You are on page 1of 107

Chapter 4

Partial Differential Equation

4.1 Introduction
4.1.1 Motivation
Most physical phenomena, whether in the domain of fluid dynamics, electricity, magnetism,
mechanics, optics or heat flow, can be in general (and actually are) described by partial differ-
ential equations.

This area has a long-standing close relationship with the physical sciences, especially physics,
thermodynamics, and quantum mechanics: for many of the topics in the field, the origins of
the problem and the qualitative nature of the solutions are best understood by describing the
corresponding result in physics, as we shall do below.

Roughly corresponding to the initial values in an ODE problem, PDEs are usually solved in
the presence of boundary conditions. For example, the Dirichlet problem (actually introduced
by Riemann) asks for the solution of the Laplace condition on an open subset D of the plane,
with the added condition that the value of u on the boundary of D was to be some prescribed
function f. (Physically this corresponds to asking, for example, for the steady-state distribution
of electrical charge within D when prescribed voltages are applied around the boundary.) It is
a nontrivial task to determine how much boundary information is appropriate for a given PDE.

Linear differential equations occur perhaps most frequently in applications (in settings in which
a superposition principle is appropriate.) When these differential equations are first-order, they
share many features with ordinary differential equations. (More precisely, they correspond to
families of ODEs, in which considerable attention must be focused on the dependence of the
solutions on the parameters.)

Historically, three equations were of fundamental interest and exhibit distinctive behavior.
These led to the clarification of three types of second-order linear differential equations of great
interest. The Laplace equation
∂ 2u ∂ 2u
+ = 0 (4.1.1)
∂x2 ∂y 2
applies to potential energy functions u = u(x, y) for a conservative force field in the plane.
PDEs of this type are called elliptic. The Heat Equation
∂ 2u ∂ 2u ∂u
2
+ 2 = (4.1.2)
∂x ∂y ∂t

143
CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

applies to the temperature distribution u(x, y) in the plane when heat is allowed to flow from
warm areas to cool ones. PDEs of this type are parabolic. The Wave Equation

∂ 2u ∂ 2u ∂ 2u
+ = (4.1.3)
∂x2 ∂y 2 ∂t2

applies to the heights u(x, y) of vibrating membranes and other wave functions. PDEs of this
type are called hyperbolic. The analysis of these three types of equations are quite distinct in
character. Allowing non-constant coefficients, we see that the solution of a general second-order
linear PDE may change character from point to point. These behaviors generalize to nonlinear
PDEs as well. Modern approaches seek methods applicable to non-linear PDEs as well as linear
ones. In this context existence and uniqueness results, and theorems concerning the regularity
of solutions, are more difficult. Since it is unlikely that explicit solutions can be obtained for
any but the most special of problems, methods of ”solving” the PDEs involve analysis within
the appropriate function space – for example, seeking convergence of a sequence of functions
which can be shown to approximately solve the PDE, or describing the sought-for function
as a fixed point under a self-map on the function space, or as the point at which some real-
valued function is minimized. Some of these approaches may be modified to give algorithms
for estimating numerical solutions to a PDE.

4.1.2 Definition
In mathematics and engineering, partial differential equations (PDE) are a type of differential
equation, i.e., a relation involving an unknown function (or functions) u of several independent
variables and its (or their) partial derivatives with respect to those variables.

Partial differential equations (PDEs) are of vast importance in applied mathematics and engi-
neering since so many real physical situations can be modeled by them. PDEs are made up of
partial derivatives . PDEs tend to be divided into three categories - hyperbolic, parabolic
and elliptic.

Standard hyperbolic equations include the wave equation and the advection (convection or
transport) equation. These equations are time-dependent; they model the transient movement
of signals. The wave equation models acoustic and electromagnetic fields, the advection equa-
tion models the translation of waves such as water waves.

Partial differential equations are used to formulate, and thus aid the solution of, problems in-
volving functions of several variables; such as the propagation of sound or heat, electrostatics,
electrodynamics, fluid flow, and elasticity. Seemingly distinct physical phenomena may have
identical mathematical formulations, and thus be governed by the same underlying dynamic.

A Partial Differential Equation (PDE) is an equation relating a function of two or more inde-
pendent variables and its partial derivatives, that is

Definition 4.1.1 A partial differential equation (PDE) is an equation that

1.) has an unknown function (dependent variable) depending on at least two independent
variables,

2.) contains some partial derivatives of the unknown function (dependent variable).

js, ddw de ii: Lecture Notes Page 144 of 249


4.1. INTRODUCTION

4.1.3 Notation
1.) t, x, y, z - the independent variables (here, t represents time while the other variables are
space coordinates),
2.) u = u(t, x, . . .) - the dependent variable (the unknown function),
3.) The partial derivatives will be denoted as follows
∂ 2u
 
∂u ∂u ∂ ∂u
ut = , ux = , uxy = =
∂t ∂x ∂x∂y ∂y ∂x

Example 4.1.1 The following are examples of PDEs.

1.) uuxy + ux = y 3.) uxx − uyy = 0

2.) uxx + 2yuxy + 3xuyy = 4 sin x 4.) (ux )2 + (uy )2 = 1

4.1.4 Existence and Uniqueness


Although the issue of the existence and uniqueness of solutions of ordinary differential equa-
tions has a very satisfactory answer with the Picard-Lindelöf theorem, that is far from the case
for partial differential equations. There is a general theorem (the Cauchy-Kovalevskaya theo-
rem) that states that the Cauchy problem for any partial differential equation that is analytic
in the unknown function and its derivatives have a unique analytic solution. Although this
result might appear to settle the existence and uniqueness of solutions, there are examples of
linear partial differential equations whose coefficients have derivatives of all orders (which are
nevertheless not analytic) but which have no solutions at all: See Lewy (1957). Even if the
solution of a partial differential equation exists and is unique, it may nevertheless have unde-
sirable properties. The mathematical study of these questions is usually in the more powerful
context of weak solutions.

4.1.5 Basic Classification of PDEs.


A PDE can be classified by

1.) Order of the equation 4.) Kinds of coefficients 7.) Kind of second order PDEs
(hyperbolic, parabolic, or el-
2.) Number of variables 5.) Homogeneity
liptic)
3.) Linearity 6.) Degree of a PDE

Definition 4.1.2 The order of a PDE is the highest ordered partial derivative appearing in
the equation.
Example 4.1.2 uxx + 2xuxy + uyy = ey is a second order PDE.
Example 4.1.3 uxxy + xuyy + 8u = 7y is a third order PDE.
Example 4.1.4
first order : ut = ux ,
second order : ut = uxx , uxy = 0,
third order : ut + uuxxx = sin x,
fourth order : uxxxx = utt

js, ddw de ii: Lecture Notes Page 145 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

Definition 4.1.3 PDEs may be classified by the number of their independent variables, that
is, the number of variables the unknown function depends on.
Example 4.1.5
PDE in two variables : ut = uxx , (u = u(t, x))
1 1
PDE in three variables : ut = urr + ur + 2 uθθ , (u = u(t, r, θ)) ,
r r
PDE in four variables : ut = uxx + uyy + uzz , (u = u(t, x, y, z))
Definition 4.1.4 A PDE is said to be linear if it is linear in the unknown function u and all
its derivatives with coefficients depending on the independent variables (all the independent
coefficients are linear). In other words, if a linear combination of the unknown function and its
derivatives u, ux , uy , uz , uxx , . . ..

In other words, a PDE is linear if the dependent variable and all its derivatives appear in a
linear fashion (highest power as one).
Definition 4.1.5 A PDE is said to be nonlinear if it is not linear.
Example 4.1.6 The following are non-linear equations

1.) ut = uxx + u(1 − u) 2.) ut + (u + c)ux = 0

Example 4.1.7 Examples of linear and nonlinear PDEs

1.) Linear: utt − tuxx = sin t 3.) Linear: xuxx + yuyy = 0

2.) Nonlinear: uuxx + ut = 0 4.) Nonlinear: ux + uy + u2 = 0

Definition 4.1.6 PDE can be with constant or variable coefficients (if at least one of the
coefficients is a function of (some of) independent variables) - kinds of coefficients.
Example 4.1.8
constant coefficients : utt + 5uxx − 3uxy = cos x,
variable coefficients : ut + exp(−t)uxx = 0.
Definition 4.1.7 PDE is homogeneous if the free term (the right-hand side term) is zero.
homogeneous : utt − uxx = 0,
nonhomogeneous : utt − uxx = x2 sin t.
Definition 4.1.8 A degree of a PDE is the highest power of the highest partial derivative in
the equation.
Example 4.1.9
4  6
∂ 3u ∂ 2u

∂u
first degree : + −5 = 0,
∂x3 ∂x 2 ∂x
 3 2  4 3
∂ 2u ∂ u ∂ u
fourth order, third degree : + 2
+ (x + 1) 4
= xy 2 .
∂x∂y ∂x ∂y ∂x
Example 4.1.10 Second order Partial Differential Equations
∂ 2u 2
2x ∂ u
2
− e 2
= u3 ,
∂x ∂t
is non-linear PDE.

js, ddw de ii: Lecture Notes Page 146 of 249


4.1. INTRODUCTION

4.1.6 Kinds of Non-Linearity


The types of non-linear partial differential equations are
1.) Semi-linear, 2.) Quasi-linear, 3.) Full nonlinear.

Definition 4.1.9 A PDE is said to be semi-linear if the highest derivatives appear in a linear
fashion and their coefficients do not depend on the unknown function or its derivatives. In
other words, a nonlinear equation is semi-linear if the coefficients of the highest derivative are
functions of the independent variables only.
Example 4.1.11 The following are semi-linear equations
1.) (x + 3)ux + xy 2 uy = u3 3.) ut + ux + u2 = 0

2.) xuxx + (xy + y 2 )uyy + uux + u2 uy = u4 4.) ut + uxxx + uux = 0

5.)
∂u ∂u
+ (x3 + y)y = u3
∂x ∂y
Example 4.1.12 The partial differential equation
ut + uux = 0
is not semi-linear.
Definition 4.1.10 A PDE is said to be quasi-linear if the highest derivatives m appear in a
linear fashion and their coefficients depend on the independent variables, the unknown function
and derivatives of order < m.
Example 4.1.13 The following are quasi-linear equations
1.) " 2 # "  2 # 2
∂ 2u ∂u ∂u ∂ 2 u

∂u ∂u ∂ u
1+ 2
−2 + 1+ =0
∂y ∂x ∂x ∂y ∂x∂y ∂x ∂y 2

2.)
∂u ∂u
x2 u
+ (y + u) = u3
∂x ∂y
Example 4.1.14 The partial differential equations ut + uux = 0 is quasi-linear but not semi-
linear.
Definition 4.1.11 A PDE is said to be fully nonlinear if the highest derivatives appear in a
nonlinear fashion.
Example 4.1.15 The following are fully non-linear equations
1.) uxx uyy − (uxy )2 = x + y 2.) u2x + u2y = 1

Example 4.1.16 yuxx + 2xyuyy + u = 1 is a second order linear PDE.


Example 4.1.17 ux uxx + xuuy = sin y is a second order quasi-linear PDE.
Example 4.1.18
ux + uy = 3uz − 2x2 − 5z (first-order linear)
uxx + u = 4x2 (second-order linear)
5xyuxy − 3zuy + 2u = 0 (second-order linear)
uxz + 2uuy − 4z = 0 (second-order quasilinear)

js, ddw de ii: Lecture Notes Page 147 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

Example 4.1.19 For each of the following PDEs, state whether it is linear, quasi-linear or
non-linear. If it is linear state whether it is homogeneous or non-homogeneous.

1.) uxx + xuy = y: Linear, non-homogeneous

2.) uux − 2xyuy = 0: Quasi-linear

3.) u2x + uuy = 1: Fully Non-linear

4.) uxxxx + 2uyxxyy + uyyyy = 0: Linear, homogeneous

5.) uxx + 2uxy + uyy = sin x: Linear, non-homogeneous

6.) uxx + uxyy + loge u = 0: Fully Non-Linear

7.) u2xx + u2x + sin u = ey : Fully Non-Linear

Remark 4.1.1 Non-linear PDEs are not necessarily fully non-linear.

Example 4.1.20 First order Partial Differential Equations.


∂u ∂u
1.) ∂x
− ∂t
= 0 is linear. 6.) uxx + uyy = x2 + y 2 is inhomogeneous linear.
2.) u ∂u − ∂u
= 0 is non-linear.
∂x ∂t 7.) ut + x2 ux = 0 is homogeneous linear.
3.) ex ∂u
∂x
+ 4 ∂u
∂t
= t is linear, inhomogeneous.
8.) ut + uxxx + uux = 0 is not linear.
4.) ut + ux = 0 is homogeneous linear.
5.) uxx + uyy = 0 is homogeneous linear. 9.) u2x + u2y = 1 is not linear.

4.1.7 Mathematical Problems


A mathematical problem consists of finding an unknown function of the PDE satisfying appro-
priate supplementary conditions. For example

ut − uxx = 0, 0 < x < l, t > 0


IC : u(x, 0) = sin x, 0 < x < l
BC : u(0, t) = 0, t ≥ 0
BC : u(l, t) = 0, t ≥ 0

Definition 4.1.12 A mathematical PDE is said to be properly posed if it satisfies the following
requirements

1.) Existence: The problem has a solution. There is at least one solution.

2.) Uniqueness: This solution is unique. There is at most one solution.

3.) Continuity or Stability: The solution depends continuously on data. In other wards, a small
change in the given data produces a small change in the solution.

Otherwise, it is ill-posed.

Remark 4.1.2 A PDE with initial and boundary conditions constitutes the so-called initial-
boundary-value problem (IBVP). Such problems are mathematical models of most physical
phenomena.

js, ddw de ii: Lecture Notes Page 148 of 249


4.1. INTRODUCTION

4.1.8 General Second Order Classification of PDEs


The general second order linear partial differential equation (PDEs) in two variables x and y is

Auxx + Buxy + Cuyy + Dux + Euy + F u = G

Where A, B, · · · , G are functions of x, y and u or constants.

A second order PDE can be classified as parabolic, elliptic or hyperbolic. If


1). B 2 − 4AC > 0 ⇒ Hyperbolic PDE
2). B 2 − 4AC = 0 ⇒ Parabolic PDE
3). B 2 − 4AC < 0 ⇒ Elliptic PDE
Classify whether the following PDEs are elliptic, hyperbolic or parabolic?
Example 4.1.21 Classify the nature of the equation below

utt − c2 uxx = 0
⇒ A = −c2 , B = 0, C = 1
B 2 − 4AC = 0 − (4)(−c2 )(1) = 4c2 > 0 ⇒ Hyperbolic

Example 4.1.22 Show that the equation is parabolic

ut − kuxx = 0
⇒ A = −k, B = 0, C = 0
2
B − 4AC = 0 − (4)(−k)(0) = 0 ⇒ Parabolic

Example 4.1.23 Determine the conditions on y for the equation to be hyperbolic, parabolic
or elliptic

uxx − yuxy + xuyy + (2x + y)ux − 3yuy + 4u = sin(x2 − 2y)


⇒ A = 1, B = − y, C = x
B 2 − 4AC = y − (4)(1)(x) = y − 4x

If y > 4x ⇒ Hyperbolic
If y = 4x ⇒ P arabolic
If y < 4x ⇒ Elliptic

Example 4.1.24 Determine the regions in the xy plane where the following equation is hy-
perbolic, parabolic, or elliptic.
1
uxx + yuyy + uy = 0
2
Since B 2 − 4AC = −4y

If y < 0 ⇒ Hyperbolic
If y = 0 ⇒ P arabolic
If y > 0 ⇒ Elliptic

js, ddw de ii: Lecture Notes Page 149 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

Example 4.1.25 uxx − uyy = 0 is Hyperbolic since A = 1, B = 0, C = −1

Example 4.1.26 uxx + uyy = 0 is Elliptic since A = 1, B = 0, C = 1

Example 4.1.27 ut = uxx which is uxx − ut = 0 is Parabolic since A = 1, B = 0, C = 0

Example 4.1.28 uy + uux − (uxx ) = 0 is a non-linear but quasi-linear.

Example 4.1.29 (uxx )2 − x2 ux = 0 is both non-linear and not quasi-linear.

Example 4.1.30 Consider the wave equation with a source.

utt − c2 uxx = s(x, t)

Since 0 − 4(1)(−c2 ) > 0, the equation is hyperbolic.

js, ddw de ii: Lecture Notes Page 150 of 249


4.1. INTRODUCTION

4.1.9 Eigen Space Classification of PDEs


Classification of a system of linear or quasi linear 1st order PDEs
∂u ∂u
A +B = C (4.1.4)
∂x ∂y
depend on the characteristic equation
 
B − λA X = 0

λ the eigen values and X eigen vectors.

4.1.9.1 Hyperbolic partial differential equations


Definition 4.1.13 A system of PDEs is said to be Hyperbolic if it has n eigen values and n
linearly independent eigen vectors a.
¯
Example 4.1.31 Examples of Hyperbolic equations are the
1. Wave equation
∂ 2u ∂ 2u
− = 0
∂x2 ∂y 2

2. Advection equation
∂u ∂u
+c = 0
∂t ∂x

3. Compressible Navier-Stokes equations


∂uj
= 0
∂xj
 
∂ρui ∂ρui ∂p ∂ ∂ui
+ uj = − + µ + ρFi
∂t ∂xj ∂xi ∂xj ∂xj
Where

u, ρ, p : is velocity, density and pressure respectively,


µ : coefficient of viscosity,
i : represents the direction of interest,

Hyperbolic PDEs describe transport, advection process, and have a wave like character (wave
propagation). Still a wave, but has moved (advected- some acceleration due to velocity gradi-
ents)

Hyperbolic Physical modeling include, Fluid flow, electro-magnetic, acoustic and elastic waves.

js, ddw de ii: Lecture Notes Page 151 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

To achieve well-posedness, we need to prescribe conditions on the boundaries according to the


number of in-going characteristics.

Hyperbolic equations;

1.) the equations are time-dependent

2.) Characteristics cross, implying only weak solutions

3.) are advection (velocity gradients) dominant

4.1.9.2 Parabolic partial differential equations


Definition 4.1.14 A system of PDEs is said to be Parabolic if it has n eigen values and less
than n linearly independent eigen vectors a.
¯
Example 4.1.32 Examples of Parabolic equations is the Heat or diffusion equation

∂u ∂ 2u
= ν 2
∂t ∂x
Typical behavior of parabolic equations are

1. Equilibrium processes.

2. Time dependent.

3. Smoothing - diffusive - smearing of the wave. Acceleration as a result of diffusion.

js, ddw de ii: Lecture Notes Page 152 of 249


4.1. INTRODUCTION

4.1.9.3 Elliptic partial differential equations


Definition 4.1.15 A system of PDEs is said to be Elliptic if it has n complex (non-real) eigen
values.
Example 4.1.33 Examples of Elliptic equations are the Poisson, Laplacian equations. Equa-
tions with steady diffusion processes - No marching algorithm.

4u = f
∂ u ∂ 2u
2
+ = 0, Laplacian
∂x2 ∂y 2
∂ 2u ∂ 2u
+ = f (x, y), Poisson
∂x2 ∂y 2
Physical applications are in Electric potential, structural mechanics, potential flow.

For an Elliptic problem, we need BC all along the domain.

Characteristics of Elliptic PDEs include


(a) Solutions u is smoother than f .
(b) Well behaved solutions (no discontinuities in the flow).
(c) Memory consuming since all points depend on each other.

4.1.9.4 Mixed partial differential equations


Definition 4.1.16 A system of PDEs is said to be Mixed if it has both real and complex
eigen values

Example 4.1.34 Consider a 2D, inviscid, linearized, compressible Prandth-Glauert fluid flow
equation over a thin body

2
 ∂ 2φ ∂ 2φ
1 − M∞ + = 0 (4.1.5)
∂x2 ∂y 2
where
U∞
M∞ = : Mach number in the free stream
C∞
C∞ : Speed of sound
φ : Velocity pottential

Classify the PDE as Hyperbolic, Parabolic or Elliptic.

Rewrite as a system of 1st order differential PDE

Let

u = φx
v = φy

Then Equation (4.1.5) will become


2

1 − M∞ ux + vy = 0

js, ddw de ii: Lecture Notes Page 153 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

But we do not have enough equations, we also use a property

φxy = φyx
⇒ uy − vx = 0

To have the system of equation as


2

1 − M∞ ux + vy = 0
uy − vx = 0

∂u ∂u
+B A = C
∂x ∂y
 2
      
1 − M∞ 0 ux 0 1 uy 0
+ =
0 −1 vx 1 0 vy 0

For eigen values


B − λA = 0

p −1
λ1 = 2
M∞ − 1
p −1
λ2 = − 2
M∞ − 1

and eigen vectors


 
B − λA X = 0

 
  1
η1
=  p 
η2 2 −1
M∞
 
  1
ζ1
=  p 
ζ2 2 −1
− M∞

2
Thus the system depends on whether M∞ is bigger, equal or smaller than 1.

If
2
(i) M∞ < 1 ⇒ λ1 , λ2 are Complex numbers, thus an elliptic system, a subsonic flow. If
disturb a flow on one point, there is a change/disturbance in all points of flow.
2
(ii) M∞ > 1 ⇒ Either Hyperbolic, a supersonic flow. Convective behavior, disturbance will
convect downstream but not upstreams, do not affect all system.
2
(iii) M∞ = 1 ⇒ λ1 , λ2 are real numbers but both equal to zero, with only one eigen vector
thus a parabolic system, a transonic flow.

Note 4.1.1 Thus characteristics can tell a lot about the flow.

js, ddw de ii: Lecture Notes Page 154 of 249


4.1. INTRODUCTION

Example 4.1.35 For the one dimensional heat equation

∂u ∂ 2u
= ν 2
∂t ∂x
Rewrite as a system of 1st order differential PDE

Let

v = ux

Then the first order system for heat equation will be

ut − νvx = 0
ux = v

∂u ∂u
+B A = C
    ∂t   ∂x  
1 0 ut 0 −ν ux 0
+ =
0 0 vt 1 0 vx 0

For eigen values

B − λA = 0
   
0 −ν 1 0
−λ = 0
1 0 0 0
−λ −ν
= 0
1 0

λ1 , λ2 are undefined (λi = ∞), but since a 2 × 2 matrix, they are two of them. and eigen vectors
 
B − λA X = 0

That gives
   
η1 0
=
η2 1

Has only one eigen vector. Since two eigen values with only one corresponding eigen vector,
the system is Parabolic.
1 λν
Note 4.1.2 If we interchange matrices A and B, we will have = 0 to have λ1 =
    −λ 0
η1 1
λ2 = 0 and = , thus still a parabolic partial differential equation.
η2 0

js, ddw de ii: Lecture Notes Page 155 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

4.1.10 Analytical Solutions of Partial Differential Equations


Classical, analytical or exact solutions of PDEs exist, although they might not be unique or
even very hard and sometimes impossible to generate.

Definition 4.1.17 A solution to PDE is, generally speaking, any function (in the independent
variables) that satisfies the PDE.

Example 4.1.36 It is not difficult to show that u(x, y) = (x + y)3 is a solution to

uxx − uyy = 0

But also u(x, y) = sin(x − y) is also a solution.

For u(x, y) = (x + y)3

ux = 3(x + y)2
uxx = 6(x + y)
uy = 3(x + y)2
uyy = 6(x + y)

Such that uxx − uyy = 6(x + y) − 6(x + y) = 0, thus u(x, y) = (x + y)3 is a solution.

For u(x, y) = sin(x − y)

ux = cos(x − y)
uxx = − sin(x − y)
uy = − cos(x − y)
uyy = − sin(x − y)

Such that uxx − uyy = − sin(x − y) − (− sin(x − y)) = 0, thus u(x, y) = sin(x − y) is a solution.

Note 4.1.3 The example above indicates that PDE classical solution can exist.

Example 4.1.37 Show that u = f (x, y) where f is an arbitrary differentiable functions


satisfies
xux − yuy = 0
and verify that the functions sin(xy), cos(xy), log(xy), exy and (xy)3 are solutions

Example 4.1.38 By setting ux = ν, find the general solution uxy + ux = 0

Exercise 4.1 Show that the given function is a solution to the given partial differential equa-
tion. Assume that κ, ω, a and c are constants.

