New J. Phys. 26 (2024) 093041 https://doi.org/10.

1088/1367-2630/ad7c72

PAPER

New angular momentum conservation laws for electromagnetic

OPEN ACCESS
waves interacting with dirac fields
RECEIVED
3 May 2024 Farhad Khosravi1,3, Li-Ping Yang2,3, Pronoy Das3,4 and Zubin Jacob3,4,∗
REVISED 1
30 August 2024 Irreversible Inc., 1950 rue Roy, Sherbrooke QC J1K 1B7, Canada
2
Center for Quantum Sciences and School of Physics, Northeast Normal University, Changchun 130024, People’s Republic of China
ACCEPTED FOR PUBLICATION 3
18 September 2024
Elmore Family School of Electrical and Computer Engineering, Birck Nanotechnology Center,Purdue University, West Lafayette, IN
47907, United States of America
PUBLISHED 4
Birck Nanotechnology Center,and Purdue Quantum Science and Engineering Institute,Purdue University, West Lafayette, IN 47907,
27 September 2024
United States of America
∗
Author to whom any correspondence should be addressed.
Original Content from
this work may be used E-mail: [email protected]
under the terms of the
Creative Commons Keywords: QED, structured light, quantum optics, angular momentum, electromagnetic, gauge field
Attribution 4.0 licence.
Any further distribution
of this work must
maintain attribution to Abstract
the author(s) and the title
of the work, journal
Global conservation laws of angular momentum (AM) are well-known in the theory of
citation and DOI. light–matter interaction. However, local conservation laws, i.e. the conservation law of AM at every
point in space, remain unexplored especially in the context of relativistic Dirac–Maxwell fields.
Here, we use the QED Lagrangian and Noether’s theorem to derive a new local conservation law of
AM for Dirac–Maxwell fields in the form of the continuity relation for linear momentum. We
separate this local conservation law into four coupled motion equations for spin and orbital AM
(OAM) densities. We introduce a helicity current tensor, OAM current tensor, and spin–orbit
torque in the motion equations to shed light on the local dynamics of spin-OAM interaction and
AM exchange between Maxwell and Dirac fields. We elucidate how our results translate to classical
electrodynamics using the example of plane wave interference as well as a dual-mode optical fiber.
Our results shine light on AM phenomena related to the relativistic interaction of electromagnetic
waves and Dirac fields.

1. Introduction

Global conservation laws for angular momentum (AM) are well known in the theory of light matter
interaction [1]. However, the spatial distribution of AM, i.e. AM density, is relatively unexplored especially in
the relativistic interaction between electromagnetic waves and Dirac fields. This interaction is the building
block for quantum electrodynamics (QED). Figure 1 highlights the difference between global and local
conservation of total AM. It also highlights unique physical quantities related to AM for both Dirac fields
and Maxwell fields. We emphasize that in the presence of sources, a comprehensive, gauge-independent,
locally applicable conservation equation becomes essential in understanding the local dynamics of the AM
(figure 1(b)).
On the other hand, the developments in classical electrodynamics have focused on both global and local
AM. The local approach towards field quantities such as helicity, chirality, and AM has proven useful in the
study of conservation laws as well as geometrical properties of EM fields [2–10]. However, the procedure
employed in these derivations is based on the duality symmetry of the source-free EM Lagrangian. Since the
duality symmetry is only maintained in source-free regions, this form of Lagrangian neglects any form of
interaction with fermionic fields and cannot be employed to present a local conservation law for the AM of
gauge fields in the QED Lagrangian describing classical Dirac-Maxwell fields. These properties are captured
in the manifestly covariant construction of the Dirac equation [11]. Connections between the fermionic field
of the Dirac equation and the bosonic fields of Maxwell’s equations is the focus of Dirac-Maxwell
correspondence where the relativistic parallels between electromagnetic and Dirac fields are studied [6, 12,
13]. These studies show that, although different in nature, electronic and electromagnetic fields can exhibit

© 2024 The Author(s). Published by IOP Publishing Ltd on behalf of the Institute of Physics and Deutsche Physikalische Gesellschaft
New J. Phys. 26 (2024) 093041 F Khosravi et al

Figure 1. Conservation laws of angular momentum in light-matter interacting systems (a) conventional conservation of global
total electromagnetic and electronic angular momentum. This conservation law applies to closed systems and does not include
angular momentum exchange due to near-field interactions. (b) Local conservation of angular momentum applicable to all
regions of interaction. The conventional conservation of sum of angular momenta is replaced by the conservation equation (8).
(c) Table of quantities defined in this paper, representing the spatial and temporal densities of the scalar, vector, or tensor
observables pertinent to angular momentum. Highlighted quantities are the new terms defined in this paper. The local
conservation of angular momentum equation (equation (8)) connects the well-known terms such as spin density, OAM density,
helicity and chirality to these newly defined quantities.

many analogous properties. This correspondence is evident in phenomena such as spin-momentum locking
which emerges in the Dirac equation [14] as well as the Maxwell’s equations [15, 16].
In this paper, we study AM of light using the Lorentz symmetry of the QED Lagrangian and local U(1)
gauge invariance [17]. With the application of Noether’s theorem [18], we find unique local conservation
laws pertinent to the AM of the gauge field interacting with fermionic fields. We show that, in the near-field,
the electronic AM can be transferred not only to the optical AM, but also to other field quantities that
represent AM current [19]. These extra terms include electromagnetic helicity density [20], fermionic
chirality density [11], electromagnetic helicity current tensor, as well as the electromagnetic and electronic
orbital AM (OAM) current tensors. We further study these conservation laws and
angular-momentum-carrying terms for the classical electromagnetic solutions of a dual-mode optical fiber.
This connection between electromagnetic waves and Dirac fields will be of interest for future experiments
related to OAM and SAM of light.

2. Noether’s theorem, QED Lagrangian, and Lorentz transformation

An important problem in QED relates to the AM of the gauge field. Particularly, the goal is to decompose the
total AM of gauge bosons into spin and orbital contributions [21]. In early work, in both Belinfante’s [22]
and Ji’s [23] decompositions, only the total AM of photons has been explored in detail. Jaffe and Manohar
split the photonic AM into spin and orbital parts [24]. However, their decomposition does not satisfy the
gauge-invariant requirement. By splitting the vector potential into transverse and longitudinal parts, Chen
et al [25] and Wakamatsu [26] propose gauge-invariant decompositions of electromagnetic AM. However,
these existing decompositions have not resolved the conflict between the gauge-invariant requirement and
canonical AM commutation relations. Recently, this problem has been solved by quantizing the U(1) gauge
field in the Lorenz gauge and merging the spin and OAM of virtual photons with the OAM of Dirac
fermions [27, 28]. Our present results are consistent with the results presented in these works, after the
quantization of the electromagnetic field.
The application of Noether’s theorem to the Dirac Lagrangian has been used in the Quantum
Chromodynamics (QCD) and QED communities to derive the expressions for the AM of the nuclei and the
gauge fields [23, 25]. In QCD and QED interactions, conservation of the total integrated AM suffices to

2
New J. Phys. 26 (2024) 093041 F Khosravi et al

describe the AM transfer in scattering processes. On the other hand, for a fast moving electron, its spin and
momentum degrees of freedom are highly correlated [29]. Explorations of a generalized definition for the
spin operator of an electron as well as for gauge bosons is still a problem of widespread interest [30, 31] in
QED and QCD [21, 23, 25]. While the focus has mainly been on global conservation laws for AM, our goal in
this paper is to focus on local spatially dependent AM density effects. In nanophotonic and condensed matter
systems, atoms can interact locally with an external EM field. Therefore, for such systems, a local
conservation equation for the AM density is necessary in the realization of nanoscale applications.
In this paper, we start with the real (also called symmetrized) Lagrangian density of a Dirac field coupled
to EM field [32],
   
µ 1 ← → 1
L = ψ̄ cγ ih̄ ∂ µ − eAµ − mc ψ −
2
Fµν Fµν , (1)
2 4µ0

where ψ̄ = ψ † γ 0 and γ µ are the gamma matrices, F µν the electromagnetic tensor, Aµ = (ϕ/c, −A) the
electromagnetic four-potential, ψ Dirac fields, h̄ Planck’s constant, and µ0 the vacuum permeability (see
appendix A). Note that summation over repeated indices is assumed throughout this paper.
In order to derive the angular momenta conservation law, we consider the the change in the Maxwell and
Dirac fields under Lorentz transformations:
i
µ ν i
δAµ (x) = − ωκσ M̂κσ
em ν
A (x) , δψ (x) = − ωµν M̂µν
D ψ (x) . (2)
2 2
Lorentz transformations are defined as the coordinate transformations such that the coordinates transform
as xµ → Λµ ν xµ , with
µ
Λµ ν = e− 2 ωκσ (Ŝ ) ν ,
i κσ
(3)

where (Ŝκσ )µ ν = i (η κµ η σ ν − η σµ η κ ν ) are the generators of rotation and boost in the four-dimensional real
space [33] and ω κσ are the rotation and boost parameters. These expressions can be used to find the AM
operators of the Maxwell, M̂em , and Dirac, M̂D , fields (see appendix A). By applying Noether’s theorem [18],
we obtain the conserved current for the AM for the Lagrangian in equation (1) by finding,

∂L ∂L ∂L
Mλ = δψ + δ ψ̄
+ δAµ + Lδxλ , (4)
∂ (∂λ ψ) ∂ ∂λ ψ̄ ∂ (∂λ Aµ )

under Lorentz transformations.

