Discrete Loaded String and the Continuum Limit

Almost every classical field — a vibrating string, sound in a solid, the displacement of an elastic rod — can be understood as the continuum limit of a microscopic mechanical lattice. The loaded string is the cleanest such model: a finite chain of identical masses joined by springs or strung along a line under tension. Because the system has only finitely many degrees of freedom, we can solve it exactly with the tools of coupled-oscillator theory, and then watch the algebra reorganise itself into the calculus of fields as the spacing shrinks. The central observation is structural. The force on a given bead depends on the differences of displacements between it and its two neighbours. The combination $y_{j+1}-2y_j+y_{j-1}$ is, up to a factor of $a^2$ , the finite-difference approximation to the second derivative $\partial_x^2 y$ . So the discrete Newton equations are not merely like the wave equation — in the limit $a\to0$ they literally become it, provided we hold the physically meaningful macroscopic quantities (mass density $\mu=m/a$ and tension $T$ ) fixed while $m$ and $a$ individually vanish. Two parallel routes lead to the continuum. The Newtonian route expands the neighbour displacements in a Taylor series and discards higher-order terms in $a$ . The Lagrangian route rewrites the kinetic and potential energy sums as Riemann sums that converge to integrals, turning the discrete Lagrangian into $\int \mathcal{L}\,\mathrm{d}x$ with a Lagrangian density. The second route is more powerful because it survives into nonlinear, relativistic, and higher-dimensional generalisations, and because applying the Euler–Lagrange machinery to the density reproduces the wave equation automatically. Studying the discrete system before the limit is not pedantic bookkeeping: it reveals what the continuum hides. The lattice has a finite number of normal modes, a maximum (cutoff) frequency, and a dispersion relation that is nonlinear in $k$ . Real crystals behave this way; the wave equation is only the long-wavelength face of a richer microscopic theory. The loaded string is therefore both the historical and the conceptual bridge from particle mechanics to field theory, to linear elasticity, and ultimately to the relativistic field theories described in the action-principle capstone.

Key takeaways

A loaded string is $N$ point masses of mass $m$ spaced a distance $a$ apart along a line under tension $T$ ; transverse displacements obey a coupled set of $N$ Newton equations — the $N\to\infty$ generalisation of the two-mass systems solved in the coupled-oscillators article.
The equation of motion for bead $j$ , $m\ddot{y}_j = T(y_{j+1}-2y_j+y_{j-1})/a$ , is exactly the second difference of the displacement — the discrete analogue of a second spatial derivative.
Taking $a\to 0$ with the linear mass density $\mu=m/a$ and tension $T$ held fixed turns the second difference into $\partial^2 y/\partial x^2$ , yielding the one-dimensional wave equation with speed $c=\sqrt{T/\mu}$ .
The discrete normal-mode dispersion $\omega(k)=2\sqrt{T/(ma)}\,|\sin(ka/2)|$ reduces to the linear continuum law $\omega=ck$ only for long wavelengths $ka\ll 1$ ; the lattice has a maximum frequency $\omega_{\max}=2\sqrt{T/(ma)}$ and is dispersive.
The discrete Lagrangian becomes a spatial integral of a Lagrangian density $\mathcal{L}=\tfrac12\mu\dot y^2-\tfrac12 T y'^2$ , the gateway to classical field theory.
This single example shows how fields, the wave equation, dispersion, and continuum elasticity all emerge from the ordinary mechanics of many coupled oscillators.
Key idea: A second spatial difference of bead displacements, $y_{j+1}-2y_j+y_{j-1}$ , equals $a^2\,\partial_x^2 y$ to leading order; holding $\mu=m/a$ and $T$ fixed as $a\to0$ converts the coupled ODEs into the wave equation with $c=\sqrt{T/\mu}$ .

