Condensed Matter: Lattice Dynamics & Phonons

Condensed Matter Physics » Lattice Dynamics & Phonons

Lattice Dynamics & Phonons

The big idea: vibrations are particles. A crystal is not a rigid scaffold of frozen atoms. At any nonzero temperature the ions are perpetually in motion, oscillating about their equilibrium positions. Because the lattice is periodic, those oscillations organize themselves into collective normal modes — travelling waves of displacement, each labelled by a wavevector $\mathbf{k}$ and a polarization. Quantizing a normal mode turns it into a harmonic oscillator whose energy comes in discrete quanta $\hbar\omega_{\mathbf{k}}$. These quanta are phonons: emergent particles that carry the crystal’s heat, scatter its electrons, and — by gluing electrons into Cooper pairs — make conventional superconductivity possible. The phonon is the cleanest example of a quasiparticle on this site: a genuinely collective excitation of $10^{23}$ atoms that you can nonetheless count, scatter, and assign a momentum, just like a free particle.

This page builds the theory of lattice vibrations from the ground up: from the geometry of the lattice, through the harmonic approximation and its normal modes, to the phonon dispersion relations that determine a solid’s thermal properties, and finally to the electron–phonon coupling that underlies conventional superconductivity. It assumes the crystallography recap on the hub page (Bravais lattices, reciprocal space, Brillouin zones) and expands it where lattice dynamics needs more.

Bravais Lattices and Reciprocal Space (Recap)

A Bravais lattice is the set of points

generated by three primitive vectors $\mathbf{a}_1, \mathbf{a}_2, \mathbf{a}_3$. Every point looks identical to every other — this discrete translational symmetry is the single fact from which everything below follows. A real crystal is a Bravais lattice decorated with a basis of one or more atoms per primitive cell; the basis is what distinguishes, say, diamond (two-atom basis on an FCC lattice) from a monatomic FCC metal.

The reciprocal lattice is spanned by vectors $\mathbf{b}_i$ defined by

with $\mathbf{b}_2, \mathbf{b}_3$ obtained by cyclic permutation. A reciprocal lattice vector is $\mathbf{G} = m_1\mathbf{b}_1 + m_2\mathbf{b}_2 + m_3\mathbf{b}_3$, and it satisfies $e^{i\mathbf{G}\cdot\mathbf{R}} = 1$ for every direct-lattice $\mathbf{R}$. This is the algebraic engine of lattice dynamics: any function with the periodicity of the lattice — the potential, the charge density, a normal-mode pattern — can be expanded in plane waves $e^{i\mathbf{G}\cdot\mathbf{r}}$ over the reciprocal lattice only.

The first Brillouin zone (BZ) is the Wigner–Seitz cell of the reciprocal lattice: the set of wavevectors closer to the origin than to any other reciprocal lattice point. Two consequences drive everything that follows:

• Wavevectors live in the BZ. A wave $e^{i\mathbf{k}\cdot\mathbf{R}}$ sampled only at lattice points is unchanged by $\mathbf{k} \to \mathbf{k} + \mathbf{G}$, because $e^{i\mathbf{G}\cdot\mathbf{R}} = 1$. So $\mathbf{k}$ and $\mathbf{k}+\mathbf{G}$ describe the same physical mode, and we may restrict $\mathbf{k}$ to one BZ. There is no such thing as a lattice vibration with a wavelength shorter than twice the lattice spacing.
• Mode counting. For a crystal of $N$ primitive cells under periodic (Born–von Kármán) boundary conditions, the allowed $\mathbf{k}$ form a uniform mesh of exactly $N$ points filling the BZ. This finite count is what makes thermodynamic sums tractable.

Crystal momentum is conserved only modulo G. Because the lattice breaks continuous translational symmetry down to discrete translations, momentum is not strictly conserved in a crystal — only crystal momentum $\hbar\mathbf{k}$, and only up to a reciprocal lattice vector $\hbar\mathbf{G}$. A scattering process that transfers $\mathbf{G}$ to the lattice as a whole is called an Umklapp (“flip-over”) process, and it is precisely these Umklapp events — not ordinary momentum-conserving collisions — that give a perfect insulating crystal a finite thermal resistance.

