08.11.02 · stat-mech / quantum-gases

Debye theory of specific heats of solids

shipped3 tiersLean: none

Anchor (Master): Landau & Lifshitz, *Statistical Physics, Part 1* 3e (Pergamon, 1980), §§63-66; Ashcroft & Mermin, *Solid State Physics* (Holt, 1976), Chs. 22-23; Kittel, *Introduction to Solid State Physics* 8e (Wiley, 2005), Ch. 5

Intuition Beginner

A solid is a crowd of atoms held in place by springs. Pluck one atom and it jiggles; its neighbours jiggle in sympathy; the whole crystal supports collective vibration waves that travel through it like ripples through a stretched fabric. The waves come in countably many flavours, each labelled by a wavelength and a direction, and each flavour stores some thermal energy. The specific heat of the solid is the total rate at which energy piles into these vibrations as you warm the crystal.

Classical physics predicted one universal answer for every solid at room temperature: the Dulong-Petit value, about $25$ joules per mole per kelvin, found empirically by Dulong and Petit in 1819. Experiments confirmed this at high temperature but found a striking failure at low temperature, where the specific heat of every solid drops smoothly toward zero. Classical equipartition could not explain the drop. Einstein in 1907 gave the first quantum fix by treating each atom as a single oscillator with frequency $ω_{E}$ , but his model decayed too fast at low temperature.

Debye in 1912 supplied the right picture. The vibrations are sound waves with a continuous range of frequencies from zero up to a material-dependent cutoff $ω_{D}$ . The Bose-Einstein occupation of each wave gives an energy per mode that freezes out below its own frequency. The number of low-frequency waves grows as the cube of frequency, so the low-temperature specific heat scales as $T^{3}$ — the famous Debye law, vindicated to staggering precision in clean crystals.

Visual Beginner

Three curves on the same axes show the specific heat $C_{V} /3 N k_{B}$ as a function of temperature $T / Θ_{D}$ in dimensionless form, where $Θ_{D}$ is the material-specific Debye temperature. The Debye curve rises smoothly from zero at low temperature, follows a $T^{3}$ rule near the origin, and saturates at the Dulong-Petit value $3 N k_{B}$ for $T$ much larger than $Θ_{D}$ . The Einstein 1907 curve, drawn next to it, drops to zero exponentially fast at low temperature and disagrees with experiment in that regime. A few real materials are overlaid: aluminium ( $Θ_{D} \approx 396 K$ ), diamond ( $Θ_{D} \approx 2230 K$ ), and lead ( $Θ_{D} \approx 105 K$ ).

The picture captures three universal features: the $T^{3}$ rise from zero, the Dulong-Petit saturation, and the role of $Θ_{D}$ as the single material parameter that rescales every solid onto one common curve.

Worked example Beginner

Compute the Debye specific heat of aluminium at $T = 30 K$ and check that it follows the low-temperature $T^{3}$ scaling.

Step 1. Aluminium has a measured Debye temperature $Θ_{D} = 396 K$ (from low-temperature calorimetry). At $T = 30 K$ , the ratio $T / Θ_{D} = 30/396 \approx 0.0758$ is well below one, so the low-temperature limit applies.

Step 2. The low-temperature Debye formula is $C_{V} = (12 π^{4} /5) N k_{B} (T / Θ_{D})^{3}$ . The numerical coefficient is $12 π^{4} /5 \approx 233.78$ . With $T / Θ_{D} = 0.0758$ , the cube is $(0.0758)^{3} \approx 4.35 \times 1 0^{- 4}$ .

Step 3. Plug in. For one mole of aluminium, $N = N_{A} \approx 6.022 \times 1 0^{23}$ and $N k_{B} = R \approx 8.314 J/ (mol \cdot K)$ . The result is $C_{V} \approx 233.78 \times 8.314 \times 4.35 \times 1 0^{- 4} \approx 0.845 J/ (mol \cdot K)$ .

Step 4. Check against tabulated calorimetry. The measured molar specific heat of aluminium at $T = 30 K$ is about $0.78 J/ (mol \cdot K)$ (lattice contribution; the electronic Sommerfeld piece is tiny here). The Debye prediction $0.85$ agrees with the data to better than ten percent — the residual is the electronic contribution plus phonon-dispersion deviations that the pure Debye approximation does not capture.

Step 5. Verify the $T^{3}$ rule. At twice the temperature, $T = 60 K$ with $T / Θ_{D} = 0.1515$ , the cube is eight times larger and the Debye prediction is $C_{V} \approx 6.76 J/ (mol \cdot K)$ . Measured value at $60 K$ : about $5.8$ , again within ten percent. The doubling of $T$ raised $C_{V}$ by a factor near eight, in line with the $T^{3}$ rule.

What this tells us: a single number $Θ_{D} = 396 K$ , fitted once for aluminium, predicts the lattice specific heat across the whole low-temperature range. The Debye formula is not a fit at each temperature; it is one universal curve with one material parameter, and the prediction matches calorimetric data to within the corrections expected from electrons and from the deviations of real phonon dispersions from the linear-Debye idealisation.

Check your understanding Beginner

Exercise (easy, true-false).

True or false: the Debye temperature $Θ_{D}$ of a material can be determined from a single low-temperature specific-heat measurement and then used to predict $C_{V} (T)$ at all temperatures.

Hint

The Debye formula has one material parameter, and the universal shape of the curve is the same for every solid once temperatures are rescaled by $Θ_{D}$ .

Answer

True. Within the Debye approximation, all solids share one universal $C_{V} (T / Θ_{D})$ curve; the material parameter $Θ_{D}$ fixes the temperature scale. One measurement, say at $T = 30 K$ in the $T^{3}$ regime, pins $Θ_{D}$ , and the same number then predicts the full curve. The agreement with experiment is excellent for elemental crystals at low and high temperature; in the intermediate range, dispersion details require the Born-von Kármán refinement.

Formal definition Intermediate+

Let $M$ be a crystalline solid of $N$ atoms in a box of volume $V$ at temperature $T$ . In the harmonic approximation, atomic displacements about equilibrium are described by $3 N$ normal-mode coordinates, each behaving as an independent quantum harmonic oscillator with angular frequency $ω_{α}$ and energy spectrum $ℏ ω_{α} (n_{α} + \frac{1}{2})$ for $n_{α} \in N$ . The canonical partition function factorises: $$ Z(T) = \prod_{\alpha=1}^{3N} \frac{e^{-\hbar\omega_\alpha / 2 k_B T}}{1 - e^{-\hbar\omega_\alpha / k_B T}}, $$ and the internal energy per mode is the Bose-Einstein occupation $$ \langle \epsilon(\omega) \rangle = \frac{\hbar\omega}{2} + \frac{\hbar\omega}{e^{\hbar\omega / k_B T} - 1}. $$ Define the vibrational density of states $g (ω)$ so that $g (ω) d ω$ counts the normal modes with frequency in $[ω, ω + d ω]$ . The total mode count is $\int_{0}^{\infty} g (ω) d ω = 3 N$ .

