Professional Documents
Culture Documents
The study of partial differential equations (PDEs)is a part of mathematics that has vast applications
in the field of Applied Science and technology. This course aims to teach the basics of Partial
differential equations (PDEs), a subject that touches on many branches of pure mathematics,
applied mathematics, as well as physics and applied science. It is addressed to first and second-
year graduate students, or anyone with an interest in the topic. Partial differential equations are a
very rich subject; so much so that at a research level most workers in the field specialize in one of
the many sub-fields. The aim of this one-semester course is both to give an overview of the subject
as much as possible and introduce some tools that are used throughout. This should prepare
students adequately for the many more advanced courses in PDEs that are (and will be) offered in
the department. In this course as we know PDEs appear as mathematical models for many physical
phenomena. Closed-forum solutions to most of the PDEs cannot found. One of the possible ways
to understand the models is by studying the qualitative properties exhibited by their solutions.
Overall, In this course, we study first order linear and nonlinear PDEs, and the properties of the
three important types of second order linear PDEs (Wave, Laplace, Heat) would be studied and
compared.
Pre-requisites: The main pre-requisite is some basic real analysis, primarily Lebesgue measure on
Rn, the basics of the Fourier transform on Rn and some functional analysis, along with
multivariable calculus (Stokes' theorem). Familiarity with the general theory of Ordinary
Differential equations is desirable, but not a prerequisite. Exposure to partial derivates will be
helpful.
Focus: The focus on this course is on the general theory of PDEs. There will be some discussion
of finding explicit solutions to specific equations with specific methods. There will be no
discussion of numerical analysis related to PDEs.
The Partial Differential Equations course contains Four Module
Hyperbolic differential equations, One dimensional wave equation, Solution of the wave
equation by Method of separation of variables, Boundary and initial value problem of two
dimensional wave equation. Parabolic differential equations, One dimensional diffusion
equation, Boundary conditions; Method of separation of variables, Solutions in cylindrical
and spherical equation
Table of contents
Module -1
Unit 1.1 Basics
Unit 1.2 Definitions
Module -2
Second order partial differential equations and its classifications
Unit 2.1 Classification of Second order PDEs
Unit 2.2 Constant coefficient to find homogenous and non-homogenous solution
Module -3
One dimensional elliptic equation
Unit 3.1 One dimensional Elliptic formation
Unit 3.2 Solution of Elliptic equations
Module -4
ONE DIMENSIONAL HYPERBOLIC AND PARABOLIC EQUATIONS
Unit 4.1 Formation of one dimensional hyperbolic and parabolic equations
Unit 4.2 Solution of One dimensional Hyperbolic PDEs
Unit-4.3 Solution of One dimensional parabolic PDEs
Partial Differential Equations
The reader has, already been introduced to the notion of partial differential equations.
Here, we shall begin by studying the ways in which partial differential equations are formed.
Then we shall investigate the solutions of special types of partial differential equations of the
first and higher orders.
In what follows x and y will, usually be taken as the independent variables and z, the
dependent variable so that z = f (x, y) and we shall employ the following notation:
𝜕𝑧 𝜕𝑧 𝜕 2𝑧 𝜕𝑧2 𝜕 2𝑧
= 𝑝, = 𝑞, 2 = 𝑟 = 𝑠, =𝑠
𝜕𝑥 𝜕𝑦 𝜕𝑥 𝜕𝑥𝜕𝑦 𝜕𝑥𝜕𝑦
𝜕𝑧 2𝑥 1 1 𝜕𝑧 𝑝
2 = 2 𝑜𝑟 2 = = 𝑥
𝜕𝑥 𝑎 𝑎 𝑥 𝜕𝑥
2𝜕𝑧 2𝑦 1 1 𝜕𝑧 𝑞
= 2 𝑜𝑟 2 = =
𝜕𝑦 𝑏 𝑏 𝑦 𝜕𝑦 𝑦
and
Substituting these values of I/a2 and I/b2 in (i), we get
2z=x p + yq
𝜕𝑧
𝑝= = (𝑥 + 𝑦)𝜙 ′ (𝑥 2 − 𝑦 2 ) ⋅ 2𝑥 + 𝜙(𝑥 2 + 𝑦 2 )
𝜕𝑥
𝜕𝑧
𝑞= = (𝑥 + 𝑦)𝜙 ′ (𝑥 2 − 𝑦 2 ). (−2𝑦) + 𝜙(𝑥 2 − 𝑦 2 )
𝜕𝑥
2
𝑝− = 2𝑥(𝑥 + 𝑦)𝜙 ′ (𝑥 2 − 𝑦 2 )
𝑥+𝑦
2
𝑞− = −2𝑦(𝑥 + 𝑦)𝜙 ′ (𝑥 2 − 𝑦 2 )
𝑥+𝑦
𝑝−2⁄(𝑥+𝑦) 𝑥
Division gives 𝑞−𝑧⁄(𝑥+𝑦) = 𝑦
i.e., [p(x+y)-z]y+[q(x+y)-z]x
(x+y)(py+qx)-z(x+y)=0
Hence, py+qz=z is required equation.
𝜕2𝑧 𝜕2𝑧
Thus, the desired partial differential equation is𝜕𝑡 2 = 𝑎2 𝜕𝑥 2
Example 2 Find the differential equation of all planes which are at a constant distance a from the
origin.
Solution. The equation of the plane in ‘normal form’ is
𝑙𝑥 + 𝑚𝑦 + 𝑛𝑧 = 𝑎
where l, m, n are the d.c.s of the normal! from the origin to the plane.
Then
𝑙 2 + 𝑚2 + 𝑛2 = 𝑙 𝑜𝑟 𝑛 = √(1 − 𝑙 2 − 𝑚2 )
(i) becomes
𝑙𝑥 + 𝑚𝑦 + √(1 − 𝑙 2 − 𝑚2 )𝑧 = 𝑎
𝑙 + √(1 − 𝑙 2 − 𝑚2 ). 𝑝 = 0
Differentiating partially with respect to y, we get
𝑚 + √(1 − 𝑙 2 − 𝑚2 ). 𝑞 = 0
𝑝2 +𝑞 2 1
(1 + 𝑝2 + 𝑚2 )(𝑙 2 + 𝑚2 ) = 𝑝2 + 𝑞 2 𝑜𝑟 1 − 𝑙 2 − 𝑚2 = 1 − = 1+𝑝2+𝑞2
1+𝑝2 +𝑞2
580
Also
−𝑝 −𝑞
𝑙= 𝑎𝑛𝑑 𝑚 =
√1 + 𝑝2 + 𝑞 2 √1 + 𝑝2 + 𝑞 2
Substituting the values of l,m and 1 − 𝑙 2 − 𝑚2 in (ii), we obtain
−𝑝𝑥 𝑞 1
− + 𝑧=𝑎
√1 + 𝑝2 + 𝑞 2 √1 + 𝑝2 + 𝑞 2 √1 + 𝑝2 + 𝑞 2
PROBLEMS 3
Form the partial differential equations (by eliminating the arbitrary constants) from:
5.Find the differential equation of all spheres of fixed radius having their centres in the xy-plane,
(Madras, 2000 S )
6. Find the differential equation of all spheres whose centres lie on the z-axis. {Kerela, 2005)
Form the partial differential equations (by eliminating the arbitrary constants) from
1
Thus 𝑧 = 4 𝑐𝑜𝑠(2𝑥 − 𝑦) − 𝑥 3 𝑦 3 + 𝑥𝑢(𝑦) + 𝑣(𝑦) = 𝑤(𝑥) where u,v,w are arbitrary functions.
𝜕2 𝑧 𝜕𝑧
Example . Solve𝜕𝑥 2 + 𝑧 = 0 ,given that when x=0, z= ⅇ 𝑦 and 𝜕𝑥 = 1
Solution. If z were function of x alone, the solution would have been z = A sin x + B cos x,
where A and B are constants. Since z is a function of x and y, A and B can be arbitrary functions
of y. Hence the solution of the given equation is 𝑧 = 𝑓(𝑦) 𝑠𝑖𝑛 𝑥 + 𝜙(𝑦) 𝑐𝑜𝑠 𝑥
𝜕𝑧
= 𝑓(𝑦) 𝑐𝑜𝑠 𝑥 − 𝜙(𝑦) 𝑠𝑖𝑛 𝑥
𝜕𝑥
When x=0;z=ⅇ 𝑦 ,ⅇ 𝑦 = ϕ(y).
𝜕𝑧
When x=0, 𝜕𝑥 = 1,1= 𝑓(𝑦)
𝜕2𝑧 𝜕𝑧
Example Solve = 𝑠𝑖𝑛𝑥𝑠𝑖𝑛𝑦 ,for which = −2𝑠𝑖𝑛𝑦 when x=0 and z=0 when y is an odd
𝜕𝑥𝜕𝑦 𝜕𝑦
multiple of π/2.
𝜕2𝑧
Solution. Given equation is 𝜕𝑥𝜕𝑦 = 𝑠𝑖𝑛𝑥𝑠𝑖𝑛𝑦
𝜕𝑧
(i). becomes 𝜕𝑦 = −𝑐𝑜𝑠𝑥𝑠𝑖𝑛𝑦 − 𝑠𝑖𝑛𝑦
0 = 0 + 0 + 𝑔(𝑥) 𝑜𝑟 𝑔(𝑥) = 0
PROBLEM-17.2
𝜕𝑢 𝜕𝑣 𝜕𝑢 𝜕𝑣 𝜕𝑢 𝜕𝑣 𝜕𝑢 𝜕𝑣 𝜕𝑢 𝜕𝑣 𝜕𝑢 𝜕𝑣
Which simplifies to(𝜕𝑦 𝜕𝑧 − 𝜕𝑧 𝜕𝑦) 𝑝 + ( 𝜕𝑧 𝜕𝑥 − 𝜕𝑥 𝜕𝑧 ) 𝑞 = (𝜕𝑥 𝜕𝑦 − 𝜕𝑥 𝜕𝑦)
𝑦2𝑧 2 𝑧𝑞
Example Solve 𝑝+𝑥 = 𝑦2
𝑥
𝜕𝑧 𝜕𝑧
Example Solve (𝑚𝑧 − 𝑛𝑦) 𝜕𝑥 + (𝑛𝑥 − 𝑙𝑧) 𝜕𝑦 = 𝑙𝑦 − 𝑚𝑥
Solution. Here, the subsidiary equations are
𝑑𝑥 𝑑𝑦 𝑑𝑧
= =
𝑚𝑧 − 𝑛𝑦 𝑛𝑥 − 𝑙𝑧 𝑙𝑦 − 𝑚𝑥
Using multipliers x, y, and z, we get each fraction =
𝑥 𝑑𝑥 + 𝑦𝑑𝑦 + 𝑧𝑑𝑧
0
2𝑥 𝑑𝑥 + 2𝑦𝑑𝑦 + 2𝑧𝑑𝑧 𝑑𝑧
=
𝑥(𝑥 2 + 𝑦 2 + 𝑧 2 ) 𝑧
Which on integration gives log(𝑥 2 + 𝑦 2 + 𝑧 2 ) = 𝑙𝑜𝑔𝑧 + 𝑙𝑜𝑔𝑏
𝑥2 + 𝑦2 + 𝑧2
=𝑏
𝑧
Hence from (i) and (ii), the required solution is
𝑦
𝑥 2 + 𝑦 2 + 𝑧 2 = 𝑧𝑓 ( )
𝑧
ⅆ𝑥−ⅆ𝑦 ⅆ𝑦−ⅆ𝑧
Each of these equations=𝑥 2 −𝑦 2−(𝑦−𝑥)𝑧 = 𝑦 2−𝑧 2−𝑥(𝑧−𝑦)
𝑑(𝑥 − 𝑦) 𝑑(𝑦 − 𝑧) 𝑑(𝑥 − 𝑦) 𝑑(𝑦 − 𝑧)
= 𝑜𝑟 =
(𝑥 − 𝑦)(𝑥 + 𝑦 + 𝑧) (𝑦 − 𝑧)(𝑥 + 𝑦 + 𝑧) (𝑥 − 𝑦) (𝑦 − 𝑧)
𝑥−𝑦
Which on integration gives, log(x-y)=log(y-z)+log c or 𝑦−𝑧 = 𝑐
Each of the subsidiary equations (i)=
𝑥 𝑑𝑥 + 𝑦𝑑𝑦 + 𝑧𝑑𝑧
𝑥3 + 𝑦 3 + 𝑧 3 − 3𝑥𝑦𝑧
𝑥 𝑑𝑥 + 𝑦𝑑𝑦 + 𝑧𝑑𝑧
(𝑥 + 𝑦 + 𝑧)(𝑥 2 + 𝑦 2 + 𝑧 2 − 𝑦𝑧 − 𝑧𝑥 − 𝑥𝑦)
Also, each of the subsidiary equations =
𝑥 𝑑𝑥 + 𝑦𝑑𝑦 + 𝑧𝑑𝑧
𝑥2 + 𝑦 2 + 𝑧 2 − 𝑦𝑧 − 𝑧𝑥 − 𝑥𝑦
Equating (iii) and (iv) and cancelling the common factor, we get
𝑥 𝑑𝑥 + 𝑦𝑑𝑦 + 𝑧𝑑𝑧
= 𝑑𝑥 + 𝑑𝑦 + 𝑑𝑧
𝑥+𝑦+𝑧
Or 𝑥 2 + 𝑦 2 + 𝑧 2 = (𝑥 + 𝑦 + 𝑧)2 + 2𝑐 ′ 𝑜𝑟 𝑥𝑦 + 𝑦𝑧 + 𝑧𝑥 + 𝑐 ′ = 0
Expressing (2) as b = ϕ (a) and substituting this value of b in (I), we get the required
solution as z = ax + ϕ (a)y+c in which a and c are arbitrary constants.
