You are on page 1of 156

YASHWANTRAO CHAVAN MAHARASHTRA OPEN UNIVERSITY

SCHOOL OF SCIENCES
(FORMERLY SCHOOL OF ARCHITECTURE, SCIENCE & TECHNOLOGY)

V151: M.Sc. Mathematics


2023 {As per NEP 2020}
Pattern

MAT503
ORDINARY DIFFERENTIAL
EQUATIONS
(2 Credits)

Semester - I

Email: director.ast@ycmou.ac.in
Website: www.ycmou.ac.in
Phone: +91-253-2231473
Yashwantrao Chavan S25013
Maharashtra Open University
Ordinary Differential
Equations
Brief Contents
Vice Chancellor’s Message ............................................... 3
Forward By The Director .................................................. 4
Credit 01 ................................................................................ 5
Unit 01-01: Linear equations with constant coefficients 6
Unit 01-02: Dependence and independence of solutions 22
Unit 01-03: Applications of second order linear equations 40
Unit 01-04: The homogeneous equation of higher order 52
Credit 02 .................................................................................
Unit 02-01: The non-homogeneous equation of higher order 67
Unit 02-02: Linear Equations with variable Coefficients 84
Unit 02-03: Reduction of the order 103
Unit 02-04: Homogeneous equations with analytic coefficients 124

Answers to Self-Tests and End of Unit Exercises ......................... 146

S25013: Ordinary Differential Equations Page 1


6Ordinary Diffe re ntial Equations
Yas hwantrao Chavan M aharas htra Ope n Unive rs ity
Vice -Chance llor: Prof. Dr. E. Vayunandan
School of Archite cture , Scie nce and Te chnology
Programme Advis ory Committe e (PAC)
Dr Sunanda M ore Dr M anoj Kille dar Dr Che tana Kamlas k ar
Director(I/c) & Associate Associate P rofessor, Assist. P rofessor,
P rofessor, School of School of Architecture, School of Architecture,
Architecture, Science & Science & Technology, Science & Technology,
Technology, YCMOU, Nashik YCMOU, Nashik YCMOU, Nashik
Dr. T.M . Karade Prof. Dr Shivdas D Katore Prof. Dr J N Salunk e
Retired P rofessor, P rofessor, P rofessor, Swami Ramanand
R.T.M. Nagpur Sant Gadge Baba,
Teerth Marathwada
University, Nagpur Amravati University
University, Nanded
Prof. Dr M e e nak s hi Prof. Dr S R Chaudhari, --
Was adik ar, P rofessor and HOD,
P rofessor and HOD, KBC North Maharashtra
Dr Babasaheb Ambedkar University, Jalgaon
Marathwada University

Development Team
Co u rs e Co o rd in at o r an d Bo o k Writ er Bo o k Ed it o r
In s t ru ctio n al Tech n o lo g y Ed it o r
Dr. Che tana Kamlas k ar Dr. L. N. Katk ar Dr. J. N. Salunk e
B.E., M. Tech., P h.D., M. Sc. P h.D. M. Sc. M.P hil., P h.D.,
Assistance P rofessor, Ex P rofessor and Head, P rofessor,
School of Architecture, Depart. of Mathematics, SRTM University,
Science & Technology, Nanded
Shivaji University, Kolhapur
YCMOU, Nashik Experience: 33+ Yrs
Experience: 35+ Yrs.
Experience@YCMOU: Credits written: 01 and 02
20+ Yrs

This work by Y CMOU is licensed under a Creative Commons


Attribution-NonCommercial-ShareAlike 4.0 I nternational
L icense.
 Firs t B ook Publication : 27 Mar 2021 , Publication Number: 2366
 Publis he r: Dr. Dinesh Bhonde, Registrar, YCMOU, Nashik - 422 222, MH, India
 ISB N Numbe r : 978-93-91514-52-5

This SLM Book V130: M.Sc. (Mathematics) {2021 Pattern}, dtd. 27/03/2021
used in V151: M.Sc. (Mathematics) {2023 Pattern}, dtd. 31/08/2023
V ICE C HANCELLOR ’ S M ESSAGE
Dear Students, Greetings!!!
I offer cordial welcome to all of you for the Master’s degree programme of Yashwantrao Chavan
Maharashtra Open University. As a post graduate student, you must have autonomy to learn, have
information and knowledge regarding different dimensions in the fie ld of Mathematics and at the
same time intellectual development is necessary for application of knowledge wisely. The process
of learning includes appropriate thinking, understanding important points, describing these points on the basis
of experience and observation, explaining them to others by speaking or writing about them. The science
of education today accepts the principle that it is possible to achieve excellence and knowledge in
this regard.
The syllabus of this course has been structured in this book in such a way, to give you autonomy to study easily
without stirring from home. During the counseling sessions, scheduled at your respective study
center, all your doubts will be clarified about the course and you will get guidance from some qualified and
experienced counsellors / professors. This guidance will not only be based on lectures, but it will also
include various techniques such as question-answers, doubt clarification. We expect your active participation
in the contact sessions at the study center. Our emphasis is on ‘self-study’. If a student learns how to study, he will
become independent in learning throughout life. This course book has been written with the objective of helping in self-
study and giving you autonomy to learn at your convenience.
During this academic year, you have to give assignments, complete laboratory activities, field
visits and the Project work wherever required. You may have to opt for specialization as per
programme structure. You will get experience and joy in personally doing above activities. This will enable you
to assess your own progress and there by achieve a larger educational objective.
We wish that you will enjoy the courses of Yashwantrao Chavan Mahar ashtra Open University,
emerge successful and very soon become a knowledgeable and honorable Master’s degree holder
of this university. I congratulate “Development Team” for the development of this excellent high
quality “Self- Learning Material (SLM)” for the students. I hope and believe that this SLM will
be immensely useful for all students of this program.
Best Wishes!
- Prof. Dr. E. Vayunandan
Vice-Chancellor, YCMOU

S25013: Ordinary Differential Equations Page 3


F ORWARD B Y T HE D IRECTOR

This book aims at acquainting the students with advance Mathematics required at post graduate
degree level.
The book has been specially designed for Science students. It has a comprehensive coverage of
Mathematical concepts and its application in practical life. The book contains numerous
mathematical examples to build understanding and skills.
The book is written with self- instructional format. Each chapter is prepared with articulated
structure to make the contents not only easy to understand but also i nteresting to learn.
Each chapter begins with learning objectives which are stated using Action Verbs as per the
Bloom’s Taxonomy. Each Unit is started with introduction to arouse or stimulate curiosity of
learner about the content/ topic. Thereafter the unit contains explanation of concepts supported
by tables, figures, exhibits and solved illustrations wherever necessary for better effectiveness
and understanding.
This book is written in simple language, using spoken style and short sentences. Topics of each
unit of the book presents from simple to complex in logical sequence. This book is appropriate
for low achiever students with lower intellectual capacity and covers the syllabus of the course.
Exercises given in the chapter include MCQs, conceptual questions and practical questions so as
to create a ladder in the minds of students to grasp each and every aspect of a particular concept.
I thank the students who have been a constant motivation for us. I am grateful to the writers,
editors and the School faculty associated in this SLM development of the Programme.

- Dr. Sunanda More


Director (I/C) & Associate Professor,
School of Architecture, Science and Technology
-

S25013: Ordinary Differential Equations Page 4


C REDIT 01

S25013: Ordinary Differential Equations Page 5


UNIT 01-01: LINEAR EQUATIONS WITH CONSTANT COEFFICIENTS
LEARNING OBJECTIVES
After successful completion of this unit, you will be able to
❖ explain linear ODEs with constant coefficients
❖ apply the method to solve second order homogeneous linear ODE with constant
coefficients.

BASIC BACKGROUNDS:
Any real situation is described either by a system of ordinary differential equations or partial
differential equations. For example, if a particle is projected with initial velocity u, then its motion
under gravity is described by a second order differential equation. Similarly, the motion of a planet
in our solar system under the action of a central force is also described by the second order
differential equation. The solution of the differential equation determines the trajectory of the
particle. Hence the study of differential equations is cru cial.
Polynomial: A polynomial is a real (complex) valued function 𝑝 whose domain is the set of all
real (complex) numbers, and which has from
𝑝(𝑥) = 𝑎0 𝑥 𝑛 + 𝑎1 𝑥 𝑛−1 +. . . +𝑎𝑛−1 𝑥 + 𝑎𝑛 , (1)
where 𝑛 is a non-negative integer and 𝑎0 , 𝑎1 , . . . , 𝑎𝑛 are constants. The highest power of 𝑥
with non-zero coefficient is called the degree of the polynomial. A root of the polynomial is a real
(complex) number 𝑟 such that 𝑝(𝑟) = 0.
Fundamental Theorem of algebra: If 𝑝 is a polynomial such that 𝑑𝑒𝑔(𝑝) ≥ 1, then 𝑝 has
at least one root.
Corollary 1: Let 𝑝(𝑧) = 𝑧 𝑛 + 𝑎1 𝑧 𝑛−1 +. . . +𝑎𝑛−1 𝑧 + 𝑎𝑛 be a complex polynomial of degree
𝑛 ≥ 1 with leading coefficient 1 (i.e coefficient of 𝑧 𝑛 ) and let r be a root of 𝑝.
Then
𝑃(𝑧) = (𝑧 − 𝑟)𝑞(𝑟), (2)
where 𝑞 is a polynomial of degree 𝑛 – 1, with leading coefficient 1.
Proof: Proof is left to the readers.
Note: If 𝑧 = 𝑟 is a root of 𝑝(𝑧) = 0, then 𝑝(𝑟) = 0. If (𝑛 − 1) ≥ 1. The polynomial 𝑞(𝑧)
has a root and this root is also a root of 𝑝(𝑧). By applying the fundamental theorem of
algebra 𝑛 – times together with Corollary 1 we have the following.
Corollary 2: If 𝑝 is a polynomial of degree 𝑛 ≥ 1 with leading coefficient 𝑎0 ≠ 0, then 𝑝
has exactly 𝑛 roots. If 𝑟1 , 𝑟2 , . . . , 𝑟𝑛 are the roots of 𝑝(𝑧), then
𝑝(𝑧) = 𝑎0 (𝑧 − 𝑟1 )(𝑧 − 𝑟2 ) . . . (𝑧 − 𝑟𝑛 ). (3)
Note: If the root 𝑟 of 𝑝(𝑧) repeats 𝑚 times, then 𝑟 is called the root of multiplicity 𝑚.
⇒ 𝑝(𝑧) = 𝑎0 (𝑧 − 𝑟)𝑚 𝑞(𝑧), 𝑞(𝑟) ≠ 0. (4)
Theorem 1: If 𝑟 is a root of multiplicity 𝑚 of a polynomial 𝑝 of degree ≥ 1, then 𝑝(𝑟) =
𝑝′ (𝑟) = ⋯ = 𝑝(𝑚−1) (𝑟) = 0 and 𝑝(𝑚) (𝑟) ≠ 0.
Proof: - Let 𝑝(𝑧) = 𝑧 𝑛 + 𝑎1 𝑧 𝑛−1 + ... + 𝑎𝑛−1 𝑧 + 𝑎𝑛 (5)
be a polynomial with 𝑎0 = 1 and of degree ≥ 1. Suppose 𝑟 is a root of 𝑝(𝑧) of multiplicity
𝑚, then we have
𝑝(𝑧) = (𝑧 − 𝑟)𝑚 𝑞(𝑧), 𝑞(𝑟) ≠ 0, (6)

S25013: Ordinary Differential Equations Page 6


where 𝑞(𝑧) is a polynomial of degree 𝑛 – 𝑚. Differentiating 𝑝(𝑧), 𝑚 – 1 times, we get
𝑝′ (𝑧) = (𝑧 − 𝑟)𝑚 𝑞 ′ (𝑧) + 𝑚(𝑧 − 𝑟)𝑚−1 𝑞(𝑧)
⇒ 𝑝′(𝑧) = (𝑧 − 𝑟)𝑚−1 [𝑞′(𝑧)(𝑧 − 𝑟) + 𝑚𝑞(𝑧)]
i.e. 𝑝′(𝑧) = (𝑧 − 𝑟)𝑚−1 [𝑎 polynomial of degree 𝑛 − 𝑚].
Similarly,
𝑝′′(𝑧) = (𝑧 − 𝑟)𝑚𝑞′′(𝑧) + 𝑚(𝑧 − 𝑟)𝑚−1 𝑞(𝑧) + 𝑚(𝑧 − 𝑟)𝑚−1 𝑞(𝑧) + 𝑚(𝑚 − 1)(𝑧 − 𝑟)𝑚−2 𝑞(𝑧) ⇒
𝑝′′(𝑧) = (𝑧 − 𝑟)𝑚−2 [(𝑧 − 𝑟)2 𝑞′′(𝑧) + 2𝑚(𝑧 − 𝑟)𝑞′(𝑧) + 𝑚(𝑚 − 1)𝑞(𝑧)]
i.e. p′′(𝑧)=(𝑧 − 𝑟)𝑚−2 [polynomial of degree 𝑛 − 𝑚].
Continuing in this way, we obtain
𝑃(𝑚−1) (𝑧) = (𝑧 − 𝑟) [𝑎 polynomial of degree n − 𝑚].
This shows that if 𝑟 is a root of 𝑝(𝑧) of multiplicity 𝑚, then
𝑝(𝑟) = 𝑝′(𝑟) = 𝑝′′(𝑟) = ⋯ = 𝑝(𝑚−1) (𝑟) = 0 and 𝑝(𝑚) (𝑟) = 𝑚! 𝑞(𝑟) ≠ 0.
𝑃(𝑚) (𝑧) = 𝑚! 𝑞(𝑧) + (𝑧 − 𝑟) [𝑎 polynomial of degree n − 𝑚 − 1].
⇒ 𝑝(𝑚) (𝑟) = 𝑚! 𝑞(𝑟) ≠ 0.
Determinant: There always exists the connection between the determinant and the
solution of system of linear equations. This connection will facilitate to understand the
nature of the solutions of differential equation. To understand this connection , consider a
system of 𝑛 – equations in 𝑛 – unknowns.
𝑎11 𝑥1 + 𝑎12 𝑥2 + ... + 𝑎1𝑛 𝑥𝑛 = 𝑏1 ,
𝑎21 𝑥1 + 𝑎22 𝑥2 + ... + 𝑎2𝑛 𝑥𝑛 = 𝑏2 ,

𝑎𝑛1 𝑥1 + 𝑎𝑛2 𝑥2 + ... + 𝑎𝑛𝑛 𝑥𝑛 = 𝑏𝑛 , (7)
where 𝑎𝑖𝑗 and 𝑏𝑖 are all constants. A set of n numbers 𝑥𝑖 , 𝑖 = 1,2, ..., 𝑛 satisfying these
equations is called a solution of (7).
If 𝑏𝑖 = 0, for all 𝑖, then the system of equations (7) is called homogeneous otherwise non -
homogeneous system of 𝑛 – linear equations.
Let 𝛥 = |𝑎𝑖𝑗 | be the determinant of the coefficients of the equation (7) and 𝛥 ≠ 0, then by
Cramer’s rule, we know the unique solution of the system is given by
𝛥𝑖
𝑥𝑖 = , ∀ 𝑖 = 1,2, ..., 𝑛 (8)
𝛥
where 𝛥𝑖 is the determinant obtained from 𝛥 by replacing 𝑖 th column of 𝛥 by 𝑏1 , 𝑏2 , ..., 𝑏𝑛 .
Nature of solutions:
1) If 𝑏𝑖 = 0, ∀ 𝑖 = 1,2, ..., 𝑛, then Δ𝑖 = 0 for all 𝑖. If 𝛥 ≠ 0, then the system of equations
(7) has unique solution 𝑥𝑖 = 0, ∀ 𝑖 and the solution is trivial.
2) If 𝛥 = 0 and 𝛥𝑖 ≠ 0, then the system has no solution.
3) If 𝛥 = 0 and 𝛥𝑖 = 0, ∀ 𝑖 due to all 𝑏𝑖 = 0, ∀ 𝑖, then there exists infinitely many solutions
and the solution is not unique. Equivalently, we say that not all 𝑥𝑖 are zero.

S25013: Ordinary Differential Equations Page 7


01-01: INTRODUCTION
A linear differential equation of order 𝑛 with constant coefficients is given by
𝑎0 𝑦 (𝑛) + 𝑎1 𝑦 (𝑛−1) + 𝑎2 𝑦 (𝑛−2) + ... + 𝑎𝑛 𝑦 = 𝑏(𝑥), (9)
where 𝑎0 ≠ 0, 𝑎1 , 𝑎2 , ..., 𝑎𝑛 are complex constants and 𝑏 is some complex valued function
on an interval 𝐼. Dividing equation (9) by 𝑎0 , we can obtain an equation of the same form
(9) with 𝑎0 replaced by 1. Therefore, we always assume 𝑎0 = 1 and write the equation (9)
as
𝑦 (𝑛) + 𝑎1 𝑦 (𝑛−1) + 𝑎2 𝑦 (𝑛−2) + ... + 𝑎𝑛 𝑦 = 𝑏(𝑥). (10)
We write this equation (10) as
𝐿(𝑦) = 𝑏(𝑥), (11)
where 𝐿(𝑦) = 𝑦 (𝑛) + 𝑎1 𝑦 (𝑛−1) + 𝑎2 𝑦 (𝑛−2) + ⋯ + 𝑎𝑛 𝑦. (12)
If 𝑏(𝑥) = 0, ∀ 𝑥 ∈ 𝐼, then corresponding equation (11) is 𝐿(𝑦) = 0. (13)
This is called the homogeneous equation. If 𝑏(𝑥) ≠ 0 for some 𝑥 ∈ 𝐼, then 𝐿(𝑦) = 𝑏(𝑥).
This is called non-homogeneous equation.

01-02: THE SECOND ORDER HOMOGENEOUS EQUATION


The second order homogeneous linear differential equation with constant coefficient is of
the form
𝐿(𝑦) = 𝑦′′ + 𝑎1 𝑦′ + 𝑎2 𝑦 = 0, (14)
where 𝑎1 , 𝑎2 are constants. Let us assume that 𝑦 = 𝑒 , for some value of 𝑟, be a solution
𝑟𝑥

of the equation (14). Then we find


𝐿(𝑒 𝑟𝑥 ) = (𝑒 𝑟𝑥 )′′ + 𝑎1 (𝑒 𝑟𝑥 )′ + 𝑎2 (𝑒 𝑟𝑥 ) = (𝑟 2 + 𝑎1 𝑟 + 𝑎2 )𝑒 𝑟𝑥 .
We denote this as
𝐿(𝑒 𝑟𝑥 ) = 𝑝(𝑟)𝑒 𝑟𝑥 , (15)
where 𝑝(𝑟) = 𝑟 + 𝑎1 𝑟 + 𝑎2
2
(16)
is the characteristic polynomial of the equation (14). We see that if 𝑟 satisfies 𝑝(𝑟) = 0,
then 𝐿(𝑒 𝑟𝑥 ) = 0 which shows that 𝑦 = 𝑒 𝑟𝑥 is a solution of 𝐿(𝑦) = 0.
Theorem 2: Let 𝑎1 , 𝑎2 be constants and consider the equation 𝐿(𝑦) = 𝑦′′ + 𝑎1 𝑦′ + 𝑎2 𝑦 =
0. If 𝑟1 and 𝑟2 are distinct roots of the characteristics polynomial 𝑝(𝑟), where 𝑝(𝑟) = 𝑟 2 +
𝑎1 𝑟 + 𝑎2, then the functions 𝜙1 = 𝑒 𝑟1𝑥 and 𝜙2 = 𝑒 𝑟2𝑥 are solutions of 𝐿(𝑦) = 0. If 𝑟1 is a
repeated root of 𝑝(𝑟), then 𝜙1 (𝑥) = 𝑒 𝑟1𝑥 and 𝜙2 (𝑥) = 𝑥𝑒 𝑟1𝑥 are solutions of 𝐿(𝑦) = 0.
Proof: We shall find 𝑟 such that 𝑒 𝑟𝑥 will be solution of the equation (14). Thus, we find
𝐿(𝑒 𝑟𝑥 ) = 𝑝(𝑟)𝑒 𝑟𝑥 , where 𝑝(𝑟) = 𝑟 2 + 𝑎1 𝑟 + 𝑎2 is the characteristic polynomial of degree 2.
Hence by the fundamental theorem of algebra 𝑝(𝑟) has two roots.
Let 𝑟1 and 𝑟2 be two roots of 𝑝(𝑟) = 0. This implies that
𝐿(𝑒 𝑟1𝑥 ) = 0 and 𝐿(𝑒 𝑟2𝑥 ) = 0.
This shows that the functions 𝜙1 , 𝜙2 defined by 𝜙1 (𝑥) = 𝑒 𝑟1𝑥 and 𝜙2 (𝑥) = 𝑒 𝑟2𝑥 are solutions
of the equation (14).
Case 1: If 𝑟1 and 𝑟2 two distinct roots of 𝑝(𝑟) = 0, then 𝑒 𝑟1𝑥 and 𝑒 𝑟2 𝑥 are two distinct
solutions of 𝐿(𝑦) = 0.

S25013: Ordinary Differential Equations Page 8


Case 2: Let 𝑟1 = 𝑟2 = 𝑟 . If the root of 𝑟 = 𝑟1 of the characteristic polynomial is of
multiplicity 2, then it is possible to find two distinct solutions of the equation 𝐿(𝑦) = 0.
We have
𝐿(𝑒 𝑟1𝑥 ) = 𝑝(𝑟1 ) 𝑒 𝑟1𝑥 , ∀ 𝑥. (17)
If 𝑟1 is a root of 𝑝(𝑟) , then 𝑝(𝑟1 ) = 0 . This gives 𝐿(𝑒 1 ) = 0 ⇒ 𝑒 1 is one solution of
𝑟 𝑥 𝑟 𝑥

𝐿(𝑦) = 0.
Now to find other solution of the 𝐿(𝑦) = 0, we differentiate 𝐿(𝑒 𝑟𝑥 ) = 𝑝(𝑟)𝑒 𝑟𝑥 with respect
to 𝑟, to get
𝐿(𝑥𝑒 𝑟𝑥 ) = [𝑝′(𝑟) + 𝑥𝑝(𝑟)]𝑒 𝑟𝑥 .
Since the operator 𝐿 involves only differentiation with respect to 𝑥, we simplify the left
hand side of the above equation as
𝜕
𝐿 [ (𝑒 𝑟𝑥 )] = [𝑝′(𝑟) + 𝑥 𝑝(𝑟)]𝑒 𝑟𝑥
𝜕𝑟
⇒ 𝐿(𝑥𝑒 𝑟𝑥 ) = [𝑟 ′ (𝑟) + 𝑥 𝑝(𝑟)]𝑒 𝑟𝑥 . (18)
If 𝑟1 is a repeated root of 𝑝(𝑟), i.e. 𝑝(𝑟1 ) = 𝑝′(𝑟1 ) = 0, so we have 𝐿(𝑥𝑒 ) = 0.
𝑟𝑥

This proves that the function 𝜙2 defined by 𝜙2 (𝑥) = 𝑥𝑒 𝑟1𝑥 is also a solution of the equation
(14). Thus if 𝑟 = 𝑟1 is a repeated root of the characteristic polynomial, then 𝜙1 (𝑥) =
𝑒 𝑟1𝑥 and 𝜙2 (𝑥) = 𝑥𝑒 𝑟1 𝑥 are two solutions of 𝐿(𝑦) = 0.
Now we will find, in the following theorem, all the solutions of the equation 𝐿(𝑦) = 0.
Theorem 3: If 𝜙1 and 𝜙2 are any two solutions of 𝐿(𝑦) = 0 and 𝑐1 and 𝑐2 are any two
constants, then 𝜙(𝑥) = 𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥) is also a solution of 𝐿(𝑦) = 0.
Proof: Given that 𝜙1 and 𝜙2 are two solutions of 𝐿(𝑦) = 0.
⇒ 𝐿(𝜙1 ) = 0 and 𝐿(𝜙2 ) = 0,
where 𝐿(𝜙1 ) = 𝜙1′′ + 𝑎1 𝜙′1 + 𝑎2 𝜙1 = 0 and 𝐿(𝜙2 ) = 𝜙2′′ + 𝑎1 𝜙′2 + 𝑎2 𝜙2 = 0. (19)
Define a function 𝜙(𝑥) = 𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥), where 𝑐1 and 𝑐2 are constants. Consider
𝐿(𝜙) = 𝐿(𝑐1 𝜙1 + 𝑐2 𝜙2 )
= (𝑐1 𝜙1 + 𝑐2 𝜙2 )′′ + 𝑎1 (𝑐1 𝜙1 + 𝑐2 𝜙2 )′ + 𝑎2 (𝑐1 𝜙1 + 𝑐 + 𝑐2 𝜙2 )
= 𝑐1 (𝜙′′1 + 𝑎1 𝜙′1 + 𝑎2 𝜙1 ) + 𝑐2 (𝜙2 ′′ + 𝑎1 𝜙2 ′ + 𝑎2 𝜙2 )
𝐿(𝜙) = 𝑐1 𝐿(𝜙1 ) + 𝑐2 𝐿(𝜙2 ) (20)
On using the equation (19), we have
𝐿(𝜙) = 0.
This prove that 𝜙 (𝑥) = 𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥) is a solution of the equation 𝐿(𝑦) = 0. Since 𝑐1
and 𝑐2 are arbitrary constants, for each value of 𝑐1 and 𝑐2 , we obtain different solutions of
𝐿(𝑦) = 0. In this way we find all the solutions of 𝐿(𝑦) = 0.
Note: We observe from the equation (20) that the operator 𝐿 is a linear operator.

01-03: INITIAL VALUE PROBLEMS FOR SECOND ORDER EQUATIONS


An initial value problem for 𝐿(𝑦) = 0 is a problem of finding a solution 𝜙 satisfying
𝜙(𝑥0 ) = 𝛼 and 𝜙 ′ (𝑥0 ) = 𝛽,
where 𝑥0 is some real number and 𝛼, 𝛽 are two constants. Thus, the IVP is denoted by
𝐿(𝑦) = 0, 𝑦(𝑥0 ) = 𝛼 and 𝑦′(𝑥0 ) = 𝛽.
S25013: Ordinary Differential Equations Page 9
Note: 1) An initial condition is a condition on the solution at one point.
2) A boundary condition is a condition on the solution at two or more points.
In the following, we prove the existence theorem which assets that every IVP has a solution
and its uniqueness.
Theorem 4: (Existence theorem) For any real 𝑥0 and constants 𝛼, 𝛽, there exists a solution
𝜙 of the initial value problem on −∞ < 𝑥 < ∞,
𝐿(𝑦) = 𝑦 ′′ + +𝑎1 𝑦 ′ + 𝑎2 𝑦 = 0, 𝑦(𝑥0 ) = 𝜙 , 𝑦′(𝑥0 ) = 𝛽.
Proof: Consider the IVP
𝐿(𝑦) = 0, 𝑦(𝑥0 ) = 𝛼, 𝑦′(𝑥0 ) = 𝛽, (21)
where 𝑥0 is a real number and 𝛼, 𝛽 are constants.
Let 𝑃(𝑟) = 𝑟 2 + 𝑎1 𝑟 + 𝑎2 be the characteristic polynomial of 𝐿 and 𝑟1 , 𝑟2 be the roots of
𝑝(𝑟).
Case 1: Let 𝑟1 ≠ 𝑟2 i. e. 𝑟1 and 𝑟2 be two distinct roots of 𝑝(𝑟). Then the functions 𝜙1 (𝑥) =
𝑒 𝑟1𝑥 , 𝜙2 (𝑥) = 𝑒 𝑟2𝑥 be two solutions of 𝐿(𝑦) = 0. Then we know
𝜙(𝑥) = 𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥)
is also a solution of 𝐿(𝑦) = 0.
Now to show there exists a solution 𝜙 of the IVP. We show that there exist unique constants
𝑐1 and 𝑐2 such that 𝜙 = 𝑐1 𝜙1 + 𝑐2 𝜙2 , satisfies 𝜙(𝑥0 ) = 𝛼, 𝜙’(𝑥0 ) = 𝛽 . This gives
𝑐1 𝜙1 (𝑥0 ) + 𝑐2 𝜙2 (𝑥0 ) = 𝛼,
𝑐1 𝜙′1 (𝑥0 ) + 𝑐2 𝜙′2 (𝑥0 ) = 𝛽. (22)
This is a non-homogenous system of simultaneous equations in 𝑐1 and 𝑐2 , where the
determinant of the coefficients is given by
𝜙 (𝑥 ) 𝜙2 (𝑥0 )
𝛥=| 1 0 |
𝜙1 ′(𝑥0 ) 𝜙2 ′(𝑥0 )
𝑒 𝑟1𝑥0 𝑒 𝑟2𝑥0
= | 𝑟1𝑥0 |
𝑟1 𝑒 𝑟2 𝑒 𝑟2𝑥0
𝛥 = (𝑟2 − 𝑟1 )𝑒 (𝑟1+𝑟2)𝑥0 . (23)
Since 𝑟1 ≠ 𝑟2 , we have 𝛥 ≠ 0. This implies that the system of equations (22) has unique
solution and is given by Cramer’s rule
𝛼 𝑒 𝑟2𝑥0 𝑒 𝑟1 𝑥0 𝛼
| | | 𝑟 1 𝑥0 |
𝛽 𝑟2 𝑒 𝑟2 𝑥0 𝑟1 𝑒 𝛽
𝑐1 = and 𝑐2 = . (24)
𝛥 𝛥
For this choice of 𝑐1 and 𝑐2 , we have 𝜙(𝑥) = 𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙(𝑥) is the solution of IVP.
Case 2: Let 𝑟1 = 𝑟2. Let r be the repeated root of the characteristic equation. In this case
the two functions 𝜙1 (𝑥) = 𝑒 𝑟𝑥 and 𝜙2 (𝑥) = 𝑥𝑒 𝑟𝑥 are solutions of 𝐿(𝑦) = 0. For this choice
of solutions, we solve the simultaneous equations (22) for 𝑐1 and 𝑐2 .
The determinant of the coefficients in this case is given by
𝑒 𝑟𝑥0 𝑥0 𝑒 𝑟𝑥0
𝛥 = | 𝑟𝑥0 |
𝑟𝑒 𝑟𝑥0 𝑒 𝑟𝑥0 + 𝑒 𝑟𝑥0
⇒ 𝛥 = 𝑒 2𝑟𝑥0 ≠ 0.

S25013: Ordinary Differential Equations Page 10


In this case, also the system of equations (22) has unique solutions 𝑐1 and 𝑐2 , where 𝑐1 and
𝑐2 are given by Cramer’s rule. For this choice of 𝑐1 and 𝑐2 , we have
𝜙(𝑥) = 𝑐1 𝜙(𝑥) + 𝑐2 𝜙2 (𝑥)
is a solution of the IVP.

Remark: It is not obvious that the solutions obtained in the above theorems turn out to be
only one. We prove in the following Theorem 6 that the solution is unique. But to prove
uniqueness of the solutions of IVP, we need the following Theorem 5, we first prove it.
However, we recall the following formulae which need in the course to derive the proof of
the following theorem.
Recall: If 𝑧1 , 𝑧2 are complex numbers, then
i) |𝑧| = |𝑧|,
ii) |𝑧|2 = 𝑧𝑧,
iii) |𝑧1 + 𝑧2 | ≤ |𝑧1 | + |𝑧2 |,
iv) |𝑧1 𝑧2 | = |𝑧1 |. |𝑧2 |.
If 𝑏 and 𝑐 are two constants, then we have
(|𝑏| − |𝑐|)2 ≥ 0
⇒ |𝑏 2 | + |𝑐|2 − 2|𝑏||𝑐| ≥ 0
v) ⇒ 2|𝑏||𝑐| ≤ |𝑏|2 + |𝑐|2. (25)
Size of 𝑳: Consider the equation
𝐿(𝑦) = 𝑦′′ + 𝑎1 𝑦′ + 𝑎2 𝑦 = 0
and 𝜙 is a solution of 𝐿(𝑦) = 0.
Then ‖𝜙(𝑥)‖ = [|𝜙(𝑥)|2 + |𝜙′(𝑥)|2 ]1/2 (26)
is called a measure the “size” of 𝜙 and 𝜙’, and the size of the operator 𝐿 will be measured
by
𝑘 = 1 + |𝑎1 | + |𝑎2 |.
Theorem 5: Let 𝜙(𝑥) be any solution of the equation 𝐿(𝑦) = 𝑦′′ + 𝑎1 𝑦′ + 𝑎2 𝑦 = 0 on an
interval 𝐼 containing a point 𝑥0 . Then
‖𝜙(𝑥0 )‖𝑒 −𝑘|𝑥−𝑥0| ≤ ‖𝜙(𝑥)‖ ≤ ‖𝜙(𝑥0 )‖𝑒 𝑘|𝑥−𝑥0| , ∀ 𝑥 ∈ 𝐼
where ‖𝜙(𝑥)‖ = [|𝜙(𝑥)|2 + |𝜙′(𝑥)|2 ]1/2 and 𝑘 = 1 + |𝑎1 | + |𝑎2 |.
Proof: Let 𝜙 be a solution of 𝐿(𝑦) = 0 on an interval 𝐼 containing a point 𝑥0 . Then we
have 𝐿(𝜙) = 0
⇒ 𝜙 ′′ (𝑥) + 𝑎1 𝜙 ′ (𝑥) + 𝑎2 𝜙(𝑥) = 0
⇒ 𝜙 ′′ (𝑥) = −[𝑎1 𝜙 ′ (𝑥) + 𝑎2 𝜙(𝑥)]
⇒ |𝜙′′(𝑥)| ≤ |𝑎1 ||𝜙 ′ (𝑥)| + |𝑎2 ||𝜙(𝑥)|. (27)
2
Set 𝑢(𝑥) = ‖𝜙(𝑥)‖2 = |𝜙(𝑥)|2 + |𝜙 ′ (𝑥)|
⇒ 𝑢(𝑥) = 𝜙𝜙 + 𝜙′𝜙′.
Differentiating this with respect to 𝑥, we get
′ ′ ′′
𝑢′ (𝑥) = 𝜙 ′ 𝜙 + 𝜙𝜙 + 𝜙 ′′ 𝜙 + 𝜙 ′ 𝜙 .
Using the result |𝜙(𝑥)| = |𝜙(𝑥)|,we find

S25013: Ordinary Differential Equations Page 11


|𝑢′ (𝑥)| ≤ 2|𝜙(𝑥)||𝜙′(𝑥)| + 2|𝜙 ′ (𝑥)||𝜙 ′′ (𝑥)|. (∵ |𝜙| = |𝜙| etc.)
Using the inequality (25), we have
2|𝜙(𝑥)||𝜙′(𝑥)| ≤ |𝜙(𝑥)|2 + |𝜙′(𝑥)|2.
Hence, we have
|𝑢′ (𝑥)| ≤ 2|𝜙(𝑥)||𝜙 ′ (𝑥)| + 2|𝜙 ′ (𝑥)|[|𝑎1 ||𝜙 ′ (𝑥)| + |𝑎2 ||𝜙(𝑥)|]
≤ (1 + |𝑎2 |)(|𝜙(𝑥)|2 + |𝜙 ′ (𝑥)|2 ) + 2|𝑎1 ||𝜙 ′ (𝑥)|2
2
≤ (1 + |𝑎2 |)|𝜙(𝑥)|2 + (1 + 2|𝑎1 | + |𝑎2 |)|𝜙 ′ (𝑥)| .
Replacing each coefficient on the right-hand side by 2(1 + |𝑎1 | + |𝑎2 |), we still have the
inequality
|𝑢′ (𝑥)| ≤ 2(1 + |𝑎1 | + |𝑎2 |)[|𝜙(𝑥)|2 + |𝜙 ′ (𝑥)|2 ]
⇒ |𝑢′(𝑥)| ≤ 2𝑘 𝑢(𝑥),
for 𝑘 = 1 + |𝑎1 | + |𝑎2 | ,
−2𝑘 𝑢(𝑥) ≤ 𝑢′ (𝑥) ≤ 2𝑘 𝑢(𝑥). (28)
Consider first the right inequality of the equation (28)
𝑢′ (𝑥) ≤ 2𝑘 𝑢(𝑥)
⇒ 𝑢′ (𝑥) − 2𝑘 𝑢(𝑥) ≤ 0.
As 𝑥 −2𝑘𝑥 > 0, we have
𝑒 −2𝑘𝑥 (𝑢′ (𝑥) − 2𝑘 𝑢(𝑥)) ≤ 0

⇒ (𝑒 −2𝑘𝑥 𝑢(𝑥)) ≤ 0.
Integrating this between the limits 𝑥0 to 𝑥 (𝑥 > 𝑥0 ), we get
[𝑒 −2𝑘𝑥 𝑢(𝑥)]𝑥𝑥0 ≤ 0
⇒ 𝑒 −2𝑘𝑥 𝑢(𝑥) − 𝑒 −2𝑘𝑥0 𝑢(𝑥0 ) ≤ 0
or 𝑢(𝑥) ≤ 𝑢(𝑥0 )𝑒 2𝑘(𝑥−𝑥0)
or ‖𝜙(𝑥)‖2 ≤ ‖𝜙(𝑥0 )‖2 𝑒 2𝑘(𝑥−𝑥0)
‖𝜙(𝑥)‖ ≤ ‖𝜙(𝑥0 )‖𝑒 𝑘(𝑥−𝑥0 ) . (29)
Similarly, by taking the left-hand inequality of the equation (28), we arrive at the result
‖𝜙(𝑥0 )‖𝑒 −𝑘(𝑥−𝑥0) ≤ ‖𝜙(𝑥)‖. (30)
From equation (29) and (30), we have
‖𝜙(𝑥0 )‖𝑒 −𝑘(𝑥−𝑥0) ≤ ‖𝜙(𝑥)‖ ≤ ‖𝜙(𝑥0 )‖𝑒 𝑘(𝑥−𝑥0) , ∀ 𝑥 > 𝑥0 . (31)
Similarly, for all 𝑥 < 𝑥0 , we obtain in the same way
‖𝜙(𝑥0 )‖𝑒 𝑘(𝑥−𝑥0) ≤ ‖𝜙(𝑥)‖ ≤ ‖𝜙(𝑥0 )‖𝑒 −𝑘(𝑥−𝑥0) , ∀ 𝑥 < 𝑥0 . (32)
The two inequalities together can be written as
‖𝜙(𝑥0 )‖𝑒 −𝑘|𝑥−𝑥0| ≤ ‖𝜙(𝑥)‖ ≤ ‖𝜙(𝑥0 )‖𝑒 𝑘|𝑥−𝑥0| , ∀ 𝑥 (33)
where 𝑘 = 1 + |𝑎1 | + |𝑎2 |.

Remark: Geometrically the inequality (33) shows that ‖𝜙(𝑥)‖ always lies between the two
curves given by
𝑦 = ‖𝜙(𝑥0 )‖𝑒 −𝑘|𝑥−𝑥0| and 𝑦 = ‖𝜙(𝑥0 )‖𝑒 𝑘|𝑥−𝑥0|
and is shown in the following figure.

S25013: Ordinary Differential Equations Page 12


𝑦 = ||𝜙(𝑥0 )||𝑒 𝑘|𝑥−𝑥0 |

||𝜙(𝑥0 )||

||𝜙(𝑥0 )||

𝑦 = ||𝜙(𝑥0 )||𝑒 −𝑘|𝑥−𝑥0 |

01-04: UNIQUENESS THEOREM


Theorem 6: - (Uniqueness theorem) Let 𝛼, 𝛽 be any two constants and let 𝑥0 be any real
number. Then on any interval 𝐼 containing 𝑥0 there exists at most one solution 𝜙 of the
IVP.
𝐿(𝑦) = 𝑦 ′′ + 𝑎1 𝑦 ′ + 𝑎2 𝑦 = 0, 𝑦(𝑥0 ) = 𝛼, 𝑦′(𝑥0 ) = 𝛽.
Proof: We have proved in the Theorem 4 that there exists a solution of the IVP
𝐿(𝑦) = 0, 𝑦(𝑥0 ) = 𝛼, 𝑦′(𝑥0 ) = 𝛽.
Let 𝜙(x) and 𝜓(𝑥) be two solutions of the IVP 𝐿(𝑦) = 0, 𝑦(𝑥0 ) = 𝛼, 𝑦′(𝑥0 ) = 𝛽. Define a
function 𝜒 = 𝜙 − 𝜓.
Therefore,
𝐿(𝜒) = 𝐿(𝜙 − 𝜓)
𝐿(𝜒) = 𝐿(𝜙) − 𝐿(𝜓).
As 𝜙 and 𝜓 are solutions, we have
𝐿(𝜙) = 0 and 𝐿(𝜓) = 0
⇒ 𝐿(𝜒) = 0.
Further, we see that
𝜒(𝑥0 ) = 𝜙(𝑥0 ) - 𝜓(𝑥0 )
=𝛼–𝛼
= 0. (34)
Also,
𝜒 ′ (𝑥0 ) = 𝜙 ′ (𝑥0 ) - 𝜓′ (𝑥0 )
=𝛽 –β
= 0. (35)
This shows that 𝜒(𝑥) is also a solution of the IVP. Hence, from the Theorem 5, we have
‖𝜒(𝑥0 )‖𝑒 −𝑘|𝑥−𝑥0| ≤ ‖𝜒(𝑥)‖ ≤ ‖𝜒(𝑥0 )‖𝑒 𝑘|𝑥−𝑥0| , ∀ 𝑥 ∈ 𝐼.
From above equations (34) and (35), we have
|𝜒(𝑥0 )| = 0 and |𝜒′(𝑥0 )| = 0
2
Hence, ‖𝜒(𝑥0 )‖2 = |𝜒(𝑥0 )|2 + |𝜒 ′ (𝑥0 )| = 0 ⇒ ‖𝜒(𝑥0 )‖ = 0.
Hence, the above inequality yields
0 ≤ ‖𝜒(𝑥)‖ ≤ 0, ∀ 𝑥 ∈ 𝐼

S25013: Ordinary Differential Equations Page 13


⇒ 𝜒(𝑥) = 0, ∀ 𝑥 ∈ 𝐼
⇒ 𝜙(𝑥) − 𝜓(𝑥) = 0, ∀ 𝑥 ∈ 𝐼.
This proves the uniqueness of the solution of the IVP.
Theorem 7: Let 𝜙1 , 𝜙2 be the two solutions of 𝐿(𝑦) = |𝑥 − 𝑥0 | = 0 given by
𝜙1 (𝑥) = 𝑒 𝑟1 𝑥 , 𝜙2 (𝑥) = 𝑒 𝑟2𝑥 , for 𝑟1 ≠ 𝑟2 ,
and 𝑟𝑥 𝑟𝑥
𝜙1 (𝑥) = 𝑒 , 𝜙2 (𝑥) = 𝑥𝑒 , for 𝑟1 = 𝑟2 = 𝑟,
where 𝑟1 and 𝑟2 are the roots the characteristic polynomial 𝑝(𝑟) = 𝑟 2 + 𝑎1 𝑟 + 𝑎2 . If 𝑐1 and
𝑐2 are any two constants, then the function 𝜙(𝑥) = 𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙(𝑥) is a solution of
𝐿(𝑦) = 0 on −∞ < 𝑥 < ∞.
Conversely, if 𝜙 is any solution of 𝐿(𝑦) = 0 on −∞ < 𝑥 < ∞ , then there are unique
constants 𝑐1 and 𝑐2 such that 𝜙(𝑥) = 𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥).
Proof: Let 𝜙1 and 𝜙2 be two solutions of 𝐿(𝑦) = 0, then the first part of the proof is proved
in the Theorem 3 that 𝜙(𝑥) = 𝑐1 𝜙1 (𝜙) + 𝑐2 𝜙2 (𝑥) is also a solution of the equation 𝐿(𝑦) =
0 on −∞ < 𝑥 < ∞.
Conversely, suppose 𝜙 is a solution of 𝐿(𝑦) = 0 on −∞ < 𝑥 < ∞. Let 𝑥0 ∈ (−∞, ∞) be a
real number. Let 𝜙(𝑥0 )and 𝜙′(𝑥0 ) be defined by
𝜙(𝑥0 ) = 𝛼 and 𝜙′(𝑥0 ) = 𝛽.
Consider the IVP 𝐿(𝑦) = 0, 𝑦(𝑥0 ) = 𝛼, 𝑦′(𝑥0 ) = 𝛽.
Therefore, by existence theorem every IVP has a solution. Let the solution be given by
𝜓(𝑥)
⇒ 𝐿(𝜓(𝑥)) = 0 and 𝜓(𝑥0 ) = 𝛼, 𝜓′(𝑥0 ) = 𝛽.
Then the solution 𝜓(𝑥) is of the form
𝜓(𝑥) = 𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥).
Therefore,
𝜓(𝑥0 ) = 𝛼 ⇒ 𝑐1 𝜙1 (𝑥0 ) + 𝑐2 𝜙2 (𝑥0 ) = 𝛼
and 𝜓′(𝑥0 ) = 𝛽 ⇒ 𝑐1 𝜙′1 (𝑥0 ) + 𝑐2 𝜙′2 (𝑥0 ) = 𝛽.
This system of equations has solutions given by
𝛼 𝜙2 (𝑥0 ) 𝜙1 (𝑥0 ) 𝛼
| | | |
𝛽 𝜙′2 (𝑥0 ) 𝜙′1 ′(𝑥0 ) 𝛽
𝑐1 = , 𝑐2 = (36)
𝛥 𝛥
where 𝛥 = (𝑟2 − 𝑟1 )𝑒 (𝑟1+𝑟2)𝑥0 ≠ 0, when 𝑟1 ≠ 𝑟2 and 𝛥 = 𝑒2𝑟𝑥0 ≠ 0 when 𝑟1 = 𝑟2 = 𝑟.
By uniqueness theorem, we have 𝜙 = 𝜓 for − ∞ < 𝑥 < ∞
⇒ 𝜙(𝑥) = 𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥),
where 𝑐1 and 𝑐2 are given in (36).

SOLVED PROBLEMS 01
Problem 01-01-01:
Show that every solution of constant coefficient equation 𝑦′′ + 𝑎1 𝑦′ + 𝑎2 𝑦 = 0 tends to zero
as 𝑥 → ∞, if and only if, the real parts of the roots of the characteristic polynomial are
negative.

S25013: Ordinary Differential Equations Page 14


Solution: DE is 𝐿(𝑦) = 𝑦′′ + 𝑎1 𝑦′ + 𝑎2 𝑦 = 0. Characteristic polynomial of 𝐿 is 𝑝(𝑟) = 𝑟 2 +
𝑎1 𝑟 + 𝑎2 . Let We know from the Theorem 3 that every solution 𝜙 of the constant
coefficient equation 𝐿(𝑦) = 0 is of the form
𝜙 = 𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥) (37)
where from the Theorem 2, the functions 𝜙1 and 𝜙2 , are
𝜙1 (𝑥)=𝑒 𝑟1𝑥 , 𝜙2 (𝑥) = 𝑒 𝑟2𝑥 , for 𝑟1 ≠ 𝑟2 ,
and 𝜙1 (𝑥)=𝑒 𝑟1𝑥 , 𝜙2 (𝑥) = 𝑥𝑒 𝑟1𝑥 , for 𝑟1 = 𝑟2 .
Let 𝑟1 ≠ 𝑟2, the equation (37) becomes
𝜙(𝑥) = 𝑐1 𝑒 𝑟1𝑥 + 𝑐2 𝑒 𝑟2(𝑥)
where 𝑟1 and 𝑟2 are the roots of characteristic polynomial 𝑝(𝑟) = 𝑟 2 + 𝑎1 𝑟 + 𝑎2 . Let 𝑟1 =
𝛼 + 𝑖𝛽 and 𝑟2 = 𝛼 − 𝑖𝛽, where 𝛼 and 𝛽 are real, be the complex roots of 𝑝(𝑟) = 0. Then we
have
𝜙 = 𝑒 ∝𝑥 (𝑐1 𝑒 𝑖𝛽𝑥 + 𝑐2 𝑒 −𝑖𝛽𝑥 ).
⇒ |𝜙(𝑥)| ≤ |𝑒 ∝𝑥 |(|𝑐1 | + |𝑐2 |), as |𝑒 ±𝑖𝛽𝑥 | = 1
lim |𝜙(𝑥)| ≤ lim |𝑒 𝛼𝑥 |(|𝑐1 | + |𝑐2 |) → 0, if 𝛼 < 0.
𝑥→∞ 𝑥→∞
⇒ |𝜙(𝑥)| → 0 as 𝑥 → ∞ if real parts of the roots of characteristic polynomial are negative.
As 𝜙(𝑥) ≤ |𝜙(𝑥)|, ∀ 𝑥 ∈ 𝐼 ⇒ 𝜙(𝑥) → 0 as 𝑥 → ∞ provided real parts of the roots of the
characteristic polynomial are negative.
Now consider the case that 𝑟1 = 𝑟2 be the repeated root of the characteristic polynomial
𝑝(𝑟), then from the equation (21) we have
𝜙(𝑥) = 𝑐𝑒 𝑟1𝑥 + 𝑐2 𝑥𝑒 𝑟1 𝑥 = 𝑒 𝑟1𝑥 (𝑐1 + 𝑐2 𝑥)
|𝜙(𝑥)| ≤ |𝑒 𝑟1𝑥 |(|𝑐1 | + |𝑐2 ||𝑥|)
⇒ |𝜙(𝑥)| ≤ |𝑒 𝛼𝑥 |(|𝑐1 | + |𝑐2 ||𝑥|)
lim |𝜙(𝑥)| → 0, if 𝛼 < 0.
𝑥→∞
Because the rate of decrease of 𝑒 −𝑥 (exponential function is much higher than the rate of
increase of a polynomial in 𝑥
𝑥𝑚
e.g. lim 𝑥 𝑚 𝑒 −𝑥 =lim .
𝑥→∞ 𝑥→∞ 𝑒 𝑥
By using the L ’Hospital rule repeatedly, we find
𝑚!
lim 𝑥 𝑚 𝑒 −𝑥 = lim
𝑥→∞ 𝑥→∞ 𝑒𝑥
𝑚 −𝑥
⇒ lim 𝑥 𝑒 = 0.
𝑥→∞
In this case also |𝜙(𝑥)| → 0, as 𝑥 → ∞, if ∝< 0 i.e. 𝜙(𝑥) → 0 as 𝑥 → ∞ if the real part of
the roots of the characteristic polynomial are negative.
Conversely, suppose every solution of the equation 𝑦′′ + 𝑎1 𝑦′ + 𝑎2 𝑦 = 0 tendts to zero as
𝑥 → ∞ . Let 𝑟1 , 𝑟2 be roots of the characteristic polynomial 𝑝(𝑟) = 𝑟 2 + 𝑎1 𝑟 + 𝑎2 . Then
𝜙1 (𝑥) = 𝑒 𝑟1𝑥 is a solution of the equation 𝑦′′ + 𝑎1 𝑦′ + 𝑎2 𝑦 = 0 and by assumption
𝜙1 (𝑥) → 0 as 𝑥 → ∞ . So, we have |𝜙1 (𝑥)| = |𝑒 𝑟1𝑥 | = 𝑒 Re(𝑟1)𝑥 → 0 𝑥 → ∞ ⇒ Re(𝑟1 ) < 0 .
Similarly, Re(𝑟2 ) < 0. Thus, the real parts of the roots of the characteristic polynomial are
negative.

S25013: Ordinary Differential Equations Page 15


Problem 01-01-02:
Consider the equation 𝑦′′ + 𝑎1 𝑦′ + 𝑎2 𝑦 = 0, where the constants 𝑎1 , 𝑎2 area real. Suppose
𝛼 + 𝑖𝛽 is a complex root of the characteristic polynomial, where 𝛼, 𝛽 are real, 𝛽 ≠ 0. Then
show that every solution of this equation tends to zero as 𝑥 → ∞ iff 𝑎1 > 0.
Solution: Let 𝑟1 = 𝛼 + 𝑖𝛽 and 𝑟2 = 𝛼 − 𝑖𝛽 be two roots of the characteristic polynomial
𝑝(𝑟) of the differential equation 𝑦′′ + 𝑎1 𝑦′ + 𝑎2 𝑦 = 0. Then we have
(𝛼 + 𝑖𝛽) + (𝛼 − 𝑖𝛽) = −𝑎1
⇒ 2𝛼 = −𝑎1
𝑎
⇒ 𝛼 = − 21 .
We see from this equation that 𝛼 is negative provided 𝑎1 > 0. From the Problem 01-01,
we have therefore
𝜙(𝑥) → 0 as 𝑥 → ∞ iff 𝑎1 > 0
i.e. every solution of such equation tends to zero as 𝑥 → ∞ iff 𝑎1 > 0.
Problem 01-01-03:
Consider the equation with constant coefficients 𝐿(𝑦) = 𝑦′′ + 𝑎1 𝑦′ + 𝑎2 𝑦 = 0. Let 𝜙1 be
the solution satisfying 𝜙1 (𝑥0 ) = 1, 𝜙1′ (𝑥0 ) = 0, and let 𝜙2 (𝑥) be the solution satisfying
𝜙2 (𝑥0 ) = 0, 𝜙2 ′(𝑥0 ) = 1. If 𝜙 is a solution satisfying 𝜙(𝑥0 ) = 𝛼, 𝜙′(𝑥0 ) = 𝛽 , then show
that 𝜙(𝑥) = 𝛼𝜙1 (𝑥) + 𝛽𝜙2 (𝑥), ∀ 𝑥.
Solution: - Let 𝜙1 (𝑥) and 𝜙2 (𝑥) be the solution of 𝐿(𝑦) = 0 satisfying the conditions
𝜙1 (𝑥0 ) = 1, 𝜙′1 (𝑥0 ) = 0 (38)
and
𝜙2 (𝑥0 ) = 0, 𝜙′2 (𝑥0 ) = 1. (39)
Let also 𝜙(𝑥) = 𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥) (40)
be a solution of 𝐿(𝑦) = 0 satisfying the conditions
𝜙(𝑥0 ) = 𝛼, 𝜙 ′ (𝑥0 ) = 𝛽 (41)
⇒ 𝑐1 𝜙1 (𝑥0 ) + 𝑐2 𝜙2 (𝑥0 ) = 𝛼,
𝑐1 𝜙1′ (𝑥0 ) + 𝑐2 𝜙2′ (𝑥) = 𝛽. (42)
These are the non-homogeneous simultaneous equations in 𝑐1 and 𝑐2 where the
determinant of the coefficients is given by
𝜙 (𝑥 ) 𝜙2 (𝑥0 )
𝛥 = | 1′ 0 |
𝜙1 (𝑥0 ) 𝜙2′ (𝑥0 )

1 0
𝛥=| | ⇒ 𝛥 = 1 ≠ 0.
0 1
Solving the set of equations (42) by Cramer’s rule, we have
∝ 0 1 ∝
| | | |
𝛽 1 0 𝛽
𝑐1 = and 𝑐2 =
𝛥 𝛥
⇒ 𝑐1 = 𝛼 and 𝑐2 = 𝛽.
Substituting this in the equation (42), we get 𝜙(x) = 𝛼𝜙1 (𝑥) + 𝛽𝜙2 (𝑥) as the solution of
𝐿(𝑦) = 0 satisfying (41).

S25013: Ordinary Differential Equations Page 16


Problem 01-01-04:
Find the function 𝜙 which has continuous derivative on 0 ≤ 𝑥 ≤ 2 which satisfies 𝜙(0) =
0, 𝜙 ′ (0) = 1 and 𝑦′′ − 𝑦 = 0 for 0 ≤ 𝑥 ≤ 1 and 𝑦′′ − 9𝑦 = 0 for 1 ≤ 𝑥 ≤ 2.
Solution: - Consider the differential equation
𝐿(𝑦) = 𝑦′′ − 𝑦 = 0, for 0 ≤ 𝑥 ≤ 1. (43)
The characteristic polynomial of (43) is 𝑟 − 1. It has roots 𝑟 = ±1. Hence the functions
2

𝜙1 (𝑥) = 𝑒 𝑥 and 𝜙2 (𝑥) = 𝑒 −𝑥 are solutions of 𝐿(𝑦) = 0 in 0 ≤ 𝑥 ≤ 1.


Hence the general solution of (43) is given by
𝜙(𝑥) = 𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥), ∀ 𝑥 ∈ [0, 1] (44)
satisfying the conditions
𝜙(0) = 0 and 𝜙 ′ (0) = 1
⇒ 𝑐1 𝜙1 (0) + 𝑐2 𝜙2 (0)=0,
and 𝑐1 𝜙′1 (0) + 𝑐2 𝜙′2 (0)=1,
where 𝜙1 (0) = 1, 𝜙2 (0) = 1, 𝜙1 (0) = 1, 𝜙2′ (0) = −1.

Thus, we have
𝑐1 + 𝑐2 = 0 and 𝑐1 − 𝑐2 = 1.
1 1
Solving these equations, we get 𝑐1 = and 𝑐2 = − . Hence the solution (44) becomes
2 2
1
𝜙(𝑥) = 2 (𝑒 𝑥 − 𝑒 −𝑥 ), in 0 ≤ 𝑥 ≤ 1. (45)
Now consider the equation
𝑦′′ − 9𝑦 = 0 for 1 ≤ 𝑥 ≤ 2 (46)
with initial conditions 𝜙(1) = 𝛼 and 𝜙′(1) = 𝛽 . The characteristic polynomial of the
equation is 𝑟 2 − 9. It has roots 𝑟 =  3 . Hence the new functions 𝜙1 (𝑥) = 𝑒 3𝑥 and 𝜙2 (𝑥) =
𝑒 −3𝑥 are solutions of (46) in 1 ≤ 𝑥 ≤ 2. The general solution of (46) is given by
𝜙(𝑥) = 𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥), in 1 ≤ 𝑥 ≤ 2 (47)
satisfying 𝜙(1) = 𝛼 and 𝜙′(1) = 𝛽, where from equation (47), we have
1 1 1 1
𝜙(1) = 2 (𝑒 − 𝑒) and 𝜙′(1) = 2 (𝑒 + 𝑒)
1 1 1 1
⇒ 𝛼 = 2 (𝑒 − 𝑒) and 𝛽 = 2 (𝑒 + 𝑒). (48)
Also, the new solution in 1 ≤ 𝑥 ≤ 2 is given by
𝜙1 (1) = 𝑒 3 , 𝜙2 (1) = 𝑒 −3
𝜙′1 (1) = 3𝑒 3 , 𝜙′2 (1) = −3𝑒 −3 .
Thus, from equation (47) and (48), we have
1 1
𝑐1 𝑒 3 + 𝑐2 𝑒 −3 = 2 (𝑒 − 𝑒)
1 1
𝑐1 𝑒 3 − 𝑐2 𝑒 −3 = 6 (𝑒 + 𝑒).
Solving these equations for 𝑐1 and 𝑐2 we obtain
1 1
𝑐1= 6 (2𝑒 −2 − 𝑒 −4 ) and 𝑐2 = 6 (𝑒 4 − 2𝑒 2 ).
Thus, the solution (47) becomes
1 1
𝜙(𝑥) = 6 (2𝑒 −2 − 𝑒 −4 )𝑒 3𝑥 + 6 (𝑒 4 − 2𝑒 2 )𝑒 −3𝑥 in 1 ≤ 𝑥 ≤ 2
1
⇒ 𝜙(𝑥) = 6 [2𝑒 3𝑥−2 − 𝑒 3𝑥−4 + 𝑒 −3𝑥+4 − 2𝑒 −3𝑥+2 ] for 1 ≤ 𝑥 ≤ 2. (49)

S25013: Ordinary Differential Equations Page 17


From equation (45) and (49), we see that the two solutions are identical at 𝑥 = 1, i. e. 𝜙(1)
1 1
from equation (45) and 𝜙(1) from equation (49) are identical and equal to (𝑒 − 𝑒). This
2
shows that 𝜙(x) is continuous on 0 ≤ 𝑥 ≤ 2. Now to derivative of 𝜙(𝑥) is continuous over
0 ≤ 𝑥 ≤ 2. We find from equation (45)
1
𝜙 ′ (𝑥) = (𝑒 𝑥 + 𝑒 −𝑥 )
2
′ (1) 1 1
⇒𝜙 = 2 (𝑒 + 𝑒).
Similarly, we find from the equation (49)
1
𝜙 ′ (𝑥) = 6 [6𝑒 3𝑥−2 − 3𝑒 3𝑥−4 − 3𝑒 −3𝑥+4 + 𝑒 −3𝑥+2 ]
1 1
⇒ 𝜙 ′ (1) = 2 (𝑒 + 𝑒).
⇒ 𝜙 ′ (𝑥) is continuous over 0  x  2 .

SELF-TEST 01
MCQ 01- 01-01
All solutions of the equation 𝑦 ′′ + 16𝑦 = 0 are given by
A. 𝜙(𝑥) = 𝑐1 cos 4𝑥 + 𝑐2 sin 4𝑥, where 𝑐1 , 𝑐2 are arbitrary constants
B. 𝜙(𝑥) = 𝑐1 𝑒 4𝑥 + 𝑐2 𝑒 −4𝑥 , where 𝑐1 , 𝑐2 are arbitrary constants
C. 𝜙(𝑥) = 𝑐1 cos 16𝑥 + 𝑐2 sin 16𝑥, where 𝑐1 , 𝑐2 are arbitrary constants
D. 𝜙(𝑥) = 𝑐1 𝑒 16𝑥 + 𝑐2 𝑒 −16𝑥 , where 𝑐1 , 𝑐2 are arbitrary constants.
MCQ 01-01-02
Consider the initial value problem 𝑦 ′′ + 𝑦 ′ − 6𝑦 = 0, 𝑦(0) = 1, 𝑦 ′ (0) = 0. Then the value
of 𝑦(1) is
2 3
A. 𝑦(1) = 5 𝑒 2 + 5 𝑒 −3
2 3
B. 𝑦(1) = 5 𝑒 2 − 5 𝑒 −3
3 2
C. 𝑦(1) = 5 𝑒 2 + 5 𝑒 −3
5 5
D. 𝑦(1) = 𝑒 2 + 𝑒 −3 .
2 3
MCQ 01-01-03
1
Consider the equation 𝐿𝑦 ′′ + 𝑅𝑦 ′ + 𝐶 𝑦 = 0 , where 𝑅, 𝐿 and 𝐶 are positive constants. If
𝑅2 4
− 𝐿𝐶 = 0, then for any constants 𝑐1 , 𝑐2 , all solutions of the given equation are given by
𝐿2
𝑅
A. 𝑦(𝑥) = (𝑐1 + 𝑐2 𝑥)𝑒 2𝐿𝑥
𝑅
B. 𝑦(𝑥) = (𝑐1 + 𝑐2 𝑥)𝑒 −2𝐿𝑥
𝑅
C. 𝑦(𝑥) = (𝑐1 + 𝑐2 𝑥)𝑒 − 𝐿 𝑥
𝑅
D. 𝑦(𝑥) = (𝑐1 + 𝑐2 𝑥)𝑒 𝐿 𝑥 .
MCQ 01-01-04
Which one of the following is the unique solution of IVP 𝑦 ′′ − 2𝑦 ′ − 3𝑦 = 0, 𝑦(0) = 0,
𝑦 ′ (0) = 1?

S25013: Ordinary Differential Equations Page 18


1 1
A. 𝑦(𝑥) = 4 𝑒 3𝑥 + 4 𝑒 −𝑥
1 1
B. 𝑦(𝑥) = 4 𝑒 −3𝑥 − 4 𝑒 −𝑥
1 1
C. 𝑦(𝑥) = 4 𝑒 −3𝑥 + 4 𝑒 −𝑥
1 1
D. 𝑦(𝑥) = 4 𝑒 3𝑥 − 4 𝑒 −𝑥 .

SHORT ANSWER QUESTIONS 01


SAQ 01- 01-01
Let 𝜙 be a solution of the equation 𝑦 ′′ + 𝑎1 𝑦 ′ + 𝑎2 𝑦 = 0 , where 𝑎1 , 𝑎2 are constants. If
𝑎1
𝜓(𝑥) = 𝑒 2 𝑥 𝜙(𝑥), then show that 𝜓(𝑥) satisfies an equation 𝑦 ′′ + 𝑘𝑦 = 0, where 𝑘 is some
constant.
Solution: Differentiating the function 𝜓 twice, we have
𝑎1 ′′
𝜓′′ = [𝑒 2 𝑥 𝜙(𝑥)]
𝑎1 𝑎1 ′
𝑎1
= [𝑒 2 𝑥 𝜙 ′ (𝑥) + 𝑒 2 𝑥 𝜙(𝑥)]
2
𝑎1
𝑎1 𝑎1
𝑎1 𝑎1
𝑎 2 𝑎1
𝑥
=𝑒 2 𝜙 ′′ (𝑥) + 𝑒 2 𝑥 𝜙 ′ (𝑥) + 𝑒 2 𝑥 𝜙 ′ (𝑥) + ( 21 ) 𝑒 2 𝑥 𝜙(𝑥)
2 2
𝑎1
𝑥 𝑎1 2
=𝑒 2 [𝜙 ′′ (𝑥) + 𝑎1 𝜙 ′ (𝑥) + ( 2 ) 𝜙(𝑥)]
𝑎 2 𝑎1
= [−𝑎2 + ( 21 ) ] 𝑒 2 𝑥 𝜙(𝑥) ( ∵ 𝜙 implies 𝜙 ′′ + 𝑎1 𝜙 ′ = −𝑎2 𝜙 )
𝑎12
= [−𝑎2 + ] 𝜓(𝑥)
4
𝑎12
⟹ 𝜓 ′′ + (𝑎2 − ) 𝜓(𝑥) = 0
4
𝑎12
⟹ 𝜓 ′′ + 𝑘𝜓(𝑥) = 0, where 𝑘 = (𝑎2 − ) is some constant, which proves the required.
4
SAQ 01-01-02
Consider the equation 𝑦 ′′ + 𝑘 2 𝑦 = 0, where 𝑘 is a non-negative constant. For what values
of 𝑘 will there exist non-trivial solutions 𝜙 satisfying 𝜙(0) = 0, 𝜙(𝜋) = 0?
Solution: Given that 𝑘 is a non-negative constant, i. e. 𝑘 ≥ 0. We discuss following two
cases:
Case-I: 𝑘 = 0. The given equation takes the form 𝑦 ′′ = 0 and hence its general solution is
given by 𝜙(𝑥) = 𝑐1 + 𝑐2 𝑥, where 𝑐1 , 𝑐2 are any constants. Using given conditions 𝜙(0) =
0, 𝜙(𝜋) = 0, it is seen that 𝜙(𝑥) = 0 for all 𝑥, which is trivial solution. Therefore, 𝑘 = 0
is not the required value.
Case-II: 𝑘 > 0. In this case, roots of the characteristic polynomial are complex conjugate
and hence the general solution is given by
𝜙(𝑥) = 𝑐1 cos 𝑘𝑥 + 𝑐2 sin 𝑘𝑥, (50)
where 𝑐1 , 𝑐2 are any constants. The first condition 𝜙(0) = 0 implies 𝑐1 = 0 and therefore
from (50), we have
𝜙(𝑥) = 𝑐2 sin 𝑘𝑥. (51)

S25013: Ordinary Differential Equations Page 19


Now, by using the second condition 𝜙(𝜋) = 0, it is observed that
𝑐2 sin 𝑘𝜋 = 0. (52)
If 𝑐2 = 0, then (50) gives 𝜙(𝑥) = 0, ∀ 𝑥. We, therefore, assume that for non-zero (non-
trivial) solution of the given equation, 𝑐2 ≠ 0. Then (52) gives
𝑘𝜋 = 𝑛𝜋, for 𝑛 = 1, 2, 3, ⋯
⟹ 𝑘 = 𝑛, for 𝑛 = 1, 2, 3, ⋯, i. e. 𝑘 = 1, 2, 3, ⋯.
Therefore, for 𝑘 = 1, 2, 3, ⋯ and 𝑐2 ≠ 0, the given problem has non-trivial solutions.
SAQ 01-01-03
Solve the IVP 2𝑦 ′′ = 𝑦 − 𝑦 ′ , 𝑦(0) = 1, 𝑦 ′ (0) = 2.
Solution: First, we rewrite the given differential equation in standard form:
2𝑦 ′′ + 𝑦 ′ − 𝑦 = 0. (53)
The corresponding auxiliary polynomial equation is 2𝑟 + 𝑟 − 1 = 0, This equation has
2
1
real and distinct roots, namely 𝑟 = 2 , −1. Therefore, the general solution of (53) is
1
𝑦(𝑥) = 𝑐1 𝑒 2𝑥 + 𝑐2 𝑒 −𝑥 , (54)
where 𝑐1 , 𝑐2 are constants. All that remains is to use the two given initial conditions to
determine the values of the constants 𝑐1 , 𝑐2 :
1
𝑦(0) = 1 ⟹ [𝑐1 𝑒 2𝑥 + 𝑐2 𝑒 −𝑥 ] = 1 ⟹ 𝑐1 + 𝑐2 = 1 (55)
𝑥=0
1
1 1
𝑦 ′ (0) = 2 ⟹ [ 𝑐1 𝑒 2𝑥 − 𝑐2 𝑒 −𝑥 ] = 2 ⟹ 𝑐1 − 𝑐2 = 2. (56)
2 𝑥=0 2
These two equations for 𝑐1 , 𝑐2 can be solved and we obtain
𝑐1 = 2, 𝑐2 = −1.
Substituting the values of 𝑐1 , 𝑐2 in the equation (54), the solution of the given IVP is,
1
therefore, 𝑦(𝑥) = 2𝑒 2𝑥 − 𝑒 −𝑥 .
SAQ 01-01-04
Show that 𝑒 −5𝑥 and 𝑒 2𝑥 are solutions to the second order linear differential equation
𝑦 ′′ + 3𝑦 ′ − 10𝑦 = 0.
Solution: We first check for the function 𝑒 −5𝑥 :
LHS= 𝑦 ′′ + 3𝑦 ′ − 10𝑦
= [𝑒 −5𝑥 ]′′ + 3[𝑒 −5𝑥 ]′ − 10𝑒 −5𝑥
= 25𝑒 −5𝑥 -15𝑒 −5𝑥 -10𝑒 −5𝑥
= 25𝑒 −5𝑥 -25𝑒 −5𝑥
= 0 =RHS.
This shows that the function 𝑒 −5𝑥 is a solution of the given equation. Similarly, one can
show that the second function 𝑒 2𝑥 is also a solution of the given equation.

S25013: Ordinary Differential Equations Page 20


SUMMARY
The concept of linear differential equation with constant coefficients is introduced. The
method to obtain the solution of second order linear homogeneous differential equation is
discussed and solved both initial and boundary value problems.
For constants 𝑎1 , 𝑎2 and for roots 𝑟1 , 𝑟2 of 𝑝(𝑟) = 𝑟 2 + 𝑎1 𝑟 + 𝑎2 , the linear homogeneous
differential equation 𝑦 ′′ + 𝑎1 𝑦 ′ + 𝑎2 𝑦 = 0 has the general solution of the form
𝜙(𝑥) = 𝑐1 𝑒 𝑟1𝑥 + 𝑐2 𝑒 𝑟2𝑥 , if 𝑟1 ≠ 𝑟2
or 𝜙(𝑥) = (𝑐1 + 𝑐2 𝑥)𝑒 𝑟1𝑥 , if 𝑟1 = 𝑟2,
where 𝑐1 , 𝑐2 are arbitrary constants.

END OF UNIT EXERCISES


1) Show that every solution of the constant coefficient equation 𝑦′′ + 𝑎1 𝑦′ + 𝑎2 𝑦 = 0 is
bounded on 0 ≤ 𝑥 < ∞ if and only if the real parts of the roots of the characteristic
polynomial are non-positive and the roots with zero real part have multiplicity one.
2) Consider the differential equation 𝑦 ′′ − (2𝛼 − 1)𝑦 ′ + 𝛼(𝛼 − 1)𝑦 = 0. Determine the
values of 𝛼 for which all solutions tend to zero and for which all solutions become
unbounded as 𝑥 → ∞.
3) Suppose 𝜙 and 𝜓 are two solutions of the constant coefficient equation
𝐿(𝑦) = 𝑦′′ + 𝑎1 𝑦′ + 𝑎2 𝑦 = 0
on a finite integral 𝐼 including a point 𝑥0 . Let
𝜙(𝑥0 ) = 𝛼1 , 𝜙′(𝑥0 ) = 𝛽1 ,
𝜓(𝑥0 ) = 𝛼2 , 𝜓′ (𝑥0 ) = 𝛽2 ,
and suppose (𝛼1 − 𝛼2 )2 + (𝛽1 − 𝛽2 )2 = 𝜀 2 .
(a) If 𝜒 = 𝜙 − 𝜓 show that 𝜒 satisfies 𝐿(𝑦) = 0 , and 𝜒(𝑥0 ) = 𝛼1 − 𝛼2 , 𝜒′(𝑥0 ) = 𝛽1 −
𝛽2.
(b) Show that |𝜙(𝑥) − 𝜓(𝑥)| ≤ 𝜀𝑒 𝑘|𝐼| for all 𝑥 in 𝐼 where 𝑘 = 1 + |𝑎1 | + |𝑎2 |, and |𝐼| is
the length of 𝐼.

KEY WORDS
Polynomial, characteristic polynomial, root (zero), ordinary differential equation, solution
of the differential equation, size of 𝜙 and 𝜙 ′ , length of operator, existence theorem and
uniqueness theorem.

S25013: Ordinary Differential Equations Page 21


UNIT 01-02: DEPENDENT AND INDEPENDENT OF SOLUTIONS
LEARNING OBJECTIVES
After successful completion of this unit, you will be able to
❖ explain linearly dependent / independent, Wronskian of functions and second order
linear nonhomogeneous differential equation with constant coefficients
❖ apply to find particular solution of a linear nonhomogeneous differential equation
with constant coefficients using linearly independent solutions of the associated
homogeneous equation

02-01: LINEARLY DEPENDENT AND INDEPENDENT


Definition: Two functions 𝜙1 (𝑥) and 𝜙2 (𝑥) are said to be independent of 𝐼 if the only
constants 𝑐1 and 𝑐2 such that 𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥) = 0, ∀ 𝑥 ∈ 𝐼 , then 𝑐1 = 𝑐2 =. Otherwise,
the functions are said to be linearly dependent.
Example 01: Consider a differential equation with constant coefficients 𝐿(𝑦) = 𝑦′′ +
𝑎1 𝑦′ + 𝑎2 𝑦 = 0. Show that there exist two solutions of 𝐿(𝑦) = 0 are linearly independent
on any interval 𝐼.
Solution: - Let 𝑟1 and 𝑟2 be the roots of the characteristic polynomial 𝑃(𝑟) = 𝑟 2 + 𝑎1 𝑟 +
𝑎2 . Then we know the functions
𝜙1 (𝑥) = 𝑒 𝑟1𝑥 , 𝜙2 (𝑥) = 𝑒 𝑟2𝑥 , 𝑟1 ≠ 𝑟2
and
𝜙1 (𝑥) = 𝑒 𝑟𝑥 , 𝜙2 (𝑥) = 𝑥𝑒 𝑟𝑥 , 𝑟1 = 𝑟2 = 𝑟
be the solutions of 𝐿(𝑦) = 0.
Now consider the sum for 𝑟1 ≠ 𝑟2
𝑐1 𝑒 𝑟1𝑥 + 𝑐2 𝑒 𝑟2𝑥 =0, ∀ 𝑥 ∈ 𝐼
⇒ 𝑐1 + 𝑐𝑒 (𝑟2−𝑟1)𝑥 =0. (1)
Differentiating this equation with respect to 𝑥, we get
𝑐2 (𝑟2 − 𝑟1 )𝑒 (𝑟2−𝑟1)𝑥 = 0, ∀ 𝑥 ∈ 𝐼 .
Since 𝑟2 ≠ 𝑟1 ⇒ 𝑐2 =0, consequently, from the equation (1), we have 𝑐1 = 0. Hence,
𝜙(𝑥) = 𝑒 𝑟1𝑥 and 𝜙2 (𝑥) = 𝑒 𝑟2𝑥 are linearly independent.
Now consider the case when 𝑟 is a repeated root of 𝑝(𝑟) = 0. Consider the sum
𝑐1 𝑒 𝑟𝑥 + 𝑐2 𝑥𝑒 𝑟𝑥 = 0, ∀ 𝑥 ∈ 𝐼
⇒ 𝑐1 + 𝑐2 𝑥 = 0. (2)
Differentiating with respect to 𝑥 , we get 𝑐2 = 0 . Consequently, 𝑐1 = 0 . Then the sum
𝑐1 𝑒 𝑟𝑥 + 𝑐2 𝑥𝑒 𝑟𝑥 = 0 ⇒ 𝑐1 = 𝑐2 = 0 . Hence the solutions 𝜙1 (𝑥) = 𝑒 𝑟𝑥 , 𝜙2 (𝑥) = 𝑥𝑒 𝑟𝑥 are
linearly independent.
Example 02: Prove the following sets of functions defined on −∞ < 𝑥 < ∞ are linearly
independent or linearly dependent and why?
i) 𝜙1 (𝑥) = 𝑥, 𝜙2 (𝑥) = |𝑥|
ii) 𝜙1 (𝑥) = 𝑒 𝑖𝑥 , 𝜙2 (𝑥) = sin 𝑥 , 𝜙3 (𝑥) = 2 cos 𝑥
Solution: i) Let 𝜙1 (𝑥) = 𝑥 and 𝜙2 (𝑥) = |𝑥|. Consider the sum
𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥) = 0, ∀ 𝑥 in − ∞ < 𝑥 < ∞

S25013: Ordinary Differential Equations Page 22


i.e. 𝑐1 𝑥 + 𝑐2 |𝑥| = 0, for 𝑥 in −∞ < 𝑥 < ∞
⇒ 𝑐1 𝑥 + 𝑐2 𝑥 = 0, for 𝑥 in 0 ≤ 𝑥 < ∞
⇒ 𝑐1 + 𝑐2 =0,
and
𝑐1 𝑥 − 𝑐2 𝑥 = 0, for 𝑥 in −∞ < 𝑥 ≤ 0
⇒ 𝑐1 − 𝑐2 = 0.
Solving these equations, we get 𝑐1 = 𝑐2 = 0 ⇒ 𝜙1 (𝑥) = 𝑥 and 𝜙2 (𝑥) = |𝑥| are linearly
independent in − ∞ < 𝑥 < ∞.
ii) For given functions 𝜙1 (𝑥) = 𝑒 𝑖𝑥 , 𝜙2 (𝑥) = sin 𝑥 and 𝜙3 (𝑥) = 2 cos 𝑥 , we consider the
sum
𝑐1 𝑒 𝑖𝑥 + 𝑐2 sin 𝑥 + 2𝑐3 cos 𝑥 = 0, ∀ 𝑥 in − ∞ < 𝑥 < ∞. (3)
1
This is true for nonzero constants 𝑐1 = −1, 𝑐2 = 𝑖, 𝑐3 = 2 and hence the functions
𝜙1 (𝑥), 𝜙2 (𝑥), 𝜙3 (𝑥) are linearly dependent.
Wronskian: We will now introduce a simple test which enables us to tell whether the
solution of the equation 𝐿(𝑦) = 0 on an interval 𝐼 are linearly independent or not. It
involves the concept of determinant of the solutions of 𝐿(𝑦) = 0 and their derivatives.
This determinant is called the Wronskian of the solutions.
Definition: Let 𝜙1 (𝑥) and 𝜙2 (𝑥) be two solutions of the equation 𝐿(𝑦) = 0 on an interval
𝜙 (𝑥) 𝜙2 (𝑥)
𝐼. Then the determinant | 1′ | is called the Wronskian of 𝜙1 and 𝜙2 , is denoted
𝜙1 (𝑥) 𝜙2′ (𝑥)
by 𝑊(𝜙1 , 𝜙2 )(𝑥). It values at 𝑥 is given by
𝜙 (𝑥) 𝜙2 (𝑥)
𝑊(𝑥) = 𝑊(𝜙1 , 𝜙2 )(𝑥) = | 1′ | = 𝜙1 (𝑥)𝜙2′ (𝑥) − 𝜙1′ (𝑥)𝜙2 (𝑥) (4)
𝜙1 (𝑥) 𝜙2′ (𝑥)
Similarly, the Wronskian of the solutions 𝜙1 , 𝜙2 , 𝜙3 of the equation 𝐿(𝑦) = 0 on interval 𝐼
is defined as
𝜙1 (𝑥) 𝜙2 (𝑥) 𝜙3 (𝑥)
𝑊(𝑥) = 𝑊(𝜙1 , 𝜙2 , 𝜙3 )(𝑥) = | 𝜙1′ (𝑥) 𝜙2′ (𝑥) 𝜙3′ (𝑥) |. (5)
𝜙1′′ (𝑥) 𝜙2′′ (𝑥) 𝜙3′′ (𝑥)
Theorem 1: Two solutions 𝜙1 (𝑥) and 𝜙2 (𝑥) of the equation 𝐿(𝑦) = 0 are linearly
independent an on interval 𝐼, if and only if, 𝑊(𝜙1 , 𝜙2 )(𝑥) ≠ 0, ∀ 𝑥 ∈ 𝐼.
Proof: Let 𝜙1 (𝑥) and 𝜙2 (𝑥) be two solutions of the equation 𝐿(𝑦) = 0 and
𝑊(𝜙1 , 𝜙2 )(𝑥) ≠ 0, ∀ 𝑥 ∈ 𝐼. (6)
We prove that the solutions 𝜙1 (𝑥) and 𝜙2 (𝑥) are linearly independent in 𝐼. Let 𝑐1 and 𝑐2 be
two constants such that the sum
𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥) = 0, ∀ 𝑥 ∈ 𝐼. (7)
Differentiating (7) with respect to 𝑥, we get
𝑐1 𝜙′1 (𝑥) + 𝑐2 𝜙′2 (𝑥) = 0, ∀ 𝑥 ∈ 𝐼. (8)
If 𝑥0 ∈ 𝐼 is any fixed point, then we have
𝑐1 𝜙1 (𝑥0 ) + 𝑐2 𝜙2 (𝑥0 ) = 0
and
𝑐1 𝜙1′ (𝑥0 ) + 𝑐2 𝜙2′ (𝑥0 ) = 0.

