Wave equation ($u_{tt}=c^2\Delta u$)

The wave equation is a second-order linear partial differential equation describing waves with finite propagation speed. In one spatial dimension its homogeneous, constant-coefficient form isHere $x$ is position, $t$ is time, $u(x,t)$ is the disturbance, and subscripts denote partial derivatives. For a wave equation on a string, $u$ is transverse displacement; in linear acoustics, it can be a pressure perturbation. In several spatial dimensions, $u_{xx}$ becomes the Laplacian $\Delta u$. The same mathematical structure connects vibrating strings and acoustic waves, although their physical assumptions and wave speeds differ. ^[ref]

For the whole-line Cauchy problem, prescribe both displacement $u(x,0)=f(x)$ and velocity $u_t(x,0)=g(x)$. A classical solution has continuous partial derivatives through order two; $f\in C^2(\mathbb{R})$ and $g\in C^1(\mathbb{R})$ suffice for D'Alembert's formula to produce one. A finite string additionally needs boundary conditions. These statements concern the unforced model with constant $c$; damping, spatially varying coefficients and nonlinear terms change the solution formula.

A travelling wave $F(x-ct)$ carries its entire profile to the right at wave speed $c$; $G(x+ct)$ travels left. Their sum is again a solution by the superposition principle. A standing wave instead has a fixed mode shape whose size and sign oscillate: in $\sin (kx)\cos (ckt)$, each node of a standing wave stays stationary. The distinction is especially clear in the pulse and string diagrams below. For this ideal model, the dispersion relation is ${\omega }^2=c^2{k}^2$: each propagation branch is nondispersive, so a localized profile translates without spreading. This makes the equation a useful reference against which to understand wave dispersion and damping.

The appealing combination is local interaction with global predictability: neighbouring displacements determine acceleration through $u_{xx}$, yet D'Alembert's formula explicitly identifies exactly which initial locations can influence a later observation. Finite propagation speed turns that observation into a domain of dependence, rather than requiring knowledge of the entire initial string.

For $E^{\prime \prime }+k^2(x)E=0$ with real $k^2$, differentiation gives ${\mathcal{J}}^{\prime }=\mathrm{Im}(E^{\ast }{E}^{\prime \prime })=0$. This conserved Wronskian measures directed wave flux. In a one-dimensional ordinary plasma wave, the mean Poynting flux is $c^2\mathcal{J}/(8\pi \omega )$ in Gaussian units. It proves $R+T=1$ for a lossless two-channel scattering problem.

A string element has extra length $[\sqrt{1+y_x^2}-1]dx\simeq y_x^{2}dx/2$. At fixed tension T, work done supplies elastic energy density $T{y}_x^2/2$. Transverse velocity supplies kinetic-energy density $\rho y_t^2/2$. For a single travelling profile $y=f(x\pm ct)$ with $c^2=T/\rho$, the two densities agree pointwise. Their sum obeys a local energy balance with flux $-T{y}_x{y}_t$. Standing waves generally exchange kinetic and elastic energy instead of maintaining equality at every instant.

A physical sheet under uniform isotropic tension per unit length T and with mass per unit area $\rho$ has small transverse displacement satisfying ${\partial }_t^2\psi =c^2{\nabla }^2\psi$, where $c^2=T/\rho$. Fixed edges impose zero displacement. This mechanical membrane model neglects bending stiffness, large slopes and dissipation.

For a rectangle of sides a and b with fixed edges, separation of variables gives $\psi =A\sin (p\pi x/a)\sin (q\pi y/b)\cos (\omega t+\varphi )$ with positive integers p and q. Its angular frequency is $\omega =c\pi \sqrt{p^2/a^2+q^2/b^2}$. Zero integers produce the trivial displacement, not a lower mode. The p−1 and q−1 interior nodal lines divide alternating-sign lobes.

