AM364

Week 3
Chapter 11 cont.
Chapter 12
Simple Harmonic Motion (SHM)
𝑥ሷ + 𝛼 2 𝑥 = 0
Anti Simple Harmonic Motion (anti-SHM)
𝑥ሷ − 𝛼 2 𝑥 = 0

2
SHM anti-SHM
The ODE: The ODE:
𝑥ሷ + 𝛼 2 𝑥 = 0 𝑥ሷ − 𝛼 2 𝑥 = 0

Solving: Trial solution Solving: Trial solution

𝑥 = 𝑒 𝑚𝑡 𝑥 = 𝑒 𝑚𝑡
Substitute into ODE Substitute into ODE
𝑚2 𝑒 𝑚𝑡 + 𝛼 2 𝑒 𝑚𝑡 = 0 𝑚2 𝑒 𝑚𝑡 − 𝛼 2 𝑒 𝑚𝑡 = 0
֜ 𝑚2 = −𝛼 2 ֜ 𝑚2 = 𝛼 2
𝑚 = ±𝑖𝛼 𝑚 = ±𝛼
𝑥1 = 𝑒 𝑖𝛼𝑡 𝑥1 = 𝑒 𝛼𝑡
𝑥2 = 𝑒 −𝑖𝛼𝑡 𝑥2 = 𝑒 −𝛼𝑡
SHM anti-SHM
The ODE: The ODE:
𝑥ሷ + 𝛼 2 𝑥 = 0 𝑥ሷ − 𝛼 2 𝑥 = 0

Solving: Trial solution Solving: Trial solution

𝑥 = 𝑒 𝑚𝑡 𝑥 = 𝑒 𝑚𝑡
Substitute into ODE Substitute into ODE
𝑚2 𝑒 𝑚𝑡 + 𝛼 2 𝑒 𝑚𝑡 = 0 𝑚2 𝑒 𝑚𝑡 − 𝛼 2 𝑒 𝑚𝑡 = 0
֜ 𝑚2 = −𝛼 2 ֜ 𝑚2 = 𝛼 2
𝑚 = ±𝑖𝛼 𝑚 = ±𝛼

𝑥 = 𝐷𝑒 𝑖𝛼𝑡 + 𝐸𝑒 −𝑖𝛼𝑡 𝑥 = 𝐷𝑒 𝛼𝑡x+ 𝐸𝑒 −𝛼𝑡

SHM anti-SHM
Solution: Solution:
𝑥 = 𝐷𝑒 𝑖𝛼𝑡 + 𝐸𝑒 −𝑖𝛼𝑡 𝑥 = 𝐷𝑒 𝛼𝑡 x+ 𝐸𝑒 −𝛼𝑡
Alternative form: Alternative form:
Substituting the identity Substituting the identity
𝑒 ±𝑖𝛼𝑡 = cos(𝛼𝑡) ± 𝑖 sin(𝛼𝑡) 𝑒 ±𝛼𝑡 = cosh (𝛼𝑡) ± sinh(𝛼𝑡)
into the solution above yields: into the solution above yields:

𝑥 = 𝐵 cos(𝛼𝑡) + 𝐶 sin(𝛼𝑡) 𝑥 = 𝐵 cosh(𝛼𝑡) + 𝐶 sinh(𝛼𝑡)

where where
𝐵 =𝐷+𝐸 𝐵 =𝐷+𝐸
𝐶 =𝑖 𝐷−𝐸 𝐶 =𝐷−𝐸
 SHM anti-SHM
Solutions:
𝑥 = 𝐷𝑒 𝑖𝛼𝑡 + 𝐸𝑒 −𝑖𝛼𝑡 𝑥 = 𝐷𝑒 𝛼𝑡 x+ 𝐸𝑒 −𝛼𝑡

𝑥 = 𝐵 cos(𝛼𝑡) + 𝐶 sin(𝛼𝑡) 𝑥 = 𝐵 cosh(𝛼𝑡) + 𝐶 sinh(𝛼𝑡)

𝛼: angular frequency, s −1
- number of oscillations in 2𝜋 seconds The ODE:
2𝜋
𝛼
: period, s 𝑥ሷ = 0
- time to complete one oscillation
The Solution:
𝑥 = 𝐵𝑡 + 𝐶
Sturm-Liouville Problem
You may omit:
• Singular Sturm-Liouville Problem
• Self-Adjoint Form
• Example 3: Parametric Bessel Equation
• Example 4: Legendre’s Equation

7
 Sturm-Liouville problem: An Example
Solve
𝑦 ′′ + 𝜆𝑦 = 0
on
𝑥 ∈ [0, 𝐿]
subject to
𝑦(0) = 0
and
𝑦 𝐿 =0

Consider
• 𝜆=0
• 𝜆 < 0 = −𝛼 2
• 𝜆 > 0 = 𝛼2
 Sturm-Liouville problem: An Example
Solve
𝑦 ′′ + 𝜆𝑦 = 0 • 𝜆=0
on
𝑥 ∈ [0, 𝐿] 𝑦′′ = 0
subject to 𝑦′ = 𝐴
𝑦(0) = 0 and 𝑦 𝐿 = 0 𝑦 = 𝐴𝑥 + 𝐵

Substitute boundary conditions:

𝑦 0 = 0: 𝑦 𝐿 = 0:
0 = 𝐴(0) + 𝐵 0 = 𝐴(𝐿)
֜ 𝐵=0 ֜ 𝐴=0
֜ 𝑦 = 𝐴𝑥 ֜ 𝑦=0

The trivial solution!

 Sturm-Liouville problem: An Example
Solve
𝑦 ′′ + 𝜆𝑦 = 0 • 0 > 𝜆 = −𝛼 2
on
𝑥 ∈ [0, 𝐿] 𝑦 ′′ − 𝛼 2 𝑦 = 0
subject to 𝑦 = 𝐴 cosh 𝛼𝑥 + 𝐵 sinh 𝛼𝑥
𝑦(0) = 0 and 𝑦 𝐿 = 0

Substitute boundary conditions:

𝑦 0 = 0: 𝑦 𝐿 = 0:
0 = 𝐴(1) + 𝐵 0 0 = 𝐵 sinh 𝛼𝐿
֜ 𝐴=0 ֜ 𝐵=0
֜ 𝑦 = 𝐵 sinh 𝛼𝑥 ֜ 𝑦=0

The trivial solution!

 Sturm-Liouville problem: An Example
Solve
𝑦 ′′ + 𝜆𝑦 = 0 • 0 < 𝜆 = 𝛼2
on
𝑥 ∈ [0, 𝐿] 𝑦 ′′ + 𝛼 2 𝑦 = 0
subject to 𝑦 = 𝐴 cos 𝛼𝑥 + 𝐵 sin 𝛼𝑥
𝑦(0) = 0 and 𝑦 𝐿 = 0

Substitute boundary conditions:

𝑦 0 = 0: 𝑦 𝐿 = 0:
0 = 𝐴(1) + 𝐵 0 0 = 𝐵 sin 𝛼𝐿
֜ 𝐴=0 ֜ 𝐵=0 ?
֜ 𝑦 = 𝐵 sin 𝛼𝑥 ֜ 𝑦=0

The trivial solution!

 Sturm-Liouville problem: An Example
Solve
𝑦 ′′ + 𝜆𝑦 = 0 • 0 < 𝜆 = 𝛼2
on
𝑥 ∈ [0, 𝐿] 𝑦 ′′ + 𝛼 2 𝑦 = 0
subject to 𝑦 = 𝐴 cos 𝛼𝑥 + 𝐵 sin 𝛼𝑥
𝑦(0) = 0 and 𝑦 𝐿 = 0

Substitute boundary conditions:

𝑦 0 = 0: 𝑦 𝐿 = 0:
0 = 𝐴(1) + 𝐵 0 0 = 𝐵 sin 𝛼𝐿
֜ 𝐴=0 ֜ 𝛼𝐿 = 𝑘𝜋 𝑘 ∈ ℤ
֜ 𝑦 = 𝐵 sin 𝛼𝑥 𝑘𝜋
֜ 𝛼=
𝐿
𝑘𝜋 2
Eigenvalues: 𝜆𝑘 = , 𝑘 = 1,2,3, … 𝑘𝜋
𝐿 ֜ 𝑦 = 𝐵 sin 𝑥
Eigenfunctions: 𝑦𝑘 = 𝐵 sin
𝑘𝜋
𝑥 , 𝑘 = 1,2,3, … 𝐿
𝐿
 Sturm-Liouville problem: An Example
Solve
𝑦 ′′ + 𝜆𝑦 = 0 • 0 < 𝜆 = 𝛼2
on
𝑥 ∈ [0, 𝐿] 𝑦 ′′ + 𝛼 2 𝑦 = 0
subject to 𝑦 = 𝐴 cos 𝛼𝑥 + 𝐵 sin 𝛼𝑥
𝑦(0) = 0 and 𝑦 𝐿 = 0

Substitute boundary conditions:

𝑦 0 = 0: 𝑦 𝐿 = 0:
0 = 𝐴(1) + 𝐵 0 0 = 𝐵 sin 𝛼𝐿
֜𝐴 = 0 ֜ 𝛼𝐿 =are 𝑘𝜋defined
𝑘∈ℤ
Similar to eigenvectors in Linear Algebra, eigenfunctions up
֜ 𝑦 = 𝐵 the
to scale (therefore, sin constant
𝛼𝑥 𝐵 is arbitrary, as long 𝑘𝜋
֜ 𝛼 = is not zero).
as it
𝐿
𝑘𝜋 2
Eigenvalues: 𝜆𝑘 = , 𝑘 = 1,2,3, … 𝑘𝜋
𝐿 ֜ 𝑦 = 𝐵 sin 𝑥
Eigenfunctions: 𝑦 = sin
𝑘𝜋
𝑥 , 𝑘 = 1,2,3, … 𝐿
𝑘 𝐿
 Sturm-Liouville problem
Reference: Britannica

In the mid-1830s the French mathematicians Charles-François

Sturm and Joseph Liouville independently worked on the problem
of heat conduction through a metal bar, in the process developing
techniques for solving a large class of PDEs, the simplest of which
take the form
𝑟 𝑥 𝑦 ′ ′ + 𝑞 𝑥 − 𝜆𝑝 𝑥 𝑦 = 0
where 𝑦 is some physical quantity and 𝜆 is a parameter, or
eigenvalue

eigenvalue, that constrains the equation so that 𝑦 satisfies the

boundary values at the endpoints of the interval over which the
variable 𝑥 ranges. If the functions 𝑟(𝑥), 𝑞(𝑥), and 𝑝(𝑥) satisfy
suitable conditions, the equation will have a family of solutions,
called eigenfunctions, corresponding to each of the eigenvalues.
Regular Sturm-Liouville problem
Consider
𝑟 𝑥 𝑦 ′ ′ + 𝑞 𝑥 + 𝜆𝑝 𝑥 𝑦 = 0
on
𝑥 ∈ [𝑎, 𝑏]
subject to
𝐴1 𝑦 𝑎 + 𝐵1 𝑦′(𝑎) = 0
and
𝐴2 𝑦 𝑏 + 𝐵2 𝑦 ′ 𝑏 = 0
Sturm-Liouville problem: An Example
Solve
(1)𝑦 ′ ′ + 0 + 𝜆(1) 𝑦 = 0
on
𝑥 ∈ [𝑎, 𝑏]
subject to
(1)𝑦 𝑎 + 0 𝑦′(𝑎) = 0
and
(1)𝑦 𝑏 + 0 𝑦 ′ 𝑏 = 0
Sturm-Liouville problem: An Example
Solve
(1)𝑦 ′ ′ + 0 + 𝜆(1) 𝑦 = 0
on
𝑥 ∈ [0, 𝐿]
subject to
(1)𝑦 0 + 0 𝑦′(0) = 0
and
(1)𝑦 𝐿 + 0 𝑦 ′ 𝐿 = 0

(Previous Example)
 Regular Sturm-Liouville problem
Consider
𝑟 𝑥 𝑦 ′ ′ + 𝑞 𝑥 + 𝜆𝑝 𝑥 𝑦 = 0
on
𝑥 ∈ [𝑎, 𝑏]
subject to
𝐴1 𝑦 𝑎 + 𝐵1 𝑦′(𝑎) = 0
and
𝐴2 𝑦 𝑏 + 𝐵2 𝑦 ′ 𝑏 = 0

Each 𝜆 for which there exists a non-trivial solution is called an

eigenvalue. The solution 𝑦(𝑥) is called the eigenfunction.
Infinite number of eigenvalues, each with its own unique non-trivial
eigenfunction.
Eigenfunctions form an orthogonal set with respect to the weight
function, 𝑝(𝑥).
 Sturm-Liouville problem
Consider
Eigenfunctions form an orthogonal set 𝑟 𝑥 𝑦 ′ ′ + 𝑞 𝑥 + 𝜆𝑝 𝑥 𝑦 = 0
on
with respect to the weight function, 𝑝 𝑥 . 𝑥 ∈ [𝑎, 𝑏]
subject to
𝐴1 𝑦 𝑎 + 𝐵1 𝑦′(𝑎) = 0
DEFINITION 11.1.4 and
𝐴2 𝑦 𝑏 + 𝐵2 𝑦 ′ 𝑏 = 0

A set of real valued functions {𝑦1 𝑥 , 𝑦2 𝑥 , 𝑦3 𝑥 , … } is said

to be orthogonal with respect to a weight function 𝑝(𝑥) on an
interval [𝑎, 𝑏] if
𝑏
න 𝑝 𝑥 𝑦𝑚 𝑥 𝑦𝑛 𝑥 𝑑𝑥 = 0, 𝑚≠𝑛
𝑎
 Sturm-Liouville problem
Consider
Eigenfunctions form an orthogonal set 𝑟 𝑥 𝑦 ′ ′ + 𝑞 𝑥 + 𝜆𝑝 𝑥 𝑦 = 0
on
with respect to the weight function, 𝑝 𝑥 . 𝑥 ∈ [𝑎, 𝑏]
subject to
𝐴1 𝑦 𝑎 + 𝐵1 𝑦′(𝑎) = 0
PROOF and
𝐴2 𝑦 𝑏 + 𝐵2 𝑦 ′ 𝑏 = 0

