Introduction to Partial Differential Equations

Weijiu Liu
Department of Mathematics
University of Central Arkansas
201 Donaghey Avenue, Conway, AR 72035, USA
1 Opening
• Welcome to your PDEs class! My name is Weijiu Liu. I will guide you to navigate
through PDEs. We will share the learning task together. You will be the major players
and I will be a just facilitator. You will need to read your textbook, attend classes,
practice in class, do your homework, and so on. I will just tell you what to read, what
to do, and how to do. Whenever you cannot hear me acoustically due to my Chinese
accent, please feel free to ask me to repeat. If you cannot hear me mathematically,
please read your textbook.
• Syllabus.

• Overview of PDEs.

2
2 Lecture 1 – PDE terminology and Derivation of 1D
heat equation
Today:

• PDE terminology.

• Classification of second order PDEs.

• Derivation of 1D heat equation.

Next:
• Boundary conditions

• Derivation of higher dimensional heat equations

Review:

• Classification of conic section of the form:

Ax2 + Bxy + Cy 2 + Dx + Ey + F = 0,

where A, B, C are constant. It is

a hyperbola if B 2 − 4AC > 0,
a parabola if B 2 − 4AC = 0,
an ellipse if B 2 − 4AC < 0.
• Conservation of heat energy:
Rate of change of heat energy in time = Heat energy flowing across boundaries per
unit time + Heat energy generated insider per unit time
• Fourier’s Law: the heat flux is proportional to the temperature gradient

φ = −K0 ∇u. (1)

• Fick’s Law of diffusion: the chemical flux is proportional to the gradient of chemical
concentration
φ = −K0 ∇u. (2)

Teaching procedure:

1. PDE terminology

• PDE: an equation containing an unknown function and its derivatives, e.g.,

∂ 2u ∂ 2u ∂ 2u
+ + = 0.
∂x2 ∂y 2 ∂z 2

3
 • Order of a PDE: the highest order of the derivative in the equation.
• Linear PDEs: linear in the unknown function and its derivatives, e. g.,

∂ 2u ∂ 2u ∂ 2u ∂u ∂u
A(x, y) 2 +B(x, y) +C(x, y) 2 +D(x, y) +E(x, y) +F (x, y)u = G(x, y).
∂x ∂x∂y ∂y ∂x ∂y

• Nonlinear PDEs: not linear, e.g., Burgers’ equation

∂u ∂u ∂ 2u
+u = k 2.
∂t ∂x ∂x
• Solution of a PDE: a function that has all required derivatives and that satisfies
the equation.
∂ 2u ∂ 2u ∂ 2u
+ + = 0.
∂x2 ∂y 2 ∂z 2
Solution: u = a + bx + cy + dz, where a, b, c, d are constants.

2. Classification of second order equations:

∂ 2u ∂ 2u ∂ 2u ∂u ∂u
A 2 +B +C 2 +D +E + F u = G(x, y),
∂x ∂x∂y ∂y ∂x ∂y
Where A, B, C are constant. It is said to be
hyperbolic if B 2 − 4AC > 0,
parabolic if B 2 − 4AC = 0,
elliptic if B 2 − 4AC < 0.

• The wave equation – hyperbolic

∂ 2u 2
2∂ u
= c .
∂t2 ∂x2
A = 1, B = 0, C = −c2 , B 2 − 4AC = 4c2 > 0.
• The heat equation – parabolic

∂u ∂2u
= k 2.
∂t ∂x
A = 0, B = 0, C = −k, B 2 − 4AC = 0.
• The Laplace’s equation – elliptic

∂ 2u ∂2u
+ = 0.
∂x2 ∂y 2

A = 1, B = 0, C = 1, B 2 − 4AC = −4 < 0.

3. List of important PDEs.

4
• The Laplace’s equation describing potentials of a physical quantity

2 ∂ 2u ∂ 2u ∂2u
∇ u= + + = 0.
∂x2 ∂y 2 ∂z 2

• The heat equation describing heat conduction (u denoting temperature or con-

centration of chemical)
µ 2 ¶
∂u ∂ u ∂2u ∂ 2u
=k + + .
∂t ∂x2 ∂y 2 ∂z 2

• The convection-diffusion equation describing heat convection and heat conduction

(u denoting temperature or concentration of chemical)
µ 2 ¶
∂u ∂u ∂u ∂u ∂ u ∂ 2u ∂ 2u
+ v1 + v2 + v3 =k + + .
∂t ∂x ∂y ∂z ∂x2 ∂y 2 ∂z 2

• The wave equation describing vibration of string or membrane (u denoting dis-

placement) µ 2 ¶
∂ 2u 2 ∂ u ∂ 2u ∂2u
=c + + .
∂t2 ∂x2 ∂y 2 ∂z 2
• Euler-Bernoulli beam equation describing the buckling of a beam (u denoting
displacement) µ ¶
∂2u ∂2 ∂ 2u
+ 2 EI 2 = 0.
∂t2 ∂x ∂x
• Korteweg-de Vries equation describing nonlinear shallow water waves (u denoting
displacement)
∂u ∂u ∂ 3u
+u + k 3 = 0.
∂t ∂x ∂x
• Schrodinger equation describing quantum mechanical behavior (Ψ denoting wave-
functions) µ ¶
∂Ψ ~2 ∂ 2 Ψ ∂ 2 Ψ ∂ 2 Ψ
i~ =− + + + V Ψ.
∂t 2m ∂x2 ∂y 2 ∂z 2
• Maxwell’s equations describing electromagnetic fields (E denoting the intensity
of electric field and H denoting the intensity of magnetic field)
∂(ε0 E) ∂(µ0 H)
∇×H− ∂t
= 0, ∇×E+ ∂t
= 0,
(3)
∇ · (ε0 E) = 0, ∇ · (µ0 H) = 0,

• Reaction-diffusion system describing reaction diffusion processes of chemicals (u

denoting the concentration vector of chemicals)
µ 2 ¶
∂(u) ∂ u ∂ 2u ∂2u
=k + 2 + 2 + f (u).
∂t ∂x2 ∂y ∂z

5
 • Thermoelastic system
µ 2 ¶
∂ 2u ∂ u ∂2u ∂ 2u
= µ + 2 + 2 + (λ + µ)∇divu − α∇θ,
∂t2 ∂x2 ∂y ∂z
µ 2 ¶ µ ¶
∂θ ∂ θ ∂ θ ∂2θ
2
∂u
= k + + − βdiv .
∂t ∂x2 ∂y 2 ∂z 2 ∂t

• The incompressible Navier-Stokes equations describing dynamics of fluid flows

(u, v, w denoting the components of a velocity field of fluid flows)
µ 2 ¶
∂u ∂u ∂u ∂u 1 ∂p ∂ u ∂2u ∂ 2u
+u +v +v = − +ν + + ,
∂t ∂x ∂y ∂z ρ ∂x ∂x2 ∂y 2 ∂z 2
µ 2 ¶
∂v ∂v ∂v ∂v 1 ∂p ∂ v ∂ 2v ∂ 2v
+u +v +v = − +ν + + ,
∂t ∂x ∂y ∂z ρ ∂y ∂x2 ∂y 2 ∂z 2
µ 2 ¶
∂w ∂w ∂w ∂w 1 ∂p ∂ w ∂ 2w ∂ 2w
+u +v +v = − +ν + + ,
∂t ∂x ∂y ∂z ρ ∂z ∂x2 ∂y 2 ∂z 2
∂u ∂v ∂w
+ + = 0.
∂x ∂y ∂z
One million dollars are offered by Clay Mathematics Institute for solutions
of the most basic questions one can ask: do solutions exist, and are they unique?
Why ask for a proof?

4. Derivation of 1D heat equation.

Physical problem: describe the heat conduction in a rod of constant cross section area
A.

Physical quantities:

• Thermal energy density e(x, t) = the amount of thermal energy per unit vol-
Energy
ume = .
Volume
• Heat flux φ(x, t) = the amount of thermal energy flowing across boundaries per
Energy
unit surface area per unit time = Area·Time .
• Heat sources Q(x, t) = heat energy per unit volume generated per unit time
Energy
= .
Volume·Time
6
 • Temperature u(x, t).
• Specific heat c = the heat energy that must be supplied to a unit mass of a
Energy
substance to raise its temperature one unit = Mass·Temperature .

• Mass density ρ(x) = mass per unit volume = Mass .

Volume
Conservation of heat energy:
Rate of change of heat energy in time = Heat energy flowing across boundaries per
unit time + Heat energy generated insider per unit time

• heat energy = Energy density × Volume = e(x, t)A∆x.

• Heat energy flowing across boundaries per unit time = flow-in flux × Area −
flow-out flux × Area = φ(x, t)A − φ(x + ∆x, t)A.
• Heat energy generated insider per unit time = Q(x, t) × Volume = Q(x, t)A∆x.

Then
∂
[e(x, t)A∆x] = φ(x, t)A − φ(x + ∆x, t)A + Q(x, t)A∆x.
∂t
Dividing it by A∆x and letting ∆x go to zero give
∂e ∂φ
=− + Q. (4)
∂t ∂x

Energy
Heat energy = Mass·Temperature · Temperature · Mass · Volume = c(x)u(x, t)ρA∆x.
Volume
So
e(x, t)A∆x = c(x)u(x, t)ρA∆x,
and then
e(x, t) = c(x)u(x, t)ρ.
It then follows from Fourier’s law that
µ ¶
∂u ∂ ∂u
cρ = K0 + Q. (5)
∂t ∂x ∂x

and then the heat equation

∂u ∂ 2u Q
=k 2 + , (6)
∂t ∂x cρ
K0
where k = cρ
is called the thermal diffusivity.

5. Practice: Exercise 1.2.3.

7
3 Lecture 2 – Derivation of higher dimensional heat
equations and Initial and boundary conditions
Today:
• Derivation of higher dimensional heat equations
• Initial conditions and boundary conditions
Next:
• Equilibrium
Review:
• Conservation of heat energy:
Rate of change of heat energy in time = Heat energy flowing across boundaries per
unit time + Heat energy generated insider per unit time
• Fourier’s Law: the heat flux is proportional to the temperature gradient

φ = −K0 ∇u. (7)

• Divergence Theorem: Z I
∇ · AdV = A · ndS (8)
R

Teaching procedure:

1. Derivation of 2D or 3D heat equation.

Physical problem: describe the heat conduction in a region of 2D or 3D space.

Physical quantities:

• Thermal energy density e(x, t) = the amount of thermal energy per unit vol-
Energy
ume = .
Volume
• Heat flux φ(x, t) = the amount of thermal energy flowing across boundaries per
Energy
unit surface area per unit time = Area·Time .

8
 • Heat sources Q(x, t) = heat energy per unit volume generated per unit time
Energy
= .
Volume·Time
• Temperature u(x, t).
• Specific heat c = the heat energy that must be supplied to a unit mass of a
Energy
substance to raise its temperature one unit = Mass·Temperature .

• Mass density ρ(x) = mass per unit volume = Mass .

Volume
Conservation of heat energy:
Rate of change of heat energy in time = Heat energy flowing across boundaries per
unit time + Heat energy generated insider per unit time
R
• heat energy = R e(x, t)dV .
H
• Heat energy flowing across boundaries per unit time = φ · ndS.
R
• Heat energy generated insider per unit time = R Q(x, t)dV .

Then Z I Z
∂
e(x, t)dV = − φ · ndS + Q(x, t)dV.
∂t R R
The divergence theorem give
Z Z Z
∂
e(x, t)dV = − ∇ · φdV + Q(x, t)dV.
∂t R R R

and then
∂e
= −∇ · φ + Q. (9)
∂t
Heat energy per unit volume = c(x)u(x, t)ρ. So

e(x, t) = c(x)u(x, t)ρ.

It then follows from Fourier’s law that

∂u
cρ = ∇ · (K0 ∇u) + Q. (10)
∂t
and then the heat equation
∂u Q
= k∇2 u + , (11)
∂t cρ
K0
where k = cρ
is called the thermal diffusivity.

2. Laplace’s equation
∇2 u = 0.

3. Poisson’s equation
∇2 u = q.

9
4. Initial conditions
u(x, y, z, 0) = f (x, y, z).

5. boundary conditions

• Prescribed temperature boundary condition (Dirichlet boundary condition, math-

ematically)
u(x, y, z, t) = T (x, y, z, t) on ∂Ω.
u(0, t) = T1 (t), u(L, t) = T2 (t).
• Prescribed heat flux boundary condition (Neumann boundary condition, math-
ematically)

∂u
K0 ∇u(x, y, z, t) · n = K0 = φ(x, y, z, t) on ∂Ω.
∂n
∂u ∂u
−K0 (0, t) = φ1 (t), K0 (L, t) = φ2 (t).
∂x ∂x
• Newton’s law of cooling boundary condition (Robin boundary condition, math-
ematically)
∂u
K0 ∇u · n = K0 = −h(u − ub ) on ∂Ω.
∂n
∂u ∂u
−K0 (0, t) = −h(u(0, t) − u1 (t)), K0 (L, t) = −h(u(L, t) − u2 (t)).
∂x ∂x
6. Practice Exercise 1.5.1.

10
4 Lecture 3 – Equilibrium
Today:
• Equilibrium
Next:
• Eigenvalue problems
Review:
Rb ¯b R
¯ b
• Integration by Parts: a
udv = uv ¯ − a vdu.
a

• Hölder’s inequality:
Z b µZ b ¶1/2 µZ b ¶1/2
2 2
|f (x)g(x)|dx ≤ |f (x)| dx |g(x)| dx . (12)
a a a

• Differential inequality: If
df (t)
≤ af (t),
dt
then
f (t) ≤ f (0)eat .

Teaching procedure:

1. Diruchlet BC.

∂u
= k∇2 u, (13)
∂t
u(x, y, z, t) = T (x, y, z) on ∂Ω, (14)
u(x, y, z, 0) = f (x, y, z). (15)

Steady equation

∇2 w = 0,
w(x, y, z) = T (x, y, z) on ∂Ω,

w is called equilibrium or steady state solution of (13)-(15).

Example. Consider 1D problem:
∂u d2 u
= k 2, (16)
∂t dx
u(0, t) = T1 , u(0, t) = T2 , (17)
u(x, 0) = f (x). (18)

11
and its corresponding steady equation

d2 w
= 0,
dx2
w(0) = T1 , w(L) = T2 .

