For each energy (z)(𝑧), this gives a constant energy curve.

Plotting the
family of energy curves we obtain the phase portrait shown in
Figure 7.5.127.5.12.

Fi
gure 1.1We begin the study of partial differential equations with the
problem of heat flow in a uniform bar of length L𝐿, situated on the x𝑥 axis
with one end at the origin and the other at x=L𝑥=𝐿 (Figure 12.1.1 ).
We assume that the bar is perfectly insulated except possibly at its
endpoints, and that the temperature is constant on each cross section and
therefore depends only on x𝑥 and t𝑡. We also assume that the thermal
properties of the bar are independent of x𝑥 and t𝑡. In this case, it can be
shown that the temperature u=u(x,t)𝑢=𝑢(𝑥,𝑡) at time t𝑡 at a point x𝑥 units
from the origin satisfies the partial differential equation
ut=a2uxx,0<x<L,t>0,𝑢𝑡=𝑎2𝑢𝑥𝑥,0<𝑥<𝐿,𝑡>0,
where a𝑎 is a positive constant determined by the thermal properties. This
is the heat equation.

Figure 12.1.1 : A uniform bar

of length L𝐿
To determine u𝑢, we must specify the temperature at every point in the bar
when t=0𝑡=0, say
u(x,0)=f(x),0≤x≤L.𝑢(𝑥,0)=𝑓(𝑥),0≤𝑥≤𝐿.
We call this the initial condition. We must also specify boundary
conditions that u𝑢 must satisfy at the ends of the bar for all t>0𝑡>0. We’ll
call this problem an initial-boundary value problem.
We begin with the boundary conditions u(0,t)=u(L,t)=0𝑢(0,𝑡)=𝑢(𝐿,𝑡)=0,
and write the initial-boundary value problem as
ut=a2uxx,0<x<L,t>0,u(0,t)=0,u(L,t)=0,t>0,u(x,0)=f(x),0≤x≤L.(12.1.1)
(12.1.1)𝑢𝑡=𝑎2𝑢𝑥𝑥,0<𝑥<𝐿,𝑡>0,𝑢(0,𝑡)=0,𝑢(𝐿,𝑡)=0,𝑡>0,𝑢(𝑥,0)=𝑓(𝑥),0≤𝑥≤𝐿
.
Our method of solving this problem is called separation of variables (not to
be confused with method of separation of variables used in Section 2.2 for
solving ordinary differential equations). We begin by looking for functions
of the form
v(x,t)=X(x)T(t)𝑣(𝑥,𝑡)=𝑋(𝑥)𝑇(𝑡)
that are not identically zero and satisfy
vt=a2vxx,v(0,t)=0,v(L,t)=0𝑣𝑡=𝑎2𝑣𝑥𝑥,𝑣(0,𝑡)=0,𝑣(𝐿,𝑡)=0
for all (x,t)(𝑥,𝑡). Since
vt=XT′andvxx=X′′T,𝑣𝑡=𝑋𝑇′and𝑣𝑥𝑥=𝑋″𝑇,
vt=a2vxx𝑣𝑡=𝑎2𝑣𝑥𝑥 if and only if
XT′=a2X′′T,𝑋𝑇′=𝑎2𝑋″𝑇,
which we rewrite as
T′a2T=X′′X.𝑇′𝑎2𝑇=𝑋″𝑋.
Since the expression on the left is independent of x𝑥 while the one on the
right is independent of t𝑡, this equation can hold for all (x,t)(𝑥,𝑡) only if the
two sides equal the same constant, which we call a separation constant,
and write it as −λ−𝜆; thus,
X′′X=T′a2T=−λ.𝑋″𝑋=𝑇′𝑎2𝑇=−𝜆.
This is equivalent to
X′′+λX=0𝑋″+𝑋=0
and
T′=−a2λT.(12.1.2)(12.1.2)𝑇′=−𝑎2𝑇.
Since v(0,t)=X(0)T(t)=0𝑣(0,𝑡)=𝑋(0)𝑇(𝑡)=0 and v(L,t)=X(L)T(t)=0𝑣(𝐿,𝑡)=
𝑋(𝐿)𝑇(𝑡)=0 and we don’t want T𝑇 to be identically
zero, X(0)=0𝑋(0)=0 and X(L)=0𝑋(𝐿)=0. Therefore λ𝜆 must be an
eigenvalue of the boundary value problem
X′′+λX=0,X(0)=0,X(L)=0,(12.1.3)(12.1.3)𝑋″+𝑋=0,𝑋(0)=0,𝑋(𝐿)=0,
and X𝑋 must be a λ𝜆-eigenfunction. From Theorem 11.1.2, the
eigenvalues of Equation 12.1.312.1.3 are λn=n2π2/L2𝑛=𝑛2𝜋2/𝐿2, with
associated eigenfunctions
Xn=sinnπxL,n=1,2,3,….𝑋𝑛=sin⁡𝑛𝑥𝐿,𝑛=1,2,3,….
Substituting λ=n2π2/L2𝜆=𝑛2𝜋2/𝐿2 into Equation 12.1.212.1.2 yields
T′=−(n2π2a2/L2)T,𝑇′=−(𝑛2𝜋2𝑎2/𝐿2)𝑇,
which has the solution
Tn=e−n2π2a2t/L2.𝑇𝑛=𝑒−𝑛2𝜋2𝑎2𝑡/𝐿2.
Now let
vn(x,t)=Xn(x)Tn(t)=e−n2π2a2t/L2sinnπxL,n=1,2,3,…
𝑣𝑛(𝑥,𝑡)=𝑋𝑛(𝑥)𝑇𝑛(𝑡)=𝑒−𝑛2𝜋2𝑎2𝑡/𝐿2sin⁡𝑛𝑥𝐿,𝑛=1,2,3,…
Since
vn(x,0)=sinnπxL,𝑣𝑛(𝑥,0)=sin⁡𝑛𝑥𝐿,
vn𝑣𝑛 satisfies Equation 12.1.112.1.1 with f(x)=sinnπx/L𝑓(𝑥)=sin⁡𝑛𝑥/𝐿.
More generally, if α1,…,αm𝛼1,…,𝑚 are constants and
um(x,t)=∑n=1mαne−n2π2a2t/
L2sinnπxL,𝑢𝑚(𝑥,𝑡)=∑𝑛=1𝑚𝑛𝑒−𝑛2𝜋2𝑎2𝑡/𝐿2sin⁡𝑛𝑥𝐿,
then um𝑢𝑚 satisfies Equation 12.1.112.1.1 with
f(x)=∑n=1mαnsinnπxL.𝑓(𝑥)=∑𝑛=1𝑚𝑛sin⁡𝑛𝑥𝐿.