Let 𝑦𝑚 (𝑥) and 𝑦𝑛 (𝑥) be the non-trivial eigenfunctions of

𝜆𝑚 and 𝜆𝑛 respectively, where 𝑚 ≠ 𝑛.
Since both 𝑦𝑚 and 𝑦𝑛 solve the boundary value problem, it
follows that

′ ′
𝑟 𝑥 𝑦𝑚 + 𝑞 𝑥 + 𝜆𝑚 𝑝 𝑥 𝑦𝑚 = 0 𝑟 𝑥 𝑦𝑛′ ′ + 𝑞 𝑥 + 𝜆𝑛 𝑝 𝑥 𝑦𝑛 = 0
′ (𝑎) = 0
𝐴1 𝑦𝑚 𝑎 + 𝐵1 𝑦𝑚 𝐴1 𝑦𝑛 𝑎 + 𝐵1 𝑦𝑛′ (𝑎) = 0
′ 𝑏 =0
𝐴2 𝑦𝑚 𝑏 + 𝐵2 𝑦𝑚 𝐴2 𝑦𝑛 𝑏 + 𝐵2 𝑦𝑛′ 𝑏 = 0
 Sturm-Liouville problem
′ ′+ 𝑞 𝑥 +𝜆 𝑝 𝑥 𝑦
𝑟 𝑥 𝑦𝑚 𝑚 𝑚 = 0 − 𝑟 𝑥 𝑦𝑛′ ′ + 𝑞 𝑥 + 𝜆𝑛 𝑝 𝑥 𝑦𝑛 = 0
× 𝑦𝑛 × 𝑦𝑚
′ ′
𝐴1 𝑦𝑛 𝑎 + 𝐵1 𝑦𝑛 (𝑎) = 0
𝐴1 𝑦𝑚 𝑎 + 𝐵1 𝑦𝑚 (𝑎) = 0
′
𝐴2 𝑦𝑚 𝑏 + 𝐵2 𝑦𝑚 𝑏 =0 𝐴2 𝑦𝑛 𝑏 + 𝐵2 𝑦𝑛′ 𝑏 = 0
 Sturm-Liouville problem
′ ′+ 𝑞 𝑥 +𝜆 𝑝 𝑥 𝑦
𝑟 𝑥 𝑦𝑚 𝑚 𝑚 = 0 − 𝑟 𝑥 𝑦𝑛′ ′ + 𝑞 𝑥 + 𝜆𝑛 𝑝 𝑥 𝑦𝑛 = 0
× 𝑦𝑛 × 𝑦𝑚
′ ′
𝐴1 𝑦𝑛 𝑎 + 𝐵1 𝑦𝑛 (𝑎) = 0
𝐴1 𝑦𝑚 𝑎 + 𝐵1 𝑦𝑚 (𝑎) = 0
′
𝐴2 𝑦𝑚 𝑏 + 𝐵2 𝑦𝑚 𝑏 =0 𝐴2 𝑦𝑛 𝑏 + 𝐵2 𝑦𝑛′ 𝑏 = 0

′ ′𝑦 + 𝑞 𝑥 + 𝜆 𝑝 𝑥 𝑦 𝑦 − 𝑟 𝑥 𝑦′ ′𝑦 − 𝑞 𝑥 + 𝜆 𝑝 𝑥 𝑦 𝑦 = 0
𝑟 𝑥 𝑦𝑚 𝑛 𝑚 𝑚 𝑛 𝑛 𝑚 𝑛 𝑛 𝑚
 Sturm-Liouville problem
′ ′+ 𝑞 𝑥 +𝜆 𝑝 𝑥 𝑦
𝑟 𝑥 𝑦𝑚 𝑚 𝑚 = 0 − 𝑟 𝑥 𝑦𝑛′ ′ + 𝑞 𝑥 + 𝜆𝑛 𝑝 𝑥 𝑦𝑛 = 0
× 𝑦𝑛 × 𝑦𝑚
′ ′
𝐴1 𝑦𝑛 𝑎 + 𝐵1 𝑦𝑛 (𝑎) = 0
𝐴1 𝑦𝑚 𝑎 + 𝐵1 𝑦𝑚 (𝑎) = 0
′
𝐴2 𝑦𝑚 𝑏 + 𝐵2 𝑦𝑚 𝑏 =0 𝐴2 𝑦𝑛 𝑏 + 𝐵2 𝑦𝑛′ 𝑏 = 0

′ ′𝑦 + 𝑞 𝑥 + 𝜆 𝑝 𝑥 𝑦 𝑦 − 𝑟 𝑥 𝑦′ ′𝑦 − 𝑞 𝑥 + 𝜆 𝑝 𝑥 𝑦 𝑦 = 0
𝑟 𝑥 𝑦𝑚 𝑛 𝑚 𝑚 𝑛 𝑛 𝑚 𝑛 𝑛 𝑚

′ ′𝑦 + 𝜆 𝑝 𝑥 𝑦 𝑦 − 𝑟 𝑥 𝑦′ ′𝑦 − 𝜆 𝑝 𝑥 𝑦 𝑦 = 0
𝑟 𝑥 𝑦𝑚 𝑛 𝑚 𝑚 𝑛 𝑛 𝑚 𝑛 𝑛 𝑚

′ ′𝑦 − 𝑟 𝑥 𝑦′ ′𝑦 = 𝜆 − 𝜆
𝑟 𝑥 𝑦𝑚 𝑛 𝑛 𝑚 𝑛 𝑚 𝑝 𝑥 𝑦𝑚 𝑦𝑛
𝑏 𝑏 𝑏
න ′ ′ 𝑦 𝑑𝑥 − න
𝑟 𝑥 𝑦𝑚 𝑟 𝑥 𝑦𝑛′ ′ 𝑦𝑚 𝑑𝑥 = 𝜆𝑛 − 𝜆𝑚 න 𝑝 𝑥 𝑦𝑚 𝑦𝑛 𝑑𝑥
𝑛
𝑎 𝑎 𝑎
 Sturm-Liouville problem
′ ′+ 𝑞 𝑥 +𝜆 𝑝 𝑥 𝑦
𝑟 𝑥 𝑦𝑚 𝑚 𝑚 = 0 − 𝑟 𝑥 𝑦𝑛′ ′ + 𝑞 𝑥 + 𝜆𝑛 𝑝 𝑥 𝑦𝑛 = 0
× 𝑦𝑛 × 𝑦𝑚
′ ′
𝐴1 𝑦𝑛 𝑎 + 𝐵1 𝑦𝑛 (𝑎) = 0
𝐴1 𝑦𝑚 𝑎 + 𝐵1 𝑦𝑚 (𝑎) = 0
′
𝐴2 𝑦𝑚 𝑏 + 𝐵2 𝑦𝑚 𝑏 =0 𝐴2 𝑦𝑛 𝑏 + 𝐵2 𝑦𝑛′ 𝑏 = 0

′ ′𝑦 + 𝑞 𝑥 + 𝜆 𝑝 𝑥 𝑦 𝑦 − 𝑟 𝑥 𝑦′ ′𝑦 − 𝑞 𝑥 + 𝜆 𝑝 𝑥 𝑦 𝑦 = 0
𝑟 𝑥 𝑦𝑚 𝑛 𝑚 𝑚 𝑛 𝑛 𝑚 𝑛 𝑛 𝑚

′ ′𝑦 + 𝜆 𝑝 𝑥 𝑦 𝑦 − 𝑟 𝑥 𝑦′ ′𝑦 − 𝜆 𝑝 𝑥 𝑦 𝑦 = 0
𝑟 𝑥 𝑦𝑚 𝑛 𝑚 𝑚 𝑛 𝑛 𝑚 𝑛 𝑛 𝑚

′ ′𝑦 − 𝑟 𝑥 𝑦′ ′𝑦 = 𝜆 − 𝜆
𝑟 𝑥 𝑦𝑚 𝑛 𝑛 𝑚 𝑛 𝑚 𝑝 𝑥 𝑦𝑚 𝑦𝑛
𝑏 𝑏 𝑏
න ′ ′ 𝑦 𝑑𝑥 − න
𝑟 𝑥 𝑦𝑚 𝑟 𝑥 𝑦𝑛′ ′ 𝑦𝑚 𝑑𝑥 = 𝜆𝑛 − 𝜆𝑚 න 𝑝 𝑥 𝑦𝑚 𝑦𝑛 𝑑𝑥
𝑛
𝑎 𝑎 𝑎

Integration by parts
𝑏 𝑏 𝑏 𝑏
′ 𝑦 ′ 𝑦′
𝐿𝐻𝑆 = 𝑟 𝑥 𝑦𝑚 𝑛 ቚ − න 𝑟 𝑥 𝑦𝑚 𝑛 𝑑𝑥 − 𝑟 𝑥 𝑦𝑛′ 𝑦𝑚 ቚ − න 𝑟 𝑥 𝑦𝑛′ 𝑦𝑚
′ 𝑑𝑥
𝑎 𝑎 𝑎 𝑎
 Sturm-Liouville problem
′ ′+ 𝑞 𝑥 +𝜆 𝑝 𝑥 𝑦
𝑟 𝑥 𝑦𝑚 𝑚 𝑚 = 0 𝑟 𝑥 𝑦𝑛′ ′ + 𝑞 𝑥 + 𝜆𝑛 𝑝 𝑥 𝑦𝑛 = 0
′ (𝑎) = 0
𝐴1 𝑦𝑚 𝑎 + 𝐵1 𝑦𝑚 𝐴1 𝑦𝑛 𝑎 + 𝐵1 𝑦𝑛′ (𝑎) = 0
′
𝐴2 𝑦𝑚 𝑏 + 𝐵2 𝑦𝑚 𝑏 =0 𝐴2 𝑦𝑛 𝑏 + 𝐵2 𝑦𝑛′ 𝑏 = 0
𝑏
𝑅𝐻𝑆 = 𝜆𝑛 − 𝜆𝑚 න 𝑝 𝑥 𝑦𝑚 𝑦𝑛 𝑑𝑥
𝑎
𝑏 𝑏 𝑏 𝑏
′ ′ ′
𝐿𝐻𝑆 = 𝑟 𝑥 𝑦𝑚 𝑦𝑛 ቚ − න 𝑟 𝑥 𝑦𝑚 𝑦𝑛 𝑑𝑥 − 𝑟 𝑥 𝑦𝑛′ 𝑦𝑚 ቚ − න 𝑟 𝑥 𝑦𝑛′ 𝑦𝑚
′
𝑑𝑥
𝑎 𝑎 𝑎 𝑎
′ 𝑏 𝑦 𝑏 − 𝑟 𝑎 𝑦′ 𝑎 𝑦 𝑎
= 𝑟 𝑏 𝑦𝑚 − 𝑟 𝑏 𝑦𝑛′ 𝑏 𝑦𝑚 𝑏 − 𝑟 𝑎 𝑦𝑛′ 𝑎 𝑦𝑚 𝑎
𝑛 𝑚 𝑛

= 𝑟 𝑎 𝑦𝑛′ 𝑎 𝑦𝑚 𝑎 − 𝑦𝑚
′ 𝑎 𝑦 𝑎
𝑛
′ 𝑏 𝑦 𝑏 − 𝑦′ 𝑏 𝑦 𝑏
+ 𝑟 𝑏 𝑦𝑚 𝑛 𝑛 𝑚
 Sturm-Liouville problem
′ ′+ 𝑞 𝑥 +𝜆 𝑝 𝑥 𝑦
𝑟 𝑥 𝑦𝑚 𝑚 𝑚 = 0 𝑟 𝑥 𝑦𝑛′ ′ + 𝑞 𝑥 + 𝜆𝑛 𝑝 𝑥 𝑦𝑛 = 0
′ (𝑎) = 0
𝐴1 𝑦𝑚 𝑎 + 𝐵1 𝑦𝑚 𝐴1 𝑦𝑛 𝑎 + 𝐵1 𝑦𝑛′ (𝑎) = 0
′
𝐴2 𝑦𝑚 𝑏 + 𝐵2 𝑦𝑚 𝑏 =0 𝐴2 𝑦𝑛 𝑏 + 𝐵2 𝑦𝑛′ 𝑏 = 0
𝑏
𝑅𝐻𝑆 = 𝜆𝑛 − 𝜆𝑚 න 𝑝 𝑥 𝑦𝑚 𝑦𝑛 𝑑𝑥
𝑎
𝑏 𝑏 𝑏 𝑏
′ ′ ′
𝐿𝐻𝑆 = 𝑟 𝑥 𝑦𝑚 𝑦𝑛 ቚ − න 𝑟 𝑥 𝑦𝑚 𝑦𝑛 𝑑𝑥 − 𝑟 𝑥 𝑦𝑛′ 𝑦𝑚 ቚ − න 𝑟 𝑥 𝑦𝑛′ 𝑦𝑚
′
𝑑𝑥
𝑎 𝑎 𝑎 𝑎
′ 𝑏 𝑦 𝑏 − 𝑟 𝑎 𝑦′ 𝑎 𝑦 𝑎
= 𝑟 𝑏 𝑦𝑚 − 𝑟 𝑏 𝑦𝑛′ 𝑏 𝑦𝑚 𝑏 − 𝑟 𝑎 𝑦𝑛′ 𝑎 𝑦𝑚 𝑎
𝑛 𝑚 𝑛

