Partial Differential Equations Lecture Notes

Contents
1 Introduction 2

2 Linear and nonlinear waves 2

2.1 Transport equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.1.1 Uniform transport . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.1.2 Nonuniform transport . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 The wave equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2.1 d’Alembert’s formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2.2 External forcing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

3 Fourier series 5
3.1 Evolution equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.2 Fourier series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.3 Convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.3.1 Pointwise convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.3.2 Uniform convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
3.4 Change of scale . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

4 Separation of variables 8
4.1 The heat equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
4.2 Boundary conditions on the wave equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
4.3 Planar Laplace equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
4.3.1 Laplace equation on a disk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
4.3.2 Average and maximum principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
4.4 Classification of PDEs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

5 Generalized functions 12
5.1 Dirac delta ”function” . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
5.1.1 Generalized derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
5.1.2 Fourier series of the delta function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
5.2 Green’s functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
5.2.1 1-dimensional boundary value problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
5.2.2 Line integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
5.2.3 2-dimensional boundary value problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

6 Fourier transforms 17
6.1 Fourier transforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
6.2 Properties of Fourier transforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
6.2.1 Algebraic properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
6.2.2 Derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
6.2.3 Some ”illegal” examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
6.3 Convolutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
6.4 Table of Fourier transforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
6.5 Fundamental solution of the heat equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
6.5.1 Maximum principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

7 Nonlinear PDEs 21
7.1 Nonlinear transport equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
7.2 Nonlinear diffusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
7.3 Dispersion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
 2 2 LINEAR AND NONLINEAR WAVES

1 Introduction
Notation
∂u ∂2u ∂ ∂
ut := uxx := uxy = u
∂t ∂x2 ∂x ∂y

Definition Classical solution

A classical solution of a PDE in n variables is a function u : D ⊂ Rn → R which:
• satisfies the PDE at every point of D
• is sufficiently smooth (continuously differentiable up to the order of the PDE)

Definition Linear differential operator

A linear differential operator on Rn of order m is an expression of the form
X ∂ k1 +...+kn u
L[u] = ak1 ,...,kn (x1 , . . . , xn )
k1 +···+kn <m
∂xk11 · · · ∂xknn

Definition Homogeneous linear PDE

A homogeneous linear PDE is of the form L[u] = 0 where L is a linear differential operator.

Proposition Superposition principle

If u1 , . . . , uk are solutions, then so is
u = c1 u1 + · · · + ck uk
where c1 , . . . , ck are constant.

Definition Inhomogeneous linear PDE

An inhomogeneous linear PDE is of the form L[u] = f where L is a linear diff. operator and f a given function.

Theorem
If up is a solution to L[u] = f , then all solutions of L[u] = f are of the form u = un + up where L[uh ] = 0.

2 Linear and nonlinear waves

2.1 Transport equations
2.1.1 Uniform transport

Proposition Stationary transport

Let D ⊂ R2 and Da := D ∩ (R × {a}).
∂u
If Da is empty or connected for all a ∈ R, and u is a classical solution of ∂t on D, then u only depends on x.

Proposition Uniform transport

If u : R2 → R is a classical solution of
∂u ∂u
+c =0
∂t ∂x
and is defined on all of R2 , then
u(t, x) = v(x − ct)
1
where v is a C function of the characteristic variable ξ = x − ct
2.1 Transport equations 3 2 LINEAR AND NONLINEAR WAVES

Theorem
For a C 1 function f : R → R the initial value problem
∂u ∂u
+c =0 u(0, x) = f (x)
∂t ∂x
has solution u(t, x) = f (x − ct).
This solution u is a travelling wave with velocity c, and is constant along characteristic lines x = ct + k, k ∈ R

Corollary
The initial value problem
∂u ∂u
+c + au = 0 u(0, x) = f (x)
∂t ∂x
has solution u(t, x) = f (x − ct)e−at . Note: u is not constant along characteristic lines.

2.1.2 Nonuniform transport

Definition Characteristic curve

Assume c : R → R is continuous and consider
∂u ∂u
+ c(x) =0 (∗)
∂t ∂x
∂x
The graph of a solution of = c(x) is called a characteristic curve for (∗).
∂t

∂x
Properties of the equation ∂t = c(x)

• Horizontal translations of solution curves are again solution curves

• Nonconstant solutions are strictly monotone functions of t
• Nonconstant solutions can be expressed as both x(t) and t(x).

Proposition Classification of characteristic curves

∂x
Nonconstant solutions of ∂t = c(x) are of the form:
ˆ
1
β(x) := dx = t + k k∈R
c(x)

Characteristic curve: t 7→ (t, x(t)) = (t, β −1 (t + k)) Characteristic variable: ξ = β(x) − t

These solutions can be computed using separation of variables:
ˆ ˆ
dx 1
= c(x) =⇒ dx = 1 dt
dt c(x)

Note: if c(x) = 0, then t 7→ (t, x) is also a characteristic curve.

Proposition
A solution u : R2 → R of (∗) is constant along characteristic curves.

Corollary
A solution u : R2 → R of (∗), is of the form

u(t, x) = v(ξ) = v(β(x) − t)

for some C 1 function v : R → R.

The initial condition u(0, x) = f (x) gives u(t, x) = f (β −1 (ξ)) = f (β −1 (β(x) − t)).

Corollary
If a characteristic curve passes through (t, x) and (0, y), then u(t, x) = f (y).
2.2 The wave equation 4 2 LINEAR AND NONLINEAR WAVES

2.2 The wave equation

Definition Wave operator
The wave operator is the differential operator given by

∂2 ∂2
□= 2
− c2 2 c>0
∂t ∂x
The 1-dimensional wave equation is given by

∂u2 ∂u2
□u = 0 ⇐⇒ 2
= c2 2
∂t ∂x

2.2.1 d’Alembert’s formula

Lemma
If u is C 2 , then
□u = (∂t − c∂x )(∂t + c∂x )u

Theorem Solutions of the wave equation

Every classical solution of □u = 0 can we written as

u(t, x) = p(x − ct) + q(x + ct)

where p and q are C 2 functions.

