Systems modeling

Partial differential equations (PDEs)

Master of Interdisciplinary & Innovative Engineering

2023

Departament de mecànica de fluids

Escola d’Enginyeria de Barcelona Est (EEBE)
Universitat Politècnica de Catalunya (UPC)

This work is licensed under the Creative Commons

Attribution-ShareAlike 4.0 International License.
Fields

In gneral, systems are described by fields (functions), i.e. a scalar, vector or

tensor at each point in space and time:

• Scalar, e.g. ρ(x, t) = ρ(x1 , x2 , x3 , t)

• Vector, e.g. v(x, t) = v(x1 , x2 , x3 , t)
• Tensor, e.g. σ(x, t) = σ(x1 , x2 , x3 , t)

Vector calculus is treated in Marsden and Tromba [2012].

Derivatives can be taken in any direction and also in time and are denoted
by ∂, e.g.
∂u ∂u
= ∂i u or = ∂t u
∂xi ∂t
The spatial derivatives of a scalar form a vector, the gradient
∂u
∇u = ei
∂xi

1/21
Gradient

The variation of u along a curve x(s) is obtained as

du ∂u ∂xi
= = v · ∇u
ds ∂xi ∂s
where
dx
v=
ds
A level set surface of a field, is the set of points where u(x1 , x2 , x3 ) = c for a
given constant c.

For any curve x(s) contained in that surface

du
=0
ds
and therefore v · ∇u = 0, where v = dx/ds is tangent to the curve.

Therefore, the gradient is orthogonal to the level set surface.

2/21
Gradient

This permits to find the normal to the surface u(x1 , x2 , x3 ) = c as

∇u
n̂ =
k∇uk

The gradient of a vector is a second order tensor, e.g. l = ∇v, whose

components are
∂vi
lij =
∂xj

Example
A force of magnitude F acts in the x direction at the point (−1, 1) on the
surface of a circle described by

x21 + x22 = 2

Determine the normal and tangential forces acting on the circle.

3/21
Divergence and curl

• Divergence: given a vector field v, ∇ · v = ∂i v i is a salar.

• A vector field v is solenoidal when ∇ · v = 0.
• Curl: given a a vector v, ∇ × v is a vector whose compoements are
ijk ∂i vk .
• A vector field v is irrotational when ∇ × v = 0.
• The gradient of scalar field φ is irrotational, i.e.

∇ × ∇φ = 0

• Conversely, if ∇ × v = 0 in a simply connected region there exists φ

such that v = ∇φ

4/21
Integral theorems

The multidimensional analogous of the fundamental theorem of calculus is

the Green theorem: if f ∈ C 1 (it has continuous partial derivatives)
Z Z
∂f
f ni dS = dV
S V ∂xi

See Malvern [1969] section 5.1 for a proof in 2D.

A direct consequence is the Gauss theorem (also called divergence
theorem): for a C 1 a vector field v
Z Z
v · ndS = ∇ · vdV
S V

Another consequence is the Stokes theorem (also called circulation

theorem): for a C 1 a vector field v
Z Z
n · (∇ × v)dS = t · vds
S C
where S is an orientable surface defined by a closed curve C.
5/21
Integral theorems

Example
Consider the field v = (x, y). Show that it is irrotational and verify Gauss
and Stokes theorems.

Example
Consider the field v = (−y, x). Show that it is solenodial and verify Gauss
and Stokes theorems.

6/21
Partial differential equations (PDEs)

Definition
A differential equation is an equation relating an unknown function and one
or more of its derivatives.

Two main families:

• Ordinary differential equations (ODEs): the unknown function is a

process, a function of one independent variable, e.g. u(t) (the notation
y(x) is common in math books). ⇐= chapter 2.

• Partial differential equaations (PDEs) Strauss [2008]; Evans [1998]: the

unknown is a field, a function of several independent variables, e.g.
u(x1 , x2 , x3 , t).

Definition
The order of differential equation is the maximum derivative order present.

7/21
Examples

1. Transport
∂u ∂u ∂u
+a +b =0
∂t ∂x ∂y
2. Shock
∂u ∂u
+u =0
∂t ∂x
3. Diffusion
∂u ∂ 2 u ∂ 2 u ∂ 2 u
+ + 2 + 2 =0
∂t ∂x2 ∂y ∂z
4. Wave
∂2u ∂2u ∂2u
− − 2 =0
∂t2 ∂x2 ∂y
5. Vibrations
∂2u ∂4u
+ =0
∂t2 ∂x4
Example
Find the order and the number of independent variables of PDEs 1 to 5.

8/21
Linear PDEs

Let us write PDEs as Lu = g where g is a given function and L the

differential operator. When g = 0 the equation is homogeneous.
In previous example 3
∂ ∂2 ∂2
L= + +
∂t ∂x2 ∂y 2
and g = 0.
Definition
A differential operator L is linear if for any functions u, v and a ∈ R,

L(u + v) = Lu + Lv and L(au) = aL(u)

Example
Find the order and the number of independent variables of PDEs 1 to 6.

9/21
Linear PDEs

Superposition principle
If u1 and u2 are solutions of the (homogeneous) equation Lu = 0 then
c1 u1 + c2 u2 with c1 , c2 ∈ R are also solutions. Solutions of linear PDEs can
be linearly combined (superposed).

