Electronic Heat Capacity of Metals

Heat a metal and its conduction electrons barely warm up. Classical physics says they should: a gas of $N$ free electrons ought to carry $\tfrac32 N k_B$ of heat capacity by equipartition, comparable to the entire crystal lattice. Yet the measured electronic contribution at room temperature is smaller by roughly two orders of magnitude, and — the decisive clue — it vanishes linearly as $T\to0$ , not as a constant. Explaining that near-invisibility was one of the earliest and cleanest triumphs of Fermi-Dirac statistics. The resolution is Pauli blocking. An electron deep in the filled Fermi sea has no empty state to jump into, so it cannot absorb a quantum of thermal energy; only the thin shell within $\sim k_BT$ of the Fermi surface is free to be excited. That shell holds a fraction $\sim T/T_F$ of the electrons, and since the Fermi temperature of a metal is $T_F\sim10^4$ – $10^5\,$ K, the fraction is tiny — a few tenths of a percent at room temperature. Count the active electrons ( $\sim g(\varepsilon_F)k_BT$ ), give each an excess energy $\sim k_BT$ , and the thermal energy grows as $T^2$ ; its derivative, the heat capacity, is linear in $T$ . The precise coefficient comes from the low-temperature Sommerfeld expansion of the internal energy, owned by aq-sm-sommerfeld-expansion. The answer is strikingly simple: $C_{\mathrm{el}}=\gamma T$ with $\gamma=\tfrac{\pi^2}{3}k_B^2\,g(\varepsilon_F)$ , depending on the electron gas only through the density of states at the Fermi level, $g(\varepsilon_F)$ (owned by aq-sm-density-of-states-free-particles). This makes the humble specific heat a direct thermodynamic probe of the Fermi surface, and a deviation of the measured $\gamma$ from the free-electron value is read as a thermal effective mass $m^*/m$ carrying band-structure and correlation physics.

Key takeaways

A classical electron gas would add $\tfrac32 N k_B$ to a metal's heat capacity; experiment finds an electronic term that is $\sim100\times$ smaller and vanishes linearly in $T$ .
Pauli exclusion freezes out all but a thermal shell of width $\sim k_BT$ at the Fermi surface, holding a fraction $\sim T/T_F$ of the electrons.
The Sommerfeld expansion turns this into $U-U_0\propto T^2$ , so $C_{\mathrm{el}}=\partial U/\partial T=\gamma T$ is linear in temperature.
The Sommerfeld coefficient is $\gamma=\tfrac{\pi^2}{3}k_B^2\,g(\varepsilon_F)$ ; for a free gas $\gamma=\tfrac{\pi^2}{2}\,N k_B/T_F$ , suppressed below the classical value by $\sim T/T_F$ .
With the Debye phonon term, the low- $T$ law is $C=\gamma T+\beta T^3$ , so a plot of $C/T$ versus $T^2$ is a straight line whose intercept gives $\gamma$ and slope gives the Debye temperature.
Key idea: only the $\sim g(\varepsilon_F)k_BT$ electrons in a shell of width $k_BT$ at the Fermi surface can absorb heat, each gaining $\sim k_BT$ , so $U-U_0\propto T^2$ and $C_{\mathrm{el}}=\gamma T$ with $\gamma=\tfrac{\pi^2}{3}k_B^2 g(\varepsilon_F)$ — the electronic heat capacity is a thermometer for the Fermi-surface density of states.

The core insight

Raising the temperature can only promote electrons from just below the Fermi surface to just above it; states deep in the sea are Pauli-blocked. The number of mobile electrons is therefore $\sim g(\varepsilon_F)k_BT$ , not $N$ , and each gains an energy $\sim k_BT$ . The excess energy $U-U_0\sim g(\varepsilon_F)(k_BT)^2$ is quadratic in $T$ , so the heat capacity $C=\partial U/\partial T=\gamma T$ is linear, with $\gamma=\tfrac{\pi^2}{3}k_B^2 g(\varepsilon_F)$ fixed entirely by the density of states at $\varepsilon_F$ .

