MA3512: Partial Differential Equations Fall 2021

Lecture 1: Introduction to PDEs

Lecturer: Shun ZHANG Dates:

Disclaimer: These notes have not been subjected to the usual scrutiny reserved for formal
publications. They may be distributed outside this class only with the permission of the
Instructor.

Textbook: We will follow this lecture notes, which is based on (Strauss). Since the
material is very basic, I also recommend several other textbooks. during the class, I will
write on the pdf files and the annotated versions of the pdf file will be uploaded to canvas.
1. (Strauss), Partial Differential Equations, An introduction, by Walter A.
Strauss. This is the textbook used by most of top/good universities for the undergrad-
uate PDE course.
Both 1st and 2nd editions are fine (they are quite similar). I have reserved 3 copies of the
book in the library. Selected solutions can be found online. Cheap copies (english version
imported by Chinese publishing house) can be bought from mainland online websites. You
can also buy used copies online from overseas. Or ask senior students.
2. (Logan), Applied Partial Differential Equations, Third Edition, Springer, by
J. David Logan. Selected solutions of the exercises are also available.
3. (Olver), Peter J. Olver, Introduction to Partial Differential Equations, Springer
4. (Essential), David F. Griffiths, John W. Dold, David J. Silvester, Essential
Partial Differential Equations, Springer.
Books 2,3, and 4 are published by Springer and we have free access to them. I have uploaded
them to canvas. Since we are studying the very basic PDEs, although we use the Strauss as
our main reference, I will point out the sections and chapters in the other 3 books for the
material we will study.
Weights for the final letter grade: 21% Midterm, 9% HWs (3% each), and 70% Final.

1-1
1-2 Lecture 1: Introduction to PDEs

Study Plan. We have 13 teaching weeks. But We will miss one week due to public holiday
and one week due to midterm. Thus we will have 11 teaching weeks. Here is the plan.
1. Introduction to PDEs. Transport equation. IC and BC. Classification of 2nd order PDEs.
2. Wave equation. Cauchy problem
3. Wave equation. Separation of Variables
4. Heat Equation. Cauchy Problem
5. Heat Equation. Separation of Variables, Maximum principle, Stability
6. Fourier Series I
7. Fourier Series II
8. Poisson equation. Maximum principle, Separation of Variables
9. Poisson equation. Poisson’s formula. Dirichlet’s Principle. Green’s first identity
10. Green’s Functions
11. Fourier Transform
Lecture 1: Introduction to PDEs 1-3

1.1 What’s a PDE?

Sec. 1.1 of Strauss, see also Section 1.1 of Logan, Chap 1 of Olver, Chap 1 of
Essential
We study functions u = u(x1 , x2 , · · · , xn ) and their partial derivatives. Here x1 , x2 , · · · , xn
are standard Cartesian coordinates on Rn . We sometimes use the alternative notation u(x, y),
u(x, y, z), etc. We also write u(r, θ, φ) for spherical coordinates on R3 , etc. We sometimes
also have a “time” coordinate t, in which case t, x1 , x2 , · · · , xn denotes standard Cartesian
def
coordinates in R1+n . We also use the alternate notation x0 = t.
We use lots of different notations for partial derivatives:

∂
∂i u = ∂xi u = u, 1 ≤ i ≤ n,
∂xi

∂ 2u
∂ij u = ∂i ∂j u = , 1 ≤ i, j ≤ n.
∂xi ∂xj

Definition 1.1.1 (Partial Differential Equation) A PDE in a single unknown u is an

equation involving u and its partial derivatives. All such equations can be written as

F (u, uxi , uxi xj , uxi1 ···xiN , x1 , · · · , xn ) = f (x1 , · · · , xn ), i1 , · · · , iN ∈ {1, 2, · · · , n}

for some function F . Here N is called the order of the PDE. N is the maximum number of
derivatives appearing in the equation. If f = 0, the PDE is homogeneous. Otherwise, it is
inhomogeneous.
A solution of PDE is a function u(x1 , · · · , xn ) that satisfies the equation identically, at least
in some region of coordinate variables.

Example 1.1.1 • ut + ux = 0, Transport equation, firt-order, homogeneous

• ∂t u − ∂x2 u = 0, Heat equation, second-order, homogeneous

• ∂t2 u − ∂x2 u = 0, Wave equation, second-order, homogeneous

def
• ∆u = ∂x2 u + ∂y2 u + ∂z2 u = 0, Laplace’s equation, second-order, homogeneous.

