Normal Modes of Continuous and Many-Body Systems

A piano string contains something like $10^{20}$ atoms, every one of them shoving its neighbours — and yet when the hammer falls, the string does not produce noise. It produces a note: one fundamental and a ladder of clean overtones, as if the unimaginable many-body chaos had secretly organized itself into a handful of pure tones. It has. Those tones are normal modes — special collective patterns in which every particle of the system oscillates at one shared frequency — and any motion, however messy, is just a chord built from them. This article scales mode thinking up from two coupled carts to $N$ masses, and then to infinity. The payoff of the climb is a single curve: the dispersion relation $\omega(k)$ , which tells you the frequency of the collective wave at every wavelength. For a uniform chain of masses and springs it reads $\omega = 2\sqrt{k/m}\,|\sin(ka/2)|$ — a formula that quietly contains the speed of sound, the existence of a maximum vibration frequency in any crystal, the difference between phase and group velocity, and, at long wavelengths, the wave equation of a continuous string. The chain is also the front door to twentieth-century physics. Quantize its decoupled mode oscillators and each becomes a phonon; give the chain two alternating masses and the spectrum splits into acoustic and optical branches, explaining why salt crystals drink infrared light. Solid-state physics, molecular spectroscopy, and quantum field theory all begin with the calculation you are about to do.

Key takeaways

A system of $N$ coupled masses has exactly $N$ normal modes — collective motions in which every coordinate oscillates at a single shared frequency; the general motion is a superposition of them.
Finding modes means solving the generalized eigenvalue problem $K\mathbf{a} = \omega^2 M\mathbf{a}$ ; for a uniform chain, translational symmetry hands you the eigenvectors for free: plane waves $e^{ik_n ja}$ .
The whole dynamics collapses to the dispersion relation $\omega(k_n) = 2\sqrt{k/m}\,|\sin(k_n a/2)|$ — linear at long wavelength, saturating at the cut-off $\omega_{\max} = 2\sqrt{k/m}$ at the zone boundary.
As $N \to \infty$ , $a \to 0$ , the chain becomes a continuous string obeying the wave equation $u_{tt} = c^2 u_{xx}$ with $c = a\sqrt{k/m}$ ; fixed ends quantize the standing waves at $\omega_n = n\pi c/L$ .
Boundary conditions count the modes: periodic (Born–von Kármán) conditions allow exactly $N$ wavenumbers in the first Brillouin zone — one mode per degree of freedom, never more.
Two alternating masses split the spectrum into an acoustic branch ( $\omega \to 0$ as $k \to 0$ ) and an optical branch (finite $\omega$ at $k = 0$ ) with a forbidden gap between them.
In normal coordinates the energy is a sum of independent oscillators, $E = \sum_n \tfrac12(\dot Q_n^2 + \omega_n^2 Q_n^2)$ — the classical seed of the phonon.
Key idea: translational symmetry diagonalizes the coupled equations — the modes of a uniform chain are plane waves, and the entire dynamics is captured by the dispersion relation $\omega(k)$ .

The core insight

Translational symmetry diagonalizes the coupled equations: the normal modes of a uniform chain are plane waves $e^{i(k_n ja - \omega t)}$ , and the entire dynamics is captured by the dispersion relation $\omega(k)$ linking frequency to wavenumber.

Where this lives today

Quantize equation (7) and each mode becomes a phonon; the dispersion curves you derived are measured directly, point by point, by inelastic neutron and X-ray scattering at facilities like the ESS and ORNL's SNS — the experimental output is literally the $\omega(k)$ plot of this article with error bars. Engineering now runs the logic in reverse: phononic crystals pattern a material's masses and stiffnesses on purpose to sculpt the dispersion, opening gaps that block sound or heat at chosen frequencies. Products range from acoustic cloaking panels to thermoelectric materials whose engineered phonon gaps strangle heat conduction while leaving electrons free — and, at metre scale, trial seismic metamaterials: buried lattices of resonators designed to put destructive earthquake frequencies inside a forbidden gap. The sharpest 2026 frontier is quantum acoustics. Superconducting qubits are now routinely coupled to individual gigahertz phonon modes of bulk and surface acoustic-wave resonators — single quanta of the chain's normal modes, created, stored, and swapped on demand. Because acoustic wavelengths are five orders of magnitude shorter than electromagnetic ones at the same frequency, phonon modes make compact quantum memories, and mechanical resonators of macroscopic mass prepared in near-ground states have become test masses for probing gravity's interface with quantum mechanics. Every one of those experiments begins by diagonalizing $K\mathbf{a} = \omega^2 M\mathbf{a}$ for a real device — Derivation A with boundary conditions from a lithography mask.

The physical picture

Line up a long row of identical beads joined by identical springs and disturb it at random. Watched bead by bead, the motion is bewildering — energy sloshing everywhere at once, as in the two-oscillator exchange of Energy transfer between coupled oscillators, but multiplied by a hundred. Yet certain special starting patterns produce something utterly calm: displace the beads along a smooth sinusoidal envelope and release, and the entire row breathes in unison, every bead in perfect sync at one pure frequency, the pattern standing still while only its amplitude oscillates. Those patterns are the normal modes, and there are exactly as many of them as there are beads.