The Debye approximation replaces the true spectrum by three branches of linear acoustic dispersion $ω (k) = c_{s} ∣ k ∣$ in an isotropic medium with sound speed $c_{s}$ , truncated at a cutoff $ω_{D}$ (the Debye frequency) chosen so that the truncated mode count matches $3 N$ . Counting modes in a box with periodic boundary conditions, the number of allowed wave vectors $∣ k ∣ < k$ is $V k^{3} /6 π^{2}$ , so per polarisation the density of states is $V ω^{2} / (2 π^{2} c_{s}^{3})$ . Summing over three polarisations, $$ g(\omega) = \frac{3 V}{2\pi^2 c_s^3}, \omega^2, \qquad 0 \leq \omega \leq \omega_D. $$ The cutoff is fixed by $\int_{0}^{ω_{D}} g (ω) d ω = 3 N$ , which gives $$ \omega_D^3 = \frac{6 \pi^2 N c_s^3}{V}, \qquad \Theta_D := \frac{\hbar \omega_D}{k_B}. $$ The dimensionless ratio $T / Θ_{D}$ controls every Debye prediction.

The internal energy (dropping the temperature-independent zero-point piece $\sum_{α} ℏ ω_{α} /2$ , which shifts $F$ by a constant) is $$ U(T) = \int_0^{\omega_D} g(\omega), \frac{\hbar\omega}{e^{\hbar\omega / k_B T} - 1}, d\omega. $$ Change variable to $x := ℏ ω / k_{B} T$ with upper limit $y := Θ_{D} / T$ : $$ U(T) = 9 N k_B T \left(\frac{T}{\Theta_D}\right)^3 \int_0^{\Theta_D / T} \frac{x^3}{e^x - 1}, dx. $$ The Debye function is $D_{3} (y) := (3/ y^{3}) \int_{0}^{y} x^{3} / (e^{x} - 1) d x$ , so $U = 3 N k_{B} T \cdot D_{3} (Θ_{D} / T)$ .

The constant-volume specific heat $C_{V} := (\partial U / \partial T)_{V}$ is obtained by differentiating under the integral sign: $$ C_V(T) = 9 N k_B \left(\frac{T}{\Theta_D}\right)^3 \int_0^{\Theta_D / T} \frac{x^4 e^x}{(e^x - 1)^2}, dx. $$ This is the Debye specific heat formula in the Sykes-Kearsley form ^{[Landau Vol 5 §64]}.

Counterexamples to common slips

The Debye cutoff is a mode-counting device, not a physical maximum-phonon-frequency. Real phonon dispersions have van Hove singularities and Brillouin-zone boundaries; the Debye cutoff $ω_{D}$ approximates the zone-boundary cutoff by demanding that the linear spectrum integrate to the correct mode count. Born-von Kármán 1912, published the same year as Debye ^{[Born-von-Karman 1912]}, retains the full dispersion and is the modern derivation; in the very-low-temperature regime where only long-wavelength modes contribute, the two derivations agree because the long-wavelength dispersion is linear by translation invariance.
The Einstein 1907 model is not a special case of Debye. Einstein took a single delta-function spectrum $g (ω) = 3 N δ (ω - ω_{E})$ ; Debye took a continuous spectrum on $[0, ω_{D}]$ . The two agree at high temperature (Dulong-Petit) but differ qualitatively at low temperature, where Einstein gives $C_{V} \sim e^{- ℏ ω_{E} / k_{B} T}$ and Debye gives $C_{V} \sim T^{3}$ . The data side with Debye in clean crystals.
The Debye temperature varies with temperature when extracted as a fitting parameter from finite-temperature data. A "Debye temperature" reported from room-temperature measurements need not match the low-temperature limit, because the simple Debye dispersion misses the optical-phonon branches and the dispersion bending near the zone boundary. The low-temperature $T^{3}$ -extrapolated $Θ_{D}$ is the cleanest definition; quoted values in tables usually refer to that limit.

Key derivation Intermediate+

Theorem (Debye specific-heat formula and its two limits). Let a solid of $N$ atoms in volume $V$ be modelled by the linear-dispersion isotropic phonon spectrum truncated at $ω_{D} = (6 π^{2} N c_{s}^{3} / V)^{1/3}$ , with $Θ_{D} = ℏ ω_{D} / k_{B}$ . The constant-volume specific heat is $$ C_V(T) = 9 N k_B \left(\frac{T}{\Theta_D}\right)^3 \int_0^{\Theta_D / T} \frac{x^4 e^x}{(e^x - 1)^2}, dx, $$ and the two universal limits are: $$ T \ll \Theta_D : \qquad C_V(T) \sim \frac{12 \pi^4}{5} N k_B \left(\frac{T}{\Theta_D}\right)^3, $$ $$ T \gg \Theta_D : \qquad C_V(T) \to 3 N k_B. $$

Proof. Start from the formal internal energy $$ U(T) = \int_0^{\omega_D} g(\omega), \frac{\hbar \omega}{e^{\hbar \omega / k_B T} - 1}, d\omega $$ with $g (ω) = (3 V /2 π^{2} c_{s}^{3}) ω^{2}$ on $[0, ω_{D}]$ . The mode-count constraint $\int_{0}^{ω_{D}} g d ω = 3 N$ gives $ω_{D}^{3} = 6 π^{2} N c_{s}^{3} / V$ , equivalently $g (ω) = 9 N ω^{2} / ω_{D}^{3}$ on the truncated interval. Substituting, $$ U(T) = \frac{9 N}{\omega_D^3} \int_0^{\omega_D} \frac{\hbar \omega^3}{e^{\hbar \omega / k_B T} - 1}, d\omega. $$

Change variable to $x = ℏ ω / k_{B} T$ , so $ω = k_{B} T x /ℏ$ and $d ω = (k_{B} T /ℏ) d x$ , with the upper limit becoming $y = ℏ ω_{D} / k_{B} T = Θ_{D} / T$ . Substitution yields $$ U(T) = \frac{9 N}{\omega_D^3}, \hbar \cdot \frac{(k_B T)^4}{\hbar^4}, \int_0^{y} \frac{x^3}{e^x - 1}, dx = 9 N k_B T \left(\frac{T}{\Theta_D}\right)^3 \int_0^{\Theta_D / T} \frac{x^3}{e^x - 1}, dx. $$ Differentiating under the integral sign (the integrand and its derivative in $T$ are uniformly bounded on compact subsets of $T > 0$ because the integrand decays as $x^{3} e^{- x}$ for large $x$ , and the upper limit $Θ_{D} / T$ contributes a boundary term that cancels by direct computation), $$ \frac{\partial U}{\partial T} = 9 N k_B \left(\frac{T}{\Theta_D}\right)^3 \int_0^{\Theta_D / T} \frac{x^4 e^x}{(e^x - 1)^2}, dx, $$ which is the stated specific-heat formula. The replacement of the derivative $\partial_{T} [x^{3} / (e^{x} - 1)]$ at fixed $x$ by $x^{4} e^{x} / (e^{x} - 1)^{2}$ uses $\partial_{T} (1/ (e^{x} - 1)) = \partial_{T} x \cdot \partial_{x} (1/ (e^{x} - 1))$ with $\partial_{T} x = - x / T$ in the integrand and a cancelling factor from $\partial_{T} y$ at the upper limit; the boundary term is $- x^{3} / (e^{x} - 1) ∣_{x = y} \cdot (\partial_{T} y) \cdot (prefactor adjustment)$ , and the combination simplifies to the integrand $x^{4} e^{x} / (e^{x} - 1)^{2}$ after collecting.

