Classification of PDEs

Classification of Quasi-Linear PDEs

Second order PDE

MA 201:Lecture 5
Classification of Quasilinear PDE on the basis of
characteristic lines
Introduction to II order PDE

IIT Guwahati MA201(2019):PDE

 Classification of PDEs
Classification of Quasi-Linear PDEs
Second order PDE

Classification of PDEs

ˆ An ODE is classified according to its order and whether it is linear or

nonlinear.

ˆ PDE models are more difficult to classify.

ˆ Order and linearity are not the only issues.

ˆ In fact the structure of the PDE dictates what type of boundary and
initial conditions are to be imposed.

IIT Guwahati MA201(2019):PDE

 Classification of PDEs
Classification of Quasi-Linear PDEs
Second order PDE

Consider the system of quasi-linear equations given below:

∂u ∂u ∂v ∂v
a1 + b1 + c1 + d1 = f1 (x, y , u, v ) (1)
∂x ∂y ∂x ∂y
∂u ∂u ∂v ∂v
a2 + b2 + c2 + d2 = f2 (x, y , u, v ) (2)
∂x ∂y ∂x ∂y
where u and v are the dependent variables, x and y are independent
variables and ai , bi , ci , di ∈ R for i = 1, 2 are the coefficients.
At any given point in the xy −space, there is a unique value of u and
a unique value of v ; moreover, the derivatives of u and v ie. ux , uy ,
vx , vy , are finite at this given point.

IIT Guwahati MA201(2019):PDE

 Classification of PDEs
Classification of Quasi-Linear PDEs
Second order PDE

ˆ Consider any point P in the xy plane.

ˆ Let us seek the lines (or directions) through P (if any exist) along
which the derivatives of u and v are indeterminant and across which
may be discontinuous.
ˆ These special lines are called characteristic lines.
To find characteristic lines: Note as the total differentials of u, v are
given by
∂u ∂u
du = dx + dy (3)
∂x ∂y
∂v ∂v
dv = dx + dy (4)
∂x ∂x

IIT Guwahati MA201(2019):PDE

 Classification of PDEs
Classification of Quasi-Linear PDEs
Second order PDE

How to calculate du, dv , dx, dy :

ˆ Let P and Q are at infinetismal distance apart. As we move from P
to Q, along a given curve, then the change in x−coordinates is
represented by dx and change in y −coordinates is represented by dy .
ˆ As u, v are continuous functions of x, y , the corresponding changes
in u, v are represented by du and dv , respectively.

e
c Characteristic curve
y

Q b
P dy
dx

f
d
a

IIT Guwahati MA201(2019):PDE

 Classification of PDEs
Classification of Quasi-Linear PDEs
Second order PDE

ˆ Note that equations (1)-(4) constitute a system of four linear

equations with four unknowns, viz., ux , uy , vx , vy which can be
written in matrix form as
    
a1 b1 c1 d1 ux f1
 a2 b2 c2 d2   uy   f2 
  vx  =  du
    
 dx dy 0 0 
0 0 dx dy vy dv
Set
   
a1 b1 c1 d1 f1 b1 c1 d1
 a2 b2 c2 d2   f2 b2 c2 d2 
[A] = 
 dx
 and [B] =  
dy 0 0   du dy 0 0 
0 0 dx dy dv 0 dx dy

IIT Guwahati MA201(2019):PDE

 Classification of PDEs
Classification of Quasi-Linear PDEs
Second order PDE

|B|
ˆ Using Cramer’s rule, we find ux = . (Similarly we get ux , vx , vy )
|A|
ˆ Now what happens if we choose to move in a direction away from
point P such that |A| = 0, and |B| = 0.
(Say along the direction ef as shown in the Fig. 1)
ˆ Then ux ( and similarly ux , vx , vy ) is indeterminant along this
direction (i.e., direction ef is a characteristic curve) and may even
be discontinuous.
ˆ Expanding the expression for |A| = 0 we get

(a1 c2 −a2 c1 )(dy )2 −(a1 d2 −a2 d1 +b1 c2 −b2 c1 )(dy )(dx)+(b1 d2 −b2 d1 )(dx)2 = 0.