S25013: Ordinary Differential Equations Page 23


These are linear homogeneous equations satisfied by 𝑐1 and 𝑐2 . We see that the determinant
of the coefficients of the system is just the Wronskian of the solutions at 𝑥0 in 𝐼 and is the
given by
𝜙 (𝑥 ) 𝜙2 (𝑥0 )
𝑊(𝜙1 , 𝜙2 )(𝑥0 ) = | 1′ 0 |≠0 (by (6))
𝜙1 (𝑥0 ) 𝜙2′ (𝑥0 )
Hence the system of equations has only trivial solutions 𝑐1 = 𝑐2 = 0. This proves that the
two solutions are linearly independent.
Conversely, let us assume that 𝜙1 (𝑥) and 𝜙2 (𝑥) are linearly independent solutions of
𝐿(𝑦) = 0 on 𝐼. We prove that 𝑊(𝜙1 , 𝜙2 )(𝑥) ≠ 0, ∀ 𝑥 ∈ 𝐼
Let us assume that there exists a point 𝑥0 in 𝐼 such that 𝑊(𝜙1 , 𝜙2 )(𝑥0 ) = 0
𝜙 (𝑥 ) 𝜙2 (𝑥0 )
⇒ | 1′ 0 |=0
𝜙1 (𝑥0 ) 𝜙2′ (𝑥0 )
This implies that the system of equations
𝑐1 𝜙1 (𝑥0 ) + 𝑐2 𝜙2 (𝑥0 ) = 0
and
𝑐1 𝜙1′ (𝑥0 ) + 𝑐2 𝜙2′ (𝑥0 ) = 0
is consistent and has solutions 𝑐1 and 𝑐2 not both zero simultaneously. For this choice of
𝑐1 and 𝑐2 consider the function.
𝜓(𝑥) = 𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥) (9)
⇒ 𝐿(𝜓(𝑥)) = 𝑐1 𝐿(𝜙1 ) + 𝑐2 𝐿(𝜙2 ) = 0
⇒ 𝐿(𝜓) = 0.
Also, 𝜓(𝑥0 ) = 𝑐1 𝜙1 (𝑥0 ) + 𝑐2 𝜙2 (𝑥0 ) = 0 and 𝜓′(𝑥0 ) = 𝑐1 𝜙′1 (𝑥0 ) + 𝑐2 𝜙2′ (𝑥0 ) = 0 ⇒ 𝜓(𝑥) is
a solution of the IVP 𝐿(𝑦) = 0, 𝑦(𝑥0 ) = 0, 𝑦′(𝑥0 ) = 0. Hence, ‖𝜓(𝑥0 )‖ = [|𝜓(𝑥0 )|2 +
1
|𝜓′ (𝑥0 )|2 ]2 = 0, and we have from the Theorem 4 of Unit 01-01.
‖𝜓(𝑥0 )‖𝑒 −𝑘|𝑥−𝑥0| ≤ ‖𝜓(𝑥)‖ ≤ ‖𝜓(𝑥0 )‖𝑒 𝑘|𝑥−𝑥0| , ∀ 𝑥 ∈ 𝐼
⟹ 0 ≤ ‖𝜓(𝑥)‖ ≤ 0, ∀ 𝑥 ∈ 𝐼
⟹ 𝜓(𝑥) = 0, ∀ 𝑥 ∈ 𝐼.
Hence from the equation (9), we have 𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥) = 0, ∀ 𝑥 ∈ 𝐼 and 𝑐1 , 𝑐2 are not
both zero. This implies 𝜙1 (𝑥) and 𝜙2 (𝑥) are linearly dependent on 𝐼. This contradicts the
fact that 𝜙1 (𝑥) and 𝜙2 (𝑥) and linearly independent on 𝐼. This shows that the assumption
that 𝑊(𝜙1 , 𝜙2 )(𝑥0 ) = 0 for some 𝑥0 ∈ 𝐼 is false, consequently, we have 𝑊(𝜙1 , 𝜙2 )(𝑥) ≠ 0,
∀ 𝑥 ∈ 𝐼.
Remarks:
1) In order to prove the first part of the Theorem 1, we need not have the functions
𝜙1, (𝑥), 𝜙2 (𝑥) to be the solutions of the equation 𝐿(𝑦) = 0. This part holds good for
any differentiable functions 𝜙1 (𝑥), 𝜙2 (𝑥) on 𝐼. Thus if 𝜙1 , 𝜙2 are any functions defined
on 𝐼 then 𝑊(𝜙1 , 𝜙2 )(𝑥) ≠ 0, ∀ 𝑥 ∈ 𝐼 ⇒ 𝜙1 , 𝜙2 are linearly independent.
2) However, to prove the converse part, the functions 𝜙1 , 𝜙2 must be the solutions of the
equation 𝐿(𝑦) = 0.
3) The converse part is not true for any differentiable functions defined on 𝐼 i. e. we have,
𝜙1 , 𝜙2 are linearly independent functions on 𝐼 ⇏ 𝑊(𝜙1 , 𝜙2 )(𝑥) ≠ 0, ∀ 𝑥 ∈ 𝐼. We will
illustrate this in the sequel by giving an example.
S25013: Ordinary Differential Equations Page 24
4) We will see in the following theorem that we need to compute 𝑊(𝜙1 , 𝜙2 ) at only one
convenient point to test the linearly independent of the s olutions.
Theorem 2: Let 𝜙1 , 𝜙2 be two solutions of 𝐿(𝑦) = 0 on an interval 𝐼 and let 𝑥0 be any
point in 𝐼. Then 𝜙1 , 𝜙2 are linearly independent on 𝐼 if and only if 𝑊(𝜙1 , 𝜙2 )(𝑥0 ) ≠ 0.
Proof: Let 𝜙1 (𝑥), 𝜙2 (𝑥) be linearly independent solutions of 𝐿(𝑦) = 0 in 𝐼. Then by the
Theorem 1 we have 𝑊(𝜙1 , 𝜙2 )(𝑥) ≠ 0, ∀ 𝑥 ∈ 𝐼. In particular 𝑊(𝜙1 , 𝜙2 )(𝑥0 ) ≠ 0, for some
𝑥0 ∈ 𝐼.
Conversely, suppose 𝑊(𝜙1 , 𝜙2 )(𝑥0 ) ≠ 0, for 𝑥0 ∈ 𝐼 and let 𝑐1 , 𝑐2 be two constants such
that
𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥)=0, ∀ 𝑥 ∈ 𝐼
⟹ 𝑐1 𝜙1 (𝑥0 ) + 𝑐2 𝜙2 (𝑥0 )=0 for 𝑥0 ∈ 𝐼 (10)
and
𝑐1 𝜙1′ (𝑥0 ) + 𝑐2 𝜙′2 (𝑥0 ) = 0 for 𝑥0 ∈ 𝐼. (11)
The determinant of the coefficients of a system of linear homogeneous equations is just
the Wronskian of 𝜙1 , 𝜙2 at 𝑥0 ∈ 𝐼. i.e.
𝜙 (𝑥 ) 𝜙2 (𝑥0 )
𝑊(𝜙1 , 𝜙2 )(𝑥0 ) = | 1′ 0 | ≠ 0.
𝜙1 (𝑥0 ) 𝜙2′ (𝑥0 )
Hence the solution of the system of linear homogeneous equations (10) -(11) is trivial given
by 𝑐1 = 𝑐2 = 0. This shows that 𝜙1 and 𝜙2 are linearly independent on 𝐼.
Theorem 3: Let 𝜙1 , 𝜙2 be any two linearly independent solutions of 𝐿(𝑦) = 0 on an
interval 𝐼 . Then every solution 𝜙 of 𝐿(𝑦) = 0 can be written uniquely as
𝜙(𝑥) = 𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥), where 𝑐1 and 𝑐2 are constants.
Proof: - Let 𝜙1 and 𝜙2 be two linearly independent solutions of 𝐿(𝑦) = 0 on 𝐼. Then we
know by the Theorem 1 that
𝑊(𝜙1 , 𝜙2 )(𝑥) ≠ 0, ∀ 𝑥 ∈ 𝐼. (12)
Let 𝑥0 be any point in 𝐼 and 𝜙(𝑥) be a solution of 𝐿(𝑦) = 0 on 𝐼 satisfying the conditions
𝜙(𝑥0 ) =∝ , 𝜙 ′ (𝑥0 ) = 𝛽. (13)
From equation (12), we have 𝑊(𝜙1 , 𝜙2 )(𝑥0 ) ≠ 0, for 𝑥0 ∈ 𝐼  the system of linear non-
homogenous equations
𝑐1 𝜙1 (𝑥0 ) + 𝑐2 𝜙2 (𝑥0 ) =∝, (14)
and
𝑐1 𝜙1′ (𝑥0 ) + 𝑐2 𝜙2′ (𝑥0 ) = 𝛽, (15)
has a unique constants 𝑐1 and 𝑐2 satisfying (14)-(15). Choose these constants 𝑐1 and 𝑐2 and
define a function
𝜓(𝑥) = 𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥) ∀ 𝑥 ∈ 𝐼. (16)
We see that
𝐿(𝜓) = 𝑐1 𝐿(𝜙1 ) + 𝑐2 𝐿(𝜙2 )
⇒ 𝐿(𝜓) = 0 on 𝐼.
Further, we see from equations (14) and (15) that
𝜓(𝑥0 ) = 𝑐1 𝜙1 (𝑥0 ) + 𝑐2 𝜙2 (𝑥0 ) =∝
and
𝜓′ (𝑥0 ) = 𝑐1 𝜙′1 (𝑥0 ) + 𝑐2 𝜙2′ (𝑥0 ) = 𝛽.

S25013: Ordinary Differential Equations Page 25


⇒ 𝜓(𝑥) is also a solution of the IVP 𝐿(𝑦) = 0 , 𝑦(𝑥0 ) =∝ , 𝑦 ′ (𝑥0 ) = 𝛽 . It follows from
uniqueness theorem that 𝜓(𝑥) = 𝜙(𝑥) on 𝐼 . This is the solution 𝜙(𝑥) of 𝐿(𝑦) = 0 is
uniquely expressed as 𝜙(𝑥) = 𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥), where 𝑐1 , 𝑐2 are constants. This proves
the theorem.

02-02: A FORMULA FOR THE WRONSKIAN


For a given differential equation, we will derive in the following theorem that the required
formula for the Wronskian of two solutions of 𝐿(𝑦) = 0. With help of the theorem, we
will find the nature of the Wronskian just by working at the values of the c oefficients of
𝐿(𝑦) = 0. We will also understand by the theorem that the Wronskian of the solutions of
𝐿(𝑦) = 0 is identically zero or never zero depending upon the solutions are linearly
dependent or linearly independent.
Theorem 4: If 𝜙1 , 𝜙2 are two solutions of 𝐿(𝑦) = 𝑦 ′′ + 𝑎1 𝑦 ′ + 𝑎2 𝑦 = 0 on an interval 𝐼
containing a point 𝑥0 , then
𝑊(𝜙1 , 𝜙2 )(𝑥) = 𝑒 −𝑎1 (𝑥−𝑥0) 𝑊(𝜙1 , 𝜙2 )(𝑥0 ).
Proof: Let 𝜙1 (𝑥) and 𝜙2 (𝑥) be two linearly independent solutions of 𝐿(𝑦) = 0 on an
interval 𝐼. Then we have
⇒ 𝜙1′′ (𝑥) + 𝑎1 𝜙′1 (𝑥) + 𝑎2 𝜙1 (𝑥) = 0 (17)
and
𝜙′′2 (𝑥) + 𝑎1 𝜙′2 (𝑥) + 𝑎2 𝜙2 (𝑥) = 0 (18)
The Wronskian of the solution 𝜙1 (𝑥), 𝜙2 (𝑥) of 𝐿(𝑦) = 0 is given by
𝜙 (𝑥) 𝜙2 (𝑥)
𝑊(𝜙1 , 𝜙2 )(𝑥) = | 1′ | = 𝜙1 𝜙′2 − 𝜙1′ 𝜙2 . (19)
𝜙1 (𝑥) 𝜙′2 (𝑥)
Differentiating equation (19) with respect to 𝑥, we get
𝑊′(𝜙1 , 𝜙2 )(𝑥) = 𝜙1 𝜙′′2 − 𝜙1′′ 𝜙2 . (20)
Eliminating 𝜙1 ′′ (𝑥)
and 𝜙2 ′′ (𝑥)
from the equation (20) by using the equations (17) and (18),
we get
𝑊 ′ (𝜙1 , 𝜙2 )(𝑥) = 𝜙1 [−𝑎1 𝜙2 ′ − 𝑎2 𝜙2 ] − 𝜙2 [−𝑎1 𝜙1′ − 𝑎2 𝜙1 ]
⇒ W ′ (𝜙1 , 𝜙2 )(𝑥) = −𝑎1 (𝜙1 𝜙2′ − 𝜙1′ 𝜙2 )
= −𝑎1 𝑊(𝜙1 , 𝜙2 )(𝑥)
or
𝑊 ′ (𝜙1 , 𝜙2 )(𝑥) + 𝑎1 𝑊(𝜙1 , 𝜙2 )(𝑥) = 0. (21)
We see from the equation (21) that the Wronskian 𝑊(𝜙1 , 𝜙2 )(𝑥) satisfies the linear
differential equation of first order whose solution is given by
𝑥
𝑊(𝜙1 , 𝜙2 )(𝑥) = 𝑐 𝑒𝑥𝑝 [− ∫𝑥 𝑎1 𝑑𝑡]
0

𝑊(𝜙1 , 𝜙2 )(𝑥) = 𝑐 𝑒𝑥𝑝[−𝑎1 (𝑥 − 𝑥0 )], (22)


where 𝑐 is a constant of integration and is to be determined. At 𝑥 = 𝑥0 , we have from
equation (22) that
𝑊(𝜙1 , 𝜙2 )(𝑥0 ) = 𝑐. (23)
Consequently, equation (22) becomes
𝑊(𝜙1 , 𝜙2 )(𝑥) = 𝑒 −𝑎1 (𝑥−𝑥0) 𝑊(𝜙1 , 𝜙2 )(𝑥0 ). (24)

S25013: Ordinary Differential Equations Page 26


This is the required formula for the Wronskian of solutions of the equation 𝐿(𝑦) = 0 and
some time it is also called Abel’s formula for Wronskian.
Remarks: We derive the following observations from the equation (24) which help s the
readers to prepare for the SET and NET examinations.
1) If 𝜙1 , 𝜙2 are linearly independent solutions of 𝐿(𝑦) = 0 on 𝐼 , then we have
𝑊(𝜙1 , 𝜙2 )(𝑥) ≠ 0, ∀ 𝑥 ∈ 𝐼.
2) If 𝜙1 , 𝜙2 are linearly dependent solution of 𝐿(𝑦) = 0 on 𝐼 , then we have
𝑊(𝜙1 , 𝜙2 )(𝑥) = 0, ∀ 𝑥 ∈ 𝐼.
3) If 𝑎1 = 0 i. e. the coefficient of 𝑦’ in the equation 𝐿(𝑦) = 0 is zero, then we have from
the equation (24) that 𝑊(𝜙1 , 𝜙2 )(𝑥) = 𝑊(𝜙1 , 𝜙2 )(𝑥0 ), ∀ 𝑥 ∈ 𝐼 which is always constant
4) To find the Wronskian of the solutions of the higher order different ial equation at a
point 𝑥 in 𝐼 is complicated. However, we choose a point 𝑥0 = 0 (say) in 𝐼 and find the
Wronskian of the solutions at 𝑥0 = 0 which is easy. Then the Wronskian at any point 𝑥
of 𝐼 is given by
𝑊(𝜙1 , 𝜙2 )(𝑥) = 𝑒 −𝑎1 𝑥 𝑊(𝜙1 , 𝜙2 )(0).
Example 03: Obtain the unique solution of the IVP 𝑦 ′′ + 6𝑦 ′ + 8𝑦 = 0, 𝑦(0) = 1, 𝑦 ′ (0) =
0.
Solution: The given problem is 𝑦 ′′ + 6𝑦 ′ + 8𝑦 = 0, 𝑦(0) = 1 , 𝑦 ′ (0) = 0 . The characteristic
equation of the given equation has real and distinct roots −2, −4. Therefore, its general
solution is given by
𝜙(𝑥) = 𝑐1 𝑒 −2𝑥 + 𝑐2 𝑒 −4𝑥 .
The initial conditions imply
𝑐1 + 𝑐2 = 1
and
−2𝑐1 − 4𝑐2 = 0 ⟹ 𝑐1 + 2𝑐2 = 0.
Solving these equations, we get 𝑐1 = 2, 𝑐2 = −1 and hence the unique solution of the given
equation is
𝜙(𝑥) = 2𝑒 −2𝑥 − 𝑒 −4𝑥 .

02-03: THE NON-HOMOGENEOUS EQUATION OF ORDER TWO :


Introduction: In this article we will study the method of finding solutions of the non -
homogeneous equation of order two given by
𝐿(𝑦) = 𝑦′′ + 𝑎1 𝑦′ + 𝑎2 𝑦 = 𝑏(𝑥), (25)
where 𝑏(𝑥) is some continuous function on an interval 𝐼. Let us suppose that 𝜓𝑝 (𝑥) be a
particular solution of the equation (25) and 𝜓(𝑥) be any other solution of (25)
Consider
𝐿(𝜓 − 𝜓𝑝 ) = 𝐿(𝜓) − 𝐿(𝜓𝑝 ) = 𝑏 − 𝑏 = 0, on 𝐼.
⇒ 𝜓 − 𝜓𝑝 is a solution of the homogeneous equation 𝐿(𝑦) = 0. Now if 𝜙1 (𝑥), 𝜙2 (𝑥) are
linearly independent solution of 𝐿(𝑦) = 0, then there are constants 𝑐1 and 𝑐2 such that
𝜓 − 𝜓𝑝 = 𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥)
⇒ 𝜓(𝑥) = 𝜓𝑝 (𝑥) + 𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥) (26)

S25013: Ordinary Differential Equations Page 27


is the solution of 𝐿(𝑦) = 𝑏(𝑥).
Hence to find all solutions of the non-homogeneous equation 𝐿(𝑦) = 𝑏(𝑥) , we find a
particular solution 𝜓𝑝 (𝑥) and two linearly independent solutions 𝜙1 (𝑥), 𝜙2 (𝑥) of 𝐿(𝑦) = 0.
Theorem 5: Let 𝑏(𝑥) be a continuous function on an interval 𝐼. Then every solution 𝜓 of
𝐿(𝑦) = 𝑏(𝑥) on 𝐼 can be written as 𝜓(𝑥) = 𝜓𝑝 (𝑥) + 𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥) , where 𝜓𝑝 is a
particular solution, and 𝜙1 (𝑥), 𝜙2 (𝑥) are two linearly independent solutions of 𝐿(𝑦) =
0 and 𝑐1 , 𝑐2 are constants, and 𝜓𝑝 is given by
𝑥 [𝜙1 (𝑡)𝜙2 (𝑥)−𝜙1 (𝑥)𝜙2 (𝑡)]𝑏(𝑡)
𝜓𝑝 = ∫𝑥 𝑑𝑡.
0 𝑊(𝜙1 ,𝜙2 )(𝑡)
Proof: (By variation of constant method):
If 𝜙1 (𝑥), 𝜙2 (𝑥) are two linearly independent solutions of 𝐿(𝑦) = 0 and 𝑐1 , 𝑐2 are constants,
then we know every solution of 𝐿(𝑦) = 0 is of the form 𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥) . Such a
function 𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥) can not be a solution of 𝐿(𝑦) = 0 unless 𝑏(𝑥) = 0 on 𝐼 .
However, if we assume 𝑐1 , 𝑐2 to become functions say 𝑢1 and 𝑢2 of 𝑥, and not necessarily
constants on 𝐼 and then ask whether 𝑢1 𝜙1 (𝑥) + 𝑢2 𝜙2 (𝑥) is a solution of a non-
homogeneous equation 𝐿(𝑦) = 𝑏(𝑥) on 𝐼. The answer is yes, and the remarkable thing is
that it works, and the procedure is known as variation of constants.
We assume therefore, the function 𝑢1 𝜙1 (𝑥) + 𝑢2 𝜙2 (𝑥) is a solution of non-homogeneous
equation 𝐿(𝑦) = 𝑏(𝑥) on 𝐼, where 𝑢1 and 𝑢2 are functions of 𝑥 and are to be determined. ⇒
𝐿(𝑢1 𝜙1 + 𝑢2 𝜙2 ) = 𝑏(𝑥) on 𝐼.
⇒ (𝑢1 𝜙1 + 𝑢2 𝜙2 )′′ + 𝑎1 (𝑢1 𝜙1 + 𝑢2 𝜙2 )′ + 𝑎2 (𝑢1 𝜙1 + 𝑢2 𝜙2 ) = 𝑏(𝑥)
⇒ (𝑢1′ 𝜙1 + 𝑢1 𝜙1 ′ + 𝑢2′ 𝜙2 + 𝑢2 𝜙2′ )′ + 𝑎1 (𝑢1′ 𝜙1 + 𝑢1 𝜙1′ + 𝑢2′ 𝜙2 + 𝑢2 𝜙2′ )
+𝑎2 (𝑢1 𝜙1 + 𝑢2 𝜙2 ) = 𝑏(𝑥)
⇒ (𝑢1′′ 𝜙1 + 2𝑢1′ 𝜙 ′ + 𝑢1 𝜙1′′ + 𝑢2′′ 𝜙2 + 2𝑢′2 𝜙2′ + 𝑢2 𝜙2′′ ) + 𝑎1 (𝑢1′ 𝜙1 + 𝑢1 𝜙1′ + 𝑢2′ 𝜙2 + 𝑢2 𝜙2′ )
+𝑎2 (𝑢1 𝜙1 + 𝑢2 𝜙2 ) = 𝑏(𝑥).
We write this as
(𝜙1 𝑢1′′ + 𝜙2 𝑢2′′ ) + 2(𝑢1′ 𝜙1′ + 𝑢2′ 𝜙2′ ) + 𝑢1 (𝜙1′′ + 𝑎1 𝜙1′ + 𝑎2 𝜙1 ) + 𝑢2 (𝜙2′′ + 𝑎1 𝜙2′ + 𝑎2 𝜙2 )
+𝑎1 (𝑢1′ 𝜙1 + 𝑢2′ 𝜙2 ) = b(x).
i.e. (𝜙1 𝑢1′′ + 𝜙2 𝑢2′′ ) + 2(𝑢1′ 𝜙1′ + 𝑢2′ 𝜙2′ ) + 𝑎1 (𝑢1′ 𝜙1 + 𝑢2′ 𝜙2 ) + 𝑢1 𝐿(𝜙1 ) + 𝑢2 𝐿(𝜙2 ) = b(𝑥)
Since 𝜙1 and 𝜙2 are linearly independent solutions of 𝐿(𝑦) = 0, on 𝐼, we have 𝐿(𝜙1 ) = 0
and 𝐿(𝜙2 ) = 0.
Therefore, we get
(𝜙1 𝑢1′′ + 𝜙2 𝑢2′′ ) + 2(𝑢1′ 𝜙1′ + 𝑢2′ 𝜙2′ ) + 𝑎1 (𝑢1′ 𝜙1 + 𝑢2′ 𝜙2 ) = b(𝑥). (27)
If 𝑢1 𝜙1 + 𝑢2 𝜙2 = 0,
′ ′
(28)
then differentiating this with respect to 𝑥, we have
(𝑢1′ 𝜙1 + 𝑢2′ 𝜙2 )′ = 𝑢1′′ 𝜙1 + 𝑢′1 𝜙1′ + 𝑢′′2 𝜙2 + 𝑢′2 𝜙2′ = 0
⇒ (𝑢1′ 𝜙1 + 𝑢2′ 𝜙2 )′ = (𝜙1 𝑢1′′ + 𝜙2 𝑢2′′ ) + (𝑢1′ 𝜙1′ + 𝑢2′ 𝜙2′ ) = 0
Using equation (27), we have
𝑢1′ 𝜙1′ + 𝑢2′ 𝜙2′ = 𝑏(𝑥). (29)
Equation (28) and (29) together is a system of linear equations in 𝑢1 and 𝑢2 as 𝜙1 , 𝜙2 are
′ ′

known solutions of 𝐿(𝑦) = 0.

S25013: Ordinary Differential Equations Page 28


Solving equations (28) and (29) by Cramer’s rule, we find
𝑢1′ 𝑢2′ 1
0 𝜙2 = 𝜙1 0 = 𝜙1 𝜙2 .
| | | ′ | | ′ |
𝑏 𝜙2′ 𝜙1 𝑏 𝜙1 𝜙2′
0 𝜙2 𝜙1 0
| ′| | ′ |
𝑏 𝜙 𝜙 𝑏
⇒ 𝑢1′ = 𝑊 2 , 𝑢2′ = 1𝑊 ,
𝜙1 𝜙2
where 𝑊 = 𝑊(𝜙1 , 𝜙2 )(𝑥) = | ′ | (30)
𝜙1 𝜙2′
𝑊 𝑊
or 𝑢1′ = 𝑊(𝑥)
1
𝑏(𝑥), 𝑢2′ = 𝑊(𝑥)
2
𝑏(𝑥), (31)
where 𝑊1 and 𝑊2 are the determinants obtained from the Wronskian 𝑊(𝑥) given in (30) by
replacing the corresponding column of 𝑊 indicated by the suffix by the column vector
(0, 1). Obviously in this case
𝑊1 = −𝜙2 (𝑥) and 𝑊2 = 𝜙1 (𝑥).
Hence equations (31) become
−𝜙2 (𝑥) 𝜙1 (𝑥)
𝑢1′ = 𝑏(𝑥), 𝑢2′ = 𝑏(𝑥). (32)
𝑊(𝑥) 𝑊(𝑥)
Integrating equation in (32) between the limits 𝑥0 to 𝑥, we get
𝑥 𝜙2 (𝑡)𝑏(𝑡) 𝑥 𝜙1 (𝑡)𝑏(𝑡)
𝑢1 (𝑥) = − ∫𝑥 𝑑𝑡, 𝑢2 (𝑥) = ∫𝑥 𝑑𝑡. (33)
0 𝑊(𝜙1 ,𝜙2 )(𝑡) 0 𝑊(𝜙1 ,𝜙2 )(𝑡)
Hence, the particular solution 𝜓𝑝 = 𝑢1 𝜙1 + 𝑢2 𝜙2 becomes
𝑥 [𝜙1 (𝑡)𝜙2 (𝑥)−𝜙2 (𝑡)𝜙1 (𝑥)]
𝜓𝑝 (𝑥) = ∫𝑥 𝑏(𝑡)𝑑𝑡 (34)
0 𝑊(𝜙1 ,𝜙2 )(𝑡)
Hence, the general solution of 𝐿(𝑦) = 𝑏(𝑥) becomes
𝜓(𝑥) = 𝜓𝑝 (𝑥) + 𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥). (35)
Example 04: Find all solutions of the non-homogeneous equation 4𝑦′′ − 𝑦 = 𝑒 𝑥 .
1 1
Solution: We rewrite the given equation in standard form 𝑦′′ − 4 𝑦 = 4 𝑒 𝑥 , ⇒ 𝐿(𝑦) = 𝑏(𝑥),
1 1
where 𝑏(𝑥) = 4 𝑒 𝑥 . The characteristic equation of the equation is 𝑟 2 − 4 = 0. It has roots
1 −1
𝑟1 = 2 , 𝑟2 = . Hence, the two linearly independent solutions 𝜙1 , 𝜙2 of the homogeneous
2
equation 𝐿(𝑦) = 0 are
1 1
𝜙1 (𝑥) = 𝑒 2𝑥 , 𝜙2 (𝑥) = 𝑒 −2𝑥 . (36)
Therefore, the particular solution 𝜓𝑝 (𝑥) of the non-homogeneous equation 𝐿(𝑦) = 𝑏(𝑥) is
of the form
𝜓𝑝 (𝑥) = 𝑢1 𝜙1 (𝑥) + 𝑢2 𝜙2 (𝑥)
1 1
i. e. 𝜓𝑝 (𝑥) = 𝑢1 𝑒 2𝑥 + 𝑢2 𝑒 −2𝑥 , (37)
where 𝑢1 and 𝑢2 are respectively given in the equation (33). To determine these functions,
we first find the Wronskian of the solutions 𝜙1 , 𝜙2 of 𝐿(𝑦) = 0.
Therefore,
1 1
𝑒 2𝑥 𝑒 −2 𝑥
𝑊(𝜙1 , 𝜙2 )(𝑥) = |1 1
1 1 |
𝑒 2𝑥 − 2 𝑒 −2𝑥
2
1 1 1
⇒ 𝑊(𝜙1 , 𝜙2 )(𝑥) = 2 | |
1 −1
S25013: Ordinary Differential Equations Page 29
⇒ 𝑊(𝜙1 , 𝜙2 )(𝑥) = −1. (38)
Hence, from equation (33), we have
𝑥 𝑏(𝑡)𝜙2 (𝑡)
𝑢1 (𝑥) = − ∫0 𝑑𝑡
𝑊(𝑡)
1
1 𝑥
= 4 ∫0 𝑒 2𝑡 𝑑𝑡
1
1
⇒ 𝑢1 (𝑥) = 2 (𝑒 2𝑥 − 1).
Similarly, we find
𝑥 𝑏(𝑡)𝜙1 (𝑡)
𝑢2 (𝑥) = ∫0 𝑑𝑡
𝑊(𝑡)
3
1 𝑥
= − 4 ∫0 𝑒 2𝑡 𝑑𝑡
3
1
⇒ 𝑢2 (𝑥) = − 6 (𝑒 2𝑥 − 1).
Substituting these values of 𝑢1 (𝑥) and 𝑢2 (𝑥) in the equation (37), we get
1 1 1 1 1
𝜓𝑝 (𝑥) = 3 𝑒 𝑥 − 2 𝑒 2𝑥 + 6 𝑒 − 2𝑥 (39)
1 1
1 1
The expression − 2 𝑒 2
𝑥
+ 6𝑒 − 𝑥
2 is a solution of 𝐿(𝑦) = 0 , so it can be absorbed in the
1 1
general solution 𝑐1 𝑒 2𝑥 + 𝑐2 𝑒 − 2𝑥 of 𝐿(𝑦) = 0. Hence, we take particular solution of 𝐿(𝑦) =
𝑏(𝑥) as
1
𝜓𝑝 (𝑥) = 3 𝑒 𝑥 .
Therefore, the most general solution represented by (35) becomes
1 1
1
𝜓(𝑥) = 𝑐1 𝑒 2𝑥 + 𝑐2 𝑒 − 2𝑥 + 3 𝑒 𝑥 .

SOLVED PROBLEMS 01
Problem 01- 02-01:
Prove or disprove the following.
i) If 𝜙1 , 𝜙2 are linearly independent functions on interval 𝐼 , then they are linearly
independent on some interval 𝐽 contained inside 𝐼.
ii) If 𝜙1 , 𝜙2 are linearly dependent functions on an interval 𝐼 , then they are linearly
dependent on any interval 𝐽 contained inside 𝐼.
Solution: i) The proof is illustrated by considering an example. Let 𝜙1 (𝑥) = 𝑥 2 , 𝜙2 (𝑥) =
𝑥|𝑥| be two functions defined on 𝐼: − ∞ < 𝑥 < ∞. Consider the sum
𝑐1 𝑥 2 + 𝑐2 𝑥|𝑥| = 0 ∀ 𝑥 in − ∞ < 𝑥 < ∞.
⇒ 𝑐1 𝑥 2 + 𝑐2 𝑥 2 =0, ∀ 0 ≤ 𝑥 < ∞
⇒ 𝑐1 + 𝑐2 = 0
and
𝑐1 𝑥 2 − 𝑐2 𝑥 2 = 0, ∀ − ∞ < 𝑥 < 0
⇒ 𝑐1 − 𝑐2 = 0.
This gives 𝑐1 = 𝑐2 = 0.
This shows that the functions 𝜙1 (𝑥) = 𝑥 2 , 𝜙2 (𝑥) = 𝑥|𝑥| are linearly independent in 𝐼: −∞ <
𝑥 < ∞.

S25013: Ordinary Differential Equations Page 30


Now take 𝐽: 0 ≤ 𝑥 < ∞ be interval in 𝐼: −∞ < 𝑥 < ∞. Then clearly, we have 𝜙1 (𝑥) = 𝑥 2 and
𝜙2 (𝑥) = 𝑥 2 in 𝐽 ⊆ 𝐼 which are linearly dependent functions in 𝐽  𝐼 . Thus if 𝜙1 , 𝜙2 are
linearly independent functions on an interval 𝐼, then they are not linearly independent on
some interval 𝐽 inside 𝐼.
ii) We prove that if 𝜙1 , 𝜙2 are linearly dependent functions on an interval 𝐼, then they are
linearly dependent in any interval 𝐽 contained inside 𝐼.
Let 𝜙1 (𝑥) and 𝜙2 (𝑥) be linearly dependent functions on 𝐼. ⇒ if the linear sum
𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥) = 0, ∀ 𝑥 ∈ 𝐼 (40)
such that not both 𝑐1 and 𝑐2 are zero.
Now, if 𝐽 is any subinterval contained in 𝐼, then it follows from the equation (40) that
𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥) = 0, ∀ 𝑥 ∈ 𝐽 ⊆ 𝐼
such that not both 𝑐1 and 𝑐2 are zero. This proves that 𝜙1 and 𝜙2 are linearly dependent in
𝐽  𝐼.
,

Remark: However, if the functions 𝜙1 (𝑥) and 𝜙2 (𝑥) are linearly independent (or
dependent) solutions of a differential equation 𝐿(𝑦) = 0 on an interval 𝐼, then we know
form the Theorem 1 that their Wronskian
𝑊(𝜙1 , 𝜙2 )(𝑥) ≠ 0, ∀ 𝑥 ∈ 𝐼 (or 𝑊(𝜙1 , 𝜙2 )(𝑥) = 0, ∀ 𝑥 ∈ 𝐼)
⇒ 𝑊(𝜙1 , 𝜙2 )(𝑥) ≠ 0, ∀ 𝑥 ∈ 𝐽 ⊆ 𝐼 (or 𝑊(𝜙1 , 𝜙2 )(𝑥) = 0, ∀ 𝑥 ∈ 𝐽 ⊆ 𝐼)
⇒ 𝜙1 (𝑥) and 𝜙2 (𝑥) are linearly independent (or linearly dependent) solution of 𝐿(𝑦) = 0
in any sub interval 𝐽 contained inside 𝐼.
Problem 01-02-02:
Let 𝜙1 , 𝜙2 be two different functions on an interval 𝐼 which are not necessarily solutions
of an equation 𝐿(𝑦) = 0. Prove the following
i) If 𝜙1 , 𝜙2 are linearly dependent on 𝐼, then 𝑊(𝜙1 , 𝜙2 )(𝑥) = 0, ∀ 𝑥 ∈ 𝐼.
ii) 𝑊(𝜙1 , 𝜙2 )(𝑥) = 0 for all 𝑥 in 𝐼 does not imply that 𝜙1 , 𝜙2 are linearly dependent on 𝐼.
iii) If 𝑊(𝜙1 , 𝜙2 )(𝑥0 ) ≠ 0 for some 𝑥0 ∈ 𝐼, then 𝜙1 , 𝜙2 are linearly independent on 𝐼.
iv) If 𝜙1 , 𝜙2 are linearly independent functions on 𝐼 does not imply that 𝑊(𝜙1 , 𝜙2 )(𝑥) ≠
0.
v) 𝑊(𝜙1 , 𝜙2 )(𝑥) = 0 for all 𝑥 in 𝐼 and 𝜙2 (𝑥) ≠ 0 on 𝐼 imply that 𝜙1 and 𝜙2 are linearly
dependent.
Solution: i) Let 𝜙1 and 𝜙2 be linearly dependent functions on 𝐼. This means that there exist
some constants 𝑐1 and 𝑐2 not both zero such that
𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥)=0
𝑐
⇒ 𝜙1 = − 𝑐2 𝜙2 , for 𝑐1 ≠ 0.
1

Thus, the Wronskian of the functions becomes


𝜙 𝜙2
𝑊(𝜙1 , 𝜙2 )(𝑥) = | 1′ |
𝜙1 𝜙2′
= 𝜙1 (𝑥)𝜙2′ (𝑥) − 𝜙1′ (𝑥)𝜙2 (𝑥)
𝑐 𝑐
= − 𝑐2 𝜙2 (𝑥)𝜙2′ (𝑥) + 𝑐2 𝜙2′ (𝑥)𝜙2 (𝑥)
1 1

= 0, ∀ 𝑥 ∈ 𝐼 .

S25013: Ordinary Differential Equations Page 31


⇒ if 𝜙1 , 𝜙2 are linearly dependent on 𝐼, then 𝑊(𝜙1 , 𝜙2 )(𝑥) = 0, ∀ 𝑥 ∈ 𝐼.
ii) This is the converse of i), and the converse is not true. We prove the result by consider
the example. Let the two functions 𝜙1 (𝑥) = 𝑥 2 , 𝜙2 (𝑥) = 𝑥|𝑥| defined on −∞ < 𝑥 < ∞.
For 𝑥 > 0 , we have |𝑥| = 𝑥 ⟹ 𝜙1 (𝑥) = 𝑥 2 , 𝜙2 (𝑥) = 𝑥 2 .
2
⟹ 𝑊(𝜙1 , 𝜙2 )(𝑥) = | 𝑥 𝑥 2 | ⟹ 𝑊(𝜙 , 𝜙 )(𝑥) = 0, 𝑥 > 0.
1 2
2𝑥 2𝑥
For 𝑥 = 0 ⟹ 𝜙1 (𝑥) = 𝜙2 (𝑥) = 0 ⇒ 𝑊(𝜙1 , 𝜙2 )(𝑥) = 0, for 𝑥 = 0.
Now let 𝑥 < 0 ⇒ |𝑥| = −𝑥 ⟹ 𝜙1 (𝑥) = 𝑥 2 , 𝜙2 (𝑥) = −𝑥 2 .
2
⟹ 𝑊(𝜙1 , 𝜙2 )(𝑥) = | 𝑥 −𝑥 2 |
2𝑥 −2𝑥
⟹ W(𝜙1 , 𝜙2 )(𝑥) = 0, ∀ 𝑥 < 0.
Thus, from above equations, we have 𝑊(𝜙1 , 𝜙2 )(𝑥)=0, in −∞ < 𝑥 < ∞.
Now consider the sum
𝑐1 𝑥 2 + 𝑐2 𝑥|𝑥| = 0, in − ∞ < 𝑥 < ∞
⟹ 𝑐1 𝑥 2 + 𝑐2 𝑥 2 = 0, for 0 ≤ 𝑥 < ∞
and
𝑐1 𝑥 2 − 𝑐2 𝑥 2 = 0 for −∞ < 𝑥 < 0
⟹ 𝑐1 + 𝑐2 = 0, and 𝑐1 − 𝑐2 = 0 ⟹ 𝑐1 =𝑐2 = 0.
Thus 𝑐1 𝑥 + 𝑐2 𝑥|𝑥| = 0 ⇒ 𝑐1 = 𝑐2 = 0 ⟹ 𝜙1 (𝑥) = 𝑥 2 and 𝜙2 (𝑥) = 𝑥|𝑥| are linearly
2

independent in 𝐼: −∞ < 𝑥 < ∞. Thus, we have 𝑊(𝜙1 , 𝜙2 )(𝑥) = 0, ∀ 𝑥 ∈ 𝐼 does not imply
𝜙1 (𝑥), 𝜙2 (𝑥) are linearly dependent on 𝐼.
iii) Let 𝑊(𝜙1 , 𝜙2 )(𝑥0 ) ≠ 0 for some 𝑥0 ∈ 𝐼.
Consider the equation
𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥) = 0, ∀ 𝑥 ∈ 𝐼
⟹ c1 𝜙1′ (𝑥) + 𝑐2 𝜙2′ (𝑥) = 0, ∀ 𝑥 ∈ 𝐼.
This gives the following
𝑐1 𝜙1 (𝑥0 ) + 𝑐2 𝜙2 (𝑥0 ) = 0
and
𝑐1 𝜙1′ (𝑥0 ) + 𝑐2 𝜙2′ (𝑥0 ) = 0, for some 𝑥0 ∈ 𝐼.
The determinant of the coefficients of the system of linear equations is just the Wronskian
of 𝜙1 , 𝜙2 ,
𝜙 (𝑥 ) 𝜙2 (𝑥0 )
i. e. 𝑊(𝜙1 , 𝜙2 )(𝑥0 ) = | 1′ 0 |≠0
𝜙1 (𝑥0 ) 𝜙2′ (𝑥0 )
⇒ 𝑐1 = 0 and 𝑐2 = 0 are the only solution of the system of linear equations. Thu s, the sum
𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥) = 0, ∀ 𝑥 ∈ 𝐼
⇒ 𝑐1 = 0 and 𝑐2 = 0.
Hence, 𝜙1 (𝑥) and 𝜙2 (𝑥) are linearly independent. Thus, we have proved that
𝑊(𝜙1 , 𝜙2 )(𝑥0 ) ≠ 0, for some 𝑥0 ∈ 𝐼 ⇒ 𝜙1 (𝑥) and 𝜙2 (𝑥) are linearly independent on 𝐼.
iv) This is converse of iii) and it is not true. The proof of this result is evident from the
example illustrated in ii). We have proved that the functions 𝜙1 (𝑥) = 𝑥 2 , 𝜙2 (𝑥) =
𝑥|𝑥| defined on 𝐼: − ∞ < 𝑥 < ∞ are linearly independent in 𝐼, where as 𝑊(𝜙1 , 𝜙2 )(𝑥) =

S25013: Ordinary Differential Equations Page 32


0 ∀ 𝑥 ∈ 𝐼. This implies that, if 𝜙1 , 𝜙2 are linearly independent functions on 𝐼, then it does
not imply 𝑊(𝜙1 , 𝜙2 )(𝑥) ≠ 0.
v) Let 𝑊(𝜙1 , 𝜙2 )(𝑥) = 0, ∀ 𝑥 ∈ 𝐼 and 𝜙2 (𝑥) ≠ 0, on 𝐼.
𝜙 𝜙2 (𝑥)
⟹| ′ 1 |=0
𝜙1 (𝑥) 𝜙2′ (𝑥)
⟹ 𝜙1 (𝑥)𝜙′2 (𝑥) − 𝜙2 (𝑥)𝜙′1 (𝑥) = 0.
Since 𝜙2 (𝑥) ≠ 0, hence dividing this equation by ϕ22 (𝑥), we get
𝜙2 (𝑥)𝜙1′ (𝑥)−𝜙1 (𝑥)𝜙2′ (𝑥)
= 0, ∀ 𝑥 ∈ 𝐼
ϕ22 (𝑥)
𝜙 (𝑥) ′
⟹ (𝜙1(𝑥)) = 0, ∀ 𝑥 ∈ 𝐼.
2
𝜙1 (𝑥)
Integrating, we get = 𝑘, where 𝑘 is some constant. Hence, we have
𝜙2 (𝑥)
𝜙1 (𝑥) = 𝑘2 𝜙2 (𝑥).
This shows that 𝜙1 (𝑥) and 𝜙2 (𝑥) and linearly independent on 𝐼.
Remarks:
1) The result i) is also true for the solutions 𝜙1 (𝑥) and 𝜙2 (𝑥) of 𝐿(𝑦) = 0 on 𝐼.
i.e. If 𝜙1 , 𝜙2 are linearly dependent solutions of 𝐿(𝑦) = 0 on 𝐼, then 𝑊(𝜙1 , 𝜙2 )(𝑥) = 0,
∀ 𝑥 ∈ 𝐼.
2) If 𝜙1 (𝑥) and 𝜙2 (𝑥) are solution of 𝐿(𝑦) = 0 on 𝐼 , then 𝑊(𝜙1 , 𝜙2 )(𝑥) = 0, ∀ 𝑥 ∈ 𝐼 ⟹
𝜙1 , 𝜙2 are linearly dependent on 𝐼.
3) The result iii) is also true for the solution 𝜙1 , 𝜙2 of the equation 𝐿(𝑦) = 0.
4) If 𝜙1 (𝑥) and 𝜙2 (𝑥) are linearly independent solutions of 𝐿(𝑦) = 0 on 𝐼 ,then
𝑊(𝜙1 , 𝜙2 )(𝑥) ≠ 0, ∀ 𝑥 ∈ 𝐼.
Problem 01-02-03:
Let 𝜙𝑛 (𝑥) be any function satisfying the boundary value problem 𝑦 ′′ + 𝑛2 𝑦 = 0, 𝑦(0) =
𝑦(2𝜋), 𝑦 ′ (0) = 𝑦 ′ (2𝜋), 𝑛 = 0, 1, 2, 3, ⋯. Show that
2𝜋
∫0 𝜙𝑛 (𝑥)𝜙𝑚 (𝑥)𝑑𝑥 = 0, for 𝑛 ≠ 𝑚.
Solution: Given that 𝜙𝑛 (𝑥) is a solution of the boundary value problem
𝑦′′ + 𝑛2 𝑦 = 0, 𝑦(0) = 𝑦(2𝜋), 𝑦 ′ (0) = 𝑦 ′ (2𝜋). (41)
⟹ −𝜙𝑛 ′′ (𝑥) 2
= 𝑛 𝜙𝑛 (𝑥) (42)
Similarly, −𝜙𝑚 ′′ (𝑥) 2
= 𝑛 𝜙𝑚 (𝑥), (43)
for another solution 𝜙𝑚 (𝑥) of the equation (41). Multiply equation (42) by 𝜙𝑚 (𝑥) and (43)
by 𝜙𝑛 (𝑥) and subtracting, we get
′′ (𝑥)
𝜙𝑛 (𝑥)𝜙𝑚 − 𝜙𝑚 (𝑥)𝜙𝑛′′ (𝑥) = (𝑛2 − 𝑚2 )𝜙𝑛 (𝑥)𝜙𝑚 (𝑥)
⟹ (𝑛2 − 𝑚2 )𝜙𝑛 (𝑥)𝜙𝑚 (𝑥) = (𝜙𝑛 (𝑥)𝜙𝑚 ′ (𝑥)
− 𝜙𝑚 (𝑥)𝜙𝑛′ (𝑥))′
Integrating this between the limits 0 to 2𝜋, with respect to 𝑥, we get
2𝜋
(𝑛2 − 𝑚2 ) ∫0 𝜙𝑛 (𝑥) 𝜙𝑚 (𝑥)𝑑𝑥 = [𝜙𝑛 (𝑥)𝜙𝑚
′ (𝑥)
− 𝜙𝑚 (𝑥)𝜙𝑛′ (𝑥)]2𝜋
0
= 0 by the boundary conditions
2𝜋
⟹ ∫0 𝜙𝑛 (𝑥)𝜙𝑚 (𝑥)𝑑𝑥 = 0, for 𝑛 ≠ 𝑚 in {0, 1, 2, 3, ⋯ }.

S25013: Ordinary Differential Equations Page 33


Note: For 𝑛 = 0, 1, 2, ⋯ sin 𝑛𝑥 , cos 𝑛𝑥 are solutions of the boundary value problem
𝑦 ′′ + 𝑛2 𝑦 = 0, 𝑦(0) = 𝑦(2𝜋), 𝑦 ′ (0) = 𝑦 ′ (2𝜋); so by Problem 02-03,
2𝜋 2𝜋
∫0 sin 𝑛𝑥 sin 𝑚𝑥 𝑑𝑥 = ∫0 cos 𝑛𝑥 cos 𝑚𝑥 𝑑𝑥 = 0,
2𝜋
for 𝑛 ≠ 𝑚 and ∀ 𝑚, 𝑛; ∫0 sin 𝑛𝑥 cos 𝑚𝑥 𝑑𝑥 = 0.
Problem 01-02-04:
Find all solutions of the following
i) 𝑦′′ − 7𝑦′ + 6𝑦 = 𝑠𝑖𝑛 𝑥.
ii) 𝑦′′ − 3𝑦 + 2𝑦 = 𝑠𝑖𝑛 𝑒 −𝑥 .
Solution: i) Consider the equation 𝑦 ′′ − 7𝑦 ′ + 6𝑦 = sin 𝑥, where 𝑏(𝑥) = sin 𝑥.
The characteristic equation of 𝐿(𝑦) = 0 is 𝑟 2 − 7𝑟 + 6 = 0 . It has roots 𝑟1 = 1, 𝑟2 = 6 .
Hence the two linearly independent solutions of 𝐿(𝑦) = 0 are 𝜙1 (𝑥) = 𝑒 𝑥 and 𝜙2 (𝑥) =
𝑒 6𝑥 . Thus, the Wronskian of the solutions is
𝑥 6𝑥
𝑊(𝜙1 , 𝜙2 )(𝑥) = |𝑒 𝑥 𝑒 6𝑥 | ⇒ 𝑊(𝜙1 , 𝜙2 )(𝑥) = 5𝑒 7𝑥 . (44)
𝑒 6𝑒
Now the particular solution 𝜓𝑝 of the non-homogeneous equation 𝐿(𝑦) = b(𝑥) is given by
𝜓𝑝 (𝑥) = 𝑢1 𝜙1 (𝑥) + 𝑢2 𝜙2 (𝑥) (45)
where 𝑢1 and 𝑢2 are defined in (33) and are given by
1 𝑥 1 𝑥
𝑢1 (𝑥) = − 5 ∫0 𝑒 −𝑡 sin 𝑡 𝑑𝑡 and 𝑢2 (𝑥) = 5 ∫0 𝑒 −6𝑡 sin 𝑡 𝑑𝑡.
Integrating the integrals by parts, we obtain after solving the expression
1
𝑢1 (𝑥) = − 10 [1 − 𝑒 −𝑥 (𝑠𝑖𝑛 𝑥 + 𝑐𝑜𝑠 𝑥)]
and
1
𝑢2 (𝑥) = 185 [1 − 𝑒 −6𝑥 (𝑐𝑜𝑠 𝑥 + 6 𝑠𝑖𝑛 𝑥)].
Substituting these values in the equation (45), we obtain the particular solution 𝜓𝑝 (𝑥) as
𝑒𝑥 𝑒 6𝑥
𝜓𝑝 (𝑥) = − 10 [1 − 𝑒 −𝑥 (sin 𝑥 + cos 𝑥)] + 185 [1 − 𝑒 −6𝑥 (cos 𝑥 + 6 sin 𝑥)]
𝑒𝑥 sin 𝑥+cos 𝑥 cos 𝑥+6 sin 𝑥 𝑒 6𝑥
= − 10 + − + 185
10 185
𝑒𝑥 𝑒 6𝑥 5sin 𝑥+7 cos 𝑥
= − 10 + 185 + ,
74
𝑒𝑥 𝑒 6𝑥 5sin 𝑥+7 cos 𝑥
and − 10 + 185 is a solution of 𝐿(𝑦) = 0. So, take 𝜓𝑝 (𝑥) = . Hence, the most
74
general solution of 𝐿(𝑦) = 𝑏(𝑥) becomes
1
𝜓(𝑥) = 𝑐1 𝑒 𝑥 + 𝑐2 𝑒 6𝑥 + 74 (5 sin 𝑥 + 7 cos 𝑥).
ii) Now consider the equation
𝑦′′ − 3𝑦 + 2𝑦 = sin 𝑒 −𝑥 .
This is of the form 𝐿(𝑦) = 𝑏(𝑥), where 𝑏(𝑥) = sin 𝑒 −𝑥 . The characteristic polynomial is
𝑝(𝑟) = 𝑟 2 − 3𝑟 + 2 . It has roots 𝑟1 = 1, 𝑟2 = 2 . Hence the two linearly independent
solutions of 𝐿(𝑦) = 0 are 𝜙1 (𝑥) = 𝑒 𝑥 , 𝜙2 (𝑥) = 𝑒 2𝑥 . The Wronskian of the solutions is
𝑥 2𝑥
𝑊(𝜙1 , 𝜙2 )(𝑥) = |𝑒 𝑥 𝑒 2𝑥 | ⇒ 𝑊(𝜙1 , 𝜙2 )(𝑥) = 𝑒 3𝑥 .
𝑒 2𝑒
The particular solution 𝜓𝑝 of the non-homogenous equation 𝐿(𝑦) = 𝑏(𝑥) becomes
𝜓𝑝 (𝑥) = 𝑢1 (𝑥)𝑒 𝑥 + 𝑢2 (𝑥)𝑒 2𝑥 ,

S25013: Ordinary Differential Equations Page 34


where 𝑢1 (𝑥) and 𝑢2 (𝑥) are defined in the equation (33). Using the equation (33), we find
𝑢1 (𝑥) = − ∫ 𝑒 −𝑥 sin 𝑒 −𝑥 𝑑𝑥
Put 𝑒 −𝑥 = 𝑡 ⇒ −𝑒 −𝑥 𝑑𝑥 = 𝑑𝑡
⇒ 𝑢1 (𝑥)= ∫ sin 𝑡 𝑑𝑡 ⇒ 𝑢1 (𝑥) = − cos 𝑥
or 𝑢1 (𝑥) = − cos 𝑒 −𝑥 .
Similarly, 𝑢2 (𝑥) = ∫ 𝑒 −2𝑥 sin 𝑒 −𝑥 𝑑𝑥
Put 𝑒 −𝑥 = 𝑡 ⇒ −𝑒 −𝑥 𝑑𝑥 = 𝑑𝑡
⇒ 𝑢2 (𝑥)=− ∫ 𝑡 sin 𝑡 𝑑𝑡.
Integrating this by parts, we get
𝑢2 (𝑥) = 𝑡 cos 𝑡 − sin 𝑡
⇒ 𝑢2 (𝑥)=𝑒 −𝑥 cos 𝑒 −𝑥 − sin 𝑒 −𝑥 .
Using these values of 𝑢1 and 𝑢2 , we obtain the particular solution 𝜓𝑝 (𝑥) as
𝜓𝑝 (𝑥) = 𝑒 𝑥 [− cos 𝑒 −𝑥 ] + 𝑒 2𝑥 [𝑒 −𝑥 cos 𝑒 −𝑥 − sin 𝑒 −𝑥 ] = −𝑒 2𝑥 sin 𝑒 −𝑥 .
Hence, the general solution as 𝜓(𝑥) = 𝜓𝑝 + 𝑐1 𝑒 𝑥 + 𝑐2 𝑒 2𝑥 becomes
𝜓(𝑥) = 𝑐1 𝑒 𝑥 + 𝑐2 𝑒 2𝑥 − 𝑒 2𝑥 sin 𝑒 −𝑥 .
Problem 01-02-05:
Consider the equation 𝑦′′ + 𝑤 2 𝑦 = 𝐴 cos 𝑤𝑥, where 𝐴 and 𝑤 are positive constants.
i) Find all solutions on 0 ≤ 𝑥 < ∞
ii) Show that every solution 𝜓 is such that |𝜓(𝑥)| assume arbitrary large values as 𝑥 → ∞
Solution: Consider the equation
𝑦 ′′ + 𝑤 2 𝑦 = 𝐴 cos 𝑤 𝑥, (46)
where 𝐴 and 𝑤 are positive constants. The characteristic polynomial of the equation
𝐿(𝑦) = 0 is 𝑝(𝑟) = 𝑟 2 + 𝑤 2 . It has roots 𝑟 = ±𝑖𝑤 . Hence the functions 𝜙1 (𝑥) =
cos 𝑤𝑥 and 𝜙2 (𝑥) = sin 𝑤𝑥 are the linearly independent solutions of 𝐿(𝑦) = 0. Then also
we have solution of 𝐿(𝑦) = 0 as
𝜙(𝑥) = 𝑐1 cos 𝑤 𝑥 + 𝑐2 sin 𝑤 𝑥, (47)
where 𝑐1 and 𝑐2 are arbitrary constants. Now the particular solution of the non –
homogenous equation 𝐿(𝑦) = 𝑏(𝑥) is given by
𝜓𝑝 (𝑥) = 𝑢1 (𝑥) cos 𝑤 𝑥 + 𝑢2 (𝑥) sin 𝑤 𝑥, (48)
where the function 𝑢1 (𝑥) and 𝑢2 (𝑥) are defined in the equation (33). The Wronskian of the
solutions of 𝐿(𝑦) = 0 is given by
cos 𝑤 𝑥 sin 𝑤 𝑥
𝑊(𝜙1 , 𝜙2 )(𝑥) = | | = 𝑤.
−𝑤 sin 𝑤 𝑥 𝑤 cos 𝑤 𝑥
Hence, we find from equation (33), the values of
𝑥 𝑏(𝑡)𝜙2 (𝑡)
𝑢1 (𝑥) = − ∫0 𝑑𝑡
𝑊(𝑡)
𝐴 𝑥
= − 2𝑤 ∫0 sin 2 𝑤𝑡 𝑑𝑡
𝐴
= 4𝑤2 (cos 2 𝑤𝑥 − 1)
and
𝑥 𝑏(𝑡)𝜙1 (𝑡)
𝑢2 (𝑥) = ∫0 𝑑𝑡
𝑊(𝑡)
𝐴 𝑥
= 𝑤 ∫0 cos 2 𝑤𝑡 𝑑𝑡

S25013: Ordinary Differential Equations Page 35


𝐴 sin 2𝑤𝑥
= 2𝑤 [𝑥 + ].
2𝑤
Substituting this in the equation (48), we obtain
𝐴
𝜓𝑝 (𝑥) = 4𝑤2 [(cos 2 𝑤𝑥 − 1) cos 𝑤 𝑥 + (2𝑤𝑥 + sin 2 𝑤𝑥) sin 𝑤 𝑥]
𝐴
= 4𝑤2 [cos 2 𝑤𝑥. cos 𝑤 𝑥 + sin 2 𝑤𝑥 sin 𝑤 𝑥 − cos 𝑤 𝑥 + 2𝑤𝑥 sin 𝑤 𝑥]
𝐴
= 4𝑤2 (cos 𝑤 𝑥 − cos 𝑤 𝑥 + 2𝑤𝑥 sin 𝑤 𝑥)
𝐴
= 2𝑤 𝑥 sin 𝑤 𝑥.
Hence the general solution represented in the equation (46) becomes
𝐴
𝜓(𝑥) = 𝑐1 cos 𝑤 𝑥 + 𝑐2 sin 𝑤 𝑥 + 2𝑤 𝑥 sin 𝑤 𝑥, ∀ 𝑥 in 0 ≤ 𝑥 < ∞
1 𝜋
Since sin 𝑤𝑥 , cos 𝑤𝑥 are bounded functions, hence for 𝑥 = 𝑤 [2𝑛𝜋 + 2 ] , i. e. 𝑤𝑥 =
𝜋 𝐴 𝜋
[2𝑛𝜋 + 2 ], for 𝑛 ∈ ℕ 𝜓(𝑥) = 𝑐2 + 2𝑤2 (2𝑛𝜋 + 2 ) → ∞ as 𝑛 → ∞ i. e. as 𝑥 → ∞. This proves
|𝜓(𝑥)| assume arbitrary large value 𝑥 → ∞.

SELF-TEST 01
MCQ 01-02-01
The complete solution of the equation 𝑦 ′′ + 6𝑦 ′ + 9𝑦 = 5𝑒 3𝑥 is given by
5
A. 𝑦(𝑥) = (𝑐1 + 𝑐2 𝑥)𝑒 3𝑥 + 36 𝑒 3𝑥
5
B. 𝑦(𝑥) = (𝑐1 + 𝑐2 𝑥)𝑒 −3𝑥 + 36 𝑒 −3𝑥
5
C. 𝑦(𝑥) = (𝑐1 + 𝑐2 𝑥)𝑒 −3𝑥 + 36 𝑒 3𝑥
5
D. 𝑦(𝑥) = (𝑐1 + 𝑐2 𝑥)𝑒 −3𝑥 − 36 𝑒 3𝑥 .
MCQ 01-02-02
Two linearly independent solutions of the differential equation 𝑦 ′′ − 6𝑦 ′ + 25𝑦 = 0 are
A. 𝜙1 (𝑥) = 𝑒 3𝑥 𝑐𝑜𝑠 4𝑥 , 𝜙2 (𝑥) = 𝑒 3𝑥 sin 4𝑥
B. 𝜙1 (𝑥) = 𝑒 −3𝑥 , 𝜙2 (𝑥) = 𝑒 −4𝑥
C. 𝜙1 (𝑥) = 𝑒 4𝑥 𝑐𝑜𝑠 3𝑥 , 𝜙2 (𝑥) = 𝑒 4𝑥 sin 3𝑥
D. 𝜙1 (𝑥) = 𝑒 3𝑥 , 𝜙2 (𝑥) = 𝑒 4𝑥 .
MCQ 01-02-03
The value of Wronskian 𝑊(𝜙1 , 𝜙2 )(𝑥) of solutions of the differential equation 𝑦 ′′ + 𝑦 = 0
is
A. 𝑊(𝜙1 , 𝜙2 )(𝑥) = 0
B. 𝑊(𝜙1 , 𝜙2 )(𝑥) = 1
C. 𝑊(𝜙1 , 𝜙2 )(𝑥) = −1
D. 𝑊(𝜙1 , 𝜙2 )(𝑥) = 2.
MCQ 01-02-04
Which one of the followings is the particular integral / solution 𝜓𝑝 (𝑥) of the differential
equation 𝑦 ′′ − 4𝑦′ + 4𝑦 = 𝑥 3 𝑒 2𝑥 ?
𝑥3
A. 𝜓𝑝 (𝑥) = 20 𝑒 2𝑥

S25013: Ordinary Differential Equations Page 36


𝑥3
B. 𝜓𝑝 (𝑥) = − 20 𝑒 2𝑥
𝑥5
C. 𝜓𝑝 (𝑥) = 20 𝑒 2𝑥
𝑥5
D. 𝜓𝑝 (𝑥) = − 20 𝑒 2𝑥 .
MCQ 01-02-05
The Wronskian of the solutions of the equation 𝑦′′ − 21𝑦′ − 𝑦 = 𝑒 𝑖𝑥 − 2𝑒 −𝑖𝑥 at point 𝑥 =
0 ∈ 𝐼 is given by
A. 0
B. 1
C. 2
D. -2.
1 )

SHORT ANSWER QUESTIONS 01


SAQ 01- 02-01
Find all solutions of the equation 𝑦′′ + 4𝑦 = cos 𝑥.
Solution: This is of the form 𝐿(𝑦) = 𝑏(𝑥), where 𝑏(𝑥) = cos 𝑥. The characteristic equation
of 𝐿(𝑦) = 0 is given by 𝑃(𝑟) = 𝑟 2 + 4 = 0 . It has roots 𝑟 = ±2i. Hence, the functions
𝜙1 (𝑥) = 𝑒 2𝑖𝑥 , 𝜙2 (𝑥) = 𝑒 −2𝑖𝑥 are two linearly independent solutions of 𝐿(𝑦) = 0 . The
particular solution of 𝐿(𝑦) = 𝑏(𝑥) is given by
𝜓𝑝 (𝑥) = 𝑢1 𝜙1 (𝑥) + 𝑢2 𝜙2 (𝑥), (49)
where the function 𝑢1 (𝑥) and 𝑢2 (𝑥) are defined in the equation (33). The Wronskian
𝑊(𝜙1 , 𝜙2 )(𝑥) of the solutions of 𝐿(𝑦) = 0 is obtain as
2𝑖𝑥
𝑊(𝜙1 , 𝜙2 )(𝑥) = | 𝑒 2𝑖𝑥 𝑒 −2𝑖𝑥 | ⇒ 𝑊(𝜙 , 𝜙 )(𝑥) = −4𝑖, (50)
1 2
2𝑖𝑒 −2𝑖𝑒 −2𝑖𝑥
Hence from equation (33), we find
i 𝑥
𝑢1 (𝑥) = − 4 ∫0 cos 𝑡 . 𝑒 −2𝑖𝑡 𝑑𝑡. (51)
Integrating this by parts, we get after simplifying
i 1
𝑢1 (𝑥) = 12 𝑒 −2𝑖𝑥 (− 2𝑖 cos 𝑥 + sin 𝑥) − 6. (52)
Similarly, we find by using formula (33)
−i 1
𝑢2 (𝑥) = 12 𝑒 2𝑖𝑥 (sin 𝑥 + 2𝑖 cos 𝑥) − 6. (53)
Substituting these values in the given equation and simplifying we obtain
1 1
𝜓𝑝 (𝑥) = cos 𝑥 − (𝑒 2𝑖𝑥 + 𝑒 −2𝑖𝑥 )
3 6
Hence, the general solution of 𝐿(𝑦) = 𝑏(𝑥) represented in the equation (49) becomes
1
𝜓(𝑥) = 𝑐1 𝑒 2𝑖𝑥 + 𝑐2 𝑒 −2𝑖𝑥 + 3 cos 𝑥.
SAQ 01-02-02
Show that the set of functions {1, 𝑥, 𝑥 2 } is linearly independent.
Solution: In view of Theorem 1, we compute the Wronskian of the functions

S25013: Ordinary Differential Equations Page 37


𝑓1 𝑓2 𝑓3
𝑊(𝑓1 , 𝑓2 , 𝑓3 )(𝑥) = | 𝑓1′ 𝑓2′ 𝑓3′ |
𝑓1′′ 𝑓2′′ 𝑓3′′
1 𝑥 𝑥2
= |0 1 2𝑥 |
0 0 2
= 1(2 − 0) − 𝑥(0 − 0) + 𝑥 2 (0 − 0)
= 2.
Since, 𝑊(1, 𝑥, 𝑥 = 2 ≠ 0, then the set of functions {1, 𝑥, 𝑥 2 } is linearly independent.
2)

SAQ 01-02-03
If 𝜙1 (𝑥) = 𝑒 𝑥 is a fundamental solution to the differential equation 𝑦 ′′ − 2𝑦 ′ + 𝑦 = 0, then
find the fundamental solution 𝜙2 .
Solution: Since the given differential equation 𝑦 ′′ − 2𝑦 ′ + 𝑦 = 0 is already in the standard
form, we know that the Wronskian of fundamental solutions is given by
𝑊(𝜙1 , 𝜙2 )(𝑥) = 𝑐𝑒 −∫ 𝑎1 𝑑𝑥 = 𝑐𝑒 − ∫ −2𝑑𝑥 = 𝑐𝑒 2𝑥 ,
where we have used the fact that 𝜙1 (𝑥) = 𝑒 𝑥 and 𝜙1′ (𝑥) = 𝑒 𝑥 . Equating the two
expressions for the Wronskian, we have
𝑒 𝑥 𝜙2′ − 𝑒 𝑥 𝜙2 = 𝑐𝑒 2𝑥
⟹ 𝜙2′ − 𝜙2 = 𝑐𝑒 𝑥 . (54)
The equation (54) represents a first order linear differential equation for the second
fundamental solution 𝜙2 .
To solve this equation, we use a linear integrating factor which is given by I. F. = 𝑒 −𝑥 .
This allows us to write the equation (54) as
(I. F. 𝜙2 )′ = 𝑐𝑒 𝑥 . 𝑒 −𝑥 =c
⟹ (𝑒 −𝑥 𝜙2 )′ =c . (55)
Integrating the equation (55) and solving for 𝜙2 , we find
𝜙2 (𝑥) = 𝑐𝑥𝑒 𝑥 + 𝑑𝑒 𝑥 , (56)
where 𝑑 is the arbitrary constant. Since 𝑑 is multiplying our first fundamental solution
only, we can ignore this term. Thus, the second fundamental solution is 𝜙2 (𝑥) = 𝑥𝑒 𝑥 .
SAQ 01- 02-04
If 𝜙1 (𝑥), 𝜙2 (𝑥) are the solutions of the differential equation 𝑦 ′′ + 𝑎1 𝑦 ′ + 𝑎2 𝑦 = 0 with
1
constant coefficients in 𝐼 and 𝑊(2) = 0 , then find the value of the Wronskian of the
solutions of the differential equation at any point 𝑥 ∈ 𝐼.
Solution: According to the Theorem 3, we can write the formula for the Wronskian of
solutions of the differential equation with constant coefficients as
𝑥
− ∫𝑥 𝑎1 𝑑𝑥
𝑊(𝜙1 , 𝜙2 )(𝑥) = 𝑒 0 𝑊(𝜙1 , 𝜙2 )(𝑥0 ), ∀ 𝑥, 𝑥0 ∈ 𝐼.
1 1
Since 𝑊(2) = 0, we have for 𝑥0 = 2,
𝑥 𝑥
− ∫1 𝑎1 𝑑𝑥 1 − ∫1 𝑎1 𝑑𝑥
⟹ 𝑊(𝜙1 , 𝜙2 )(𝑥) = 𝑒 2 𝑊(𝜙1 , 𝜙2 ) (2) = 𝑒 2 .0 ∀ 𝑥 ∈ 𝐼
⟹ 𝑊(𝜙1 , 𝜙2 )(𝑥) = 0, ∀ 𝑥 ∈ 𝐼.

S25013: Ordinary Differential Equations Page 38


SUMMARY
In this Unit, we have defined linearly dependent and linearly independent functions of real
variables, Wronskian of functions and it is used to verify linearly dependent and linearly
independent of functions on given interval.
Further, we have derived the general solution of nonhomogen eous linear differential
equation 𝐿(𝑦) = 𝑦 ′′ + 𝑎1 𝑦′ + 𝑎2 𝑦 = 𝑏(𝑥), where 𝑎1 , 𝑎2 are constants in form
𝜓(𝑥) = 𝜙(𝑥) + 𝜓𝑝 (𝑥),
where 𝜙(𝑥) = 𝑐1 𝜙1 + 𝑐2 𝜙2 is the complementary function of 𝐿(𝑦) = 0 (𝜙1 , 𝜙2 are solutions
of 𝐿(𝑦) = 0) and 𝜓𝑝 is the particular solution of 𝐿(𝑦) = 𝑏(𝑥) and is given by
𝑥 [𝜙1 (𝑡)𝜙2 (𝑥)−𝜙1 (𝑥)𝜙2 (𝑡)]𝑏(𝑡)
𝜓𝑝 (𝑥) = ∫𝑥 𝑑𝑡.
0 𝑊(𝜙1 ,𝜙2 )(𝑡)

END OF UNIT EXERCISES


1) State the following sets of functions defined on −∞ < 𝑥 < ∞ are linearly independent
or linearly dependent:
(i) 𝜙1 (𝑥) = 𝑥, 𝜙2 (𝑥) = |𝑥|
(ii) 𝜙1 (𝑥) = 𝑒 𝑖𝑥 , 𝜙2 (𝑥) = sin 𝑥 , 𝜙3 (𝑥) = 2 cos 𝑥
(iii) 𝜙1 (𝑥) = sin 𝑥 , 𝜙2 (𝑥) = 𝑒 𝑖𝑥
(iv) 𝜙1 (𝑥) = 𝑥, 𝜙2 (𝑥) = 𝑒 2𝑥 , 𝜙3 (𝑥) = |𝑥|
(v) 𝜙1 (𝑥) = 𝑒 𝑥 , 𝜙2 (𝑥) = 𝑒 2𝑥 , 𝜙3 (𝑥) = 𝑒 3𝑥 .
2) Let 𝜙1 (𝑥) and 𝜙2 (𝑥) be two linearly independent solutions of the equation 𝑦′′ + 𝑎2 𝑦 =
0 with constant coefficients. Then prove the Wronskian of 𝜙1 and 𝜙2 at any point in
the interval 𝐼 is constant for all 𝑥 ∈ 𝐼.
3) Let 𝜙1 (𝑥) = 𝑥 2 , 𝜙2 (𝑥) = 𝑥|𝑥| be two linearly independent functions of 𝑥 ∈ 𝐼: −∞ < 𝑥 <
∞. Then prove that 𝑊(𝜙1 , 𝜙2 )(𝑥) = 0 ∀ 𝑥 ∈ 𝐼 and hence 𝑊(𝜙1 , 𝜙2 )(𝑥) =
𝑊(𝜙1 , 𝜙2 )(𝑥0 ).
4) If the Wronskian of the functions 𝜙1 (𝑥), 𝜙2 (𝑥) at 𝑥 = 𝑥0 in an interval 𝐼 is zero, then
show that the systems of the equations
𝑐1 𝜙1 (𝑥0 ) + 𝑐2 𝜙2 (𝑥0 ) = 0,
𝑐1 ϕ1′ (𝑥0 ) + 𝑐2 ϕ′2 (𝑥0 ) = 0,
has infinitely many solutions.

KEY WORDS
Linearly dependent, linearly independent, Wronskian, general solution, particular solution,
nonhomogeneous differential equation.

S25013: Ordinary Differential Equations Page 39


UNIT 01-03: APPLICATIONS OF SECOND ORDER LINEAR EQUATIONS
LEARNING OBJECTIVES
After successful completion of this unit, you will be able to
❖ explain mass-spring system, electric circuit problems
❖ apply Newton’s second law of motion, Hooke’s law, Kirchhoff ’s voltage laws to form
second order ODEs and obtain their solutions

03-01: THE VIBRATIONS OF A MASS ON A SPRING (THE BASIC PROBLEM):


Differential equations having a great variety of important applications. In particular,
second-order linear differential equations with constant coefficients have numerous
applications in physics and in electrical and mechanical engineering .
Let us formulate the problem systematically. Let the coil spring have natural (unstreched)
length 𝐿. The mass 𝑚 is attached to its lower end and comes to rest in its equillibrium
position, thereby stretching the spring an amount 𝑙 so that its streched length is 𝐿 + 𝑙. We
choose the axis along the line of the spring, with the origin 𝑂 at the equillibrium position
and the positive direction downward. Thus, letting 𝑥 denote the displacement of the mass
from 𝑂 along this line, we see that 𝑥 is positive, zero or negative according to whether the
mass is below, at or above its equillibrium position (see figures).

S25013: Ordinary Differential Equations Page 40


In order to set up the differential equation for this problem we shall need two laws of
physics Newton’s second law and Hooke’s law.
Hooke’s Law:
The magnitude of the force needed to produce a certain elongation of a spring is directly
prortional to the amount of this elongation, provided this elongation is not too great .
Hence, this can be fomulated mathematically as
|𝐹| = 𝛼𝑒, (1)
where 𝐹 is the magnitude of the force, 𝑒 is the amount of elongation, and 𝛼 is a constant
of proportionality, and is called spring constant.
Newton’s Second Law:
When a force 𝐹 acts upon a body, the acceleration 𝑎 is produced in the direction of the
force whose magnitude is proportional to the magnitude of force. Therefore, we have
𝐹 = 𝑚𝑎, (2)
where 𝑚 is constant of proportionality and it represents mass of the body.

03-02: FORCE ACTING UPON THE MASS


In this section, we derive the DE describing the position of a mass on a spring as a function
of time. We consider the force of gravity, the restoring force of the spring, the resisting
force of the medium, and other external forces. Forces tending to pull the mass downward
are postive, while those tending to pull it upward are negative. If the object is located at
𝑥, then the forces are
1. 𝐹1 , the force of gravity, of magnitude 𝑚𝑔, where 𝑔 is the acceleration due to gravity.
Since this acts in the downward direction, it is positive, and so
𝐹1 = 𝑚𝑔 (3)
2. 𝐹2 , the restoring force of the spring. Since 𝑥 + 𝑙 is the total amount of elongation by
Hooke’s law the magnitude of this force is 𝑘(𝑥 + 𝑙). When the mass is below the end of
the unstreched spring, this force acts in the upward direction and so is negative.
Therefore, one can have
𝐹2 = −𝑘(𝑥 + 𝑙) (4)
This also gives the restoring force when the mass is above the end of the unstreched
spring, as one can see by replacing each italicized word in the three preceding sentences
by its opposite. When the mass is at rest in its equillibrium position the restoring force
𝐹2 is equal in magnitude but opposite in direction to the force of gravity and so is given
by −𝑚𝑔. Since in the position 𝑥 = 0, equation (4) gives −𝑚𝑔 = −𝑘(0 + 𝑙) or 𝑚𝑔 = 𝑘𝑙.
Replacing 𝑘𝑙 by 𝑚𝑔 in equation (4) we see that the restoring force can thus be written
as
𝐹2 = −𝑘𝑥 − 𝑚𝑔. (5)
3. 𝐹3 , the resisting force of the medium, called the damping force. Although the magnitude
of this not known exactly, it is known that for small velocities it is approximately
proportional to the magnitude of the velocity:

S25013: Ordinary Differential Equations Page 41


𝑑𝑥
|𝐹3 | = 𝑎 | |, where 𝑎 > 0 is called the damping constant. When the mass is moving
𝑑𝑡
downward, 𝐹3 acts in the upward direction (opposite to that of the motion) and so 𝐹3 <
𝑑𝑥
0. Also, since 𝑚 is moving downward, 𝑥 is increasing and is positive. Thus, assuming
𝑑𝑡
equation (5) to hold, when the mass is moving downward, the damping force is given
by
𝑑𝑥
𝐹3 = −𝑎 𝑑𝑡 (𝑎 > 0). (6)
4. 𝐹4 , any external impressed foreces that act upon the mass. Let us denote the resultant of
all such external forces at time 𝑡 simply by 𝐹(𝑡) and write
𝐹4 = 𝐹(𝑡). (7)
We now apply Newton’s second law, 𝐹 = 𝑚𝑎 , where 𝐹 = 𝐹1 + 𝐹2 + 𝐹3 + 𝐹4 and
𝑎 =acceleration. Using (3), (5), (6) and (7), we find
𝑑2 𝑥 𝑑𝑥
𝑚 𝑑𝑡 2 = 𝑚𝑔 − 𝑘𝑥 − 𝑚𝑔 − 𝑎 𝑑𝑡 + 𝐹(𝑡)
or
𝑑2 𝑥 𝑑𝑥
𝑚 𝑑𝑡 2 + 𝑎 𝑑𝑡 + 𝑘𝑥 = 𝐹(𝑡). (8)
This we take as the differential equation for the motion of the mass on the spring.

Note:
1) From the equation (8), we observe that it is a nonhomogeneous second-order linear
differential equation with constant coefficients.
2) If 𝑎 = 0 the motion is called undamped otherwise it is called damped.
3) If there are no external impressed forces, 𝐹(𝑡) = 0 for all 𝑡 and the motion is called
free otherwise it is called forces.
4) In the following sections, we consider the solution of (8) in each of these cases.

03-03: FREE, UNDAMPED MOTION


We now consider the special case of free, undamped motion, that is, the case in which both
𝑎 = 0 and 𝐹(𝑡) = 0 for all 𝑡. The differential equation (8), then reduces to
𝑑2 𝑥
𝑚 𝑑𝑡 2 + 𝑘𝑥 = 0, (9)
where 𝑚(> 0) is the mass and 𝑘(> 0) is the spring constant.
It is convenient to rearrange this equation and introduce a new variable, called the angular
𝑘
frequency, 𝜆. Letting 𝜆 = √𝑚, we can write the equation (9) as follows
𝑑2 𝑥
+ 𝜆2 𝑥 = 0. (10)
𝑑𝑡 2
This is of second order homogeneous differential equation and hence has the general
solution
𝑥(𝑡) = 𝑐1 sin𝜆𝑡 + 𝑐2 cos𝜆𝑡, (11)
where 𝑐1 and 𝑐2 are arbitrary constants. The equation (11) gives the position of the mass at
any point in time. The motion of the mass is called simple harmonic motion. The period of
this motion (the time it takes to complete one oscillation) is

S25013: Ordinary Differential Equations Page 42


2𝜋
𝑇= 𝜆
and the frequency is given by
1 𝜆
𝑓 = 𝑇 = 2𝜋.
Let us now assume that the mass was initially displaced a distance 𝑥0 from its equillibrium
position and released from that point with initial velocity 𝜈0 . Then, in addition to the
differential equation (10), we have initial conditions
𝑥(0) = 𝑥0 ,
} (12)
𝑥 ′ (0) = 𝜈0 .
Using initial conditions (12) and (11), we see at once that
𝑐2 = 𝑥0 ,
𝑐1 = 𝜆0 .}
𝜈 (13)

Substituting the values of 𝑐1 and 𝑐2 in (11) gives the particular solution of the differential
equation (10) in the form
𝜈0
𝑥(𝑡) = sin𝜆𝑡 + 𝑥0 cos𝜆𝑡. (14)
𝜆
We put this in an alternative form by first writing it as
𝜈0 /𝜆 𝑥0
𝑥(𝑡) = [ 𝑐
sin𝜆𝑡 + 𝑐
cos𝜆𝑡], (15)
where
𝜈 2
𝑐 = √( 𝜆0) + 𝑥02 > 0. (16)
𝜈0
𝑥0
Then, letting 𝜆
= − sin 𝜙 and = cos 𝜙, the equation (14) reduces to
𝑐 𝑐
𝑥(𝑡) = 𝑐 cos(𝜆𝑡 + 𝜙), (17)
𝑘
Since 𝜆 = √𝑚 , we now write the solution (7) in the form
𝑘
𝑥(𝑡) = 𝑐 cos (√𝑚 𝑡 + 𝜙). (18)
This, then, gives the displacement 𝑥 of the mass from the equillibrium position 𝑂 as a
function of the time 𝑡 (𝑡 > 0). We see at once that the free, undamped motion of the mass
is a simple harmonic motion. The constant 𝑐 is called the amplitude of the motion and gives
the maximum (positive) displacement of the mass from its equillibrium position. The
number 𝜙 is called the phase constant (or phase angle).
The motion is a periodic motion, and the mass oscillates back and forth between 𝑥 = 𝑐 and
𝑥 = −𝑐. We have 𝑥 = 𝑐 if and only if
𝑘
√ 𝑡 + 𝜙 = ±2𝑛𝜋,
𝑚

𝑛 = 0,1,2,3, ⋯ , 𝑡 > 0. Thus the maximum (positive) displacement occurs if and only if
𝑚
𝑡 = √ 𝑘 (±2𝑛𝜋 − 𝜙) > 0, (19)

where 𝑛 = 0,1,2,3, ⋯.

