On Magnetic Forces and Work

Jacob A. Barandes1, ∗
1
Jefferson Physical Laboratory, Harvard University, Cambridge, MA 02138
(Dated: October 10, 2020)
We address a long-standing debate over whether classical magnetic forces can do work, ultimately
answering the question in the affirmative. In detail, we couple a classical particle with intrinsic spin
and elementary dipole moments to the electromagnetic field, derive the appropriate generalization
of the Lorentz force law, show that the particle’s dipole moments must be collinear with its spin
axis, and argue that the magnetic field does mechanical work on the particle’s elementary magnetic
dipole moment. As consistency checks, we calculate the overall system’s energy-momentum and
angular momentum, and show that their local conservation equations lead to the same force law
and therefore the same conclusions about magnetic forces and work. We also compute the system’s
Belinfante-Rosenfeld energy-momentum tensor.

I. INTRODUCTION A. The Phase Space for a Relativistic Massive

Particle with Spin

Textbook treatments and research articles on classi-

Following [4–6], we model the particle’s kinematics us-
cal electromagnetism, such as [1, 2], often suggest that
ing spacetime coordinates X µ = (c T, X)µ , energy E,
magnetic fields cannot do mechanical work. On the other
four-momentum pµ = (E/c, p)µ , positive inertial mass
hand, everyday examples of bar magnets lifting other bar
m > 0, and antisymmetric spin tensor S µν by identi-
magnets would seem to suggest otherwise. In this paper,
fying the particle’s phase space as a transitive or “irre-
we show that there exists a classical way to understand
ducible” group action (or homogeneous space) of the or-
how magnetic fields can indeed do work.[3]
thochronous Poincaré group. The states in this phase
We start in Section II with a review of the kinematics of space take the form (X, p, S) and are each obtained
classical relativistic point particles with intrinsic spin and from the reference state (0, (mc, 0), S0 ) by an appropriate
permanent, elementary dipole moments. In Section III, Poincaré transformation (a, Λ) ∈ R4 n O(1, 3) according
we couple a particle of this kind to the electromagnetic to
field and derive its dynamics, showing, in particular, that
magnetic forces can classically do work on the particle via (X, p, S) = (a, Λ(mc, 0), ΛS0 ΛT ). (1)
its elementary magnetic dipole moment. We also show as
a matter of self-consistency that the particle’s elementary Here the coordinates X µ = aµ and the variable Lorentz-
dipole moments must be collinear with the particle’s in- transformation matrix Λµν are treated as the particle’s
trinsic spin. In Section IV, we derive expressions for the fundamental phase-space variables, with the condition
overall system’s energy-momentum and angular momen- that ΛT ηΛ = η = diag(−1, +1, +1, +1).
tum, and show that their associated conservation laws
lead to the same equations of motion as before, thereby
providing further confirmation that magnetic fields can B. Charge and Elementary Dipole Moments
do work on a particle with elementary dipole moments.
We conclude with one more new result by calculating the
system’s Belinfante-Rosenfeld energy-momentum tensor. We can couple the particle to the electromagnetic field
by assigning the particle an electric-monopole charge q
and an antisymmetric elementary dipole tensor mµν , so
that the particle is an elementary dipole.
We note that elementary dipoles of this kind are nei-
ther of the Ampère model, which consist of loops of mov-
II. THE KINEMATICS OF A RELATIVISTIC
ing electric monopoles, nor of the Gilbert model, which
ELEMENTARY DIPOLE
consist of pairs of hypothetical magnetic monopoles. In
particular, the elementary dipoles that we examine here
To start, we will need a relativistic description of the represent a classical extension of Maxwell’s original the-
kinematics of a classical particle with intrinsic spin. ory of electromagnetism, as Maxwell’s theory includes
dipoles only of the Ampère type.[7]
We let uµ ≡ dX µ /dλ denote the particle’s four-velocity
and γ ≡ u0 /c denote the particle’s associated Lorentz
factor, where uµ is not generically normalized to u2 =
−c2 unless the worldline parameter λ is taken to be the
∗ barandes@physics.harvard.edu particle’s proper time τ . The particle’s four-velocity then
2

takes the form III. THE DYNAMICS OF A RELATIVISTIC

ELEMENTARY DIPOLE
uµ = (γc, γv)µ . (2)
Next, we turn to a discussion of the particle’s dynam-
The particle has four-dimensional electric-monopole cur-
ics.
rent density

