PARTIAL DIFFERENTIAL EQUATIONS

1
basic concepts of partial differential equations 1
This is the first of series of lecture
notes primarily taken from the text
Lecturer: Dr. Peter Amoako-Yirenkyi book (Partial Differential Equations-An
introduction by Walter A. Strauss. After
Recommended Textbook:Partial Differential Equations-An going through this lecture notes, you
would be able to:
introduction by Walter A. Strauss2
• know what is a Partial Differential
Equation
• identify a first-Order Linear Equa-
tions
One of the main goals of the theory of partial differential equations is to
express the unknown function of several independent variables from an • describe flows, Vibrations, and
Diffusions
identity where this function appears together with its partial derivatives.
Ordinary and Partial differential equations occur in many applications. • understand and classify Initial and
Boundary Conditions
An ordinary differential equation is a special case of a partial differential
equation but the behaviour of solutions is quite different in general. It 2
W.A. Strauss. Partial Differential
is much more complicated in the case of partial differential equations Equations: An Introduction. Wiley,
2007. ISBN 9780470054567. URL
caused by the fact that the functions for which we are looking at are
http://books.google.com.gh/books?
functions of more than one independent variable. id=PihAPwAACAAJ

Classification, Types of Equations, Boundary and Initial Conditions

In the sequel, we keep the following notation: t denotes the time vari-
able, x, y, z... stand for the spatial variables. The general partial differen-
tial equation in 3-dimension can be written as:

F ( x, y, z, t, u, u x , uy , uz , ut , u xy , u xt , ...) = 0

where ( x, y, z) ∈ Ω R3 , t ∈ I, Ω is a given domain in R3 and I ⊂ R is

a time interval. If F is a vector-valued function, F= ( F1 , ..., Fm ), and we
look for several unknown functions u = u( x, y, z, t), v = v( x, y, z, t), ...,
then
F( x, y, z, t, u, u x , ..., v, v x ) = 0
is a system of partial differential equations. It is clear that these relations
can be, in general, very complicated, and only some of their particu-
lar cases can be successfully studied by a mathematical theory. That
is why it is important to know how to recognize these types and to
distinguish them.

Basic Types of Equations, Boundary and Initial Conditions

Partial differential equations can be classified from various points of view. If time t is one of the independent
variables of the searched-for function, we speak about evolution equations. If it is not the case (the equation
contains only spatial independent variables), we speak about stationary equations. The highest order of the
derivative of the unknown function in the equation determines the order of the equation. If the equation
consists only of a linear combination of u and its derivatives (for example, it does not contain products a
math366 - basic concepts of partial differential equations 2

uu x , u x u xy , etc.), we speak about a linear equation. Otherwise, it is a nonlinear equation. A linear equation
can be written symbolically by means of a linear differential operator L, i.e., the operator with the property
L(αu + βv) = αL(u) + βL(v),
where α, β are real constants and u, v are real functions. The equation

L(u) = 0

is called homogeneous, the equation

L(u) = f ,

where f is given function, is called nonhomogeneous. The function f

represents the "right-hand side" of the equation.
According to the above-mentioned aspects, we can classify the fol-
lowing equation:

1. The transport equation3 has one spatial variable given by : 3

its an evolution equation, first order
linear with L(u) := ut + u x and
ut + u x = 0 homogeneous.

2. The Laplace equation4 in three spatial variables is given by: 4

its a stationary equation, the second
order, linear with L(u) := u xx + uyy +
∆u := u xx + uyy + uzz = 0 uzz and homogeneous.

3. The Poisson equation5 in two spatial variables is given by: 5

its stationary equation, linear second
order with L(u) := u xx − uyy and non
∆u := u xx − uyy = f ( x, y) homogeneous.

where f = f ( x, y) is a given function.

4. The wave equation6 with interaction in one spatial variable is given

by: 6
its evolution of the second order and
utt − u xx + u3 = 0 nonlinear. The interaction is repre-
sented by the term u3 .

5. The diffusion equation7 is one spatial variable is given by: 7

its evolution, of the second order,
linear with L(u) := ut − u xx and non
ut − u xx = f ( x, t) homogeneous.

6. The equation of the vibrating beam8 is given by: 8

its evolution, of the fourth order,
linear with L(u) := utt + u xxxx and
utt + u xxxx = 0 homogeneous.

