Normal Modes and Standing Waves on a String

A taut string obeys the one-dimensional wave equation. By itself this equation admits travelling waves of any frequency, but a real string is finite and tied down at its ends. Imposing boundary conditions selects a privileged, discrete family of solutions: the normal modes. Each mode is a standing wave whose every point oscillates in phase with a common frequency, the spatial profile remaining fixed while only the overall amplitude breathes in time. Mathematically the problem is a Sturm–Liouville eigenvalue problem. Writing $y(x,t)=\phi(x)T(t)$ separates the wave equation into an ordinary differential equation for $\phi$ subject to the boundary conditions, plus a harmonic equation for $T$ . The boundary conditions admit nontrivial $\phi$ only for a discrete set of eigenvalues, each fixing an allowed frequency. The corresponding eigenfunctions are orthogonal under the natural inner product, which is precisely what makes a normal-mode expansion possible. Physically the normal modes are the continuum limit of the normal modes of a chain of beads connected by springs: as the number of masses goes to infinity the discrete dispersion relation collapses onto the linear $\omega = ck$ of the string, and the bead displacement patterns become smooth sine functions. The string therefore behaves like an infinite collection of independent simple harmonic oscillators, one per mode. Because the modes decouple, the general motion is a superposition with constant amplitudes and phases fixed by the initial displacement and velocity. The energy of the string is the sum of the energies stored in each mode, with no cross terms — a direct consequence of orthogonality. This decomposition is the prototype for vibrational analysis of solids, acoustic resonators, electromagnetic cavities, and the mode expansion used to quantize fields.

Key takeaways

A string fixed (or otherwise constrained) at its ends supports a discrete set of standing-wave solutions called normal modes, each oscillating at a single frequency $\omega_n$ with a fixed spatial shape $\phi_n(x)$ .
Separating variables in the wave equation $\partial_t^2 y = c^2\partial_x^2 y$ reduces it to a spatial eigenvalue problem whose allowed eigenvalues are quantized by the boundary conditions.
For a string of length $L$ fixed at both ends, $\phi_n(x)=\sin(n\pi x/L)$ with frequencies $\omega_n = n\pi c/L$ forming an exact harmonic series.
The eigenfunctions are mutually orthogonal, so any initial shape and velocity expand as a Fourier sum over modes — each mode then evolving independently like an uncoupled oscillator, with energy $E=\sum_n E_n$ and no cross terms.
Every standing wave is secretly two equal counter-propagating travelling waves; the nodes are where they perpetually cancel.
This is the continuum limit of the discrete loaded string, and the same eigenvalue/orthogonality machinery underlies quantum bound states, waveguides, and field quantization.
Key idea: Boundary conditions turn the wave equation into an eigenvalue problem: only a discrete spectrum of standing-wave shapes and frequencies survives, and these orthogonal eigenfunctions form a complete basis in which the general motion is a sum of independent oscillators.

Where this lives today

Normal-mode thinking is the daily language of precision measurement. The flagship 2026-era example is quantum optomechanics: silicon-nitride membranes and phononic-crystal beams — literally millimetre-scale drumheads and strings — are laser-cooled into their quantum ground state, with single normal modes engineered for quality factors above $10^9$ so that one chosen $\phi_n$ stays energetically isolated for seconds. The design tools are exactly this article's ideas: node placement (clamp where the mode does not move, and the support cannot steal energy), mode orthogonality (neighbouring modes must not hybridise), and the mode-as-oscillator reduction that lets a continuum device be treated as a single quantum harmonic oscillator. The same eigenvalue machinery, in three dimensions, is structural engineering's modal analysis — the computed mode shapes and frequencies of bridges, turbine blades, and launch vehicles, with sensors placed at antinodes and dampers tuned to the offending $\omega_n$ — and it is the classical scaffold on which field quantization stands: each mode's $\ddot q_n+\omega_n^2q_n=0$ becomes a quantum oscillator with levels $(n+\tfrac12)\hbar\omega_n$ , the construction carried out via the Hamiltonian-density article and completed in the action-principle capstone.