6) u(x, y) = tan−1 xy for uxx + uyy = 0



1) u(x, y) = x3 − 3xy 2 for uxx + uyy = 0
2 κt
2) u(x, y) = 3x2 y − y 3 for uxx + uyy = 0 7) u(x, t) = e−ω cos(ωx) for ut = κuxx
3) u(x, t) = 2t + x2 for ut = uxx 8) u(x, t) = sin(ωx) cos(aωt) for utt = a2 uxx

4) u(x, t) = x2 + t2 for utt = uxx 9) u(x, t) = f (x + ct) for utt = c2 uxx

5) u(x, y) = ex sin y for uxx + uyy = 0 10) u(x, t) = f (x − ct) for utt = c2 uxx

js, ddw de ii: Lecture Notes Page 156 of 249


4.1. INTRODUCTION

Exercise 4.2 Show that


1 2
u(x, t) = √ e−x /(4t)
t
is a solution to the heat equation ut = uxx .
−1/2
Exercise 4.3 Show that u(x, y, z) = (x2 + y 2 + z 2 ) is a solution to the 3 dimensional
Laplace equation
∂ 2u ∂ 2u ∂ 2u
+ + =0
∂x2 ∂y 2 ∂z 2
Exercise 4.4 Let i2 = −1 and suppose that u(x, y) and v(x, y) are such that

(x + iy)2 = u(x, y) + iv(x, y)

Find u and v and show that both satisfy Laplace’s equation-that is, that

∂ 2u ∂ 2u ∂ 2v ∂ 2v
+ = 0 and + =0
∂x2 ∂y 2 ∂x2 ∂y 2
In addition, show that u and v satisfy the Cauchy-Riemann Equations

ux = vy , uy = −vx

Example 4.1.39 Show that u(x, y) = x2 − y 2 and u(x, y) = ex sin y are a solution to

uxx + uyy = 0

For u(x, y) = ex sin y

ux = ex sin y
uxx = ex sin y
uy = ex cos y
uyy = −ex sin y

Such that uxx + uyy = ex sin y + (−ex sin y) = 0, thus u(x, y) = ex sin y is a solution.

Note 4.1.4 For a continuous function u,

uxy = uyx

Remark 4.1.3 We have so far seen how to show that a function is a solutions, however,
analytically, first and second order PDEs can be solved using the method of

1). Characteristics (Canonical forms), as will be seen in Section 4.2.4.

2). Separation of variables, as discussed in Section 4.3.

3). Laplace transforms, as covered in Section 1.8.2.

4). Fourier transforms and Fourier series, as applied in Section 4.3.

js, ddw de ii: Lecture Notes Page 157 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

4.2 Method of Characteristic Equations - - First Order


PDEs
A linear PDE of order one involving a dependent variable u and two independent variables x
and y has the form
∂u ∂u
P +Q = R (4.2.1)
∂x ∂y
Where P, Q and R are functions of x, y and u or constants.

If P = 0 or Q = 0, then the equation (4.2.1) is easy to solve, for example


Example 4.2.1
∂u
= 2x + 3y
∂x
⇒ u(x, y) = x2 + 3xy + φ(y)

Definition 4.2.1 Consider the equation


∂u ∂u
P +Q = R
∂x ∂y
Lagrange reduced equation (4.2.1) to that of solving the auxilliary system
dx dy du
= = (4.2.2)
P Q R
The method of Characteristic equation of a first order PDE (4.2.1)
Example 4.2.2 Solve the first order PDE
∂u ∂u
x +y = 3u
∂x ∂y
using the characteristic equations
dx dy du
= =
x y 3u

dx dy
= (4.2.3)
x y
Z Z
dx dy
⇒ =
x y
⇒ ln x = ln y + C
x
⇒ = eC = A = φ(B)
y

dx du
= (4.2.4)
x Z3u Z
dx du
⇒ =
x 3u
1
⇒ ln x = ln u + C
3
js, ddw de ii: Lecture Notes Page 158 of 249
4.2. METHOD OF CHARACTERISTIC EQUATIONS - - FIRST
ORDER PDES
x
⇒ 1 = eC
u 3

x3
⇒ = e3C = B = φ(A)
u
But for any constants A, B, C, D can express one constant in terms of the others.
D = AC + B
Therefore, the solution is
x3
 
x
=A +B
u y
Example 4.2.3 Solve the PDE
∂u ∂u
2 +3 =1
∂x ∂y
using the characteristic equations
dx dy du
= =
2 3 1
dx du
= (4.2.5)
2 1
⇒ x − 2u = 2C = φ(B)

dx dy
= (4.2.6)
2 3
⇒ 3x − 2y = 6C = φ(A)
The final solution is
(x − 2u) = A (3x − 2y) + B
Remark 4.2.1 You always equate those to have better integration
Example 4.2.4 Solve the PDE
∂u ∂u
y2u − x2 u = x2 y
∂x ∂y
using the characteristic equations.
dx dy du
= =
y2u −x2 u x2 y
With
dx dy
= (4.2.7)
y2u −x2 u
x3 y 3
⇒ − − =A
3 3
⇒ x3 + y 3 = −3A = φ(B)
Using
dy du
= (4.2.8)
−x2 u x2 y
y 2 u2
⇒ + =C
2 2
⇒ y 2 + u2 = 2C = φ(A)
The solution is thus
(y 2 + u2 ) = A x3 + y 3 + B


js, ddw de ii: Lecture Notes Page 159 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

Example 4.2.5 Solve the PDE


∂u ∂u
a +b + cu = 0
∂x ∂y
cx
u(x, y) = e− a φ(ay − bx)

Exercise 4.5 Solve the linear first order partial differential equation
∂u
=0
∂x
Solving the equation
x − y = const
Consequently, the general form of the solution is

u(x, y) = F (x − y)

where F is an arbitrary function

Exercise 4.6 Solve the linear first order partial differential equation
∂u ∂u
+ =0
∂x ∂y
Exercise 4.7 Solve the linear first order partial differential equation
1 ∂u 1 ∂u
+ =0
x ∂x y ∂y

such that u(x, 0) = µx4 where µ is a constant.

Solving the equation we get


x2 − y 2 = −const
Consequently, the general form of the solution is

u(x, y) = F (x2 − y 2 )

where F is an arbitrary function.

Substituting in the problem, it is evident that the latter is satisfied for all choices of F . With
the initial condition, a specific form of the solution can be obtained as

u(x, y) = µ(x2 − y 2 )

js, ddw de ii: Lecture Notes Page 160 of 249


4.2. METHOD OF CHARACTERISTIC EQUATIONS - - FIRST
ORDER PDES

Exercise 4.8 Solve the Quasi-linear equation

∂u ∂u
2y +u = 2yu2
∂x ∂y

1 y2 2 2
u = e = Aey ⇒ A = ue−y
A
1 1
x = Be− u ⇒ B = xe u , ⇒
A = φ(B) + C
 1
−y 2
ue = φ xe u + C

Or
2 2 2 2
u = Aey , & Ax + e−y = B, ⇒ B = Ax + e−y = (1 + xu)e−y
to have h i
2 2
u(x, y) = ey F (1 + xu)e−y

Since with two arbitrary constants A and B, and F, φ an arbitrary function.

A = F (B)

Exercise 4.9 Solve the first order partial differential equation


 2  2
∂u ∂u
x + = 2xyu2
∂x ∂y

Exercise 4.10 Find the general solution of

xux + 2xuuy = u

and then seek those solutions (if they exist) subject to:

1.) u = 2x on y = 2x2 + 1;

2.) u = 2x2 on y = 3x3

3.) u = x2 on y = x3 − 1.

dy
x = Au ⇒ u = Ax & = 2u = 2Ax ⇒ y = Ax2 + B ⇒ A = F (B)
dx
Example 4.2.6 Find the general solution of

ux + 2xuy = u2
1
y = x2 + A & − u
=x−B

Exercise 4.11 Show that


u = c1 cos x + c2 sin x
is a general solution to the partial differential equation

uxx = −u

js, ddw de ii: Lecture Notes Page 161 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

Example 4.2.7 Solve the partial differential equation


∂u ∂u
+ (u + y) + u = 0, u(x, 0) = x.
∂y ∂x
Our two ordinary differential equations become
du dx
= −dy, u(0) = ξ; = dy, x(0) = ξ.
u u+y
We note that the first differential equation does not involve x. Using the initial condition we
quickly obtain
u = ξe−y .
Substituting this into the second differential equation we have
dx dx
= dy ⇔ dx = ξe−y + y dy ; x(0) = ξ

= dy ⇔ −y
u+y ξe + y
Since the right hand side does not involve the unknown x we can integrate to obtain
y2  y2
Z
ξe−y + y dy = −ξe−y + + A ; x(0) = ξ ⇒ A = 2ξ ⇔ x = ξ 2 − e−y +

x(y) =
2 2
Solving for ξ in terms of x and substituting into the formula for u we have
2
x − y2
ξ=
2 − e−y
To have 2 2
x − y2 −y x − y2
u = u(x, y) = e =
2 − e−y 2ey − 1
The basic solution pattern (given initial conditions) is:
(i) Solve the two differential equations for x and u (if possible) with
x(0) = ξ, u(ξ, 0) = u0 (ξ)
In terms of ξ, which can even be u(0) = ξ or any other function in ξ.
(ii) Use one the solutions from i) to solve for ξ and u in terms of x and y.
Example 4.2.8 We revise partial differentiation by looking at a few examples.
1.)
u = x2 sin t,
∂u ∂u
⇒ = 2x sin t, = x2 cos t.
∂x ∂t
2.)
2
u = x2 + t2 + ex ,
∂u
= 2 x2 + t2 2x + ex = 4x x2 + t2 + ex .
 

∂x
3.) x
−1
u = tan ,
t
∂u 1 1 t
⇒ = 2 2
. = 2 ,
∂x 1 + x /t t x + t2
∂u 1 −x −x
and = . = .
∂t 1 + x2 /t2 t2 x2 + t2

js, ddw de ii: Lecture Notes Page 162 of 249


4.2. METHOD OF CHARACTERISTIC EQUATIONS - - FIRST
ORDER PDES

Example 4.2.9 Consider the linear, equation


∂u ∂u
+
c = −u.
∂x ∂t
Show that u = e−t F (x − ct) is a solution, where F is an arbitrary function.
∂u ∂u
= e−t F 0 (x − ct), = −e−t F (x − ct) − ce−t F 0 (x − ct).
∂x ∂t
Hence,
∂u ∂u
= −u − c
∂t ∂x
and the equation is satisfied.
Example 4.2.10 Consider the linear equation without non-constant coefficients,
∂u ∂u
x
+ = 0, x > 0.
∂x ∂t
The solution is given by u = F (xe−t ), where F is again an arbitrary function but this time
z = xe−t .
∂u ∂u
= e−t F 0 (xe−t ), = −xe−t F 0 (xe−t ),
∂x ∂t
and so the equation is clearly satisfied.
Example 4.2.11 Verify that u = sin(x − ct) is a solutions to first order PDE
∂u ∂u
+ c =0
∂x ∂t
where c is a constant called the wave speed.
u = sin(x − ct)
∂u
= cos(x − ct)
∂x
∂u
= cos(x − ct)(−c)
∂t
∂u ∂u
c + = 0
∂x ∂t
Example 4.2.12 Verify that

 1 − (x − ct)2 0 ≤ (x − ct)2 ≤ 1,
u=
0 (x − ct)2 ≥ 1,

is a solutions to first order PDE


∂u ∂u
c + =0
∂x ∂t

∂u  −2(x − ct), 0 ≤ (x − ct)2 ≤ 1,
=
∂x 
0, (x − ct)2 ≥ 1,

∂u  2c(x − ct), 0 ≤ (x − ct)2 ≤ 1,
=
∂t
0, (x − ct)2 ≥ 1,

js, ddw de ii: Lecture Notes Page 163 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

Example 4.2.13 Verify that u = F (x − ct) where F is an arbitrary function is a solutions to

∂u ∂u
c + =0
∂x ∂t
It is easier to understand what is happening if we define z = x − ct, so that u = F (z). Hence,
the partial derivatives, on using the chain rule, are
∂u ∂u
= F 0 (z).1, = F 0 (z).(−c).
∂x ∂t
Therefore,
∂u ∂u
c + = 0,
∂x ∂t

for any functional form for F . This illustrates an important point. Ordinary differential
equations have arbitrary constants but Partial differential equations have arbitrary functions.
The function F is determined by an initial condition.

Example 4.2.14 Assume that the initial condition, at t = 0, is u(x, 0) = f (x), where f is a
prescribed function. Now if

u(x, t) = F (x − ct) ⇒ u(x, 0) = F (x).


Hence, the unknown function, F is the prescribed function, f but the argument is replaced
by x − ct. Hence, the solution is

u(x, t) = f (x − ct).
−x2
As an illustration, if f (x) = e , then
2
u(x, t) = e−(x−ct) ,
satisfies Equation
∂u ∂u
c + =0
∂x ∂t
and the initial condition.

Example 4.2.15 Solve the first order PDE with initial condition using the method of charac-
teristic equations
∂u ∂u
2 + = 0, −∞ < x < ∞, t > 0,
∂x ∂t
with the initial condition
1
u(x, 0) = ,
1 + x2
The solution is
1
u(x, t) =
1 + (x − 2t)2

js, ddw de ii: Lecture Notes Page 164 of 249


4.2. METHOD OF CHARACTERISTIC EQUATIONS - - FIRST
ORDER PDES

Example 4.2.16 Solve the PDE


∂u ∂u
− + = 0, −∞ < x < ∞, t > 0,
∂x ∂t
with the initial condition

 1 − |x|, |x| ≤ 1
u(x, 0) =
0, |x| > 1

using characteristic curves , the solutions is



 1 − |x + t|, |x + t| ≤ 1
u(x, t) =
0, |x + t| > 1

Example 4.2.17 Solve the boundary problem


∂u ∂u
2 + = 0,
∂x ∂t
with 
 1 − x 0 < x ≤ 1,
u(x, 0) =
0 x > 1,

u(0, t) = e−t , t>0


The characteristics are given by
x0 = x − 2t.
Therefore the plane is split into three regions and the solution is


 0 x − 2t > 1,



u(x, t) = 1 − (x − 2t) 0 < x − 2t ≤ 1,




e−(t−x/2) x − 2t < 0.

A little bit of thought about this last example will soon illustrate a major problem. If the wave
speed is negative, i.e. c < 0, then there is in general no solution to the initial-boundary-value
problem. This is because the characteristics intersect both boundaries and u cannot be defined
by two different values (resulting from values of f (x) and g(t)) on the same characteristics.

Example 4.2.18 u = sin(r + s2 ) and r = tex and s = x + t. Hence,


∂r ∂s ∂r ∂s
= ex t, = 1, = ex , = 1.
∂x ∂x ∂t ∂t
and so
∂u
= cos tex + (x + t)2 {ex + 2(x + t)} ,
 
∂t
using the chain rule .

js, ddw de ii: Lecture Notes Page 165 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

Example 4.2.19 Consider the initial-boundary-value problem


∂u ∂u
u2 + = 0, x > 0, t > 0,
∂x ∂t
with √
u(x, 0) = x x > 0,
u(0, t) = 0 t > 0.
So √
u= x − u2 t, x − u2 t > 0.
Squaring both sides, we can manipulate this solution into an explicit solution for u in terms of
x and t.

u2 = x − u2 t,
u2 (1 + t) = x,
x
u2 = ,
1+t
and so the final solution is r
x
u= , x > 0, t > 0.
1+t
We can easily check that this is the solution to the problem by calculating the partial derivatives,
∂u 1
= x−1/2 (1 + t)−1/2 ,
∂x 2
∂u 1
= − x1/2 (1 + t)−3/2 ,
∂t 2
Therefore,
2 ∂u x 1 1 x1/2 ∂u
u = 1/2 1/2
= 3/2
=−
∂x (1 + t) 2x (1 + t) 2 (1 + t) ∂t
Example 4.2.20 Obtain a solution, u(x, t), to the equation

∂ 2u
+ u = 0.
∂t2
Note that there are only derivatives with respect to t and none with respect to x. Thus, we
can treat the equation like an ordinary differential equation and use first year work to write
the solution.
u(x, t) = A cos t + B sin t
However, unlike the ordinary differential equation case A and B are not constants but are, in
fact, functions of the other independent variable. Hence, the actual solution is

u(x, t) = A(x) cos t + B(x) sin t

as can be verified by differentiating and substituting into the differential equation.


Exercise 4.12 Solve the equation
dx dy du
2
= 2 = 2
uy ux y x

x 2 − y 2 = c1 , y 3 − u 3 = c2

js, ddw de ii: Lecture Notes Page 166 of 249


4.2. METHOD OF CHARACTERISTIC EQUATIONS - - FIRST
ORDER PDES

Example 4.2.21 Consider the equation


∂ 2u
 
∂ ∂u ∂u
≡ =−
∂t∂x ∂t ∂x ∂x
To solve this we set p = ∂u/∂x so that the equation becomes
∂p
= −p
∂t
Thus, the solution is
p = c(x)e−t
Now we must integrate with respect to x to get the solution u(x, t). Therefore, we get

u(x, t) = f (x)e−t + g(t),


R
where f (x) = c(x)dx and g(t) is an arbitrary function of integration when we integrate with
respect to x

Example 4.2.22 Solve the problem


∂u ∂u
2xt + = u, −∞ < x < ∞,
∂x ∂t
with the initial condition
u(x, 0) = x
The initial condition is in fact
u(x, 0) = x(0) = ξ,
The characteristic equations are
dx dt du
= =
2xt 1 u

du dt
= ⇔ ln u = t + ln A ⇔ u = Aet ; u(0) = ξ ⇒ A = ξ
u 1
u = ξet (4.2.9)

dx dt dx 2
= ⇔ = 2tdt ⇔ ln x = t2 + ln A ⇔ x = Aet ; x(0) = ξ ⇒ A = ξ
2xt 1 x
2
x = ξet (4.2.10)
2
From (4.2.10), ξ = xe−t to have
2 2
u(x, t) = ξet = xe−t · et = xet−t

Exercise 4.13 Solve the partial differential equation


∂u ∂u
+x + u = 0, x(0) = ξ, u(ξ, 0) = ξ 2 .
∂y ∂x
Here the initial data is given on y = 0.

x = ξey
u = ξ 2 e−y

u(x, y) = x2 e−3y

js, ddw de ii: Lecture Notes Page 167 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

4.2.1 Linear Equations Method of Characteristics


Goal: Develop a technique to solve the (somewhat more general) first order PDE
∂u ∂u
+ p(x, y) = 0 (4.2.11)
∂x ∂y
Idea: Look for characteristic curves in the xy-plane along which the solution u satisfies
an ODE.
Consider u along a curve y = y(x). On this curve we have
d ∂u ∂u dy
u(x, y(x)) = + (4.2.12)
dx ∂x ∂y dx
From (4.2.11) and (4.2.12) we have
d
u(x, y(x)) = 0 (4.2.13)
dx
dy
= p(x, y) (4.2.14)
dx
Example 4.2.23 Solve the transport equation
∂u ∂u
+5 = x, u(x, 0) = sin 2πx, − ∞ < x < ∞.
∂x ∂t
Idea: Perform a linear change of variables to eliminate one partial derivative by use of multi-
variable chain rule on a special characteristic.

Consider u along a curve t = t(x) (the independent variable should be the first partial derivative,
whose coefficient is 1 in the PDE). On this curve we have
d ∂u ∂u dt
u(x, t(x)) = +
dx ∂x ∂t dx
such that the ordinary differential equations
d
u(x, t(x)) = x
dx
dt
= p(x, t) = 5
dx
Then on solutions of
du 1 1
= x ⇔ u = x2 + A; u(x, 0) = sin 2πx = sin 2πx(0) = sin 2πξ ⇒ A = sin 2πξ − ξ 2
dx 2 2
to have
1 1
u = x2 + sin 2πξ − ξ 2 .
2 2
and
dt
= 5 ⇔ t = 5x + B; x(0) = ξ ⇒ B = −5ξ ⇒ t = 5x − 5ξ
dx
Solving for ξ in terms of (x, t) to have
(5x − t)
ξ=
5
and substituting into the formula for u.
   2
1 1 1 5x − t 1 5x − t
u = u(x, t) = x2 + sin 2πξ − ξ 2 = x2 + sin 2π −
2 2 2 5 2 5

js, ddw de ii: Lecture Notes Page 168 of 249


4.2. METHOD OF CHARACTERISTIC EQUATIONS - - FIRST
ORDER PDES

Example 4.2.24 Solve the transport equation


∂u ∂u 2
+3 = 0, u(x, 0) = xe−x .
∂t ∂x
Idea: Perform a linear change of variables to eliminate one partial derivative by use of multi-
variable chain rule on a special characteristic.

Consider u along a curve x = x(t) (the independent variable should be the first partial deriva-
tive, whose coefficient is 1 in the PDE). On this curve we have
d ∂u ∂u dx
u(x(t), t) = +
dt ∂t ∂x dt
such that the ordinary differential equations
d
u(x(t), t) = 0
dt
dx
= p(x, t) = 3
dt
Then on solutions of
du 2 2 2
= 0 ⇔ u = A; u(0) = xe−x = x(0)e−x(0) = ξe−ξ
dt
to have
2
u = ξe−ξ .
and
dx dx
=3 ⇔ = 3; x(0) = ξ
dt dt
Since the right hand side does not involve the unknown x we can integrate to obtain
Z
x(t) = 3dt = 3t + A ; x(0) = ξ ⇒ A = ξ ⇔ x = 3t + ξ

Solving for ξ in terms of x to have


ξ = (x − 3t)
and substituting into the formula for u.
2 2
u = u(x, t) = ξe−ξ = (x − 3t)e−(x−3t)

Example 4.2.25 Solve the partial differential equation


∂u ∂u
+u = 0, u(x, 0) = u0 (x) = x2 .
∂y ∂x
As is usual by now we change the initial condition to u(ξ, 0) = u0 (ξ) = ξ 2 .

Consider u along a curve x = x(y). On this curve we have


d ∂u ∂u dx
u(x(y), y) = +
dy ∂y ∂x dy
such that the ordinary differential equations
d
u(x(y), y) = 0
dy

js, ddw de ii: Lecture Notes Page 169 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

dx
= p(x, y) = u
dy
Then on solutions of
du
= 0 ⇔ u = A; u(0) = ξ 2
dy
to have
u = ξ2.
and
dx dx
=u ⇔ = ξ 2 ; x(0) = ξ
dy dy
Since the right hand side does not involve the unknown x we can integrate to obtain
Z
x(y) = ξ 2 dy = ξ 2 y + A ; x(0) = ξ ⇒ A = ξ ⇔ x = ξ 2 y + ξ

Solving for ξ in terms of x


x = ξ2y + ξ ⇔ ξ2y + ξ − x = 0
a quadratic in ξ with solution √
−1 ± 1 + 4xy
ξ=
2y
and substituting into the formula for u. Now u = ξ 2 so the solution could be obtained by squar-
x−ξ
ing the above formula for ξ. But it is better to use the quadratic equation to get ξ 2 = ,
y
after which we have
 √  √
x 1 −1 ± 1 + 4xy 2xy + 1 1 + 4xy
u = u(x, y) = − = 2
±
y y 2y 2y 2y 2

Example 4.2.26 Solve the first order partial differential equation


∂u ∂u
+ (2x + u) − (x + 2u) = 0, u(x, 0) = −x + 1.
∂y ∂x
As is usual by now we change the initial condition to u(ξ, 0) = u0 (ξ) = −ξ + 1.

Consider u along a curve x = x(y). On this curve we have


d ∂u ∂u dx
u(x(y), y) = +
dy ∂y ∂x dy
such that the ordinary differential equations
d dx
u(x(y), y) = (x + 2u) , = p(x, y) = (2x + u)
dy dy
That is
dx
= (2x + u)
dy
du
= (x + 2u)
dy
with x(0) = ξ, u(ξ, 0) = u0 (ξ) = −ξ + 1.

js, ddw de ii: Lecture Notes Page 170 of 249


4.2. METHOD OF CHARACTERISTIC EQUATIONS - - FIRST
ORDER PDES

This is a linear system of two first order differential equations


   
dX x 2 1
= AX, where X = , and A = (4.2.15)
dy u 1 2

That can be solved by either Matrix Method (Undetermined coefficient or order reduction if
non-homogeneous system).

The matrix A with the characteristic polynomial of P (λ) = λ2 − 4λ + 3 = (λ − 1)(λ − 3) = 0


whose eigen values and their corresponding eigen vectors given by
   
η1 −1
λ1 = 1 , =
η2 1
   
η1 1
λ2 = 3 , =
η2 1

Real and distinct eigen values, to have solutions as


     
x y −1 3y 1
= c1 e + c2 e
u 1 1

Applying the initial conditions x(0) = ξ, u(ξ, 0) = u0 (ξ) = −ξ + 1,


     
ξ −1 1 −c1 + c2 = ξ
= c1 + c2 ⇒
−ξ + 1 1 1 c1 + c2 = −ξ + 1

with the solution  


1 1 1 y 1 3y
c1 = −ξ + , c2 = ⇒ x= ξ− e + e .
2 2 2 2
Then
1 x − e3y /2
ξ− = ,
2 ey
To have  
1 y 1 3y
u = u(x, y) = − ξ − e + e = e3y − x
2 2
Example 4.2.27 Consider the PDE, with initial and boundary data,
∂u ∂u
+ (x + 1) + (u − y) = 0, u(ξ, 0) = ξ, u(0, η) = η − 1.
∂y ∂x

Consider u along a curve x = x(y). On this curve we have

d ∂u ∂u dx
u(x(y), y) = +
dy ∂y ∂x dy
such that the ordinary differential equations
d dx
u(x(y), y) = (−u + y) , = p(x, y) = (x + 1)
dy dy
That is
dx
= (x + 1)
dy

js, ddw de ii: Lecture Notes Page 171 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

du
= (−u + y)
dy
This is a non-homogeneous linear system of two first order differential equations
     
dX x 1 0 1
= AX + f (y), where X = , and A = , f (y) =
dy u 0 −1 y
That can be solved by either method of Undetermined Coefficient or Order Reduction technique.

