Byeongho Ban bban@umass.

edu

MATH 534H : Introduction to Partial Differential

Equations Homework Solutions
:Taught by Andrea Nahmod

Mathematics & Statistics

University of Massachusetts, Amherst
Byeong Ho Ban

1
MATH 534H Editor : Byeongho Ban

Due date : January 31st, 2018

1.1.2
Which of the following operators are linear?

(a) Lu = ux + xuy
(b) Lu = ux + uuy
(c) Lu = ux + u2y
(d) Lu = √
ux + uy + 1
(e) Lu = 1 + x2 (cos y)ux + uyxy − [arctan (x/y)]u
Proof. .

Let a constant a be given. And let the n times differentiable functions u and v be given.(for appropriate n with respect
to the problem.)
(a)
Observe
L(u + v) = (u + v)x + x(u + v)y = (u)x + (v)x + x[(u)y + (v)y ] = [ux + xuy ] + [vx + xvy ] = Lu + Lv,
L(au) = (au)x + x(au)y = aux + axuy = aLu.
Thus, L is a linear operator.

(b)
Observe, for nonzero constant c 6= 1,
L(cu) = (cu)x + (cu)(cu)y = cux + c2 uuy 6= cux + cuuy = cLu.
Thus, L is a nonlinear operator.

(c)
Observe, for a nonzero constant c 6= 1,
L(cu) = (cu)x + (cu)2y = cux + c2 u2y 6= cux + cu2y = cLu.
Thus, L is a nonlinear operator.

(d)
Observe, for a nonzero constant c 6= 1,
L(cu) = (cu)x + (cu)y + 1 = cux + cuy + 1 6= cux + cuy + c = cLu.
Thus, L is a nonlinear operator.

(e)
Observe
p
L(au) = 1 + x2 (cos y)(au)x + (au)yxy − [arctan (x/y)](au)
p
= a 1 + x2 (cos y)ux + auyxy − a[arctan (x/y)]u
= aLu
and
p
L(u + v) = = 1 + x2 (cos y)(u + v)x + (u + v)yxy − [arctan (x/y)](u + v)
p p
= 1 + x2 (cos y)ux + uyxy − [arctan (x/y)]u + 1 + x2 (cos y)vx + vyxy − [arctan (x/y)]v
= Lu + Lv.
Thus, L is a linear operator.

Page 2
MATH 534H Editor : Byeongho Ban

1.1.3
For each of the following equations, state the order and whether it is nonlinear, linear inhomogeneous, or linear
homogeneous; provide reasons.

(a) ut − uxx + 1 = 0
(b) ut − uxx + xu = 0
(c) ut − uxxt + uux = 0
(d) utt − uxx + x2 = 0
(e) iut − uxx + u/x = 0
1 1
(f) ux (1 + u2x )− 2 + uy (1 + u2y )− 2 = 0
y
(g) ux + e uy = 0√
(h) ut + uxxxx + 1 + u = 0
Proof. (u, w and v are function n times differentiable for appropriate n with respect to the problem. )

(a) Order 2, Linear inhomogeneous

Note that Lu = ut − uxx is a linear operator since, for any constant a and b,
L(aw + bv) = (aw + bv)t − (aw + bv)xx = a(wt − wxx ) + b(vt − vxx ) = aLw + bLv.
And note that the equation can be represented as Lu = −1. So it is linear inhomogeneous equation.

(b) Order 2, Linear homogeneous

Note Lu = ut − uxx + xu is a linear operator since, for any constant a and b
L(aw + bv) = (aw + bv)t − (aw + bv)xx + x(aw + bv) = a(wt − wxx + xw) + b(vt − vxx + xv) = aLw + bLv.
Then note that the equation can be written as Lu = 0. Thus, it is linear homogeneous.

(c)Order 3, Nonlinear
Note that Lu = ut − uxxt + uux is nonlinear operator since, for any nonzero constant c 6= 1,
L(cu) = (cu)t − (cu)xxt + (cu)(cu)x = c(ut − uxxt + cuux ) 6= c(ut − uxxt + uux ) = cLu.
Since every terms is related to u, the equation is nonlinear.

(d) Order 2, Linear inhomogeneous

Note that Lu = utt − uxx is linear operator since, for any constant a, b,
L(aw + bv) = (aw + bv)tt − (aw + bv)xx = (awtt − awxx ) + (bvtt − bvxx ) = aLw + bLv.
And note that the equation can be written as Lu = −x2 , the equation is Linear inhomogeneous.

(e) Order 2, Linear homogeneous

Note that Lu = iut − uxx + u/x is linear since, for any constant a and b,
L(aw + bv) = i(aw + bv)t − (aw + bv)xx + (aw + bv)/x = aiwt − awxx + aw/x + bivt − bvxx + bv/x = aLw + bLv.
And note that the equation can be written as Lu = 0. Thus, the equation is linear homogeneous.

(f ) Order 1, Nonlinear
1 1
Note that Lu = ux (1 + u2x )− 2 + uy (1 + u2y )− 2 is nonlinear since, for any nonzero constant c 6= 1,
1 1 1 1 1 1
L(cu) = cux (1 + cu2x )− 2 + cuy (1 + cu2y )− 2 = c(ux (1 + cu2x )− 2 + uy (1 + cu2y )− 2 ) 6= c(ux (1 + u2x )− 2 + uy (1 + u2y )− 2 ) = cLu.
Since every terms contain u, the equation is nonlinear.

(g) Order 1, Linear homogeneous

Note that Lu = ux + ey uy is linear since, for any constant a and b
L(aw + bv) = (aw + bv)x + ey (aw + bv)y = a(wx + ey wy ) + b(vx + ey vy ) = aLw + bLv.
Since the equation can be written as Lu = 0, this equation is linear homogeneous equation.

(h) Order 4, Nonlinear√

Note that ut + uxxxx + 1 + u is nonlinear since, for any nonzero constant c 6= 1,
√
L(cu) = (cu)t + (cu)xxxx + 1 + cu 6= cLu.


Page 3
MATH 534H Editor : Byeongho Ban

1.1.4
Show that the difference of two solutions of an inhomogeneous linear equation Lu = g with the same g is a solution of
the homogeneous equation Lu = 0.
Proof. .

Suppose that w and v are given solution of the inhomogeneous equation so that Lw = g and Lv = g. Then observe that
L(w − v) = Lw − Lv = g − g = 0.
Thus, w − v is a solution of Lu = 0. Since w and v were arbitrary, the difference of any two solution of the inhomogeneous
equation is a solution of the homogeneous equation.

1.1.10
Show that the solutions of the differential equation u000 − 3u00 + 4u = 0 form a vector space. Find a basis of it.
Proof. .

Let S be the set of solution of the equation. And let w, v ∈ S be given. Then observe that, for any constant c,
(w + v)000 − 3(w + v)00 + 4(w + v) = (w000 − 3w00 + 4w) + (v 000 − 3v 00 + 4v) = 0,
(cv)000 − 3(cv)00 + 4(cv) = c(v 000 − 3v 00 + 4v) = 0.
Thus, S is closed under addition and scalar multiplication. Since the set of all functions is a vector space and S is a
subset of the vector space which is closed under addition and scalar multiplication, S is a vector space.

Now note that the characteristic equation for the ODE is x3 − 3x2 + 4 = (x + 1)(x − 2)2 = 0. Note that, clearly,
{e−x , e2x , xe2x } is linearly independent. Therefore, the basis of the solution is {e−x , e2x , xe2x }.

1.1.11
Verify that u(x, y) = f (x)g(y) is a solution of the PDE uuxy = ux uy for all pairs of (differentiable) functions f and g of
one variable.
Proof. .

Let two one variable differentiable functions f (x) and g(y) be given. And let u(x, y) = f (x)g(y). Then observe that
uuxy = f g(f g)xy = (f g)(fx gy ) = (fx g)(f gy ) = ux uy .

1.1.12
Verify by direct substitution that
un (x, y) = sin(nx) sinh(ny)
is a solution of uxx + uyy = 0 for every n > 0.
Proof. .

Let n > 0 be given and let un = w. Observe that

wxx + wyy = −n2 sin(nx) sinh(ny) + n2 sin(nx) sinh(ny) = (−n2 + n2 )w = 0
since (sin(nx))xx = (−n cos(nx))x = −n2 sin(nx) and (sinh(nx))xx = (n sinh(nx))x = n2 sinh(nx). Thus, w = un is the
solution of the equation for any n > 0.


Page 4
MATH 534H Editor : Byeongho Ban

Due date: February 14th ByeongHo Ban

Proof. 1.2.1
Observe that the characteristic line of the PDE,
2ut + 3ux = 0
is 3t − 2x = 0. Thus, the solution is the form of u(t, x) = f (3t − 2x) for some differntiable one variable function f . Note
that, by the auxiliary condition, u(0, x) = sin(x) ,we have
sin(x) = u(0, x) = f (−2x).
Let w = −2x, then w
f (w) = − sin .
2
Therefore, our solution is  
3t − 2x
u(t, x) = f (3t − 2x) = − sin .
2
Clearly,    
3t − 2x 3t − 2x
2ut + 3ux = −3 sin + 3 sin =0
2 2
so it solves the PDE and u(0, x) = − sin −2x


2 = − sin (−x) = sin x.

Proof. 1.2.2
Let v = uy , then we have an ODE
dv
= vx = −3v.
dx
Then, by separation of variable, for any one variable differentiable function f (y), we have
uy (x, y) = v(x, y) = f (y)e−3x .
Then, letting F (y) be a primitive function of f (F 0 = f ), we have ,
u(x, y) = F (y)e−3x .
Observe that
3uy + uxy = 3(F (y)e−3x )y + (F (y)e−3x )xy = 3f (y)e−3x + (−3)f (y)e−3x = 0.
Thus, our u solves the P DE.
Thus, our solution to the PDE is u(x, y) = F (y)e−3x for any differntiable one variable function F .


Page 5
MATH 534H Editor : Byeongho Ban

Proof. 1.2.3
Note that, the PDE is the directional derivative of u in the direction of (1 + x2 , 1). Thus, by the ODE
dy 1
= =⇒ y = arctan(x) + C for any constant C.
dx 1 + x2
And y = arctan(x) + C is the characteristic curve for the PDE. Then, our solution to the PDE is
u(x, y) = f (y − arctan(x))
for any differentiable function f . Clearly, our u solves the PDE.
As for the graph of characteristic curves, note the below.


Proof. 1.2.4
Note
(7) → u(x, y) = f (e−x y)
(4) → ux + yuy = 0.
Observe that
(f (e−x y))x + y(f (e−x y))y = (−e−x yf 0 ) + y(e−x f 0 ) = 0.
Thus, (7) solves (4).

Proof. 1.2.5
Note that the PDE means the directional derivative of u in the direction of (x, y) is zero. Thus, we have an ODE and its
solution(by the separation of the variable),
dy y y0 1
= =⇒ = =⇒ ln y = ln x + A =⇒ y = Cx for any constant C and A.
dx x y x
Thus, xy = C is the characteristic curves. Thus, our solution to the PDE is
y
u(x, y) = f ( ) for any differentible function, f.
x
Observe that
y y
xux + yuy = − f 0 + f 0 = 0.
x x
Thus, our solution solves the PDE.


Page 6
MATH 534H Editor : Byeongho Ban

Proof. 1.2.6 √
Observe that the given PDE means that the directional derivative of u in the direction of ( 1 − x2 , 1) is zero. Thus, by
the ODE with the solution
dy 1
=√ =⇒ y = arcsin(x) + C for any constant C,
dx 1 − x2
we have the characteristic curves
y − arcsin(x) = C for any constant C.
Thus, the solution to the PDE is
u(x, y) = f (y − arcsin(x)) for a differentiable function f.
Then by the auxiliary condition, u(0, y) = y, observe that
y = u(0, y) = f (y − arcsin(0)) = f (y).
Therefore, our solution to the PDE is
u(x, y) = y − arcsin(x).
Observe that
p
1 − x2 ux + uy = −1 + 1 = 0.
Thus, our u solves the PDE.

Additional Problem 1
Solve the transport equation 5ux − 6uy = 0 together with the auxiliary condition that u(x, 0) = 4x3
Proof. Note that the PDE is presented as
vu=0
D→
− where →
−
v = (5, −6).
Then observe that the characteristic line is
(x, y) · (6, 5) = 6x + 5y = const.
Thus, our PDE solution is
u(x, y) = f (6x + 5y) for some differentiable fucntion f.
Note that, by the auxiliary condition,
f (6x) = u(x, 0) = 4x3 .
Then observe that, by using w = 6x,
 w 3 w3
f (w) = 4 = .
6 54
Therefore, our solution to the PDE is
(6x + 5y)3
u(x, y) = f (6x + 5y) = .
54
Observe that

18(6x + 5y)2 15(6x + 5y)2

   
5ux − 6uy = 5 −6 =0
54 54
and that
(6x)3
u(x, 0) = = 4x3 .
54


Page 7
MATH 534H Editor : Byeongho Ban

Additional Probem 2
Solve the inhomogeneous transport equation 2ux + 3uy = 1.
Proof. Let (
ζ = 2x + 3y
.
η = −3x + 2y.
And observe that
∂u ∂u ∂ζ ∂u ∂η
ux = = + = 2uζ − 3uη
∂x ∂ζ ∂x ∂η ∂x
∂u ∂u ∂ζ ∂u ∂η
uy = = + = 3uζ + 2uη .
∂y ∂ζ ∂y ∂η ∂y

Then observe that

1 = 2ux + 3uy = 2(2uζ − 3uη ) + 3(3uζ + 2uη ) = 13uζ
Then our particular solution is
1 1
= uζ =⇒ up = ζ + f (η) for some differentiable function f .
13 13
By changing the variable back, we have
1
up (x, y) = (2x + 3y) + f (−3x + 2y).
13
And let’s find the general solution for the PDE by solving
2ux + 3uy = 0.
Through the same method as above, we get
13uζ = 0 =⇒ uζ = 0.
Thus, u = h(η) for some differentiable function h. Then, by changing the variable back, our general solution would be
ug (x, y) = h(−3x + 2y).
Then our solution is
1
u(x, y) = ug (x, y) + up (x, y) = (2x + 3y) + F (−3x + 2y),
13
where F ≡ f + h.
Observe that    
2 3
2ux + 3uy = 2 − 3F 0 +3 + 2F 0 = 1.
13 13
Thus, our u solves the PDE. 