Low-temperature limit. When $T ≪ Θ_{D}$ , the upper limit $Θ_{D} / T \to \infty$ . The integrand $x^{4} e^{x} / (e^{x} - 1)^{2}$ decays as $x^{4} e^{- x}$ for large $x$ , so the integral converges to the value at infinity. A standard evaluation uses the integral identity $$ \int_0^\infty \frac{x^{s-1}}{e^x - 1}, dx = \Gamma(s), \zeta(s) \qquad (s > 1). $$ With $s = 4$ this gives $\int_{0}^{\infty} x^{3} / (e^{x} - 1) d x = Γ (4) ζ (4) = 6 \cdot (π^{4} /90) = π^{4} /15$ . Differentiating the internal-energy expression with respect to $T$ at the convergent integral gives $$ C_V(T) = 9 N k_B \left(\frac{T}{\Theta_D}\right)^3 \cdot \frac{4\pi^4}{15} = \frac{12 \pi^4}{5} N k_B \left(\frac{T}{\Theta_D}\right)^3, $$ where the factor $4 π^{4} /15$ comes from $\int_{0}^{\infty} x^{4} e^{x} / (e^{x} - 1)^{2} d x = 4 \int_{0}^{\infty} x^{3} / (e^{x} - 1) d x = 4 π^{4} /15$ (use integration by parts: $\int x^{4} e^{x} / (e^{x} - 1)^{2} d x = - x^{4} / (e^{x} - 1) + 4 \int x^{3} / (e^{x} - 1) d x$ , with the boundary at infinity vanishing exponentially).

High-temperature limit. When $T ≫ Θ_{D}$ , the upper limit $y = Θ_{D} / T \to 0$ , and the integrand is dominated by small $x$ where $e^{x} - 1 \approx x$ and $x^{4} e^{x} / (e^{x} - 1)^{2} \approx x^{2}$ . Then $$ \int_0^{y} x^2, dx = \frac{y^3}{3}. $$ Plugging in, $$ C_V(T) \to 9 N k_B \left(\frac{T}{\Theta_D}\right)^3 \cdot \frac{1}{3} \left(\frac{\Theta_D}{T}\right)^3 = 3 N k_B. $$ This is the Dulong-Petit value. The next correction is $- \frac{1}{20} (Θ_{D} / T)^{2}$ (from the next term in the Taylor expansion of $x^{4} e^{x} / (e^{x} - 1)^{2} - x^{2}$ about $x = 0$ ), so $C_{V} /3 N k_{B} = 1 - (1/20) (Θ_{D} / T)^{2} + O ((Θ_{D} / T)^{4})$ . $□$

Bridge. The Debye derivation builds toward every quantum-statistical model of a continuous phonon spectrum and appears again in 08.10.01, where the same Bose-Einstein occupation $1/ (e^{ℏ ω / k_{B} T} - 1)$ governs the grand-canonical partition function on the bosonic Fock space. The foundational reason the $T^{3}$ law holds is exactly that the long-wavelength phonon spectrum is linear in three dimensions and the density of states is $ω^{2}$ ; this is the central insight, and it generalises to any massless bosonic excitation with linear dispersion. The Stefan-Boltzmann law for photons is dual to the Debye law in this sense: blackbody radiation in 11.05.01 is the same calculation without the cutoff, because photons have no upper-frequency mode-counting constraint, and the corresponding energy density goes as $T^{4}$ rather than $C_{V}$ as $T^{3}$ . Putting these together, the Debye model, the Planck blackbody spectrum, and the magnon contribution to ferromagnetic specific heats are different specialisations of one phonon-counting calculation, parametrised by the dispersion relation and the spatial dimension. The bridge is that the dimensionless ratio $Θ_{D} / T$ controls the entire $C_{V}$ curve through one universal function $D_{3}$ , and this is exactly the universality that identifies aluminium with diamond once their Debye temperatures are matched.

Exercises Intermediate+

Exercise 3 (medium, symbolic).

Derive the next-to-leading correction to the high-temperature Dulong-Petit limit. Show that $C_{V} /3 N k_{B} = 1 - (1/20) (Θ_{D} / T)^{2} + O ((Θ_{D} / T)^{4})$ .

Hint

Taylor-expand the integrand $x^{4} e^{x} / (e^{x} - 1)^{2}$ about $x = 0$ to one extra order beyond $x^{2}$ , then integrate from $0$ to $y = Θ_{D} / T$ and multiply by the $T^{3}$ prefactor.

Answer

Expand $e^{x} - 1 = x + x^{2} /2 + x^{3} /6 + O (x^{4})$ , so $(e^{x} - 1)^{2} = x^{2} (1 + x /2 + x^{2} /6 + O (x^{3}))^{2} = x^{2} (1 + x + 7 x^{2} /12 + O (x^{3}))$ . Then $x^{4} e^{x} / (e^{x} - 1)^{2} = x^{2} (1 + x + x^{2} /2 + O (x^{3})) / (1 + x + 7 x^{2} /12 + O (x^{3})) = x^{2} (1 - x^{2} /12 + O (x^{3}))$ , where the cancellation of linear terms is exact. (One can also see this by writing $x^{4} e^{x} / (e^{x} - 1)^{2} = x^{2} e^{x} / (1 - e^{- x})^{2} \cdot e^{- x} = x^{2} \cdot (x / (e^{x /2} - e^{- x /2}))^{2} / x^{- 2}$ and using the Taylor expansion of $x / sinh (x /2) = 2 - x^{2} /12 + O (x^{4})$ .) Integrating: $\int_{0}^{y} x^{2} (1 - x^{2} /12) d x = y^{3} /3 - y^{5} /60 + O (y^{7})$ . Multiplying by $9 N k_{B} (T / Θ_{D})^{3} = 9 N k_{B} / y^{3}$ gives $C_{V} = 3 N k_{B} (1 - y^{2} /20 + O (y^{4})) = 3 N k_{B} (1 - (Θ_{D} / T)^{2} /20 + O ((Θ_{D} / T)^{4}))$ . Rubric: full credit for the $- (Θ_{D} / T)^{2} /20$ coefficient.

Exercise 4 (medium, symbolic).

Show that the Debye free energy per atom at temperature $T$ is $f (T) = (9/8) k_{B} Θ_{D} + 3 k_{B} T ln (1 - e^{- Θ_{D} / T}) - k_{B} T \cdot D_{3} (Θ_{D} / T)$ , where $D_{3}$ is the Debye function and the first term is the zero-point energy per atom.

Hint

The Helmholtz free energy of an oscillator at frequency $ω$ is $f_{ω} = (ℏ ω /2) + k_{B} T ln (1 - e^{- ℏ ω / k_{B} T})$ . Integrate against $g (ω) = 9 N ω^{2} / ω_{D}^{3}$ from $0$ to $ω_{D}$ , change variable to $x = ℏ ω / k_{B} T$ , then use integration by parts to relate the resulting integral to $D_{3}$ .

