Unit 12, lesson 72

The Wave Equation, Normal Modes, and d'Alembert's Formula

How the wave equation uses initial displacement and velocity, how standing modes solve it with fixed or free ends, and how traveling waves and d'Alembert's formula give the domain of dependence and finite propagation speed.

  • Video 72
  • Heat, Wave and Laplace Equations

A string held in the shape of a single arch can be released from rest, or it can be released while it is also given a flick. The two motions differ even though the starting shape is the same, so a vibrating string needs more initial information than a cooling rod. Here u(x,t)u(x, t) is the displacement of the string at position xx and time tt, and subscripts denote partial derivatives, so utt=∂2u/∂t2u_{tt} = \partial^2 u/\partial t^2 and uxx=∂2u/∂x2u_{xx} = \partial^2 u/\partial x^2. This lesson solves the wave equation in two ways, as a sum of standing modes and as a pair of traveling waves, and it uses the second form to show that disturbances move at a finite speed.

Key terms

Wave equation

Definition The one-dimensional wave equation with wave speed c>0c > 0 is

utt=c2 uxx.u_{tt} = c^2\,u_{xx}.

For small transverse vibrations of a string with tension τ\tau and mass per unit length ρ\rho, it follows from ρ utt=τ uxx\rho\,u_{tt} = \tau\,u_{xx}, so c=τ/ρc = \sqrt{\tau/\rho}. The equation is second order in time, so an initial-value problem needs two initial functions, the initial displacement u(x,0)=f(x)u(x, 0) = f(x) and the initial velocity ut(x,0)=g(x)u_t(x, 0) = g(x). On a bounded interval it also needs one boundary condition at each end.

In plain words The acceleration at each point is proportional to the curvature of the string there. Unlike heat flow, the motion does not die out, and knowing where the string is does not tell you how it is moving.

Example On 0≤x≤π0 \le x \le \pi with fixed ends and c=1c = 1, the data f(x)=sin⁡xf(x) = \sin x and g=0g = 0 give u=sin⁡xcos⁡tu = \sin x\cos t. The same displacement with the velocity g(x)=sin⁡xg(x) = \sin x gives

u=sin⁡x (cos⁡t+sin⁡t).u = \sin x\,(\cos t + \sin t).

Both start in the same shape and both satisfy utt=uxxu_{tt} = u_{xx}. The first never exceeds 11, while the second reaches the amplitude 2\sqrt{2} at t=π/4t = \pi/4, because cos⁡t+sin⁡t=2sin⁡(t+π/4)\cos t + \sin t = \sqrt{2}\sin(t + \pi/4).

Standing modes

Definition On 0≤x≤L0 \le x \le L with fixed ends u(0,t)=u(L,t)=0u(0, t) = u(L, t) = 0, the product u=X(x) T(t)u = X(x)\,T(t) separates the equation into X′′+λX=0X'' + \lambda X = 0 with X(0)=X(L)=0X(0) = X(L) = 0 and T′′+c2λT=0T'' + c^2\lambda T = 0, where primes are derivatives in each factor’s own variable. With λn=(nπ/L)2\lambda_n = (n\pi/L)^2 and the angular frequencies ωn=nπc/L\omega_n = n\pi c/L, the standing modes are the products Xn(x) Tn(t)X_n(x)\,T_n(t) with

Xn(x)=sin⁡nπxL,Tn(t)=Ancos⁡ωnt+Bnsin⁡ωnt,\begin{aligned} X_n(x) &= \sin\frac{n\pi x}{L}, \\[0.4em] T_n(t) &= A_n\cos\omega_n t \\[0.4em] &\quad + B_n\sin\omega_n t, \end{aligned}

for n=1,2,3,…n = 1, 2, 3, \dots. The solution with displacement ff and velocity gg is the sum of these modes with

An=fn,Bn=gnωn,A_n = f_n, \qquad B_n = \frac{g_n}{\omega_n},

where fnf_n and gng_n are the half-range sine coefficients of ff and gg. When ff and gg are smooth enough and compatible with the fixed ends, this series converges to the classical solution. The division by ωn\omega_n is needed because the derivative of sin⁡ωnt\sin\omega_n t is ωncos⁡ωnt\omega_n\cos\omega_n t. Each mode oscillates with the angular frequency ωn\omega_n and stays at rest at its nodes x=kL/nx = kL/n for k=0,1,…,nk = 0, 1, \dots, n. For a smooth solution with fixed ends the energy

E=12∫0L(ut2+c2ux2)dxE = \frac{1}{2}\int_0^L \left(u_t^2 + c^2u_x^2\right)dx

is constant in time.