Figure

Three normal modes of a bead chain — each is a frozen sinusoidal pattern that oscillates in unison at its own single frequency

Why do smooth sinusoidal patterns get this privilege? Symmetry. In a uniform chain every bead sees an identical environment — same mass, same springs, same neighbours. The equations of motion therefore do not care if you shift the whole chain by one site, and the natural motions must respect that indifference: they are the patterns that shift into themselves up to a phase factor, which is precisely a plane wave $e^{ikja}$ . Symmetry, not luck, diagonalizes the problem — the same reason plane waves diagonalize every translationally invariant system from crystals to quantum fields. Each mode has a wavelength, and the relationship between wavelength and frequency is the chain's fingerprint — the dispersion relation:

Figure

The dispersion relation of a bead chain — linear like a string at long wavelengths, bending over to a maximum frequency at the shortest

Read the curve from left to right. Near the origin it is a straight line: long-wavelength modes do not notice that the chain is grainy, so they behave like waves on a continuous string, all travelling at one speed $c = a\sqrt{k/m}$ — this is where sound comes from. As the wavelength shrinks toward the bead spacing, the curve bends over: short waves travel slower, the medium has become dispersive. And at the shortest wavelength the chain can express — $\lambda = 2a$ , neighbouring beads exactly out of phase — the curve goes flat. Flat means zero group velocity: that pattern is a standing wave that transports no energy at all, and its frequency $\omega_{\max} = 2\sqrt{k/m}$ is the highest note the chain can play. No crystal in the universe vibrates faster than its own version of this cut-off.

From N equations to one relation

For a general small-oscillation system the machinery is the matrix eigenproblem built in Normal modes and dissected in The normal-mode eigenvalue problem: linearizing about equilibrium gives

M\ddot{\mathbf{q}} = -K\mathbf{q} \qquad (1)

where: $\mathbf{q}$ — the column of $N$ displacements from equilibrium; $M$ — the symmetric, positive-definite mass matrix; $K$ — the symmetric stiffness matrix of second derivatives of the potential. Seeking a synchronous solution $\mathbf{q}(t) = \mathbf{a}\,e^{i\omega t}$ , in which every coordinate shares one frequency, turns the differential system into algebra:

K\mathbf{a} = \omega^2 M\mathbf{a} \qquad (2)

where: $\mathbf{a}$ — the mode shape (eigenvector), the frozen amplitude pattern; $\omega^2$ — its eigenvalue. Because $M$ and $K$ are symmetric with $M$ positive-definite, all $N$ eigenvalues are real and non-negative: no coupled mechanical system near stable equilibrium can have a complex or runaway mode frequency. For a handful of masses you solve (2) by determinant; for $N = 10^{23}$ that is hopeless — unless the system has a symmetry that hands you the eigenvectors in advance. The uniform chain has exactly that. Take $N$ identical masses $m$ at spacing $a$ , nearest neighbours joined by springs of stiffness $k$ . Mass $j$ feels only the stretch of its two adjacent springs:

m\ddot u_j = k\,(u_{j+1} - u_j) - k\,(u_j - u_{j-1}) = k\,(u_{j+1} + u_{j-1} - 2u_j) \qquad (3)

where: $u_j$ — the displacement of the $j$ -th mass; the combination $u_{j+1} + u_{j-1} - 2u_j$ — a discrete second derivative, measuring how much mass $j$ is out of line with the average of its neighbours. A mass exactly midway between its neighbours feels nothing; the force is proportional to the local curvature of the displacement pattern. Keep that reading in mind — it is why the continuum limit will produce a second space derivative.

Derivation ADispersion relation of the uniform chain

We are deriving the dispersion relation $\omega(k) = 2\sqrt{k/m}\,|\sin(ka/2)|$ — the single algebraic formula that replaces all $N$ coupled equations of the uniform chain. The strategy is to let symmetry do the diagonalizing: because every mass sees an identical environment, the equations commute with the shift $j \to j+1$ , and the simultaneous eigenfunctions of the shift are plane waves whose phase advances by a fixed $ka$ per site. Substituting that ansatz must collapse the infinite system to one scalar condition relating $\omega$ to $k$ . Watch two things: the symbol collision between the spring stiffness $k$ and the wavenumber $k$ (an unfortunate but universal convention), and the final counting step — boundary conditions restrict the allowed wavenumbers to exactly $N$ values, so no modes are gained or lost in the transformation.

Step 1. Try the plane-wave ansatz, the same frequency on every mass but a phase that winds uniformly along the chain:

u_j(t) = A\,e^{i(kja - \omega t)}

Why this is the right guess: discrete translational symmetry means the dynamics commutes with the shift operator $j \to j+1$ ; the eigenfunctions of the shift are exactly the patterns multiplied by a constant phase $e^{ika}$ under one step — plane waves. Symmetry, not trial and error, picks the ansatz.

Step 2. Substitute into the equation of motion (3). The time derivatives bring down $-\omega^2$ ; each neighbour contributes a phase factor $e^{\pm ika}$ :

-m\omega^2\,A\,e^{i(kja - \omega t)} = k\,A\left(e^{ika} + e^{-ika} - 2\right)e^{i(kja - \omega t)}

Why the system collapses: the common exponential — carrying all the $j$ -dependence — cancels from both sides. One equation now stands for all $N$ ; this is the entire miracle of the method.

