Electromagnetic Stress, Momentum, and Energy in Media

Everything this chapter has built — energy flow, momentum density, stress, angular momentum — was built in vacuum, where the answers are unique and the theorems are clean. Send the machinery into matter and one of physics' longest-running honest disagreements is waiting. Energy survives the trip intact: Poynting's theorem generalizes to media with $\mathbf{S} = \mathbf{E}\times\mathbf{H}$ and $u = \tfrac12(\mathbf{E}\cdot\mathbf{D} + \mathbf{B}\cdot\mathbf{H})$ , derivable in five lines and disputed by no one. Momentum does not. In 1908 Hermann Minkowski proposed the momentum density $\mathbf{g}_M = \mathbf{D}\times\mathbf{B}$ ; within a year Max Abraham countered with $\mathbf{g}_A = \mathbf{E}\times\mathbf{H}/c^2$ . In a medium of refractive index $n$ they disagree by a factor of $n^2$ — for a single photon entering glass, Minkowski says its momentum rises to $n\hbar\omega/c$ , Abraham says it falls to $\hbar\omega/nc$ . Both derivations are competent. Both have experiments on their side. The argument ran for a century. This article does the controversy the honest way, because the resolution is genuinely instructive — more instructive than either "answer" alone. The total momentum of light-plus-matter is unambiguous and conserved; what Abraham and Minkowski disagree about is how to split the total between "field" and "medium," and the split is a convention until you say which experiment you are doing. The modern settlement (Barnett, 2010): Abraham's is the kinetic momentum — what pushes the medium's center of mass, vindicated by center-of-energy arguments and force measurements; Minkowski's is the canonical momentum — conjugate to position, what wave mechanics and every recoil, diffraction, and Doppler experiment in a medium measures. Two questions, two answers, one consistent physics — and a working rule for which to use when.

Key takeaways

Energy bookkeeping in linear media is undisputed: $\partial_t u + \nabla\cdot\mathbf{S} = -\mathbf{J}_f\cdot\mathbf{E}$ with $\mathbf{S} = \mathbf{E}\times\mathbf{H}$ , $u = \tfrac12(\mathbf{E}\cdot\mathbf{D} + \mathbf{B}\cdot\mathbf{H})$ — derived here, five lines from Maxwell-in-matter.
Momentum splits into rival candidates: Minkowski $\mathbf{g}_M = \mathbf{D}\times\mathbf{B}$ vs Abraham $\mathbf{g}_A = \mathbf{E}\times\mathbf{H}/c^2$ ; per photon in index $n$ : $n\hbar\omega/c$ vs $\hbar\omega/nc$ — a factor $n^2$ apart ( $2.25$ in glass).
Both are "right": the total momentum is unique; the field/matter split is convention-plus-context. Abraham = kinetic momentum (center-of-mass motion, Balazs box argument); Minkowski = canonical momentum (wave mechanics, recoil, diffraction).
Experiments sort accordingly: submerged-mirror pressure and atom recoil measure $\propto n$ (canonical/Minkowski); the Balazs-box logic and time-averaged Abraham-force measurements see the kinetic side.
The matter's co-traveling mechanical momentum (including hidden-momentum-style terms) is the reconciliation fund: count it and every experiment is consistent with one total.
KEY IDEA: in media the total momentum of light plus matter is unambiguous; "Abraham vs Minkowski" is the choice between kinetic and canonical momentum of the light subsystem — ask which your experiment measures, and both centuries of results cohere.

The core insight

In material media the total momentum of field plus matter is unique and conserved, but its split is convention: Abraham's $\mathbf{E}\times\mathbf{H}/c^2$ is the light's kinetic momentum (what moves mass), Minkowski's $\mathbf{D}\times\mathbf{B}$ its canonical momentum (what recoil and interference measure) — a factor $n^2$ apart, both correct for their own question.

Where this lives today

The kinetic/canonical sorting is now working infrastructure in several precision fields. Optomechanics of liquids and soft matter — laser-induced surface deformations (Ashkin and Dziedzic's 1973 observation that a laser exiting water pulls the surface outward), optical stretching of cells, radiation-pressure microrheology — is analyzed with total-stress methods where the old controversy would have paralyzed sign choices. Atom optics is canonical momentum engineering: every atom interferometer, Bragg beamsplitter, and photon-recoil measurement of $h/m$ (input to the fine-structure constant) uses $\hbar k_{\text{medium}}$ bookkeeping, with the 2005 BEC experiment as the calibration-grade demonstration. Photon-drag detectors (semiconductors converting light momentum to current) and slow-light media (where $n_{\text{group}}$ can reach $10^7$ , brutally amplifying the splits' difference) are standing testbeds; proposals to weigh light in dispersive media continue to sharpen the fine print. Conceptually, the episode became the canonical (pun long since surrendered) example of a pseudo-controversy resolved by question hygiene — cited in quantum information (kinetic vs canonical in mechanical systems), in the Casimir-physics literature (stress splits in media), and in general relativity (where "gravitating momentum" of light in media must be the total). The 2010 Barnett–Loudon papers function as the settlement documents; disputes since have been about terms in the medium account (electrostriction, dispersion corrections) rather than about the structure — physics arguing about the third decimal of a reconciliation rather than the existence of one.

