You are on page 1of 7

MIDTERM EXAM II FOR MATH 423, SPRING 2017

Name: ID Number: –00 −

Reminder: Show your work. Make your presentation comprehensible. Only correct, relevant
work will be given partial credits. Erase or cross out parts that you don’t want us to grade.

Problem Maximum Points Your Points

1 20

2 20

3 20

4 20

5 20

Total 100

You may use the following intergrals/trigonometric identities without having to derive them.

π π
(−1)n+1 π (−1)n − 1
Z Z
x sin(nx) dx = , x cos(nx) dx = for integer n ≥ 1.
0 n 0 n2

2 cos A cos B = cos(A + B) + cos(A − B), 2 sin A cos B = sin(A + B) + sin(A − B).

1
(1) (20 points) Construct a solution to the initial value problem
2

 utt (x, t) − c uxx (x, t) = sin t, x ∈ R, t > 0,

u(x, 0) = sin x, x ∈ R,

ut (x, 0) = 0.

Solution. We construct a solution u(x, t) as v(x, t) + w(x, t), where


2

 vtt (x, t) − c vxx (x, t) = 0, x ∈ R, t > 0,

v(x, 0) = sin x, x ∈ R,

vt (x, 0) = 0,

and
2

 wtt (x, t) − c wxx (x, t) = sin t, x ∈ R, t > 0,

w(x, 0) = 0, x ∈ R,

wt (x, 0) = 0.

v(x, t) = 21 [sin(x − ct) + sin(x + ct)] according to d’Almbert’s formula, and by Duhamel’s
Rt
principle w(x, t) = 0 η(x, t − τ, τ )dτ , where
2

 ηtt (x, t, τ ) − c ηxx (x, t, τ ) = 0, x ∈ R, τ > 0,

η(x, 0, τ ) = 0, x ∈ R,

ηt (x, 0, τ ) = sin τ.

Since
Z x+c(t−τ )
1
η(x, t − τ, τ ) = sin(τ )dy
2c x−c(t−τ )
Z x+c(t−τ )
sin(τ )
= 1 dy
2c x−c(t−τ )
[x + c(t − τ )] − [x − c(t − τ )]
= sin(τ )
2c
= (t − τ ) sin(τ ), a common error is to confuse sin(τ ) into sin y,
we have
Z t
w(x, t) = (t − τ ) sin(τ )dτ
0
Z t
= −(t − τ ) cos(τ )|ττ =t
=0 − cos(τ )dτ
0
= t − sin(t).
1
Thus u(x, t) = 2 [sin(x − ct) + sin(x + ct)] + t − sin(t).


2
(2) (20 points) Given φ(x) = x for 0 < x < π. Calculate the full Fourier series of φ(x) on
(0, π),

a0 X  nπx nπx 
+ an cos( ) + bn sin( ) ,
2 l l
n=1
—you need to determine an appropriate choice for l, and sketch a plot on [−3π, 3π] for
the sum of the first three non-zero terms of its full Fourier series. (You may use the
integral formula provided on the cover page.)
Solution. The full Fourier series of φ(x) on (0, π) will have period equal to π, so
they would have the form

a0 X
+ (an cos(2nx) + bn sin(2nx)) ,
2
n=1
π
where we have chosen l = 2 , half the length of the interval (0, π), so that each term in
the above series has period π (the family {1, cos(2nx), sin(2nx)}∞ n=1 forms a complete
and orthogonal family on (0, π), while the family {1, cos(nx), sin(nx)}∞ n=1 does not form
an orthogonal family on (0, π)—it forms a complete and orthogonal family on (0, 2π)).
Then
2 π
Z
an = φ(x) cos(2nx) dx
π 0
Z π
2
= x cos(2nx) dx
π 0
 Z π 
1 x=π
= x sin(2nx) x=0 − sin(2nx) dx
nπ 0
1 x=π
= 2 cos(2nx) x=0
2n π
=0

for n 6= 0; while for n = 0, a0 = π2 0 x dx = π1 .
2 π
Z
bn = φ(x) sin(2nx) dx
π 0
2 π
Z
= x sin(2nx) dx
π 0
 Z π 
1 x=π
= −x cos(2nx) x=0 + cos(2nx) dx
nπ 0
1
=− .
n
So the full Fourier series of φ(x) on (0, π) is

1 X sin(2nx)
− .
2π n
n=1
1
The first three non-zero terms are 2π − sin(2x) − sin(4x)
2 . Its plot should provide an
approximation for the π-periodic extension of the given φ(x) on (0, π). Consult the
maple program posted on sakai for a plot. 