Example Solve 𝑥 2 𝑝2 + 𝑦 2 𝑞 2 = 𝑧 2
Solution. Given equation can be reduced to the above form by writing it as-
𝑥 𝜕𝑧 2 𝑦 𝜕𝑧 2
( ⋅ ) +( ⋅ ) =1
𝑧 𝜕𝑥 𝑧 𝜕𝑦
ⅆ𝑥 ⅆ𝑦 ⅆ𝑧
And setting = 𝑑𝑢, = 𝑑𝑣, = 𝑑𝜔 𝑠𝑜 𝑡ℎ𝑎𝑡 𝑢 = 𝑙𝑜𝑔𝑥, 𝑣 = log 𝑦, 𝑤 = log 𝑧
𝑥 𝑦 𝑧
𝜕𝑤 2 𝜕𝑤 2
Then (i) becomes ( 𝜕𝑢 ) + ( 𝜕𝑣 ) = 1
𝜕𝜔 𝜕𝜔
𝑃2 + 𝑄 2 = 1 𝑤ℎⅇ𝑟ⅇ 𝑃 = ,𝑄 =
𝜕𝑢 𝜕𝑣
Its complete solution is 𝜔 = 𝑎𝑢 + 𝑏𝑣 + 𝑐
Where 𝑎2 + 𝑏 2 = 1 𝑜𝑟 𝑏 = √(1 − 𝑎2 )
(ii) becomes 𝜔 = 𝑎𝑢 + √(1 − 𝑎2 )𝑣 + 𝑐
Or 𝑙𝑜𝑔 𝑧 = 𝑎 𝑙𝑜𝑔 𝑥 + √(1 − 𝑎2 ) 𝑙𝑜𝑔 𝑦 + 𝑐 which is the required solution
𝑑𝑧 𝑎 𝑑𝑧 𝑑𝑧 𝑎 𝑑𝑧 𝑎 𝑑𝑧
(1 + ) = 𝑎𝑧 𝑜𝑟 = 𝑎𝑧 − 1 𝑜𝑟 ∫ = ∫ 𝑑𝑢 + 𝑏
𝑑𝑢 𝑑𝑢 𝑑𝑢 𝑑𝑢 𝑎𝑧 − 1
𝑙𝑜𝑔(𝑎𝑧 − 1) = 𝑢 + 𝑏 𝑜𝑟 𝑙𝑜𝑔(𝑎𝑧 − 1) = 𝑥 + 𝑎𝑦 + 𝑏
Example Solve 𝑞 2 = 𝑧 2 𝑝2 (1 − 𝑝2 )
Solution. Setting 𝑢 = 𝑦 + 𝑎𝑥 𝑎𝑛𝑑 𝑧 = 𝑓(𝑢) , 𝑤ⅇ 𝑔ⅇ𝑡
𝜕𝑧 𝑑𝑧 𝜕𝑢 𝑑𝑧 𝑑𝑧 𝜕𝑢 𝑑𝑧
𝑝= = ⋅ =𝑎 𝑎𝑛𝑑 𝑞 = ⋅ =
𝜕𝑥 𝑑𝑢 𝜕𝑥 𝑑𝑢 𝑑𝑢 𝜕𝑦 𝑑𝑢
ⅆ𝑧 2 ⅆ𝑧 2 ⅆ𝑧 2
The given equation becomes (ⅆ𝑢) = 𝑎2 𝑧 2 (ⅆ𝑢) {1 − 𝑎2 (ⅆ𝑢) }
𝑑𝑧 2 𝑑𝑧 √(𝑎2 𝑧 2 ) − 1
𝑎 𝑧 ( ) = 𝑎2 𝑧 2 − 1
4 2
𝑜𝑟 =
𝑑𝑢 𝑑𝑢 𝑎2 𝑧
Integrating,
√(𝑎2 𝑧 2 ) − 1 1
∫ 𝑑𝑧 = ∫ 𝑑𝑢 + 𝑐 𝑜𝑟 (𝑎2 𝑧 2 − 1)2 = 𝑢 + 𝑐
𝑎2 𝑧
𝑎2 𝑧 2 = (𝑦 + 𝑎𝑥 + 𝑐)2 + 1
ⅆ𝑧
The second factor (i) is ⅆ𝑢=0. Its solution is z=c’
ⅆ𝑧 ⅆ𝑧
Putting X = log x, so that x ⅆ𝑥 = ⅆ𝑋 takes the standard form
𝜕𝑧 𝑧
2
𝜕𝑧 2
𝑧 [( ) + ( ) ] = 1
𝜕𝑋 𝜕𝑦
𝜕𝑧 𝜕𝑧 𝜕𝑧 𝜕𝑧
Let u=X+ay and put 𝜕𝑋 = 𝜕𝑢 and 𝜕𝑦 = 𝑎 𝜕𝑢 in (ii), so that
ⅆ𝑧 𝑧 𝜕𝑧 2
𝑧 2 [(ⅆ𝑢) + 𝑎2 (𝜕𝑢) ] = 1 or √(1 + 𝑎2 )𝑧𝑑𝑧 = ±𝑑𝑢
𝑑𝑧 = ϕ(x)𝑑𝑥 + ѱ(y)𝑑𝑦
Integrating,
𝑧 = ∫ ϕ(x)𝑑𝑥 + ∫ ѱ(y)𝑑𝑦 + 𝑏
which is the desired complete solution containing two constants a and b.
Example Solve 𝑧 2 (𝑝2 + 𝑞 2 ) = 𝑥 + 𝑦
Solution. Given equation is 𝑝2 − 𝑥 = 𝑦 − 𝑞 2 = 𝑎, 𝑠𝑎𝑦
𝑝2 − 𝑥 = 𝑎 gives p=√(𝑎 + 𝑥)
𝑦 − 𝑞 2 = 𝑎 gives q=√(𝑦 − 𝑎)
Substituting these values of p and q in dz=pdx+qdy, we get
𝜕𝑧 2 𝜕𝑧 2
(𝑧 ) + (𝑧 ) = 𝑥 2 + 𝑦 2
𝜕𝑥 𝜕𝑦
1
And putting 𝑧𝑑𝑥 = 𝑑𝑍, 𝑖. ⅇ. , 𝑍 = 2 𝑧 2
𝜕𝑍 𝜕𝑍 𝜕𝑧 𝜕𝑧
= . =𝑧 =𝑃
𝜕𝑥 𝜕𝑧 𝜕𝑥 𝜕𝑥
𝜕𝑍 𝜕𝑍 𝜕𝑧 𝜕𝑧
= . =𝑧 =𝑄
𝜕𝑦 𝜕𝑧 𝜕𝑦 𝜕𝑦
(i)becomes
𝑃2 + 𝑄 2 = 𝑥 2 + 𝑦 2
𝑃2 − 𝑥 2 = 𝑦 2 − 𝑄 2
𝑥 + √(𝑥 2 + 𝑎)
𝑧 2 = 𝑥√(𝑥 2 + 𝑎) + √(𝑦 2 − 𝑎) + 𝑎 𝑙𝑜𝑔 + 2𝑏
𝛾 + √(𝑦 2 − 𝑎)
This is the required solution.
Example Solve (𝑥 + 𝑦)(𝑝 + 𝑞)2 + (𝑥 − 𝑦)(𝑝 − 𝑞)2 = 1
Solution This equation can be reduced to the form 𝑓 (x,q)=F(y,q) by putting u=x+y, v=x-y and
taking z=(z(u,v)
𝜕𝑧 𝜕𝑧 𝜕𝑢 𝜕𝑧 𝜕𝑣
Then, 𝑝 = 𝜕𝑥 = 𝜕𝑢 . 𝜕𝑥 + 𝜕𝑣 . 𝜕𝑥 = 𝑃 + 𝑄
𝜕𝑧 𝜕𝑧 𝜕𝑢 𝜕𝑧 𝜕𝑣 𝜕𝑧 𝜕𝑧
And 𝑄 = 𝜕𝑦 = 𝜕𝑢 . 𝜕𝑦 + 𝜕𝑣 . 𝜕𝑦 = 𝑃 − 𝑄, 𝑤ℎⅇ𝑟ⅇ 𝑃 = 𝜕𝑢 , 𝑄 = 𝜕𝑣
1 𝑎 1 1−𝑎
𝑃 = ± .√ ,𝑄 = ± .√
2 𝑢 2 𝑣
𝜕𝑧 𝜕𝑧
𝑑𝑧 = 𝑑𝑢 + = 𝑝𝑑𝑢 + 𝑄𝑑𝑣
𝜕𝑢 𝜕𝑣
1 √𝑎 𝑑𝑢 √1 − 𝑎 𝑑𝑣
=± . . ± .
2 2 √𝑢 2 √𝑣
Integrating , we have 𝑧 = ±√𝑎√𝑢 ± √1 − 𝑎√𝑣 + 𝑏
Solution . Given equation is of the form z=px+qy+ 𝑓 (p,q) where 𝑓 (p,q)= √(1 + 𝑝2 + 𝑞 2 )
CHARPIT'S METHOD*
We now explain a general method for finding the complete integral of a non-linear partial
differential equation which is due to Charpit.
Consider the equation
𝑓(𝑥, 𝑦, 𝑧, 𝑝, 𝑞) = 0
...(1)
Since z depends on x and y, we have
𝜕𝑧 𝜕𝑧
𝑑𝑧 = 𝑑𝑥 + 𝑑𝑦 = 𝑝𝑑𝑥 + 𝑞𝑑𝑦
𝜕𝑢 𝜕𝑦
...(2)
Now if we can find another relation involving x, y, z, p, q such as +ϕ(x, y, z, p, q) = O
...(3) then we can solve (I) and (3) for p and q and substitute in (2). This will give the solution
provided (2) is integrable.
To determine ϕ, we differentiate (1) and (3) w.r.t x and y giving ,
𝜕𝑓 𝜕𝑓 𝜕𝑓 𝜕𝑝 𝜕𝑓 𝜕𝑞
+ 𝑝+ + =0
𝜕𝑥 𝜕𝑧 𝜕𝑝 𝜕𝑥 𝜕𝑞 𝜕𝑥
𝜕ϕ 𝜕ϕ 𝜕ϕ 𝜕𝑝 𝜕ϕ 𝜕𝑞
+ 𝑝+ + =0
𝜕𝑥 𝜕𝑧 𝜕𝑝 𝜕𝑥 𝜕𝑞 𝜕𝑥
𝜕𝑓 𝜕𝑓 𝜕𝑓 𝜕𝑝 𝜕𝑓 𝜕𝑞
+ 𝑝+ + =0
𝜕𝑦 𝜕𝑧 𝜕𝑝 𝜕𝑦 𝜕𝑞 𝜕𝑦
𝜕ϕ 𝜕ϕ 𝜕ϕ 𝜕𝑝 𝜕ϕ 𝜕𝑞
+ 𝑝+ + =0
𝜕𝑦 𝜕𝑧 𝜕𝑝 𝜕𝑦 𝜕𝑞 𝜕𝑦
𝜕𝑝
Eliminating 𝜕𝑥 between the equations (4) and (5), we get
𝜕𝑓 𝜕ϕ 𝜕ϕ 𝜕𝑓 𝜕𝑓 𝜕ϕ 𝜕ϕ 𝜕𝑓 𝜕𝑓 𝜕ϕ 𝜕ϕ 𝜕𝑓 𝜕𝑞
( − )+( − )𝑝 + ( − ) =0
𝜕𝑥 𝜕𝑝 𝜕𝑥 𝜕𝑝 𝜕𝑧 𝜕𝑝 𝜕𝑧 𝜕𝑝 𝜕𝑞 𝜕𝑝 𝜕𝑞 𝜕𝑝 𝜕𝑥
𝜕𝑞
Also eliminating 𝜕𝑦 between the equations (6) and (7) , we obtain
𝜕𝑓 𝜕ϕ 𝜕ϕ 𝜕𝑓 𝜕𝑓 𝜕ϕ 𝜕ϕ 𝜕𝑓 𝜕𝑓 𝜕ϕ 𝜕ϕ 𝜕𝑓 𝜕𝑞
( − )+( − )𝑞 + ( − ) =0
𝜕𝑦 𝜕𝑝 𝜕𝑦 𝜕𝑞 𝜕𝑧 𝜕𝑞 𝜕𝑧 𝜕𝑞 𝜕𝑝 𝜕𝑞 𝜕𝑝 𝜕𝑞 𝜕𝑦
𝜕𝑞 𝜕2𝑧 𝜕𝑞
Adding (8) and (9) by using = =
𝜕𝑥 𝜕𝑥𝜕𝑦 𝜕𝑦
We mean that the last terms in both cancel and the other terms , on rearrangement , give
𝜕𝑓 𝜕𝑓 𝜕ϕ 𝜕𝑓 𝜕𝑓 𝜕ϕ 𝜕𝑓 𝜕𝑓 𝜕ϕ 𝜕𝑓 𝜕ϕ 𝜕𝑓 𝜕ϕ
( +𝑝 ) +( +𝑞 ) + (−𝑝 −𝑞 ) + (− ) + (− ) =0
𝜕𝑥 𝜕𝑧 𝜕𝑝 𝜕𝑦 𝜕𝑧 𝜕𝑞 𝜕𝑝 𝜕𝑞 𝜕𝑧 𝜕𝑝 𝜕𝑥 𝜕𝑞 𝜕𝑦
𝜕𝑓 𝜕ϕ 𝜕𝑓 𝜕ϕ 𝜕𝑓 𝜕𝑓 𝜕ϕ 𝜕𝑓 𝜕𝑓 𝜕ϕ 𝜕𝑓 𝜕𝑓 𝜕ϕ
(− ) + (− ) + (−𝑝 −𝑞 ) +( +𝑝 ) +( +𝑞 ) =0
𝜕𝑝 𝜕𝑥 𝜕𝑞 𝜕𝑦 𝜕𝑝 𝜕𝑞 𝜕𝑧 𝜕𝑥 𝜕𝑧 𝜕𝑝 𝜕𝑦 𝜕𝑧 𝜕𝑞
𝑑𝑥 𝑑𝑦 𝑑𝑧 𝑑𝑝 𝑑𝑞 𝜕ϕ
= = = = =
𝜕𝑓 𝜕𝑓 𝜕𝑓 𝜕𝑓 𝜕𝑓 𝜕𝑓 𝜕𝑓 𝜕𝑓 0
− − −𝑝 −𝑞 +𝑝 +𝑞
𝜕𝑝 𝜕𝑞 𝜕𝑝 𝜕𝑞 𝜕𝑥 𝜕𝑧 𝜕𝑦 𝜕𝑧
1
. 𝑑(𝑧 2 − 𝑐 2 𝑦 2 )
𝑧𝑑𝑧 = 𝑐 𝑦𝑑𝑦 = 𝑐√(𝑧 2 − 𝑐 2 𝑦 2 )𝑑𝑥 𝑜𝑟 2
2
= 𝑐𝑑𝑥
√(𝑧 2 − 𝑐 2 𝑦 2 )
𝑑𝑞 = 0 𝑜𝑟 𝑞 = 𝑎
2𝑥(𝑧−𝑎𝑦)
Putting q=a in (i) we get p= 𝑥 2 −𝑎
Example 2𝑧 + 𝑝2 + 𝑞𝑦 + 2𝑦 2 = 0
Solution. 𝑓(𝑥, 𝑦, 𝑧, 𝑝, 𝑞) = 2𝑧 + 𝑝2 + 𝑞𝑦 + 2𝑦 2
Charpit’s subsidiary equations are
𝑑𝑥 𝑑𝑦 𝑑𝑧 𝑑𝑝 𝑑𝑞
= = = =
−2𝑝 −𝑦 (−2𝑝2 + 𝑞𝑦) 2𝑝 4𝑦 + 3𝑞
Its symbolic operator equated to zero, 𝑖. ⅇ., 𝐷2 + 𝑘1 𝐷𝐷′ + 𝑘2 𝐷2 = 0 is called the auxiliary
equation (A.E)
Let its root be 𝐷/𝐷’ = 𝑚1 , 𝑚2 .
(𝐷 − 𝑚1 𝐷′ )(𝐷 − 𝑚2 𝐷′ )𝑧 = 0
(𝐷 − 𝑚2 𝐷′ )𝑧 = 0, 𝑖. ⅇ. , 𝑝 − 𝑚2 𝑞 = 0
(𝐷 − 𝑚1 𝐷′)2 𝑧 = 0
𝑢 = 𝜙(𝑦 + 𝑚1 𝑥)
which gives
𝑚 = −2, −1
1
Here the complete solution is 𝑧 = 𝑓1 (𝑦 − 2𝑥) + 𝑓2 (𝑦 − 2 𝑥)
which may be written as 𝑧 = 𝑓1 (𝑦 − 2𝑥) + 𝑓2 (2𝑦 − 𝑥).
Example Solve 4𝑟 + 12𝑠 + 9𝑡 = 0.
for
𝜕2𝑧 𝜕2𝑧 𝜕2𝑧
𝑟 = 𝜕𝑥 2 = 𝐷2 𝑧, 𝑠 = 𝜕𝑥𝜕𝑦 = 𝐷𝐷′𝑧 and 𝑡 = 𝜕𝑦 2 = 𝐷′2 𝑧.
Consider the equation (𝐷2 + 𝑘1 𝐷𝐷′ + 𝑘2 𝐷′2 )𝑧 = 𝐹(𝑥, 𝑦) 𝑖. ⅇ. , 𝑓(𝐷, 𝐷′ )𝑧 = 𝐹(𝑥, 𝑦).
