Permanent Dipoles and Orientational Polarization

The previous article stretched atoms and got parts-per-million dipoles for its trouble. Yet water's dielectric constant is 80 while hexane's — a bigger, more polarizable molecule — is 1.9. Clearly water is playing a different game. It is: the water molecule is born polar, its bent geometry locking in a permanent moment $p_0 = 1.85\,\mathrm{D} = 6.2\times10^{-30}\,\mathrm{C\cdot m}$ — thirty thousand times anything a laboratory field can induce. A field applied to water need not create dipoles; it need only turn them, using the torque $\mathbf{N} = \mathbf{p}\times\mathbf{E}$ derived two articles ahead in study order and recalled here by name. But turning molecules is a fight. Every collision — at room temperature, every fraction of a picosecond — re-randomizes orientations, and the thermal energy scale $k_BT \approx 4\times10^{-21}\,\mathrm{J}$ towers over the alignment energy $p_0E \approx 6\times10^{-24}\,\mathrm{J}$ that even a megavolt-per-meter field offers. The result is a statistical compromise: orientations remain almost random, biased toward the field by a fraction of a percent, and that whisper of bias — multiplied by $10^{25}$ – $10^{28}$ molecules — is orientational polarization, the second and stronger engine of dielectric response. The mathematics of the compromise is one clean Boltzmann average, and it yields the Langevin function $L(u) = \coth u - 1/u$ with $u = p_0E/k_BT$ — derived here in full, including the factor $\tfrac13$ that governs its linear regime and the saturation plateau no static laboratory field can reach. The payoff formulas carry a signature the induced mechanism lacks: temperature in the denominator. That $1/T$ is Debye's fingerprint — measure a dielectric's response against temperature and the two mechanisms confess separately — and it is also why your microwave oven heats the water in food and nothing else.

Key takeaways

Polar molecules (water $1.85\,\mathrm{D}$ , HCl $1.08\,\mathrm{D}$ ) carry permanent dipole moments thousands of times larger than any induced moment; a field aligns them via the torque $\mathbf{N} = \mathbf{p}\times\mathbf{E}$ while collisions randomize them.
The Boltzmann average over orientations gives $\langle\cos\theta\rangle = L(u) = \coth u - 1/u$ , the Langevin function, with $u = p_0E/k_BT$ the alignment-to-thermal energy ratio.
Ordinary conditions sit deep in the linear regime: $u \sim 10^{-3}$ , so $L(u) \approx u/3$ and $\langle p\rangle = p_0^2E/3k_BT$ — alignment is a fraction of a percent, yet it dominates the response.
The orientational susceptibility $\chi = np_0^2/3\varepsilon_0k_BT$ falls as $1/T$ (Curie behavior); induced polarization is temperature-blind — Debye's separation trick reads both mechanisms off one $\chi(T)$ plot.
Saturation ( $L \to 1$ , all dipoles aligned) needs $u \sim 1$ , i.e. $E \sim 7\times10^8\,\mathrm{V/m}$ at room temperature — far beyond static breakdown; only cryogenics or laser fields get near it.
For water the orientational "polarizability" $p_0^2/3k_BT$ beats the induced $\alpha$ by a factor $\approx 19$ at room temperature — and would keep winning up to a fictitious crossover near 5700 K.
KEY IDEA: orientational polarization is a thermal tug-of-war — field energy $p_0E$ against $k_BT$ — giving alignment $L(p_0E/k_BT)$ , linear with slope $p_0^2/3k_BT$ at ordinary fields and saturating at complete alignment only at fields no static laboratory reaches.

The core insight

Orientational polarization is a thermal tug-of-war: the field energy $p_0E$ tries to align permanent dipoles while thermal energy $k_BT$ randomizes them, giving a net alignment $\langle\cos\theta\rangle=L(p_0E/k_BT)$ that is linear in $E$ (susceptibility $\propto 1/T$ ) at ordinary fields and saturates at $np_0$ in strong fields.