The matrix A with the characteristic polynomial of (1 − λ)(−1 − λ) = 0 whose eigen values
and their corresponding eigen vectors given by
   
η1 1
λ1 = 1 , =
η2 0
   
η1 0
λ2 = −1 , =
η2 1
Real and distinct eigen values, to have the complimentary (homogeneous part) solutions as
     
x y 1 −y 0
= c1 e + c2 e
u 0 1
 
1
Since f (y) = only a polynomial of degree 1, then let the particular solutions Xp be
y
defined as    
a1 y + a2 0 a1
Xp = ⇒ Xp =
b1 y + b2 b1
A particular solution is a solution, so satisfys the linear system of ordinary differential equations.
Substituting in the formula X 0 = AX + f (y) to have
Xp0 = AXp + f (y)
      
a1 1 0 a1 y + a2 1
= +
b1 0 −1 b1 y + b2 y
Equating the coefficients of y and coefficients of a constant respectively are :
ay + (a1 − a2 + 1) = 0 a1 = 0 , a2 = −1

(1 − b1 )y + (−b2 − b1 ) = 0 b1 = 1 , b2 = −1
The particular solutions is therefore given by
   
a1 y + a2 −1
Xp = =
b1 y + b2 y−1
The general solutions (sum of complimentary and particular solutions) is given by
X G = Xc + Xp
       
x y 1 −y 0 −1
= c1 e + c2 e +
u 0 1 y−1
So that
x = c1 ey − 1 (4.2.16)
u = c2 e−y + y − 1 (4.2.17)
Applying the initial and boundary conditions u(ξ, 0) = ξ, u(0, η) = η − 1, we gets the specific
solutions.

js, ddw de ii: Lecture Notes Page 172 of 249


4.2. METHOD OF CHARACTERISTIC EQUATIONS - - FIRST
ORDER PDES

Case i)

Looking at base curves x(y) which pass through initial points u(ξ, 0) = ξ ⇒ x(0) = ξ. From
(4.2.16) and (4.2.17), we must have

ξ = c1 − 1 ⇒ c1 = ξ + 1
ξ = c2 (1) + (0) − 1 ⇒ c2 = ξ + 1

Therefore,
x − ey + 1
x = c1 ey − 1 = (ξ + 1)ey − 1 ⇒ ξ =
ey
x − ey + 1
 
−y −y
u = c2 e + y − 1 = (ξ + 1)e + y − 1 = + 1 e−y + y − 1 = (x + 1)e−2y + y − 1
ey

Case ii)

Applying the boundary conditions u(0, η) = η − 1, we have x = 0 when y = η that is


u(0, η) = η − 1, x(0) = η to have

c1 = e−η
x+1
e−η =
ey
c2 = 0
u(x, y) = y − 1

The difference between the two cases, (x + 1)e−2y , corresponds to the discontinuity in u prop-
agating along the curve x = ey − 1.
Example 4.2.28 Using the method of characteristics, solve the first order equation
∂u ∂u
2y + (3x2 − 1) =0
∂x ∂y
We first divide the PDE by 2y obtaining
∂u 3x2 − 1 ∂u
+ =0
∂x 2y ∂y
Consider u along a curve y = y(x) (the independent variable should be the first partial deriva-
tive, whose coefficient is 1 in the PDE). On this curve we have
d ∂u ∂u dy
u(x, y(x)) = +
dx ∂x ∂y dx
such that the ordinary differential equations
du
= 0
dx
dy 3x2 − 1
=
dx 2y
Then on solutions of
du
=0 ⇔ u=A
dx
js, ddw de ii: Lecture Notes Page 173 of 249
CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

and
dy 3x2 − 1
= ⇔ 2ydy = (3x2 − 1)dx ⇒ y 2 = x3 − x + B
dx 2y
Hence
u(x, y) = φ y 2 − x3 + x


Example 4.2.29 Solve


∂u ∂u
+x = u.
∂x ∂y
Consider u along a curve y = y(x) (the independent variable should be the first partial deriva-
tive, whose coefficient is 1 in the PDE). On this curve we have
d ∂u ∂u dy
u(x, y(x)) = +
dx ∂x ∂y dx
such that the ordinary differential equations
du
= u
dx
dy
= x
dx
Then on solutions of
du
= u ⇔ u = Aex ; ⇒ A = ue−x
dx
and
dy x2 x2
=x ⇔ y= +B ⇒ B =y−
dx 2 2
Thus
x2
 
−x
ue = φ y −
2
Such that
x2
 
u(x, y) = φ y − ex
2
Exercise 4.14 Burger’s Equation. Solve the Cauchy problem

 ut + uux = 0,

u(x, 0) = x.

Exercise 4.15 Solve the first order equation


∂u ∂u
x2 +y + xyu = 1.
∂x ∂y
Exercise 4.16 Solve the constant coefficient equation
∂u ∂u
a +b + cu = 0.
∂x ∂y
Exercise 4.17 Solve the partial differential equation
∂u ∂u
2 +3 + 8u = 0, u(x, 0) = sin x
∂x ∂y
2y 8
e− 3 y

u(x, y) = sin x − 3

js, ddw de ii: Lecture Notes Page 174 of 249


4.2. METHOD OF CHARACTERISTIC EQUATIONS - - FIRST
ORDER PDES

Exercise 4.18 Solve the partial differential equation


∂u ∂u
x −y = u, with u = x2 on y = x, 1 ≤ y ≤ 2
∂x ∂y

u(x, y) = x xy
Exercise 4.19 Solve the particular advection (transport) equation
∂u 1 ∂u
+ 2 =0
∂t x + 1 ∂x
Exercise 4.20 Solve
ut − xux = 0
Exercise 4.21 Show how we may apply the Method of Characteristics to a real and given
problem as for a quasi-linear PDE below
1
uux1 + ux2 = 1 with u(x1 , x1 ) = x1 .
2

4.2.2 Quasilinear Equations Method of Characteristics


Consider the first order quasilinear PDE
∂u ∂u
a(x, y, u) + b(x, y, u) = c(x, y, u) (4.2.18)
∂x ∂y
where the functions a, b and c can involve u but not its derivatives.

The dependent variable can be written as u = u(x(t), y(t)) in terms of a dummy variable t. On
the characteristic curve we have
d ∂u dx ∂u dy
u(x(t), y(t)) = +
dt ∂x dt ∂y dt
which is
du ∂u dx ∂u dy
= + (4.2.19)
dt ∂x dt ∂y dt
Comparing (4.2.18) and (4.2.19) will generate a set of ordinary differential equations
dx
= a(x, y, u)
dt
dy
= b(x, y, u)
dt
du
= c(x, y, u)
dt
Note 4.2.1 Note that, for a linear first order PDE, the dummy variable t is one of the inde-
pendent variable to have (4.2.19) for u = u(x, y(x)) written as (where t = x)
du ∂u ∂u dy
= + (4.2.20)
dx ∂x ∂y dx
or for u = (x(y), y) to get (where t = y)
du ∂u ∂u dx
= + (4.2.21)
dy ∂y ∂x dy
that have been used in the previous section 4.2

js, ddw de ii: Lecture Notes Page 175 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

Example 4.2.30 Solve the partial differential equation


∂u ∂u
(y − u) + (x − y) = u−x
∂x ∂y

Comparing with (4.2.19)

du ∂u dx ∂u dy
= +
dt ∂x dt ∂y dt
The system ode is given by
dx
= y−u (4.2.22)
dt
dy
= x−y (4.2.23)
dt
du
= u−x (4.2.24)
dt
1). Adding (4.2.22), (4.2.23) and (4.2.24)

d
(x + y + u) = 0
Z dt Z
d
(x + y + u)dt = 0 dt
dt
(x + y + u) = A (4.2.25)

2).

d du dy
(yu) = y +u
dt dt dt
d d
(yu) = y(u − x) + u(x − y) ⇒ (yu) = −x(y − u)
dt dt
d dx
(yu) = −x
Z dt Z dt Z Z
d dx d
(yu)dt = −x dt ⇒ (yu)dt = −xdx
dt dt dt
x2
(yu) = − + B
2
to have

(2yu) + x2 = B (4.2.26)

From (4.2.25) and (4.2.26) we get the general solutions

(x + y + u) = φ 2yu + x2


js, ddw de ii: Lecture Notes Page 176 of 249


4.2. METHOD OF CHARACTERISTIC EQUATIONS - - FIRST
ORDER PDES

Example 4.2.31 Solve the partial differential equation


∂u ∂u
(y + u) +y = x−y
∂x ∂y
Comparing with (4.2.19)
du ∂u dx ∂u dy
= +
dt ∂x dt ∂y dt
The characteristic equations are
dx
= y+u (4.2.27)
dt
dy
= y (4.2.28)
dt
du
= x−y (4.2.29)
dt
1). Adding (4.2.27) and (4.2.29)
d
(x + u) = x + u
dt
and from equation (4.2.28)
1 dy
= 1
y dt
Now, consider
 
d x+u 1d x + u dy
= (x + u) −
dt y y dt y 2 dt
 
d x+u x+u x+u
= −
dt y y y
 
d x+u
= 0
dt y

 
x+u
= c1 (4.2.30)
y
2). Also, equations (4.2.27) and (4.2.28) give
d
(x − y) = u,
dt
and equation (4.2.29)
d du
(x − y) (x − y) = u .
dt dt
Now, consider
d  d   d
(x − y)2 − u2 = (x − y)2 − (u2 )

dt dt dt
d  d du
(x − y)2 − u2 = 2(x − y) (x − y) − 2u

dt dt dt
d 
(x − y)2 − u2 = 0

dt

(x − y)2 − u2 = c2 (4.2.31)

js, ddw de ii: Lecture Notes Page 177 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

From (4.2.30) and (4.2.31) we get the general solutions


 
x+u 2 2
F , (x − y) − u = 0
y
or  
2 2 x+u
(x − y) − u = φ
y
Now to get a particular solution, lets apply the initial conditions
u = 1 + x, when y = 1
This means that, at
y = 1, x(1) = ξ, ⇒ u(1) = 1 + x(1) = 1 + ξ
From (4.2.31)
c2 = (x − y)2 − u2
= (x(1) − 1)2 − u(1)2
= (ξ − 1)2 − (1 + ξ)2
= −4ξ
To have
(x − y)2 − u2 = −4ξ (4.2.32)
Also from (4.2.30)
 
x+u
c1 =
y
 
x(1) + u(1)
= = 2ξ + 1
1
To generate
 
x+u
= 2ξ + 1 (4.2.33)
y
From (4.2.33)
 
x+u−y
ξ =
2y
Thus from (4.2.32) the solutions as
2
(x − y)2 − u2 = (y − x − u)
y
We can rewrite this as a quadratic equation for u
 
2 2 x−y 2
u − u− 2 + (x − y) = 0
y y
 2
2 2 1 1
u − u− x−y+ + 2 = 0
y y y
Then, s  2  
1 1 1 1 1 1
u= ± 2
+ x−y+ − 2 = ± x−y+
y y y y y y

js, ddw de ii: Lecture Notes Page 178 of 249


4.2. METHOD OF CHARACTERISTIC EQUATIONS - - FIRST
ORDER PDES

Example 4.2.32 Solve the partial differential equation


∂u ∂u
x(y − u) + y(x + u) = (x + y)u
∂x ∂y

Comparing with (4.2.19)

du ∂u dx ∂u dy
= +
dt ∂x dt ∂y dt
The characteristic equations are
dx
= x(y − u) (4.2.34)
dt
dy
= y(x + u) (4.2.35)
dt
du
= (x + y)u (4.2.36)
dt
1). Equations (4.2.34) and (4.2.35) give

dx dy
y +x = xy 2 − xyu + yx2 + xyu
dt dt
dx dy
y +x = xy(x + y)
dt dt
dx dy 1 du
y +x = xy
dt dt u dt
using (4.2.36).

Now, consider
1 d 1 dy 1 du
+ =
x dt y dt u dt

d  xy 
ln = 0
dt u

xy
= c1 (4.2.37)
u

2). On the other hand,

dx dy du
− = xy − xu − xy − yu = −u(x + y) = −
dt dt dt

d
(x + u − y) = 0
dt

x + u − y = c2 (4.2.38)

js, ddw de ii: Lecture Notes Page 179 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

From (4.2.37) and (4.2.38) we get the general solutions


xy
= φ (x + u − y)
u
Now to get a particular solution, lets apply the initial conditions

u = x2 + 1, on y = x

This means that, at

y(0) = ξ, x(0) = ξ, ⇒ u(0) = x(0)2 + 1 = ξ 2 + 1

From (4.2.38)

c2 = x + u − y
= ξ + ξ2 + 1 − ξ


= ξ2 + 1

To have

x + u − y = ξ2 + 1 (4.2.39)

Also from (4.2.37)


xy
c1 =
u
ξξ
= 2
ξ +1
ξ2
= 2
ξ +1
To generate

xy ξ2
= 2 (4.2.40)
u ξ +1

Using (4.2.39) into (4.2.40)

xy ξ2
= 2
u ξ +1
xy x+u−y−1
=
u x+u−y
So, finally the solution is
xy x+u−y−1
= . Rearrange to finish!
u x+u−y
Exercise 4.22 Solve the equation
dx dy du
= =
x(y 2 2
−u ) 2 2
−y(u + x ) u(x + y 2 )
2

x2 + y 2 + u2 = c1 , yu = xc2 .

js, ddw de ii: Lecture Notes Page 180 of 249


4.2. METHOD OF CHARACTERISTIC EQUATIONS - - FIRST
ORDER PDES

Example 4.2.33 Solve the characteristic equations


∂u ∂u
x2 +u = 1, with u = 0 on x + y = 1
∂x ∂y
Comparing with (4.2.19)
du ∂u dx ∂u dy
= +
dt ∂x dt ∂y dt
The characteristic equations are
dx
= x2 (4.2.41)
dt
dy
= u (4.2.42)
dt
du
= 1 (4.2.43)
dt
The Ordinary differential equations (the characteristics) (4.2.41) to (4.2.43) can be easily solved
to have
1
x =
c1 − t
t2
y = + c2 t + c3
2
u = c2 + t

Remark 4.2.2 Equation (4.2.43) was solved before (4.2.42) in order to have a u to substitute
into equation (4.2.42).

Parameterising the initial conditions,

At t = 0, y(0) = ξ, x(0) = 1 − ξ, u(0) = 0

Remark 4.2.3 When solving such characteristics directly, we will need not only to eliminate
the constants, but also the parameter t.
1
c1 =
1−ξ
c2 = 0
c3 = ξ

To have
u2 u2
 
t = u, ⇒ y = + ξ, ⇒ ξ = y−
2 2
 
u2
1−ξ 1− y− 2
x = = u2
1 − u (1 − ξ)

1−u 1− y− 2
3 2
u u
x − ux + uyx − x = 1 − y +
2 2
To have the final solution as
u2 x
y= +1−
2 1 + ux

js, ddw de ii: Lecture Notes Page 181 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

Example 4.2.34 Alternative approach to Example 4.2.31.


∂u ∂u
(y + u) +y = x−y
∂x ∂y
The characteristic equations are
dx
= y+u (4.2.44)
dt
dy
= y (4.2.45)
dt
du
= x−y (4.2.46)
dt
Solve with respect to the dummy variable t; (4.2.45) gives,
y = c1 et ,
(4.2.44) and (4.2.46) give
d
(x + u) = x + u ⇒ x + u = c2 et ,
dt
and (4.2.44) give
dx
= c1 et + c2 et − x,
dt
so,
1
x = c3 e−t + (c1 + c2 ) et
2
−t 1
u = −c3 e + (c2 − c1 )et
2
Now, at s = 0, y = 1 and x = ξ, u = 1 − ξ (parameterising initial line Γ),
c1 = 1, c − 2 = 1 + 2ξ and c3 = −1
Hence, the parametric form of the surface integral is,
x = −e−t + (1 + ξ)et , y = et , and u = e−t + ξet .
Then eliminate ξ and t:
 
1 1 1
x = − + (1 + ξ)y ⇒ ξ = x−y+
y y y
so  
1 1 1
u= + x−y+ y.
y y y
Finally,
2
u=x−y+
y
as before.

To find the invariants, return to solved characteristics equations and solve for constants in
terms of x, y and u. We only need two, so put for instance c1 = 1 and so y = et . Then,
c3 1 c3 1
x= + (1 + c2 )y and u = − + (c2 − 1)y
y 2 y 2
Solve for c2 and c3
x+u x+u 1
c2 = ⇒ φ= . c3 = (x − u − y)y ⇒ ψ = (x − u − y)y
y y 2

js, ddw de ii: Lecture Notes Page 182 of 249


4.2. METHOD OF CHARACTERISTIC EQUATIONS - - FIRST
ORDER PDES

Remark 4.2.4 Solving the characteristic equations-two approaches.


1. Manipulate the equations to get them in a ‘directly integrable’ form such as
1 d
(x + u) = 1
x + u dt
and find some combination of the variable which differentiates to zero, such as
 
d x+u
= 0.
dt y

2. Solve the equation with respect to the dummy variable t, and apply the initial data (param-
eterised by ξ) at t = 0. Eliminate ξ and t; find invariants by solving for constants.

Exercise 4.23 Solve the non-linear wave equation


∂u ∂u
+ (u + c) = 0
∂t ∂x
The characteristics are defined by
dt dx du
= 1, = c + u, and =0
ds ds ds
To have
φ = u, and ψ = x − (u + c)t
Exercise 4.24 Solve the Cauchy problem

ut + ux − 3u = t, −∞ < x < ∞, t > 0


u(x, 0) = x2 . −∞ < x < ∞.

Exercise 4.25 Solve the following initial value problem using method of characteristics

x2 ∂u
∂x
+ y 2 ∂u
∂y
= u2 ,
u(x, 2x) = 1

The characteristic equations are


∂x ∂y ∂u
= x2 , = y2, = u2
∂t ∂t ∂t
with initial conditions
x(0, ξ) = ξ, y(0, ξ) = 2ξ, u(0, ξ) = 1.
The solutions of the system are given by
ξ 2ξ 1
x(t, ξ) = , y(t, ξ) = , u(t, ξ) = .
1 − ξt 1 − 2ξt 1−t
Solving the first two equations above for t in terms of x and y gives
y − 2x
t=
xy
so that
1 xy
u(x, y) = =
1−t 2x − y + xy

js, ddw de ii: Lecture Notes Page 183 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

4.2.3 Fully Non-linear Equations Method of Characteristics


Definition 4.2.2 We consider fully nonlinear equation of the first order

F (x, y, u, ux , uy ) = 0 (4.2.47)

together with the initial value problem

u(x0 (ξ), y0 (ξ)) = uξ (ξ).

We denote the coordinates of F as (x, y, z, p, q). As before we want to find values of u by


calculating them along curves (x(t), y(t)).

To find (x(t), y(t)), we abbreviating α(t) = (x(t), y(t)) and assuming that we have a solution u
of (4.2.47). We let

z(t) = u (α(t)) = u (x(t), y(t)) , p(t) = ux (α(t))) , and q(t) = uy (α(t)) .

Then differentiating p and q we get

p0 (t) = uxx (α(t)) · x0 (t) + uxy (α(t)) · y 0 (t)


(4.2.48)
q 0 (t) = uxy (α(t)) · x0 (t) + uyy (α(t)) · y 0 (t).

Next we differentiate the equation (4.2.47) with respect to x and then with respect to y to get

Fx + Fz · ux + Fp · uxx + Fq · uyx = 0
Fy + Fz · uy + Fp · uxy + Fq · uyy = 0

so that abbreviating γ(t) = (x(t), y(t), z(t), p(t), q(t)) we have

Fx (γ(t)) + Fz (γ(t)) ux (α(t)) + Fp (γ(t)) uxx (α(t)) + Fq (γ(t)) uyx (α(t)) = 0

and
Fy (γ(t)) + Fz (γ(t)) uy (α(t)) + Fp (γ(t)) uxy (α(t)) + Fq (γ(t)) uyy (α(t)) = 0
To eliminate the second partial derivatives uxx , uxy , uyy from the last two equations we set

x0 (t) = Fp (γ(t)) and y 0 (t) = Fq (γ(t))

so that, using equation in (4.2.48), the above equations become

Fx (γ(t)) + Fz (γ(t)) · p(t) + p0 (t) = 0

and
Fy (γ(t)) + Fz (γ(t)) · uy (α(t)) + q 0 (t) = 0.
Hence the equations for p and q are

p0 (t) = −Fx (γ(t)) − Fz (γ(t)) · p(t)


(4.2.49)
q 0 (t) = −Fy (γ(t)) − Fz (γ(t)) · q(t).

Finally, differentiating z(t) = u(α(t)), we get

z 0 (t) = ux (α(t)) · x0 (t) + uy (α(t)) · y 0 (t)


= Fp (γ(t)) · p(t) + Fq (γ(t)) · q(t).

js, ddw de ii: Lecture Notes Page 184 of 249


4.2. METHOD OF CHARACTERISTIC EQUATIONS - - FIRST
ORDER PDES

We have arrived at the system of characteristic equations,

x0 (t) = Fp (γ(t))
y 0 (t) = Fq (γ(t))
z 0 (t) = Fp (γ(t)) · p(t) + Fq (γ(t)) · q(t) (4.2.50)
p0 (t) = −Fx (γ(t)) − Fz (γ(t)) · p(t)
q 0 (t) = −Fy (γ(t)) − Fz (γ(t)) · q(t)

To incorporate initial condition we write x(t, ξ), y(t, ξ), z(t, ξ)), p(t, ξ), q(t, ξ) to denote solu-
tions of (4.2.50) which satisfy the initial condition

x(0, ξ) = x0 (ξ), y(0, ξ) = y0 (ξ), z(0, ξ) = u0 (ξ), p(0, ξ) = p0 (ξ), q(0, ξ) = q0 (ξ)

at t = 0. The initial values for x, y and z are given by the initial curve Γ(ξ). However, the
initial values p0 (ξ) and q0 (ξ) for p and q have to be found. They can’t be prescribed arbitrarily
and they must satisfy the following condition. First they satisfy the equation,

F [x0 (ξ), y0 (ξ), u0 (ξ), p0 (ξ), q0 (ξ)] = 0. (4.2.51)

Further, since u0 (ξ) = u(x0 (ξ), y0 (ξ)), we get after differentiating this with respect to ξ we get

u̇0 (ξ) = ux [x0 (ξ), y0 (ξ)] · ẋ0 (ξ) + uy [x0 (ξ), y0 (ξ)] · ẋ0 (ξ)

which shows that p0 (ξ) and q0 (ξ) should satisfy

p0 (ξ) · ẋ0 (ξ) + q0 (ξ) · ẏ0 (ξ) = u̇0 (ξ). (4.2.52)

If the curves p0 (ξ) and q0 (ξ) satisfy (4.2.51)-(4.2.52), we say that the initial data

(x0 (ξ), y0 (ξ), u0 (ξ), p0 (ξ), q0 (ξ))

are compatible with the PDE (4.2.47). Having solve the system of characteristic equation
(4.2.50) with the compatible initial condition, we have to invert the map

(t, ξ) 7→ [x(t, ξ), y(t, ξ)]

and express t and ξ as functions of x and y, t = t(x, y) and ξ = ξ(x, y). If this can be done,
then the solution u is given by

u(x, y) = z [t(x, y), ξ(x, y)] .

Exercise 4.26 Solve the following initial vale problem


1

2
u2x + u2y + (ux − x)(uy − y) = u
u(x, 0) = 0

Let F (x, y, u, p, q) = 12 (p2 + q 2 ) + (p − x)(q − y) − u. The characteristic equations are

∂x
= Fp = p + q − y
∂t
∂y
= Fq = p + q − x
∂t
∂u
= pFp + qFq = p(p + q − y) + q(p + q − x)
∂t
js, ddw de ii: Lecture Notes Page 185 of 249
CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

∂p
= p+q−y
∂t
∂q
= p+q−x
∂t
The initial conditions for x, y, u are
x(0, ξ) = ξ, y(0, ξ) = 0, u(0, ξ) = 0.
The initial condition p0 (ξ) and q0 (ξ) for p and q satisfy
1 2
p0 (ξ) + q02 (ξ) + [p0 (ξ) − ξ] q0 (ξ) = 0

2
p0 (ξ) · 1 + q0 (ξ) · 0 = 0
So, p0 = 0 and either q0 = 0 or q0 (ξ) = 2ξ. Consequently, we have two sets of initial conditions:
x(0, ξ) = ξ, y(0, ξ) = 0, u(0, ξ) = 0, p0 (ξ) = 0, q0 (ξ) = 2ξ
and
x(0, ξ) = ξ, y(0, ξ) = 0, u(0, ξ) = 0, p0 (ξ) = 0, q0 (ξ) = 0.
We consider only the first set of the initial conditions. From the equations we get
(x − p)0 = 0 and (y − q)0 = 0
which after integration gives
x = ξ + p and y = q − 2ξ.
Adding the fourth and the fifth equation and then subtracting the first gives
(p + q − x)0 = p + q − x
at the initial condition (p + q − x)(0, ξ) = ξ so that
p + q − x = ξet .
Adding the fourth and the fifth equation and subtracting the second gives
(p + q − y)0 = p + q − y, (p + q − y)(0, ξ) = 2ξ
which gives
p + q − y = 2ξet .
Hence
x = ξ(e2t − 1)
y = ξ(et − 1)
p = 2ξ(et − 1)
q = ξ(et + 1)
Substituting to the third equation and integrating gives
5
u(t, ξ) = ξ 2 e2t − 1 − 3ξ 2 (et − 1)

2
Solving the first two equations above for t and ξ in terms of x and y, one gets
y−x
et = and ξ = x − 2y
2y − x
Consequently,
y(4x − 3y)
u(x, y) = u (t(x, y), ξ(x, y)) = .
2
js, ddw de ii: Lecture Notes Page 186 of 249
4.2. METHOD OF CHARACTERISTIC EQUATIONS - - FIRST
ORDER PDES

Example 4.2.35 Solve the initial value problem

ux uy − u = 0, (x, y) ∈ (0, ∞) × R, u(0, y) = y 2 , y ∈ R.