Origins · 2 minMax Abraham and how this was discovered

Max Abraham

1875–1922 · Göttingen, Milan

Abraham was electron theory's most uncompromising craftsman and its most stubborn casualty. Planck's first doctoral student, he built (1902–03) the era's definitive "rigid electron" — a purely electromagnetic particle whose mass was entirely field energy, with momentum $\mathbf{E}\times\mathbf{H}/4\pi c$ bookkeeping he defended with ferocious rigor — and his Theorie der Elektrizität trained a generation (it remained a standard text for decades). His momentum density for light in media (1909) came from demanding the field's stress–energy tensor be symmetric — the very property this chapter tied to angular-momentum conservation — against Minkowski's asymmetric but relativistically elegant tensor; that both demands were defensible is the controversy's whole anatomy. Abraham never accepted special relativity — his rigid electron was its direct competitor, and he fought Einstein in print (and, on gravitation, in a famous exchange of increasingly sharp papers) to the end — yet he fought so honestly that Born called him relativity's most valuable opponent: his objections had to be answered, and answering them strengthened the theory. Exiled from Italy by the First World War, he died in 1922 of a brain tumor, before the experiments that would put his name on half of physics' most instructive tie. The resolution — his momentum is the kinetic one, forced by the center-of-energy theorem of the relativity he rejected — is an irony he would have appreciated only after checking every tensor index himself.

History

1908

— Minkowski's covariant electrodynamics of media assigns light the momentum density D cross B

1909

— Abraham counters with E cross H over c-squared, demanding a symmetric stress-energy tensor

1953

— Balazs's box argument: center-of-energy bookkeeping favors the kinetic (Abraham) momentum in transit

1954

— Jones and Richards measure submerged-mirror radiation pressure proportional to n (canonical side); Jones-Leslie refine to 0.05% in 1978

1973

— Ashkin and Dziedzic observe laser light pulling a water surface outward — interface optomechanics joins the evidence

1975

— Walker and Lahoz detect the oscillatory Abraham force in a torsion experiment

2005

— Campbell et al.: atom recoil in a BEC equals n h-bar k — canonical momentum, interferometrically clean

2010

— Barnett resolves the dilemma: Abraham = kinetic, Minkowski = canonical; totals unique, century closed

The physical picture

Why does matter break the uniqueness that vacuum enjoyed? Because in matter, "the field's momentum" stops being a well-posed phrase. A light pulse in glass is not a field configuration — it is a polariton: an inseparable choreography of electromagnetic field and oscillating bound charges, each carrying energy and momentum, exchanging both continuously at the wave's own frequency. Any statement of the form "the light carries momentum $p$ " secretly draws a boundary through the middle of that choreography — how much of the swaying medium do you count as "light"? Draw it one way, you get Abraham; another, Minkowski; the total — field plus every last swaying dipole plus the slab's bulk recoil — is the same for everyone, conserved by the vacuum theorem applied to the complete system (microscopic fields plus charges, where this chapter's results hold without amendment).

Figure

One photon enters glass: the two honest ledgers — kinetic momentum falls, canonical momentum rises, and the slab bookkeeps the difference