Where this lives today

The most-used appliance application of this article runs at $2.45\,\mathrm{GHz}$ . A microwave oven drives water's permanent dipoles to reorient billions of times per second; at that frequency the molecules in liquid water can almost keep up — they lag the field (the rotational drag called Debye relaxation, time constant $\sim10\,\mathrm{ps}$ for water), and the lag dissipates field energy as heat. Fats and sugars, feebly polar, barely couple; ice, rotation-locked, barely couples — which is why the water in food heats while the plate stays cool and frozen cores thaw last. The same lossy-reorientation physics, viewed as a nuisance, sets atmospheric absorption windows for radar and satellite links (water vapor lines) and drives the search for low-loss board materials in 5G hardware. Dielectric spectroscopy has meanwhile grown from Debye's single-molecule method into a probe of everything that reorients: sweep frequency from Hz to THz and each polarization mechanism drops out at its own characteristic rate — protein side chains, hydration shells, polymer segments, ionic atmospheres — each leaving a Debye-like step in $\kappa(\omega)$ . Biophysicists read protein dynamics from these steps; food scientists certify moisture content; battery engineers watch electrolyte solvation. And below it all, the tabulated molecular dipole moments that parameterize modern force fields for molecular-dynamics simulation — every water model in every protein simulation carries a $p_0$ near $1.85$ – $2.3\,\mathrm{D}$ — descend directly from Curie-law slopes measured on gases a century ago.

Origins · 1 minPaul Langevin and how this was discovered

Paul Langevin

1872–1946 · ESPCI and Collège de France, Paris

Langevin derived the function bearing his name in 1905 for magnetism — permanent molecular magnetic moments fighting thermal chaos — and the identical average, with $p_0E$ for $\mu B$ , became the foundation of orientational dielectric theory when Debye imported it in 1912. It was one strike in a career of them: the Langevin equation founded the theory of Brownian motion in a form physicists still prefer; his analysis of paramagnetism gave Weiss the platform for ferromagnetic mean-field theory; and during the First World War he invented sonar, driving quartz slabs piezoelectrically to chase submarines with ultrasound. A student of Pierre Curie (whose $1/T$ law his statistics explained) and teacher of half of French physics, Langevin was also the century's model of the engaged scientist: an early, vocal champion of Einstein's relativity in France, president of the Human Rights League, dismissed and placed under house arrest by the Vichy regime — he survived to see the Liberation, and the École where he taught now bears his name on its physics laboratories.

History

1785

— Coulomb's torsion balance measures the torques on magnetized needles — alignment physics before molecules

1905

— Langevin's Boltzmann average over orientations yields L(u) for paramagnetic gases

1912

— Debye transplants the Langevin average to electric dipoles, founding the theory of polar molecules

1929

— Debye's Polar Molecules systematizes dipole-moment measurement from temperature-swept permittivity

1936

— Onsager repairs the dense-liquid theory, explaining why water's 80 defies the dilute formula

1947

— The first commercial microwave oven turns Debye relaxation of water into a kitchen appliance

The physical picture

Watch one water molecule in steam for a nanosecond. Its dipole arrow tumbles chaotically — struck by neighbors thousands of times, pointing every direction in rapid succession. Now switch on a field. The torque $\mathbf{N} = \mathbf{p}\times\mathbf{E}$ (anatomy in Torque and Force on a Dipole in a Field) twists the arrow toward alignment during every free flight, and every collision undoes the work. The molecule does not align; it leans. Averaged over tumbles, orientations along the field are visited slightly more often than orientations against it, because they cost less energy and the Boltzmann factor $e^{-U/k_BT}$ pays energy discounts in exponential coin.