Specializing to the rotations (µ, ν = i, j = 1, 2, 3), we obtain the general continuity equation associated
with rotational symmetry of the QED Lagrangian [33](see appendix A),
 
ij,λ ij,λ
∂λ Mij,λ = ∂λ SD + LD + Sem ij,λ
+ Lijem
,λ
= 0, (5)

where Mij,λ is the total angular momentum current tensor. Note that the roman indices take the values
ij,λ
i, j = 1, 2, 3 while the Greek indices are λ = 0, 1, 2, 3. This tensor can be split into spin SD(em) (related to the
ij,λ
rotation of the internal degrees of freedom) and OAM LD(em) (related to the coordinate dependence of the
fields) parts of the Dirac and EM fields, respectively, with individual components defined as (figure 1),

h̄c

SDµν,λ = ψ̄ γ λ σ µν + σ µν γ λ ψ, (6a)
4


Lµν,λ
D = h̄cR ψ̄γ λ (xµ ∂ ν − xν ∂ µ ) ψ + η λµ xν − η λν xν LD , (6b)
1 λµ ν

Sem
µν,λ
=− F A − Fλν Aµ , (6c)
µ0
1  λ 

Lµν,λ
em = − F κ (xµ ∂ ν − xν ∂ µ ) Aκ + η λµ xν − η λν xµ Lem , (6d)
µ0

where LD and Lem are the Dirac and electromagnetic contributions to the total Lagrangian in equation (1),
respectively. Here σ µν = 2i [γ µ , γ ν ], ∂ µ = ( ∂ct , −∇), Fµν = ∂ µ Aν − ∂ ν Aµ is the anyti-symmetric
electromagnetic tensor, and R{· · · } takes the real part of its argument. It is important to note that splitting
the tensor Mij,λ into the four components for the electromagnetic and Dirac spin and OAM, as
demonstrated in equations (5) and (6), is arbitrary since Noether’s theorem does not prescribe how spin and
OAM should be defined individually. Instead it defines the conservation law for the total AM of the
combined fields. We use this fact to derive gauge-independent expressions for the spin and OAM of the Dirac

3
New J. Phys. 26 (2024) 093041 F Khosravi et al

and Maxwell fields in the next section by rearranging the longitudinal (labelled by the symbol ∥) and the
transverse (labelled by the symbol ⊥) components of the electric field and the vector potential between these
four equations (see appendix A). For non-interacting fields, however, we obtain separate continuity
equations for the Dirac and EM fields.
The time-components (λ = 0) of these four AM tensors in equation (6) give the common spin and OAM
densites for the Dirac and EM fields, which are generally gauge dependent.
 † The electronic part of the spin
ij,0 † ij,0
and OAM respectively are SD = ε h̄ψ Σk ψ/2 and LD = −iε h̄R ψ (r × ∇)k ψ , where Σ is the spin
ijk ijk
ij,0
 part of the spin and OAM, on the other hand, are Sem = ε (ϵ0 E × A)k
ijk
operator in Dirac equation. The EM
ij,0
and Lem = −ε ϵ0 E (r × ∇)k Al , respectively, where ε0 is vacuum permittivity. An important observation
ijk l

is that OAM densities of the Dirac fields as well as the spin and OAM of the EM field are gauge dependent
and thus do not represent observable physical quantities in a local frame. In fact, one can see that most of the
ij,λ ij,λ
terms in the AM tensors SD(E) and LD(E) are gauge dependent. Due to this problem, we rearrange the
expressions in these four tensors to write the conservation law in equation (5) in terms of gauge-independent
and physically observable quantities.
We note that throughout this paper, εijk represents the Levi–Civita symbol with the convention that
ijk
ε = εijk , and is normalized such that ε123 = 1 together with all of its even permutations; this means that any
of its indices can be lowered or raised without any sign change. Also, our convention for outer product is
(a × b)k = εijk ai bj for arbitrary vectors a and b.

3. Local conservation law of AM

The essence of Noether’s theorem´in field theory lies in the local conservation law with a general form
∂µ jµ = 0. The global charge Q ≡ d3 xj0 (x) conserves only if all fields vanish at spatial infinity, i.e. the
vanishing boundary condition. As we mentioned above, the local conservation law of AM has not received
the attention it deserves both in experiments and in theory. We note that the density of the conserved charge
in the original continuity equation (5) of AM is a rank-2 tensor and the corresponding current is a rank-3
tensor. This local conservation law is difficult to apply to practical systems and challenging to verify in
experiments. In this section, we re-express this relation as a continuity equation of a vector density similar to
the local conservation law of linear momentum. The corresponding current reduces to a rank-2 tensor akin
to the Maxwell stress tensor. We then go one step further by splitting our local conservation law into four
coupled motion equations, which describe the spin-OAM interaction and AM exchanging between
Maxwell–Dirac fields.
By splitting the electromagnetic vector potential into transverse and longitudinal parts, A = A⊥ + A∥ ,
defined as ∇ · A⊥ = 0 and ∇ × A∥ = 0 [1], Chen et al defined the gauge-independent AM densities for both
the Dirac field and EM field [25]. However, a gauge-independent form for the continuity equation has not
been addressed. Here, following the same approach used in [25], and using equation (6) and Dirac and
Maxwell’s equations, we obtain the following gauge-independent spin and OAM continuity equations (see
appendix A):
 
ij,λ ∂ †

h̄c † 5


∂λ SD = ε h̄ ijk
ψ Σψ + ∇ ψ γ ψ , (7a)
∂t 2 k
  o
∂ n †  o n 
R ψ r × p∥ ψ + c∇ · R ψ̄γ r × p∥ ψ
ij,λ
∂λ LD = εijk , (7b)
∂t k
 
∂ ⊥ ⊥

1 ⊥

1 ⊥


∂λ Sem = ε ϵ
ij,λ ijk
E ×A − ∇· A B + ∇ A ·B , (7c)
∂t µ0 µ0 k
( " !#
∂  ⊥ ⊥
 1 ⊥ ⊥ ∂E∥
∂λ Lem = ε
ij,λ ijk
ϵ E · (r × ∇) A − ∇ · B × (r × ∇) A + ϵ0 A r×
∂t µ0 ∂t
" #)
1 ϵ0 ∂E∥ ⊥
+ (r × ∇) − B · B + E ⊥ · E ⊥ + ϵ0 ·A . (7d)
2µ0 2 ∂t
k

The longitudinal and transverse components of the electric field can be defined similarly to those of the
∥ ⊥
vector potential. In the Lorenz gauge, they take the form E∥ = −∇ϕ − ∂∂At and E⊥ = − ∂∂At . Adding these
equations together, we obtain a standard (vector) continuity equation for the AM density of the combined
system,

∂Mj
+ ∇i T i j = 0, (8)
∂t

4
New J. Phys. 26 (2024) 093041 F Khosravi et al

where Mj are the components of the AM density vector, written as

h̄ †
n   o
 
M= ψ Σψ + R ψ † r × p∥ ψ + ϵ0 E⊥ × A⊥ − ϵ0 Ei,⊥ (r × ∇) A⊥
i , (9)
2

with gauge-independent
 transverse
 vector potential A⊥ and gauge-independent electronic momentum
∥ ∥
operator p = −ih̄∇ − eA . The first two terms in equation (9) are the spin and OAM densities of the
Dirac field, while the last two terms are the gauge-independent spin and OAM densities of the EM field [21,
25]. This shows that only the transverse part of the vector potential contributes to the physically observable
spin and OAM of EM field. Note that here A⊥ i = −A
⊥,i
= (−A) for i = 1, 2, 3.
The AM current tensor, Ti j , is a second rank tensor, similar to the Maxwell stress tensor (EM momentum
current) [14, 34]. The tensor Ti j is composed of three parts :∇i Ti j = ∇i (χi j + Ji j + Ni j ) with
 
chirality helicity
z }| { z }| {
 h̄
1

 
χi j =  ψ † γ 5 ψ + A⊥ ·B  δ i j ≡ χ δ i j , (10a)
 2 µ 0 

OAM current tensor (Dirac)

z  }| { helicity current
z }| {
tensor (Maxwell)
   1

J i j = R ψ̄γ i r × p∥ ψ − A⊥,i Bj
j µ0
!
h
1 ikl i ∂E ∥
+ ε Bk (r × ∇)j A⊥
l − ϵ0 A⊥,i r × , (10b)
µ0 ∂t
j
| {z }
OAM current tensor (Maxwell)

ϵ0 ⊥ ⊥ 1 ∂E∥ ⊥
∇i N i j = (r × ∇)j Nem , Nem = E ·E − B · B +ϵ0 ·A . (10c)
2 2µ0 ∂t
| {z }
Maxwell Lagrangian in free space

where γ 5 is the chirality operator in Dirac equation (see appendix A).

The term χ in equation (10a) describes the chirality/helicity of the fields. The first term, h̄(ψ † γ 5 ψ)/2, is
the density of expectation value (in the classical sense of quantum mechanics) of the chirality operator of the
Dirac field. The operator γ 5 is widely used to find the projections of fermionic fields into left-handed and
right-handed chiral states [17]. Therefore, the expecation value ψ † γ 5 ψ represents the chirality of the Dirac
field; a negative (postivie) value means a left-handed (right-handed) chiral state. The second term in
equation (10a), on the other hand, is widely known as the helicity of EM field [35–37]. Interestingly,
equation (10a) shows that the chirality of the fermionic field is identified with the helicity of the
electromagnetic field and the gradient of these quantities enter the continuity equation.
Note that the chirality of EM field defined in [38] does not appear in the continuity equation of AM.
This shows that the helicity is more fundamental for interacting systems. It is also important to note that
helicity can be defined for the Dirac field. Being defined as the projection of spin on the momentum of the
field, it is fundamentally different from chirality as it is momentum dependent, and unlike chirality, it is not
Lorentz covariant. Interestingly, however, for the massless Dirac field, the definition of chirality and helicity
become identical [39]. This points to a similar property in the massless EM fields of the Maxwell’s equations.
In the next three sections we define the new terms encountered in the conservation equation.

