ANALYSIS III FOR CIVIL ENGINEERS-LECTURE 1

MENNY AKA

Zusammenfassung. First definitions are given and the classification of

2nd order linear PDEs into Parabolic, Elliptic, Hyperbolic is explained.

Please send in typos and remarks to [email protected]

1. First definitions and examples

1.1. What is a partial differential equation? Let us first fix notation. Let
Ω be an open subset of Rn . For a function f : Ω ⊂ Rn → R, all the following
notations are commonly used in order to denote the partial derivative with
respect to xi :
∂f ∂
, f, ∂xi f, fxi
∂xi ∂xi
We will mainly use the last notation, fxi , but it is important to know them all
in order to be able to read other texts. Similarly,
∂ 2f ∂ 3f
fxy = , fttt = 3 .
∂x∂y ∂t
A Partial Differential Equation 1 or in short PDE, is an equation involving
partial derivatives of some unknown function (which is called the dependent va-
riable). In contrast to Ordinary Differential Equations2 (ODE for short) where
the unknown function is a function of one variable, in a PDE the unknown
function depends on several variables. A solution for a PDE is a function
satisfying the relations inscribed by the equation.
Before giving more definitions let us get the ball rolling with some examples.
1.2. The simplest example. Arguably, the simplest ODE is f 0 = 0. Its
solutions are f = C where C is an arbitrary constant. Similarly, the simplest
PDE one can think of is
∂u(x, y)
= 0.
∂x
This just means that u is independent of x so its solutions are arbitrary functi-
ons of y. That is u(x, y) = f (y) for some arbitrary function f depending only
on y.
1Partielle Differentialgleichung, Abkürzung PDG oder PDGL
2Gewöhnliche Differentialgleichung Abkürzung GDGL
1
 LECTURE 1 2

Remark 1.1. We can already learn two important things from this simple
example. The first is that arbitrary functions play the same role that arbitrary
constants play in the study of ODE’s. The second is that typically, without
posing more conditions on the solution, the space of the solution for a given
PDE is huge.
1.3. Playing with the one-dimensional Heat Equation. The following
PDE is called the one-dimensional Heat equation:
(1.1) ut = uxx .
We will explain in detail later why it is named like that. For now let’s just
say that t stands for time, and x is a spatial3 coordinate. The zero function
u(x, t) ≡ 0 is evidently a solution for (1.1) and the simplest non-zero solution
that comes to mind is u(x, t) = 12 x2 + t. Indeed, uxx = 1 and ut = 1. Can you
find more polynomial solutions?
Exercise 1.2. A polynomial satisfying Equation (1.1) is called heat-polynomial
or caloric polynomial. Can you find other such polynomials?
Exercise 1.3. Find more solutions. You may want to try solutions of the form
eax+bt for a, b ∈ C.
Solution: We did this in class. Substituting u(x, t) = eax+bt in Equation (1.1)
we get
ut = beax+bt = a2 eax+bt = uxx
so u(x, t) = eax+bt is a solution of Equation (1.1) if and only if b = a2 . In other
words, any function of the form
2
u(x, t) = eax+a t , a∈C
if a solution for the PDE (1.1).
Remarks 1.4. (1) The bold attempt to use functions of the form u(x, t) =
eax+bt and hope for the best is usually called Ansatz (also in non
German-speaking countries). This approach is very common in this
field of research as after arriving to a possible solution one may check
if it solves the problem or not.
(2) Sometime the Ansatz will be just a simplifying assumption. For ex-
ample, as we will see, a common simplifying assumption is that the
solution has separated variables, that is, u(x, t) is of the form
u(x, t) = X(x)T (t)

3meaning related to space.

 LECTURE 1 3

where X(x) and T (t) are functions of one variable. The solutions we
found above are such solutions:
2 2
u(x, t) = eax+a t = ea t eax .
Just in order to impress upon the reader how much mathematics one PDE
can generate, one can take a look at this book [Can84] whose topic is equa-
tion (1.1). We will use equation (1.1) in order to introduce some important
mathematical tools.

