Clausius–Mossotti Relation

Dielectric Sphere in a Uniform Field ended with a promise. Pure macroscopic geometry — a ball of permittivity $\varepsilon_r$ in a uniform field — produced a response dial, the combination $(\varepsilon_r-1)/(\varepsilon_r+2)$ , and we pledged that its microscopic meaning would be earned later. This is later. The Clausius–Mossotti relation declares that this exact combination equals $N\alpha/3\varepsilon_0$ : number density times molecular polarizability, the two most microscopic quantities a dielectric owns. One bulk measurement of $\varepsilon_r$ , and you have weighed the electron cloud of a single molecule. The obstacle that makes the bridge nontrivial is a bookkeeping subtlety with teeth. The macroscopic field $\mathbf{E}$ of Maxwell's equations is a spatial average that includes the contribution of every molecule — including the one whose response you are trying to compute. A molecule does not polarize itself; it responds to the field of everything else, the local field $\mathbf{E}_{\mathrm{loc}}$ , which in dense matter is genuinely larger than $\mathbf{E}$ . Ignore the difference and you get a formula that fails by twenty percent for liquid argon and worse for anything denser. Respect it, and a two-line self-consistency argument delivers a relation that tracks a molecule's polarizability from vapor to liquid across a density jump of a factor of 372 — we will run that test on real carbon disulfide data. Mastering this article buys you the vocabulary of every micro-to-macro argument in condensed matter: self-consistent response, mean-field feedback, and the seed of effective-medium theory. It also hands you the chapter's cliffhanger — a denominator that can reach zero.

Key takeaways

The Clausius–Mossotti relation is the micro↔macro bridge: $\dfrac{\varepsilon_r-1}{\varepsilon_r+2}=\dfrac{N\alpha}{3\varepsilon_0}$ , connecting bulk permittivity $\varepsilon_r$ to molecular polarizability $\alpha$ and number density $N$ .
Its one physical input beyond $\mathbf{p}=\alpha\mathbf{E}_{\mathrm{loc}}$ is the Lorentz local field $\mathbf{E}_{\mathrm{loc}}=\mathbf{E}+\mathbf{P}/3\varepsilon_0$ — the field a molecule actually feels, derived carefully in Local Field and the Lorentz Cavity.
Solving the self-consistent loop gives $\chi_e=\dfrac{N\alpha/\varepsilon_0}{1-N\alpha/3\varepsilon_0}$ : the naive answer $N\alpha/\varepsilon_0$ , amplified by feedback from the polarized neighbourhood.
The molar polarization $P_M=\dfrac{\varepsilon_r-1}{\varepsilon_r+2}\dfrac{M}{\rho}=\dfrac{N_A\alpha}{3\varepsilon_0}$ is a molecular constant — density drops out. CS $_2$ vapor and liquid CS $_2$ agree to 2% across a ×372 density jump.
The same $(\varepsilon_r-1)/(\varepsilon_r+2)$ ruled the dielectric sphere: Clausius–Mossotti is secretly the statement that each molecule responds like a tiny dielectric ball filling its own share of the volume — the same spherical $1/3$ both times.
When $N\alpha/3\varepsilon_0\to1$ the denominator dies and $\varepsilon_r\to\infty$ : the polarization catastrophe, this chapter's road sign toward Ferroelectrics and Spontaneous Polarization.
Key idea: a molecule responds to the local field, not the macroscopic average; closing the loop $\mathbf{P}=N\alpha\,\mathbf{E}_{\mathrm{loc}}$ self-consistently yields $(\varepsilon_r-1)/(\varepsilon_r+2)=N\alpha/3\varepsilon_0$ .

The core insight

Each molecule responds to the local field $\mathbf{E}_{\mathrm{loc}}=\mathbf{E}+\mathbf{P}/3\varepsilon_0$ , not the macroscopic average field; self-consistently closing $\mathbf{P}=N\alpha\mathbf{E}_{\mathrm{loc}}$ yields $\frac{\varepsilon_r-1}{\varepsilon_r+2}=\frac{N\alpha}{3\varepsilon_0}$ .