7. The Schrödinger equation9 is given by: 9

its evolution, of the second order,
linear with L(u) := ut − iu xx and
ut − iu xx = 0 homogeneous

where i is the imaginary unit: i2 = −1

8. The equation of a dispersed wave10 : its evolution, of the third order,

10

nonlinear.
ut + uu x + u xxx = 0
math366 - basic concepts of partial differential equations 3

Solution of a partial Differential Equation

A function u is called a solution of a partial difference equation if, when substituted (together with its partial
derivatives) into the equation, the latter becomes an identity. As in the case of ordinary differential equations,
a solution of the partial difference equation is not determined uniquely. Thus we come to the notion of the
general solution.
Note that while the general solution of an ODE includes arbitrary constants, the general solution of PDE
includes arbitrary functions. This fact is illustrated by the examples below (see. W. Strauss).

example 1. Let us search for a function of two variables u = u( x, y) satis-

fying the equation
u xx = 0 (1)

The problem can be solved by direct integration of equation (1). Since

we integrate with respect to x, the integration "constant" can depend,
in general, on y. From (1) it follows that

u x ( x, y) = f (y)

and further integration gives:

u( x, y) = f (y) x + g(y).

We thus obtain a general solution of equation (1), and f and g are

arbitrary functions of the variable y.

example 2. Let us search for another function u = u( x, y) satisfying the

equation
uyy + u = 0. (2)

Similarly to the case of the following ODE

v00 + v = 0

, the general solution 11 is given by: 11

you should explain why

v(t) = A cos t + B sin t

with arbitrary constants A, B ∈ R , the general solution of the equation

(2) has the form

u( x, y) = f ( x ) cos y + g( x ) sin y,

where f and g are arbitrary functions of the variable x.

math366 - basic concepts of partial differential equations 4

example 3. Let us search for a function u = u( x, y) satisfying the equation

u xy = 0 (3)

Integrating (3) with respect to y, we obtain

ux = f (x)

(f is an arbitrary "constant" depending on x). Further integration with

respect to x then leads to

u( x, y) = F ( x ) + G (y),

where F 0 = f . Functions F and G are again arbitrary. If we look for

twice continuously differentiable function u, then its second partial
derivatives are exchangeable. Hence we can integrate (2.3) first with
respect to x and then with respect to y. Thus both F and G must be
differentiable.

Boundary conditions

As in ODEs, a single PDE does not provide sufficient information to enable us to determine its solution
uniquely. For the unique determination of a solution, we need further information. In the case of stationary
equations, it is usually boundary conditions which, together with the equation, form a boundary value problem.
For example, 
u + u = 0 , ( x, y) ∈ B(0, 1) = {( x, y) ∈ R2 : x2 + y2 < 1},
xx yy
u( x, y) = 0 , ( x, y) ∈ ∂B(0, 1) = {( x, y) ∈ R2 : x2 + y2 = 1}

forms a homogeneous Dirichlet boundary value problem for the Laplace equation. If, in general, Ω is a
bounded domain in R3 , we distinguish the following basic types of (linear) boundary conditions.

1. Dirichlet boundary condition

u( x, y, z) = g( x, y, z), ( x, y, z) ∈ ∂Ω

2. Neumann boundary condition:

∂u
( x, y, z) = g( x, y, z), ( x, y, z) ∈ ∂Ω,
∂n

3. Robin (or Newton) boundary condition:

∂u
A ( x, y, z) + Bu( x, y, z) = g( x, y, z), ( x, y, z) ∈ ∂Ω
∂n
∂u
where ∂n denotes the derivative with respect to the outer normal n to the boundary (surface) of the domain
Ω; A, B, A2 + B2 6= 0, are given constants. If on various parts of the boundary ∂Ω different types of boundary
conditions are given, we speak about a problem with mixed boundary conditions. In our case g ≡ 0, the
boundary conditions are called homogeneous, otherwise they are nonhomogeneous. In one dimension, that is
math366 - basic concepts of partial differential equations 5

in the case of problems on the interval Ω = ( a, b), the boundary ∂Ω consists of two points x = a, x = b .
Then, for example, the nonhomogeneous Neumann boundary conditions reduce to

u x ( a ) = g1 , u x ( b ) = g2

On an unbounded domain, for example, on the interval Ω = (0, +∞), where it is not possible to speak about
a value of the given function at the point "infinity", the homogeneous Dirichlet boundary condition has the
form
u(0) = 0, lim u( x ) = 0
x →∞
In the case of evolution equations, we usually deal, besides boundary conditions, also with initial conditions
which, together with the equation and the boundary conditions, form an initial boundary value problem. For
example,

utt = u xx , t ∈ B(0, ∞), x ∈ (0, 1),


u(0, t) = u(1, t) = 0,


u( x, 0) = ϕ( x ), ut ( x, 0) = ψ( x ),


forms an initial boundary value problem for the one-dimensional wave equation. Here the boundary condi-
tions are the homogeneous Dirichlet ones. The function ϕ denotes the initial displacement and ψ stands for
the initial velocity at a given point x . The derivative ut at time t = 0 is understood as the derivative from the
right. If we look for the so-called classical solution, the functions ϕ and ψ are supposed to be continuous and
also the function u is continuous (in fact, even the partial derivatives of the second order are continuous).
That is why the boundary and initial conditions must satisfy the compatibility conditions

ϕ(0) = ϕ(1) = 0.