In plain words Each mode is a fixed shape whose size swings back and forth at its own frequency. The initial displacement sets the cosine part of each swing, and the initial velocity, divided by the frequency, sets the sine part.

Example Take c=2c = 2 on 0≤x≤π0 \le x \le \pi with f=sin⁡xf = \sin x and g=sin⁡2xg = \sin 2x. The only nonzero coefficients are f1=1f_1 = 1 and g2=1g_2 = 1, and ω2=4\omega_2 = 4, so

u=sin⁡xcos⁡2t+14sin⁡2xsin⁡4t.\begin{aligned} u &= \sin x\cos 2t \\[0.4em] &\quad + \tfrac{1}{4}\sin 2x\sin 4t. \end{aligned}

For the first term both uttu_{tt} and 4uxx4u_{xx} equal −4-4 times the term, and for the second both equal −16-16 times it. At t=0t = 0 the displacement is sin⁡x\sin x and the velocity ut=−2sin⁡xsin⁡2t+sin⁡2xcos⁡4tu_t = -2\sin x\sin 2t + \sin 2x\cos 4t equals sin⁡2x\sin 2x, and both ends stay fixed. The energy equals 5π/45\pi/4 at every time. Leaving out the factor 14\tfrac{1}{4} would still give a solution of the equation, but with the initial velocity 4sin⁡2x4\sin 2x.

Fixed and free ends

Definition A fixed end holds the displacement at zero, so u=0u = 0 there, a Dirichlet condition. A free end, such as a ring that slides on a frictionless vertical rod, feels no vertical force from its support, so ux=0u_x = 0 there, a Neumann condition. With free ends at x=0x = 0 and x=Lx = L, the modes are cos⁡(nπx/L)\cos(n\pi x/L) for n=0,1,2,…n = 0, 1, 2, \dots, with the same ωn=nπc/L\omega_n = n\pi c/L. The zero mode n=0n = 0 has λ0=0\lambda_0 = 0 and the time equation T0′′=0T_0'' = 0, so

T0=A0+B0 t,T_0 = A_0 + B_0\,t,

where A0A_0 and B0B_0 are the mean values of ff and gg over the interval. The mean displacement of the string is A0+B0tA_0 + B_0 t, a uniform drift. A fixed end at 00 and a free end at LL give the modes sin⁡((2n−1)πx/(2L))\sin\big((2n - 1)\pi x/(2L)\big) for n=1,2,3,…n = 1, 2, 3, \dots.

In plain words A fixed end cannot move. A free end can move up and down but stays flat. With two free ends the whole string may also drift, and that rigid motion is the zero mode, which a sum of nonconstant cosines would miss.

Example With free ends on 0≤x≤π0 \le x \le \pi, c=1c = 1, f=0f = 0 and g=1+cos⁡xg = 1 + \cos x, the mean of gg is 11, and the mode n=1n = 1 has ω1=1\omega_1 = 1 and the velocity coefficient 11. So

u=t+sin⁡tcos⁡x.u = t + \sin t\cos x.

To check, utt=−sin⁡tcos⁡x=uxxu_{tt} = -\sin t\cos x = u_{xx}, u(x,0)=0u(x, 0) = 0, ut(x,0)=1+cos⁡xu_t(x, 0) = 1 + \cos x, and ux=−sin⁡tsin⁡xu_x = -\sin t\sin x vanishes at 00 and at π\pi. The string rocks while it drifts upward at the average speed 11.