1
∴ P.I. = 𝑓(𝐷,𝐷′ ) 𝐹(𝑥, 𝑦)
𝑎𝑥=𝑏𝑦
Since 𝐷ⅇ 𝑎𝑥+𝑏𝑦 = 𝑎ⅇ 𝑎𝑥+𝑏𝑦 ; 𝐷′𝑒 = 𝑏ⅇ 𝑎𝑥+𝑏𝑦
𝑎𝑥=𝑏𝑦
∴ 𝐷2 ⅇ 𝑎𝑥+𝑏𝑦 = 𝑎2 ⅇ 𝑎𝑥+𝑏𝑦 ; 𝐷𝐷′𝑒 = 𝑎𝑏ⅇ 𝑎𝑥+𝑏𝑦
1 1
P.I. = 𝑓(𝐷,𝐷′) ⅇ 𝑎𝑥+𝑏𝑦 = 𝑓(𝑎,𝑏) ⅇ 𝑎𝑥+𝑏𝑦
∴ 𝑓(𝐷2 , 𝐷𝐷′ , 𝐷′2 ) sin(𝑚𝑥 + 𝑛𝑦) = 𝑓(−𝑚2 , −𝑚𝑛, −𝑛2 )sin (𝑚𝑥 + 𝑛𝑦)
1 1
P.I. = sin(𝑚𝑥 + 𝑛𝑦) = 𝑓(−𝑚2 ,−𝑚𝑛,−𝑛2) sin (𝑚𝑥 + 𝑛𝑦)
𝑓(𝐷 2 ,𝐷𝐷 ′ ,𝐷′2 )
1
∴ P.I. = 𝑥 𝑚 𝑦 𝑛 = [𝑓(𝐷, 𝐷′)]−1 𝑥 𝑚 𝑦 𝑛 .
𝑓(𝐷,𝐷′)
1
∴ P.I. = 𝐹(𝑥, 𝑦)
𝑓(𝐷,𝐷 ′ )
To evaluate it, we resolve 1/𝑓(𝐷, 𝐷′) into partial fractions treating 𝑓(𝐷, 𝐷′ ) as a function
of D alone and operate each partial fraction on 𝐹(𝑥, 𝑦) remembering that
1
𝐹(𝑥, 𝑦) = ∫ 𝐹(𝑥, 𝑐 − 𝑚𝑥)𝑑𝑥
𝐷−𝑚𝐷 ′
𝜕𝑛 𝑧 𝜕𝑛 𝑧 𝜕𝑛 𝑧
+ 𝑘1 𝜕𝑥 𝑛−1 𝜕𝑦 + ⋯ + 𝑘𝑛 𝜕𝑦 𝑛 = 𝐹(𝑥, 𝑦).
𝜕𝑥 𝑛
or briefly
𝑓(𝐷, 𝐷 ′ )𝑧 = 𝐹(𝑥, 𝑦)
Step I. To find the C.F.
1
(i) When 𝐹(𝑥, 𝑦) = ⅇ 𝑎𝑥+𝑏𝑦 P.I. = 𝑓(𝐷,𝐷′) ⅇ 𝑎𝑥+𝑏𝑦 [𝑃𝑢𝑡 𝐷 = 𝑎 𝑎𝑛𝑑 𝐷’ = 𝑏]
1
P.I. = 𝑓(𝐷2,𝐷𝐷′ ,𝐷′2) sin 𝑜𝑟 cos (𝑚𝑥 + 𝑛𝑦)
[𝑃𝑢𝑡 𝐷2 = −𝑚2 , 𝐷𝐷′ = −𝑚𝑛, 𝐷′2 = −𝑛2]
1
(iii) When 𝐹(𝑥, 𝑦) = 𝑥 𝑚 𝑦 𝑛 , P.I. = 𝑓(𝐷,𝐷′) 𝑥 𝑚 𝑦 𝑛 = [𝑓(𝐷, 𝐷′)]−1 𝑥 𝑚 𝑦 𝑛 .
1
(iv) When 𝐹(𝑥, 𝑦) is any function of x and y P.I. = 𝐹(𝑥, 𝑦).
𝑓(𝐷,𝐷 ′ )
1
𝐹(𝑥, 𝑦) = ∫ 𝐹(𝑥, 𝑐 − 𝑚𝑐)𝑑𝑥
𝐷−𝑚𝐷 ′
∴ C.F. = 𝑓1 (𝑦 + 𝑥) + 𝑓2 (𝑦 − 5𝑥)
1
P.I. = 𝐷2+4𝐷𝐷′ −5𝐷′2 sin (2𝑥 + 3𝑦)
[𝑃𝑢𝑡 𝐷2 = −22 , 𝐷𝐷′ = −2 × 3, 𝐷′2 = −32 ]
1 1
= −4+4(−6)−5(−9) sin (2𝑥 + 3𝑦) = 17 sin (2𝑥 + 3𝑦)
1
Hence the C.S. is 𝑧 = 𝑓1 (𝑦 + 𝑥) + 𝑓2 (𝑦 − 5𝑥) + 17 sin(2𝑥 + 3𝑦).
𝜕2𝑧 𝜕2𝑧
Example Solve 𝜕𝑥 2 − 𝜕𝑥𝜕𝑦 = 𝑐𝑜𝑠 𝑥 𝑐𝑜𝑠 2𝑦.
Solution. Given equation in symbolic form is (𝐷2 − 𝐷𝐷′ )𝑧 = 𝑐𝑜𝑥 𝑐𝑜𝑠 2𝑦.
∴ C.F. = 𝑓1 (𝑦) + 𝑓2 (𝑦 + 𝑥)
1 1 1
P.I. = 𝐷2−𝐷𝐷′ cos 𝑥 cos 2𝑦 = 2 𝐷2−𝐷𝐷′ [cos(𝑥 + 2𝑦) + cos (𝑥 − 2𝑦)]
1 1
= 2 [𝐷2−𝐷𝐷′ cos(𝑥 + 2𝑦)
1
+ 𝐷2−𝐷𝐷′ cos(𝑥 − 2𝑦)] [Put 𝐷2 = −1, 𝐷𝐷′ = 2]
1 1 1 1 1
= 2 [−1+2 cos(𝑥 + 2𝑦) + −1−2 cos(𝑥 − 2𝑦)] = 2 cos(𝑥 + 2𝑦) − 6 cos (𝑥 − 2𝑦)
1 1
Hence the C.S. is 𝑧 = 𝑓1 (𝑦) + 𝑓2 (𝑦 + 𝑥) + cos(𝑥 + 2𝑦) − cos(𝑥 − 2𝑦).
2 6
𝜕3𝑧 𝜕3 𝑧
Example. Solve 𝜕𝑥 3 − 2 𝜕𝑥 2𝜕𝑦 = 2ⅇ 2𝑥 + 3𝑥 2 𝑦.
(𝐷2 − 2𝐷2 𝐷′ )𝑧 = 2ⅇ 2𝑥 + 3𝑥 2 𝑦
1 1 1
P.I. = 𝐷3−2𝐷2𝐷′(2ⅇ 2𝑥 + 3𝑥 2 𝑦) = 2𝐷3−2𝐷2𝐷′ ⅇ 2𝑥 + 3 𝐷3 (1−2𝐷′ /𝐷) 𝑥 2 𝑦
1 3 1 3 2𝐷′ 4𝐷 ′2
= 2 23 −2.22(0) ⅇ 2𝑥 + 𝐷3 (1 − 2𝐷′ /𝐷)−1 𝑥 2 𝑦 = 4 ⅇ 2𝑥 + 𝐷3 (1 + + + ⋯ ) 𝑥2𝑦
𝐷 𝐷2
1 3 2 1 3 2
= 4 ⅇ 2𝑥 + 𝐷3 (𝑥 2 𝑦 + 𝐷 𝑥 2 . 1) = 4 ⅇ 2𝑥 + 𝐷2 (𝑥 2 𝑦 + 3 𝑥 3 )
1
[∵ 𝑓(𝑥) = ∫ 𝑓(𝑥)𝑑𝑥]
𝐷
1 𝑥5 𝑥6
= 4 ⅇ 2𝑥 + 3𝑦 3 . + 2. 4 .
4. 5 5. 6
1
[∵ 𝑓(𝑥) = ∫[∫(∫ 𝑓(𝑥)𝑑𝑥)𝑑𝑥]𝑑𝑥]
𝐷3
𝑒 2𝑥 𝑥5𝑦 𝑥6
= + + 60
4 20
𝑒 2𝑥 𝑥5𝑦 𝑥6
Hence the C.S. is 𝑧 = 𝑓1 (𝑦) + 𝑥𝑓2 (𝑦) + 𝑓3 (𝑦 + 2𝑥) + + + 60
4 20
1
P.I. = (𝐷−2𝐷′)2 ⅇ 2𝑥+𝑦
[∵ 𝑦 = 𝑐 − 𝑚𝑥 = 𝑐 − 2𝑥]
1
Hence the C.S. is 𝑧 = 𝑓1 (𝑦 + 2𝑥) + 𝑥𝑓2 (𝑦 + 2𝑥) + 2 𝑥 2 ⅇ 2𝑥+𝑦 .
Solution. Given equation in symbolic form is (𝐷2 + 𝐷𝐷′ − 6𝐷′2 )𝑧 = 𝑐𝑜𝑠 (2𝑥 + 𝑦)
1 1
P.I. = 𝐷2−𝐷𝐷′ −6𝐷′2 cos(2𝑥 + 𝑦) = (𝐷+3𝐷′ )(𝐷−2𝐷′ ) cos (2𝑥 + 𝑦)
1 1
= (𝐷+3𝐷′ ) [∫ cos(2𝑥 + ̅̅̅̅̅̅̅̅̅
𝑐 − 2𝑥 ) 𝑑𝑥]𝑐→𝑦−3𝑥 = (𝐷+3𝐷′) [∫ cos 𝑐 𝑑𝑥]𝑐→𝑦−3𝑥
[∵ 𝑦 = 𝑐 − 𝑚𝑥 = 𝑐 − 2𝑥]
1
= (𝐷+3𝐷′ ) 𝑥𝑐𝑜𝑠(𝑦 + 2𝑥) = [∫ 𝑥 cos(𝑐̅̅̅̅̅̅̅̅̅
+ 3𝑥 + 2𝑥) 𝑑𝑥]𝑐→𝑦−3𝑥 = [∫ 𝑥 cos(5𝑥 + 𝑐) 𝑑𝑥]𝑐→𝑦−3𝑥
𝑥 1 𝑥 1
= 5 sin(5𝑥 + ̅̅̅̅̅̅̅̅̅
𝑦 − 3𝑥 ) + 25 cos(5𝑥 + ̅̅̅̅̅̅̅̅̅
𝑦 − 3𝑥) = 5 sin(2𝑥 + 𝑦) + 25 cos (2𝑥 + 𝑦)
𝑥 1
𝑧 = 𝑓1 (𝑦 − 3𝑥) + 𝑓2 (𝑦 + 2𝑥) + 5 sin(2𝑥 + 𝑦) + 25 cos (2𝑥 + 𝑦)
𝑥 1
𝑧 = 𝑓1 (𝑦 − 3𝑥) + 𝑓2 (𝑦 + 2𝑥) + 5 sin(2𝑥 + 𝑦) + 25 cos(2𝑥 + 𝑦).
1 1
P.I. = (𝐷−2𝐷′ )(𝐷+3𝐷′ ) 𝑦 cos 𝑥 = (𝐷−2𝐷′ ) [∫(𝑐 + 3𝑥) cos 𝑥 𝑑𝑥]𝑐→𝑦−3𝑥
[∵ 𝑦 = 𝑐 − 𝑚𝑥 = 𝑐 + 3𝑥]
1
= (𝐷−2𝐷′ ) [(𝑐 + 3𝑥) sin 𝑥 + 3 cos 𝑥]𝑐→𝑦−3𝑥
[Integrating by parts]
1
= (𝐷−2𝐷′ ) (𝑦 sin 𝑥 + 3 cos 𝑥) = [∫{(𝑐 + 3𝑥) sin 𝑥 + 3 cos 𝑥} 𝑑𝑥]𝑐→𝑦−2𝑥
= −𝑦 cos 𝑥 + sin 𝑥
1 1 1
P.I. = (2𝐷−𝐷′ )2 16 𝑙𝑜𝑔(𝑥 + 2𝑦) = 4 1 { 1 𝑙𝑜𝑔(𝑥 + 2𝑦)}
(𝐷− 𝐷 ′ ) (𝐷− 𝐷 ′ )
2 2
1 𝑥
=4 1 [∫ 𝑙𝑜𝑔 {𝑥 + 2 (𝑐 − 2)} 𝑑𝑥]
𝐷− 𝐷 ′ 𝑐→𝑦+𝑥⁄2
2
[∵ 𝑦 = 𝑐 − 𝑚𝑥 = 𝑐 + 3𝑥]
1 1
=4 1 [∫ log (2𝑐)𝑑𝑥]𝑐→𝑦+𝑥⁄2 = 4 1 [𝑥 log (𝑥 + 2𝑦)]
𝐷− 𝐷 ′ 𝐷− 𝐷 ′
2 2
𝑥
= 4 [∫ {𝑥𝑙𝑜𝑔 [𝑥 + 2 (𝑐 − 2)]} 𝑑𝑥] = 4[log 2𝑐 ∫ 𝑥 𝑑𝑥]𝑐→𝑦+𝑥⁄2 = 2𝑥 2 𝑙𝑜𝑔(𝑥 + 2𝑦)
𝑐→𝑦+𝑥⁄2
1 1
Hence the C.S. is 𝑧 = 𝑓1 (𝑦 + 2 𝑥) + 𝑓2 (𝑦 + 2 𝑥) + 2𝑥 2 𝑙𝑜𝑔(𝑥 + 2𝑦)
PROBLEMS 17.6
If in the equation
𝑓(𝐷, 𝐷′ )𝑧 = 𝐹(𝑥, 𝑦)
the polynomial expression 𝑓(𝐷, 𝐷′) is not homogeneous, then (1) is a non-homogeneous linear
partial differential equation. As in the case of homogeneous linear partial differential equations,
its complete solution = C.F. + P.I.
The methods to find P.I. are the same as those for homogeneous linear equations.
To find the C.F., we factorize 𝑓(𝐷, 𝐷′) into factors of the form 𝐷 − 𝑚𝐷′ − 𝑐. To find the
solution of (𝐷 − 𝑚𝐷′ − 𝑐)𝑧 = 0, we write it as
𝑝 − 𝑚𝑞 = 𝑐𝑧
as the solution of (2). The solution corresponding to various factors added up, give the C.F. of
(1).
𝑧 = ⅇ 𝑐𝑥 𝜙(𝑦 + 𝑚𝑥)
∴ C.F. = 𝜙1 (𝑦 − 𝑥) + ⅇ 2𝑥 𝜙2 (𝑦 − 𝑥)
1
∴ P.I. = 𝐷2+2𝐷𝐷′ +𝐷′2−2𝐷−2𝐷′ sin (𝑥 + 2𝑦)
1
= −1+2(−2)+(−4)−2𝐷−2𝐷′ sin (𝑥 + 2𝑦)
1 2(𝐷+𝐷 ′ )−9
= − 2(𝐷+𝐷′)+9 sin(𝑥 + 2𝑦) = − 4(𝐷2+2𝐷𝐷′ +𝐷′2)−81 sin (𝑥 + 2𝑦)
1
= 39 [2 cos(x + 2y) − 3 sin (𝑥 + 2𝑦)]
1
𝑧 = 𝜙1 (𝑦 − 𝑥) + ⅇ 2𝑥 𝜙2 (𝑦 − 𝑥) + 39 [2 cos(x + 2y) − 3 sin (𝑥 + 2𝑦)].