The solution is
T2 − T1
w(x) = T1 + x.
L
Theorem 4.1. For any continuous function f (x), the solution u(x, t) of (16)-(18)
converges w(x) as t tends to infinity. More precisely,
Z L
lim |u(x, t) − w(x)|2 dx = 0 (19)
t→∞ 0

Lemma 4.1. (Poincare’s inequality) For any continuous function f (x) with f (0) = 0,
Z L Z L¯ ¯
¯ df (x) ¯2
2
|f (x)| dx ≤ L 2 ¯ ¯
¯ dx ¯ dx. (20)
0 0

Proof.
¯Z ¯ Z ¯ ¯ µZ x ¶1/2 ÃZ x ¯ ¯2 !1/2
¯ x
df (s) ¯¯ x ¯ df (s) ¯ ¯ ¯
|f (x)| = ¯¯ ds¯ ≤ ¯ ¯ ds ≤ 12
ds ¯ df (s) ¯ ds
ds ¯ ds ¯ ¯ ds ¯
0 0 0 0
ÃZ ¯ ¯ !1/2
L ¯ df (s) ¯2
≤ L 1/2 ¯ ¯
¯ ds ¯ ds .
0

Squaring the inequality and integrating from 0 to L give (20).

Proof of Theorem 4.1. Let v = u − w. Then

∂v d2 v
= k 2, (21)
∂t dx
v(0, t) = 0, v(0, t) = 0, (22)
v(x, 0) = f (x) − w(x). (23)

Multiplying (21) by v and integrating from 0 to L give

Z L Z L
1d 2 d2 v(x, t)
v (x, t)dx = k v(x, t) dx
2 dt 0 0 dx2
Z Lµ ¶2
dv(x, t)
= −k dx
0 dx
Z L
2
= −kL |v(x, t)|2 dx.
0

12
 Solving this inequality gives
Z L Z L Z L
2 −2kL2 t 2 −2kL2 t
|v(x, t)| dx ≤ e |v(x, 0)| dx = e |f (x) − w(x)|2 dx. (24)
0 0 0

2. Neumann BC.

∂u
= k∇2 u, (25)
∂t
∂u
(x, y, z, t) = φ(x, y, z) on ∂Ω, (26)
∂n
u(x, y, z, 0) = f (x, y, z). (27)

Steady equation

∇2 w = 0,
∂w
(x, y, z, t) = φ(x, y, z) on ∂Ω,
∂n

w is called equilibrium or steady state solution of (25)-(27).

Example. Consider 1D problem:
∂u d2 u
= k 2, (28)
∂t dx
∂u ∂u
(0, t) = 0, (0, t) = 0, (29)
∂x ∂x
u(x, 0) = f (x). (30)
and its corresponding steady equation
d2 w
= 0,
dx2
∂w ∂w
(0) = 0, (L) = 0.
∂x ∂x
The solution is
w(x) = C. (any constant)

• Conservation of thermal energy. Integrating (28) from 0 to L gives

Z
d L
u(x, t)dx = 0.
dt 0
So
Z L Z L Z L
u(x, t)dx = u(x, 0)dx = f (x)dx. (31)
0 0 0

13
 • Convergence.
Theorem 4.2. For any continuous function f (x), the solution u(x, t) of (28)-(30)
RL
converges w(x) = C = L1 0 f (x)dx as t tends to infinity. More precisely,
Z L Z L
1
lim |u(x, t) − f (s)ds|2 dx = 0 (32)
t→∞ 0 L 0

Lemma 4.2. (Poincaré’s inequality) For any continuous function f (x),

Z ¯ Z L ¯2 Z L¯ ¯
L ¯ 1 ¯ ¯ df (x) ¯2
¯f (x) − f (s)ds¯¯ dx ≤ M ¯ ¯
¯ L 0 ¯ dx ¯ dx, (33)
0 0

where M is a positive constant.

1
RL
Proof of Theorem 4.2. Let v = u − L 0
f (s)ds. Then

∂v d2 v
= k 2, (34)
∂t dx
∂v ∂v
(0, t) = 0, (0, t) = 0, (35)
∂x ∂x
Z
1 L
v(x, 0) = f (x) − f (s)ds. (36)
L 0

Multiplying (34) by v and integrating from 0 to L give

Z L Z L
1d 2 d2 v(x, t)
v (x, t)dx = k v(x, t) dx
2 dt 0 0 dx2
Z Lµ ¶2
dv(x, t)
= −k dx
0 dx
Z L
= −kM |v(x, t)|2 dx.
0

Solving this inequality gives

Z Z Z ¯ Z L ¯2
L L L ¯ 1 ¯
2
|v(x, t)| dx ≤ e −2kM t 2
|v(x, 0)| dx = e −2kM t ¯f (x) − f (s)ds¯ dx.
¯ L 0 ¯
0 0 0

3. Practice. Exercise 1.4.1 (b), (c).

14
5 Lecture 4 – Eigenvalue problems
Today:
• Eigenvalue problems
Next:
• Linear differential operators
• Separation of variables
Review:
• The method of solving second-order homogeneous linear equations with constant coef-
ficients:

ay 00 + by 0 + cy = 0
am2 + bm + c = 0

– Distinct real roots m1 and m2 :

y = c1 em1 x + c2 em2 x ;

– Repeated real roots m1 = m2 :

y = c1 em1 x + c2 xem1 x ;

– Conjugate complex roots m1 = α + iβ and m2 = α − iβ:

y = eαx (c1 cos(βx) + c2 sin(βx));

Teaching procedure:

1. Separation of variables.

∂u ∂ 2u
= k 2, (37)
∂t ∂x
u(0, t) = 0, u(L, t) = 0, (38)
u(x, 0) = f (x). (39)

Look for a solution of the form of separation of variables:

u(x, t) = φ(x)G(t), (40)

The method of finding such a solution is called the method of separation of vari-
ables.

15
 Substitute the above expression into the equation (37), we obtain

φ(x)G0 (t) = kφ00 (x)G(t),

and then
G0 (t) φ00 (x)
= = −λ, (41)
kG(t) φ(x)
where λ is constant to be determined. The boundary condition yields that

φ(0) = φ(L) = 0.

We then have an eigenvalue problem

d2 φ
− = λφ,
dx2
φ(0) = 0, φ(L) = 0,

2. Eigenvalue problems with Dirichlet BC: Find a complex number λ such that the prob-
lem

d2 φ
− = λφ,
dx2
φ(0) = 0, φ(L) = 0.

has non-zero solutions φ. λ is called eigenvalue and φ called the corresponding

eigenfunction.
Auxiliary equations:
m2 = −λ.
√ √
• Case 1: λ < 0. Distinct real roots m1 = −λ and m2 = − −λ:
√ √
−λx
φ(x) = c1 e + c2 e− −λx
.

The boundary conditions imply that c1 = c2 = 0. So no non-zero solutions exist

and then λ < 0 is not an eigenvalue.
• Case 2: λ = 0. Repeated real roots m1 = m2 = 0:

φ = c1 + c2 x.

The boundary conditions imply that c1 = c2 = 0. So no non-zero solutions exist

and then λ < 0 is not an eigenvalue.
√ √
• Case 3: λ > 0. Conjugate complex roots m1 = i λ and m2 = −i λ:
√ √
φ = c1 cos( λx) + c2 sin( λx).

16
 φ(0) = 0 implies that c1 = 0. φ(L) = 0 gives
√
sin( λL) = 0.
√
So λL = nπ (n = 1, 2, · · · ) and then we obtain the eigenvalues
n2 π 2
λn = , n = 1, 2, · · · (42)
L2
and the corresponding eigenfunctions
³ nπx ´
φn = sin , n = 1, 2, · · · (43)
L
3. Eigenvalue problems with Neumann BC:

d2 φ
− = λφ,
dx2
∂φ ∂φ
(0) = 0, (L) = 0.
∂x ∂x
Auxiliary equations:
m2 = −λ.
√ √
• Case 1: λ < 0. Distinct real roots m1 = −λ and m2 = − −λ:
√ √
−λx
φ(x) = c1 e + c2 e− −λx
.
The boundary conditions imply that c1 = c2 = 0. So no non-zero solutions exist
and then λ < 0 is not an eigenvalue.
• Case 2: λ = 0. Repeated real roots m1 = m2 = 0:
φ = c1 + c2 x.
The boundary conditions imply that c2 = 0 and c2 can be any constants. So
λ0 = 0 is an eigenvalue with the eigenfunction
φ0 = 1.
√ √
• Case 3: λ > 0. Conjugate complex roots m1 = i λ and m2 = −i λ:
√ √
φ = c1 cos( λx) + c2 sin( λx).
φ0 (0) = 0 implies that c2 = 0. φ0 (L) = 0 gives
√
sin( λL) = 0.
√
So λL = nπ (n = 1, 2, · · · ) and then we obtain the eigenvalues
n2 π 2
λn = , n = 1, 2, · · · (44)
L2
and the corresponding eigenfunctions
³ nπx ´
φn = cos , n = 1, 2, · · · (45)
L

17
4. Eigenvalue problems with periodic BC:

d2 φ
− = λφ,
dx2
∂φ ∂φ
φ(−L) = φ(L), (−L) = (L).
∂x ∂x
Auxiliary equations:
m2 = −λ.
√ √
• Case 1: λ < 0. Distinct real roots m1 = −λ and m2 = − −λ:
√ √
−λx
φ(x) = c1 e + c2 e− −λx
.
The boundary conditions imply that c1 = c2 = 0. So no non-zero solutions exist
and then λ < 0 is not an eigenvalue.
• Case 2: λ = 0. Repeated real roots m1 = m2 = 0:
φ = c1 + c2 x.
The boundary conditions imply that c2 = 0 and c2 can be any constants. So
λ0 = 0 is an eigenvalue with the eigenfunction
φ0 = 1.
√ √
• Case 3: λ > 0. Conjugate complex roots m1 = i λ and m2 = −i λ:
√ √
φ = c1 cos( λx) + c2 sin( λx).
φ(−L) = φ(L) gives
√ √ √ √
c1 cos(− λL) + c2 sin(− λL) = c1 cos( λL) + c2 sin( λL).
√
c2 sin( λL) = 0.
∂φ ∂φ
∂x
(−L) = ∂x
(L)
gives
√ √ √ √ √ √
λ[−c1 sin(− λL) + c2 cos(− λL)] = λ[−c1 sin( λL) + c2 cos( λL)].
√
c1 sin( λL) = 0.
Hence we obtain the eigenvalues
n2 π 2
λn = 2 , n = 1, 2, · · · (46)
L
√ √
since, for these λn , c1 cos( λn x) + c2 sin( λn x) are non-zero solutions. So the
corresponding eigenfunctions
³ nπx ´ ³ nπx ´
φ2n−1 = cos , φ2n = sin , n = 1, 2, · · · (47)
L L

18
6 Lecture 5 – Separation of variables
Today:
• Linear differential operators
• Dirichlet boundary value problems
Next:
• Continue with Dirichelt BVP.
Review:
• the heat, wave, and Laplace’s equations
• solution of linear ode
dG
= aG
dt
is
G(t) = Ceat .

• The eigenvalue problem

d2 φ
− 2 = λφ,
dx
φ(0) = 0, φ(L) = 0

has the eigenvalues

n2 π 2
λn = , n = 1, 2, · · · (48)
L2
and the corresponding eigenfunctions
³ nπx ´
φn = sin , n = 1, 2, · · · (49)
L

Teaching procedure:

1. Linear differential operators.

• Linear operator. An operator L that maps a function to another function is

linear if it satisfies

L(c1 u1 + c2 u2 ) = c1 L(u1 ) + c2 L(u2 )

for any two functions u1 , u2 and any constants c1 , c2 .

Example. The heat operator
∂ ∂2
−k 2
∂t ∂x
is a linear differential operator.

19
 The wave operator
∂2 2 ∂
2
− c
∂t2 ∂x2
is a linear differential operator.
The Laplace’s operator
∂2 ∂2
+
∂x2 ∂y 2
is a linear differential operator.
• Linear differential equation. If L is a linear differential operator, then

L(u) = f

is a linear differential equation, where f is a known function.

Example. The heat equation

∂u ∂ 2u
−k 2 =f
∂t ∂x
is a linear differential equation.
The wave equation
∂ 2u 2
2∂ u
−c =f
∂t2 ∂x2
is a linear differential equation.
The Poisson’s equation
∂ 2u ∂2u
+ =f
∂x2 ∂y 2
is a linear differential equation.
The Burgers’ equation
∂u ∂u ∂ 2u
+u =k 2
∂t ∂x ∂x
is a nonlinear equation
• Principle of superposition. If u1 and u2 are solutions of a linear equation

L(u) = f,

then c1 u1 + c2 u2 is also a solution, where c1 , c2 are any constants.

Proof.
L(c1 u1 + c2 u2 ) = c1 L(u1 ) + c2 L(u2 ) = 0.

20
2. Dirichlet boundary value problems:

∂u ∂ 2u
= k 2, (50)
∂t ∂x
u(0, t) = 0, u(L, t) = 0, (51)
u(x, 0) = f (x). (52)

Look for a solution of the form of separation of variables:

u(x, t) = φ(x)G(t), (53)
The method of finding such a solution is call the method of separation of variables.
Substitute the above expression into the equation (50), we obtain
φ(x)G0 (t) = kφ00 (x)G(t),
and then
G0 (t) φ00 (x)
= = −λ, (54)
kG(t) φ(x)
where λ is constant to be determined. The boundary condition (51) yields that
φ(0) = φ(L) = 0.
We then have an eigenvalue problem

d2 φ
− = λφ,
dx2
φ(0) = 0, φ(L) = 0,
which has the eigenvalues
n2 π 2
λn = , n = 1, 2, · · · (55)
L2
and the corresponding eigenfunctions
³ nπx ´
φn = sin , n = 1, 2, · · · (56)
L
On the other hand, it follows from (54) that
dG
= −λkG, (57)
dt
which has solutions
kn2 π 2 t
G(t) = ce−λkt = ce− L2 .
We then derive that
³ nπx ´ kn2 π 2 t
un (x, t) = cn sin e−
, n = 1, 2, · · · ,
L2 (58)
L
These functions do satisfy the equation (50) and the boundary condition (51), but,
unfortunately, they DO NOT satisfy the initial condition (52)!

21
7 Lecture 6 – Infinite Series Solutions
Today:
• Superposition

• Infinite series solutions

Next:
• Neumann and Periodic boundary value problems.
Review:
• Trigonometric identities:
1
sin u sin v = [cos(u − v) − cos(u + v)].
2
1
sin2 u = [1 − cos(2u)].
2
• Trigonometric integral: Z
cos xdx = sin x + C.
Z
sin xdx = − cos x + C.

• Integration by parts:
Z b Z b
u(x)dv(x) = uv|ba − v(x)du(x).
a a

• Orthogonal vectors.
• Dirichlet boundary value problems:

∂u ∂ 2u
= k 2, (59)
∂t ∂x
u(0, t) = 0, u(0, t) = 0, (60)
u(x, 0) = f (x). (61)

Functions ³ nπx ´ kn2 π 2 t

un (x, t) = cn sin e−
, n = 1, 2, · · · ,
L2 (62)
L
do satisfy the equation and the boundary condition, but DO NOT satisfy the initial
condition.
Teaching procedure:

22
1. Superposition:
N
X ³ nπx ´ kn2 π 2 t
u(x, t) = cn sin e− L2 .
n=1
L
If there are N and constants c1 , c2 , · · · , cN such that
N
X ³ nπx ´
f (x) = cn sin ,
n=1
L

then u(x, t) is a solution of (50), (51), and (52).