For a rectangular membrane mode with peak antinode displacement A, the integrals of each squared sine factor are half the corresponding side length. Integrating local kinetic energy density $\rho ({\partial }_t\psi {)}^2/2$ gives $K=\rho ab{\omega }^2{A}^2{\sin }^2(\omega t+\varphi )/8$. The total energy is $\rho ab{\omega }^2{A}^2/8$, and potential energy has the complementary cosine-squared factor. Its time-averaged kinetic energy is $\rho ab{\omega }^2{A}^2/16$. At equal peak displacement, frequency and density, such mean energies are proportional to area.

A string of length $L$ with fixed endpoints, initially displaced into a triangular tent of height $k$ at its midpoint and released from rest, has a Fourier sine series with the coefficients in the title. Integrating the two linear halves gives $a_n$, and symmetry cancels the even modes. Each coefficient evolves as $a_n\cos ({\omega }_{n}t)$ under the wave equation. The displacement vanishes everywhere at $t=L/(2c)$, reverses sign at $t=L/c$, and returns after $2L/c$. A straight configuration at an intermediate instant does not mean the string is at rest: its kinetic energy is then nonzero. The initial slope corner gives a finite-energy weak solution whose corner lines propagate and reflect at the ends.

A fixed circular elastic membrane of radius $a$ has radial normal modes satisfying the displayed Sturm-Liouville problem, with weight $r$, $R(a)=0$ and regularity at the origin. The regular solutions are Bessel functions of the first kind $J_0(\sqrt{\lambda }r)$. Integration by parts gives the positive Rayleigh quotient $\lambda ={\int }_0^{a}r|R^{\prime }{|}^{2}dr/{\int }_0^{a}r|R{|}^{2}dr$. Distinct modes are orthogonal in the weighted inner product ${\int }_0^{a}r\overline{R}Sdr$.

For a fixed circular elastic membrane, the Rayleigh-Ritz variational principle applied to $f(r)=1-(r/a{)}^p$, $p>0$, gives ${\lambda }_1\le (p+3+2/p)/a^2$. The numerator of its Rayleigh quotient is $p/2$, and the denominator is $a^2{p}^2/[2(p+1)(p+2)]$. Minimization gives $p=\sqrt{2}$, ${\lambda }_1\le (3+2\sqrt{2})/a^2$, and the displayed upper bound for the ordinary frequency ${\nu }_1=c\sqrt{{\lambda }_1}/(2\pi )$. Every $p>0$ has finite radial Dirichlet energy, which is the required variational admissibility condition.

A wave equation with a linear damping term and positive wave speed $c$ separates into spatial normal modes and damped harmonic oscillators. For a spatial eigenvalue $k^2$, the time factor obeys $T^{\prime \prime }+\alpha T^{\prime }+c^2{k}^{2}T=0$. Its characteristic roots are $-\alpha /2\pm \sqrt{{\alpha }^2/4-c^2{k}^2}$. Thus underdamping requires $|\alpha |<2c|k|$, critical damping gives a repeated root and the second solution $t{e}^{-\alpha t/2}$, and overdamping gives two real exponential rates. Negative $\alpha$ amplifies rather than damps the motion.

For operator $c^{-2}{\partial }_t^2+2\alpha {\partial }_t-{\partial }_x^2$, define $\gamma =\alpha c^2$ and $d_k^2={\gamma }^2-c^2{k}^2$. Retarded support, continuity at zero time and the derivative jump $G_k^{\prime }(0^{+})=c^2$ fix the displayed normalization. Imaginary $d_k$ produces a damped sine; zero $d_k$ gives $c^2\tau e^{-\gamma \tau }$. The spatially uniform mode tends to $1/(2\alpha )$ rather than decaying to zero.

At small wavenumber compared with $\alpha c$, the slow decay rate of the damped wave operator is $k^2/(2\alpha )$ and the other root decays on time $1/(2\alpha c^2)$. Long-time, long-wavelength motion therefore approaches heat diffusion with the displayed diffusivity. The original hyperbolic equation nevertheless has finite signal speed $c$, unlike the parabolic diffusion approximation.