PROBLEMS
𝜕2𝑧 𝜕2𝑧 𝜕𝑧
1. 𝜕𝑥 2 + 𝜕𝑥𝜕𝑦 + 𝜕𝑦 2 − 𝑧 = ⅇ −𝑥 . 2. (𝐷 − 𝐷′ − 1)(𝐷 − 𝐷′ −
2)𝑧 = ⅇ 2𝑥−𝑦 .
𝜕2 𝑧 𝜕2𝑧 𝜕𝑧
3. (𝐷 + 𝐷′ − 1)(𝐷 + 2𝐷′ − 3)𝑧 = 4 + 3𝑥 + 6𝑦. 4. 𝜕𝑥 2 − 𝜕𝑥𝜕𝑦 + 𝜕𝑦 = 𝑥 2 + 𝑦 2 .
5. (𝐷2 + 𝐷𝐷′ + 𝐷′ − 1)𝑧 = sin(𝑥 + 2𝑦). 6. (2𝐷𝐷′ + 𝐷′2 + 3𝐷′)𝑧 =
3 cos(3𝑥 − 2𝑦).
𝑅𝑟 + 𝑆𝑠 + 𝑇𝑡 = 𝑉
Since
𝜕𝑝 𝜕𝑝
𝑑𝑝 = 𝜕𝑥 𝑑𝑥 + 𝜕𝑦 𝑑𝑦 = 𝑟𝑑𝑥 + 𝑡𝑑𝑦, and 𝑑𝑞 = 𝑠𝑑𝑥 + 𝑡𝑑𝑦,
we have
Substituting these values of 𝑟 and 𝑡 in (1), and rearranging the terms, we get
𝑥+𝑦
𝑥𝑑𝑝𝑑𝑦 + 𝑦𝑑𝑞𝑑𝑥 − 𝑥−𝑦 (𝑝 − 𝑞) 𝑑𝑦𝑑𝑥 = 0
Taking 𝑥𝑦 = 𝑐 and dividing each term of (ii) by 𝑥𝑑𝑦 or its equivalent −𝑦𝑑𝑥, we get
ⅆ𝑥 ⅆ𝑦 ⅆ𝑧
= −1 = (𝑥−𝑦)𝜙(𝑥𝑦)
1
𝑑𝑧 = −𝜙(𝑎𝑥 − 𝑥 2 ). (𝑎 − 2𝑥)𝑑𝑥
𝑧 = 𝜙1 (𝑥𝑦) + 𝜙2 (𝑥 + 𝑦).
Obs. Had we started with the integral 𝑥 + 𝑦 = 𝑐 and divided each term of (ii) by
𝑑𝑥 𝑜𝑟 − 𝑑𝑦, we would have arrived at the same solution.
whose integral is 𝑝𝑦 − 𝑞 + 3𝑦 2 = 𝑎
ⅆ𝑥 ⅆ𝑦 ⅆ𝑧
= −1 = 𝜙(𝑦 2+2𝑥)−3𝑦2
𝑦
PROBLEMS
Solve :
1. (𝑞 + 1)𝑠 = (𝑝 + 1)𝑡.
2. 𝑟 − 𝑡 𝑐𝑜𝑠 2 𝑥 + 𝑝 tan 𝑥 = 0.
3. 2𝑥 2 2𝑟 − 5𝑥𝑦𝑠 + 2𝑦 2 𝑡 + 2(𝑝𝑥 + 𝑞𝑦) = 0.
4. 𝑥𝑦(𝑡 − 𝑟) + (𝑥 2 − 𝑦 2 )(𝑠 − 2) = 𝑝𝑦 − 𝑞𝑥.
5. 𝑞 2 𝑟 − 2𝑝𝑞𝑠 + 𝑝2 𝑡 + 𝑝𝑞 2 .
6. (1 + 𝑞)2 𝑟 − 2(1 + 𝑝 + 𝑞 + 𝑝𝑞)𝑠 + (1 + 𝑝)2 𝑡 = 0.
OBJECTIVE TYPE OF QUESTIONS
PROBLEMS
Fill up the blanks or choose the correct answer in each of the following problems:
𝜕2𝑧 𝜕𝑧 2 𝜕𝑧
1. The equation 𝜕𝑥 2 + 2𝑥𝑦 (𝜕𝑥) + 𝜕𝑦 = 5 is of order ………. and degree ………..
2. The complementary function of (𝐷2 − 4𝐷𝐷′ + 4𝐷′2 )𝑧 = 𝑥 + 𝑦 is ……..
𝜕2 𝑧
3. The solution of 𝜕𝑦 2 = sin (𝑥𝑦) is ……..
4. A solution of (𝑦 − 𝑧)𝑝 + (𝑧 − 𝑥)𝑞 = 𝑥 − 𝑦 is ……….
5. The particular integral of (𝐷2 + 𝐷𝐷′)𝑧 = sin (𝑥 + 𝑦) is ………
6. The partial differential equation obtained from 𝑧 = 𝑎𝑥 + 𝑏𝑦 + 𝑎𝑏 by eliminating 𝑎
and 𝑏 is ………..
7. Solution of √𝑝 + √𝑞 = 1 is ………
8. Solution of 𝑝√𝑥 + 𝑞 √𝑦 = √𝑧 is ………
9. Solution of 𝑝 − 𝑞 = log(𝑥 + 𝑦).
10. The order of the partial differential equation obtained by eliminating 𝑓 from
𝑧 = 𝑓(𝑥 2 + 𝑦 2 ), is ……….
𝜕𝑧
11. The solution of 𝑥 𝜕𝑥 = 2𝑥 + 𝑦 is
12. By eliminating 𝑎 and 𝑏 from 𝑧 = 𝑎(𝑥 + 𝑦) + 𝑏, the p.d.e. formed is ……….
13. The solution of (𝐷3 − 3𝐷2 𝐷′ + 2𝐷𝐷′2 )𝑧 = 0 is ………
14. By eliminating the arbitrary constants from 𝑧 = 𝑎2 𝑥 + 𝑎𝑦 2 + 𝑏, the partial
differential equation formed is.
15. A solution of 𝑢𝑥𝑦 = 0 is of the form …………
𝜕2 𝑢 𝜕2𝑢
16. If 𝑢 = 𝑥 2 + 𝑡 2 is a solution of 𝑐 2 𝜕𝑥 2 = , then 𝑐 = …….
𝜕𝑡 2
17. The general solution of 𝑢𝑥𝑦 = 𝑥𝑦 is ………….
18. The complementary function of 𝑟 − 7𝑠 + 6𝑡 = ⅇ 𝑥+𝑦 is …………
19. The solution of 𝑥𝑝 + 𝑝𝑞 = 𝑧 is
(i) 𝑓(𝑥 2 , 𝑦 2 ) = 0 (ii) 𝑓(𝑥𝑦, 𝑦𝑧) =0
𝑥 𝑦
(iii) 𝑓(𝑥, 𝑦) = 0 (iv) 𝑓 (𝑦 , 𝑧 ) = 0.
20. The solution of (𝑦 − 𝑧)𝑝 + (𝑧 − 𝑥)𝑞 = 𝑥 − 𝑦, is
(i) 𝑓(𝑥 2 + 𝑦 2 + 𝑧 2 ) = 𝑥𝑦𝑧
(ii) 𝑓(𝑥 + 𝑦 + 𝑧) = 𝑥𝑦𝑧
(iii) 𝑓(𝑥 + 𝑦 + 𝑧) = 𝑥² + 𝑦² + 𝑧²
(iv) 𝑓(𝑥 2 + 𝑦 2 + 𝑧 2 , 𝑥𝑦𝑧) = 0.
21. The partial differential equation from 𝑧 = (𝑐 + 𝑥)² + 𝑦 is
𝜕𝑧 2 𝜕𝑧 2
(i) 𝑧 = (𝜕𝑥) + 𝑦 (ii) 𝑧 = (𝜕𝑦) + 𝑦
1 𝜕𝑧 2 1 𝜕𝑧 2
(iii) 𝑧 = 4 (𝜕𝑥) + 𝑦 (iv) 𝑧 = 4 (𝜕𝑦) + 𝑦
22. The solution of 𝑝 + 𝑞 = 𝑧 is
(i) 𝑓(𝑥𝑦, 𝑦 𝑙𝑜𝑔 𝑧) = 0
(ii) 𝑓(𝑥 + 𝑦, 𝑦 + 𝑙𝑜𝑔 𝑧) = 0
𝜕3𝑧
24. The solution of 𝜕𝑥 3 = 0 is
(i) 𝑧 = (1 + 𝑥 + 𝑥 2 )𝑓(𝑦)
(ii) 𝑧 = (1 + 𝑦 + 𝑦 2 )𝑓(𝑥)
(iii) 𝑧 = 𝑓1 (𝑥) + 𝑦𝑓2 (𝑥) + 𝑦 2 𝑓3 (𝑥)
(iv) 𝑧 = 𝑓1 (𝑦) + 𝑥𝑓2 (𝑦) + 𝑥 2 𝑓3 (𝑦).
At this point we are ready to now resume our work on solving the three
main equations: the heat equation, Laplace’s equation and the wave equa-
tion using the method of separation of variables.
XT 0 = X 00 T,
T0 X 00
= , (4.4)
T X
implying that
T 0 = λT, X 00 = λX, (4.5)
where λ is a constant. From (4.2) and (4.3), the boundary conditions be-
comes
X(0) = X(1) = 0. (4.6)
Integrating the X equation in (4.5) gives rise to three cases depending on
the sign of λ but as seen in the last chapter, only the case where λ = −k2
for some constant k is applicable which we have as the solution
which leads to
At this point, we recognize that we have a Fourier sine series and that the
coefficients bn are chosen such that
Z 1
bn = 2 (x − x2 ) sin nπx dx
0
· µ ¶ ¸¯1
1 − 2x x2 − x 2 ¯
= 2 cos nπx + − 3 3 cos nπx ¯¯
n2 π 2 nπ n π 0
4
= (1 − (−1)n ).
n π3
3
subject to
u(x, 0) = x − x2 , u x (0, t) = u x (1, t) = 0. (4.12)
Integrating the X equation in (4.14) gives rise to again three cases depend-
ing on the sign of λ but as seen earlier, only the case where λ = −k2 for
some constant k is relevant. Thus, we have
which leads to
2 π2 t
T(t) = c3 e−n ,
∞
2 2
u(x, t) = ∑ an e−n π t cos nπx,
n=0
∞
2
u(x, 0) = x − x = ∑ an cos nπx
n=0
∞
a0
= + ∑ an cos nπx.
2 n=1
We again recognize that we have a Fourier cosine series and that the coef-
ficients an are chosen such that
Z 1 ¸¯1
·
2 x2 x3 ¯¯ 1
a0 = 2 (x − x ) dx = 2 − ¯ = ,
0 2 3 0 3
Z 1
an = 2 (x − x2 ) cos nπx dx
0
· µ ¶ ¸¯1
1 − 2x x2 − x 2 ¯
= 2 sin nπx − − 3 3 sin nπx ¯¯
n2 π 2 nπ n π 0
2
= − 3 3 (1 + (−1)n ).
n π
∞
1 2 (−1)( n + 1) − 1 −n2 π2 t
u(x, t) = + 3
3 π ∑ n3
e cos nπx. (4.19)
n=1
0.3 y 0.3 y
t=0 t=0
0.2 0.2
t = .05
t = 0.1
0.1 0.1
t = .025
t = 0.2
0.0 0.0
0.0 0.5
x 1.0 0.0 0.5
x 1.0
Figure 1. The solution of the heat equation with the same initial condition with
fixed and no flux boundary conditions.
Example 2
Solve
ut = u xx , 0 < x < 2, t>0 (4.20)
subject to
(
x if 0 < x < 1,
u(x, 0) = u(0, t) = u x (2, t) = 0. (4.21)
2−x if 1 < x < 2,
gives rise to
T 0 = λT, X 00 = λX, (4.23)
Integrating the X equation in (4.23) with λ = −k2 for some constant k gives
which leads to
π 3π 5π (2n − 1)π
c2 = 0, cos 2k = 0 ⇒ k = , , ,..., ,
4 4 4 4
for integer n. From (4.23), we then deduce that
(2n−1)2 2
T(t) = c3 e− 16 π t
,
Figure 2 shows the solution both in short time t = 0, 0.1, 0.2 and long time
t = 10, 20, 30.
2 y 0.8 y
t=0 t = 10
t=2 t = 20
1 0.4
t = 30
t=1
x
0 0.0
x
0 1 2 0 1 2
Figure 2. Short time and long time behavior of the solution (4.26).
Example 3
Solve
ut = u xx , 0 < x < 2, t>0 (4.27)
subject to
In this problem, we have a fixed left endpoint and a radiating right end
point. Assuming separable solutions
gives rise to
T 0 = λT, X 00 = λX, (4.30)
Integrating the X equation in (4.30) with λ = −k2 for some constant k gives
c1 sin 0 + c2 cos 0 = 0, ⇒ c2 = 0
and the second boundary condition of (4.31) gives
It is very important that we recognize that the solutions of (4.33) are not
equally spaced as seen in earlier problems. In fact, there are an infinite
number of solutions of this equation. Figure 3 shows graphically the curves
y = −k and y = tan 2k. The first three intersection points are the first three
solutions of (4.33).
0 1 2 3 4
k
−2
k k3
1
−4
k 2
Therefore, we have
X(x) = c1 sin k n x. (4.34)
2
T(t) = c3 e−kn t (4.35)
and together with X, we have the solution to the PDE as
∞
2
u(x, t) = ∑ cn e−kn t sin kn x, (4.36)
n=1
Imposing the boundary conditions (4.28) gives
∞
u(0, t) = 2x − x2 = ∑ cn sin kn x, (4.37)
n=1
It is important to know that the cn ’s are not given by the formula
Z
2 2
cn = (2x − x2 )sink n x, dx
2 0
as usual. The reason for this is that the k n ’s are not equally spaced. So
it is necessary to examine (4.37) on its own. Multiplying by sin k m x and
integrating over [0,2] gives
Z 2 ∞ Z 2
2
0
(2x − x ) sin k m x dx = ∑ cn
0
sin k m x sin k n x dx,
n=1
For n 6= m, we have
Z 2
k m sin 2k n cos 2k m − k n sin 2k m cos 2k n
sin k m x sin k n x dx = (4.38)
0 k2n − k2m
and imposing (4.33) for each of k m and k n shows (4.38) to be identically
satisfied. Therefore, we obtain the following when n = m
Z 2 Z 2
2
(2x − x ) sin k n x dx = cn sin2 k n x dx,
0 0
or R2
0 (2x − x2 ) sin k n x dx
cn = R2 (4.39)
2
0 sin k n x dx,
y
1.0 t = 0
0.5
t = 1
t = 2
0.0
x
0 1 2
Example 4
Solve
ut = u xx , 0 < x < 3, t>0 (4.40)
subject to
u(x, 0) = 4x − x2 u(0, t) = 0, u(3, t) = 3. (4.41)
u = v + x. (4.44)
We notice that under the transformation (4.44), the original equation doesn’t
change form, i.e.
ut = u xx ⇒ vt = v xx
v(x, 0) = 3x − x2 .
subject to
v(x, 0) = 3x − x2 v(0, t) = 0, v(3, t) = 3. (4.46)
nπ
X(x) = c1 sin x, (4.48)
3
and
n2 π 2
T(t) = c3 e− 9 t
where n is an integer. Therefore, we have the solution of (4.45) as
∞ n2 π 2 nπ
v(x, t) = ∑ bn e − 9 t sin
3
x.
n=1
Z
2 3 nπ
bn = (3x − x2 ) sin x dx
3 0 3
· ¸¯3
6(2x − 3) nπ 3(n2 π 2 x2 − 3n2 π 2 x − 18) nπ ¯¯
= − sin x+ cos x ¯
n2 π 2 3 n3 π 3 3 0
32(1 − (−1) ) n
= .
n3 π 3
This gives
∞
32 (1 − (−1)n ) − n2 π2 t nπ
v(x, t) =
π3 ∑ n 3
e 9 sin
3
x.
n=1
and since u = v + x, we obtain the solution for u as
∞
32 (1 − (−1)n ) − n2 π2 t nπ
u(x, t) = x +
π3 ∑ n 3
e 9 sin
3
x. (4.49)
n=1
4
y
t = 0
t = 1
2 t = 2
x
0
0 1 2 3
Example 5
Solve
ut = u xx , 0 < x < 1, t>0 (4.50)
subject to
u(x, 0) = 0 u x (0, t) = −1, u x (1, t) = 0. (4.51)
Unfortunately, the trick u = v + ax + b won’t work since u x = v x + a and
choosing a to fix the right boundary to zero only makes the left boundary
nonzero. To overcome this we might try u = v + ax2 + bx but the original
equation changes
ut = u xx , ⇒ vt = v xx + 2a (4.52)
noting now
ut = u xx , ⇒ vt = v xx . (4.54)
subject to
1
v(x, 0) = x − x2 , v x (0, t) = v x (1, t) = 0 (4.57)
2
A separation of variables v = XT leads to X 00 = −k2 X and T 0 = −k2 T
from which we obtain
c1 = 0, k = nπ, (4.59)
Finally we arrive at
∞
a0 2 2
v(x, t) = + ∑ an e−n π t cos nπx.
