Energy and momentum in the electromagnetic field: Poynting vector, Maxwell stress tensor, conservation laws

Intuition Beginner

Electric and magnetic fields are not just bookkeeping devices for forces on charges — they themselves carry energy and momentum, in the same operational sense that a moving baseball carries energy and momentum. You can extract energy from a field (a microwave oven heats food using oscillating electromagnetic fields), and you can push an object purely by hitting it with light (a solar sail accelerates because sunlight delivers momentum to its absorbing surface).

The quantity that tracks energy flow through space is the Poynting vector $S=E\times B/{\mu }_0$. It measures how many joules per second pass through one square metre of imaginary surface, oriented so that the surface normal points along $S$. The units are watts per square metre. At the surface of the Earth at noon, sunlight delivers about $1361W/{\mathrm{m}}^2$ — this is the solar constant, and it is the magnitude of the Sun's Poynting vector evaluated at Earth's orbit.

Where does the energy in a capacitor come from when you charge it through a wire? Common sense says: the electrons carry it down the wire. The Poynting picture says something stranger and more accurate: the energy flows in radially from the side, through the empty space around the wire, guided by the field that surrounds the wire. The wire is the channel; the field is what does the carrying.

Light also pushes. Each photon carries momentum equal to its energy divided by the speed of light, $p=E/c$. When light strikes a black absorber, the momentum is deposited as a small but real force. Maxwell predicted this radiation pressure in 1873; Lebedev measured it in 1900 in a torsion-balance experiment, finding the predicted micro-newton-per-square-metre forces.

A practical example: a solar sail with one square metre of surface area, sitting at the same distance from the Sun as Earth, experiences a force of about nine micro-newtons if it absorbs all the light, and twice that if it reflects all the light. This is tiny per unit area but builds momentum steadily over time; NASA's IKAROS spacecraft launched in 2010 was the first vehicle to demonstrate this propulsion in space.

Visual Beginner

Picture a wire running horizontally with a current flowing through it. Around the wire, the magnetic field $B$ forms circles in the plane perpendicular to the wire. Along the wire, an electric field $E$ points in the direction of current flow (because the wire has a small resistive voltage drop). The Poynting vector $S=E\times B/{\mu }_0$ at every point near the wire points inward toward the wire — perpendicular to its surface. Energy is flowing into the wire from all around it, then being dissipated as heat in the metal.

The picture captures the central counter-intuitive fact: the energy that heats the wire is not flowing through the metal alongside the electrons; it is flowing through the empty space around the wire and entering the metal sideways. Removing the surrounding field stops the energy delivery, even with the same current.

Worked example Beginner

A perfectly absorbing solar sail with area $A=1.0{\mathrm{m}}^2$ is positioned at one astronomical unit from the Sun (Earth's orbital distance), facing the Sun. The local intensity of sunlight is $I=1361{W/m}^2$. What force does the sunlight exert on the sail?

Step 1. The momentum carried by light is its energy divided by the speed of light. Each second, the sail absorbs energy $E=I\cdot A\cdot t$ from the incoming sunlight (with $t=1\mathrm{s}$).

Step 2. The momentum delivered per second is $p/t=(IA)/c$, where $c=3.0\times 10^{8}m/s$ is the speed of light.

Step 3. By Newton's second law, the rate of momentum transfer is the force on the sail:

$$F=\frac{IA}{c}=\frac{(1361)(1.0)}{3.0\times 10^8}=4.5\times 10^{-6}\mathrm{N}.$$

What this tells us: a one-square-metre absorbing sail feels a force of about $4.5\mu \mathrm{N}$. If the sail were a perfect mirror, it would feel twice this — about $9\mu \mathrm{N}$ — because each photon's momentum reverses direction, so the impulse is doubled. Forces this small are unusable for short trips but accumulate steadily; over months, a large sail can gain kilometres per second of velocity without consuming fuel.

Check your understanding Beginner

Formal definition Intermediate+

The electromagnetic field carries energy density and energy flux. The energy density is the scalar field

$$u=\frac{{ϵ}_0}{2}|E{|}^2+\frac{1}{2{\mu }_0}|B{|}^2,$$

with units of joules per cubic metre. The energy flux is the Poynting vector

$$S=\frac{1}{{\mu }_0}E\times B,$$

with units of watts per square metre. These two objects together satisfy the local conservation law known as Poynting's theorem.

Definition (Poynting's theorem). For Maxwell's equations in vacuum (or in a linear medium with constant $ϵ,\mu$), the energy density $u$ and Poynting vector $S$ obey

$$\frac{\partial u}{\partial t}+\nabla \cdot S=-J\cdot E.$$

The right-hand side is the rate at which the field does work on the free currents per unit volume. The integral form, obtained by integrating over a region $V$ with boundary $\partial V$ carrying outward area element $dA$, is

$$-\frac{d}{dt}{\int }_{V}u{d}^{3}r={\oint }_{\partial V}S\cdot dA+{\int }_{V}J\cdot E{d}^{3}r.$$

Conservation interpretation: the decrease of field energy in $V$ equals the energy that flowed out across $\partial V$ plus the energy delivered as work to the charges in $V$.

Definition (electromagnetic momentum density). The momentum per unit volume carried by the field is

$$g={ϵ}_{0}E\times B=\frac{S}{c^2}.$$

The identification of momentum density with $S/c^2$ is the field analogue of the relation $p=E/c$ for a single photon.

Definition (Maxwell stress tensor). The symmetric tensor

$$T_{ij}={ϵ}_0\left(E_i{E}_j-\frac{1}{2}{\delta }_{ij}|E{|}^2\right)+\frac{1}{{\mu }_0}\left(B_i{B}_j-\frac{1}{2}{\delta }_{ij}|B{|}^2\right)$$

is the Maxwell stress tensor. Its $ij$ component gives the $i$-component of force per unit area transmitted across a surface whose outward normal is in the $j$ direction.

The total force on the charges in a region $V$ is then

$$F_i={\oint }_{\partial V}{T}_{ij}d{A}_j-\frac{d}{dt}{\int }_V{g}_i{d}^{3}r.$$

The first term is the flux of momentum into $V$ through the boundary; the second is the rate at which momentum stored in the field inside $V$ is changing.