Hold two concrete experiments in mind, because each "side" owns one. Experiment K (kinetic): shine a pulse through a free-floating glass slab and watch the slab's position. The relativistic center-of-energy theorem — the hidden-momentum article's stricter ledger, valid for the total system — forces a specific answer for how the slab shifts, and working it backward (Derivation B) assigns the light in transit momentum $\hbar\omega/nc$ : Abraham. Light in glass pushes less mass-motion per joule than in vacuum, and the slab inches forward while the pulse is inside, exactly compensating. Experiment C (canonical): let an atom in a Bose–Einstein condensate absorb a photon from a laser passing through the condensate (index $n$ ), and measure the atom's recoil interferometrically. Measured answer (Campbell et al., 2005): recoil $= n\hbar\omega/c$ — Minkowski, larger than vacuum, because the photon's wavelength in the medium is shorter ( $\lambda/n$ ) and recoil in wave mechanics is $\hbar k = h/\lambda_{\text{medium}}$ . Same medium, same light; two well-posed questions, two different momenta — and no contradiction, because the medium's own mechanical account differs between the two protocols (in the recoil experiment, the surrounding fluid exchanges momentum with the atom's neighborhood; in the slab experiment, the whole block's motion is the observable). That is the honest shape of the resolution before any formalism: "the momentum of light in a medium" is under-specified until you say what you will measure. Kinetic questions (does mass move?) get Abraham's answer; canonical questions (what does the wave's phase gradient impart per quantum?) get Minkowski's. The century of controversy was, in retrospect, a century of experiments correctly answering differently-posed questions — plus a bookkeeping fog around the medium's co-traveling momentum (including relativistic terms of exactly the hidden-momentum family) that the 2010 settlement finally organized.

The undisputed part: energy in media

Before the contested ledger, the clean one — it is also the template for how media versions of vacuum theorems are made. The macroscopic Maxwell equations (E6, Maxwell's Equations in Matter) carry the free sources and the auxiliary fields; mirror the vacuum Poynting derivation with them (the vacuum theorem itself belongs to the sibling article Poynting's Theorem, recalled by name):

-\frac{\partial u}{\partial t} - \nabla\cdot\mathbf{S} = \mathbf{J}_f\cdot\mathbf{E}, \qquad \mathbf{S} = \mathbf{E}\times\mathbf{H}, \qquad u = \tfrac12\left(\mathbf{E}\cdot\mathbf{D} + \mathbf{B}\cdot\mathbf{H}\right), \qquad (1)

where: $\mathbf{J}_f$ — free current only (the bound charges' work is inside $u$ , which is the point); validity — linear, non-dispersive, lossless media, the fine print Derivation A makes explicit. Physically: $u$ counts field energy plus the energy parked in stretched polarization and aligned magnetization (the E3/E5 energy articles' storage terms, absorbed into one quadratic), and $\mathbf{E}\times\mathbf{H}$ — not $\mathbf{E}\times\mathbf{B}/\mu_0$ — is the flux that makes the books balance at material interfaces (its tangential components are continuous where fields cross boundaries, so energy does not spuriously pile up at surfaces). In strongly dispersive media $u$ generalizes (Brillouin's $\tfrac{d(\omega\varepsilon)}{d\omega}$ form) and in lossy media part of the budget drains irreversibly to heat — both owned downstream by E8's Absorption and Dispersion, deferred by name.

The contested part: two momenta

The candidates, with their pedigrees:

\mathbf{g}_M = \mathbf{D}\times\mathbf{B} \qquad\text{(Minkowski, 1908)}, \qquad \mathbf{g}_A = \frac{\mathbf{E}\times\mathbf{H}}{c^2} \qquad\text{(Abraham, 1909)}. \qquad (2)

where: in vacuum both collapse to $\varepsilon_0\mathbf{E}\times\mathbf{B}$ — the dispute is invisible until matter appears. For a plane wave in a linear medium of index $n$ (non-magnetic, non-dispersive), the ratio of their magnitudes is exactly $n^2$ : per photon of energy $\hbar\omega$ ,

p_M = \frac{n\hbar\omega}{c}, \qquad p_A = \frac{\hbar\omega}{nc}. \qquad (3)

where: the two scalings are memorable through their native pictures — Minkowski rides the wavevector ( $k = n\omega/c$ grows in the medium: canonical $\hbar k$ ), Abraham rides the speed (energy transport at $c/n$ slows: kinetic $\sim$ energy × velocity / $c^2$ , the transport identity of the momentum article with $v = c/n$ ). Each candidate drags a stress tensor and a force density along with it (their difference, $\mathbf{f}_A - \mathbf{f}_M \supset \frac{n^2-1}{c^2}\partial_t(\mathbf{E}\times\mathbf{H})$ , is the Abraham force — oscillatory, tiny, and measured in low-frequency torsion experiments by Walker and Lahoz in 1975); Minkowski's tensor is asymmetric, which historically counted against it (angular momentum bookkeeping — the previous articles' symmetry requirement) until the canonical reading explained why it may be: canonical momentum includes medium co-motion whose angular bookkeeping sits elsewhere. The settlement, stated as the theorem it now is (Barnett 2010, closing a literature stretching from von Laue to Gordon to Nelson): the total momentum is unique; Abraham's is the kinetic momentum of the light subsystem, Minkowski's is its canonical momentum, and the difference is momentum carried by the medium's response —

\mathbf{p}_{\text{total}} = \mathbf{p}_A + \mathbf{p}_{\text{med,kinetic}} = \mathbf{p}_M + \mathbf{p}_{\text{med,canonical-complement}}, \qquad (4)

where: the medium terms include the bulk recoil of Derivation B and, in configurations with crossed static fields, contributions of exactly the hidden-momentum family (several claimed "experimental discriminations" between the rivals dissolved when that account was included — the previous article's warning, cashing out). Equation (4) is the media edition of this chapter's constant refrain: totals are physics, splits are bookkeeping, and a split becomes physics only when an experiment's question selects it.