This motivates the next definition.
Definition 12.1.1
The formal solution of the initial-boundary value problem
ut=a2uxx,0<x<L,t>0,u(0,t)=0,u(L,t)=0,t>0,u(x,0)=f(x),0≤x≤L(12.1.4)
(12.1.4)𝑢𝑡=𝑎2𝑢𝑥𝑥,0<𝑥<𝐿,𝑡>0,𝑢(0,𝑡)=0,𝑢(𝐿,𝑡)=0,𝑡>0,𝑢(𝑥,0)=𝑓(𝑥),0≤𝑥≤𝐿
is
u(x,t)=∑n=1∞αne−n2π2a2t/L2sinnπxL,(12.1.5)
(12.1.5)𝑢(𝑥,𝑡)=∑𝑛=1∞𝑛𝑒−𝑛2𝜋2𝑎2𝑡/𝐿2sin⁡𝑛𝑥𝐿,
where
S(x)=∑n=1∞αnsinnπxL𝑆(𝑥)=∑𝑛=1∞𝑛sin⁡𝑛𝜋𝑥𝐿
is the Fourier sine series of f𝑓 on [0,L][0,𝐿]; that is,
αn=2L∫L0f(x)sinnπxLdx.𝑛=2𝐿∫0𝐿𝑓(𝑥)sin⁡𝑛𝑥𝐿𝑑𝑥.
We use the term “formal solution” in this definition because it is not in
general true that the infinite series in Equation 12.1.512.1.5 actually
satisfies all the requirements of the initial-boundary value problem
Equation 12.1.412.1.4 when it does, we say that it is an actual solution of
Equation 12.1.412.1.4.
Because of the negative exponentials in
Equation 12.1.512.1.5, u𝑢 converges for all (x,t)
(𝑥,𝑡) with t>0𝑡>0 (Exercise 12.1.54). Since each term in
Equation 12.1.512.1.5 satisfies the heat equation and the boundary
conditions in Equation 12.1.412.1.4, u𝑢 also has these properties
if ut𝑢𝑡 and uxx𝑢𝑥𝑥 can be obtained by differentiating the series in
Equation 12.1.512.1.5 term by term once with respect to t𝑡 and twice with
respect to x𝑥, for t>0𝑡>0. However, it is not always legitimate to
differentiate an infinite series term by term. The next theorem gives a
useful sufficient condition for legitimacy of term by term differentiation of
an infinite series. We omit the proof.
Theorem 12.1.2
A convergent infinite series
W(z)=∑n=1∞wn(z)𝑊(𝑧)=∑𝑛=1∞𝑤𝑛(𝑧)
can be differentiated term by term on a closed interval [z1,z2][𝑧1,𝑧2] to
obtain
W′(z)=∑n=1∞w′n(z)𝑊′(𝑧)=∑𝑛=1∞𝑤𝑛′(𝑧)
((where the derivatives at z=z1𝑧=𝑧1 and z=z2𝑧=𝑧2 are one-
sided)) provided that w′n𝑤𝑛′ is continuous on [z1,z2][𝑧1,𝑧2] and
|w′n(z)|≤Mn,z1≤z≤z2,n=1,2,3,…,|𝑤𝑛′(𝑧)|≤𝑀𝑛,𝑧1≤𝑧≤𝑧2,𝑛=1,2,3,…,
where M1,𝑀1, M2,𝑀2, …, Mn,𝑀𝑛, …, are constants such that the
series ∑∞n=1Mn∑𝑛=1∞𝑀𝑛 converges.
Theorem 12.1.2 , applied twice with z=x𝑧=𝑥 and once with z=t𝑧=𝑡, shows
that uxx𝑢𝑥𝑥 and ut𝑢𝑡 can be obtained by differentiating u𝑢 term by term
if t>0𝑡>0 (Exercise 12.1.54). Therefore u𝑢 satisfies the heat equation and
the boundary conditions in Equation 12.1.412.1.4 for t>0𝑡>0. Therefore,
since u(x,0)=S(x)𝑢(𝑥,0)=𝑆(𝑥) for 0≤x≤L0≤𝑥≤𝐿, u𝑢 is an actual solution
of Equation 12.1.412.1.4 if and only
if S(x)=f(x)𝑆(𝑥)=𝑓(𝑥) for 0≤x≤L0≤𝑥≤𝐿. From Theorem 11.3.2, this is true
if f𝑓 is continuous and piecewise smooth on [0,L][0,𝐿],
and f(0)=f(L)=0𝑓(0)=𝑓(𝐿)=0.
In this chapter we’ll define formal solutions of several kinds of problems.
When we ask you to solve such problems, we always mean that you should
find a formal solution.
Example 12.1.1
Solve
Equation 12.1.412.1.4 with f(x)=x(x2−3Lx+2L2)𝑓(𝑥)=𝑥(𝑥2−3𝐿𝑥+2𝐿2).
Solution
From Example [example:11.3.6}, the Fourier sine series of f𝑓 on [0,L]
[0,𝐿] is
S(x)=12L3π3∑n=1∞1n3sinnπxL𝑆(𝑥)=12𝐿3𝜋3∑𝑛=1∞1𝑛3sin⁡𝑛𝑥𝐿
Therefore
u(x,t)=12L3π3∑n=1∞1n3e−n2π2a2t/
L2sinnπxL𝑢(𝑥,𝑡)=12𝐿3𝜋3∑𝑛=1∞1𝑛3𝑒−𝑛2𝜋2𝑎2𝑡/𝐿2sin⁡𝑛𝑥𝐿
If both ends of bar are insulated so that no heat can pass through them,
then the boundary conditions are
ux(0,t)=0,ux(L,t)=0,t>0.𝑢𝑥(0,𝑡)=0,𝑢𝑥(𝐿,𝑡)=0,𝑡>0.
We leave it to you (Exercise 12.1.1) to use the method of separation of
variables and Theorem 11.1.3 to motivate the next definition.