Where this lives today

The loaded string is not a museum piece; it is the working model of every crystal lattice. The dispersion $\omega\propto|\sin(ka/2)|$ , the acoustic branch, the Brillouin-zone edge with $v_g=0$ — all are measured daily by inelastic neutron and X-ray scattering, and the "beads" of 2026-era phononic engineering are literal: phononic crystals and acoustic metamaterials pattern millimetre-to-micron lattices to sculpt exactly this dispersion, opening band gaps that block sound the way the zone boundary blocks short lattice waves. Carbon-nanotube and graphene nanomechanical resonators push the same mode physics to the molecular scale, where a resonator is a loaded string with a countable number of unit cells and the discrete corrections of this article are the signal, not the error. The other living face is computational. Every finite-difference solver of the wave equation — seismology codes, room-acoustics simulators, the numerical-relativity grids behind gravitational-wave templates — is a loaded string by construction, and its leading inaccuracy is precisely the $\omega\approx ck[1-(ka)^2/24]$ lattice dispersion derived here. Grid-resolution rules of thumb ("at least 10–20 points per wavelength") are direct restatements of $ka\ll1$ , and the worked example above is, in miniature, the error budget of such a code.

The physical picture

Picture beads on a taut, massless string, each free to move transversely. If a bead sits exactly on the straight line connecting its two neighbours, the two tension pulls cancel and the net transverse force vanishes; it is only the curvature of the bead profile that pulls a bead back. The discrete measure of curvature is precisely $y_{j+1}-2y_j+y_{j-1}$ — the amount by which the middle bead overshoots the average of its neighbours — which is why this combination governs the dynamics and why it becomes a second derivative in the limit.

Figure

The hero picture — a bead that rises above the chord joining its neighbours feels a net restoring pull; zooming out, the same chain reads as a smooth string with density mu = m/a and tension T

Why must $\mu=m/a$ stay fixed? Imagine zooming out from the chain. From far away you cannot resolve individual beads; you only perceive how much mass sits in a centimetre of string. That ratio — the linear density — is the physical macroscopic input, together with the tension. The microscopic numbers $m$ and $a$ separately are unobservable in the smooth limit, so only their ratio (and $T$ ) can enter the wave speed. The lattice imposes a shortest meaningful wavelength: nothing shorter than twice the spacing $2a$ can be represented by discrete beads, because a profile that flips sign from bead to bead is already the most wiggly shape the chain can hold. This is the origin of the Brillouin-zone cutoff and the maximum frequency $\omega_{\max}=2\sqrt{T/(ma)}$ . The continuum, having no smallest length, has no cutoff — which is exactly why naive continuum field theories sometimes need regularisation at short distances.

Figure

The payoff picture — the exact lattice dispersion bends away from the straight continuum line and saturates at the zone boundary, where neighbouring beads move in antiphase and the group velocity vanishes

The same construction with longitudinal springs gives a chain whose continuum limit is an elastic rod: $T$ is replaced by Young's modulus times cross-section, and the transverse wave speed becomes the speed of sound. Thus one toy model unifies vibrating strings, sound, and the foundations of field theory. The two-bead and three-bead versions of this chain are solved completely in the coupled-oscillators and normal-modes articles; here we let $N$ grow without bound and watch particle mechanics turn into field theory.

From beads to the wave equation

Everything starts from the force law for one bead. For small transverse displacements the tension magnitude is constant along the string, and the net transverse force on bead $j$ is the tension times the difference between the slope on its right and the slope on its left — each segment pulls the bead along its own direction, and only the transverse components survive at small angles. With slopes $(y_{j+1}-y_j)/a$ and $(y_j-y_{j-1})/a$ this gives

m\,\ddot{y}_j = \frac{T}{a}\big(y_{j+1}-2y_j+y_{j-1}\big),\qquad j=1,\dots,N \qquad (1)

where: $y_j(t)$ — transverse displacement of the $j$ -th bead; $m$ — mass of each bead; $a$ — equilibrium spacing; $T$ — tension in the connecting string. The right-hand side is the second difference of the displacement profile: it vanishes when the three beads are collinear, is negative when bead $j$ crests above its neighbours' average, and so always drives the profile back toward straightness. Equation (1) is $N$ coupled linear oscillators — exactly the structure diagonalised for $N=2$ in the coupled-oscillators article, and we will not re-derive that machinery here. To find the modes of the infinite chain we try a travelling lattice wave,

y_j(t)=A\,e^{i(kja-\omega t)} \qquad (2)

where: $A$ — complex amplitude; $k$ — wavenumber; $x_j=ja$ — equilibrium position of bead $j$ ; $\omega$ — mode angular frequency. The ansatz respects the discrete translation symmetry of the chain: shifting $j\to j+1$ multiplies the whole pattern by the fixed phase $e^{ika}$ , so if one bead oscillates harmonically, all do.