1 A. The Action Functional for a Relativistic

jeν (x, t) = (ρe (x, t)c, Je (x, t))ν = quν δ 3 (x − X) (3) Massive Particle with Spin
γ
and elementary-dipole density In the absence of external interactions, as shown in [4–
1 6], we can encode the dynamics of a particle with intrinsic
M µν = mµν δ 3 (x − X), (4) spin in terms of the manifestly covariant action functional
γ
Z
1
with overall current density Sparticle [X, Λ] = dλ Jµν θ̇µν
2
j ν (x, t) = jeν (x, t) + ∂µ M µν (x, t),
 
(5)
Z
1
= dλ pµ Ẋ µ + Tr[S Λ̇Λ−1 ] , (13)
2
where (1/γ)δ 3 (x−X) is the Lorentz-invariant form of the
three-dimensional Dirac delta function. where λ is a smooth and monotonic but otherwise ar-
It follows immediately from (3) that the parti- bitrary parameter along the particle’s worldline, Jµν =
cle’s electric-monopole density ρe = jet /c, its electric- Lµν + Sµν is the particle’s total angular-momentum ten-
monopole current density Je = (jex , jey , jez ), and its three- sor, Lµν ≡ Xµ pν −Xν pµ is its orbital angular-momentum
velocity v ≡ dX/dt satisfy the basic relationship tensor, θµν is an antisymmetric tensor of boost and angu-
lar degrees of freedom, and we ignore irrelevant bound-
Je = ρe v. (6) ary terms. Consistency of the particle’s dynamics with
the required invariance of the quantities p2 ≡ −m2 c2
We emphasize that no such relationship holds for the par-
and s2 ≡ (1/2)Sµν S µν requires the auxiliary phase-space
ticle’s elementary dipole moments, which, again, are not
condition
assumed to arise from any underlying motion of electric
monopoles.
pµ S µν = 0. (14)
As in [2], by introducing suitable four-vectors π µ and
µµ and antisymmetric tensors
1 µ ν B. The Particle’s Equations of Motion
π µν ≡ (p π − pν π µ ), (7)
mc
1 µνρσ Our next step will be to couple the particle to the elec-
µµν ≡  pρ µσ , (8)
mc tromagnetic field and obtain the particle’s equations of
motion, from which we will be able to infer the appropri-
we can write the particle’s elementary dipole tensor in ate generalization of the Lorentz force law.
terms of an electric part π µν and a magnetic part µµν as Given the charge and elementary dipole moments out-
lined above, the overall action functional for the elemen-
mµν = π µν + µµν , (9)
tary dipole and the electromagnetic field is given by
or, equivalently, as
µν S[X, Λ, A] ≡ Sparticle [X, Λ] + Sfield [A] + Sint [X, Λ, A]
0 cπx cπy cπz

−cπx 0 −µz µy 
mµν ≡ . (10)
−cπy µz 0 −µx  Z 
1

−cπz −µy µx 0 = dλ pµ Ẋ µ + Tr[S Λ̇Λ−1 ] (Sparticle )
2
Here µνρσ is the four-dimensional Levi-Civita symbol
Z Z  
1 µν
+ dt d3 x − F Fµν (Sfield )
(with txyz ≡ +1), and π ν (λ) and µµ (λ) are related to 4µ0
their reference values π0µ ≡ (0, π 0 )µ and µµ0 ≡ (0, µ0 )µ Z Z
and the particle’s variable Lorentz-transformation matrix + dt d3 x j ν Aν (Sint ),
Λµν (λ) according to (15)
π µ (λ) ≡ Λµν (λ)π0ν , (11) where Fµν ≡ ∂µ Aν − ∂ν Aµ is the usual Faraday tensor,
µ
µ (λ) ≡ Λµν (λ)µν0 . (12) j ν = jeν + ∂µ M µν is the particle’s overall current density,
 3