Page 8
MATH 534H Editor : Byeongho Ban

Additional Problem 3
Solve the linear homogeneous equation ux + uy + u = 0.
Proof. Let (
ζ =x+y
.
η = x − y.
And observe that
∂u ∂u ∂ζ ∂u ∂η
ux = = + = uζ + uη
∂x ∂ζ ∂x ∂η ∂x
∂u ∂u ∂ζ ∂u ∂η
uy = = + = uζ − uη .
∂y ∂ζ ∂y ∂η ∂y

By plugging them into the PDE, we have

ux + uy + u = (uζ + uη ) + (uζ − uη ) + u = 2uζ + u = 0.
Then observe that
uζ 1 1
2uζ = −u =⇒ = − =⇒ ln u = − ζ + f (η) for some differentiable function f .
u 2 2
By letting G(η) = ef (η) , we have
1
u(ζ, η) = G(η)e− 2 ζ .
By changing the variable back, our solution is
1
u(x, y) = G(x − y)e− 2 (x+y) .
Observe that
1 1 1 1 1 1 1
ux + uy + u = G0 e− 2 (x+y) − Ge− 2 (x+y) − G0 e− 2 (x+y) − Ge− 2 (x+y) + Ge− 2 (x+y) = 0.
2 2
Thus, our u solves the PDE.

Additional Problem 4-(a)
Check that
1 x+2y
− ex−2y


u(x, y) = e
4
solves the inhomogeneous equation
ux + uy + u = ex+2y

Proof. Note that

1 x+2y 1
− ex−2y 2ex+2y + 2ex−2y .



ux = e & uy =
4 4
Now, observe that
1 x+2y
1
1 x+2y
− ex−2y + 2ex+2y + 2ex−2y + − ex−2y = ex+2y .


ux + uy + u = e e
4 4 4
Thus, the given u(x, y) solves the given PDE.

Additional Problem 4-(b)
Next use the additional problem 3 to write the general form of the solution to
ux + uy + u = ex+2y.
Proof. Note that the solution for the given PDE is the summation of particular solution and general solution. And note
that the solution we found in the Additional problem 3 is the general solution. And the solution given in Additional
Problem 4-(a) is the particular solution. Therefore, our solution to the PDE would be
1 1 x+2y
u(x, y) = G(x − y)e− 2 (x+y) + − ex−2y


e for some differentiable function G.
4
And clearly from previous steps, it solve our PDE.

Additional Problem 4-(c)
Find the solution to
ux + uy + u = ex+2y
that also satisfies u(x, 0) = 1.
Page 9
MATH 534H Editor : Byeongho Ban

Proof. We only need to find specific function G from the general form in 4-(b).

From the general form, observe that

1
1 = u(x, 0) = G(x)e− 2 x .
1 1
Thus, only possibility is when G(x) = e 2 x . Therefore, we have G(x − y) = e 2 (x−y) then our specific solution would be
1 1 1 x+2y 1 x+2y
u(x, y) = e 2 (x−y) e− 2 (x+y) + − ex−2y = e−y + − ex−2y .



e e
4 4
And observe that
1
u(x, 0) = e0 + (ex − ex ) = 1.
4
Thus, it is correct one.

Proof. 1.3.6

Note that the Heat flow equation is

∂u
cρ = ∇ · (κ∇u) = κ∆u,
∂t
where κ, c and ρ are constants.
Then observe that
1 ∂2u ∂2u
 
∂u 1 ∂u ∂u 1 ∂ ∂u
∆u = ∇ · ∇u = ∇ · , , = r + 2 2 + 2.
∂r r ∂θ ∂z r ∂r ∂r r ∂θ ∂z
Since the temperature u does not depend on z and θ, derivatives with respect to z and θ are 0. Therefore, observe that
∂2u
   
∂u 1 ∂ ∂u 1 ∂u u
r

cρut = cρ =κ r =κ +r 2 =κ + urr .
∂t r ∂r ∂r r ∂r ∂r r
κ
Setting k = cρ , we get
 ur 
ut = k urr +
r
as desired.


Page 10
MATH 534H Editor : Byeongho Ban

Proof. 1.3.9
Suppose that D =the ball of radius a centered at the origin and F = r2 x with x = (x, y, z). We should verify
ZZZ ZZ
∇ · Fdx = F · ndS.
D bdyD
∂r ∂r ∂r
Let’s verify the LHS first. Observe that, noting that =
∂x 2x2 , ∂y
= 2y 2 and ∂z = 2z 2 ,
ZZZ ZZZ
∇ · r2 x, r2 y, r2 z dx


∇ · Fdx =
D
Z Z ZD
r2 + 2x2 + r2 + 2y 2 + r2 + 2z 2 dx


=
Z Z ZD
= (3r2 + 2r2 )dx
D
ZZZ
= 5r2 r2 sin φdrdφdθ
Z Da
= 4π 5r4 dr
0
= 4πa5 .
On the other hands,
ZZ ZZ
(x, y, z)
F · ndS = r2 (x, y, z) · dS
bdyD bdyD r
ZZ
= r(x2 + y 2 + z 2 )dS
bdyD
ZZ
= r3 dS
bdyD
Z π Z 2π
= a3 a2 sin φdφdθ
φ=0 θ=0
5
= 4πa .
Therefore, the left and right hand sides are equal so divergence theorem is valid in this case.

Proof. 1.3.10
Let r > 0 be given and let Dr be a ball of radius r. Then observe that, by the divergence theorem,
Z Z Z  Z Z 
   

 ∇ · f dx =
  f · ndS 

Dr bdyDr
ZZ
≤ |f |dS
bdyDr
ZZ
1
≤ 3+1
dS
bdyDr |x|
ZZ
1
= 3+1
dS
bdyDr r
4πr2
= .
1 + r3
Observe that
 
 ZZZ  
4πr2
   ZZZ 

∇ · f dx =  lim ∇ · f dx ≤ lim = 0.
  
r→∞ 1 + r 3

  r→∞ Dr
all space 
Therefore,
 
 ZZZ  ZZZ
 
∇ · f dx = 0 =⇒ ∇ · f dx = 0.
 

 
all space  all space D

Page 11
MATH 534H Editor : Byeongho Ban

Proof. 1.4.1

Let u(x, t) = x2 + 2t. And note that

ut = (2t + x2 )t = 2 = (2t + x2 )xx = uxx .
Thus, u solves the PDE. Also, observe that
u(x, 0) = x2 + 2 · 0 = x2 .
Thus, u also satisfies the initial condition. Therefore, the u is the function we are finding.


Page 12
MATH 534H Editor : Byeongho Ban

Due Date : February 28th Byeongho Ban

Additional Problem 1- (a)
Find the general solution to Problem 8 in Section 1.2. Specify what method you are using and explain step by step your
work. Show all your work.

Problem 8 of Section 1.2

Solve aux + buy + cu = 0.
Proof. Let’s use variation method. Let
(
ζ = ax + by
η = −bx + ay.
Then observe that
∂u ∂u ∂ζ ∂u ∂η
ux = = + = auζ − buη
∂x ∂ζ ∂x ∂η ∂x
∂u ∂u ∂ζ ∂u ∂η
uy = = + = buζ + auη .
∂y ∂ζ ∂y ∂η ∂y

By plugging them into the PDE, we have

aux + buy + cu = a(auζ − buη ) + b(buζ + auη ) + cu = (a2 + b2 )uζ + cu = 0.
Then observe that
uζ c c
(a2 + b2 )uζ = −cu =⇒ =− 2 =⇒ ln u = − 2 ζ + f (η) for some differentiable function f .
u a + b2 a + b2
By letting G(η) = ef (η) , we have
− c ζ
u(ζ, η) = G(η)e a2 +b2 .
By changing the variable back, our solution is
c
− a2 +b 2 (ax+by)
u(x, y) = G(−bx + ay)e .
Observe that

c
− a2 +b 2 (ax+by)
ac − c (ax+by)
aux + buy + cu = −abG0 e −a Ge a2 +b2
a2 +b 2
c
− a2 +b 2 (ax+by)
bc − c (ax+by) − c (ax+by)
+ baG0 e −b Ge a2 +b2 + cGe a2 +b2
a2 + b2
= 0.
Thus, our u solves the PDE.

Additional Problem 1-(b)
Choose a = 2, b = 5, and c = 29 and find the solution u(x, y) to part (a) that also satisfies u(x, 0) = e−3x .
Proof. From part (a), we know that the solution for 2ux + 5uy + 29u = 0 is
− 2229 (2x+5y)
u(x, y) = G(−5x + 2y)e +52 = G(−5x + 2y)e−(2x+5y) ,
where G(x) = ef (x) for some differntiable function f . Then observe that, by given condition,
e−3x = u(x, 0) = G(−5x)e−2x .
Then we have, with α = −5x,
α
e−x = G(−5x) = G(α) = e 5 .
Therefore,
−5x+2y 23
u(x, y) = G(−5x + 2y)e−(2x+5y) = e 5 e−(2x+5y) = e−3x− 5 y .
Then note that
2ux + 5uy + 29u = −6u − 23u + 29u = 0.
Thus, our solution is
23
u(x, y) = e−3x− 5 y .


Page 13
MATH 534H Editor : Byeongho Ban

Additional Problem 1-(c)

Check that 61 (ex+y − e3x−y ) is a particular solution to the inhomogeneous equation
2ux + 5uy + 29u = (6ex+y − 5e3x−y ).
Proof. Observe that

1 1 1
2ux + 5y + 29u = 2 ex+y − 3e3x−y + 5 ex+y + e3x−y + 29 ex+y − e3x−y




6  6
  6
2 5 29 x+y 5 29 3x−y
= + + e + −1 + − e
6 6 6 6 6
= 6ex+y − 5e3x−y .
Thus, 16 (ex+y − e3x−y ) solves the inhomogeneous equation so it is a particular solution.

Additional Problem 1-(d)
Use part (a) a = 2, b = 5, and c = 29 together with part (c) to find the general solution to the inhomogeneous equation
2ux + 5uy + 29u = (6ex+y − 5e3x−y ).
Proof. Let
∂ ∂
L=2 +5 + 29.
∂x ∂y
Then let
uh (x, y) = G(−5x + 2y)e−(2x+5y)
for any differentiable function G and
1 x+y
ui (x, y) = (e − e3x−y ).
6
Then observe that
L(uh + ui ) = Luh + Lui = 0 + (6ex+y − 5e3x−y )
since it is clear that L is linear.
Therefore, for any differentiable function G, our general solution to the PDE is
1
G(−5x + 2y)e−(2x+5y) + (ex+y − e3x−y ).
6

Proof. 1.5.1

The solution to the PDE is not unique depending on the value of L. Consider the function, when L = nπ for any n ∈ Z,
u(x) = eix − e−ix .
Observe that
d2 u
+ u = i2 eix − (−i)2 e−ix + eix − e−ix = −eix + e−ix + eix − e−ix = 0
dx2
and that
u(0) = ei0 − e−i0 = 1 − 1 = 0 and u(L) = u(nπ) = enπi − e−nπi = cos(nπ) − cos (−nπ) = 0.
Thus, u solves the ODE.

Note that the general solution for the ODE is

ψ(x) = Aeix + Be−ix .
And observe that
ψ(0) = 0 ⇐⇒ Ae0 + Be0 = 0
⇐⇒ A = −B
so that
ψ(L) = 0 = ψ(0) ⇐⇒ A(eiL − e−iL ) = 0
⇐⇒ cos(2L) + i sin(2L) = e2iL = 1
⇐⇒ 2L = 2πk
⇐⇒ L = kπ ∀k ∈ Z.
Thus, in order for any nonzero function to solve the given ODE, L should be integer multiple of π. Therefore, the solution
of the ODE is unique if and only if L is not an integer multiple of π.

Page 14
MATH 534H Editor : Byeongho Ban

Proof. 1.5.4

(a)
Let u be the solution to the given Neumann problem. Then, for arbitrary constant C, observe that
∆(u + C) = ∆(u) + ∆(C) = f + 0 = f in D
and
∂ ∂u ∂C
(u + C) = + = 0 + 0 on ∂D.
∂n ∂n ∂n
Thus, as long as C is constant, we can add C to the solution of the problem to make another different solution. Therefore,
the solution to the PDE is not unique so this problem is ill-posed.

(b)
Suppose that the problem has solution, u. Then, by the divergence theorem, observe that
ZZZ ZZZ
f (x, y, z)dxdydz = ∆udxdydz
D
Z ZD
= (∇u) · →
−
n dS
∂D
ZZ
∂u
= dS
∂n
∂D
= 0.
Therefore, ZZZ
f (x, y, z)dxdydz = 0
D
is a necessary condition for the Neumann problem to have a solution.

(c)
Note that, for heat flow or diffusion,
∂u
f (x, y, x) = k
∂t
for some constant k > 0.

The physical interpretation from (a) is that the diffusion or heat flow only depends on the difference of concentration or
temperature level between two regions. Thus, when we increase a certain amount of heat or concentration equally, even
if the increased level is very high, the diffusion and flow would be same as before the increasing.

Note from (b) that

ZZZ ZZZ
∂u ∂ ∂
k dxdydz = k udxdydz = k M =0
∂t ∂t ∂t
D D
∂u
where M is the total mass over all D. Note that this hold when ∂n = 0 on ∂D which means there is no lose of mass in D.
Thus, physical interpretation from (b) is that, when there is no mass going out from ∂D, the total mass is preserved over
time.


Page 15
MATH 534H Editor : Byeongho Ban

Proof. 1.5.5

Note that
0 = ux + yuy = (1, y) · (ux , uy ) = (1, y) · ∇u.
Then solve the ODE
dy y y0
= = y =⇒ = 1 =⇒ ln y = x + C =⇒ y = Aex =⇒ A = ye−x
dx 1 y
for some constant C and A = eC . Thus, the characteristic curve is
A = ye−x .
Then the solution to the PDE is
u(x, y) = f (ye−x ) for any differentiable function f.
(a)
Suppose that, when φ(x) = x, the solution exists.
Then f (0) = u(x, 0) = φ(x) = x. Note f (0) should be constant and x is not constant. Thus, it is a contradiction.
Therefore, there does not exist such f .

(b)
Suppose that φ(x) = 1. Then observe that
f (0) = u(x, 0) = φ(x) = 1.
ax
Note that ga (x) = e , for any constant a, satisfies g(0) = 1 which says for any constant a, f can be ga . Therefore, for
infinitely many a, u(x, y) = ga (ye−x ) is a solution. So there are infinitely many solution to the problem.

Problem 6(modified)
Solve the equation ux + 2xy 2 uy = 0 and find a solution that satisfies the auxiliary condition u(0, y) = y.
Proof. Note that
ux + 2xy 2 uy = (1, 2xy 2 ) · (ux , uy ) = (1, 2xy 2 ) · ∇u = 0.
Thus, the tangent of characteristic curve is
dy 2xy 2
= = 2xy 2 .
dx 1
Then observe that
y0 1
2
= 2x =⇒ − = x2 + C
y y
where C is some constant. Then our characteristic curve is
1
x2 + = C.
y
Thus, our solution is  
2 1
u(x, y) = f x + for some differentiable function f .
y
By the given auxiliary condition, we get  
1
y = u(0, y) = f .
y
Then we have
1
f (x) = .
x
Therefore, our final solution is  
2 1 1 y
u(x, y) = f x + = 2 1 = 2 .
y x +y yx + 1
Additionally, note that, if y were 0, then our PDE is
ux = 0
which implies that u(x, y) = g(y) for some differentiable function g. Then, by the auxiliary condition,
g(y) = u(0, y) = y = 0.
Thus, u(x, y) = y = 0 which is consistent with our result. 