Definition 12.1.3 : Formal Solution
The formal solution of the initial-boundary value problem
ut=a2uxx,0<x<L,t>0,ux(0,t)=0,ux(L,t)=0,t>0,u(x,0)=f(x),0≤x≤L(12.1.6)
(12.1.6)𝑢𝑡=𝑎2𝑢𝑥𝑥,0<𝑥<𝐿,𝑡>0,𝑢𝑥(0,𝑡)=0,𝑢𝑥(𝐿,𝑡)=0,𝑡>0,𝑢(𝑥,0)=𝑓(𝑥),0≤𝑥
≤𝐿
is
u(x,t)=α0+∑n=1∞αne−n2π2a2t/
L2cosnπxL,𝑢(𝑥,𝑡)=𝛼0+∑𝑛=1∞𝑛𝑒−𝑛2𝜋2𝑎2𝑡/𝐿2cos⁡𝑛𝑥𝐿,
where
C(x)=α0+∑n=1∞αncosnπxL𝐶(𝑥)=𝛼0+∑𝑛=1∞𝑛cos⁡𝑛𝜋𝑥𝐿
is the Fourier cosine series of f𝑓 on [0,L];[0,𝐿]; that is,
α0=1L∫L0f(x)dxandαn=2L∫L0f(x)cosnπxLdx,n=1,2,3,
….𝛼0=1𝐿∫0𝐿𝑓(𝑥)𝑑𝑥and𝑛=2𝐿∫0𝐿𝑓(𝑥)cos⁡𝑛𝑥𝐿𝑑𝑥,𝑛=1,2,3,….
Example 12.1.2
Solve Equation 12.1.612.1.6 with f(x)=x𝑓(𝑥)=𝑥.
Solution
From Example 11.3.1, the Fourier cosine series of f𝑓 on [0,L][0,𝐿] is
C(x)=L2−4Lπ2∑n=1∞1(2n−1)2cos(2n−1)πxL𝐶(𝑥)=𝐿2−4𝐿2∑𝑛=1∞1(2𝑛
−1)2cos⁡(2𝑛−1)𝑥𝐿
Therefore
u(x,t)=L2−4Lπ2∑n=1∞1(2n−1)2e−(2n−1)2π2a2t/
L2cos(2n−1)πxL.𝑢(𝑥,𝑡)=𝐿2−4𝐿2∑𝑛=1∞1(2𝑛−1)2𝑒−(2𝑛−1)2𝜋2𝑎2𝑡/
𝐿2cos⁡(2𝑛−1)𝑥𝐿.
We leave it to you (Exercise 12.1.2) to use the method of separation of
variables and Theorem 11.1.4 to motivate the next definition.
Definition 12.1.4
The formal solution of the initial-boundary value problem
ut=a2uxx,0<x<L,t>0,u(0,t)=0,ux(L,t)=0,t>0,u(x,0)=f(x),0≤x≤L(12.1.7)
(12.1.7)𝑢𝑡=𝑎2𝑢𝑥𝑥,0<𝑥<𝐿,𝑡>0,𝑢(0,𝑡)=0,𝑢𝑥(𝐿,𝑡)=0,𝑡>0,𝑢(𝑥,0)=𝑓(𝑥),0≤𝑥≤
𝐿
is
u(x,t)=∑n=1∞αne−(2n−1)2π2a2t/
4L2sin(2n−1)πx2L,𝑢(𝑥,𝑡)=∑𝑛=1∞𝑛𝑒−(2𝑛−1)2𝜋2𝑎2𝑡/4𝐿2sin⁡(2𝑛−1)𝑥2𝐿,
where
SM(x)=∑n=1∞αnsin(2n−1)πx2L𝑆𝑀(𝑥)=∑𝑛=1∞𝑛sin⁡(2𝑛−1)𝑥2𝐿
is the mixed Fourier sine series of f𝑓 on [0,L];[0,𝐿]; that is,
αn=2L∫L0f(x)sin(2n−1)πx2Ldx.𝑛=2𝐿∫0𝐿𝑓(𝑥)sin⁡(2𝑛−1)𝑥2𝐿𝑑𝑥.
Example 12.1.3
Solve Equation 12.1.712.1.7 with f(x)=x𝑓(𝑥)=𝑥.
Solution
From Example 11.3.4, the mixed Fourier sine series of f𝑓 on [0,L][0,𝐿] is
SM(x)=−8Lπ2∑n=1∞(−1)n(2n−1)2sin(2n−1)πx2L.𝑆𝑀(𝑥)=−8𝐿2∑𝑛=1∞(
−1)𝑛(2𝑛−1)2sin⁡(2𝑛−1)𝑥2𝐿.
Therefore
u(x,t)=−8Lπ2∑n=1∞(−1)n(2n−1)2e−(2n−1)2π2a2t/
4L2sin(2n−1)πx2L.𝑢(𝑥,𝑡)=−8𝐿2∑𝑛=1∞(−1)𝑛(2𝑛−1)2𝑒−(2𝑛−1)2𝜋2𝑎2𝑡/
4𝐿2sin⁡(2𝑛−1)𝑥2𝐿.
Figure 12.1.2 shows a graph of u=u(x,t)𝑢=𝑢(𝑥,𝑡) plotted with respect
to x𝑥 for various values of t𝑡. The line y=x𝑦=𝑥 corresponds to t=0𝑡=0. The
other curves correspond to positive values of t𝑡. As t𝑡 increases, the graphs
approach the line u=0𝑢=0.
 Figure
12.1.2
We leave it to you (Exercise 12.1.3) to use the method of separation of
variables and Theorem 11.1.5 to motivate the next definition.
Definition 12.1.5
The formal solution of the initial-boundary value problem
ut=a2uxx,0<x<L,t>0,ux(0,t)=0,u(L,t)=0,t>0,u(x,0)=f(x),0≤x≤L(12.1.8)
(12.1.8)𝑢𝑡=𝑎2𝑢𝑥𝑥,0<𝑥<𝐿,𝑡>0,𝑢𝑥(0,𝑡)=0,𝑢(𝐿,𝑡)=0,𝑡>0,𝑢(𝑥,0)=𝑓(𝑥),0≤𝑥≤
𝐿
is
u(x,t)=∑n=1∞αne−(2n−1)2π2a2t/
4L2cos(2n−1)πx2L,𝑢(𝑥,𝑡)=∑𝑛=1∞𝑛𝑒−(2𝑛−1)2𝜋2𝑎2𝑡/
4𝐿2cos⁡(2𝑛−1)𝑥2𝐿,
where
CM(x)=∑n=1∞αncos(2n−1)πx2L𝐶𝑀(𝑥)=∑𝑛=1∞𝑛cos⁡(2𝑛−1)𝑥2𝐿
is the mixed Fourier cosine series of f𝑓 on [0,L][0,𝐿]; that is,
αn=2L∫L0f(x)cos(2n−1)πx2Ldx.𝑛=2𝐿∫0𝐿𝑓(𝑥)cos⁡(2𝑛−1)𝑥2𝐿𝑑𝑥.
Example 12.1.4
Solve Equation 12.1.812.1.8 with f(x)=x−L𝑓(𝑥)=𝑥−𝐿.