Derivation APoynting's theorem in linear media — the clean ledger, with its fine print

The goal is equation (1): the energy continuity law for macroscopic electrodynamics, identifying the flux $\mathbf{E}\times\mathbf{H}$ and density $\tfrac12(\mathbf{E}\cdot\mathbf{D}+\mathbf{B}\cdot\mathbf{H})$ , with every assumption surfaced. The strategy mirrors the vacuum derivation (sibling article, recalled not repeated): start from the work rate on free charges, substitute the matter-form curl equations, and watch the constitutive relations decide whether the leftover terms form a total time derivative. The fine print lives exactly there — linearity and non-dispersion are what make the ledger close; watch them earn their keep.

Step 1. Work on free charges. The fields pay wages only through $\mathbf{E}$ (magnetic forces do no work): rate per volume $= \mathbf{J}_f\cdot\mathbf{E}$ . Substitute Ampère–Maxwell in matter, $\mathbf{J}_f = \nabla\times\mathbf{H} - \partial_t\mathbf{D}$ :

\mathbf{J}_f\cdot\mathbf{E} = \mathbf{E}\cdot(\nabla\times\mathbf{H}) - \mathbf{E}\cdot\frac{\partial\mathbf{D}}{\partial t}.

Step 2. Symmetrize with the E0 identity $\nabla\cdot(\mathbf{E}\times\mathbf{H}) = \mathbf{H}\cdot(\nabla\times\mathbf{E}) - \mathbf{E}\cdot(\nabla\times\mathbf{H})$ and Faraday $\nabla\times\mathbf{E} = -\partial_t\mathbf{B}$ :

\mathbf{J}_f\cdot\mathbf{E} = -\nabla\cdot(\mathbf{E}\times\mathbf{H}) - \mathbf{H}\cdot\frac{\partial\mathbf{B}}{\partial t} - \mathbf{E}\cdot\frac{\partial\mathbf{D}}{\partial t}.

Why $\mathbf{E}\times\mathbf{H}$ and not some other pairing: the curl equations in matter couple $\mathbf{E}$ with $\mathbf{H}$ (and $\mathbf{B}$ with $\mathbf{D}$ under time derivatives); only this pairing lets the identity absorb both curls with no stray terms. The flux candidate is chosen by the equations, not by taste.

Step 3. The constitutive moment. For linear, non-dispersive media — $\mathbf{D} = \varepsilon\mathbf{E}$ , $\mathbf{B} = \mu\mathbf{H}$ with $\varepsilon$ , $\mu$ real constants in time —

\mathbf{E}\cdot\partial_t\mathbf{D} = \tfrac12\partial_t(\mathbf{E}\cdot\mathbf{D}), \qquad \mathbf{H}\cdot\partial_t\mathbf{B} = \tfrac12\partial_t(\mathbf{B}\cdot\mathbf{H}),

and the two work terms become the time derivative of $u = \tfrac12(\mathbf{E}\cdot\mathbf{D} + \mathbf{B}\cdot\mathbf{H})$ : equation (1) follows. Why the fine print is load-bearing: with dispersion, $\varepsilon = \varepsilon(\omega)$ and $\mathbf{E}\cdot\partial_t\mathbf{D}$ is no longer a total derivative — part of the exchange is energy lent to the medium's oscillators and returned a cycle later (recoverable: Brillouin's $u = \tfrac12\frac{d(\omega\varepsilon)}{d\omega}|E|^2$ time-averaged form), and with loss, part never returns (absorption). The clean $u$ is the zero-frequency-response idealization; E8 owns the general case by name.

Step 4. Read the ledger. $u$ contains field energy plus polarization/magnetization storage — the capacitor-and-inductor energies of E3/E5, unified; $\mathbf{S} = \mathbf{E}\times\mathbf{H}$ is continuous in its normal component across source-free interfaces (tangential $\mathbf{E}$ and $\mathbf{H}$ are continuous — the E6 boundary conditions), so energy flows through material boundaries without fictitious surface deposits. No rival candidates, no controversy: energy's ledger in media is settled physics. The momentum ledger, built by the same moves in the next derivation's shadow, is not — and seeing exactly where the analogy breaks (the stress term's split between field and matter has no canonical choice) is the cleanest way to understand why.