Step 3. Simplify the bracket with the half-angle identity $2\cos\theta - 2 = -4\sin^2(\theta/2)$ :

e^{ika} + e^{-ika} - 2 = 2\cos(ka) - 2 = -4\sin^2\!\left(\frac{ka}{2}\right)

Step 4. Solve for the frequency:

\omega^2 = \frac{4k}{m}\,\sin^2\!\left(\frac{ka}{2}\right) \quad\Longrightarrow\quad \omega(k) = 2\sqrt{\frac{k}{m}}\,\left|\sin\!\left(\frac{ka}{2}\right)\right|

Why the absolute value: frequencies are positive by convention; the sign of $k$ only encodes the direction the wave runs.

Step 5. Count the modes with periodic (Born–von Kármán) boundary conditions $u_{j+N} = u_j$ , which force $e^{ikNa} = 1$ :

k_n = \frac{2\pi n}{Na}, \qquad n = 0, 1, \dots, N-1

Why exactly $N$ values: wavenumbers differing by $2\pi/a$ produce identical displacements of the discrete masses, so all distinct physics lives in one window of width $2\pi/a$ — the first Brillouin zone $-\pi/a < k \le \pi/a$ — and the periodicity condition slices that window into precisely $N$ allowed wavenumbers. One mode per degree of freedom, as the eigenproblem (2) demands. *Two quick checks confirm the formula behaves correctly at its limits: as $ka \to 0$ , $\omega \to a\sqrt{k/m}\,|k|$ — linear in $k$ , matching the wave equation derived below; at $k = \pi/a$ , $\sin(\pi/2) = 1$ and $\omega = 2\sqrt{k/m} = \omega_{\max}$ . The mode count of exactly $N$ allowed wavenumbers matches the dimension of the eigenproblem (2), confirming no modes are gained or lost.*

The dispersion relation deserves its own frame:

\omega(k_n) = 2\sqrt{\frac{k}{m}}\,\left|\sin\!\left(\frac{k_n a}{2}\right)\right| \qquad (4)

where: $k_n = 2\pi/\lambda_n$ — the mode wavenumber, an inverse length, living in the first Brillouin zone; $k$ under the square root — the spring stiffness. Two different physical quantities, one unlucky letter; every solid-state course makes this apology. The formula's two ends carry the physics: the linear foot (sound) and the flat top (cut-off). From any dispersion relation two velocities follow, and they agree only when the curve is straight:

v_p = \frac{\omega}{k}, \qquad v_g = \frac{\mathrm{d}\omega}{\mathrm{d}k} = a\sqrt{\frac{k}{m}}\,\cos\!\left(\frac{ka}{2}\right) \qquad (5)

where: $v_p$ — the phase velocity, the speed of an individual crest; $v_g$ — the group velocity, the speed of a wave packet and of the energy it carries. On the linear foot of the curve both equal $c = a\sqrt{k/m}$ and a pulse holds its shape. Elsewhere they part company, and at the zone boundary $v_g = \cos(\pi/2) = 0$ exactly: the shortest wave stands still. A packet built from modes near the cut-off spreads and creeps; this is dispersion in the literal, shape-destroying sense.

The continuum limit: where the wave equation comes from

Zoom out until you can no longer see the beads. Formally: let the spacing $a \to 0$ and the number $N \to \infty$ while the total length and mass stay fixed. The discrete label $j$ becomes a continuous position $x = ja$ , and the displacement list $u_j(t)$ becomes a field $u(x,t)$ .

Go deeper · Derivation B · 1 minFrom the bead chain to the wave equation

Derivation BFrom the bead chain to the wave equation

We are deriving the one-dimensional wave equation $u_{tt} = c^2 u_{xx}$ , with wave speed $c = a\sqrt{k/m}$ , as the long-wavelength limit of the bead chain — establishing that the continuous string is not a separate theory but the coarse-grained chain. The route is a Taylor expansion of the neighbouring displacements about the point $x$ : the discrete second difference in (3) will become a literal second derivative, with corrections that vanish as $a^2$ . Watch what survives the limit: neither $m$ nor $k$ alone remains meaningful (both are properties of vanishing beads), only the combinations mass-per-length and stiffness-times-spacing — the density and the tension.

Step 1. Promote $u_j(t) \to u(x,t)$ with $x = ja$ , and Taylor-expand the neighbours about $x$ :

u_{j\pm 1} = u(x \pm a) = u \pm a\,u_x + \tfrac12 a^2\,u_{xx} \pm \tfrac16 a^3\,u_{xxx} + \cdots

Step 2. Form the discrete second difference. The odd-order terms cancel in the sum:

u_{j+1} + u_{j-1} - 2u_j = a^2\,u_{xx} + O(a^4)

Why the first derivative disappears: the chain's force compares each mass with the average of its neighbours, a symmetric stencil — uniform tilts produce no force, only curvature does. The continuum inherits this: strings feel $u_{xx}$ , not $u_x$ .