Answer

The total Helmholtz free energy is $F = \int_{0}^{ω_{D}} g (ω) [ℏ ω /2 + k_{B} T ln (1 - e^{- ℏ ω / k_{B} T})] d ω$ . The zero-point piece is $(9 N / ω_{D}^{3}) \int_{0}^{ω_{D}} (ℏ ω^{3} /2) d ω = (9 N ℏ ω_{D} /8) = (9 N k_{B} Θ_{D} /8)$ , i.e. $9/8 \cdot k_{B} Θ_{D}$ per atom. For the thermal piece, change variable to $x = ℏ ω / k_{B} T$ : $$ F_{\mathrm{th}} = 9 N k_B T (T/\Theta_D)^3 \int_0^{\Theta_D/T} x^2 \ln(1 - e^{-x}), dx. $$ Integrate by parts: $\int x^{2} ln (1 - e^{- x}) d x = (x^{3} /3) ln (1 - e^{- x}) - \int (x^{3} /3) \cdot (e^{- x} / (1 - e^{- x})) d x = (x^{3} /3) ln (1 - e^{- x}) - (1/3) \int x^{3} / (e^{x} - 1) d x$ . Evaluating between $0$ and $y = Θ_{D} / T$ and combining with the prefactor produces $$ F_{\mathrm{th}} = 3 N k_B T \ln(1 - e^{-\Theta_D/T}) - N k_B T \cdot D_3(\Theta_D / T). $$ Dividing by $N$ gives the per-atom free energy quoted. Rubric: full credit for the integration-by-parts derivation plus the identification of the zero-point and $D_{3}$ pieces.

Exercise 6 (hard, short-answer).

Generalise the Debye derivation to a solid in $d$ spatial dimensions. Show that the low-temperature specific heat scales as $C_{V} \propto T^{d}$ , and identify the analogue of the Debye function.

Hint

In $d$ dimensions, the number of wave vectors with $∣ k ∣ < k$ scales as $k^{d}$ , so the density of states per polarisation is proportional to $ω^{d - 1}$ . Repeat the Debye change of variable.

Answer

In $d$ dimensions with $d$ polarisations per atom, the truncated density of states per polarisation is $g_{p} (ω) = (V_{d} / S_{d} c_{s}^{d}) ω^{d - 1}$ on $[0, ω_{D}]$ , where $V_{d}$ is the volume and $S_{d}$ depends on the geometry of the $d$ -dimensional Brillouin zone (e.g., $S_{d} = (2 π)^{d} / vol (B_{1}^{d})$ with $B_{1}^{d}$ the unit ball). Mode-counting gives $ω_{D}^{d} = d \cdot S_{d} N c_{s}^{d} / V_{d}$ , fixing $Θ_{D} = ℏ ω_{D} / k_{B}$ . The internal energy is $$ U(T) = d N k_B T \cdot d \left(\frac{T}{\Theta_D}\right)^d \int_0^{\Theta_D/T} \frac{x^d}{e^x - 1}, dx, $$ modulo combinatorial constants, where $d$ in front counts polarisations and the prefactor $d (T / Θ_{D})^{d}$ replaces the $9 (T / Θ_{D})^{3}$ of the three-dimensional case. The low-temperature limit uses $\int_{0}^{\infty} x^{d} / (e^{x} - 1) d x = Γ (d + 1) ζ (d + 1)$ , giving $C_{V} \propto T^{d}$ . The Debye function generalises to $D_{d} (y) = (d / y^{d}) \int_{0}^{y} x^{d} / (e^{x} - 1) d x$ . In $d = 1$ (a phonon chain): $C_{V} \propto T$ at low temperature. In $d = 2$ (a graphene-like sheet, with the caveat that real two-dimensional acoustic spectra include flexural modes with quadratic dispersion that modify the leading exponent): $C_{V} \propto T^{2}$ . Rubric: full credit for the scaling exponent $d$ plus the generalised Debye function.

Exercise 7 (hard, short-answer).

For a metal at low temperature, the total specific heat is $C_{V} = γ T + A T^{3}$ , where $γ T$ is the Sommerfeld electronic contribution and $A T^{3}$ is the Debye lattice contribution. Explain how to extract both $γ$ and $Θ_{D}$ from a measured plot of $C_{V} / T$ versus $T^{2}$ .

Hint

Divide the equation by $T$ and observe how the two contributions separate.

Answer

Dividing by $T$ gives $C_{V} / T = γ + A T^{2}$ . A plot of $C_{V} / T$ versus $T^{2}$ is a straight line with intercept $γ$ (the Sommerfeld coefficient, proportional to the electronic density of states at the Fermi level) and slope $A = (12 π^{4} /5) N_{A} k_{B} / Θ_{D}^{3}$ for one mole. The Debye temperature is then $Θ_{D} = ((12 π^{4} /5) N_{A} k_{B} / A)^{1/3} = (1944/ A)^{1/3} K$ with $A$ in units of $J/ (mol \cdot K^{4})$ . This linear-plot extraction is the standard experimental method ^{[Ashcroft-Mermin Ch. 2 and Ch. 23]}: it cleanly separates the electronic and phononic contributions by exploiting their different temperature exponents. Below $T_{c}$ in a superconductor, the electronic piece changes to $γ T \cdot exp (- Δ/ k_{B} T)$ from the BCS gap, and the linear plot bends. Rubric: full credit for the linearisation argument plus identification of intercept and slope.

Exercise 8 (hard, short-answer).

Show that the integral identity $\int_{0}^{\infty} x^{3} / (e^{x} - 1) d x = π^{4} /15$ follows from $ζ (4) = π^{4} /90$ , using the geometric-series expansion of $1/ (e^{x} - 1)$ .

Hint

Write $1/ (e^{x} - 1) = \sum_{n = 1}^{\infty} e^{- n x}$ and integrate term by term against $x^{3}$ .

Answer

For $x > 0$ , the geometric-series expansion gives $1/ (e^{x} - 1) = e^{- x} / (1 - e^{- x}) = \sum_{n = 1}^{\infty} e^{- n x}$ . Multiplying by $x^{3}$ and integrating from $0$ to $\infty$ , $$ \int_0^\infty \frac{x^3}{e^x - 1}, dx = \sum_{n=1}^\infty \int_0^\infty x^3 e^{-n x}, dx = \sum_{n=1}^\infty \frac{3!}{n^4} = 6 \zeta(4) = 6 \cdot \frac{\pi^4}{90} = \frac{\pi^4}{15}. $$ Term-by-term integration is justified by monotone convergence (each $x^{3} e^{- n x}$ is positive). The closed-form value $ζ (4) = π^{4} /90$ is Euler's 1735 evaluation, originally obtained from the Weierstrass product for $sin (π z) / (π z)$ . Rubric: full credit for the series expansion plus the Euler-zeta identification.

Advanced results Master

Theorem (Einstein 1907 single-frequency model and its failure at low $T$ ). In Einstein's 1907 model, all $3 N$ atomic oscillators have a single frequency $ω_{E}$ , so $g (ω) = 3 N δ (ω - ω_{E})$ . The resulting specific heat is $$ C_V^{\mathrm{Einstein}}(T) = 3 N k_B \left(\frac{\Theta_E}{T}\right)^2 \frac{e^{\Theta_E / T}}{(e^{\Theta_E / T} - 1)^2}, \qquad \Theta_E = \frac{\hbar \omega_E}{k_B}. $$ At high $T$ , $C_{V} \to 3 N k_{B}$ as in Debye. At low $T$ , $C_{V} \sim 3 N k_{B} (Θ_{E} / T)^{2} e^{- Θ_{E} / T}$ , an exponential decay that disagrees with the experimental $T^{3}$ behaviour ^{[Einstein 1907]}.