Traveling waves

Definition For any twice continuously differentiable functions FF and GG on R\mathbb{R}, the function

u=F(x−ct)+G(x+ct)u = F(x - ct) + G(x + ct)

solves the wave equation. The term F(x−ct)F(x - ct) is a profile moving right at speed cc, and G(x+ct)G(x + ct) is one moving left at speed cc. Conversely, every twice continuously differentiable solution for x∈Rx \in \mathbb{R} and t≥0t \ge 0 has this form. To see it, set ξ=x−ct\xi = x - ct and η=x+ct\eta = x + ct and write u(x,t)=U(ξ,η)u(x, t) = U(\xi, \eta). The chain rule gives

utt−c2uxx=−4c2 Uξη,u_{tt} - c^2u_{xx} = -4c^2\,U_{\xi\eta},

so the equation says Uξη=0U_{\xi\eta} = 0. Then UξU_\xi depends on ξ\xi alone, and integrating once more gives U=F(ξ)+G(η)U = F(\xi) + G(\eta).

In plain words On the whole line every smooth solution is two shapes sliding in opposite directions at the same speed, passing through each other without interacting.

Example By the product formulas for sines and cosines, the first mode of the example above splits as sin⁡xcos⁡2t=12sin⁡(x−2t)+12sin⁡(x+2t)\sin x\cos 2t = \tfrac{1}{2}\sin(x - 2t) + \tfrac{1}{2}\sin(x + 2t), and the second as 14sin⁡2xsin⁡4t=18cos⁡(2x−4t)−18cos⁡(2x+4t)\tfrac{1}{4}\sin 2x\sin 4t = \tfrac{1}{8}\cos(2x - 4t) - \tfrac{1}{8}\cos(2x + 4t). So the two-mode solution is F(x−2t)+G(x+2t)F(x - 2t) + G(x + 2t) with

F(ξ)=12sin⁡ξ+18cos⁡2ξ,G(η)=12sin⁡η−18cos⁡2η.\begin{gathered} F(\xi) = \tfrac{1}{2}\sin\xi + \tfrac{1}{8}\cos 2\xi, \\[0.4em] G(\eta) = \tfrac{1}{2}\sin\eta - \tfrac{1}{8}\cos 2\eta. \end{gathered}

A standing wave is two traveling waves of equal size moving in opposite directions.

d’Alembert’s formula

Definition Let c>0c > 0, let ff be twice continuously differentiable and let gg be continuously differentiable on R\mathbb{R}. The problem utt=c2uxxu_{tt} = c^2u_{xx} for x∈Rx \in \mathbb{R} and t≥0t \ge 0, with u(x,0)=f(x)u(x, 0) = f(x) and ut(x,0)=g(x)u_t(x, 0) = g(x), has exactly one twice continuously differentiable solution. Its value at each point (x,t)(x, t) is given by d’Alembert’s formula

u=12f(x−ct)+12f(x+ct)+12c∫x−ctx+ctg(s) ds.\begin{aligned} u &= \tfrac{1}{2}f(x - ct) \\[0.4em] &\quad + \tfrac{1}{2}f(x + ct) \\[0.4em] &\quad + \frac{1}{2c}\int_{x - ct}^{x + ct} g(s)\,ds. \end{aligned}

It comes from the traveling form u=F(x−ct)+G(x+ct)u = F(x - ct) + G(x + ct). The data give F+G=fF + G = f and −cF′+cG′=g-cF' + cG' = g, and together with F′+G′=f′F' + G' = f' this yields

F′=12f′−12c g,G′=12f′+12c g.\begin{gathered} F' = \tfrac{1}{2}f' - \tfrac{1}{2c}\,g, \\[0.4em] G' = \tfrac{1}{2}f' + \tfrac{1}{2c}\,g. \end{gathered}

With an antiderivative HH of gg, this gives F=12f−H/(2c)+kF = \tfrac{1}{2}f - H/(2c) + k and G=12f+H/(2c)−kG = \tfrac{1}{2}f + H/(2c) - k for a constant kk, which cancels in the sum, and H(x+ct)−H(x−ct)H(x + ct) - H(x - ct) is the integral of gg. On a string 0≤x≤L0 \le x \le L with fixed ends, the same formula holds after ff and gg are replaced by their odd 2L2L-periodic extensions, and with free ends by their even 2L2L-periodic extensions. For the odd extensions of ff and gg to be twice and once continuously differentiable, the values f(0)f(0), f′′(0)f''(0) and g(0)g(0) and the same values at LL must vanish.