3.1. Helicity current density tensor

The second terms in equation (10b) is reminiscent of the EM helicity in equation (10a) with the difference
that this quantity is a tensor, while helicity is a scalar. In fact, helicity density in equation (10a) is equal to the
trace of µ10 A⊥,i Bj . We thus name this tensor quantity helicity current density tensor since it transforms as a
helicity. Furthermore, the divergence of the helicity current tensor enters the conservation of AM equation,
meaning that it serves as a current between different forms of Dirac or Maxwell AM.
One important fact regarding this term is that it has no analog in the Dirac field representation, as one
might expect. The equivalent expression for helicity current tensor, in the Dirac field, is normally considered
to be spin current represented by a tensor involving γ matrices and spin operators. However, our derivation
shows that this term vanishes due to the spinor nature of the Dirac fields (see appendix A). This, of course,
does not put the numerous works on spin currents in condensed matter physics into question [40–43], since
spin currents are generated by the collective motion of electrons through spatially separating electrons with

5
New J. Phys. 26 (2024) 093041 F Khosravi et al

opposite spins in a quantum cavity [44] or using ferromagnets with strong electron–electron interactions
[45]. The relativistic treatment of the Dirac equation in this paper, however, only incorporates the wave
function of a single electron in vacuum and thus cannot account for the observed electronic spin currents.
This shows that a single electron in vacuum does not exhibit spin current and that many-body Dirac
equation should be considered for a relativistic account of electronic spin currents [46, 47].

3.2. OAM current density tensor

Equation (10b) specifies the AM current tensors for the fields. This current tensor is specified by two
directions: direction of AM and direction of the current. For instance, in the first term of equation (10b), the
direction of propagation of the current is specified by the gamma matrices γ i , while the direction of the AM
is specified by the operator (r × p∥ ). Thus we call this term OAM current density tensor of Dirac field.
Similarly, the third term in equation (10b) is the OAM current tensor density of the EM field. The last term is
the AM current tensor due to longitudinal component of electric field and we incorporate that into the OAM
current tensor of the EM field. Equation (10c) shows that the electromagnetic Lagrangian also contributes to
the continuity equation (8).
Barnett introduced the AM carrying terms in the source-free electromagnetism [19] as quantities that
behave as the currents in the conservation equations written for AM of light. The terms helicity and OAM
current tensors introduced here for the electromagnetic field have a similar tensorial nature and extend to the
QED interactions. These quantities are gauge-invariant and connect to similar terms coming from the Dirac
fields.
Note that the expression for the AM due to the electromagnetic field in equation (9) originally has
contribution from the longitudinal electric field, E∥ . This is the contribution due to the free charges
(∇ · E∥ = ρ/ϵ0 ) which decays as a function of 1/r2 outside of the region of charges [34]. We have shown that
the contribution from this component, namely the term E∥ × A⊥ − E∥,i (r × ∇)A⊥ i , can be written as the
divergence of some quantities involving E∥ (see appendix A). Therefore, when integrated over the entire
space, these terms vanish and thus the total global AM of the interacting field does not involve longitudinal
electric field components. Therefore, we have shown that the AM density of the EM field only depends on
the transverse component of the electric field E⊥ since it is only these terms that give nonzero contribution
to the total AM.
It should be emphasized that, when working with global quantities (integrated over entire space), the
contribution from the longitudinal electric field as well as the contribution from Ti j in equation (10) become
surface terms and vanish, assuming that the field quantities are zero on the surface of the integration volume.
In such cases, the global AM can be written in equivalent forms which are different only by the divergence of
a function of the fields. In the local case, however, all of the terms in equation (10) are observable quantities
and each have a different interpretation.

3.3. Spin–orbit torque

We now study the spin-OAM exchange and the AM transfer between the Dirac and EM fields. The continuity
equation in equations (5) and (8) only give the dynamics of the total AM density. To obtain insight into the
detailed interaction between spin and OAM of the Dirac and EM fields, we separately write the
ij,λ ij,λ
four-divergence of the gauge-independent spin and OAM tensors of the Dirac (SD and LD ) as well as the
ij,λ ij,λ
EM fields (Sem and Lem ). By incorporating the Dirac and Maxwell’s equations into the gauge- invariant AM
densities in equations (9) and (10), we obtain their detailed conservation equations (see appendix B),
 
∂λ SD = εijk τ D + jc × A⊥ k ,
ij,λ
(11a)
h   i
∂λ LD = εijk −τ D +ρ r × E∥ −jlc (r × ∇) A⊥
ij,λ
l , (11b)
k
 
∂λ Sem
ij,λ
= εijk τ em − jc × A⊥ k , (11c)
h   i
∂λ Lijem
,λ
= εijk −τ em −ρ r × E∥ +jlc (r × ∇) A⊥l , (11d)
k

where we have defined

n   o
τ D = −cR ψ̄ γ × p∥ ψ (12)

and
!
1 ∂E∥
τ em = (B · ∇) A⊥ − ϵ0 × A⊥ (13)
µ0 ∂t

6
New J. Phys. 26 (2024) 093041 F Khosravi et al

as the Dirac and Maxwell spin–orbit torque, respectively, since it gives the amount of torque exerted on the
spin from the OAM of the fields and vice versa. This nomenclature is further motivated by the resemblance of
the Dirac spin–orbit torque (equation (12)) of the Rashba spin–orbit coupling Hamiltonian [49]. The direct
connection between these terms, however, is out of scope of this article and is the focus of a future work. Note
that γ = γ 1 x̂ + γ 2 ŷ + γ 3 ẑ and jc = ecψ̄γψ is the electric charge current density in equations (11) and (12).
Equation (11) clearly shows that the spin–orbit torques contribute to the spin-OAM exchange in both
Dirac and EM fields. Moreover, it is evident from equations (11a) – (11c) and (11d) that the charge-field
coupling terms, jc × A⊥ , jlc (r × ∇)A⊥ ∥
l , and ρ(r × E ), are responsible for the AM transfer between the Dirac
and EM fields [50]. In fact, the first terms gives rise to an optical torque exerted on dipoles due to a circularly
polarized optical field [51]. Our results are significantly different from scalar continuity equations of EM
helicity in [4] and [5], which derive a scalar conservation law for the dual-symmetric expressions of spin and
helicity of EM field. The helicity continuity equation obtained from the free-space Maxwell equations cannot
characterize the spin-OAM exchange and specifically the AM transfer between Dirac and EM fields.

4. Source-free problems

We now show the importance of these continuity equations by demonstrating the spin-OAM exchange via
the spin–orbit torque for propagating EM fields. We evaluate the terms in ∂λ Sijem,λ
and the EM spin–orbit
torque τ em (equation (13)) for two simple EM problems in source-free regions. Similar spin–orbit signature
can also be observed in a cylindrical geometry for the Dirac fields [14]. However, a thorough study of each
individual term in equation (10) for Dirac and EM fields is outside the scope of this paper.
The general form of ∂λ Sijem
,λ
, in a source-free regions, can be found from equations (7c) – (11c) and by
taking jc = 0:

∂ ⊥
1
1

ϵ E × A⊥ − ∇i A⊥,i B + ∇ A⊥ · B = τ em . (14)
∂t µ0 µ0

Here, the extra term on the right hand side results from the coupling to OAM. The spin–orbit torque in the
source-free case reduces to τ em = (B · ∇)A⊥ /µ0 . We emphasize that different from free-space case [4, 5], the
near-field spin-OAM exchange can still exist in the presence of sources.

4.1. Dual-mode optical fiber

We first consider a dual-mode optical fiber, which is placed along the z axis with radius a. The two modes
have the same propagation constant, β, and different frequencies of ω 1 and ω 2 . Standard solutions of
source-free Maxwell equations in cylindrical coordinates have been used to evaluate the local angular
momenta distributions [48]. Figure 2 shows the four terms in equation (14). The three rows show the
components of these quantities along the radius ρ̂, azimuthal ϕ̂, and z axis of the fiber. Note that for all three
rows, the sum of the first three terms is equal to τ em ; thus satisfying equation (14). These figures show that,
for a dual-mode optical fiber, the helicity current current tensor, helicity density, spin density, as well as the
spin–orbit torque are all non-zero and can play a role in a local interaction between optical modes and
atomic sources. Note that all three components become zero as the fiber radius a → ∞, which confirms that
the spin–orbit torque τ em becomes zero for plane wave solutions.
Figure 3 shows the dynamics of the time-derivative of spin, divergence of helicity current density, and
gradient of helicity (the terms in equation (14)) as a function of z (figure 3(a)) and time (figure 3(b)). The
curves show the value of these terms at a point 0.05 a from the fiber as shown in the insets. The black curve
with the circle markers show the EM spin–orbit torque, τ em , which is essentially the torque that is
transferred to the OAM of the fields. Note that the sum of other curves equals the black curve τ em ; satisfying
the continuity equation (14).