Counterexamples to common slips

• The Poynting vector $S$ is not unique. Adding the curl of any vector field $\nabla \times G$ to $S$ leaves the conservation equation $\partial u/\partial t+\nabla \cdot S=-J\cdot E$ unchanged, because $\nabla \cdot (\nabla \times G)=0$. The convention $S=E\times B/{\mu }_0$ is the standard choice because it transforms covariantly under Lorentz transformations and matches the off-diagonal components of the four-tensor $T^{\mu \nu }$; alternative forms exist in dielectric media (the Abraham-Minkowski ambiguity).
• Energy density $u$ is non-negative, but the local force per unit area $T_{ij}{n}_j$ can point either way. The diagonal of the stress tensor has the form $T_{ii}={ϵ}_0(E_i^2-|E{|}^2/2)+(B_i^2-|B{|}^2/2)/{\mu }_0$, which is positive along field lines (tension) and negative perpendicular to them (pressure). Faraday described the field as having "tension along the lines of force and pressure transverse to them"; the modern stress tensor makes this precise.
• The Poynting energy flux is non-zero in static crossed E and B fields, even when nothing is moving. A perpendicular static $E$ and $B$ pair gives $S\ne 0$, indicating field momentum even though no energy is being transported in any operational sense. The Feynman disk paradox uses this static field momentum to construct a non-conservation of mechanical angular momentum that is restored only when the field's angular momentum is included.

Key theorem with proof Intermediate+

Theorem (Poynting's theorem, Poynting 1884). Let $E,B$ satisfy Maxwell's equations in vacuum with free current density $J$. Then

$$\frac{\partial u}{\partial t}+\nabla \cdot S=-J\cdot E,$$

where $u={ϵ}_0|E{|}^2/2+|B{|}^2/(2{\mu }_0)$ and $S=E\times B/{\mu }_0$.

Proof. Start from the work done on charges. The rate at which the electromagnetic field does work on charges in a region per unit volume is

$$J\cdot E,$$

since the magnetic force $qv\times B$ does no work (it is perpendicular to $v$). Our goal is to rewrite this expression entirely in terms of fields, then identify it as $-\partial u/\partial t-\nabla \cdot S$.

Use the Ampère-Maxwell law to eliminate $J$:

$$J=\frac{1}{{\mu }_0}\nabla \times B-{ϵ}_0\frac{\partial E}{\partial t}.$$

Substituting,

$$J\cdot E=\frac{1}{{\mu }_0}E\cdot (\nabla \times B)-{ϵ}_{0}E\cdot \frac{\partial E}{\partial t}.$$

Apply the vector identity $\nabla \cdot (A\times B)=B\cdot (\nabla \times A)-A\cdot (\nabla \times B)$ with $A=E,B=B$ to get $E\cdot (\nabla \times B)=B\cdot (\nabla \times E)-\nabla \cdot (E\times B)$:

$$J\cdot E=\frac{1}{{\mu }_0}B\cdot (\nabla \times E)-\frac{1}{{\mu }_0}\nabla \cdot (E\times B)-{ϵ}_{0}E\cdot \frac{\partial E}{\partial t}.$$

Use Faraday's law $\nabla \times E=-\partial B/\partial t$:

$$J\cdot E=-\frac{1}{{\mu }_0}B\cdot \frac{\partial B}{\partial t}-\nabla \cdot \left(\frac{1}{{\mu }_0}E\times B\right)-{ϵ}_{0}E\cdot \frac{\partial E}{\partial t}.$$

Recognise the two time derivatives as derivatives of squared field magnitudes:

$${ϵ}_{0}E\cdot \frac{\partial E}{\partial t}=\frac{\partial }{\partial t}\left(\frac{{ϵ}_0}{2}|E{|}^2\right),\frac{1}{{\mu }_0}B\cdot \frac{\partial B}{\partial t}=\frac{\partial }{\partial t}\left(\frac{1}{2{\mu }_0}|B{|}^2\right).$$

Therefore

$$J\cdot E=-\frac{\partial }{\partial t}\left(\frac{{ϵ}_0}{2}|E{|}^2+\frac{1}{2{\mu }_0}|B{|}^2\right)-\nabla \cdot \left(\frac{E\times B}{{\mu }_0}\right)=-\frac{\partial u}{\partial t}-\nabla \cdot S.$$

Rearranging gives the conservation form $\partial u/\partial t+\nabla \cdot S=-J\cdot E$. $■$

Bridge. Poynting's theorem is the foundational reason that electromagnetic radiation carries energy: a propagating wave packet in vacuum has $J=0$, so $\partial u/\partial t+\nabla \cdot S=0$ — a pure continuity equation for field energy with no source. This is exactly the relation that builds toward 10.07.01 radiation patterns and the Larmor formula, where an accelerating charge produces a far-field whose Poynting flux integrated over a large sphere is the radiated power. The momentum form of the theorem — the identification of $g=S/c^2$ as momentum density and the Maxwell stress tensor as momentum flux — is dual to the energy form; putting these together they form the spatial $0\nu$ and $i\nu$ components of the four-tensor that appears again in 10.06.02 pending as $T^{\mu \nu }$, the covariant stress-energy of the electromagnetic field. The same divergence identity in covariant form expresses both energy and momentum conservation as the single equation ${\partial }_{\mu }{T}^{\mu \nu }=-F^{\nu \alpha }{J}_{\alpha }$, identifying the relativistic four-force on the charge current with minus the divergence of the field's stress-energy.

Exercises Intermediate+

Advanced results Master

Theorem 1 (covariant electromagnetic stress-energy tensor, Minkowski 1908). In a Lorentzian spacetime with metric signature $(+,-,-,-)$, the symmetric traceless rank-2 tensor

$$T_{\mathrm{EM}}^{\mu \nu }=\frac{1}{{\mu }_0}\left[F^{\mu \alpha }{F}^{\nu }{}_{\alpha }-\frac{1}{4}{\eta }^{\mu \nu }{F}_{\alpha \beta }{F}^{\alpha \beta }\right]$$

has components $T^{00}=u$ (energy density), $T^{0i}=S^i/c$ (momentum density times $c$, equivalently energy flux divided by $c$), and $T^{ij}$ equals the negative of the spatial Maxwell stress tensor $-T_{ij}^{\mathrm{Maxwell}}$ in standard conventions.

The conservation law is ${\partial }_{\mu }{T}_{\mathrm{EM}}^{\mu \nu }=-F^{\nu \alpha }{J}_{\alpha }$, expressing four-momentum conservation for the field plus minus-the-Lorentz-four-force on the charge current. The trace vanishes: $T^{\mu }{}_{\mu }=0$. This tracelessness reflects the conformal invariance of source-free electromagnetism in four dimensions.

Theorem 2 (symmetry and Belinfante-Rosenfeld procedure, Belinfante 1940). The canonical Noether stress tensor for the electromagnetic Lagrangian $\mathcal{L}=-F_{\mu \nu }{F}^{\mu \nu }/(4{\mu }_0)$ is not symmetric, but it differs from the symmetric tensor $T_{\mathrm{EM}}^{\mu \nu }$ above by the divergence of an antisymmetric tensor ${\partial }_{\lambda }{\Sigma }^{\lambda \mu \nu }$. Belinfante's symmetrisation procedure adds this divergence-free term to recover symmetry.