The physical picture

Think of the string as infinitely many beads on springs — the loaded string of the previous article, pushed to its $N\to\infty$ limit. A normal mode is a collective motion in which every bead reaches its extreme at the same instant and passes through zero at the same instant — the whole pattern pulses up and down without travelling. The nodes (points that never move) stay pinned in place, which is why a bowed string photographed under a strobe shows fixed motionless points between breathing loops.

Figure

The mode gallery — the first three normal modes of a fixed-fixed string; each shape breathes in place at its own frequency, with nodes that never move

The boundary conditions act as a quantization rule: the string must have zero displacement at each fixed end, so only wavelengths fitting a whole number of half-bumps between the walls survive. And every standing wave is secretly a superposition of two equal travelling waves moving in opposite directions, their interference building the stationary node–antinode pattern — the bridge between this article and the d'Alembert picture of the wave-equation article. Because different modes are orthogonal, exciting one does not feed energy into another — the string is a set of decoupled tuning forks. Plucking simply sets the initial amplitudes of all modes at once; thereafter each rings at its own pitch, and the timbre of the sound is the recipe of those amplitudes.

Figure

The payoff picture — a midpoint pluck decomposed: the triangular shape is a definite recipe of odd modes with amplitudes falling as one over n squared, and every even mode is absent because the pluck point is its node

The core equations

The starting point is the wave equation of the previous article, now confined to a finite domain:

\frac{\partial^2 y}{\partial t^2} = c^2\,\frac{\partial^2 y}{\partial x^2},\qquad c=\sqrt{\frac{T}{\mu}} \qquad (1)

where: $y(x,t)$ — transverse displacement; $c$ — wave speed; $T$ — tension; $\mu$ — linear density. Nothing about the equation itself is discrete; every quantization below enters through the walls. Seeking solutions whose shape does not change — only their overall scale — we substitute a product form. Dividing $\phi\,T''=c^2\phi''\,T$ by $c^2\phi T$ isolates a pure function of $t$ on one side and of $x$ on the other; two functions of different variables can be identically equal only if both equal one constant, written $-k^2$ so the solutions oscillate rather than grow:

y(x,t)=\phi(x)\,T(t)\;\Rightarrow\; \frac{T''}{c^2 T}=\frac{\phi''}{\phi}=-k^2 \qquad (2)

where: $\phi(x)$ — spatial mode shape (eigenfunction); $T(t)$ — temporal factor; $k$ — the separation constant, which will become the wavenumber. The spatial half of the problem, with fixed ends, is the eigenvalue problem

\phi''(x)+k^2\phi(x)=0,\qquad \phi(0)=\phi(L)=0 \qquad (3)

where: $k^2$ — eigenvalue; $L$ — string length. This is a Sturm–Liouville problem with Dirichlet conditions: the same mathematical species as the quantum particle in a box, and the source of every "discrete spectrum" in linear physics.

Derivation AQuantization of the spectrum from the boundary conditions

We solve the spatial eigenvalue problem and show that clamping both ends admits only a discrete ladder of wavenumbers and frequencies. The strategy is elementary: write the general solution of the harmonic ODE, then let each boundary condition kill part of the freedom — the first fixes a coefficient, the second can only be satisfied at special values of $k$ . Watch the logic of the last step: we refuse the trivial solution, and that refusal is where discreteness comes from.

Step 1. The general solution of $\phi''+k^2\phi=0$ for real $k$ is

\phi(x)=a\sin(kx)+b\cos(kx)

Step 2. The left boundary gives $\phi(0)=b=0$ , leaving $\phi=a\sin(kx)$ . Why: the cosine does not vanish at the origin, so a clamped left end simply forbids it.