Derivation AExact dispersion relation of the bead lattice

We derive the exact relation between frequency and wavenumber for the infinite loaded string. The strategy is direct substitution: insert the plane-wave ansatz (2) into the equation of motion (1) and use the discrete translation structure — each neighbour contributes a phase factor $e^{\pm ika}$ — to reduce the coupled system to a single algebraic condition. Watch that no small- $ka$ approximation is made anywhere: the result is exact at every wavelength the lattice can represent, which is what makes the later comparison with the continuum honest.

Step 1. Substitute $y_j=A\,e^{i(kja-\omega t)}$ into (1). Time derivatives bring down $-\omega^2$ ; the neighbours acquire phases:

-m\omega^2=\frac{T}{a}\big(e^{ika}-2+e^{-ika}\big)

Why: the common factor $A\,e^{i(kja-\omega t)}$ divides out of every term — this is precisely why the plane wave diagonalises the chain: it is an eigenvector of the shift operation.

Step 2. Combine the exponentials with $e^{ika}+e^{-ika}=2\cos ka$ :

m\omega^2=\frac{2T}{a}\,(1-\cos ka)

Why: the sign flip makes both sides manifestly positive — a stability statement: every mode of the chain oscillates rather than grows.

Step 3. Use the half-angle identity $1-\cos ka=2\sin^2(ka/2)$ and take the positive root:

\omega(k)=2\sqrt{\frac{T}{ma}}\,\left|\sin\frac{ka}{2}\right|

Why: the half-angle form exposes the two limits at a glance — linear for $ka\ll1$ , saturating at $ka=\pi$ . *Why: the two limits are immediate — for $ka\ll1$ , $\sin(ka/2)\approx ka/2$ gives $\omega\approx\sqrt{T/(ma)}\,ka=\sqrt{T/\mu}\,k=ck$ , recovering the continuum law; at $k=\pi/a$ the frequency reaches $\omega_{\max}=2\sqrt{T/(ma)}$ , which also follows directly from (1) with the antiphase substitution $y_{j+1}=-y_j$ .*

The dispersion relation is the single most information-dense statement about the chain:

\omega(k)=2\sqrt{\frac{T}{ma}}\,\left|\sin\!\frac{ka}{2}\right| \qquad (3)

where: $\omega(k)$ — frequency of the lattice mode with wavenumber $k$ ; only $|k|\le\pi/a$ (the first Brillouin zone) labels distinct motions, because adding $2\pi/a$ to $k$ changes no bead's phase. Read physically: long waves ride the linear part and feel a smooth string; short waves near $k=\pi/a$ feel the graininess, slow down, and finally stand still. Two consequences follow by inspection. Differentiating (3) gives the group velocity $v_g(k)=\mathrm{d}\omega/\mathrm{d}k=c\cos(ka/2)$ with $c=\sqrt{T/\mu}$ : energy transport is fastest for long waves and ceases entirely at the zone boundary, where the mode is a standing antiphase oscillation. And confining the chain between fixed walls quantises $k$ to $k_n=n\pi/[(N+1)a]$ , giving the finite mode spectrum

\omega_n=2\sqrt{\frac{T}{ma}}\,\sin\!\frac{n\pi}{2(N+1)},\quad n=1,\dots,N \qquad (4)

where: $N$ — number of beads; $n$ — mode index. Exactly $N$ modes exist — no more coordinates, no more modes. The sine-shaped standing-wave patterns that carry these frequencies are the discrete samples of the continuum mode shapes derived in the standing-waves article; the counting, node structure, and orthogonality all carry over.

The continuum limit, twice

The limit can be taken at the level of the equation of motion or at the level of the action. Both must agree — and their agreement is the first consistency test of classical field theory.