Figure

The tug-of-war — field bias versus thermal chaos

How strong is the lean? The whole physics is in one dimensionless ratio, and it is worth computing before any calculus. Alignment energy at stake: $p_0E$ , about $6\times10^{-24}\,\mathrm{J}$ for water in a strong laboratory field of $10^6\,\mathrm{V/m}$ . Randomizing energy available: $k_BT \approx 4\times10^{-21}\,\mathrm{J}$ at room temperature. Their ratio $u = p_0E/k_BT \approx 1.5\times10^{-3}$ . Thermal chaos outguns the field seven hundred to one, and the average alignment — computed properly below — is $u/3 \approx 0.05\%$ . Every formula in this article is a statement about that biased chaos; the miracle, as always in matter physics, is the census that multiplies a $10^{-4}$ bias into $\kappa = 80$ . Why does this mechanism dominate wherever it operates? Compare budgets with the induced mechanism. Stretching a nonpolar molecule buys a moment $\alpha E \sim 10^{-34}\,\mathrm{C\cdot m}$ ; leaning a water molecule contributes $p_0\langle\cos\theta\rangle \sim 6\times10^{-30}\times5\times10^{-4} \sim 3\times10^{-33}\,\mathrm{C\cdot m}$ — an order of magnitude more, per molecule, from a mechanism that merely redirects a pre-built moment. The effective orientational polarizability works out to $p_0^2/3k_BT$ , which for water beats the electronic $\alpha$ nineteen-fold at room temperature. And uniquely, it carries a thermometer: heat the sample and the denominator grows, the lean fades, the response weakens as $1/T$ — while the induced response, a stiff spring with no entropy in the game, does not budge. That contrast is experimental gold, and Debye mined it first.

The Langevin average

Everything reduces to one question: given the orientation energy $U(\theta) = -p_0E\cos\theta$ (from Torque and Force on a Dipole in a Field), what is the thermal-equilibrium average of $\cos\theta$ ? Statistical mechanics answers with the Boltzmann-weighted average over the sphere of orientations:

\langle\cos\theta\rangle = \frac{\displaystyle\int \cos\theta\; e^{\,u\cos\theta}\,\mathrm{d}\Omega}{\displaystyle\int e^{\,u\cos\theta}\,\mathrm{d}\Omega} = \coth u - \frac{1}{u} \equiv L(u), \qquad u = \frac{p_0E}{k_BT} \qquad (1)

where: $L(u)$ — the Langevin function; $u$ — the alignment-to-thermal energy ratio; $\mathrm{d}\Omega = \sin\theta\,\mathrm{d}\theta\,\mathrm{d}\phi$ — the solid-angle element. Landmark values: $L(0.1) = 0.033$ , $L(1) = 0.313$ , $L(3) = 0.672$ , $L(10) = 0.900$ — the function climbs with slope $\tfrac13$ from the origin and creeps toward complete alignment as $1 - 1/u$ . The average molecular moment along the field is $\langle p\rangle = p_0L(u)$ , and the macroscopic polarization of $n$ dipoles per volume is $P = np_0L(u)$ . At every static field a laboratory can sustain, $u \ll 1$ and only the linear zone matters:

\langle p\rangle = \frac{p_0^2}{3k_BT}\,E \qquad\Longrightarrow\qquad \alpha_{\rm orient} = \frac{p_0^2}{3k_BT}, \qquad \chi_{\rm orient} = \frac{n\,p_0^2}{3\,\varepsilon_0k_BT} \qquad (2)

where: $\alpha_{\rm orient}$ — the effective orientational polarizability, playing exactly the role $\alpha$ played for induced dipoles but built from statistics rather than springs; $\chi_{\rm orient}$ — the corresponding susceptibility of a dilute gas ( $\kappa = 1 + \chi$ ; dense liquids need the local-field corrections deferred to Local Field and the Lorentz Cavity). The $1/T$ is the Curie law of dielectrics. Add the temperature-blind induced part and you have the Debye equation, $\alpha_{\rm total} = \alpha_{\rm ind} + p_0^2/3k_BT$ : plot measured $\alpha_{\rm total}$ against $1/T$ and the slope hands you the molecule's permanent moment while the intercept hands you its electronic polarizability — molecular structure read off a capacitance bridge.