The Harmonic Crystal

Let the equilibrium position of the atom in cell $\mathbf{R}$ on basis site $s$ be $\mathbf{R} + \boldsymbol{\tau}s$, and let $\mathbf{u}{\mathbf{R}s}$ be its small displacement away from equilibrium. The crystal’s potential energy $U$ is some complicated function of all the displacements. The key move is to Taylor-expand $U$ about the equilibrium configuration and keep only the leading nontrivial term:

The linear term vanishes because we expand about a minimum (zero net force at equilibrium). The constant $U_0$ is just the static binding energy. The quadratic term defines the force-constant matrix

where $\alpha,\beta \in {x,y,z}$. Truncating here is the harmonic approximation: every atom feels a restoring force that is linear in the displacements, so the whole crystal is a giant coupled system of harmonic oscillators. The classical equation of motion for each atom is then

The force constants are constrained by the crystal’s symmetry. Translational invariance — moving the whole crystal rigidly costs no energy — forces the acoustic sum rule

which we will see guarantees the existence of gapless acoustic phonons. Point-group symmetry further relates different $\Phi$ components, drastically reducing the number of independent constants.

Why “harmonic” is a good first approximation — and where it fails. Near a potential minimum every smooth potential looks parabolic, so the harmonic model is exact in the limit of small amplitude. It captures the existence of well-defined, infinitely long-lived normal modes and gives the correct low-temperature heat capacity. What it cannot capture are effects that require phonons to interact: thermal expansion, finite phonon lifetimes, and the temperature dependence of the thermal conductivity all come from the cubic and higher (anharmonic) terms we dropped. A purely harmonic crystal would have infinite thermal conductivity and would never expand on heating.

Normal Modes and Phonon Dispersion

Translational symmetry lets us solve the coupled equations exactly with a plane-wave ansatz. Seek solutions of the form

where $\boldsymbol{\epsilon}_s$ is a (mass-weighted) polarization vector to be determined. Substituting into the equation of motion turns the infinite set of coupled differential equations into a single small eigenvalue problem at each $\mathbf{k}$:

where the dynamical matrix is the mass-weighted Fourier transform of the force constants:

For a crystal with $p$ atoms in the basis, $D(\mathbf{k})$ is a $3p \times 3p$ Hermitian matrix. Diagonalizing it at each $\mathbf{k}$ yields $3p$ eigenvalues $\omega_j^2(\mathbf{k})$ and eigenvectors $\boldsymbol{\epsilon}_j(\mathbf{k})$. The functions $\omega_j(\mathbf{k})$ are the phonon dispersion relations — the central output of lattice dynamics — and the index $j = 1,\dots,3p$ labels the branches. The eigenvector tells you the pattern of atomic motion: longitudinal (along $\mathbf{k}$), transverse, or some mixture.

The monatomic 1D chain

The smallest nontrivial example fixes intuition. Take a chain of identical atoms of mass $M$, spacing $a$, connected by springs of constant $C$ to nearest neighbours only. The equation of motion for atom $n$ is

The plane-wave ansatz $u_n = \epsilon\,e^{i(kna - \omega t)}$ gives, after using $e^{ika} + e^{-ika} = 2\cos(ka)$,

This single curve is the whole dispersion. Note its features, which generalize to every acoustic branch:

• Gapless and linear at small $k$: $\omega \approx \sqrt{C/M}\,a\,

  k

  = v_s

  k

  $, where $v_s = a\sqrt{C/M}$ is the sound speed. Long-wavelength vibrations are ordinary sound waves. The gaplessness is the acoustic sum rule at work.

• Zone-boundary flattening: at $k = \pi/a$ the group velocity $d\omega/dk$ vanishes — the wave is a standing wave (neighbouring atoms exactly out of phase) and carries no energy. This is a van Hove singularity in the density of states.
• Periodicity: $\omega(k)$ is periodic in $k \to k + 2\pi/a$, confirming that only the first BZ $(-\pi/a, \pi/a]$ contains distinct modes.

From classical wave to quantum phonon. Each normal mode $(\mathbf{k},j)$ is an independent harmonic oscillator of frequency $\omega_j(\mathbf{k})$. Quantizing it gives energy levels $E_{n} = \hbar\omega_j(\mathbf{k})\,(n + \tfrac{1}{2})$. We reinterpret the integer $n$ as the number of phonons in that mode: adding a quantum of energy $\hbar\omega_j(\mathbf{k})$ means creating one phonon of crystal momentum $\hbar\mathbf{k}$. Phonons are bosons — a mode can hold any number — so their thermal population follows the Bose–Einstein distribution $\bar{n}_j(\mathbf{k}) = \big[e^{\hbar\omega_j(\mathbf{k})/k_BT} - 1\big]^{-1}$, the fact that powers all the thermodynamics below.