Step 3. The right boundary demands $\phi(L)=a\sin(kL)=0$ . Taking $a=0$ would leave no string motion at all, so the sine itself must vanish:

kL=n\pi\quad\Rightarrow\quad k_n=\frac{n\pi}{L},\qquad n=1,2,3,\dots

Why: $n=0$ is the trivial flat string and negative $n$ merely flips the sign of the shape — the physically distinct solutions are the positive integers.

Step 4. Each allowed $k_n$ feeds the temporal equation $T''+c^2k_n^2\,T=0$ , an oscillator at

\omega_n=c\,k_n=\frac{n\pi c}{L}

Verified: the spectrum is the exact harmonic series, matching $f_1=c/2L$ of string instruments; the same two-boundary logic with $\phi'(L)=0$ instead yields the odd-harmonic spectrum of the limiting-cases table — Goldstein ch. 13, Morin waves chapter.

The eigenfunctions and their spectrum, assembled:

\phi_n(x)=\sin\!\left(\frac{n\pi x}{L}\right),\qquad \omega_n = \frac{n\pi c}{L},\qquad f_n=\frac{n c}{2L},\qquad \lambda_n=\frac{2L}{n} \qquad (4)

where: $n$ counts the half-wavelengths between the walls; mode $n$ has $n-1$ interior nodes. The frequencies are exactly integer multiples of the fundamental — a special gift of the uniform string, not a general law of vibrating systems. The modes are orthogonal under the natural inner product,

\int_0^L \phi_m(x)\,\phi_n(x)\,\mathrm{d}x = \frac{L}{2}\,\delta_{mn} \qquad (5)

where: $\delta_{mn}$ — Kronecker delta. The quick proof uses $\sin a\sin b=\tfrac12[\cos(a-b)-\cos(a+b)]$ : each cosine integrates to zero over $[0,L]$ unless $m=n$ , when the $\cos0$ term contributes $L/2$ . Orthogonality is what lets initial data be projected mode by mode, exactly like resolving a vector along perpendicular axes. With completeness (Fourier's theorem), the general clamped-string motion is

y(x,t)=\sum_{n=1}^{\infty}\sin\!\left(\frac{n\pi x}{L}\right)\big[A_n\cos\omega_n t + B_n\sin\omega_n t\big] \qquad (6)

where: the coefficients are read off the initial data by projection:

A_n=\frac{2}{L}\int_0^L y(x,0)\sin\!\frac{n\pi x}{L}\,\mathrm{d}x,\quad B_n=\frac{2}{L\omega_n}\int_0^L \dot y(x,0)\sin\!\frac{n\pi x}{L}\,\mathrm{d}x \qquad (7)

where: $A_n$ encodes the initial shape and $B_n$ the initial velocity; both are constants for all time. Everything a string will ever do is decided at $t=0$ , one mode at a time.

Go deeper · Derivation B · 1 minModes as independent oscillators by diagonalizing the Lagrangian

Derivation BModes as independent oscillators by diagonalizing the Lagrangian

We insert the mode expansion into the string Lagrangian and show the string is literally an infinite set of uncoupled harmonic oscillators — the statement that gives "normal mode" its meaning. The strategy is to treat the mode amplitudes $q_n(t)$ as generalized coordinates and let orthogonality kill every cross term in both the kinetic and potential energies. The step to watch is the potential term, where differentiation turns sines into cosines: a different orthogonal family, but orthogonal all the same.

Step 1. Expand the field in mode coordinates:

y(x,t)=\sum_n q_n(t)\,\sin\frac{n\pi x}{L}

Why: this is a change of coordinates in configuration space, from "displacement at each point" to "amplitude of each shape" — the continuum version of normal coordinates for coupled oscillators.