Derivation AThe Langevin function from the Boltzmann average

The goal is equation (1): the mean alignment of a classical permanent dipole in thermal equilibrium, exactly, for any field strength. The strategy is direct evaluation of the Boltzmann average over orientations — one substitution turns both integrals into elementary ones. Watch two things: the azimuthal symmetry that kills all components transverse to the field (only $\langle\cos\theta\rangle$ survives), and the honest classical assumptions — continuous orientations, no dipole-dipole interactions — which bound where the result applies (gases and dilute solutions, not ice).

Step 1. Set up the average. Orientation energy $U = -p_0E\cos\theta$ ; each orientation is weighted by $e^{-U/k_BT} = e^{u\cos\theta}$ . Transverse components average to zero by symmetry (for every molecule leaning east there is a statistical twin leaning west), so the only survivor is

\langle\cos\theta\rangle = \frac{\int_0^{2\pi}\!\mathrm{d}\phi\int_0^\pi \cos\theta\,e^{u\cos\theta}\sin\theta\,\mathrm{d}\theta}{\int_0^{2\pi}\!\mathrm{d}\phi\int_0^\pi e^{u\cos\theta}\sin\theta\,\mathrm{d}\theta}.

Step 2. Substitute $s = \cos\theta$ , $\mathrm{d}s = -\sin\theta\,\mathrm{d}\theta$ ; the $\phi$ integrals cancel between numerator and denominator:

\langle\cos\theta\rangle = \frac{\int_{-1}^{1} s\,e^{us}\,\mathrm{d}s}{\int_{-1}^{1} e^{us}\,\mathrm{d}s}.

The denominator is elementary: $\int_{-1}^{1}e^{us}\mathrm{d}s = (e^{u} - e^{-u})/u = 2\sinh u/u$ .

Step 3. The numerator, by parts (or by differentiating the denominator with respect to $u$ — the standard partition-function trick, $\langle s\rangle = \mathrm{d}\ln Z/\mathrm{d}u$ ):

\int_{-1}^{1}s\,e^{us}\,\mathrm{d}s = \frac{\mathrm{d}}{\mathrm{d}u}\left[\frac{2\sinh u}{u}\right] = \frac{2\cosh u}{u} - \frac{2\sinh u}{u^2}.

Divide:

\langle\cos\theta\rangle = \frac{\cosh u}{\sinh u} - \frac{1}{u} = \coth u - \frac{1}{u} = L(u).

Why the answer must have this shape: as $u\to\infty$ , $\coth u\to1$ and $L\to1 - 1/u$ — alignment saturates, approaching completeness only asymptotically because the last stragglers fight pure geometry (there are vanishingly few orientations exactly along the field). As $u\to0$ , both terms diverge and their difference must be extracted with care — Step 4.

Step 4. The linear regime. Expand $\coth u = 1/u + u/3 - u^3/45 + \cdots$ (the Laurent series), so

L(u) = \frac{u}{3} - \frac{u^3}{45} + \cdots \approx \frac{u}{3} \quad (u \ll 1).

Hence $\langle p\rangle = p_0L(u) \approx p_0^2E/3k_BT$ — equation (2). The cubic correction is down by $u^2/15 \sim 10^{-7}$ at laboratory fields: the linear formula is not an approximation you should ever hesitate over in statics. Why 1/3 and not 1: even at favorable Boltzmann odds, a dipole spends most of its time pointing substantially off-axis — the sphere of orientations has most of its area near the equator. The 1/3 is the same geometric dilution that makes $\langle\cos^2\theta\rangle = 1/3$ for random orientations; one factor of the projection is spent just living on a sphere. Verified: L(u) landmarks recomputed (0.0333, 0.3130, 0.6716, 0.9000 at u = 0.1, 1, 3, 10); the u/3 slope against direct numerical integration; the identical average with identical algebra gives paramagnetism (Langevin 1905), with mu B in place of p0 E — one function, two fields of physics.