Where this lives today

The relation is working infrastructure, not history. Every benchtop refractometer that turns a refractive-index reading into a concentration — sugar in grape juice, protein in serum, salinity at sea — is silently inverting Lorenz–Lorenz, banking on the density-linearity of $(n^2-1)/(n^2+2)$ . Atmospheric physicists use the same linearity in reverse: the radio refractivity of air that GPS signals feel is a Clausius–Mossotti sum over N $_2$ , O $_2$ , and water vapor, which is how GPS occultation measurements become weather data. And the response dial found a second career at the nanoscale. The polarizability of a small sphere carries the factor $(\varepsilon_p-\varepsilon_m)/(\varepsilon_p+2\varepsilon_m)$ , so a particle whose permittivity passes through $\varepsilon_p=-2\varepsilon_m$ has a diverging response — the Fröhlich resonance. Metals at optical frequencies do exactly that, which is why 20-nanometre gold spheres blaze ruby-red in suspension: the localized surface-plasmon resonance that powers biosensors and stained-glass windows alike is the Clausius–Mossotti denominator reaching zero in its optical costume. Effective-medium design — metamaterials, photonic composites, exoplanet-cloud models — still starts from Maxwell Garnett's 1904 generalization of the formula you derived above.

Origins · 1 minRudolf Clausius and how this was discovered

Rudolf Clausius

1822–1888 · Zürich, Würzburg, Bonn

Clausius is the nineteenth century's great bookkeeper of hidden microscopic quantities: entropy's inventor, co-founder of kinetic theory, coiner of the virial. The dielectric relation bearing his name appears in his 1879 treatise on the mechanical theory of heat, derived from the picture of molecules as small conducting spheres — thirty years after Ottaviano-Fabrizio Mossotti, working in Pisa in 1850, published essentially the same formula from essentially the same picture. The double name is history's compromise, and the physics community's habit of calling the optical version "Lorentz–Lorenz" (for two more independent discoverers, one Dutch, one Danish, publishing within months of each other in 1880) makes this modest algebraic identity perhaps the most repeatedly discovered equation in electromagnetism. What none of the four could know: their mean-field feedback denominator would become the template for twentieth-century self-consistent field theories, from Weiss ferromagnetism to plasmon resonances.

History

1850

— Mossotti, in Pisa, relates dielectric constant to the polarizability of molecular "conducting spheres" — the relation's first appearance

1879

— Clausius derives the same relation independently in Die mechanische Wärmetheorie, giving it its second parent

1878

— Lorentz's doctoral-era work constructs the local field E + P/3e0 for cubic media, supplying the missing microscopic justification

1880

— Lorenz (Copenhagen) and Lorentz (Leiden) independently publish the refractive-index form, the Lorenz–Lorenz equation

1904

— Maxwell Garnett generalizes the argument to metal-particle composites: effective-medium theory is born

1912

— Debye extends the relation to polar molecules with the p squared over 3kT orientational term, turning dielectric data into dipole-moment tables

1936

— Onsager repairs the polar-liquid failure with the reaction-field cavity, closing the classical local-field saga

The physical picture

Stand inside a polarized dielectric, at the site of one particular molecule, and ask the only question that matters to it: what field is stretching me? Not the applied field — that was screened and reshaped long ago. Not the macroscopic $\mathbf{E}$ either, because that average democratically includes the molecule's own dipole field, and no molecule polarizes itself. The honest answer is the field of everything else: the distant polarized medium, plus the handful of discrete neighbours breathing down its neck. Hendrik Lorentz found the clean way to organize that bookkeeping: carve an imaginary sphere around the molecule, big enough that everything outside blurs into a continuum, small enough that everything inside must be counted molecule by molecule. The continuum outside contributes the macroscopic $\mathbf{E}$ plus the field of the bound charge lining the imaginary cavity wall, which works out to $+\mathbf{P}/3\varepsilon_0$ — precisely the negative of the interior field of a uniformly polarized sphere, an old friend from Uniformly Polarized Sphere. The discrete neighbours inside contribute, for cubic or statistically isotropic arrangements, exactly zero. The full construction, its shape-dependence, and its failure modes are the business of the next article, Local Field and the Lorentz Cavity; here we take its headline result and spend it:

\mathbf{E}_{\mathrm{loc}}=\mathbf{E}+\frac{\mathbf{P}}{3\varepsilon_0} \qquad (1)

where: $\mathbf{E}_{\mathrm{loc}}$ — the field acting on one molecule (its self-field excluded); $\mathbf{E}$ — the macroscopic, spatially averaged field of Maxwell's equations; $\mathbf{P}$ — the polarization density. The correction is positive: a polarized medium pushes its members to polarize harder than the average field suggests. That plus sign sets up a feedback loop, and the feedback is the whole story. Molecules polarize; their polarization raises the local field; the raised local field polarizes them further. In a dilute gas the loop is negligible and the naive estimate survives. In a liquid it is a 20–70% effect. And if the medium is polarizable enough, the loop can in principle feed itself without any applied field at all — a runaway this chapter will meet again under the name polarization catastrophe.