= 𝑟 𝑎 𝑦𝑛′ 𝑎 𝑦𝑚 𝑎 − 𝑦𝑚
′ 𝑎 𝑦 𝑎
𝑛
′ 𝑏 𝑦 𝑏 − 𝑦′ 𝑏 𝑦 𝑏
+ 𝑟 𝑏 𝑦𝑚 𝑛 𝑛 𝑚

Case 1:
𝑟 𝑎 =𝑟 𝑏 =0
𝑏
֜ න 𝑝 𝑥 𝑦𝑚 𝑦𝑛 𝑑𝑥 = 0 (𝜆𝑚 ≠ 𝜆𝑛 )
𝑎
 Sturm-Liouville problem
′ ′+ 𝑞 𝑥 +𝜆 𝑝 𝑥 𝑦
𝑟 𝑥 𝑦𝑚 𝑚 𝑚 = 0 𝑟 𝑥 𝑦𝑛′ ′ + 𝑞 𝑥 + 𝜆𝑛 𝑝 𝑥 𝑦𝑛 = 0
′ (𝑎) = 0
𝐴1 𝑦𝑚 𝑎 + 𝐵1 𝑦𝑚 𝐴1 𝑦𝑛 𝑎 + 𝐵1 𝑦𝑛′ (𝑎) = 0
′
𝐴2 𝑦𝑚 𝑏 + 𝐵2 𝑦𝑚 𝑏 =0 𝐴2 𝑦𝑛 𝑏 + 𝐵2 𝑦𝑛′ 𝑏 = 0
𝑏
𝑅𝐻𝑆 = 𝜆𝑛 − 𝜆𝑚 න 𝑝 𝑥 𝑦𝑚 𝑦𝑛 𝑑𝑥
𝑎
𝑏 𝑏 𝑏 𝑏
′ ′ ′
𝐿𝐻𝑆 = 𝑟 𝑥 𝑦𝑚 𝑦𝑛 ቚ − න 𝑟 𝑥 𝑦𝑚 𝑦𝑛 𝑑𝑥 − 𝑟 𝑥 𝑦𝑛′ 𝑦𝑚 ቚ − න 𝑟 𝑥 𝑦𝑛′ 𝑦𝑚
′
𝑑𝑥
𝑎 𝑎 𝑎 𝑎
′ 𝑏 𝑦 𝑏 − 𝑟 𝑎 𝑦′ 𝑎 𝑦 𝑎
= 𝑟 𝑏 𝑦𝑚 − 𝑟 𝑏 𝑦𝑛′ 𝑏 𝑦𝑚 𝑏 − 𝑟 𝑎 𝑦𝑛′ 𝑎 𝑦𝑚 𝑎
𝑛 𝑚 𝑛

= 𝑟 𝑎 𝑦𝑛′ 𝑎 𝑦𝑚 𝑎 − 𝑦𝑚
′ 𝑎 𝑦 𝑎
𝑛
′ 𝑏 𝑦 𝑏 − 𝑦′ 𝑏 𝑦 𝑏
+ 𝑟 𝑏 𝑦𝑚 𝑛 𝑛 𝑚

Case 2: 𝑟 𝑥 ≠ 0 at 𝑎 and 𝑏

𝐴1 𝑦𝑛 𝑎 𝐴1 𝑦𝑚 𝑎 𝐴2 𝑦𝑚 𝑏 𝐴2 𝑦𝑛 𝑏
=𝑟 𝑎 − 𝑦𝑚 𝑎 − − 𝑦𝑛 𝑎 +𝑟 𝑏 − 𝑦𝑛 𝑏 − − 𝑦𝑚 𝑏
𝐵1 𝐵1 𝐵2 𝐵2
𝑏
֜ න 𝑝 𝑥 𝑦𝑚 𝑦𝑛 𝑑𝑥 = 0 (𝜆𝑚 ≠ 𝜆𝑛 )
𝑎
 Regular Sturm-Liouville problem
THEOREM 11.4.1: Properties of the Regular Sturm-Liouville Problem
• There exist an infinite number of real eigenvalues that can be arranged in
increasing order 𝜆1 < 𝜆2 < 𝜆3 < ⋯ such that 𝜆𝑘 → ∞ as 𝑘 → ∞.
• For each eigenvalue there is only one eigenfunction (except for nonzero constant
multiples, i.e. up to scale) and the eigenfunctions are linearly independent.
• The set of eigenfunctions corresponding to the set of eigenvalues is orthogonal
with respect to the weight function 𝑝(𝑥) on the interval [𝑎, 𝑏].

Sturm-Liouville problem: Example 1

Solve Eigenvalues:
(1)𝑦 ′ ′ + 0 + 𝜆(1) 𝑦 = 0 𝑘𝜋 2
on 𝜆𝑘 = , 𝑘 = 1,2,3, …
𝐿
𝑥 ∈ [0, 𝐿]
subject to
(1)𝑦 0 + 0 𝑦′(0) = 0 Eigenfunctions:
and 𝑘𝜋
(1)𝑦 𝐿 + 0 𝑦 ′ 𝐿 = 0 𝑦𝑘 = sin 𝑥 , 𝑘 = 1,2,3, …
𝐿
Sturm-Liouville problem: Example 2 (DIY)
Solve
𝑦 ′′ + 𝜆𝑦 = 0
on
𝑥 ∈ [0, 𝐿]
subject to
𝑦′(0) = 0
and
𝑦′ 𝐿 = 0

Eigenvalue 𝜆 = 0 with Eigenfunction 𝑦 = 𝐶

𝑘𝜋 2
Eigenvalues: 𝜆𝑘 = , 𝑘 = 1,2,3, …
𝐿
𝑘𝜋
Eigenfunctions: 𝑦𝑘 = 𝐵 cos 𝑥 , 𝑘 = 1,2,3, …
𝐿
Sturm-Liouville problem: Example 2 (DIY)
Solve
𝑦 ′′ + 𝜆𝑦 = 0
on
𝑥 ∈ [0, 𝐿]
subject to
𝑦′(0) = 0
and
𝑦′ 𝐿 = 0

Eigenvalue 𝜆 = 0 with Eigenfunction 𝑦 = 1

𝑘𝜋 2
Eigenvalues: 𝜆𝑘 = , 𝑘 = 1,2,3, …
𝐿
𝑘𝜋
Eigenfunctions: 𝑦𝑘 = cos 𝑥 , 𝑘 = 1,2,3, …
𝐿
Sturm-Liouville problem: Example 2 (DIY)
Solve
𝑦 ′′ + 𝜆𝑦 = 0
on
𝑥 ∈ [0, 𝐿]
subject to
𝑦′(0) = 0
and
𝑦′ 𝐿 = 0

𝑘𝜋 2
Eigenvalues: 𝜆𝑘 = , 𝑘 = 0,1,2,3, …
𝐿
𝑘𝜋
Eigenfunctions: 𝑦𝑘 = cos 𝑥 , 𝑘 = 0,1,2,3, …
𝐿
Sturm-Liouville problem: Example 3
Solve
𝑦 ′′ + 𝜆𝑦 = 0
on
𝑥 ∈ [0,1]
subject to
𝑦(0) = 0
and
𝑦(1) + 𝑦′ 1 = 0
 Sturm-Liouville problem: Example 3
Solve
𝑦 ′′ + 𝜆𝑦 = 0 • 𝜆=0
on
𝑥 ∈ [0,1] 𝑦′′ = 0
subject to 𝑦′ = 𝐴
𝑦(0) = 0 and 𝑦 = 𝐴𝑥 + 𝐵
𝑦 1 + 𝑦′(1) = 0
Substitute boundary conditions:
𝑦 0 = 0: 𝑦 1 + 𝑦′(1) = 0:
0 = 𝐴(0) + 𝐵 0=𝐴 1 +𝐴
֜𝐵=0 ֜𝐴=0
֜ 𝑦 = 𝐴𝑥 ֜𝑦=0

The trivial solution!

 Sturm-Liouville problem: Example 3
Solve
𝑦 ′′ + 𝜆𝑦 = 0 • 0 > 𝜆 = −𝛼 2
on
𝑥 ∈ [0,1] 𝑦 ′′ − 𝛼 2 𝑦 = 0
subject to 𝑦 = 𝐴 cosh 𝛼𝑥 + 𝐵 sinh 𝛼𝑥
𝑦(0) = 0 and
𝑦 1 + 𝑦′(1) = 0
Substitute boundary conditions:
𝑦 0 = 0: 𝑦 1 + 𝑦′(1) = 0:
0 = 𝐴(1) + 𝐵 0 0 = 𝐵(sinh 𝛼 + 𝛼 cosh 𝛼 )
֜𝐴=0 ֜𝐵=0
֜ 𝑦 = 𝐵 sinh 𝛼𝑥 ֜𝑦=0

The trivial solution!

 Sturm-Liouville problem: Example 3
Solve
𝑦 ′′ + 𝜆𝑦 = 0 • 0 < 𝜆 = 𝛼2
on
𝑥 ∈ [0,1] 𝑦 ′′ + 𝛼 2 𝑦 = 0
subject to 𝑦 = 𝐴 cos 𝛼𝑥 + 𝐵 sin 𝛼𝑥
𝑦(0) = 0 and
𝑦 1 + 𝑦′(1) = 0
Substitute boundary conditions:
𝑦 0 = 0: 𝑦 1 + 𝑦′(1) = 0:
0 = 𝐴(1) + 𝐵 0 0 = 𝐵(sin 𝛼 + 𝛼 cos 𝛼 )
֜𝐴=0 ֜ 𝐵 =0?
֜ 𝑦 = 𝐵 sin 𝛼𝑥 ֜ 𝑦=0

The trivial solution!

 Sturm-Liouville problem: Example 3
Solve
𝑦 ′′ + 𝜆𝑦 = 0 • 0 < 𝜆 = 𝛼2
on
𝑥 ∈ [0,1] 𝑦 ′′ + 𝛼 2 𝑦 = 0
subject to 𝑦 = 𝐴 cos 𝛼𝑥 + 𝐵 sin 𝛼𝑥
𝑦(0) = 0 and
𝑦 1 + 𝑦′(1) = 0
Substitute boundary conditions:
𝑦 0 = 0: 𝑦 1 + 𝑦′(1) = 0:
0 = 𝐴(1) + 𝐵 0 0 = 𝐵(sin 𝛼 + 𝛼 cos 𝛼 )
֜𝐴=0 sin 𝛼 + 𝛼 cos 𝛼 = 0
֜ 𝑦 = 𝐵 sin 𝛼𝑥
sin 𝛼
= −𝛼
cos 𝛼
tan(𝛼) = −𝛼
Sturm-Liouville problem: Example 3
• 0 < 𝜆 = 𝛼2

tan(𝛼) = −𝛼
 Sturm-Liouville problem: Example 3
• 0 < 𝜆 = 𝛼2

𝛼1 𝛼2 𝛼3 𝛼4 𝛼5 𝛼6

tan(𝛼) = −𝛼
 Sturm-Liouville problem: Example 3
• 0 < 𝜆 = 𝛼2

𝛼2
𝛼1 𝛼3 𝛼4 𝛼5 𝛼6

sin 𝛼 + 𝛼 cos 𝛼 = 0
Sturm-Liouville problem: Example 3

sin 𝛼 + 𝛼 cos 𝛼 = 0
 Sturm-Liouville problem: Example 3
Solve
𝑦 ′′ + 𝜆𝑦 = 0 • 0 < 𝜆 = 𝛼2
on
𝑥 ∈ [0,1] 𝑦 ′′ + 𝛼 2 𝑦 = 0
subject to 𝑦 = 𝐴 cos 𝛼𝑥 + 𝐵 sin 𝛼𝑥
𝑦(0) = 0 and
𝑦 1 + 𝑦′(1) = 0
Substitute boundary conditions:
𝑦 0 = 0: 𝑦 1 + 𝑦′(1) = 0:
0 = 𝐴(1) + 𝐵 0 0 = 𝐵(sin 𝛼 + 𝛼 cos 𝛼 )
֜𝐴=0 𝑓 𝛼 = sin 𝛼 + 𝛼 cos 𝛼
֜ 𝑦 = 𝐵 sin 𝛼𝑥

Eigenvalues: 𝜆𝑘 = 𝛼𝑘2 , 𝑘 = 1,2,3, … (𝛼𝑘 are the positive roots of 𝑓(𝛼))

Eigenfunctions: 𝑦𝑘 = 𝐵 sin 𝛼𝑘 𝑥 , 𝑘 = 1,2,3, …
End of Chapter 11
You may omit:
• 11.5.1 Fourier-Bessel series
• 11.5.2 Fourier-Legendre series

42
PDEs
General concepts

43
Ordinary Derivatives
𝑦(𝑥) 𝑑𝑦
𝑦 ′ 𝑥1 = ቤ
𝑑𝑥 𝑥=𝑥
1

𝑥1 𝑥
Ordinary Derivatives
𝑦(𝑥) 𝑑𝑦
𝑦 ′ 𝑥1 = ቤ
𝑑𝑥 𝑥=𝑥
1

𝑥1 𝑥

𝑑𝑦 𝑦 𝑥1 + Δ𝑥 − 𝑦(𝑥1 )
ቤ = lim
𝑑𝑥 𝑥=𝑥 Δ𝑥→0 Δ𝑥
1

𝑥1 𝑥1 + Δ𝑥
Ordinary Derivatives
𝑦(𝑥) 𝑑𝑦
𝑦 ′ 𝑥1 = ቤ
𝑑𝑥 𝑥=𝑥
1

𝑥1 𝑥

𝑑𝑦 𝑦 𝑥1 + Δ𝑥 − 𝑦(𝑥1 )
ቤ = lim
𝑑𝑥 𝑥=𝑥 Δ𝑥→0 Δ𝑥
1

𝑥1 − Δ𝑥 𝑥1
𝑑𝑦 𝑦 𝑥1 − 𝑦(𝑥1 − Δ𝑥)
ቤ = lim
𝑑𝑥 𝑥=𝑥 Δ𝑥→0 Δ𝑥
1
Ordinary Derivatives
𝑦(𝑥) 𝑑𝑦
𝑦 ′ 𝑥1 = ቤ
𝑑𝑥 𝑥=𝑥
1

𝑥1 𝑥

𝑑𝑦 𝑦 𝑥1 + Δ𝑥 − 𝑦(𝑥1 )
ቤ = lim
𝑑𝑥 𝑥=𝑥 Δ𝑥→0 Δ𝑥
1

Δ𝑥 Δ𝑥
𝑑𝑦 𝑦 𝑥1 + 2 − 𝑦(𝑥1 − 2 )
ቤ = lim
𝑑𝑥 𝑥=𝑥 Δ𝑥→0 Δ𝑥
1 Δ𝑥 𝑥1 Δ𝑥
𝑥1 − 𝑥1 +
𝑑𝑦 𝑦 𝑥1 − 𝑦(𝑥1 − Δ𝑥) 2 2
ቤ = lim
𝑑𝑥 𝑥=𝑥 Δ𝑥→0 Δ𝑥
1
Ordinary Derivatives
𝑦(𝑥) 𝑑𝑦
𝑦 ′ 𝑥1 = ቤ
𝑑𝑥 𝑥=𝑥
1