Example
As u1 (x, y) = x and u2 (x, u) = y are solutions of the difussion equation,
u(x, y) = ax + by is also a solution.

Observe that if u1 is a solution of the (non-homogeneous) equation Lu = g

and u2 is a solution of the (homogeneous) equation Lu = 0, then u1 + u2 is
a solution of the (non-homogeneous) equation Lu = g. Why?

10/21
First order linear PDEs

The general first order linear PDE with constant coefficients and two
independent variables is

∂u ∂u
a +b =0
∂x ∂y

where a, b ∈ R and its solution is u(x, y) = f (bx − ay) for some function f ,
which requires an extra condition to be determined.

Example
Solve the equation
∂u ∂u
4 −3 =0
∂x ∂y
with u(0, y) = y 3 .

11/21
First order linear PDEs

The general first order linear PDE with variable coefficients and two
independent variables is

∂u ∂u
a(x, y) + b(x, y) =0
∂x ∂y

and its solution requires finding the characteristic curves defined (when
a 6= 0) by
dy
= b/a
dx
which is an ODE.
Example
Solve the equation
∂u ∂u
2y + =0
∂x ∂y
with u(x, 0) = sin(x).

12/21
First order linear PDEs

The characteristic curves are dy/dx = 1/2y whose solution is y 2 = x + c

1.5

0.5

-0.5

-1

-1.5

-2
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2

and the solution of the equation is u(x, y) = sin(x − y 2 ).

Note that is very easy to change the condition, e.g. by u(x, 0) = x3 .
However, not any condition works, try u(0, y) = y 3 . Why?
13/21
First order linear PDEs

The method of characteristics works with more than two independent

variables. Typically, independent variables are t and x.
The solution of the transport equation

∂u
+ v · ∇u = 0
∂t
is u(x, t) = f (ξ) obtained from the characteristic curves

dx
= v(x, t)
dt
with x(0) = ξ. In 1D

∂u ∂u
+v =0
∂t ∂x

14/21
First order linear PDEs

ξ = x − at

L x
15/21
Initial and boundary conditions

The auxliar conditions required to determine a unique solution are

• an initial condition, the value of the function at the initial time

u(x, 0) = u0 (x) ∀x ∈ Ω

where u0 is given.
• a Dirichlet boundary condition at the inlet, the value of the function
at the inlet
u(x, t) = uin (x) ∀x ∈ Γ−
where uin is given.

16/21
Well-posed problems

A well-posed problem is a PDE together with appropriate initial and/or

boundary conditions with the following properties

• Existence: there is exists at least one solution

• Uniqueness: the solution is unique
• Stability: the solution depends continuously on data

Not any combination initial and boundary conditions results in a well-posed

problem.
Appropriate combinations depend on the particular PDE (there is no a
unique recipe).

17/21
Second order linear PDEs

The general second order linear PDE with constant coefficients and two
independent variables is
∂2u ∂2u ∂2u ∂u ∂u
a11 + a12 + a22 + a1 + a2 + a0 u = g
∂x2 ∂x∂y ∂y 2 ∂x ∂y
where a, b, c ∈ R and a2 + b2 + c2 > 0.
Examples:
1. the Poisson equation is
∂2u ∂2u
+ 2 =g
∂x2 ∂y
2. the wave equation
∂2u ∂2u
− =g
∂t2 ∂x2
3. the diffusion equation is
∂2u ∂2u
+ =g
∂t2 ∂x2
18/21
Second order linear PDEs

Classification theorem (two independent variables only!): there is a

change of variables such that
1. if a212 < 4a11 a22 the equation is elliptic and can be reduced to the
generalized Poisson equation
∂2u ∂2u
+ 2 + ... = 0
∂x2 ∂y
2. if a212 > 4a11 a22 the equation is hyperbolic and can be reduced to the
generalized wave equation
∂2u ∂2u
− 2 + ... = 0
∂x2 ∂y
3. if a212 = 4a11 a22 the equation is parabolic and can be reduced to the
generalized diffusion equation
∂2u
+ ... = 0
∂x2
where dots represent one or zero order terms.
19/21
Nonlinear PDEs

• Superposition principle does not hold.

• More difficult to analyze than linear ones.

• No general mathematical theory.

• May present weak (discontinuous) solutions.

20/21
Bibliography i

L. C. Evans. Partial differential equations. American Mathematical Society,

1998. URL
https://discovery.upc.edu/iii/encore/record/C__Rb1152629.
L. E. Malvern. Introduction to the Mechanics of a Continuous Medium.
Prentice-Hall, Inc., 1969. URL https://discovery.upc.edu/permalink/
34CSUC_UPC/rdgucl/alma991000032449706711.
J. E. Marsden and A. Tromba. Vector calculus. W.H. Freeman, 2012. URL
https://discovery.upc.edu/iii/encore/record/C__Rb1421252.
W. A. Strauss. Partial differential equations : an introduction. John Wiley &
Sons, 2008. URL https://discovery.upc.edu/permalink/34CSUC_
UPC/rdgucl/alma991003867079706711.

21/21