For real constants $\gamma ,c$, the substitution $v=e^{\gamma t}u$ turns the displayed equation into the wave equation $v_{tt}=c^2{v}_{xx}$. If $u(x,0)=f(x)$ and $u_t(x,0)=g(x)$, then $v(x,0)=f(x)$ and $v_t(x,0)=g(x)+\gamma f(x)$. For $v_0=|c|>0$, D'Alembert's formula consequently givesThe extra $\gamma f$ term is necessary even when the original initial velocity is zero. For $c=0$, take the limit to obtain $u=e^{-\gamma t}[f+t(g+\gamma f)]$. The formula gives classical solutions for suitably smooth data and weak solutions for less regular data. Positive $\gamma$ damps the transformed wave, whereas negative $\gamma$ amplifies it.

For $y_{tt}-y_{xx}=F(t)\delta (x-x_0)$ with zero endpoint displacement, a Fourier sine series gives $q_n^{\prime \prime }+k_n^2{q}_n=2F(t)\sin (k_n{x}_0)/L$. With the $e^{-i\omega t}$ Fourier transform convention, the causal inverse of this oscillator uses the displayed denominator. Its inverse transform is $H(t)\sin (k_{n}t)/k_n$, giving $q_n(t)=2\sin (k_n{x}_0)/(L{k}_n){\int }_{-\infty }^{t}F(s)\sin [k_n(t-s)]ds$ when the past integral exists. The plain denominator $k_n^2-{\omega }^2$ requires a pole prescription: its Cauchy principal value gives the time-symmetric kernel $\sin (k_n|t|)/(2k_n)$ instead. Homogeneous sine and cosine modes are additional freedom unless temporal data or causality are specified.

A strip of membrane with fixed edges $y=0,b$ supports transverse normal modes proportional to $\sin (n\pi y/b)$, $n\ge 1$. Separation of the two-dimensional wave equation gives the displayed dispersion relation. Confinement produces longitudinal wave dispersion even when the membrane's underlying wave speed $v$ is independent of frequency.

A fixed-edge membrane strip has dispersion ${\omega }_n^2=v^2[k_x^2+(n\pi /b{)}^2]$, n≥1. Differentiating gives $v_{g,x}=v^2{k}_x/{\omega }_n$, whereas $v_{p,x}={\omega }_n/k_x$. Their product is $v^2$. In this lossless scalar mode the integrated energy flux divided by energy density agrees with group speed. At cutoff the longitudinal group speed vanishes while formal phase speed diverges; an infinite phase speed therefore indicates no superfast energy transport.

For a plate under uniform tension per length $T$, with surface mass density $\sigma$ and flexural rigidity $D_b>0$, the small-displacement equation is $\sigma z_{tt}=T{\nabla }^{2}z-D_b{\nabla }^{4}z$. If the edge conditions admit a transverse sine mode of wavenumber $q_n$, its longitudinal dispersion relation isBending raises mode cutoff frequencies of a membrane waveguide and changes large-$k$ behaviour from $\omega \propto |k|$ to $\omega \propto k^2$. A genuinely clamped plate also fixes the edge slope, so its transverse mode shapes need not remain simple sine functions. The qualitative increase in restoring force survives that change of boundary conditions.

The cutoff frequency of a membrane waveguide is the lowest frequency at which a specified transverse normal mode can propagate longitudinally. For fixed edges separated by $b$, ${\omega }_{c,n}=n\pi v/b$ and $f_{c,n}={\omega }_{c,n}/(2\pi )$. Below cutoff the longitudinal wavenumber is imaginary and a bounded excitation is an evanescent wave. Above it, $v_g=v\sqrt{1-({\omega }_{c,n}/\omega {)}^2}$ and $v_p{v}_g=v^2$.