2 n=1
Figure 6 shows plots at time t = 0.01, 0.5, 1.0 and 1.5. It is interesting to
note that at the left boundary u x = −1 and since the flux φ = −ku x implies
that φ = k > 0 which gives that the flux is position and that heat is being
added at the left boundary. Hence the profile increase at the left while
insulated at the right boundary (no flux).
2 y
t = 1.5
1
t = 1.0
t = 0.5
t = 0.1
0
x
0.0 0.5 1.0
where
16(1 − (−1)n ) − n2 π2 t
Tn (t) = e 4 (4.71)
n3 π 3
We note that even if this Tn wasn’t known, we could find it as substitution
of (4.70) into the heat equation and isolating coefficients of sin nπx
2 , would
lead to
n2 π 2
Tn0 (t) = − Tn (t),
4
leading to the solution (4.71). For the problem with a source term we look
for solutions of the same form i.e. (4.70). However, in order that this tech-
nique works, it is necessary to expand the source term also in terms of a
Fourier sine series, i.e.
∞
nπx
Q(x) = ∑ qn sin 2
. (4.72)
n=1
For (
x if 0 < x < 1,
Q(x) = (4.73)
2−x if 1 < x < 2,
where
Z 2
nπx
qn = Q(x) sin dx.
0 2
Z 1 Z 2
nπx nπx
= x sin dx + (2 − x) sin dx,
0 2 x 2
· ¸¯
4 nπx nπx ¯¯1
= sin − 2xnπ cos
n2 π 2 2 2 ¯0
· ¸¯
4 nπx 2x4 nπx ¯¯2
+ − 2 2 sin + cos ,
n π 2 nπ 2 ¯1
8 nπx
= sin . (4.74)
n2 π 2 2
Substituting both (4.70) and (4.72) into (4.69) gives
∞
nπx ∞ ³ nπx ´2 nπx ∞
nπx
∑ Tn0 (t) sin = ∑− Tn (t) sin + ∑ qn sin ,
n=1
2 n=1
2 2 n=1
2
n2 π 2
Tn0 (t) + Tn (t) = qn , (4.75)
4
a linear ODE in Tn (t)! On solving (4.75) we obtain
4 nπ 2
Tn (t) = qn + bn e−( 2 ) t ,
n2 π 2
giving the final solution
∞ µ ¶
4 −( nπ )2 t nπx
u(x, t) = ∑ 2
n π 2
q n + bn e 2 sin
2
. (4.76)
n=1
If we set
4
cn = qn + bn ,
n2 π 2
then we have
∞
2 nπx
2x − x = ∑ cn sin 2
.
n=1
a regular Fourier sine series. Therefore
Z 2³ ´ nπx
cn = 2x − x2 sin dx.
0 2
16 ³ nπ ´
= 1 − cos , (4.77)
n3 π 3 2
which in turn, gives
4
bn = c n − qn ,
n2 π 2
and finally, the solution as
∞ µ µ ¶ ¶
4 4 −( nπ ) 2t nπx
u(x, t) = ∑ 2 2
qn + cn − 2 2 qn e 2 sin , (4.78)
n=1
n π n π 2
where qn and cn are given in (4.74) and (4.77), respectively. Typical plots
are given in figure 7 at times t = 0, 1, 2 and 3.
1.0 y
t = 0
t = 1
0.5
t = 2
t = 3
0.0
x
0 1 2
and ask “Is is possible to find A such that the source term in (4.83) can be
removed?” Substituting of (4.84) in (4.83) gives
Avt + At v = Av xx + 2A x v x + A xx v + αAv,
and dividing by A and expanding and regrouping gives
µ ¶
Ax A xx At
vt = v xx + 2 v x + − + α v.
A A A
In order to target the standard heat equation, we choose
A xx At
A x = 0, − + α = 0. (4.85)
A A
From the first we obtain that A = A(t) and from the second we obtain
A0 = αA which has the solution A(t) = A0 eαt for some constant A0 . The
boundary conditions becomes
t = 0.2
0.3 0.4 y
t= 0
t = 0.1
0.2 t= 0
t = 0.1
0.2
t = 0.2
0.1
0.0 0.0
x x
0.0 0.5 1.0 0.0 0.5 1.0
Figure 8. The solution (4.88) of the heat equation with a source (4.83) with α = 5
and α = 12.
Example 6
Solve
and so the insulted boundary conditions also remain insulated! Thus, the
problem (4.90) becomes
where
Z 2 · ¸¯2
2 3 x4 ¯¯
2
a0 = (4x − x ) dx = 2x − = 4,
2 0 4 ¯0
Z
2 2 nπ
an = (4x − x3 ) cos x dx (4.95)
2 0 2
·µ ¶ µ ¶ ¸¯2
2(4x − x3 ) 48 nπ 4(4 − 3x2 ) 96 nπ ¯¯
= + 3 3 sin x+ + 4 4 cos x ¯
nπ n π 2 n2 π 2 n π 2 0
µ ¶ µ ¶
16 96 6 48 nπ 4 96 nπ
= − 2 2− 4 4+ + 3 3 sin + 2 2
+ 4 4 cos .
n π n π nπ n π 2 n π n π 2
where an is given in (4.95). Figure 9 show plots at t = 0, 0.2, 0.4 and 0.6
when α = −2 and 2. It is interesting to note that the sign of α will deter-
mine whether the solution will grow or decay exponentially.
y y
3 t= 0 8
t = 0.6
6
2
t = 0.4
4
t = 0.2
t = 0.2
1 t = 0.4
2
t = 0.6
t= 0
0 x 0
x
0 1 2
0 1 2
Figure 9. The solution (4.96) of the heat equation with a source (4.90) with no
flux boundary condition with α = −2 and α = 2.
and ask “Is is possible to find A such that the convection term can be re-
moved?” Substituting of (4.98) in (4.97) gives
Avt + At v = Av xx + 2A x v x + A xx v + β (Av x + A x v) ,
and dividing by A and expanding and regrouping gives
2A x + βA A xx − At + βA x
vt = v xx + vx + v. (4.99)
A A
In order to target the standard heat equation, we choose
2A x + βA = 0, A xx − At + βA x = 0. (4.100)
1
From the first we obtain that A(x, t) = C(t)e− 2 βx and from the second
β2 1 2
we obtain C 0 + 4 C = 0 which has the solution C(t) = A0 e− 4 β t for some
constant A0 . This then gives
1 1 2
A(x, t) = A0 e− 2 βx− 4 β t
(4.101)
1 1 2
∞
2 2
u(x, t) = e− 2 βx− 4 β t
∑ bn e−n π t sin nπx, (4.109)
n=1
For β = −12
Z 1
bn = 2 (x − x2 )e−6x sin nπx dx
0
7n2 π 2 + 108 + (5n2 π 2 + 324)e−6 cos nπ
= 2nπ . (4.111)
(36 + n2 π 2 )3
1 1 2
u(x, t) = 4πe− 2 βx− 4 β t (4.112)
∞ 2 2 2 2 3
27 + n π + 2n π e cos nπ −n2 π2 t
× ∑n e sin nπx,
n=1
(9 + n2 π 2 )3
1 1 2
u(x, t) = 2πe− 2 βx− 4 β t (4.113)
∞ 2 2 2 2 −6
7n π + 108 + (5n π + 324)e cos nπ −n2 π2 t
× ∑n 2 π 2 )3
e sin nπx.
n=1
(36 + n
t = 0.025
0.2 0.2
t = 0.05
t = 0.05
0.1 t = 0.1 0.1
0.0 x 0.0 x
0.0 0.5 1.0 0.0 0.5 1.0
Figure 10. The solution (4.112) of the heat equation with convection with fixed
boundary conditions with β = 6 and β = −12.
Example 7
As a final example, we consider
subject to
u(x, 0) = 4x − x3 u x (0, t) = 0, u x (2, t) = 0. (4.115)
1 1 2 β − 1 βx− 1 β2 t
u x = e− 2 βx− 4 β t v x − e 2 4 v, (4.116)
2
and so
β
u x (0, t) = 0 ⇒ v x (0, t) − v(0, t) = 0, (4.117)
2
β
u x (2, t) = 0 ⇒ v x (2, t) − v(2, t) = 0. (4.118)
2
Thus, the insulated boundary condition become radiating boundary con-
ditions. As for the initial condition
1
u(x, 0) = 4x − x3 ⇒ v(x, 0) = (4x − x3 )e 2 βx , (4.119)
u xx + uyy = 0 (4.120)
We will show that a separation of variables also works for this equation
As an example consider the boundary conditions
X 00 Y + XY 00 = 0. (4.123)
4(1 − (−1)n )
an = . (4.135)
n3 π 3 sinh nπ
Thus, the solution to Laplace’s equation with the boundary conditions
given in (4.121) is
∞
4 1 − (−1)n sinh nπy
u(x, y) = 3
π ∑ n 3
sin nπx
sinh nπ
. (4.136)
n=1
y u
1.0 0.2
u = 0.2
u=
0.1
0.5
u = 0.05
u = 0.025 0.0
u = 0.001 0.5
1.0
0.0 1.0
x
0.0 0.5 1.0
y
x
Figure 11. The solution (4.120) with the boundary conditions (4.121)
subject to
is nπy
∞
nπx sinh Ly
u = ∑ bn sin , (4.139)
n=1
L x sinh nπ
Ly
30
where Z Lx
2 nπx
bn = f (x) sin dx. (4.140)
Lx 0 Lx
Example 8
Solve
u xx + uyy = 0, (4.141)
subject to
X 00 Y + XY 00 = 0. (4.144)
X 00 Y 00
+ = 0, (4.145)
X Y
X 00 Y 00
= λ, = −λ (4.146)
X Y
16
An = (1 − cos nπ), (4.155)
n3 π 3
and further that the solution to Laplace’s equation with the boundary con-
ditions given in (4.141) subject to (4.186) is
∞
4 1 − (−1)n sinh nπx
u= 3
π ∑ n3 sinh nπ
sin nπy. (4.156)
n=1
Figure 12 show both a top view and a 3 − D view of the solution. In com-
paring the solutions (4.136) and (4.156) shows that if we interchange x and
y they are the same. This should not be surprising because if we con-
sider Laplace equations with the boundary conditions given in (4.121) and
(4.186), that if we interchange x and y, the problems are transformed to
each other.
In general, using separation of variables, the solution of
subject to
u(x, 0) = 0, u(x, 1) = 0
u(0, y) = 0, u(1, y) = g(y).
is
∞ sinh nπx
Lx nπy
u= ∑ bn sinh nπ sin
Ly
(4.159)
n=1 Lx
where Z Ly
2 nπy
bn = g(y) sin dy (4.160)
Ly 0 Ly
Example 9
Solve
u xx + uyy = 0 (4.161)
subject to
leads to
X 00 Y + XY 00 = 0, (4.164)
noting that the last boundary condition is different than the boundary con-
dition considered at the beginning of this section (i.e. Y(0) = 0). The re-
maining boundary condition in (4.162a) will be used later. In order to solve
the X equation in (4.166) subject to the boundary conditions (4.167), it is
necessary to set λ = −k2 . The X equation (4.166) as the general solution
subject to
Example 10
As the final example, we consider
u xx + uyy = 0 (4.182)
subject to
subject to
u(x, 0) = 0, u(x, 1) = 0
u(0, y) = g(y), u(1, y) = 0.
is
∞ sinh nπ(1−x)
Ly nπy
u= ∑ bn sinh nπ sin
Ly
. (4.187)
n=1 Ly
where Z Ly
2 nπy
bn = g(y) sin dy (4.188)
Ly 0 Ly
Second-Order Partial Differential Equations
The most general case of second-order linear partial differential equation (PDE) in two inde-
pendent variables is given by
∂ 2u ∂ 2u ∂ 2u ∂u ∂u
A + B + C + D + E + Fu = G (1)
∂ x2 ∂ x∂ y ∂ y2 ∂x ∂y
where the coefficients A, B, and C are functions of x and y and do not vanish simultaneously,
because in that case the second-order PDE degenerates to one of first order. Further, the
coefficients D, E, and F are also assumed to be functions of x and y. We shall assume that the
function u(x, y) and the coefficients are twice continuously differentiable in some domain Ω.
The classification of second-order PDE depends on the form of the leading part of the
equation consisting of the second order terms. So, for simplicity of notation, we combine the
lower order terms and rewrite the above equation in the following form
∂ 2u ∂ 2u ∂ 2u ∂u ∂u
A(x, y) 2 + B(x, y) + C(x, y) 2 = Φ x, y, u, , (2a)
∂x ∂ x∂ y ∂y ∂x ∂y
As we shall see, there are fundamentally three types of PDEs – hyperbolic, parabolic, and
elliptic PDEs. From the physical point of view, these PDEs respectively represents the wave
propagation, the time-dependent diffusion processes, and the steady state or equilibrium pro-
cesses. Thus, hyperbolic equations model the transport of some physical quantity, such as
fluids or waves. Parabolic problems describe evolutionary phenomena that lead to a steady
state described by an elliptic equation. And elliptic equations are associated to a special state
of a system, in principle corresponding to the minimum of the energy.
Mathematically, these classification of second-order PDEs is based upon the possibility of
reducing equation (2) by coordinate transformation to canonical or standard form at a point. It
may be noted that, for the purposes of classification, it is not necessary to restrict consideration
to linear equations. It is applicable to quasilinear second-order PDE as well. A quasilinear
second-order PDE is linear in the second derivatives only.
The type of second-order PDE (2) at a point (x0 , y0 ) depends on the sign of the discriminant
defined as
B 2A
∆(x0 , y0 ) ≡ = B(x0 , y0 )2 − 4 A(x0 , y0 )C(x0 , y0 ) (3)
2C B
The classification of second-order linear PDEs is given by the following: If ∆(x0 , y0 ) > 0, the
equation is hyperbolic, ∆(x0 , y0 ) = 0 the equation is parabolic, and ∆(x0 , y0 ) < 0 the equation
is elliptic. It should be remarked here that a given PDE may be of one type at a specific point,
and of another type at some other point. For example, the Tricomi equation
∂ 2u ∂ 2u
+ x =0
∂ x2 ∂ y2
is hyperbolic in the left half-plane x < 0, parabolic for x = 0, and elliptic in the right half-plane
x > 0, since ∆ = −4x. A PDE is hyperbolic (or parabolic or elliptic) in a region Ω if the PDE
is hyperbolic (or parabolic or elliptic) at each point of Ω.