Solution
From Example 11.3.3, the mixed Fourier cosine series of f𝑓 on [0,L]
[0,𝐿] is
CM(x)=−8Lπ2∑n=1∞1(2n−1)2cos(2n−1)πx2L.𝐶𝑀(𝑥)=−8𝐿2∑𝑛=1∞1(2𝑛
−1)2cos⁡(2𝑛−1)𝑥2𝐿.
Therefore
u(x,t)=−8Lπ2∑n=1∞1(2n−1)2e−(2n−1)2π2a2t/
4L2cos(2n−1)πx2L.𝑢(𝑥,𝑡)=−8𝐿2∑𝑛=1∞1(2𝑛−1)2𝑒−(2𝑛−1)2𝜋2𝑎2𝑡/
4𝐿2cos⁡(2𝑛−1)𝑥2𝐿.
Nonhomogeneous Problems
A problem of the form
ut=a2uxx+h(x),0<x<L,t>0,u(0,t)=u0,u(L,t)=uL,t>0,u(x,0)=f(x),0≤x≤L(12.
1.9)
(12.1.9)𝑢𝑡=𝑎2𝑢𝑥𝑥+ℎ(𝑥),0<𝑥<𝐿,𝑡>0,𝑢(0,𝑡)=𝑢0,𝑢(𝐿,𝑡)=𝑢𝐿,𝑡>0,𝑢(𝑥,0)=𝑓(
𝑥),0≤𝑥≤𝐿
can be transformed to a problem that can be solved by separation of
variables. We write
u(x,t)=v(x,t)+q(x),(12.1.10)(12.1.10)𝑢(𝑥,𝑡)=𝑣(𝑥,𝑡)+𝑞(𝑥),
where q𝑞 is to be determined. Then
ut=vtanduxx=vxx+q′′𝑢𝑡=𝑣𝑡and𝑢𝑥𝑥=𝑣𝑥𝑥+𝑞″
so u𝑢 satisfies Equation 12.1.912.1.9 if v𝑣 satisfies
vt=a2vxx+a2q′′(x)
+h(x),0<x<L,t>0,v(0,t)=u0−q(0),v(L,t)=uL−q(L),t>0,v(x,0)=f(x)
−q(x),0≤x≤L.𝑣𝑡=𝑎2𝑣𝑥𝑥+𝑎2𝑞″(𝑥)
+ℎ(𝑥),0<𝑥<𝐿,𝑡>0,𝑣(0,𝑡)=𝑢0−𝑞(0),𝑣(𝐿,𝑡)=𝑢𝐿−𝑞(𝐿),𝑡>0,𝑣(𝑥,0)=𝑓(𝑥)
−𝑞(𝑥),0≤𝑥≤𝐿.
This reduces to
vt=a2vxx,0<x<L,t>0,v(0,t)=0,v(L,t)=0,t>0,v(x,0)=f(x)
−q(x),0≤x≤L(12.1.11)
(12.1.11)𝑣𝑡=𝑎2𝑣𝑥𝑥,0<𝑥<𝐿,𝑡>0,𝑣(0,𝑡)=0,𝑣(𝐿,𝑡)=0,𝑡>0,𝑣(𝑥,0)=𝑓(𝑥)
−𝑞(𝑥),0≤𝑥≤𝐿
if
a2q′′+h(x)=0,q(0)=u0,q(L)=uL.𝑎2𝑞″+ℎ(𝑥)=0,𝑞(0)=𝑢0,𝑞(𝐿)=𝑢𝐿.
We can obtain q𝑞 by integrating q′′=−h/a2𝑞″=−ℎ/𝑎2 twice and choosing
the constants of integration so
that q(0)=u0𝑞(0)=𝑢0 and q(L)=uL𝑞(𝐿)=𝑢𝐿. Then we can solve
Equation 12.1.1112.1.11 for v𝑣 by separation of variables, and
Equation 12.1.1012.1.10 is the solution of Equation 12.1.912.1.9.
Example 12.1.5
Solve
ut=uxx−2,0<x<1,t>0,u(0,t)=−1,u(1,t)=1,t>0,u(x,0)=x3−2x2+3x−1,0≤x≤1.
𝑢𝑡=𝑢𝑥𝑥−2,0<𝑥<1,𝑡>0,𝑢(0,𝑡)=−1,𝑢(1,𝑡)=1,𝑡>0,𝑢(𝑥,0)=𝑥3−2𝑥2+3𝑥−1,0≤
𝑥≤1.
Solution
We leave it to you to show that
q(x)=x2+x−1𝑞(𝑥)=𝑥2+𝑥−1
satisfies
q′′−2=0,q(0)=−1,q(1)=1.𝑞″−2=0,𝑞(0)=−1,𝑞(1)=1.
Therefore
u(x,t)=v(x,t)+x2+x−1,𝑢(𝑥,𝑡)=𝑣(𝑥,𝑡)+𝑥2+𝑥−1,
where
vt=vxx,0<x<1,t>0,𝑣𝑡=𝑣𝑥𝑥,0<𝑥<1,𝑡>0,
v(0,t)=0,v(1,t)=0,t>0,𝑣(0,𝑡)=0,𝑣(1,𝑡)=0,𝑡>0,
and
v(x,0)=x3−2x2+3x−1−x2−x+1=x(x2−3x+2).𝑣(𝑥,0)=𝑥3−2𝑥2+3𝑥−1−𝑥2−𝑥
+1=𝑥(𝑥2−3𝑥+2).
From Example 12.1.1 with a=1𝑎=1 and L=1𝐿=1,
v(x,t)=12π3∑n=1∞1n3e−n2π2tsinnπx.𝑣(𝑥,𝑡)=12𝜋3∑𝑛=1∞1𝑛3𝑒−𝑛2𝜋2𝑡s
in⁡𝑛𝑥.
Therefore
u(x,t)=x2+x−1+12π3∑n=1∞1n3e−n2π2tsinnπx.𝑢(𝑥,𝑡)=𝑥2+𝑥−1+12𝜋3∑𝑛
=1∞1𝑛3𝑒−𝑛2𝜋2𝑡sin⁡𝑛𝑥.
A similar procedure works if the boundary conditions in
Equation 12.1.1112.1.11 are replaced by mixed boundary conditions
ux(0,t)=u0,u(L,t)=uL,t>0𝑢𝑥(0,𝑡)=𝑢0,𝑢(𝐿,𝑡)=𝑢𝐿,𝑡>0
or
u(0,t)=u0,ux(L,t)=uL,t>0;𝑢(0,𝑡)=𝑢0,𝑢𝑥(𝐿,𝑡)=𝑢𝐿,𝑡>0;
however, this isn’t true in general for the boundary conditions
ux(0,t)=u0,ux(L,t)=uL,t>0.𝑢𝑥(0,𝑡)=𝑢0,𝑢𝑥(𝐿,𝑡)=𝑢𝐿,𝑡>0.