For a membrane waveguide branch ${\omega }^2=v^2[k_x^2+(n\pi /b{)}^2]$, the exact cutoff has zero longitudinal wavevector and zero group velocity. It is not a traveling channel carrying finite longitudinal energy flux. Modes strictly below cutoff are evanescent; amplitudes of the permitted above-cutoff modes depend on the transverse excitation profile.

For the wave equation ${\psi }_{tt}=c^2{\psi }_{xx}$ on $0<x<L$, impose homogeneous Neumann boundary conditions ${\psi }_x(0,t)={\psi }_x(L,t)=0$, where $c,L>0$. Separation of variables gives $X^{\prime \prime }+\lambda X=0$ and $T^{\prime \prime }+c^2\lambda T=0$. Integration by parts gives ${\int }_0^L(X^{\prime }{)}^{2}dx=\lambda {\int }_0^L{X}^{2}dx$, so every nontrivial spatial eigenfunction has $\lambda \ge 0$.

For $\lambda =0$, $X$ is constant and $T=A_0+B_{0}t$. For $\lambda >0$, the boundary conditions select ${\lambda }_n=(n\pi /L{)}^2$ and $X_n=\cos (n\pi x/L)$. The superposition principle gives the displayed expansion when the initial data have sufficient regularity. The constant zero mode must be retained: homogeneous Neumann boundary conditions do not force the spatial average of the displacement to vanish.

With zero initial velocity, all $B_n$ vanish, including $B_0$. Initial displacement $a+b\cos (n\pi x/L)$ then evolves as $a+b\cos (cn\pi t/L)\cos (n\pi x/L)$, a standing wave above a constant offset.

For a spherically symmetric function, $v=ru$ reduces the radial wave equation to $v_{tt}=c^2{v}_{rr}$. A time-harmonic separated solution has radial numerator $C\sin (kr)+D\cos (kr)$, with $k=\omega /c$. Finiteness at the origin excludes $D$, and the prescribed finite origin value fixes $C$. This is a standing combination of incoming and outgoing waves. A purely outgoing point-source wave instead has a $1/r$ singularity and must be specified by its source strength, not by a finite value at the origin.

This is the causal three-dimensional fundamental solution of ${\partial }_t^2-c_0^2{\nabla }^2$. It concentrates on the outgoing sound cone and gives the retarded convolution of acoustic sources.

For $u_{tt}=\Delta u+\alpha u$ with a homogeneous Dirichlet boundary condition, let ${\lambda }_1>0$ be the first Dirichlet Laplacian eigenvalue. On a bounded one-dimensional interval, finite-energy displacement remains bounded for all time and all initial data exactly when $\alpha <{\lambda }_1$. The conserved energy $\frac{1}{2}(\Vert u_t{\Vert }_2^2+\Vert \nabla u{\Vert }_2^2-\alpha \Vert u{\Vert }_2^2)$ controls the gradient and, by the Poincaré inequality and one-dimensional integration, the maximum norm. At equality a mode can grow linearly; above it a mode can grow exponentially. Bounded oscillations need not have a long-time limit.

The vector field method commutes a wave equation with spacetime symmetry generators, estimates the resulting commuted wave energies, and converts weighted L2 norms to pointwise decay using a Klainerman-Sobolev inequality. The commutation vector fields for the wave equation include translations, rotations, Lorentz boost vector fields and the scaling vector field. It is useful for almost global existence for wave equations and for small-data global existence under the classical null condition for wave equations.

A commuted wave energy controls derivatives of a solution after applying the commutation vector fields for the wave equation. A convenient equivalent norm is $A_m(t)={\sum }_{|I|\le m}\Vert \partial Z^{I}u(t){\Vert }_2$. The wave energy estimate controls its growth through commuted sources, while the Klainerman-Sobolev inequality gives pointwise decay for low-order derivatives.