The terminology hyperbolic, parabolic, and elliptic chosen to classify PDEs reflects the anal-
ogy between the form of the discriminant, B2 −4AC, for PDEs and the form of the discriminant,
B2 − 4AC, which classifies conic sections given by
The type of the curve represented by the above conic section depends on the sign of the
discriminant, ∆ ≡ B2 − 4AC. If ∆ > 0, the curve is a hyperbola, ∆ = 0 the curve is an parabola,
and ∆ < 0 the equation is a ellipse. The analogy of the classification of PDEs is obvious. There
is no other significance to the terminology and thus the terms hyperbolic, parabolic, and elliptic
are simply three convenient names to classify PDEs.
In order to illustrate the significance of the discriminant ∆ and thus the classification of the
PDE (2), we try to reduce the given equation (2) to a canonical form. To do this, we transform
the independent variables x and y to the new independent variables ξ and η through the change
of variables
ξ = ξ (x, y), η = η (x, y) (4)
where both ξ and η are twice continuously differentiable and that the Jacobian
∂ (ξ , η ) ξx ξy
J= = 6= 0 (5)
∂ (x, y) ηx ηy
in the region under consideration. The nonvanishing of the Jacobian of the transformation
ensure that a one-to-one transformation exists between the new and old variables. This simply
means that the new independent variables can serve as new coordinate variables without any
ambiguity. Now, define w(ξ , η ) = u(x(ξ , η ), y(ξ , η )). Then u(x, y) = w(ξ (x, y), η (x, y)) and,
apply the chain rule to compute the terms of the equation (2) in terms of ξ and η as follows:
ux = wξ ξx + wη ηx
uy = wξ ξy + wη ηy
uxx = wξ ξ ξx2 + 2wξ η ξx ηx + wηη ηx2 + wξ ξxx + wη ηxx (6)
uyy = wξ ξ ξy2 + 2wξ η ξy ηy + wηη ηy2 + wξ ξyy + wη ηyy
uxy = wξ ξ ξx ξy + wξ η (ξx ηy + ξy ηx ) + wηη ηx ηy + wξ ξxy + wη ηxy
Substituting these expressions into equation (2) we obtain the transformed PDE as
where Φ becomes φ and the new coefficients of the higher order terms a, b, and c are expressed
via the original coefficients and the change of variables formulas as follows:
At this stage the form of the PDE (7) is no simpler than that of the original PDE (2), but this
is to be expected because so far the choice of the new variable ξ and η has been arbitrary.
However, before showing how to choose the new coordinate variables, observe that equation
(8) can be written in matrix form as
T
ξx ξy ξx ξy
a b/2 A B/2
=
b/2 c ηx ηy B/2 C ηx ηy
Recalling that the determinant of the product of matrices is equal to the product of the de-
terminants of matrices and that the determinant of a transpose of a matrix is equal to the
determinant of that matrix, we get
a b/2 A B/2 2
=
b/2 c B/2 C J
where J is the Jacobian of the change of variables given by (5). Expanding the determinant
and multiplying by the factor, −4, to obtain
where δ = b2 − 4ac is the discriminant of the equation (7). This shows that the discriminant
of the transformed equation (7) has the same sign as the discriminant of the original equation
(2) and therefore it is clear that any real nonsingular (J 6= 0) transformation does not change
the type of PDE. Note that the discriminant involves only the coefficients of second-order
derivatives of the corresponding PDE.
Canonical forms
Let us now try to construct transformations, which will make one, or possibly two of the
coefficients of the leading second order terms of equation (7) vanish, thus reducing the equation
to a simpler form called canonical from. For convenience, we reproduce below the original PDE
We again mention here that for the PDE (2) (or (7)) to remain a second-order PDE, the
coefficients A, B, and C (or a, b, and c) do not vanish simultaneously.
By definition, a PDE is hyperbolic if the discriminant ∆ = B2 − 4AC > 0. Since the sign of
discriminant is invariant under the change of coordinates (see equation (9)), it follows that for
a hyperbolic PDE, we should have b2 − 4ac > 0. The simplest case of satisfying this condition
is a = c = 0. So, if we try to chose the new variables ξ and η such that the coefficients a and
c vanish, we get the following canonical form of hyperbolic equation:
wξ η = ψ ξ , η , w, wξ , wη
(10a)
where ψ = φ /b. This form is called the first canonical form of the hyperbolic equation. We
also have another simple case for which b2 − 4ac > 0 condition is satisfied. This is the case
when b = 0 and c = −a. In this case (9) reduces to
wαα − wβ β = ψ α , β , w, wα , wβ
(10b)
which is the second canonical form of the hyperbolic equation, where ψ = φ /a.
By definition, a PDE is parabolic if the discriminant ∆ = B2 − 4AC = 0. It follows that for
a parabolic PDE, we should have b2 − 4ac = 0. The simplest case of satisfying this condition is
c (or a) = 0. In this case another necessary requirement b = 0 will follow automatically (since
b2 − 4ac = 0). So, if we try to chose the new variables ξ and η such that the coefficients b
and c vanish, we get the following canonical form of parabolic equation:
wξ ξ = ψ ξ , η , w, wξ , wη
(11)
where ψ = φ /a.
By definition, a PDE is elliptic if the discriminant ∆ = B2 − 4AC < 0. It follows that for
a elliptic PDE, we should have b2 − 4ac < 0. The simplest case of satisfying this condition is
b = 0 and c = a. So, if we try to chose the new variables ξ and η such that b vanishes and
c = a, we get the following canonical form of elliptic equation:
wξ ξ + wηη = ψ ξ , η , w, wξ , wη
(12)
where ψ = φ /a.
In summary, equation (7) can be reduced to a canonical form if the coordinate transformation
ξ = ξ (x, y) and η = η (x, y) can be selected such that:
• a = c = 0 corresponds to the first canonical form of hyperbolic PDE given by
wξ η = ψ ξ , η , w, wξ , wη
(10a)
wαα − wβ β = ψ α , β , w, wα , wβ
(10b)
wξ ξ = ψ ξ , η , w, wξ , wη
(11)
wξ ξ + wηη = ψ ξ , η , w, wξ , wη
(12)
We will now examine the kind of transformation required to reduce the PDE to its canonical
form.
Hyperbolic equations
For a hyperbolic PDE the discriminant ∆(= B2 − 4AC) > 0. In this case, we have seen that, to
reduce this PDE to canonical form we need to choose the new variables ξ and η such that the
coefficients a and c vanish in (7). Thus, from (8), we have
Dividing equation (13a) and (13b) throughout by ξy2 and ηy2 respectively to obtain
ξx 2 ξx
A +B +C = 0 (14a)
ξy ξy
ηx ηx
2
A +B +C = 0 (14b)
ηy ηy
Equation (14a) is a quadratic equation for (ξx /ξy ) whose roots are given by
√
−B + B2 − 4AC
µ1 (x, y) =
√2A
−B − B2 − 4AC
µ2 (x, y) =
2A
The roots of the equation (14b) can also be found in an identical manner, so as only two distinct
roots are possible between the two equations (14a) and (14b). Hence, we may consider µ1 as
the root of (14a) and µ2 as that of (14b). That is,
√
ξx −B + B2 − 4AC
µ1 (x, y) = = (15a)
ξy 2A
√
ηx −B − B2 − 4AC
µ2 (x, y) = = (15b)
ηy 2A
The above equations lead to the following two first-order differential equations
These are the equations that define the new coordinate variables ξ and η that are necessary
to make a = c = 0 in (7).
Along the coordinate line ξ (x, y) = constant, we have the total derivative of ξ , d ξ = 0. It
follows that
d ξ = ξx dx + ξy dy = 0
and hence, the slope of such curves is given by
dy ξx
=−
dx ξy
We also have a similar result along coordinate line η (x, y) = constant, i.e.,
dy ηx
=−
dx ηy
This is called the characteristic polynomial of the PDE (2) and its roots are given by
√
dy B + B2 − 4AC
= = λ1 (x, y) (18a)
dx √ 2A
dy B − B2 − 4AC
= = λ2 (x, y) (18b)
dx 2A
The required variables ξ and η are determined by the respective solutions of the two ordinary
differential equations (18a) and (18b), known as the characteristic equations of the PDE (2).
They are ordinary differential equations for families of curves in the xy-plane along which ξ =
constant and η = constant. Clearly, these families of curves depend on the coefficients A, B,
and C in the original PDE (2).
Integration of equation (18a) leads to the family of curvilinear coordinates ξ (x, y) = c1 while
the integration of (18b) gives another family of curvilinear coordinates η (x, y) = c2 , where c1
and c2 are arbitrary constants of integration. These two families of curvilinear coordinates
ξ (x, y) = c1 and η (x, y) = c2 are called characteristic curves of the hyperbolic equation (2) or,
more simply, the characteristics of the equation. Hence, second-order hyperbolic equations have
two families of characteristic curves. The fact that ∆ > 0 means that the characteristic are real
curves in xy-plane.
If the coefficients A, B, and C are constants, it is easy to integrate equations (18a) and
(18b) to obtain the expressions for change of variables formulas for reducing a hyperbolic PDE
to the canonical form. Thus, integration of (18) produces
√ √
B + B2 − 4AC B − B2 − 4AC
y= x + c1 and y= x + c2 (19a)
2A 2A
or √ √
B + B2 − 4AC B − B2 − 4AC
y− x = c1 and y− x = c2 (19b)
2A 2A
Thus, when the coefficients A, B, and C are constants, the two families of characteristic curves
associated with PDE reduces to two distinct families of parallel straight lines. Since the families
of curves ξ = constant and η = constant are the characteristic curves, the change of variables
are given by the following equations:
√
B + B2 − 4AC
ξ = y− x = y − λ1 x (20)
√ 2A
B − B2 − 4AC
η = y− x = y − λ2 x (21)
2A
The first canonical form of the hyperbolic is:
wξ η = ψ ξ , η , w, wξ , wη
(22)
α = ξ + η = (y − λ1 x) + (y − λ2 x)
= 2y − (λ1 + λ2 )x
β = ξ − η = (y − λ1 x) − (y − λ2 x)
= (λ2 + λ1 )x
Example 1
Show that the one-dimensional wave equation
∂ 2u 2∂ u
2
− c =0
∂ t2 ∂ x2
is hyperbolic, find an equivalent canonical form, and then obtain the general solution.
Solution To interpret the results for (2) that involve the independent variables x and y in
terms of the wave equation utt − c2 uxx = 0, where the independent variables are t and x, it will
be necessary to replace x and y in (2) and (6) by t and x. It follows that the wave equation is
a constant coefficient equation with
A = 1, B = 0, C = −c2
We calculate the discriminant, ∆ = 4c2 > 0, and therefore the PDE is hyperbolic. The roots of
the characteristic polynomial are given by
√ √
B+ ∆ B− ∆
λ1 = =c and λ2 = = −c
2A 2A
Therefore, from the characteristic equations (18a) and (18b), we have
dx dx
= c, = −c
dt dt
Integrating the above two ODEs to obtain the characteristics of the wave equation
x = ct + k1 , x = −ct + k2
where k1 and k2 are the constants of integration. We see that the two families of characteristics
for the wave equation are given by x − ct = constant and x + ct = constant. It follows, then,
that the transformation
ξ = x − ct, η = x + ct
reduces the wave equation to canonical form. We have,
∆
a = 0, c = 0, b=− = −4c2
A
So in terms of characteristic variables, the wave equation reduces to the following canonical
form
wξ η = 0
For the wave equation the characteristics are found to be straight lines with negative and
positive slopes as shown in Fig. 1. The characteristics form a natural set of coordinates for the
hyperbolic equation.
t
x + ct = a x − ct = a
b
x
a
The canonical forms are simple because they can be solved directly by integrating twice.
For example, integrating with respect to ξ gives
Z
wη = 0 d ξ = h(η )
wαα − wβ β = 0
Example 2
In steady or unsteady transonic flow around wings and airfoils with thickness to chord ratios
of a few percent, we can generally consider that the flow is predominantly directed along
the chordwise direction, taken as the x-direction. In this case, the velocities in the transverse
direction can be neglected and the potential equation reduces to the so-called small disturbance
potential equation:
∂ 2φ ∂ 2φ
1 − M∞2 + =0
∂ x2 ∂ y2
Historically, this was the form of equation used by Murman and Cole (1961) to obtain the first
numerical solution for a transonic flow around an airfoil with shocks.
Show that, depending on the Mach number, the small disturbance potential equation is
elliptic, parabolic, or hyperbolic. Find the characteristic variables for the hyperbolic case and
hence write the equation in canonical form.
Solution The given equation is of the form (2) where
A = 1 − M∞2 , B = 0, C=1
The discriminant, ∆ = B2 − 4AC = −4(1 − M∞2 ). Therefore, the PDE is hyperbolic for M > 1,
elliptic for M < 1, and parabolic for M = 1 (along the sonic line). For the case of supersonic
flow (M > 1), the roots of the characteristic polynomial are given by
√
B+ ∆
p
4(M∞2 − 1) 1
λ1 = = 2
= −p
2A 2(1 − M∞ ) M∞2 − 1
√
B− ∆
p
4(M∞2 − 1) 1
λ2 = =− 2
= p
2A 2(1 − M∞) M∞2 − 1
Therefore, from the characteristic equations (18a) and (18b), we have
dy 1 1
= p , −p
dx M∞2 − 1 M∞2 − 1
Integrating the above two ODEs to obtain the characteristics of the wave equation
x x
y= p + c1 , y = −p + c2
M∞2 − 1 M∞2 − 1
where c1 and c2 are the constants of integration.
p We see that the two families
p of characteristics
for the wave equation are given by y − x/ M∞ − 1 = constant and y + x/ M∞2 − 1 = constant.
2
wξ η = 0
This is the canonical form of the given hyperbolic PDE.p Here ξ = const. and η = const. lines
represent two families of straight lines with slopes, ±1/ M∞2 − 1.
Parabolic equations
For a parabolic PDE the discriminant ∆ = B2 − 4AC = 0. In this case, we have seen that, to
reduce this PDE to canonical form we need to choose the new variables ξ and η such that the
coefficients a and b vanish in (7). Thus, from (8), we have
ξx ξx
2
A +B +C = 0 (25)
ξy ξy
As the total derivative of ξ along the coordinate line ξ (x, y) = const., d ξ = 0. It follows that
d ξ = ξx dx + ξy dy = 0
and hence, the slope of such curves is given by
dy ξx
=−
dx ξy
Using this result, equation (25) can be written as
2
dy dy
A −B +C = 0 (26)
dx dx
This is called the characteristic polynomial of the PDE (2). Since B2 − 4AC = 0 in this case,
the characteristic polynomial (25) has only one root, given by
dy B
= = λ (x, y) (27)
dx 2A
Hence we see that for a parabolic PDE there is only one family of real characteristic curves.
The required variables ξ is determined by the ordinary differential equation (27), known as the
characteristic equations of the PDE (2). This is an ordinary differential equation for families of
curves in the xy-plane along which ξ = constant.
To determine the second transformation variable η , we set b = 0 in (8) so that
2Aξx ηx + Bξx ηy + ξy ηx ) + 2Cξy ηy =0
ξx ξx
2A ηx + B ηy + ηx + 2Cηy =0
ξy ξy
B B
2A − ηx + B − ηy + ηx + 2Cηy =0
2A 2A
B2
−Bηx − ηy + Bηx + 2Cηy =0
2A
B2 − 4AC ηy
=0
Since B2 − 4AC = 0 for a parabolic PDE, ηy could be an arbitrary function of (x, y) and con-
sequently the transformation variable η can be chosen arbitrarily, as long as the change of
coordinates formulas define a non-degenerate transformation.