Page 16
MATH 534H Editor : Byeongho Ban

Proof. 1.6.1

(a)
Note that we can reduce the PDE as
uxx − uxy + 2uy + uyy − 3uyx + 4u = 0 =⇒ uxx − 4uxy + uyy + 2uy + 4u = 0.
Then observe that
a11 a22 − a212 = 1 · 1 − (−2)2 = 1 − 4 = −3 < 0.
Thus, the equation is Hyperbolic.

(b)
Observe that
a11 a22 − a212 = 9 · 1 − (3)2 = 9 − 9 = 0.
Thus, the equation is Parabolic.

Proof. 1.6.2
Observe that
a11 a22 − a212 = (1 + x)(−y 2 ) − (xy)2 = −y 2 − xy 2 − x2 y 2 = −y 2 (1 + x + x2 ).
Let
E = {(x, y) ∈ R2 : −y 2 − xy 2 − x2 y 2 > 0} = {(x, y) ∈ R2 : 1 + x + x2 > 0 & y 6= 0}
H = {(x, y) ∈ R2 : −y 2 − xy 2 − x2 y 2 < 0} = {(x, y) ∈ R2 : 1 + x + x2 < 0 & y 6= 0}
P = {(x, y) ∈ R2 : −y 2 − xy 2 − x2 y 2 = 0} = {(x, y) ∈ R2 : 1 + x + x2 = 0 or y = 0}
Then the equation is Elliptic in E, Hyperbolic in H and Parabolic in P .
Further note that, since
 2
2 1 3
x +x+1= x+ + > 0 ∀x ∈ R,
2 4
we have E = {(x, y) ∈ R2 : y 6= 0}, H = ∅, and P = {(x, y) ∈ R2 : y = 0}. Therefore, the equation is Elliptic in R2 except
x−axis and is Parabolic in x− axis. And the equation is never hyperbolic.

Page 17
MATH 534H Editor : Byeongho Ban

Proof. 1.6.4

Suppose the general form of second order PDE

a11 uxx + 2a12 uxy + a22 uyy + a1 ux + a2 uy + a0 u+ = 0

From given PDE, note that
a11 a22 − a212 = 1 · 4 − (−2)2 = 4 − 4 = 0.
Thus, the equation is Parabolic.
Let f and g be arbitrary differentiable function and let
u(x, y) = f (y + 2x) + xg(y + 2x).
Then observe that
uxx − 4uxy + 4uyy = (4f 00 + 2g 0 + 2g 0 + 4xg 00 ) − 4(2f 00 + g 0 + 2xg 00 ) + 4(f 00 + xg 00 )
= (4 − 8 + 4)f 00 + (2 + 2 − 4)g 0 + (4x − 8x + 4x)g 00
= 0.
Thus, u(x, y) = f (y + 2x) + xg(y + 2x) for arbitrary functions f and g is a solution to the PDE.

Additional Problem 2
Find the regions in R2 where x2 uxx + 4uxy + y 2 uyy = 0 is respectively elliptic, parabolic, hyperbolic. Plot these regions.
Proof. Let the general form of second order PDE is
a11 uxx + a12 uxy + a22 uyy + (low order terms) = 0.
Then, in our case, a11 = x , a12 = 2, and a22 = y 2 . Then note that
2

a11 a22 − a212 = x2 y 2 − 4.

Then let
E = {(x, y) ∈ R2 : x2 y 2 − 4 > 0}
H = {(x, y) ∈ R2 : x2 y 2 − 4 < 0}
P = {(x, y) ∈ R2 : x2 y 2 − 4 = 0}.
Then the PDE is Elliptic in E, Hyperbolic in H and Parabolic in P . The graph of x2 y 2 = 4 and the region of E, H and
P are drawn indicated at below. (P is just the union of the lines.)

Page 18
MATH 534H Editor : Byeongho Ban

Proof. 2.1.1
Note that ex ∈ C 2 and sin x ∈ C 1 . Consider the function
Z x+ct
1  x+ct x−ct 1 1 1
sin sds = [ex+ct + ex−ct ] − [cos(x + ct) − cos(x − ct)].

u(x, t) = e +e +
2 2c x−ct 2 2c
Observe that
1 x 1
u(x, 0) = [e + ex ] − [cos(x) − cos(x)] = ex
2 2c
and that
1 x+ct 1
ut = [ce − cex−ct ] − [−c sin(x + ct) − c sin(x − ct)]
2 2c
so
1 1
ut (x, 0) = [cex − cex ] − [−c sin(x) − c sin(x)] = sin x.
2 2c
Thus, it satisfies the auxiliary condition. Also, observe that
1 1
utt = (ut )t = [c2 ex+ct + c2 ex−ct ] − [−c2 cos(x + ct) + c2 cos(x − ct)]
2 2c
and that
1 1
ux = [ex+ct + ex−ct ] − [− sin(x + ct) + sin(x − ct)]
2 2c
so
1 x+ct 1
uxx = [e + ex−ct ] − [− cos(x + ct) + cos(x − ct)].
2 2c
Therefore, utt = c2 uxx so u solves the PDE. Therefore, u(x, y) is the solution to the IVP.


Page 19
MATH 534H Editor : Byeongho Ban

Proof. 2.1.2
Note that log(1 + x2 ) ∈ C 2 and 4 + x ∈ C 1 . Then consider the function
Z x+ct
1 1
u(x, t) = [log(1 + (x + ct)2 ) + log(1 + (x − ct)2 )] + (4 + s)ds
2 2c x−ct
 
1 1 1
= [log(1 + (x + ct)2 ) + log(1 + (x − ct)2 )] + 8ct + [(x + ct)2 − (x − ct)2 ]
2 2c 2
 
1 1 1
= [log(1 + (x + ct)2 ) + log(1 + (x − ct)2 )] + 8ct + [4xct]
2 2c 2
1
= [log(1 + (x + ct)2 ) + log(1 + (x − ct)2 )] + 4t + xt
2
1
= [log(1 + x2 + c2 t2 + 2xct) + log(1 + x2 + c2 t2 − 2xct)] + 4t + xt.
2
Note that
1
u(x, 0) = [log(1 + (x)2 ) + log(1 + (x)2 )] + 0 = log(1 + x2 ).
2
And note that
2c2 t + 2xc 2c2 t − 2xc
 
1
ut = + +4+x
2 1 + x2 + c2 t2 + 2xct 1 + x2 + c2 t2 − 2xct
and that  
1 2xc −2xc
ut (x, 0) = + + 4 + x = 4 + x.
2 1 + x2 1 + x2
Thus, u satisfies the initial conditions. Furthermore, observe that

2c2 t + 2xc 2c2 t − 2xc

 
∂ 1
utt = + +0
∂t 2 1 + x2 + c2 t2 + 2xct 1 + x2 + c2 t2 − 2xct
1 2c2 (1 + x2 + c2 t2 + 2xct) − (2c2 t + 2xc)2 2c2 (1 + x2 + c2 t2 − 2xct) − (2c2 t − 2xc)2
 
= +
2 (1 + x2 + c2 t2 + 2xct)2 (1 + x2 + c2 t2 − 2xct)2
c2 2(1 + x2 + c2 t2 + 2xct) − (2x + 2ct)2 2(1 + x2 + c2 t2 − 2xct) − (2x − 2ct)2
 
= +
2 (1 + x2 + c2 t2 + 2xct)2 (1 + x2 + c2 t2 − 2xct)2
and that
   
∂ 1 2x + 2ct 2x − 2ct
uxx = + +t
∂x 2 1 + x2 + c2 t2 + 2xct 1 + x2 + c2 t2 − 2xct
1 2(1 + x2 + c2 t2 + 2xct) − (2x + 2ct)2 2(1 + x2 + c2 t2 − 2xct) − (2x − 2ct)2
 
= + .
2 (1 + x2 + c2 t2 + 2xct)2 (1 + x2 + c2 t2 − 2xct)2
Thus, c2 uxx = utt so u satisfies the PDE. Therefore, u(x, t) is the solution to the IVP.


Page 20
MATH 534H Editor : Byeongho Ban

Proof. 2.1.8 (Hint : Recall a function f on R is odd if f (x) = −f (−x).)

(a)
Let v = ru, then observe that
    
2 2 vtt 2 v 2 v
utt = c urr + ur =⇒ =c +
r r r rr r r r
   
vtt 2 rvr − v 2 rvr − v
=⇒ =c +
r r2 r r r2
 2   
vtt 2 (r )(rvrr + vr − vr ) − 2r(rvr − v) 2 rvr − v
=⇒ =c +
r r4 r r2
   
vtt (r)(rvrr ) − (2rvr − 2v) 2rvr − 2v
=⇒ = c2 3
+
r r r3
vtt v 
rr
=⇒ = c2
r r
=⇒ vtt = c2 vrr .

(b)
We use general solution of the wave equation
v(r, t) = f (r + ct) + g(r − ct)
for arbitrary functions f and g. Observe that
vtt = c2 f 00 + c2 g 00 = c2 (f 00 + g 00 ) = c2 vxx .
Thus, v solves the PDE. Then note that
v 1
u= = (f (r + ct) + g(r − ct)).
r r
Then observe that v 1 2 00
c f (r + ct) + c2 g 00 (r − ct)


utt = =
r tt r
and that

rf 0 (r + ct) + rg 0 (r − ct) − f (r + ct) − g(r − ct)

ur =
r2
∂ur
urr =
∂r
r2 (f 0 (r + ct) + g 0 (r − ct) + rf 00 (r + ct) + rg 00 (r − ct) − f 0 (r + ct) − g 0 (r − ct))
=
r4
0 0
−2r(rf (r + ct) + rg (r − ct) − f (r + ct) − g(r − ct))
+
r4
(r3 )f 00 (r + ct) + (r2 − 2r2 − r2 )f 0 (r + ct) + 2rf (r + ct)
=
r4
(r )g (r − ct) + (r − r − 2r2 )g 0 (r − ct) + 2rg(r − ct)
3 00 2 2
+
r4 
(f (r + ct) − g (r − ct)) 2 −2rf 0 (r + ct) − 2rg 0 (r − ct) + 2(f (r + ct) + g(r − ct))
00 00

= +
r r r2
0 0
 
utt 2 r(f (r + ct) + g (r − ct)) − (f (r + ct) + g(r − ct))
= 2 + −
c r r2
utt 2
= 2 − ur .
c r
Therefore,  
2
utt = c2 urr + ur .
r
Thus,
v
u=
r
solves our wave equation.


Page 21
MATH 534H Editor : Byeongho Ban

Proof. 2.1.8
(c)
Note that vtt = c2 vrr and that v(r, 0) = ru(r, 0) = rφ(r) and vt (r, 0) = rut (r, 0) = rψ(r). Then
Z r+ct
1 1
v(r, t) = [(r + ct)φ(r + ct) + (r − ct)φ(r − ct)] + sψ(s)ds
2 2c r−ct
v
is the solution to the PDE vtt = c2 vrr . Then u(r, t) = r solves the PDE
 
2 2
utt = c urr + ur
r
by the part (b).
Now note that
     Z r+ct
v(r, t) 1 ct ct 1
u(r, t) = = 1+ φ(r + ct) + 1 − φ(r − ct) + sψ(s)ds.
r 2 r r 2cr r−ct
Then Z r
1 1
u(r, 0) = [(1 + 0) φ(r + 0) + (1 − 0) φ(r − 0)] + sψ(s)ds = φ(r) + 0 = φ(r).
2 2cr r
Also, note that
1 nc o 1
ut (r, t) = (φ(r + ct) − φ(r − ct)) + c(φ0 (r + ct) − φ0 (r − ct)) + (c(r + ct)ψ(r + ct) + c(r − ct)ψ(r − ct)) .
2 r 2cr
Then
1 nc o 1
ut (r, 0) = (0) + c(0) + (crψ(r) + crψ(r)) = ψ(r).
2 r 2cr
Thus, our u(r, t) satisfies the given initial condition. Therefore, u(r, t) solves the PDE
 
2 2
utt = c urr + ur .
r


Page 22
MATH 534H Editor : Byeongho Ban

Due date: March 7th ByeongHo Ban

Additional Problem 1
First find the solution to the linear homogeneous wave equation with wave speed 1 with initial condition u(x, 0) = sin x,
ut (x, 0) = 0. Then calculate ut (0, t).
Proof. We must solve the following IVP:

utt − uxx = 0

u(x, 0) = sin x

ut (x, 0) = 0.


The solution for IVP of wave equation is given by

1 x+t
Z
1
u(x, t) = [u(x + t, 0) + u(x − t, 0)] + ut (s, 0)ds
2 2 x−t
1 x+t
Z
1
= [sin(x + t) + sin(x − t)] + 0ds
2 2 x−t
1
= [sin(x + t) + sin(x − t)]
2
Let’s double check:
Note that, clearly,
1  1
utt = − sin(x + t) − (−1)2 sin(x − t) = [− sin(x + t) − sin(x − t)] = uxx .
2 2
Now, observe that
1 1
u(x, 0) = [sin(x + 0) + sin(x − 0)] = [2 sin x] = sin x
2 2
and that
1 1
ut (x, t) = [cos(x + t) − cos(x − t)] =⇒ ut (x, 0) = [cos x − cos x] = 0.
2 2
Thus, our solution solves the IVP.

Lastly, note that

1
ut (0, t) = [cos t − cos(−t)] = 0
2
since cos is even function.


Page 23
MATH 534H Editor : Byeongho Ban

Additional Problem 2
Find the solution to the wave equation utt − 4uxx = 0 and with initial conditions u(x, 0) = sin x, ut (x, 0) = 10. Calculate
then ut (0, t).
Proof. We must solve the following IVP:

utt − 4uxx = 0

u(x, 0) = sin x

ut (x, 0) = 10.


The solution for IVP of wave equation is given by

1 x+2t
Z
1
u(x, t) = [u(x + 2t, 0) + u(x − 2t, 0)] + ut (s, 0)ds
2 4 x−2t
1 x+2t
Z
1
= [sin(x + 2t) + sin(x − 2t)] + 10ds
2 4 x−2t
1 1
= [sin(x + 2t) + sin(x − 2t)] + (4t)(10)
2 4
1
= [sin(x + 2t) + sin(x − 2t)] + 10t.
2
Let’s double check:
Note that, clearly,
1  4
utt = −(2)2 sin(x + 2t) − (−2)2 sin(x − 2t) = [− sin(x + t) − sin(x − t)] = 4uxx .
2 2
Now, observe that
1 1
u(x, 0) = [sin(x + 0) + sin(x − 0)] + 0 = [2 sin x] = sin x
2 2
and that
1 1
ut (x, t) = [2 cos(x + 2t) − 2 cos(x − 2t)] + 10 =⇒ ut (x, 0) = [2 cos x − 2 cos x] + 10 = 10.
2 2
Thus, our solution solves the IVP.