Figure

Two descriptions of one material — the bridge the relation builds

Closing the loop

The microscopic side of the bridge is the constitutive law each molecule obeys, established in Induced Dipoles and Atomic Polarizability: a molecule in field $\mathbf{E}_{\mathrm{loc}}$ acquires

\mathbf{p}=\alpha\,\mathbf{E}_{\mathrm{loc}},\qquad \mathbf{P}=N\mathbf{p}=N\alpha\,\mathbf{E}_{\mathrm{loc}} \qquad (2)

where: $\alpha$ — the molecular polarizability in SI form ( $\mathrm{C\,m^2/V}$ ; the intuitive "polarizability volume" is $\alpha/4\pi\varepsilon_0$ , of order an ångström cubed); $N$ — molecules per cubic metre. Note carefully which field sits in the law: the local one. Writing $\mathbf{P}=N\alpha\mathbf{E}$ instead is the single most common error this topic produces, and everything below is the price of avoiding it. The macroscopic side is the definition of susceptibility from Linear Dielectrics, Susceptibility and Permittivity: $\mathbf{P}=\varepsilon_0\chi_e\mathbf{E}$ with $\varepsilon_r=1+\chi_e$ , where $\mathbf{E}$ is the total macroscopic field. The relation now follows from forcing both descriptions to describe the same material at once.

Derivation AThe Clausius–Mossotti relation from the self-consistent loop

We derive the bridge between $\alpha$ and $\varepsilon_r$ by closing the feedback loop: polarization creates local field, local field creates polarization. The strategy is to substitute the Lorentz expression (1) into the molecular law (2), notice that $\mathbf{P}$ then appears on both sides — the signature of self-consistency — and solve for it exactly rather than iteratively. The step to watch is the last one: a change of variable that turns an ugly rational expression into the clean, density-linear form the relation is famous for.

Step 1. Insert the local field into the polarization law:

\mathbf{P}=N\alpha\left(\mathbf{E}+\frac{\mathbf{P}}{3\varepsilon_0}\right).

Why is this legitimate? Every molecule in a uniform region sees the same statistics of neighbours, so the single-molecule relation scales up by $N$ unchanged — the mean-field move.

Step 2. Collect the two appearances of $\mathbf{P}$ and solve:

\mathbf{P}\left(1-\frac{N\alpha}{3\varepsilon_0}\right)=N\alpha\,\mathbf{E} \quad\Longrightarrow\quad \mathbf{P}=\frac{N\alpha}{1-N\alpha/3\varepsilon_0}\,\mathbf{E}.

Why solve exactly rather than iterate? Iterating $\mathbf{P}\to N\alpha(\mathbf{E}+\mathbf{P}/3\varepsilon_0)$ generates the geometric series $N\alpha E\,[1+u+u^2+\cdots]$ with $u=N\alpha/3\varepsilon_0$ — each pass adds one more round of "my neighbours polarized, so I polarize more." The closed form is that series summed; it converges precisely when $u<1$ , and the algebra quietly knows it.

Step 3. Compare with the macroscopic definition $\mathbf{P}=\varepsilon_0\chi_e\mathbf{E}$ to read off

\chi_e=\varepsilon_r-1=\frac{N\alpha/\varepsilon_0}{1-N\alpha/3\varepsilon_0}.

This is already the physics: the naive susceptibility $N\alpha/\varepsilon_0$ , amplified by the feedback denominator.

Step 4. Define $u\equiv N\alpha/3\varepsilon_0$ and tidy both combinations:

\varepsilon_r-1=\frac{3u}{1-u},\qquad \varepsilon_r+2=\frac{3}{1-u}.

Step 5. Divide. The messy $(1-u)$ cancels wholesale:

\boxed{\ \frac{\varepsilon_r-1}{\varepsilon_r+2}=u=\frac{N\alpha}{3\varepsilon_0}\ }

The left side is a bench measurement; the right side is molecular census data. The combination $(\varepsilon_r-1)/(\varepsilon_r+2)$ — nonlinear in $\varepsilon_r$ — is exactly linear in density, which is what makes the relation testable. Verified: with $u=N\alpha/3\varepsilon_0$ , $[3u/(1-u)]\,/\,[3/(1-u)]=u$ identically; dilute limit $u\ll1$ gives $\chi_e\approx N\alpha/\varepsilon_0(1+u+\cdots)$ , recovering the naive result to first order. Matches Jackson Eq. (4.70) and Griffiths Problem 4.41.