𝑥1 𝑥

𝑑𝑦 𝑦 𝑥1 + Δ𝑥 − 𝑦(𝑥1 ) Forward differencing

ቤ ≈
𝑑𝑥 𝑥=𝑥 Δ𝑥
1

Δ𝑥 Δ𝑥
𝑑𝑦 𝑦 𝑥1 + 2 − 𝑦(𝑥1 − 2 )
ቤ ≈ Central differencing
𝑑𝑥 𝑥=𝑥 Δ𝑥
1 𝑥1 − Δ𝑥 𝑥1 𝑥1 + Δ𝑥
𝑑𝑦 𝑦 𝑥1 − 𝑦(𝑥1 − Δ𝑥)
ቤ ≈ Backward differencing
𝑑𝑥 𝑥=𝑥 Δ𝑥
1
Ordinary Derivatives
𝑦(𝑥) 𝑑𝑦
𝑦 ′ 𝑥1 = ቤ
𝑑𝑥 𝑥=𝑥
1

𝑥1 𝑥

𝑑𝑦 𝑦 𝑥1 + Δ𝑥 − 𝑦(𝑥1 ) Forward differencing

ቤ ≈
𝑑𝑥 𝑥=𝑥 Δ𝑥
1

𝑑𝑦 𝑦 𝑥1 + Δ𝑥 − 𝑦(𝑥1 − Δ𝑥)
ቤ ≈ Central differencing
𝑑𝑥 𝑥=𝑥 2Δ𝑥
1
𝑥1 − Δ𝑥 𝑥1 𝑥1 + Δ𝑥
𝑑𝑦 𝑦 𝑥1 − 𝑦(𝑥1 − Δ𝑥)
ቤ ≈ Backward differencing
𝑑𝑥 𝑥=𝑥 Δ𝑥
1
Partial Derivatives
Partial Derivatives
Partial Derivatives
Partial Derivatives

𝑓 𝑥, 𝑦 = 2𝑦𝑥 2 + 3𝑦 3
𝜕𝑓
= 4𝑦𝑥
𝜕𝑥
𝜕𝑓
= 2𝑥 2 + 9𝑦 2
𝜕𝑦
Notation

𝜕2𝑢 𝜕 2𝑢
Wave: = 𝑐 2 𝑢𝑡𝑡 = 𝑐 2 𝑢𝑥𝑥
𝜕𝑡 2 𝜕𝑥 2

𝜕𝑢 𝜕2𝑢
Heat: =𝜎 2 𝑢𝑡 = 𝜎𝑢𝑥𝑥
𝜕𝑡 𝜕𝑥

𝜕2𝑢 𝜕2𝑢
Laplace: + 2=0 𝑢𝑥𝑥 + 𝑢𝑦𝑦 = 0
𝜕𝑥 2 𝜕𝑦
Order
The highest order derivative in the PDE determines the
order of the PDE

First order: 𝑢𝑥 𝑢 + 𝑢𝑦2 = 0

Second order: 𝑢𝑥𝑦 + 𝑢𝑥 𝑢 + 𝑢𝑦2 = 0

Third order: 𝑢𝑥𝑥 𝑢 + 𝑢𝑥𝑥𝑦 + 𝑢𝑦2 = 0

Classification
For the discussion we are only going to consider general
second order PDEs
Fully non-linear: 𝑓 𝑥, 𝑦, 𝑢, 𝑢𝑥 , 𝑢𝑦 , 𝑢𝑥𝑥 , 𝑢𝑥𝑦 , 𝑢𝑦𝑦 = 0
𝑎 𝑥, 𝑦, 𝑢, 𝑢𝑥 , 𝑢𝑦 , 𝑢𝑥𝑥 , 𝑢𝑥𝑦 , 𝑢𝑦𝑦 𝑢𝑥𝑥 + 𝑏 𝑥, 𝑦, 𝑢, 𝑢𝑥 , 𝑢𝑦 , 𝑢𝑥𝑥 , 𝑢𝑥𝑦 , 𝑢𝑦𝑦 𝑢𝑥𝑦
+𝑐 𝑥, 𝑦, 𝑢, 𝑢𝑥 , 𝑢𝑦 , 𝑢𝑥𝑥 , 𝑢𝑥𝑦 , 𝑢𝑦𝑦 𝑢𝑦𝑦 + 𝑑 𝑥, 𝑦, 𝑢, 𝑢𝑥 , 𝑢𝑦 , 𝑢𝑥𝑥 , 𝑢𝑥𝑦 , 𝑢𝑦𝑦 = 0
Fully non-linear: 𝑓 𝑥, 𝑦, 𝑢, 𝑢𝑥 , 𝑢𝑦 , 𝑢𝑥𝑥 , 𝑢𝑥𝑦 , 𝑢𝑦𝑦 = 0
𝑎 𝑥, 𝑦, 𝑢, 𝑢𝑥 , 𝑢𝑦 , 𝑢𝑥𝑥 , 𝑢𝑥𝑦 , 𝑢𝑦𝑦 𝑢𝑥𝑥 + 𝑏 𝑥, 𝑦, 𝑢, 𝑢𝑥 , 𝑢𝑦 , 𝑢𝑥𝑥 , 𝑢𝑥𝑦 , 𝑢𝑦𝑦 𝑢𝑥𝑦
+𝑐 𝑥, 𝑦, 𝑢, 𝑢𝑥 , 𝑢𝑦 , 𝑢𝑥𝑥 , 𝑢𝑥𝑦 , 𝑢𝑦𝑦 𝑢𝑦𝑦 + 𝑑 𝑥, 𝑦, 𝑢, 𝑢𝑥 , 𝑢𝑦 , 𝑢𝑥𝑥 , 𝑢𝑥𝑦 , 𝑢𝑦𝑦 = 0

Quasi-linear: the coefficients of the highest order derivative must be

at least one order lower
Fully non-linear: 𝑓 𝑥, 𝑦, 𝑢, 𝑢𝑥 , 𝑢𝑦 , 𝑢𝑥𝑥 , 𝑢𝑥𝑦 , 𝑢𝑦𝑦 = 0
𝑎 𝑥, 𝑦, 𝑢, 𝑢𝑥 , 𝑢𝑦 , 𝑢𝑥𝑥 , 𝑢𝑥𝑦 , 𝑢𝑦𝑦 𝑢𝑥𝑥 + 𝑏 𝑥, 𝑦, 𝑢, 𝑢𝑥 , 𝑢𝑦 , 𝑢𝑥𝑥 , 𝑢𝑥𝑦 , 𝑢𝑦𝑦 𝑢𝑥𝑦
+𝑐 𝑥, 𝑦, 𝑢, 𝑢𝑥 , 𝑢𝑦 , 𝑢𝑥𝑥 , 𝑢𝑥𝑦 , 𝑢𝑦𝑦 𝑢𝑦𝑦 + 𝑑 𝑥, 𝑦, 𝑢, 𝑢𝑥 , 𝑢𝑦 , 𝑢𝑥𝑥 , 𝑢𝑥𝑦 , 𝑢𝑦𝑦 = 0

Quasi-linear: the coefficients of the highest order derivative must be

at least one order lower
Semi-linear: terms containing highest order derivatives must be
linear (in 𝑢 and its derivatives)
Fully non-linear: 𝑓 𝑥, 𝑦, 𝑢, 𝑢𝑥 , 𝑢𝑦 , 𝑢𝑥𝑥 , 𝑢𝑥𝑦 , 𝑢𝑦𝑦 = 0
𝑎 𝑥, 𝑦, 𝑢, 𝑢𝑥 , 𝑢𝑦 , 𝑢𝑥𝑥 , 𝑢𝑥𝑦 , 𝑢𝑦𝑦 𝑢𝑥𝑥 + 𝑏 𝑥, 𝑦, 𝑢, 𝑢𝑥 , 𝑢𝑦 , 𝑢𝑥𝑥 , 𝑢𝑥𝑦 , 𝑢𝑦𝑦 𝑢𝑥𝑦
+𝑐 𝑥, 𝑦, 𝑢, 𝑢𝑥 , 𝑢𝑦 , 𝑢𝑥𝑥 , 𝑢𝑥𝑦 , 𝑢𝑦𝑦 𝑢𝑦𝑦 + 𝑑 𝑥, 𝑦 𝑢𝑥 + 𝑒 𝑥, 𝑦 𝑢𝑦 + 𝑓 𝑥, 𝑦 𝑢
+ 𝑔(𝑥, 𝑦) = 0

Quasi-linear: the coefficients of the highest order derivative must be

at least one order lower
Semi-linear: terms containing highest order derivatives must be
linear (in 𝑢 and its derivatives)

Linear: all term must be linear (in 𝑢 and its derivatives)

Fully non-linear: 𝑓 𝑥, 𝑦, 𝑢, 𝑢𝑥 , 𝑢𝑦 , 𝑢𝑥𝑥 , 𝑢𝑥𝑦 , 𝑢𝑦𝑦 = 0
𝑎 𝑥, 𝑦, 𝑢, 𝑢𝑥 , 𝑢𝑦 , 𝑢𝑥𝑥 , 𝑢𝑥𝑦 , 𝑢𝑦𝑦 𝑢𝑥𝑥 + 𝑏 𝑥, 𝑦, 𝑢, 𝑢𝑥 , 𝑢𝑦 , 𝑢𝑥𝑥 , 𝑢𝑥𝑦 , 𝑢𝑦𝑦 𝑢𝑥𝑦
+𝑐 𝑥, 𝑦, 𝑢, 𝑢𝑥 , 𝑢𝑦 , 𝑢𝑥𝑥 , 𝑢𝑥𝑦 , 𝑢𝑦𝑦 𝑢𝑦𝑦 + 𝑑 𝑥, 𝑦 𝑢𝑥 + 𝑒 𝑥, 𝑦 𝑢𝑦 + 𝑓 𝑥, 𝑦 𝑢
+ 𝑔(𝑥, 𝑦) = 0

Quasi-linear: the coefficients of the highest order derivative must be

at least one order lower
Semi-linear: terms containing highest order derivatives must be
linear (in 𝑢 and its derivatives)

Linear: all term must be linear (in 𝑢 and its derivatives)

Homogeneous: no terms in 𝑥 or 𝑦 only

Classification of semi-linear 2nd order PDEs:

𝑎 𝑥, 𝑦 𝑢𝑥𝑥 + 𝑏 𝑥, 𝑦 𝑢𝑥𝑦 + 𝑐 𝑥, 𝑦 𝑢𝑦𝑦 + 𝑑 𝑥, 𝑦, 𝑢, 𝑢𝑥 , 𝑢𝑦 = 0

∆ = 𝑏(𝑥, 𝑦)2 − 4𝑎(𝑥, 𝑦)𝑐(𝑥, 𝑦)

Classification of semi-linear 2nd order PDEs:

𝑎 𝑥, 𝑦 𝑢𝑥𝑥 + 𝑏 𝑥, 𝑦 𝑢𝑥𝑦 + 𝑐 𝑥, 𝑦 𝑢𝑦𝑦 + 𝑑 𝑥, 𝑦, 𝑢, 𝑢𝑥 , 𝑢𝑦 = 0

∆ = 𝑏(𝑥, 𝑦)2 − 4𝑎(𝑥, 𝑦)𝑐(𝑥, 𝑦)

Elliptic: Δ < 0

Parabolic: Δ = 0

Hyperbolic: Δ > 0
Classification of semi-linear 2nd order PDEs:

𝑎 𝑥, 𝑦 𝑢𝑥𝑥 + 𝑏 𝑥, 𝑦 𝑢𝑥𝑦 + 𝑐 𝑥, 𝑦 𝑢𝑦𝑦 + 𝑑 𝑥, 𝑦, 𝑢, 𝑢𝑥 , 𝑢𝑦 = 0

∆ = 𝑏(𝑥, 𝑦)2 − 4𝑎(𝑥, 𝑦)𝑐(𝑥, 𝑦)

Elliptic: Δ < 0 Laplace’s equation:

𝑢𝑥𝑥 + 𝑢𝑦𝑦 = 0
Parabolic: Δ = 0 𝑎 = 1; 𝑏 = 0; 𝑐 = 1

Δ = −4 < 0
Hyperbolic: Δ > 0
Classification of semi-linear 2nd order PDEs:

𝑎 𝑥, 𝑦 𝑢𝑥𝑥 + 𝑏 𝑥, 𝑦 𝑢𝑥𝑦 + 𝑐 𝑥, 𝑦 𝑢𝑦𝑦 + 𝑑 𝑥, 𝑦, 𝑢, 𝑢𝑥 , 𝑢𝑦 = 0

∆ = 𝑏(𝑥, 𝑦)2 − 4𝑎(𝑥, 𝑦)𝑐(𝑥, 𝑦)

Elliptic: Δ < 0 Wave equation:

𝑢𝑡𝑡 − 𝐶 2 𝑢𝑥𝑥 = 0
Parabolic: Δ = 0 𝑎 = 1; 𝑏 = 0; 𝑐 = −𝐶 2

Δ = 4𝐶 2 > 0
Hyperbolic: Δ > 0
Classification of semi-linear 2nd order PDEs:

𝑎 𝑥, 𝑦 𝑢𝑥𝑥 + 𝑏 𝑥, 𝑦 𝑢𝑥𝑦 + 𝑐 𝑥, 𝑦 𝑢𝑦𝑦 + 𝑑 𝑥, 𝑦, 𝑢, 𝑢𝑥 , 𝑢𝑦 = 0

∆ = 𝑏(𝑥, 𝑦)2 − 4𝑎(𝑥, 𝑦)𝑐(𝑥, 𝑦)

Elliptic: Δ < 0 Heat equation:

𝑢𝑡 − 𝜎𝑢𝑥𝑥 = 0
Parabolic: Δ = 0 𝑎 = 0; 𝑏 = 0; 𝑐 = −𝜎

Δ=0
Hyperbolic: Δ > 0
The superposition principle
Applicable to linear, homogeneous PDEs
A general second order, linear, homogeneous PDE:
𝑎 𝑥, 𝑦 𝑢𝑥𝑥 + 𝑏 𝑥, 𝑦 𝑢𝑥𝑦 + 𝑐 𝑥, 𝑦 𝑢𝑦𝑦 + 𝑑 𝑥, 𝑦 𝑢𝑥 + 𝑒 𝑥, 𝑦 𝑢𝑦 + 𝑓 𝑥, 𝑦 𝑢 = 0
- Linear: all term must be linear (in 𝑢 and its derivatives)
- Homogeneous: no terms in 𝑥 or 𝑦 only