Here the initial curve Γ(ξ) = (0, ξ, ξ 2 ) and F (x, y, z, p, q) = pq − z. So the system of character-
istic equations is as follows,

x0 = q
y0 = p
z0 = 2pq
p0 = p
q0 = q

Here x = x(t, ξ), y = y(t, ξ), z = z(t, ξ), p = p(t, ξ) and q = q(t, ξ), and ’ stands for the
derivative with respect to t. We have to find the initial condition p0 (ξ), q0 (ξ) for p(t, ξ) and
q(t, ξ). The curve (p0 (ξ), q0 (ξ)) has to satisfy the equations

F (x0 (ξ), y0 (ξ), z0 (ξ), p0 (ξ), q0 (ξ)) = 0 i.e., p0 (ξ)q0 (ξ) = ξ 2

and
u00 (ξ) = p0 (ξ)ẋ0 (ξ) + q0 (ξ)ẏ0 (ξ) i.e., 2ξ = q0 (ξ).
ξ
Hence q0 (ξ) = 2ξ and p0 (ξ) = 2
and full set of the initial condition for the system of character-
istic equations above is
ξ
x(0, ξ) = 0, y(0, ξ) = ξ, z(0, ξ) = ξ 2 , p(0, ξ) = , q(0, ξ) = 2ξ.
2
Solving the system for p and q we find that
ξ
p(t, ξ) = et , q(t, ξ) = 2set .
2
Then z 0 = 2ξ 2 e2t so that
z(t, ξ) = ξ 2 e2t .
Finally, x0 (t, ξ) = 2ξet and y 0 (t, ξ) = ξ 2 et together with the initial condition give

ξ t
x(t, ξ) = 2ξ et − 1 , y(t, ξ) =
  
e +1 .
2
ξ
Next we have to solve the system x = 2ξ [et − 1] and y = 2
[et + 1] for (t, ξ) in terms of (x, y).
After this the solution u is given by
2
u(x, y) = z(t, ξ) = ξ 2 e2t = ξet .

So we only have to find set. The calculation shows that


x + 4y
ξet =
4
so that
(x + 4y)2
u(x, y) = .
16

js, ddw de ii: Lecture Notes Page 187 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

Example 4.2.36 Find the solution of the eikonal equation

u2x + u2y = 1

satisfying u = 0 on the circle x2 + y 2 = 1. The initial curve Γ(ξ) is given by

Γ(ξ) = (x0 (ξ), y0 (ξ), u0 (ξ)) = (cos ξ, sin ξ, 0).

Moreover, let
F (x, y, z, p, q) = p2 − q 2 − 1.
Then if u is a solution of the eikonal equation, we have F (x, y, u, ux , uy ) = 0. Hence the system
of characteristic equations is given by

x0 = Fp = 2p
y0 = Fq = 2q
z0 = pFp + qFq = 2p2 + 2q 2
p0 = −Fx − pFz = 0
q0 = −Fy − qFz = 0,

where x = x(t, ξ), y = y(t, ξ), z = z(t, ξ), p = p(t, ξ), and q = q(t, ξ). The initial condition for
solutions x, y, and z are

x(0, ξ) = cos ξ, y(0, ξ) = sin ξ, z(0, ξ) = 0.

If the initial conditions (p0 (ξ), q0 (ξ)) for p and q should satisfy

F (x0 (ξ), y0 (ξ), u0 (ξ), p0 (ξ), q0 (ξ)) = 0,

which gives
p0 (ξ)2 + q0 (ξ)2 = 1
and
u00 (ξ) = p0(ξ) · x00 (ξ) + q0 (ξ) · y00 (ξ),
which gives
0 = p0 (ξ) · (− sin ξ) + q0 (ξ) · cos ξ.
Geometrically, by the first equation the point (p0 (ξ), q0 (ξ)) lies on the unit circle and the
second equation means that the dot product

(p0 (ξ), q0 (ξ)) · (− sin ξ, cos ξ) = 0

so that the vector (p0 (ξ), q0 (ξ)) is perpendicular to the vector (− sin ξ, cos ξ). It follows that
there are two sets of solutions (p0 (ξ), q0 (ξ)), namely,

p0 (ξ) = cos ξ, q0 (ξ) = sin ξ

and
p0 (ξ) = − cos ξ, q0 (ξ) = − sin ξ.
Consider the first case. Then the initial condition for the characteristic system above is

x(0, ξ) = cos ξ, y(0, ξ) = sin ξ, z(0, ξ) = 0, p0 (ξ) = cos ξ, q0 (ξ) = sin ξ.

Then
p(t, ξ) = cos ξ and q(t, ξ) = sin ξ.

js, ddw de ii: Lecture Notes Page 188 of 249


4.2. METHOD OF CHARACTERISTIC EQUATIONS - - FIRST
ORDER PDES

Using theses two solutions we get


x(t, ξ) = 2t cos ξ + cos ξ = (2t + 1) cos ξ,
y(t, ξ) = t sin ξ + sin ξ = (2t + 1) sin ξ,
z(t, ξ) = 2t.
We have to solve the equations x = (2t + 1) cos ξ and y = (2t + 1) sin ξ for (t, ξ). Note that
x2 + y 2 = (2t + 1)2
so that p
|2t + 1| = x2 + y 2
But 
 2t + 1, t ≥ − 21
|2t + 1| =
−2t − 1, t < − 12

Since our solution has to satisfy the initial condition, we must take consider t ≥ −1/2. So
p
2t + 1 = x2 + y 2 .
Then p
u(x, y) = z [t(x, y), ξ(x, y)] = 2t(x, y) = −1 + x2 + y 2 .
Let us consider the case p0 (ξ) = − cos ξ, q0 (ξ) = − sin ξ. Then the characteristic curves are
x(t, ξ) = −2t cos ξ + cos ξ = (1 − 2t) cos ξ,
y(t, ξ) = −2t sin ξ + sin ξ = (1 − 2t) sin ξ,
z(t, ξ) = 2t,
p(t, ξ) = − cos ξ,
q(t, ξ) = − sin ξ.
Set x = (1 − 2t) cos ξ and y = (1 − 2t) sin ξ. Then
x2 + y 2 = (1 − 2t)2 and
p
x2 + y 2 = |1 − 2t|.

 1 − 2t, t ≤ 21
|1 − 2t| =
−1 + 2t, t > 12

1
and the interval t > 2
doesn’t contain t = 0, we see that
p
x2 + y 2 = 1 − 2t.
So, p
u(x, y) = z [t(x, y), ξ(x, y)] = 2t(x, y) = 1 − x2 + y 2
the general solutions for the fully non-linear first order PDE.
Remark 4.2.5 z can be replaced by u straight from the start of derivations.

4.2.4 Method of Characteristic Equations / Canonical Forms - -


Second Order PDEs
The method of characteristics which is also commonly referred to as a technique into canonical
forms has been left out in this volume but considerably handled in last year lecture notes.

js, ddw de ii: Lecture Notes Page 189 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

4.3 Method of Separation of Variables - - Second Order


PDEs
For the method of separation of variables, also known as the Fourier method for

Auxx + Buxy + Cuyy + Dux + Euy + F u = G

We let the solution u(x, y) to be

u(x, y) = X(x)Y (y) (4.3.1)

a product of two solutions and substitute into the general equation

A PDE in n independent variables is reduced to n ODEs - Separation of variables.

Step I Generate the partial derivatives to substitute back into the PDE

ux = X 0Y
uxx = X 00 Y
uy = XY 0
uxy = X 0Y 0
uyy = XY 00

Step II Separate the variables, X on one side and Y on the other and equal to a constant.
Solve the odes in one variables.

Step III Substitute back solutions for X(x) and Y (y) into the general solutions

u(x, y) = X(x)Y (y)

Step IV Apply the principal of superposition to sum the solutions as the general solutions.

Step V Use the Boundary and Initial conditions. Apply Fourier series or Fourier Sine se-
ries or Fourier cosine series (depending on the interval and the u(x, t) generated) to
determine the coefficients.

Step VI Rewrite the general solutions.

js, ddw de ii: Lecture Notes Page 190 of 249


4.3. METHOD OF SEPARATION OF VARIABLES - - SECOND
ORDER PDES

4.3.1 Homogeneous Heat Equation


The general heat equation of an insulated wire or conductor

in one dimension is of the form

ut = kuxx , 0<x<L
u(0, t) = u(L, t) = 0 , t≥0 (4.3.2)
u(x, 0) = f (x) , 0<x<L (4.3.3)

Since a PDE in (x, t) not (x, y), by separation of variables, we assume a solution

u(x, t) = X(x)T (t) = XT (4.3.4)

where X(x) is a solution and T (t) is also a solution that satisfy the original problem.

ux = X 0 T
uxx = X 00 T
ut = XT 0

Substituting in the general heat equation given,

ut = kuxx
XT 0 = kX 00 T
X 00 T0
=
X kT
Note 4.3.1 For easy computations, all constants must be on the side of T s not Xs.
Note 4.3.2 Since the left hand side is a function of x alone and the right hand side is a function
of t alone, this is impossible (we could change the value of x, at a fixed time, and the left hand
side would have a different value but the right hand side would remain the same) unless they
are both equal to a constant say −α2 (the selection of the constant is special for the nature
of equation to generate, if choose a positive constant, the solution might not exist).. Thus, we
have
X 00 T0
= = − α2
X kT
to imply
X 00
= −α2 (4.3.5)
X
T0
= −α2 (4.3.6)
kT
Solving the two equations for X(x) and T (t), From Equation (4.3.5)
X 00
= −α2
X
js, ddw de ii: Lecture Notes Page 191 of 249
CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

X 00 + α2 X = 0
X(x) = A cos αx + B sin αx

Lets now concentrate on X(x). The boundary conditions (4.3.2) imply that

u(0, t) = X(0)T (t) = 0 and u(L, t) = X(L)T (t) = 0

If we let T (t) = 0, then we will obtain the solution u(x, t) = 0 for all t. This is called the trivial
solution since it is the solution corresponding to the string not moving at all. To avoid the
trivial solution, we thus assume that

X(0) = 0 and X(L) = 0 (4.3.7)

using X(0) = 0

X(x) = A cos αx + B sin αx


0 = A cos 0 + B sin 0 ⇒ A = 0
⇒ X = B sin αx

Using X(L) = 0

X(x) = B sin αx

0 = B sin αL ⇒ αL = nπ ⇒ α =
L


Xn (x) = Bn sin x (4.3.8)
L
Using Equation (4.3.6)
T0 dT 2
= −α2 ⇔ = −α2 kdt ⇔ T (t) = Ce−α kt
kT T

nπ 2
Tn (t) = Cn e−( L ) kt
(4.3.9)

using the principle of superposition with



X
u(x, t) = X(x)T (t)
n=1

nπ 2 nπ
bn e−( L )
X
kt
u(x, t) = sin x (4.3.10)
n=1
L

Remark 4.3.1 If we used a positive scalar, say λ > 0, however, then temperature T (t) = Ceλkt
would grow to ∞ which is not physically possible (Negative is ok, since exponential bounded
by zero below). Thus, we assume that λ is negative, which is to say that λ = −α2 for some
number α. Another reason for Note 4.3.2 is to develop a sine-cosine solutions of X which is an
orthogonal set whose coefficients can easily be by Fourier series.

Using the initial condition u(x, 0) = f (x)



X  nπ 
u(x, 0) = bn sin x = f (x)
n=1
L

js, ddw de ii: Lecture Notes Page 192 of 249


4.3. METHOD OF SEPARATION OF VARIABLES - - SECOND
ORDER PDES

This is a Fourier ”Sine” series form with


Z L
2  nπ 
bn = f (x) sin x dx (4.3.11)
L 0 L
The general solution will be
∞  Z L 
X 2  nπ 
−( nπ )
2
kt
 nπ 
u(x, t) = f (x) sin x dx e L sin x (4.3.12)
n=1
L 0 L L

Note 4.3.3 We solve for X(x), α = L
, T (t), bn and finally the general solution.

Review: Solutions of Second order constant coefficient, homogeneous


Ordinary differential equations
Consider a second order homogeneous linear differential equation with constant coefficients

ay 00 + by 0 + cy = 0 (4.3.13)

where a, b, c are all real constants; a 6= 0. Let us denote by m2 = y 00 , m = y 0 , m0 = y. Then


equation (4.3.13) is
am2 + bm + c = 0 (4.3.14)
Definition 4.3.1 Equation (4.3.14) is commonly known as the characteristic ( auxiliary )
equation of the differential equation (4.3.13). The roots, r1 and r2 of the equation (4.3.14) are
called characteristic (or auxiliary) roots of the differential equation.

Equation (4.3.14) is clearly a quadratic equation and in principle, if r1 and r2 are roots of such
an equation we can write (4.3.14) as

m2 − (r1 + r2 )m + r1 r2 = 0. (4.3.15)

If the auxiliary roots r1 and r2 are real and distinct


The general solution is
y = c1 er1 x + c2 er2 x

If the auxiliary roots r1 and r2 are real and equal/repeated


The general solution is
y = c1 erx + c2 xerx

If the auxiliary roots are complex conjugates


Let

r1 = α + iβ
r2 = α − iβ

be two complex conjugate roots of the characteristic equation am2 + bm + c = 0. Then general
solution is then
y = eαx (A cos βx + B sin βx).

js, ddw de ii: Lecture Notes Page 193 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

Example 4.3.1 Find the formal solution to the given initial-boundary value problem.

ut = 5uxx 0<x<1
u(0, t) = u(1, t) = 0 t≥0
u(x, 0) = (1 − x)x2 0<x<1

By separation of variables

u = XT, ux = X 0 T, uxx = X 00 T, ut = XT 0
X 00 T0
= = − α2
X 5T

X 00
= −α2
X
X 00 + α2 X = 0
X(x) = A cos αx + B sin αx

using the boundary conditions

X(0, t) = 0 ⇒ X(0) = 0 ⇒ A = 0

⇒ X(x) = B sin αx
using the other boundary condition

u(1, t) = 0 ⇒ X(1) = 0 ⇒ B sin α = 0 ⇒ sin α = 0 ⇒ α = nπ

Xn (x) = Bn sin nπx


T0 2 2
= −α2 ⇒ T (t) = Ce−α 5t = Cn e−(nπ) 5t
5T
2
Tn (t) = Cn e−(nπ) 5t

The solution will be

u(x, t) = T (t)X(x)
X∞
u(x, t) = T (t)X(x)
n=1

X 2
u(x, t) = Cn e−(nπ) 5t Bn sin nπx
n=1

X 2
u(x, t) = bn e−(nπ) 5t
sin nπx
n=1

Using the initial conditions

u(x, 0) = (1 − x)x2
X∞
2
(1 − x)x = bn sin nπx
n=1

js, ddw de ii: Lecture Notes Page 194 of 249


4.3. METHOD OF SEPARATION OF VARIABLES - - SECOND
ORDER PDES

and by the Fourier sine series (the equation above is in that form)
2 L 2 1 [−8(−1)n − 4]
Z  nπ  Z
bn = f (x) sin x dx = (1 − x)x2 sin nπx dx =
L 0 L 1 0 (nπ)3
Thus the general solution is given by

X
u(x, t) = bn T (t)X(x)
n=1

X [−8(−1)n − 4] 2
u(x, t) = e−(nπ) 5t
sin nπx
n=1
(nπ)3
Example 4.3.2 Solve the partial differential equation
∂u 2
= 3 ∂∂xu2
∂t
0<x<1
u(0, t) = u(1, t) = 0 t≥0
u(x, 0) = e−2x 0<x<1

Xn = Bn sin nπx
2
Tn = Cn e−3(nπ) t

X 2nπ  −2 n
 −3(nπ)2 t
u(x, t) = 2π2 + 4
1 − e (−1) e sin nπx
n=1
n
Example 4.3.3 Solve the second order equation
∂u 2
∂t
= 4 ∂∂xu2 0<x<p
u(0, t) = u(p, t) = 0 t≥0
u(x, 0) = x(p − x) 0<x<p

p2 [−1 + (−1)n ]
bn = −4
 π 3 n3
0, n even;
= 8p2
π 3 n3
, n odd.

8p2 X 1 −4n2 π 2 t
 

u(x, t) = exp sin x
π 3 n=1 n3 p2 p
n odd
or

p2 [−1 + (−1)n ] −4n2 π 2 t
 
X nπ
u(x, t) = −4 3 3
exp 2
sin x
n=1
π n p p
Example 4.3.4 Solve the partial differential equation
ut = 0.003uxx
u(0, t) = u(1, t) = 0
u(x, 0) = 50 x (1 − x) 0 < x < 1.
1
200 (−1)n
Z 
200 0 if n even,
bn = 2 50 x (1 − x) sin(nπx) dx = 3 3 − = 400
0 π n π 3 n3 π 3 n3
if n odd.
The solution for 0 ≤ t ≤ 100, is given by the series:

X 400 2 2
u(x, t) = 3 3
sin(nπx) e−n π 0.003 t
n=1
π n
n odd

js, ddw de ii: Lecture Notes Page 195 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

Example 4.3.5 Solve the differential equation


∂u 2
∂t
= 9 ∂∂xu2 0<x<π
ux (0, t) = ux (π, t) = 0 t≥0
u(x, 0) = 1 − sin x 0<x<π

Note 4.3.4 For Boundary conditions given after differentiating

Xn = 6 Bn sin nx
Xn = An cos nx
2
Tn = Cn e−9n t

2
X
u(x, t) = an cos nx e−9n t
n=0

a0 X 2
u(x, t) = + an cos nx e−9n t , to have a Fourier cosine form
2 n=1


P
Note 4.3.5 We can start with n = 0, since for cos 0 6= 0, unlike when u(x, t) = sin · · ·
n=1

But u(x, 0) = f (x), the initial conditions



a0 X
(1 − sin x) = + an cos nx
2 n=1

But for a Fourier cosine series



a0 X nπ
For f (x) = + an cos x
2 n=1
L
2 L 2 π
Z Z
2
⇒ a0 = f (x) dx = (1 − sin x) dx = [π − 2]
L 0 π 0 π
Z L Z π
2 1 + (−1)n
 
2 nπ 2
⇒ an = f (x) cos x dx = (1 − sin x) cos nx dx =
L 0 L π 0 π −1 + n2

∞ ∞ 
π − 2 2 X 1 + (−1)n

a0 X −9n2 t 2
u(x, t) = + an cos nx e = + 2
cos nx e−9n t
2 n=1
π π n=1 −1 + n

Exercise 4.27 Find a series solution for


ut = 31 uxx
ux (0, t) = ux (π, t) = 0
10x
u(x, 0) = 0 < x < π.
π
Exercise 4.28 Find a series solution for
ut = uxx
u(0, t) = 0, u(1, t) = 100
u(x, 0) = sin(πx) 0 < x < 1.

Hint: Use the fact that u(x, t) = 100x is a solution satisfying ut = uxx , u(0, t) = 0, u(1, t) = 100.
See Section 4.3.3.

js, ddw de ii: Lecture Notes Page 196 of 249


4.3. METHOD OF SEPARATION OF VARIABLES - - SECOND
ORDER PDES

Example 4.3.6 Solve the partial differential equation


∂u 2
= 7 ∂∂xu2
∂t
0<x<p
ux (0, t) = ux (p, t) = 0   t≥0
u(x, 0) = T0 sin2 πx
p
0<x<p
For Boundary conditions given after differentiating

Xn = Bn cos x
p
nπ 2
T = C e−7( p ) t
n n
∞ ∞
nπ 2 nπ a0 X nπ 2 nπ
an e−7( p ) t cos an e−7( p ) t cos
X
u(x, t) = x= + x
n=0
p 2 n=1
p
But u(x, 0) = f (x)
  ∞
2 π a0 X nπ
T0 sin x = + an cos x ⇒ a0 = T0 , an = 0
p 2 n=1
p
In order not to have a constant solution (not natural that heat will remain same all time), we
try to expand f (x) by trigonometric identities.
     ∞  
2 π T0 2π a0 X nπ
T0 sin x = 1 − cos x = + an cos x
p 2 p 2 n=1
p
       
T0 T0 2π a0 π 2π 3π
− cos x = + a1 cos x + a2 cos x + a3 cos x + ···
2 2 p 2 p p p
To have coefficients
T0
a0 = T0 , a1 = 0, a2 = − , a3 = 0, a4 = 0, · · ·
2
The general solution is

a0 X nπ −7( nπ p )
2
t
u(x, t) = + an cos xe
2 n=1
p
     
a0 2
−7( πp ) t π −7( 2π )
2
t 2π −7( 3π 2
) t 3π
= + a1 e cos x + a2 e p cos x + a3 e p cos x + ···
2 p p p
 
T0 T0 −7( 2πp )2 t 2π
= +0− e cos x + 0 + 0 + ···
2 2 p
   
T0 2π −7( 2π
p )
2
t
= 1 − cos x e
2 p
Exercise 4.29 Suppose you have a wire of length 2, with k = 0.001 and an initial temperature
distribution of u(x, 0) = 50x. Suppose that both the ends are embedded in ice (temperature
0). Find the solution as a series.
Exercise 4.30 Find a series solution of
ut = uxx
u(0, t) = u(1, t) = 0
u(x, 0) = 100 0 < x < 1.
Exercise 4.31 Find a series solution of
ut = uxx
ux (0, t) = ux (π, t) = 0
u(x, 0) = 3 cos(x) + cos(3x) 0 < x < π.

js, ddw de ii: Lecture Notes Page 197 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

Example 4.3.7 Solve the initial valued problem


ut = 6uxx t>0
ux (0, t) = ux (1, t) = 0 t≥0
u(x, 0) = 6 + 4 cos(3πx)
For Boundary conditions given after differentiating
Xn = 6 Bn sin nπx
Xn = Bn cos nπx
2
Tn = Cn e−6(nπ) t

a0 X 2
u(x, t) = + an e−6(nπ) t cos nπx
2 n=1

But u(x, 0) = f (x)


∞ ∞
X a0 X
f (x) = an cos nπx = + an cos nπx
n=0
2 n=1

a0 X
6 + 4 cos(3πx) = + an cos nπx ⇒ a0 = 12, an = 0
2 n=1

Like in Example 4.3.6, we expand the relation to find any non-zero ai


a0
6 + 4 cos(3πx) = + a1 cos πx + a2 cos 2πx + a3 cos 3πx + a4 cos 4πx + · · ·
2
To generate
a0 = 12, a1 = 0, a2 = 0, a3 = 4, a4 = 0, a5 = 0, · · ·

a0 X 2
u(x, t) = + an e−6(nπ) t cos nπx
2 n=1
a0 2 2 2
= + a1 cos(πx) · e−6(π) t + a2 cos(2πx) · e−6(2π) t + a3 cos(3πx) · e−6(3π) t
2
2
+a4 cos(4πx) · e−6(4π) t + · · ·
2t
u(x, t) = 6 + 4 cos(3πx) · e−6(3π)
A solution, function of x and t, and not just a constant.
Example 4.3.8 Solve the following PDE problem
ut = 0.003uxx
ux (0, t) = ux (1, t) = 0
u(x, 0) = 50 x (1 − x) 0 < x < 1.
For this problem, we must find the Cosine series of u(x, 0). For 0 < x < 1 we have
∞  
25 X −200
50 x (1 − x) = + cos(nπx).
3 n=2
π 2 n2
n even

Hence, the solution to the PDE problem, plotted in is given by the series
∞  
25 X −200 2 2
u(x, t) = + 2 2
cos(nπx) e−n π 0.003 t
3 n=2
π n

js, ddw de ii: Lecture Notes Page 198 of 249


4.3. METHOD OF SEPARATION OF VARIABLES - - SECOND
ORDER PDES

Exercise 4.32 Find the steady state temperature solution as a function of x alone, by letting
t → ∞ in the solution from exercises 4.27 and 4.28. Verify that it satisfies the equation uxx = 0.