Example 1.1.2 u = u(t, x)

−∂t2 u + 2∂x2 u + u = t
is a second-order linear PDE. We say that this equation is an inhomogeneous PDE.
1-4 Lecture 1: Introduction to PDEs

Example 1.1.3 u = u(t, x)

∂t u + 2(1 + x2 )∂x3 u + u = t

is a third-order linear PDE. This is a variable coefficient inhomogeneous PDE.

Definition 1.1.2 A linear differential operator L is a differential operator such that

L(au + bv) = aLu + bLv

for all constants a, b ∈ R and all functions u, v.

A PDE is linear if it can be written as

Lu = f (x1 , · · · , xn )

for some linear operator L and some function f .

Example 1.1.4 • L = −∂t2 + (t2 − x2 )∂x2 : linear

• Lu = ∂x2 u + u2 ∂x2 u: nonlinear.

Example 1.1.5 u = u(t, x)

−∂t2 u + (1 + cos u)∂x3 u = 0

is a third-order nonlinear PDE. Also Monge-Ampere equation

det(D2 u) = f

and minimal surface equation

!
Du
div p = 0.
1 + |Du|2

There follows a direct consequence of the linearity of PDE.

Proposition 1.1.1 (Superposition principle) If uP 1 , · · · , um are solutions to the homo-

geneous linear PDE Lu = 0 and c1 , · · · , cm ∈ R, then m i=1 ci ui is also a solution.

The relation between the homogeneous PDE and the inhomogeneous PDE is stated as follows.
Lecture 1: Introduction to PDEs 1-5

Proposition 1.1.2 Let A be the set of all solutions to the homogeneous linear PDE

Lu = 0,

and let u0 be a solution to the inhomogeneous linear PDE

Lu = f.

Then the set of all solutions to the inhomogeneous equation is the translation of A by u0 ,

{u0 + u : u ∈ A}.

The statement above is quite similar that for solving linear algebraic equations.
We recall the way to solve ODE.

Example 1.1.6 Solve

• y 0 (t) + y(t) = 0, y(0) = 1;

• y 00 (t) + 5y 0 (t) + 4y(t) = 0, y(0) = y 0 (0) = 0.

Some PDEs can be solved quite easily, for example

Example 1.1.7 • Find all u(x, y) with ∂x u = 0. u(x, y) = f (y).

• Find all u(x, y) satisfying the equation uxx + u = 0. u(x, y) = f (y) cos x + g(y) sin x.
• Find all u(x, y) satisfying the equation uxy = 0. u(x, y) = F (x) + G(y).

1.2 Simple PDEs (I): Transport Equations

Sec. 1.2 of Strauss, see also Section 1.2 of Logan, 2.1, 2.2 of Olver, 4.1 of Essential

1.2.1 Physical Interpretation

Consider a fluid flowing at a constant rate c along a horizontal pipe of fixed cross section in the
positive x direction. A substance is suspended in the water. Let u(x, t) be its concentration
Rb
at time t. The amount of pollutant in the interval [0, b] at the time t is M = 0 u(x, t)dx.
At the later time t + h, the same molecules of pollutant have moved to the right by c · h.
Hence Z b Z b+ch
M= u(x, t)dx = u(x, t + h)dx.
0 ch
1-6 Lecture 1: Introduction to PDEs

Differentiating with respect to b, we get

u(b, t) = u(b + ch, t + h).

Differentiating with respect to h and putting h = 0, we get

0 = cux + ut .

This is a first order PDE. Its general form, the transport equation, takes

a(t, x)∂t u(t, x) + b(t, x)∂x u(t, x) = 0,

where a(t, x) and b(t, x) are two functions.

1.2.2 Constant coefficient transport equations

Recall that transport equations with constant coefficients

a∂x u(x, y) + b∂y u(x, y) = 0,

where a, b are two constants.

Geometric Method. Geometrically, this equation implies that

(∂x u, ∂y u) · (a, b) = 0,

or equivalently the gradient (∂x u, ∂y u) is perpendicular to the vector (a, b). Thus, the deriva-
tive of u in the direction (a, b) is 0, which implies that u is constant along lines pointing
in the direction of (a, b). The slope of such a line is ab . Therefore, every such line can be
described as the set of solutions to bx − ay = c with c ∈ R. Since u is constant along these
lines, we know that u is a “function that depends only on the line c”. Therefore

u(x, y) = f (c) = f (bx − ay) (1.2.1)

for some function f .

Change of Variables. Change variables to

x0 = ax + by, y 0 = bx − ay.