Go deeper · Derivation B · 1 minNewtonian continuum limit via Taylor expansion

Derivation BNewtonian continuum limit via Taylor expansion

We show explicitly how the second difference becomes a second derivative and the wave equation emerges with the correct speed. The strategy is to promote the discrete index to a continuous coordinate — write $y_j(t)\to y(x,t)$ with $x=ja$ — and Taylor-expand the neighbour displacements about $x$ , keeping careful track of which powers of $a$ survive when $\mu=m/a$ and $T$ are held fixed. The thing to watch is the order of limits: $m$ and $a$ must go to zero together, tied by the fixed density, or the wave speed degenerates.

Step 1. Expand both neighbours to fourth order:

y_{j\pm1}=y(x\pm a)=y\pm a\,\partial_x y+\tfrac12 a^2\partial_x^2 y\pm\tfrac16 a^3\partial_x^3 y+\tfrac1{24}a^4\partial_x^4 y+\cdots

Why: we need one order beyond the leading term to know the size of what we are about to throw away.

Step 2. Form the second difference; odd orders cancel by the symmetry $+a\leftrightarrow-a$ :

y_{j+1}-2y_j+y_{j-1}=a^2\,\partial_x^2 y+\tfrac{1}{12}a^4\,\partial_x^4 y+\cdots

Why: the cancellation of odd derivatives is the discrete counterpart of the second difference being symmetric under left-right reflection — no first-derivative (advective) term can appear.

Step 3. Substitute into (1) and divide by $a$ to expose the density:

\frac{m}{a}\,\partial_t^2 y=T\,\partial_x^2 y+\frac{T a^2}{12}\,\partial_x^4 y+\cdots

Step 4. Hold $\mu=m/a$ and $T$ fixed and let $a\to0$ : the fourth-derivative correction dies as $a^2$ and

\mu\,\partial_t^2 y=T\,\partial_x^2 y,\qquad c=\sqrt{T/\mu}

Why: fixing $\mu$ rather than $m$ is forced by physics, not taste — a metre of real string has a definite mass; the bead bookkeeping underneath is invisible to any long-wavelength experiment. *Why: the discarded term $\tfrac{Ta^2}{12\mu}\,\partial_x^4 y$ reproduces, for a plane wave, the correction $\omega\approx ck\,[1-(ka)^2/24]$ obtained by expanding the exact dispersion (3) to order $(ka)^3$ — the two calculations agree term by term.*

The result is the one-dimensional wave equation, studied in its own right in the wave-equation article:

\mu\,\frac{\partial^2 y}{\partial t^2}=T\,\frac{\partial^2 y}{\partial x^2},\qquad c=\sqrt{\frac{T}{\mu}} \qquad (5)

where: $\mu$ — linear mass density (held fixed in the limit); $y(x,t)$ — continuous displacement field replacing the discrete $y_j(t)$ ; $c$ — propagation speed. Note what disappeared: the cutoff frequency, the dispersion, the Brillouin zone. The continuum equation treats all wavelengths identically because it has forgotten that a smallest length ever existed. The second route starts from energy bookkeeping. Summing kinetic energies of the beads and the stretching energies of the segments gives the discrete Lagrangian

L=\sum_{j}\left[\tfrac12 m\dot y_j^2-\tfrac12\frac{T}{a}(y_{j+1}-y_j)^2\right] \qquad (6)

where: $\tfrac12 m\dot y_j^2$ — kinetic energy of bead $j$ ; $\tfrac12(T/a)(y_{j+1}-y_j)^2$ — potential energy stored in the segment between $j$ and $j+1$ , which is the work $T\,\delta\ell$ done against tension when the segment lengthens by $\delta\ell\approx(y_{j+1}-y_j)^2/2a$ . The effective transverse spring constant $T/a$ comes from geometry, not from any material elasticity of the string.

Go deeper · Derivation C · 1 minContinuum limit of the Lagrangian and the field route to the wave equation

Derivation CContinuum limit of the Lagrangian and the field route to the wave equation

We convert the discrete Lagrangian into a spatial integral of a Lagrangian density and recover the wave equation from the field form of the Euler–Lagrange equation. The strategy is to factor one power of the cell width $a$ out of every term of the sum, so that what multiplies it is built purely from intensive quantities — then the sum over cells is a Riemann sum, and the limit is an integral. The point to watch: this requires the displacement profile to vary slowly on the scale $a$ , exactly the regime where the previous derivation was honest too.