Step 2. Kinetic energy: squaring the sum and integrating, cross terms die by (5):

\int_0^L \tfrac12\mu\,\dot y^2\,\mathrm{d}x=\frac{\mu L}{4}\sum_n \dot q_n^2

Step 3. Potential energy: $y'=\sum_n q_n(n\pi/L)\cos(n\pi x/L)$ , and the cosines obey the same $L/2$ orthogonality:

\int_0^L \tfrac12 T\,y'^2\,\mathrm{d}x=\frac{T L}{4}\sum_n \Big(\frac{n\pi}{L}\Big)^2 q_n^2=\frac{\mu L}{4}\sum_n\omega_n^2q_n^2

Why: the last equality uses $c^2=T/\mu$ and $\omega_n=n\pi c/L$ — the potential stiffness of mode $n$ grows as $n^2$ , which is why high modes are expensive to excite.

Step 4. The Lagrangian is a sum of independent oscillator Lagrangians, so Euler–Lagrange gives, mode by mode,

\ddot q_n+\omega_n^2 q_n=0

Verified: energy $E=\sum_n\tfrac{\mu L}{4}(\dot q_n^2+\omega_n^2q_n^2)$ is constant term by term — modes never exchange energy in the linear string; matches Goldstein and Landau–Lifshitz normal-coordinate treatments.

The energy statement that falls out of the diagonalization is worth displaying on its own:

E=\sum_{n=1}^\infty E_n,\qquad E_n=\frac{\mu L}{4}\,\omega_n^2\big(A_n^2+B_n^2\big) \qquad (8)

where: $E_n$ — energy locked into mode $n$ forever (in the ideal string); there are no cross terms, thanks to orthogonality. A string is an energy ledger with one indestructible account per mode.

Go deeper · Derivation C · 1 minA standing wave is two counter-propagating travelling waves

Derivation CA standing wave is two counter-propagating travelling waves

We reconcile this article with d'Alembert's theorem, which insists every solution is a right-mover plus a left-mover — standing waves included. The strategy is one trigonometric identity read in reverse, and the payoff is conceptual: nodes are points of perpetual cancellation, and the "non-travelling" wave is two waves travelling oppositely in perfect balance.

Step 1. Take a single mode and split it with the product-to-sum identity:

\sin(k_nx)\cos(\omega_nt)=\tfrac12\sin\!\big(k_n(x-ct)\big)+\tfrac12\sin\!\big(k_n(x+ct)\big)

Why: with $\omega_n=ck_n$ the arguments combine exactly into the characteristic variables $x\mp ct$ — the mode is manifestly of d'Alembert form.

Step 2. Interpret: two sinusoidal wavetrains of equal amplitude run in opposite directions; where their phases always oppose, the sum vanishes for all time — the nodes; where they always agree, the motion doubles — the antinodes.

Step 3. Check the energy flux: for the standing mode $S=-T\,\partial_ty\,\partial_xy\propto\sin(2k_nx)\sin(2\omega_nt)$ , which vanishes at every node and averages to zero everywhere. Why: the two half-waves carry equal and opposite energy currents; energy sloshes locally between kinetic (antinodes) and potential (around nodes) twice per period but never crosses a node. *Verified: reflection at a clamped end converts a right-mover into the inverted left-mover with exactly the equal amplitude required — a standing wave is what a wave and its own reflection build; consistent with the method-of-images treatment in the wave-equation article.*

Limiting cases

How the spectrum depends on the boundary conditions and parameters:

Case	Result	Meaning
Both ends fixed	$\omega_n = \dfrac{n\pi c}{L},\quad n=1,2,3,\dots$	$\phi_n=\sin(n\pi x/L)$ , nodes at both ends; a perfect harmonic series — the fundamental plus integer overtones.
One end fixed, one free	$\omega_n = \dfrac{(2n-1)\pi c}{2L},\quad n=1,2,\dots$	$\phi_n=\sin\!\big((2n-1)\pi x/2L\big)$ ; antinode at the free end. Only odd harmonics — like a closed organ pipe.
Both ends free	$\omega_n = \dfrac{n\pi c}{L},\quad n=0,1,2,\dots$	$\phi_n=\cos(n\pi x/L)$ ; the $n=0$ zero-frequency mode is uniform translation of the whole string.
Finite bead chain, low modes $n\ll N$	$\omega_n = 2\sqrt{\tfrac{k}{m}}\,\sin\!\Big(\tfrac{n\pi}{2(N+1)}\Big)\;\approx\;\tfrac{n\pi c}{L}$	The loaded-string spectrum linearizes to the harmonic series when the mode wavelength spans many beads — the continuum-limit check done quantitatively in the loaded-string article.