The Einstein model was the first quantum-statistical treatment of a solid and resolved the rough Dulong-Petit failure at moderately low temperature, but at very low $T$ the exponential gap left over by the single frequency is wrong: a real solid has acoustic modes of arbitrarily low frequency, contributing power-law-not-exponential thermal energy. The Debye 1912 fix is to give the spectrum a low-frequency tail, which then dominates at low temperature. Einstein's model remains useful as an approximation for optical phonon branches in materials with a clear optical-acoustic separation (NaCl, GaAs).

Theorem (Born-von Kármán 1912 lattice-dynamics derivation). Born and von Kármán, in their 1912 paper ^{[Born-von-Karman 1912]} published the same year as Debye, derived the specific heat from the full lattice dynamics of a crystal without the linear-dispersion approximation. The vibrational normal modes are found by diagonalising the dynamical matrix $$ D_{\alpha\beta}(k) = \frac{1}{m} \sum_{R} \Phi_{\alpha\beta}(R), e^{i k \cdot R}, $$ where $Φ_{α β}$ is the force-constant matrix and $R$ runs over Bravais lattice vectors. The phonon dispersion $ω_{s} (k)$ , indexed by polarisation $s$ , gives the exact density of states $g (ω) = \sum_{s} \int [d^{3} k / (2 π)^{3}] δ (ω - ω_{s} (k))$ , integrated over the Brillouin zone.

In the long-wavelength limit, the dispersion is linear and Born-von Kármán reduces to Debye; near the Brillouin-zone boundary, the dispersion bends and produces van Hove singularities in $g (ω)$ that the Debye approximation misses. For a typical elemental crystal at $T ≲ Θ_{D} /10$ , only long-wavelength modes contribute, and the two derivations agree numerically. At intermediate temperatures the difference is visible: the "effective Debye temperature" $Θ_{D}^{*} (T)$ extracted from a one-parameter fit varies smoothly with $T$ between the low-temperature ( $T^{3}$ ) and high-temperature (Dulong-Petit) limits.

Theorem (Sommerfeld 1928 electronic contribution in metals). For a metal modelled as a degenerate ideal Fermi gas of $N$ conduction electrons in volume $V$ , the electronic specific heat at $T ≪ T_{F}$ is $$ C_V^{\mathrm{el}}(T) = \frac{\pi^2}{2} N k_B \frac{T}{T_F} + O(T^3), $$ where $T_{F} = ϵ_{F} / k_{B}$ is the Fermi temperature. The total low-temperature specific heat of a metal is the sum $C_{V} = γ T + (12 π^{4} /5) N k_{B} (T / Θ_{D})^{3} + \dots$ , and the two contributions are separated by the linearisation procedure in Exercise 7 ^{[Sommerfeld 1928]}.

The Sommerfeld result follows from a Sommerfeld expansion of the Fermi-Dirac integral about $T = 0$ . In a clean metal at $T \approx Θ_{D} /100$ , the electronic and lattice contributions are comparable; below that, the electronic linear- $T$ term dominates the lattice $T^{3}$ term. The Sommerfeld coefficient $γ = (π^{2} /2) N k_{B} / T_{F} = (π^{2} /3) k_{B}^{2} ρ (ϵ_{F})$ is proportional to the electronic density of states at the Fermi level $ρ (ϵ_{F})$ , and is a key diagnostic of band-structure renormalisations in correlated metals (heavy-fermion compounds carry $γ$ values orders of magnitude above the free-electron prediction).

Theorem (BCS-superconducting exponential specific-heat anomaly). For a phonon-mediated BCS superconductor with gap $Δ (T)$ , the electronic specific heat below $T_{c}$ decays as $$ C_V^{\mathrm{el, BCS}}(T) \sim a \cdot \gamma T_c \left(\frac{\Delta(0)}{k_B T}\right)^{3/2} e^{-\Delta(0) / k_B T} $$ for $T ≪ T_{c}$ , where $a$ is a numerical constant and $Δ (0) \approx 1.76 k_{B} T_{c}$ in the BCS weak-coupling limit. The BCS jump $Δ C_{V} / C_{V}^{normal} = 1.43$ at $T_{c}$ is the universal weak-coupling prediction ^{[Bardeen-Cooper-Schrieffer 1957]}.

The phonon contribution remains $\sim T^{3}$ across $T_{c}$ , so the exponential anomaly is observable only in the electronic part. The mechanism that produces the gap — phonon-mediated electron-electron attraction — is itself a direct consequence of the Debye phonon spectrum: the BCS pairing kernel is cut off at $ℏ ω_{D}$ , and the resulting $T_{c} \sim Θ_{D} exp (- 1/ N (0) V)$ formula identifies the Debye temperature as the energy scale controlling superconducting transition temperatures across the conventional-superconductor periodic table (Pb, Nb, Al, Sn).

Theorem (anharmonic corrections via the Grüneisen parameter). Beyond the harmonic approximation, the phonon frequencies depend on volume through the Grüneisen parameter $γ_{G} := - d ln ω / d ln V$ (assumed approximately constant across the spectrum). The thermal-expansion coefficient $α$ is related to the specific heat by the Grüneisen relation $$ \alpha = \gamma_G \frac{\kappa_T C_V}{V}, $$ where $κ_{T}$ is the isothermal compressibility. The Debye temperature itself becomes a slow function of volume, $Θ_{D} (V)$ , with $d ln Θ_{D} / d ln V = - γ_{G}$ .

The Grüneisen relation links specific heat to thermal expansion and is one of the cleanest cross-checks of the Debye model. For aluminium, $γ_{G} \approx 2.2$ , $κ_{T} \approx 1.4 \times 1 0^{- 11} P a^{- 1}$ , $V \approx 1 0^{- 5} m^{3} /mol$ , and the predicted $α$ matches the measured thermal-expansion coefficient $α \approx 23 \times 1 0^{- 6} K^{- 1}$ at room temperature to within ten percent.

Theorem (universality of the dimensionless Debye curve). The function $C_{V} /3 N k_{B}$ expressed in terms of the dimensionless variable $T / Θ_{D}$ is one universal curve $f (T / Θ_{D})$ , independent of material parameters. Two different solids with the same $T / Θ_{D}$ have the same fractional Dulong-Petit. The collapse of specific-heat data onto this universal curve, after scaling each material's temperature by its $Θ_{D}$ , is one of the cleanest universality observations in condensed-matter physics.

The universality follows from the Debye formula being a function of $Θ_{D} / T$ only, after extracting the $3 N k_{B}$ Dulong-Petit prefactor. The cleanness of the collapse in real crystals testifies to the smallness of dispersion corrections at low temperature. Anomalies near the Brillouin-zone boundary, optical-phonon contributions, and impurity scattering produce sub-leading deviations from the universal curve at intermediate $T$ .