Acoustic vs. Optical Branches

When the basis contains more than one atom, the $3p$ branches split into two qualitatively different families. The cleanest illustration is the diatomic 1D chain: alternating masses $M_1$ and $M_2$ on a lattice of period $a$, joined by identical springs $C$. There are now two atoms per cell, so $p=2$ and we expect $3p = 6$ branches in 3D, or $2$ branches in this 1D model. Solving the $2\times 2$ dynamical matrix gives

The two signs give two branches with sharply different character:

• Acoustic branch ($\omega_-$). As $k\to 0$, $\omega_- \to v_s

  k

  $ — gapless and linear, exactly like the monatomic chain. Here the two atoms in a cell move in phase (the cell translates as a unit), so this is the long-wavelength sound mode. Every crystal has exactly 3 acoustic branches (one longitudinal, two transverse), guaranteed gapless by the acoustic sum rule.

• Optical branch ($\omega_+$). As $k\to 0$, $\omega_+ \to \sqrt{2C(1/M_1 + 1/M_2)}$ — a finite frequency at zero wavevector. Here the two atoms move out of phase (against each other), so the cell’s centre of mass stays fixed while the bond stretches and compresses. The remaining $3p - 3$ branches are optical. In an ionic crystal these out-of-phase motions create an oscillating electric dipole that couples directly to infrared light — hence “optical.”

In general, for $p$ atoms per primitive cell in 3D there are $3p$ branches: 3 acoustic and $3p-3$ optical. Silicon ($p=2$) has 6 branches; a complex oxide with a large unit cell has dozens. The optical branches sit at higher frequency and, when polar, dominate a crystal’s infrared optical response and its dielectric behaviour.

The Debye and Einstein Models

The full dispersion $\omega_j(\mathbf{k})$ is generally known only numerically, but two classic idealizations capture the thermal physics with a single parameter each. Both target the long-standing puzzle that classical equipartition predicts a constant heat capacity (Dulong–Petit, $C_V = 3Nk_B$) while experiments show $C_V \to 0$ as $T \to 0$.

Einstein model (optical phonons)

Einstein (1907) made the crudest possible assumption: every one of the $3N$ modes has the same frequency $\omega_E$. This is a good caricature of a narrow optical branch, where $\omega_j(\mathbf{k})$ is nearly flat across the BZ. The energy of $3N$ identical quantum oscillators gives

This was the first model to derive the fall-off of $C_V$ at low $T$ from quantization, a landmark result. But it fails quantitatively at low temperature: it predicts an exponential freeze-out $C_V \sim e^{-\theta_E/T}$, because every mode has a finite excitation gap $\hbar\omega_E$. Real solids have gapless acoustic modes, so their heat capacity dies as a power law instead.

Debye model (acoustic phonons)

Debye (1912) fixed the low-$T$ behaviour by focusing on the gapless acoustic modes. He approximated the three acoustic branches by a single linear dispersion $\omega = v_s k$ extending isotropically, and cut off the $\mathbf{k}$-sum at a Debye wavevector $k_D$ chosen so that the sphere of radius $k_D$ contains exactly $3N$ modes:

This defines the Debye frequency $\omega_D = v_s k_D$ and Debye temperature $\theta_D = \hbar\omega_D/k_B$ — a single material constant (typically $100$–$2000\,\mathrm{K}$) that sets the scale separating “cold” quantum behaviour from “hot” classical behaviour. The heat capacity becomes

The two limits are exactly what experiment demands:

• Low temperature ($T \ll \theta_D$): the integral saturates to a constant ($\tfrac{4\pi^4}{15}$), giving the famous Debye $T^3$ law

The power law (not exponential) is a direct fingerprint of gapless, linearly dispersing phonons in 3D. In a metal this lattice $T^3$ adds to the electronic $\gamma T$, so plotting $C_V/T$ against $T^2$ gives a straight line whose slope and intercept separate the two contributions.

• High temperature ($T \gg \theta_D$): every mode is classically excited and $C_V \to 3Nk_B$, recovering Dulong–Petit.