2. Infinite series solution:

∞
X ³ nπx ´ kn2 π 2 t
u(x, t) = cn sin e− L2 .
n=1
L

The initial condition gives

∞
X ³ nπx ´
f (x) = u(x, 0) = cn sin . (63)
n=1
L
¡ ¢
To determine cn , we multiply (63) by sin nπx
L
and integrate from 0 to L. We then
find Z
2 L ³ nπx ´
cn = f (x) sin dx.
L 0 L
cn is called Fourier coefficient. The trigonometric series (63) with these coefficients
is called Fourier series.
RL
3. Orthogonality. Two functions f (x), g(x) are said to be orthogonal if 0 f (x)g(x)dx =
0. A set of functions is called an orthogonal set of functions if each member of the
set is orthogonal to every other member.
¡ ¢ ∞
Example. The set {sin nπx L
}n=1 is orthogonal.

4. Example. Solve the following initial boundary value problem

∂u d2 u
= k 2, (64)
∂t dx
u(0, t) = 0, u(L, t) = 0, (65)

with the initial condition

¡ ¢ ¡ ¢
(a) u(x, 0) = 3 sin πx
L
− sin 3πx
L
.
(b) u(x, 0) = x.

23
Solution. (a) The Fourier coefficient are
Z · ³ πx ´ µ ¶¸ µ ¶
2 L 3πx 1πx
c1 = 3 sin − sin sin dx
L 0 L L L
= 3,
Z · ³ πx ´ µ ¶¸ µ ¶
2 L 3πx 2πx
c2 = 3 sin − sin sin dx
L 0 L L L
= 0,
Z · ³ πx ´ µ ¶¸ µ ¶
2 L 3πx 3πx
c3 = 3 sin − sin sin dx
L 0 L L L
= −1,
Z · ³ πx ´ µ ¶¸ ³ nπx ´
2 L 3πx
cn = 3 sin − sin sin dx
L 0 L L L
= 0, n = 4, 5, · · · ,

So the solution is
³ πx ´ µ ¶
− kπ2 t
2
3πx 9kπ 2 t
u(x, t) = 3 sin e L − sin e− L2 .
L L

(b) The Fourier coefficient are

Z L ³ nπx ´
2
cn = x sin
dx
L 0 L
³ nπx ´ ¯ Z L ³ nπx ´
2 ¯ L 2
= − x cos 0
+ cos dx
nπ L nπ 0 L
2L 2L ³ nπx ´ ¯
= cos (nπ) + 2 2 sin ¯L
nπ nπ L 0
2L
= cos (nπ) , n = 1, 2, · · · ,
nπ
So the solution is
∞
X 2L ³ nπx ´ kn2 π 2 t
u(x, t) = cos (nπ) sin e− L2 .
n=1
nπ L

24
8 Lecture 7 – Neumann and Periodic boundary value
problems
Today:
• Neumann boundary value problems
• Periodic boundary value problems
Next:
• Laplace’s equation.
Review:
• Trigonometric identities:
1
cos u cos v = [cos(u − v) + cos(u + v)].
2
1
cos2 u = [1 + cos(2u)].
2
• The Neumann eigenvalue problem
d2 φ
− = λφ,
dx2
dφ dφ
(0) = 0, (L) = 0
dx dx
has the eigenvalues
n2 π 2
λn = , n = 0, 1, 2, · · · (66)
L2
and the corresponding eigenfunctions
³ nπx ´
φn = cos , n = 0, 1, 2, · · · (67)
L

• The periodic eigenvalue problem

d2 φ
− = λφ,
dx2
dφ dφ
φ(−L) = φ(L), (−L) = (L)
dx dx
has the eigenvalues
n2 π 2
λn = , n = 0, 1, 2, · · · (68)
L2
and the corresponding eigenfunctions
³ nπx ´ ³ nπx ´
φ0 = 1, φ2n−1 = cos , φ2n = sin , n = 1, 2, · · · (69)
L L

25
Teaching procedure:
¡ nπx ¢ ¡ nπx ¢ ¡ nπx ¢
1. Orthogonality of the sets {1, cos L
, n = 1, 2, · · · , } and {1, cos L
, sin L
,n =
1, 2, · · · , }.
2. Neumann boundary value problem:

∂u ∂2u
= k 2, (70)
∂t ∂x
∂u ∂u
(0, t) = 0, (L, t) = 0, (71)
dx dx
u(x, 0) = f (x). (72)

Look for a solution of the form of separation of variables:

u(x, t) = φ(x)G(t), (73)
Substitute the above expression into the equation (70), we obtain
φ(x)G0 (t) = kφ00 (x)G(t),
and then
G0 (t) φ00 (x)
= = −λ, (74)
kG(t) φ(x)
where λ is constant to be determined. The boundary condition (71) yields that
φu φ
(0) = (L) = 0.
dx dx
We then have an eigenvalue problem

d2 φ
− = λφ,
dx2
φ φ
(0) = 0, (L) = 0,
dx dx
which has the eigenvalues
n2 π 2
λn = , n = 0, 1, 2, · · · (75)
L2
and the corresponding eigenfunctions
³ nπx ´
φn = cos , n = 0, 1, 2, · · · (76)
L

On the other hand, it follows from (74) that

dG
= −λkG, (77)
dt
26
which has solutions
kn2 π 2 t
G(t) = ce−λkt = ce− L2 .
We then derive the infinite series solution:
∞
X ³ nπx ´ kn2 π 2 t
u(x, t) = a0 + an cos e− L2 .
n=1
L

The initial condition gives

∞
X ³ nπx ´
f (x) = u(x, 0) = a0 + an cos . (78)
n=1
L
¡ nπx ¢
To determine an , we multiply (78) by cos L and integrate from 0 to L. We then
find
Z
1 L
a0 = f (x)dx, (79)
L 0
Z ³ nπx ´
2 L
an = f (x) cos dx, n ≥ 1. (80)
L 0 L

Example. Solve the following initial boundary value problem

∂u ∂2u
= k 2, (81)
∂t dx
∂u ∂u
(0, t) = 0, (L, t) = 0, (82)
dx dx
u(x, 0) = x. (83)

Solution. The Fourier coefficient are

Z
1 L L
a0 = xdx = ,
L 0 2
Z L ³
2 nπx ´
an = x cos dx
L 0 L
³ nπx ´ ¯ Z L ³ nπx ´
2 ¯L 2
= x sin 0
− sin dx
nπ L nπ 0 L
2L ³ nπx ´ ¯
= − 2 2 cos ¯L
nπ L 0

2L(1 − cos (nπ))

= , n = 1, 2, · · · ,
n2 π 2
So the solution is
∞
L X 2L(1 − cos (nπ)) ³ nπx ´ kn2 π2 t
u(x, t) = + cos e− L2 .
2 n=1 n2 π 2 L

27
3. Periodic boundary value problem:

∂u d2 u
= k 2, (84)
∂t dx
du du
u(−L, t) = u(L, t), (−L, t) = (L, t), (85)
dx dx
u(x, 0) = f (x). (86)

Look for a solution of the form of separation of variables:

u(x, t) = φ(x)G(t), (87)

Substitute the above expression into the equation (84), we obtain

φ(x)G0 (t) = kφ00 (x)G(t),

and then
G0 (t) φ00 (x)
= = −λ, (88)
kG(t) φ(x)
where λ is constant to be determined. The boundary condition (85) yields that
φ φ
φ(−L) = φ(L), (−L) = (L).
dx dx
We then have an eigenvalue problem

d2 φ
− = λφ,
dx2
φ φ
φ(−L) = φ(L), (−L) = (L),
dx dx
which has the eigenvalues
n2 π 2
λn = , n = 0, 1, 2, · · · (89)
L2
and the corresponding eigenfunctions
³ nπx ´ ³ nπx ´
φ0 = 1, φ2n−1 = cos , φ2n = sin , n = 1, 2, · · · (90)
L L

On the other hand, it follows from (88) that

dG
= −λkG, (91)
dt
which has solutions
kn2 π 2 t
G(t) = ce−λkt = ce− L2 .

28
 We then derive the infinite series solution:
X∞ ³ nπx ´ kn2 π2 t
u(x, t) = a0 + an cos e− L2 .
n=1
L
The initial condition gives
∞ h
X ³ nπx ´ ³ nπx ´i
f (x) = u(x, 0) = a0 + an cos + bn sin . (92)
n=1
L L
¡ ¢ ¡ nπx ¢
To determine the coefficients, we multiply (92) by cos nπx L
or cos L
and integrate
from −L to L. We then find
Z L
1
a0 = f (x)dx, (93)
2L −L
Z ³ nπx ´
1 L
an = f (x) cos dx, n ≥ 1, (94)
L −L L
Z ³ nπx ´
1 L
bn = f (x) sin dx, n ≥ 1. (95)
L −L L
Example. Solve the following initial boundary value problem
∂u ∂ 2u
= k 2, (96)
∂t dx
∂u ∂u
u(−L, t) = u(L, t), (−L, t) = (L, t), (97)
dx dx
u(x, 0) = x2 . (98)

Solution. The Fourier coefficient are

Z L
1 L2
a0 = x2 dx = ,
2L −L 3
Z ³ nπx ´
1 L 2
an = x cos dx
L −L L
³ nπx ´ ¯ Z L ³ nπx ´
1 2 L
¯ − 2
= x sin −L
x sin dx
nπ L nπ −L L
³ nπx ´ ¯ Z L ³ nπx ´
2L ¯ L 2L
= x cos −L
− cos dx
n2 π 2 L n2 π 2 −L L
4L2 cos (nπ)
= , n = 1, 2, · · · ,
n2 π 2
bn = 0, n = 1, 2, · · · .
So the solution is
∞
L2 X 4L2 cos (nπ) ³ nπx ´ kn2 π2 t
u(x, t) = + 2π2
cos e− L2 .
3 n=1
n L

4. Exercise 2.4: 2.4.1 (a).

29
9 Lecture 8 – Laplace’s equation
Today:
• Laplace’s equation.
Next:
• Review of the heat equation.
Review:
• Trigonometric identities:
1
cos u cos v = [cos(u − v) + cos(u + v)].
2
1
cos2 u = [1 + cos(2u)].
2
• The Neumann eigenvalue problem
d2 φ
− = λφ,
dx2
dφ dφ
(0) = 0, (L) = 0
dx dx
has the eigenvalues
n2 π 2
λn = , n = 0, 1, 2, · · · (99)
L2
and the corresponding eigenfunctions
³ nπx ´
φn = cos , n = 0, 1, 2, · · · (100)
L

Teaching procedure:

1. Solve Laplace’s equation

∂ 2u ∂ 2u
+ = 0, (101)
∂x2 ∂y 2
∂u ∂u
(0, y) = 0, (L, y) = 0, u(x, 0) = 0, u(x, H) = f (x), (102)
∂x ∂x

Look for a solution of the form of separation of variables:

u(x, y) = φ(x)h(y), (103)

Substitute the above expression into the equation (101), we obtain

φ00 (x)h(y) + φ(x)h00 (y) = 0,

30
 and then
h00 (y) φ00 (x)
=− = λ, (104)
h(y) φ(x)
where λ is constant to be determined. The boundary condition yields that
φ φ
(0) = (L) = 0, h(0) = 0.
dx dx
We then have an eigenvalue problem

d2 φ
− = λφ,
dx2
φ φ
(0) = 0, (L) = 0,
dx dx
which has the eigenvalues
n2 π 2
λn = , n = 0, 1, 2, · · · (105)
L2
and the corresponding eigenfunctions
³ nπx ´
φn = cos , n = 0, 1, 2, · · · (106)
L
On the other hand, it follows from (104) that
n2 π 2
h00 (y) = h(y), h(0) = 0. (107)
L2
The general solutions are nπy nπy
h(y) = c1 e L + c2 e− L .
The boundary condition h(0) = 0 gives
c1 + c2 = 0.
So ³ nπy nπy
´
h(y) = c1 e L − e− L .
We then derive the infinite series solution:
X∞ ³ nπx ´ ³ nπy nπy
´
u(x, t) = an cos e L − e− L .
n=1
L
The boundary condition u(x, H) = f (x) gives
X∞ ³ nπx ´ ³ nπH nπH
´
f (x) = u(x, H) = an cos e L − e− L . (108)
n=1
L
¡ ¢
To determine an , we multiply the above by cos nπx L
and integrate from 0 to L. We
then find
³ nπH ´
2 e L − e− L Z L
nπH
³ nπx ´
an = f (x) cos dx, n ≥ 1. (109)
L 0 L
2. Exercise 2.5.1 (b).

31
10 Review of the heat equation and Laplace’s equation
1. The heat equation

∂u ∂ 2u Q
=k 2 + . (110)
∂t ∂x cρ
2. Laplace’s equation

∂2u ∂ 2u
+ =0 (111)
∂x2 ∂y 2
3. Equilibrium:

∂ 2u Q
k + = 0. (112)
∂x2 cρ
4. The eigenvalue problem

d2 φ
− = λφ,
dx2
φ(0) = 0, φ(L) = 0

has the eigenvalues

n2 π 2
λn = , n = 1, 2, · · · (113)
L2
and the corresponding eigenfunctions
³ nπx ´
φn = sin , n = 1, 2, · · · (114)
L

5. The Neumann eigenvalue problem

d2 φ
− = λφ,
dx2
dφ dφ
(0) = 0, (L) = 0
dx dx
has the eigenvalues
n2 π 2
λn = , n = 0, 1, 2, · · · (115)
L2
and the corresponding eigenfunctions
³ nπx ´
φn = cos , n = 0, 1, 2, · · · (116)
L

6. separation of variables:
u(x, t) = φ(x)G(t). (117)

32
 7. Dirichlet boundary value problems:

∂u ∂ 2u
= k 2, (118)
∂t ∂x
u(0, t) = 0, u(L, t) = 0, (119)
u(x, 0) = f (x). (120)
has infinite series solution:
∞
X ³ nπx ´ kn2 π 2 t
u(x, t) = cn sin e− L2 ,
n=1
L

where Z
2 L ³ nπx ´
cn = f (x) sin dx.
L 0 L
8. Neumann boundary value problem:

∂u ∂2u
= k 2, (121)
∂t ∂x
∂u ∂u
(0, t) = 0, (L, t) = 0, (122)
dx dx
u(x, 0) = f (x). (123)
has the infinite series solution:
∞
X ³ nπx ´ kn2 π 2 t
u(x, t) = a0 + an cos e− L2 ,
n=1
L

where
Z
1 L
a0 = f (x)dx, (124)
L 0
Z ³ nπx ´
2 L
an = f (x) cos dx, n ≥ 1. (125)
L 0 L