If 𝑢1 and 𝑢2 are both

solutions of the PDE, then
𝑤 = 𝛼𝑢1 + 𝛽𝑢2 is also a
solution, where 𝛼 and 𝛽
are constants.
 If 𝑢1 and 𝑢2 are both
The superposition principle solutions of the PDE, then
𝑤 = 𝛼𝑢1 + 𝛽𝑢2 is also a
Example solution, where 𝛼 and 𝛽
are constants.
A third order, linear, homogeneous PDE:
𝐴 𝑥, 𝑦 𝑢𝑥𝑦𝑦 (𝑥, 𝑦) + 𝑢 𝑥, 𝑦 + 𝐵(𝑥, 𝑦)𝑢𝑦 (𝑥, 𝑦) = 0

Since 𝑢1 and 𝑢2 solves the PDE, we know that

𝐴 𝑥, 𝑦 𝑢1 𝑥𝑦𝑦 (𝑥, 𝑦) + 𝑢1 𝑥, 𝑦 + 𝐵(𝑥, 𝑦)𝑢1 𝑦 (𝑥, 𝑦) = 0
𝐴 𝑥, 𝑦 𝑢2 𝑥𝑦𝑦 (𝑥, 𝑦) + 𝑢2 𝑥, 𝑦 + 𝐵(𝑥, 𝑦)𝑢2 𝑦 (𝑥, 𝑦) = 0

Let’s check if 𝑤(𝑥, 𝑦) = 𝛼𝑢1 (𝑥, 𝑦) + 𝛽𝑢2 (𝑥, 𝑦) also solves the PDE:
𝐴 𝑥, 𝑦 𝑤𝑥𝑦𝑦 (𝑥, 𝑦) + 𝑤 𝑥, 𝑦 + 𝐵 𝑥, 𝑦 𝑤𝑦 𝑥, 𝑦

= 𝐴 𝑥, 𝑦 𝛼𝑢1 (𝑥, 𝑦) + 𝛽𝑢2 (𝑥, 𝑦) 𝑥𝑦𝑦 + 𝛼𝑢1 (𝑥, 𝑦) + 𝛽𝑢2 (𝑥, 𝑦)

+ 𝐵 𝑥, 𝑦 𝛼𝑢1 (𝑥, 𝑦) + 𝛽𝑢2 (𝑥, 𝑦) 𝑦

= 𝐴 𝑥, 𝑦 𝛼𝑢1 𝑥𝑦𝑦 𝑥, 𝑦 + 𝐴 𝑥, 𝑦 𝛽𝑢2 𝑥𝑦𝑦 𝑥, 𝑦 + 𝛼𝑢1 𝑥, 𝑦 + 𝛽𝑢2 𝑥, 𝑦

+ 𝐵 𝑥, 𝑦 𝛼𝑢1 𝑦 𝑥, 𝑦 + 𝐵 𝑥, 𝑦 𝛽𝑢2 𝑦 (𝑥, 𝑦)
 If 𝑢1 and 𝑢2 are both
The superposition principle solutions of the PDE, then
𝑤 = 𝛼𝑢1 + 𝛽𝑢2 is also a
Example solution, where 𝛼 and 𝛽
are constants.
A third order, linear, homogeneous PDE:
𝐴 𝑥, 𝑦 𝑢𝑥𝑦𝑦 (𝑥, 𝑦) + 𝑢 𝑥, 𝑦 + 𝐵(𝑥, 𝑦)𝑢𝑦 (𝑥, 𝑦) = 0

Since 𝑢1 and 𝑢2 solves the PDE, we know that

𝐴 𝑥, 𝑦 𝑢1 𝑥𝑦𝑦 (𝑥, 𝑦) + 𝑢1 𝑥, 𝑦 + 𝐵(𝑥, 𝑦)𝑢1 𝑦 (𝑥, 𝑦) = 0
𝐴 𝑥, 𝑦 𝑢2 𝑥𝑦𝑦 (𝑥, 𝑦) + 𝑢2 𝑥, 𝑦 + 𝐵(𝑥, 𝑦)𝑢2 𝑦 (𝑥, 𝑦) = 0

Let’s check if 𝑤(𝑥, 𝑦) = 𝛼𝑢1 (𝑥, 𝑦) + 𝛽𝑢2 (𝑥, 𝑦) also solves the PDE:
𝐴 𝑥, 𝑦 𝑤𝑥𝑦𝑦 (𝑥, 𝑦) + 𝑤 𝑥, 𝑦 + 𝐵 𝑥, 𝑦 𝑤𝑦 𝑥, 𝑦

= 𝐴 𝑥, 𝑦 𝛼𝑢1 𝑥𝑦𝑦 𝑥, 𝑦 + 𝐴 𝑥, 𝑦 𝛽𝑢2 𝑥𝑦𝑦 𝑥, 𝑦 + 𝛼𝑢1 𝑥, 𝑦 + 𝛽𝑢2 𝑥, 𝑦

+ 𝐵 𝑥, 𝑦 𝛼𝑢1 𝑦 𝑥, 𝑦 + 𝐵 𝑥, 𝑦 𝛽𝑢2 𝑦 (𝑥, 𝑦)

= 𝛼 𝐴 𝑥, 𝑦 𝑢1 𝑥𝑦𝑦 𝑥, 𝑦 + 𝑢1 𝑥, 𝑦 + 𝐵 𝑥, 𝑦 𝑢1 𝑦 𝑥, 𝑦
+ 𝛽 𝐴 𝑥, 𝑦 𝑢2 𝑥𝑦𝑦 𝑥, 𝑦 + 𝑢2 𝑥, 𝑦 + 𝐵 𝑥, 𝑦 𝑢2 𝑦 (𝑥, 𝑦)
 If 𝑢1 and 𝑢2 are both
The superposition principle solutions of the PDE, then
𝑤 = 𝛼𝑢1 + 𝛽𝑢2 is also a
Example solution, where 𝛼 and 𝛽
are constants.
A third order, linear, homogeneous PDE:
𝐴 𝑥, 𝑦 𝑢𝑥𝑦𝑦 (𝑥, 𝑦) + 𝑢 𝑥, 𝑦 + 𝐵(𝑥, 𝑦)𝑢𝑦 (𝑥, 𝑦) = 0

Since 𝑢1 and 𝑢2 solves the PDE, we know that

𝐴 𝑥, 𝑦 𝑢1 𝑥𝑦𝑦 (𝑥, 𝑦) + 𝑢1 𝑥, 𝑦 + 𝐵(𝑥, 𝑦)𝑢1 𝑦 (𝑥, 𝑦) = 0
𝐴 𝑥, 𝑦 𝑢2 𝑥𝑦𝑦 (𝑥, 𝑦) + 𝑢2 𝑥, 𝑦 + 𝐵(𝑥, 𝑦)𝑢2 𝑦 (𝑥, 𝑦) = 0

Let’s check if 𝑤(𝑥, 𝑦) = 𝛼𝑢1 (𝑥, 𝑦) + 𝛽𝑢2 (𝑥, 𝑦) also solves the PDE:
𝐴 𝑥, 𝑦 𝑤𝑥𝑦𝑦 (𝑥, 𝑦) + 𝑤 𝑥, 𝑦 + 𝐵 𝑥, 𝑦 𝑤𝑦 𝑥, 𝑦

= 𝐴 𝑥, 𝑦 𝛼𝑢1 𝑥𝑦𝑦 𝑥, 𝑦 + 𝐴 𝑥, 𝑦 𝛽𝑢2 𝑥𝑦𝑦 𝑥, 𝑦 + 𝛼𝑢1 𝑥, 𝑦 + 𝛽𝑢2 𝑥, 𝑦

+ 𝐵 𝑥, 𝑦 𝛼𝑢1 𝑦 𝑥, 𝑦 + 𝐵 𝑥, 𝑦 𝛽𝑢2 𝑦 (𝑥, 𝑦)
=0
Therefore, 𝑤(𝑥, 𝑦) also solves the PDE.
 If 𝑢1 and 𝑢2 are both
The superposition principle solutions of the PDE, then
𝑤 = 𝛼𝑢1 + 𝛽𝑢2 is also a
Do it yourself solution, where 𝛼 and 𝛽
are constants.

- Show that 𝑤 = 𝛼𝑢1 + 𝛽𝑢2 is not a solution of the nonhomogeneous PDE,

𝐴 𝑥, 𝑦 𝑢𝑥𝑦𝑦 𝑥, 𝑦 + 𝑢 𝑥, 𝑦 + 𝐵 𝑥, 𝑦 𝑢𝑦 𝑥, 𝑦 + 𝐶(𝑥, 𝑦) = 0.

- Show that 𝑤 = 𝛼𝑢1 + 𝛽𝑢2 is not a solution of the non-linear PDE,

𝐴 𝑥, 𝑦 𝑢𝑥𝑦𝑦 𝑥, 𝑦 + 𝑢 𝑥, 𝑦 + 𝐵 𝑥, 𝑦 𝑢 𝑥, 𝑦 𝑢𝑦 𝑥, 𝑦 = 0.
Boundary value problem
A PDE governs the physical phenomenon we are modelling.
To find a unique solution for the specific problem, we also
need to state the domain and the boundary conditions.
Boundary value problem
Under defined ODE:
ODE for SHM: 𝑢ሷ + 𝜔2 𝑢 = 0
Domain: 𝑡 ∈ [0, ∞)
BC: 𝑢 0 = 0.5 m

Most general solution:

𝑢 𝑡 = 𝐴 sin(𝜔𝑡) + 𝐵 cos(𝜔𝑡)

More specific solution (imposing BC):

𝑢 𝑡 = 𝐴 sin(𝜔𝑡) + 0.5 cos(𝜔𝑡)

Final solution…???
Boundary value problem
Under defined ODE:
𝑢

Final solution…???
Boundary value problem
Under defined ODE:
ODE for SHM: 𝑢ሷ + 𝜔2 𝑢 = 0
Domain: 𝑡 ∈ [0, ∞)
BC: 𝑢 0 = 0.5 m
…
Most general solution:
𝑢 𝑡 = 𝐴 sin(𝜔𝑡) + 𝐵 cos(𝜔𝑡)

More specific solution (imposing BC):

𝑢 𝑡 = 𝐴 sin(𝜔𝑡) + 0.5 cos(𝜔𝑡)

Final solution…???
Boundary value problem
Under defined ODE:
𝑢

Final solution…???
Boundary value problem
Under defined ODE:
𝑢

Final solution…???
Boundary value problem
Over defined ODE:
ODE for Exp. growth: 𝑢ሶ = 𝑢
Domain: 𝑡 ∈ [0, ∞)
BC: 𝑢 0 = 0.5 …[1]
𝑢ሶ 0 = 0.2 s−1 …[2]
Most general solution:
𝑢 𝑡 = 𝐶𝑒 𝑡
More specific solution (imposing BC[1]):
𝑢 𝑡 = 0.5𝑒 𝑡
But this is the “final solution” – no integration constant left
– but 𝑢ሶ 0 = 0.5, so BC[2] is not satisfied!!
Boundary value problem
Over defined ODE:
ODE for Exp. growth: 𝑢ሶ = 𝑢
Domain: 𝑡 ∈ [0, ∞)
BC: 𝑢 0 = 0.5 …[1]
𝑢ሶ 0 = 0.2 s−1 …[2]
Most general solution:
𝑢 𝑡 = 𝐶𝑒 𝑡
More specific solution (imposing BC[1]):
𝑢 𝑡 = 0.5𝑒 𝑡
But this is the “final solution” – no integration constant left
– but 𝑢ሶ 0 = 0.5, so BC[2] is not satisfied!!
Boundary value problem
A PDE governs the physical phenomenon we are modelling.
To find a unique solution for the specific problem, we also
need to state the domain and the boundary conditions.
A well-posed boundary value problem is a problem is not
under or over defined and has a unique solution.
Boundary value problem
Heat equation: 𝑢𝑡 = 𝜎𝑢𝑥𝑥 𝑢(𝑥, 𝑡)

𝑢
𝑡

𝑥
Boundary value problem
Heat equation: 𝑢𝑡 = 𝜎𝑢𝑥𝑥 𝑢(𝑥, 𝑡)
Domain: 𝑥 ∈ 0, 𝐿
𝑡 ∈ [0, ∞)

𝑢
𝑡 ∞
∞

𝑥
0 𝐿
Boundary value problem
Heat equation: 𝑢𝑡 = 𝜎𝑢𝑥𝑥 𝑢(𝑥, 𝑡)
Domain: 𝑥 ∈ 0, 𝐿
𝑡 ∈ [0, ∞)

Initial condition & Boundary conditions

𝑢
𝑡 ∞
∞
𝑢 0, 𝑡 = ?
𝑥 𝑢 𝐿, 𝑡 = ?
0 𝐿
𝑢 𝑥, 0 = ?
Heat equation: let’s play
Heat equation: 𝑢𝑡 = 𝜎𝑢𝑥𝑥 𝑢(𝑥, 𝑡)
Domain: 𝑥 ∈ 0, 𝐿
𝑡 ∈ [0, ∞)

𝑡 ∞
∞

𝑥
0 𝐿
Heat equation: let’s play
Heat equation: 𝑢𝑡 = 𝜎𝑢𝑥𝑥 𝑢(𝑥, 𝑡)
Domain: 𝑥 ∈ 0, 𝐿
𝑡 ∈ [0, ∞)

Initial condition:

𝑡 ∞
∞

𝑥
0 𝐿
Heat equation: let’s play
Heat equation: 𝑢𝑡 = 𝜎𝑢𝑥𝑥 𝑢(𝑥, 𝑡)
Domain: 𝑥 ∈ 0, 𝐿
𝑡 ∈ [0, ∞)

Initial condition: 𝑢(𝑥, 0) = 10sin ? 𝑥

𝑢

10
𝑡 ∞
∞

𝑥
0 𝐿
Heat equation: let’s play
Heat equation: 𝑢𝑡 = 𝜎𝑢𝑥𝑥 𝑢(𝑥, 𝑡)
Domain: 𝑥 ∈ 0, 𝐿
𝑡 ∈ [0, ∞)
3𝜋
Initial condition: 𝑢(𝑥, 0) = 10sin 𝑥
𝐿
Boundary conditions: 𝑢(0, 𝑡) = 0 𝑡 ∈ [0, ∞)
𝑢(𝐿, 𝑡) = 0 𝑡 ∈ [0, ∞)
𝑢