In plain words Half of the initial shape travels left and half travels right. For t>0t > 0 the velocity term equals tt times the average of gg over the interval from x−ctx - ct to x+ctx + ct.

Example For c=2c = 2, f=sin⁡xf = \sin x and g=sin⁡2xg = \sin 2x on the whole line, the displacement part is 12(sin⁡(x−2t)+sin⁡(x+2t))=sin⁡xcos⁡2t\tfrac{1}{2}\big(\sin(x - 2t) + \sin(x + 2t)\big) = \sin x\cos 2t. The velocity part is

14∫x−2tx+2tsin⁡2s ds,\frac{1}{4}\int_{x - 2t}^{x + 2t} \sin 2s\,ds,

which equals 18(cos⁡(2x−4t)−cos⁡(2x+4t))=14sin⁡2xsin⁡4t\tfrac{1}{8}\big(\cos(2x - 4t) - \cos(2x + 4t)\big) = \tfrac{1}{4}\sin 2x\sin 4t. So the formula reproduces the standing-mode solution. Restricted to 0≤x≤π0 \le x \le \pi it also has fixed ends, because sin⁡x\sin x and sin⁡2x\sin 2x are already odd and 2π2\pi-periodic. A pulse behaves differently. The data f=e−x2f = e^{-x^2}, g=0g = 0 and c=1c = 1 give u=12e−(x−t)2+12e−(x+t)2u = \tfrac{1}{2}e^{-(x - t)^2} + \tfrac{1}{2}e^{-(x + t)^2}, two bumps of half the height moving apart.

Domain of dependence

Definition By d’Alembert’s formula, the value u(x0,t0)u(x_0, t_0) with t0>0t_0 > 0 depends only on ff at the two points x0±ct0x_0 \pm ct_0 and on gg over the interval

[x0−ct0, x0+ct0],[x_0 - ct_0,\ x_0 + ct_0],

which is the domain of dependence of the point (x0,t0)(x_0, t_0). In the (x,t)(x, t) plane it is the base of the triangle cut out by the two characteristic lines x−ct=x0−ct0x - ct = x_0 - ct_0 and x+ct=x0+ct0x + ct = x_0 + ct_0. Conversely, changing the data only on an interval [a,b][a, b] changes uu only in the region a−ct≤x≤b+cta - ct \le x \le b + ct, the domain of influence of that interval. This is finite propagation speed. In particular, if ff and gg vanish outside [a,b][a, b], then u(x,t)=0u(x, t) = 0 whenever x<a−ctx < a - ct or x>b+ctx > b + ct. The heat equation behaves differently. On the whole line, the heat kernel solution with a bounded continuous initial temperature that is nonnegative and positive somewhere is positive everywhere at every time t>0t > 0.

In plain words A signal on the string cannot travel faster than cc. A point far from the initial disturbance stays at rest until the first wave front reaches it.

Example Let c=2c = 2, g=0g = 0, and let f(x)=(1−x2)3f(x) = (1 - x^2)^3 for ∣x∣≤1\lvert x \rvert \le 1 and f=0f = 0 otherwise, which is twice continuously differentiable. At x=10x = 10 the term f(10+2t)f(10 + 2t) is always zero, so u(10,t)=12f(10−2t)u(10, t) = \tfrac{1}{2}f(10 - 2t). The point is at rest for t≤4.5t \le 4.5, the right-moving half of the pulse passes during 4.5<t<5.54.5 < t < 5.5 with the peak 12\tfrac{1}{2} at t=5t = 5, and the point is at rest again for t≥5.5t \ge 5.5. An initial velocity leaves something behind. If instead f=0f = 0 and gg vanishes outside [−1,1][-1, 1], then for t≥5.5t \ge 5.5 the whole interval [−1,1][-1, 1] lies inside [10−2t,10+2t][10 - 2t, 10 + 2t], and

u(10,t)=14∫−11g(s) ds,u(10, t) = \frac{1}{4}\int_{-1}^{1} g(s)\,ds,

a constant displacement that stays after the waves have passed.