Where this lives today

Low-temperature calorimetry — plotting $C/T$ against $T^2$ to read off $\gamma$ and $\Theta_D$ — remains a first-line characterization tool for any new conductor, superconductor, or correlated material. The Sommerfeld coefficient is the experimental handle on the Fermi-surface density of states and, through the ratio $\gamma_{\text{exp}}/\gamma_{\text{free}}=m^*/m$ , on the thermal effective mass. In heavy-fermion compounds such as $\mathrm{CeAl_3}$ and $\mathrm{UPt_3}$ , $\gamma$ is enhanced a hundredfold or more, revealing effective masses of $10^2$ – $10^3\,m$ from electron correlations — a dramatic, purely thermodynamic signature. In superconductors the jump in $C/T$ at $T_c$ above the normal-state $\gamma$ baseline measures the condensation energy and tests BCS predictions, while the low- $T$ behavior of the electronic term distinguishes fully gapped from nodal (line- or point-node) order parameters. The same $g(\varepsilon_F)$ that sets $\gamma$ also fixes the Pauli paramagnetic susceptibility, and their ratio (the Wilson ratio) probes spin correlations.

The physical picture

The whole effect lives in a thin band at the Fermi energy. Think of the Fermi-Dirac occupation (owned by aq-sm-fermi-dirac-distribution) as a step at $\varepsilon_F$ that heating rounds off over a width of about $k_BT$ . Only electrons inside that rounded region change their occupation: an electron far below $\varepsilon_F$ would need to land on an already-filled level, which Pauli exclusion forbids, and there are no electrons far above to begin with. So although the metal contains $\sim10^{23}$ conduction electrons, only the fraction within one thermal energy of the surface — a sliver of order $T/T_F$ — is thermodynamically alive. The first figure is the hero: the active shell, drawn as the derivative $-\partial\langle n\rangle/\partial\varepsilon$ , a narrow bell of width $k_BT$ centered on $\varepsilon_F$ . Multiply its area ( $\sim g(\varepsilon_F)k_BT$ electrons) by the energy each gains ( $\sim k_BT$ ) and you already have $U-U_0\sim g(\varepsilon_F)(k_BT)^2$ — the $T^2$ that makes $C$ linear.

Figure

Only a shell of width kT at the Fermi surface is thermally active

Count electrons in this bell ( $\sim g(\varepsilon_F)k_BT$ ), give each $\sim k_BT$ , and the thermal energy is $\sim g(\varepsilon_F)(k_BT)^2$ — quadratic in $T$ , so $C\propto T$ .

The punchline of the article is a contrast, not a single curve: the measured electronic heat capacity is a faint linear ramp hugging the temperature axis, while the classical equipartition prediction is a flat plateau at $\tfrac32 N k_B$ sitting far above it. At room temperature the two differ by the factor $T/T_F$ , roughly a hundredfold. The second figure places both on the same axes so the suppression is visible at a glance.

Figure

The tiny linear electronic heat capacity versus the failed classical plateau

The classical plateau is what equipartition demands; the observed linear ramp is what Pauli exclusion permits. The ratio of the two is $\tfrac{\pi^2}{3}(T/T_F)$ — the entire story in one dimensionless number.

A useful corollary: because $\gamma\propto g(\varepsilon_F)$ , the heat capacity is a geometric probe. A metal whose Fermi level sits on a flat, high-density band has a large $\gamma$ ; one whose $\varepsilon_F$ falls in a density-of-states valley has a small $\gamma$ . This is why $\gamma$ is such a sensitive fingerprint of band structure and electron correlations.

The core equations

Everything begins with the Fermi-Dirac occupation, which supplies the sharp step whose thermal rounding drives the effect. Its role here is only to be differentiated: the response window $-\partial\langle n\rangle/\partial\varepsilon$ selects the active shell.

\langle n(\varepsilon)\rangle=\frac{1}{e^{(\varepsilon-\mu)/k_BT}+1} \qquad (1)

where: $\varepsilon$ — single-particle energy; $\mu$ — chemical potential (temperature dependent); $k_B$ — Boltzmann constant; $T$ — absolute temperature. Physical reading: this is a step at $\varepsilon=\mu$ smeared over a width $\sim k_BT$ ; only that smeared window changes when the metal is heated, which is the whole physical content of the linear specific heat. To turn occupations into energy we need the free-electron density of states, the count of single-particle levels per unit energy. Its value at the Fermi energy is the only property of the gas that survives into $\gamma$ .

g(\varepsilon)=\frac{V}{2\pi^2}\left(\frac{2m}{\hbar^2}\right)^{3/2}\sqrt{\varepsilon}=\frac{3N}{2\varepsilon_F}\left(\frac{\varepsilon}{\varepsilon_F}\right)^{1/2} \qquad (2)

where: $g(\varepsilon)$ — spin-summed density of states; $V$ — volume; $m$ — electron mass; $\varepsilon_F$ — Fermi energy; $N$ — number of conduction electrons. Physical reading: integrating $g$ up to $\varepsilon_F$ at $T=0$ must return $N$ , which forces the tidy identity $g(\varepsilon_F)=3N/(2\varepsilon_F)$ used everywhere below (the full $g(\varepsilon)$ is derived in aq-sm-density-of-states-free-particles; here we only need its value at the surface). The low-temperature engine is the Sommerfeld expansion, which evaluates any Fermi-weighted integral as its $T=0$ value plus a correction set by the slope of the integrand at $\mu$ . This is the machinery owned by aq-sm-sommerfeld-expansion; we quote it and use it.