9. Separation of variables
u(x, y) = φ(x)h(y), (126)
for Laplace’s equation

∂ 2u ∂ 2u
+ = 0, (127)
∂x2 ∂y 2
∂u ∂u
(0, y) = 0, (L, y) = 0, u(x, 0) = 0, u(x, H) = f (x), (128)
∂x ∂x
10. Review Exercises: 1.4.1 (d), 2.3.2(b), 2.3.3 (d), 2.5.1 (e)

33
11 Lecture 9 – Convergence of Fourier Series
Today:
• Piecewise smooth functions.
• Periodic extension.
• Convergence of Fourier Series
Next:
• Since and cosine series.
Review:
• The periodic eigenvalue problem
d2 φ
− = λφ,
dx2
dφ dφ
φ(−L) = φ(L), (−L) = (L)
dx dx
has the eigenvalues
n2 π 2
λn = , n = 0, 1, 2, · · · (129)
L2
and the corresponding eigenfunctions
³ nπx ´ ³ nπx ´
φ0 = 1, φ2n−1 = cos , φ2n = sin , n = 1, 2, · · · (130)
L L

Teaching procedure:

1. Orthogonality of the set of eigenfunctions

n ³ πx ´ ³ πx ´ ³ nπx ´ ³ nπx ´ o
1, cos , sin , · · · , cos , sin ,··· .
L L L L
Proof. Let φ1 and φ2 be two different eigenfunctions corresponding to two different
eigenvalues λ1 and λ2 . Multiplying the equation
d2 φ1
− = λ1 φ1
dx2
by φ2 and integrating from −L to L yield
Z L Z L
d2 φ1
− φ2 2 dx = λ1 φ1 φ2 dx.
−L dx −L

Integration by parts with the periodic boundary conditions gives

Z L Z L
dφ1 dφ2
dx = λ1 φ1 φ2 dx. (131)
−L dx dx −L

34
 Multiplying the equation
d2 φ2
− = λ2 φ2
dx2
by φ1 and integrating from −L to L yield
Z L Z L
d2 φ2
− φ1 2 dx = λ2 φ1 φ2 dx.
−L dx −L

Integration by parts with the periodic boundary conditions gives

Z L Z L
dφ1 dφ2
dx = λ2 φ1 φ2 dx. (132)
−L dx dx −L

Subtracting (132) from (131) gives

Z L
(λ1 − λ2 ) φ1 φ2 dx = 0,
−L

which implies that Z L

φ1 φ2 dx = 0.
−L

So φ1 is orthogonal to φ2 .

2. Fourier series of f (x) on [−L, L]:

∞ h
X ³ nπx ´ ³ nπx ´i
f (x) ∼ a0 + an cos + bn sin . (133)
n=1
L L
Fourier coefficients
Z L
1
a0 = f (x)dx, (134)
2L −L
Z ³ nπx ´
1 L
an = f (x) cos dx, n ≥ 1, (135)
L −L L
Z ³ nπx ´
1 L
bn = f (x) sin dx, n ≥ 1. (136)
L −L L

Example. ½
0, x ≤ 0,
f (x) =
x, x > 0.
Fourier series of f (x) on [π, π]:
∞ · ¸
π X (−1)n − 1 (−1)n+1
f (x) ∼ + cos(nx) + sin(nx) .
4 n=1 πn2 n

35
3. Piecewise smooth function: f (x) is piecewise function if the interval can be broken
up into pieces such that in each piece f 0 (x) is continuous.
½
0, x ≤ 0,
f (x) =
x, x > 0.
4. Periodic extension.

5. Fourier Theorem. Let f (x) be a piecewise smooth function on the interval [−L, L]
and F (x) denote the periodic extension of f with period 2L. Then the Fourier series
of f (x) converges to
1
[f (x+) + f (x−)] if − L < x < L
2
or to
1
[F (x+) + F (x−)] if x ≤ −L or x ≥ L.
2
Example. ½
0, x ≤ 0,
f (x) =
x, x > 0.
 π
∞ · ¸  2, x = −π
π X (−1) − 1 n
(−1) n+1
+ cos(nx) + sin(nx) = f (x), −π < x < π
4 n=1 πn2 n  π
2
, x=π
Taking x = π, we have the identity
∞
π X (−1)n − 1 π
+ 2
(−1)n = ,
4 n=1 πn 2
which can be simplified to
∞
X 1 π2
= .
n=1
(2n − 1)2 8
This provide a method to compute an approximate value of π.

6. Exercise 3.2.1 (f).

36
12 Lecture 10 – Fourier Since and cosine series
Today:
• Since and cosine series.
Next:
• Term-by-term differentiation.
Review:
RL
• −L f (x)dx = 0 if f is odd.
RL RL
• −L
f (x)dx = 2 0
f (x)dx if f is even.

• Fourier series of f (x) on [−L, L]:

∞ h
X ³ nπx ´ ³ nπx ´i
f (x) ∼ a0 + an cos + bn sin . (137)
n=1
L L
Fourier coefficients
Z L
1
a0 = f (x)dx, (138)
2L −L
Z ³ nπx ´
1 L
an = f (x) cos dx, n ≥ 1, (139)
L −L L
Z ³ nπx ´
1 L
bn = f (x) sin dx, n ≥ 1. (140)
L −L L

Teaching procedure:

1. Fourier sine series: If f is odd, that is, f (−x) = −f (x), then

Z L
1
a0 = f (x)dx = 0, (141)
2L −L
Z ³ nπx ´
1 L
an = f (x) cos dx = 0, n ≥ 1, (142)
L −L L
Z ³ nπx ´ Z ³ nπx ´
1 L 2 L
bn = f (x) sin dx = f (x) sin , n ≥ 1. (143)
L −L L L 0 L

So ∞
X ³ nπx ´
f (x) ∼ bn sin . (144)
n=1
L

Example.

37
 ∞
X ³ nπx ´
x∼ bn sin , (145)
n=1
L
where
Z L ³ nπx ´
2
bn = x sin , n ≥ 1. (146)
L 0 L

Sketch the series:

2. Fourier cosine series: If f is even, that is, f (−x) = f (x), then

Z L Z
1 1 L
a0 = f (x)dx = f (x)dx, (147)
2L −L L 0
Z ³ nπx ´ Z ³ nπx ´
1 L 2 L
an = f (x) cos dx = f (x) cos dx, n ≥ 1, (148)
L −L L L 0 L
Z ³ nπx ´
1 L
bn = f (x) sin dx = 0, n ≥ 1. (149)
L −L L

So ∞
X ³ nπx ´
f (x) ∼ a0 + an cos . (150)
n=1
L

Example.

∞
X ³ nπx ´
2
x ∼ a0 + an cos , (151)
n=1
L
where
Z
1 L 2
a0 = x dx, (152)
L 0
Z ³ nπx ´
2 L 2
an = x cos dx, n ≥ 1, (153)
L 0 L

Sketch the series:

38
3. Odd and even extension of a function.
½
f (x), x>0
odd extension of f (x) = (154)
−f (−x) x<0
½
f (x), x>0
even extension of f (x) = (155)
f (−x) x<0
Example. ex . Sketch the extensions:

4. Comparison among Fourier series, Fourier since series, and Fourier cosine series.

• Fourier series of ex on [−L, L]:

∞ h
X ³ nπx ´ ³ nπx ´i
x
e ∼ a0 + an cos + bn sin . (156)
n=1
L L
where
Z L
1
a0 = ex dx, (157)
2L −L
Z ³ nπx ´
1 L x
an = e cos dx, n ≥ 1, (158)
L −L L
Z ³ nπx ´
1 L x
bn = e sin dx, n ≥ 1. (159)
L −L L

Sketch the series:

• Fourier since series of ex on [0, L]:

∞
X ³ nπx ´
x
e ∼ bn sin 0 ≤ x ≤ L, (160)
n=1
L
∞
X ³ nπx ´
x
odd extension of e ∼ bn sin , −L ≤ x ≤ L, (161)
n=1
L

39
 where
Z L ³ nπx ´
2 x
bn = e sin dx, n ≥ 1. (162)
L 0 L
Sketch the series:

• Fourier cosine series of ex on [0, L]:

∞
X ³ nπx ´
ex ∼ a0 + an cos 0 ≤ x ≤ L, (163)
n=1
L
∞
X ³ nπx ´
even extension of ex ∼ a0 + an cos −L≤x≤L (164)
n=1
L
where
Z
1 L x
a0 = e dx, (165)
L 0
Z ³ nπx ´
2 L x
an = e cos dx, n ≥ 1. (166)
L 0 L
Sketch the series:

5. Physical examples.

• Fourier sine series for Dirichlet Boundary conditions:

∂u ∂ 2u
= k 2, (167)
∂t ∂x
u(0, t) = 0, u(L, t) = 0, (168)
u(x, 0) = f (x). (169)
has infinite series solution:
∞
X ³ nπx ´ kn2 π 2 t
u(x, t) = cn sin e− L2 ,
n=1
L

40
 ∞
X ³ nπx ´
f (x) = cn sin ,
n=1
L
where Z
2 L ³ nπx ´
cn = f (x) sin dx.
L 0 L
• Fourier cosine series for Neumann boundary conditions:

∂u ∂ 2u
= k 2, (170)
∂t ∂x
∂u ∂u
(0, t) = 0, (L, t) = 0, (171)
dx dx
u(x, 0) = f (x). (172)

has the infinite series solution:

∞
X ³ nπx ´ kn2 π 2 t
u(x, t) = a0 + an cos e− L2 ,
n=1
L

∞
X ³ nπx ´
f (x) = a0 + an cos ,
n=1
L
where
Z
1 L
a0 = f (x)dx, (173)
L 0
Z ³ nπx ´
2 L
an = f (x) cos dx, n ≥ 1. (174)
L 0 L

• Fourier series for periodic boundary condition:

∂u d2 u
= k 2, (175)
∂t dx
du du
u(−L, t) = u(L, t), (−L, t) = (L, t), (176)
dx dx
u(x, 0) = f (x) (177)

has the infinite series solution:

∞ h
X ³ nπx ´ ³ nπx ´i kn2 π 2 t
u(x, t) = a0 + an cos + bn sin e− L2 ,
n=1
L L

∞ h
X ³ nπx ´ ³ nπx ´i
f (x) = u(x, 0) = a0 + an cos + bn sin . (178)
n=1
L L

41
 where
Z L
1
a0 = f (x)dx, (179)
2L −L
Z ³ nπx ´
1 L
an = f (x) cos dx, n ≥ 1, (180)
L −L L
Z ³ nπx ´
1 L
bn = f (x) sin dx, n ≥ 1. (181)
L −L L

6. Even and odd parts of a function.

Even part of f (x):
1
fe (x) = [f (x) + f (−x)].
2
Odd part of f (x):
1
fo (x) = [f (x) − f (−x)].
2
Example: ex .

7. Numerical evidence for convergence of Fourier series.

∞
X ³ nπx ´
100 = cn sin ,
n=1
L
where Z ½
2 L ³ nπx ´ 0, n even
cn = 100 sin dx = 400
L 0 L nπ
, n odd.
Set L = 1: µ ¶
400 sin πx sin 3πx sin 5πx
100 = + + + ··· .
π 1 3 5

8. Exercise 3.3.2 (b).

42
13 Lecture 11 – One-dimensional wave equation (Sec-
tion 4.4)
Today:
• One-dimensional wave equation
Next:
• Two-dimensional wave equation (Section 7.3).
Review:
• The eigenvalue problem

d2 φ
− = λφ,
dx2
φ(0) = 0, φ(L) = 0

has the eigenvalues

n2 π 2
λn = , n = 1, 2, · · · (182)
L2
and the corresponding eigenfunctions
³ nπx ´
φn = sin , n = 1, 2, · · · (183)
L

Teaching procedure:

1. Physical background of the wave equation

2. Dirichlet Initial boundary value problem:

∂2u 2
2∂ u
= c , (184)
∂t2 ∂x2
u(0, t) = 0, u(L, t) = 0, (185)
∂u
u(x, 0) = f (x), t(x, 0) = g(x). (186)
∂

Look for a solution of the form of separation of variables:

u(x, t) = φ(x)h(t), (187)

43
Substituting the above expression into the equation (184), we obtain

and then
h00 (t) φ00 (x)
= = −λ, (188)
c2 G(t) φ(x)
where λ is constant to be determined. The boundary condition (185) yields that

We then have an eigenvalue problem

d2 φ
− = λφ,
dx2
φ(0) = 0, φ(L) = 0,

which has the eigenvalues

and the corresponding eigenfunctions

On the other hand, it follows from (188) that

d2 h c2 n2 π 2
= − h, (189)
dt2 L2
which has solutions

we then have infinite series solution:

44
 The initial condition gives

f (x) = u(x, 0) =
∂u
g(x) = (x, 0) =
∂t
where

an =
bn =

3. Example. Solve the following initial boundary value problem

∂ 2u 2
2∂ u
= c ,
∂t2 ∂x2
u(0, t) = 0, u(π, t) = 0,
∂u
u(x, 0) = 1, (x, 0) = x.
∂t

45
14 Lecture 12 – Higher-dimensional heat and wave equa-
tion (Section 7.3)
Today:

• Higher-dimensional heat equation

• Higher-dimensional wave equation

Next:

• Three-dimensional Laplace’s equation (Section 7.3).

Review:

• The eigenvalue problem

d2 φ
− = λφ,
dx2
φ(0) = 0, φ(L) = 0

has the eigenvalues

n2 π 2
λn = , n = 1, 2, · · · (190)
L2
and the corresponding eigenfunctions
³ nπx ´
φn = sin , n = 1, 2, · · · (191)
L

• The eigenvalue problem

d2 φ
− = λφ,
dx2
φ0 (0) = 0, φ0 (L) = 0

has the eigenvalues

n2 π 2
λn = , n = 0, 1, 2, · · · (192)
L2
and the corresponding eigenfunctions
³ nπx ´
φn = cos , n = 0, 1, 2, · · · (193)
L

Teaching procedure:

1. Solve initial boundary value problem:

46
 µ ¶
∂u ∂ 2u ∂2u
= k + , (194)
∂t ∂x2 ∂y 2
∂u ∂u
(0, y, t) = 0, (L, y, t) = 0, u(x, 0, t) = 0, u(x, H, t) = 0, (195)
∂x ∂x
u(x, y, 0) = f (x, y). (196)

Look for a solution of the form of separation of variables:

u(x, y, t) = φ(x, y)h(t). (197)

Substituting the above expression into the equation (194), we obtain

and then 2 2
∂ φ
h0 (t) ∂x2
+ ∂∂yφ2
= = −λ, (198)
kh(t) φ(x, y)
where λ is constant to be determined. The boundary condition (196) yields that

We then have an eigenvalue problem

To solve this eigenvalue problem, we have to separate x and y again by letting

φ(x, y) = α(x)β(y).