Step 3. Substitute into the chain equation (3):

m\,u_{tt} = k\,a^2\,u_{xx}

Step 4. Divide by $m$ and name the constant:

u_{tt} = c^2\,u_{xx}, \qquad c = a\sqrt{\frac{k}{m}}

Why $c$ survives the limit: write $c^2 = ka^2/m = (ka)/(m/a)$ — the numerator $T = ka$ is the effective tension (force per unit strain times spacing) and the denominator $\rho = m/a$ is the linear mass density. Both stay finite as $a \to 0$ , giving the string formula $c = \sqrt{T/\rho}$ in its familiar form. *As a consistency check, the same $c$ appears as the small- $k$ slope of the dispersion relation (4) — the two calculations agree, as they must, since the wave equation is the $ka \ll 1$ shadow of the chain.*

The wave equation and its standing-wave solutions get their own full treatment in The one-dimensional wave equation; here we need only the result that closes our circle. Clamp the string at both ends, $u(0,t) = u(L,t) = 0$ , and separation of variables forces the spatial pattern to fit whole half-wavelengths between the clamps, quantizing the spectrum:

u_n(x,t) = A_n\,\sin\!\left(\frac{n\pi x}{L}\right)\cos(\omega_n t + \phi_n), \qquad \omega_n = \frac{n\pi c}{L} \qquad (6)

where: $n = 1, 2, 3, \dots$ — the mode number, counting antinodes; $A_n, \phi_n$ — set by the initial shape and velocity. The frequencies are evenly spaced — $\omega_2 = 2\omega_1$ , $\omega_3 = 3\omega_1$ — which is precisely why a plucked string sounds musical: its overtones form the harmonic series. Compare the chain: its dispersion curve bends, so its high modes crowd together below the cut-off. Even spacing is a fingerprint of the continuum; a bent spectrum is the fingerprint of grain. Measure a solid's vibration spectrum precisely enough and you can hear that it is made of atoms.

Two atoms per cell: acoustic and optical branches

One more structural move unlocks real crystals. Let two different masses alternate along the chain — $m_1, m_2, m_1, m_2, \dots$ — as in a salt crystal's alternating Na and Cl ions. The unit cell now contains two atoms, so there are two amplitude patterns per wavenumber, and the single dispersion curve splits in two:

Figure

The diatomic chain splits the spectrum — an acoustic branch that becomes sound and an optical branch marooned at high frequency, with a forbidden gap between

Go deeper · Derivation C · 2 minAcoustic and optical branches of the diatomic chain

Derivation CAcoustic and optical branches of the diatomic chain

We are deriving the two-branch dispersion relation of a chain with alternating masses $m_1$ and $m_2$ joined by identical springs $\kappa$ — the boxed $\omega_\pm^2$ formula — and with it the origin of acoustic and optical phonon branches. The strategy repeats the monatomic calculation with one structural upgrade: since the unit cell holds two atoms, the plane-wave ansatz needs two independent amplitudes, one per sublattice, and the substitution yields a 2×2 linear system instead of a scalar identity. Its secular determinant is a quadratic in $\omega^2$ , hence two branches. Watch the two roots at $k = 0$ : one must vanish (rigid translation costs nothing — the acoustic branch) and one must not (the sublattices beating against each other costs energy even at infinite wavelength — the optical branch).

Step 1. Write the equations of motion, letting $u_j$ ride the $m_1$ sublattice and $w_j$ the $m_2$ sublattice, each mass coupled to its two unlike neighbours:

m_1\ddot u_j = \kappa\,(w_j + w_{j-1} - 2u_j), \qquad m_2\ddot w_j = \kappa\,(u_{j+1} + u_j - 2w_j)

Step 2. Insert a plane wave per sublattice, $u_j = U e^{i(kja - \omega t)}$ and $w_j = W e^{i(kja - \omega t)}$ . Why two amplitudes: translational symmetry now only guarantees the pattern repeats cell to cell; within the cell, the two atoms are free to move with different amplitudes and phases, encoded in the complex ratio $W/U$ .

Step 3. The substitution collapses the chain to a 2×2 homogeneous system:

\begin{pmatrix} 2\kappa - m_1\omega^2 & -\kappa\,(1 + e^{-ika}) \\ -\kappa\,(1 + e^{ika}) & 2\kappa - m_2\omega^2 \end{pmatrix} \begin{pmatrix} U \\ W \end{pmatrix} = 0

Step 4. Nontrivial solutions require the secular determinant to vanish. With $|1 + e^{ika}|^2 = 2 + 2\cos ka = 4\cos^2(ka/2)$ :

(2\kappa - m_1\omega^2)(2\kappa - m_2\omega^2) - 4\kappa^2\cos^2\!\left(\frac{ka}{2}\right) = 0

Why a quadratic in $\omega^2$ : two atoms per cell, two branches — the mode count per wavenumber always equals the number of degrees of freedom in the unit cell.

