Maxwell's Equations in Dielectric Media

We take our starting point ready-made from electromagnetism. The companion article aq-em-maxwells-equations-in-matter has already done the hard microscopic averaging — smoothing $\sim 10^{19}$ atomic dipoles per cubic millimetre into continuous fields — and delivered the macroscopic Maxwell equations $\nabla\cdot\mathbf{D}=\rho_f$ , $\nabla\times\mathbf{H}=\mathbf{J}_f+\partial_t\mathbf{D}$ , with the displacement field $\mathbf{D}=\varepsilon_0\mathbf{E}+\mathbf{P}$ carrying all the bound charge out of sight. This article does not re-derive any of that — the source-splitting, and the vacuum equations with their speed $c=1/\sqrt{\mu_0\varepsilon_0}$ (owned by aq-em-maxwells-equations-in-vacuum), are electromagnetism's static and macroscopic machinery, recalled here by name. The de-overlap is clean: electromagnetism owns the equations and the plane wave they permit — the optical wave equation, the phase speed $c/n$ , and $n=\sqrt{\varepsilon_r}$ are recalled from it, not claimed here; optics' own contribution begins where the constant- $\varepsilon$ picture breaks down: the frequency dependence $\varepsilon(\omega)$ , the thread of every article that follows. So we take the four equations as given and ask the one question optics cares about: what wave do they permit inside a piece of glass, and how fast does it go? The answer hinges entirely on the constitutive relation — the single equation, $\mathbf{D}=\varepsilon\mathbf{E}$ , in which all of a material's optical personality is packed. Feed it into the source-free curl equations, eliminate $\mathbf{B}$ , and Maxwell's equations collapse into an ordinary wave equation whose speed is $v=c/n$ with $n=\sqrt{\varepsilon_r}$ . That is the whole of linear optics in embryo: refraction, dispersion, absorption, and reflection are all just this wave equation read out under different conditions on $\varepsilon$ . There is a sting in the tail, and it is the reason the entire O4 chapter exists. Push the relation $n=\sqrt{\varepsilon_r}$ on water and it fails spectacularly — water's static dielectric constant is about $80$ , which would predict $n\approx 9$ , yet water's optical index is $1.33$ . The constant $\varepsilon$ is a fiction. What the wave truly sees is $\varepsilon(\omega)$ , a frequency-dependent response, and unpacking that function is the work of every article that follows.

Key takeaways

Recalled from electromagnetism (aq-em-maxwells-equations-in-matter): the macroscopic equations $\nabla\cdot\mathbf{D}=\rho_f$ , $\nabla\cdot\mathbf{B}=0$ , $\nabla\times\mathbf{E}=-\partial_t\mathbf{B}$ , $\nabla\times\mathbf{H}=\mathbf{J}_f+\partial_t\mathbf{D}$ , with $\mathbf{D}=\varepsilon_0\mathbf{E}+\mathbf{P}$ . Optics takes these as the starting point.
The constitutive relation $\mathbf{D}=\varepsilon\mathbf{E}$ is where all the material physics enters; everything else is universal.
For a source-free, non-magnetic, linear medium the curl equations reduce to the wave equation $\nabla^2\mathbf{E}=(n^2/c^2)\,\partial_t^2\mathbf{E}$ with phase speed $v=c/n$ .
The refractive index is $n=\sqrt{\varepsilon_r\mu_r}\approx\sqrt{\varepsilon_r}$ : glass $\varepsilon_r=2.25\Rightarrow n=1.50$ , and $v=c/n=2.0\times10^8\ \mathrm{m\,s^{-1}}$ .
Light slows because the polarization $\mathbf{P}$ cannot respond instantly; that lag, made frequency-dependent, is dispersion.
Key idea: the macroscopic Maxwell equations plus one constitutive relation $\mathbf{D}=\varepsilon\mathbf{E}$ turn all of electromagnetism-in-matter into a wave equation with speed $c/n$ , $n=\sqrt{\varepsilon_r}$ — and the failure of that formula on water ( $\varepsilon_r^{\text{static}}=80$ but $n=1.33$ ) is the proof that $\varepsilon=\varepsilon(\omega)$ .