Worked examples

Example 1 — reading a guitar string's spectrum. The A-string of a guitar has $L=0.648\ \mathrm{m}$ , sounds $f_1=110.0\ \mathrm{Hz}$ , and (being wound) has $\mu=5.2\ \mathrm{g/m}$ . Find the wave speed, the tension, and the frequency and node positions of the third mode.

Step 1. Wave speed from the fundamental: $c=2Lf_1=2\times0.648\times110.0=142.6\ \mathrm{m/s}$ .

Step 2. Tension: $T=\mu c^2=5.2\times10^{-3}\times(142.6)^2=5.2\times10^{-3}\times2.03\times10^4$ , so $T\approx106\ \mathrm{N}$ — about $10.8\ \mathrm{kg}$ -force.

Step 3. Third mode: $f_3=3f_1=330.0\ \mathrm{Hz}$ (the E above middle C), wavelength $\lambda_3=2L/3=0.432\ \mathrm{m}$ .

Step 4. Nodes of $\sin(3\pi x/L)$ sit where the argument is a multiple of $\pi$ : $x=L/3=0.216\ \mathrm{m}$ and $x=2L/3=0.432\ \mathrm{m}$ , plus the two clamped ends. Touching the string lightly at $0.216\ \mathrm{m}$ while plucking isolates exactly this $330\ \mathrm{Hz}$ harmonic — the guitarist's "natural harmonic" is applied Sturm–Liouville theory.

Example 2 — the mode recipe of a midpoint pluck. A string of length $L=0.65\ \mathrm{m}$ is pulled up by $h=4.0\ \mathrm{mm}$ at its midpoint into a triangle and released from rest. Find the mode amplitudes, evaluate the first three nonzero ones, and compute the fraction of the total energy carried by the fundamental.

Step 1. Release from rest means every $B_n=0$ ; only the shape integrals (7) survive. The triangular shape is $y(x,0)=2hx/L$ on $[0,L/2]$ and $2h(L-x)/L$ on $[L/2,L]$ .

Step 2. The projection integral (done by parts, or looked up as the Fourier sine series of a triangle) gives

A_n=\frac{8h}{\pi^2n^2}\sin\!\frac{n\pi}{2}

so $A_n=0$ for every even $n$ — those modes have a node exactly where the string was displaced — and $|A_n|=8h/\pi^2n^2$ for odd $n$ , alternating in sign.

Step 3. Numbers with $h=4.0\ \mathrm{mm}$ : $A_1=8h/\pi^2=3.24\ \mathrm{mm}$ , $A_3=-A_1/9=-0.36\ \mathrm{mm}$ , $A_5=A_1/25=0.13\ \mathrm{mm}$ . The recipe collapses fast: the fundamental dominates by an order of magnitude by mode 3.

Step 4. Energy fractions from (8): $E_n\propto\omega_n^2A_n^2\propto n^2\times n^{-4}=n^{-2}$ over odd $n$ . The sum $\sum_{\rm odd}n^{-2}=\pi^2/8$ , so the fundamental's share is $1/(\pi^2/8)=8/\pi^2$ . The fundamental carries $8/\pi^2\approx81\%$ of the pluck's energy, mode 3 carries $9.0\%$ , mode 5 carries $3.2\%$ — and all even modes carry exactly zero, which is why midpoint-plucked strings sound hollow and flute-like.

