Professional Documents
Culture Documents
Introduction To The Wave Equation(s)
Introduction To The Wave Equation(s)
@ @
+c =0 (1.1)
@t @x
for real, constant c > 0. As (1.1) is first order in time, we need a single initial condition of
which implies that the initial condition simply propagates to the right with permanent form.
Why is this the case? We plug the solution (1.3) back into (1.1), but first: let u = x ct
so that any function f (x ct) = f (u) and the derivative of the function with respect to u is
notated as df /du = f 0 . Now
@ @ 0 (x ct) @ o (u)
d 0 @u d 0
= = =
= c
@t @t @t du @t du
@ @ 0 (x ct) @ o (u) d 0 @u d 0
= = = =
@x @x @x du @x du
Note the di↵erence in the partial and total derivative above. Thus the first-order wave
equation becomes
d 0 d 0
c +c =0
du du
3
1.1.1 Plane wave solution and dispersion relationship
A common practice is to plug in a propagating wave solution such as cos(kx !t) or sin(kx
!t) into the governing equations and hunting for a solution and dispersion equation. For
linear systems it is often more convenient to use complex notation. For example, let have
a wave like solutions where
= ei(kx !t)
(1.4)
where k is a wavenumber with units [rad/m] and a wavelength = 2⇡/k. Similarly the
radian frequency ! has units [rad/s] and is associated with the wave period T = 2⇡/!.
Plug the wave solution (1.4) into (1.1) gives us a dispersion relationship
[ i! + ikc]ei(kx !t)
=) ! = kc.
Note that this dispersion relationship has !(k) as a linear function of k. Or that !/k = c
is a constant and does not depend upon k. The type of waves that propagate by such a
dispersion relationship are called non-dispersive waves. This means that all frequencies or
wavenumbers propagate at the same speed - in this case c.
where the integral is over all k. Now plug the Fourier-integral representation (2.15) into
(1.1) and one gets
Z Z
@ ˆ ikx
e dk + ikc ˆeikx dk = 0 (1.7)
@t
2 3
Z
6@ ˆ 7
4 + ikc ˆ5 eikx dk = 0 (1.8)
|@t {z }
=0
4
Thus ˆ(k) has solutions of the form ˆ(k) = Ce ikct
and with the initial condition (2.11) at
t = 0, it then is clear that the full solution is
Z
(x, t) = ˆ0 (k)eik(x ct)
dk (1.9)
which is back to the old familiar form of 0 (x ct) a solution which just translates.
This is a fairly simple example of using Fourier analysis and substitution to get solutions
to PDEs. However, this is a mainstay of wave analysis.
that is two permanent forms propagating right and left. We can show this the same way as
with the first-order wave equation with variables u = x ct and v = x + ct, then
@2 2@
2 2
2d f 2 df
2
2d g dg
2
c 2
= c 2
c = c 2
c2 =0 (1.18)
@t @x du du du du
As (1.13) is linear, one can also linearly superimpose solutions.
where the prime denotes the derivative. The second equation can be integrated to
Z x
1
g(x) f (x) = c b(x̃)dx̃ (1.21)
x0
where x̃ is a dummy variable of integration. Adding and subtracting (1.19) and (1.21) gives
Z x
1 1
f (x) = a(x) c b(x̃)dx̃ (1.22)
2 x0
Z x
1 1
g(x) = a(x) + c b(x̃)dx̃ (1.23)
2 x0
6
and rewritting Z Z
x ct x0
b(x̃)dx̃ = b(x̃)dx̃
x0 x ct
[ ! 2 + c2 k 2 ]ei(kx !t)
=) ! 2 = c2 k 2 .
= ei(kx+!t) (1.28)
which results in
[ ! 2 + c2 k 2 ]ei(kx !t)
=) ! 2 = c2 k 2 .
Analogous to g(x+ct) being a solution. They are the same! Here, the dispersion relationship
is also non-dispersive as |!|/k = c is a constant. Thus, all frequencies and wavenumbers
propagate at the same speed - and if you know c and k, you also know the frequency.