Figure

The Balazs box: the photon exits late, so the slab must have inched forward — center-of-energy bookkeeping forces the kinetic momentum

Go deeper · Derivation B · 2 minThe Balazs box — the kinetic momentum of light in glass, by center-of-energy bookkeeping

Derivation BThe Balazs box — the kinetic momentum of light in glass, by center-of-energy bookkeeping

The goal is the kinetic side's cornerstone: a photon (energy $\hbar\omega$ ) traversing a free transparent slab (mass $M$ , length $L$ , index $n$ ) must carry momentum $\hbar\omega/nc$ as kinetic momentum, with the slab picking up the difference — forced by the center-of-energy theorem, no model of the microphysics required. The strategy is the hidden-momentum article's Derivation A applied as a constraint: the total system is isolated, so its center of energy moves uniformly; the photon is delayed by the slab; something must therefore advance. Watch the logic run entirely on kinematics plus conservation — this is why the kinetic answer is considered theorem-grade.

Step 1. The race. Compare two histories: photon passes beside the slab (pure vacuum, speed $c$ ), or through it (speed $c/n$ inside). Through-transit takes longer by $\Delta t = \dfrac{nL}{c} - \dfrac{L}{c} = \dfrac{(n-1)L}{c}$ : the photon exits behind its vacuum twin by a distance $\Delta z_\gamma = c\,\Delta t = (n-1)L$ .

Step 2. The constraint. The system (photon + slab) is isolated with total momentum $\hbar\omega/c$ (its vacuum value — nothing external acted); the center-of-energy theorem says $\dot{\mathbf{R}}_E$ is the same constant in both histories. The photon's energy-parcel $\hbar\omega$ arriving late means the slab must have moved forward to keep $\mathbf{R}_E$ on schedule:

M c^2\,\Delta z_{\text{slab}} = \hbar\omega\,\Delta z_\gamma \qquad\Longrightarrow\qquad \Delta z_{\text{slab}} = \frac{\hbar\omega}{Mc^2}\,(n-1)L.

The slab advances while the light is inside — a forward push at entry, a backward push at exit, zero net impulse after transit (the slab ends at rest, displaced).

Step 3. Extract the photon's transit momentum. During the crossing (duration $nL/c$ ), the slab moved $\Delta z_{\text{slab}}$ , so it carried momentum $p_{\text{slab}} = M\,\Delta z_{\text{slab}}/(nL/c) = \dfrac{\hbar\omega}{c}\cdot\dfrac{n-1}{n}$ . Momentum conservation assigns the light the remainder:

p_\gamma = \frac{\hbar\omega}{c} - \frac{\hbar\omega}{c}\frac{n-1}{n} = \boxed{\ \frac{\hbar\omega}{nc}\ }

— Abraham. Why this is specifically the KINETIC momentum: every quantity in the argument is a mass-motion observable — displacements of energy, velocity of the center of energy. The derivation never mentions wavelength, phase, or wavevector; it cannot see canonical structure. It answers exactly one question — how does mass move? — and Abraham is that answer.

Step 4. Now let the other experiment ask its question. An absorber embedded in the medium (an atom in the BEC experiment; a submerged mirror) interacts with the wave locally, and wave mechanics pays recoil in $\hbar k$ with $k = n\omega/c$ — the wavelength in the medium is what diffracts, interferes, and Doppler-shifts: recoil $= n\hbar\omega/c$ , Minkowski, measured (Jones–Richards 1954 and Jones–Leslie 1978: submerged-mirror radiation pressure $\propto n$ to 0.05%; Campbell et al. 2005: atom recoil in a BEC at $n\hbar k$ ). No contradiction with Step 3: in these protocols the medium around the absorber exchanges momentum too (the fluid surges, interface stresses act), and the total — absorber plus local medium response — reconciles with the kinetic ledger. Kinetic question → Abraham; canonical question → Minkowski; total → unique. That three-line summary is the resolved controversy.