The core insight

Take the four macroscopic Maxwell equations from electromagnetism as given. Close them with the constitutive relation $\mathbf{D}=\varepsilon\mathbf{E}$ , and they become the wave equation $\nabla^2\mathbf{E}=(n^2/c^2)\partial_t^2\mathbf{E}$ with $n=\sqrt{\varepsilon_r}$ . All of linear optics is this one equation under different conditions on $\varepsilon(\omega)$ .

Where this lives today

Every simulation that designs a lens coating, a photonic waveguide, or a metamaterial starts from exactly these equations: a finite-difference or finite-element solver integrates the macroscopic Maxwell equations forward in time with a constitutive relation $\mathbf{D}=\varepsilon(\omega)\mathbf{E}$ supplied for each material. The entire discipline of computational electromagnetics — the FDTD method behind antenna design, integrated-photonics layout, and cloaking studies — is this article's content turned into an algorithm. The frequency dependence flagged by the water anomaly is now the central design variable. Optical engineers specify $\varepsilon(\omega)$ (equivalently $n(\omega)$ ) over the whole spectrum from a Sellmeier fit; radio and microwave engineers use the static $\varepsilon_r$ ; and the gap between them is precisely the dispersion that must be managed in every broadband fibre link and femtosecond laser. The one equation $\mathbf{D}=\varepsilon\mathbf{E}$ , read at the right frequency, underwrites all of it.

The physical picture

The macroscopic equations were built for us in electromagnetism, so we can be brief about where they come from and lavish about what they do. A cubic millimetre of glass holds too many oscillating dipoles to track, so the field is averaged over volumes tiny against a wavelength but huge against an atom, and the bound charge and current are swept into the polarization $\mathbf{P}$ and magnetization $\mathbf{M}$ . Two auxiliary fields, $\mathbf{D}=\varepsilon_0\mathbf{E}+\mathbf{P}$ and $\mathbf{H}=\mathbf{B}/\mu_0-\mathbf{M}$ , are defined precisely so that only the free sources — the charges an experimenter actually places — appear on the right-hand side. The physics is untouched; the accounting is transformed.

Figure

The recall-and-payoff chain: four macroscopic Maxwell equations (from aq-em-maxwells-equations-in-matter) closed by the constitutive relation D = εE collapse to a wave equation with speed c/n

Once you accept $\mathbf{D}=\varepsilon\mathbf{E}$ , the medium looks exactly like vacuum with the swap $\varepsilon_0\to\varepsilon$ . Light slows because $\varepsilon>\varepsilon_0$ : the bound electrons take time to re-radiate, their contribution lags the driving field, and the superposition of driver and re-radiation crawls along at $c/n$ . Everything interesting about a transparent solid — that it bends light, splits colours, and glints at its faces — is the single number $n=\sqrt{\varepsilon_r}$ doing its work, and the moment we admit that $\varepsilon$ depends on frequency, that single number becomes a whole spectrum.

Figure

The punchline: n = √ε_r nails glass, diamond, and optical water — but water's STATIC ε_r = 80 would demand n = 8.94, the famous failure that forces ε = ε(ω)

The core equations

Optics inherits four equations and adds one. The four macroscopic Maxwell equations are quoted from the electromagnetism owner aq-em-maxwells-equations-in-matter, which derives them by averaging the microscopic fields and defining the auxiliary fields to carry the bound sources; here they are simply the platform we build on. Written for the general medium, Gauss's laws read as follows, with only free charge as a source.

\nabla\cdot\mathbf{D}=\rho_f,\qquad \nabla\cdot\mathbf{B}=0 \qquad (1)

where: $\mathbf{D}=\varepsilon_0\mathbf{E}+\mathbf{P}$ — electric displacement field ( $\mathrm{C\,m^{-2}}$ ); $\rho_f$ — free charge density; $\mathbf{B}$ — magnetic flux density. The bound charge $-\nabla\cdot\mathbf{P}$ has been absorbed into $\mathbf{D}$ , which is why only $\rho_f$ appears (the mechanism is owned by aq-em-polarization-density-p). The two curl equations couple $\mathbf{E}$ and $\mathbf{B}$ and, with a time-varying field, are what allow a self-sustaining wave. Faraday's law is unchanged from vacuum; the Ampère–Maxwell law carries the displacement current $\partial_t\mathbf{D}$ that Maxwell added.