The people behind the modes

Origins · 1 minDaniel Bernoulli and how this was discovered

Daniel Bernoulli

1700–1782 · Basel; St Petersburg Academy

Daniel Bernoulli was the first to insist — in 1753, against both d'Alembert and Euler — that the general motion of a vibrating string is a superposition of its sinusoidal modes, "an infinity of simple vibrations coexisting." He had physics on his side (he could hear overtones ringing above a fundamental) but not yet the mathematics: he could not prove an arbitrary initial shape expandable in sines, and Euler's kinked pluck shapes seemed to defeat the claim. The dispute simmered for seventy years until Fourier's heat-equation work supplied the missing expansion theorem, vindicating Bernoulli completely. His superposition principle is now the default grammar of linear physics, from concert halls to quantum field theory.

History

1715

— Brook Taylor derives the fundamental frequency of a vibrating string from first principles

1746

— d'Alembert publishes the wave equation and its travelling-wave general solution

1753

— Daniel Bernoulli claims every string motion is a superposition of sinusoidal modes

1787

— Chladni's sand figures make two-dimensional nodal lines directly visible

1822

— Fourier's expansion theorem settles the superposition controversy in Bernoulli's favour

Going further

References. Morin, Introduction to Classical Mechanics, ch. on waves and normal modes · Goldstein, Classical Mechanics, 3rd ed., ch. 13 (Continuous Systems and Fields) · Landau & Lifshitz, Mechanics / Theory of Elasticity (oscillations of continua) · Taylor, Classical Mechanics, ch. 16 (Continuum Mechanics)

Provenance. Equations are standard Sturm–Liouville/Fourier results. A verifier should recompute: separation of variables giving $\phi''+k^2\phi=0$ ; quantization $k_n=n\pi/L$ from Dirichlet conditions; orthogonality integral $\int_0^L\sin(m\pi x/L)\sin(n\pi x/L)dx=(L/2)\delta_{mn}$ ; Lagrangian diagonalization to $\ddot q_n+\omega_n^2 q_n=0$ ; and the bead-chain limit $\omega_n=2\sqrt{k/m}\sin(n\pi/2(N+1))\to n\pi c/L$ . Cross-check against Morin (waves chapter), Goldstein 3rd ed. ch.13, Landau–Lifshitz Mechanics, and Taylor ch.16. The 220→660 Hz numeric example uses $f_n=nf_1$ ; the $1/\sqrt2$ tension scaling uses $c=\sqrt{T/\mu}$ . No nonstandard numbers introduced.

Builds on: The One-Dimensional Wave Equation · Discrete Loaded String and the Continuum Limit · Lagrangian Density and Continuum Lagrangian Formalism Leads to: Euler–Lagrange Equations for Fields · Hamiltonian Density and Canonical Field Momentum · Elastic Waves in Solids

Are the nodes really perfectly motionless? How does energy get from one side of a node to the other?

In a single ideal mode, yes — and energy does not cross a node at all. The flux $S=-T\,\partial_ty\,\partial_xy$ vanishes identically at nodes (the velocity is zero there for all time), so each half-wavelength cell is energetically sealed: its energy sloshes between kinetic form at the antinode and potential form near the nodes, twice per period. Real strings blur this in two ways: any superposition of several modes has no strictly motionless points (the "nodes" of the sum wander), and damping or stiffness couples cells weakly. But for one mode, a node is a wall.

In the separation of variables, why must the constant be negative? What goes wrong with a positive constant?

Try it: $\phi''=+\kappa^2\phi$ gives exponentials $e^{\pm\kappa x}$ , and no combination of growing and decaying exponentials can vanish at two points without vanishing everywhere — check: $a e^{\kappa x}+be^{-\kappa x}=0$ at $x=0$ and $x=L$ forces $a=b=0$ . A zero constant gives a straight line, killed the same way. Only the negative constant produces oscillatory solutions that can thread two zeros. This is the general Sturm–Liouville fact that Dirichlet eigenvalues of $-\mathrm{d}^2/\mathrm{d}x^2$ are strictly positive — the same reason quantum bound-state energies in a box are positive.