Step 1. Factor out the spacing:

L=\sum_j a\left[\tfrac12\frac{m}{a}\dot y_j^2-\tfrac12 T\Big(\frac{y_{j+1}-y_j}{a}\Big)^2\right]

Why: each bracket now contains only $\mu=m/a$ , $T$ , and difference quotients — quantities with finite continuum meaning; the lone prefactor $a$ is the width of one cell of the emerging integral.

Step 2. Let $a\to0$ : the difference quotient tends to the slope $\partial_x y$ and the sum becomes an integral:

L\to\int \mathrm{d}x\,\left[\tfrac12\mu\,\dot y^2-\tfrac12 T\,y'^2\right]\equiv\int\mathcal{L}\,\mathrm{d}x

Why: this is the definition of the Riemann integral, applied to mechanics — the Lagrangian density $\mathcal{L}$ is simply the Lagrangian per unit length of string.

Step 3. Read off the derivatives of $\mathcal{L}(\dot y, y')$ :

\frac{\partial\mathcal{L}}{\partial\dot y}=\mu\dot y,\qquad \frac{\partial\mathcal{L}}{\partial y'}=-T y',\qquad \frac{\partial\mathcal{L}}{\partial y}=0

Step 4. Assemble the field Euler–Lagrange equation (the field generalisation, derived in full in the Euler–Lagrange-for-fields article, of the particle equation from W3):

\partial_t\!\left(\frac{\partial\mathcal{L}}{\partial \dot y}\right)+\partial_x\!\left(\frac{\partial\mathcal{L}}{\partial y'}\right)-\frac{\partial\mathcal{L}}{\partial y}=0\;\Rightarrow\;\mu\,\ddot y-T\,y''=0

Why: one term per independent variable that $y$ is differentiated with respect to — time contributes the momentum-density term, space the tension term. *Why: both the direct Newtonian limit and the Lagrangian route yield $\mu\,\partial_t^2 y=T\,\partial_x^2 y$ with $c=\sqrt{T/\mu}$ , confirming mutual consistency; the canonical momentum density $\pi=\partial\mathcal{L}/\partial\dot y=\mu\dot y$ also matches the continuum limit of the bead momentum $m\dot y_j/a$ .*

The object that survives the limit,

\mathcal{L}=\tfrac12\mu\left(\frac{\partial y}{\partial t}\right)^2-\tfrac12 T\left(\frac{\partial y}{\partial x}\right)^2 \qquad (7)

where: $\mathcal{L}$ — Lagrangian density (energy per unit length); $\partial y/\partial x$ — local slope, the one-dimensional ancestor of strain, is the true protagonist of this chapter. Its structure — kinetic term minus gradient term — recurs unchanged in the Lagrangian-density article, in linear elasticity, and (with $c$ reinterpreted as the speed of light) in relativistic field theory en route to the action-principle capstone.

Limiting cases

The exact discrete dispersion $\omega(k)=2\sqrt{T/(ma)}\,|\sin(ka/2)|$ contains every regime as a special case:

Case	Result	Meaning
Long wavelength $ka\ll1$	$\omega\approx \sqrt{\tfrac{T}{ma}}\,ka=\sqrt{\tfrac{T}{\mu}}\,k=ck$	Linear, non-dispersive: the continuum wave equation. All wavelengths travel at the same speed $c=\sqrt{T/\mu}$ .
Zone boundary $k=\pi/a$	$\omega=\omega_{\max}=2\sqrt{\tfrac{T}{ma}}$	Shortest wavelength $\lambda=2a$ ; adjacent beads move oppositely. Group velocity $v_g=0$ : this mode does not propagate energy.
Continuum limit $a\to0,\;m\to0,\;\mu=m/a$ fixed	$m\ddot y_j=\tfrac{T}{a}(y_{j+1}-2y_j+y_{j-1})\;\longrightarrow\;\mu\,\partial_t^2 y=T\,\partial_x^2 y$	The coupled ODE system becomes the wave equation; the cutoff $\omega_{\max}\to\infty$ disappears.
Two beads $N=2$ (fixed ends)	$\omega_1=\sqrt{\tfrac{T}{ma}},\;\omega_2=\sqrt{\tfrac{3T}{ma}}$	Equation (4) with $N=2$ reproduces the symmetric and antisymmetric modes of the two-mass chain from the coupled-oscillators article.
Single bead $N=1$ (fixed ends)	$\omega=\sqrt{\tfrac{2T}{ma}}$	A consistency check: the chain reduces to one mass between two string segments of combined transverse stiffness $2T/a$ .