Synthesis. The Debye derivation is the foundational reason every quantum-statistical model of a continuous massless bosonic spectrum predicts a power-law specific heat at low temperature, and the central insight is that the long-wavelength dispersion is linear in three dimensions by translation invariance, producing a density of states proportional to $ω^{2}$ on the truncated interval. Putting these together, the Debye law, the Stefan-Boltzmann photon law, the magnon $T^{3/2}$ contribution in ferromagnets, and the $C_{V} \propto T$ behaviour of the one-dimensional phonon chain are different specialisations of one density-of-states calculation, parametrised by the dispersion exponent and the spatial dimension. The bridge is that the dimensionless ratio $Θ_{D} / T$ controls every Debye prediction through one universal function, which identifies aluminium with diamond once their Debye temperatures are matched. This is exactly the universality that appears again in 11.05.01 (Bose-Einstein distribution), where the Planck blackbody spectrum is the same calculation without the upper cutoff and the energy density scales as $T^{4}$ for the same density-of-states reason. The Debye derivation generalises to any bosonic excitation with a known dispersion relation: phonons, photons, magnons, and the rotons and phonons of superfluid $^{4} He$ are all instances of one phonon-counting framework.

The same framework supports the deeper analysis of metals. The Sommerfeld electronic contribution adds a linear-in- $T$ term that dominates below $T \approx Θ_{D} /100$ and is the experimental fingerprint of the Fermi level; the Bardeen-Cooper-Schrieffer superconducting transition replaces this linear term with the exponential anomaly $C_{V} \sim e^{- Δ/ k_{B} T}$ from the superconducting gap. The Debye temperature itself controls the BCS pairing kernel cutoff and sets the energy scale for $T_{c}$ in conventional superconductors. The bridge from a single-element parameter to a quantitative prediction of transition temperatures across the periodic table is one of the central conceptual achievements of twentieth-century solid-state physics. Anharmonic corrections enter through the Grüneisen parameter and connect the specific heat to the thermal-expansion coefficient by one relation, and this is exactly the integration of the Debye model into the wider phenomenology of crystalline solids that Ashcroft and Mermin develop systematically in Chapters 22-23.

Full proof set Master

Proposition (mode-counting cutoff). Given the linear-dispersion density of states $g (ω) = (3 V /2 π^{2} c_{s}^{3}) ω^{2}$ on $[0, ω_{D}]$ , the cutoff $ω_{D}$ chosen by $\int_{0}^{ω_{D}} g (ω) d ω = 3 N$ satisfies $ω_{D}^{3} = 6 π^{2} N c_{s}^{3} / V$ .

Proof. Direct integration: $$ \int_0^{\omega_D} \frac{3 V}{2 \pi^2 c_s^3}, \omega^2, d\omega = \frac{3 V}{2 \pi^2 c_s^3} \cdot \frac{\omega_D^3}{3} = \frac{V \omega_D^3}{2 \pi^2 c_s^3}. $$ Setting this equal to $3 N$ gives $V ω_{D}^{3} / (2 π^{2} c_{s}^{3}) = 3 N$ , hence $ω_{D}^{3} = 6 π^{2} N c_{s}^{3} / V$ . $□$

Proposition (low-temperature limit). As $T / Θ_{D} \to 0$ , $$ \int_0^{\Theta_D / T} \frac{x^4 e^x}{(e^x - 1)^2}, dx \to \frac{4 \pi^4}{15}, $$ and consequently $C_{V} \sim (12 π^{4} /5) N k_{B} (T / Θ_{D})^{3}$ .

Proof. Integration by parts with $u = x^{4}$ , $d v = e^{x} / (e^{x} - 1)^{2} d x = - d (1/ (e^{x} - 1))$ : $$ \int_0^Y \frac{x^4 e^x}{(e^x - 1)^2}, dx = -\frac{x^4}{e^x - 1}\bigg|0^Y + 4 \int_0^Y \frac{x^3}{e^x - 1}, dx. $$ The boundary term at $x = 0$ vanishes because $x^{4} / (e^{x} - 1) \sim x^{3} \to 0$ , and at $x = Y$ it scales as $Y^{4} e^{- Y} \to 0$ as $Y \to \infty$ . The remaining integral evaluates via the geometric-series expansion $1 / (e^x - 1) = \sum{n=1}^\infty e^{-nx}$: $$ \int_0^\infty \frac{x^3}{e^x - 1}, dx = \sum_{n=1}^\infty \int_0^\infty x^3 e^{-nx}, dx = \sum_{n=1}^\infty \frac{6}{n^4} = 6 \zeta(4) = \frac{\pi^4}{15}. $$ Multiplying by $4$ gives $4 π^{4} /15$ . Then $C_{V} = 9 N k_{B} (T / Θ_{D})^{3} \cdot 4 π^{4} /15 = (12 π^{4} /5) N k_{B} (T / Θ_{D})^{3}$ . $□$

Proposition (high-temperature limit). As $T / Θ_{D} \to \infty$ , $$ 9 N k_B \left(\frac{T}{\Theta_D}\right)^3 \int_0^{\Theta_D / T} \frac{x^4 e^x}{(e^x - 1)^2}, dx \to 3 N k_B, $$ with leading correction $- 3 N k_{B} \cdot (Θ_{D} / T)^{2} /20$ .

Proof. For small $x$ , expand the integrand. Using $e^{x} - 1 = x + x^{2} /2 + x^{3} /6 + O (x^{4})$ and $e^{x} = 1 + x + x^{2} /2 + x^{3} /6 + O (x^{4})$ , factor $x^{2}$ from the denominator: $$ \frac{x^4 e^x}{(e^x - 1)^2} = \frac{x^4 e^x}{x^2 (1 + x/2 + x^2/6 + O(x^3))^2} = x^2 \cdot \frac{1 + x + x^2/2 + O(x^3)}{1 + x + 7x^2/12 + O(x^3)}. $$ The ratio inside expands as $1 + (x^{2} /2 - 7 x^{2} /12) + O (x^{3}) = 1 - x^{2} /12 + O (x^{3})$ . So $x^{4} e^{x} / (e^{x} - 1)^{2} = x^{2} (1 - x^{2} /12 + O (x^{3}))$ . Integrating from $0$ to $y = Θ_{D} / T$ gives $y^{3} /3 - y^{5} /60 + O (y^{7})$ . Multiplying by $9 N k_{B} (T / Θ_{D})^{3} = 9 N k_{B} / y^{3}$ yields $$ C_V = 3 N k_B - \frac{3 N k_B}{20} \left(\frac{\Theta_D}{T}\right)^2 + O((\Theta_D / T)^4). \qquad \square $$

Proposition (Einstein-model low-temperature exponential decay). In Einstein's model, $C_{V}^{Einstein} (T) \sim 3 N k_{B} (Θ_{E} / T)^{2} e^{- Θ_{E} / T}$ as $T / Θ_{E} \to 0$ .

