 0, and the scalar potential can be represented by familiar Fourier series in rectangular, cylindrical and spherical coordinates (and 8 other coordinate systems as well). If the interface between the dielectric medium and vacuum (or another dielectric medium) supports no free charge, then the normal component of the displacement field D and the tangential component of the electric field E are continuous across that interface. The latter condition is equivalent to the requirement that the potential V be continuous across the interface.
2.1.1 Point Charge Outside a Dielectric Half Space
This section is a variant of sec. 4.4 of [3].^2 We first consider the case of a dielectric medium with relative permittivity in the half space z < 0 with a point charge q at (x, y, z) = (0, 0 , a), where otherwise the region z > 0 has relative permittivity ′. The image method is to suppose that the potential in the region
(^1) For a review, see [2]. (^2) For a point charge outside an infinite, dielectric plate of finite thickness, see [4, 5].
z > 0 is that due to the original point charge q at (0, 0 , a) plus an image charge q′^ at (0, 0 , −b), and that the potential in the region z < 0 is that due to the original point charge plus a point charge q′′^ at (0, 0 , c). We first recall that a “free” electric charge q in an infinite dielectric medium of relative permittivity ′^ has fields D = ′E = q ˆr/r^2 , because charge −q( − 1)/ is induced on the adjacent surface of the medium surrounding the charge. This suggests that for the proposed image method, the electric scalar potential at (x, 0 , z > 0) has the form,
V (x, 0 , z > 0) =
q ′[x^2 + (z − a)^2 ]^1 /^2
q′ ′[x^2 + (z + b)^2 ]^1 /^2
and that at (x, 0 , z < 0) could be written as,
V (x, 0 , z < 0) =
q ′[x^2 + (z − a)^2 ]^1 /^2
q′′ ′[x^2 + (z − c)^2 ]^1 /^2
Then, continuity of the potential V across the plane z = 0 requires that,
b = c, and q′′^ = q′. (3)
Continuity of Dz across the plane z = 0 requires that ′Ez (x, 0 , 0 +) = Ez (x, 0 , 0 −), i.e.,
′^
∂V (x, 0 , 0 +) ∂z
∂V (x, 0 , 0 −) ∂z
qa [x^2 + a^2 ]^3 /^2
q′b [x^2 + b^2 ]^3 /^2
qa [x^2 + a^2 ]^3 /^2
q′b [x^2 + b^2 ]^3 /^2
which implies that,^3
a = b (= c), and (q′′^ =) q′^ = −q
+ ′^
The potential and electric field E = D/ in the region z < 0 are as if the media were vacuum and the original charge q were replaced by charge (q + q′′)/′^ = 2q/( + ′). In the region z > 0 the potential and E field are as for the original effective charge q/′^ plus an effective image charge q′/′^ = q′′/′, both in vacuum.^4 In the limit that a = 0 (such that charge q lies on the interface between the two dielectric media) the electric field is as if the media were vacuum but the charge were 2q/( + ′). Note that the electrical field is radial, and the same in both media despite their differing
(^3) For a metamaterial with relative permeability ′ (^) = −, eq. (6) diverges (as could its magnetic equiva-
lents (37) and (43) for certain negative relative permeabilities). However, metamaterials can have negative permittivity (and/or permeability) only for nonzero frequencies [6], so technically this divergence cannot occur. It remains possible that metamaterials could lead to large “image” forces at low frequencies [7]. (^4) The electric field of the image charge q′ (^) at z = −a does not vary as q′/r (^2) as might be expected for an
actual charge at this location, because the image charge represents effects at z > 0 of induced charges at the interface z = 0. Likewise, the electric field for z < 0 can be written as 2q ˆrq/(′^ + )r^2 q = q ˆrq/aver q^2.