Theorem d’Alembert’s formula

The solution of the initial value problem
∂u
□u = 0 u(0, x) = f (x) (0, x) = g(x)
∂t
is given by ˆ x+ct
f (x − ct) + f (x + ct) 1
u(t, x) = + g(z) dz
2 2c x−ct

2.2.2 External forcing

Proposition
The solution of the initial value problem:
∂u
□u = F (t, x) u(0, x) = 0 (0, x) = 0
∂t
is given by
ˆ tˆ x+c(t−s)
1
u(t, x) = F (s, y) dy ds
2c 0 x−c(t−s)

Corollary
The solution of the initial value problem:
∂u
□u = F (t, x) u(0, x) = f (x) (0, x) = g(x)
∂t
is given by
ˆ x+ct ˆ tˆ x+c(t−s)
f (x − ct) + f (x + ct) 1 1
u(t, x) = + g(z) dz + F (s, y) dy ds
2 2c x−ct 2c 0 x−c(t−s)
 5 3 FOURIER SERIES

3 Fourier series
3.1 Evolution equations
Definition Linear evolution equation
A linear evolution equation is of the form
∂u
= L[u]
∂t
where the operator L satisfies

L[u + v] = L[u] + L[v] L[c(t)u] = c(t)L[u]

Eigenfunctions and eigenvalues

The educated guess u(t, x) = eλt v(x) gives

∂u
= L[u] ⇐⇒ λv = L[v]
∂t
v is called an eigenfunction corresponding to the eigenvalue λ.

3.2 Fourier series

Definition Inner product
Let X be a linear space over K. A map ⟨·, ·⟩ : X × X → K is called an inner product if:
1. ⟨x, x⟩ ≥ 0
2. ⟨x, x⟩ = 0 ⇐⇒ x = 0
3. ⟨λx + µy, z⟩ = λ⟨x, z⟩ + µ⟨y, z⟩ for all λ, µ ∈ K
4. ⟨x, y⟩ = ⟨y, x⟩

Definition Orthonormal basis

Let X be a Hilbert space. The set {ek : k ∈ N} is called an orthonormal basis for X if
(
1 if i = j
span{ek : k ∈ N} = X ⟨ei , ej ⟩ =
0 if i ̸= j

Definition L2 (−π, π)
L2 (−π, π) is the completion of the inner product space
ˆ π
{f : [−π, π] → C : f is continuous } ⟨f, g⟩ = f (x)g(x) dx
−π

Proposition
The functions {1, sin(x), cos(x), sin(2x), cos(2x), . . .} for k ∈ N form an orthogonal basis for L2 .
The same is true for the functions {eikx : k ∈ Z}.

Theorem Fourier series

Any f ∈ L2 (−π, π) can be written as a Fourier series:
∞ ˆ π ˆ π
a0 X 1 1
f (x) = + (ak cos(kx) + bk sin(kx)) ak = f (x) cos(kx) dx bk = f (x) sin(kx) dx
2 π −π π −π
k=1

ˆ π 1/2
2 2
The Fourier series converges with respect to the L norm: lim |f (x) − sn (x)| dx =0
n→∞ −π
3.3 Convergence 6 3 FOURIER SERIES

Corollary Complex Fourier series

The Fourier series can also be written as:
∞ ˆ π
X 1
f (x) = ck eikx ck = f (x)e−ikx dx
2π −π
k=−∞

Lemma
ak → 0 and bk → 0 as k → ∞.

Lemma Periodic extensions

Let f : (−π, π] → C be any function. Then there exists a function f˜ : R → C such that
• f˜|(−π,π] = f
• f˜ is 2π-periodic

3.3 Convergence
3.3.1 Pointwise convergence

Definition Left-hand and right-hand limits

For f : R → R we say that lim− f (x) = L, denoted L = f (a− ), if:
x→a

for all ε > 0 there exists δ > 0 such that a − δ < x < a =⇒ |f (x) − L| < ε

We say that lim+ f (x) = R, denoted R = f (a+ ), if:

x→a

for all ε > 0 there exists δ > 0 such that a < x < a + δ =⇒ |f (x) − R| < ε

Definition Piecewise continuity

f : [a, b] → R is piecewise continuous if it is defined and continuous except at finitely many points

a ≤ x1 < x2 < · · · < xn ≤ b

and at each xk the left-hand and right-hand limits of f exist.

A function f : R → R is piecewise continuous if it is piecewise continuous on any compact interval.

Definition Piecewise smoothness

f : [a, b] is piecewise C 1 if it is defined, continuous and continuously differentiable except at finitely many points

a ≤ x1 < x2 < · · · < xn ≤ b

and at each xk the left-hand and right-hand limits of f and f ′ exist.

A function f : R → R is piecewise C 1 if it is piecewise C 1 on any compact interval.

Theorem Pointwise convergence of Fourier series

Assume f : R → R is 2π-periodic and piecewise C 1 . Let sn denote the partial sums of the Fourier series.
Then for all fixed x ∈ R we have
f (x+ ) + f (x− )
lim sn (x) =
n→∞ 2
If f is continuous at x, then lim sn (x) = f (x).
n→∞
3.4 Change of scale 7 3 FOURIER SERIES

3.3.2 Uniform convergence

Theorem Uniform convergence of Fourier series

Let f : [−π, π] → R have Fourier coefficients ak and bk , and
∞
X
(|ak | + |bk |) < ∞
k=1

Then the Fourier series of f converges uniformly on R.

The limit f˜ : R → R is continuous and 2π-periodic and has the same Fourier coefficients as f .