S25013: Ordinary Differential Equations Page 43


03-04: FREE, DAMPED MOTION
We now consider the effect of the resistance of the medium upon the mass on the spring.
Still assuming that ni external focus are present, this is then the case of free, damped
motion. Hence with the damping coefficient 𝑎 > 0 and 𝐹(𝑡) = 0 for all 𝑡 , the basic
differential equation (8) reduces to
𝑑2 𝑥 𝑑𝑥
𝑚 𝑑𝑡 2 + 𝑎 𝑑𝑡 + 𝑘𝑥 = 0 (20)
𝑘
Dividing through by 𝑚 and putting 𝜆 = √𝑚 and 𝑎/𝑚 = 2𝑏 (for convenience), we have the
differential equation (20) in the form
𝑑2 𝑥 𝑑𝑥
+ 2𝑏 𝑑𝑡 + 𝜆2 𝑥 = 0. (21)
𝑑𝑡 2
Observe that since 𝑎 is positive, 𝑏 is also positive. We find roots of the corresponding
auxiliary equation are
−2𝑏±√4𝑏 2 −4𝜆2
= −𝑏 ± √𝑏 2 − 𝜆2 . (22)
2
Three distinct cases occur, depending upon the nature of these roots, which in turn depends
upon the sign of 𝑏 2 − 𝜆2 .
Case 1: Damped Oscillatory Motion. Here we consider the case in which 𝑏 < 𝜆, which
implies that 𝑏 2 − 𝜆2 < 0 . Then the roots are the conjugate complex numbers −𝑏 ±
𝑖√𝜆2 − 𝑏 2 and the general solution of the equation (20) or (21) is therefore,
𝑥(𝑡) = 𝑒 −𝑏𝑡 (𝑐𝑡 sin√𝜆2 − 𝑏 2 𝑡 + 𝑐2 cos√𝜆2 − 𝑏 2 𝑡), (23)
where 𝑐1 and 𝑐2 are arbitrary constants.
Case 2: Critical Damping. This is the case in which 𝑏 = 𝜆, which implies that 𝑏 2 − 𝜆2 =
0. The both roots are equal to the real negative number −𝑏, and hence the general solution
of the equation (20) or (21) is
𝑥(𝑡) = (𝑐1 + 𝑐2 𝑡)𝑒 −𝑏𝑡 . (24)
The motion is no longer oscillatory, the damping is just great enough to prevent
oscillations.
Case 3: Overcritical Damping. Finally, we consider here the case which 𝑏 > 𝜆 which
implies that 𝑏 2 − 𝜆2 > 0 . Here the roots are the distinct real negative numbers and
therefore, the general solution of (20) or (21) is
𝑥(𝑡) = 𝑐1 𝑒 𝑟1𝑡 + 𝑐2 𝑒 𝑟2𝑡 . (25)
The damping is now so great that no oscillations can occur. Further, we can longer say that
every decrease in the amount of damping will result in oscillations, as we could in Case 2.
The motion here is said to be overcritically damped (or simply overdamped).

03-05: ELECTRIC CIRCUIT PROBLEMS


In this section, we consider the application of differential equations to series circuits
containing (1) an elextromotive force, and (2) resistors, indicators, and capacitors. We
assume that the reader is somewhat familiar with these items and so we shall avoid an

S25013: Ordinary Differential Equations Page 44


extensive discussion. Let us simply recall that the electromotive force (for example, a
battery generator) produces a flow of current in a closed circuit and that this current
produces a so-called voltage drop across each resistor, inductor and capacitor. Further, the
following three laws concerning the voltage drops across these various elements are known
to hold:
1. The voltage drop across a resistor is given by
𝐸𝑅 = 𝑅𝑖 , (26)
where 𝑅 is a constant of proportionality called the resistance, and 𝑖 is the current.
2. The voltage drop across an inductor is given by
𝑑𝑖
𝐸𝐿 = 𝐿 𝑑𝑡, (27)
where 𝐿 is a consistent of proportionality called the inductance, and 𝑖 again denotes the
current.
3. The voltage drop across a capacitor is given by
1
𝐸𝐶 = 𝐶 𝑞, (28)
where 𝐶 is a constant of proportionality called the capacitance and 𝑞 is the
instantaneous charge on the capacitor. Since 𝑖 = 𝑑𝑞/𝑑𝑡, this is often written as
1
𝐸𝐶 = 𝐶 ∫ 𝑖𝑑𝑡. (29)
The fundamental law in the study of electric circuits in the following:
Kirchhoff ’s Voltage Law (Form 1). The algebraic sum of the instantaneous volatge drops
arounds a close circuit in a specific direction is zero.
Since voltage drop across resistor, inductors, and capacitors have the opposite sign from
voltages arising from electromotive forces, we may state this law in the following
alternative form:
Kirchhoff ’s Voltage Law (Form 2). The sum of the voltage drops across resistors, inductors,
and capacitors is equal to the total electromotive force in a closed circuit. We now consider
the cicuit shown in figure:

Let us apply Kirchhoff’s law to the circuit of the figure. Letting 𝐸 denote the electromotive
force, and using the laws 1,2, and 3 for voltage drops that were given above, we are let at
once to the equation
𝑑𝑖 1
𝐿 𝑑𝑡 + 𝑅𝑖 + 𝐶 𝑞 = 𝐸. (30)

S25013: Ordinary Differential Equations Page 45


This equation contains two dependent variables 𝑖 and 𝑞. However, we recall that these two
𝑑𝑞
variables are related to each other by the equation 𝑖 = .
𝑑𝑡
Using this, we may eliminate 𝑖 from the equation (30) and write it in the form
𝑑2 𝑞 𝑑𝑞 1
𝐿 𝑑𝑡 2 + 𝑅 𝑑𝑡 + 𝐶 𝑞 = 𝐸. (31)
The equation (31) is a second-order linear differential equation in the single dependent
variable 𝑞. On the other hand, if we differetiate the equation (30) with respect to 𝑡 and
𝑑𝑞
make use of 𝑖 = , we may eliminate 𝑞 from Equation (30) and write
𝑑𝑡
𝑑2 𝑖 𝑑𝑖 1 𝑑𝐸
𝐿 𝑑𝑡 2 + 𝑅 𝑑𝑡 + 𝐶 𝑖 = . (32)
𝑑𝑡
This is a second-order linear differential equation in the single dependent variable 𝑖.
Thus we have the two second-order linear differential equations (31) and (32) for the
charge 𝑞 and current 𝑖 , respectively. Further observe that in two very simple cases the
problem, the equation (30) itself reduces directly to
𝑑𝑖
𝐿 𝑑𝑡 + 𝑅𝑖 = 𝐸; (33)
while if no inductor is present, the equation (5.) reduces to
𝑑𝑞 1
𝑅 𝑑𝑡 + 𝐶 𝑞 = 𝐸. (34)
The differential equation (31) for the charge is exactly the same as the differential equation
(8) of previous section for the vibrtations of a mass on a coil spring, except for the
notations used. That is, the electrical system is analogous to the mechanical system .

SOLVED PROBLEMS 01
Problem 01-03-01:
An 8 − 𝑙𝑏 weight is placed upon the lower end of a coil spring suspended from the ceiling.
The weight comes to rest in its equillibrium position, thereby stretching the spring 6 𝑖𝑛.
The weight is then pulled down 3 𝑖𝑛. below its equillibrium position and released at 𝑡 = 0
with an initial velocity of 1 𝑓𝑡/𝑠𝑒𝑐, directed downward. Neglecting the resistance of the
medium and assuming that no external forces are present, determine the amplitude, period
and frequency of the resulting motion.
Solution: This is clearly an example of free, undamped. Since the 8 − 𝑙𝑏 weight stretches
1 1
the spring 6 𝑖𝑛. = 2 𝑓𝑡. Hooke’s law 𝐹 = 𝑘𝑠 gives 8 = 𝑘 (2) and so 𝑘 = 16 lb/ft. Also, 𝑚 =
8
𝑤/𝑔 = 32(slugs) and so the equation (8) gives
8 𝑑2 𝑥
+ 16𝑥 = 0
32 𝑑𝑡 2
or
𝑑2 𝑥
+ 64𝑥 = 0. (35)
𝑑𝑡 2
Since the weight was released with a downward initial velocity of 1 𝑓𝑡/𝑠𝑒𝑐 from a point
1
3 𝑖𝑛. (= 4 𝑓𝑡) below its equillibrium position, we also have the initial conditions
1
𝑥(0) = 4 , 𝑥′(0) = 1 (36)

S25013: Ordinary Differential Equations Page 46


Thus, the general solution of the differential equation ( 35) is given by
𝑥(𝑡) = 𝑐1 sin 8𝑡 + 𝑐2 cos 8𝑡, (37)
1
where 𝑐1 and 𝑐2 are arbitrary comstants. Applying conditions given in (36) , we find 𝑐2 = 4
1
and 𝑐1 = 8 . Therefore, the solution of the differential equation (35) satisfying the
conditions (36) is
1 1
𝑥(𝑡) = 8 sin 8𝑡 + 4 cos 8𝑡. (38)
Let us put this in the form (18), we find
√5 √5 2√5
𝑥= ( 5 sin 8𝑡 + cos 8𝑡) (39)
8 5
2√5 √5
Thus letting cos 𝜙 = , sin 𝜙 = − 5 , we write the solution in the form
5
√5
𝑥(𝑡) = cos(8𝑡 + 𝜙), (40)
8
√5
where 𝜙 is determined by sin 𝜙 = − . From these equations we find that 𝜙 = −0.46
5
radians. Taking √5 ≈ 2.236, the solution (35) is thus given approximately by
𝑥(𝑡) = 0.280cos(8𝑡 − 0.46). (41)
2𝜋
The ammplitude of the motion √5/8 ≈ 0.280(ft.). Also, the period is = 𝜋/4(sec), and
8
the frequency is 4/𝜋 oscillations/sec.
Problem 01-03-02:
A circuit has in series an electromotive force given by 𝐸 = 100sin40𝑡 𝑉, a resistor of 10Ω
and an inductor of 0.5 𝐻. If the initial current is 0, find the current at time 𝑡 > 0.
Solution: Let 𝑖 denote the current in amperes at time 𝑖. The total electromotive force is
100sin40𝑡 𝑉. Using the laws 1 and 2, we find the voltage drop are as follows:
1. Across the resistor: 𝐸𝑅 = 𝑅𝑖 = 10𝑖.
𝑑𝑖 1 𝑑𝑖
2. Across the inductor: 𝐸𝐿 = 𝐿 𝑑𝑡 = 2 𝑑𝑡.
Applying Kirchhoff’s law, have the differential equation
1 𝑑𝑖
+ 10𝑖 = 100 sin 40𝑡, (42)
2 𝑑𝑡
or
𝑑𝑖
+ 20𝑖 = 200 sin 40𝑡. (43)
𝑑𝑡
Since the initial current is 0, the intial condition is 𝑖(0) = 0. Thus the solution is
𝑖(𝑡) = 2(sin40𝑡 − 2cos40𝑡) + 4𝑒 −20𝑡 . (44)
Problem 01-03-03:
Find the charge on the capacitor in an RLC series circuit where 𝐿 = 5/3 𝐻, 𝑅 = 10𝛺, 𝐶 =
1
𝐹, and 𝐸(𝑡) = 300 𝑉. Assume the initial charge on the capacitor is 0 𝐶 and the initial
30
current is 9 𝐴. What happens to the charge on the capacitor over time?
Solution: The given RLC series circuit can be modelled as
𝑑2 𝑞 𝑑𝑞
+ 6 𝑑𝑡 + 18𝑞 = 180, (45)
𝑑𝑡 2
𝑑𝑞
with initial conditions: 𝑞(0) = 0 and 𝑖(0) = ( 𝑑𝑡 ) = 9. On solving the general solution is
𝑡=0
obtained as

S25013: Ordinary Differential Equations Page 47


𝑞(𝑡) = 𝑒 −3𝑡 (𝑐1 cos(3𝑡) + 𝑐2 sin(3𝑡)) + 10. (46)
Now we apply the conditions, we find 𝑐1 = −10 and 𝑐2 = −7 . So, the charge on the
capacitor is
𝑞(𝑡) = −10𝑒 −3𝑡 cos(3𝑡) − 7𝑒 −3𝑡 sin(3𝑡) + 10. (47)
By looking at first two terms of the equation (47), one will observe that there is decay over
time (as a result of the negative exponent in the exponential function. Therefore, the
capacitor eventually approaches a steady-state charge of 10 𝐶.

SELF-TEST 01
MCQ 01-03-01
Assume an object weighing 2 𝑙𝑏 stretches a spring 6 𝑖𝑛ch. If the spring is released from the
equilibrium position with an upward velocity of 16 𝑓𝑡/𝑠𝑒𝑐, then what is the period of the
motion?
4
A. second
𝜋
𝜋
B. second
4

C. second
3
3
D. second.
2𝜋
MCQ 01-03-02
A 200 − 𝑔 mass stretches a spring 5 𝑐𝑚. Suppose mass released from rest from a position
10 𝑐𝑚 below the equilibrium position. What is the frequency of this motion?
7
A. Hz
π
14
B. Hz
π
𝜋
C. Hz
7
𝜋
D. Hz.
14
MCQ 01-03-03
Suppose 𝑥 (𝑡 ) = 2 cos(3𝑡) + sin(3𝑡) represents the solution of simple harmonic motion.
The amplitude is given by
A. 5
B. √3
C. √5
D. 3..
MCQ 01-03-04
𝑑2 𝑥 𝑑𝑥
Consider spring-mass system model + 2𝑏 𝑑𝑡 + 𝜆2 𝑥 = 0, where damping coefficient 𝑎 >
𝑑𝑡 2
𝑘 𝑎
0 and mass 𝑚 with 𝜆 = √𝑚 and = 2𝑏. Then the system is overcritically damped if .
𝑚

A. 𝑏 2 − 𝜆2 > 0
B. 𝑏 2 − 𝜆2 < 0

S25013: Ordinary Differential Equations Page 48


C. 𝑏 2 − 𝜆2 = 0
D. 𝜆2 − 𝑏 2 > 0.
MCQ 01-03-05
In RLC series circuit where 𝐿 = 1/5 𝐻, 𝑅 = 2/5𝛺, 𝐶 = 1/2 𝐹, and 𝐸(𝑡) = 50 𝑉. Assume the
initial charge on the capacitor is 0 𝐶 and the initial current is 4 𝐴. Then the charge 𝑞(𝑡) on
the capacitor is
A. 𝑞(𝑡) = −25𝑒 −𝑡 cos(3𝑡) − 7𝑒 −𝑡 sin(3𝑡) + 25
B. 𝑞(𝑡) = 25𝑒 −𝑡 cos(3𝑡) − 7𝑒 −𝑡 sin(3𝑡) + 25
C. 𝑞(𝑡) = −25𝑒 −𝑡 cos(3𝑡) + 7𝑒 −𝑡 sin(3𝑡) + 25
D. 𝑞(𝑡) = −25𝑒 −𝑡 cos(3𝑡) − 7𝑒 −𝑡 sin(3𝑡) − 25.

SHORT ANSWER QUESTIONS 01


SAQ 01- 03-01
𝑑2 𝑥 𝑑𝑥 1
Consider the equation + 4 𝑑𝑡 + 16𝑥 = 0 , with initial conditions 𝑥(0) = 2 , 𝑥′(0) = 0
𝑑𝑡 2
describing the motion of the weight on the spring. Determine the resulting motion of the
weight on the spring.
Solution: The given equation is
𝑑2 𝑥 𝑑𝑥
+ 4 𝑑𝑡 + 16𝑥 = 0, (48)
𝑑𝑡 2
1
with initial conditions 𝑥(0) = 2, 𝑥′(0) = 0. Thus the general solution of (48) may be written
𝑥(𝑡) = 𝑒 −2𝑡 (𝑐1 sin 2√3𝑡 + 𝑐2 cos 2√3𝑡), (49)
where 𝑐1 and 𝑐2 are arbitrary constants. Applying the initial conditions, we obtain
√3 1
𝑐1 = , 𝑐2 =2
6
and the solution is
√3 1
𝑥(𝑡) = 𝑒 −2𝑡 ( 6 sin 2√3𝑡 + 2 cos 2√3𝑡). (50)
This is the motion of the weight on the spring.
SAQ 01- 03-02
A circuit has in series an electromotive force given by 𝐸 = 100sin60𝑡 𝑉, a resistor of 2Ω,
1
an inductor of 0.1𝐻, and a capacitor of the farads. If the initial current and the initial
240
charge on the capacitor are both zero, find the charge on the capacitor at any time 𝑡 > 0.
Solution: Let 𝑖 denote the current and 𝑞 the charge on the capacitor at time 𝑡. The total
electromotive force is 100sin60𝑡(volts). Using the voltage drop laws 1,2, and 3 we find
that the voltage drops are as follows:
1. Across the resistor: 𝐸𝑅 = 𝑅𝑖 = 2𝑖.
𝑑𝑖 1 𝑑𝑖
2. Across the induction: 𝐸𝐿 = 𝐾 𝑑𝑡 = 10 𝑑𝑡.
1
3. Across the capacitor: 𝐸𝑐 = 𝐶 𝑞 = 260𝑞.
Now applying Kirchhoff’s law we have at once:
1 𝑑𝑖
+ 2𝑖 + 260𝑞 = 100 sin 60𝑡. (51)
10 𝑑𝑡

S25013: Ordinary Differential Equations Page 49


Since 𝑖 = 𝑑𝑞/𝑑𝑡, this reduces to
1 𝑑2 𝑞 𝑑𝑞
+ 2 𝑑𝑡 + 260𝑞 = 100 sin 60𝑡. (52)
10 𝑑𝑡 2
Since the charge 𝑞 is initially zero, we have as a first initial condition
𝑞(0) = 0 (53)
Since the current 𝑖 is also initially zero and 𝑖 = 𝑑𝑞/𝑑𝑡, we take the second initial condition
in the form
𝑞′(0) = 0. (54)
Therefore, the general solution of the equation (52) is
25 30
𝑞(𝑡) = 𝑒 −10𝑡 (𝑐1 sin 50𝑡 + 𝑐2 cos 50𝑡) − 61 sin 60𝑡 − 61 cos 60𝑡. (55)
Applying conditions, we have
36 30
𝑐1 = 61 and 𝑐2 = 61. (56)
Thus the solution of the problem is
6𝑒 −10𝑡 5
𝑞(𝑡) = (6 sin 50𝑡 + 5 cos 50𝑡) − 61 (5 sin 60𝑡 + 6 cos 60𝑡). (57)
61

SUMMARY
In this Unit, the physical applications based on second order ODEs are considered.
Mass-spring system is an oscillation system where a mass 𝑚 is attached at loer end of a
vertical spring of natural length 𝑙 the mass is pulled down a certain distance and released
it. It undergoes motion which take it strictly vertical and leads to three types of possible
motions: free motion, damped motion, and forced motion.
Equation of motion of a particle is given by Newton’s secon d law of motion 𝑚𝑥 ′′ = 𝐹 or
𝑑𝑣
𝑚 𝑑𝑡 = 𝐹. Hence for given force 𝐹, one can integrate above equation and obtain 𝑥 in terms
of 𝑡.
Electric circuit: The second order ODEs governing the LCR electric circuit are:
𝑑2 𝐼 𝑑𝐼 𝐼 𝑑𝐸
𝐿 𝑑𝑡 2 + 𝑅 𝑑𝑡 + 𝐶 = 𝑑𝑡
and
𝑑2 𝑄 𝑑𝑄 𝑄
𝐿 𝑑𝑡 2 + 𝑅 𝑑𝑡 + 𝐶 = 𝐸.
The quantities 𝐿, 𝐶, 𝑅 are assumed to be constants.

END OF UNIT EXERCISES

1) A 32 − 𝑙𝑏 weight is attached to the lower end of a coil spring suspended from the
ceiling. The weight comes to rest in its equillibrium position, thereby stretching the
spring 2 𝑓𝑡 . The weight is then ulled down 6 𝑖𝑛 below its equilibrium position and
released at 𝑡 = 0. No external forces are present but the resistance of the medium in
𝑑𝑥 𝑑𝑥
pounds is numerically equal to 4( 𝑑𝑡 ), where is the instantaneous velocity in feet per
𝑑𝑡
second. Determine the resulting motion of the weight on the spring.

S25013: Ordinary Differential Equations Page 50


2) A 16 − 𝑙𝑏 weight is attached to the lower end of a coil spring suspended from the
ceiling, the spring constant of the spring being 10 𝑙𝑏/𝑓𝑡. The weight comes to rest in
its equillibrium position. Beginning at 𝑡 = 0 an external force given by 𝐹(𝑡) = 5 cos 2𝑡
is applied to the system. Determine the resulting motion if the damping force in pounds
𝑑𝑥 𝑑𝑥
is numerically equal to 2( 𝑑𝑡 ), where is the instantaneous velocity in feet per second.
𝑑𝑡

KEY WORDS
Spring, equilibrium, resisting force, Hooke’s law, frequency, amplitude, period, undamped
motion, damped motion, forced motion, resister, inductor, capacitor, voltage, Kirchhoff’s
voltage laws, resonance phenomena, electric circuit.

S25013: Ordinary Differential Equations Page 51


UNIT 01-04: THE HOMOGENEOUS EQUATION OF HIGHER ORDER
LEARNING OBJECTIVES
After successful completion of this unit, you will be able to
❖ explain homogeneous linear differential equation with constant coefficients of order 𝑛 ≥ 2
❖ apply the method of solution by using roots of characteristic polynomial
04-01: HOMOGENEOUS EQUATION OF ORDER 𝒏
The 𝑛th order homogeneous equation with constants coefficients is given by
𝐿(𝑦) = 𝑦 (𝑛) + 𝑎1 𝑦 (𝑛−1) 𝑎2 𝑦 (𝑛−2) + ⋯ + 𝑎𝑛 𝑦 = 0, (1)
where 𝑎1 , 𝑎2 , ⋯ , 𝑎𝑛 are constants.
In the previous units, we have proved some theorems regarding with the second order
equation with constant coefficients and solved some examples. The procedure adopted in
solving the second order homogeneous equation with constant coefficie nts can be carried
out to the case of equation (1)
As before we assume that 𝑒 𝑟𝑥 is the solution of (1), then we have
𝐿(𝑒 𝑟𝑥 ) = 𝑝(𝑟)𝑒 𝑟𝑥 , (2)
where 𝑛
𝑝(𝑟) = 𝑟 + 𝑎1 𝑟 𝑛−1
+ 𝑎2 𝑟 𝑛−2
+ ⋯ + 𝑎𝑛 (3)
is the characteristic polynomial of 𝐿(𝑦) = 0 . If 𝑟 = 𝑟1 is the root of characteristic
polynomial 𝑝(𝑟), then we have 𝑝(𝑟1 ) = 0 and consequently 𝐿(𝑒 𝑟1𝑥 ) = 0 and 𝜙(𝑥) = 𝑒 𝑟1𝑥 is
a solution of 𝐿(𝑦) = 0.
If 𝑟1 is a root of multiplicity 𝑚1 , then we have
𝑝(𝑟1 ) = 0, 𝑝′ (𝑟1 ) = 0, ⋯ , 𝑝(𝑚1−1) (𝑟1 ) = 0 and 𝑝(𝑚) (𝑟1 ) ≠ 0.
Then, we have, each 𝑥 𝑘 𝑒 𝑟1𝑥 , 𝑘 = 0,1,2, ⋯ , 𝑚1 − 1 is a solution of 𝐿(𝑦) = 0.
We prove the result in detail in the following theorem.
Theorem 1: Let 𝑟1 , 𝑟2 , ⋯ , 𝑟𝑠 be the distinct roots of the characteristic polynomial 𝑝(𝑟) =
𝑟 𝑛 + 𝑎1 𝑟 𝑛−1 + 𝑎2 𝑟 𝑛−2 + ⋯ + 𝑎𝑛 and suppose 𝑟𝑖 , 𝑖 = 1,2, ⋯ , 𝑠 has multiplicity 𝑚𝑖 (𝑚1 +
𝑚2 + ⋯ + 𝑚𝑠 = 𝑛). Then the 𝑛 functions
𝑥 𝑘 𝑒 𝑟𝑖 𝑥 , 𝑘 = 0,1,2, ⋯ , 𝑚𝑖−1 , 𝑖 = 1,2, ⋯ , 𝑠
are solutions of 𝐿(𝑦) = 0.
Proof: Consider the nth order with constant coefficients given by
𝐿 (𝑦) = 𝑦 (𝑛) + 𝑎1 𝑦 (𝑛−1) + 𝑎2 𝑦 (𝑛−2) + ⋯ + 𝑎𝑛 𝑦 = 0
The characteristic polynomial of 𝐿(𝑦) = 0 is
𝑝(𝑟) = 𝑟 𝑛 + 𝑎1 𝑟 𝑛−1 + 𝑎2 𝑟 𝑛−2 + ⋯ + 𝑎𝑛 .
Suppose 𝑟𝑖 for each value of i= 1, 2, ⋯ , 𝑠 is a root of 𝑝(𝑟) of multiplicity 𝑚𝑖 . Hence we
write 𝑝(𝑟) as
𝑝(𝑟)=(𝑟 − 𝑟𝑖 )𝑚𝑖 𝑞(𝑟), (4)
where 𝑞(𝑟) is a polynomial of degree 𝑛 – 𝑚𝑖 and 𝑞(𝑟𝑖 ) ≠ 0.
Differentiating (4) 𝑚𝑖 − 1 times with respect to 𝑟, we get
𝑝′ (𝑟) = (𝑟 − 𝑟𝑖 )𝑚𝑖 𝑞 ′ (𝑟) + 𝑚𝑖 (𝑟 − 𝑟𝑖 )𝑚𝑖 −1 𝑞(𝑟)
⇒ p′ (𝑟)=(𝑟 − 𝑟𝑖 )𝑚𝑖 −1 [𝑞′(𝑟)(𝑟 − 𝑟𝑖 ) + 𝑚𝑖 𝑞(𝑟)]
⇒ 𝑝′(𝑟)=(𝑟 − 𝑟𝑖 )𝑚𝑖 −1 × [polynomial of degree 𝑛 − 𝑚𝑖 ]

S25013: Ordinary Differential Equations Page 52


Similarly, we obtain
𝑝′′ (𝑟) = (𝑟 − 𝑟𝑖 )𝑚𝑖 𝑞′′(𝑟) + 𝑚𝑖 (𝑟 − 𝑟𝑖 )𝑚𝑖 −1 𝑞 ′(𝑟) + 𝑚𝑖 (𝑟 − 𝑟𝑖 )𝑚𝑖 −1 𝑞 ′(𝑟)
+𝑚𝑖 (𝑚𝑖 − 1)(𝑟 − 𝑟𝑖 )𝑚𝑖 −2 𝑞(𝑟)
⇒ 𝑝′′(𝑟)=(𝑟 − 𝑟𝑖 )𝑚𝑖 −2 [(𝑟 − 𝑟𝑖 )2 𝑞′′(𝑟) + 2𝑚𝑖 (𝑟 − 𝑟𝑖 )𝑞 ′ (𝑟) + 𝑚𝑖 (𝑚𝑖 − 1)𝑞(𝑟)]
⇒ 𝑝′′(𝑟)=(𝑟 − 𝑟𝑖 )𝑚𝑖 −2 × [polynomial of degree 𝑛 − 𝑚𝑖 ].
Continuing in this way, we obtain
𝑝(𝑚𝑖 −1) (𝑟)=(𝑟 − 𝑟𝑖 )𝑚𝑖−(𝑚𝑖 −1) × [polynomial of degree 𝑛 − 𝑚𝑖 ]
i.e. 𝑝(𝑚𝑖 −1) (𝑟)=(𝑟 − 𝑟𝑖 )× [polynomial of degree 𝑛 − 𝑚𝑖 ].
We see that if 𝑟𝑖 is a root of 𝑝(𝑟) of multiplicity 𝑚𝑖 , then
𝑝(𝑟𝑖 )=𝑝′ (𝑟𝑖 ) = 𝑝′′ (𝑟𝑖 ) = ⋯ = 𝑝(𝑚𝑖 −1) (𝑟𝑖 ) = 0 and 𝑝(𝑚𝑖 ) (𝑟𝑖 ) ≠ 0. (5)
We look for some 𝑟𝑖 , for which 𝑒 is a solution of 𝐿(𝑦) = 0
𝑟𝑥

⇒ 𝐿(𝑒 𝑟𝑥 )=𝑝(𝑟)𝑒 𝑟𝑥 . (6)


If 𝑟𝑖 is a root of 𝑝(𝑟) , then we have 𝐿(𝑒 𝑒 is a solution of 𝐿(𝑦) = 0.
𝑟𝑖 𝑥 )=0⇒ 𝑟𝑖 𝑥

Differentiating equation (6) 𝑘 − times with respect to 𝑟, we get


𝜕𝑘 𝜕𝑘
(𝐿(𝑒 𝑟𝑥 ))= 𝜕𝑟 𝑘 [𝑝(𝑟)𝑒 𝑟𝑥 ]
𝜕𝑟 𝑘
𝜕𝑘 𝜕𝑘
⇒ 𝐿 [𝜕𝑟 𝑘 (𝑒 𝑟𝑥 )]= 𝜕𝑟 𝑘 [𝑝(𝑟)𝑒 𝑟𝑥 ]
𝜕𝑘
⇒ 𝐿(𝑥 𝑘 𝑒 𝑟𝑥 )= 𝜕𝑟 𝑘 [𝑝(𝑟)𝑒 𝑟𝑥 ]. (7)
Now, we recall the formula
𝑘(𝑘−1) (𝑘−2)
(𝑓𝑔)𝑘 =𝑓 (𝑘) 𝑔 + 𝑘𝑓 (𝑘−1) 𝑔′ + 𝑓 𝑔′′ + ⋯ + 𝑓𝑔(𝑘) . (8)
2!
Using the formula (8), we obtain from (7)
𝑘(𝑘−1)
𝐿(𝑥 𝑘 𝑒 𝑟𝑥 )=[𝑝(𝑘) (𝑟) + 𝑘𝑝(𝑘−1) (𝑟)𝑥 + 𝑝(𝑘−2) (𝑟)𝑥 2 + ⋯ + 𝑝(𝑟)𝑥 𝑘 ] 𝑒 𝑟𝑥 .
2!
For 𝑟 = 𝑟𝑖 and 𝑘 = 0, 1, 2, ⋯ , 𝑚𝑖−1 and using equation (5), we find that 𝑥 𝑘 𝑒 𝑟𝑖 𝑥 is a solution
of 𝐿(𝑦) = 0. This is true for all values of 𝑖 = 1, 2, ⋯ , 𝑠
⇒ 𝑥 𝑘 𝑒 𝑟𝑖 𝑥 , 𝑘 = 0,1,2, ⋯ , 𝑚𝑖 − 1, 𝑖 = 1,2, ⋯ , 𝑠 are solutions of 𝐿(𝑦) = 0. This proves the
theorem.
Theorem 2: The 𝑛 solutions 𝑥 𝑘 𝑒 𝑟𝑖 𝑥 , 𝑘 = 0,1,2, ⋯ , 𝑚𝑖 − 1 , 𝑖 = 1, 2, ⋯ , 𝑠 of 𝐿(𝑦) = 0 are
linearly independent on any interval 𝐼.
Proof: Consider the nth order differential equation with constant coefficients given by
𝐿(𝑦) = 𝑦 (𝑛) + 𝑎1 𝑦 (𝑛−1) + 𝑎2 𝑦 (𝑛−2) + ⋯ + 𝑎𝑛 𝑦 = 0.
The characteristic polynomial of 𝐿(𝑦) = 0 is given by
𝑝(𝑟) = 𝑟 𝑛 + 𝑎1 𝑟 𝑛−1 + 𝑎2 𝑦 (𝑛−2) + ⋯ + 𝑎𝑛 𝑦
Let 𝑟1 , 𝑟2 , ⋯ , 𝑟𝑠 be the distinct roots of the characteristic polynomial 𝑝(𝑟) and suppose each
r i has multiplicity 𝑚𝑖 , 𝑖 = 1,2, ⋯ , 𝑠. In this case we have shown in the Theorem 1 that
𝑥 𝑘 𝑒 𝑟𝑖 𝑥 , 𝑘 = 0,1,2, ⋯ , 𝑚𝑖 − 1, 𝑖 = 1,2, ⋯ , 𝑠
are the solution of 𝐿(𝑦) = 0 on 𝐼.
Now we show that these 𝑛 solutions are linearly independent.
Consider the constants 𝑐𝑖𝑘 , 𝑖 = 1,2, ⋯ , 𝑠 and 𝑘 = 0,1,2, ⋯ , 𝑚𝑖 − 1 such that the algebraic
sum

S25013: Ordinary Differential Equations Page 53


∑𝑠𝑖=1 ∑𝑚𝑖 −1 𝑘 𝑟𝑖 𝑥
𝑘=0 𝑐𝑖𝑘 𝑥 𝑒 = 0, on 𝐼 (9)
𝑚 −1
⇒ ∑𝑠𝑖=1[∑𝑘=0 𝑖
𝑐𝑖𝑘 𝑥 𝑘 ]𝑒 𝑟𝑖𝑥 = 0 (10)
or ∑𝑠𝑖=1 𝑝𝑖 (𝑥)𝑒 ri 𝑥 = 0
𝑚 −1
where 𝑝𝑖 (𝑥) = ∑𝑘=0
𝑖
𝑐𝑖𝑘 𝑥 𝑘 ⇒ 𝑝𝑖 (𝑥) = 𝑐𝑖0 + 𝑐𝑖1 𝑥 + 𝑐𝑖2 𝑥 2 + ⋯ + 𝑐𝑖 𝑚 −1 𝑥 𝑚𝑖−1, (11)
𝑖

is a polynomial of degree 𝑚𝑖 − 1.
Expanding the sum defined in (10), we get
𝑝1 (𝑥)𝑒 𝑟1𝑥 + 𝑝2 (𝑥)𝑒 𝑟2𝑥 + ⋯ + 𝑝𝑠 (𝑥)𝑒 𝑟𝑠 𝑥 = 0. (12)
Assume that not all the constants 𝑐1𝑘 are zero. This implies that there will be at least one
of the polynomials 𝑝𝑖 (𝑥) which is not identically zero on I. Assume that 𝑝𝑠 (𝑥) is not
identically zero on I. Multiplying equation (12) by 𝑒 −𝑟1𝑥 we get
𝑝1 (𝑥) + 𝑝2 (𝑥)𝑒 (𝑟2−𝑟1)𝑥 + ⋯ + 𝑝𝑠 (𝑥)𝑒 (𝑟𝑠 −𝑟1)𝑥 = 0, on 𝐼. (13)
Differentiating the equation (13) sufficiently many times (at most 𝑚1 times) we can reduce
𝑝1 (𝑥) to 0. In this process of differentiation, the degrees of the polynomials multiplying
𝑒 (𝑟2 −𝑟1)𝑥 remains unchanged. Hence, we can obtain an expression of the form
𝑄2 (𝑥)𝑒 (𝑟2−𝑟1)𝑥 + ⋯ + 𝑄𝑠 (𝑥)𝑒 (𝑟𝑠 −𝑟1)𝑥 = 0
⇒ 𝑄2 (𝑥)𝑒 𝑟2𝑥 + ⋯ + 𝑄𝑠 (𝑥)𝑒 𝑟𝑠 𝑥 = 0. (14)
Multiplying the equation (14) by 𝑒 −𝑟2 𝑥
, we get
𝑄2 (𝑥) + ⋯ + 𝑄𝑠 (𝑥)𝑒 (𝑟𝑠 −𝑟2)𝑥 = 0, (15)
where 𝑄𝑖 (𝑥) are polynomials such that 𝑑𝑒𝑔 𝑄𝑖 (𝑥) = 𝑑𝑒𝑔 𝑝𝑖 (𝑥), and 𝑄𝑠 (𝑥) does not vanish
identically zero on 𝐼 . Continuing this way, we finally arrive at a situation where
𝑅𝑠 (𝑥)𝑒 𝑟𝑠 𝑥 = 0 on 𝐼 and 𝑅𝑠 (𝑥) is a polynomial such that 𝑑𝑒𝑔 𝑅𝑠 (𝑥) = 𝑑𝑒𝑔 𝑝𝑠 (𝑥)which does
not vanish identically on 𝐼. But the equation 𝑅𝑠 (𝑥)𝑒 𝑟𝑠 𝑥 = 0 yields 𝑅𝑠 (𝑥) = 0, ∀ 𝑥 ∈ 𝐼. This
contradict that 𝑅𝑠 (𝑥) ≠ 0 on 𝐼.
This contradiction forces us to assume
𝑝𝑠 (𝑥) = 0, ∀ 𝑥 ∈ 𝐼, i. e. 𝑝𝑠 (𝑥) is zero polynomial
⇒ 𝑐𝑖𝑘 = 0, ∀ 𝑖, 𝑘
⇒ all 𝑛 solutions of 𝐿(𝑦) = 0 are linearly independent on any interval 𝐼.

04-02: INITIAL VALUE PROBLEM FOR 𝒏𝒕𝒉 ORDER EQUATIONS:


Definition: Let 𝐿(𝑦) = 𝑦 (𝑛) + 𝑎1 𝑦 (𝑛−1) + 𝑎2 𝑦 (𝑛−2) + ⋯ + 𝑎𝑛 𝑦 = 0, where 𝑎1 , 𝑎2 , ⋯ , 𝑎𝑛 are
constants be a 𝑛𝑡ℎ order differential equation. An IVP for 𝐿(𝑦) = 0 is a problem of finding
a solution 𝜙 which has prescribed values for it and its first 𝑛 – 1 derivatives, at some initial
points 𝑥0 . Thus if ∝1 , ∝2 , ⋯ , ∝𝑛 are given constants and 𝑥0 is some real number, then the
problem of finding of solution 𝜙 of 𝐿(𝑦) = 0 satisfying 𝜙(𝑥0 ) =∝1 , 𝜙′(𝑥0 ) =
∝2 , ⋯ , 𝜙 (𝑛−1) (𝑥0 ) =∝𝑛 is called on IVP.
Mathematically, the problem: 𝐿(𝑦) = 0 such that
𝑦(𝑥0 ) =∝1 , 𝑦′(𝑥0 ) =∝2 , ⋯ , 𝑦 (𝑛−1) (𝑥0 ) =∝𝑛
is called an initial value problem. There exists only one solution of such IVP and the
demonstration of the proof will depend on the following Theorem 3.

S25013: Ordinary Differential Equations Page 54


Theorem 3: Let 𝜙(𝑥) be any solution of the equation
𝐿(𝑦) = 𝑦 (𝑛) + 𝑎1 𝑦 (𝑛−1) + 𝑎2 𝑦 (𝑛−2) + ⋯ + 𝑎𝑛 𝑦 = 0
on interval 𝐼 containing a point 𝑥0 . Then for all 𝑥 in 𝐼, we have
‖𝜙(𝑥0 )‖𝑒 −𝑘|𝑥−𝑥0| ≤ ‖𝜙(𝑥)‖ ≤ ‖𝜙(𝑥0 )‖𝑒 𝑘|𝑥−𝑥0| ,
where 𝑘 = 1 + |𝑎1 | + |𝑎2 | + ⋯ + |𝑎𝑛 |.
Proof: (Note the proof of the theorem is analogous is the proof of the theorem in Unit 01 -
01). Let 𝜙 (x) be a solution of 𝐿(𝑦) = 0 on an interval 𝐼 containing a point 𝑥0 ⇒ 𝐿(𝜙) = 0
⟹ 𝜙 (𝑛) (𝑥) + 𝑎1 𝜙 (𝑛−1) (𝑥) + 𝑎2 𝜙 (𝑛−2) (𝑥) + ⋯ + 𝑎𝑛 𝜙(𝑥) = 0 (16)
(𝑛)
⟹ 𝜙 (𝑥) = −[𝑎1 𝜙 (𝑛−1)
(𝑥) + 𝑎2 𝜙 (𝑛−2)
(𝑥) + ⋯ + 𝑎𝑛 𝜙(𝑥)] (17)
⟹ |𝜙 (𝑛) (𝑥)| ≤ |𝑎1 ||𝜙 (𝑛−1) (𝑥)| + |𝑎2 ||𝜙 (𝑛−2) (𝑥)| + ⋯ + |𝑎𝑛 ||𝜙(𝑥)|. (18)
2
Let 𝑢(𝑥)=‖𝜙(𝑥)‖2 = [|𝜙(𝑥)|2 + |𝜙′(𝑥)|2 + ⋯ + |𝜙 (𝑛−1) (𝑥)| ], where |𝜙(𝑥)|2 = 𝜙𝜙.

⟹ 𝑢(𝑥) = 𝜙𝜙 + ϕ′ 𝜙 ′ + ⋯ + 𝜙 (𝑛−1) 𝜙 (𝑛−1) .


Differentiating this equation with respect to 𝑥, we get
𝑢′(𝑥)=𝜙′𝜙 + 𝜙𝜙′ + 𝜙′′𝜙′ + 𝜙′𝜙′′+. . . . . +𝜙 (𝑛−1) 𝜙 (𝑛)
⇒ |𝑢′(𝑥)| ≤ 2|𝜙(𝑥)||𝜙′(𝑥)| + 2|𝜙′(𝑥)||𝜙′′(𝑥)| + ⋯ + 2|𝜙 (𝑛−1) (𝑥)||𝜙 (𝑛) (𝑥)| as |𝜙| = |𝜙|.
Now using the equation (18) in the above equation, we get
|u′ (𝑥)| ≤ 2|𝜙(𝑥)||𝜙 ′ (𝑥)| + 2|𝜙 ′ (𝑥)||𝜙 ′′ (𝑥)| + ⋯ + 2|𝜙 (𝑛−2) (𝑥)||𝜙 (𝑛−1) (𝑥)|
+2|𝜙 (𝑛−1) (𝑥)| [|𝑎1 ||𝜙 (𝑛−1) (𝑥)| + |𝑎2 |𝜙 (𝑛−2) (𝑥) + ⋯ + |𝑎𝑛−1 ||𝜙′(𝑥)| + |𝑎𝑛 ||𝜙(𝑥)|]
⇒ |𝑢′ (𝑥)| ≤ 2|𝜙(𝑥)||𝜙 ′ (𝑥)| + 2|𝜙 ′ (𝑥)||𝜙 ′′ (𝑥)| + ⋯ + 2|𝜙 (𝑛−2) (𝑥)||𝜙 (𝑛−1) (𝑥)|
2
+2|𝑎1 ||𝜙 (𝑛−1) (𝑥)| + 2|𝑎2 ||𝜙 (𝑛−1) (𝑥)||𝜙 (𝑛−2) (𝑥)|
+ ⋯ + 2|𝑎𝑛−1 ||𝜙 (𝑛−1) (𝑥)||𝜙 ′(𝑥) | + 2|𝑎𝑛 ||𝜙 (𝑛−1) (𝑥)||𝜙(𝑥)|.
We know the inequality 2|𝜙||𝜙′| ≤ |𝜙|2 + |𝜙′|2 .
Using this inequality in the above equation, we find
|𝑢′ (𝑥)| ≤ (|𝜙(𝑥)|2 + |𝜙 ′ (𝑥)|2 ) + (|𝜙 ′ (𝑥)|2 + |𝜙 ′′ (𝑥)|2 ) + ⋯
2 2 2
+ (|𝜙 (𝑛−2) (𝑥)| + |𝜙 (𝑛−1) (𝑥)| ) + 2|𝑎1 ||𝜙 (𝑛−1) (𝑥)|
2 2
+|𝑎2 | (|𝜙 (𝑛−1) (𝑥)| + |𝜙 (𝑛−1) (𝑥)| )
2 2
+ ⋯ + |𝑎𝑛−1 | (|𝜙 (𝑛−1) | + |𝜙′(𝑥)|2 ) + |𝑎𝑛 | (|𝜙 (𝑛−1) (𝑥)| + |𝜙(𝑥)|2 )
2
⇒ |𝑢′(𝑥)| ≤ (1 + |𝑎𝑛 |)|𝜙(𝑥)|2 + (2 + |𝑎𝑛−1 |)|𝜙′(𝑥)|2 + ⋯ +(2 + |𝑎2 |)|𝜙 (𝑛−2) (𝑥)|
+2(1 + |𝑎1 | + |𝑎2 | + ⋯ + |𝑎𝑛 |) 𝜙 (𝑛−1) (𝑥).
The above inequality will hold good if we change each coefficient by the greater
coefficient 2(1 + |𝑎1 | + |𝑎2 | + ⋯ + |𝑎𝑛 |), we obtain
2
|𝑢′(𝑥)| ≤ 2(1 + |𝑎2 |+. . . . . +|𝑎𝑛 |) [|𝜙(𝑥)|2 + |𝜙′(𝑥)|2 + ⋯ + |𝜙(𝑥)(𝑛−1) (𝑥)| ].
Let 𝑘 = 1 + |𝑎1 | + |𝑎2 | + ⋯ + |𝑎𝑛 |, therefore, we have
|𝑢′(𝑥)| ≤ 2𝑘𝑢(𝑥)
⇒ −2𝑘𝑢(𝑥) ≤ 𝑢′ (𝑥) ≤ 2𝑘𝑢(𝑥), ∀ 𝑥 ∈ 𝐼. (19)
This is exactly same as the equation (28) of the Theorem 4 of Unit 01-01. Proceeding
exactly in the same way, we arrive at an inequality
S25013: Ordinary Differential Equations Page 55
‖𝜙(𝑥)‖𝑒 −𝑘|𝑥−𝑥0| ≤ ‖𝜙(𝑥)‖ ≤ ‖𝜙(𝑥0 )‖𝑒 𝑘|𝑥−𝑥0| , ∀ 𝑥 ∈ 𝐼. (20)
This completes the proof.

Wronskian of 𝒏 –functions:
Definition: Let 𝜙1 , 𝜙2 , ⋯ , 𝜙𝑛 be 𝑛 functions having at least (𝑛 − 1) derivatives on an
𝜙1 𝜙2 𝜙𝑛−1 𝜙𝑛
′ ′ ⋯ ′
𝜙1 𝜙2 𝜙𝑛−1 𝜙𝑛′
interval I. Then the determinant | ⋮ ⋱ ⋮ | is called the
(𝑛−1) (𝑛−1) (𝑛−1) (𝑛−1)
𝜙1 𝜙2 ⋯ 𝜙𝑛−1 𝜙𝑛
Wronskian of the functions 𝜙1 , 𝜙2 , ⋯ , 𝜙𝑛 and it is denoted by 𝑊(𝜙1 , 𝜙2 , ⋯ , 𝜙𝑛 )(𝑥).
Thus,
𝜙1 𝜙2 𝜙𝑛−1 𝜙𝑛
′ ′ ⋯ ′
𝜙1 𝜙2 𝜙𝑛−1 𝜙𝑛′
𝑊(𝜙1 , 𝜙2 , . . . . . . . . . . . 𝜙𝑛 )(𝑥) = | ⋮ ⋱ ⋮ |. (21)
(𝑛−1) (𝑛−1) (𝑛−1) (𝑛−1)
𝜙1 𝜙2 ⋯ 𝜙𝑛−1 𝜙𝑛
Theorem 4: If 𝜙1 , 𝜙2 , ⋯ , 𝜙𝑛 are 𝑛 –solutions of
𝐿(𝑦) = 𝑦 (𝑛) + 𝑎1 𝑦 (𝑛−1) + 𝑎2 𝑦 (𝑛−2) + ⋯ + 𝑎𝑛 𝑦 = 0
on an interval 𝐼 , then they are linearly independent on 𝐼 if, and only if,
𝑊(𝜙1 , 𝜙2 , ⋯ , 𝜙𝑛 )(𝑥) ≠ 0 for all 𝑥 in 𝐼.
Proof: The proof of the Theorem is omitted and left as an exercise for two reasons. First
the proof is entirely similar to the proof of the Theorem 1, the case 𝑛 = 2 of the Unit 01-
01 and second, we will prove this theorem for the general homogeneous equation with
variable coefficients of order 𝑛. Since, the method does not depend on the fact that the
coefficient 𝑎1 , 𝑎2 , ⋯ , 𝑎𝑛 are constants or variables, hence avoid repetition.

04-03: EXISTENCE AND UNIQUENESS THEOREMS


Theorem 5: (Existence theorem) Let ∝1 , ∝2 , ⋯ , ∝𝑛 be any 𝑛 constants and let 𝑥0 be any
real number. Then there exists a solution 𝜙 of
𝐿(𝑦) = 𝑦 (𝑛) + 𝑎1 𝑦 (𝑛−1) + 𝑎2 𝑦 (𝑛−2) + ⋯ + 𝑎𝑛 𝑦 = 0
on −∞ < 𝑥 < ∞ satisfying 𝜙(𝑥0 ) =∝1 , 𝜙 ′ (𝑥0 ) =∝2 , ⋯ , 𝜙 (𝑛−1) (𝑥0 ) =∝𝑛 .
Proof: Let 𝜙1 , 𝜙2 , ⋯ , 𝜙𝑛 be any set of 𝑛 linearly independent solutions of 𝐿(𝑦) = 0 on
−∞ < 𝑥 < ∞. We will prove that there exist unique constants 𝑐1 , 𝑐2 , ⋯ , 𝑐𝑛 such that
𝜙(𝑥) = 𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥) + ⋯ + 𝑐𝑛 𝜙𝑛 (𝑥) (22)
is a solution of 𝐿(𝑦) = 0 satisfying
𝜙(𝑥0 ) =∝1 , 𝜙1′ (𝑥0 ) =∝2 , ⋯ , 𝜙 (𝑛−1) (𝑥0 ) =∝𝑛 . (23)
To satisfy the conditions (23), these constants would have to satisfy the equations
𝑐1 𝜙1 (𝑥0 ) + 𝑐2 𝜙2 (𝑥0 ) + ⋯ + 𝑐𝑛 𝜙𝑛 (𝑥0 ) =∝1
𝑐1 𝜙1′ (𝑥0 ) + 𝑐2 𝜙2′ (𝑥0 ) + ⋯ + 𝑐𝑛 𝜙′𝑛 (𝑥0 ) =∝2 ,

(𝑛−1) (𝑛−1)
𝑐1 𝜙1 + 𝑐2 𝜙2 (𝑥0 ) + ⋯ + 𝑐𝑛 𝜙𝑛(𝑛−1) (𝑥0 ) =∝𝑛 . (24)

S25013: Ordinary Differential Equations Page 56


We see that the equation (24) is a system of 𝑛 linear non-homogeneous equations in
𝑐1 , 𝑐2 , ⋯ , 𝑐𝑛 . Since the function 𝜙1 (𝑥), 𝜙2 (𝑥), ⋯ , 𝜙𝑛 (𝑥) are linearly independent solutions of
𝐿(𝑦) = 0, then the determinant of the coefficients is just the Wronskian
𝑊(𝜙1 , 𝜙2 , ⋯ , 𝜙𝑛 )(𝑥0 ) ≠ 0. (25)
There exists unique set of constants 𝑐1 , 𝑐2 , ⋯ , 𝑐𝑛 as the solutions of the equation (24). For
this choice 𝑐1 , 𝑐2 , ⋯ , 𝑐𝑛 , the functions 𝜙(𝑥) = 𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥) + ⋯ + 𝑐𝑛 𝜙𝑛 (𝑥) will be
the desired solution of the equation 𝐿(𝑦) = 0.
Theorem 6: (Uniqueness theorem): Let ∝1 , ∝2 , ⋯ , ∝𝑛 be any n constants and let 𝑥0 be any
real number on any interval 𝐼 containing 𝑥0 . Then there exists at most one solution 𝜙 of
𝐿(𝑦) = 𝑦 (𝑛) + 𝑎1 𝑦 (𝑛−1) + 𝑎2 𝑦 (𝑛−2) + ⋯ + 𝑎𝑛 𝑦 = 0 satisfying
𝜙(𝑥0 ) =∝1 , 𝜙′(𝑥0 ) =∝2 , ⋯ , 𝜙 (𝑛−1) (𝑥0 ) =∝𝑛 .
Proof: By existence theorem there is a solution of the IVP
𝐿(𝑦) = 0, 𝑦(𝑥0 ) = ∝1 , 𝑦′(𝑥0 ) =∝2 , ⋯ , 𝑦 (𝑛−1) (𝑥0 ) =∝𝑛 . (26)
Let 𝜙(𝑥) and 𝜓(𝑥) be two solutions of the IVP (26). Then we have
𝐿(𝜙)=0, 𝜙 ′ (𝑥0 )=∝1 𝜙1′ (𝑥0 )=∝2 , ⋯ , 𝜙 (𝑛−1) (𝑥0 ) =∝𝑛 (27)
and
𝐿(𝜓) = 0, 𝜓(𝑥0 ) =∝1 , 𝜓′(𝑥0 ) =∝2 , ⋯ , 𝜓 (𝑛−1) (𝑥0 ) =∝𝑛 (28)
Consider a function χ(𝑥) = ϕ(𝑥) − ψ(𝑥).
⇒ 𝐿(𝜒) = L(ϕ(𝑥)) − L(ψ(𝑥)) as 𝐿 is linear operator.
⇒ 𝐿(𝜒) = 0.
⇒ χ(𝑥) is a solution of 𝐿(𝑦) = 0.
Further, we see that 𝜒(𝑥0 ) = 𝜙(𝑥0 ) − 𝜓(𝑥0 )
= ∝1 −∝1
⇒ 𝜒(𝑥0 ) = 0.
Similarly, we have by using the equation (27) and (28) that
𝜒 ′ (𝑥0 ) = 𝜙 ′ (𝑥0 ) − 𝜓′ (𝑥0 ) =∝2 −∝2 = 0,
𝜒 ′′ (𝑥0 ) = 𝜙 ′′ (𝑥0 ) − 𝜓′′ (𝑥0 ) =∝3 −∝3 = 0,

and 𝜒 (𝑛−1)
(𝑥0 )=𝜙 (𝑛−1) (𝑥0 ) − 𝜓 (𝑛−1) (𝑥0 ) =∝𝑛 −∝𝑛 = 0.
Thus, we have proved that 𝜒 = 𝜙 − 𝜓 is a solution of the IVP (26). Since 𝜒(𝑥0 ) = 𝜒 ′ (𝑥0 ) =
⋯ = 𝜒 (𝑛−1) (𝑥0 ) = 0, therefore we have
2
‖𝜒(𝑥0 )‖2 = |𝜒(𝑥0 )|2 + |𝜒 ′ (𝑥0 )|2 + ⋯ + |𝜒 (𝑛−1) (𝑥0 )| = 0 ⟹ ‖𝜒(𝑥0 )‖ = 0.
Hence from the Theorem 3, the function 𝜒 satisfies the inequality
‖𝜒(𝑥0 )‖𝑒 −𝑘|𝑥−𝑥0| ≤ ‖𝜒(𝑥)‖ ≤ ‖𝜒(𝑥0 )‖𝑒 𝑘|𝑥−𝑥0| , ∀ 𝑥 ∈ 𝐼.
⟹ 𝜒(𝑥) = 0, ∀ 𝑥 ∈ 𝐼
⟹ 𝜙(𝑥) − 𝜓(𝑥)=0, ∀ 𝑥 ∈ 𝐼
⟹ 𝜙(𝑥) = 𝜓(𝑥), ∀ 𝑥 ∈ 𝐼.
This proves that the solution of the IVP (26) is unique.
Note: To avoid the repetition of the proof, we omit the proofs of the following theorems
as these are entirely similar to the proofs of the case 𝑛 = 2 (See corresponding proofs in

S25013: Ordinary Differential Equations Page 57


the Unit 01-01) and moreover, we will prove these theorems for the more general case of
a homogenous equation with variable coefficients. The method does not depend on the fact
that the coefficients 𝑎1 , 𝑎2 , ⋯ , 𝑎𝑛 of 𝐿(𝑦) = 0 are constants or variables. Students are
advised to see the proofs for a more general case in the next Unit.
Theorem 7: Let 𝜙1 , 𝜙2 , ⋯ , 𝜙𝑛 be 𝑛 linearly independent solutions of
𝐿(𝑦) = 𝑦 (𝑛) + 𝑎1 𝑦 (𝑛−1) + 𝑎2 𝑦 (𝑛−2) + ⋯ + 𝑎𝑛 𝑦 = 0
on an interval 𝐼. If 𝑐1 , 𝑐2 , ⋯ , 𝑐𝑛 are any constants, then every solution of 𝐿(𝑦) = 0 may be
represented by 𝜙(𝑥) = 𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥) + ⋯ + 𝑐𝑛 𝜙𝑛 (𝑥).
Proof: - Refer the proof of the theorem of next Unit.
Theorem 8: Let 𝜙1 , 𝜙2 , ⋯ , 𝜙𝑛 be 𝑛 solutions of
𝐿(𝑦) = 𝑦 (𝑛) + 𝑎1 𝑦 (𝑛−1) + 𝑎2 𝑦 (𝑛−2) + ⋯ + 𝑎𝑛 𝑦 = 0
on an interval 𝐼 containing a point 𝑥0 . Then we have
𝑊(𝜙1 , 𝜙2 , ⋯ , 𝜙𝑛 )(𝑥) = 𝑒 −𝑎1(𝑥−𝑥0) 𝑊(𝜙1 , 𝜙2 , ⋯ , 𝜙𝑛 )(𝑥0 ).
Proof: For proof refer the theorem of next Unit. This theorem is a corollary of a more
general theorem of the next Unit.

04-04: EQUATIONS WITH REAL CONSTANTS


The roots of polynomials can be real or non-real complex numbers. (We need to be a little
careful with our language because a real number is also a complex number with imaginary
part 0.) Roots can also be repeated.
The following theorem plays an important role in many practical problems presented by
differential equations with real coefficients
Theorem 9: Suppose the constants 𝑎1 , 𝑎2 , ⋯ , 𝑎𝑛 in the equation
𝐿(𝑦) = 𝑦 (𝑛) + 𝑎1 𝑦 (𝑛−1) + 𝑎2 𝑦 (𝑛−2) + ⋯ + 𝑎𝑛 𝑦 = 0
are all real. Let 𝑟1 , 𝑟2 , ⋯ 𝑟𝑗 , 𝑟2𝑗+1 , ⋯ 𝑟𝑠 be distinct roots of the characteristic polynomial
𝑝(𝑟) = 𝑟 𝑛 + 𝑎1 𝑟 𝑛−1 + 𝑎2 𝑟 𝑛−2 ⋯ + 𝑎𝑛 of multiplicities 𝑚1 , 𝑚2 , ⋯ 𝑚𝑗 , 𝑚2𝑗+1 , ⋯ 𝑚𝑠 , where
𝑟𝑘 = 𝜎𝑘 + 𝑖𝜏𝑘 , (𝜏𝑘 ≠ 0), 𝜎𝑘 , 𝜏𝑘 and 𝑟2𝑗+1 , ⋯ , 𝑟𝑠 are real for 𝑘 = 1,2, ⋯ , 𝑗. There exists a set of
𝑛 linearly independent real-valued solutions,
𝑒 𝜎1𝑥 cos 𝜏1 𝑥 , 𝑥𝑒 𝜎1𝑥 cos 𝜏1 𝑥 , ⋯ , 𝑥 𝑚1−1 𝑒 𝜎1𝑥 cos 𝜏1 𝑥,
𝑒 𝜎1𝑥 sin 𝜏1 𝑥 , 𝑥𝑒 𝜎1𝑥 sin 𝜏1 𝑥 , ⋯ , 𝑥 𝑚1 −1 𝑒 𝜎1𝑥 sin 𝜏1 𝑥,



𝑒 𝜎𝑗 𝑥 cos 𝜏𝑗 𝑥 , 𝑥𝑒 𝜎𝑗 𝑥 cos 𝜏𝑗 𝑥 , ⋯ , 𝑥 𝑚𝑗 −1 𝑒 𝜎𝑗 𝑥 cos 𝜏𝑗 𝑥,
𝑒 𝜎𝑗 𝑥 sin 𝜏𝑗 𝑥 , 𝑥𝑒 𝜎𝑗 𝑥 sin 𝜏𝑗 𝑥 , ⋯ , 𝑥 𝑚𝑗 −1 𝑒 𝜎𝑗 𝑥 sin 𝜏𝑗 𝑥,
𝑒 𝑟2𝑗+1 𝑥 , 𝑥𝑒 𝑟2𝑗+1𝑥 , ⋯ , 𝑥 𝑚2𝑗+1 −1 𝑒 𝑟2𝑗+1 𝑥 ,



𝑒 𝑟𝑠 𝑥 , 𝑥𝑒 𝑟𝑠 𝑥 , ⋯ , 𝑥 𝑚𝑠 −1 𝑒 𝑟𝑠 𝑥 ,
and every real-valued solution is a linear combination of these with real coefficients. If a
solution satisfies real initial conditions, then it is real -valued.

S25013: Ordinary Differential Equations Page 58


Proof: Suppose that the constants 𝑎1 , 𝑎2 , ⋯ , 𝑎𝑛 in
𝐿(𝑦) = 𝑦 (𝑛) + 𝑎1 𝑦 (𝑛−1) + 𝑎2 𝑦 (𝑛−2) + ⋯ + 𝑎𝑛 𝑦 = 0 (29)
are all real numbers. The characteristic polynomial
𝑝(𝑟) = 𝑟 𝑛 + 𝑎1 𝑟 𝑛−1 + 𝑎2 𝑟 𝑛−2 + ⋯ + 𝑎𝑛 , (30)
then has all real coefficients. This implies that
𝑝(𝑟) = 𝑝(𝑟), (31)
for all 𝑟, since
𝑝(𝑟) = 𝑟 𝑛 + 𝑎1 𝑟 𝑛−1 + 𝑎2 𝑟 𝑛−2 + ⋯ + 𝑎𝑛
= 𝑟 𝑛 + 𝑎1 𝑟 𝑛−1 + 𝑎2 𝑟 𝑛−2 + ⋯ + 𝑎𝑛
𝑛 𝑛−1 𝑛−2
= 𝑟 + 𝑎1 𝑟 + 𝑎2 𝑟 + ⋯ + 𝑎𝑛
= 𝑝(𝑟).
From (31), it follows that if 𝑟1 is root of 𝑝 , then so is 𝑟1 . Thus, the roots of 𝑝 whose
imaginary parts do not vanish occur in conjugate pairs. A slight extension of this argument
shows that if 𝑟1 is root of 𝑝 of multiplicity 𝑚1 , then so is 𝑟1 with same multiplicity 𝑚1 . If
there are 𝑠 distinct roots of 𝑝, let us enumerate them as follows:
𝑟1 , 𝑟1 , 𝑟2 , 𝑟2 , ⋯ , 𝑟𝑗 , 𝑟𝑗 , 𝑟2𝑗+1 , ⋯ , 𝑟𝑠 ,
where
𝑟𝑘 = 𝜎𝑘 + 𝑖𝜏𝑘 , (𝑘 = 1,2, ⋯ , 𝑗; 𝜎𝑘 , 𝜏𝑘 are real 𝜏𝑘 ≠ 0),
and 𝑟2𝑗+1 , ⋯ , 𝑟𝑠 are real. Suppose that 𝑟𝑘 has multiplicity 𝑚𝑘 .Then we have
2(𝑚1 + 𝑚2 + ⋯ + 𝑚𝑗 ) + 𝑚2𝑗+1 + ⋯ + 𝑚𝑠 = 𝑛.
Corresponding to these roots, we have the 𝑛 linearly independent solutions
𝑒 𝑟1𝑥 , 𝑥𝑒 𝑟1𝑥 , ⋯ , 𝑥 𝑚1 −1 𝑒 𝑟1𝑥 , ; 𝑒 𝑟1𝑥 , 𝑥𝑒 𝑟1𝑥 , ⋯ , 𝑥 𝑚1 −1 𝑒 𝑟1𝑥 ; ⋯ ; 𝑒 𝑟𝑠 𝑥 , 𝑥𝑒 𝑟𝑠 𝑥 , ⋯ , 𝑥 𝑚𝑠 −1 𝑒 𝑟𝑠 𝑥 (32)
of 𝐿(𝑦) = 0. Every solution is a linear combination, with constant coefficients, of these.
We now note that if 1 ≤ 𝑘 ≤ 𝑗; 0 ≤ ℎ ≤ 𝑚𝑘 − 1,
𝑥 ℎ 𝑒 𝑟𝑘𝑥 = 𝑥 ℎ 𝑒 (𝜎𝑘 +𝑖𝜏𝑘)𝑥 = 𝑥 ℎ 𝑒 𝜎𝑘 𝑥 (cos 𝜏𝑘 𝑥 + 𝑖 sin 𝜏𝑘 𝑥),
(33)
𝑥 ℎ 𝑒 𝑟𝑘 𝑥
=𝑥 𝑒 ℎ (𝜎𝑘 −𝑖𝜏𝑘 )𝑥
=𝑥 𝑒 ℎ 𝜎𝑘 𝑥 (cos
𝜏𝑘 𝑥 − 𝑖 sin 𝜏𝑘 𝑥).
Thus, every solution is a linear combination, with constant coefficients, of the 𝑛 functions:
𝑒 𝜎1 𝑥 cos 𝜏1 𝑥 , 𝑥𝑒 𝜎1𝑥 cos 𝜏1 𝑥 , ⋯ , 𝑥 𝑚1 −1 𝑒 𝜎1𝑥 cos 𝜏1 𝑥
𝑒 𝜎1𝑥 sin 𝜏1 𝑥 , 𝑥𝑒 𝜎1𝑥 sin 𝜏1 𝑥 , ⋯ , 𝑥 𝑚1 −1 𝑒 𝜎1𝑥 sin 𝜏1 𝑥

⋅ (34)

𝑒 𝑟𝑠 𝑥 , 𝑥𝑒 𝑟𝑠 𝑥 , ⋯ , 𝑥 𝑚𝑠 −1 𝑒 𝑟𝑠 𝑥 .
Each of the functions in (34) is a solution of 𝐿(𝑦) = 0 since, from (33),
1 ℎ 𝑟 𝑥
𝑥 ℎ 𝑒 𝜎𝑘 𝑥 cos 𝜏𝑘 𝑥 = 𝑥 (𝑒 𝑘 + 𝑒 𝑟𝑘𝑥 ),
2
(35)
ℎ 𝜎𝑘 𝑥 1 ℎ 𝑟𝑘 𝑥 𝑟𝑘 𝑥
𝑥 𝑒 cos 𝜏𝑘 𝑥 = 2𝑖 𝑥 (𝑒 −𝑒 ).
The solutions in (34) are all real-valued, and they are linearly independent. For suppose
we have a linear combination of these functions equal to zero. Let us denote the terms in
this sum which involve
S25013: Ordinary Differential Equations Page 59
𝑥 ℎ 𝑒 𝜎𝑘 𝑥 cos 𝜏𝑘 𝑥, 𝑥 ℎ 𝑒 𝜎𝑘 𝑥 cos 𝜏𝑘 𝑥
by
𝑐𝑥 ℎ 𝑒 𝜎𝑘 𝑥 cos 𝜏𝑘 𝑥 + 𝑑𝑥 ℎ 𝑒 𝜎𝑘 𝑥 cos 𝜏𝑘 𝑥,
where 𝑐 and 𝑑 are constants. Using (35), we find that a linear combination of the functions
(32) equal to zero, and the terms involving 𝑥 ℎ 𝑒 𝑟𝑘𝑥 , 𝑥 ℎ 𝑒 𝑟𝑘𝑥 will be
(𝑐−𝑖𝑑) (𝑐+𝑖𝑑)
𝑥 ℎ 𝑒 𝑟𝑘 𝑥 + 𝑥 ℎ 𝑒 𝑟𝑘 𝑥 .
2 2
Since the functions (32) are linearly independent, we must have all the coefficients in this
sum equal to zero. In particular
𝑐 − 𝑖𝑑 = 0, 𝑐 + 𝑖𝑑 = 0,
from which it follows that 𝑐 = 0, 𝑑 = 0. Thus, the solutions (34) are linearly independent.
If 𝜙 is any real-valued solution of 𝐿(𝑦) = 0, then 𝜙 is a linear combination of the real
solutions (34) with real coefficients. Indeed, if we denote the solutions in (34) by
𝜙1 , 𝜙2 , ⋯ , 𝜙𝑛 , we have
𝜙(𝑥) = 𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥) + ⋯ + 𝑐𝑛 𝜙𝑛 (𝑥),
for some constants 𝑐1 , 𝑐2 , ⋯ , 𝑐𝑛 . Since 𝜙1 , 𝜙2 , ⋯ , 𝜙𝑛 are all real-valued, we have
0 = Im𝜙(𝑥) = (Im 𝑐1 )𝜙1 (𝑥) + (Im 𝑐2 )𝜙2 (𝑥) + ⋯ + (Im 𝑐𝑛 )𝜙𝑛 (𝑥),
and since 𝜙1 , 𝜙2 , ⋯ , 𝜙𝑛 are linearly independent, we must have
Im 𝑐1 = Im 𝑐2 = ⋯ = Im 𝑐𝑛 = 0.
This shows that 𝑐1 , 𝑐2 , ⋯ , 𝑐𝑛 are all real numbers.
We remark that if 𝜙 is a solution of 𝐿(𝑦) = 0 which is such that
𝜙(𝑥0 ) = 𝛼1 , 𝜙 ′ (𝑥0 ) = 𝛼2 , ⋯ , 𝜙 (𝑛−1) (𝑥0 ) = 𝛼𝑛 , (36)
where 𝛼1 , 𝛼2 , ⋯ , 𝛼𝑛 are real constants, then 𝜙 is real-valued. One way to see this is to note
that since
𝐿(𝜙) = 𝐿(𝜙) = 0,
𝜙 is also solution, and hence so is
1
𝜓(𝑥) = 2𝑖 [𝜙(𝑥) − 𝜙(𝑥)] = Im 𝜙.
But, from (36), we see that
𝜓(𝑥0 ) = 0, 𝜓′ (𝑥0 ) = 0, ⋯ , 𝜓 (𝑛−1) (𝑥0 ) = 0.
The uniqueness theorem implies that 𝜓(𝑥) = 0 for all 𝑥 , or Im 𝜙 = 0, showing that 𝜙 is
real-valued.
Example 05: Find the real valued general solution of 𝑦 ′′ + 5 𝑦 = 0.
Solution: Let 𝐿(𝑦) = 𝑦 ′′ + 5 𝑦 = 0. The characteristic polynomial is 𝑝(𝑟) = 𝑟 2 + 5, with
roots 𝑟 = ±5𝑖 . Therefore, the real valued fundamental solutions are 𝜙1 (𝑥) = cos 5𝑥 ,
𝜙2 (𝑥) = sin 5𝑥 . Thus, the real valued general solution is 𝜙(𝑥) = 𝑐1 cos 5𝑥 + 𝑐2 sin 5𝑥 ,
𝑐1 , 𝑐2 ∈ ℝ.

SOLVED PROBLEMS 01
Problem 01-04-01:
Find the Wronskian of the solutions of the 𝑦′′′ − 6𝑦′′ + 12𝑦′ − 8 = 0, on 𝐼: −∞ < 𝑥 < ∞.

S25013: Ordinary Differential Equations Page 60


Solution: Let L(y)= 𝑦′′ − 6𝑦′′ + 12𝑦′ − 8 = 0 . The characteristic polynomial of the
equation is 𝑝(𝑟)=𝑟 3 − 6𝑟 2 + 12𝑟 − 8 = (𝑟 − 2)3 . It has roots r=2 of multiplicity 3. Hence,
the solutions of L(y)=0 are given by 𝜙1 (𝑥) =𝑒 2𝑥 , 𝜙2 (𝑥) =𝑥𝑒 2𝑥 𝜙3 (𝑥) =𝑥 2 𝑒 2𝑥 . Hence, the
Wronskian of the solutions is given by
𝜙1 𝜙2 𝜙3
𝑊(𝜙1 , 𝜙2 , 𝜙3 )(𝑥) = |𝜙1 ′ 𝜙2′ 𝜙3′ |
𝜙1′′ 𝜙2′′ 𝜙3′′
2𝑥
𝑒 2𝑥 𝑥𝑒 2𝑥 𝑥 2𝑒
⇒ 𝑊(𝜙1 , 𝜙2 , 𝜙3 )(𝑥) = |2𝑒 2𝑥 (1 + 2𝑥)𝑒 2𝑥 2𝑥(1 + 𝑥)𝑒 2𝑥 |.
4𝑒 2𝑥 4(1 + 𝑥)𝑒 2𝑥 2(1 + 4𝑥 + 2𝑥 2 )𝑒 2𝑥
Though the determinant is of order 3, we see that it is tedious to evaluate. Hence, we take
1 0 0
a point 𝑥0 = 0 in 𝐼, 𝑊(𝜙1 , 𝜙2 , 𝜙3 )(0)=|2 1 0| = 2 ≠ 0.
4 4 2
Hence by using the Theorem 8 viz,
𝑊(𝜙1 , 𝜙2 , 𝜙3 )(𝑥)=𝑒 −𝑎1 (𝑥−𝑥0) 𝑊(𝜙1 , 𝜙2 , 𝜙3 )(𝑥0 ),
We have 𝑊(𝜙1 , 𝜙2 , 𝜙3 )(𝑥)=2𝑒 6𝑥 , ∀ 𝑥 ∈ 𝐼.

Problem 01-04-02:
Consider the equation 𝑦 ′′′ − 4𝑦 ′ = 0.
a) Compute three linearly independent solutions.
b) Compute the Wronskian of the solutions found in (a).
c) Find that the unique solution 𝜙 satisfying 𝜙(0) = 0, 𝜙 ′ (0) = 1, 𝜙 ′′ (0) = 0.
Solution: Let 𝐿(𝑦) = 𝑦 ′′′ − 4𝑦 ′ = 0 . The characteristic polynomial of the equation is
𝑝(𝑟)=𝑟 3 − 4𝑟 = 𝑟(𝑟 − 2)(𝑟 + 2). It has roots 𝑟 = 0, 2, −2.
(a) Hence, the three linearly independent solutions of 𝐿(𝑦) = 0 are given by
𝜙1 (𝑥)=1, 𝜙2 (𝑥)=𝑒 2𝑥 𝜙3 (𝑥) = 𝑒 −2𝑥 .
(b) The Wronskian of the solutions is given by
𝜙1 𝜙2 𝜙3
𝑊(𝜙1 , 𝜙2 , 𝜙3 )(𝑥) = |𝜙1 ′ 𝜙2′ 𝜙3′ |
𝜙1′′ 𝜙2′′ 𝜙3′′
1 𝑒 2𝑥 𝑒 −2𝑥
⇒ 𝑊(𝜙1 , 𝜙2 , 𝜙3 )(𝑥) = |0 2𝑒 2𝑥 −2𝑒 −2𝑥 |
0 4𝑒 2𝑥 4𝑒 −2𝑥
= 1(8𝑒 2𝑥−2𝑥 + 8𝑒 2𝑥−2𝑥 ) − 0 + 0
= 16.
(c) Therefore, all solutions of the given equation are given by
𝜙(𝑥) = 𝑐1 1 + 𝑐2 𝑒 2𝑥 + 𝑐3 𝑒 −2𝑥 , (37)
where 𝑐1 , 𝑐2 , 𝑐3 are constants.
Now, by using the initial conditions, we have
𝜙(0) = 0 ⟹ 𝑐1 + 𝑐2 + 𝑐3 = 0, (38)
𝜙 ′ (0) = 1 ⟹ 2𝑐2 − 2𝑐3 = 1, (39)
𝜙 ′′ (0)
= 0 ⟹ 4𝑐2 + 4𝑐3 = 0. (40)

S25013: Ordinary Differential Equations Page 61


1 1
On solving these equations, we get 𝑐1 = 0, 𝑐2 = 4 , 𝑐3 = − 4. Therefore, the unique solution
of the given problem is
1 1
𝜙(𝑥) = 4 𝑒 2𝑥 − 4 𝑒 −2𝑥
1 𝑒 2𝑥 −𝑒 −2𝑥
= 2. 2
sinh 2𝑥
= . (41)
2
Problem 01-04-03:
Find four linearly independent solutions of the equation 𝑦 (4) + 𝜆𝑦 = 0, in the each of the
following case:
(d) 𝜆 = 0, (b) 𝜆 > 0, (c) 𝜆 < 0.
Solution: The given equation is
𝑦 (4) + 𝜆𝑦 = 0. (42)
Case (a): 𝜆 = 0 . Then we have L(y)= 𝑦 = 0 . The characteristic polynomial of the
(4)

equation is 𝑝(𝑟)=𝑟 4 . It has roots 𝑟 = 0 is of multiplicity 4. Hence, all four solutions are
given by
𝜙1 (𝑥) = 1, 𝜙2 (𝑥) = 𝑥, 𝜙3 (𝑥) = 𝑥 2 , 𝜙4 (𝑥) = 𝑥 3 . (43)
Case (b): 𝜆 > 0, i. e. 𝜆 = 𝑘 , 𝑘 ≠ 0. Then the equation (42) becomes L(y)=𝑦 + 𝑘 𝑦 = 0.
2 (4) 2

The characteristic polynomial of the equation is 𝑝(𝑟)=𝑟 4 + 𝑘 2 . It has roots 𝑟 = ±√𝑖𝑘, ±𝑖√𝑖𝑘
1+𝑖 1+𝑖
and since √𝑖 = or √𝑖 = − , therefore, we can rewrite the roots in either case as 𝑟 =
√2 √2
𝑘 𝑘 𝑘 𝑘 𝑘 𝑘 𝑘 𝑘
√ + 𝑖√ , √ − 𝑖√ , −√ + 𝑖√ , −√ − 𝑖√ . Hence, all four solutions are given by
2 2 2 2 2 2 2 2

𝑘 𝑘 𝑘 𝑘
𝜙1 (𝑥) = cosh √2 𝑥 cos √2 𝑥, 𝜙2 (𝑥) = cosh √2 𝑥 sin √2 𝑥,

𝑘 𝑘 𝑘 𝑘
𝜙3 (𝑥) = sinh √2 𝑥 cos √2 𝑥, 𝜙4 (𝑥) = sinh √2 𝑥 sin √2 𝑥. (44)

Case (c): 𝜆 < 0, i. e. 𝜆 = −𝑘 2 , 𝑘 ≠ 0. Then the equation (42) becomes L(y)=𝑦 (4) − 𝑘 2 𝑦 =
0 . The characteristic polynomial of the equation is 𝑝(𝑟) = 𝑟 4 − 𝑘 2 . It has roots 𝑟 =
±√𝑘, ±𝑖√𝑘. Hence, all four solutions are given by
𝜙1 (𝑥) = cosh √𝑘 𝑥, 𝜙2 (𝑥) = sinh √𝑘 𝑥, 𝜙3 (𝑥) = cos √𝑘 𝑥, 𝜙4 (𝑥) = sin √𝑘 𝑥. (45)
Problem 01-04-04:
Find the real-valued general solution of the equation 𝑦 ′′ − 2𝑦 ′ + 6𝑦 = 0.
Solution: Let L(y)=𝑦 ′′ − 2𝑦 ′ + 6𝑦 = 0. The characteristic polynomial of the equation is
𝑝(𝑟)=𝑟 2 − 2𝑟 + 6. It has roots 𝑟 = 1 + 𝑖√5, 1 − 𝑖√5.
A fundamental solution set is
𝑧1 (𝑥) = 𝑒 (1+𝑖√5)𝑥 , 𝑧2 (𝑡) = 𝑒 (1−𝑖√5)𝑥 .
These are complex-valued functions. The general solution is
z(𝑡) = 𝑐1 𝑒 (1+𝑖√5)𝑥 + c2 𝑒 (1−𝑖√5)𝑥 , c1, c2 ∈ ℂ.
Any linear combination of these functions is solution of the differential equation. In
particular,

S25013: Ordinary Differential Equations Page 62


1 1
𝜙1 (𝑥) =
[𝑧1 (𝑥) + 𝑧2 (𝑥)], 𝜙2 (𝑥) = [𝑧 (𝑥) − 𝑧2 (𝑥)].
2 2𝑖 1
The Euler formula and its complex-conjugate formula, we obtain a real-valued
fundamental set as
𝜙1 (𝑥) = 𝑒 𝑥 cos √5 𝑥, 𝜙2 (𝑥) = 𝑒 𝑥 sin √5 𝑥.
Hence, the real-valued general solution is simple to obtain
𝜙(𝑥) = 𝑒 𝑥 [𝑐1 cos √5 𝑥 + 𝑐2 sin √5 𝑥 ], 𝑐1 , 𝑐2 ∈ ℝ.
We just restricted the coefficients 𝑐1 , 𝑐2 to be real-valued.