The Belinfante symmetric tensor coincides with the metric (Hilbert) stress tensor obtained by varying the gravitational action $\int \mathcal{L}\sqrt{-g}{d}^{4}x$ with respect to the metric: $T^{\mu \nu }=(2/\sqrt{-g})\delta (\sqrt{-g}\mathcal{L})/\delta g_{\mu \nu }$. Symmetry of $T^{\mu \nu }$ is what makes the angular-momentum tensor $M^{\mu \nu \lambda }=x^{\mu }{T}^{\nu \lambda }-x^{\nu }{T}^{\mu \lambda }$ conserved: ${\partial }_{\lambda }{M}^{\mu \nu \lambda }=T^{\nu \mu }-T^{\mu \nu }=0$.

The asymmetry of the canonical Noether stress tensor reflects an additional spin-angular-momentum current that mixes orbital and spin contributions; symmetrising by adding ${\partial }_{\lambda }{\Sigma }^{\lambda \mu \nu }$ (with $\Sigma$ antisymmetric in its first two indices, hence divergence-free) reorganises the conserved angular momentum into a purely orbital form $r\times p$ where $p$ is the symmetric momentum density. This Belinfante-Rosenfeld procedure is the prototype for coupling matter to gravity: the same machinery applied to Dirac fermions yields the symmetric stress tensor that sources the Einstein equations, including the half-integer spin tied to spinor representations.

Theorem 3 (Feynman disk paradox, Feynman Lectures Vol. II §17.4 + §27.6). A static perpendicular pair of $E$ and $B$ fields carries non-zero EM angular momentum density $\ell =r\times g$, even though the fields do not appear to be transporting energy in any conventional sense. When the source of one field is removed quasi-statically, the EM angular momentum is transferred to mechanical motion of the charges, preserving total angular momentum.

Feynman's concrete example: a thin charged plastic disk free to rotate on its axis, threaded by a solenoid producing a uniform axial $B$ field. Static $E$ from the charges, perpendicular static $B$ from the solenoid, $E\times B$ tangential to the disk, integrated angular momentum $L_{\mathrm{EM}}$ non-zero. Switching off the solenoid current induces an azimuthal $E$ by Faraday's law that torques the disk; the disk starts spinning, picking up exactly $L_{\mathrm{EM}}$ of mechanical angular momentum. The paradox is resolved: the apparent missing angular momentum was sitting in the field, not in any moving charges, and is recovered by integrating $r\times ({ϵ}_{0}E\times B)$ over space.

A quantitative version: a uniformly charged spherical shell of radius $R$ with total charge $Q$ sits inside a long solenoid producing $B=B_0\hat{z}$ along its axis. The electric field outside the sphere is radial, $E=(Q/(4\pi {ϵ}_0{r}^2))\hat{r}$; inside the solenoid the magnetic field is uniform. The angular momentum density at a point with cylindrical radius $\rho$ from the solenoid axis is ${\ell }_z={ϵ}_0\rho [E\times B{]}_{\varphi }$. Integrating over the volume between the sphere and the solenoid wall (using the fact that the sphere's external field is exactly that of a point charge) yields $L_{\mathrm{EM}}=(2/3)Q{B}_0{R}^2$ when the integral is regulated appropriately. Switching off $B_0$ over a finite time induces an azimuthal $E$ that delivers torque $\tau =-d{L}_{\mathrm{EM}}/dt$ to the charges on the sphere, spinning the sphere up to exactly the angular momentum the field originally carried. The bookkeeping closes only when the field's angular momentum is included; mechanical angular momentum alone is not conserved.

Theorem 4 (Abraham-Lorentz radiation-reaction force, Abraham 1903; Lorentz 1916). For a non-relativistic charged particle of charge $q$ and acceleration $a$, the radiation-reaction force in the leading-order approximation is

$$F_{\mathrm{rad}}=\frac{{\mu }_0{q}^2}{6\pi c}\dot{a}.$$

Time-averaging $F_{\mathrm{rad}}\cdot v$ over periodic motion recovers the Larmor radiated power $P={\mu }_0{q}^2|a{|}^2/(6\pi c)$.

The Abraham-Lorentz equation $m\dot{v}=F_{\mathrm{ext}}+({\mu }_0{q}^2/6\pi c)\ddot{v}$ exhibits pre-acceleration and runaway solutions, signalling its breakdown at frequencies above the inverse Compton time. The full covariant version (the Lorentz-Dirac equation) was derived by Dirac in 1938 Proc. R. Soc. A 167, 148; its more practical reduced form (Landau-Lifshitz) avoids the pathological solutions by treating the radiation-reaction term perturbatively. Modern quantum-electrodynamic treatment is exact and finite, recovering Abraham-Lorentz only as the non-relativistic, soft-photon limit.

The runaway problem: the Abraham-Lorentz equation admits a solution $v(t)=v_0\exp (t/{\tau }_e)$ for $F_{\mathrm{ext}}=0$, where ${\tau }_e={\mu }_0{q}^2/(6\pi mc)\approx 6\times 10^{-24}\mathrm{s}$ for an electron. This unphysical exponentially growing self-acceleration exists because the equation is third-order in position. Standard remedies impose the constraint $a\to 0$ at $t\to +\infty$, selecting the unique physical solution; the resulting integral form features acausal pre-acceleration over times $\sim {\tau }_e$. The conceptual difficulty signalled by these pathologies is resolved in QED: the self-energy of a point charge is divergent classically but renormalisable in quantum field theory, and the finite radiation-reaction effects emerge as a perturbative loop correction to the Dirac equation that does not produce runaway solutions. The Abraham-Lorentz formula remains useful as the leading semiclassical approximation in laser-plasma physics, where modern petawatt-class lasers approach intensities at which classical radiation-reaction effects become observable in electron-beam scattering experiments (CALA Garching, Rutherford Appleton Laboratory Astra-Gemini, ELI Beamlines).

Theorem 5 (radiation pressure measurements: Lebedev 1900; Nichols-Hull 1903). Maxwell's prediction $P_{\mathrm{rad}}=|S|/c$ for an absorber and $2|S|/c$ for a perfect reflector was confirmed experimentally to within $20\%$ by Lebedev in 1900 (Moscow torsion balance with thin mica vanes), and independently by Nichols and Hull in 1903 (refined balance with control for radiometer-effect convection currents).

Lebedev's apparatus used a fine quartz fibre suspension with a torsion period of $\sim 30\mathrm{s}$, illuminated by an arc lamp focused on thin platinum vanes; he reported the predicted micropascal-level forces after subtracting thermal radiometer effects with elaborate gas-pressure controls. Nichols and Hull's Phys. Rev. paper was the first to push the precision below $10\%$, using a Crookes-style apparatus with thin silvered vanes mounted on a horizontal torsion fibre in a partial vacuum at $\sim 1\mathrm{mPa}$, where radiometer (gas-driven thermal-creep) forces are suppressed but residual gas absorption is still adequate to provide damping. Both groups confirmed the central prediction that light pressure is a real, measurable effect — closing the 27-year gap between Maxwell's 1873 prediction and its experimental verification.