\int_0^\infty H(\varepsilon)\,\langle n(\varepsilon)\rangle\,\mathrm{d}\varepsilon=\int_0^\mu H(\varepsilon)\,\mathrm{d}\varepsilon+\frac{\pi^2}{6}(k_BT)^2\,H'(\mu)+O(T^4) \qquad (3)

where: $H(\varepsilon)$ — any smooth function (we use $g$ and $\varepsilon g$ ); $H'$ — its energy derivative; the coefficient $\pi^2/6$ is the standard Sommerfeld constant (its origin, the Fermi integral $\int x^2\,\mathrm{sech}^2(x/2)/4\,\mathrm{d}x=\pi^2/3$ , halved). Physical reading: because $-\partial\langle n\rangle/\partial\varepsilon$ is a symmetric bell at $\mu$ , the leading thermal correction probes only the curvature/slope of $H$ there — nothing deep in the sea contributes. Fixing the particle number as the metal warms forces the chemical potential to drift slightly downward; the shift is second order in temperature.

\mu(T)=\varepsilon_F\left[\,1-\frac{\pi^2}{12}\left(\frac{k_BT}{\varepsilon_F}\right)^{\!2}+\cdots\right] \qquad (4)

where: $\mu(T)$ — chemical potential; $\varepsilon_F=\mu(0)$ . Physical reading: the $\sqrt{\varepsilon}$ density of states rises with energy, so the smearing promotes slightly more electrons upward than it leaves holes below; holding $N$ fixed pulls $\mu$ down by $\propto(k_BT/\varepsilon_F)^2$ (this drift is owned by aq-sm-chemical-potential-fermi-gas-temperature). Feeding $H=\varepsilon g$ into the expansion, and using the $\mu$ -drift to keep $N$ fixed, the internal energy acquires a single clean thermal term.

U(T)=U_0+\frac{\pi^2}{6}(k_BT)^2\,g(\varepsilon_F)+O(T^4) \qquad (5)

where: $U_0=\tfrac35 N\varepsilon_F$ — the zero-temperature energy of the Fermi sea (owned by aq-sm-degenerate-fermi-gas-zero-temperature); the $g'$ -dependent pieces cancel exactly. Physical reading: remarkably, only the bare density of states at $\varepsilon_F$ survives — the $\mu$ -drift contributes nothing net to the coefficient. Differentiating at fixed $N,V$ gives the central result: a heat capacity strictly linear in temperature.

C_{\mathrm{el}}=\left(\frac{\partial U}{\partial T}\right)_{N,V}=\frac{\pi^2}{3}k_B^2\,g(\varepsilon_F)\,T\equiv\gamma T \qquad (6)

where: $\gamma$ — Sommerfeld coefficient (units $\mathrm{J\,K^{-2}}$ ); the factor $\pi^2/6$ in $U$ becomes $\pi^2/3$ after $\partial/\partial T$ of a $T^2$ term. Physical reading: the heat capacity measures $g(\varepsilon_F)$ and nothing else — the cleanest thermodynamic window onto the Fermi surface. For a free-electron gas, substituting $g(\varepsilon_F)=3N/(2\varepsilon_F)$ exposes the suppression relative to the classical value.

\frac{C_{\mathrm{el}}}{N k_B}=\frac{\pi^2}{2}\,\frac{T}{T_F},\qquad k_B T_F=\varepsilon_F \qquad (7)

where: $T_F$ — Fermi temperature. Physical reading: the classical equipartition result is $\tfrac32 N k_B$ ; the quantum answer is smaller by $\tfrac{\pi^2}{3}(T/T_F)$ , which is $\sim1\%$ for a metal at room temperature — exactly the observed near-invisibility. Finally, a real metal adds the lattice (phonon) heat capacity, which is Debye's $T^3$ at low temperature. The two terms separate cleanly in a single plot.