and the interaction term in the final line ensures that ex- equations of motion (20)–(21) reduce to
tremizing the action functional with respect to the elec-
tromagnetic gauge field Aµ yields the Maxwell equations dE
≈ v · (qEext + ∇(π · Eext + µ · Bext )), (22)
in their usual form. The first line in this action func- dt
tional (Sparticle ) is fixed by group theory, the second line dp
(Sfield ) defines the vacuum in the pure Maxwell theory, ≈ q(Eext + v × Bext ) + ∇(π · Eext + µ · Bext ),
dt
and the third line (Sint ) provides the canonical coupling (23)
between the particle and the electromagnetic field in a dJ dp
manner consistent with the Maxwell equations and the ≈X× + π × Eext + µ × Bext , (24)
dt dt
particle’s features as laid out in the previous section.
After an integration by parts, we can write the inter- where the particle’s four-momentum in this limit is
action term in the final line as
Z Z   pµ = (E/c, p)µ ≈ (mc2 + (1/2)mv2 , p)µ , (25)
1
Sint [X, Λ, A] = dt d3 x jeν Aν − M µν Fµν . (16)
2 and the particle’s overall angular-momentum pseudovec-
Collecting together all the terms that involve the parti- tor J is made up of orbital and spin contributions accord-
cle’s degrees of freedom, we obtain ing to

J ≡ L + S = (Lyz , Lzx , Lxy ) + (S yz , S zx , S xy ).

Z  
µ 1 −1 (26)
Sparticle+int [X, Λ, A] = dλ pµ Ẋ + Tr[S Λ̇Λ ]
2
Z Z Z Z   The dynamical equation (23) tells us that the electro-
3 ν 3 1
+ dt d x je Aν + dt d x − M µν Fµν , magnetic force on the particle is
2
(17) F = qEext + qv × Bext + ∇(π · Eext ) + ∇(µ · Bext ). (27)
which we can further reduce to the form
Z We observe that the usual Lorentz force law, qEext +
Sparticle+int [X, Λ, A] = dλ Lparticle+int , (18) qv × Bext , is enhanced in the presence of the particle’s
elementary dipole moments by the appearance of two ad-
ditional dipole terms, ∇(π · Eext ) + ∇(µ · Bext ), in which
for a manifestly covariant Lagrangian defined by
the magnetic field appears on an equal footing with the
1 electric field. Accordingly, the magnetic field contributes
Lparticle+int ≡ pµ Ẋ µ + Tr[S Λ̇Λ−1 ]
2 to theR work done by the external electromagnetic field,
ν 1p W ≡ dX · F:
+ q Ẋ Aν − −Ẋ 2 mµν Fµν . (19)
2c Z B
It follows from a straightforward calculation that the par- W = dt (qv · Eext ) + ∆(π · Eext ) + ∆(µ · Bext ). (28)
ticle’s equations of motion, expressed in terms of the par- A
ticle’s proper time τ , are then
Moreover the rate at which work is done is in agreement
with the dynamical equation (22).
dp µ 1 1 d We have reached the key conclusion of this pa-
= −quν F νµ − mρσ ∂ µ Fρσ − 2 (uµ mρσ Fρσ )
dτ 2 2c dτ per—namely, that magnetic forces can do work on clas-
1 sical particles with elementary dipole moments.[11] We
= −quν F νµ − mρσ (η µν + uµ uν )∂ν Fρσ
2 next turn to a detailed treatment of self-consistency con-
1 d µ ρσ ditions on the particle’s dynamics, as well as obtain the
− 2 (u m )Fρσ , (20) necessary formulas for determining the particle’s four-
2c dτ
velocity uµ in the presence of a nonzero electromagnetic
as obtained in [6, 8, 9], and field. Later on, we will analyze electromagnetic forces
dS µν and work done on the particle from the standpoint of
= −(uµ pν − uν pµ ) − (mµρ F νρ − mνρ F µρ ), (21) local conservation laws.
dτ
which generalizes the results of [6, 8, 10].
D. Implications of Self-Consistency