SELF-TEST 01
MCQ 01-04-01
Let 𝜙1 , 𝜙2 and 𝜙3 be three LI solutions of the equation 𝑦′′′ − 𝑦′′ + 12𝑦′ − 8𝑦 = 0 for 𝑥 ∈
[0,1] satisfying the condition
𝜙1 (0) = 1, 𝜙2 (0) = 0, 𝜙3 (0) = 0,
𝜙1′ (0) = 2, 𝜙2′ (0) = 1, 𝜙3′ (0) = 0
𝜙1′′ (0) = 1, 𝜙2′′ (0) = 4, 𝜙3′′ (0) = 2,
Then the Wronskian 𝑊(𝜙1 , 𝜙2 , 𝜙3 )(𝑥) of 𝜙1 , 𝜙2 , 𝜙3 at point 𝑥 = 1 is
A. −𝑒
B. 𝑒
C. 2𝑒
D. −2𝑒.
MCQ 01-04-02
Let 𝜙1 , 𝜙2 , 𝜙3 be three LI solutions of the equations 𝑦′′′ + 𝑦′′ + 𝑦′ + 𝑦 = 1 for 𝑥 ∈ [0,1]
satisfying the condition
𝜙1 (0) = 1, 𝜙1′ (0) = −1, 𝜙1′′ (0) = 1,
𝜙2 (0) = 1, 𝜙2′ (0) = 0, 𝜙2′′ (0) = −1
𝜙3 (0) = 0, 𝜙3′ (0) = 1, 𝜙3′′ (0) = 0.
Then the Wronskian 𝑊(𝜙1 , 𝜙2 )(𝑥) of 𝜙1 , 𝜙2 and 𝜙3 of point 𝑥 = 1 is
A. −𝑒
2
B. 𝑒
C. 2𝑒
1
D. 𝑒.
MCQ 01-04-03
Which one of the following sets of functions is linearly dependent?
A. {1, 𝑥, 𝑥 2 }
B. {𝑒 𝑖𝑥 , sin 𝑥, 2 cos 𝑥}
C. {𝑥, 𝑒 2𝑥 , |𝑥|}
D. {1, 𝑒 2𝑥 , 𝑒 −2𝑥 }.
MCQ 01-04-04
For 𝜆𝑖 ∈ ℝ, let 𝜙𝑖 : ℝ → ℝ be 𝜙𝑖 (𝑥) = 𝑒 𝜆𝑖𝑥 , 𝑖 = 1, 2, 3. Then

S25013: Ordinary Differential Equations Page 63


A. 𝜙1 , 𝜙2 , 𝜙3 are linearly independent
B. 𝜙1 , 𝜙2 , 𝜙3 are linearly independent if, and only if, 𝜆1 ≠ 𝜆2
C. 𝜙1 , 𝜙2 , 𝜙3 are linearly independent if, and only if, 𝜆1 ≠ 𝜆2 and 𝜆2 ≠ 𝜆3
D. 𝜙1 , 𝜙2 , 𝜙3 are linearly independent if, and only if, 𝜆1 ≠ 𝜆2 , 𝜆2 ≠ 𝜆3 and 𝜆3 ≠ 𝜆1 .

SHORT ANSWER QUESTIONS 01


SAQ 01- 04-01
Find the Wronskian of the solutions of the equation 𝑦′′′ + 𝑦′′ + 𝑦′ + 𝑦 = 1 at any point 𝑥 ∈
𝐼, containing zero.
Solution: The corresponding homogenous equation of the given equation is 𝐿(𝑦) = 0 ⟹
𝑦 ′′′ + 𝑦 ′′ + 𝑦 ′ + 𝑦 = 0. Hence, its solutions are 𝜙1 (𝑥) = 𝑒 −𝑥 , 𝜙2 (𝑥) = cos 𝑥, 𝜙3 (𝑥) = sin 𝑥.
Therefore,
𝑒 −𝑥 cos 𝑥 sin 𝑥
𝑊(𝜙1 , 𝜙2 , 𝜙3 )(𝑥) = |−𝑒 −𝑥 − sin 𝑥 cos 𝑥 | , ∀ 𝑥 ∈ 𝐼.
𝑒 −𝑥 − cos 𝑥 − sin 𝑥
But, for 𝑥0 = 0 ∈ 𝐼, we have
1 1 0
𝑊(𝜙1 , 𝜙2 , 𝜙3 )(0) = |−1 0 1| = 2.
1 −1 0
Hence by Wronskian formula, we have
𝑥
− ∫𝑥 𝑎1 𝑑𝑡
𝑊(𝜙1 , 𝜙2 , 𝜙3 )(𝑥) = 𝑒 0 𝑊(𝜙1 , 𝜙2 , 𝜙3 )(𝑥0 ), ∀ 𝑥 ∈ 𝐼
𝑥
− ∫0 1𝑑𝑡
=𝑒 𝑊(𝜙1 , 𝜙2 , 𝜙3 )(0)
−𝑥
= 2𝑒 , ∀ 𝑥 ∈ 𝐼.
SAQ 01- 04-02
Find all solutions of the equation 𝑦 ′′′ − 𝑖𝑦 ′′ + 4𝑦 ′ − 4𝑖𝑦 = 0.
Solution: Let L(y)= 𝑦 ′′′ − 𝑖𝑦 ′′ + 4𝑦 ′ − 4𝑖𝑦 = 0 . The characteristic polynomial of the
equation is 𝑝(𝑟)=𝑟 3 − 𝑖𝑟 2 + 4𝑟 − 4𝑖. It has roots 𝑟 = 𝑖, 2𝑖, −2𝑖. Hence, all solutions of the
given equations are given by
𝜙(𝑥) = 𝑐1 𝑒 𝑖𝑥 + 𝑐2 𝑒 2𝑖𝑥 + 𝑐3 𝑒 −2𝑖𝑥 ,
where 𝑐1 , 𝑐2 , 𝑐3 are any constants.
SAQ 01- 04-03
If a complex valued function 𝑧 is a solution of the equation 𝐿(𝑧) = 𝑧 ′′ + a1 𝑧 ′ + 𝑎2 𝑧 = 0,
where a1 , a2 are real, then show that the real and imaginary parts of 𝑧 are also solutions of
𝐿(𝑧) = 0.
Solution: Let 𝑦1 (𝑥) be the real part of 𝑧 and 𝑦2 (𝑥) the imaginary part. Then we can write
𝑧(𝑥) = 𝑦1 (𝑥) + 𝑖𝑦2 (𝑥).
Now, then we have,
𝐿(𝑧) = 𝑦1′′ (𝑥) + 𝑖𝑦2′′ (𝑥) + 𝑎1 𝑦1′ + 𝑖𝑎1 𝑦2′ + 𝑎2 𝑦1 + 𝑖𝑎2 𝑦2
= [𝑦1′′ (𝑥) + 𝑎1 𝑦1′ + 𝑎2 𝑦1 ] + 𝑖[𝑦2′′ (𝑥) + 𝑎1 𝑦2′ + 𝑎2 𝑦2 ]. (46)
Since 𝑧(𝑥) is a solution to the homogeneous 𝐿(𝑧) = 0, gives
𝐿(𝑧) = [𝑦1′′ (𝑥) + 𝑎1 𝑦1′ + 𝑎2 𝑦1 ] + 𝑖[𝑦2′′ (𝑥) + 𝑎1 𝑦2′ + 𝑎2 𝑦2 ] = 0. (47)

S25013: Ordinary Differential Equations Page 64


Both expressions in parentheses are real, so the only way the sum can be zero is for both
to be zero. That is,
[𝑦1′′ (𝑥) + 𝑎1 𝑦1′ + 𝑎2 𝑦1 ] = 0
and
[𝑦2′′ (𝑥) + 𝑎1 𝑦2′ + 𝑎2 𝑦2 ] = 0. (48)
This shows that both real part 𝑦1 (𝑥) and imaginary part 𝑦2 (𝑥) of complex solution 𝑧(𝑥)
are solutions of 𝐿(𝑧) = 0.

SUMMARY
In this Unit, we have discussed the method to solve the 𝑛-th order differential equation of
the form 𝐿(𝑦) = 𝑦 (𝑛) + 𝑎1 𝑦 (𝑛−1) + 𝑎2 𝑦 (𝑛−2) + ⋯ + 𝑎𝑛 𝑦 = 0. If 𝑟1 , 𝑟2 , ⋯ , 𝑟𝑛 are distinct roots
of 𝑝(𝑟) = 𝑟 𝑛 + 𝑎1 𝑟 𝑛−1 + 𝑎2 𝑟 𝑛−2 + ⋯ + 𝑎𝑛 , then 𝑒 𝑟1𝑥 , 𝑒 𝑟2𝑥 , ⋯ , 𝑒 𝑟𝑛𝑥 are linearly independent
solutions of the differential equation 𝐿(𝑦) = 0 . If 𝑟1 is a repeated root of 𝑝(𝑟) of
multiplicity 𝑚1 , then 𝑒 𝑟1𝑥 , 𝑥𝑒 𝑟1 𝑥 , ⋯ , 𝑥 𝑚1 −1 𝑒 𝑟1𝑥 are linearly independent solutions of the
differential equation 𝐿(𝑦) = 0.
For 𝑎1 , 𝑎2 , ⋯ , 𝑎𝑛 ∈ ℝ, if 𝛼 + 𝑖𝛽 is a root of 𝑝(𝑟), where 𝛼, 𝛽 ∈ ℝ and 𝛽 ≠ 0, then 𝑝(𝑟) has
also root 𝛼 − 𝑖𝛽 and in this case, the differential equation has 𝑒 𝛼𝑥 cos 𝛽𝑥 , 𝑒 𝛼𝑥 sin 𝛽𝑥 are
linearly independent solutions of the differential equation 𝐿(𝑦) = 0. Further, the linear
combination of these solutions is also a solution of the equation 𝐿(𝑦) = 0.

END OF UNIT EXERCISES


1) Suppose the roots with multiplicity of a certain homogeneous const ant coefficient
linear equation are 3, 4, 4, 4, 5 ± 2𝑖, 5 ± 2𝑖. Give the general real solution to the
equation. What is the order of the equation?
i )

2) Find all real-valued solutions of the equation 𝑦 ′′ + 𝑦 = 0.


3) Find the value of the Wronskian of three functions 𝑥 2 , 3𝑥 + 2, 2𝑥 + 3.
4) Find a homogeneous linear differential equation with real constant coefficients, which
has 𝜙(𝑥) = 𝑥𝑒 −3𝑥 cos 2𝑥 + 𝑒 −3𝑥 sin 2𝑥 as one of its solution.

KEY WORDS
Homogeneous differential equation of 𝑛 -th order, Wronskian of solutions, initial value
problem, existence theorem and uniqueness theorem.

S25013: Ordinary Differential Equations Page 65


C REDIT 02

S25013: Ordinary Differential Equations Page 66


UNIT 02-01: THE NON-HOMOGENEOUS EQUATION OF HIGHER ORDER
LEARNING OBJECTIVES
After successful completion of this unit, you will be able to
❖ explain particular solution of a linear nonhomogeneous differential equation with
constant coefficients
❖ apply the methods of variation of constants and annihilator for solving such
differential equations
01-01: NON-HOMOGENEOUS EQUATION OF ORDER 𝒏
The non – homogeneous equation with constant coefficients of order 𝑛 is given by
𝐿(𝑦) = 𝑦 (𝑛) + 𝑎1 𝑦 (𝑛−1) + 𝑎2 𝑦 (𝑛−2) + ⋯ + 𝑎𝑛 𝑦 = 𝑏(𝑥), (1)
where 𝑎1 , 𝑎2 , ⋯ ,𝑎𝑛 are constants and 𝑏(𝑥) is continuous function defined on an interval 𝐼.
Theorem 1: Let 𝑏(𝑥) be a continuous function on an interval 𝐼 and 𝜙1 , 𝜙2 , ⋯ , 𝜙𝑛 be n
linearly independent solutions of 𝐿(𝑦) = 𝑦 (𝑛) + 𝑎1 𝑦 (𝑛−1) + 𝑎2 𝑦 (𝑛−2) + ⋯ + 𝑎𝑛 𝑦 = 0 on 𝐼 .
Then every solution 𝜓 of 𝐿(𝑦) = 𝑏(𝑥) is given by
𝜓 = 𝜓𝑝 + 𝑐1 𝜙1 + 𝑐2 𝜙2 + ⋯ + 𝑐𝑛 𝜙𝑛 (2)
where 𝜓𝑝 is a particular solution of 𝐿(𝑦) = 𝑏(𝑥) and 𝑐1 , 𝑐2 , ⋯ , 𝑐𝑛 are constants and 𝜓𝑝 is
given by
𝑥 𝑊𝑘 (𝑡)𝑏(𝑡)
𝜓𝑝 (𝑥) = ∑𝑛𝑘=1 𝜙𝑘 (𝑥) ∫𝑥 𝑑𝑡, (3)
0 𝑊(𝜙1 ,𝜙2 ,⋯,𝜙𝑛 )(𝑡)
where 𝑊𝑘 is the determinant obtained from the Wronskain 𝑊(𝜙1 , 𝜙2 , ⋯ , 𝜙𝑛 ) by replacing
the its 𝑘 𝑡ℎ column by the column vector (0, 0, … ,0, 1).
Proof: For case 2 refer the Theorem 1 in Unit 01-01 and for the general case with variable
coefficients refer the Theorem 10 in the next Unit 02-02.

01-02: A SPECIAL METHOD FOR SOLVING THE NON-HOMOGENEOUS EQUATION


Introduction:
We have used the variation of constants method to find a solution of the non–homogeneous
differential equation 𝐿(𝑦) = 𝑏(𝑥). The method involves more labor than necessary. There
is another method which is often faster than the variation of constants method, of solving
the non-homogeneous differential equation 𝐿(𝑦) = 𝑏(𝑥). The method is called the
annihilator method.
The method involves the calculation of the homogeneous equation 𝑀(𝑦) = 0 with
constants coefficients of which 𝑏(𝑥) is a solution. Once such 𝑀 has been found the
problem becomes algebraic and no integrations are involved. Thus, the functions 𝑏(𝑥) must
be a sum of terms of the form 𝑃(𝑥)𝑒 𝑎𝑥 where 𝑃 is a polynomial and 𝑎 is a constant.
The differential operator 𝑀 is called annihilator and which annihilates the function 𝑏(𝑥),
i. e. 𝑀(𝑏(𝑥)) = 0.
Let 𝐿(𝑦) = 𝑏(𝑥) be a given differential equation with constants coefficients of order n. let
us assume, we could find the differential equation 𝑀(𝑦) = 0 with constants coefficients
of order 𝑚, that is satisfied by the function 𝑏(𝑥).
i.e. 𝑀(𝑏) = 0 ⇒ 𝑀(𝐿(𝜓)) = 0,

S25013: Ordinary Differential Equations Page 67


⇒ 𝜓(𝑥) is a solution of a homogeneous equation 𝐿(𝐿(𝑦)) = 0 with constant coefficients
of order 𝑛 + 𝑚. This implies that 𝜓(𝑥) can be expressed as a linear combination of 𝑚 +
𝑛 linearly independent solutions of 𝑀(𝐿(𝑦) = 0 . The method is illustrated in the
following example. However, note that not every linear combination will be a solution of
𝐿(𝑦) = 𝑏(𝑥).
Example 01: Use annihilator method and hence find a particular solution of the equation
𝐿(𝑦) = 𝑦′′ − 3𝑦′ + 2𝑦 = 𝑥 2 .
Solution: Let 𝐿(𝑦) = 𝑦′′ − 3𝑦′ + 3𝑦 = 𝑥 2 . Here 𝑏(𝑥) = 𝑥 2 . We find a homogeneous
differential equation 𝑀(𝑦) = 0 with constants coefficients such that 𝑏(𝑥) is a solution of
𝑀(𝑦) = 0. Such equation is 𝑀(𝑦) = 𝑦′′′ = 0. We see that 𝑦 = 𝑥 2 is a solution. Hence every
solution 𝜓 of the equation 𝐿(𝑦) = 𝑏(𝑥) is a solution of
𝑀(𝐿(𝑦)) = 𝑀(𝑏) = 0
Operating 𝑀 on 𝐿 we find that
𝑦 (5) − 3𝑦 (4) + 2𝑦 (3) = 0.
The characteristic polynomial of this equation is 𝑟 3 (𝑟 2 − 3𝑟 + 2). This is the product of
the characteristic polynomial of 𝐿 and 𝑀. The roots of the characteristic polynomial are
𝑟 = 0, 0, 0, 1, 2. Hence the solution 𝜓 is of the form
𝜓(𝑥) = 𝑐0 + 𝑐1 𝑥 + 𝑐2 𝑥 2 + 𝑐3 𝑒 𝑥 + 𝑐4 𝑒 2𝑥 .
We observe that 𝑐3 𝑒 𝑥 + 𝑐4 𝑒 2𝑥 a solution of L(y) = 0. Hence, we write equation the above
expression as
𝜓(𝑥)=𝜓𝑝 (𝑥) + 𝑐3 𝑒 𝑥 + 𝑐4 𝑒 2𝑥 ,
where, we have to find 𝑐0 , 𝑐1 , 𝑐2 such that where 𝜓𝑝 (𝑥)=𝑐0 + 𝑐1 𝑥 + +𝑐2 𝑥 2 is a particular
solution of 𝐿(𝑦) = 𝑥 2 .
Our aim now is to find the constants 𝑐0 , 𝑐1 and 𝑐2 such that 𝜓𝑝 (𝑥) =𝑐0 + 𝑐1 𝑥 + 𝑐2 𝑥 2 is a
solution of 𝐿(𝑦) = 𝑥 2 .
⇒ 𝐿(𝑐0 + 𝑐1 𝑥 + 𝑐2 𝑥 2 )=𝑥 2
⇒ (𝑐0 + 𝑐1 𝑥 + 𝑐2 𝑥 2 )′′ − 3(𝑐0 + 𝑐1 𝑥 + 𝑐2 𝑥 2 )′ + 2(𝑐0 + 𝑐1 𝑥 + 𝑐2 𝑥 2 ) = 𝑥 2
⇒ (𝑐1 + 2𝑐2 𝑥)′ − 3(𝑐1 + 2𝑐2 𝑥) + 2(𝑐0 + 𝑐1 𝑥 + 𝑐2 𝑥 2 ) = 𝑥 2
⇒ (2𝑐2 − 3𝑐1 + 2𝑐0 ) + (−6𝑐2 + 2𝑐1 )𝑥 + 2𝑐2 𝑥 2 = 𝑥 2 .
Equating the corresponding coefficients on both the sides, we get
1
2𝑐2 =1 ⇒ 𝑐2 = 2
3
2(𝑐1 − 3𝑐2 ) = 0 ⇒ 𝑐1 = 2
and
2𝑐2 − 3𝑐1 + 2𝑐0=0
1 1 9 7
⇒ 𝑐0 = 2 (3𝑐1 − 2𝑐2 ) = 2 (2 − 1) = 4.
1
Thus 𝜓𝑝 (𝑥)= 4 (7 + 6𝑥 + 2𝑥 3 ) is a particular solution of 𝐿(𝑦) = 𝑥 2 .
Example 02: Find the particular solution of 𝑦′′ + 4𝑦 = cos 𝑥 by annihilator method.
Solution: Let 𝐿(𝑦) = 𝑦′′ + 4𝑦 = cos 𝑥 were 𝑏(𝑥) = cos 𝑥 . We see that 𝑏(𝑥) = cos 𝑥 is a
solution of the constant coefficients equation 𝑀(𝑦) = 𝑦′′ + 𝑦 = 0. Hence, every solution of
the equation 𝐿(𝑦) = 𝑏(𝑥) is a solution of the equation

S25013: Ordinary Differential Equations Page 68


𝑀 (𝐿(𝑦)) = 𝑀(𝑏(𝑥)) = 0
⇒ 𝑀(𝐿(𝑦))=0.
Operating 𝑀 on 𝐿 we obtain (𝐿(𝑦))′′ + 𝐿(𝑦)=0
⇒ (𝑦′′ + 4𝑦)′′ + 𝑦′′ + 4𝑦=0
⇒ 𝑦 (4) + 4𝑦′′ + 𝑦′′ + 4𝑦=0
⇒ 𝑦 (4) + 5𝑦′′ + 4𝑦=0.
The characteristic polynomial of the equation is 𝑝(𝑟)=𝑟 4 + 5𝑟 2 + 4 ⇒ 𝑝(𝑟)=(𝑟 2 + 1)(𝑟 2 +
4) ⇒ 𝑝(𝑟) = product of the characteristic polynomial of 𝐿 and 𝑀. The root of 𝑝(𝑟) are 𝑟 =
±𝑖 and 𝑟 = ±2𝑖. Hence the solution 𝜓 is given by
𝜓(𝑥) = 𝑐1 𝑒 𝑖𝑥 + 𝑐2 𝑒 −𝑖𝑥 + 𝑐3 𝑒 2𝑖𝑥 + 𝑐4 𝑒 −2𝑖𝑥 .
We see that 𝑐3 𝑒 2𝑖𝑥 + 𝑐4 𝑒 −2𝑖𝑥 is a solution of 𝐿(𝑦) = 0. We write the above equation as
𝜓(𝑥) = 𝜓𝑝 (𝑥) + 𝑐3 𝑒 2𝑖𝑥 + 𝑐4 𝑒 −2𝑖𝑥 .
Now our aim is to find the constants 𝑐1 and 𝑐2 such that 𝜓𝑝 (𝑥)=𝑐1 𝑒 𝑖𝑥 + 𝑐2 𝑒 −𝑖𝑥 is particular
solution of 𝐿(𝑦) = 𝑏(𝑥)
i.e. 𝐿 (𝜓𝑝 (𝑥)) = 𝜓𝑝′′ (𝑥) + 4𝜓𝑝 (𝑥) = cos 𝑥
⇒ −𝑐1 𝑒 𝑖𝑥 − 𝑐2 𝑒 −𝑖𝑥 + 4(𝑐1 𝑒 𝑖𝑥 + 𝑐2 𝑒 −𝑖𝑥 ) = cos 𝑥
⇒ 3(𝑐1 𝑒 𝑖𝑥 + 𝑐2 𝑒 −𝑖𝑥 ) = cos 𝑥
⇒ 3[𝑐1 cos 𝑥 + 𝑐1 𝑖 sin 𝑥 + 𝑐2 cos 𝑥 − 𝑐2 𝑖 sin 𝑥] = cos 𝑥
⇒ 3[𝑐1 cos 𝑥 + 𝑐1 𝑖 sin 𝑥 + 𝑐2 cos 𝑥 − 𝑖𝑐2 sin 𝑥] = cos 𝑥
⇒ 3(𝑐1 + 𝑐2 ) cos 𝑥 + 3𝑖(𝑐1 − 𝑐2 ) sin 𝑥 = cos 𝑥
⇒ 3(𝑐1 + 𝑐2 ) =1, 3𝑖(𝑐1 − 𝑐2 ) = 0
⇒ 𝑐1 =𝑐2 and 3(2𝑐2 ) = 1
1
⇒ 𝑐1 =𝑐2 = 6.
Hence the particular solution 𝜓𝑝 (𝑥) becomes
1 1
𝜓𝑝 (𝑥) = 6 (𝑒 𝑖𝑥 + 𝑒 −𝑖𝑥 ) = 6 (2 cos 𝑥)
1
⇒ 𝜓𝑝 (𝑥) = 3 cos 𝑥 is a particular solution of 𝐿(𝑦) = cos 𝑥.
𝑑
Example 03: Show that the constant coefficient operator 𝐷𝑛+1, where 𝐷 = 𝑑𝑥, annihilates
each of the functions 1, 𝑥, 𝑥 2 , ⋯ , 𝑥 𝑛 .
Solution: Using the fact that differentiation can be done term by term, a polynomial
𝑐0 + 𝑐1 𝑥 + ⋯ + 𝑐𝑛 𝑥 𝑛
can be annihilated by finding an operator that annihilates the highest power of 𝑥 .
Therefore, the functions that are annihilated by a linear operator 𝑀 = 𝐷𝑛+1 are simply
those functions that can be obtained from the general solution of the homogeneous
differential equation 𝑀(𝑦) = 0.
Example 04: Show that the constant coefficient operator (𝐷 − 𝑎)𝑛+1 annihilates each of
the functions 𝑒 𝑎𝑥 , 𝑥𝑒 𝑎𝑥 , ⋯ , 𝑥 𝑛 𝑒 𝑎𝑥 .
Solution: To see this, note that the characteristic equation of the homogenous differential
equation (𝐷 − 𝑎)𝑛+1 𝑦 = 0 is (𝑟 − 𝑎)𝑛+1 = 0 . Since 𝑎 is a root of multiplicity 𝑛 + 1 , the
general solution is
S25013: Ordinary Differential Equations Page 69
𝑦(𝑥) = (𝑐1 + 𝑐2 𝑥 + ⋯ + 𝑐𝑛 𝑥 𝑛 )𝑒 𝑎𝑥
⟹ 𝑦(𝑥) = 𝑐1 𝑒 𝑎𝑥 + 𝑐2 𝑥𝑒 𝑎𝑥 + ⋯ + 𝑐𝑛 𝑥 𝑛 𝑒 𝑎𝑥 .
Hence, all functions 𝑒 𝑎𝑥 , 𝑥𝑒 𝑎𝑥 , ⋯ , 𝑥 𝑛 𝑒 𝑎𝑥 are obtained solutions of the equation 𝑀(𝑦) = 0
with 𝑀 = (𝐷 − 𝑎)𝑛+1 and consequently all are annihilated by (𝐷 − 𝑎)𝑛+1.
Example 05: Show that the constant coefficient operator [𝐷2 − 2𝑎𝐷 + 𝑎2 + 𝑏 2 ]𝑛+1
annihilates each of the functions (for 𝑎 and 𝑏, 𝑏 > 0 are real numbers):
𝑒 𝑎𝑥 cos 𝑏𝑥 , 𝑥𝑒 𝑎𝑥 cos 𝑏𝑥 , ⋯ , 𝑥 𝑛 𝑒 𝑎𝑥 cos 𝑏𝑥,
𝑒 𝑎𝑥 sin 𝑏𝑥 , 𝑥𝑒 𝑎𝑥 sin 𝑏𝑥 , ⋯ , 𝑥 𝑛 𝑒 𝑎𝑥 sin 𝑏𝑥.
Solution: To see this, note that the characteristic equation of the homoge nous differential
equation [𝐷2 − 2𝑎𝐷 + 𝑎2 + 𝑏 2 ]𝑛+1 𝑦 = 0 is [𝑟 2 − 2𝑎𝑟 + 𝑎2 + 𝑏 2 ]𝑛+1 = 0 . When 𝑎 and 𝑏,
𝑏 > 0 are real numbers, the quadratic formula reveals that [𝑟 2 − 2𝑎𝑟 + 𝑎2 + 𝑏 2 ]𝑛+1 = 0 has
complex roots 𝑎 + 𝑖𝑏, 𝑎 − 𝑖𝑏, both of multiplicity 𝑛 + 1, therefore the general solution is
𝑦(𝑥) = 𝑒 𝑎𝑥 [(𝑐1 + 𝑐2 𝑥 + ⋯ + 𝑐𝑛+1 𝑥 𝑛 ) cos 𝑏𝑥 + (𝑐1′ + 𝑐2′ 𝑥 + ⋯ + 𝑐𝑛+1 ′
𝑥 𝑛 ) sin 𝑏𝑥].
⟹ 𝑦(𝑥) = 𝑐1 𝑒 𝑎𝑥 cos 𝑏𝑥 + 𝑐2 𝑥𝑒 𝑎𝑥 cos 𝑏𝑥 + ⋯ + 𝑐𝑛+1 𝑥 𝑛 𝑒 𝑎𝑥 cos 𝑏𝑥
+𝑐1′ 𝑒 𝑎𝑥 sin 𝑏𝑥 + 𝑐2′ 𝑥 𝑒 𝑎𝑥 sin 𝑏𝑥 + ⋯ + 𝑐𝑛+1 ′
𝑥 𝑛 𝑒 𝑎𝑥 sin 𝑏𝑥,
where 𝑐1 , 𝑐2 , ⋯ , 𝑐𝑛+1 , 𝑐1′ , 𝑐2′ , ⋯ , 𝑐𝑛+1

are arbitrary constants.
This proves that the operator [𝐷 − 2𝑎𝐷 + 𝑎2 + 𝑏 2 ]𝑛+1 annihilates all given functions.
2

Example 06: Consider the constant coefficient operator 𝐿 with characteristic polynomial
𝑝 . If 𝑎 is not a root of the characteristic polynomial 𝑝 of 𝐿 , then prove that there is a
𝐴𝑒 𝑎𝑥
solution 𝜓𝑝 (𝑥) of the equation 𝐿(𝑦) = 𝐴𝑒 𝑎𝑥 of the form 𝜓𝑝 (𝑥) = , where 𝐴, 𝑎 are
𝑝(𝑎)
constants.
Solution: Let 𝐿 be the constant coefficient operator with characteristic polynomial 𝑝 .
Suppose 𝑎 is not a root of the characteristic polynomial 𝑝 of 𝐿, i. e. 𝑝(𝑎) ≠ 0. According
to the annihilator method, one can easily find that the operator 𝑀 given by 𝑀(𝑦) = 𝑦 ′ −
𝑎𝑦, with characteristic polynomial 𝑟 − 𝑎, annihilates the function 𝐴𝑒 𝑎𝑥 .
Therefore, the characteristic polynomial of the equation 𝑀(𝐿(𝑦)) = 0 or simply 𝑀𝐿 is
(𝑟 − 𝑎)𝑝(𝑟). Since 𝑎 is not a root of the polynomial 𝑝(𝑟), so we have 𝑎 as a simple root
(multiplicity one) of the polynomial (𝑟 − 𝑎)𝑝(𝑟). Thus, any solution the given equation has
the form
𝜓(𝑥) = 𝑐0 𝑒 𝑎𝑥 + 𝜙(𝑥),
where 𝐿(𝜙) = 0 and 𝑐0 is a constant.
The problem is to determine the constant 𝑐0 so that 𝐿(𝜓𝑝 ) = 𝐴𝑒 𝑎𝑥 , where 𝜓𝑝 (𝑥) = 𝑐0 𝑒 𝑎𝑥 .
i.e. 𝐿 (𝜓𝑝 (𝑥)) = 𝐴𝑒 𝑎𝑥 ⇒ 𝐿(𝑐0 𝑒 𝑎𝑥 ) = 𝐴𝑒 𝑎𝑥
⇒ 𝑐0 𝐿(𝑒 𝑎𝑥 ) = 𝐴𝑒 𝑎𝑥
⇒ 𝑐0 𝑝(𝑎)𝑒 𝑎𝑥 = 𝐴𝑒 𝑎𝑥
⇒ 𝑐0 𝑝(𝑎) = 𝐴
𝐴
⇒ 𝑐0 = 𝑝(𝑎) (since 𝑝(𝑎) ≠ 0).
Substituting the value of 𝑐0 in the function 𝜓𝑝 (𝑥), we get
𝐴𝑒 𝑎𝑥
𝜓𝑝 (𝑥) = .
𝑝(𝑎)

S25013: Ordinary Differential Equations Page 70


Therefore, we have shown that, if 𝑎 is not a root of the characteristic polynomial 𝑝 of
𝐿(𝑦) = 0, there is a solution (is called particular solution) of 𝐿(𝑦) = 𝐴𝑒 𝑎𝑥 of the form
𝐴𝑒 𝑎𝑥
𝜓𝑝 (𝑥) = .
𝑝(𝑎)

Example 07: Consider the constant coefficient operator 𝐿 with characteristic polynomial
𝑝 . Consider the equation 𝐿(𝑦) = 𝑒 𝑎𝑥 , where 𝑎 is a constant. If 𝑎 is a root of the
characteristic polynomial 𝑝 of 𝐿 with multiplicity 𝑘, then show by the annihilator method
𝑥 𝑘 𝑒 𝑎𝑥
that a solution 𝜓𝑝 (𝑥) of the equation 𝐿(𝑦) = 𝑒 𝑎𝑥 is given by 𝜓𝑝 (𝑥) = 𝑝(𝑘)(𝑎) .
Solution: Similarly, as discussed in the above Example 06 and from Theorem 1 of the
Unit 01-01, we have the characteristic polynomial of the equation 𝑀(𝐿(𝑦)) = 0 or simply
𝑀𝐿 is (𝑟 − 𝑎)𝑝(𝑟) with 𝑝(𝑘) (𝑎) ≠ 0 . But 𝑎 is a root of the polynomial 𝑝(𝑟) with
multiplicity 𝑘 , therefore 𝑎 is a root of the polynomial (𝑟 − 𝑎)𝑝(𝑟) with multiplicity
𝑘 + 1. Thus, any solution the given equation has the form
𝜓(𝑥) = (𝑐1 + 𝑐2 𝑥 + 𝑐3 𝑥 2 + ⋯ + 𝑐𝑘−1 𝑥 𝑘−1 + 𝑐𝑘+1 𝑥 𝑘 )𝑒 𝑎𝑥
+ combination of solutions corresponding to the remaining roots
of the characteristic polynomial 𝑝(𝑟) of the equation 𝐿(𝑦) = 0.
We separate it and write as
𝜓(𝑥) = 𝑐𝑘+1 𝑥 𝑘 𝑒 𝑎𝑥 + 𝜙(𝑥),
where 𝐿(𝜙) = 0 and 𝑐𝑘+1 is a constant is to be determined.
Find the value of the constant 𝑐𝑘 so that 𝐿(𝜓𝑝 ) = 𝑒 𝑎𝑥 , where 𝜓𝑝 (𝑥) = 𝑐𝑘+1 𝑥 𝑘 𝑒 𝑎𝑥 and as 𝐿
annihilates the function (𝑐1 + 𝑐2 𝑥 + 𝑐3 𝑥 2 + ⋯ + 𝑐𝑘 𝑥 𝑘−1 )𝑒 𝑎𝑥 + 𝜙(𝑥)
i.e. 𝐿 (𝜓𝑝 (𝑥)) = 𝑒 𝑎𝑥 ⇒ 𝐿(𝑐𝑘+1 𝑥 𝑘 𝑒 𝑎𝑥 ) = 𝑒 𝑎𝑥
⇒ 𝑐𝑘+1 𝐿(𝑥 𝑘 𝑒 𝑎𝑥 ) = 𝑒 𝑎𝑥
𝑘(𝑘−1)
⇒ 𝑐𝑘+1 [𝑝(𝑘) (𝑎) + 𝑘𝑝(𝑘−1) (𝑎)𝑥 + 𝑝(𝑘−2) (𝑎)𝑥 2 + ⋯ + 𝑝(𝑎)𝑥 𝑘 ] 𝑒 𝑎𝑥 = 𝑒 𝑎𝑥
2!
⇒ 𝑐𝑘+1 𝑝(𝑘) (𝑎)𝑒 𝑎𝑥 = 𝑒 𝑎𝑥 (by Theorem 1, 𝑝(𝑎) = 𝑝′′ (𝑎) = ⋯ = 𝑝(𝑘−1) (𝑎) = 0)
⇒ 𝑐𝑘+1 𝑝(𝑘) (𝑎) = 1
1
⇒ 𝑐𝑘+1 = (Theorem 1⇒ 𝑝(𝑘) (𝑎) ≠ 0).
𝑝(𝑘) (𝑎)
Substituting the value of 𝑐𝑘+1 in the function 𝜓𝑝 (𝑥), we get
𝑒 𝑎𝑥
𝜓𝑝 (𝑥) = 𝑝(𝑘)(𝑎).

Remark: To find the particular solution 𝐿(𝑦) = 𝑏(𝑥) by the annihilator method, we have
seen from the above examples that what we require is the characteristic polynomial 𝑞(𝑟)
of the equation 𝑀(𝑦) = 0. Students are advised to remember the characteristic polynomial
of 𝑀(𝑦) = 0 for some functions which are given in the following table.

Functions 𝑏(𝑥) 𝑥𝑘 𝑒 𝑎𝑘 𝑥 𝑘 𝑒 𝑎𝑥 sin 𝑎𝑥 or cos 𝑎𝑥 𝑥 𝑘 sin 𝑥 or 𝑥 𝑘 cos 𝑎 𝑥


Characteristic
𝑟 𝑘+1 r–a (𝑟 − 𝑎)𝑘+1 𝑟 2 + 𝑎2 (𝑟 2 + 𝑎2 )𝑘+1
polynomial

S25013: Ordinary Differential Equations Page 71


01-03: ALGEBRA OF CONSTANT COEFFICIENT OPERATORS
In order to justify the annihilator method, we study the algebra of constant coefficient
operators a little more carefully. For the type of equation, we have in mind
𝑎0 𝑦 (𝑛) + 𝑎1 𝑦 (𝑛−1) + ⋯ + 𝑎𝑛 𝑦 = 𝑏(𝑥),
where 𝑎0 ≠ 0, 𝑎1 , 𝑎2 , ⋯ , 𝑎𝑛 are constants, and 𝑏 is a sum of products of polynomials and
exponentials, every solution 𝜓 has all derivatives on −∞ < 𝑥 < ∞. This follows from the
fact that 𝜓 has 𝑛 derivatives there, and
𝑎 𝑎 𝑎
𝜓 (𝑛) = 𝑏 − 𝑎1 𝜓 (𝑛−1) − 𝑎2 𝜓 (𝑛−2) − ⋯ − 𝑎𝑛 𝜓,
0 0 0

where 𝑏 has all derivatives on −∞ < 𝑥 < ∞.


All the operators we now define will be assumed to be defined on the set of all functions
𝜙 on −∞ < 𝑥 < ∞ which have all derivatives there. Let 𝐿 and 𝑀 denote the operators
given by
𝐿(𝜙) = 𝑎0 𝜙 (𝑛) + 𝑎1 𝜙 (𝑛−1) + ⋯ + 𝑎𝑛 𝜙,
𝑀(𝜙) = 𝑏0 𝜙 (𝑚) + 𝑏1 𝜙 (𝑚−1) + ⋯ + 𝑏𝑚 𝜙,
where 𝑎0 ≠ 0, 𝑎1 , 𝑎2 , ⋯ , 𝑎𝑛 ; 𝑏0 ≠ 0, 𝑏1 , 𝑏2 , ⋯ , 𝑏𝑚 are constants. It will be convenient in what
follows to consider 𝑎0 , 𝑏0 which are not necessary 1. The characteristic polynomials of 𝐿
and 𝑀 are thus
𝑝(𝑟) = 𝑎0 𝑟 𝑛 + 𝑎1 𝑟 𝑛−1 + ⋯ + 𝑎𝑛 ,
and
𝑞(𝑟) = 𝑏0 𝑟 𝑚 + 𝑏1 𝑟 𝑚−1 + ⋯ + 𝑏𝑚 ,
respectively.
We define the sum 𝐿 + 𝑀 to be the operator given by
(𝐿 + 𝑀)(𝜙) = 𝐿(𝜙) + 𝑀(𝜙)
and the product 𝑀𝐿 to be the operator given by
(𝐿𝑀)(𝜙) = 𝑀(𝐿(𝜙)).
If 𝛼 is a constant, we define 𝛼𝐿 by
(𝛼𝐿)(𝜙) = 𝛼𝐿(𝜙).
We note that 𝐿 + 𝑀, 𝑀𝐿 and 𝛼𝐿 , are all linear differential operators with constant
coefficients.
Two operators 𝐿 and 𝑀 are said to be the equal if
𝐿(𝜙) = 𝑀(𝜙)
for all 𝜙 which have an infinite number of derivatives on −∞ < 𝑥 < ∞.
Suppose 𝐿, 𝑀 have characteristic polynomials 𝑝, 𝑞 respectively. Since 𝑒 𝑟𝑥 , for any constant
𝑟, has an infinite number of derivatives on −∞ < 𝑥 < ∞, we see that if 𝐿 = 𝑀, then
𝐿(𝑒 𝑟𝑥 ) = 𝑝(𝑟)𝑒 𝑟𝑥 = 𝑀(𝑒 𝑟𝑥 ) = 𝑞(𝑟)𝑒 𝑟𝑥 ,
and hence 𝑝(𝑟) = 𝑞(𝑟) for all 𝑟. This implies that 𝑚 = 𝑛 and 𝑎𝑘 = 𝑏𝑘 , 𝑘 = 0,1,2 ⋯ 𝑛. Thus
𝐿 = 𝑀 if and only if 𝐿 and 𝑀 have the same order and the same coefficients, if and only if
𝑝 = 𝑞.
If 𝐷 is the differential operator
𝐷(𝜙) = 𝜙 ′ ,
We define 𝐷2 = 𝐷𝐷, and successively

S25013: Ordinary Differential Equations Page 72


𝐷𝑘 = 𝐷𝐷𝑘−1 , (𝑘 = 2,3, ⋯ ).
For completeness we define 𝐷0 by 𝐷0 (𝜙) = 𝜙 but not usually write it explicitly. If 𝛼 is a
constant we understand by 𝛼 operating on a function 𝜙 just multiplication by 𝛼. Thus
𝛼(𝜙) = (𝛼𝐷0 𝜙) = 𝛼𝜙.
Now, using our definitions, it is clear that
𝐿 = 𝑎0 𝐷𝑛 + 𝑎1 𝐷𝑛−1 + ⋯ + 𝑎𝑛 ,
and
𝑀 = 𝑏0 𝐷𝑚 + 𝑏1 𝐷𝑚−1 + ⋯ + 𝑏𝑚 .
Theorem 2: The correspondence which associates with each
𝐿 = 𝑎0 𝐷𝑛 + 𝑎1 𝐷𝑛−1 + ⋯ + 𝑎𝑛 ,
its characteristic polynomial 𝑝 given by
𝑝(𝑟) = 𝑎0 𝑟 𝑛 + 𝑎1 𝑟 n−1 + ⋯ + 𝑎𝑛
is a one-to-one correspondence between all linear differential operators with constant
coefficients and all polynomials. If 𝐿, 𝑀 are associated with characteristic polynomials 𝑝, 𝑞
respectively, then 𝐿 + 𝑀 is associated with 𝑝 + 𝑞 , 𝑀𝐿 is associated with 𝑝𝑞 , and 𝛼𝐿 is
associated with 𝛼𝑝 (𝛼 a constant).
Proof: We have already seen that the correspondence is one-to-one since 𝐿 = 𝑀 if and only
if 𝑝 = 𝑞. The remainder of the theorem can be shows directly, or by noting
(𝐿 + 𝑀)(𝑒 𝑟𝑥 ) = 𝐿(𝑒 𝑟𝑥 ) + 𝑀(𝑒 𝑟𝑥 ) = 𝑝(𝑟)𝑒 𝑟𝑥 + 𝑞(𝑟)𝑒 𝑟𝑥 ,
(𝐿𝑀)(𝑒 𝑟𝑥 ) = 𝑀(𝐿(𝑒 𝑟𝑥 )) = 𝑀(𝑝(𝑟)𝑒 𝑟𝑥 ) = 𝑝(𝑟)𝑀(𝑒 𝑟𝑥 ) = 𝑝(𝑟)𝑞(𝑟)𝑒 𝑟𝑥 ,
(𝛼𝐿)(𝑒 𝑟𝑥 ) = 𝛼𝐿(𝑒 𝑟𝑥 ) = 𝛼𝑝(𝑟)𝑒 𝑟𝑥 .
This result implies that the algebraic properties of the constant coefficient operators are
the same as those of the polynomials. For example, si nce 𝐿𝑀 and 𝑀𝐿 both have the
characteristic polynomial 𝑝𝑞 , we have 𝐿𝑀 = 𝑀𝐿 . We remark that if 𝐿 and 𝑀 are not
constant coefficient operators, then it may not be true that 𝐿𝑀 = 𝑀𝐿 . For example, if
𝐿(𝜙) = 𝜙 ′ , 𝑀(𝜙) = 𝑥𝜙′, then
(𝐿𝑀 − 𝑀𝐿)(𝜙)(𝑥) = 𝐿(𝑀(𝜙)) − 𝑀(𝐿(𝜙))
= 𝐿(𝑥𝜙 ′ ) − 𝑀(𝜙 ′ )
= (𝑥𝜙)′′ − 𝑥𝜙 ′′
= 𝑥𝜙 ′′ + 2𝜙 ′ − 𝑥𝜙′′
= 2𝜙 ′ (𝑥),
where 𝜙 is a function of 𝑥, differentiable upto order 2. This proves that 𝐿𝑀 ≠ 𝑀𝐿.
Theorem 3: Consider the equation with constant coefficients
𝐿(𝑦) = 𝑃(𝑥)𝑒 𝑎𝑥 ,
where 𝑃 is the polynomial given by
𝑃(𝑥) = 𝑏0 𝑥 𝑚 + 𝑏1 x m−1 + ⋯ + 𝑏𝑚 (𝑏0 ≠ 0).
Suppose 𝑎 is root of the characteristic polynomial 𝑝 of 𝐿 of multiplicity 𝑗. Then there is a
unique solution 𝜓 of 𝐿(𝑦) = 𝑃(𝑥)𝑒 𝑎𝑥 of the form
𝜓(𝑥) = 𝑥 𝑗 (𝑐0 𝑥 𝑚 + 𝑐1 𝑥 𝑚−1 + ⋯ + 𝑐𝑚 )𝑒 𝑎𝑥 ,
where 𝑐0 , 𝑐1 , ⋯ , 𝑐𝑚 are constants determined by the annihilator method.
Proof: Let 𝐿(𝑦) = 𝑃(𝑥)𝑒 𝑎𝑥 (4)

S25013: Ordinary Differential Equations Page 73


be the equation with constant coefficients, where 𝑃 is the polynomial given by
𝑃(𝑥) = 𝑏0 𝑥 𝑚 + 𝑏1 x m−1 + ⋯ + 𝑏𝑚 (𝑏0 ≠ 0). (5)
Suppose 𝑎 is root of the characteristic polynomial 𝑝 of the equation 𝐿(𝑦) = 0 of
multiplicity 𝑗. The proof makes use of the formula
𝑘(𝑘−1) ′′
𝐿(𝑥 𝑘 𝑒 𝑟𝑥 ) = [𝑝(𝑟)𝑥 𝑘 + 𝑘𝑝′ (𝑟)𝑥 𝑘−1 + 2!
𝑝 (𝑟)𝑥 𝑘−2 + ⋯ + 𝑘𝑝(𝑘−1) (𝑟)𝑥 + 𝑝(𝑘) (𝑟)] 𝑒 𝑟𝑥 . (6)
The coefficients of 𝑝(𝑙) (𝑟)𝑥 𝑘−𝑙 in the bracket is the binomial coefficient
𝑘!
(𝑘𝑙) = (𝑘−𝑙)! 𝑙!.
Thus, we may write
𝐿(𝑥 𝑘 𝑒 𝑟𝑥 ) = [∑𝑘𝑙=0(𝑘𝑙) 𝑝(𝑙) (𝑟)𝑥 𝑘−𝑙 ]𝑒 𝑟𝑥 ,
where we understand 0! = 1.
An annihilator of the right side of (4) is
𝑀 = (𝐷 − 𝑎)𝑚+1 ,
with characteristic polynomial given by
𝑞(𝑟) = (𝑟 − 𝑎)𝑚+1 .
Since 𝑎 is a root of 𝑝 with multiplicity 𝑗 , it is a root of 𝑝𝑞 with multiplicity 𝑗 + 𝑚 + 1.
Thus, solution of 𝑀(𝐿(𝑦)) = 0 is of the form
𝜓(𝑥) = (𝑐0 𝑥 𝑗+𝑚 + 𝑐1 𝑥 𝑗+𝑚−1 + ⋯ + 𝑐𝑗+𝑚 )𝑒 𝑎𝑥 + 𝜙(𝑥), (7)
where 𝐿(𝜙) = 0, and 𝜙 involves exponentials of the form 𝑒 𝑠𝑥 with 𝑠 a root of 𝑝, 𝑠 ≠ 𝑎 .
Since 𝑎 is a root of 𝑝 with multiplicity 𝑗, we have that
(𝑐𝑚+1 𝑥 𝑗−1 + 𝑐𝑚+2 𝑥 𝑗−2 + ⋯ + 𝑐𝑗+𝑚 )𝑒 𝑎𝑥 (8)
is a solution of 𝐿(𝑦) = 0. Consequently, we see that there is a solution 𝜓 of (4) having the
form
𝜓(𝑥) = 𝑥 𝑗 (𝑐0 𝑥 𝑚 + 𝑐1 𝑥 𝑚−1 + ⋯ + 𝑐𝑚 )𝑒 𝑎𝑥 , (9)
where𝑐0 , 𝑐1 , ⋯ , 𝑐𝑚 are constants.
We now show that these constants are uniquely determined by the requirement that 𝜓
satisfy (4). We substitute (9) into 𝐿, we obtain
𝐿(𝜓) = 𝑐0 𝐿(𝑥 𝑗+𝑚 𝑒 𝑎𝑥 ) + 𝑐1 𝐿(𝑥 𝑗+𝑚−1 𝑒 𝑎𝑥 ) + ⋯ + 𝑐𝑚 𝐿(𝑥 𝑗 𝑒 𝑎𝑥 ). (10)
The terms in this sum can be computed using (6). We note that
𝑝(𝑎) = 𝑝′ (𝑎) = ⋯ = 𝑝(𝑗−1) (𝑎) = 0, 𝑝(𝑗) (𝑎) ≠ 0,
since 𝑎 is root of 𝑝 with multiplicity 𝑗. Thus, if 𝑘 ≥ 𝑗,
𝑘
𝐿(𝑥 𝑘 𝑒 𝑎𝑥 ) = [(𝑘−𝑗 𝑘
) 𝑝(𝑗) (𝑎)𝑥 𝑘−𝑗 + (𝑘−𝑗−1) 𝑝(𝑗+1) (𝑎)𝑥 𝑘−𝑗−1 + ⋯ + 𝑝(𝑘) (𝑎)] 𝑒 𝑎𝑥 .
We then have
𝐿(𝑥 𝑗+𝑚 𝑒 𝑎𝑥 ) = [(𝑗+𝑚
𝑚
𝑗+𝑚
)𝑝(𝑗) (𝑎)𝑥 𝑚 + (𝑚−1 )𝑝(𝑗+1) (𝑎)𝑥 𝑚−1 + ⋯ + 𝑝(𝑗+𝑚) (𝑎)]𝑒 𝑎𝑥 ,
𝐿(𝑥 𝑗+𝑚−1 𝑒 𝑎𝑥 ) = [(𝑗+𝑚−1
𝑚−1
)𝑝(𝑗) (𝑎)𝑥 𝑚−1 + (𝑗+𝑚−1
𝑚−2
)𝑝(𝑗+1) (𝑎)𝑥 𝑚−2 + ⋯ + 𝑝(𝑗+𝑚−1) (𝑎)]𝑒 𝑎𝑥 ,



𝑗
𝐿(𝑥 𝑗 𝑒 𝑎𝑥 ) = ( ) 𝑝(𝑗) (𝑎)𝑒 𝑎𝑥 = 𝑝(𝑗) (𝑎)𝑒 𝑎𝑥 .
0
Using these computations in (10), and noting (5), we see that 𝜓 satisfies (4) if and only if

S25013: Ordinary Differential Equations Page 74


𝑐0 (𝑗+𝑚
𝑚
)𝑝(𝑗) (𝑎) = 𝑏0 ,
𝑐0 𝑝(𝑗+1) (𝑎) + c1 (𝑗+𝑚−1
𝑚−1
)𝑝(𝑗) (𝑎) = 𝑏1 ,



𝑐0 𝑝(𝑗+𝑚) (𝑎) + 𝑐1 𝑝(𝑗+𝑚−1) (𝑎) + ⋯ + 𝑐𝑚 𝑝(𝑗) (𝑎) = 𝑏𝑚 .
This is a set of 𝑚 + 1 linear equations for the constants 𝑐0 , 𝑐1 , 𝑐2 , ⋯ , 𝑐𝑚 . They have a unique
solution, which can be obtained by solving the equations in succession since 𝑝(𝑗) (𝑎) ≠ 0.
Alternatively, we see that the determinant of the coefficients is just
𝑚+1
(𝑗+𝑚
𝑚
)(𝑗+𝑚−1
𝑚−1
) ⋯ 1[𝑝(𝑗) (𝑎)] ≠ 0.
This completes the proof of theorem.
Remark: The justification of the annihilator method when the right side of
𝐿(𝑦) = 𝑏(𝑥)
is a sum of terms of the form 𝑃(𝑥)𝑒 can be reduced to Theorem 3, by noting that if 𝜓1 , 𝜓2
𝑎𝑥

satisfy
𝐿(𝜓1 ) = 𝑏1 , 𝐿(𝜓2 ) = 𝑏2 ,
respectively, then 𝜓1 + 𝜓2 satisfies
𝐿(𝜓1 + 𝜓2 ) = 𝑏1 + 𝑏2 .
Example 08: Show that if 𝑓, 𝑔 are two functions with derivatives, then
𝐷𝑘 (𝑓𝑔) = ∑𝑘𝑙=0(𝑘𝑙) 𝐷𝑙 (𝑓)𝐷𝑘−𝑙 (𝑔),
𝑘!
where (𝑘𝑙) = (𝑘−𝑙)! 𝑙!. This is called Leibnitz theorem.
Solution: We prove this result by induction on 𝑘. Therefore for 𝑘 = 1, hence by directly
product formula, we have
𝐷1 (𝑓𝑔) = 𝐷(𝑓𝑔) = 𝑓𝐷(𝑔) + 𝐷(𝑓)𝑔
1! 1!
= (1−0)!0! 𝑓𝐷(𝑔) + (1−1)!1! 𝐷(𝑓)𝑔
1 1
= ( ) 𝑓𝐷(𝑔) + ( ) 𝐷(𝑓)𝑔
0 1
1 𝑘
= ∑𝑙=0( 𝑙 ) 𝐷 (𝑓)𝐷1−𝑙 (𝑔).
𝑙

This shows the result is true for 𝑘 = 1. Now, suppose the result is true for 𝑘 = 𝑚. Then
we have
𝐷𝑚 (𝑓𝑔) = ∑𝑚 𝑚 𝑙
𝑙=0( 𝑙 ) 𝐷 (𝑓)𝐷
m−𝑙
(𝑔). (11)
We shall show that the result is true for 𝑚 + 1. Consider
𝐷𝑚+1 (𝑓𝑔) = D(𝐷 𝑚 (𝑓𝑔))
𝑚
= 𝐷[∑𝑚 𝑙
𝑙=0( 𝑙 ) 𝐷 (𝑓)𝐷
m−𝑙 (𝑔)]

𝑚
= [∑𝑚 𝑙
𝑙=0( 𝑙 ) 𝐷(𝐷 (𝑓)𝐷
m−𝑙 (𝑔))
]
𝑚 𝑚
= [∑𝑚 𝑙
𝑙=0( 𝑙 ) 𝐷 (𝑓)𝐷(𝐷
m−𝑙 (𝑔))
+ ∑𝑚 𝑙
𝑙=0( 𝑙 ) 𝐷(𝐷 (𝑓))𝐷
m−𝑙 (𝑔)]

𝑚 𝑚
= [∑𝑚 𝑙
𝑙=0( 𝑙 ) 𝐷 (𝑓)𝐷
m+1−𝑙 (𝑔)
+ ∑𝑚
𝑙=0( 𝑙 ) 𝐷
𝑙+1 (𝑓)𝐷 m−𝑙 (𝑔)]

𝑚 𝑚 𝑚
= ( ) 𝑓𝐷𝑚+1 (𝑔) + ( ) 𝐷(𝑓)𝐷𝑚 (𝑔) + ( ) 𝐷2 (𝑓)𝐷𝑚−1 (𝑔)
0 1 2

S25013: Ordinary Differential Equations Page 75


+ ⋯ + (𝑚
𝑚
)𝐷𝑚 (𝑓)𝐷(𝑔)
+(𝑚
0
)𝐷(𝑓)𝐷𝑚 (𝑔) + (𝑚
1
)𝐷2 (𝑓)𝐷𝑚−1 (𝑔) + (𝑚
2
)𝐷 3 (𝑓)𝐷𝑚−2 (𝑔)
+ ⋯ + (𝑚
𝑚
)𝐷𝑚+1 (𝑓)𝑔
𝑚 𝑚 𝑚 𝑚 𝑚
= ( ) 𝑓𝐷𝑚+1 (𝑔) + [( ) + ( )] 𝐷(𝑓)𝐷𝑚 (𝑔) + [( ) + ( )] 𝐷2 (𝑓)𝐷𝑚−1 (𝑔)
0 0 1 1 2
𝑚
+ ⋯ + (𝑚)𝐷 𝑚+1 (𝑓)𝑔
(12)
But (𝑚
0
) = (𝑚+1
0
) = 1, [(𝑚
0
) + (𝑚
1
)] = (𝑚+1
1
), [(𝑚
1
) + (𝑚
2
)] = (𝑚+1
2
𝑚
), ⋯, [(𝑟−1) + (𝑚
𝑟
)] = (𝑚+1
𝑟
),
⋯, and (𝑚
𝑚
) = (𝑚+1
𝑚+1
) = 1. Therefore, using these corresponding values, we obtain
𝑚+1 𝑚+1 𝑚+1 2
𝐷𝑚+1 (𝑓𝑔) = ( ) 𝑓𝐷𝑚+1 (𝑔) + ( ) 𝐷(𝑓)𝐷𝑚 (𝑔) + ( ) 𝐷 (𝑓)𝐷𝑚−1 (𝑔)
0 1 2
+ ⋯ + (𝑚+1
𝑚+1
)𝐷𝑚+1 (𝑓)𝑔.
= ∑𝑚+1 𝑚+1 𝑙
𝑙=0 ( 𝑙 ) 𝐷 (𝑓)𝐷
m−𝑙
(𝑔). (13)
Hence the result is true for 𝑘 = 𝑚 + 1. Thereby principle of mathematical induction,
theorem is true for all positive integral value of 𝑘.

SOLVED PROBLEMS 01
Problem 02-01-01:
Compute the solution of 𝜓 𝑜𝑓 𝑦′′′ + 𝑦′′ + 𝑦′ + 𝑦 = 1 which satisfies 𝜓(0) = 0, 𝜓′(0) =
1, 𝜓′′(0) = 0.
Solution: Let 𝐿(𝑦) = 𝑦′′′ + 𝑦′′ + 𝑦′ + 𝑦 = 1. (14)
The characteristic polynomial is 𝑝(𝑟) = 𝑟 + 𝑟 + 𝑟 + 1, i.e. 𝑝(𝑟) = (𝑟 + 1)(𝑟 + 1). The
3 2 2

characteristic equation has roots 𝑟 = −1 and 𝑟 = ±𝑖 . Hence the three independent


solutions of 𝐿(𝑦) = 0 are
𝜙1 (𝑥) = 𝑒 −𝑥 , 𝜙2 (𝑥) = 𝑒 𝑖𝑥 , 𝜙3 (𝑥) = 𝑒 −𝑖𝑥 .
The general solution is given by 𝜙(𝑥) = 𝑐1 𝑒 −𝑥 + 𝑐2 𝑒 𝑖𝑥 + 𝑐3 𝑒 −𝑖𝑥 . We can also represent this
solution as
𝜙(𝑥) = 𝑑1 𝑒 −𝑥 + 𝑑2 cos 𝑥 + 𝑑3 sin 𝑥 (15)
with some new constants 𝑑1 , 𝑑2 , 𝑑3 .
Thus, we take 𝜙1 (𝑥)=𝑒 −𝑥 , 𝜙2 (𝑥) = cos 𝑥 , 𝜙3 (𝑥) = sin 𝑥 as three independent solutions of
𝐿(𝑦) = 0.
The particular solution of the non-homogeneous equation 𝐿(𝑦) = 𝑏(𝑥) is given by
𝜓𝑝 (𝑥) = 𝑢1 𝜙1 (𝑥) + 𝑢2 𝜙2 (𝑥) + 𝑢3 𝜙3 (𝑥), (16)
where the functions 𝑢𝑘 (𝑥) are given by (refer Theorem 1)
𝑥 𝑊𝑘 (𝑡)𝑏(𝑡)
𝑢𝑘 (𝑥) = ∫0 𝑑𝑡. (17)
𝑊(𝜙1 ,𝜙2 ,𝜙3 )(𝑡)
The Wronskian of the solutions of 𝐿(𝑦) = 0 can be determined by using the formula
𝑊(𝜙1 , 𝜙2 , 𝜙3 )(𝑥)=𝑒 −𝑎1 (𝑥−𝑥0) 𝑊(𝜙1 , 𝜙2 , 𝜙3 )(𝑥0 ). (18)
We take 𝑥0 = 0 ∈ 𝐼 and compute

S25013: Ordinary Differential Equations Page 76


𝜙1 (0) 𝜙2 (0) 𝜙3 (0)
𝑊(𝜙1 , 𝜙2 , 𝜙3 )(0) = | 𝜙1′ (0) 𝜙2′ (0) 𝜙3′ (0) |
𝜙1′′ (0) 𝜙2′′ (0) 𝜙3′′ (0)
1 1 0
𝑊(𝜙1 , 𝜙2 , 𝜙3 )(0) = |−1 0 1| = 2. (19)
1 −1 0
Hence, by using the result (18), we find
𝑊(𝜙1 , 𝜙2 , 𝜙3 )(𝑥)=2𝑒 −𝑥 . (20)
Since 𝑊𝑘 is the determinant obtained from the 𝑊(𝜙1 , 𝜙2 , 𝜙3 )(𝑥) by replacing the 𝑘 𝑡ℎ
column of 𝑊 by the column vector (0, 0, 1). Thus, we have
0 cos 𝑥 sin 𝑥
𝑊1 (𝑥)=|0 − sin 𝑥 cos 𝑥 | ⇒ 𝑊1 (𝑥) = 1,
1 − cos 𝑥 − sin 𝑥
𝑒 −𝑥 0 sin 𝑥
𝑊2 (𝑥) = |−𝑒 −𝑥 0 cos 𝑥 | ⇒ 𝑊2 (𝑥) = −𝑒 −𝑥 (sin 𝑥 + cos 𝑥)
𝑒 −𝑥 1 − sin 𝑥
and
𝑒 −𝑥 cos 𝑥 0
𝑊3 (𝑥) = |−𝑒 −𝑥 − sin 𝑥 0| ⇒ 𝑊3 (𝑥) = 𝑒 −𝑥 (cos 𝑥 − sin 𝑥).
𝑒 −𝑥 − cos 𝑥 1
Using these values, we compute by using (17) the function 𝑢𝑘 (𝑥) as
𝑥1 1
𝑢1 (𝑥) = ∫0 𝑒 𝑡 𝑑𝑡 ⇒ 𝑢1 (𝑥) = (𝑒 𝑥 − 1)
2 2
1 𝑥 1
𝑢2 (𝑥) = − 2 ∫0 (sin 𝑡 + cos 𝑡) 𝑑𝑡 ⇒ 𝜓2 (𝑥) = − 2 (1 + sin 𝑥 − cos 𝑥)
and
1 𝑥 1
𝑢3 (𝑥) = 2 ∫0 (cos 𝑡 − sin 𝑡) 𝑑𝑡 ⇒ 𝑢3 (𝑥) = 2 (sin 𝑥 + cos 𝑥 − 1).
Substituting these values together with the solutions in the equation (14) we get after
simplification the particular solution of 𝐿(𝑦) = 1 as
1
𝜓𝑝 (𝑥) = 2 (2 − 𝑒 −𝑥 − sin 𝑥 − cos 𝑥).
Hence the most general solutions represented in (2) becomes
𝜓(𝑥) = 𝑎1 𝑒 −𝑥 + 𝑎2 cos 𝑥 + 𝑎3 sin 𝑥 + 1, (21)
where 𝑎1 , 𝑎2 , 𝑎3 are constants. Since the solution satisfies 𝜓(0) = 0, 𝜓′(0) = 1 and 𝜓′′(0) =
0 which gives respectively 𝑎1 + 𝑎2 + 1 =0, −𝑎1 + 𝑎3 =1 and 𝑎1 − 𝑎2 = 0 . Solving these
1 1 1
equations, we find 𝑎1 =− 2 𝑎2 =− 2 and 𝑎3 = 2. Hence the solution (21) becomes
1
𝜓(𝑥) = (−𝑒 −𝑥 − cos 𝑥 + sin 𝑥) + 1
2
1
⇒ 𝜓(𝑥) = 1 − (cos 𝑥 − sin 𝑥 + 𝑒 −𝑥 ).
2
Problem 02-01-02:
Find all solutions of 𝑦′′′ − 𝑦′ = 𝑥.
Solution: Let (𝑦)=𝑦′′′ − 𝑦′ = 𝑥. We see that 𝑝(𝑟)=𝑟 3 − 𝑟 = 𝑟(𝑟 2 − 1) is the characteristic
equation. The roots of the equation 𝑝(𝑟)=0 are 𝑟 = 0, 𝑟 = ±1. Hence the functions 𝜙1 (𝑥) =
1 , 𝜙2 (𝑥) = 𝑒 𝑥 and 𝜙3 (𝑥) = 𝑒 −𝑥 are the linearly independent solutions of 𝐿(𝑦) = 0 . The
Wronskian of the solutions is given by

S25013: Ordinary Differential Equations Page 77


𝜙1 (𝑥) 𝜙2 (𝑥) 𝜙3 (𝑥) 1 𝑒 𝑥 𝑒 −𝑥
′ (𝑥)
𝑊(𝜙1 , 𝜙2 , 𝜙3 )(𝑥)=| 𝜙1 𝜙2′ (𝑥)
𝜙3′ (𝑥)
| = |0 𝑒 𝑥 −𝑒 −𝑥 | = 2. (22)
𝜙1′′ (𝑥)
𝜙2′′ (𝑥)
𝜙3′′ (𝑥)
0 𝑒 𝑥 𝑒 −𝑥
The general solution of 𝐿(𝑦) = 𝑏(𝑥) is given by
𝜓(𝑥) = 𝜓𝑝 (𝑥) + 𝑐1 𝜙1 + 𝑐2 𝜙2 + 𝑐3 𝜙3 , (23)
where 𝜓𝑝 (𝑥) = 𝑢1 𝜙1 + 𝑢2 𝜙2 + 𝑢3 𝜙3 . (24)
𝑥 𝑊 (𝑡)𝑏(𝑡)
and 𝑢𝑘 (𝑥) = ∫0 𝑊(𝜙 𝑘,𝜙 ,𝜙 )(𝑡) 𝑑𝑡. (25)
1 2 3

With usual meaning to 𝑊𝑘 we find


0 𝑒 𝑥 𝑒 −𝑥
𝑊1 (𝑥)=|0 𝑒 𝑥 −𝑒 −𝑥 | ⇒ 𝑊1 (𝑥) = −2
1 𝑒 𝑥 𝑒 −𝑥
1 0 1
𝑊2 (𝑥)=𝑒 −𝑥 |0 0 −1| ⇒ 𝑊2 (𝑥) = 𝑒 −𝑥
0 1 1
and
1 1 0
𝑊3 (𝑥)=𝑒 𝑥 |0 1 0| ⇒ 𝑊3 (𝑥) = 𝑒 𝑥 .
0 1 1
Hence the function 𝑢𝑘 (𝑥) from equation (25) become
𝑥 1
𝑢1 (𝑥)=− ∫0 𝑡 𝑑𝑡 ⇒ 𝑢1 (𝑥) = − 2 𝑥 2 ,
1 𝑥
𝑢2 (𝑥)= 2 ∫0 𝑡 𝑒 −𝑡 𝑑𝑡.
1
Integrating by parts we get 𝑢2 (𝑥) = 2 (1 − 𝑒 −𝑥 − 𝑥𝑒 −𝑥 )
and
1 𝑥
𝑢3 (𝑥) = 2 ∫0 𝑡 𝑒 𝑡 𝑑𝑡.
This gives after integrating by parts
1
𝑢3 (𝑥)= 2 (1 − 𝑒 𝑥 + 𝑥𝑒 𝑥 ).
1
Hence the particular solution (24) becomes 𝜓𝑝 (𝑥) = − 2 (𝑥 2 + 𝑒 𝑥 + 𝑒 −𝑥 − 2) .
Consequently, the general solution (23) reduces to
1
𝜓(𝑥) = 𝑎1 + 𝑎2 𝑒 𝑥 + 𝑎3 𝑒 −𝑥 − 2 𝑥 2 .
Problem 02-01-03:
Find all solution of 𝑦′′ − 2𝑖𝑦 ′ − 𝑦 = 𝑒 𝑖𝑥 − 2𝑒 −𝑖𝑥 .
Solution: - Let L(y)=𝑦′′ − 2𝑖𝑦′ − 𝑦 = 𝑒 𝑖𝑥 − 2𝑒 −𝑖𝑥 . The characteristic polynomial of L(y)=0
is 𝑝(𝑟) = 𝑟 2 − 2𝑖𝑟 − 1 ⇒ 𝑝(𝑟) = (𝑟 − 𝑖)2 . The root of 𝑝(𝑟) = 0 is 𝑟 = 𝑖 of multiplicity 2 .
Then the two linearly independent solutions of L(y) = 0 are given by 𝜙1 (𝑥) =𝑒 𝑖𝑥 and
𝜙2 (𝑥) = 𝑥𝑒 𝑖𝑥 . The general solution of the non – homogenous equation 𝐿(𝑦) = 𝑏(𝑥) is given
by
𝜓(𝑥)=𝜓𝑝 (𝑥) + 𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥),
where 𝜓𝑝 (𝑥) = 𝑢1 𝜙1 (𝑥) + 𝑢2 (𝑥)𝜙2 (𝑥) and the functions 𝑢1 (𝑥) and 𝑢2 (𝑥) are determined
from the formula

S25013: Ordinary Differential Equations Page 78


𝑥 𝜙2 (𝑡)𝑏(𝑡) 𝑥 𝑏(𝑡)𝜙1 (𝑡)
𝑢1 (𝑥) = − ∫0 𝑑𝑡 and 𝑢2 (𝑥) = ∫0 𝑑𝑡.
𝑊(𝜙1 ,𝜙2 )(𝑡) 𝑊(𝜙1 ,𝜙2 )(𝑡)
The Wronskian of the solution at any point 𝑥 in 𝐼 is given by the formula
𝑊(𝜙1 , 𝜙2 )(𝑥)=𝑒 −𝑎1𝑥 𝑊(𝜙1 , 𝜙2 )(0),
where 𝑎1 = −2𝑖.
𝜙 (0) 𝜙2 (0)
⇒ 𝑊(𝜙1 , 𝜙2 )(𝑥) = 𝑒 2𝑖𝑥 | 1′ |
𝜙1 (0) 𝜙2′ (0)
1 0
= 𝑒 2𝑖𝑥 | |
𝑖 1
⇒ 𝑊(𝜙1 , 𝜙2 )(𝑥) = 𝑒 2𝑖𝑥 .
Hence,
𝑥 (𝑒 𝑖𝑡 −2𝑒 −𝑖𝑡 )𝑡𝑒 𝑖𝑡 𝑥
𝑢1 (𝑥) = − ∫0 𝑑𝑡 = − ∫0 (𝑡 − 2𝑡𝑒 −2𝑖𝑡 )𝑑𝑡.
𝑒 2𝑖𝑡
Integrating, we get
1
𝑢1 (𝑥)=− 2 (𝑥 2 − 𝑒 −2𝑖𝑥 − 2𝑖𝑥𝑒 −2𝑖𝑥 + 1).
Similarly, we find
𝑥 (𝑒 𝑖𝑡 −2𝑒 −𝑖𝑡 )𝑒 𝑖𝑡 𝑥
𝑢2 (𝑥)= ∫0 𝑑𝑡 = ∫0 (1 − 2𝑒 −2𝑖𝑡 ) 𝑑𝑡.
𝑒 2𝑖𝑡
Integrating, we get
𝑢2 (𝑥) = 𝑥 − 𝑖(𝑒 −2𝑖𝑥 − 1).
Substituting all these values in the expression for the general solution of 𝐿(𝑦) = 𝑏(𝑥), we
𝑥2
obtain after simplifying 𝜓(𝑥) = (𝑎1 + 𝑎2 𝑥)𝑒 𝑖𝑥 + 2 𝑒 𝑖𝑥 , where 𝑎1 and 𝑎2 are some new
constants.
Problem 02-01-04:
Use annihilator method to find the solution of the equation 𝐿(𝑦) = 𝑦′′ + 9𝑦 = 𝑥 2 𝑒 3𝑥 .
Solution: Let 𝐿(𝑦) = 𝑦′′ + 9𝑦 = 𝑥 2 𝑒 3𝑥 , (26)
where 𝑏(𝑥) = 𝑥 𝑒 . We see that 𝑏(𝑥) is a solution of the equation with constant
2 3𝑥

coefficients given by
𝑀(𝑦) = 𝑦′′′ + 9𝑦′′ + 27𝑦′ + 27𝑦 = 0. (27 )
The characteristic polynomial of 𝑀(𝑦) = 0 is 𝑞(𝑟)=[𝑟 − 3] . Hence every solution 𝜓 of the
3

equation is the solution of 𝐿(𝑦) = 𝑏(𝑥)


⇒𝑀(𝐿(𝑦)) = 𝑀(𝑏)
⇒ 𝑀 (𝐿 (𝑦)) = 0. (28)
The characteristic polynomial of the equation (28) is the product of the characteristic
polynomials and 𝑀 and 𝐿 . ⇒ characteristic polynomial of (28) is (𝑟 − 3)3 (𝑟 2 + 9) . This
equation has roots 𝑟 = 3 three times, 𝑟 = ±3𝑖. Hence the solution of (28) is given by
𝜓(𝑥) = (𝑐0 + 𝑐1 𝑥 + 𝑐2 𝑥 2 )𝑒 3𝑥 + 𝑐3 𝑒 3𝑖𝑥 + 𝑐4 𝑒 −3𝑖𝑥 . (29)
We see that 𝑐3 𝑒 + 𝑐4 𝑒
3𝑖𝑥 −3𝑖𝑥
is a solution of 𝐿(𝑦) = 0. We write equation (29) as
𝜓(𝑥) = 𝜓𝑝 (𝑥) + 𝑐3 𝑒 3𝑖𝑥 + 𝑐4 𝑒 −3𝑖𝑥 . (30)
Now we find the constants 𝑐0 , 𝑐1 , 𝑐2 such that 𝜓𝑝 (𝑥) = (𝑐0 + 𝑐1 𝑥 + 𝑐2 𝑥 2 ) 𝑒 3𝑥 ) is a particular
solution of 𝐿(𝑦) = 𝑏(𝑥)
i.e. 𝐿 (𝜓𝑝 (𝑥)) = 𝑥 2 𝑒 3𝑥
⇒ 𝜓𝑝′′ + 9𝜓𝑝 = 𝑥 2 𝑒 3𝑥

S25013: Ordinary Differential Equations Page 79


⇒ [(𝑐0 + 𝑐1 𝑥 + 𝑐2 𝑥 2 )𝑒 3𝑥 ]′′ + 9(𝑐0 + 𝑐1 𝑥 + 𝑐2 𝑥 2 )𝑒 3𝑥 = 𝑥 2 𝑒 3𝑥
⇒ [(𝑐1 + 2𝑐2 𝑥) 𝑒 3𝑥 + 3𝑒 3𝑥 [𝑐0 + 𝑐1 𝑥 + 𝑐2 𝑥 2 ])] + 9(𝑐0 + 𝑐1 𝑥 + 𝑐2 𝑥 2 )𝑒 3𝑥 = 𝑥 2 𝑒 3𝑥
⇒ 2𝑐2 𝑒 3𝑥 + (𝑐1 + 2𝑐2 𝑥)3𝑒 3𝑥 + 9𝑒 3𝑥 (𝑐0 + 𝑐1 𝑥 + 𝑐2 𝑥 2 ) + 3𝑒 3𝑥 (𝑐1 + 2𝑐2 𝑥)
+9(𝑐0 + 𝑐1 𝑥 + 𝑐2 𝑥 2 )𝑒 3𝑥 = 𝑥 2 𝑒 3𝑥
⇒ 𝑒 3𝑥 [2𝑥2 + 3𝑐1 + 6𝑐2 𝑥 + 9𝑐0 + 9𝑐1 𝑥 + 9𝑐2 𝑥 2 + 3𝑐1 + 6𝑐2 𝑥 + 9(𝑐0 + 𝑐1 𝑥 + 𝑐2 𝑥 2 )]
= 𝑥 2 𝑒 3𝑥
⇒ (18𝑐0 + 𝑐1 + 2𝑐2 ) + (18𝑐1 + 12𝑐2 ) + 18𝑐2 𝑥 2 = 𝑥 2
⇒ 18𝑐2 = 1, 18𝑐1 + 12𝑐2 = 0, 18𝑐0 + 6𝑐1 + 2𝑐2 = 0.
Solving these equations, we obtain
1 1 1
𝑐0 = 162 , 𝑐1 = − 27, 𝑐2 = 18.
Substituting these values in the particular solution 𝜓𝑝 (𝑥), we get
1
𝜓𝑝 (𝑥) = 162 (1 − 6𝑥 + 9𝑥 2 )𝑒 3𝑥 .
Problem 02-01-05:
Solve by using annihilator method the equation 𝑦′′ − 4𝑦 = 3𝑒 2𝑥 + 4𝑒 −𝑥 .
Solution: Let 𝐿(𝑦) = 𝑦′′ − 4𝑦 = 3𝑒 2𝑥 + 4𝑒 −𝑥 , (31)
where 𝑏(𝑥) = 3𝑒 2𝑥 + 4𝑒 −𝑥 . Looking at the nature of 𝑏(𝑥) we can estimate that (𝑟 – 2) (𝑟 +
1) is a characteristic polynomial of 𝑀(𝑦) = 0 where obviously 𝑀(𝑦) = 0 is of the form
𝑦 ′′ − 𝑦 ′ − 2𝑦 =0. Hence every solution 𝜓 of the given non-homogeneous equation 𝐿(𝑦) =
𝑏(𝑥) is the solution of
𝑀 (𝐿(𝑦)) = 𝑀 (𝑏(𝑥)) = 0. (32)
The characteristic polynomial of the equation (32) is the product of the characteristic
polynomial of 𝐿 and 𝑀.
⇒ the characteristic equation of the equation (32) is (𝑟 − 2)(𝑟 + 1)(𝑟 2 − 4) = 0 . This
equation has roots 2, −1, 2, −2. Hence, the solution 𝜓 of the equation (32) is given by
𝜓(𝑥) = (𝑐1 + 𝑐2 𝑥)𝑒 2𝑥 + 𝑐3 𝑒 −𝑥 + 𝑐4 𝑒 −2𝑥 . (33)
We separate it and write as
𝜓(𝑥) = 𝑐2 𝑥𝑒 2𝑥 + 𝑐3 𝑒 −𝑥 + 𝑐1 𝑒 2𝑥 + 𝑐4 𝑒 −2𝑥 . (34)
⇒ 𝜓(𝑥) = 𝜓𝑝 (𝑥) + 𝑐1 𝑒 + 𝑐4 𝑒 ,
2𝑥 −2𝑥
(35)
where 𝜓𝑝 (𝑥)=𝑐2 𝑥𝑒 2𝑥 + 𝑐3 𝑒 −2𝑥 , (36)
is a particular solution of 𝐿(𝑦) = 𝑏(𝑥).
Hence, we find constants 𝑐2 and 𝑐3 such that 𝜓𝑝 (𝑥)=𝑐2 𝑥𝑒 2𝑥 + 𝑐3 𝑒 −2𝑥 , is a solution of the
equation 𝐿(𝑦) = 𝑏(𝑥).
⇒ 𝐿 (𝜓𝑝 (𝑥)) = 3𝑒 2𝑥 + 4𝑒 −𝑥
⇒ (𝑐2 𝑥𝑒 2𝑥 + 𝑐3 𝑒 −𝑥 )′′ − 4(𝑐2 𝑥𝑒 2𝑥 + 𝑐3 𝑒 −𝑥 ) = 3𝑒 2𝑥 + 4𝑒 −𝑥
⇒ (𝑐2 𝑒 2𝑥 + 2𝑐2 + 𝑒 2𝑥 − 𝑐3 𝑒 −𝑥 )′ − 4(𝑐2 𝑥𝑒 2𝑥 + 𝑐3 𝑒 −𝑥 ) = 3𝑒 2𝑥 + 4𝑒 −𝑥
⇒ 2𝑐2 𝑒 2𝑥 + 2𝑐2 𝑒 2𝑥 + 𝑐3 𝑒 −𝑥 − 4𝑐2 𝑥𝑒 2𝑥 − 4𝑐3 𝑒 −𝑥 = 3𝑒 2𝑥 + 4𝑒 −𝑥
⇒ 4𝑐2 𝑒 2𝑥 − 3𝑐3 𝑒 −𝑥 = 3𝑒 2𝑥 + 4𝑒 −𝑥
3 4
⇒ 4𝑐2=3 ⇒ 𝑐2 = 4 and −3𝑐3 = 4 ⇒ 𝑐3 = − 3.
Substituting these values in the equation (36), we get

S25013: Ordinary Differential Equations Page 80


3 4
𝜓𝑝 (𝑥) = 4 𝑥𝑒 2𝑥 − 3 𝑒 −𝑥 as the particular solution of 𝐿(𝑦) = 𝑏(𝑥).