3
(3) (20 points) Solve the following boundary value problem for X(x) on [0, π].
X 00 (x) = −λX(x) for 0 < x < π, and X 0 (0) = 0, X(π) = 0,
In particular, verify that the set of eigenvalues is the set of numbers {λn = (n+ 21 )2 : n =
0, 1, 2, 3, · · · }, and for each eigenvalue λn = (n + 12 )2 , the eigenfunctions are non-zero
scalar multiples of cos (n + 12 )x .


00 (x) = −λX(x) is X(x) = A cos( λx)+
Solution.
√ For λ > 0, the general solution to X
B sin( λx) for some constants √ A and B. In order to satisfy X 0 (0) = 0, we must have
B = √ 0. Thus X(x) = A cos( λx). In order to satisfy√X(π) = 0, we must have
A
√ cos( λπ) = 0. Since we want A 6= 0, we must have cos( λπ) = 0, which imlies that
λ must be equal to n + 12 for some integer n. Thus λ = (n + 21 )2 for some integer
n. Since when n is a negative integer, (n + 21 )2 = (−1 − n + 12 )2 , and −1 − n would
be a positive integer, so we could limit n to be from 0, 1, 2, · · · in λ = (n + 12 )2 . The
corresponding solution for X(x) has been identified to be X(x) = A cos (n + 12 )x .
For λ = 0, the general solution to X 00 (x) = −λX(x) is X(x) = A + Bx. In order to
satisfy X 0 (0) = 0, we must haveB = 0. Thus X(x) = A. In order to satisfy X(π) = 0,
we must have A = 0. So we now have A = B = 0, which means that the only solution
is X ≡ 0 in this case, so λ = 0 is not an eigenvalue.
For λ < 0, we write λ = −β 2 for some real number β. Then the general solution to
X 00 (x)
= −λX(x) is X(x) = A cosh(βx) + B sinh(βx). In order to satisfy X 0 (0) = 0,
we must have B = 0. Thus X(x) = A cosh(βx). In order to satisfy X(π) = 0, we
must have A cosh(βπ) = 0. Since cosh(βπ) > 0, we must have A = 0. So we now have
A = B = 0, which means that the only solution is X ≡ 0 in this case, so λ = −β 2 is
not an eigenvalue. Similar analysis can rule out non-real eigenvalues: we may assume
λ = α + iβ, with α and β real, and analyze the solution forms to X 00 (x) = −λX(x),
together with the boundary conditions. This also follows from the argument on the
bottom of p.171 — note that the displayed formulae in the 3rd line from the bottom
Rb
are missing the factor (λ̄ − λ), and that the notation (y1 , y2 ) = a y1 (x)y2 (x)dx as
the inner product (over the interval (a, b)) between functions y1 and y2 is valid only
when both are real-valued fucntions; when they are not real-valued, the inner product
Rb
between them needs to be defined as (y1 , y2 ) = a y1 (x)y2 (x)dx instead, and it’s not
Rb
appropriate to call a y1 (x)y2 (x)dx the inner product between functions y1 and y2 in
such a case.


4
(4) (20 points) Look for separable, non-zero solutions of the form u(x, t) = X(x)T (t) to the
equation utt (x, t) − c2 uxx (x, t) = −rut (x, t) in 0 < x < π, together with the boundary
conditions u(0, t) = 0, ux (π, t) = 0. Show that the boundary value problem that X(x)
has to satisfy on [0, π] is
X 00 (x) = −λX(x) for 0 < x < π, and X(0) = 0, X 0 (π) = 0,
and also derive the equation that the factor T (t) has to satisfy.
Solution. Plugging ut = X(x)T 0 (t), utt = X(x)T 00 (t), and uxx = X 00 (x)T (t) into
the equation utt − c2 uxx = −rut , we obtain X(x)T 00 (t) − c2 X 00 (x)T (t) = −rX(x)T 0 (t).
00 (t) 00 (x) 0 (t)
At (x, t) such that X(x) 6= 0 and T (t) 6= 0, we must have TT (t) − c2 XX(x) = −r TT (t) .
T 00 (t) 0 00
Rewriting this equation as T (t) + r TT (t)
(t)
= c2 XX(x)
(x)
. Thus both sides must be equal to
X 00 (x)
a constant independent of x or t. Let X(x) = −λ. This gives rise to T 00 (t) + rT 0 (t) +
c2 λT (t) = 0.
To satisfy the condition u(0, t) = 0, we must have X(0)T (t) = 0 for all t > 0, which
imlies that X(0) = 0 in order for T (t) not ≡ 0. Similarly, to satisfy the condition
ux (π, t) = 0, we must have X 0 (π)T (t) = 0 for all t > 0, which imlies that X 0 (π) = 0. So
we conclude that X(x) has to satisfy X 00 (x) = −λX(x) for 0 < x < π, and X(0) = 0,
X 0 (π) = 0.
Note that the problem here didn’t ask to solve the equation or the boundary value
problem. 