4.2. Double plane wave interference

The second example is the interference of two circularly polarized plane waves at two different frequencies.
For the plane wave solutions, the current term ∇i (A⊥
i B) and the spin–orbit torque τ em vanish (see
appendix C). Therefore, we arrive at a simpler conservation law for the second problem,
 
∂
1 ⊥
ϵ0 E⊥ × A⊥ + c∇ A · B = 0, (15)
∂t µ0 c

which shows the spin density propagating in space with light speed c. For circularly polarized plane waves
propagating along z direction, the electric field is written as E = R{E 1 e−i(ω1 t−k1 z) + E 2 e−i(ω2 t−k2 z) }, where

7
New J. Phys. 26 (2024) 093041 F Khosravi et al

Figure 2. Local conservation law for the spin in a dual-mode optical fiber. The individual terms in equation (14) are plotted on
the first three columns from left: the first column is the time-derivative of spin ( ∂∂t (E⊥ × A⊥ )), the second column is the
divergence of helicity current tensor (−∇i (A⊥ i B)/µ0 ) , and the third column is the gradient of the helicity density
(∇(A⊥ · B)/µ0 ) . The fourth column is the EM spin–orbit torque given by equation (13) (in this case τem = (B · ∇)A⊥ /µ0 ).
The three rows show the local value of each vector along the three axes of optical fiber problem: ρ̂ radial direction, ϕ̂ azimuthal
direction, and ẑ axis of the fiber. Note that adding the first three column on each row together gives the last column τ em ; thus
confirming equation (14). The results are for an optical fiber of radius 50 µm with the two modes at the wavelengths 4.3 µm and
4.29 µm. Solutions of source-free Maxwell equations in cylindrical coordinates have been used to evaluate these terms [48].

√
E i = Ei (x̂ + iŷ)/ 2 are the complex electric field amplitudes of the two modes with frequencies ωi /c = ki .
For these fields, we find (see appendix C),

1
∂
ω 2 − ω22 n o
∇ A⊥ · B = −ϵ0 E × A ⊥ = ϵ0 1 I E1 E2∗ e−i[(ω1 −ω2 )t−(k1 −k2 )z] , (16)
µ0 ∂t 2ω1 ω2

hence satisfying the conservation equation (15). This clearly shows that the change in time of the spin is
compensated by the gradient of the helicity of the EM field.

5. Experimental perspective

The interaction between the electronic spin and optical AM has been widely used in devices such as optical
circulators, insulators, electro-optic modulators, and Faraday rotators. In these devices, conservation of AM
combined with the field properties of the electromagnetic (EM) radiation gives rise to non-reciprocal effects
that can be manipulated for a wide variety of applications. These devices incorporate the interaction between
EM radiation and the magnetic materials in the far-field regions where a conservation law based on the total
electronic and EM angular momenta suffices to explain the underlying physical phenomena [52].
Due to the possibility of optical control of the quantum spin states, the dynamics of AM between the
electronic and optical fields has recently gained attention for the systems interacting in the near-field region.
Coupling to the local spin of EM field has been observed for cold atoms in the vicinity of an optical fiber
[53], magnons interacting with spherical whispering gallery modes [54], quantum dots in photonic crystals
[55], as well as quantum sources coupling to other waveguide systems [9, 56, 57]. Non-classical spin texture
and non-local photonic spin noise has been shown in quantum structured light [58, 59], which can be
measured in experiment via nano-scale quantum sensors for photonic spin density [60]. In such systems,
since the interaction between the source and the EM field occurs in the near-field rather than the far-field
regions, a local approach to the governing dynamical equations of AM becomes important. These studies
reveal the new insight that the local interaction with classical EM field has to offer.
With recent advances in the cold atom and quantum dot communities, local interactions between atomic
sources and optical fields are gaining more attention due to the emergent new phenomena. These
experiments emphasize the need for a local conservation law describing the dynamics of AM in light-matter
interactions.

8
New J. Phys. 26 (2024) 093041 F Khosravi et al

Figure 3. Dynamics of the terms in equation (14) versus (a) z and (b) time. The plots only show the ẑ component of each term.
These plots show that the spin–orbit torque (τ em ) is equal to the sum of the other terms. The inset in panel (a) shows the location
where the terms are evaluated for both of the figures.

We applied the Noether’s theorem to the Dirac Lagrangian interacting with the EM field to find the local
conservation laws of AM for the most general electrodynamics problem. The results developed here can be
applied to near-field as well as far-field to study the transformation of AM between different fields in these
regions. Our results show that, in consideration of local conservation laws, other quantities including helicity
and OAM current tensors, EM helicity, and electronic chirality should also be considered in addition to the
spin and OAM of the EM and Dirac fields.
Equation (8) holds everywhere in space and time. This shows that the total AM density M is not locally
conserved. In other words, ∂∂Mt ̸= 0 and the conservation law can only be written after including all the other
terms in equations (8) and (10). Integrating equation (8) over some volume V, on the surface´ of which both
Dirac and EM fields become zero, ψ → 0 , E → 0, gives the usual global conservation law, ∂∂t V Md3 x = 0.
This states that the integrated values of the spin and OAM densities of electronic and EM field over the entire
space is a conserved quantity (figure 1(a)).
We can also get a simpler continuity equation than equation (8) if we limit ourselves to regions outside
the Dirac fields. To do so, we take the integral of equation (8) in the volume V ′ on which only the Dirac fields
become zero ψ = 0. Doing so we get the semi-local conservation law (see appendix D),
 ˆ 
∂ M̃D ∂ M̃em
=− + J̃ + h̃ + n̂ × (rNem ) da (17)
∂t ∂t S′

which expresses the time-evolution of the AM of the Dirac field M̃D in terms of EM dependent quantities.
M̃em is the EM AM in the region of the source, while J̃ and h̃ are the EM AM current tensor (projected onto
the normal of the surface) and helicity (multiplied by the normal of the surface), integrated on the surface of
the volume surrounding the current charges. Also, Nem is given by equation (10c). Equation (17) can be used
to find the conservation of AM in the near-field of current sources and how is it transferred to the EM fields.
It is also important to note that the conservation of AM derived here is based on the spatial components
of Lorentz transformations (i.e. rotations). Time-space components of the Lorentz transformation can also

9
New J. Phys. 26 (2024) 093041 F Khosravi et al

give conservation equations related to the boost operators in the relativistic equations of motion. These
equations provide a new conservation equation for the Dirac–Maxwell fields which can be of interest to
applications investigating relativistic behaviour of the particles and their pertinent conserved quantities
when interacting. A brief discussion about these conservation equations is given in appendix E.
The method presented in this paper can be further extended to find the dynamics of magnetization in
different materials. The simplest system can be regarded as the interaction between an externally applied EM
field and the electrons in a non-magnetic metal. A weak probe signal can then be used to study spin
dynamics of the metal. Such a system can be closely modeled as a non-interacting electron gas whose
dynamics can be described by the Dirac equation. Although the solutions of the Dirac equation interacting
with an externally applied plane wave can be rigorously found [61, 62], these solutions are extremely
complicated and only apply to the particular case of free Dirac field described by plane wave solutions. Our
method, circumvents the problem of solving the Dirac Hamiltonian to find the spin dynamics of electronic
fields, by using Noether’s theorem and finding the conservation equation governing the AM dynamics. By
knowing the properties of the externally applied EM field, a simpler equation such as equation (17) can be
used to find the dynamics of AM without the need of electronic wavefunctions. Since in our derivations we
make no simplifying assumption on the EM field, these equations can be easily used for interactions that take
place in the near-field region and where EM fields cannot be regarded as plane waves.
The Dirac equation can be further extended to model ferromagnetic materials. Dirac–Kohn–Sham
(DKS) equation is an extension to the Dirac equation which accounts for the Kohn–Sham potential as well as
the spin-polarized part of the exchange correlation potential inside a magnetic material [63, 64]. Corrections
from DKS equation can be added to the usual Pauli Hamiltonian to account for the terms in the
Landau–Lifshitz–Gilbert equation and to add corrections accounting for higher order terms dependent on
external and internal parameters [65]. The method presented in this paper can also be applied to the DKS
Hamiltonian to derive the conservation equation for the AM of electrons in magnetic materials.

Data availability statement

All data that support the findings of this study are included within the article (and any supplementary files).

Acknowledgments

This work is partially supported by the funding from Army Research Office (W911NF-21-1-0287). L P Y is
funded by National Key R&D Program of China (No. 2021YFE0193500). L P Y was a post-doctoral scholar
and Farhad Khosravi was a visiting scholar at Purdue University when this collaborative research work was
performed.

Appendix A. Conserved angular momentum tensor

Symmetrized Dirac Lagrangian, with the minimal coupling term [32], is given by equation (1), where
→ ←
← − − → ←− −
→
∂ = ∂ + ∂ with ∂ and ∂ acting only on ψ̄ and ψ, respectively,

 
0 Ex /c Ey /c Ez /c
−Ex /c 0 −Bz By 
Fµν = ∂µ Aν − ∂ν Aµ = 
−Ey /c Bz
 (A.1)
0 −Bx 
−Ez /c −By Bx 0

is the EM tensor, and γ µ are the Dirac gamma matrices with the property {γ µ , γ ν } = 2η µν , where

 
1 0 0 0
0 −1 0 0
η µν = 
0 0 −1 0  (A.2)
0 0 0 −1

10
New J. Phys. 26 (2024) 093041 F Khosravi et al

is the Minkowski metric tensor with the signature (+ − −−). We can get the conserved currents related to
the rotational symmetry of the Lagrangian, using the Noether’s theorem, as [18, 32]:
  !
i ∂L i ∂L
Mµν,λ = − µν
M̂ ψ + − ψ̄ M̂D µν


2 ∂ (∂λ ψ) D 2 ∂ ∂λ ψ̄
 
i ∂L
κ σ 1 λµ ν

+ − κ
M̂µν
em σ
A + η Lx − η λν Lxµ (A.3)
2 ∂ (∂λ A ) 2

where

M̂µν
D = L̂
µν
+ Σ̂µν (A.4)

is the AM operator for the Dirac fields with

1 i
Σ̂µν = σ µν , σ µν = [γ µ , γ ν ] (A.5)
2 2
L̂µν = xµ ∂ ν − xν ∂ µ , (A.6)

and

κ  κ
M̂µν
em σ
= L̂ µν κ
δ σ + Ŝµν (A.7)
σ

is the AM operator for the EM fields with L̂µν given by equation (A.6), δσκ being the Kronecker delta
function, and
 κ
Ŝµν = i (η µκ η ν σ − η µ σ η νκ ) . (A.8)
σ

Plugging these equations into equation (A.3), we get for the AM currents

Mµν,λ = Mµν,λ
D + Mµν,λ
em (A.9)

where

Mµν,λ
D = SDµν,λ + Lµν,λ
D (A.10)

and

Mµν,λ
em = Sem + Lem
µν,λ µν,λ
(A.11)

are the contributions to the spin and OAM from the Dirac and Maxwell’s fields, respectively, given in
equation (6). In these equations, LD and Lem are contributions due to the Dirac field and the EM field to the
total Lagrangian, given by,
 
1 µ←→
LD = ψ̄ ih̄c γ ∂ µ − mc ψ − ceψ̄γ µ ψ Aµ
2
(A.12)
2

and
 
1 1 E·E
Lem = − Fµν Fµν = − B · B . (A.13)
4µ0 2µ0 c2

As a consequence of Noether theorem, the AM tensor Mµν,λ is conserved. In other words,

∂λ Mµν,λ = 0. (A.14)

For the tensors given in equation (6), and using Maxwell and Dirac equations, one can show that
equation (A.14) holds for the total AM tensor.