Proposition Convergence rate

If f : R → R is 2π-periodic and C n , then there exists M such that |f (n) (x)| ≤ M for all x ∈ R, and

2M 2M
|ak | ≤ |bk | ≤ for all k ∈ N
kn kn

Theorem
Let f : [−π, π] → R have Fourier coefficients ak and bk , and
∞
X
k n (|ak | + |bk |) < ∞
k=1

Then the limit f˜ : R → R is a C n function.

3.4 Change of scale

Proposition Change of scale (real)
Any f ∈ L2 (−ℓ, ℓ) can be written as
∞     
a0 X kπx kπx
f (x) = + ak cos + bk sin
2 ℓ ℓ
k=1

ˆ ℓ   ˆ ℓ  
1 kπx 1 kπx
ak = f (x) cos dx bk = f (x) sin dx
ℓ −ℓ ℓ ℓ −ℓ ℓ

Proposition Change of scale (complex)

Any f ∈ L2 (−ℓ, ℓ) can be written as
∞ ˆ ℓ
X 1
f (x) = ck ekπix/ℓ ck = f (x)e−kπix/ℓ dx
2ℓ −ℓ
k=−∞

Proposition Fourier series of even and odd functions

If f (x) is even, then bk = 0, and so f (x) can be represented by a Fourier cosine series
∞ ˆ π
a0 X 2
f (x) ∼ + ak cos kx ak = f (x) cos(kx) dx
2 π 0
k=1

If f (x) is odd, then ak = 0, and so f (x) can be represented by a Fourier sine series
∞ ˆ π
X 2
f (x) ∼ bk sin kx bk = f (x) sin(kx) dx
π 0
k=1
 8 4 SEPARATION OF VARIABLES

4 Separation of variables
4.1 The heat equation
Derivation of the heat equation
We have the following physical quantities and material properties:
• ε(t, x) = thermal energy densitiy
• w(t, x) = heat flux
• ρ(x) = mass density
• χ(x) = specific heat
• κ(x) = thermal conductivity
Conservation law: ˆ b
∂ε ∂w d
+ = 0 =⇒ ε(t, x) dx = w(t, a) − w(t, b)
∂t ∂x dt a
Constitutive assumption:
ε(t, x) = ρ(x)χ(x)u(t, x)
Fourier’s law of cooling:
∂u
w(t, x) = −κ(x)
∂x

Definition Heat equation

1-dimensional heat equation:  
∂ ∂ ∂u
[ρ(x)χ(x)u(t, x)] + −κ(x)
∂t ∂x ∂x
If ρ, χ, κ are constant, then
∂u ∂2u κ
=γ 2 γ=
∂t ∂x ρχ
We have the following initial condition:
u(0, x) = f (x)
and we choose one or multiple of the following homogeneous boundary conditions:
∂u
Dirichlet condition: u(t, a) = 0 Neumann condition: (t, a) = 0
∂x
∂u
Robin condition: (t, a) + βu(t, a) = 0
∂x
We can also impose periodic boundary conditions:
∂u ∂u
u(t, a) = u(t, b) (t, a) = (t, b)
∂x ∂x

Proposition Homogenisation trick (Dirichlet condition)

Assume u satisfies the heat equation and

u(0, x) = f (x) u(t, a) = ua u(t, b) = ub

Define the function

ub − ua
u∗ (t, x) = ua + (x − a)
b−a
Then v − u − u∗ satisfies the heat equation and

v(0, x) = f (x) − u∗ (0, x) v(t, a) = 0 v(t, b) = 0

4.1 The heat equation 9 4 SEPARATION OF VARIABLES

Proposition Homogenisation trick (Neumann condition)

Assume u satisfies the heat equation and
∂u ∂u
u(0, x) = f (x) (t, x) = ϕa (t, b) = ϕb
∂x ∂x
Define the function
ϕb (x − a)2 − ϕa (x − b2 ) + 2(ϕb − ϕa )γt
u∗ (t, x) =
2(b − a)
Then v − u − u∗ satisfies the heat equation and
∂v ∂v
v(0, x) = f (x) − u∗ (0, x) (t, a) = 0 (t, b) = 0
∂x ∂x

Definition Hyperbolic functions

ex + e−x ex − e−x sinh(x)
cosh(x) = sinh(x) = tanh(x) =
2 2 cosh(x)
d d
sinh x = cosh x cosh x = sinh x
dx dx

Solution method
For a ≤ x ≤ b and t ≥ 0, consider

∂u ∂2u
=γ 2 u(0, x) = f (x) + homogeneous boundary conditions
∂t ∂x
The educated guess u(t, x) = eλt v(x) gives

γv ′′ = λv + homogeneous boundary conditions

Check for nontrivial solutions:

√ √
• λ > 0 =⇒ v(x) = Ae− λ/γx + Be λ/γx = Ã cosh( λ/γx) + B̃ sinh( λ/γx)
p p

• λ = 0 =⇒ v(x) = A + Bx
• λ < 0 =⇒ v(x) = A cos( −λ/γx) + B sin( −λ/γx)
p p

The superposition of all nontrivial solutions gives:

∞ ´b
X f (x)vk (x) dx
u(t, x) = ck eλk t vk (x) ck = ´a b
k=1 a
|vk (x)|2 dx

Proposition Green’s function

Consider for fixed σ ∈ R the boundary value problem
(
γv ′′ (x) − σv(x) = f (x)
homogeneous boundary conditions at x = a, b

Assume that the homogeneous boundary value problem only has the trivial solution:

f = 0 =⇒ v = 0

Then the boundary value problem has a Green’s function G : [a, b] × [a, b] → R such that
ˆ b
v(x) = G(x; ξ)f (ξ) dξ
a

T f (x) = v(x) is a Fredholm operator, therefore if G is continuous, then T is compact.