Check yourself

  1. Solve utt=9uxxu_{tt} = 9u_{xx} on 0≤x≤π0 \le x \le \pi with fixed ends, u(x,0)=0u(x, 0) = 0 and ut(x,0)=sin⁡2xu_t(x, 0) = \sin 2x. Check the initial velocity of your answer.
Show the answer

Here c=3c = 3, the only nonzero coefficient is g2=1g_2 = 1, and ω2=3⋅2=6\omega_2 = 3 \cdot 2 = 6. So A2=0A_2 = 0 and B2=g2/ω2=1/6B_2 = g_2/\omega_2 = 1/6, which gives

u=16sin⁡2xsin⁡6t.u = \tfrac{1}{6}\sin 2x\sin 6t.

Its velocity is ut=sin⁡2xcos⁡6tu_t = \sin 2x\cos 6t, which equals sin⁡2x\sin 2x at t=0t = 0, and the displacement starts at 00. Writing sin⁡2xsin⁡6t\sin 2x\sin 6t without the factor 16\tfrac{1}{6} is a common slip. It solves the equation and the end conditions, but its initial velocity is 6sin⁡2x6\sin 2x.

  1. Solve utt=uxxu_{tt} = u_{xx} on 0≤x≤π0 \le x \le \pi with free ends, u(x,0)=cos⁡xu(x, 0) = \cos x and ut(x,0)=2u_t(x, 0) = 2. What is the mean displacement at t=3t = 3? What happens if the zero mode is left out?
Show the answer

The displacement cos⁡x\cos x is the mode n=1n = 1 with ω1=1\omega_1 = 1, and the constant velocity 22 belongs entirely to the zero mode, with A0=0A_0 = 0 and B0=2B_0 = 2. So

u=2t+cos⁡xcos⁡t.u = 2t + \cos x\cos t.

To check, utt=−cos⁡xcos⁡t=uxxu_{tt} = -\cos x\cos t = u_{xx}, ux=−sin⁡xcos⁡tu_x = -\sin x\cos t vanishes at both ends, u(x,0)=cos⁡xu(x, 0) = \cos x and ut(x,0)=2u_t(x, 0) = 2. The cosine integrates to zero over the interval, so the mean displacement is 2t2t, which is 66 at t=3t = 3. Without the zero mode, the constant 22 has no cosine coefficient with n≥1n \ge 1 to go into, and the result cos⁡xcos⁡t\cos x\cos t starts at rest, so it does not meet the velocity condition.

  1. On the whole line with c=4c = 4, the data ff (twice continuously differentiable) and gg (continuously differentiable) vanish outside [0,1][0, 1]. For which times can u(9,t)u(9, t) be nonzero, and what is u(9,t)u(9, t) for t≥9/4t \ge 9/4?
Show the answer

The domain of dependence of (9,t)(9, t) is [9−4t,9+4t][9 - 4t, 9 + 4t], and the data are nonzero only inside (0,1)(0, 1). The two intervals overlap in more than one point only when 9−4t<19 - 4t < 1, that is t>2t > 2, so u(9,t)=0u(9, t) = 0 for 0≤t≤20 \le t \le 2. For t≥9/4t \ge 9/4 the left end 9−4t9 - 4t is at most 00, so both values f(9±4t)f(9 \pm 4t) are zero and the interval contains all of [0,1][0, 1]. Then

u(9,t)=18∫01g(s) dsu(9, t) = \frac{1}{8}\int_0^1 g(s)\,ds

for every t≥9/4t \ge 9/4. The point does not return to rest unless the integral of gg is zero, which is a property of the wave equation in one space dimension.

How this lesson connects