Figure

The Langevin function — a long linear ramp and a distant plateau

Go deeper · Derivation B · 2 minThe Curie law, the Debye separation, and where the classical theory ends

Derivation BThe Curie law, the Debye separation, and where the classical theory ends

The goal is threefold: turn the linear Langevin result into the measurable susceptibility with its $1/T$ signature, show how one temperature sweep separates orientational from induced response (Debye's method, still the workhorse of molecular-moment measurement), and mark honestly where the assumptions fail. The only new input is the census: $n$ independent dipoles per unit volume.

Step 1. From molecule to medium. With each molecule contributing $\langle p\rangle = p_0^2E/3k_BT$ along the field, a dilute gas of $n$ per volume polarizes to $P = np_0^2E/3k_BT$ , and since $\chi_e$ is defined by $P = \varepsilon_0\chi_eE$ (formalized next chapter-step in Linear Dielectrics, used here as a label):

\chi_{\rm orient} = \frac{n\,p_0^2}{3\,\varepsilon_0\,k_BT}.

Inverse temperature, straight from the Boltzmann denominator: the Curie law. Doubling $T$ halves the orientational response — no analogue exists for the spring-stiff induced part.

Step 2. Debye's separation. Real molecules both stretch and lean; the responses add:

\alpha_{\rm total}(T) = \alpha_{\rm ind} + \frac{p_0^2}{3k_B}\cdot\frac{1}{T}.

Measured $\alpha_{\rm total}$ (from gas-phase $\kappa$ via a capacitance bridge) plotted against $1/T$ is a straight line: intercept $= \alpha_{\rm ind}$ , slope $= p_0^2/3k_B$ . From the slope, the molecular dipole moment — this is where "water: 1.85 D" and every other tabulated $p_0$ historically came from. A nonpolar molecule betrays itself by a flat line; a polar one by its tilt.

Step 3. The fine print, honestly. Three assumptions bound the theory. Classical orientations: molecular rotation is quantized, but at room temperature rotational level spacing ( $\sim$ meV) is far below $k_BT$ , so the classical continuum is excellent — the quantum correction matters only for light molecules at cryogenic temperatures. Independence: each dipole must feel only the applied field. In dense polar liquids, neighbors' fields and hydrogen bonding violate this badly — water's $\kappa = 80$ is not $1 + \chi_{\rm orient}$ with gas-phase numbers (that formula, over-applied, would confidently predict nonsense; the local-field and association physics is deferred to Local Field and the Lorentz Cavity and beyond). Mobility: the dipoles must be free to rotate. Freeze the sample and orientational response switches off almost discontinuously — ice at $-10\,^\circ$ C has a static $\kappa$ near 100 at low frequency (protons still hop) but loses its orientational response entirely at microwave frequencies, which is why you can microwave-thaw food only from the liquid film outward.

Step 4. Also honestly: saturation exists. At $u \gtrsim 1$ the linear formula overshoots ( $L(1) = 0.313$ , not $0.333$ ), and complete alignment $P \to np_0$ is approached only as $1 - 1/u$ . For liquid-water density, full alignment would mean $P_{\rm sat} = np_0 = 0.21\,\mathrm{C/m^2}$ — a fantastical polarization corresponding to a bound-charge sheet of $0.21\,\mathrm{C}$ per square meter. Static fields never get there ( $u = 1$ needs $6.6\times10^8\,\mathrm{V/m}$ ); intense laser pulses and the permanent internal fields of ferroelectric crystals (where dipole-dipole interaction does the aligning spontaneously below a Curie temperature — the chapter's Ferroelectrics article) are the places saturation physics actually lives. Verified: chi for steam at 1 atm, 373 K computed in worked example 1 (0.0055, vs measured kappa − 1 = 0.0059 — 7% low, the expected quality for an independent-dipole model of a hydrogen-bonding gas); the Debye-plot logic matches Griffiths §4.1.3 and Debye's Polar Molecules; saturation field recomputed.