If the coefficients A, B, and C are constants, it is easy to integrate equation (27) to obtain
the expressions for change of variable formulas for reducing a parabolic PDE to the canonical
form. Thus, integration of (27) produces
B
y= x + c1 (28a)
2A
or
B
y− x = c1 (28b)
2A
Since the families of curves ξ = constant are the characteristic curves, the change of variables
are given by the following equations:
B
ξ = y− x (29)
2A
η =x (30)
where we have set η = x. The Jacobian of this transformation is
ξx ξy
−B/2A 1
J = = = −1 6= 0
ηx ηy 1 0
In these new coordinate variables given by (29) and (30), equation (7) reduces to following
canonical form:
wηη = ψ ξ , η , w, wξ , wη
(31)
where ψ = φ /c. As the choice of η is arbitrary, the form taken by ψ will depend on the choice
of η . We have from (8)
c = Aηx2 + Bηx ηy + Cηy2 = A (32)
Equation (7) may also assume the form
wξ ξ = ψ ξ , η , w, wξ , wη
(33)
if we choose c = 0 instead of a = 0.
Example 3
Show that the one-dimensional heat equation
∂ 2u ∂u
α =
∂x 2 ∂t
is parabolic, choose the appropriate characteristic variables, and write the equation in equivalent
canonical form.
Solution It follows that the heat equation is a constant coefficient equation with
A = α, B = 0, C=0
We calculate the discriminant, ∆ = 0, and therefore the PDE is parabolic. The single root of
the characteristic polynomial is given by
λ = B/2A = 0
Elliptic equations
For an elliptic PDE the discriminant ∆ = B2 − 4AC < 0. In this case, we have seen that, to
reduce this PDE to canonical form we need to choose the new variables ξ and η to produce
b = 0 and a = c, or b = 0 and a − c = 0. Then, from (8) we obtain the following equations:
A ξx2 − ηx2 + B(ξx ξy − ηx ηy ) + C ξy2 − ηy2 = 0
(34a)
2Aξx ηx + B(ξx ηy + ξy ηx ) + 2Cξy ηy = 0 (34b)
For hyperbolic and parabolic PDEs, ξ and η are satisfied by equations that are not coupled
each other (see (13) and (25)). However, equations (34) are coupled since both unknowns ξ
and η appear in both equations. In an attempt to separate them, we add the first of these
equation to complex number i times the second to give
A (ξx + iηx )2 + B(ξx + iηx )(ξy + iηy ) + C (ξy + iηy )2 = 0
Dividing the above equation throughout by (ξy + iηy )2 to obtain
ξx + iηx 2 ξx + iηx
A +B +C = 0 (35)
ξy + iηy ξy + iηy
This equation can be solved for two possible values of the ratio
√ √
ξx + iηx −B ± B2 − 4AC −B ± i 4AC − B2
= = (36)
ξy + iηy 2A 2A
Clearly, these two roots are complex conjugates and are given by
√
αx −B + i 4AC − B2
= (37a)
αy 2A
√
βx −B − i 4AC − B2
= (37b)
βy 2A
where β (x, y) is the complex conjugate of α (x, y). They are given by
α (x, y) = ξ (x, y) + iη (x, y) (38a)
β (x, y) = ξ (x, y) − iη (x, y) (38b)
We will now proceed in a purely formal fashion. As the total derivative of α along the coordinate
line α (x, y) = constant, d α = 0. It follows that
d α = αx dx + αy dy = 0
and hence, the slope of such curves is given by
dy αx
=−
dx αy
We also have a similar result along coordinate line β (x, y) = constant, i.e.,
dy βx
=−
dx βy
From the foregoing discussion it follows that
√
dy B − i 4AC − B2
= λ1 = (39a)
dx √2A
dy B + i 4AC − B2
= λ2 = (39b)
dx 2A
Equations (39a) and (39b) are called the characteristic equation of the PDE (2). Clearly, the
solution of this differential equations are necessarily complex-valued and as a consequence there
are no real characteristic exist for an elliptic PDE.
The complex variables α and β are determined by the respective solutions of the two
ordinary differential equations (39a) and (39b). Integration of equation (39a) leads to the
family of curvilinear coordinates α (x, y) = c1 while the integration of (39b) gives another family
of curvilinear coordinates β (x, y) = c2 , where c1 and c2 are complex constants of integration.
Since α and β are complex function the characteristic curves of the elliptic equation (2) are
not real.
Now the real and imaginary parts of α and β give the required transformation variables ξ
and η . Thus, we have
α +β α −β
ξ = η = (40)
2 2i
With the choice of coordinate variables (40), equation (7) reduces to following canonical form:
wξ ξ + wηη = ψ ξ , η , w, wξ , wη
(41)
where ψ = φ /a.
Note: It may be noted that the quasilinear second-order equations in two independent variables
can also be classified in a similar way according to rule analogous to those developed above for
semilinear equations. However, since A, B, and C are now functions of ux , uy , and u its type
turns out to depend in general on the particular solution searched and not just on the values
of the independent variables.
Example 4
Show that the equation
uxx + x2 uyy = 0
is elliptic everywhere except on the coordinate axis x = 0, find the characteristic variables and
hence write the equation in canonical form.
Solution The given equation is of the form (2) where
A = 1, B = 0, C = x2
The discriminant, ∆ = B2 − 4AC = −4x2 < 0 for x 6= 0, and therefore the PDE is elliptic. The
roots of the characteristic polynomial are given by
√ √
B − i 4AC − B2 B + i 4AC − B2
λ1 = = −ix and λ2 = = ix
2A 2A
Therefore, from the characteristic equations (18a) and (18b), we have
dy dy
= −ix, = ix
dx dx
Integrating the above two ODEs to obtain the characteristics of the wave equation
x2 x2
y = −i + c1 , y=i + c2
2 2
where c1 and c2 are the complex constants. We see that the two families of complex charac-
teristics for the elliptic equation are given by y + ix2 /2 = constant and y − ix2 /2 = constant. It
follows, then, that the transformation
x2 x2
α = y+i , β = y−i
2 2
The real and imaginary parts of α and β give the required transformation variables ξ and η .
Thus, we have
α +β α −β x2
ξ = =y η = =
2 2i 2
With this choice of coordinate variables, equation (7) reduces to following canonical form. From
the relations (6), we have
Substituting these relations in the given PDE and noting that x2 = 2η , we obtain
1
wξ ξ + wηη = − wη
2η
This is the canonical form of the given hyperbolic PDE. Therefore, the PDE
uxx + x2 uyy = 0
in rectangular coordinate system (x, y) has been transformed to PDE
1
wξ ξ + wηη = − wη
2η
in curvilinear coordinate system (ξ , η ). Here ξ = const. lines represents a family of straight
lines parallel to x axis and η = const. lines represents family of parabolas.
Example 5
Consider the Tricomi equation
uxx − xuyy = 0
This is simple model of a second-order PDE of mixed elliptic-hyperbolic type with two inde-
pendent variables. The Tricomi equation is a prototype of the Chaplygin’s equation for study
of transonic flow.
The Tricomi equation is of the form (2) where
A = 1, B = 0, C = −x
The discriminant, ∆ = B2 − 4AC = 4x. Therefore, the Tricomi equation is hyperbolic for x > 0,
elliptic for x < 0 and degenerates to an equation of parabolic type on the line x = 0. Assuming
x > 0, the roots of the characteristic polynomial are given by
√ √
B+ ∆ √ B− ∆ √
λ1 = = x and λ2 = =− x
2A 2A
Therefore, from the characteristic equations (18a) and (18b), we have
dy √ dy √
= x, =− x
dx dx
Integrating the above two ODEs to obtain the characteristics of the wave equation
y = 32 x3/2 + c1 , y = − 32 x3/2 + c2
where c1 and c2 are the constants of integration. We see that the two families of characteristics
for the wave equation are given by y − 2/(3x3/2 ) = constant and y + 2/(3x3/2 ) = constant. It
follows, then, that the transformation
ξ = y − 32 x3/2 , η = y + 32 x3/2
reduces the wave equation to canonical form. The derivatives of ξ and η are given by
√
ξx = − x ξy = 1
ξxx = − 2√ 1
ξyy = 0
√ x
ηx = x ηy = 1
ηxx = 2√ 1
x
ηyy = 0
17
From the relations (6), we have
3(η − ξ )
x3/2 =
4
Substituting these relations in the given PDE to obtain
1 wξ − wη
wξ η =
6 ξ −η
This is the canonical form of the Tricomi equation in the hyperbolic region.
Example 6
An interesting example is provided by the stationary potential flow equation in two dimensions,
defined by equation (where c designates the speed of sound):
u2 ∂ 2 φ uv ∂ 2 φ v2 ∂ 2 φ
1− 2 −2 2 + 1− 2 =0
c ∂ x2 c ∂ x∂ y c ∂ y2
with
u2 v2
uv
A = 1− 2 , B = −2 2 , C = 1− 2
c c c
we can write the potential equation under the form (2). In this
√ particular case the discriminant
2
(B − 4AC) becomes, introducing the Mach number, M(= u2 + v2 /c)
2
u + v2
2 2
B − 4AC = 4 − 1 = 4 M − 1
c2
and hence the stationary potential equation is elliptic for subsonic flows and hyperbolic for
supersonic flows. Along the sonic line M = 1, the equation is parabolic. This mixed nature of the
potential equation has been a great challenge for the numerical computation of transonic flows
since the transition line between the subsonic and the supersonic regions is part of the solution.
An additional complication arises from the presence of shock waves which are discontinuities of
the potential derivatives and which can arise in the supersonic regions.
Classification of Second-Order Equations in n Variables
We now consider the classification to second-order PDEs in more than two independent vari-
ables. To extend the examination of characteristics for more than two independent variables is
less useful. In n dimension, we need to consider (n − 1) dimensional surfaces. In three dimen-
sions, it is necessary to obtain transformations, ξ = ξ (x, y, z), η = η (x, y, z), and ζ = ζ (x, y, z)
such that all cross derivatives in (ξ , η , ζ ) disappear. However, this approach will fail for more
than three independent variables and hence it is not usually possible to reduce the equation to
a simple canonical form. Consider a general second-order semilinear partial differential equation
in n independent variables
∂ 2u ∂ 2u ∂ 2u ∂ 2u
a11 + a12 + a13 + · · · + a1n +
∂ x1 ∂ x1 ∂ x1 ∂ x2 ∂ x1 ∂ x3 ∂ x1 ∂ xn
∂ 2u ∂ 2u ∂ 2u ∂ 2u
a21 + a22 + a23 + · · · + a2n +
∂ x2 ∂ x1 ∂ x2 ∂ x2 ∂ x2 ∂ x3 ∂ x2 ∂ xn
.. .. .. .. ..
. . . . .
∂ 2u ∂ 2u ∂ 2u ∂ 2u
an1 + an2 + an3 + · · · + ann +
∂ xn ∂ x1 ∂ xn ∂ x2 ∂ xn ∂ x3 ∂ xn ∂ xn
∂u ∂u ∂u ∂u
b1 + b2 + b3 + ··· + bn + cu + d = 0
∂ x1 ∂ x2 ∂ x3 ∂ xn
For more than three independent variables it is convenient to write the above PDE in the
following form:
n n
∂ 2u n
∂u
∑ ∑ i j ∂ xi ∂ x j ∑ bi ∂ xi + cu + d = 0
a + (42)
i=1 j=1 i=1
The coefficient matrix of the terms ∂ 2 u/(∂ ξk ∂ ξl ) in this transformed expression is seen to be
equal to QT AQ. That is,
(qki ai j ql j ) ≡ QT AQ
From linear algebra, we know that for any real symmetric matrix A, there is an associate
orthogonal matrix P such that PT AP = Λ. Here P is called diagonalizing matrix of A and Λ is
a diagonal matrix whose element are the eigenvalues, λi , of A and the columns of P the linearly
independent eigenvectors of A, ei = (e1i , e2i , . . ., eni ). So, we have
P = (ei j ) and Λ = (λi δi j ), i, j = 1, 2, . . ., n
where δi j is the Kronecker delta. Now if the transformation is such that Q is taken to be a
diagonalizing matrix of A, it follows that
λ1
λ2
Q AQ = Λ =
T
(44)
. . .
λn
The numbers λ1 , λ2 , . . . , λn are real, because the eigenvalues of a real symmetric matrix are
always real. It is instructive to note that the previously mentioned transformation (for second-
order PDE with two independent variables) to remove cross derivatives is equivalent to finding
eigenvalues λi of the coefficient matrix A.
We are now in a position to classify the equation (42).
• Equation is called elliptic if all eigenvalues λi of A are non-zero and have the same sign.
• Equation is called hyperbolic if all eigenvalues λi of A are non-zero and have the same
sign except for one of the eigenvalues.
• Equation is called parabolic if any of the eigenvalues λi of A is zero. This means that the
coefficient matrix A is singular.
When more than two independent variables are involved, there are other intermediate classifi-
cations exist which depends on the number of zero eigenvalues and the pattern of signs of the
non-zero eigenvalues. These sub classification has not much practical importance and will not
be discussed here.
Example 7
Classify the three-dimensional Laplace equation
uxx + uyy + uzz = 0
Example 8
Classify the two-dimensional wave equation
utt − c2 (uxx + uyy ) = 0
ut − α (uxx + uyy ) = 0
As the coefficient matrix is already in diagonalized form it can be seen immediately that it has
a zero eigenvalue. Hence, according to the classification rule the given PDE is parabolic.
Example 10
Classify the two-dimensional equation
∂ 2u ∂ 2u ∂ 2u
− + = f (x, y)
∂ x2 ∂ x∂ y ∂ y2
∂ 2u 1 ∂ 2u 1 ∂ 2u ∂ 2u
− − + = f (x, y)
∂ x2 2 ∂ x∂ y 2 ∂ y∂ x ∂ y2
The coefficient matrix is then given by
1 − 21
A=
− 12 1
Since the coefficient matrix not in diagonalized form, we solve the eigenvalue problem det(A −
λ I) = 0. That is,
1 − λ −1
2
−1 1 − λ = 0
2
Expanding the determinant to obtain
1 3
(1 − λ )2 − =0 =⇒ λ 2 − 2λ + =0
4 4
Hence the two eigenvalues are λ1 = 1/2 and λ2 = 3/2. The equation is elliptic.
Example 11
Classify the following small disturbance potential equation for compressible flows:
∂ 2φ ∂ 2φ
1 − M∞2 + =0
∂ x2 ∂ y2
1 − M∞2 0
A=
0 1
As the coefficient matrix is already in diagonalized form it can be seen immediately that its
eigenvalues are given by
λ1 = 1, λ2 = 1 − M∞2
It follows that: If M∞2 < 1 then all eigenvalues are nonzero and of same sign thus small distur-
bance potential equation is elliptic, if M∞2 = 1 then one of the eigenvalues is zero thus small
disturbance potential equation is parabolic, and if M∞2 > 1 then all eigenvalues are nonzero and
are of opposite sign thus small disturbance potential equation is hyperbolic.
The system of first-order PDE of three independent variables can be generalized for n dependent
variables, u j , as follows:
n ∂uj ∂uj ∂uj
∑ ai j ∂ x + bi j ∂ y + ci j ∂ z = fi, i = 1, 2, . . ., n (45a)
j=1
or in matrix form
∂U ∂U ∂U
A +B +C =F (45b)
∂x ∂y ∂z
where
a11 · · · a1n b11 · · · b1n c11 · · · c1n
A = ... . . . ... B = ... . . . ... C = ... . . . ...