Exercise 4.33 Suppose that one end of the wire is insulated (say at x = 0) and the other end
is kept at zero temperature. That is, find a series solution of

ut = kuxx
ux (0, t) = u(L, t) = 0
u(x, 0) = f (x) 0 < x < L.

Express any coefficients in the series by integrals of f (x).

Exercise 4.34 [See Section 4.3.3] Suppose that the wire is circular and insulated, so there are
no ends. You can think of this as simply connecting the two ends and making sure the solution
matches up at the ends. That is, find a series solution of

ut = kuxx
u(0, t) = u(L, t)
ux (0, t) = ux (L, t)
u(x, 0) = f (x) 0 < x < L.

Express any coefficients in the series by integrals of f (x).

Exercise 4.35 Find a series solution of


ut = 3uxx
u(0, t) = u(π, t) = 0
u(x, 0) = 5 sin(x) + 2 sin(5x) 0 < x < π.

u(x, t) = 5 sin(x) e−3t + 2 sin(5x) e−75t

Exercise 4.36 Find a series solution of


ut = 0.1uxx
ux (0, t) = ux (π, t) = 0
u(x, 0) = 1 + 2 cos(x) 0 < x < π.

u(x, t) = 1 + 2 cos(x) e−0.1t

Remark 4.3.2 If the initial and boundary conditions are insufficient, solving a partial differ-
ential equation becomes harder.

js, ddw de ii: Lecture Notes Page 199 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

Example 4.3.9 Solve


∂u ∂ 2u
= ,
∂t ∂x2
subject to the boundary conditions

u(0, t) = u(1, t) = 0,

and the initial condition



 x if 0 < x < 12 ,
u(x, 0) = f (x) =
1 − x if 1/2 < x < 1.

with, "Z #
1/2 Z 1
nπ  nπ 
bn = 2 x sin xdx + (1 − x) sin x dx
0 l 1/2 l
Integration by parts yields, for even n = 2m,
1
b2m = (−1)m ,
m2 π 2
and
b2m+1 = 0,
for odd n = 2m + 1.

For problems involving heat flow there is no reason why the temperature should have the same
value at each end of the rod. The next few examples will show the effect of different boundary
conditions.

Example 4.3.10 Solve

∂u ∂ 2u
= ,
∂t ∂x2
subject to

u(0, t) = 1, u(l, t) = 0, u(x, 0) = 0.


The solution of the non-homogeneous heat equation is

x X 2
  nπ  2 2 2 2
u(x, t) = 1 − − sin x e−n π c t/l
l n=1
nπ l

Instead of specifying the value of the temperature at the ends of the rod we could fix ∂u
∂x
instead.
This corresponds to fixing the heat flux that enters or leaves the system. For example, if ∂u
∂x
= 0,
then no heat enters the system and the ends are said to be insulated.

Example 4.3.11 Second order Partial Differential Equations

∂ 2u
 
1 ∂u
∂∂x x − 2 = x2 ,
x ∂x ∂t

is linear, inhomogeneous PDE.

js, ddw de ii: Lecture Notes Page 200 of 249


4.3. METHOD OF SEPARATION OF VARIABLES - - SECOND
ORDER PDES

Example 4.3.12 The following example illustrates the case when one end is insulated and
the other has a fixed temperature.

Solve the heat equation subject to the boundary conditions


∂u
u(0, t) = 0, (1, t) = 0,
∂x
and the initial condition

u(x, 0) = 1.
In this case the steady state solution must satisfy the boundary conditions and is simply

ν(x) = 0.
Again we look for a seperable solution of the form

u(x, t) = X(x)T (t),


and the solution reduces to
2 c2 t
u(x, t) = (A cos px + B sin px) e−p
Applying the boundary condition at x = 0 gives

A = 0,
and the condition at x = 1 gives

Bp cos p = 0,
so that, in this case p must be an odd integer times π/2. Thus,
π
p = (2n + 1) .
2
Hence, the solution is

X π 2 2
u(x, t) = an sin(2n + 1) xe−(2n+1) π t/4 .
n=0
2
Finally, the initial condition is given by inverting the Fourier series at t = 0. Thus,

X π
u(x, 0) = an sin(2n + 1) x = 1.
n=0
2
Again multiply by sin(2m + 1) π2 x and integrate between x = 0 and 1 to get
Z 1
π 4
am = 2 sin(2m + 1) xdx = .
0 2 (2m + 1)π
Example 4.3.13 Second order Partial Differential Equations

∂ 2u ∂u
2
+4 = 0,
∂x ∂t
is linear PDE.

js, ddw de ii: Lecture Notes Page 201 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

4.3.2 Separation of Variables - - First Order PDEs.


The idea is still the same,
1.) How do we solve PDEs? One approach is to reduce a PDE into ODEs. How do we reduce it
to ODEs? Remember that the difference between an ODE and a PDE is that, the unknown
function in an ODE has only one variable, while that in a PDE has more than one. So
naturally we try to “separate the variables”
u(x, t) = X(x)T (t)
and hope that X, T can be obtained by solving ODEs.
2.) Substitute u = XT into the equation
3.) The two new unknown functions X, T are still coupled together. Fortunately, we can obtain
a single equation for each.
Example 4.3.14 Find the general solution of
∂u ∂u
+ = 0 (4.3.16)
∂x ∂y
The variables are x and y. Let the general solution be
u(x, y) = X(x)Y (y)
To have
ux = X 0 Y
uy = XY 0
with X 0 = dX dx
and Y 0 = dY
dy
since X = X(x) and Y = Y (y). Substituting into the partial
differential equation,
∂u ∂u
+ = 0
∂x ∂y
X 0 Y + XY 0 = 0
X0 Y 0
+ = 0
X Y
To have the ordinary differential equations
X0 Y0
=− = β
X Y
Solved with where c1 and c2 are just two arbitrary constants that can be combined later.
X0 dX dX
=β ⇔ =β ⇔ = βdx ⇔ ln X = βx + ln c1 ⇔ X = c1 eβx
X Xdx X
and
Y0 dY dY
= −β ⇔ = −β ⇔ = −βdy ⇔ ln Y = −βy + ln c2 ⇔ Y = c2 e−βy
Y Y dy Y
To reconstruct the general solution as
u(x, y) = X(x)Y (y) = c1 eβx · c2 e−βy = ce(βx−βy)
where c is an arbitrary constant. It can be readily verified that this solution satisfies the original
PDE eqn (4.3.16).

js, ddw de ii: Lecture Notes Page 202 of 249


4.3. METHOD OF SEPARATION OF VARIABLES - - SECOND
ORDER PDES

Example 4.3.15 Find the general solution of


∂u ∂u
y −x = 0 (4.3.17)
∂x ∂y
The variables are x and y. Let the general solution be

u(x, y) = X(x)Y (y)

To have the separable ordinary differential equations


X0 Y0
= = β
xX yY
Solved with where c1 and c2 are just two arbitrary constants that can be combined later.
dX 1 1 2
= xβdx ⇔ ln X = x2 β + ln c1 ⇔ X = c1 e 2 x β
X 2
and
dY 1 1 2
= yβdy ⇔ ln Y = y 2 β + ln c2 ⇔ Y = c2 e 2 y β
Y 2
We recombine them to generate the general solution as
 
1 2
x β 1 2
y β 1 2 2

u(x, y) = X(x)Y (y) = c1 e 2 · c2 e 2 = c exp β x + y
2
where c is an arbitrary constant. It can be readily verified that this solution satisfies the original
PDE eqn (4.3.17).
Example 4.3.16 Solve the first order PDE
∂u ∂u
x +y = 3u (4.3.18)
∂x ∂y
using the separation of variables.

Let the general solution be

u(x, y) = X(x)Y (y)

substituting into PDE (4.3.18) generates the separable ordinary differential equations
xX 0 yY 0
=3− = β
X Y
which can be solved as
dX dx
=β ⇔ ln X = β ln x + ln c1 ⇔ X = c1 xβ
X x
and
dY dy
= (3 − β) ⇔ ln Y = (3 − β) ln y + ln c2 ⇔ Y = c2 y 3−β
Y y
We recombine them to generate the general solution as

u(x, y) = X(x)Y (y) = c1 xβ · c2 y 3−β = cxβ y 3−β

where c is an arbitrary constant. Not identical to solutions for Example 4.2.2 on page (p. 158)
but can be readily verified that this solution satisfies the original PDE eqn (4.3.18).

js, ddw de ii: Lecture Notes Page 203 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

Example 4.3.17 Solve the problem


∂u ∂u
2xt + = u, −∞ < x < ∞, (4.3.19)
∂x ∂t
with the initial condition
u(x, 0) = x
The initial condition is in fact
u(x, 0) = x(0) = ξ,
as with Example (4.2.22) on page (p. 167).

Let the general solution be


u(x, y) = X(x)T (t)
substituting into PDE (4.3.19) generates the separable ordinary differential equations
X0 T0
 
1
x = 1− = β
X 2t T
which can be solved as
dX dx
=β ⇔ ln X = β ln x + ln c1 ⇔ X = c1 xβ
X x
and
dT
= (1 − 2βt)dt ⇔ ln T = (t − βt2 ) + ln c2 ⇔ T = c2 exp (t − βt2 )
T
We recombine them to generate the general solution as
u(x, t) = X(x)T (t) = c1 xβ · c2 exp (t − βt2 ) = c xβ exp (t − βt2 )
where c is an arbitrary constant.

Applying the initial conditions,


u(x, t) = c xβ exp (t − βt2 ), u(x, 0) = x(0) = ξ
ξ = c ξ β iff c = 1, β = 1
The general solution become
2
u(x, t) = X(x)T (t) = xet−t
Identical to solutions for Example 4.2.22 and can be readily verified that this solution satisfies
the original PDE eqn (4.3.19). Also
2
u(x, t) = X(x)T (t) = x2 et−2t
satisfy the equation - Solutions are not unique.
Exercise 4.37 Solve the problem in Example 4.2.15 on page (p. 164) using separation of
variables.
Exercise 4.38 Solve the problem in Example 4.13 on page (p. 167) using separation of
variables.
Exercise 4.39 Solve the first order PDE
∂u
= 2x + 3y
∂x
An alternative solution is done in Exercise 4.2.1 on page (p. 158).
Remark 4.3.3 Each method has advantages and limitations. This technique will always fail
if we cannot separate variables.

js, ddw de ii: Lecture Notes Page 204 of 249


4.3. METHOD OF SEPARATION OF VARIABLES - - SECOND
ORDER PDES

Example 4.3.18 Solve the transport equation


∂u ∂u 2
+3 = 0, u(x, 0) = xe−x .
∂t ∂x
Alternative approach is in Example 4.2.24 on page (p. 169).

By separation of variables, we let

u(x, y) = X(x)T (t)

substituting into PDE,


X0 1 T0
=− = β
X 3T
to have
dX
= βdx ⇔ ln X = βx + ln c1 ⇔ X = c1 eβx
X
and
dT
= −3βdt ⇔ ln T = −3βt + ln c2 ⇔ T = c2 e−3βt
T
We recombine them to generate the general solution as

u(x, t) = X(x)T (t) = c1 eβx · c2 e−3βt = ce(x−3t)β

where c is an arbitrary constant.

Applying the initial conditions,


2
u(x, t) = ce(x−3t)β , x(0) = ξ, u(0) = ξe−ξ
2
ξe−ξ = c eξβ iff c = ξ, β = −ξ

The general solution become


2
u(x, t) = X(x)T (t) = (x − 3t)e−(x−3t)

If let ξ = (x − 3t). Notice that, with separation of variables, we need more conditions in order
to get out the all the scalars.
Example 4.3.19 Solve the partial differential equation
∂u ∂u
+u =0
∂y ∂x
By separation of variables, we let

u(x, y) = X(x)Y (y)

substituting into PDE,


Y0
X0 = − = β
Y2
to have
dY 1
dX = βdx ⇔ X = βx + c1 and − 2
= βdy ⇔ Y =
Y βy + c2

js, ddw de ii: Lecture Notes Page 205 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

We recombine them to generate the general solution as


1 βx + c1
u(x, t) = X(x)T (t) = βx + c1 · =
βy + c2 βy + c2
where c1 , c2 and β are arbitrary constants.

Alternatively, see Example 4.2.25 on page (p. 169). Try to apply the initial conditions on page
(p. 169), we realise, more is needed.
Example 4.3.20 Find the separated solution to

ux + 2xuy = 0

Using
X0 Y0
= = β
−2xX Y
Generates
2
X = c1 e−βx
Y = c2 eβy

To have general solutions


u(x, t) = X(x)Y (y) = ceβ (y−x )
2

where c is an arbitrary constant.

Notice that there are similarities between the separated solution

u(x, t) = ceβ (y−x )


2

and the other solutions using a different method, say

u(x, y) = y − x2

However, the two solutions are clearly not the same.

Still the separated solution will again be different if −2 was on side of Y and not X.
Exercise 4.40 Find the separated solution to each of the following partial differential equa-
tions. Assume that κ, a, c and τ are constant.
∂u ∂u
1) ∂t
= ∂x
5) ux + 3x2 uy = 0
∂u
2) ∂t
= −κ ∂u
∂x
6) ux + ut = u
∂u ∂u ∂u ∂u
3) ∂x
+ ∂y
=0 7) ∂x ∂y
=u

4) ux + e−x uy = 0 8) ∂u
∂x
= −2x ∂u
∂y

Exercise 4.41 Find a separated solution of the following nonlinear wave equation:
∂u ∂u
= cu
∂t ∂x

js, ddw de ii: Lecture Notes Page 206 of 249


4.3. METHOD OF SEPARATION OF VARIABLES - - SECOND
ORDER PDES

4.3.3 Non-Homogeneous Heat Equation


A general heat equation is said to be non-homogeneous if

Either ut = kuxx + h
or u(0, t) 6= 0
or u(L, t) 6= 0
or all

Solution for Non-Homogeneous Heat Equation


We get a linear function ν(x) or ν(t) or ν(x, t) to transform back to a homogeneous equation.
Thus our general solution will be sum of the complimentary and particular solution

ug = uc + up
u(x, t) = w(x, t) + ν(x) if h = h(x)
u(x, t) = w(x, t) + ν(t) if h = h(t)
u(x, t) = w(x, t) + ν(x, t) if h = h(x, t)

1.) The general solution satisfy the partial differential equation. Substitute u(x, t) into the
PDE.

2.) Generate PDEs for both complimentary w and particular solutions ν.

3.) The particular solutions ν do satisfy the boundary conditions. The particular solution can
be solved fully.

4.) The complimentary solutions (after replacing u with u = w + ν) will therefore satisfy a
homogeneous PDE
wt = wtt
w(0, t) = w(L, t) = 0
w(x, 0) = f (x) − ν
which can be solved using separation of variables.

w(x, t) = X(x)T (t)


∞  nπ 
2
−( nπ )
X
kt
w(x, t) = an e L sin x
n=1
L
2 L
Z  nπ 
an = [f (x) − ν(x)] sin x dx
L 0 L

5.) The final solution will be the sum of complimentary and particular solutions.

u(x, t) = w(x, t) + ν(x)

Remark 4.3.4 For


ν = ν(x) ⇒ νt = 0

js, ddw de ii: Lecture Notes Page 207 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

Example 4.3.21 Find a solution to the given initial-boundary value problem


ut = uxx 0<x<π
u(0, t) = 0, u(π, t) = 3π t≥0
u(x, 0) = π − x 0<x<π
Note 4.3.6 This is a non-homogeneous Heat equation since u(π, t) = 3π 6= 0

Using u = w(x, t) + ν(x)


wt + 0 = wxx + νxx 0<x<π
w(0, t) + ν(0) = 0, w(π, t) + ν(π) = 3π t≥0 (4.3.20)
w(x, 0) = (π − x) − ν(x) 0<x<π
Thus for particular solution
0 = νxx 0<x<π
ν(0) = 0, ν(π) = 3π
To have
ν 00 (x) = 0 ⇒ ν 0 (x) = c1 ⇒ ν(x) = c1 x + c2
With the boundary conditions ν(0, t) = 0, ν(π, t) = 3π
c2 = 0 , c1 = 3 ⇒ ν(x) = 3x + 0 = 3x
With the ν transformation, the homogenous (4.3.20) problem (Put all in w not u) becomes
wt = wxx 0<x<π
w(0, t) = 0, w(π, t) = 0 t≥0
w(x, 0) = f (x) − ν(x) = π − 4x 0<x<π
By separation of variables w = XT, wx = X 0 T, wxx = X 00 T, wt = XT 0
Xn (x) = Bn sin nx
2
Tn (t) = Cn e−n t
2 L
Z
an = [f (x) − ν(x)] sin nx dx
L 0
2 π 2 π
Z Z
= [(π − x) − 3x] sin nx dx = [π − 4x] sin nx dx
π 0 π 0
2
= [3(−1)n + 1]
n

X
w(x, t) = X(x)T (t) = an Xn Tn
n=1

2
X
= an e−n t sin nx
n=1

X 2 2
= [3(−1)n + 1] e−n t sin nx
n=1
n
The general solution
u(x, t) = ν(x) + w(x, t)

X 2 2
= 3x + [3(−1)n + 1] e−n t sin nx
n=1
n

js, ddw de ii: Lecture Notes Page 208 of 249


4.3. METHOD OF SEPARATION OF VARIABLES - - SECOND
ORDER PDES

Example 4.3.22 Find a solution to the given initial-boundary value problem


ut = 4uxx 0 < x < 10
u(0, t) = 10, u(10, t) = 30 t≥0
u(x, 0) = 0 0 < x < 10
Note 4.3.7 This is a non-homogeneous Heat equation since u(0, t) = 10 6= 0 and also
u(10, t) = 30 6= 0.

Using general solution as complimentary plus particular, u = w(x, t) + ν(x)


wt = 4wxx + 4νxx 0 < x < 10
w(0, t) + ν(0) = 10, w(10, t) + ν(10) = 30 t≥0 (4.3.21)
w(x, 0) + ν(x) = 0 0 < x < 10
Such that for the particular solutions (particular solutions satisfy boundary conditions, not
initial conditions)
0 = 4νxx 0 < x < 10
ν(0) = 10, ν(10) = 30
To have
ν 00 (x) = 0 ⇒ ν 0 (x) = c1 ⇒ ν(x) = c1 x + c2
Since ν(x) is also a solution, we use the boundary conditions ν(0, t) = 10, ν(10, t) = 30
c2 = 10, c1 = 2 ⇒ ν(x) = 2x + 10
A transformation that generates (4.3.21) into a homogenous problem in w.
wt = 4wxx 0 < x < 10
w(0, t) = 0, w(10, t) = 0 t≥0
w(x, 0) = f (x) − ν(x) = −(2x + 10) 0 < x < 10
By separation of variables w = XT, wx = X 0 T, wxx = X 00 T, wt = XT 0
 nπ   nπ 
Xn (x) = Bn sin x = Bn sin x
L 10  2 2
2 nπ 2
−( nπ ) kt −( ) 4t − n 25π t
Tn (t) = Cn e L = Cn e 10 = Cn e
Z L Z 10
2  nπ  2  nπ 
an = [f (x) − ν(x)] sin x dx = [(0) − (2x + 10)] sin x dx
L 0 10 10 0 10
Z 10
2  nπ  20
= − [2x + 10] sin x dx = [3(−1)n − 1]
10 0 10 nπ
X∞
w(x, t) = X(x)T (t) = an Xn Tn
n=1

n2 π 2
   nπ 
− t
X
= an e 25
sin x
n=1
10

20
 2 2  nπ 
− n 25π t
X
n
= [3(−1) − 1] e sin x
n=1
nπ 10

The general solution


u(x, t) = ν(x) + w(x, t)

20
 2 2  nπ 
− n 25π t
X
n
= (2x + 10) + [3(−1) − 1] e sin x
n=1
nπ 10

js, ddw de ii: Lecture Notes Page 209 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

Example 4.3.23 Solve the initial-boundary value problem


ut = 2uxx + 4x 0<x<π
u(0, t) = u(π, t) = 0 t≥0
u(x, 0) = sin x 0<x<π
Note 4.3.8 This is a non-homogeneous Heat equation since ut = 2uxx + 4x

Using u(x, t) = w(x, t) + ν(x)


wt + 0 = 2wxx + 2νxx + 4x 0<x<π
w(0, t) + ν(0) = w(π, t) + ν(π) = 0 t≥0 (4.3.22)
w(x, 0) + ν(x) = sin x 0<x<π
To have the particular as
0 = 2νxx + 4x 0<x<π
ν(0) = ν(π) = 0
To generate the particular solutions as
x3
ν 00 (x) = −2x, ν 0 (x) = −x2 + c1 , ν(x) = − + c1 x + c2
3
With boundary conditions ν(0, t) = ν(π, t) = 0
c2 = 0
π2
c1 =
3
x3 π 2
ν(x) = − + x
3 3
The steady (does not depend on time) state solution ν(x) can be used in (4.3.22) to generate
a homogeneous PDE
wt = 2uxx 0<x<π
w(0, t) = w(π, t) = 0 t≥0
x3 π2
w(x, 0) = f (x) − ν(x) = sin x + 3
− 3
x 0<x<π
To be solved by separation of variables w = XT, wx = X 0 T, wxx = X 00 T, wt = XT 0
 nπ 
Xn (x) = Bn sin x = Bn sin nx
L
nπ 2
Tn (t) = Cn e−( L ) kt = Cn e−2n t
2

2 L
Z  nπ 
an = [f (x) − ν(x)] sin x dx
L 0 π
2 π
 3
π2
Z  
x
= (sin x) − − + x sin nx dx
π 0 3 3
Z π 3 2
4(−1)n
  
2 x π
= sin x + − x sin nx dx =
π 0 3 3 n3
∞ ∞ ∞ 
4(−1)n

−2n2 t 2
X X X
w(x, t) = X(x)T (t) = an Xn Tn = an e sin nx = 3
sin nxe−2n t
n=1 n=1 n=1
n

The general solution


 3 ∞ 
π2 4(−1)n
 X 
x 2
u(x, t) = ν(x) + w(x, t) = − + x + 3
sin nxe−2n t
3 3 n=1
n

js, ddw de ii: Lecture Notes Page 210 of 249


4.3. METHOD OF SEPARATION OF VARIABLES - - SECOND
ORDER PDES

Example 4.3.24 Solve the initial-boundary value problem


ut = uxx + sin(3πx) 0<x<1
u(0, t) = u(1, t) = 0 t≥0
u(x, 0) = sin(πx) 0<x<1
Note 4.3.9 This is a non-homogeneous Heat equation since ut = uxx + sin(3πx)

For ν 00 (x) = − sin(3πx), we solve for ν(x)


ν 00 (x) = − sin(3πx)
1
ν 0 (x) = cos(3πx) + c1

1
ν(x) = sin(3πx) + c1 x + c2
9π 2
Since ν(x) is also a solution, we use the boundary conditions ν(0, t) = ν(1, t) = 0
c2 = 0
c1 = 0
1
ν(x) = sin(3πx)
9π 2
The steady state solution. Since we have got a transformation, we now solve the homogenous
problem (Put all in w not u)
wt = wxx 0<x<1
w(0, t) = w(1, t) = 0 t≥0
1
w(x, 0) = f (x) − ν(x) = sin πx − 9π 2
sin(3πx) 0<x<1
By separation of variables
w = XT
wx = X 0T
wxx = X 00 T
wt = XT 0
 nπ 
Xn (x) = Bn sin x = Bn sin nπx
L
nπ 2
Cn e−( L ) kt = Cn e−(nπ) t
2
Tn (t) =
2 L
Z
an = [f (x) − ν(x)] sin nπx dx
L 0
Z 1 
1
= 2 sin πx − 2 sin(3πx) sin nπx dx
0 9π
X∞
w(x, t) = X(x)T (t) = an Xn Tn
n=1

2
X
= an e−(nπ) t sin nπx
n=1

The general solution


u(x, t) = ν(x) + w(x, t)

1 X 2
= 2
sin(3πx) + an e−(nπ) t sin nπx
9π n=1

js, ddw de ii: Lecture Notes Page 211 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

Example 4.3.25 Solve the initial-boundary value problem

ut = 3uxx + 5 0<x<π
u(0, t) = u(π, t) = 1 t≥0
u(x, 0) = 1 0<x<π

Note 4.3.10 This is a non-homogeneous Heat equation since ut = 3uxx + 5 and


u(0, t) = u(π, t) = 1

With general solutions u(x, t) = w(x, t) + ν(x)

wt = 3wxx + 3νxx + 5 0<x<π


w(0, t) + ν(0) = w(π, t) + ν(π) = 1 t≥0 (4.3.23)
w(x, 0) + ν(x) = 1 0<x<π

in order to remain with a homogeneous one (particular solutions always satisfy boundary con-
ditions)
0 = 3νxx + 5 0<x<π
ν(0) = ν(π) = 1
To have
5 5 5
ν 00 (x) = − ⇒ ν 0 (x) = − x + c1 ⇒ ν(x) = − x2 − c1 x + c2
3 3 6
With the boundary conditions ν(0, t) = ν(π, t) = 1
5π 5 5π
c2 = 1, c1 = ⇒ ν(x) = − x2 + x+1
6 6 6
The steady state solution and a transformation to simplify (4.3.23) to the homogeneous problem

wt = 3wxx 0<x<π
w(0, t) = w(π, t) = 0 t≥0
w(x, 0) = 65 x2 − 5π
6
x 0<x<π

By separation of variables
 nπ 
Xn (x) = Bn sin x = Bn sin nx
L
nπ 2
Tn (t) = Cn e−( L ) kt = Cn e−3n t
2

2 L 2 π
Z Z  
5 2 5π
an = [f (x) − ν(x)] sin nπx dx = 1+ x − x − 1 sin nπx dx
L 0 π 0 6 6
2 π 5 2 5π
Z   
10 n 0 n even
= x − x sin nπx dx = [(−1) − 1] = 20
π 0 6 6 3πn 3 − 3πn3 n is odd
∞ ∞ ∞
X 20
2 2
X X
w(x, t) = X(x)T (t) = an Xn Tn = an e−3n t sin nx = − 3
e−3n t sin nx
n=1 n=1 n=1
3πn

The general solution

u(x, t) = ν(x) + w(x, t)


  X ∞
5 2 5π 20 −3n2 t
= − x + x+1 − e sin nx
6 6 n=1
3πn3

js, ddw de ii: Lecture Notes Page 212 of 249


4.3. METHOD OF SEPARATION OF VARIABLES - - SECOND
ORDER PDES

Example 4.3.26 Solve the following heat equation

ut − kuxx = A cos αt 0 < x < 1, t > 1


ux (0, t) = ux (1, t) = 0 t≥0
u(x, 0) = 1 + cos2 πx 0 ≤ x ≤ 1.