11
New J. Phys. 26 (2024) 093041 F Khosravi et al

For the case that µν = ij, where i, j = 1, 2, 3, we find the AM currents due to rotations. We find for the
spin and OAM currents of the Dirac field
 
ij,λ ∂ †

h̄c † 5


∂λ SD = ε h̄ijk
ψ Σψ + ∇ ψ γ ψ , (A.15a)
∂t 2
 k

∂  
∂λ LD = εijk −h̄ R iψ † (r × ∇) ψ − h̄c∇ · R iψ̄γ (r × ∇) ψ
ij,λ
(A.15b)
∂t k

where
 
1 jk i  j k  1 σi 0
Σi = εijk σ = εijk γ , γ = , (A.16)
2 4 2 0 σi

with σ i being the Pauli matrices, R{· · · } takes the real part of its argument, γ 5 = iγ 0 γ 1 γ 2 γ 3 is the chirality
operator in the Dirac equation [32], and

γ = γ 1 x̂ + γ 2 ŷ + γ 3 ẑ. (A.17)

Note that we have used the fact that, using the Dirac equation,

(ih̄γ µ ∂µ − eγ µ Aµ − mc) ψ = 0, (A.18)

we get LD = 0 for the fields that follow Dirac equation of motion. This is straightforward to show by
multiplying equation (A.18) from left by cψ̄. For the spin and OAM currents of the EM field we find
 
∂ 1 1
∂λ Sem = ε ϵ (E × A) − ∇ · (AB) + ∇ (A · B)
ij,λ ijk
(A.19a)
∂t µ0 µ0
  k
 
∂ 1
∂λ Lijem
,λ
= εijk ϵ [E · (r × ∇) A] − ∇ · B × (r × ∇) A − ϵE (r × ∇ϕ) + (r × ∇) Lem . (A.19b)
∂t µ0 k

The problem with equations (A.15b) and (A.19) is that they are gauge dependent, which means that
i
under the transformations ψ → ψ e e ζ and Aµ → Aµ + ∂µ ζ they do not remain invariant. Therefore, the
individual terms do not represent any physically meaningful quantity. The fundamental equation to hold is
equation (A.14). Therefore, as long as this relation is satisfied, we can cast equations (A.15) and (A.19) into
gauge-independent forms. This means that by rearranging some terms between equations (A.15) and (A.19)
we can produce gauge-independent terms while still satisfying the continuity equation in equation (A.14). To
do so, we break A into two longitudinal and transverse parts as A = A∥ + A⊥ , where ∇ · A⊥ = 0 and
∇ × A∥ = 0 by definition. From equations (A.15) and (A.19), the time component of AM tensor Mij,0 is
given by:

h̄ †

M= ψ Σψ − h̄R ψ † (r × ∇) ψ + ϵ (E × A) + ϵ [E · (r × ∇) A] . (A.20)
2

Splitting A into transverse and longitudinal parts, we can write the terms containing A∥ as:
h  i  
E × A∥ + E · (r × ∇) A∥ = ∇ · E r × A∥ − (∇ · E) r × A∥ (A.21)

where we have used the following identity for arbitrary vectors X and X:

∇ · [X (r × X)] = (∇ · X) (r × X) + (X × X) + X · (r × ∇) X + (X · r) (∇ × X) − X [r · (∇ × X)] , (A.22)

and the fact that ∇ × A∥ = 0. Using the fact that ∇ · E = ρ/ϵ = ϵe ψ † ψ, we can write the last term in
equation (A.21) as:
  e  
(∇ · E) r × A∥ = ψ † r × A∥ ψ (A.23)
ϵ
Plugging these into equation (A.20) we get

h̄ †
 h  i  
M= ψ Σψ − h̄R ψ † (r × ∇) ψ + ϵ∇ · E r × A∥ − eψ † r × A∥ ψ.
2
h̄ †
n h i o h  i
= ψ Σψ + R ψ † r × p∥ ψ + ϵ∇ · E r × A∥ , (A.24)
2

12
New J. Phys. 26 (2024) 093041 F Khosravi et al

where p∥ = −ih̄∇ − eA∥ is the gauge-independent covariant momentum operator of the electronic field.
Note that the first two terms are now gauge invariant. The last term is written as the divergence of a vector
and can thus be absorbed into the rest of the continuity equations written as the divergence of the space
components of the AM tensor Mij,λ .
Repeating a similar procedure as above for the other components of the continuity equation, namely
∇k Mij,k , we find for the new spin and OAM tensors of Dirac and EM field after some algebra:
 
∂
h̄c

ψ † Σψ + ∇ ψ † γ 5 ψ ,
ij,λ
∂λ SD = εijk h̄ (A.25a)
∂t 2
 k
 o
∂ n †  o n 
R ψ r × p∥ ψ + c∇ · R ψ̄γ r × p∥ ψ
ij,λ
∂λ LD = εijk , (A.25b)
∂t
  k
∂ ⊥

1 ⊥

1 ⊥


∂λ Sem = ε ϵ
ij,λ ijk
E×A − ∇· A B + ∇ A ·B , (A.25c)
∂t µ0 µ0
  k
   
∂   1
∂λ Lijem
,λ
= εijk ϵ E · (r × ∇) A⊥ − ∇ · B × (r × ∇) A⊥ + ϵE r × E∥ + (r × ∇) Lem . (A.25d)
∂t µ0 k

We can further separate the contribution of electric field to the AM into its transverse and longitudinal
components as E = E⊥ + E∥ . Making use of the vector identity in equation (A.22) again, this contribution
can be written as:
h  i h  i
E∥ × A⊥ −E∥,i (r × ∇) A⊥
i = −∇ × r E · A
∥ ⊥
− ∇ · A⊥ r × E∥ . (A.26)

When integrated over the entire space, both of these terms on r.h.s of this expression become zero due to the
Stokes theorem and thus the longitudinal electric field does not contribute to the global AM of the EM field.
For this reason, we move this term into the AM current terms so that the global AM represents the integrated
AM density. Making this change, we get for the new components of the EM AM currents:
 
∂ ⊥ ⊥

1 ⊥

1 ⊥


∂λ Sem = ε ϵ
ij,λ ijk
E ×A − ∇· A B + ∇ A ·B , (A.27a)
∂t µ0 µ0 k

( " !#
∂  ⊥ ⊥
 1 ⊥ ⊥ ∂E∥
∂λ Lijem
,λ
=ε ijk
ϵ E · (r × ∇) A − ∇ · B × (r × ∇) A + ϵ0 A r×
∂t µ0 ∂t
" #)
1 ϵ0 ∂E∥ ⊥
+ (r × ∇) − B · B + E ⊥ · E ⊥ + ϵ0 ·A . (A.27b)
2µ0 2 ∂t
k

Adding these four equations together we get equation (5),

∂Mj
+ ∇j J i j + ∇i χ + ∇i N i j = 0, (A.28)
∂t
where the terms are given in equations (9) and (10). Equation (A.28) can be written as

∂M ←→
+∇· T = 0 (A.29)
∂t
where
←
→ ←→ ← → ← →
T =χ I + J + N (A.30)
←→
with I = x̂x̂ + ŷŷ + ẑẑ = δ i j x̂i x̂j , and
 
0 z −y
←
→ ϵ0 ⊥ ⊥ 1 ∂E∥ ⊥
N = εi jk x̂i x̂j xk Nem = −z 0 x  Nem , Nem = E ·E − B · B + ϵ0 ·A . (A.31)
2 2µ0 ∂t
y −x 0

Note that the first two terms of Nem describe the Lagrangian due to the transverse electric field, which can be
interpreted as the Lagrangian of the free EM field [66], while the last term shows the interaction between the
transverse vector potential and the currents due to the longitudinal electric fields. Note that the term
ϵ0 ∥ ∥
2 E · E in the Lagrangian of the EM field cancels out with the same term coming from ∇i Jij . The
←→
expressions for J and χ are given in the main manuscript.