C. The Non-Relativistic Limit with

Time-Independent External Fields Taking a derivative of the phase-space condition pµ S µν
from (14) yields the self-consistency requirement
In the non-relativistic limit and ignoring self-field ef-
fects—so that we can replace the overall electric and mag- dpµ µν dS µν
netic field with the external fields Eext and Bext —the S + pµ = 0,
dτ dτ
4

which entails that the particle’s four-momentum pµ and Here Ξ is a pseudoscalar and Γ is the particle’s scalar gy-
its four-velocity uµ = dX µ /dτ (now normalized to u2 = romagnetic ratio. We can understand these relationships
−c2 ) are related by physically as telling us that if the particle’s elementary-
dipole vectors were not collinear with the particle’s spin
pµ = meff uµ + bµ . (29) axis, then torques exerted on the particle by the electro-
magnetic field would cause the particle’s overall spin to
Here meff , which plays the role of an effective mass, is speed up or slow down, in violation of the constancy of
defined by s2 .
m2 c2
meff ≡ − , (30)
p·u IV. CONSERVATION LAWS

and the four-vector bµ , which is orthogonal to the parti-

cle’s four-momentum, b · p = 0, is given by For completeness, we verify that the equations of mo-
tion (20)–(21) also follow from local conservation of
1

dpν νµ
 energy-momentum and angular momentum. To begin,
bµ ≡ S − pν (mνρ F µρ − mµρ F νρ ) . (31) we recall the relevant version of Noether’s theorem, which
p · u dτ
states that if a system’s dynamics has a continuous sym-
As in [6], we regard (29) as an implicit formula for the metry,
particle’s four-velocity uµ . This formula ensures, in par-
ticular, that the particle’s four-momentum pµ has con- qα 7→ qα0 = qα + δ qα ,
stant norm-squared p2 = −m2 c2 .
X
δ qα = gqα ,b b , (37)
For vanishing field, Fµν = 0, the relationship (29) re- b
duces to the familiar equation pµ = muµ , as expected.
On the other hand, when the electromagnetic field is where the quantities b parameterize the symmetry and
nonzero, Fµν 6= 0, (29) has the form the quantities gqα ,b characterize its precise form, then we
have the following conservation law:
pµ = muµ + (terms of order 1/c2 ). (32)
X ∂L dQ
Qb ≡ gq ,b − fb , = 0. (38)
This relation ensures that there is no ambiguity over α
∂ q̇α α dt
whether we should identify the particle’s energy E as pt c
or ut mc2 for the purposes of quantifying the work done Here Qb are a set of conserved quantities, L is the sys-
by the field on the particle in the non-relativistic regime. tem’s Lagrangian, qα are its degrees of freedom, and the
Invoking the spin tensor’s equation of motion (21), to- functions fb are related to the change in the Lagrangian
gether with the phase-space condition (14), pµ S µν = 0, according to
and the constancy of the particle’s spin-squared scalar
s2 ≡ (1/2)Sµν S µν , we find L 7→ L + δ L,
X 
  d X dfb
d 2 d 1 δ L = fb b = b . (39)
(s ) = Sµν S µν dt dt
dτ dτ 2 b b

= (S ρµ mµσ − S σµ mµρ )Fρσ = 0, (33)

A. Local Conservation of Energy-Momentum
which yields the condition

S ρµ mµσ = S σµ mµρ . (34) In order to employ Noether’s theorem to obtain the

overall system’s energy-momentum tensor, we examine
In the particle’s reference state, this equality produces the behavior of the system under a translation in space-
the relations time by an infinitesimal four-vector µ . The particle’s
phase-space variables transform as
π 0 × S0 = 0,