Limiting cases

Case	Result	Meaning
vacuum, $n \to 1$	$\mathbf{g}_M = \mathbf{g}_A = \varepsilon_0\mathbf{E}\times\mathbf{B}$	the dispute needs matter to exist
dilute medium, $n = 1 + \delta$	rivals differ at order $\delta$	controversy grows continuously from zero
glass, $n = 1.5$	$p_M/p_A = n^2 = 2.25$	not a small bookkeeping nicety
kinetic question (slab displacement)	Abraham: $\hbar\omega/nc$	center-of-energy theorem's answer
canonical question (recoil, diffraction, Doppler)	Minkowski: $n\hbar\omega/c$	wave mechanics' $\hbar k$ answer
total system audit	unique, conserved	microscopic vacuum theorem, always
dispersive/lossy media	$u$ → Brillouin form; energy drains to heat	E8's absorption-dispersion territory, by name
static crossed fields in media	hidden-momentum-family terms enter the split	several "discriminating" experiments dissolved here

Common trapDo not ask which momentum is right — ask which question you are asking

The classic failure mode, a century old and still committed in submitted manuscripts: treating "the momentum of light in glass" as a single well-posed quantity, adopting whichever of (3) a favorite derivation produces, and then declaring the experiments that disagree "anomalous." Every term of that procedure is broken. The quantity is under-specified (kinetic or canonical?); both derivations are correct for their own object; and the experiments disagree with each other only if you insist they measured the same thing — the submerged mirror and the BEC atom measured canonical recoil; the balance of a deflected fiber and the Balazs logic constrain kinetic transport; all are consistent with one conserved total once the medium's co-traveling momentum (bulk recoil, interface stresses, and hidden-momentum-family terms in crossed static fields) is counted. Working rules: for forces on and motion of media — bulk displacement, optical tweezers on droplets, fiber recoil — audit with the total stress tensor or the kinetic (Abraham) split plus explicit medium forces. For quantum optics and wave mechanics in media — recoil, phase matching, Doppler, photon drag — use canonical (Minkowski) $\hbar k$ without guilt. Never mix splits mid-calculation, and never claim an experiment "proves Abraham" or "proves Minkowski" without stating the question it posed; the literature's confusions are almost entirely sentences missing that clause.

Worked examples

1 · One joule into water — every ledger at once (confidence builder)

A $1.0$ J light pulse enters a large tank of water ( $n = 1.33$ ). Tabulate: vacuum momentum, canonical and kinetic momenta inside, and the mirror force if it reflects off a submerged mirror (per watt of continuous power).

Step 1. Vacuum: $p = E/c = 1.0/2.998\times10^8 = \mathbf{3.34\times10^{-9}}$ kg·m/s — the unambiguous total, before entry and after exit.

Step 2. Inside, the two splits: canonical $p_M = np/1 = 1.33\times3.34\times10^{-9} = \mathbf{4.44\times10^{-9}}$ kg·m/s; kinetic $p_A = p/n = \mathbf{2.51\times10^{-9}}$ kg·m/s. The water's mechanical account holds the difference from the total in each bookkeeping — $+0.83\times10^{-9}$ (kinetic split: water surges forward with the pulse) or $-1.10\times10^{-9}$ (canonical split's complement) — same physics, two filing systems.

Step 3. Submerged mirror, canonical question (the wave reverses its $\hbar k$ in the medium): force per watt $= 2n/c = \mathbf{8.9\times10^{-9}}$ N/W, versus $6.7\times10^{-9}$ in air. Jones and Richards (1954), refined by Jones and Leslie (1978), measured exactly this $n$ -proportionality — to $0.05\%$ — on a mirror-bearing torsion balance submerged in various liquids. The medium-enhanced push is real, repeatable, and larger than vacuum: the single most counterintuitive settled fact in this subject.

Step 4. Reconcile with kinetics in one sentence: the extra push on the mirror is supplied through the liquid (pressure and flow fields near the mirror), whose own momentum budget completes the kinetic ledger — total in, total out, unique. A tabulated version of this example is worth memorizing as the controversy in miniature: one pulse, four numbers ( $p$ , $p_M$ , $p_A$ , $2n/c$ ), zero contradictions.

2 · The Balazs displacement and the BEC recoil, in numbers (exam level)

(a) A $1.0$ J pulse crosses a free-floating glass slab, $M = 1.0$ g, $L = 10$ cm, $n = 1.5$ . Find the slab's permanent displacement. (b) A rubidium atom ( $m = 1.44\times10^{-25}$ kg) in a BEC with effective index $n = 1.2$ absorbs one 780 nm photon. Find its recoil velocity, and compare with vacuum.

Step 1. (a) From Derivation B, Step 2: $\Delta z = \dfrac{E}{Mc^2}(n-1)L = \dfrac{1.0}{10^{-3}\times8.99\times10^{16}}\times0.5\times0.10 = \mathbf{5.6\times10^{-16}\ m}$ — half a proton radius. Forward, permanent, and (currently) unmeasurable — the kinetic side's cornerstone is theorem-grade logic, not laboratory-grade signal, which is why the controversy lasted: the clean kinetic experiment is brutally small.