13
New J. Phys. 26 (2024) 093041 F Khosravi et al

Also note that the EM Lagrangian does not include the interaction term jµc Aµ because it appears in the
Dirac part of the Lagrangian LD . Dirac Lagrangian, as mentioned earlier, vanishes from the expression for
the OAM of the Dirac field since it satisfies the Dirac equation. For this reason, the contribution from the
Dirac Lagrangian LD disappears from the conservation equations.
←
→
Equation (A.29) describes the local conservation law for the AM currents T and the AM density (charge)
M. When integrated over the entire space, and assuming that the fields vanish on the boundary of the this
surface, the second term becomes an integral over this surface and thus vanishes. In this case, we arrive at the
usual global conservation of AM equation which states that the total AM of the Dirac and Maxwell fields is a
constant. However, in situations where the problem under consideration is an open dissipative system, this
simplification cannot be made and surface terms of the AM current can carry AM out of the system.
Appendix B. Spin–orbit torque

In this section we show that equation (7) leads to equation (11). Starting with the equation for the spin of the
Dirac fields, we get for the z component for instance,

h̄
h̄c

∂λ SD12,λ = ∂t ψ † Σz ψ + ∂z ψ † γ 5 ψ (B.1)
2 2

where ∂t ≡ ∂
∂t and ∂i ≡ ∂
∂ xi . Using Dirac equation we get

h̄∂t ψ = −h̄cγ 0 (γ · ∇ψ) − iceAµ γ 0 γ µ ψ − imc2 γ 0 ψ (B.2a)

h̄∂t ψ † = h̄c ∇ψ † · γ γ 0 + iceAµ ψ † γ 0 γ µ + imc2 ψ † γ 0 (B.2b)

Using these equations we find

h̄ 


∂λ SD12,λ = ∂t ψ † Σz ψ + ψ † Σz (∂t ψ) + c∂z ψ † γ 5 ψ
2
1

+ h̄c ∇ψ † · γ iγ 0 γ 1 γ 2 ψ − ceAµ ψ † γ 0 γ µ γ 1 γ 2 ψ − mc2 ψ † γ 0 γ 1 γ 2 ψ
2

−i h̄cψ † γ 1 γ 2 γ 0 (γ · ∇ψ) + ceAµ ψ † γ 1 γ 2 γ 0 γ µ ψ + mc2 ψ † γ 1 γ 2 γ 0 ψ + h̄c∂z ψ † γ 5 ψ
1


= −h̄c∂z ψ † γ 5 ψ +i h̄c ∂x ψ † γ 0 γ 2 ψ − i h̄cψ † γ 0 γ 2 (∂x ψ)
2


−i h̄c ∂y ψ † γ 0 γ 1 ψ + i h̄cψ † γ 0 γ 1 ∂y ψ


+2ceA1 ψ † γ 0 γ 2 ψ − 2ceA2 ψ † γ 0 γ 1 ψ + h̄c∂z ψ † γ 5 ψ

   
= h̄cR i ψ̄ γ 1 ∂y − γ 2 ∂x ψ − ce A∥x ψ̄γ 2 ψ − A∥y ψ̄γ 1 ψ − A⊥ ⊥
x jc,y − Ay jc,x
n h    i o  
= −cR ψ̄ γ 1 −ih̄∂y − eA∥y − γ 2 −ih̄∂x − eA∥x ψ − A⊥ j
x c,y − A ⊥
j
y c,x
n   o

= −cR ψ̄ γ × p∥ ψ + jc × A⊥ z . (B.3)
z

In this derivation we have used the facts that {γ µ , γ ν } = 2η µν , (γ i )2 = −1, A1 = −Ax , A2 = −Ay , and
ij,λ
jc = ceψ̄γψ. We can do a similar derivation for other components of ∂λ SD . Doing so we find
equation (11a). Note that the first term on the r.h.s. of equation (B.3) is nothing but the spin–orbit torque of
the Dirac field τ D given in equation (12).
We now turn into the equation for the OAM of the Dirac field equation (7b). We get, for instance, for the
z component,
n o n o
∂λ L12
D
,λ
= ∂t R ψ † (xp∥y − yp∥x )ψ + c∇ · R ψ̄γ(xp∥y − yp∥x )ψ
n
= R (∂t ψ † )(xp∥y − yp∥x )ψ + ψ † (xp∥y − yp∥x )(∂t ψ) + c∇ · ψ̄γ(xp∥y − yp∥x )ψ
o
− eψ † ψ (x∂t A∥y − y∂t A∥x )
n i ce imc2 † 0 ∥
= R c(∇ψ † · γ)γ 0 (xp∥y −yp∥x )ψ + Aµ ψ † γ 0 γ µ (xp∥y −yp∥x )ψ + ψ γ (xpy −yp∥x )ψ
h̄ h̄
ice imc2 † ∥
− cψ † (xp∥y −yp∥x )γ 0 (γ · ∇ψ) − ψ † (xp∥y −yp∥x )γ 0 γ µ (Aµ ψ) − ψ (xpy −yp∥x )γ 0 ψ
h̄ o h̄
+ c∇ · ψ̄γ(xp∥y − yp∥x )ψ − eψ † ψ (x∂t A∥y − y∂t A∥x )

14
New J. Phys. 26 (2024) 093041 F Khosravi et al

n h i
= R −c∇ · ψ̄γ(xp∥y − yp∥x )ψ + cψ̄ γ · ∇(xp∥y −yp∥x ) ψ − ceψ̄γ µ ψ (x∂y − y∂x )Aµ
o
+ c∇ · ψ̄γ(xp∥y − yp∥x )ψ − eψ † ψ (x∂t A∥y − y∂t A∥x )
n o
= R cψ̄(γ 1 p∥y − γ 2 p∥x )ψ − ceψ̄γψ (x∇A∥y − y∇A∥x ) − ceψ̄γ µ ψ (x∂y − y∂x )Aµ
− eψ † ψ (x∂t A∥y − y∂t A∥x )
n o h
= R cψ̄(γ 1 p∥y − γ 2 p∥x )ψ − ceψ̄ γ 1 (x∂x A∥y − y∂x A∥x ) + γ 2 (x∂y A∥y − y∂y A∥x )
+ γ 3 (x∂z A∥y − y∂z A∥x ) − γ 1 (x∂y Ax − y∂x Ax ) − γ 2 (x∂y Ay − y∂x Ay ) − γ 3 (x∂y Az − y∂x Az )
1 1 i
+ γ 0 (x∂y ϕ − y∂x ϕ) + ψ † (x∂t A∥y − y∂t A∥x ) ψ
c
n oc
= R cψ̄(γ 1 p∥y − γ 2 p∥x )ψ − ceψ̄γ 1 ψ x(∂x A∥y − ∂y A∥x ) + ceψ̄γ 1 ψ (x∂y − y∂x )A⊥
x

− ceψ̄γ 2 ψ y(∂x A∥y − ∂y A∥x ) + ceψ̄γ 2 ψ (x∂y − y∂x )A⊥ ∥ ∥

y − ceψ̄γ ψ x(∂z Ay − ∂y Az )
3
h i
+ ceψ̄γ 3 ψ y(∂z A∥x −∂x A∥z ) +ceψ̄γ 3 ψ (x∂y −y∂x )A⊥ † ∥ ∥
z + eψ ψ x(−∂y ϕ−∂t Ay )−y(−∂x ϕ − ∂t Ax )
n o
= cR ψ̄(γ × p∥ )z ψ − jic (r × ∇)z A⊥ ∥
i + ρ(r × E )z (B.4)

where we have used the facts that ∇ × A∥ = 0 and E∥ = −∇ϕ − ∂t A∥ . Writing similar equations for the
other components we get equation (11b).
We can repeat this derivation for the EM spin and OAM currents as well. Using equation (7c) we find,
again for the z component for instance,

1
1

∂λ Sem
12,λ
= ϵ0 ∂ t E ⊥ × A ⊥ z
−∇ · A⊥ Bz + ∂z A⊥ · B
µ0 µ0
⊥

1 ⊥

1

⊥
= ϵ0 ∂ t E × A z − B z ∇ · A − A⊥ · ∇ Bz
µ0 µ0
1  ⊥
 ⊥
  ⊥


+ A · ∇ Bz + (B · ∇) A⊥ z + A × (∇ × B) z + B × ∇ × A z
µ0

1  
(B.5)
= ϵ0 ∂ t E ⊥ × A ⊥ + (B · ∇) A⊥
z − ϵ0 (∂t E) × A
⊥
z
− jc × A ⊥ z
µ0
" #
 ⊥ ⊥

 1 ∂E∥

= ϵ0 E × ∂ t A z
+ ⊥
(B · ∇) Az − ϵ0 × A − jc × A⊥ z
⊥
µ0 ∂t
!
1 ∂E∥

= (B · ∇) A⊥
z − ϵ0 × A ⊥ − jc × A ⊥ z ,
µ0 ∂t
z

where we have used the facts that ∇ · A⊥ = 0, B = ∇ × A⊥ , E⊥ = −∂t A⊥ , and ∇ × B = c12 ∂t E + µ0 jc . By

repeating this derivation for the other components we get equation (11c). Note that the first two terms are
the spin–orbit torque of the EM field, τ em , given in equation (13).
We can repeat this derivation to find the equation for ∂λ Lijem
,λ
. However, using the continuity condition,
 
ij,λ ij,λ
∂λ SD + LD + Sem
ij,λ
+ Lijem
,λ
= 0, (B.6)

it is straightforward to show that equation (11d) holds.