Step 5. Solve the quadratic:

\omega_\pm^2 = \kappa\!\left(\frac{1}{m_1} + \frac{1}{m_2}\right) \pm \kappa\sqrt{\left(\frac{1}{m_1} + \frac{1}{m_2}\right)^{\!2} - \frac{4\sin^2(ka/2)}{m_1 m_2}}

Why the branch characters: at $k \to 0$ the minus root gives $\omega_- \to 0$ with $U = W$ — both sublattices gliding together, a long sound wave. The plus root gives $\omega_+^2 \to 2\kappa(1/m_1 + 1/m_2)$ with $m_1 U = -m_2 W$ — the sublattices beating in opposition about a fixed center of mass. In an ionic crystal that counter-motion of opposite charges is an oscillating dipole, which couples directly to light: hence "optical." *Two limiting cases confirm the formula: setting $m_1 = m_2 = m$ merges the two branches into the folded monatomic dispersion, closing the gap; at $k = \pi/a$ the two roots are $\sqrt{2\kappa/m_1}$ and $\sqrt{2\kappa/m_2}$ , which are distinct — and hence leave a gap — whenever $m_1 \neq m_2$ .*

Between $\omega_-(\pi/a)$ and $\omega_+(\pi/a)$ lies a forbidden gap: no travelling mode exists at those frequencies. Shake the chain inside the gap and the disturbance dies exponentially with distance instead of propagating. Mass-contrast has turned the chain into a mechanical band-stop filter — the design principle behind phononic crystals and seismic shields, and the classical ancestor of the electronic band gap.

The energy ledger and the road to phonons

Project any motion onto the eigenvectors and define one normal coordinate $Q_n$ per mode (the mass-weighted amplitude of pattern $n$ in the current motion). In these variables the energy of the entire many-body system falls apart into independent pieces:

E = \sum_{n=1}^{N} \tfrac12\left(\dot Q_n^2 + \omega_n^2 Q_n^2\right) \qquad (7)

where: each summand — the energy of a single fictitious harmonic oscillator at frequency $\omega_n$ ; no cross terms $Q_m Q_n$ survive. This is a stronger statement than energy conservation: each mode's energy is separately conserved, and modes never exchange energy in the harmonic approximation. (What weak nonlinearity does to that statement is the Fermi-Pasta-Ulam surprise told in Energy transfer between coupled oscillators.) Equation (7) is also the door to quantum theory: quantize each independent oscillator and its energy comes in lumps $\hbar\omega_n$ — a mode with $n$ quanta "contains $n$ phonons." Every sentence of solid-state physics that mentions a phonon is leaning on the classical decoupling you see here.

Which equation answers which question

you want	use	notes
mode frequencies of any small-oscillation system	$K\mathbf{a} = \omega^2 M\mathbf{a}$ , eq. (2)	general but brute-force
chain mode frequencies by symmetry	$\omega = 2\sqrt{k/m}\,\vert\sin(ka/2)\vert$ , eq. (4)	plane waves, Brillouin zone
speed of long waves / sound	$c = a\sqrt{k/m} = \sqrt{T/\rho}$	slope of the dispersion at $k = 0$
energy transport speed	$v_g = \mathrm{d}\omega/\mathrm{d}k$ , eq. (5)	zero at the zone boundary
string standing-wave spectrum	$\omega_n = n\pi c/L$ , eq. (6)	evenly spaced — the harmonic series
two-atom-basis branches	$\omega_\pm^2$ boxed in Derivation C	acoustic + optical, gap between
total energy by modes	$E = \sum_n \tfrac12(\dot Q_n^2 + \omega_n^2 Q_n^2)$ , eq. (7)	modes never trade energy

Limiting cases

Case	Result	Meaning
long wavelength $ka \ll 1$	$\omega \approx c\,\vert k\vert$ , $c = a\sqrt{k/m}$	non-dispersive string; sound emerges; $v_p = v_g$
zone boundary $k = \pi/a$	$\omega = \omega_{\max} = 2\sqrt{k/m}$ , $v_g = 0$	neighbours anti-phase; standing wave; no energy flow
continuum $N \to \infty$ , $a \to 0$	$u_{tt} = c^2 u_{xx}$ , $\omega_n = n\pi c/L$	wave equation; evenly spaced harmonics
two-mass basis	branches $\omega_\pm$ with a gap	acoustic + optical; gap frequencies evanescent
$m_1 = m_2$ in the diatomic chain	gap closes; folded monatomic curve	the gap is pure mass contrast

Common trapThe spring constant k and the wavenumber k are the same kind of thing

They are not — and the shared letter is this subject's most reliable source of nonsense answers. The spring constant $k$ is a stiffness measured in newtons per metre; the wavenumber $k = 2\pi/\lambda$ is an inverse length measured in radians per metre. They meet in one formula, $\omega = 2\sqrt{k/m}\,|\sin(ka/2)|$ , where the first $k$ is stiffness and the second is wavenumber; swap them mentally and every dimensional check fails. The second classic confusion is between the two velocities: the phase velocity $\omega/k$ tracks a single crest, the group velocity $\mathrm{d}\omega/\mathrm{d}k$ tracks the packet and the energy. They coincide only where the dispersion is a straight line through the origin. At the zone boundary the chain has $v_g = 0$ with $v_p$ perfectly finite — a pattern whose crests "move" while its energy goes nowhere. Quoting $v_p$ as the speed of energy transport there is not an approximation; it is wrong by a factor of infinity.