The feedback denominator deserves to be seen, not just derived. Below, the naive straight line $\chi_e=N\alpha/\varepsilon_0$ against the Clausius–Mossotti curve: they kiss at the origin, part company around $N\alpha/\varepsilon_0\sim0.5$ , and the curve runs away to its vertical asymptote at $N\alpha/\varepsilon_0=3$ .

Figure

Feedback in one curve — susceptibility versus density of polarizability

Read the two plotted materials off the curve. Liquid argon sits at $N\alpha/\varepsilon_0=0.43$ , where feedback boosts the naive $\chi_e$ by 17% — measurable, decisive (Worked Example 2 makes it a verdict). Liquid carbon disulfide sits at $1.08$ , where feedback is a 56% amplification: the naive line predicts $\varepsilon_r\approx2.08$ , the curve predicts $2.68$ , and the bench says $2.63$ .

The sphere echo, and why the same dial appears twice

Dielectric Sphere in a Uniform Field promised that the microscopic meaning of its response dial would be earned here. Watch closely — this is one of those coincidences that isn't.

Go deeper · Derivation B · 1 minEach molecule is a tiny dielectric sphere of the bulk material

Derivation BEach molecule is a tiny dielectric sphere of the bulk material

We show that the Clausius–Mossotti relation is algebraically identical to a startling statement: a molecule's polarizability equals that of a small sphere made of the bulk dielectric, exactly filling the molecule's allotted share of volume. The strategy is to invert the relation for $\alpha$ , then express the number density through the volume-per-molecule. The payoff is structural: the spherical $1/3$ of the Lorentz cavity and the spherical $1/3$ that built the sphere's $(\varepsilon_r-1)/(\varepsilon_r+2)$ are the same geometry wearing two costumes.

Step 1. Invert the boxed relation for the molecular polarizability:

\alpha=\frac{3\varepsilon_0}{N}\,\frac{\varepsilon_r-1}{\varepsilon_r+2}.

Step 2. Give each molecule its fair share of space. If each occupies volume $1/N$ , define the radius $a$ of the equivalent sphere by $\tfrac{1}{N}=\tfrac{4}{3}\pi a^{3}$ , so that $3\varepsilon_0/N=4\pi\varepsilon_0 a^{3}$ .

Step 3. Substitute:

\alpha=4\pi\varepsilon_0 a^{3}\,\frac{\varepsilon_r-1}{\varepsilon_r+2}.

Why is this remarkable? Compare with the induced moment of a macroscopic dielectric sphere of radius $R$ in an applied field, derived in Dielectric Sphere in a Uniform Field: $p=4\pi\varepsilon_0R^{3}\,\frac{\varepsilon_r-1}{\varepsilon_r+2}\,E_0$ , i.e. polarizability $4\pi\varepsilon_0R^{3}(\varepsilon_r-1)/(\varepsilon_r+2)$ . Identical, with $R\to a$ . Clausius–Mossotti says the bulk is consistent with each molecule acting as a miniature of the whole.

Step 4. Trace both factors of $1/3$ to their common source. The sphere problem's dial came from the depolarizing field $-\mathbf{P}/3\varepsilon_0$ of Uniformly Polarized Sphere; the Lorentz local field's $+\mathbf{P}/3\varepsilon_0$ is the complement of the same interior field (the cavity is the sphere problem inside-out). One geometric fact — a uniformly polarized sphere's interior field is $-\mathbf{P}/3\varepsilon_0$ — feeds both derivations, so the same combination of $\varepsilon_r$ must emerge. Verified: substituting $N=1/(\tfrac{4}{3}\pi a^3)$ into $\alpha=(3\varepsilon_0/N)(\varepsilon_r-1)/(\varepsilon_r+2)$ gives $4\pi\varepsilon_0a^3(\varepsilon_r-1)/(\varepsilon_r+2)$ exactly (checked symbolically). In the $\varepsilon_r\to\infty$ limit it reduces to the conducting-sphere polarizability $4\pi\varepsilon_0a^3$ — the crude model of Induced Dipoles and Atomic Polarizability recovered as a limiting case.