2. General Classification of PDEs

4
In this course we will mainly learn methods to study the following class of
PDEs:

2nd order linear equations with constant coefficients in small number of variables

Thus we begin by defining the (colored) notions appearing in this title:

2.1. The order of a PDE. The order of the highest derivative appearing in
the equation is called the order of the PDE. For example, the Beam Equation
(2.1) utt = −αuxxxx
has order four, while Equation (1.1), the 3-dimensional Wave Equation
(2.2) utt = c2 (uxx + uyy + uzz ) ,
and equation
uxy = 0
all have order 2. Do not get confused with degree of polynomials! The following
equation
u4x + uy u18 = 0
has order 1.

2.2. The number of variables. This is the numbers of independent varia-

bles of the unknown function u. For example equation (1.1) has 2 variables,
equation (2.2) has 4 variables and equation (2.1) has 2 variables. The only con-
fusion that may arise is that as the above examples show the n-dimensional
Heat/Wave equation has in fact n + 1 variables as n stands only of the spatial
(i.e. space-related) variables and there is another time variable t.

4In this section we loosely follow [Far93, Ch.1] and [KKN11, §12.1].
 LECTURE 1 4

2.3. Linear PDEs. One should think on PDE as equation relating the de-
pendent variable u and its partial derivative. A PDE is called linear if the
dependent variable u and its partial derivatives appear in a linear manner.
This means that u and its partial derivatives are not multiplied, squared, ta-
ken square root of, etc. For understanding this notion, one can also think
about u and its partial derivatives as variables. Then a linear PDE is a linear
equation in these variables.
Remark 2.1. As the examples below show, the ’scalars’ or ’coefficients’ can be
functions (and not just constants) of the independent variables.
These examples are from [Far93, Ch.1]
(1) utt = e−t uxx + sin t is linear. Note that the coefficients multiplying uxx
and the constant function 1 are e−t and sin t respectively.
(2) uuxx + ut = 0 is not linear as u and uxx are multiplied.
(3) uxx + yuyy = 0 is linear. The scalars multiplying uxx and uyy are 1 and
y respectively.
(4) xux + yuy + u2 is not linear as u is multiplied with itself.
We are now in position to further classify the most important class of PDEs
for our course.
3. Further classification of 2nd order linear equations
The class of 2nd order linear PDEs in two variables is the class of equations
of the form
(3.1) Auxx + Buxy + Cuyy + Dux + Euy + F u = G
where A, B, C, D, E, F, G are functions in x and y. If all A, B, C, D, E, F, G
are constants (respectively not all are constants) then equation (3.1) is said to
have constant coefficients (respectively non-constant coefficients).
5
Equation (3.1) is called
homogeneous if G=0
non-homogeneous if G 6= 0
The importance of this notion lies in:
Proposition 3.1 (The superposition principle- first form). If u1 and u2 are
solutions for a homogeneous linear equation Φ and α1 , α2 are arbitrary con-
stants, then
α1 u1 + α2 u2
is also a solution Φ. In other words, the space of solutions of a linear homo-
geneous equation is a vector space.
5The following definition and proposition were not discussed in class yet-we will talk
about it next week
 LECTURE 1 5

Here are some examples to practice these notions: the examples above
utt = e−t uxx + sin t
and
uxx + yuyy = 0
are in the class of 2nd order linear equations. Both have non-constant co-
efficients. The first is non-homogeneous (because of sin t) and the second is
homogeneous. Equation (1.1) is homogeneous with constant coefficients.

3.1. Parabolic, Elliptic and Hyperbolic equations. Equation of the form

(3.1) is said to be
Parabolic if B 2 − 4AC = 0
Hyperbolic if B 2 − 4AC > 0
Elliptic if B 2 − 4AC < 0
The term Parabolic/Hyperbolic/Elliptic is also called the type of the equation.
Remarks 3.2. (1) Note that the type of Equation (3.1) does not depend on
the coefficients D, E, F, G.
(2) If A, B, C are non-constants, i.e., they are functions of x and y, they
may have different type for different values of x and y (see an example
below).
(3) We will see and explain later that the type of the equation stays inva-
riant under coordinate change.
(4) Why is this notion important? It turns out that the type of the equa-
tion tells us quite a lot about it. For example, equation of the same
type usually model similar phenomenas, their solutions share similar
properties, one can apply similar methods in order to
solve them, etc.
B 2
(5) Technical remark- Some books and teachers use 2 − AC or AC −
B 2


2
in order to determine the type. Even worse, some assume that
the coefficient of uxy in Equation (3.1) is 2B and evaluate AC − B 2
for determining the type. So be alert to this issue when you are using
different notes and books.