4.2 Boundary conditions on the wave equation 10 4 SEPARATION OF VARIABLES

Theorem Spectral theorem for compact operators

If T is a compact operator on a Banach space (for example L2 ), then
1. For every ε > 0, the number of eigenvalues λ of T with |λ| > ε is finite.
2. If λ ̸= 0 is an eigenvalue of T , then dim ker(T − λ) < ∞

4.2 Boundary conditions on the wave equation

Recap
∂2u 2
2∂ u ∂u
Wave equation: = c u(0, x) = f (x) (0, x) = g(x)
∂t2 ∂x2 ∂t
ˆ x+ct
f (x − ct) + f (x + ct) 1
d’Alembert’s formula: u(t, x) = + g(z) dz
2 2c x−ct

Proposition Boundary conditions

We apply the following Dirichlet boundary conditions to the wave equation:

u(t, 0) = u(t, ℓ) = 0

This gives the following solutions:

∞       
X kπct kπct kπx
u(t, x) = ak cos + bk sin sin
ℓ ℓ ℓ
k=1

ˆ ℓ   ˆ ℓ  
2 kπx 2 kπx
ak = f (x) sin dx bk = g(x) sin dx
ℓ 0 ℓ kπc 0 ℓ

Periodic extension of d’Alembert’s formula

Consider f and g from d’Alembert’s formula. We have the following odd periodic extension:

f (x)
 if 0 < x < ℓ
f˜(x) = −f (x) if − ℓ < x < 0 f˜(x + 2ℓ) = f˜(x) for all x ∈ R

0 if x ∈ {−ℓ, 0, ℓ}


We construct g̃ in the exact same way as f˜.

If we replace f and g with f˜ and g̃, then d’Alembert’s formula satisfies the boundary conditions.
f˜ and g̃ have the same Fourier expansions as f and g.
Note: in some cases the boundary conditions and initial conditions are incompatible.

4.3 Planar Laplace equations

Definition Laplace operator
The 2-dimensional Laplace operator is given by

∂2 ∂2
∆= 2
+ 2
∂x ∂y

A function u : R2 → R2 is called harmonic if ∆u = 0.

Assuming Ω ⊂ R2 is connected, open and bounded, we have the following boundary value problem:

∆u = 0 on Ω u = h on ∂Ω

Note that we can also replace the Dirichlet boundary condition with a different one.
4.3 Planar Laplace equations 11 4 SEPARATION OF VARIABLES

Laplace equation on a rectangle

Consider the following problem:
∆u = 0 on Ω = (0, a) × (0, b)
u(x, 0) = f (x) u(x, b) = g(x) u(0, y) = h(y) u(a, y) = k(y)
Without loss of generality, we assume g = h = k = 0. We then have the following solution:
∞ ˆ a
X 2 kπ
u(x, y) = ck sin(ωk x) sinh(ωk (b − y)) ck = f (x) sin(ωk x) dx ωk =
a sinh(ωk b) 0 a
k=1

4.3.1 Laplace equation on a disk

Laplace equation on a disk

Consider the following problem:

∆u = 0 on x2 + y 2 < 1 u(x, y) = h(x, y) on x2 + y 2 = 1

We replace every occurrence of x and y by r cos θ and r sin θ respectively. This gives the following polar equation:

∂ 2 u 1 ∂u 1 ∂2u
∆u = 2
+ + 2 2 =0 u(1, θ) = h(θ)
∂r r ∂r r ∂ θ
Superposition of solutions without singularities at r = 0 gives:
ˆ π ∞
" #
1 1 X k
u(r, θ) = h(ϕ) + r cos(k(θ − ϕ)) dϕ
π −π 2
k=1

Theorem Poisson’s formula

For the Laplace equation on a disk, we have the following solutions:
ˆ π
1 1 − r2
u(r, θ) = 2
h(ϕ) dϕ
2π −π 1 + r − 2r cos(θ − ϕ)

In particular, ˆ π
1
u(0, θ) = h(ϕ) dϕ = average value of h
2π −π

4.3.2 Average and maximum principle

Theorem Average principle

If u is harmonic inside the disk

D = {(x, y) ∈ R2 : (x − a)2 + (y − b)2 ≤ ρ2 }

then we have ˛ ˆ π
1 1
u(a, b) = u ds = u(a + ρ cos θ, b + ρ sin θ) dθ
2πρ ∂D 2π −π

Theorem Maximum principle

Assume that
• Ω is bounded, open and connected
• u is harmonic on Ω and continuous on ∂Ω
• for all (x, y) ∈ Ω we have u(x, y) ≤ M If u(x0 , y0 ) = M for some (x0 , y0 ) ∈ Ω,
then u is constant on Ω, and u attains its maximum value on ∂Ω.
4.4 Classification of PDEs 12 5 GENERALIZED FUNCTIONS

Corollary
Poisson’s equation with Dirichlet conditions

−∆u = f on Ω u = h on ∂Ω

has a unique solution.

4.4 Classification of PDEs

Theorem
The solutions of
Ax2 + Bxy + Cy 2 + Dx + Ey + F = 0
trace a curve whose type is determined by the discriminant

∆ = B 2 − 4AC

The following classification applies:

• ∆ > 0: hyperbolic
• ∆ = 0: parabolic
• ∆ < 0: elliptic

Definition Classification of PDEs

For the linear, 2nd-order PDE
Autt + Butx + Cuxx + Dut + Eux + F = 0
we define the discriminant
∆(t, x) = B 2 − 4AC
At a point (t, x), the PDE is called
• a hyperbolic PDE if ∆(t, x) > 0
• a parabolic PDE if ∆(t, x) = 0
• a elliptic PDE if ∆(t, x) < 0
• a singular PDE if A = B = C = 0

5 Generalized functions
5.1 Dirac delta ”function”
Lemma
Define: (
1
n if |x| ≤ 2n n 2 2
rn (x) = fn (x) = √ e−n x
0 otherwise π
If u is continuous and a < 0 < b then ˆ b
lim rn (x)u(x) dx = u(0)
n→∞ a
If u is continuous and bounded, then for all ξ ∈ R we have
ˆ ∞
lim fn (x − ξ)u(x) dx = u(ξ)
n→∞ −∞
5.1 Dirac delta ”function” 13 5 GENERALIZED FUNCTIONS

Definition Dirac delta function

The Dirac delta function is defined by
ˆ b
δξ (x)u(x) dx = u(ξ)
a
whenever u is continuous and a < ξ < b
Note: this is not actually a function, but it is a linear functional u 7→ u(ξ) on a suitable space.