Worked examples

Example 1 — three beads, by hand and against the continuum. A string of tension $T=40\ \mathrm{N}$ carries $N=3$ beads of mass $m=10\ \mathrm{g}$ spaced $a=10\ \mathrm{cm}$ apart, with fixed ends (total length $L=(N+1)a=0.4\ \mathrm{m}$ ). Find the three mode frequencies and compare with the continuum prediction for the same string.

Step 1. Compute the frequency scale. $\sqrt{T/(ma)}=\sqrt{40/(0.010\times0.10)}=\sqrt{40{,}000}=200\ \mathrm{s^{-1}}$ , so $\omega_{\max}$ -scale $2\sqrt{T/(ma)}=400\ \mathrm{s^{-1}}$ .

Step 2. Apply equation (4) with $N=3$ : $\omega_n=400\,\sin\big(\tfrac{n\pi}{8}\big)$ . The sines are $\sin22.5^\circ=0.3827$ , $\sin45^\circ=0.7071$ , $\sin67.5^\circ=0.9239$ , giving $\omega_1=153.1$ , $\omega_2=282.8$ , $\omega_3=369.6\ \mathrm{rad/s}$ — i.e. $f_1=24.4$ , $f_2=45.0$ , $f_3=58.8\ \mathrm{Hz}$ .

Step 3. Continuum comparison. The equivalent smooth string has $\mu=m/a=0.10\ \mathrm{kg/m}$ and $c=\sqrt{T/\mu}=\sqrt{400}=20\ \mathrm{m/s}$ , so $f_n=nc/2L=25n\ \mathrm{Hz}$ : $25$ , $50$ , $75\ \mathrm{Hz}$ .

Step 4. Compare. The continuum overestimates by $2.6\%$ , $11\%$ , and $27\%$ for $n=1,2,3$ : the coarser the mode (relative to the spacing), the worse the smooth description. The exact spectrum is $\{24.4,\ 45.0,\ 58.8\}\ \mathrm{Hz}$ against the continuum's $\{25,\ 50,\ 75\}\ \mathrm{Hz}$ .

Example 2 — a discretised piano string and the error budget of a simulation. A steel piano string ( $L=1.00\ \mathrm{m}$ , $\mu=6.5\ \mathrm{g/m}$ , $T=700\ \mathrm{N}$ ) is modelled numerically by $N=99$ beads with spacing $a=1\ \mathrm{cm}$ . How accurate are mode 10 and mode 50, and what is the highest frequency the model can represent at all?

Step 1. Continuum baseline. $c=\sqrt{T/\mu}=\sqrt{700/0.0065}=\sqrt{107{,}692}=328.2\ \mathrm{m/s}$ , so $f_1=c/2L=164.1\ \mathrm{Hz}$ (close to the E below middle C) and $f_n=164.1\,n\ \mathrm{Hz}$ .

Step 2. Discrete frequencies. With $m=\mu a=6.5\times10^{-5}\ \mathrm{kg}$ , $\sqrt{T/(ma)}=c/a=32{,}817\ \mathrm{s^{-1}}$ , and $k_na=n\pi/100$ , equation (4) gives $f_n^{\rm disc}=f_n\times\frac{\sin(n\pi/200)}{\,n\pi/200\,}$ .

Step 3. Mode 10: $x=10\pi/200=0.1571$ , $\sin x/x=0.99589$ , so $f_{10}^{\rm disc}=1641\times0.99589=1634\ \mathrm{Hz}$ — only $0.41\%$ low. Mode 50: $x=50\pi/200=0.7854$ , $\sin x/x=0.9003$ , so $f_{50}^{\rm disc}=8205\times0.9003=7387\ \mathrm{Hz}$ — a full $10\%$ below the true $8205\ \mathrm{Hz}$ .