Lastly, note that

1
ut (0, t) = [2 cos 2t − 2 cos(−2t)] + 10 = 10
2
since cos is even function.


Page 24
MATH 534H Editor : Byeongho Ban

2.1.9

We must solve following IVP:


uxx − 3uxt − 4utt = 0

u(x, 0) = x2

ut (x, 0) = ex .


Observe that

uxx − 3uxt − 4utt = (∂x + ∂t ) (∂x − 4∂t ) u.

Let v = (∂x − 4∂t ) u. Then since the solution of the transport equation vx +vt = 0 is v(x, t) = f (x−t) for any differentiable
function f : R → R, we have a following PDE
(∂x − 4∂t ) u = f (x − t).
Note that the solution of homogeneous transport PDE, ux − 4ut = 0 is g x + 41 t for any function g.

And observe that, letting q = h(x − t)

qx − 4qt = h0 (x − t) + 4h0 (x − t).
Since RHS is a function of x − t and f is arbitrary, h(x − t) satisfy the PDE and so is particular solution. Thus,
 
1
u(x, t) = g x + t + h(x − t).
4
Now, observe that
u(x, 0) = g(x) + h(x) = x2 =⇒ g 0 (x) + h0 (x) = 2x
1
ut (x, 0) = g 0 (x) − h0 (x) = ex .
4
Thus, g 0 (x) = 54 (2x + ex ) and h0 (x) = 15 (2x − 4ex ). Then observe that
 
1
u(x, t) = g x + t + h(x − t)
4
Z x+ 41 t Z x−t
= g 0 (s)ds + h0 (s)ds + g(0) + h(0)
0 0
Z x+ 41 t Z x−t
4 1
= (2s + es )ds + (2s − 4es )ds + 02
0 5 0 5
 2
4 1 4 x+ 1 t 4 1 4 4
= x + t + e 4 − + (x − t)2 − ex−t +
5 4 5 5 5 5 5
"  2 #
4 1 1 4 x+ 1 t
h i
= x + t + (x − t)2 + e 4 − ex−t .
5 4 5 5
Thus, the u(x, t) above is the our solution to the IVP.

Page 25
MATH 534H Editor : Byeongho Ban

2.1.10

We should solve the following IVP :


uxx + uxt − 20utt = 0

u(x, 0) = φ(x)

ut (x, 0) = ψ(x)


Observe that

uxx + uxt − 20utt = (∂x + 5∂t ) (∂x − 4∂t ) u.

Let v = (∂x − 4∂t ) u. Then since the solution of the transport equation vx + 5vt = 0 is v(x, t) = f (x − 51 t) for any
differentiable function f : R → R, we have a following PDE
 
1
(∂x − 4∂t ) u = f x − t .
5
Note that the solution of homogeneous
transport PDE, ux − 4ut = 0 is g x + 14 t for any function g.

And observe that, letting q = h x − 15 t

   
0 1 0 1
qx − 4qt = h x − t + 4h x − t .
5 5
1 1


Since RHS is a function of x − 5 t and f is arbitrary, h x − 5 t satisfy the PDE and so is particular solution. Thus,
   
1 1
u(x, t) = g x + t + h x − t .
4 5
Now, observe that
u(x, 0) = g(x) + h(x) = φ(x) =⇒ g 0 (x) + h0 (x) = φ0 (x)
1 1
ut (x, 0) = g 0 (x) − h0 (x) = ψ(x).
4 5
Thus, g 0 (x) = 94 (φ0 (x) + 5ψ(x)) and h0 (x) = 59 (φ0 (x) − 4ψ(x)). Then observe that
   
1 1
u(x, t) = g x + t + h x − t
4 5
Z x+ 14 t Z x− 15 t
= g 0 (s)ds + h0 (s)ds + g(0) + h(0)
0 0
Z x+ 14 t Z x− 15 t
4 0 5 0
= (φ (s) + 5ψ(s))ds + (φ (s) − 4ψ(s))ds + φ(0)
0 9 0 9
1 1
20 x+ 4 t 20 x− 5 t
  Z   Z
4 1 4 5 1 5
= φ x+ − φ(0) + ψ(s)ds + φ x − t − φ(0) − ψ(s)ds + φ(0)
9 4 9 9 0 9 5 9 9 0
1
20 x+ 4 t
     Z
4 1 5 1
= φ x+ t + φ x− t + ψ(s)ds.
9 4 9 5 9 x− 51 t
Thus, the u(x, t) above is the our solution to the IVP.

Page 26
MATH 534H Editor : Byeongho Ban

2.1.11

1
Firstly, observe that up (x, t) = − 16 sin(x + t) is the particular solution to the PDE since
     
1 1 1
3(up )tt + 10(up )xt + 3(up )xx = 3 sin(x + t) + 10 sin(x + t) + 3 sin(x + t) = sin(x + t).
16 16 16
We must find the general solution, the solution of homogeneous linear PDE 3utt + 10uxt + 3uxx = 0.
Note that is it hyperbolic since
( coefficient of uxx )(coefficient of utt ) − (coefficient of uxt )2 = 9 − 100 = −91 < 0.
Observe that

3utt + 10uxt + 3uxx = (3∂x + ∂t ) (∂x + 3∂t ) u.

Let v = (∂x + 3∂t ) u. Then since the solution of the transport equation 3vx + vt = 0 is v(x, t) = f (x − 3t) for any
differentiable function f : R → R, we have a following PDE
(∂x + 3∂t ) u = f (x − 3t) .
Note that the solution of homogeneous transport PDE, ux + 3ut = 0 is g x − 31 t for any function g.

And observe that, letting q = h (x − 3t)

qx + 3qt = h0 (x − 3t) + 3h0 (x − 3t) .
Since RHS is a function of x − 3t and f is arbitrary, h(x − 3t) satisfy the PDE and so is particular solution. Thus,
 
1
u(x, t) = g x − t + h (x − 3t) .
3
Thus,  
1 1
u(x, t) = g x − t + h (x − 3t) − sin(x + t)
3 16
is the our solution to the IVP.

Page 27
MATH 534H Editor : Byeongho Ban

Additional Problem 3
Find the general solution to uxx + uxt − 10utt = 0 (Check whether it is hyperbolic first.)
Proof. Firstly, let’ check if it is Hyperbolic.
Let the general second order PDE be

a11 uxx + a12 uxt + a22 utt + [lower order terms] = 0

Then note that, by comparing with our PDE,
a11 a22 − a212 = −10 − (1)2 = −11 < 0.
Thus, the PDE is Hyperbolic.
We will change the coordinate (x, t) → (x0 , t0 ) so that
 2 √ !2
1 41
uxx + uxt − 10utt = ∂x + ∂t u − ∂t u = ut0 t0 − ux0 x0 .
2 2
We need to specify α, β, γ and δ satisfying following system
(
x = αx0 + βt0
t = γx0 + δt0 .
Observe that
1 ∂t ∂x 1
ux + ut = ut0 = ut 0 + ux 0 = δut + βux =⇒ δ = and β = 1
√ 2 ∂t ∂t 2√
41 ∂t ∂x 41
ut = ux0 = ut 0 + ux 0 = γut + αux =⇒ γ = and α = 0.
2 ∂x ∂x 2
Thus, we get
(
x =√t0 =⇒ t0 = x
t = 241 x0 + 21 t0 =⇒ x0 = √2 − 12 x .


41
t
Note that the general solution for ut0 t0 − ux0 x0 = 0 is u(x0 , t0 ) = f (x + t ) + g(x0 − t0 ) = 0 for any functions f and g which
0 0

are at least two times differentiable.

Then note that
               
2 1 2 1 2 1 2 1
u(x, t) = f √ t − x + x +g √ t− x −x =f √ t+ 1− √ x +g √ t − √ + 1 x .
41 2 41 2 41 41 41 41
√   √ 
41−1
By setting fe(z) = f √
41
z and ge(z) = g − 1+ √ 41 z , our solution is
41
   
2 2
u(x, t) = fe x + √ t + ge x − √ t .
41 − 1 41 + 1
(Actually, fe and ge are any twoice differentiable functions. )


Page 28
MATH 534H Editor : Byeongho Ban

Additional Problem 4
Find the general solution to uxx + 2uxt − 20utt = 0. (Check whether it is hyperbolic first.)
Proof. Firstly, let’ check if it is Hyperbolic.
Let the general second order PDE be

a11 uxx + a12 uxt + a22 utt + [lower order terms] = 0

Then note that, by comparing with our PDE,
a11 a22 − a212 = −20 − (2)2 = −44 < 0.
Thus, the PDE is Hyperbolic.
We will change the coordinate (x, t) → (x0 , t0 ) so that
√ 2
2
uxx + 2uxt − 20utt = (∂x + ∂t ) u − 21∂t u = ut0 t0 − ux0 x0 .
We need to specify α, β, γ and δ satisfying following system
(
x = αx0 + βt0
t = γx0 + δt0 .
Observe that
∂t ∂x
ux + ut = ut0 = ut 0
+ ux 0 = δut + βux =⇒ δ = 1 and β = 1
∂t ∂t
√ ∂t ∂x √
21ut = ux0 = ut 0 + ux 0 = γut + αux =⇒ γ = 21 and α = 0.
∂x ∂x
Thus, we get
(
x = t0 =⇒ t0 = x
√
t = 21x0 + t0 =⇒ x0 = √1 (t
21
− x) .
Note that the general solution for ut0 t0 − ux0 x0 = 0 is u(x0 , t0 ) = f (x + t ) + g(x0 − t0 ) = 0 for any functions f and g which
0 0

are at least two times differentiable.

Then note that
           
1 1 1 1 1 1
u(x, t) = f √ (t − x) + x + g √ (t − x) − x = f √ t + 1 − √ x +g √ t− √ +1 x .
21 21 21 21 21 21
√   √ 
21−1 1+ 21
By setting fe(z) = f √
21
z and ge(z) = g − √21 z , our solution is
   
1 1
u(x, t) = f x + √
e t + ge x − √ t .
21 − 1 21 + 1
(Actually, fe and ge are any twoice differentiable functions. )


Page 29
MATH 534H Editor : Byeongho Ban

Due Date : March 21st, 2019 ByeongHo Ban

Additional Problem 1
Find the solution to the IVP

uxx − 6uxt + 5utt = 0,
 x ∈ R, t > 0
u(x, 0) = x2

ut (x, 0) = 0


Check first whether the second order PDE is hyperbolic.

Proof. Firstly, let’ check if it is Hyperbolic.
Let the general second order PDE be

a11 uxx + a12 uxt + a22 utt + [lower order terms] = 0

Then note that, by comparing to our PDE,
a11 a22 − a212 = 5 − (−6)2 = −31 < 0.
Thus, the PDE is Hyperbolic.

Observe that

uxx − 6uxt + 5utt = (∂x − ∂t ) (∂x − 5∂t ) u.

Let v = (∂x − 5∂t ) u. Then since the solution to the transport equation vx −vt = 0 is v(x, t) = f (x+t) for any differentiable
function f : R → R, we have a following PDE
(∂x − 5∂t ) u = f (x + t).
Note that the solution of homogeneous transport PDE, ux − 5ut = 0 is g x + 15 t for any function g.

And observe that, letting q = h(x + t)

qx − 5qt = h0 (x + t) − 5h0 (x + t).
Since RHS is a function of x + t and f is arbitrary, h(x + t) satisfy the PDE and so is particular solution. Thus,
 
1
u(x, t) = g x + t + h(x + t).
5
Now, observe that
u(x, 0) = g(x) + h(x) = x2 =⇒ g 0 (x) + h0 (x) = 2x
1
ut (x, 0) = g 0 (x) + h0 (x) = 0.
5
Thus, g 0 (x) = 45 (2x) = 5x
2 and h0 (x) = − 14 (2x) = x2 . Then observe that
 
1
u(x, t) = g x + t + h(x + t)
5
Z x+ 51 t Z x+t
= g 0 (s)ds + h0 (s)ds + g(0) + h(0)
0 0
Z x+ 51 t x+t
Z
5 1
= (2s)ds + − (2s)ds + 02
0 4 0 4
 2
5 1 1
= x + t − 0 − (x + t)2 − (0)
4 5 4
"  2 #
5 1 1
= x + t − (x + t)2 .
4 5 4
Thus, the u(x, t) above is the our solution to the IVP.


Page 30
MATH 534H Editor : Byeongho Ban

Additional Problem 2
Consider the wave equation in 1D with damping
utt = c2 uxx − ku − rut , k, r > 0.
Show that the energy functional
1 ∞
Z
E(t) = |ut |2 + c2 |ux |2 + k|u|2 dx
2 −∞
dE
satisfies dt ≤ 0; that is energy decrease. Assume u and its derivatives vanish as x → ±∞.
Proof. Observe that
 Z 
dE d 1 2 2 2 2
= (ut ) + c (ux ) + k(u) dx
dt dt 2 R
Z
= utt ut + c2 uxt ux + kut udx
R
Z
 2
(c uxx − ku − rut )ut + c2 uxt ux + kut u dx

=
ZR
 2
(c uxx − rut )ut + c2 uxt ux dx,

=
R
2
sine utt = c uxx − ku − rut .
Now, by the integration by part, we have
Z Z Z
∞
uxt ux dx = ut ux −∞ −
 uxx ut dx = − uxx ut dx
R R R
since limx→±∞ ut = 0 and limx→±∞ ux = 0 by our assumption. Thus,
Z
dE  2
(c uxx − rut )ut + c2 uxt ux dx

=
dt
ZR
 2
(c uxx − rut )ut − c2 uxx ut dx

=
R
Z
= −r |ut |2 dx ≤ 0
R
since R |ut |2 dx ≥ 0 and r > 0. And we are done.
R

Page 31
MATH 534H Editor : Byeongho Ban

Proof. 2.2.1
Suppose that u(x, 0) = φ(x) ≡ 0 and ut (x, 0) = ψ(x) ≡ 0. And note that
Z
1
E(t) = (ρu2t + T u2x )dx.
2 R
Then by the conservation of energy(proved in book), dE dt = 0. Thus, E(t) is constant in time t. Then observe that
Z Z Z
1 2 2 1 2 2 1
E(0) = (ρut (x, 0) + T ux (x, 0))dx = (ρψ (x) + T φx (x))dx = (0)dx = 0,
2 R 2 R 2 R
∂ ∂
since φx = ∂x φ = ∂x 0 = 0. Therefore, since E is constant in time t, 0 = E(0) = E(t) for all t. So E ≡ 0. Then by the
vanishing theorem,
Z
1
E(t) = (ρu2t + T u2x )dx ≡ 0 =⇒ ρu2t + T u2x = 0 =⇒ ut = 0 and ux = 0 =⇒ ∇u = (ux , ut ) = 0.
2 R
Thus, u is constant in x and t. Since u(x, 0) = φ(x) = 0, we have u(x, t) ≡ 0 for all x ∈ R and t > 0.