(See Exercise 12.1.47.)
Using Technology
Numerical experiments can enhance your understanding of the solutions of
initial-boundary value problems. To be specific, consider the formal
solution
u(x,t)=∑n=1∞αne−n2π2a2t/
L2sinnπxL,𝑢(𝑥,𝑡)=∑𝑛=1∞𝑛𝑒−𝑛2𝜋2𝑎2𝑡/𝐿2sin⁡𝑛𝑥𝐿,
of Equation 12.1.412.1.4, where
S(x)=∑n=1∞αnsinnπxL𝑆(𝑥)=∑𝑛=1∞𝑛sin⁡𝑛𝑥𝐿
is the Fourier sine series of f𝑓 on [0,L][0,𝐿]. Consider the m𝑚-th partial
sum
um(x,t)=∑n=1mαne−n2π2a2t/L2sinnπxL.(12.1.12)
(12.1.12)𝑢𝑚(𝑥,𝑡)=∑𝑛=1𝑚𝑛𝑒−𝑛2𝜋2𝑎2𝑡/𝐿2sin⁡𝑛𝑥𝐿.
For several fixed values of t𝑡 (including t=0𝑡=0),
graph um(x,t)𝑢𝑚(𝑥,𝑡) versus t𝑡. In some cases it may be useful to graph
the curves corresponding to the various values of t𝑡 on the same axes in
other cases you may want to graph the various curves successively (for
increasing values of t𝑡), and create a primitive motion picture on your
monitor. Repeat this experiment for several values of m𝑚, to compare
how the results depend upon m𝑚 for small and large values of t𝑡.
However, keep in mind that the meanings of “small” and “large” in this
case depend upon the constants a2𝑎2 and L2𝐿2. A good way to handle this
is to rewrite Equation 12.1.1212.1.12 as
um(x,t)=∑n=1mαne−n2τsinnπxL,𝑢𝑚(𝑥,𝑡)=∑𝑛=1𝑚𝑛𝑒−𝑛2𝜏sin⁡𝑛𝑥𝐿,
where
τ=π2a2tL2,(12.1.13)(12.1.13)𝜏=𝜋2𝑎2𝑡𝐿2,
and graph um𝑢𝑚 versus x𝑥 for selected values of τ𝜏.
These comments also apply to the situations considered in Definitions
12.1.3 -12.1.5 , except that Equation 12.1.1312.1.13 should be replaced by
τ=π2a2t4L2,𝜏=𝜋2𝑎2𝑡4𝐿2,
in Definitions 12.1.4 and 12.1.5 .

This page tit is shared under a CC BY-NC-SA 3.0 license and was
authored, remixed, and/or curated by William F. Trench via source
content that was edited to the style and standards of the LibreTexts
platform; a detailed edit history is available upon request.
 A family of energy curves in the phase plane
for 12θ˙2−ω2cosθ=z12𝜃˙2−𝜔2cos⁡𝜃=𝑧. Here we
took ω=1.0𝜔=1.0 and z∈[−5,15]𝑧∈[−5,15].
Wave Equation
Imagine we have a tensioned guitar string of length L𝐿. Suppose we only
consider vibrations in one direction. That is, let x𝑥 denote the position
along the string, let t𝑡 denote time, and let y𝑦 denote the displacement of
the string from the rest position. See Figure 4.7.14.7.1.

figure 4.7.14.7.1: Vibrating string of length L𝐿, x𝑥 is position, y𝑦 is

displacement.
The equation that governs this setup is the so-called one-dimensional wave
equation:
ytt=a2yxx,𝑦𝑡𝑡=𝑎2𝑦𝑥𝑥,
for some constant a>0𝑎>0. The intuition is similar to the heat equation,
replacing velocity with acceleration: the acceleration at a specific point is
proportional to the second derivative of theshape of the string. In other
words when the string is concave down then uxx𝑢𝑥𝑥 is negative and the
string wants to accelerate downwards, so utt𝑢𝑡𝑡 should be negative. And
vice versa. The wave equation is an example of a hyperbolic PDE.
Assume that the ends of the string are fixed in place:
y(0,t)=0andy(L,t)=0.𝑦(0,𝑡)=0and𝑦(𝐿,𝑡)=0.
Note that we have two conditions along the x𝑥 axis as there are two
derivatives in the x𝑥 direction.
There are also two derivatives along the t𝑡 direction and hence we need
two further conditions here. We need to know the initial position and the
initial velocity of the string. That is, for some known
functions f(x)𝑓(𝑥) and g(x)𝑔(𝑥), we impose
y(x,0)=f(x)andyt(x,0)=g(x).𝑦(𝑥,0)=𝑓(𝑥)and𝑦𝑡(𝑥,0)=𝑔(𝑥).
As the equation is again linear, superposition works just as it did for the
heat equation. And again we will use separation of variables to find
enough building-block solutions to get the overall solution. There is one
change however. It will be easier to solve two separate problems and add
their solutions.
The two problems we will solve are
wtt=a2wxx,w(0,t)=w(L,t)=0,w(x,0)=0wt(x,0)=g(x)for 0<x<L,for 0<x<L,
(4.7.1)(4.7.1)𝑤𝑡𝑡=𝑎2𝑤𝑥𝑥,𝑤(0,𝑡)=𝑤(𝐿,𝑡)=0,𝑤(𝑥,0)=0for 0
<𝑥<𝐿,𝑤𝑡(𝑥,0)=𝑔(𝑥)for 0<𝑥<𝐿,
and
ztt=a2zxx,z(0,t)=z(L,t)=0,z(x,0)=f(x)zt(x,0)=0for 0<x<L,for 0<x<L.(4.7.2)
(4.7.2)𝑧𝑡𝑡=𝑎2𝑧𝑥𝑥,𝑧(0,𝑡)=𝑧(𝐿,𝑡)=0,𝑧(𝑥,0)=𝑓(𝑥)for 0<𝑥<𝐿,𝑧𝑡(𝑥,0)=0for 0<
𝑥<𝐿.
The principle of superposition implies that y=w+z𝑦=𝑤+𝑧 solves the wave
equation and
furthermore y(x,0)=w(x,0)+z(x,0)=f(x)𝑦(𝑥,0)=𝑤(𝑥,0)+𝑧(𝑥,0)=𝑓(𝑥) and yt(
x,0)=wt(x,0)+zt(x,0)=g(x)𝑦𝑡(𝑥,0)=𝑤𝑡(𝑥,0)+𝑧𝑡(𝑥,0)=𝑔(𝑥). Hence, y𝑦 is a
solution to
ytt=a2yxx,y(0,t)=y(L,t)=0,y(x,0)=f(x)yt(x,0)=g(x)for 0<x<L,for 0<x<L.