\nabla\times\mathbf{E}=-\frac{\partial\mathbf{B}}{\partial t},\qquad \nabla\times\mathbf{H}=\mathbf{J}_f+\frac{\partial\mathbf{D}}{\partial t} \qquad (2)

where: $\mathbf{H}=\mathbf{B}/\mu_0-\mathbf{M}$ — magnetic field ( $\mathrm{A\,m^{-1}}$ ); $\mathbf{J}_f$ — free current density; $\mathbf{M}$ — magnetization. For a transparent optical dielectric there are no free charges or currents, so $\rho_f=0$ and $\mathbf{J}_f=0$ . These four equations are underdetermined: four field quantities, not enough relations. The missing ingredient — and the only place a specific material enters — is the constitutive relation linking $\mathbf{D}$ to $\mathbf{E}$ . For a linear, isotropic, local medium it is a simple proportionality.

\mathbf{P}=\varepsilon_0\chi\,\mathbf{E},\qquad \mathbf{D}=\varepsilon_0(1+\chi)\mathbf{E}=\varepsilon\mathbf{E} \qquad (3)

where: $\chi$ — electric susceptibility (dimensionless); $\varepsilon=\varepsilon_0\varepsilon_r$ — permittivity; $\varepsilon_r=1+\chi$ — relative permittivity (dielectric constant). The relation $\mathbf{P}=\varepsilon_0\chi\mathbf{E}$ is the electrostatic form owned by aq-em-linear-dielectrics-susceptibility-and-permittivity; optics generalises $\chi\to\chi(\omega)$ , developed in the companion articles on susceptibility and the refractive index. Closing Maxwell's equations with (3) and eliminating $\mathbf{B}$ yields the equation optics actually uses: a wave equation, one for each Cartesian component of $\mathbf{E}$ , whose coefficient is the refractive index.

\nabla^2\mathbf{E}=\varepsilon\mu\,\frac{\partial^2\mathbf{E}}{\partial t^2}=\frac{n^2}{c^2}\,\frac{\partial^2\mathbf{E}}{\partial t^2},\qquad v=\frac{1}{\sqrt{\varepsilon\mu}}=\frac{c}{n} \qquad (4)

where: $n=\sqrt{\varepsilon_r\mu_r}$ — refractive index, $=\sqrt{\varepsilon_r}$ for non-magnetic media ( $\mu_r=1$ ); $c=1/\sqrt{\varepsilon_0\mu_0}$ — vacuum light speed; $v$ — phase velocity in the medium. A dielectric is optically just vacuum with $\varepsilon_0\to\varepsilon$ , and the wave slows by exactly the factor $n$ .

Derivation AThe optical wave equation and n = √ε_r

The aim is to show that the four macroscopic Maxwell equations, once closed with $\mathbf{D}=\varepsilon\mathbf{E}$ in a source-free medium, are not a new law of optics but the ordinary wave equation in disguise — and that its single coefficient is the refractive index. We assume a linear, isotropic, homogeneous, non-magnetic, non-conducting dielectric: $\mathbf{D}=\varepsilon\mathbf{E}$ , $\mathbf{B}=\mu_0\mathbf{H}$ , $\rho_f=0$ , $\mathbf{J}_f=0$ , with $\varepsilon$ a spatial constant. The strategy is to eliminate $\mathbf{B}$ by taking the curl of Faraday's law and feeding in Ampère's.

Step 1. Take the curl of Faraday's law, $\nabla\times(\nabla\times\mathbf{E})=-\partial_t(\nabla\times\mathbf{B})$ . Why: the two curl equations are coupled first-order equations; curling one lets us substitute the other and close the loop on a single field.

Step 2. Substitute Ampère's law. With $\mathbf{J}_f=0$ and $\mathbf{B}=\mu_0\mathbf{H}$ , $\nabla\times\mathbf{B}=\mu_0\nabla\times\mathbf{H}=\mu_0\varepsilon\,\partial_t\mathbf{E}$ , so $\nabla\times(\nabla\times\mathbf{E})=-\mu_0\varepsilon\,\partial_t^2\mathbf{E}$ . Why: the displacement current $\partial_t\mathbf{D}=\varepsilon\,\partial_t\mathbf{E}$ is what supplies the second time derivative that makes a wave.