Worked examples

1 · A bead chain you could build (confidence builder)

A chain of beads, each of mass $m = 0.10\ \mathrm{kg}$ , sits at spacing $a = 0.20\ \mathrm{m}$ , joined by springs of stiffness $k = 25\ \mathrm{N/m}$ . Find the sound speed, the cut-off frequency, and — for the mode with $ka = \pi/2$ — the phase and group velocities.

Step 1. The frequency scale: $\sqrt{k/m} = \sqrt{25/0.10} = 15.8\ \mathrm{rad/s}$ .

Step 2. Sound speed (long-wave slope): $c = a\sqrt{k/m} = 0.20 \times 15.8 = \boxed{3.16\ \mathrm{m/s}}$ — a wave you could jog alongside.

Step 3. Cut-off: $\omega_{\max} = 2\sqrt{k/m} = 31.6\ \mathrm{rad/s}$ , i.e. $f_{\max} = 31.6/2\pi = \boxed{5.03\ \mathrm{Hz}}$ . Shake the end of the chain faster than five times a second and no travelling wave carries it away — the disturbance stays near your hand and dies off exponentially.

Step 4. Mode at $ka = \pi/2$ (wavelength $\lambda = 4a = 0.80$ m): $\omega = 31.6 \times \sin(\pi/4) = 22.4\ \mathrm{rad/s}$ , and $k = \pi/(2a) = 7.85\ \mathrm{rad/m}$ , so $v_p = \omega/k = 22.4/7.85 = \boxed{2.85\ \mathrm{m/s}}$ .

Step 5. Group velocity: $v_g = c\cos(ka/2) = 3.16 \times \cos(\pi/4) = \boxed{2.24\ \mathrm{m/s}}$ . Already 30% below $c$ , and 21% below $v_p$ — halfway across the zone, dispersion is no longer a small correction.

2 · From a guitar string to why salt blocks infrared (exam level)

(a) A guitar's A string has length $L = 0.648\ \mathrm{m}$ , linear density $\rho = 5.2\ \mathrm{g/m}$ , and fundamental $f_1 = 110\ \mathrm{Hz}$ . Find the wave speed and tension. (b) Model rock salt as a diatomic chain of Na ( $m_1 = 23.0\,\mathrm{u}$ ) and Cl ( $m_2 = 35.5\,\mathrm{u}$ ) ions with effective spring constant $\kappa = 12\ \mathrm{N/m}$ . Find the optical-branch frequency at $k = 0$ and the vacuum wavelength of light it absorbs.

Step 1. The fundamental of a fixed-fixed string is $f_1 = c/2L$ , so $c = 2Lf_1 = 2 \times 0.648 \times 110 = \boxed{143\ \mathrm{m/s}}$ .

Step 2. From $c = \sqrt{T/\rho}$ : $T = \rho c^2 = 0.0052 \times (142.6)^2 = \boxed{106\ \mathrm{N}}$ — about 11 kg-force, which is why guitar necks need truss rods.

Step 3. Overtones: $f_n = nf_1 = 220, 330, 440\ \mathrm{Hz}\dots$ — evenly spaced because the string's dispersion is linear. (A real string's slight stiffness bends the dispersion and sharpens high overtones — piano tuners call it inharmonicity, and it is exactly the chain-versus-string correction of this article.)

Step 4. The optical branch at $k = 0$ : $\omega_+ = \sqrt{2\kappa\left(\tfrac{1}{m_1} + \tfrac{1}{m_2}\right)}$ . With $m_1 = 23.0 \times 1.66\times10^{-27} = 3.82\times10^{-26}\ \mathrm{kg}$ and $m_2 = 35.5 \times 1.66\times10^{-27} = 5.89\times10^{-26}\ \mathrm{kg}$ : $\tfrac{1}{m_1} + \tfrac{1}{m_2} = 2.62\times10^{25} + 1.70\times10^{25} = 4.32\times10^{25}\ \mathrm{kg^{-1}}$ , so $\omega_+ = \sqrt{2 \times 12 \times 4.32\times10^{25}} = \boxed{3.2\times10^{13}\ \mathrm{rad/s}}$ — about 5.1 THz.

Step 5. The light it absorbs: $\lambda = 2\pi c_{\mathrm{light}}/\omega_+ = 2\pi \times 3.00\times10^8 / 3.2\times10^{13} = \boxed{59\ \mathrm{\mu m}}$ — deep infrared, and strikingly close to rock salt's measured reststrahlen band near 61 μm. A two-mass chain with one effective spring constant predicts, to within a few percent, which color of infrared a salt crystal refuses to transmit.