IIT Guwahati MA201(2019):PDE

 Classification of PDEs
Classification of Quasi-Linear PDEs
Second order PDE

ˆ Dividing by (dx)2 , and setting

a = (a1 c2 −a2 c1 ), b = −(a1 d2 −a2 d1 +b1 c2 −b2 c1 ), c = (b1 d2 −b2 d1 )

we get
 2
dy dy
a +b + c = 0. (5)
dx dx
ˆ (5) can, in principle, be integrated to give y = y (x), which is the
equation of a characteristic curve in the xy plane and
√
dy −b ± b 2 − 4ac
=
dx 2a
gives the slope of the characteristic curve.

IIT Guwahati MA201(2019):PDE

 Classification of PDEs
Classification of Quasi-Linear PDEs
Second order PDE

Denote D := b 2 − 4ac, then we have

ˆ two real and distinct characteristic curves if D > 0. Hyperbolic
ˆ only one real characteristic curve if D = 0. Parabolic
ˆ no real characteristic curves if D < 0. Elliptic
Remark: The existence of characteristic curve(s) determines the domain
of the solution of the given system of PDE (1)-(2).
Remark:This classification of quasilinear PDE is applicable for higher
order equations also.

IIT Guwahati MA201(2019):PDE

 Classification of PDEs
Classification of Quasi-Linear PDEs
Second order PDE

Characteristic curve
y
Domain of
existence of Domain of existence of
P
solutions solutions Hyperbolic PDE
Characterstic curve

y Characterstic
Parabolic PDE
curve P Domain of existence of
solution

y
No real characteristic
Domain of curve
existence of
P Elliptic PDE
solution

IIT Guwahati MA201(2019):PDE

 Classification of PDEs
Classification of Quasi-Linear PDEs
Second order PDE

Second order PDE

ˆ Physical problems are governed by many different partial differential

equations. A few problems are governed by a single first-order PDE.
Numerous problems are governed by higher order PDEs.
ˆ Solutions of PDEs of order higher than the first exhibit distinctive
types of behaviour.
ˆ Physical problems fall into one of the following three general
classifications:
1 Propagation problems: The solution u(x, t) tends to propagate with
some coherence.
2 Diffusion problems: The solution u(x, t) tends to smear out signals.
3 Steady-state problems: The solution here is in steady-state and do
not vary in time at all.

IIT Guwahati MA201(2019):PDE

 Classification of PDEs
Classification of Quasi-Linear PDEs
Second order PDE

Propagation Problems

Wave Equation
ˆ The simplest situation to give rise to a one-dimensional wave
equation is the motion of a stretched string- specifically the
transverse vibrations of a string such as the string of a musical
instrument.
ˆ Assume that a string is placed along the x-axis, and then fixed at
ends x = 0 and x = L; it is then deflected and at some instant,
which we call t = 0, is released and allowed to vibrate.
ˆ The quantity of interest is the deflection u of the string at any point
x, 0 ≤ x ≤ L and at any time t > 0.
ˆ We write u = u(x, t).

IIT Guwahati MA201(2019):PDE

 Classification of PDEs
Classification of Quasi-Linear PDEs
Second order PDE

The following diagram shows a possible displacement of the string at a

fixed time t.

u(x,t)

0 L
x

Figure: x versus u(x,t)

IIT Guwahati MA201(2019):PDE

 Classification of PDEs
Classification of Quasi-Linear PDEs
Second order PDE

Subject to various assumptions...

1. neglecting damping forces such as air resistance.
2. neglecting the weight of the string.
3. assuming that the tension in the string is tangential to the curve of
the string at any time.
4. assuming that the string performed small transverse oscillations i.e.,
every particle of the string moves strictly vertically and such that its
deflection and slope s every point on the string is small.
..., it can be shown, by applying Newton’s Law of motion to a small
segment of the string, that u satisfies the PDE

∂2u 2
2∂ u
= c (6)
∂t 2 ∂x 2
where c 2 = Tρ , ρ being the mass per unit length and T being the
horizontal component of the tension in the string.