(35)
µ0 × S0 = 0,
X µ (λ) 7→ X 0µ (λ) ≡ X µ (λ) + µ ,
)

which dictate that the particle’s elementary electric and (40)

Λµν (λ) 7→ Λ0µν (λ) ≡ Λµν (λ),
magnetic dipole moments must be collinear with the par-
ticle’s spin pseudovector S0 : and the electromagnetic gauge potential transforms as

1
π 0 = Ξ S0 ,  Aµ (x) 7→ A0µ (x) ≡ Aµ (x − )
c (36)
µ0 = Γ S0 .
 = Aµ (x) − ∂ν Aµ (x)ν . (41)
 5

By an application of Noether’s theorem to the parti- tive of the particle’s four-momentum pν :

cle’s manifestly covariant Lagrangian L ≡ Lparticle+int
dpν
Z
defined by (18) and the Lagrangian density L for the 1 d
= d3 x Tcan,particle
tν
overall systemR defined in terms of the action functional dt c dt
S[X, Λ, A] ≡ dt d3 x L from (15), one finds that the
R Z
1 d
overall system’s conserved four-momentum is expressible =− d3 x Tcan,field
tν
c dt
as Z   
1 2
= d3 x − ∂µ H µρ F νρ − η µν F
4µ0
Z
∂L ∂L
Pν = gX ρ ,ν + d3 x (−nµ ) gA ,ν − fν
∂ Ẋ ρ ∂(c∂µ Aρ ) ρ 1 d ν ρσ
− 2 (u m Fρσ )
1 2c dt
= pν + qAν + 2 uν mστ Fστ 1 d
2c  = −quµ F µν + mρµ ∂ µ F νρ − 2 (uν mρσ Fρσ ).
2c dτ
Z  
1 1 ρσ
+ d x (−nµ ) H µρ ∂ν Aρ − δνµ
3
F Fρσ
c 4µ0 After invoking the electromagnetic Bianchi identity
Z
1 ∂ µ F νρ + ∂ ρ F µν + ∂ ν F ρµ = 0, we obtain the equation
= d3 x (−nµ )Tcan,ν
µ
, (42)
c of motion (20).
Our formulas above for the overall system’s canonical
where nµ ≡ (−1, 0)µ is a unit timelike vector orthogonal energy-momentum tensor are new results. By replicating
to the three-dimensional spatial hypersurface of integra- the particle’s equation of motion (20), they provide fur-
tion. In this expression, the overall system’s canonical ther support for the key claim of this paper—that mag-
energy-momentum tensor is given by netic forces can classically do work on particles with ele-
mentary dipole moments.
µν µν µν
Tcan = Tcan,particle + Tcan,field , (43)

with the contributions from the particle and the field B. Local Conservation of Angular Momentum
given respectively by[12]
Next, we use Noether’s theorem to examine the overall
µν 1 system’s angular momentum and its local conservation.
Tcan,particle ≡ u p δ 3 (x − X)
µ ν
(44)
γ Under an infinitesimal Lorentz transformation

and
i
1 2 Λinf = 1 + ρσ σρσ , (47)
µν
Tcan,field ≡ H µρ F νρ − η µν F 2
4µ0
1 1 the particle’s phase-space variables transform as
+ 2 uµ uν mρσ Fρσ δ 3 (x − X)
2c γ
X µ (λ) 7→ X 0µ (λ) ≡ (Λinf X(λ))µ

+ ∂ρ (H µρ Aν ). (45)


i 

= X µ (λ) + ρσ [σρσ ]µν X ν (λ),


2

Here H µν is the auxiliary Faraday tensor: (48)
Λµν (λ) 7→ Λ0µν (λ) ≡ (Λinf Λ(λ))µν 



1 µν i 
H µν ≡ F + M µν = Λµν (λ) + ρσ [σρσ ]µλ Λλν (λ).