Debye for acoustics, Einstein for optics — use both. Neither model is “right”; each idealizes one part of the spectrum. The acoustic branches, being gapless, dominate the low-temperature heat capacity and are well described by Debye’s linear dispersion. The optical branches, being high-frequency and flat, are well described by one or more Einstein terms and contribute mainly above their characteristic temperature. A practical fit to a real solid’s $C_V(T)$ often superposes a Debye term for the acoustic modes with one Einstein term per optical branch.

Phonon Density of States

To turn dispersion relations into thermodynamics we need the phonon density of states (DOS) $g(\omega)$ — the number of modes per unit frequency interval:

normalized so that $\int_0^\infty g(\omega)\,d\omega = 3N$. Any phonon thermodynamic average becomes a one-dimensional integral over $\omega$ weighted by $g(\omega)$.

The DOS inherits two signature features from the dispersion:

• Low-frequency behaviour. Near $\omega = 0$ the acoustic branches give $g(\omega) \propto \omega^2$ in 3D (the surface area of a sphere of radius $k = \omega/v_s$). This $\omega^2$ density of low-energy modes is the microscopic origin of the Debye $T^3$ law.
• Van Hove singularities. Wherever the group velocity $\nabla_{\mathbf{k}}\,\omega_j(\mathbf{k})$ vanishes — at zone boundaries, band edges, and saddle points — the DOS develops a sharp peak or kink. In 3D these appear as square-root cusps; in lower dimensions they diverge. Measured phonon DOS spectra (by inelastic neutron or X-ray scattering) are dominated by these van Hove peaks, which serve as fingerprints of the dispersion.

Within the two simple models the DOS is explicit:

Debye’s $\omega^2$ ramp captures the acoustic modes; Einstein’s delta function captures a single flat optical branch.

Heat Capacity from the Full Spectrum

With $g(\omega)$ in hand, the lattice internal energy is a sum over all modes of the average oscillator energy (zero-point plus thermally excited):

and the heat capacity is its temperature derivative,

The bracketed factor is the heat capacity of a single oscillator: it equals $k_B$ when $k_BT \gg \hbar\omega$ (classically excited) and is exponentially small when $k_BT \ll \hbar\omega$ (frozen out). So at temperature $T$ only modes with $\hbar\omega \lesssim k_BT$ contribute — raising $T$ progressively “switches on” higher-frequency modes until, above $\theta_D$, all $3N$ are active and $C_V$ saturates at Dulong–Petit. The Debye and Einstein formulas above are just this master integral evaluated with their respective model DOS.

What heat capacity measures. The temperature dependence of $C_V$ is a thermometer for the phonon spectrum. The low-$T$ exponent (here $T^3$) reveals the dimensionality and dispersion of the gapless modes; the value of $\theta_D$ extracted from a fit measures the stiffness of the lattice ($\theta_D \propto v_s \propto \sqrt{C/M}$); and any deviation from a single Debye curve flags optical branches or, in a metal, the electronic $\gamma T$ term. Heat capacity was historically the first quantitative evidence that lattice vibrations are quantized.

Electron–Phonon Coupling

So far the lattice has vibrated in isolation. But the same ionic displacements that constitute a phonon also shift the periodic potential that the conduction electrons feel — so electrons and phonons are inevitably coupled. When an ion moves by $\mathbf{u}_{\mathbf{R}s}$, the crystal potential changes, and to linear order this gives an interaction in which an electron can absorb or emit a phonon, scattering from Bloch state $\mathbf{k}$ to $\mathbf{k}’$ while the lattice gains or loses crystal momentum $\hbar\mathbf{q} = \hbar(\mathbf{k}’-\mathbf{k})$ (modulo $\mathbf{G}$):

where $c^\dagger, c$ create/annihilate electrons, $a^\dagger, a$ create/annihilate phonons, and $g_{\mathbf{q}j}$ is the electron–phonon coupling matrix element. This single interaction controls a remarkable range of observable physics:

• Electrical resistivity of metals. Phonons are the dominant scatterers of conduction electrons at ordinary temperatures. The number of available phonons grows with $T$, so the resistivity rises — linearly at high $T$ ($\rho \propto T$, since the phonon population $\bar{n} \propto T$) and as the Bloch–Grüneisen $\rho \propto T^5$ at low $T$, where both the phonon number and the small-angle scattering geometry suppress momentum relaxation.
• Indirect optical transitions. In an indirect-gap semiconductor like silicon, the band extrema sit at different $\mathbf{k}$, so a photon (which carries negligible momentum) cannot bridge the gap alone. A phonon must supply the momentum difference, making silicon a poor light emitter — a direct consequence of electron–phonon coupling.
• Temperature-dependent band structure. Electron–phonon interaction renormalizes electron energies and lifetimes (the “kinks” seen in ARPES near $E_F$) and shifts band gaps with temperature.
• Polarons. A strongly coupled electron drags a cloud of lattice distortion with it, forming a heavier composite quasiparticle — the polaron — with an enhanced effective mass.