3.2. Actual ’recipe’ for finding the type of a given equation. This
explanation will be reviewied during your first exercise class. We will exemplify
the steps below using the equation uss = upp for u = u(s, p).
(1) rearrange the equation by moving all the terms (or at least the terms
with the second order partial derivatives) to one side (it does not matter
which one). (e.g. change uss = upp into uss − upp = 0 or upp − uss = 0.)
(2) Find out the coefficient of the mixed partial derivative. This is B. Note
that the names of your variable can vary but we assume that there are
 LECTURE 1 6

only two variables so there is a unique mixed partial derivative of order

2. In our example B = 0 as the coefficient of usp is 0.
(3) Choose one of the variables to be the first and the other to be second,
so you can see who is A and who is C. It doesn’t matter which variable
you choose as the first or second.
3.3. Three main examples.
These examples are the usual representatives of each class of equations. We
will study these equations and their variants in details during our course.
The 1-dimentional Heat Equation ut = αuxx , α > 0.
Here B = 0 as there is no mixed term and A = α, C = 0 (or if we reorder
the variables, A = 0, C = α). In any case B 2 − 4AC = 0 so it is a parabolic
equation.
The 1-dimensional wave equation utt = c2 uxx .
We rewrite utt −c2 uxx = 0 so B = 0 and A = 1, C = −c2 . Therefore B 2 −4AC =
4c2 > 0 so it is a hyperbolic equation.
The 2-dimensional Laplace equation uxx + uyy = 0.
Here B = 0 and A = 1, C = 1. Therefore B 2 − 4AC = −4 < 0 so it is a elliptic
equation.
3.4. Further examples.
The ‘disguised‘ wave equation uξη = 0. We mention, without proof for now,
that this 1-dimensional wave equation utt = c2 uxx under the change of coor-
dinates ξ = x + ct, η = x − ct. By the remark above, this means that both
equations have the same type, so this equation is Hyperbolic. Checking this
directly, we see that B = 1, A = C = 0 so B 2 − 4AC = 1 > 0 so it is indeed
Hyperbolic.
The Euler-Tricomi Equation
(3.2) yuxx + uyy = 0.
Here A = y, B = 0, C = 1 so B 2 − 4AC = −4y. Therefore, Equation (3.2) is
Parabolic if y=0
Hyperbolic if y<0
Elliptic if y>0
3.5. Solution to the Clicker questions. These were taken from the 2013
exam of D–MATL, D–MAVT in analysis III.
Exercise 3.3. Determine the type of
(3.3) uxx + 2uxy + uyy + 3ux + xu = 0

(3.4) uxx + 2uxy + 2uyy + uy = 0

 LECTURE 1 7

(3.5) uxx + 8uxy + 2uyy + ex u = 0

Solution 3.4. Here there is no need to rearrange the equation. For Equation
(3.3), A = 1, B = 2, C = 1, B 2 − 4AC = 0 so it is parabolic. For Equation
(3.4), A = 1, B = 2, C = 2, B 2 − 4AC = −4 so it is elliptic. For Equation
(3.5), A = 1, B = 8, C = 2, B 2 − 4AC = 56 so it is hyperbolic.
Literatur
[Can84] J. R. Cannon, The one-dimensional heat equation, Cambridge University Press,
1984. ↑3
[Far93] S. J Farlow, Partial differential equations for scientists and engineers, Courier
Corporation, 1993. ↑3, 4
[KKN11] E. Kreyszig, H. Kreyszig, and E. J. Norminton, Advanced engineering mathe-
matics, Tenth, Wiley, Hoboken, N.J., 2011. ↑3

M.A. Departement Mathematik, ETH Zürich, Rämistrasse 101, 8092 Zurich,

Switzerland
E-mail address: [email protected]