7
Now plug the Fourier-integral representation (2.15) into (1.14) and one gets
Z Z
@ 2 ˆ ikx
e dk + c2 k 2 ˆeikx dk = 0 (1.31)
@t2
2 3
Z
6 @2 ˆ 2 2 7 ikx
4 2 + c k ˆ5 e dk = 0 (1.32)
|@t {z }
=0
The quantity in brackets is the good-old fashioned harmonic oscillator equation. It has
solutions
ˆ(k, t) = A(k)ei!t + B(k)e i!t
(1.33)
which is back to the familiar form of f (x ct) and g(x + ct) solutions. One can also do this
analysis for a general initial condition, but it is just more algebra.
Why a sum and not an integral? Because only particular sin() that match the boundary
conditions are allowed. On an infinite domain, we use integrals instead of sums. The initial
8
conditions must also be similarly written
1
X
(x, t = 0) = a(x) = ân sin(n⇡x/L). (1.37)
n=1
The quantity in brackets is again the harmonic oscillator equation with solution
Now recall that @ /@t = 0 at t = 0 which means that all c2n = 0. Matching the initial
condition implies that c1n = ân and thus the full solution is written as
1
X
(x, t) = ân cos(n⇡ct/L) sin(n⇡x/L). (1.40)
n=1
Now, this solution is interesting. It is not obviously in a form of f (x ct and g(x + ct) as
were the earlier solutions on an infinite domain. Such solutions are progressive in that the
waves move. Instead the solution has the form cos(ct) sin(x) - representing harmonic motions
that is standing in space and time. Such waves are known as standing waves. Think of a
guitar string with modes n = 1, n = 2, etc.
How to think about the distinction between progressive and standing? Departing from
exact mathematical solutions and using heuristic thinking, Suppose you have a wave train
+x propagating towards a perfectly reflective wall at x = L. A little later, there will be waves
propagating towards and away from the wall, i.e., = cos(x ct) + cos(x + ct) (supposing
k = 1). Using trigonometric identies1 we can rewrite
Thus, we see that standing waves are superposition of waves propagating in the ±x direction
at the same frequency and wavenumber. This provides an interpretation of the finite domain
solution. There really are f (x ct) and g(y + ct) modes propagating within 0 < x < L, but
they are constantly reflecting back and forth.
Many wave types allow for both progressive and standing components. In summary,
progressive waves have a form cos(kx !t) or exp(ikx !t). Standing waves have a form
cos(kx) sin(!t).
1
cos(a + b) = cos(a) cos(b) sin(a) sin(b)
9
1.3 Problem Set
1. Imagine you are lowering a thermistor (temperature sensor) to measure the vertical
(z) temperature profile. You don’t have a pressure sensor on the instrument but you
know that the instrument falls at a steady velocity w [m/s]. From the time-series of
temperature that the thermistor measures, how would you estimate the T (z)?
t + xxx =0 (1.42)
t + xxxx =0 (1.43)
For each, determine (a) if plane wave solutions (i.e., exp[i(kx !(k)t]) are permissible
or if not why not, and (b) if so, what is the dispersion relationsihp ! = !(k).
4. Consider the 3D second-order wave equation (1.13) with plane wave solutions / exp[i(kx+
ly + mz)]. What is the dispersion relation ! = !(k, l, z)?
5. For the 1D 2nd-order wave equation on 1 < x < 1, consider the initial condition
(x, t = 0) = (x) and @ (x, t = 0)/@t = 0, what is the solution for (x, t)?
6. Similar to above but with initial conditions (x, t = 0) = 0 and @ (x, t = 0)/@t = (x),
what is the solution for (x, t)? Use the fact that
Z
(x) = eikx dk
(a) Use the initial conditions to come up with an expression for A(k) and B(k) in
(1.33).
(b) Which are larger in amplitude, longer or shorter wavelengths?
(c) (EXTRA CREDIT) What is going on at k = 0?
10
section (3.2), what would the form be for a solution for (x, y, t)? Plug it into (1.44),
does it work as a solution?
11