SELF-TEST 01
MCQ 02-01-01
Consider the following statements:
𝑑𝑛
I. The function 𝑥 𝑛 is annihilated by the differential operator 𝑀 = 𝑑𝑥 𝑛.
II. The differential operator that annihilates a function is not unique.
Then
A. only (I) is true
B. only (II) is true
C. both (I) and (II) are true
D. both (I) and (II) are false.
MCQ 02-01-02
Suppose 𝑏 = 𝑏1 + 𝑏2 + ⋯ + 𝑏𝑛 , where 𝑏𝑘 is annihilated by the constant coefficient
differential operator 𝑀𝑘 . Then the function 𝑏 is annihilated by
A. 𝑀1 + 𝑀2 + ⋯ + 𝑀𝑛
B. 𝑀1 − 𝑀2 − ⋯ − 𝑀𝑛
C. 𝑀1 𝑀2 ⋯ 𝑀𝑛
D. 𝑀1 𝑀2 + 𝑀3 𝑀4 + ⋯ + 𝑀𝑛−1 𝑀𝑛 .
MCQ 02-01-03
If 𝑀 is constant coefficient operator that annihilates both 𝑥 𝑛 cosh 𝑎𝑥 and 𝑥 𝑛 sinh 𝑎𝑥, then
the operator 𝑀 is given by
A. 𝑀 = (𝐷 − 𝑎)𝑛+1
B. 𝑀 = (𝐷2 − 𝑎2 )𝑛+1
C. 𝑀 = (𝐷2 + 𝑎2 )𝑛+1
D. 𝑀 = (𝐷 + 𝑎)𝑛+1
MCQ 02-01-04
Which one of the followings is the particular solution of the equation 𝑦 ′′′ + 3𝑦 ′′ + 3𝑦 ′ +
𝑦 = 𝑥 2 𝑒 −𝑥 ?
𝑥 4 𝑒 −𝑥
A. 𝜓𝑝 (𝑥) = .
30
𝑥 3 𝑒 −𝑥
B. 𝜓𝑝 (𝑥) = .
18
𝑥 2 𝑒 −𝑥
C. 𝜓𝑝 (𝑥) = .
4
𝑥 5 𝑒 −𝑥
D. 𝜓𝑝 (𝑥) = .
60

SHORT ANSWER QUESTIONS 01


SAQ 02-01-01
Find the corresponding differential operator 𝑀 that annihilates each of the following
functions

S25013: Ordinary Differential Equations Page 81


i) 𝑏(𝑥) = 𝑐
ii) 𝑏(𝑥) = 𝑥
iii) 𝑏(𝑥) = 𝑥 4 .
Solution: i) From the above table for characteristic polynomial of an annihilator for 𝑘 =
𝑑 𝑑
0 , we have [𝑏(𝑥)] = [𝑘] = 0. Consequently, the differential operator 𝑀 is 𝑀(𝑦) =
𝑑𝑥 𝑑𝑥
𝑑𝑦
= 𝑦 ′.
𝑑𝑥
ii) Similarly, from the above table for characteristic polynomial of an ann ihilator for 𝑘 =
𝑑2 𝑑2 𝑑2 𝑦
1, we get 2
[𝑏(𝑥)] = [𝑥] = 0. Hence the annihilator 𝑀 is 𝑀(𝑦) = 2 = 𝑦 ′′ .
𝑑𝑥 𝑑𝑥 2 𝑑𝑥
iii) For 𝑘 = 4, the same table implies that
𝑑5 𝑑5
[𝑏(𝑥)] = [𝑥 4 ] = 0.
𝑑𝑥 5 𝑑𝑥 5
𝑑5 𝑦
Therefore, we have 𝑀(𝑦) = 𝑑𝑥 5 = 𝑦 (5) .
SAQ 02-01-02
If 𝑀1 annihilates 𝑏1 (𝑥) and 𝑀2 annihilates 𝑏2 (𝑥) , then prove that the product of the
differential operators 𝑀1 . 𝑀2 will annihilate the linear combination 𝑐1 𝑏1 (𝑥) + 𝑐2 𝑏2 (𝑥).
Solution: Let 𝑀1 and 𝑀2 be two differential operators that annihilate 𝑏1 (𝑥) and 𝑏2 (𝑥)
respectively. Then we can have
𝑀1 (𝑏1 (𝑥)) = 0
(37)
𝑀2 (𝑏2 (𝑥)) = 0.
Now, by linearity and commutative property of the differential operators, we have
𝑀1 . 𝑀2 (𝑐1 𝑏1 (𝑥) + 𝑐2 𝑏2 (𝑥)) = 𝑀2 [𝑀1 (𝑐1 𝑏1 (𝑥))] + 𝑀1 [𝑀2 (𝑐2 𝑏2 (𝑥))]
= 𝑀2 [𝑐1 𝑀1 (𝑏1 (𝑥))] + 𝑀1 [𝑐2 𝑀2 (𝑏2 (𝑥))]
= 𝑀2 [𝑐1 . 0] + 𝑀1 [𝑐2 . 0] (by using (37))
= 0.
This proves the required.
SAQ 02-01-03
Let 𝐿 be a constant coefficient operator and suppose 𝜓𝑘 is a solution of
𝐿(𝑦) = 𝑏𝑘 (𝑥), 𝑘 = 1,2, ⋯ , 𝑛,
where the 𝑏𝑘 are continuous functions on some interval 𝐼. Show that 𝜓 = 𝜓1 + 𝜓2 + ⋯ +
𝜓𝑛 is a solution of
𝐿(𝑦) = 𝑏(𝑥), 𝑏 = 𝑏1 + 𝑏2 + ⋯ + 𝑏𝑛 .
Solution: Let 𝐿 be a constant coefficient operator and suppose 𝜓𝑘 is a solution of
𝐿(𝑦) = 𝑏𝑘 (𝑥), 𝑘 = 1,2, ⋯ , 𝑛, (38)
where the 𝑏𝑘 are continuous functions on some interval 𝐼 and 𝑏 = 𝑏1 + 𝑏2 + ⋯ + 𝑏𝑛 .
Therefore, we have
𝐿(𝜓𝑘 ) = 𝑏𝑘 (𝑥), 𝑘 = 1,2, ⋯ , 𝑛. (39)
Now, we have using (39) and linear operator 𝐿 that
𝐿(𝜓(𝑥)) = 𝐿(𝜓1 (𝑥) + 𝜓2 (𝑥) + ⋯ + 𝜓𝑛 (𝑥))
= 𝐿(𝜓1 (𝑥)) + 𝐿(𝜓2 (𝑥)) + ⋯ + 𝐿(𝜓𝑛 (𝑥))

S25013: Ordinary Differential Equations Page 82


= 𝑏1 (𝑥) + 𝑏2 (𝑥) + ⋯ + 𝑏𝑛 (𝑥)
= 𝑏(𝑥).
Hence, we have proved as the required.

SUMMARY
In this Unit, we consider the nonhomogeneous equation with constant coefficients of order
𝑛 : 𝐿(𝑦) = 𝑦 (𝑛) + 𝑎1 𝑦 (𝑛−1) + 𝑎2 𝑦 (𝑛−2) + ⋯ + 𝑎𝑛 𝑦 = 𝑏(𝑥) , where 𝑎1 , 𝑎2 , ⋯ ,𝑎𝑛 are constants
and 𝑏(𝑥) is continuous function, defined on an interval 𝐼 . For 𝑛 linearly independent
solutions 𝜙1 , 𝜙2 , ⋯ , 𝜙𝑛 of 𝐿(𝑦) = 0 on 𝐼, with usual notations, the particular solution 𝜓𝑝 of
𝐿(𝑦) = 𝑏(𝑥) is given by
𝑥 𝑊𝑘 (𝑡)𝑏(𝑡)
𝜓𝑝 (𝑥) = ∑𝑛𝑘=1 𝜙𝑘 (𝑥) ∫𝑥 𝑑𝑡.
0 𝑊(𝜙1 ,𝜙2 ,⋯,𝜙𝑛 )(𝑡)

If 𝑀 is a differential operator with constant coefficients such that 𝑀(𝑏(𝑥)) = 0 , then


solving 𝑀(𝐿(𝑦)) = 0, we can obtain the solution of 𝐿(𝑦) = 𝑏(𝑥).

END OF UNIT EXERCISES


1) For real 𝑎, show that cosh 𝑎𝑥 and sinh 𝑎𝑥 satisfy 𝑦 ′′ − 𝑎2 𝑦 = 0.
2) Show that if 𝑔 has 𝑘 derivatives, and 𝑟 is a constant, then 𝐷𝑘 (𝑒 𝑟𝑥 𝑔) = 𝑒 𝑟𝑥 (𝐷 + 𝑟)𝑘 (𝑔).

KEY WORDS
Nonhomogeneous differential equation, Wronskian of solutions, particular solution,
annihilator, Leibnitz theorem.

S25013: Ordinary Differential Equations Page 83


UNIT 02-02: LINEAR EQUATIONS WITH VARIABLE COEFFICIENTS
LEARNING OBJECTIVES
After successful completion of this unit, you will be able to
❖ explain linear equations with variable constants
❖ apply the method for determination of Wronskian of solutions of a linear
homogeneous differential equation at any point 𝑥 ∈ 𝐼, if it is known at some 𝑥0 ∈ 𝐼

02-01: INTRODUCTION
A linear differential equation of order n with variable coefficients is given by
𝑎0 (𝑥)𝑦 (𝑛) + 𝑎1 (𝑥)𝑦 (𝑛−1) + 𝑎2 (𝑥)𝑦 (𝑛−2) + ⋯ + 𝑎𝑛 (𝑥)𝑦 = 𝑏(𝑥) (1)
where 𝑎0 , 𝑎1 , ⋯ , 𝑎𝑛 and 𝑏 are complex valued functions on some real interval 𝐼. Singular
points of the differential equation (1) are given by 𝑎0 (𝑥) = 0 and requires special
consideration of such points.
We will study such equation in the next units dividing the equation ( 1) by 𝑎0 (𝑥) we obtain
the equation of the form
𝑦 (𝑛) + 𝑎1 (𝑥)𝑦 (𝑛−1) + ⋯ + 𝑎𝑛 (𝑥)𝑦 = 𝑏(𝑥) (2)
Let
𝐿(𝑦) = 𝑦 (𝑛) + 𝑎1 (𝑥)𝑦 (𝑛−1) + ⋯ + 𝑎𝑛 (𝑥)𝑦. (3)
Then we write equation (1) as
𝐿(𝑦) = 𝑏(𝑥). (4)
If 𝑏(𝑥) = 0 for all 𝑥 on 𝐼, then we have 𝐿(𝑦) = 0 which is a homogeneous equation of
order 𝑛 with variable coefficients. If 𝑏(𝑥) ≠ 0 for some 𝑥 in 𝐼, then the equation (4) is
called a non-homogeneous equation.
A solution of (4) on 𝐼 is a function 𝜙(𝑥) on 𝐼 which has n derivatives in 𝐼 and which
satisfies 𝐿(𝑦) = 𝑏(𝑥).

Remark: Most of the results developed in the Unit 1 for the case 𝑎1 , 𝑎2 , ⋯ , 𝑎𝑛 are constants
continue to valid in more general case, when 𝑎1 (𝑥), 𝑎2 (𝑥), ⋯ , 𝑎𝑛 (𝑥) are continuous
function on 𝐼. The readers can refer the proofs of the theorems in this Unit , for the proofs
of the theorems omitted in the Unit 01-04.
The major difficulty with linear equations with variable coefficients is th at it is rare that
we can solve the equations in terms of elementary functions such as exponential and
trigonometric functions. There is no analogue of some of the theorems proved in the earlier
units.
However, in case 𝑎1 , 𝑎2 , ⋯ , 𝑎𝑛 and 𝑏 have convergent power series expansions the series
solution can be obtained. Although in many case, it is not possible to express a solution of
𝐿(𝑦) = 𝑏(𝑥) in terms of elementary functions, it can be proved that solutions always exist.

S25013: Ordinary Differential Equations Page 84


02-02: INITIAL VALUE PROBLEMS FOR THE HOMOGENEOUS EQUATION :
We assume the following existence theorem without proof.
Theorem 1: (Existence theorem) Let 𝑎1 (𝑥), 𝑎2 (𝑥), ⋯ , 𝑎𝑛 (𝑥) be continuous functions on an
interval 𝐼 containing a point 𝑥0 . If ∝1 , ∝2 , ⋯ , ∝𝑛 are any 𝑛 constants, there exists a solution
𝜙 of 𝐿(𝑦) = 𝑎0 (𝑥)𝑦 (𝑛) + 𝑎1 (𝑥)𝑦 (𝑛−1) + 𝑎2 (𝑥)𝑦 (𝑛−2) + ⋯ + 𝑎𝑛 (𝑥)𝑦 = 0 on 𝐼 satisfying
𝜙(𝑥) =∝1 , 𝜙(𝑥0 ) =∝1 , 𝜙′(𝑥0 ) =∝2 , ⋯ , 𝜙 (𝑛−1) (𝑥0 ) =∝𝑛 .
Remarks: The theorem assets that
i) the solution exists on the entire interval 𝐼 and
ii) every initial value problem has a solution.
iii) Proof of Theorem 1 is given in Unit 04-04 for more general case.
Note: If the coefficients of 𝑦 (𝑛) vanishes somewhere in 𝐼, then neither of these results may
be true.
The uniqueness of the solution of the IVP is demonstrated with the aid of the following
theorem. However, to prove the following theorem we recall here some results from
analysis.
Recall:
i) Continuous functions defined on an interval 𝐼 which is not closed and bounded, need
not be bounded there.
ii) Continuous functions defined on a closed and bounded interval attains its bounds.
iii) Closed bounded interval is of the form 𝑎 ≤ 𝑥 ≤ 𝑏 with 𝑎 and 𝑏 are reals.
iv) If 𝐼 is a closed bounded interval and if the functions 𝑎𝑗 (𝑥) are continuous on 𝐼 then
there always exists finite constants 𝑏𝑗 such that |𝑎𝑗 (𝑥)| ≤ 𝑏𝑗 , on 𝐼.
1
e.g. i) Consider the function 𝑓(𝑥) = 𝑥 defined on 𝐼: 0 < 𝑥 ≤ 1
1
We see that 𝑓(𝑥) is continuous on 𝐼 but not bounded on 𝐼 as lim 𝑓(𝑥) = lim = ∞.
𝑥→0 𝑥→0+ 𝑥
However, in the closed bounded interval 𝐽 ⊆ 𝐼 . Defined by 𝐽: [0+∈ ,1], ∈> 0 and small
1 1
then we see that 𝑓(𝑥) = 𝑥 is bounded on 𝐽 ⊆ 𝐼, since |𝑓(𝑥)| ≤ 𝜖 , ∀ 𝑥 ∈ 𝐽.
ii) Similarly, let 𝑓(𝑥) = 𝑥, defined on 0 ≤ 𝑥 < ∞, is continuous but not bounded because
lim 𝑥 = ∞.
𝑥→∞
However, if we choose a closed bounded interval 𝐽: [0, 𝑎], 0 < 𝑎 < ∞ in which 𝑓(𝑥) = 𝑥
is bounded since |𝑓(𝑥)| ≤ 𝑎, ∀ 𝑥 ∈ 𝐽.
Theorem 2: Let 𝑏1 , 𝑏2 , ⋯ , 𝑏𝑛 be non negative constants such that for all 𝑥 in 𝐼 |𝑎𝑗 (𝑥)| ≤
𝑏𝑗 , 𝑗 = 1, 2, ⋯ , 𝑛 and define 𝑘 by 𝑘 = 1 + 𝑏1 + 𝑏2 + ⋯ + 𝑏𝑛 . If 𝑥0 is point in 𝐼 and 𝜙(𝑥) is a
solution of 𝐿(𝑦) = 0 on 𝐼, then
‖𝜙(𝑥0 )‖𝑒 −𝑘|𝑥−𝑥0| ≤ ‖𝜙(𝑥)‖ ≤ ‖𝜙(𝑥0 )‖𝑒 𝑘|𝑥−𝑥0| , ∀ 𝑥 ∈ 𝐼
where,
1
2 2 2
‖𝜙(𝑥)‖=[|𝜙(𝑥)|2 + |𝜙′(𝑥)|2 + |𝜙 ′′ (𝑥)| + ⋯ + |𝜙 (𝑛−1) (𝑥)| ] .
Proof: Let 𝐿(𝑦) = 𝑦 (𝑛) + 𝑎1 (𝑥)𝑦 (𝑛−1) + ⋯ + 𝑎𝑛 (𝑥)𝑦 = 0. (5)
Let 𝜙(𝑥) be a solution of 𝐿(𝑦) = 0 on an interval 𝐼 containing a point 𝑥0 .

S25013: Ordinary Differential Equations Page 85


⇒ 𝐿(𝜙) = 0
⇒ 𝜙 (𝑛) (𝑥) + 𝑎1 (𝑥)𝜙 (𝑛−1) (𝑥) + 𝑎2 (𝑥)𝜙 (𝑛−1) (𝑥) + ⋯ + 𝑎𝑛 (𝑥)𝜙(𝑥)= 0
⇒ 𝜙 (𝑛) (𝑥)=−[𝑎1 (𝑥)𝜙 (𝑛−1) (𝑥) + 𝑎2 (𝑥)𝜙 (𝑛−2) (𝑥) + ⋯ + 𝑎𝑛 (𝑥)𝜙(𝑥)]
⇒ |𝜙 (𝑛) (𝑥)| ≤ |𝑎1 (𝑥)||𝜙 (𝑛−1) (𝑥)| + |𝑎2 (𝑥)||𝜙 (𝑛−2) (𝑥)| + ⋯ + |𝑎𝑛 (𝑥)||𝜙(𝑥)|. (6)
Since |𝑎𝑗 (𝑥)| ≤ 𝑏𝑗 , ∀ 𝑗 and 𝑏𝑗 are non-negative constants
⇒ |𝜙 (𝑛) (𝑥)| ≤ 𝑏1 |𝜙 (𝑛−1) (𝑥)| + 𝑏2 |𝜙 (𝑛−2) (𝑥)| + ⋯ + 𝑏𝑛 |𝜙(𝑥)| (7)
2
Let 𝑢(𝑥) = ‖𝜙(𝑥)‖2 = [|𝜙(𝑥)|2 + |𝜙′(𝑥)|2 + ⋯ + |𝜙 (𝑛−1) (𝑥)| ]

⇒ 𝑢(𝑥) = 𝜙𝜙 + 𝜙 ′ 𝜙′ + ⋯ + 𝜙 (𝑛−1) 𝜙 (𝑛−1) .


Differentiating we get
𝑢′(𝑥)=𝜙′𝜙 + 𝜙𝜙′ + 𝜙 ′′ 𝜙′ + 𝜙′𝜙′′ + ⋯ + 𝜙 (𝑛) 𝜙 (𝑛−1) + 𝜙 (𝑛−1) 𝜙 (𝑛)
We rearrange the terms and write this as
𝑢′(𝑥)=𝜙′𝜙 + 𝜙′′𝜙′ + ⋯ + 𝜙 (𝑛) 𝜙 (𝑛−1) + 𝜙𝜙′ + 𝜙′𝜙′′ + ⋯ + 𝜙 (𝑛−1) 𝜙 (𝑛)
𝑢′ (𝑥) ≤ 2|𝜙(𝑥)||𝜙′(𝑥)| + 2|𝜙 ′ (𝑥)||𝜙 ′′ (𝑥)| + ⋯ + 2|𝜙 (𝑛−1) ||𝜙 (𝑛) | (∵ |𝜙| = |𝜙|) (8)
Now using equation (7) in (8) we get
|𝑢′ (𝑥)| ≤ 2|𝜙(𝑥)||𝜙 ′ (𝑥)| + 2|𝜙 ′ (𝑥)||𝜙 ′′ (𝑥)| + ⋯
+2|𝜙 (𝑛−1) (𝑥)|[𝑏1 |𝜙 (𝑛−1) (𝑥)| + 𝑏2 |𝜙 (𝑛−2) (𝑥)| + ⋯ + 𝑏𝑛 |𝜙(𝑥)|]
≤ 2|𝜙(𝑥)||𝜙 ′ (𝑥)| + 2|𝜙 ′ (𝑥)||𝜙 ′′(𝑥) | + ⋯ + 2|𝜙 (𝑛−2) (𝑥)||𝜙 (𝑛−1) (𝑥)|
2
+ 2𝑏1 |𝜙 (𝑛−1) (𝑥)| + 2𝑏2 |𝜙 (𝑛−1) (𝑥)||𝜙 (𝑛−2) (𝑥)| + ⋯ + 2𝑏𝑛 |𝜙(𝑥)||𝜙 (𝑛−1) (𝑥)|.
We know
2|𝑏||𝑐| ≤ |𝑏 2 | + |𝑐 2 |.
Using this we obtain
2 2
𝑢′ (𝑥) ≤ |𝜙(𝑥)|2 + |𝜙 ′ (𝑥)|2 + |𝜙 ′ (𝑥)|2 + |𝜙 ′′ (𝑥)|2 + ⋯ + |𝜙 (𝑛−2) (𝑥)| + |𝜙 (𝑛−1) (𝑥)|
2 2 2 2
+2𝑏1 |𝜙 (𝑛−1) (𝑥)| + 𝑏2 (|𝜙 (𝑛−2) (𝑥)| + |𝜙 (𝑛−1) (𝑥)| ) + ⋯ + 𝑏𝑛 (|𝜙(𝑥)|2 + |𝜙 (𝑛−1) (𝑥)| )
2
≤ (1 + 𝑏𝑛 )|𝜙(𝑥)|2 + (2 + 𝑏𝑛−1 )|𝜙 ′ (𝑥)|2 + ⋯ + (2 + 𝑏2 )|𝜙 (𝑛−2) (𝑥)|
2
+(1 + 2𝑏1 + 𝑏2 + ⋯ + 𝑏𝑛 )|𝜙 (𝑛−1) (𝑥)| .
Since 𝑏𝑗 ∀ 𝑗 is nonnegative and hence replacing each bracket by 2(1 + 𝑏1 + 𝑏2 + ⋯ + 𝑏𝑛 ),
we get
2
|𝑢′ (𝑥)| ≤ 2(1 + 𝑏1 + 𝑏2 + ⋯ + 𝑏𝑛 ) (|𝜙(𝑥)|2 + |𝜙′(𝑥)|2 + ⋯ + |𝜙 (𝑛−1) (𝑥)| ). (9)
or |𝑢′ (𝑥)| ≤ 2𝑘𝑢(𝑥), for 𝑘 = 1 + 𝑏1 + 𝑏2 + ⋯ + 𝑏𝑛
⇒ −2𝑘𝑢(𝑥) ≤ 𝑢′ (𝑥) ≤ 2𝑘𝑢(𝑥). (10)
This is exactly same as the equation (28) of the Theorem 4 of Unit 01 -01. Proceeding
exactly in the same way (Readers are asked to refer the proof), we arrive at an inequality
‖(𝑥0 )‖𝑒 −𝑘|𝑥−𝑥0| ≤ ‖𝜙(𝑥)‖ ≤ ‖𝜙(𝑥0 )‖𝑒 𝑘|𝑥−𝑥0| , ∀ 𝑥 ∈ 𝐼. (11)
This completes the proof.
Theorem 3: (Uniqueness theorem) Let 𝑎1 (𝑥), 𝑎2 (𝑥), ⋯ , 𝑎𝑛 (𝑥) be continuous functions on
an interval 𝐼 containing a point 𝑥0 . If ∝1 , ∝2 , ⋯ , ∝𝑛 are any 𝑛 constants, then there is at

S25013: Ordinary Differential Equations Page 86


most one solution 𝜙 of 𝐿(𝑦) = 𝑦 (𝑛) + 𝑎1 (𝑥)𝑦 (𝑛−1) + ⋯ + 𝑎𝑛 (𝑥)𝑦 = 0 on 𝐼 , satisfying
𝜙(𝑥0 ) =∝1 , 𝜙′(𝑥0 ) =∝2 , ⋯ , 𝜙 (𝑛−1) (𝑥0 ) =∝𝑛 .
Proof: By existence theorem, there exists a solution of the IVP
𝐿(𝑦) = 0, 𝑦(𝑥0 ) =∝1 , 𝑦 ′ (𝑥0 ) =∝2 , ⋯ , 𝑦 (𝑛−1) (𝑥0 ) =∝𝑛 . (12)
Let 𝜙 (𝑥) and 𝜓(𝑥) be two solutions of the IVP (12) on 𝐼.
⇒ 𝐿(𝜙(𝑥)) = 0 and 𝐿(𝜓(𝑥)) = 0, (13)
satisfying
𝜙(𝑥0 ) =∝1 , 𝜓(𝑥0 ) =∝2 ,

𝜙 (𝑥0 ) =∝2 , 𝜓′ (𝑥0 ) =∝2

𝜙 (𝑛−1)
(𝑥0 ) =∝𝑛 , 𝜓 (𝑛−1) (𝑥0 ) =∝𝑛 (14)
Define the function
𝜒(𝑥) = 𝜙(𝑥) − 𝜓(𝑥) (15)
⇒ 𝜒(𝑥0 ) = 𝜙(𝑥0 ) − 𝜓(𝑥0 ) = 0
𝜒 ′ (𝑥0 ) = 𝜙 ′ (𝑥0 ) − 𝜓′ (𝑥0 ) = 0,

𝜒 (𝑛−1) (𝑥0 ) = 𝜙 (𝑛−1) (𝑥0 ) − 𝜓 (𝑛−1) (𝑥0 ) = 0. (16)
Since the function 𝑎1 (𝑥), 𝑎2 (𝑥), ⋯ , 𝑎𝑛 (𝑥) are continuous on 𝐼 . We know continuous
function on an interval 𝐼 need not be bounded there. Therefore, let 𝑥 be any point on 𝐼
other than 𝑥0 and let 𝐽 be any closed bounded interval in 𝐼 which contains 𝑥0 and 𝑥. The
𝑎𝑗 (𝑥)′𝑠 being continuous and bounded on 𝐽  𝐼 hence we have
|𝑎𝑗 (𝑥)| ≤ 𝑏𝑗 , 𝑗 = 1,2, ⋯ , 𝑛 on 𝐽 ⊆ 𝐼, (17)
where 𝑏𝑗 are some constants which may depend on 𝐽. From equations (15) and (16), we
have 𝜒(𝑥) is a solution of the IVP: 𝐿(𝑦) = 0, 𝑦(𝑥0 ) = 𝑦 ′ (𝑥0 ) = ⋯ = 𝑦 (𝑛−1) (𝑥0 ) = 0. Further
the functions 𝑎𝑗 (𝑥), ∀ 𝑗 are bounded on 𝐽, hence by using the Theorem 2, 𝜒(𝑥) must satisfy
the inequality (11).
i.e. ‖𝜒(𝑥0 )‖𝑒 −𝑘|𝑥−𝑥0| ≤ ‖𝜒(𝑥)‖ ≤ ‖𝜒(𝑥0 )‖𝑒 𝑘|𝑥−𝑥0| , ∀ 𝑥 ∈ 𝐽 ⊆ 𝐼.
⇒ 0 ≤ |𝜒(𝑥)| ≤ ‖𝜒(𝑥)‖ ≤ 0, ∀ 𝑥 ∈ 𝐽
⇒ 𝜒(𝑥) = 0, ∀ 𝑥 ∈ 𝐽 ⊆ 𝐼
⇒ 𝜙(𝑥) − 𝜓(𝑥) = 0, ∀ 𝑥 ∈ 𝐽 ⊆ 𝐼
⇒ 𝜙(𝑥) = 𝜓(𝑥), ∀ 𝑥 ∈ 𝐽 ⊆ 𝐼.
Since 𝑥 was chosen to be any arbitrary on 𝐼 other than 𝑥0 . It follows that is is true for all
∀ 𝑥 in 𝐼
⇒ 𝜙(𝑥) = 𝜓(𝑥), ∀ 𝑥 ∈ 𝐼
⇒ The solution of the IVP (12) is unique.

02-03: SOLUTION OF THE HOMOGENOUS EQUATION :


Definition: The 𝑛 – functions 𝜙1 , 𝜙2 , ⋯ , 𝜙𝑛 defined on an interval 𝐼 are said to be linearly
independent if and only if for constants 𝑐1 , 𝑐2 , ⋯ , 𝑐𝑛 such that the sum
𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥) + ⋯ + 𝑐𝑛 𝜙𝑛 (𝑥) = 0, ∀ 𝑥 ∈ 𝐼

S25013: Ordinary Differential Equations Page 87


⇒ 𝑐1 = 𝑐2 = ⋯ = 𝑐𝑛 = 0.
Linearly dependent means not linearly independent.
In this section, using the existence theorem, we prove there exist 𝑛 linearly independent
solutions of 𝐿(𝑦) = 𝑦 (𝑛) + 𝑎1 (𝑥)𝑦 (𝑛−1) + ⋯ + 𝑎𝑛 (𝑥)𝑦 = 0 on 𝐼 and also show that every
solution is a linear combination of these 𝑛 linearly independent solutions.
Theorem 4: There exist 𝑛 linearly independent solutions of
𝐿(𝑦) = 𝑦 (𝑛) + 𝑎1 (𝑥)𝑦 (𝑛−1) + ⋯ + 𝑎𝑛 (𝑥)𝑦 = 0
on 𝐼.
Proof: Let 𝐿(𝑦) = 𝑦 𝑛 + 𝑎1 (𝑥)𝑦 (𝑛−1) + ⋯ + 𝑎𝑛 (𝑥)𝑦 = 0 (18)
be a differential equation with variable coefficients.
Let 𝑥0 be a point in 𝐼. According to the existence theorem, there is a solution 𝜙1 of 𝐿(𝑦) =
(𝑛−1)
0 satisfying 𝜙1 (𝑥0 ) = 1, 𝜙1 ′(𝑥0 ) = 0, ⋯ , 𝜙1 (𝑥0 ) = 0.
In general, for each 𝑖 = 1, 2, ⋯ , 𝑛 , there is a solution 𝜙𝑖 (𝑥) of 𝐿(𝑦) = 0 satisfying the
conditions
𝜙1 (𝑥0 ) = 1, 𝜙1′ (𝑥0 ) = 0, ⋯ , 𝜙1 (𝑛−1) (𝑥0 ) = 0
𝜙2 (𝑥0 ) = 0, 𝜙2′ (𝑥0 ) = 1, ⋯ , 𝜙2 (𝑛−1) (𝑥0 ) = 0,
… (19)
𝜙𝑛 (𝑥0 ) = 0, 𝜙𝑛′ (𝑥0 ) = 0, ⋯ , 𝜙𝑛 (𝑛−1) (𝑥0 ) = 1
That is for each 𝑖 = 1, 2, ⋯ , 𝑛 , there is a solution 𝜙𝑖 (𝑥) on 𝐼 satisfying the conditions
𝜙𝑖 (𝑖−1) (𝑥0 ) = 1, 𝜙𝑖 (𝑗−1) (𝑥0 ) = 0, ∀ 𝑗 ≠ 𝑖 (20)
We shall show that these solutions are linearly independent on 𝐼 . Let 𝑐1 , 𝑐2 , ⋯ , 𝑐𝑛 be 𝑛
constants such that the sum
𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥) + ⋯ + 𝑐𝑛 𝜙𝑛 (𝑥) = 0, ∀ 𝑥 ∈ 𝐼. (21)
Differentiating the equation (21) we obtain
𝑐1 𝜙′1 (𝑥) + 𝑐2 𝜙′2 (𝑥) + ⋯ + 𝑐𝑛 𝜙𝑛′ (𝑥)=0
𝑐1 𝜙1′′ (𝑥) + 𝑐2 𝜙2′′ (𝑥) + ⋯ + 𝑐𝑛 𝜙𝑛′′ (𝑥)=0
⋯ (22)
(𝑛−1)
𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑛−1) (𝑥) + ⋯ + 𝑐𝑛 𝜙𝑛 (𝑛−1) (𝑥)=0, ∀ 𝑥 ∈ 𝐼.
In particular, for 𝑥=𝑥0 ∈ 𝐼, we have
𝑐1 𝜙1 (𝑥0 ) + 𝑐2 𝜙2 (𝑥0 ) + ⋯ + 𝑐𝑛 𝜙𝑛 (𝑥0 ) = 0
𝑐1 𝜙′1 (𝑥0 ) + 𝑐2 𝜙′2 (𝑥0 ) + ⋯ + 𝑐𝑛 𝜙𝑛′ (𝑥0 )=0
𝑐1 𝜙1′′ (𝑥0 ) + 𝑐2 𝜙2′′ (𝑥0 ) + ⋯ + 𝑐𝑛 𝜙𝑛′′ (𝑥0 )=0
⋯ (23)
𝑐1 𝜙1 (𝑛−1) (𝑥0 ) + 𝑐2 𝜙2 (𝑛−1) (𝑥0 ) + ⋯ + 𝑐𝑛 𝜙𝑛 (𝑛−1) (𝑥0 )=0, ∀ 𝑥 ∈ 𝐼.
Using conditions (20) in (23), we see that 𝑐1=0, 𝑐2 = 0, ⋯ = 𝑐𝑛 = 0.
Hence, we conclude that the 𝑛 – solutions 𝜙1 , 𝜙2 , ⋯ , 𝜙𝑛 are linearly independent on 𝐼.
Theorem 5: Let 𝜙1 , 𝜙2 , ⋯ , 𝜙𝑛 be the solutions of 𝐿(𝑦) = 0 on 𝐼 satisfying
𝜙𝑖 (𝑖−1) (𝑥0 ) = 1, 𝜙𝑖 (𝑗−1) (𝑥0 ) = 0, 𝑗 ≠ 𝑖, 𝑖, 𝑗 = 1,2, ⋯ , 𝑛

S25013: Ordinary Differential Equations Page 88


If 𝜙 is any solution of 𝐿(𝑦) = 0 on 𝐼, then there exist 𝑛 constants 𝑐1 , 𝑐2 , ⋯ , 𝑐𝑛 such that
𝜙 = 𝑐1 𝜙2 + 𝑐2 𝜙2 + ⋯ + 𝑐𝑛 𝜙𝑛 . (i.e. there exists 𝜙 is expressed as a linear combination of
𝑛 linearly independent solutions of 𝐿(𝑦) = 0).
Proof: Let 𝐿(𝑦) = 𝑦 (𝑛) + 𝑎1 (𝑥)𝑦 (𝑛−1) + ⋯ + 𝑎𝑛 (𝑥)𝑦 = 0, (24)
where 𝑎1 (𝑥), 𝑎2 (𝑥), ⋯ , 𝑎𝑛 (𝑥) are continuous functions defined on 𝐼 . Let 𝑥0 ∈ 𝐼 and
𝜙1 , 𝜙2 , ⋯ , 𝜙𝑛 be 𝑛 solutions of 𝐿(𝑦) = 0 on 𝐼 such that
𝜙𝑖 (𝑖−1) = 1, 𝜙𝑖 (𝑗−1) (𝑥0 ) = 0, 𝑗 ≠ 𝑖, 𝑖, 𝑗 = 1,2, ⋯ , 𝑛. (25)
⇒ 𝜙1 , 𝜙2 , ⋯ , 𝜙𝑛 are linearly independent on 𝐼 by Theorem 4.
Let 𝜙 be any solutions of 𝐿(𝑦) = 0 in 𝐼 satisfying the conditions
𝜙(𝑥0 ) =∝1 , 𝜙 ′ (𝑥0 ) =∝2 , ⋯ , 𝜙 (𝑛−1) (𝑥0 ) =∝𝑛 . (26)
Now consider the function
𝜓(𝑥) =∝1 𝜙1 (𝑥) +∝2 𝜙2 (𝑥) + ⋯ + ∝𝑛 𝜙𝑛 (𝑥), (27)
⇒ 𝐿(𝜓(𝑥)) = 0.
⇒ 𝜓 is a solution of 𝐿(𝑦) = 0 and further on using (25), we have
𝜓(𝑥0 ) =∝1 𝜙1 (𝑥0 ) +∝2 𝜙2 (𝑥0 ) + ⋯ + ∝𝑛 𝜙𝑛 (𝑥0 ) =∝1
𝜓 ′ (𝑥0 ) =∝1 𝜙2′ (𝑥0 ) +∝2 𝜙2′ (𝑥0 ) + ⋯ +∝𝑛 𝜙𝑛′ (𝑥) =∝2

𝜓 (𝑛−1) (𝑥0 ) =∝1 𝜙1 (𝑛−1) (𝑥0 ) +∝2 𝜙2 (𝑛−1) (𝑥0 ) + ⋯ +∝𝑛 𝜙𝑛 (𝑛−1) (𝑥0 ) =∝𝑛 .
⇒ 𝜓 is a solution of 𝐿(𝑦) = 0 satisfying the same initial conditions at 𝑥0 as 𝜙. Therefore,
by uniqueness theorem, we have
𝜙(𝑥) = 𝜓(𝑥), ∀ 𝑥 ∈ 𝐼
⇒ 𝜙(𝑥) =∝1 𝜙1 (𝑥) +∝2 𝜙2 (𝑥) + ⋯ +∝𝑛 𝜙𝑛 (𝑥)
or for ∝1 = 𝑐1 , ∝2 = 𝑐2 , ⋯ , ∝𝑛 = 𝑐𝑛 .
⇒ 𝜙(𝑥) = 𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥) + ⋯ + 𝑐𝑛 𝜙𝑛 (𝑥).
This proves the theorem.

02-04: THE WRONSKIAN AND LINEAR INDEPENDENT


Definition: Let 𝜙1 , 𝜙2 , … , 𝜙𝑛 be 𝑛 functions of 𝑥 each posses at least 𝑛 − 1 derivatives
𝜙1 𝜙2 … 𝜙𝑛
𝜙′1 𝜙′2 … 𝜙′𝑛
with respect to 𝑥 on an interval 𝐼. The determinant |⋮ ⋮ ⋮ ⋮
| is called
(𝑛−1) (𝑛−1) (𝑛−1)
𝜙1 𝜙2 … 𝜙𝑛
the Wronskian of the functions and it is denoted by 𝑊(𝜙1 , 𝜙2 , … , 𝜙𝑛 )(𝑥).
Theorem 6: Let 𝜙1 , 𝜙2 , … , 𝜙𝑛 be 𝑛 solutions of 𝐿(𝑦) = 𝑦 (𝑛) + 𝑎1 (𝑥)𝑦 (𝑛−1) + 𝑎2 (𝑥)𝑦 (𝑛−2) +
⋯ + 𝑎𝑛 (𝑥)𝑦 = 0. Then they are linearly independent if and only if
𝑊(𝜙1 , 𝜙2 , … , 𝜙𝑛 )(𝑥) ≠ 0, ∀ 𝑥 ∈ 𝐼.
Proof: Let 𝐿(𝑦) = 𝑦 (𝑛) + 𝑎1 (𝑥)𝑦 (𝑛−1) + 𝑎2 (𝑥)𝑦 (𝑛−2) + ⋯ + 𝑎𝑛 (𝑥)𝑦 = 0. (28)
Let 𝜙1 , 𝜙2 , … , 𝜙𝑛 be 𝑛 solutions of 𝐿(𝑦) = 0. Let
𝑊(𝜙1 , 𝜙2 , … , 𝜙𝑛 )(𝑥) ≠ 0, ∀ 𝑥 ∈ 𝐼. (29)
Claim: We prove that 𝜙1 , 𝜙2 , … , 𝜙𝑛 are linearly independent. Consider 𝑐1 , 𝑐2 , … , 𝑐𝑛 be 𝑛
constants such that

S25013: Ordinary Differential Equations Page 89


𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥) + ⋯ + 𝑐𝑛 𝜙𝑛 (𝑥) = 0, ∀ 𝑥 ∈ 𝐼. (30)
Differentiating this with respect to 𝑥, (𝑛 − 1) times we get respectively
𝑐1 𝜙1′ (𝑥) + 𝑐2 𝜙2′ (𝑥) + ⋯ + 𝑐𝑛 𝜙𝑛′ (𝑥) = 0,
𝑐1 𝜙1′′ (𝑥) + 𝑐2 𝜙2′′ (𝑥) + ⋯ + 𝑐𝑛 𝜙𝑛′′ (𝑥) = 0,

(𝑛−1) (𝑛−1) (𝑛−1)
𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥) + ⋯ + 𝑐𝑛 𝜙𝑛 (𝑥) = 0. (31)
If 𝑥0 ∈ 𝐼 is a fixed point, then from above equations we have
𝑐1 𝜙1 (𝑥0 ) + 𝑐2 𝜙2 (𝑥0 ) + ⋯ + 𝑐𝑛 𝜙𝑛 (𝑥0 ) = 0,
𝑐1 𝜙1′ (𝑥0 ) + 𝑐2 𝜙2′ (𝑥0 ) + ⋯ + 𝑐𝑛 𝜙𝑛′ (𝑥0 ) = 0,
𝑐1 𝜙1′′ (𝑥0 ) + 𝑐2 𝜙2′′ (𝑥0 ) + ⋯ + 𝑐𝑛 𝜙𝑛′′ (𝑥0 ) = 0,

(𝑛−1) (𝑛−1) (𝑛−1)
𝑐1 𝜙1 (𝑥0 ) + 𝑐2 𝜙2 (𝑥0 ) + ⋯ + 𝑐𝑛 𝜙𝑛 (𝑥0 ) = 0. (32)
This is a system of 𝑛 homogeneous equations in 𝑐1 , 𝑐2 , … , 𝑐𝑛 . The determinant of the
coefficients is just Wronskian given by
𝜙1 (𝑥0 ) 𝜙2 (𝑥0 ) … 𝜙𝑛 (𝑥0 )
𝜙′1 (𝑥0 ) 𝜙′2 (𝑥0 ) … 𝜙′𝑛 (𝑥0 )
𝑊(𝜙1 , 𝜙2 , … , 𝜙𝑛 )(𝑥0 ) = || || ≠ 0,

(𝑛−1) (𝑛−1) (𝑛−1)
𝜙1 (𝑥0 ) 𝜙2 (𝑥0 ) … 𝜙𝑛 (𝑥0 )
⟹ That the system of 𝑛 homogeneous linear equations has only the trivial solutions 𝑐1 =
0, 𝑐2 = 0, … , 𝑐𝑛 = 0. Then 𝜙1 , 𝜙2 , … , 𝜙𝑛 are linearly independent functions.
Conversely, suppose that 𝜙1 , 𝜙2 , … , 𝜙𝑛 are linearly independent solutions of 𝐿(𝑦) = 0 on 𝐼.
We prove that
𝑊(𝜙1 , 𝜙2 , … , 𝜙𝑛 )(𝑥) ≠ 0 ∀ 𝑥 ∈ 𝐼
Let us suppose that there exists a points 𝑥0 in 𝐼 such that
𝑊(𝜙1 , 𝜙2 , … , 𝜙𝑛 )(𝑥0 ) = 0
𝜙1 (𝑥0 ) 𝜙2 (𝑥0 ) … 𝜙𝑛 (𝑥0 )
′ ′
𝜙1 (𝑥0 ) 𝜙2 (𝑥0 ) … 𝜙𝑛′ (𝑥0 )
||⋮ ⋮ ⋮ || = 0
(𝑛−1) (𝑛−1) (𝑛−1)
𝜙1 (𝑥0 ) 𝜙2 (𝑥0 ) … 𝜙n (𝑥0 )
This implies that the system of 𝑛 homogeneous linear equations
𝑐1 𝜙1 (𝑥0 ) + 𝑐2 𝜙2 (𝑥0 ) + ⋯ + 𝑐𝑛 𝜙𝑛 (𝑥0 ) = 0,
𝑐1 𝜙1′ (𝑥0 ) + 𝑐2 𝜙2′ (𝑥0 ) + ⋯ + 𝑐𝑛 𝜙𝑛′ (𝑥0 ) = 0,

(𝑛−1) (𝑛−1)
𝑐1 𝜙1𝑛−1 (𝑥0 ) + 𝑐2 𝜙2 (𝑥0 ) + ⋯ + 𝑐𝑛 𝜙𝑛 (𝑥0 ) = 0, (33)
is consistent and has a solutions 𝑐1 , 𝑐2 , ⋯ , 𝑐𝑛 , where all 𝑐i ’s are not zero simultaneously.
We see that
𝐿(𝜓) = 𝐿(𝑐1 𝜙1 + 𝑐2 𝜙2 + ⋯ + 𝑐𝑛 𝜙𝑛 )
= 𝑐1 𝐿(𝜙1 ) + 𝑐2 𝐿(𝜙2 ) + ⋯ + 𝑐𝑛 𝐿(𝜙𝑛 )
𝐿(𝜓) = 0.
⟹ 𝜓 is a solution of 𝐿(𝑦) = 0. Also from equation (33), we see that
𝜓(𝑥0 ) = 𝑐1 𝜙1 (𝑥0 ) + 𝑐2 𝜙2 (𝑥0 ) + ⋯ + 𝑐𝑛 𝜙𝑛 (𝑥0 ) = 0,

S25013: Ordinary Differential Equations Page 90


𝜓′(𝑥0 ) = 𝑐1 𝜙′1 (𝑥0 ) + 𝑐2 𝜙′2 (𝑥0 ) + ⋯ + 𝑐𝑛 𝜙′𝑛 (𝑥0 ) = 0,

𝑛−1 𝑛−1
𝜓 (𝑥0 ) = 𝑐1 𝜙1 (𝑥0 ) + 𝑐2 𝜙2 (𝑥0 ) + ⋯ + 𝑐𝑛 𝜙𝑛𝑛−1 (𝑥0 ) = 0,
𝑛−1
(34)
⟹ 𝜓 is a solution of IVP: 𝐿(𝑦) = 0, 𝑦(𝑥0 ) = 𝑦 ′ (𝑥0 ) = ⋯ = 𝑦 (𝑛−1) (𝑥0 ) = 0.
Hence, we have
||𝜓(𝑥0 )||𝑒 −𝑘|𝑥−𝑥0| ≤ ||𝜓(𝑥0 )|| ≤ ||𝜓(𝑥0 )||𝑒 𝑘|𝑥−𝑥0| , ∀ 𝑥, (35)
where ||𝜓(𝑥0 )||2 = ||ψ(𝑥0 )||2 + ||𝜓′ (𝑥0 )||2 + ⋯ + ||𝜓𝑛−1 (𝑥0 )||2 = 0, i. e ||𝜓(𝑥0 )|| = 0. The
equation (35) ⟹ 0 ≤ |𝜓(𝑥)| ≤ ||𝜓(𝑥)|| ≤ 0, ∀ 𝑥 ∈ 𝐼
⟹ 𝜓(𝑥) = 0, ∀ 𝑥 ∈ 𝐼.
⟹ 𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥) + ⋯ 𝑐𝑛 𝜙𝑛 (𝑥) = 0 such that not all 𝑐𝑖 ’s are zero simultaneously.
⟹ 𝜙1 , 𝜙2 , … , 𝜙𝑛 are not linearly independent. This contradiction arises because of our
wrong assumptions 𝑊(𝜙1 , 𝜙2 , … , 𝜙𝑛 )(𝑥0 ) = 0, for some 𝑥0 ∈ 𝐼.
⟹ 𝑊(𝜙1 , 𝜙2 , … , 𝜙𝑛 )(𝑥) ≠ 0 ∀ 𝑥 ∈ 𝐼.
Theorem 7: Let 𝜙1 , 𝜙2 , … , 𝜙𝑛 be any 𝑛 linearly independent solutions of
𝐿(𝑦) = 𝑦 (𝑛) + 𝑎1 (𝑥)𝑦 (𝑛−1) + 𝑎2 (𝑥)𝑦 (𝑛−2) + ⋯ + 𝑎𝑛 (𝑥)𝑦 = 0
on an interval 𝐼. If 𝜙 is any solution of 𝐿(𝑦) = 0 on 𝐼 then it can be represented by
𝜙(𝑥) = 𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥) + ⋯ + 𝑐𝑛 𝜙𝑛 (𝑥)
where 𝑐1 , 𝑐2 , … , 𝑐𝑛 are constants.
i.e. any set of 𝑛 linearly independent solutions of 𝐿(𝑦) = 0 on 𝐼 is a basis for the solution
of 𝐿(𝑦) = 0 on 𝐼.
Proof: Let 𝜙1 , 𝜙2 , … , 𝜙𝑛 be 𝑛 linearly independent solutions of 𝐿(𝑦) = 0, on 𝐼, where
𝐿(𝑦) = 𝑦 (𝑛) + 𝑎1 (𝑥)𝑦 (𝑛−1) + ⋯ 𝑎𝑛 (𝑥)𝑦 = 0.
⟹ 𝑊(𝜙1 , 𝜙2 , … , 𝜙𝑛 )(𝑥) ≠ 0, ∀ 𝑥 ∈ 𝐼. (36)
Let 𝑥0 be a point in 𝐼 and 𝜙 be any solution of 𝐿(𝑦) = 0 on 𝐼 satisfying the conditions
𝜙(𝑥0 ) = 𝛼1 , 𝜙′(𝑥0 ) = 𝛼2 , … , 𝜙 (𝑛−1) (𝑥0 ) = 𝛼𝑛 . (37)
Since from equation (36), we have for 𝑥0 ∈ 𝐼,
𝑊(𝜙1 , 𝜙2 , … , 𝜙𝑛 )(𝑥0 ) ≠ 0.
⟹ the system of equations,
𝑐1 𝜙1 (𝑥0 ) + 𝑐2 𝜙2 (𝑥0 ) + ⋯ + 𝑐𝑛 𝜙𝑛 (𝑥0 ) = 𝛼1 ,
𝑐1 ϕ1′ (𝑥0 ) + 𝑐2 𝜙′2 (𝑥0 ) + ⋯ + 𝑐𝑛 𝜙′𝑛 (𝑥0 ) = 𝛼2 ,

(𝑛−1) (𝑛−1) (𝑛−1)
𝑐1 𝜙1 (𝑥0 ) + 𝑐2 𝜙2 (𝑥0 ) + ⋯ + 𝑐𝑛 𝜙𝑛 (𝑥0 ) = 𝛼𝑛
has unique solutions for 𝑐1 , 𝑐2 , … , 𝑐𝑛 . For this choice of 𝑐1 , 𝑐2 , … , 𝑐𝑛 , define a new function
𝜓(𝑥) = 𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥) + ⋯ + 𝑐𝑛 𝜙𝑛 (𝑥), ∀ 𝑥 ∈ 𝐼 (38)
⟹ 𝐿(𝜓(𝑥)) = 𝐿(𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥) + 𝑐𝑛 𝜙𝑛 (𝑥))
= 𝑐1 𝐿(𝜙1 ) + 𝑐2 𝐿(𝜙2 ) + ⋯ + 𝑐𝑛 𝐿(𝜙𝑛 )
= 0, on 𝐼.
Also from equation (37) and (38), we see that
𝜓(𝑥0 ) = 𝛼1 , 𝜓 ′ (𝑥0 ) = α2 , … , 𝜓𝑛−1 (𝑥0 ) = 𝛼𝑛 . (39)
⟹ 𝜓 is a solution of IVP 𝐿(𝑦) = 0, satisfying

S25013: Ordinary Differential Equations Page 91


𝑦(𝑥0 ) = 𝛼1 , 𝑦′(𝑥0 ) = 𝛼2 , … , 𝑦 𝑛−1 (𝑥0 ) = 𝛼𝑛 .
Therefore by uniqueness theorem, we have
𝜙(𝑥) = 𝜓(𝑥), ∀ 𝑥 ∈ 𝐼
⟹ 𝜙(𝑥) = 𝑐1 𝜙1 (𝑥) + 𝑐2 𝜙2 (𝑥) + ⋯ + 𝑐𝑛 𝜙𝑛 (𝑥), ∀ 𝑥 ∈ 𝐼.
⟹ Any solution of 𝐿(𝑦) = 0 is uniquely expressed as a linear combinations of linearly
independent solutions. This implies that {𝜙1 , 𝜙2 , … , 𝜙𝑛 } from a basis of 𝐿(𝑦) = 0 on 𝐼.
Theorem 8: Let 𝜙1 , 𝜙2 , … , 𝜙𝑛 be 𝑛 solutions of
𝐿(𝑦) = 𝑦 (𝑛) + 𝑎1 (𝑥)𝑦 (𝑛−1) + 𝑎2 (𝑥)𝑦 (𝑛−2) + ⋯ + 𝑎𝑛 (𝑥)𝑦 = 0
on an interval 𝐼 and let 𝑥0 be any point in 𝐼. Then
𝑥
𝑊(𝜙1 , 𝜙2 , … , 𝜙𝑛 )(𝑥) = exp [− ∫𝑥 𝑎1 (𝑡)𝑑𝑡] 𝑊(𝜙1 𝜙2 , … , 𝜙𝑛 )(𝑥0 ).
0

Proof: Let 𝑛
𝐿(𝑦) = 𝑦 + 𝑎1 (𝑥)𝑦 𝑛−1
+ ⋯ + 𝑎𝑛 (𝑥)𝑦 = 0, (40)
where 𝑎1 (𝑥), 𝑎2 (𝑥), … , 𝑎𝑛 (𝑥) are continuous functions in 𝐼 . Let also 𝜙1 , 𝜙2 , … , 𝜙𝑛 be 𝑛
linearly independent solutions of 𝐿(𝑦) = 0. Let 𝑊(𝜙1 , 𝜙2 , … , 𝜙𝑛 ) be the Wronskian of the
solutions 𝜙1 , 𝜙2 , … , 𝜙𝑛 where
𝜙1 (𝑥) 𝜙2 (𝑥) … 𝜙𝑛 (𝑥)
′ ′
𝜙1 (𝑥) 𝜙2 (𝑥) … 𝜙n′ (𝑥)
𝑊(𝜙1 , 𝜙2 , … , 𝜙𝑛 )(𝑥) = ⋮| |. (41)
⋮ ⋮ ⋮
(𝑛−1) (𝑛−1) (𝑛−1)
𝜙1 (𝑥) 𝜙2 (𝑥) … 𝜙𝑛 (𝑥)
By the well known property of the determinant we know 𝑊′ is the sum of 𝑛 determinants
of the form
𝑊′(𝑥) = 𝑉1 + 𝑉2 + ⋯ + 𝑉𝑘 + ⋯ + 𝑉𝑛 , (42)
where the determinant 𝑉𝑘 is of order 𝑛 × 𝑛 and it is differs from 𝑊 in the 𝑘 row. The 𝑘 th th

row of 𝑉𝑘 is obtained by differentiating the 𝑘 th row of 𝑊. Then we have


𝜙1 𝜙2 … 𝜙𝑛
𝜙1′ ϕ′2 … ϕ′n ′′ ′′
′ ′ ′ 𝜙 1 𝜙 2 … 𝜙𝑛′′
𝜙 𝜙 … 𝜙𝑛 | |
𝑊 ′ = |⋮ 1 2
| + 𝜙1′′ 𝜙2′′ … 𝜙𝑛′′
⋮ ⋮ ⋮ |⋮ |
(𝑛−1) (𝑛−1) (𝑛−1) ⋮ ⋮ ⋮
𝜙1 𝜙2 … 𝜙𝑛 (𝑛−1) (𝑛−1) (𝑛−1)
𝜙1 𝜙2 … 𝜙𝑛
𝜙1 𝜙2 … 𝜙𝑛
′ ′
𝜙 𝜙2 … 𝜙𝑛′
|⋮ 1 |
+ ⋯ + (𝑛−2) ⋮ (𝑛−2) ⋮ ⋮ (𝑛−2) . (43)
|𝜙1 𝜙2 … 𝜙𝑛 |
(𝑛) (𝑛) (𝑛)
𝜙1 𝜙2 … 𝜙𝑛
We see that the first (𝑛 − 1) determinant 𝑉1 , 𝑉2 , … , 𝑉𝑛−1 on the R.H.S. of (43) are zero as
each of them have two identical rows. Hence
𝜙1 𝜙2 … 𝜙𝑛
𝜙1′ 𝜙2′ … 𝜙𝑛′
| |
𝑊′ = ⋮ (𝑛−2) ⋮ (𝑛−2) ⋮ ⋮ (𝑛−2) . (44)
|𝜙1 𝜙2 … 𝜙𝑛 |
(𝑛) (𝑛) (𝑛)
𝜙1 𝜙2 … 𝜙𝑛
Since 𝜙𝑖 (𝑥) ∀ 𝑖 = 1,2, … , 𝑛 is a solution of 𝐿(𝑦) = 0

S25013: Ordinary Differential Equations Page 92


(𝑛) (𝑛−1) (𝑛−2)
⟹ 𝜙𝑖 + 𝑎1 (𝑥)𝜙𝑖 + 𝑎2 (𝑥)𝜙𝑖 + ⋯ + 𝑎𝑛 (𝑥)𝜙𝑖 (𝑥) = 0
(𝑛) (𝑛−1) (𝑛−2)
⟹ 𝜙𝑖 (𝑥) = −[𝑎1 (𝑥)𝜙𝑖 + 𝑎2 (𝑥)𝜙𝑖 + ⋯ + 𝑎𝑛 (𝑥)𝜙𝑖 (𝑥)], ∀ 𝑖 = 1,2, ⋯ , 𝑛.
We write this as
(𝑛) (𝑛−2) (𝑛−1)
𝜙𝑖 (𝑥) = −[𝑎𝑛 (𝑥)𝜙𝑖 (𝑥) + 𝑎𝑛−1 𝜙′𝑖 (𝑥) + ⋯ + 𝑎2 (𝑥)𝜙𝑖 (𝑥) + 𝑎1 (𝑥)𝜙𝑖 (𝑥)]
(𝑛) (𝑗)
𝜙𝑖 (𝑥) = − ∑𝑛−1
𝑗=0 𝑎𝑛−𝑗 𝜙𝑖 (𝑥), ∀ 𝑖 = 1,2, … , 𝑛 (45)
Hence equations (44) becomes
𝜙1 𝜙2 … 𝜙𝑛
𝜙′ 𝜙′2 … 𝜙′𝑛
| 1 |
𝑊′ = ⋮ (𝑛−2) ⋮
(𝑛−2)
⋮ ⋮
(𝑛−2)
. (46)
|𝜙1 𝜙2 … 𝜙n |
(𝑗) (𝑗) (𝑗)
− ∑𝑛−1
𝑗=0 𝑎𝑛−𝑗 𝜙1 − ∑𝑛−1
𝑗=0 𝑎𝑛−𝑗 𝜙2 … − ∑𝑛−1
𝑗=0 𝑎𝑛−𝑗 𝜙𝑛
We know the value of the determinant is unchanged if we multiply any row by a number
and add it to other row. Thus, we multiply the first row by 𝑎𝑛 , the second row by 𝑎𝑛−1 , ⋯,
(𝑛 − 1)𝑠𝑡 row by 𝑎2 and add these to the last row, we get
(i. e. perform in the operation 𝑅𝑛 + 𝑎𝑛 𝑅1 + 𝑎𝑛−1 + 𝑅2 + ⋯ + 𝑎2 𝑅𝑛−1 , we get)
𝜙1 𝜙2 … 𝜙𝑛
′ ′
𝜙 𝜙2 … 𝜙′𝑛
| 1 |
𝑊′ = ⋮ ⋮ ⋮ ⋮
|𝜙1(𝑛−2) 𝜙2
(𝑛−2)
… 𝜙n
(𝑛−2)
|
(n−1) (n−1) (n−1)
−𝑎1 𝜙1 −𝑎1 𝜙2 … −𝑎1 𝜙n
𝜙1 𝜙2 … 𝜙𝑛

𝜙 𝜙′2 … 𝜙′𝑛
| 1 |
= −a1 ⋮ ⋮ ⋮ ⋮
|𝜙1(𝑛−2) 𝜙2(𝑛−2) … 𝜙n
(𝑛−2)
|
(n−1) (n−1) (n−1)
𝜙1 −𝜙2 … 𝜙n
= −𝑎1 𝑊. (47)
Equation (47) shows that 𝑊 satisfies the linear first order equation
𝑦′ + 𝑎1 (𝑥)𝑦=0.
Thus, we write
𝑊′
= −𝑎1 (𝑥).
𝑊
𝑑𝑊
⟹ = −𝑎1 (𝑥)𝑑𝑥.
𝑊
On integrating between the limits 𝑥0 to 𝑥, we get
𝑥
(log𝑊)𝑥𝑥0 = − ∫𝑥 𝑎1 (𝑡)𝑑𝑡
0
𝑥
or log 𝑊(𝑥) − log 𝑊(𝑥0 ) = − ∫𝑥 𝑎1 (𝑡)𝑑𝑡
0
𝑊(𝑥) 𝑥
⟹ log (𝑊(𝑥 )) = − ∫𝑥 𝑎1 (𝑡)𝑑𝑡
0 0
𝑥
⟹ 𝑊(𝑥) = exp [− ∫𝑥 𝑎1 (𝑡)𝑑𝑡] 𝑊(𝑥0 )
0
𝑥
i. 𝑒. 𝑊(𝜙1 , 𝜙2 , ⋯ , 𝜙𝑛 )(𝑥) = exp [− ∫𝑥 𝑎1 (𝑡)𝑑𝑡] 𝑊(𝜙1 , 𝜙2 , … , 𝜙𝑛 )(𝑥0 ). (48)
0

S25013: Ordinary Differential Equations Page 93


This proves the theorem.
Corollary: If the coefficients 𝑎𝑖 (𝑥), ∀ 𝑖 = 1,2, … , 𝑛 in equation (40) are constants, then
equation (48) gives
𝑊(𝜙1 , 𝜙2 , … , 𝜙𝑛 )(𝑥) = 𝑒 −𝑎1 (𝑥−𝑥0) 𝑊(𝜙1 , 𝜙2 , … , 𝜙𝑛 )(𝑥0 ). (49)
This is the proof of the theorem of linear differential equation of order 𝑛 of constant
coefficients.
Note: 𝑊(𝜙1 , 𝜙2 , … , 𝜙𝑛 )(𝑥0 ) ≠ 0 ⟺ 𝑊(𝜙1 , 𝜙2 , … , 𝜙𝑛 )(𝑥) ≠ 0 ∀ 𝑥 ∈ 𝐼 or
𝑊(𝜙1 , 𝜙2 , … , 𝜙𝑛 )(𝑥0 ) = 0 ⇔ 𝑊(𝜙1 , 𝜙2 , … , 𝜙𝑛 )(𝑥) = 0 ∀ 𝑥 ∈ 𝐼 as exponential function
never zero. The first is true provided 𝜙1 , 𝜙2 , … , 𝜙𝑛 are linearly independent solutions of
𝐿(𝑦) = 𝑦 (𝑛) + 𝑎1 (𝑥)𝑦 (𝑛−1) + 𝑎2 (𝑥)𝑦 (𝑛−2) + ⋯ + 𝑎𝑛 (𝑥)𝑦 = 0 and the second is true provided
𝜙1 , 𝜙2 , … , 𝜙𝑛 are solutions of 𝐿(𝑦) = 0.
Example 01: Let 𝜙 be a real valued non-trivial solution of 𝑦 ′′ + 𝛼(𝑥)𝑦 = 0 on 𝑎 < 𝑥 < 𝑏,
and let 𝜓 be a real valued non-trivial solution of 𝑦 ′′ + 𝛽(𝑥)𝑦 = 0 on 𝑎 < 𝑥 < 𝑏. Here 𝛼, 𝛽
are real valued continuous functions. Suppose that 𝛽(𝑥) > α(𝑥), (𝑎 < 𝑥 < 𝑏). Show that
if 𝑥1 and 𝑥2 are successive zeros of 𝜙 on 𝑎 < 𝑥 < 𝑏, then 𝜓 must vanish at some point 𝜉,
𝑥1 < 𝜉 < 𝑥2 .
Solution: Assume that 𝜙 and 𝜓 are real solutions of
𝑦 ′′ + 𝛼(𝑥)𝑦 = 0, (50)
and
𝑦 ′′ + 𝛽(𝑥)𝑦 = 0 (51)
respectively on (𝑎, 𝑏). The proof is by the method of contradiction. Suppose 𝜓 does not
vanish in (𝑥1 , 𝑥2 ). Then either 𝜓 is positive in (𝑥1 , 𝑥2 ) or 𝜓 is negative in (𝑥1 , 𝑥2 ). Suppose
𝜓 is positive in (𝑥1 , 𝑥2 ). Without loss of generality, let us assume that 𝜙(𝑥) > 0 on (𝑥1 , 𝑥2 ).
Multiplying (50) and (51) by 𝜓 and 𝜙 respectively and subtraction leads to
(𝜙 ′ )′ 𝜓 − (𝜓′ )′ 𝜙 − (𝛽(𝑥) − 𝛼(𝑥)) 𝜙𝜓 = 0,
which, on integration gives us
𝑥 𝑥
∫𝑥1 [(𝜙 ′ )′ 𝜓 − (𝜓′ )′ 𝜙]𝑑𝑥 = ∫𝑥1 (𝛽(𝑥) − 𝛼(𝑥)) 𝜙𝜓𝑑𝑥. (52)
2 2

Since 𝛽(𝑥) > 𝛼(𝑥) and both 𝜙 > 0, 𝜓 > 0 in the interval (𝑥1 , 𝑥2 ), then we have
𝑥
∫𝑥1 [(𝜙 ′ )′ 𝜓 − (𝜓′ )′ 𝜙]𝑑𝑥 > 0. (53)
2

Using the identity


𝑑
[(𝜙 ′ 𝜓 − 𝜙𝜓′ )] = (𝜙 ′ )′ 𝜓 − (𝜓′ )′ 𝜙,
𝑑𝑥
now the inequality (53) implies
[(𝜙 ′ 𝜓 − 𝜙𝜓′ )]𝑥𝑥12 > 0
𝑥 𝑥
⟹ [𝜙 ′ 𝜓]𝑥12 − [𝜙𝜓′ ]𝑥12 > 0
𝑥
⟹ [𝜙 ′ 𝜓]𝑥12 − 0 > 0 (∵ 𝜙(𝑥1 ) = 𝜙(𝑥2 ) = 0)
⟹ 𝜙 ′ (𝑥2 )𝜓(𝑥2 ) − 𝜙 ′ (𝑥1 )𝜓(𝑥1 ) > 0. (54)
Now 𝜙(𝑥) > 0, ∀ 𝑥 ∈ (𝑥1 , 𝑥2 ) and 𝜙(𝑥1 ) = 0 = 𝜙(𝑥2 ), as 𝜙 is strictly increasing at 𝑥1 and
strictly decreasing at 𝑥2 , we have 𝜙 ′ (𝑥1 ) > 0. Similarly, 𝜙 ′ (𝑥2 ) < 0.
In view above discussion, the inequality (54) leads to a contradiction. [Because
𝜙 ′ (𝑥1 )𝜓(𝑥1 ) > 0 and 𝜙 ′ (𝑥2 )𝜓(𝑥2 ) < 0 due to 𝜙 ′ (𝑥2 ) < 0, and hence the complete left-hand

S25013: Ordinary Differential Equations Page 94


side of inequality (54) is negative, which is a contradiction]. Therefore 𝜓 takes both
positive and negative values in the interval (𝑥1 , 𝑥2 ) and hence there exists 𝜉 ∈ (𝑥1 , 𝑥2 ) such
that 𝜓(𝜉) = 0.
Example 02: Show that any solution 𝜓 of 𝑦 ′′ + 𝑥𝑦 = 0 on 0 < 𝑥 < ∞ has an infintiy of
zeros there.
Solution: Given that 𝜓 is any solutin of
𝑦 ′′ + 𝑥𝑦 = 0 on 0 < 𝑥 < ∞, with 𝛽(𝑥) = 𝑥. (55)
Now we consider the equation
𝑦 ′′ + 𝑦 = 0 with 𝛼(𝑥) = 1. (56)
The function 𝜙(𝑥) = sin 𝑥 is the one of solutions of the equation (56) and has infinitely
many zeros 𝑥 = 𝑛𝜋 , 𝑛 = 0, 1, 2,3, ⋯ . Since 𝛽(𝑥) > 𝛼(𝑥) for all 𝑥 ∈ (1, ∞) and by
application of Example 01, one can easily conclude that the solution 𝜓 of (55) also has
infinitely many zeros on (1, ∞) and consequently on (0, ∞).

SOLVED PROBLEMS 01
Problem 02-02-01:
Let 𝜙1 (𝑥) and 𝜙2 (𝑥) be two solutions of (1 − 𝑥 2 )𝑦 ′′ − 2𝑥𝑦 ′ + sec 𝑥 𝑦 = 0, with Wronskian
1 1
𝑊(𝑥). If 𝜙1 (0) = 1, 𝜙1′ (0) = 0 and 𝑊 (2) = 3, then find the value 𝜙2′ (0).
1 1
Solution: Given that 𝜙1 (0) = 1, 𝜙1′ (0) = 0 and 𝑊 ( ) = . (57)
2 3
By definition of Wronskian,
𝜙 (0) 𝜙2 (0)
𝑊(0) = | 1′ |
𝜙1 (0) 𝜙2′ (0)
1 𝜙2 (0)
=| |
0 𝜙2′ (0)
= 1. 𝜙2′ (0) − 0
= 𝜙2′ (0).
⟹ 𝜙2′ (0) = 𝑊(0). (58)
Re-write the given equation as follows
(1 − 𝑥 2 )𝑦 ′′ − 2𝑥𝑦 ′ + sec 𝑥 𝑦 = 0
2𝑥 sec 𝑥
⟹ 𝑦 ′′ − (1−𝑥 2) 𝑦 ′ + (1−𝑥 2) 𝑦 = 0. (59)
Comparing it with standard differential equation 𝑦 ′′ + 𝑎1 (𝑥)𝑦 ′ + 𝑎2 (𝑥)𝑦 = 0 , we get
2𝑥
𝑎1 (𝑥) = − (1−𝑥 2).
From a fromula for Wronskian (Abel’s formula)
𝑥
𝑊(𝜙1 , 𝜙2 )(𝑥) = exp (− ∫𝑥 𝑎1 (𝑠)𝑑𝑠) 𝑊(𝜙1 , 𝜙2 )(𝑥0 ). (60)
0
1
Therefore, replacing 𝑥 by and 𝑥0 by 0 in (60) and use the equation (58), we get
2
1
1 2𝑠
𝑊(𝜙1 , 𝜙2 ) (2) = exp (− ∫02 (− 1−𝑠2 ) 𝑑𝑠) 𝑊(𝜙1 , 𝜙2 )(0)
1
2𝑠
= exp (∫02 (1−𝑠2 ) 𝑑𝑠) 𝜙2′ (0)

S25013: Ordinary Differential Equations Page 95


1
= exp (− log(1 − 𝑠 2 )⌉20 ) 𝜙2′ (0)
1
= exp (− log (1 − 4) + log 1) 𝜙2′ (0)
3
= exp (− log (4 )) 𝜙2′ (0)
3 −1
= exp (log (4 ) ) 𝜙2′ (0)
3 −1
= (4 ) 𝜙2′ (0)
4
= 3 𝜙2′ (0).
1 4
⟹ 𝑊 (2) = 3 𝜙2′ (0)
3 1
⟹ 𝜙2′ (0) = 4 𝑊 (2)
3 .1 1 1
⟹ 𝜙2′ (0) = 4 3 (∵ 𝑊 (2) = 3)
1
⟹ 𝜙2′ (0) = 4.
Problem 02-02-02:
Let 𝜙1 (𝑥) and 𝜙2 (𝑥) be complete set of solutions to the differential equation
𝑦 ′′ − 2𝑥𝑦 ′ + sin 𝑥 𝑒 2𝑥 𝑦 = 0 , 𝑥 ∈ [0,1] , with 𝜙1 (0) = 0, 𝜙1′ (0) = 1, 𝜙2 (0) = 1, 𝜙2′ (0) = 1 .
Then fnd value of Wronskian at 𝑥 = 1.
Solution: By definition of Wronskian,
𝜙 (0) 𝜙2 (0)
𝑊(0) = | 1′ |
𝜙1 (0) 𝜙2′ (0)
0 1
=| |
1 1
= 0.1 − 1.1
= −1. (61)
Comparing the given equation with 𝑦 + 𝑎1 (𝑥)𝑦 + 𝑎2 (𝑥)𝑦 = 0, we get 𝑎1 (𝑥) = −2𝑥.
′′ ′

Now, by Abel’s formula for Wronskian, we have


𝑥
𝑊(𝜙1 , 𝜙2 )(𝑥) = exp (− ∫𝑥 𝑎1 (𝑠)𝑑𝑠) 𝑊(𝜙1 , 𝜙2 )(𝑥0 ). (62)
0

Therefore, replacing 𝑥 by 1 and 𝑥0 by 0 in (62) and use the equation (61), we get
1
𝑊(𝜙1 , 𝜙2 )(1) = exp (− ∫0 (−2𝑠)𝑑𝑠) 𝑊(𝜙1 , 𝜙2 )(0)
1
= exp (∫0 2𝑠 𝑑𝑠) (−1)
1
𝑠2
= − exp (2 2 ⌉ )
0
= − exp(𝑠 2 ⌉10 )
= − exp(12 − 0)
= − exp(1)
= −𝑒 1
⟹ 𝑊(1) = −𝑒.

S25013: Ordinary Differential Equations Page 96


Problem 02-02-03:
Let 𝑎1 (𝑥) and 𝑎2 (𝑥) be continuous real valued functions defined on [−1,1] and let
𝜙𝑖 : [−1,1] → ℝ , for 𝑖 = 1, 2 be solutions of the ordinary differential equation 𝑦 ′′ +
𝑎1 (𝑥)𝑦 ′ + 𝑎2 (𝑥)𝑦 = 0, 𝑥 ∈ [−1,1] satisfying 𝜙1 ≥ 0, 𝜙2 ≤ 0 and 𝜙1 (0) = 𝜙2 (0) = 0. Let 𝑊
denotes the Wronskian of 𝜙1 , and 𝜙2 . Then which of the followings is / are true?
(a) 𝜙1 and 𝜙2 are linearly independent
(b) 𝜙1 and 𝜙2 are linearly dependent
(c) 𝑊(𝑥) = 0, ∀ 𝑥 ∈ [−1,1]
(d) 𝑊(𝑥) ≠ 0, for some 𝑥 ∈ [−1,1].
Solution: Using definition of Wronskian of two solutions 𝜙1 and 𝜙2 ,
𝜙 (𝑥) 𝜙2 (𝑥)
𝑊(𝑥) = | 1′ |
𝜙1 (𝑥) 𝜙2′ (𝑥)
= 𝜙1 (𝑥)𝜙2′ (𝑥) − 𝜙1′ (𝑥)𝜙2 (𝑥). (63)
Hence, in particular, we have from given results
𝜙 (𝑥) 𝜙2 (𝑥)
𝑊(𝜙1 , 𝜙2 )(0) = | 1′ |
𝜙1 (𝑥) 𝜙2′ (𝑥)
0 0
= | ′ (0) |
𝜙1 𝜙2′ (0)
=0
⟹ 𝑊(0) = 0.
Now, by Abel’s formula for Wronskian, we have
𝑥
𝑊(𝜙1 , 𝜙2 )(𝑥) = exp (− ∫𝑥 𝑎1 (𝑠)𝑑𝑠) 𝑊(𝜙1 , 𝜙2 )(𝑥0 ). (64)
0

Therefore, replacing 𝑥0 by 0 in (62) and use the equation (64), we get


𝑥
𝑊(𝜙1 , 𝜙2 )(𝑥) = exp(− ∫0 𝑎1 (𝑠)𝑑𝑠)𝑊(𝜙1 , 𝜙2 )(0)
𝑥
= exp(∫0 𝑎1 (𝑠)𝑑𝑠) × 0
= 0, ∀ 𝑥 ∈ [−1,1]
This implies two solutions 𝜙1 and 𝜙2 are linearly dependent on [−1,1].
Problem 02-02-04:
1 1
Consider the equation 𝑦 ′′ + 𝑥 𝑦 ′ − 𝑥 2 𝑦 = 0, for 𝑥 > 0.
(a) Show that there is a solution of the form 𝑥 𝑟 , where 𝑟 is a constant.
(b) Find two linearly independent solutions for 𝑥 > 0, and prove that they are lineraly
independent.
(c) Find the two solutions 𝜙1 , 𝜙2 satisfying
𝜙1 (1) = 1, 𝜙2 (1) = 0
𝜙1′ (1) = 0, 𝜙2′ (1) = 1.
1 1
Solution: (a) Let 𝐿(𝑦) = 𝑦 ′′ + 𝑥 𝑦 ′ − 𝑥 2 𝑦 = 0, for 𝑥 > 0. (65)
Now
1 1
𝐿(𝑥 𝑟 ) = (𝑥 𝑟 )′′ + 𝑥 (𝑥 𝑟 )′ − 𝑥 2 𝑥 𝑟
1 1
= (𝑟𝑥 𝑟−1 )′ + 𝑥 (𝑟𝑥 𝑟−1 ) − 𝑥 2 𝑥 𝑟

S25013: Ordinary Differential Equations Page 97


= 𝑟(𝑟 − 1)𝑥 𝑟−2 + 𝑟𝑥 𝑟−2 − 𝑥 𝑟−2
= [𝑟(𝑟 − 1) + (𝑟 − 1)]𝑥 𝑟−2
= (𝑟 + 1)(𝑟 − 1)𝑥 𝑟−2 . (66)
This shows that the function of the form 𝑥 is a solution of (65) if and only if 𝑟 satisfies
𝑟

the polynomial equation (𝑟 + 1)(𝑟 − 1) = 0 for 𝑥 > 0. That is we have


𝐿(𝑥 𝑟 ) = (𝑟 + 1)(𝑟 − 1)𝑥 𝑟−2 = 0 ⟺ (𝑟 + 1)(𝑟 − 1) = 0 for 𝑥 > 0.
Therefore, we have (𝑟 + 1)(𝑟 − 1) = 0 ⟹ 𝑟 = 1, −1. So for these value of constant 𝑟, we
have the solutions of the equation (65) are of the form
1
𝑦1 (𝑥) = 𝑥1 = 𝑥 and 𝑦2 (𝑥) = 𝑥 −1 = 𝑥.
1
(b) By (a), we assume that 𝑦1 (𝑥) = 𝑥 and 𝑦2 (𝑥) = are two solutions of the equation (65).
𝑥
Hence, we have
1
𝑦1 𝑦2 𝑥 1 1 1 1 2
𝑊(𝑥) = |𝑦 ′ 𝑦2′ | = | 𝑥
1| = −𝑥 𝑥 2 − 1 𝑥 = − 𝑥 − 𝑥 = − 𝑥 ≠ 0 for all 𝑥 > 0.
1 1 − 𝑥2
1
This implies that 𝑦1 (𝑥) = 𝑥 and 𝑦2 (𝑥) = 𝑥 are two solutions of the equation (65) are
linearly independent for 𝑥 > 0.
(c) Let us find two solutions satisfying the given conditions.
(1) We know that any solution of the equation (65) can be written as
𝑦(𝑥) = 𝑐1 𝑦1 (𝑥) + 𝑐2 𝑦2 (𝑥). (67)
Now, from first pair of initiial conditions 𝜙1 (1) = 1, 𝜙1′ (1) = 0, we have
𝜙1 (1) = 1 ⟹ 𝑐1 𝑦1 (1) + 𝑐2 𝑦2 (1) = 1 ⟹ 𝑐1 + 𝑐2 = 1, (68)
and
𝜙1′ (1) = 0 ⟹ 𝑐1 𝑦1′ (1) + 𝑐2 𝑦2′ (1) = 0 ⟹ 𝑐1 − 𝑐2 = 0. (69)
Solving (68) and (69), we get
1 1
2𝑐1 = 1 ⟹ 𝑐1 = 2 and 2𝑐2 = 1 ⟹ 𝑐2 = 2.
Therefore, the first solution is given by
1 1 1
𝜙1 (𝑥) = 𝑦1 (𝑥) + 𝑦2 (𝑥) = [𝑥 + 𝑥 −1 ].
2 2 2
(2) Now, from second pair of initiial conditions 𝜙2 (1) = 0, 𝜙2′ (1) = 1, we have
𝜙2 (1) = 0 ⟹ 𝑐1 𝑦1 (1) + 𝑐2 𝑦2 (1) = 0 ⟹ 𝑐1 + 𝑐2 = 0, (70)
and
𝜙2′ (1) = 1 ⟹ 𝑐1 𝑦1′ (1) + 𝑐2 𝑦2′ (1) = 1 ⟹ 𝑐1 − 𝑐2 = 1. (71)
Solving (70) and (71), we get
1 1
2𝑐1 = 1 ⟹ 𝑐1 = 2 and 2𝑐2 = −1 ⟹ 𝑐2 = − 2.
Hence, the second solution is given by
1 1 1
𝜙2 (𝑥) = 𝑦1 (𝑥) − 𝑦2 (𝑥) = [𝑥 − 𝑥 −1 ].
2 2 2
Problem 02-02-05:
Find two linearly independent solutions of the equation (3𝑥 − 1)2 𝑦 ′′ + (9𝑥 − 3)𝑦 ′ − 9𝑦 =
1
0, for 𝑥 > 3.