(4.7.3)(4.7.3)𝑦𝑡𝑡=𝑎2𝑦𝑥𝑥,𝑦(0,𝑡)=𝑦(𝐿,𝑡)=0,𝑦(𝑥,0)=𝑓(𝑥)for 0
<𝑥<𝐿,𝑦𝑡(𝑥,0)=𝑔(𝑥)for 0<𝑥<𝐿.
The reason for all this complexity is that superposition only works for
homogeneous conditions such
as y(0,t)=y(L,t)=0𝑦(0,𝑡)=𝑦(𝐿,𝑡)=0, y(x,0)=0𝑦(𝑥,0)=0,
or yt(x,0)=0𝑦𝑡(𝑥,0)=0. Therefore, we will be able to use the idea of
separation of variables to find many building-block solutions solving all
the homogeneous conditions. We can then use them to construct a solution
solving the remaining nonhomogeneous condition.
Let us start with (4.7.1)(4.7.1). We try a solution of the
form w(x,t)=X(x)T(t)𝑤(𝑥,𝑡)=𝑋(𝑥)𝑇(𝑡) again. We plug into the wave
equation to obtain
X(x)T′′(t)=a2X′′(x)T(t).𝑋(𝑥)𝑇″(𝑡)=𝑎2𝑋″(𝑥)𝑇(𝑡).
Rewriting we get
T′′(t)a2T(t)=X′′(x)X(x).𝑇″(𝑡)𝑎2𝑇(𝑡)=𝑋″(𝑥)𝑋(𝑥).
Again, left hand side depends only on t𝑡 and the right hand side depends
only on x𝑥. Therefore, both equal a constant, which we will denote
by −λ−𝜆.
T′′(t)a2T(t)=−λ=X′′(x)X(x).𝑇″(𝑡)𝑎2𝑇(𝑡)=−𝜆=𝑋″(𝑥)𝑋(𝑥).
We solve to get two ordinary differential equations
X′′(x)+λX(x)T′′(t)+λa2T(t)=0,=0.(4.7.4)(4.7.4)𝑋″(𝑥)+𝜆𝑋(𝑥)=0,𝑇″(𝑡)
+𝜆𝑎2𝑇(𝑡)=0.
The
conditions 0=w(0,t)=X(0)T(t)0=𝑤(0,𝑡)=𝑋(0)𝑇(𝑡) implies X(0)=0𝑋(0)=0 an
d w(L,t)=0𝑤(𝐿,𝑡)=0 implies that X(L)=0𝑋(𝐿)=0. Therefore, the only
nontrivial solutions for the first equation are
when λ=λn=n2π2L2𝜆=𝜆𝑛=𝑛2𝜋2𝐿2 and they are
Xn(x)=sin(nπLx).𝑋𝑛(𝑥)=sin⁡(𝑛𝜋𝐿𝑥).
The general solution for T𝑇 for this particular λn𝜆𝑛 is
Tn(t)=Acos(nπaLt)+Bsin(nπaLt).𝑇𝑛(𝑡)=𝐴cos⁡(𝑛𝜋𝑎𝐿𝑡)+𝐵sin⁡(𝑛𝜋𝑎𝐿𝑡).
Again, left hand side depends only on t𝑡 and the right hand side depends
only on x𝑥. Therefore, both equal a constant, which we will denote
by −λ−𝜆.
T′′(t)a2T(t)=−λ=X′′(x)X(x).𝑇″(𝑡)𝑎2𝑇(𝑡)=−𝜆=𝑋″(𝑥)𝑋(𝑥).
We solve to get two ordinary differential equations
X′′(x)+λX(x)T′′(t)+λa2T(t)=0,=0.(4.7.4)(4.7.4)𝑋″(𝑥)+𝜆𝑋(𝑥)=0,𝑇″(𝑡)
+𝜆𝑎2𝑇(𝑡)=0.
The
conditions 0=w(0,t)=X(0)T(t)0=𝑤(0,𝑡)=𝑋(0)𝑇(𝑡) implies X(0)=0𝑋(0)=0 an
d w(L,t)=0𝑤(𝐿,𝑡)=0 implies that X(L)=0𝑋(𝐿)=0. Therefore, the only
nontrivial solutions for the first equation are
when λ=λn=n2π2L2𝜆=𝜆𝑛=𝑛2𝜋2𝐿2 and they are
Xn(x)=sin(nπLx).𝑋𝑛(𝑥)=sin⁡(𝑛𝜋𝐿𝑥).
The general solution for T𝑇 for this particular λn𝜆𝑛 is
Tn(t)=Acos(nπaLt)+Bsin(nπaLt).𝑇𝑛(𝑡)=𝐴cos⁡(𝑛𝜋𝑎𝐿𝑡)+𝐵sin⁡(𝑛𝜋𝑎𝐿𝑡).
We also have the condition
that w(x,0)=0𝑤(𝑥,0)=0 or X(x)T(0)=0𝑋(𝑥)𝑇(0)=0. This implies
that T(0)=0𝑇(0)=0, which in turn forces A=0𝐴=0. It is convenient to
pick B=Lnπa𝐵=𝐿𝑛𝜋𝑎 (you will see why in a moment) and hence
Tn(t)=Lnπasin(nπaLt).𝑇𝑛(𝑡)=𝐿𝑛𝜋𝑎sin⁡(𝑛𝜋𝑎𝐿𝑡).
Our building-block solutions are
wn(x,t)=Lnπasin(nπLx)sin(nπaLt).𝑤𝑛(𝑥,𝑡)=𝐿𝑛𝜋𝑎sin⁡(𝑛𝜋𝐿𝑥)sin⁡(𝑛𝜋𝑎𝐿𝑡).
We differentiate in t𝑡, that is
∂wn∂t(x,t)=sin(nπLx)cos(nπaLt).∂𝑤𝑛∂𝑡(𝑥,𝑡)=sin⁡(𝑛𝜋𝐿𝑥)cos⁡(𝑛𝜋𝑎𝐿𝑡).
Hence,
∂wn∂t(x,0)=sin(nπLx).∂𝑤𝑛∂𝑡(𝑥,0)=sin⁡(𝑛𝜋𝐿𝑥).
We expand g(x)𝑔(𝑥) in terms of these sines as
g(x)=∑n=1∞bnsin(nπLx).𝑔(𝑥)=∑𝑛=1∞𝑏𝑛sin⁡(𝑛𝜋𝐿𝑥).