Replace all x and y derivatives by x0 and y 0 derivatives. By the chain rule,

ux = aux0 + buy0

and
uy = bux0 − auy0 .
Lecture 1: Introduction to PDEs 1-7

Hence
aux + buy = (a2 + b2 )ux0 .
Since a2 + b2 6= 0, the equation takes the form ux0 = 0 in the new (primed) variables. Thus
the solution is u = f (y 0 ) = f (bx − ay) with f an arbitrary function of one variable.
In order to provide more details about u, we set for example u(x, 0) = x2 , then

x2 = f (bx).

Thus
f (c) = b−2 c2 ,
and
u(x, y) = (x − b−1 ay)2 .

Example 1.2.1 Solve the PDE 4ux − 3uy = 0, together with the auxiliary condition that
u(0, y) = y 3 .

By (1.2.1), the solution takes the form u(x, y) = f (−3x − 4y). Setting x = 0 yields the
equation y 3 = f (−4y). Changing of variables gives f (y) = −y 3 /64. Thus

u(x, y) = (3x + 4y)3 /64.

1.2.3 Variable coefficient transport equations

Consider
y∂x u + x∂y u = 0.
Let X denote a point X = (x, y), and let V denote the vector V = (y, x). Using vector
calculus notation, the equation above can be written as ∇(x,y) u(X)·V = 0, i.e., the derivative
of u at X in the direction of V is 0. Thus, that equation implies that u is constant along
the curve C passing through X that points in the same direction as V . This vector can
be viewed as a line segment with slope xy . Therefore, if the curve C is parameterized by
dy
x 7→ (x, y(x)), where we are viewing y as a function of x along C, then C has slope dx , and
y(x) is therefore a solution to the following ODE:

dy x
= ,
dx y

which can be solved as

y2 x2
= + C, C = constant.
2 2
1-8 Lecture 1: Introduction to PDEs

These curves are called the characteristic curves. As C is changed, the curves fill out the xy
plane perfectly without intersecting. On each of the curves u(x, y) is a constant because

d √ x 1
u(x, x2 + 2C) = ∂x u + √ ∂y u = (y∂x u + x∂y u) = 0.
dx x2 + 2C y
√ √ 2 2
Thus u(x, x2 + 2C) = u(0, 2C) is independent of x. Putting C = y −x
2
, we have
p
u(x, y) = u(0, y 2 − x2 ).

It follows that p
u(x, y) = f ( y 2 − x2 ).

Let’s consider two examples.

Example 1.2.2 Solve the equation

• ux + yuy = 0 with u(0, y) = y 3 . u(x, y) = e−3x y 3 .

 
• ux + 2xy 2 uy = 0. u(x, y) = f x2 + y1 .

Proof:[Solution] (1): The characteristic curves are given by the ODE

dy
= y,
dx
and hence y(x) = Cex for some constant C. On each of the curves u(x, Cex ) is a constant
def
and u(x, Cex ) = u(0, Ce0 ) = u(0, C) = f (C). Therefore we have

u(x, y) = f (C) = f (e−x y).

Since u(0, y) = y 3 , there follows f (y) = y 3 . Thus u(x, y) = (e−x y)3 = e−3x y 3 .
dy
(2): The characteristic curves satisfy the ODE dx = 2xy 2 /1 = 2xy 2 . Solving this ODE gives
y = (C − x2 )−1 . Again u(x, y) is a constant on each such curve. So u(x, y) = f (C), where f
is an arbitrary function. Therefore the general solution is obtained by writing

C = x2 + y −1 ,

and u(x, y) = f (x2 + y −1 ).

More generally, we can carry out the same procedure for a PDE of the form

a(x, y)∂x u + b(x, y)∂y u = 0, (1.2.2)

Lecture 1: Introduction to PDEs 1-9

as long as we can figure out how to integrate the ODE

dy b(x, y)
= . (1.2.3)
dx a(x, y)
Keeping in mind, this ODE is solvable whenever the function in the right hand side can be
separated.
Note: The solvability of the transport equation (1.2.2) is just an application of Chain Rule.
Indeed, since a2 (x, y) + b2 (x, y) 6= 0, one assumes that a(x, y) 6= 0 and the equation (1.2.2)
becomes
b(x, y)
∂x u + ∂y u = 0. (1.2.4)
a(x, y)
Thus if one can solve ODE (1.2.3), then the left hand side of (1.2.4) is just an expression of
Chain Rule for the derivative of g(x) = u(x, y(x)) as
dg dy b(x, y)
= ∂x u + ∂y u = ∂x u + ∂y u = 0.
dx dx a(x, y)
Again, by using the Chain Rule and solving ODE (1.2.3), one reduces the original PDE to
an easy ODE.
Exercise.