This little theorem is more than a party trick. Promote the "host" from vacuum to a medium of permittivity $\varepsilon_m$ and the same algebra yields the Maxwell Garnett effective-medium formula for composites — the response dial generalizing to $(\varepsilon_p-\varepsilon_m)/(\varepsilon_p+2\varepsilon_m)$ , exactly the factor Forces on Dielectrics used to steer cells and beads in dielectrophoresis and optical tweezers.

The experimental teeth: density scaling

A relation earns its keep by being falsifiable. Clausius–Mossotti's sharpest prediction: the combination $(\varepsilon_r-1)/(\varepsilon_r+2)$ , unlike $\chi_e$ itself, is strictly proportional to density, with the proportionality constant a molecular property. Package it per mole and the density cancels entirely.

Go deeper · Derivation C · 1 minMolar polarization — the number that must not change

Derivation CMolar polarization — the number that must not change

We construct the quantity experimentalists actually tabulate: the molar polarization $P_M$ . The strategy is to rewrite the relation's right side in terms of measurable bulk quantities — mass density $\rho$ and molar mass $M$ — and observe that everything environmental cancels, leaving a pure molecular constant. This is the form in which the relation is tested across phase changes, and the form in which it usually appears on data sheets (at optical frequency it is called molar refractivity).

Step 1. Express the number density through bench quantities: $N=N_A\rho/M$ , with $N_A$ Avogadro's number.

Step 2. Substitute into the relation and isolate the material-independent side:

\frac{\varepsilon_r-1}{\varepsilon_r+2}=\frac{N_A\rho\,\alpha}{3M\varepsilon_0} \quad\Longrightarrow\quad P_M\equiv\frac{\varepsilon_r-1}{\varepsilon_r+2}\,\frac{M}{\rho}=\frac{N_A\,\alpha}{3\varepsilon_0}.

Step 3. Read the result. The right side contains only constants of nature and $\alpha$ — no density, no temperature (for nonpolar molecules), no phase. Vapor, compressed gas, liquid: same $P_M$ , with units of volume per mole (numerically, about the actual volume of a mole of molecules — matter is polarizable roughly in proportion to how much electron cloud it contains). Why is this a stringent test and not a tautology? Because $\varepsilon_r$ itself changes violently with density — for CS $_2$ , from 1.0029 in the vapor to 2.63 in the liquid — while the packaged combination is predicted to stay fixed. Any failure of the local-field picture shows up as a drift in $P_M$ . *Verified: dimensional check — $[\alpha]=\mathrm{C\,m^2/V}$ , so $N_A\alpha/3\varepsilon_0$ has units $\mathrm{mol^{-1}\cdot C\,m^2 V^{-1}}/(\mathrm{C^2N^{-1}m^{-2}})=\mathrm{m^3/mol}$ ; matches $[M/\rho]$ . Worked Example 1 runs the CS $_2$ test numerically: agreement to 2.0% across a ×372 density jump.*

Figure

The density test — a molecular constant surviving condensation

One more identity closes the toolkit. At optical frequencies the permittivity that matters is $\varepsilon_r=n^2$ (a statement about the measuring frequency, unpacked properly when electromagnetic waves arrive in this track), and the relation becomes the Lorentz–Lorenz equation of refractometry:

\frac{n^{2}-1}{n^{2}+2}=\frac{N\alpha(\omega)}{3\varepsilon_0} \qquad (3)

where: $n$ — refractive index; $\alpha(\omega)$ — the polarizability at that frequency, which is the electronic part only: nuclei and permanent-dipole orientation cannot follow an optical-frequency field. That is why water has $n^2\approx1.77$ but static $\varepsilon_r\approx80$ — same molecule, different census of mechanisms. Dispersion, the full $\omega$ -dependence, belongs to the waves chapter; here the equation's daily job is humbler and mightier: measure $n$ and density, invert for $\alpha$ , and you have performed molecular spectroscopy with a refractometer.