We write the solution u is of the form

u(x, t) = w(x, t) + ν(t)

where ν is a solution of the equation ν 0 (t) = A cos αt satisfying ν(0) = 0. Find the equation for
w. Then " #t
Z t
A A
ν(t) = ν(0) + A cos αs ds = sin αs = sin αt.
0 α α
0

The function w(x, t) = u(x, t) − ν(t) satisfies

wt − kvxx = 0 0 < x < 1, t > 1


wx (0, t) = wx (1, t) = 0 t≥0
w(x, 0) = 1 + cos2 πx 0 ≤ x ≤ 1.
The formal solution is given by
2 π2 t
X
w(x, t) = Bn cos(nπx)e−kn ,
n≥0

where Z 1 Z 1
2
B0 = (1 + cos πx) dx and Bn = 2 (1 + cos2 πx) cos nπx dx.
0 0
We have Z 1
3
B0 = (1 + cos2 πx)dx =
0 2
and for n ≥ 1,
Z 1 Z 1
2
Bn = 2 (1 + cos πx) cos nπx dx = (3 + cos 2πx) cos nπx dx
0 0
Z 1 Z 1  1
2
, n=2
= 3 cos nπx dx + cos 2πx cos nπx dx =
0 0 0, n = 2.

Hence
3 1 2 A
u(x, t) = + cos(2πx)e−4kπ t + sin αt.
2 2 α
Remark 4.3.5 ν 6= ν(x), but ν = ν(t) since RHS is function in t and not x.

js, ddw de ii: Lecture Notes Page 213 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

4.3.4 Wave Equation Solution


The wave equation is an important second-order linear partial differential equation that de-
scribes the propagation of a variety of waves, such as sound waves, light waves and water
waves. It arises in fields such as acoustics, electromagnetic, and fluid dynamics. Historically,
the problem of a vibrating string such as that of a musical instrument was studied by Jean le
Rond d’Alembert, Leonhard Euler, Daniel Bernoulli, and Joseph-Louis Lagrange.

The general wave equation is of the form

utt = c2 uxx 0<x<L


u(0, t) = u(L, t) = 0 t≥0
u(x, 0) = f (x) 0<x<L
ut (x, 0) = g(x) 0<x<L

Since a PDE in (x, t) not (x, y), by separation of variables, we assume a solution

u(x, t) = X(x)T (t) = XT (4.3.24)

where X(x) is a solution and T (t) is also a solution that satisfy the original problem.

ux = X 0T
uxx = X 00 T
ut = XT 0
utt = XT 00

Substituting in the general wave equation given,

utt = c2 uxx
XT 00 = c2 X 00 T
X 00 T 00
= 2
X cT
Note 4.3.11 All constants must be on the side of T s not Xs.
X 00 T 00
= = − α2
X c2 T

X 00
= −α2 (4.3.25)
X
T 00
= −α2 (4.3.26)
c2 T
Solving the two equations for X(x) and T (t),

X 00
= −α2
X
X 00 + α2 X = 0
X(x) = A cos αx + B sin αx

using X(0, t) = 0

X(x) = A cos αx + B sin αx

js, ddw de ii: Lecture Notes Page 214 of 249


4.3. METHOD OF SEPARATION OF VARIABLES - - SECOND
ORDER PDES

0 = A cos 0 + B sin 0
⇒ A=0
⇒ X = B sin αx

using X(L, t) = 0

X(x) = B sin αx
0 = B sin αL
⇒ αL = nπ

⇒ α =
L  nπ 
⇒ Xn (x) = Bn sin x (4.3.27)
L

T 00
= −α2
c2 T
T 00 + α2 c2 T = 0
T (t) = D cos cαt + E sin cαt

h  nπ  i h  nπ  i
Tn (t) = D cos c t + E sin c t (4.3.28)
L L
using the principle of superposition -Now, we have a solution for every integer value of n. Thus,
to form the most general solution we must add together all the possible solutions - with

u(x, t) = X(x)T (t)

∞ h
X  nπ   nπ  i  nπ 
u(x, t) = Dn cos c t + En sin c t sin x (4.3.29)
n=1
L L L

using the initial condition u(x, 0) = f (x)



X  nπ 
u(x, 0) = Dn sin x = f (x)
n=1
L

This is a ”Sine” series with


Z L
2  nπ 
Dn = f (x) sin x dx (4.3.30)
L 0 L
using the initial condition ut (x, 0) = g(x)

X  nπc   nπ 
ut (x, 0) = En sin x = g(x)
n=1
L L

This is a ”Sine” series with


 nπc  Z L
2  nπ 
En = g(x) sin x dx
L L 0 L

Z L
2  nπ 
En = g(x) sin x dx (4.3.31)
nπc 0 L

js, ddw de ii: Lecture Notes Page 215 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

using the definitions of Dn and En , the final solution to the wave equation, that satisfies both
the initial and boundary conditions is

X nπ  nπc nπc 
u(x, t) = sin x Dn cos t + En sin t .
n=1
L L L
Notice that, although the solution was obtained by looking for solutions that were separable in
x and t, the final answer is NOT separable (unless all the coefficients are zero except one).
Example 4.3.27 Using the geometric series
1
= 1 + x + x2 + x3 + · · · ,
1−x
we see that we can express

1 X 1
1 = n
sinn x cosn t.
1 − 2 sin x cos t n=0 2
Each term in the sum is separable, having the form sinn x cosn t, but the left hand side cannot
be expressed as a function of x times a function of t.
Example 4.3.28 We can relate our solution to d’Alembert’s solution as follows. using the
trigonometric identities
nπ nπc 1  nπ nπ 
sin x cos t= sin (x − ct) + sin (x + ct) ,
L L 2 L L
and
nπ nπc 1 nπ nπ 
sin x sin t= cos (x − ct) − cos (x + ct) .
L L 2 L L
These expressions are both of the form of F (x − ct) and G(x + ct). The solution that was
obtained in (4.3.29) corresponds to standing waves whereas d’Alembert’s solution corresponds
to travelling waves we see that standing waves can be considered as the superposition of an
inifinite number of travelling waves travelling to the left and right but interacting in such a way
as to produce standing waves.We now illustrate the method of separation with a few examples
corresponding to different initial conditions. We choose different forms for f (x) and g(x). In
all these examples we choose L = π and c = 1.
Example 4.3.29 Consider
∂u
u(x, 0) = f (x) = sin x, and (x, 0) = g(x) = 0.
∂t
From (4.3.31) we have En = 0 for all n and from (4.3.30) we have D1 = 1 and Dn = 0 for
n > 1. We can verify this quite easily. Consider
1
sin x sin nx = [cos(1 − n)x − cos(1 + n)x] ,
2
for n 6= 1. Clearly,
Z π  π
1 sin(1 − n)x sin(1 + n)x
sin x sin nxdx = − + ,
0 2 1−n 1+n 0
and so π   
sin(1 − n)π sin(1 + n)π
Z
1
sin x sin nxdx = − + − (0) = 0.
0 2 1−n 1+n
Finally,
π
1 π
Z Z
2 π
sin xdx = (1 − cos 2x) dx = .
0 2 0 2
Thus, the only term in the Fourier series that is non-zero is the first term and so D1 = 1.

js, ddw de ii: Lecture Notes Page 216 of 249


4.3. METHOD OF SEPARATION OF VARIABLES - - SECOND
ORDER PDES

Example 4.3.30 Consider



2x π
if 0 < x <


2
π


f (x) =

 2 π
 (π − x) if

2
<x<π
π
This is a triangular profile for u(x, 0) = f (x) and again we take
∂u
(x, 0) = g(x) = 0
∂t
∂u
Since ∂t
= 0 we have En = 0 and Dn are given by
"Z #
π/2 Z π
22
Dn = x sin nxdx + (π − x) sin nxdx
ππ 0 π/2

Adding the two terms together gives


  nπ 
4 2
Dn = 2 sin
π n2 2
Notice that when n is even we have the sine of a multiple of π and so the even coefficients are
zero. For n = 2m + 1, so that n is odd, we have
 
1
sin m + π = (−1)m
2
Thus,
8
D2m+1 = (−1)m
(2m + 1)2 π 2
Hence the solution to the wave equation, for these initial conditions, is

8 X (−1)m
u(x, t) = sin [(2m + 1)x] cos [(2m + 1)t]
π 2 m=0 (2m + 1)2

Note 4.3.12 We solve for X(x), α = L
, T (t), Dn , En and finally the general solution.
Example 4.3.31 Here we take the initial conditions as

u(x, 0) = f (x) = sin x

and
∂u
= g(x) = − sin x
∂t
so that there is an initial velocity and well as an initial displacement. However, since the
functions only involve a single sine term, the solution will simply be

u(x, t) = D sin x cos t + E sin x sin t

Thus, D = 1 and E = −1 so that

u(x, t) = sin x cos t − sin x sin t

It is easily verified that this solution does indeed satisfy the initial conditions.

js, ddw de ii: Lecture Notes Page 217 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

Example 4.3.32 Consider u(x, 0) = f (x) = 0 and



2x
if 0 < x < π2


π

∂u 
(x, 0) = g(x) =
∂t  2
 (π − x) if π2 < x < π


π
Now we have
Dn = 0
and "Z #
π/2 Z π
2 2
En = x sin nxdx + (π − x) sin nxdx
nπ π 0 π/2

we obtain
8  nπ 
En = sin
n3 π 2 2
Hence, the solution to the wave equation is

8 X (−1)n
u(x, t) = 2 sin [(2n + 1)x] sin [(2n + 1)t]
π n=0 (2n + 1)3

Example 4.3.33 Find a formal solution to the given boundary-value problem

utt = 4uxx −∞ < x < ∞, t > 0


2
u(x, 0) = e−x −∞ < x < ∞
ut (x, 0) = sin x −∞ < x < ∞

Example 4.3.34 Solve the wave equation problem below

utt = 4uxx 0 < x < 1, t > 0


u(0, t) = u(1, t) = 0 t>0
u(x, 0) = x(1 − x) 0<x<1
ut (x, 0) = sin 7πx 0<x<1

js, ddw de ii: Lecture Notes Page 218 of 249


4.3. METHOD OF SEPARATION OF VARIABLES - - SECOND
ORDER PDES

4.3.5 Laplace Equation Solutions


In mathematics, Laplace’s equation is a partial differential equation named after Pierre-Simon
Laplace who first studied its properties. The solutions of Laplace’s equation are important in
many fields of science, notably the fields of electromagnetism, astronomy, and fluid dynamics,
because they describe the behavior of electric, gravitational, and fluid potentials. The general
theory of solutions to Laplace’s equation is known as potential theory. In the study of heat
conduction, the Laplace equation is the steady-state heat equation.

The general Laplace equation 52 u is of the form

uxx + uyy = 0 0 < x < a, 0 < y < b


u(x, 0) = 0
u(x, b) = f (x) 0<x<a
ux (0, y) = 0 0<y<b
uy (0, y) = 0 0<y<b

Since a PDE in (x, y), by separation of variables, we assume a solution

u(x, y) = X(x)Y (y) = XY (4.3.32)

where X(x) is a solution and Y (y) is also a solution that satisfy the original problem.

ux = X 0Y
uxx = X 00 Y
uy = XY 0
uyy = XY 00

Substituting in the general Laplace equation given,


∞ 
X nπy  nπx
u(x, y) = A0 y + An sinh cos (4.3.33)
n=1
a a
Z a
1
Where A0 = f (x) dx (4.3.34)
ab 0
Z a
2 nπx
and An = nπb
f (x) cos dx (4.3.35)
a sinh a 0 a

Example 4.3.35 Find a formal solution to the given boundary-value problem

uxx + uyy = 0 0 < x < π, 0 < y < 1


u(0, y) = 0, u(π, y) = 0 0≤y≤1
u(x, 0) = 0, u(x, 1) = sin x 0≤x≤π

Example 4.3.36 Find a formal solution to the given boundary-value problem

uxx + uyy = 0 0 < x < 1, 0 < y < 1


ux (0, y) = 0, ux (1, y) = 0 0≤y≤1
u(x, 0) = x2 , uy (x, 1) = 0 0≤x≤1

js, ddw de ii: Lecture Notes Page 219 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

Exercise 4.42 Outline the assumptions under problems for


1.) the heat flow in a rod.
2.) the one dimensional wave equation.
Exercise 4.43 Sate the various situations out which boundary conditions arise for wave equa-
tions.
Exercise 4.44 Find a formal solution to the following initial-boundary value problems.
1.)
ut = 2uxx 0<x<π
u(0, t) = 5, u(π, t) = 10 t≥0
u(x, 0) = sin 3x − sin 5x 0<x<π
2.)
ut = 3uxx + 5 0<x<π
u(0, t) = u(π, t) = 1 t≥0
u(x, 0) = 1 0<x<π
3.)
utt = uxx 0<x<1
u(0, t) = u(1, t) = 0 t≥0
u(x, 0) = x(1 − x) 0<x<1
ut (x, 0) = sin 7πx 0<x<1
4.)
utt = 9uxx 0<x<π
u(0, t) = u(π, t) = 0 t≥0
u(x, 0) = sin 4x + 7 sin 5x 0<x<π
x, 0 < x < π2
ut (x, 0) =
π − x , π2 < x < π
Example 4.3.37 Find the general solution of
1.) zxx − 2zx + z = 5x2 − 7x + 4x2 ex
2.) ztt − 3zt + 2z = 3e−t − 10 cos 3t
Since in one variable, not to consider it as PDE but as ode and we solve using the method of
”undetermined coefficients” (Recall: Principle of superposition and expected difficulties)
1.)
zc = c1 ex + c2 xex
zp1 = Ax2 + Bx + c = 5x2 + 13x + 16
zp2 = x2 (Ax2 + Bx + c)ex

2.)
zc = c1 et + c2 e2t
1
zp1 = Ae−t = e−t
2
7 9
zp2 = A cos 3t + B sin 3t = cos 3t + sin 3t
13 13
zg = zc + zp
1 7 9
= c1 et + c2 e2t + e−t + cos 3t + sin 3t
2 13 13
js, ddw de ii: Lecture Notes Page 220 of 249
4.3. METHOD OF SEPARATION OF VARIABLES - - SECOND
ORDER PDES

Note 4.3.13 The basic idea of separation of variables is to:

1. Apply the method of separation to obtain two ordinary differential equations.

2. Determine the solutions that satisfy the boundary conditions.

3. use Fourier series to superimpose the solutions to get the final solution that satisfies both
the heat,wave or Laplace equation and the given initial conditions.

We assume that u(x, t) can be expressed as a product of a function of x and a function of t.


This is an assumption and if we end up with a contradiction along the line then the assumption
was wrong. If there is no contradiction then the assumption is valid. Thus, we seek a solution
of the form
u(x, t) = X(x)T (t).

Example 4.3.38 Consider the equation

∂ 2u
− u = 0.
∂y∂x
Again we set

u(x, y) = X(x)Y (y).


so that the equation becomes

dX dY
= XY.
dx dy
Dividing by (dX/dx)Y , instead of XY , we obtain

1 dY 1
=X .
Y dy dX/dx
Since the left hand side is a function of y alone and the right hand side is function of x
alone, the only possible solution is if they are both equal to a constant, say k. Thus, we get
two ordinary differential equations of the form

dY
= kY,
dy
and

dX 1
= X.
dx k
These are easily solved in terms of exponential functions to give

Y (y) = Aeky ,
and

X(x) = Bex/k .
Thus, the general solution to the partial differnetial equation is

u(x, y) = Ceky+x/k .

js, ddw de ii: Lecture Notes Page 221 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

Example 4.3.39 This example shows how the method of separation of variables can be ap-
plied to problems for which the coefficients are functions of x and y. For the separation to work
the coefficients do need to have particular forms. Here

∂ 2u
x2 + 3y 2 u = 0.
∂y∂x
Again we set
u(x, y) = X(x)Y (y),
substitute into the equation and divide by X(dY /dy) to obtain

dX 1 1
x2 = −3y 2 Y .
dx X dY /dy

Here the left hand side is a function of x alone, whereas the right hand side is a function of y
alone and so they must both equal a constant, k. Thus, the two equations are
dX
x2 = kX,
dx
dY y2
= −3 Y
dy k
These equations are both separable first order equations and so we obtain, for the x dependence
dX k k
= 2 dx ⇒ ln X = − + c,
X x x
and hence
X(x) = Ae−k/x .
The y dependence gives

dY y2 y3
= −3 Y ⇒ log Y = − + c,
Y k k
so that
3 /k
Y (y) = Be−y .
Thus, the general solution to the equation is
3 /k
u(x, y) = Ce−k/x−y .

js, ddw de ii: Lecture Notes Page 222 of 249


4.4. METHOD OF CONSTANT COEFFICIENT - - SECOND ORDER
PDES

4.4 Method of Constant Coefficient - - Second Order


PDEs
Like ordinary DEs, partial DEs with constant coefficients can be solved explicitly.

4.4.1 Homogeneous Partial DEs


For partial DEs of the form

uxx + k1 uxy + k2 uyy = 0, (4.4.1)


∂r ∂r
where k1 and k2 are constants, we write D1r = ∂x
and D2r = ∂y
, so that in symbolic form it can
be written as

D12 + k1 D1 D2 + k2 D22 u = 0.

(4.4.2)

Its symbolic operator equated to zero, i.e.,

D12 + k1 D1 D2 + k2 D22 = 0. (4.4.3)

is called the auxiliary equation. Let its roots be

D1
= m1 , m2 . (4.4.4)
D2

1.) Case I. If the roots are distinct, then (4.4.2) is equivalent to

(D1 − m1 D2 )(D1 − m2 D2 )u = 0. (4.4.5)

Which will be satisfied by

1.
dx dy du
(D1 − m2 D2 )u = 0 ⇔ ux − m2 uy = 0 ⇔ = =
1 −m2 0
⇔ y + m2 x = a and u = b
⇔ u = φ(y + m2 x).

2.
dx dy du
(D1 − m1 D2 )u = 0 ⇔ ux − m1 uy = 0 ⇔ = =
1 −m1 0
⇔ y + m1 x = a and u = b
⇔ u = f (y + m1 x).

Hence, in this case the general solution of (4.4.1) is

u = f (y + m1 x) + φ(y + m2 x).

2.) Case II. If the roots are equal, i.e., m1 = m2 , then (4.4.5) is equivalent to

(D1 − m1 D2 )2 u = 0. (4.4.6)

js, ddw de ii: Lecture Notes Page 223 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

Putting (D1 − m1 D2 )u = v, it becomes (D1 − m1 D2 )v = 0, which as earlier gives v =


φ(y + m1 x). Therefore, (4.4.6) takes the form (D1 − m1 D2 )u = φ(y + m1 x), or

ux − m1 uy = φ(y + m1 x).

Which can be solved as


dx dy du
= =
1 −m1 φ(y + m1 x)
giving y + m1 x = a and du = φ(a)dx, i.e., u = φ(a)x + b. Thus, the general solution of
(4.4.1) in this case is
u = f (y + m1 x) + xφ(y + m2 x).

Example 4.4.1 For the partial DE

2uxx + 5uxy + 2uyy = 0

the auxiliary equation is 2m2 + 5m + 2 = 0, m = D1 /D2 , which gives m1 = −2, m2 = −1/2.


Hence, its general solution can be written as

u = f1 (y − 2x) + f2 (2y − x).

Example 4.4.2 For the partial DE

uxx + 6uxy + 9uyy = 0

the auxiliary equation is m2 + 6m + 9 = 0, m = D1 /D2 , which gives m1 = −3, m2 = −3. Hence,


its general solution can be written as

u = f1 (y − 3x) + xf2 (y − 3x).

js, ddw de ii: Lecture Notes Page 224 of 249


4.4. METHOD OF CONSTANT COEFFICIENT - - SECOND ORDER
PDES

4.4.2 Nonhomogeneous Partial DEs


Consider nonhomogeneous partial DEs of the form
d(D1 , D2 )[u] = (D12 + k1 D1 D2 + k 2 D22 )u = F (x, y). (4.4.7)
As in the case of ordinary DEs a particular solution up (x, y) of (4.4.7) can be obtained by
employing the operator method, i.e.,
1
up (x, y) = F (x, y). (4.4.8)
d(D1 , D2 )
Definition 4.4.1 An inverse differential operator D−1 = 1/D correspond to the integration
of a function i.e. Z
1
g(x) = g(x)dx
D
and
Z Z 
1 1 1
g(x) = = g(x)dx dx (4.4.9)
D2 DD
and
Z
1
g(x) = eax e−ax g(x)dx (4.4.10)
D−a
1.) Case 1. F (x, y) = sin(mx + ny) or cos(mx + ny). Since
d(D1 , D2 ) sin(mx + ny) = D12 + k1 D1 D2 + k2 D22 sin(mx + ny)


= (−m2 − k1 mn − k2 n2 ) sin(mx + ny)


operating both sides by 1/d(D1 , D2 ), we find
1 1
up (x, y) = sin(mx + ny) = sin(mx + ny)
d(D1 , D2 ) −m − k1 mn − k2 n2
2

provided m2 + k1 mn + k2 n2 6= 0. Similarly, we have


1 1
up (x, y) = cos(mx + ny) = cos(mx + ny).
d(D1 , D2 ) −m − k1 mn − k2 n2
2

2.) Case 2. F (x, y) = eax+by . Since


d(D1 , D2 )eax+by = D12 + k1 D1 D2 + k2 D22 eax+by


= (a2 + k1 ab + k2 b2 )eax+by = d(a, b)eax+by


operating both sides by 1/d(D1 , D2 ), we find
1 1
up (x, y) = eax+by = 2 eax+by
d(D1 , D2 ) (a + k1 ab + k2 b2 )
provided a2 + k1 ab + k2 b2 6= 0.
3.) Case 3. F (x, y) = xm y n , where m and n are nonnegative integers. Since
1
up (x, y) = xm y n = d(D1 , D2 )−1 xm y n
d(D1 , D2 )
we expand d(D1 , D2 )−1 in ascending powers of D1 or D2 by the binomial theorem and then
operate on xm y n term by term.

js, ddw de ii: Lecture Notes Page 225 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

4.) Case 4. F (x, y) is any function of x and y. To evaluate (4.4.8) we resolve 1/d(D1 , D2 )
into partial fractions treating d(D1 , D2 ) as a function of D1 alone and operate each partial
fraction on F (x, y), remembering that
Z
1
F (x, y) = F (x, c − mx)dx (4.4.11)
D1 − mD2
where c is replaced by y + mx after integration. To show this,

Proof: We let
1
F (x, y) = φ(x, y)
D1 − mD2
so that (D1 − mD2 )φ(x, y) = F (x, y) for which
dx dy dφ
= =
1 −m F (x, y)
and hence y + mx = c, and
Z

dx = ⇔ φ(x, y) = F (x, c − mx)dx.
F (x, c − mx)

Example 4.4.3 For the partial DE

uxx − uxy = cos x cos 2y

the auxiliary equation is m2 − m = 0, m = D1 /D2 , which gives m1 = 0, m2 = 1. Hence, its


complementary solution is uc (x, y) = f1 (y) + f2 (y + x). Now for its particular solution, we have
1 1 1
up (x, y) = cos x cos 2y = [cos(x + 2y) + cos(x − 2y)]
D12 − D1 D2 2
2 D1 − D1 D2
 