• Verify by direct substitution that

un (x, y) = sin(nx) sinh(ny)

is a solution of uxx + uyy = 0 for every n > 0.

• Solve the first-order equation 2ut + 3ux = 0 with the auxiliary condition u = cos x
when t = 0.
2
• Solve the equation yux + xuy = 0 with u(0, y) = e−y .

1.3 Initial and Boundary Condition

Sec. 1.4 of Strauss, see also page 6 of Olver, chap 2 of Essential
Typically PDEs have many solutions, as we see from the transport equations, we single out
one solution by imposing auxiliary conditions: initial condition and boundary condition.
These conditions are motivated by the physics.
An initial condition specifies the physical state at a particular time t0 , say for example

u(t0 , x) = φ(x) or/and ∂t u(x, t0 ) = ψ(x).

1-10 Lecture 1: Introduction to PDEs

For the diffusion/heat equation the initial condition is

u(x, t0 ) = φ(x),

where φ(x) is a given function. For the wave equation there is a pair of initial conditions

u(x, t0 ) = φ(x) and ∂t u(x, t0 ) = ψ(x),

where φ is the initial position and ψ is the initial velocity.

In each physical problem there is a domain Ω in which the PDE is valid. For the vibrating
string Ω is the interval 0 < x < l, so its boundary ∂Ω consists only two points x = 0 and
x = l. It is clear, again from our physical intuition, that it is necessary to specify some
boundary condition in order to specify the solution. The three most important kinds of
boundary conditions are

1. Dirichlet Condition: u is specified;

∂u
2. Neumann Condition: the normal derivative ∂n
= ∇u · n is specified;
∂u
3. Robin Condition: ∂n
+ au is specified where a is given function of x.

Each is to hold for all t and for x belonging to the boundary ∂Ω. In one-dimensional problems
where Ω is just an interval 0 < x < l, the boundary consists of just two endpoints, and these
boundary conditions take the simple form

1. Dirichlet: u(0, t) = g(t) and u(l, t) = h(t)

2. Neumann: ∂x u(0, t) = g(t) and ∂x u(l, t) = h(t)

and similarly for the Robin condition. If g and h are equal to zero, the boundary condition
is called homogeneous. Otherwise, it is called inhomogeneous.

1.4 Goals and Properties of PDE

Sect 1.5 of Strauss
Suppose that we are interested in some physical system. A very fundamental question is:

• Which PDEs are good models for the system?

Lecture 1: Introduction to PDEs 1-11

A major goal of modeling is to answer this question. There is no general recipe for answering
it!
One of main goals of PDE theory is to figure out which kind of data lead to a unique solution.
It is not always obvious which kind of data we are allowed to specify in order to solve the
equation. When we have a PDE and a notion of data such that the data always lead to
a unique solution, and the solution depends “continuously” on the data, we say that the
problem is well-posed. Three key questions of a PDE:

• Existence: Does the PDE have any solutions?

• Uniqueness: Are the solutions corresponding to the given data unique?

• Continuous dependence: What happens if we slightly vary the data? Does the solution
then also vary only sightly?

1.4.1 Types of second-order equations

Sect 1.6 of Strauss, see also section 1.9 of Logan, 4.2 of essential
Let’s consider the PDE

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

Theorem 1.4.1 By a linear transformation of independent variables, the equation can be

reduced to one of three forms

1. Elliptic case: if a212 < a11 a22 , it is reducible to

uxx + uyy + · · · = 0.

2. Hyperbolic case: if a212 > a11 a22 , it is reducible to

uxx − uyy + · · · = 0.

3. Parabolic case: if a212 = a11 a22 , it is reducible to

uxx + · · · = 0.

For simplicity, let’s suppose that a11 = 1 and a1 = a2 = a0 = 0. By completing the square,
we then have
(∂x + a12 ∂y )2 + (a22 − a212 )∂y2 u = 0.
1-12 Lecture 1: Introduction to PDEs

p
In the elliptic case, a212 < a22 . Let b = a22 − a212 > 0. Introduce the new variables ξ and η
by
x = ξ, y = a12 ξ + bη.
Then ∂ξ = ∂x + a12 ∂y , ∂η = b∂y , so that the equation becomes

∂ξ2 u + ∂η2 u = 0,

which is Laplace’s.

Example 1.4.1 Classify each of the equations

• uxx − 5uxy = 0. Hyperbolic

• 4uxx − 12uxy + 9uyy + uy = 0. Parabolic

• 4uxx + 6uxy + 9uyy = 0. Elliptic.

Proof:[Solution] Check the sign of the quantity a212 − a11 a22 .