Limiting cases

Case	Result	Meaning
Dilute gas, $N\alpha\ll3\varepsilon_0$	$\chi_e\approx N\alpha/\varepsilon_0$	Feedback negligible; naive and Clausius–Mossotti coincide to first order in density.
Dense nonpolar liquid	$\chi_e=\dfrac{N\alpha/\varepsilon_0}{1-N\alpha/3\varepsilon_0}$	Feedback amplification of tens of percent; the difference is measurable and the relation wins (liquid argon, Worked Example 2).
Optical frequency	$\dfrac{n^2-1}{n^2+2}=\dfrac{N\alpha(\omega)}{3\varepsilon_0}$	Lorentz–Lorenz form; only electronic polarizability responds, so $n^2\ll\varepsilon_r^{\mathrm{static}}$ for polar substances.
$N\alpha/3\varepsilon_0\to1$	$\varepsilon_r\to\infty$	Polarization catastrophe: feedback self-sustains, predicting spontaneous polarization — the mean-field road to Ferroelectrics and Spontaneous Polarization.
Polar molecules (water)	relation fails	Orientational response and dipole–dipole correlations break the Lorentz construction; Local Field and the Lorentz Cavity performs the autopsy and names the Onsager repair.
$\varepsilon_r\to\infty$ in the sphere echo	$\alpha\to4\pi\varepsilon_0a^3$	Conducting-sphere polarizability recovered — the crude model of Induced Dipoles and Atomic Polarizability as a limit.

Common trapThe field in p equals alpha E is not the field in P equals epsilon-zero chi E

The molecular law uses the local field; the macroscopic definition of $\chi_e$ uses the averaged field. They differ by $\mathbf{P}/3\varepsilon_0$ , and confusing them collapses Clausius–Mossotti back to the naive $\chi_e=N\alpha/\varepsilon_0$ — wrong by 17% for liquid argon, 56% for liquid CS $_2$ , and catastrophically for anything approaching ferroelectricity. Two smaller traps ride along. First, units: SI polarizability ( $\mathrm{C\,m^2/V}$ ) versus the Gaussian "polarizability volume" ( $\mathrm{cm^3}$ ) differ by $4\pi\varepsilon_0$ , and tables mix them freely — check units before substituting. Second, frequency: $n^2=\varepsilon_r$ links the same-frequency quantities only; inserting water's static $\varepsilon_r=80$ into the Lorenz–Lorenz form predicts $n\approx9$ , off by a factor of seven, because the orientational mechanism that produces the 80 cannot follow light.

Worked examples

Example 1 — the CS $_2$ density test: one molecule, two phases. Carbon disulfide vapor at $0^{\circ}\mathrm{C}$ and 1 atm has $\varepsilon_r=1.0029$ ; the liquid at $20^{\circ}\mathrm{C}$ has $\varepsilon_r=2.63$ and density $\rho=1263\ \mathrm{kg/m^3}$ . Molar mass $M=76.13\ \mathrm{g/mol}$ . Does one polarizability explain both?

Step 1. Vapor side. The molar volume at $0^{\circ}\mathrm{C}$ , 1 atm is $V_m=RT/P=22.41\times10^{-3}\ \mathrm{m^3/mol}$ , so

P_M^{\mathrm{gas}}=\frac{0.0029}{3.0029}\times22.41\times10^{-3}=2.165\times10^{-5}\ \mathrm{m^3/mol}.

Step 2. Liquid side. The molar volume is $M/\rho=6.028\times10^{-5}\ \mathrm{m^3/mol}$ — the density jump from vapor is a factor of $372$ . Then

P_M^{\mathrm{liq}}=\frac{1.63}{4.63}\times6.028\times10^{-5}=2.122\times10^{-5}\ \mathrm{m^3/mol}.

Step 3. Compare: the molar polarization drops by only 2.0% while the density multiplies by 372 and $\varepsilon_r$ nearly triples. The molecular constant survives condensation. Extracting the polarizability from the vapor value, $\alpha=3\varepsilon_0P_M^{\mathrm{gas}}/N_A=9.55\times10^{-40}\ \mathrm{C\,m^2/V}$ , a polarizability volume of $8.6\ \mathrm{\AA^3}$ — a plausibly molecule-sized answer for a 3-atom molecule with fat sulfur electron clouds.

Step 4. Run the prediction forward. Using the gas-phase $\alpha$ at liquid density, $u=N_{\mathrm{liq}}\alpha/3\varepsilon_0=0.359$ , so Clausius–Mossotti predicts $\varepsilon_r=(1+2u)/(1-u)=2.68$ , against the measured $2.63$ — a 1.9% overshoot. The naive formula predicts $\varepsilon_r=1+N\alpha/\varepsilon_0=2.08$ , a 21% miss. The local field is not a refinement; it is most of the answer.