Plugging it into the eigenvalue problems yields

and then

47
which gives two eigenvalue problems

The first one has the eigenvalues

and the corresponding eigenfunctions

The second one has the eigenvalues

and the corresponding eigenfunctions

Thus the original two-dimensional eigenvalue problem has the eigenvalues

and the corresponding eigenfunctions

On the other hand, we have

µ ¶
dh n2 π 2 m2 π 2
= −k + h, (199)
dt L2 H2

which has solutions

48
 we then have infinite series solution:

The initial condition gives

f (x) = u(x, y, 0) =

where
amn =

2. Example. Solve initial boundary value problem:

µ ¶
∂u ∂ 2u ∂ 2u
= 0.01 + , (200)
∂t ∂x2 ∂y 2
∂u ∂u
(0, y, t) = 0, (L, y, t) = 0, u(x, 0, t) = 0, u(x, H, t) = 0, (201)
∂x ∂x
u(x, y, 0) = y(1 − y) sin(πx). (202)
See Figure 1.
3. Solve Initial boundary value problem:

µ ¶
∂ 2u ∂ 2u ∂2u
2
= c2 + , (203)
∂t ∂x2 ∂y 2
∂u ∂u
(0, y, t) = 0, (L, y, t) = 0, u(x, 0, t) = 0, u(x, H, t) = 0, (204)
∂x ∂x
∂u
u(x, y, 0) = f (x, y), (x, y, 0) = g(x, y). (205)
∂t
Look for a solution of the form of separation of variables:
u(x, y, t) = φ(x, y)h(t). (206)

Substituting the above expression into the equation (203), we obtain

and then 2 2
∂ φ
h00 (t) ∂x2
+ ∂∂yφ2
= = −λ, (207)
c2 h(t) φ(x, y)

49
 Time =0 Time =2

0.3 0.3

0.25 0.25

0.2 0.2

0.15 0.15
u

u
0.1 0.1

0.05 0.05

0 0
1 1
0.8 1 0.8 1
0.6 0.8 0.6 0.8
0.4 0.6 0.4 0.6
0.4 0.4
0.2 0.2
0.2 0.2
0 0 0 0
y x y x

Time =19.5 Time =39.5

0.3 0.3

0.25 0.25

0.2 0.2

0.15 0.15
u

0.1 0.1

0.05 0.05

0 0
1 1
0.8 1 0.8 1
0.6 0.8 0.6 0.8
0.4 0.6 0.4 0.6
0.4 0.4
0.2 0.2
0.2 0.2
0 0 0 0
y x y x

Figure 1: Solution of 2D heat equation (200).

where λ is constant to be determined. The boundary condition (205) yields that

We then have an eigenvalue problem

To solve this eigenvalue problem, we have to separate x and y again by letting

φ(x, y) = α(x)β(y).

Plugging it into the eigenvalue problems yields

50
and then

which gives two eigenvalue problems

The first one has the eigenvalues

and the corresponding eigenfunctions

The second one has the eigenvalues

and the corresponding eigenfunctions

Thus the original two-dimensional eigenvalue problem has the eigenvalues

and the corresponding eigenfunctions

51
 Time =0 Time =0.25

0.3 0.3

0.2 0.2

0.1 0.1

0 0

u
−0.1 −0.1

−0.2 −0.2

1 1
0.8 1 0.8 1
0.6 0.8 0.6 0.8
0.4 0.6 0.4 0.6
0.4 0.4
0.2 0.2
0.2 0.2
0 0 0 0
y x y x

Time =0.5 Time =0.75

0.3 0.3

0.2 0.2

0.1 0.1

0 0
u

u
−0.1 −0.1

−0.2 −0.2

1 1
0.8 1 0.8 1
0.6 0.8 0.6 0.8
0.4 0.6 0.4 0.6
0.4 0.4
0.2 0.2
0.2 0.2
0 0 0 0
y x y x

Time =1 Time =1.25

0.3 0.3

0.2 0.2

0.1 0.1

0 0
u

−0.1 −0.1

−0.2 −0.2

1 1
0.8 1 0.8 1
0.6 0.8 0.6 0.8
0.4 0.6 0.4 0.6
0.4 0.4
0.2 0.2
0.2 0.2
0 0 0 0
y x y x

Figure 2: Solution of 2D wave equation (209).

On the other hand, we have

µ ¶
d2 h n2 π 2 m2 π 2
2
= −c2 + h, (208)
dt L2 H2

which has solutions

we then have infinite series solution:

52
 The initial condition gives

f (x) = u(x, y, 0) =
∂u
g(x) = (x, y, 0) =
∂t

where

amn =
bmn =

4. Example. Solve the initial boundary valume problem

∂ 2u ∂ 2u ∂ 2u
= + , (209)
∂t2 ∂x2 ∂y 2
∂u ∂u
(0, y, t) = 0, (L, y, t) = 0, u(x, 0, t) = 0, u(x, H, t) = 0, (210)
∂x ∂x
∂u
u(x, y, 0) = y(1 − y) sin(πx), (x, y, 0) = y(1 − y) sin(πx). (211)
∂t

See Figure 2.

53
15 Lecture 13 – Higher-dimensional eigenvalue prob-
lems (Section 7.4)
Today:

• Higher-dimensional eigenvalue problems

Next:

• Problem-solving (Assignment 6).

Teaching procedure:

1. Eigenvalue problem in rectanglar domain

∂ 2φ ∂ 2φ
+ = −λφ(x, y), (212)
∂x2 ∂y 2
∂φ ∂φ
(0, y) = (L, y) = φ(x, 0) = φ(x, H) = 0, (213)
∂x ∂x
has the eigenvalues
n2 π 2 m2 π 2
λm,n = +
L2 H2
and the corresponding eigenfunctions
³ nπx ´ ³ mπx ´
φm,n = cos sin , m = 1, 2, · · · , n = 0, 1, 2, · · · .
L H

2. Eigenvalue problem in any domain

∂ 2φ ∂ 2φ
+ = −λφ(x, y), (214)
∂x2 ∂y 2
φ = 0 on the boundary (215)

See the result on page 290.

54
16 Main Technique Practice
1. Useful trig identies
1
sin u sin v = [cos(u − v) − cos(u + v)],
2
1
cos u cos v = [cos(u − v) + cos(u + v)],
2
1
sin u cos v = [sin(u − v) + sin(u + v)].
2
2. Integrals:

Z
− cos ax
sin axdx = + C,
a
Z
sin ax
cos axdx = + C.
a

3. The solution of h0 = ah is
4. The solution of h00 = −a2 h is
5. Compute
RL ¡ ¢
• 0 sin2 nπx L
dx =
RL ¡ ¢
• 0 cos2 nπx L
dx =
RL ¡ ¢ ¡ ¢
• If m 6= n, then 0 sin nπx L
sin mπxL
dx =
RL ¡ nπx ¢ ¡ mπx ¢
• If m 6= n, then 0 cos L cos L dx =
RL ¡ ¢ ¡ ¢
• −L sin nπx L
cos mπx
L
dx =
¡ ¢ ¡ ¢ ¡ ¢
6. If f (x) = a1 sin πx
L
+ a2 sin 2πxL
+ · · · + an sin nπx
L
+ · · · , then an =

¡ πx ¢ ¡ 2πx ¢ ¡ nπx ¢
7. If f (x) = a0 + a1 cos L
+ a2 cos L
+ · · · + an cos L
+ · · · , then an =

55
8. Let u(x, t) = φ(x)h(t).

• If u(0, t) = 0, then φ(0) =

• If u(L, t) = 0, then φ(L) =
∂u ∂φ
• If ∂x
(0, t) = 0, then ∂x
(0) =
∂u ∂φ
• If ∂x
(L, t) = 0, then ∂x
(L) =
∂u
• ∂t
=
∂2u
• ∂x2
=
∂u 2
• If ∂t
= k ∂∂xu2 , then the φ equation and h equation are

∂2u 2
• If ∂t2
= c2 ∂∂xu2 , then the φ equation and h equation are

9. Let φ(x, y) = α(x)β(y).

• If φ(0, y) = 0, then α(0) =

• If φ(L, y) = 0, then α(L) =
∂φ ∂α
• If ∂x
(0, y) = 0, then ∂x
(0) =
∂φ ∂α
• If ∂x
(L, y) = 0, then ∂x
(L) =
• If φ(x, 0) = 0, then β(0) =
• If φ(x, H) = 0, then β(H) =
∂φ ∂β
• If ∂y
(x, 0) = 0, then ∂y
(0) =
∂φ ∂β
• If ∂y
(x, H) = 0, then ∂y
(H) =
∂2φ
• ∂x2
=
∂2φ
• ∂y 2
=
∂2φ ∂2φ
• If ∂x2
+ ∂y 2
= −λφ, then the α equation and β equation are

56
17 Lecture 14 – Nonhomogeneous Problems for the
Heat Equation (Section 8.2)
Today:
• Time-independent boundary conditions
• Steady nonhomogeneous equation
• Time-depedendent nonhomogeneous terms
Next:
• Nonhomogeneous Problems for the Wave Equation.
Review:
• Dirichlet boundary value problems:

∂u ∂ 2u
= k 2, (216)
∂t ∂x
u(0, t) = 0, u(L, t) = 0, (217)
u(x, 0) = f (x). (218)
has infinite series solution:
∞
X ³ nπx ´ kn2 π 2 t
u(x, t) = an sin e− L2 .
n=1
L
where Z
2 L ³ nπx ´
an = f (x) sin dx.
L 0 L
Teaching procedure:
1. Time-independent boundary conditions:

∂u ∂2u
= k 2, (219)
∂t ∂x
u(0, t) = A, u(L, t) = B, (220)
u(x, 0) = f (x). (221)

2. Steady-state heat equation

d2 ue
k = 0, (222)
dx2
ue (0) = A, ue (L) = B. (223)
has the unique solution
B−A
ue (x) = A + x.
L

57
3. Conversion of the nonhomogeneous to the homogeneous: Define v(x, t) = u(x, t) −
ue (x). Then v satisfies

∂v ∂ 2v
= k 2, (224)
∂t ∂x
v(0, t) = 0, v(L, t) = 0, (225)
v(x, 0) = f (x) − ue (x). (226)

which has a unique solution

∞
X ³ nπx ´ kn2 π 2 t
v(x, t) = an sin e− L2 .
n=1
L

where Z
2 L ³ nπx ´
an = [f (x) − ue (x)] sin dx.
L 0 L
So the solution of the original nonhomogeneous equation is

∞
X ³ nπx ´ kn2 π 2 t
u(x, t) = ue (x) + an sin e− L2 .
n=1
L

4. Steady nonhomogeneous equation:

∂u ∂2u
= k 2 + Q(x), (227)
∂t ∂x
u(0, t) = A, u(L, t) = B, (228)
u(x, 0) = f (x). (229)

5. Steady-state heat equation

d2 ue
k = −Q(x), (230)
dx2
ue (0) = A, ue (L) = B. (231)

has the unique solution ue (x).

6. Conversion of the nonhomogeneous to the homogeneous: Define v(x, t) = u(x, t) −

ue (x). Then v satisfies

∂v ∂ 2v
= k 2, (232)
∂t ∂x
v(0, t) = 0, v(L, t) = 0, (233)
v(x, 0) = f (x) − ue (x). (234)

58
 which has a unique solution
∞
X ³ nπx ´ kn2 π 2 t
v(x, t) = an sin e− L2 .
n=1
L

where Z
2 L ³ nπx ´
an = [f (x) − ue (x)] sin dx.
L 0 L
So the solution of the original nonhomogeneous equation is

∞
X ³ nπx ´ kn2 π 2 t
u(x, t) = ue (x) + an sin e− L2 .
n=1
L

B−A−sin L
7. Example. If Q(x) = k sin x, then ue (x) = A + L
x + sin x.

8. Time-dependent nonhomogeneous problems:

∂u ∂ 2u
= k 2 + Q(x, t), (235)
∂t ∂x
u(0, t) = A(t), u(L, t) = B(t), (236)
u(x, 0) = f (x). (237)

9. Find a reference temperature distribution r(x, t) such that

r(0, t) = A(t), r(L, t) = B(t).

For example,
B(t) − A(t)
r(x, t) = A(t) + x.
L
10. Conversion of the nonhomogeneous to the homogeneous: Define v(x, t) = u(x, t) −
r(x, t). Then v satisfies

∂v ∂2v ∂r ∂ 2r
= k 2 + Q(x, t) − + k 2, (238)
∂t ∂x ∂t ∂x
v(0, t) = 0, v(L, t) = 0, (239)
v(x, 0) = f (x) − r(x, 0). (240)

11. Exercises 8.2.1. (c).

59
18 Lecture 15 – Nonhomogeneous Problems for the
Wave Equation and Eigenfunction expansion (Sec-
tion 8.3)
Today:
• Nonhomogeneous Problems for the Wave Equation
• Eigenfunction expansion

Next:
• Eigenfunction expansion.

Review:

• Dirichlet boundary value problems:

∂ 2u 2
2∂ u
= c , (241)
∂t2 ∂x2
u(0, t) = 0, u(L, t) = 0, (242)
∂u
u(x, 0) = f (x), (x, 0) = g(x). (243)
∂t
has infinite series solution:
X∞ µ µ ¶ µ ¶¶ ³ nπx ´
cnπt cnπt
u(x, t) = an cos + bn sin sin .
n=1
L L L

where Z
2 L ³ nπx ´
an = f (x) sin
dx,
L 0 L
Z ³ nπx ´
cnπ 2 L
bn = g(x) sin dx,
L L 0 L
Teaching procedure:

1. Time-independent nonhomogeneous problem:

∂ 2u 2
2∂ u
= c + Q(x), (244)
∂t2 ∂x2
u(0, t) = A, u(L, t) = B, (245)
∂u
u(x, 0) = f (x), (x, 0) = g(x). (246)
∂t

60
2. Steady-state wave equation
d2 ue
c2 = −Q(x), (247)
dx2
ue (0) = A, ue (L) = B. (248)
has the unique solution ue (x).
3. Conversion of the nonhomogeneous to the homogeneous: Define v(x, t) = u(x, t) −
ue (x). Then v satisfies
∂ 2v 2
2∂ v
= c , (249)
∂t2 ∂x2
v(0, t) = 0, v(L, t) = 0, (250)
∂v
v(x, 0) = f (x) − ue (x), (x, 0) = g(x). (251)
∂t
which has a unique solution
X∞ µ µ ¶ µ ¶¶ ³ nπx ´
cnπt cnπt
v(x, t) = an cos + bn sin sin .
n=1
L L L

where Z
2 L ³ nπx ´
an = [f (x) − ue (x)] sin dx,
L 0 L
Z ³ nπx ´
cnπ 2 L
bn = g(x) sin dx,
L L 0 L
So the solution of the original nonhomogeneous equation is

∞ µ
X µ ¶ µ ¶¶ ³ nπx ´
cnπt cnπt
u(x, t) = ue (x) + an cos + bn sin sin .
n=1
L L L

4. Time-dependent nonhomogeneous problems:

∂ 2u 2
2∂ u
= c + Q(x, t), (252)
∂t2 ∂x2
u(0, t) = A(t), u(L, t) = B(t), (253)
∂u
u(x, 0) = f (x), (x, 0) = g(x). (254)
∂t

5. Find a reference vibration r(x, t) such that

r(0, t) = A(t), r(L, t) = B(t).
For example,
B(t) − A(t)
r(x, t) = A(t) + x.
L

61
6. Conversion of the nonhomogeneous to the homogeneous: Define v(x, t) = u(x, t) −
r(x, t). Then v satisfies
∂ 2v 2
2∂ v ∂2r 2
2∂ r
= c + Q(x, t) − + c , (255)
∂t2 ∂x2 ∂t2 ∂x2
v(0, t) = 0, v(L, t) = 0, (256)
∂v ∂r
v(x, 0) = f (x) − r(x, 0), (x, 0) = g(x) − (x, 0). (257)
∂t ∂t

7. Eigenfunction expansion

∂v ∂ 2v
= k 2 + Q(x, t), (258)
∂t ∂x
v(0, t) = 0, v(L, t) = 0, (259)
v(x, 0) = f (x). (260)
The corresponding eigenvalue problem is
d2 φ
− = λφ, (261)
dx2
φ(0) = 0, φ(L) = 0. (262)

Assume that v(x, t) can be expanded as the series of the eigenfunctions

∞
X
v(x, t) = an (t)φn (x).
n=1

Plugging it into the heat equation gives

∞
X ∞
X ∞
X
a0n (t)φn (t) =k an (t)φ00n (x) + Q(x, t) = −k λn an (t)φn (x) + Q(x, t).
n=1 n=1 n=1

Multiplying the equation by φm (x) and integrating from 0 to L gives

RL
0 Q(x, t)φm (x)dx
am (t) = −kλm am (t) + 0 R L .
φ 2 (x)dx
0 m

To get the initial condition for am , we use the initial condition v(x, 0) = f (x) to get
∞
X
f (x) = an (0)φn (x).
n=1

So RL
0
f (x)φm (x)dx
am (0) = RL , m = 1, 2, · · · ,
φ 2 (x)dx
0 m

Once solving the initial value problem for am , we obtain the solution v(x, t).

