Lecture 17: Waves in Two and Three Dimensions

1. Revision of the Wave Equation in 1 Dimension

A string with a tension T and density µ kg/m is displaced sideways by an amount f (x, t). Consider a
short element δx. It is pulled transversely at each end by an amount T ∂f ∂x . These forces almost cancel, the
imbalance is due to the difference in ∂x between the two ends of the element. The net force is ∂(T (∂f
∂f
∂x
/∂x))
δx =
2
∂ f
p
T ∂x2 δx. The mass is µδx. So Newton II gives the wave equation, with v = T /µ

∂2f µ ∂2f
= (1)
∂x2 T ∂t2

2. The Wave Equation in 2 Dimensions

For 2D, consider an elastic sheet, like the surface of a balloon. It has a mass density we can again call
µ, but it’s in kg/m2 not kg/m. Stretch it uniformly. Then the force that keeps it stretched is applied not
just at the two ends but all round the edge. The force along a length L of the edge will be proportional to
L. Again we describe it by a tension T but it’s not in Newtons but in Newtons/metre.
Consider a rectangle of sides δx and δy. The force along one of the y edges (left and right) is T δy.
Looking at the projection on the x axis i.e. from the y direction, the transverse component is T δy ∂f
∂x . There
is one of these forces on both edges, at x and x + δx, and the imbalance that moves the rectangular element
∂(T δy ∂f
∂x )
2
sideways is δx, which is T ∂∂xf2 δxδy
∂x ³ 2 ´
∂2f
There is another force from the x edges (top and bottom), making a total of T ∂∂xf2 + ∂y 2 δxδy.
The mass is µδxδy so Newton’s 2nd law gives the wave Equation in 2 dimensions

∂2f ∂2f 2 µ ∂2f

+ = ∇ f = (2)
∂x2 ∂y 2 T ∂t2
q
T
Again we can write v = µ, remembering that the units are different.

3. Solving the equation

We need to find a function f (x, y, t) which satisfies this equation. This is done using a clever trick called
separation of variables. Suppose that this function can be written as the product of 3 separate functions of
the 3 variables. (There is no justification for this. We just try it and see.) f (x, y, t) = fx (x)fy (y)ft (t)
∂
The ∂x operator affects only fx . The others are constant as far as x is concerned. So the equation is
2 ∂2f 2
fy ft ∂∂xf2x + fx ft ∂y2y = Tµ fx fy ∂∂tf2t .
Now (!) divide by the function we first thought of, that is fx fy ft . The terms cancel leaving
2
1 ∂ fx ∂2f 2

fx ∂x2 + f1y ∂y2y = Tµ f1t ∂∂tf2t .

The RHS does not depend on x or y. The only variable in it is t. But it is equal to the LHS. And the
LHS has no t in it. As the LHS is independent of t, and the RHS is equal to the LHS, the RHS must also
be independent of t (as well as x and y). So it must be constant. Call this constant −k 2 .
Looking at the LHS, the first term contains x and only x. But there are no xs in the second term, or
the RHS. So this first term is constant - call it −kx2 . Similarly the second term can’t depend on y, and must
be constant - call it −ky2
So we have 3 equations

∂ 2 fx ∂ 2 fy 1 ∂ 2 ft
= −kx2 fx = −ky2 fy = −k 2 ft (3)
∂x2 ∂y 2 v 2 ∂t2
These have solutions, writing ω for vk:

fx = Cx e±ikx x fy = Cy e±iky y ft = Ct e±iωt (4)

We also have to remember
k 2 = kx2 + ky2 (5)
Lecture 17 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Waves in Two and Three Dimensions

4. Forms of the solution

Multiplying all these together to give the total f (x, y, t), the various arbitrary constants can all be
combined. The plus or minus choices correspond to directions of motion. We pick out

f (x, y, t) = Cei(kx x+ky y−ωt) (6)

What does this look like? The Real part is cos(kx x + ky y − ωt). Let’s start at t = 0. Along the x axis,
where y = 0, there is a wave with wave number kx . The variation is like wavelength λx = 2π kx . Along the y
axis where x = 0, there is a wave with wavelength 2π/ky .
The first peak on each axis occurs at kx x = 2π and ky y = 2π. Generally it occurs where kx x + ky y = 2π.
So this line describes a peak of the wave. Other peaks are along straight lines given by kx x + ky y = N π (N
even) and the troughs are given by the same straight lines with N odd. As time goes by these lines move.
(2π/k ) k
The gradient of the line is − (2π/kxy ) = −kx /ky . Motion is normal to these lines, with gradient kxy . The
distance between them is given by λ = 2π/k.

5. Wave vectors
~
We can write (x, y) = ~r. We can also write (kx , ky ) = ~k. Then the wave is ei(k.~r−ωt)
This is more than just notation. The direction of k gives the direction of travel of the wave. Also the
equation kx2 + ky2 = k 2 appears naturally, and we can write k 2 = ω 2 /v 2 . So ~k is a real vector, represented by
pointy arrow, transforming like x and y

6. Three dimensions
2
Waves in 3 dimensions are similar. The wave equation is the same ∇2 f = v12 ∂∂t2f (with an extra ∂ 2 f /∂z 2
~
term.) Separation of variables works again, giving solutions ei(k.~r−ωt) where ~k = kx , ky , kz .

7. Other solutions
Are these all the solutions? Yes and no.
7.1 Standing waves
Combining solutions with different ± signs in Eq 4, we get waves like sin(kx x)sin(ky y)sin(ωt). These
are the standing waves as opposed to the traveling waves. These are not of the form of Eq 6, but can be
made up from them. They describe the modes of vibration of rectangular sheets.
7.2 Solution in polar co-ordinates
This is just for fun, and I probably won’t have time for it in the lecture.
Suppose we work in terms of polar co-ordinates (r, θ) where
∂2f 1 ∂f 1 ∂2f 1 ∂2f
∇2 f = 2
+ + 2 2 = 2 2
∂r r ∂r r ∂θ v ∂t
Separation of variables is a little more tricky. First we separate off the time. f (r, θ, t) = frθ (r, θ)ft (t).
As before, there are eiωt solutions and we are left with

∂ 2 frθ 1 ∂frθ 1 ∂ 2 frθ

+ + = −k 2 frθ
∂r2 r ∂r r2 ∂θ2
Now multiply by r2 and separate into fr (r)fθ (θ)

r 2 ∂ 2 fr r ∂fr 1 ∂ 2 fθ
2
+ + k2 r2 + 2 =0
fr ∂r fr ∂r fθ ∂θ2
The θ part is interesting. Call the constant −m2 , then the solutions of f 00 (θ) = −m2 f (θ) are just
sin(mθ), or cos(mθ), or eimθ . m is an arbitrary constant, but it has to be integer as f (θ) = f (θ + 2π)
This leaves us with

r 2 ∂ 2 fr r ∂frθ
2
+ + k 2 r 2 = m2 r2 fr00 (r) + rfr0 (r) + (k 2 r2 − m2 )fr (r) = 0
fr ∂r fr ∂r
Solutions of this are messy (Bessel functions). At large r the small terms are neglected and the radial
solution is like eikr . So there are solutions ei(kr−ωt) sin(mθ) which describe circular (or spiral) waves. They
are not new: they can be made from the xy solutions by Fourier transforms. But they are appropriate for
situations with cylindrical symmetry.

–2–