Step 3. Expand the double curl with the identity $\nabla\times(\nabla\times\mathbf{E})=\nabla(\nabla\cdot\mathbf{E})-\nabla^2\mathbf{E}$ . In a homogeneous source-free medium $\nabla\cdot\mathbf{E}=\nabla\cdot\mathbf{D}/\varepsilon=\rho_f/\varepsilon=0$ , so the first term drops. Why: constant $\varepsilon$ lets $\nabla\cdot\mathbf{D}=0$ pass to $\nabla\cdot\mathbf{E}=0$ ; in an inhomogeneous medium the surviving $\nabla\varepsilon$ term couples polarizations and this clean split fails.

Step 4. Assemble: $\nabla^2\mathbf{E}=\mu_0\varepsilon\,\partial_t^2\mathbf{E}$ . Since $\mu_0\varepsilon=\mu_0\varepsilon_0\varepsilon_r=\varepsilon_r/c^2=n^2/c^2$ , the wave travels at $v=c/n$ with $n=\sqrt{\varepsilon_r}$ . Why: matching the coefficient $\mu_0\varepsilon$ to $1/v^2$ identifies the phase speed, and its ratio to $c$ defines the refractive index — the one measurable that a whole material collapses into. *Verified symbolically ( $\mu_0\varepsilon_0\varepsilon_r=\varepsilon_r/c^2$ ) and numerically (glass $\varepsilon_r=2.25\Rightarrow n=1.50$ , $v=2.0\times10^8\ \mathrm{m\,s^{-1}}$ ) in verify_aq-optics-maxwell-equations-dielectric-media.py.*

Go deeper · Derivation B · 1 minWhy a static dielectric constant cannot be the optical index

Derivation BWhy a static dielectric constant cannot be the optical index

The goal is to make precise the single most important fact in this chapter: that the constant $\varepsilon$ of electrostatics is not what an optical wave responds to. We do it by pushing $n=\sqrt{\varepsilon_r}$ onto a polar liquid, water, where the static and optical responses diverge by a factor of sixty — a contradiction that can only be resolved by making $\varepsilon$ a function of frequency. The static value is a genuine, measured electrostatic quantity; the optical value is a genuine, measured index; both are right, so the assumption that links them must be wrong.

Step 1. Take water's static (DC) relative permittivity, $\varepsilon_r^{\text{static}}\approx 80$ , a textbook electrostatics number set by the reorientation of permanent molecular dipoles. Applying $n=\sqrt{\varepsilon_r}$ naively predicts $n=\sqrt{80}=8.94$ . Why: this is exactly the formula just derived, applied without asking at what frequency $\varepsilon$ was measured.

Step 2. Measure water's actual optical index at visible frequencies: $n=1.33$ , i.e. an effective $\varepsilon_r^{\text{opt}}=n^2=1.77$ . The prediction is off by a factor of nearly seven in $n$ , forty-five in $\varepsilon$ . Why: a factor-of-seven miss is not experimental error — it is a signal that a physical assumption has broken.

Step 3. Resolve it: the permanent dipoles that give the huge static response are far too sluggish to follow a $5\times10^{14}\ \mathrm{Hz}$ optical field; at optical frequencies only the light electron clouds respond, and the permittivity has fallen to $1.77$ . The constitutive "constant" is therefore a frequency-dependent function $\varepsilon(\omega)$ , and $n=\sqrt{\varepsilon_r(\omega)}$ must be evaluated at the optical frequency. Why: response and frequency are inseparable — every mechanism (dipole rotation, ionic motion, electronic clouds) switches off above its own resonance, so $\varepsilon$ steps down as $\omega$ climbs. *Verified: $\sqrt{80}=8.94$ vs measured $n=1.33$ ( $\varepsilon_r^{\text{opt}}=1.77$ ); the discrepancy is asserted in verify_aq-optics-maxwell-equations-dielectric-media.py as the motivation for $\varepsilon(\omega)$ .*