The Lebedev-Nichols-Hull measurements completed a conceptual loop. Maxwell's prediction was a consequence of the field's stress tensor: an electromagnetic wave's momentum-flux component $T_{ii}$ along the propagation direction is exactly the energy density $u$, and absorbing the wave deposits this flux as force per unit area on the absorber. Photon-recoil-momentum measurements in atomic spectroscopy half a century later (Frisch 1933 Z. Phys. 86, 42, observing atomic-beam deflection from absorbed sodium D-line photons) recovered the same conclusion at the single-quantum level, confirming that the classical Maxwell stress-tensor prediction holds atom-by-atom and photon-by-photon when reinterpreted quantum-mechanically.

Theorem 6 (optical tweezers and the gradient force, Ashkin et al. 1986). Focused laser light exerts a gradient force on a small dielectric particle proportional to $\nabla |E{|}^2$, drawing the particle toward the focus of the beam.

The force decomposes into a scattering component (along the beam, $\propto |S|$ from absorption/Rayleigh scattering cross section) and a gradient component ($\propto \alpha \nabla |E{|}^2$ for polarizability $\alpha$, drawing the particle toward maximum intensity). For sub-wavelength dielectric beads with $n>n_{\mathrm{medium}}$, the gradient force is stable and three-dimensional, enabling single-cell manipulation and single-molecule force spectroscopy. Ashkin's 1986 Opt. Lett. paper trapped 25-nm dielectric particles using a single objective-focused beam; subsequent developments include the optical-stretcher (Guck et al. 2001), DNA-stretching experiments (Smith-Cui-Bustamante 1996), and atomic optical lattices (Bloch et al. 2002). Ashkin shared the 2018 Nobel Prize in Physics for the technique. Modern dual-trap setups achieve sub-piconewton force resolution.

The mathematical statement: a dielectric particle of polarizability $\alpha$ in an inhomogeneous electric field $E(r)$ feels a gradient force $F_{\mathrm{grad}}=(\alpha /2)\nabla |E{|}^2$. In a focused Gaussian laser beam, $|E{|}^2$ peaks at the focus, so the particle is pulled in three dimensions toward the focus along the gradient. The trap stiffness — the linear restoring-force coefficient near the focus — depends on the beam waist, the wavelength, and the dielectric contrast $(n_p-n_m)/n_m$. For a typical biophysics setup (1064-nm Nd laser, 1.2-NA water-immersion objective, polystyrene beads of $\sim 1\mu \mathrm{m}$ in water), trap stiffness $\kappa \approx 0.1pN/nm$ is achievable, allowing single-molecule force resolution at sub-piconewton level over millisecond timescales. The same machinery scales down to single-atom traps and up to micromirrors suspended on optical fields in optomechanics experiments probing the quantum-classical boundary.

Theorem 7 (electromagnetic angular momentum). The angular momentum density of the electromagnetic field is $\ell =r\times g={ϵ}_{0}r\times (E\times B)$. For a configuration of stationary sources the total field angular momentum $L_{\mathrm{EM}}=\int \ell d^{3}r$ is conserved together with the mechanical angular momentum of the sources, with substantive transfer between the two when the field configuration changes.

The Aharonov-Bohm phase shift acquired by a charged particle encircling a solenoid is, in one viewpoint, the geometric record of the canonical EM angular momentum coupling through the vector potential $A$ even in the absence of any local field acting on the particle. Coleman-Van Vleck identities relate the Aharonov-Bohm geometric phase to integrated angular momentum; the same machinery underlies the modern theory of Berry phases in quantum mechanics (Berry 1984 Proc. R. Soc. A 392, 45) and the topological pumping of charge in periodic insulators (Thouless 1983 Phys. Rev. B 27, 6083). Optical orbital-angular-momentum-carrying beams (Allen-Beijersbergen-Spreeuw-Woerdman 1992 Phys. Rev. A 45, 8185) deliver discrete units of angular momentum per photon, $\hslash \ell$ for the $\ell$-th Laguerre-Gaussian mode, observed experimentally in mechanical rotation of absorbing micrometre-scale particles in focused-laser traps.

Theorem 8 (Eddington luminosity limit). For a spherical accreting source of mass $M$ radiating isotropically with luminosity $L$, the maximum radiative luminosity at which gravitational attraction on infalling hydrogen plasma balances outward radiation pressure (via Thomson scattering on electrons, which drag their proton partners by Coulomb coupling) is

$$L_{\mathrm{Edd}}=\frac{4\pi GM{m}_{p}c}{{\sigma }_T},$$

where ${\sigma }_T=(8\pi /3)r_e^2=6.65\times 10^{-29}{\mathrm{m}}^2$ is the Thomson scattering cross-section and $m_p$ is the proton mass. Above $L_{\mathrm{Edd}}$, radiation pressure overwhelms gravity and matter is blown outward.

The derivation matches force balances: gravitational force per electron-proton pair is $GM{m}_p/r^2$; radiation force per electron is ${\sigma }_{T}L/(4\pi r^{2}c)$ (intensity $L/4\pi r^2$ times Thomson cross-section divided by $c$ for momentum flux). Setting them equal gives the threshold $L_{\mathrm{Edd}}/L_{\odot }\approx 3.2\times 10^4(M/M_{\odot })$. Massive stars (O-type, $M\gtrsim 20M_{\odot }$) approach Eddington; supermassive black holes accreting near Eddington produce the brightest persistent astrophysical sources (quasars at $L\sim 10^{47}erg/s$). The Eddington argument was introduced by Eddington 1916 Monthly Not. R. Astron. Soc. 77, 16 in the first stellar-structure treatment incorporating radiation pressure, and remains the canonical upper bound on steady accretion luminosity in modern observational astrophysics — for example, in interpreting AGN luminosity functions and the slim-disk transition in stellar-mass black-hole binaries above their Eddington rate.

Theorem 9 (Abraham-Minkowski momentum-of-light controversy; resolution by Barnett 2010 PRL 104, 070401). For light in a medium of refractive index $n$, two distinct definitions of EM momentum density coexist:

$$g_{\mathrm{Abraham}}=\frac{E\times H}{c^2},g_{\mathrm{Minkowski}}=D\times B.$$

Barnett 2010 showed that the Abraham momentum corresponds to the field's kinetic momentum (governing centre-of-mass motion) and the Minkowski momentum corresponds to its canonical momentum (governing wavelength and the de Broglie relation $\lambda =h/p$).