Origins · 1 minMax Born and how this was discovered

Max Born

1882–1970 · Göttingen → Edinburgh, Nobel Prize 1954

Born is remembered for the probabilistic interpretation of the quantum wavefunction, but his first great edifice was classical: lattice dynamics. In 1912, with the young Theodore von Kármán, he wrote down the equations of a crystal as a periodic array of masses and springs, introduced the periodic boundary conditions that now bear their names, and derived the mode counting and dispersion relations of this article — independently of, and more generally than, Einstein's and Debye's contemporaneous models of specific heat. His Göttingen school then made the harmonic crystal a factory of results: his assistants and students included Heisenberg, Pauli, and Fermi, and his 1954 treatise with Kun Huang, Dynamical Theory of Crystal Lattices, remains the standard scripture of the field. Driven from Germany in 1933, he rebuilt his school in Edinburgh. The Born-von Kármán boundary condition you used in Derivation A is the workhorse of every band-structure calculation run today.

History

1753

— Daniel Bernoulli asserts that a vibrating string's general motion is a superposition of its simple modes — the first statement of the normal-mode principle, decades before the mathematics could justify it.

1822

— Joseph Fourier's Théorie analytique de la chaleur supplies that justification: arbitrary functions expand in trigonometric series, making mode superposition a theorem.

1877

— Lord Rayleigh's Theory of Sound systematizes normal modes, mode counting, and perturbation of vibrating systems — still the classical reference.

1912

— Max Born and Theodore von Kármán write the dynamics of a crystal lattice with periodic boundary conditions, deriving dispersion relations and founding solid-state physics.

1929

— Felix Bloch generalizes the plane-wave-times-periodic-function ansatz to electrons in crystals, exporting the chain's symmetry argument to quantum mechanics.

Going further

Texts. Taylor, Classical Mechanics, ch. 11 (Coupled Oscillators and Normal Modes) · Goldstein, Poole & Safko, Classical Mechanics, 3rd ed., ch. 6 and ch. 13 (continuous systems) · Landau & Lifshitz, Mechanics, §23-28 · Morin, Introduction to Classical Mechanics, ch. 4 and appendix on coupled oscillators · Ashcroft & Mermin, Solid State Physics, ch. 22 (classical theory of the harmonic crystal). The key results of this article can be traced through four self-contained calculations: (1) substituting $u_j=Ae^{i(kja-\omega t)}$ into $m\ddot u_j=k(u_{j+1}+u_{j-1}-2u_j)$ and applying $2\cos\theta-2=-4\sin^2(\theta/2)$ yields $\omega=2\sqrt{k/m}|\sin(ka/2)|$ ; (2) a Taylor expansion of the discrete second difference gives the continuum limit $u_{tt}=c^2u_{xx}$ with $c=a\sqrt{k/m}$ ; (3) separation of variables with fixed ends produces the string spectrum $\omega_n=n\pi c/L$ ; (4) the diatomic secular determinant is a quadratic in $\omega^2$ whose two roots $\omega_\pm$ satisfy $\omega_-\to0$ and $\omega_+^2\to2\kappa(1/m_1+1/m_2)$ at $k=0$ . The worked-example numbers ( $c = 3.16$ m/s and $f_{\max} = 5.03$ Hz for the bead chain; $T = 106$ N for the string; $\omega_+ = 3.2\times10^{13}$ rad/s and $\lambda = 59\ \mathrm{\mu m}$ for the salt model) follow directly from these formulas.

Builds on: Small oscillations about equilibrium · Simple harmonic motion · Complex exponential representation of oscillations · Energy in simple harmonic motion. Leads to: Driven oscillations · Resonance · The one-dimensional wave equation · Phonons and lattice dynamics.

If the modes never exchange energy, how does a plucked string ever reach a steady sound or decay?

Two separate answers. First, the pluck itself distributes energy across many modes at the initial instant — a triangular pluck shape has a Fourier series full of harmonics, so no exchange is needed to make them all sound; they were all funded at $t = 0$ . Second, the decay you hear is not modes trading energy with each other but each mode separately losing energy to things outside the harmonic model: air drag, internal friction in the string, and energy leaking through the bridge into the soundboard (which is precisely what makes the guitar audible). Within the ideal harmonic chain, the no-exchange statement is exact. Add nonlinearity and it fails in fascinating slow ways — the Fermi-Pasta-Ulam recurrences — but for a real string those effects are far smaller than the dissipative ones.

Where exactly does the "maximum frequency" of the chain come from physically? A single mass on a spring has no such limit

A single oscillator's frequency is set by its own $\sqrt{k/m}$ , and you can drive it at any frequency you like — it just responds off-resonance. The chain's cut-off is about propagation: for a wave to travel, each mass must hand the disturbance to its neighbour, and the fastest possible handoff pattern is neighbours moving exactly out of phase ( $\lambda = 2a$ ). In that pattern each mass feels the maximum restoring force geometry allows — both springs fully opposed — giving $\omega_{\max} = 2\sqrt{k/m}$ , exactly twice the single-mass value. Ask for a still-shorter wavelength and the discrete chain cannot even represent it: a "wave" with $\lambda < 2a$ sampled at the mass positions is indistinguishable from a longer-wavelength one (the same aliasing that limits digital audio). The cut-off is the physics of graininess, which is why continuous media, having none, have no cut-off.

Is the wavenumber k physically meaningful between the allowed k_n values, or only at them?