Limiting cases

Case	Result	Meaning
Vacuum	$\mathbf{P}=\mathbf{M}=0\Rightarrow\mathbf{D}=\varepsilon_0\mathbf{E}$	Recovers the vacuum equations; wave speed $c=1/\sqrt{\varepsilon_0\mu_0}$
Linear non-magnetic dielectric	$\mathbf{D}=\varepsilon\mathbf{E}$ , $\mu_r=1$	The optics workhorse: $\nabla^2\mathbf{E}=(n^2/c^2)\partial_t^2\mathbf{E}$ , $n=\sqrt{\varepsilon_r}$
Static vs optical	$\varepsilon_r^{\text{static}}\neq\varepsilon_r^{\text{opt}}$	Water $80$ vs $1.77$ : $\varepsilon=\varepsilon(\omega)$ — the whole reason O4 exists
Conductor	$\mathbf{J}_f=\sigma\mathbf{E}\neq 0$	Ohmic term makes $n$ complex — skin depth and metallic reflection (the conducting-media regime, developed later in the track)
Anisotropic crystal	$D_i=\varepsilon_0\varepsilon_{ij}E_j$	$\varepsilon$ is a tensor, $\mathbf{D}\nparallel\mathbf{E}$ — birefringence and double refraction

Common trap“The dielectric constant is a fixed number you can look up”

The phrase "dielectric constant" is a historical trap. The single-number $\varepsilon_r$ tabulated in a chemistry handbook is the static value, measured in a capacitor at essentially zero frequency, and it is dominated by whatever slow mechanism a material happens to have — for water, the tumbling of permanent dipoles, giving $\varepsilon_r\approx 80$ . An optical wave oscillates a hundred trillion times a second; the dipoles cannot keep up, and the permittivity the light sees is an entirely different, much smaller number set only by the electrons. Use the static $\varepsilon_r$ in $n=\sqrt{\varepsilon_r}$ and you predict water's index as $9$ instead of $1.33$ . The correct statement is $n(\omega)=\sqrt{\varepsilon_r(\omega)}$ , evaluated at the frequency of the light — and the frequency dependence is dispersion.

Worked examples

1 · Refractive index and phase velocity of glass (confidence builder)

A borosilicate glass has $\varepsilon_r=2.25$ at optical frequencies and $\mu_r=1$ . Find its refractive index and the phase velocity of light inside it.

Step 1. Refractive index straight from the wave equation: $n=\sqrt{\varepsilon_r\mu_r}=\sqrt{2.25\times 1}=\mathbf{1.50}$ — the value for ordinary crown glass. The material's entire optical identity, at this frequency, is that one number.

Step 2. Phase velocity: $v=c/n=(2.998\times10^8)/1.50=\mathbf{2.0\times10^8\ m\,s^{-1}}$ . Light crawls through the glass at two-thirds of its vacuum speed.

Step 3. Sanity check on the wavelength: the frequency is fixed across the vacuum–glass boundary, so the wavelength contracts, $\lambda=\lambda_0/n$ . Green light at $\lambda_0=550$ nm becomes $367$ nm inside — the same colour, a shorter ruler. n = 1.50, v = 2.0 × 10⁸ m s⁻¹.

2 · The water anomaly (exam level)

Water's static dielectric constant is $\varepsilon_r\approx 80$ . Predict its refractive index from $n=\sqrt{\varepsilon_r}$ , compare with the measured optical index $1.33$ , and explain the discrepancy.

Step 1. Naive prediction: $n=\sqrt{80}=\mathbf{8.94}$ . This is wildly larger than any transparent liquid's index.

Step 2. Measured optical index: $n=1.33$ , so the effective optical permittivity is $\varepsilon_r^{\text{opt}}=n^2=1.77$ , not $80$ . The formula overpredicts $n$ by a factor of $6.7$ .

Step 3. Resolution: the static $80$ comes from permanent water dipoles reorienting in the field — a mechanism with a response time around $10^{-11}$ s, hopeless at optical periods of $2\times10^{-15}$ s. At optical frequencies only the electron clouds respond, and $\varepsilon_r$ drops to $1.77$ . The dielectric "constant" is a function $\varepsilon_r(\omega)$ that steps down each time the frequency outruns another polarization mechanism. √80 = 8.94 predicted vs 1.33 measured ⟹ ε = ε(ω).