Proof. 2.2.2
(a)
Note that
1
e = (u2t + u2x ) and p = ut ux .
2
Then observe that
∂e
= ut utt + uxt ux
∂t
and that
∂p
= uxt ux + ut uxx ,
∂x
Since utt = uxx , note that
∂e ∂p
= ut utt + uxt ux = ut uxx + uxt ux = .
∂t ∂x

(b)
Observe that
∂
ett = (ut utt + uxt ux ) = u2tt + ut uttt + uxtt ux + u2xt
∂t
and that
∂
exx = (uxt ut + uxx ux ) = u2xt + uxxt ut + uxxx ux + u2xx .
∂x
Since uxx = utt , we have ett = exx so e satisfies the wave equation.

Nextly, observe that

∂
ptt = (utt ux + ut uxt ) = uttt ux + utt uxt + utt uxt + ut uxtt
∂t
and that
∂
pxx = (uxt ux + ut uxx )
∂x
= uxxt ux + uxt uxx + utx uxx + ut uxxx
= uxxt ux + uxt utt + utx utt + ut uxxx (∵ uxx = utt )
= ptt .
Thus, ptt = pxx and so p also satisfies the wave equation.


Page 32
MATH 534H Editor : Byeongho Ban

Proof. 2.2.3
Let our wave equation be utt = c2 uxx and u is the solution.
(a)
Let u(x − y, t) = v(x, t) and observe that
∂2 2 2 ∂
2
vtt (x, t) = u(x − y, t) = utt (x − y, t) = c uxx (x − y, t) = c u(x − y, t) = c2 vxx (x, t).
∂t2 ∂x2
Thus, the wave equation is translation invariant.

(b)
Let w = ux and observe that
∂2 ∂ ∂ 2 ∂2 2
wtt (x, t) = 2 ux (x, t) = uxtt (x, t) = utt (x, t) = (c uxx ) = c ux (x, t) = c2 wxx (x, t).
∂t ∂x ∂x ∂x2
Thus, the wave equation is derivative invariant.

(c)
Let a ∈ R be given and let q(x, t) = u(ax, at). Then observe that
∂2 ∂2
qtt (x, t) =
2
u(ax, at) = a2 utt (ax, at) = a2 c2 uxx (ax, at) = c2 2 u(ax, at) = c2 qxx (x, t).
∂t ∂x
Therefore, the wave equation is dilation invariant.

Additional Problem 3
Let u = u(x, t) be a solution to the wave equation utt − ∆u = 0 in R2 . Assuming that ∇u → 0 fast enough as |x| → ∞
prove that Z
E(t) = |ut |2 + |∇u|2 dxdy
R2
is constant in t for all time t.
Proof. Observe that
Z
dE
=2 [utt ut + ∇ut · ∇u] dx
dt R2
Z
=2 [ut ∆u + ∇ut · ∇u] dx
R2
since utt = ∆u in R2 .
Now, by integration by part,
Z Z Z
∇ut · ∇udx = ut ∇u · dS − ut ∆udx
R2 ∂R2 R2
Z
=− ut ∆udx
R2
since lim|x|→∞ ∇u = 0 and so
Z
ut ∇u · dS = 0.
∂R2
Therefore,
Z 2
dE
=2 [ut ∆u + ∇ut · ∇u] dx
dt R
Z
=2 [ut ∆u − ut ∆u] dx
R2
= 0.
Thus, E(t) is constant in time t for all t.


Page 33
MATH 534H Editor : Byeongho Ban

Due Date : Match 28th,2019 ByeongHo Ban

Proof. 2.3.2

Consider ut = uxx and

M (T ) = max u(x, t) m(T ) = min u(x, t)
ΩT ΩT
where ΩT = {(x, t) : 0 ≤ x ≤ l 0 ≤ t ≤ T }.

(a)
Suppose that T ≤ S, then ΩT ⊆ ΩS which implies that
m(T ) = min u(x, t) ≥ min u(x, t) = m(S)
ΩT ΩS

since if m(T ) = u(x0 , t0 ) then (x0 , t0 ) ∈ ΩT ⊆ ΩS so m(T ) = u(x0 , t0 ) ≤ m(S). Thus, m is non-increasing or decreasing
function of T .

(b)
Suppose that T ≤ S, then ΩT ⊆ ΩS which implies that
M (T ) = max u(x, t) ≤ max u(x, t) = M (S)
ΩT ΩS
0 0 0 0
since if M (T ) = u(x , t ) then (x , t ) ∈ ΩT ⊆ ΩS . Thus, M is non-decreasing or increasing function of T .


Page 34
MATH 534H Editor : Byeongho Ban

Proof. 2.3.4 (Hint for 4(b) : Do not solve explicitly. Rather prove that u(1 − x, t) also solves the equation and then
apply the uniqueness theorem.)

Consider ut = uxx in Ω = {(x, t) : 0 < x < l and 0 < t < ∞} with u(0, t) = u(1, t) = 0 and u(x, 0) = 4x(1 − x).

(a)
Note that the solution of this diffusion equation does not assumes the maximum and the minimum in the interior of the
domain by the stronger version of maximum principle since the solution u is not constant due to u(x, 0) = 4x(1 − x). And
observe that
1
ux (x, 0) = 4(1 − 2x) = 0 ⇐⇒ x = and uxx (x, 0) = −8 < 0.
2
Thus, u(x, 0) takes its maximum, u(1/2, 0) = 1, at 21 , 0 and minimum, u(1, 0) = u(0, 0) = 0, at x = 0 and x = 1. Thus,

u should be bigger than 0 and smaller than 1 in the interior. i.e.0 < u(x, t) < 1 ∀(x, t) ∈ Ω.

(b)
Suppose that u(x, t) is the solution of

ut = uxx

u(x, 0) = 4x(1 − x)

u(0, t) = u(1, t) = 0.


Now, let w(x, t) = u(1 − x, t). Firstly, observe that

wt (x, t) = ut (1 − x, t) = (−1)2 uxx (1 − x, t) = wxx (x, t).
Nextly, observe that
w(x, 0) = u(1 − x, 0) = 4(1 − x)(1 − (1 − x)) = 4(1 − x)x = 4x(1 − x)
w(0, t) = u(1, t) = 0 = u(0, 0) = u(1 − 1, t) = w(1, t).
Thus, w(x, t) = u(1 − x, t) is also the solution of the equation. Since the solution of the equation is unique,
u(x, t) = u(1 − x, t).

(c)
Observe that
1 2
ut = uxx =⇒ (u )t = uut = uuxx
2
1 d 1 2
Z Z 1 Z 1 Z 1
1
=⇒ u dx = uuxx dx = uux 0 − u2x dx = − u2x dx ≤ 0 (∵ u(0, t) = u(1, t) = 0.)
2 dt 0 0 0 0
Assume that
Z 1
u2x dx = 0.
0
Then note that ux = 0 which implies that u is constant in x. Then, since u(0, t) = u(1, t) = 0, u(x, t) = 0 ∀(x, t) ∈ Ω.
However, u(x, 0) = 4x(1 − x) implies discontinuity of u(x, t) in t at t = 0. Thus, it is a contradiction. Therefore,
d 1 2
Z Z 1
u dx = −2 u2x dx < 0
dt 0 0
which implies that u2 dx is strictly decreasing in t.
R


Page 35
MATH 534H Editor : Byeongho Ban

Proof. 2.3.5

(a)
Observe that
ut = (−2xt − x2 )t = −2x = x(−2) = x(−2xt − x2 )xx = xuxx .
Thus, u solves the PDE.
And observe that
ux (x, t) = −2t − 2x = 0 ⇐⇒ x = −t uxx (x, t) = −2 < 0.
Also, note that u(−t, t) = t2 , u(x, 0) = −x2 , u(−2, t) = 4t − 4, and u(2, t) = −4t − 4. Since 0 ≤ t ≤ 1 and −2 ≤ x ≤ 2,
the maximum of u(x, t) in the region is 1 at each (−t, t) ∀t ∈ [0, 1]. Thus, the location of its maximum in the closed
region is {(−t, t) : t ∈ [0, 1]} .

(b)
The result of (a) breaks the maximum principle because maximum does not happen on the boundary since u(x, 0) ≤ 0,
u(−2, t) ≤ 0, and u(2, t) ≤ 0.
In the proof, the problem of variable coefficient of uxx is that the value of the coefficient can be negative number which
says this problem is not well posed.

Proof. 2.3.6
Let w = u − v and Ω = {(x, t) : 0 ≤ x ≤ l 0 ≤ t < ∞}. Then w is the solution of the diffusion equation since
wt = ut − vt = kuxx − kvxx = kwxx .
Then we have the following PDE

wt = kwxx
 (x, t) ∈ Ω
w(0, t) ≤ 0 and w(l, t) ≤ 0 0≤t<∞

w(x, 0) ≤ 0 0 ≤ x ≤ l.


By the maximum principle, w assumes its maximum on the parabolic boundary (P = {(x, t) ∈ Ω : t = 0 or x = 0 or x =
l}) but w ≤ 0 on P by the given condition. Therefore,
w(x, t) ≤ max w(x, t) ≤ 0 ∀(x, t) ∈ Ω.
(x,t)∈Ω

Thus, u − v = w ≤ 0 ∀(x, t) ∈ Ω implies u ≤ v ∀(x, t) ∈ Ω. Then we are done.


Proof. 2.3.7(a)
Let w = u − v and Ω = {(x, t) : 0 ≤ x ≤ l 0 ≤ t < ∞}. Then observe that
wt − kwxx = (ut − kuxx ) − (vt − kvxx ) = f − g ≤ 0.
Thus, by the maximum principle in the class, w takes maximum on its parabolic boundary (P = {(x, t) ∈ Ω : t = 0 or x =
0 or x = l}). However, w = u − v ≤ 0 on P by the given condition. Therefore, w(x, t) = u(x, t) − v(x, t) ≤ 0 ∀(x, t) ∈ Ω
which is saying u ≤ u on Ω.


Page 36
MATH 534H Editor : Byeongho Ban

Proof. 2.4.3

Note that Z ∞
1 2
u(x, t) = √ e−(x−y) /4kt
φ(y)dy
4πkt −∞
which is given as (8) is the solution to the
(
ut = kuxx −∞ < x < ∞, 0 < t < ∞
u(x, 0) = φ(x).
Since φ(x) = e3x as given, our solution to the equation should be
Z ∞
1 2
u(x, t) = √ e−(x−y) /4kt φ(y)dy
4πkt −∞
Z ∞
1 2
=√ e−(x−y) /4kt e3y dy
4πkt −∞
Z ∞
1 2
=√ e−z /4kt e3(x+z) dz (z = y − x)
4πkt −∞
Z ∞
1 2
=√ e3x e(−z +12ktz)/4kt dz
4πkt −∞
Z ∞
1 3x 2 2 2
=√ e e[−(z−6kt) +36k t ]/4kt dz
4πkt −∞
Z ∞
1 3x+9kt 2
=√ e e[−(z−6kt) ]/4kt dz
4πkt −∞
1 3x+9kt ∞ −u2
 
z − 6kt
Z
=√ e e du u= √
π −∞ 4kt
Z ∞
1 √ 2 √
= √ e3x+9kt π ∵ e−u du = π
π −∞
= e3x+9kt .
R∞ 2
As for the last equality, let’s calculate α = 0 e−x dx. Observe that, by using polar coordinate,
Z ∞  Z ∞ 
2 −x2 −y 2
α = e dx e dy
0 0
Z ∞ Z ∞
2 2
= e−(x +y ) dxdy
0 0
1 2π ∞ −r2
Z Z
= e rdrdθ
4 0
Z ∞0
2π 2
= e−r rdr
4 0
π ∞ −z
Z
= e dz (r2 = z)
4 0
πh i π
= lim (1 − e−t ) = .
4 t→∞ 4
√ R ∞ −x2 √
π −x2
Therefore, α = 2 and −∞ e dx = π since e is an even function.
As a consequence, our solution is
u(x, t) = e3x+9kt .


Page 37
MATH 534H Editor : Byeongho Ban

Proof. 2.4.5(do only (a),(b),(c))

(a) Let u(x, t) be the solution of diffusion equation ut = kuxx and let y be given constant. And let v(x, t) = u(x − y, t).
Then observe that
vt (x, t) = ut (x − y, t) = kuxx (x − y, t) = kvxx (x, t).
Thus, v(x, t) = u(x − y, t) is another solution.

(b)
Let u be the solution of ut = kuxx . Then observe that
(ut )t = (kuxx )t = kuxxt = k(ut )xx
(ux )t = (ut )x = (kuxx )x = kuxxx = k(ux )xx
(uxx )t = (ut )xx = (kuxx )xx = kuxxxx = k(uxx )xx .
Thus, ut , ux and uxx are also the solution to the diffusion equation.

(c)
Let a, b be given constant and u and v be two solutions to the diffusion equation ut = kuxx . Letting w = au + bv, observe
that
wt = (au + bv)t = aut + bvt = akuxx + bkvxx = k(au + bv)xx = kwxx .
Thus, w is again a solution to the equation. Therefore, any linear combination of solutions to the equation is also a
solution.

Proof. 2.4.8

We should prove that, for any δ > 0,

lim max S(x, t) = 0
t→0 δ≤|x|<∞

where
1 x2
S(x, t) = √ e− 4kt .
2 πkt
Let δ > 0 be given. And observe that
1  x  − x2
Sx (x, t) = √ − e 4kt .
2 πkt 2kt
Note that Sx < 0 when x > 0 and Sx > 0 when x < 0 . Thus,
(
is decreasing if x > 0
S(x, t) =
is increasing if x < 0
having maximum √1 at x = 0. Thus,
2 πkt
1 δ2
max S(x, t) = S(δ, t) = √ e− 4kt .
δ≤|x|<∞ 2 πkt
Now, by L’Hospital’s rule,
1 δ2
lim max S(x, t) = lim √ e− 4kt
t→0 δ≤|x|<∞ t→0 2 πkt
√
1/2 πkt
= lim 2
t→0 exp δ
4kt
√ 3
(1/2 πk)t− 2
= lim 2
t→0 (δ 2 /2k)t−2 exp δ
4kt
√
(1/2 πk) δ2 √
= 2
lim e− 4kt t = 0.
(δ /2k) t→0
Thus, we are done.