62
8. Example. Solve

∂v ∂2v
= k 2 + t, (263)
∂t ∂x
v(0, t) = 0, v(π, t) = 0, (264)
v(x, 0) = sin x. (265)

63
19 Review of Fourier Series and 1D Wave Equation
19.1 Main Topics
1. The basic techniques in the hand-out 16 that must be known.

2. Orthogonality of the set of eigenfunctions

n ³ πx ´ ³ πx ´ ³ nπx ´ ³ nπx ´ o
1, cos , sin , · · · , cos , sin ,··· .
L L L L
n ³ πx ´ ³ nπx ´ o
1, cos , · · · , cos ,··· .
L L
n ³ πx ´ ³ nπx ´ o
sin , · · · , sin ,··· .
L L
3. Fourier series of f (x) on [−L, L]:

∞ h
X ³ nπx ´ ³ nπx ´i
f (x) ∼ a0 + an cos + bn sin . (266)
n=1
L L
Fourier coefficients
Z L
1
a0 = f (x)dx, (267)
2L −L
Z ³ nπx ´
1 L
an = f (x) cos dx, n ≥ 1, (268)
L −L L
Z ³ nπx ´
1 L
bn = f (x) sin dx, n ≥ 1. (269)
L −L L

4. Fourier sine series: If f is odd, that is, f (−x) = −f (x), then

Z L
1
a0 = f (x)dx = 0, (270)
2L −L
Z ³ nπx ´
1 L
an = f (x) cos dx = 0, n ≥ 1, (271)
L −L L
Z ³ nπx ´ Z ³ nπx ´
1 L 2 L
bn = f (x) sin dx = f (x) sin , n ≥ 1. (272)
L −L L L 0 L

So ∞
X ³ nπx ´
f (x) ∼ bn sin . (273)
n=1
L

64
 5. Fourier cosine series: If f is even, that is, f (−x) = f (x), then

Z L Z
1 1 L
a0 = f (x)dx = f (x)dx, (274)
2L −L L 0
Z ³ nπx ´ Z ³ nπx ´
1 L 2 L
an = f (x) cos dx = f (x) cos dx, n ≥ 1, (275)
L −L L L 0 L
Z ³ nπx ´
1 L
bn = f (x) sin dx = 0, n ≥ 1. (276)
L −L L
So ∞
X ³ nπx ´
f (x) ∼ a0 + an cos . (277)
n=1
L

6. Fourier Convergence Theorem.

7. Calculation of Fourier series.
8. Sketch of Fourier series.
9. How to solve Dirichlet initial boundary value problem:

∂ 2u 2
2∂ u
= c , (278)
∂t2 ∂x2
u(0, t) = 0, u(L, t) = 0, (279)
∂u
u(x, 0) = f (x), (x, 0) = g(x). (280)
∂t
10. How to solve Neumann initial boundary value problem:

∂2u 2
2∂ u
= c , (281)
∂t2 ∂x2
∂u ∂u
(0, t) = 0, (L, t) = 0, (282)
∂x ∂x
∂u
u(x, 0) = f (x), (x, 0) = g(x). (283)
∂t
19.2 Review Exercises
Exercises 3.2: 3.2.2. (b), (d); Exercises 3.3: 3.3.1. (b), 3.3.2 (c); 3.3.5 (a).
Solve Dirichlet initial boundary value problem:

∂ 2u 2
2∂ u
= c , (284)
∂t2 ∂x2
u(0, t) = 0, u(1, t) = 0, (285)
∂u
u(x, 0) = x(1 − x), (x, 0) = 0. (286)
∂t
65
Solve Neumann initial boundary value problem:

∂ 2u 2
2∂ u
= c , (287)
∂t2 ∂x2
∂u ∂u
(0, t) = 0, (L, t) = 0, (288)
∂x ∂x
∂u
u(x, 0) = 0, (x, 0) = 1. (289)
∂t
Evaluate
Z L ³ nπx ´ ³ mπx ´
sin sin dx
0 L L
Z L ³ nπx ´ ³ mπx ´
cos cos dx
0 L L
Z L ³ nπx ´ ³ mπx ´
sin cos dx
−L L L

66
20 Lecture 16 – Eigenfunction expansion for the wave
equation
Today:

• Eigenfunction expansion for the wave equation

Next:

• Characteristics for first-order wave equations (Section 12.2).

Review:
• Eigenfunction expansion of the heat equation

∂v ∂ 2v
= k 2 + Q(x, t), (290)
∂t ∂x
v(0, t) = 0, v(L, t) = 0, (291)
v(x, 0) = f (x). (292)

The solution v(x, t) can be expanded as the series of the eigenfunctions

∞
X
v(x, t) = an (t)φn (x),
n=1
¡ nπx ¢ n2 π 2
where φn (x) = sin L
, λn = L2
and
RL
Q(x, t)φn (x)dx
a0n (t) = −kλn an (t) + 0
RL .
φ 2 (x)dx
0 m

RL
0
f (x)φn (x)dx
an (0) = RL , n = 1, 2, · · · ,
φ2 (x)dx
0 n

Teaching procedure:

1. Solve
a00n (t) = −c2 n2 an (t) + bt.

an (0) = b1 , a0n (0) = b2 .

67
2. Eigenfunction expansion for
∂ 2v 2
2∂ v
= c + Q(x, t), (293)
∂t2 ∂x2
v(0, t) = 0, v(L, t) = 0, (294)
∂v
v(x, 0) = f (x), (x, 0) = g(x). (295)
∂t
The corresponding eigenvalue problem is
d2 φ
− = λφ, (296)
dx2
φ(0) = 0, φ(L) = 0, (297)
which has the eigenvalues and eigenfunctions
n2 π 2 ³ nπx ´
λn = , φn (x) = sin .
L2 L
Assume that v(x, t) can be expanded as the series of the eigenfunctions
∞
X
v(x, t) = an (t)φn (x).
n=1

Plugging it into the wave equation gives

∞
X ∞
X ∞
X
a00n (t)φn (t) 2
=c an (t)φ00n (x) + Q(x, t) = −c 2
λn an (t)φn (x) + Q(x, t).
n=1 n=1 n=1

Multiplying the equation by φm (x) and integrating from 0 to L gives

RL
Q(x, t)φm (x)dx
a00m (t) = −c2 λm am (t) + 0 R L .
0
φ2m (x)dx
To get the initial condition for am , we use the initial condition v(x, 0) = f (x) and
∂v
∂t
(x, 0) = g(x) to get
∞
X
f (x) = an (0)φn (x),
n=1
X∞
g(x) = a0n (0)φn (x).
n=1
So RL
0
f (x)φm (x)dx
am (0) = RL , m = 1, 2, · · · ,
0
φ2m (x)dx
RL
g(x)φm (x)dx
a0m (0) = 0
RL , m = 1, 2, · · · ,
φ 2 (x)dx
0 m

Once solving the initial value problem for am , we obtain the solution v(x, t).

68
3. Example. Solve

∂ 2v ∂ 2v
= 4 + t, (298)
∂t2 ∂x2
v(0, t) = 0, v(π, t) = 0, (299)
∂v
v(x, 0) = sin x, (x, 0) = sin(x). (300)
∂t

69
21 Lecture 17 – Method of Characteristics for the first
order wave equation (Section 12.2 and Chapter 2 of
Dr. Arrigo’s book)
Today:
• Method of Characteristics for the first order wave equation
Next:

• Characteristics for first-order quasi-linear equations (Section 12.2).

Review:

• Normal vector of a surface.

• parametric equations.

• Chain rule for partial derivatives

Teaching procedure:

1. Solution of The first-order wave equation

∂u ∂u
+c = 0, (301)
∂t ∂x
u(x, 0) = f (x). (302)

Please try the method of separation of variable to see why the method does not work
for this so simple equation.
The idea of solving this equation is to reduce the PDE to an ODE. For this we construct
a curve in the x − t plane
x = x(s), t = t(s).

Then consider the PDE along this curve by setting

z(s) = u(x(s), t(s)).

70
 To construct such a curve, we compute

dz ∂u dx ∂u dt
= + .
ds ∂x ds ∂t ds
If we take
dx dt
= c, = 1,
ds ds
we obtain
dz
= 0,
ds
a simple ODE we can solve easily! Before we solve it, let us add initial conditions to
these equations. From the initial condition u(x, 0) = f (x), we can see x(0) can be
at any point on the x-axis and t(0) is always equal to 0. So we have x(0) = r (any
number) and t(0) = 0. Then z(0) = u(x(0), t(0)) = u(r, 0) = f (r). We then have the
following initial value problem
dx dt dz
= c, = 1, = 0,
ds ds ds
x(0) = r, t(0) = 0, z(0) = f (r).
Solving it gives
x = cs + r, t = s, z(s) = f (r),
and then
z(s) = f (x − cs) = f (x − ct).
So we obtain the solution
u(x, t) = z = f (x − ct).
2. Important notions.

• Characteristic curve (line): for any fixed r, the parametric equation

x = cs + r, t = s, z(s) = f (r),
gives a curve (straight line in this case) in space, called Characteristic curve.
• Characteristic equations:
dx dt dz
= c, = 1, = 0.
ds ds ds
• Characteristic method: The method of using Characteristic equations to
solve first PDEs is called Characteristic method.
• The Domain of dependence of u(x, t) on the initial values is the single point
r = x − ct.

71
 • The Range of Influence of the initial values at a particular point r is the
characteristic line x = ct + r.

3. Geometrical interpretation of the first-order wave equation. The surface z = u(x, t)

in xtz-space corresponding to solutions of the PDE is called integral surface. The
equation implies that the vector v = (c, 1, 0), called characteristic vector, is per-
pendicular to the normal vector of the surface n = ( ∂u , ∂u , −1).
∂x ∂t

4. Example. Solve
∂u ∂u
+4 = 0,
∂t ∂x
u(x, 0) = cos x.

72
22 Lecture 18 – Method of Characteristics for the first-
order quasi-linear equations (Section 12.6.1, hand-
outs, and Chapter 2 of Dr. Arrigo’s book)
Today:
• Method of Characteristics for first-order quasi-linear equations
Next:
• Characteristics for the second-order wave equations (Section 12.3).
Review:
• Method of Characteristics for the first-order wave equation
Teaching procedure:
1. First-order quasi-linear equations
∂u ∂u
a(x, y, u) + b(x, y, u) = c(x, y, u), (303)
∂x ∂y
u(x, 0) = f (x). (304)
Consider the curve x = x(s), y = y(s) defined by the equations
dx dy
= a(x, y, u), = b(x, y, u).
ds ds
Set z(s) = u(x(s), y(s)). Then along this curve, we have
dz ∂u dx ∂u dy ∂u ∂u
= + = a(x, y, u) + b(x, y, u) = c(x, y, u) = c(x, y, z).
ds ∂x ds ∂y ds ∂x ∂x
Thus we obtain Characteristic equations:
dx dy dz
= a(x, y, z), = b(x, y, z), = c(x, y, z)
ds ds ds
with the initial conditions
x(0) = r, y(0) = 0, z(0) = f (r).
Solving it gives
x = x(s, r), y = y(s, r), z = z(s, r).
The curve of this parametric equation for any fixed r is called Characteristic curve.
Solving the first two equation
x = x(s, r), y = y(s, r)
for s, r gives
s = s(x, y), r = r(x, y).
We then obtain the solution
u = z = z(s(x, y), r(x, y)).

73
2. Geometrical interpretation. The surface z = u(x, y) in xyz-space corresponding to
solutions of the PDE is called integral surface. The equation implies that the vector
v = (a(x, y, u), b(x, y, u), c(x, y, u)), called characteristic vector, is perpendicular to
the normal vector of the surface n = ( ∂u , ∂u , −1).
∂x ∂y

3. Example 1. Solve
∂u ∂u
y + = 1,
∂x ∂y
u(x, 0) = cos x.

The Characteristic equations are

dx dy dz
= y, = 1, =1
ds ds ds
with the initial conditions

x(0) = r, y(0) = 0, z(0) = cos(r).

The solution is
1
x = s2 + r, y = s, z = s + cos(r).
2
So the solution of the PDE is
µ¶
1 2
u = z = y + cos x − y .
2

4. Example 2. Solve
∂u ∂u
+ = u2 ,
∂x ∂y
u(x, 0) = ex .

The Characteristic equations are

dx dy dz
= 1, = 1, = z2
ds ds ds
with the initial conditions

x(0) = r, y(0) = 0, z(0) = er .

The solution is
1
x = s + r, y = s, z= .
e−r −s
So the solution of the PDE is
1
u=z= .
ey−x − y

74
23 Exercises of Characteristics
Use the method of characteristics to solve
(1)

∂u ∂u
y +x = xu,
∂x ∂y
2
u(x, 0) = e−x .

(2)

∂u ∂u
x + 2u = u,
∂x ∂y
u(x, 0) = x2 .