\frac{C}{T}=\gamma+\beta T^2,\qquad \beta=\frac{12\pi^4}{5}\,\frac{N k_B}{\Theta_D^{\,3}} \qquad (8)

where: $\beta$ — Debye coefficient; $\Theta_D$ — Debye temperature (the phonon $\beta T^3$ term is owned by aq-sm-debye-model-phonons and aq-sm-debye-temperature-t-cubed-law; here it is quoted, not derived). Physical reading: plotting $C/T$ against $T^2$ yields a straight line — intercept $\gamma$ (electronic), slope $\beta$ (lattice) — the standard low-temperature calorimetry of metals.

Derivations

Derivation ALinear heat capacity from the Sommerfeld expansion of the energy

The goal is to derive $C_{\mathrm{el}}=\tfrac{\pi^2}{3}k_B^2 g(\varepsilon_F)\,T$ from first principles by expanding the internal energy to order $T^2$ at fixed particle number and differentiating. The subtle point we must handle carefully is the temperature drift of the chemical potential: naively evaluating the energy integral at fixed $\mu$ gives the wrong answer, so we track the $\mu$ -shift and watch the density-of-states-slope terms cancel. We take the Sommerfeld expansion itself as given (owned by aq-sm-sommerfeld-expansion) and the ground-state energy $U_0=\tfrac35 N\varepsilon_F$ as given (owned by aq-sm-degenerate-fermi-gas-zero-temperature); this card owns the assembly into $\gamma$ .

Step 1. Write the energy as a Fermi-weighted integral. The internal energy of the electron gas is $U=\int_0^\infty \varepsilon\,g(\varepsilon)\,\langle n(\varepsilon)\rangle\,\mathrm{d}\varepsilon$ , so we apply the Sommerfeld formula with $H(\varepsilon)=\varepsilon g(\varepsilon)$ . Why: the total energy is the sum over occupied levels of energy times occupation; converting the sum to an integral with $g(\varepsilon)$ is the standard step.

Step 2. Apply the expansion. With $H=\varepsilon g$ , one has $H'(\mu)=g(\mu)+\mu g'(\mu)$ , so $U=\int_0^\mu \varepsilon g\,\mathrm{d}\varepsilon+\tfrac{\pi^2}{6}(k_BT)^2\big[g(\mu)+\mu g'(\mu)\big]$ . Why: the Sommerfeld correction is proportional to the derivative of the integrand at $\mu$ , by Eq. (3).

Step 3. Split the leading integral about $\varepsilon_F$ . Since $\mu$ is close to $\varepsilon_F$ , write $\int_0^\mu \varepsilon g\,\mathrm{d}\varepsilon=\int_0^{\varepsilon_F}\varepsilon g\,\mathrm{d}\varepsilon+\varepsilon_F g(\varepsilon_F)(\mu-\varepsilon_F)$ ; the first term is $U_0$ and the boundary term uses the integrand value $\varepsilon g\big|_{\varepsilon_F}=\varepsilon_F g(\varepsilon_F)$ . Why: to order $T^2$ we only need the integral to linear order in the small shift $\mu-\varepsilon_F$ , and $\mu-\varepsilon_F$ is already $O(T^2)$ .

Step 4. Insert the chemical-potential shift. Fixing $N$ gives (from the number integral) $g(\varepsilon_F)(\mu-\varepsilon_F)=-\tfrac{\pi^2}{6}(k_BT)^2 g'(\varepsilon_F)$ , so the boundary term becomes $\varepsilon_F g(\varepsilon_F)(\mu-\varepsilon_F)=-\tfrac{\pi^2}{6}(k_BT)^2\,\varepsilon_F g'(\varepsilon_F)$ . Why: the same Sommerfeld expansion applied to $N=\int g\langle n\rangle\,\mathrm{d}\varepsilon$ pins the $\mu$ -drift that keeps the electron number constant.

Step 5. Collect the $T^2$ terms. Adding the boundary term to the Sommerfeld correction, $U=U_0-\tfrac{\pi^2}{6}(k_BT)^2\varepsilon_F g'(\varepsilon_F)+\tfrac{\pi^2}{6}(k_BT)^2\big[g(\varepsilon_F)+\varepsilon_F g'(\varepsilon_F)\big]$ . The two $\varepsilon_F g'(\varepsilon_F)$ pieces cancel identically, leaving $U=U_0+\tfrac{\pi^2}{6}(k_BT)^2 g(\varepsilon_F)$ . Why: this cancellation is exactly why the $\mu$ -drift does not change the coefficient — only the bare $g(\varepsilon_F)$ survives, a robust feature of any smooth density of states.