10
𝑡 ∞
∞

𝑥
0 𝐿
Heat equation: let’s play
𝑢𝑡 = 𝜎𝑢𝑥𝑥 on 𝑥 ∈ 0, 𝐿 , 𝑡 ∈ [0, ∞)
with: 3𝜋𝑥
[1] 𝑢(𝑥, 0) = 10 sin 𝑥 ∈ 0, 𝐿
𝐿
[2] 𝑢(0, 𝑡) = 0 𝑡 ∈ [0, ∞)
[3] 𝑢(𝐿, 𝑡) = 0 𝑡 ∈ [0, ∞)
[4] lim 𝑢(𝑥, 𝑡) is finite 𝑥 ∈ 0, 𝐿
𝑡→∞
Let’s just assume we know that the solution is of the form
(separation of variables):
−𝜎𝜆 2𝑡
𝑢 𝑥, 𝑡 = 𝑒 𝐴 cos 𝜆𝑥 + 𝐵 sin 𝜆𝑥
𝜎𝜆 2𝑡
+𝑒 𝐶 cosh 𝜆𝑥 + 𝐷 sinh 𝜆𝑥 + 𝐸𝑥 + 𝐹
Heat equation: let’s play
𝑢𝑡 = 𝜎𝑢𝑥𝑥 on 𝑥 ∈ 0, 𝐿 , 𝑡 ∈ [0, ∞)
with: 3𝜋𝑥
[1] 𝑢(𝑥, 0) = 10 sin 𝑥 ∈ 0, 𝐿
𝐿
[2] 𝑢(0, 𝑡) = 0 𝑡 ∈ [0, ∞)
[3] 𝑢(𝐿, 𝑡) = 0 𝑡 ∈ [0, ∞)
[4] lim 𝑢(𝑥, 𝑡) is finite 𝑥 ∈ 0, 𝐿
𝑡→∞
Let’s just assume we know that the solution is of the form
(separation of variables):
−𝜎𝜆 2𝑡
𝑢 𝑥, 𝑡 = 𝑒 𝐴 cos 𝜆𝑥 + 𝐵 sin 𝜆𝑥
𝜎𝜆 2𝑡
+𝑒 𝐶 cosh 𝜆𝑥 + 𝐷 sinh 𝜆𝑥 + 𝐸𝑥 + 𝐹
We now solve these 7 constants by imposing the boundary
conditions one by one.
Heat equation: let’s play
𝑢𝑡 = 𝜎𝑢𝑥𝑥 on 𝑥 ∈ 0, 𝐿 , 𝑡 ∈ [0, ∞)
with: 3𝜋𝑥
[1] 𝑢(𝑥, 0) = 10 sin 𝑥 ∈ 0, 𝐿
𝐿
[2] 𝑢(0, 𝑡) = 0 𝑡 ∈ [0, ∞)
[3] 𝑢(𝐿, 𝑡) = 0 𝑡 ∈ [0, ∞)
[4] lim 𝑢(𝑥, 𝑡) is finite 𝑥 ∈ 0, 𝐿
𝑡→∞

Most general solution:

−𝜎𝜆 2𝑡
𝑢 𝑥, 𝑡 = 𝑒 𝐴 cos 𝜆𝑥 + 𝐵 sin 𝜆𝑥
𝜎𝜆 2𝑡
+𝑒 𝐶 cosh 𝜆𝑥 + 𝐷 sinh 𝜆𝑥 + 𝐸𝑥 + 𝐹

Impose [4]: 𝐶 = 0, 𝐷 = 0
Heat equation: let’s play
𝑢𝑡 = 𝜎𝑢𝑥𝑥 on 𝑥 ∈ 0, 𝐿 , 𝑡 ∈ [0, ∞)
with: 3𝜋𝑥
[1] 𝑢(𝑥, 0) = 10 sin 𝑥 ∈ 0, 𝐿
𝐿
[2] 𝑢(0, 𝑡) = 0 𝑡 ∈ [0, ∞)
[3] 𝑢(𝐿, 𝑡) = 0 𝑡 ∈ [0, ∞)
[4] lim 𝑢(𝑥, 𝑡) is finite 𝑥 ∈ 0, 𝐿
𝑡→∞

More specific solution:

−𝜎𝜆 2𝑡
𝑢 𝑥, 𝑡 = 𝑒 𝐴 cos 𝜆𝑥 + 𝐵 sin 𝜆𝑥 + 𝐸𝑥 + 𝐹
Impose [2]:
−𝜎𝜆 2𝑡
𝑢 0, 𝑡 = 0 = 𝑒 𝐴+0 +𝐹
𝐴 = 0, 𝐹 = 0
Heat equation: let’s play
𝑢𝑡 = 𝜎𝑢𝑥𝑥 on 𝑥 ∈ 0, 𝐿 , 𝑡 ∈ [0, ∞)
with: 3𝜋𝑥
[1] 𝑢(𝑥, 0) = 10 sin 𝑥 ∈ 0, 𝐿
𝐿
[2] 𝑢(0, 𝑡) = 0 𝑡 ∈ [0, ∞)
[3] 𝑢(𝐿, 𝑡) = 0 𝑡 ∈ [0, ∞)
[4] lim 𝑢(𝑥, 𝑡) is finite 𝑥 ∈ 0, 𝐿
𝑡→∞

More specific solution:

−𝜎𝜆 2𝑡
𝑢 𝑥, 𝑡 = 𝑒 𝐵 sin 𝜆𝑥 + 𝐸𝑥
Impose [3]:
−𝜎𝜆 2𝑡
𝑢 𝐿, 𝑡 = 0 = 𝑒 𝐵 sin 𝜆𝐿 + 𝐸𝐿
𝐸 = 0, sin 𝜆𝐿 = 0 ֜ 𝜆𝐿 = 𝑘𝜋 𝑘 ∈ ℤ
Heat equation: let’s play
𝑢𝑡 = 𝜎𝑢𝑥𝑥 on 𝑥 ∈ 0, 𝐿 , 𝑡 ∈ [0, ∞)
with: 3𝜋𝑥
[1] 𝑢(𝑥, 0) = 10 sin 𝑥 ∈ 0, 𝐿
𝐿
[2] 𝑢(0, 𝑡) = 0 𝑡 ∈ [0, ∞)
[3] 𝑢(𝐿, 𝑡) = 0 𝑡 ∈ [0, ∞)
[4] lim 𝑢(𝑥, 𝑡) is finite 𝑥 ∈ 0, 𝐿
𝑡→∞

More specific solution:

2
𝑘𝜋
−𝜎 𝑡 𝑘𝜋𝑥
𝑢 𝑥, 𝑡 = 𝑒 𝐿 𝐵 sin
𝐿
Impose [1]:
3𝜋𝑥 𝑘𝜋𝑥
𝑢 𝑥, 0 = 10 sin = 𝐵 sin
𝐿 𝐿
𝐵 = 10, 𝑘 = 3
Heat equation: let’s play
𝑢𝑡 = 𝜎𝑢𝑥𝑥 on 𝑥 ∈ 0, 𝐿 , 𝑡 ∈ [0, ∞)
with: 3𝜋𝑥
[1] 𝑢(𝑥, 0) = 10 sin 𝑥 ∈ 0, 𝐿
𝐿
[2] 𝑢(0, 𝑡) = 0 𝑡 ∈ [0, ∞)
[3] 𝑢(𝐿, 𝑡) = 0 𝑡 ∈ [0, ∞)
[4] lim 𝑢(𝑥, 𝑡) is finite 𝑥 ∈ 0, 𝐿
𝑡→∞

Final solution:
3𝜋 2
−𝜎 𝑡 3𝜋𝑥
𝑢 𝑥, 𝑡 = 𝑒 𝐿 10 sin
𝐿
Final solution:
3𝜋 2
−𝜎 𝑡 3𝜋𝑥
𝑢 𝑥, 𝑡 = 𝑒 𝐿 10 sin
𝐿
But… what are the odds??
Most general solution:
−𝜎𝜆 2𝑡
𝑢 𝑥, 𝑡 = 𝑒 𝐴 cos 𝜆𝑥 + 𝐵 sin 𝜆𝑥
𝜎𝜆2 𝑡
+𝑒 𝐶 cosh 𝜆𝑥 + 𝐷 sinh 𝜆𝑥 + 𝐸𝑥 + 𝐹

𝑢 10𝑥 + 10 𝑥 ∈ [−1,0]
??? 𝑢(𝑥, 0) = ቊ
10 −10𝑥 + 10 𝑥 ∈ (0,1]
𝑡 ∞
∞

𝑥
−1 1
Most general solution:
2
𝑢 𝑥, 𝑡 = 𝑒 −𝜎𝜆 𝑡 𝐴 cos 𝜆𝑥 + 𝐵 sin 𝜆𝑥
𝜎𝜆 2𝑡
+𝑒 𝐶 cosh 𝜆𝑥 + 𝐷 sinh 𝜆𝑥 + 𝐸𝑥 + 𝐹

The plan of action:

Before we impose the boundary / initial condition, first
expand this function as a Fourier series.

𝑢 10𝑥 + 10 𝑥 ∈ [−1,0]
??? 𝑢(𝑥, 0) = ቊ
10 −10𝑥 + 10 𝑥 ∈ (0,1]
𝑡 ∞
∞

𝑥
−1 1
PDEs

Heat equation Wave equation

𝑢𝑡 = 𝜎𝑢𝑥𝑥 𝑢𝑡𝑡 = 𝑐 2 𝑢𝑥𝑥
Domain
𝑥 ∈ 0, 𝐿 , 𝑡 ∈ [0, ∞)

Heat equation Wave equation

𝑢𝑡 = 𝜎𝑢𝑥𝑥 𝑢𝑡𝑡 = 𝑐 2 𝑢𝑥𝑥

𝑢 (℃) 𝑢 (𝑚)

𝑡 ∞ 𝑡 ∞

𝑥 𝑥
0 𝐿 0 𝐿
Initial Conditions
Domain: 𝑥 ∈ 0, 𝐿 , 𝑡 ∈ [0, ∞)

Heat equation Wave equation

𝑢𝑡 = 𝜎𝑢𝑥𝑥 𝑢𝑡𝑡 = 𝑐 2 𝑢𝑥𝑥
subject to subject to
𝑢(𝑥, 0) = 𝑓(𝑥) 𝑢(𝑥, 0) = 𝑓(𝑥)
𝑢𝑡 (𝑥, 0) = 𝑔(𝑥)

𝑢 𝑢 (𝑚)

𝑡 ∞ 𝑡 ∞

𝑥 𝑥
0 𝐿 0 𝐿
 Boundary Conditions
Domain: 𝑥 ∈ 0, 𝐿 , 𝑡 ∈ [0, ∞)

Heat equation Wave equation

𝑢𝑡 = 𝜎𝑢𝑥𝑥 𝑢𝑡𝑡 = 𝑐 2 𝑢𝑥𝑥

(i) Dirichlet condition: 𝑢

𝜕𝑢
(ii) Neumann condition: , where 𝑛 is directed perpendicular to the boundary
𝜕𝑛
𝜕𝑢
(iii) Robin condition: + 𝐶𝑢 (𝐶 is constant)
𝜕𝑛

𝑢 𝑢 (𝑚)

𝑡 ∞ 𝑡 ∞

𝑥 𝑥
0 𝐿 0 𝐿
Boundary Conditions
Domain: 𝑥 ∈ 0, 𝐿 , 𝑡 ∈ [0, ∞)

Heat equation Wave equation

𝑢𝑡 = 𝜎𝑢𝑥𝑥 𝑢𝑡𝑡 = 𝑐 2 𝑢𝑥𝑥

(i) Dirichlet condition: 𝑢 0, 𝑡 = 𝑢0

(ii) Neumann condition: 𝑢𝑥 𝐿, 𝑡 = 0

𝑡 ∞

𝑥 Insulated
0 𝐿 (adiabatic BC)
 Boundary Conditions
Domain: 𝑥 ∈ 0, 𝐿 , 𝑡 ∈ [0, ∞)

Heat equation Wave equation

𝑢𝑡 = 𝜎𝑢𝑥𝑥 𝑢𝑡𝑡 = 𝑐 2 𝑢𝑥𝑥

𝜕𝑢
(iii) Robin condition: + 𝐶𝑢 (𝐶 is constant) Newton’s law of cooling:
𝜕𝑛 𝑑𝑢(𝑡)
= −ℎ 𝑢 𝑡 − 𝑢𝑚
−𝐾𝑢𝑥 0, 𝑡 = ℎ 𝑢 0, 𝑡 − 𝑢𝑚 (ℎ > 0) 𝑑𝑡
Fourier’s law of heat conduction:
−𝐾𝑢𝑥 𝐿, 𝑡 = ℎ 𝑢 𝐿, 𝑡 − 𝑢𝑚 (ℎ > 0) 𝑑𝑢
𝑞 = −𝐾
𝑑𝑥
𝑢
These boundary conditions says that
the heat flowing out of the rod by
conduction at the tip equals heat 𝑡 ∞
lost to air by convection — it's a 𝑢𝑚
balance of fluxes. 𝑢𝑚
𝑥
0 𝐿
Boundary Conditions
Nonhomogeneous boundary conditions
(i) Dirichlet condition: 𝑢 ≠ 0
(ii) Neumann condition: 𝑢𝑥 ≠ 0
(iii) Robin condition: 𝑢𝑥 = ±ℎ 𝑢 − 𝑢𝑚 (𝑢𝑚 ≠ 0)

Homogeneous boundary conditions (zero boundary conditions)

(i) Dirichlet condition: 𝑢 = 0
(ii) Neumann condition: 𝑢𝑥 = 0
(iii) Robin condition: 𝑢𝑥 ± ℎ𝑢 = 0
Separation of Variables

105
The HEAT equation
Simple Example

106
Appendix A (3 ODEs) 𝜓(𝜉):
Exponential growth/decay:
𝑑𝜓
= 𝐴𝜓 𝜓 𝜉 = 𝐶𝑒 𝐴𝜉
𝑑𝜉
SHM:
𝑑2𝜓 2𝜓 = 0
+ 𝐴 𝜓 𝜉 = 𝐶1 sin(𝐴𝜉) + 𝐶2 cos(𝐴𝜉)
𝑑𝜉 2
𝜓 𝜉 = 𝐶3 𝑒 𝑖𝐴𝜉 + 𝐶4 𝑒 −𝑖𝐴𝜉
Anti-SHM:
𝑑2𝜓 2𝜓 = 0
− 𝐴 𝜓 𝜉 = 𝐶1 sinh(𝐴𝜉) + 𝐶2 cosh(𝐴𝜉)
𝑑𝜉 2
𝜓 𝜉 = 𝐶3 𝑒 𝐴𝜉 + 𝐶4 𝑒 −𝐴𝜉
 Separation of variables:
As an example, consider the 1D HEAT equation
Seek solutions of the form of a product of a function of 𝑥 and a function of 𝑡.
 Separation of variables:
As an example, consider the 1D HEAT equation
Seek solutions of the form of a product of a function of 𝑥 and a function of 𝑡.