Lemma
Any continuous function f is a ”superposition” of delta functions:
ˆ ∞
f (x) = f (ξ)δ(x − ξ)dξ
−∞

5.1.1 Generalized derivatives

Definition Unit step function
ˆ x (  
0 if x < 0 1 1 dσ
σ(x) = δ0 (t) dt = σ(0) = lim− σ(x) + lim+ σ(x) = = δ0
−1 1 if x > 0 x x→0 x→0 2 dx

Definition Ramp function

ˆ x
(
0 if x < 0 dρ
ρ(x) = σ(t) dt = =σ
−1 1 if x ≥ 0 dx

Derivatives of the absolute value

(
′ −1 if x < 0
f (x) = |x| =⇒ f (x) = =⇒ f ′′ (x) = 2δ0
1 if x > 0

General rule for generalized derivatives

If f has a discontinuity at x = ξ such that

r := f (ξ + ) − f (ξ − ) ̸= 0

then include the following term in the expression for f ′ :

rδ(x − ξ)

5.1.2 Fourier series of the delta function

Fourier coefficients of the delta function
ˆ ˆ
1 π 1 1 1 π
1
ak = δ(x) cos(kx) dx = cos(0) = bk = δ(x) sin(kx) dx = sin(0) = 0
π −π π π π −π π

Definition Dirac comb

The Dirac comb is the 2π-periodic extension of δ:
∞ ∞
!
X 1 1X 1 X
δ̃(x) = δ(x − 2kπ) = + cos(kx) = 1+2 cos(kx)
2π π 2π
k∈Z k=1 k=1
5.2 Green’s functions 14 5 GENERALIZED FUNCTIONS

5.2 Green’s functions

General solution of a 2nd-order ODE
Consider the following ODE:
p(x)u′′ + q(x)u′ + r(x)u = f (x)
The general form of the solution is
u(x) = c1 u1 (x) + c2 u2 (x) + up (x)
where
• u1 , u2 are linearly independent solutions of the homogeneous solution
• up is a particular solution of the inhomogeneous equation
• c1 , c2 are constants

Variation of parameters
For a particular solution, try the ansatz up = c1 u1 + c2 u2 , where c1 , c2 depend on x.
By the product rule, we have
u′p = c1 u′1 + c2 u′2 + c′1 u1 + c′2 u2
We then take a second ansatz: c′1 u1 + c′2 u2 = 0. Then we have:

u′p = c1 u′1 + c2 u′2 u′′p = c1 u′′1 + c2 u′′2 + c′1 u′1 + c′2 u′2

Since u1 and c2 are solutions of the homogeneous equation, we get the following expression for f

u1 u2 c′1
    
′′ ′ ′ ′ ′ ′ ′ ′ ′ ′ f 0
f = pup + qup + rup = p(c1 u1 + c2 u2 ) =⇒ c1 u1 + c2 u2 = =⇒ =
p u′1 u′2 c′2 f /p

Define the Wronskian determinant w = u1 u′2 − u′1 u2 . Then we have

 ′
1 u′2 −u2
    
c1 0 −u2 /W p
= =
c′2 W −u′1 u1 f /p u1 f /W p

c1 and c2 can then be found by integration.

Proposition
A particular solution to
p(x)u′′ + q(x)u′ + r(x)u = f (x)
is given by ˆ ˆ
u2 (x)f (x) u1 (x)f (x)
up (x) = −u1 (x) dx + u2 (x) dx
W (x)p(x) W (x)p(x)
where u1 , u2 are solutions to the homogeneous case.

Proposition Green’s function for u′′ (x) = f (x)

Consider the boundary value problem:

u′′ (x) = f (x) u(0) = u(1) = 0

This has the following solution:

ˆ 1
(
(x − 1)ξ if ξ ≥ x
G(x, ξ)f (ξ) dξ G(x, ξ) =
0 (ξ − 1)x if ξ ≥ x
5.2 Green’s functions 15 5 GENERALIZED FUNCTIONS

Properties of Green’s function

Consider the following Green’s function and its derivatives:
( (
(x − 1)ξ if ξ ≥ x ∂ ξ if ξ < x ∂2
G(x, ξ) = G(x, ξ) = G(x, ξ) = δ(x − ξ)
(ξ − 1)x if ξ ≥ x ∂x ξ−1 if ξ > x ∂x2

This Green’s function satisfies the following properties:

• G solves the homogeneous equation u′′ (x) = 0 when x ̸= ξ.
• G solves the homogeneous boundary conditions u(0) = u(1) = 0.
• G is continuous in (x, ξ).
• ∂G
∂x has a jump discontinuity at x = ξ.
Note: this Green’s function can easily be derived by integrating its second derivative twice.