Step 6. Differentiate at fixed $N,V$ . Then $C_{\mathrm{el}}=\partial U/\partial T=\tfrac{\pi^2}{3}k_B^2 g(\varepsilon_F)\,T\equiv\gamma T$ , linear in temperature, with $\gamma=\tfrac{\pi^2}{3}k_B^2 g(\varepsilon_F)$ . Why: differentiating a term $\propto T^2$ produces the factor $2$ that turns $\pi^2/6$ into $\pi^2/3$ and gives the single power of $T$ . Verified: the $\pi^2/6$ Sommerfeld constant (from $\int x^2/[4\cosh^2(x/2)]\,\mathrm{d}x=\pi^2/3$ ), the exact cancellation of the $g'(\varepsilon_F)$ terms, and $C_{\mathrm{el}}=\tfrac{\pi^2}{3}k_B^2 g(\varepsilon_F)T$ are checked symbolically in verify_electronic-heat-capacity-metals.py.

Go deeper · Derivation B · 1 minFree-electron form and the classical-suppression factor

Derivation BFree-electron form and the classical-suppression factor

The goal is to convert the general result $\gamma=\tfrac{\pi^2}{3}k_B^2 g(\varepsilon_F)$ into the free-electron expression $\gamma=\tfrac{\pi^2}{2}N k_B/T_F$ and to show explicitly why the electronic heat capacity is invisible next to the classical equipartition value. This card owns the free-gas specialization and the dimensionless suppression ratio; it uses the identity $g(\varepsilon_F)=3N/(2\varepsilon_F)$ , which itself follows from requiring the $\sqrt{\varepsilon}$ density of states to integrate to $N$ at $T=0$ .

Step 1. Evaluate the Fermi-surface density of states. Integrating Eq. (2), $N=\int_0^{\varepsilon_F} g(\varepsilon)\,\mathrm{d}\varepsilon=\tfrac23 g(\varepsilon_F)\varepsilon_F$ , so $g(\varepsilon_F)=3N/(2\varepsilon_F)$ . Why: the $T=0$ sea is filled up to $\varepsilon_F$ , and for a $\sqrt{\varepsilon}$ DOS two-thirds of the "height" $g(\varepsilon_F)\varepsilon_F$ equals the count $N$ .

Step 2. Substitute into $\gamma$ . Then $\gamma=\tfrac{\pi^2}{3}k_B^2\cdot\tfrac{3N}{2\varepsilon_F}=\tfrac{\pi^2}{2}\dfrac{N k_B^2}{\varepsilon_F}=\tfrac{\pi^2}{2}\dfrac{N k_B}{T_F}$ , using $\varepsilon_F=k_B T_F$ . Why: replacing $g(\varepsilon_F)$ by its free-electron value trades the microscopic DOS for the two measurable numbers $N$ and $T_F$ .

Step 3. Form the dimensionless heat capacity. Dividing by $N k_B$ , $C_{\mathrm{el}}/(N k_B)=\tfrac{\pi^2}{2}(T/T_F)$ . Why: this isolates the temperature dependence and shows the natural expansion parameter is $T/T_F$ , small for any metal below melting.

Step 4. Compare to classical equipartition. The classical gas would give $C_{\mathrm{el}}^{\text{cl}}/(N k_B)=\tfrac32$ , so the ratio of quantum to classical is $\dfrac{C_{\mathrm{el}}}{\tfrac32 N k_B}=\dfrac{\pi^2}{3}\dfrac{T}{T_F}$ . Why: the quantum result is suppressed by the fraction of electrons that escape Pauli blocking, $\sim T/T_F$ ; the factor $\pi^2/3$ is the exact coefficient, and for a metal with $T_F\approx8\times10^4\,$ K this ratio is about $1\%$ at $300\,$ K. Verified: $g(\varepsilon_F)=3N/(2\varepsilon_F)$ , $\gamma=\tfrac{\pi^2}{2}N k_B/T_F$ , $C_{\mathrm{el}}/(N k_B)=\tfrac{\pi^2}{2}(T/T_F)$ , and the suppression ratio $\tfrac{\pi^2}{3}(T/T_F)$ are checked symbolically, with the copper numbers checked numerically, in verify_electronic-heat-capacity-metals.py.