In three spatial dimensions, with $Z$ the eleven translations, spatial rotations, boosts and scaling commutation vector fields for the wave equation, a sufficiently decaying smooth function satisfies $|u(t,x)|\le C(1+t+|x|{)}^{-1}(1+|t-|x||{)}^{-1/2}{\sum }_{|I|\le 2}\Vert Z^{I}u(t){\Vert }_2$. This weighted Sobolev inequality converts control of commuted wave energy into decay. It holds independently of any wave equation satisfied by $u$.

A first-order vector field $Z$ is a commutation field for the flat wave equation when $[\square ,Z]$ is zero or a controlled multiple of $\square$. Translations, spatial rotations and Lorentz boost vector fields commute with $\square$; the scaling vector field satisfies $[\square ,S]=2\square$. Repeated commutation preserves the differential order of a semilinear wave equation.

The vector field $S=t{\partial }_t+x\cdot \nabla$ generates simultaneous spacetime dilation. For the flat d'Alembert operator, $[\square ,S]=2\square$. Consequently it commutes with the homogeneous wave equation at the level of its solution set, although its operator commutator is nonzero.

For Minkowski spacetime with unit light speed, $L_i=t{\partial }_i+x_i{\partial }_t$ generates a Lorentz boost. It commutes with the d'Alembert operator and controls derivatives transverse to time slices in the vector field method for wave equations.

The vector field ${\Omega }_{ij}=x_i{\partial }_j-x_j{\partial }_i$ is an infinitesimal spatial rotation. It commutes with the flat d'Alembert operator and supplies angular derivatives in the Klainerman-Sobolev inequality.

The vector fields ${\partial }_t$ and ${\partial }_i$ generate time and space translations and commute with the flat d'Alembert operator. They form the unweighted part of the commutation vector fields for the wave equation.

For a real scalar wave equation, the kinetic and gradient energy is $E_{\mathrm{lin}}(t)=\frac{1}{2}\int (|u_t{|}^2+|\nabla u{|}^2)dx$. A potential term is added for a semilinear wave equation $u_{tt}-\Delta u+V^{\prime }(u)=0$, giving the conserved wave energy $E=E_{\mathrm{lin}}+\int V(u)dx$.

For $y_{tt}=c^2{y}_{xx}$ on $0<x<L$, initially at rest with zero displacement, apply $y(0,t)=a$ for $t>0$ and keep $y(L,t)=0$. A geometric series in the Laplace transform givesHere $H$ is the Heaviside step function, $L,c>0$, and only finitely many terms contribute at any finite time. The second step in each pair is a sign-reversed reflection from the fixed endpoint. The formula solves the wave equation classically away from the fronts and distributionally across them; the midpoint is a square wave of period $2L/c$.

For the homogeneous one-dimensional wave equation $u_{tt}=c^2{u}_{xx}$ with constant $c>0$, wave speed means the speed at which a travelling wave profile translates. In $F(x-ct)$, a feature initially at $x=s$ follows $x=s+ct$, giving signed velocity $+c$; in $G(x+ct)$ it follows $x=s-ct$, giving signed velocity $-c$. The characteristic lines for the one-dimensional wave equation therefore have slopes $dx/dt=\pm c$. The speed of either profile is $c$, even though the displacement at a fixed point can change sign or pass through zero.^[ref]

The same $c$ bounds the causal cone for the constant-speed wave equation. For a wave equation on a string, $c=\sqrt{T/\mu }$ depends on tension and linear mass density; for small acoustic waves in stationary uniform air it is the speed of sound. This interpretation assumes the stated ideal wave equation. In media with wave dispersion, phase velocity, group velocity and the earliest signal arrival have different meanings: a measured wave crest speed need not be the speed of a wave envelope, and neither alone establishes a causal cone for the constant-speed wave equation.