Step 4. Absolute ceiling. $\omega_{\max}=2c/a=65{,}634\ \mathrm{rad/s}$ , i.e. $f_{\max}=10{,}446\ \mathrm{Hz}$ . The grid is excellent below $\sim2\ \mathrm{kHz}$ ( $<1\%$ error), degrades to $10\%$ by mode 50, and cannot represent any physics above $10.4\ \mathrm{kHz}$ — this artificial dispersion, not round-off, is the leading error of finite-difference wave codes.

The people behind the limit

Origins · 1 minJoseph-Louis Lagrange and how this was discovered

Joseph-Louis Lagrange

1736–1813 · Turin, Berlin and Paris

Lagrange solved the loaded string exactly at age 23 in his 1759 Turin memoir on the propagation of sound — all $N$ modes, for arbitrary $N$ — and used the solution to arbitrate the fierce d'Alembert–Euler–Bernoulli dispute over what counts as a legitimate solution of the wave equation. His finite chain made the question concrete: with $N$ beads there are exactly $N$ modes and no ambiguity; the mystery lay entirely in the limit. The same memoir contains, in embryo, the passage from a sum over particles to an integral over a density that his Mécanique analytique (1788) would turn into the systematic method this chapter now calls the Lagrangian route to field theory.

History

1727

— Johann Bernoulli studies the discrete massed string, finding the shape of its lowest mode

1746

— d'Alembert derives the continuum wave equation and its travelling-wave solution

1753

— Daniel Bernoulli asserts every string motion is a superposition of sine modes

1759

— Lagrange solves the N-bead chain completely and examines the passage to the continuum

1912

— Born and von Kármán quantise the same lattice dynamics, founding the theory of phonons

Going further

References. Morin, Introduction to Classical Mechanics, ch. on waves and coupled oscillators · Goldstein, Poole & Safko, Classical Mechanics 3rd ed., §13.1 (transition from discrete to continuous systems) · Landau & Lifshitz, Mechanics Vol.1, §23–24 (oscillations of systems with many degrees of freedom) · Taylor, Classical Mechanics, ch. 16 (continuum mechanics and waves) The key results can be checked by direct calculation: the bead equation $m\ddot y_j=(T/a)(y_{j+1}-2y_j+y_{j-1})$ follows from small-angle tension components; substituting a plane wave confirms $\omega=2\sqrt{T/(ma)}|\sin(ka/2)|$ and its $ka\ll1$ limit $\omega=ck$ with $c=\sqrt{T/\mu}$ ; the second-difference Taylor expansion gives $y_{j+1}-2y_j+y_{j-1}=a^2 y''+\tfrac{1}{12}a^4 y''''+\cdots$ ; and the Lagrangian density $\mathcal{L}=\tfrac12\mu\dot y^2-\tfrac12 T y'^2$ yields $\mu\ddot y=Ty''$ via the field Euler–Lagrange equation. No empirical constants are quoted.

Builds on: Normal Modes and Standing Waves on a String · The One-Dimensional Wave Equation · Lagrangian Density and Continuum Lagrangian Formalism Leads to: Euler–Lagrange Equations for Fields · Hamiltonian Density and Canonical Field Momentum · Elastic Waves in Solids · Linear Elasticity and Hooke's Law for Solids

The string segments are not springs — where does the potential energy formula for each segment actually come from?

From geometry, not elasticity. When neighbouring beads are displaced by $y_j$ and $y_{j+1}$ , the connecting segment lengthens from $a$ to $\sqrt{a^2+(y_{j+1}-y_j)^2}\approx a+(y_{j+1}-y_j)^2/2a$ . Pulling that extra length against the constant tension $T$ costs work $T\times(y_{j+1}-y_j)^2/2a$ — exactly the quadratic "spring" energy in the Lagrangian, with effective stiffness $T/a$ even though nothing material stretches appreciably. This is why a chain under higher tension is transversely stiffer, and why the transverse stiffness disappears entirely if the tension is released.

Why does the sum in the Lagrangian route converge to an integral — and when would it fail?

Factoring $L=\sum_j a[\ldots]$ makes the sum a Riemann sum: cell width $a$ times an intensive integrand sampled at $x_j=ja$ . Riemann sums converge to the integral when the integrand varies slowly across one cell, i.e. when the displacement profile has wavelength $\lambda\gg a$ . It fails precisely for zone-boundary-scale profiles ( $\lambda\sim2a$ ), where the difference quotient $(y_{j+1}-y_j)/a$ is nothing like a derivative of any smooth function. So the field description is not wrong but scoped: it is the effective theory of the chain's long-wavelength sector.