Limiting cases

The linear law is the leading term of a degenerate expansion; its validity and crossovers are:

Case	Result	Meaning
Deep degeneracy $T\ll T_F$	$C_{\mathrm{el}}=\gamma T,\ \ \gamma=\tfrac{\pi^2}{3}k_B^2 g(\varepsilon_F)$	The Sommerfeld regime. Metals have $T_F\sim10^4$ – $10^5\,$ K, so $T/T_F\sim10^{-2}$ already at room temperature.
Classical limit $T\gg T_F$	$C_{\mathrm{el}}\to\tfrac32 N k_B$	Maxwell-Boltzmann equipartition — never reached, since metals melt far below $T_F$ .
With phonons, low $T$	$C=\gamma T+\beta T^3$	Electronic term dominates below the crossover $T^\star=\sqrt{\gamma/\beta}$ , typically a few kelvin.
Free-electron value	$\gamma=\tfrac{\pi^2}{2}\,N k_B/T_F$	Uses $g(\varepsilon_F)=3N/(2\varepsilon_F)$ ; measured $\gamma$ deviates by $m^*/m$ .

Common trapThe heat capacity is taken at fixed particle number, not fixed chemical potential

The most common error is to differentiate the energy at fixed $\mu$ and forget that $\mu(T)$ itself drifts. Because the number of electrons is conserved, warming the metal pulls the chemical potential down as $\mu(T)=\varepsilon_F[1-\tfrac{\pi^2}{12}(k_BT/\varepsilon_F)^2+\cdots]$ . That shift feeds a term $\propto g'(\varepsilon_F)$ into the energy — but it is exactly cancelled by a matching $\varepsilon_F g'(\varepsilon_F)$ piece from the energy integral, so only the bare $g(\varepsilon_F)$ survives in $\gamma$ . Omitting either piece gives a wrong coefficient. A second trap: $C_{\mathrm{el}}=\gamma T$ is the electronic heat capacity; the total measured $C$ also contains the phonon $\beta T^3$ , and comparing raw $C$ to $\gamma T$ without subtracting the lattice term will overstate $\gamma$ .

Worked examples

Example 1 — Count the active electrons and estimate the suppression.

Step 1. Setup. Take copper: one conduction electron per atom, Fermi temperature $T_F=8.2\times10^4\,$ K (from $\varepsilon_F=7.0\,$ eV, since $\varepsilon_F/k_B=7.0\times1.602\times10^{-19}/1.381\times10^{-23}$ ). Ask what fraction of electrons is thermally active at room temperature, $T=300\,$ K.

Step 2. Active fraction. Only electrons within $\sim k_BT$ of $\varepsilon_F$ participate, a fraction $\sim g(\varepsilon_F)k_BT/N=(3/2)(k_BT/\varepsilon_F)=(3/2)(T/T_F)$ . Numerically $(3/2)(300/8.2\times10^4)=\mathbf{5.5\times10^{-3}}$ — about half a percent.

Step 3. Suppression of the heat capacity. The exact ratio to the classical value is $C_{\mathrm{el}}/(\tfrac32 N k_B)=\tfrac{\pi^2}{3}(T/T_F)=\tfrac{\pi^2}{3}(300/8.2\times10^4)=\mathbf{0.0120}$ . So the electronic heat capacity is about $1.2\%$ of the classical equipartition value.

Step 4. Interpret. Classical physics overestimates the electronic heat capacity by a factor $1/0.0120\approx\mathbf{83\times}$ . This is precisely the "missing" specific heat that puzzled pre-quantum physicists: the electrons are there, but Pauli exclusion locks all but a thin surface shell out of the thermal budget.

Example 2 — Extract $\gamma$ and $\Theta_D$ from low-temperature calorimetry.

Step 1. Setup. Suppose measurements of a metal give $C/T=0.70+2.6\,T^2$ in $\mathrm{mJ\,mol^{-1}K^{-2}}$ (with $T$ in kelvin), i.e. a straight line in $C/T$ versus $T^2$ . Identify the electronic and lattice parameters.

Step 2. Read off the coefficients. The intercept is the Sommerfeld coefficient, $\gamma=\mathbf{0.70\ mJ\,mol^{-1}K^{-2}}$ ; the slope is the Debye coefficient, $\beta=\mathbf{2.6\ mJ\,mol^{-1}K^{-4}}$ . These are the copper-scale values (free-electron $\gamma\approx0.50$ ; measured $\approx0.69$ ).