Appendix C. Two plane wave interference

We now evaluate the terms in equation (11c) for the interference of two plane waves at different frequencies.
For a plane wave propagating along k/|k|, with the wavevector k we have k · E = k · A⊥ = k · B = 0. Therefore
we get

∇ · A⊥ B = k · A⊥ B = 0. (C.1)

15
New J. Phys. 26 (2024) 093041 F Khosravi et al

Note also that the EM spin–orbit torque τ em also vanishes for planes waves because

(B · ∇) A⊥ = (B · k) A⊥ = 0. (C.2)

Thus the only relevant terms in finding the conservation law for the spin current of the two plane wave
interference are the time-derivative of spin and gradient of helicity.
The electric field for two plane waves propagating along z direction can be written as
n o
E = R E1 e−i(ω1 t−k1 z) + E2 e−i(ω2 t−k2 z) (C.3)

Ei
where E i = √
2
(x̂ + iŷ) are the complex electric field amplitudes of the two modes with frequencies
ωi /c = ki . Using Maxwell equation ∇ × E = − ∂∂Bt and E = − ∂∂At , we get
 
k1 k2
B=R − E1 e−i(ω1 t−k1 z) − E2 e−i(ω2 t−k2 z) (C.4a)
iω1 i ω2
 
1 1
A⊥ = R E1 e−i(ω1 t−k1 z) + E2 e−i(ω2 t−k2 z) . (C.4b)
iω1 iω2

Using these equations we get for the spin of the two plane waves
   n o
⊥ ⊥

ϵ 1 ∗ 1 ∗ 1 1 ∗ +i[(ω1 −ω2 )t−(k1 −k2 )z]
ϵ E ×A = I {E 1 × E 1 } + I {E 2 × E 2 } + + I E 1 × E 2e
2 ω1 ω2 ω1 ω2
(C.5)

and thus we get

∂ ⊥
ω 2 − ω22 n ∗ o
ϵ E × A⊥ = − ϵ 1 R E 1 × E 2 e+i[(ω1 −ω2 )t−(k1 −k2 )z]
∂t 2ω1 ω2
(C.6)
ω 2 − ω22 n o
= −ϵ 1 I E1 E2∗ e−i[(ω1 −ω2 )t−(k1 −k2 )z] ẑ.
2ω1 ω2
For the helicity density we find
k1 k2 k1 + k2 n o
A⊥ · B = − |E1 | 2
− |E 2 | 2
− R E1 E ∗ −i[(ω1 −ω2 )t−(k1 −k2 )z]
2 e (C.7)
ω12 ω22 2ω1 ω2

and thus
1
1 k21 − k22 n o
∇ A⊥ · B = I E1 E2∗ e−i[(ω1 −ω2 )t−(k1 −k2 )z] ẑ
µ0 µ0 2ω1 ω2
(C.8)
ω 2 − ω22 n o
=ϵ 1 I E1 E2∗ e−i[(ω1 −ω2 )t−(k1 −k2 )z]
2ω1 ω2
which confirms the conservation equation (16).

Appendix D. Semi-local conservation law

We can get the semi-local conservation laws by integrating the terms in equation (7) over the volume V ′ , on
which surface the Dirac eigenfunctions ψ become zero. Using Gauss’s theorem, the integral of the terms
∇(ψ † γ 5 ψ) and ∇ · R{ψ̄γ(r × p)ψ } vanish because they become surface integrals of functions of ψ. We
therefore arrive at the semi-local conservation law
ˆ
∂ M̃D ∂ M̃em
+ + J̃A + h̃ + ∇ × (rNem ) dV ′ = 0 (D.1)
∂t ∂t V′

where M̃D is the total AM of the electronic field given by

ˆ  
1 †
 †
M̃D = h̄ ψ Σψ + R ψ (r × p) ψ dV ′ , (D.2)
V′ 2

M̃em is the EM total AM in the volume V ′ ,

ˆ ˆ

 
M̃em = ϵ E⊥ × A⊥ d3 x + ϵ E⊥ · (r × ∇) A⊥ d3 x, (D.3)
V′ V′

16
New J. Phys. 26 (2024) 093041 F Khosravi et al

and
ˆ   
1 ⊥ 1 ⊥ ⊥ ∂E∥
J̃A = − ∇· A B + B × (r × ∇) A + ϵA r× dV ′ , (D.4a)
V′ µ0 µ0 ∂t
ˆ


h̃ = ∇ A⊥ · B dV ′ . (D.4b)
V′

Using Gauss’s theorem the volume integrals in equations (D.1), (D.4a) and (D.4b) can be converted into
surface integrals. We find:
ˆ   
∂E∥
J̃A = − n̂i A⊥,i B + ϵij k Bj (r × ∇) A⊥,k + ϵA⊥,i r × da, (D.5a)
′ ∂t
ˆ S


h̃ = n̂ A⊥ · B da, (D.5b)
S′

where n̂ is the unit vector normal to the surface of the volume V ′ , and da its surface element. Equation (D.1)
presents an equation for the time evolution of AM of the electronic field in terms of the EM fields.

Appendix E. Boost conservation relations

We started our derivation by applying the Noether’s theorem to the Dirac-Maxwell fields under the Lorentz
transformation. To derive the AM conservation equations, however, we only focused on the spatial rotations
of the coordinates i.e. the space-space components of the AM tensor current Mµν,λ . In this section, we look
at the conservation equations for the time-space components of the Lorentz transformations, namely the
boosts, of the Dirac–Maxwell fields.
Using a similar approach to the one used in the first section, the conservation equation resulting from the
boost components of the Lorentz transformation is given by,

i0,λ i0,λ
∂λ Mi0,λ = ∂λ (MD ) + ∂λ (ME )
 
∂ n † hr ∥
i o h̄ † n hr i o
= R iψ p0 − ctp ψ − c∇ × ψ Σψ + c∇.R iψ̄γ p0 − ctp∥ ψ
∂t c 2 c
∂ h  r  i 1 h  r  i 1 1
−ϵ E. ∂t + ct∇ A⊥ + ∇. B × ∂t + ct∇ A⊥ + B × E⊥ + E · (∇) A⊥
∂t c µ0 c µ0 c µ0 c
r r  r 
+ cρA⊥ − J.E∥ + ctρE∥ + ϵE. ∂t + ct∇ E∥ + c 2 ∂t + t∇ LE = 0,
c c c
(E.1)

where p0 = ih̄∂t − eϕ and p∥ = −ih̄∇ − eA∥ are the gauge-independent time-derivative and momentum
operators of the Dirac field, respectively. It is a matter of straightforward algebra to show that
r × ∂λ M0,λ = εijk xj ∂λ Mk0,λ gives the conservation equation of the AM in equation (8).

Appendix F. Symmetrized angular momentum tensor

In our derivation, we have used the canonical form of the AM tensor which is derived directly from the
application of Noether theorem to the QED Lagrangian in equation (1). The AM tensor can be written in
terms of the canonical energy-momentum tensor, T µν , as

Mµν,λ = xµ Tλν − xν Tλµ + Sµν,λ . (F.1)

Conservation of AM implies that, when ∂λ Sµν,λ ̸= 0,

Tµν ̸= Tνµ . (F.2)

This in unpleasant because in general relativity, the energy-momentum tensor is directly proportional to the
metric tensor which is symmetric in µ and ν. To overcome this problem, the energy momentum tensor can
be modified to the so-called Bellifante–Resenfeld energy-momentum tensor as [22, 67]

′ 1

T µν = Tµν + ∂λ Sνλ,µ + Sµλ,ν − Sνµ,λ . (F.3)
2

17
New J. Phys. 26 (2024) 093041 F Khosravi et al

It can be shown that this new energy-momentum tensor is symmetric and does not change the conservation
′
law of the energy-momentum tensor, ∂µ T µν = 0. We can therefore write a new symmetrized AM tensor,
′
M µν,λ as
′ ′ ′
M µν,λ = xµ T λν − xν T λµ (F.4)

where we do not need to include the additional spin tensor because it is already present in the symmetric
energy-momentum tensor. This symmetrized AM tensor is of course different from the canonical one we
derived in equation (6). However, it is still gauge invariant and the conservation law of AM still holds. In
other words,
′
∂λ M µν,λ = ∂λ Mµν,λ = 0. (F.5)

We can follow a similar procedure as we did in the previous section to derive the expression for the
gauge-independent forms of the symmetrized AM tensor for the EM and Dirac fields. We find, setting
µ, ν = i, j,

∂M
′
←→′
+ ∇ · J + ∇χ ′ − ∇ × (rUem ) = 0 (F.6)
∂t
where
′ h̄ †
n   o

M = ψ Σψ + R ψ † r × p∥ ψ + ϵ0 r × (E × B) − ρ r × A⊥ (F.7a)
2
←→′
n   o 1 

∇ · J = c∇i R ψ̄γ i r × p∥ ψ + ϵ0 ∇i Ei (r × E) + ∇i [Bi (r × B)] − ∇i Ji r × A⊥ (F.7b)
µ0
h̄c † 5

χ̃ = ψ γ ψ (F.7c)
2 
1 1
Uem = ϵ0 E · E + B · B . (F.7d)
2 µ0
We emphasize that equation (F.6) is identical to the conservation equation (8). In fact, the terms related
to the Dirac field are exactly the same as the one with the four-divergence of the canonical AM tensor. The
main difference is that we lose separate physically observable expressions for the spin and OAM densities and
currents of the EM field and instead we get expressions for the total AM,


ϵ0 r × (E × B) − ρ r × A⊥ , (F.8)

and total AM currents,

1

ϵ0 E (r × E) + B (r × B) − J r × A⊥ , (F.9)
µ0
of the EM field. Assuming that the symmetric energy-momentum tensor is more fundamental than the
canonical one, however, does not prevent us from writing conservation laws that involve canonical AM
tensor and taking its gauge-independent terms as physically meaningful quantities.

Appendix G. Meaning of the notations

Throughout this paper, the usual expressions for dot and cross product are assumed. Terms like A⊥ B are
tensorial expressions which can be expanded as

A⊥ B = A⊥ ⊥ ⊥
x Bx x̂x̂ + Ax By x̂ŷ + Ax Bz x̂ẑ

+ A⊥ ⊥ ⊥
y Bx ŷx̂ + Ay By ŷŷ + Ay Bz ŷẑ (G.1)
+ A⊥ ⊥ ⊥
z Bx ẑx̂ + Az By ẑŷ + Az Bz ẑẑ.