Using superposition we can just write down the solution to (4.7.1)
(4.7.1) as a series
w(x,t)=∑n=1∞bnwn(x,t)=∑n=1∞bnLnπasin(nπLx)sin(nπaLt).𝑤(𝑥,𝑡)=∑𝑛=
1∞𝑏𝑛𝑤𝑛(𝑥,𝑡)=∑𝑛=1∞𝑏𝑛𝐿𝑛𝜋𝑎sin⁡(𝑛𝜋𝐿𝑥)sin⁡(𝑛𝜋𝑎𝐿𝑡).
Check that w(x,0)=0𝑤(𝑥,0)=0 and wt(x,0)=g(x)𝑤𝑡(𝑥,0)=𝑔(𝑥).
Similarly we proceed to solve (4.7.2)(4.7.2). We again
try z(x,y)=X(x)T(t)𝑧(𝑥,𝑦)=𝑋(𝑥)𝑇(𝑡). The procedure works exactly the
same at first. We obtain
X′′(x)+λX(x)T′′(t)+λa2T(t)=0,=0.(4.7.5)(4.7.5)𝑋″(𝑥)+𝜆𝑋(𝑥)=0,𝑇″(𝑡)
+𝜆𝑎2𝑇(𝑡)=0.
and the conditions X(0)=0𝑋(0)=0, X(L)=0𝑋(𝐿)=0. So
again λ=λn=n2π2L2𝜆=𝜆𝑛=𝑛2𝜋2𝐿2 and
Xn(x)=sin(nπLx).𝑋𝑛(𝑥)=sin⁡(𝑛𝜋𝐿𝑥).
This time the condition on T𝑇 is T′(0)=0𝑇′(0)=0. Thus we get
that B=0𝐵=0 and we take
Tn(t)=cos(nπaLt).𝑇𝑛(𝑡)=cos⁡(𝑛𝜋𝑎𝐿𝑡).
Our building-block solution will be
zn(x,t)=sin(nπLx)cos(nπaLt).𝑧𝑛(𝑥,𝑡)=sin⁡(𝑛𝜋𝐿𝑥)cos⁡(𝑛𝜋𝑎𝐿𝑡).
As zn(x,0)=sin(nπLx)𝑧𝑛(𝑥,0)=sin⁡(𝑛𝜋𝐿𝑥), we expand f(x)𝑓(𝑥) in terms of
these sines as
f(x)=∑n=1∞cnsin(nπLx).𝑓(𝑥)=∑𝑛=1∞𝑐𝑛sin⁡(𝑛𝜋𝐿𝑥).
And we write down the solution to (4.7.2)(4.7.2) as a series
z(x,t)=∑n=1∞cnzn(x,t)=∑n=1∞cnsin(nπLx)cos(nπaLt).𝑧(𝑥,𝑡)=∑𝑛=1∞𝑐𝑛𝑧
𝑛(𝑥,𝑡)=∑𝑛=1∞𝑐𝑛sin⁡(𝑛𝜋𝐿𝑥)cos⁡(𝑛𝜋𝑎𝐿𝑡).
Exercise 4.7.24.7.2
Fill in the details in the derivation of the solution of (4.7.2)(4.7.2). Check
that the solution satisfies all the side conditions.
Putting these two solutions together, let us state the result as a theorem.
Theorem 4.7.14.7.1
Take the equation
ytt=a2yxx,y(0,t)=y(L,t)=0,y(x,0)=f(x)yt(x,0)=g(x)for 0<x<L,for 0<x<L,
(4.7.6)(4.7.6)𝑦𝑡𝑡=𝑎2𝑦𝑥𝑥,𝑦(0,𝑡)=𝑦(𝐿,𝑡)=0,𝑦(𝑥,0)=𝑓(𝑥)for 0
<𝑥<𝐿,𝑦𝑡(𝑥,0)=𝑔(𝑥)for 0<𝑥<𝐿,
where
f(x)=∑n=1∞cnsin(nπLx),𝑓(𝑥)=∑𝑛=1∞𝑐𝑛sin⁡(𝑛𝜋𝐿𝑥),
and
 g(x)=∑n=1∞bnsin(nπLx).𝑔(𝑥)=∑𝑛=1∞𝑏𝑛sin⁡(𝑛𝜋𝐿𝑥).
Then the solution y(x,t)𝑦(𝑥,𝑡) can be written as a sum of the solutions of
(4.7.4) and (4.7.5). In other words,
y(x,t)=∑n=1∞bnLnπasin(nπLx)sin(nπaLt)
+cnsin(nπLx)cos(nπaLt)=∑n=1∞sin(nπLx)[bnLnπasin(nπaLt)
+cncos(nπaLt)].(4.7.7)(4.7.7)𝑦(𝑥,𝑡)=∑𝑛=1∞𝑏𝑛𝐿𝑛𝜋𝑎sin⁡(𝑛𝜋𝐿𝑥)sin⁡(𝑛𝜋𝑎𝐿𝑡)
+𝑐𝑛sin⁡(𝑛𝜋𝐿𝑥)cos⁡(𝑛𝜋𝑎𝐿𝑡)=∑𝑛=1∞sin⁡(𝑛𝜋𝐿𝑥)[𝑏𝑛𝐿𝑛𝜋𝑎sin⁡(𝑛𝜋𝑎𝐿𝑡)
+𝑐𝑛cos⁡(𝑛𝜋𝑎𝐿𝑡)].
Example 4.7.14.7.1
Let us try a simple example of a plucked string. Suppose that a string of
length 2 is plucked in the middle such that it has the initial shape given in
Figure 4.7.24.7.2. That is
f(x)={0.1x0.1(2−x)if 0≤x≤1,if 1<x≤2.𝑓(𝑥)={0.1𝑥if 0≤𝑥≤1,0.1(2−𝑥)if 1<𝑥≤
2.
Check that w(x,0)=0𝑤(𝑥,0)=0 and wt(x,0)=g(x)𝑤𝑡(𝑥,0)=𝑔(𝑥).

Fi
gure 4.7.24.7.2: Initial shape of a plucked string from Example 4.7.14.7.1.
The string starts at rest (g(x)=0𝑔(𝑥)=0). Suppose that a=1𝑎=1 in the wave
equation for simplicity. In other words, we wish to solve the problem:
ytt=yxx,y(0,t)=y(2,t)=0,y(x,0)=f(x)andyt(x,0)=0.(4.7.8)
(4.7.8)𝑦𝑡𝑡=𝑦𝑥𝑥,𝑦(0,𝑡)=𝑦(2,𝑡)=0,𝑦(𝑥,0)=𝑓(𝑥)and𝑦𝑡(𝑥,0)=0.