Proof. From $C_{V}^{Einstein} = 3 N k_{B} (Θ_{E} / T)^{2} e^{Θ_{E} / T} / (e^{Θ_{E} / T} - 1)^{2}$ , write $e^{Θ_{E} / T} = u$ , so the formula is $3 N k_{B} (Θ_{E} / T)^{2} u / (u - 1)^{2}$ . For $u ≫ 1$ (low $T$ ), $u / (u - 1)^{2} = u / (u^{2} (1 - 1/ u)^{2}) = (1/ u) (1 + O (1/ u)) \sim 1/ u = e^{- Θ_{E} / T}$ . So $C_{V}^{Einstein} \sim 3 N k_{B} (Θ_{E} / T)^{2} e^{- Θ_{E} / T}$ , an exponential decay rather than the $T^{3}$ Debye law. $□$

Proposition (Sommerfeld electronic specific heat). For a degenerate ideal Fermi gas with chemical potential $μ \approx ϵ_{F} = k_{B} T_{F}$ at $T ≪ T_{F}$ , the internal energy admits the Sommerfeld expansion $U (T) = U_{0} + (π^{2} /6) ρ (ϵ_{F}) (k_{B} T)^{2} + O (T^{4})$ , where $ρ (ϵ)$ is the electronic density of states. Differentiation gives $C_{V}^{el} = (π^{2} /3) ρ (ϵ_{F}) k_{B}^{2} T + O (T^{3}) = (π^{2} /2) N k_{B} (T / T_{F}) + O (T^{3})$ after using $ρ (ϵ_{F}) = (3 N) / (2 ϵ_{F})$ for the three-dimensional free Fermi gas.

Proof sketch — stated without full proof, see Ashcroft-Mermin Ch. 2 ^{[Ashcroft-Mermin]}. The Sommerfeld expansion is the Mellin-Barnes / Bernoulli-number expansion of the Fermi-Dirac integral $\int H (ϵ) f (ϵ) d ϵ$ about $ϵ = μ$ at low $T$ . The leading $T^{2}$ correction comes from the second Bernoulli number $B_{2} = 1/6$ and a factor of $π^{2}$ from the Riemann zeta values $ζ (2) = π^{2} /6$ . The full derivation is standard and is included in any solid-state-physics textbook. The result $γ = (π^{2} /2) N k_{B} / T_{F}$ is the Sommerfeld coefficient, proportional to the density of states at the Fermi level. $□$

Connections Master

Bosonic Fock space and second quantisation 08.10.01. The Debye model is the simplest physical realisation of a bosonic Fock space with finitely many modes truncated at the Debye frequency. The Bose-Einstein occupation $\overset{n}{ˉ} (ω) = 1/ (e^{ℏ ω / k_{B} T} - 1)$ for each phonon mode is the thermal-state expectation value of the number operator $a^{†} (ω) a (ω)$ on the Fock space, and the partition function over $3 N$ oscillators is the trace of $e^{- β H}$ over the truncated bosonic Fock space.
Bose-Einstein distribution 11.05.01. Phonons are bosons, and their thermal occupation is the Bose-Einstein distribution with vanishing chemical potential. The Debye derivation parallels the Planck blackbody derivation: both replace a sum over modes by an integral over a density of states and substitute the Bose-Einstein occupation. The two calculations differ in the upper cutoff: Debye truncates at $ω_{D}$ to enforce a finite mode count of $3 N$ , while the photon spectrum extends to infinity because photons have no mode-counting constraint.
Partition function (statistical mechanics) 08.01.01. The Debye derivation starts from the canonical partition function for $3 N$ independent harmonic oscillators, computed exactly from the single-oscillator partition function $Z_{ω} = e^{- β ℏ ω /2} / (1 - e^{- β ℏ ω})$ . The internal energy and specific heat are derived by differentiation of $ln Z$ with respect to $β = 1/ k_{B} T$ , the standard route for any canonical ensemble.
Boltzmann distribution and canonical ensemble 08.01.03. The canonical-ensemble framework supplies the thermal weights $e^{- β E_{α}}$ on each oscillator state, from which the Bose-Einstein occupation follows by geometric-series summation. The Debye theory is the worked example of a canonical-ensemble calculation for a system of independent quantum oscillators with a continuous spectrum.
Free energy 08.01.04. The Debye free energy $F (T) = (9/8) N k_{B} Θ_{D} + 3 N k_{B} T ln (1 - e^{- Θ_{D} / T}) - N k_{B} T D_{3} (Θ_{D} / T)$ is the standard derivation of a phonon-system Helmholtz free energy. Differentiation $S = - \partial F / \partial T$ and $C_{V} = T \partial S / \partial T = - T \partial^{2} F / \partial T^{2}$ recovers entropy and specific heat in one stroke, illustrating the general role of $F$ as the canonical generating function.

Historical & philosophical context Master

Dulong and Petit reported in 1819 ^{[Dulong-Petit 1819]} the empirical law that the molar specific heat of every solid element they measured was approximately $3 R \approx 25 J/ (mol \cdot K)$ at room temperature. Their list included copper, silver, gold, lead, and a dozen other metals; the agreement was excellent at high temperature but, as later refrigeration techniques pushed measurements below $100 K$ , dramatic discrepancies appeared. James Dewar and Heinrich Friedrich Weber in the 1870s and 1880s found that diamond's specific heat at room temperature was far below $3 R$ , and that the specific heats of all solids dropped toward zero as temperature was lowered. Classical equipartition could not account for the drop, and the failure of Dulong-Petit at low temperature was one of the standing puzzles of late-nineteenth-century thermodynamics.

Albert Einstein's 1907 paper ^{[Einstein 1907]} applied Planck's 1900 quantisation of oscillator energies to the atomic vibrations of a solid, treating each atom as a single oscillator with frequency $ω_{E}$ . The resulting formula $C_{V} = 3 N k_{B} (Θ_{E} / T)^{2} e^{Θ_{E} / T} / (e^{Θ_{E} / T} - 1)^{2}$ correctly recovered the Dulong-Petit value at high temperature and decayed toward zero at low temperature, the first quantum-statistical resolution of the Dulong-Petit puzzle. But the exponential decay at low $T$ did not match the data — by 1911, Walther Nernst's calorimetric programme in Berlin had measured enough low-temperature specific heats to show that the decay was a power law, not exponential.

Peter Debye's 1912 paper Zur Theorie der spezifischen Wärmen (Ann. Phys. 39, 789-839) ^{[Debye 1912]} supplied the right model. Debye replaced Einstein's single oscillator with a continuous spectrum of acoustic vibrations, capped at a frequency $ω_{D}$ fitted by the mode-counting constraint. The resulting $T^{3}$ law at low temperature matched Nernst's data immediately. The same year, Max Born and Theodore von Kármán published a more general lattice-dynamics treatment (Phys. Z. 13, 297-309) ^{[Born-von-Karman 1912]} that retained the exact phonon dispersion; for long-wavelength modes their result agrees with Debye's, and for shorter wavelengths it adds dispersion corrections. The Born-von Kármán framework is now the modern standard, with Debye's linear-dispersion approximation serving as the analytically tractable simplification.