∂uj n ∂uj
+ ∑ di j = ei , i = 1, 2, . . ., n
∂ x j=1 ∂y
∂ u1 ∂ u1 ∂ u2 ∂ un
+ d11 + d12 + · · · + d1n = e1
∂x ∂y ∂y ∂y
∂ u2 ∂ u1 ∂ u2 ∂ un
+ d21 + d22 + · · · + d2n = e2
∂x ∂y ∂y ∂y
.. .. .. .. .. ..
. . . . . .
∂ un ∂ u1 ∂ u2 ∂ un
+ dn1 + dn2 + · · · + dnn = en
∂x ∂y ∂y ∂y
Just as in the case of a single partial differential equation, the important properties of
solutions of the system (48) depend only on its principal part Ux + DUy . Since this principal
part is completely determined by the coefficient matrix D(x, y) = A−1 B, this matrix plays a
fundamental role in the study of (47).
V = P−1U ⇒ U = PV
∂U ∂V ∂P ∂U ∂V ∂P
=P + V, =P + V
∂x ∂x ∂x ∂y ∂y ∂y
Substituting these into (48) to obtain
∂V ∂P ∂V ∂P
P + V + DP +D V = E
∂x ∂x ∂y ∂y
Rearranging,
∂V ∂V ∂P ∂P
P + DP =E− +D V =G
∂x ∂y ∂x ∂y
We multiply the above equation by P−1 to obtain
∂V ∂V
+ P−1 DP =H (49)
∂x ∂y
where
∂P ∂P
−1 −1
H =P G=P E− +D V .
∂x ∂y
Let Λ be the n × n diagonal matrix with diagonal entries the eigenvalues of D, and P is taken
to be the n × n diagonalizing matrix of D with columns the corresponding eigenvectors of D,
i.e.,
λ1 · · · 0
p11 · · · p1n
Λ = (λi δi j ) = ... . . . ... , P = (pi j ) = ... . . . ...
0 · · · λn pn1 · · · pnn
where δi j is the Kronecker delta. It follows that
P−1 DP = Λ. (50)
It is instructive to note that the previously mentioned transformations (for second-order PDE
with two independent variables) to remove cross derivatives is equivalent to diagonalizing the
matrix D = A−1 B and then finding eigenvalues λi of the matrix D of the system of PDEs
(47). Thus, we can write the system of equations (49) [and consequently (47)] in the following
canonical form:
∂V ∂V
+ Λ(x, y) = H(x, y,V ) (51a)
∂x ∂y
The simplicity of the canonical form (51a) becomes apparent if we write it in component form,
∂ vi ∂ vi
+ λi (x, y) = hi (x, y, v1 , . . . , vn ), i = 1, 2, . . ., n (51b)
∂x ∂y
or
∂ v1 ∂ v1
+ λ1 = h1
∂x ∂y
∂ v2 ∂ v2
+ λ2 = h2
∂x ∂y
..
. (51c)
∂ vn ∂ vn
+ λn = hn
∂x ∂y
It is clear that the principal part of the ith equation involves only the ith unknown vi .
The classification of the system of first-order PDEs (47) is done based on the nature of the
eigenvalues λi of the matrix P−1 DP, which are exactly the eigenvalues values of D = A−1 B.
Recall that an eigenvalue of D is a root λ of the characteristic equation
|D − λ I| = 0.
The system (47) based on the nature of its eigenvalues is classified as follows:
• If all the n eigenvalues of D are real and distinct the system is called hyperbolic type.
• If all the n eigenvalues of D are complex the system is called elliptic type.
• If some of the n eigenvalues are real and other complex the system is considered as hybrid
of elliptic-hyperbolic type.
• If the rank of matrix D is less than n, i.e., there are less than n real eigenvalues (some of
the eigenvalues are repeated) then the system is said to be parabolic type.
For a system of PDE having only two dependent variables, we can determine the eigenvalues
of matrix D analytically and state the conditions classification in an explicit manner as follows.
Consider the system of two equations (with dependent variables u1 and u2 ) in two dimensions
(x, y):
∂ u1 ∂ u2 ∂ u1 ∂ u2
a11 + a12 + b11 + b12 = f1 (52a)
∂x ∂x ∂y ∂y
∂ u1 ∂ u2 ∂ u1 ∂ u2
a21 + a22 + b21 + b22 = f2 (52b)
∂x ∂x ∂y ∂y
In matrix form,
∂ ∂
a11 a12 u1 b11 b12 u1 f1
+ =
a21 a22 ∂x u2 b21 b22 ∂y u2 f2
or
∂ ∂
u1 u1
A +B =F
∂x u2 ∂y u2
where
a11 a12 b11 b12 f1
A= , B= , F =
a21 a22 b21 b22 f2
The inverse of A is
−1 1 a22 −a12
A =
|A| −a21 a11
where |A| is the determinant of matrix A. We now compute the matrix D as
−1 1 a22 −a12 b11 b12 1 a22 b11 − a12 b21 a22 b12 − a12 b22
D=A B= =
|A| −a21 a11 b21 b22 |A| a11 b21 − a21 b11 a11 b22 − a21 b12
so that the system (52) may be written as
|D − λ I| = 0
• If |b|2 − 4|A||B| > 0, there exists two real and distinct eigenvalues and thus the system
is hyperbolic.
• If |b|2 − 4|A||B| < 0, there exists two complex eigenvalues and thus the system is elliptic.
• If |b|2 − 4|A||B| = 0, there is only one real eigenvalue and thus the system is parabolic.
A special case
When A is an identity matrix, the system of equation (47) takes the form
∂U ∂U
+ B(x, y) = F(x, y,U ) (55)
∂x ∂y
so that, we have
D=B and E =F
For a system of PDE having only two dependent variables, (55) becomes
∂ u1 ∂ u1 ∂ u2
+ b11 + b12 = f1 (56a)
∂x ∂y ∂y
∂ u2 ∂ u1 ∂ u2
+ b21 + b22 = f2 (56b)
∂x ∂y ∂y
or, in matrix form
∂ ∂
u1 b11 b12 u1
+ =F
∂x u2 b21 b22 ∂y u2
To determine eigenvalues of D, we solve the eigenvalue problem |B − λ I| = 0. The corre-
sponding characteristic equation is
λ 2 − |b|λ + |B| = 0
where the determinants |B| and |b| are given by
b11 b12
|B| = = b11 b22 − b12 b21
b21 b22
b11 0 1 b12
|b| =
+ = b11 + b22
b21 1 0 b22
The two roots of the quadratic equation for λ are given by
p
|b| ± |b|2 − 4|B|
λ1,2 = (57)
2
Example 12
Classify the single first-order equation
∂u ∂u
a +b = f
∂x ∂y
where a and b are real constants.
Solution In the standard matrix form the above equation may be written as
∂U ∂U
A +B =F
∂x ∂y
where
A= a , B= b , U = u , F = f
The D matrix, in this case, can be easily found:
D = A−1 B = a−1
b = b/a
The matrix D has the eigenvalue, λ = b/a. This is always real and hence, a single first-order
PDE is always hyperbolic in the space (x, y). Note that here we have only a single eigenvalue
and thus a characteristic direction.
Example 13
Classify the following system of first-order equation:
∂φ ∂ψ
a +c = f1
∂x ∂y
∂ψ ∂φ
b +d = f2
∂x ∂y
The matrix form of the system is:
a 0 ∂ φ 0 c ∂ φ
f1
+ =
0 b ∂x ψ d 0 ∂y ψ f2
∂ φ 0 c/a ∂ φ
e1
+ =
∂x ψ d/b 0 ∂y ψ e2
∂φ ∂ψ
+ =0
∂x ∂y
∂ψ ∂φ
+ =0
∂x ∂y
By eliminating the variable ψ and replacing y by t, we obtain the well-known wave equation in
φ:
∂ 2φ ∂ 2φ
=
∂ t2 ∂ x2
which is a hyperbolic equation as seen previously.
If cd/ab < 0 then the eigenvalues are complex and the system is elliptic in the space (x, y).
For instance, a = b = 1; c = −d = −1 and vanishing right-hand side, the system of equation
becomes the well-known Cauchy–Riemann equation:
∂φ ∂ψ
− =0
∂x ∂y
∂ψ ∂φ
+ =0
∂x ∂y
By eliminating the variable ψ , we obtain the Laplace equation in φ :
∂ 2φ ∂ 2φ
+ 2 =0
∂ x2 ∂y
which is the standard form of elliptic equations and describes steady-state diffusion phenomena.
Note that we could also obtain the Laplace equation in ψ by eliminating the variable φ .
Finally, if one of the coefficients is equal zero, say c, then there is only one real eigenvalue
and the system is parabolic. For instance, with a = −b = 1, c = 0, d = 1 and f1 = ψ , f2 = 0,
the system of equation becomes
∂φ
=ψ
∂x
∂φ ∂ψ
− =0
∂y ∂x
which on eliminating the variable φ and replacing y by t leads to the standard form for a
parabolic equation:
∂φ ∂ 2φ
=
∂t ∂ x2
This is recognizable by the fact that the equation presents a combination of first and second-
order derivatives.
Example 14
Let us find out the canonical form the system of first-order equations
∂ u1 ∂ u2
+c =0
∂x ∂y
(58)
∂ u2 ∂ u1
+d =0
∂x ∂y
Solution The given system is of the form (55) and can be written as
∂ u1 0 c ∂
u1 0
+ =
∂ x u2 d 0 ∂ y u2 0
or
∂U ∂U
+B =F
∂x ∂y
where,
0 c u1 0
B= , U = , F =
d 0 u2 0
Therefore, we have
Therefore, √
cd 0
P −1
BP = Λ = √
0 − cd
The canonical form is then given by
∂V ∂V
+Λ =0
∂x ∂y
or √
∂ ∂
v1 cd 0
√ v1 0
+ =
∂x v2 0 − cd ∂y v2 0
or in component form
∂ v1 √ ∂ v1
+ cd =0
∂x ∂y
(59)
∂ v2 √ ∂ v2
− cd =0
∂x ∂y
The canonical form (59) is particularly simple. Each equation involves only one unknown and
can be easily solved by the methods of characteristics. The general solution of the system (59)
is √
v1 = f (x − cdy)
√ (60)
v2 = g(x + cdy)
where f and g are arbitrary functions of a single variable.
The vectors U and V are related through the following transformation:
√ √ √
u1 c c v1 c(v1 + v2 )
U = PV ⇒ = √ √ = √ (61)
u2 d − d v2 d(v1 − v2 )
Now the general solution of (58) can be obtained using the relation (61) as
√ √ √
u1 = c[ f (x − cdy) + g(x + cdy)]
√ √ √ (62)
u2 = d[ f (x − cdy) − g(x + cdy)]
Example 15
The one-dimensional form of the time-dependent shallow water equations can be written as
∂h ∂h ∂u
+u +h =0
∂t ∂x ∂x
∂u ∂u ∂h
+u +g =0
∂t ∂x ∂x
where h represents the water height, g is the gravity acceleration and u the horizontal velocity.
Solution Since A is a unit matrix, the given system in matrix form
∂ h u h ∂
h
+ = 0
∂t u g u ∂x u
Introducing the vector
h
U =
u
the system is written in the condensed form:
∂U ∂U
+B =0
∂t ∂x
where,
u h h 0
B= , U = , F =
g u u 0
Therefore, we have
Now, the diagonal matrix Λ by definition is P−1 BP. As we have already found the eigenvalues
of B, the matrix Λ can be directly obtained as
√
u + gh 0
Λ= √
0 u − gh
The canonical form is then given by
∂V ∂V
+Λ =0
∂t ∂x
or √
∂ ∂ v1
v1 u + gh 0 0
+ √ =
∂t v2 0 u − gh ∂ x v2 0
or in component form
∂ v1 √ ∂ v1
+ (u + gh) =0
∂t ∂x
∂ v2 √ ∂ v2
+ (u − gh) =0
∂t ∂x
Each equation involves only one unknown and can be easily solved by the methods of charac-
teristics. The general solution of the system is
√
v1 = f1 [x − (u + gh)t]
√ (63)
v2 = f2 [x − (u − gh)t]
Now the general solution can be obtained using the relation (63) as
√ √ √
h = h[ f1 (x − (u + gh)t) − f2 (x − (u − gh)t)]
√ √ √ (64)
u = g[ f1 (x − (u + gh)t) + f2 (x − (u − gh)t)]
The procedure for classification of semilinear PDEs is equally applicable for quasilinear PDEs.
We illustrate this with an example from fluid dynamics.
Example 16
Classify the Euler equations for unsteady, one-dimensional flow:
∂ρ ∂ρ ∂u
+u +ρ =0
∂t ∂x ∂x
∂u ∂u 1 ∂p c2 ∂ ρ
+u =− =−
∂t ∂x ρ ∂x ρ ∂x
where the speed of sound c is given by the isentropic relation between pressure and density as
∂p
2
c =
∂ρ s
35
Solution In matrix form the Euler equation can be written as
1 0 ∂ ρ ρ ∂ ρ
u
+ = 0
0 1 ∂t u c2 /ρ u ∂ x u
∂U ∂U
A +B =0
∂t ∂x
where,
ρ ρ
1 0 u 0
A= , B= , U = , F =
0 1 c /ρ u
2 u 0
Therefore, we have
Since A is a unit matrix, the inverse of A is A itself. We now compute the matrix D:
ρ
−1 u
D=A B=B=
c2 /ρ u
∂U ∂U
+D =0
∂t ∂x
The eigenvalues of D = A−1 B are given by
p p
|b| ± |b|2 − 4|A||B| 2u ± 4u2 − 4(u2 − c2 )
λ1,2 = = = u±c
2|A| 2
ρ ρ
u e11 e11 e11
= (u + c) =⇒ =
c2 /ρ u e12 e12 e12 c
ρ −ρ
u e21 e21 e21
= (u − c) =⇒ =
c2 /ρ u e22 e22 e22 c
ρ −ρ
P=
c c
Therefore,
u+c 0
P DP = Λ =
−1
0 u−c
The canonical form is then given by
∂V ∂V
+Λ =0
∂t ∂x
or
∂ ∂
v1 u+c 0 v1 0
+ =
∂t v2 0 u−c ∂x v2 0
or in component form
∂ v1 ∂ v1
+ (u + c) =0
∂t ∂x
∂ v2 ∂ v2
+ (u − c) =0
∂t ∂x
Each equation involves only one unknown and can be easily solved by the standard methods of
characteristics. The general solution of the system is
v1 = f1 [x − (u + c)t]
(65)
v2 = f2 [x − (u − c)t]
ρ −ρ ρ (v1 − v2 )
h v1
U = PV ⇒ = =
u c c v2 c(v1 + v2 )
Now the general solution can be obtained using the relation (65) as
ρ = ρ [ f1 (x − (u + c)t) − f2 (x − (u − c)t)]
(66)
u = c[ f1 (x − (u + c)t) + f2 (x − (u − c)t)]
Note: Since both the eigenvalues are real, for all values of the velocity u, the system is always
hyperbolic in space and time. This is an extremely important property that the steady isentropic
Euler equations are elliptic in the space (x, y) for subsonic velocities and hyperbolic in the space
(x, y) for supersonic velocities.
Here, in space and time, the inviscid isentropic equations are always hyperbolic independently
of the subsonic or supersonic state of the flow. As a consequence, the same numerical algorithms
can be applied for all flow velocities. On the other hand, dealing with the steady state equations,
the numerical algorithms will have to adapt to the flow regime, as the mathematical nature of
the system of equations is changing when passing from subsonic to supersonic, or inversely.
38