Limiting cases

Case	Result	Meaning
$u \ll 1$ (all static lab fields)	$L \approx u/3$ , $\chi = np_0^2/3\varepsilon_0k_BT$	linear response; alignment $\sim0.05\%$ at 1 MV/m
$u \gg 1$	$L \to 1 - 1/u$ , $P \to np_0$	saturation; needs $\sim7\times10^8$ V/m at 300 K
high $T$	$\chi_{\rm orient} \propto 1/T \to 0$	chaos wins; only induced response survives
low $T$ (still fluid)	$\chi_{\rm orient}$ grows	Curie law — until the sample freezes and rotation locks
optical frequency	orientational response drops out	molecules can't follow $10^{15}$ Hz: water $\kappa$ 80 → 1.77
nonpolar molecule ( $p_0 = 0$ )	orientational term absent	flat Debye plot; induced mechanism only
dense polar liquid	independent-dipole formula fails	local fields + H-bonds; water's 80 needs more machinery

Common trapKappa = 80 does not mean water's dipoles are aligned — they are 99.95 percent random

The seductive misreading of a big dielectric constant is a picture of neatly ranked molecules. Reality at 1 MV/m: $u \approx 10^{-3}$ , mean alignment $u/3 \approx 0.05\%$ — orientations remain almost perfectly isotropic, and the enormous response is the census multiplying a whisper. The error has teeth in both directions: it makes saturation look nearby (it is a factor of several hundred away in field strength), and it makes the linear regime look fragile (it is bulletproof — corrections enter at $u^2/15 \sim 10^{-7}$ ). A companion trap: do not compute water's $\kappa = 80$ from the dilute-gas formula — liquid water's dipoles are neither independent nor feeling the applied field alone, and the honest agreement you can get (steam, worked example 1) becomes confident nonsense when extrapolated to the liquid.

Worked examples

1 · The dielectric constant of steam, from first principles (confidence builder)

Water vapor at $1.0\,\mathrm{atm}$ and $T = 373\,\mathrm{K}$ : predict the orientational susceptibility and compare with the measured $\kappa = 1.0059$ . Use $p_0 = 1.85\,\mathrm{D} = 6.17\times10^{-30}\,\mathrm{C\cdot m}$ .

Step 1. Number density from the ideal-gas law: $n = P/k_BT = 101325/(1.381\times10^{-23}\times373) = 1.97\times10^{25}\,\mathrm{m^{-3}}$ .

Step 2. Check the regime: at any sensible measuring field, $u = p_0E/k_BT \sim 10^{-3} \ll 1$ — linear zone, formula (2) applies.

Step 3. Susceptibility: $\chi = np_0^2/3\varepsilon_0k_BT = (1.97\times10^{25})(6.17\times10^{-30})^2/[3(8.854\times10^{-12})(1.381\times10^{-23})(373)] = 5.5\times10^{-3}$ .

Step 4. Compare: predicted $\kappa = 1.0055$ versus measured $1.0059$ — agreement to 7%, from nothing but the molecular dipole moment and a Boltzmann average. (The shortfall is real physics: the neglected induced part contributes its small share, and steam molecules are not perfectly independent.) Run backwards, this is Debye's method: the measured $\kappa(T)$ of a vapor determines $p_0$ — molecular geometry from a capacitor.

2 · Orientational versus induced — who wins, and when? (exam level)

For a single water molecule at $T = 293\,\mathrm{K}$ , compare the orientational polarizability $p_0^2/3k_BT$ with the electronic $\alpha_{\rm ind} = 4\pi\varepsilon_0\times1.45\,\text{Å}^3$ ; find the alignment fraction at air's breakdown field; and locate the (fictitious) temperature where the mechanisms would cross.

Step 1. Orientational: $\alpha_{\rm orient} = (6.17\times10^{-30})^2/[3(1.381\times10^{-23})(293)] = 3.1\times10^{-39}\,\mathrm{F\cdot m^2}$ .