The dispersion curve $\omega(k)$ is a property of the infinite chain's local dynamics and is smooth and meaningful at every $k$ — it answers "if a plane wave of this wavelength existed, at what frequency would it oscillate?" The boundary conditions then select which wavelengths actually fit your finite system: $N$ discrete points $k_n$ sprinkled along the curve. Make the chain longer and the points crowd together, approaching a continuum; that is why solid-state physics freely draws $\omega(k)$ as a curve even though a real crystal has finitely many modes. The distinction matters when you count things — heat capacity, for instance, needs the actual number of modes per frequency interval, which comes from the spacing $2\pi/(Na)$ of the allowed points, not from the curve alone.

Why do periodic boundary conditions get used so much when real crystals obviously have surfaces, not wrap-around ends?

Convenience plus a theorem-grade excuse. Periodic (Born-von Kármán) conditions keep the system perfectly translation-invariant, so the modes stay clean travelling plane waves and the counting is trivial — $N$ wavenumbers evenly spaced in the Brillouin zone. Fixed or free ends break the symmetry and give standing-wave modes with slightly different wavenumbers. The excuse: for bulk properties, the difference is a surface effect, of relative size $\sim 1/N$ per dimension — for $10^8$ atoms across a crystal, parts in a hundred million. The mode density, which controls thermodynamics, is identical in the limit. Where the choice genuinely matters is when the surface is the physics: surface phonon modes, edge states, and nanoscale systems where $1/N$ is not small. There you must solve with the real boundaries.

What does it mean for a mode in the diatomic gap to "decay evanescently" — where does the energy go?

Nowhere — that is the point. If you shake one end of the chain at a gap frequency, the secular equation has no real wavenumber solution; the mathematics instead returns a complex $k$ , meaning a displacement pattern whose amplitude falls off exponentially with distance from your hand. No travelling wave forms, so no energy is transported down the chain; the near region oscillates in place, storing and returning energy to the driver each cycle, like a mass driven far off resonance. The chain behaves as a reactive load, not a transmission line. This is precisely how band-stop vibration filters and phononic shields work: they do not absorb the offending frequency (absorption needs dissipation), they refuse to conduct it, reflecting the energy back toward the source.

The article says short waves travel slower on the chain. Doesn't a stiffer, "harder" medium usually carry high frequencies faster?

The intuition you are quoting is about material stiffness (steel carries sound faster than rubber), which sets the overall scale $c = a\sqrt{k/m}$ — and indeed stiffer springs raise every speed on the chain proportionally. Dispersion is a different, geometric effect within one material: it compares wavelengths to the lattice spacing. A long wave barely bends between neighbouring masses, so each mass feels nearly the continuum restoring force. A short wave bends sharply, but the discrete second difference $u_{j+1} + u_{j-1} - 2u_j$ saturates — it can never exceed what fully anti-phase neighbours produce — so the restoring force stops growing as fast as the continuum $k^2$ would demand, and $\omega$ falls below $ck$ . Slower short waves are thus a signature of discreteness, not softness. (Media can also show the opposite, "anomalous" dispersion — bending stiffness in real piano strings raises high-mode speeds, sharpening overtones.)

How literally should I take "a phonon is a quantized normal mode"? Is a phonon a particle or a wave pattern?

Take the decoupling literally and the particle language operationally. Classically, equation (7) says the chain is exactly equivalent to $N$ independent harmonic oscillators, one per mode. Quantum mechanics then quantizes each oscillator: mode $n$ can hold energy only in steps of $\hbar\omega_n$ above its ground state. "The mode holds three quanta" is abbreviated to "there are three phonons in mode $n$ ." Each phonon is thus a delocalized excitation of a collective pattern spanning the whole crystal — nothing like a marble. The particle picture earns its keep because these quanta scatter off electrons, neutrons, and each other with conserved energy $\hbar\omega$ and crystal momentum $\hbar k$ , bookkeeping exactly like collisions. Localized, particle-ish phonon wave packets are superpositions of many modes, moving — inevitably — at the group velocity from equation (5).

For a molecule rather than a chain, how do I know how many normal modes to expect?

Count degrees of freedom and subtract the motions that cost no potential energy. $N$ atoms in 3D have $3N$ coordinates. Three combinations are rigid translations and (for a nonlinear molecule) three are rigid rotations — these leave all interatomic distances unchanged, so the stiffness matrix gives them $\omega = 0$ ; they are "modes" only formally. That leaves $3N - 6$ genuine vibrational normal modes ( $3N - 5$ for a linear molecule, which has one less rotation). Water: $3\times3 - 6 = 3$ modes — symmetric stretch, asymmetric stretch, bend — exactly the three infrared bands you see. CO₂, linear: $3\times3 - 5 = 4$ modes, with the bend doubly degenerate. The same subtraction happens in the chain with free ends: its $k = 0$ "mode" is a rigid translation at zero frequency, which is also why the acoustic branch must start at $\omega = 0$ .

Normal Modes of Continuous and Many-Body Systems

Where this lives today

The physical picture

From N equations to one relation

The continuum limit: where the wave equation comes from

Two atoms per cell: acoustic and optical branches

The energy ledger and the road to phonons

Which equation answers which question

Limiting cases

Worked examples

1 · A bead chain you could build (confidence builder)

2 · From a guitar string to why salt blocks infrared (exam level)

Going further

Before this

Where this leads