IIT Guwahati MA201(2019):PDE

 Classification of PDEs
Classification of Quasi-Linear PDEs
Second order PDE

ˆ To determine u(x, t) uniquely, we must also know

(a) the initial displacement of the string at the time t = 0 at which it is
released.
(b) the initial velocity of the string.
ˆ Thus we must be given

u(x, 0) = f (x)( Initial position)

∂u
(x, 0) = g (x)( Initial velocity)
∂t
The PDE (6) is the (undamped) wave equation. We will discuss solution
of it for various initial conditions. We also consider the following wave
equations which arise under different assumptions:
∂2u 2
(a) ∂t 2= c 2 ∂∂xu2 − g if the weight of the string is also taken into
consideration.
2 2
(b) ∂∂tu2 = c 2 ∂∂xu2 − α ∂u
∂t if a damping force proportional to the velocity of
the string (with a damping constant α) is included.

IIT Guwahati MA201(2019):PDE

 Classification of PDEs
Classification of Quasi-Linear PDEs
Second order PDE

Diffusion Problems
The Heat Conduction Equation

ˆ Consider a long thin bar, or wire, of length L and constant

cross-section and of homogeneous material oriented along x-axis.
ˆ Imagine that the bar is thermally insulated laterally and is
sufficiently thin that heat flows (by conduction) only in x-direction.
ˆ Then the temperature u at any point in the bar depends only on the
x-coordinate of the point and the time t.

x=0 Heat conduction in a thin rod of length L x=L

IIT Guwahati MA201(2019):PDE

 Classification of PDEs
Classification of Quasi-Linear PDEs
Second order PDE

ˆ By applying the principle of conservation of energy, it can be shown

that the temperature at point x and at time t, given by u(x, t),
satisfies the following PDE:

∂u ∂2u
= k 2 , 0 < x < L, t > 0. (7)
∂t ∂x
ˆ where k is a positive constant. In fact k, also known as thermal
diffusivity of the bar, is given by
κ
k=
sρ
where

κ = thermal conductivity of the material of the bar,

s = specific heat capacity of the material of the bar and
ρ = density of the material of the bar.

The PDE (7) is called the one dimensional heat conduction equations
(or, in other context where it arises, diffusion equation).
IIT Guwahati MA201(2019):PDE
 Classification of PDEs
Classification of Quasi-Linear PDEs
Second order PDE

ˆ The fact that (7) is first order in time variable t means that only one
initial condition at t = 0 is needed, together with two boundary
conditions, to obtain a unique solution.
ˆ The usual initial condition u(x, 0) = f (x), where f (x) is a known
function, specifies the initial temperature distribution in the bar.
ˆ Various types of boundary conditions at x = 0 and x = L is possible.
For example,
(a) u(0, t) = T1 and u(L, t) = T2 (i.e., the ends of the bar are at
constant temperatures T1 and T2 )
(b) ∂u
∂x
(0, t) = 0 = ∂u
∂x
(L, t) which are insulation condition as there is no
heat flow through the ends of the bar.

IIT Guwahati MA201(2019):PDE

 Classification of PDEs
Classification of Quasi-Linear PDEs
Second order PDE

Two dimensional heat conduction problem

ˆ Consider a thin plate, insulated on its top and bottom surfaces so
that no heat flow occurs other than Oxy plane as shown in the
figure.
ˆ The two dimensional heat conduction problem in the plate is
governed by the following equation:
∂ 2 u(x, y , t) ∂ 2 u(x, y , t)
 
∂u(x, y , t)
=k 2
+ , 0 < x < a, 0 < y < b, ; t > 0.
∂t ∂x ∂y 2
(8)
ˆ Here u(x, y , t) represents the temperature on the plate at the point
(x, y ) and at the time t.

y=b

x
0 x=a
IIT Guwahati MA201(2019):PDE
 Classification of PDEs
Classification of Quasi-Linear PDEs
Second order PDE