Step 2. Induced: $\alpha_{\rm ind} = 4\pi(8.854\times10^{-12})(1.45\times10^{-30}) = 1.6\times10^{-40}\,\mathrm{F\cdot m^2}$ . Ratio: $\alpha_{\rm orient}/\alpha_{\rm ind} = 19$ . The lean beats the stretch nineteen to one — and this is per molecule, before the liquid's cooperative effects widen the gap.

Step 3. Alignment at breakdown ( $E = 3\times10^6\,\mathrm{V/m}$ ): $u = p_0E/k_BT = (6.17\times10^{-30})(3\times10^6)/(4.05\times10^{-21}) = 4.5\times10^{-3}$ , so $\langle\cos\theta\rangle = u/3 = 0.15\%$ . The strongest field air tolerates aligns one water molecule in seven hundred.

Step 4. Crossover temperature from $p_0^2/3k_BT^* = \alpha_{\rm ind}$ : $T^* = p_0^2/(3k_B\alpha_{\rm ind}) = (6.17\times10^{-30})^2/[3(1.381\times10^{-23})(1.6\times10^{-40})] \approx 5700\,\mathrm{K}$ . Water would need to be solar-surface hot before stretching catches leaning — i.e., for any temperature at which water exists, orientation rules. The practical moral: polar molecules are orientational responders, full stop; the induced part is a correction — except at frequencies too high for rotation, where it is suddenly all that's left.

Going further

References. Griffiths, Introduction to Electrodynamics (4th ed.), §4.1.3 (Polarization, polar molecules) · Jackson, Classical Electrodynamics (3rd ed.), §4.5–4.6 (Molecular polarizability, Clausius–Mossotti, Debye) · Purcell & Morin, Electricity and Magnetism (3rd ed.), Ch. 10 · Zangwill, Modern Electrodynamics, Ch. 6 (Dielectric matter) · Debye, Polar Molecules (1929).

Provenance. Equations and the Langevin/Debye treatment follow Griffiths §4.1.3, Jackson §4.5–4.6, and Debye's Polar Molecules. A verifier should recompute: the Langevin integral $\langle\cos\theta\rangle=\coth u-1/u$ ; its small- $u$ expansion $L(u)\approx u/3$ ; the orientational polarizability $p_0^2/3k_BT$ and susceptibility $np_0^2/3\varepsilon_0 k_BT$ ; and the numerical estimate $u\approx8\times10^{-4}$ using $1\text{ D}=3.336\times10^{-30}$ C·m, $k_B=1.381\times10^{-23}$ J/K. Dipole moments (water 1.85 D, HCl 1.08 D) and water's $\varepsilon_r\approx80$ are standard textbook values. Worked numbers recomputed for this edition: steam at 1 atm/373 K — $n = 1.97\times10^{25}$ m⁻³, $\chi = 5.5\times10^{-3}$ vs measured 0.0059; water at 293 K — $\alpha_{\rm orient} = 3.1\times10^{-39}$ F·m², ratio to induced 19, alignment 0.15% at 3 MV/m, crossover $T^* \approx 5700$ K, saturation field $6.6\times10^8$ V/m, $P_{\rm sat} = 0.21$ C/m².

Builds on: Electric Dipole Moment · Torque and Force on a Dipole in a Field · Induced Dipoles and Atomic Polarizability · Dielectric Materials and Insulators

Leads to: Polarization Density P · Linear Dielectrics, Susceptibility and Permittivity · Local Field and the Lorentz Cavity · Ferroelectrics and Spontaneous Polarization

Why does the average involve cos theta and not just count aligned versus anti-aligned molecules?

Because a dipole in three dimensions can point anywhere on a sphere, not just up or down, and its contribution to the polarization is its projection $p_0\cos\theta$ along the field. The Boltzmann weight biases the whole continuous distribution, and the sphere's geometry — most solid angle lives near the equator — is what dilutes the response by the famous factor of 3. A genuinely two-state dipole (some defects in solids behave this way) gives $\tanh u$ instead of $L(u)$ : same physics, steeper by that geometric factor at small $u$ .