potential has the symmetry V (r, −θ) = V (r, θ), so the Fourier expansion for the potential contains terms in cos nθ, but not sin nθ. The potential due to the wire in the absence of the dielectric cylinder has the general form,
Vwire(r, θ) =
a 0 +
n=1 an
( (^) r b
)n cos nθ (r < b), a 0 + b 0 ln r b +
n=1 an
( (^) b r
)n cos nθ (r > b),
since the potential should not blow up at the origin, should be continuous at r = b, and can have a logarithmic divergence at infinity. For large r the electric field due to the wire is Ewire = 2q ˆr/′r, and the corresponding asymptotic potential is defined to be Vwire = −(2q/′) ln r. Thus,
a 0 = − 2 q ′^
ln b, b 0 = − 2 q ′^
The remaining coefficients an are determined the Maxwell equation ∇ · D = 4πρfree by considering a Gaussian surface (of unit length in z) that surrounds the cylindrical shell (b, θ),
4 πqfree,in =
D · dArea = b′
dθ (Er+ − Er− ). (10)
From this we learn that,
4 πqδ(θ) = b′(Er+^ − Er−^ ) = b′
∂Vwire(b+) ∂r
∂Vwire(b−) ∂r
= 2 q + 2′^
n
nan cos nθ. (11)
Multiplying eq. (11) by cos nθ and integrating over θ we find that,
an =
2 q ′n
Then, the potential due to the wire can be written as,
Vwire(r, θ) =
2 q ′^ [−^ ln^ b^ +^
n=
1 n
( (^) r b
)n cos nθ] (r < b), 2 q ′^ [−^ ln^ r^ +^
n=
1 n
( (^) b r
)n cos nθ] (r > b).
When the dielectric cylinder is present it supports a polarization charge density that results in an additional scalar potential which can be expanded in a Fourier series similar to eq. (8),
Vcylinder(r, θ) =
n=1 An
( (^) r a
)n cos nθ (r < a), ∑ n=1 An
( (^) a r
)n cos nθ (r > a),
noting that the dielectric cylinder has zero total charge, so its potential goes to zero at large r. The coefficients An can be evaluated by noting that the radial component of the total electric displacement field D = E is continuous across the boundary r = a,
∂V (a−) ∂r
= ′^
∂V (a+) ∂r
n=
[
2 q ′a
( (^) a b
)n
nAn a
]
cos nθ = ′^
n=
[
2 q ′a
( (^) a b
)n −
nAn a
]
cos nθ, (16)
An = −
2 q ′n
( (^) a
b
)n , (17)
Vcylinder(r, θ) =
− (^2) q′^ − ′ +′
n=
1 n
( (^) r b
)n cos nθ (r < a), 2 q ′
−′ +′^ (−^ ln^ r)^ −^
2 q ′
−′ +′
[
− ln r +
n=
1 n
a^2 /b r
)n cos nθ
]
(r > a),
Outside the dielectric cylinder of radius a the total potential Vwire + Vcylinder is the same as if the media were vacuum and the cylinder were replaced by a wire at the origin of charge q( − ′)/′( + ′) per unit length, and an oppositely charged wire at (r, θ) = (a^2 /b, 0). Inside the cylinder, the potential is, to within a constant, as if charge per unit length −q( − ′)/′( + ′) had been added to the effective charge q/′^ per unit length of the original wire, bringing its effective linear charge density to 2q/( + ′). That is, an image-method holds for this example. In the limit → ∞, while ′^ → 1, we obtain the image prescription for a charged wire outside an neutral, conducting cylinder; the potential outside the cylinder is as if it were replaced by a wire of linear charge density −q at (r, θ) = (a^2 /b, 0) plus a wire of line charge density q along the z-axis, and the potential inside the cylinder is constant. Another limit of interest is that the radius a of the dielectric cylinder goes to infinity while distance d = b − a remains constant, such that the dielectric cylinder becomes a half space. For the potential outside the dielectric, the line charge is at depth a − a^2 /b → d below the planar surface of the dielectric, as previously found in sec. 2.1.
2.1.3 Point Charge Outside a Dielectric Sphere
We finally consider the case of a dielectric sphere of radius a and relative permittivity when a point charge q per unit length is located in vacuum at distance b > a from the center of the cylinder.
In this three dimensional problem we take the center of the sphere to be the origin, and take the position of the point charge to be (r, θ, φ) = (b, 0 , 0) in a spherical coordinate system. The geometry is azimuthally symmetric, so the potential does not depend on coordinate φ. The potential due to the point charge q in the absence of the dielectric sphere has the general form,
Vq(r, θ, φ.) =
n=0 an
(r b
)n Pn(cos θ) (r < b), ∑ n=0 an
( (^) b r
)n+ Pn(cos θ) (r > b),
The potential of the dielectric sphere does not quite have the form of the potential due to point charges. Hence, there is no general image method for the case of a point charge outside a dielectric sphere.^6 In the limit → ∞ while ′^ → 1 we do obtain the image prescription for a point charge outside an neutral, conducting sphere; the potential outside the cylinder is as if it were replaced by a point charge −qa/b at (r, θ, φ) = (a^2 /b, 0 , 0) plus a charge qa/b at the origin, and the potential inside the sphere is constant.