5
(5) Consider the boundary-initial value problem
ut (x, t) − uxx (x, t) = 0 0 < x < 1, t > 0



u(1, t) = 1 t > 0



 ux (0, t) = 0 t > 0
πx



 u(x, 0) = 1 − cos( ) 0 < x < 1
2
(a) (5 points) Look for a solution of the form u(x, t) = 1+v(x, t), and reduce the prob-
lem to a boundary-initial value problem for v(x, t)—it should have homogenuous
boundary conditions.
Solution. Using the substitution u(x, t) = 1 + v(x, t), we will have ut (x, t) =
vt (x, t), ux (0, t) = vx (0, t), and uxx (x, t) = vxx (x, t). Therefore
vt (x, t) − vxx (x, t) = 0 0 < x < 1, t > 0



v(1, t) = 0 t > 0



 vx (0, t) = 0 t>0
πx



 v(x, 0) = − cos( ) 0<x<1
2


(b) (15 points) Solve the reduced boundary-initial value problem for v(x, t) from the
previous part, and use the solution v(x, t) to produce a solution to the original
boundary-initial value problem for u(x, t). Then determine limt→∞ u(x, t). (Re-
mark: The corresponding eigenvalue problem will end up to be directly related to
that in Problem 3; you may use the result there directly without further deriva-
tions.)
Solution. If we look for separable solutons v(x, t) = X(x)T (t) to

 vt (x, t) − vxx (x, t) = 0 0 < x < 1, t > 0

v(1, t) = 0 t > 0

vx (0, t) = 0 t > 0

we would find T 0 (t)X(x) − X 00 (x)T (t) = 0, and X 0 (1)T (t) = 0, X(0)T (t) = 0.
00 (x)
Thus XX(x) must be a constant, which we call −λ. Then we have
( 00
X (x) = − λX(x) 0 < x < 1
X(1) =0 X 0 (0) = 0.
T (t) satisfies T 0 (t) = −λT (t).
We note that cos( πx π 2
2 ) is a solution to the above eigenvalue problem with λ = ( 2 ) .
π 2
The corresponding solution for T 0 (t) = −( π2 )2 T (t) is T (t) = Ae−( 2 ) t for some
π 2
constant A. Thus v(x, t) = Ae−( 2 ) t cos( πx
2 ) is a solution to the reduced boundary-
initial value problem, if we choose A = −1. And the solution to the original
π 2
boundary-initial value problem is u(x, t) = 1 + e−( 2 ) t cos( πx 2 ), which → 1 as
t → ∞.
The solution v(x, t) can also be found by using solutions to the above eigenvalue
problem, which is identical to the eigenvalue problem in 1(a), except for the change
of scale from 0 < x < π to 0 < x < 1 here. A simple change of variables shows
that the eigenvalues are λ = (n + 21 )2 π 2 , and the corresponding eigenfunctions are

6
cos (n + 21 )πx . We can then solve for T (t) from T 0 (t) = −(n + 21 )2 π 2 T (t) to get

1 2 2
T (t) = An e−(n+ 2 ) π t . Finally, we can construct a series solution of the form
∞  
X
−(n+ 21 )2 π 2 t 1
An e cos (n + )πx ,
2
n=0
and use it to fit the initial condition v(x, 0) = − cos( πx
2 ), we will find that A0 = −1,
R1
and An = 0 for all n ≥ 1, as An = −2 0 cos( πx 1

2 ) cos (n + 2 )πx dx. 

You might also like