Where exactly does the torque N = p cross E enter, if the derivation only used energy?

Through the energy itself: $U = -p_0E\cos\theta$ is the potential whose gradient in $\theta$ is the torque, both derived in Torque and Force on a Dipole in a Field. Statistical mechanics then needs only $U(\theta)$ — equilibrium doesn't care about the dynamics that enforce it. The torque earns its explicit keep in the dynamics: how fast the distribution relaxes toward the Boltzmann form (the Debye relaxation time, torque against rotational friction), which sets the gigahertz response and your microwave oven's operating frequency, but never changes the equilibrium average.

Is the classical treatment of orientations legitimate for a quantum molecule?

Remarkably, yes, at ordinary temperatures. Molecular rotation is quantized with level spacings of order $10^{-4}$ – $10^{-3}$ eV, well below $k_BT \approx 0.025$ eV at room temperature, so many rotational states are populated and the classical continuum average is an excellent approximation — the quantum and classical results merge exactly in that limit (the correspondence principle doing honest work). The classical picture fails for light rotors at low temperature (H $_2$ below ~100 K) and for the tunneling regimes of trapped dipoles in solids, where discrete-level statistics replace $L(u)$ .

If thermal motion opposes alignment, why does cooling a polar liquid eventually reduce its response?

Because the Curie law needs mobile dipoles, and cooling eventually freezes rotation. Above freezing, $\chi_{\rm orient}$ indeed grows as $T$ drops. At the phase transition the molecules lock into a lattice: in ordinary ice the dipoles can still rearrange sluggishly via proton hopping (giving ice a surprisingly large but slow static response), while at microwave frequencies the orientational channel is effectively dead — frozen food is nearly transparent at 2.45 GHz. In some materials cooling does something more interesting: dipole-dipole interactions take over and align the dipoles spontaneously below a Curie temperature — ferroelectricity, this chapter's later business.

Why does the steam prediction land within 7 percent but the liquid-water prediction fails by orders of magnitude?

The formula assumes each dipole responds independently to the applied field. In steam at 1 atm, molecules are ~30 Å apart and interact weakly — the assumption is decent, hence the 7%. In the liquid, each molecule sits 3 Å from hydrogen-bonded neighbors whose fields rival the applied one; the "field at the molecule" is not the applied field (the local-field problem), and clusters of correlated dipoles respond collectively (Kirkwood's correlation factor). Getting from molecular $p_0$ to liquid water's 80 took Onsager (1936) and Kirkwood (1939), and the machinery starts in Local Field and the Lorentz Cavity. Knowing which regime you're in — dilute and honest versus dense and correlated — is half of dielectric competence.

Can a strong field permanently align the dipoles and leave the material polarized after switching off?

Not in a fluid — remove the field and collisions erase the bias within picoseconds (the Debye relaxation time). But freeze or solidify the material while the field is on and you can trap the bias: that is an electret, the electrostatic analogue of a permanent magnet, made by cooling molten polar wax or corona-charging polymers under field. Electret microphones in phones and laptops hold trapped polarization for decades. The equilibrium theory of this article says the polarization wants to decay; the electret's trick is kinetic — the rotational relaxation time in the solid is effectively infinite.

Does the saturation regime ever matter in practice if static fields cannot reach u of order 1?

Yes, three ways. Cryogenics: $u \propto 1/T$ , so polar molecules in helium-temperature matrices or on cold surfaces reach substantial alignment at modest fields. Laser fields: a $10^{16}\,\mathrm{W/m^2}$ pulse carries fields of order $10^9\,\mathrm{V/m}$ — molecules can be transiently "brute-force aligned," a working technique in ultrafast molecular physics (though at optical frequencies the alignment follows $\langle E^2\rangle$ via the polarizability anisotropy rather than $L(u)$ directly). And internal fields: inside ferroelectric crystals the effective aligning fields from neighboring dipoles are of exactly this magnitude, which is why those materials sit in the saturated, spontaneously ordered regime with no external field at all.