5.2.1 1-dimensional boundary value problems

Green’s function for a 1-dimensional boundary value problem

Define the operator
L[u] = pu′′ + qu′ + ru
Goal: find a Green’s function G(x, ξ) such that
( ˆ b
L[u] = f
=⇒ u(x) = G(x, ξ)f (ξ) dξ
u(a) = u(b) = 0 a

We first find linearly independent solutions u1 , u2 such that

L[u1 ] = L[u2 ] = 0 u1 (a) = u2 (b) = 0

Then we have the following ansatz: (

c1 u1 (x) if x ≤ ξ
G(x, ξ) =
c2 u2 (x) if x ≥ ξ
We find c1 and c2 by requiring
• G is continuous at x = ξ
∂G 1
• has a jump discontinuity of magnitude at x = ξ
∂x p(ξ)

5.2.2 Line integrals

Definition Line integral

Assume t 7→ (x(t), y(t)) with t ∈ [a, b] parametrizes a curve C.
The line integral of a scalar function f is
ˆ ˆ b p
f ds := f (x(t), y(t)) x′ (t)2 + y ′ (t)2 dt
C a

The line integral of a vector field is

ˆ ˆ b
P dx + Q dy := P (x(t), y(t))x′ (t) + Q(x(t), y(t))y ′ (t) dt
C a

Theorem Green’s theorem

ˆ ¨
∂Q ∂P
P dx + Q dy = − dx dy
∂Ω Ω ∂x ∂y
Note: the parametrization of ∂Ω must satisfy the left-hand rule.
5.2 Green’s functions 16 5 GENERALIZED FUNCTIONS

Proposition Green’s identities

Let n denote the outward unit normal vector along ∂Ω.
¨ ˆ
∂v
Green’s first identity: u∆v + ∇u · ∇v dx dy = u ds
Ω ∂Ω ∂n
¨ ˆ
∂v ∂u
Green’s second identity: u∆v − v∆u dx dy = u −v ds
Ω ∂Ω ∂n ∂n

5.2.3 2-dimensional boundary value problems

Laplace operator (recap)

∂2 ∂2
∆= + f harmonic ⇐⇒ ∆f = 0
∂x2 ∂y 2

Definition G0
1 1
G0 (x, y, ξ, η) = −
log ∥(x, y) − (ξ, η)∥ = − log[(x − ξ)2 + (y − η)2 ]
2π 4π
From now on, consider (x, y) fixed and (ξ, η) variable.

Lemma

1. G0 is harmonic on R2 \ {(x, y)}

2. If Cr is a circle of radius r around (x, y), then

1 ∂G0 1
G0 (x, y, ξ, η) = − log r (x, y, ξ, η) = − for all (ξ, η) ∈ Cr
2π ∂n 2πr

Theorem Representation formula

Let n denote the outward unit normal vector along ∂Ω. If u is harmonic in Ω, then for (x, y) ∈ Ω we have
ˆ
∂u ∂G0
u(x, y) = G0 −u
∂Ω ∂n ∂n

Definition G
G = G0 + z ∆z = 0 on Ω z = −G0 on ∂Ω

Corollary
If u is a solution of
∆u = 0 on Ω u = h on ∂Ω
Then we have the representation formula
ˆ
∂G
u(x, y) = − h ds
∂Ω ∂n

Theorem
If u is a solution of
−∆u = f on Ω u = 0 on ∂Ω
Then we have the representation formula
¨
u(x, y) = f (ξ, η)G(x, y, ξ, η) dξ dη
Ω
 17 6 FOURIER TRANSFORMS

Finding G using the method of images

To each point (ξ, η) ∈ Ω associate an image point (ξ ′ , η ′ ) ∈ R2 \ Ω.

The following ansatz guarantees ∆z = 0 on Ω:
a b
z(x, y, ξ, η) = log ∥(x − y) − (ξ ′ , η ′ )∥ +
2π 2π
Then we determine a, b and (ξ ′ , η ′ ) such that

z(x, y, ξ, η) = −G0 (x, y, ξ, η) for all (x, y) ∈ ∂Ω

(for geometric examples, see the slides of Lecture 09.)

6 Fourier transforms
6.1 Fourier transforms
Definition L1
 ˆ ∞ 
1
L (R) = f :R→C: |f (x)| dx < ∞
−∞

Theorem Fourier integral representation

If f ∈ L1 (R) is piecewise C 1 , then
ˆ ∞
f (x+ ) + f (x− )
= [A(k) cos(kx) + B(k) sin(kx)] dk
2 0
ˆ ˆ
1 ∞ 1 ∞
A(k) = f (y) cos(ky) dy B(x) = f (y) sin(ky) dy
π −∞ π −∞

Definition Fourier transform

The Fourier transform of f ∈ L1 (R) is given by
ˆ ∞
1
fb(k) = F[f (x)] = √ f (x)e−ikx dx
2π −∞

The inverse Fourier transform is given by

ˆ ∞
1
F −1 [fb(x)] = √ fb(k)eikx dk
2π −∞

Theorem
If f ∈ L1 is piecewise C 1 , then
f (x+ ) + f (x− )
F −1 [F[f (x)]] =
2

6.2 Properties of Fourier transforms

6.2.1 Algebraic properties

Lemma
The Fourier transform is linear.

Lemma
If f is real and even, then fb is real and even.
If f is real and odd, then fb is purely imaginary and odd.
6.2 Properties of Fourier transforms 18 6 FOURIER TRANSFORMS

Lemma Fourier transform of translations

g(x) = f (x − ξ) =⇒ gb(k) = e−ikξ fb(k) g(x) = eiξx f (x) =⇒ gb(k) = fb(k − ξ)

Lemma Fourier transform of dilations

 
1 b k
g(x) = f (cx) =⇒ gb(k) = f
|c| c

Lemma Symmetry principle

F[f (x)] = fb(k) =⇒ F[fb(x)] = f (−k)

6.2.2 Derivatives
Lemma
If f ∈ L1 (R) is C 1 and lim f (x) = 0, then
|x|→∞
(f
d ′ )(k) = ik fb(k)

Lemma
If f, xf ∈ L1 (R), then
F[xf (x)] = i(fb)′ (k)

Note
These two properties can be used to solve ODEs and PDEs, namely by applying a Fourier transform to the entire
equation. Examples can be found at the end of the Lecture 10 slides.