Page 38
MATH 534H Editor : Byeongho Ban

Due Date : April 9th ByeongHo Ban

Proof. 2.4.2

Note that the solution is

Z ∞
1 (x−y)2
u(x, t) = √ e− 4kt φ(y)dy
4πkt −∞
where
(
1 x>0
φ(x) =
3 x < 0.
Then observe that
Z ∞ Z 0
1 (x−y)2
− 4kt 3 (x−y)2
u(x, t) = √ e dy + √ e− 4kt dy
4πkt 0 4πkt −∞
   
1 1 x 3 3 x
= + erf √ + − erf √
2 2 4kt 2 2 4kt
 
x
= 2 − erf √
4kt
since
∞ ∞
y−x
Z Z
1 (x−y)2 1 2
√ e− 4kt dy = √ e−p dp ∵p= √
4πkt 0 π − √x 4kt
4kt
!
Z ∞ Z 0
1 2 2
=√ e−p dp + e−p dp
π 0 − √x
4kt
 
1 1 x 2
= + erf √ ∵ e−p is even
2 2 4kt
and since

0 − √x
y−x
Z Z
1 (x−y)2 1 4kt 2
√ e− 4kt dy = √ e−p dp ∵p= √
4πkt −∞ π −∞ 4kt
!
Z 0 Z 0
1 −p2 −p2
=√ e dp − e dp
π −∞ − √x
4kt
!
Z ∞ Z √x
1 2 4kt 2 2
=√ e−p dp − e−p dp ∵ e−p is even
π 0 0
 
1 1 x
= − erf √
2 2 4kt
Rx 2
where erf (x) = √2
π 0
e−p dp.

Note that
   (
x 2 − erf (∞) = 1 x>0
u(x, 0) = lim+ 2 − erf √ =
t→0 4kt 2 − erf (−∞) = 3 x < 0
since
  Z √x Z ∞
x 2 4kt 2 2 2
lim+ erf √ = lim+ √ e−p dp = √ e−p dp = 1
t→0 4kt t→0 π 0 π 0
−|x|
  √ −∞ 0
−|x|
Z Z Z
2 4kt 2 2 2 2 2
lim erf √ = lim √ e−p dp = √ e−p dp = − √ e−p dp = −1.
t→0+ 4kt t→0+ π 0 π 0 π −∞
Thus, u(x, t) is the solution to the IVP. 

Page 39
MATH 534H Editor : Byeongho Ban

Proof. 2.4.4

Note that
Z ∞
1 (x−y)2
u(x, t) = √ e− 4kt φ(y)dy
4πkt −∞
where
(
e−x x>0
φ(x) =
0 x < 0.
Then observe that
Z ∞
1 (x−y)2
u(x, t) = √ e− 4kt e−y dy
4πkt 0
Z ∞
1 x2 −2xy+y 2 +4kty
=√ e− 4kt dy
4πkt 0
Z ∞
1 (y−(x−2kt))2 −(x−2kt)2 +x2
=√ e− 4kt dy
4πkt 0
Z ∞
1 −4ktx+4k2 t2 (y−(x+2kt))2
=√ e 4kt e− 4kt dy
4πkt 0
Z ∞
1 (y−(x−2kt))2
=√ e−x+kt e− 4kt dy
4πkt 0
  
1 1 x − 2kt
= e−x+kt + erf √
2 2 4kt
is the solution to the IVP. To ensure that is satisfies the initial condition, observe that
  
1 1 x − 2kt
u(x, 0) = e−x + lim+ erf √
2 2 t→0 4kt
  
1 1 x − 2kt
= e−x + erf lim √
2 2 t→0+ 4kt
  
1 1 x 2kt
= e−x + erf lim+ √ ∵ lim+ √ =0
2 2 t→0 4kt t→0 4kt
(
e−x 12 + 12 erf (∞) = e−x x > 0


=
e−x 21 + 12 erf (−∞) = 0 x < 0.

Thus, we are done.



Page 40
MATH 534H Editor : Byeongho Ban

Proof. 2.4.9

We consider IVP, with k > 0,

(
ut = kuxx x ∈ R t > 0,
(†)
u(x, 0) = x2 x ∈ R.
Firstly, observe that, since u solves the IVP,
(uxxx )t = uxxxt = utxxx = (ut )xxx = (kuxx )xxx = kuxxxxx k(uxxx )xx
and that
uxxx (x, 0) = (x2 )xxx = 0.
Thus, uxxx satisfies an IVP (††) with initial condition uxxx (x, 0) = 0. Now, note that w ≡ 0 on R × (0, ∞) solves (††).
Since the solution to diffusion/heat IVP is unique, uxxx ≡ 0 in R × (0, ∞). By integrating uxxx three times with respect
to x, we get
u(x, t) = A(t)x2 + B(t)x + C(t).
Then by plugging it to (†), we get
ut − kuxx = (A0 (t)x2 + B 0 (t)x + C 0 (t)) − k(2A(t)) = A0 (t)x2 + B 0 (t)x + (C 0 (t) − 2kA(t)) = 0.
Since 1, x, x2 are linearly independent, we get

0
A (t) = 0

B 0 (t) = 0 († † †).
 0

C (t) − 2kA(t) = 0
Also, by the initial condition,

A(0) = 1

2 2
x = u(x, 0) = A(0)x + B(0)x + C(0) =⇒ B(0) = 0 († † ††)

C(0) = 0.


Then by († † ††) and († † †),

A0 (t) = 0 =⇒ A(t) = const =⇒ A(t) = 1 ∵ A(0) = 1
=⇒ B 0 (t) = 0 =⇒ B(t) = const =⇒ B(t) = 0 ∵ B(0) = 0
=⇒ C 0 (t) = 2k =⇒ C(t) = 2kt + const =⇒ C(t) = 2kt ∵ C(0) = 0.
Therefore,
u(x, t) = x2 + 2kt.

Proof. 2.4.11(a)(b)

(a)
Let u(x, t) be the solution to the IVP. Then consider w(x, t) = u(x, t) + u(−x, t). Then observe that
(
wt (x, t) = ut (x, t) + ut (−x, t) = kuxx (x, t) + kuxx (−x, t) = k(u(x, t) + u(−x, t))xx = kwxx (x, t)
(†)
w(x, 0) = u(x, t) + u(−x, 0) = φ(x) + φ(−x) = φ(x) − φ(x) = 0 ∵ φ(−x) = −φ(x).
Then w is the solution to (†). Also, note that q ≡ 0 in R × (0, ∞) also solves (†). Then by the uniqueness of diffusion
IVP, u(x, t) + u(−x, t) = 0 ∀(x, t) ∈ R × (0, ∞). Thus, u(x, t) = −u(−x, t) ∀(x, t) ∈ R × (0, ∞) so u(x, t) is odd function
in R × (0, ∞).

(b)
Now, let’s suppose that φ is even, i.e. φ(x) = φ(−x). Let u(x, t) be the solution to the IVP. Then consider w(x, t) =
u(x, t) − u(−x, t). Then observe that
(
wt (x, t) = ut (x, t) − ut (−x, t) = kuxx (x, t) − kuxx (−x, t) = k(u(x, t) − u(−x, t))xx = kwxx (x, t)
(†)
w(x, 0) = u(x, t) − u(−x, 0) = φ(x) − φ(−x) = φ(x) − φ(x) = 0 ∵ φ(−x) = φ(x).
Then w is the solution to (†). Also, note that q ≡ 0 in R × (0, ∞) also solves (†). Then by the uniqueness of diffusion
IVP, u(x, t) − u(−x, t) = 0 ∀(x, t) ∈ R × (0, ∞). Thus, u(x, t) = u(−x, t) ∀(x, t) ∈ R × (0, ∞) so u(x, t) is even function
in R × (0, ∞).


Page 41
MATH 534H Editor : Byeongho Ban

Proof. 2.4.15

Suppose that there are two solution u and v to the Neumann boundary condition. And let’s consider w = u − v. Then
observe that


 wt − kwxx = (u − v)t − k(u − v)xx = (ut − kuxx ) − (vt − kvxx ) = f − f = 0 for 0 < x < l, t > 0,

w(x, 0) = u(x, 0) − v(x, 0) = φ(x) − φ(x) = 0,
(†)


 wx (0, t) = ux (0, t) − vx (0, t) = g(t) − g(t) = 0
wx (l, t) = ux (l, t) − vx (l, t) = h(t) − h(t) = 0.


Let’s define
Z l
E[w](t) = |w(x, t)|2 dx.
0
Observe that E[w](t) is a non-negative function since |w(x, t)|2 ≥ 0 ∀(x, t).
Then observe that
Z l
d
E[w](t) = wt (x, t)w(x, t)dx
dt 0
Z l
= kwxx (x, t)w(x, t)dx
0
Z l
= kwx (l, t)w(l, t) − kwx (0, t)w(0, t) − k |wx (x, t)|2 dx ∵ The integration by parts
0
Z l
= −k |wx (x, t)|2 dx ∵ (†)
0
≤0 ∵ k > 0 and |wx (x, t)|2 ≥ 0 ∀(x, t).
Thus, E[w](t) is a decreasing function in t so 0 ≤ E[w](t) ≤ E[w](0). Now, observe that
Z l
E[w](0) = |w(x, 0)|2 dx = 0 ∵ (†).
0
Thus, E[w](t) = 0 ∀t > 0. It implies that
Z l
|w(x, t)|2 dx = 0 =⇒ |w(x, t)|2 = 0 =⇒ w(x, t) = 0 ∀(x, t).
0
Therefore, 0 ≡ w ≡ u − v so u ≡ v and the solution is unique.

Proof. 2.4.16

Let v(x, t) = ebt u(x, t). Then observe that

(
0 = ut − kuxx + bu = (e−bt v)t − k(e−bt v)xx + b(e−bt v) = −be−bt v + e−bt vt − ke−bt vxx + be−bt v = e−bt (vt − kvxx )
v(x, 0) = e0 u(x, 0) = u(x, 0) = φ(x).
Since b > 0, vt − kvxx = 0. Then
Z
1 (x−y)2
v(x, t) = √ e− 4kt φ(y)dy.
4πkt R
Then we have
e−bt
Z
(x−y)2
−bt
u(x, t) = e v(x, t) = √ e− 4kt φ(y)dy.
4πkt R


Page 42
MATH 534H Editor : Byeongho Ban

Additional Problem 1
a) Show that the function u(x, t) = e−kt sin(x) solves the heat equation ut − kuxx = 0.

b) Find a relationship between the constants a and b so that u(x, t) = e−at cos(bx) is a solution to ut − kuxx = 0(assume
cos(bx) 6= 0.)
Proof. .

(a)
Observe that
ut (x, t) − kuxx (x, t) = (e−kt sin(x))t − k(e−kt sin(x))xx
= −e−kt sin(x) − k(e−kt (− sin(x)))
= 0.
−kt
Thus, u(x, t) = e sin(x) solves the heat equation.

(b)
Observe that
0 = ut − kuxx
= (e−at cos(bx))t − k(e−at cos(bx))xx
= −ae−at cos(bx) − k(e−at (−b2 cos(bx)))
= (−a + kb2 )e−at cos(bx).
Note that, if a = 0, u(x, t) = cos(bx) and
ut − kuxx = 0 − k(−b2 ) cos(bx) 6= 0
so it does not solve the diffusion equation so it should be a 6= 0.
From here, since cos(bx) 6= 0 and e−at 6= 0, kb2 = a.


Page 43
MATH 534H Editor : Byeongho Ban

Additional Problem 2
This problem pertains the heat equation on the whole real line. Recall Gauss’ error function
Z x
2 2
erf(x) := √ e−z dz.
π 0
 
a) Prove that all solutions to the equation ut − kuxx = 0 of the form u(x, t) = v √xt , for x ∈ R and t > 0 have the form
 
x
(†) u(x, t) = C1 + C2 erf √
4kt
b) Prove that by choosing C2 suitably in (†) the fundamental solution Γk (x, t) = ux (x, t).
Proof. (a)

Let y = √xt .
Observe that
 2
∂y 0 ∂y
ut − kuxx = v −k v 00
∂t ∂x
x 1
= − √ v 0 − k v 00
2t t t
 
1 x 0 00
=− √ v + kv .
t 2 t
Since t > 0, we get an ODE
x 0 y 0
v 00 + √ v = v 00 + v = 0.
2k t 2k
Let w = v 0 then
y 0 y
v 00 = − v =⇒ w0 = − w
2k 2k
w0 y
=⇒ =−
w 2k
y2
=⇒ ln |w| = − + const
4k
2
=⇒ w(y) = C2 e−y /4k .
Further observe that
2 2
w(y) = C2 e−y /4k
=⇒ v 0 = C2 e−y /4k
Z y
2
=⇒ v(y) = C1 + C2 e−p /4k
dp
0
√
√ Z y/ 4k
2 √
=⇒ v(y) = C1 + C2 4k e−z dz ∵p= 4kz
0
√  
4kπ y
=⇒ v(y) = C1 + C2 erf √
2 4k
√
 
y
=⇒ v(y) = C1 + C2 erf √ ∵ C2 = kπC2 (redefine).
4k
Then, lastly, we observe that
   
x x
u(x, t) = v √ = C1 + C2 erf √
t 4kt
so we are done.

(b)
Let C2 = √1 then observe that
π
   Z √x
∂ x ∂ 4kt 2 1 − x2 1 |x|2
ux (x, t) = C1 + C2 erf √ = C2 e−z dz = C2 √ e 4kt = √ e− 4kt = Γk (x, t).
∂x 4kt ∂x 0 4kt 4πkt
√
Therefore, if C2 = 1/ π then Γk = ux .


Page 44
MATH 534H Editor : Byeongho Ban

Additional Problem 3
Consider the Cauchy IVP for the heat equation:
(
ut − kuxx = 0, x ∈ R, t > 0
(††)
u(x, 0) = H(x), x∈R
where H(x) is the Heaviside function, H(x) = 0 if x < 0 and H(x) = 1 for x ≥ 0.
Using the general formula given the additional problem 3a) above, in find the constants C1 and C2 so that u(x, t) also
satisfies the initial condition and hence is a solution to (††). In other words find C1 and C2 so that for x < 0
 
x
0 = u(x, 0+ ) = lim C1 + C2 erf √
t→0+ 4kt
as for x > 0,
 
+ x
1 = u(x, 0 ) = lim+ C1 + C2 erf √ .
t→0 2kt
Note in the first case get an integral between 0 and −∞ while on the second you get an integral between 0 and ∞.
After finding C1 and C2 write down the final expression for u(x, t) solving (††) in terms of the erf function and the explicit
constants C1 and C2 .
Proof. From Additional Problem 2, we have
 
x
u(x, t) = C1 + C2 erf √ .
4kt
We will choose C1 and C2 so that u satisfies the given initial condition u(x, 0) = H(x). Observe that
(
1 x > 0,
lim u(x, t) = u(x, 0+ ) =
t→0+ 0 x < 0.
Then, when x > 0,
   Z ∞ √
x 2 C2 π
1 = u(x, 0+ ) = lim C1 + C2 erf √ = C1 + C2 e−z dz = C1 +
t→0+ 4kt 0 2
and when x < 0,
   Z −∞ Z 0 √
x 2 2 C2 π
0 = u(x, 0+ ) = lim C1 + C2 erf √ = C1 + C2 e−z dz = C1 − C2 e−z dz = C1 −
t→0+ 4kt 0 −∞ 2
2
since e−z is even so since
Z 0 Z ∞
−z 2 2
e dz = e−z dz.
−∞ 0
By adding first and second equations,
1
1 = 2C1 =⇒ C1 = .
2
By plugging it back to first equation we get
√
1 C2 π 1
1= + =⇒ C2 = √ .
2 2 π
Therefore, our solution to this Cauchy IVP heat equation is
 
1 1 x
u(x, t) = + √ erf √ .
2 π 4kt


Page 45
MATH 534H Editor : Byeongho Ban

Proof. 2.5.1
Hint: Consider the example in homework 5 section 2.1 (the hammer blow) and note that since there is no boundary, if
there was a maximum principle it would assert that it has to be attained initially. But...