𝑢𝑡 = 𝜎𝑢𝑥𝑥 …(1)

Let 𝑢 𝑥, 𝑡 = 𝑋 𝑥 𝑇(𝑡) and substitute into (1):

𝑋 𝑥 𝑇′(𝑡) = 𝜎𝑋′′ 𝑥 𝑇(𝑡) …(2)

Now we separate the terms containing 𝑡’s and 𝑥 ′ 𝑠 in (2), and we integrate:

𝑋 𝑥 𝑇′ 𝑡 𝜎𝑋 ′′ (𝑥)𝑇 𝑡
=
𝜎𝑋 𝑥 𝑇(𝑡) 𝜎𝑋 𝑥 𝑇(𝑡)
𝑇′ 𝑡 𝑋 ′′ (𝑥)
= …(3)
𝜎𝑇(𝑡) 𝑋 𝑥
 Separation of variables:
As an example, consider the 1D HEAT equation
Seek solutions of the form of a product of a function of 𝑥 and a function of 𝑡.
𝑇′ 𝑡 𝑋 ′′ (𝑥)
= …(3)
𝜎𝑇(𝑡) 𝑋 𝑥
We need a function in terms of only 𝑡 to be equal to a function in terms of only
𝑥. The only way this is possible, is if both functions are equal to the same
constant.

Find the general solution and impose the boundary conditions one-by-one:
• Find the most general solution of the form 𝑢 𝑥, 𝑡 = 𝑋 𝑥 𝑇(𝑡) of the PDE
by considering a positive, a negative, and a separation constant of zero.
• Solve all the constants of integration (but one) by imposing the boundary
conditions one by one.
• Impose the initial condition to obtain the final solution.
 Separation of variables:
As an example, consider the 1D HEAT equation
Seek solutions of the form of a product of a function of 𝑥 and a function of 𝑡.
𝑇′ 𝑡 𝑋 ′′ (𝑥)
= …(3)
𝜎𝑇(𝑡) 𝑋 𝑥
We need a function in terms of only 𝑡 to be equal to a function in terms of only
𝑥. The only way this is possible, is if both functions are equal to the same
constant.
Or first solve the Strum-Liouville problem (bypass solving the general solution):
• Set the RHS of (3) equal to −𝜆 and solve the regular Sturm-Liouville problem
in 𝑋(𝑥) by imposing the boundary conditions. Therefore, solve the
eigenfunctions 𝑋𝑘 𝑥 and the corresponding eigenvalues 𝜆𝑘 by considering
the 3 cases 𝜆 = 0, 𝜆 = 𝛼 2 , 𝜆 = −𝛼 2 .
• Solve 𝑇𝑘 𝑡 by setting the LHS of (3) equal to the eigenvalues −𝜆𝑘 .
• 𝑢 𝑥, 𝑡 = 𝑋 𝑥 𝑇(𝑡)
• Impose the initial condition to obtain the final solution.
Find the general solution
𝑇′ 𝑡 𝑋 ′′ (𝑥)
and solve BVP. = …(3)
𝜎𝑇(𝑡) 𝑋 𝑥
Separation constant zero:
𝑇′ 𝑡 = 0 𝑋 ′′ 𝑥 = 0

𝑇1 𝑡 = 𝐴 𝑋1 𝑥 = 𝐵𝑥 + 𝐶

First solution: 𝑢1 𝑥, 𝑡 = 𝐴 𝐵𝑥 + 𝐶 = 𝐴𝐵 𝑥 + 𝐴𝐶 = 𝐷𝑥 + 𝐸

This is a solution to a linear, homogeneous PDE: from superposition principle any

linear combination of this solution is also a solution.
∞

𝑢1 𝑥, 𝑡 = ෍ 𝐷𝑘 𝑥 + 𝐸𝑘
𝑘=1
= 𝐷1 + 𝐷2 + 𝐷3 + ⋯ 𝑥 + (𝐸1 + 𝐸2 + 𝐸3 + ⋯ )

= 𝐹𝑥 + 𝐺
 𝑇′ 𝑡 𝑋 ′′ (𝑥)
= …(3)
𝜎𝑇(𝑡) 𝑋 𝑥
𝑢1 𝑥, 𝑡 = 𝐹𝑥 + 𝐺

Separation constant positive: 𝛽 2

𝑇 ′ 𝑡 = 𝜎𝛽 2 𝑇(𝑡) 𝑋 ′′ 𝑥 = 𝛽 2 𝑋(𝑥)
2𝑡
𝑋 ′′ 𝑥 − 𝛽 2 𝑋 𝑥 = 0
𝑇2 𝑡 = 𝐻𝑒 𝜎𝛽
𝑋2 𝑥 = 𝐼𝑒 𝛽𝑥 + 𝐽𝑒 −𝛽𝑥
𝑋2 𝑥 = 𝐾 sinh(𝛽𝑥) + 𝑀 cosh(𝛽𝑥)

𝜎𝛽 2𝑡
Second solution: 𝑢2 𝑥, 𝑡 = 𝐻𝑒 𝐾 sinh 𝛽𝑥 + 𝑀 cosh 𝛽𝑥
2
= 𝑒 𝜎𝛽 𝑡 𝑃 sinh 𝛽𝑥 + 𝑄 cosh 𝛽𝑥
This is a solution to a linear, homogeneous PDE: from superposition principle any
linear combination of this solution is also a solution.
∞
2
𝑢2 𝑥, 𝑡 = ෍ 𝑒 𝑡
𝜎𝛽𝑘 𝑃𝑘 sinh 𝛽𝑘 𝑥 + 𝑄𝑘 cosh 𝛽𝑘 𝑥
𝑘=1
 𝑇′ 𝑡 𝑋 ′′ (𝑥)
= …(3)
𝜎𝑇(𝑡) 𝑋 𝑥
𝑢1 𝑥, 𝑡 = 𝐹𝑥 + 𝐺
∞
2
𝑢2 𝑥, 𝑡 = ෍ 𝑒 𝜎𝛽𝑘 𝑡 𝑃𝑘 sinh 𝛽𝑘 𝑥 + 𝑄𝑘 cosh 𝛽𝑘 𝑥
𝑘=1
Separation constant negative: −𝜇2
𝑇 ′ 𝑡 = −𝜎𝜇2 𝑇(𝑡) 𝑋 ′′ 𝑥 = −𝜇2 𝑋(𝑥)
2𝑡
𝑋 ′′ 𝑥 + 𝜇2 𝑋 𝑥 = 0
𝑇3 𝑡 = 𝑅𝑒 −𝜎𝜇
𝑋3 𝑥 = 𝑆𝑒 𝑖𝜇𝑥 + 𝑈𝑒 −𝑖𝜇𝑥
𝑋3 𝑥 = 𝑉 sin(𝜇𝑥) + 𝑊 cos(𝜇𝑥)

−𝜎𝜇 2𝑡
Third solution: 𝑢3 𝑥, 𝑡 = 𝑅𝑒 𝑉 sin 𝜇𝑥 + 𝑊 cos 𝜇𝑥
2
= 𝑒 −𝜎𝜇 𝑡 𝑌 sin 𝜇𝑥 + 𝑍 cos 𝜇𝑥
This is a solution to a linear, homogeneous PDE: from superposition principle any
linear combination of this solution is also a solution.
∞
2
−𝜎𝜇 𝑡
𝑢3 𝑥, 𝑡 = ෍ 𝑒 𝑘 𝑌𝑘 sin 𝜇𝑘 𝑥 + 𝑍𝑘 cos 𝜇𝑘 𝑥
𝑘=1
 𝑢𝑡 = 𝜎𝑢𝑥𝑥 …(1)

𝑢1 𝑥, 𝑡 = 𝐹𝑥 + 𝐺
∞
2
𝑢2 𝑥, 𝑡 = ෍ 𝑒 𝜎𝛽𝑘 𝑡 𝑃𝑘 sinh 𝛽𝑘 𝑥 + 𝑄𝑘 cosh 𝛽𝑘 𝑥
𝑘=1
∞
2
𝑢3 𝑥, 𝑡 = ෍ 𝑒 −𝜎𝜇𝑘𝑡 𝑌𝑘 sin 𝜇𝑘 𝑥 + 𝑍𝑘 cos 𝜇𝑘 𝑥
𝑘=1

Three solution to linear, homogeneous PDE: from superposition principle a

linear combination of the three solutions is also a solution.

Most general solution:

∞ ∞
𝜎𝛽𝑘2 𝑡 2
𝑢 𝑥, 𝑡 = ෍ 𝑒 𝑃𝑘 sinh 𝛽𝑘 𝑥 + 𝑄𝑘 cosh 𝛽𝑘 𝑥 + ෍ 𝑒 −𝜎𝜇𝑘𝑡 𝑌𝑘 sin 𝜇𝑘 𝑥 + 𝑍𝑘 cos 𝜇𝑘 𝑥 + 𝐹𝑥 + 𝐺
𝑘=1 𝑘=1
 𝑢𝑡 = 𝜎𝑢𝑥𝑥 …(1)

Most general solution:

∞ ∞
𝜎𝛽𝑘2 𝑡 2
𝑢 𝑥, 𝑡 = ෍ 𝑒 𝑃𝑘 sinh 𝛽𝑘 𝑥 + 𝑄𝑘 cosh 𝛽𝑘 𝑥 + ෍ 𝑒 −𝜎𝜇𝑘𝑡 𝑌𝑘 sin 𝜇𝑘 𝑥 + 𝑍𝑘 cos 𝜇𝑘 𝑥 + 𝐹𝑥 + 𝐺
𝑘=1 𝑘=1

Impose boundary conditions:

𝑃𝑘 = 𝑄𝑘 = 0 ∀ 𝑘

Most general feasible solution:

∞
2
𝑢 𝑥, 𝑡 = + ෍ 𝑒 −𝜎𝜇𝑘 𝑡 𝑌𝑘 sin 𝜇𝑘 𝑥 + 𝑍𝑘 cos 𝜇𝑘 𝑥 + 𝐹𝑥 + 𝐺
𝑘=1
 𝑢𝑡 = 𝜎𝑢𝑥𝑥 …(1)

Most general feasible solution:

∞
2
𝑢 𝑥, 𝑡 = + ෍ 𝑒 −𝜎𝜇𝑘 𝑡 𝑌𝑘 sin 𝜇𝑘 𝑥 + 𝑍𝑘 cos 𝜇𝑘 𝑥 + 𝐹𝑥 + 𝐺
𝑘=1

∞
2
𝑢 0, 𝑡 = ෍ 𝑒 −𝜎𝜇𝑘 𝑡 𝑍𝑘 + 𝐺
𝑘=1
𝑍𝑘 = 0 ∀ 𝑘, 𝐺=0

New more specific solution:

∞
2
𝑢 𝑥, 𝑡 = ෍ 𝑒 −𝜎𝜇𝑘 𝑡 𝑌𝑘 sin 𝜇𝑘 𝑥 + 𝐹𝑥
𝑘=1

∞
2
𝑢 𝐿, 𝑡 = ෍ 𝑒 −𝜎𝜇𝑘 𝑡 𝑌𝑘 sin 𝜇𝑘 𝐿 + 𝐹𝐿
𝑘=1

𝐹 = 0, sin 𝜇𝑘 𝐿 = 0 ∀ 𝑘
𝑘𝜋
𝜇𝑘 = , 𝑘 = 1,2,3, …
𝐿
 𝑢𝑡 = 𝜎𝑢𝑥𝑥 …(1)

New more specific solution:

∞
𝑘𝜋 2 𝑘𝜋
−𝜎 𝑡
𝑢 𝑥, 𝑡 = ෍𝑒 𝐿 𝑌𝑘 sin 𝑥
𝐿
𝑘=1

∞
𝑘𝜋
𝑢 𝑥, 0 = ෍ 𝑌𝑘 sin 𝑥
𝐿
𝑘=1

Write the initial condition in term of a Fourier series that is completed as an odd function over
𝑥 ∈ [−𝐿, 𝐿], i.e.:
∞
𝑘𝜋𝑥
𝑓 𝑥 = ෍ 𝑏𝑘 sin
𝐿
𝑘=1
where
2 𝐿 𝑘𝜋𝑥
𝑏𝑘 = න 𝑓 𝑥 sin 𝑑𝑥
𝐿 0 𝐿
𝑌𝑘 = 𝑏𝑘 ∀ 𝑘

Final solution: ∞
𝑘𝜋 2 𝑘𝜋
−𝜎 𝐿 𝑡
𝑢 𝑥, 𝑡 = ෍𝑒 𝑏𝑘 sin 𝑥
𝐿
𝑘=1
Solve the Sturm-Liouville
𝑇′ 𝑡 𝑋 ′′ (𝑥)
problem, then solve BVP. = …(3)
𝜎𝑇(𝑡) 𝑋 𝑥
Separation constant: −𝜆

𝑇 ′ 𝑡 + 𝜎𝜆𝑇(𝑡) = 0 𝑋′′ 𝑥 + 𝜆𝑋(𝑥) = 0

𝑢 0, 𝑡 = 𝑋 0 𝑇 𝑡 = 0 ֜ 𝑋 0 =0
𝑢 0, 𝑡 = 𝑋 𝐿 𝑇 𝑡 = 0 ֜ 𝑋 𝐿 =0

Sturm-Liouville problem

Setting 𝜆 = 0, −𝛼 2 and 𝛼 2 yield respectively

𝑋1 𝑥 = 𝐴𝑥 + 𝐵 Trivial solution

𝑋2 𝑥 = 𝐶 sinh(𝛼𝑥) + 𝐷 cosh(𝛼𝑥) Trivial solution

𝑘𝜋
𝑋3 𝑥 = 𝐸 sin(𝛼𝑥) + 𝐹 cos(𝛼𝑥) 𝐹 = 0, 𝛼= , 𝑘 = 1,2,3
𝐿
 𝑇′ 𝑡 𝑋 ′′ (𝑥)
= …(3)
𝜎𝑇(𝑡) 𝑋 𝑥
Separation constant: −𝜆