By a solution (classical solution) of the initial boundary (or boundary) value problem, we understand a
function which satisfies the equation as well as the boundary and initial conditions pointwise. In particular,
the solution must be differentiable up to the order of the equation. These requirements can be too strong and
thus the notion of a solution of a PDE (or of a system of PDEs) is often understood in another (generalized)
sense. In this course, we concentrate mainly on searching for classical solutions.
Another notion which we introduce in this part, is that of a well-posed
boundary (or initial boundary) value problem. The problem is called
well-posed if the following three conditions are satisfied:

(i) a solution of the problem exists;

(ii) the solution of the problem is determined uniquely;

(iii) the solution of the problem is stable with respect to the given data,
which means that a “small change" of initial or boundary condi-
tions, right-hand side, etc., causes only a “small change" of the so-
lution.

The last condition concerns especially models of physical problems, since the given data can never be mea-
sured with absolute accuracy. However, the question left in the definition of stability is what does “very
small" or “small" change mean. The answer depends on the particular problem and, at this moment, we put
up with only an intuitive understanding of this notion.
math366 - basic concepts of partial differential equations 6

The contrary of a well-posed problem is the ill-posed problem, i.e., a problem which does not satisfy at
least one of the three previous requirements. If the solution exists but the uniqueness is not ensured, the
problem can be underdetermined. Conversely, if the solution does not exist, it can be an overdetermined
problem. For example, the wave equation with non-smooth initial conditions is an ill-posed problem, since
its classical solution does not exist. However, if we consider the solution in a more general sense, the problem
becomes well-posed, the generalized solution exists, it is unique and stable with respect to “small" changes
of given data.

Classification of Linear Equation of the Second Order

In this section we state the classification of the basic types of PDEs of

the second order that can be found most often in practical models.
The basic types(based on the idea of conic section12 ) of linear evolu-
tion equations of the second order are the wave equation (in one spatial
variable): 12
If we expand the following system:
  
A B/2 D/2 x
utt − u xx = 0 ( c = 1), x y 1  B/2 C E/2   y  = 0
D/2 E/2 F 1
which is of hyperbolic type, and the diffusion equation (in one spatial
We will get:
variable):  
Ax + B/2 y + D/2
ut − u xx = 0 ( k = 1), x y

1  B/2 x + Cy + E/2  = 0
which is of parabolic type. The basic type of the linear stationary D/2 x + E/2 y + F

equation of the second order (in two spatial variables) is the Laplace and finally:
equation: Ax2 + Bxy + Cy2 + Dx + Ey + F = 0 (4)
u xx − uyy = 0, In the same way, we can rewrite
equation (4) as follows:
which is of elliptic type. Formal analogues of these PDEs are equations
A B/2 x
of conics in the plane: the equation of a hyperbola, t2 − x2 = 1 the x y + D E +F = 0
B/2 C y
equation of a parabola, t2 − x2 = 1, and the equation of an ellipse
In order to determine the discriminant
(here we mention its special case - a circle), x2 + y2 = 1. of the quadratic equation, we take the
Let us consider a general linear homogeneous PDE of the second determinant of the above equation
order using the coefficient of the higher
powers given by:
a11 u xx + 2a12 u xy + a22 uyy + a1 u x + a2 uy + a0 u = 0 (5) A B/2 1
= AC − B2
B/2 C 4
with two independent variables x, y and with six real coefficient that
AC − 41 B2 can either be greater, less or
can depend on x and y. Let us denote by equal to zero.
" #
a11 a12 For example when AC − 14 B2 > 0
A= B2 − 4AC < 0 representing an elliptic
a12 a22
equation. Note that when A = C and
B = 0 we get a circle
the matrix formed by the coefficients at the partial derivatives of the
When AC − 41 B2 < 0
second order. It is possible to show that there exists a linear trans- B2 − 4AC > 0 representing a Hyper-
formation of variables x, y which reduces equation(2.4)to one of the bola.
following forms. In this respect, an important role is played by the And finally when AC − 14 B2 = 0
determinant det A of the matrix A. B2 − 4AC > 0 representing a Parabolic
equation
math366 - basic concepts of partial differential equations 7

(i) Elliptic form: If det A >0, that is a11 a22 > a212 ,the equation is re-
ducible to the form
u xx + uyy + . . . = 0,

where the dots represents terms with derivatives of lower orders.