2.2 Conducting Dielectric Media
In general a dielectric medium has a small electrical conductivity σ. Then, if an electric field exists inside the medium, a free-current density flows according to,
Jfree = σE =
σD
where is the (relative) permittivity.^7 Conservation of free charge can then be expressed as,
∂ρfree ∂t
= −∇ · Jfree = −
σ∇ · D
4 πσ
ρfree , (30)
such that in the absence of an energy flow to maintain the electric field, the free charge distribution decays according to,
ρfree(t) = ρ 0 e−t/τ^. (31)
with time constant τ = / 4 πσ. For metals this time constant is so short that the above calculation should be modified to include wave motion [18], but for dielectric with low con- ductivity this approximation is reasonable. If an external charge is brought near a conducting dielectric medium, the potentials and associated charge distributions found in sec. 2 apply only for times small compared to the relaxation time τ. At longer times the medium behaves like a static conductor, with zero internal electric field, for which the potentials and charge distributions are the familiar versions for good conductors. For example, glass has (relative) dielectric constant ≈ 4 and electrical conductivity ≈ 10 −^4 Gaussian units, and hence τ (^) glass ≈ 1 hour. Rock has conductivity σ ≈ 106 Gaussian units, and hence τrock ≈ 1 μs. If the time structure of the external charges and currents is known, it is preferable to perform a full time-domain analysis. For a review, see [19]. For an incident wave of angular frequency ω the medium can be approximated as a good conductor if ω 1 /τ and as a nonconductor if ω 1 /τ. For frequencies high enough that the medium is a “good conductor,” the waves are attenuated as they penetrate into the medium over the skin depth,
(^6) It was found by Neumann in 1883 [12] that the problem of a point charge inside/outside a dielectric
sphere can be solved by an image point charge together with an image line charge outside/inside the sphere. However, this result was little known until rediscovered by Lindell (1992) [13, 14], and then re-rediscovered by Norris in 1995 [15]. (^7) For a steady-state example of a conducting dielectric, see [16]. For examples of steady currents in two-
and three-dimensional conducting media in which image methods are relevant, see [17].
d = c/
2 πμωσ, where μ if the relative permeability of the medium. When ω ≈ 1 /τ , δ ≈
/μ c/ 2 πσ, which provides an estimate of the maximum distance a transient field can penetrate into the medium.^8 For glass, this maximum depth is very large, but for rock is of order 1 m.
2.3 Permeable Media
So far as is presently known, (Gilbertian) magnetic charges do not exist in Nature,^9 and all magnetic fields can be associated with (Amperian) electrical currents. The magnetic moments of elementary particles, nuclei and atoms are quantum effects, not well described in “classical” electrodynamics, where one characterizes their presence in macroscopic media by the density M of magnetic moments. Noting that B = H + 4πM and that ∇ · B = 0, one can define an effective magnetic charge density according to ρm,eff = −∇ · M, such that ∇ · H = 4πρm,eff , but isolated effective magnetic charges do not exist and the effective magnetic charge density is only an alternative representation of Amperian currents in bulk matter.^10 Furthermore, ρm,eff = 0 inside linear, isotropic magnetic media, where B = μH, and one can speak only of an surface density of effective magnetic charges. We first consider (secs. 2.3.1-2) image methods for permeable media in two-dimensional cases with a line conduction current I parallel to the z-axis, and then we turn (sec. 2.3.3) to three-dimensional examples with permanent magnetism. Away from the line current, ∇ × H = 0, so we could write H = −∇ΦM where ΦM is a scalar potential. However, when conduction currents are present it seems better to work with the vector potential A, such that B = ∇ × A. We recall that for a current I along the z-axis in a medium of relative permeability μ′, the magnetic field is B = μ′H = 2μ′I θˆ/cr, so the vector potential is A = −(2μ′I/c) ln r ˆz.