75
 (1) Solution. The Characteristic equations are

dx dy dz
= y, = x, = xz
ds ds ds
with the initial conditions
2
x(0) = r, y(0) = 0, z(0) = e−r .

We then have
dy x
= .
dx y
ydy = xdx.
y 2 = x2 + c.
The initial condition gives c = −r2 . So

y 2 = x2 − r 2 .

Also
dz xz
= = z.
dy x
2
So z = ey−r . We then have the solution
2 −x2
u = ey+y .

(2) Solution. The Characteristic equations are

dx dy dz
= x, = 2z, =z
ds ds ds
with the initial conditions

x(0) = r, y(0) = 0, z(0) = r2 .

Solving it gives
x = res , z = r2 es , y = 2r2 (es − 1).
We then have
z r 2 es
= s = r.
x re
x x x2
es = = z = .
r x
z
Thus the solution is µ ¶
z2 x2
y=2 2 −1 .
x z

76
24 Lecture 19 – Reduction of Second Order Equations
to Canonical Form (Chapter 3 of Dr. Arrigo’s book)
Today:
• Type of second order equations
• Canonical form
• Reduction of hyperbolic equations
Next:
• Reduction of parabolic equations
• Reduction of elliptic equations
Review:
• chain rule of partial derivative
• Classification of conic section of the form:
Ax2 + 2Bxy + Cy 2 + Dx + Ey + F = 0,
where A, B, C are constant. It is
a hyperbola if B 2 − AC > 0,
a parabola if B 2 − AC = 0,
an ellipse if B 2 − AC < 0.
Teaching procedure:
1. Classification of second order equations:

∂ 2u ∂2u ∂ 2u ∂u ∂u
A 2
+ 2B + C 2
+D +E + F u = G(x, y), (305)
∂x ∂x∂y ∂y ∂x ∂y
where A, B, C are given functions of x, y. It is said to be
hyperbolic if B 2 − AC > 0,
parabolic if B 2 − AC = 0,
elliptic if B 2 − AC < 0.
2. Canonical form of second order equations:

• Hyperbolic Canonical form

∂2u ∂u ∂u
+D +E + F u = G(x, y),
∂x∂y ∂x ∂y
or

∂ 2u ∂2u ∂u ∂u
2
− 2 +D +E + F u = G(x, y),
∂x ∂y ∂x ∂y

77
 • Parabolic Canonical form

∂2u ∂u ∂u
+ D + E + F u = G(x, y),
∂x2 ∂x ∂y
• Elliptic Canonical form

∂ 2u ∂ 2u ∂u ∂u
+ + D + E + F u = G(x, y),
∂x2 ∂y 2 ∂x ∂y

3. Reduction of hyperbolic equations (B 2 − AC > 0). We now want to find a change

of variables
ξ = ξ(x, y), η = η(x, y)
to reduce the hyperbolic equation to its canonical form. We assume that the inverse
transform of the above
x = x(ξ, η), y = y(ξ, η)
exits. Then
u(x, y) = u(x(ξ, η), y(ξ, η)) = u(ξ, η).
To find the above desired change of variables, we now compute derivatives of u and
then substitute them into the equation (305).

∂u ∂u ∂ξ ∂u ∂η
= + ,
∂x ∂ξ ∂x ∂η ∂x
∂u ∂u ∂ξ ∂u ∂η
= + ,
∂y ∂ξ ∂y ∂η ∂y
∂ 2u ∂ 2 u ∂ξ ∂ξ ∂ 2 u ∂η ∂ξ ∂u ∂ 2 ξ ∂ 2 u ∂η ∂η ∂u ∂ 2 η
= + 2 + + + ,
∂x2 ∂ξ 2 ∂x ∂x ∂ξ∂η ∂x ∂x ∂ξ ∂x2 ∂η 2 ∂x ∂x ∂η ∂x2
∂ 2u ∂ 2 u ∂ξ ∂ξ ∂ 2 u ∂η ∂ξ ∂u ∂ 2 ξ ∂ 2 u ∂η ∂η ∂ 2 u ∂η ∂ξ ∂u ∂ 2 η
= + + + + + ,
∂x∂y ∂ξ 2 ∂x ∂y ∂ξ∂η ∂x ∂y ∂ξ ∂x∂y ∂η 2 ∂x ∂y ∂ξ∂η ∂y ∂x ∂η ∂x∂y
∂ 2u ∂ 2 u ∂ξ ∂ξ ∂ 2 u ∂η ∂ξ ∂u ∂ 2 ξ ∂ 2 u ∂η ∂η ∂u ∂ 2 η
= + 2 + + + .
∂y 2 ∂ξ 2 ∂y ∂y ∂ξ∂η ∂y ∂y ∂ξ ∂y 2 ∂η 2 ∂y ∂y ∂η ∂y 2

Substituting them into the equation (305) gives

µ ¶
∂ 2u ∂2u ∂2u ∂u ∂u
A1 2 + 2B1 + C1 2 + F ξ, η, u, , η = 0, (306)
∂ξ ∂ξ∂η ∂η ∂ξ ∂
where
∂ξ ∂ξ ∂ξ ∂ξ ∂ξ ∂ξ
A1 = A + 2B +C ,
∂x ∂x ∂x ∂y ∂y ∂y
µ ¶
∂ξ ∂η ∂ξ ∂η ∂ξ ∂η ∂ξ ∂η
B1 = A +B + +C ,
∂x ∂x ∂x ∂y ∂y ∂x ∂y ∂y
∂η ∂η ∂η ∂η ∂η ∂η
C1 = A + 2B +C .
∂x ∂x ∂x ∂y ∂y ∂y

78
 To reduce the equation (306) to the canonical form, we need to choose ξ and η such
that A1 = C1 = 0. This motivate us to consider the equation
∂φ ∂φ ∂φ ∂φ ∂φ ∂φ
A + 2B +C = 0. (307)
∂x ∂x ∂x ∂y ∂y ∂y
∂φ ∂φ
Dividing the equation by ∂y ∂y
gives
µ ¶2
∂φ ∂φ ∂φ ∂φ
A / + 2B / + C = 0.
∂x ∂y ∂x ∂y
Solving the equation gives
∂φ √
∂x −B ± B 2 − AC
∂φ
= .
∂y
A

We then obtain two equations

∂φ √ ∂φ
A + (B + B 2 − AC) = 0, (308)
∂x ∂y
and
∂φ √ ∂φ
A + (B − B 2 − AC) = 0. (309)
∂x ∂y
Let φ1 (x, y) and φ2 (x, y) are two solutions of these two equation, respectively. If we
choose
ξ = φ1 (x, y), η = φ2 (x, y),
then we have A1 = C1 = 0 because φ1 (x, y) and φ2 (x, y) satisfy the equation (307).
The equation (307) is called the characteristic equation of (305) and the curves
on the xy plane defined by φ1 (x, y) = constant and φ2 (x, y) = constant is called the
characteristic curve of (305).

4. Example. Determine the type of each of the following equation, reduce it to canonical
form, and then find its general solution.

∂2u ∂ 2u ∂ 2u
3 + 10 + 3 = 0.
∂x2 ∂x∂y ∂y 2

(a) A = C = 3 and B = 5. So B 2 − AC = 25 − 9 = 16 > 0. Thus the equation is

hyperbolic.
(b) Solving
∂φ ∂φ
3 +9 = 0,
∂x ∂y
and
∂φ ∂φ
3 + =0
∂x ∂y

79
 we obtain
1
ξ = φ1 = y − 3x, η = φ2 = y − x.
3
Then we have
∂ξ ∂ξ ∂ξ ∂ξ ∂ξ ∂ξ
A1 = A + 2B +C
∂x ∂x ∂x ∂y ∂y ∂y
2
= 3 · (−3) + 10 · (−3) + 3
= 0,
µ ¶
∂ξ ∂η ∂ξ ∂η ∂ξ ∂η ∂ξ ∂η
B1 = A +B + +C
∂x ∂x ∂x ∂y ∂y ∂x ∂y ∂y
= 3 · (−3)/(−3) + 5(−3 − 1/3) + 3
= −32/5,
∂η ∂η ∂η ∂η ∂η ∂η
C1 = A + 2B +C
∂x ∂x ∂x ∂y ∂y ∂y
2
= 3(−1/3) − 10/3 + 3
= 0.

So the canonical form is

∂ 2u
= 0.
∂ξ∂η
(c) Integrating the above canonical equation with respect to η, we obtain

∂u
= f (ξ).
∂ξ
Integrating it with respect to ξ gives the general solution

u = F (ξ) + G(η) = F (y − 3x) + G(y − x/3).

80
5. Reduction of parabolic equations (B 2 = AC). In this case, equations (308) and
(309) become one equation
∂φ ∂φ
A +B = 0.
∂x ∂y
Let φ(x, y) be the solution of the equation and η(x, y) be any other twice differentiable
function. If we choose
ξ = φ(x, y), η = η(x, y),
then we have A1 = 0 because φ(x, y) satisfies the equation (307). Also
µ ¶
∂ξ ∂η ∂ξ ∂η ∂ξ ∂η ∂ξ ∂η
B1 = A +B + +C
∂x ∂x ∂x ∂y ∂y ∂x ∂y ∂y
µ ¶ µ ¶
∂φ ∂φ ∂η ∂φ ∂φ ∂η
= A +B + B +C
∂x ∂y ∂x ∂x ∂y ∂y
µ 2
¶
∂φ B ∂φ ∂η
= B +
∂x A ∂y ∂y
µ ¶
B ∂φ ∂φ ∂η
= A +B
A ∂x ∂y ∂y
= 0.

6. Example. Determine the type of each of the following equation, reduce it to canonical
form, and then find its general solution.

∂ 2u ∂ 2u ∂ 2u
+ 2 + = 0.
∂x2 ∂x∂y ∂y 2

(a) A = C = 1 and B = 1. So B 2 − AC = 1 − 1 = 0. Thus the equation is parabolic.

(b) Solving
∂φ ∂φ
+ = 0,
∂x ∂y
we obtain
ξ =φ=x−y
Let
η = x.

81
 Then we have
∂ξ ∂ξ ∂ξ ∂ξ ∂ξ ∂ξ
A1 = A + 2B +C
∂x ∂x ∂x ∂y ∂y ∂y
2 2
= 1 + 2 · (−1) + (−1)
= 0,
µ ¶
∂ξ ∂η ∂ξ ∂η ∂ξ ∂η ∂ξ ∂η
B1 = A +B + +C
∂x ∂x ∂x ∂y ∂y ∂x ∂y ∂y
= 1 + (−1) + 0
= 0,
∂η ∂η ∂η ∂η ∂η ∂η
C1 = A + 2B +C
∂x ∂x ∂x ∂y ∂y ∂y
= 1.
So the canonical form is
∂ 2u
= 0.
∂η 2
(c) Integrating the above canonical equation with respect to η, we obtain
∂u
= f (ξ).
∂η
Integrating it with respect to η gives the general solution
u = ηf (ξ) + g(ξ) = xf (x − y) + g(x − y).

7. Reduction of elliptic equations (B 2 − AC < 0). In this case, equations (308) and
(309) become complex equation
∂φ √ ∂φ
A + (B + i AC − B 2 ) = 0.
∂x ∂y
Let φ(x, y) = φ1 (x, y)+iφ2 (x, y) be the complex solution of the equation and substitute
it into the equation (307). We then obtain
∂(φ1 + iφ2 ) ∂(φ1 + iφ2 ) ∂(φ1 + iφ2 ) ∂(φ1 + iφ2 ) ∂(φ1 + iφ2 ) ∂(φ1 + iφ2 )
A +2B +C = 0.
∂x ∂x ∂x ∂y ∂y ∂y
Separating the real and imaginary parts gives
∂φ1 ∂φ1 ∂φ1 ∂φ1 ∂φ1 ∂φ1 ∂φ2 ∂φ2 ∂φ2 ∂φ2 ∂φ2 ∂φ2
A + 2B +C =A + 2B +C
∂x ∂x ∂x ∂y ∂y ∂y ∂x ∂x ∂x ∂y ∂y ∂y
and µ ¶
∂φ1 ∂φ2 ∂φ1 ∂φ2 ∂φ1 ∂φ2 ∂φ1 ∂φ2
A +B + +C =0
∂x ∂x ∂x ∂y ∂y ∂x ∂y ∂y
So we can choose
ξ = φ1 (x, y), η = φ2 (x, y),
then we have A1 = C1 and B1 = 0.

82
8. Example. Determine the type of each of the following equation, reduce it to canonical
form, and then find its general solution if possible.

∂ 2u ∂ 2u ∂ 2u
5 + 4 + = 0.
∂x2 ∂x∂y ∂y 2

(a) A = 5, C = 1 and B = 2. So B 2 − AC = 4 − 5 = −1. Thus the equation is

elliptic.
(b) Solving
∂φ ∂φ
5 + (2 + i) = 0,
∂x ∂y
we obtain
φ = x − 2y + iy
Let
ξ = x − 2y, η = y.
Then we have
∂ξ ∂ξ ∂ξ ∂ξ ∂ξ ∂ξ
A1 = A + 2B +C
∂x ∂x ∂x ∂y ∂y ∂y
2
= 5 + 4 · (−2) + (−2)
= 1,
µ ¶
∂ξ ∂η ∂ξ ∂η ∂ξ ∂η ∂ξ ∂η
B1 = A +B + +C
∂x ∂x ∂x ∂y ∂y ∂x ∂y ∂y
= 5·0+2−2
= 0,
∂η ∂η ∂η ∂η ∂η ∂η
C1 = A + 2B +C
∂x ∂x ∂x ∂y ∂y ∂y
= 1.

So the canonical form is

∂ 2u ∂2u
+ 2 = 0.
∂ξ 2 ∂η
(c) Can we find its general solution?

83
25 Lecture 20 – Method of Characteristics for the one-
dimensional wave equation)
Today:

• General solutions

• initial value problems

• Initial boundary value problems

Next:
• Review

Review:

• Characteristic equation
∂φ √ ∂φ
A + (B + B 2 − AC) = 0.
∂x ∂y

∂φ √ ∂φ
A + (B − B 2 − AC) = 0.
∂x ∂y
Teaching procedure:

1. General solutions of the wave equation

∂ 2u 2
2∂ u
= c .
∂t2 ∂x2
Solving
∂φ ∂φ
c − =0
∂x ∂t

∂φ ∂φ
c + =0
∂x ∂y
gives
ξ = x − ct, η = x + ct
. Then the wave equation reduces to

∂ 2u
= 0.
∂ξ∂η
Integrating twice gives the general solution

u = F (x − ct) + G(x + ct).

84
2. Initial value problem

∂ 2u 2
2∂ u
= c ,
∂t2 ∂x2
∂u
u(x, 0) = f (x), (x, 0) = g(x).
∂t
Initial condition gives

f (x) = F (x) + G(x),

g(x) = −cF 0 (x) + cG0 (x).