(ii) Hyperbolic form: If det A <0, that is a11 a22 < a212 ,the equation is
reducible to the form

u xx − uyy + . . . = 0.

The dots stand for the terms with derivatives of lower orders.

(iii) Parabolic form: If det A =0, that is a11 a22 = a212 ,the equation is
reducible to the form

u xx + . . . = 0, (or uyy + . . . = 0)

unless a11 = a12 = a22 = 0. Here, again, the dots represent terms
with derivatives of lower orders.

Finding the corresponding transformation relations and reducing

the equation is based on the same idea as the analysis of conics in
analytic geometry. For simplicity, let us consider only the principal
terms in the equation, that is, let a1 = a2 = a0 = 0, and let us normalize
∂2
the equation by a11 = 1. If, moreover, we denote ∂ x = ∂x ∂
, ∂2y = ∂y 2 , etc.,
we can write equation (2.4) as

(∂2x + 2a12 ∂ x ∂y + a22 ∂2y )u = 0

and, by formally completing the square, we convert it to

(∂ x + a12 ∂y )2 u + ( a22 − a212 )∂2y u = 0. (6)

Further, let us consider the elliptic case a22 > a212 , denote b= ( a22 −
1
a212 ) 2 , which means that b ∈ R. We introduce new independent vari-
ables ξ and η by
x = ξ, y = a12 ξ + bη
. The transformed derivatives assume the form

∂ξ = ∂ x + a12 ∂y , ∂η = b∂y

(you can prove it using the chain rule), and equation (2.5) becomes

∂2ξ u + ∂2η u = 0
math366 - basic concepts of partial differential equations 8

or, equivalently,
uξξ + uηη = 0

In the remaining two cases, we would proceed analogously (in the

parabolic case we have b = 0 and in the hyperbolic case, b = i( a212 −
1
a22 ) 2 ∈ C.

example 4. We determine types of the following equations:

(a) u xx − 3u xy = 0,

(b) 3u xx − 6u xy + 3uyy + u x = 0,

(c) 2u xx + 2u xy + 3uyy = 0.

In terms of the previous explanation, we decide according to the

sign of det A= a11 a22 − a212 . Thus, in case ( a), we obtain det A= 94 < 0
and the equation is of hyperbolic type. In case (b), we have det A = 0
, and thus the equation is of parabolic type. In case (c), we have det
A = 5 > 0 , and the equation is of elliptic type.

If A is a function of x and y (i.e., the equation has nonconstant co-

efficients), then the type of the equation may be different in different
parts of the xy -plane. See the following two examples.

example 5. We find regions of the xy -plane where the equation

xu xx − u xy + yuyy = 0

is of elliptic, hyperbolic, or parabolic type, respectively.

In this case the coefficients depend on x and y and we obtain det A=

a11 a22 − a212 = xy − 14 . The equation is thus of parabolic type on the
hyperbola xy = 14 , of elliptic type in two convex regions xy > 14 ,
and of hyperbolic type in the connected region xy < 14 . The reader is
invited to sketch a picture of corresponding regions.

example 6. Again, we find regions of the xy -plane where the equation

− x2 u xx + 2xyu xy + (1 + y)uyy = 0

is of elliptic, hyperbolic, or parabolic type, respectively.

math366 - basic concepts of partial differential equations 9

This time we have det A = a11 a22 − a212 = − x2 (1 + y) − x2 y2 . The

equation is thus of hyperbolic type in the whole plane except the axis
y , where it is of parabolic type.

remark 1. In a similar way we can classify linear PDEs of the second order with an arbitrary finite number of variables
N. The coefficient matrix A is then of type N × N. The type of the equation is related to definitenesses of the matrix A
and can be determined by signs of its eigenvalues:
(i) the equation is of elliptic type, if the eigenvalues of A are all positive or all negative (i.e., A is positive or negative
definite);
(ii) the equation is of parabolic type, if A has exactly one zero eigenvalue and all the other eigenvalues have the same
sign (i.e., A is a special case of a positive or negative semidefinite matrix);
(iii) the equation is of hyperbolic type, if A has only one negative eigenvalue and all the others are positive, or A has
only one positive eigenvalue and all the others are negative (i.e., A is a special case of an indefinite matrix);
(iv) the equation is of ultrahyperbolic type, if A has more than one positive eigenvalue and more than one negative
eigenvalue, and no zero eigenvalues (i.e., A is indefinite).
Notice that the matrix A is symmetric, since we consider exchangeable second partial derivatives, and thus all its
eigenvalues have to be real.

References

W.A. Strauss. Partial Differential Equations: An Introduction. Wiley,

2007. ISBN 9780470054567. URL http://books.google.com.gh/
books?id=PihAPwAACAAJ.