2.3.1 Line Current Outside a Permeable Half Space
We first consider the case of a medium with relative permeability μ in the half space x < 0 with a line current I at (x, y) = (a, 0), where otherwise the region x > 0 has relative permeability μ′. The image method is to suppose that the vector potential in the region x > 0 is that due to the original current I at (a, 0) plus an image current I′^ at (−a, 0), and that the potential in the region x < 0 is that due to the original point current plus additional current I′′^ also at (a, 0). According to the suggested image method, the vector potential at x > 0 can be written as,
Az (x > 0) = −
μ′I c
ln[(x − a)^2 + y^2 ] −
μ′I′ c
ln[(x + a)^2 + y^2 ] , (32)
and that at x < 0 as,
Az(x < 0) = −
μ′(I + I′′) c
ln[(x − a)^2 + y^2 ]. (33)
(^8) Compare eq. (7.70) of [3]. This argument seems to have disappeared from the 3rd edition. (^9) For a discussion of electrodynamics if magnetic charges existed, see, for example, [20]. (^10) However, one should not associate a magnetization or effective magnetic charges with “free” conduction
currents. Attempts to do this lead to confusions as exhibited in [21]. See also [22, 23].
In the limit μ → 0 with μ′^ → 1 we obtain the image prescription for a linear current I outside a perfectly conducting cylinder; the vector potential and magnetic fields outside the cylinder is as if it were replaced by a current −I at (r, θ) = (a^2 /b, 0) plus a current I along the z-axis, and the potential inside the cylinder is constant.^11
2.3.3 Point Magnetic Charge, and Point Magnetic Dipole, Outside a Permeable Half Space
In this section we consider permanent magnets in vacuum that are outside a permeable half space z < 0 in which the (relative) permeability is μ. We begin with the case of a point magnetic charge qM (even though these don’t seem to exist in Nature) at (x, y, z) = (0, 0 , a). As in sec. 2.1.1, the image method is to suppose that the magnetic scalar potential ΦM in the region z > 0 is that due to the original point charge qM at (0, 0 , a) plus an image charge q′ M at (0, 0 , −b),
ΦM (x, 0 , z > 0) =
qM [x^2 + (z − a)^2 ]^1 /^2
q M′ [x^2 + (z + b)^2 ]^1 /^2
and that the potential in the region z < 0 (inside the permeable half space) is that due to the original point charge plus a point charge q′′ M at (0, 0 , c),
ΦM (x, 0 , z < 0) = qM [x^2 + (z − a)^2 ]^1 /^2
q M′′ [x^2 + (z − c)^2 ]^1 /^2
Continuity of the potential ΦM across the plane z = 0 requires that,
b = c, and q M′′ = q′ M. (40)
Continuity of Bz across the plane z = 0 requires that Bz (x, 0 , 0 +) = Hz (x, 0 , 0 +) = Bz (x, 0 , 0 −) = μHz (x, 0 , 0 −), i.e., ∂ΦM (x, 0 , 0 +) ∂z
= μ
∂ΦM (x, 0 , 0 −) ∂z
qM a [x^2 + a^2 ]^3 /^2
q′ M b [x^2 + b^2 ]^3 /^2
= μ
qM a [x^2 + a^2 ]^3 /^2
q M′ b [x^2 + b^2 ]^3 /^2
which implies that,
a = b = c, and q′′ M = q M′ = −q
μ − 1 μ + 1
The potential and magnetic field H = B/μ in the permeable region z < 0 are as if the media were vacuum and the original magnetic charge qM were replaced by charge qM + q M′′ = 2 q/(μ + 1). In the region z > 0 the potential and B = H fields are as for the original charge qM plus an image charge q′/′^ = q′′/′, both in vacuum. In the limit that a = 0 (such that charge qM lies on the surface of the permeable medium the H field is as if the media were vacuum but the charge were 2qM /(μ + 1). Note that
(^11) This justifies the choice of signs for the image currents inside the cylinder, as made in the previous
paragraph.
the H field is radial, and the same in both media despite their differing permeabilities, as is consistent with the requirement that the tangential component of H be continuous at the interface z = 0. In the limit that μ → ∞, the potential above the plane is that due to the original charge plus an image charge −q at (0, 0 , −a), and the potential below the plane is zero. However, this is not the image prescription for a point magnetic charge above a perfectly conducting plane, where the B field must be tangential to the surface of the conduction plane such that the image charge is +qM rather than −qM. That is, a magnetic charge is attracted to a half space of infinite permeability, but repelled from a perfectly conducting (or superconducting) plane.