We leave it to the reader to compute the sine series of f(x)𝑓(𝑥). The series
will be
f(x)=∑n=1∞0.8n2π2sin(nπ2)sin(nπ2x).𝑓(𝑥)=∑𝑛=1∞0.8𝑛2𝜋2sin⁡(𝑛𝜋2)sin⁡(𝑛
𝜋2𝑥).
Note that sin(nπ2)sin⁡(𝑛𝜋2) is the sequence 1,0,−1,0,1,0,−1,…
1,0,−1,0,1,0,−1,… for n=1,2,3,4,…𝑛=1,2,3,4,…. Therefore,
f(x)=0.8π2sin(π2x)−0.89π2sin(3π2x)+0.825π2sin(5π2x)−
⋯𝑓(𝑥)=0.8𝜋2sin⁡(𝜋2𝑥)−0.89𝜋2sin⁡(3𝜋2𝑥)+0.825𝜋2sin⁡(5𝜋2𝑥)−⋯
The solution y(x,t)𝑦(𝑥,𝑡) is given by
y(x,t)=∑n=1∞0.8n2π2sin(nπ2)sin(nπ2x)cos(nπ2t)=∑m=1∞0.8(−1)m+1(2
m−1)2π2sin((2m−1)π2x)cos((2m−1)π2t)=0.8π2sin(π2x)cos(π2t)
 −0.89π2sin(3π2x)cos(3π2t)+0.825π2sin(5π2x)cos(5π2t)−⋯(4.7.9)
(4.7.9)𝑦(𝑥,𝑡)=∑𝑛=1∞0.8𝑛2𝜋2sin⁡(𝑛𝜋2)sin⁡(𝑛𝜋2𝑥)cos⁡(𝑛𝜋2𝑡)=∑𝑚=1∞0.8(−
1)𝑚+1(2𝑚−1)2𝜋2sin⁡((2𝑚−1)𝜋2𝑥)cos⁡((2𝑚−1)𝜋2𝑡)=0.8𝜋2sin⁡(𝜋2𝑥)cos⁡(𝜋2𝑡
)−0.89𝜋2sin⁡(3𝜋2𝑥)cos⁡(3𝜋2𝑡)+0.825𝜋2sin⁡(5𝜋2𝑥)cos⁡(5𝜋2𝑡)−⋯
See Figure 4.7.34.7.3 for a plot for 0<t<30<𝑡<3. Notice that unlike the heat
equation, the solution does not become "smoother," the "sharp edges"
remain. We will see the reason for this behavior in the next section where
we derive the solution to the wave equation in a different way.

Figure 4.7.34.7.3: Shape of the plucked string for 0<t<30<𝑡<3.

Make sure you understand what the plot, such as the one in the figure, is
telling you. For each fixed t𝑡, you can think of the function y(x,t)𝑦(𝑥,𝑡) as
just a function of x𝑥. This function gives
you the shape of the string at time t𝑡. See Figure 4.7.44.7.4 for plots of
at y𝑦 as a function of x𝑥 at several different values of t𝑡. On this plot you
can see the sharp edges remaining much better.
 Fi
gure 4.7.44.7.4: Plucked string for t=0𝑡=0, t=0.4𝑡=0.4, t=0.8𝑡=0.8,
and t=1.2𝑡=1.2.
One thing to take away from all this is how a guitar sounds. Notice that the
(angular) frequencies that come up in the solution are nπaL𝑛𝜋𝑎𝐿. That is,
there is a certain base fundamental frequency πaL𝜋𝑎𝐿, and then we also
get all the multiples of this frequency, which in music are called
the overtones. Which overtones appear and with what amplitude is what
musicians call the timbre of the note. Mathematicians usually call this
the spectrum. Because all the frequencies are multiples of one frequency
(the fundamental) we get a nice pleasing sound.
The fundamental frequency πaL𝜋𝑎𝐿 increases as we decrease length L𝐿.
That is, if we place a finger on the fingerboard and then pluck a string we
get a higher note. The constant a𝑎 is given by
a=Tρ−−√,𝑎=𝑇𝜌,
where T𝑇 is tension and ρ𝜌 is the linear density of the string. Tightening
the string (turning the tuning peg on a guitar) increases a𝑎 and hence
produces a higher fundamental frequency (a higher note). On the other
hand using a heavier string reduces a𝑎 and produces a lower fundamental
frequency (a lower note). A bass guitar has longer thicker strings, while a
ukulele has short strings made of lighter material.
Something rather interesting is the almost symmetry between space and
time. In its simplest form we see this symmetry in the solutions
sin(nπLx)sin(nπaLt).sin⁡(𝑛𝜋𝐿𝑥)sin⁡(𝑛𝜋𝑎𝐿𝑡).
Except for the a𝑎, time and space are just the same.
In general, the solution for a fixed x𝑥 is a Fourier series in t𝑡, for a
fixed t𝑡 it is a Fourier series in x𝑥, and the coefficients are related. If the
shape f(x)𝑓(𝑥) or the initial velocity have lots of corners, then the sound
wave will have lots of corners. That is because the Fourier coefficients of
the initial shape decay to zero (as n→∞𝑛→∞) at the same rate as the
Fourier coefficients of the wave in time (for some fixed x𝑥). So if you use
a sharp object to pick the string, you get a sharper sound with lots of high
frequency components, while if you use your thumb, you get a softer
sound without so many high overtones. Similarly if you pluck close to the
bridge, you are getting a pluck that looks more like the sawtooth, and you
get an even sharper sound.
In fact, if you look at the formula for the solution, you see that for any
fixed x𝑥 we get an almost arbitrary Fourier series in t𝑡, everything except
the constant term. You can essentially obtain any
sound you want by plucking the string in just the right way. Of course we
are considering an ideal string of no stiffness and no air resistance. Those
variables clearly impact the sound as well.
We next investigate the stability of the equilibrium solutions of the
nonlinear pendulum which we first encountered in section 2.3.2 . Along
the way we will develop some basic methods for studying the stability of
equilibria in nonlinear systems in general.
Recall that the derivation of the pendulum equation was based upon a
simple point mass m𝑚 hanging on a string of length L𝐿 from some
support as shown in Figure 7.5.17.5.1. One pulls the mass back to some
starting angle, θ0𝜃0, and releases it. The goal is to find the angular
position as a function of time, θ(t)𝜃(𝑡).
In Chapter 2 we derived the nonlinear pendulum equation,
Lθ¨+gsinθ=0𝐿¨+𝑔sin⁡𝜃=0
There are several variations of Equation 7.5.17.5.1 which we have used in
this text. The first one is the linear pendulum, which was obtained using a
small angle approximation,
Lθ¨+gθ=0𝐿¨+𝑔=0