µ0 2
1 µν 1
= F + mµν δ 3 (x − X). (46) The second of these two transformation laws is equiva-
µ0 γ lent to the following transformation rule for the particle’s
Lorentz parameters θµν (λ):
The last term in (45) is a total spacetime divergence with
vanishing divergence ∂µ ∂ρ (H µρ Aν ) = 0 on its µ index, θµν (λ) 7→ θ0µν (λ) ≡ θµν (λ) + µν . (49)
and its temporal component ∂ρ (H tρ Aν ) has vanishing in-
tegral over three-dimensional space under the assumption Meanwhile, the gauge field Aµ (x) transforms as
that the fields go to zero sufficiently rapidly at spatial in-
finity. We emphasize that in our approach, all the terms
Aµ (x) 7→ A0µ (x) ≡ (A(Λ−1 −1
inf x)Λinf )µ
in the overall system’s canonical energy-momentum ten-
sor follow from the systematic application of Noether’s ≡ Aλ ((1 − (i/2)ρσ σρσ )x)(δµλ − (i/2)ρσ [σρσ ]λµ )
theorem to the relevant action functionals. = Aµ (x) − ∂ν Aµ (x)(i/2)ρσ [σρσ ]ν λ xλ
µν
We can integrate the local conservation law ∂µ Tcan =0
over three-dimensional space to compute the time deriva- − Aλ (x)(i/2)ρσ [σρσ ]λµ . (50)
6

Noether’s theorem (38) then yields the system’s overall tive of the particle’s spin tensor as follows:
angular-momentum tensor, up to an overall minus sign: dS νρ d
Z
tνρ
= d3 x Sparticle
dt dt
Z
∂L 1 ∂L d 1
− Jνρ = gX α ,νρ + g αβ =− d3 x (xν Tcan
tρ
− xρ Tcan
tν
+ H tν Aρ − H tρ Aν )
α 2 ∂ θ̇αβ θ ,νρ dt c
Z ∂ Ẋ Z
∂L = − d3 x ∂µ (xν Tcan µρ
− xρ Tcan
µν
+ H µν Aρ − H µρ Aν )
+ d3 x (−nµ ) gA ,νρ − fνρ
∂(c∂µ Aα ) α
1 1
= − (uν pρ − uρ pν ) − (mνσ F ρσ − mρσ F νσ ).
 
1
= − pα + qAα − (−uα /c )m Fσλ (Xν δρα − Xρ δνα )
2 σλ
γ γ
2
We therefore see that local conservation of angular mo-
− Sνρ
Z    mentum yields the equation of motion (21).
1 3 µα µ 1 2 C. The Belinfante-Rosenfeld Energy-Momentum
− d x (−nµ ) H − δσ F
c 4µ0 Tensor
× ∂σ Aα (xν δρσ − xρ δνσ )
1
Z The overall system’s canonical energy-momentum ten-
− d3 x (−nµ )(H µν Aρ − H µρ Aν ) sor (43) is not symmetric on its two indices, a feature that
c is required of the energy-momentum tensor that locally
Z
= − d3 x (−nµ )Jcan,νρµ
. (51) sources the gravitational field in general relativity. To
conclude this paper, we follow the standard Belinfante-
Rosenfeld construction[13] to construct a properly sym-
metric energy-momentum tensor, which will likewise rep-
Here we have identified the system’s canonical angular- resent a new result.
momentum flux tensor as We start by introducing a new tensor