A magnetic dipole m can be regarded (insofar as we are only considered with its effect outside the dipole) as due to a pair of equal and opposite magnetic charges ±qM separated by distance d = m/qM. If this dipole is in vacuum outside a permeable half space, then the image method of point magnetic charges tells us that the potential and fields of the dipole in vacuum are as if the permeably medium were also vacuum but with an image magnetic dipole,
m′^ = (m⊥ − m‖)
μ − 1 μ + 1
, where m = m⊥ + m‖. (44)
And, the H field inside the permeable medium is as if the medium were vacuum but the original magnetic dipole had strength 2m/(μ + 1).^12 For a high-permeability medium (μ 1) the image dipole is m′^ = m⊥ − m‖, such that for a permanent magnet with magnetization perpendicular to the surface of the permeable medium, the image magnet has the same magnetization m = m⊥ as the actual magnet.
3 Forces
When an object is embedded in a surrounding medium there is an ambiguity as to whether or not forces on the adjacent surface of the medium surrounding the object should be counted at part of the force on the object itself. So, we first consider the dielectric examples of sec. 2.1 where ′^ = 1.
3.1 Point Charge Outside a Dielectric Half Space
3.1.1 ′^ = 1
It seems obvious that the force on the charge q at z = a in the presence of a medium of relative permittivity at z < 0 can be computed as that due to the electric field of the image charge q′,
Fq = qEq′ = −
q^2 4 a^2
ˆz, (45)
(^12) These results are discussed in [2, 24].
towards positive and negative z, respectively. Since the stress tensor varies as 1/r^4 on these infinite hemispheres there is no contribution to the forces from these surfaces, so the force on the charge q is equal and opposite to that on the medium at z < 0. The force on the latter is,
Fz (z < 0) =
z=0+
Tzz dAreaz =
4 π
z=0+
E z^2 (z = 0+) −
E^2 (z = 0+) 2
dArea
4 π
0
π dr^2
4 q^2 ^2 a^2 ( + 1)^2 (r^2 + a^2 )^3
2 q^2 (r^2 + ^2 a^2 ) ( + 1)^2 (r^2 + a^2 )^3
q^2 2( + 1)^2
0
dr^2
^2 a^2 − r^2 (r^2 + a^2 )^3
q^2 2( + 1)^2
[
^2 a^2 2 a^4
a^2
a^2 2 a^4
)]
q^2 4 a^2
using Dwight 91.3, which is the same result as found in eq. (51).^13 This reinforces the author’s view that the stress tensor is the most reliable method of computation of forces in electromagnetic examples, in that only the total electromagnetic fields need be used (without need for discussion of effective charge densities and effective average fields as in eqs. (46)- (51)).^14
3.1.2 ′^ = 1
When both and ′^ are different from 1 the electric field on both sides of the interface z = 0 are, recalling sec. 2.1.1,
E(z = 0−) =
q (r^2 + a^2 )^3 /^2
′^ +
(r ˆr − a ˆz) , (54)
and,
E(z = 0+) =
q ′(r^2 + a^2 )^3 /^2
(r ˆr − a ˆz) −
q ′(r^2 + a^2 )^3 /^2
+ ′^
(r ˆr + a ˆz)
2 q ′( + ′)(r^2 + a^2 )^3 /^2
(′r ˆr − a ˆz), (55)
so that Er is continuous across the interface. The bound surface charge densities are then,
σbound(z = 0−) = P(z = 0−) · nˆ−^ =
4 π
E(z = 0−) · ˆz = −
qa( − 1) 2 π( + ′)(r^2 + a^2 )^3 /^2
and,
σbound(z = 0+) = P(z = 0+) · ˆn+^ = −
′^ − 1
4 π
E(z = 0+) · ˆz =
qa(′^ − 1) 2 π′( + ′)(r^2 + a^2 )^3 /^2
(^13) A peculiar use of the stress tensor was made in sec. IIIC of [25], which led to the erroneous conclusion
that a new form of the stress tensor must be invented to compute forces on dielectrics. (^14) For other examples that illustrate this view, see [26, 27].