Step 2. (b) Canonical recoil: $v = \dfrac{n\hbar k}{m} = n\times\dfrac{h}{m\lambda} = 1.2\times\dfrac{6.63\times10^{-34}}{1.44\times10^{-25}\times7.8\times10^{-7}} = \mathbf{7.1\times10^{-3}\ m/s}$ , versus $5.9\times10^{-3}$ m/s in vacuum — a 20% enhancement, resolved cleanly by atom interferometry (Campbell et al. 2005, using the condensate itself as the interferometer).

Step 3. Why (b) is measurable and (a) is not: canonical experiments amplify by using wave observables (interference phases accumulate the $n$ -fold $k$ over many cycles), while kinetic experiments fight the $1/Mc^2$ of macroscopic matter. The experimental record of the controversy is largely a record of canonical-question experiments — which is why "Minkowski keeps winning" was the folklore until the questions were sorted.

Step 4. The exam-grade synthesis: (a) and (b) are both correct predictions of one theory. State the question (mass motion vs wave recoil), pick the matching split (Abraham vs Minkowski), keep the medium's account when totalling — three sentences that would have settled a hundred years of argument, which is precisely what makes this chapter's ledger discipline worth the pain.

Going further

References. Griffiths, Introduction to Electrodynamics, 4th ed., §8.4 (media momentum discussion) · Jackson, Classical Electrodynamics, 3rd ed., §6.7–6.8 · Zangwill, Modern Electrodynamics, §17.3, §17.9 · Barnett, Phys. Rev. Lett. 104, 070401 (2010) · Barnett & Loudon, Phil. Trans. R. Soc. A 368, 927 (2010) · Pfeifer, Nieminen, Heckenberg & Rubinsztein-Dunlop, Rev. Mod. Phys. 79, 1197 (2007) Key checks for this article: the constitutive step in equation (1) requires time-constant linear $\varepsilon, \mu$ (the factor $\tfrac12$ comes directly from linearity); the ratio $p_M/p_A = n^2$ holds for any non-dispersive plane wave. Anchor numbers: $p = 3.34\times10^{-9}$ kg·m/s per joule in vacuum; water canonical/kinetic splits $4.44/2.51\times10^{-9}$ kg·m/s; submerged-mirror force $2n/c = 8.9\times10^{-9}$ N/W at $n = 1.33$ ; Balazs slab displacement $5.6\times10^{-16}$ m (1 J, 1 g, 10 cm, $n = 1.5$ ); Rb recoil $7.1\times10^{-3}$ m/s at $n = 1.2$ .

Builds on: Poynting's Theorem · The Maxwell Stress Tensor · Momentum Density of the Electromagnetic Field · Conservation of Linear Momentum in Electrodynamics · Hidden Momentum in Magnetic Systems

Leads to: Newton's Third Law in Electrodynamics · Global vs. Local Conservation Laws

If the total momentum is what's conserved, why not abolish the split and only ever compute totals?

Because subsystem questions are real experimental questions, and answering them requires a split. "What phase does the atom interferometer accumulate?" is a question about the light–atom exchange alone — canonical split, Minkowski, $\hbar k n$ . "Does the droplet move toward or away from the beam focus?" is a mass-motion question — kinetic split plus explicit medium stresses. Insisting on totals-only would be like banning potential energy because only total energy is conserved: correct in principle, paralyzing in practice. The discipline the resolution actually imposes is subtler: declare the split, use it consistently, and remember that the medium's account is part of your system whether or not it is part of your detector. The pseudo-controversy arose not from using splits but from using them silently — each side attaching the unqualified words "the momentum of light" to its own bookkeeping. Physics is full of such convention-plus-context quantities (gauge-dependent potentials, coordinate energy in GR, quark masses at a scale); the Abraham–Minkowski affair is the classic training example precisely because both conventions are so respectable.

Which momentum does a solar sail inside an atmosphere, or a photon-drag detector, actually respond to?

Work the question hygiene. A photon-drag detector (a semiconductor slab where absorbed light drives a measurable current along the beam) responds to the momentum delivered to its charge carriers per absorbed photon — a canonical-type local exchange, and indeed drag voltages track $n\hbar\omega/c$ bookkeeping in the standard analyses (with material corrections). A sail — any macroscopic reflector whose bulk motion you want — is a kinetic question, but with a subtlety: if the sail is immersed (submerged mirror), the surrounding medium participates, and the measured mirror force is the canonical-looking $2nP/c$ (Jones–Leslie) precisely because the liquid's local response delivers momentum to the mirror along with the light; the sail-plus-nearby-fluid system's kinetic budget still reconciles with the total. In dilute air ( $n - 1 \sim 3\times10^{-4}$ ) all splits agree to a part in $10^4$ and nobody budgets the difference. The operational habit: identify what your transducer touches (carriers? bulk mass? interface?), write the momentum exchange for that touching, and let the medium's account absorb the rest. Every published "anomaly" in this field is a transducer misassigned.