Role in superconductivity

The deepest consequence of electron–phonon coupling is conventional (BCS) superconductivity. An electron moving through the lattice attracts the positive ions, leaving a transient region of excess positive charge in its wake. A second electron is drawn to this region — so the lattice mediates an effective attraction between electrons. Because the ions respond slowly (on the phonon timescale $1/\omega_D \sim 10^{-13}\,\mathrm{s}$) compared to the fast electrons, the attraction is retarded: the second electron arrives long after the first has passed, which lets the attraction overcome the screened Coulomb repulsion.

This phonon-mediated attraction binds electrons of opposite momentum and spin into Cooper pairs, the bosonic objects that condense into the superconducting ground state. Within BCS theory the critical temperature reflects the coupling strength and the relevant phonon energy scale:

where $N(E_F)$ is the electronic DOS at the Fermi level and $V$ measures the (phonon-derived) pairing interaction. Two predictions of this expression are experimental cornerstones:

• Isotope effect. Since $\omega_D \propto M^{-1/2}$, the prefactor predicts $T_c \propto M^{-1/2}$ — heavier isotopes superconduct at lower temperature. Measuring $T_c$ across isotopes of mercury and tin confirmed that phonons are the glue, a decisive clue that led to BCS theory.

• Coupling-strength control. $T_c$ rises with both $N(E_F)$ and the coupling $\lambda \sim N(E_F)\,

  g

  ^2/\omega$. The search for high $T_c$ in phonon-mediated superconductors is largely a search for materials with strong, high-frequency phonons — culminating in the hydrogen-rich hydrides (e.g. $\mathrm{H_3S}$, $\mathrm{LaH_{10}}$) where light hydrogen atoms give enormous $\omega_D$ and near-room-temperature $T_c$ under pressure.

Phonons: heat carriers and superconducting glue, all at once. The same vibrations that carry a crystal’s heat and scatter its electrons (creating resistance) can, at low temperature, mediate the attraction that eliminates resistance entirely. There is no contradiction: above $T_c$ phonons relax electron momentum and cause resistivity; below $T_c$ the retarded phonon-mediated attraction binds electrons into a coherent condensate that flows without dissipation. Phonons are simultaneously the villain of normal-state transport and the hero of conventional superconductivity.

Key Takeaways

• Vibrations are quantized. A crystal’s normal modes, quantized, become phonons — bosonic quasiparticles with crystal momentum $\hbar\mathbf{k}$ and energy $\hbar\omega_j(\mathbf{k})$.
• The dynamical matrix gives the dispersion. Diagonalizing the $3p\times 3p$ dynamical matrix at each $\mathbf{k}$ yields the $3p$ phonon branches of the spectrum.
• Acoustic vs. optical. Three gapless acoustic branches (sound) plus $3p-3$ finite-frequency optical branches (atoms beating within a cell).
• Heat capacity probes the spectrum. The Debye $T^3$ law at low $T$ and Dulong–Petit at high $T$ both follow from counting thermally excited phonon modes.
• Electrons feel the lattice. Electron–phonon coupling sets the temperature-dependent resistivity of metals and enables indirect optical transitions.
• Phonons glue Cooper pairs. The retarded, phonon-mediated attraction binds electrons into Cooper pairs — the microscopic origin of BCS superconductivity.

See Also

• Condensed Matter Physics (Hub) — crystal structure, reciprocal space, band theory, and magnetism.
• Superconductivity, Quantum Hall & Topological Phases — how Cooper pairs and BCS theory build the superconducting state.
• Graduate-Level Formalism & Experiment — second quantization, Green’s functions, and the inelastic-scattering probes that map phonon dispersions.
• Statistical Mechanics — Bose–Einstein statistics and the thermodynamics behind phonon heat capacity.
• Quantum Mechanics — the harmonic oscillator whose quanta are phonons.
• Thermodynamics — heat capacity, thermal expansion, and the macroscopic thermal behaviour of solids.
• Physics Hub — browse all physics topics.