Physics of the portrait

The whole edifice — light as an electromagnetic wave whose speed is fixed by a material's electrical response — rests on one constitutive relation and one man's audacious identification of two constants of nature that no one had thought to compare.

Origins · 1 minconstitutive relation and how this was discovered

constitutive relation

the equation D = εE that closes Maxwell's system · electromagnetic theory of light

A constitutive relation is the bridge between the universal field equations and a specific piece of matter: Maxwell's four equations hold in every medium, but they do not determine the fields until you say how $\mathbf{D}$ responds to $\mathbf{E}$ . For a linear dielectric that response is $\mathbf{D}=\varepsilon\mathbf{E}$ , and the single number $\varepsilon$ carries everything the material contributes to optics. The relation's deepest payoff was James Clerk Maxwell's in 1865: having shown that his equations support waves travelling at $1/\sqrt{\varepsilon_0\mu_0}$ , and finding that this equalled the measured speed of light, he concluded that light is an electromagnetic wave — and that a transparent medium's index should therefore be $n=\sqrt{\varepsilon_r}$ . Boltzmann confirmed it for simple non-polar gases in the 1870s. The relation also flagged its own limits: for polar liquids like water the static $\varepsilon_r$ gave the wrong index, the first sign that $\varepsilon$ is a frequency-dependent response and that the constitutive relation, not the field equations, is where all the subtle physics of light in matter lives.

History

1865

— Maxwell shows his equations support waves at

1/\sqrt{\varepsilon_0\mu_0}=c

and identifies light as electromagnetic, predicting

n=\sqrt{\varepsilon_r}

1873

— Maxwell's Treatise on Electricity and Magnetism systematises the field theory

1874

— Boltzmann measures

\varepsilon_r

for non-polar gases and confirms

n^2=\varepsilon_r

1880

— Lorentz and Lorenz relate the index to molecular polarizability, explaining why polar liquids deviate

1900

— Drude's electron theory (Lehrbuch der Optik) systematizes the optical constants of matter

Going further

Texts. Hecht, Optics, 5th ed., ch. 3 (electromagnetic theory) · Jackson, Classical Electrodynamics, 3rd ed., ch. 6–7 · Griffiths, Introduction to Electrodynamics, 4th ed., ch. 4 & 9 (macroscopic fields, waves in matter) · Saleh & Teich, Fundamentals of Photonics, 3rd ed., ch. 5 · Born & Wolf, Principles of Optics, §1.1 (foundations).

Provenance. The macroscopic Maxwell equations, the auxiliary fields $\mathbf{D}=\varepsilon_0\mathbf{E}+\mathbf{P}$ , $\mathbf{H}=\mathbf{B}/\mu_0-\mathbf{M}$ , and the bound-charge bookkeeping are recalled by name from the electromagnetism owner aq-em-maxwells-equations-in-matter (and -in-vacuum), which own their derivation — including the plane-wave reduction to $v=c/n$ and $n=\sqrt{\varepsilon_r}$ , reprised here in optical notation as the starting point. This article's own contribution is the frequency dependence $\varepsilon(\omega)$ the static treatment lacks (the water anomaly that opens O4). A verifier should confirm: (1) $\mu_0\varepsilon_0\varepsilon_r=\varepsilon_r/c^2=n^2/c^2$ so $v=c/n$ ; (2) glass $\varepsilon_r=2.25\Rightarrow n=1.50$ , $v=2.0\times10^8\ \mathrm{m\,s^{-1}}$ ; (3) the water anomaly $\sqrt{80}=8.94$ vs $n=1.33$ . Historical values (Maxwell 1865, Boltzmann's gas confirmation) are standard.

Prerequisites. Maxwell's Equations in Dielectric Media builds on Electric Polarization and the Displacement Field · Dielectric Susceptibility and Permittivity. Leads to. Plane Waves in Homogeneous Isotropic Media · The Refractive Index · The Lorentz Oscillator Model · Complex Refractive Index and the Extinction Coefficient.

Why bother with D and H if the "real" fields are E and B?