1 1 1
= cos(x + 2y) + cos(x − 2y)
2 −1 + 2 −1 − 2
1 1
= cos(x + 2y) − cos(x − 2y)
2 6
Alternatively, using partial fractions for the particular solutions
1 1
up (x, y) = 2
[cos(x + 2y) + cos(x − 2y)]
2 D1 − D1 D2
1 1
= [cos(x + 2y) + cos(x − 2y)]
2 D1 (D1 − D2 )
 
1 1 1
= · [cos(x + 2y) + cos(x − 2y)]
2 D1 (D1 − D2 )
Z 
1 1
= · [cos[x + 2(c − x)] + cos[x − 2(c − x)]] dx by (4.4.11)
2 D1
Z 
1 1
= · [cos(2c − x) + cos(3x − 2c)] dx
2 D1
 
1 1 1
= − sin(2c − x) + sin(3x − 2c)
2 D1 3

js, ddw de ii: Lecture Notes Page 226 of 249


4.4. METHOD OF CONSTANT COEFFICIENT - - SECOND ORDER
PDES
 
1 1 1
= − sin(x + 2y) + sin(x − 2y)
2 D1 3
 
1 1 1
= − sin(x + 2y) + sin(x − 2y)
2 (D1 − 0D2 ) 3
Z  
1 1
= · − sin(x + 2c) + sin(x − 2c) dx by (4.4.11)
2 3
   
1 1 1 1
= · cos(x + 2c) − sin(x − 2c) = · cos(x + 2y) − sin(x − 2y)
2 3 2 3

Hence, the general solution is


1 1
ug = uc + up ⇔ u(x, y) = f1 (y) + f2 (y + x) + cos(x + 2y) − cos(x − 2y)
2 6
Example 4.4.4 For the partial differential equation

uxx − 4uxy + 4uyy = e2x+y

the auxiliary equation is (m − 2)2 = 0, m = D1 /D2 , which gives m1 = 2, m2 = 2. Hence, its


complementary function is uc (x, y) = f1 (y + 2x) + xf2 (y + 2x). Now for its particular solution
1 1
up(x, y) = e2x+y = e2x+y
D12 2
− 4D1 D2 + 4D2 (D1 − 2D2 )2

clearly Case 2 fails as (D1 − 2D2 ) = 0. However, we can apply Case 4. For this we note that
for the equation (D1 − 2D2 )v = e2x+y the solution is
Z Z
v(x, y) = F (x, c − mx)dx = e2x+(c−2x) dx = xec = xe2x+y

and since (D1 − 2D2 )up = v = xe2x+y , the particular solution is


Z Z
1 1
up (x, y) = F (x, c − mx)dx = xe2x+(c−2x) dx = x2 ec = x2 e2x+y .
2 2
Alternatively, using partial fractions for the particular solutions
 
1 2x+y 1 1 2x+y
up (x, y) = e = · e
(D1 − 2D2 )2 (D1 − 2D2 ) (D1 − 2D2 )
Z 
1 2x+(c−2x)
= · e dx by (4.4.11)
(D1 − 2D2 )
1  2x+y
= xe
(D1 − 2D2 )
Z
= xe2x+(c−2x) dx by (4.4.11)
1 2 2x+y
= xe
2

Hence, the general solution is


1
u(x, y) = f1 (y + 2x) + xf2 (y + 2x) + x2 e2x+y .
2

js, ddw de ii: Lecture Notes Page 227 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

Example 4.4.5 For the partial differential equation


uxx + uxy − 6uyy = y cos x
the auxiliary equation is m2 + m − 6 = 0, m = D1 /D2 , which gives m1 = −3, m2 = 2. Hence, its
complementary solution is uc (x, y) = f1 (y − 3x) + f2 (y + 2x). Now for its particular solution,
we have
1
up (x, y) = y cos x
(D1 − 2D2 )(D1 + 3D2 )
first we solve the equation (D1 + 3D2 )v = y cos x, to find
Z
v(x, y) = [(c + 3x) cos x] dx = (c + 3x) sin x + 3 cos x = y sin x + 3 cos x.

Now since (D1 − 2D2 )up = v = y sin x + 3 cos x, the particular solution is
Z
up (x, y) = [(c − 2x) sin x + 3 cos x] dx = (c − 2x)(− cos x) − (−2)(− sin x) + 3 sin x
= sin x − y cos x.
Alternatively, using partial fractions for the particular solutions
 
1 1 1
up (x, y) = y cos x = · y cos x
(D1 − 2D2 )(D1 + 3D2 ) (D1 − 2D2 ) (D1 + 3D2 )
Z 
1
= · [(c + 3x) cos x] dx by (4.4.11)
(D1 − 2D2 )
1
= {y sin x + 3 cos x}
(D1 − 2D2 )
Z
= [(c − 2x) sin x + 3 cos x] dx by (4.4.11)
= sin x − y cos x
Hence, the general solution is
ug = uc + up
u(x, y) = f1 (y − 3x) + f2 (y + 2x) + sin x − y cos x
Example 4.4.6 Solve the inhomogeneous partial differential equation
uxxx − 2uxxy = 2e2x + 3x2 y
the auxiliary equation is m3 − 2m2 = 0, m = D1 /D2 , which gives m1 = 0, m2 = 0, m3 = 2.
Hence, its complementary function is
uc (x, y) = f1 (y) + xf2 (y) + f3 (y + 2x).
For the particular solutions
1 1
up (x, y) = 3 2
2e2x + 3x2 y = 2 2e2x + 3x2 y
(D1 − 2D1 D2 ) D1 (D1 − D2 )
   
1 1 2x 2 1 2x 3 3 4
= · 2e + 3x y = 2 · e + (x + y)x − x
D12 (D1 − D2 ) D 4
   1 
1 1 1 1 1 2x 1 4 1 1 1 1 6
= · e2x + x3 y + x4 = e + x y + x5 = e2x + x5 y + x
D1 D1 4 D1 2 4 20 4 20 120
Hence, the general solution is
1 1 1 6
u(x, y) = f1 (y) + xf2 (y) + f3 (y + 2x) + e2x + x5 y + x
4 20 120

js, ddw de ii: Lecture Notes Page 228 of 249


4.4. METHOD OF CONSTANT COEFFICIENT - - SECOND ORDER
PDES

Exercise 4.45 Solve the following second, third and fourth-order linear partial DEs:

1.) uxx + uxy − 2uyy = 0 7.) uxxx + uxxy − uxyy − uyyy = 0


2.) uxx − 5uxy + 6uyy = ex+y
8.) uxxx − 3uxxy + 4uyyy = ex+2y
3.)
uxx − 2uxy + uyy = sin x 9.)
uxxx − 2uxxy = 2e2x + 3x2 y
4.)
uxx + 4uxy − 5uyy = y 2
10.)
5.) uxxxx − uyyyy = 0
uxx − uxy − 6uyy = xy
6.) 11.)
uxx − uyy = sin x cos 2y uxxxx − 2uxxyy + uyyyy = 0

1.) u = f1 (y + x) + f2 (y − 2x) 2.) u = f1 (y + 2x) + f2 (y + 3x) + 21 ex+y

3.) u = f1 (y + x) + xf2 (y + x) − sin x

4.) u = f1 (y + x) + f2 (y − 5x) + 12 x2 y 2 − 43 x3 y + 74 x4

5.) u = f1 (y − 2x) + f2 (y + 3x) + 16 x3 y + 1 4


24
x

6.) u = f1 (y) + f2 (y + x) + 13 (sin x cos 2y + 2 cos x sin 2y)

7.) u = f1 (y + x) + f2 (y − x) + xf3 (y − x)
1 x+2y
8.) u = f1 (y − x) + f2 (y + 2x) + xf3 (y + 2x) + 27
e

9.) u = f1 (y) + xf2 (y) + f3 (y + 2x) + 41 e2x + 1 5


20
xy + 1 6
60
x

10.) u = f1 (y + x) + f2 (y − x) + f3 (y + ix) + f4 (y − ix)

11.) u = f1 (y + x) + xf2 (y + x) + f3 (y − x) + xf4 (y − x)

js, ddw de ii: Lecture Notes Page 229 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

4.5 Method of Laplace Transforms - - Second Order PDEs


With appropriate regularity assumptions
  Z ∞ Z ∞ 
∂f −st ∂f ∂ −st ∂ ˆ ∂
L = e dt = e f dt = f= F (s)
∂x 0 ∂x ∂x 0 ∂x ∂x

Thus with L {f (t)} = F (s) = fˆ(s)

1. L {fx } = Fx = fˆx 4. L {fyy } = Fyy = fˆyy


2. L {fxx } = Fxx = fˆxx
3. L {fy } = Fy = fˆy 5. L {fxy } = Fxy = fˆxy

Example 4.5.1 Use the Laplace transform to solve the following partial differential equation
problem
∂ 2u 2
2∂ u
− c = 0 x ≥ 0, t ≥ 0
∂t2 ∂x2
∂u
(0, t) = 0 t≥0
∂x
∂u
u(x, 0) = 0, (x, 0) = e−ax x ≥ 0, a ≥ 0
∂t
Taking Laplace transforms on both sides
∂ 2 û  ∂ 2 û ∂ 2 û
s û − su(0) − u0 (0) − c2 2 = s2 û − 0 − e−ax − c2 2 = s2 û − e−ax − c2 2 = 0
 2  
∂x ∂x ∂x
∂ û
This can be viewed as a second order ODE with boundary conditions (0, s) = 0 and û(x, 0) = 0.
∂t
Rewriting the equation as
d2 û s2 1 −ax
− û = − e ,
dx2 c2 c2
the general solution can be seen to take the form
 sx   sx  e−ax
û(x, s) = c1 exp + c2 exp − + 2 2 .
c c c a − s2
For H, the heaviside function, with Laplace inverse, it can be shown that
1  x h   x  i e−ax
u(x, t) = H t − cosh ca t − −1 − sinh(cat).
ca c c ca
Exercise 4.46 Consider the initial value problem:
utt + ω 2 u = 0, u(0) = 0, ut (0) = ω
 
2 2
 −1 ω
s +ω û = ω ⇒ u = L = sin(ωt)
s + ω2
2

Exercise 4.47 Compute the Laplace solution of the equation


∂ 2u ∂ 2u
− 2 = 0 x ≥ 0, t ≥ 0
∂t2 ∂x
∂u
(0, t) = 0 t≥0
∂x
∂u
u(x, 0) = 0, (x, 0) = 1 − H(x − 1) x≥0
∂t

js, ddw de ii: Lecture Notes Page 230 of 249


4.5. METHOD OF LAPLACE TRANSFORMS - - SECOND ORDER
PDES

Example 4.5.2 Solve the second order differential equation

∂u ∂ 2u
= 0 < x < 2, t > 0
∂t ∂x2
u(0, t) = u(2, t) = 0 t>0
u(x, 0) = 3 sin(2πx) 0<x<2

Take the Laplace transform and apply the initial condition



û(x, s) = sû(x, s) − u(x, 0) = sû(x, s) − 3 sin(2πx)
∂t
to generate equation as a non-homogeneous, second order linear constant coefficient equation

∂2
û(x, s) − sû(x, s) = 3 sin(2πx)
∂x2
Equivalent to an ordinary differential equation

d2
û(x, s) − sû(x, s) = 3 sin(2πx)
dx2
The general solution can be written as

ûg (x, s) = ûc (x, s) + ûp (x, s)

where ûc (x, s) is the general solution of the homogeneous problem


√ √
ûc (x, s) = c1 e sx
+ c2 e− sx

and ûp (x, s) is any particular solution of the non-homogeneous problem


3
ûp (x, s) = A cos(2πx) + B sin(2πx) = sin(2πx)
s + 4π 2
using the Laplace transform of the boundary conditions ûp (0, s) = 0, ûp (2, s) = 0 would imply
that c1 = 0, c2 = 0.
√ √ 3 3
ûg (x, s) = c1 e sx
+ c2 e− sx
+ sin(2πx) = sin(2πx)
s + 4π 2 (s + 4π 2 )
To find our solution we apply the inverse Laplace transform
 
−1 3 2
u(x, t) = L 2
sin(2πx) = 3e−4π t sin(2πx).
(s + 4π )

Exercise 4.48 Solve the partial differential equation

∂ 2u ∂ 2u
= + sin(πx) 0 < x < 1, t > 0
∂t2 ∂x2
u(0, t) = u(1, t) = 0 t>0
u(x, 0) = ut (x, 0) = 0 0<x<1

∂2 d2
û(x, s) = û(x, s) = s2 û(x, s) − su(x, 0) − ut (x, 0) = s2 û(x, s)
∂t2 dt2
d2 sin(πx)
⇒ s2 û(x, s) = 2 +
dx s
js, ddw de ii: Lecture Notes Page 231 of 249
CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

sin(πx)
û(x, s) = c1 esx + c2 e−sx +
s(s2 + π 2 )
sin(πx)
=
s(s2 + π 2 )
     
−1 sin(πx) −1 1 1 −1 1 s
u(x, t) = L =L sin(πx) = 2 L − sin(πx)
s(s2 + π 2 ) s(s2 + π 2 ) π s (s2 + π 2 )
1
= [1 − cos(πt)] sin(πx)
π2

js, ddw de ii: Lecture Notes Page 232 of 249


4.6. CHAPTER QUESTIONS

4.6 Chapter Questions


4.6.1 Solving First Order PDEs Questions
Exercise 4.49 Show that there exists a unique solution for the system
ux = 3x2 y + y
uy = x3 + x
satisfying u(0, 0) = 0 but the the system
ux = 2.99x2 y + y
u y = x3 + x
has no solution at all.

Integrating the second equation with respect y, one finds that u(x, y) = (x3 + x)y + A(x). Then
ux = (3x2 + 1)y + A0 (x) and using the first equation, one finds that A0 (x) = 0. Hence the
function A is constant. say A(x) ≡ a. Hence u(x, y) = (x3 + x)y + a, Then u(0, 0) = a = 0,
implies that the solution of the system is given by u(x, y) = (x3 + x)y.

As in first system after integrating the second equation one finds that u(x, y) = (x3 +x)y+A(x).
ux = (3x2 + 1)y + A0 (x) and using the first equation, one finds that A0 (x) = −0.01x2 y. Now,
this is impossible since 0.01x2 y is a function of two variables x and y while A0 (x) only depends
on x. Consequently, the system doesn’t have a solution.
Exercise 4.50 Consider the equation
yux − xuy = 0 for (x, y) ∈ R × (0, ∞).
Check for each of the following initial conditions whether the problem is solvable. If it is
solvable, find a solution. If not, explain why.
(a) u(x, 0) = x2 .
u(x, y) = x2 + y 2 the only solution of the P.D.E.
(b) u(x, 0) = x.
There are no solution for this problem
(c) u(x, 0) = x, x > 0.
p
u(x, y) = x2 + y 2 is a solution of the P.D.E.

Exercise 4.51 Consider the initial value problem


∂u ∂u
x2 + y2 = u2 , u(x, 2x) = x2 for all x ∈ R
∂x ∂y
Find a function u which solves the initial value problem using the method of characteristics.

The characteristic curves are


ξ 2ξ ξ2
x(t, ξ) = , y(t, ξ) = , x(t, ξ) =
1 − ξt 1 − 2ξt 1 − ξ2t
which leads to
x2 y 2
u(x, y) =
4(y − x)2 − xy(y − 2x)

js, ddw de ii: Lecture Notes Page 233 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

Exercise 4.52 Let λ be a real number. Consider the PDE’s

xux + yuy = λu, (x, y) ∈ R2

(a) Find the characteristic curves for the equations.

The system of characteristic equations is given by

x0 = x, y 0 = y, u0 = λu

Consequently, the characteristic curves are

x(t) = c1 et , y(t) = c2 et , u(t) = c3 eλt .

(b) Let λ = 4. Find an explicit solution that satisfies u = 1 on the circle x2 + y 2 = 1.

Here λ = 4 and the initial curve is given by ξ 7→ (cos ξ, sin ξ, 1). So, the characteristic
curves satisfying the initial condition are

x(t, ξ) = et cos ξ, y(t, ξ) = et sin ξ, u(t, ξ) = e4t .

Now, one needs to solve x = et cos ξ and y = et sin ξ for t and ξ in terms of x and y.

Squaring both equations and then adding, one gets x2 + y 2 = e2t . Since a solution u is
defined by u(x, y) = e4t , it follows that

u(x, y) = (x2 + y 2 )2 .

(c) Let λ = 2. Find two solutions that satisfy u(x, 0) = x2 for x > 0.

Here λ = 2 and the initial curve is given by (ξ, 0, ξ 2 ). So, the characteristic curves
satisfying the initial condition are

x(t, ξ) = ξet , y(t, ξ) = 0, u(t, ξ) = ξ 2 e2t .

Note that the projection of the characteristic curves and of the initial curve are the same
(the x axis in the (x, y) plane). This means that there is no unique solution.

Also not that the characteristic curve (ξet , 0, ξ 2 e2t ) and the initial curve (ξ, 0, ξ 2 ) are the
same in R3 . This means that the P.D.E. has infinitely many solutions.

Exercise 4.53 Solve the following initial vale problems:


1.)
yux + xuy = 0
u = sin x on the circle x2 + y 2 = 1.

2.)
xux − uuy = y
u(1, y) = y.
Is the solution unique? What is the maximal domain where it is defined?
3.)
ux + u2 uy = 0
u(x, 0) = 1

js, ddw de ii: Lecture Notes Page 234 of 249


4.6. CHAPTER QUESTIONS

4.)
x2 ux + y 2 uy = u2
u(x, 2x) = 1

Exercise 4.54 Solve the following Cauchy problem,

u2x + uy + u = 0
u(x, 0) = x.

Here F (x, y, u, p.q) = p2 + p + u and the initial curve is given by

(x0 (ξ), y0 (ξ), u0 (ξ)) = (ξ, 0, ξ).

The system of characteristic equations is given by

x0 = Fp = 2p
y0 = Fq = 1
u0 = pFp + qFq = 2p2 + q (4.6.1)
p0 = −Fx − pFz = −p
q0 = −Fy − qFz = −q.

The initial conditions for the first three equations are

x(0, ξ) = ξ, y(0, ξ) = 0, u(0, ξ) = ξ.

The initial conditions p0 (ξ) and q0 for p and q should satisfy the following equations

F [x0 (ξ), y0 (ξ), u0 (ξ), p0 (ξ), q0 (ξ)] = 0, i.e., p0 (ξ)2 + q0 (ξ) + ξ = 0,

and
u00 (ξ) = x00 (ξ)p0 (ξ) + y00 (ξ)q0 (ξ), i.e., 1 = p0 (ξ).
Thus, the initial conditions for p and q are

p(0, ξ) = 1, q(0, ξ) = −(1 + ξ).

Solving the second and the last two equations of (4.6.1) with the initial conditions, we obtain

y(t, ξ) = t, p(t, ξ) = e−t , q(t, ξ) = −(1 + ξ)e−t .

Substituting to the first and the third equation and integrating, we find that

x(t, ξ) = ξ + 2 1 − e−t , u(t, ξ) = (1 + ξ)e−t − e−2t .


 

Finally, we have to invert the map (t, ξ) 7→ (x(t, ξ), y(t, ξ)). We get

t = t(x, y) = y and ξ = ξ(x, y) = x − 2 [1 − e−y ].

Hence
u(x, y) = (x − 1)e−y + e−2y .

js, ddw de ii: Lecture Notes Page 235 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

Exercise 4.55 Consider the following Cauchy problem,


yux + xuy = αu (x, y) ∈ R × R
u(x, 0) = φ(x) x∈R
Write down the characteristics and find the solution. What is the domain of the solution?
Consider the same equation but with the initial condition
u(x, x) = φ(x), x ∈ R.
For what functions φ is the problem solvable?
Exercise 4.56 Solve the following initial value problems:
1.) 
u2x − 3u2y − u = 0
u(x, 0) = x2 .

2.) 
u2y ux = 1
u(x, 0) = x.

3.)
1
 
2
u2x + u2y + (ux − x)(uy − y) = u
u(x, 0) = 1.
Exercise 4.57 Solve the following problem

ut + uux = 0 (x, t) ∈ R × (0, ∞)
u(x, 0) = sin x.
What are the characteristics of the equation? Do the characteristics intersect? For what values
of t does the classical solution exist?
Exercise 4.58 Solve the linear equation
ux + uy + u = 1
subject to the initial condition
u(x, x + x2 ) = sin x for x > 0.
The solutions (characteristics) of the equations are
x(t, ξ) = t + ξ, y(t, ξ) = t + ξ + ξ 2 , u(t, ξ) = 1 − [1 − sin ξ]e−t .
with p √ √
ξ = |ξ| = ξ 2 = y − x, t = x − ξ = x − y − x.
Consequently, √ √
u(x, y) = 1 − 1 − sin y − x e−x+ y−x
 

Exercise 4.59 Solve the semi-linear equation


aux + uy = u2
subject to the initial condition u(x, 0) = cos x for x > 0. ( a is assumed to be nonzero constant).
cos ξ cos(x − ay)
u(x, y) = =
1 − t cos ξ 1 − y cos(x − ay)

js, ddw de ii: Lecture Notes Page 236 of 249


4.6. CHAPTER QUESTIONS

Exercise 4.60 Solve the following problem


ut + uux = 0, (x, t) ∈ R × (0, ∞)
with the initial condition 
 −1 x≤0
u(x, 0) = 2x − 1 0 ≤ x ≤ 1
1 x ≥ 1.

What are the characteristics of the equation? For what values of t does the solution exist?

The characteristics are given by


x = −t + ξ : ξ ≤ 0
x=t+ξ : ξ ≥0
x = (2ξ − 1)t + ξ : 0 < ξ < 1
The characteristics don’t intersect for positive values of t, so the classical solution exists for all
t > 0 and it is given by 
 −1 x ≤ −t
2x−1
u(x, t) = −t
 2t+1
1 x ≥ t.
Exercise 4.61 Solve the Burger’s conservation equation
 
1 2
ut + u = 0, (x, t) ∈ R × R+ (4.6.2)
2 x

with respect to the the initial condition u(x, 0) = φ(x).

The conservative equation (4.6.2) can be rewritten as


ut + uux = 0, (x, t) ∈ R × R+
Example 4.6.1 Solve
xux + uy = y, u(x, 0) = x2
Our characteristic equations are
dx dy du
= =
x 1 y
dx
x
= dy
1
dy
1
= du
y
y2
ln x = y + ln c1 2
= u + c2
y2
x = c1 e y 2
= u + c2
2
⇒ c1 = xe−y c2 = y2 − u

c2 = c1 A1 + A2
y2
− u = A1 xe−y + A2

2
y2
+ A3 xe−y + A4

u =
2
Using the initial conditions, we get
y2
⇒ u(x, y) = + x2 e−2y
2
which holds ∀x, y, z.

js, ddw de ii: Lecture Notes Page 237 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

Example 4.6.2 Solve


y 
xux + (x2 + y)uy = 1 − −x u
x
u(1, y) = 0
The solution is
1  y
x− x

⇒ u(x, y) = y 1 − x
x
−x
which holds ∀x > 0, y. Note that the solution is restricted to x > 0 because t = ln x.
Example 4.6.3 Solve
xux + yuy + uz = u
u(x, y, 0) = h(x, y)
⇒ u(x, y, z) = h(xe−z , ye−z )ez ∀x, y, z.
Example 4.6.4 Compute the general solution of
(x + u)ux + (y + u)uy = 0
x+u
= const
y+u
x+u
Our second integral surface is ψ = = const. A general solution is give by
y+u
 
x+u
F u, =0
y+u
or using the implicit function theorem
 
x+u
u=f
y+u
where f is an arbitrary C 1 function.
Example 4.6.5 Find a general solution to

ux + uy = u (4.6.3)
Note that there are no initial conditions, so we will be finding integral surfaces.
Our characteristic equations are
dx dy du √
=1 =1 = u
dt dt dt
du √
Combining the first and the third gives dx = √ which gives x = 2 u + C1 . So our first
√ u
integral surface is φ = x − 2 u = const. Solving the first pair of characteristic equations gives
x = y + C2 , so our second integral surface is ψ = x − y = const. By a theorem in McOwen, we
may write one integral surface as an arbitrary C 1 function with the other integral surface as
an argument. So we have
φ = f (ψ)

x − 2 u = f (x − y)
√ f (x − y) + x
u =
2
(f (x − y) + x)2
⇒ u(x, y) =
4
where f ∈ C 1 .

js, ddw de ii: Lecture Notes Page 238 of 249


4.6. CHAPTER QUESTIONS

Exercise 4.62 Solve the following simultaneous differential equations:


dx dy du dx dy du
1.) 2
= 2 = 4.) = = 2
x y nxy u(x + y) u(x − y) x + y2
dx dy du dx dy du
2.) = = 5.) = =
mu − ny nx − lu ly − mx x(y 2 − u2 ) y(u2 − x2 ) u(x2 − y 2 )
dx dy du dx dy du
3.) = 2 = 2 6.) = =
x2 − yu y − ux u − xy x2 2
−y −u2 2xy 2xu

y−u
6.) y = c1 u, x2 + y 2 + u2 = uc2 = c1 , xy + yu + xu = c2 x−y
3.)