Similar expressions can be written the terms like E(r × E), E(r × E), J(r × A⊥ ), and so on. Therefore, the
expression ∇ · (A⊥ B) means

h
 
i
∇ · A⊥ B = ∂x A⊥ ⊥
x Bx + ∂y Ay Bx + ∂z Az Bx
⊥
x̂
h
 
i
+ ∂x A⊥ B
x y + ∂ y A ⊥
B
y y + ∂ z A ⊥
B
z y ŷ (G.2)
h
 
i
+ ∂x A⊥ ⊥
x Bz + ∂y Ay Bz + ∂z Az Bz
⊥
x̂

18
New J. Phys. 26 (2024) 093041 F Khosravi et al

which is a vector. We can similar expand

h

⊥i
B × (r × ∇) A⊥ = By y∂z − z∂y A⊥ z − Bz y∂z − z∂y Ay x̂x̂


⊥
+ Bz y∂z − z∂y A⊥ x − Bx y∂z − z∂y Az ŷx̂
h

⊥i
+ Bx y∂z − z∂y A⊥ y − B y y∂ z − z∂ y Ax ẑx̂
h i
+ By (z∂x − x∂z ) A⊥ ⊥
z − Bz (z∂x − x∂z ) Ay x̂ŷ
 
+ Bz (z∂x − x∂z ) A⊥ ⊥
x − Bx (z∂x − x∂z ) Az ŷŷ (G.3)
h i
+ Bx (z∂x − x∂z ) A⊥
y − B y (z∂ x − x∂ z ) A ⊥
x ẑŷ
h

⊥i
+ By x∂y − y∂x A⊥ z − Bz x∂y − y∂x Ay x̂ẑ


⊥
+ Bz x∂y − y∂x A⊥ x − Bx x∂y − y∂x Az ŷẑ
h

⊥i
+ Bx x∂y − y∂x A⊥ y − By x∂y − y∂x Ax ẑẑ

and we can take its divergence similar to equation (G.2).

ORCID iDs

Li-Ping Yang  https://orcid.org/0000-0002-7266-4901

Pronoy Das  https://orcid.org/0000-0002-6441-7177
Zubin Jacob  https://orcid.org/0009-0005-5016-9175

References
[1] Cohen-Tannoudji C, Dupont-Roc J and Grynberg G 1997 Photons and Atoms-Introduction to Quantum Electrodynamics
[2] Willner A E, Pang K, Song H, Zou K and Zhou H 2021 Appl. Phys. Rev. 8 041312
[3] Bustamante C J, Chemla Y R, Liu S and Wang M D 2021 Rev. Dis. Primers 1 25
[4] Crimin F, Mackinnon N, Götte J B and Barnett S M 2019 Appl. Sci. 9 828
[5] Bliokh K Y, Bekshaev A Y and Nori F 2013 New J. Phys. 15 033026
[6] Berry M V 2009 J. Opt. A: Pure Appl. Opt. 11 094001
[7] Nienhuis G 2016 Phys. Rev. A 93 023840
[8] Fernandez-Corbaton I, Zambrana-Puyalto X, Tischler N, Vidal X, Juan M L and Molina-Terriza G 2013 Phys. Rev. Lett. 111 060401
[9] Tang Y and Cohen A E 2010 Phys. Rev. Lett. 104 163901
[10] Philbin T G 2013 Phys. Rev. A 87 043843
[11] Greiner W et al 1990 Relativistic Quantum Mechanics vol 3 (Springer)
[12] Barnett S M 2014 New J. Phys. 16 093008
[13] Bialynicki-Birula I 1996 Prog. Opt. 36 245–94
[14] Khosravi F, Van Mechelen T and Jacob Z 2019 Phys. Rev. B 100 155105
[15] Van Mechelen T and Jacob Z 2019 Phys. Rev. B 99 205146
[16] Van Mechelen T and Jacob Z 2019 Nanophotonics 8 1399–416
[17] Greiner W, Müller B and Bromley D A 1996 Gauge Theory of Weak Interactions vol 5 (Springer)
[18] Noether E 1971 (Reprint) Transport Theory and Statistical Physics 1 186–207
[19] Barnett S M 2002 J. Opt. B 4 S7–S16
[20] Calkin M G 1965 Am. J. Phys. 33 958–60
[21] Leader E and Lorcé C 2014 Phys. Rep. 541 163–248
[22] Belinfante F 1940 Physica 7 449–74
[23] Ji X 1997 Phys. Rev. Lett. 78 610–3
[24] Jaffe R L and Manohar A 1990 Nucl. Phys. B 337 509–46
[25] Chen X S, Lü X F, Sun W M, Wang F and Goldman T 2008 Phys. Rev. Lett. 100 232002
[26] Wakamatsu M 2010 Phys. Rev. D 81 114010
[27] Yang L P, Khosravi F and Jacob Z 2022 Phys. Rev. Res. 4 023165
[28] Das P, Bharadwaj S and Jacob Z 2024 New J. Phys 26 083008
[29] Foldy L L and Wouthuysen S A 1950 Phys. Rev. 78 29–36
[30] Terno D R 2003 Phys. Rev. A 67 014102
[31] Fujikawa K, Oh C H and Zhang C 2014 Phys. Rev. D 90 025028
[32] Greiner W and Reinhardt J 2013 Field Quantization (Springer Science & Business Media)
[33] Kleinert H 2016 Particles and Quantum Fields (World Scientific Singapore)
[34] Jackson J D 1999 Classical Electrodynamics 3rd edn (Wiley) ch 6.7
[35] Woltjer L 1958 Proc. Natl. Acad. Sci. USA 44 489
[36] Coles M M and Andrews D L 2012 Phys. Rev. A 85 063810
[37] Leeder J M, Haniewicz H T and Andrews D L 2015 J. Opt. Soc. Am. B 32 2308–13
[38] Lipkin D M 1964 J. Math. Phys. 5 696–700
[39] Meredith L 2018 Helicity, chirality, and the Dirac equation in the non-relativistic limit (available at: https://www.sas.rochester.edu/
pas/assets/pdf/undergraduate/helicity_chirality_and_the_dirac_equation_in_the_non-relativistic_limit.pdf)

19
New J. Phys. 26 (2024) 093041 F Khosravi et al

[40] McCreary K M, Swartz A G, Han W, Fabian J and Kawakami R K 2012 Phys. Rev. Lett. 109 186604
[41] Zhao H, Loren E J, van Driel H M and Smirl A L 2006 Phys. Rev. Lett. 96 246601
[42] Wolf S A, Awschalom D D, Buhrman R A, Daughton J M, von Molnár S, Roukes M L, Chtchelkanova A Y and Treger D M 2001
Science 294 1488–95
[43] Sharma P 2005 Science 307 531–3
[44] Watson S K, Potok R M, Marcus C M and Umansky V 2003 Phys. Rev. Lett. 91 258301
[45] Kashiwaya S, Tanaka Y, Yoshida N and Beasley M R 1999 Phys. Rev. B 60 3572–80
[46] Crater H W and Alstine P V 1983 Ann. Phys., NY 148 57–94
[47] Salpeter E E and Bethe H A 1951 Phys. Rev. 84 1232–42
[48] Okamoto K 2006 Optical Fibers Fundamentals of Optical Waveguides 2nd edn, ed K Okamoto (Academic) ch 3, pp 57–158
[49] Manchon A, Koo H C, Nitta J, Frolov S and Duine R 2015 Nat. Mater. 14 871–82
[50] Enk S V and Nienhuis G 1994 J. Mod. Opt. 41 963–77
[51] Nieto-Vesperinas M 2015 Opt. Lett. 40 3021–4
[52] Malinowski G, Dalla Longa F, Rietjens J, Paluskar P, Huijink R, Swagten H and Koopmans B 2008 Nat. Phys. 4 855
[53] Mitsch R, Sayrin C, Albrecht B, Schneeweiss P and Rauschenbeutel A 2014 Nat. Commun. 5 5713
[54] Schmiegelow C T, Schulz J, Kaufmann H, Ruster T, Poschinger U G and Schmidt-Kaler F 2016 Nat. Commun. 7 12998
[55] Young A B, Thijssen A C T, Beggs D M, Androvitsaneas P, Kuipers L, Rarity J G, Hughes S and Oulton R 2015 Phys. Rev. Lett.
115 153901
[56] Söllner I et al 2015 Nat. Nanotechnol. 10 775
[57] Barik S Karasahin A Mittal S Waks E and Hafezi M 2019 Chiral quantum optics using a topological resonator (arXiv:1906.11263)
[58] Das P, Yang L P and Jacob Z 2024 J. Opt. Soc. Am. B 41 1764
[59] Yang L P and Jacob Z 2021 Commun. Phys. 4 1–9
[60] Kalhor F, Yang L P, Bauer L and Jacob Z 2021 Phys. Rev. Res. 3 043007
[61] Wolkow D M 1935 Z. Phys. 94 250–60
[62] Redmond P J 1965 J. Math. Phys. 6 1163–9
[63] MacDonald A H and Vosko S H 1979 J. Phys. C 12 2977–90
[64] Eschrig H and Servedio V D P 1999 J. Comput. Chem. 20 23–30
[65] Mondal R, Berritta M and Oppeneer P M 2018 Phys. Rev. B 98 214429
[66] Yang L P, Khosravi F and Jacob Z 2020 arXiv:2004.03771
[67] Rosenfeld L J H C 1940 Sur le Tenseur d’Impulsion-éNergie (Palais des Académies)

20