The century-long controversy began with Minkowski 1908 and Abraham 1909, with each definition supported by different experiments. Photon recoil in dilute atomic vapours measures the canonical Minkowski momentum; the Jones-Richards 1954 experiment (Nature 174, 707) on radiation pressure in a liquid measured the kinetic Abraham momentum. Barnett's PRL paper, building on earlier work by Brevik, Garrison, and Loudon, identified both as valid in their respective regimes (kinetic vs canonical), resolving the apparent paradox. The lesson — that "momentum" in a medium has multiple inequivalent definitions corresponding to different operational measurements — extends to phonon-photon coupling in optomechanics and to the photon-drag effect in semiconductors.

Synthesis. The Poynting vector and Maxwell stress tensor are the foundational reason that the electromagnetic field is itself a dynamical entity, on the same footing as matter. The central insight is that local conservation of energy ($\partial u/\partial t+\nabla \cdot S=-J\cdot E$) and momentum ($\partial g_i/\partial t+{\partial }_j{T}_{ji}=-[\rho E+J\times B{]}_i$) together compose into the divergence identity ${\partial }_{\mu }{T}_{\mathrm{EM}}^{\mu \nu }=-F^{\nu \alpha }{J}_{\alpha }$, identifying the field's stress-energy as the bookkeeping that closes Newton's third law for the field-matter system.

Putting these together with the Belinfante symmetrisation, the bridge is between two viewpoints on the electromagnetic stress tensor: the canonical Noether tensor from $\partial \mathcal{L}/\partial ({\partial }_{\mu }{A}_{\nu })$ and the symmetric metric-variation tensor from $\delta (\sqrt{-g}\mathcal{L})/\delta g_{\mu \nu }$. This is exactly the structure that generalises to general relativity, where $T_{\mathrm{EM}}^{\mu \nu }$ sources spacetime curvature through Einstein's equations $G^{\mu \nu }=(8\pi G/c^4)T^{\mu \nu }$, identifies the Reissner-Nordström charged black hole as the gravitational image of a point charge, and builds toward 13.04.01 as the canonical worked example of an Einstein-Maxwell coupled system.

The pattern recurs: radiation pressure measured by Lebedev in 1900 is dual to the photon-recoil cooling of atoms in modern laser-cooling traps (Chu-Cohen-Tannoudji-Phillips 1997 Nobel); the Maxwell-Faraday tension along field lines appears again in magnetohydrodynamic flux freezing and astrophysical jet collimation; the Abraham-Minkowski controversy identifies the conceptual difference between kinetic and canonical momentum that recurs in optomechanics, in photonic crystals, and in the quantum theory of light-matter interaction. Modern applications — solar sails (IKAROS 2010, LightSail-2 2019, Breakthrough Starshot 2016 concept), optical tweezers (Ashkin 1986, Nobel 2018), and laser cooling of atoms (Nobel 1997) — all run through the same Maxwell-Poynting machinery developed in the 1873-1908 period.

The conformal-invariance trace identity $T^{\mu }{}_{\mu }=0$ is the foundational reason that black-body radiation in four spacetime dimensions has a strictly traceless stress-energy tensor at the classical level; the quantum trace anomaly, whose breakdown of conformal invariance is computed via heat-kernel methods in 03.09.05 (eta invariant) and appears again in the gravitational anomaly of Hawking radiation, identifies the bridge between classical EM stress-energy and the quantum field-theory framework. The Poynting theorem in its covariant form is the prototype conservation law for any classical gauge field; replacing $A_{\mu }$ by a non-Abelian connection $A_{\mu }^a$ generalises to the Yang-Mills stress-energy tensor that underlies modern QCD/electroweak physics. The pattern of energy density + flux + stress, organised into a symmetric rank-2 conserved tensor, recurs in fluid mechanics (Navier-Stokes momentum equation), in elasticity (Cauchy stress tensor), in continuum dislocation theory (Eshelby stress tensor), and in quantum field theory (vacuum-state stress-energy as the cosmological constant). Every physical theory with a continuous translation symmetry produces a Noether-derived stress-energy tensor, and the electromagnetic case worked out in this unit is the historically first and conceptually clearest example.

Full proof set Master

Proposition 1 (energy conservation in source-free vacuum). For Maxwell's equations in vacuum with no free charges or currents, the energy density $u$ and Poynting vector $S$ satisfy

$$\frac{\partial u}{\partial t}+\nabla \cdot S=0.$$

Proof. Setting $J=0$ in Poynting's theorem $\partial u/\partial t+\nabla \cdot S=-J\cdot E$ gives the source-free conservation form directly. Integrating over all space and using the divergence theorem with the assumption that $E,B$ decay sufficiently fast at infinity (any localised initial-data wave packet has compact spatial support that propagates outward at finite speed, so $E,B\to 0$ at the spatial boundary of a sufficiently large region) yields $d{U}_{\mathrm{total}}/dt=0$, conservation of total field energy. $■$

Proposition 2 (force from Maxwell stress tensor in static field). The force per unit area transmitted across a planar surface in a static electric field with normal component $E_n$ and tangential magnitude $E_t$ is

$$T_{nn}=\frac{{ϵ}_0}{2}(E_n^2-E_t^2),$$

the difference between a tension ${ϵ}_0{E}_n^2/2$ along the field lines (when they cross the surface) and a pressure ${ϵ}_0{E}_t^2/2$ perpendicular to the lines.

Proof. Choose local coordinates so that the surface normal is along $\hat{x}$. Then $T_{xx}={ϵ}_0(E_x^2-|E{|}^2/2)$ with $|E{|}^2=E_x^2+E_y^2+E_z^2=E_n^2+E_t^2$. Direct substitution: $T_{xx}={ϵ}_0(E_n^2-(E_n^2+E_t^2)/2)={ϵ}_0(E_n^2-E_t^2)/2$. The first term ($+{ϵ}_0{E}_n^2/2$) acts as outward tension along field lines crossing the surface; the second term ($-{ϵ}_0{E}_t^2/2$) is a transverse pressure squeezing parallel field lines together. This formalises Faraday's "lines of force have tension along them and pressure between them" picture in modern stress-tensor language. $■$

Proposition 3 (electromagnetic momentum density equals $S/c^2$ in vacuum). The momentum per unit volume of an electromagnetic field in vacuum is $g={ϵ}_0(E\times B)=S/c^2$.

Proof. The relativistic identification follows from the four-tensor structure: $T^{0i}/c=({ϵ}_{0}c)(E\times B{)}^i=(1/({\mu }_{0}c))(E\times B{)}^i=S^i/c$. Since $T^{0i}$ is the momentum density times $c$ in the $(+,-,-,-)$ convention (equivalent: energy flux density divided by $c$), the momentum density is $T^{0i}/c=S^i/c^2={ϵ}_0(E\times B{)}^i$.