Step 3. Debye temperature. Invert $\beta=\tfrac{12\pi^4}{5}R/\Theta_D^3$ with $R=8.314\,\mathrm{J\,mol^{-1}K^{-1}}=8314\,\mathrm{mJ\,mol^{-1}K^{-1}}$ : $\Theta_D^3=\tfrac{12\pi^4}{5}R/\beta=233.8\times8314/2.6=7.48\times10^5\,\mathrm{K^3}$ , so $\Theta_D=(7.48\times10^5)^{1/3}=\mathbf{331\ K}$ — a typical metallic Debye temperature.

Step 4. Crossover temperature. The electronic and lattice terms are equal at $\gamma T^\star=\beta T^{\star3}$ , i.e. $T^\star=\sqrt{\gamma/\beta}=\sqrt{0.70/2.6}=\mathbf{0.52\ K}$ . Below about half a kelvin the linear electronic term dominates the specific heat; above it the phonon $T^3$ takes over. This is why $\gamma$ measurements demand sub-kelvin cryogenics.

Origins · 1 minelectronic heat capacity and how this was discovered

electronic heat capacity

puzzle resolved 1927–1928 · The "missing" electronic specific heat was a genuine crisis for the classical Drude theory: a gas of conduction electrons should have added a huge (3/2)Nk_B to every metal's heat capacity, yet it was never seen. Arnold Sommerfeld resolved it in 1927–1928 by treating the electrons as a degenerate Fermi-Dirac gas: Pauli exclusion freezes out all but a thermal shell at the Fermi surface, leaving only the tiny linear-in-T term γT. It was one of the first quantitative victories of Fermi-Dirac statistics in the solid state, and it made the Sommerfeld coefficient γ a permanent diagnostic of the Fermi-surface density of states.

History

1900

— Drude models metals as a classical electron gas but cannot explain the missing electronic heat capacity

1926

— Fermi and, independently, Dirac formulate the statistics of exclusion-obeying particles

1927

— Pauli applies Fermi-Dirac statistics to the electron gas, explaining its weak paramagnetism

1928

— Sommerfeld derives the linear electronic heat capacity, resolving the classical specific-heat crisis

1928

— Bloch extends the free-electron gas to periodic lattices, launching band theory

1930

— de Haas–van Alphen and related low-temperature measurements begin mapping Fermi surfaces

1975

— heavy-fermion compounds are found with γ enhanced by factors of hundreds, signaling huge effective masses

Going further

Texts: Ashcroft & Mermin, Solid State Physics, Ch. 2 (Sommerfeld theory of metals) and Ch. 23 (Debye term); Pathria & Beale, Statistical Mechanics (3rd ed.), §8.1; Kardar, Statistical Physics of Particles, Ch. 7; Reif, Fundamentals of Statistical and Thermal Physics, §9.16–9.17; Kittel, Introduction to Solid State Physics, Ch. 6.
Builds on: Sommerfeld Expansion (owns the $\pi^2/6$ machinery used here) · Degenerate Fermi Gas at Zero Temperature (owns $U_0=\tfrac35 N\varepsilon_F$ ) · Density of States for Free Particles (owns $g(\varepsilon)$ ) · Fermi Energy, Momentum and Fermi Temperature · Chemical Potential of a Fermi Gas vs Temperature (owns the $\mu(T)$ drift).
Leads to: Degeneracy Pressure of Fermions · Fermi Surface and Fermi Sphere · Photon Gas and Planck Radiation Law; and, for the lattice term, the Debye Model of Phonons and the Debye Temperature and T-cubed Law (the $\beta T^3$ quoted here).
Provenance: AllQFT `sm` export (Pathria & Beale, Ch. 8), aq-sm-electronic-heat-capacity-metals, ref STAT.IQGAS.ECVM.1223. All coefficients ( $\pi^2/6\to\pi^2/3$ , $g(\varepsilon_F)=3N/2\varepsilon_F$ , $\gamma=\tfrac{\pi^2}{2}N k_B/T_F$ , the $g'$ cancellation, the $\tfrac{\pi^2}{3}(T/T_F)$ suppression) and the copper/Debye numbers are python-verified (22 checks, all PASS).

Why don't the conduction electrons obey the Dulong–Petit / equipartition rule?

Equipartition assumes every degree of freedom can freely absorb $\tfrac12 k_BT$ , but that requires empty states nearby. In a degenerate electron gas the states below $\varepsilon_F$ are full and Pauli exclusion forbids an electron there from accepting a small energy — there is nowhere to go. Only electrons within $\sim k_BT$ of the Fermi surface have empty neighbours, a fraction $\sim T/T_F\sim1\%$ . So instead of $\tfrac32 N k_B$ you get roughly $\tfrac32 N k_B\times(T/T_F)$ , linear in $T$ and about a hundred times smaller at room temperature.