Is the motion of a plucked string periodic in time? A general superposition of oscillators usually is not.

For the ideal fixed-fixed string, yes — a hidden miracle of the harmonic spectrum. All frequencies are integer multiples of $\omega_1$ , so after one fundamental period $2L/c$ every mode has completed a whole number of cycles and the exact initial shape recurs. This is why a plucked note has a stable pitch. A general system with incommensurate mode frequencies (a drumhead, a stiff piano string) never exactly repeats: its motion is quasi-periodic. The audible difference is real — drums have no definite pitch precisely because their spectrum is not harmonic.

A plucked string has a kink, but every mode shape is a smooth sine. How can smooth functions add up to a corner?

Pointwise, they genuinely do: the Fourier sine series of the triangle converges to the triangle at every point, kink included — but non-uniformly in its derivatives. The slope series converges to the true slope on either side of the kink and to the average at it, and the convergence is slow ( $A_n\sim n^{-2}$ means slope coefficients $\sim n^{-1}$ ). Truncate the series and the corner is rounded over a width $\sim L/n_{\max}$ . Nothing pathological happens for the string because the energy (8) is finite; a genuine discontinuity in displacement, by contrast, would carry infinite potential energy and is forbidden.

Do the modes really never exchange energy, even after minutes of ringing?

In the linear, undamped, uniform string: never — the diagonalized Lagrangian has strictly no coupling terms. Real strings violate the assumptions slightly, and each violation opens a specific channel: air damping and bridge losses drain every mode (fast for high $n$ ), bending stiffness shifts frequencies without coupling modes, but nonlinearity at finite amplitude genuinely couples them, feeding energy between harmonics (audible as the evolving timbre of a hard-struck piano note in its first second). The clean statement is: mode independence is exact at linear order and fails at second order in the slope, the same order at which the wave equation itself was truncated.

What exactly distinguishes a normal mode from a standing wave — the article seems to use both words?

Every normal mode of the clamped string is a standing wave, but not every standing wave is a normal mode. "Standing wave" is the local phenomenon: two equal counter-propagating waves producing a stationary node-antinode pattern, which can happen on any stretch of any medium (microwave ovens, laser cavities, a rope shaken at one end). "Normal mode" adds the global requirement: the pattern also satisfies the system's boundary conditions at an eigenfrequency, so it can persist alone forever. On the clamped string the two notions coincide because reflection at the walls supplies the counter-propagating partner automatically.

The string's frequencies are quantized. Is this quantum mechanics sneaking into classical physics?

No — and the distinction is worth stating sharply. What is discrete here is the set of allowed frequencies (eigenvalues of a boundary-value problem), a purely classical fact known since the 18th century; the amplitude, hence the energy, of each mode varies continuously. Quantum mechanics does something additional: it quantizes the energy of each mode in units of $\hbar\omega_n$ . The two discretenesses then compound — a cavity field has classically discrete mode frequencies and quantum-discrete photon numbers per mode. The string is the reason the Schrödinger box problem feels familiar: same eigenvalue mathematics, different physical object being quantized.

Where should I pluck to make a chosen harmonic as loud as possible, or to remove it?

The projection integral answers both. Plucking at position $x_0$ weights mode $n$ by $\sin(n\pi x_0/L)$ : to kill harmonic $n$ , pluck at one of its nodes ( $x_0=mL/n$ ); to maximise its share, pluck at one of its antinodes. Guitarists use both directions of the rule — plucking near the bridge ( $x_0\to L$ ) keeps many high modes appreciable, giving a bright twang, while the light-touch "harmonic" technique is the complementary trick: touching (not pressing) at $L/2$ or $L/3$ damps every mode without a node there, isolating the chosen harmonic.