Integrating the second equation gives

F (x) + G(x) = f (x),

Z
1 x
−F (x) + G(x) = g(z)dz + k.
c 0
Solving it gives
Z
1 1 x k
F (x) = f (x) − g(z)dz − ,
2 2c 0 2
Z x
1 1 k
G(x) = f (x) + g(z)dz + .
2 2c 0 2
and then d’Alembert’s solution
·Z x+ct Z x−ct ¸
f (x + ct) + f (x − ct) 1
u(x, t) = + g(z)dz − g(z)dz
2 2c 0 0
Z
f (x + ct) + f (x − ct) 1 x+ct
= + g(z)dz,
2 2c x−ct

3. Example. Solve the initial value problem

∂ 2u ∂ 2u
= ,
∂t2 ∂x2
∂u
u(x, 0) = x, (x, 0) = x.
∂t
Solution.
Z
x + t + x − t 1 x+t
u(x, t) = + zdz
2 2 x−t
1£ ¤
= x + (x + t)2 − (x + t)2 ,
4

85
4. An important function equation: If

u(x, t) = F (x − ct) + G(x + ct).

then

u(x0 , t0 ) − u(x0 + cξ, t0 + ξ) − u(x0 − cη, t0 + η) + u(x0 + cξ − cη, t0 + ξ + η) = 0

for any x0 and t0 . Geometrically, for any parallelogram A(x0 , t0 ), B(x0 + cξ, t0 +
ξ), C(x0 + cξ − cη, t0 + ξ + η), D(x0 − cη, t0 + η) we have

u(A) + u(C) = u(B) + u(D). (310)

5. Example. Solve the initial boundary value problem using (310)

∂ 2u ∂ 2u
= ,
∂t2 ∂x2
u(0, t) = 0, u(π, t) = 1,
∂u
u(x, 0) = 1, (x, 0) = 0.
∂t

86
26 Review of Nonhomogeneous Problems
1. Procedure for solving Steady nonhomogeneous equation:

∂u ∂2u
= k 2 + Q(x), (311)
∂t ∂x
u(0, t) = A, u(L, t) = B, (312)
u(x, 0) = f (x). (313)

(a) Find Steady-state solution

d2 ue
k = −Q(x), (314)
dx2
ue (0) = A, ue (L) = B. (315)

(b) Convert the nonhomogeneous to the homogeneous by defining v(x, t) = u(x, t) −

ue (x)

∂v ∂ 2v
= k 2, (316)
∂t ∂x
v(0, t) = 0, v(L, t) = 0, (317)
v(x, 0) = f (x) − ue (x). (318)

(c) Solve the homogeneous equation to obtain

∞
X ³ nπx ´ kn2 π 2 t
v(x, t) = an sin e− L2 .
n=1
L
where Z
2 L ³ nπx ´
an = [f (x) − ue (x)] sin dx.
L 0 L
(d) Write down the solution of the original nonhomogeneous equation
∞
X ³ nπx ´ kn2 π 2 t
u(x, t) = ue (x) + an sin e− L2 .
n=1
L

2. Procedure for solving time-dependent nonhomogeneous problems:

∂u ∂ 2u
= k 2 + Q(x, t), (319)
∂t ∂x
u(0, t) = A(t), u(L, t) = B(t), (320)
u(x, 0) = f (x). (321)

87
 (a) Find a reference temperature distribution r(x, t)

B(t) − A(t)
r(x, t) = A(t) + x.
L

(b) Convert the nonhomogeneous to the homogeneous by defining v(x, t) = u(x, t) −

r(x, t)

∂v ∂2v ∂r ∂ 2r ∂2v
= k 2 + Q(x, t) − + k 2 = k 2 + Q1 (x, t), (322)
∂t ∂x ∂t ∂x ∂x
v(0, t) = 0, v(L, t) = 0, (323)
v(x, 0) = f (x) − r(x, 0) = f1 (x). (324)

(c) Use eigenfunction expansion

∞
X
v(x, t) = am (t)φm (x),
m=1

where ³ mπx ´
φm (x) = sin
L
RL
Q1 (x, t)φm (x)dx
a0m (t) = −kλm am (t) + 0
RL .
φ 2 (x)dx
0 m
RL
0
f1 (x)φm (x)dx
am (0) = RL , m = 1, 2, · · · ,
0
φ2m (x)dx

3. All these procedures can be applied to the wave equation.

4. Exercises.

(a) Solve

∂u ∂ 2u
= + sin x, (325)
∂t ∂x2
u(0, t) = 1, u(π, t) = 2, (326)
u(x, 0) = 0. (327)

(b) Solve

∂ 2u ∂ 2u
= + 1, (328)
∂t2 ∂x2
u(0, t) = 1, u(π, t) = 2, (329)
∂u
u(x, 0) = 0, (x, 0) = 0. (330)
∂t

88
(c) Solve

∂u ∂ 2u
= + sin xe−2t , (331)
∂t ∂x2
u(0, t) = 1, u(π, t) = 0, (332)
u(x, 0) = 0. (333)

(d) Solve

∂ 2u ∂ 2u
= + sin xe−2t , (334)
∂t2 ∂x2
u(0, t) = 1, u(π, t) = 0, (335)
∂u
u(x, 0) = 0, (x, 0) = 0. (336)
∂t

89
27 Review of First-order quasi-linear equations
1. Procedure for solving First-order quasi-linear equations
∂u ∂u
a(x, y, u) + b(x, y, u) = c(x, y, u), (337)
∂x ∂y
u(x, 0) = f (x). (338)

(a) Solve the characteristic equations:

dx dy du
= a(x, y, u), = b(x, y, u), = c(x, y, u)
ds ds ds
with the initial conditions

x(0) = r, y(0) = 0, u(0) = f (r)

to obtain
x = x(s, r), y = y(s, r), u = u(s, r).
(b) Solve the first two equations

x = x(s, r), y = y(s, r)

for s, r to obtain
s = s(x, y), r = r(x, y),
and then the solution
u = u(s(x, y), r(x, y)).

2. Working problems. Solve the following equations (from Dr. Arrigo’s tests or sample
tests):

(a) xux − yuy = 2u.

(b) yux − uy = 1.
(c) tut − ux = 2u.
2
(d) yux + xuy = xu, u(x, 0) = e−x .
(e) xux + 2uuy = u, u(x, 0) = x2 .
x2
(f) xux + (x + 2y)uy = 2u, u(x, 0) = x+1
.

90
28 Final Review
28.1 Eigenvalue Problems
1. The eigenvalue problem with Dirichlet boundary conditions
d2 φ
− = λφ,
dx2
φ(0) = 0, φ(L) = 0
has the eigenvalues
n2 π 2
λn = , n = 1, 2, · · · (339)
L2
and the corresponding eigenfunctions
³ nπx ´
φn = sin , n = 1, 2, · · · (340)
L
2. The eigenvalue problem with Neumann boundary conditions
d2 φ
− = λφ,
dx2
dφ dφ
(0) = 0, (L) = 0
dx dx
has the eigenvalues
n2 π 2
λn = , n = 0, 1, 2, · · · (341)
L2
and the corresponding eigenfunctions
³ nπx ´
φn = cos , n = 0, 1, 2, · · · (342)
L
3. Further Study Problems:

(a) Prove Poincaré’s inequality: For any

∞
X ³ nπx ´
u(x) = cn sin ,
i=1
L

we have Z Z
L L
0 2 π2
|u (x)| dx ≥ 2 |u(x)|2 dx.
0 L 0

(b) (Ref. page 64 of the textbook) Show that the eigenvalue problem with periodic
boundary conditions
d2 φ
− = λφ,
dx2
dφ dφ
φ(−L) = φ(L), (−L) = (L)
dx dx
91
 has the eigenvalues
n2 π 2
λn = , n = 0, 1, 2, · · · (343)
L2
and the corresponding eigenfunctions
³ nπx ´ ³ nπx ´
φ0 = 1, φ2n−1 = cos , φ2n = sin , n = 1, 2, · · · (344)
L L

(c) (Ref. page 199 of the textbook) Discuss the eigenvalue problem with boundary
conditions of the third kind
d2 φ
− = λφ,
dx2
dφ
φ(0) = 0, (L) = −hφ(L) (h > 0).
dx

(d) Find the eigenvalues and eigenfunctions of the eigenvalue problem

d2 φ dφ
−k 2
+ v0 = λφ,
dx dx
φ(0) = 0, φ(L) = 0.

(e) Solve the eigenvalue problem 5.3.9. (c) on page 169 of the textbook.
(f) Study the Sturm-Liouville eigenvalue problems, Section 5.3 (optional)
(g) Study the Bessel’s differential equation, section 7.7.4 (optional)

28.2 The Heat Equation

1. Dirichlet boundary value problems:

∂u ∂ 2u
= k 2, (345)
∂t ∂x
u(0, t) = 0, u(L, t) = 0, (346)
u(x, 0) = f (x). (347)

has infinite series solution:

∞
X ³ nπx ´ kn2 π 2 t
u(x, t) = cn sin e− L2 ,
n=1
L

where Z
2 L ³ nπx ´
cn = f (x) sin dx.
L 0 L

92
2. Neumann boundary value problem:

∂u ∂2u
= k 2, (348)
∂t ∂x
∂u ∂u
(0, t) = 0, (L, t) = 0, (349)
dx dx
u(x, 0) = f (x). (350)

has the infinite series solution:

∞
X ³ nπx ´ kn2 π 2 t
u(x, t) = a0 + an cos e− L2 ,
n=1
L

where
Z
1 L
a0 = f (x)dx, (351)
L 0
Z ³ nπx ´
2 L
an = f (x) cos dx, n ≥ 1. (352)
L 0 L

3. Method of eigenfunction expansion for the nonhomogeneous problem

∂v ∂ 2v
= k 2 + Q(x, t), (353)
∂t ∂x
v(0, t) = 0, v(L, t) = 0, (354)
v(x, 0) = f (x). (355)

∞
X
v(x, t) = am (t)φm (x),
m=1

where ³ mπx ´
φm (x) = sin
L
RL
Q(x, t)φm (x)dx
a0m (t) = −kλm am (t) + 0
RL .
φ 2 (x)dx
0 m

RL
0
f (x)φm (x)dx
am (0) = RL , m = 1, 2, · · · ,
0
φ2m (x)dx

4. Further study problems:

(a) Problem 5.3.4. (b) on page 168 of the textbook.

93
 (b) Solve

∂u ∂ 2u
= k 2 + au, (356)
∂t ∂x
u(0, t) = 0, u(L, t) = 0, (357)
u(x, 0) = f (x), (358)

where a is a constant. Determine the range of a for which the solution may grow
exponentially and the range of a for which the solution decay to zero exponentially.
(c) Consider the heat equation

∂u ∂ 2u
= , (359)
∂t ∂x2
∂u 1
u(0, t) = 0, (1, t) = − u(1, t), (360)
∂x 2
u(x, 0) = f (x), (361)

Show that
π2 t
ku(t)k ≤ kf ke− 8 ,
where k · k denotes the L2 norm defined by
s
Z 1
ku(t)k = |u(x, t)|2 dx.
0

(d) Problem 1.5.3. (b) and (e) on page 29 of the textbook.

(e) Problem 1.5.4. (b) and (e) on page 30 of the textbook.
(f) Problem 5.4.3. on page 173 of the textbook (optional).
(g) Problem 8.3.3. on page 359 of the textbook (optional).

28.3 The Wave Equation

1. The Dirichlet initial boundary value problem:

∂ 2u 2
2∂ u
= c , (362)
∂t2 ∂x2
u(0, t) = 0, u(L, t) = 0, (363)
∂u
u(x, 0) = f (x), (x, 0) = g(x) (364)
∂t
has infinite series solution:
X∞ µ µ ¶ µ ¶¶ ³ nπx ´
nπct nπct
u(x, t) = an cos + bn sin sin ,
n=1
L L L

94
 where
Z ³ nπx ´
2 L
an = f (x) sin dx,
L 0 L
Z ³ nπx ´
nπc 2 L
bn = g(x) sin dx
L L 0 L

2. The Neumann initial boundary value problem:

∂2u 2
2∂ u
= c , (365)
∂t2 ∂x2
∂u ∂u
(0, t) = 0, (L, t) = 0, (366)
∂x ∂x
∂u
u(x, 0) = f (x), (x, 0) = g(x). (367)
∂t
has infinite series solution:
X∞ µ µ ¶ µ ¶¶ ³ nπx ´
nπct nπct
u(x, t) = a0 + b0 t + an cos + bn sin cos ,
n=1
L L L

where
Z
1 L
a0 = f (x)dx,
L 0
Z ³ nπx ´
2 L
an = f (x) cos dx, n≥1
L 0 L
Z
1 L
b0 = g(x)dx.
L 0
Z ³ nπx ´
nπc 2 L
bn = g(x) cos dx, n ≥ 1.
L L 0 L

3. The Method of Eigenfunction expansion for

∂ 2v 2
2∂ v
= c + Q(x, t), (368)
∂t2 ∂x2
v(0, t) = 0, v(L, t) = 0, (369)
∂v
v(x, 0) = f (x), (x, 0) = g(x). (370)
∂t

The corresponding eigenvalues and eigenfunctions:

n2 π 2 ³ nπx ´
λn = 2 , φn (x) = sin .
L L

95
 Solution: ∞
X
v(x, t) = an (t)φn (x).
n=1
where RL
Q(x, t)φm (x)dx
a00m (t) 2
= −c λm am (t) + 0
RL .
φ 2 (x)dx
0 m
RL
0
f (x)φm (x)dx
am (0) = RL , m = 1, 2, · · · ,
0
φ2m (x)dx
RL
g(x)φm (x)dx
a0m (0) = 0
RL , m = 1, 2, · · · .
φ 2 (x)dx
0 m

4. Further study problems:

(a) Problem 5.8.6. (c) on page 210 of the textbook.
(b) Solve
∂ 2v ∂ 2v
= , (371)
∂t2 ∂x2
v(0, t) = 0, v(1, t) = sin t, (372)
∂v
v(x, 0) = 0, (x, 0) = 0. (373)
∂t
(c) Problem 8.5.2. on page 371 of the textbook.
(d) Problem 7.7.1. on page 315 of the textbook (optional).

28.4 Laplace’s equation and Poisson’s equation

Further study problems:
1. Problem 2.5.1. (e) on page 85 of the textbook.
2. Problem 7.3.7. (c) on page 288 of the textbook.
3. Problem 8.6.6. on page 379 of the textbook (use eigenfunction expansion).
4. Problem 2.5.8. (a) on page 87 of the textbook (optional).
5. Problem 8.6.3. (a) on page 378 of the textbook (optional).

28.5 Method of Characteristics

Further study problems:
1. Problem 12.6.8. (a) and (e) on page 582 of the textbook.
2. Problem 12.6.9. (a) and (c) on page 582 of the textbook.
3. Problem 12.4.2. on page 556 of the textbook (optional).

96