ˆ Boundary conditions and initial condition are needed to determine

the solution of (8) uniquely.
ˆ A typical boundary condition might be:
u(x, 0, t) = T1 , 0 ≤ x ≤ a( bottom side is kept at a constant temperature
∂u
(a, y , t) = 0, 0 ≤ y ≤ b( the right hand side is being insulated )
∂x
u(x, b, t) = T2 , 0 ≤ x ≤ a( top side is kept at a constant temperature )
u(0, y , t) = 0, 0 ≤ y ≤ b( the left hand side is kept at zero temperature)
ˆ An initial condition would be of the form u(x, y , 0) = f (x, y ) where
f is a given function.

u(x,b,t)=T2

y=b

ux(a,y,t)=0
u(0,y,t)=0

0
IITu(x,0,t)=T
Guwahati x=a
MA201(2019):PDE
1
 Classification of PDEs
Classification of Quasi-Linear PDEs
Second order PDE

Steady state problems

Laplace’s equation
ˆ Recall the two dimensional heat conduction problem:

∂ 2 u(x, y , t) ∂ 2 u(x, y , t)
 
∂u(x, y , t)
=k 2
+ , 0 < x < a, 0 < y < b, ; t > 0.
∂t ∂x ∂y 2

ˆ If the heat conduction is steady, i.e., ∂u

= 0, then the temperature
∂t
distribution across the plate is independent of the time t and hence
u solves the equation

∂2u ∂2u
+ = 0, 0 < x < a, 0 < y < b. (9)
∂x 2 ∂y 2
ˆ (9) is known as Laplace’s equation.

IIT Guwahati MA201(2019):PDE

 Classification of PDEs
Classification of Quasi-Linear PDEs
Second order PDE

Both equation (9) and its three-dimensional analogous form, viz.,

∂2u ∂2u ∂2u

+ + = 0, 0 < x < a, 0 < y < b, 0 < z < c, (10)
∂x 2 ∂y 2 ∂z 2
arise in wide variety of applications, quite apart from the steady state of
heat conduction theory.
Since time does not arise in (8) or (9) it is evident that Laplace’s
equation is always a model for equilibrium situations. Some areas in
which Laplace’s equation arises are
(a) electrostatics (u being the electrostatic potential in a charge free
region)
(b) gravitation (u being the gravitational potential in free space)
(c) steady state flow of inviscid fluids
(d) steady state heat conduction.

In any problem involving Laplace’s equation, we are interested in solving

it in a specific region R for given boundary conditions.
IIT Guwahati MA201(2019):PDE
 Classification of PDEs
Classification of Quasi-Linear PDEs
Second order PDE

Boundary Condition
ˆ The boundary condition is a set of constraints that define the behavior of
unknown functions on the spatial boundary of the domain.
ˆ A PDE with a boundary condition is also called a boundary value problem.
ˆ Four main types of boundary conditions exist:
(i) Dirichlet Condition: The Dirichlet condition specifies the values of unknown
function u on the boundary of the domain. For example,

u(0, t) = T1 (t) and u(L, t) = T2 (t).

(ii) Neumann Condition: The Neumann condition specifies the values of the normal
derivative on the boundary of the domain. For example,

ux (0, t) = T1 (t) and ux (L, t) = T2 (t).

(iii) Robin Condition: The Robin condition is a linear combination of the Dirichlet and
Neumann conditions. For example,

a1 u(0, t) + a2 ux (0, t) = T1 (t) and c1 u(L, t) + c2 ux (L, t) = T2 (t).

(iv) Mixed Condition: This condition occurs when Dirichlet condition is specified on
some part of the boundary and Neumann condition is specified on the remaining
part of the boundary. For example,

u(0, t) = T1 (t), ux (L, t) = T2 (t).

IIT Guwahati MA201(2019):PDE

 Classification of PDEs
Classification of Quasi-Linear PDEs
Second order PDE

Initial Condition

For an evolutionary( or time-dependent) equation, a boundary condition

does not guarantee a unique solution. We also need to define the initial
condition, which specifies the value of unknown function u(x, t) at the
initial time on the domain. This is called the Cauchy initial condition.
For example,
u(x, 0) = f (x).
If the evolutionary equation contains the higher order derivatives with
respect to time t, specify the additional values of derivatives up to the
highest order of the derivative of t.

IIT Guwahati MA201(2019):PDE