6.2.3 Some ”illegal” examples

Fourier transform of constant functions

√ √
F[1] = 2πδ(k) F[e−ixξ ] = 2πδ(k + ξ)

Sign function

1
 if x > 0 r
21
sign(x) = 0 if x = 0 F[sign(x)] = −i
 πk
−1 if x < 0


Step function

1
 if x > 0 r
π i
1
σ(x) = if x = 0 F[σ(x)] = δ(k) − √
2 2 2πk
0 if x < 0


Inverse Fourier transform of rational functions

 ( √
−i 2πe(−b+ai)x (1 − σ(x)) if b < 0
  r 
−1 1 π iax −1 1
F =i e sign(x) if a ∈ R F = √
k−a 2 k − (a + bi) i 2πe(−b+ai)x σ(x) if b > 0
6.3 Convolutions 19 6 FOURIER TRANSFORMS

6.3 Convolutions
Definition Convolution product
ˆ ∞
(f ∗ g)(x) = f (x − ξ)g(ξ) dξ
−∞

Theorem Properties of convolutions

For functions f, g and constants a, b we have
1. f ∗ g = g ∗ f
2. f ∗ (g ∗ h) = (f ∗ g) ∗ h
3. f ∗ (ag + bh) = a(f ∗ g) + b(f ∗ h)
4. f ∗ 0 = 0
5. f ∗ δ = f
√
6. f = f ∗ g =⇒ b
h= 2π fb · gb
7. f = f · g =⇒ b
h= √1 fb ∗ gb
2π

Note: properties 6 and 7 are very useful for solving PDEs.

6.4 Table of Fourier transforms

Table of Fourier transforms

f (x) fˆ(k) f (x) fˆ(k) f (x) fˆ(k)

√ 1
e−ax σ(x) √ eikd/c ˆ k
 
1 2π δ(k)
2π (a + ik) f (cx + d) f
1 1 |c| c
δ(x) √ ax
e (1 − σ(x)) √
2π fˆ(x) f (−k)
r
π i r2π (a − ik)
σ(x) δ(k) − √ 2 a fˆ(−x) f (k)
2 2π k e−a|x|
π k22 + a2 ′
ik fˆ(k)
r
2 1 e−k /(4a) f (x)
sign x −i e−ax
2
√
r π k 2ar xf (x) i fˆ ′ (k)
2 sin ak π 3/2
π e−|k| √
σ(x + a) − σ(x − a) −1
tan x √ δ(k) − i f ∗ g(x) 2π fˆ(k) ĝ(k)
π k 2 2 k

6.5 Fundamental solution of the heat equation

Nonhomogeneous initial value problem
Consider the following problem:

∂u ∂2u
= γ 2 + h(t, x) t>0 −∞<x<∞ u(0, x) = f (x)
∂t ∂x
We can split it into two problems as follows:

∂v ∂2v ∂w ∂2w
=γ 2 v(0, x) = f (x) = γ 2 + h(t, x) w(0, x) = 0
∂t ∂x ∂t ∂x
and then obtain the solutions by doing Fourier transforms.
6.5 Fundamental solution of the heat equation 20 6 FOURIER TRANSFORMS

Unforced equation
In the case f (x) = δ(x − ξ) we have the following fundamental solution for the v problem:
1 2
F (t, x; ξ) = √ e−(x−ξ) /4γt
2 πγt

Let F0 (t, x) = F (t, x; 0). For general f , we have the following solution

v(t, x) = (F0 ∗ f )(x, t)

Forced equation
In the case h(t, x) = δ(t − τ )δ(x − ξ) we have the following fundamental solution for the w problem:

σ(t − τ ) (x−ξ)2
G(t, x; τ, ξ) = p e− 4γ(t−τ ) = σ(t − τ )F (t − τ, x; ξ)
2 πγ(t − τ )

For general h(x, t), we have the following solution

ˆ tˆ ∞
h(τ, ξ) (x−ξ)2
w(t, x) = p e− 4γ(t−τ ) dξ dτ
0 −∞ 2 πγ(t − τ )

General solution of the forced heat equation

ˆ ∞ ˆ tˆ ∞
1 2 h(τ, ξ) 2
u(t, x) = √ e−(x−ξ) /4γt f (ξ) dξ + p e−(x−ξ) /4γ(t−τ ) dξ dτ
2 γπt −∞ 0 −∞ 2 γπ(t − τ )

6.5.1 Maximum principle

Definition Rectangle with (partial) boundary

R = {(t, x) ∈ R2 : 0 < t < c and a < x < b}

R1 = {(t, x) ∈ R2 : t = 0 and a < x < b} R2 = {(t, x) ∈ R2 : 0 < t < c and x = a}
R3 = {(t, x) ∈ R2 : 0 < t < c and x = b} B = R1 ∪ R2 ∪ R3

Theorem Maximum principle

Assume γ > 0, h(t, x) ≤ 0 for all (t, x) ∈ R, and u satisfies the following PDE on R:

∂u ∂2u
= γ 2 + h(t, x)
∂t ∂x
Then the maximum of u on R is attained on the set B.

Corollary
Assume that u satisfies
∂u ∂2u
=γ 2
∂t ∂x
and define
m = min{u(t, x) : (t, x) ∈ B} M = max{u(t, x) : (t, x) ∈ B}
Then for all (t, x) ∈ R, we have
m ≤ u(t, x) ≤ M

Theorem
The problem
∂u ∂2u
= γ 2 + h(t, x)
∂t ∂x
with initial condition and Dirichlet boundary conditions at x = a, b has at most 1 solution.
 21 7 NONLINEAR PDES

7 Nonlinear PDEs
7.1 Nonlinear transport equation
Note
Nonlinear PDEs do not satisfy the superposition principle.

Nonlinear transport equation

∂u ∂u
+u u(0, x) = f (x)
∂t ∂x

Definition Characteristic curve (nonlinear)

Characteristic curves of the nonlinear transport equation are graphs of solutions of
dx
= u(t, x(t))
dt

Proposition
A solution u of the nonlinear transport equation is constant along characteristic curves.