Where does the factor π²/3 actually come from?

It descends from the Sommerfeld expansion (owned by aq-sm-sommerfeld-expansion). Any Fermi-weighted integral equals its $T=0$ value plus $\tfrac{\pi^2}{6}(k_BT)^2 H'(\mu)$ ; the $\pi^2/6$ is the standard Fermi integral $\int_{-\infty}^{\infty}x^2/[4\cosh^2(x/2)]\,\mathrm{d}x=\pi^2/3$ , halved. The energy inherits $U-U_0=\tfrac{\pi^2}{6}(k_BT)^2 g(\varepsilon_F)$ , and differentiating the $T^2$ brings down a factor of $2$ , turning $\pi^2/6$ into $\pi^2/3$ . So the $\pi^2/3$ in $\gamma$ is really $2\times\pi^2/6$ .

Doesn't the chemical potential change with temperature? Why doesn't that spoil the answer?

It does change: $\mu(T)=\varepsilon_F[1-\tfrac{\pi^2}{12}(k_BT/\varepsilon_F)^2+\cdots]$ , dropping because the $\sqrt{\varepsilon}$ density of states rises with energy. That drift contributes a term proportional to $g'(\varepsilon_F)$ to the internal energy. Remarkably, an equal and opposite $\varepsilon_F g'(\varepsilon_F)$ term arises when you expand the energy integral about $\varepsilon_F$ , and the two cancel exactly. The net effect is that only the bare $g(\varepsilon_F)$ enters $\gamma$ — the coefficient is insensitive to the $\mu$ -drift for any smooth density of states.

What is the Sommerfeld coefficient γ good for experimentally?

It is the cleanest thermodynamic measurement of the Fermi-surface density of states $g(\varepsilon_F)$ . Because $\gamma=\tfrac{\pi^2}{3}k_B^2 g(\varepsilon_F)$ , a low-temperature $C/T$ -vs- $T^2$ plot gives $g(\varepsilon_F)$ from the intercept. Comparing to the free-electron value yields the thermal effective mass $m^*/m=\gamma_{\text{exp}}/\gamma_{\text{free}}$ , which encodes band structure and many-body renormalizations. In heavy-fermion metals $\gamma$ is enhanced by factors of hundreds, flagging enormous effective masses; in superconductors the jump in $C/T$ at $T_c$ measures the condensation energy.

How do I actually separate the electronic and lattice contributions?

Measure $C(T)$ at low temperature and plot $C/T$ against $T^2$ . The total is $C=\gamma T+\beta T^3$ (electrons plus Debye phonons), so $C/T=\gamma+\beta T^2$ is a straight line. The intercept at $T^2=0$ is the electronic $\gamma$ ; the slope is the phonon $\beta$ , from which the Debye temperature follows via $\beta=\tfrac{12\pi^4}{5}Nk_B/\Theta_D^3$ . The two terms cross at $T^\star=\sqrt{\gamma/\beta}$ , typically below one kelvin, which is why $\gamma$ measurements need sub-kelvin cryogenics.

Is this the same physics as Pauli paramagnetism?

Very nearly — both are Fermi-surface effects governed by the same $g(\varepsilon_F)$ . In heat capacity, only the thermal shell of width $k_BT$ absorbs energy, giving $\gamma\propto g(\varepsilon_F)$ . In Pauli paramagnetism, only the shell near $\varepsilon_F$ can repopulate between spin-up and spin-down, giving a susceptibility $\chi_P=\mu_0\mu_B^2 g(\varepsilon_F)$ , again $\propto g(\varepsilon_F)$ . Their dimensionless ratio, the Wilson ratio, is a constant for free electrons and deviates when spin correlations matter — a useful cross-check on the free-electron picture.

Why does the electronic term matter only at very low temperature if it is "always there"?

It is always present, but the lattice heat capacity dwarfs it except in the cold. The phonon term scales as $\beta T^3$ and the electronic term as $\gamma T$ , so their ratio is $\gamma/(\beta T^2)$ , which grows as $T$ falls. The two are comparable only at $T^\star=\sqrt{\gamma/\beta}$ , usually a fraction of a kelvin. Above that the metal's heat capacity is essentially all phonons; below it the linear electronic term wins and dominates as $T\to0$ , since $T$ beats $T^3$ .