The electric field component Ez at z = 0 can be defined as,
Ez (z = 0) = Ez (z = 0−) + 4πσbound(z = 0−) = Ez (z = 0+) − 4 πσbound(z = 0+)
= −
2 qa ( + ′)(r^2 + a^2 )^3 /^2
so the electric field exactly at the interface is,
E(z = 0) =
2 q ( + ′)(r^2 + a^2 )^3 /^2
(r ˆr − a ˆz). (59)
The force on the medium at z < 0 can be computed as,
Fz (z < 0) = σbound(z = 0−)
Ez (z = 0−) + Ez (z = 0) 2
q^2 a^2 2 π
( + ′)^2
dArea (r^2 + a^2 )^3
=
q^2 a^2 2
^2 − 1
( + ′)^2
0
d r^2 (r^2 + a^2 )^3
q^2 4 a^2
^2 − 1
( + ′)^2
and also via the stress tensor at the interface (where D = E) as,
Fz (z < 0) =
z=
Tzz dAreaz =
4 π
z=0+
E^2 z (z = 0) −
E^2 (z = 0) 2
dArea
4 π
0
π dr^2
4 q^2 ^2 a^2 ( + ′)^2 (r^2 + a^2 )^3
2 q^2 (r^2 + ^2 a^2 ) ( + ′)^2 (r^2 + a^2 )^3
q^2 2( + ′)^2
0
dr^2
^2 a^2 − r^2 (r^2 + a^2 )^3
q^2 2( + ′)^2
[
^2 a^2 2 a^4
a^2
a^2 2 a^4
)]
q^2 4 a^2
^2 − 1
( + ′)^2
The force on all material at z > 0, i.e., the original charged wire plus the medium with relative permittivity ′, is equal and opposite to that of eqs. (60)-(61). The issue of the force on the “point” charge q is delicate in macroscopic electrodynamics, as the size of a “point” charge is smaller than the length scale over which the macroscopic averages are taken. Following Lord Kelvin (for example, p. 397 of [28]), we can suppose the charge resides in a cavity that is otherwise vacuum and compute the electric field in this cavity due to the bound charge densities outside it. Of course, the result is famously dependent on the shape of the cavity. For a spherical cavity, the field inside is given by,^15
Ein =
2 ′^ + 1
Eout, (62)
where Eout is the field at the position of the cavity (in a medium with relative permittivity ′) in its absence. In the present example, it seems reasonable to take Eout to be the field at the position of the charge q due to the image charge q′^ at z = −a,
Eout = −
q 4 ′a^2
+ ′^
ˆz, (63)
(^15) See, for example, sec. 4.4 of [3].
3.5 Permanent Magnet in Contact with a High Permeability Medium
(Refrigerator Magnet)
If the permanent magnet is magnetized perpendicular to the surface of the high permeability medium, then the magnetization of the image magnet is the same as that of the original, according to sec. 2.3.3. The magnet is attracted to the medium with a force
B^2 dArea/ 8 π, where B is the magnetic field at the symmetry plane of the original magnet plus its image. For example, if the original magnet is a hemisphere of radius a and uniform magnetization M perpendicular to its base, the combination of original plus image magnet is spherical, with well-known fields, and the attractive force is π^2 a^2 M^2 [29].
References
[1] W. Thomson, Geometrical Investigations with Reference to the Distribution of Elec- tricity on Spherical Conductors, Camb. Dublin Math. J. 3 , 141 (1848), http://physics.princeton.edu/~mcdonald/examples/EM/thomson_cdmj_3_131_48.pdf
[2] P. Hammond, Electric and Magnetic Images, Proc. IEE 107 , 306 (1960), http://physics.princeton.edu/~mcdonald/examples/EM/hammond_piee_107_306_60.pdf
[3] J.D. Jackson, Classical Electrodynamics, 2nd^ ed. (Wiley, 1975), http://physics.princeton.edu/~mcdonald/examples/EM/jackson_ce2_75.pdf http://physics.princeton.edu/~mcdonald/examples/EM/jackson_ce3_99.pdf
[4] T. Sometani, Image method for a dielectric plate and a point charge, Eur. J. Phys. 21 , 549 (2000), http://physics.princeton.edu/~mcdonald/examples/EM/sometani_ejp_21_549_00.pdf