Proof start
Let a > 0 be given.
Consider the WAVE IVP

2
utt − c uxx = 0
 (x, t) ∈ R × (0, ∞)
u(x, 0) = φ(x) = 0

ut (x, 0) = ψ(x)


where
(
1 x ≤ |a|
ψ(x) = χ[−a,a] (x) =
0 x > |a|.
Then, due to De’Alambert, the solution to the wave IVP is
Z x+ct
1 1 1
u(x, t) = [0 + 0] + χ[−a,a] dx = m([−a, a] ∩ [x − ct, x + ct])
2 2c x−ct 2c
where m is Lebesgue measure (measuring length of the interval).

Assume that the maximum principle holds for this Cauchy wave equation. Then the maximum value of u(x, t) should be
assumes either initially or on the lateral side. But note that there is no lateral side of R × (0, ∞). Therefore, the maximum
value should be assumed initially (when t = 0). However, when t = 0,
1 1 1
u(x, 0) = m([−a, a] ∩ [x − ct, x + ct]) = m([−a, a] ∩ [x, x]) ≤ m([x, x]) = 0.
2c 2c 2c
Thus, the maximum value of u should be 0. Now, let x0 ∈ R be given. And choose t0 ∈ (0, ∞) so that [−a, a] ⊆
[x0 − ct0 , x0 + ct0 ]. Then observe that
1 1 2a a
u(x0 , t0 ) = m([−a, a] ∩ [x0 − ct0 , x0 + ct + 0]) = m([−a, a]) = = > 0.
2c 2c 2c c
It shows that u assumes a value bigger than 0 not initially. Therefore, for this problem the maximum principle does not
hold.


Page 46
MATH 534H Editor : Byeongho Ban

Due Date: April 23rd ByeongHo Ban

Proof. 4.1.2

Our equation is heat equation which is


ut − kuxx = 0
 0 < x < l, 0 < t < ∞
u(0, t) = 0 = u(l, t)

u(x, 0) = 1.


Then our solution would be

∞
2 nπx
An e−( )
X nπ
kt
u(x, t) = l sin .
n=1
l
We only need to specify the constants An .
Note that, by the initial condition, we have
∞
X nπx
1 = u(x, 0) = An sin .
n=1
l
Then by the given equality
 
4 πx 1 3πx 1 5πx
1= sin + sin + sin + ··· ,
π l 3 l 5 l
we can see
∞  
X nπx 4 πx 1 3πx 1 5πx
An sin =1= sin + sin + sin + ··· .
n=1
l π l 3 l 5 l
Therefore,
(
4 1
πn n : odd
An =
0 Otherwise.

Proof. 4.1.3

Let’s guess that u(x, t) = X(x)T (t). Then observe that

X 00 T0
0 = ut − iuxx = XT 0 − iX 00 T =⇒ =
X iT
Since each sides depend on only x and t respectively, it should be
X 00 T0
= = −λ =⇒ X 00 = −λX and T 0 = −iλT.
X iT
for some positive constant λ.

Recall that we have Dirichlet boundary condition.

Note that we have same condition for X with the situation we had in wave equation. Therefore, by using (8), we have
 nπ 2
λn = Xn = A en sin nπx .
l l
And note that the solution for T is
nπ 2
T (t) = Ce−i( l ) t .
Then, by setting An = A
en C, our solution representation would be
∞
X nπx −i( nπ 2
l ) t.
u(x, t) = An sin e
n=1
l


Page 47
MATH 534H Editor : Byeongho Ban

Additional Problem 1 Separate variables to solve utt − uxx = 0 in 0 < x < 3 with boundary conditions u(0, t) =
u(3, t) = 0.
Proof. Let’s guess that the solution is form of u(x, t) = X(x)T (t). Then observe that
X 00 T 00
0 = utt − uxx = XT 00 − X 00 T =⇒ = .
X T
Since left and right hand sides depend on different variables, it should be
X 00 T 00
= = λ =⇒ X 00 = −λX and T 00 = −λT
X T
for some positive constant λ.

Now, since λ > 0, we can set λ = α2 we get

X(x) = A sin αx + B cos αx.
Then, by the boundary condition,
0 = X(0) = B =⇒ B = 0
nπ
0 = X(3) = A sin 3α =⇒ α =
3
for some n ∈ Z+ .
Let Xn (x) = An sin nπx
3 . Also, note that
nπt nπt
T (t) = C sin + D cos .
3 3
Then finally we have
∞  
X nπt nπt nπx
u(x, t) = X(x)T (t) = Cn sin + Dn cos sin
n=1
3 3 3
where we reset Cn = An C and Dn = An D.

Additional Problem 2.
Separate variables to solve ut = uxx in 0 < x < π with boundary conditions u(0, t) = u(π, t) = 0
Proof. Let’s guess that u(x, t) = X(x)T (t). Then observe that
X 00 T0
0 = ut − uxx = XT 0 − X 00 T =⇒ = .
X T
Since each sides depend on different variables, it should be
X 00 T0
= = −λ =⇒ X 00 = −λX and T 0 = −λT.
X T
By the result from class, λ = β 2 for some β ∈ R. Then note that
X(x) = A sin βx + B cos βx
2
T (t) = Ce−β t .
Then by the boundary condition we should have
0 = u(0, t) = X(0)T (t) ∀t =⇒ 0 = X(0) = B
0 = u(π, t) = X(π)T (t) ∀t =⇒ 0 = X(π) = A sin βπ.
Then it must be βπ = nπ for some n ∈ Z+ so β = n. Therefore,
2
Xn (x) := X(x) = An sin nx T (t) = Ce−n t .
Setting CAn = Bn , we get
∞
X 2
u(x, t) = Bn e−n t sin nx.
n=1


Page 48
MATH 534H Editor : Byeongho Ban

Additional Problems 3

Let g be a smooth function on [0, 1], g(0) = g(1) = 0. Consider the Dirichlet BIVP:
ut − uxx + 3u = 0 in 0 < x < 1 with u(0, t) = u(1, t) = 0 and u(x, 0) = g(x) (⊕)

a) Consider the change of variables u(x, t) = e−3t v(x, t) and prove that v solves vt − vxx = 0 in 0 < x < 1 with
v(0, t) = v(1, t) = 0 and v(x, 0) = g(x). (†)

b) Use the method of separation of variables to find v, the solution to (†).

c) Use a) and b) to find u, the solution to (⊕).

Proof. a)
Note that v(x, t) = e3t u(x, t) and observe that
vt (x, t) − vxx (x, t) = 3e3t u(x, t) + e3t ut (x, t) − e3t uxx (x, t) = e3t [3u(x, t) + ut (x, t) − uxx (x, t)] = 0
since u satisfies the given equation ut − uxx + 3u = 0 in x ∈ (0, 1). And also observe that
v(0, t) = e3t u(0, t) = 0 = e3t u(1, t) = v(x, 1)
since u(0, t) = 0 = u(1, t) ∀t and also observe that
v(x, 0) = e0 u(x, 0) = u(x, 0) = g(x).
Thus, v satisfies all the condition so we are done.

b)
Let v(x, t) = X(t)T (t). Then observe that
X 00 T0
0 = vt − vxx = XT 0 − X 00 T =⇒ = .
X T
Since each sides depend on different variables, the fraction should be a constant. Thus,
( (
X 00 T0 2 X 00 = −β 2 X X(x) = A sin βx + B cos βx
= = −β =⇒ =⇒ 2
X T T 0 = −β 2 T T (t) = Ce−β t .
for some β ∈ R from the class. Since it satisfies zero Dirichlet condition, we have
0 = X(0) = B =⇒ B = 0
0 = X(1) = A sin β =⇒ β = nπ
for some n ∈ Z. Letting A := An according to the n A depends, we have
∞
X 2 2
v(x, t) = [An sin nπx] Ce−n π t
n=1
Letting An := An C for convenience, we have
∞
2
π2 t
X
v(x, t) = An e−n sin nπx.
n=1
Finally, by applying initial condition,
∞
X
g(x) = v(x, 0) = An sin(nπx)
n=1

c)
Note that u(x, t) = e−3t v(x, t), then by b), we have
∞ ∞
X 2 2 X 2 2
u(x, t) = e−3t v(x, t) = e−3t An e−n π t sin nπx = An e−(n π +3)t sin nπx
n=1 n=1
with An specified in b).


Page 49
MATH 534H Editor : Byeongho Ban

Additional Problem 4
Separate variables to solve the following problem for the heat equation with mixed boundary conditions : ut − 5uxx = 0
in 0 < x < L with boundary conditions ux (0, t) = u(L, t) = 0.
Proof. Let’s guess that u(x, t) = X(x)T (t). Then observe that
X 00 T0
0 = ut − 5uxx = XT 0 − 5X 00 T =⇒ = .
X 5T
Since each side depend on different variables, it should be
( (
X 00 T0 X 00 + λX = 0 X(x) = A sin βx + B cos βx
= = −λ =⇒ =⇒ 2
X 5T T 0 + 5λT = 0 T (t) = Ce−5β t
Here, we suppose λ = β 2 > 0 for some β ∈ R. I will prove it at the end.

Then by the boundary condition,

0 = ux (0, t) = X 0 (0)T (t) ∀t =⇒ 0 = X 0 (0) = Aβ =⇒ A = 0
0 = u(L, t) = X(L)T (t) ∀t =⇒ 0 = X(L) = B cos βL
for some constants A, B, C, since if β = 0, then X(t) = B and T (t) = C so by u(L, t) = 0, B = 0 or C = 0 but then
u = XT = 0.
Then, since B 6= 0 by same reason,
(2n − 1)π
B cos βL = 0 =⇒ β = , n ∈ Z.
2L
Then with
∞
X (2n − 1)πx
X(x) = Bn cos ,
n=1
2L
we have

∞
X 5(2n−1)2 π 2 t (2n − 1)πx
u(x, t) = X(x)T (t) = Bn e− 4L2 cos
n=1
2L
resetting Bn := CBn .

Now, suppose λ = 0, then

(
X(x) = Ax + B
T (t) = C
for some constants A, B and C. Then by the boundary condition,
(
ux (0, t) = 0 =⇒ X 0 (x) = A = 0
u(L, t) = 0 =⇒ X(L) = AL + B = 0 =⇒ B = 0
which implies X(x) = 0 so u(x, t) = X(x)T (t) = 0.

Suppose λ < 0, then, by setting λ = −β 2

(
X(x) = Aeβx + Be−βx
2
T (t) = Ceβ t .
then
ux (0, t) = 0 =⇒ X 0 (0) = Aβeβx − Bβe−βx = 0 =⇒ A = B
u(L, t) = 0 =⇒ A(eβL + e−βL ) = 0 =⇒ A = 0
since if β = 0 then X(x) = A + B and T (t) = C so u(x, t) = (A + B)C but by u(L, t) = 0, u ≡ 0 and since eβL + e−βL > 0.
Then we have X(x) = 0 so u ≡ 0.

Also, it cannot be a complex number since then by same procedure we can consider λ = −γ 2 where γ is a square root of
−λ and can have same result with when λ < 0. Therefore, λ > 0.


Page 50
MATH 534H Editor : Byeongho Ban

Proof. 5.1.9
(Hint : Use the trig. identities cos(2x) = cos2 (x) − sin2 (x) = 2 cos2 (x) − 1 to re-express the initial velocity and
immediately obtain its cosine ”expansion.”)

Recall the solution to the wave equation with Neumann boundary condition is represented by
∞  
1 1 X nπct nπct nπx
u(x, t) = A0 + B0 t + An cos + Bn sin cos .
2 2 n=1
l l l
Since, in this case, l = π, we have
∞
1 1 X
u(x, t) = A0 + B0 t + (An cos nct + Bn sin nct) cos nx.
2 2 n=1
Note that we have the initial condition
u(x, 0) = 0 ut (x, 0) = cos2 x.
Then observe that
∞
1 X
0 = u(x, 0) = A0 + (An ) cos ncx =⇒ An = 0 ∀n ∈ Z+ ∪ {0}
2 n=1

1 cos 2x 1
∞
X 1
 n=0
+ = cos2 x = ut (x, 0) = B0 + (Bn nc) cos nx =⇒ Bn = 4c1
n=2
2 2 2 n=1

0 , Otherwise.


Then our solution would be

1 1
u(x, t) = t+ sin 2ct cos 2x.
2 4c

Proof. 5.2.3

Let f and g are odd and even function respectively. Then for some given l ∈ R, observe that
Z l Z l Z 0
f (x)dx = f (x)dx + f (x)dx
−l 0 −l
Z l Z −l
= f (x)dx − f (x)dx
0 0
Z l Z −l
= f (x)dx − (−f (−x))(−d(−x)) ∵ f is odd
0 0
Z l Z l
= f (x)dx − f (y)d(y) change of variables y = x
0 0
= 0.
And also, observe that
Z l Z l Z 0
g(x)dx = g(x)dx + g(x)dx
−l 0 −l
Z l Z −l
= g(x)dx − g(x)dx
0 0
Z l Z −l
= g(x)dx − g(−x)(−d(−x)) ∵ g is even
0 0
Z l Z l
= g(x)dx + g(y)d(y) change of variables y = x
0 0
Z l
=2 g(x)dx.
0


Page 51
MATH 534H Editor : Byeongho Ban

Proof. 5.2.10

(a)
In order for φ to have continuous odd extension, we should have
lim φ(x) = lim −φ(−x) = lim −φ(x).
x→0+ x→0− x→0+
In particular, since
lim −φ(−x) = lim+ −φ(x),
x→0− x→0
we should have limx→0+ φ(x) = 0. Thus, if it is continuous, it must be φ(0) = 0.

(b)
Let ψ be the odd extension of φ. In order for ψ to be differentiable we should have continuity and, since −φ(−x) is
continuous in (−l, 0), it suffices to have
φ() − φ(0) ψ() − ψ(0) ψ(0) − ψ() φ(0) + φ(−)
lim = lim = lim = lim .
→0+  →0 +  →0 − − →0 − −
Then from (a), we must have
φ() φ(−)
lim = lim− .
→0+  →0 −
However, by changing  → 0− to  → 0+ , we have
φ() φ(−) φ()
lim+ = lim− = lim+
→0  →0 − →0 
so it is always true. Therefore, if left derivative of φ at x = 0 exits, then it is automatically differentiable.