Non-covariantly, one can derive the same identification from the force balance on a region: by Poynting's theorem and momentum conservation, the only consistent local momentum density that yields the Lorentz force as the divergence of the stress tensor minus the time derivative of $g$ is $g={ϵ}_0(E\times B)=S/c^2$. The cross-check is the Einstein box gedanken: a photon of energy $E$ traversing a box of length $L$ deposits its momentum $E/c$ on the absorbing wall; total system momentum must be conserved, so the field carried momentum $E/c$ over distance $L$, giving momentum density $E/(L\cdot c)\cdot V$ where $V$ is the box volume — equivalent to the local relation $g=S/c^2$ when the box is large. $■$

Proposition 4 (covariant divergence identity). In Minkowski spacetime with metric ${\eta }_{\mu \nu }=\mathrm{diag}(+1,-1,-1,-1)$, the symmetric electromagnetic stress-energy tensor

$$T_{\mathrm{EM}}^{\mu \nu }=\frac{1}{{\mu }_0}\left[F^{\mu \alpha }{F}^{\nu }{}_{\alpha }-\frac{1}{4}{\eta }^{\mu \nu }{F}_{\alpha \beta }{F}^{\alpha \beta }\right]$$

satisfies ${\partial }_{\mu }{T}_{\mathrm{EM}}^{\mu \nu }=-F^{\nu \alpha }{J}_{\alpha }$, where $J^{\alpha }=(\rho c,J)$ is the four-current.

Proof. Compute ${\partial }_{\mu }{T}^{\mu \nu }$ directly. The first term:

$$\frac{1}{{\mu }_0}{\partial }_{\mu }(F^{\mu \alpha }{F}^{\nu }{}_{\alpha })=\frac{1}{{\mu }_0}({\partial }_{\mu }{F}^{\mu \alpha })F^{\nu }{}_{\alpha }+\frac{1}{{\mu }_0}{F}^{\mu \alpha }{\partial }_{\mu }{F}^{\nu }{}_{\alpha }.$$

The inhomogeneous Maxwell equation in covariant form is ${\partial }_{\mu }{F}^{\mu \alpha }={\mu }_0{J}^{\alpha }$, so the first piece becomes $J^{\alpha }{F}^{\nu }{}_{\alpha }=-F^{\nu }{}_{\alpha }{J}^{\alpha }$. The second piece: relabel and use ${\partial }_{\mu }{F}^{\nu }{}_{\alpha }=-{\partial }^{\nu }{F}_{\mu \alpha }-{\partial }_{\alpha }{F}^{\nu }{}_{\mu }/?$ — actually, use the Bianchi identity ${\partial }_{\mu }{F}_{\nu \rho }+{\partial }_{\rho }{F}_{\mu \nu }+{\partial }_{\nu }{F}_{\rho \mu }=0$ (the homogeneous Maxwell equation ${\partial }_{[\mu }{F}_{\nu \rho ]}=0$) to rewrite ${\partial }_{\mu }{F}^{\nu }{}_{\alpha }$ in terms of ${\partial }^{\nu }{F}_{\mu \alpha }+{\partial }_{\alpha }{F}^{\nu }{}_{\mu }$.

Substituting and using antisymmetry of $F^{\mu \alpha }$ together with the Bianchi-derived identity, the second piece combines with the second term of $T^{\mu \nu }$ (the ${\eta }^{\mu \nu }{F}^2$ trace term):

$$-\frac{1}{4{\mu }_0}{\partial }^{\nu }(F_{\alpha \beta }{F}^{\alpha \beta })=-\frac{1}{2{\mu }_0}{F}^{\alpha \beta }{\partial }^{\nu }{F}_{\alpha \beta }.$$

After careful index manipulation using the Bianchi identity, the $F^{\mu \alpha }{\partial }_{\mu }{F}^{\nu }{}_{\alpha }$ contribution from the first term and the trace-term contribution cancel exactly, leaving only $-F^{\nu \alpha }{J}_{\alpha }$. This is the relativistic statement of energy-momentum balance: the divergence of the field's stress-energy equals minus the four-Lorentz-force on the current, identifying the source of field momentum loss with the Lorentz force pumping momentum into the matter. $■$

Proposition 5 (Larmor formula from radiation reaction). A non-relativistic charged particle of charge $q$ undergoing acceleration $a$ radiates energy at the rate $P={\mu }_0{q}^2|a{|}^2/(6\pi c)$.

Proof. The far-field radiation pattern of an accelerating charge is the standard Liénard-Wiechert result: at distance $r$ in direction $\hat{n}$, the radiated electric field has magnitude

$$E_{\mathrm{rad}}(r,\hat{n},t)=\frac{q}{4\pi {ϵ}_0{c}^{2}r}|a_{\perp }|,$$

where $a_{\perp }$ is the component of acceleration perpendicular to $\hat{n}$ (evaluated at retarded time). The corresponding Poynting flux is

$$|S_{\mathrm{rad}}|=\frac{|E_{\mathrm{rad}}{|}^2}{{\mu }_{0}c}=\frac{q^2|a{|}^2{\sin }^2\theta }{16{\pi }^2{ϵ}_0{c}^3{r}^2},$$

where $\theta$ is the angle between $\hat{n}$ and $a$. Total radiated power: integrate $|S_{\mathrm{rad}}|$ over a sphere of radius $r$, using $\int {\sin }^2\theta d\Omega =8\pi /3$:

$$P=\frac{q^2|a{|}^2}{16{\pi }^2{ϵ}_0{c}^3}\cdot \frac{8\pi }{3}=\frac{q^2|a{|}^2}{6\pi {ϵ}_0{c}^3}=\frac{{\mu }_0{q}^2|a{|}^2}{6\pi c},$$

using $1/({ϵ}_0{c}^3)={\mu }_0/c$. This is the Larmor formula. The momentum balance — radiated four-momentum equals minus the Abraham-Lorentz self-force times time — yields Theorem 4 above. $■$

Proposition 6 (momentum-conservation form of Maxwell's equations). The Lorentz force per unit volume on charges is $f_i=(\rho E+J\times B{)}_i$. Using Maxwell's equations, this can be rewritten as

$$f_i={\partial }_j{T}_{ij}-\frac{\partial g_i}{\partial t},$$

where $g_i={ϵ}_0(E\times B{)}_i$ is the field momentum density and $T_{ij}$ is the Maxwell stress tensor.