Example 2 — liquid argon as judge and jury. Argon's polarizability volume is $1.641\ \mathrm{\AA^3}$ ( $\alpha=1.83\times10^{-40}\ \mathrm{C\,m^2/V}$ , known precisely from gas refractometry). Liquid argon at 87 K has $\rho=1396\ \mathrm{kg/m^3}$ , $M=39.95\ \mathrm{g/mol}$ . Predict its dielectric constant both ways.

Step 1. Number density: $N=N_A\rho/M=2.104\times10^{28}\ \mathrm{m^{-3}}$ . The dimensionless load is $N\alpha/\varepsilon_0=0.434$ , i.e. $u=0.145$ .

Step 2. Naive prediction (no local-field correction): $\varepsilon_r=1+0.434=1.434$ .

Step 3. Clausius–Mossotti: $\chi_e=0.434/(1-0.145)=0.507$ , so $\varepsilon_r=1.507$ .

Step 4. The measured value is $\varepsilon_r\approx1.52$ – $1.53$ . Clausius–Mossotti lands within about 1.5%; the naive formula misses by 6% — twelve times the experimental uncertainty. For a monatomic, nonpolar, nearly-random liquid (the best-case medium for the Lorentz assumptions), the feedback term is exactly what the data demand.

Step 5. For perspective, ask how far argon sits from the catastrophe: $N_{\mathrm{crit}}=3\varepsilon_0/\alpha=1.45\times10^{29}\ \mathrm{m^{-3}}$ , about 6.9 times the liquid density. Ordinary van-der-Waals matter cannot pack anywhere near its own catastrophe — a first hint that reaching the instability requires a different mechanism (soft ionic lattices), which is the ferroelectrics story.

Going further

References. Griffiths, Introduction to Electrodynamics, 4th ed., §4.4.3 (Problem on Clausius–Mossotti) · Jackson, Classical Electrodynamics, 3rd ed., §4.5 (Molecular Polarizability and Electric Susceptibility) · Purcell & Morin, Electricity and Magnetism, 3rd ed., Ch. 10 · Zangwill, Modern Electrodynamics, §6.5 · Ashcroft & Mermin, Solid State Physics, Ch. 27 (Dielectric Properties of Insulators)

Provenance. Equations and derivations cross-checked against Jackson (Classical Electrodynamics, 3rd ed., §4.5, Eqs. 4.68–4.70), Griffiths (Introduction to Electrodynamics, 4th ed., Problem 4.41 and §4.4.3), and Ashcroft & Mermin (Ch. 27). A verifier should confirm: (1) the interior field of a uniformly polarized sphere is $-\mathbf{P}/3\varepsilon_0$ ; (2) the algebraic identity $(\varepsilon_r-1)/(\varepsilon_r+2)=N\alpha/3\varepsilon_0$ follows from $\mathbf{P}=N\alpha(\mathbf{E}+\mathbf{P}/3\varepsilon_0)$ with $\varepsilon_r=1+\chi_e$ ; (3) constants $\varepsilon_0=8.854\times10^{-12}\,\mathrm{F/m}$ , $N_A=6.022\times10^{23}\,\mathrm{mol^{-1}}$ . No experimental polarizability values are quoted to avoid unverified numbers.

Builds on: Polarization Density P · Induced Dipoles and Atomic Polarizability · Uniformly Polarized Sphere · Linear Dielectrics, Susceptibility and Permittivity · Field of a Polarized Object Leads to: Dielectric Materials and Insulators · Permanent Dipoles and Orientational Polarization · Linear Dielectrics, Susceptibility and Permittivity

Why doesn't every dielectric run away, if each molecule's polarization increases its neighbours' local field?

Because the feedback loop has a gain below one. Each round trip multiplies the polarization by $u=N\alpha/3\varepsilon_0$ , and the total response is the geometric series $1+u+u^2+\cdots=1/(1-u)$ : finite whenever $u<1$ . For real van-der-Waals matter $u$ is stuck well below one — liquid argon manages 0.145, dense CS $_2$ about 0.36 — because packing molecules more tightly than their own electron clouds is exactly what interatomic repulsion forbids. Reaching $u=1$ needs a lattice whose ions are anomalously easy to displace, which is the ferroelectric mechanism, not ordinary compression.

Is the local field something I could measure, or is it a theoretical bookkeeping device?