S25013: Ordinary Differential Equations Page 98


(3𝑥−1) 9
Solution: Let 𝐿(𝑦) = 𝑦 ′′ + 3 (3𝑥−1)2 𝑦 ′ − (3𝑥−1)2 𝑦 = 0
3 9
⟹ 𝐿(𝑦) = 𝑦 ′′ + (3𝑥−1) 𝑦 ′ − (3𝑥−1)2 𝑦 = 0 (72)
Recall (a) of Problem 02-09 with 𝑥 replaced by (3𝑥 − 1), we have
3 9
𝐿[(3𝑥 − 1)𝑟 ] = ((3𝑥 − 1)𝑟 )′′ + (3𝑥−1) ((3𝑥 − 1)𝑟 )′ − (3𝑥−1)2 (3𝑥 − 1)𝑟
3 9
= (𝑟(3𝑥 − 1)𝑟−1 )′ . 3 + (3𝑥−1) 𝑟(3𝑥 − 1)𝑟−1 . 3 − (3𝑥−1)2 (3𝑥 − 1)𝑟
= 9𝑟(𝑟 − 1)(3𝑥 − 1)𝑟−2 + 9 𝑟(3𝑥 − 1)𝑟−2 − 9(3𝑥 − 1)𝑟−2
= [9𝑟(𝑟 − 1) + 9 𝑟 − 9](3𝑥 − 1)𝑟−2
= [9𝑟(𝑟 − 1) + 9( 𝑟 − 1)](3𝑥 − 1)𝑟−2
= [(9𝑟 + 9)( 𝑟 − 1)](3𝑥 − 1)𝑟−2
= [9(𝑟 + 1)( 𝑟 − 1)](3𝑥 − 1)𝑟−2 . (73)
This shows that the function of the form (3𝑥 − 1) is a solution of (72) if and only if 𝑟
𝑟
1
satisfies the polynomial equation 9(𝑟 + 1)(𝑟 − 1) = 0 for 𝑥 > 3. That is we have
1
𝐿[(3𝑥 − 1)𝑟 ] = 9(𝑟 + 1)(𝑟 − 1)(3𝑥 − 1)𝑟−2 = 0 ⟺ 9(𝑟 + 1)(𝑟 − 1) = 0 for 𝑥 > 3.
Therefore, we have (𝑟 + 1)(𝑟 − 1) = 0 ⟹ 𝑟 = 1, −1. So for these value of constant 𝑟, we
have the solutions of the equation (72) are of the form
1
𝑦1 (𝑥) = (3𝑥 − 1)1 = 3𝑥 − 1 and 𝑦2 (𝑥) = (3𝑥 − 1)−1 = 3𝑥−1.
Hence, we also have
𝑦1 𝑦2
𝑊(𝑥) = |𝑦 ′ 𝑦2′ |
1
1
3𝑥 − 1 3𝑥−1
=| 3 |
3 − (3𝑥−1)2
3 3
= −(3𝑥 − 1). (3𝑥−1)2 − 3. 3𝑥−1
3 9
= − 3𝑥−1 − 3𝑥−1
12 1
= − 3𝑥−1 ≠ 0 for all 𝑥 > 3.
1
This implies that 𝑦1 (𝑥) = 3𝑥 − 1 and 𝑦2 (𝑥) = 3𝑥−1 are two solutions of the equation (72)
1
are linearly independent for 𝑥 > 3.

SELF-TEST 01
MCQ 02-02-01
Suppose 𝜙1 and 𝜙2 are solutions of the equation 𝑥 2 𝑦 ′′ + 2𝑥 3 𝑦 ′ − 𝑥 −2 𝑦 = 0. Then for some
constant 𝑐, Wronskian 𝑊(𝜙1 , 𝜙2 )(𝑥) =
2
A. cex
2
B. ce−x
C. cex
D. ce−x

S25013: Ordinary Differential Equations Page 99


MCQ 02-02-02
2
Assume that the solutions of homogeneous equation 𝐿(𝑦) = 𝑦 ′′ − 𝑥 2 𝑦 = 0 is of the form
𝑥 𝑟 . Then what are values of 𝑟?
A. r = 0, 1
B. r = 2, 1
C. r = 2, −1
D. r = 1, −1.
MCQ 02-02-03
Consider the equation 𝑦 ′′ + 𝛼(𝑥)𝑦 = 0, where 𝛼 is a continuous function on the domain
−∞ < 𝑥 < ∞. Let 𝜙1 and 𝜙2 be the fundamental solutions satisfying 𝜙1 (0) = 1, 𝜙2 (0) =
0, 𝜙1′ (0) = 1, 𝜙2′ (0) = 1. Then the Wronskian 𝑊(𝜙1 , 𝜙2 )(𝑥) =
A. 4
B. 3
C. 2
D. 1.
MCQ 02-02-04
The normal form of Bessel’s equation 𝑥 2 𝑦 ′′ + 𝑥𝑦 ′ + (𝑥 2 − 𝑛2 )𝑦 = 0 is given by
1−4n2
A. y ′′ − (1 + 4x2
)y = 0
1−4x2
B. y ′′ − (1 + )y = 0
4n2
1−4x2
C. y ′′ + (1 + 4n2
)y = 0
1−4n 2
D. y ′′ + (1 + 4x2
) y = 0.

SHORT ANSWER QUESTIONS 01


SAQ 02-02-01
Find the longest interval 𝐼 ⊆ ℝ such that there exists a unique solution to the initial value
problem (𝑥 − 1)𝑦 ′′ − 3𝑥𝑦 ′ + 4𝑦 = 𝑥(𝑥 − 1), 𝑦(−2) = 2, 𝑦 ′ (−2) = 1.
Solution: Rewritting the equation in standard form:
(𝑥 − 1)𝑦 ′′ − 3𝑥𝑦 ′ + 4𝑦 = 𝑥(𝑥 − 1), 𝑦(−2) = 2, 𝑦 ′ (−2) = 1
3𝑥 4
⟹ 𝑦 ′′ − (𝑥−1) 𝑦 ′ + (𝑥−1) 𝑦 = 𝑥, 𝑦(−2) = 2, 𝑦 ′ (−2) = 1.
The intervals where the hypotheses in the Existence and Uniqueness Theorem are satisfied,
i. e. where the coefficients of the given equations are continous, are 𝐼1 = (−∞, 1) and 𝐼2 =
(1, ∞). Since the initial conditions given at point belong to 𝐼1 , the solution domain is 𝐼1 =
(−∞, 1).
SAQ 02-02
Find the Wronskian 𝑊(𝑥) of two solutions of the equation (1 − 𝑥 2 )𝑦 ′′ − 2𝑥𝑦 ′ +
𝛼(𝛼 + 1)𝑦 = 0.
Solution: Rewritting the equation in standard form,
2𝑥 𝛼(𝛼+1)
𝑦 ′′ − (1−𝑥2) 𝑦 ′ + (1−𝑥2 )
𝑦 =0

S25013: Ordinary Differential Equations Page 100


2𝑥
and we find that 𝑎1 (𝑥) = − (1−𝑥 2). The Wronskian, by Abel’s formula, is given by
𝑊(𝜙1 , 𝜙2 )(𝑥) = 𝑐 exp(−∫ 𝑎1 (𝑥)𝑑𝑥)
2𝑥
= 𝑐 exp (− ∫ − 1−𝑥 2 𝑑𝑥)
= 𝑐 exp(− log(1 − 𝑥 2 ))
= 𝑐(|1 − 𝑥 2 |)−1
𝑐
⟹ 𝑊(𝜙1 , 𝜙2 )(𝑥) = |1−𝑥 2|,
where 𝑐 is a some constant.
SAQ 02-03
If 𝜙1 and 𝜙2 are linearly independent solutions of 𝑥 2 𝑦 ′′ − 2𝑥𝑦 ′ + (3 + 𝑥)𝑦 = 0, and if
𝑊(𝜙1 , 𝜙2 )(2) = 3, then find the value of 𝑊(𝜙1 , 𝜙2 )(5).
Solution: Rewritting the equation in standard form,
2𝑥 (3+𝑥) 2 (3+𝑥)
𝑦 ′′ − 𝑥 2 𝑦 ′ + 𝑦 = 0 ⟹ 𝑦 ′′ − 𝑥 𝑦 ′ + 𝑦 = 0.
𝑥2 𝑥2
2
On comparing, we find that 𝑎1 (𝑥) = − 𝑥. The Wronskian, by Abel’s formula, is given by
𝑊(𝜙1 , 𝜙2 )(𝑥) = 𝑐 exp(−∫ 𝑎1 (𝑥)𝑑𝑥)
2
= 𝑐 exp (− ∫ − 𝑥 𝑑𝑥)
= 𝑐 exp(2 log 𝑥)
⟹ 𝑊(𝜙1 , 𝜙2 )(𝑥) = 𝑐𝑥 2 , (74)
where 𝑐 is a some constant. Since 𝑊(2) = 3, thus from (74), we have
3
𝑐22 = 3 ⟹ 𝑐. 4 = 3 ⟹ 𝑐 = 4.
3
Therefore, we obatin 𝑊(𝜙1 , 𝜙2 )(𝑥) = 4 𝑥 2 and consequently we have
3 75
𝑊(𝜙1 , 𝜙2 )(5) = 4 52 = .
4

SUMMARY
In this Unit, the differential equation is considered on an interval 𝐼 of the type
𝐿(𝑦) = 𝑦 (𝑛) + 𝑎1 (𝑥)𝑦 (𝑛−1) + 𝑎2 (𝑥)𝑦 (𝑛−2) + ⋯ + 𝑎𝑛 (𝑥)𝑦 = 0.
For its solutions 𝜙1 , 𝜙2 , ⋯ , 𝜙𝑛 and 𝑥0 ∈ 𝐼, the Wronskian 𝑊(𝑥) is given by
𝑥
𝑊(𝜙1 , 𝜙2 , ⋯ , 𝜙𝑛 )(𝑥) = exp [− ∫𝑥 𝑎1 (𝑡)𝑑𝑡] 𝑊(𝜙1 , 𝜙2 , … , 𝜙𝑛 )(𝑥0 ), ∀ 𝑥 ∈ 𝐼
0

and some examples are solved on it. Further, the uniqueness theorem is proved for IVP.

END OF UNIT EXERCISES


1
1) Show that 𝜙1 = √𝑥 and 𝜙2 = are fundamental solutions of the differential equation
𝑥
𝐿(𝑦) = 2𝑥 2 𝑦 ′′ + 3𝑥𝑦 ′ − 𝑦 = 0.
2) Consider the equation 𝐿(𝑦) = 𝑦 ′′ + 𝑎1 (𝑥)𝑦 ′ + 𝑎2 (𝑥)𝑦 = 0 . Show that 𝑎1 (𝑥) nad 𝑎2 (𝑥)
are uniquely determined by the fundamental solutions 𝜙1 and 𝜙2 of the equation 𝐿(𝑦) =

S25013: Ordinary Differential Equations Page 101


𝜙1 𝜙2 𝜙1′ 𝜙2′
| ′′ | | ′′ |
𝜙1 𝜙2′′ 𝜙1 𝜙2′′
0 and are given by 𝑎1 (𝑥) = − 𝑊(𝜙 ,𝜙 )(𝑥) and 𝑎2 (𝑥) = , where 𝑊 denote the
1 2 𝑊(𝜙1 ,𝜙2 )(𝑥)
Wroskian of solutions.
3) Investigate whether the functions 𝑦1 (𝑥) = 𝑥 + 2 and 𝑦2 (𝑥) = 2𝑥 − 1 are linearly
independent.
4) Write a homogeneous linear differential equation if its fundamental system of solutions
is known: 𝑥, 𝑒 𝑥 .
5) One solution of the differential equation 𝐿(𝑦) = 𝑥 2 𝑦 ′′ − 2𝑥𝑦 ′ + 2𝑦 = 0, 𝑥 > 0 is
𝜙1 (𝑥) = 𝑥. Obtain second independent solution.

KEY WORDS
Linear differential equation, Determinant, Wronskian of solutions, IVP.

S25013: Ordinary Differential Equations Page 102


UNIT 02-03: REDUCTION OF THE ORDER
LEARNING OBJECTIVES
After successful completion of this unit, you will be able to
❖ explain reduction of the order of a differential equation by order one using its non -
vanishing solution
❖ apply it to obtain remaining linearly independent solution of the differential
equation 𝐿(𝑦) = 0 and hence using these basic solutions to obtain a particular
solution of 𝐿(𝑦) = 𝑏(𝑥)
03-01: INTRODUCTION:
In the case of a linear homogeneous differential equation if one solution is known, then it
is possible to find another solution by reducing the order of the equation.

03-02: REDUCTION OF THE ORDER OF A HOMOGENEOUS EQUATION :


The method of reduction of the order of a homogeneous equation is illustrated in the
following theorem.
Theorem 1: Let 𝜙1 be a solution of 𝐿(𝑦) = 𝑦 (𝑛) + 𝑎1 (𝑥)𝑦 (𝑛−1) + 𝑎2 (𝑥)𝑦 (𝑛−2) + ⋯ +
𝑎𝑛 (𝑥)𝑦 = 0 on an interval 𝐼 and suppose 𝜙1 (𝑥) ≠ 0 on 𝐼. If 𝑣2 , 𝑣3 , … , 𝑣𝑛 is any basis on 𝐼
for the solutions of the linear differential equation of order (𝑛 − 1),
(𝑛−1) (𝑛−2)
𝑀(𝑦) ≡ 𝜙1 𝑣 (𝑛−1) + ⋯ + [𝑛𝜙1 + (𝑛 − 1)𝜙1 + ⋯ + 𝑎𝑛−1 𝜙1 ]𝑣 = 0
and if 𝑣𝑘 = 𝑢′𝑘 , 𝑘 = 2,3, … , 𝑛, then 𝜙1 , 𝑢2 𝜙1 , 𝑢3 𝜙1 , … , 𝑢𝑛 𝜙1 is a basis for the solutions of
𝐿(𝑦) = 0 on 𝐼.
Proof: Let 𝐿(𝑦) = 𝑦 (𝑛) + 𝑎1 (𝑥)𝑦 (𝑛−1) + 𝑎2 (𝑥)𝑦 (𝑛−2) + ⋯ + 𝑎𝑛 (𝑥)𝑦 = 0. (1)
Suppose we have obtained by some means one solution 𝜙1 (𝑥) ≠ 0 of 𝐿(𝑦) = 0 on 𝐼.
⟹ 𝐿(𝜙1 ) = 0 (2)
Now with this information, it is possible to reduce the order of the equation by one. (This
idea is the same one employed in the variation of constant method.)
We try to find solutions 𝜙 of 𝐿(𝑦) = 0 of the form 𝜙 = 𝑢𝜙1 , where 𝑢 is some function.
Now 𝜙 = 𝑢𝜙1 is to be a solution, we must have
𝐿(𝑢𝜙1 ) = 0 (3)
(𝑛) (𝑛−1)
⟹ (𝑢𝜙1 ) + 𝑎1 (𝑥)(𝑢𝜙1 ) + ⋯ + 𝑎𝑛 (𝑥)(𝑢𝜙1 (𝑥)) = 0
𝑛 𝑛 𝑛 (𝑛−1) 𝑛 (𝑛)
⟹ [𝑢(𝑛) 𝜙1 + ( ) 𝑢(𝑛−1) ϕ1′ + ( ) 𝑢(𝑛−2) ϕ1′′ + ⋯ + ( ) 𝑢′ 𝜙1 + ( ) 𝑢𝜙1 ]
1 2 𝑛−1 𝑛
(𝑛−1) 𝑛 − 1 (𝑛−2) ′ 𝑛 − 1 (𝑛−1)
+𝑎1 (𝑥) [𝑢 𝜙1 + ( )𝑢 ϕ1 + ⋯ + ( ) 𝑢𝜙1 ] + ⋯ + 𝑎𝑛 (𝑥)(𝑢𝜙) = 0
1 𝑛−1
𝑛 𝑛 𝑛−1 ′
⟹ 𝑢(𝑛) 𝜙1 + [( ) 𝜙1′ + 𝑎1 (𝑥)𝜙1 ] 𝑢(𝑛−1) + [( ) 𝜙1′′ + 𝑎1 (𝑥) ( ) 𝜙1 + 𝑎2 (𝑥)𝜙1 ] 𝑢(𝑛−2)
1 2 1
(𝑛−1) (𝑛−2)
+ ⋯ + 𝑢′ [𝑛𝜙1 + (𝑛 − 1)𝑎1 (𝑥)𝜙1 + ⋯ + 𝑎𝑛−1 (𝑥)𝜙1 ]
(𝑛) (𝑛−1)
+𝑢[𝜙1 + 𝑎1 (𝑥)𝜙1 + ⋯ + 𝑎𝑛−1 (𝑥)𝜙1′ + 𝑎𝑛 (𝑥)𝜙1 ] = 0. (4)
We see that the coefficient of 𝑢 in the expression of 𝐿(𝑢𝜙1 ) is 𝐿(𝜙1 ) which is zero as 𝜙1 is
a solution of 𝐿(𝑦) = 0. So that the equation (4) shows that 𝑦 = 𝑢 satisfies the equation
𝑏1 𝑦 (𝑛) + 𝑏2 𝑦 (𝑛−1) + ⋯ + 𝑏𝑛−1 𝑦′ = 0, (5)

S25013: Ordinary Differential Equations Page 103


𝑛 𝑛 𝑛−1 ′
where 𝑏1 = 𝜙1 (𝑥) ≠ 0, 𝑏2 = ( ) 𝜙1′ + 𝑎1 𝜙1 , 𝑏3 = ( ) 𝜙1′′ + 𝑎1 (𝑥) ( ) 𝜙1 + 𝑎2 (𝑥)𝜙1 ,
1 2 1
𝑏4 , … , 𝑏𝑛−2 are function of 𝑥 and
(𝑛−1) (𝑛−2)
𝑏𝑛−1 = 𝑛𝜙1 + (𝑛 − 1)𝑎1 (𝑥)𝜙1 + ⋯ + 𝑎𝑛−1 (𝑥)𝜙1.
We write 𝑣 = 𝑢′ (6)
𝑥
⟹ 𝑢(𝑥) = ∫𝑥0 𝑣(𝑡)𝑑𝑡. (7)
By using the transformation (6), the equation (5) reduces to
𝑏1 𝑣 (𝑛−1) + 𝑏2 𝑣 (𝑛−2) + ⋯ + 𝑏𝑛−1 𝑣 = 0
𝑛(𝑛−1)
𝜙1 𝑣 (𝑛−1) + [𝑛ϕ1′ + 𝑎1 𝜙1 ]𝑣 (𝑛−2) + [ 𝜙1′′ + 𝑎1 (𝑥)(𝑛 − 1)𝜙1′ + 𝑎2 (𝑥)𝜙1 ] 𝑣 (𝑛−3)
2
(𝑛−1) (𝑛−1)
+ ⋯ + [𝑛𝜙1 + 𝑎1 (𝑥)(𝑛 − 1)𝜙1 +
⋯ + 𝑎𝑛−1 (𝑥)𝜙1 ]𝑣 = 0. (8)
Since 𝜙1 ≠ 0 , equation (8) being a linear equation of order (𝑛 − 1) and hence it has
(𝑛 − 1) linearly independent solutions 𝑣2 , 𝑣2 , … , 𝑣𝑛 on 𝐼 . Hence from equation (7), the
corresponding values of 𝑢 are given by
𝑥
𝑢𝑘 (𝑥) = ∫𝑥 𝑣𝑘 (𝑡)𝑑𝑡, 𝑘 = 2,3, … , 𝑛.
0

The functions 𝜙1 , 𝑢2 𝜙1 , … , 𝑢𝑛 𝜙1 are solutions of 𝐿(𝑦) = 0 on 𝐼.


Now we will show that the solutions 𝜙1 , 𝑢2 𝜙1 , … , 𝑢𝑛 𝜙1 are linearly independent. Let
𝑐1 , 𝑐2 , ⋯ , 𝑐𝑛 be constants such that
𝑐1 𝜙1 + 𝑐2 𝑢2 𝜙1 + ⋯ + 𝑐𝑛 𝑢𝑛 𝜙1 = 0. (9)
Since 𝜙1 ≠ 0, ⟹ 𝑐1 + 𝑐2 𝑢2 + ⋯ + 𝑐𝑛 𝑢𝑛 = 0. (10)
Differentiating (10) with respect to 𝑥, we get
𝑐2 𝑢′2 + 𝑐3 𝑢′3 + ⋯ + 𝑐𝑛 𝑢′𝑛 = 0.
In view of equation (7) this is equivalent to
𝑐2 𝑣2 + 𝑐3 𝑣3 + ⋯ + 𝑐𝑛 𝑣𝑛 = 0.
Since 𝑣2 , 𝑣3 , … , 𝑣𝑛 are linearly independent on 𝐼.
⟹ 𝑐2 = 𝑐2 = ⋯ = 𝑐𝑛 = 0. Hence from equation (10) we have 𝑐1 = 0.
This shows that 𝜙1 , 𝑢2 𝜙1 , … , 𝑢𝑛 𝜙1 are linearly independent, hence is a basis for the solution
of 𝐿(𝑦) = 0 on 𝐼. This proves the theorem.
Our aim is to find another linearly independent solution of second order differential
equation and the formula is derived in the following theorem.
Theorem 2: If 𝜙1 is a solution of 𝐿(𝑦) = 𝑦′′ + 𝑎1 (𝑥)𝑦′ + 𝑎2 (𝑥)𝑦 = 0 on an interval 𝐼 and if
𝜙1 (𝑥) ≠ 0 on 𝐼, a second solution 𝜙2 (𝑥) of 𝐿(𝑦) = 0 is given by
𝑠
𝑥 exp[− ∫𝑥0 𝑎1 (𝑡)𝑑𝑡]
𝜙2 (𝑥) = 𝜙1 (𝑥) ∫𝑥 𝑑𝑠.
0 [𝜙1 (𝑠)]2
The functions 𝜙1 , 𝜙2 form a basis for the solutions of 𝐿(𝑦) = 0 on 𝐼.
Proof: Let 𝜙1 (𝑥) be a solution of 𝐿(𝑦) = 0 on 𝐼 and 𝜙1 (𝑥) ≠ 0, ∀ 𝑥 ∈ 𝐼.
⟹ 𝜙1′′ (𝑥) + 𝑎1 (𝑥)𝜙1′ (𝑥) + 𝑎2 (𝑥)𝜙1 = 0. (11)
Now we try to find the solution of 𝐿(𝑦) = 0 of the form 𝜙 = 𝑢𝜙1 , where 𝑢 is some
function. If 𝜙 = 𝑢𝜙1 is to be a solution of 𝐿(𝑦) = 0, we must of have 𝐿(𝑢𝜙1 ) = 0.
(𝑢𝜙1 )′′ + 𝑎1 (𝑥)(𝑢𝜙1 )′ + 𝑎2 (𝑥)(𝑢𝜙1 ) = 0
⟹ (𝑢′𝜙1 + 𝑢𝜙1′ )′ + 𝑎1 (𝑥)(𝑢′𝜙1 + 𝑢𝜙1′ ) + 𝑎2 (𝑥)(𝑢𝜙1 ) = 0

S25013: Ordinary Differential Equations Page 104


⟹ (𝑢′′𝜙1 + 2𝑢′𝜙1′ + 𝑢𝜙1′′ ) + 𝑎1 (𝑥)(𝑢′𝜙1 + 𝑢𝜙1′ ) + 𝑎2 (𝑥)(𝑢𝜙1 ) = 0
⟹ 𝑢′′𝜙1 + 𝑢′ (2𝜙 ′ + 𝑎1 (𝑥)𝜙1 ) + 𝑢(𝜙1′′ + 𝑎1 (𝑥)𝜙 ′1 + 𝑎2 (𝑥)𝜙1 ) = 0.
By virtue of equation (11) it is equivalent to
𝑢′′𝜙1 + 𝑢′(2𝜙1′ + 𝑎1 (𝑥)𝜙1 ) = 0. (12)
Let 𝑣 = 𝑢′. Then the equation reduces to
𝑣′𝜙1 + (2𝜙′ + 𝑎1 (𝑥)𝜙1 )𝑣 = 0, (13)
which a linear equation of order one.
We have reduced the order of the equation by one.
𝑣′𝜙1 + 2𝜙′1 𝑣 + 𝑎1 (𝑥)𝜙1 𝑣 = 0.
Multiplying this by 𝜙1 , we get
𝑣′𝜙12 + 2𝜙1 𝜙1′ 𝑣 + 𝑎1 (𝑥)𝜙12 𝑣 = 0
⟹ (𝑣𝜙12 )′ + 𝑎1 (𝑥)(𝑣𝜙12 ) = 0

(𝑣𝜙12 )
⟹ = −𝑎1 (𝑥)
(𝑣𝜙12 )
𝑑(𝑣𝜙12 )
or = −𝑎1 (𝑥)𝑑𝑥.
(𝑣𝜙12 )
Integrating this, we get
𝑥
log(𝑣𝜙12 ) = − ∫𝑥 𝑎1 (𝑡)𝑑𝑡 + log𝑐,
0

where x0 is a fixed point in 𝐼, and 𝑐 is constant.


𝑥
− ∫𝑥 𝑎1 (𝑡)𝑑𝑡
⟹ 𝑣𝜙12 = 𝑐𝑒 0

𝑐 𝑥
or 𝑣(𝑥) = 𝜙2 exp [− ∫𝑥 𝑎1 (𝑡)𝑑𝑡].
1 0

Since any constant multiple of a solutions of (13) is again a solution, therefore we see that
1 𝑥
𝑣(𝑥) = 𝜙2(𝑥) exp [− ∫𝑥 𝑎1 (𝑡)𝑑𝑡]
1 0
is a solution of (13).
1 𝑥
⟹ 𝑢′(𝑥) = 𝑣 = 𝜙2 exp [− ∫𝑥 𝑎1 (𝑡)𝑑𝑡]
1 0
Integrating, we get
𝑠
𝑥 exp[− ∫𝑥0 𝑎1 (𝑡)𝑑𝑡]
𝑢(𝑥) = ∫𝑥0 𝑑𝑠. (14)
[𝜙1 (𝑠)]2
We write 𝜙2 (𝑥) = 𝑢(𝑥)𝜙1 (𝑥). (15)
As 𝐿(𝑢(𝑥)𝜙1 (𝑥)) = 0, so 𝜙2 (𝑥) is a solution of 𝐿(𝑦) = 0 on 𝐼.
Now we show that 𝜙1 , 𝜙2 are linearly independent. Consider therefore for constants 𝑐1 and
𝑐2 such that
𝑐1 𝜙1 + 𝑐2 𝜙2 = 0
⟹ 𝑐1 𝜙1 (𝑥) + 𝑐2 𝑢(𝑥)𝜙1 (𝑥) = 0
⟹ 𝑐1 + 𝑐2 𝑢(𝑥) = 0 (∵ 𝜙1 (𝑥) ≠ 0, ∀ 𝑥 ∈ 𝐼). (16)
Differentiating with respect to 𝑥, we get
𝑐2 𝑢′(𝑥) = 0
⟹ 𝑐2 𝑣(𝑥) = 0
⟹ 𝑐2 = 0 as 𝑣(𝑥) ≠ 0 on 𝐼.
From equation (16), we have 𝑐1 = 0. Thus we have proved that

S25013: Ordinary Differential Equations Page 105


𝑐1 𝜙1 + 𝑐2 𝜙2 = 0 ⟹ 𝑐1 = 0, 𝑐2 = 0.
Consequently, we have 𝜙1 and 𝜙2 are linearly independent on 𝐼.
Thus from equation (14) and (15) we have conclude that the second solution 𝜙2 of 𝐿(𝑦) =
0 is given by
𝑠
𝑥 exp[− ∫𝑥0 𝑎1 (𝑡)𝑑𝑡]
𝜙2 (𝑥) = 𝜙1 (𝑥) ∫𝑥 𝑑𝑠,
0 [𝜙1 (𝑠)]2
and the function 𝜙1 and 𝜙2 from a basis for the solution of 𝐿(𝑦) = 0 on 𝐼.

03-03: THE NON-HOMOGENEOUS EQUATION


We first prove the case for 𝑛 = 2 and then the general case.
Theorem 3: Consider the equation
𝐿(𝑦) = 𝑦′′ + 𝑎1 (𝑥)𝑦′ + 𝑎2 (𝑥)𝑦 = 𝑏(𝑥),
where 𝑎1 (𝑥), 𝑎2 (𝑥) and 𝑏(𝑥) are continuous function on an interval 𝐼. Then every solution
𝜓 of 𝐿(𝑦) = 𝑏(𝑥) on 𝐼 can be written as
𝜓 = 𝜓𝑝 + 𝑐1 𝜙1 + 𝑐2 𝜙2 ,
where 𝜓𝑝 is a particular solution of 𝐿(𝑦) = 𝑏(𝑥); 𝜙1 , 𝜙2 are two linearly independent
solutions of 𝐿(𝑦) = 0 and 𝑐1 , 𝑐2 are constants. A particular solution 𝜓𝑝 is given by
𝑥 [𝜙1 (𝑡)𝜙2 (𝑥)−𝜙1 (𝑥)𝜙2 (𝑡)]
𝜓𝑝 = ∫𝑥 𝑏(𝑡)𝑑𝑡.
0 𝑊(𝜙1 ,𝜙2 )(𝑡)
Proof: Consider 𝐿(𝑦) = 𝑦′′ + 𝑎1 (𝑥)𝑦′ + 𝑎2 (𝑥)𝑦 = 𝑏(𝑥), (17)
where 𝑎1 (𝑥), 𝑎2 (𝑥), 𝑏(𝑥) are continuous functions on 𝐼 . We know that every solution of
𝐿(𝑦) = 0 is of the form 𝑐1 𝜙1 + 𝑐2 𝜙2 , where 𝑐1 and 𝑐2 are constants and 𝜙1 , 𝜙2 are linearly
independent solutions of 𝐿(𝑦) = 0. Obviously, such a function 𝑐1 𝜙1 + 𝑐2 𝜙2 can not be a
solution of 𝐿(𝑦) = 𝑏(𝑥) unless 𝑏(𝑥) = 0 on 𝐼.
Suppose we allow 𝑐1 , 𝑐2 to become functions, say 𝑢1 , 𝑢2 on 𝐼 and ask whether there is
solution of 𝐿(𝑦) = 𝑏(𝑥) of the form𝑐1 𝜙1 + 𝑐2 𝜙2 on 𝐼. The answer is yes, and the procedure
is known as the variation of constant.
If 𝑢1 𝜙2 + 𝑢2 𝜙2 is a solution of the equation 𝐿(𝑦) = 𝑏(𝑥), where 𝑢1 and 𝑢2 are functions of
𝑥, then the equation (17) becomes
(𝑢1 𝜙1 + 𝑢2 𝜙2 )′′ + 𝑎1 (𝑥)(𝑢1 𝜙1 + 𝑢2 𝜙2 )′ + 𝑎2 (𝑢1 𝜙1 + 𝑢2 𝜙2 ) = 𝑏(𝑥)
⟹ (𝑢′1 𝜙1 + 𝑢1 ϕ1′ + 𝑢′ 2 𝜙2 + 𝑢2 ϕ′2 )′ + 𝑎1 (𝑥)(𝑢′1 𝜙1 + 𝑢1 ϕ1′ + 𝑢′ 2 𝜙2 + 𝑢2 ϕ′2 )
+𝑎2 (𝑢1 𝜙1 + 𝑢2 𝜙2 ) = 𝑏(𝑥)
⟹ u1′′ 𝜙1 + 2u1′ ϕ1′ + 𝑢1 ϕ′′ ′′ ′ ′ ′
2 + u2 𝜙2 + u2 ϕ2 + 𝜙2 + 𝑢2 ϕ2
′′

+𝑎1 (𝑥)(u1′ 𝜙1 + 𝑢1 ϕ1′ + 𝑢2′ 𝜙2 + 𝑢2 ϕ′2 ) +𝑎2 (𝑢1 𝜙1 + 𝑢2 𝜙2 ) = 𝑏(𝑥)


⟹ (𝜙1 u1′′ + 𝜙2 𝑢2′′ ) + 2(ϕ1′ 𝑢1′ + 𝑢2′ ϕ′2 ) + 𝑢1 (ϕ1′′ + 𝑎1 (𝑥)ϕ1′ + 𝑎2 (𝑥)𝜙1 ) + 𝑢2 (𝜙2′′
+𝑎1 (𝑥)𝜙2′ + 𝑎2 (𝑥)𝜙2 ) + 𝑎1 (𝑥)(𝑢1′ 𝜙1 + 𝑢2′ 𝜙2 ) = 𝑏(𝑥)
⟹ (𝜙1 u1′′ + 𝜙2 𝑢2′′ ) + 2(ϕ1′ 𝑢1′ + 𝑢2′ ϕ′2 ) + 𝑎1 (𝑥)(𝑢1′ 𝜙1 + 𝑢2′ 𝜙2 ) + 𝑢1 L(𝜙1 ) + 𝑢2 L(𝜙2 ) = 𝑏(𝑥).
Since 𝐿(𝜙1 ) = 𝐿(𝜙2 ) = 0 as 𝜙1 , 𝜙2 are independent solutions L(y) = 0, we get therefore
(𝜙1 u1′′ + 𝜙2 𝑢2′′ ) + 2(ϕ1′ 𝑢1′ + 𝑢2′ ϕ′2 ) + 𝑎1 (𝑥)(𝑢1′ 𝜙1 + 𝑢2′ 𝜙2 ) = 𝑏(𝑥). (18)
If we choose 𝑢1 𝜙1 + 𝑢2 𝜙2 = 0,
′ ′
(19)
then (𝑢1 𝜙1 + 𝑢2 𝜙2 )′ = u1 𝜙1 + 𝑢1 𝜙1 + 𝑢2 𝜙2 + 𝑢2 𝜙2 = 0
′ ′ ′′ ′ ′ ′′ ′ ′

S25013: Ordinary Differential Equations Page 106


⟹ (𝑢1′ 𝜙1 + 𝑢2′ 𝜙2 )′ = (𝑢1′ 𝜙1′ + 𝑢2′ 𝜙2′ ) + (𝜙1 u1′′ + 𝜙2 𝑢2′′ ) = 0
Hence, we have from (18)
𝑢1′ 𝜙1′ + 𝑢2′ 𝜙2′ = 𝑏(𝑥). (20)
Thus, we see that, if we can find two functions 𝑢1 , 𝑢2 satisfying (19) and (20) then indeed
𝑢1 𝜙1 + 𝑢2 𝜙2 will also satisfy 𝐿(𝑦) = 𝑏(𝑥). Then solving (19) and (20) for 𝑢1′ and 𝑢2′ , we
get
𝑢1′ 𝑢2′ 1
0 𝜙2 = 𝜙1 0 = 𝜙1 𝜙2
| | | ′ | | ′ |
𝑏 𝜙2′ 𝜙1 𝑏 𝜙1 𝜙2′
0 𝜙2
𝑏(𝑥)| |
1 𝜙2′ −𝑏(𝑥)𝜙 (𝑥)
𝑢1′ = ⟹ 𝑢′1 (𝑥) = 𝑊(𝜙 ,𝜙 2)(𝑥) (21)
𝑊(𝜙1 ,𝜙2 )(𝑥) 1 2
and
𝜙1 0
𝑏(𝑥)| ′ |
𝜙2 1 𝑏(𝑥)𝜙1 (𝑥)
𝑢2′ = ⟹ 𝑢2′ (𝑥) = 𝑊(𝜙 , (22)
𝑊(𝜙1 ,𝜙2 )(𝑥) 1 ,𝜙2 )(𝑥)
where 𝑊(𝜙1 , 𝜙2 )(𝑥) ≠ 0.
Integrating equations (21) and (22), we get
𝑥 𝜙2 (𝑡)𝑏(𝑡) 𝑥 𝜙1 (𝑡)𝑏(𝑡)
𝑢1 (𝑥) = − ∫𝑥 𝑑𝑡, 𝑢2 (𝑥) = ∫𝑥 𝑑𝑡.
0 𝑊(𝜙1 ,𝜙2 )(𝑡) 0 𝑊(𝜙1 ,𝜙2 )(𝑡)

Then the particular solution 𝜓𝑝 = 𝑢1 𝜙1 + 𝑢2 𝜙2 takes the form


𝑥 [𝜙1 (𝑡)𝜙2 (𝑥)−𝜙1 (𝑥)𝜙2 (𝑡)]
𝜓𝑝 (x) = ∫𝑥 𝑏(𝑡)𝑑𝑡.
0 𝑊(𝜙1 ,𝜙2 )(𝑡)
This completes the proof.
Now prove the general case.
Theorem 4: Consider the equation
𝐿(𝑦) = 𝑦 (𝑛) + 𝑎1 (𝑥)𝑦 (𝑛−1) + 𝑎2 (𝑥)𝑦 (𝑛−2) + ⋯ + 𝑎𝑛 (𝑥)𝑦 = 𝑏(𝑥),
where 𝑏(𝑥) is a continuous function on an interval 𝐼. Let 𝜙1 , 𝜙2 , … , 𝜙𝑛 be a basis for the
solution of 𝐿(𝑦) = 0 on 𝐼. Then every solution 𝜓 of 𝐿(𝑦) = 𝑏(𝑥) on 𝐼 can be written as
𝜓 = 𝜓𝑝 + 𝑐1 𝜙1 + 𝑐2 𝜙2 + ⋯ + 𝑐𝑛 𝜙𝑛 ,
where 𝜓𝑝 is a particular solution of 𝐿(𝑦) = 𝑏(𝑥); 𝑐1 , 𝑐2 , … , 𝑐𝑛 are constants. A particular
solution 𝜓𝑝 is given by
𝑥 𝑊𝑘 (𝜙1 ,𝜙2 ,⋯,ϕn )(𝑡)
𝜓𝑝 (𝑥) = ∑𝑛𝑘=1 𝜙𝑘 (𝑥) ∫𝑥 𝑏(𝑡)𝑑𝑡,
0 𝑊(𝜙1 ,𝜙2 ,⋯,𝜙𝑛 )(𝑡)

where 𝑊𝑘 is the determinant obtained from 𝑊 by replacing the 𝑘 th column


(𝑛−1)
(𝜙𝑘 , 𝜙𝑘′ , … , 𝜙𝑘 ) by (0,0, ⋯ ,0,1).
Proof: Let 𝐿(𝑦) = 𝑦 (𝑛) + 𝑎1 (𝑥)𝑦 (𝑛−1) + 𝑎2 (𝑥)𝑦 (𝑛−2) + ⋯ + 𝑎𝑛 (𝑥)𝑦 = 𝑏(𝑥), (23)
where 𝑎1 (𝑥), 𝑎2 (𝑥), … , 𝑎𝑛 (𝑥) and 𝑏(𝑥) are continuous of functions of 𝑥 on 𝐼. Set 𝜓 be any
solution of 𝐿(𝑦) = 𝑏(𝑥) and 𝜓𝑝 is a particular solution of 𝐿(𝑦) = 𝑏(𝑥).
Consider
𝐿(𝜓 − 𝜓𝑝 ) = 𝐿(𝜓) − 𝐿(𝜓𝑝 )
= 𝑏(𝑥) − 𝑏(𝑥)
𝐿(𝜓 − 𝜓𝑝 ) = 0,
which implies 𝜓 − 𝜓𝑝 is solution of 𝐿(𝑦) = 0.

S25013: Ordinary Differential Equations Page 107


Since 𝜙1 , 𝜙2 , … , 𝜙𝑛 are linearly independent solution of 𝐿(𝑦) = 0 . If 𝑐1 , 𝑐2 , … , 𝑐𝑛 are 𝑛
constants, then any solution of 𝐿(𝑦) = 0 can be written as 𝑐1 𝜙1 + 𝑐2 𝜙2 + ⋯ + 𝑐𝑛 𝜙𝑛 .
⟹ 𝜓 − 𝜓𝑛 = 𝑐1 𝜙1 + 𝑐2 𝜙2 + ⋯ + 𝑐𝑛 𝜙𝑛
⟹ 𝜓 = 𝜓𝑝 + 𝑐1 𝜙1 + 𝑐2 𝜙2 + ⋯ + 𝑐𝑛 𝜙𝑛 , (24)
where 𝜓𝑝 is a particular solution and is to be determined. To find the particular solution
𝜓𝑝 , we use the variation of constant method. We try to find 𝑛 functions 𝑢1 , 𝑢2 , … , 𝑢𝑛 so that
𝜓𝑝 = ∑𝑛𝑘=1 𝜙𝑘 𝑢𝑘 = 𝑢1 𝜙1 + 𝑢2 𝜙2 + ⋯ + 𝑢𝑛 𝜙𝑛 (25)
is a solution of 𝐿(𝑦) = 𝑏(𝑥), i.e. we claim that 𝜓𝑝 is a solution of 𝐿(𝑦) = 𝑏(𝑥).
That is we prove that
(𝑛) (𝑛−1)
𝜓𝑝 (𝑥) + 𝑎1 (𝑥)𝜓𝑝 (𝑥) + 𝑎2 (𝑥)𝜓𝑝(𝑛−2) (𝑥) + ⋯ + 𝑎𝑛 (𝑥)𝜓𝑝 (𝑥) = 𝑏(𝑥). (26)
Differentiating equation (25) with respect to 𝑥 we get
𝜓p′ = 𝑢1′ 𝜙1 + 𝑢2′ 𝜙2 + ⋯ + 𝑢𝑛′ 𝜙𝑛 + 𝑢1 𝜙1′ + 𝑢2 𝜙2′ + ⋯ + 𝑢𝑛 𝜙𝑛′
If we choose 𝑢1 , 𝑢2 , … , 𝑢𝑛 such that
𝑢1′ 𝜙1 + 𝑢2′ 𝜙2 + ⋯ + 𝑢𝑛′ 𝜙𝑛 = 0, (27)
then we have
𝜓p′ = 𝑢1 𝜙′1 + 𝑢2 𝜙′2 + ⋯ + 𝑢𝑛 𝜙′𝑛 (28)
⟹ 𝜓p′′ = 𝑢1′ 𝜙1′ + 𝑢2′ 𝜙2′ + ⋯+ 𝑢n′ 𝜙𝑛′ + 𝑢1 𝜙1′′ + 𝑢2 𝜙2′′ + ⋯+ 𝑢𝑛 𝜙𝑛′′
Further if we choose
𝑢1′ 𝜙1′ + 𝑢2′ 𝜙2′ + ⋯ + 𝑢n′ 𝜙𝑛′ = 0 (29)
⟹ 𝜓p′′ = 𝑢1 𝜙1′′ + 𝑢2 𝜙2′′ + ⋯ + 𝑢𝑛 𝜙𝑛′′ . (30)
We obtain
(𝑛) (𝑛−1) (𝑛−1) (𝑛−1) (𝑛) (𝑛) (𝑛)
𝜓𝑝 = 𝑢1′ 𝜙1 + 𝑢2′ 𝜙2 + ⋯ + 𝑢𝑛′ 𝜙𝑛 + 𝑢1 𝜙1 + 𝑢2 𝜙2 + ⋯ + 𝑢𝑛 𝜙𝑛 . (31)
Continuing in this way, in general, if we choose
(𝑛−1) (𝑛−1) (𝑛−1)
𝑢1′ 𝜙1 + 𝑢2′ 𝜙2 + ⋯ + 𝑢𝑛′ 𝜙𝑛 = 𝑏(𝑥), (32)
(𝑛) (𝑛) (𝑛) (𝑛)
then we have 𝜓𝑝 = 𝑢1 𝜙1 + 𝑢2 𝜙2 +⋯+ 𝑢𝑛 𝜙𝑛 + 𝑏(𝑥). (33)
Thus if 𝑢1′ , 𝑢2′ , … , 𝑢𝑛′ satisfy
𝑢1′ 𝜙1 + 𝑢2′ 𝜙2 + ⋯ + 𝑢𝑛′ 𝜙𝑛 = 0,
𝑢1′ 𝜙1′ + 𝑢2′ 𝜙2′ + ⋯ + 𝑢𝑛′ 𝜙𝑛′ = 0
… (34)
(𝑛−2) (𝑛−2) (𝑛−2)
𝑢1′ 𝜙1 + 𝑢2′ 𝜙2 + ⋯ + 𝑢𝑛′ 𝜙𝑛 =0
(𝑛−1) (𝑛−1) (𝑛−1)
𝑢1′ 𝜙1 + 𝑢2′ 𝜙2 + ⋯ + 𝑢𝑛′ 𝜙𝑛 = 𝑏(𝑥),
then we have
𝜓𝑝 = 𝑢1 𝜙1 + 𝑢2 𝜙2 + ⋯ + 𝑢𝑛 𝜙𝑛 (35)
𝜓𝑝′ = 𝑢1 𝜙1′ + 𝑢2 𝜙2′ + ⋯ + 𝑢𝑛 𝜙𝑛′ , (36)
𝜓𝑝′′ = 𝑢1 𝜙1′′ + 𝑢2 𝜙2′′ + ⋯ + 𝑢𝑛 𝜙𝑛′′ , (37)

(𝑛−1) (𝑛−1) (𝑛−1) (𝑛−1)
𝜓𝑝 = 𝑢1 𝜙1 + 𝑢2 𝜙2 + ⋯ + 𝑢𝑛 𝜙𝑛 , (38)
(𝑛) (𝑛) (𝑛) (𝑛)
𝜓𝑝 = 𝑢1 𝜙1 + 𝑢2 𝜙2 + ⋯ + 𝑢𝑛 𝜙𝑛 + 𝑏(𝑥). (39)

S25013: Ordinary Differential Equations Page 108


Multiplying equation (35) by 𝑎𝑛 (𝑥), (36) by 𝑎𝑛−1 (𝑥), ⋯ , (38) by 𝑎1 (𝑥) and (39) by 1 and
adding, we get
𝑢1 𝐿(𝜙1 ) + 𝑢2 𝐿(𝜙2 ) + ⋯ + 𝑢𝑛 𝐿(𝜙𝑛 ) + 𝑏(𝑥) = 𝑏(𝑥).
Since 𝜙1 , 𝜙2 , … , 𝜙𝑛 are independent solution of 𝐿(𝑦) = 0. ⟹ 𝐿(𝜙𝑖 ) = 0 ∀ 𝑖 = 1,2, … , 𝑛.
Thus we have proved that 𝜓𝑝 = 𝑢1 𝜙1 + 𝑢2 𝜙2 + ⋯ + 𝑢𝑛 𝜙𝑛 is a solution of 𝐿(𝑦) = 𝑏(𝑥).
Now only the problems left is to find the functions 𝑢1 , 𝑢2 , … , 𝑢𝑛 . To find 𝑢1 , 𝑢2 , … , 𝑢𝑛 , we
solve the system of equations (34) for 𝑢1′ , 𝑢2′ , … , 𝑢𝑛′ .
Equation (34) is the system of 𝑛 linear equation in 𝑛-unknowns 𝑢1′ , 𝑢2′ , … , 𝑢𝑛′ . We see that
the determinant of the coefficients is just Wronskian 𝑊(𝜙1 , 𝜙2 , … , 𝜙𝑛 )(𝑥) which is never
zero ∀ 𝑥 ∈ 𝐼 as 𝜙1 , 𝜙2 , … , 𝜙𝑛 are linearly independent solutions of 𝐿(𝑦) = 0. Hence there
are unique solutions 𝑢′1 , 𝑢′2 , … , 𝑢′𝑛 satisfying equations (34). We know those solutions are
given by Cramer’s rule.
Then we have
𝛥 (𝑥)
𝑢𝑘′ (𝑥) = 𝑊(𝜙 ,𝜙𝑘,⋯,𝜙 )(𝑥) , 𝑘 = 1,2, … , 𝑛, (40)
1 2 𝑛
𝜙1 𝜙2 … 𝜙𝑘 … 𝜙𝑛
𝜙′ 𝜙2′ … 𝜙𝑘′ … 𝜙𝑛′
| 1 |
where 𝑊(𝜙1 , 𝜙2 , … , 𝜙𝑛 )(𝑥) = … … … … … … (41)
| (𝑛−1) (𝑛−1) (𝑛−1) (𝑛−1) |
𝜙1 𝜙2 … 𝜙𝑘 … 𝜙𝑛

and
𝜙1 𝜙2 … 0 … 𝜙𝑛
𝜙1′ 𝜙2′ … 0 … 𝜙𝑛′
| |
𝛥𝑘 (𝑥) = … … … … … …
| (𝑛−1) (𝑛−1) (𝑛−1) |
𝜙1 𝜙2 … 𝑏 … 𝜙𝑛

is determinant obtained from 𝑊 by replacing the 𝑘 th column of 𝑊 by 0,0, ⋯ ,0, 𝑏. We write


this as
𝜙1 𝜙2 … 0 … 𝜙𝑛
′ ′
𝜙 𝜙2 … 0 … 𝜙𝑛′
|…1 |
𝛥𝑘 (𝑥) = 𝑏(𝑥) … … … … … = 𝑏(𝑥)𝑊(𝑥).
| (𝑛−1) (𝑛−1) (𝑛−1) |
𝜙1 𝜙2 … 1 … 𝜙𝑛

⟹ 𝛥𝑘 (𝑥) = 𝑏(𝑥)𝑊𝑘 (𝜙1 , 𝜙2 , ⋯ , 𝜙𝑛 )(𝑥), (42)


where 𝑊𝑘 is the determinant obtained from the Wronskian by replacing the 𝑘 column by
th

(0,0, … ,0,1). Substituting this in the equations (40) we get


𝑊𝑘 (𝜙1 ,𝜙2 ,⋯,𝜙𝑛 )(𝑥)𝑏(𝑥)
𝑢𝑘′ (𝑥) = , 𝑘 = 1,2, ⋯ , 𝑛. (43)
𝑊(𝜙1 ,𝜙2 ,⋯,𝜙𝑛 )(𝑥)
Integrating (43) between the limits 𝑥0 to 𝑥 we get
𝑥 𝑊𝑘 (𝑡)𝑏(𝑡)
𝑢𝑘 (𝑥) = ∫𝑥 𝑑𝑡; 𝑘 = 1,2, … , 𝑛. (44)
0 𝑊(𝜙1 ,𝜙2 ,⋯,𝜙𝑛 )(𝑡)
Substituting the values of 𝑢𝑘 (𝑥) ∀ 𝑘, in the equation (25), we get
𝑥 𝑊𝑘 (𝑡)𝑏(𝑡)
𝜓𝑝 = ∑𝑛𝑘=1 𝜙𝑘 (𝑥) ∫𝑥 𝑑𝑡. (45)
0 𝑊(𝜙1 ,𝜙2 ,⋯,𝜙𝑛 )(𝑡)
This completes the proof.

S25013: Ordinary Differential Equations Page 109


SOLVED PROBLEMS 01
Problem 02-03-01:
One solution of 𝑥 2 𝑦′′ − 7𝑥𝑦′ + 15𝑦 = 0, 𝑥 > 0 is 𝜙1 (𝑥) = 𝑥 3 . Find a second solution.
7 15
Solution: The given equation can be written as 𝑦′′ − 𝑥 𝑦′ + 𝑥 2 𝑦 = 0, (46)
7
where 𝑎1 (𝑥) = − 𝑥.
𝑠 𝑠 1
∴ − ∫1 𝑎1 (𝑡)𝑑𝑡 = 7 ∫1 𝑑𝑡 = 7(log𝑡)1𝑠
𝑡
𝑠
⟹ − ∫1 𝑎1 (𝑡)𝑑𝑡 = 7log𝑠
𝑠
⟹ exp[− ∫1 𝑎1 (𝑡)𝑑𝑡] = 𝑒 7log𝑠 = 𝑠 7 .
Since 𝜙1 (𝑥) = 𝑥 3 is one solution of (46). Then the second solution is given by
𝜙2 (𝑥) = 𝑢𝜙1 (𝑥), (47)
where
𝑠 𝑥
𝑥 𝑒 − ∫1 𝑎1 (𝑡)𝑑𝑡 𝑥 𝑠7 𝑥 𝑠2 𝑥 2 −1
𝑢(𝑥) = ∫1 [𝜙 (𝑠)]2 𝑑𝑠 = ∫1 𝑠6 𝑑𝑠 = ∫1 𝑠𝑑𝑠 = ( 2 ) = ( ).
1 1 2
1 1
⟹ 𝜙2 (𝑥) = (𝑥 − 1)𝑥 ⟹ 𝜙2 = (𝑥 5 − 𝑥 3 ).
2 3
2 2
We know that 𝜙(𝑥) = 𝑐1 𝜙1 + 𝑐2 𝜙2 is also a solution of (46), therefore we have
𝜙(𝑥) = 𝑐1 𝜙1 + 𝑐2 𝜙2
1
= 𝑐1 𝑥 3 + 𝑐2 2 (𝑥 5 − 𝑥 3 )
1 1
= 𝑐1 𝑥 3 − 𝑐2 2 𝑥 3 + 𝑐2 2 𝑥 5
1 1
= (𝑐1 − 𝑐2 2) 𝑥 3 + (𝑐2 2) 𝑥 5
= 𝑑1 𝑥 3 + 𝑑2 𝑥 5 .
Hence the solution of the equation (46) is 𝜙2 (𝑥) = 𝑥 5 .
Now to show that 𝜙1 , 𝜙2 are linearly independent. We show that their Wronskian is
different from zero.
𝜙 𝜙2
∴ 𝑊(𝜙1 , 𝜙2 )(𝑥) = | 1′ |
𝜙1 𝜙2′
3 5
= |𝑥 2 𝑥 4 |
3𝑥 5𝑥
= 5𝑥 7 − 3𝑥 7 = 2𝑥 7 ≠ 0 for 𝑥 > 0.
⟹ 𝜙1 , 𝜙2 are linearly independent.
Problem 02-03-02:
One solution of 𝑥𝑦 ′′ − (𝑥 + 1)𝑦 ′ + 𝑦 = 0, 𝑥 > 0 is 𝜙1 (𝑥) = 𝑒 𝑥 . Find another independent
solution of the equation.
Solution: The given equation can be written as
𝑥+1 1
𝐿(𝑦) = 𝑦 ′′ − ( ) 𝑦 ′ + 𝑥 𝑦 = 0, 𝑥 > 0, (48)
𝑥
𝑥+1 1
where 𝑎1 (𝑥) = − ( ) = − (1 + ).
𝑥 𝑥
If 𝜙2 (𝑥) is another solution of (48), then it is given by
𝑠
𝑥 𝑒 − ∫1 𝑎1 (𝑡)𝑑𝑡
𝜙2 (𝑥) = 𝜙1 (𝑥) ∫1 𝑑𝑠
[𝜙1 (𝑠)]2

S25013: Ordinary Differential Equations Page 110


𝑠 1
(1+ )𝑑𝑡
𝑥 𝑥 𝑒 ∫1 𝑥
= 𝑒 ∫1 2𝑠
𝑑𝑠
𝑒
𝑥 𝑒 𝑠−1 𝑒 log𝑠
= 𝑒 𝑥 ∫1 𝑑𝑠
𝑒 2𝑠
𝑥
𝜙2 (𝑥) = 𝑒 𝑥 ∫1 𝑠𝑒 −𝑠−1 𝑑𝑠.
Put 𝑢 = −(𝑠 + 1) ⟹ 𝑑𝑢 = −𝑑𝑠. When 𝑠 = 1, 𝑢 = −2 and 𝑠 = 𝑥, u = −(1 + 𝑥). Then
−1−𝑥
∴ 𝜙2 (𝑥) = 𝑒 𝑥 ∫−2 (−1 − 𝑢)𝑒 𝑢 (−𝑑𝑢)
−1−𝑥
= 𝑒 𝑥 ∫−2 (1 + 𝑢)𝑒 𝑢 𝑑𝑢
= 𝑒 𝑥 [(1 + 𝑢)𝑒 𝑢 − 𝑒 𝑢 ]−1−𝑥
−2
𝑥 [𝑢𝑒 𝑢 ]−1−𝑥
=𝑒 −2
= 𝑒 [−(1 + 𝑥)𝑒 −(1+𝑥) − (−2)𝑒 −2 ]
𝑥

= [−(1 + 𝑥)𝑒 −1 + 2𝑒 𝑥−2 ].


Now 𝜙(𝑥) = 𝑐1 𝜙1 + 𝑐2 𝜙2 is also a solution of 𝐿(𝑦) = 0.
𝜙(𝑥) = 𝑐1 𝑒 𝑥 + 𝑐2 [−𝑒 −1 (1 + 𝑥) + 2𝑒 −2 𝑒 𝑥 ]
= (𝑐1 + 2𝑐2 𝑒 −2 )𝑒 𝑥 + (−𝑐2 𝑒 −1 )(1 + 𝑥)
= 𝑑1 𝑒 𝑥 + 𝑑2 (1 + 𝑥) = 𝑑1 𝜙1 (𝑥) + 𝑑2 𝜙2 (𝑥).
⟹ 𝜙1 (𝑥) = 𝑒 𝑥 and 𝜙2 (𝑥) = (1 + 𝑥) are solutions of 𝐿(𝑦) = 0.
To show there are linearly independent, we find
𝜙 𝜙2 𝑒 𝑥 (1 + 𝑥)
𝑊(𝜙1 , 𝜙2 )(𝑥) = | 1′ | = | | = 𝑒 𝑥 − 𝑒 𝑥 (1 + 𝑥) = −𝑒 𝑥 𝑥
𝜙1 𝜙2′ 𝑒𝑥 1
⟹ 𝑊(𝜙1 , 𝜙2 )(𝑥) = −𝑥𝑒 𝑥 ≠ 0, 𝑥 > 0
⟹ 𝜙1 and 𝜙2 are linearly independent solutions of 𝐿(𝑦) = 0.
Problem 02-03-03:
One solution of (1 − 𝑥 2 )𝑦 ′′ − 2𝑥𝑦 ′ + 2𝑦 = 0, 0 < 𝑥 < 1 is 𝜙(𝑥) = 𝑥 , obtain the second
independent solution.
Solution: We write the given differential equation
2𝑥 2
𝐿(𝑦) = 𝑦 ′′ − (1−𝑥 2) 𝑦 ′ + (1−𝑥 2 ) 𝑦 = 0, 0 < 𝑥 < 1, (49)
−2𝑥 2
where 𝑎1 (𝑥) = , 𝑎2 (𝑥) = .
1−𝑥 2 1−𝑥 2
Here 𝜙1 (𝑥) = 𝑥 is one known solution of (49). If 𝜙2 is another solution of 𝐿(𝑦) = 0, then
it is given that
𝑠
−∫ 𝑎 (𝑡)𝑑𝑡
𝑥 𝑒 𝑥0 1
𝜙2 (𝑥) = 𝜙1 (𝑥) ∫𝑥 [𝜙 (𝑠)]2 𝑑𝑠
0 1
𝑠 2𝑡
∫ 𝑑𝑡
𝑥 𝑒 𝑥0 1−𝑡2
= 𝑥 ∫𝑥 𝑑𝑠
0 𝑠2
2 𝑠
𝑥 𝑒 −log(1−𝑡 )|x0
= 𝑥 ∫𝑥 𝑑𝑠
0 𝑠2
𝑥 1−𝑥02 1
= 𝑥 ∫𝑥 ( ) 𝑑𝑠
0 1−𝑠2
𝑠2
𝑥 1 1
= 𝑥(1 − 𝑥02 ) ∫𝑥 [1−𝑠2 + 𝑠2 ] 𝑑𝑠
0
𝑥 1 1 1 𝑥 1
=𝑥(1 − 𝑥0 ) [∫𝑥 2 (1+𝑠 + 1−𝑠) 𝑑𝑠
2
+ ∫𝑥 𝑑𝑠]
0 0 𝑠2

S25013: Ordinary Differential Equations Page 111


1 1 𝑥
= 𝑥(1 − 𝑥02 ) [ (log(1 + 𝑠) − log(1 − 𝑠)) − ]
2 𝑠 𝑥0
1 1+𝑥 1 1 1+𝑥0 1
=𝑥(1 − 𝑥02 ) [2 log (1−𝑥) − 𝑥 − 2 log (1−𝑥 ) + 𝑥 ]
0 0

𝑥 1+𝑥 1 1+𝑥 1
= (1 − 𝑥02 ) [2 log (1−𝑥) − 1 − 𝑥 [2 log (1−𝑥0) − 𝑥 ]].
0 0

𝑥 1+𝑥
⟹ 𝜙2 (𝑥) = 𝑐1 [2 log (1−𝑥) − 1] + 𝑐2 𝑥
1 1+𝑥 1
where 𝑐1 = (1 − 𝑥02 ) and 𝑐2 = − [2 log (1−𝑥0) − 𝑥 ].
0 0

We know if 𝜙1 , 𝜙2 are solution of 𝐿(𝑦) = 0, then 𝜙(𝑥) = 𝑑1 𝜙1 + 𝑑2 𝜙2 is also a solution.


Therefore, we have
𝑥 1+𝑥
𝜙(𝑥) = 𝑑1 𝑥 + 𝑑2 [𝑐1 (2 log (1−𝑥) − 1) + 𝑐2 𝑥]
𝑥 1+𝑥
= 𝑒1 𝑥 + 𝑒2 [2 log (1−𝑥) − 1] , 𝑒1 = 𝑑1 + 𝑑2 𝑐2 and 𝑒2 = 𝑐1 𝑑2
𝑥 1+𝑥
⟹ 𝜙2 (𝑥) = 2 log (1−𝑥) − 1 is another independent solution of 𝐿(𝑦) = 0.
Note, 𝑊(𝜙1 , 𝜙2 )(𝑥) ≠ 0.
Problem 02-03-04:
Suppose 𝜙1 and 𝜙2 are two linearly independent solutions of the folowing equation
𝐿(𝑦) = 𝑦′′′ + 𝑎1 (𝑥)𝑦′′ + 𝑎2 (𝑥)𝑦′ + 𝑎3 (𝑥)𝑦 = 0.
(a) Let 𝜙 = 𝑢𝜙1 and compute the equation of order two satisfied by 𝑢′ in order that 𝐿(𝜙) =
𝜙 ′
0. Show that (𝜙2 ) is a solution of this equation of order two.
1
𝜙 ′
(b) Use the fact that (𝜙2 ) satisfies the equation of order two to reduce the order of this
1

equation by one.
Solution: Let 𝐿(𝑦) = 𝑦′′′ + 𝑎1 (𝑥)𝑦′′ + 𝑎2 (𝑥)𝑦′ + 𝑎3 (𝑥)𝑦 = 0. (50)
Let 𝜙 = 𝑢𝜙1 be a solution of (50). That is 𝐿(𝜙) = 0.
⟹ 𝐿(𝑢𝜙1 ) = 0
⟹ (𝑢𝜙1 )′′′ + 𝑎1 (𝑥)(𝑢𝜙1 )′′ + 𝑎2 (𝑢𝜙1 )′ + 𝑎3 (𝑥)(𝑢𝜙1 ) = 0
⟹ (𝑢′′′𝜙1 + 3𝑢′′𝜙1 + 3𝑢′𝜙1′′ + 𝑢𝜙1′′′ ) + 𝑎1 (𝑥)(𝑢′′𝜙1 + 2𝑢′𝜙1′ + 𝜙1′′ )
+𝑎2 (𝑥) (𝑥)(𝑢′𝜙1 + 𝑢𝜙1′ ) + 𝑎3 (𝑥)𝑢𝜙1 = 0
⟹ 𝑢′′′ 𝜙1 + (3ϕ1′ + 𝑎1 (𝑥)𝜙1 )𝑢′′ + (3𝜙1′′ + 2𝑎1 (𝑥)𝜙1′ + 𝑎2 (𝑥)𝜙1 )𝑢′
+𝑢(𝜙1′′′ + 𝑎1 (𝑥)𝜙1′′ + 𝑎2 (𝑥)𝜙1′ + 𝑎3 (𝑥)𝜙1 ) = 0
⟹ 𝜙1 𝑢′′′ + (3ϕ1′ + 𝑎1 (𝑥)𝜙1 )𝑢′′ + (3𝜙1′′ + 2𝑎1 (𝑥)𝜙1′ + 𝑎2 (𝑥)𝜙1 )𝑢′ = 0
⟹ 𝜙1 (𝑢′ )′′ + (3ϕ1′ + 𝑎1 (𝑥)𝜙1 )(𝑢′ )′ + (3𝜙1′′ + 2𝑎1 (𝑥)𝜙1′ + 𝑎2 (𝑥)𝜙1 )𝑢′ = 0
Thus 𝐿(𝜙) = 0 is an equation of order two in 𝑢′.
Let us denote
𝑣 = 𝑢′ (51)
𝜙1 𝑣′′ + (3𝜙1 + 𝑎1 (𝑥)𝜙1 )𝑣′ + (3𝜙1′′ + 2𝑎1 (𝑥)𝜙1′ + 𝑎2 (𝑥)𝜙1 )𝑣 = 0,

(52)
which is a differentiable equation of order 2.

S25013: Ordinary Differential Equations Page 112


𝜙 ′
Now we claim that ( 2 ) satisfies the equation (52). Then, we simply prove that
𝜙1
𝜙2 ′
𝑣 = (𝜙 ) . Then we write
1
𝜙 ′ 𝜙1 𝜙2′ −𝜙1′ 𝜙2
𝑣 = (𝜙2 ) = ,
1 𝜙12
′′ ′
𝜙 𝜙1 𝜙2′ −𝜙1′ 𝜙2
𝑣 ′ = (𝜙2 ) = ( )
1 𝜙12
′ ′
𝜙2′ 𝜙1′ 𝜙2
= (𝜙 ) − ( )
1 𝜙12

𝜙1 𝜙2′′ −𝜙2′ 𝜙1′ 𝜙12 (𝜙1′ 𝜙2 ) −2𝜙1 𝜙1′ 𝜙1′ 𝜙2
= −
𝜙12 𝜙14
2
𝜙2′′ 𝜙1′ 𝜙2′ 𝜙12 (𝜙1′′ 𝜙2 +𝜙1′ 𝜙2′ )−2𝜙1 𝜙2 𝜙1′
= − −
𝜙1 𝜙12 𝜙14
2
𝜙2′′ 𝜙1′ 𝜙2′ 𝜙1′′ 𝜙2 2𝜙2 𝜙1′
= −2 − + ,
𝜙1 𝜙12 𝜙12 𝜙13
and
𝜙1 𝜙2′′′ −𝜙1′ 𝜙2′′ 𝜙12 (𝜙1′ 𝜙2′′ +𝜙1′′ 𝜙2′ )−2𝜙1 𝜙1′ 𝜙1′ 𝜙2′
𝑣 ′′ = −2
𝜙12 𝜙14
2 2
𝜙12 (𝜙1′′ 𝜙2′ +𝜙1′′′ 𝜙2 )−2𝜙1 𝜙1′ 𝜙1′′ 𝜙2 𝜙13 (𝜙2 2𝜙1′ 𝜙1′′ +𝜙2′ 𝜙1′ )−3𝜙12 𝜙1′ 𝜙2 𝜙1′
− +2
𝜙14 𝜙16
2
𝜙2′′′ 𝜙1′ 𝜙2′′ 𝜙1′ 𝜙2′′ 𝜙1′′ 𝜙2′ 𝜙1′ 𝜙2′ 𝜙1′′ 𝜙2′ 𝜙1′′′ 𝜙2 𝜙1′ 𝜙1′′ 𝜙2
= − −2 −2 +4 − − +2
𝜙1 𝜙12 𝜙12 𝜙12 𝜙13 𝜙12 𝜙12 𝜙13
2 3
𝜙2 𝜙1′ 𝜙1′′ 𝜙2′ 𝜙1′ 𝜙1′ 𝜙2
+4 +2 −6
𝜙13 𝜙13 𝜙14
2 3
𝜙2′′′ 𝜙1′ 𝜙2′′ 𝜙1′′ 𝜙2′ 𝜙1′ 𝜙2′ 𝜙2 𝜙1′ 𝜙1′′ 𝜙1′ 𝜙2 𝜙1′′′ 𝜙2
= −3 −3 +6 +6 −6 −
𝜙1 𝜙12 𝜙12 𝜙13 𝜙13 𝜙14 𝜙12
Substituting the values of 𝑣, 𝑣 ′ , 𝑣′′ in (52), we have
𝜙1 𝑣′′ + (3𝜙1′ + 𝑎1 (𝑥)𝜙1 )𝑣′ + (3𝜙1′′ + 2𝑎1 (𝑥)𝜙1′ + 𝑎2 (𝑥)𝜙1 )𝑣
2 3
𝜙1′ 𝜙2′′ 𝜙1′′ 𝜙2′ 𝜙1′ 𝜙2′ 𝜙2 𝜙1′ 𝜙1′′ 𝜙1′ 𝜙2 𝜙1′′′ 𝜙2
= 𝜙2′′′ − 3 −3 +6 +6 −6 −
𝜙1 𝜙1 𝜙12 𝜙12 𝜙13 𝜙1
2 3
3𝜙1′ 𝜙2′′ 𝜙1′ 𝜙2′ 𝜙2 𝜙1′ 𝜙1′′ 𝜙2 𝜙1′
+ −6 −3 +6
𝜙1 𝜙12 𝜙12 𝜙13
2
𝜙1′ 𝜙2′ 𝜙1′′ 𝜙2 𝜙2 𝜙1′
+𝑎1 (𝑥)𝜙2′′ − 2𝑎1 (𝑥) − 𝑎1 (𝑥) + 2𝑎1 (𝑥)
𝜙1 𝜙1 𝜙12
2
𝜙1′′ 𝜙2′ 𝜙1′ 𝜙1′′ 𝜙2 𝜙1′ 𝜙2′ 𝜙1′ 𝜙2 𝜙1′ 𝜙2
+3 −3 + 2𝑎1 (𝑥) − 2𝑎1 (𝑥) + 𝑎2 (𝑥)𝜙2′ − 𝑎2 (𝑥)
𝜙1 𝜙12 𝜙1 𝜙12 𝜙1
𝜙1′′′ 𝜙2 𝜙1′′ 𝜙2 𝜙1′ 𝜙2
= (𝜙2′′′ + 𝑎1 (𝑥)𝜙2′′ + +𝑎2 (𝑥)𝜙2′ ) − − 𝑎1 (𝑥) − 𝑎2 (𝑥)
𝜙1 𝜙1 𝜙1
𝜙2
= (𝜙2′′′ + 𝑎1 (𝑥)𝜙2′′ + +𝑎2 (𝑥)𝜙2′ ) −𝜙 (𝜙1′′′ + 𝑎1 (𝑥)𝜙1′′ + 𝑎2 (𝑥)𝜙1′ )
1
𝜙
= −𝑎3 (𝑥)𝜙2 − 𝜙2 (−𝑎3 (𝑥)𝜙1 ) (∵ 𝐿(𝜙1 ) = −𝑎3 (𝑥)𝜙1 , 𝐿(𝜙2 ) = −𝑎3 (𝑥)𝜙2 )
1

= −𝑎3 (𝑥)𝜙2 + 𝑎3 (𝑥)𝜙2


= 0.
𝜙 ′
This proves 𝑣 = (𝜙2 ) is a solution of redueced order equation (52).
1

S25013: Ordinary Differential Equations Page 113


(b) Consider the equation
𝜙1 𝑣 ′′ + (3𝜙1′ + 𝑎1 (𝑥)𝜙1 )𝑣 ′ + (3𝜙1′′ + 2𝑎1 (𝑥)𝜙1′ + 𝑎2 (𝑥)𝜙1 )𝑣 = 0
1 1
⟹ 𝑣 ′′ + 𝜙 (3𝜙1′ + 𝑎1 (𝑥)𝜙1 )𝑣′ + 𝜙 (3𝜙1′′ + 2𝑎1 (𝑥)𝜙1′ + 𝑎2 (𝑥)𝜙1 )𝑣 = 0 (53)
1 1

⟹ 𝑣′′ + 𝐴1 (𝑥)𝑣′ + 𝐴2 (𝑥)𝑣 = 0, (54)


1
𝐴1 = 𝜙 (3𝜙 ′1 + 𝑎1 (𝑥)𝜙1 ),
where 1
1
} (55)
𝐴2 = 𝜙 (3𝜙′′1 + 2𝑎1 (𝑥)𝜙′1 + 𝑎2 (𝑥)𝜙1 )
1
𝜙2 ′
Let 𝜙 = (𝜙 ) which satisfy equation (54). Hence we assume 𝜙 is a known solution of
1

equation (54)
𝜙 ′′ + 𝐴1 (𝑥)𝜙′ + 𝐴2 (𝑥)𝜙 = 0. (56)
Hence we know another solution of (54) is given by
𝑣 = 𝑤𝜙. (57)
Hence equation (54) reduces to
(w𝜙)′′ + 𝐴1 (𝑥)(𝑤𝜙)′ + 𝐴2 (𝑥)(𝑤𝜙) = 0
⟹ 𝑤′′𝜙 + 2𝜙′𝑤′ + 𝜙′′𝑤 + 𝐴1 (𝑥)(w′𝜙 + 𝑤𝜙′) + 𝐴2 (𝑥)(𝑤𝜙) = 0
⟹ 𝜙𝑤′′ + (2𝜙 ′ + 𝐴1 (𝑥)𝜙)𝑤′ + (𝜙 ′′ + 𝐴1 (𝑥)𝜙 ′ + 𝐴2 (𝑥)𝜙)𝑤 = 0
⟹ 𝜙𝑤 ′′ + (2𝜙′ + 𝐴1 (𝑥)𝜙)𝑤′ = 0 due to (56).
Let 𝑡 = 𝑤′
⟹ 𝜙𝑡′ + (2𝜙 ′ + 𝐴1 (𝑥)𝜙)𝑡 = 0, (58)
which is the first order differentiable equation.
Problem 02-03-05:
Two solutions of 𝑥 3 𝑦 ′′′ − 3𝑥𝑦 ′ + 3𝑦 = 0, 𝑥 > 0 are 𝜙1 (𝑥) = 𝑥, 𝜙2 (𝑥) = 𝑥 3 . Find the third
independent solution.
3 3
Solution: Let 𝐿(𝑦) = 𝑦 ′′′ − 𝑥 2 𝑦 ′ + 𝑥 3 𝑦 = 0, 𝑥 > 0, (59)
2 3
where 𝑎1 (𝑥) = 0, 𝑎2 (𝑥) = − 𝑥 2 , 𝑎3 (𝑥) = 𝑥 3 .
Given that
𝜙1 (𝑥) = 𝑥, 𝜙2 (𝑥) = 𝑥 3 . (60)
are solutions of 𝐿(𝑦) = 0. To find the third solution of 𝐿(𝑦) = 0, let 𝜙 = 𝑢𝜙1 be the third
solution of 𝐿(𝑦) = 0, where 𝑢 is to be determined.
3 3
⟹ (𝑢𝜙1 )′′′ − 𝑥 2 (𝑢𝜙1 )′ + 𝑥 3 (𝑢𝜙1 ) = 0
⟹ 𝜙1 𝑢′′′ + (3𝜙1′ + 𝑎1 (𝑥)𝜙1 )𝑢′′ + (3𝜙1′′ + 2𝑎1 (𝑥)𝜙1′ + 𝑎2 (𝑥)𝜙1 )𝑢′ + 𝑢𝐿(𝜙1 ) = 0.
−3
Since 𝐿(𝜙1 ) = 0 and 𝑎1 (𝑥) = 0, 𝑎2 (𝑥) = − 𝑥 2 , then
3
𝜙1 𝑢′′′ + 3𝜙1′ 𝑢′′ + (3𝜙1′′ − 𝑥 2 𝜙1 )𝑢′ = 0.
Define a new function 𝑣 = 𝑢′ (61)
3
𝜙1 𝑣′′ + 3𝜙1′ 𝑣′ + (3𝜙1′′ − 𝑥 2 𝜙1 ) 𝑣 = 0 (62)
Since 𝜙1 = 𝑥 ⟹ 𝜙1′ = 1, 𝜙1′′ = 0
3
⟹ 𝑥𝑣′′ + 3𝑣′ − 𝑥 2 (𝑥)𝑣 = 0
3 3
⟹ 𝑣′′ + 𝑥 𝑣′ − 𝑥 2 𝑣 = 0. (63)

S25013: Ordinary Differential Equations Page 114


𝜙 ′
This is a second order differential equations. We know that 𝜙 = ( 2 ) is a solution of
𝜙1
equation (63). We consider 𝜙 is a known solution of (63) and assume that 𝑣 = 𝑤𝜙 be
another solution of (63) where 𝑤 is to be determined.
Here
′ ′
𝜙 𝑥3
𝜙 = (𝜙2 ) = ( 𝑥 ) = 2𝑥 ⟹ 𝜙 = 2𝑥. (64)
1

Substituting v = w𝜙 in(63), we get


3 3
(𝑤𝜙)′′ + 𝑥 (𝑤𝜙)′ − 𝑥 2 (𝑤𝜙) = 0
3 3
⟹ 𝑤′′𝜙 + 2𝑤′𝜙′ + 𝑤𝜙′′ + 𝑥 (𝑤′𝜙 + 𝑤𝜙′) − 𝑥 2 𝑤𝜙 = 0
3 3 3
⟹ 𝜙𝑤′′ + (2𝜙 ′ + 𝑥 𝜙) 𝑤′ + w (𝜙 ′′ + 𝑥 𝜙 ′ − 𝑥 2 𝜙) = 0
3
⟹ 𝜙𝑤′′ + (2𝜙 ′ + 𝑥 𝜙) 𝑤′ = 0 (65)
Since 𝜙 = 2𝑥, 𝜙′ = 2 ⟹ 𝑥𝑤′′ + 5𝑤′ = 0 (66)
Define 𝑡 = 𝑤′ ⟹ 𝑥𝑡′ + 5𝑡 = 0.
This is the first order differential equation
𝑑𝑡 5
𝑡
= − 𝑥 𝑑𝑥
Integrating, we get
𝑐
log 𝑡 = −5 log 𝑥 + log 𝑐 ⟹ 𝑡 = 𝑥 5 , (𝑐 constant) (67)
𝑐 𝑐 𝑐
⟹ 𝑤′ = 𝑥 5 ⟹ 𝑤 = 𝑥14 , (𝑐1 = − 4). (68)
𝑐
Hence the equation 𝑣 = 𝑤𝜙 becomes 𝑣 = 𝑥23, (𝑐2 = 2𝑐1 ). This gives
𝑐
𝑢′ = 𝑣 = 𝑥23 .
𝑐 𝑐 1
Integrating, we get 𝑢 = 𝑥32 , (𝑐3 = − 22 ) . ⟹ 𝑢 = 𝑥 2 is also solution of
3
𝜙1 𝑢′′′ + 3𝜙1′ 𝑢′′ + (3𝜙1′′ − 𝑥 2 𝜙1 )𝑢′ = 0.
Consequently the solution of (59) is given by 𝜙3 = 𝑢𝜙1 .
1
⟹ 𝜙3 = 𝑥 is the third independent solution of (59).
Problem 02-03-06:
2
One solution of 𝑦 ′′ − 𝑥 2 𝑦 = 0 (0 < 𝑥 < ∞) is 𝜙1 (𝑥) = 𝑥 2 . Find all solutions of the
2
equation 𝑦′′ − 𝑥 2 𝑦 = 𝑥.
Solution: Consider the equation
2
𝐿(𝑦) = 𝑦 ′′ − 𝑥 2 𝑦 = 𝑥, 0 < 𝑥 < ∞. (69)
Given that 𝜙1 (𝑥) = 𝑥 2 is one solution of 𝐿(𝑦) = 0. (70)
Let 𝜙(𝑥) be another solution of (70), where 𝜙(𝑥) is given by
𝑠
−∫ 𝑎 (𝑡)𝑑𝑡
𝑥 𝑒 𝑥0 1
𝜙(𝑥) = 𝜙1 (𝑥) ∫𝑥 𝑑𝑠,
0 [𝜙1 (𝑠)]2
where we see from equation (70) that 𝑎1 (𝑥) = 0,
𝑥 1
⟹ 𝜙(𝑥) = 𝜙1 (𝑥) ∫𝑥 2 𝑑𝑠
0 (𝜙 (𝑠)) 1

S25013: Ordinary Differential Equations Page 115


𝑥 1
= 𝑥 2 ∫𝑥 𝑑𝑠
0 𝑠4
1 𝑥
= 𝑥 2 (− 3𝑠3 ) 𝜙2 (𝑥)
𝑥0
𝑥2 1 1
=− (𝑥 2 − 𝑥 3 )
3 0
1 𝑥2
= − 3𝑥 + 3𝑥 2 .
0
1 1 1
This is of the form 𝜙(𝑥) = 𝑐1 𝜙1 + 𝑐2 𝜙2 i.e. where 𝑐2 = − 3 and 𝑐1 = 3𝑥 2 and 𝜙2 (𝑥) = 𝑥 and
0
1
𝜙1 (𝑥) = 𝑥 , ⟹ other
2
independent solution of 𝐿(𝑦) = 0 is 𝜙2 (𝑥) = 𝑥 . Because
1
𝑥2 1
𝑊(𝜙1 , 𝜙2 )(𝑥) = | 𝑥
1| = −1 − 2 = −3 ≠ 0 . Hence 𝜙1 (𝑥) = 𝑥 2 and 𝜙2 (𝑥) = 𝑥 from a
2𝑥 − 𝑥2
basis for the solution of (70). To find the solution of 𝐿(𝑦) = 𝑏(𝑥) , we know that its
particular solution is given by
𝜓𝑝 = 𝑢1 𝜙1 + 𝑢2 𝜙2
𝑢2
𝜓𝑝 = 𝑢1 𝑥 2 + (71)
𝑥
𝑏(𝑥)𝜙2 (𝑥) 𝑏(𝑥)𝜙1 (𝑥)
where 𝑢1 = − ∫ 𝑊(𝜙 𝑑𝑥, and 𝑢2 = ∫ 𝑊(𝜙 𝑑𝑥,
1 ,𝜙2 )(𝑥) 1 ,𝜙2 )(𝑥)

where 𝑊(𝜙1 , 𝜙2 )(𝑥) = −3 and 𝑏(𝑥) = 𝑥.