Because inside matter the bound charge and current — from $\sim 10^{19}$ atomic dipoles per cubic millimetre — are enormous and unknowable in detail. The definitions $\mathbf{D}=\varepsilon_0\mathbf{E}+\mathbf{P}$ and $\mathbf{H}=\mathbf{B}/\mu_0-\mathbf{M}$ absorb all of that bound response, so the source equations reference only the free charges and currents you actually control. It is pure bookkeeping, not new physics: $\mathbf{E}$ and $\mathbf{B}$ remain the fundamental fields, but $\mathbf{D}$ and $\mathbf{H}$ let you write the equations without tracking every atom. The electromagnetism owner article derives the split; optics simply uses it.

Is n = √ε_r ever exactly true?

It is exact for a linear, isotropic, non-magnetic, non-absorbing medium at the frequency where $\varepsilon_r$ is measured. The catches are all in that clause. If the medium is magnetic, $n=\sqrt{\varepsilon_r\mu_r}$ ; at optical frequencies $\mu_r=1$ to superb accuracy, so the $\sqrt{\varepsilon_r}$ form holds. If it absorbs, $\varepsilon_r$ is complex and so is $n$ . And crucially $\varepsilon_r$ must be the value at the optical frequency of interest, not the static one — that is the entire content of the water anomaly. Used correctly, $n=\sqrt{\varepsilon_r(\omega)}$ is one of the most reliable relations in optics.

Where does the light actually "slow down" — what is physically lagging?

The bound electrons. An incoming wave drives each atom's electron cloud, which oscillates and re-radiates its own little wave. Because the electrons have inertia, their re-radiated field lags the driver in phase. The total field — original plus all the re-radiated wavelets — is what propagates, and that superposition advances more slowly than $c$ , by exactly the factor $n$ . So nothing travels faster or slower than $c$ microscopically; the net wave crawls because it is continually being rebuilt from delayed contributions. When the lag becomes frequency-dependent (electrons respond differently near a resonance), the slowing becomes dispersion.

Why can I drop ∇(∇·E) in the wave-equation derivation but not always?

Because $\nabla\cdot\mathbf{E}=0$ requires two things: no free charge ( $\nabla\cdot\mathbf{D}=0$ ) and a spatially uniform $\varepsilon$ , so that $\nabla\cdot\mathbf{E}=\nabla\cdot\mathbf{D}/\varepsilon=0$ . In a homogeneous transparent dielectric both hold and the term vanishes, leaving a clean $\nabla^2\mathbf{E}$ . In a graded-index medium, near an interface, or in an anisotropic crystal, $\varepsilon$ varies in space and the surviving $\nabla(\nabla\cdot\mathbf{E})$ term couples the field components — which is exactly how graded-index lenses guide light and how birefringent crystals split it. The clean wave equation is the homogeneous-medium special case.

How does Snell's law fall out of these equations?

From the boundary conditions plus the phase-matching requirement. The tangential field components must be continuous across an interface, and for that to hold at every point along the boundary and for all time, the phase of the wave must match on both sides. The frequency is continuous across the interface but the wavelength is not ( $\lambda=\lambda_0/n$ ), so matching the tangential wavevector $k\sin\theta=(n\omega/c)\sin\theta$ gives $n_1\sin\theta_1=n_2\sin\theta_2$ immediately. Feeding the four continuity conditions into the plane-wave amplitudes then produces the Fresnel coefficients — so refraction and interface reflectivity both descend from this article's equations, developed in the polarization-on-reflection articles.

Does this framework handle metals and absorbers, or only transparent glass?

The same equations, with a complex or conducting constitutive relation. For a conductor, Ohm's law $\mathbf{J}_f=\sigma\mathbf{E}$ adds a term that makes the effective permittivity complex, and the wavevector $k=(n+i\kappa)\omega/c$ acquires an imaginary part that attenuates the wave — the skin effect and metallic reflection, the subject of the conducting-media articles later in the track. For a tinted glass or a dye, $\varepsilon(\omega)$ itself is complex near an absorption band, giving the Beer–Lambert decay $I=I_0e^{-\alpha z}$ with $\alpha=2\kappa\omega/c$ . Transparent glass, coloured filters, and polished metal are one framework with different constitutive relations — the unifying power of $\mathbf{D}=\varepsilon\mathbf{E}$ once $\varepsilon$ is allowed to be complex and frequency-dependent.