𝑇 ′ 𝑡 + 𝜎𝜆𝑇(𝑡) = 0 𝑋′′ 𝑥 + 𝜆𝑋(𝑥) = 0

𝑢 0, 𝑡 = 𝑋 0 𝑇 𝑡 = 0 ֜ 𝑋 0 =0
𝑢 0, 𝑡 = 𝑋 𝐿 𝑇 𝑡 = 0 ֜ 𝑋 𝐿 =0

2 Sturm-Liouville problem
′
𝑘𝜋
𝑇 𝑡 +𝜎 𝑇(𝑡) = 0
𝐿
Eigenvalues: 2
𝑘𝜋 2 𝑘𝜋
−𝜎 𝐿 𝑡 2
𝑇𝑘 𝑡 = 𝐹𝑘 𝑒 𝜆=𝛼 =
𝐿

𝑢𝑘 𝑥, 𝑡 = 𝑇𝑘 𝑡 𝑋𝑘 𝑥 Eigenfunctions:
𝑘𝜋 2 𝑘𝜋 𝑘𝜋
𝑢𝑘 𝑥, 𝑡 = 𝑒
−𝜎 𝐿 𝑡
𝑌𝑘 sin 𝑥 𝑋𝑘 (𝑥) = 𝐸𝑘 sin 𝑥
𝐿 𝐿
 𝑢𝑡 = 𝜎𝑢𝑥𝑥 …(1)

New more specific solution:

∞
𝑘𝜋 2 𝑘𝜋
−𝜎 𝑡
𝑢 𝑥, 𝑡 = ෍𝑒 𝐿 𝑌𝑘 sin 𝑥
𝐿
𝑘=1

∞
𝑘𝜋
𝑢 𝑥, 0 = ෍ 𝑌𝑘 sin 𝑥
𝐿
𝑘=1

Write the initial condition in term of a Fourier series that is completed as an odd function over
𝑥 ∈ [−𝐿, 𝐿], i.e.:
∞
𝑘𝜋𝑥
𝑓 𝑥 = ෍ 𝑏𝑘 sin
𝐿
𝑘=1
where
2 𝐿 𝑘𝜋𝑥
𝑏𝑘 = න 𝑓 𝑥 sin 𝑑𝑥
𝐿 0 𝐿
𝑌𝑘 = 𝑏𝑘 ∀ 𝑘

Final solution: ∞
𝑘𝜋 2 𝑘𝜋
−𝜎 𝐿 𝑡
𝑢 𝑥, 𝑡 = ෍𝑒 𝑏𝑘 sin 𝑥
𝐿
𝑘=1
The HEAT equation
Derivation

123
Diffusion equations Δ𝑄 = 𝑐𝑚Δ𝑇
𝑐 (J/(K∙kg)): specific heat
𝑇 (Kelvin) 1
𝑐𝑚
Temperature

The quantity of heat Q in

Δ𝑇 an element of mass m is
Δ𝑄 ❶ 𝑄 = 𝑐𝑚𝑇 + 𝑄0

𝑄 (Joule)
𝑄0
Heat
 Diffusion equations: Mass flow
Fick’s first law:
Flux of a substance in the positive coordinate direction is directly proportional to
negative the concentration gradient
#𝐴 𝑑𝑐𝐴
𝑗= = −𝐷𝐴𝐵
Area × Time 𝑑𝑧

Mass flux 𝑗 (kg. m−2 . s −1 )

How much of A crosses 1 m2 per second

𝑧
 Diffusion equations: Mass flow
Fick’s first law:
Flux of a substance in the positive coordinate direction is directly proportional to
negative the concentration gradient
#𝐴 𝑑𝑐𝐴
𝑗= = −𝐷𝐴𝐵
Area × Time 𝑑𝑧

#𝐴 (kg)

𝑧
 Diffusion equations: Mass flow
Fick’s first law:
Flux of a substance in the positive coordinate direction is directly proportional to
negative the concentration gradient
#𝐴 𝑑𝑐𝐴
𝑗= = −𝐷𝐴𝐵
Area × Time 𝑑𝑧

Diffusion coefficient (or diffusivity) of A into B (m2 /s)

- temperature dependent

𝑧
 Diffusion equations: Mass flow
Fick’s first law:
Flux of a substance in the positive coordinate direction is directly proportional to
negative the concentration gradient
#𝐴 𝑑𝑐𝐴
𝑗= = −𝐷𝐴𝐵
Area × Time 𝑑𝑧

Concentration 𝑐𝐴 (kg/m3 )
Concentration gradient < 0: Δ𝑧 = 𝑧2 − 𝑧1 > 0 and Δ𝑐 = 𝑐2 − 𝑐1 < 0

𝑧
 Diffusion equations
Fick’s first law (mass flow) (mid 1800’s):
Flux of a substance in the positive coordinate direction is directly proportional to
negative the concentration gradient
#𝐴 𝑑𝑐𝐴
𝑗= = −𝐷𝐴𝐵
Area × Time 𝑑𝑧

Heat flow (Fourier’s law of heat conduction) (early 1800’s):

𝜕𝑇
𝑞 = −𝐾
𝜕𝑥
𝑞 (J/(m2 ∙s)): heat flux
𝐾 (W/(m∙K)): material’s thermal conductivity
The rate of heat flow 𝑄𝑡 (Joule per second) through a cross-sectional area A is proportional to
the area of the cross section and the partial derivative with respect to x of the temperature:

𝜕𝑄 𝜕𝑇
❷ = −𝐾𝐴
𝜕𝑡 𝜕𝑥
 The 1D Heat equation
conductivity not constant
• Let the temperature (our dependent variable) be denoted by 𝑢(𝑥, 𝑡).

Assumptions:
• The flow of heat within the rod takes place only in the 𝑥-direction.
• The lateral, or curved, surface of the rod is insulated; that is, no heat
escapes from this surface.
• Rod is homogeneous: Constant density, 𝜌 (kg∙m−3 );
Constant cross-sectional area, 𝐴 (m2 )
• Constant specific heat, 𝑐 (J∙K −1 ∙kg −1 ) Let’s play –
−1 −1
• Thermal conductivity, 𝐾(𝑥) (W∙m ∙K ) differ here from textbook
The 1D Heat equation
conductivity not constant

Take a small element

Δ𝑄1 𝑡 Δ𝑄2 𝑡
𝑥

𝑥 𝑥 + Δ𝑥
The 1D Heat equation
conductivity not constant
“a little bit of…” Δ pertaining
Take a small element to small element

Δ𝑄1 𝑡 Δ𝑄2 𝑡
𝑥

𝑥 𝑥 + Δ𝑥
The 1D Heat equation
conductivity not constant

Take a small element

Δ𝑄1 𝑡 Δ𝑄2 𝑡
𝑥

𝑥 𝑥 + Δ𝑥

“The rate of heat flowing into the element – The rate of heat flowing out of the element
= The rate of change of heat in the element”

❶ 𝑄 = 𝑐𝑚𝑢 + 𝑄0

❷ 𝑄𝑡 = −𝐾𝐴𝑢𝑥
 The 1D Heat equation
conductivity not constant

Δ𝑄1 𝑡 Δ𝑄2 𝑡
𝑥

𝑥 𝑥 + Δ𝑥

“The rate of heat flowing into the element – The rate of heat flowing out of the element”

❷ 𝑄𝑡 = −𝐾𝐴𝑢𝑥

(Δ𝑄1 )𝑡 − Δ𝑄2 𝑡

= −𝐾(𝑥)𝐴𝑢𝑥 𝑥, 𝑡 − [−𝐾(𝑥 + Δ𝑥)𝐴𝑢𝑥 (𝑥 + Δ𝑥, 𝑡)]

= 𝐾(𝑥 + Δ𝑥)𝐴𝑢𝑥 (𝑥 + Δ𝑥, 𝑡) − 𝐾(𝑥)𝐴𝑢𝑥 𝑥, 𝑡

 The 1D Heat equation
conductivity not constant

Δ𝑄1 𝑡 Δ𝑄2 𝑡
𝑥

𝑥 𝑥 + Δ𝑥

OR – note change in diagram, “The rate of heat flowing into the element”

❷ 𝑄𝑡 = −𝐾𝐴𝑢𝑥

(Δ𝑄1 )𝑡 + Δ𝑄2 𝑡

= −𝐾(𝑥)𝐴𝑢𝑥 𝑥, 𝑡 + [− −𝐾 𝑥 + Δ𝑥 𝐴𝑢𝑥 𝑥 + Δ𝑥, 𝑡 ]

= 𝐾(𝑥 + Δ𝑥)𝐴𝑢𝑥 (𝑥 + Δ𝑥, 𝑡) − 𝐾(𝑥)𝐴𝑢𝑥 𝑥, 𝑡

The 1D Heat equation
conductivity not constant

Δ𝑄
𝑥
Δ𝑥
𝑥+
2
“a little bit of…” Δ
“= The rate of change of heat in the element”
pertaining to small
element The quantity of heat Q in
❶ 𝑄 = 𝑐𝑚𝑢 + 𝑄0 an element of mass m

Heat in the element

Δ𝑥 Δ𝑥
Δ𝑄 𝑥 + , 𝑡 = 𝑐 Δ𝑚 𝑢 𝑥 + ,𝑡 + 𝑄0
2 2

The rate of change of heat in the element

Δ𝑥 Δ𝑥
Δ𝑄𝑡 𝑥 + ,𝑡 = 𝑐 Δ𝑚 𝑢𝑡 𝑥 + ,𝑡
2 2
The 1D Heat equation
conductivity not constant
Δ𝑥
𝑥+
2
Δ𝑄 𝑡 Δ𝑄 Δ𝑄 𝑡
𝑥

𝑥 𝑥 + Δ𝑥

“The rate of heat flowing into the element – The rate of heat flowing out of the element
= The rate of change of heat in the element”

Δ𝑥
𝑐 Δ𝑚 𝑢𝑡 𝑥 + , 𝑡 = 𝐾(𝑥 + Δ𝑥)𝐴𝑢𝑥 (𝑥 + Δ𝑥, 𝑡) − 𝐾 𝑥 𝐴𝑢𝑥 𝑥, 𝑡
2
Δ𝑥 1
𝑢𝑡 𝑥 + ,𝑡 = 𝐾(𝑥 + Δ𝑥)𝐴𝑢𝑥 𝑥 + Δ𝑥, 𝑡 − 𝐾 𝑥 𝐴𝑢𝑥 𝑥, 𝑡
2 𝑐 𝜌Δ𝑉
1
= 𝐾 𝑥 + Δ𝑥 𝐴𝑢𝑥 𝑥 + Δ𝑥, 𝑡 − 𝐾(𝑥)𝐴𝑢𝑥 𝑥, 𝑡
𝑐𝜌𝐴Δ𝑥
The 1D Heat equation
conductivity not constant

Δ𝑄 𝑡 Δ𝑄 𝑡
𝑥

𝑥 𝑥 + Δ𝑥
Δ𝑥 1
𝑢𝑡 𝑥+ ,𝑡 = 𝐾 𝑥 + Δ𝑥 𝐴𝑢𝑥 𝑥 + Δ𝑥, 𝑡 − 𝐾 𝑥 𝐴𝑢𝑥 𝑥, 𝑡
2 𝑐𝜌𝐴Δ𝑥

Take the limit where Δ𝑥 → 0:

1 𝐾 𝑥 + Δ𝑥 𝑢𝑥 𝑥 + Δ𝑥, 𝑡 − 𝐾(𝑥) 𝑢𝑥 𝑥, 𝑡
𝑢𝑡 𝑥, 𝑡 = lim
𝑐𝜌 Δ𝑥→0 Δ𝑥
1 𝜕
= 𝐾(𝑥) 𝑢𝑥 𝑥, 𝑡
𝑐𝜌 𝜕𝑥
The 1D Heat equation
conductivity not constant

1 𝜕
𝑢𝑡 𝑥, 𝑡 = 𝐾 𝑥 𝑢𝑥 𝑥, 𝑡
𝑐𝜌 𝜕𝑥
1
= 𝐾 𝑥 𝑢𝑥 𝑥, 𝑡 + 𝐾(𝑥) 𝑢𝑥𝑥 𝑥, 𝑡
𝑐𝜌 𝑥

1
𝑢𝑡 = 𝐾𝑥 𝑢𝑥 + 𝐾𝑢𝑥𝑥
𝑐𝜌
The 1D Heat equation
conductivity not constant

1
𝑢𝑡 = 𝐾𝑥 𝑢𝑥 + 𝐾𝑢𝑥𝑥
𝑐𝜌

constant conductivity

𝐾
𝑢𝑡 = 𝑢𝑥𝑥
𝑐𝜌
The 1D Heat equation 𝜎: thermal diffusivity
constant conductivity
𝐾
𝑢𝑡 = 𝑢𝑥𝑥 = 𝜎𝑢𝑥𝑥
𝑐𝜌

Textbook: modification could be adding a convection term

(Newton’s law of heating/cooling added)

𝑢𝑡 = 𝜎𝑢𝑥𝑥 − ℎ 𝑢 − 𝑢𝑚

The 2D Heat equation

𝑢𝑡 = 𝜎(𝑢𝑥𝑥 + 𝑢𝑦𝑦 )

Note:
• Steady state: 𝑢𝑥𝑥 + 𝑢𝑦𝑦 = 0 → Laplace’s equation
The 1D Heat equation 𝜎: thermal diffusivity
constant conductivity
𝐾
𝑢𝑡 = 𝑢𝑥𝑥 = 𝜎𝑢𝑥𝑥
𝑐𝜌
The 1D Heat equation 𝜎: thermal diffusivity
constant conductivity
𝐾
𝑢𝑡 = 𝑢𝑥𝑥 = 𝜎𝑢𝑥𝑥
𝑐𝜌

Heat diffusion spreads temperature

out and smooths the bumps.