Therefore, we have
1 1
𝑢1 = ∫ 3 𝑑𝑥 ⟹ 𝑢1 = 3 𝑥,
and
1 𝑥4
𝑢2 = − ∫ 3 𝑥 3 𝑑𝑥 ⟹ − 12.
Hence
1 1 −𝑥 4
𝜓𝑝 = (3 𝑥) 𝑥 2 + 𝑥 ( 12 ),
1 𝑥3
⟹ 𝜓𝑝 = 3 𝑥 3 − 12,
𝑥3
⟹ 𝜓𝑝 = .
4
Hence any solution 𝜓 of equation (69) has the form
𝜓 = 𝜓𝑝 + 𝑐1 𝜙1 + 𝑐2 𝜙2
𝑥3 𝑐
⟹𝜓= + 𝑐1 𝑥 2 + 𝑥2 ,
4
where 𝑐1 and 𝑐2 are constants.
Problem 02-03-07:
One solution of 𝑥 2 𝑦′′ − 2𝑦 = 0 on 0 < 𝑥 < ∞ is 𝜙1 (𝑥) = 𝑥 2 . Find all solutions of the
equation 𝑥 2 𝑦′′ − 2𝑦 = 2𝑥 − 1 on 0 < 𝑥 < ∞.
2
Solution: Let 𝐿(𝑦) = 𝑦′′ − 𝑥 2 𝑦 = 0, (72)
2
where 𝑎1 (𝑥) = 0 and 𝑎2 (𝑥) = − 𝑥.
Given that 𝜙1 (𝑥) = 𝑥 2 is a solution of 𝐿(𝑦) = 0. The other solution 𝜙(𝑥) of 𝐿(𝑦) = 0 is
given by

S25013: Ordinary Differential Equations Page 116


𝑠
𝑥 exp[− ∫𝑥0 𝑎1 (𝑡)𝑑𝑡]
𝜙(𝑥) = 𝜙1 (𝑥) ∫𝑥 𝑑𝑠
0 [𝜙1 (𝑠)]2
𝑥 1
= 𝑥 2 ∫𝑥−0 𝑠4 𝑑𝑠
1 𝑥 𝑥2 1 1
= 𝑥 2 (− 3𝑠3 ) = − 3 (𝑥 3 − 𝑥 3 )
𝑥0 0
1 1
= − 3𝑥 + 3𝑥 3 𝑥 2
0
1
= 𝑐1 𝑥 + 𝑐2 𝑥 ,
2

1
where 𝜙1 (𝑥) = 𝑥 2 and 𝜙2 (𝑥) = 𝑥 form a basis for the solutions of (72). Now let 𝜓𝑝 be the
particular solution of the equation
𝑥 2 𝑦′′ − 2𝑦 = 2𝑥 − 1
2 2 1
⟹ 𝑦 ′′ − 𝑥 2 𝑦 = 𝑥 − 𝑥 2
2 2𝑥−1
⟹ 𝑦 ′′ − 𝑥 2 𝑦 = . (73)
𝑥2
2𝑥−1
One can easily find that 𝑊(𝜙1 , 𝜙2 )(𝑥) = −3 and 𝑏(𝑥) = . Then 𝜓p is given by
𝑥2
𝜓𝑝 = 𝑢1 𝑥 2 + 𝑢2 𝑥 −1 , (74)
𝑏(𝑥)𝜙2 (𝑥) 𝑏(𝑥)𝜙1 (𝑥)
where 𝑢1 = − ∫ 𝑊(𝜙 𝑑𝑥, and 𝑢2 = ∫ 𝑊(𝜙 𝑑𝑥.
1 ,𝜙2 )(𝑥) 1 ,𝜙2 )(𝑥)
Therefore, we have
1 2 1
𝑢1 = ∫ (2𝑥 − 1)𝑑𝑥 ⟹ 𝑢1 = − + 2
3𝑥 3 3𝑥 6𝑥
and
1 −1 1
𝑢2 = ∫ − 3 (2𝑥 − 1)𝑑𝑥 ⟹ 𝑢2 = 𝑥 2 + 3 𝑥.
3
Substituting the values of 𝑢1 and 𝑢2 in (74), we get
2 1 𝑥2 𝑥
𝜓𝑝 = (− 3𝑥 + 6𝑥 2 ) 𝑥 2 + (− + 3) 𝑥 −1
3
−2 1 𝑥 1
= 𝑥+6−3+3
3
1
= 2 − 𝑥.
Then every solution of (73) has the form
𝜓 = 𝜓𝑝 + 𝑐1 𝜙1 + 𝑐2 𝜙2
1
⟹ 𝜓 = 2 − 𝑥 + 𝑐1 𝑥 2 + 𝑐2 𝑥 −1 .
Problem 02-03-08:
One solution of 𝑥 2 𝑦′′ − 𝑥𝑦′ + 𝑦 = 0, 𝑥 > 0 is 𝜙1 (𝑥) = 𝑥 . Find the solution 𝜓 of 𝑥 2 𝑦′′ −
𝑥𝑦′ + 𝑦 = 𝑥 2 satisfying 𝜓(1) = 1, 𝜓′(1) = 0.
Solution: Let 𝐿(𝑦) = 𝑥 2 𝑦′′ − 𝑥𝑦′ + 𝑦 = 0 be the given equation.
1 1
⟹ 𝐿(𝑦) = 𝑦′′ − 𝑥 𝑦′ + 𝑥 2 𝑦 = 0, (75)
1 1
where 𝑎1 (𝑥) = − 𝑥 , 𝑎2 (𝑥) = 𝑥 2 .
Since 𝜙1 (𝑥) = 𝑥 is one solution of 𝐿(𝑦) = 0, then another solution is given by
𝑠
−∫ 𝑎 (𝑡)𝑑𝑡
𝑥 𝑒 𝑥0 1
ϕ(𝑥) = 𝜙1 (𝑥) ∫𝑥 [𝜙 (𝑠)]2 𝑑𝑠
0 1

S25013: Ordinary Differential Equations Page 117


𝑠 1
∫ 𝑑𝑡
𝑥 𝑒 𝑥0 𝑡
= 𝑥 ∫𝑥 𝑠2 𝑑𝑠
0
𝑠1
𝑑𝑡
𝑥 𝑒 ∫1 𝑡
= 𝑥 ∫𝑥 𝑠2 𝑑𝑠
0
𝑥 𝑠
= 𝑥 ∫1 𝑠2 𝑑𝑠
= 𝑥(log 𝑠)1𝑥
= 𝑥 log 𝑥. (76)
⟹ 𝜙1 (𝑥) = 𝑥 and 𝜙2 (𝑥) = 𝑥 log 𝑥 are independent solution of 𝐿(𝑦) = 0 which form a basis
for the solution of (75) since
𝑥 𝑥 log 𝑥
𝑊(𝜙1 , 𝜙2 )(𝑥) = | | = 𝑥 + 𝑥 log 𝑥 − 𝑥 log 𝑥 = 𝑥 > 0.
1 1 + log 𝑥
Now the particular solution 𝜓𝑝 of 𝐿(𝑦) = 𝑏(𝑥) is given by
𝜓𝑝 = 𝑢1 𝜙1 + 𝑢2 𝜙2
𝜓𝑝 = 𝑢1 𝑥 + 𝑢2 𝑥 log 𝑥,
𝑏(𝑥)𝜙2 (𝑥) 𝑏(𝑥)𝜙1 (𝑥)
where 𝑢1 = − ∫ 𝑊(𝜙 ,𝜙 )(𝑥) 𝑑𝑥, and 𝑢2 = ∫ 𝑊(𝜙 𝑑𝑥.
1 2 1 ,𝜙2 )(𝑥)

In this case 𝑏(𝑥) = 1 and hence, 𝑢1 and 𝑢2 are given by


𝑥log𝑥
𝑢1 = − ∫ 𝑑𝑥 = − ∫ log 𝑥 𝑑𝑥 = −(𝑥 log 𝑥 − 𝑥),
𝑥
𝑥
𝑢2 = ∫ 𝑥 𝑑𝑥 = ∫ 𝑑𝑥 = 𝑥.
Substituting these values in the equation for 𝜓𝑝 , we get
𝜓𝑝 = (𝑥 − 𝑥 log 𝑥)𝑥 + 𝑥 ⋅ 𝑥 log 𝑥
𝜓𝑝 = 𝑥 2 .
Thus any solution 𝜓 of the equation 𝐿(𝑦) = 𝑏(𝑥) is given by
𝜓 = 𝜓𝑝 + 𝑐1 𝜙1 + 𝑐2 𝜙2
𝜓 = 𝑥 2 + 𝑐1 𝑥 + 𝑐2 𝑥 log 𝑥, (77)
where 𝑐1 and 𝑐2 are constants. This solution satisfies the conditions 𝜓(1) = 1, 𝜓′(1) = 0.
Then 𝜓(1) = 1 ⟹ 1 = 1 + 𝑐1 ⟹ 𝑐1 = 0 and 𝜓′ (𝑥) = 2𝑥 + 𝑐1 + 𝑐2 (1 + log 𝑥)
⟹ 𝜓 ′ (1) = 0 ⟹ 0 = 2 + 0 + 𝑐2 ⟹ c2 = −2
⟹ ψ(x) = 𝑥 2 − 2𝑥 log 𝑥.
Problem 02-03-09:
Show that there is a basis 𝜙1 , 𝜙2 for the solution of 𝑥 2 𝑦′′ + 4𝑥𝑦′ + (2 + 𝑥 2 )𝑦 = 0, 𝑥 > 0 of
𝜓1 (𝑥) 𝜓2 (𝑥)
the form 𝜙1 (𝑥) = , 𝜙2 (𝑥) = . Hence find all solutions of 𝑥 2 𝑦 ′′ + 4𝑥𝑦 ′ +
𝑥2 𝑥2
(2 + 𝑥 2 )𝑦 = 𝑥 2 , 𝑥 > 0.
Solution: Let 𝐿(𝑦) = 𝑥 2 𝑦 ′′ + 4𝑥𝑦 ′ + (2 + 𝑥 2 )𝑦 = 0 , 𝑥 > 0. (78)
Let 𝜙 be a solution of the equation 𝑥 2 𝑦′′ + 4𝑥𝑦′ + (2 + 𝑥 2 )𝑦 = 0.
⟹ 𝑥 2 𝜙′′ + 4𝑥𝜙′ + (2 + 𝑥 2 )𝜙 = 0 (79)
Let 𝑣 = 𝑣(𝑥) be a function such that
𝑣(𝑥)
𝑣 = 𝑥2𝜙 ⟹ 𝜙 = (80)
𝑥2
where the function 𝑣 is to be determined. Since 𝜙 is a solution of the equation (78).
Therefore, from (80), we have
S25013: Ordinary Differential Equations Page 118
𝜙′ = 𝑣′𝑥 −2 − 2𝑥 −3 𝑣
𝜙′′ = 𝑣′′𝑥 −2 + 𝑣′(−2𝑥 −3 ) − 2𝑥 −3 𝑣′ + 6𝑥 −4 𝑣.
Substituting this in the equation (79), we get
𝑥 2 (𝑣′′𝑥 −2 − 4𝑥 −3 𝑣′ + 6𝑥 −4 𝑣) + 4𝑥(𝑣′𝑥 −2 − 2𝑥 −3 𝑣) + (2 + 𝑥 2 )𝑣𝑥 −2 = 0
4 6 𝑣′ 8
⟹ 𝑣′′ − 𝑣′ + 2 𝑣 + 4 − 2 𝑣 + 2𝑣𝑥 −2 + 𝑣 = 0
𝑥 𝑥 𝑥 𝑥
2 2
⟹ 𝑣 ′′ − 2 𝑣 + 2 𝑣 + 𝑣 = 0
𝑥 𝑥
⟹ 𝑣′′ + 𝑣 = 0. (81)
sin𝑥 cos𝑥
The equation (81) has solutions 𝑣1 = sin 𝑥 , 𝑣2 = cos 𝑥. Thus 𝜙1 (𝑥) = and 𝜙2 = are
𝑥2 𝑥2
solutions of (78). To show that 𝜙1 , 𝜙2 are linearly independent, consider constants 𝑐1 , 𝑐2
such that
𝑐1 𝜙2 + 𝑐2 𝜙2 = 0 ∀ 𝑥 > 0
⟹ 𝑐1 sin 𝑥 + 𝑐2 cos 𝑥 = 0 ∀ 𝑥 > 0
𝜋 𝜋
𝑐1 sin + 𝑐2 cos = 0
⟹ 2 2 } ⟹ 𝑐1 = 0, 𝑐2 = 0
𝑐1 sin 𝜋 + 𝑐2 cos 𝜋 = 0
⟹ 𝜙1 , 𝜙2 are linearly independent.
sin 𝑥 cos 𝑥
⟹ 𝜙1 = , 𝜙2 (𝑥) = form a basis for the solutions of the equation (78).
𝑥2 𝑥2
Now to find solution of the equation
𝑥 2 𝑦′′ + 4𝑥𝑦′ + (2 + 𝑥 2 )𝑦 = 𝑥 2 (82)
Let 𝜓𝑝 = 𝑢1 𝜙1 + 𝑢2 𝜙2 be the particular solution of (82) where 𝑢′1 and 𝑢′2 satisfies the
equation
𝑢′1 𝜙1 + 𝑢′2 𝜙2 = 0,
𝑢′1 𝜙′1 + 𝑢′2 𝜙′2 = 𝑏(𝑥).
sin 𝑥 cos 𝑥
⟹ 𝑢′1 2 + 𝑢′2 2 = 0.
𝑥 𝑥
𝑢′1 (𝑥 −2 cos 𝑥 − 2𝑥 −3 sin 𝑥) + u′2 (−𝑥 −2 sin 𝑥 − 2𝑥 −3 cos 𝑥) = 1.
Solving then equation we get
𝑢1′ 𝑢2′ 1
−2 = −2 = −2 𝑥 −2 cos 𝑥
|0 𝑥 cos 𝑥 | |𝑥 −2 sin 𝑥 0| |𝑥 −2 sin 𝑥 |
1 −𝑥 −2 sin 𝑥−2𝑥 −3 cos 𝑥 𝑥 cos 𝑥−2𝑥 −3 sin 𝑥 1 𝑥 cos 𝑥−2𝑥 −3 sin 𝑥 −𝑥 −2 sin 𝑥−2𝑥 −3 cos 𝑥
−𝑥 −2 cos𝑥
⟹ 𝑢1′ = 𝑥 −2 sin 𝑥(−𝑥 −2sin𝑥−2𝑥 −3 cos 𝑥)−𝑥 −2 cos 𝑥(𝑥 −2 cos 𝑥−2𝑥 −3 sin 𝑥)
−𝑥 −2 cos𝑥
⟹ u1′ = ⟹ 𝑢′1 = 𝑥 2 cos 𝑥
−𝑥 −4
𝑥 −2 sin 𝑥
and 𝑢2′ = ⟹ u′2 = −𝑥 2 sin 𝑥.
−𝑥 −4
Thus we obtain
𝑢1 = ∫ (𝑥 2 cos 𝑥)𝑑𝑥 = 𝑥 2 sin 𝑥 − ∫ 2 sin 𝑥 𝑑𝑥
⟹ 𝑢1 (𝑥) = 𝑥 2 sin 𝑥 + 2𝑥 cos 𝑥 − 2 sin 𝑥.
Similarly
𝑢2 (𝑥) = − ∫ 𝑥 2 sin 𝑥 𝑑𝑥
= −[−𝑥 −2 cos 𝑥 + 2 ∫ 𝑥 cos 𝑥 𝑑𝑥]

S25013: Ordinary Differential Equations Page 119


= 𝑥 2 cos 𝑥 − 2(𝑥 sin 𝑥 − ∫ sin 𝑥 𝑑𝑥)
𝑢2 (𝑥) = 𝑥 2 cos 𝑥 − 2𝑥 sin 𝑥 − 2 cos 𝑥.
Thus, particular solution becomes
sin 𝑥 cos 𝑥
𝜓𝑝 = (𝑥 2 sin 𝑥 + 2𝑥 cos 𝑥 − 2 sin 𝑥) 2 + (𝑥 2 cos 𝑥 − 2𝑥 sin 𝑥 − 2 cos 𝑥) 2
𝑥 𝑥
𝜓𝑝 = 1 − 2𝑥 . −2

Thus any solution 𝜓 of (88) is of the form


𝜓 = 𝜓𝑝 + 𝑐1 𝑥 −2 sin 𝑥 + 𝑐2 𝑥 −2 cos 𝑥,
𝜓 = 1 − 2𝑥 −2 + 𝑐1 𝑥 −2 sin 𝑥 + 𝑐2 𝑥 −2 cos 𝑥,
where 𝑐1 and 𝑐2 are constants.

SELF-TEST 01
MCQ 02-03-01
Let 𝑃 be a continuous function on ℝ and 𝑊 be the Wronskian of two linearly independent
solutions 𝑦1 , 𝑦2 of the ODE 𝑦 ′′ + (1 + 𝑥 2 )𝑦 ′ + 𝑃(𝑥)𝑦 = 0, 𝑥 ∈ ℝ. Let 𝑊(1) = 𝑎, 𝑊(2) = 𝑏
and 𝑊(3) = 𝑐. Then which of the following is true?
A. 𝑎 < 0 and 𝑏 > 0
B. 𝑎 < 𝑏 < 𝑐 or 𝑎 > 𝑏 > 𝑐
𝑎 𝑏 𝑐
C. |𝑎|
= |𝑏| = |𝑐|
D. 0 < 𝑎 < 𝑏 and 𝑏 > 𝑐 > 0.
MCQ 02-03-02
Let 𝑊 be the Wronskian of two linearly independent solutions 𝜙1 , 𝜙2 of the ordinary
differential equation 2𝑦 ′′ (𝑥) + 𝑦 ′ (𝑥) + 𝑥 2 𝑦 = 0, ∀𝑥 ∈ ℝ . Let 𝑊(2) = 𝑎, 𝑊(4) = 𝑏 . Then
which of the following is true?
A. 𝑎 < 0 and 𝑏 > 0
B. 𝑎 > 𝑏
𝑎 𝑏
C. |𝑎|
= |𝑏|
D. 0 < 𝑎 < 𝑏.
MCQ 02-03-03
One solution of the equation is 𝑥 2 𝑦 ′′ − 7𝑥𝑦 ′ + 15𝑦 = 0, x > 0 is 𝜙1 (𝑥) = x 3 . Then the
another independent solution of the equation is
A. 𝜙2 (𝑥) = 𝑥
B. 𝜙2 (𝑥) = 𝑥 2
C. 𝜙2 (𝑥) = 𝑥 4
D. 𝜙2 (𝑥) = 𝑥 5
MCQ 02-03-04
Let 𝜙1 (𝑥) = 𝑥 and 𝜙2 (𝑥) = 𝑥 log 𝑥 be two solutions of the differential equation is
𝐿(𝑦) = 𝑥 2 𝑦 ′′ − 𝑥𝑦 ′ + 𝑦 = 0, 𝑥 > 0. Then the particular intergal of 𝑥 2 𝑦 ′′ − 𝑥𝑦 ′ + 𝑦 = log 𝑥
is
A. ψp (𝑥) = log 𝑥 + 2
B. ψp (𝑥) = 2

S25013: Ordinary Differential Equations Page 120


C. ψp (𝑥) = log 𝑥
D. ψp (𝑥) = log 𝑥 − 2.
MCQ 02-03-05
One of the solutions of 𝐿(𝑦) = 𝑦 ′′ − 2𝑥𝑦 ′ + 2𝑦 = 0 is 𝜙1 (𝑥) = 𝑥, 𝑥 ≠ 0.Then the another
independent solution is
x 2
A. 𝜙2 (𝑥) = 𝑥 ∫x 𝑠 −2 𝑒 −𝑠 𝑑𝑠, 𝑥, 𝑥0 > 0
0
x 2
B. 𝜙2 (𝑥) = x ∫x 𝑠 2 𝑒 −𝑠 𝑑𝑠, 𝑥, 𝑥0 > 0
x 2
C. 𝜙2 (𝑥) = 𝑥 ∫x 𝑠 −2 𝑒 𝑠 𝑑𝑠, 𝑥, 𝑥0 > 0
0
x 2
D. 𝜙2 (𝑥) = 𝑥 ∫x 𝑠 2 𝑒 𝑠 𝑑𝑠, 𝑥, 𝑥0 > 0.
0

SHORT ANSWER QUESTIONS 01


SAQ 02-03-01
One solution of 𝑥 2 𝑦′′ − 𝑥𝑦′ + 𝑦 = 0, 𝑥 > 0 is 𝜙1 (𝑥) = 𝑥. Find another independent solution
of the equation.
Solution: Consider 𝑥 2 𝑦′′ − 𝑥𝑦′ + 𝑦 = 0, 𝑥 > 0 (83)
1 𝑦
⟹ 𝑦 ′′ − x y ′ + 𝑥 2 = 0, (84)
1 1
where 𝑎1 (𝑥) = − 𝑥 , 𝑎2 (𝑥) = 𝑥 2 . (85)
Since 𝜙1 (𝑥) = 𝑥 is a solution of the given equation (84). If 𝜙2 is another independent
solution of the given equation (84), then it is given by
𝑠
𝑥 𝑒 − ∫1 𝑎1 (𝑡)𝑑𝑡
𝜙2 (𝑥) = 𝜙1 (𝑥) ∫1 𝑑𝑠,
[𝜙1 (𝑠)]2
𝑡1
𝑑𝑡
𝑥 𝑒 ∫1 𝑡
= 𝑥 ∫1 𝑠2 𝑑𝑠
𝑠
𝑥 𝑒 (log𝑡)1
= 𝑥 ∫1 𝑑𝑠
𝑠2
𝑥 𝑠
= 𝑥 ∫1 𝑠2 𝑑𝑠
= 𝑥(log𝑠)1𝑥
= 𝑥 log 𝑥.
⟹ 𝜙2 = 𝑥 log 𝑥 is another independent solution of the equation (84).
SAQ 02-03-02
2
One solution of 𝑦′′ − 4𝑥𝑦 ′ + (4𝑥 2 − 2)𝑦 = 0 is 𝜙1 (𝑥) = 𝑒 𝑥 . Find another independent
solution of the equation.
Solution: Let 𝐿(𝑦) = 𝑦′′ − 4𝑥𝑦′ + (4𝑥 2 − 2)𝑦 = 0, (86)
where 𝑎1 (𝑥) = −4𝑥.
2
Since 𝜙1 (𝑥) = 𝑒 𝑥 is one solution of (86). If 𝜙2 (𝑥) is another independent solutions of the
equation (86), then we have
𝑠
𝑥 𝑒 ∫0 𝑎1 (𝑡)𝑑𝑡
𝜙2 (𝑥) = 𝜙1 (𝑥) ∫0 [𝜙 (𝑠)]2 𝑑𝑠
1

S25013: Ordinary Differential Equations Page 121


𝑠
𝑥2 𝑥 𝑒 ∫0 4𝑡𝑑𝑡
=𝑒 ∫0 𝑠2 2 𝑑𝑠
(𝑒 )
2
𝑥2 𝑥 𝑒 2𝑠
= 𝑒 ∫0 2𝑠2 𝑑𝑠
𝑒
𝑥2 𝑥
= 𝑒 ∫0 𝑑𝑠
2 𝑥
= 𝑒 𝑥 ∫0 𝑑𝑠
2
= 𝑒 𝑥 (𝑠)0𝑥
𝑥2
= 𝑒 𝑥.
𝑥2
⟹ 𝜙2 (𝑥) = 𝑥𝑒 is a nother independent solution of 𝐿(𝑦) = 0.
SAQ 02-03-03
One solution of 𝐿(𝑦) = 𝑥 3 𝑦′′′ − 3𝑥 2 𝑦′′ + 6𝑥𝑦′ − 6𝑦 = 0 for 𝑥 > 0 is 𝜙1 (𝑥) = 𝑥, find a basis
for the solution of 𝐿(𝑦) = 0 for 𝑥 > 0.
Solution: Let 𝐿(𝑦) = 𝑥 3 𝑦′′′ − 3𝑥 2 𝑦′′ + 6𝑥𝑦′ − 6𝑦 = 0 (87)
and 𝜙1 (𝑥) = 𝑥 is a known solution of (87). We try to find solution of (87) of the form
𝜙 = 𝑢𝜙1 ⟹ 𝜙 = 𝑢𝑥 ⟹ 𝐿(𝑢𝑥) = 0
⟹ 𝑥 3 (𝑢𝑥)′′′ − 3𝑥 2 (𝑢𝑥)′′ + 6𝑥(𝑢𝑥)′ − 6(𝑢𝑥) = 0
⟹ 𝑥 3 [𝑢′𝑥 + 𝑢]′′ − 3𝑥 2 (𝑢′𝑥 + 𝑢)′ + 6𝑥(𝑢′𝑥 + 𝑢) − 6𝑢𝑥 = 0
⟹ 𝑥 3 [𝑢′′𝑥 + 𝑢′ + 𝑢′]′ − 3𝑥 2 (𝑢′′𝑥 + 2𝑢′) + 6𝑢′ + 6𝑢𝑥 − 6𝑢𝑥 = 0
⟹ 𝑥 2 (𝑢′′′𝑥 + 3𝑢′′) − 3𝑥 2 (𝑢′′𝑥 + 2𝑢′) + 6𝑢′𝑥 = 0
⟹ 𝑥 3 𝑢′′′ = 0
𝑑3 𝑢
⟹ 𝑑𝑥 3 = 0. (88)
The solutions of (87) are obtained by integrating, 𝑢1 (𝑥) = 𝑥 and 𝑢2 (𝑥) = 𝑥 2 . Hence the
other solutions of (88) are
𝜙2 (𝑥) = 𝑢1 𝜙1 and 𝜙3 (𝑥) = 𝑢2 𝜙1
⟹ 𝜙2 (𝑥) = 𝑥 2 , 𝜙3 (𝑥) = 𝑥 3 .
Thus, 𝜙1 (𝑥) = 𝑥, 𝜙2 (𝑥) = 𝑥 2 and 𝜙3 (𝑥) = 𝑥 3 are linearly independent solutions of the
given equation 𝐿(𝑦) = 0.
SAQ 02-03-04
Consider the equation 𝑦′′ + 𝑦 = 𝑏(𝑥), where 𝑏(𝑥) is a continuous function on 1 ≤ 𝑥 < ∞

satisfying ∫1 |𝑏(𝑡)|𝑑𝑡 < ∞. Show that a particular solution 𝜓𝑝 is given by
𝑥
𝜓𝑝 (𝑥) = ∫1 sin(𝑥 − 𝑡)𝑏(𝑡)𝑑𝑡.
Solution: Let 𝐿(𝑦) = 𝑦′′ + 𝑦 = 𝑏(𝑥) . The two independent solution of 𝐿(𝑦) = 0 are
𝜙1 (𝑥) = cos 𝑥 , 𝜙2 (𝑥) = sin 𝑥. Hence the particular solution of 𝐿(𝑦) = 𝑏(𝑥) is given by
𝜓𝑝 = 𝑢1 𝜙1 + 𝑢2 𝜙2 ,
𝑥 −𝜙2 (𝑡)𝑏(𝑡) 𝑥 𝜙1 (𝑡)𝑏(𝑡)
where 𝑢1 = ∫1 𝑑𝑡, 𝑢2 (𝑥) = ∫1 𝑑𝑡,
𝑊(𝜙1 ,𝜙2 )(𝑡) 𝑊(𝜙1 ,𝜙2 )(𝑡)
cos 𝑥 sin 𝑥
Now we have 𝑊(𝜙1 , 𝜙2 )(𝑥) = | | = 1 and hence the functions 𝑢1 , 𝑢2 are given
− sin 𝑥 cos 𝑥
by
𝑥 −𝜙 (𝑡)𝑏(𝑡) 𝑥 − sin 𝑡𝑏(𝑡) 𝑥
𝑢1 = ∫1 𝑊(𝜙2 ,𝜙 )(𝑡) 𝑑𝑡 = ∫1 𝑑𝑡 = − ∫1
sin 𝑡 𝑏(𝑡)𝑑𝑡,
1 2 1

S25013: Ordinary Differential Equations Page 122


and
𝑥 𝜙1 (𝑡)𝑏(𝑡) 𝑥 cos 𝑡𝑏(𝑡) 𝑥
𝑢2 (𝑥) = ∫1 𝑑𝑡 = ∫1 𝑑𝑡 = ∫1 cos 𝑡 𝑏(𝑡)𝑑𝑡.
𝑊(𝜙1 ,𝜙2 )(𝑡) 1
⟹ 𝜓𝑝 = 𝑢1 𝜙1 + 𝑢2 𝜙2
𝑥 𝑥
= − cos 𝑥 ∫1 sin 𝑡 𝑏(𝑡)𝑑𝑡 + sin 𝑥 ∫1 cos 𝑡 𝑏(𝑡)𝑑𝑡
𝑥
= ∫1 𝑏(𝑡)[sin 𝑥 cos 𝑡 − cos 𝑥 sin 𝑡]𝑑𝑡
𝑥
= ∫1 sin(𝑥 − 𝑡)𝑏(𝑡)𝑑𝑡.

SUMMARY
If one non-vanishing solution of linear homogeneous differential equation is known, then
we can find another linearly independent solution by reducing the order of the differential
equation. In particular, if 𝜙1 is a non-vanishing solution on 𝐼 of a differential equation of order
two: 𝐿(𝑦) = 𝑦 ′′ + 𝑎1 (𝑥)𝑦 ′ + 𝑎2 𝑦 = 0, then second linearly independent solution 𝜙2 of this equation
is given by
𝑠
∫ 𝑎 (𝑡)𝑑𝑡
𝑥 𝑒 𝑥0 1
𝜙2 (𝑥) = 𝜙1 (𝑥) ∫𝑥 [𝜙 (𝑠)]2 𝑑𝑠
0 1

and the particular solution 𝜓𝑝 of 𝐿(𝑦) = 𝑏(𝑥) is given by


𝑥 [𝜙1 (𝑡)𝜙2 (𝑥)−𝜙1 (𝑥)𝜙2 (𝑡)]
𝜓𝑝 = ∫𝑥 𝑏(𝑡)𝑑𝑡.
0 𝑊(𝜙1 ,𝜙2 )(𝑡)
Further, we let 𝜙1 , 𝜙2 , … , 𝜙𝑛 be a basis for the solution of the differential equation of order
𝑛: 𝐿(𝑦) = 𝑦 (𝑛) + 𝑎1 (𝑥)𝑦 (𝑛−1) + 𝑎2 (𝑥)𝑦 (𝑛−2) + ⋯ + 𝑎𝑛 (𝑥)𝑦 = 0 on 𝐼. Then we have derived
the formula for particular solution 𝜓𝑝 of non-homogeneous differential equation of order
𝑛: 𝐿(𝑦) = 𝑏(𝑥) and is given by
𝑥 𝑊𝑘 (𝜙1 ,𝜙2 ,⋯,ϕn )(𝑡)
𝜓𝑝 (𝑥) = ∑𝑛𝑘=1 𝜙𝑘 (𝑥) ∫𝑥 𝑏(𝑡)𝑑𝑡,
0 𝑊(𝜙1 ,𝜙2 ,⋯,𝜙𝑛 )(𝑡)
where 𝑊𝑘 is the determinant obtained from 𝑊 by replacing the 𝑘 th column
(𝑛−1)
(𝜙𝑘 , 𝜙𝑘′ , … , 𝜙𝑘 ) by (0,0, ⋯ ,0,1).

END OF UNIT EXERCISES


2 2
1) Find the general solution of 𝑦 ′′ − 𝑥 𝑦 ′ + 𝑥 2 𝑦 = 𝑥 sin 𝑥 , 𝑥 ∈ [1, ∞) , given that 𝜙1 (𝑥) = 𝑥
and 𝜙2 (𝑥) = 𝑥 2 are linearly independent solutions of the corresponding homogeneous
equation.
1
2) Consider the equation 𝑥 2 𝑦 ′′ + 𝑥𝑦 ′ − 𝑦 = 𝑥. Show that 𝜙1 (𝑥) = 𝑥 and 𝜙2 (𝑥) = − 2𝑥 are
solutions of the homogeneous equation. Use the variation of parameters method to show
1 𝑥 log 𝑥
that a solution of the given equation is 𝑦(𝑥) = 𝑘1 𝑥 + 𝑘2 (− 2𝑥) + 2 .

KEY WORDS
Linear differential equation, particular solution, general solution, Wronskian of solutions,
reduction of order.

S25013: Ordinary Differential Equations Page 123


UNIT 02-04: HOMOGENEOUS EQUATIONS WITH ANALYTIC COEFFICIENTS
LEARNING OBJECTIVES
After successful completion of this unit, you will be able to
❖ explain singular point, ordinary point of a DE and power series with radius of convergence.
❖ apply power series solution with unknown coefficients to a DE such as Legendre equation
and obtain its general solution.
04-01: INTRODUCTION:
In the Unit 02-03, we have discussed reduction of the order of a homogeneous DE with
variable coefficients and the solutions were expressed in terms of some elementary
functions such as polynomial, exponential and trignometric functions.
However, there is a class of linear equations with variable coefficients which can not be
solved by analytic methods. One of the effective methods to obtain solutions of such
equations is the use of power series. Such a class of equations do occur in many physical
problems of engineering and science.
The central fact about such class of equations is that the behavior of the solutions near a
point 𝑥0 depends on the behavior of its coefficients functions. That is if the coefficients
𝑎1 (𝑥) and 𝑎2 (𝑥) of an equation 𝑦′′(𝑥) + 𝑎1 (𝑥)𝑦′ + 𝑎2 (𝑥)𝑦 = 0 with variable coefficients are
analytic at certain point, then the solutions of the equations are also analytic there.
Thus to study the solutions of an equation with variable coefficients in power series, we
define some terms, recall some results and record it below for the use in the sequel.
Definition: Analytic Function: A function 𝑓(𝑥) is said to be analytic in a neighborhood
of a point 𝑥 = 𝑥0 , if the derivatives of 𝑓(𝑥) exist and so 𝑓(𝑥) posses a convergent power
series expansion about 𝑥0 . That is, if the function 𝑓(𝑥) is analytic at 𝑥 = 𝑥0 , then it can be
represented in the form
𝑓(𝑥) = ∑∞ 𝑘=0 𝑐𝑘 (𝑥 − 𝑥0 )
𝑘
(1)
where 𝑐𝑘 are constants and the series converges for |𝑥 − 𝑥0 | < 𝑟0 , for some 𝑟0 > 0.
Definition: The series ∑∞ 𝑘=0 𝑐𝑘 (𝑥 − 𝑥0 ) is said to be convergent if lim ∑𝑘=0 𝑐𝑘 (𝑥 − 𝑥0 )
𝑘 𝑛 𝑘
𝑛→∞
exists.
Test for convergence of series
Ratio Test : Consider a series ∑∞
𝑘=0 𝑐𝑘 , where the 𝑐𝑘 are complex. If |𝑐𝑘 | > 0 for all 𝑘 and
𝑐𝑘+1
| | → 𝐿 as 𝑘 → ∞, then the series is convergent if 𝐿 < 1 and divergent for 𝐿 > 1.
𝑐𝑘
𝑧𝑘 𝑧𝑘 𝑐𝑘+1 |𝑧 |
Ex (i) Consider a series ∑∞
𝑘=0 we see that here 𝑐𝑘 = .| 𝑘
| = 𝑘+1 → 0 as 𝑘 → ∞
𝑘! 𝑘! 𝑐𝑘
𝑧𝑘
The series ∑∞
𝑘=0 is convergent for every 𝑧, such that |𝑧| < ∞. (i.e. for all complex 𝑧).
𝑘!
Some more examples are
(ii) The series ∑∞𝑘=0 𝑧 converges for |𝑧| < 1 and diverges for |𝑧| > 1.
𝑘

(iii) The series ∑∞𝑘=0 𝑘! 𝑧 fails to converge for all 𝑧 ≠ 0.


𝑘

Radius of convergence: Each power series in 𝑧 has a radius of convergence 𝑅, where 0 ≤


𝑅 ≤ ∞, with the property that the series converges if |𝑧| < 𝑅 and diverges if |𝑧| > 𝑅, when
𝑅 < ∞.

S25013: Ordinary Differential Equations Page 124


Note that it is customary to put 𝑅 = 0 when the series converges only for 𝑧 = 0 and 𝑅 = ∞
when it converges for all 𝑧.
For a given series ∑∞𝑘=0 𝑐𝑘 𝑧 , 𝑅 is obtained by the formula
𝑘
1
1 𝑐𝑘
= lim sup|ck |𝑘 and 𝑅 = lim |𝑐 |, if limit. Exits.
𝑅 𝑘→∞ 𝑘→∞ 𝑘+1

Singularities of a differential equation:


Consider the DE
𝑃(𝑥)𝑦′′ + 𝑄(𝑥)𝑦′ + 𝑅(𝑥)𝑦 = 0. (2)
The point 𝑥 = 𝑥0 is called the singular point of the equation (2) if 𝑃(𝑥0 ) = 0.
If we write equation (2) as
𝑦′′ + 𝑎1 (𝑥)𝑦′ + 𝑎2 (𝑥)𝑦 = 0, (3)
𝑄(𝑥) 𝑅(𝑥)
where 𝑎1 (𝑥) = , 𝑎2 (𝑥) = 𝑃(𝑥). (4)
𝑃(𝑥)
Then in this case if either 𝑎1 (𝑥) or 𝑎2 (𝑥) or both of the coefficients of (3) fails to be
analytic at 𝑥 = 𝑥0 . Hence it is difficult to determine the nature of the solutions in the
vicinity of such singular points.
However, there is a large class of equations for which the singularity is rather weak. In
these case the coefficients (𝑥 − 𝑥0 )𝑎1 (𝑥) and (𝑥 − 𝑥0 )2 𝑎2 (𝑥) are analytic. The singularity
𝑥 = 𝑥0 is called regular singular point otherwise irregular singular point. Then we
define
Regular Singular Point: A point 𝑥 = 𝑥0 of the equation 𝑦′′ + 𝑎1 (𝑥)𝑦′ + 𝑎2 (𝑥)𝑦 = 0 is said
to be regular singular point (rsp) if lim (𝑥 − 𝑥0 )𝑎1 (𝑥) and lim (𝑥 − 𝑥0 )2 𝑎2 (𝑥) exist and
𝑥→𝑥0 𝑥→𝑥0
finite. At such points the functions (𝑥 − 𝑥0 )𝑎1 (𝑥) and (𝑥 − are analytic.
𝑥0 )2 𝑎2 (𝑥)
Irregular Singular Point: If a singular point 𝑥 = 𝑥0 is not a regular singular point, then it
is called an irregular singular point (irsp).
A second order differential equation with regular singular point at 𝑥 = 𝑥0 has the form
(𝑥 − 𝑥0 )2 𝑦′′ + 𝑎(𝑥)(𝑥 − 𝑥0 )𝑦′ + 𝑏(𝑥)𝑦 = 0, (5)
where 𝑎(𝑥) and 𝑏(𝑥) are analytic at 𝑥 = 𝑥0 and have power series expansions
𝑎(𝑥) = ∑∞ 𝑘=0 𝑐𝑘 (𝑥 − 𝑥0 ) and 𝑏(𝑥) = ∑𝑘=0 𝑑𝑘 (𝑥 − 𝑥0 ) ,
𝑘 ∞ 𝑘

which are convergent on some interval |𝑥 − 𝑥0 | < 𝑟0 , 𝑟0 > 0.

04-02: EXISTENCE THEOREM FOR HOMOGENEOUS EQUATIONS WITH ANALYTIC COEFFICIENTS:


Ordinary Point: The points which are not singular are called ordinary point . When the
coefficients of a given differential equation are analytic then there always exists a
convergent power series solution for any initial value problem. We state below the
existence theorem for analytic coefficients.
Example 01: Consider the equation 𝑦 ′′ + 𝑥𝑦 ′ + (𝑥 2 + 2)𝑦=0.
Solution: Here 𝑎1 (𝑥) = 𝑥 , and 𝑎2 (𝑥) = 𝑥 2 + 2 . Both the functions 𝑎1 (𝑥) and 𝑎2 (𝑥) are
polynomial functions and they are analytic everywhere. Thus all points are ordinary points
of this differential equation.
1
Example 02: Consider the equation (𝑥 − 1)𝑦 ′′ + 𝑥𝑦 ′ + 𝑥 𝑦=0.

S25013: Ordinary Differential Equations Page 125


Solution: We rewrite the equation in the standard form 𝑦 ′′ + 𝑎1 (𝑥)𝑦 ′ + 𝑎2 (𝑥)𝑦 = 0 as
follows:
𝑥 1
𝑦 ′′ + (𝑥−1) 𝑦 ′ + 𝑥(𝑥−1) 𝑦=0. (6)
𝑥 1
Here 𝑎1 (𝑥) = (𝑥−1), and 𝑎2 (𝑥) = 𝑥(𝑥−1). The function 𝑎1 (𝑥) is analytic,except at 𝑥 = 1, and
𝑎2 (𝑥) is analytic except at 𝑥 = 0 and 𝑥 = 1. Thus 𝑥 = 0 and 𝑥 = 1 are singular points of
the differential equation under consideration. All other ponits are ordinary points. Note
clearly that 𝑥 = 0 is a singular ponit, even though 𝑎1 is analytic at 𝑥 = 0. We mention this
fact to emphasize that both 𝑎1 and 𝑎2 must be analytic at 𝑥0 in order for 𝑥0 to be an
ordinary point. Here both 0 and 1 are regular singular points of the DE.
Theorem 1: (Existence Theorem for analytic coefficients): Let 𝑥0 be a real number and
suppose that the coefficients 𝑎1 , 𝑎2 , … , 𝑎𝑛 in the equation
𝐿(𝑦) = 𝑦 (𝑛) + 𝑎1 (𝑥)𝑦 (𝑛−1) + 𝑎2 (𝑥)𝑦 (𝑛−2) + ⋯ + 𝑎𝑛 (𝑥)𝑦 = 0
have convergent power series expansions in powers of 𝑥 − 𝑥0 on an interval |𝑥 − 𝑥0 | <
𝑟0 , 𝑟0 > 0. If 𝛼1 , 𝛼2 , … , 𝛼𝑛 are any 𝑛 constants, there exists a solution 𝜙 of the IVP.
𝐿(𝑦) = 0, 𝑦(𝑥0 ) = 𝛼1 , 𝑦′(𝑥0 ) = 𝛼2 , … , 𝑦 (𝑛−1) (𝑥0 ) = 𝛼𝑛
with a power series expansion
𝜙(𝑥) = ∑∞ 𝑘=0 𝑐𝑘 (𝑥 − 𝑥0 )
𝑘

convergent for |𝑥 − 𝑥0 | < 𝑟0 .


This theorem gives us a sufficient condition for the existence of power series solutions of
the differential equation 𝐿(𝑦) = 0. It states that if 𝑥0 is an ordinary point of the equation
𝐿(𝑦) = 0, then this equaton has power series solutions in powers of (𝑥 − 𝑥0 ) and that these
series solutions are linearly independent. Thus, if 𝑥0 is an ordinary point of 𝐿(𝑦) = 0, we
may obtain the general solution of 𝐿(𝑦) = 0 as a linear combination of these linearly
independent power series solutions.
Note: (i) In Example 01, we observed that all points are ordinary points of the differetnial
equation 𝑦 ′′ + 𝑥𝑦 ′ + (𝑥 2 + 2)𝑦 =0. Thus this differential equation has two linearly
independent power series solutions about any point 𝑥0 .
(ii) In Example 02, we noted that 𝑥 = 0 and 𝑥 = 1 are the only singular points of the
1
differential equation (𝑥 − 1)𝑦 ′′ + 𝑥𝑦 ′ + 𝑥 𝑦 =0. Thus this differential equation has two
linearly independent power series solutions about any point 𝑥0 ≠ 0 or 1.
The method of solution near an ordinary point:
Note that we are assumed that under appropriate hypotheses, equation
𝑃(𝑥)𝑦′′ + 𝑄(𝑥)𝑦′ + 𝑅(𝑥)𝑦 = 0 (7)
actually does have power series solutions of the form
ϕ(𝑥) = ∑∞𝑘=0 𝑐𝑘 (𝑥 − 𝑥0 ) . (8)
𝑘

How do we proceed to find thses solutions? In other words, how do we determine the
coefficients 𝑐0 , 𝑐1 , ⋯ , 𝑐𝑛 , ⋯ in the expression (8) so that this expression become a solution
of (7)? We shall first give a brief outline of the procedure for finding th ese coefficients
and shall then illustrate the procedure in detail by considering specific examples.

S25013: Ordinary Differential Equations Page 126


Assuming that 𝑥0 is an ordinary point of the differential equation (7), so that solutions in
powers of (𝑥 − 𝑥0 ) actually do exist, we denote such a solution by
𝑦(𝑥) = 𝑐0 + 𝑐1 (𝑥 − 𝑥0 ) + 𝑐2 (𝑥 − 𝑥0 )2 + ⋯ = ∑∞ 𝑘=0 𝑐𝑘 (𝑥 − 𝑥0 ) . (9)
𝑘

Since the series in (9) converges on an interval |𝑥 − 𝑥0 | < 𝑟0 about 𝑥0 , it may be


differentiated term by term on this interval twice in succession to obtain
𝑦 ′ = 𝑐1 + 2𝑐2 (𝑥 − 𝑥0 ) + 3𝑐3 (𝑥 − 𝑥0 )2 + ⋯ = ∑∞ 𝑘=1 k 𝑐𝑘 (𝑥 − 𝑥0 )
𝑘−1
(10)
and
𝑘−2
𝑦 ′′ = 2𝑐2 + 6𝑐3 (𝑥 − 𝑥0 ) + 12𝑐4 (𝑥 − 𝑥0 )2 + ⋯ = ∑∞ 𝑘=2 k (𝑘 − 1)𝑐𝑘 (𝑥 − 𝑥0 ) (10)
respectively. We now substitute the series expansions for 𝑦 and its derivatives,respectively,
in the differential equation (7). We then simplify the resulting expression so that it takes
the form
𝑑0 + 𝑑1 (𝑥 − 𝑥0 ) + 𝑑2 (𝑥 − 𝑥0 )2 + ⋯ = 0, (11)
where the coefficients 𝑑𝑘 , (𝑘 = 0, 1,2, ⋯ ) are functions of certain coefficients 𝑐𝑘 of the
solution (8), which are all zeros. This leads to a set of conditions that must be satisfied by
the various coefficients 𝑐𝑘 in the series (8) in order to that (8) be a solution of the
differential equation (7). If the 𝑐𝑘 are chosen to satisfy the set of conditions that thus
occurs, then the resulting series (8) is the desired solution of the differential equation (7).
We shall illustrate this procedure in detail through following examples.
Example 03: Find two linearly independent power series solutions of the differential
equation 𝑦′′ + 𝑦 = 0 near the ordinary point 𝑥 = 0.
Solution: We see that the equation
𝑦′′ + 𝑦 = 0 (12)
has no singularities in the finite plane. Since the coefficients 𝑎1 = 0 and 𝑎2 = 1 are
analytic for all 𝑥0 . It turns on that the solution are also analytic at the ordinary point 𝑥0 =
0.
Let
𝑦 = ∑∞ 𝑘=0 𝑐𝑘 𝑥
𝑘
(13)
be a power series solution of (12).
⟹ 𝑦′(𝑥) = ∑∞ 𝑘=1 𝑐𝑘 𝑘𝑥
𝑘−1
,

𝑦′′(𝑥) = ∑𝑘=2 𝑐𝑘 𝑘(𝑘 − 1)𝑥 𝑘−2
⟹ 𝑦′′(𝑥) = ∑∞ 𝑘=0 𝑐𝑘+2 (𝑘 + 2)(𝑘 + 1)𝑥 .
𝑘

Substituting this in the equation (12), we get


∑∞ 𝑘
𝑘=0 [𝑐𝑘+2 (𝑘 + 1)(𝑘 + 2) + 𝑐𝑘 ]𝑥 = 0
Equating to zero the coefficients of all powers of 𝑥 we get
−𝑐
𝑘
𝑐𝑘+2 = (𝑘+1)(𝑘+2) ∀ 𝑘≥0 (14)
This equation (14) is called the recurrence relation. By giving 𝑘 = 0,1,2, … , 𝑛, ⋯ , we obtain
the constants
𝑐0 𝑐1
𝑐2 = − , 𝑐3 = − ,
1⋅2 1⋅2⋅3
(−1)2 𝑐0 (−1)2 𝑐0 (−1)2 𝑐1
𝑐4 = − ⟹ 𝑐4 = and 𝑐5 = .
1⋅2⋅3⋅4 4! 5!
In general we obtain

S25013: Ordinary Differential Equations Page 127


(−1)𝑘 𝑐0 (−1)𝑘 𝑐1
𝑐2𝑘 = and 𝑐2𝑘+1 =
(2𝑘)! (2𝑘+1)!
Substituting these values in the solution (13), we get
𝑦 = 𝑐0 + ∑∞
𝑘=1 𝑐2𝑘 𝑥
2𝑘
+ 𝑐1 𝑥 + ∑∞
𝑘=1 𝑐2𝑘+1 𝑥
2𝑘+1

(−1)𝑘 𝑥 2𝑘 (−1)𝑘 𝑥 2𝑘+1


𝑦 = 𝑐0 + 𝑐0 [∑∞
𝑘=1 ] + 𝑐1 𝑥 + 𝑐1 [∑∞
𝑘=1 ]
(2𝑘)! (2𝑘+1)!
(−1)𝑘 𝑥 2𝑘 (−1)𝑘 𝑥 2𝑘+1
or 𝑦 = 𝑐0 ∑∞
𝑘=0 + 𝑐1 ∑∞
𝑘=0 . (15)
(2𝑘)! (2𝑘+1)!
The series converges for all values of 𝑥.
Remark: By analytic method the solution of equation (13) is given by
𝑦 = 𝑐0 cos 𝑥 + 𝑐1 sin 𝑥 (16)
where we know the power series expansion of the cos 𝑥 and sin 𝑥 are given by
(−1)𝑘 𝑥 2𝑘
cos 𝑥 = ∑∞
𝑘=0 ,
(2𝑘)!
and
(−1)𝑘 𝑥 2𝑘+1
sin 𝑥 = ∑∞
𝑘=0 .
(2𝑘+1)!

Example 04: Find two linearly independent power series solutions of the equation
𝑦′′ − 𝑥 2 𝑦 = 0,
at ordinary point 𝑥 = 0.
Solution: The differential equation is given by
𝑦′′ − 𝑥 2 𝑦 = 0. (17)
Here 𝑎1 (𝑥) = 0 and 𝑎2 (𝑥) = −𝑥 are analytic for all 𝑥0 . It turns out that the solutions are
2

also analytic at the ordinary point 𝑥0 = 0. Therefore, but


𝑦 = ∑∞ 𝑘=0 𝑐𝑘 𝑥
𝑘
(18)
be the series solution of the equation (17). Differentiating the equation (18) twice and
substituting in the equation (17) and collecting the coefficients of like powers of 𝑥, we get
2𝑐2 + 6𝑐3 𝑥 + ∑∞ 𝑘
𝑘=2 [𝑐𝑘+2 (𝑘 + 1)(𝑘 + 2) − 𝑐𝑘−2 ]𝑥 = 0.
This is true if and only if all the coefficients of the powers of 𝑥 are zero.
1
𝑐2 = 0, 𝑐3 = 0 and 𝑐𝑘+2 = (𝑘+1)(𝑘+2) 𝑐𝑘−2 , ∀ 𝑘 ≥ 2. (19)
This is the recurrence relations. Giving different values of 𝑘 ≥ 2, we get
𝑐0
𝑘 = 2 ⟹ 𝑐4 = ,
3⋅4
𝑐1
𝑘 = 3 ⟹ 𝑐5 = ,
4⋅5
𝑐2
𝑘 = 4 ⟹ 𝑐6 = 5⋅6 ⟹ 𝑐6 = 0 as 𝑐2 = 0
𝑐
3
𝑘 = 5 ⟹ 𝑐7 = 6⋅7 ⟹ 𝑐7 = 0 as 𝑐3 = 0,
𝑐4 𝑐0
𝑘 = 6 ⟹ 𝑐8 = ⟹ 𝑐8 = ,
7⋅8 3⋅4⋅7⋅8
𝑐5 𝑐1
𝑘 = 7 ⟹ 𝑐9 = ⟹ 𝑐9 = ,
8⋅9 4⋅5⋅8⋅9
𝑐6
𝑘 = 8 ⟹ 𝑐10 = 9⋅10 ⟹ 𝑐10 = 0 as 𝑐6 = 0,
𝑐
7
𝑘 = 9 ⟹ 𝑐11 = 10⋅11 ⟹ 𝑐11 = 0 as 𝑐7 = 0,

S25013: Ordinary Differential Equations Page 128


Looking at the pattern, we have,
𝑐2 = 𝑐3 = 𝑐6 = 𝑐7 = 𝑐10 = 𝑐11 = ⋯ = 0 ⟹ 𝑐4𝑘−1 = 0 and 𝑐4𝑘−2 = 0 ∀ 𝑘 ≥ 1
𝑐 𝑐 𝑐
and 𝑐4 = 3⋅4
0 0
, 𝑐8 = 3⋅4⋅7⋅8 0
, 𝑐12 = 3⋅4⋅7⋅8⋅11⋅12⋯ ,
In general, we have
0 𝑐
𝑐4𝑘 = 3⋅4⋅7⋅8⋯(4𝑘−1)4𝑘 , ∀ 𝑘 ≥ 1.
Also,
𝑐
1 1 𝑐 1 𝑐
𝑐5 = 4⋅5 , 𝑐9 = 4⋅5⋅8⋅9 , 𝑐13 = 4⋅5⋅8⋅9⋅12⋅13⋯ , ⋯.
In general we have
1 𝑐
𝑐4𝑘+1 = 4⋅5⋅8⋅9⋯4𝑘(4𝑘+1) , 𝑘 ≥ 1.
Putting this in the equation (18), we get,
𝑥4 𝑥8 𝑥5 𝑥9
𝜙(𝑥) = 𝑐0 [1 + 3⋅4 + 3⋅4⋅7⋅8 + ⋯ ] + 𝑐1 [𝑥 + 4⋅5 + 4⋅5⋅8⋅9 + ⋯ ]
𝑥 4𝑘 𝑥 4𝑘+1
⟹ 𝜙(𝑥) = 𝑐0 [1 + ∑∞
𝑘=1 ] + 𝑐1 [𝑥 + ∑∞
𝑘=1 ],
3⋅4⋅7⋅8⋯(4𝑘−1)4𝑘 4⋅5⋅8⋅9⋯(4𝑘)(4𝑘+1)
is the solutuion of (17).
⟹ 𝜙(𝑥) = 𝑐0 𝜙1 (𝑥) + 𝑐1 𝜙2 (𝑥),
where
𝑥 4𝑘
𝜙1 (𝑥) = 1 + ∑∞
𝑘=1 3⋅4⋅7⋅8⋯(4𝑘−1)4𝑘
and
𝑥 4𝑘+1
𝜙2 (𝑥) = 𝑥 + ∑∞
𝑘=1 4⋅5⋅8⋅9⋯(4𝑘)(4𝑘+1)
are two solutions of equation (17), we see that,
1 0
𝑊(𝜙1 , 𝜙2 )(0) = | |=1≠0
0 1
⟹ the solutions 𝜙1 , 𝜙2 are linearly independent.
𝑑𝑘+1
Also if we represent 𝜙1 (𝑥) as 𝜙1 (𝑥) = 1 + ∑∞
𝑘=1 𝑑𝑘 (𝑥). Then | | → 0 as 𝑘 → ∞ for every
𝑑𝑘
finite 𝑥, so 𝜙1 (𝑥) is convergent for all 𝑥. Similarly, 𝜙2 (𝑥) is also convergent ∀𝑥.

04-03: THE LEGENDRE EQUATION:


One of the second order linear equations with analytic coefficients is the Legendre
equation. It is given by
𝐿(𝑦) = (1 − 𝑥 2 )𝑦′′ − 2𝑥𝑦′ + 𝑛(𝑛 + 1)𝑦 = 0, (20)
where 𝑛 is a constant. We see that 𝑥 = 1, −1, are singular points of the Legendre equation.
We write equation (20) as
2𝑥 𝑛(𝑛+1)
𝑦′′ − 1−𝑥 2 𝑦′ + 𝑦=0 (21)
1−𝑥 2
i.e. 𝑦′′ + 𝑎1 (𝑥)𝑦′ + 𝑎2 (𝑥)𝑦 = 0,
−2x 𝑛(𝑛+1)
where 𝑎1 (𝑥) = 1−𝑥 2 , 𝑎2 (𝑥) = .
1−𝑥 2
We see that lim (𝑥 ∓ 1)𝑎1 (𝑥) = 1 and lim (𝑥 ∓ 1)2 𝑎2 (𝑥) = 0 exist and finite. This shows
𝑥→±1 𝑥→±1
that the singular points 𝑥 = ±1 are regular singular points of the Legendre equation.
⟹ 𝑎1 (𝑥) and 𝑎2 (𝑥) have convergent power series expansion on |𝑥| < 1. It follows that the
solutions of the Legendre equation on |𝑥| < 1 have convergent power series expansions.
S25013: Ordinary Differential Equations Page 129
We shall show below the Legendre equation has convergent power series solutions at
ordinary point 𝑥 = 0.
Example 05: Compute two linearly independent power series solutions of the Legendre
equation (1 − 𝑥 2 )𝑦′′ − 2𝑥𝑦′ + 𝑛(𝑛 + 1)𝑦 = 0.
Solution: Let
(1 − 𝑥 2 )𝑦′′ − 2𝑥𝑦′ + 𝑛(𝑛 + 1)𝑦 = 0. (22)
Let ∞
𝑦 = ∑𝑘=0 𝑐𝑘 𝑥 𝑘
(23)
be any power series solution of the Legendre equation at ordinary point 𝑥0 = 0 .
Differentiating equation (23) twice and then putting the values of 𝑦′ and 𝑦′′ in the equation
(22), we obtain after collecting the coefficients of like powers of 𝑥,
[2𝑐2 + 𝑛(𝑛 + 1)𝑐0 ] + [2 ⋅ 3𝑐3 + (𝑛(𝑛 + 1) − 2)𝑐1 ]𝑥

+ ∑𝑘=0 [𝑐𝑘+4 (𝑘 + 3)(𝑘 + 4) + (𝑛(𝑛 + 1) − (𝑘 + 2)(𝑘 + 3))𝑐𝑘+2 ]𝑥 𝑘 = 0
This equation is true if and only if all the coefficients of the powers of 𝑥 are zero. Hence
equating to zero the coefficients of all powers of 𝑥, we get
1 (𝑛−1)(𝑛+2)
𝑐2 = − 2 𝑛(𝑛 + 1)𝑐0 , 𝑐3 = − 𝑐1 ,
2⋅3
and
𝑛(𝑛+1)−(𝑘+2)(𝑘+3)
𝑐𝑘+4 = − (𝑘+3)(𝑘+4)
𝑐𝑘+2 ∀ 𝑘 ≥ 0. (24)
This the recurrence relation. By giving different values to 𝑘 = 0, 1, 2 ⋯, we obtain
(𝑛+3)(𝑛+1)𝑛(𝑛−2)
𝑐4 = (−1)2 0𝑐 ,
4!
2 (𝑛+4)(𝑛+2)(𝑛−1)(𝑛−3)
𝑐5 = (−1) 1 𝑐 ,
5!
3 (𝑛+5)(𝑛+3)(𝑛+1)𝑛(𝑛−2)(𝑛−4)
𝑐6 = (−1) 0 𝑐 ,
6!
3 (𝑛+6)(𝑛+4)(𝑛+2)(𝑛−1)(𝑛−3)(𝑛−5)
𝑐7 = (−1) 𝑐1 .
7!
The pattern is now clear. It follows by induction that
(𝑛+2𝑘−1)(𝑛+2𝑘−3)⋯(𝑛+1)𝑛(𝑛−2)⋯(𝑛−2𝑘+2)
𝑐2𝑘 = (−1)𝑘 (2𝑘)!
𝑐0
and
(𝑛+2𝑘)(𝑛+2𝑘−2)⋯(𝑛+2)(𝑛−1)(𝑛−3)⋯(𝑛−2𝑘+1)
𝑐2𝑘+1 = (−1)𝑘 𝑐1 .
(2𝑘+1)!
All these coefficients are determined in terms of 𝑐0 and 𝑐1 . Thus we have the general
solution
𝜙(𝑥) = 𝑐0 𝜙1 (𝑥) + 𝑐1 𝜙2 (𝑥),
where the functions 𝜙1 (𝑥) and 𝜙2 (𝑥) are given by
(𝑛+2𝑘−1)⋯(𝑛+1)𝑛(𝑛−2)⋯(𝑛−2𝑘+2)
𝜙1 (𝑥) = 1 + ∑∞
𝑘=1 (−1)
𝑘
𝑥 2𝑘 (25)
(2𝑘)!
and
(𝑛+2𝑘)(𝑛+2𝑘−2)⋯(𝑛+2)(𝑛−1)⋯(𝑛−2𝑘+1)
𝜙2 (𝑥) = 𝑥 + ∑∞
𝑘=1 (−1)
𝑘
𝑥 2𝑘+1 (26)
(2𝑘+1)!
Both 𝜙1 (𝑥) and 𝜙2 (𝑥) are the solutions of the Legendre equation. We see from the
solutions that
𝜙1 (0) = 1, 𝜙1′ (0) = 0,
and
𝜙2 (0) = 0, ϕ′2 (0) = 1

S25013: Ordinary Differential Equations Page 130


1 0
⟹ 𝑊(𝜙1 , 𝜙2 )(0) = | |≠0
0 1
⟹ 𝜙1 (𝑥) and 𝜙2 (𝑥) are linearly independent solutions of the Legendre's equation.
Observations:
Thus, the two independent solutions are
(𝑛+1)𝑛 (𝑛+3)(𝑛+1)𝑛(𝑛−2) (𝑛+5)(𝑛+3)(𝑛+1)𝑛(𝑛−2)(𝑛−4) 6
𝜙1 (𝑥) = 1 − 𝑥2 + 𝑥4 − 𝑥
2! 4! 6!
𝑘 (𝑛+2𝑘−1)⋯(𝑛+1)𝑛(𝑛−2)⋯(𝑛−2𝑘+2) 2𝑘
+ ⋯ + (−1) (2𝑘)!
𝑥
𝑘+1 (𝑛+2(𝑘+1)−1)⋯(𝑛+1)𝑛(𝑛−2)⋯(𝑛−2(𝑘+1)+2)
+(−1) (2(𝑘+1))!
𝑥 2(𝑘+1)
+⋯ , (27)
and
(n+2)(n−1) 3 (𝑛+4)(n+2)(n−1)(𝑛−3) 5 (𝑛+6)(𝑛+4)(n+2)(n−1)(𝑛−3)(𝑛−5) 7
𝜙2 (𝑥) = 𝑥 − 𝑥 + 𝑥 − 𝑥
3! 5! 7!
(𝑛+2𝑘)(𝑛+2𝑘−2)⋯(𝑛+2)(𝑛−1)⋯(𝑛−2𝑘+1)
+ ⋯ + (−1)𝑘 (2𝑘+1)!
𝑥 2𝑘+1
(𝑛+2 (𝑘+1))(𝑛+2(𝑘+1)−2)⋯(𝑛+2)(𝑛−1)⋯(𝑛−2(𝑘+1)+1)
+(−1)𝑘+1 (2(𝑘+1)+1)!
𝑥 2(𝑘+1)+1
+ ⋯. (28)
Case: 1. If 𝑛 = 2𝑚, 𝑚 = 0, 1, 2, 3, ⋯, we note that
𝑛(𝑛 − 2)(𝑛 − 4) ⋯ (𝑛 − 2𝑘 + 2) = 2𝑚(2𝑚 − 2)(2𝑚 − 4) ⋯ (2𝑚 − 2𝑘 + 2)
= 2𝑚2(𝑚 − 1)2(𝑚 − 2) ⋯ 2(𝑚 − 𝑘 + 1)
= 2𝑘 𝑚(𝑚 − 1)(𝑚) − 2) ⋯ (𝑚 − 𝑘 + 1)
2𝑘 𝑚(𝑚−1)(𝑚)−2)⋯ (𝑚−𝑘+1)(𝑚−𝑘)!
= (𝑚−𝑘)!
2𝑘 𝑚!
= (𝑚−𝑘)! ,
and
(𝑛 + 1)(𝑛 + 3) ⋯ (𝑛 + 2𝑘 − 1) = (2𝑚 + 1)(2𝑚 + 3) ⋯ (2𝑚 + 2𝑘 − 1)
(2𝑚+1)(2𝑚+2)(2𝑚+3)(2𝑚+4)⋯(2𝑚+2𝑘−1)(2𝑚+2𝑘)
= (2𝑚+2)(2𝑚+4)⋯(2𝑚+2𝑘)
(2𝑚)!(2𝑚+1)(2𝑚+2)(2𝑚+3)(2𝑚+4)⋯(2𝑚+2𝑘−1)(2𝑚+2𝑘)
= (2𝑚)!(2𝑚+2)(2𝑚+4)⋯(2𝑚+2𝑘)
(2𝑚+2𝑘)!
= (2𝑚)!(2𝑚+2)(2𝑚+4)⋯(2𝑚+2𝑘)
(2𝑚+2𝑘)!
= (2𝑚)!2(𝑚+1)2(𝑚+2)⋯2(𝑚+𝑘)
(2𝑚+2𝑘)!
= (2𝑚)!2𝑘(𝑚+1)(𝑚+2)⋯(𝑚+𝑘)
(2𝑚+2𝑘)!𝑚!
= (2𝑚)!2𝑘𝑚!(𝑚+1)(𝑚+2)⋯(𝑚+𝑘)
(2𝑚+2𝑘)!𝑚!
= 2𝑘(2𝑚)!(𝑚+𝑘)!.
Also, from (27), one of factors of the coefficient of 𝑥 2(𝑘+1) implies for 𝑘 = 𝑚 that
(2𝑚 − 2(𝑘 + 1) + 2) = 2𝑚 − 2(𝑚 + 1) + 2 = 2𝑚 − 2𝑚 − 2 + 2 = 0 , so all coefficients in
the expansion of 𝜙1 (𝑥) of 𝑥 2𝑡 are zeros, for 𝑛 = 2𝑚, 𝑡 > 𝑚. Then, therefore in this case,
𝜙1 (𝑥) becomes

S25013: Ordinary Differential Equations Page 131


(2𝑚+2𝑘)!𝑚! 2𝑘 𝑚!
𝑘 (2𝑚)!(𝑚+𝑘)! (𝑚−𝑘)!
𝜙1 (𝑥) = 1 + ∑m
𝑘=1 (−1)𝑘 2 𝑥 2𝑘
(2𝑘)!
(2𝑚+2𝑘)!
(𝑚!)2
= 1 + (2𝑚)! ∑𝑚
𝑘=1 (−1)
𝑘 (𝑚−𝑘)!(𝑚+𝑘)! 2𝑘
𝑥
(2𝑘)!
(𝑚!)2 (2𝑚+2𝑘)!
= 1 + (2𝑚)! ∑𝑚 𝑘
𝑘=1 (−1) (𝑚−𝑘)!(𝑚+𝑘)!(2𝑘)! 𝑥
2𝑘

(𝑚!)2 (2𝑚+2𝑘)!
= 1 + (2𝑚)! ∑m 𝑘 2𝑘
𝑘=1 (−1) (𝑚−𝑘)!(𝑚+𝑘)!(2𝑘)! 𝑥 ,

which is a polynomial of degree 𝑛 = 2𝑚 and it is also even function. In particular, for


𝑚 = 0, 1, 2 ⟹ 𝑛 = 0, 2, 4, the corresponding polynomials are
𝜙1 (𝑥) = 1, 𝑛=0
2
𝜙1 (𝑥) = 1 − 3𝑥 , 𝑛=2
35
𝜙1 (𝑥) = 1 − 10𝑥 2 + 𝑥4, 𝑛 = 4.
3
Note that the series 𝜙2 (𝑥) is not a polynomial when 𝑛 is even because the coefficients of
𝑥 2𝑘+1 is never zero.
Case: 2. If 𝑛 = 2𝑚 + 1, 𝑚 = 0, 1, 2, 3, ⋯, we note that
(𝑛 − 1)(𝑛 − 3)(𝑛 − 5) ⋯ (𝑛 − 2𝑘 + 1)
= (2𝑚 + 1 − 1)(2𝑚 + 1 − 3)(2𝑚 + 1 − 5) ⋯ (2𝑚 + 1 − 2𝑘 + 1)
= (2𝑚)(2𝑚 − 2)(2𝑚 − 4) ⋯ (2𝑚 − 2𝑘 + 2)
= 2(𝑚)2(𝑚 − 1)2(𝑚 − 2) ⋯ 2(𝑚 − 𝑘 + 1)
= 2(𝑚)2𝑘−1 (𝑚 − 1)(𝑚 − 2) ⋯ (𝑚 − 𝑘 + 1)
2𝑘 𝑚(𝑚−1)(𝑚−2)⋯(𝑚−𝑘+1)(𝑚−𝑘!)
= (𝑚−𝑘)!
2𝑘 𝑚!
= (𝑚−𝑘)! ,
and
(𝑛 + 2)(𝑛 + 4) ⋯ (𝑛 + 2𝑘) = (2𝑚 + 1 + 2)(2𝑚 + 1 + 4) ⋯ (2𝑚 + 1 + 2𝑘)
= (2𝑚 + 3)(2𝑚 + 5) ⋯ (2𝑚 + 2𝑘 + 1)
(2𝑚+2)(2𝑚+3)(2𝑚+4)(2𝑚+5)⋯(2𝑚+2𝑘)(2𝑚+2𝑘+1)
= (2𝑚+2)(2𝑚+4)⋯(2𝑚+2𝑘)
(2𝑚+1)!(2𝑚+2)(2𝑚+3)(2𝑚+4)(2𝑚+5)⋯(2𝑚+2𝑘)(2𝑚+2𝑘+1)
= (2𝑚+1)!(2𝑚+2)(2𝑚+4)⋯(2𝑚+2𝑘)
(2𝑚+2𝑘+1)!
= (2𝑚+1)!(2𝑚+2)(2𝑚+4)⋯(2𝑚+2𝑘)
(2𝑚+2𝑘+1)!
= (2𝑚+1)!2(𝑚+1)2(𝑚+2)⋯2(𝑚+𝑘)
(2𝑚+2𝑘+1)!
= (2𝑚+1)!2𝑘(𝑚+1)(𝑚+2)⋯(𝑚+𝑘)
(2𝑚+2𝑘+1)!𝑚!
= (2𝑚+1)!2𝑘𝑚!(𝑚+1)(𝑚+2)⋯(𝑚+𝑘)
(2𝑚+2𝑘+1)!𝑚!
= (2𝑚+1)!2𝑘(𝑚+𝑘)!
𝑚!(2𝑚+2𝑘+1)!
= 2𝑘(2𝑚+1)!(𝑚+𝑘)!.
Similarly, from (28), one of factors of the coefficient of 𝑥 2(𝑘+1)+1 implies for 𝑘 + 1 = 𝑚 +
1 that (2𝑚 + 1 − 2(𝑘 + 1) + 1) = 2𝑚 + 1 − 2(𝑚 + 1) + 1 = 2𝑚 + 1 − 2𝑚 − 2 + 1 = 0.

S25013: Ordinary Differential Equations Page 132


Then, therefore in this case, 𝜙2 (𝑥) becomes
𝑚!(2𝑚+2𝑘+1)! 2𝑘 𝑚!
𝑘 (2𝑚+1)!(𝑚+𝑘)! (𝑚−𝑘)!
𝜙2 (𝑥) = 𝑥 + ∑m
𝑘=1 (−1)𝑘 2 𝑥 2𝑘+1
(2𝑘)!
(2𝑚+2𝑘+1)!
(𝑚!)2 (𝑚−𝑘)!(𝑚+𝑘)!
= 1+ ∑𝑚 (−1)𝑘 (2𝑘+1)! 𝑥 2𝑘+1
(2𝑚+1)! 𝑘=1
(𝑚!)2 (2𝑚+2𝑘+1)!
= 1 + (2𝑚+1)! ∑𝑚 𝑘
𝑘=1 (−1) (𝑚−𝑘)!(𝑚+𝑘)!(2𝑘+1)! 𝑥
2𝑘+1
,
which is a polynomial of degree 𝑛 = 2𝑚 + 1 and it is also odd function. In particular, for
𝑚 = 0,1, 2 ⟹ 𝑛 = 1, 3, 5, the corresponding polynomials are
𝜙2 (𝑥) = 𝑥, 𝑛=1
5
𝜙2 (𝑥) = 𝑥 − 𝑥 3 , 𝑛=3
3
14 21
𝜙2 (𝑥) = 𝑥 − 3 𝑥 3 + 5 𝑥 5 , 𝑛 = 5.
Note that the series 𝜙1 (𝑥) is not a polynomial when 𝑛 is odd because the coefficients of
𝑥 2𝑘 is never zero.
Thus, if 𝑛 is nonnegative integer, then exactly one solution is a polynomial of degree 𝑛
and other is an infinite power series.
Legendre Polynomials:
The general solution 𝜙(𝑥) = 𝑐0 𝜙1 (𝑥) + 𝑐1 𝜙2 (𝑥) is called a Legendre function.
If 𝑛 is nonnegative integer, then precisely one of 𝜙1 or 𝜙2 is a polynomial of degree 𝑛. The
polynomial solution 𝑃𝑛 (𝑥) of degree 𝑛 of
(1 − 𝑥 2 )𝑦 ′′ − 2𝑥𝑦 ′ + 𝑛(𝑛 + 1)𝑦 = 0,
satisfying 𝑃𝑛 (1) = 1 is called 𝑛𝑡ℎ -Legendre polynomial.
For 𝑛 ≥ 0, note that 𝑃𝑛 (𝑥) is a polynomial solution of degree 𝑛 and it is an even function
if 𝑛 is even, and an odd function if 𝑛 is odd.
In order to justify this definition, we must show that there is just one such solution for
each non-negative integer 𝑛. This will be established in the following example.
Example 5: Let us define a function 𝜙 be the polynomial of degree 𝑛 defined by
𝑑𝑛
𝜙(𝑥) = 𝑑𝑥 𝑛 (𝑥 2 − 1)𝑛
satisfies the Legendre’s equation (1 − 𝑥 2 )𝑦′′ − 2𝑥𝑦′ + 𝑛(𝑛 + 1)𝑦 = 0 . Hence show that
𝜙(1) = 2𝑛 𝑛!.
Solution: Let 𝜙 be the polynomial of degree 𝑛 and defined by
𝑑𝑛
𝜙(𝑥) = 𝑑𝑥 𝑛 (𝑥 2 − 1)𝑛 . (29)
Assume 𝑢(𝑥) = (𝑥 2 − 1)𝑛 . (30)
Differentiating with respect to 𝑥, we get
𝑢′(𝑥) = 2𝑛(𝑥 2 − 1)𝑛−1 𝑥
⟹ (𝑥 2 − 1)𝑢′(𝑥) = 2𝑛(𝑥 2 − 1)𝑛 𝑥
or (𝑥 2 − 1)𝑢′ − 2𝑛𝑥𝑢 = 0.
If 𝑓(𝑥) and 𝑔(𝑥) are functions of 𝑥, then we know the formula
𝐷𝑛 (𝑓𝑔) = 𝐷𝑛 𝑓 ⋅ 𝑔+𝑛 𝒞1 𝐷𝑛−1 𝑓 ⋅ 𝐷𝑔+𝑛 𝒞2 𝐷𝑛−2 𝑓𝐷2 𝑔 + ⋯ +𝑛 𝒞𝑛−1 𝐷𝑓𝐷𝑛−1 𝑔 + 𝑓𝐷𝑛 𝑔 (31)

S25013: Ordinary Differential Equations Page 133


Using this formula, differentiate the equation (𝑥 2 − 1)𝑢′ − 2𝑛𝑥𝑢 = 0, (𝑛 + 1) times, we get
(𝑥 2 − 1)𝑢(𝑛+2) +𝑛+1 𝒞1 2𝑥𝑢(𝑛+1) +𝑛+2 𝒞2 2 ⋅ 𝑢(𝑛) − 2𝑛𝑥𝑢(𝑛+1) − 2𝑛 𝑛+1 𝒞1 𝑢(𝑛) = 0
⟹ (𝑥 2 − 1)𝑢(𝑛+2) + 2𝑥(𝑛 + 1)𝑢(𝑛+1) + 𝑛(𝑛 + 1)𝑢(𝑛) − 2𝑛𝑥𝑢(𝑛+1) − 2𝑛(𝑛 + 1)𝑢(𝑛) = 0.
Using the notations (29) and (30), we have
𝜙(𝑥) = 𝑢(𝑛) , 𝜙 ′ (𝑥) = 𝑢(𝑛+1) , 𝜙′′(𝑥) = 𝑢(𝑛+2)
⟹ (𝑥 2 − 1)𝜙 ′′ (𝑥) + 2𝑥𝜙 ′ (𝑥) − 𝑛(𝑛 + 1)𝜙(𝑥) = 0
⟹ (1 − 𝑥 2 )𝜙 ′′ (𝑥) − 2𝑥𝜙′(𝑥) + 𝑛(𝑛 + 1)𝜙(𝑥) = 0.
This shows that 𝜙(𝑥) defined in (29) satisfies the Legendre’s equation.
Now we prove 𝜙(1) = 2𝑛 𝑛!.
We have 𝜙(𝑥) = [(𝑥 2 − 1)𝑛 ](𝑛)
⟹ 𝜙(𝑥) = [(𝑥 − 1)𝑛 (𝑥 + 1)𝑛 ](𝑛) .
Using the formula (31), we obtain
𝜙(𝑥) = [(𝑥 − 1)𝑛 ](𝑛) (𝑥 + 1)𝑛 +𝑛 𝒞1 [(𝑥 − 1)𝑛 ](𝑛−1) 𝐷(𝑥 + 1)𝑛
+ ⋯ + (𝑥 − 1)𝑛 [(𝑥 + 1)𝑛 ](n) (32)
where
𝑑𝑛
[(𝑥 − 1)𝑛 ](𝑛) = 𝑑𝑥 𝑛 (𝑥 − 1)𝑛
𝑑𝑛−1
= 𝑛 𝑑𝑥 𝑛−1 (𝑥 − 1)𝑛−1
𝑑𝑛−2
= 𝑛(𝑛 − 1) 𝑑𝑥 𝑛−2 (𝑥 − 1)𝑛−2

𝑑
= 𝑛(𝑛 − 1)(𝑛 − 2) ⋯ (𝑛 − (𝑛 − 1)) 𝑑𝑥 (𝑥 − 1)
= 𝑛(𝑛 − 1)(𝑛 − 2) ⋯ 1
𝑑𝑛
(𝑥 − 1)𝑛 = 𝑛!. (33)
𝑑𝑥 𝑛
Hence equation (32) becomes
𝜙(𝑥) = 𝑛! (𝑥 + 1)𝑛 + terms containing (𝑥 − 1)
𝜙(1) = 2𝑛 𝑛!.
It is now clear that the function, given by
1 𝑑𝑛
𝑃𝑛 (𝑥) = 2𝑛 𝑛! 𝑑𝑥 𝑛 (𝑥 2 − 1)𝑛 (34)
is the 𝑛𝑡ℎ −Legendre polynomial, provided we can show that there is no other polynomial
solution of Legendre equation which is 1 at 𝑥 = 1.
Suppose 𝜓 is any polynomial solution of Legendre equation.Then for some constant 𝑐 we
must have 𝜓(𝑥) = 𝑐𝜙1 or 𝜓(𝑥) = 𝑐𝜙2 according as 𝑛 is even or odd. Here 𝜙1 , 𝜙2 are two
linearly independent solutions are given by (25) and (26) respectively. Suppose 𝑛 is even,
for example. Then for |𝑥| < 1,
𝜓(𝑥) = 𝑐𝜙1 (𝑥) + 𝑑𝜙2 (𝑥)
for some constants 𝑐, 𝑑, since 𝜙1 , 𝜙2 form a basis for the solutions on |𝑥| < 1. But then
𝜓 − 𝑐𝜙1 is a polynomial, whereas 𝑑𝜙2 is not a polynomial in case 𝑑 ≠ 0. Hence 𝑑 = 0. In
particular the function 𝑃𝑛 given by (34) satisfies 𝑃𝑛 (1) = 1 and 𝑃𝑛 (𝑥) = 𝑐𝜙1 (𝑥) for some
constant 𝑐, if 𝑛 is even. Since 1 = 𝑃𝑛 (1) = 𝑐𝜙1 (1), we see that 𝜙1 (1) ≠ 0. A similar result

S25013: Ordinary Differential Equations Page 134


is valid if 𝑛 is odd. Thus no non-trivial polynomial solution of the Legendre equation can
be zero at 𝑥 = 1 . From this it follows that there is only one polynomial 𝑃𝑛 satisfying
Legendre equation and 𝑃𝑛 (1) = 1, for if 𝑃𝑛 was another, then 𝑃𝑛 − 𝑃𝑛 , would be a
polynomial solution, and 𝑃𝑛 (1) − 𝑃𝑛 (1) = 1 − 1 = 0, which is a contradiction. This proves
that 𝑃𝑛 is the only such solution.