Proof. Start from the Lorentz force density and substitute Maxwell's equations:

$$f_i=\rho E_i+(J\times B{)}_i={ϵ}_0(\nabla \cdot E)E_i+(J\times B{)}_i,$$

using Gauss's law $\rho ={ϵ}_0\nabla \cdot E$. For the second term, substitute $J=(1/{\mu }_0)\nabla \times B-{ϵ}_0\partial E/\partial t$ from the Ampère-Maxwell law:

$$(J\times B{)}_i=\frac{1}{{\mu }_0}[(\nabla \times B)\times B{]}_i-{ϵ}_0{\left[\frac{\partial E}{\partial t}\times B\right]}_i.$$

Use the identity $(\nabla \times B)\times B=(B\cdot \nabla )B-\nabla (|B{|}^2/2)$ — equivalently, $[(\nabla \times B)\times B{]}_i=(B_j{\partial }_j)B_i-{\partial }_i(|B{|}^2/2)$. Add and subtract the term $(\nabla \cdot B)B_i=0$ (since $\nabla \cdot B=0$ identically) to make the magnetic contribution take the divergence form ${\partial }_j[B_i{B}_j-{\delta }_{ij}|B{|}^2/2]$.

For the electric contribution, by the same identity $(\nabla \cdot E)E_i+((\nabla \times E)\times E{)}_i={\partial }_j[E_i{E}_j-{\delta }_{ij}|E{|}^2/2]$ (using $(\nabla \times E)\times E=(E\cdot \nabla )E-\nabla (|E{|}^2/2)$ similarly). Add and subtract using Faraday's law $\nabla \times E=-\partial B/\partial t$:

$${ϵ}_0(\nabla \cdot E)E_i={ϵ}_0{\partial }_j(E_i{E}_j-{\delta }_{ij}|E{|}^2/2)-{ϵ}_0((\nabla \times E)\times E{)}_i={ϵ}_0{\partial }_j(E_i{E}_j-{\delta }_{ij}|E{|}^2/2)+{ϵ}_0{\left[\frac{\partial B}{\partial t}\times E\right]}_i.$$

Combining all pieces:

$$f_i={\partial }_j\left[{ϵ}_0(E_i{E}_j-{\delta }_{ij}|E{|}^2/2)+\frac{1}{{\mu }_0}(B_i{B}_j-{\delta }_{ij}|B{|}^2/2)\right]-{ϵ}_0{\left[\frac{\partial E}{\partial t}\times B+E\times \frac{\partial B}{\partial t}\right]}_i.$$

The two cross-product time derivatives combine into ${\partial }_t(E\times B{)}_i={\partial }_t(g_i/{ϵ}_0)$, so the bracket becomes $\partial g_i/\partial t$. The square-bracketed divergence is exactly the Maxwell stress tensor $T_{ij}$. Therefore $f_i={\partial }_j{T}_{ij}-\partial g_i/\partial t$.

Integrating over a region $V$ and applying the divergence theorem gives the force on charges in $V$:

$$F_i={\int }_V{f}_i{d}^{3}r={\oint }_{\partial V}{T}_{ij}d{A}_j-\frac{d}{dt}{\int }_V{g}_i{d}^{3}r,$$

identifying the momentum delivered to the charges as the flux of momentum into $V$ across the boundary minus the rate of change of field momentum stored in $V$. This is the local form of momentum conservation for the combined field-plus-charge system. $■$

Connections Master

• Faraday's law and EM induction 10.03.01. Supplies the time-varying-field framework that makes Poynting's theorem dynamical: the derivation uses $\nabla \times E=-\partial B/\partial t$ to convert $E\cdot (\nabla \times B)$ into the energy-flow form. Without Faraday's law there is no way to write the energy-density time derivative as $\partial u/\partial t$; the two laws are complementary parts of the conservation identity.

• Maxwell in differential forms 10.04.01. Provides the covariant geometric setting for the four-tensor $T_{\mathrm{EM}}^{\mu \nu }$: in differential-forms language, $T^{\mu \nu }$ arises as the symmetric tensor associated to the Faraday two-form $F=dA$ via the Hodge dual and the metric, and the conservation identity ${\partial }_{\mu }{T}^{\mu \nu }=-F^{\nu \alpha }{J}_{\alpha }$ is the covariant divergence of the stress-energy form.

• Conductors, capacitance, electrostatic energy 10.01.03. The static limit of Poynting's theorem reduces to the electrostatic energy formula $U=({ϵ}_0/2)\int |E{|}^2{d}^{3}r$, and the attractive pressure ${ϵ}_0{E}^2/2$ between capacitor plates derived there is exactly the relevant diagonal component of the Maxwell stress tensor in the static case.

• EM stress-energy tensor 10.06.02 pending (pending). The full covariant generalisation: $T_{\mathrm{EM}}^{\mu \nu }$ unifies energy density, momentum density, and stress into a single rank-2 tensor whose divergence is the four-Lorentz-force on charges. This unit provides the non-covariant pieces; the covariant assembly is the subject of the pending unit and connects to the gravitational source term in the next bullet.

• Einstein field equations 13.04.01. $T_{\mathrm{EM}}^{\mu \nu }$ appears on the right-hand side of Einstein's equations as a source of gravitational curvature; charged black holes (Reissner-Nordström) source spacetime curvature exactly through the EM stress-energy derived here. The conformal invariance ($T^{\mu }{}_{\mu }=0$) plays a role in the trace anomaly of black-hole thermodynamics.

Historical & philosophical context Master

Maxwell's Treatise on Electricity and Magnetism (1873) ^{[Maxwell1873]} Volume 2 Chapter XVIII introduced the stress in the electromagnetic field in the form he attributed to Faraday's lines-of-force picture: tension along field lines, pressure between them. The same volume's Chapter XX predicted radiation pressure, deriving $P=u$ (energy density equals pressure for a propagating wave) and noting that sunlight should push absorbing surfaces with a measurable force. Maxwell estimated the magnitude of solar radiation pressure at Earth's orbit as approximately one micropascal — beyond the precision of mid-Victorian gravimetry but in principle measurable with sufficiently sensitive torsion-balance instruments.

The Poynting vector itself was introduced in Poynting's 1884 paper ^{[Poynting1884]} in the Phil. Trans. R. Soc. — "On the transfer of energy in the electromagnetic field" — where Poynting derived the conservation theorem from Maxwell's equations and applied it to charging capacitors and current-carrying wires. Heaviside independently introduced the same vector around the same period (sometimes called the Poynting-Heaviside vector); his contributions appear scattered through his 1893-1912 Electromagnetic Theory volumes and were less widely cited than Poynting's because of Heaviside's eccentric operational-calculus notation. The conceptual content is identical.

Lebedev's experimental confirmation in 1900 ^{[Lebedev1901]} (Annalen der Physik 6, 433-458) used a torsion balance with platinum-vane targets in a partial vacuum, achieving micropascal-level sensitivity by integrating over a long quartz fibre suspension. Nichols and Hull repeated the experiment independently in 1903 ^{[NicholsHull1903]} in Phys. Rev., with improved gas-pressure controls eliminating the radiometer-effect convection bias. Their concordance closed the experimental gap between Maxwell's 1873 prediction and physical verification.