Corollary Implicit expression for u

u(t, f (y)t + y) = f (y)

7.2 Nonlinear diffusion

Burgers’ equation
∂u ∂u ∂2u
+u =γ 2 γ>0
∂t ∂x ∂x

Solution of Burgers’ equation

The ansatz u(t, x) = v(ξ) = v(x − ct) where v is bounded leads to the following solution:

ae(b−a)(x−ct−δ)/(2γ) + b
u(t, x) =
e(b−a)(x−ct−δ)/(2γ) + 1

Proposition Hopf-Cole transformation

All solutions to Burgers’ equation can be obtained via the transformation

∂ vx (t, x)
u(t, x) = [−2γ log v(t, x)] = −2γ
∂x v(t, x)

where v is a positive solution to the heat equation.

Proposition Solution of Burgers’ equation with initial condition

The solution of Burgers’ equation with initial condition u(0, x) = f (x) is given by
ˆ ∞ ˆ ξ
vex (t, x) (x − ξ)2 1
u(t, x) = −2γ ṽ(t, x) = e−H(t,x;ξ) dξ H(t, x; ξ) = + f (η) dη
ve(t, x) −∞ 4γt 2γ 0
7.3 Dispersion 22 7 NONLINEAR PDES

7.3 Dispersion
Dispersive wave equation with boundary conditions
∂u ∂ 3 u
+ =0 t>0 −π <x<π u(0, x) = f (x)
∂t ∂x3
u(t, −π) = u(t, π) ux (t, −π) = ux (t, π) uxx (t, −π) = uxx (t, π)
We have the following Fourier series solution:
∞ ˆ π
X
i(kx+k3 t) 1
u(t, x) = ck (0)e ck (0) = f (x)e−ikx dx
2π −π
k=−∞

Definition Airy function of the first kind

ˆ ∞
1 1
Ai(x) = cos(kx + k 3 ) dk
π 0 3

Linearized Korteweg-de Vries equation

∂u ∂ 3 u
+ =0 u(0, x) = f (x) = δξ (x)
∂t ∂x3
By taking Fourier transforms we get a fundamental solution:
 
h 3
i 1 x−ξ
F (t, x; ξ) = F −1 δbξ (k)eik t = √ 3
Ai √3
3t 3t

The solution for arbitrary f (x) is

ˆ ∞  
1 x−ξ
u(t, x) = (F0 ∗ f )(t, x) = √
3
Ai √3
f (ξ) dξ
3t −∞ 3t

Korteweg-de Vries equation

∂u ∂ 3 u ∂u
+ 3
+u =0 t>0 −∞<x<∞
∂t ∂x ∂x
As an ansatz, we take a traveling wave with some conditions on the asymptotic behavior.

u(t, x) = v(ξ) = v(x − ct) lim v(ξ) = 0 lim v ′ (ξ) = 0 lim v ′′ (ξ) = 0
ξ→±∞ ξ→±∞ ξ→±∞

This leads to the following ODE:  

′ 2 1 2
(v ) = v c − v
3
A solution is then given by

1√
 
2
u(t, x) = 3c sech2 c(x − ct) + δ sech(y) =
2 e−y + ey
Index
1-dimensional heat equation, 8 image point, 17
Implicit expression for u, 21
Average principle, 11 inhomogeneous linear PDE, 2
initial value problem, 3
Boundary conditions, 10 inner product, 5
inverse Fourier transform, 17
Change of scale (complex), 7
Change of scale (real), 7 Korteweg-de Vries equation, 22
characteristic curve, 3
Characteristic curves, 21 Laplace operator, 10
characteristic lines, 3 Left-hand and right-hand limits, 6
characteristic variable, 2 line integral, 15
classical solution, 2 linear differential operator, 2
Classification of characteristic curves, 3 linear evolution equation, 5
Classification of PDEs, 12 Linearized Korteweg-de Vries equation, 22
Complex Fourier series, 6
Conservation law, 8 Maximum principle, 11, 20
Constitutive assumption, 8 Method of images, 17
Convergence rate, 7
Neumann condition, 8
d’Alembert’s formula, 4, 10
Dirac comb, 13 odd periodic extension, 10
Dirac delta function, 13 orthonormal basis, 5
Dirichlet condition, 8
parabolic PDE, 12
discriminant, 12
periodic boundary conditions, 8
Dispersive wave equation with boundary conditions,
Periodic extensions, 6
22
piecewise C 1 , 6
eigenfunction, 5 piecewise continuous, 6
eigenvalue, 5 Pointwise convergence of Fourier series, 6
elliptic PDE, 12 Poisson’s equation, 12
Poisson’s formula, 11
Fourier cosine series, 7 Properties of convolutions, 19
Fourier integral representation, 17
Fourier series, 5 Ramp function, 13
Fourier series of even and odd functions, 7 Representation formula, 16
Fourier sine series, 7 Robin condition, 8
Fourier transform, 17
singular PDE, 12
Fourier transform of dilations, 18
Solution of Burgers’ equation with initial condition,
Fourier transform of translations, 18
21
Fourier’s law of cooling, 8
Solutions of the wave equation, 4
fundamental solution, 20
Spectral theorem for compact operators, 10
Stationary transport, 2
General solution of a 2nd-order ODE, 14
Superposition principle, 2
Green’s first identity, 16
Symmetry principle, 18
Green’s function, 9, 15
Green’s function for u′′ (x) = f (x), 14 Table of Fourier transforms, 19
Green’s identities, 16
Green’s second identity, 16 Uniform convergence of Fourier series, 7
Green’s theorem, 15 Uniform transport, 2
Unit step function, 13
harmonic, 10
homogeneous boundary conditions, 8 Variation of parameters, 14
homogeneous linear PDE, 2
Homogenisation trick (Dirichlet condition), 8 Wave equation, 10
Homogenisation trick (Neumann condition), 9 wave equation, 4
Hopf-Cole transformation, 21 wave operator, 4
hyperbolic PDE, 12 Wronskian determinant, 14

23