Does the lattice cutoff have anything to do with the ultraviolet divergences of field theory?

Conceptually, yes — and the loaded string is the standard cautionary tale. The continuum string has infinitely many modes with no upper frequency, so quantities summed over all modes (like thermal energy at high temperature, or zero-point energy after quantisation) can diverge. The lattice regulates this automatically: only $N$ modes exist, none above $\omega_{\max}$ . Condensed-matter systems really have such a cutoff (the atomic spacing); relativistic field theories may not, which is why they need renormalisation — a systematic way of parametrising ignorance of the short-distance "beads". The Debye model of solids is exactly this article's spectrum used as a regulator.

Is the wavenumber of a lattice wave even unique? Adding a large number to k seems to give the same bead motion.

Correct, and this is a real physical statement, not a technicality. Since only the values $y_j=y(ja)$ exist, two waves with wavenumbers differing by $2\pi/a$ assign identical displacements to every bead: $e^{i(k+2\pi/a)ja}=e^{ikja}$ . All distinct motions are labelled by $k$ in the first Brillouin zone $(-\pi/a,\pi/a]$ . This is the same aliasing that makes wagon wheels spin backwards on film and forces audio engineers to sample above the Nyquist rate — the lattice is a sampled string, at sampling interval $a$ .

Why does the group velocity vanish at the zone boundary while the phase velocity stays finite?

At $k=\pi/a$ the phase velocity is $\omega_{\max}/k=2c/\pi\approx0.64c$ — finite. But the mode there is neighbouring beads in perfect antiphase, and a pattern that merely flips sign bead-to-bead has nowhere to go: shifting it by one site reproduces itself with a sign, so no profile feature can translate. Formally, $\omega(k)$ is symmetric about $\pi/a$ (waves with $k$ slightly above and below the boundary are degenerate), and a maximum of $\omega$ forces $v_g=\mathrm{d}\omega/\mathrm{d}k=0$ . Energy in that mode sloshes locally; it does not travel.

What happens if the beads interact beyond nearest neighbours?

The structure survives; only the dispersion changes. With springs to second neighbours of stiffness ratio $\beta$ , the dispersion becomes $\omega^2=\frac{4T}{ma}[\sin^2(ka/2)+\beta\sin^2(ka)]$ — still periodic in $k$ with period $2\pi/a$ , still linear at small $k$ (with a renormalised speed), still $v_g=0$ at the zone boundary. The continuum limit again yields the wave equation, because any short-ranged coupling looks like a second derivative at long wavelength. This robustness — microscopic details washing out into a few macroscopic constants — is the first appearance of universality, the idea behind effective field theory.

In the numerical-analysis reading, is the lattice dispersion the same thing as the Courant instability?

No — they are the two distinct ways a finite-difference wave code fails. Lattice dispersion is a spatial discretisation error, present even with exact time integration: short waves lag, as computed in this article. The Courant condition ( $c\,\Delta t\le a$ for the standard explicit scheme) is a temporal stability constraint: violate it and zone-boundary modes grow exponentially rather than oscillate. A code can be stable yet dispersive (usual), or accurate in space yet unstable in time. The loaded string gives you the vocabulary for the first; von Neumann stability analysis of the time-stepped lattice gives the second.

Equation (4) has the beads' frequencies saturating, yet the continuum modes climb linearly forever. Which is right for a real string?

The lattice, always — but the crossover is far above anything audible. A real string is atomically discrete ( $a\sim10^{-10}\ \mathrm{m}$ ), so its zone-boundary frequency is in the terahertz range, ten orders above the fundamental. For the first thousands of modes the continuum formula $f_n=nc/2L$ is superb. The saturation matters in genuinely coarse lattices: bead-spring metamaterials, coupled-pendulum arrays, and numerical grids, where $a$ is macroscopic and the sine in equation (4) bends visibly within the measured spectrum — as in Example 1, where mode 3 of a 3-bead chain sits $27\%$ below the continuum guess.