The Debye temperature emerged as a material parameter of fundamental importance: it is the energy scale separating the quantum and classical regimes of lattice vibrations, and it appears as the cutoff in the BCS pairing interaction (Bardeen, Cooper and Schrieffer 1957 ^{[Bardeen-Cooper-Schrieffer 1957]}) that produces phonon-mediated superconductivity. Sommerfeld's 1928 derivation of the electronic linear-in- $T$ contribution to metallic specific heat ^{[Sommerfeld 1928]} completed the modern decomposition $C_{V} = γ T + A T^{3} + \dots$ that experimentalists use to extract both the Sommerfeld coefficient and the Debye temperature from a single low-temperature plot. Landau and Lifshitz's textbook treatment in Volume 5 of the Course of Theoretical Physics §§63-66 ^{[Landau Vol 5]} established the Russian-school canonical presentation of the derivation; Ashcroft and Mermin's Solid State Physics Chapters 22-23 ^{[Ashcroft-Mermin]} developed the dispersion-corrected modern version with detailed comparison to experiment.

Bibliography Master

@article{Debye1912SpecificHeats,
  author  = {Debye, Peter},
  title   = {Zur {T}heorie der spezifischen {W}{\"a}rmen},
  journal = {Annalen der Physik},
  volume  = {344},
  number  = {14},
  pages   = {789--839},
  year    = {1912}
}

@article{BornVonKarman1912,
  author  = {Born, Max and von K{\'a}rm{\'a}n, Theodore},
  title   = {{\"U}ber {S}chwingungen in {R}aumgittern},
  journal = {Physikalische Zeitschrift},
  volume  = {13},
  pages   = {297--309},
  year    = {1912}
}

@article{Einstein1907Specific,
  author  = {Einstein, Albert},
  title   = {Die {P}lancksche {T}heorie der {S}trahlung und die {T}heorie der spezifischen {W}{\"a}rme},
  journal = {Annalen der Physik},
  volume  = {327},
  number  = {1},
  pages   = {180--190},
  year    = {1907}
}

@article{DulongPetit1819,
  author  = {Dulong, Pierre Louis and Petit, Alexis Th{\'e}rese},
  title   = {Recherches sur quelques points importants de la {T}h{\'e}orie de la {C}haleur},
  journal = {Annales de Chimie et de Physique},
  volume  = {10},
  pages   = {395--413},
  year    = {1819}
}

@article{Sommerfeld1928Electron,
  author  = {Sommerfeld, Arnold},
  title   = {Zur {E}lektronentheorie der {M}etalle auf {G}rund der {F}ermischen {S}tatistik},
  journal = {Zeitschrift f{\"u}r Physik},
  volume  = {47},
  pages   = {1--32},
  year    = {1928}
}

@article{BardeenCooperSchrieffer1957,
  author  = {Bardeen, John and Cooper, Leon N. and Schrieffer, John Robert},
  title   = {Theory of Superconductivity},
  journal = {Physical Review},
  volume  = {108},
  pages   = {1175--1204},
  year    = {1957}
}

@book{LandauLifshitzStatPhys1,
  author    = {Landau, Lev Davidovich and Lifshitz, Evgeny Mikhailovich},
  title     = {Statistical {P}hysics, {P}art 1},
  publisher = {Pergamon},
  edition   = {3},
  series    = {Course of Theoretical Physics},
  volume    = {5},
  year      = {1980},
  note      = {Translated by J. B. Sykes and M. J. Kearsley; revised and enlarged by E. M. Lifshitz and L. P. Pitaevskii}
}

@book{AshcroftMermin1976,
  author    = {Ashcroft, Neil W. and Mermin, N. David},
  title     = {Solid {S}tate {P}hysics},
  publisher = {Holt, Rinehart and Winston},
  year      = {1976}
}

@book{Kittel2005,
  author    = {Kittel, Charles},
  title     = {Introduction to {S}olid {S}tate {P}hysics},
  publisher = {John Wiley \& Sons},
  edition   = {8},
  year      = {2005}
}

@book{Reif1965,
  author    = {Reif, Frederick},
  title     = {Fundamentals of {S}tatistical and {T}hermal {P}hysics},
  publisher = {McGraw-Hill},
  year      = {1965}
}

@book{PathriaBeale2011,
  author    = {Pathria, R. K. and Beale, Paul D.},
  title     = {Statistical {M}echanics},
  publisher = {Elsevier},
  edition   = {3},
  year      = {2011}
}

@book{Kardar2007Particles,
  author    = {Kardar, Mehran},
  title     = {Statistical {P}hysics of {P}articles},
  publisher = {Cambridge University Press},
  year      = {2007}
}

@book{Schroeder2000,
  author    = {Schroeder, Daniel V.},
  title     = {An {I}ntroduction to {T}hermal {P}hysics},
  publisher = {Addison-Wesley},
  year      = {2000}
}

Prerequisites

08.01.01
08.01.03
08.01.04
08.10.01

Tier anchors

beginner: Schroeder, *An Introduction to Thermal Physics* (Addison-Wesley, 2000), §7.5; Feynman *Lectures on Physics* Vol. 1 Ch. 40
intermediate: Reif, *Fundamentals of Statistical and Thermal Physics* (McGraw-Hill, 1965), §10.2; Pathria & Beale, *Statistical Mechanics* 3e (Elsevier, 2011), §7.4; Kardar, *Statistical Physics of Particles* (Cambridge, 2007), §7.5
master: Landau & Lifshitz, *Statistical Physics, Part 1* 3e (Pergamon, 1980), §§63-66; Ashcroft & Mermin, *Solid State Physics* (Holt, 1976), Chs. 22-23; Kittel, *Introduction to Solid State Physics* 8e (Wiley, 2005), Ch. 5

References

Debye, P. — Zur Theorie der spezifischen Wärmen · Annalen der Physik 39, 789-839 (1912) — original Debye paper introducing the linear-dispersion / cutoff-frequency model and the T^3 law
Born, M. & von Kármán, T. — Über Schwingungen in Raumgittern · Physikalische Zeitschrift 13, 297-309 (1912) — same-year lattice-dynamics derivation retaining the full phonon dispersion
Einstein, A. — Die Plancksche Theorie der Strahlung und die Theorie der spezifischen Wärme · Annalen der Physik 22, 180-190 (1907) — single-frequency oscillator model, the first quantum-statistical treatment of a solid
Dulong, P. L. & Petit, A. T. — Recherches sur quelques points importants de la théorie de la chaleur · Annales de Chimie et de Physique 10, 395-413 (1819) — empirical law C_V = 3R per mole for solids at room temperature
Sommerfeld, A. — Zur Elektronentheorie der Metalle auf Grund der Fermischen Statistik · Zeitschrift für Physik 47, 1-32 (1928) — degenerate electron gas, linear-in-T electronic specific heat in metals
Landau, L. D. & Lifshitz, E. M. — Statistical Physics, Part 1, Volume 5 of the Course of Theoretical Physics · Pergamon / Butterworth-Heinemann 3e (Sykes-Kearsley translation, 1980), §§63-66 (solid as oscillator system, Debye approximation, T^3 law, Dulong-Petit limit)
Ashcroft, N. W. & Mermin, N. D. — Solid State Physics · Holt, Rinehart and Winston (1976), Chs. 22-23 — modern derivation, Debye-Einstein interpolation, comparison with experiment
Bardeen, J., Cooper, L. N. & Schrieffer, J. R. — Theory of superconductivity · Physical Review 108, 1175-1204 (1957) — phonon-mediated pairing and exponential specific-heat anomaly below T_c

Estimated time

beginner: 18m
intermediate: 45m
master: 75m