(c)
Let ψ is the even extension of φ. In order to have continuity at x = 0, we should have
lim φ(x) = lim− φ(−x).
x→0+ x→0

But note that LHS and RHS are always equal after we change x → 0− to x → 0+ . Thus, once we have left limit of φ(x)
at x = 0 exits, we automatically have continuity of φ(x).

(d)
Similarly, we should investigate the differentiability of even extension at x = 0 so we must have
φ() − φ(0) ψ() − ψ(0) ψ(0) − ψ() φ(0) − φ(−) φ(0) − φ()
lim = lim+ = lim− = lim− = lim+ .
→0+  →0  →0 − →0 − →0 
Thus, we have
φ() − φ(0) φ(0) − φ() φ() − φ(0)
lim+ = lim+ = − lim+ .
→0  →0  →0 
Thus, in order ψ to be differenttiable in (−l, l), the left derivative of φ at x = 0 should exist and its value should be 0.


Page 52
MATH 534H Editor : Byeongho Ban

Additional Problem 5
Let φ : R → R be a periodic function with period p; that is φ(x + p) = φ(x), ∀x ∈ R. Assume that φ is integrable on any
finite interval.

(a) Prove that for any a, b ∈ R

Z b Z b+p Z b−p
φ(x)dx = φ(d)dx = φ(x)dx
a a+p a−p

(b) Prove that for any a ∈ R

Z p/2 Z p/2+a Z p/2
φ(x + a)dx = φ(x)dx = φ(x)dx
−p/2 −p/2+a −p/2
Note then that for any a ∈ R,
Z p/2 Z p/2
φ(x + a)dx = φ(x)dx
−p/2 −p/2
R a+p
and thus in particular that a
φ(x)dx does not depend on a, as we discussed in class(section 5.2, Strauss).
Proof. Claim
Firstly, we claim that
Z α Z p Z α+p
φ(x)dx = φ(x)dx = φ(x)dx ∀α ∈ R.
α−p 0 α

Since φ(x) = φ(x + p) ∀x ∈ R, it suffices to show it only when α ∈ (0, 1). Now, let α ∈ (0, 1) be given. Then observe that
Z α+p Z p Z α+p
φ(x)dx = φ(x)dx + φ(x)dx
α α p
Z p Z α
= φ(x)dx + φ(y + p)dy ∵ by the change of variable y = x − p
Zαp Z0 α
= φ(x)dx + φ(x)dx φ(x) = φ(x + p)
Zαp 0

= φ(x)dx.
0
Similarly,
Z α Z p Z α
φ(x)dx = φ(x)dx + φ(x)dx
α−p α−p p
Z p Z α−p
= φ(x)dx + φ(y + p)dy ∵ by the change of variables y = x − p
α−p 0
Z p Z α−p
= φ(x)dx + φ(x)dx φ(x) = φ(x + p)
α−p 0
Z p
= φ(x)dx.
0
Thus, we are done for this claim.


Page 53
MATH 534H Editor : Byeongho Ban

Proof. (a)
Observe that
Z b Z b Z a+p
φ(x)dx = φ(x)dx + φ(x)dx
a a+p a
Z b+p Z b+p Z a+p
= φ(x)dx − φ(x)dx + φ(x)dx
a+p b a
Z b+p Z p Z p
= φ(x)dx − φ(x)dx + φ(x)dx ∵ by our claim above.
a+p 0 0
Z b+p
= φ(x)dx.
a+p
Similarly,
Z b Z b Z a−p
φ(x)dx = φ(x)dx + φ(x)dx
a a−p a
Z b−p Z b−p Z a−p
= φ(x)dx − φ(x)dx + φ(x)dx
a−p b a
Z b−p Z p Z p
= φ(x)dx − φ(x)dx + φ(x)dx ∵ by our claim above.
a−p 0 0
Z b−p
= φ(x)dx.
a−p

(b)
Observe that
Z p/2 Z a+p/2
φ(x + a)dx = φ(y)dy ∵ by the change of variables y = x + a
−p/2 a−p/2
Z a+p/2
= φ(x)dx ∵ by the change of variables y = x
a−p/2
Z a+p/2 Z −p/2
= φ(x)dx + φ(x)dx
−p/2 a−p/2
Z p/2 Z p/2+a Z −p/2
= φ(x)dx + φ(x)dx + φ(x)dx
−p/2 p/2 a−p/2
Z p/2 Z p/2+a Z −p/2+p
= φ(x)dx + φ(x)dx + φ(x)dx ∵ by (a)
−p/2 p/2 a−p/2+p
Z p/2 Z p/2 Z p/2
= φ(x)dx − φ(x)dx + φ(x)dx
−p/2 p/2+a a+p/2
Z p/2
= φ(x)dx
−p/2

Actually, the last statement of (b) can be observed in our claim above.


Page 54
MATH 534H Editor : Byeongho Ban

Due date : April 25th ByeongHo Ban

Proof. 5.1.2a)
We can represent φ(x) = x2 as
∞
X
φ(x) = (An sin nπx) .
n=1
And observe that
Z 1
An = 2 x2 sin nπxdx
0
 Z 1 
1 2
= 2 − (x2 cos nπx)|1x=0 + x cos nπxdx
nπ nπ 0
 Z 1 
2 2 2
=− cos nπ − x sin nπx|1x=0 + sin nπxdx
nπ nπ nπ 0
 
2 2
=− (−1)n − 2 2 (cos nπ − 1)
nπ n π
2 4
= (−1)n+1 + 3 3 ((−1)n − 1)
nπ n π
Thus,
(
− 2 , n : even
An = 2 nπ 8
nπ − n3 π 3 , n : odd.
Thus, by plugging the coefficient An back to the series, we get the sine Fourier series of x2 .


Page 55
MATH 534H Editor : Byeongho Ban

Proof. 5.1.5

We can represent φ(x) by

∞
X nπx
x = φ(x) = An sin .
n=1
l
Observe that
" l # " l #
2 l
Z Z l
nπx 2 l nπx  l nπx 2 n+1 l nπx  2l
An = x sin = − x cos + cos = l(−1) + sin = (−1)n+1 .
l 0 l l nπ l x=0 nπ 0
 l nπ nπ l x=0
 nπ
Thus,
∞
2l X (−1)n+1 nπx
x= sin .
π n=1 n l

(a)
Now, assuming we can integrate it term by term, we integrate both sides and get
∞
x2 2l2 X (−1)n nπx
=C+ 2 cos
2 π n=1 n2 l
where C is integration constant.

x2
We can represent ψ(x) = 2 as
∞
1 X
ψ(x) = A0 + (An cos nπx) .
2 n=1
And observe that
l
x2
Z
2 1
A0 = dx = l2 .
l 0 2 3
l2
Thus, C = 6 so
∞
x2 l2 2l2 X (−1)n nπx
ψ(x) = = + 2 cos .
2 6 π n=1 n2 l

(b) Let’s set x = 0, then

∞ ∞
02 l2 2l2 X (−1n ) π2 X (−1)n+1
0= = ψ(0) = + 2 2
=⇒ =
2 6 π n=1 n 12 n=1 n2
So we found the sum of the series and we are done.


Page 56
MATH 534H Editor : Byeongho Ban

Proof. 5.2.4

(a) We know we can represent φ(x) by

∞ h
1 X nπx nπx i
φ(x) = A0 + An cos + Bn sin .
2 n=1
l l
Now observe that
Z l
1
A0 = φ(x) dx = 0 φ : odd
l −l
Z l
1 nπx
An = φ(x) cos dx = 0
l −l l
since φ is odd function and cos is even function so their multiplication is odd function. Therefore, the Fourier series of φ
has only sine terms if φ is nonzero function.

(b)
If φ is even function, when observe that
Z l
1 nπx
Bn = φ(x) sin dx = 0
l −l l
since φ is even and sin is odd function so their multiplication is odd function. Therefore, the Fourier series of φ would
have only cosine terms if φ is nonzero function.

Proof. 5.3.2 a)

Let a constant function C be given. Then observe that

Z 1 Z 1
Cx dx = C x dx = 0
−1 −1
since x is odd function. Therefore, any constant function is orthogonal to x vice versa.

Proof. 5.3.2 b)

Let f (x) = x2 + ax + b be the quadratic polynomial orthogonal to x and 1 on the interval [−1, 1]. (it suffice to investigate
monic polynimial since inner product is linear functional.) Then we should have
Z 1 Z 1
2 1
0= f (x)dx = x2 + ax + b dx = + 2b =⇒ b = −
−1 −1 3 3
Z 1 Z 1
2a
0= xf (x)dx = x3 + ax2 + bx dx = =⇒ a = 0
−1 −1 3
Thus,
1
f (x) = x2 − .
3
Then we are done.


Page 57
MATH 534H Editor : Byeongho Ban

Proof. 5.3.3 (You should do this problem for zero Neumann boundary conditions instead of the mixed ones. That
is consider instead the given wave equations with ux (0, t) = 0 = ux (`, t) and the same initial data u(x, 0) = x and
ut (x, 0) = 0.)

Note that the solution representation for this problem with the zero Neumann condition is
∞  
1 t X nπct nπct nπx
u(x, t) = A0 + B0 + An cos + Bn sin cos .
2 2 n=1
l l l
Observe that
∞
1 X nπx
x = u(x, 0) = A0 + An cos .
2 n=1
l
Note that the Fourier cosine series of x is given by the coefficient
2 l
Z
A0 = xdx = l
l 0
(
2 l − n24lπ2
Z
nπx 2l n n : odd
An = x cos dx = 2 2 [(−1) − 1] = .
l 0 l n π 0 n : even
Also, observe that
∞ h
1 X nπc i nπx
0 = ut (x, 0) = B0 + Bn cos .
2 n=1
l l
Thus, all Bn should be 0. Therefore, our solution would be
∞
l 4l X 1 (2n − 1)πct (2n − 1)πx
u(x, t) = − 2 cos cos .
2 π n=1 (2n − 1)2 l l


Page 58
MATH 534H Editor : Byeongho Ban

Proof. 5.4.5

(a)
Note the representation of Fourier cosine series of φ(x) would be
∞
1 X nπx
ψ(x) = A0 + An cos .
2 n=1
3
Then the coefficients are
Z 3 Z 3
2 2 4
A0 = φ(x)dx = dx =
3 0 3 1 3
Z 3
2 nπx
An = φ(x) cos dx
3 0 3
Z 3
2 nπx
= cos
dx
3 1 3
3
2 3 nπx 
= sin
3 nπ 3 x=1
2 nπ
=− sin
nπ 3
Then note that
√
2 π 3
A1 = − sin = −
π 3 π√
2 2π 3
A2 =− sin =−
2π 3 2π
2 3π
A3 =− sin =0
3π 3 √
2 4π 3
A4 =− sin =
4π 3 4π

(b)
Thus, for each x ∈ [0, 3] the sum of this series is

∞ √ √ √ 0
 , x ∈ (0, 1)
2 X 2 nπ nπx 2 3 πx 3 2πx 3 4πx
ψ(x) = − sin cos = − cos − cos +0+ cos + · · · = 12 ,x = 1
3 n=1 nπ 3 3 3 π 3 2π 3 4π 3 
1 , x ∈ (1, 3)


since ψ(x) = 21 [φ(x− ) + φ(x+ )] and there is a jump at x = 1. Also, since this series is cosine(even) series, we have
even extension which is symmetric with respect to y−axis. Thus, at x = 0, the formula above indicates that ψ(0) = 0.
And at x = 3, since the extension has period 2l, left and right limit is same value 3. Thus, by the above formula, ψ(3) = 3.

(c)
Observe that
Z 3 Z 3
2
|φ(x)| dx = dx = 2 < ∞.
0 1
Since the integration is finite, the Fourier series converges in the L2 sense.

(d)
Observe that
√ √ √ √ √
2 3 3 3 3 3
0 = lim+ φ(0) = − − +0+ + +0− ···
x→0 3 π 2π 4π 5π 7π
√  
2 3 1 1 1 1
= − 1 + − − + + ··· .
3 π 2 4 5 7
Thus,
 
1 1 1 1 2π
1 + − − + + ··· = √ .
2 4 5 7 3 3


Page 59
MATH 534H Editor : Byeongho Ban

Proof. 5.4.6

We should find a coefficient An in

∞
X
cos x = An sin nx.
n=1
Observe that
Z π
2
An = cos x sin nxdx
π 0
and that
Z π π
Z π
cos x sin nxdx = sin x sin nxx=0 − n sin x cos nxdx
0 0
 Z π 
π
= −n − cos x cos nxx=0 − n cos x sin nxdx
0
Z π
= −n(1 + (−1)n ) + n2 cos x sin nxdx (†)
0
so
π
n(1 + (−1)n )
Z
cos x sin nx dx = .
0 n2 − 1
(When n = 1, from (†), the A1 = 0. )
Therefore,
A1 = 0
(
2n(1 + (−1)n ) 0 , n : odd
An = =⇒ 4n
π(n2 − 1) π(n2 −1) , n : even.
and
∞ ∞
X 2n(1 + (−1)n ) X 8n
cos x = 2 − 1)
sin nx = 2 − 1)
sin 2nx ∀x ∈ (0, π).
n=2
π(n n=1
π(4n

Since cos is continuously differentiable, the series pointwisely converges to cos x.

Let
∞
X 8n
ψ(x) = 2 − 1)
sin 2nx,
n=1
π(4n
Then note that, since each term is sine function so is odd function, it must be ψ(−x) = −ψ(x). Thus, ψ(x) = − cos x
∀x ∈ (−π, 0). As for the points ±π and 0 we compute directly. Observe that
∞
X 8n
ψ(0) = sin 0 = 0
n=1
π(4n2 − 1)
∞
X 8n
ψ(±π) = 2 − 1)
sin ±2nπ = 0.
n=1
π(4n
Therefore,

cos x
 , x ∈ (0, π)
ψ(x) = − cos x , x ∈ (−π, 0) .

0 , x ∈ {0, ±π}.


Page 60
MATH 534H Editor : Byeongho Ban

Proof. 5.4.8a)

We consider f (x) = x3 on (0, l) with Fourier sine series

∞
X πnx
ψ(x)= An sin .
n=1
l
Note that f is infinitely continuously differentiable in [0, l]. And also note that ψ satisfies the given Dirichlet boundary
condition. But also note that f (l) = l3 6= 0 so f does not satisfies the Dirichlet boundary condition. Thus, we cannot
use Theorem 2 to determine the uniform convergence of ψ to f .

Now, observe that

l l
l7
Z Z
|f (x)|2 dx = x6 dx = < ∞.
0 0 7
Thus, by the theorem 3, ψ converges to f in the mean square (L2 ) sense in (0, l) as long as l is finite.

Lastly, since f is continuously differentiable in [0, l] which means f itself is continuous, by theorem 4 , the series
converges pointwisely in [0, l].


Page 61