04-04: JUSTIFICATION OF THE POWER SERIES METHOD


We now consider the proof of Theorem 1. In order not to be complicate matters too much,
we shall give a proof for the case 𝑛 = 2 and 𝑥0 = 0. All the essential ideas in this case. We
shall make use of two results concerning power series. The first is that if we have two
power series
∑∞𝑘=0 𝑐𝑘 𝑥 , ∑𝑘=0 𝐶𝑘 𝑥 ,
𝑘 ∞ 𝑘

and we know that


|𝑐𝑘 | ≤ 𝐶𝑘 , 𝐶𝑘 ≥ 0, (𝑘 = 0, 1, 2, ⋯ )
and that the series
∑∞𝑘=0 𝐶𝑘 𝑥
𝑘

converges for |𝑥| < 𝑟, for some 𝑟 > 0, then the series
∑∞𝑘=0 𝑐𝑘 𝑥
𝑘

also converges for |𝑥| < 𝑟. This is usually called the comparison test for convergence. The
second result we require is that if a series
∑∞𝑘=0 𝛼𝑘 𝑥
𝑘
(35)
is convergent for |𝑥| < 𝑟0 , then for any 𝑥, |𝑥| = 𝑟 < 𝑟0 , there is a constant 𝑀 > 0 such that
𝑟 𝑘 |𝛼𝑘 | ≤ 𝑀, (𝑘 = 0, 1, 2, ⋯ ). (36)
This is not difficult to show. Since the series (35) is convergent |𝑥| = 𝑟, its terms must
tends to zero,
|𝛼𝑘 𝑥 𝑘 | = |𝛼𝑘 |𝑟 𝑘 → 0, as 𝑘 → ∞.
In particular there is an integer 𝑁 > 0 such that
|𝛼𝑘 |𝑟 𝑘 ≤ 1, (𝑘 > 𝑁).
Let 𝑀 be the largest number among
|𝛼0 |, |𝛼1 |𝑟, |𝛼1 |𝑟, ⋯ |𝛼𝑁 |𝑟 𝑁 , 1.
Then clearly (36) is valid for this 𝑀.
We now consider the equation
𝐿(𝑦) = 𝑦 ′′ + 𝑎(𝑥)𝑦 ′ + 𝑏(𝑥)𝑦 = 0, (37)
where 𝑎 and 𝑏 are functions having expansions
𝑎(𝑥) = ∑∞ 𝑘=0 𝛼𝑘 𝑥 , 𝑏(𝑥) = ∑𝑘=0 𝛽𝑘 𝑥 , (38)
𝑘 ∞ 𝑘

which converge for |𝑥| < 𝑟0 for some 𝑟0 > 0 . Given any constants 𝑎1 , 𝑎2 we want to
produce a solution 𝜙 of (37) satisfying
𝜙(0) = 𝑎1 , 𝜙1′ (0) = 𝑎2
and which can be written in the form
𝜙(𝑥) = ∑∞ 𝑘=0 𝑐𝑘 𝑥 , (39)
𝑘

where the series converges for |𝑥| < 𝑟0 . If this series is convergent, we must have

S25013: Ordinary Differential Equations Page 135


𝑐0 = 𝑎1 , 𝑐1 = 𝑎2 ,
and the constants 𝑐𝑘 (𝑘 ≥ 2) must satisfy a recursion relation, which we now compute. We
have
𝜙 ′ (𝑥) = ∑∞ 𝑘=0(𝑘 + 1)𝑐𝑘+1 𝑥 ,
𝑘

and
𝜙 ′′ (𝑥) = ∑∞𝑘=0(𝑘 + 2)(𝑘 + 1)𝑐𝑘+2 𝑥 . (40)
𝑘

Now from (38), we obtain


𝑎(𝑥)𝜙 ′ (𝑥) = (∑∞ 𝑘 ∞
𝑘=0 𝛼𝑘 𝑥 )(∑𝑘=0(𝑘 + 1)𝑐𝑘+1 𝑥 )
𝑘

= ∑∞ 𝑘 𝑘
𝑘=0(∑𝑗=0 𝛼𝑘−𝑗 (𝑗 + 1)𝑐𝑗+1 )𝑥 , (41)
and
𝑏(𝑥)𝜙(𝑥) = (∑∞ 𝑘 −∞ 𝑘
𝑘=0 𝛽𝑘 𝑥 )(∑𝑘=0 𝑐𝑘 𝑥 )
= ∑∞ 𝑘 𝑘
𝑘=0(∑𝑗=0 𝛽𝑘−𝑗 𝑐𝑗 )𝑥 . (42)
Adding (40), (41), and (42),we get
𝑘 𝑘
𝐿(𝑦) = ∑∞ 𝑘
𝑘=0[(𝑘 + 2)(𝑘 + 1)𝑐𝑘+2 + ∑𝑗=0 𝛼𝑘−𝑗 (𝑗 + 1)𝑐𝑗+1 + ∑𝑗=0 𝛽𝑘−𝑗 𝑐𝑗 ] 𝑥 = 0 .
Thus the 𝑐𝑘 must satisfy
(𝑘 + 2)(𝑘 + 1)𝑐𝑘+2 = − ∑𝑘𝑗=0[𝛼𝑘−𝑗 (𝑗 + 1)𝑐𝑗+1 + 𝛽𝑘−𝑗 𝑐𝑗 ] , (𝑘 = 0, 1, 2, ⋯ ). (43)
Our job now is to show that if the 𝑐𝑘 for 𝑘 ≥ 2, are defined by (43), then the series
∑∞𝑘=0 𝑐𝑘 𝑥 , (44)
𝑘

is convergent for |𝑥| < 𝑟0 . To do this we make use of the two results concerning power
series we mentioned earlier. Let 𝑟 be any number satisfying 0 < 𝑟 < 𝑟0 .Since the series in
(38) are convergent for |𝑥| = 𝑟, we have a constant 𝑀 > 0 such that
|𝛼𝑗 |𝑟 𝑗 ≤ 𝑀, |𝛽𝑗 |𝑟 𝑗 ≤ 𝑀, (𝑗 = 0, 1, 2, ⋯ ).
Using this in (43), we find that
(𝑘 + 2)(𝑘 + 1)|𝑐𝑘+2 | ≤ ∑𝑘𝑗=0[|𝛼𝑘−𝑗 |(𝑗 + 1)|𝑐𝑗+1 | + |𝛽𝑘−𝑗 | |𝑐𝑗 |]
𝑀 𝑀
≤ ∑𝑘𝑗=0 [𝑟 𝑘−𝑗 (𝑗 + 1)|𝑐𝑗+1 | + 𝑟 𝑘−𝑗 |𝑐𝑗 |]
𝑀
= 𝑟 𝑘 ∑𝑘𝑗=0[(𝑗 + 1)|𝑐𝑗+1 | + |𝑐𝑗 |]𝑟 𝑗
𝑀
≤ 𝑟 𝑘 ∑𝑘𝑗=0[(𝑗 + 1)|𝑐𝑗+1 | + |𝑐𝑗 |]𝑟 𝑗 + 𝑀|𝑐𝑘+1 |𝑟. (45)
Now let us define
𝐶0 = |𝑐0 |, 𝐶1 = |𝑐1 |,
and 𝐶𝑘 for 𝑘 ≥ 2 by
𝑀
(𝑘 + 2)(𝑘 + 1)𝐶𝑘+2 = 𝑘 ∑𝑘𝑗=0[(𝑗 + 1)𝐶𝑗+1 + 𝐶𝑗 ]𝑟 𝑗 + 𝑀𝐶𝑘+1 𝑟. (46)
𝑟
Comparing (46) with (45), we see that an induction yields
|𝑐𝑘 | ≤ 𝐶𝑘 , 𝐶𝑘 ≥ 0, (𝑘 = 0, 1, 2, ⋯ ). (47)
We now investigate for what 𝑥 the series
∑∞𝑘=0 𝐶𝑘 𝑥
𝑘
(48)
is convergent. From (46) we find that
𝑀
(𝑘 + 1)(𝑘)𝐶𝑘+1 = 𝑘−1 ∑𝑘−1 𝑗
𝑗=0 [(𝑗 + 1)𝐶𝑗+1 + 𝐶𝑗 ]𝑟 + 𝑀𝐶𝑘 𝑟,
𝑟
and
𝑀
(𝑘)(𝑘 − 1)𝐶𝑘 = 𝑘−2 ∑𝑘−2 𝑗
𝑗=0 [(𝑗 + 1)𝐶𝑗+1 + 𝐶𝑗 ]𝑟 + 𝑀𝐶𝑘−1 𝑟,
𝑟

S25013: Ordinary Differential Equations Page 136


for large 𝑘. From these expressions, we obtain
𝑀 𝑀
𝑟(𝑘 + 1)𝑘𝐶𝑘+1 = 𝑟 𝑘−2 ∑𝑘−2 𝑗
𝑗=0 [(𝑗 + 1)𝐶𝑗+1 + 𝐶𝑗 ]𝑟 + 𝑟 𝑘−2 [𝑘𝐶𝑘 + 𝐶𝑘−1 ]𝑟
𝑘−1
+ 𝑀𝐶𝑘 𝑟 2
= 𝑘(𝑘 − 1)𝐶𝑘 − 𝑀𝐶𝑘−1 𝑟 + 𝑀[𝑘𝐶𝑘 + 𝐶𝑘−1 ]𝑟 + 𝑀𝐶𝑘 𝑟 2
= 𝑘(𝑘 − 1)𝐶𝑘 − 𝑀𝐶𝑘−1 𝑟 + 𝑀𝑘𝐶𝑘 𝑟 + 𝑀𝐶𝑘−1 𝑟 + 𝑀𝐶𝑘 𝑟 2
= 𝑘(𝑘 − 1)𝐶𝑘 + 𝑀𝑘𝐶𝑘 𝑟 + 𝑀𝐶𝑘 𝑟 2
= [𝑘(𝑘 − 1) + 𝑀𝑘𝑟 + 𝑀𝑟 2 ]𝐶𝑘 .
Hence
𝐶𝑘+1 𝑥 𝑘+1 [𝑘(𝑘−1)+𝑀𝑘𝑟+𝑀𝑟 2 ]
| |= |𝑥|
𝐶𝑘 𝑥𝑘 𝑟(𝑘+1)𝑘
1 𝑀𝑟 𝑀𝑟2
𝑘 2 [(1− )+ + 2 ]
𝑘 𝑘 𝑘
= |𝑥|
𝑟(𝑘+1)𝑘
1 𝑀𝑟 𝑀𝑟2
𝑘 2 [(1− )+ + 2 ]
𝑘 𝑘 𝑘
= 2 1 |𝑥|
𝑟𝑘 (1+ )
𝑘
1 𝑀𝑟 𝑀𝑟2
[(1− )+ + 2 ] |𝑥|
.
𝑘 𝑘 𝑘
= 1
(1+ ) 𝑟
𝑘
1 𝑀𝑟 𝑀𝑟2
𝐶𝑘+1 𝑥 𝑘+1 [(1− )+ + 2 ] |𝑥| |𝑥|
as 𝑘 → ∞.
𝑘 𝑘 𝑘
⟹| |= 1 →
𝐶𝑘 𝑥𝑘 (1+ ) 𝑟 𝑟
𝑘

Thus, by the ration test, the series (48) converges for |𝑥| < 𝑟. This implies that the series
(44) converges for |𝑥| < 𝑟 , and since 𝑟 was any number satisfying 0 < 𝑟 < 𝑟0 , we have
shown at last the series (44) converges for |𝑥| < 𝑟0 . This completes our justification of
Theorem 1.

SOLVED PROBLEMS 01
Problem 02-04-01:
Find two linearly independent solutions of 𝑦′′ + 3𝑥 2 𝑦′ − 𝑥𝑦 = 0
Solution: Given equation is
𝑦′′ + 3𝑥 2 𝑦′ − 𝑥𝑦 = 0. (49)
Here 𝑎1 (𝑥) = 3𝑥 , 𝑎2 (𝑥) = −𝑥 and both are analytic for all 𝑥0 . This shows that the
2

solutions are also analytic at the ordinary point 𝑥0 = 0. Therefore, we assume


𝑦 = ∑∞ 𝑘=0 𝑐𝑘 𝑥
𝑘
(50)
be the power series solution of the equation (49). As usual we find 𝑦′ and 𝑦′′ and
substituting in the equation (49), we get
2𝑐2 + (6𝑐3 − 𝑐0 )𝑥 + ∑∞ 𝑘
𝑘=2 [𝑐𝑘+2 (𝑘 + 1)(𝑘 + 2) + 𝑐𝑘−1 (3𝑘 − 4)]𝑥 = 0.
This is true if and only if all the coefficients of the power series of 𝑥 are zero.
𝑐 4−3𝑘
0
𝑐2 = 0, 𝑐3 = 2⋅3 and 𝑐𝑘+2 = (𝑘+1)(𝑘+2) 𝑐𝑘−1 ∀ 𝑘 ≥ 2 (51)
This is the recurrence relation. By giving the values to 𝑘 = 2, 3, ⋯, we obtain respectively,
−2 −8 (−1)2 2⋅11 (−1)2 8⋅17
𝑐4 = 3⋅4 𝑐1 , 𝑐5 = 0, 𝑐6 = 2⋅3⋅5⋅6 𝑐0 , 𝑐7 = 𝑐1 , 𝑐8 = 0, 𝑐9 = 3⋅4⋅6⋅7⋅9⋅11 𝑐1 , ⋯
3⋅4⋅6⋅7
Looking at this pattern, we find in general
(−1)𝑘+1 𝑐0 8⋅17⋯(9𝑘−10) (−1)𝑘 𝑐1 2⋅11⋅11⋅20⋯(9𝑘−7)
𝑐3𝑘 = , and 𝑐3𝑘+1 = , ∀ 𝑘 = 1,2, ⋯
2⋅3⋅5⋅6⋅8⋅9⋯(3𝑘−1)3𝑘 3⋅4⋅6⋅7⋅9⋅11⋯(3𝑘)(3𝑘+1)

S25013: Ordinary Differential Equations Page 137


and all 𝑐3𝑘+2 = 0 ∀ 𝑘 = 1,2, ⋯.
Substituting these values in the solution (50), we get
𝑥3 8𝑥 6 2𝑥 4 (−1)2 2⋅11 7
𝜙(𝑥) = 𝑐0 [1 + 2⋅3 − 2⋅3⋅5⋅6 + ⋯ ] + 𝑐1 [𝑥 − + x +⋯]
3⋅4 3⋅4⋅6⋅7
(−1)𝑘+1 8⋅17⋯(9𝑘−10) (−1)𝑘 2⋅11⋅20⋯(9𝑘−7)
⟹ 𝜙(𝑥) = 𝑐0 [1 + ∑∞
𝑘=1 𝑥 3𝑘 ] + 𝑐1 [𝑥 + ∑∞
𝑘=1 𝑥 3𝑘+1 ]
2⋅3⋅5⋅6⋅8⋅9⋯(3𝑘−1)(3𝑘) 3⋅4⋅6⋅7⋅9⋅11⋯3𝑘(3𝑘+1)
or 𝜙(𝑥) = 𝑐0 𝜙1 (𝑥) + 𝑐1 𝜙2 (𝑥),
(−1)𝑘+1 8⋅17⋯(9𝑘−10)
where 𝜙1 (𝑥) = 1 + ∑∞
𝑘=1 𝑥 3𝑘
2⋅3⋅5⋅6⋅7⋅9⋅11⋯3𝑘(3𝑘+1)
(−1)𝑘 2⋅11⋅20⋯(9𝑘−7)
and 𝜙2 (𝑥) = 𝑥 + ∑∞
𝑘=1 𝑥 3𝑘+1
3⋅4⋅6⋅7⋅9⋅11⋯3𝑘(3𝑘+1)
are the two solutions of (49). We notice that the Wronskian of the solution at 𝑥 = 0 is given
by
1 0
𝑊(𝜙1 , 𝜙2 )(0) = | | ≠ 0.
0 1
⟹ the solutions 𝜙1 (𝑥) and 𝜙2 (𝑥) are linearly independent.
Problem 02-04-02:
Compute the solution 𝜙 of 𝑦′′′ − 𝑥𝑦 = 0 which satisfies 𝜙(0) = 1, 𝜙′(0) = 0, 𝜙′′(0) = 0.
Solution: For the given differential equations, we see that 𝑎1 (𝑥) = 0, 𝑎2 (𝑥) = 0, 𝑎3 (𝑥) =
𝑥 are all analytic for all real 𝑥0 = 0. This shows that the solutions are also analytic.
Let
𝜙(𝑥) = ∑∞ 𝑘=0 𝑐𝑘 𝑥
𝑘
(52)
be the power series solution of the equation 𝑦′′′ − 𝑥𝑦 = 0.
Differentiating equation (52) thrice, we get
𝜙′′′(𝑥) = ∑∞𝑘=3 𝑐𝑘 𝑘(𝑘 − 1)(𝑘 − 2)𝑥
𝑘−3

Substituting this in the given equation and collecting the terms of the like powers of 𝑥 we
get
6𝑐3 + ∑∞𝑘=0 [𝑐𝑘+4 (𝑘 + 2)(𝑘 + 3)(𝑘 + 4) − 𝑐𝑘 ]𝑥
𝑘+1
= 0.
Now equating to zero, all the powers of 𝑥 we get
𝑐
𝑐3 = 0 and 𝑐𝑘+4 = (𝑘+2)(𝑘+3)(𝑘+4)
𝑘
, ∀ 𝑘≥0 (53)
This is the recurrence relation. By giving differet values to 𝑘 = 0,1,2, ⋯, we have the values
of all constants and are given by, 𝑐3 = 0 and
𝑐0 2𝑐1 2⋅3 5𝑐0
𝑐4 = , 𝑐5 = , 𝑐6 = 𝑐2 , 𝑐7 = 0, 𝑐8 = ,⋯
4! 5! 6! 8!
Substituting these constants in the solution,
𝜙(𝑥) = 𝑐0 + 𝑐1 𝑥 + 𝑐2 𝑥 2 + 𝑐3 𝑥 3 + 𝑐4 𝑥 4 + 𝑐5 𝑥 5 + 𝑐6 𝑥 6 + ⋯.
𝑥4 5 5⋅9
We get, 𝜙(𝑥) = 𝑐0 [1 + + 8! 𝑥 8 + 12! 𝑥12 + ⋯ ]
4!
2 2⋅6 2⋅6⋅10
+𝑐1 [𝑥 + 5! 𝑥 5 + 𝑥9 + 𝑥13 + ⋯ ]
9! 13!
2⋅3 2⋅3⋅7
+𝑐3 [𝑥 2 + 𝑥6 + 𝑥10 + ⋯ ].
6! 10!
Looking at the pattern of the values of ck , we write this as
1⋅5⋅9⋯(4𝑘−3) 4𝑘 2⋅6⋅10⋯(4𝑘−2)
𝜙(𝑥) = 𝑐0 [1 + ∑∞
𝑘=1 (4𝑘)!
𝑥 ] +𝑐1 [𝑥 + ∑∞
𝑘=1 𝑥 4𝑘+1 ]
(4𝑘+1)!

S25013: Ordinary Differential Equations Page 138


3⋅7⋅11⋯(4𝑘−1)
+𝑐2 [𝑥 2 + 2 ∑∞
𝑘=1 𝑥 4𝑘+2 ]. (54)
(4𝑘+2)!
This is the required solution. This solution satisfies
𝜙(0) = 1 ⟹ 𝑐0 = 1
𝜙 ′ (0) = 0 ⟹ 𝑐1 = 0,
𝜙′′(0) = 0 ⟹ 𝑐2 = 0
Hence the solution (54) becomes
5⋅9⋯(4𝑘−3)
𝜙(𝑥) = 1 + ∑∞
𝑘=1 𝑥 4𝑘 .
(4𝑘)!
Problem 02-04-03:
1 𝑑𝑛
Show that if 𝑃𝑛 (𝑥) = 2𝑛 𝑛! 𝑑𝑥 𝑛 (𝑥 2 − 1)𝑛 , then 𝑃𝑛 (−𝑥) = (−1)𝑛 𝑃𝑛 (𝑥), 𝑃𝑛 (1) = 1, ∀ 𝑛 ∈ ℕ and
𝑃𝑛 (−1) = (−1)𝑛 .
Solution: The Legendre’s polynomial is given by
1 𝑑𝑛
𝑃𝑛 (𝑥) = 2𝑛 𝑛! 𝑑𝑥 𝑛 (𝑥 2 − 1)𝑛 . (55)
1
⟹ 𝑃𝑛 (𝑥) = 2𝑛𝑛! 𝐷𝑛 [(𝑥 − 1)𝑛 (𝑥 + 1)𝑛 ].
Using the formula (31), we find
1
𝑃𝑛 (𝑥) = 2𝑛 𝑛! [(𝑥 − 1)𝑛 𝐷𝑛 (𝑥 + 1)𝑛 +𝑛 𝒞1 𝐷(𝑥 − 1)𝑛 𝐷𝑛−1 (𝑥 + 1)𝑛 + ⋯ + (𝑥 + 1)𝑛 𝐷𝑛 (𝑥 − 1)𝑛 ] (56)
Since 𝐷𝑛 (𝑥 + 1)𝑛 = 𝑛!
𝐷 𝑛−1 (𝑥 + 1)𝑛 = 𝑛(𝑛 − 1) … 2(𝑥 + 1) = 𝑛! (𝑥 + 1)
𝐷 𝑛−2 (𝑥 + 1)𝑛 = 𝑛(𝑛 − 1) … 3(𝑥 + 1)2 and so on.
Hence equation (56) becomes
1 𝑛(𝑛+1)
𝑃𝑛 (𝑥) = 2𝑛 𝑛! [(𝑥 − 1)𝑛 𝑛! + 𝑛2 (𝑥 − 1)𝑛−1 𝑛! (𝑥 + 1) + (𝑥(𝑛 − 1)(𝑥 − 1)𝑛−2 )
2
1
+ 2 𝑛! (𝑥 + 1)2 + ⋯ + (𝑥 + 1)𝑛 𝑛!] (57)
Since each term except the first in the above equation contains (𝑥 + 1). If we put 𝑥 = −1
in (57) all terms become zero except the first. This gives
1
𝑝𝑛 (−1) = 2𝑛𝑛! (−2)𝑛 𝑛!
⟹ 𝑃𝑛 (−1) = (−1)𝑛 (58)
Similarly each term except the last of the equation (57) contains (𝑥 − 1). Hence if we put
𝑥 = 1 in the equation (57) all terms become zero, except the last term in the bracket.
1
𝑃𝑛 (𝑥) = 2𝑛 𝑛! ⟹ 𝑃𝑛 (1) = 1. (59)
Now replacing 𝑥 by −𝑥 in the equation (57) we get
1
𝑃𝑛 (−𝑥) = 2𝑛𝑛! [(−1)𝑛 (𝑥 + 1)𝑛 𝑛! + 𝑛2 𝑛! (−1)𝑛−1 (𝑥 + 1)𝑛−1 (𝑥 − 1) + ⋯ + (−1)𝑛 (𝑥 − 1)𝑛 𝑛!]
1
⟹ 𝑃𝑛 (−𝑥) = 2𝑛𝑛! (−1)𝑛 [(𝑥 + 1)𝑛 𝑛! + 𝑛2 𝑛! (𝑥 + 1)𝑛−1 (𝑥 − 1) + ⋯ + (𝑥 − 1)𝑛 𝑛!]
⟹ 𝑃𝑛 (−𝑥) = (−1)𝑛 𝑃𝑛 (𝑥). (60)
Problem 02-04-04:
1
Show that ∫−1 𝑃𝑛 (𝑥)𝑃𝑚 (𝑥)𝑑𝑥 = 0 if 𝑛 ≠ 𝑚 in ℕ.
Solution: We know the Legendre polynomial 𝑃𝑛 (𝑥) satisfies
(1 − 𝑥 2 )𝑦′′ − 2𝑥𝑦′ + 𝑛(𝑛 + 1)𝑦 = 0 (61)

S25013: Ordinary Differential Equations Page 139


⟹ (1 − 𝑥 2 )𝑃′′𝑛 (𝑥) − 2𝑥𝑃′𝑛 (𝑥) + 𝑛(𝑛 + 1)𝑃𝑛 (𝑥) = 0. (62)
Similarly 𝑃𝑚 (𝑥) also satisfies the equation (61)
(1 − 𝑥 2 )𝑃′′ 𝑚 (𝑥) − 2𝑥𝑃′ 𝑚 (𝑥) + 𝑚(𝑚 + 1)𝑃𝑚 (𝑥) = 0. (63)
Multiply equation (62) by 𝑃𝑚 (𝑥) and equation (63) by 𝑃𝑛 (𝑥) and subtracting the results, we
get
(1 − 𝑥 2 )[𝑃′′ 𝑛 (𝑥)𝑃𝑚 (𝑥) − 𝑃′′ 𝑚 (𝑥)𝑃𝑛 (𝑥)] − 2𝑥[𝑃′ 𝑛 (𝑥)𝑃𝑚 (𝑥) − 𝑃′ 𝑚 (𝑥)𝑃𝑛 (𝑥)]
+[𝑛(𝑛 + 1) − 𝑚(𝑚 + 1)] 𝑃𝑛 (𝑥)𝑃𝑚 (𝑥) = 0.
This is nothing but
𝑑
[(1 − 𝑥 2 )(𝑃′ 𝑛 (𝑥)𝑃𝑚 (𝑥) − 𝑃′ 𝑚 (𝑥)𝑃𝑛 (𝑥))] = [𝑚(𝑚 + 1) − 𝑛(𝑛 + 1)]𝑃𝑛 𝑃𝑚
𝑑𝑥
Intgerating this between the limits −1 to 1 we get
1
[𝑚(𝑚 + 1) − 𝑛(𝑛 + 1)] ∫−1 𝑃𝑛 𝑃𝑚 𝑑𝑥 = [(1 − 𝑥 2 )(𝑃′ 𝑛 (𝑥)𝑃𝑚 (𝑥) − 𝑃′ 𝑚 (𝑥)𝑃𝑛 (𝑥))]1−1 = 0
1
⟹ (𝑚 − 𝑛)(𝑚 + 𝑛 + 1) ∫−1 𝑃𝑛 (𝑥)𝑃𝑚 (𝑥)𝑑𝑥 = 0
1
⟹ ∫−1 𝑃𝑛 (𝑥)𝑃𝑚 (𝑥)𝑑𝑥 = 0 for 𝑛 ≠ 𝑚.
This is known as orthogonality property of Legendre polynomials.
Problem 02-04-05:
1 2
Show that ∫−1 𝑃𝑛2 (𝑥)𝑑𝑥 = 2𝑛+1.
Solution: We know the Legendre polynomial 𝑃𝑛 (𝑥) is given by
1 𝑑𝑛
𝑃𝑛 (𝑥) = 2𝑛 𝑛! 𝑑𝑥 𝑛 (𝑥 2 − 1)𝑛 .
1 1
Let 𝑢(𝑥) = (𝑥 2 − 1)𝑛 . Hence 𝑃𝑛 (𝑥) = 2𝑛𝑛! 𝐷𝑛 𝑢(𝑥) = 2𝑛𝑛! 𝑢(𝑛) (𝑥), (64),
where 𝐷𝑛 𝑢(𝑥) = 𝐷𝑛 [(𝑥 − 1)𝑛 (𝑥 + 1)𝑛 ].
Then from equation (57), we have
𝐷 𝑛 𝑢(𝑥)|𝑥=1 = 2𝑛 𝑛!
i.e. 𝑢(𝑛) (1) = 2𝑛 𝑛! and 𝑢(𝑘) (−1) = 𝑢(𝑘) (1) = 0 ∀ 𝑘 < 𝑛. (65)
1
Consider now the integral ∫−1 𝑢(𝑛) (𝑥)𝑢(𝑛) (𝑥)𝑑𝑥. Integrating by parts, we get
1 1
∫−1 𝑢(𝑛) (𝑥)𝑢(𝑚) (𝑥)𝑑𝑥 = 𝑢(𝑛) (𝑥)𝑢(𝑛−1) (𝑥)]1−1 − ∫−1 𝑢(𝑛+1) 𝑢(𝑛−1) (𝑥)𝑑𝑥
Using the equation (65) the first term in the right hand side vanishes at both the limits. We
have therefore
1 1
∫−1 𝑢(𝑛) (𝑥)𝑢(𝑛) (𝑥)𝑑𝑥 = − ∫−1 𝑢(𝑛+1) (𝑥)𝑢(𝑛−1) (𝑥)𝑑𝑥.
Integrating it by parts and using the equation (65), we get
1 1
∫−1 𝑢(𝑛) (𝑥)𝑢(𝑛) (𝑥)𝑑𝑥 = (−1)2 ∫−1 𝑢(𝑛+2) (𝑥)𝑢(𝑛−2) (𝑥)𝑑𝑥.
Continuing this way 𝑛 times we get
1 1
∫−1 𝑢(𝑛) (𝑥)𝑢(𝑛) (𝑥)𝑑𝑥 = (−1)𝑛 ∫−1 𝑢(2𝑛) (𝑥)𝑢(𝑥)𝑑𝑥. (66)
Note the 0th derivative of 𝑢(𝑥) is 𝑢 i.e. 𝑢(0) = 𝑢(𝑥) and it does not mean 𝑢(𝑥) raised to
zero.
Using the equation (64) in the equation (66), we get
1 1
∫−1 (2𝑛 𝑛!)2 𝑃𝑛2 (𝑥)𝑑𝑥 = (−1)𝑛 ∫−1 𝑢(2𝑛) (𝑥)𝑢(𝑥)𝑑𝑥

S25013: Ordinary Differential Equations Page 140


1 (−1)𝑛 1
⟹ ∫−1 𝑃𝑛2 (𝑥)𝑑𝑥 = (2𝑛𝑛!)2 ∫−1 𝑢(2𝑛) (𝑥)𝑢(𝑥)𝑑𝑥, (67)
where 𝑢(2𝑛) (𝑥) = [(𝑥 − 1)𝑛 (𝑥 + 1)𝑛 ](2𝑛) .
Using the formula (31), the only non-vanishing term in the above expansion is
2𝑛
𝐶𝑛 [(𝑥 + 1)𝑛 ](𝑛) [(𝑥 − 1)𝑛 ](𝑛)
(2𝑛)! 𝑑𝑛 𝑑𝑛
i.e. 𝑢(2𝑛) (𝑥) = (𝑛!)2 𝑑𝑥 𝑛 (𝑥 + 1)𝑛 𝑑𝑥 𝑛 (𝑥 − 1)𝑛 ,
where we have
𝑑𝑛 𝑑𝑛
(𝑥 + 1)𝑛 = 𝑛! and (𝑥 − 1)𝑛 = 𝑛!
𝑑𝑥 𝑛 𝑑𝑥 𝑛
Thus 𝑢(2𝑛) (𝑥) = (2𝑛)!. (68)
Hence, the equation (67) becomes
1 (−1)n (2𝑛)! 1
∫−1 𝑃𝑛2 (𝑥)𝑑𝑥 = (2𝑛 𝑛!)2
∫−1 (𝑥 2 − 1)𝑛 𝑑𝑥
(−1)2𝑛 (2𝑛)! 1
= (2𝑛 𝑛!)2
∫−1 (1 − 𝑥 2 )𝑛 𝑑𝑥
1 (2𝑛)! 1
⟹ ∫−1 𝑃𝑛2 (𝑥)𝑑𝑥 = (2𝑛𝑛!)2 ∫−1 (1 − 𝑥 2 )𝑛 𝑑𝑥. (69)
1 1
Consider ∫−1 (1 − 𝑥 2 )𝑛 𝑑𝑥 = 2 ∫0 (1 − 𝑥 2 )𝑛 𝑑𝑥.
𝜋
Put 𝑥 = sin 𝜃 , d𝑥 = cos 𝜃 d𝜃 when 𝑥 = 0 ⟹ 𝜃 = 0, 𝑥 = 1 ⟹ 𝜃 = 2
𝜋
1
Therefore 2 ∫0 (1 − 𝑥 2 )𝑛 𝑑𝑥 = 2 ∫0 cos 2𝑛+1 𝜃𝑑𝜃 2

𝜋
1 𝑚+1 𝑛+1
Using the formula ∫02 sin𝑚 𝜃cos𝑛 𝜃𝑑𝜃 = 2 𝛽 ( , ), we find
2 2
1 1 1
2 ∫0 (1 − 𝑥 2 )𝑛 𝑑𝑥 = 2 ⋅ 2 𝛽 (2 , 𝑛 + 1)
Γ𝑝,Γ𝑞
Also we have 𝛽(𝑝, 𝑞) = Γ(𝑝+𝑞).
This gives
1
1 Γ Γ(𝑛+1)
2 ∫0 (1 − 𝑥 2 )𝑛 𝑑𝑥 = 2
2𝑛+3
Γ( )
2
1
1 Γ 𝑛!
⟹ 2 ∫0 (1 − 𝑥 ) 𝑑𝑥 =2 𝑛 2
2𝑛+3 .
Γ
2
2𝑛+3 2𝑛+1 2𝑛+1
We have Γ =( )Γ
2 2 2
2𝑛+3 2𝑛+1 2𝑛−1 2𝑛−3 3 1 1
⟹Γ =( )( )( ) ⋯ 2 ⋅ 2 Γ 2.
2 2 2 2
1
1 Γ 𝑛!
Therefore we have 2 ∫0 (1 − 𝑥 2 )𝑛 𝑑𝑥 = 1 3 2
2𝑛−3 2𝑛−1 2𝑛+1 1 .
⋅ ⋯( )( )( Γ )
22 2 2 2 2

We write this by multiplying both numerator and denominator of the right had side of the
above equation by 1 ⋅ 2 ⋅ 3 ⋯ 𝑛 we get
1 (𝑛!)2
2 ∫0 (1 − 𝑥 2 )𝑛 𝑑𝑥 = 1 2 3 5 6 2𝑛−3 2𝑛−2 2𝑛−1 2𝑛 2𝑛+1
⋅ ⋅ ⋅ ⋅ ⋅( )⋅( )( )⋅ ⋅( )
22222 2 2 2 2 2
(𝑛!)2 22𝑛+1
= (2𝑛+1)!
2⋅22𝑛 (𝑛!)2
= (2𝑛+1)!

S25013: Ordinary Differential Equations Page 141


1 2(2𝑛 𝑛!)2
⟹ 2 ∫0 (1 − 𝑥 2 )𝑛 𝑑𝑥 = (70)
(2𝑛+1)!
Substituting this in the equation (69) we get
1 2(2𝑛)! (2𝑛 𝑛!)2
∫−1 𝑃𝑛2 (𝑥)𝑑𝑥 = (2𝑛𝑛!)2 ⋅ (2𝑛+1)!
2(2𝑛)!
= (2𝑛+1)!
1 2
∫−1 𝑃𝑛2 (𝑥)𝑑𝑥 = 2𝑛+1.

SELF-TEST 01
MCQ 02-04-01
Consider the following two statements:
I. If 𝑛 is non-negative even integer, then any polynomial solution of Legendre equation
which has only even powers of 𝑥, is a multiple of Legendre polynomial 𝑃𝑛 (𝑥).
II. If 𝑛 is positive odd integer, then any polynomial solution of Legendre equation which
has only odd powers of 𝑥, is a multiple of Legendre polynomial 𝑃𝑛 (𝑥).
Then
A. only (I) is true
B. only (II) is true
C. both (I) and (II) are true
D. neither (I) or nor (II) is true.
MCQ 02-04-02
𝑑𝑛
For the 𝑛𝑡ℎ −Legendre polynomial 𝑐𝑛 𝑑𝑥 𝑛 (𝑥 2 − 1)𝑛 , the value of 𝑐𝑛 is
1
A. 2𝑛 𝑛!
𝑛!
B. 2𝑛
C. 2 𝑛!
𝑛

2𝑛
D. .
𝑛!
MCQ 02-04-03
The Legendre equation (1 − 𝑥 2 )𝑦 ′′ − 2𝑥𝑦 ′ + 𝑛(𝑛 + 1)𝑦 = 0 on the interval [0,1] has
A. both 0 and 1 are regular singular points
B. both 0 and 1 are regular points
C. a regular singular point at 0 and a regular point 1
D. a regular point at 0 and a regular singular point 1.

SHORT ANSWER QUESTIONS 01


SAQ 02-04-01
2 𝑑𝑛 2
Show that 𝐻𝑛 (𝑥) = (−1)𝑛 𝑒 𝑥 (𝑒 −𝑥 ) is a solution of the equation 𝑦 ′′ − 2𝑥𝑦 ′ + 2𝑛𝑦 = 0,
𝑑𝑥 𝑛
𝑛 =constant.
2
Solution: Let 𝑢(𝑥) = 𝑒 −𝑥 . (71)
S25013: Ordinary Differential Equations Page 142
Differentiating (71) with respect to 𝑥, we get
𝑑𝑢 2
= −2𝑥𝑒 −𝑥 ,
𝑑𝑥
𝑑𝑢
or + 2𝑥𝑢(𝑥) = 0. (72)
𝑑𝑥
Using the formula given in the equation (31), we differentiate equation (72) 𝑛 times we
get
𝑑𝑛+1 𝑢 𝑑𝑛 𝑢 𝑑𝑛−1 𝑢
+ 2𝑥 𝑑𝑥 𝑛 + 2𝑛 𝑑𝑥 𝑛−1 = 0. (73)
𝑑𝑥 𝑛+1
2
Multiplying the equation (73) by (−1)𝑛 𝑒 𝑥 , we get
2 𝑑𝑛+1 𝑢 2 𝑑𝑛 𝑢 2 𝑑𝑛−1 𝑢
(−1)𝑛 𝑒 𝑥 + 2𝑥(−1)𝑛 𝑒 𝑥 + 2𝑛(−1)𝑛 𝑒 𝑥 = 0. (74)
𝑑𝑥 𝑛+1 𝑑𝑥 𝑛 𝑑𝑥 𝑛−1
Since it is given that
2 𝑑𝑛 𝑢
𝐻𝑛 (𝑥) = (−1)𝑛 𝑒 𝑥 . (75)
𝑑𝑥 𝑛
Using this we write equation (74) as
−𝐻𝑛+1 (𝑥) + 2𝑥𝐻𝑛 (𝑥) − 2𝑛𝐻𝑛−1 (𝑥) = 0
or 𝐻𝑛+1 (𝑥) − 2𝑥𝐻𝑛 (𝑥) + 2𝑛𝐻𝑛−1 (𝑥) = 0, 𝑛 ≥ 1 (76)
Now differentiating equation (75) with respect to 𝑥, we get
2 𝑑𝑛 𝑢 2 𝑑𝑛+1 𝑢
𝐻′𝑛 (𝑥) = (−1)𝑛 2𝑥𝑒 𝑥 + (−1)𝑛 𝑒 𝑥
𝑑𝑥 𝑛 𝑑𝑥 𝑛+1
⟹ 𝐻′𝑛 (𝑥) = 2𝑥𝐻𝑛 (𝑥) − 𝐻𝑛+1 (𝑥). (77)
Using the equation (76) in the equation (77) we get
𝐻′𝑛 (𝑥) = 2𝑛𝐻𝑛−1 (𝑥). (78)
Differentiating equation (77) with respect to 𝑥, we get
𝐻 ′′ 𝑛 (𝑥) = 2𝐻𝑛 (𝑥) + 2𝑥𝐻 ′ 𝑛 (𝑥) − 𝐻 ′ 𝑛+1 (𝑥)
⟹ 𝐻′′𝑛 (𝑥) − 2𝑥𝐻′𝑛 (𝑥) = 2𝐻𝑛 (𝑥) − 𝐻′𝑛+1 (𝑥)
Using the equation (78) for𝑛 + 1 in place of 𝑛, we eliminate the term 𝐻′𝑛+1 (𝑥), we have
𝐻′′𝑛 (𝑥) − 2𝑥𝐻′𝑛 (𝑥) = 2𝐻𝑛 (𝑥) − 2(𝑛 + 1)𝐻𝑛 (𝑥)
⟹ 𝐻′′𝑛 (𝑥) − 2𝑥𝐻′𝑛 (𝑥) + 2𝑛𝐻𝑛 (𝑥) = 0. (79)
This shows that 𝐻𝑛 (𝑥) is a solution of the equation
𝑦′′ − 2𝑥𝑦′ + 2𝑛𝑦 = 0.
SAQ 02-04-02
Find first four Legendre polynomials.
Solution: We know the Legendre polynomial 𝑃𝑛 (𝑥) is given by
1 𝑑𝑛
𝑃𝑛 (𝑥) = 2𝑛 𝑛! 𝑑𝑥 𝑛 (𝑥 2 − 1)𝑛 . (80)
So first polynomial is obtained form (80) by putting 𝑛 = 0
1 𝑑0 2
𝑃0 (𝑥) = 0 (𝑥 − 1)0 .
2 0! 𝑑𝑥 0
Note the 0th derivative of (𝑥 2 − 1)0 is (𝑥 2 − 1)0 i.e. [(𝑥 2 − 1)0 ](0) = (𝑥 2 − 1)0 and it does
not mean (𝑥 2 − 1)0 raised to zero. Therefore, we have
1
⟹ 𝑃0 (𝑥) = 1.1 (𝑥 2 − 1)0 ⟹ 𝑃0 (𝑥) = 1.1 ⟹ 𝑃0 (𝑥) = 1. This is first polynomial.
Now we put 𝑛 = 1, we have

S25013: Ordinary Differential Equations Page 143


1 𝑑1
𝑃1 (𝑥) = (𝑥 2 − 1)1
21 1! 𝑑𝑥1
1 d 2
= (𝑥 − 1)
2.1 dx
1
= (2𝑥 − 0)
2
= 𝑥.
⟹ 𝑃1 (𝑥) = 𝑥 is the second polynomial.
Similarly, we have for 𝑛 = 2
1 𝑑2
𝑃2 (𝑥) = 22 2! 𝑑𝑥 2 (𝑥 2 − 1)2
1 d2 2
= (𝑥 − 1)2
4.2 dx 2
1 𝑑
= [2(𝑥 2 − 1)2𝑥]
8 𝑑𝑥
1 𝑑
= [4(𝑥 3 − x)]
8 𝑑𝑥
1
= [4(3𝑥 2 − 1)]
8
1
= 2 (3𝑥 2 − 1).
This is the third polynomial.
Finally for 𝑛 = 3, we have
1 𝑑3
𝑃3 (𝑥) = 23 3! 𝑑𝑥 3 (𝑥 2 − 1)3
1 d3
= [(𝑥 2 − 1)3 ]
8.6 dx 3
1 𝑑
= 48 [𝑑𝑥 (𝑥 2 − 1)3 ]
1 𝑑2
= 48 𝑑𝑥 2 [3(𝑥 2 − 1)2 2𝑥]
1 𝑑2
= 48 𝑑𝑥 2 [6𝑥(𝑥 2 − 1)2 ]
6 𝑑 𝑑
= 48 [ {𝑥(𝑥 2 − 1)2 }]
𝑑𝑥 𝑑𝑥
6 𝑑
= 48 [{1. (𝑥 2 − 1)2 + 𝑥. 2(𝑥 2 − 1)2𝑥}]
𝑑𝑥
1 𝑑
=8 [(𝑥 2 − 1)2 + 4(𝑥 4 − 𝑥 2 )]
𝑑𝑥
1
= [(4𝑥 3 − 4𝑥) + (16𝑥 3 − 8𝑥)]
8
20 12
= 8 𝑥 3 − 8 𝑥.
5 3
⟹ 𝑃3 (𝑥) = 2 𝑥 3 − 2 𝑥, which is the required fourth polynomial.
SAQ 02-04-03
1 𝑑𝑛
Assume Legendre polynomial 𝑃𝑛 (𝑥) of degree 𝑛 , is given by 𝑃𝑛 (𝑥) = 2𝑛𝑛! 𝑑𝑥 𝑛 (𝑥 2 − 1)𝑛 .
𝑛(𝑛+1)
Then show that 𝑃𝑛′ (1) = .
2
Solution: We know the Legendre polynomial 𝑃𝑛 (𝑥) satisfying the Legendre equation

S25013: Ordinary Differential Equations Page 144


(1 − 𝑥 2 )𝑦 ′′ − 2𝑥𝑦 ′ + 𝑛(𝑛 + 1)𝑦 = 0
for all non-negative integer 𝑛.
⟹ (1 − 𝑥 2 )𝑃𝑛′′ (𝑥) − 2𝑥𝑃𝑛′ (𝑥) + 𝑛(𝑛 + 1)𝑃𝑛 (𝑥) = 0. (81)
Now setting 𝑥 = 1 in (81) and using the result 𝑃𝑛 (1) = 1, we have
(1 − 12 )𝑃𝑛′′ (𝑥) − 2.1𝑃𝑛′ (1) + 𝑛(𝑛 + 1)𝑃𝑛 (1) = 0
⟹ 0. 𝑃𝑛′′ (𝑥) − 2𝑃𝑛′ (1) + 𝑛(𝑛 + 1). 1 = 0
⟹ −2𝑃𝑛′ (1) + 𝑛(𝑛 + 1) = 0
⟹ −2𝑃𝑛′ (1) = −𝑛(𝑛 + 1)
⟹ 2𝑃𝑛′ (1) = 𝑛(𝑛 + 1)
𝑛(𝑛+1)
⟹ 𝑃𝑛′ (1) = .
2
This is the required result.
1)
2)

SUMMARY
Learning apart a few cases, in general linear Des with variable coefficients do not have solutions
that are expressible in closed form. In such cases, we opt for power series method which gives the
solution in terms of series.
For 𝑛 ∈ ℕ , the Legendre equation (1 − 𝑥 2 )𝑦 ′′ − 2𝑥𝑦 ′ + 𝑛(𝑛 + 1)𝑦 = 0 has a general solution
1 𝑑𝑛
𝑦(𝑥) = 𝑐1 𝑃𝑛 (𝑥) + 𝑐2 𝑄𝑛 (𝑥), where 𝑃𝑛 (𝑥) = (𝑥 2 − 1)𝑛 and 𝑃𝑛 (𝑥) has properties such as:
2𝑛 𝑛! 𝑑𝑥 𝑛
1 0, for 𝑛 ≠ 𝑚
1. ∫−1 𝑃𝑛 (𝑥)𝑃𝑚 (𝑥)𝑑𝑥 = { 2 ,
2𝑛+1
,for 𝑛 = 𝑚
2. 𝑃𝑛 (−𝑥) = (−1)𝑛 𝑃𝑛 (𝑥),
3. 𝑃𝑛 (1) = 1,
4. 𝑃𝑛 (−1) = (−1)𝑛 ,
(2𝑛)!
5. Coefficient of 𝑥 𝑛 in 𝑃𝑛 (𝑥) is 2𝑛 (𝑛!)2
.

END OF UNIT EXERCISES


1) Let 𝑢(𝑥) = (𝑥 2 − 1)𝑛 . Then find the derivative 𝑢(2𝑛) (𝑥).
2) Show that the Legendre polynomial 𝑃𝑛 (𝑥) satisfy the following recurrence relation
′ ′ (𝑥)
(𝑥) − 𝑃𝑛−1
𝑃𝑛+1 = (2𝑛 + 1)𝑃𝑛 (𝑥).
3) Express the function 𝑓(𝑥) = 5𝑥 + 𝑥 in the Legendre polynomials.
3
1
4) Prove that ∫−1 𝑥 𝑘 𝑃𝑛 (𝑥)𝑑𝑥 = 0, (𝑘 < 𝑛).
𝜋
5) Find the value of ∫02 cos 𝑥 𝑃3 (sin 𝑥)𝑑𝑥.
6) Using the fact that 𝑃0 (𝑥) = 1 is a solution of (1 − 𝑥 2 )𝑦 ′′ − 2𝑥𝑦 ′ = 0, find a second
independent solution by the method of reduction order.

KEY WORDS
Power series, radius of convergence, singular point, regular -singular point, irregular singular
point, ordinary point, Analytic function, Legendre equation, Legendre polynomial, orthogonality
property.

S25013: Ordinary Differential Equations Page 145


A NSWERS TO S ELF -T ESTS AND E ND OF U NIT E XERCISES

UNIT 01-01: LINEAR EQUATIONS WITH CONSTANT COEFFICIENTS


SELF-TEST 01
MCQ 01-01-01: A; MCQ 01-01-02: C; MCQ 01-01-03: B; MCQ 01-01-04: D.

END OF UNIT EXERCISES


1) Hint: Let 𝑟1 , 𝑟2 be roots of the characteristic polynomial 𝑝(𝑟) = 𝑟 2 + 𝑎1 𝑟 + 𝑎2 of the DE 𝑦 ′′ + 𝑎1 𝑦 ′ + 𝑎2 𝑦 =
0. We prove this result by considering two cases: Case 1: Re(𝑟1 ) < 0, Re(𝑟2 ) < 0: (roots may or not be
repeated) and Case 2: Re(𝑟1 ) = 0, Re(𝑟2 ) = 0: (roots 𝑟1 ≠ 𝑟2 ).

2) Hint: With roots 𝑟 = 𝛼, 𝛼 − 1 and the general solution 𝜙 of the given equation is given by 𝜙(𝑥) = 𝑐1 𝑟 𝛼𝑥 +
𝑐2 𝑒 (𝛼−1)𝑥 . By considering various possibilities: 𝛼 < 0, 𝛼 > 1, 0 < 𝛼 < 1, one concludes the behaviour of
solution.
3) Hint: (a) We substitute 𝜒 = 𝜙 − 𝜓 in given equation 𝐿(𝑦) = 0 and use 𝐿(𝜙) = L(𝜓) = 0 . (b) We know the
theorem of this unit, ||𝜙(𝑥0 )||𝑒 −𝑘|𝑥−𝑥0 | ≤ ||𝜙(𝑥)|| ≤ ||𝜙(𝑥0)||𝑒 𝑘|𝑥−𝑥0 | . Hence applying the to our problem,
we have ||𝜒(𝑥)|| ≤ ||𝜒(𝑥0 )||𝑒 𝑘|𝑥−𝑥0 |

UNIT 01-02: DEPENDENT AND INDEPENDENT OF SOLUTIONS


SELF-TEST 01
MCQ 01-02-01: C; MCQ 01-02-02: A; MCQ 01-02-03: B, C;
MCQ 01-02-04: C; MCQ 01-02-05: B.

END OF UNIT EXERCISES


1) Hint: Use linear combination and solve equations for scalars and check whether they zero or non-zero.
Therefore, we have (i) is LI, (ii), (iii) , (iv), and (v) .
2) Hint: Here we have two possibilities: In the first possibility, the roots are 𝑟 = ±√−𝑎2 , a2 < 0 and hence
two solutions are given by 𝜙1 = 𝑒 √−𝑎2𝑥 , 𝜙2 = 𝑒 −√−𝑎2𝑥 . Therefore, we have 𝑊(𝜙1 , 𝜙2 ) = −2√𝑎2 . On other
hand, the roots are 𝑟 = ±i√𝑎2 , √𝑎2 > 0 and the corresponding solutions 𝜙1 = cos √𝑎2 𝑥, 𝜙2 = sin √𝑎2 𝑥.
One can have 𝑊(𝜙1 , 𝜙2 ) = √𝑎2 for all 𝑥 ∈ 𝐼, where √𝑎2 is constant.
3) Hint: Using the definition of Wronskian 𝑊(𝜙1 , 𝜙2 )(𝑥) , we can show that 𝑊(𝜙1, 𝜙2 )(𝑥) = 0 ∀ 𝑥 ∈
𝐼 : −∞ < 𝑥 < ∞ and in particular, we put 𝑥 = 𝑥0 , we have 𝑊(𝜙1 , 𝜙2 )(𝑥0 ) = 0.
4) Hint: The determinant of coefficient matrix of the homogeneous system of equations 𝑐1 𝜙1 (𝑥0 ) + 𝑐2 𝜙2 (𝑥0 ) =
0 , 𝑐1 ϕ1′ (𝑥0 ) + 𝑐2 ϕ′2 (𝑥0 ) = 0 , is identically to Wronskian 𝑊(𝜙1 , 𝜙2 )(𝑥0 ) , but it is given that
𝑊(𝜙1 , 𝜙2 )(𝑥0 ) = 0. Therefore, this homogeneous system has infinitely many solutions.
.

S25013: Ordinary Differential Equations Page 146


UNIT 01-03: APPLICATIONS OF SECOND ORDER LINEAR EQUATIONS
SELF-TEST 01
MCQ 01-03-01: B; MCQ 01-03-02: A; MCQ 01-03-03: C;
MCQ 01-03-04: A; MCQ 01-03-05: A.

END OF UNIT EXERCISES


𝑑2 𝑥 𝑑𝑥
1) Hint: This is a free, damped motion and therefore its formulation is as 𝑑𝑡 2 + 4 𝑑𝑡 + 16𝑥 = 0, with initial
1
conditions 𝑥(0) = , 𝑥 ′ (0) = 0. Thus the general solution of free, damped motion equation is given by
2
𝑥(𝑡) = 𝑒 −2𝑡 (𝑐1 sin2√3𝑡 + 𝑐2 cos2√3𝑡). Applying the initial conditions, we obtain the resulting motion of the
√3 −2𝑡 𝜋
weight on the spring 𝑥(𝑡) = 3
𝑒 cos (2√3𝑡 − 6 ).
𝑑2 𝑥 𝑑𝑥
2) Hint: The basic differential equation for the motion is 𝑑𝑡 2 + 4 𝑑𝑡 + 20𝑥 = 10 cos 2𝑡 with initial conditions
𝑥(0) = 0, 𝑥′(0) = 0. We can find the complementary function and particular integral and hence the general
1 1
solution is 𝑥(𝑡) = 𝑒 −2𝑡 (𝑐1 sin4𝑡 + 𝑐2 cos4𝑡) + 2 cos2𝑡 + 4 sin2𝑡. After using initial conditions, one find the
3 1 1 1
required solution 𝑥(𝑡) = 𝑒 −2𝑡 (− 8 sin4𝑡 − 2 cos4𝑡) + 2 cos2𝑡 + 2 sin2𝑡.

UNIT 01-04: THE HOMOGENEOUS EQUATION OF HIGHER ORDER


SELF-TEST 01
MCQ 01-04-01: C; MCQ 01-04-02: B; MCQ 01-04-03: A, B, C, D; MCQ 01-04-04: D.

END OF UNIT EXERCISES


1) Hint: The roots of the polynomial equation of a certain homogeneous constant coefficient linear equation are
3, 4, 4, 4, 5 ± 2𝑖, 5 ± 2𝑖. So, looking at these roots, the basic solutions are 𝑒 3𝑥 , 𝑒 4𝑥 , 𝑥𝑒 4𝑥 , 𝑥 2 𝑒 4𝑥 ,
𝑒 5𝑥 cos 2𝑥 , 𝑒 5𝑥 sin 2𝑥, 𝑥𝑒 5𝑥 cos 2𝑥 , 𝑥𝑒 5𝑥 sin 2𝑥 . Using superposition, the general real valued solution is
𝜙(𝑥) = 𝑐1 𝑒 3𝑥 + 𝑒 4𝑥 (𝑐2 + 𝑐3 𝑥 + 𝑐4 𝑥 2 ) + 𝑒 5𝑥 [(𝑐5 + 𝑐6 𝑥) cos 2𝑥 + (𝑐7 + 𝑐8 𝑥) sin 2𝑥] and there are 8 roots,
so the order of the differential equation is 8.
2) Hint: In this case, roots are 𝑟 = ±𝑖 and hence complex-valued solutions are 𝑧1 (𝑥) = 𝑒 𝑖𝑥 , 𝑧2 (𝑡) = 𝑒 −𝑖𝑥 . The
general solution is z(𝑡) = 𝑐1 𝑒 𝑖𝑥 + c2 𝑒 −𝑖𝑥 , c1 , c2 ∈ ℂ. The Euler formula and its complex-conjugate formula,
we obtain a real-valued fundamental set as 𝜙1 (𝑥) = cos 𝑥 , 𝜙2 (𝑥) = sin 𝑥. Thus, all real-valued solutions
are given by 𝜙(𝑥) = [𝑐1 cos 𝑥 + 𝑐2 sin 𝑥 ], 𝑐1 , 𝑐2 ∈ ℝ.
3) Hint: Use definition of the Wronskian, we have
𝑥 2 3𝑥 + 2 2𝑥 + 3
2
𝑊(x , 3x + 2, 2x + 3) = |2𝑥 3 2 | = −(3𝑥 + 2)(−4) + (2𝑥 + 3)(−6) = −10.
2 0 0
4) Hint: Looking at given solution, there are trigonometric functions and one term containing 𝑥 in multiplication.
Therefore, the corresponding auxiliary equation has repeated complex conjugate roots which means that the
differential equation is order at least four and hence the equation is (𝐷2 + 6𝐷 + 13)2𝑦 = 0.

S25013: Ordinary Differential Equations Page 147


UNIT 02-01: THE NON-HOMOGENEOUS EQUATION OF HIGHER ORDER
SELF-TEST 01
MCQ 02-01-01: B; MCQ 02-01-02: C; MCQ 02-01-03: B; MCQ 02-01-04: D.

END OF UNIT EXERCISES


1) Hint: Note that the characteristic equation of the homogenous equation (𝐷 2 − 𝑎2 )𝑦 = 0 is (𝑟 2 − 𝑎2 ) = 0.
Since both 𝑎, −𝑎 are simple roots, the general solution is 𝑦(𝑥) = 𝑐1 𝑒 𝑎𝑥 + 𝑐2 𝑒 −𝑎𝑥 or 𝑦(𝑥) = 𝑐3 cosh 𝑎𝑥 +
𝑐4 sinh 𝑎𝑥, where 𝑐3 = 𝑐1 + 𝑐2 , 𝑐4 = 𝑐1 − 𝑐2 are constants ( 𝑒 𝑎𝑥 = cosh 𝑎𝑥 + sinh 𝑎𝑥 , 𝑒 −𝑎𝑥 = cosh 𝑎𝑥 −
sinh 𝑎𝑥 ). Hence, all functions cosh 𝑎𝑥 , sinh 𝑎𝑥 are obtained the general solution of the equation 𝑀(𝑦) = 0
with 𝑀 = (𝐷 2 − 𝑎2 ) and are annihilated by the operator 𝑀 = (𝐷 2 − 𝑎2 ). This proves that both functions
cosh 𝑎𝑥 , sinh 𝑎𝑥 are solutions of the given equation 𝑦 ′′ − 𝑎2 𝑦 = 0.
2) Hint: Let 𝑔 be function with 𝑘 derivatives, and 𝑟 is a constant. Recalling the formula, which was proved in
Example 08, therefore we have
𝐷 𝑘 (𝑒 𝑟𝑥 𝑔) = ∑𝑘𝑙=0(𝑘𝑙) 𝐷 𝑙 (𝑒 𝑟𝑥 )𝐷𝑘−𝑙 (𝑔)
= ∑𝑘𝑙=0(𝑘𝑙) 𝑟 𝑙 𝑒 𝑟𝑥 𝐷𝑘−𝑙 (𝑔)
= 𝑒 𝑟𝑥 ∑𝑘𝑙=0(𝑘𝑙) 𝑟 𝑙 𝐷 𝑘−𝑙 (𝑔)
= 𝑒 𝑟𝑥 ∑𝑘𝑙=0(𝑘𝑙) (𝑟 𝑙 𝐷𝑘−𝑙 )(𝑔) (𝑟 𝑙 as constant and just it is multiplication)
= 𝑒 𝑟𝑥 (𝐷 + 𝑟)𝑘 (𝑔) (by binomial expansion)
⟹𝐷 𝑘 (𝑒 𝑟𝑥
𝑔) = 𝑒 𝑟𝑥 (𝐷 + 𝑟)𝑘 (𝑔). This completes the required result.

UNIT 02-02: LINEAR EQUATIONS WITH VARIABLE COEFFICIENTS


SELF-TEST 01
MCQ 02-02-01: B; MCQ 02-02-02: C; MCQ 02-02-03: D; MCQ 02-02-04: D.

END OF UNIT EXERCISES

1) Hint: Let 𝐿(𝑦) = 2𝑥 2 𝑦 ′′ + 3𝑥𝑦 ′ − 𝑦 = 0. First show that 𝜙1 = √𝑥 is a solution of 𝐿(𝑦) = 0 Then for 𝜙1 =
1 3
1 1
√𝑥, 𝜙1′ = 2 𝑥 2 , 𝜙1′′ = − 4 𝑥 2 and 𝜙2 = 𝑥 −1, 𝜙2′ = −𝑥 −2 , 𝜙2′′ = 2𝑥 −3 . Putting these,one can easily
− −

check 𝐿(𝜙1) = 0 and 𝐿(𝜙2) = 0 and hence, both 𝜙1 and 𝜙2 are solutions of 𝐿(𝑦) = 0. Finally, we have
1
𝑥2 𝑥 −1 3 3
𝑊(𝜙1 , 𝜙2 )(𝑥) == |1 1
−2 −2
| = − 2 𝑥 −2 ≠ 0. This implies that 𝜙1 and 𝜙2 are linearly independent
2
𝑥 −𝑥
solutions.
2) Hint: As 𝜙1 and 𝜙2 are fundamental solutions, we have 𝐿(𝜙1 ) = 𝜙1′′ + 𝑎1 (𝑥)𝜙1′ + 𝑎2 (𝑥)𝜙1 = 0 , and
𝐿(𝜙2 ) = 𝜙2′′ + 𝑎1 (𝑥)𝜙2′ + 𝑎2 (𝑥)𝜙2 = 0. Hence, we re-write the above equations as 𝑎1 (𝑥)𝜙1′ + 𝑎2 (𝑥)𝜙1 =
−𝜙1′′, and 𝑎1 (𝑥)𝜙2′ + 𝑎2 (𝑥)𝜙2 = −𝜙2′′. This is a system of two equations for 𝑎1 and 𝑎2 with 𝑊(𝜙1 , 𝜙2 )(𝑥) =
−Δ, where Δ is the determinant of coefficient matrix of the system. Hence by Cramer’s rule, the coefficients
𝜙1 𝜙2 𝜙1′ 𝜙2′
| | | ′′ |
Δ1 𝜙1′′ 𝜙2′′ Δ2 𝜙1 𝜙2′′
𝑎1 and 𝑎2 are uniquely determined 𝑎1 (𝑥) = Δ
= − 𝑊(𝜙 ,𝜙 )(𝑥) and 𝑎2 (𝑥) = Δ
= 𝑊(𝜙1 ,𝜙2 )(𝑥)
.
1 2

S25013: Ordinary Differential Equations Page 148


𝑦1 (𝑥) 𝑥+2 1 5
3) Hint: We form the quotient of two functions, one can show that 𝑦2 (𝑥)
= 2𝑥−1 = 2 + 4𝑥−2. It is seen that this
ratio is not equal to constant but depends on 𝑥. Hence these functions are linearly independent.
4) Hint: Let 𝜙1 (𝑥) = 𝑥 and 𝜙2 (𝑥) = 𝑒 𝑥 be two fundamental solutions. Then recalling Exercise 2), we have
𝜙1 𝜙2 𝜙1′ 𝜙2′
| ′′ | | |
𝜙1 𝜙2′′ 𝑥 𝜙1′′ 𝜙2′′ 1
𝑎1 (𝑥) = − = − (𝑥−1), and 𝑎2 (𝑥) = = (𝑥−1). Hence, the corresponding homogeneous
𝑊(𝜙1 ,𝜙2 )(𝑥) 𝑊(𝜙1 ,𝜙2 )(𝑥)
𝑥 1
linear differential equation is 𝑦 ′′ + 𝑎1 (𝑥)𝑦 ′ + 𝑎2 (𝑥)𝑦 = 0 ⟹ 𝑦 ′′ − (𝑥−1) 𝑦 ′ + (𝑥−1) 𝑦 = 0 or (𝑥 − 1)𝑦 ′′ −
𝑥𝑦 ′ + 𝑦 = 0.
2 2
5) Hint: Let 𝜙1 (𝑥) = 𝑥 be one of the solutions of the equation 𝑦 ′′ − 𝑦 ′ + 𝑦 = 0, 𝑥 > 0. Hence by Abel’s
𝑥 𝑥2
𝜙 𝜙2
formula for Wronskian 𝑊(𝑥), we have 𝑊(𝜙1, 𝜙2 )(𝑥) = | ′1 | = 𝑐1 exp(−∫ 𝑎1(𝑥) 𝑑𝑥) ⟹ 𝜙1 𝜙2′ −
𝜙1 𝜙2′
𝜙1 𝜙2′ −𝜙1′ 𝜙2 𝑐1 𝑥 2 𝑐1 𝑥 2
𝜙1′ 𝜙2 = 𝑐1 𝑥 2. Dividing both sides of the above equation by 𝜙12, we get 𝜙12
= 𝜙12
= 𝑥2
= 𝑐1 ⟹
𝜙 ′
(𝜙2 ) = 𝑐1 . After integration, we get 𝜙2 = (𝑐1 𝑥 2 + 𝑐2 𝑥). Now 𝜙(𝑥) = 𝑑1 𝜙1 + 𝑑2 𝜙2 is also a solution of
1

𝐿(𝑦) = 0. Thus, 𝜙(𝑥) = 𝑑1 𝑥 + 𝑑2 (𝑐1 𝑥 2 + 𝑐2 𝑥) ⟹ 𝜙(𝑥) = 𝑘1 𝑥 + 𝑘2 𝑥 2 ⟹ 𝜙1 (𝑥) = 𝑥 and 𝜙2 (𝑥) = 𝑥 2 are


2
solutions of 𝐿(𝑦) = 0. Finally, 𝑊(𝜙1 , 𝜙2 )(𝑥) = | 𝑥 𝑥 | = 𝑥 2 ≠ 0, 𝑥 > 0.⟹ 𝜙1 (𝑥) = 𝑥 and 𝜙2 (𝑥) = 𝑥 2
1 2𝑥
are linearly independent solutions of 𝐿(𝑦) = 0.

UNIT 02-03: REDUCTION OF THE ORDER


SELF-TEST 01
MCQ 02-03-01: B, C; MCQ 02-03-02: C; MCQ 02-03-03: D;
MCQ 02-03-04: A; MCQ 02-03-05: C.

END OF UNIT EXERCISES


1) Hint: The complementary function of equation is 𝑦𝑐 (𝑥) = 𝑐1 𝜙1 + 𝑐2 𝜙2 = 𝑐1 𝑥 + 𝑐2 𝑥 2 . We therefore by
method of variation of parameters, let 𝑦𝑝 (𝑥) = 𝑢1 (𝑥)𝜙1 + 𝑢2 (𝑥)𝜙2 = 𝑢1 (𝑥)𝑥 + 𝑢2 (𝑥)𝑥 2 , where 𝑢1 (𝑥) =
𝑥 𝑊1 (𝑡)𝑏(𝑡) 𝑥 𝑊2 (𝑡)𝑏(𝑡)
∫𝑥 𝑊(𝑡)
𝑑𝑡 and 𝑢2 (𝑥) = ∫𝑥 𝑊(𝑡)
𝑑𝑡. For this, we calculate first Wronskian 𝑊, is given by 𝑊(𝑥) =
0 0
2 2 𝑥 0
|𝑥 𝑥 | = 𝑥 2 ≠ 0, 𝑥 ∈ [1, ∞). Similarly, we have 𝑊1 = |0 𝑥 | = 0 − 𝑥 2 = −𝑥 2 and 𝑊2 = | |=
1 2𝑥 1 2𝑥 1 1
𝑥 𝑊1 (𝑡)𝑏(𝑡) 𝑥 −𝑡 2 .𝑡 sin 𝑡
𝑥 − 0 = 𝑥. Therefore, we find 𝑢1 and 𝑢2 : 𝑢1 (𝑥) = ∫𝑥 𝑊(𝑡)
𝑑𝑡 = ∫1 𝑡2
𝑑𝑡 = 𝑥 cos 𝑥 − sin 𝑥 −
0
𝑥 𝑊2 (𝑡)𝑏(𝑡) 𝑥 𝑡.𝑡 sin 𝑡
cos 1 + sin 1 and 𝑢2 (𝑥) = ∫𝑥 𝑑𝑡 = ∫1 𝑑𝑡 = − cos 𝑥 + cos 1. Hence, the particular integral
0 𝑊(𝑡) 𝑡2
is given by 𝑦𝑝 (𝑥) = 𝑢1 𝜙1 + 𝑢2 𝜙2 = −𝑥 sin 𝑥 + 𝑥[− cos 1 + sin 1] + 𝑥 2 cos 1. Thus, the general solution of
the equation is given by 𝑦(𝑥) = 𝑦𝑐 (𝑥) + 𝑦𝑝 (𝑥) = 𝑘1 𝑥 + 𝑘2 𝑥 2 − 𝑥 sin 𝑥,where 𝑘1 = [𝑐1 − cos 1 + sin 1] and
𝑘2 = [𝑐2 + cos 1]; 𝑘1 , 𝑘2 are arbitrary constants.
2) Hint: The corresponding homogeneous equation is 𝐿(𝑦) = 𝑥 2 𝑦 ′′ + 𝑥𝑦 ′ − 𝑦 = 0. Now we have 𝐿(𝜙1 ) =
𝑥 2 𝜙1′′ + 𝑥𝜙1′ − 𝜙1 = 𝑥 2 (𝑥)′′ + 𝑥(𝑥)′ − 𝑥 = 𝑥 2 . 0 + 𝑥. 1 − 𝑥 = 𝑥 − 𝑥 = 0, which shows that 𝜙1 (𝑥) = 𝑥 is a
1
solution of the equation 𝐿(𝑦) = 0.. Similarly, we can get 𝜙2 = − 2𝑥 as second solution. We therefore by
method of variation of parameters and let 𝑦𝑝 (𝑥) = 𝑢1 (𝑥)𝜙1 + 𝑢2 (𝑥)𝜙2 = 𝑢1 (𝑥)𝑥 + 𝑢2 (𝑥)𝑥 2 , where
𝑥 𝑊1 (𝑡)𝑏(𝑡) 𝑥 𝑊2 (𝑡)𝑏(𝑡) 1
𝑢1 (𝑥) = ∫𝑥 𝑊(𝑡)
𝑑𝑡 and 𝑢2 (𝑥) = ∫𝑥 𝑊(𝑡)
𝑑𝑡. Following Exercise 1, we have 𝑢1 (𝑥) = 2 log 𝑥 and
0 0

S25013: Ordinary Differential Equations Page 149


𝑥2 1 1 𝑥2
𝑢2 (𝑥) = 2
− 2 . Hence, the particular integral is given by 𝑦𝑝 (𝑥) = 𝑢1 𝜙1 + 𝑢2 𝜙2 = 2 log 𝑥 𝑥 + [ 2 −
1 1 1 1 1 log 𝑥
2
] (− 2𝑥) = − 4 . 𝑥 − 4 (− 2𝑥) + 𝑥 ( 2
). Thus, the general solution of the equation is given by 𝑦(𝑥) =
1 log 𝑥 1 1 1 1 log 𝑥
𝑦𝑐 (𝑥) + 𝑦𝑝 (𝑥) = 𝑐1 𝑥 + 𝑐2 (−
2𝑥
)+𝑥( 2
) − 4 . 𝑥 − 4 (− 2𝑥) = 𝑘1 𝑥 + 𝑘2 (− 2𝑥) + 𝑥 ( 2
) , where 𝑘1 =
1 1
[𝑐1 − ] , 𝑘2 = [𝑐2 − ]; 𝑘1 , 𝑘2 are arbitrary constants.
4 4

UNIT 02-04: HOMOGENEOUS EQUATIONS WITH ANALYTIC COEFFICIENTS


SELF-TEST 01
MCQ 02-04-01: C; MCQ 02-04-02: A; MCQ 02-04-03: D.

END OF UNIT EXERCISES


1) Hint: Using the formula (31), the only non-vanishing term in the expansion of derivative 𝑢(2𝑛) (𝑥)
(2𝑛)! 𝑑𝑛 𝑑𝑛 𝑑𝑛
is 2𝑛
𝐶𝑛 [(𝑥 + 1)𝑛 ](𝑛) [(𝑥 − 1)𝑛 ](𝑛) i.e. 𝑢(2𝑛) (𝑥) = (𝑥 + 1)𝑛 (𝑥 − 1)𝑛 , where (𝑥 + 1)𝑛 = 𝑛!
(𝑛!)2 𝑑𝑥 𝑛 𝑑𝑥 𝑛 𝑑𝑥 𝑛
𝑑𝑛
and (𝑥 − 1)𝑛 = 𝑛!. Thus 𝑢(2𝑛) (𝑥) = (2𝑛)!.
𝑑𝑥 𝑛
1 𝑑𝑛+1
2) Hint: Using the definition of polynomial 𝑃𝑛 , we can have 𝑃𝑛+1 (𝑥) = 2𝑛+1 (𝑛+1)! 𝑑𝑥 𝑛+1 (𝑥 2 − 1)𝑛+1 .
1 𝑑𝑛 2𝑛𝑥 2
Differentiating w. r. t. 𝑥 on both side, we get 𝑃𝑛+1
′ (𝑥)
= [(1 + (𝑥 2 ) (𝑥 2 − 1)𝑛 ] . Replacing 𝑛 by
2𝑛 n! 𝑑𝑥 𝑛 −1)
1 𝑑𝑛 2n
𝑛 − 2 in the above equation, we obtain the relation for 𝑃𝑛−1 as 𝑃𝑛−1
′ ′ (𝑥)
= 2𝑛 n! 𝑑𝑥 𝑛 [(𝑥 2 −1) (𝑥 2 − 1)𝑛 ]. Hence,
we have
′ (𝑥) ′ (𝑥)
1 𝑑𝑛 2𝑛𝑥 2 2 𝑛
1 𝑑𝑛 2n
𝑃𝑛+1 − 𝑃𝑛−1 = 𝑛 𝑛
[(1 + 2
) (𝑥 − 1) ] − 𝑛 𝑛
[ 2 (𝑥 2 − 1)𝑛 ]
2 n! 𝑑𝑥 (𝑥 − 1) 2 n! 𝑑𝑥 (𝑥 − 1)
1 𝑑𝑛
= (2𝑛 + 1) [(𝑥 2 − 1)𝑛 ]
2𝑛 n! 𝑑𝑥 𝑛
′ (𝑥)
⟹ 𝑃𝑛+1 ′ (𝑥)
− 𝑃𝑛−1 = (2𝑛 + 1)𝑃𝑛 (𝑥) . This is the required result.
1 𝑑𝑛
3) Hint: We know the Legendre polynomial 𝑃𝑛 (𝑥) of degree 𝑛 , is given by 𝑃𝑛 (𝑥) = (𝑥 2 − 1)𝑛 .
2𝑛 𝑛! 𝑑𝑥 𝑛
1 1
Therefore, we have 𝑃0 (𝑥) = 1, 𝑃1 (𝑥) = 𝑥, 𝑃2 (𝑥) = (3𝑥 2 − 1), 𝑃3 (𝑥) = (5𝑥 3 − 3𝑥) . ⟹ 𝑥 = 𝑃1(𝑥),
2 2
2𝑃3(𝑥) + 3𝑥 = 5𝑥 3 ⟹ 𝑥 = 𝑃1(𝑥), 5𝑥 3 = 2𝑃3 (𝑥) + 3𝑃1 (𝑥) ⟹ 𝑓(𝑥) = 5𝑥 3 + 𝑥 = 2𝑃3 (𝑥) + 3𝑃1 (𝑥) +
𝑃1 (𝑥) = 2𝑃3 (𝑥) + 4𝑃1 (𝑥).
4) Hint: By parts integration, we have
1 1 1
∫−1 𝑥 𝑘 𝑃𝑛 (𝑥)𝑑𝑥 = 2𝑛 𝑛! [0 − 𝑘 ∫−1 𝑥 𝑘−1 Dn−1 (𝑥 2 − 1)𝑛 𝑑𝑥] (∵ D𝑘 (𝑥 2 − 1)𝑛 |±1 = 0 for 0 ≤ 𝑘 < 𝑛) .

Continuing 𝑘 −times, we obtain


1 (−1)𝑘 𝑘!
∫−1 𝑥 𝑘 𝑃𝑛 (𝑥)𝑑𝑥 = [0] (∵ D𝑛−𝑘−1 (𝑥 2 − 1)𝑛 | = 0 for 0 ≤ 𝑛 − 𝑘 − 1 < 𝑛)
2𝑛 𝑛! ±1

1 (−1)𝑘 𝑘!
⟹ ∫−1 𝑥 𝑘 𝑃𝑛 (𝑥)𝑑𝑥 = [0]
2𝑛 𝑛!
1
⟹ ∫−1 𝑥 𝑘 𝑃𝑛 (𝑥)𝑑𝑥 = 0 for (𝑘 < 𝑛).

S25013: Ordinary Differential Equations Page 150


𝜋
5) Hint: We use the substitution sin 𝑥 = 𝑡 ⟹ cos 𝑥 𝑑𝑥 = 𝑑𝑡 and when 𝑥 = 0 ⟹ 𝑡 = 0 ; 𝑥 = 2
⟹ 𝑡 = 1.
𝜋
1 1
Hence, we have ∫0 cos 𝑥 𝑃3 (sin 𝑥)𝑑𝑥 = ∫0 𝑃3(𝑡)𝑑𝑡.
2 But 𝑃3 (𝑡) = 2 (5𝑡 3 − 3𝑡) , we therefore have
𝜋
11 1 1 1 −1 1
∫02 cos 𝑥 𝑃3(sin 𝑥)𝑑𝑥 = ∫0 2 (5𝑡 3 − 3𝑡)𝑑𝑡 = 2 ∫0 (5𝑡 3 − 3𝑡)𝑑𝑡 = 4 ( 2 ) = − 8.
2𝑥 −2𝑥
6) Hint: We write the given differential equation 𝐿(𝑦) = 𝑦 ′′ − (1−𝑥 2) 𝑦 ′ = 0, where 𝑎1 (𝑥) = 1−𝑥 2 , 𝑎2 (𝑥) =
0. Here 𝜙1 (𝑥) is one known solution of the equation. If 𝜙2 is another solution of 𝐿(𝑦) = 0, then it is given
that
𝑠
− ∫ 𝑎 (𝑡)𝑑𝑡
𝑥 𝑒 𝑥0 1 1 1+𝑥 1 1+𝑥
𝜙2 (𝑥) = 𝜙1 (𝑥) ∫𝑥 𝑑𝑠 = (1 − 𝑥02 ) [ log ( ) − [2 log (1−𝑥0)]]
0 [𝜙1 (𝑠)]2 2 1−𝑥 0
1 1+𝑥 1 1+𝑥
⟹ 𝜙2 (𝑥) = 𝑐1 [2 log (1−𝑥)] + 𝑐2 1, where 𝑐1 = (1 − 𝑥02 ) and 𝑐2 = −(1 − 𝑥02 ) [2 log (1−𝑥0 )].
0
We know if 𝜙1 , 𝜙2 are solution of 𝐿(𝑦) = 0, then 𝜙(𝑥) = 𝑑1 𝜙1 + 𝑑2 𝜙2 is also a solution. Therefore,
1 1+𝑥 1 1+𝑥
we have 𝜙(𝑥) = 𝑑1 1 + 𝑑2 [𝑐1 ( log ( )) + 𝑐2 1] = 𝑒1 1 + 𝑒2 [2 log (1−𝑥)] , 𝑒1 = 𝑑1 + 𝑑2 𝑐2 and 𝑒2 =
2 1−𝑥
1 1+𝑥
𝑐1 𝑑2 ⟹ 𝜙2 (𝑥) = log (
2 1−𝑥
) is another independent solution of 𝐿(𝑦) = 0. Note, 𝑊(𝜙1 , 𝜙2 )(𝑥) ≠ 0

S25013: Ordinary Differential Equations Page 151


Feedback Form
School of Sciences
(Formerly School of Architecture, Science & Technology)
Yashwantrao Chavan Maharashtra Open University,
Nashik – 422 222

FEEDBACK SHEET FOR THE STUDENT

Dear Student,
You have gone through this book, it is time for you to do some thinking for us.
Please answer the following questions sincerely. Your response will help us to
analyse our performance and make the future editions of this book more useful.
Your response will be completely confidential and will in no way affect your
examination results. Your suggestions will receive prompt attention from us.

Please submit your feedback online at this QR Code or at following link


https://forms.gle/rpDib9sy5b8JEisQ9

or email at: director.ast@ycmou.ac.in


or send this filled “Feedback Sheet” by post to above address.

(Please tick the appropriate box)

Pl. write your Program Code Course Code & Name

Style
01. Do you feel that this book enables you to learn the subject independently without
any help from others?

Yes No Not Sure


02. Do you feel the following sections in this book serve their purpose? Write the
appropriate code in the boxes.
Code 1 for “Serve the purpose fully”
Code 2 for “Serve the purpose partially”
Code 3 for “Do not serve any purpose”
Code 4 for “Purpose is not clear”
Warming up Check Point Answer to Check Points

To Begin With Summary References


Objectives Key Words

03. Do you feel the following sections or features, if included, will enhance self - learning
and reduce help from others?
Yes No Not Sure

Index
Glossary
List of “Important Terms Introduced”
Two Colour Printing

Content

04. How will you rate your understanding of the contents of this Book?

Very Bad Bad Average Good Excellent

05. How will you rate the language used in this Book?
Very Simple Simple Average Complicated Extremely Complicated

06. Whether the Syllabus and content of book complement to each other?
Yes No Not Sure

07. Which Topics you find most easy to understand in this book?
Sr.No. Topic Name Page No.
08. Which Topics you find most difficult to understand in this book?
Sr.No. Topic Name Page No.

09. List the difficult topics you encountered in this Book. Also try to suggest how
they can be improved.
Use the following codes:
Code 1 for “Simplify Text”
Code 2 for “Add Illustrative Figures”
Code 3 for “Provide Audio-Vision (Audio Cassettes with companion Book)”
Code 4 for “Special emphasis on this topic in counseling”

Sr.No. Topic Name Page No. Required Action Code

10. List the errors which you might have encountered in this book.

Sr.No. Page Line Errors Possible Corrections


No. No.
11. Based on your experience, how would you place the components of distance
learning for their effectiveness?
Use the following code.
Code 1 for “Most Effective” Code 3 for “Average” Code 5 fo r “Least Effective”

Code 2 for “Effective” Code 4 for “less Effective”

Printed Book Counseling Lab Journal

Audio Lectures Home Assignment YouTube Videos

Video Lectures Lab-Experiment Online Counseling

12. Give your overall rating to this book?

1. 2. 3. 4. 5.

13. Any additional suggestions:

Thank you for your co-operation!

You might also like