2.) lx + my + nu = c1 , x2 + y 2 + u2 = c2 5.) xyu = c1 , x2 + y 2 + u2 = c2

4.) x2 − y 2 − 2xy = c1 , x2 − y 2 − u2 = c2 ln xy y−x


nxy
+ c1 , u = c2 + y
1
= x
1
1.)

Exercise 4.62
Exercise 4.63 Solve the following first-order linear partial differential equations:

1.) x2 − y 2 − u2 ux + 2xyuy = 2xu 4.) y 2 ux − xyuy = x(u − 2y)




2.) (y − u)ux + (x − y)uy = u − x 5.) x2 ux + y 2 uy = (x + y)u


2y + u 4y + 2u
3.) (x2 − yu)ux + (y 2 − ux)uy = u2 − xy 6.) ux + uy =
x x
x4 y−u
2y − u = 0.
(2y+u)
, 6.) f xy + yu + xu = 0 x−y
, 3.) φ
   
u u
5.) φ xy , x−y = 0 2.) 21 x2 + yu = φ(x + y + u)

4.) x2 + y 2 = f 2 ln y + uy u
1.) x2 + y 2 + u2 = uf
   y

Exercise 4.63
Exercise 4.64 Solve the following first-order nonlinear partial differential equations:

1.) x(ux )2 + y(uy )2 = u 4.) xy(ux ) + (ux )(uy ) + y(uy ) = yu

2.) xy(ux )(uy ) = u2 5.) u2 (ux )2 + (uy )2 = x2 + y 2


 

3.) 1 + (ux )2 = u(uy ) 6.) (ux )2 + (uy )2 y = u(uy )


 

Exercise 4.64 hints and solutions


√ √ 2 √ 2 1
1.) u = ( a + x ) + b + y 2.) u = axb y y

u2
u√ √ 
3.) 2
± 2
u2 − 4a2 − 2a2 ln u + u2 − 4a2 = 2ax + 2y + b
4.) ln(a − ax) = y − a ln(a + y) + b
p  √ p n p o
5.) Put u2 = U, u2 = b+x (x2 + a2 )+a ln x + x2 + a2 +y (y 2 − a2 )+a ln y + y 2 − a2

1 d (u2 − c2 y 2 )
6.) p = cdx ⇒ u2 = (a + cx)2 + c2 y 2
2 u −c y
2 2 2

js, ddw de ii: Lecture Notes Page 239 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

4.6.2 Solving Second Order PDEs Questions


Example 4.6.6 Find a solution of the Cauchy problem,

utt + 3uxt − 4uxx = 0 (x, t) ∈ R × (0, ∞)


u(x, 0) = φ(x) x∈R
ut (x, 0) = ψ(x) x ∈ R.

using the method of characteristics.

The discriminant of the equation is positive since

B 2 − 4AC = 32 − 4(−4)(1) = 25 > 0,

hence the equation is hyperbolic. Solve µ2 + 3µ − 4 = 0 to get two real solutions, µ1 = −4 and
µ2 = 1. There are two characteristic equations
dx dx
= 4 and = −1
dt dt
whose solutions are
x − 4t = c1 and x + t = c2 .
Define the change of coordinates by ξ = x − 4t and η = x + t. Then

ξx = 1, ξt = −4, ηx = 1, ηt = 1.

Setting u(x, t) = ν(ξ, η) = ν(x − 4t, x + t), one finds that

ux = νξ + νη
ut = −4νξ + νη
uxx = νξξ + 2νξη + νηη
uxt = −4νξξ + 2νξη + νηη
utt = 16νξξ − 3νξη + νηη .

Substituting into the equation,

0 = utt + 3uxt − 4uxx


= (16νξξ − 3νξη + νηη ) + 3(4νξξ + 2νξη + νηη ) − 4(νξξ + 2νξη + νηη )
= −5νξη .

The general solution of the equation νξη = 0 is equal to

ν(ξ, η) = F (ξ) + G(η)

for any two twice continuously differentiable functions F and G. Hence, the solution u is equal
to
u(x, t) = F (x − 4t) + G(x + t)
where F and G are as above. The derivative ut (x, t) is equal to

ut (x, t) = −4F 0 (x − 4t) + G0 (x + t).

Hence the initial conditions give

F (x) + G(x) = φ(x)

js, ddw de ii: Lecture Notes Page 240 of 249


4.6. CHAPTER QUESTIONS

−4F 0 (x) + G0 (x) = ψ.


Integrating the second equation one gets
Z x
−4F (x) + G(x) = ψ(y)dy + C
0

where C = −4F (0) + G(0). So we have a system of equations for F and G


F (x) + G(x) = φ(x)
Z x
−4F (x) + G(x) = ψ(y)dy + C.
0
Subtracting the second equation from the first and dividing by 5 gives
Z x
1 C
F (x) = φ(x) − ψ(y)dy − .
5 0 5
Multiplying the first equation by 4, adding to the second and dividing by 5 gives,
1 x
Z
1 C
G(x) = φ(x) + ψ(y)dy + .
5 5 0 5
Consequently,
Zx+t
φ(x − 4t) + φ(x + t) 1
u(x, t) = F (x − 4t) + G(x + t) = + ψ(y)dy.
5 5
x−4t

Example 4.6.7 By Canonical method (method of characteristics), solve the following Cauchy
problem:
utt − 4uxx = ex + sin t (x, t) ∈ R × (0, ∞)
u(x, 0) = 0 x∈R
1
ut (x, 0) = 1+x2 x∈R
The solution u of is given by
x+2t Z t "Z x+2(t−s)
#
φ(x + 2t) + φ(x − 2t) 1
Z
1
u(x, t) = + ψ(y)dy + f (y, s)dy ds
2 4 x−2t 4 0 x−2(t−s)

1
where φ = 0, ψ = 1+x2
and f (y, s) = ey + sin s. So,
,
Z "Z x+2(t−s) #
1 x+2t 1 1 t
Z
u(x, t) = dy + ey + sin s dy ds.
4 x−2t 1 + y 2 4 0 x−2(t−s)

The first integral is equal to


x+2t
1 x+2t 1 tan−1 (x + 2t) − tan−1 (x − 2t)
Z
1
2
dy = tan−1 y = .
4 x−2t 1 + y 4 x−2t 4
The second integral is equal to
Z "Z x+2(t−s) #
1 t 1 t
Z
y
4(t − s) sin s ds + ex+2(t−s) − ex−2(t−s) ds

e + sin s dy ds =
4 0 x−2(t−s) 4 0
ex 2t
e + e−2t − et − e−t .

= t − sin t +
4

So the solution is equal to


tan−1 (x + 2t) − tan−1 (x − 2t) ex 2t
e + e−2t − et − e−t .

u(x, t) = + t − sin t +
4 4
js, ddw de ii: Lecture Notes Page 241 of 249
CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

Example 4.6.8 Find the formal solution of the problem by Fourier series (separation of
variables) method


 utt − uxx = 0 0 < x < π, t > 0
u(0, t) = u(π, t) = 0 t≥0


 u(x, 0) = sin3 x 0≤x≤π
ut(x, 0) = sin 2x 0 ≤ x ≤ π.

Show that the above solution is a classical solution.

The formal solution u is given by


X
u(x, t) = [An cos nt + Bn sin nt] sin nx.
n≥1

Note that
X X
u(x, 0) = sin3 x = An sin nx and ut (x, 0) = sin 2x = (nBn ) sin x.
n≥1 n≥1

From the second equality we get that 2B2 = 1 and nBn = 0 for all n ≥ 2. So, B2 = 1/2 and
Bn = 0 for all n 6= 2. The coefficient An is given by
2 π 3
Z
An = sin x sin nxdx.
π 0
Using sin3 x = 14 [3 sin x − sin 3x]
Z π Z π
2 3 1 2
An = sin x sin nx dx − [3 sin x − sin 3x] sin x sin nx dx
π 0 0 4 π
 3
3
Z π
1
Z π  4, n = 1,
= sin x sin nx dx − sin 3x sin x sin nx dx = −1, n = 3,
2π 0 2π 0  4
0, otherwise.
Hence
3 1 1
cos t sin x + sin 2t sin 2x − cos 3t sin 3x.
u(x, t) =
4 2 4
Clearly u is smooth and after differentiating twice with respect to t and with respect to x and
then substituting to the equations, one sees that u solves the equation. Hence u is a classical
solution.
Exercise 4.65 Solve the heat equation with Dirichlet (constant) boundary conditions by
separation of variables
ut − uxx = 0, 0 < x < π, t > 0
0, t ≥ 0
u(0, t) = u(π, t) = 
x, 0 ≤ x ≤ π/2
u(x, 0) = φ(x) =
π − x, π/2 ≤ x ≤ π

∞ ∞
X 4 X (−1)n+1
−n2 t 2
u(x, t) = Bn (sin nx) e = 2
sin [(2n − 1)x] e−(2n−1) t
n=1
π n=1 (2n − 1)
Z π
2 π/2 2 π
Z Z
2 4 nπ
Bn = φ(x) sin nxdx = x sin nxdx + (π − x) sin nxdx = 2
sin .
π 0 π 0 π π/2 πn 2

nπ 0, n = 2k
sin = k+1 (4.6.4)
2 (−1) , n = 2k − 1

js, ddw de ii: Lecture Notes Page 242 of 249


4.6. CHAPTER QUESTIONS

Example 4.6.9 By separation of variables, solve




 ut − uxx = 0 −π < x < π, t > 0
u(−π, t) = u(π, t) = 0 t≥0


 ux (−π, t) = ux (π, t) t≥0
u(x, 0) = φ(x) −π < x < π.

where 
1, − π < x < 0,
φ(x) =
0, − 0 < x < π.
The formal solution u has the form
2
X
u(x, t) = A0 + [Bn cos nt + Cn sin nt] e−n t ,
n≥1

where
Z π Z π Z π
1 1 1
A0 = φ(x)dx, Bn = φ(x) cos nx dx, Cn = φ(x) sin nx dx.
2π −π 2π −π 2π −π

We compute A0 = 21 ,

1 π 1 0
Z Z
Bn = φ(x) cos nx dx = cos nx dx = 0,
π −π π π
1 π 1 0
Z Z
1
Cn = φ(x) sin nx dx = sin nx dx = − [cos nx]0−π
π −π π −π nπ

1 0, n is even,
= [(−1n ) − 1] = 2
nπ − nπ , n is odd.
Consequently,
1 2X 1 2
u(x, t) = − sin(2n − 1)xe−(2n−1) t .
2 π n≥1 2n − 1

Remark 4.6.1 Periodic boundary conditions are of the form

u(−L, t) = u(L, t) and ux (−L, t) = ux (L, t) for all t ≥ 0. (4.6.5)

Exercise 4.66 Solve the problem


utt − 4uxx = (1 − x) cos t 0 < x < π, t > 0,
ux (0, t) = cos t − 1, ux (π, t) = cos t t ≥ 0,
x2
u(x, 0) = 2π 0 ≤ x ≤ π,
ut (x, 0) = cos 3x 0 ≤ x ≤ π.

x2 2t2 1
u(x, t) = + (1 − cos t) + + sin 6t cos 3x.
2π π 6
Remark 4.6.2 The particular solution is not ν = ν(x) nor ν = ν(t) but for ν = ν(x, t). See
Section 4.4.2.
Exercise 4.67 Solve the problem
ut − uxx = e−t sin 3x 0 < x < π, t > 0
u(0, t) = 0, u(π, t) = 1 t≥0
u(x, 0) = φ(x) 0≤x≤π

js, ddw de ii: Lecture Notes Page 243 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

Example 4.6.10 Use the method of separation of variables to find the formal solution of the
following problem.
ut − uxx − hu = 0 0 < x < π, t > 0
u(0, t) = u(π, t) = 0 t≥0
u(x, 0) = x(π − x) 0≤x≤π
where h is a real constant.

Write u(x, t) = X(x)T (t). Then


T 0 (t) X 00 (x)
−h= .
T (t) X(x)
The left side depends only on t while the right hand-side depends on x. Since they are equal,
they must be equal to some constant −λ. Moreover, the Dirichlet’s boundary condition implies
that X(0) = X(π) = 0. Thus, X and T have to be solutions of
T 0 (t) + (λ − h)T (t) = 0 (4.6.6)
and
(X 00 (x) + λX(x) = 0


X(0) = X(π) = 0
The eigenvalues of the later and the corresponding eigenfunctions are
λn = n2 , Xn (x) = sin nx, n ≥ 1.
Given λn , the solution of (4.6.6) is
2
Tn (t) = An e−(λn −h)t = An eht e−n t .
Thus we obtain a sequence of product solutions
2
un (x, t) = An eht e−n t sin nx, n ≥ 1
which we combine to form the series
2
X
u(x, t) = eht An e−n t sin nx.
n≥1

At t = 0, we have X
x(π − x) = u(x, 0) = An sin nx.
n≥1

The coefficients An are computed via the formula


" Z #
2 π π
Z Z π
2
An = x(π − x) sin nx dx = π x sin nx dx − x2 sin nx dx
π 0 π 0 0
" #π " #
4 4
= − 3 cos nx = − 3 (−1)n − 1 .
nπ nπ
0

Thus, 
0 n = 2k,
An = 8
n3 π
n = 2k − 1,
so that 2
8eht X e−(2n−1) t
u(x, t) = sin(2n − 1)x.
π n≥1 (2n − 1)3

js, ddw de ii: Lecture Notes Page 244 of 249


4.6. CHAPTER QUESTIONS

Example 4.6.11 Use the method of separation of variables to solve the telegraphic equation
with the initial boundary conditions
utt + ut − uxx = 0 0 < x < 2, t > 0
u(0, t) = u(2, t) = 0 t≥0
u(x, 0) = 0 0≤x≤2
ut (x, 0) = x 0 ≤ x ≤ 2.
We let u(x, t) = X(x)T (t). Then
T 00 (t) T 0 (t) X 00 (x)
+ = = −λ
T (t) T (t) X(x)
for some constant λ. This gives to equations
T 00 + T 0 + λT = 0 (4.6.7)
and
X 00 + λX = 0X(0) = X(2) = 0 (4.6.8)
The problem (4.6.8) has the following eigenvalues and corresponding eigenfunctions
!2
nπ nπx
λn = , Xn (x) = sin , n ≥ 1.
2 2
For given λn , we have to solve (4.6.7). Its characteristic equation is s2 + s + λs = 0. Since
1 − 4λn = 1 − n2 π 2 < 0 for all n ≥ 1, the characteristic equation has two complex solutions
√ √
1 4n 1 4n
s1 = − − i and s2 = − + i
2 2 2 2
2 2
where 4n = n π − 1. So the solutions Tn of (4.6.7) with λ = λn is given by
√ √
−t/2 4n −t/2 4n
Tn (t) = An e cos t + Bn e sin t
2 2
and with the product solution u(n(x, t) = Xn (x)Tn (x), the proposed solution for the problem
is " √ √ #
X X 4n 4n nπx
u(x, t) = un (x, t) = e−t/2 An cos t + Bn sin t sin .
n≥1 n≥1
2 2 2
At t = 0, √
−t/2
X nπx
0 = u(x, 0) = e An sin
n≥1
2
so that An = 0 for all n ≥ 1. Differentiating u with respect to t at t = 0, we get
X √ 4n √
nπx
x = ut (x, 0) = Bn sin
n≥1
2 2
form which we obtain
√ !
4n 2 2 4(−1)n+1
Z
nπx
Bn = x sin dx = , n ≥ 1.
2 2 0 2 nπ
Hence
8(−1)n+1
Bn = √
n 4n π
and so,
√ !
8e−t/2 X 1 4n t nπx
u(x, t) = √ sin sin .
π n≥1 n 4n π 2 2

js, ddw de ii: Lecture Notes Page 245 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

Example 4.6.12 Use the method of separation of variables to solve


ut = uxx − 4u 0 < x < π, t > 0
ux (0, t) = u(π, t) = 0 t≥0
u(x, 0) = x2 − π 2 0≤x≤π
Is the solution found above a classical solution?

Let u(x, t) = X(x)T (t). Then


T 0 (t) X 00 (x)
+4= = −λ.
T (t) X(x)
This gives two equations
T 0 + (4 + λ)T = 0 (4.6.9)
and
X 00 + λX = 0
X 0 (0) = X(π) = 0 (4.6.10)
For λ < 0, the general solution is
√ √
X(x) = A cosh −λx + B sinh −λx.
√ √ √
Since X 0 (x) = −λ A sinh −λx + B cosh −λx , it follows that 0 = X 0 (0) = B and 0 =
 

X(π) = A cosh −λπ so that also A = 0.

If λ = 0, then X(x) = Ax + B and the boundary conditions imply that A = B = 0. If A > 0,


then √ √
X(x) = A cos λx + B sin λx.
√ h √ √ i
Since X 0 (x) = λ −A sin λx + B cos λx , the boundary conditions imply that 0 = X 0 (0) =
√ √
λB so √that B = 0 and 0 = X(π) = A cos λπ. To get the nontrivial solution we need B 6= 0
2
and cos λπ = 0. This is the case when λ = 2n+1 2
for all n ≥ 0. So, the eigenvalues and
corresponding eigenfunctions are
 2  
2n + 1 (2n + 1)x
λn = , Xn (x) = cos for all n ≥ 0.
2 2
For given λn , the solution of (4.6.9) is
Tn (t) = e−4t e−λn t .
The product solution un (x, t) = Xn (x)Tn (t) = e−4t e−λn t cos (2n+1)x
2
so that the formal solution
of the problem is
X (2n + 1)x
u(x, t) = e−4t An e−λn t cos .
n≥1
2
At t = 0,
X (2n + 1)x
x2 − π 2 = u(x, 0) = An cos
n≥1
2
So,
π
16(−1)n 16(−1)n+1
Z
2 (2n + 1)x
An = (x2 − π 2 ) cos =− =
π 0 2 (2n + 1)3 π (2n + 1)3 π
Hence
16e−4t X (−1)n+1 −( 2n+1
2 )
2
t (2n + 1)x
u(x, t) = 3
e cos .
π n≥1 (2n + 1) 2

js, ddw de ii: Lecture Notes Page 246 of 249


4.6. CHAPTER QUESTIONS

Exercise 4.68 Solve the following problem by separation of variables


ut − uxx = 1 + x cos t 0 < x < 1, t > 0
ux (0, t) = ux (1, t) = sin t t≥0
u(x, 0) = 1 + cos(2πx) 0 < x < 1, t > 0.
We look for the solution uG = uC + uP of the form u = w(x, t) + ν(x, t) where ν solves the
Neumann (derivative equal to zero) boundary conditions. Hence we take (a function that
satisfies the BC)
νx (0, t) = νx (1, t) = sin t ⇒ ν(x, t) = x sin t.
Then w satisfies
wt − wxx = 1 + x cos t − x cos t = 1 0 < x < 1, t > 0
wx (0, t) = wx (1, t) = 0 t≥0
w(x, 0) = 1 + cos(2πx) 0 < x < 1, t > 0.
We look for eigenvalues and eigenfunctions of the problem
X 00 + λX = 0, 0<x<1
X (0) = X 0 (1) = 0.
0

The eigenvalues and the corresponding eigenfunctions of this problem are


λn = (nπ)2 , Xn (x) = cos nπx, n ≥ 0.
Then we look for the solution u in the form of an infinite series
X X
w(x, t) = Tn (t)Xn (x) = Tn (t) cos nπx.
n≥0 n≥0

Differentiating and substituting into the equation we find that


X
T 0 n(t) + (nπ)2 Tn (t) cos nπx = 1

wt − wxx = (4.6.11)
n≥0

and at t = 0
X
1 + cos 2πx = w(x, 0) = Tn (0) cos nπx. (4.6.12)
n≥0

The equation (4.6.11) gives



1, n = 0
Tn0 (t) 2
+ (nπ) Tn (t) = (4.6.13)
6 0
0, n =
and (4.6.12) gives

1, n = 0 and n = 2
Tn (0) = (4.6.14)
0, otherwise.
From (4.6.13) and (4.6.14), if n = 0, then T00 (t) = 1 implying that T0 (t) = t + 1. If n ≥ 1, then
2 π2 t
Tn (t) = An e−n
2
which, in view of (4.6.14), is equal to T2 (t) = e−4π t and Tn (t) = 0 for n ≥ 1 and different than
2. So,
2
w(x, t) = 1 + t + e−4π t cos(2πx),
and
2
u(x, t) = 1 + t + e−4π t cos(2πx) + x sin t.

js, ddw de ii: Lecture Notes Page 247 of 249


CHAPTER 4. PARTIAL DIFFERENTIAL EQUATION

Exercise 4.69 Solve the following problem (by separation of variables)


utt − c2 uxx = 0 0 < x < 1, t > 0
u(0, t) = 1, u(1, t) = 2π t≥0
u(x, 0) = x + π, ut (x, 0) = 0 0 ≤ x ≤ 1.
( )
4(1 − π) X 1
u(x, t) = w(x, t) + ν(x) = − cos(2nπct) sin(2nπx) + {1 + x[2π − 1]} .
π n≥1
2n
Exercise 4.70 Find the solution u of the reduced Helmholtz equation by separation of vari-
ables,
∆u − ku = 0 0 < x <, y < π
u(0, y) = 1, u(π, y) = 0 0≤y≤π
u(x, 0) = u(x, π) = 0 0 ≤ x ≤ π.
where k is a positive constant.
X 00 Y 00
=− + k = n2
X Y
Yn (y) = sin ny α ≥ 1
√ √
Xn (n) = An cosh n2 + k x + Bn sinh n2 + k x

An = −Bn tanh n2 + k π
√ 2 π
Z
2
2
−Bn tanh n + k π = sin nydy = [1 − cos nπ]
π o nπ
Exercise 4.71 Consider the equation
uxx + 4uxy + 4uyy = 0.
1.) Find the canonical form of the equation and the general solution u.
B 2 − 4AC = 42 − 4(1)(4) = 0 ⇒ parabolic equation

νss = 0
ν(r, s) = sφ(r) + ψ(r)
u(x, y) = xφ(y − 2x) + ψ(y − 2x)

2.) Find the solution u which satisfies u(x, 0) = sin x and uy (x, 0) = cos x.
u(x, y) = xφ(y − 2x) + ψ(y − 2x)
y − 2x y − 2x y − 2x y − 2x 3 y − 2x y − 2x
= 3x cos +3 cos − sin = y cos − sin .
2 2 2 2 2 2 2
Exercise 4.72 Consider the equation
y 5 uxx − yuyy + 2uy = 0.
1.) Find the canonical form of the equation and the general solution u.
B 2 − 4AC = 02 − 4(y 5 )(−y) = y 6 > 0 ⇒ hyperbolic equation

−2y 5 νrs = 0 ⇒ νrs = 0 ⇒ ν(r, s) = φ(r) + ψ(s)


 3   3 
y y
u(x, y) = φ +x +ψ −x
3 3
2.) Find the solution u which satisfies u(0, y) = 8y 3 and ux (0, y) = 6 for all y > 0.
 3   3   3   3 
y y y y
u(x, y) = φ +x +ψ − x = 15 +x +9 − x = 8y 3 + 6x.
3 3 3 3

js, ddw de ii: Lecture Notes Page 248 of 249


4.6. CHAPTER QUESTIONS

Exercise 4.73 Consider the equation

y 2 uxx + x2 yuyy + u = 0.

Find the canonical form of the equation.

B 2 − 4AC = 02 − 4(y 2 )(x2 y) = −4x2 y 3 ⇒ equation depends on value of y since x2 > 0

1.) y < 0 ⇒ hyperbolic equation


dy x dy x
=√ and = −√
dx −y dx −y
2 3 1 2 3 1
r = (−y) 2 − x2 and s = (−y) 2 + x2
3 2 3 2
to have
x2 x2 1
νrs + 3 νr + 3 νs − ν = 0
8(−y) 2 8(−y) 2 3x2 y 2
1
1 1 43
νrs + νr + νs − 4 4 ν = 0
6(r + s) 6(r + s) 3 3 (s − r)(r + s) 3
3
since x2 = s − r and r + s = 43 (−y) 2 .

2.) y = 0 ⇒ parabolic equation

The equation is no longer a differential equation

u = 0.

3.) y > 0 ⇒ elliptic equation


dy x dy x
= −i √ and = i√
dx y dx y
2 3 1 2 3 1
c1 = y 2 + i x2 and c2 = y 2 − i x2
3 2 3 2
we set
4 3
r(x, y) = y 2 and s(x, y) = x2
3
to have
1 1 1
νrr + νss + 2
νr + 3/2 νs + 2 2 ν = 0.
2x 4y 4x y

Finally, since r(x, y) = 34 y 3/2 and s(x, y) = x2 , we get

1 1 41/3
νrr + νss + νr + νs + 4/3 4/3 ν = 0.
2s 3r 3 r s

Remark 4.6.3 If f (x, t) is u or a constant, is still separable. That is why for Example 4.6.10
and Exercise 4.68 never bothered to take it to zero. However for Exercise 4.68 we could use
ν(x, t) = 12 x2 + x sin t but will not satisfy the boundary conditions.

js, ddw de ii: Lecture Notes Page 249 of 249

You might also like