[5] V. Ivchenko, On the interaction force between a point charge and an infinite dielectric plate of finite thickness, Eur. J. Phys. 41 , 015201 (2020), http://physics.princeton.edu/~mcdonald/examples/EM/ivchenko_ejp_41_015201_20.pdf
[6] K.T. McDonald, Doubly Negative Metamaterials (Apr. 14, 20010), http://physics.princeton.edu/~mcdonald/examples/metamaterials.pdf
[7] Y. Urzhumov et al., Magnetic levitation of metamaterial bodies enhanced with magne- tostatic surface resonances, Phys. Rev. B 85 , 054430 (2012), http://physics.princeton.edu/~mcdonald/examples/EM/urzhumov_prb_85_054430_12.pdf
[8] D.J. Griffiths, Introduction to Electrodynamics, 4th^ ed. (Pearson, 2012), http://physics.princeton.edu/~mcdonald/examples/EM/griffiths_em3.pdf
[9] W.R. Smythe, Static and Dynamic Electricity, 3rd^ ed. (McGraw-Hill, 1968), http://physics.princeton.edu/~mcdonald/examples/EM/smythe_50.pdf
[10] L.D. Landau, E.M. Lifshitz and L.P. Pitaevskii, Electrodynamics of Continuous Media, 2 nd^ ed. (Butterworth-Heinemann, 1984), http://physics.princeton.edu/~mcdonald/examples/EM/landau_ecm2.pdf
[11] I.V. Lindell and K.I. Nikoskinen, Two-dimensional image method for time-harmonic line current in front of a material cylinder, Elec. Eng. 81 , 357 (1999), http://physics.princeton.edu/~mcdonald/examples/EM/lindell_ee_81_357_99.pdf
[12] C. Neumann, Hydrodynamische Untersuchungen nebst einem Anhange ¨uber die Prob- leme der Elektrostatik und der Magnetischen Induction (Teubner, 1883), pp. 272-282, http://physics.princeton.edu/~mcdonald/examples/EM/neumann_83.pdf
[13] I.V. Lindell, Electrostatic image theory for the dielectric sphere, Radio Sci. 27 , 1 (1992), http://physics.princeton.edu/~mcdonald/examples/EM/lindell_rs_27_1_92.pdf
[14] I.V. Lindell, Image theory for electrostatic and magnetostatic problems involving a material sphere, Am. J. Phys. 61 , 39 (1993), http://physics.princeton.edu/~mcdonald/examples/EM/lindell_ajp_61_39_93.pdf
[15] W.T. Norris, Charge images in a dielectric sphere IEE Proc. Sci. Meas. Tech. 142 , 142, 495 (1995), http://physics.princeton.edu/~mcdonald/examples/EM/norris_ieepsmt_142_142_95.pdf
[16] K.T. McDonald and T.J. Maloney, Leaky Capacitors (Oct. 7, 2001), http://physics.princeton.edu/~mcdonald/examples/leakycap.pdf
[17] K.T. McDonald, Currents in a Conducting Sheet with a Hole (Feb. 22, 2006), http://physics.princeton.edu/~mcdonald/examples/panofsky_7-3.pdf
[18] H.C. Ohanian, On the approach to electro- and magneto-static equilibrium, Am. J. Phys. 51 , 1020 (1983), http://physics.princeton.edu/~mcdonald/examples/EM/ohanian_ajp_51_1020_83.pdf
[19] R.T. Shuey and M. Johnson, On the Phenomenology of Electrical Relaxation in Rocks, Geophysics 38 , 37 (1973), http://physics.princeton.edu/~mcdonald/examples/EM/shuey_geophysics_38_37_73.pdf
[20] K.T. McDonald, Poynting’s Theorem with Magnetic Monopoles (May 1, 2013), http://physics.princeton.edu/~mcdonald/examples/poynting.pdf
[21] M. Mansuripur, Trouble with the Lorentz Law of Force: Incompatibility with Special Relativity and Momentum Conservation, Phys. Rev. Lett. 108 , 193901 (2012), http://physics.princeton.edu/~mcdonald/examples/EM/mansuripur_prl_108_193901_12.pdf
[22] K.T. McDonald, Mansuripur’s Paradox (May 2, 2012), http://physics.princeton.edu/~mcdonald/examples/mansuripur.pdf
[23] K.T. McDonald, Biot-Savart vs. Einstein-Laub Force Law (May 21, 2013), http://physics.princeton.edu/~mcdonald/examples/laub.pdf
[24] M. Beleggia, D. Vokoun and M. De Graef, Forces Between a Permanent Magnet and a Soft Magnetic Plate, IEEE Mag. Lett. 3 , 0500204 (2012), http://physics.princeton.edu/~mcdonald/examples/EM/beleggia_ieeeml_3_0500204_12.pdf