The experiment in A time of flight method to measure the speed of sound using a stereo sound card by Carlos C. Carvalho and others makes wave speed visible as a delay. Two powered microphones feed separate channels of one stereo line input. A clap beyond one end of their common axis reaches the nearer microphone first. The displayed signals show two separated onsets when enlarged; count their sample separation $N$ to obtain $\Delta t=N/f_s$ and infer speed of sound $c=d/\Delta t$ from the measured microphone spacing $d$. The reported sampling rate is $44100{\mathrm{s}}^{-1}$. Repeating at different spacings gives a distance-versus-delay plot whose slope estimates speed of sound; dependence on distance helps identify acoustic travel rather than a fixed electronic delay. These are the paper's apparatus and observed readout, not measurements made for this article.^[ref]

A controlled implementation also needs timing calibration for acoustic speed measurements. Put the microphones side by side to estimate the channel-delay offset ${\tau }_0$, swap channels to check its sign, and check the acquisition clock against an independent timing reference. Measure spacings between consistent points on the microphones, keep the source outside their span and on their axis, and fit $\Delta t=d/c+{\tau }_0$ by linear regression rather than forcing a zero intercept. Repeat pulses and compare the same onset criterion before reflections arrive. These are proposed calibration and control checks, not additional procedures claimed for the paper.

One sample corresponds to $1/f_s\simeq 22.7\mu \mathrm{s}$ for the reported sampling rate, so a delay of only a few samples has large relative measurement error. Background measurement noise, different microphone responses and threshold choice can move the apparent onset; reflections can supply a later competing pulse. Record temperature, keep air motion small, and compare several spacings. The result estimates speed of sound under those conditions; it cannot prove an exactly zero precursor below the instrument's sensitivity or establish finite propagation speed for every real medium. Both new diagrams in these sections show explicitly synthetic ideal wave equation solutions.^[ref]

Timing calibration for acoustic speed measurements separates travel delay from the offset ${\tau }_0$ between acquisition channels and checks the clock defining the sampling rate. A constant offset changes the intercept of a delay-versus-distance linear regression; a clock-scale error changes the inferred speed of sound itself.

A microphone is an acoustic transducer that converts an acoustic wave into an electrical signal. In an acoustic time-of-flight speed measurement, its response affects which recorded feature is identified as arrival.

The sampling rate is the number of recorded samples per unit time. A sample-index difference $N$ represents a delay $N/f_s$ when the acquisition clock is correctly calibrated.

Slowness is the reciprocal of phase velocity, measured in time per distance. A plane wave with wavevector $\mathbf{k}$ and angular frequency $\omega$ has vector slowness $\mathbf{p}=\mathbf{k}/\omega$. Its direction is the wavefront normal, which need not be the group velocity direction in an anisotropic medium.

In three spatial dimensions, the solution with initial position $u_0$ and velocity $u_1$ iswhere $M_t$ is the spherical mean.

Extend lower-dimensional data independently of an extra coordinate and apply the higher-dimensional wave equation solution. Projecting a three-dimensional sphere to the planar disk yields the kernel $[2\pi \sqrt{t^2-r^2}{]}^{-1}$ for the two-dimensional solution with zero initial displacement. This method of descent for the wave equation explains why three-dimensional wave propagation can depend on a sphere while two-dimensional propagation depends on its interior.

The strong Huygens principle says that the three-dimensional wave solution at $(t,x)$ depends only on initial data on the sphere $|y-x|=|t|$. A compactly supported disturbance therefore leaves no tail in the interior of its light cone.

For the unforced, constant-$c$ one-dimensional wave equation, finite propagation speed says that changing Cauchy data outside $[x_{\ast }-c{t}_{\ast },x_{\ast }+c{t}_{\ast }]$ cannot change the solution at $(x_{\ast },t_{\ast })$, where $c>0$ and $t_{\ast }\ge 0$. This interval is the initial-data domain of dependence. Its endpoints are found by following the two characteristic lines for the one-dimensional wave equation backwards from the observation. D'Alembert's formula makes the dependence sharper: initial displacement is sampled at the endpoints, while initial velocity is integrated over the entire interval.^[ref]