It is real and it has measurable consequences, but no probe you insert measures it directly — any probe large enough to average is macroscopic, and any probe small enough is a molecule with its own local-field problem. Its reality shows indirectly: the 17% argon discrepancy between naive and corrected susceptibility is experiment discriminating between the two candidate fields. Modern simulations can compute the field at a molecular site explicitly, and for cubic crystals they confirm the Lorentz $+\mathbf{P}/3\varepsilon_0$ ; for polar liquids they confirm its failure. The next article does that autopsy honestly.

Which alpha goes into the relation — the static one or the optical one?

Whichever matches the frequency of the $\varepsilon_r$ on the other side; the relation is a same-frequency identity. With static $\varepsilon_r$ you must use the full static polarizability (electronic plus ionic displacement plus, for polar molecules, the orientational term — though for those the relation itself is failing). With $\varepsilon_r=n^2$ at optical frequency, only the electronic cloud responds and $\alpha$ is the electronic polarizability. Comparing the two extractions for one substance is actually a diagnostic tool: the difference between static and optical molar polarization measures the ionic and orientational contributions.

My table lists polarizability in cubic ångströms. What do I do with that?

That is the polarizability volume, $\alpha'=\alpha/4\pi\varepsilon_0$ , the Gaussian-unit habit that survived in data tables because it is intuitive: $\alpha'$ is comparable to the molecular volume ( $1.64\ \mathrm{\AA^3}$ for argon, $8.6\ \mathrm{\AA^3}$ for CS $_2$ ). Convert with $\alpha=4\pi\varepsilon_0\alpha'$ before using SI formulas, or rewrite the relation directly as $(\varepsilon_r-1)/(\varepsilon_r+2)=\tfrac{4\pi}{3}N\alpha'$ . Half of all wrong factors of $10^{-40}$ in homework trace to skipping this conversion.

Does the relation assume the molecules sit on a cubic lattice? Liquids obviously don't.

The derivation needs the discrete near-neighbour contribution to the local field to vanish, which is exact for cubic lattices and true on average for statistically isotropic arrangements — which is precisely what a nonpolar liquid is. That is why liquid argon obeys the relation so well. What kills it is not the absence of a lattice but the presence of correlations: polar molecules torque their neighbours into aligned clusters, the near-field sum stops averaging to zero, and the Lorentz construction loses its licence. Structure, not disorder, is the enemy.

Is the polarization catastrophe ever actually reached, or is it purely hypothetical?

It is closely approached, and its approach is visible. In perovskites like barium titanate, a soft lattice mode makes the effective ionic polarizability grow as temperature falls, driving $N\alpha/3\varepsilon_0$ toward one; the diverging Clausius–Mossotti susceptibility is observed as the Curie–Weiss growth of $\varepsilon_r$ (into the tens of thousands) near the transition. At the transition the material sidesteps a true infinity by spontaneously polarizing and restructuring — the mean-field catastrophe rounded into a phase transition. Ferroelectrics and Spontaneous Polarization takes it from here.

How is Maxwell Garnett's composite formula related to what we derived?

Replace the vacuum host by a background medium of permittivity $\varepsilon_m$ and rerun the sphere-echo logic: inclusions of permittivity $\varepsilon_p$ occupying volume fraction $f$ act like molecules with polarizability proportional to $(\varepsilon_p-\varepsilon_m)/(\varepsilon_p+2\varepsilon_m)$ , and the effective permittivity obeys a Clausius–Mossotti relation with $\varepsilon_0\to\varepsilon_m$ and $N\alpha/3\varepsilon_0\to f\,(\varepsilon_p-\varepsilon_m)/(\varepsilon_p+2\varepsilon_m)$ . That is the workhorse first estimate for any dilute composite — gold nanoparticles in glass, water droplets in cloud, voids in ceramic — and it inherits both the successes and the dilute-limit caveats of its parent.

Why does the exam answer sometimes use eps_r minus 1 over eps_r plus 2 and sometimes just chi over 3?

They are the same object in different clothes: $(\varepsilon_r-1)/(\varepsilon_r+2)=\chi_e/(\chi_e+3)$ , and for small $\chi_e$ this is $\approx\chi_e/3$ . In the dilute regime all the forms collapse together, which is why gas-phase problems tolerate sloppiness. In the dense regime only the full combination is linear in density — that linearity is the relation's actual content, so when in doubt, write the ratio form and let the algebra protect you.

Clausius–Mossotti Relation

Where this lives today

The physical picture

Closing the loop

The sphere echo, and why the same dial appears twice

The experimental teeth: density scaling

Limiting cases

Worked examples

Going further

Before this

Where this leads