The covariant four-tensor formulation emerged in the wake of Einstein's 1905 special relativity. Abraham's 1903 paper ^{[Abraham1903]} in Annalen der Physik anticipated a four-tensor structure for electrodynamics in a pre-relativistic frame; Minkowski's 1908 lecture notes ^{[Minkowski1908]} in Nachr. Königl. Ges. Wiss. Göttingen — published shortly before his death — gave the modern $T_{\mathrm{EM}}^{\mu \nu }$ in the form used today, embedded in the Minkowski-spacetime four-vector calculus that he had introduced one year earlier. Belinfante's 1940 Physica paper ^{[Belinfante1940]} resolved the asymmetry of the canonical Noether stress tensor by adding a divergence of an antisymmetric tensor, recovering the symmetric Hilbert stress tensor needed for general-relativistic coupling. Rosenfeld's 1940 Mém. Acad. R. Belg. paper independently performed the same construction; the procedure is now standard under the joint name Belinfante-Rosenfeld.

Modern applications run through the same machinery. Ashkin's optical-tweezer paper ^{[AshkinEtal1986]} in Optics Letters 11 (1986) used the gradient force derived from the Maxwell stress tensor in a high-numerical-aperture focus to trap dielectric particles, enabling biophysical force spectroscopy on single molecules; he shared the 2018 Nobel Prize for the technique. The Chu-Cohen-Tannoudji-Phillips 1997 Nobel for laser cooling exploits the photon-recoil momentum derived from Poynting flux to cool atoms below microkelvin temperatures. The Abraham-Minkowski momentum-of-light controversy — distinct definitions of EM momentum density in a refractive medium — was resolved by Barnett 2010 Phys. Rev. Lett. ^{[Barnett2010]} who showed the two definitions correspond to kinetic and canonical momenta and are both physically meaningful in their respective regimes.

Solar-sail propulsion entered demonstrated reality with the Japanese IKAROS spacecraft (2010), the first interplanetary vehicle to use radiation pressure as primary propulsion, deploying a 196-square-metre sail toward Venus; LightSail-2 (Planetary Society, 2019) demonstrated controlled solar-sail flight in Earth orbit using a 32-square-metre sail to raise its apogee through repeated solar-pressure burns. The conceptual Breakthrough Starshot programme (announced 2016) targets laser-driven interstellar gram-scale probes accelerated to twenty percent of light speed by a hundred-gigawatt ground-based phased laser array — all driven by the same Maxwell-Poynting radiation pressure measured for the first time by Lebedev in Moscow in 1900. Spacecraft precision orbit determination (GPS satellites, Cassini at Saturn, the LISA Pathfinder drag-free testbed) requires accounting for solar-radiation-pressure forces at the micronewton level alongside gravitational and atmospheric perturbations. The same machinery underlies astrophysical processes ranging from radiation-pressure-driven stellar winds in massive O-stars to the Eddington luminosity limit for accreting black holes, where outward radiation pressure on infalling gas balances inward gravity at $L_{\mathrm{Edd}}=4\pi GM{m}_{p}c/{\sigma }_T$.

Bibliography Master

@article{Maxwell1865,
  author    = {Maxwell, James Clerk},
  title     = {A dynamical theory of the electromagnetic field},
  journal   = {Philosophical Transactions of the Royal Society of London},
  volume    = {155},
  year      = {1865},
  pages     = {459--512},
}

@book{Maxwell1873,
  author    = {Maxwell, James Clerk},
  title     = {A Treatise on Electricity and Magnetism},
  publisher = {Clarendon Press, Oxford},
  year      = {1873},
}

@article{Poynting1884,
  author    = {Poynting, John Henry},
  title     = {On the transfer of energy in the electromagnetic field},
  journal   = {Philosophical Transactions of the Royal Society of London},
  volume    = {175},
  year      = {1884},
  pages     = {343--361},
}

@article{Lebedev1901,
  author    = {Lebedev, Pyotr},
  title     = {Untersuchungen {\"u}ber die Druckkr{\"a}fte des Lichtes},
  journal   = {Annalen der Physik},
  volume    = {6},
  year      = {1901},
  pages     = {433--458},
}

@article{NicholsHull1903,
  author    = {Nichols, Ernest Fox and Hull, Gordon Ferrie},
  title     = {The pressure due to radiation},
  journal   = {Physical Review (Series 1)},
  volume    = {17},
  year      = {1903},
  pages     = {26--50},
}

@article{Abraham1903,
  author    = {Abraham, Max},
  title     = {Prinzipien der Dynamik des Elektrons},
  journal   = {Annalen der Physik},
  volume    = {10},
  year      = {1903},
  pages     = {105--179},
}

@article{Minkowski1908,
  author    = {Minkowski, Hermann},
  title     = {Die Grundgleichungen f{\"u}r die electromagnetischen Vorg{\"a}nge in bewegten K{\"o}rpern},
  journal   = {Nachrichten von der K{\"o}niglichen Gesellschaft der Wissenschaften zu G{\"o}ttingen},
  year      = {1908},
  pages     = {53--111},
}

@article{Belinfante1940,
  author    = {Belinfante, Frederik J.},
  title     = {On the spin angular momentum of mesons},
  journal   = {Physica},
  volume    = {7},
  year      = {1940},
  pages     = {449--474},
}

@article{AshkinEtal1986,
  author    = {Ashkin, A. and Dziedzic, J. M. and Bjorkholm, J. E. and Chu, S.},
  title     = {Observation of a single-beam gradient force optical trap for dielectric particles},
  journal   = {Optics Letters},
  volume    = {11},
  year      = {1986},
  pages     = {288--290},
}

@article{Barnett2010,
  author    = {Barnett, Stephen M.},
  title     = {Resolution of the Abraham-Minkowski dilemma},
  journal   = {Physical Review Letters},
  volume    = {104},
  year      = {2010},
  pages     = {070401},
}

@book{Jackson1999,
  author    = {Jackson, John David},
  title     = {Classical Electrodynamics},
  edition   = {3},
  publisher = {Wiley},
  year      = {1999},
}

@book{Griffiths2013,
  author    = {Griffiths, David J.},
  title     = {Introduction to Electrodynamics},
  edition   = {4},
  publisher = {Cambridge University Press},
  year      = {2013},
}

@book{LandauLifshitzVol2,
  author    = {Landau, Lev Davidovich and Lifshitz, Evgeny Mikhailovich},
  title     = {The Classical Theory of Fields},
  series    = {Course of Theoretical Physics},
  volume    = {2},
  edition   = {4},
  publisher = {Pergamon},
  year      = {1980},
}