Minkowski's tensor is asymmetric — didn't the angular momentum articles prove that's fatal?

They proved it is fatal for the total closed system's tensor — an asymmetric total stress–energy would break angular momentum conservation with the density $\mathbf{r}\times\mathbf{g}$ , as the field-angular-momentum article showed. Minkowski's tensor is not the total: it is a subsystem tensor (the canonical light part), and subsystem tensors owe no symmetry — the medium's complementary tensor carries a compensating asymmetry, and the total (light + matter, microscopically the vacuum tensor plus particles) is symmetric and conserves angular momentum exactly. Historically the asymmetry was Abraham's strongest argument and it is decisive against reading $\mathbf{D}\times\mathbf{B}$ as the whole story; the canonical interpretation dissolves the objection by demoting Minkowski's object from "the momentum of the field" to "the canonical momentum of the light subsystem," which was the only sustainable reading anyway. There is a satisfying moral for tensor-minded readers: symmetry requirements attach to closed systems; the moment you split, bookkeeping asymmetries may appear in each part, constrained only to cancel in the sum. The same grammar returns in general relativity, where only the total stress–energy sources gravity.

What happens to the controversy in dispersive or absorbing media, where n depends on frequency?

The clean $n^2$ dichotomy blurs into a genuinely richer problem — this is where live research still happens. With dispersion, phase index $n(\omega)$ and group index $n_g = n + \omega\,dn/d\omega$ part company: the canonical momentum keeps $\hbar k = n\hbar\omega/c$ (wavevector is wavevector), but energy transport moves at the group velocity, so the kinetic ledger acquires $n_g$ factors, and the medium's co-traveling excitation (the polariton's matter share) carries a frequency-dependent slice of both energy and momentum. Slow-light experiments ( $n_g$ up to $10^7$ in atomic vapors) make these distinctions enormous and have become the sharpest testbeds. With absorption, momentum is transferred to the medium irreversibly (photon drag currents, radiation forces on absorbing droplets), and the split question becomes a transfer-rate question with electrostriction and thermal effects competing at the same order — experimental disentangling is delicate and occasionally still contentious. The settled structure survives: totals unique, kinetic vs canonical per the question; what dispersion adds is more terms in the medium's account, each measurable, none free. E8's Absorption and Dispersion builds the $n(\omega)$ machinery these refinements run on.

Was the century-long fight really necessary — could 1910s physics have resolved it?

Mostly yes, and the near-misses are instructive. Von Laue and others understood by the 1910s–20s that field-only tensors in media were incomplete without matter terms, and the "total is unique, split is convention" position appears explicitly in mid-century literature (Penfield–Haus's 1967 monograph essentially states it; Gordon's 1973 analysis of optical pressure in fluids computes reconciliations in detail). What kept the fight alive was threefold: experiments kept probing different questions without a shared taxonomy to say so; the hidden-momentum-family terms in the medium account weren't systematically understood until the late 1960s (the same Shockley-era clarification this chapter met an article ago); and the community lacked a crisp name for the two momenta — "kinetic vs canonical," which finally made the resolution teachable, is the 2010 settlement's real contribution as much as any calculation. The episode is now a standard case study in how physics disputes persist: not through anyone being wrong, but through a missing distinction that every party was implicitly straddling. This chapter's ledger discipline — totals, splits, questions — is that distinction, generalized.

Does gravity care which momentum light carries in a medium?

Gravity settles it the only way it can: by coupling to the total. In general relativity the source of curvature is the complete stress–energy tensor — microscopic fields plus matter — which is unique, symmetric, and conserved; a light pulse crossing a glass block gravitates via that total, with no Abraham/Minkowski ambiguity even in principle (the split never enters Einstein's equations). This is more than a formal remark: proposals to test the splits by weighing light in media (a slab on a balance while a pulse transits) are really tests of the medium account — the balance reads the slab's weight plus the transiting pulse's total energy-momentum contribution, and careful analyses show the reading is split-independent, as it must be. The same totality principle polices the momentum articles' earlier results: field momentum, hidden momentum, and their media refinements all gravitate only in the sum. If a future quantum-gravity experiment ever seemed to distinguish kinetic from canonical gravitationally, the correct first response — in the spirit this chapter has trained — would be to audit the medium's account for the term that went missing.