c µνρ
µνρ B µρν ≡ (S + S νµρ + S ρµν )
Jcan = Lµνρ + S µνρ , (52) 2
1 1
= −H µρ Aν + (uµ S νρ + uν S µρ + uρ S µν ) δ 3 (x − X).
2 γ
with orbital contribution (57)
We then obtain a symmetric, locally conserved energy-
1 µρ 1 µν momentum tensor T µν for the overall system from the
Lµνρ ≡ xν Tcan − xρ Tcan (53) relation T µν = Tcan
µν
+ ∂ρ B µρν :[14]
c c
1 µ ν 1
T µν = (u p + uν pµ ) δ 3 (x − X)
2 γ
and spin contribution 1 1 1 ρσ
+ H µρ F νρ + H νρ F µρ − η µν F Fρσ
2 2 4µ0
1 1
1 µ νρ 1 3 1 + 2 uµ uν mρσ Fρσ δ 3 (x − X)
S µνρ = u S δ (x − X) + (H µν Aρ − H µρ Aν ). (54) 2c γ
c γ c
1 µνρ νµρ
+ ∂ρ (Sparticle + Sparticle ). (58)
2
We naturally read off the spin flux tensors for the particle In the free-field limit, this energy-momentum tensor re-
and the field respectively as duces to the standard gauge-invariant Maxwell energy-
momentum tensor, as expected:
1 µρ ν 1 ρσ
µνρ 1 1 T µν = F F ρ − η µν F Fρσ . (59)
Sparticle = uµ S νρ δ 3 (x − X), (55) µ0 4µ0
c γ
µνρ 1 µν ρ
Sfield = (H A − H µρ Aν ). (56) ACKNOWLEDGMENTS
c

J. A. B. has benefited from personal communications

µνρ
Integrating the local conservation law ∂µ Jcan = 0 over with Gary Feldman, Howard Georgi, Andrew Stro-
three-dimensional space and taking advantage of the local minger, Bill Phillips, David Griffiths, David Kagan,
µρ
conservation ∂µ Tcan = 0 of the overall canonical energy- David Morin, Logan McCarty, Monica Pate, Alex Lup-
µρ
momentum tensor Tcan , we can compute the time deriva- sasca, and Sebastiano Covone.
 7

[1] D. J. Griffiths, Introduction to Electrodynamics, 4th ed. system of particles, when treated classically, always has a
(Cambridge University Press, 2017). vanishing average magnetization at thermal equilibrium.
[2] S. E. Gralla, A. I. Harte, and R. M. Wald, Physical The theorem’s implication is that phenomena like dia-
Review D 80, 024031 (2009). magnetism can only be understood in terms of quantum
[3] For a more extensive treatment of the results in this pa- effects, a view challenged by our results, at least in prin-
per, see [15]. ciple.
[4] J. A. Barandes, (2019), arXiv:1911.02515. [12] The authors of [2] decompose the overall energy-
[5] J. A. Barandes, (2019), arXiv:1911.08892. momentum tensor by including the interaction terms
[6] B.-S. Skagerstam and A. Stern, Physica Scripta 24, 493 with the energy-momentum tensor for the particle, an
(1981). approach that obscures the work being done by the elec-
[7] We thank David Griffiths for pointing out this defining tromagnetic field on the particle.
feature of Maxwell’s original theory. [13] For a review, see [18].
[8] H. van Dam and T. W. Ruijgrok, Physica A 104, 281 [14] This formula differs from the corresponding result in [8],
(1980). whose energy-momentum tensor yields the correct equa-
[9] R. Geroch and J. O. Weatherall, Communications in tions of motion for the particle only after an unjustified
Mathematical Physics 364, 607 (2018), arXiv:1707.04222 four-dimensional integration by parts.
[gr-qc]. [15] J. A. Barandes, (2019), arXiv:1911.08890.
[10] V. Bargmann, L. Michel, and V. L. Telegdi, Physical [16] N. Bohr, in Early Work (1905–1911), Niels Bohr Col-
Review Letters 2, 435 (1959). lected Works, Vol. 1, edited by L. Rosenfeld and J. R.
[11] We thank Sebastiano Covone for suggesting the rele- Nielsen (Elsevier, 1972) pp. 163–393.
vance of these results to the Bohr-van Leeuwen theo- [17] H. J. van Leeuwen, Journal de Physique et le Radium 2,
rem [16, 17]. The Bohr-van Leeuwen theorem assumes the 361 (1921).
original Lorentz force law without contributions from ele- [18] P. Di Francesco, P. Mathieu, and Sénéchal, Conformal
mentary dipole moments, and asserts that a non-rotating Field Theory, 1st ed. (Springer, 1997).